Datasets:

medarc

/

clinical_pile_v1_minhash_deduped

like 1

solvent-dependent_conductance_decay_constants_in_single_cluster_junctions

## Abstract: Single-molecule conductance measurements have focused primarily on organic molecular systems. Here, we carry out scanning tunneling microscope-based break-junction measurements on a series of metal chalcogenide Co 6 Se 8 clusters capped with conducting ligands of varying lengths. We compare these measurements with those of individual free ligands and find that the conductance of these clusters and the free ligands have different decay constants with increasing ligand length. We also show, through measurements in two different solvents, 1-bromonaphthalene and 1,2,4-trichlorobenzene, that the conductance decay of the clusters depends on the solvent environment. We discuss several mechanisms to explain our observations. ## Introduction Controlling charge transport through molecular junctions is critical to the realization of nanoscale electronic devices. 1,2 While numerous organic molecules have been studied as connecting wires for single-molecule junction studies, very little is known about the effect of metal complexes in these types of junctions. We recently reported that we could incorporate metal chalcogenide molecular clusters in single-molecule electrical circuits. 15 In this study, in order to determine how transport through such systems depends on molecular length, we connect these same clusters to conducting ligands of varying lengths. We have found that the inclusion of the cluster in the molecular circuit reduces the effect of ligand length on conductance decay with apparent molecular size. Moreover, we have found that the decay constant is impacted greatly by changing the solvent from 1,2,4-trichlorobenzene (TCB) to 1bromonaphthalene (BrN). Specifcally, the decay constant of the cluster is 0.04 1 in BrN, while it is 0.12 1 in TCB. We consider two possible mechanisms to explain these remarkable observations. Our work demonstrates, for the frst time, a molecular system where the tunneling decay constant can be modifed by altering the environment around the molecule. ## Results and discussion The single cluster circuits that we have designed, assembled, and studied consist of an atomically defned Co 6 Se 8 molecular cluster 16,17 (Fig. 1a) wired between nanoscale electrodes. The wiring is formed from bifunctional, conjugated ligands (Fig. 1b) that bind specifcally and directionally to the electrode and to the cluster. We employ an atomically defned segment of polyacetylene 18 that has an arylphosphine group on one terminus that coordinates to a cobalt atom on the clusters and an arylthiomethyl group on the other terminus that attaches to the Au electrode. 19,20 The mono-, di-, and triene ligands are L1, L2, and L3 and the corresponding clusters are 1, 2, and 3, respectively. Fig. 1c shows the molecular structure of 1 as determined by single crystal X-ray diffraction (SCXRD). 15 We measured the conductance of both the individual molecular clusters (1-3) and the free conducting ligands (L1-L3) using a scanning tunneling microscope-based break-junction (STM-BJ) technique. 21 In this technique, an Au STM tip and substrate are repeatedly brought into and out of contact to form and break Au-Au point contacts in solutions of the target compounds. During this process, a bias voltage is applied across the junction while current is measured in order to determine conductance (G ¼ I/V) of the junction. The measurements are repeated thousands of times, and the data is analyzed to reveal statistically signifcant results. The data is processed by compiling thousands of individual conductance traces into one-dimensional, logarithmically-binned conductance histograms. 22 We further generate two-dimensional (2D) histograms of the conductance versus displacement by aligning each conductance trace after the point contact ruptures (at a conductance of 0.5 G 0 ) and overlaying all conductance traces. In order to characterize transport through the molecular clusters and the free ligands, we measured the conductance of 1-3 and L1-L3 in two different solvents, BrN and TCB. These solvents were chosen taking into consideration the solubility of both the ligand and the cluster systems as well as for their varied affinity to gold electrodes. 23 Fig. 2a and b contain the onedimensional histograms for the measurements in BrN, and Fig. 2d and e show the same for the measurements in TCB. The 2D histograms for 1 and L1 in each solvent are insets in the respective fgures. The 2D histograms show a signifcant difference in length of the molecular feature for 1 and for L1. Moreover, the cluster junction lengths measured from the 2D histograms correlate with the molecular lengths of the cluster with the ligands fully extended (measured in BrN: 9 and 21 , and expected from SCXRD: 13 and 32 , for L1 and 1 respectively). Despite the additional complexity of the cluster system, we conclude that we are indeed probing transport through Au-ligand-cluster-ligand-Au junctions based on this large difference in the observed lengths. Furthermore, the histograms in Fig. 2a and d show shoulders, with an increasing prominence for the longer-ligand systems. Comparing the conductance of these shoulders with the ligand conductance in Fig. 2b and e, we attribute these shoulders to free ligands, that is, ligands that have detached from the clusters. We ft the peaks of all conductance histograms for both solvents with a Gaussian function and plot the peak conductance values versus the number of C]C units or "enes" in each molecule (Fig. 2c and f). In both solvents, and for both free ligand and cluster, we observe that the conductance decreases exponentially with increasing molecular length following the relationship G $ e bn , where n is the number of "ene" units in the backbone and b is the decay constant. We report the decay constant per Angstrom using a length of 2.48 per "ene" unit. The decay constant for the free ligand series is essentially independent of the solvent (b ¼ 0.15 1 in TCB and 0.17 1 BrN). The unexpected result is the factor of 3 difference in the decay constants of the cluster series in different solvents as can be seen comparing Fig. 2c and f. In TCB, the b of the cluster system is 0.12 1 , and in BrN it is 0.04 1 . We note that the difference between the decay constant of the ligand and that of the cluster is greater in BrN than in TCB. Furthermore, the absolute values of the conductance of the cluster series are signifcantly higher when measured in BrN than in TCB, with the conductance of 3 being almost an order of magnitude higher in BrN compared to TCB. Such a solvent-induced effect on the conductance has been observed in other systems, and this has been attributed to the solvent's ability to modulate the electrode work function. 23,24 These fndings are summarized: regardless of solvent the effect of C]C chain-length on conductance is less pronounced in 1-3 than in L1-L3. Furthermore, the conductance and the decay of L1-L3 are essentially insensitive to the choice of solvent, while the solvent signifcantly influences those of 1-3. To understand these results we consider several possible mechanisms of charge transport through these junctions. Charge transport can occur via a coherent off-resonance process through an orbital on the ligand-cluster-ligand assembly that is coupled to both electrodes. In that situation, the conductance depends on at least two related factors: (1) the energy of this conducting orbital relative to the metal E F , and (2) the coupling between this orbital and both electrodes. 25 As the length of the molecule increases, the HOMO-LUMO gap narrows, and if conductance were just related to energy level alignment, one would naively expect conductance to actually increase. However, transport through the junction is also related to how well the conducting orbital overlaps with the leads, and since the orbital is more delocalized over a longer conjugated molecule, this overlap decreases with increasing length. The conductance thus typically decays exponentially with increasing molecular length. Specifcally, as the conjugated backbone gets longer, the molecular orbital is delocalized over a longer molecule, and since the orbital is normalized, a smaller fraction of its amplitude resides on the sulfur atoms; therefore the coupling between the molecule and the electrodes decreases. If we assume that the conducting orbitals of the cluster and of the ligand for a given length are similar in both character and energy, we can develop a simple tight-binding model of the molecular junctions. The objective of this model is not to reproduce the experimental data but to examine and illustrate how the additional electronic structure of the cluster impacts transport through the system. Our tight-binding model is schematically presented in the insets of Fig. 3a and b for the ligand and the cluster respectively. For the conducting ligands, we assign a single energy level, 3, for each unit, and allow nearest neighbors to be coupled by d. The terminal units are coupled to the Au electrodes using an imaginary self-energy, iG/2. We apply a similar model for the cluster, adding an additional energy level, E 0 , between two ligands and coupling this site to its nearest neighbor ligand states with s. We compute the transmission functions for these model systems using a Green's function approach (see the ESI for a detailed description †). 25,26 Sample computed transmission functions are shown in Fig. 3a and b using the same values for 3, d and G for the ligand and the cluster series. The transmission functions display resonances at energy values corresponding to the molecular orbitals of the system where the probability of an electron being transmitted through the system is unity. The transmission function for L1 contains one resonance at energy 3, while longer ligands have resonances equal to the number of sites in the corresponding model. As the length of the molecule increases, the frontier resonance moves closer to E F but also narrows, which is a consequence of the frontier orbital being delocalized over a longer molecular backbone. Upon comparing the transmission functions for the ligands with those of the clusters, we see that the clusters contain resonances that are closer to E F than their ligand counterparts, but with narrower full widths at half max (i.e. they are more poorly coupled to the leads). This observation leads to a lower transmission at E F ; more importantly, it also leads to a conductance that is more sensitive to the exact location of the E F . In Fig. 3c, we show the conductances that are determined from the tight-binding model for each molecule versus the number of ligand levels in the molecule (using the same model parameter values for both series). From the ft to these values, it is clear that the predicted decay constants are essentially the same for the ligand and cluster series. We use one set of E 0 and s values to calculate the representative transmission/conductance functions shown in Fig. 3a and b. Regardless of what value is assigned to E 0 and s, we fnd that this model predicts very similar decay constants for the two systems (Fig. 3d). Our tight-binding calculations suggest that the addition of a cluster level E 0 between two ligands cannot explain the observed change in b. In other words, the ligand and cluster series should have the same b values, unless the energy alignment of the cluster resonance is altered relative to the electrode Fermi level in this model. We have three sets of observations that are consistent with a change in E F : (1) b of the cluster in BrN is signifcantly lower than in TCB, (2) b values measured in both solvents are almost the same for the ligand series, and (3) b of the cluster is lower than that of the ligand in both solvents. The steeper transmission curves of the cluster series in Fig. 3 indicate that the resonance energies are closer to E F . Within this coherent transport model, we can see that a small change in E F will result in a large shift in b for the cluster relative to the ligand. For instance, changing E F by 0.5 eV shifts the b value to 0.1 1 for the clusters while a similar change in E F for the ligand changes b to 0.3 1 . These results, when viewed in light of the known ability of solvent-binding to produce changes in E F , 23 point to BrN shifting E F closer to resonance relative to TCB. This effect is compounded by the sensitivity of the metal. The free ligand and the cluster have very different characteristics (e.g., size, steric hindrance, redox behaviour, dielectric constant polarizability and binding ability) that will result in different shifts in E F . We also consider a hopping mechanism for charge transfer, a process generally mediated by an activation-controlled reaction (e.g., a thermally induced conformational change or an electron transfer reaction). 27,28 We frst rule out the possibility that such a conformational change can occur within the ligand. 3 We also refute the process involving a direct through-space charge transfer from the electrode to an unoccupied molecular level on the cluster through a resonant transfer process. 14 In this picture, the cluster does not have to be chemically attached to the electrodes to form a conducting junction and the charge transfer efficiency depends on the core-electrode spacing. We discount this mechanism based on a previously published study in which we demonstrated that our clusters form molecular junctions by bonding their terminal thiomethyl groups to the Au electrodes. 15 By varying the substitution pattern or removing the aurophilic functionality, we can modulate or completely shut down the conductivity of these molecular junctions, suggesting that there is an orbital pathway for the transport of charge in these cluster systems. These fndings refute the idea of direct through-space charge transfer mediated by an orbital localized on the core. We are left with a hopping mechanism in which the charge tunnels from the source electrode across the ligand to the cluster core and then transfers to the drain electrode through a second coherent tunneling process. Such a transport process requires that the cluster can reversibly change its oxidation state with each charge transfer. Since the applied bias in these measurements is not small ($0.5 V) and the cluster core Co 6 Se 8 is redox active, it is plausible that such a hopping process is at play. In this case, the activation energy arises from the charge transfer process reorganization energy, which can be strongly influenced by the solvent. This mechanism is consistent with our observation that b changes with solvent. Within our experimental constraints, it is therefore difficult to conclusively establish which process (off-resonance tunneling or hopping) is at work in our single cluster junction system. ## Conclusions In summary, we measured charge transport through molecular clusters with ligands of different lengths and showed that the conductance decay depends on the solvent used for these measurements. Our results illustrate a novel effect that allows the environment to alter the conductance decay constants. This study opens up the possibility to carry out conductance measurements in which clusters can be controllably gated by changing the environment. 20 While the conducting ligands alone are limited to a one-dimensional system, the threedimensional architecture of the metal chalcogenide cluster allows us to envision novel electronic devices where a molecular cluster is contacted by electrodes at multiple locations.

chemsum

{"title": "Solvent-dependent conductance decay constants in single cluster junctions", "journal": "Royal Society of Chemistry (RSC)"}

characterization_of_frequency-chirped_dynamic_nuclear_polarization_in_rotating_solids

## Abstract: Continuous wave (CW) dynamic nuclear polarization (DNP) is used with magic angle spinning (MAS) to enhance the typically poor sensitivity of nuclear magnetic resonance (NMR) by orders of magnitude. In a recent publication we show that further enhancement is obtained by using a frequency-agile gyrotron to chirp incident microwave frequency through the electron resonance frequency during DNP transfer. Here we characterize the effect of chirped MAS DNP by investigating the sweep time, sweep width, center-frequency, and electron Rabi frequency of the chirps. We show the advantages of chirped DNP with a tritylnitroxide biradical, and a lack of improvement with chirped DNP using AMUPol, a nitroxide biradical. Frequency-chirped DNP on a model system of urea in a cryoprotecting matrix yields an enhancement of 142, 21% greater than that obtained with CW DNP. We then go beyond this model system and apply chirped DNP to intact human cells. In human Jurkat cells, frequency-chirped DNP improves enhancement by 24% over CW DNP. The characterization of the chirped DNP effect reveals instrument limitations on sweep time and sweep width, promising even greater increases in sensitivity with further technology development. These improvements in gyrotron technology, frequency-agile methods, and incell applications are expected to play a significant role in the advancement of MAS DNP. ## Introduction Dynamic nuclear polarization (DNP) is commonly used to improve the inherent insensitivity of nuclear magnetic resonance (NMR) spectroscopy . Typically, only continuous wave (CW) microwave methods have been employed with magic angle spinning (MAS) DNP. The solid effect and the cross effect are the primary DNP mechanisms used in moderate magnetic field strengths of 5-14 Tesla (T) . While CW approaches can significantly increase NMR sensitivity, they have limitations. Except in certain model systems , the solid effect and cross effect are inefficient at room temperature due to short longitudinal electron relaxation times. To perform CW DNP, samples are commonly cooled to <120 K, which adds complexity not only to the instrumentation, but also often leads to a loss of spectral resolution . Arrested molecular motion at these temperatures can cause substantial line broadening in most samples [3, . The cross effect and solid effect also exhibit worse performance at higher magnetic field, with cross effect efficiency decreasing as 1/B0 and that of solid effect as 1/B0 2 . Therefore new mechanisms will be required for efficient DNP at magnetic fields of 28 T and higher. Frequency-chirped DNP techniques, such as the frequency-swept integrated solid effect (FS-ISE) , nuclear orientation via electron spin locking (NOVEL) , and timeoptimized pulsed (TOP) DNP show promise to perform well both at high magnetic field and room temperature. For instance, ISE yields DNP enhancements of ~150 at room temperature and is predicted to be unaffected by the strength of the external magnetic field . However, these experiments have been performed without MAS and at magnetic fields <3 T , primarily due to the difficulty of implementing MAS with the microwave resonators required to generate considerable electron nutation frequencies. Frequency-swept DNP at higher magnetic fields has also been shown to improve DNP performance , but has only recently been implemented with MAS . MAS improves the sensitivity and resolution of solid-state NMR by partially averaging anisotropic interactions of the magnetic resonance Hamiltonian, and is a crucial aspect of applying DNP to systems of interest. Here we characterize the behavior of frequency-chirped DNP experiments performed with MAS, expanding on our recent work . We optimize frequency chirps from a custom-built frequency-agile high-power gyrotron to produce large gains in intensity beyond those obtained with CW DNP. In addition to measuring its performance on a model system, we conduct optimized chirped experiments on intact human Jurkat cells to demonstrate frequency-chirped DNP in a biologically complex environment. ## NMR Experiments MAS DNP NMR experiments were performed using a custom-built DNP spectrometer at a magnetic field of 7.1584 T . 13 C and 1 H Larmor frequencies were 75.4937 MHz and 300.1790 MHz, respectively. A CPMAS, rotor synchronized, Hahn echo sequence with TPPM decoupling was used for all experiments (Fig. 1a). The initial magnetization of 1 H and 13 C spins was destroyed using a saturation train. 1 H and 13 C pulses were performed with nutation frequencies of 77 kHz and 100 kHz, respectively. The Hartmann-Hahn matching condition (γB1) for 1 H and 13 C was 30 kHz. Frequency chirps were applied over the DNP polarization period (τpol), and CW microwaves were employed over the rest of the experiment. The spinning frequency was 4.5 kHz for all experiments, and the sample temperature was 90 K. Typical polarization times (τpol) for optimized spectra were 5-times the T1 of the sample in the absence of microwaves, in order to remove contamination of the data by differences in the nuclear T1 and the T1DNP. Microwaves were generated using a frequency-agile gyrotron, whose output frequency was adjusted by varying the electron acceleration potential at the electron gun anode. An arbitrary waveform generator (AWG) integrated into the NMR spectrometer (Redstone, Tecmag Inc. Houston, TX) was used to generate a waveform, which ramped the output frequency of the gyrotron in a linear fashion through 197.670 GHz, the frequency of maximum DNP enhancement of the TEMTriPol-1 radical . The frequency chirps were a triangular waveform, which was repeated over the entire polarization period. For frequency chirp optimization the incident microwave power, the center DNP microwave frequency, and the sweep width and sweep time of the individual chirps were varied. The center frequency of the sweeps was varied by changing the voltage at the gyrotron anode with the AWG amplified by a high-voltage amplifier (TREK, Inc. Lockport, NY). The sweep width corresponded to the frequency range of one sweep/chirp (either up or down) in MHz, and sweep time was the time to complete a sweep/chirp. Microwave power was attenuated from full power by inserting copper foil with slits cut in it into a gap in the waveguide to partially pass the microwave beam. The optimal power of 7 W incident on the sample was used for most experiments, which provided an estimated electron Rabi frequency of 0.43 MHz . ## 7 The 13 C carbonyl resonance was fit using DMfit to determine resulting enhancement increases. For all optimization spectra, the magnitude of the Hahn echo was used to calculate the percent increase in intensity. All experiments were repeated four times to acquire adequate error values for the measurements. ## Sample Preparation Experiments were performed on 4 M [U-13 C, 15 N] urea mixed with 5 mM TEMTriPol-1 or 5 mM AMUPol in a cryoprotecting matrix consisting of 60% d8 glycerol, 30% D2O, and 10% H2O by volume. Intact Jurkat cells (ATCC, Manassas, VA) were cultured in [U- 13 C, 98%; U-15 N, 98%] BioExpress-6000 mammalian cell growth medium (Cambridge Isotope Laboratories, Tewksbury, MA) at a concentration of 3 × 10 6 cells/mL in a six-well plate at 37°C and 5% CO2 for 48 hr. 3.6 × 10 7 cells were collected, spun at 170 g for 5 min, washed with 1× phosphate-buffered saline (PBS), and spun again at 170 g for 5 min to remove extracellular NMR labels (g = 9.8 m/s 2 ). 40 µL of 20 mM TEMTriPol-1 in 1×PBS with 10% DMSO was added to a cell pellet containing 36 million Jurkat cells. This suspension was centrifuged directly into the 3.2 mm zirconia rotor at 800 g for 30 s and immediately frozen in liquid nitrogen as detailed in our previous work . ## Results and Discussion Frequency-chirped DNP refers to a change in the microwave frequency or intensity throughout the course of an experiment. The frequency-chirped DNP pulse sequence is shown in Fig. 1A. Frequency chirps (represented by the rainbow gradient) are applied over the DNP polarization period and the resulting NMR signal is detected through a cross polarization (CP) Hahn echo sequence. We emphasize that microwave frequency chirps result in better manipulation of the electron spin polarization, yet the active DNP mechanism is still the cross effect. Selection of appropriate radicals for frequency-chirped DNP is crucial due to drastic differences in electron spin g-anisotropy and relaxation properties. In our previous demonstrations of electron decoupling using chirped microwave pulses with MAS, we employed trityl rather than nitroxide radicals . Those successes led us to explore the use of trityl-nitroxide biradicals, with the rational that the narrow trityl resonance would be easier to manipulate and the tethered nitroxide would provide greater DNP enhancements through the cross effect mechanism. TEMTriPol-1 is such a biradical, consisting of a Finland trityl radical covalently linked to a 4-amino TEMPO radical, which is used for cross effect DNP . TEMTriPol-1 improves cross effect efficiency at high magnetic fields. Where other biradicals, such as AMUPol, depolarize nuclear spins at 100 K in the absence of microwave irradiation, TEMTriPol-1 preserves nuclear polarization . ## Frequency-chirped DNP on a Model System CW DNP CPMAS experiments were performed at various microwave frequencies to record a 1 H DNP enhancement profile with TEMTriPol-1 . The enhancement profile shows the trityl resonance frequency as the optimal frequency for CW DNP enhancement. This will be the target for the center of the frequency chirps. In a 7.1584 T magnetic field, the microwave frequency for maximum CW DNP enhancement was 197.670 GHz (Fig. 1B). Experiments were performed to determine the effect of frequency-chirped microwave pulses during the polarization period of MAS DNP (Fig. 2). For comparison, cross effect DNP experiments were performed with CW microwave irradiation. CW DNP experiments on a model system of urea with TEMTriPol-1 resulted in an enhancement of 118 (Fig. 2, red). Enhancements herein are defined as the NMR signal intensity recorded with DNP compared to that without DNP . For frequency-chirped DNP experiments, the microwave frequency was linearly chirped with a triangular waveform over 197.670 GHz, with a 28 µs sweep time and a 120 MHz sweep width. These optimized chirps yielded a 21% increase over CW DNP and an enhancement of 142 (Fig. 2, blue). Polarization times of 53 s (5×T1DNP, Fig. S1) were used to ensure that >99% of the polarization had built up for both experiments, allowing for direct comparison of the CW and frequency-chirped experiments. To determine the necessity of a narrow-line radical, such as trityl, for frequency-chirped DNP, experiments were performed on a sample containing the nitroxide-nitroxide biradical, AMUPol. The frequency chirps were centered at 197.674 GHz (maximum with 1 H-enhancement for AMUPol) the previously optimized sweep time of 28 µs and sweep width of 120 MHz were used. Frequency chirps over the polarization period resulted in a decrease in signal intensity of 3% compared to CW DNP (Fig. 3). These frequency chirps do not yield the same improved electron spin control over the nitroxide biradical, AMUPol, as they do over TEMTriPol-1. This implies that a narrow-line radical is required for implementation of frequency-chirped DNP. ## Frequency-chirped DNP in Intact Jurkat Cells The performance of frequency-chirped DNP was then examined within intact human Jurkat cells (Fig. 4). Frequency chirps improved the NMR signal by 24%, yielding an enhancement of 6 (versus 4.8 for CW DNP). These results display the application of frequency-chirped DNP to more complex samples of biological interest. ## Power Dependence of CW and Frequency-chirped DNP To determine the dependence of CW and frequency-chirped enhancement on microwave power, CPMAS experiments were performed with varying microwave attenuation on the TEMTriPol-1/urea sample (Fig. 5). For frequency-chirped DNP the optimized triangle waveform (28 µs sweep time and 120 MHz sweep width) was repeated over a polarization time of 20 s. 35 W of microwave power incident on the sample (Rabi frequency of 0.95 MHz) produced a 123% increase in signal with frequency-chirped DNP compared to CW, yielding enhancements of 17 and 8, respectively (Fig. 5a, b). We note that such high microwave powers place the cross effect in the oversaturated regime, leading to less overall enhancement. 7 W of microwave power resulted in the highest sensitivity and an improvement of 25% with frequency-chirped DNP compared to CW. Higher microwave power yielded greater improvements with frequency-chirped DNP over CW DNP, but the overall signal intensity obtained was suboptimal due to saturation of the cross effect . ## Characterization of Frequency-chirped DNP The effects of sweep time, sweep width, and center frequency on the improvement with frequencychirped DNP over CW irradiation are shown in Fig. 6. For this dependence the polarization time was 20 s; the sweep width was held constant at 80 MHz, the incident microwave power at 7 W, and the center frequency at 197.670 GHz. Shorter sweep times increased the sensitivity to a greater extent than longer sweep times, with the greatest improvement over CW (15%) occurring with a 20 μs sweep time (Fig. 6a). Sweep times below 20 μs were not achievable with the current microwave frequency agility circuit, as the frequency output waveform became distorted. A sweep time of 150 µs resulted in only a 1% improvement in signal intensity over CW. We suspect that at longer sweep times electron spin saturation is lost through relaxation mechanisms. The dependence of frequency-chirped DNP sensitivity on the sweep width of the frequency chirps is shown in Fig. 6b. For this dependence the polarization time was 20 s; the sweep time was held constant at 28 μs, the incident microwave power at 7 W, and the center frequency at 197.670 GHz. The improvement from the frequency chirps increased as the sweep width increased. A 120 MHz sweep width resulted in an improvement of 21%, while the signal intensity decreased by 1% with a sweep width of 10 MHz. Due to instrument limitations, sweep widths greater than 120 MHz could not be attained. This width is roughly that of the base of the trityl lineshape in the enhancement profile (Fig. 1b). We previously reported a similar optimal sweep width in electron decoupling experiments involving the Finland trityl radical . Larger sweep widths provide microwave irradiation that is resonant with a greater number of trityl electron spins, enabling better electron spin control and improving the efficiency of frequency-chirped DNP. During characterization it is important to consider multiple points on the enhancement profile. Fig. 6d provides a clear picture of the effect of frequency chirping, whereas Fig. 6c shows the potential for misinformation. The choice of irradiation frequency can lead to suspiciously high improvements due to difference in positive and negative enhancement regions between CW and frequency-chirped DNP. The CW enhancement profile shows maximum positive and negative enhancements at 197.670 GHz and 197.850 GHz, respectively (Fig. 6d). Frequency chirping at microwave frequencies lower than 197.750 GHz (positive enhancement), yielded greater enhancements than CW (Fig. 6d). However, at frequencies greater than 197.750 GHz (negative enhancement), the frequency-chirped DNP provided lower signal intensity than CW DNP. Note that at this point we have simply demonstrated the methodology of performing frequencychirped DNP experiments with TEMTriPol-1. To compare the sensitivity of the experiments with TEMTriPol-1 and AMUPol, we can divide the signal-to-noise from each experiment by the square root of the polarization time for the respective experiments. In doing so, we obtain a sensitivity of 79 with AMUPol (Fig. 3) and 73 with TEMTriPol-1 (Fig. 2). Thus, while the sensitivity of the experiments performed on each radical are similar at this stage, advances in instrumentation that enable greater sweep times and sweep widths will make frequency-chirped DNP experiments with TEMTriPol-1 more sensitive than AMUPol, and thus more feasible for sensitivity-demanding, multidimensional experiments. ## Conclusion To date, frequency-chirped DNP experiments, such as FS-ISE, NOVEL, and TOP DNP, have been largely restricted to static samples due to the difficulties of housing microwave resonators with the instrumentation required for magic angle spinning (MAS). Here, we have characterized the optimal experimental conditions for frequency-chirped MAS DNP. At a magnetic field of 7 T and with 7 W of microwave power, frequency-chirped microwaves over the polarization period improved DNP enhancements by 21%. Greater microwave powers resulted in up to 123% improvements with frequency-chirped DNP, but saturation of the cross effect resulted in less overall signal intensity. These optimized frequency-chirped experiments were applied to a more biologically complex sample: intact Jurkat cells. This resulted in an improvement in signal intensity of 24% over CW DNP. Characterization of the parameters of frequency-chirped DNP revealed areas for further improvements to elicit even greater sensitivity. More powerful gyrotrons with larger frequency bandwidths, and gating mechanisms for chirps can be developed to increase sweep widths and shorten the sweep times, thus improving electron spin control. To take full advantage of frequencychirped DNP at high power and high electron Rabi frequencies, duty cycling of the microwaves can be implemented to reduce dielectric heating . We expect optimization of the waveform, with respect to both intensity and phase, to result in improved frequency-chirped DNP MAS performance. Future studies could analyze the effect of the spinning frequency on the enhancement achieved by frequency chirped DNP over CW DNP. Both the solid effect and cross effect are driven by interactions between the spin system, the microwave field, and the spinning rotor. Understanding these effects will prove crucial in the future development of DNP, as MAS frequencies and magnetic fields are pushed to ever higher values. New radicals composed of tethered broad and narrow line radicals are currently being investigated with useful electronic properties such as long longitudinal relaxation times. Longer relaxation times will afford even more electron spin control with frequency-chirped DNP. Although the precise mechanism governing the improvement in sensitivity will require further investigation, it is possible that it is governed by an adiabatic process. As such, future experiments could focus on maintaining a constant sweep rate by simultaneously varying the sweep time and sweep width in an inverse manner. This could prove important, as adiabatic processes often show a remarkable resilience to microwave inhomogeneities and frequency offsets arising from difference in molecular orientation and conformations in a solid sample. These techniques can be paired with other advances in instrumentation such as higher power microwave sources and microwave lenses for improved microwave intensity and high frequency MAS for 1 H detected spectra in future experiments. These could allow for the implementation of pulsed DNP mechanisms such as electron-nuclear cross polarization at high magnetic fields in the foreseeable future.

chemsum

{"title": "Characterization of Frequency-chirped Dynamic Nuclear Polarization in Rotating Solids", "journal": "ChemRxiv"}

physico-chemical_properties_and_catalytic_activity_of_the_sol-gel_prepared_ce-ion_doped_lamno3_perov

## Abstract: Ce-doped LaMno 3 perovskite ceramics (La 1−x Ce x Mno 3 ) were synthesized by sol-gel based coprecipitation method and tested for the oxidation of benzyl alcohol using molecular oxygen. Benzyl alcohol conversion of ca. 25-42% was achieved with benzaldehyde as the main product. X-ray diffraction (XRD), thermogravimetric analysis (TGA), BET surface area, transmission electron microscopy (teM), X-ray photoelectron spectroscopy (Xps), temperature-programmed reduction (H 2 -tpR), temperature-programmed oxidation (o 2 -tpo), Ft-IR and UV-vis spectroscopic techniques were used to examine the physiochemical properties. XRD analysis demonstrates the single phase crystalline high purity of the perovskite. the Ce-doped LaMno 3 perovskite demonstrated reducibility at low-temperature and higher mobility of surface o 2 -ion than their respective un-doped perovskite. the substitution of Ce 3+ ion into the perovskite matrix improve the surface redox properties, which strongly influenced the catalytic activity of the material. The LaMnO 3 perovskite exhibited considerable activity to benzyl alcohol oxidation but suffered a slow deactivation with time-on-stream. Nevertheless, the insertion of the A site metal cation with a trivalent Ce 3+ metal cation led to an enhanced in catalytic performance because of atomic-scale interactions between the A and B active site. La 0.95 Ce 0.05 Mno 3 catalyst demonstrated the excellent catalytic activity with a selectivity of 99% at 120 °C. Presently, perovskite-based materials are gaining immense popularity in the field of material science due to their extraordinary optical, electro-magnetic properties. Perovskite materials mostly applied for removing common exhaust pollutants including carbon monoxide, hydrocarbon, ammonia oxidation, water dissociation, and NOx, etc. . Amongst different perovskite, Mn-containing oxide materials have been growing a considerable interest from the researchers because of the large specific external area, high thermo-chemical durability and extraordinary catalytic performance even at environmental conditions . These excellent physicochemical properties of Mn-based perovskite materials made them an ideal candidate for their applications in the decomposition of customary use pollutants including carbon monoxide, NOx, and poisonous hydrocarbons. In this regard, various types of catalytic conversion technologies were developed 4,5,8,10,11 . Besides that, in order to make the catalytic combustion widely applicable, the development of reliable technologies is highly desirable. Amongst various catalytic active perovskite materials, lanthanide (Ln 3+ ) ion substituted perovskite demonstrated superior activities 10,12 . Such materials revealed higher catalytic activity and superior thermal stability for hydrocarbon combustion than their respective un-substituted perovskites 2,7,10 . Owing to the outstanding catalytic activity of perovskite-type oxide ABO 3 , where A is 12 coordinated and larger cation in size, whereas B is 6 fold coordination and smaller cation in size with oxygen anion. The partial co-doping of the A-site by the transition metal ions with dissimilar valance generate a structural defect because of bond stretching and amend the valence of the B-site to meet the chemical charge balance of the perovskite structure; actually, it is the prime origin for extraordinary catalytic oxidation performance of the ABO 3 based oxides. Therefore, doping of similar valence state ions at A or B sites might be altered the crystal structure, geometrical symmetry and disturb the oxidation states of the cations without altering the structure. Besides that, the variation of Mn 4+/ Mn 3+ ratio has the main effect on the catalytic activities of ABO 3 materials. The partial doping of Ce 3+ ion into LaMnO 3 altered the catalytic activity because of an increase in specific surface area, surface defects, oxygen mobility, and redox ability. Ceria has the capability to absorb and release the oxygen vacancies, and these oxygen species play a crucial role in the overall catalytic activities of the CeO 2 -based perovskites . Owing to the oxidation state transformation behavior of ceria between Ce 3+ and Ce 4+ dependent on the O 2 partial pressure in the nearby atmosphere 13,14 . Usually, the redox behavior of Ce 3+ is determined by morphology, size, and dissemination of oxygen species as the utmost appropriate surface defects 13 . This unique property of Ce 3+ revealed high thermo-chemical robustness and large O 2 species movement, and thus displays improved performance in catalytic oxidation of hydrocarbons and nitrogen oxides. So far, nonstoichiometric perovskite materials demonstrated some specific physical properties including evolution in surface defects, oxygen ion mobility, and redox property. In this article, we proposed the synthesis of Ce 3+ ion substituted LaMnO 3 nanoparticles via sol-gel based co-precipitation process. We inspected the impact of Ce 3+ ion doping in LaMnO 3 nanoparticles on physiochemical properties and oxidation performance of C 6 H 5 CH 2 OH to C 6 H 5 CHO. For characterization various techniques were applied including X-ray diffraction pattern (XRD), transmission electron microscope (TEM), energy dispersive x-ray analysis (EDX), N 2 adsorption, Fourier transform infrared (FTIR), optical absorption (UV-Vis), thermogravimetric analysis (TGA), temperature program reduction (TPR), temperature program oxidation (TPO) and X-ray photoelectron spectroscopy (XPS) techniques. These techniques revealed the role of Ce 3+ ion substitution on the crystal structure, crystallinity, surface properties, thermal stability, optical, redox behavior, oxygen adsorption properties and catalytic activities of the as-prepared nonstoichiometric LaMnO 3 materials. experimental section synthesis of perovskites (La 1−x Ce x Mno 3 ). Analytical grade chemicals were procured and used directly without any extra distillation. In a typical synthesis of LaMnO 3 perovskite, 4.3 g La(NO 3 ) 3 .6H 2 O (99.99%), and 2.4 g Mn(NO 3 ) 3 .3H 2 O (99.99%, BDH Chemicals Ltd, UK), were dissolved in 50 ml H 2 O along with C 6 H 8 O 7 .H 2 O (E-Merck, Germany). Citric acid was used as a chelating agent for complexation with lanthanum and manganese nitrates. The resulting mixed aqueous solution was magnetically stirred on a hot plate at 100 °C until the transparent solution was achieved. Aqueous ammonia solution was quickly added to precipitation under constant mechanical stirring. The occurrence of the willing product was dried at 100 °C for overnight and further annealed at 700 °C in the air for 5 hrs. A similar procedure was repeated for synthesis of La 1−x Ce x MnO 3 oxides (x = 0.05, 0.07 and 0.10 mol %). ## Catalyst characterization. Powder X-ray diffraction measurement was performed on a PANalytical X'PERT (X-ray diffractometer) furnished with Ni filter and using CuKα (λ = 1.5406 ). Morphology was obtained from Field emission Transmission Electron Microscope (FE-TEM, JEM-2100F JEOL, Japan) furnished with energy dispersive x-ray analysis (EDX) functioned at an accelerating voltage of 200 kV. Thermal analysis was measured on (TGA/DTA Mettler, Toledo, AG, Analytical CH-8603, Schwerzenbach, Switzerland). UV/Vis absorption spectra were measured by using Perkin-Elmer Lambda-40 Spectrophotometer. Fourier transforms Infrared (FT-IR) spectra were recorded on Perkin-Elmer 580B IR spectrometer. Temperature program reduction (TPR) and Temperature program oxidation (TPO) spectra were recorded on chemisorption Micromeritics AutoChem model 2910 analyzer furnished with a thermal conductivity indicator. Before the experiment, 100 mg material sample was treated with 10 vol % O 2 /He stream at 500 °C for 30 min to get complete oxidation. Then materials were cooled at room temperature and a mixture of 10 vol% H 2 /Ar gas with flow rate 20 mL/min was introduced and the reactor was heated from ambient temperature to 900 °C and maintained this temperature up to 20 min. For the O 2 -TPO experiments, helium(He, 30 mL/min) gas was applied for drying the perovskite samples at 150 °C and cooled down to room temperature, followed by an increase of temperature under O 2 /He (30 mL/min) flow with a temperature slope of 10 °C/min to 900 °C on the same instrument. The textural properties of the perovskites were recorded on a Micromeritics TriStar 3000 BET Analyzer, taking a value of 0.162 nm 2 for the cross-sectional area of the N 2 molecule adsorbed at 77 K. Powder samples were dried and degassed by heating gently to 90 °C for 1 h, then at 200 °C for 3 h under flowing N 2 before measurement. The free space in each sample tube was determined with He, which was assumed not absorb. Catalytic studies. Liquid-phase oxidation of benzyl alcohol was carried out in a glass vessel equipped with a magnetic stirrer, reflux condenser, and thermometer. Briefly, a mixture containing benzyl alcohol (2 mmol), toluene (10 mL) and the perovskite (0.3 g) was vigorously stirred in a three-necked round-bottomed flask (100 mL) and then heated up to 120 °C. The O 2 -gas was introduce in the reaction mixture through bubbling to start the oxidation experiment with a 20 mL/min flow rate. After completion of reaction solid catalyst extracted from the solution by centrifugation and reaction mixture was analyzed by gas chromatography to examine the conversion of the alcohol and product selectivity by (GC, 7890 A) Agilent Technologies Inc, equipped with a flame ionization detector (FID) and a 19019S-001 HP-PONA column. The specific activity of the catalyst was calculated using the equation ## Specific activity Moles of substrate (mmol) Product formed/Amount of catalyst(g) Reactiontime(h) The turnover number and turnover frequency of the catalyst were calculated using ## Results and Discussion Crystallographic and morphological structure. Figure 1 demonstrates the XRD pattern to observe the chemical composition, crystallographic structure and grain size of the as-synthesized perovskite. As observed in Fig. 1. the distinct diffraction lines of perovskite in XRD pattern can be assigned to the (012), (110), (104), (202), (024), (122), (116), (214), (018), ( 208) and (128) lattice planes, which are attributed to the hexagonal structure of LaMnO 3 nanoparticles(Fig. 1) (JCPDS card No. 032-0484) 6,19 . Any other diffraction line associated with MnO or CeO 2 is not identified over the whole XRD range specifies the homogeneous dispersion into the crystal lattice and formation of perfect single phase LaMnO 3 perovskite. An observed diffraction line at 30.27° corresponds to La 2 O 3 , which is weaker than the reflection lines of LaMnO 3 perovskite. All diffractograms of the perovskite materials revealed the similar trigonal symmetry in the crystallographic space group with marginally dissimilar cell parameters. As shown in Fig. 1 diffraction lines in trivalent Ce 3+ substituted perovskite are slightly shifted towards longer angle along with reduced intensity in respect to the un-substituted LaMnO 3 perovskite, it could be due to the effect of Ce 3+ ion doping into the crystal matrix. Owing to the small radius of Ce 3+ ions, they are highly mobile and easily migrate from surface to crystal lattice within the crystal matrix of perovskite materials at environment conditions 13,14,20 . The broadening of reflection lines in perovskite materials suggested the nanocrystalline nature of the as-prepared nanomaterials. As shown in Fig. 1, on substituted of small radius Ce 3+ (1.25 ) in place of La(1.27 ), the reflection lines slightly shifted to higher 2θ, signifying that the crystal arrangement becomes distorted 13,21 , resulting the transformation is occurring in the symmetry of crystallographic structure 7,10,22 . The experimentally calculated lattice parameters for LaMnO 3 , La 0.95 Ce 0.05 MnO 3 , La 0.93 Ce 0.07 MnO 3, and La 0.90 Ce 0.10 MnO 3 are a = 5.527 , 5.463 , 5.449 and 5.436 , respectively, are decreased on increasing the substitution concentrations of the Ce 3+ ion into the LaMnO 3 crystal lattice in respect to un-substituted LaMnO 3 perovskite. These variations in lattice parameters and shifts in peak positions endorse the substitution of modified ions into the crystal lattice structure. TEM micrograph clearly shows the irregular hexagonal structure, smooth surface, uncontrolled size, highly aggregated, well-distributed nanoparticles. Figure 2a illustrates the typical image of Ce 3+ ion substituted LaMnO 3 perovskite nanoproduct with size ranging from 25-31 nm. Energy dispersive x-ray analysis in Fig. 2b revealed the existence of all substituted elements including La 3+ , Mn 3+ , Ce 3+ and oxygen elements in the as-prepared LaMnO 3 perovskite. The appearance of intense peaks of Cu 2+ and C belong to the carbon coated copper grid. It confirmed the efficacious doping of Ce 3+ into the crystal matrix. textural properties and thermal stability. The structural parameters after calcination of Ce substituted LaMnO 3 catalysts, Specific surface area (BET), pore volume (PV) and average pore size (PD) are summarized in Table 1. The PV and PD were obtained from the adsorption branch of the respective N 2 isotherm by put on the BJH method. Surface area (Single point BET and Multipoint BET), PV and PD drop with increasing Ce ion concentrations from 5 to 10 mol% (Table 1). Thermogravimetric (TGA) analysis of the as-prepared LaMnO 3 perovskite and Ce-substituted materials exhibit a similar decomposition trend in all thermograms (Fig. 3). TGA spectra were recorded from 0-900 °C in N 2 -atmosphere with a heating rate of 10 °C/min (Fig. 3). First big exothermic peak (DTA) in all samples are observed at around 400 °C resemble the crystalline H 2 O molecules or complexation form surface attached organic impurities. The surface attached OH groups or organic moieties are coordinated to the central metal ion in different attachment form in the existing complex precursor system 23,24 . Generally, -OH groups attached on the surface of metal ions in two forms either terminal Ln-OH or in the bridge from Ln-(OH)-Mn 25 . In both cases, the dissociation of surface OH groups contrasts from each other depending on the surrounding chemical environment. So that, the reduction ii molar mass occurs in a rather varied range of temperature. No decomposition peaks signifying further crystallization are found in TGA, specifying that the perovskite materials are in crystalline form, as verified by XRD results. All four thermograms illustrate the sluggish weight loss (~6-8%) in between 400-900 °C, which is assigned to the removal or combustion of carbon dioxide at high temperature. optical properties. Figure 4 displays the infrared spectra of the as-synthesized LaMnO 3 and different Ce ion substituted LaMnO 3 perovskite nanoparticles. All samples exhibited a diffused band in between 3160-3653 cm −1 assigned to the νO-H stretching vibration originating from surface adsorbed H 2 O molecules (Fig. 4) 25 . Two additional strong intensity infrared bands are observed positioned at 1486 and 1375 cm −1 attributed to the δOH and γOH vibrational modes of H 2 O molecules. These observed infrared spectral results are in accord with TGA observations. The observed infrared band at 644 cm −1 is allotted to the νM-O stretching vibrational mode which certified the formation of metal oxide framework 26,27 . www.nature.com/scientificreports www.nature.com/scientificreports/ Optical absorption spectra were carried out to determine the optical characteristics of the as-synthesized perovskites (Fig. 5a,b). The direct energy band gap (E g ) is estimated by fitting the absorption spectral data to the straight transition equation by extrapolating the linear portions of the curve into αhν = A(hν − E g )½, where α is optical absorption coefficient, hν is the photon energy, E g is the direct bandgap and A is constant (Fig. 5b) 25,28,29 . The experimentally assessed direct energy band gaps of all perovskite nanomaterials are 1.15, 1.31, 1.34 and 1.32 eV for LaMnO 3 , La 0.95 Ce 0.05 MnO 3 , La 0.93 Ce 0.07 MnO 3 , and La 0.90 Ce 0.10 MnO 3 perovskites, respectively. An observed increase band gap energy with increasing the Ce 3+ ion substitution quantity into the LaMnO 3 crystal lattice, which is attributable to the Burstein-Moss effect 28, . ## Redox properties (tpR/tpo). Redox properties of the as-prepared LaMnO 3 perovskite and their Ce 3+ ion substituted LaMnO 3 perovskites are determined by H 2 -TPR and the observed results are presented in Fig. 6a and tabulated in Table 2. TPR and TPO studies are performed to examine the role of Ce 3+ ion-doping on redox behavior of LaMnO 3 perovskite within the range from 50-800 °C. The TPR spectra were recorded within the temperature range from 50 to 800 °C temperature. TPR spectra exhibited two typical characteristic reduction peaks, first one in between 280-600 °C and second started from 645 °C5 . The observed peak at low reduction temperature (280-600 °C) is correspond to the reduction of Mn 4+ to Mn 3+ and elimination of surface adsorbed oxygen vacancies, and the second reduction band is observed at a higher temperature (645 °C), which correspond to the reduction of Mn 3+ to Mn 2+ 4,6,7,33,34 . The first broadband occurred at lower reduction temperature indicate the largest H 2 -consumption, it suggesting the better initiative catalytic activities of LaMnO 3 perovskite at a lower temperature. The higher oxidation state of Mn 3+/4+ ions is accountable for more oxygen species because of lacking ligand amounts of Mn 3+/4+ ion. The occurrence of Mn 4+ ion is associated with the fact that Mn 3+ has a permitted electron, and have the ability to adsorb molecular O 2 and convert it into an electrophilic form 6 . Reversed transformation of manganese ion oxidation states is observed by the TPO analysis (Fig. 6b), in which the oxidation peak www.nature.com/scientificreports www.nature.com/scientificreports/ at low temperature (205-310 °C) suggest the transition of Mn 2+ to Mn 3+ and the oxidation peak at 445-717 °C exhibit the oxidation from Mn 3+ to Mn 4+ . These observations are in accord with published reports 4,5,34 . Additionally, the H 2 -TPR profile shape of LaMnO 3 is altered after doping of different Ce 3+ ion concentrations into the LaMnO 3 crystal lattice as seen in Fig. 6a. The incorporation of Ce 3+ ion into the LaMnO 3 matrix strongly modified the reduction behavior of LaMnO 3 perovskite. As shown in Fig. 6a, the Ce 3+ ions-substituted sample revealed three peaks at 330-345, 440-450 and ~800 °C, the first band looks very minute and the second band occurs very robustly 35 . The occurrence of two peaks in Ce 3+ ion substituted LaMnO 3 TPR profiles indicates the existence of at least two species in the LaMnO 3 crystal lattice, which became stronger and shifted towards high temperature after increasing the doping concentrations of Ce 3+ . An observed band between 330-345 °C, ascribed to the replacement of Mn 2+ by Ce 3+ in LaMnO 3 crystal matrix. Because of this charge disparity lattice alteration would arise that promote to the construction of La-O-Mn-O-Ce solid solution form, resulting the reactive O 2 vacancies are produced that may be reduced simply at low temperature. Generally, the elimination of oxygen vacancies at low temperatures associated with higher oxygen mobility (oxygen reacts more easily) and oxygen reactivity 4,6 . An observed reduction band at 448 °C ascribed to the dissociation of powerfully interactive MnO 2 type with Ce 3+ supports, whereas weak intensity reduction band observed at ~800 °C consigned to the high-temperature dissociation band because of bulk MnO 2 24 . Owing to the variation in balance of both metal (Mn 3+/4+ and Ce 3+/4+ ) cations from 4+ to 3+ or from 3+ to 2+, the up-down swings of O 2 imperfections escorted with valence alteration is observed 6,35 . Therefore, the high O 2 storage capacity of 10 mol% Ce substituted LaMnO 3 perovskite because of the simultaneous occurrence of transportable O 2 vacancies and analogous (Mn 2+/3+/4+ /Ce 3+/4+ ) redox couples. Consequently, the La 0.90 Ce 0.10 MnO 3 sample revealed an excellent catalytic activity at a lower temperature, so that, the highest redox properties, these results are in accord with previous literature reports 7,24,33 . Comparatively the intensity of the high-temperature components is remarkably varied on increasing the Ce ion concentrations, whereas peak positions (decomposition temperature) are almost similar. It suggested the similar type of species is reduced at the same temperature, which enhanced by Ce 3+ ion substitution. As shown in Fig. 6a, La 0.90 Ce 0.10 MnO 3 sample revealed high reducibility at high temperature. So that, the replacement of La 3+ by Ce 3+ ion would effect in enhanced concentrations of Mn 3+ ions and oxygen vacancies because of charge discrepancy accomplished by oxidation of Mn 2+ to Mn 3+ and by the construction of an oxygen-deficient perovskite La 0.90 Ce 0.10 MnO 3 , which would enhance the reducibility character of the perovskite. These observations are well consistent with XRD and XPS results, in which non-Ce ion substituted Mn 2+ species are oxidized and transform into Mn 3+ valence states. It inferred that the reducibility behavior of the perovskites in the following sequence LaMnO 3 ≤ La 0.95 Ce 0.05 MnO 3 ≤ La 0.90 Ce 0.10 MnO 3 ≤ La 0.90 Ce 0.10 MnO 3 , according to the H 2 consumption at 446 °C and 800 °C. Generally, oxygen species are attached with metal ion into two different bonding forms including non-crystalline and crystalline bonding forms. In the non-crystalline bonding form, the oxygen species are present in the outer coordination sphere and is referred to as surface adsorbed oxygen species. Whereas in case of crystalline bonding form, the oxygen species entered into the inner coordination sphere and compensate its valence state. These crystalline form oxygen species can be typically eliminated in metal oxide products at higher temperature 36,37 . Temperature program oxidation or desorption was performed to evaluate the catalytic affinity towards oxygen. Figure 6b illustrates the TPO profile of the as-prepared LaMnO 3 and different Ce 3+ ion concentration substituted LaMnO 3 perovskites. The TPO-profile of blank LaMnO 3 perovskite in Fig. 6b, illustrate three oxygen desorption regions, at three different temperatures including 266, 533 and ~799 °C, respectively. An observed first band at 266 °C is attributed to the weakest oxygen vacancies (superficial O 2 species), which are physiochemically adsorbed/chemisorbed O 2 species and are eliminated at low-temperature. The appearance of broadband between 350-725 °C assigned to the non-stoichiometric oxygen (interfacial oxygen) vacancies and reduction of Mn 4+ to Mn 3+ , which are desorbed at high temperature. Whereas the oxygen vacancies desorbed at a higher temperature (≥725 °C) can be attributed to the relocation of lattice O 2 in the bulk perovskite phase and reduction of Mn 3+ to Mn 2+ 7,10,33,35 . Generally, surface adsorbed O 2 vacancies desorbed at low temperatures and interfacial oxygen in non-stoichiometric form desorbing at high temperature 33,35,36,38 . As seen in Fig. 6b, when the Ce 3+ ion is replaced in the La 3+ site of LaMnO 3 perovskite a charge balance is desired to attain the neutrality of the perovskite. It can either achieved by O 2 defects or the swing of the Mn ion towards higher valance states (Mn 3+ to Mn 4+ ). As illustrated in Fig. 6b, on the substitution of 5 mol% Ce 3+ ion doping the strong low-temperature peak is shifted towards slightly higher temperature, which corresponds to surface desorbed oxygen species. While high-temperature peak assigned to interfacial oxygen species is split into two peaks observed at 390 and 490 °C. However, on increasing the substitution concentration of Ce 3+ ion in LaMnO 3 crystal lattice, the low temperature desorption peaks are moved towards higher temperature with significant enhanced integral area, indicating the homogeneous substitution of Ce 3+ ion into crystal lattice which increase the oxygen ion mobility of both surface (superficial) oxygen species and non-stoichiometric (interfacial) lattice oxygen species, it could be due to the effect of small ionic size Ce 3+ ion substitution 13,24,25 . As observed previously, the Ce 3+/4+ ions have high oxygen species motilities because of their multiple oxidation states. The high-temperature O 2 desorption of LaMnO 3 is typically denoted to as the removal of non-stoichiometric surplus oxygen. It could be due to the creation of Mn 3+ in LaMnO 3 to reduce the Jahn-Teller distortion, although the charge stability advocates that Mn should be in 3+ oxidation state. In La 0.90 Ce 0.10 MnO 3 the Mn 3+ state is highly stable because of the existence of Ce 3+ ions in the crystal lattice (charge compensation) 33 . Xps studies. The surface chemical components, phase purity, and their oxidation states are inspected by XPS analysis. Figures 7 and 8 demonstrated the XPS spectra of La(3d & 4d), Mn(2p) and O(1 s) for the different Ce ion concentration substituted perovskites. XPS spectra of the La 3d in the LaMnO 3 and La x Ce 1−x MnO 3 displayed two binding energies (BE) bands located at 844 and 860 eV which correspond to the La 3d 5/2 and La 3d 3/2 , respectively. The existence of these valence band indicates that lanthanum in La 3+ ion form(Fig. 7a) 1 . Additionally, each band has additional satellite band along with core band, owing to the relocation of electrons from O2p to the vacant orbital of La 5 f orbital. These observations are similar to the previous values observed for La 2 O 3 1,39 , it suggested www.nature.com/scientificreports www.nature.com/scientificreports/ the trivalent state of La 3+ ions in the perovskite materials. The increased La 4d binding energy is interpreted as due to the displacement of the electron density toward nearest neighbors. The oxygen (O1s) signal in XPS spectra shows two peaks, the first one is centered at 531 eV and second at around 436 eV in La 0.95 Ce 0.05 MnO 3 sample (Fig. 7b). As shown in Fig. 7b, the low BE band is due to the lattice oxygen, whereas broader band with high BE band is associated with the surface adsorbed oxygen or surface hydroxyl groups. Peng et al. observed that the surface adsorbed O 2 is the most active oxygen because of higher mobility in respect of lattice oxygen, which plays a crucial role in conversion process through migration from the surface to lattice sites 1,3,13 . As seen in Fig. 7b, on increasing the dopant concentration (Ce 3+ ions) the peaks are varied along with broadening, it indicates the existence of several types of oxygen vacancies such as oxygen of hydroxyl (-OH − )/carbonate(-CO 3 2− ) groups on the surface of matrices 2,7,8,10 and it is in accord with the TPO results. According to the TPO results the observed low-temperature desorption band(surface O 2 species) is directly related to the quantity of O 2 species are in very small, while the high quantity of O 2 species evolved at a higher temperature(chemisorbed O 2 species). An observed an increase in core-level binding energy indicates that all of the cations in the samples (La, Ce, and Mn) are bonded to the oxygen. Most importantly, we are unable to observe the Ce ion peak in the current perovskites matrixes due to the Ce ion in LaCeMnO 3 perovskites are mostly in the tetravalent state 40 . An observed XPS peak located at around 655 eV is assigned to 2p 1/2 of Mn ions, although the band of Mn 2p 3/2 is composed of multiple bands it implies the presence of multivalence states such as Mn 2+ (641), Mn 3+ (644) and Mn 4+ (648) (Fig. 8) . Qureshi et al. observed that the splitting in Mn 2p peak is due to the asymmetric nature of the metal, which suggests Mn exists in the mixed valence state 46,47 . However, satellite structure at higher BE divided by ~4 eV, it could be due to the strong columbic interaction in between hybridization of Mn 3d electrons and other valence sub-shells 42,44,47 . No Mn 2p 3/2 band for Mn (~639 eV) is detected in the spectrum, it implies that no metallic form of Mn is presented in the as-prepared perovskites (Fig. 8). The impact of the catalytic activity on MnOx is related to its oxidation states which are MnO 2 > Mn 2 O 3 > MnO as reported by Thirupathi & Smirniotis 4,10,48,49 . According to them, MnO 2 is a highly reactive compound in all Mn-based compounds including MnO 2 , Mn 5 O 8 , Mn 2 O 3 , and Mn 3 O 4 . Therefore, Mn 4+ has higher catalytic performance, and this resembled the finest catalytic denitration activity of La 90 Ce 10 MnO 3 . The peaks of the Mn 2p 1/2 and Mn 2p 3/2 of the applied materials are moved towards longer BE, observed at ~2 eV and 3 eV, respectively. As shown in Fig. 8, the binding energies are significantly varied upon increasing the Ce ion concentration into the perovskite matrix, it indicates the variation in valence states of Mn ions. ## Catalytic reaction. The prepared materials were exposed to catalytic assessment and the conversion of benzyl alcohol into benzaldehyde is taken up as a typical reaction. It was observed that the prepared catalysts are active against the substrate benzyl alcohol. Adding Ce in the LaMnO 3 catalyst is found to impact on catalytic aerobic oxidation of benzyl alcohol due to the synergetic effect between Ce 3+/4+ and Mn 3+/4+ ions. The C 6 H 5 CHO is the core constituent, with an insignificant quantity of C 6 H 5 COOH as a byproduct. The perovskite LaMnO 3 is found to yield a 29% benzaldehyde within 12 hours, while conversion yield is improved on increasing the Ce ion substitution concentration in the perovskite, as shown in Table 3 (Fig. 9). As demonstrated in Fig. 9, on the substitution of 0.05% Ce in the La 0.95 Ce 0.05 MnO 3 catalyst yielded 10% more benzaldehyde i.e. 40% which is better than their parent or blank perovskite. Further modification of the catalyst with further increase in the percentage content of Ce in the catalytic system, yielded La 0.93 Ce 0.07 MnO 3 and La 0.9 Ce 0.1 MnO 3 respectively, it indicates that the catalytic activity decreases as the % of Ce 3+ ion concentration increase in the catalyst composition. The catalyst La 0.93 Ce 0.07 MnO 3 and La 0.9 Ce 0.1 MnO 3 yielded 37% and 32% oxidation product, i.e. benzaldehyde, respectively. Furthermore, the selectivity towards benzaldehyde was found to be >99% in all the cases. The graphical representation of the results obtained for all the catalysts tested is given in Fig. 9. When the catalytic activity is compared to the external area of the as-synthesized perovskite, it was observed that the catalyst La 0.95 Ce 0.05 MnO 3 which displayed the best catalytic performance has a surface area of 7.7922 m 2 /g, and it found to be lower than the surface area of the perovskite LaMnO 3 i.e. 8.3410 m 2 /g, which yielded a 29% benzaldehyde within 12 hours lower than the catalyst La 0.95 Ce 0.05 MnO 3 which yielded a 40% benzaldehyde. However, as the % of Ce in the catalyst composition is increased in the perovskites i.e. La 0.93 Ce 0.07 MnO 3 and La 0.9 Ce 0.1 MnO 3 the surface area further decreases to 7.7554 and 6.9371 respectively and the catalytic performance also depreciates. This indicates that the catalytic activity is not only dependent on the specific surface area it also depends on the doping concentration of the Ce 3+ ion in the materials. An un-doped perovskite possesses Mn in +3 state, while upon the inclusion of the Ce 3+ ions and the Mn oxidation state +4 (excess) and +2 is obtained as indicated by the XPS. Noticeably, Ce 3+ ion concentration plays a crucial part in the enhancement of the catalytic performance as it induces a high surface oxygen mobility than their un-doped perovskite, and the Mn oxidation state +4 (excess) and +2 is obtained, which enhances the surface redox properties of the perovskites as confirmed by the XPS. However, further increase of the Ce 3+ ions in the perovskite was found to result in the diminution in the catalytic performance, it specifies may be the depreciation in Mn 4+ and Mn 2+ sites and increase in the Mn 3+ ion. Apart from the oxidation states of Mn, the decrease in the La 3+ which results due to the increase of Ce 3+ in the catalytic www.nature.com/scientificreports www.nature.com/scientificreports/ system may also be accountable for the depreciation in the catalytic activity. The specific catalytic activity of the as-designed materials is calculated based on the turnover number and turnover frequency as presented in Table 3. From the values obtained, it is found that the catalyst La 0.95 Ce 0.05 MnO 3 has the highest TON and TOF among all the catalysts prepared. Further studies are determined in order to optimize the reaction temperature for the best catalytic performance, the catalyst La 0.95 Ce 0.05 MnO 3 , is utilized for the oxidation of C 6 H 5 CH 2 OH at various temperatures ranging from 40 °C to reflux temperature, and it was found that the catalyst performance is best at the reflux temperature, while at other temperatures, a slight decrease in catalytic performance was observed, observed results are illustrated in Fig. 10. ## Conclusions We successfully synthesized and characterized the Ce 3+ ion substituted lanthanum magnetite perovskites materials by co-precipitation method and applied for conversion of benzyl alcohol into benzaldehyde. Chemical composition and phase purity of the as-synthesized materials were validated from XRD, EDX, TGA and FTIR analysis. The values of optical energy band gaps were varied because of discrepancy in the grain size of the perovskite materials. The increase in doping quantity of Ce 3+ ions altered the redox (TPR and TPO) behavior of the perovskite oxides. The insertion of co-dopant Ce 3+ ion in perovskite lattice enhanced the quantity of Mn 4+ and chemisorbed oxygen positions on the surface of perovskite lattice to increase the catalytic performance. The XPS spectra of La 3d, Mn 2p, and O 1 s clearly revealed the influence of Ce ion substitution, which confirms the transformation of the Mn oxidation state from 3+ to 4+ due to the substitution of trivalent Ce 3+ ions at the La 3+ site in LaMnO 3 perovskite. The surface Ce 3+ ion in the perovskite matrix simplifies in oxidation and reduction of oxygen species which stimulates the oxy-dehydrogenation of benzyl alcohol to benzaldehyde. The Mn 2p 3/2 core level XPS analysis suggests that due to oxygen vacancies, Mn 2+ ions were generated from the Mn 3+ transformation in perovskites. It is observed that La 0.95 Ce 0.05 MnO 3 catalyst shows the highest TON and TOF among all prepared perovskites. According to our observed results the Ce 3+ ion -doped LaMnO 3 materials could serve as potential heterogeneous catalysts for hydrocarbon conversion. Besides that, trivalent cerium ion doping stimulate the synergistic effect

chemsum

{"title": "Physico-chemical properties and catalytic activity of the sol-gel prepared Ce-ion doped LaMnO3 perovskites", "journal": "Scientific Reports - Nature"}

protein_conjugation_with_triazolinediones:_switching_from_a_general_tyrosine-selective_labeling_meth

## Abstract: Selective labeling of tyrosine residues in peptides and proteins can be achieved via a 'tyrosine-click' reaction with triazolinedione reagents (TAD). We have found that tryptophan residues are in fact often also labeled with this reagent. This off-target labeling is only observed at very low levels in protein bioconjugation but remains under the radar due to the low relative abundance of tryptophan compared to tyrosines in natural proteins, and because of the low availability and accessibility of their nucleophilic positions at the solvent-exposed protein surface. Moreover, because TAD-Trp adducts are known to be readily thermoreversible, it can be challenging to detect these physiologically stable but thermally labile modifications using several MS/MS techniques. We have found that fully solvent-exposed tryptophan side chains are kinetically favored over tyrosines under almost all conditions, and this selectivity can even be further enhanced by modifying the pH of the aqueous buffer to effect selective Trp-labeling. This new site-selective bioconjugation method does not rely on unnatural amino acids and has been demonstrated for peptides and for recombinant proteins. Thus, the TAD-Tyr click reaction can be turned into a highly site-specific labeling method for tryptophans. Site-selective protein modification is of utmost importance for many applications from fundamental biology (fluorescent tagging) to therapeutic development (antibody-drug conjugates). 1,2,3,4 While amino acid selectivity can be achieved by exploiting the nucleophilic functionalities of e.g. lysines and cysteines, 5,6 genuine site selectivity depends on their representation density on the protein surface. In this regard, tryptophan (Trp) is an interesting target for native conjugation strategies, with an abundance of just over 1% in proteins. 7 Despite the indole side chain not being the most chemically tractable target, several groups have reported methodologies for selective modification of tryptophan in peptides and proteins. 8,9,10 ,11 Many of these strategies employ transition metal catalyzed reactions and/or conditions limiting downstream biochemical applications. These reactions are typically alkynylations and C-H arylations of the indole. 12,13,14,15,16 Also, Trp sulfenylation was demonstrated for peptide ligation. 17 While Francis and coworkers showed rhodium carbenoid-based Trp labeling at mild pH, 18 this method is dependent on transition metal catalysis and requires long reaction times. An organoradical Trp conjugation was demonstrated on peptides and proteins 19 and even if the method is devoid of transition metals, it requires acidic conditions and is not compatible with buffers. Very recently, a novel biomimetic approach for the selective conjugation of tryptophan was developed, this method however employs UV irradiation and needs to be performed in absence of oxygen. 20 Scheme 1. Prototype reactions for the TAD-Y click, organoradical tryptophan modification (previous work) and TAD tryptophan labeling (this work). In 2010, Barbas and co-workers reported a click like reaction for the more abundant tyrosine (Tyr, 3.3% abundance 7 ) using triazolinedione chemistry, 21 and several ap-plications for protein conjugation followed. 22,23,24,25,26 Interestingly, when exploring this powerful Tyr click reaction on Trp-containing peptides, we observed a high degree of offtarget labeling on Trp residues, even in aqueous buffers. While the swift reaction of indoles with triazolinediones was reported by Baran, Guerrero and Corey in 2003, 27,28 Barbas and co-workers demonstrated that tyrosine labeling is kinetically favored in buffers. However, we now surmise that this competitive Trp-labeling in protein bioconjugation remained under the radar, likely due to a combination of the low abundance and low solvent accessibility of Trp residues. Moreover, in line with observations of Baran and coworkers, we found that indole-TAD modifications have limited thermal stability and can reverse under MS/MS conditions rendering their detection more tedious. We thus cided to more closely examine the competition between Trp and Tyr labeling by TADs in order to probe the potential of TAD reagents for selective Trp-bio-conjugation (scheme 1). For that purpose, tetrapeptides NWAS 1a and 1b were tested in intermolecular competition experiments with phenyltriazolinedione (PTAD 2a) in PBS-buffer at two different pH values, allowing for head to head comparison between Tyr and Trp side chains embedded in the exact same chemical environment (figure 1). Signals for peptide conjugates 2aa and 2ba overlap on the HPLC UV chromatogram, therefore extracted ion chromatograms (XIC's) were used for the analysis. When analyzing the XIC's of the starting peptide-ions NWAS 1a (green) and NYAS 1b (pink) and conjugated peptide-ions NWAS-PTAD 2aa (orange) and NYAS-PTAD 2ba (blue), a pronounced difference can be observed between the reaction at pH 4 and pH 7. Indeed, at pH 4 Trp conjugate 2aa was detected nearly exclusively while at pH 7 a mixture of conjugates was obtained. This observed pH-dependent reactivity of TADs with Tyr is in accord with previous mechanistic studies of the tyrosine-TAD click reaction, which indicate the phenolate species as the prevalent nucleophile. 29 Lowering the pH will effectively decrease the amount of tyrosine-phenolate form and thus decrease the extent of reaction of Tyr with TAD. This was further confirmed using additional peptides (1a-1h, table 1) and TAD-propanol 2b, PTAD-alkyne 2c and fluorescent DMEQ-TAD 2d (Section S2.2.2). It was also observed that, even without competing Trp-peptide present, lowering of pH causes a significant reduction in Tyr-conjugate formation (Section S2.2.1). These findings at peptide level prompted us to look in more detail to earlier reports on the tyrosine click protein modification. Indeed, off-target Trp-labeling was observed earlier at protein level. In the initial study of Ban et al. modification on tryptophan was observed upon myoglobin labeling albeit in a very low amount. 21,25 Furthermore, careful reinterpretation of the MALDI-TOF MS spectra (kindly provided by the authors) of Vandewalle et al., 23 who labeled BSA with butyl-TAD, showed that Trp-modification was indeed noticeable (Section S3.1). These findings demonstrate that researchers can incorrectly assume that tryptophan will not react with TAD-reagents in protein conjugation reactions, possibly leading to flawed interpretation of data. Table 1. Peptide sequences used in this study, structures of TAD reagents 2b, 2c and 2d. ## Entry Sequence 1a Asn Intermolecular competition between 1a and 1c clearly demonstrates the position-sensitivity of the Trp-TAD reaction: the C-terminal tryptophan in 1c is labeled to a 3 times higher extent, as calculated via HPLC peak integration at 214 nm, compared to its internal tryptophan 1a counterpart. This reactivity difference can be attributed to the more exposed reactive center as well as to the presence of the carboxylic acid which can transiently donate a proton to the TAD moiety rendering it even more electrophilic. A second striking difference resides in the nature of the formed adducts. For the C-terminal tryptophan, two peaks for the labeled product 2cb are observed, indicating the formation of isomers. Indeed, we found this adduct had undergone an additional annulation caused by the reaction of the lone pair on the backbone nitrogen with the indole C2 after reaction of TAD with the indole C3. These findings were confirmed via NMR analysis of Boc-Trp-OH and N-Ac-Trp-OMe adducts with TAD-propanol 2b (Section S4) and are in agreement with the results reported by Baran et al. 27 on non-peptide related TAD-indole reactions. In a subsequent series of experiments, we investigated if the observed intermolecular selectivity, translates into intramolecular Trp versus Tyr selectivity. To this end, competition experiments were performed with peptides containing both tyrosine and tryptophan (1i-1l, table 1). MS/MS analyses were done to determine the modification site. We found that the modification on tryptophan is unstable in all tested MS/MS conditions except for ESI in combination with electron transfer dissociation (ETD). ESI-HCD, ESI-CID as well as MALDI-TOF/TOF all largely lead to the loss of the TAD modification on tryptophan. The TAD modification on tyrosine was found to be stable in all tested conditions. These findings are in agreement with earlier work on the thermoreversibility of indole-TAD reactions. 30 Peptide VWSQKRHFGY 1k was labeled using TAD-propanol In conclusion, we report that competitive tryptophan labeling is liable to have so far been systematically overlooked in the current use of triazolinedione (TAD) chemistry for putative tyrosine-selective protein conjugation, a technique which is growing in popularity. The reversibility of the TADtryptophan in MS/MS analysis, in combination with the low abundance and low accessibility of tryptophan side chains likely caused this off-target effect to have remained under the radar. We have found that an exposed tryptophan is in fact kinetically favored over tyrosine in most conditions. Lowering the buffer pH further enhanced the selectivity resulting in a transition metal free, buffer-compatible amino acid specific labeling method for the least abundant natural amino acid tryptophan. Thus, in addition to a better understanding of the factors that govern the click-like TAD-based protein conjugation, its scope has been expanded, and a very interesting new option for native amino acid selective modification has been revealed. The implementation of Trpsubstitutions at protein surfaces or loops can thus be an interesting rational design strategy for fully site-selective labeling of native proteins.

chemsum

{"title": "Protein Conjugation with Triazolinediones: Switching from a General Tyrosine-Selective Labeling Method to a Highly Specific Tryptophan Bioconjugation Strategy", "journal": "ChemRxiv"}

diels–alder_reactions_of_myrcene_using_intensified_continuous-flow_reactors

## Abstract: This work describes the Diels-Alder reaction of the naturally occurring substituted butadiene, myrcene, with a range of different naturally occurring and synthetic dienophiles. The synthesis of the Diels-Alder adduct from myrcene and acrylic acid, containing surfactant properties, was scaled-up in a plate-type continuous-flow reactor with a volume of 105 mL to a throughput of 2.79 kg of the final product per day. This continuous-flow approach provides a facile alternative scale-up route to conventional batch processing, and it helps to intensify the synthesis protocol by applying higher reaction temperatures and shorter reaction times. ## Introduction Over the past years, great attention has been devoted to finding alternative, renewable feedstocks to fossil oil for the production of fuel and industrial chemicals. Especially, high value added products from fine chemicals, specialty chemicals or the pharmaceuticals sector allow for a 'drop-in' replacement of existing, fossil resources based synthesis routes with economic alternatives based on renewable sources. Besides chemical platforms based on sugar, lignin or fatty acid containing feedstocks, terpenes present another plant derived feedstock which is of great interest for a variety of industrial applications, first and foremost in the fragrance and flavor industries, but also in the pharmaceutical and chemical industries . Myrcene is a naturally occurring, acyclic monoterpene which is used industrially for the manufacture of flavoring substances and fragrances; in research it is used as a model compound for a series of different reactions and in the synthesis of complex natural products, including several pheromones . Myrcene is a colorless oil and exists as two isomers, the synthetic α-myrcene, containing an isopropenyl group, and the naturally occurring β-myrcene (which will be referred to in the following only as "myrcene" Scheme 1: Diels-Alder reaction of myrcene (1), with various dienophiles 2. (1), see Scheme 1, vide infra). It can be found in significant quantities (up to 39%) in the essential oils of several plants, such as wild thyme , ylang-ylang , bay leaf , juniper berries , lemongrass , or parsley , and in smaller percentages (<5%) in hops , celery , dill , rosemary , tarragon and nutmeg to name but a few. A review by Behr and Johnen describes the manufacture of myrcene from other terpenes, as well as several synthetic routes based on this versatile and reactive starting material to form alcohols, esters, amines, chlorides, dimers, polymers and even complex natural products, amongst others. At present myrcene (1) is manufactured industrially from turpentine; the distillate of pine resin . One of the main components of turpentine is β-pinene, from which myrcene can be synthesized upon thermal isomerization at temperatures between 400 and 600 °C. This was first described by Goldblatt and Palkin in 1947 . Myrcene is a very versatile molecule that can act as the starting material for several valuable compounds. The industrial production of a series of top-selling flavors and fragrances are based on myrcene, such as geraniol, nerol, linalool, menthol, citral, citronellol or citronellal . The terminal diene moiety present in myrcene allows for a reaction with a suitable dienophile following the Diels-Alder reaction mechanism. Dahill et al. describe the synthesis of the Diels-Alder adduct of myrcene and acrylonitrile for the use as an odorant in the perfume industry . A series of Diels-Alder reactions of myrcene (1) and another sesquiterpene, farnesene, with various dienophiles have been reported by Tabor et al. for the use as solvents and surfactants. The emergence of compact continuous-flow reactors has begun to transform the way chemical synthesis is conducted in research laboratories and small manufacturing over the past few years . In several applications, where reaction times are short and heat management is important, intensified continuous processes inside tubular or plate-type flow reactors can successfully replace batch methodologies classically carried out in stirred glass vessels. We have demonstrated the benefits of this superior heat management in previous work looking at exothermic radical polymerizations in continuous flow . Over the past years, Diels-Alder reactions of isoprene using laboratory-scale flow reactors were studied by different research groups . A continuous-flow reactor can offer a range of benefits over batch processing, with the enhanced heat and mass transfer arguable being one of the most important. In many cases increased control over the process and improvements in product quality are the result. Herein, we describe the synthesis of several Diels-Alder adducts made from myrcene (1) and a series of dienophiles, which contain carboxylic acids, esters or acid anhydrides. In particular, the reaction of myrcene (1) with acrylic acid (2b) was investigated in detail, through batch and continuous-flow methods. The intensified flow process presents a more compact and efficient alternative to classic batch manufacture for the production of Diels-Alder adduct surfactants from myrcene. ## Results and Discussion The solution-phase Diels-Alder reactions presented herein follow the general reaction pathway shown in Scheme 1. The 1 for entries 2.1 to 2.6, as in these experiments R was close to 1 (between 0.9 and 1.1). conjugated diene myrcene (1) was reacted with a series of dienophiles 2 to form the Diels-Alder adducts 3. Before investigating this reaction for continuous-flow processing, we first undertook a series of batch experiments to explore the reactivity of the different dienophiles shown in Scheme 1. These experiments were carried out on a batch microwavereactor system (see experimental section) at temperatures between 100 and 140 °C, and the results are presented in Table 1. a Entries 1.1 to 1.3 were reacted with an initial myrcene concentration, c MYR,0 , of 2.8 mol/L; all entries were reacted with a myrcene to dienophile ratio, R, of 0.9; b conversions were calculated based on NMR. Maleic anhydride (2a) proved to be the most reactive of the dienophiles used in this study with reaction completion occurring after a few minutes at 100 °C. Other activated dienophiles such as acrylic acid (2b) and ethyl acrylate (2f) reached high conversions in excess of 90% after 1 to 5 h and the maleates 2d and 2e required up to 10 h reaction time at 140 °C to reach nearcompletion. The slowest reactions were observed using itaconic acid (2c) and the PEG containing acrylate 2g. Acrylic acid (2b) was selected for further study given our interest in products with surfactant properties, and the preferable reaction kinetics of the acrylic acid-myrcene system. Table 2 presents a set of experiments using this system, at different process conditions and in different solvents; samples were analyzed over time in order to establish kinetic profiles of these reactions. Figure 1 shows the kinetic profiles of the reactions presented in Table 2. All reactions followed an expected trend, asymptotically approaching full conversion with increasing reaction time. While both EtOAc and toluene produced similarly fast kinetic data with conversions around 95% after 40 to 60 min toluene was preferred due to its higher boiling point. Figure 1b shows the influence of temperature and the ratio of starting materials. These experiments also showed trends as were expected. Values for the reaction rate constant, k, calculated from these experiments, are presented in Table 2 and are within expected limits when compared to literature values. More details on the derivation of the k values and the literature references can be found in Supporting Information File 1. After the Diels-Alder reaction was optimized in batch on a small scale (typically 2 mL reaction volume) the process was scaled-up first on a Vapourtec R2/R4 tubular flow reactor to a reaction volume of typically 20 mL and then on a Chemtrix Plantrix ® MR260 plate flow reactor to a reaction volume of typically 200 mL (see also experimental section). The results from these continuous-flow experiments are shown in Table 3. The 10-times scale-up in the tubular flow reactor and the 100 times scale-up in the plate flow reactor resulted in similar, if not slightly higher conversions than the batch experiments (see Figure 2). The two continuous reactors produced highquality material at steady state conditions. The reaction profile in the plate flow reactor was quantified by taking samples at the outlet of the reactor over the entire duration of one experiment. These profiles are very uniform with steep fronts and tails and a flat steady state region, suggesting that the residence time distribution inside the reactor is narrow and close to plug flow. One of these profiles is shown in Figure S4 (Supporting Information File 1). The fastest conditions investigated herein were 30 min in the plate reactor at 160 °C giving 99% conversion of 2b and a yield of 94% of a semi-crystalline product (Table 3, entry 3.9). As part of the scale-up investigations, we also performed the Diels-Alder reaction of myrcene (1) and 2b in a 6 mm i.d. stainless steel tubular flow reactor with a reaction volume of 108 mL. A few minutes after start of the reaction, however, we observed a pressure increase in the reactor which was caused by fouling occurring in the reactor entrance section and ultimately led to complete blockage of the tube at this point. This is believed to be caused by a side reaction of 2b and myrcene (1) forming polymeric material, which built up on the metal walls of the reactor, ultimately leading to the complete blockage. The mechanism and circumstances of this side-reaction are unknown; it only occurred in the stainless steel reactor and not in the PFA tubing of the Vapourtec R-series flow reactor or the silicon carbide module of the plate flow reactor. Hence, it was postulated that a metal catalyzed polymerisation on the stainless steel reactor tubes might have occurred, however, this could not be confirmed. Further details on these observations can be found in Supporting Information File 1. Using 13 C NMR an approximate ratio of the two isomers, 3-3 and 3-4 (see Figure 2), was calculated for the continuous-flow reactions performed between 140 and 160 °C (see Table 3). The amount of Diels-Alder adduct with the carboxylic acid located in the 3-substituted position, 3-3, was always larger than the 4-substitituted adduct, 3-4, with an average 3-3/3-4 ratio of 7:3 (3-substituited adduct was between 68 and 71%). For Table 3, entry 3.9, the yield of the semi-crystalline product after solvent removal was 94%. The production capacity (PC) and the space time yield (S.T.Y.) can be calculated based on the amount of isolated product, m P , using Equations 1 and 2. ( Here, is the total volumetric flow rate through the reactor, V SS the combined volume of both stock solutions and V R the volume of the flow reactor. Running the plate reactor at 160 °C (Table 3, entry 3.9), we managed to achieve a production capacity of 116.3 g/h, which equates to an S.T.Y. of 1.11 kg L −1 h −1 . Parallel to the scale-up in the plate flow reactor, we also scaled up the process in batch to a 6 L scale using a jacketed stirred tank reactor. Here, the reaction was run for ~10 h at 100 °C in order to reach completion, compared to only 30 min at 160 °C in continuous flow. Preliminary experiments were carried out looking at the surfactant properties of the Diels-Alder adduct of myrcene (1) and 2b. The results were promising and showed that the product was able to stabilize emulsions for several hours compared to several seconds or minutes in the control experiments without the Diels-Alder adduct. Further details on these surfactant tests are presented in Supporting Information File 1. ## Conclusion We have investigated the Diels-Alder reaction of myrcene (1) with a range of different dienophiles at temperatures between 100 and 160 °C. The Diels-Alder reaction of myrcene (1) with acrylic acid (2b), yielding a carboxylic acid containing surfactant, was scaled-up in a plate-type continuous-flow reactor and a batch stirred tank. The use of continuous-flow processing allows for an efficient synthesis of large quantities of the Diels-Alder adduct and we managed to scale-up the reaction of myrcene (1) with acrylic acid (2b) inside the 105 mL flow reactor to a throughput of 2.79 kg of the final product per day. The small dimensions of the fluidic channels inside the tubular and the plate-type flow reactors ensured that heat and mass transfer were efficient and fast, and that the reaction could be operated under 'quasi isothermal' conditions (i.e., with negligible deviations from the set temperature in the entire bulk reaction volume of the reactor). This resulted in a much more uniform reaction profile than in batch stirred tanks, allowing for a much shorter reaction time than classically applied in batch operations. ## Experimental Materials and analysis The reactants myrcene (1, 90% purity), maleic anhydride (2a), acrylic acid (2b), itaconic acid (2c), dimethyl maleate (2d), ethyl acrylate (2f) and poly(ethylene glycol) methyl ether acrylate (PEGA, 2g) were obtained from Sigma-Aldrich; bis(2ethylhexyl) maleate was provided by TriTech Lubricants. The solvents tetrahydrofuran (THF), ethyl acetate (EtOAc), toluene, dichloromethane (DCM) and isopropanol (iPrOH) were obtained from Merck KGaA. All reagents and solvents were used without further purification. Reaction conversions were calculated from 1 H NMR spectra, which were recorded on a Bruker AC-400 spectrometer in deuterated chloroform (from Cambridge Isotope Laboratories Inc.). Conversion calculations were based on clearly identifiable and non-convoluted peaks of remaining starting material and generated product. The residual solvent peak at δ = 7.26 ppm was used as an internal reference. Product compositions were analyzed by GC-FID and GC-MS; details for both can be found in Supporting Information File 1. The GC-FID results were also used to confirm NMR conversions and to calculate GC-based yields. ## Batch Diels-Alder reaction The following procedure is typical for the preparation of the Diels-Alder adduct of myrcene (1) and a series of different dienophiles. A reactant solution of myrcene (1, 811 mg of myrcene stock solution with a 90% purity, 5.36 mmol of myrcene), 2b (429 mg, 5.95 mmol), in EtOAc (0.49 mL), was premixed and filled into a sealed microwave vial. The reaction was conducted in a laboratory microwave reactor (Biotage Initiator) at 140 °C with a reaction time of 2 h. A transparent, faintly yellow solution was obtained after reaction, from which the conversion was determined by 1 H NMR. The solvent was evaporated under reduced pressure to yield a yellow semi-crystalline paste. Detailed reaction conditions and reagent compositions for each batch experiment can be found in Table 1 and Table 2. For kinetic studies, small samples of the reaction mixture for 1 H NMR were withdrawn through the septum of the microwave reactor glass vial using a syringe. For this the microwave reaction was stopped at various points in time over the course of the reaction, namely at 20, 40, 60 and 120 min. ## Continuous-flow Diels-Alder reaction using a Vapourtec R2/R4 flow reactor The following procedure is typical for the preparation of the Diels-Alder adduct of myrcene (1) and acrylic acid (2b) in a tubular flow reactor. Two reactant solutions were prepared, one containing myrcene (16.22 g of myrcene stock solution with a 90% purity, 107.16 mmol of myrcene) in EtOAc (1.98 mL), and the other containing 2b (8.58 g, 119.06 mmol), in EtOAc (7.75 mL). The two solutions were continuously mixed in a T-piece and then fed into a Vapourtec R2/R4 flow reactor setup , consisting of two 1.0 mm i.d. perfluoroalkoxy alkane (PFA) reactor coil modules in series (10 mL each -total reactor volume: 20 mL). The pump flow rate of the myrcene solution was set to 0.3 mL•min −1 , the pump flow rate of the acrylic acid solution was set to 0.2 mL•min −1 . This resulted in a total flow rate of 0.5 mL•min −1 and a mean hydraulic residence time of 40 min inside the two PFA reactor coils (the mean hydraulic residence time is defined as 'flow rate/reactor volume'). The reaction was conducted at 140 °C. The product, a transparent, faintly yellow solution, was collected at the reactor outlet, after passing through a 75 psi back-pressure regulator. From this solution, the reaction conversion was determined by 1 H NMR. Afterwards, the solvent was evaporated under reduced pressure to yield a yellow semi-crystalline paste. Detailed reaction conditions and reagent compositions for each experiment in the tubular flow reactor can be found in Table 3. ## Continuous-flow Diels-Alder reaction using a Chemtrix MR260 flow reactor The following procedure is typical for the preparation of the Diels-Alder adduct of myrcene (1) and acrylic acid (2b) in a silicon carbide plate-type flow reactor. Two reactant solutions were prepared, one containing myrcene (208.2 g of myrcene stock solution with a 90% purity, 1.375 mol of myrcene) in toluene (21.2 mL), and the other containing 2b (90.1 g, 1.250 mol), in toluene (80.1 mL). The two feed solutions were pumped using two Teledyne Isco D-series dual syringe pumps (100 DX, with Hastelloy™ syringes) and were continuously mixed in a T-piece. After mixing, the combined starting material solution was fed into a Chemtrix Plantrix ® MR260 plate-type flow reactor. This plate flow reactor configuration consisted of a series of 3M™ silicon carbide microstructured plates (see also Figures S2 and S3 in Supporting Information File 1), which was thermally regulated by a Lauda Integral XT 150 heater/chiller unit. The total reactor volume was 105 mL. An SSI Prep 100 dual piston pump with PEEK pump heads was used to flush the reactor before and after the reaction with toluene. The pump flow rate of the myrcene solution was set to 2.21 mL•min −1 , the pump flow rate of the acrylic acid solution was set to 1.30 mL•min −1 . This resulted in a total flow rate of 3.51 mL•min −1 and a reaction time of 30 min inside the plate flow reactor. The reaction was conducted at 160 °C. The product, a transparent, faintly yellow solution, was collected at the reactor outlet, after passing through a stainless steel Swagelok ® R3A series adjustable high pressure valve. This valve was used as a back pressure regulator, in order to set the pressure inside the reactor to between 8 and 10 bar (116 to 145 psi) during operation. From the resulting product solution, the reaction conversion was determined by 1 H NMR. Afterwards, the solvent was evaporated under reduced pressure to yield a yellow semi-crystalline paste. Detailed reaction conditions and reagent compositions for each experiment in the plate-type flow reactor can be found in Table 3.

chemsum

{"title": "Diels\u2013Alder reactions of myrcene using intensified continuous-flow reactors", "journal": "Beilstein"}

organocatalytic_reductive_coupling_of_aldehydes_with_1,1-diarylethylenes_using_an_<i>in_situ</i>_gen

## Abstract: Organocatalytic reductive coupling of aldehydes with 1,1-diarylethylenes using an in situ generated pyridine-boryl radical A pyridine-boryl radical promoted reductive coupling reaction of aldehydes with 1,1-diarylethylenes has been established via a combination of computational and experimental studies. Density functional theory calculations and control experiments suggest that the ketyl radical from the addition of the pyridine-boryl radical to aldehyde is the key intermediate for this C-C bond formation reaction. This metal-free reductive coupling reaction features a broad substrate scope and good functional compatibility. Organocatalytic reductive coupling of aldehydes with 1,1-diarylethylenes using an in situ generated pyridine-boryl radical † Introduction Carbon-carbon bond formation is the most important transformation in organic synthesis. 1 The catalytic reductive coupling of olefns with carbonyl compounds is one of the most economical C-C bond constructing methods, due to the abundant source of olefns and carbonyl compounds. 2 Traditionally, transition metal catalysts have played privileged roles in these transformations, including metal-catalyzed C]O reductive coupling (Scheme 1, top) and redox-triggered C]O coupling via H 2 transfer (Scheme 1, middle). 4 However, sensitive organometallic reagents or transition-metal catalysts are usually required in these reactions. In contrast, organocatalytic reductive coupling of olefns with carbonyl derivatives for C-C bond formation in the presence of sensitive functional groups or congested structural environments is still rare. 5d Boron containing radicals are important reactive intermediates in organic synthesis. In this context, our group recently revealed that the pyridine-ligated boryl radical (Py-Bpinc) could be readily generated from (pinacolato)diboron (B 2 pin 2 ) through a cooperative catalysis involving two 4-cyanopyridine molecules. 11 This kind of pyridine-boryl radical was used for the catalytic reduction of azo-compounds 11 or as a carbon-centered radical for the synthesis of 4-substituted pyridines. 12 Moreover, the pyridine-boryl radical can act as a persistent radical 13 for the synthesis of organoboronate derivatives. 14 Because the precursors (pyridines and B 2 pin 2 ) of these pyridine-boryl radicals are inexpensive and stable, 15 the development of new chemical transformations with these pyridine-boryl radicals is attractive. In this work, we further explored pyridine-boryl radical chemistry in the organocatalytic reductive coupling of aldehydes with 1,1-diarylalkenes (Scheme 1, bottom), which, to the best of our knowledge, has not been reported previously. ## Results and discussion It will be shown that the reductive coupling of aldehydes and olefns can be promoted by an in situ generated pyridine-boryl radical, following the proposed pathway as shown in Scheme 2. The proposed catalytic cycle consists of the following four steps: (1) activation of the B-B bond of B 2 pin 2 by pyridines to form a pyridine-boryl radical (Int1); (2) the addition of the pyridineboryl radical to aldehyde 1a to generate a new ketyl radical (Int3), with the regeneration of the pyridine catalyst; (3) the Scheme 1 Reductive coupling of carbonyl compounds with olefins. a Key Laboratory of Mesoscopic Chemistry of Ministry of Education, Institute of Theoretical and Computational Chemistry, School of Chemistry and Chemical Engineering, Nanjing University, Nanjing, 210093, P. R. China. E-mail: shuhua@ nju.edu.cn addition of the new ketyl radical to 1,1-diphenylethylene to yield a diaryl-stabilized radical species (Int4); and (4) the hydrogen abstraction of Int4 from an appropriate H-source to yield the fnal reductive coupling product. In addition, one molecule of Int4 may also abstract a hydrogen atom from another molecule of Int4 to give the reductive coupling product and another disproportionation product. To make this catalytic cycle happen, it is necessary to inhibit the possible radical-radical C-C coupling reaction between the pyridine-boryl radical and the ketyl radical, as observed between a,b-unsaturated ketones and 4-cyanopyridine in the presence of B 2 pin 2 . 12 Thus, other pyridines with different substituents may be better catalysts than 4cyanopyridine for the proposed reaction. With a pyridine-boryl radical bearing a suitable substituent, its reactivity might be tuned so that the newly generated ketyl radical could react with 1,1-diphenylethylene to yield a diaryl-stabilized radical species, which then undergoes a hydrogen atom abstraction from an appropriate hydrogen source to produce the reductive coupling product. To fnd suitable pyridines which can react with B 2 pin 2 to form the corresponding pyridine-boryl radical under mild conditions, we frst performed density functional theory (DFT) calculations with the M06-2X 16 functional to screen a series of pyridines. A careful analysis of stationary points revealed that the formation of the pyridine-boryl radical proceed through a -sigmatropic rearrangement/homolytic C-C bond cleavage pathway 17 rather than via the direct homolytic cleavage of the B-B bond 11,18 In order to determine a suitable combination of a pyridine catalyst and a hydrogen source, we conducted an initial investigation on the reaction between isobutyraldehyde 1a and 1,1diphenylethylene 2 (see Tables S1 to S3 † for details). As shown in Table 1, by heating a mixture of isobutyraldehyde 1a (1.0 equiv.), 1,1-diphenylethylene (2.0 equiv.), and B 2 pin 2 (1.0 equiv.) in the presence of 1,3,5-trimethyl-1,4-cyclohexadiene (a hydrogen source, 1.0 equiv.) and 4-cyanopyridine A (0.2 equiv.) in tert-butyl methyl ether (MTBE) at 120 C, the desired reductive coupling product 3a was observed in 28% yield (entry 1), together with a small amount of pyridine-aldehyde adducts (12% yield, see the ESI † for details). When 4-(4-pyridinyl)benzonitrile B was used as the catalyst (entry 2), the NMR yield of 3a improved to 78%, and the yield of a byproduct 3a 0 from the disproportionation of the diaryl radical intermediate (Int4) was 6%. However, when other pyridines (for example C, D, or E, entries 3-5) were adopted, the yield of 3a decreased signifcantly. If Et 3 SiH was chosen as the hydrogen source, the yield of 3a is somewhat lower than that with 1,3,5-trimethyl-1,4-cyclohexadiene as a hydrogen source (entry 6). In the absence of a hydrogen source (entry 7), the ratio of 3a/3a 0 was 52% : 16%, suggesting that the addition of a hydrogen source is important for improving the yield of 3a (see Table S2 Under the optimum conditions (Table 1, entry 2), we explored the generality of this transformation with a series of alkyl and aryl aldehydes. As shown in Table 2, the reductive coupling reactions of several fully aliphatic aldehydes proceeded with good efficiency (1a-1d). It was noteworthy that aldehydes with C]C double bond (1e), methylthio (1f), or furyl (1h) functionalities on the alkyl chain were tolerated, giving the reductive coupling products in moderate to good yields. The abranched aldehydes (1i-1r), in particular, pivaldehyde (1q) and 1-adamantylcarboxaldehyde (1r), also reacted well to afford the desired products in good yields. It should be mentioned that the substrates with a congested structure environment show less reactivity in transition-metal catalyzed reductive coupling of olefns and aldehydes, possibly because the coordination between the metal centre and the corresponding substrates is difficult. 4c However, our method is also suitable for butyl aldehydes (1q and 1r). Beside alkyl aldehydes, aryl aldehydes (1s and 1t) bearing electron-donating groups (CH 3 and CH 3 O) could also serve as the coupling partners, furnishing corresponding products in moderate yields. In addition to aldehydes, alkyl ketones (1u-1w) also reacted smoothly to provide the desired alcohols in 27-42% yield. Diarylalkanes are important pharmacophores in drugs. 19 It would be attractive to apply this metal-free method in the late stage functionalization of medicinally related molecules. As shown in Table 2(C), an abietic acid derivative (1x) and gemfbrozil derivative (1y) reacted smoothly with 1,1-diphenylethylene to form 3x and 3y in acceptable yields, respectively. Next, the scope of 1,1-diarylalkenes (4) was examined (Table 3(a)). Both symmetrical (4a-d) and unsymmetrical (4e-o) 1,1diarylalkenes were converted into the corresponding products 5 in moderate to good yields with modest diastereoselectivities. The reaction tolerated substrates bearing various functional groups on the benzene ring, such as halogen functionalities (4c and 4d), CF 3 (4e), CN (4f and 4g), MeO (4h), CH 3 S (4i), CO 2 Me (4j), and tBu (4k). More importantly, 1,1-diarylalkenes containing heterocyclic structures (4m-o), such as benzofuran (4o) and thioxanthene (4p), also reacted smoothly to give the expected products in reasonable yields. Additionally, we also tested the reactivity of other alkenes with pivaldehyde 1q (Table 3(b)). However, our results show that other alkenes, including ethyl 2phenylacrylate (4q), styrenes (4r and 4s) or aliphatic olefn (4t), generally gave little or no desired product. The reason why 1,1diarylalkenes are suitable coupling partners of ketyl radicals may be due to (1) the radical stabilization effect of two aryl groups, and (2) the less nucleophilicity of present boron-ketyl radicals (compared with typical ketyl radicals). 5b Thus, this protocol provides a metal-free reductive coupling method of 1,1diarylalkenes with aldehydes (via the radical addition mechanism), which traditionally requires transition metal catalysts or organometallic reagents. To understand the mechanism of the reductive coupling of 1,1-diarylalkenes with aldehydes, we have performed DFT calculations with the M06-2X functional to explore the free energy profle of the proposed mechanism for the reaction between isobutyraldehyde (1a) and 1,1-diphenylethylene (2) in the presence of Int1 as a reactive intermediate. Our theoretical studies have shown that the generation of Int1 from B 2 pin 2 and 4-(4-pyridinyl)benzonitrile is exergonic by 13.4 kcal mol 1 (see Fig. S4 †). The calculated free energy profle and transition state structures are displayed in Fig. 1 (the optimized structures of all minimum species are shown in Fig. S12 †). First, the coordination of the oxygen atom of isobutyraldehyde to the boron atom of the pyridine-boryl radical Int1 generates a boron-containing intermediate (Int2) via TS1, with a barrier of 13.3 kcal mol 1 . Then, the breaking of the B-N bond in Int2 yields a ketyl radical (Int3) and regenerates the 4-(4-pyridinyl) benzonitrile catalyst. This process is exothermic by 4.5 kcal mol 1 , with a barrier of 3.2 kcal mol 1 (relative to Int2), suggesting that the formation of the ketyl radical (Int3) from Int2 is possible. Next, the addition of Int3 to the b-position of 1,1diphenylethylene to form a diaryl-stabilized radical (Int4) via TS3 is exothermic by 15.6 kcal mol 1 , with a barrier of 15.5 a Reaction conditions: isobutyraldehyde (0.2 mmol), B 2 (pin) 2 (0.2 mmol), catalyst (0.04 mmol), 1,1-diphenylethylene (0.4 mmol), H-donor (0.2 mmol), 24 hours, 120 C, and MTBE (1 mL). b Yields were determined by 1 H-NMR analysis of the crude mixture using CH 2 Br 2 as the internal standard. c Isolated yield of 3a. TMe-1,4-CHD ¼ 1,3,5-trimethyl-1,4-cyclohexadiene. kcal mol 1 (with respect to the radical Int3). Finally, the fnal product is obtained with a hydrogen atom abstraction from 1,3,5-trimethyl-1,4-cyclohexadiene via TS4 with a barrier of 26.3 kcal mol 1 (relative to the radical Int4). The whole reductive coupling reaction is exergonic by 11.7 kcal mol 1 (with respect to the reactants 1a and Int1). These results suggest that the studied reaction is thermodynamically and kinetically feasible under the experimental conditions. In addition, our calculations suggest that the direct single electron transfer (SET) process between the pyridine-boryl radical and isobutyraldehyde is highly endergonic (see details in Fig. S13 and S14 †). Thus, the SET mechanism for the present reaction can be excluded. Table 2 Substrate scope for the reductive coupling of aldehydes or ketones with 1,1-diphenylethylene a a Reaction conditions: aldehyde (0.2 mmol), B 2 (pin) 2 (0.2 mmol), 4-(4pyridinyl)benzonitrile (0.04 mmol), 1,1-diphenylethylene (0.4 mmol), 1,3,5-trimethyl-1,4-cyclohexadiene (0.2 mmol), MTBE (1.0 mL), 24 h, and 120 C. Isolated yield. The diastereoselectivities (d. r.) were determined by 1 H-NMR analysis of the crude mixture. Boc ¼ tertbutoxycarbonyl. Table 3 Substrate scope for the reductive coupling of pivaldehyde with 1,1-diarylethylenes a a Reaction conditions: pivaldehyde (0. In addition to DFT calculations described above, we also conducted several experiments to verify the proposed pathway. First, the EPR signal was observed for the reaction of 4-(4-pyridinyl)benzonitrile and B 2 (pin) 2 , which supports the formation of the proposed pyridine-boryl radical, as shown in Scheme 3a. Second, the involvement of the ketyl radical was confrmed by a competition experiment (Scheme 3b). It has been reported that thiols are quick hydrogen atom donors that can interfere with the radical reaction. 20 When the hydrogen source 1,3,5-trimethyl-1,4-cyclohexadiene was replaced by 3methylbenzenethiol, the ketyl radical quickly abstracted a hydrogen atom from 3-methylbenzenethiol to yield the reductive product, 3-phenyl-1-propanol, so that its addition to 1,1-diphenylethylene (to form the reductive coupling product) was inhibited (see Page S20 †). This result clearly indicated the involvement of the ketyl radical. Third, the generation of the radical species Int4 (or its analogues) via the addition of the ketyl radical to the b-position of arylethene was confrmed by an intermolecular trapping experiment (Scheme 3c). When 2vinylpyridine and trimethylacetaldehyde were subjected to the standard reaction conditions, species 6 could be detected by HRMS analysis for the crude reaction mixture (see Page S21 †). This result suggests that in this reaction, the radical species Int4-like was further trapped by another 2-vinylpyridine molecule. However, in the presence of 2-vinylpyridine as a substrate, the yield of 6 is quite low and its isolation from the reaction mixture was not successful. Besides, we further conducted analysis of the 11 B-NMR spectrum and HRMS to detect the formation of the proposed O-boron intermediate (Int6, Fig. S17 †). The 11 B-NMR of the crude reaction mixture displays resonances at $21 ppm, which is consistent with the signal of a boron atom bound to three oxygen atoms. 21 In addition, our HRMS analysis (with 4-vinylpyridine as the substrate) also indicates the formation of the O-boron intermediate (Int7), as shown in Fig. S18. † Moreover, we have performed a radical-clock study using cyclopropanecarboxaldehyde as the substrate. The experimental results indicate that some ketyl radicals frst convert into the corresponding carbon radicals (via a ring-opening process) and then add to the alkene to form the ring-opening product (Scheme 3d). The experiments described above provide strong evidence on the involvement of a radical addition step between the ketyl radical and 1,1-diarylethylene in this reaction. , 2018, 9, 3664-3671 This journal is © The Royal Society of Chemistry 2018 ## Conclusions In summary, we have established the organocatalytic reductive coupling of aldehydes with 1,1-diarylalkenes via a combination of computational and experimental studies. This study showed that 4-(4-pyridinyl)benzonitrile is a suitable catalyst for cleaving the B-B bond of B 2 pin 2 , and the ketyl radical from the addition of an in situ generated pyridine-boryl radical to aldehydes is a key intermediate for the C-C bond formation. The reaction is practical and applicable to a broad range of aldehydes and 1,1diarylalkenes with good functional group tolerance. DFT calculations and control experiments were conducted to verify the proposed mechanism. This pyridine-boryl radical promoted radical addition mechanism represents a metal-free reductive coupling reaction of aldehydes with 1,1-diarylalkenes. Further studies will be directed toward the development of new transformations involving readily formed pyridine-boryl radicals with the aid of combined theoretical and experimental studies. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Organocatalytic reductive coupling of aldehydes with 1,1-diarylethylenes using an <i>in situ</i> generated pyridine-boryl radical", "journal": "Royal Society of Chemistry (RSC)"}

evaluating_performance_and_cycle_life_improvements_in_the_latest_generations_of_prismatic_lithium-io

## Abstract: The last decade has seen an enormous improvement of energy density for lithium-ion battery cells, particularly for automotive grade cells intended for use in electrified vehicles. This has led to vastly improved range for battery electric vehicles as well as for plug-in hybrids. However, the challenge of uncertain battery lifetime remains. The ageing effect due to fast charging is especially difficult to predict due to its non-linear dependence on charge rate, state-of-charge and temperature. We here present results from fast charging (1C and 3C in a 20 % to 80 % SOC-level) of several energy-optimized, prismatic lithium-ion battery cell generations utilizing NMC/graphite chemistry through comparison of capacity retention, resistance and dQ/dV analysis. Considerable improvements are observed throughout cell generations and the results imply that acceptable cycle life can be expected, even under fast charging, when restricting the usage of the available battery capacity. Even though this approach reduces the useable energy density of a battery system, this trade-off could still be acceptable for vehicle applications where conventional overnight charging is not possible. The tested cell format (the VDA PHEV2-standard) has been used for a decade in different electrified vehicles. The ongoing development and improvement of this cell format by several battery cell manufacturers suggests it will continue to be a good choice for future vehicles. ## Introduction Electrified vehicles such as hybrid electric vehicles (HEV), plug-in hybrid electric vehicles (PHEV), battery electric vehicles (BEV) and fuel cell electric vehicles (FCEV) are now established on the market. The powertrain technology has been proven for both passenger cars and commercial vehicles such as buses and trucks, despite relatively high component costs when compared to conventional vehicles. Despite very rapid development in terms increasing energy density and decreasing cell cost, the battery remains the critical component in EVs when evaluating total cost, service life and performance. Li-ion battery cells experience ageing, i.e. loss of capacity and power, during both storage (calendar ageing) and usage (cycle ageing) due to several ageing mechanisms for which the individual reaction rates depend strongly on operating conditions such as temperature, charge level (state-of-charge, SOC), charge/discharge currents and load cycle characteristics. In addition, many of the ageing mechanisms are inter-dependent. The charge rate can in many cases be regarded as the dominant ageing factor, especially for energy-optimized cells cycled within large state-of-charge (SOC) windows, as are typical for EVs and PHEVs. At the same time, fast-charging capability is widely desired for both passenger cars and commercial vehicles. Subsequently, the vehicle industry needs to develop accurate ageing models to develop robust battery systems and to forecast actual service life for the batteries. The Swedish automotive industry and several universities have, within continuous research collaboration, carried out a number of studies on ageing of automotive grade Li-ion battery cells, through cycling and post-mortem analysis. In the first phase, focused on LFP/Graphite battery cells, it was shown that ageing may be severely non-uniformly distributed within the battery cells, especially for those that have been exposed to high charge rates. This distribution was observed both laterally across the jelly roll surface and throughout the depth of the electrodes . In this cell type, the graphite electrode was particularly affected by severe ageing. . The uneven ageing was argued to be a consequence of inhomogeneous current distribution due to the tab positions, as well as internal variations in temperature, electrolyte wetting and pressure. In the second phase of the research collaboration, the ageing effects of fast charging on NMC111/Graphite cells were studied. From careful post-mortem analysis it could be concluded that different ageing mechanisms dominate, depending on the charging rate . While NMC particle cracking was observed at 1-2C charging rate, lithium plating likely reduced lifetime at 3C charging rate, and gas evolution rapidly killed the cells cycled with 4C charging rate. In a separate study of a similar cell , it was observed by on-line mass spectrometry that the high ageing rate under fast charging was associated with the evolution of large quantities of ethylene gas. Even for NMC111/graphite cells, heterogeneous ageing within the cell was observed. For instance, a difference in direct current internal resistance (DCIR) was observed among samples harvested from the curved and the flat regions of the prismatic cell jelly roll, pointing towards a nonuniform distribution of mechanical pressure affecting the local ageing . The negative electrode is known to be a major bottleneck for practical fast charging rates (i.e. less than 1 hour charging for energy-optimized cells) . This is mainly due to the slow kinetics of lithium insertion into graphite, the typical negative electrode material of choice for energy-optimized cells. If graphite particles are unable to intercalate lithium ions at sufficiently high rate, then lithium deposits onto the particles as tree-like, highly-reactive, metallic dendrites, a process referred to as lithium plating. These dendrites can grow through the separator and hence cause internal short-circuits in the cell or, in less severe cases, cause accelerated electrolyte degradation and subsequent loss of cyclable lithium and increase of cell resistance. In the past, large-format cells have been seldom studied, probably due to the high cost and scarce availability of this type of automotive grade cells. In this work however, we compare performance and ageing for several generations of automotive grade cells. Results are summarized from cycle ageing of three different NMC111/graphite cells performed within this research consortium over the course of several years. The comparisons herein emphasize the progress between the different generations of energy optimized cells. The capacity retention curves after cycling under 1C and 3C charge (which correspond to fully charging the battery in 60 and 20 minutes, respectively) are compared, as well as corresponding DCIR and dQ/dV analyses. Finally, we discuss the progress in performance and durability of this type of cells and include preliminary comparison with a nickel-rich NMC811/graphite cell type cycled within the present third phase of the collaboration. ## Experimental Three different types of prismatic battery cells of the VDA (Verband der Automobilindustrie) PHEV2 format (148 x 26.5 x 91 mm) were cycled at the same temperature and in the same SOC-region. The electrode active material is NMC111/graphite for the three tested cell types. Cycling was done in a 20 % to 80 % SOC window, with constant current charge and discharge currents. The charge and discharge time required to stay in the SOC-window was recalculated every 200 th cycle from periodically performed capacity measurements. In addition to the periodic capacity measurements, DCIR measurements were also performed, followed by constant voltage adjustment to 80 % SOC before the next cycling period. Two cells of each cell type were cycled at 1C/1C (charge/discharge) current and two cells per cell type at 3C/1C. One additional cell type, called D, in the same format but with the electrode active material NMC811/graphite is used as reference but was not included in the original test matrix. This cell type has been cycled under other conditions and detailed results from that work will be published separately. Additional information about all cell types is shown in Table 1. For the cell type C, reference cells were also stored in different temperatures during the testing period in order to measure the calendar ageing separately. The conditions for the cell cycling and reference performance testing of cell types A and B, which were cycled at the same test facility, are described in detail in previous work . The cycling of cell type C was done at another test facility under the following conditions. Cycling and performance testing were done on a Maccor Series 4000 cell tester. All cells were placed in a climate chamber operating at an average temperature of +33 ± 1 °C during testing to obtain an average cell skin temperature of +35 °C (measured individually with surface-mounted thermocouples). Steel plates were attached to all cells during cycling to maintain external cell pressure according to supplier recommendation. The applied test protocol was the same as for cell types A and B. ## Capacity The capacity fade behaviour of cell types A, B and C differs from each other in several ways. A clear difference in ageing characteristics is seen in the appearance of the capacity retention curves. Comparison of different slopes in the capacity retention curves can be done by fitting curves based on coulombic efficiency (CE) calculations. Coulombic efficiency has been used in earlier research as a tool for analysing ageing . If a constant CE is assumed, the capacity retention can be estimated according to following equation: where Q is the capacity retention, ƞ is the coulombic efficiency and Neq is the corresponding number of equivalent full cycles. Using this approach, it is possible to identify three different CEslopes for the capacity retention of cell type A at 1C charge current, as seen in Figure 1a. Also for cell type B, three different CE-slopes for the capacity retention are identified under 1C charging, depicted in Figure 1c. Both cell types experience an increasing coulombic efficiency under 1C charging, hence showing a regressive or decelerating aging behaviour beginning after a couple hundred equivalent full cycles. The capacity retention for cell type A can be fitted with a single CEcurve for the 3C charging case, as seen in Figure 1b. It can be noted that the estimated coulombic efficiency in this case is similar to the estimated CE1 for the 1C cycled cell type A in Figure 1a. A similar behaviour is also seen for cell type B under 3C charge; its fitted CE-curve also has an efficiency corresponding to the first fitted CE-curve for the 1C cycled cells. However, in this case there is also a sudden change in slope at around 80% remaining capacity, where the estimated coulombic efficiency decreases drastically (CE2 in Figure 1d). The capacity retention curves for cell type C are very similar to that of cell type B (Figure 1e and f), i.e. a decelerated ageing rate for the first few hundred cycles, followed by ageing at a constant CE-rate between 90 % and 80 % capacity retention. In addition, there are signs of accelerating ageing towards EOL for the cells cycled with the 3C charge strategy, where the overall lifetime is only about one third of that for the corresponding cells cycled with 1C charging. In summary, the results from the capacity fade analysis show three typical regimes: decelerated, constant and accelerated ageing, i.e. sudden fade close to end-of-life, which is in line with similar published work by Waldmann et al. . Using CE measurements on cells at the beginning of cell life could hence in some cases be helpful to predict battery cell cycle life, but there are cases where this type of extrapolation could either overestimate or underestimate battery lifetime. Regarding the test cases in this study, CE measurements at the beginning of the test could have been useful for predicting ageing of the cells cycled with the 3C charging regime, but they would have underestimated lifetime for cells cycled with the 1C charging regime. The accuracy and relevance of these CE measurements are greatly affected by the temperature and charge rate of each test . Such characterization at BOL would also be very limited in its ability to predict subsequent changes in CE in the different ageing regimes. The effective lifetime of the cells can be discussed as their cyclability, or, the number of cycles to reach 80 % SOH. A general improvement in cyclability was observed through the three subsequent cell candidate generations, with cell type A having the lowest cyclability and cell type C having the highest. However, cell type B seems to have the largest spread between tested cell pairs at 3C charging which could be related to cell manufacturing or slight variations in test conditions. Common for all three cell types is that 3C charging rates result in lower cyclability relative to 1C charging rates. The largest difference can be seen for cell type C that only shows around one third of the cycle life at 3C charging compared to 1C charging. In addition, for cell type C, a time dependent aging mechanism was also measured by means of a calendar aging test at 50 % SOC from which the results are presented in Figure 2. All cycled cells also showed swelling at end of testing, especially cells cycled at high charge C-rates. As expected, calendar ageing of this type of lithium-ion cell has a capacity loss dependency related to the square root of time , with an Arrhenius temperature correction. ## Direct current internal resistance Direct current internal resistance (DCIR) was also measured on all cells throughout the testing. These measurements were done at a constant current pulse for short duration which captures information about electronic and ionic resistance of the cells. Cell type A has a steadily increasing DCIR throughout cycling, both for 1C and 3C charging, while cell types B and C reach an DCIR minimum after some hundred equivalent full cycles. This decrease in DCIR during early cycles is seldom reported in literature and could be related to several possible causes, of which one could be how the formation process differ between cell candidates . Certain electrolyte additives or protective layers may also be used to stabilize the cell for storage before sale and use. The decrease may be related to the consumption of these sacrificial additives or some other activation behaviour inside the cell (e.g. porosity increase). As this minimum was not observed for the oldest cell type A, this likely reflects recent advances in cell materials, design, and production. For cell type A it hence seems possible to correlate capacity loss to DCIR rise with a linear fit under the conditions applied. However, it is seen from Figure 3 that despite this correlation, the spread between corresponding cells of cell type A is large. This makes estimation of remaining battery capacity from DCIR data challenging in a real-life vehicle application. The relationship between DCIR and capacity retention is not linear for cell types B and C. However, after the point where the cell DCIR reached the minimum, it could be possible to find a correlation between DCIR and capacity retention. This behaviour makes it considerably more complicated to apply a model for battery capacity estimation from DCIR measurements alone in a real-life vehicle application. ## Qualitative capacity loss analysis Incremental capacity analysis, i.e. analysing the inverse derivative (dQ/dV) of charge and discharge voltage curves versus voltage, has been demonstrated as a valuable tool to analyse ageing of the negative and positive electrode respectively . This method was applied to all test cases and is presented for beginning of life and end of life in Figure 4. Peaks dQ/dV curves relate to phase equilibrium of active electrode material (voltage plateaus in voltage versus capacity plots) . For the cell type NMC111/graphite, two peaks are dominant, one at around 3.5 V mostly related to the negative active electrode material graphite and one at around 3.65 V mostly related to the positive active electrode material NMC111. For the 1C cycled type A cells depicted in Figure 4a, the low-voltage peak has moved to higher cell voltage at EOL, indicating loss of cyclable lithium However, the behaviour of the peak at 3.65 V is quite complicated, and could be linked to several different ageing phenomena. The loss of peak area could be related to a loss of NMC active material, though cursory analysis of the corresponding dV/dQ curves shows no strong evidence for this. The decrease and shift in this 3.65 V peak is likely due primarily to the loss of cyclable lithium. (a) (b) ## dQ/dV-plots for a) cell type A at 1C charge, b) cell type A at 3C charge, c) cell type B at 1C charge, d) cell type B at 3C charge, e) cell type C at 1C charge and f) cell type C at 3C charge and calendar aged (red curves) On the other hand, for the 3C cycled type A cells depicted in Figure 4b, both main peak features have more or less disappeared at EOL. This could again indicate several different ageing phenomena or a combination thereof. Capacity attributed to the solid solution behaviour of NMC111 at cell voltages above 3.7 V appears to be retained . This capacity is observed as a constant, non-zero dQ/dV value at high voltage. This implies that while much of the positive electrode material remains intact, it may be incompletely lithiated on discharge due to limitations of the negative electrode, lithium inventory, or other factors. The marked decrease in the lowvoltage peak further suggests that loss of graphite active material at the negative electrode may be responsible, alongside loss of cyclable lithium. The behaviour of cell type B is depicted in Figure 4c and d. In this case, it is seen that both 1C and 3C cycled cells experience loss of area under both main peaks at EOL. Many of the peak features are better preserved under 3C cycling, indicating better rate capability compared to cell type A. Again, there appeared only slight losses at high voltage, indicating that loss of NMC performance may not be caused by loss of active material. Peak shifts indicating loss of cyclable lithium are observed, alongside a broadening of several peaks, particularly for the negative electrode. The sharp, well-defined peaks at BOL represent the electrochemical reactions in relatively homogenous electrodes. The broadening and smudging of these peaks at EOL indicate the occurrence of these reactions over wider windows of cell voltage. In this way, peak broadening indicates increasing heterogeneity and local gradients of both SOC and overpotential within the cell. Broadening without loss of integrated peak area does not entail loss of capacity. Lastly, cell type C is depicted in Figure 4e and f. As with cell type B, features are not eliminated under 3C cycling, indicating acceptable rate capability. Broadening indicative of electrode heterogeneity is seen, but it is difficult to assess the relative impacts of loss of active material at each electrode. Both 1C and 3C cycled cases continue to show shift of the peaks indicating loss of cyclable lithium. ## Discussion and summary The energy density of prismatic lithium-ion battery cells of the PHEV2 VDA-format has increased significantly from the market introduction until the present. Our results on cell aging for NMC111/graphite PHEV2-size cells show that the evolution towards higher energy density is accompanied by an increased slow charging (1C) cyclability, while the fast (3C) charging cyclability has a much lower increase throughout cell generations (in a 20 % to 80 % SOC-window, Fig. 5). Regarding DCIR, the linear relationship between capacity retention and DCIR rise seen for the early version of this cell size has evolved to a more complex non-monotonic relationship. This may be a consequence of additives used to enhance shelf stability or cell performance. In any case, the increased complexity of DCIR data reveals an increasingly complex field of electrochemical phenomena within the cell and increases the difficulty of meaningful cell monitoring. Both early and more recent versions of the PHEV2-format cells show tendencies of swelling towards EOL, especially at higher (3C) charging currents. Such swelling can be due to expansion of the solid-phase electrode materials as well as gas generation from breakdown of the liquid electrolyte. Cell type A shows more severe aging of the NMC111 material compared to graphite when cycled at slower charging currents. However, the graphite seems to be more affected by cycling at higher charging currents. When compared to cell types B and C, which in turn show no severe loss of active material, it appears that optimizations have been implemented to improve the performance and longevity of the NMC and graphite electrodes. All three cell types also show loss of cyclable lithium upon cycling, both at slow and fast charging currents. One possible explanation of the different aging behaviours between cell types could be that different rates of loss of cyclable lithium affect the final outcome of electrode active material loss at EOL . In some cases, loss of cyclable lithium by formation of a solid-electrolyte interphase could help to passivate certain mechanisms and hold particles together. However, after a certain amount lost, local overpotentials could increase to the point that lithium plating or other catastrophic mechanisms are induced. One such mechanism could be electrode dry out due to long-term electrolyte degradation . In this way, the internal electrochemical environment is dynamic and changing with age of the cell. These changing mechanisms of passivation and aging could contribute to fitted changes in coulombic efficiency during the life of a cell. For each cell type, the initial coulombic efficiency appears to be similar for both slow charging and fast charging. When cycling with high charging currents, a constant coulombic efficiency is seen until EOL; in some cases there is a shift to sudden fade (decreased coulombic efficiency) close to EOL. When cycling with low charging currents, the initial low coulombic efficiency improves after a few hundred equivalent full cycles, sometimes in several steps. This decelerated ageing behaviour is well known for lithium-ion batteries but still hard to predict in models, especially if it is followed by a sudden fade. Overall, our results indicate that this type of cell could be suitable for applications such as PHEV distribution trucks, where there are demands for zero tail pipe emissions and silent driving during night delivery. However, to obtain a reasonable lifetime from a corresponding battery pack, the charging rate should be limited to around 1C. For example, a PHEV distribution truck that cycles the battery between 20 % and 80 % SOC two times per day for 250 days per year should, in the best case, be able to achieve a battery service life of around 10 years (until 80 % capacity retention), as estimated from the data obtained for cell type C. With a 3C charging regime this figure would instead be less than 4 years, which could be challenging for the customer. For heavyduty BEV applications where all traction and auxiliary energy needs to come from the battery, a traction battery with very high energy density is needed. In applications where charging can be done during longer periods (< 1C), a pure energy-optimized cell type should be a suitable choice. Today, more energy-optimized cells in the PHEV2-size are also available. For example, it has from around 2019 been possible to obtain PHEV2-size cells with around 50 Ah from selected cell suppliers . These cells seem to utilize a nickel-rich NMC positive active electrode material, sometimes also in combination with a silicon-containing graphite negative active electrode to reach high capacity. Looking into the future, from a simple model based on current cell designs, we project that a high-nickel-content cathode combined with a solid-state electrolyte could push future PHEV2 battery cell capacities towards 100 Ah, corresponding to a very high energy density as well as specific energy (>350 Wh/kg, >1000 Wh/L). Since the trend in automotive electrification increasingly points towards pure BEVs, many battery suppliers are focusing on developing more energy-dense battery cells. This development over the course of cell types A (2012), B (2014), C (2016), and beyond is shown in Figure 5. In addition to the improvement in energy density throughout cell generations, there is also an improvement in specific energy, though this is less pronounced. This means that for each generation of cells in this format, more capacity can be fit into the same cell housing, but at a greater weight. This effective densification of the cells has interesting implications for the automotive industry. A battery pack built with a certain size specification in 2018 may deliver about 90 % more energy than a visually-identical pack built in 2012, but will also be around 30 % heavier, assuming a gravimetric cell-to-pack ratio of 60 %. Hence, the same weight of batteries would in 2018 have given a 38 % increase in capacity compared to 2012. The impact of this tradeoff between pack energy and pack weight can be far-reaching for mobile applications. Trucks designed with heavier battery packs would in some cases have to sacrifice payload in favour of range. However, alongside these increases in energy density and specific energy, it can also be noted from Figure 5 that the cyclability using the 1C-charge strategy is vastly improved throughout the generations while cyclability using 3C-charge has had a slower rate of improvement. ## Conclusions We have compared three different lithium-ion battery cell generations of the prismatic VDAstandard PHEV2 regarding lifetime, with focus on the usage in electrified heavy-duty vehicles. The energy density has increased by almost 50 % over a four-year period and three cell generations, while the specific energy has increased by a more moderate 18 %. The equivalent full cycle throughput, under 1C/1C charge/discharge in a 20 % to 80 % SOC-window and at +35 °C, is also much improved through cell generations while a more moderate increase in cyclability at 3C/1C is seen. The DCIR behaviour changes throughout the cell generations from a linear relationship with capacity retention to a non-monotonic one. Present versions of the VDA PHEV2 cell format offer almost 50 Ah capacity, which means an impressively doubled capacity in 8 years. Altogether, the results from this study point out that the VDA PHEV2 cell format is still a viable choice for electrified heavy-duty vehicles such as inner-city distribution PHEV trucks or even BEVs.

chemsum

{"title": "Evaluating performance and cycle life improvements in the latest generations of prismatic lithium-ion batteries", "journal": "ChemRxiv"}

the_potential_of_jak/stat_pathway_inhibition_as_a_new_treatment_strategy_to_control_cytokine_release

## Abstract: COVID-19, a pandemic affecting virus, which is caused by the current SARS-CoV2 coronavirus. The present research is performed on anti virus and immune-modulating therapies. Cytokine storms are the toxic drivers and mortality caused by various human viral infections. In addition, the intensity was linked to an elevated risk of acute respiratory failure, myocardial injury, and mortality in SARS-CoV-2-infected patients. The Janus kinase (JAK) therapeutic inhibitor class showed significant clinical benefits in anti-inflammatory and anti-viral effects. Among them, filgotinib has been approved as an active JAK inhibitor by decreasing biomarkers with main immune reaction functions and markers supporting matrix-degradation, angiogenesis, leukocyte adhesion, and recruitment in both research trials. In this study, we tried to get an insight into the choice of this drug in controlling the jack, to treat Covid 19 using drug design methods will be discussed. ## Introduction: A new coronavirus disease with high mortality, emerging as pandemic disease, is the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Despite many public health initiatives, pharmacological therapies remain desperately required to treat patients affected, to decrease mortality, and to limit virus shedding and eventual transmission optimally. The infection with SARS-CoV-2 pushes the host to a deep cytokine response, which involves a sequence of mediators aimed at Immune-mediated inflammatory diseases (IMIDs). No specific therapy for COVID-19 is available up to now, cause many of the current treatments are symptomatic. (1) , (2) Efficient prevention and care products are an immediate imperative, in particular in difficult, life-threatening situations. Pathogen infection with coronavirus (e.g., SARS, SARS-CoV-2) also contributes to the development of acute respiratory distress syndrome (ARDS) through severe cytokine and chemokine activity. Anti-viral cytokine signaling develops like a secondary haemophagocytic lymphohistiocytosis in some patients with moderate to severe COVID-19, hyper inflammatory status triggered by viral infections. Maintains the unacceptable levels of acute lung injury, chronic interferon activation, and deteriorate resistance to T cells and antibodies, perhaps for inadequate viral clearance. While the inflammation must mount most cytokines induced by SARS-CoV-2 infection and those attacked in some \IMIDs, they do not regulate the virus clearance. The role of immunosuppressive drugs widely used in immunemediated diseases in the susceptibility and natural history of COVID-19 can be appropriately taken and concern expressed. (3)(4)(5) The cytokine excess associated with the SARS-CoV-2 reaction may also affect both viral clearance and defensive immune responses. Patients with autoimmune disorders have a high risk of infection as a result of endogenous and external factors such as immunosuppressants (dysfunctional immune system). One of the primary deficiencies of COVID-19 infection is the control of the cytokine storm. ( 6) , (7) As described above, current COVID-19 management is mostly positive, and medically validated therapies are not available. ARDS and cytokine storms are the leading causes of death. In addition, 50% of cytokine storm syndrome patients suffer from ARDS. Considering the exceptionally rapid progression of systemic and pulmonary inflammation in a subset of COVID-19 patients, it is highly necessary to recognize and control the immune reactions that are disrupted at an early stage. However, this must be checked, and other biomarkers that are more sensitive and more precise can be identified. ( 8) , ( 9) Several explosive cytokines that include automatic diseases associated with their receptors have activated a JAK based phosphorylation cascade, which constitutes signals of gene transcription. Thus, medications block signals of cytokine that impede the action of JAK. These antagonists target medical treatment to HIV, RA, Psoriasis, Psoriatics, and inflammatory bowel diseases. (10) Signal transduction plays a significant role in having and blocking the cytokine releases of the JAK family of enzymes and JAK inhibitors. Inhibitors of JAK can handle a cytokine storm by the induction of many inflammatory cytokines. Most inhibitors for JAK 1, JAK 2, less JAK 3, and the Tyrosine kinase 2 (TYK 2) are immensely successful for inhibition of Interleukin 6 (IL-6) and interferon, but also inhibit the signal cascade of both Interleukin 2 (IL-2) and Interferon alfa or beta (IFN-α / β). JAK inhibitors have been efficient in inhibition. (11) Inhibitions of small molecules in JAK are quite a recent concept for systemic autoimmune/inflammatory conditions. JAK Inhibitors are biological inhibitors that interact with the Adenosine triphosphate (ATP) binding domain by inhibition of type I / II cytokine receptors. Jak inhibitors have provided targeted synthetic immunosuppressant products that interfere with JAK Signal transducer and activator of transcription (JAK-STAT) signaling by inhibiting one or more members of the JAK family (JAK1, JAK2, JAK3, TYK2). These molecules mediate the transcription factors of the STAT family, which contribute to pro-inflammatory cytokine release. Thus, cytokine expression can be decreased by them and help regulate cytokine storms. Additional inhibition of JAK by small modules can also be detrimental as they further restrict the isolation and clearance of the pathogen and can cause unexpected complications. ( 12) On the other hand, these compounds are provided as medications orally, with highly trained pharmacodynamics and pharmacokinetics. They can provide a more practical approach to calm down the cytokine storm transiently to avoid ARDS and fulminating myocarditis. In addition to blocking IL-6, JAK1 inhibitors not only block inflammatory pathways in a cytokine storm. ( 13) The aims of JAK1 and JAK3 affect some cytokines involved in anti-viral reactions such as interferons, IL-2, IL-15, IL-21, and IFNβ. Thus, potentially, JAK1 inhibitors can inhibit SARS-CoV-2 clearance. Inhibition of the SARS-CoV-2 or the IL-17-induced cytokine inhibits viral induction. In particular, it seems to be very promising to apply interleukin 6 (IL-6) and GM-CSF blockers to manage the massive cytokine storm that is linked to the development of lung damage typically and resulting ARDS in the most attacking patterns of SAR S-CoV disease. Both of whom rely on SARS-CoV-2 signaling in part or entirely (Figure 1). ( 14) Therefore, clinical trials have been JAK1 inhibitors are also suggested as a safe treatment in hospitalized patients with COVID-19. Based on recent analyzes of the COVID-19 inflammatory markers and previous knowledge of inflammatory responses in other mortal lung infections, the potential strategy for anti-ARDS, brilliant myocarditis, organ failure, and mortality at an advanced stage of the condition has been assessed . (15) One of the significant challenges in this regard is the replacement of more effective drugs. It has been indicated to be given to patients with COVID-19 in the late inflammatory process by baricitinib or other JAK inhibitors. Also, using different anti-viral medications, as a result of a growing understanding of infection pathophysiology, other drugs widely used in RA diagnosis have been proposed as alternative therapies for COVID-19. Baricitinib were tested for their anticytokine and anti-inflammatory function. Baricitinib induces cytokines with a lower IC50 value, which indicates that cytokine-induced JAK1 / STAT signalling becomes more impaired in the dosing period. It implies a more potent overall inhibition of cytokine-induced JAK1 / STAT signalling during dosing. But finding a drug with better therapeutic properties can help in the treatment of this disease. ( 16) Finding new drugs is a challenging, expensive, and time-consuming task because there is no structured way to immediately discover a drug even though the drug activity's disorder, targets, and molecular mechanisms are well understood. There are millions of candidate molecules, and because of prohibitive costs, both in terms of time and energy, individual tests cannot be performed on any candidate. Reasonable drug design strategies have been introduced in recent months, particularly for In silico-based solutions, and this strategy has been backed by a recent study as a promising substitute or complementary method for efficient screening of potential drugs. Here, we used a bioinformatics approach to repurpose medication to classify the active antagonists of SARS-CoV2 Key Jak1-inhibitors. ( 17) , ( 18) EXPERIMENTAL : In this study, the three-dimensional structures of ligand and proteins were obtained from PubChem and PDB database, respectively. Density functional theory at B3LYP/631+G (d, p) level implemented was used for 3D and geometry optimizations with energy minimization of each molecule. The protein-ligand interaction calculations were done by Autodock 4.2 and 2D ligand-protein interaction was calculated by Ligplot software. DRAGON software was used for molecular descriptors calculation. Genetic Algorithm and Partial least squares regression were used for feature selection. The evaluation of the active site, surface, and volume of protein was done by Computed Atlas for Surface Topography of Proteins (CASTp). Swiss ADME and target prediction were used to determine the pharmacokinetics properties and target analysis of molecules. GROMACS-2019 version using OPLS force field during was used for Molecular dynamic simulations during 20 ns by selecting periodic boundary conditions and the TIP3P water model for solvating complexes, followed by addition of ions to neutralize. Energy minimization was Tolerance for energy minimization was 1000 kJ/mol/nm. (19)(20)(21) ## Results and discussion The interaction of the JAK1 inhibitor drugs (clinical and pre-clinical) using Autodock software was studied, and the results are given in Table 1. The binding force of molecular docking demonstrates the affinity of a specific ligand and energy, by which a compound interacts and binds to the pocket of a target protein. As a potential drug choice, a compound with fewer binding energy is favoured. The results showed that the drug Filgotinib is more stable with Jack1. ## Name Filgotinib (GLPG0634 / GS-6034) is an active and selective inhibitor of JAK1 that under investigation in the treatment of RAs and inflammatory bowel disease. Filgotinib has demonstrated promising efficacy and is well tolerated for the treatment of rheumatoid arthritis. It is an orally delivered, potent, and selective seed inhibitor of JAK1. The pharmacokinetics and active metabolite of filgotinib in safe volunteers and the usage and analysis of pharmacokineticpharmaceutical models to help the design of the dosage for Phase IIB for patients with rheumatoid arthritis were addressed here. Two-phase II tests of another treatment for JAK1filgotinib in 2018 showed effectiveness in both patients with psoriatic arthritis1 and ankylosing spondylitis in patients2. Compared to Upadacitinib 3,4, Filgotinib therapy provided a mean positive improvement in hemoglobin and platelet counts. A study of the interaction of 33 derivatives of this drug using Autodock software showed that compared to the drugs studied in the previous section, ΔG shifted to more stable values (Table 2) (Figure -2). QSAR calculations performed using algorithm-PLS genetics showed that the number of benzene ring and polarity is an essential factor in molecule-JAK1 interaction (Figure -2). Also, compared to filgotinib, only entry five has created a more stable complex. Pharmacophore analysis showed that the behaviour of this substance is similar to that of filgotinib, and changes in volume and area are similar (Table -3) (Figure 4) . But the results of 2d interaction result showed more hydrophobic interaction with amino acids (Table-4) (Figure -6). Target prediction results show that the A 93% composition targets the LAK, while the B 73.3% composition interacts (Figure -7). ADME studies also showed similar behavioural similarities to filgotinib (Table -5). As a result of these two compounds can be the alternative of Barticinib drug. Still, need supports a more in-depth study on JAK-1 inhibits as the mechanism for therapeutic prevention of a cytokine storm and the downstream organ failure under this situation. ## Compliance with ethical guidelines Conflict of interest: The Autors have no financial or non-financial conflict of interest to declare. For this article, no studies with human participants or animals were performed by any of the authors. All studies conducted were in accordance with the ethical standards indicated in each case.

chemsum

{"title": "The potential of JAK/STAT pathway inhibition as a New Treatment Strategy to Control Cytokine Release Syndrome in COVID-19", "journal": "ChemRxiv"}

enantioselective_phase-transfer_catalyzed_alkylation_of_1-methyl-7-methoxy-2-tetralone:_an_effective

## Abstract: In order to prepare asymmetrically (R)-(+)-1-(5-bromopentyl)-1-methyl-7-methoxy-2-tetralone (3a), a key intermediate of dezocine, 17 cinchona alkaloid-derived catalysts were prepared and screened for the enantioselective alkylation of 1-methyl-7methoxy-2-tetralone with 1,5-dibromopentane, and the best catalyst (C7) was identified. In addition, optimizations of the alkylation were carried out so that the process became practical and effective. ## Introduction The preparation of enantiomerically pure compounds has become a stringent requirement for pharmaceutical synthesis . In this context, asymmetric catalysis is probably one of the most attractive procedures for the synthesis of active pharmaceutical ingredients (APIs) due to environmental, operational, and economic benefits. Dezocine, (5R,11S,13S)-13-amino-5-methyl-5,6,7,8,9,10,11,12octahydro-5-methyl-5,11-methanobenzocyclodecen-3-ol (1, Scheme 1), a typical opioid analgesic developed by AstraZeneca, was extensively used in China recently. Because of its effectiveness and safety , it would have a very good marketing prospect. However, the cost of dezocine was very high since the commercial synthesis process involved the tradi-tional resolution : alkylation of 1-methyl-7-methoxy-2tetralone (2) with 1,5-dibromopentane gave the designed (R)-(+)-1-(5-bromopentyl)-1-methyl-7-methoxy-2-tetralone (3a) and an equal amount of the S-isomer 3b, both 3a and 3b underwent the following cyclization, oximation and reduction, and then, (5R,11S,13S)-3-methoxy-5-methyl-5,6,7,8,9,10,11,12octahydro-5,11-methanobenzocyclodecen-13-amine (6a) and (5S,11R,13R)-3-methoxy-5-methyl-5,6,7,8,9,10,11,12octahydro-5,11-methanobenzocyclodecen-13-amine (6b) were separated by two times of resolution with L-tartaric acid and D-tartaric acid (Scheme 1). the topics in stereoselective synthesis in both industry and academia . It was reported that the alkylation of 2 in the catalysis of N-(p-trifluoromethylbenzyl)cinchonidinium bromide in a two-phase system gave the enantioselective product 3a, although the ee value of the product was 60%, determined by 1 H NMR. And so far, no further report on the stereoselective alkylation of 2a was found. (Some reports on the nonstereoselective alkylation of 2 were given in references ). In this paper, several cinchona-derived phase-transfer catalysts were screened for this reaction, and the structure-activity relationship for the catalysis was studied. In addition, optimizations had been made to make the process efficient. ## Results and Discussion A series of the quaternary ammonium bromides from cinchonidine or quinine as phase-transfer catalysts was prepared (Scheme 2). Cinchonidine was reacted with the benzyl bromides (R 1 Br) in THF to obtain catalysts C1-C11 . And then C7 reacted with allyl or propargyl bromide to obtain C12 and C13. In another way, cinchonidine was reduced by H 2 /Pd/C to yield dihydrocinchonidine, and then reacted with 4-trifluoromethylbenzyl bromide to obtained C14. C15 was prepared from cinchonidine via bromination, debromination and condensation with 4-trifluoromethylbenzyl bromide . Quinine was reacted with 4-trifluoromethylbenzyl bromide or 3,5-bis(trifluoromethyl)benzyl bromide to obtain C16 or C17. In the beginning, the alkylation of 2 in the catalysis of C1 in the two-phase system (toluene and 50% NaOH aqueous solution) was tested, although the yield was moderate (60.1%, entry 1 in Table 1), the enantiomeric ratio (3a:3b) was only 55:45. When the benzyl in C1 was replaced by the bulky groups, such as methylnaphthalene or methylanthracene, neither the enantiomeric ratio was improved (Table 1, entry 2) nor the reaction took place (Table 1, entry 3). Subsequently, when the groups substituted at the para-position on the benzyl group were investigated, the structure-activity relationship showed that catalyst C4 (with methyl substituent) did not work for the reaction (Table 1, entry 4) and those with Cl or F (C5 and C6) worked well with an improvement in enantiocontrol (Table 1, entries 5 and 6). Fortunately, the p-CF 3 derivative (C7) promoted the reaction to give a enantiomeric ratio of 83:17 (Table 1, entry 7). These findings suggested that the presence of electron-withdrawing groups on the benzyl group was favourable for the enantioselective reaction except the case of a nitro group (Table 1, entry 8). And then, the catalysts with a di-substituted benzyl group were examined. C9 with 3.4-dichlorobenzyl resulted in a slightly higher enantiomeric ratio (68:32) than C5 (Table 1, entry 9). But, neither C10 nor C11 (Table 1, entries 10 and 11) were as good as the mono-substituted counterparts (C6 and C7). The derivatives (C12-C15) of C7, the best one so far, were further studied. When the hydroxy group in C7 was protected by an allyl or a propargyl group, racemic product was obtained a The reaction was performed with 0.045 mol/L of 2 in toluene (24 mL), 3.0 equiv of 1.5-dibromopentane and 50% aq NaOH (2.4 mL) in the presence of 10 mol % of catalyst at 15-25 °C for 48 h under N 2 . b Isolated yield including 3a and 3b. c The enantiomeric ratio was determined by HPLC using a chiral column (Daicel chiral AY-H) with hexane/isopropyl alcohol 90:10 as the eluent, detected at 280 nm. (Table 1, entries 12 and 13). This suggested that the free hydroxy group in C7 was crucial to guarantee the stereoselectivity. Meanwhile, the good catalysis was maintained with both dihydrocinchonidine-derived C14 and dehydro compound C15. Finally, the quaternary ammonium group from quinine was examined (Table 1, entries 16 and 17), and C16 and C17 gave the result inferior to the cinchonidine derivatives (C7 and C11). After a suitable catalyst (C7) was identified, further reaction optimization was performed (Table 2). In general, dichloromethane (DCM) was the common solvent for the two-phase reaction, but to our surprise, when the reaction was run in DCM (entry 2 in Table 2), it resulted in the racemic product. When other solvents, such as benzene, bromobenzene and fluorobenwere used, neither the enantiomeric ratio nor the yield was compared with toluene as the solvent (Table 2, entries 1, 3-5). But, the reaction in chlorobenzene gave a slightly improved yield at a substrate concentration of 0.045 mol/L (Table 2, entry 1 and 6). Surprisingly, when the concentration increased to 0.07 mol/L, the improvement became more significant (Table 2, entries 7 and 8). However, further increasing the substrate concentration (Table 2, entry 9) decreased the stereoselectivity. For the screening of the base, the reduction of volume or concentration of 50% aq NaOH resulted in a decreased yield (Table 2, entries 11 and 12). If NaOH was replaced by K 2 CO 3 , no reaction took place (Table 2, entry 13). As far as the reaction temperature was concerned (Table 2, entry 7, 14 and 15), it was found that the reaction at 15-25 °C gave the best result. Finally, the reaction was scaled up (90 g of 2) according to the conditions in entry 7, a similar outcome was obtained (Table 2, entry 16). On the base of the above experimental results, a catalytic mechanism was proposed (Scheme 3). Compound 2 is deprotonated by sodium hydroxide into an anion in the organic layer. The anion goes to the interface between chlorobenzene and water, where it interacts with the quaternary ammonium group of catalyst C7. The distance between two molecules is getting close by the attraction between charges, then two additional interaction forces in the complex are produced on the same plane, including: 1) the carbonyl of 2 makes a hydrogen bond with the hydroxy group of C7; 2) the phenyl group of 2 forms a face-to- The volume ratio of aqueous solution and organic solvent was 1:10. c Isolated yield including 3a and 3b. d The enantiomeric ratio was determined by HPLC using a chiral column (Daicel chiral AY-H) with hexane/isopropyl alcohol 90:10 as the eluent, detected at 280 nm. e The volume of 50% aq NaOH decreased to 5% of volume of PhCl. f 90 g of 2 was added. ## Scheme 3: The proposed catalytic mechanism of stereoselective alkylation. face π-stacking interaction with the benzyl moiety of C7. The complex of 2 with C7 goes to the organic phase. Due to the sterical hindrance from the benzyl group, the alkylation by 1,5dibromopentane takes place at the opposite side of the benzyl group of C7 to afford 3a. ## Conclusion In summary, an enantioselective synthesis of (R)-(+)-1-(5bromopentyl)-1-methyl-7-methoxy-2-tetralone (3a), a key intermediate of dezocine, in the catalysis of the quaternary ammonium benzyl bromides from cinchonidine was investigated and the best catalyst (C7) was identified. In addition, the preparation of 3a with the optimized conditions was performed and the product was isolated in 77.8% yield with an enantiomeric ratio of 79:21. This method can be easily performed in large scale. In addition, the structure-activity relationships for the cinchona alkaloids catalysts were elucidated. ## Experimental All solvents and reagents were of commercial sources and used without further purification. Melting points were determined on a Büchi Melting Point M-565 apparatus and are uncorrected. 1 H and 13 C NMR spectra were recorded using a Bruker 400 MHz spectrometer with TMS as an internal standard. Mass spectra were recorded with a Q-TOF mass spectrometer using electrospray positive ionization (ESI + ). The enantiomeric ratio was determined by HPLC using a chiral column (Daicel chiral AY-H) with (hexane/isopropyl alcohol 90:10) as eluents, detected at 280 nm. Specific rotations were determined on a Rudolph Research Analytical automatic polarimeter IV. All reactions were monitored by TLC, which were carried out on silica gel GF254. Column chromatography was carried out on silica gel (200-300 mesh) purchased from Qindao Ocean Chemical Company of China. General procedure for the preparation of (R)-(+)-1-(5-bromopentyl)-1-methyl-7-methoxy-2tetralone (3a) To a stirred mixture of 2 (90.0 g, 0.47 mol), C7 (25.2 g, 0.047 mol) and 1,5-dibromopentane (326.3 g, 1.4 mol) in chlorobenzene (6750 mL) was added 50% aq NaOH solution (675 mL) at 0 °C. The mixture was allowed to warm up slowly to 15-25 °C and stirred for 48 h under N 2 , and then aqueous layer was separated and extracted with chlorobenzene (700 mL). The combined organic layers were washed with 1 M HCl aqueous solution (2 L) and water (2 L), then the solvent and excess of 1,5-dibromopentane were recovered, respectively, under reduced pressure and then in vacuo. The above-obtained product underwent subsequent cyclization, oximation and reduction according to the literature (without resolution) to get compound 6a, and then 6a was trans-formed to dezocine with 23.0% overall yield and 100% purity. The mp, optical rotation value, MS and 1 H NMR of the product were consistent with those in the literature . ## Supporting Information Supporting Information File 1 Synthesis of catalysts C1-C17, synthesis of dezocine, 1 H NMR and MS spectra of catalysts C1-C17 and chiral HPLC diagrams of 3. 1 H NMR, 13 C NMR, MS spectra of 3. 1 H NMR, MS spectra HPLC diagrams of dezocine. [https://www.beilstein-journals.org/bjoc/content/ supplementary/1860-5397-14-119-S1.pdf] ## ORCID ® iDs Ruipeng Li -https://orcid.org/0000-0001-9520-0635

chemsum

{"title": "Enantioselective phase-transfer catalyzed alkylation of 1-methyl-7-methoxy-2-tetralone: an effective route to dezocine", "journal": "Beilstein"}

efficient_luminescence_control_in_a_dithienylethene_functionalized_cyclen_macrocyclic_complex

## Abstract: We report the synthesis of an original ligand scaffold based on a dimethyl-cyclen platform Medo2pa with two dithienylethene units attached to each picolinate arms and the corresponding yttrium(III), europium(III) and ytterbium(III) complexes. All three complexes show reversible photochromism with high photo-conversions. Photoluminescence experiments demonstrate that this design is versatile and adapted for both europium and ytterbium emission switching when measured in frozen organic glasses at 77 K. The OFF/ON luminescence ratio are excellent in the case of europium (4 to 8 %) and still quite good in the case of ytterbium (around 13 %). ## Introduction Responsive materials in which a key property can be modulated by an external stimulus in a controlled way are a great achievement in the field of molecular materials. Among them "alloptical" systems, that are triggered by light to change their optical (absorption, emission) properties, combine fast response, remote control and a low level of technical requirements for their implementation in real life applications. Applications could be as diverse as labels for cell imaging, 3 super resolution imaging, 4 anti-counterfeiting dyes, optical data-storage 7 and many others. In this context, several research groups have explored the photo-modulation of lanthanide-based luminescent systems, 5, mainly focusing on the association of photochromic compounds with the red-emitting europium(III) ion. 5, The ubiquitous diarylethene (DAE) photochromic units, 17 on top of their excellent photo-physical properties, fatigue resistance and thermal stability of both open and closed isomers, is perfectly suited. Indeed, DAE scaffolds can be easily designed so that the closed isomers show strong absorptions around 610 nm, matching the narrow emission lines of europium(III) and then favoring emission quenching typically via an energy transfer. However, according to this strategy, a complete quenching of europium luminescence in the closed form has not been realized yet. The only total quenching of europium luminescence by a photochromic unit reported to date consist of a tris(dipicolinate)europium core decorated with three N^C chelate four coordinate organoboron T type (reversible upon heating) photoswitches. 15 Therefore, it is highly desirable to achieve a complete optical control of ON/OFF switching of europium luminescence with the P (thermally stable) photochromic DAE. Recently, some of us reported an example of partial photo-modulation in a dithienylethene (DTE) appended dipicolinic amide europium complex (Chart 1), 18 and we hypothesize that a partial lability of the complex could be a factor contributing to the moderate efficiency of the quenching in the closed form. At the same time, surprisingly, this previous paper showed that DTE photochromic units could actually be more versatile modulators of lanthanide luminescence than initially thought since ytterbium(III) NIR emission could be sensitized by the 580 nm absorption of the closed isomer. Based on this, two important goals remain to be achieved in this field: i) the improvement of the efficiency of europium(III) emission quenching by closed DTE system in order to reach real ON/OFF switching, and ii) the generalization and optimization of photo-modulation of ytterbium(III) ion by DTE units. These two goals thus require a better understanding of the underlying photo-physical mechanisms and the exploration of new systems combining DTE and lanthanide ions. between the open and closed state (TTA is 2-thenoyltrifluoroacetonate). 18 (middle) Medo2pa provides water soluble and stable lanthanide complexes and M-Medo2pa-2P chlorine salts enable cell imaging in the NIR range in the case of the ytterbium(III) complex. (bottom) Target complexes. In parallel, macrocyclic lanthanide complexes have been widely studied as imaging bioprobes in general, 3, and as luminescent systems in particular. Among them, the cyclen based Medo2pa platform (Chart 1) has provided complexes of various lanthanide ions displaying high stability constants, 25 that are typically stable in water solutions. 26 This cyclen platform Nfunctionalized by two picolinate pendants and two methyl groups, when modified with two photon active conjugated antennas, provides bright luminescent complexes of europium(III) and ytterbium(III) that are spontaneously internalized into live cells, the latter remaining highly luminescent in biological media (Chart 1). 20 Based on these convincing results, and complementary to another strategy on based DTE modified acetyl acetonate ligands that we are developing in parallel, 27 we thought that the association of the Medo2pa platform with appropriate DTE units could lead to "all optical" switches with improved stability and, therefore, better switching ratio between the open and closed state, as well as to provide a new efficient ytterbium based switch in the NIR range through the closed DTE unit sensitization. We therefore targeted the synthesis of a new Medo2pa platform bearing two DTE units (on each picolinate arms) as shown in Chart 1. First motivated by the ease of synthesis, the presence of two photochromic units within the same scaffold could also be anticipated as an advantage to improve i) quenching efficiency in the case of the europium(III) complex, and ii) sensitization through the closed DTE unit in the case of the ytterbium(III) complex. In this paper, we report on the synthesis of this new ligand and of the corresponding europium(III), ytterbium(III) and yttrium(III) complexes. We study in detail the photo-switching of these three complexes by absorption and ( 1 H, 19 F) NMR spectroscopies to illustrate that a reversible and complete isomerization occurs, the two DTE units behaving independently. Our strategy is proved effective in improving the quenching efficiency of europium luminescence as shown by a residual intensity of 4-8 % of the initial one for the closed form as compared to the open one when measured at 77 K. We also show that the ytterbium complex luminescence can be modulated at 77 K although it does not exhibit any sensitization through the closed DTE. ## Results and Discussion Complex synthesis. Synthesis of the target complexes [MLoo]Cl (M = Y, Eu, Yb) is described in scheme 1. The DTE-photochromic-picolinate arm 1 was obtained by Sonogashira coupling from the alkyne terminated DTE and methyl 6-(hydroxymethyl)-4-iodopicolinate 28 (see SI). Mesylation of the latter was performed under usual conditions and trans-dialkylation of the dimethyl-cyclen macrocycle with two equivalents of compound 2 in the presence of K2CO3 led to the desired diester 3 with an excellent yield of 95%. Saponification of compound 3 in the presence of KOH in THF led to the ligand Loo as a potassium salt which was purified, thanks to a precipitation in an EtOAc/hexane mixture. The synthesis of the complexes was further performed in MeOH at pH around 7. Washings with water and precipitations in CH2Cl2/hexane gave the desired [MLoo]Cl complexes with yields comprised between 61% and 90%. These new compounds were fully characterized (see experimental section and SI). As characteristic features in its 1 H NMR spectrum, the diamagnetic complex [YLoo]Cl exhibit shielded pyridine protons chemical shifts, similarly to other yttrium(III) dimethyl cyclen complexes, 29 while the signals from the cyclen moiety become significantly broadened upon coordination (Figure S12). In the case of the [EuLoo]Cl complex, additional paramagnetic shifts (pseudo contact shifts) are observed. Typically, the photochromic moiety shows small paramagnetic shifts, of around -0.1/-0.2 ppm as compared with the yttrium(III) complex, while the pyridine protons are observed at  = 38.4 and 25.8 ppm and the cyclen protons give broad signals down to -16 ppm as expected (Figure S21). 19 For [YbLoo]Cl complex, in line with the greater magnetic anisotropy tensor of ytterbium(III) compared with europium(III), 30 shifts of the same sign but of greater magnitude are observed, the pyridine protons being observed at  = 83.8 and 55.5 ppm and the cyclen ones down to  = -40.5 ppm (Figure S18). The paramagnetic shifts observed for the photochromic moiety are also larger with, for instance, the thiophene protons shielded to  = 6.97 and 6.37 ppm instead of  = 7.47 and 7.28 ppm in [YLoo]Cl. ## Electronic absorption spectra and photochromism of 3oo and [MLoo]Cl complexes. The absorption spectrum of 3oo in DCM shows several intense bands in the UV range (Figure 1) that can be assigned to local -* transition of the picolyl unit (275 nm) overlapping with one of the DTE open form (315 nm). Upon irradiation at 330 nm, a decrease of absorption is observed at max =272 nm while two new bands appear at max = 382 and 607 nm (Figure 1). The initial spectrum can be recovered by 580 nm irradiation. This is in line with the usual photochromic behavior of DTE units 18 and consistent with the above mentioned assignment of the bands. Photo-cyclisation is evidenced by the characteristic lower energy band (max = 607 nm) ascribed to an intra-ligand (IL) transition centered on the closed DTE moiety. 17 In this system with two DTE units, isomerization proceeds through the intermediate 3oc compound with one closed ring. However, at intermediate photo-conversions, no shifting of the lower energy transition was observed, suggesting that the two DTE units are electronically decoupled and behave independently in that case (Figure S24). 31 The isomerization was also studied by 1 H NMR that proved that a high photoisomerization conversion (up to 94 % of 3cc and 6 % of 3oc) can be reached in the photo-stationary state (PSS) (Figure S29). Typically, the thienyl protons chemical shifts change from  = 7.25 and 7.31 ppm in 3oo to 6.72 and 6.43 ppm in 3cc. In the NMR conditions ([c] = 1.2×10 -3 M), the cycloreversion process is almost quantitative with the recovery of 3oo in 94 % yield accompanied by unknown species, probably coming from partial degradation upon prolonged exposure to light. This behavior is in contrast to the more diluted UV-vis experiment that displays quantitative recovering. Scheme 2. Synthetic pathway yielding the target complexes. 300 400 500 600 700 800 0,0 2,0x10 4 4,0x10 4 6,0x10 4 e (M -1 cm -1 ) 1 and Figures S27 and S28). The initial spectra were recovered after bleaching at 580 nm. Once the cyclen group is coordinated, clean photochromic behavior, exempt of photo-degradation was observed as evidenced by the presence of isobestic points. The absorption spectra of the three complexes are very similar with two main transitions at max = 269 nm and 350-360 nm and Figure 1 shows the representative behavior of the europium complex (the cases of Y and Yb complexes are depicted in figures S27 and S28 respectively). Both bands are strongly modified upon UV irradiations and subsequent ring closure, and new transitions appear with max values of 330 and 627 nm. The lower energy transition is slightly red shifted upon coordination as compared with 3cc. Under visible light irradiation (max = 580 nm), the cycloreversion process is triggered as attested by the quantitative recovery of the initial spectra. Further 1 H NMR monitoring of the process unambiguously shows that the photochromic process upon UV irradiation is almost complete with the reaching of a photo-stationary state composed of ca. 95 % of closed DTE units and a recovery of the initial spectra upon 580 nm irradiation, in contrast to the organic precursor. Details of the changes in the NMR spectra are highlighted in Figures 2 and S31 Photoluminescence of [MLoo]Cl complexes. We further studied the photoluminescence of all three complexes (M = Y, Eu, Yb). The yttrium complex serves as a reference to understand the photo-physics of the ligand since no metal-based emission is expected for this compound. Thus, upon excitation at ex = 350 nm of [YLoo]Cl in an ethanol:methanol glass (77 K), a ligand-based fluorescence centered at em = 395 nm was observed (Figure S32) with the presence of additional peaks in its tail. A time-gated measurement performed with a 1 ms delay allows us to assign unambiguously these features to a simultaneous structured phosphorescence with maximum at em = 517 nm and a corresponding lifetime of 14 ms at 77K (Figure S33). This phosphorescence process corresponds to a ligand-centered triplet at around 19 000 cm -1 . Upon continuous irradiation at 350 nm and closing of the DTE units, both fluorescence and phosphorescence progressively disappeared, and at the PSS the closed yttrium(III) complex was almost non-emissive (Figure S32). Concerning the spectroscopy of the europium complex, [EuLoo]Cl was studied at room temperature and at 77 K. At room temperature, excitation at 350 nm induces both emission and competitive closing of the DTE units. The spectrum is actually dominated by an intense ligandcentered emission at em = 395 nm accompanied by a weak europium emission at 616 nm (Figure S34). In contrast, at 77 K in a methanol/ethanol organic glass, ligand centered emission is drastically decreased as compared with the sharp f-f transitions. The difference in the response of the system with temperature could be ascribed to the occurrence of thermally activated back energy transfer that is hampered at 77 K. We also observed a drastic slowing down of the closing reaction by this lowering of temperature and immobilization in an organic glass that allows to measure the emission spectrum of pure [EuLoo]Cl with an excellent resolution. Therefore, the characteristic europium(III) emission profile assigned to the 5 D0  7 FJ (J = 0-4) transitions were detected at em = 580 (J = 0), 588, 593, 595 (J = 1), 610, 613, 622, 627 (J = 2), 646, 650 , 652, 658, 673 (J = 3), and 694, 704, 711 nm (J = 4) (Figure 4) and overall, the spectrum and particularly the crystal field splitting, is very similar to the one of a previously published europium complexes with a similar Medo2pa ligand for which a C2 symmetry was calculated by DFT. 19 The same measurement at 77 K was performed on [EuLcc]Cl (PSS state) and showed that an impressive quenching of europium luminescence occurs after closing of the DTE since only very weak emission (about 8 % of the original intensity determined by integration of the open state more intense band (J = 2), see Figure 3) was detected. It is also possible to follow the emission quenching in the glass at 77K upon successive scans, highlighting the progressive closing of the DTE during each luminescence measurement (Figure S35). An attempt to reach the PSS was performed upon irradiation of the glass during 1000 s. A 90% quenching was achieved after only 40 s but the complete closing was not reached at the end of the experiment where less than 4% of the initial emission was still observed. A perfect reproducibility of the behavior was observed after several re-opening performed with white light irradiation (Figure 4). 4,0x10 5 6,0x10 5 8,0x10 5 1,0x10 6 1,2x10 6 1,4x10 6 1,6x10 4,0x10 5 6,0x10 5 8,0x10 5 1,0x10 6 1,2x10 6 1,4x10 For complex [YbLoo]Cl, no ytterbium emission was detected upon 350 nm excitation at room temperature. In contrast, in an ethanol/methanol organic glass at 77 K, the typical emission of ytterbium(III) was detected in the NIR. In order to avoid distortion of the signal due to concomitant closing, the emission was detected with a CCD camera. First, a resolved spectrum can be obtained, clearly showing the different lines expected for the 2 F5/2 → 2 F7/2 transition and again very similar to previously reported complexes with C2 symmetry, 20 with the main crystal field splitting lines at 971, 996, 1025 and 1040 nm (Figure 5). In order to follow the effect of photo-isomerization on ytterbium emission, fast-acquired successive spectra were obtained, clearly showing a 10 fold quenching of luminescence due to the closing reaction (Figure S36). Note that the quenching ratio is not rendered by figure 5 because the initial intensity actually corresponds to a system already undergoing a significant amount of closing. Rather, the ratio between the initial and final states can be obtained from integration of the fast acquired data (Figure 6), giving a 13 % ratio. Finally, we have addressed the possibility of sensitization by excitation at 600 nm, and unlike complex Yb-DTEc (Scheme 1) no ytterbium emission was detected in such case. 18 For both europium and ytterbium complexes, it is unclear whether the remaining emission after closing arises from the closed species or whether a PSS composition different from the one in DCM solutions at room temperature (95 % of closed units, no remaining oo isomer) is reached due to immobilization in a frozen organic glass. Discussion. Altogether, and in light with the objectives mentioned in the introduction, the results of the photoluminescence experiments deserve a few comments. First, temperature/medium dependence of the response is very spectacular for both systems and in both cases, no lanthanide based emission can be detected at room temperature. In the case of europium, this is probably because of thermally activated back-transfer, hence causing ligand-centered emission as suggested by the presence of the open form ligand triplet state at 19000 cm -1 . In the case of ytterbium, it is more likely that luminescence is inherently weak due to efficient non-radiative processes and therefore difficult to detect without causing the closing of the DTE. At 77 K in an organic glass, the non-radiative processes are drastically slowed down as well as the closing reaction and both factors favor the observation of ytterbium emission. Second, when measured in appropriate conditions, the contrast between the responses of the two states for our europium complex is much higher than in previous photoswitchable systems based on europium and diarylethene combinations (Table 2) and only one example relying on N^C chelate four coordinate organoboron photoswitches of T type previously showed better quenching ratio. 15 Provided that back transfer and non-radiative processes are reduced by further chemical engineering, our design with a macrocycle bearing two DTE units could lead to very efficient RT europium luminescence switches. Nonetheless, this design leads to the second example of efficient ytterbium luminescence photo-control reported so far. In that case, the mechanism for emission quenching does not rely on spectral overlap between the closed DTE and the lanthanide emission lines and we are currently investigating the possibility of a low lying triplet state quenching the emission in the closed state. We also postulate that the position of this state is not favorable to sensitization of ytterbium emission through the visible transition of the closed DTE unit unlike in Yb-DTEc. This leaves room for improvement of ligand design in order to obtain optimized positioning of this state depending on the targeted behavior ie UV sensitization with quenching by a low lying state or controllable visible light sensitization. ## Conclusion. With this work, we report the synthesis of an original ligand scaffold with two DTE units attached to a cyclen based macrocycle designed for luminescence switching and the corresponding complexes of yttrium(III), europium(III) and ytterbium(III). All three complexes show reversible photochromism with high photo-conversions. Our design proved to be versatile and adapted for both europium and ytterbium emission switching, when measured in frozen organic glasses. The OFF/ON luminescence ratio are excellent in the case of europium compared to all previously published compounds and still quite good in the case of ytterbium, that represents the second example of such behavior. More important, our study, combined with on-going in depth photophysical studies, will contribute to the understanding of important factors for the design of further improved molecular switches with custom switching, excitation and emission wavelengths. Complex [YLoo]Cl. To a solution of compound Loo (60 mg, 41 µmol) in MeOH (HPLC grad, 10 mL) was added YCl3.6H2O (37 mg, 122 µmol, 3 eq). The pH was controlled at 7 and the reaction mixture was stirred at room temperature for 3.5 days. Solvents were evaporated to dryness and water was added to the residue. Water was then filtered on cotton and the solid kept on the cotton was dissolved with CH3CN (HPLC grad). CH3CN was evaporated to dryness and the residue was dissolved in the minimum of CH2Cl2. A large amount of hexane was added and the precipitated was filtered, washed with hexane and dried under vacuum to yield [YLoo]Cl (38 mg, 25 µmol, 61%) as a pale yellow solid.

chemsum

{"title": "Efficient Luminescence Control in a Dithienylethene Functionalized Cyclen Macrocyclic Complex", "journal": "ChemRxiv"}

insights_into_the_role_of_noncovalent_interactions_in_distal_functionalization_of_the_aryl_c(sp<sup>

## Abstract: Burgeoning interest in distal functionalization of aryl C-H bonds led to the development of iridiumcatalyzed borylation reactions. The significance and inadequate mechanistic understanding of C(sp 2 )-H borylations motivated us to investigate the key catalytic steps and the origin of a directing-group-free regiocontrol in the reaction between aryl amides and B 2 pin 2 (bis(pinacolato)diboron). An Ir(III)(ubpy) tris(boryl) complex, generated from the pre-catalyst [Ir(OMe)(cod)] 2 by the action of a bipyridine-urea ligand (ubpy) and B 2 pin 2 , is considered as the most likely active catalyst. The meta C-H activation of N,N-dihexylbenzamide is energetically more favorable over the para isomer. The origin of this preference is traced to the presence of a concerted action of noncovalent interactions (NCIs), primarily between the catalyst and the substrate, in the regiocontrolling transition states (TSs). Molecular insights into such TSs revealed that the N-H/O interaction between the tethered urea moiety of the Ir-bound ubpy ligand of the catalyst and the amide carbonyl of the substrate is a critical interaction that helps orient the meta C-H bond nearer to iridium. Other NCIs such as C-H/p between the substrate and the catalyst, C-H/O involving the substrate C-H and the oxygen of the B 2 pin 2 ligand and C-H/N between the substrate and the N atom of the Ir-bound ubpy confirm the significance of such interactions in providing the desirable differential energies between the competing TSs that form the basis of the extent of regioselectivity. ## Introduction Selective activation of thermodynamically strong and kinetically inert C-H bonds has garnered the attention of chemists for decades. Among the several activation strategies available, functionalization via C-H bond activation using a borylation reaction is a promising one due to the wider utility of the borylated products. 1 The C-H borylation reactions witnessed a number of interesting developments encompassing a range of transition metals such as Co, Ni, Ru, Rh, Pd and Ir. 2 Some of the most important examples in this genre employ iridium catalysts in conjunction with the prototypical B 2 pin 2 (bis(pinacolato)diboron) as the borylating agent. In this regard, use of pre-catalysts such as [Ir(X)(cod)] 2 (where X ¼ Cl and OMe) and bipyridine ligands in the activation of the C(sp 2 )-H bond is noteworthy. 3 A lot of effort has been expended toward developing selective activation of aryl C-H bonds, wherein one typically strives to achieve control over ortho, meta and para functionalization. The functional group (FG) directed borylation is an effective protocol for imparting ortho selectivity. 4 Along the similar lines, efforts for achieving meta selectivity 5 continued to receive attention due to the synthetic value of the meta functionalized products. 6 Development of a functionalization strategy without having to use an additional directing group (DG) on the substrate, is certainly a great advantage. 7 It will, therefore, be of inherent value if the catalyst could perform the directing role such that the method can be utilized for a broader range of substrates. 8 Such an approach would help reduce the proportion of a DG from stoichiometric to catalytic levels. A number of such endeavors where the catalyst is tailored to perform the role of a directing group rely on the careful control/use of weak noncovalent interactions (NCIs). 9 Harnessing NCIs as a handle to gain regiocontrol in transition metal catalyzed C-H activation reactions remains much less explored at this stage of development. NCIs when operating in a concerted manner are known to impact the stereochemical outcome of reactions, 10 which is an idea that could be exploited in the catalyst design for regioselective transformations as well. Within the NCI directed C-H functionalization strategies, two distinct methods employing [Ir(OMe)(cod)] 2 and bipyridine-derived ligands as the catalytic system have been reported very recently. While the approach developed concurrently by Kanai 11 and Chattopadhyay 12 proposes a secondary interaction as responsible for directed C-H functionalization, the other one by Phipps demonstrated an ion-pair directed regiocontrol. 13 In keeping with our continued efforts in probing the mechanism and selectivity controlling factors in transition metal catalyzed C-H functionalization reactions, 14 we became interested in examining the important meta C(sp 2 )-H borylation of aryl amides using [Ir(OMe)(cod)] 2 (Scheme 1). The observed regioselective borylation was proposed to arise from a hydrogen bonding interaction between the catalyst and the substrate. 15 Although the experimental studies on iridium catalyzed ortho borylation reactions are widely available, the current mechanistic understanding of meta C-H borylation is inadequate. Furthermore, rationalization of regioselectivity in these reactions typically invoked qualitative geometric features of certain putative intermediates or that of a transition state (TS) in a proposed mechanistic pathway. Hence, a number of vital geometric and energetic aspects responsible for the observed product distribution remain vague at this stage. Using DFT computations (SMD (p-xylene) /B3LYP-D3/6-31G**,SDD(Ir)), we aim to gain insights into (a) the energetic details of the catalytic cycle, (b) molecular geometry as well as electronic features of the critical intermediates and TSs, (c) the origin of regioselectivity, and (d) how changes in the substituents of the catalyst and/or substrate could impact the regiochemical outcome. The knowledge on the origin of regioselectivity would help make more rational modifcations to the substrate and/or ligand as well as to expand the scope of such catalytic reactions. ## Computational methods All computations were performed using the Gaussian09 (Revision D.01) suite of quantum chemical programs. 16 We employed the hybrid density functional B3LYP 17 with Grimme's dispersion correction (D3) 18 in combination with the Stuttgart-Dresden double-z zeta basis set (SDD) 19 with an effective core potential for 60 inner electrons out of 77 total electrons for iridium and the 6-31G** basis set 20 for all the other elements. Similar computational methods were successfully employed in the study of transition metal catalyzed reactions. 21 All stationary points identifed as TSs were characterized by one and only one imaginary frequency that corresponds to the expected reaction coordinate. The intrinsic reaction coordinate (IRC) calculations 22 were also done at the same level of theory to examine whether the TS is connected to the reactant/product as desired. The effect of the solvent was taken into account using the SMD solvation model in para-xylene as the continuum dielectric (3 ¼ 2.27). 23 The free energies of all the TSs and intermediates reported in the manuscript were obtained by adding the thermal and entropic corrections with the quasi rigid-rotor harmonic oscillator approximation 24 to the electronic energies in the condensed phase. Thus, the results and discussion are presented using the Gibbs free energies obtained at the SMD (pxylene) /B3LYP-D3/6-31G**,SDD(Ir) level of theory at 298.15 K and 1 atm pressure, unless stated otherwise. Topological analyses of the electron densities were performed using Bader's Atoms-in-Molecules (AIM) using the AIM2000 software so as to analyze the weak inter-atomic interactions in various TSs. 25 Further, the regions of attractive and repulsive interactions are identifed through the generation of NCI plots. 26 The energy span of the catalytic cycle has been calculated using the energetic span model developed by Shaik and Kozuch. 27 ## Results and discussion The catalytic regioselective borylation of C(sp 2 )-H bonds of aromatic amides using B 2 pin 2 (bis(pinacolato)diboron), employing [Ir(OMe)(cod)] 2 as the pre-catalyst (cod ¼ 1,5-cyclooctadiene) in the presence of a bipyridine-derived ligand, is examined (Scheme 1). The ligand employed here is a urea-bpy (ubpy) system tethered via an ortho-phenylene linker. The Ir(III)(ubpy)tris(boryl) complex formed by the action of the ubpy ligand and the borylating agent B 2 pin 2 on the pre-catalytic Ir(I) species is considered as the active catalyst. 3b,28 While different possible confgurations of the active catalyst as well as that of the catalyst-substrate complex are frst examined using a model system, all species involved in the catalytic cycle presented in the manuscript employ only the real system. 29 ## Important details of the catalytic cycle The key mechanistic steps in the overall catalytic cycle, starting with the formation of a catalyst-substrate complex are shown in Scheme 2. To generate adequate space to accommodate the substrate near the iridium center, a higher energy confguration (A2) of the active catalyst is considered. 30 Depending on the site of interaction of the substrate with the catalyst, three distinct coordination modes, such as an Ir/p (when the aromatic amide is coordinated through the aryl p ring), Ir/N (nitrogen of the amide) and Ir/O (carbonyl oxygen of the amide) binding, are examined. 31 The computed energies suggest that the substrate coordination to the iridium center via oxygen or nitrogen atom of the amide group is equally feasible as they differ only by a kcal mol 1 . 32 The optimized geometries of the catalyst-substrate complexes in both these binding modes convey that only the ortho aryl C-H bond is close enough to the iridium center for any effective interaction. In other words, when N,N-dihexylbenzamide is bound to the catalyst through its amide moiety, the meta and para positions remain far from the iridium center to afford C-H bond activation. 33 In the p-binding mode, on the other hand, all the three aryl C-H bonds, including the meta enjoy enhanced proximity to the iridium center that can lead to effective functionalization. 34 Thus, the p-binding mode in the catalyst-substrate complex is considered as the reactive conformer in our study as shown in Scheme 2. Since the N-hexyl chains are conformationally flexible, as many as 8 conformers of all the important TSs are identifed. 35 The results presented herein are on the basis of the most favorable geometry for both the meta and para C-H activation TSs. The key mechanistic events in the catalytic cycle, as shown in Scheme 2, start with an oxidative addition in the catalystsubstrate complex 1 wherein Ir(III) inserts into the meta C-H bond via transition state TS[1-2] m . The ensuing Ir(V)-aryl intermediate 2 then undergoes a reductive elimination (RE) through TS[2-3] m to generate the borylated product and an Ir(III)-hydride intermediate ( 3). Uptake of one molecule of B 2 pin 2 by 3 can then lead to a weakly interacting complex between 3 and B 2 pin 2 , which is denoted as 4. In the following step, insertion of B 2 pin 2 into 4 generates a hepta-coordinate Ir(V) species ( 5 The Gibbs free energy profle for the formation of the meta borylated product is provided in Fig. 1. The formation of the borylated product exhibits notable barriers of 21.6 and 21.5 kcal mol 1 , respectively, for the C-H activation and the RE. According to the energetic span model, 27 intermediate 1 and TS [1-2] m are, respectively, the turnover determining intermediate (TDI) and turnover determining transition state (TDTS). The activation span (dE), calculated as the energy difference between the TDTS and TDI, is found to be 21.6 kcal mol 1 . 37 The subsequent steps in the mechanism, such as the B 2 pin 2 insertion and the catalyst regeneration, are much less energy demanding, as indicated by their relatively lower barriers. ## Factors controlling regioselectivity The knowledge of energetics, geometry, and electronic features of the important TSs involved in a catalytic cycle would be valuable toward developing a better understanding of the catalytic transformation. For the present example, such features of the regio-controlling TSs enabling meta C-H activation are not established yet. Kanai's insightful working hypothesis, 15 on the other hand, placed a signifcant emphasis on the H-bonding between N,N-dialkylbenzamide and the urea moiety of the Irbound ubpy ligand as the key factor in their proposed TSs. The observed meta to para ratio of 27 corresponds to a selectivity of 93% in favor of meta borylation. To probe this important observation in greater depth, TSs for the meta and para C-H activation of substrates such as N,N-dihexylbenzamide (S0) and N,N-dimethylbenzamide (S1) are identifed. Different variations in the catalyst (ubpy), and the borylating agent (B0) are also considered in this study. The original catalyst-substrate combination is referred to as ubpy-S0 while that with a modifed substrate is denoted as ubpy-S1. Optimized geometries of the regiocontrolling TSs for S0 and S1 are provided in Fig. S2 Although the above-mentioned analysis, based on the difference in the number of NCIs, offers a qualitative insight into the origin of meta selectivity, it does not provide room for quantitative assessment. For instance, a lesser number of more efficient interactions can outweigh the influence of a greater number of weaker interactions. Hence, we endeavored to quantify the important NCIs using the topological parameters such as the electron density at the bond critical point (r bcp ), the corresponding Laplacian of the electron density (V 2 r), and kinetic energy (G) using Espinosa's formulation. 38 Although Espinosa's formulation was proposed for the quantifcation of isolated pairwise intermolecular H/F interactions, we have extended the same to various intramolecular weak NCIs operating in important TSs in the present study. Even though the formulation has not been applied to complex intramolecular interactions such as that prevail in regiocontrolling TSs, we believe that it could provide a reasonable measure of the relative Since the dihexyl chain of the substrate engages in a good number of C-H/p interactions with the catalyst, we wanted to examine whether changes in such interactions might affect the extent of regioselectivity. To this end, a modifed substrate S1 is considered wherein the N,N-dihexyl group on the amido nitrogen is replaced with a N,N-dimethyl substituent. The optimized transition state geometries are devoid of C-H/p interactions, except one such prominent contact between the cyclohexane of the urea moiety and the aryl group of the substrate (shown as f in Fig. 2). This interaction is present only in the case of TS[1-2] m(S1) but not in TS[1-2] p(S1) . Interestingly, a number of other NCIs such as C-H/O and C-H/N interactions are found to be present in both the meta and para TSs for N,N-dimethyl amide as the substrate. It is also of interest to note that the NCIs in the case of the S1 system are not as pronounced as those in S0 (Fig. 3). Further, the total strength of all the important NCIs is estimated to be 70.9 kcal mol 1 for TS[1-2] m(S1) and only 63.5 kcal mol 1 for TS[1-2] p(S1) . The regioselectivity, calculated using the energy difference of 7.4 kcal mol 1 between the competing meta and para C-H activation TSs for the N,N-dimethylamide is strongly in favor of the meta isomer. The summary of noncovalent interactions shown in Fig. 3 conveys that except for the C-H/O contacts, all the other NCIs Fig. 2 Optimized geometries of the TSs for the meta and para C-H activation of ubpy-S0 and ubpy-S1 catalyst-substrate combinations. The distances are in and electron densities (r 10 2 au) at the bond critical points are given in parentheses. The hydrogen atoms that are not involved in any noticeable interaction are removed for improved clarity. are more dominant in TS[1-2] m(S0) (shown in green) than in TS [1-2] p(S0) (blue) for the ubpy-S0 catalyst-substrate pair. It should be noted that TS[1-2] m(S0) is 6.6 kcal mol 1 lower in energy than TS[1-2] p(S0) , suggesting that the NCIs do have a direct influence on the extent of regioselectivity. In the case of the S1 system, though the number of C-H/p contacts is much smaller, other NCIs such as C-H/O and N-H/O interactions are able to effectively make TS[1-2] m(S1) lower in energy by 7.4 kcal mol 1 as compared to TS[1-2] p(S1) . Thus, it appears that the C-H/p interactions may not solely be responsible for the observed high regioselectivity if the cumulative effect of other weak interactions is able to compensate. It is also interesting to note that the difference in the total strength of NCIs between meta and para C-H activation TSs exhibits a good correlation with the predicted regioselectivity. The predicted preference toward the meta C-H activation for ubpy-S0 and ubpy-S1 is, respectively, 6.6 and 7.4 kcal. This is in line with the experimentally observed meta to para ratio of 27 for both the catalyst-substrate pairs. Appreciable difference in the collective strength of NCIs between the meta and para C-H activation TSs is noted in the case of ubpy-S0 (23.4 kcal mol 1 ) and ubpy-S1 (7.4 kcal mol 1 ), thus favoring meta C-H activation over the alternative para pathway. Hence, a combination of C-H/p, C-H/N and C-H/O interactions together with the N-H/O H-bonding makes the meta C-H activation TS lower in energy than the corresponding para position. 40 While it is prudent to acknowledge that various noncovalent interactions described above impact the predicted relative energy order between the regiocontrolling TSs, analysis of the effect of distortion in such TSs is equally important. The Distortion-Interaction/Activation-Strain (DI-AS) analysis 41 was therefore carried out on these TSs to gain additional insights into the factors responsible for the observed regioselectivity. It can be noticed from the data provided in Table 1 that in both ubpy-S0 and ubpy-S1 catalyst-substrate combinations, the para C-H activation TSs experience a higher total distortion relative to the corresponding meta analogue. The extent of distortion in the substrate, as well as the catalyst in the para TS, is found to be higher than that in the meta case for ubpy-S1. 42 Thus, due to the combined effect of the higher number of NCIs operating in concert as well as the relatively lower distortion experienced by both the meta TSs, we could rationalize the preference toward the high meta regioselectivity in both the above-mentioned examples. ## Rational modications to change the pattern of NCIs After having understood the critical role of various NCIs as well as the N-H/O H-bonding in the C-H activation step, we considered two new modifcations of the parent system. As the N-H/O interactions were also found to play a vital role in imparting selectivity, the parent ligand ubpy is modifed by removing the phenylene-urea linker to a simple bpy. Similarly, instead of B 2 pin 2 (B0), we have considered B 2 (OMe) 4 (B1) as the borylating agent. 43 The change in the regioselectivity due to these modifcations is predicted for the substrate S0. The bpy ligand led to no energy difference between the meta and para C-H activation TSs, implying no regioselectivity. In other words, turning off the vital N-H/O interactions between the substrate and the catalyst, by way of removing the urea moiety diminishes the energetic advantage toward the meta C-H activation as compared to the competing para analogue. 44 Similar to the case with the bpy ligand, the B1 modifcation also leads to a relatively smaller energy difference between the meta and para TSs (2.6 kcal mol 1 ) compared to the unmodifed system (6.6 kcal mol 1 ). 45 A similar trend in selectivity is also noticed when computed using relative enthalpy differences between the meta and para TSs for the aforesaid modifcations. 46 A detailed analysis of the meta and para C-H activation TSs of these modifed systems based on Espinosa's formulation 47 and the DI-AS analysis 48 is thus performed to assess how NCIs impact the regiochemical outcome of this reaction. We note that the prominent NCIs other than the N-H/O Hbonding in these modifed systems are the C-H/O and C-H/ N interactions. A quantifed NCI, as given in Table 2, conveys that for the bpy modifcation, the total strength of the C-H/O interactions in TS[1-2] p(bpy) (36.7 kcal mol 1 ) is higher than that in the meta TS (31.3 kcal mol 1 ), while TS[1-2] m(bpy) enjoys improved C-H/N interaction (8.9 kcal mol 1 ) than in the para counterpart (5.5 kcal mol 1 ) (Table 2). 47 While the predicted strength of the C-H/O interactions appears to be overestimated, it serves the present purpose wherein we intend to compare the relative strengths of such interactions in chemically identical meta and para C-H activation TSs. Another interaction, namely, the N-H/O interaction is absent in both the TS geometries, and four more C-H/p interactions (9.8 kcal mol 1 ) are found in the meta TS than that in the para TS. a Sum of the strength of key NCIs (in kcal mol 1 ) calculated using Espinosa's method. b This interaction is absent. Thus, the absence of some of the important NCIs in TS[1-2] m(bpy) and TS[1-2] p(bpy) appears to result in comparable energies for both these TSs, which in turn, leads to very low regioselectivity. 47 In the case of B1 modifcation, equal numbers of C-H/O interactions are identifed in both meta and para TSs, albeit the cumulative strength of such interactions is higher in para (48.1 kcal mol 1 ) than that in meta (37.8 kcal mol 1 ) TS. However, a greater number of stronger N-H/O interactions are found in the meta TS than in para (Table 2). The C-H/N interactions, on the other hand, are found to be of comparable strengths in the meta (11.9 kcal mol 1 ) and para C-H activation TSs (9.3 kcal mol 1 ). The lack of notable differences in the noncovalent interactions can be regarded as the origin of the small energy difference (2.6 kcal mol 1 ) between the two competing TSs. 47 The ortho conundrum. The experiments suggested the formation of meta and para borylated products, but no ortho product. 15 This is somewhat surprising as no particular rationale was offered as to why an ortho borylated product was not observed. In principle, all the three C-H bonds at ortho, meta and para positions could be accessible for the oxidative addition. In line with this expectation, our computed data indicate that the C-H activation at the ortho position is more favorable than that at the para position. Importantly, the meta C-H activation is energetically the most favorable possibility, followed by ortho and then para activations. The barriers for the C-H activation step with respect to the respective preceding intermediates are found to be 21. 6, 23.8 and 25.3 kcal mol 1 , respectively, for meta, ortho, and para positions. To inspect whether this predicted preference arises due to the use of a particular computational approach, we have also computed the Gibbs free energies using a range of different density functionals and basis sets. 49 All such additional computations yielded similar energetic trends, suggesting that the ortho C-H activation cannot be ruled out on the basis of the frst key step in the catalytic cycle. Since clarity on whether ortho C-H activation is more likely than that at the para position under the experimental conditions could not be sought on the basis of the computed energetics of the C-H activation step, we have carefully examined the ensuing steps of the catalytic cycle in greater detail. Interestingly, the Ir(V)-aryl intermediate (2) formed as a result of the C-H activation at the ortho as well as para positions is found to be more reversible than the one derived at the meta position. 50 Furthermore, a comparative study of the reductive elimination (RE) leading to the formation of the borylated product is undertaken for the ubpy-S0 catalyst-substrate combination. An elementary step barrier of 21.5 kcal mol 1 is found for the RE to a meta borylated product, which is in concert with the experimental preference toward the meta product. However, the RE elimination barrier at the para position is 27.4 kcal mol 1 while that at the ortho position is found to be 32.8 kcal mol 1 , suggesting that the para product would be the next most likely product other than the meta, in accord with the experimental observations. Though the pathways appear to compete in the initial oxidative addition step, the application of the energetic span model also helped us understand the regioselectivity more convincingly. 51 While the TDI is the respective catalystsubstrate complex (1) in all these pathways, the TDTS for the meta pathway is found to be the oxidative addition and that for the para and ortho pathways it is the reductive elimination. The corresponding energetic span dE for para and ortho is, respectively, 27.3 and 32.8 kcal mol 1 . In other words, the catalytic turnover toward ortho is the least favored, followed by the para product. It is also interesting to note that the relative enthalpies of the relevant TSs also convey a similar trend in the predicted selectivities. 52 Thus, the overall energetic features of various borylation pathways suggest the formation of the meta product as the major and para as the minor regioisomer in this catalytic transformation. 53 ## Conclusion Density functional theory investigation of important Ir(III)catalyzed meta selective aryl C(sp 2 )-H borylation revealed that the key mechanistic steps in the lower energy pathway are (i) oxidative addition (C-H activation), (ii) reductive elimination (borylation), and (iii) catalyst regeneration. A comparison of energies and stereoelectronic factors operating in the C-H activation transition states for meta and para functionalization of N,N-dihexylbenzamide helped us gain signifcant new insights into the role of various noncovalent interactions, particularly between the catalyst and the substrate. The catalyst Ir(III)(ubpy)tris(boryl), decorated with a phenylene-urea tether on the bpy ligand is found to play an important role in positioning the aryl amide through N-H/O H-bonding interactions such that the C-H activation at the meta position is rendered energetically more favorable over that at para. However, we noted that the presence or absence of this H-bonding interaction could not solely account for the regioselectivity. A good number of noncovalent interactions between the catalyst (Irbound ligands) and the substrate are found to be vital toward bringing about the energy difference between the meta and para C-H activation TSs. These interactions, operating between the C-H bonds of the substrate (hexyl and aryl moieties) and (i) the bipyridyl nitrogen atoms as well as the p-face of the Ir-bound ubpy ligand and (ii) the oxygen atoms of B 2 pin 2 , are found to be more prominent in the meta C-H activation transition state, thereby making it 6.6 kcal mol 1 lower in energy than the para analogue. This energy difference is fully consistent with the experimental observation of the high meta to para ratio of 27. Additional series of computations on modifed systems obtained by changing the substrate (replacing dihexyl with dimethyl), catalysts without the urea moiety on the Ir-bound ubpy ligand, and the use of B 2 (OMe) 4 instead of B 2 pin 2 as the borylating agent further helped us conclude that the balance between C-H/p, C-H/O and C-H/N NCIs that operate between the catalyst and the substrate is more important than the primary N-H/O H-bonding contact that binds the substrate to the catalyst. For instance, the meta C-H activation TS for the N,N-dimethylbenzamide is noted to enjoy a larger number of relatively better NCIs thereby maintaining high regioselectivity even though the C-H/p interactions are not as much as those in N,N-dihexylbenzamide. For the catalysts devoid of N-H/O interactions (bpy and dtbpy), the low selectivity could be attributed to the absence of differentiating NCIs between the meta and the para TSs. With the modifed borylating agent, the predicted lower selectivity relative to the parent system is found to be due to the presence of similar efficiencies in the C-H/O and C-H/N interactions in the para TS to those in the meta TS. The regioselectivity of the borylation reaction thus hinges upon a set of NCIs that operate in concert and hence could be fne-tuned by making a rational choice of the ligand on the catalyst and suitable reactants. These conclusions are expected to have broader applicability in developing catalytic regioselective protocols using noncovalent interactions. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Insights into the role of noncovalent interactions in distal functionalization of the aryl C(sp²)\u2013H bond", "journal": "Royal Society of Chemistry (RSC)"}

reinterpreting_π_π_π-stacking

## Abstract: The nature of π-π interactions has long been debated. The term "π-stacking" is considered by some to be a misnomer, in part because overlapping π-electron densities are thought to incur steric repulsion, and the physical origins of the widely-encountered "slip-stacked" motif have variously been attributed to either sterics or electrostatics, in competition with dispersion. Here, we use quantum-mechanical energy decomposition analysis to investigate π-π interactions in supramolecular complexes of polycyclic aromatic hydrocarbons, ranging in size up to realistic models of graphene, and for comparison we perform the same analysis on stacked complexes of polycyclic saturated hydrocarbons, which are cyclohexane-based analogues of graphane. Our results help to explain the short-range structure of liquid hydrocarbons that is inferred from neutron scattering, trends in melting-point data, the interlayer separation of graphene sheets, and finally band gaps and observation of molecular plasmons in graphene nanoribbons. Analysis of intermolecular forces demonstrates that aromatic π-π interactions constitute a unique and fundamentally quantum-mechanical form of non-bonded interaction. Not only do stacked π-π architectures enhance dispersion, but quadrupolar electrostatic interactions that may be repulsive at long range are rendered attractive at the intermolecular distances that characterize π-stacking, as a result of charge penetration effects. The planar geometries of aromatic sp 2 carbon networks lead to attractive interactions that are "served up on a molecular pizza peel", and adoption of slip-stacked geometries minimizes steric (rather than electrostatic) repulsion. The slip-stacked motif therefore emerges not as a defect induced by electrostatic repulsion but rather as a natural outcome of a conformation landscape that is dominated by van der Waals interactions (dispersion plus Pauli repulsion), and is therefore fundamentally quantum-mechanical in its origins. This reinterpretation of the forces responsible for π-stacking has important implications for the manner in which nonbonded interactions are modeled using classical force fields, and for rationalizing the prevalence of the slip-stacked π-π motif in protein crystal structures. ## Introduction Is there a special type of dispersion associated with π-π interactions? Some studies suggest that there is, citing the relationship between the π-stacking distance in aromatic π-π systems and the strength of the dispersion interaction. 1 Others point out that aromaticity is not a necessary condition for obtaining augmented dispersion in π-electron systems, and in fact can sometimes lead to additional Pauli (steric) repulsion that diminishes the attractive interaction. 2,3 In view of this, structural rigidity of the interacting a Department of Chemistry & Biochemistry, The Ohio State University, Columbus, OH, moieties may be a more incisive metric for predicting enhanced attraction. 4 Complicating the picture is the fact that aromatic rings often possess large quadrupole moments, 5,6 bringing an electrostatic angle to the problem, and this consideration has fomented a suggestion that the term "π-stacking" should be reconsidered altogether. 7 Arguments based on classical multipole moments, however, seem ill-suited to explain the prevalence of the slip-stacked motif between aromatic side chains in protein crystal structures, where the data presumably sample a broad range of electrostatic environments. Nevertheless, quadrupolar electrostatics is a recurring theme in discussion of π-π interactions, and has long been the principle paradigm through which paralleldisplaced π-stacking has been rationalized. This conventional wisdom persists despite considerable evidence that charge penetration effects, which nullify or at least complicate classical electrostatic arguments, are significant at typical π-stacking dis- tances. Benzene dimer is the archetypal π-stacked system and its conformational preferences are traditionally discussed in terms of several geometric isomers that are depicted in Fig. 1. The cofacial geometry (Fig. 1a) represents canonical π-stacking, although for (C 6 H 6 ) 2 in the gas phase this geometry is an energetic saddle point along a sliding coordinate leading to the parallel-offset (or slip-stacked) geometry in Fig. 1b. 30 The slip-stacked isomer is a local minimum, and is nearly iso-energetic with the T-shaped isomer depicted in Fig. 1c. 13, The perfectly perpendicular Tshaped geometry is a saddle point in the gas phase, 32 and tilts by a few degrees along the pendular CH• • • π coordinate to lower the energy by 0.2 kcal/mol, 32,33 but this will not concern us here. In fact, we will argue that benzene dimer is not representative of π-π interactions in larger polycyclic aromatic hydrocarbons (PAHs), and thus undeserving of its paradigmatic status. The traditional explanation for the geometry preferences of (C 6 H 6 ) 2 , as formalized long ago by Hunter and Sanders, 11 is based on a competition between attractive dispersion and repulsive quadrupolar electrostatics. While the Hunter-Sanders model correctly predicts a slip-stacked structure for (C 6 H 6 ) 2 , in agreement with ab initio calculations, it does not explain the fact that (C 6 H 6 )• • • (C 6 F 6 ) also exhibits a parallel-offset structure, despite quadrupolar electrostatic interactions that are attractive in the cofacial arrangement. Various studies have since suggested that the Hunter-Sanders model exaggerates the role of electrostatics, 3, 32, however this model remains a widelydiscussed paradigm for π-π interactions, highlighted in contemporary textbooks. 14,15 We have recently provided a clear and concise demonstration that the importance of electrostatics in π-π interactions has been misconstrued, and that the Hunter-Sanders model does not simply "overemphasize" electrostatics, 1 but is in fact qualitatively wrong and represents a fundamentally flawed framework for understanding π-π interactions. 29 Rather than being dictated by quadrupolar electrostatics, conformational preferences in systems such as (C 6 H 6 ) 2 and (C 6 H 6 )• • • (C 6 F 6 ) are instead driven by van der Waals (vdW) interactions, by which we mean a combination of dispersion and Pauli repulsion. The vdW model provides a unified explanation for the emergence of a slip-stacked geometry in both cofacial (C 6 H 6 ) 2 , where the quadrupolar interaction is re-pulsive, but also in (C 6 H 6 )• • • (C 6 F 6 ), were the polarity of the C-F bonds reverses the sign of the C 6 F 6 quadrupole moment, relative to that of C 6 H 6 . 5,6 The problem with the classical quadrupole model is that it fail to account for charge penetration at short range. 24,27,29,42 Note that charge penetration is a fundamentally different concept than intermolecular charge transfer. 21 The latter describes a particular form of polarization, whose definition can be quite sensitive to the choice of orbitals and basis set, 43 but which is rather small for the systems considered here. This may be inferred experimentally by the absence of significant vibrational frequency shifts upon complexation, even in systems like (C 6 H 6 )• • • (C 6 F 6 ), 6 and also theoretically by the rather small induction energies that are reported in this work. Instead, the term "charge penetration" describes the fact that a low-order multipole expansion may misrepresent the electrostatic interaction energy, specifically at short intermolecular distances where the monomer charge densities interpenetrate. Unlike the problematic definition of charge transfer, 43 however, there is no ambiguity in the definition of E elst because ρ A (r) and ρ B (r) are isolated monomer densities. Any deviation between eq. 1 and a multipolar approximation is a manifestation of charge penetration. In benzene dimer, charge penetration effects largely mitigate the electrostatic preference for a cofacial versus a slip-stacked arrangement, and the latter emerges as the preferred geometry due to a competition between dispersion and Pauli repulsion, rather than between dispersion and electrostatics. 29 This can be modeled using a simple vdW (repulsion + dispersion) potential that reproduces ab initio geometries for benzene dimer, naphthalene dimer, and (C 6 H 6 )• • • (C 6 F 6 ). 29 Offset π-stacking can thus be understood without appeal to electrostatics at all! This helps to rationalize the persistence of the offset-stacked motif in the π-π side-chain interactions in proteins, which are revealed by data-mining studies of the protein data bank. In view of this new interpretation of π-stacking, it seems pertinent to revisit old questions regarding whether π-π interactions truly constitute a unique form of dispersion. The concept of π-stacking has elicited controversy, perhaps due to an incomplete definition of the phenomenon. The terminology seems to suggest significant overlap between π-electron clouds of two moieties in a cofacial arrangement. From the standpoint of dispersion, which varies with distance as ∼ ᾱ/R 6 where ᾱ denotes the isotropic polarizability, the cofacial arrangement minimizes interatomic distances and therefore maximizes the attraction due to dispersion. On the other hand, exchange repulsion (i.e., steric or Pauli repulsion) is proportional to the overlap integral S between molecular orbitals and decays as ∼ S 2 /R. 44,45 Any overlap between π clouds is therefore repulsive to some extent. Recent work by Tkatchenko and co-workers has also highlighted the role of charge-density fluctuations in stabilizing nanoscale π-π interactions. 46,47 Grimme 1 and others 48 have examined stacking of both aro-matic and saturated hydrocarbons as a function of size, concluding that for larger acene dimers there is a clear enhancement of the interaction energy in cofacial arrangements, beyond what is seen in perpendicular orientations that are analogous to the Tshaped isomer of (C 6 H 6 ) 2 . Interaction energies between saturated hydrocarbons exhibit size dependence that is much closer to that of perpendicular acene dimers. 1,48 One goal of the present work is to reexamine these size-dependent trends in view of our new understanding of the role of vdW forces. The role of electrostatics is more complicated. Grimme's analysis is framed against the backdrop of the Hunter-Sanders model, 1 with its assumption that electrostatic interactions are repulsive in cofacial π-stacked arrangements and that this repulsion drives offset-stacking. In fact, charge penetration effects are significant at typical π-stacking distances, as documented by Sherrill and by others. 27,28 In acene dimers, for example, the exact electrostatic interaction energy computed using eq. 1 deviates from the leading-order quadrupolar result by as much as 50% at crystal-packing distances. 27 That said, previous ab initio studies of electrostatic effects in π-stacked systems have focused on single-point energy decompositions or on the intermolecular separation coordinate. As we showed previously for (C 6 H 6 ) 2 , 29 the role of vdW forces in determining the conformational landscape emerges only upon consideration of the potential energy surface for sliding one molecule across the other. In the present work, we extend this analysis to acene dimers up to (pentacene) 2 , to benzene on the surface of a C 96 H 24 graphene nanoflake, and to corannulene dimer, which is less structurally rigid and bows significantly in its equilibrium geometry. In the course of this analysis, we also make the first detailed examination of the effects of charge penetration in these larger π-stacked systems. We revisit the question of whether π-stacking constitutes an exceptional form of dispersion, using quantum-mechanical energy decomposition analysis based on symmetry-adapted perturbation theory (SAPT). 21, Side-by-side comparison of results for PAHs with their saturated polycyclic analogues (fused cyclohexane ring systems) reveals that there are indeed unique aspects of dispersion interactions in aromatic systems. These feature ultimately originate in the fact that PAHs are planar and structurally rigid, which facilitates exceptionally close-contact interactions via vdW forces. In this close-contact regime, electrostatic interactions become attractive even in cofacial geometries where they might be asymptotically repulsive. At the intermolecular separations that typify π-stacking, the interaction potential is dominated by vdW effects that drive charge penetration, nullifying the classical electrostatic picture. This implies that π-stacking is not solely attributable to a unique form of dispersion, but conspires with molecular geometry to afford a unique combination of electrostatic attraction and the vdW interactions in flat, rigid molecules. ## Computational Details Interaction energies are calculated using the extended "XSAPT" version of second-order SAPT, which includes a variational description of polarization for electrostatics. 55 Monomer wave functions were computed using the LRC-ωPBE functional, 56,57 tuning the range-separation parameter ω as described in previ-ous work. 52,57 Tuned values of ω can be found in Table S1. Induction energies reported here include a "δ E HF " correction, 50 in which a Hartree-Fock calculation for the dimer is used to estimate polarization beyond second order in perturbation theory. In results presented below, the induction (or polarization) energy is defined as where E (2) ind and E (2) exch-ind are the second-order SAPT induction and exchange-induction components. First-order SAPT electrostatics (E (1) elst , eq. 1) and exchange (E exch ) energies will simply be reported as E elst and E exch , respectively. In place of the usual second-order SAPT dispersion terms, which tend to be the least accurate contributions to the secondorder version of the theory, 53,57,58 we use a self-consistent manybody dispersion (MBD) method, 59 which is a modified form of the MBD approach introduced by Tkatchenko et al. for use with density functional theory. The MBD formalism goes beyond an atomic-pairwise description of dispersion to include screening effects between multiple polarizable atomic centers in a selfconsistent fashion, which is likely to be important for conjugated π-electron systems. 63 Combined with electrostatic, induction, and Pauli repulsion energies computed using SAPT, the resultant XSAPT+MBD method is a computationally efficient way to calculate benchmark-quality noncovalent interaction energies in large supramolecular complexes. 55,59 These calculations were performed using Q-Chem v. 5.3. 64 Geometries for all complexes were optimized at the TPSS-D3/def2-TZVP level of theory, 65,66 and are unconstrained except where noted. (Constrained optimizations are reported for corannulene dimer and these were performed using the ORCA software, v. 4.1.1. 67 ) In order to account for deformation in the large graphene flake that is considered here, geometries of the (C 96 H 24 )• • • (C 6 H 6 ) complex were optimized at each point on a two-dimensional potential energy surface, essentially scanning the center position of C 6 H 6 over the two-dimensional plane of C 96 H 24 . Potential energy surfaces for the naphthalene and decalin (perhydronaphthalene) dimers, computed along a twodimensional cofacial sliding coordinate, do not include monomer deformation. In these cases, a parallel configuration is used with a face-to-face separation of 3.4 for (naphthalene) 2 and 4.6 for (decalin) 2 . ## Results and Discussion A large body of research on π-π interactions has focused on benzene dimer, both because it is amenable to high-level ab initio calculations and because it is regarded as emblematic of πstacking. Conformational preferences in (C 6 H 6 ) 2 are framed as a competition between London dispersion, favoring the cofacial π-stacked arrangement (Fig. 1a), and quadrupolar electrostatics that favor a perpendicular configuration (Fig. 1c). 10,11,15 Accurate calculations suggest that these two configurations are nearly iso-energetic, 13, and indeed the short-range structure of liquid benzene that is inferred from neutron diffraction experiments is consistent with the coexistence of both orientations. 68 It happens that the interaction energy (stacking energy) in benzene dimer is nearly identical to that in of cyclohexane dimer. 69 This raises the question of whether the former is representative of π-π interactions more generally, or indeed whether such interactions are genuinely distinct from "ordinary" (and ubiquitous) London dispersion. 1,7 In arguing that they are not, it is sometimes pointed out that C 6 H 12 has a larger (isotropic) polarizability as compared to C 6 H 6 , 7 although this argument misses the point that polarizability is an extensive quantity and the polarizability per electron is slightly larger in C 6 H 6 than it is in C 6 H 12 . 70 This observation suggest that in comparing aromatic to saturated hydrocarbons, a comparison of size-dependent trends may afford insight, and this is what we consider first. ## Size-Dependent Trends We first examine size-dependent trends amongst dimers of linear acenes, (C 4n+2 H 2n+4 ) 2 , with the number of rings ranging up to n = 5 (pentacene). Both perpendicular and parallel-offset geometries are considered, as shown in Fig. 2. We also consider dimers of the complementary polycyclic saturated hydrocarbon (PSH) molecules, the perhydroacenes, ranging from cyclohexane dimer through perhydropentacene dimer, (C 22 H 36 ) 2 . Interaction energies of the linear acenes have been reported elsewhere, 1,71 and our XSAPT+MBD interaction energies are in line with previous computational work. The present calculations capture the energetic similarities that are expected in the single-ring systems, 69 as the stacking energies of benzene and cyclohexane dimers are within 0.1 kcal/mol of one another. This degeneracy is lifted when just one more ring is added, as the parallel-offset geometry of naphthalene dimer emerges as the most stable of the two-ring structures depicted in Fig. 2, by 1.3 kcal/mol. This prediction is corroborated by experimental neutron diffraction data for liquid naphthalene, which exhibit a clear propensity for parallel-offset configurations, 72 unlike the corresponding data for liquid benzene. 68 Evolution of the size dependence of the interaction energies is presented in Fig. 3a. These data demonstrate that enhanced attraction with respect to the length of the acene nanoribbon is unique to the cofacial arrangement of these aromatic dimers; interaction energies for perpendicular configurations of the PAH dimers remain nearly identical to those for the stacked PSH dimers even as the size of the monomer unit is increased. In contrast, intermolecular attractions in the parallel-offset PAHs is amplified with the addition of each ring until the energy difference between parallel-offset (pentacene) 2 and the other two n = 5 ring systems (perpendicular pentacene dimer and stacked perhydropentacene dimer) exceeds 6 kcal/mol. All else being equal, stronger intermolecular attraction means larger enthalpy of vaporization and this is reflected in the boilingpoint data presented in Fig. 3b. Remarkably, these experimental data capture the similarity between interaction energies for the benzene and cyclohexane dimers, as well as the fact that adding just one ring lifts the degeneracy; the boiling point of naphthalene exceeds that of perhydronaphthalene (decalin) by 27 K. The boiling points of the aromatics increase more rapidly versus monomer size as compared to those for the saturated hydrocarbons. In view of the neutron diffraction data for liquid benzene 68 and liquid naphthalene, 72 which provide evidence for both parallel and perpendicular orientations in the former case but only parallel con-figurations in the latter, it seems reasonable to hypothesize that the boiling point increases for larger PAHs evidence a continued propensity for parallel-offset geometries in aromatics larger than benzene. Together, these data suggest that (C 6 H 6 ) 2 , rather than being a paradigmatic example, is actually a poor surrogate for aromatic π-π interactions more generally. This is consistent with studies of the size-dependent trends in (benzene) 2 , (naphthalene) 2 , and (pyrene) 2 interaction energies, 42 where it was determined that extrapolations based on smaller PAHs produces misleading results. Grimme has also suggested that any "special" aspects of dispersion in π-π interactions manifest only in aromatic moieties larger than a single benzene ring. 1 The present results are consistent with that idea but suggest that the aromatic moiety need not be much larger. Cofacial π-stacking rapidly comes to dominate the intermolecular landscape of the acene dimers as the length of the nanoribbon increases, with a widening energetic gap between the parallel-offset and the perpendicular arrangement. ## Benzene on Graphene We next investigate π-stacking in a system with disparate monomer sizes, examining the two-dimensional potential energy surface for scanning C 6 H 6 over the surface of a graphene nanoflake (C 96 H 24 ), in both cofacial and perpendicular orientations. There are no near-degeneracies in this case (see Fig. 4), and a comparison between the minimum-energy structure obtained in either orientation reveals that the cofacial arrangement is 6 kcal/mol more stable than the perpendicular configuration. The cofacial benzene probe is more stable when the center of the ring is directly atop an atom or bond of the underlying C 96 H 24 molecule, because these configurations minimize the effects of exchange repulsion. This is the benzene-graphene analogue of the parallel-offset geometry in the acene dimers, and it arises for the same reasons that we have previously discussed for the benzene and anthracene dimers. 29 In the perpendicular orientation, benzene on C 96 H 24 adopts a minimum-energy geometry in which the C-H bond of benzene points to the center of a ring on C 96 H 24 , analogous to the T-shaped isomer of (C 6 H 6 ) 2 . In previous work, 29 we developed an analytic model potential for describing π-π interactions, to serve as a replacement for the conventional Hunter-Sanders model. 11 Whereas the latter consists of an attractive London dispersion term along with point charges arranged to provide repulsive quadrupolar electrostatics, we called our analytic model a "vdW potential" because it replaces the electrostatics with an overlap-based model of Pauli repulsion. (Short-range repulsion plus long-range dispersion are the intermolecular forces that compete to yield the vdW equation of state for gases, so the nomenclature is consistent.) For the dispersion component of this vdW model, we used a pairwise atomic dispersion potential fit to ab initio dispersion data. 52 Potential surfaces for the (C 6 H 6 ) • • • (C 96 H 24 ) system, generated by both of these model potentials, can be found in Figs. S1 and S2. The Hunter-Sanders model erroneously predicts an energy minimum in which benzene sits directly above the center ring of C 96 H 24 (i.e., cofacial rather than offset stacking), at odds with the XSAPT+MBD results. In contrast, the vdW model correctly predicts that this configuration is a saddle point. The remainder of the two-dimensional XSAPT+MBD potential surface is also reproduced with high fidelity by the vdW model. Although in its present form this model is a simple parameterization designed for physical insight, it has a functional form amenable to use with classical force fields, to obtain interaction potentials for πstacking with correct underlying physics. Note that the minimum-energy point on both the parallel and perpendicular (C 6 H 6 ) • • • (C 96 H 24 ) potential energy surfaces places the benzene molecule near the center of the graphene flake. In this sense, there is no analogue of the parallel-offset structure in (C 6 H 6 ) 2 , where one aromatic molecule extends beyond the edge of the other, although the driving force for offcenter stacking in benzene-graphene is the same as that which drives the benzene dimer into a parallel-displaced geometry. Relative to the perpendicular arrangement, the cofacial orientation is strongly preferred in benzene-graphene, just as it was in acene dimers larger than (C 6 H 6 ) 2 . A similar preference for π-stacked geometries has been noted in the case of 74 implying that benzene's interactions with larger aromatic molecules more generally favor a π-stacked arrangement. Whereas the T-shaped and paralleloffset geometries of (C 6 H 6 ) 2 have nearly identical interaction energies, this degeneracy is a finite-size effect because the slipstacked arrangement must sacrifice attractive dispersion, which falls off rapidly as the π-electron clouds of the two monomers are displaced from one another. On the surface of the graphene nanoflake, however, a small offset can be introduced without loss of dispersion, and the cofacial orientation becomes strongly preferred with respect to the perpendicular arrangement. Charge penetration effects, and therefore electrostatic interactions, are also essentially unchanged by this small displacement, which serves to reduce Pauli repulsion and thus to enhance the total interaction energy. The interaction potential of perpendicular benzene on C 96 H 24 is not enhanced by parallel offsets. Maximizing the surface area of closely-interacting π-electron densities, "serving up the interactions on a platter", seems to be highly beneficial when extended π networks are considered, a fact that could not have been inferred from (C 6 H 6 ) 2 . ## Size-Intensive Energy Decomposition Dispersion is intimately tied to polarizability but this connection has sometimes been misconstrued in the context of π-π interactions, with the somewhat larger polarizability of C 6 H 12 as compared to C 6 H 6 taken as evidence that dispersion interactions in benzene dimer should not be larger than those in cyclohexane dimer. 7 Setting aside the fact that the polarizability per electron is actually larger in C 6 H 6 , 70 even this simple argument fails to generalize to monomers with more than one ring: the isotropic polarizability ᾱ of naphthalene is (slightly) larger than that of perhydronaphthalene. 70 Furthermore, XSAPT+MBD calculations afford a dispersion energy of E disp = −5.8 kcal/mol for (C 6 H 12 ) 2 , which is less attractive than the value E disp = −6.7 kcal/mol that is obtained for (C 6 H 6 ) 2 . Clearly, polarizability is not the whole story when it comes to dispersion. Normalizing to the number of electrons (n elec ), so as to obtain a size-intensive property ᾱ/n elec , isotropic polarizabilities per electron in benzene and cyclohexane are within 5% of one another, yet the dispersion energy in (C 6 H 6 ) 2 is 16% larger than that in (C 6 H 12 ) 2 . This means that the dispersion per electron, The size-extensive nature of dispersion is familiar to any chemist in the guise of melting and boiling points for the n-alkanes that increase as a function of molecular weight, and this extensivity means that it is imperative to analyze dispersion on a per-electron basis when assessing trends versus molecular size. Only then can one make a valid comparison that might reveal whether π-π interactions constitute a unique form of dispersion. Before doing so, let us define several relevant energy components. As in previous work, 29 we group together the SAPT electrostatic and induction energies, E elst+ind = E elst + E ind . This "elst+ind" energy represents the sum of permanent and induced electrostatics. We also define the vdW energy to be the sum of the SAPT exchange and MBD dispersion energies, This is the part of the interaction potential that drives offset πstacking. 29 The total interaction energy is To make a valid side-by-side comparison of energy components in homologous systems of increasing size, however, we must normalize by the number of particles. As we did for dispersion in eq. 3, we therefore we define a normalized (per-electron) vdW energy, and also a normalized elst+ind energy, In eq. 7, we normalize by the total number of charged particles, because E elst+ind contains contributions from both nuclei and electrons. These normalized energy components are plotted in Fig. 5 for both acene and perhydroacene dimers. As the size of the system increases, E vdW converges rapidly to a constant in all three cases considered: cofacial PAHs, perpendicular PAHs, and stacked PSHs. For acenes larger than naphthalene, it is perhaps surprising to observe that the value of E vdW is the same in both parallel and perpendicular orientations, even though the dispersion per electron ( E disp , Fig. 5b) is significantly larger in the parallel orientation. This is a result of significant cancellation between the dispersion and exchange-repulsion energies, as has been noted in other work on π-stacking, where this observation is sometimes used to conclude that the geometry must be controlled by electrostatics. 16,75 However, our work suggests that it is often E vdW , not E elst or E elst+ind , that dictates the geometry. 29 Because the attractive dispersion and repulsive exchange energies are the largest energy components for closecontact π-π interactions, and because the forces on the nuclei must be zero at the equilibrium geometry, it is essentially a requirement that dispersion and exchange repulsion cancel to a significant extent at the equilibrium geometry, meaning that their sum (E vdW ) is small. As such, the fact that E vdW is small for equilibrium geometries should not be misconstrued to mean that vdW forces do not play an important role in dictating geometries. That assessment can properly be made only by examining potential energy surfaces, not simply by performing energy decomposition analysis at stationary points. Whereas the per-electron vdW interactions effectively contribute a constant to the normalized (per-particle) interaction energy, the elst+ind energy makes a significant contribution in cofacial acenes that is absent in the perpendicular orientation, and also absent in the stacked perhydroacenes (Fig. 5a). In the latter two cases, E elst+ind converges rapidly to a limiting value as a function of molecular size, and in fact for the perpendicular acene dimers the value of E elst+ind has reached its converged value already in the case of benzene dimer. For the cofacial acene dimers, however, E elst+ind continues to grow as a function of molecular size and may not yet have reached its converged value even for (pentacene) 2 . Note that E elst+ind is attractive for the cofacial PAHs even though the classical quadrupole-quadrupole energy would be repulsive in this configuration. Apparently, this leading-order multipolar contribution is offset by charge-penetration effects arising from the close proximity of the two monomers at the vdW contact distance of the supramolecular complex. The quadrupolar electrostatic picture, and with it the Hunter-Sanders model, is therefore qualitatively wrong for these systems, as we documented previously for benzene dimer. 29 Charge penetration decays exponentially with distance, in proportion to density overlap, which is smallest in the stacked PSHs due to their larger intermolecular separation (≈ 4.6 , independent of monomer size). The average intermolecular separation in the cofacial acenes is ≈ 3.5 , making charge penetration much more significant. This is underscored by the normalized elst+ind energy, E elst+ind , which changes by 67% between the cofacial benzene and naphthalene dimers. The corresponding change in the saturated systems is only 31% between cyclohexane and perhydronaphthalene. A significant orientational effect is observed as well in the case of the acene dimers. The perpendicular configuration exhibits far less density overlap, and this manifests as a mere 3% change in E elst+ind between T-shaped (benzene) 2 and perpendicular (naphthalene) 2 . Smaller charge-penetration effects in the perpendicular orientation explain why E elst+ind is essentially the same regardless of the length of the acene nanoribbon. This strong orientational dependence is imposed by the exponential dependence of charge penetration on density overlap, and dramatically alters the elst+ind interaction as the system moves from perpendicular to cofacial geometries. Dispersion may be the dominant intermolecular force in π-stacking, and its competition with exchange repulsion explains the emergence of offset-stacking, but the contributions of electrostatics and induction to the stability of π-π interactions cannot be ignored in larger aromatic systems. (We have argued that electrostatics can be ignored in benzene dimer, 29 which is one reason why this system is not representative of π-π interactions more generally.) Augmentation of E elst+ind in cofacial PAH dimers is a unique stabilization effect brought about by the interpenetration of π-electron densities, consistent with the notion that π-stacking constitutes a unique form of intermolecular interaction. ## Role of HOMO/LUMO Gaps An alternative hypothesis to explain the increase in E elst+ind , as a function of monomer size, for the cofacial acene dimers is that it results from a narrowing of the gap between highest occupied and lowest unoccupied molecular orbitals. HOMO/LUMO gap for both the acenes and their perhydro analogues are plotted as a function of size in Fig. S3. While the gap decreases monotonically with size in both cases, it does so much more rapidly for the aromatic molecules. The calculated HOMO/LUMO gaps extrapolate to 0.6 eV (acene) and 8.8 eV (perhydroacene) for infinitely-long nanoribbons. The former value is consistent with a measured band gap of 0.2 eV PAH nanoribbons as thin as 15 nm, 76 and while these computed values cannot be equated directly with the band gaps of graphene and graphane, these extrapolations are at least suggestive of the difference between these materials. Experimentally measured band gaps are zero for graphene and 4 eV for graphane. 77,78 Induction can be understood as occupied → virtual excitations engendered by the perturbing influence of the electrostatic potential from a neighboring molecule, and such excitations become more accessible as the HOMO/LUMO gap decreases. Therefore one might ask whether the growth in E elst+ind as a function of size (Fig. 5a) results from a gap-induced increase in E ind . We address this hypothesis by separating E elst+ind = E elst + E ind and examining these components separately, in Fig. 5c. For the cofacial acene dimers, the per-particle electrostatic energy E elst is significantly larger than the per-particle induction energy E ind , and also grows faster as a function of molecular size. This suggests that charge penetration effects, integrated over an increasingly long molecule and rendering the electrostatic energy increasingly attractive, are more important than the gap-induced increase in the induction energy. Size-dependent changes in E elst+ind therefore have less to do with band gaps and more to do with interpenetration of π-electron clouds. While dispersion is exceptionally strong in cofacial PAHs, its influence is exhausted in the determination of the geometry of the system. The two monomers approach closely enough to balance dispersion with Pauli repulsion, and not and the elst+ind interactions exist under the constraints of a vdW-driven geometry. For complexes consisting of flat, rigid, two-dimensional molecules, these constraints can be satisfied while retaining large density overlap. This observation bolsters the case that it is charge penetration, not dispersion, that provides the exceptional attraction in π-π systems. This should not, however, be misconstrued to mean that dispersion is less important than electrostatics in πstacking. Without exceptionally strong dispersion, the vdW force would reach equilibrium at larger intermolecular separation, reducing charge penetration and making electrostatic interactions less favorable, even tending towards repulsive in the cofacial ar-rangement when the intermolecular separation is large. Instead, the π-stacking phenomenon should be understood as a dramatic increase in the electrostatic interaction that is facilitated by the unique vdW force and is only possible in complexes composed of rigid, two-dimensional molecules. The importance of planar geometries in facilitating strong dispersion is discussed in more detail in the next section. Like induction, dispersion also relies on occupied → virtual excitations, and we have noted above that the the per-electron dispersion interactions increase nonlinearly with monomer size. Even though dispersion is largely cancelled by exchange repulsion, E disp is slightly more attractive than E exch is repulsive, for all three sets of systems considered; see Fig. 5b. The change in E disp with monomer size is most pronounced for the cofacial acene dimers and is not due to any reduction in the intermolecular separation, which is 3.4 in both the benzene and pentacene dimers. HOMO/LUMO gaps, on the other hand, are 11.3 eV for benzene and 5.4 eV for pentacene. We conclude that the change in E disp is attributable primarily to the significant reduction in the gap, rather than any change in the intermolecular separation. These intermolecular separations are consistent with the interlayer separation of 3.35 in graphene, 79 suggesting that the intermolecular distance between PAHs converges rapidly with monomer size. Such strong similarity between the intermolecular separation in a system as small as (benzene) 2 with the interlay spacing in graphitic carbon provides further evidence that the dominant effect of a parallel offset is to mitigate exchange repulsion. If electrostatics were the driving force for offset-stacking, then the intermolecular separation in (C 6 H 6 ) 2 would likely be very different from that in graphene. ## Role of Collective Density Oscillations Electron energy-loss spectroscopy and atomic force microscopy reveal that the π electrons in PAHs behave like plasmons. 80,81 In graphene, these surface plasmons obey the typical dispersion relation for an ideal two-dimensional electron gas. 82 The twodimensional collective motion of the plasmons in graphene is captured in the quantum harmonic oscillator model that is used in the MBD approach, 62 where it manifests as in-plane displacements of the oscillators. 46,47 These plasmon modes are the lowest-energy dispersive modes in π-stacked systems, and can be related to the HOMO → LUMO transition of PAHs in the molecular orbital picture. The lowest π → π * transitions in PAHs are in-plane excitations that lead to the diffusion of charge across the plane of the molecule, and the delocalized nature of the π electrons leads to low-energy HOMO → LUMO transitions at energies that vary inversely with the size of the PAH. Conversely, in the stacked perhydroacene dimers, the nodal structure of the σ orbitals prevents such delocalization, and the corresponding σ → σ * transition is out-of-plane and much higher in energy. Due to its threedimensional shape, electrodynamic screening in graphane differs from that in graphene, causing the dispersion of plasmon waves through quasi-two-dimensional materials to be slowed. 83 model. Comparison of these modes suggests that dispersion in benzene dimer is facilitated by in-plane oscillations of the electrons, whereas in cyclohexane dimer the fluctuations are much more disordered. Disordered implies less in-plane charge mobility, consistent with dispersion interactions in (C 6 H 12 ) 2 that arise from coupled out-of-plane σ → σ * excitations. ## Figures 6a and 6b These are qualitative comparisons based on just the lowestenergy eigenmode of the MBD Hamiltonian for each system, whereas in total there are 3×N atoms separate modes in the spectrum, each of which contributes to the dispersion energy. In order to generalize and quantify the analysis above, we introduce a normalized planarity index (NPI) to assess the maximum planarity of each eigenmode. The NPI quantifies how much the atomic oscillator displacements for a given eigenmode deviate from the intermolecular planes that are suggested in Fig. 6c, with limiting values NPI = 1 if all of the displacements are parallel to the prescribed plane and NPI = 0 if they are all perpendicular to it. (Mathematical details are provided in the Supplementary Information.) The individual NPIs for each of the 3N atoms eigenmodes of the MBD Hamiltonian are plotted in Fig. 6d, for both (C 6 H 6 ) 2 and (C 6 H 12 ) 2 , and these values demonstrate that on average the excitations in (C 6 H 6 ) 2 are 27% more planar than those in (C 6 H 12 ) 2 , supporting the notion of greater in-plane charge displacement for for dispersion interactions in benzene dimer. The distribution of NPIs for both systems is plotted in histogram form in Fig. 6e, from which we observe that the distribution of values is unimodal and centered around the mean in the case of (C 6 H 12 ) 2 but bimodal for (C 6 H 6 ) 2 . In the latter case, the distribution favors in-plane fluctuations, characterized by larger values of the NPI, although with a moderate preference for values NPI ≈ 0 and fewer data points at intermediate values. This hints that the dispersion-induced charge mobility in acenes is largely comprised of strongly in-plane and out-of-plane shifts, with little intermediate motion unlike the charge fluctuations that characterize (C 6 H 12 ) 2 . The absence of these intermediate values of the NPI in the case of (C 6 H 6 ) 2 is indicative of collective oscillation of charge, as modes that lie at the either end of the NPI distribution require oscillator displacements that are largely coplanar. In contrast, the charge displacements in cyclohexane dimer vary strongly from atom to atom. Thus, this simple metric therefore allows for an assessment of collective charge fluctuations induced by dispersion, considering all eigenmodes on an equal footing. Dispersion-induced charge fluctuations in (C 6 H 6 ) 2 have significantly more in-plane character as compared to those in (C 6 H 12 ) 2 . the former are much more collective as well, implying that the charge distribution about each atom changes in the same way, in contrast to the disordered atomic-density perturbations in (C 6 H 12 ) 2 . Lastly, the per-electron dispersion energies for the PAHs ( E disp , Fig. 5b) are also suggestive of collective excitations. A dispersion interaction requires creating an excitation, which creates a dipole moment even no permanent dipole moment is present and gives rise to the induced-dipole picture of London dispersion. As such, larger values of E disp reflect enhanced probability of collective excitation, even allowing for normalization for the size of the π system. The nonlinear increase in E disp versus system size that is observed for the cofacial PAHs (Fig. 5b) can be understood to result from collective excitations that generate the aforementioned molecular plasmons. Even small graphene flakes (i.e., acenes) thus appear to exhibit plasmon-like couplings in their dispersion interaction, whereas in the saturated hydrocarbons the planarity of the plasmon modes is disrupted. This result suggests that the dispersion in two-dimensional systems is unique, and changes as a function of the molecular geometry, adding additional evidence to support π-stacking as a unique form of noncovalent interaction. ## Reduced Density Isosurfaces In order to study the influence of dimensionality on intermolecular interactions, we next examine so-called "noncovalent interaction plots", 84,85 i.e., isosurfaces of the reduced density gradient The function s(r) encodes information about intermolecular interactions because noncovalent interactions are characterized by regions where the density ρ(r) is small (i.e., away from the nuclei and the covalent bonds) yet rapidly varying, as a result of an antisymmetry requirement imposed by the existence of the molecule's noncovalent partner. Isosurfaces of s(r) are plotted as in Fig. 7 for the cofacial and perpendicular acene dimers and for the stacked PSH dimers. For the cofacial acenes these isosurface plots reveal an incredibly flat landscape, and for the other two systems these isosurfaces bear a strong resemblance to plots of a vdW molecular surface. This is no accident, and results from the fact that the interactions are dominated by short-range exchange repulsion, without significant modulation by either electrostatics or induction. Note that the density ρ(r) that is used to obtain the plots in Fig. 7 does not include a self-consistent treatment of dispersion, so it is possible that these plots miss subtle changes in the density that are induced by dispersion. These self-consistent effects are found to be significant at metallic surfaces and interfaces, 86 but in the present cases the NCI plots are dominated by short-range vdW effects. For that reason, the NCI plots in Fig. 7 closely resemble molecular surface plots, i.e., they resemble the contours of molecular shape. For these systems, the vdW interactions are maximally repulsive in the regions that correspond to the oscillations in the reduced density gradient. For the cofacial PAHs, small ripples appear in the reduced density isosurface directly over the ring centers, indicating that the perfectly cofacial arrangement (with no offset) is less favorable as compared to a slip-stacked geometry. In contrast, a completely flat s(r) isosurface would imply that the noncovalent interactions were such that the monomers have complete flexibility in their relative orientation, and indeed the isosurfaces for the cofacial PAHs are relatively flat as compared to those for either the perpendicular acene dimers or for the stacked PSH dimers. This reflects the fact that the cofacial acene monomers have the flexibility to adopt parallel-offset geometries that are sterically inaccessible to the PSH dimers, which are instead conformationally locked into place, as can be inferred from the highly corrugated s(r) isosurfaces for the latter species. Parallel-offset geometries in cofacial PAHs minimize exchange repulsion, allowing for a slight decrease in intermolecular separation, e.g., 3.8 for the cofacial benzene dimer saddle point (Fig. 1a) versus 3.4 for the paralleloffset minimum (Fig. 1b). 29,32 This maximizes stabilization from charge penetration and dispersion. 29 In contrast, stacked PSHs exhibit a single low-energy conformation characterized by interlocking C-H moieties on opposite monomers. This severely limits geometric flexibility along the parallel sliding coordinate but also along the intermolecular coordinate, thus preventing the exploration of any closer-contact or slip-stacked geometries, which do not exist for these systems. Small corrugations can also be seen directly over C-C bonds in the perpendicular acene dimers, implying that exchange repulsion dominates the vdW interaction when the hydrogen atoms of one monomer are directly above the C-C bond density of the other monomer. These conclusions are consistent with the sawtooth potential energy surface of perpendicular anthracene dimer that we reported previously, using a vdW model potential. 29 In the saturated hydrocarbons, the three-dimensional nature of the atomic framework results in geometric constraints that are more pronounced and that limit the capacity for intermolecular attraction, whereas the two-dimensional aromatic molecules can sidestep this steric hindrance by adopting a parallel-offset in the cofacial arrangement. In this sense, the geometry of the molecule (driven by aromaticity or lack thereof), along with the relative orientation of the π-electron densities, conspires with dispersion to afford a unique type of stacking interaction for the cofacial acene dimers that is not available to their perhydroacene analogues. This line of argument suggests that it is the planarity of the PAHs, and not necessarily their aromaticity per se, that facilitates stacking interactions. This is consistent with other work suggesting that aromaticity is not a prerequisite for π-stacking, which can instead be driven other factors leading to a reduction in exchange repulsion. 2 Of course, aromatic molecules tend to be planar and rigid, which accounts for the close association between aromaticity and π-stacking. Planar molecules are better able to circumvent geometric constraints imposed by vdW interactions. ## Energy Landscapes for Stacked Polycyclic Hydrocarbons Isosurface plots of s(r) in Fig. 7 afford a qualitative picture of the energy landscape along the cofacial sliding coordinate in these molecules. To obtain a more quantitative picture, we have computed the two-dimensional potential energy surface for cofacial sliding of (naphthalene) 2 and (perhydronaphthalene) 2 ; see cofacial acene perpendicular acene stacked perhydroacene Fig. 7 Isosurfaces of the reduced density gradient s(r) defined in eq. 8. These isosurfaces indicate regions of space where the electron density is small but rapidly varying, which is the signature of a noncovalent interaction. as the other potential energy surface in Fig. 8. Surfaces on the far left in Fig. 8 correspond to the left side of eq. 9. Taken by itself, the E elst + E ind + E disp potential surface for (naphthalene) 2 exhibits a preference for perfect cofacial stacking with no offset. (The E elst + E ind potential surface, which is shown in Fig. S4, has a saddle point at the cofacial geometry but this disappears when dispersion is added.) This is perfectly consistent with the vdW model of π-π interactions: 29 absent exchange repulsion, the interaction potential at fixed intermolecular separation is featureless and there is no driving force towards a parallel-offset geometry. Interestingly, the E elst + E ind + E disp surface of (perhydronaphthalene) 2 exhibits three local minima corresponding to various parallel-offset structures. Each of these minima corresponds to a geometry that places hydrogen atoms from one monomer directly atop hydrogen atoms from the other. Geometries with overlapping out-of-plane atoms significantly amplify electrostatic charge penetration effects, but are also strongly prohibited by exchange repulsion. A similar phenomenon can be seen in the potential energy surface of perpendicular benzene dimer, where the L-shaped isomer (i.e., the parallel-offset version of the T-shaped isomer) is a minimum on the E elst + E ind potential surface. 29 This nuanced structure is absent in the E elst + E ind + E disp surface of naphthalene dimer, as a result of enhanced dispersion and charge-penetration, both brought about by shorter intermolecular separation. In contrast to the Hunter-Sanders model, the sum of electrostatics (including induction) and dispersion predicts a qualitatively wrong minimum-energy geometry for this system! Exchange repulsion must be included to obtain the correct geometric structure, both for (naphthalene) 2 , but also for its perhydro analogue. The exchange potentials in Fig. 8 highlight the importance of steric repulsion on the intermolecular geometry, as the most repulsive regions of E exch are precisely the regions where E elst + E ind + E disp is most favorable. In (naphthalene) 2 , Pauli repulsion shifts the geometry in a manner that corresponds to slip-stacking, whereas in (perhydronaphthalene) 2 , E exch shifts the geometry from a parallel-offset one to a structure with interlocking C-H moieties directed towards the centers of the rings on the other monomer. In this way, Pauli repulsion can be viewed as the sculptor of intermolecular orientation, especially in the shortrange regime where classical electrostatic arguments are invalid. A complementary point of view comes in noting that the E elst + E ind + E disp contribution to the interaction energy of naphthalene dimer is 14% less attractive at the actual minimum-energy (slipstacked) geometry of the complex that it is at the perfectly cofacial geometry that the system would adopt in the absence of Pauli repulsion. For perhydronaphthalene dimer, the corresponding reduction is 21%. In other words, simply accounting for changes in geometry induced by exchange repulsion reduces the attractive components of the potential by these amounts, even before the repulsion energy itself is added into the mix. We find it significant that this geometric effect is less significant for the aromatic dimer. The relatively featureless nature of the E elst + E ind + E disp surface for naphthalene dimer means that the geometric displacement that is forced upon the system by the introduction of E exch has a smaller impact on the attractive components of the interaction. The unsaturated system is more sensitive to the displacements produced by addition of Pauli repulsion. Furthermore, the featureless nature of the E elst + E ind + E disp potential for (naphthalene) 2 enhances the attractive interactions due to the uniformity of the charge penetration across the potential surface. For the PSH systems, the monomers adopt threedimensional shapes because the spatial variation of ρ(r), and thus the electrostatic interactions, is more complicated, and the monomers use their flexibility to conform to the contours of the repulsive interactions. In this sense, molecules that "serve up their attractive interactions on a platter" (i.e., a rigid twodimensional shape driven by aromaticity, that can be rotated in space but not deformed) are more likely to engage in especially strong attractive interactions because the attractive components of the interaction potential are less perturbed by the influence of exchange on the geometry of the system. ## Influence of Monomer Distortion: Corannulene Dimer In an effort to more directly correlate the flat geometries of PAH monomers with their tendency to adopt parallel-offset π stacks, we have investigated the interaction energies of corannulene dimer, (C 20 H 10 ) 2 , along a "flexing" coordinate corresponding to curvature of the monomers. Corannulene monomer is naturally bowl-shaped, and its dimer adopts a geometry consisting of concentric (or stacked) bowls with no offset. We optimized the geometry of the dimer under dihedral constraints, fixing the curvature of each monomer in the constrained system in increments, starting from the unconstrained bowl-shaped equilibrium structure and ending with completely planar monomers, corresponding to cofacial π-stacking. All of the optimized structures were initially in a cofacial, stacked arrangement, and to prevent optimization to saddle points we manually nudged one molecule in each structure to a small offset and re-optimized with constraints. All structures whose curvature was constrained at < 80% of the equilibrium value optimized to parallel-offset geometries whose offset increased as the curvature was was reduced toward pla-nar monomers. The optimized structures are depicted at the top of Fig. 9 where the "flex coordinate" indicates the degree of curvature, with 0% corresponding to planar monomers and 100% corresponding to the fully-relaxed geometry of (corannulene) 2 . Figure 9 also reports interaction energies along this flexing coordinate, which are then further decomposed into an elst+ind component (permanent electrostatics + induction) and a vdW component (dispersion + Pauli repulsion), according to eq. 5. These data reveal that the intermolecular attraction is actually most favorable (E int = −18.9 kcal/mol) in the coplanar geometry, whereas the equilibrium bowl-shaped structure of the complex has a slightly less attractive interaction energy (E int = −17.7 kcal/mol). The resolution to this apparent paradox is that the monomer deformation energy, which is not considered in the analysis shown in Fig. 9, is larger in the coplanar geometry. Contrary to a previous assertion, 87 the large dipole moment of bowl-shaped corannulene (measured experimentally at 2.07 D 88 ) does not appear to have a dominant effect on the behavior of the electrostatic interaction along the flexing coordinate. While this may seem surprising, it again speaks to the breakdown of the classical multipole picture at length scales representative of vdW close-contact distances. If the dipole moments of the corannulene monomers were the dominant effect, then the bowl-shaped equilibrium structure would have the largest interaction energy or at least the largest elst+ind energy component. In fact, E elst+ind is Elst+Ind Elst+Ind vdW Fig. 9 Interaction energies for (corannulene) 2 along a "flexing" coordinate corresponding to curvature of the monomers. Illustrative geometries are shown, optimized at fixed curvature, with 100% flex corresponding to the fully-relaxed geometry of the dimer and 0% flex corresponding to enforced planarity of the monomers. The total length of each bar (red + blue) represents the total interaction energy, which is decomposed as more attractive (by 0.5 kcal/mol) in the coplanar, parallel-offset structure than it is in the fully-relaxed equilibrium geometry. The coplanar structure has quadrupole-quadrupole interactions but the monomer dipole moments are zero (by symmetry) in this configuration. As such, the enhanced elst+ind energy in the coplanar geometry signifies charge penetration effects leading to a breakdown of the classical dipolar picture along the flexing coordinate. The dipole moment of corannulene in the equilibrium structure of the dimer is likely a consequence of the curvature of the monomers, rather than a driving force for adopting a curved geometry. There is a crossing point where sufficiently flat molecules will adopt a parallel offset, and at this point the balance of forces favors the formation of an offset. After this point there is also a monotonic increase in charge penetration as a function of flatness, as reflected by the additional electrostatic attraction. In this way, the formation of parallel offsets is a key feature of πstacking, rather than some defect as the Hunter-Sanders picture would have it. ## Conclusions We have shown that π-stacking interactions in cofacial PAH dimers, the finite-size analogues of graphene layers, are stronger than the interactions in the corresponding polycyclic saturated hydrocarbons, which are analogues of graphane. The question is sometimes asked, 1,7 "does π-stacking constitute a unique form of dispersion?". Our answer is unequivocally "yes". That said, energetic stabilization due to dispersion is largely canceled by exchange repulsion in the determination of the geometry of the πstacked complexes, which we believe should be a general feature of these systems. The exceptional strength of π-stacking interactions is better attributed to a special form of electrostatic at-traction, caused by charge penetration and thus not captured by classical multipole moments, and which is furthermore unique to molecules with flat geometries. In PAHs, the planar geometry of the molecule acts in concert with the electrostatic interaction to enhance the attraction in a manner that is not available to polycyclic alkanes. The geometric flexibility of the latter causes them to hew closely to the contours of the vdW molecular surface that are established by the Pauli repulsion interaction, leading to a strong preference for structures with interlocking C-H moieties. The PAHs, in contrast, are characterized by π-electron densities that are effectively "served up on a pizza peel" that can be rotated but not distorted, and where closer intermolecular approach is possible, leading to significant enhancement of the electrostatic interaction. The lateral offsets ("slip-stacking"), by means of which the PAH dimers reduce Pauli repulsion, are unavailable to polycyclic molecules with three-dimensional geometries. The role of charge penetration is especially important to acknowledge, and arguments based on classical multipoles badly misrepresent the interactions in π-electron systems. According to the widely-used Hunter-Sanders paradigm, 10,11,15 quadrupolar repulsion in cofacial π-stacked geometries competes with London dispersion, with the slip-stacked motif emerging as a compromise structure. At intermolecular distances characteristic of π-stacking interactions, however, the classical multipole description of electrostatics breaks down, and in fact there is no electrostatic driving force for offset-stacking. 29 This is true even in the corannulene dimer, which adopts the structure of concentric bowls whose curvature endows the monomers with sizable dipole moments of 2.07 D each. For benzene dimer, corannulene dimer, and numerous systems in between, we find that it is Pauli repulsion rather than electrostatics (or even electrostatics plus induction) that is responsible for offset-stacking. This explains, in particular, the frequent occurrence of offset-stacked geometries between nearby aromatic residues in protein structures, across what must certainly be myriad electrostatic environments. Whatever may be happening with local electrostatics, Pauli repulsion is ever-present. As we observed previously in smaller aromatic dimers, 29 the πstacking interaction can be understood as a competition between dispersion (a fundamentally quantum-mechanical type of interaction, originating in electron correlation effects) and Pauli repulsion (also quantum-mechanical in origin, as a result of the exclusion principle). This, combined with the failure of any classical multipole description to rationalize either the geometric preferences of these systems or their strong electrostatic attraction, suggests that π-stacking is unique and intimately quantummechanical. Moreover, the parallel-offsets adopted by supramolecular PAH architectures should not be viewed as defects or perturbations away from the π-stacked picture, but rather intrinsic to that picture. Interpenetration of the π-electron densities, driven by dispersion, is key to making electrostatics attractive rather than repulsive in the cofacial orientations of these systems, but this comes at a price of increased Pauli repulsion. Offset-stacking mitigates that repulsion. This is facilitated by the planar geometries of PAHs, which also support collective excitations (plasmons) that are reflected in the nonlinear growth of the dispersion interaction in PAHs as a function of molecular size, even when normalized according to the number of electrons. Theory and experiment both suggest strong interactions in π systems that ought to be considered unique in their own right, as interactions that are "served up" on flat molecular architectures. ## Conflicts of interest J.M.H. serves on the board of directors of Q-Chem, Inc.

chemsum

{"title": "Reinterpreting \u03c0 \u03c0 \u03c0-Stacking", "journal": "ChemRxiv"}

automated_fitting_of_transition_state_force_fields_for_biomolecular_simulations

## Abstract: The generation potential energy functions (PEF) that are orders of magnitude faster to compute but as accurate as the underlying training data from high-level electronic structure methods is one of the most promising applications of machine learning (ML) in chemistry. In contrast to such studies in materials and small molecules, which parameterizes the entire system without constraints on the functional form of the PEF, the simulation of biomolecular systems requires that the PEF is compatible with one of the extensively validated biomolecular force fields. Here, we describe the application of the quantum guided molecular mechanics (Q2MM) method to transition states of enzymatic reactions to generate a transition state force field (TSFF) with the functional form of the well-established AMBER force field. The differences to fitting small molecule TSFFs and the similarities of the approach to transfer learning are discussed. Finally, the application of the to the transition state of the second hydride transfer in HMGCoA Reductase from Pseudomonas mevalonii is demonstrated. ## Introduction Understanding how enzymes achieve their catalytic function is one of the grand challenges of chemistry and biology. Studying enzymes using computational methods has produced highly impactful work, as highlighted by the award of the Nobel Prize in 2014 1 for the development of multiscale methods such as the Quantum Mechanics/Molecular Mechanics (QM/MM) method. 2 Because enzymes consist of tens of thousands of atoms, using even low level electronic structure methods is cost prohibitive for the full system. Furthermore, extensive sampling of the conformational space, e.g. by molecular dynamics simulation at the microsecond time scale for the enzyme, possible ligands, and the surrounding water molecules, is necessary to obtain physically meaningful results. To enable such simulations, a range of classical force fields that approximate atoms and bonds as masses connected by springs have been developed. 3,4 The accuracy of these simulations is dependent on the quality of the force field used. 5 As a result, extensive validation studies of the force field functional form as well as the parameters themselves have been performed. The use of machine learning (ML) methods in science and technology has expanded exponentially in recent years, in part due to the rapid expansion in computational power and available datasets. In chemistry, applications of ML range from basic research through material research to drug discovery. 6 More pertinent to the topic of the present study, ML has been applied to force field and PEF parameterization given its strengths in pattern recognition. There are numerous examples in materials chemistry, where the accurate description of large systems to predict material properties demanded a cheap method at high accuracy. 11 Another well-recognized example is the ANI-1 potentias 12 that use active learning and neural network algorithms to take high-level QM data to create transferable ML potentials. 13 Even though the development of ML methods for the treatment of enzymatic reactions provides an alternative to the computationally expensive QM/MM methods, there have been comparatively few ML applications for force field development reactions and/or biomolecular systems. One reason is that in most cases, the new potential energy surfaces created break away from the restrictions of a predefined functional form. This is less likely to be successful for the case of the study of reactions in biomolecular systems, as exemplified by Pseudomonas mevalonii 3-hydroxy-3-methylglutaryl-CoA Reductase (PmHMGR) shown in Figure 1. Here, the vast majority of the system (shown in grey) is well described by extensively validated classical force fields. However, these cannot describe the substrate, cofactor, or residues involved in the transition state the reaction (show in color). The large dataset needed for training of an ML PEF for the reactive center is not available from experimental data and cannot be generated from high-level electronic structure calculations due to their high computational cost. Here, we propose an alternative approach that is reminiscent of transfer learning where the functional form and extensively validated parameters of a classical force field (in the present case, AMBER) are used and retrained for a subset of the structure that includes the bond breaking and making atoms as well as key active site residues and cofactors (shown in color in Figure 1) using the quantum-guided molecular mechanics (Q2MM) method that was originally developed for the parameterization of small molecule force fields, especially TSFFs. 14,15 As mentioned before, most of the work done on the use of ML for all-atom force fields has been focused on small molecules or solvents using functional forms determined by e.g. a neural network. 16, There are a few examples of the use of ML for fitting predefined functional forms using both linear and non-linear regression algorithms in the literature to reproduce training data from appropriate electronic structure methods. ML in the form of a genetic algorithm was used to optimize a polarizable force field from ab initio QM data 18 as well as the parameterization of reactive force fields. 19 The Parsely force field for small molecules uses QM data for parameterization of an AMBER-lineage with SMIRNOFF atom specification. 20 Similarly, the AMBER-15 Force Balance force field 21 for use with the TIP3P-Force Balance water model 22 is fitted using a weighted least-squares method. The AMOEBA-2013 force field also was optimized using automated techniques to obtain a general polarizable protein force field. However, these studies concern ground state (GS) force fields that are not able to describe bond breaking and making steps of an enzymatic reaction where a TSFF is needed. One of the best established 24 automated fitting procedures for the parameterization of both ground state and transition state force fields (TSFF) is the Quantum Guided Molecule Mechanics (Q2MM) approach that has been used extensively for the development of TSFF for the prediction of stereoselectivity of small molecule reactions. To the best of our knowledge, the only application of Q2MM to biomolecular systems is a TSFF for transition-state docking of small molecular drugs to P450 enzymes to identify potential sites of hydroxylation. 29,30 However, the code used for this fitting procedure is to the best of our knowledge not widely available. Q2MM uses training data from electronic structure (usually density functional theory) reference calculations to automatically parameterize molecular mechanics TSFF based on the MM3* PEF. The details of this process for asymmetric catalysis by small molecules have been covered elsewhere 14, 15, and will not be elaborated on here. Here, we will describe the application of the Q2MM method to derive TSFFs of a predefined functional form compatible with the AMBER-family force fields with particular attention to the differences to the fitting of small molecule TSFFs. We will also discuss the interfacing of the Q2MM tools to the AMBER suite of molecular dynamics programs and demonstrate this workflow for the case of a TSFF for the second hydride transfer of PmHMGR. ## Fitting Methods Q2MM fits the FF parameters by minimizing the value of objective or loss function, where x i 0 is the reference data point, x i is the FF data point, and w i is the weight for that data point. The minimization step in the parameter space is calculated using gradient-based method such as the Newton-Raphson technique and simplex method. 32 The gradient-based method is general and utilizes the Jacobian matrix J where and p j is j-th parameter, which is calculated in many programs using numerical differentiation and therefore the rate-determining step. Thus, the simplex method is often used to avoid the high cost of numerical derivatives. 33 The simplex method in Q2MM is modified to move toward the best point(s) in the parameter space using a bias of reflection point. 32 The modified simplex method has shown to have faster convergence than the Raphson-type methods up to ca. 40 parameters. 31 Thus, it is used to optimize a medium-sized parameter set or a subset of the larger parameter set. Q2MM, unlike most traditional methods for fitting system-specific FF parameters, 22,34,35 uses the Hessian Matrix for the fitting of force constants of bonded parameters with geometric data for reference structures. 32,36,37 The Hessian matrix is the second partial derivative of the energy with respect to the xyz coordinates of atoms, which gives the matrix size of 3N x 3N where N is the number of atoms. It can be obtained by appropriate electronic structure calculations of suitable model systems including, in the case of the large biomolecular systems discussed here, QM/MM calculations. In the later case, the calculation of the Hessian Matrix usually needs to be limited to a subsystem due to the memory demands of such calculations. The Hessian matrix's eigenvalues and eigenvectors provides information on the vibrational frequencies and normal modes, respectively. Normally, eigenvalues of the Hessian matrix are positive, but at the transition state geometry, the eigenvalues contain one significantly negative value with its eigenvector representing the reaction vector. By providing Hessian matrix information in the objective function, Q2MM uses information on both the transition state geometry and the shape of the potential energy surface around it when fitting the FF parameters. However, because Q2MM fits these parameters to represent the transition state, which is a saddle point, as a minimum on the potential energy surface, the matrix element that corresponds to the negative eigenvalue is, inevitably, altered during the fitting process. This leads to an increase of the objective function value. To address this and the fact that the algorithms in most molecular force field-based programs 38, 39 38 are designed to optimize towards minima rather than transition states, a small modification to Q2MM is made. Traditionally, in the Cartesian Hessian fitting method, all indices of the Hessian matrix are accounted for in the objective function with respect to the reference values. However, different weights are assigned to each element of the Hessian matrix to correctly represent the transition state as a minimum. The indices of Hessian matrix are given a weight of 0.0 to 1-1 interactions, 0.031 to 1-2 and 1-3 interactions, 0.31 to 1-4 interactions and 0.031 to all other interactions. 40 The Cartesian Hessian matrix fitting method is used for large molecule systems such as an enzyme, where only one reference structure is used to fit the parameters. Alternatively, users can use the eigenmode fitting method in Q2MM. In this method, 41 the reference Hessian matrix H=V T EV is decomposed into eigenvector V and eigenvalue E . Then the objective function includes the calculated eigenvalue matrix E ' where E ' =VH ' V T and H ' is the Hessian matrix of the FF calculated Hessian matrix. By preserving the original eigenvector V , all of the originally positive eigenvalues are preserved and only the negative eigenvalue is converted into a positive value by zeroing the weight of the eigenvalue to represent a transition state as a minimum. This method has yielded an FF that is stable to unnatural distortions and is used for smallmolecule systems such as metal-ligand-substrates, where multiple reference data are used to fit the parameters. It should be noted that this this inversion of the potential energy surface in the reaction coordinate is done to allow the use of simple energy minimization techniques available in all force field packages to locate the stationary point. However, it is not absolutely required and alternative approaches have been developed. 42 The Q2MM Flow Scheme The following parameterization scheme is specific towards the implementation of the AMBER20 39 interface of Q2MM and its use for large large biomolecular systems. Details of the method regarding parameterization of TSFF for asymmetric catalysis using other programs such as Macromodel have been documented elsewhere. 15 As an example of using Q2MM for a large biomolecular system, the development of a TSFF for the second hydride transfer transition state of PmHMGR, shown in Figure 1, will be discussed. Examples of the files discussed in this section as well as the final TSFF are given in the Supporting Information. The Q2MM code itself, which contains the interface to the AMBER Suite of programs, and several published TSFFs are freely available on the Q2MM/CatVS github repository (github.com/q2mm). In order to develop a TSFF for an enzyme, the first step is to define a model system that includes the reactive species and the relevant parts of the protein involved in catalysis to generate the training data for the TS of this model system. For the example discussed here, the QM/MM or theozyme 46 model incorporated the relevant residues in the QM region derived from our previous studies 43,44 of the mechanism of HMGR and shown in Figure 2, though other model systems were also explored. 43 Since this model system is derived from electronic structure calculations, only the most essential atoms should be included for efficiency of the fitting procedure even though the methodology is equally applicable to larger numbers of refitted atoms. A fixedatoms.txt file is created to include any atoms frozen in the electronic structure calculation (Figure 2, green atoms). Because the frozen atoms create unphysical Hessian elements, the weight of the Hessian values associated with these atoms are set to zero during the parameterization. Results of transition state optimizations, in a .log file, contain the energetic and geometric data that are used by Q2MM in the parametrization and are thereby included in the Q2MM input as reference. Currently, Q2MM supports interfaces to Gaussian 47 and Jaguar 38 .log files as training data for the parameterization process. The .log file is also used to create a .mol2 file of the model system using the RESP protocol in AMBER. The .mol2 file contains updated partial charges of all the atoms in the model system at the TS and is for used throughout the parametrization. Figure 2. Flow scheme of the Q2MM method for the parameterization of TSFFs for enzymatic reactions using the AMBER interface At this point, new atom types should be assigned to the atoms directly involved in the reaction, as their properties will be sufficiently different from that of the parent force field. This allows for the parameters defined by the TSFF to be restricted to a specific atom in the entire system. The atoms to be reparametrized in the case discussed here are shown in Fig. 3A. It should be noted that this procedure is analogous to transfer learning in that parameters trained to a much larger dataset (standard parameters of the Amber force field) and extensively validated in the literature are used as a starting point for retraining a much smaller subset for which smaller training data sets are available. It is a key difference from the development of TSFFs for transition metal catalyzed reactions 14,15,25,26 where there are usually no parameters available for the transition metal environment. As a result, a much larger training set is needed in those cases to achieve a reliable TSFF. Even though the number of atoms to be retrained is usually larger for the case of enzyme catalyzed reactions, the use of a transfer learning approach makes the fitting procedure much more effective because the vast majority of atoms only undergoes minor perturbations in proceeding from the ground state to the transition state of the reaction. The .mol2 file should also be used to generate the force field modification (frcmod) file, using antechamber program of AMBER. 39 The .frcmod file needs to be updated accordingly to be used in Q2MM, examples of which can be found in the documentation on github. All parameters such as bonds, angles, and dihedrals for atoms directly involved in the reaction should be included in the .frcmod file. Transition state parameters are different from the ground state ones, so initial guesses of the bond lengths and angles should be for the system at the TS as described by the QM reference data. The estimation of the parameters prevents optimization to local nonphysical minima of the objective function and decreases the number of iterations required for parameterization. Force constants are initially set to standard values based on the Generalized Amber Force Field (GAFF), 48 and initial estimations for dihedrals should in our experience be avoided to prevent over-parameterization. The parameter.py module of Q2MM generates a list of a specified parameter type to be optimized that references the .frcmod file line and includes the range of values acceptable for that parameter type. The input file, loop.in, for Q2MM files should contain all of the relevant information for an optimization cycle. The FFLD being read every cycle should be the given AMBER .frcmod file and the RDAT being read should be the Gaussian or Jaguar .log file. For CDAT, a tleap input file that calls the mol2 file and frcmod file of the model system and relevant Amber force fields should be created to generate a prmtop and inpcrd file that is used during the parameterization process. The optimization criteria of the penalty function are set in the loop.in file under the LOOP flag. Initially the penalty function can be set to a 10% convergence criterion. The loop.in file can be submitted by >python loop.py loop.in . Partial charges should remain unchanged throughout the course of the parameterization. The order of parameterization (Figure 2) is largely the same as discussed earlier. 15 The force constants should be optimized first while ensuring that the optimized value stays above 32.2 kcal mol -1 -2 for bond distances and angles and 3.2 kcal mol -1 -2 for dihedrals. Subsequently, the bond length parameter can be refined to reflect the reference data. Bond angles can be optimized after the bond lengths while ensuring that the optimized values are within reasonable ranges. If the optimized angles deviate towards unreasonable values, then this angle parameters value should be "tethered" to the reference data to prevent major deviations during optimization. The tether is defined as a weight value associated that would thereby control the deviation of the parameter being optimized. A higher tether weight should be used in the first round of optimization, then slowly decreased to zero in subsequent optimizations cycles. Finally, the Vn terms for the torsional potentials are fit to the Hessian data first before being further refined. A second round of optimizations should be performed with a 1% convergence criterion for the penalty function to allow parameter refinement to be closer to the reference data parameters. Additional optimization cycles can be performed as needed until a working transition state force field has been obtained. For enzymatic systems, a working TSFF is obtained when an optimization step changes the objective function by less than a 1% and the values and parameters are deemed realistic by the given user. Additionally, the resulting force field should be tested in a large-scale molecular dynamics simulations in conjunction with the ground state force field to describe the remainder of the protein (shown in gray in Figure 1). The TSFF will have to be parsed to generate new residue types that contain reparametrized atoms and new library files will need to be created to read into the leap module of AMBER20. This could also involve setting conditions that allow the reacting atoms to have more than the standard amount of bonds in a system. Other important considerations are adjusting the time step of the simulation to account for the vibrations of the reacting atoms and potentially removing the SHAKE algorithm for hydrogens in the TSFF. A short MD simulation should then be performed to ensure that the total energy of the system remains stable with the TSFF in combination with the ground state FF that would be used for the rest of the enzyme. ## Application to PmHMGR This method described above was employed for the second hydride transfer TS of PmHMGR. Here, the reference data for the training of the TSFF were obtained from QM/MM calculations where the atoms indicated in Figure 2 were treated at the ONIOM-(B3LYP/6-31G(d,p):AMBER) level of theory. 43,44 This includes the side chains of H381, K267, D283 and E83 as well as the substrates and cofactor as shown in Figure 3 and the hmgrqm.log example file in the Supporting Information. As the functional form of the underlying force file to which to fit the TSFF to, AMBER99SB and GAFF for atoms on residues and substrates were used, respectively, as seen in the ts2.frcmod file. During parameterization, the full size of the substrates and cofactor, along with the backbone and sidechains of the residues mentioned above, were included while calculating the MM data (Figure 3B). As discussed earlier, the bonding character and partial charges of the atoms directly involved in the TS change in going from the ground to the transition state. Furthermore, the nicotinamide ring of the cofactor is dearomatized. To describe these perturbations, new atom types were introduced as indicated in Figure 3A. It is worth reemphasizing that the initial parameters for these new atom types were derived from the standard ground state AMBER99SB parameters and then trained to reproduce the electronic structure results in the training data. In this specific case, only parameters directly associated with these atoms (within 3 bonds) were reparametrized for the TSFF. As shown in Figure 3 B, the TSFF successfully reproduced the geometries around the reacting center of the active site and could successfully be incorporated into the rest of the enzyme that is treated with a traditional ground state force field. Using this, the enzyme could be simulated at the transition state on the microsecond timescale. The results of these studies will be discussed elsewhere. ## Conclusions In this contribution, we have discussed an automated workflow that combines the Q2MM method with transfer learning-type approaches for the generation of fast and accurate TSFFs for large biomolecular systems. Application of the workflow to the second hydride transfer of HMGR, an enzyme of high biomedical importance, shows that the transition state of this reaction can be accurately reproduced by the TSFF derived by this workflow. The use of machine learning to generate potential energy functions that are orders of magnitude faster to compute than their training data, which often are derived from accurate but slow electronic structure calculations, is a promising application of ML in chemistry. The work presented here uses the philosophy of transfer learning and applies it to the parametrization of TSFF by retraining of well validated existing force fields as oppose to creating completely new atom types and parameters, as is done in the generation of small molecule TSFF that cover transition metal catalyzed reactions. The results are an early example for using only electronic structure reference data and a much larger number of parameters adjusted in the biomolecular TSFF than in the earlier cases of small molecule TSFFs. They show that idea derived from ML can be used to parameterize a TSFF to simulate enzymes at the transition state ~10 4 times faster than the underlying electronic structure methods, allowing for molecular dynamics simulation for system sizes and timescales well beyond the accessibility of DFT-based methods. 13

chemsum

{"title": "AUTOMATED FITTING OF TRANSITION STATE FORCE FIELDS FOR BIOMOLECULAR SIMULATIONS", "journal": "ChemRxiv"}

efficient_solid-state_photoswitching_of_methoxyazobenzene_in_a_metal–organic_framework_for_thermal_e

## Abstract: Efficient photoswitching in the solid-state remains rare, yet is highly desirable for the design of functional solid materials. In particular, for molecular solar thermal energy storage materials high conversion to the metastable isomer is crucial to achieve high energy density. Herein, we report that 4methoxyazobenzene (MOAB) can be occluded into the pores of a metal-organic framework Zn 2 (BDC) 2 (DABCO), where BDC ¼ 1,4-benzenedicarboxylate and DABCO ¼ 1,4-diazabicyclo[2.2.2] octane. The occluded MOAB guest molecules show near-quantitative E / Z photoisomerization under irradiation with 365 nm light. The energy stored within the metastable Z-MOAB molecules can be retrieved as heat during thermally-driven relaxation to the ground-state E-isomer. The energy density of the composite is 101 J g À1 and the half-life of the Z-isomer is 6 days when stored in the dark at ambient temperature. ## Introduction Molecular photoswitches are currently receiving signifcant interest for molecular solar thermal (MOST) energy storage applications. MOST materials convert photon energy to thermal energy through reversible isomerization between ground and metastable isomeric states. While many classes of photoswitch have been investigated for MOST applications, azobenzene (AB) derivatives are among the most widely studied 4 due to their high quantum yield, high-fatigue resistance and appreciable energy separation between the ground-state E and metastable Z isomers. However, in their pure solid form, photoswitching of AB derivatives is often limited due to dense crystal packing. Several strategies have been proposed to address this problem by increasing the molecular free volume; these include templating AB on nanotubes and graphene, 8,9 incorporating as sidegroups within polymers, and using bulky functional groups to form amorphous flms. 10 Recently, confnement within metal-organic frameworks (MOFs) has been shown as an effective way to impart conformational freedom to photoswitches within bulk solid materials. Using this approach, solid-state photoswitching has been demonstrated for AB derivatives as well as other molecules including dithienylethenes, 2-phenylazopyridine, 28 and spiropyrans. In addition to spatial considerations, another requirement for solid-state MOST materials is to ensure a high degree of photoconversion to the metastable state. For AB, overlap of the p-p* absorption bands for the E and Z isomers means that the photostationary state (PSS) is limited to approximately 78% Z isomer under 365 nm irradiation. 36 The effects of MOF confnement on the achievable PSS are not well understood but recent studies have shown it can be detrimental depending on the structural properties of the MOF. 20,37 One way to increase the intrinsic PSS is to alter the electronic structure through chemical modifcation. This has been demonstrated for ortho-functionalised AB derivatives, for which the E and Z isomers have well-separated n-p* bands. 38 However, while this can lead to more efficient photoswitching, including quantitative switching within MOFs, 39 it can also reduce the energy difference between the ground and metastable states, thereby reducing the energy density of the resulting MOST material. 1,37,39,40 The design of both the photoswitch and MOF architecture therefore need to be carefully considered in order to achieve a high degree of conversion as well as a high energy density. Herein, we report the structural and photothermal properties of a composite comprising the breathable MOF Zn 2 (-BDC) 2 (DABCO) (1) and 4-methoxyazobenzene (MOAB) which shows promising properties for solid-state MOST applications. This composite (1IMOAB), has the advantages that 1 is straightforwardly prepared by solvothermal synthesis, and MOAB is widely commercially available. Unlike bulk MOAB which photoconverts to 70% Z-isomer, MOAB exhibits a photostationary state of nearly 100% when confned within the composite, resulting in an energy density of 101 J g 1 . The half-life of the Z-isomer within the composite is 6 days when stored in the dark at ambient temperature. ## Results and discussion MOAB was occluded within the pores of 1 by a previously described melt-infltration procedure. 13,37 The maximum loading level obtained was 1.25 MOAB molecules per pore (Fig. S1, S2 and Tables S1-S3 †). DSC measurements confrm the absence of residual MOAB outside the pores (Fig. S3 †). Surprisingly, this loading level is comparable to the 1.3 molecules per pore observed for AB within 1 (1IAB), despite the increased length of the guest molecule. 37 1 is known to undergo guest-induced breathing, 41 and X-ray powder diffraction (XRPD) confrmed that the framework within 1IMOAB is contracted compared to the guest-free form (Fig. 1a). Profle ftting shows a single phase corresponding to the narrow pore (np) orthorhombic (Cmmm) structure (Table 1, Fig. S4 and S5 †). This symmetry contrasts with the tetragonal phase previously reported 13 for 1IAB when one molecule of AB is occluded per pore, though it is less contracted, presumably due to the high density of occluded guest molecules. This is notable because the cell parameters of this orthorhombic phase are highly sensitive to changes in temperature (Fig. 1b), which indicates flexibility in the low temperature phase of the framework. Additionally, variable-temperature (VT-) XRPD shows that 1IMOAB undergoes a np to large pore (lp) phase transition at 160 C, where the lp structure is equivalent to the guest-free phase (Fig. 1b). Table 1 summarizes the lattice parameters for the different phases of 1IMOAB studied. VT-XRPD experiments below 140 C show a reversible temperature-dependent expansion and contraction of the a and b cell lengths, respectively (Fig. 1b, Tables S4 and S5 †); however, the c cell length (defned by the DABCO pillar group) remains consistent across the temperature range (Fig. S6-S13 †). For an alkoxy-functionalised analogue of 1, Henke et al. reported expansion of the a and b cell lengths with increasing temperature below the np / lp phase transition, with a concomitant increase in the unit cell volume. 42 However, for 1IMOAB we observe a small volume contraction between 80 and 140 C, which is driven by the contraction of the b axis. This highlights that the arrangement of guest-molecules can signifcantly affect the flexibility of 1. In the 13 C cross-polarisation (CP) magic-angle spinning (MAS) NMR spectrum of 1IMOAB (Fig. 2a and S14-S16 †), single DABCO and carbonyl resonances at 47.6 and 171.2 ppm are consistent with those previously observed for the np orthorhombic structure of 1IAB. 13 Considering the MOAB guest molecules, the methoxy and C-N carbons each show three distinct resonances, suggesting three crystallographically inequivalent MOAB molecular conformations within the pores. Comparison with DFT chemical shift calculations (Tables S6 and S7 †) suggests fast-rotational dynamics around the molecular axis, as was previously observed for 1IAB. 13 A spectrum recorded at 34 C showed a pronounced broadening of the MOAB ring resonances consistent with a reduction in the timescale of this motion (Fig. S17 and S18 †). The dynamics of the guest molecules indicates that despite the dense packing, a signifcant degree of conformational freedom is retained. Thermal analysis of 1IMOAB between 0-200 C (Fig. 2b, Fig. S19, Tables S8 and S9 The solution-state UV-vis spectrum of E-MOAB (Fig. 3a) shows characteristic absorptions at 348 nm (p-p*) and 440 nm (n-p*). The p-p* absorption is signifcantly red-shifted relative to E-AB (320 nm); this is attributed to the electron donating effect of the methoxy group in MOAB. Irradiation with 365 nm light causes E / Z isomerization, resulting in a decrease and shift of the p-p* absorption to 305 nm and producing a photostationary state containing 98% Z-MOAB, as measured by 1 H NMR. The near-quantitative photoswitching can be attributed to the signifcant redshift of the p-p* absorption in E-MOAB which effectively separates the absorptions for the E and Z isomers. Irradiation of 1IMOAB at 365 nm causes E / Z isomerization within the framework and a 1 H NMR shows that a photostationary state (PSS) of 98% is reached after 480 minutes (Fig. 3b). Crystalline MOAB photomelts under 365 nm light; 43 however, 1IMOAB remained solid and thermal measurements show MOAB is not lost from the pores. The PSS of the composite is equivalent to the solution-state value (Fig. S20-S25 †) and higher than the $70% PSS obtained when photo-melting pure MOAB (Fig. S22 †). It is noteworthy that the high PSS of MOAB is maintained when confned within 1, whereas other guest molecules show a signifcantly reduced PSS (AB -40%, PAP -28%). 13,22,28 However, recent work has shown that fluorinated ABs can also be quantitatively switched inside MOFs. 21 This highlights that further investigation is required to understand the precise factors controlling the PSS which may include the density and arrangement of the guest molecules as well as guest-induced breathing of the framework. Thermal analysis of the irradiated composite between 0-200 C reveals a large exotherm upon the frst heating branch (Fig. 3c), which is attributed to thermally-driven reconversion of the Z-MOAB molecules to the ground-state E isomer. The exotherm magnitude agrees well with the calculated the Z / E reconversion enthalpy based on the PSS and a DFT-calculated E-Z energy difference of 68.7 kJ mol 1 (Fig. S26-S28 and Tables S10-S13 †). 43 On the cooling branch, another exothermic feature is observed corresponding to the lp / np phase transition (Fig. S29 †). This feature is expected due to contraction of the lp framework around the E-MOAB formed due to Z / E reconversion that has taken place on the heating branch. Over one full heating and cooling cycle, the exotherms for Z / E reconversion and the lp / np phase transition give a combined energy density of 86.4 kJ mol 1 or 101 J g 1 . The cyclability of the composite was investigated using a reduced irradiation time of 300 minutes which results in a conversion to 83% Z-MOAB, with a corresponding energy density of 86.4 J g 1 (Fig. 4a). This energy density was maintained over fve full cycles of irradiation and thermally-driven discharge with no degradation of the composite or loss of MOAB from the pores of 1 (Fig. 4b). The gravimetric energy density of 101 J g 1 for 1IMOAB represents an increase by a factor greater than 3 in comparison to the previously reported energy density of 28.9 J g 1 for 1IAB. 13 This is due to a combination of the larger energy E-Z difference (68.7 kJ mol 1 for MOAB vs. $50 kJ mol 1 for AB) and the almost quantitative conversion to the Z-isomer of MOAB within the composite. The energy density of Z-MOAB is also comparable to other azobenzene-based solid-state MOST materials including functionalised polymers with reported energy densities between 90 and 176 J g 1 , 10,11 and an amorphous flm formed from a bulky azobenzene derivative which showed an energy density of 135 J g 1 . 12 Complementing work on azopolymers and molecular AB derivatives, surfacetemplated azobenzene derivatives have been demonstrated to store up to a remarkable 540 J g 1 , and in some cases also report half-lives of nearly 2 months. 9 However, reports of energy densities around 300 J g 1 are more typical, with associated half-lives ranging from hours to days. 44 One of the advantageous properties of 1IMOAB is nearquantitative isomerisation is achieved in the bulk material without the requirement to suspend in solution or cast into flms or surface-based architectures. While near quantitative conversion within the bulk of a MOF has been previously reported for an ortho-fluoroazobenzene derivative confned within 1; however, DFT calculations (Table S14 †) show that the functionalisation of the azobenzene moiety in this case is predicted to result in a signifcant decrease in the E-Z energy difference and therefore a marked reduction in energy density as compared to 1IMOAB. To rationalize the high PSS within the composite, XRPD measurements were performed to monitor structural changes during UV irradiation. With increasing irradiation time, the reflections of the np orthorhombic phase shift and decrease in intensity and new reflections emerge (Fig. S30 †). After 240 minutes the new phase dominates the pattern. Profle ftting shows this phase is consistent with the lp tetragonal (P4/mmm) structure (Table 1 and Fig. S31 †). This is consistent with a previously reported phase change from orthorhombic np to tetragonal lp for irradiated of 1IAB. 22 For 1IMOAB, the expansion of the framework by irradiation (3.5%) is lower than that of the temperature-induced np / lp phase transition (4.4%). We note that a minor component of the orthorhombic np structure is retained even after the PSS is reached. This suggests that a small proportion of the occluded MOAB molecules isomerize within pores that remain contracted. However, 1IMOAB shows greater overall flexibility than 1IAB, which is highlighted by the temperature-dependent distortion of the framework below the phase transition as well as the larger expansion upon irradiation. The differences in the guestinduced flexibility are presumably related to the ordering of the guest molecules and/or host-guest interactions, which may be key to achieving efficient photoconversion to the Z isomer. A 13 C CPMAS NMR spectrum of irradiated 1IMOAB (Fig. 3d, S32 and S33 †) shows no changes in the chemical shifts of the framework resonances, although the carbonyl resonance sharpens. Considering the MOAB resonances, there is considerable shifting for each carbon site. The most noticeable of these is the methoxy carbon which reduces to a single resonance at 55.1 ppm. Comparison of the experimental chemical shifts with DFT-calculated values (Tables S15 and S16 †) suggests that Z-MOAB molecules also undergo rapid rotational motion within the pores. At ambient temperature in the dark, the occluded Z-MOAB molecules thermally reconvert to the E isomer with a half-life of approximately 6 days (Fig. 5a and Table S17 †). The best ft to the thermal reconversion data was obtained when the process was modelled as following third-order kinetics (Fig. 5a and S34 †). This implies a complex cooperative mechanism where motion or rearrangement of multiple Z-MOAB molecules is required during the thermal reconversion. This is further supported by DSC thermograms for samples with low Z-MOAB populations (Fig. 4b and c), which show complex multicomponent features. The data suggest that there are at least two separate processes characterized by a residual exotherm of 9.3 J g 1 (Fig. 5b) and a small composite feature reminiscent of 1IAB which remains when the Z-MOAB population is further reduced (Fig. 5c). The 6-day half-life of Z-MOAB within the composite is longer than azobenzene-based polymer MOST materials which have been reported in the range 12-75 hours. 45,46 However, it is signifcantly shorter than that of Z-AB when occluded within 1 (4.5 years). 13 While the reason for this marked difference requires further investigation, it is likely that the flexible nature of the orthorhombic framework for 1IMOAB allows greater freedom of the guest to revert to the E-form. This is further supported by the observation by DSC that the thermal reversion begins before the np / lp phase transition in 1IMOAB, whereas no thermal reversion is observed in 1IAB until the onset of the np / lp phase transition. ## Conclusions High-efficiency photoswitching in the solid-state remains highly desirable, and the 1IMOAB system demonstrates a PSS of >98% for the Z-isomer. The composite constitutes a solidstate MOST system that can store 101 J g 1 of thermal energy. It also displays a useful half-life of around 6 days at ambient temperature. Reducing the mass of the host MOF or structural modifcation of the photoswitch could further increase the gravimetric energy density or the half-life of the Z-isomer. Both these approaches are currently being targeted and we are also examining the feasibility and efficiency of light-triggered, as well as thermally-triggered, energy release in 1IMOAB and other systems.

chemsum

{"title": "Efficient solid-state photoswitching of methoxyazobenzene in a metal\u2013organic framework for thermal energy storage", "journal": "Royal Society of Chemistry (RSC)"}

oxidation_of_a_wood_extractive_betulin_to_biologically_active_oxo-derivatives_using_supported_gold_c

## Abstract: Betulin (90-94%) was extracted from birch with a non-polar solvent and recrystallized from 2-propanol.Liquid-phase oxidation of betulin aimed at obtaining its biologically active oxo-derivatives (betulone, betulonic and betulinic aldehydes), exhibiting e.g. antitumor, anti-inflammatory, antiparasitic, anticancer and anti-HIV properties, was demonstrated for the first time over gold-based catalysts. Gold was deposited on pristine TiO 2 and the same support modified with ceria and lanthana, followed by pretreatment with a H 2 or O 2 atmosphere. The catalysts were characterized by XRD, BET, ICP, TEM, XPS, DRIFT CO, TPD of NH 3 and CO 2 methods. The nature of the support, type of modification and the pretreatment atmosphere through the metal-support interactions significantly influenced the average particle size of gold, its distribution and the electronic state of gold, as well as the acid-base properties and, thereby, the catalytic performance (activity and selectivity) in betulin oxidation. Au/La 2 O 3 /TiO 2 pretreated in H 2 displayed the highest catalytic activity in betulin oxidation among the studied catalysts with selectivities to betulone, betulonic and betulinic aldehydes of 42, 32 and 27%, respectively, at 69% conversion. Side reactions resulting in oligomerization/polymerization products occurred on the catalyst surface with the participation of strong acid sites, diminishing the yield of the desired compounds. The latter was improved by adding hydrotalcite with the basic properties to the reaction mixture containing the catalyst. Kinetic modelling through numerical data fitting was performed to quantify the impact of such side reactions and determine the values of rate constants. † Electronic supplementary information (ESI) available. See ## Introduction Utilisation of natural compounds for chemical transformations to obtain biologically active compounds has become one of the promising and actively developing areas of fine organic synthesis and pharmaceutical chemistry. Triterpenoids are a class of compounds that combine accessibility, i.e., availability in nature and easiness of isolation, with valuable biological activity. Betulin (lup-20 (29)-ene-3, 28-diol, C 30 H 50 O 2 , CAS: 473-98-3)a pentacyclic triterpenoid of the lupane series, is found in almost two dozen plants belonging to different genera and families, wherein the main source of betulin is birch bark with the content varying from 10 to 35%. The methods of betulin extraction from birch bark are widely reported in the literature including extraction of the bark outer layer using various solvents, bark alkaline hydrolysis followed by ethanol extraction of betulin, "explosive" autohydrolysis, etc. Betulin and especially its oxo-derivatives (betulone, betulinic and betulonic aldehydes, and betulinic and betulonic acids) have valuable biologically active properties, and are of exceptional interest for the pharmaceutical, cosmetic and food industries. For example, betulinic acid and its derivatives exhibit anti-cancer, anti-HIV, antiviral, anti-inflammatory, anti-septic, antimicrobial, anti-malarial, anti-leishmaniasis, anthelmintic and fungicidal activities, while betulonic acid shows pronounced anti-inflammatory, antimelanoma and antiviral effects. Antiviral and anti-leukemic activities have also been reported for betulonic aldehyde, which is, moreover, active against diseases of the liver and the digestive tract and disorders of reproductive function. The 3-oxo derivative of betulin betulone and its derivatives, exhibiting antitumor, antiinflammatory, aniparasitic, and anti-HIV properties, also demonstrate in vitro cytotoxic activity against different cancer cell lines. Recent studies indicate a clear demand for betulone as a building block for creating effective anticancer agents with minimal side effects. Currently, the main method of synthesis of betulin oxoderivatives is its oxidation. The work of Csuk et al. 23 described the formation of betulinic acid via betulinic aldehyde by oxidation of betulin with a mixture comprising TEMPO (2,2,6,6tetramethylpiperidine-1-oxyl)-NaClO 2 -NaOCl at 35 °C with 92% yield. Synthesis of betulinic acid from betulin can be carried out in one stage using butyl acetate with 4-acetamido-TEMPO and Bu 4 NBr•H 2 O as oxidants in an aqueous solution of NaClO 2 and NaOCl at 50 °C, regulating the pH with phosphate buffer. Methods of obtaining betulinic acid and betulinic aldehyde by betulin oxidation with chromium oxide(VI), chlorochromate or pyridinium dichromate fixed on a solid silica gel or alumina have also been reported. Betulinic acid was synthesized by the oxidation of betulinic aldehyde with potassium permanganate. 24 A method for selective oxidation of betulin with pyridinium dichromate (PDC), pyridinium chlorochromate (PCC) or K 2 Cr 2 O 7 -9 M H 2 SO 4 in the presence of tetrabutylammonium bromide (TBAB) to betulinic aldehyde, or its mixture with betulonic aldehyde and ketol was developed by Komissarov et al. 25 The yield of oxidation products did not exceed 75%. The method of obtaining betulonic acid by oxidation of betulin, in the first step with the Jones reagent in acetone, should also be mentioned. 26,27 Alternatives include application of the pyridine dichromate complex and acetic anhydride in dimethylformamide, 28 and chromium(VI) oxide in acetic acid, followed by reduction to betulinic acid. 29 Twostage methods for the synthesis of betulinic acid have significant shortcomings. The low solubility of betulin in acetic acid, acetone and methylene chloride and of betulonic acid and its salts in alcohols, tetrahydrofuran and water, imposes several limitations, preventing oxidation and reduction and, thus, resulting in poor yields and purity. In the majority of the proposed oxidation methods, highly toxic Cr(VI) is used. Moreover, separation of the products containing toxic Cr(III) ions is very laborious and time consuming. In addition to chemical modifications, attempts to transform betulin using microorganisms were carried out, mainly for the synthesis of betulone. 10, However, such betulin biotransformation processes using conditionally pathogenic yeasts, fungi, etc. have significant drawbacks, requiring complex nutrient media, long duration, and low concentration levels of products as the biocatalyst might not tolerate higher concentrations. Recently, 34,35 some of the authors of this work demonstrated for the first time the possibility to selectively oxidize betulin to betulinic aldehyde using Ru/C as a catalyst mixed with the basic hydrotalcite and SiO 2 as a dehydrating agent at 108 °C in toluene, with air as an oxidant. Under these conditions, the conversion of betulin after 24 hours was 41% with 67% selectivity to betulinic aldehyde, whereas without the addition of SiO 2 , the conversion and selectivity were 20% and 66%, respectively. A higher conversion was achieved when the reaction was carried out in an acidic medium, giving, however, allobetulin as the main product. It was found that the presence of a basic agent and elimination of water are crucial for selective oxidation of betulin to betulinic aldehyde on Ru catalysts. Selective oxidation of betulin to betulone has also been reported using silver supported on unmodified titania and modified with ceria under mild conditions, e.g. atmospheric pressure, relatively low temperature (140 °C), synthetic air as an oxidant, and mesitylene as the solvent. Conversion of betulin over a Ag/CeO 2 /TiO 2 catalyst reached 27% after 6 hours, which was substantially larger than 11% obtained for Ag/TiO 2 . In all cases the main product was betulone with the selectivity exceeding 60%. Based on the analysis of published studies on betulin oxidation, it is obvious that currently there are no economically and ecologically acceptable methods for producing oxo-betulin derivatives. Such methods based on heterogeneous catalysis should replace the existing stoichiometric processes which lead to the formation of large amounts of toxic waste, being able to provide preferably a quantitative yield of the desired product. In the present work, the possibility of heterogeneous catalytic oxidation of betulin with synthetic air, using catalysts based on gold nanoparticles supported on unmodified and modified titania, will be demonstrated for the first time. Some of these catalysts have been synthesized before and used for oxidation of octanol 37 which has, however, chemical and physical properties different from those for betulin. No reports on betulin oxidation over gold catalysts are available in the literature; thus it was interesting to explore the possibility of their utilization for a much more complex case than oxidation of octanol. The aim of the present study is thus to evaluate the applicability of gold-based catalysts in the liquid phase selective oxidation of betulin, to elucidate the influence of the support and nature of the additives, and the impact of redox pretreatment on catalytic properties. Moreover, a comparative analysis of the catalytic properties of ruthenium, silver and gold catalysts in betulin oxidation was performed showing that the latter ones are more active and stable. ## Results and discussion XRD was used to study the phase composition of the investigated catalysts (ESI Fig. S1 †). The XRD patterns showed the absence of the reflections characteristic of gold and modifiers, indicating small sizes of gold and metal oxide particles (lower than the sensitivity XRD threshold of 3-4 nm) or their X-ray amorphous structure. Table 1 shows the specific surface area of supports and catalysts (S BET ), the Au content and gold particle size data. The surface area of pristine TiO 2 was diminished by 13% after modification (48 m 2 g −1 ) with both modifiers. Further gold deposition did not significantly change the specific surface area of the supports, except Au/La 2 O 3 /TiO 2 , for which there was a noticeable decrease by 10% (Table 1). ICP analysis showed that Au contents were close to the nominal ones. The average size of gold nanoparticles is lower than 3 nm for most of the studied materials, except Au/TiO 2 _pO 2 (Table 1). The largest size of gold nanoparticles and the broadest distribution were observed in the case of gold supported on unmodified titania (ESI Fig. S2 †). In contrast, for La-modified materials, the size of Au particles and the range of their distribution are the smallest. These values for Ce-modified materials were in between those for other catalysts. In addition to the nature of the support, the pretreatment atmosphere (H 2 or O 2 ) also affects the uniformity of particles and their relative size. At the same time, the impact of pretreatment depends also on the support. For unmodified and lanthana-modified materials, smaller particles were obtained after pretreatment in H 2 (300 °C), and for ceria-modified materials after pretreatment in O 2 (300 °C). These effects can be attributed to the specificity of gold interactions with different supports during catalyst preparation, as previously confirmed, 36 and to the different nature of the gold precursor decomposition under reducing and oxidizing pretreatments, previously revealed by TPR. 37 It should also be taken into account that a certain fraction of gold is in the form of highly dispersed oxidized gold species, which are quite difficult to be detected by electron microscopy because of the lower contrast of oxidized species compared to that of the reduced ones. The presence of such oxidized species was previously validated by TPR. 37 Moreover, gold in the ionic state (Au + or Au 3+ ) not detected by TEM could still be present, according to DRIFT CO and XPS (Fig. 1 and Table 2). The amount of these gold species (ionic and oxidized gold species) depends on the support and pretreatment. Table 2 shows the relative atomic concentrations of various electronic states of gold, calculated according to XPS. As can be seen, the relative values of different gold states depend strongly on the support and pretreatment conditions. On the surface of all studied catalysts, most of the gold (68-89%) is in a metallic state with BE(Au 4f 7/2 ) in the range of 84.2-84.3 eV, but also a part of gold (11-20%) is in the form of singly charged ions (Au + ) with BE(Au 4f 7/2 ) in the range of 85.2-85.5 eV. In the case of unmodified and Ce-modified samples pretreated in H 2 , another state related to three-charged gold (Au 3+ ) with BE(Au 4f 7/2 ) equal to 86.5 and 86.3 eV appears in the XPS spectrum (11 and 12%, respectively). These data confirm that the formation of the active surface on different supports, under the action of various pretreatments, occurs differently, and is in good agreement with TEM (Table 1). For a more detailed study of the electronic state of gold in the investigated catalysts, and also as for the evaluation of the strength and stability of the adsorption centers, DRIFT spectroscopy of adsorbed CO was applied. CO adsorption was carried out at different pressures: 5, 20, and 50 Torr, making it possible to evaluate the strength of the centers. Pure supports did not exhibit bands of adsorbed CO in this region of the spectrum under the studied conditions. From Fig. 1, it can be concluded that for all catalysts, regardless of the pretreatment, one absorption band with the maximum in the range of 2100-2120 cm −1 , attributed to the surface carbonyl groups of gold atoms Au 0 -CO, 38 was observed. The intensity of this band increased with increasing CO pressure. Differences in the signal positions are caused by CO adsorption on the metal clusters of different sizes, and bands with a low-frequency are associated with larger nanoparticles. CO starts to adsorb on larger gold clusters as the pressure increases. Moreover, carbon monoxide is very weakly adsorbed on metallic gold because of some features of σ-π binding in M 0 -CO for Au, in comparison with other noble metals (Pt, Pd, Ru, Rh, and Cu). 39 Only highly dispersed gold clusters or atoms can be sites for CO adsorption. This explains the different intensities of the absorption bands corresponding to Au 0 -CO at a CO pressure of 50 Torr. Subsequently, it can be assumed that there are larger particles in Au/TiO 2 _pH 2 , Au/CeO 2 /TiO 2 _pH 2 , Au/CeO 2 /TiO 2 _pO 2 , and especially Au/La 2 O 3 /TiO 2 _pO 2 cata- lysts, which were not taken into account when analyzing TEM images because of their relatively low abundance. In our previous study, 40 there was a similar discrepancy between the average particle size obtained by TEM and SR-XRD. The average particle size of gold obtained from SR-XRD was in good agreement with the catalytic results, and for some catalysts it was larger than the values determined by TEM. Another absorption band with the maximum in the range of 2140-2185 cm −1 , related to the complexes of ions Au + -CO, 41,42 was observed in almost all cases, except Au/TiO 2 _pH 2 . However, the intensity of this absorption band and its change with pressure variation are different. This absorption band is less intense than that attributed to Au 0 -CO, and also strongly depends on CO pressure. The intensity increases with the pressure increase. It is interesting to note that, for Au/La 2 O 3 / TiO 2 _pO 2 (Fig. 1f ), reduction of Au + sites is observed under a CO atmosphere, which indicates their very low stability. The absence of this absorption band for Au/TiO 2 _pH 2 can be due to the presence of only weak Au + sites, and even a CO pressure of 50 Torr is not enough for their identification by DRIFT CO, while according to XPS (Table 2), Au + is 15% of the total amount of gold. It should also be noted that the XPS method determines the ionic states of gold in the near-surface layer, some of which may not be accessible for adsorbed molecules. At the same time, the method of DRIFT adsorbed CO allows the identification of the active sites on the surface available for the reactions. The TPD of NH 3 was used to determine the acidity of supports and respective gold catalysts, namely the concentration and strength of acid sites (Table 3 and ESI Fig. S †). Physical adsorption can take place in the case of ammonia TPD, being, however, typical of low temperatures. Therefore, to avoid the contribution of physical adsorption, the analysis started from 100 °C. Three types of acid sites are detected for the initial supports, but their concentration and strength are different (Table 3). The pristine titania showed the highest acidity among the used supports with the majority of acid sites being of weak strength, while the concentrations of the medium and strong acid sites are 2.6 and 6.9 fold lower than the previous one. Therewith, they are all Brønsted acid sites (acidic OH groups). However, it is possible that the strong acid sites are of aprotic nature and are Lewis acid sites (e.g. tetrahedral coordinated Ti 4+ ). 46 Modification of titania with ceria and lanthana led to a decrease in the concentration of weak and medium acid sites. This is most likely a consequence of surface dehydration after calcination at 550 °C during preparation. Alongside that, the amount of strong acid sites increased for Ce-modified titania, but decreased for La-modified titania. In the case of ceriamodified TiO 2 , this can be explained by the appearance of new Lewis sites, due to the presence of Ce 4+ /Ce 3+ , whose existence was indirectly confirmed by TPR. 37 For the lanthana-modified material, no hydrogen consumption was observed in TPR pro-files. Thus, it can be assumed that lanthana blocked the acid sites on the pristine titania surface leading to a decrease in acidity. After gold deposition, in all cases, there was a redistribution of acid sites. Regardless of the pretreatment for unmodified and Ce-modified materials, an increase in the concentration of weak acid sites and a significant decrease of strong acid sites were observed. The amount of medium sites in the case of Au/CeO 2 /TiO 2 was increased, while for Au/TiO 2 it remained almost unchanged. The distribution of acid sites is noticeably different for the La-modified catalyst, compared with the other materials. This is most clearly seen for strong acid sites, whose concentration significantly increased. Moreover, there was a decrease of weak acid sites in comparison with unmodified and Ce-modified materials. Such changes in acidity after metal deposition may originate from several reasons. One of the options can be associated with a change in the support properties during catalyst preparation resulting in the mutual influence of the support and the metal precursor, as previously discussed. 36, Another possibility is blocking the acid sites, previously existing on the surface, by newly formed metal nanoparticles. It should be noted that, in the case of the La-modified catalyst, a part of the strong acid sites (90 × 10 −4 mol m −2 ) may be associated with the Lewis acid sites, namely Au + . This is confirmed by comparing XPS (Table 2), DRIFT CO (Fig. 1) and NH 3 -TPD (Table 3) data. Considering that lanthana is a non-reducible oxide, it can be suggested that another part of the strong acid sites is associated with Brønsted acidity and belongs to the support or the modifier. In order to assess the basic properties of the studied materials, the TPD of CO 2 was used. Based on the literature, depending on the temperature range in which CO 2 desorption occurs, the basic sites are divided into three types: weak, medium and strong, reflecting their nature. The weak basic sites (25-200 °C) are usually attributed to surface hydroxyl groups, medium ones (200-400 °C) to metal oxide pairs, and strong sites (400-600 °C) to low-coordinated oxygen anions. All types of basic sites mentioned above were observed for the supports studied in this work (Table 4 and ESI Fig. S4 †). Pristine titania exhibited the average total basicity among the studied supports, with the dominance of the basic sites of medium strength and almost absent strong sites. A similar distribution of the basic sites was also observed for Cemodified titania, while the amount of these sites was lower. After modification of titania by lanthana, there was an increase in the concentration of weak and strong basic sites, while the amount of medium sites remained almost unchanged. Table 4 also presents the results for hydrotalcite, the basicity of which is 2-3 fold higher than that of the used supports. The gold deposition on the support surface led to a redistribution of the basic sites similar to the acidic ones (Table 3). For almost all studied catalysts, there was an increase in the amount of basic sites while, in all cases, the strong basic sites increased. The reasons for the changes in basicity after gold deposition are apparently the same as for acidity in a sense that they originate from the exposure of the support to the metal precursor during preparation, mutual influence of the support and the precursor, and base site blocking. In addition, as shown in ref. 54, 55, CO 2 is also capable of being adsorbed on small gold nanoparticles, with the abstraction of oxygen by Au 0 , giving CO and Au 2 + O 2− species, which cannot be considered as the basic sites. When comparing CO 2 -TPD (Table 4) and TEM (Table 1), it can be concluded that the highest increase in the amount of basic sites was observed for samples with the smallest particle sizes. Subsequently, a part of the strong basic sites can be associated with CO 2 adsorption on small gold nanoparticles. The catalytic behavior of Au supported catalysts in the betulin liquid phase oxidation (Fig. 2) was studied at 140 °C and a synthetic air pressure of 1 bar in mesitylene. To assess the influence of the nature of the support and pretreatment atmosphere, gold was deposited on pristine titania and TiO 2 modified with ceria and lanthana. The obtained materials were pretreated in H 2 or O 2 . It was found that both the nature of the support and the pretreatment atmosphere have a significant influence on the catalytic behavior of the studied catalysts (Table 5). Among the studied gold containing materials, Au/TiO 2 pretreated in H 2 (Table 5, entry 1) showed the lowest activity. Betulin (A) conversion was 25%, with selectivity to betulone (B) and betulinic aldehyde (C) of 40 and 53%, respectively, at this conversion level. Herewith, the total yield of products was only 15%, which is 1.7 fold lower than the observed conversion (Table 5 entry 1, Fig. 3a and d). This difference is due to incomplete mass balance (the sum of the masses of the reactants and products visible in GC and GCLPA). This is most likely caused by the strong adsorption of reactants or products on the catalyst surface. The mass balance closure was different for the various catalysts, being determined by the catalytic properties. In this particular case, the GCLPA was 90%. Betulin conversion and selectivity for the same material (Au/TiO 2 ), but pretreated in O 2 , were almost the same (Table 5, entry 2). However, due to a better GCLPA -96%, the total yield of products for this catalyst turned out to be 1.7 fold higher than that for the one pretreated in H 2 . In contrast to the unmodified material, betulin conversion for the Ce-modified catalyst was higher after the pretreatment in H 2 (betulin conversion -45%, Table 5, entry 3). However, the mass balance closure in this case was the worst -77%, ultimately giving only 22% total yield of the main products. For the same catalyst, but pretreated in oxygen, a lower conversion -33% was achieved (Table 5, entry 4). The best 95% GCLPA was, however, reached for this case, with the total yield of products being 1.2 fold higher than that after pretreatment in H 2 . The pretreatment atmosphere even affected the selectivity for the primary products, such as betulone (B) and betulinic aldehyde (C), which should be less sensitive to conversion levels. For Au/CeO 2 /TiO 2 _pO 2 in particular, the main product was betulinic aldehyde (C), while for Au/CeO 2 /TiO 2 _pH 2 betulone (B) was mainly obtained. Au/La 2 O 3 /TiO 2 pretreated in hydrogen showed the highest activity among the studied catalysts. Betulin conversion was 69% and the main products were betulone (B), betulonic (D) and betulinic (C) aldehydes, with selectivities of 42, 32 and 27%, respectively (Table 5 entry 5, Fig. 3b and d). It should be noted that due to a GCLPA of 80%, the total yield of the main products turned out to be 1.4 fold lower than the observed conversion, being 48%. Betulin conversion for the same material (Au/La 2 O 3 /TiO 2 ), but pretreated in O 2 , was 2.8 fold lower than that after treatment in H 2 (Table 5, entry 6). For this catalyst, the product distribution was also different. It can also be related to the conversion as betulinic aldehyde (C) can be transformed to betulonic aldehyde (D). In this case, betulonic aldehyde (D) was practically not formed and the main products were betulone (B) and betulinic aldehyde (C), with a higher GCLPA (97%). Table 5 also presents the results of the previous studies on betulin oxidation over Ru and Ag catalysts. 34,35 As can be seen from the data (Table 5, entries 4-9, 15, 17 and 18), under the same experimental conditions, the activity of supported gold catalysts significantly exceeds the activity of Ru and Ag counterparts. Selectivity of Ru is significantly different from that for Au and Ag catalysts. Over the majority of Au(Ag)/(modifier)/TiO 2 catalysts the main reaction products were betulone (B) and betulinic aldehyde (C), while betulonic acid (F) and betulinic aldehyde (C) were obtained for Ru/C. It is also worth noting that allobetulin was not observed in the reaction products for Au(Ag)/(modifier)/TiO 2 in comparison with Ru. Moreover in the previous work the mass balance closure was not explicitly accounted for, making a direct comparison of the yields difficult. In the work, 34 it was also shown that the catalytic behavior (activity and selectivity) of Ru catalysts depends strongly on the reaction conditions. In toluene as a solvent at 108 °C, betulin conversion over Ru/C (entry 14) was 54% after 5 hours with allobetulin (77%), a structural isomer of betulin, as the main product. To evaluate how the reaction conditions affect the catalytic behavior of gold materials, a similar experiment was carried out using Au/La 2 O 3 /TiO 2 _pH 2 in toluene at 108 °C (entry 13). Compared with entry 8, betulin conversion decreased 1.2 fold; however, the main reaction products were still betulone (B), betulonic (D) and betulinic (C) aldehydes. It is worth noting that in this case, the mass balance closure was higher (88%) compared with entry 8 (80%); therefore, the difference in ∑Y product between entries 8 and 13 was only 2%. Selectivity to a more desired product (betulinic aldehyde as opposed to allobetulin) was increased 34 by adding basic hydrotalcite to the reaction mixture, even if there was a negative influence on activity decreasing the betulin conversion from 54% to 16% (Table 5, entry 15). The betulin conversion reached 41% with a selectivity to betulinic aldehyde of 67% (Table 5, entry 17) with an increase in the reaction time up to 24 hours, adding hydrotalcite and silica as dehydrating agents. In the present work, a similar experiment was carried out, and hydrotalcite was added to the reaction mixture containing Au/ La 2 O 3 /TiO 2 _pH 2 (Table 5, entry 10, Fig. 3c and d). However, there were no significant changes in betulin conversion or in the product distribution. Betulin conversion increased by only 1% after adding hydrotalcite. Despite a slight increase in betulin conversion, the TOF increased 1.4 fold and was 0.010 s −1 , compared to the experiment without hydrotalcite, for which it was 0.007 s −1 . Moreover, due to the better GCLPA -85%, the total yield of the main products increased to 55% compared with entry 5, being 48% (Table 5). Addition of silica to the reaction mixture, along with Au/La 2 O 3 /TiO 2 _pH 2 and hydrotalcite, additionally increased the betulin conversion by 1% and the product yield by 3%, as well as the GCPLA to 87% (Table 5, entry 11). Thus, it can be assumed that the addition of hydrotalcite, leading to apparently local changes in concentrations of a proton and hydroxyl groups in the vicinity of the catalyst surface, can thereby affect the properties of the catalyst surface. In turn, silica prevents the inhibitory action of water. However, in the case of betulin oxidation over gold catalysts, inhibition by water is much less pronounced than that for Ru catalysts. 34 In general, gold materials were less sensitive to changes in the reaction conditions than ruthenium ones. When Au/La 2 O 3 /TiO 2 _pH 2 was recycled (Table 5, entry 12), a 30% drop in activity compared to entry 8 was observed, indicating some catalyst deactivation. However, the gold catalyst was still more stable than ruthenium, for which the activity decreased by 44% in the second run (from 41% to 23%). 34 As mentioned above in Introduction betulone is equally important as betulinic aldehyde or betulinic acid. Selectivity to betulone can be increased by adding to the reaction mixture besides the catalyst also hydrotalcite and silica or by replacing the solvent and lowering the reaction temperature. When comparing the GCLPA for the same catalyst, but pretreated under different atmospheres (e.g. Au/La 2 O 3 /TiO 2 _pH 2 and Au/La 2 O 3 /TiO 2 _pO 2 , Table 5, entries 4-8) with acid-base properties (Tables 3 and 4), it can be assumed that the GCLPA is determined by the acidity of the materials, namely the concentration of medium and strong acid sites. The lower this concentration, the higher the mass balance closure (Fig. 4), which can be explained by side reactions, promoted on stronger sites leading to a lower GCLPA. This is also confirmed by comparing the catalytic and ammonia TPD data for the supports applied in this work (Tables 3 and 5 entries 1-3). For Ce-modified TiO 2 , the concentration of medium and strong acid sites was the highest among supports, and the GCPLA was the lowest. For La-modified TiO 2 , the opposite situation was observed. From this point of view, hydrotalcite indirectly affected the catalytic properties, in particular the strong acid sites, preventing the side reactions and increasing the product yield. Along with the acidity of the materials, their basicity also plays an important role (Table 4). Betulin conversion was higher for catalysts with more pronounced basic properties. Herewith, H2-pretreated materials were more basic, but at the same time, they also demonstrated higher acidity. The exception was Au/ TiO 2 , for which conversion or the product distribution was almost independent of the pretreatment atmosphere. This can be explained based on the fact that the acid-base properties of this material vary only slightly upon different pretreatments. It should be noted that while a certain correlation between the acid-base properties and catalytic performance was seen, the role of gold in betulin oxidation is decisive. Moreover, correlations between the TPD of ammonia and CO 2 made in the gas-phase with catalytic properties should be taken with caution when the catalysts are employed in the liquid-phase processes. Nevertheless, TPD methods provide general information on the nature of solid surfaces and the types of sites and are often applied for characterizing the acid-base properties of solid catalysts even for the liquid phase reactions. In order to find out the reason for a decrease in the GCPLA, namely, what was adsorbed on the catalyst surface, size exclusion chromatography (SEC) was used. After carrying out the extraction and SEC analysis, it was found that polymers and oligomers with a molecular weight of 5000 and 1000 Da, respectively, were formed on the catalyst surface (ESI Fig. S5 †). The weight of oligomers/polymers on the catalyst was not quantified, being, however, related to the mass imbalance between the theoretical GCLPA (100%) and the corresponding values of GCLPA reported in Table 5. Thus, it can be concluded that a decrease in the GCPLA is associated with the side reactions of the oligomerization/polymerization of betulin or its derivatives on the catalyst surface, with the participation of strong acid sites. Formation of oligomers/ polymers on the catalyst surface is also likely to cause partial deactivation of the recycled catalyst (Table 5, entry 12). Despite washing the catalyst after the first run in hot acetone, a part of the oligomers/polymers could remain on the surface thereby blocking partially active sites. This is also confirmed by an increase in the GCPLA after the second run (Table 5, entry 12). In order to quantify the kinetic significance of various steps comprising the reaction network and a contribution of side reactions leading to oligomers/polymers, kinetic modelling was performed for betulin oxidation in the presence of Au/La 2 O 3 / TiO 2 _pH 2 and hydrotalcite (Fig. 3c). The reaction scheme given in Fig. 2 was somewhat modified (Fig. 5) to incorporate formation of oligomers (O) and finally polymers (P) and account for a clear lack of mass balance closure in Fig. 3c. The reaction scheme was simplified as the concentration of acids was negligible for Au/La 2 O 3 /TiO 2 _pH 2 . In general, oligomers can originate not only from the reactant as in Fig. 5 but also from the products, as mentioned above. However, a lack of mass balance closure in some cases was already seen at the beginning of experiments justifying that the main contribution for the formation of oligomers comes from the reactants. To keep a more general character of the model, formation of oligomers was considered to be reversible, while generation of polymers as terminal species was supposed to be irreversible. The equations for the reaction rates presented in Fig. 5 can be easily written: These equations correspond to the adsorption of all organic compounds and subsequent oxidation with noncompetitively adsorbed oxygen. Dependence of the oxygen concentration is thus implicitly incorporated in the rate constants k i . In the preliminary development of the kinetic model (eqn (1)), adsorption of all reactants was considered. However, the initial parameter estimation showed that the calculated terms in the denominator involving adsorption coefficients, for all substances and their concentrations apart from betulonic aldehyde, are very low. This allows assuming that the coverage of these species is rather low. The constants in eqn (1) are lumped ones comprising implicitly also dependence on oxygen pressure. The reactor mass balances for each component in the reaction system are as follows: In eqn (1)-( 6) C i denotes the concentration of respective compounds, mol L −1 , and ρ is the catalyst bulk density given in g L −1 . Modified constants, etc. contain also the respective adsorption coefficients. Differential eqn (2)-( 6) were solved using the backward difference method and the parameter estimation was performed with the simplex and Levenberg-Marquardt methods. The numerical tools are inbuilt in the optimization software ModEst, 56 in which the objective function Q is defined through experimental y i and calculated ŷ i concentrations of the components in the reacting system: The results (Fig. 6) show that this model can describe the experimental data rather well. For 24 data there were initially 8 adjustable parameters, namely 7 rate constants (k′ 1 to k′ 6 and k′ −5 ) and one adsorption constant (K D ). During the parameter estimation it turned out that some of these constants, namely k′ 3 and k′ 6 , are negligible. Thus the final model comprised 6 parameters. Their values are given in Table S1. † Even if the number of data points is much larger than the number of parameters they were somewhat correlated with each other, preventing a detailed analysis of their physicochemical significance. Apparently, a separate kinetic study accounting for catalyst deactivation and more rigorous chemical analysis of oligomers/polymers is required, being, however, outside of the scope of the current work. The degree of explanation R 2 : was 99.3% reflecting the applicability of the model. ## Conclusions The current work is the first study dealing with the liquidphase oxidation of betulin over gold-based catalysts. As a support, titania per se, or modified with CeO 2 or La 2 O 3 , was used. The nature of the support and pretreatment atmosphere (H 2 or O 2 ) significantly affected the uniformity of gold particle distribution and their mean size, the electronic state of gold and acid-base properties and, as a consequence, the catalytic behavior (activity and selectivity) of the studied materials in betulin oxidation. The smallest gold nanoparticles with their narrow distribution and the strongest and most stable adsorption sites (Au 0 and Au + ) were formed on the La-modified TiO 2 surface after H 2 pretreatment. Additionally, this material exhibited the highest basicity and the highest concentration of medium and strong acid sites among the studied catalysts, and as a consequence the best catalytic results. Betulin conversion was 69% for 6 h at 140 °C, and the main products were betulone, betulonic and betulinic aldehydes, with selectivities of 42, 32 and 27%, respectively. However, the total yield of products was 1.4 fold lower than the observed conversion, which was due to an incomplete mass balance and was caused by the side reactions of oligomerization/polymerization on the catalyst surface, promoted on stronger acid sites. The product yield was increased by adding basic hydrotalcite to the reaction medium along with the catalyst. Such results can be explained by an indirect influence of hydrotalcite on the surface properties of the catalyst, in particular strong acid sites which, in turn prevents the side reactions and increases the product yield. Kinetic modelling was performed to quantify the significance of such side reactions. ## Experimental Catalyst preparation TiO 2 P25 (nonporous, 70% anatase and 30% rutile, particle size: 21 nm, purity: 99.5%, Evonik Degussa GmbH) was used as the starting support. For comparative studies, titania was modified with ceria and lanthana by impregnation with solutions of the corresponding nitrates (molar ratio Ti/M = 40, where M = Ce or La). After impregnation, the samples were dried at room temperature for 48 h, then at 110 °C for 4 h, followed by calcination at 550 °C for 4 h. Gold catalysts (Au/TiO 2 , Au/CeO 2 /TiO 2 and Au/La 2 O 3 /TiO 2 ) were prepared by deposition-precipitation with urea, according to the procedure previously described. 36, The nominal gold content in all catalysts was 4 wt%. The gold precursor (HAuCl 4 •3H 2 O, Merck) and urea (Merck) were dissolved in distilled water, and thereafter the support was added to the solution. The resulting mixture was heated to 80 °C and kept at constant temperature for 16 h, with stirring. Thereafter, the catalysts were pretreated at 300 °C for 1 hour under a H 2 or O 2 atmosphere. The catalysts are denoted hereinafter as Au/(M x O y )/TiO 2 _P, where M x O y is CeO 2 or La 2 O 3 and P indicates the pretreatment atmosphere (O 2 or H 2 ). ## Catalyst characterization The specific surface area (S BET ) of supports and catalysts was measured by nitrogen adsorption with a "TriStar 3000" analyzer (Micromeritics, USA). Prior to measurements, the samples were subjected to thermal vacuum treatment at 300 °C for 5 hours. To calculate the S BET , a multipoint BET method with linearization of the adsorption isotherm for the relative pressure between 0.005 to 0.25 was used. The phase composition of supports and catalysts was studied by the step-scanning procedure (step size: 0.02°; 0.5 s) with a Philips XPert PRO diffractometer, using CuKα radiation (λ = 0.15406 nm) and a Ni-filter. The measured diffractograms were analyzed with the ICDD-2013 powder diffraction database. The morphology of catalysts and the size of gold particles were investigated by transmission electron microscopy (TEM) and STEM-HAADF (scanning transmission electron microscopy-High Angle Annular Dark Field) using a JEOL JEM-2100F. The samples were ground to a fine powder and sonicated in hexane at room temperature. Then a part of the suspension was placed on a lacey carbon-coated Cu grid. In order to obtain micrographs that most fully reflect the real structure of the samples, a thorough examination of the samples was carried out, after which the selected area was scanned at various resolutions. For each sample, at least 150 particles were registered. The metal loading of the catalysts was determined by inductively coupled plasma optical emission spectrometry (ICP-OES) Conditions and notation of components are given in Fig. 3. using a PerkinElmer ICP-OES Optima 3300 DV spectrometer. The solids were dissolved by acid dissolution, digested in a microwave oven, diluted to 100 mL and analyzed in the spectrometer. The catalysts were characterized by X-ray photoelectron spectroscopy (XPS) with a SPECS GmbH custom made system using a PHOIBOS 150 WAL hemispherical analyzer and a nonmonochromated X-ray source. All the data were acquired using Al Kα X-rays (1486.6 eV, 200 W). A pass-energy of 50 eV, a step size of 0.1 eV per step and a high-intensity lens mode were selected. The diameter of the analysed area was 3 mm. Charging shifts were referenced against the Ti 2p 3/2 peak of TiO 2 at 458.8 eV. The pressure in the analysis chamber was kept lower than 1 × 10 −8 mbar. The accuracy of the binding energy (BE) values was about ±0.1 eV. Peak areas were estimated by calculating the integral of each peak after subtracting a Shirley type background, fitting the experimental peak to a combination of Lorentzian/Gaussian lines with a 30/70 proportion and keeping the same width on all lines. Deconvolution of spectra was performed with the program CasaXPS. Diffuse Reflectance Fourier Transform Infrared (DRIFT) spectra of CO adsorbed on the catalysts were recorded by using a Bruker EQUINOX 55/S FTIR spectrometer with a homemade accessory at 4 cm −1 resolution at room temperature. The powdery fraction of an oxide was placed in a quartz ampoule with a window of CaF 2 . The samples were preliminarily calcined at 100 °C under vacuum not less than 10 −4 Torr for 1 h. For each catalyst three samples were investigated: as-prepared, and after pretreatments either in H 2 or in O 2 (100 Torr) at 300 °C for 1 h and then cooled down to room temperature. Then, H 2 or O 2 was evacuated and CO adsorption (>99%) was carried out. The spectra of adsorbed CO were recorded at several pressures -5, 20, 50 Torr, at room temperature, with the pressure measurement accuracy of 5%. The obtained spectra were recalculated into Kubelka-Munk units (KMU). The background spectrum was subtracted from the spectrum of the sample with adsorbed CO and the baseline was corrected. All calculations were performed using the OPUS 6.0 software (Bruker). Acidic and basic properties of the catalysts and corresponding supports were studied by the temperature-programmable desorption (TPD) of ammonia ("Chemosorb" chemical adsorption instrument) and CO 2 (Autochem 2900 apparatus), respectively. The procedures in both cases were almost the same apart from the starting desorption temperature, which was 100 °C for ammonia TPD and 25 °C in the case of CO 2 and the carrier gas, in the former case, was helium, and the latter argon. Prior to the analysis, the samples were treated at 300 °C under an inert atmosphere (helium or argon) for 1 h to remove the impurities adsorbed on the surface. Thereafter, the temperature was decreased to 100 °C (25 °C) followed by saturation with NH 3 (CO 2 ) for 60 min and flushing with He (Ar) for 1 h to remove physisorbed NH 3 (CO 2 ). The temperature was increased to 600 °C with a 10 °C min −1 ramp under a helium (argon) atmosphere. For comparative analysis, NH 3 and CO 2 desorption profiles of the supports and corresponding catalysts are demarcated into temperature ranges: 100-200 °C (for TPD CO 2 the starting temperature is 25 °C), 200-400 °C and 400-600 °C and are designated as weak, medium and strong acid or basic sites, respectively. ## Catalytic testing Betulin (90-94%) was extracted from birch with a non-polar solvent and recrystallized from 2-propanol in bo Akademi University. 2 Betulin oxidation was performed over supported Au catalysts under atmospheric pressure with synthetic air (AGA, 20% oxygen, 80% nitrogen) as an oxidant in mesitylene at 140 °C or in toluene at 108 °C (Sigma Aldrich, >99%). Synthetic air was bubbled through the liquid with an inlet for the gas (flow rate: 50 ml min −1 ) located at the bottom of the reactor to enhance the gas-liquid mass transfer. Moreover, a metallic sinter was applied to diminish the size of air bubbles. Typically oxidation of betulin was carried out using 200 mg of the reagent in 100 ml of the solvent (the initial betulin concentration was 4.5 mmol l −1 ) using 200 mg of the catalyst. The reaction started when the desired temperature was reached, via turning on the stirring (450 rpm). Small catalyst particles (<63 µm) and a high stirring rate of 450 rpm were used to suppress the internal and external mass transfer limitations. In some experiments, hydrotalcite (Merck) was used together with the catalyst as a base-additive. Hydrotalcite was calcined for 3 h at 500 °C prior to its use. The samples for analysis were withdrawn from the reactor at regular intervals. Prior to GC-analyses, the samples (150 µL) were silylated by adding 150 µL of a mixture of pyridine (VWR International, Fontenay-sous-Bois, France), N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA, Supelco Analytical, Bellefonte, PA, USA), and trimethylsilyl chloride (TMCS, Merck KGaA, Darmstadt, Germany) in a 1 : 4 : 1 volume ratio, and the mixture was heated in an oven at 70 °C for 45 min. GC analysis was performed on a PerkinElmer AutosystemXL gas chromatograph using an Agilent HP-1 capillary column, 25 m (L) × 0.2 mm (ID), film thickness: 0.11 mm. Hydrogen was used as a carrier gas, with a flow rate of 0.8 ml min −1 . Betulinic aldehyde and betulinic acid (90% purity), used as standards, were purchased from MedChem Express and Merck, respectively. The products were confirmed by GC-MS. The conditions of betulin oxidation and the analytical procedure were previously published. 34,35 Size exclusion chromatography was performed to investigate oligomer and polymer formation on the spent catalyst surface. 60 20 mg of the spent catalysts was added to a round flask together with 20 ml of the solvent heptane and a condenser. The flask was placed in an oil bath and heated to 98 °C. Thereafter, extraction occurred for four hours with a stirring rate of 400 rpm. The flow rate of the inert gas, consisting of 5% Ar in 95% N 2 , was set to 100 ml min −1 . The solution obtained after the 4 h extraction was then kept at 40 °C, until complete evaporation of heptane. The resulting residue was then dissolved in 10 ml of tetrahydrofuran, and thereafter fil-tered for analysis. The resulting concentration of the residue was 2 mg ml −1 . The analysis was carried out using a SEC-HPLC system equipped with two columns, a Guard column with the dimensions of 50 mm × 7.8 mm and a Jordi Gel DVB 500A column with the dimensions of 300 mm × 7.8 mm. The TOF values were calculated as the number of converted moles of betulin per mole of exposed catalytic site per unit time, during the first 15 min, taking into account the metal dispersion: where n Betulin is the number of converted moles of betulin, n Metal is the number of moles of the metal, D is dispersion and t is time. The number of surface metal atoms was calculated knowing the average gold particle size measured by transmission electron microscopy (TEM). ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Oxidation of a wood extractive betulin to biologically active oxo-derivatives using supported gold catalysts", "journal": "Royal Society of Chemistry (RSC)"}

palladium-catalyzed_<i>ortho</i>-halogenations_of_acetanilides_with_<i>n</i>-halosuccinimides_via_di

## Abstract: A solvent-free palladium-catalyzed ortho-iodination of acetanilides using N-iodosuccinimide as the iodine source has been developed under ball-milling conditions. This present method avoids the use of hazardous organic solvents, high reaction temperature, and long reaction time and provides a highly efficient methodology to realize the regioselective functionalization of acetanilides in yields up to 94% in a ball mill. Furthermore, the current methodology can be extended to the synthesis of ortho-brominated and ortho-chlorinated products in good yields by using the corresponding N-halosuccinimides. ## Introduction Aryl halides have been widely utilized in organic syntheses, which give access to a range of complex natural products . However, traditional halogenations of aromatic compounds by direct electrophilic halogenation and Sandmeyer reaction have several drawbacks such as low regioselectivities, complicated reaction procedures and even a risk of danger. Thus, it is necessary to discover new approaches to the regioselective construction of C-X bonds. With the development of transition-metal-catalyzed cross-coupling reactions, a series of halogenations at the ortho-position of the directing groups have been disclosed . Nevertheless, from the viewpoint of green chemistry, the reduction or even elimination of organic solvents, shorter reaction times, simplification of work-up procedures and improvement of product yields are highly demanding. In recent years, the application of mechanochemical techniques in organic syntheses has attracted increasing attention . A few mechanochemical ortho-C-H bond activation reactions under the catalysis of rhodium and palladium salts have been reported . Hernández and Bolm reported the rhodium-catalyzed bromination and iodination of 2-phenylpyridine using N-bromosuccinimide (NBS) and N-iodosuccinimide (NIS), respectively, as the halogen source . However, the mechanochemical ortho-halogenation using the cheaper palladium catalysts has not been reported yet. In continuing our interest in mechanochemistry [21,22, and C-H activation reactions , we have independently investigated the solvent-free ortho-iodination of acetanilides under ball-milling conditions . In addition, the current reaction can be extended to ortho-bromination and ortho-chlorination by using the corresponding N-halosuccinimides. Herein, we report these regioselective ortho-halogenations in detail. ## Results and Discussion To begin our study, N-(p-tolyl)acetamide (1a) was chosen as the model substrate to react with NIS using Pd(OAc) 2 as the catalyst to optimize reaction parameters such as additive, reaction time and reagent ratio. The reaction of 1a (0.4 mmol) with NIS (0.4 mmol) was initially performed under the catalysis of Pd(OAc) 2 (10 mol %) in a Spex SamplePrep 8000 mixer mill at a frequency of 875 cycles per minute at room temperature for 3 h. Unfortunately, the desired iodinated product was not detected (Table 1, entry 1). Then, various acids were examined because the addition of acids into the reaction system could promote the C-H bond halogenation according to the previous literature . As desired, compound 2a was isolated in 87% yield when p-toluenesulfonic acid (PTSA) was employed (Table 1, entry 2). A control experiment was conducted for the reaction of 1a with NIS in the absence of Pd(OAc) 2 , yet still with PTSA as the promoter, and no iodinated product was furnished (Table 1, entry 3). The use of D-camphorsulfonic acid (D-CSA) or mesitylenesulfonic acid dihydrate provided inferior results than that obtained in the presence of PTSA (Table 1, entries 4 and 5 vs entry 2). Furthermore, no desired product was obtained when pyridine-2-sulfonic acid, 2-nitrobenzoic acid, 2-aminoethanesulfonic acid or tungstophosphoric acid hydrate (HPA) was used in the reaction (Table 1, entries 6-9). Thus, the combination of Pd(OAc) 2 with PTSA was essential for the reaction to take place effectively. Subsequently, the ratio of substrates was investigated, and the results demonstrated that the amount of both NIS and PTSA affected the product yield. Decreasing or increasing the amount of PTSA was not beneficial to the reaction (Table 1, entries 10 and 11). When the amount of NIS was increased from 1.0 equiv to 1.5 equiv and 2.0 equiv, the yield of the iodinated product did not further go up (Table 1, To demonstrate the generality of this protocol, the regioselective iodination of a series of acetanilides was then examined in the presence of Pd(OAc) 2 and PTSA under the ball-milling conditions (Table 2). Gratifyingly, the ortho-iodinated acetanilides were obtained in moderate to good isolated yields. Both p-Me and m-Me-substituted acetanilides provided products 2a and 2b in excellent yields of 87% and 80%, respectively (Table 2, entries 1 and 2). As expected, 3,4-dimethylacetanilide underwent iodination successfully at the less sterically hindered ortho-position and gave product 2c in 85% yield (Table 2, entry 3). The unsubstituted acetanilide provided the desired product 2d in 77% yield (Table 2, entry 4). It is worth mentioning that the presence of a potentially reactive group, such as fluoro, chloro, and bromo substituents in the acetanilides was tolerable, and products 2e-i were isolated in 51-94% yields (Table 2, entries 5-9), highlighting the functional group compatibility of the current protocol. The presence of an acetyl group at the para-position of the phenyl ring of acetanilide 1j decreased the yield of the corresponding product 2j to 11% (Table 2, entry 10). Unfortunately, substrates bearing a strong electron-donating methoxy group and a strong electronwithdrawing nitro group could not afford any desired products, and the reason is not quite clear right now. In an aim to investigate the influence of the milling frequency, the model reaction of 1a with NIS was conducted by employing different types of mixer mills with different milling frequencies. Ortho-iodized acetanilide 2a was furnished in 90% yield after milling for 2 h by using a Retsch MM 200 mixer mill (30 Hz, Scheme 1a). At a milling frequency of 50 Hz in a Spex SamplePrep 5100 mixer mill, the iodination was accomplished within 1.5 h to afford the corresponding product 2a in 92% yield (Scheme 1b). According to the above experimental results, it could be concluded that the higher milling frequency had a beneficial effect on the reaction efficiency in terms of product yield and reaction time. ## Scheme 1: The influence of the milling frequency on the reaction of 1a with NIS. To illustrate the superiority of the ball-milling technique, the reaction was also investigated in an organic solvent. The reaction of 1a with NIS conducted in toluene at room temperature for 3 h provided the desired product 2a in only 49% yield, which was inferior to those obtained by our mechanochemical approaches (Scheme 2). Scheme 2: Palladium-catalyzed ortho-iodination of 1a in toluene. The plausible mechanism is proposed and depicted in Scheme 3. The addition of PTSA was essential for the present reaction. It is believed the more active Pd(OTs) 2 is formed in situ from Pd(OAc) 2 and TsOH . The formed Pd(OTs) 2 inserts into the ortho C-H bond of the anilides after coordination to the oxygen atom of the amide moiety, affording the species A. Oxidative addition of the species A with NIS generates the Pd(IV) complex B. Finally, the iodinated product is provided by reductive elimination along with regeneration of Pd(OTs) 2 in the presence of TsOH. It was intriguing to find that N-bromosuccinimide (NBS) and N-chlorosuccinimide (NCS) could also be used as reaction partners to react with the representative acetanilide 1a under identical ball-milling conditions. The corresponding ortho-brominated and ortho-chlorinated products 3a and 4a were obtained in 73% and 77% yields, respectively (Scheme 4). ## Conclusion In summary, we have developed a solvent-free and efficient protocol to synthesize ortho-iodinated acetanilide derivatives with Pd(OAc) 2 as the catalyst and N-iodosuccinimide as the halogen source under mechanical milling conditions. This protocol shows its advantages in terms of high regioselectivity, simple operation and environmentally friendliness. In addition, the present protocol can be extended to the synthesis of orthobrominated and chlorinated acetanilides delivering good yields by using the corresponding N-halosuccinimides.

chemsum

{"title": "Palladium-catalyzed <i>ortho</i>-halogenations of acetanilides with <i>N</i>-halosuccinimides via direct sp² C\u2013H bond activation in ball mills", "journal": "Beilstein"}

chemical_reaction_within_a_compact_non-porous_crystal_containing_molecular_clusters_without_the_loss

## Abstract: The very rare occurrence of a gas-solid chemical reaction has been found to take place on a molecule within a compact non-porous crystal without destroying its long-range structural order and retaining similar crystal structures when yellow crystals of Fe II 4 (mbm) 4 Cl 4 (MeOH) 4 were exposed to air to give black [Fe III 4 (mbm) 4 Cl 4 (OH) 4 ]$2H 2 O. The latter cannot be synthesised directly. The original cluster underwent an exchange of methanol to hydroxide, an oxidation of Fe(II) to Fe(III), a change in stereochemistry and hydration while the packing and space-group remained unaltered. ## Introduction One of the requirements for a chemical reaction to take place is that the reactants should be within close proximity for electronic interactions to promote bond breaking and bond formation. 1 Thus, the reactants have high probability of getting close together when they are in liquid or gas states. Reactions in the solid state require repetitive grinding and mixing before heat treatment, and the crystalline state can then be obtained at high temperatures. 2 With the exception of oligomerization via irradiation or heat treatment, there are rarely reactions that take place in the solid state with retention of the crystallinity. 3 Gassolid reactions take place at the surface and invariably destroy the crystalline state of the solid if the structure is non-porous. However, when it is porous, the gaseous reactants can penetrate the structure and modify it without destruction. 4 Therefore, gas-solid reactions remain exotic, and the advances made in the past twenty years in the feld of porous metal-organic frameworks (MOFs) are slowly changing our perceptions and have introduced several new conceptual synthetic approaches leading to good quality crystals. One prominent advance is the post-synthetic modifcation (PSM) of a crystalline solid without destroying its crystalline state, which has given rise to a new feld of single-crystal-to-single-crystal (SC-SC) transformations. Some notable in situ advances include (a) desolvation, (b) solvent exchange, (c) coordination at the naked metal sites, and (d) the reaction of the organic moiety. The most remarkable advance is the replacement of the metal centres of a MOF without dissolving the crystals, which has major importance for the development of smart and intelligent materials, as it evidences the process of auto-repairing. All of these advances are possible due to strong connectivity through a combination of covalent and dative bonds within the frameworks and, most importantly, their porous character, which provides space for the reactants to get to the reaction sites. However, when the crystals are non-porous, the compactness of the building units limits the reaction to the surface and, consequently, the crystallinity is destroyed by any form of modifcation of the crystals through chemical reactions. This is even more likely if the crystals contain molecular units that are held by weak supramolecular interactions. 22 Two interesting examples have been reported where discrete clusters transform into one-dimensional chains 23 and layers 24 and maintain their crystallinity. Notably, Atwood et al. reported some non-porous organic solids absorbing various gases without chemical reactions in an SC-SC manner. Two features were involved during the SC-SC transformation of nonporous crystals containing discrete molecules. One is guest transport through the crystal lattice, such as coordinated ligand exchange 20,30,31 and the addition of H 2 to a coordinated ligand. 32,33 The other is charge reorganisation between the metal ion and ligand, such as metal complexes involving tautomerism and hydrogen-atom transfer. 39 Note that a dimolybdenum molecular pair with a [Mo 2 ](m-OH) 2 [Mo 2 ] core undergoes a deprotonation process, 40 and a dicobalt core is known to serve as an active site for oxygen chemisorption/ desorption in a reversible SC-SC transformation. 41 Therefore, it is considered that chemical reactions that produce molecules within a compact crystalline solid, involving not only guest transport, but also a change in the metal ion charge without destroying the crystal or perturbing the crystalline long-range order, are really very rare. Here, we present a unique gas-solid reaction that can be considered as a different form of rusting, not of iron metal but of a tetranuclear Fe(II) molecular complex, without destroying its crystalline state, and the gaseous reactants are H 2 O and O 2 from the atmosphere. ## Results and discussion The yellow crystals of Fe 1-2d underwent an annealing process in the frst 8 days to form 1-8d, which then lost its crystallinity slowly to form 1-180d via hydration (Fig. 1 and S1 †). The results that were obtained are quite unique and reveal a balance of stability as a function of time. Following the determination of the structures from several crystals under noncontrollable conditions, we selected three similarly sized virgin yellow crystals of 1 from one batch for a systematic study under ambient conditions (27 C and 56% RH). The frst crystal was used for diffraction data collection within four hours, which reproduced the structure that was found in several other crystals that were studied independently and gave consistent geometrical parameters. The second and third crystals were exposed to air under ambient conditions for 2 days (1-2d) and 8 days (1-8d), respectively, before collecting the data (Table S1 †). The presence of an {Fe 4 O 4 } cubic core is the key feature of the three structures, in which Fe and O atoms occupy alternate corners of a slightly distorted cube (Fig. 2 and Table S2 †). They all adopt the non-centrosymmetric space group P N atom and one O atom from a chelating mbm ligand in an orthogonal plane (Fig. 2a and b). In contrast, the asymmetric unit of 1-2d contains the same atoms, except that an OH group has replaced the MeOH molecule (Fig. 2c and d). The Fe centre of 1-2d has the same coordination sphere as that of 1, but the Cl atom is now in the place of the methanol and the hydroxide is in the position of the Cl atom. The replacement of the neutral methanol by the charged hydroxide increases the oxidation state of Fe from two to three. The Fe(II)-O distances of 1 lie in the narrow range 2.107-2.181 . The Fe(III)-O distances of 1-2d were found to lie in the wide range 2.019-2.230 . The Fe-O-Fe bridging angles for 1 fall in the narrow range 96.32-99.66 , but again they fall in the wide range 97.64-103.32 for 1-2d. The distances and angles are comparable to those reported for other {Fe 4 O 4 } n+ complexes in the literature. 1-2d and 1-8d have the same Fe coordination sphere but with slightly different structural parameters, with Fe-O distances of 2.054-2.251 and Fe-O-Fe bridging angles of 97.94-103.90 . The Fe-OH distances are 2.019 (1-2d) and 2.055 (1-8d). Given the ease of the deprotonation of the terminal hydroxide, leading to the formation of iron oxides, 46,47 the stability of 1-2d and 1-8d is quite remarkable. 48 To the best of our knowledge, it also has the highest number (4) of terminal hydroxides in a discrete coordination complex. Surprisingly, a discrete cluster containing the fully oxidised Fe III 4 O 4 cubane has not been reported, but it exists for a series of octanuclear {Fe 8 } complexes with central Fe III 4 O 4 . 49,50 The 24 known discrete {Fe 4 O 4 } cubanes are either divalent Fe(II) or mixed-valent Fe(II)/(III) (Table S3 †). Therefore, we attempted to synthesize 1-2d directly using Fe III Cl 3 $6H 2 O as the starting material, but all of the attempts have so far resulted in only yellow crystals of 1. This could be due to poor stability under ambient conditions or the fact that ferric ions tend to connect via oxo ligands to form {Fe III 4 O 6 } layers in solution. 43 Consequently, 1-2d and 1-8d represent the frst discrete ferric cubes with terminal hydroxide ligands. The terminal hydroxide ligand is possibly stabilized by a combination of steric hindrance provided by the bulky ligand and the H-bond between the clusters (Fig. S2 and S3 †). It is interesting to note that the coordinated Cl atom changes its position during the transformation from 1 to 1-2d and 1-8d, while the O and N atoms from mbm ligands remain in their original positions. 51 This suggests that the mbm ligand is strongly bonded compared to the methanol. It also implies that an intermediate fvecoordinated Fe is formed by the initial departure of the methanol. It is logical to assume that the transformation proceeds gradually from the nucleation sites at the surface to the entire crystal without a loss of crystallinity. However, all of the endeavours to remove MeOH from crystals of 1 under vacuum while retaining its crystallinity have been unsuccessful. The slow post-synthetic SC-SC modifcation provides us with an opportunity to track this progressive gas-solid reaction closely. Given the change in colour, our frst conclusion is that the compound was being oxidized, which was subsequently confrmed using crystallography. We then performed an experiment to show that water is also required. When 1 was kept in anhydrous methanol, no change in colour was observed after 2 days, but when it was exposed to air, it darkened (Fig. S4 †). This suggests that the departed methanol is frst replaced by water followed by oxidation leading to Fe III and hydroxide. 52 Powder X-ray diffraction (PXRD) patterns suggest that the reaction can be stopped by keeping the samples under a nitrogen atmosphere. Time-dependent PXRD patterns with exposure times in air of up to 240 days (Fig. S5 †) reveal that the samples diffract well for the frst 8 days, but very poorly by 20 days. The results suggest that the crystallinity and long-range order of the cluster in the structure of 1 are maintained up to at least 8 days. Although the diffracting power progressively weakens as a function of time, leading to an almost amorphous solid, the presence of the peak at 2q ¼ 9.0 for 1-180d indicates the existence of short-range order in the structure up to 180 days after the annealing process. In addition, time dependent crystallography was used on one single crystal. A yellow crystal of 1 was used to collect oscillation frames in air for one orientation as a function of exposure time for up to 50 h (Fig. 3 and S6 †) while it was in the enclosure of the Bruker diffractometer under a controlled atmosphere (27 C; 56% humidity). Zooming in on a selected area of the frames shows a pair of Bragg reflections, one weak and one strong; the intensity ratio changes slowly during 24 h, but by 48 h, only one Bragg reflection is present (Fig. 3b). This suggests the progressive transformation of the phases and the existence of two diffracting lattices from one crystal at the intermediate times without the loss of the crystalline state. 53,54 We should also note that the presence of the weak peak in the frst frame indicates that the crystal has already been partially oxidised during the mounting of the crystal. There are different effects involving SC-SC transformations that have been reported. These are the loss of solvents, the exchange of solvents, the reaction of the organic ligands, and the exchange of the metal. In this context, the SC-SC transformation of 1 to 1-2d possesses the most chemical changes (Table S4 †). Crystallography was used to fnd both 1 and 1-2d adopt the same space-group and possess {Fe 4 O 4 } cores with four chemical changes. An interesting question that follows the above observations is: how do the water and oxygen molecules go through a nonporous crystalline structure? Indeed, this is a very rarely observed process. For the case of the solvation of the crystals containing calixarene, Atwood et al. suggested that the guest can be transported through the non-porous solid via dynamic van der Waals cooperativity and the expansion of the entire solid. 27,29 This is not the case here. Therefore, we propose the following plausible process for our case. Because the intercluster interactions between the clusters in the structure of 1 are weak, the MeOH molecule at the surface of the crystal can easily be dissociated, leaving a vacant site at the metal centre for reaction with O 2 and H 2 O. Since the new oxidised molecular unit is smaller than the original one, the surface is hydrated and the water can move further to neighbouring molecules provoking further reaction, which then propagates through the whole crystal. The expected exothermic energy from the oxidation reaction drives the removal of further MeOH molecules. In contrast, the strong intercluster interactions in the structure of 1-2d help to maintain the long-range order of the single-crystal lattice. It is also different from the iron rusting case, for example, where the iron crystals are eroded due to the presence of H 2 O and O 2 (often catalysed by acidic gases) to form ironoxide crystals at the surface. The oxidation of Fe(II) to Fe(III) introduces a change in the spin and orbital states of the magnetic ions. Therefore, we have followed the changes in the magnetic properties using a SQUID magnetometer, and HF-EPR and Mössbauer spectroscopy (further details are given in the ESI †). The high temperature susceptibility data reveal a change from dominant ferromagnetic (q ¼ +2.8(2) K for 1, from Curie-Weiss law ftting) to strong antiferromagnetic exchange (q ¼ 53.0(2) K for 1-2d) with time (Fig. 4a and Table S5 †). The results from the ftting of the magnetic data (Fig. S7 †) correlate well with those observed for related compounds and those calculated using DFT (Table S6 †). 49,50 From the HF-EPR spectra at low temperatures and different frequencies, g-values of 1.49, 2.92, 3.61 and 5.50 and three energy gaps of 27, 46, and 190 GHz were extracted (Fig. S8 †), which confrm the ZFS of the Fe(II) atom in 1. These gaps are in good agreement with the values of D (14.3(1) cm 1 ) and E (2.1(1) cm 1 ) obtained from modelling the high temperature data. 1-2d, 1-8d and 1-180d only have two resonances that correspond to g-values of 2.19 and 2.11 for 1-2d, 2.08 and 2.04 for 1-8d, and 2.09 and 1.99 for 1-180d, but they have small energy gaps of $20 GHz that are consistent with those of singlet Fe(III) ions. 41 Due to the increasing AF exchange energy with time, the isothermal magnetization is harder to saturate with a feld (Fig. 4b). Moreover, Mössbauer spectroscopy also confrmed that all of the Fe(II) ions in the molecular cluster completely oxidised to Fe(III), and gives a more accurate proportion of the different valences (Fig. 4c and S9-S10 and Tables S8-S10 †). The temperature dependence of the acsusceptibility for 1 and 1-2d indicates there is no singlemolecule magnetic behaviour above 1.8 K (Fig. S11 †). ## Conclusions In summary, the progressive post-synthetic transformation of a yellow ferrous cubane cluster, Fe II 4 (mbm) 4 Cl 4 (MeOH) 4 , into its dark ferric congener, [Fe III 4 (mbm) 4 (OH) 4 Cl 4 ]$2H 2 O, as a function of exposure to air has been observed and explored using crystallography, magnetometry, and HF-EPR and Mössbauer spectroscopy. This unique single-crystal-to-single-crystal transformation prevails up to 8 h as the Fe II ions are oxidised to Fe III ions, but the crystallinity degrades slowly afterwards due to disorder induced by water intake. Although SC-SC transformations involving non-porous molecular materials have been reported, to the best of our knowledge, no material has such abundant guest transport through the crystal lattice, with dioxygen entering and methanol departing. In particular, four chemical changes were noted: (a) the replacement of the methanol by the hydroxide (Fig. S12 †), (b) a coordination site swap of the chlorine atom within the Fe octahedron, (c) the oxidation of Fe and (d) hydration. The consequence of these changes is reflected in the SQUID magnetometry results, where a progressive change from ferromagnetic coupling with considerable single-ion anisotropy for the virgin yellow crystals to strongly antiferromagnetic coupling and weak anisotropy for the oxidised dark crystals is observed. Mössbauer spectra confrm the complete oxidation of Fe II to Fe III and gave more accurate proportions of the two valences at different times. This astonishing retention of the crystalline state through the three chemical changes on a molecule can be regarded as a gas-solid state reaction. ## Experimental All of the reagents were obtained from commercial sources and used without further purifcation. Elemental analyses for C, H, Fig. 4 (a) The temperature dependence of c g T in 1 kOe for the samples that were exposed to air for different periods of time. The solid lines represent the theoretical fits using the parameters given in the ESI; † (b) their isothermal magnetisation at 2 K; and (c) zero-field 57 Fe M össbauer spectra at 80 K for the fresh sample (1), and samples of 1-8h, 1-15h, 1-2d, and 1-180d that were exposed to air. The simulations shown in red correspond to the sum of all of the components. and Hmbm (1.0 mmol, 162 mg), triethylamine (0.1 mL) and methanol (8 mL) under similar conditions resulted in light yellow rhombic crystals of 1 (75 mg, yield 30%). Fe(III) is reduced to Fe(II) under solvothermal conditions in the presence of methanol. 1-2d, 1-8d and 1-180d were obtained by exposing crystals of 1 to air at ambient temperature for 2, 8 and 180 days, respectively (Fig. S1 †). The reaction rate somehow varied upon the change of ambient temperature and humidity. The rate of this blackening appears to have been faster for the smaller crystals. The fnal phase of the black crystals was then identifed using X-ray diffraction, while the solvent was confrmed using TG-IR spectroscopy and EA.

chemsum

{"title": "Chemical reaction within a compact non-porous crystal containing molecular clusters without the loss of crystallinity", "journal": "Royal Society of Chemistry (RSC)"}

benchmarking_density_functionals,_basis_sets,_and_solvent_models_in_predicting_thermodynamic_hydrici

## Abstract: Many renewable energy technologies, such as hydrogen gas synthesis and carbon dioxide reduction, rely on chemical reactions involving hydride anions (H − ). When selecting molecules to be used in such applications, an important quantity to consider is the thermodynamic hydricity, which is the free energy required for a species to donate a hydride anion. Theoretical calculations of thermodynamic hydricity depend on several parameters, mainly the density functional, basis set, and solvent model. In order to assess the effects of the above three parameters, we carry out hydricity calculations for a set of molecules with known experimental hydricity values, generate linear fits, and compare the R-squared, root-mean-squared error (RMSE), and Akaike Information Criterion (AIC) across different combinations of density functionals, basis sets, and solvent models. Based on these results we are able to quantify the accuracy of theoretical predictions of hydricity and recommend the parameters with the best compromise between accuracy and computational cost. ## Introduction The hydricity of a molecule is given by meaning it measures the free energy difference before and after a hydride transfer reaction. Hydride transfer reactions play a crucial role in various renewable energy technologies, such as the electrochemical reduction of CO 2 into carbon-based fuels or H 2 synthesis . Traditionally, transition metal hydrides have been the most popular candidates for such applications , but many of these metals are expensive, unsustainable, and toxic . Organic hydrides such as dihydropyridine and benzimidazoles are promising metal-free, renewable alternatives to their costly counterparts. Therefore, studying the hydricities of organic compounds in the hopes of selecting more metal-free catalysts for CO 2 reduction would be a valuable effort towards closing the carbon cycle. However, the hydricity is expensive and laborious to measure experimentally; it involves summing equilibrium constants, acid dissociation constants, and free energies over several thermochemical reactions , not to mention having to synthesize the molecule of interest in the first place. There have been several works that use Kohn-Sham density functional theory (DFT) to theoretically predict hydricities. Ref. calculated the hydricities of several metal-free hydrides via two different approaches specified in and . Ref. calculated the hydridicies of various p-and o-quinones in DMSO with geometry optimizations done using B3LYP/6-31+G*, single point calculations using B3LYP/6-311++G and MP2/6-311++G**, and correction terms calculated using B3LYP/6-31+G*, with solvent model IEFPCM for all steps of the calculation. Ref. calculated the hydricities of 6d transition metal hydrides using B3LYP as the density functional, LACVP** and LACV3P++** as the basis set for the geometry optimization/frequency analysis and single point calculations, respectively, and the Poisson-Boltzmann solvent model. These works all use 2 selected methods for the entire set of tested molecules, rather than testing several different methods and observing the effects. A benchmark of DFT methods for calculating hydricities did not exist until July 2021 , and this benchmark was exclusively for 3d transition metal complexes, while we provide a benchmark for organic hydrides. When using density functional theory to calculate hydricity, the three most important parameters to consider are the density functional, basis set, and solvent model. In this paper, we test different combinations of these three parameters, all stemming from the "base" level of theory, which uses B3LYP as the functional, TZVP as the basis set, and PCM as the solvent model. We will first summarize the systems we studied and the methods we used to calculate their hydricities. Then we will introduce the statistical measures we utilized to compare performance across different models. From these results we were able to formulate a set of guidelines for carrying out theoretical calculations of thermodynamic hydricities for organic hydrides. We end by discussing concluding thoughts and future directions. Centre and Leibniz Institute for Information Infrastructure. Any structures not available on WebCSD were built by hand on GaussView, a graphical interface used for preparing input files for quantum chemistry calculations. In such cases, we took special care to make the initial structures as close to the expected final structures as possible by manually adjusting bond angles. ## Methods For molecules numbered 20-27, we ran geometry optimizations on different acceptor structures to determine the hydridic hydrogen, i.e. the H that is donated in a hydride transfer reaction. For molecules 1 and 10, the donor structures have charge -1 and acceptor structures charge 0. For the others the donor structures have charge 0 and acceptor structures charge +1. All molecules are solvated in either acetonitrile or dimethylsulfoxide (DMSO). Table 1 gives the combinations of density functionals, basis sets, and solvent models used for our calculations. BP86, which has the lowest computational cost of the density functionals we used, is a generalized gradient approximation (GGA) functional, meaning it only uses the local electron density and gradient. B3LYP is the most widely used hybrid functional, i.e. it mixes the DFT exchange-correlation energy and Hartree-Fock (HF) exchange energy with a fixed ratio . B3LYP has 20% HF exchange, while B3LYP* has 15% . ωB97X-D3, the most computationally expensive out of the three functionals used, is a range separated hybrid functional, meaning the mixing ratio of the DFT and HF contributions vary depending on the distance between electrons . B3LYP has 3 emperical parameters fitted to experiment, while ωB97X-D3 has 17 . For the basis sets, we selected 6-31G*, TZVP, and TZVP+ (short for ma-def2-TZVP(f)-LTZ+), which have 14, 19, and 28 basis functions per carbon atom, respectively . For any atoms beyond potassium (K), 6-31G* is replaced by the LANL2DZ (LDZ) basis set and an effective core potential (ECP), while and TZVP and TZVP+ are replaced by the LANL2TZ (LTZ) basis set and an ECP . TZVP+ is not only larger than TZVP, but also has the f functions removed and diffuse functions added to non-hydrogen atoms . Lastly, we used two continuum solvent models for our calculations: C-PCM ISWIG (PCM for short) and SMD. C-PCM ISWIG is a conductor-like polarizable continuum solvent model (C-PCM) with a "smooth discretization" via the Improved Switching/Gaussian (ISWIG) method. Polarizable continuum models represent the solvent by placing the solute in a cavity with an apparent charge distribution over the surface of that cavity. Boundary-element methods are used to discretize the solute/continuum interface, but this often leads to a discontinuous potential energy surface for the solute, leading to singularities. The ISWIG method is a discretization scheme that overcomes such limitations . SMD is another type of polarizable continuum solvent model, but it accounts for short-range solvent-solute interactions such as dispersion and solvent structural effects (e.g. hydrogen bonding or exchange repulsion), whereas regular polarizable continuum models only account for bulk electrostatic interactions . Model ID For models that have two columns, the first column indicates the level of theory used for the geometry optimization and frequency analysis calculations, while the second column indicates the level of theory used for the single point calculations. This composite approach allows for using more expensive levels of theory while keeping computational cost low, as geometry optimization and frequency analysis calculations involve taking many gradients while single point calculations do not. If a model has a single column, the same level of theory was used for all calculations. Levels of theory using SMD as the solvent model were calculated using Q-Chem , while others were calculated using TeraChem . The free energies of the acceptor and donor can easily be calculated on TeraChem and Q-Chem, but the free energy of the solvated hydride is difficult to compute using such methods. A hydride anion has complicated interactions with the solvent that a continuum solvent model cannot account for. Even if we were to use an explicit solvent model in an attempt to calculate the free energy of the hydride, the hydride will quickly react with surrounding molecules, making it extremely difficult to obtain a reasonable estimate of its free energy in solvent. There are several ways to circumvent this problem. Ref. calculated the free energy of the hydricity half reaction then used a reference reaction to evaluate G(H − ) and construct the thermodynamic hydricity via an isodesmic reaction scheme where AH − and BH − are the donor structures of species A and B, respectively. Meanwhile, Ref. used a thermochemical cycle given by and modeled the protons as a complex with discrete solvent molecules. In this paper, we chose to calculate the free energy of the hydricity half reaction and treat the free energy of the solvated hydride as a fitting parameter. More specifically, we compute the ∆G HHR for all molecules using one model, create a linear fit with the slope fixed at 1, and apply an overall vertical shift to all the data points so that the linear fit goes through the origin. This vertical shift corresponds to the free energy of the hydride, and the final linear fit for all our data sets are y=x. Then we calculate the R-squared, root-mean-squared error, and Akaike information criterion (AIC) for each data set to quantify the accuracy of the given model. For this benchmark, we used the AIC as a tool to verify that each model's performance is statistically significant. For instance, when model A gives a better R-squared and RMSE compared to model B, we want to confirm that this result is due to model A truly being more accurate than model B, and not because of random statistical fluctuations. That is, if we sample some noise from a Gaussian distribution and apply it to the data set produced by model B, the new linear fit better not give a better R-squared and RMSE value compared to that of model A. This is in theory what a P-value measures, but it is flawed in that the choice of cutoff for determining statistical significance is rather arbitrary. Bootstrapping could be another alternative, but we would have to perform many iterations to get the desired degree of confidence. This is why we chose the AIC, whose difference across different models serves the same purpose as P-values , as our measure of statistical significance. For our data, we used the second order AIC for small sample sizes which is applicable for n K < 40, where n is the sample size and K is the number of model parameters. Since our model is a linear fit with data points that do not lie exactly on the line, K = 3. −2 ln L is given by where a i are the values that the model predicts, x i are the actual values, and σ is the uncertainty. Since the uncertainty is unknown for our models, we perform an iteration process to find the uncertainty that maximizes the likelihood function L for each model. To compare the AIC across the different models, we calculate the exponential of the AIC differences, or the relative likelihood: which is proportional to the i th model's probability for minimizing information loss. We have disregarded the normalization factor that would give us the AIC weights for convenience. The model with the lowest AIC (AIC min ) has a relative likelihood of exactly 1. Since the n and K of all our models are the same, the 2K + 2K(K+1) n−K−1 terms cancel when computing the relative likelihood, leaving only the log-likelihood terms in the exponent. [ From Figure 3, we can see that ωB97X-D3 gives the best accuracy, as expected. The result for ωB97X-D3/TZVP/PCM in DMSO seems to be inconclusive at first sight, since it gives an R-squared value closest to 1 while giving the largest RMSE. However, the relative likelihood plots (Figures 8 and 9) confirm that ωB97X-D3 is indeed the most accurate density functional. It is also apparent that the effect of using B3LYP* instead B3LYP is minimal; this is because B3LYP* was originally developed to more accurately predict low-spin/high spin energy splitting in molecules with multiple spin configurations, i.e. transition metal complexes . Since we are only working with or-ganic hydrides, changing the percentage of HF exchange will not impact the calculated hydricity in a meaningful manner. Looking at Figures 4 and 5, it is clear that using a bigger basis set does not necessarily give better results. When using B3LYP as the density functional, TZVP performed worse than 6-31G*, and using TZVP for the geometry optimization/frequency analysis and TZVP+ for the single point calculation also gave a slightly worse accuracy than using 6-31G* and TZVP. With ωB97X-D3 as the density functional, TZVP performed best for acetonitrile while 6-31G*, the smallest basis set, gave the best accuracy. This is not surprising given that our test molecules were all organic hydrides, meaning we rarely have atoms beyond K. For users looking to calculate the hydricity of an organic molecule, we would recommend using a basis set no larger than TZVP, considering the trade-off between computational cost and accuracy. From Figures 6 and 7, we can conclude that the solvent model can be turned off for geometry optimization and frequency analysis calculations. We can also see that turning off the solvent model for the single point calculation has a minimal impact on accuracy for molecules in DMSO compared to those in acetonitrile. This is because the effect of using a continuum solvent model is largely influenced by the net charge of the solute, and we apply the same vertical shift to all data points in a single data set. If all molecules in a data set have the same charge, such as those in DMSO, the near-identical effect of turning off the solvent model is effectively "cancelled out" by the vertical shift. However, if any molecules have a different charge compared to the rest of the data set, like molecules 1 and 10 in acetonitrile, their calculated hydricities differ drastically from the other molecules, leading to the result shown in Figure 2b. Comparing the effects of using PCM versus SMD, SMD does perform slightly better than PCM but the difference is minimal. SMD calculations take considerably longer than PCM calculations to both set up and run, and they are also more prone to convergence failures. This leads us to recommend using PCM when calculating the hydricity of an organic hydride. Lastly, we examine the relative likelihoods of each method to test the statistical significance of our results. Most methods give a log-scale relative likelihood on the order of 10 −2 . Looking at models that used ωB97X-D3 (D03, D12, D13), we can see that all such models gave a relative likelihood noticeably closer to 1 (closer to 0 in log-scale) compared to other methods, giving us confidence that the improved accuracy coming from using ωB97X-D3 over other density functionals is statistically significant. The relative likelihoods of methods using no solvent model at all (D05, D06) show that the improvements coming from using a solvent model during the single point calculation are also statistically significant. In contrast, the effect of changing the basis set generally does not seem to be statistically significant, comparing across D01, D02, D08, D09 (B3LYP/X/PCM). Comparing across D03, D12, D13 (ωB97X-D3/X/PCM), 6-31G* performed the best with statistically significant improvement, even though it was the smallest basis set tested. This confirms our statement above that using a larger basis set is not recommended. The effect of using SMD (D14, D15) over PCM is also not statistically significant, again leading us to recommend using PCM, the less computationally expensive solvent model. ## Conclusions When calculating the hydricity of organic hydrides, we recommend using ωB97X-D3 for the density functional, a basis set no larger than TZVP, and PCM as the solvent model. We also generally advise turning PCM off for geometry optimization and frequency analysis calculations, and keeping PCM on for the single point calculations. If the molecule of interest has a net neutral charge, or if one is interested only in the relative hydricities across different molecules with the same charge, then it is safe to turn off PCM for the single point calculations as well. A future direction of this project is running ab-initio molecular dynamics (AIMD) simulations to study the hydration of solutes in water. Continuum solvent models such as PCM or SMD fail to model strong solvent-solvent and solvent-solute interactions such as hydrogen bonding, so for hydrides in water, a simple DFT calculation with just a continuum solvent will not suffice. A diagnostic algorithm that can determine the importance of certain solvent-solute interactions and give the user a recommendation as to which solvent model to use is also a future direction that would be valuable to theoretical chemistry literature. Such an algorithm could be developed using machine learning approaches. ## Acknowledgements We thank Louise Berben for inspiring this project. We would also like to recognize former and current group members Yudong Qiu for developing a large fraction of our group's cluster tools that we still use and benefit from to this day, Hyesu Jang for her helpful modifications to the TeraChem output file format, and Nathan Yoshino for helping us figure out the difference between B3LYP and B3LYP*. We would also like to acknowledge Maria Fernanda Guizar for answering our questions about statistical hypothesis testing.

chemsum

{"title": "Benchmarking Density Functionals, Basis Sets, and Solvent Models in Predicting Thermodynamic Hydricities of Organic Hydrides", "journal": "ChemRxiv"}

radical–anion_coupling_through_reagent_design:_hydroxylation_of_aryl_halides

## Abstract: The design and development of an oxime-based hydroxylation reagent, which can chemoselectively convert aryl halides (X ¼ F, Cl, Br, I) into phenols under operationally simple, transition-metal-free conditions is described. Key to the success of this approach was the identification of a reducing oxime anion which can interact and couple with open-shell aryl radicals. Experimental and computational studies support the proposed radical-nucleophilic substitution chain mechanism. ## Introduction Arene hydroxylation reactions are powerful enabling synthetic methods which are routinely used in the preparation of highvalue pharmaceuticals, agrochemicals, polymers and natural products. 1 Many different synthetic approaches have been developed to form aryl C(sp 2 )-OH bonds, 2 but in terms of cost, operational simplicity and toxicity, nucleophilic aromatic substitution (S N Ar) 3 represents one of the most attractive and frequently used methods. 4 However, the broad application and selectivity of this approach is limited by the high basicity and low nucleophilicity of the hydroxide anion. Hydroxide surrogates have been developed to improve these aspects, but their reactivity is still mostly limited to aryl fluorides or chlorides bearing strong electron-withdrawing groups in either the ortho or para positions. 5 The development of more general, transition-metal-free 6 substitution reactions for arene hydroxylation is therefore a topic of signifcant importance with wide-reaching synthetic potential. It has long been known that aryl halides that are not activated with strong electron-withdrawing groups can be substituted with a variety of different nucleophiles through the radical-nucleophilic substitution (S RN 1) chain mechanism. 7 However, hydroxide anions do not participate in S RN 1 mechanisms since such processes are driven by electron transfer (ET) and hydroxide anions are poor electron donors. Consequently, the activation barrier for radical-anion coupling is insurmountably high. This is a general problem with oxygen nucleophiles as, to the best of our knowledge, there is no known oxygen-based anion which can engage in intermolecular coupling with aryl radicals to form new C(sp 2 )-O bonds. 7b,8 Our efforts in solving this limitation are outlined herein. In particular, we rationalised that oxime anions could not only be electronically tuned to initiate and favour an S RN 1 process, but also serve as hydroxide surrogates. Indeed, based on literature precedent with perfluoroalkyl iodides, 9 it was envisaged that oxime anions 1 may readily form charge-transfer complexes 10 (CTCs, 2) with aryl halides 3, which could be activated under mild conditions to promote the formation of aryl radical intermediates 4 (Scheme 1a). Radical-anion coupling could then be rendered kinetically favourable by employing a sufficiently reducing oxime anion (Scheme 1b). In addition, it was anticipated that the oxime p-system could also alleviate the need for the aromatic coupling partner to accommodate the unpaired electron in this coupling process (e.g. 5 vs. 6), and therefore enable coupling with a broader range of substrates. Finally, ET from the coupled radical anions 6 to the aryl halides 3 could propagate a radical chain and afford O-aryl oxime intermediates 7 (Scheme 1c), which as demonstrated by Fier and Maloney 11 can readily fragment under basic conditions to afford phenols 8. In this paper, using the design rationale set out in Scheme 1, we report the development of an easily handled oxime-based nucleophile which can selectively substitute an array of electronically diverse arenes bearing every common halide (F, Cl, Br, I) to form phenols under operationally simple, transitionmetal-free conditions. The proposed S RN 1 chain mechanism is supported by experimental and DFT computational studies. ## Results and discussion Our studies commenced by reacting aryl bromide 3a Br with a range of electronically diverse oximes (9a-d are representative) using KOt-Bu in anhydrous DMSO (0.2 M) at 30 C for 16 h under nitrogen (Table 1, entries 1-4). In all cases, we observed the formation of phenol 8a in modest to excellent yield, with electron-rich pyrrole-based oxime 9d proving optimal (75%, entry 4). The compatibility of oxime 9d with different bases was also demonstrated (KOH and Cs 2 CO 3 ), but phenol 8a was obtained in diminished yields (entries 5 and 6). Notably, strongly coloured solutions were observed in every reaction, which can indicate the formation of CTCs. To investigate this possibility further, the reaction using oxime 9d was irradiated with blue LEDs (l max ¼ 455 nm) for 1 h, which gave phenol 8a in 65% yield instead of 38% yield in the dark or 44% yield when exposed to ambient light from the laboratory (entries 7-9). However, under these photochemical conditions the yield of 8a was partially diminished by the formation of the hydrodehalogenated byproduct 10, which suggested that aryl radicals may be potential intermediates in this reaction. Indeed, reactivity was signifcantly inhibited by the addition of galvinoxyl or DPPH (1 equiv.) as electron accepting radical scavengers, which reduced the yield of phenol 8a to #10% (entries 10 and 11). The addition of TEMPO had a relatively small effect on the yield of phenol 8a (entry 12, no trapped product was detected by high-resolution mass spectrometry but consumption of TEMPO was observed by EPR spectroscopy). However, it should be noted that the coupling of nitroxyl radicals with aryl radicals is known to be relatively slow in polar solvents. 12 The acceleration of this reaction by light, its inhibition by galvinoxyl and DPPH, and the detection of hydrodehalogenated product 10 all strongly indicated that a radical chain mechanism consistent with an S RN 1 reaction was in operation. UV/vis spectroscopic analysis of the reaction mixture and computational studies both supported the formation of a 1 : 1 CTC 2a (formed between anion 1d and aryl bromide 3a Br ), which may be activated with light or heat 10c,d to promote the formation of aryl radical 4a (Scheme 2). The envisaged coupling of 4a with oxime anion 1d was also theoretically explored by DFT computational analysis. 13 These studies suggest that radical-anion coupling is exergonic (DG ¼ 17.2 kcal mol 1 ) and there is only a modest activation barrier for radical-anion coupling (DG ‡ ¼ 15.0 kcal mol 1 ), which is almost entirely entropic in nature (DH ‡ ¼ 0.4 kcal mol 1 ). Considering this, any attractive interaction between the oxime anion and aryl radical could dramatically accelerate the rate of coupling. Indeed, we observed the formation of a weak two-centre three-electron (2c, 3e) s bonded species 11a in the gas phase. 14 In addition, when accounting for concentration effects, the large excess of the oxime anion relative to the radical-anion product will likely lower the activation barrier by $4 kcal mol 1 (see the ESI † for details). The calculated redox potential of the coupled radical anion 6a (E 1/2 ¼ 2.14 vs. SCE) indicates that propagation of a radical chain by ET to aryl bromide 3a Br (E 1/2 ¼ 1.89 vs. SCE) 15 would also be exergonic. The resultant neutral O-aryl oxime could then fragment under the basic reaction conditions to afford the observed phenol product. A polar S N Ar pathway was considered unlikely to proceed at 30 C due to the signifcant activation barrier calculated for the addition of the oxime anion (DG ‡ ¼ 32.4 kcal mol 1 ). Importantly, oxime reagent 9d is an easily handled white solid that is prepared on a gram-scale simply by condensing commercial aldehyde 12 with hydroxylamine in the presence of Na 2 CO 3 (Scheme 3). To showcase the utility of designed reagent oxime 9d, the scope of this new arene hydroxylation reaction was fully explored (Table 2). We frst sought to determine if halides other than bromine could be substituted by examining a variety of para-and ortho-substituted aromatic carbonyl derivatives (3a-e). Pleasingly, these derivatives could all be converted into the corresponding phenols in good to excellent yields, which demonstrates the compatibility of this reagent with every common halide nucleofuge. However, of the metasubstituted carbonyl derivatives, only fluoride 3f F could be efficiently substituted and that was at elevated temperature (60 C), which may be due to a switch to a complementary polar S N Ar mechanism. Benzonitrile and sulfone derivatives (3g-j) were also examined and the same reactivity pattern was observed: para-substituted derivatives (3g, i) reacted smoothly at 30 C, whilst the meta-isomers (3h, j) required prolonged reaction times or heating at 60 C. This reactivity pattern may directly correspond to the rate of radical-anion fragmentation, which is typically ortho > para > meta for aryl halides. 7b More strongly electronically activated trifluoromethyl-and nitrosubstituted aryl halides (3k-n) were all hydroxylated in typically excellent yields at 30 C. Relatively unactivated 1-naphthyl and 4-biphenyl halides (3o, p) could also be substituted to afford the desired phenols in modest to excellent yields, although they generally required more forcing reaction conditions (100 C) and the use of NaOt-Bu as the base. These harsher conditions may be required to overcome higher activation barriers associated with polar pathways (S N Ar or benzyne 16 ) or challenging ET initiation events (e.g. from the oxime anion to the arene). However, the ortho-fluorine substituent of dihalogenated biphenyl 3q F could be easily and selectively substituted at 30 C to afford the phenol in 78% yield. This remarkable reactivity may be due to the sterics of the phenyl ring forcing the fluorine atom to bend out of plane, which could facilitate either a S N Ar mechanistic switch or accelerate the rate of radical anion C-F bond fragmentation. 17,18 The ortho-fluorine substituent of dihalogenated acetophenone 3r F was also selectively substituted under these reaction conditions. Next, heteroaryl halides were studied (3s-v), and pleasingly activated pyridine 3s Br could be hydroxylated in excellent yield at 30 C. Unactivated bromo quinolines 3t, u could also be substituted to afford the corresponding phenols in 44-73% yield. Interestingly, as previously observed for dihalogenated arenes, the fluorine atom of pyridine 3v F could also be selectively substituted. Finally, the wider synthetic utility of oxime reagent 9d was demonstrated through the functionalization of aryl halide containing drugs; pleasingly, fenofbrate 3w Cl , iloperidone 3x F , etoricoxib 3y Cl and blonanserin 3z F were all successfully hydroxylated (47-83% yield). Intrigued by the reactivity and selectivity of some of the aryl fluorides, which could in theory also be substituted via a polar S N Ar pathway, their reactions were also studied in the presence of galvinoxyl (Scheme 4a). Interestingly, clear inhibition was observed for every example, which indicates that these reactions are at least partially radical in nature. Alternatively, it is possible that galvinoxyl may disrupt CTC formation, which can theoretically facilitate both polar 19 and open-shell reactivity. In this regard, it should also be noted that the formation of strongly coloured reaction mixtures was observed for almost every substrate described in Table 2, which suggests that CTC formation with oxime reagent 9d could be a general process. Thus, considering these results and our previous observations, it is reasonable to assume that many of the substitution reactions described herein likely proceed via an open-shell mechanism. We therefore propose that an electron-catalysed 7c S RN 1 chain is initiated by either the formation and activation of a CTC, or a slow thermal (concerted) dissociative ET 20 from an anionic electron donor 21 (e.g. the oxime anion 1) to the aryl halide 3 (Scheme 4b). The resultant aryl radical 4 can then interact with an oxime anion 1 to form a weakly interacting cluster that may be viewed as a 2c, 3e s bonded species 11. 22 As this bond shortens, a delocalised radical anion 6 (and a standard 2c, 2e bond) is then formed by intramolecular ET from species 11 into a nearby p* orbital (on either the oxime or the aryl ring). Radical anion 6 then reduces another equivalent of 3 through intermolecular ET to regenerate aryl radical 4 and release the coupled product 7, which fragments in situ to afford the observed phenol product. 23 However, the contribution of a polar S N Ar pathway for some substrates cannot be completely excluded. ## Conclusions In summary, we have reported the design and development of a new oxime-based hydroxylation reagent, which can be used to chemoselectively convert aryl halides into phenols under remarkably simple, transition-metal-free conditions. These reactions are proposed to primarily proceed via the unprecedented intermolecular coupling of an oxygen-based anion with aryl radicals to form new C(sp 2 )-O bonds. We believe that the synthetic utility of this reagent is likely enhanced by its ability to substitute nucleofuges through complementary polar pathways. It is hoped that these fndings will facilitate the rational design of other such anionic reagents and enable new unconventional retrosynthetic strategies to be realised.

chemsum

{"title": "Radical\u2013anion coupling through reagent design: hydroxylation of aryl halides", "journal": "Royal Society of Chemistry (RSC)"}

controlled_synthesis_of_spion@sio_2_nanoparticles_using_design_of_experiments

## Abstract: The synthesis of single-core superparamagnetic iron oxide nanoparticles (SPIONs) coated with a silica shell of controlled thickness remains a challenge, due to the dependence on a multitude of experimental variables. Herein, we utilise design of experiment (DoE) to study the formation of SPION@SiO 2 nanoparticles (NPs) via reverse microemulsion. Using a 3 3 full factorial design, the influence of reactant concentration of tetraethyl orthosilicate (TEOS) and ammonium hydroxide (NH 4 OH), as well as the number of fractionated additions of TEOS on the silica shell was investigated with the aim of minimising polydispersity and increasing the population of SPION@SiO 2 NPs formed. This investigation facilitated a reproducible and controlled approach for the high yield synthesis of SPION@SiO 2 NPs with uniform silica shell thickness. Application of a multiple linear regression analysis established a relationship between the applied experimental variables and the resulting silica shell thickness. These experimental variables were similarly found to dictate the monodispersity of the SPION@SiO 2 NPs formed. The overall population of single-core@shell particles, was dependent on the interaction between the number of moles of TEOS and NH 4 OH, with no influence from the number of fractionated additions of TEOS. This work demonstrates the complexity of the preparative method, and produces an accessible and flexible synthetic model to achieve monodisperse SPION@SiO 2 NPs with controllable shell thickness. ## Introduction Superparamagnetic iron oxide nanoparticles (SPIONs) have been extensively researched due to their unique magnetic properties when compared to bulk iron oxide including: superparamagnetic behaviour with a large magnetic susceptibility; low Curie temperature; and high coercivity. 1,2 SPIONs are inverse spinel type iron oxide nanoparticles (NPs) including Fe 3 O 4 (magnetite) and γ-Fe 2 O 3 (maghemite) and are typically below 20 nm in size. 3,4 The distinctive properties of SPIONs have made them useful for a range of applications in catalysis, purification, and biomedicine. Most notably, these materials have shown excellent potential in biomedical diagnostic and therapeutic applications including, but not limited to, point-of-care diagnostics, magnetic resonance imaging contrast agents, hyperthermic cancer treatments, and targeted drug delivery. Synthetic pathways to SPIONs are diverse, including co-† Electronic Supplementary Information (ESI) available: Tables and figures containing DoE experimental factors and response variables, lack of fit test for population of SPION@SiO 2 , SPION TEM and χ m , TEM images at Level 2 TEOS, TEM images at Level 3 TEOS, TEM at treatment 222, box-plot of size distribution of the measured NP diameter from TEM analysis (d TEM ) of all treatments, size distribution of validation experiments, mass susceptibility (χ m ) of all treatments. See DOI: 00.0000/00000000. precipitation, hydrothermal or thermal decomposition, solvothermal, and sol-gel reactions, with choice of reaction depending on the desired particle size, shape, hydrophilicity/phobicity and surface functionalities. As the magnetic properties of SPIONs are dependent on their size and shape, control of such properties, as well as uniformity of SPIONs, is critical. This is particularly imperative for biomedical applications, where highly polydisperse SPIONs can result in variable net magnetisation, leading to unreliable and poorly reproducible diagnostic and therapeutic capabilities. 25,26 Co-precipitation of Fe 2+ /Fe 3+ salts is one of the most popular methods due to its high yield and low cost of manufacture; however, this typically produces SPIONs that are highly polydisperse, often with poor control over particle morphology, crystallinity, and aggregation. 6,22,26 In contrast, thermal decomposition has greater control over size and morphology, generating monodisperse NPs with high crystallinity. In this approach, an organic-iron precursor is heated in situ with an amphiphilic surfactant or ligand such as oleic acid, a hydrophobic fatty acid, forming hydrophobic coated SPIONs which typically avoid aggregation. 23, To obtain biocompatible SPIONs that are colloidally and physically stable, it is necessary to coat and functionalise the nanoparticle surface. Without this coating, the poor colloidal stability of SPIONs often leads to agglomeration under physiological conditions, leading to increased toxicity via red blood cell damage and haemolysis. 14,30,31 Furthermore, uncoated SPIONs are susceptible to oxidation, which can contribute to changes in magnetic properties and chemical behaviour. 2,8,32 There are various coating strategies to achieve this, including both covalent and noncovalent synthetic methods such as: surface stabilisation with cit-ric acid; capping with oleic acid; adsorption of polymers; or coating with inorganic material, such as silica (SiO 2 ). 14,15,29,33 The latter is of particular interest as SiO 2 coated SPIONs are able to undergo subsequent surface modification with a variety of silanes or ligands due to readily modifiable silica chemistry. 29, This provides the opportunity for functionalisation with an array of molecules such as fluorescent dyes, bio-compatible, biological or targeting ligands. 35,37, There are several coating routes to attain SPION@SiO 2 core@shell NPs, with the option of either a non-porous or a mesoporous silica shell, the latter providing an opportunity to utilise the pores for cargo storage and release. 34,46,47 Both sol-gel and reverse microemulsion routes are popular methods for the synthesis of non-porous SPION@SiO 2 NPs. The sol-gel method typically uses hydrophilic, small moleculestabilised, magnetic nanoparticles which are usually prepared by co-precipitation. Seeds of SiO 2 form on the SPION surface in-situ and grow through a Stöber mechanism under relatively mild conditions. However, SPION@SiO 2 NPs prepared using this route tend to have high polydispersity and poor control over morphology, often exhibiting multiple SPION cores within a single shell. Additionally, high yields of by-product SiO 2 , with no SPION core, are often observed. 26 Enhanced control over morphology and reduced polydispersity can be achieved using a reverse microemulsion route. 29, In this approach, hydrophobic magnetite (usually oleic acid stabilised) is dispersed in a nonpolar solvent (such as cyclohexane) along with a surfactant, e.g. IGEPAL-co-520, leading to ligand exchange. Ammonium hydroxide (NH 4 OH) and tetraethyl orthosilicate (TEOS) are added to the organic solution, forming a water-in-oil (w/o) emulsion. The SiO 2 precursor nucleates within the aqueous domain and forms around the SPION core generating SPION@SiO 2 core@shell NPs (Fig. 1). In comparison, this approach has been reported as more favourable in terms of forming single core SPION@SiO 2 NPs with controlled shell thickness. While SPION@SiO 2 NPs are a prominent area of research, there are still difficulties associated with the reverse microemulsion method. Mainly, the route produces a large fraction of noncore SiO 2 NPs, a non-magnetic by-product that requires additional processing to separate. Since reproducibility as well as particle size and morphology control are vital when considering such particles for biomedical applications, understanding the relationship between the reaction conditions and resulting SPION@SiO 2 NPs is critical. To date, optimisation of SPION@SiO 2 NPs has relied on 'one factor at a time' (OFAT) variation, where each experimental factor is changed individually. 29,34,56, Consequently, a narrow range of the experimental domain is typically examined and variable interactions overlooked. This limits the understanding of how experimental variables affect the outcome of a synthesis, often leading to inaccurate conclusions and sub-optimal results. 67 Design of experiment (DoE) is a powerful statistical tool for understanding and optimising the relationship between the experimental variables and the outcome of a process. The use of DoE for nanoparticle synthesis is gaining popularity and has provided insight into reaction mechanisms. 68 For example, Lak et al. have used a DoE approach for the optimisation of SPIONs, with tailorable magnetic properties. In two separate studies, interactions were identified between experimental factors responsible for determining the dispersity and magnetisation of SPIONs formed via thermal decomposition and non-hydrolytic synthesis. 23,69 Similarly, Roth et al. have employed DoE to study the formation of SPIONs using co-precipitation. 22 The magnetisation of the SPI-ONs was deemed highly dependent on the experimental variables used, such as the molar ratio and concentration of Fe 2+ /Fe 3+ ions. Arafa et al. used DoE to tune the properties of pregabalinloaded niosomes for therapeutic delivery. This was made possible by examining the combined effect of different factors simultaneously. 70 There have been several successful examples of the optimisation of nanoparticle synthesis and maximising their functionality using a DoE approach in the literature. 23, While DoE has proved a powerful tool for nanoparticle optimisation, this approach has not been applied in the context of SPION@SiO 2 NP synthesis. Herein, we applied a statistical approach to understand and predict the influence experimental factors have on the properties of SPION@SiO 2 NPs formed via a reverse microemulsion method. The number of moles of the reactants NH 4 OH and TEOS, as well as the number of fractionated additions of TEOS were selected for investigation (vide infra 2.4 and 3.2). The size and stability of SPION@SiO 2 NPs formed were characterised using transmission electron microscopy (TEM) and dynamic light scattering (DLS). Magnetic properties were evaluated using vibrating sample magnetometry (VSM). The experimental campaign was designed using a 3 3 full-factorial design and modelled using a response surface model (RSM) and analysed via the software JMP. 80 In this way, we aim to provide a comprehensive understanding of the influence of important preparation parameters on resulting SPION@SiO 2 NPs, and predictive capability to produce optimal single-core@shell particles. ## Chemicals All chemicals were used as purchased with no additional purification. Cyclohexane (99.5%), and n-hexane were purchased from Fisher Scientific UK. Ammonium hydroxide (28-30 %), IGEPAL-CO-520, iron (III) chloride hexahydrate, oleic acid (90 %), nitric acid (69 %) 1-octadecane (90 %), tetraethyl orthosilicate (TEOS, 99 %) and sodium hydroxide pellets were purchased from Merck. Ultrapure water was obtained using a Millipore filtration system, operated at 18.2 MOhm. ## SPION NPs Hydrophobic SPION NPs were prepared using a thermal decomposition method in accordance with the literature. 29 FeCl 3 • 6 H 2 O (0.54 g, 2 mmol) was dissolved in water (6 mL, 333 mmol), ethanol (8 mL, 137 mmol), and hexane (14 mL, 106 mmol) at room temperature. Oleic acid (1.9 mL, 6 mmol) was added to the solution and stirred for 30 minutes. Sodium hydroxide (0.24 g, 6 mmol) was added to the stirred solution and sealed in a closed vessel and heated to 70 • C for 4 hours. The resulting solution was cooled to room temperature, forming two distinct layers. The top organic layer, containing Fe(oleate) 3 , was separated, collected and washed with ultrapure water three times. To remove hexane, Fe(oleate) 3 was heated in a vacuum oven at 80 • C overnight. The dried Fe(oleate) 3 precursor was resuspended in oleic acid (0.32 mL, 1 mmol) and 1-octadecane (12.5 mL, 38 mmol) in a sealed vessel and oxygen was degassed with N 2 gas for 1 hour at room temperature. Subsequently, the solution was heated at 320 • C for 30 minutes under the inert atmosphere in a sealed vessel. The solution was cooled to room temperature, and excess ethanol was added to precipitate NPs. Hydrophobic SPIONs were collected by centrifugation and the supernatant decanted. The isolated solid was re-dispersed in hexane (approx. 1-2 mL) and subsequently precipitated in excess ethanol. This precipitation-re-dispersion process was repeated three times to purify the magnetic NPs. The SPIONs were stored in cyclohexane (5.0 mg/mL). ## SPION@SiO 2 NPs Core@shell SPION@SiO 2 NPs were prepared using a reverse microemulsion method adapted from the literature. 29 IGEPAL-co-520 (0.5 g, 1 mmol) was dispersed in cyclohexane (11 mL, 101 mmol) and sonicated for 10 min at ambient conditions. SPIONs in cyclohexane (1 mL, 5 mg/mL) were added to the stirred solution followed by the addition of ammonium hydroxide. The controlled addition of TEOS was performed using a fractionated drop-method; where a known volume of TEOS was added every 4 hours. A total of 3 additions were carried out per day and reacted overnight. For 6 and 8 additions, the 4 th (and 7 th ) addition were carried out 24 (and 48) hours after the first addition, and subsequent additions were carried out following the 4 hour interval. A schematic of the synthesis can be seen in Fig. 1, detailing how the the shell is formed within the reverse micelle. Once complete, SPION@SiO 2 NPs were washed in ethanol using centrifugation three times and redispersed in ethanol. During the study, the number of moles of TEOS and NH 4 OH, and the number of fac-torial drop-wise additions of TEOS were varied according to the design shown in Table S1, ESI. ## Characterisation Dynamic light scattering (DLS) was carried out using a Beckman Coulter DelsaMax Pro, measured at 22 • C, with samples dispersed in ultrapure water (1 mg mL -1 ). Transmission electron microscopy (TEM) was performed with a JEOL 2100 transmission electron microscope operated at 200 kV. Samples were prepared by depositing NPs dispersed in water (0.05 mg mL −1 ) onto a carbon coated 300 mesh copper grid (Agar Scientific). NP diameter and population were measured using ImageJ (version v1.53) with a sample size of approximately 300 NPs. By assuming the particles were spherical, the diameter was calculated by measuring the area of the particles, with an estimated resolution of 0.1 nm. 81 This resolution was used as the bin width to produce a histogram of each population. It was then possible to determine the dispersity of the NPs using the concept of information entropy, as first introduced by Shannon, and modified to suit nanoparticle analysis as follows: where n is the number of non-empty bins and p i is the probability of a given nanoparticle size occurring within the population. 82 A vibrating sample magnetometer-physical property measurement system (VSM-PPMS, Quantum Design) was employed to measure the mass susceptibility (χ m ) of SPION@SiO 2 NPs. Samples were prepared as dry powders and recordings were performed at 25 • C, using a maximum magnetic field strength of 20,000 Oe and the minimum magnetic field strength of -20,000 Oe, with a sweep rate of 50 Oe/sec. Graphs were analysed in Origin, and smoothed with a percentile filter. ## Experimental Design The experimental design was produced and studied using the statistical software JMP (version 15). 80 To establish the local re- sponse surface, a 3 3 full-factorial design was chosen to study the synthesis of SPION@SiO 2 (Fig. 2). This design allowed for the identification of main effects and interactions, in addition to nonlinear effects, which were modelled as quadratic functions. Three experimental variables were selected for investigation: moles of TEOS, moles of NH 4 OH, and the number of fractionated additions of TEOS, as stated in Table 1. These experimental factors and their upper and lower limits were chosen based on literature evidence, being previously identified as important parameters in the hydrolysis and condensation mechanism for the silica shell formation. The concentration and size of the SPION NP core and the concentration of the core stabilising agent, IGEPAL-co-520 were fixed, and not included in the study, as the growth of a silica shell is primarily controlled by the aforementioned experimental factors. The centre point treatment, labelled 222, was repeated five times, totalling 31 experimental runs. These were randomised to avoid bias. It is worth noting that for ease of comparison between factors and the responses, the experimental treatments have been arranged based on the DoE pattern of the experimental factors (1,2,3). In this study, the silica shell thickness, monodispersity and population of SPION@SiO 2 NP formed were used as response variables. The null hypothesis, H 0 , states that the experimental factors (number of mole of TEOS, number of moles of NH 4 OH, and the number of fractionated additions of TEOS) have no effect on the SPION@SiO 2 NP formed. The alternative hypothesis, H a , states that the experimental factors do effect the SPION@SiO 2 NP formed. Following the completion and characterisation of the experimental runs, regression analysis was performed for each of the response variables: shell thickness (described as radius, t shell ), monodispersity (described using E n ), and population (% of SPION@SiO 2 NPs compared to all particles formed). Using relevant statistics including the analysis of variance, lack of fit test, and goodness of fit metrics such as the r 2 and root mean squared error (RMSE), the most suitable models for each response were identified. Using the model, a regression equation was generated for each response. A factor was considered active when p ≤ 0.05. Terms greater than this were deemed inactive and were iteratively removed from the model. 3 Results and Discussion ## Synthesis of SPION NPs Superparamagnetic iron oxide nanoparticles (SPIONs) stabilised with oleic acid were prepared using thermal decomposition. The NPs were highly monodisperse with a size of 12 ± 2 nm (as measured from TEM analysis, Fig. 1, ESI †) and χ m of 38 emu.g −1 at 20,000 Oe, normalised against total mass of particle, and confirmed to be superparamagnetic (Fig. 1, ESI †). While hydrophilic SPIONs usually report a χ m of approx. 70 emu.g −1 at 20,000 Oe, in this case the χ m is lower due to the presence of non-magnetic oleic acid contributing to the sample mass. 6,83 ## Experimental Design of Synthesis of SPION@SiO 2 NPs SPIONs were coated with silica to form SPION@SiO 2 NPs, using a reverse microemulsion method, as described in the Experimental Section. The formation of a silica coating on NPs using reverse microemulsion is reliant on a number of factors, 29,56, 65,84 with the concentrations of the silica source and the basic catalyst being reported as the most important in controlling shell thickness, presence of undesirable (non-magnetic) by-products, and uniform populations. The SiO 2 shell thickness, for example, can be controlled by altering the number of moles of TEOS, with low concentrations resulting in ultra-thin 2 nm SiO 2 shell radii, due to the small amount of silica precursor available for coating. 61 A similar effect can be achieved by changing the number of moles of NH 4 OH and/or the ratio of NH 4 OH-to-surfactant. In the former, adjusting the quantity of NH 4 OH affects the rate of silica hydrolysis and nucleation, where higher concentrations increase the rate of TEOS hydrolysis and increases the shell thickness; however this can also encourage the formation of non-magnetic silica by-product. 29 The ratio of NH 4 OH-to-surfactant, on the other hand, determines the size and number of micelles formed, influencing the ultimate size of particles formed (since the coating step occurs within the micelle). 29,60 It has also been found that the fractionated addition of TEOS can reduce particle polydispersity, simultaneously reducing the formation of by-product SiO 2 . 29 Despite their identification as important experimental conditions in the formation of SPION@SiO 2 NPs, these parameters have not been investigated through a DoE approach, which allows for the simultaneous variation of experimental factors. 71, This not only enables the evaluation of interactions between these factors, but also reduces the impact of random variation. Furthermore, by implementing 3 levels (low, centre, high), non-linear effects can also be measured. The resulting model can be manip- ulated to maximise a response, such as the size of core@shell NPs. Therefore, in the study herein, the numbers of moles of TEOS and NH 4 OH, and fractionated additions of TEOS were investigated as primary variables of importance in the production of core@shell particles. The responses of SiO 2 shell thickness, monodispersity, and population (where population is described as the percentage of single-core@shell particles with respect to the total number of particles) were analysed in order to optimise and control the properties of the NPs formed. A 3 3 full-factorial model was employed, and experimental conditions were generated according to Table1. ## 3.3 Response surface analysis: SPION@SiO 2 particle design SPION@SiO 2 NPs prepared according to the conditions outlined by the DoE were characterised using TEM, and PPMS-VSM. TEM analysis determined the NP size, monodispersity, and population of SPION@SiO 2 NPs formed. Figure 3 shows TEM images of the NPs generated at level 1 of number of moles of TEOS (0.16 mmol); TEM images of samples prepared at TEOS levels 2 and 3, and centre point conditions can be found in Fig. 2-4, ESI †. Boxplots showing the size distribution of the measured total diameter (d TEM ) of the core@shells produced in this study can be observed in Fig. 5, ESI †. Response surface analysis was used to explore relationships and interactions between experimental factors and their effect on the response variables (Table 2 and Table S1, ESI †). Across all particles produced, the d TEM of particles ranged from 15 nm to 75 nm; the thickness of the silica shell (t shell ) ranged 2 nm to 32 nm; the normalised entropy (E n ) ranged 0.15 to 0.65 (characterising all particles to be either monodisperse or near monodisperse (vide infra)); and a single-core@shell population of 35 % to 99 % was observed. The SPION@SiO 2 were found to be weakly superparamagnetic; the mass susceptibility plots can be found in Fig. 6, ESI †. ## Shell thickness of SPION@SiO 2 nanoparticles The silica shell thickness, t shell , of the SPION@SiO 2 NPs was determined from TEM images, Fig. 3. 81 It was observed that as the level of the experimental factors increased, the shell thickness of the particles also increased. Nanoparticles produced using lower levels of the factors (i.e. treatments 111 to 122) were observed to have an ultra-thin silica shell, however the particles themselves appeared embedded within a non-uniform aggregated silica matrix, lacking discrete individual core@shell structures. It is thought that when a lower number of moles of TEOS (0.16 mmol) and NH 4 OH (0.13 mmol) and fewer fractionated additions of TEOS are used, the formation of by-product silica is Fig. 3 TEM images of SPION@SiO 2 NPs produced using conditions determined from the 3 3 factorial model, at level 1 of the number of moles of TEOS (0.16 mmol). Scale bar is 50 nm. The x-axis (row) is the number of fractionated additions of TEOS (Frac. Add. TEOS), which is increasing from left to right in levels of fractionated additions of TEOS, from levels 1 to 3. The y -axis (column) is the number of moles of NH 4 OH, and is increasing from top to bottom in levels 1 to 3 for the number of moles of NH 4 OH. The z-axis is the number of moles of TEOS. Each 3 digit code is the xyz coordinates for each treatment condition generated from the experimental domain, as described in Table 1. The remaining TEM images, at TEOS levels 2 and 3 can be found in Fig. 2-3, ESI †. Additionally, repeats of the centre point condition (treatment 222) can be found in Fig. 4, ESI †. Table 3 Parameter estimates that were deemed to be active in effecting the silica shell thickness of SPION@SiO 2 nanoparticles formed. The parameter estimates were calculated from the regression analysis and used in the prediction expression. In the case below, the fractionated addition of TEOS is included as a parameter, regardless of its p-value, due to its presence in the interaction term. Note that the t-ratio is the estimate divided by the standard error. If ≥ 1.96 (absolute value) the parameter is statistically significant. If the absolute value is < 1.96 the parameter is not statistically significant. The p-value is the probability of the null hypothesis (H 0 ) being true. The lower the p-value the less likely the H 0 is true. Herein, the null hypothesis assumes the experimental factors have no impact on the outcome. OH reduces the aqueous domains present which reduces the micelle size and quantity, therefore limiting places for shell growth to occur. Fewer fractionated additions of TEOS increase the amount of TEOS within each addition available for hydrolysis. With more hydrolysed TEOS present, there is competition between the formation of silica byproduct and growth of the silica shell. Due to the limited micelles available, this increased presence of hydrolysed TEOS therefore favours the formation of silica by-product. As such, these conditions encouraged the formation of non-defined core@shell structures, (as seen in the top row of Fig. 3). As the number of moles of NH 4 OH and fractionated additions of TEOS increased, more defined core@shells were formed, and increasing t shell was observed. ## Parameter The prediction expression of t shell , as determined by regression analysis, is given by: Where α is number of moles of TEOS, β is the number of moles of NH 4 OH and γ is the number of fractionated additions of TEOS. The parameter estimates for the equation can be seen in Table 3. The model was found to agree well with the data with an r 2 of 0.74 and RMSE of 5.00 (Table 2). Analysis of variance (ANOVA) demonstrated the model was highly significant with a p-value of < 0.0001. As such, the model explains the variance in the outcome, and H 0 is rejected. A plot of actual vs. predicted response for shell thickness, determined from the regression analysis, is presented in Fig. 4a and demonstrates the success of the model. The shell thickness was observed to be linearly dependent on the number of moles of TEOS. The number of moles of NH 4 OH similarly impacted the size. Additionally, an interaction between the number of moles of NH 4 OH and the number of fractionated additions was identified. It is important to note here, that the fractionated addition of TEOS alone did not have a significant effect (p-value = 0.9263), however it is included due to its presence in the interaction term. A larger mean shell thickness was achieved for higher number of moles and more additions, while the inverse produced smaller particles. These relationships are summarised in Equation 2 and the contour maps presented in Fig. 4b-d. Overall, it is seen that increasing each of the terms increases the size of particles: increasing the number of moles of TEOS means more material present to form silica shells; increasing the number of moles of NH 4 OH increases the presence of aqueous domains, resulting in larger micelles as the number of moles increases; and increasing fractionated additions favours the growth of the silica shell around the SPION core. These observations correlate with literature descriptions of silica shell growth, 29 however this study illustrates and quantifies the relationship of the interaction between NH 4 OH and fractionated addition of TEOS for the first time. By combining the number of moles of each parameter and alternating the fractionated additions of TEOS, exceptional control over shell thickness can be achieved through careful manipulation of experimental conditions. It is important to clarify that at 8 fractionated additions of TEOS between 0.1 to 0.3 mmol of NH 4 OH and 0.25 and 0.75 mmol of TEOS, it is predicted that no coating of the SPION core would occur. This is a limitation of the experimental design model, where the resulting models are often less reliable at the extremities of the experimental domain. In practice, these conditions are likely to produce SPION embedded in a silica matrix, similar to those formed at treatments 111, 112 and 113, for reasons discussed previously. Table 4 Parameter estimates that were deemed to be active in effecting the monodispersity of SPION@SiO 2 nanoparticles formed, described as nanoparticle entropy (E n ). The parameter estimates were calculated from the regression analysis and used in the prediction expression. In the case below, the fractionated addition of TEOS is included as a parameter, regardless of its p-value, due to its presence in the interaction term. ## Monodispersity of SPION@SiO 2 nanoparticles The monodispersity of the NPs, as determined from nanoparticle size distribution was assessed using nanoparticle entropy (E n ). 82 Size distribution was analysed using a modified Shannon entropy, equation 1. By using this approach, a system can be described as either highly monodisperse, monodisperse, near-monodisperse, or polydisperse, based on the normalised nanoparticle entropy, E n . If E n falls in the range 0 to 0.125, it is classified as highly monodisperse, it is monodisperse if E n is between 0.125 and 0.206, it is near-monodisperse between 0.206 and 0.618, and anything above 0.618 is classified as polydisperse. 82 Following this classification, all SPION@SiO 2 NPs formed were classified as monodisperse or near-monodisperse, excluding treatments 111 and 113, which were classified as polydisperse. Treatments 112 and 331 were excluded from analysis due to an insufficient sample size. The prediction expression of E n , as determined by regression analysis, is given by: Where α is number of moles of TEOS, β is the number of moles of NH 4 OH and γ is the number of fractionated additions of TEOS. The parameter estimates for the equation are shown in Table 4. The model was found to agree well with the data, with an r 2 of 0.68 and RMSE of 0.089 (Table 2). ANOVA demonstrated the model was highly significant with a p-value of <0.0001, illustrating again that, overall, the experimental factors directly impact the monodispersity of the produced particles, hence, H 0 was rejected. A plot of actual vs. predicted response of particle dispersity, determined from the regression analysis, is presented in Fig. 5a. The E n was observed to be negatively dependent on the number of moles of TEOS; with increasing number of moles, there was a decrease in E n . Similar behaviour was observed for NH 4 OH. An interaction between the number of moles of NH 4 OH and the number of fractionated additions of TEOS was identified. As discussed for the shell thickness, the fractionated addition of TEOS term was also included in the model, due to the presence in the interaction term, despite not exhibiting a significant effect itself (p-value = 0.821). Monodisperse populations were achieved at higher numbers of moles and more additions, while the inverse produced populations classified as near-monodisperse. These relationships are summarised in Equation 3 and the contour maps presented in Fig. 5b-d. It is of interest to note that the terms active in effecting E n were also active in influencing t shell . This is due to the commonality between the mechanism, outlined in the previous section; namely that the number of moles of NH 4 OH influences the size of the micelle formed and is used to hydrolyse TEOS. Using fractionated additions of TEOS influences the silica shell growth or the formation of silica by-product. By controlling the micelle size (where shell growth occurs) and rate of hydrolysis of TEOS, it is possible to create monodisperse particles; as showcased by the interaction of number of moles of NH 4 OH and fractionated additions of TEOS. Monodispersity of particle population is vital in the consideration of such materials for biomedical applications and hence this observation is important, as it clearly demonstrates how both the size and dispersity of particles can be carefully controlled through manipulation of experimental conditions. ## Population of SPION@SiO 2 nanoparticles formed Population of SPION@SiO 2 nanoparticles refers to the percentage of the total measured population of particles which exist as single-core@shell particles. The regression analysis of population of SPION@SiO 2 NPs was performed using the data collected from TEM analysis (Table 2 and Table S1, ESI †). The regression model correlation coefficient (r 2 ) was 0.32, indicating that the experi-Table 5 Parameter estimates deemed to be active in effecting the population of SPION@SiO 2 formed. The parameter estimates were calculated from the regression analysis and used in the prediction expression. In the case below, the number of moles of TEOS and number of moles of NH 4 OH are still included as a parameter they are active in the (TEOS-1.43 The prediction expression for population, as determined by regression analysis, is given by: Pop. = 51.97 + 1.80 Where α is number of moles of TEOS, and β is the number of moles of NH 4 OH. The parameter estimates for the equation can be seen in Table 5. Across all samples prepared, there was a mean population of 53 % and the RMSE was 18.94, Table 2. ANOVA indicated that the model was significant as the p-value was 0.014 and the F-value was 4.25, meaning H 0 was rejected. A plot of actual vs. predicted population responses, determined from the regression analysis, is presented in Fig. 6a. Here the high experimental error can be observed, from the scattering around the regression line of fit, and wide confidence bands. The population of SPION@SiO 2 was observed to be dependent on the interaction between the number of moles of TEOS and the number of moles of NH 4 OH. The term was estimated to have a negative effect on the population if the number of moles of both simultaneously increased or decreased. If the number of moles of either TEOS or NH 4 OH increased, while the other decreased, an increase in population of SPION@SiO 2 would be observed. These trends can be found in the regression equation 4 and contour-map in Fig. 6b, where the contour planes are curved and the map has a paraboloid-structure. It is interesting to note that the fractionated addition of TEOS was determined to have no effect on the population of core@shell particles formed, contrary to what is reported in literature. 85 Instead, we found that fractionated additions of TEOS contributed only to the shell growth and monodispersity and hence the mechanism of growth, rather than population of SPION@SiO 2 formed. ## Optimisation of SPION@SiO 2 NPs Using the regression analysis for each of the response variables, the models shown herein should allow the production of single- Table 6 Conditions of validation experiments with predicted (pred.) and observed (obs.) responses, guided by prediction expression of regression analysis. Note that the predicted error was taken from the prediction profiler, from JMP software, which uses the prediction expression to predict the outcome of a response using the conditions outlined above. The observed error was calculated from the standard deviation of each sample set. core@shell NPs with controlled shell thickness, good monodispersity and with high populations. In order to test the models, the prediction expressions were combined and three validation experiments were conducted, aiming to produce SPION@SiO 2 NPs with total sizes of 50 nm, having a shell thickness of 18 nm, classified as monodisperse, and with a high population of core@shell particles. For these studies, the experimental conditions were taken at different coordinates of the experimental design, as seen in Table 6 and Fig. 8, demonstrating that different experimental conditions may lead to similar particles being produced. Size distribution of the measured nanoparticle diameter from TEM (d TEM ), can be found in Fig. 7, ESI. The mass susceptibility of the nanoparticles is given in Fig. 6, ESI. The silica shell thickness of the three validation experiments were found to be 14±3 nm, 28±3 nm, and 22±3 nm, respectively, as seen in Fig 7 . Run 1 and 3 were in good agreement with the predicted thickness, as observed in Fig. 8b, whereas run 2 was larger than anticipated. There was strong agreement between the predicted and observed nanoparticle entropy, with all particles observed to be monodisperse. Furthermore there was also strong agreement between the predicted and observed population, which achieved populations of SPION@SiO 2 >50 % for the three samples. Run 1 used 2.69 mmol of TEOS, across 3 fractionated additions, using 0.40 mmol NH 4 OH. This supports the regression analysis for each of the responses: the silica shell thickness was dependent on the number of moles of TEOS and the combined interaction of number of moles of NH 4 OH and fractionated addition of TEOS, where the shell thickness increased as one of these terms increased. The particle dispersity was also dependent on the aforementioned factors, which decreased as the terms increased (becoming more monodisperse). Section 3.3.3 described the population of SPION@SiO 2 to be dependent on the interaction of TEOS number of moles and the number of moles of NH 4 OH, where an increase in population when one number of moles was high, and the other number of moles was low resulted in high population yield, reflective of the conditions used in run 1. Similar observations were also observed for run 2 and 3, which used higher number of moles of NH 4 OH (1.16 mmol), and lower number of moles of TEOS (0.98 mmol and 0.63 mmol, respectively) across more fractionated additions of TEOS (7 and 8 respectively). Interestingly here, particles of similar morphology, size and characteristics have been produced using different experimental conditions. These validation experiments demonstrate that DoE is valuable for the optimisation of this nanoparticle synthesis and allow flexibility for users to achieve their desired optimised NP. DoE allows accurate prediction across a variety of parameter combinations which can be of use in scenarios where experimental set ups are limited, or can aid in considering scaleup of particle synthesis. ## Conclusions Regression analysis of SPION@SiO 2 indicated that the number of moles of TEOS, number of moles of NH 4 OH and fractionated addition of TEOS were important parameters in determining the size, monodispersity and overall quality of the population of particles formed. Throughout each of the different response models, the experimental parameters of TEOS and NH 4 OH number of moles were statistically relevant, regardless of the nature of their effect. On the other hand, the number of fractionated additions of TEOS was found to be significant only as part of an interaction effect in combination with either the number of moles of TEOS or number of moles of NH 4 OH. The cause of these dependencies lies in the mechanism of the growth of NPs. SPION@SiO 2 NPs were synthesised via a reverse microemulsion method, as seen in Fig. 1 and discussed in section 2.2.2. Through understanding the reaction mechanism, it is clear why the regression analysis determined that the experimental factors were highly dependent on each other. TEOS is clearly needed for the growth of silica shell, hence the strong linear effect for shell thickness, and monodispersity of particles formed. The NH 4 OH number of moles is responsible for both the hydrolysis of TEOS and the quantity and size of aqueous domains (hence micelles present), therefore also exhibiting a linear effect in controlling shell thickness and monodispersity. With these interactions in mind, it is clear why the population of SPION@SiO 2 NPs is dependent on both the number of moles of TEOS and NH 4 OH. These trends were identified and modelled through the use of DoE. From understanding these findings and their interplay in the mechanism of core@shell particle formation, they can be applied to optimise the characteristics of desired SPION@SiO 2 NPs, or can be used to match specifications for a given application or experimental setup. For example, monodisperse SPION@SiO 2 NPs with a yield greater than 70 % which are 50 nm in size could be synthesised using low number of moles of NH 4 OH, and high number of moles of TEOS, over 6 fractionated additions. If altering the size of NPs are of interest, on the other hand, the protocol could be tuned through altering the amount of TEOS and NH 4 OH used, and using 3 fractions of TEOS. Alternatively, the (population) yield and monodispersity of NPs may be increased through using 6 fractionated additions, instead of 3 or 8. This approach has therefore yielded exploitation of the mechanism of formation of the particles to produce a desired goal. It must be emphasised, from the model there are a number of possible routes to the desired result, as observed from the validation experiments, which all produced monodisperse SPION@SiO 2 NPs that had a desired shell thickness of 18 nm, at high populations. The synthesis of NPs can be a complicated and demanding process to understand. Through using DoE, the intricate reaction process can be studied and modelled, allowing for the re-action outcome to be predicted in relation to the experimental domain used. In this study, the synthesis of SPION@SiO 2 NPs through a microemulsion method has been modelled using a 3 3 full-factorial design. Following regression analysis, the silica shell thickness was found to be linearly dependent on the number of moles of TEOS, and the interaction between the number of moles of NH 4 OH and fractionated additions of TEOS, whereby increasing these terms were observed to increase the size of particles formed. Similarly the nanoparticle dispersity was dependent on the linear effect of the number of moles of TEOS, and the interaction between the number of moles of NH 4 OH and fractionated additions of TEOS. In this case, the increase in one of these terms resulted in the reduction in nanoparticle dispersity. The population of SPION@SiO 2 NPs was effected by the interaction between TEOS and NH 4 OH, and not impacted by the fractionated addition of TEOS. The complexity of the model was reflective of the synthesis mechanism, where each of the reagents hold multiple roles that are dependent on each other in controlling the properties of the produced NPs. Through using a DoE approach, these underlying trends were identified and modelled and could be used for the optimisation of or tailoring of SPION@SiO 2 NPs with controlled properties.

chemsum

{"title": "Controlled synthesis of SPION@SiO 2 nanoparticles using design of experiments", "journal": "ChemRxiv"}

compatible_ferroelectricity,_antiferroelectricity_and_broadband_emission_for_a_multi-functional_2d_o

## Abstract: Two-dimensional (2D) organic-inorganic hybrid perovskites with multifunctional characteristics have potential applications in many fields, such as, solar cells, microlasers and light-emitting diodes (LEDs), etc. Here, a 2D organic-inorganic lead halide perovskite, [Br(CH2)3NH3]2PbBr4 (BPA-PbBr4, BPA = Br(CH2)3NH3, 3-Bromopropylamine), is examined for its photophysical properties. Interestingly, BPA-PbBr4 reveals five successive phase transitions with decreasing temperature, including successive paraelectric-ferroelectric-antiferroelectric phases. Besides, BPA-PbBr4 displays ferroelectricity and antiferroelectricity throughout a wide temperature range (<376.4 K) with accompanying saturation polrization (Ps) values of 4.35 and 2.32 μC/cm 2 , respectively, and energy storage efficiency of 28.2%, and also exhibits superior second harmonic generation (SHG) with maximum value accounts for 95 % of the standard KDP due to the great deformation of structure (3.2302*10 -4 ). In addition, the photoluminescence (PL) of the BPA-PbBr4 exhibits abnormal red-shift and blue-shift in different phases due to a consequence of competition between electron-phonon interaction and the lattice expansion. Further, BPA-PbBr4 reveals a broadband emission accompanied by bright white light at room temperature (293 K), which is supposed to be due to self-trapped excitons. In short, the versatility of BPA-PbBr4 originates from molecular reorientation of dynamic organic cations, as well as significant structural distortion of PbBr6 octahedra. This work paves an avenue to design new hybrid multifunctional perovskites for potential applications in the photoelectronic field. ## ■ INTRODUCTION Electroactive substances, as an indispensable basic component of advanced electronic devices, have the ability of energy stroage and conversion. Ferroelectrics are one of the important electroactive substances, which can realize spontaneous polarization (Ps) storage and switching. 3 The appearance of ferroelectricity leads to miscellaneous optoelectric effects, while the coupling of ferroelectric polarity with other physical properties brings new concepts for electronic and optoelectronic applications. Additional, antiferroelectrics have achieved an indispensable status in electroactive substances due to their inherent advantages of large storage capacity, which reveals a polarization versus electric field (P−E) double hysteresis loops derived from the switching of antiparallel dipoles when subjected to a strong electric field, thus they are become one of the most promising candidates for high performance capacitors with high-energy storage density and fast discharging rates. Significant efforts have been focused on pushing the boundaries of ferroelectric and antiferroelectric design through deepening the understanding of structure-property effects. Engineering crystal structures of low-dimensional (0D to 2D) perovskites by employing suitable organic ammonium cations is the predominant methods for the tuning of structure and physical performance. As a booming multifunctional materials, two-dimensional (2D) perovskites, which feature natural quantum-well structures formed by homogeneous integration, 11 alternating and periodic arrangement of semiconducting inorganic layers and capped organic layers at a molecular level, are not subject to the Goldschmidt's factor, that is, they relax structural constrains, and have a wide range of selectivity for the larger, high-aspect ratio, 12 and potentially functional organic cations, 13 therefore, they become an ideal choice for ferroelectrics and antiferroelectric. In addition, they also show superior luminescence performance, which inspires further rational structural and property optimization to realize the desirable performance, thus they are widely used in the light-emitting diodes (LEDs), solar cells, lasers and other fields. Generally, the emergence of ferroelectricity or antiferroelectricity is inseparable from phase transitions, changing from a highsymmetry paraelectric phase to a low-symmetry ferroelectric phase, or from a high-symmetry paraelectric phase to a low-symmetry antiferroelectric phase. However, it is scare to achieve a sequence of changes in the identical 2D perovskite from high-symmetry paraelectric phase to a low-symmetry ferroelectric phase to an even lower-symmetric antiferroelectric phase. So far, only a few 2D perovskites have been reported to have such successive ferroelectric-antiferroelectric-paraelectric phase transitions, for example, (BA)2(EA)2Pb3I10 (BA = nbutylammonium and EA = ethylammonium) and (3pyrrolinium)CdBr3. Herein, we have achieved a successive antiferroelectric-ferroelectric-paraelectric transformation covering a wide temperature range in a 2D hybrid perovakite [Br(CH2)3NH3]2PbBr4 (BPA-PbBr4), accompanied by a prominent saturation polrization (Ps) value and a signifcant SHG effect comparable to the standard KDP. Meanwhile, the PL characteristics realized the regulation of broadband emission from blue to white to violet light with temperature stimulation and accompany with the lifetime of microsecond level. This work sheds light on the design of new ## ■ RESULTS AND DISCUSSION The thermodynamic and structural reversibility of BPA-PbBr4 were detected by thermal analysis. As shown in Figure 1, the heat flow curves display five pairs of abnormal endothermic and exothermic peaks during heating and cooling runs, respectively, indicating an occurrence of five phase transitions for BPA-PbBr4. The sharp phase transition peaks between heating and cooling process from high temperature to low temperature indicating that all the peaks exhibited first order phase transition feature except the second order phase transition indicated by the slow peak at 225.7 K. We label the phase below 149.1 K as low-temperature (LT) phase Ⅰ, the phases from 149.1 to 170.2 K and from 170.2 to 225.7 K and from 225.7 to 376.4 K and from 376.4 to 412.0 K as intermediate-temperature phasesⅡ~Ⅴ(IT1 ~ IT4), respectively, and the phase above 412.0 K as the high-temperature (HT) phase Ⅵ. Variable temperature single crystal X-ray diffractions were performed to determine the evolution of crystal structure of each phase with temperature. The crystal structures of BPA-PbBr4 consist of infinite layers of corner-sharing PbBr6 octahedra charge-compensated by the bilayered organic [Br(CH2)3NH3] + cations, forming a 2D organic-inorganic layered structure (Figure 2). What's interesting is that the first four phases (Ⅰ~ Ⅳ) of BPA-PbBr4 with temperature below 376.4 K crystalize in noncentrosymmetric polar space groups, while the other two high temperature phases ( Ⅴ ~ Ⅵ ) belong to centrosymmetric space groups. Concretely, at LT phase Ⅰ (130 K), the symmetry of BPA-PbBr4 depresses to an orthorhombic antiferroelectric space group Pca21 with cell parameters of aLT = 27.1474 (8), bLT = 16.9201 (5), cLT = 7.9011(2) and VLT = 3629.3(2) 3 (Table S1). Both the crystallographically independent Pb1 and Pb2 ions are sixcoordinated and linked together by sharing one bromine ion. Each ordered [Br(CH2)3NH3] + cation is hydrogenbonded to the PbBr6 octahedra by four hydrogen bonds with N•••Br distances in the range of 3.3425(204)-3.5139(197) (Table S2), and the corresponding N-H•••Br angles are in the range of 121.26(126)-179.28(103) o (Table S3). Meanwhile, PbBr6 octahedra show a clearly distorted configuration as indicated by the discrepant Pb1-Br bond distances (2.9252(26)-3.0726(26) ) and Br-Pb1-Br bond angles (82.35(7)-174.72(8) o ); and the Pb2-Br lengths spannning from 2.9255(28) to 3.0707(26) with corresponding Br-Pb2-Br bond angles in the range of 82.34(7)-174.71(8) o ). The magnitude of the structural distortion (Δd) can be further quantitatively estimated based on the Pb-Br bond lengths by the following equation: Δd = ( )Σ[ where 𝑑 ̅ is the mean Pb-Br distance and dn is the six individual Pb-Br distances. Calculated Δd is 3.2302*10 -4 for Pb1Br6 and 3.1711*10 -4 for Pb2Br6, which is two orders of magnitude higher than the congeneric hybrid perovskites (C4H9NH3)2PbCl4 (4.3447*10 -6 ) 26 and (2meptH2)PbBr4 (2mept = 2-methyl-1,5-diamino-pentane) (4.3447*10 -6 ), 27 indicating that PbBr6 octahedron is significant distorted with Pb atoms deviating from the balanced sites (Figure S1a). In addition, the Pb-Br-Pb angles (157.543 and 157.531 o for θ1 and 138.389 o for θ2 (Tables S2-3)) exhibit large deviation compared with the ideal 180 o . At the same time, the two adjacent octahedrons rotated 61.929(59) o relative to each other, further proofs that the framework is highly distorted (Figure S1b-c). An important characteristic is that the amino groups on the [Br(CH2)3NH3] + cations are arranged in reverse parallel along the a-axis leading to total polarization to zero, which is shown by the arrows in the Figure 2a. Such a remarkable structural distortion synergistically leads to local polarization generation together with the ordering dynamic of organic cation. As the temperature increases to phase Ⅱ (IT1, 160 K), BPA-PbBr4 turns to another orthorhombic antiferroelectric space group Pna21 with cell parameters of aIT1 = bLT = 16.8757(4), bIT1 = cLT = 7.9609(2), cIT1 = aLT = 27.1272(5) and VIT1 = 3644.4(1) 3 (Table S1). Figure 2a depicts that its perovskite motif is still identified as an octahedral tilting architecture, leading to the emergence of local polarization. Identical as LT phase, [Br(CH2)3NH3] + cations are arrangeded in antiparallel is assumed to satisfy S2-3). Meanwhile, Δd is 1.3106*10 -4 , and the relative rotation degree of two adjacent octahedrons is further increased to 62.771(48) o (Figure S1c). Interestingly, the ammonium heads of [Br(CH2)3NH3] + cation still arrangement stretch down the positive c-axis, resulting in the polarization in this positive direction as well in a wide temperature range that spans 150 K (Figure 2b). S1), which is inconsistent with literature report, where reports that the space group is Cmca. 28 We tried to refine this into the space group Cmca, but the data is poor. Ordered [Br(CH2)3NH3] + cations are arranged in reverse to support the disappearance of electric polarization along crystallographic c-axis direction (Figure 2c S2-3). Meanwhile, Δd is 1.9740*10 -5 , which is an order of magnitude smaller than the octahedral distortion of the other phases indicating a lower degree of distortion. The relative rotation degree of two adjacent octahedrons is further increased to 64.908(40) o (Figure S1c), suggesting a minimum distorted geometry. As the temperature goes up and there's a big phase transition at 412 K (phase Ⅵ, HT), however, the acquisition of single crystal data fails due to the high temperature. Other physical properties have been measured to prove that phase Ⅵ is still in the paraelectric phase. All the bond lengths and bond angles as well as Δd increase as the temperature increases, indicating that the distortion of the structure decreases from phases Ⅰ to Ⅴ. The emergence of polarization relates to the reorientation of organic [Br(CH2)3NH3] + cations and tilting of inorganic PbBr6 octahedrons, which contributes to the driving force of the phase transition. During this series of phase transitions (Figure 2d), symmetry breaking occurs accompanied with an Aizu notation of mmmFmm2Fm from phase Ⅴto phase Ⅰ according to Landau theory, namely, eight symmetric elements (E, C2, 2Cʹ2, i, σh, 2σv) at phase Ⅴ are halved into four (E, C2, 2σv) at phases Ⅲ and Ⅳ and are further reduced by a quarter into two (E, σh) at phases Ⅰ and Ⅱ (Figure 2e). 19, Since front four phases Ⅰ~ Ⅳ of the BPA-PbBr4 belong to the non-centrosymmetric polar ferroelectric and antiferroelectric space groups, variable temperature SHG and polarization-electric field (P-E) hysteresis loop measurements are used to further verify the correctness of the structures. Figure 3a standard potassium dihydrogen phosphate (KDP) at 293 K. The results indicate that the SHG intensities of BPA-PbBr4 account for 64.9, 93.9 and 95.4 % of the KDP at 293, 150 and 120 K, respectively, demonstrating the SHG intensity of the BPA-PbBr4 is equivalent to that of the KDP in the LT and IT phases, which is attributed to the significant structural distortion of PbBr6 octahedra. Therefore BPA-PbBr4 can be widely used as a superior nonlinear material with good purity (Figure S2). However, SHG signals tend to zero in phases Ⅴ and Ⅵ, which is consistent with the assignment of the centrosymmetric structures in these phases. In addition, the measurement of P−E hysteresis loop is one of the most direct methods to determine the ferroelectricity of materials. Thus the P−E hysteresis loops of BPA-PbBr4 in six phases were measured. Results indicate that the polarization response to the applied field is linear in phases Ⅴ and Ⅵ, indicating that these phases belong to paraelectric phase. However, P−E hysteresis loops at 293 and 184 K behave as ferroelectric phases and accompanying saturation polarization (Ps) values are 4.35 and 2.71 μC/cm 2 , respectively which is consistent with the literature reported (4.8 μC/cm 2 ) 28 and calculated value (4.88 μC/cm 2 ) according to a point charge model (Figure S3). Such spontaneous polarization is consistent with those in other analogous organic-inorganic hybrid perovskite ferroelectrics, such as, (CHA)2PbBr4−4xI4x (x = 0-1) (CHA = cyclohexylammonium) (Ps = 5.8, 8.5, and 7.5 µC cm -2 when x = 0, 0.1125, and 0.175, respectively), 31 and significantly less than some organic ferroelectrics (20 ~ 55 μC/cm 2 ) and hybrid ferroelectrics (C2H5NH3)2CuCl4 (37 μC/cm 2 ) and (MV)[BiI3Cl2] (MV 2+ = methylviologen) (80 μC/cm 2 ). What makes BPA-PbBr4 more intriguing is the emergence of antiferroelectricity as temperature drops evidenced by the notable double P-E hysteresis loops in two low temperature phases, along with the polarization of 2.25 and 2.32 μC/cm 2 , respectively. Such a curious phenomenon contradicts the Kittel prediction that antiferroelectrics generally crystallize in centrosymmetric space group without remnant polarization and SHG response. However, this unusual feature has also been observed in K3Nb3B2O12 (KNBO) and TCMBI. Meanwhile, we notice that the cell volume of BPA-PbBr4 in antiferroelectric phases Ⅰ and Ⅱ increase to about twice that of ferroelectric and paraelectric phases Ⅲ ~Ⅴ, which is an important microstructure feature to judge antiferroelectric materials. To the best of our knowledge, these polarization values are a b the slightly less than the antiferroelectrics reported so far, including (BA)2(EA)2Pb3I10 (5.6 μC/cm 2 ), ((CH3)2CHCH2NH3)2CsPbBr7 (6.3 μC/cm 2 ) and (3pyrrolinium)CdBr3 (7.0 μC/cm 2 ). 39 Besides, BPA-PbBr4 enables an evident energy storage efficiency of 28.2%, making it a potential candidate for energy storage materials. Furthermore, we have measured the antiferroelectric double P-E hysteresis loops of phases Ⅰ~ Ⅱ at various temperatures during the cooling (Figure S4a) and heating (Figure S4b) sequences respectively, and ferroelectric P-E hysteresis loops of phases Ⅲ and Ⅳ(Figure S4c) at various temperatures during the heating sequence. Results indicate that Ps in antiferroelectric phases increases slightly as temperature decreases during cooling in the range of 108-164 K, but remains almost unchanged during heating in the range of 117-168 K, while Ps increases with the raise of temperature in the whole ferroelectric phases in the range of 197-369 K. Depending on the excellent ferroelectric and antiferroelectric properties of BPA-PbBr4, and Pb-based 2D organic-inorganic hybrid perovskites are expected to have potential semiconductiong properties, UV-Vis absorption spectrum at room temperature was carried out to investigate the electronic structures and photophysical properties of BPA-PbBr4. Result reveals a steep absorption edge at 410 nm, indicating BPA-PbBr4 is a direct band gap semiconductor with corresponding band gap of 2.998 eV by fitting absorption curve with Tauc equation, 40 which is consistent with the literature report. 28 In addition, a shoulder peak appears at around 400 nm in the absorption spectrum, indicating BPA-PbBr4 has distinct excition features near the absorption edge, where the low dielectric constant of the organic layer and 2D structure of the inorganic layers leads to enhancement of the attraction between the electron and hole in an exciton. To uncover the electronic origin micro-mechanism of band gap, we calculated the electronic band structures and partial density of states (PDOS) of BPA-PbBr4 based on density functional theory (DFT) calculations. As shown in Figure 4b, BPA-PbBr4 shows a direct band gap at the Brillouin zones G point with energy value of 2.937 eV, which is slightly lower than the experimental value and that's because the known generalized gradient approximation (GGA) functional underestimates the band gap. Meanwhile, PDOS plot shows that the valence band maximum of BPA-PbBr4 except dominated by Br-4p orbitals, and bits of Pb-6s, N-2p, C-2p and H-1s are also involved, while the Pb-6p orbital spans the entire band gap and conduction band minimum, which indicates that the inorganic PbBr6 octahedron is responsible for band gap (Figure S5). These results are obviously identical to the most of the reported metal-halide perovskites. High distortion of lead-halide octahedron in organicinorganic hybrid pervoskites not only contributes to the prominent ferroelectricity, but also associates with the great potential of photoluminescence (PL) emission, which induces the self-trapped excitons generaged from recombination of excitons-hole pairs through strong electron-phonon coupling. PL spectra of BPA-PbBr4 at six phases are measured and shown in Figure 5a, results express that the emission wavelength centers are 404 and 405 nm at 130 and 160 K, respectively, with the full width at half-maximum (FWHM) are 17 and 18 nm, respectively, which is almost unchanged in the two low temperature phases (Ⅰ and Ⅱ). However, the emission wavelength shows a broadband emission with center wavelength redshifted to 431 nm at 293 K (phase Ⅳ), which proves that the emission peak position red-shifted 27 nm due to temperature stimulation. Meanwhile, the FWHM is broaden to 80 nm at 293 K, which is attributed to the increase of thermal filling vibration dynamics at high temperature. Oddly enough, as the temperature rises to the phases Ⅴ and Ⅵ, the emission wavelength centers blue-shifted back to 373 and 372 nm with FWHM of 38 and 20 nm, respectively. The temperature dependence of the PL emission wavelength is a consequence of both the electron-phonon interaction and the lattice expansion. Generally, the lattice expansion leads to a red-shift with increasing temperature, while the electron-phonon interaction causes a blue-shift. Thus, BPA-PbBr4 occurs a red-shift between phases Ⅰ and Ⅳ, where the influence of lattice expansion exceeds the contribution of the electron-phonon interaction, due to the lattice exhibits a nonlinear expansion. On the contrary, the electronphonon interaction becomes the dominating process leading to a blue-shift in phases Ⅴ and Ⅵ attributed to the lattice expansion slows down and becomes linear. As shown in the image inserted in Figure 5c, BPA-PbBr4 displays a bright white light in the center surrounded by blue light at 293 K, which has great potential for application in light emitting diode (LED) due to 2D perovskites feature natural quantum-well structures formed by alternating and periodic arrangement of bulky organic layers and inorganic layers. In addition, BPA-PbBr4 exhibits a moderate photoluminescence quantum efficiency of 1.79 % at 293 K, which is comparable to other analogous broadband emmision organic-inorganic hybrid perovskites including (C4H9NH3)2PbCl4 (1%), 26 (N-MEDA = N1-methylethane-1,2diammonium)[PbBr4] (0.5%), 44 but much lower than some similar organic-inorganic hybrid perovskite (N1methylethane-1,2-diammonium)[PbBr4-xClx] and (2,2′ -(ethylenedioxy)bis(ethylammonium))[PbX4] with X = Cl or Br (9 %). 45 This low PLQYs may be attributed to insufficient confinement of Wannier type excitons within the inorganic layers. ## ■ CONCLUSION In summary, five successive phase transitions occur in a 2D organic-inorganic hybrid perovskite, [Br(CH2)3NH3]2PbBr4 (BPA-PbBr4, BPA = Br(CH2)3NH3), accompanied by a series of changes from paraelectric to ferroelectric to antiferroelectric phase transitions, and saturation polarization (Ps) values also change from 0 to 4.35 to 2.25 μC/cm 2 , respectively. Further, BPA-PbBr4 has superior SHG characteristics, accounting for 95 % of the standard KDP. In addition, the photoluminescence (PL) properties of BPA-PbBr4 also undergo a peculiar transformation under the influence of both the electron-phonon interaction and the lattice expansion, which occurs red-shift at the beginning as the temperature increases from 130 to 376 K, and followed by blue-shifted as temperature increases further from 376 to 420 K. Besides, BPA-PbBr4 exhibits a broadband emission with a bright white light at room temperature, accompanied by quantum efficiency is 1.79 %. This discovery paves an avenue to search for multifunctional hybrid perovskites and provides the impetus for further optoelectronic industrial applications. The Supporting Information is available free of charge via the Internet at http://pubs.acs.org. Figures S1-5: the schematic of octahedral distortion, Pb-Br-Pb angles and rotation degree of two adjacent octahedrons in antiferroelectric, ferroelectric and paraelectric phases; powder XRD pattern; distribution of Pb and N atoms in a unit cell; P-E hysteresis loops of antiferroelectric and ferroelectric phases at various temperatures; the calculated partial density of states (PDOS) for BPA-PbBr4. Tables S1-3: crystal data and structure refinement details for BPA-PbBr4; the bond lengths () and bond angles ( o ) of BPA-PbBr4 at different temperature.

chemsum

{"title": "Compatible Ferroelectricity, Antiferroelectricity and Broadband Emission for a multi-functional 2D Organic-Inorganic Hybrid Perovskite", "journal": "ChemRxiv"}

an_environmentally_benign_and_selective_electrochemical_oxidation_of_sulfides_and_thiols_in_a_contin

## Abstract: A practical and environmentally benign electrochemical oxidation of thioethers and thiols in a commercially-available continuous-flow microreactor is presented. Water is used as the source of oxygen to enable the oxidation process. The oxidation reaction utilizes the same reagents in all scenarios and the selectivity is solely governed by the applied potential. The procedure exhibits a broad scope and good functional group compatibility providing access to various sulfoxides (15 examples), sulfones (15 examples) and disulfides (6 examples). The use of continuous flow allows the optimal reaction parameters (e.g. residence time, applied voltage) to be rapidly assessed, to avoid mass-and heat-transfer limitations and to scale the electrochemistry. † Electronic supplementary information (ESI) available. See ## Introduction Sulfoxide and sulfone moieties are widespread in a broad variety of functional organic molecules. 1 These moieties have been incorporated in numerous pharmaceutical compounds (e.g. Esomeprazole, Dapsone, Sulmazole, Methionine sulfone and Ponazuril) 2 and even in polymeric materials 3 (Fig. 1). Moreover, chiral sulfoxides have been employed as chiral auxiliaries (e.g. Ellman's sulfinamide, Oppoltzer camphorsultam) 4 and as chiral ligands in asymmetric transition-metal catalyzed transformations (e.g. Skarzewzki's and Hiroi's ligands, Fig. 1). 5 Typically, sulfoxides and sulfones can be accessed through oxidation of the corresponding thioether. 6 Hereto, a wide variety of oxidizing agents have been used, such as H 2 O 2 in combination with metal catalysts, 7 m-CPBA, 8 NaIO 4 , 9 CrO 3 , 9 KMnO 4 10 and dioxiranes. 11 Unfortunately, these strategies typically suffer from selectivity issues, e.g. overoxidation of the sulfoxide to the sulfone or oxidation of other functional groups within the molecule. And, while hydrogen peroxide is considered to be a green oxidant, its industrial synthesis via the so-called anthraquinone autooxidation process is not sustainable. 12 Fig. 1 Examples of interesting sulfoxides and sulfones, ranging from pharmaceuticals to chiral ligands. Oxidation chemistry can also be achieved using alternative and sustainable technologies, such as photochemistry or electrochemistry, which allow to carry out the desired transformation in the absence of strong oxidants. 13 Here, the desired transformation is induced by so-called traceless reagents such as photons or electrons, providing sustainable alternatives for the often hazardous and toxic oxidants. 14,15 In addition, these methods are relatively mild and provide good functional group tolerance and high chemoselectivity. Furthermore, sustainable electricity, derived from solar and wind energy, is becoming more abundantly available. Due to the transient nature of these energy sources, small scale electrochemical plants are ideally suited to directly harness this sustainable energy source. Despite the advantages provided by electrochemistry, many organic synthetic practitioners have been discouraged to adopt this technology into their laboratories. 16 This is in part due to the apparent complexity of electrochemical transformations which originates from numerous problems, such as the use of specialized equipment and large amounts of tailor-made electrolytes, mass-and heat-transfer limitations, electrodeposition of organic material on the surface of the electrode and limited scalability. However, many of these challenges can be overcome by combining electrochemistry and continuous-flow microreactor technology. 17 Due to the small dimensions (100 μm-1 mm), micro-flow technology allows to intensify the contact between the reaction mixture and the electrodes, to eliminate mass-transfer limitations and to avoid hot-spot formation. 15b,d,17,18 Furthermore, organic deposition on the electrodes can be minimized due to the continuous-flow operation of the reactor. 17 Using a combination of electrochemistry and continuousflow microreactors, we questioned whether we could selectively access either the sulfoxide or the sulfone starting from their corresponding thioethers. 16,19,20 Specifically, we hoped to develop a set of generally applicable conditions which would allow us to access both compounds simply by tuning the applied potential in the electrochemical flow cell, whilst keeping the reaction mixture the same. This would be a great advance compared to traditional synthetic approaches where each product class requires its own set of conditions with often toxic oxidants, metal catalysts and/or elevated temperatures. ## Results and discussion In order to find the optimal conditions for the direct electrochemical oxidation of thioethers, the potentiostatic oxidation of thioanisole (1) was taken as a benchmark (Table 1). The Asia Flux reactor was used as a commercially available electrochemical flow module and was equipped with cheap stainless steel electrodes. At the outset of our investigations, it was found that the use of an electrolyte was crucial to maintain a stable current. Optimal results were obtained using tetrabutylammonium perchlorate (TBAClO 4 ). We also observed that the use of an aqueous acidic solvent was mandatory to lower the pH which allowed to increase the redox potential of the Fe electrodes, as explained by Pourbaix's diagrams (Table 1, entries 3 and 4). 21 Consequently, the combination of tetrabutylammonium perchlorate in MeCN with aqueous HCl (3 : 1 v/v, 0.1 M HCl in H 2 O) with an applied potential of 2.5 V (corresponding to a current of approximately 2.1 mA cm −2 ), resulted in the oxidation of 1 toward sulfoxide 1-A (63% yield; Table 1, entry 1). Increasing the residence time while keeping the flow rate constant resulted in an increase in yield (entry 2, Table 1). Interestingly, an increase in the applied potential clearly shifted the selectivity towards the double oxidized product 1-B (Table 1, entries 7-9). Finally, a control experiment was carried out using an oxygen-enriched acetonitrile solution without additional water (Table 1, entry 6). Oxidation of the thioether was still observed, indicating that O 2 likely serves as the [O]-source. Furthermore, in the presence of water, gas formation can be noticed due to water splitting (see ESI †). With these preliminary results in hand, the effect of the applied potential on the observed selectivity was investigated in greater detail. Hereto, a single sweep voltammetry experiment was performed at a constant residence time of 5 min (Fig. 2A). The polarogram shows a clear plateau at 2.0-2.5 V, 22 which corresponds to the first oxidation step of thioanisole toward 1-A (Fig. 2B). When the applied potential reaches approximately 3.5-4.0 V, another plateau is observed, which corresponds to the second oxidation step (via 1-A → 1-B or 1 → 1-B). Further increase in the potential results into a critical oxidation of the stainless steel electrodes and should be avoided. It is clear that such polarograms represent a very useful tool to establish the optimal reaction conditions to obtain either the sulfoxide or the sulfone from a specific thioether (see Fig. 2B and ESI †). Next, the effect of flow rate and residence time was investigated (Fig. 3). These investigations were carried out to define the optimal flowrate/residence time ratio in order to avoid mass-transfer limitations. At lower flow rates, mixing is not intense enough to facilitate diffusion of the reactants from the bulk to the electrodes. At higher flow rates, the reaction time is too short to allow for complete conversion. From our data, it is clear that a residence time of 7.5 minutes effects the highest conversion. A further increase in residence time results in a drop in yield (Fig. 3, red zone). Hence, an optimum flowrate regime was found to be situated between 0.04 to 0.06 mL min −1 , which corresponds to residence times of 7.5 to 5 minutes, respectively. A residence time of 7.5 minutes gave a slightly higher yield in 1-A, but was accompanied with a low amount of the corresponding sulfone 1-B (<5%). Therefore, a standard residence time of 5 minutes per run was considered optimal and, when full conversion was not achieved within a single run, the reaction mixture was reinjected to increase the overall residence time (see ESI †). Having established insight in the governing parameters in the electrochemical oxidation of sulfides, we set out to probe the generality of our electrochemical flow protocol. Various thioethers bearing different functional groups were subjected to both Methods A and B, yielding the corresponding sulfoxide A or sulfone B, respectively (Fig. 4). For every compound, a fast potential screening was carried out to obtain the polarogram, which allows us to find the optimal condition for each compound (see ESI †). Notably, the productivity of our electrochemical flow protocol is excellent (i.e. 7.8 mmol h −1 for 1-A and 5.5 mmol h −1 for 1-B) providing means to scale this chemistry to quantities which are sufficient for Medicinal Chemistry applications. Further scale-up can be achieved using larger electrochemical flow reactors or via numbering-up of electrochemical microreators. 23 As shown in Fig. 4, a wide variety of aryl, heteroaryl and alkyl thioethers were efficiently converted into their corresponding sulfoxides and sulfones in moderate to excellent yield. Functional groups, such as halides (6, 7, 11), ketones (8), esters (10, 15) and amides (9, 15), were well tolerated under our reaction conditions. Notably, nitrogen-containing heteroaromatic compounds, e.g. pyridine (12) and benzimidazole (13), can be selectively oxidized to produce the corresponding sulfoxides and sulfones without N-oxide formation. However, in the case of benzoxazole ( 14), nitrogen oxidation was observed at higher potentials prior to the formation of the sulfone. Interestingly, biologically relevant compounds such as the amino acid Methionine (15) and a precursor of the antiprotozoal agent Toltrazuril 24 (11) were efficiently oxidized in good yield. However, it is important to note that for some substrates it was not possible to access both the sulfoxide and sulfone. As an example, dibenzothiophene (5) is directly oxidized to the sulfone. It is however plausible that a double oxidation is immediately occurring, since it is known that aromatic sulfoxides are relatively reactive. 25 The Toltrazuril precursor (11) is oxidized only to the sulfoxide form, which can be explained by the strong electron-withdrawing character of the CF 3 group, making it less prone to oxidation. Despite the broad substrate scope, some limitations do exist to our methodology (see ESI †). Free amines, alcohols and carboxylic acids do not yield any product. In addition, thioethers bearing nitriles, aldehydes and hydroxyls at the γ-position underwent retro-Michael reactions. Finally, the oxidation of thiols to symmetric disulfides was performed applying an analogous approach as for the oxidation of thioethers. Also here, a potential screening was carried out after which the proper voltage value was set to perform the reaction (Fig. 5). As a result, simple thiols such as (18), benzylthiol (19) and octanethiol (20), were all converted to the corresponding disulfide in good to excellent yield. Notably, 2-mercaptopyrimidine (21) and 2-furanylmethanethiol (23) could also be converted to their disulfide derivative and were obtained in good yield. It should be noted that disulfide 23-C is prone to overoxidation under photochemical oxidation reactions, however, this was not observed using our electrochemical method. 26 This highlights the efficiency of electrochemistry and its complementarity compared to photoredox catalysis. 27 Furthermore, poorly soluble starting materials like 4-mercaptobenzoic acid (22) could also be converted using this protocol. ## Conclusion A straightforward, green and broadly applicable electrochemical continuous-flow procedure to oxidize thioethers and thiols has been developed. Using a commercially available electrochemical flow setup (Asia Flux), a wide variety of functionalized sulfoxides (15 examples) and sulfones (15 examples) could be accessed selectively, simply by changing the applied voltage. Similarly, aryl and alkyl thiols could be efficiently oxidized to their corresponding disulfides (6 examples). Because of the sustainable nature of our developed protocol, we believe that our method is highly attractive for technical applications.

chemsum

{"title": "An environmentally benign and selective electrochemical oxidation of sulfides and thiols in a continuous-flow microreactor", "journal": "Royal Society of Chemistry (RSC)"}

dynamic_catalysis_fundamentals:_i._fast_calculation_of_limit_cycles_in_dynamic_catalysis

## Abstract: Dynamic catalysis-the forced oscillation of catalytic reaction coordinate potential energy surfaces (PES)-has recently emerged as a promising method for the acceleration of heterogeneously-catalyzed reactions. Theoretical study of enhancement of rates and supraequilibrium product yield via dynamic catalysis has, to-date, been severely limited by onerous computational demands of forward integration of stiff, coupled ordinary differential equations (ODEs) that are necessary to quantitatively describe periodic cycling between PESs. We establish a new approach that reduces, by ≳10 8 ×, the computational cost of finding the time-averaged rate at dynamic steady state (i.e. the limit cycle for linear and nonlinear systems of kinetic equations).Our developments are motivated by and conceived from physical and mathematical insight drawn from examination of a simple, didactic case study for which closed-form solutions of rate enhancement are derived in explicit terms of periods of oscillation and elementary step rate constants. Generalization of such closed-form solutions to more complex catalytic systems is achieved by introducing a periodic boundary condition requiring the dynamic steady state solution to have the same periodicity as the kinetic oscillations and solving the corresponding differential equations by linear algebra or Newton-Raphson-based approaches. The methodology is well-suited to extension to non-linear systems for which we detail the potential for multiple solutions or solutions with different periodicities. For linear and non-linear systems alike, the acute decrement in computational expense enables rapid optimization of oscillation waveforms and, consequently, accelerates understanding of the key catalyst properties that enable maximization of reaction rates, conversions, and selectivities during dynamic catalysis. ## Introduction Experimental (1-6) and theoretical (7)(8)(9)(10) reports have demonstrated that periodic input of thermodynamic work (e.g. by oscillation of applied electric potential) can effect orders-ofmagnitude improvement in catalytic turnover rates and overcome static equilibrium limits to chemical conversion akin to molecular motors/ratchets in biological systems (11,12). So-called dynamic catalysis circumvents both kinetic and thermodynamic barriers by leveraging the kinetic asymmetry of two or more energetic states of the catalytic material to, for example, promote reactant adsorption and product desorption in a cyclic, stepwise fashion. In this sense, dynamic catalysis proffers a method to surpass static limits to turnover rates prescribed by the Sabatier principle by de-coupling and separately optimizing reactant and product binding energies, which are otherwise fundamentally interdependent. The virtue of this technique has recently been demonstrated by calculation of rates and selectivities of various catalytic systems at dynamic steady state (i.e. the limit cycle). Dauenhauer and coworkers (2,13) have shown that both simple three-step sequences and industrially-relevant reactions such as ammonia synthesis are, theoretically, profoundly accelerated by oscillation of the energetic state of the catalyst (e.g. by periodic oscillation of lattice strain). The current approach for calculating dynamic steady-state rates in such studies, however, primarily involves computationally expensive numerical forward integration of coupled ordinary differential equations (ODEs) until a limit cycle is reached. Simulation of reaction kinetics in this manner requires large calculation times that increase with oscillation frequency, requiring ~1 day for threestep reaction schemes at 10 6 Hz and an expected >300 days for a frequency of 10 10 Hz (14). These onerous computational demands hinder the exploration of vast parameter spaces that describe dynamic catalytic systems and, therefore, essentially proscribe discovery of the oscillation waveforms (shape and frequency) that maximize rate, yield, and/or selectivity. In this work, we develop new strategies for the calculation of dynamic limit cycles disencumbered of the need to forward integrate stiff, coupled ODEs-the numerical solutions for which do not provide the mechanistic clarity characteristic of closed-form rate expressions. Our developments are informed by physical and mathematical intuition established from the examination of a model catalytic system, A + * → A * → B + *, oscillating between two kinetic states-each of which exclusively permits either A + * → A * or A * → B + *. The simplicity of the two-step catalytic sequence allows for an exact analytical solution of dynamic steady-state rates and coverages solely in terms of elementary step rate constants and square waveform frequencies. The derived closed-form dynamic steady-state rate law reveals that (i) the optimal oscillation waveform is uniquely determined by elementary step rate constants, (ii) the optimal waveform may be asymmetric (e.g. more time is spent promoting A + * → A* than A * → B + *), and (iii) the concept of catalytic resonance is not general; for the two-step catalytic reaction, rate is accelerated indefinitely with increase to oscillation frequencies. The learnings from this didactic example are critically enabling in the development of linear algebra and Newton-Raphson based approaches that generalize analytical methods used to derive closed-form solutions and, in doing so, calculate the limit cycle for three-step reactions in milliseconds to seconds, ≳10 8 × faster than previous methods (14). The expedience of the developed mathematical and algorithmic methods enables facile discovery of dynamic catalysis conditions that optimize both (i) the magnitude of oscillation of, for example, A * binding energy and (ii) the wavelength/duration of the oscillation in each energetic state. Linear algebra methods reveal that previously observed resonance regimes are defined by eigenvalues of the matrices that describe governing reaction ODEs; these eigenvalues formalize the concept of characteristic/resonance time scales of catalysis and, like in the two-step example, are relatable, in closed-form, to elementary step rate constants. Complex reaction sequences proceeding via non-linear elementary steps (e.g. bimolecular surface reaction) are not fully describable by matrix algebra methods and, therefore, we instead recast the description of non-linear systems as an optimization problem solved by Newton-Raphson-based approaches. Formulation of non-linear catalytic reactions in the framework of mathematical optimization enables calculation and physical characterization of non-unique steady states that we surmise are intrinsic to non-linear reactions and therefore may hinder dynamic control of industrially-relevant reactions. ## Methods All functions and scripts are written in Octave GNU and Matlab ® 2020a. The code available to download for free from https:\www.github.com/foley352/dynamic. Computational times are measured using the "tic" and "toc" functions. ## Finding analytical solutions for limit cycles in dynamic catalysis We begin our discussion by considering the simplest kinetic system suitable for rate enhancement under dynamic catalysis conditions (Scheme 1), for which we will derive an analytical solution for the time-averaged rate at dynamic steady state. In Scheme 1, there are two reaction steps in series: the adsorption of A and the desorptive conversion of A* to B. We consider the case where there is a square-wave oscillation between two kinetic states, j. Each kinetic state represents a different state of the catalyst (e.g., strain, electric potential) and has a different set of elementary step rate constants 𝑘 𝑖 [𝑗] , for state j and elementary step i. During dynamic catalysis, the catalytic state, or potential energy surface (PES), oscillates with a wavelength 𝜆 (or frequency 𝑓 = 1/𝜆). In this example, the rate constants are 𝑘 𝑖 = 𝑘 𝑖 for time 𝛿𝑡 = 𝜆/2 followed by 𝑘 𝑖 = 𝑘 𝑖 for time 𝛿𝑡 = 𝜆/2, as illustrated in Figure 1a. This oscillation repeats indefinitely. Without oscillation, the static steady-state rate, 𝑟 SS , for the reaction network in Scheme 1 is 𝑟 SS = 𝑘 1 𝑘 2 𝑎 A /(𝑘 1 𝑎 𝐴 + 𝑘 2 ), which is zero for kinetic states 1 and 2. Dynamic catalysis enables the coupling of these kinetic states to give a nonzero reaction rate, by first operating at kinetic state 1 to accumulate A* on the surface and then switching to kinetic state 2 to convert A* to B. Figure 1b illustrates the oscillatory response of A* coverage caused by the periodic switch between kinetic states 1 and 2 (Figure 1a). The surface concentration history in Figure 1b is determined by forward integration of the differential equation (eq. ( 1)): where 𝜃 * + 𝜃 A * = 1, 𝑘 𝑖 = 𝑘 𝑖 for 𝑛𝜆 ≤ 𝑡 < (𝑛 + 1/2)𝜆 and 𝑘 𝑖 = 𝑘 𝑖 for (𝑛 + 1/2)𝜆 ≤ 𝑡 < (𝑛 + 1)𝜆, with the initial condition 𝜃 𝐴 * (𝑡 = 0) = 0. After forward integration of hundreds of wavelengths, the fractional coverage of A* converges to a periodic limit cycle where (eq. ( 2)): We contend that numerical forward integration (e.g. of eq. ( 1)), while quantitatively accurate, (i) does not provide the same physical insight or mathematical clarity as an analytical solution and (ii) is needlessly computationally intensive because the differential equations in dynamic catalysis are very stiff, and much of this computational cost is for calculating unnecessary information-the transient leading up to the limit cycle. In practice, we primarily are concerned with the behavior at the "dynamic steady state", which as shown in Figure 1b, is a limit cycle. To this end, we establish a computationally efficient method for calculating the limit cycle for the reaction in Scheme 1 by finding the analytical solution to the limit cycle itself. Deriving the analytical solution is enabled by two key observations: (i) the differential equation in eq. ( 1) can be solved in piecewise fashion on the ranges from 0 to 𝛿𝑡 and from 𝛿𝑡 to 𝛿𝑡 + 𝛿𝑡 because the rate constants are time-invariant over these ranges, and (ii) instead of an initial condition, as is used for forward integration, we can introduce continuity and periodic boundary conditions that satisfy the defining behavior of a limit cycle (eq. ( 2)). The analytical solution to eq. (1) in general is eq. ( 3): where 𝑐 𝑗 are the arbitrary constants of integration. Equation (3) collapses to the static steady-state solution for 𝑡 → ∞ in the absence of oscillation. Substituting 𝑘 𝑖 and 𝑘 𝑖 into eq. ( 3) gives the piecewise solution 𝑐 2 exp (−𝑘 2 (𝑡 − 𝛿𝑡 )) 0 ≤ 𝑡 < 𝛿𝑡 𝛿𝑡 ≤ 𝑡 < 𝛿𝑡 + 𝛿𝑡 (4) where the (𝑡 − 𝛿𝑡 ) term is arbitrary and chosen for convenience when solving for the two unknown constants of integration, 𝑐 1 and 𝑐 2 . The integration constants are determined by satisfaction of the continuity condition (eq. ( 5)): 𝑐 1 exp (−𝑘 1 𝑎 A 𝛿𝑡 ) + 1 = 𝑐 2 and the periodic boundary conditions (eq. ( 6)): 𝜃 A * (0) = 𝜃 A * (𝛿𝑡 + 𝛿𝑡 ) 𝑐 1 + 1 = 𝑐 2 exp (−𝑘 2 𝛿𝑡 ) which ensure coverages are equal on either side of the switch from kinetic state 1 to 2 and from kinetic state 2 to 1. The solution to the continuity and periodic boundary conditions gives (eq. ( 7)): 𝛿𝑡 ) exp (−𝑘 1 𝑎 A 𝛿𝑡 − 𝑘 2 𝛿𝑡 ) − 1 𝑐 2 = exp (−𝑘 1 𝑎 A 𝛿𝑡 ) − 1 exp (−𝑘 1 𝑎 A 𝛿𝑡 − 𝑘 2 𝛿𝑡 ) − 1 Thus, we now have an analytical solution for 𝜃 A * (𝑡) after substitution of eq. ( 7) into eq. ( 4). The time-averaged rate during the limit cycle is defined as (eq. ( 8)): 𝛿𝑡 0 + ∫ 𝑘 2 𝜃 A * d𝑡 𝛿𝑡 +𝛿𝑡 𝛿𝑡 𝛿𝑡 + 𝛿𝑡 〈𝑟〉 = (1 − exp (−𝑘 1 𝑎 A 𝛿𝑡 )) (1 − exp (−𝑘 2 𝛿𝑡 )) (𝛿𝑡 + 𝛿𝑡 ) (1 − exp (−𝑘 1 𝑎 A 𝛿𝑡 − 𝑘 2 𝛿𝑡 )) which, as expected, is a mathematically symmetric function (i.e. interchange of terms corresponding to states 1 and 2 gives an identical equation). The functional form of eq. (8) demonstrates that, unlike previously reported dynamic catalysis case studies, there is no effect of "catalytic resonance" (Figure 2a). The only condition relevant to rate enhancement for this system is whether the oscillation is sufficiently fast, such that 𝑘 1 𝑎 A 𝛿𝑡 ≪ 1 and 𝑘 2 𝛿𝑡 ≪ 1. At these conditions, the surface coverage is approximately constant because the oscillation frequency is much faster than the time required for the surface coverages to change. We term this state the "quasi-static surface condition", at which eq. ( 8) simplifies to eq. ( 9): 𝑘 2 𝛿𝑡 (𝛿𝑡 + 𝛿𝑡 ) (𝑘 1 𝑎 A 𝛿𝑡 + 𝑘 2 𝛿𝑡 ) = 𝑘 1 𝑎 A 𝑘 2 ( 𝛿𝑡 𝛿𝑡 ) (1 + 𝛿𝑡 𝛿𝑡 ) (𝑘 1 𝑎 A + 𝑘 2 ( 𝛿𝑡 𝛿𝑡 )) At quasi-static surface conditions, the rate of the reaction in Scheme 1 depends solely on the ratio 𝛿𝑡 /𝛿𝑡 , with time-averaged rates shown in Figure 2b. Scheme 1. Simplest dynamic catalysis reaction network. Contour plot of the time-averaged rate as a function of 𝛿𝑡 and 𝛿𝑡 for the reaction in Scheme 1 with 𝑎 A = 1. (b) Time-averaged rate as a function of 𝛿𝑡 /𝛿𝑡 at quasi-static surface conditions. Examination of eq. ( 9) reveals that, in general, the optimal ratio of 𝛿𝑡 /𝛿𝑡 is , as is evidenced by maximum rate occurring for 𝛿𝑡 /𝛿𝑡 = 10 −1.5 (Fig. 2b). The discovery of this simple, consequential mathematical relationship is made possible by the analytical solution and demonstrates that (i) synergistic asymmetry in rate constants and oscillation waveform is key in determining the optimality of rate enhancement and (ii) the phenomenon of catalytic resonance frequency is not a general, or defining, feature of dynamic catalysis. In addition to the proffered physical insight, the analytical solution greatly reduces the computational time compared to forward integration. In the following, we generalize the presented analytical technique by development of an algorithmic procedure for reactions of any number of steps, network connectivity, and kinetic oscillation shape to programmatically find the time-averaged rates during dynamic catalysis. ## A programmatic method for solving for dynamic catalysis limit cycles for linear reaction schemes In reaction schemes that do not involve the reaction between two species with timedependent concentrations, the coupled differential equations that describe the dynamics of fractional coverages are written in matrix form as (eq. ( 10)): where 𝜽 is a vector of all surface species (including vacant sites) and 𝑨 is a time-dependent matrix that is a function of rate constants and chemical activities of reactants and products, which are the coefficients that multiply the fractional coverages in each differential equation. Equation (10) closely resembles a system of coupled first-order ordinary differential equations, with two exceptions: (1) the coefficient matrix 𝑨 is a function of time and (2) at any time, 𝑨 is a singular (non-invertible) matrix because the fractional coverages are not linearly independent. To resolve the second issue, we must eliminate one of the fractional coverages by substituting , which is equivalent to the following procedure: (1) remove the j th row of 𝑨 and 𝜽, ( remove the j th column of 𝑨 and rename it as a column vector 𝒃, and (3) subtract 𝒃 from each column of 𝑨. The new matrix, 𝑨 ′ , has one less row and column than 𝑨 and is no longer singular. The new form of the coupled differential equations is (eq. ( 11)): where 𝜽 ′ is 𝜽 with the j th row removed. Next, it is necessary to eliminate the time-dependence of 𝑨 ′ and 𝒃. This is accomplished by discretizing continuous waves into square waves with n steps. At the limit of 𝑛 → ∞, the n-stepped square wave converges to the continuous wave, as illustrated in Figure 3. On each of the flat terraces of the n-stepped square wave, the rate constants are not functions of time, and only change at the locations of the step discontinuities. Thus, for an nstepped square wave, equation ( 11) can be rewritten as n equations: d d𝑡 𝜽 [𝑗] = 𝑨 [𝑗] 𝜽 [𝑗] + 𝒃 [𝑗] ∀ 𝑡 ∈ [𝑡 [𝑗−1] , 𝑡 [𝑗] ] where the superscript [𝑗] refers to the j th step of the n-stepped square wave, and the primes (" ′ ") have been dropped for clarity. There is one eq. ( 12) for each step of the square wave, and each equation is valid from the end of the previous step (𝑡 [𝑗−1] ) to the end of the present step (𝑡 [𝑗] ). The utility of formulating dynamic catalytic systems in terms of eq. ( 12) is that the coefficient matrix 𝑨 [𝑗] and the vector 𝒃 [𝑗] are not functions of time, and thus eq. ( 12) is in the form of a differential equation that is easily solved with linear algebra. The general solution to eq. ( 12) is of the form (eq. ( 13)): where 𝜃 𝑚 * [𝑗] is the row of vector 𝜽 [𝑗] corresponding to species 𝑚 * , 𝒑 [𝑗] is the particular solution vector for the j th step, 𝑐 𝑠 [𝑗] is the s th constant of integration in the j th step, and 𝒗 𝒔 [𝑗] is the s th eigenvector of 𝑨 [𝑗] with the corresponding eigenvalue 𝜆 𝑠 [𝑗] . Subtraction of 𝑡 [𝑗−1] from 𝑡 in the exponential of eq. ( 13) is arbitrary and chosen for convenience such that the exponentials all equal unity at 𝑡 = 𝑡 [𝑗−1] . The solution presented in eq. ( 13) assumes no repeat eigenvalues of 𝑨 [𝑗] and includes the particular solution to 𝑑 𝑑𝑡 𝜽 [𝑗] = 𝟎, found by eq. ( 14): The only remaining unknowns in eq. ( 13) are the integration constants 𝑐 𝑠 [𝑗] , which are found by satisfying the boundary conditions analogously to the two-step reaction in Scheme 1. For a system with n steps in the square wave and m surface species, there are 𝑛 × (𝑚 − 1) boundary conditions (e.g. for the reaction in Scheme 2, the number of boundary conditions is 2 × (2 -1) = 2). The boundary conditions for a dynamic catalytic system operating at the limit cycle are illustrated in Figure 4. The fractional coverages of all surface species must be continuous in time, which in vector form is written as (eq. ( 15)): 𝜽 [𝑗] (𝑡 = 𝑡 [𝑗] ) = 𝜽 [𝑗+1] (𝑡 = 𝑡 [𝑗] ) ∀ 𝑗 < 𝑛 𝜽 [𝑛] (𝑡 = 𝑡 [𝑛] ) = 𝜽 (𝑡 = 𝑡 ) (15) where n is the total number of steps and 𝑡 is the starting time for kinetic state 1 in the limit cycle (see Figure 4). For the last step (step 3 in Figure 4), there is no "𝑗 + 1" step after, and thus a periodic boundary condition is applied here requiring that the final fractional coverages in step n are equal to the initial coverages in step 1. We emphasize that the periodic boundary conditions in eq. ( 15) assume that the solution 𝜽(𝑡) has the same periodicity as the initial coefficient matrix 𝑨(𝑡) and discuss the existence of solutions that are aperiodic or that have different periodicities at the end of this section. By substitution of eq. ( 13) into ( 15), the boundary conditions can be written in the form of algebraic equations that are linear in the unknowns, 𝑐 𝑠 [𝑗] (eq. ( 16)): ∑ 𝑐 𝑠 [𝑗] 𝑣 𝑠 𝑚 * [𝑗] exp (𝜆 𝑠 [𝑗] (𝑡 [𝑗] − 𝑡 ∑ 𝑐 𝑠 [𝑛] 𝑣 𝑠 𝑚 * [𝑛] exp (𝜆 𝑠 [𝑛] (𝑡 [𝑛] − 𝑡 Equation ( 16) represents a system of linear equations of the form given in eq. ( 17): where 𝒄 is a vector of all 𝑐 𝑠 [𝑗] , 𝑴 is a matrix of coefficients, and 𝒑 is the vector of 𝑝 𝑚 * [𝑗] − 𝑝 𝑚 * [𝑗+1] , where each row in 𝑴 and 𝒑 corresponds to a different equation in eq. ( 16). Solving eq. ( 17) is often the slowest computational step for solving the limit-cycle fractional coverages with n-step square waves, and the computational cost of this step is essentially independent of oscillation frequency. With the constants of integration solved for, we can now describe the entire time-dependence of each species during the limit cycle. The time-averaged rates are found by analytical integration of the rate as a function of time (eq. ( 18)): Ardagh et al. ( 14) investigated the kinetics of the reaction in Scheme 2 with dynamic kinetics where the binding energy of surface species B* is oscillated, and this binding energy correlates linearly with the (i) transition state energy for the A* to B* reaction and (ii) the binding energy of A* via Brønsted-Evans-Polanyi relations. The relationship between the binding energies is given by (eq. ( 19)): where Δ𝐻 ovr is the heat of the overall reaction, and BE A and BE B are the enthalpy change of sorption of species A and B, respectively (e.g. BE A = 𝐻 A + 𝐻 * − 𝐻 A * ). The definition in eq. ( 19) is such that at BE A = 𝛿, the surface reaction becomes isothermic (𝐻 A * = 𝐻 B * ), and the change in binding energy of A and B are related by 𝛾ΔBE A = ΔBE B . In this work, we reproduce a previously published example where 𝛾 = 0.5, 𝛿 = 1.4 eV, Δ𝐻 ovr = 0 eV, and the binding energy BE B is oscillated from 0.1 to 1.03 eV. The activation energy of the surface reaction is (eq. ( 20)): where Δ𝐻 sr is the enthalpy change of the surface reaction (A* to B*), and in this example 𝛼 = 0.6 and 𝛽 = 102 kJ/mol. Following the methodology described above, we reproduce the simulation reported by Ardagh et al. (14) for a square wave (n = 2) oscillation in Figure 5a, with excellent agreement between most data points. The discrepancy for frequencies 10 -2 -10 -4 Hz may be because Ardagh et al. ( 14) simulated a continuous stirred tank reactor where the chemical activities of the reactants and products are not fixed in time and the yields may vary slightly from simulation to simulation. Here, no assumption on a reactor configuration is made and rates are reported for fixed activities of reactants and products. A final difference between the simulation here and the simulation from Ardagh et al. ( 14) is that we capped the value of 𝑘 −1 to 10 25 s -1 , whereas the value from the simulation by Ardagh et al. ( 14) reached 10 29 s -1 . We capped this value because poor scaling of matrices 𝑨 [𝑗] or 𝑴 can cause them to be singular within the numerical precision of Matlab. We also note that the rate constant 𝑘 −1 >> 10 13 s -1 is nonphysical and occurs because the desorption of A * was given a negative activation energy at some conditions. This has no impact on the theoretical insights of the simulations, and we expect allowing the BEP trends to continue to artificial, or non-physical, regimes is preferred to capping rate constants if one aims to develop theoretical insights regarding the general consequences of BEP relationships in dynamic catalysis. Observed dynamic catalysis behavior, even for artificial rate constants, may be edifying and relevant for some different, more physically realistic choice of 𝛽, 𝛿, and reference state. We do not believe that adjusting 𝑘 −1 had a significant impact on the comparison of our results to Ardagh et al. (14). Figure 5b shows the computational time per dynamic steady-state calculation as a function of the number of steps in the square wave. For the n = 2 square wave, each limit cycle takes on average 0.14 ms to calculate, regardless of the oscillation frequency. This is more than 8 orders of magnitude faster than the high frequency calculations reported by Ardagh et al. ( 14) using numerical forward integration. Figure 5 The presented computationally-efficient method for finding the limit cycles in dynamic catalysis vastly expands the explorable parameter space and thereby facilitates rapid discovery of kinetic regimes and the kinetic/energetic parameters that determine their optimality and delineation. For example, there are four parameters that describe a simple square wave: the binding energy of B in each kinetic state, BE B [𝑗] , and the time spent at each kinetic state, 𝛿𝑡 [𝑗] . Employing the developed formalism, we facilely explore the effect of asymmetric square waveforms (i.e. 𝛿𝑡 ≠ 𝛿𝑡 ) in Figure 7a at the binding energies reported in Figure 5a. Figure 7a demonstrates that, by introducing asymmetry, the time-averaged rate is increased by a factor of two, and that the line 𝛿𝑡 = 𝛿𝑡 corresponding to symmetric oscillations is an edge on a larger "resonance region." Further improvement to rate could be made by brute force testing each parameter of the square wave, but we instead continue to leverage the descriptive potence and computational efficiency proffered of algorithmic methods by treating the discovery of maximum time-averaged rate as an optimization problem. The objective function to maximize the time-averaged rate is written as eq. ( 21), where 〈𝑟〉 is a function of the vector containing the times and binding energies for each kinetic state: max 〈𝑟〉(𝒙) where 𝒙 = [𝛿𝑡 , 𝛿𝑡 , BE B , BE B ] T This optimization problem is solvable by the method of gradient ascent, which computes the gradient of the time-averaged rate at the current guess, 𝛁〈𝑟〉| 𝒙 𝑛 , and calculates the next guess, 𝒙 𝑛+1 , until a convergence criteria is satisfied (eq. ( 22)): where 𝜀 is a small parameter that controls the step size. We instead utilized Matlab optimization function fminunc which uses the BFGS quasi-Newton method with a cubic line search procedure to find the optimal square wave for the three-step reaction for 0.1 < BE B < 1.03 eV. This method converges to a local maximum in 0.1 s for the initial guess 𝛿𝑡 = 𝛿𝑡 = 0.5 × 10 −3 s, BE B = 0.1 eV, BE B = 0.8 eV. The optimal square wave is depicted in Figure 8a, where the optimal wave stays at a BEB of 0.1 eV for 10 −4 𝜆, and at 0.9 eV for 0.9999𝜆, where 𝑓 = 1/𝜆 = 6.1 MHz, and gives 〈𝑟〉 = 382 s −1 , a ~14× improvement on the maximum for a symmetric square wave with ## BE B = 0.1 eV and BE B = 1.03 eV (Figure 5). Repeating the same optimization but with a 20-step square wave gives the same solution, suggesting that this asymmetric two-stepped square wave is near the global optimum for these kinetics and constraints. The fractional coverages of each species during the algorithmically optimized limit cycle is shown in Figure 8b, and the instantaneous rates are shown in Figure 8c. In the optimal square wave, the binding energy of B is decreased momentarily to rapidly remove all A* and B*, emptying the surface. The next state of the square wave has a high binding energy of B to accumulate A* on the surface and convert the A* to B*. During this phase, the rate of B formation is negative as B adsorbs on the catalyst surface from the fluid phase. The negative rate of B formation in the second state is compensated by asymmetry in the square wave which maximizes the time-averaged reaction rate by ensuring time is not needlessly spent during either the accumulation or recovery of surface-bound intermediates-which, for this particular system, corresponds to 10 4 × more time spent in the accumulation phase. The critical importance of such asymmetry is explicated by the contour plot in Figure 7b which, along with Figure 7a, illustrates the parameters which define the resonance region. In both figures, the resonance region is bounded to the right by the ratio 𝛿𝑡 /𝛿𝑡 = 1 and to the left by another ratio, 𝛿𝑡 /𝛿𝑡 , which depends on the kinetics of the system. The bottom and top of the resonance region are bound by two of the eigenvalues of the system, in this case, the eigenvalues from kinetic state 2, 𝜆 1 and 𝜆 2 . These eigenvalues bound the characteristic resonance frequencies of the system and exemplify the physical and mathematical detail conferred by formulating the analysis of dynamic catalysis in terms of the well-established relationship between linear algebra and ordinary differential equations ubiquitous in the description of catalytic reactions. Figure 7. (a) Effect of asymmetric times in a square wave where 𝛿𝑡 is the amount of time spent at the condition BE B = 0.1 eV and 𝛿𝑡 is the amount of time at the condition BE B = 1.03 eV. Maximum rate is ~ 52 s -1 . Inset: Rate as a function of frequency for a symmetric oscillation, which is a diagonal slice of the contour plot. (b) Rate as a function of 𝛿𝑡 and 𝛿𝑡 with BE B = 0.1 eV BE B = 0.9 eV. Maximum rate at these conditions is ~ 382 s -1 . The upper and lower bounds on the resonance region are determined by the eigenvalues 𝜆 1 and 𝜆 2 (solid white lines), the right bound (white dashed line) corresponds to symmetric oscillation, 𝛿𝑡 = 𝛿𝑡 , and the left bound is another line that depends on the kinetics, but corresponds to a constant 𝛿𝑡 /𝛿𝑡 ratio. Thus far, we have described the methods for finding the solution that has the same wavelength as the oscillation (𝜆), but the question remains as to whether this solution is unique. Proofs regarding the criteria for the existence and uniqueness of solutions to first-order differential equations with periodic boundary conditions are present in the literature (15,16), but we present here a logical argument for the existence and uniqueness of the solution to the periodic boundary value problem of the coupled first-order differential equations that arise in dynamic catalysis. Equation ( 12) is a linear coupled ordinary differential equation and has a unique solution for the initial value problem 𝜽(𝑡 = 𝑡 0 ) = 𝜽 𝟎 by the Picard-Lindelöf theorem (17,18). Thus, if any function is discretized into an infinite-stepped square wave, there exists one unique solution to each step of the square wave for a given initial value. Now, there must be only one initial value that satisfies the periodic boundary problem criteria 𝜽(𝑡 = 0) = 𝜽(𝑡 = 𝜆) since eqs. ( 16) and ( 17) represent a specified system of linear algebraic equations which has only one solution. Because each initial value problem gives a unique solution, and there exists only one initial value vector that satisfies the periodic boundary condition, we conclude that there exists one unique solution to this system of differential equations. If instead we searched for a solution with a periodicity 𝑛𝜆, then the periodic boundary condition becomes 𝜽(𝑡 = 0) = 𝜽(𝑡 = 𝑛𝜆), for which following the same argument as above, there must exist only one unique solution. Further, we know that the solution 𝜽(𝑡) with periodicity 𝜆 also satisfies the boundary conditions for any periodicity 𝑛𝜆, and thus the only periodic solution for linear systems will be those that have the same periodicity as the kinetic oscillation. Proving that aperiodic solutions to this system of equations do not exist is beyond the scope of this work, but this would require the existence of an initial condition 𝜽(𝑡 = 𝑡 0 ) = 𝜽 𝟎 such that lim 𝑛→∞ 𝜽(𝑡 = 𝑛𝜆) ≠ 𝜽(𝑡 = (𝑛 + 1)𝜆), which does not seem possible for this linear system of equations. The method described above can be employed to find the limit cycles for any dynamic kinetic system where no reactions occur between species that change in time. However, in nondifferential reactors, the activities of fluid-phase species may be transient, and many important reactions involve the reaction between two surface species, and thus require an alternative method to find the dynamic steady states. Further, as the arguments of the existence of unique periodic solutions above required the system of differential equations to be linear, nonlinear differential equations may allow for the possibility of multiple dynamic steady-state solutions, as discussed hereinafter. ## Finding limit cycle solutions for non-linear reaction systems The dynamic steady-state with periodicity 𝜆 is the solution to systems of differential equations where the periodic boundary condition is satisfied (eq. ( 23)): We define a function, F, that integrates the differential equations over one wavelength, 𝜆, to give the output 𝜽(𝑡 0 + 𝜆) for the initial condition, 𝜽(𝑡 0 ), such that (eq. ( 24)): After substitution of eq. ( 24) into eq. ( 23), our periodic boundary condition becomes (eq. ( 25)): Thus, to satisfy the periodic boundary condition, we need to find the fractional coverages vector, 𝜽, that outputs the same vector 𝜽 after forward integration of one wavelength (function F). One method for finding this vector is simply by forward integration until a dynamic steady-state is reached, where we guess a vector 𝜽 𝑘 , and define 𝜽 𝑘+1 = 𝐹(𝜽 𝑘 ) where 𝜽 𝑘+1 is the next guess, and iterate until 𝜽 𝑘+1 ≈ 𝜽 𝑘 is sufficiently satisfied. The efficiency of this algorithm decreases with increasing frequency, for which the method requires forward integration of an indeterminately large number of wavelengths before the periodic boundary condition criteria are satisfied. An alternative approach is using the multivariate Newton-Raphson method, which uses the Jacobian, 𝐽, to determine the next initial guess. This method involves first defining a function that we wish to minimize. For a periodic boundary condition this can be defined as minimizing the sum of the square differences between the input and the output of function 𝐹 for each surface species i (eq. ( 26)): The Jacobian for the vector function 𝒈(𝜽 𝑘 ) is given as (eq. ( 27)): and describes how the function that is being minimized changes with respect to each fractional coverage. The next guess in the Newton-Raphson method is therefore given by: such that information provided by the Jacobian guides and accelerates the iterative search for the dynamic steady-state coverages. The process is iterated until an arbitrary criterion ∑ 𝑔 𝑖 (𝜽 𝑘 ) 𝑖 < 𝜀 is satisfied. This is one of many methods for finding the local minimum of a function, and other methods may have faster convergence to the local minimum; the primary development of the presented methodology is to reformulate the periodic boundary condition as an optimization problem (eq. ( 26)), for which many algorithms can be employed to efficiently find the dynamic steady state at high oscillation frequencies. We demonstrate the computational speed of the Newton-Raphson method for finding the dynamic steady state by considering the reaction network in Scheme 3. This reaction network is nonlinear because step 3 involves the reaction between two species that are changing in time, A * and B * , and thus the differential equations are themselves nonlinear. In this example, we consider the oscillation of rate constants as simple square waves between two states 𝑗 = 1 and 𝑗 = 2, with rate constants for each state given in Table 1. The difference between the two kinetic states lies in the affinity of the catalyst to adsorb A and B, where kinetic state 1 adsorbs B and ejects A * off the surface, while kinetic state 2 does the opposite. The convergence of the Newton-Raphson and forward integration methods to the limit cycles are compared in Figure 9 for a frequency f = 10 2 Hz. The Newton-Raphson method converges to the fractional coverage of A* at the periodic boundary of the limit cycle, 𝜃 A * ,0 , in 11 iteration steps and 1.45 seconds. Forward integration requires more than 100,000 iterations to reach the same value and takes over 2,000 seconds. The computation times of the two methods are compared across decades of oscillation frequency in Figure 9b. At low frequencies, forward integrations will converge to limit cycles in as little as one oscillation, and thus can be faster than the Newton-Raphson method, which requires the numerical calculation of the Jacobian and may take smaller steps in the low frequency regime. At increasing frequencies, the Newton-Raphson method becomes faster because each integration is over a shorter length of time, while the forward integration method generally becomes slower because more oscillations are required before converging to the limit cycle. The decrease in computation time for the forward integration method at 10 2 Hz is a consequence of changing chemical dynamics, which decreases the total time required before converging to a limit cycle. For nonlinear reaction systems, such as the network shown in Scheme 3 or for ammonia synthesis (13,19), it is unclear whether one or multiple solutions exist for the periodic boundary value problem. Using a mixture of the Newton-Raphson method and forward integration, the fractional coverage of A* at the periodic boundary, 𝜃 A * ,0 , was found as a function of the squarewave oscillation frequency, as shown in Figure 10. At the limits of low and high frequencies, there was only one limit-cycle solution. However, at intermediate frequencies of 10 -1 to 10 2 Hz, three limit-cycle solutions were found, one of which was unstable and diverges with any slight perturbation. These unstable limit cycles require the Newton-Raphson solver, because unstable solutions are located at saddle points that locally minimize the criterion in eq. ( 26), but can fundamentally never be reached by forward integration. The fractional coverage of A* in the stable (solid) and unstable (dashed) limit cycles are shown in Figure 11; at all conditions, the fractional coverages of 𝜃 * and 𝜃 C * are near zero, and thus the fractional coverage 𝜃 B * (𝑡) ≈ 1 − 𝜃 A * (𝑡). 1. Solid lines are stable limit cycles. Dashed lines are unstable limit cycles. Only limit cycles that satisfy the periodic boundary condition 𝜽(𝑡 = 0) = 𝜽(𝑡 = 𝜆) were considered. In Figure 10, the limiting behaviors at high and low frequencies are connected smoothly by the unstable states, while the stable states diverge sharply at the onset of instability. The unstable states have the property that 𝜃 B * (𝑡) ≈ 1 − 𝜃 A * (𝑡) ≈ 𝜃 A * (𝑡 − 1/2𝜆), and thus the fractional coverages 𝜃 A * and 𝜃 B * oscillate symmetrically about ~0.5. This behavior is stable at the limit of low and high frequencies but becomes unstable at intermediate frequencies. At low frequencies, the oscillation frequency is sufficiently small that the catalyst surface essentially reaches static steady-state in each oscillation, reaching the bounds of 𝜃 A * ≈ 0 and 𝜃 A * ≈ 1 (Figure 11a). As the frequency increases, the fractional coverages no longer proceed via a sequence of steady states, and the stable solutions diverge at the expense of an unstable limit cycle. At sufficiently large frequencies, the stable solutions separate from the bounds at 𝜃 A * ≈ 0 and 𝜃 A * ≈ 1 and ultimately converge at the quasi-static surface coverage 𝜃 A * (𝑡) ≈ 0.5.. The Newton-Raphson method for finding the limit cycle of nonlinear periodic differential equations can be much faster than forward integration (Figure 9), but is significantly slower than the linear algebra method employed for linear reaction schemes. One method for accelerating the integration of nonlinear differential equations is by Taylor linearization, where the differential equations are linearized by the formula (eq. ( 29)): where, for example, 𝐹 is a function of two variables 𝑥 and 𝑦 and is linearized about some point (𝑥 0 , 𝑦 0 ). Linearizing the differential equation for each reaction intermediate in this way, we obtain a set of linear equations that are analytically solved following eqs. ( 10)- (17). Doing so decreases the computation time by several orders of magnitude, but will only give one solution, despite the actual differential equations having two stable and one unstable limit cycle. Furthermore, the solution is sensitive to the choice in linearization point, as shown by approximate 𝜃 A * ,0 obtained by linearizing the differential equations for the nonlinear reaction in Scheme 3 with kinetics in Table 1. Choices in linearization points were informed by the true solutions depicted in Figure 11. The linearization approximates the true solutions, but further work is necessary to understand the conditions at which multiple steady states may arise and how to choose reasonable linearization points a priori. The existence of multiple limit cycles may be problematic in practical application. First, for the reaction in Scheme 2, the stable solutions give surfaces that are much less evenly distributed between A * and B * , and thus will have lower rates than the unstable solution. Second, any perturbations in the system may result in jumping from one limit cycle to another, causing unpredictable changes in reaction rate, heat generation, optimal feed composition, and outlet composition-leading to many system controls issues (20). In practice, regimes of multiple steady states are typically best avoided. In general, for nonlinear reaction systems, we cannot determine the number of possible limit cycles during dynamic catalysis, nor is it clear at what frequencies these multiple limit cycles will arise, though they are likely related to the time scales for kinetic processes (e.g., quasi-equilibrium of reaction or quasi-steady-state of species). This problem has many similarities to Hilbert's sixteenth problem, as yet unsolved, which concerns the number of limit cycles that exist for a coupled system of two variables with time-independent polynomial differential equations (21). We can also make no justifiable comment on when solutions with different periodicities or aperiodic, chaotic solutions generally exist under dynamic catalysis conditions; however, we contend that, at the limit of low and high frequencies, there will always be one unique limit cycle solution if the reaction network gives only one static steady-state solution, as we discuss next. At the low frequency limit, if the reaction network allows for only one steady-state solution under static kinetics, as determined by chemical reaction network theory (22), then there exists only one limit cycle during dynamic kinetics. This conclusion is arrived at by recognizing that for sufficiently low frequencies, sufficient time is spent in each kinetic state such that, for most of the time spent in each state, rates and surface coverages are time-invariant. Thus, at the low frequency limit, the fractional coverages of the surface can be approximated as The quasi-static surface assumption is an excellent approximation at sufficiently high frequencies, as shown in Figure 12a. At lower frequencies, the quasi-static surface assumption is not rigorously valid over the entire transient, yet can still converge to approximately the same limit cycle, as shown in Figure 12b. This is because for the first oscillation in Figure 12b, the quasi-static surface approximation is not valid as A* quickly covers the surface. During subsequent oscillations, the quasi-static surface approximation becomes valid, which is why they ultimately converge to the same steady state condition. 1. The quasi-static surface assumption reveals that apparent rate constants of elementary steps can be favorably altered by time-averaging the rate constants of two different kinetic states when the kinetic oscillation frequency is sufficiently large. This confirms that, while resonance certainly can be a factor for enhancing the rate (Figure 7), it is not a necessary pre-condition for enhanced rate, selectivity, or conversion during dynamic catalysis. Instead, as recognized by Astumian and coworkers (7,8,23) the fundamental prerequisite for rate enhancement by dynamic catalysis is kinetic asymmetry between the energetic states through which the catalyst is cycled. Rate enhancement by time-averaging of rate constants at quasi-static surface conditions and by resonance represent two different mechanisms by which dynamic catalysis can enhance rates, selectivities, and conversions. Understanding under which conditions one mechanism is favored is a topic that warrants further research. ## Conclusion We establish methods significantly faster than numerical forward integration for finding the limit cycles and time-averaged rates for dynamic catalytic systems. These methods calculate the limit cycles for kinetic oscillations of any shape with computation times that are essentially independent of oscillation frequency and enable facile discovery of the optimal kinetic waveform that maximizes the time-averaged reaction rate using optimization methods. The approach for linear systems, where no time-dependent species react with each other, uses linear algebra to analytically solve for the limit cycles. For nonlinear systems, the coupled ODEs and corresponding periodic boundary conditions are recast as criteria in an optimization problem solved by a Newton-Raphson approach. For linear systems, it is shown that there exists only one periodic limit cycle, but for nonlinear systems, multiple limit cycles exist. Generally, if the reaction network allows for only one steady-state solution under static kinetic conditions, only one limit cycle exists under dynamic conditions in the limit of low oscillation frequency, for which the reaction proceeds via a series of steady states, and in the limit of high oscillation frequency, for which the reaction is maintained at a single quasi-static state. For intermediate oscillation frequencies, no such simplifying conditions exist, and multiple nonlinear solutions are expected. Under sufficiently fast kinetic oscillations, the activities of species are "quasi-static" in comparison to the frequency of kinetic oscillations, and thus the reaction network behaves identically to a static reaction network with rate constants that are equal to the time-averaged rate constants of the kinetic waveforms. These conditions are rapidly simulated by forward integration regardless of whether the reaction network is linear. Analysis of reaction networks under quasistatic conditions reveal that resonance is not always a necessary condition to observe enhanced kinetics during dynamic catalysis; rather, the principal requirement for rate enhancement is asymmetry of the kinetic states sampled by the oscillation waveform.

chemsum

{"title": "Dynamic Catalysis Fundamentals: I. Fast calculation of limit cycles in dynamic catalysis", "journal": "ChemRxiv"}

aggregation-induced_radical_of_donor-acceptor_organic_semiconductors

## Abstract: Narrow bandgap donor-acceptor organic semiconductors are generally considered to show closed-shell singlet ground state and their radicals are reported as impurities, polarons, charge transfer state monoradical or defects. Herein, we reported the open-shell singlet diradical electronic ground state of two diketopyrrolopyrrole-based compounds Flu-TDPP and DTP-TDPP via the combination of variable temperature NMR, variable temperature electron spin spectroscopy (ESR), superconducting quantum interference device magnetometry, and theoretical calculations. It is observed that the quinoid-diradical character is significantly enhanced in aggregation state because of the limitation of intramolecular rotation. Consequently, we propose a mechanism of aggregation-induced radical to understand the driving force of the open-shell diradical formation of DTP-TDPP based on the ESR spectroscopy test in different proportions of mixed solvents. Our results demonstrate the thermally-excited triplet state for donor-acceptor organic semiconductors, providing a novel view to comprehend the intrinsic chemical structure of donor-acceptor organic semiconductors, as well as the potential electronic transition process between ground state and excited state. ## Introduction In the recent more than 30 years, organic semiconductors (OSCs) exhibited great application potential in organic light-emitting diodes (OLEDs), 1,2 organic photovoltaics (OPVs), 3,4 organic field-effect transistors (OFETs), organic photodetectors (OPDs) 4 and other organic electronic devices 8 . Most of the researchers made efforts to develop novel material systems and focused on the photo-and electron-excited states, however, the in-depth investigations of ground-state electronic structures are relatively ignored. The ground states of OSCs can be divided as the following main several types (Fig. 1). The first type is the closed-shell singlet ground state for the most extensively studied organic semiconductors. The representative examples are the triphenylamine, 9 and the other relatively wide bandgap OSCs which have been widely applied in OLEDs possessing a definite singlet electronic ground state (Fig. 1a). The second type is open-shell singlet ground state (S=0, spin multiplicity = 1) for diradicals with two unpaired electrons. Over the past several decades, extensive researches on the diradical analogs including polycyclic aromatic hydrocarbons (PAHs) including zethrenes and oligothiophenes have achieved a great progress on the tuning of their ground states. For most of the diradical molecules such as para-quinodimethanes (p-QDMs) analogues (Fig. 1b), 11,17 they exhibit a singlet ground state because of the efficient delocalization of spins on π-conjugated system, giving birth to a relatively strong coupling between their two spins. The third type is the doublet ground state which existed in monoradical with one unpaired electron in single molecular structure, meaning that the spin quantum number (S) is 1/2, giving a ground-state spin multiplicity of doublet (Fig. 1c). Li et al. demonstrated the highly efficient OLEDs based on monoradical emitters with the spin doublet ground state. 18 The fourth type is triplet spin states (S = 1, spin multiplicity = 3) for diradicals. The previous reports on triangulene-based diradicals have also demonstrated the possibility to obtain relatively stable diradicals with triplet ground state by extending the conjugation system of phenalenyl radicals, however, this synthesis of this type of molecules are still very challenging in this field (Fig. 1d). 10,19,20 Differing from the open-shell PAH radicals and other quinoidal p-QDMs radicals, donoracceptor (D-A) type narrow bandgap OSCs are commonly viewed as closed-shell structure, and the radical species have been recognized as oxygen traps, 21 impurities or defects, 22,23 polarons, 24 or radical cation/anions 25 . In our previous work, we observed the intrinsic diradical character of the D-A conjugated small molecules based on various acceptor units including benzothiadiazole (BT), diketopyrrolopyrrole (DPP), and naphthalene diimide (NDI). 26 We proposed the quinoid-diradical resonance structure to understand the intrinsic radical ground state and the thermally-excited triplet state (Fig. 1e). 26 It was worth mentioned that Prof. mechanism of the formation of diradical ground state is not well studied and reported in previous work. To further investigate the driving force for the formation of quinoiddiradical, the underlying mechanism is still needed to be established. In this contribution, we focused on the studies of ground-state electronic property and demonstrated the open-shell quinoid-diradical character of the diketopyrrolopyrrole (DPP)based small molecules. We proposed the mechanism for the formation of this quinoid-diradical structure as aggregation-induced radical (AIR). The detailed study and discussion were presented in the following work. ## Result and discussion From the molecular design, thienyl-diketopyrrolopyrrole (TDPP) -based derivatives were demonstrated to possess diradical character, 26 and thus providing a good model to investigate the ground-state electronic structure. The donor groups dimethylfluorenyl (Flu) and dithienopyrrolyl (DTP) with different electron-donating capability result in the different degrees of intramolecular charge transfer (ICT) within these TDPP-based small molecules. Based on these considerations, the small molecules Flu-TDPP and DTP-TDPP based on D-A-D structure were prepared by one-step Suzuki or Stille coupling reaction using TDPP as electron deficient chromophore (Fig. 2b), 26 providing dark blue solids with gold-yellow lustra, whose chemical structure were characterized via the UV-vis-NIR absorption, 1 H-NMR, 13 C-NMR and MALDI-TOF-mass spectra (Fig. 3, S1-S6). The molecular configuration of π-conjugated framework and molecular orbital distribution in their relaxed ground-state geometry were predicted via density functional theory (DFT). DTP-TDPP possessed a smaller dihedral angle of 8° compared with the 28°of the relatively twist Flu-TDPP, indicating the more planar configuration for DTP-TDPP (Fig. 2c). These can be also demonstrated by the well-distributed highest occupied molecular orbitals (HOMOs) and lowest unoccupied molecular orbitals (LUMOs) observed for DTP-TDPP (Fig. S1). In the film, Flu-TDPP exhibits an absorption maximum (λmax) at 586 nm with an additional peak at 639 nm, whereas DTP-TDPP shows a broad red-shifted λmax of 640 nm with long absorption tail extended to over 850 nm (Fig. 3a). The optical bandgaps of Flu-TDPP and DTP-TDPP obtained according to the absorption edge of the UV-vis-NIR spectra are 1.77 and 1.46 eV, respectively. The smaller bandgap of DTP-TDPP implies the stronger ICT effect comparing with Flu-TDPP. The energy levels of these compounds were obtained through cyclic voltammetry measurement (Fig. S8). The HOMOs/LUMOs of Flu-TDPP and DTP-TDPP were measured to be -5.25/-3.48 eV and -4.64/-3.18 eV, respectively. The relatively high-lying HOMO of DTP-TDPP was due to the stronger electron-donating ability of DTP group. The high-lying HOMO and narrow bandgap can promote the formation of their diradical and the thermal exciting accessibility from singlet to triplet state. 10 Meanwhile, it has been widely reported that the radical character will endow the open-shell molecules with narrow HOMO-LUMO bandgap. 10,30 It is noteworthy to mention that there are many diradical systems based on PAHs with HOMOs higher than -4.8 eV, however, their ESR signals are not correlated with the oxygen doping or impurity. 31,32 Based on our previous work, 24 electron spin resonance (ESR) was applied to study the potential radical characteristics of these two compounds. DTP-TDPP displayed a one-line ESR spectrum, suggesting the presence of delocalized radical (Fig. 3b). 33 The g value of 2.003 was in good agreement with the typical carbon-based radical rather than other defects or impurities. 26,33,34 In obvious contrast, Flu-TDPP exhibited a nearly silent ESR spectrum at the same test condition. The huge differences of the ESR signal between the two compounds are related with their different molecular geometric configuration and electronic structure according to our previous work. 26 DTP-TDPP exhibited a more planar quinoidal configuration due to its much narrower bandgap and intensive aggregation character comparing with Flu-TDPP according to the DFT calculation (Fig. 2b) and the UV-vis-NIR absorption spectra (Fig. 3a). S1), respectively, indicating the weakened covalency within the bonds of these two molecules and the generation of diradicals. 35 In our previous work, we proposed that the radicals in donor-acceptor type narrow bandgap organic semiconductors originated from the resonant conversion from aromatic to planar quinoids. 26 Considering that the aggregation effect acts as an important role for the formation of quinoidal molecular conformations, the aggregation behavior will be fundamentally consistent with the quinoid-diradical character. The aggregation behavior of Flu-TDPP and DTP-TDPP was investigated by the variable temperature UV-vis-NIR absorption spectroscopy in solutions and thin films. In the dilute chlorobenzene solution (10 -4 M) solutions of the compounds, the 0-1 and 0-0 peaks that denote the intramolecular charge transfer and aggregation characteristics, respectively, exhibited a significant reduction of the from 20 to 100 °C (Fig. S9a, S9b), which is a typical behavior for disaggregation at elevated temperature. 36,37 For Flu-TDPP in thin film, the 0-0 peak weakened and exhibited a slight red-shift (nearly 7 nm) from 20 to 100 °C (Fig. S9c). However, for DTP-TDPP, the enhanced and broad red shift (over 50 nm) 1-0 peak was observed when the temperature elevated from 20 to 100 °C (Fig. S9d), demonstrating the stronger aggregation of DTP-TDPP compared with Flu-TDPP. 38 To investigate the aggregation effects on the spontaneous diradical character in narrow bandgap D-A OSCs, we conducted the ESR measurements in a mixed solution with different proportions of good and poor solvents (Table S2). With the increase of poor solvent (hexane), the ESR signals increased gradually and achieved the peak in solid sample (Fig. 3d). It was due to the different aggregation behaviors in different mixed solvents. In good solvents, the DTP rings linked with TDPP core tend to rotate due to lower kinetics energy. 39,40 The distorted geometric conformation will increase the solubility and weaken the aggregation of molecules in solution, thereby it is not conductive for the formation of quinoid-diradicals. On the other hand, when the ESR test of compounds was conducted in relatively poor solvents, strong repulsions between the solute and the solvent limited the dispersity of the molecules, which induced the formation of large-scale aggregated molecular clusters. In this way, the intramolecular rotation will be limited, leading to the more planar configuration, which theoretically enlarges the conjugated systems and stabilizes the quinoid-diradical structure (Fig. 3c). 41,42 Therefore, the enhanced radical character is attributed to the enhancement of the molecular geometric planarity and we propose aggregation-induced radical (AIR) to interpret the driving force for the formation of quinoid-diradical of DTP-TDPP. The solvent effect on the ESR signal was also eliminated by experimental result and it showed the solvent effect is a negligible factor for the ESR signal decrease. Interestingly, Flu-TDPP did not exhibited an obvious tail absorption in NIR range comparing with DTP-TDPP (Fig. 3a). The long-tail absorption extended to over 800 nm in thin film of DTP-TDPP are indicative of the diradical contribution and the low-lying double exciton state (H, H to L, L). 43 Meanwhile, it is noteworthy to mention that the benzobis(thiadiazole) (BBT) is typical quinoidal building block for the construction of near-infrared absorption and emission OSCs. 41,42 We also detected obvious ESR signal in the OSCs containing BBT core as well as almost all the NIR OSCs with narrow bandgap. 24 We would like to highlight that the typical quinoid cores such as BBT and others will produce more quinoid-diradical configuration in aggregated state than solution state. It is well-known and reported that diradicaloids always show low photoluminescence quantum yields (PLQYs) from decay of the singlet exited state due to the absorption and transition between the H, H and L, L orbitals, 43 singlet fission, 44 and non-radiative decay pathways. 45 The OSCs containing BBT core with more planar molecular structure exhibited aggregation-caused quenching (ACQ) behavior, however, the compounds with BBT core with relatively twist configuration showed aggregation-induced emission (AIE). 41,42 The ACQ behavior can be well understood with our AIR mechanism and the formation of quinoid-diradical configuration will be difficult for the AIE molecules in both solution and solid states. 26,41,42 The electronic ground states of Flu-TDPP and DTP-TDPP were investigated through variable temperature 1 H-NMR spectra. The electronic ground states of Flu-TDPP and DTP-TDPP were investigated through variable temperature ESR and NMR spectra. Both Flu-TDPP and DTP-TDPP exhibit the enhanced ESR signal with the increase of temperature (Fig. S10), implying for the diradical character with singlet ground state (S0) as well as the thermally accessible triplet state (Tt). The variable temperature 1 H-NMR on Flu-TDPP gives the further evidence. The electronic ground states of Flu-TDPP and DTP-TDPP were investigated through variable temperature 1 H-NMR spectra. For Flu-TDPP in C2D2Cl4, the sharp peaks between 7.3 -8.0 ppm in NMR spectrum broaden upon heating from 295 to 393 K (Fig. S11), which is a typical characteristic of the thermal population of triplet species. 46 It is noteworthy that DTP-TDPP exhibits sharp peaks between 6.9 -9.0 ppm in 1 H-NMR spectra and the peaks do not show a broadening trend as the temperature increase (Fig. S12), which is different from the observation in Flu-TDPP. It is due to the more distorted molecular conformation of DTP-TDPP comparing with Flu-TDPP in good solvent originating from the large intramolecular steric hindrance between TDPP core and DTP unit with n-octyl chain. The distorted geometric conformation and weakened aggregation at high temperature weakens the quinoid-diradical character. The singlet-triplet energy gap (ΔEST) of DTP-TDPP was estimated by superconducting quantum interference device (SQUID) in the temperature range from 2 to 400 K. 46,47 The product of molar magnetic susceptibility and temperature (χm• T) show a linear relationship with temperature (T) for DTP-TDPP from 2 to 400 K (Fig. 4a). The plot was fitted by modified Bleaney-Bowers equation and gave the ΔEST = -204 K (-0.406 kcal/mol), suggesting the openshell singlet ground state and the low-lying thermally accessible triplet state of DTP-TDPP. 46,47 With unusual electronic structures compared with closed-shell molecules, open-shell radical molecules show unique electronic transition in photo-, electrical-and thermal-excited processes. 18 For the closed-shell molecules, two singlet spins with different orientations locate in the highest occupied molecular orbitals. The photoexcited electrons dissipate most of the energy by means of radiative decay (Fig. 4b). For the open-shell monoradicals, the photoexcitation of monoradical molecules in their ground states (double ground state, D0) generates doublet excited state (D1). The transition from D1 to D0 is spin-allowed, and thus generates efficient fluorescence emission. 18 We proposed some distinctive electron transitions of open-shell diradical D-A molecules. Similar to the closed-shell molecules, open-shell D-A diradicals possess two reverse spins in their ground states (S0), however, the spin coupling of these two spins is quite weaker compared with that of closed-shell molecules. 33 In the ground states, the weakly electron pairing leads to the thermally-excited triplet (Tt) (Fig. 4b). The singlet-triplet splitting (ΔEST, energy gap between S0 and Tt) is determined by the strength of spin coupling and can be quantified according to SQUID. 11 The presence of thermally-excited triplet state has been demonstrated in the PAHs and p-QDMs diradical molecules, 10,46 but it has not been reported in D-A type OSCs. Furthermore, we propose that the formation of Tt in narrow bandgap D-A OSCs may show underlying interaction and electronic transition with the first triplet excited state (T1) and triplet pair (T1T1) 44 . These electronic transitions processes are closely related to singlet fission process, 48 photothermal conversion, phototheranostic, as well as the triplet excitons in room-temperature phosphorescence and organic afterglow materials. The transition between electronic ground state and excitation state for open-shell D-A diradical molecules, and the influence on the structure-property-performance relationship is still undiscovered but fascinating. ) ## Materials All the reactants and catalysts used in this work were commercially available and used after high vacuum drying. Solvents used in reaction were ultra-dry (water ≤50 ppm). The crude products were purified by silica gel (300-400 mesh) column chromatography with AR eluents and recrystallization with HPLC solvent. The 1 H NMR spectra were measured on a Bruker AV 400 MHz spectrometer in CDCl3 at room temperature. The electronic ground state was investigated by superconducting quantum interference device (SQUID) with Quantum Design 7 Tesla SQUID-VSM system. The powder samples were sealed in a plastic capsule and measured in the temperature range of 2.5 to 400 K, with an applied field of 500 oe. The magnetic susceptibility of samples was fitted with Bleaney-Bowers equation (χ= , where χ is the magnetic susceptibility, N is Avogadro's number, β is the Bohr magneton, g is the magnetic field splitting factor, k is the Boltzmann's constant, T is the temperature, and J is the exchange integral) after correction of diamagnetic signal of plastic capsule and sample holder, diamagnetism of monomer and paramagnetic contamination. Density functional theory (DFT) calculations were performed to investigate the ground-state configurations using Gaussian 09 package. The Lee Yang Parr's correlation functional (B3LYP) that had been proved to predict the geometric structures of organic molecules and was adopted to optimize the ground state (S0) geometry in conjunction with the 6-31G(d) basis set. The frontier molecular orbital energy levels (highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO)) were calculated on the optimized structures at S0. The torsion angle between the major parts of these molecules was calculated based on the optimal ground state structure. The diradical (y0) and tetraradical (y1) characters were calculated with the spin-projected unrestricted Hartree-Fock method using Yamaguchi's formula at UHF/6-31G(d, p) level of theory and basis set: where, i = 0, 1; Ti is the overlap between the two corresponding orbitals; nHONO and nLUNO are the occupation numbers of the highest occupied natural orbital and the lowest unoccupied natural orbital, respectively. Table S1. Optical, electronic properties and diradical indexes of the materials. Materials λonset [a] [nm] Eg opt [b] [eV] HOMO

chemsum

{"title": "Aggregation-Induced Radical of Donor-Acceptor Organic Semiconductors", "journal": "ChemRxiv"}

[3-n₂-o-c₂b₁₀h₁₁][bf₄]:_a_useful_synthon_for_

## Abstract: A simple and efficient method for selective cage B(3) multiple functionalization of o-carborane is described. Reaction of [3-N 2 -o-C 2 B 10 H 11 ][BF 4 ] with various kinds of nucleophiles gave a very broad spectrum of cage B(3)-substituted o-carborane derivatives, 3-X-o-C 2 B 10 H 11 (X ¼ OH, SCN, NH 2 , NO 2 , N 3 , CF 3 , PO(C 6 H 5 ) 2 , etc). This reaction may serve as another efficient [ 18 F]-radiolabeling method of carborane clusters for positron emission tomography applications. ## Introduction Carboranes, 3-dimensional relatives of benzenes, are a class of boron hydride clusters in which one or more BH vertices are replaced by CH units. 1 Carboranes and organic molecules display different electronic, physical, chemical and geometrical properties, which highlights the feasibility or necessity to produce hybrid molecules incorporating both of these two types of fragments. 2a,b Indeed, functional carboranes are now fnding a broad range of applications encompassing organic synthesis, polymers, catalysis, metal-organic frameworks, electronic devices and more. 2c-r As a result, considerable attention has been directed towards the functionalization of carborane molecules. 3 In contrast to the relatively well-studied methods for cage carbon functionalization of carboranes, 1,4 selective cage boron functionalization of carboranes still represents a challenging task and developing new methodologies for selective boron derivatization is eagerly desired. 5,6 Diazonium compounds (R-N 2 + X ) constitute an important group of intermediates that have found wide applications in organic synthesis. 7 Many prominent named reactions associated with aryl diazonium salts have been developed since their frst discovery in 1858. 8 In sharp contrast, diazonium derivatives of carboranes are little known. 9 It has been documented that o-carboranyl diazonium salts are non-isolable, can only be prepared in situ and undergo substitution reactions with the reaction solvent, usually inorganic acids, in the presence of copper salts. On the other hand, it has been reported that B(9)-carboranyl iodonium salt can react with nucleophiles. 13 Very recently, a similar approach for the functionalization of closo-borates via nucleophilic substitution reactions of the corresponding iodonium zwitterions has been developed. 14 However, in these cases, only limited nucleophiles are tolerated and the chemoselectivity of the reaction is highly dependent on the nature of the nucleophiles or the reaction conditions. 13c,14 As the most widely investigated among the carborane family, general and versatile methods for selected cage boron functionalization of o-carboranes still remain very limited. 5 Previously, our group has reported that utilizing 3-diazonium-o-carborane tetrafluoroborate as the starting material, selective B(3)-arylation of o-carborane can be achieved via the aromatic ene reaction of 1,3-dehydro-o-carborane or a visible-light mediated B-C(sp 2 ) coupling of a carboranyl boron-centered radical. However, the substrate scope is only limited to arenes. 11a,12 Considering that dinitrogen is an excellent leaving group, carboranyl diazonium salt may easily undergo a substitution reaction in the presence of a nucleophile. Moreover, compared to aryl diazonium salts, carboranyl diazonium salt may exhibit higher reactivity due to the electron defcient nature of the boron atom and lack of conjugation between the carborane cage and the diazonium group. Herein, we report a proof-of-concept study demonstrating that carboranyl diazonium salt can serve as a powerful synthon for selective cage boron functionalization of o-carboranes (Scheme 1). ## Results and discussion was prepared in 77% isolated yield, by treatment of 3-amino-o-carborane with 1.5 equivalents of nitrosonium tetrafluoroborate. 11a It is noted that the stability of 1 is dependent upon the counterions used and BF 4 offers the highest thermal stability of the salt among the anions examined, such as PF 6 and Cl . A 1.0 g batch of carboranyl diazonium salt 1 stored at 5 C showed no signs of decomposition over four months. With this stable precursor in hand, we found that precursor 1 reacted rapidly with various nucleophiles (2) in acetonitrile, providing the corresponding B(3)-substituted o-carboranes in good to excellent yields (Table 1). Treatment of 1 with 1 equivalent of strong (charged) nucleophiles, such as halide ions, gave the corresponding halogenated carboranes in excellent yields in <5 min (Table 1, entry 1). A large variety of nucleophiles, including inorganic salts, water, alcohols, acids, organometallic reagents, ketones, nitriles and phosphine oxides are compatible with this reaction, resulting in the formation of B-C, B-N, B-P, B-O, B-S and B-X (X ¼ F, Cl, Br, I) bonds. More importantly, various functional groups that were previously unable to be introduced into the carborane unit can now be installed in a very simple and efficient manner. For instance, common functional groups can be easily installed on the o-carborane cage boron position using simple inorganic salts in 5 min, affording the corresponding B(3)-functionalized o-carboranes 3-14 (Table 1, entries 1-12). Reaction of precursor 1 with Grignard reagents or lithium amides also gave the B(3)-substituted o-carboranes in moderate to good yields (Table 1, entries 13 and 14). Weak nucleophiles also work well in this reaction. For instance, in the presence of 10 equivalents of alcohols or water, B(3)-oxygenated carboranes 17 were produced in 81-98% yield (Table 1, entry 15). However, no desired product was observed for tert-butyl alcohol, probably due to the steric hindrance imposed by the tert-butyl group. Instead, 3-F-o-carborane 3a, generated via decomposition of precursor 1, was the only isolated product. Compared to other neutral nucleophiles, the reaction of nitriles is slower even at elevated temperature (Table 1, entry 17). 15 The reactivity of precursor 1 towards nucleophiles containing P]O and S]O double bonds was also examined. For example, the reaction of dimethyl sulfoxide furnished compound 20 after hydrolysis (Table 1, entry 18). Although 31 P and 11 B NMR spectra indicated high conversions, reactions with phosphine oxide nucleophiles resulted in lower yields due to the deboronation of the product during the purifcation process (Table 1, entry 19). 17 Notably, this metal-free approach provides a rare example of B-carboranyl phosphines. 18 The rich chemistry of the carboranyl diazonium salt towards various nucleophiles suggests that it can Scheme 1 Functionalization of arene and o-carborane via diazonium salt. Table 1 Reaction of nucleophiles with precursor 1 a a Reaction conditions: precursor 1 (0.1 mmol) was treated with nucleophile 2 (0.1 mmol for inorganic salt and phosphine oxide; 1.0 mmol for alcohol, acid and ketone; 0.4 mmol for Grignard reagent and lithium amide; nitriles were utilized as solvent) in CH 3 CN solution for 5 min; yields of isolated products are given. serve as a very promising synthon for selective cage boron functionalization of o-carboranes. It is noteworthy that the reaction also works well when performed on a 0.5 mmol scale. 17 All new compounds were fully characterized by 1 H, 13 C, and 11 B NMR spectroscopy as well as HRMS spectrometry. The molecular structures of compounds 4 and 6 were further confrmed by single-crystal X-ray analyses. 17 Interestingly, precursor 1 did not react with anhydrous ether (Scheme 2, eqn (1)); however, it reacted rapidly with wet ethereal solvents. For instance, upon treatment with wet diethyl ether, compound 13, resulting from the C-O bond cleavage of ether, was isolated in 95% yield (Scheme 2, eqn ( 2)). When treated with anhydrous THF, polymerization occurred, leading to gel formation, which suggests the intermediacy of cationic species (Scheme 2, eqn ( 3)). 17 When tert-butyl methyl ether was examined under the same reaction conditions, compound 17a, bearing a methoxy substituent at B(3) position, was formed quantitatively (Scheme 2, eqn (4)), which may shed some light on the reaction mechanism (vide infra). The nucleophilic reaction of the carboranyl diazonium salt was expected to proceed through an S N 1 type of mechanism (Scheme 3). 9,19 Although precursor 1 is stable in solution, it can undergo nucleophile-induced heterolytic B-N bond cleavage, producing a boronium intermediate A. 14 Similar to the reaction of the dinitrogen derivatives of closo-borates, the rate-determining step is the B-N bond cleavage. 9 The resultant reactive boronium intermediate can be trapped by various nucleophiles. For instance, when charged nucleophiles such as inorganic salts were employed as nucleophiles, the corresponding substituted compounds 3-14 were formed in very high yields within 5 min. If the nucleophiles are strong bases, the addition products 15-17 might also be produced via the intermediacy of 1,3-dehydro-o-carborane intermediates. 11 Addition of neutral nucleophiles to the boronium intermediate A, alcohols for example, generates an oxonium ion B, which is further deprotonated by the BF 4 anion to afford 17. For weakly nucleophilic ethers, such as tert-butyl methyl ether, no reaction occurs under anhydrous conditions. However, in the presence of a catalytic amount of water, the oxygenated products, such as 17a, were produced within 5 min. This reaction probably proceeds through a sequence of C-H bond cleavage/isobutylene elimination in intermediate C, which is generated by the nucleophilic addition of the ether to the naked boron vertex of intermediate A. 20 The role of the catalytic amount of water is to facilitate the isobutylene elimination that was detected by GC-MS analyses. The formation of HBF 4 was also confrmed by 11 B and 19 F NMR spectra. 17 The present strategy provides a straightforward and practical access to cage boron functionalized o-carboranes. It has been documented that 18 F-labelled (t 1/2 ¼ 109.8 min) carboranes are promising radiotracers in Positron Emission Tomography (PET). Previously, 18 [F]-fluorination of o-carborane was achieved by nucleophilic substitution of a B(9)-carboranyl iodonium bromide. 21 However, the overall synthesis time of 20 min limits its possible application, probably due to the low reactivity of the carboranyl iodonium bromide. As a proof of concept, we opted to improve the efficiency of the fluorination process by using precursor 1 as the starting material. Under similar reaction conditions to those reported in the literature, 17 the fluorinated product 3a was formed quantitatively within 1 min and it can be easily purifed (eqn ( 5)). ## Conclusions A practical method for selective cage boron functionalization of o-carborane has been developed. By utilizing B-carboranyl diazonium salt as a synthon, a large class of o-carborane derivatives bearing previously inaccessible functional groups can now be efficiently prepared, which may fnd applications in materials sciences. This work demonstrates that B-carboranyl diazonium salt can serve not only as a source of boron-centered radicals 12 or 1,3-dehydro-o-carborane, 11 but also as a source of boronium cations in the presence of nucleophiles. 9,20 These intermediates serve different purposes and are complementary to each other, building up a useful toolbox for cage boron functionalization of o-carboranes. Compared to aryl diazonium salts, the exceptionally high reactivity of B-carboranyl diazonium salt may be due to the lack of conjugation between the carborane cage and the diazonium group. Such a method may fnd useful applications in the efficient and fast synthesis of 18 F-labelled o-carborane derivatives for medical applications. 21

chemsum

{"title": "[3-N₂-o-C₂B₁₀H₁₁][BF₄]: a useful synthon for multiple cage boron functionalizations of o-carborane", "journal": "Royal Society of Chemistry (RSC)"}

cell_adherence_and_drug_delivery_from_particle_based_mesoporous_silica_films

## Abstract: Spatially and temporally controlled drug delivery is important for implant and tissue engineering applications, as the efficacy and bioavailability of the drug can be enhanced, and can also allow for drugging stem cells at different stages of development. Long-term drug delivery over weeks to months is however difficult to achieve, and coating of 3D surfaces or creating patterned surfaces is a challenge using coating techniques like spin-and dip-coating. In this study, mesoporous films consisting of SBA-15 particles grown onto silicon wafers using wet processing were evaluated as a scaffold for drug delivery. Films with various particle sizes (100 -900 nm) and hence thicknesses were grown onto OTS-functionalized silicon wafers using a direct growth method. Precise patterning of the areas for film growth could be obtained by local removal of the OTS functionalization through laser ablation. The films were incubated with the model drug DiO, and murine myoblast cells (C2C12 cells) were seeded onto films with different particle sizes. Confocal laser scanning microscopy (CLSM) was used to study the cell growth, and a vinculin-mediated adherence of C2C12 cells on all films was verified. The successful loading of DiO into the films was confirmed by UV-vis and CLSM. It was observed that the drugs did not desorb from the particles during 24 hours in cell culture. During adherent growth on the films for 4 h, small amounts of DiO and separate particles were observed inside single cells. After 24 h, a larger number of particles and a strong DiO signal were recorded in the cells, indicating a particle mediated drug uptake. A substantial amount of DiO loaded particles were however attached on the substrate after 24 making the films attractive as a long-term reservoir for drugs on e.g. medical implants. ## Introduction The development of controlled drug delivery systems that can administer drugs locally and with a regulated release profile within the human body is of great relevance for e.g. medical implants and tissue engineering applications. 1 Especially challenging is the delivery of hydrophobic drugs, which cannot be administered directly. 2 An ideal scaffold material for these applications should be nontoxic, biologically active and dissolve over time. In the last 25 years mesoporous silica has gained a lot of attention in various fields of research, from catalysis to medical applications. The possibility to control the material characteristics, e.g. pore sizes and particle morphology, in combination with large surface areas up to 1000 m 2 /g and tunable surface functionalities, are a few reasons for why this class of materials attracts extensive interest for use in drug delivery and tissue engineering applications. The high inner surface area of the particles allows the loading with a large amount of active substance molecules. The synthesis of a variety of silica-based films has been reported in literature, mainly by using evaporation-induced self-assembly in spin-or dip-coating, or e.g. electroassisted self-assembly, 11 or vapor-phase deposition. 12 Mesoporous films have previously been used as a drug delivery system, but a controlled release of the drugs is often limited to diffusion control or degradation of the silica matrix itself. 15 Wiltschka et al. reported on coating of pre-formed mesoporous silica nanoparticles with a diameter of about 400 nm on glass slides by spin-coating. 16 By a further functionalization of glass slides with hyaluronic acid it was possible to covalently link a sub-monolayer of particles to the substrate, leading to a delayed uptake of particles by cells over several days. 17 This allowed a potentially delayed particle-mediated release of small hydrophobic drugs within cells. Although spin-coating is an attractive means for preparing homogeneous, particle-based coatings, coat large, or 3D substrates, is virtually not possible. Furthermore, controlled micropatterning of surfaces remain to be demonstrated for both above described methods. In this work, we use a direct growth (DiG) method to form films consisting of mesoporous silica particles grown onto a silicon substrate. 18 Using the described method, it has previously been shown that it is possible both to grow films with controlled thickness and on 3Dsubstates. 19 This is of great importance for fabricating porous coatings on implants without compromising the fine structures of the substrate. The particles themselves in powder form have been shown to efficiently immobilize enzymes, 20 and to act as a potential drug carrier. 21 By varying the particle size using various NH4F concentration in the synthesis solution, the film thickness can be altered to preserve fine structures of the substrate. 19 We shopw in this study that it is possible to functionalize the films with ≡Si-R-COOH after synthesis and to load the films with a hydrophobic, small molecule model drug, DiO, confirming the accessibility of the pores. Cells can adhere to all films, and the particles are taken up by the cells prior to release of their cargo. However, the particle uptake after 24 h of cultivation is still small, indicating that the films can act as long-time drug reservoir on e.g. an implant. The data show further that areas for film growth can be controlled by removal of the substrate functionalization using laser pulses prior to the film growth. This enables selective patterning of the substrate which is useful when designing the implant. Thus, the results show that silica films grown with the DiG method have an excellent potential to be used as a new material for drug delivery and tissue engineering applications, especially when a long-term drug release and a controlled surface morphology are needed. ## Chemicals and cells Tetraethoxysilane (TEOS, 98%), 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC, ≥97.4%), N-hydroxysuccinimide (NHS, 98%), hydrochloric acid (HCl, ≥37%), nitric acid (HNO3, ≥65%), ammonia (NH3, 32% in water), heptane (99%), trichloro(octadecyl)silane (OTS), ammonium fluoride (NH4F, ≥98%), poly(ethylene glycol)-block-poly(propylene glycol)-block-poly(ethylene glycol) (P123, Mn~5,800), 4′,6-diamidine-2′-phenylindole dihydrochloride (DAPI), phalloidin-tetramethylrhodamine B isothiocyanate (Phalloidin-TRITC), 3,3′-Dioctadecyloxacarbocyanine perchlorate (DiO), and C2C12 cells were purchased from Sigma-Aldrich Chemie GmbH, Schnelldorf, Germany. Hydrogen peroxide (H2O2, 30%) and (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) sodium salt (HEPES, 99.3%) was purchased from Merck KGaA, Darmstadt, Germany. Carboxyethylsilanetriol di-sodium salt (25% in water) was purchased from ABCR GmbH & Co. KG, Karlsruhe, Germany. ATTO 647N-amine was purchased from ATTO-TEC GmbH, Siegen, Germany. Dulbecco´s Modified Eagle Medium (DMEM), L-glutamine, antibiotics and fetal calf serum (FCS) and Alexa488 conjugated goat anti-mouse secondary antibody (A11029) was purchased from Life Technologies -Thermo Fisher Scientific, Darmstadt, Germany. Mouse monoclonal anti vinculin antibody (ab18058/clone SPM227) was purchased at Abcam, Cambridge, UK. All chemicals were used as supplied by the manufacturer without further purification. ## Film synthesis The protocol for synthesizing the films followed the DiG method presented in literature. In the synthesis P123 (0.414 mmol) was dissolved in HCl (1.84 M, 80 mL) at 20 °C. Simultaneously, 0, 0.189, or 0.756 mmol of NH4F was added to the solution. Subsequently, 1 mL of heptane was mixed with 5.5 mL of TEOS and added to the P123 solution. After stirring for 4 min, the solution was kept under static conditions over night. OTS functionalized silica wafers, prepared as reported elsewhere, 18 were added to the synthesis solution under static conditions after 12, 7 or 0.5 min, depending on the NH4F concentration. Afterwards, a hydrothermal treatment step (100 °C, 24 h) was performed, followed by filtration and washing with deionized water, and finally all films were calcined at 550 °C for 5 h (heating rate: 1 °C/min). The films are named as DiG_X, where, X corresponds to the NH4F/P123 molar ratio. ## COOH-functionalization After the synthesis, the films were washed extensively with demineralized water in an ultrasonic bath to remove free particles from the surface. Subsequently, the films were placed in a solution of carboxyethylsilanetriol di-sodium salt (25% in water) and HEPES (25 mM, pH 7.2) and stirred at ambient temperature for 2 h (1 µg/mL). Afterwards the functionalized films were washed twice with Ethanol and dried at 60 °C. ## Labeling with ATTO dye For activation of the carboxy group, the COOH-functionalized films were incubated with a mixture of NHS (69.5 μmol in HEPES) and EDC (55.8 μmol) in HEPES (25 mM, pH 7.2) for 20 min at room temperature. After washing with water, a mixture of HEPES and ATTO647Namine (1 mg/mL dissolved in DMSO, 32.75 nmol) was added to the films and stirred for one hour. For purification the films were washed twice with water and dried at 60 °C over night ## DiO loading and release Before incubation with the model drug DiO all films (either non-labelled or ATTO-647N labelled) were dried at ambient temperature for 1 h. The dried films were incubated in a mixture of cyclohexane and DiO (2.27 µM) for 4 h at RT. After the incubation all films were washed with cyclohexane and dried at 60 °C for 40 min. To determine the release of the model drug in an aqueous environment DiO loaded films were incubated with 2 mL of DMEM + 10 % FCS solution for 24 h. The released amount of DiO was subsequently analyzed by UV-vis measurements as well as used for the incubation of cells in µ-Slide 8 wells (ibitreat μ-slide, IBIDI, Munich, Germany). ## Substrate patterning The patterning process was performed using a flash lamp pumped and q-switched Nd:YAG-Laser (Quanta Ray GCR-4, Spectra Physics) at a wavelength of 1064 nm, a pulse length of 10 ns and a pulse repetition rate of 10 Hz. For this purpose, an experimental setup was created. The laser beam was focused down to a spot diameter of approximately 1 mm using an antireflection coated plano-convex lens with a focal length of 500 mm. OTS functionalized silica wafers were moved with a high or low traversing speed of 2 mm/s or 10 mm/s over a length of 10 mm by use of a computer-controlled translation stage. The pulse energy was increased in steps until plasma formation was observed. This resulted in the pulse energy levels of 15 mJ, 19 mJ, and 29 mJ. Concerning a spot diameter of 1 mm this results in radiant exposure (He) values of 1.9 J/cm 2 , 2.4 J/cm 2 and 3.7 J/cm 2 . The pulse energy was measured using a power meter (Nova II, OPHIR) and a corresponding measurement head (30A-P-SH, OPHIR). For each set of parameters (velocity and pulse energy) a linear array of 5 separated lines was irradiated by the laser. ## Cell culture C2C12 cells were cultivated under standard cell culture conditions (37 °C, 90 % humidity, 5 % CO2) in DMEM mixed with 10 % FCS, 2 mM L-glutamine, and antibiotics. Films were incubated with a cell suspension (50K cells/mL) and cultivated for 24 h under standard cell culture conditions. The cells were then fixed with paraformaldehyde (4 %) in phosphate buffered saline (PBS, pH 7.4). Staining of cells was performed using DAPI (0.1 mg/mL) for cell nuclei and Phalloidin-TRITC (50 nM) for filamentous actin. The visualization of focal adhesion contacts (FACs) was performed by using a mouse monoclonal anti vinculin antibody and Alexa488 conjugated goat anti-mouse secondary antibody as described previously. 22 ## Material characterization The morphology of the films was visualized by scanning electron microscopy (SEM) using a Leo 1550 Gemini Scanning Electron Microscope operated at 3 kV. The working distance was 3 -5 mm. The pore characteristics were analyzed using nitrogen sorption isotherms recorded on a Quantachrome Instruments Quadrasorb-SI operated at -196 °C. The specific surface area was calculated using the BET method at the relative pressure of 0.1 -0.2. The pore size distribution was calculated using the KJS-method on the adsorption isotherm, and the total pore volume was determined at P/P0 = 0.99. Small angle x-ray diffraction (SAXRD) measurements were performed on an PANAlytical Empyrean in transmission mode using Cu Kα radiation. ## Biological assessment The growth, adhesion, and particle uptake of cells on top of films were visualized using a Leica TCS SP8 confocal laser scanning microscope and LASX software (Leica Microsystems, Wetzlar, Germany). Lasers emitting at 405 nm (used for excitation of DAPI), 552 nm (for phalloidin-TRITC) and 638 nm (for Atto647-labelled mesoporous silica nanoparticles (MSNs)) were used for the detection of integrated fluorophores. Fluorescence was detected using a HP CL APO 63x/1.40 OILCS2 oil immersion objective (Leica Microsystems) and a pinhole setting of 1 airy unit. The drug loading of the films was studied by UV-Vis spectroscopy measurements of the supernatant using an Analytik Jena AG spectrophotometer SPECORD® 50. ## Results and discussion Films were synthesized using the DiG method with three concentrations of NH4F. SEM micrographs of the films are presented in Figure 1. All films consist of mesoporous SBA-15 particles directly grown onto the substrate with the pores oriented parallel to the surface. No free particles are observed on the substrate after ultrasonication treatment. The particle width and film thickness decreases with an increasing salt concentration (Table 1), which is consistent with the data previously shown by Wu et al. 19 The micrographs clearly shows that a vast majority of the particles are well attached to the substrate. It has been suggested that this type of films is formed through nucleation of silica coated micelles on the hydrophobic substrate, and that the particles grow from these sites. However, defects in the interface between the particles and substrate can appear, as is indicated by the black arrow in Figure 1 (a). This defect yields a particle that is only partly bound to the substrate and can thus influence the particle detachment rate. The time for adding the substrates to the synthesis solution increased with decreasing salt content to gain the desired film morphology since the material formation rate is significantly increased by the addition of NH4F. 23 The salt affects the condensation rate of the silica precursor, and also the structure of the silica network, and therefore it is of great importance that the substrates are added during a time window in the formation of the siliceous network where micelles can nucleate onto the substrate, but prior to the micelle aggregation. In order to demonstrate that DiG films can also be grown on rough, 3D surfaces, a DiG_0.00 was synthesized onto an a lumina sand blasted silicon wafer. The SEM micrograph in Figure 1 (d) reveal that the film follows the 3D structure of the substrate in all directions. Hence, the films are not limited to flat substrates, and coating of e.g. an implant should be possible as long as the surface is pre-hydrophobized. Nitrogen physisorption and SAXRD analysis of the corresponding SBA-15 powders from the film syntheses were performed. The results are presented in Figure 2 and Table 1. As can be observed in Figure 2 (a), all materials give type IV nitrogen isotherms with type 1 hysteresis loops, typical for mesoporous materials with cylindrical pores. DiG_0.00 shows a small tail in the desorption branch of the isotherm, indicating plugs or constrictions in the mesopores. 24 The pore size distributions show only small alterations in the mesopore size. Figure 2 (b) illustrates the SAXRD diffractograms of the materials. All materials show three peaks that can be indexed as the p6mm structure of SBA-15. ## Cell culture and adherence To study the potential of utilizing films synthesized with the DiG method in medical applications, and to visualize the particles, fluorescence labeling of the DiG film particles was performed using an ATTO647N dye. The successful binding of the fluorescent dye was ratified by CLSM, and a homogeneous distribution of MSNs on the surface of the films, comparable to the previously shown SEM images in Figure 1, could be confirmed (Figure S1, Supplementary Information). ## 00, and (d) a silicon wafer without film. Staining: blue = nucleus, red = microfilaments, green = vinculin stained FACs. The specificity of secondary antibody binding was demonstrated in a negative control omitting the primary anti vinculin antibody (supplementary material S2). To determine the biocompatibility of the films, C2C12 cells were seeded on the film surfaces. Cells grown on blank silicon wafers were used as reference. The number of focal adhesion contacts (FACs), characterized by vinculin staining at the end of actin filaments, indicate adhesion strength on a given surface. 25 CLSM verified a regular formation of the actin cytoskeleton and vinculin-mediated FACs on all of the investigated surfaces, indicating that growth and adhesion was not compromised in presence of the films (Figure 3 and Figure 5). Furthermore, the presence of a multitude of FACs suggests a good adherence of cells on the surfaces of the films, independent of the film morphology (Figure 3 (a) -(c)). Also, the shape and appearance of the adhesion points of C2C12 cells on all surfaces are in good agreement with the work of others. 26 It can be concluded that the vinculin staining does not affect the cell adhesion. ## Drug loading and release Due to their high inner surface area, MSNs have been extensively used as drug delivery vehicles especially for drugs showing a low solubility in aqueous environments. Moreover, with such a system a premature uptake of active substances by cells could be excluded as these drugs are only released from the porous system after the particles have been taken up by cells. To show the general accessibility of the film porosity for drug loading, DiG_0.00 was loaded with DiO, a hydrophobic model drug. The DiG_0.00 was chosen for these studies since it consists of the largest particles and hence more available effective surface area and pore volume per cm 2 . However, as the conclusions drawn are naturally also fully relevant also for the thinner films. 29 This lack in the uptake of free model drug at dye concentration levels present in the supernatant after release indicates that a cellular uptake of drugs is favored by intracellular release of drugs from loaded particles from the film surface under the studied conditions, fully in line with previously reported results obtained for spin-coated particulate films. 16 The possibility of particle-mediated transport of drugs was investigated by further cell experiments. Therefore, cells were seeded on the surface of DiG_0.00 films loaded with the model drug DiO and the uptake of drug and particles was investigated by CLSM. The corresponding micrographs are presented in Figure 5. After an incubation of 4 hours only a weak signal of DiO could be detected inside C2C12 cells (Figure 5 (a)), whereas after an incubation time of 24 hours a strong signal of the model drug could be visualized within cells (Figure 5 (b)). To investigate the drug uptake in more detail, the particles forming films were labeled with ATTO647N dye prior to the loading with DiO. Hence, a co-localization of DiO (yellow) and the fluorescently labeled MSNs (red) inside cells could be observed after 8 and 24 h of cell culture (Figure 5 (c)). ## in cells after a cultivation time of (a) 4 h and (b) and 24 h, (c) co-localization of MSN (red) and DiO (yellow) signal within cells after different time of incubation (left =8 h, right = 24 h), and (d) C2C12 cells grown on film free silicon wafers incubated with free DiO-solution in cell culture media for 24 h (blue= nucleus, white = microfilaments, yellow = DiO) . This suggests a particle mediated uptake of drugs from the films. The particle uptake and particle-mediated transport of active agents shown here is in accordance with former work of us and others where the uptake of MSNs as potential drug carriers occurred from cell culture media in both monolayer and hydrogel cultures. 30 Figure 5(d) demonstrates comparatively weak staining subsequent to incubation of free DiO on film free silicon wafers prior to the cell cultivation, illustrating the need of a particle-mediated uptake for lipophilic active substances by cells. It has though been observed that lipophilic drugs can be taken up by cells through a kiss-and-run mechanism when drug loaded nanoparticles come in contact with the hydrophobic membrane surface of the cells. 31 This mechanism cannot fully be excluded based on the present results. However, the amounts of drugs available on the films surface is limited to the molecules close to the pore openings, and the CLMS micrographs only show DiO signal when particles are internalized in the cells. After a cultivation of 24 h it was additionally possible to observe that only a small fraction of the film particles had been taken up by cells while the majority of the particles remained on the substrates surface (Figure 1, Supplementary Material). No particle-free areas around adhered cells could be detected. As discussed above, a gradual detachment of particles from the substrate can be due to differences in the particle growth. The films were treated in an ultrasonic bath three times prior to the cell cultivation in order to remove any free particles from the surface. Figure 1(a) shows a defect in the particle/substrate interface formed during the film growth, e.g. only one side of the half hexagonal prism is bound to the substrate. In addition, smaller particles can be observed in the cross section. These defects are leading to a faster loosening of the particles, and hence yield the initial drug release. One can hypothesize that pa continued drug release will depend on dissolution of the silica framework in the body fluid 32 , which can lead to a continued particle loosening from the substrate or exposure of the drugs through disintegration of the pore walls at the film surface. This can be compared to the spin-coated particulate system developed by Wiltschka et al. that showed a significantly increased uptake of particles by the cells and particle-free areas were easily observed around the adhered cells. 16 A possible consequence of an increased particle uptake can be a non-desired increased cell toxicity. In addition, a longlasting uptake of particles from spin-coated films by cells,a as it is advantageous in potential later applications, can only be achieved by further time consuming surface modifications. 17 For these reasons, the presented films have the potential to be used as an implant coating material without further surface modification, providing a reservoir for a long-time uptake of drug loaded particles by surrounding cells. ## Patterning The microstructure of surfaces has in many studies been shown to be important for how cells respond to the substrate, including cellular alignment and stem cell differentiation. (See for example REFs: . Thus, as a further development, micropattern DiG films were synthesized by local removal of the hydrophobic surface film on the substrate, as it has been shown that functionalization of the substrate with hydrophobic molecules, like OTS or TMCS, is required for a densely packed film to form. 19 Here, OTS functionalized substrates were partially irradiated with a Nd:YAG-Laser using tuneable pulse energy. An irradiation with a pulse energy of 29 mJ using a traversing speed of 2 mm/s and 10 mm/s resulted in a selective removal of the hydrophobic groups on Si-wafer. Both lower pulse energies (15 mJ and 19 mJ) showed almost no removal effect. This indicates a threshold between 19 mJ and 29 mJ at a spot diameter of 1 mm or 2.4 J/cm 2 and 3.7 J/cm 2 in terms of radiant exposure. Also, the plasma formation on the sample surface during the laser irradiation starts between 19 mJ and 29 mJ. Film synthesis using the irradiated substrates revealed that the particles only grow at the nonirradiated areas on the substrates, i.e. where the functionalization was intact (Figure 7). The various sample velocities resulted in different shapes of the irradiated area, where a fast movement resulted in separate, circular domains and a slow movement gave linear structures, shown by the photographs in Figure 7 (a) and (b). The irradiated and film containing areas are indicated by red and green arrows, respectively. The film growth was confirmed using CLSM, where the films were functionalized with COOH-groups and subsequently marked with ATTO647N (Figure 7 (c) and (d)). These micrographs show film growth solely on nonirradiated areas of the substrate. The green lines in the micrographs are artefacts due to interference. The different appearance of the two sample velocities was expected and can be explained by the different overlapping areas. In case of 10 mm/s at 10 Hz two adjacent laser pulses with a diameter of approximately 1 mm shows no overlapping at all and can be clearly separated. Using a sample velocity of 2 mm/s the distance between two adjacent pulses is smaller by a factor of 5 which makes it impossible to distinguish the single pulses. With both sample velocities the irradiated areas showed a significant irregular boundary structure (Figure 7 (c) and (d)). This irregularity and the observation of a threshold between 2.4 J/cm 2 and 3.7 J/cm 2 that coincides with the beginning of plasma formation indicates a removal mechanism based on a photo-mechanical effect. Also, a formation of small dots near the center of the irradiated areas (both velocities) was observed and can be seen in (Figure 7 (c) and (d)). The distribution of the dots is very similar for each irradiated area in case of 10 mm/s and shows a repeating appearance at 2 mm/s. This indicates that at these dots the Si-substrate shows ablation induced by hot spots in the beam profile of the laser. It is hence clear that laser irradiation can be used for pattering a substrate and control the film growth on a substrate, but there are several points for further investigations, like the hot spot formation, the influence of the substrate ablation threshold or the behavior at higher levels of radiation exposure. Also, to support the hypothesis of photo-mechanical interaction some further experiments should be performed. ## Conclusions We have shown that films synthesized with the DiG method can be used as a drug delivery system. C2C12 cells adhere well on films comprising of particles with various sizes. The accessible pores make it possible to load the films with potential drugs, DiO, and functionalization of the film surfaces with e.g. COOH-groups is possible. The films are biocompatible with good growth and adherence of C2C12 cells. The DiO is distributed to the cells mainly through particle uptake where the particles release their cargo inside cells. No release of DiO from the films could be detected in the supernatant after 24 h. The particles are bound to the substrate, resulting in a slow uptake of the drugs and hence, the films are suitable as a drug-reservoir. In combination with the 3D growth of particles 19 and the possibility of a local control of the growth areas of particles on substrates by pre-treatment with lasers enables new potential applications, especially in the field of implant coatings where necessary areas can be provided with film while other areas can be omitted to enable an optimal healing process. As an outlook, one can imagine DiG film growth on other substrates than Si-wafers, e.g. titanium or flexible polymers to come closer to a medical application. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Cell adherence and drug delivery from particle based mesoporous silica films", "journal": "ChemRxiv"}

inverse_material_search_and_synthesis_verification_by_hand_drawings_via_transfer_learning_and_contou

## Abstract: Nanomaterials of various morphologies and chemistry have an extensive use as photonic devices, advanced catalysts, sorbents for water purification, agrochemicals, platforms for drug delivery as well as imaging systems to name a few. However, search for synthesis routes giving custom nanomaterials for particular needs with the desired structure, shape, and size remains a challenge and is often implemented by manual research articles screening. Here, we develop for the first time scanning and transmission electron microscopy (SEM/TEM) reverse image search and hand drawing-based search via transfer learning (TL), namely, VGG16 convolutional neural network (CNN) repurposing for image features extraction and subsequent image similarity determination. Moreover, we demonstrate case use of this platform on calcium carbonate system, where sufficient amount of data was acquired by random high throughput multiparametric synthesis, as well as on Au nanoparticles (NPs) data extracted from the articles. This approach can be not only used for advanced nanomaterials search and synthesis procedure verification, but also can be further combined with machine learning (ML) solutions to provide data-driven novel nanomaterials discovery.Nanomaterials are widely used as photonic devices, advanced catalysts, sorbents for water purification, agrochemicals, platforms for drug delivery etc. due to its ability to control the shape, size, morphology, surface chemistry, and composition, which strongly influence its physicochemical properties as well as its biological behavior. Calcium carbonate represents an inorganic material with a huge potential in the formation of complex micro-and macrostructures, 1,2 which is evident from its wide use by the living organisms in the process of biomineralization, [3][4][5] thereby it is widely used in drug delivery, 6-9 photonics 10,11 etc. At the same time, gold nanomaterials are widespread in drug delivery and photonics mainly due to surface plasmon resonance 12 , photothermal activity, [13][14][15] and surface chemistry tunability, 16 which, coupled with the ability of precise shape control, 17,18 makes it a promising nanomaterial for nanomedicine and physics. The ever-growing amount of experimental data devoted to nanomaterials properties and its synthesis procedures creates a need for fully systematized data collection, storage, and precise search. Several annotated materials synthesis-related databases exist, where the content is usually processed via natural language processing (NLP)-based text mining 22 allowing for either direct or inverse materials search from synthesis procedures to the outcome and vice versa, respectively. In the field of materials science, there is a need in inverse materials search since scientists are often puzzled over how to synthesize a material with desired properties prior to what one would get given the set of experimental conditions. To date, there are several solutions toward nanomaterials synthesis search such as, for example, Nano (https://nano.nature.com/) based on machine learning-driven automated procedures extraction from research articles, although their search is limited to the keywords. Electron microscopy (EM) remains one of the most demanded instruments for materials characterization giving the information about material morphology, size, as well as the shape. Controlling these parameters is of great importance 23 to obtain drug delivery systems (DDSs) with desired biodistribution in the organism, which were shown to depend strongly on DDS size and shape, 24 photonic crystals with low polydispersity, drastically affecting its optical properties 25,26 etc. Moreover, TEM images give insights into the material crystallinity as well as an internal structuring allowing to study core-shell and hollow structures composed of different crystalline phases. Therefore, EM images represent meaningful synthesis outcomes, which can be used for reverse material search. The main question is how to distinguish the difference between two or more synthesis outcomes represented by SEM images. Pixel-by-pixel comparison via image distance calculation 27,28 is able to find exactly the same images but fails on other images of the same objects since it does not consider the relationships between the pixels, not to mention its high computational cost. Instead of pixels, image features invariant to some geometric transformations can be used. 29 For instance, Scale Invariant Feature Transform (SIFT) can extract rotation-insensitive located image features on various scales, which then can be used for image similarity calculation via nearest neighbors. However, SIFT and other similar algorithms 30,31 suffer from low computation speed and are sensitive to brightness/contrast as well as blurring. CNNs usually outperform such algorithms in feature extraction and subsequent classification tasks while being more robust. 32 In their work, Modarres et al. have implemented TL approach on Inception-v3 model pre-trained on ImageNet 2012 dataset for SEM images supervised classification. 33 Therefore, CNNs can be used for feature extraction and subsequent image similarity determination for reverse image search. Due to the lack of large amounts of experimental data in materials science, it is rather difficult to achieve sufficient training accuracy on only materials science datasets, that is why TL approach, which refers to the use of pre-trained ML models for another task, gains the momentum. In this Article, we develop EM reverse image search based on VGG16 CNN repurposing for automated image features extraction and subsequent image similarity determination. Furthermore, we demonstrate case use of this approach on calcium carbonate system, where sufficient amount of data was acquired via random high throughput multiparametric solution chemistry synthesis. Presented approach can be not only used for custom nanomaterials search and synthesis procedure verification, but also can be further equipped with ML solutions to provide data-driven novel nanomaterials discovery. ## Results and discussion To generate meaningful experimental data to form a database of synthesis routes and its outcomes, namely, scanning electron microscope (SEM) images showing micro-/nanoparticle morphology, size, and shape, random high throughput screening, namely was introduced on inorganic calcium carbonate system as a case use including materials synthesis, evaluation using SEM, and database expansion (Fig. 1a, 1b, and 1c, respectively). In particular, randomization of reagents volumes with fixed stock concentrations, coupled with the association of samples in small arbitrary groups of random synthesis parameters e.g., temperature, synthesis time etc. was implemented. To cover the vast majority of possible materials shapes, sizes, polydispersity, and surface morphologies, such parameters as synthesis time, temperature, stirring rate, concentrations of precursors e.g. calcium, carbonate, and bicarbonate ions, mass fraction of miscible/immiscible solvents e.g. methanol, hexanol, isopropyl alcohol (IPA), dimethylformamide (DMFA), propylene glycol, ethylene glycol (EG), tert-butyl alcohol, charged and uncharged polymers of various molar weights e.g. polyethylene glycol (PEG), polystyrene sulfonate (PSS), polyvinylpyrrolidone (PVP), polyethyleneimine (PEI), polyacrylic acid (PAA), and concentrations of differently-charged surfactants e.g. sodium dodecyl sulfate, Triton X-15, myristyltrimethylammonium bromide, cetrimonium bromide (CTAB), were varied in a wide range as it can be seen from distribution box plots for each of the variables on Fig. S1. Randomization of these variables allows to exclude human bias as well as to include 'negative' outcomes, which are very important for any subsequent ML but still under-represented in the majority of research articles. Accordingly, the database of >200 individual nanomaterials was collected (Fig. S2) consisting >20 unique shapes (Fig. 1d), and SEM image was assigned to every single synthetic procedure in the database as an outcome. To achieve reverse image search on SEM images and subsequent image label-based synthesis procedure retrieval from the database, image features extraction needs to be implemented, which is usually achieved by the utilization of encoder-decoder CNNs gradually compressing the image dimensions and trying to reconstruct it with unique features extracted from the images. TL approach has been implemented in this work, namely, re-purposing of the widely used VGG16 CNN model (Fig. 2a) pre-trained on >14.000.000 images of macroscopic objects of as many as 20.000 categories for SEM image features extraction (Fig. 2b). VGG16 CNN consists of convolution, as well as pooling and fully connected dense layers. All these layers represent the mathematical transformation of the (1,224,224,3)-shaped input image pixel intensities, where convolution basically represents the application of filters to the groups of pixels thereby considering interrelations between the adjacent pixels, pooling compresses the image resulting in compressed image representation, and dense layers are usually used for further classification tasks. The last fully connected layer of shape (,4096) carries 4096 features generated for every single image, which are then compressed to 200 and used for cosine distance (eq. 1, cosine similarity of two n-dimensional vectors A and B) determination between the images represented as vectors in 200-dimensional feature space. To demonstrate that this model captures complex crystal morphologies on SEM images, several queries were made resulting in top 3 the most similar images in the database (Fig. 2c). For instance, the model was able to find as simple shapes e.g., cubes, spheres, and spikes, as more complex e.g. urchins, flowers, and sphere aggregates even of comparable sizes, since the exact match is limited to the database size. Moreover, it can be seen that shape mixes are also recognized. To check the ability of this algorithm to reflect the shape abundancies as well as particle sizes, these parameters were calculated for query containing both spheres and cubes and cubes only as well as top 3 the closest SEM images in the database (Fig. 3a and 3b, respectively). The algorithm, indeed, reflects sphere abundancy in the query image equal to 71% trying to find the best match, where top 3 similar images have this parameter equal to 87, 48, and 98% (Fig. 3c). Moreover, according to the sphere size distribution in query image of 1.66±0.19 µm as well as in top 3 most similar images, 2.15±0.26, 1.88±0.40, and 2.55±0.39 µm (Fig. 3d), respectively, query results represent the compromise between particles shape and size. To minimize the impact of shape diversity, image query with cubes only has been examined, where the results of comparable sizes were suggested by the algorithm (Fig. 3e), namely, query results containing cubes with side lengths of 5.6 and 4.5, 5.2 and 3.3, 8.35 and 5.2 µm for a query image with these parameters equal to 5.9 and 3.3 µm were obtained. It is important to note that the increased number of crystal defects going from the query image to the 3 rd query result is observed (Fig. 3b, inset), which suggest this algorithm is sensitive to the surface morphology of the material. From the abovementioned, it can be concluded that this approach allows to search for the materials with the closest shape abundancies, size distributions, as well as material surface morphologies, where all the query results are ranked given the compromise between all of these parameters. To reveal further material insights captured by the algorithm, image data analysis was implemented following image augmentation through the generation of flipped copies of SEM images existing in the database. Principal component analysis (PCA) approach performed on the features extracted from the SEM images has shown the ability of the most representative shapes e.g. cubes, spheres, and spikes, to form distinguishable clusters (Fig. S3a). PCA is a dimension reduction technique, thus some valuable information in the form of feature variance may be lost when n-dimensional space is compressed to visualizable 2-dimensional one. Cumulative explained variance ratio (CEVR) of 2 principal components equal to 28% suggests that the majority of data variance is preserved during the transformation (Fig. S4). K-nearest neighbors (kNN) algorithm implemented on full augmented SEM image dataset allowed to also identify 3 distinctive clusters, which does not correspond with the number of shapes presented in the dataset. These findings, together with the existence of calcium carbonate in 3 main stable crystalline phases as well as literature data indicating spheres are usually consist of vaterite phase, 1,34 cubes -of calcite, 35,36 and spikes -of aragonite, 35,37 suggest that this algorithm probably not only detect the material shape, size, and surface morphology, but also the crystalline phase, where the latter dictates the listed parameters. Therefore, this 2-dimensional scatter plot can be potentially interpreted as a material phase diagram, however, more thorough investigations are needed, which is out of the article scope. To show the versatility of the developed approach as well as its indifference to the material used, its size, as well as the type of the image, this concept was verified on transmission electron microscopy (TEM) images of gold nanoparticles (Au NPs) of six shapes e.g., sphere, cube, rod, dumbbell, trigonal, and amorphous (Fig. 4a) widely used mainly due to their shape-dependent surface plasmon resonance and optical properties, manually extracted from the articles. Algorithm was able to find the most similar images in the collected set of 15 TEM images (Fig. 4b). For instance, all rodlike shapes presented in the image set were found, while the 3 rd query result is turned out to be the best of the worst having the value of cosine distance equal to 1.00, while this parameter of the 1 st and 2 nd query results is equal to 0.37 and 0.52, respectively. The big cosine distance between the query image and 1 st result can be explained by the big difference in NPs lengths (74.3±6.2 and 63.7±5.5 nm, respectively) and widths (19.8±1.4 and 11.9±1.2 nm, respectively). Moreover, size sorting of 2 query results containing trigonal and spherical Au NPs of different mean size 14.7±2.1 and 9.5±1.4 nm was observed (Fig. 4c), where scale bars were included in the images, thereby being included in the image features, and considered during image similarity determination. To make step beyond the synthesis verification towards the customized inverse material queries, drawing-based inverse material search was demonstrated for the first time. First, Canny contour detection was implemented on the set of pre-processed with contrasting and Gaussian blurring calcium carbonate-based nanomaterials SEM images to generate hand drawing-like images for further image similarity determination (Fig. 5a). To examine, whether this approach is feasible, two queries comprising simple crystal shapes, namely, spheres and cubes, were made (Fig. 5b). It is important to note that the algorithm was able to find the closest SEM image in the dataset for a given query with spheres. Moreover, it can be seen that going from the 1 st to the very last query result is accompanied by the change in sphere morphology as well as its size and shape (Fig. 5b, inset) becoming less and less similar to the hand drawn query. More complex query comprising cubes with surface defects has also resulted in a successful search for similar samples even of close sizes, where facet defects were changing from the 1 st to the last query result. Therefore, hand drawing-based inverse material search is demonstrated. Hence, in this study, a novel approach towards the synthesis verification by inverse EM image search and customized drawing-based material query for custom inverse material search is introduced for the first time. TL, namely, VGG16 CNN pre-trained on >14 million images re-purposing was implemented for SEM/TEM image feature extraction and subsequent image similarity determination. Case use of this approach on >200 manually synthesized by random high-throughput screening calcium carbonate-based nanomaterials of >20 various shapes, sizes, and surface morphologies, as well as on Au NPs of >6 shapes extracted from the research articles was demonstrated, thereby proving approach versatility. It was shown that Canny contour detection enables one to implement hand drawing-based query introducing customized inverse material search with the desired shapes, sizes, and surface morphologies. Developed approach can be not only utilized for advanced nanomaterials search and synthesis procedure verification, but also can be further equipped with machine learning (ML) solutions to provide data-driven novel nanomaterials discovery. ## ABBREVIATIONS CEVR, cumulative explained variance ratio; CNN, convolutional neural network; CTAB, cetrimonium bromide; DDS, drug delivery system; DMFA, dimethylformamide; EG, ethylene glycol; EM, electron microscopy; IPA, isopropyl alcohol; kNN, k-nearest neighbors; ML, machine learning; NLP, natural language processing; NP, nanoparticle; PAA, polyacrylic acid; PCA, principal component analysis; PEG, polyethylene glycol; PEI, polyethylene imine; PSS, polystyrene sulfonate; PVP, polyvinyl pyrrolidone; SEM, scanning electron microscope; SIFT, scale invariant feature transform; TEM, transmission electron microscopy; TL, transfer learning.

chemsum

{"title": "Inverse material search and synthesis verification by hand drawings via transfer learning and contour detection", "journal": "ChemRxiv"}

cascade_radical_reaction_of_substrates_with_a_carbon–carbon_triple_bond_as_a_radical_acceptor

## Abstract: The limitation of hydroxamate ester as a chiral Lewis acid coordination moiety was first shown in an intermolecular reaction involving a radical addition and sequential allylation processes. Next, the effect of hydroxamate ester was studied in the cascade addition-cyclization-trapping reaction of substrates with a carbon-carbon triple bond as a radical acceptor. When substrates with a methacryloyl moiety and a carbon-carbon triple bond as two polarity-different radical acceptors were employed, the cascade reaction proceeded effectively. A high level of enantioselectivity was also obtained by a proper combination of chiral Lewis acid and these substrates. ## Introduction Strategies involving a cascade process offer the advantage of multiple carbon-carbon and/or carbon-heteroatom bond formations in a single operation. Radical chemistry has been developed as one of the most powerful tools for carbon-carbon bond formation in organic synthesis . Particularly, the advantages for utilizing the radical methodologies are the high functional group tolerance and the mild reaction conditions, because radical intermediates are not charged species. Therefore, a number of extensive investigations into sequential radical reactions have been reported over the past fifteen years and significant progress has been made in recent years . We have also directed our efforts toward the development of new and efficient cascade approaches for the construction of carbon-carbon/heteroatom bonds based on radical chemistry. These approaches can be classified into two categories according to their reaction mechanism (Figure 1) . Enantioselective radical reactions have been intensively studied over the past fifteen years. Compared with stereocontrol studies on intermolecular radical reactions, the enantioselective stereocontrol in radical cyclizations still remains a major challenge . We have also investigated a new type of chiral Lewis acid mediated cyclization approach for cascade bond-forming reactions via sequential radical-radical processes (Figure 2) . In these studies, the control of the enantioselectivities was achieved by the introduction of a hydroxamate ester as a two-point-binding coordination tether into the middle of substrates A, together with the control of the rotamer population of substrates . In this paper, we describe in detail the cascade addition-cyclization-trapping reaction of substrates with a carbon-carbon triple bond as a radical acceptor as well as the effect of hydroxamate ester as a Lewis acid coordination moiety. Some results have been reported in our preliminary communication . ## Results and Discussion Renaud's group showed in 2002 that hydroxamic acid derivatives are useful achiral templates in enantioselective Diels-Alder reactions . To study the effect of hydroxamate ester as an achiral template in the intermolecular radical reaction, our experiments began with the investigation of cascade radical addition-allylation of hydroxamate esters 3A-C having an acryloyl moiety (Scheme 1). The reactions were eval-Scheme 1: Effect of hydroxamate ester on intermolecular C-C bondforming reactions. uated in CH 2 Cl 2 at −78 °C by employing isopropyl iodide, allyltin reagent, and Et 3 B as a radical initiator. The enantiomeric purities of products were checked by chiral HPLC analysis. The effect of the substituents R 1 and R 2 of hydroxamate esters 3A-C on yield and selectivity was evaluated in the presence of a chiral Lewis acid prepared from box ligand L1 and Zn(OTf) 2 . The results are shown in Scheme 1. Although good enantioselectivities were not observed, the size of the substituents had an impact on enantioselectivity with the larger group leading to lower ee. These observations indicate that the formation of the rigid ternary complex of hydroxamate ester, Zn(OTf) 2 and the ligand L1 is required for enantioselective transformation. A similar trend was observed in our studies on the addition-cyclization-trapping reaction of hydroxamate esters . The chiral Lewis acid promoted the reaction of substrate 3A having a bulky 2-naphthylmethyl group as substituent R 2 to form the product 4A in 40% yield with 7% ee. Moderate enantioselectivity was observed by employing the substrate 3B having a benzyl group as R 1 and a methyl group as R 2 . Particularly, the steric factor of the fluxional substituent R 1 affected not only enantioselectivity but also the chemical efficiency. The use of 3C having a 2-naphthylmethyl group as R 1 led to a decrease in the chemical yield, probably because of the steric repulsion by a bulky substituent R 1 leading to the dissociation of the chiral Lewis acid. In these studies, the absolute configuration at newly generated stereocenters has been not determined. We recently reported in detail the cascade addition-cyclization-trapping reaction of substrates with carbon-carbon double bonds as two kinds of polarity-different radical acceptors . On the basis of these results, the possibility of the carbon-carbon triple bond as a radical acceptor and the hydroxamate ester functionality as a two-point-binding coordination tether was next studied in detail. To understand the scope and limitation of the cascade transformation of hydroxamate esters with carbon-carbon triple bonds, the substrates of choice were 5, 6A-C, 7 and 8 having hydroxamate ester functionality (Figure 3). At first, we studied the cascade reaction of 5 with an acryloyl moiety and 6A-C with a methacryloyl moiety as an electrondeficient acceptor in the absence of a chiral ligand (Scheme 2). To control the rotamer population of substrates, Zn(OTf) 2 was used as a Lewis acid to coordinate the hydroxamate ester functionality. The reactions were evaluated in CH 2 Cl 2 at 20 °C under the tin-free iodine atom transfer conditions by using isopropyl iodide and Et 3 B. The reaction of hydroxamate ester 5 did not give the desired product probably due to polymerization of 5 through the labile acrylamide moiety. In contrast, the reaction of 6A-C proceeded effectively to give the cyclic products 9Aa-9Ca in good yields. Among them, hydroxamate esters 6A and 6B, which have a small methyl or benzyl group as R 1 , have shown a high reactivity, although a 76% yield of product 9Ca was obtained even when hydroxamate ester 6C having a 2-naphthylmethyl group was used. Furthermore, the regiochemical course of the initial radical addition to 6A-C was well controlled. The nucleophilic isopropyl radical reacted selectively with the electron-deficient methacryloyl moiety to give the single isomers 9Aa-9Ca. It is also important to note that Z-isomers 9Aa-9Ca were selectively obtained without the formation of corresponding E-isomers. The E,Z-selectivities are determined by capturing the intermediate vinyl radicals with an atom-transfer reagent such as isopropyl iodide (Figure 4). These selectivities are controlled by the steric factor around vinyl radicals. The vinyl radicals are σ-radicals in a very fast equilibrium between E-isomer B and Z-isomer C. The steric hindrance between the substituents on the α-carbon atom of radical C and isopropyl iodide is assumed to lead to selective iodine atom-transfer in radical B giving 9Aa-9Ca as single Z-isomers. On the basis of the above results, we next studied the reaction of 6A-C at −78 °C in the presence of Zn(OTf) 2 and chiral box ligands L1-L3 (Scheme 3 and Table 1). A stoichiometric amount of chiral Lewis acid prepared from Zn(OTf) 2 and ligand L1 accelerated the reaction of hydroxamate ester 6A having a methyl group as substituent R 1 (Table 1, entry 1), although the reaction of 6A did not proceed effectively at −78 °C in the absence of box ligand L1. The desired product 9Aa was isolated as a single isomer in 51% yield with 60% ee after being stirred for 10 h. The use of hydroxamate ester 6B having a benzyl group led to not only an enhancement in chemical yield but also to an improvement in enantioselectivity to give the pro- 1, entry 3 and 4). Further reduction of the chiral Lewis acid load to 10 mol % resulted in a decrease of both the chemical yield and enantioselectivity (Table 1, entry 5). In the case of 10 mol % of the chiral Lewis acid, the ternary complex of the ligand, the Lewis acid and the substrate were not effectively formed, and the background reaction giving the racemic product proceeded. Additionally, the high Z-selectivity of product 9Ba indicates that the stereoselective iodine-atom transfer from isopropyl iodide to an intermediate radical proceeded effectively under these catalytic reaction conditions. The reaction using box ligand L2 instead of L1 attenuated the enantioselectivity ( The absolute configuration at the newly generated stereocenters of 9Aa-Bd was assumed by similarity between the present reaction and the previously reported reaction of substrates having the carbon-carbon double bond . In these reactions, a ternary complex of ligand, Lewis acid and substrate would control the three-dimensional arrangement of two radical acceptors. A tetrahedral or cis-octahedral geometry around the zinc center was proposed . In Figure 5, a tentative model of an octahedral complex is shown, in which two oxygen atoms of the hydroxamate ester functionality occupy two equatorial positions. To study the effect of an electron-deficient acceptor on the cascade process, the reactions of propiolic acid derivatives 7 and 8 were tested (Scheme 4). At first, the reaction of 7 was evaluated under asymmetric reaction conditions. However, the cascade addition-cyclization-trapping reaction did not proceed, and the simple adduct 10 was formed in 57% yield by the addition-trapping process. Next, the reaction of propiolic acid derivative 8 was tested, because we expected the -hydrogen shift from 1,3-dioxolane ring into the reactive vinyl radical as shown as D. However, the simple adduct 11 was only obtained Scheme 6: Cascade reaction of 14. in 78% yield. The results from these studies show that a carbon-carbon double bond, e.g., a methacryloyl group, of the electron-deficient acceptor is essential for the successful cascade transformation. To gain further insight into the stereocontrol in the cyclization step, we next studied the opposite regiochemical cyclization by using the substrate 12 via the intermediate radical F (Scheme 5). The reaction was carried out in the presence of Bu 3 SnH under asymmetric reaction conditions. Although the reaction proceeded even at −78 °C, the nearly racemic product 13 was isolated in 60% yield. This observation indicates that the regiochemical course of the cyclization step is an important factor to achieve the highly asymmetric induction. Scheme 5: Opposite regiochemical cyclization using substrate 12. We next investigated the reactivity of internal alkynes as electron-rich acceptors (Scheme 6). The internal alkyne 14 has shown a good reactivity comparable to that of the terminal alkynes 6A-C. In the absence of a chiral ligand, the zinc Lewis acid accelerated the reaction of alkyne 14 with an isopropyl radical at 20 °C to give the desired cyclic product 15a in 73% yield. Under analogous reaction conditions, both cyclohexyl iodide and cyclopentyl iodide worked well to give 15b and 15c in 65% and 68% yields, respectively. However, the reaction with a bulky tert-butyl radical did not proceed effectively, probably due to side reactions such as polymerization. We finally investigated the enantioselective reaction of internal alkynes 14 and 16 (Scheme 7). The reaction of 14 proceeded with good enantioselectivities (Table 2). When a stoichiometric amount of chiral Lewis acid was employed, the reaction with an isopropyl radical gave the desired product 15a in 86% yield with 83% ee (Table 2, entry 1). The reaction proceeded equally well with 30 mol % of chiral Lewis acid as with a stoichiometric amount (Table 2, entry 2). The secondary radicals, generated from cyclohexyl iodide or cyclopentyl iodide, reacted well to afford 15b and 15c with 85% ee and 83% ee, respectively (Table 2, entry 3 and 4). In marked contrast to the reaction in the absence of a chiral ligand (Scheme 6), the use of bulky tertbutyl iodide led to not only an enhancement in chemical yield but also to an improvement in enantioselectivity ( 2, entry 9). It is also important to note that the high Z/E-selectivity of products was observed even when internal alkynes 14 and 16 were employed. These results indicate that the iodine atom-transfer from R 2 I to the substituted vinyl radicals proceeded stereoselectively. Particularly, the substrate 16 having a phenyl group gave the intermediate linear π-radical. Thus, the capture of linear vinyl radical with atom-transfer reagent would be influenced by the steric hindrance around the quaternary carbon atom . ## Conclusion We have shown the cascade radical addition-cyclization-trapping reaction of substrates with a carbon-carbon triple bond as a radical acceptor as well as the scope and limitation of hydroxamate ester as a coordination site with a chiral Lewis acid. Synthetic strategies involving enantioselective radical cyclizations would be desirable tools for preparing functionalized cyclic compounds with multiple stereocenters. These studies offer opportunities for further exploration of fascinating possibilities in the realm of cascade radical reactions. ## Supporting Information Supporting Information File 1 General experimental procedures, characterization data of obtained compounds, and preparation of substrates. [http://www.beilstein-journals.org/bjoc/content/ supplementary/1860-5397-9-128-S1.pdf]

chemsum

{"title": "Cascade radical reaction of substrates with a carbon\u2013carbon triple bond as a radical acceptor", "journal": "Beilstein"}

a_cyclorgdf(me-v)_analog_as_chemical_probe_to_study_integrins_function_in_living_cells

## Abstract: Studying the role of integrins in cellular processes requires the ability to monitor their localization in dynamic events. We report a chemical probe that can be used to image integrins in living cells. The fluorescent probe was derived from cyclo-RGDf(Me-V), a compound selective for integrins that possess an RGD-binding domain. We describe its synthesis and we demonstrate its use to detect integrin αVβ5 in cells. The probe's dissociation constant for the integrin αVβ5 protein is 0.18 µM. The probe's activity was validated in murine BV-2 microglial cells using cell engulfment assays, flow cytometry, and confocal fluorescence imaging. This probe will provide access to spatiotemporally resolved studies of RGD-binding integrin function in living cells without the need for genetic modification. 8 0.1% DMSO 10 nM Cilengitide (1) 0.1% DMSO 10 nM Probe 2 10 nMProbe2 ## ■ INTRODUCTION Integrins are heterodimeric transmembrane proteins found at the surface of cells, where they connect the intracellular cytoskeleton to the extracellular matrix. Besides being essential structural elements of cells, they also play important signaling roles in development and in pathology. The function of integrin proteins is dictated by the differential pairing of 18 alpha and 8 beta subunits. A family of integrins-αVβ3, αVβ5, α5β1 and αIIβ3-binds extracellular matrix proteins that contain the "RGD" peptide sequence such as fibronectin, fibrinogen, vitronectin, and osteopontin. These integrins, αVβ3 in particular, have been studied extensively in cancer: their upregulation in tumour angiogenesis and metastasis has made them attractive drug targets. The complex role that integrins play in cellular signaling events is dynamic: it must be studied in living systems to reveal meaningful information about signaling cascades and how integrins respond to biochemical stimuli. This is especially true in the brain, where the physiological role of integrins αVβ3 and αVβ5 remains understudied. The αVβ3 and αVβ5 subtypes may play a crucial role in synaptic plasticity and participate in microglial activation/inflammation. However, obtaining time-resolved information at the molecular level in brain tissue is still a major challenge due to a lack of appropriate technology. The most common fluorescence microscopy methods used to visualize proteins are not ideal to study dynamic events in cells: they rely on immunocytochemistry (fast method, but not dynamic as cells must be fixed), or on genetic fusion of the protein of interest with a reporter tag, e.g., GFP fusion (very specific, but time-consuming genetic manipulations). Part of our research program on the role of integrin in neuron-glia interactions motivated us to address this challenge. Accordingly, we aimed to create a fluorescence imaging probe to monitor endogenous integrin receptors in living cells. An exogenous chemical compound that targets a conserved motif of an integrin protein in its native state would ensure that: (1) the expression levels are unperturbed, (2) the method is independent of species or cell line, and (3) the wavelength of the fluorescence reporter can be modified to allow for multiplexing (e.g., pulse-chase experiments). For our design, we selected a class of ligands discovered by Kessler that is wellestablished to be selective for RGD-binding integrins: cyclo-RGDf(Me)V, more commonly known by its commercial name cilengitide (1, Figure 1). This pseudopeptide shows nanomolar inhibitory activity for αVβ3, αVβ5 and α5β1, and micromolar activity for the platelet receptor αIIβ3. This compound has also been exploited in cancer diagnosis where the overexpression of αVβ3 in tumour tissues can be detected with cilengitide analogs conjugated with radionuclides, MRI contrast agents, fluorophores, and nanoparticles; however, several of these techniques lack high spatiotemporal resolution. Figure 1. Permissible structural variations of Kessler's cRGDf(Me-V) (1). Affinity decreases when the N-methylvaline is replaced by lysine, but it can be compensated by creating oligomers. Cyclic pentapeptide 1 is a rigid mimic of the RGD motif found on the extracellular matrix protein vitronectin that associates tightly with integrins αVβ3 and αVβ5. A crystal structure of αVβ3 bound to cyclo-RGDf(Me)V was reported in 2002; it shows that the valine and D-phenylalanine residues of 1 are pointing away from the binding site, toward the extracellular milieu (Fig. 2). Most reported compounds used for conjugation are analogs of 1 where the valine residue is replaced by lysine, with a reporter tag covalently attached to the amine of the lysine residue (typically, a PET radionuclide). Although it is synthetically simpler to replace the N-methylvaline by a lysine to attach linker chains, this substitution leads to at least a 30-fold loss of activity on integrin αVβ5. Lower binding affinity due to structural changes can be partially compensated by creating cRGDfK dimers, trimers, or tetramers. Figure 2. Crystal structure of the extracellular portion of integrin αVβ3 with bound ligand cRGDf(Me)V. The RGD binding domain is located at the apical junction of the dimer. The cyclo-RGDf(Me)V ligand (orange) is nested at the interface of the alpha (grey) and beta (blue) subunits. The ligand's phenyl sidechain points away from the protein; modifying this aromatic ring to attach a linker chain should maintain a high affinity. (PDB 1L5G). In contrast, the distal aromatic positions of the D-Phe residue have never been exploited for conjugation, despite the analog being shown to be equally active to 1. Only a tyrosine-substituted and a meta-iodinated analog have been reported, and both showed the same activity as 1 (Fig. 1). Herein, we report the design and application of molecular imaging probes cilengitide-PB405 (2) and cilengitide-A568 (3) (Figure 1). We demonstrate the use of the probes for fluorescence imaging of integrins in living cells, and for functional study of endocytosis in a bead engulfment assay. ## ■ RESULTS AND DISCUSSION The molecular imaging probes cilengitide-PB405 (2) and cilengitide-A568 (3) differ from other conjugates in that the reporter cargo is appended to the aromatic ring of the ent-Phe residue of cyclo-RGDf(Me)V (Fig. 1). A fluorescent imaging molecule has been reported, cRGDyK-A568 (22), but three structural deviations from cilengitide made it less potent: the valine was replaced by lysine, the N-methyl group was removed, and the phenylalanine was replaced by a tyrosine. Hypothesizing that the meta position of D-Phe in 1 should be amenable to substitution with a fluorescent label without loss of activity for integrin αVβ3 and αVβ5, we synthesized two analogs: blue-fluorescent probe cilengitide-PB405 (2) and red-fluorescent probe cilengitide-A568 (3). Synthesis of modified Cyclo-RGDf(Me)V Pseudopeptides. Imaging probe cilengitide-PB405 (2) was synthesized in a total of 27 steps, with 19 steps for the longest linear sequence (Fig. 3). The preparation of a novel meta-substituted aminomethyl-D-phenylalanine derivative 9 is outlined in Figure 3A. m-Toluic acid (4) was converted to benzylic bromide 5 in five steps. The stereocenter of the amino acid precursor was installed via enantioselective alkylation of glycinyl imine 6 with 0.1 mol% of Maruoka's chiral catalyst (7) under biphasic conditions. The chiral product showed a 94% ee. Three additional steps were required to obtain the modified D-phenylalanine 9 bearing protecting groups appropriate for solid-phase synthesis. The protected N-methylvaline 12 was prepared in 88% over two steps according to a reported procedure. The linear form of the pentapeptide was assembled on trityl chloride resin 10, starting with Fmoc-protected glycine. HBTU was used as the principal coupling agent, with the exception of coupling of phenylalanine 9 to N-methylamine of 12, where the more reactive HATU was used. The pentapeptide 13 was cleaved from the resin with hexafluoroisopropanol and purified by chromatography to give diastereomers 14 in a 97:3 ratio (reflecting the 94% ee of 9). Pentapeptide 14 was cyclized under high dilution conditions with diphenyl phosphoroazidate as the coupling reagent. Global deprotection in 95% trifluoroacetic acid led to cyclic pentapeptide 15 in very good yield. Blue fluorescent imaging probe cilengitide-PB405 (2) was assembled by coupling 15 to a fluorophore via a six atoms linker chain (Figure 3B). The dye Pacific Blue (16) was activated as a succinimidyl ester, coupled with 3-oxa-ω-aminovaleric acid, and converted to succinimidyl ester 17b. Coupling of 15 with activated ester 17b yielded the desired fluorescent probe 2 in 29%. The low yield is due to a challenging purification by reversephase HPLC. While the small size of the fluorescent moiety on 2 is an advantage, its excitation/emission wavelengths in the violet/blue spectrum (405/455 nm) can limit the range of possible experiments (e.g., strong autofluorescence background). A complementary red variant was therefore synthesized. Red fluorescent imaging probe cilengitide-A568 (3) was prepared from fully protected pentapeptide 18 (Fig. 3C). In this case, the N-Boc protecting group on the modified D-Phe allowed a selective deprotection, followed by coupling to the NHS-activated ester of ω-azidobutyric acid. Global deprotection afforded peptidic azide 19. Sulforhodamine Alexa568 dye (20) was reacted with HATU to obtain propargyl glycinamide 21. Both azide 19 and alkyne 21 fragments were then connected using a copper-catalyzed click reaction to yield redfluorescent cilengitide-A568 (3). To evaluate the efficiency of 2 and 3, we synthesized known probe cRGDyK-A568 (22) by preparing the cyclo(RGDyK) peptide according to the literature. Then, fluorophore 20 was directly attached to lysine sidechain to provide the closest analog with respect to fluorescent and chemical properties (See Supplementary Information). Binding Activity of Cilengitide-A568 (3) with Isolated Integrin αVβ5. The binding was investigated with isolated human integrin αVβ5 protein (in line with our interest in studying its role in glia). The change in fluorescence anisotropy was measured with probe 3 to obtain a dissociation constant (Kd) of 0.18 ± 0.13 µM (95% CI). The value was calculated by fitting the data to a non-linear regression representing a single-site binding model (see Fig. 4 and Methods). The difference between the anisotropy of the free and bound states was small, likely due to the free rotation around the long linker of 3 (propeller effect). Nevertheless, the assay demonstrates that fluorescently-labeled compound 3 retains affinity for integrin αVβ5 receptor in the midnanomolar range. The binding assay was conducted only with cilengitide-A568 probe 3 due to instrumental constraints. Yet, given that the fluorescent sidechain in 3 is much larger than that of Pacific Blue-labeled probe 2, the activity of 2 is expected to be very similar (supported by qualitative observations in other experiments, data not shown). ## Labeling Integrins in Cells With the imaging probes in hand, we turned to confirm their use in cellular contexts to study integrins that bind the RGD-motif, integrin αVβ5 more specifically. The probes' spectrophysical properties were measured using fluorescence spectroscopy and confocal microscopy in live-and fixed-cell systems. We investigated whether our modifications of cRGDf(Me)V maintained the binding affinity, selectivity, and inhibitory effect of the probes 2 and 3 towards integrin αVβ5 and its associated signaling functions ## Imaging Probes Cilengitide-PB405 (2) and Cilengitide-A568 (3) Label Endogenous Integrin αVβ5 in BV-2 Microglia. The labeling of endogenous integrin αVβ5 was investigated with probe 3 using live BV-2 microglia (for experiments with probe cilengitide-PB405, see Supp. Info.). BV-2 cells are a robust model for primary murine microglia, and αVβ5 represents close to 60% of the RGD-binding integrin proteins they express. BV-2 cells were labelled with cilengitide-A568 probe 3 within 10 min at a concentration of 1 µM. While lower concentrations can be used, they require impractical incubation times (e.g., one hour or longer, data not shown). We opted for a short incubation time of 10 minutes as the probe's intended use is for the imaging of dynamic events in living cells. Fluorescent imaging probe cilengitide-A568 3 labelled native integrin proteins in living BV-2 cells at a level significantly higher than the background, the free dye 20, or the lysine-modified analog cRGDyK-A568 (Fig. 5). Importantly, competition experiments with 10 equivalents of cilengitide (1) reduced the fluorescence to background level (Fig. 5G). In terms of selectivity, cilengitide-A568 (3) performed ca. 50% better than cRGDyK-A568 (22); the difference in fluorescence signal between the probes only and the pre-blocked cells was greater with 3 than 22. It suggests that it is indeed less perturbing to the ITG-cilengitide interaction to append the linker-dye moiety to the phenyl substituent of cilengitide in 2 and 3, instead of a lysine residue substituting for Me-Val in cRGDyK-A568. A common limitation of fluorescent imaging probes based on small molecules is false positives arising from their embedment within the cell lipid bilayer via their linker-dye component. Cilengitide-A568 and its dye component 20 are very hydrophilic and are not expected to permeate the cell membrane via passive diffusion. With the free dye 20, the background cytosolic signal observed is ascribed to normal pinocytosis of the nutrient medium. Co-incubation of the free dye with cilengitide did not cause an increase of the cellular fluorescence, supporting that only the bound probe is internalized. Large vesicles can be clearly observed in the labelled cells (Figure 5A and 5D). In contrast, both cRGDyK-A568 and cilengitide-A568 are concentrated in smaller vesicles in the cytoplasm and nearby focal adhesion points (Figure 5B, C, E and 5F). Presumably, the RGD-peptide is internalized with the integrin once 6 the ligand displaces attachment to extracellular matrix. Localization and intensity of the fluorescent signal in all the competition experiments resembled that of the free dye (see Supp. Info., Fig. S9). Probes cilengitide-PB405 (2) and cilengitide-A568 (3) Inhibit Integrin αVβ5-Mediated Engulfment Cascades in BV-2 Microglia. In the brain, synaptic pruning has been proposed to involve neuron phagocytosis mediated by microglia via upregulated αVβ5 cascades, however, it has never been observed directly. [10, As a proof of concept toward studying this important phenomenon, we developed a functional assay to assess whether our probes can be used to monitor αVβ5 in living cells. A reliable method to quantify microglial phagocytosis is the detection of internalized latex beads using confocal microscopy and flow cytometry. When beads are coated with extracellular matrix (ECM) proteins, such as collagen and fibronectin, they trigger signaling pathways that can initiate phagocytic cascades (vitronectin is the native ECM protein ligand for integrin αVβ5). These cascades are responsive to inhibitors, and bead engulfment serves as a measure of phagocytic activity. Accordingly, we conducted an assay to determine whether probes cilengitide-PB405 (2) and cilengitide-A568 (3) inhibit phagocytosis of vitronectin-coated latex beads in model glia. This assay confirmed that probes cilengitide-PB405 and cilengitide-A568 maintain integrin αVβ5-binding activity despite being covalently modified with a fluorophore. Figure 6 indicates that both cilengitide-PB405 and cilengitide-A568 block microglial engulfment of vitronectin-coated beads with similar efficiency (or better) than the parent cilengitide inhibitor. Their half-maximal inhibitory concentrations (IC50) were calculated to be: cilengitide (1) = 623 ± 235 nM, probe 2 = 903 ± 348 nM, and probe 3 = 71 ± 23 nM. It may be noted that these IC50 values for cilengitide are one to two orders of magnitude higher than previously reported: integrin αVβ3 = 0.65 nM; integrin αVβ5 = 11.7 nM. However, the assay we used measures the phenotype (blocking microglial phagocytic response), while prior reports used displacement assays with purified αVβ3 and αVβ5 proteins. Here, cilengitide and the probes showed similar efficacy in blocking phagocytosis initiated by integrin. While the data presented in Figure 6 show a slight deviation of probe 3 relative to 1 and 2, this experiment confirms that modifying cilengitide with fluorophores attached to the aromatic D-Phe residue does not adversely affect its binding affinity or selectivity. Importantly, none of the compounds reduced the viability of BV-2 cells, supporting that the reduced phagocytic response results from integrin inhibition and not cell death (Fig. 6F). Practical considerations make cilengitide-based probes 2 and 3 attractive tools to label endogenous integrin αVβ5 in cell systems. First, they can be used across multiple species as the RGD binding site is highly conserved. Second, the procedure is faster than other labeling techniques, such as immunocytochemistry or recombinant gene expression/editing. Indeed, from seeding cells to collecting confocal microscopy images, the entire workflow described herein can be completed under 3 h. Third, they also circumvent limitations of other complementary techniques: immunocytochemistry requires sample fixation prior to analysis, and the expression of recombinant genes can alter the subcellular localization of the modified proteins. Current limitations for these first generation probes include: the localization of the probe was found to differ between live cells vs fixed cells; and integrins that are labelled are also inhibited by design, thereby preventing the study of active integrin function in cells. ## ■ CONCLUSION We reported two new molecules that label endogenous integrin αVβ5 proteins in living cells. In this proof of concept study, we described the synthesis of fluorescent imaging probes based on the selective integrin inhibitor cyclo(RDGf(Me)V): Pacific Blue-linked probe 2 (cilengitide-PB405) and Alexa568-linked probe 3 (cilengitide-A568), both modified at the D-phenylalanine residue. We demonstrated that the probes maintain their activity toward integrin αVβ5 using a protein binding assay, as well as a whole-cell functional assay. The larger of the two probes, 3, was found to have a KD of 0.18 ± 0.13 µM with isolated integrin αVβ5, which confirms that modifying cilengitide (1) with fluorophores on its aromatic residue alters its potency only minimally. Importantly, we confirmed that the inhibitory function of the probes was equal to or stronger than cilengitide itself in a new phagocytotic functional assay. Finally, given that cilengitide binds both integrin αVβ5 and αVβ3 isoforms, it is highly likely that probes 2 and 3 can also be used to study integrin αVβ3 in living cells. Both integrins serve similar functions and are often co-expressed. ## ■ SIGNIFICANCE Little is known about integrin αVβ5's regulation, localization, or associated proteins in glia-mostly due to a lack of methods to track these proteins in their native environment. To elucidate integrin αVβ5's contributions to physiology, it is imperative to characterize integrin dynamics at the molecular level. Through rational design, we synthesized integrin-selective blue and red fluorescence imaging probes that are: easy to use, rapidly applicable, and organism-independent. These probes offer a well-needed alternative to recombinant genetic fusion-highly specific, but a lengthy technique to implement, currently the only method that can label endogenous proteins for longitudinal analyses through time. Access to practical means of monitoring integrin αVβ5 in live-cell systems will help define their role in molecular pathways contributing to glia-mediated synaptic elimination and neuron phagocytosis associated with cognitive deficits in neurodegenerative diseases. ■ SUPPLEMENTAL INFORMATION Supplementary information includes experimental synthetic procedures, characterization, and spectroscopic data; it can be found with this article online. ## ■ AUTHORS CONTRIBUTIONS V.K. and W.S. are co-first authors and contributed equally to this paper. V.K., W.S., and F.M. conceived the study and wrote the manuscript. V.K. synthesized all compounds. W.S. and V.K. conducted experiments and interpreted the data.

chemsum

{"title": "A CycloRGDf(Me-V) Analog as Chemical Probe to Study Integrins Function in Living Cells", "journal": "ChemRxiv"}

quantum_dynamics_study_of_energy_requirement_on_reactivity_for_the_hbr_+_oh_reaction_with_a_negative

## Abstract: A time-dependent, quantum reaction dynamics approach in full dimensional, six degrees of freedom was carried out to study the energy requirement on reactivity for the HBr + OH reaction with an early, negative energy barrier. The calculation shows both the HBr and OH vibrational excitations enhance the reactivity. However, even this reaction has a negative energy barrier, the calculation shows not all forms of energy are equally effective in promoting the reactivity. On the basis of equal amount of total energy, the vibrational energies of both the HBr and OH are more effective in enhancing the reactivity than the translational energy, whereas the rotational excitations of both the HBr and OH hinder the reactivity. The rate constants were also calculated for the temperature range between 5 to 500 K. The quantal rate constants have a better slope agreement with the experimental data than quasi-classical trajectory results.The title reaction HBr + OH → Br + H 2 O has attracted great interest with many experimental and theoretical studies during the past several decades. From the practical aspect, this reaction plays an important role in atmospheric chemistry because it produces bromine atoms, and the bromine atoms can very effectively destroy the ozone by a catalytic cycles in the stratosphere:In addition, the reaction also plays a key role in combustion chemistry as some brominated compounds act as fire retardants.From experimental studies, there is a number of measurements of rate constants mainly at room temperature (298 K) [1][2][3][4][5][6] . Moreover, the rate constants were also measured at the temperature ranges 249-416 7 , 23-295 8 , 76-242 9 , 230-360 10 , 120-224 11 , 20-350 12 , 53-135 K 13 , and at a high temperature of 1925 K 14 . Among them, four studies 4,10,11,13 have also measured the rate constants for the isotopomers system and found the primary kinetic isotope effect (KIE) is independent of temperature between 53 and 135 K 13 . The results of these investigations reveal that the HBr + OH reaction's rate constants are extremely negative temperature-dependent below 150 K and nearly independent temperature between ~400 K and room temperature. Furthermore, Butkovskaya and Setser 15 studied the vibrational distributions for H 2 O, HOD and D 2 O produced in reactions of OH and OD with HBr and DBr. Che et al. 16 observed the negative collision energy dependence of reaction cross section for the HBr + OH/OD reaction in a crossed molecular beam experiment. Tsai et al. 17,18 reported the orientation dependence of the Br formation and found that O-end attack is more favored for this reaction.From theoretical studies, Clary et al. 19 provided the upper limit of rate constant for HBr + OH at low temperatures and predicted a maximum rate constant with the value of 3.5 × 10 −10 cm 3 molecule −1 s −1 at 20 K, using the statistical adiabatic capture theory with a long-range barrierless electrostatic interaction potential. After that, Clary et al. 20 reported a three-dimensional quantum scattering calculation with the rotating bond approximation on a simple potential energy surface (PES) based on a LEPS function and an accurate H 2 O potential. The reaction cross sections are found to be dependent on (2j + 1) −1 , where j is the initial rotational quantum number of OH. And the calculated rate constant has a − T 1/2 dependence at low temperatures. Furthermore, Nizamov et al. 21 readjusted the LEPS PES 20 to fit the experimentally measured H 2 O vibrational energy and the thermal rate constant, they performed a quasi-classical trajectory (QCT) study on the mechanism for excitation of the bending mode and isotopic effects on the energy disposal. In 2001, Liu et al. 22 investigated the dynamic properties of the hydrogen abstraction reaction HBr + OH over a wide range of temperatures 23-2000 K, by employing the improved canonical variational transition-state theory (VTST) 23 with a small-curvature tunneling correction. Recently, Bowman's group 24 has developed a high-quality, full dimensional PES for the HBr + OH system based on 26,000 high-level ab initio energies. There is a van der Waals (vdW) well in the entrance channel, as well as in the product channel respectively, and a negative energy saddle-point barrier on the PES. They carried out a QCT calculation to obtain the reaction's rate constants over the temperature range from 5 to 500 K, and found an inverse temperature dependence of rates below 160 K and a nearly constant temperature dependence above 160 K. In addition, they also studied the reaction cross section, energy disposal and rate constant for the isotopomers reaction DBr + OH 25 . In 2015, Ree et al. 26 reported the temperature dependence of the title reaction using analytic forms of two-, three-, four-body and long-range interaction potentials in a QCT calculation over the temperature range of 20-2000 K. In 2016, Coutinho et al. 27 investigated the stereodirectional dynamics of the title reaction as the prominent reason for the peculiar kinetics on a multidimensional PES mechanically generated on-the-fly 28 . Till now, there have been no full-dimensional, quantum dynamics studies on the HBr + OH reaction. Thus, in this paper, we carry out the first, full-dimensional, quantum dynamics time-dependent, wave-packet study on the PES developed by Bowman's group. Our purpose of the present work is to (1) calculate the thermal rate constants over the temperature range of 5-500 K and compare our six degrees of freedom (6DOF) results with experiments and the QCT results 24 , see the relationship of the rate constants with the temperature; (2) investigate the energy efficiency of the translational, vibrational, and rotational energy on a negative-energy barrier. In recent years, studies on energy efficacy rules for more than three atoms systems show there does not exist a unified rule on the energy efficacy to reactivity regarding the location of the transition states. For example, the O + CH 4 /CD 4 /CHD 3 reaction with a slightly late barrier, studies on the reactions indicate that the translational energy is more effective than all the vibrational motions in surmounting the slightly late barrier. Similar to the O + CH 4 reaction, the reaction H + CH 4 also has a slightly late barrier. However, the quantum dynamics calculations 32,33 show that the vibrational energy is more efficient in promoting the reaction than the translational energy. The title reaction HBr + OH has a large exoergicity with an early barrier, however, the early barrier is − 0.52 kcal mol −1 lower than the reactant on the PES. There has been no quantum reaction dynamics studies before on the energy efficacy for the negative early barrier. Since the ground-state energy of the reactant is already higher than the barrier height, there is no barrier for the reactant to surmount, one wonders whether any form of the reactant energy (translational, vibrational or rotational energy) is equal to enhance the reactivity; if not, it would be interesting to find the energy efficacy in surmounting this negative early barrier and to see what energy efficacy rule governs this reaction system. ## Results and Discussion Vibrational excitation of HBr. Figure 1(a) and (b) give the integral cross sections' (ICSs) comparison for the first four vibrational excitation states of HBr (v 1 , j 1 = 0) with OH (v 2 = 0, j 2 = 0) at ground state as a function of translational energy and total energy, respectively. To converge the ICSs for the initial states: (v 1 = 0, j 1 = 0), (v 1 = 1, j 1 = 0), (v 1 = 2, j 1 = 0) and (v 1 = 3, j 1 = 0), 200, 230, 260 and 260 partial waves are calculated, respectively. For the partial waves of J ≤ 100, the reaction probability for every partial wave was calculated explicitly, and the reaction probabilities for different partial waves of J > 100 were computed using the J-shifting method 34 with a J interval of 5. The standard centrifugal sudden (CS) approximation 35,36 was employed in calculation for J > 0. This figure shows the cross sections decrease as the translational energy increases. Especially for the excited state ICSs, v 1 = 1, 2, 3, the cross sections decrease significantly about 75% as the translational energy goes from 0.05 to 0.3 eV. On the other hand, the ground state cross section drops slower only about 45% for the same energy. Among these four cross sections, the ground state ICS is the smallest. For the translational energy lower than 0.1 eV, the amplitudes of the HBr v 1 = 1, 2, 3 ICSs are about 3~4 times bigger than that of ground state; even for the collision energy is larger than 0.1 eV, the three excited-state ICSs are also about 2 times bigger than the ground state's. As this reaction has a negative barrier height, even the ground state energy is higher than the barrier, it is surprising to see that the higher of the excited state the more reactive of this reaction. In order to see the energy efficacy of the vibrational energy of HBr on the reactivity. We plot the ICS ratios of the HBr, σ(v 1 = 1)/σ(v 1 = 0), σ(v 1 = 2)/σ(v 1 = 1) and σ(v 1 = 3)/σ(v 1 = 2), on the basis at an equivalent amount of total energy in Fig. 2. The ICS ratio of σ(v 1 = 1)/σ(v 1 = 0) has a maximum ~10.8 at the initial translational energy 0.037 eV, then rapidly drops to 3.5 at the 0.145 eV until it reaches to 2.2 at 0.283 eV. In the whole energy range, the ratio is always bigger than 1.0, which means that vibrational energy is more effective to promote the reaction than translational energy. Furthermore, the ratio of σ(v 1 = 1)/σ(v 1 = 0) is much bigger than those of σ(v 1 = 2)/σ(v 1 = 1) and σ(v 1 = 3)/σ(v 1 = 2), and the ratio of σ(v 1 = 3)/σ(v 1 = 2) is just slightly larger than σ(v 1 = 2)/σ(v 1 = 1). Bases on the above results, we can conclude that the vibrational excitation from the ground state to the first excited state is the most effective one to promote the reactivity; however, there is no much reactivity change as the vibrational quantum numbers increase from v 1 = 1 to v 1 = 2 and from v 1 = 2 to v 1 = 3. Nevertheless, the comparison of the ICS ratios on the equal amount total energy indicates that the vibrational energy of HBr is more effective than translational energy on promoting the reactivity for this negative-barrier reaction. Vibrational excitations of OH. In Fig. 3(a) and (b), we also compares the ICSs for those vibrational excitation states of OH with HBr (v 1 = 0, j 1 = 0) at ground state as a function of translation energy and total energy, respectively. There are 215, 210 and 210 partial waves needed to converge the vibrational excitation state of OH (v 2 = 1, j 2 = 0), (v 2 = 2, j 2 = 0), (v 2 = 3, j 2 = 0), respectively. The results show that ICSs almost stick together in regard to the translation energy, and the ICSs decrease as the translational energy increases. The reaction path of the PES 24 we used here has a vdW minimum at the entrance channel with a structure of the O-end of OH linked to HBr, HOHBr. This means the favorite route of this reaction is the two reactants enter the entrance vdW minimum to form HOHBr, then scale the transition state to make the reaction occur. This has been confirmed by the crossed beam scattering experiment 17,18 by Tsai et al. They found the orientation dependence for the title reaction that the reaction is favored by OH re-orientating its O-end to face the HBr. Since the barrier height respect to the vdW minimum is 0.12 eV on the PES 24 , thus as seen in Fig. 3(a), the reaction ICSs are almost the same at the translational energy larger than ~0.15 eV because the reactants would overpass the vdW minimum without reorientation in this large energy range; however, for the translational energy less than ~0.15 eV, the reactants will enter the vdW minimum to re-orientate themselves then surmount the barrier, therefore the ICSs are bigger for translational energy smaller than 0.15 eV. In order to investigate the vibrational energy efficacy of the OH, we need to check the ICS ratio of the excited state over the ground state in terms of equal amount of the total energy. Figure 4 gives the ICS ratio, σ(v 2 = 1)/σ(v 2 = 0), in terms of translational energy at the equal amount of total energy. This figure shows that the ratio is larger than 1 in the whole energy range and the ICS ratio curve of σ(v 2 = 1)/σ(v 2 = 0) is very similar to the HBr's. The ratio is also inversely proportional to the energy and has a rapid decline at lower energy. This shows that the vibrational energy of OH is also much more efficient than the translational energy in promoting the reaction. Compared with the F + CH 4 37 and F + H 2 O reaction 38 , the title reaction HBr + OH has some similarities with them in regard to the PES. The three reactions all have an early saddle point located in the reactant channel, a vdW well in the entrance channel, and a relatively deep vdW minimum in the product valley. And the difference is that the HBr + OH reaction has a negative energy barrier (− 0.52 kcal mol −1 ) on the PES, while F + CH 4 barrier height is 0.5 kcal/mol and F + H 2 O's is 3.8 kcal/mol. For the early barrier reaction HBr + OH, our calculation shows both the vibrational energies of HBr and OH are more effective than translational energy in enhancing the reactivity. While for the two early barrier of F + H 2 O 39 and F + CHD 3 40,41 reactions, the study of F + H 2 O shows that the vibrational energy of H 2 O has higher efficacy in enhancing the reactivity than the translational energy; however, the vibrational excitation of C-H stretching motion of CHD 3 hinders the overall reaction rate. Thus, these investigations further prove that, Polanyi rules 42 in which the translational energy is more effective to raise the reactivity for the early-barrier tri-atomic reaction systems cannot be extended to the ploy-atomic reaction systems. Nonetheless, this study shows that, for this negative, early barrier reaction, the vibrational energy is more efficient than the translational energy in promoting the reactivity. ## Rotational excitations ICSs. In addition, we also studied the rotational excitations on the reactivity for this reaction. For all the excited rotational ICSs' calculations, 200 partial waves were needed to converge the excited rotational ICSs, and the CS approximation 35,36 was also used to calculate the partial wave reaction probabilities for J > 0. Similar to the vibrational ICSs' calculations, for the reaction probabilities of partial waves for J ≤ 100, every J partial wave was calculated; and for J > 100, the reaction probabilities were computed using the J-shifting method 34 with a J interval of 5. Figure 5 presents the first five rotational excited ICSs of the HBr (v 1 = 0, j 1 ) with OH (v 1 = 0, j 1 = 0) at the ground state as a function of translational energy. As the figure shows that the ground state has the largest ICS among the 5 ICSs, and as the rotational quantum numbers j 1 increases, these excited rotational states' ICSs significantly decrease. So the rotational-excited modes of the HBr greatly hinder the reactivity. In Fig. 6, the first four ICSs of rotational excitations and the ground state of the OH were compared. It is shown that overall the OH rotational excitations greatly inhibit the reactivity, and the faster of the rotation, the smaller of the ICS. This can be explained due to the fact that the OH plays a receiver role, the faster rotation of OH will further add difficulty for H atom in HBr to attack the O atom in OH. And our 6DOF quantum dynamics results here are in agreement with the three-dimensional quantum scattering calculations by Clary et al. 20 who found the reaction cross sections are proportional to (2j + 1) −1 , where j is the initial rotational quantum number of OH. Overall, the rotational excitations, both HBr and OH, hinders the reactivity. This indirectly proves Tsai et al. molecular beam study 17,18 that Br formation of this reaction has orientation dependence which favors the O-end attack. On one hand, the faster rotation of HBr will make H in HBr cannot attack O-end easily; on the other hand, the faster rotation of OH will make O-end having difficulty to receive the H in HBr. Therefore, both the rotational excitations of HBr and OH hinder the reactivity. Thermal rate constants. By summing over all the ro-vibrational states of HBr and OH, we can obtain the 6DOF cumulative reaction probability (CRP). In order to converge the rate constants up to 500 K to compare with the experimental results , the reaction probabilities of the ground vibrational state including 6 HBr rotational excitation states (j 1max = 5) and 3 OH rotational excitation states (j 2max = 3) were calculated. The thermal rate constants are obtained using the J-K shifting rate expression from equation ( 8) in the Method Section. Note, in the QCT calculation by Bowman's group 24 , they compared their calculated rates, with the OH spin-orbit coupling (RR/SO rates), without the spin-orbit coupling(RR/nSO rates), and with a fully coupled partition function (Coupled rates), with the experimental measured ones. They found that, with the spin-orbit coupling neglected, the RR/nSO rates have the best overall agreement with the experimental results. Therefore, in the current study, we neglected the spin-orbit coupling to calculate our 6DOF quantal rate constants. In Fig. 7, our 6DOF results are compared with experimental and QCT results (RR/SO, RR/nSO, Coupled) 24 . And the QCT calculations of RR/SO, RR/nSO and Coupled are obtained from three different approaches to treat the reactant OH rotational and associated electronic partition function. As the comparison shows, our 6DOF rate constants have a good agreement with the experimental data and demonstrate a negative temperature dependence, which is in agreement with the experimental ones and the QCT results. However, at the very low temperature range upto about 50 K, our 6DOF results are bigger than the experimental and QCT results. Nonetheless, in general, our 6DOF results have a better agreement with the slope of the experimental data than the QCT results. In addition, our 6DOF results give a maximum at about 15 K just as RR/SO and RR/nSO results do. This agrees with Clary et al.'s 19 prediction that a maximum rate constant should appear at 20 K. On the whole, Fig. 7 shows that our 6DOF rate constant has a better slope agreement with the experiments than the QCT results except at the extreme low temperature. ## Discussion In this work, we carried out a 6DOF quantum reaction dynamics, time-dependent wave packet propagation approach to study the HBr + OH → Br + H 2 O reaction system on the PES developed by Bowman's group. This is the first, full-dimensional, quantum dynamical study on the title reaction. For the HBr + OH reaction system with a negative-early barrier, this study shows that not all forms of energy are equal in enhancing the reactivity. Even the ground state energy of the reactant is higher than the barrier, the calculation still shows that vibrational excitations of both the HBr and OH vibrational enhance the reactivity. Furthermore, the HBr and OH vibrational excitations are more effective in enhancing the reactivity than the translational energy. We also studied the rotational excitations of HBr and OH. The results show that both the rotational excitations hinder the reactivity. This is due to the fact that the faster rotation of HBr makes H having difficult to attack O in OH; and the faster rotation of OH makes O having difficult to receive H from HBr. Comparing with other two early barrier reaction systems, F + H 2 O and F + CHD 3 , we can see there are no general rules so far on the energy efficacy for the ploy-atomic reaction systems as the Polanyi rules do to the tri-atomic systems. We think this is due to the complexity of the PESs of the poly-atomic systems with vdW wells usually in both the reactant and product channels. These wells, especially the entrance channel well before the transition state, might also play a role that cannot be neglected on energy efficacy on surmounting the energy barrier. Furthermore, the comparison of the thermal rate constants between our 6DOF quantum results and the experimental display that our data agree well with experimental measurement except at extreme low temperatures. ## Theoretical Methods 6DOF approach. We performed a full dimensional, 6DOF, time-dependent quantum dynamics study for the HBr + OH → Br + H 2 O reaction. The 6DOF Hamiltonian in the reactant Jacobi coordinates, as shown in Fig. 8, can be written as, where, μ R is the reduced mass of the whole reaction system; R is the center-of-mass distance between HBr and OH, r 1 is the distance of H-Br and r 2 is the distance of O-H; θ 1 and θ 2 are the two Jacobi angles between r 1 and R and r 2 and R, Φ is the torsion angle; J is the total angular momentum operator of the reaction system, j 1 and j 2 are the rotational angular momentum operators for HBr and OH, respectively, j 12 is the coupled angular momentum operator of j 1 and j 2 ; and V 6D is the interaction potential. The vibrational reference Hamiltonians h 1 (r 1 ) and h 2 (r 2 ) are defined as ##  Here V(r 1 ) and V(r 2 ) are the one-dimensional reference potentials for HBr and OH, and μ 1 and μ 2 are the corresponding masses. These potentials correspond to the reactant at the asymptotic region with other coordinates fixed at the equilibrium geometry. The split-operator method 43 is employed here to propagate the wave packet on the full-dimensional ab initio PES for the quantum scattering calculation. And we expand the time-dependent wave-function in terms of the body-fixed (BF) rovibrational eigenfunctions defined in terms of the above reactant Jacobi coordinates. After the time-dependent wave function is propagated into the product region, we perform the standard reactive flux method to extract the initial-state-selected reaction probability. To obtain the initial-state-selected ICS, we first 1/2 is the wave number and E is the translational energy, v 0 and j 0 denotes, respectively, the initial vibrational and rotational quantum numbers, K 0 is the projection of J onto BF z axis of the reaction system. Cumulative reaction probability and thermal rate constant. The CRP N J=0 (E) is defined as the sum of all the initial sate selected ro-vibrational reaction probabilities = P E ( ) Next, the J-shifting method 34 is employed here to calculate the CRP for the nonzero total angular momentum J. The total CRP, N(E) is defined as the sum of CRPs for all the open J and K channels where ‡ E JK is the rigid rotor rotational energy of the reaction system at the transition state. This energy is approximated by the expression for a symmetric top molecule, JK 2 A ‡ and B ‡ are the rotational constants of OHHBr at the transition state. Thus the thermal rate constant can be computed as where Q r (T) is the reactant partition function, which is written as a product of vibrational, rotational, and translational partition functions. The equation ( 7) simplifies under the J-K shifting approximation in terms of Equation ( 5), where ‡ Q rot is the rotational partition function of the reaction system at the transition state. Numerical aspects. To converge the above 6DOF, wave-packet, quantum dynamics calculation, we used the following numerical parameters to expand the wavefunction: for the translational coordinate R from 2.5 to 12.5 bohr, 240 sine basis functions are used, and among these, 150 are used for the interaction region; 30 potential-optimized vibrational discrete variable representation (DVR) points 48 for the r 1 coordinates in the range from 1.6 to 5.5 bohr; 40 spherical harmonic rotational functions are used for θ 1 and 15 for θ 2 , which gives 4896 coupled parity adapted total angular momentum basis. The time-dependent wave packet is propagated for a total time of about 16,000 atomic unit time with a time step of 15 a.u. For the thermal rate constant calculation in equation ( 6), the rotational constants A ‡ and B ‡ of OHHBr at the transition state are 15.46 cm −1 and 0.14 cm −1 . For the reactant HBr and OH's vibrational partition function, the used harmonic vibrational frequencies of HBr and OH are 2525 cm −1 and 3611 cm −1 . For reactant rotational partition function, the HBr's rotational constant is 8.46 cm −1 and the OH's 18.86 cm −1 .

chemsum

{"title": "Quantum dynamics study of energy requirement on reactivity for the HBr + OH reaction with a negative-energy barrier", "journal": "Scientific Reports - Nature"}

single-molecule_photoredox_catalysis

## Abstract: The chemistry of life is founded on light, so is it appropriate to think of light as a chemical substance? Planck's quantization offers a metric analogous to Avogadro's number to relate the number of particles to an effective reaction of single molecules and photons to form a new compound. A rhodamine dye molecule serves as a dehalogenating photocatalyst in a consecutive photoelectron transfer (conPET) process which adds the energy of two photons, with the first photon inducing radical formation and the second photon triggering PET to the substrate molecule. Rather than probing catalytic heterogeneity and dynamics on the single-molecule level, single-photon synthesis is demonstrated: the light quantum constitutes a reactant for the single substrate molecule in a dye-driven reaction. The approach illustrates that molecular diffusion and excited-state internal conversion are not limiting factors in conPET reaction kinetics because of catalyst-substrate preassociation. The effect could be common to photoredox catalysis, removing the conventional requirement of long excited-state lifetimes. ## Introduction Photosynthesis is the archetypal photocatalytic process. Having evolved from primordial life over billions of years, the conversion of sunlight into chemical energy remains enigmatic at once in its elegance and complexity. Whereas nature combines the energy of multiple photons to drive the conversion of carbon dioxide and water into carbohydrates, even the simplest artifcial models of consecutive photoelectron transfer (conPET) synthesis have proven challenging to realize in the laboratory. Most photocatalysts involve expensive heavy-metal elements, but recently, the potential of hydrocarbon dyes in organic photocatalysis has emerged. First reports have shown that even common organic dyestuffs such as perylene 12 or rhodamine function as effective organic photocatalysts. Since one of the goals of photocatalysis is to achieve cheap large-scale conversion of materials, single-molecule techniques have received only limited attention as an avenue to exploring and optimizing catalytic efficiency. 16 But since, ultimately, photocatalysis is a molecular process, only microscopic spectroscopic techniques can provide truly mechanistic insights for quantumchemical models. The main focus to date of the technique in the context of photocatalysis has been on exploring the spatial and temporal heterogeneity of reactions involving either single-photon mediated processes, 26 or using chemical conversion of a dye molecule to track catalytic activity. In addition, single-molecule techniques have proven versatile in imaging protein-based reactions, and single-electron transfer events in general. 36,52 Little attention has been paid to actually driving chemical reactions on the single-molecule level, with most prior interest directed at the potential of scanning-probe techniques in electrically catalysed reactions for lithographic applications. Few experiments illustrate the particle nature of light more directly than single-photon counting. Passing the fluorescence of a single molecule through a semi-transparent mirror, a beam splitter, with single-photon detectors on either side will give rise to a pronounced anticorrelation in time between the two detectors: photon antibunching occurs, since the same photon cannot be picked up by both detectors. 42 This antibunching arises on timescales of the excited-state lifetime of the molecule, i.e. typically several nanoseconds. On longer timescales, the opposite effect occurs: the fluorescent molecule undergoes quantum jumps between bright and dark states, for example between the singlet and the triplet manifolds of the excited state, leading to bunching of photons in time. 43 This cycling between emissive and non-emissive states of the fluorophore provides crucial insight into the molecular quantum jumps responsible for the photosynthetic reaction. 44 Here, we exploit the versatile method of single-molecule spectroscopy to probe the conPET process, one photon at a time. Fig. 1a illustrates a prototypical model process of aqueous organic photocatalysis exploiting consecutive photoelectron transfer (conPET). 13 Absorption of a photon of energy hn 1 by a rhodamine-6G (Rh6G) dye molecule leads to the formation of an excited singlet state. This singlet can undergo either radiative relaxation to the ground state by fluorescing, convert into a triplet by intersystem crossing, or interchange charge with reductants to form a radical. If the latter two processes occur, the dye molecule will cease to fluoresce for a short period of time, on the order of a few microseconds up to milliseconds. 52 Addition of a reducing agent, ascorbic acid, promotes formation of the rhodamine radical from either the singlet or the triplet state. Since Rh6G in water is cationic, 45 electron transfer from the reducing agent will form the neutral Rh6G radical Rc. This radical is characterised by a certain lifetime, and will ultimately relax to the cationic ground state by shedding the additional electron to the environment. The reduction potential of the Rc ground state of 1 V vs. SCE 13 is too low for electron transfer to occur to a substrate molecule to cleave stable bonds in aryl halides, such as in the dehalogenation of 2-bromobenzonitrile. Such a reaction requires a reduction potential of 1.9 V vs. SCE. 13,46,47 The additional energy necessary to achieve this is made available by re-exciting the radical with a second photon of energy hn 2 . The reduction potential of the excited radical state of Rh6G, Rc*, is 2.4 V vs. SCE, 13 which is sufficient for dehalogenation of the substrate. The second photon, in combination with the electron transfer to the substrate, therefore removes the additional electron from the dye radical, returning the dye to the ground state and thereby reactivating the S 1 / S 0 fluorescence cycle. The waterbased mechanism proposed here is, in principle, analogous to reaction cycles recently described in organic solvents. 13 Fig. 1b states the synthetic-scale C-H aromatic substitution reaction of 2-bromobenzonitrile in water, using a reaction mixture containing the dye, substrate, and reducing agent, along with an additional trapping agent, N-methylpyrrole. The reaction occurs under continuous illumination with two light-emitting diodes (LEDs). The conversion yield after 24 hours determined by gas chromatography (GC) is 94%. Chromatograms of the product of this reaction and several control reactions are shown in Fig. S1 of the ESI. † This simple cycle constitutes one of the frst reports of a C-H arylation by an organic dye in water and is therefore likely to be of interest in a range of aqueous biochemical reactions. 13 Note that the additional trapping agent is only required in the ensemble reaction, where the product yield is monitored, and not in the single-molecule experiments, where the dye acts as the reporter on the reaction. Since the absorption spectrum of the radical is broad, the conPET cycle appears to work with a range of different photon energies. For experimental reasons, different light sources and wavelengths are used for the singlemolecule and synthetic-scale reactions. Even though this conPET cycle apparently works, it is not at all clear how a conventional dye molecule actually enables consecutive photoredox catalysis. Internal conversion is the most efficient process of energy dissipation from higher-lying states in molecules, and a photoexcited radical is expected to shed excess energy to the environment within a few hundred femtoseconds, as documented by transient absorption spectroscopy. 48 Such ultrafast energy dissipation inhibits intermolecular photoreactions and would certainly prevent any diffusion-driven process from occurring in solution. ## Single-molecule imaging In order to study the conPET mechanism on the single-molecule single-photon level, the photocatalytically active dye molecules have to be immobilized on a surface to prevent diffusion in the solvent. 49 We therefore tether the dyes to DNA oligomers, functionalised with biotin-streptavidin linkers, as sketched in Fig. 2a. These linkers bind to biotinylated bovine serum albumin (BSA) covered glass substrates at sufficiently low concentration so that they can be resolved individually in ## (a) A cationic rhodamine 6G (Rh6G) dye molecule in phosphate-buffered saline (PBS) is excited by a photon of energy hn 1 and is subsequently reduced by ascorbic acid (AscA) to form a radical. A second photon hn 2 excites the radical, leading to PET to the halogenated substrate 2bromobenzonitrile (BrBN). Note that since in the single-molecule experiments it is the dye and not the product yield which is monitored, a trapping agent is not needed for this reaction. (b) Synthetic-scale C-H aromatic substitution of 2-bromobenzonitrile with an N-methylpyrrole trapping agent in an aqueous mixture of dye, substrate, reducing and trapping agent under two-colour LED illumination in the green (hn 1 ) and blue (hn 2 ). The conversion yield after 24 hours as determined by gas chromatography is 94%. Note that synthetic-scale reactions are usually carried out with LEDs rather than lasers. Lasers are necessary to focus light tightly in single-molecule experiments. Since the absorption spectrum of the radical state is broad, the reaction works for both blue wavelengths (hn 2 ) of 405 nm (laser) and 455 nm (LED). a confocal fluorescence microscope. Fig. 2b indicates the anticipated level scheme of the Rh6G dye molecules. Fluorescence is observed from the single molecules as long as they cycle between S 0 and S 1 states. Excursions to the triplet or the radical state lead to a disruption of this cycle and inhibit fluorescence. The triplet can relax back to the ground state by reverse intersystem crossing with a rate of k T1 ISC ; or else be reduced to form the radical of the dye molecule. In the presence of a reducing agent, the singlet can also be reduced to form the radical, which can re-oxidise at an intrinsic rate, returning the dye molecule to its ground state; or else the radical can be photoexcited again to form Rc*, which can transfer its electron to the substrate molecule 2-bromobenzonitrile. To test the feasibility of tracking the conPET cycle on the single-molecule level, we plot the fluorescence of a single tethered Rh6G molecule in Fig. 2c as a function of time, binned in intervals of 5 ms, with alternating application of hn 2 . The fluorescence intensity, stated in terms of the photon count rate, appears as bursts of approximately equal strength, separated by prolonged intervals of darkness. The average photon count rate, binned over intervals of 0.5 s, is superimposed in the plot as a red line. As the dye radical is re-excited by hn 2 , the number of fluorescence spikes increases and the average brightness of the single molecule (red line) doubles. The height of the individual spikes remains almost constant, implying that it is not the number of photons absorbed by the dye which increases upon simultaneous excitation at two wavelengths. Rather, the intermittency between bursts is shortened. Panel d plots a twosecond interval of the fluorescence trace of panel c, revealing distinct "on" and "off" periods of the molecular fluorescence. Such intermittency can be used to analyse the fluorescence to extract characteristic timescales s on and s off . Fitting directly to fluorescence intermittency traces is cumbersome and limited in time resolution by the fnite photon count rate. A versatile quantifcation of the fluorescence dynamics is instead offered by a single-photon correlation analysis of the fluorescence intensity. 50 As indicated in panel e, the correlation is computed by calculating the self-convolution, i.e. the time average of the product of the trace with itself, shifted by a temporal offset of Ds. Fig. 2f shows the result of such a typical cross-correlation, plotted on a logarithmic time axis. The correlation can be ftted with a single-exponential function of the form , where A is the correlation amplitude, s corr is the characteristic decay time of the correlation, and the "on" and "off" times of the molecular fluorescence are related by s on ¼ s corr (1 + 1/A) and s off ¼ s corr (1 + A). 50 By adding up s on and s off , we determine the single-molecule turnover frequency TOF SM ¼ 1/(s on + s off ). This number of cycles which one single dye molecule undergoes through the dark state sets the upper limit for synthetic-scale TOF. Details of the fluorescence microscopy, including the background correction procedure, are summarized in the ESI. † ## Single-molecule photon-correlation spectroscopy We analyse the photocatalytic cycle using fluorescence intensity correlation spectroscopy. We stress that this analysis is only possible on the single-molecule level, since in the ensemble the molecular excursions to the dark state and the associated fluctuations in fluorescence intensity are averaged out. Each singlemolecule fluorescence-intensity trace gives an individual photon correlation curve. To account for the statistical variation between different single molecules, we consider the median value of one hundred single-molecule correlation curves for each value of Ds, plotted with exponential fts in Fig. 3. We begin in panel a by examining the fluorescence correlation in nitrogen-saturated phosphate-buffered saline (PBS) for the case of excitation with photon energy hn 1 . Under these conditions, the regular transitions of the dye molecule to the triplet manifold give rise to a well-defned "off" time, which can be attributed to the triplet-state lifetime or the lifetime of a radical formed out of the triplet. The temporal excursions to such a dark state are indicated in the cartoon to the right, with s on ¼ 2.22 AE 0.02 ms, s off ¼ 5.99 AE 0.09 ms, and TOF SM ¼ 123 s 1 . The Rh6G triplet state is quenched by molecular oxygen, by saturating the solvent with air. When this quenching occurs, the dye molecule cycles solely between ground and excited singlet state: no amplitude exists in the photon correlation signal in panel b, implying the absence of a dark state. To monitor the molecular dynamics relating to PET, we carry out the following experiments under conditions where the dark state is stabilised, i.e. under nitrogen saturation. Panel c plots the photon correlation with addition of the reducing agent ascorbic acid. Now, the molecular dark state must be attributed to the radical with s on ¼ 1.73 AE 0.03 ms, s off ¼ 21.3 AE 0.5 ms, and TOF SM ¼ 44 s 1 . Adding the substrate compound 2-bromobenzonitrile in panel d has no effect on the correlation and the associated timescales. In contrast, exciting the radical with hn 2 in the absence of the substrate in panel e promotes depopulation of the radical state, shortening the dark-state lifetime to s off ¼ 8.4 AE 0.3 ms, with s on ¼ 1.5 AE 0.05 ms, and TOF SM ¼ 100 s 1 . 53 The dramatic effect arises upon simultaneous addition of the two reactantshn 2 photons and 2-bromobenzonitrile moleculesto the dye catalyst. The correlation amplitude in panel f is suppressed almost entirely, but characteristic "on" and "off" times can still be determined as s on ¼ 2.6 AE 0.2 ms, s off ¼ 4.7 AE 0.4 ms, and TOF SM ¼ 137 s 1 . The additional 37 photocycles per second undergone by the catalyst in the presence of the substrate provide a metric for the overall upper limit of the dehalogenation reaction efficiency. Under these reaction conditions, each single Rh6G molecule dehalogenates 37 2-bromobenzonitrile molecules per second. In order to prove chemical specifcity of the microscopic photocatalytic conPET cycle, it is necessary to demonstrate that the dark state of the dye is not quenched for substrate molecules which cannot be dehalogenated. The obvious material to test this is the non-halogenated compound benzonitrile. Fig. 4a plots the single-molecule correlation signal for the four conditions used in Fig. 3c-f, but with benzonitrile added as the substrate. As before, the correlation is identical with only the reducing agent ascorbic acid added (black curve) and with ascorbic acid and benzonitrile combined (red curve) in the solution. Excitation of the Rh6G radical with hn 2 shortens the dark-state lifetime by returning the dye from the radical state to the ground state (light-blue curve). However, in contrast to the situation in Fig. 3, addition of benzonitrile has no effect on the photon correlation trace (dark-blue curve). We conclude that benzonitrile does not interact with the photocatalyst since addition of it to the solution has no effect on the fluorescence cross-correlation. This conclusion is crucial since otherwise product inhibition of the catalyst would occur by the dehalogenated substrate, disrupting the photon cycling process. Once bromine is cleaved from 2-bromobenzonitrile, the molecule disassociates from the catalyst of Fig. 1. An alternative test of the reaction is performed with 4-chloroanisole, as summarized in Fig. 4b. This substrate is energetically not expected to undergo bond cleavage by the excited radical Rc*, as the reduction potential necessary amounts to 2.9 V vs. SCE. 45 Indeed, in Fig. 4b no effect is seen on the correlation curves of addition of the 4-chloroanisole substrate at the same concentration as that used in Fig. 3. ## Discussion The single-molecule conPET cycle demonstrated here effectively constitutes a single-photon chemical reaction: the frst photon hn 1 , in combination with a reducing agent, generates the photocatalystthe rhodamine radicalwhich subsequently reacts the two "compounds", the substrate 2-bromobenzonitrile and the photon hn 2 . An appealing aspect of the single-molecule single-photon double-excitation scheme is the potential ability to resolve in time the consecutive excitation processes. In a double-pulse experiment, for example, it should be possible to measure directly the lifetime of the photoexcited radical by varying the duration of the hn 2 pulse. In addition, tuning the energy hn 2 in a "photocatalytic action" experiment may even allow time-resolved probing of conformational relaxation dynamics of the catalytically active dye which would offer crucial insight for quantum-chemical modelling of the molecular dynamics of the catalyst-substrate interaction. In this context, we derive two conclusions from the observations. First, the photocatalytic reaction is not fundamentally diffusion limited. Since the lifetime of the photoexcited radical Rc* is expected to be extremely short, 48 the conPET process can only occur if the substrate molecule is preassociated with the dye catalyst. Second, to prevent product inhibition of the catalyst and enable continued observation of the photocatalytic cycle in fluorescence, the reacted species must dissociate from the catalyst to allow the reaction to begin anew. We propose that the radical exerts an attractive force on the substrate, promoting preassociation, and speculate that such an effect may be more common to photocatalytic processes than previously thought. While we cannot conclusively prove that preaggregation does not occur in the dye ground state, we reiterate the observed reduction in turnover number upon dehalogenation of the substrate, implying that interaction with either form of the dye must be weakened. We note that the substrate is an aromatic system with two electron-withdrawing substituents. The interaction of such an electron-poor aromatic should be stronger with the neutral dye radical than with the cationic dye ground state. As discussed above, depending on the protonation balance, the rhodamine ground state may actually be neutral. In this case, the interaction of the dye with the electron-poor substrate would also be stronger in the anionic radical state than in the neutral ground state. Without precise determination of the different contributions from van der Waals interactions, pi-stacking and electrostatics, such arguments, however, remain qualitative. To further explore the microscopic origins of this phenomenon will necessitate the development of time-dependent density functional theory (TD-DFT) techniques which can take into account the strong polarization effects of the surrounding medium. 51 This can be achieved by implementing new theoretical methods to account for the complex excited-state geometry optimization arising from the non-adiabatic molecular dynamics. To arrive at such a microscopic theory of organic photocatalysis it is imperative to have access to truly microscopic experimental data, which only become available on the single-molecule level. An open question is whether the trapping agent N-methylpyrrole used in the ensemble experiments also sticks to the photocatalyst. This could conceivably be expected since dispersive interactions should be of a comparable nature to those of the substrate, but such an association could in turn block the photocatalyst. Given near-unity conversion yields found in the ensemble, such blocking is apparently unlikely. Our crucial conclusion is that in mechanisms which involve preassociation of substrate and photocatalyst, diffusion no longer appears to be the limiting factor so that long excitedstate lifetimes are not necessary to ensure effective photocatalytic transformation. This is an important point since most photoredox catalytic cycles involve long-lived triplet states. Triplets, however, limit the overall catalytic potential since electronic energy is inherently lost to the quantum-mechanical exchange interaction by satisfying Pauli's exclusion principle. Our work therefore encourages a renewed search for materials supporting singlet-based photoredox cycles. The dehalogenation reaction demonstrated here on the single-molecule singlephoton level constitutes a precursor to more complex photocatalytic mechanisms. We expect the cycle to work equally well in forming carbon-carbon bonds, opening up the possibility of multicolour directed synthesis 13 on the single-molecule level. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Single-molecule photoredox catalysis", "journal": "Royal Society of Chemistry (RSC)"}

one-pot_multistep_mechanochemical_synthesis_of_fluorinated_pyrazolones

## Abstract: Solventless mechanochemical synthesis represents a technique with improved sustainability metrics compared to solvent-based processes. Herein, we describe a methodical process to run one solventless reaction directly into another through multistep mechanochemistry, effectively amplifying the solvent savings. The approach has to consider the solid form of the materials and compatibility of any auxiliary used. This has culminated in the development of a two-step, one-jar protocol for heterocycle formation and subsequent fluorination that has been successfully applied across a range of substrates, resulting in 12 difluorinated pyrazolones in moderate to excellent yields. ## Introduction Mechanochemical methods are emerging as an alternative approach to traditional solvent-based reactions for chemical synthesis. Under mechanochemical conditions reactions are performed between neat reagents and do not require a solvent. Processing chemical reactions in such a manner is desirable as reactions are consequently less wasteful and more environmentally benign than the analogous solution-based approaches, especially if the work-up and purification processes can also be made solventless or solvent minimised . As such, there is now a significant number of mechanochemical synthetic transformations reported . However, for the synthetic commu-nity, perhaps the most interesting examples of mechanochemical reactions are not those that are merely solventless but those in which different reactivity or selectivity arises, as well as those that are significantly shorter in reaction time than those conducted in solution. Indeed, there are several examples where reactions are clearly significantly faster under mechanochemical conditions . One of several challenges to be overcome for the further development of mechanochemistry as an up to date tool for synthesis is to gain a better insight into the ability to run multistep proce-Scheme 1: Factors to be considered regarding the physical form in the one-pot two-step mechanochemical procedure. dures. One-pot multistep procedures are particularly efficient, in that the same reaction vessel is used for each step, additional reagents are simply added to the reaction mixture at each stage with no isolation of intermediates or removal of side products . One-pot procedures require the conditions for each step to be compatible with succeeding steps. Typical problems encountered when attempting to multistep reactions include solvent compatibility, or, issues with side products that can inhibit future steps, e.g., by providing access to alternative reaction pathways, poisoning catalysts or altering the pH unfavourably . With regards to mechanochemistry such processing serves to amplify the sustainability metrics by running back-to-back solventless reactions. Multistep mechanochemical procedures have been successfully applied to the synthesis of O-glycosides , bioactive hydantoins , extended iptycenes and organometallics where problems can occur using solutionbased synthesis due to limited solubility. Whilst mechanochemical one-pot procedures offer the inherent ability to overcome the issue of identifying a solvent compatible with several consecutive steps, we envisaged alternative hurdles not previously described with regard to compatibility of chemical form. The state of reagents or chemical form is significant to reactions conducted under mechanochemical conditions, where liquids and solids behave differently. For instance, when liquid components are used it may be critical to add a solid auxiliary that helps the transfer of energy and mass (adequate mixing) throughout the mixture. In many cases, leaving out such an auxiliary material can result in a gum or a paste that does not mix well and results in low reaction conversions. Clearly the presence of such a material may have a knock-on effect on any multistep process. Liquid-assisted grinding (LAG) is another phenomenon that can provide enhancement to the reaction outcome and again should be considered for use in a multistep format . Having recently begun our research programme in the area of mechanochemistry, we were particularly intrigued by the compatibility of differing chemical forms and additives across a two-step, one-grinding jar solventless process. To investigate this we designed a 2-step reaction related to our recent work on liquid assisted grinding effects of the fluorination of 1,3-dicarbonyl compounds, in which the dicarbonyl will initially form a pyrazolone in the first reaction prior to undergoing difluorination in the second step (Scheme 1) . Notably this approach will likely require a grinding auxiliary in the first step where two liquid phases react and will be catalysed by an acid to afford a solid pyrazolone material. This will then be followed by a difluorination reaction between solid-solid reactants, this reaction may perform better in the presence of base in the second step. In this report, we present a systematic approach to finding the optimal conditions, which are most compatible with both steps. Notably, fluorinated pyrazolones have the potential to be useful pharmaceutical or agrochemical products, given the desirable properties that can be obtained on introduction of fluorine to a molecule . However, there have been limited reports on the synthesis of fluorinated pyrazoles, but fluorinated pyrazolones remain poorly studied . ## Results and Discussion Initially the mechanochemical pyrazolone formation was investigated as the first step of the two step process, we opted to keep the ball size, ball number, jar size and jar and ball material as in our previous studies to reduce the number of variables for this analysis . In the first instance, simply milling the two liquids in the absence of an auxiliary material resulted in a poor yield (Table 1, entry 1). Pleasingly, treatment of ethyl benzoylacetate with one equivalent of phenylhydrazine in the presence of sodium chloride afforded the desired pyrazolone product in 66% yield after milling for 10 minutes (Table 1, entry 2). The addition of a grinding auxiliary could play several roles. We propose that the key benefits are related to improved mixing, and aiding in energy transfer, specifically in mechanochemical reactions where the reaction mixture could be described as a gum, paste or liquid. Notably, the comparable reaction under solvent-based conditions (in toluene, under reflux) required 24 hours to achieve a similar yield (Table 1, entry 3). As pyrazolone formation can be catalysed by acid, a screen of both solid and liquid acids was next performed (Table 1, entries 4-9). In general, the weaker carboxylate acids performed better than mineral acids, with the highest yield obtained using acetic acid (Table 1, entry 9). The quantity of acid used was then varied. In general, the yield increased with an increase in the amount of acid used ( the larger amount of liquid altering the texture of the reaction mixture and thus reducing effective mixing. An alternative justification is that at higher acid equivalents in the solid state the 'on-off' protonation of the hydrazine is slow, meaning that the nucleophilicity is greatly retarded compared to lower acid loadings. Nonetheless, considering that the subsequent fluorination step should proceed optimally under basic conditions , the lowest amount of acid which also provided a good yield was thus chosen; 30 μL (Table 1, entry 9). Finally, the reaction time with this quantity of acid was then optimised, whereupon the reaction was found to be complete after 40 minutes producing 92% isolated yield of pyrazolone 1 (Table 1, entry 14). For comparison, these optimal conditions have been applied to a solution-based reaction, resulting in a poorer yield after 24 hours at reflux in toluene (Table 1, entry 17). Having achieved optimal conditions for the first step of the reaction, our attention turned to the second step. Initial investigation of the fluorination of the pyrazolone focused on finding the optimum reaction time for the isolated step rather than two-step, i.e., the pyrazolone material was isolated from step one and purified before subjecting to this second reaction optimisation. With no additives, the fluorination was complete after 2 hours (Table 2, entry 4), notably an extra hour returned no further improvement (Table 2, entry 5). The fluorination reaction studied here proceeds via an enolate which is aromatic and therefore is relatively facile (compared to the fluorination of other heterocyclic systems). Introduction of a mild base, such as sodium carbonate to the reaction vessel Scheme 2: Optimised conditions for the one-pot synthesis. served to enhance the rate of reaction, providing complete conversion after 1 hour (Table 2, entry 6). With an understanding of the second step we then assessed the reaction whilst mimicking aspects of the first reaction in order to look for compatibility of a two-step one-jar process. The most important difference between the two steps is the physical state of the reactants. For the first step (Table 1), both reagents are liquids, and a grinding auxiliary was required to aid mixing and energy transfer. However, for the second step (Table 2), the reagents are solids, and the presence of a grinding agent could have a diluting effect. Indeed, addition of sodium chloride does slow down the fluorination, giving a poorer yield (Table 2, entry 7). Another factor to be explored was the effect of acetic acid on the second step. Again, this resulted in a decrease in yield of the fluorination reaction achievable within a two hour reaction time (Table 2, entry 8). Pleasingly a combination of sodium carbonate with the sodium chloride grinding auxiliary resulted in complete reaction after one hour (Table 2, entry 9). The only compatibility issue remaining was the acid present from the first step. However, as a base improved the reactivity of the fluorination, the final conditions make use of enough sodium carbonate both to neutralise the remaining acid and accelerate the second step. By applying these compatible conditions to the one-pot procedure, the desired fluorinated pyrazolone was isolated in 75% yield (Scheme 2). Scheme 2 also shows the physical state descriptors and photographs of the practical experiment. With suitable conditions in hand, the scope of this one-pot mechanochemical process was explored (Scheme 3). Initially, the scope of β-ketoesters was assessed and the procedure was found to be compatible with both the electron-withdrawing and electron-donating groups. However, a poorer yield was obtained for the electron-withdrawing trifluoromethyl substituent (5). The scope of phenylhydrazines was also briefly investigated, with several examples demonstrating good isolated yields, again an electron-withdrawing trifluoromethyl substituent was an exception to this (7) . For this case, crude 19 F NMR after the first step shows a 41% conversion, suggesting that the pyrazolone formation is the limiting factor in this example. An alkyl β-ketoester (ethyl acetoacetate) was also used, affording methyl substituted difluoropyrazolone 12 in modest yield. Finally, an α-substituted β-ketoester was successfully converted to the pyrazolone before monofluorination using one equivalent of Selectfluor to prepare pyrazolone 13, also in moderate yield. In general the optimised approach seems to apply to a small range of compounds. ## Conclusion In summary, we have developed a one-pot, two-step mechanochemical synthesis of fluorinated pyrazolones. The experiments provide a logical approach to multistep solventless synthesis under milling conditions and more broadly will assist in the conversion of other processes to such a system. After careful consideration of physical form and additive compatibility the final protocol has been successfully applied to the preparation of a small library of 12 difluorinated pyrazolones, several of which are hitherto unreported.

chemsum

{"title": "One-pot multistep mechanochemical synthesis of fluorinated pyrazolones", "journal": "Beilstein"}

multiscale_simulations_identify_origins_of_differential_carbapenem_hydrolysis_by_the_oxa-48_β-lactam

## Abstract: OXA-48 β-lactamases are frequently encountered in bacterial infections caused by carbapenem-resistant Gram-negative bacteria. Due to the importance of carbapenems in treatment of healthcare-associated infections, and the increasingly wide dissemination of OXA-48-like enzymes on plasmids, these βlactamases are of high clinical significance. Notably, OXA-48 hydrolyses imipenem more efficiently than other commonly used carbapenems, such as meropenem. Here, we use extensive multiscale simulations of imipenem and meropenem hydrolysis by OXA-48 to dissect the dynamics and to explore differences in reactivity of the possible conformational substates of the respective acylenzymes. QM/MM simulations of the deacylation reaction for both substrates demonstrate that deacylation is favoured when the 6αhydroxyethyl group is able to hydrogen bond to the water molecule responsible for deacylation, but disfavoured by increasing hydration of either oxygen of the carboxylated Lys73 general base. Differences in free energy barriers calculated from the QM/MM simulations correlate well with the experimentally observed differences in hydrolytic efficiency between meropenem and imipenem. We conclude that the impaired breakdown of meropenem, compared to imipenem, which arises from a subtle change in the hydrogen bonding pattern between the deacylating water molecule and the antibiotic, is most likely induced by the meropenem 1β-methyl group. In addition to increased insights into carbapenem breakdown by OXA β-lactamases, which may aid in future efforts to design of antibiotics or inhibitors, our approach exemplifies the combined use of atomistic simulations in determining the possible different enzyme-substrate substates, and their influence on enzyme reaction kinetics. ## Introduction The World Health organization describes antibiotic resistance as "...one of the biggest threats to global health, food security, and development today." 2 Antibiotic resistance arises naturally and evolved long ago, 3 but its emergence and dissemination have been considerably accelerated by the current excessive use of antibacterial drugs. 4,5 This evolving resistance not only complicates standard medical practices, but also has additional expensive implications e.g. for the global economy and food production. Moreover, we are currently living in the so-called antibiotic discovery void 9 where discovering new and safe antibacterials, especially for Gram-negative bacteria, is difficult, time-consuming, and often unprofitable for big pharmaceutical companies. 10,11 β-Lactam antibiotics offer broad-spectrum antibacterial activity against Gram-negative bacteria and remain the most prescribed drugs in clinical practice. 12 The importance of β-lactams in healthcare has been highlighted by the World Health Organization, which includes multiple different β-lactam antibiotics in their Model List of Essential Medicine. 13 All of these antibiotics contain a four-membered β-lactam ring, which ensures antibiotic binding to penicillin-binding proteins and consequently inhibition of bacterial cell wall biosynthesis. 14,15 Clinically used β-lactam compounds can be divided into four different groups: penicillins, cephalosporins, carbapenems, and monobactams, of which carbapenems play a critical role as potent antibiotics reserved for the most serious Gram-negative infections where alternatives are limited. 16 Emerging resistance against β-lactams is evident, and especially in Gram-negative bacteria, βlactamase enzymes are the main resistance mechanism against these drugs. 17 β-Lactamases block antibiotic action by hydrolysing the β-lactam ring, which impairs efficient antibiotic binding to their ultimate target in cells. The Ambler sequence-based classification divides β-lactamases into four major subgroups: serine-β-lactamases (SBLs) comprising classes A, C, and D; and metallo-βlactamases (MBLs), class B. 18 The hydrolysis mechanism differs between SBLs and MBLs, as SBLs utilise a nucleophilic serine residue and MBLs employ zinc cofactors. 17 Class D SBLs are referred to as OXA (oxacillinase) enzymes, stemming from their activity against the isoxazolyl penicillin oxacillin, 19 and they are currently of interest due to their wide distribution and the ability of many members of the group to inactivate carbapenems. The OXA enzymes include five subgroups of recognised carbapenemases: the OXA-23, OXA24/40, OXA-51, and OXA-58 βlactamases are mainly found in Acinetobacter baumannii, while OXA-48-like β-lactamases are mostly encountered in Enterobacterales. 20 In Enterobacterales, OXA-48 β-lactamases are among the most commonly present carbapenemases in clinical samples. 21 Their activity is relatively specific towards imipenem, but other carbapenem substrates (such as meropenem and ertapenem) are also hydrolysed, albeit slowly. 22 The specific origin of this imipenemase activity is not well established, even though variations in measured hydrolysis rates between point variants of OXA-48 hint at structural moieties contributing to specific hydrolytic phenotypes (Figure 1). In OXA-163, a partial deletion of the β5-β6 loop (Arg214-Pro217) and one amino acid substitution (Ser212Asp) expands the hydrolysis profile to accommodate expanded-spectrum oxyimino cephalosporins (such as ceftazidime) at the expense of efficient imipenem breakdown. 23 Further studies show that the β5-β6 loop plays a role in acquired carbapenemase activity, as engineering the OXA-48 β5-β6 loop into the noncarbapenemase OXA-10 enhances its carbapenemase activity. 24 Conversely, replacing the β5-β6 loop in OXA-48 with that of OXA-18 also alters the measured carbapenemase activity (lower kcat values). 25 Site-directed mutagenesis studies of OXA-48 variants indicate that residue 214 (arginine in the wildtype OXA-48) is essential for efficient carbapenem hydrolysis. 26 In recent years, structural studies have yielded a variety of crystal structures of OXA-48 in complex with carbapenems, which shed new light on the acylenzyme (AC) intermediate state. 1, Intriguingly, although the β5-β6 loop is suggested to influence carbapenem activity, the only interaction observed between the substrate and residues within this loop (Thr213-Lys218) is a water-mediated contact between the imipenem 6α-hydroxyethyl hydroxyl and Thr213. 1,30 Furthermore, bound carbapenem Figure 1. Crystal structures of OXA-48 complexed with carbapenems. Acylenzyme structures of OXA-48 with imipenem (PDB ID 6P97, green sticks) and meropenem (PDB ID 6P98, light pink sticks) show a highly similar binding pose for both substrates, where main differences lie in the orientation of carbapenem C2 "tail" group. 1 The Ω-loop is highlighted in orange, the β5-β6-loop in yellow, and relevant active site interactions with dashed black lines. The carbapenem pyrroline ring is modelled as the Δ2-tautomer in both structures. tail groups (C2 substituents) seem to be dynamic and able to adopt multiple conformations, which suggests they do not form strong, specific interactions with the enzyme active site. 29 The generalized β-lactam hydrolysis mechanism for SBLs consists of acylation followed by deacylation (Scheme 1). 17 Both acylation and deacylation reactions include the formation of a shortlived tetrahedral intermediate (TI) through a nucleophilic attack; the respective TI species collapses to yield either a covalent AC structure (after acylation), or the final hydrolyzed product (after deacylation). In both reactions, the nucleophile (conserved serine (Ser70) in acylation and a water molecule (deacylating water, DW) in deacylation) is activated via proton abstraction by a general base. For OXA enzymes, this general base is a carboxylated lysine residue (Lys73). 31,32 Notably, Lys73 needs to be carboxylated for optimal activity; this carboxylation is reversible and pH dependent, i.e. more carboxylation is observed at higher pH values. 31 At lower pH values, protonation of the Lys73:Nζ would lead to decarboxylation. 33 Based on pH dependence studies of the reaction between OXA-10 and penicillin or nitrocefin, the pKa of the carboxylated Lys73 is expected to be ~5.8-6.2. 31 For carbapenems, the pyrroline ring can undergo Δ2 → Δ1 tautomerization in the AC state, the Δ1 tautomer also having two stereoisomers (R and S). For class A SBLs, the Δ2 tautomer has been suggested to be the catalytically competent form, whereas the Δ1 form would essentially inhibit the enzyme. 34 For OXA-48 enzymes, all three tautomers have Scheme 1. Top: Structures of meropenem and imipenem (with atoms numbered), the 6α-hydroxyethyl group is highlighted in red. Bottom: Deacylation mechanism in OXA-48 with a carbapenem substrate (Δ2 tautomer). Starting from the acylenzyme, the antibiotic is deacylated via tetrahedral intermediate formation (1 → 2), which collapses to yield the hydrolysed antibiotic (3). been observed in AC crystal structures, 1, 28-30 but, based on NMR studies, the hydrolysis product is suggested to be either the Δ2 or R-Δ1 tautomer. 35 Kinetic measurements suggest that for OXA-48-like β-lactamases, deacylation is the ratelimiting step in carbapenem breakdown. 30 These authors suggested that the impaired imipenemase activity in the ESBL-like OXA-163, compared to OXA-48, is due to a larger active site, which would not constrain the substrate in deacylation-compatible conformations. Molecular dynamics (MD) simulations of the non-covalent complexes of OXA-48 and OXA-163 with meropenem and imipenem suggested some differences between the substrates in mobility. However, the measured KM values for OXA-48 with imipenem and meropenem are very similar (according to one assay, 11 and 13 μM, respectively) 22 , which indicates that there is unlikely to be any significant difference in the stabilities of the respective Michaelis complexes. The difference in the inactivation efficiency of imipenem compared to meropenem is thus primarily related to differences in the rate of the deacylation step, and it is therefore essential to consider this reaction when seeking to understand and explain activity differences. To analyse differences in activity for carbapenems in atomistic detail, we here simulate TI formation in deacylation, i.e. the expected rate-limiting step, of both imipenem and meropenem by OXA-48 using combined quantum mechanics/molecular mechanics (QM/MM) simulations. Our simulations support the hypothesis that the AC state arising from carbabenem acylation is dynamic in nature. Further, we identify conformations of the 6αhydroxyethyl group that allow for efficient deacylation. Additionally, active site hydration around the carboxylated Lys73 is observed to affect the calculated free energy barriers for deacylation, as we previously observed hydrolysis of the expanded-spectrum oxyimino cephalosporin ceftazidime by OXA-48 enzymes. 36 Analysis of the reaction simulations shows that efficient carbapenem breakdown results both from a decrease in hydration around carboxy-Lys73, and from subtle changes in hydrogen bonding between the substrate and the catalytic water molecule. These results provide detailed insight into the causes of differences in enzyme activity against different antibiotics, information potentially useful in understanding and combating antimicrobial resistance. ## Methods Computational methods and details of the system setup are described in detail in the Supporting Information (SI). To summarise, models of OXA-48 with imipenem and meropenem were prepared based on corresponding acylenzyme (AC) crystal structures (PDB IDs 6P97 1 and 6P98 1 for imipenem and meropenem, respectively). The ff14SB parameter set was used for the protein, 37 parameters and partial charges for non-standard residues (acylated carbapenems and carboxylated lysine) were derived with the R.E.D. Server. 38 Both systems were energy minimised, heated from 50 K to 300 K (in 20 ps), and their dynamics in the AC state were simulated for 200 ns using Langevin dynamics (collision frequency 0.2 ps -1 ) with a 2 fs timestep. Five independent simulations for each AC system were run. All bonds involving hydrogens were restrained using the SHAKE algorithm. Starting structures for QM/MM 39 modelling were chosen from MD simulations based on visual inspection of the active site hydration pattern and the 6αhydroxyethyl orientation; this orientation was kept from changing during subsequent QM/MM US MD by applying a weak dihedral restraint (except in the case of orientation I). Free energy barriers for the first (rate-limiting) step of deacylation for the different active site conformations were determined from three separate QM/MM umbrella sampling (US) calculations for each conformation. 40 Two reaction coordinates were employed in US, one for the nucleophilic attack and one for the proton transfer, as in previous simulations of deacylation in serine β-lactamases. 36, Sampling time in each window was 2 ps, and DFTB2 (SCC-DFTB) was used as the QM method for regions consisting of 43 and 46 atoms (including link atoms) for imipenem and meropenem, respectively (Figure S1). Free energy surfaces (FESs) were constructed from 399 individual US windows. The weighted histogram analysis method (WHAM) 47,48 was used to construct the free energy surfaces, and the minimum energy paths were analysed using the Minimum Energy Path Surface Analysis (MEPSA) program 49 . All simulations and trajectory analyses were done using the Amber18 software package 50 (pmemd.cuda for MM MD, and sander for QM/MM calculations). ## Results & Discussion Conformational Dynamics of Carbapenem:OXA-48 Acylenzymes AC dynamics for both imipenem and meropenem complexed with OXA-48, each in the 2 (enamine) configuration, were explored by running five 200 ns MM MD simulations for each complex. The first 50 ns were excluded from trajectory analysis to allow time for equilibration. For both carbapenems, the salt bridge between the C3 carboxylate and Arg250 was preserved during simulations, and the C7 carbonyl stayed in the oxyanion hole formed by the backbone amides of Ser70 (nucleophile) and Tyr211. The carbapenem C2 (tail) substituents sampled a range of conformations during the simulations, consistent with previous suggestions based on structural analysis. 29 Clustering the substrate poses based on their heavy atom RMSD yielded four distinct clusters per substrate, which differ by 0.8-1.8 and 1.7-2.5 for imipenem and meropenem, respectively, from the poses in the corresponding crystal structures (Figure S2, Table S1 and SI section Acylenzyme Clustering). The main deviations between cluster centroids and the crystal structure coordinates are due to the positions of the C2 tail groups, as the pyrroline ring and its substituents are anchored in place by hydrogen bonds to the oxyanion hole and the salt bridge with Arg250. However, for the crystal structures 6P97 and 6P98 there is only limited electron density beyond the sulfur atom for both imipenem and meropenem, so the deposited coordinates may not completely reliably depict the actual substrate binding poses. Additional clustering on the active site residues (explained in further detail in the SI) implies that there may be slight differences also in the positions of active site residues Lys73, Tyr157, as well as of the substrate (Figure S3 and Table S2). During MM MD, the carbapenem 6α-hydroxyethyl group was able to rotate to occupy three different orientations, which can be distinguished by the value of the C7-C6-C-O dihedral angle: around 50°, 180°, or 290°, henceforth referred to as orientations I, II, and III, respectively (Figure 3). The 6α-hydroxyethyl orientation affects interactions in the active site, because its hydroxyl group can hydrogen bond either with the DW (I), or with the Lys73 carboxylate (III), or stay close to the crystallographically observed pose, in which its methyl group is positioned next to the DW and points towards Leu158 (II, Figure 2). The starting orientation of the 6α-hydroxyethyl for both carbapenems is II, as in the crystal structures used in model construction. During MD si mulations, this sidechain is free to move and sample all three orientations. For meropenem, orientation I is sampled more than II, while III is sampled only minimally (Figure 2). Conversely, both orientations II and III are sampled more than I for imipenem. The free energy difference between the different orientations of the 6α-hydroxyethyl group was estimated by calculating the ratio of MD trajectory frames corresponding to each orientation (Z), and using ΔG=RTln(Z), where R is the molar gas constant and T the simulation temperature (300 K). For imipenem, the lowest free energy state is orientation II, with slightly higher relative energies of 0.6 and 0.2 kcal/mol for orientations I and III, respectively. For meropenem, orientation I has the lowest free energy, orientation II is slightly higher (0.6 kcal/mol) but orientation III is significantly higher (2.2 kcal/mol). The presence of a methyl group in the 1β-position in meropenem (instead of a 1β-proton in imipenem) may explain the relatively higher penalty for orientation III, as in this orientation the 1β-substituent is located directly next to the 6α-hydroxyethyl moiety. Previously, our QM/MM simulations indicated that Leu158 may play an important role in modulating active site hydration in the deacylation of ceftazidime by OXA-48-like enzymes. 36 The orientation of Leu158 also differed initially between the two OXA-48/carbapenem systems, as the Cβ -Cγ bond has rotated by 180° in the meropenem structure. To study if Leu158 has a similar effect on carbapenem hydrolysis as observed for ceftazidime, its rotamers were first investigated by measuring the χ1 dihedral (N-Cα-Cβ-Cγ) in MM MD simulations. The distribution of sampled rotamers is presented in Figure S4. After the heating phase, Leu158 essentially always rotates away from the crystallographic g-orientation (χ1 ≈ 290°) to the t orientation (χ1 ≈ 180°) to allow space for the 6α-hydroxyethyl moiety, which in turn also permits for two water molecules to form hydrogen bonds with Lys73:OQ1. As the cephalosporin scaffold lacks a functional group similar to the 6αhydroxyethyl group of carbapenems, typically bearing larger substituents in the β orientation at the equivalent 7-position, it is likely that Leu158 does not possess a similar role in carbapenem hydrolysis to that suggested for cephalosporins. ## Deacylation efficiencies for different orientations of the 6α-hydroxyethyl group Because the interactions of the 6α-hydroxyethyl group in the active site have been suggested to play a role in modulating β-lactamase activity towards carbapenems, 32 deacylation free energy barriers were calculated separately for all three orientations of both imipenem and meropenem acylenzymes observed in MD simulations. Starting structures for US were chosen from the 200 ns MM MD simulations following two criteria: that a potential DW was at a suitable distance for nucleophilic attack, and the 6α-hydroxyethyl orientation was that desired. For orientations II and III, the sidechain dihedral was restrained close to the reference values to avoid the substrate changing between orientations during the reaction (no restraints were needed for I, as no sidechain rotation was observed during US). Overall barriers for deacylation were determined by combining sampling from three separate US calculations for each AC conformation (with different starting structures), with standard deviations calculated between the free energy barriers for individual US simulations (Table S3). More details of the US setup and analysis are available in the SI. Calculated deacylation free energy barriers for the ACs formed by imipenem and meropenem with the 6α-hydroxyethyl in each of the three orientations are shown in Figure 3. For all orientations, two barriers are shown, corresponding to two different hydration states around the general base. The lower barrier (in colour) corresponds to a state with only one water molecule hydrogen bonded to Lys73:OQ2, and one or two waters hydrogen bonded to Lys73:OQ1 while the higher barrier corresponds to a state with two water molecules hydrogen bonded to both carboxylate oxygens (Figure 4, carboxylate oxygens labelled in Scheme 1). For all hydration states, the calculated barriers follow the same trend of I < II < III, i.e. the lowest barriers are calculated for orientation I. Notably, the barriers are consistently underestimated due to the QM method used (DFTB2), as is generally found for this method for similar reactions. 42,43 This underestimation likely also causes an underestimation of the stability of the TI compared to the TS (see e.g. the small molecule benchmark calculations the SI section "Benchmarking"), but TI minima were still located in our free energy surfaces (likely due to stabilization by the enzyme environment). As the overall shape of the QM/MM PES is consistent when using DFTB2 or M06-2X/def2-TZVP as the QM method, it is reasonable to expect that the underestimation of TI stability with DFTB2 does not affect trends in reaction barriers (SI section "Benchmarking"). Taking into account an underestimation of ~8 kcal/mol, as indicated by comparison of DFTB2 to SCS-MP2/aug-cc-pVTZ (SI section "Benchmarking"), the lowest barriers are in good agreement with experiment (see further the section "Comparison with experimental data"). Importantly, we expect our protocol for obtaining free energy barriers using semi-empirical QM methods to be a reliable indicator of relative energetic trends between different enzyme active site conformations; we have demonstrated this previously in studies of deacylation of β-lactam acylenzymes for both class A (with meropenem) and D SBLs. 36,43 As discussed above and in ref. 36 , increased hydration around the proton-accepting Lys73:OQ1 impairs deacylation in ceftazidime hydrolysis. A similar phenomenon was observed for carbapenems, with the additional observation that hydration around the second carboxylate oxygen (Lys73:OQ2) also affects reactivity. In orientation I, the average number of hydrogen bonds Lys73:OQ1 accepts during the reaction is 2.4 (± 0.1 standard deviation, calculated from the US minimum free energy path trajectories), which aligns with OQ1 being hydrogen bonded to two water molecules, and partly to Trp157. The two subpopulations with different deacylation barriers arise from a change in hydration around Lys73:OQ2. For the lower barriers in Figure 3, the number of hydrogen bonds to OQ2 is 1.3 (± 0.1) and for the higher barriers 2.2 (±0.1) for orientation I. The lowest calculated deacylation barrier, 8.4 kcal/mol, is for imipenem in orientation I with one water molecule hydrogen bonded to OQ2 and two to OQ1 (Figure 4). The barrier increases by 2.0 kcal/mol when another solvent molecule donates a hydrogen bond to OQ2. For meropenem, the barrier is raised by 4.1 kcal/mol upon introduction of an additional water molecule close to OQ2. The hydration effect around Lys73:OQ2 indicated here has an apparently smaller effect on the calculated barriers than that of hydration around Lys73:OQ1, since the presence of an additional water molecule hydrogen bonded to OQ1 raised the barrier for ceftazidime deacylation by approximately 5 kcal/mol. 36 Orientation II (corresponding to a dihedral angle of between 147°-192° depending on the structure and protein chain) is observed in most OXA-48:carbapenem AC crystal structures. In this orientation, no part of the 6α-hydroxyethyl moiety interacts with either the DW or with Lys73, so the antibiotic may possibly not interfere with the reactive atoms. However, calculated deacylation barriers are increased by 2.1 kcal/mol for imipenem, and by 2.4 kcal/mol for meropenem, when comparing orientation II against I (in which only one water molecule is hydrogen bonded to OQ2). Having two water molecules donating hydrogen bonds to both OQ1 and OQ2 further raises the calculated barriers to 13.6 and 16.0 kcal/mol for imipenem and meropenem, respectively. Therefore, our simulations suggest that II is not the most deacylation-competent AC orientation. Additionally, orientation II might hinder the positioning of the DW in the active site in proximity to the electrophilic acyl carbon. For 93% and 87%, respectively, of the simulation times for the imipenem and meropenem acylenzymes in orientation II, the distance between the AC electrophilic carbon and the closest water molecule falls beyond 4 (an arbitrary threshold distance for a feasible nucleophilic attack; Figure S5). This is likely due to the 6α-hydroxyethyl methyl group partly occupying the space in the binding pocket for the deacylating water molecule, and thereby forcing this water further away from the AC. This is reflected in deposited crystal structures, as a DW candidate that is suitably positioned for nucleophilic attack is not observed in any OXA-48/carbapenem complex. 1, In a previous study (mainly based on molecular dynamics), orientation II was observed to obstruct the positioning of the DW in the active site. 32 Docquier et al. concluded that only a slight repositioning of the methyl group of the 6α-hydroxyethyl sidechain is needed to better accommodate a water molecule at a suitable distance for nucleophilic attack. However, these conclusions are based on a single 10 ns MD simulation, which likely gives insufficient time to sample all available substrate orientations. Based on our MM MD simulations, as well as the calculated free energy barriers, orientation II is less likely to contribute to efficient deacylation of the carbapenem ACs. This is due both to an increase in energy required for deacylation, as well as to a lack of sampling of active site configurations that would be suitable for the AC carbonyl to undergo nucleophilic attack by an incoming water molecule. The largest increase in energetics between the two hydration states is calculated for orientation III, where the barriers increase by 9.6 and 5.6 kcal/mol for imipenem and meropenem, respectively, when the hydration state is changed. For the lower barriers, OQ1 and O Q2 form on average 2.0 (± 0.1) and 1.4 (± 0.1) hydrogen bonds, respectively, for the imipenem and meropenem complexes, while for the higher barriers the equivalent numbers are 2.8 (± 0.1) and 2.1 (± 0.2, data not shown). For the lower barriers, Leu158 has not (yet) rotated from the g-to the t rotamer (Figure S4), as the starting structures were chosen almost directly after the heating phase. The g-rotamer of Leu158 allows space only for the DW positioned near Lys73:OQ1, which was inserted into the active site in the starting model. Further, only one water molecule is donating a hydrogen bond to OQ2. Upon MD equilibration, Leu158 rotates, allowing for active site hydration to change to two water molecules hydrogen bonding to both carboxylate oxygens each. Subsequently, only the 'high barrier' hydration state is sampled. This explains the large increase in activation free energy when comparing the two hydration substates for orientation III, as two water molecules are located near Lys73, as opposed to only one water molecule close to Lys73:OQ2 (as for orientations I and II). Therefore, our simulations indicate that III is the AC orientation that is the least competent for deacylation for the equilibrated system (in which Leu158 has rotated). Experimentally, this AC orientation is seen in the crystal structure of OXA-48 with hydrolyzed, non-covalently bound imipenem (PDB ID 6PK0) 28 , where the hydroxyethyl hydroxyl donates a hydrogen bond to the newly-formed carboxylate group. In our MM MD simulations of the AC, the exchange between 6αhydroxyethyl dihedral orientations is frequent (indicating a low energy barrier). This is probably true also for the hydrolyzed antibiotic, suggesting that rotation of this moiety can occur postdeacylation. Further analysis of the US trajectories reveals that hydration around Lys73:OQ2 correlates with the rotamer of Val120. Valine has three rotamers for the χ1 dihedral (N-Cα-Cβ-Cγ1): the g+ rotamer around 50°, t around 180°, and g-around 300° (Figure 4, Figure S6). In the starting structures for simulations, Val120 is in the t orientation for both carbapenems (for meropenem, partial occupancy for both t and g-rotamers was observed in the deposited structure, but only the t rotamer was used in the computational model building). 1 The rotameric state can switch to either g+ or g-during MD simulations (Figure S6). For the g+ rotamer, one of the methyl groups points directly towards Lys73, which only leaves space for a single water molecule next to Lys73:OQ2; this water is positioned to accept a hydrogen bond from Gln124 and to donate one to Lys73. Conversely, the t rotamer allows for a second water molecule to occupy the space between Lys73 and Val120, and this water molecule is able to donate hydrogen bonds to both Lys73:OQ2 and the Val120 backbone carbonyl. Val120 is part of motif II, which is formed by residues Ser118 -Val120 and is conserved across class D β-lactamases. 32 Together with Leu158, it forms the so-called 'deacylating water channel' in the vicinity of Lys73; this hydrophobic patch partly shields the active site from bulk solvent. 1 For other OXA enzymes, a similar water channel has been proposed to open upon substrate binding to allow for water ingress into the active site and therefore for efficient deacylation. 54,55 For OXA-48, previous comparison of apoenzyme and acylenzyme structures shows that substrate binding shifts Val120 and Leu158 only slightly, and that the water channel is more open than e.g. in OXA-23. 1 Access of water into the catalytic position next to the substrate and Lys73 is necessary for antibiotic hydrolysis, but as we indicate above, any additional solvent in the active site will impair reactivity. In OXA-48, it appears that Val120 (and the specific rotamers that it samples) is an important gateway residue controlling approach of bulk solvent to Lys73:OQ2. Our previous work (on ceftazidime hydrolysis in OXA-48-like enzymes) indicates that Leu158 modulates hydration around Lys73:OQ1. 36 Notably, Val120 is mutated to a leucine in OXA-519, a single point mutant of OXA-48; this mutation results in an increase in measured hydrolysis for some 1β-methyl carbapenems, such as meropenem and ertapenem, but decreased imipenemase activity. Compared to OXA-48, OXA-519 also increases the proportion of β-lactone reaction products, rather than conventionally formed ring-opened species, hydrolysis products of meropenem. 56 Further, the Val120Leu mutation increases both kcat and KM for meropenem, indicating opposite effects on binding and hydrolysis. 57 The exact effect of the Val120Leu mutation on carbapenem hydrolysis on the molecular level is therefore complex and remains to be determined. ## Comparison of carbapenem deacylation in orientation I As presented above, orientation I of the 6α-hydroxyethyl moiety is calculated to give the overall lowest deacylation free energy barriers for both carbapenems. The combined FESs for the hydration state with lower free energy barriers are presented in Figure S7 for all three substrate orientations. In this section, we focus further on orientation I and the 'reactive' active site configuration in which only one water molecule is hydrogen bonded to Lys73:OQ1, and two to Lys73:OQ2 (unless otherwise stated). For this AC conformation, two different hydrogen bonding arrangements in the active site are possible: the DW can donate a hydrogen bond to the 6αhydroxyethyl hydroxyl group (named configuration 1), or the hydroxyl group can donate a hydrogen bond to the DW (configuration 2), see Figure 4. In MM MD, configuration (1) is sampled for 87% and 86% of simulation time for imipenem and meropenem, respectively. In addition to donating a hydrogen bond to the DW as in (2), the 6α-hydroxyethyl hydroxyl group can also donate a hydrogen bond directly to Lys73:OQ1 if the DW is displaced. This orientation of the carbapenem 6α-hydroxyethyl group may be the relevant one for β-lactone formation, which has been characterized as a side product for OXA-48-catalysed carbapenem turnover, particularly of 1βmethyl carbapenems (such as meropenem). 56,58 The β-lactone product has been proposed to form via intramolecular cyclisation, where the hydroxyl group acts as a nucleophile and donates a proton to Lys73. If the reaction occurs without a bridging water molecule, i.e. by a direct proton transfer between -OH and Lys73, lactonization is most likely lower in energy in orientation I than in III, based on the trends observed for deacylation energetics. For imipenem deacylation, both configurations (1) and (2) were observed in umbrella sampling. The lowest free energy barrier of 8.4 kcal/mol was calculated for configuration (1), and the barrier was increased by 2.0 kcal/mol for configuration (2). In addition to raising the free energy barriers, changing from (1) to (2) shifts the location of the transition state on the FES. For (1), the TS is located approximately at values -0.1 and 1.7 for the proton transfer and nucleophilic attack reaction coordinates, respectively (Figure 5, left). However, for (2), the TS location on the FES shifts to around -0.5 and 2.0 Figure S8). With active site configuration (2), the proton transfer has progressed further at the TS, whereas the approach of the DW oxygen to the acyl carbon is less advanced. This is potentially due to the additional hydrogen bond from the 6α-hydroxyethyl moiety hydroxyl decreasing the nucleophilicity of the DW, requiring the proton transfer reaction to have progressed further from the starting structure in the TS. Notably, a similar shift in the TS position on the FES is observed also in orientation III, where a water molecule is donating a hydrogen bond to the DW instead of the 6α-hydroxyethyl group (Figure S7). Mulliken charge analysis of the key QM atoms does not reveal many significant differences for the calculated charges along the reaction when comparing US calculations with either configuration (1) or (2) (Tables S5-S8). The main difference is observed at the TS, where for Lys73:OQ1 the charge is more positive and for DW:O the charge is more negative for configuration (2), as expected by the shift in the TS location towards the TI. For meropenem, the lowest calculated deacylation barrier is 11.2 kcal/mol with an average of 2.4 (± 0.1) and 1.4 (± 0.0) hydrogen bonds accepted by K73:OQ1 and OQ2, respectively. This barrier is 2.8 kcal/mol higher than the lowest calculated barrier for imipenem, or 2.2 kcal/mol including the free energy penalty (derived from MM MD for imipenem) for orientation I. In contrast to imipenem, the hydroxyl of the 6α-hydroxyethyl moiety in meropenem always rotates during unrestrained US sampling to hydrogen bond configuration (2), donating a hydrogen bond to the DW. This rotation occurs before the TS is reached even when configuration (1) is present in the starting structure. Enforcing the donation of a hydrogen bond from DW to the 6α-hydroxyethyl -OH, i.e. restraining the reaction simulations to configuration (1), affects the location of the TS in a similar manner to that observed with imipenem. TS locations for configurations (1) and (2) are at -0.2/1.8 and -0.5/2.0 (proton transfer/nucleophilic attack), respectively. However, changing the hydrogen bonding pattern between configurations has only a minimal effect on the energetics, as the barrier for (1) is 11.9 kcal/mol. Therefore, the decrease in activation energy for configuration (1) vs. (2) does not follow the same trend for meropenem as it does for imipenem. Possible reasons for this may include the presence of a 1β-methyl group in meropenem, as this may hinder the rotation of the 6α-hydroxyethyl group to better optimise further hydrogen bonds between active site residues and water molecules nearby. Such hindrance of 6α-hydroxyethyl rotation may also explain the preference observed for configuration 2 as the DW approaches the acyl carbon. A water molecule lodged between Tyr211 and Thr213 accepts a hydrogen bond from the carbapenem -OH moiety in configuration (1) or donates a hydrogen bond to it in configuration (2) (Figure 5 and Figure S8). The 1β-methyl group occupies the space above this water and may therefore induce its displacement or the re-organization of the surrounding water molecules to optimise hydrogen bonds between them, which could subsequently lead to a change from configuration (1) to (2). Additionally, the initial nucleophilic approach of the DW (from 3.5 to 2.2 ) with the 6αhydroxyethyl moiety in orientation I and hydrogen bond configuration (1) is calculated to be slightly lower in energy for imipenem (Figure S9). The DW remains hydrogen bonded to the hydroxyethyl oxygen during this approach, with the average distance to the hydroxyethyl methyl carbon reducing to about 3.3 . Notably, the initial approach between the DW and the carbapenem is also slightly higher in energy in orientations II and III than in orientation I, which may contribute to their overall energetics being less favorable for deacylation. However, the reasons for the preference for the imipenem, but not the meropenem, complex to adopt configuration (1) during deacylation are likely subtle and can result from small structural changes between the active site, substrate, and solvent molecules. ## Comparison with experimental data Most of the variants in the OXA-48 family are carbapenemases, with elevated imipenem hydrolysis rates when compared against other carbapenems. 59 For OXA-48, experimental measurements of kcat values for imipenem hydrolysis vary between 1.5 and 22.5 s -1 , which can be converted to free energy barriers for activation (Δ ‡ G) of 15.7 to 17.3 kcal/mol, using the Eyring equation. For meropenem, the measured kcat values range between 0.07 -0.16 s -1 , which converts to barriers of 18.7-19.2 kcal/mol. Using these figures as experimental estimates of free energies of activation, the difference (ΔΔ ‡ G) between imipenem and meropenem hydrolysis is between 1.4-3.5 kcal/mol, which is approximately the same magnitude as the strength of a single hydrogen bond (1- 3 kcal/mol). 60 Hence, structural factors contributing to more efficient breakdown of imipenem, compared to 1β-methyl carbapenems, are most likely to be subtle. Our QM/MM simulations suggest that orientation I of the 6α-hydroxyethyl group is the most likely AC orientation to undergo deacylation, when this exists in a state with decreased hydration around Lys73:OQ2 (i.e., with only one water molecule donating a hydrogen bond to this carboxylate oxygen). When comparing the lowest free energy barriers calculated in orientation I for imipenem and meropenem (Figure 3), the difference (ΔΔ ‡ G) for the two substrates is 2.8 kcal/mol; including the free energy penalty for the imipenem 6α-hydroxyethyl moiety adopting orientation I (0.6 kcal/mol, as determined from our MM MD simulations), the obtained ΔΔ ‡ G value drops to 2.2 kcal/mol. This is in excellent agreement with the experimentally determined range of ΔΔ ‡ G values. This strongly supports our assumption that TI formation is the rate-limiting process for carbapenem hydrolysis by OXA-48, consistent with similar findings for ceftazidime breakdown by OXA-48-like enzymes 35 and carbapenem breakdown by a range of class A serine β-lactamases. 41,42 The agreement further implies that the difference between imipenem and meropenem deacylation in OXA-48 may indeed be caused by the subtle difference in the preferred hydrogen bonding patterns involving the DW and the 6α-hydroxyethyl sidechain reported here. In turn, the presence of the meropenem 1β-methyl group apparently contributes to this difference by influencing both the orientation of the 6αhydroxyethyl group and the organization of water molecules in the near vicinity. (We further note that the underestimation of the absolute barriers can be fully accounted for by comparison of DFTB2 to higher level QM calculations, which indicates DFTB2 underestimates barriers by ~6.3-8 kcal/mol, see Table S4 & Figure S11. Thus, combined with the free energy penalty of 0.6 kcal/mol noted above, the corrected lowest barriers would be 15.3-17.0 and 17.5-19.2 kcal/mol for imipenem and meropenem, respectively.) Based on our MD simulations, the carbapenem tail groups are highly flexible and are thus unlikely to directly affect deacylation efficiency. Differences in kcat (reflecting the rate-limiting deacylation step) for carbapenems might therefore be explained similarly to our findings here, with differences largely caused by the presence or absence of the 1 β-methyl group. This is consistent with experimental data for OXA-48, which show higher kcat values for imipenem and panipenem vs. 1β-methyl containing carbapenems. 32,61 Overall, our analysis of the effects of active site conformations on carbapenem hydrolysis activity highlights the importance of controlling water access to the active site. On the one hand, it is crucial for the enzyme active site to support the binding of the deacylating water (through the aforementioned water channel). On the other hand, partial desolvation of the catalytic base (carboxylated Lys73) is required for efficient reaction. Such intricate control of active site solvation is a common feature of enzyme activity. For example, in ketosteroid isomerase, additional water molecules hydrogen bonding to the catalytic aspartate raise the barrier of reaction significantly. 62 Notably, this increased solvation occurs through water molecules hydrogen bonding to the carboxylate oxygen that is not receiving the proton, similar to what is observed here (difference between high and low barriers in Figure 3), but different from what we observed for ceftazidime hydrolysis. 36 Such additional hydrogen bonding will decrease the pKa of the catalytic carboxylate base, weakening its proton affinity and thereby leading to higher barriers for the reaction. To avoid or limit the occurrence of additional hydrogen bonding to catalytic bases, enzymes have evolved active site architectures that can promote desolvation to increase carboxylate reactivity. Such desolvation can for example be achieved by loop closure (as in triosephosphate isomerase and dihydrofolate reductase) 66,67 or closure of the substrate binding cleft (as in ketosteroid synthase). Here, subtle control of the solvation around the carboxylated Lys73 is related to nearby hydrophobic residues (Val120 and Leu158), which can adopt conformations that allow the presence of the deacylating water but avoid more extensive solvation of the catalytic carboxylate. ## Conclusions We have modelled carbapenem hydrolysis by the OXA-48 β-lactamase using QM/MM reaction simulations. The deacylation reaction was modelled for two carbapenem substrates, imipenem and meropenem, to deduce the origin of the higher activity towards imipenem compared to other carbapenems. MM MD simulations of the acylenzyme complexes demonstrate that the carbapenem tail (C2) groups are able to adopt many different conformations. In contrast, the carbapenem 6α-hydroxyethyl group is able to rotate and to adopt three specific different orientations, where it either interacts with the DW (I), Lys73 (III), or is rotated so that the methyl group is oriented towards Leu158 (II). Subsequently, deacylation was modelled using QM/MM for both substrates in these three orientations to investigate the effect of orientation upon deacylation efficiency. Our calculated free energy barriers indicate that the most deacylation-competent orientation is I, where the hydroxyl group interacts with the DW, and that the orientation III has the highest free energy barriers. Detailed comparison of the simulations revealed two factors that significantly affect the reaction energetics: hydration around Lys73, and the hydrogen bonding pattern between the DW and substrate, specifically the 6α-hydroxyethyl group. Hydration around the general base has been proposed to affect the predicted hydrolysis rates for other β-lactam substrates; 36 here, we show that this is affected by hydration around both Lys73 carboxylate oxygens (not only the oxygen participating in proton transfer). Increased hydration around the non-reactive oxygen (Lys73:OQ2) correlates with higher calculated barriers; in turn, the orientation of Val120 correlates with the number of water molecules near this oxygen. Another aspect influencing deacylation efficiency is the pattern of hydrogen bonds in the active site that involve the DW and the carbapenem 6αhydroxyethyl sidechain. Imipenem shows a preference for a configuration in which the DW donates hydrogen bonds to Lys73 and the 6α-hydroxyethyl hydroxyl group; the free energy barrier is higher when the hydroxyl group instead rotates to donate a hydrogen bond to the DW. This preference is not observed for meropenem: simulations with both hydrogen bond configurations have comparable energy barriers, which are similar to that calculated for imipenem in the less favorable orientation. Therefore, we can conclude that the difference between hydrolytic activities for the two carbapenem substrates stems from subtle differences in the active site hydrogen bonding patterns, which affect the reactivity of the DW. Furthermore, our results indicate that active site hydration is an important determinant of catalysis in OXA-48 enzymes: increasing hydration around the general base impairs carbapenem hydrolysis. Our study highlights the importance of detailed atomistic modelling in addition to experimental research to determine the exact origins of catalytic activity. Simulation protocols such as those employed here can extend information from crystallographic studies to enable investigation of the strength and dynamics of specific active site interactions during the catalytic cycle and directly investigate determinants of activity in situ.

chemsum

{"title": "Multiscale simulations identify origins of differential carbapenem hydrolysis by the OXA-48 \u03b2-lactamase", "journal": "ChemRxiv"}

synthesis_of_disulfide-rich_heterodimeric_peptides_through_an_auxiliary_n,_n-crosslink

## Abstract: Insulins, relaxins, and other insulin-like peptides present a longstanding synthetic challenge due to their unique cysteine-rich heterodimeric structure. While their three disulfide signature is conserved within the insulin superfamily, sequences of the constituent chains exhibit considerable diversity. As a result, methods which rely on sequence-specific strategies fail to provide universal access to these important molecules. Biomimetic methods utilizing native and chemical linkers to tether the A-chain N-terminus to the B-chain Cterminus, entail complicated installation, and require a unique proteolytic site, or a two-step chemical release. Here we present a strategy employing a linkage of the A-and B-chains Ntermini offering unrestricted access to these targets. The approach utilizes a symmetrical linker which is released in a single chemical step. The simplicity, efficiency, and scope of the method are demonstrated in the synthesis of insulin, relaxin, a 4-disulfide insulin analog, two penicillamine-substituted insulins, and a prandial insulin lispro. O ver the course of the last 50 years, the efficiency in linear peptide assembly has advanced through a series of innovations, most notably stepwise solid-phase synthesis and fragment ligation . Still, precise control of higher order structure through directed disulfide bond formation remains challenging . This is particularly the case with peptides such as insulin and relaxin given the additional complexity of their heterodimeric structures 9,10 . Biomimetic linkers have been introduced to address these issues. They share the native linear order of proinsulin where the N-terminus of the A-chain is indirectly connected to the C-terminus of the B-chain. Conversion to the two-chain form initially employed enzymatic conversion and was restricted by the requirement for a unique proteolytic site . The more recent reports employing chemically labile linkers represent a leap forward by eliminating the need for a proteolytic site, and the enzyme itself. The Kent group described an insulinspecific linkage of GluA4-ThrB30, which was saponified following oxidative folding 16 . Subsequently, a sequence-agnostic approach to synthesis in the insulin-like peptide family was reported, which employed a reversible tethering of the A-chain N-terminus through a labile amide to the B-chain C-terminal ester 17,18 . In each instance, the strategies mimicked the linear order of the A-and B-chains found in proinsulin. While native, this linear orientation increases synthetic complexity in requiring either non-standard chemistry for linker installation, or a two-step linker excision. A straightforward, general synthesis would ideally align with conventional solid-phase methods and employ a single chemical step for removal. We report the synthesis of insulin and a set of related peptides by a synthetic protocol that employs a reversible crosslink of the two N-termini through parallel extension of the respective A-and B-chains made by conventional Solid Phase Peptide Synthesis (SPPS). The N-N heterodimers of these insulin-related peptides efficiently fold under standard redox conditions, and subsequently convert to the native hormone under mildly alkaline conditions by virtue of two simultaneous diketopiperazine cyclizations. The efficiency and versatility of the method are demonstrated in the synthetic yield of insulin and the direct translation to the synthesis of relaxin. The non-native N-N linkage enables the synthesis of two site-specific penicillamine-substituted analogs 4 and 5, that fail when using a high-efficiency N-C folding intermediate, named des-DI insulin 15 . This synthetic approach is based upon an orthogonal, non-native N-N linkage of individual peptide chains that is synthetically straightforward and of high efficiency in synthesis of insulin-like peptides. The approach holds promise for translation within the broader class of disulfide-rich heterodimeric peptides. ## Results Insulin synthesis. We explored the chemical synthesis of insulin, relaxin-2, and four insulin analogs (Fig. 1) through reversible crosslink of the two N-termini by parallel extension of insulin A-and B-chains made by conventional SPPS. A Lys-(iBu)Gly dipeptide extension at the N-termini was envisioned to provide a side-chain anchor for an OEG (polyethylene glycol)oxime-based crosslink (Fig. 2). Whether a suitably, sequence-extended N-N heterodimer of insulin A and B-chains might fold under standard redox conditions was a central uncertainty to be investigated. The synthesis of the insulin A-chain began with the coupling of Fmoc-Asp-OtBu to Chemmatrix Rink amide resin to introduce the C-terminal Asn, and the remaining residues added by conventional automated Fmoc-based SPPS protocol (Fig. 3). The isoacyl Thr-Ser dipeptide at A8-A9 was incorporated as a means to enhance peptide assembly, solubility, and handling 19 . The Boc-Lys(iBu)Gly-OH dipeptide was installed through sequential α-bromoacetylation and isobutylamine treatment 20 of resin-bound A-chain 7 followed by 3-(diethoxyphosphoryloxy)-1, 2, 3-benzotriazin-4(3H)-one (DEPBT)-mediated coupling of Boc-Lys (Fmoc)-OH. Successive side-chain coupling of PEG 8 and bis (Boc)amino-oxyacetic acid provided resin-bound A-chain 8, which was deprotected and cleaved from the resin under standard conditions. The crude peptide, (Supplementary Figure 13) was purified by C8 reverse phase HPLC (RP-HPLC) to provide Achain 9 in 25% yield (Fig. 4 and Supplementary Figure 14). The B-chain synthesis (Fig. 3) was initiated with loading of Fmoc-Thr(tBu)-OH on a ChemMatrix HMPB resin under Mitsunobu conditions to minimize racemization. The remaining residues (B1-B29), including the isoacyl Tyr-Thr dipeptide at B26-27 were incorporated by conventional Fmoc-based SPPS protocol. The Boc-Lys(iBu)Gly-OH dipeptide, PEG 8 (polyethylene glycol), and bis(Boc)amino-oxyacetic acid were sequentially coupled to resin 10, followed by cleavage and deprotection to afford 11. The free aminooxy-derivatized B-chain 11 was treated with Terephthalaldehyde (10 equiv.) in 0.1% TFA/70% aqueous acetonitrile (ACN) to provide the crude imino-benzaldehyde Bchain derivative 12 (Supplementary Figure 15), which was recovered in 15% total synthetic yield following RP-HPLC purification (Fig. 4 and Supplementary Figure 16). The oxime ligation of A-chain and B-chain was accomplished by combination of 9 and 12 in a 0.1% TFA containing 50% aqueous ACN, for 2 h (Figs. 3, 4 & Supplementary Figure 17). The PEG 8 was experimentally determined to be the minimal distance required between A-and B-chains to subsequently produce the properly disulfide-paired hormone. The folding of the ligated A-B intermediate 13 was performed at pH 9 in an aqueous buffer with 2 mM cysteine and 0.5 mM cystine at 4 °C, to produce a single major product 14 (Figs. 3, 4 and Supplementary Figure 18). This properly folded, single-chain insulin was obtained in a 45% combined yield for ligation, disulfide-formation, and RP-HPLC purification (Supplementary Figure 19). Single-chain insulin 14 was efficiently converted to two-chain insulin 1 using 0.5 M phosphate buffer (pH 7.0) at 56 °C (Fig. 3). The two simultaneous diketopiperzine (DKP) cleavage reactions were complete after 5 h to provide insulin 1 in 65% yield, following HPLC purification (Fig. 4, Supplementary Figures 20-21). The speed of DKP formation can be further accelerated by selection of dipeptides that favor cis-configuration, which can be achieved by alkylation at the alpha carbon of the first amino acid and more judicious N-alkylation at the second. When compared to our previous report employing an N-C insulin order, the yield was enhanced in a relative sense by 20% 17 . This improvement predominantly results from eliminating the more alkaline pH needed to cleave the ester bond. Overall, the synthetic yield of insulin was 30%, starting from purified A-chain. 1-2) began with A-and B-chains, respectively, utilizing Fmoc-Cys(Trt)-OH/ NovoSyn® TGA resin and Fmoc-Ser(tBu)-OH/ ChemMatrix HMPB esterified resin. The remaining amino acids were added by a conventional Fmoc protocol, with isoacyl dipeptides employed as Asp-Ser at B1-B2 and Ser-Thr at B26-B27. In addition, the N-terminal residue of the A-chain was introduced as Gln, which was subsequently cyclized to pGlu. The Boc-Lys(iBu)Gly-OH dipeptide, PEG 8 and bis-Boc-aminooxyacetic acid were introduced as reported in the insulin synthesis (Supplementary Figures 22-25). The oxime ligation 21 and peptide folding 22 (Supplementary Figure 3) were also conducted as previously communicated with a combined yield of 46% (Supplementary Figures 26-28). The DKP cyclization and the subsequent pGlu formation were completed in 7 h using 0.5 M phosphate buffer (pH 7.0, 56 °C) in a combined 65% yield (Supplementary Figures 29-30), and the overall synthetic yield of human relaxin-2, 2 was 30%, starting from A-chain. To minimize intermediate handling loss, we assessed the ligation, folding, and linker excision steps starting with pure A-and B-chains and chromatographically purifying only at the end (Fig. 5 and Supplementary Figure 4). This simplified protocol improved the total synthetic yield from 30 to 38%, and represents one of the most efficient chemical syntheses reported yet for human relaxin. The bioactivity of the synthetic relaxin-2 proved indistinguishable from an external native control hormone (Supplementary Figure 12). Synthesis of a four-disulfide insulin analog. To further explore the potential of the new methodology, we applied it to an insulin analog with an additional, fourth disulfide linking CysA10 and CysB4 (Supplementary Figure 5). This analog as prepared by biosynthesis is reported to possess reduced propensity to fibrillation, and full in vivo activity 21 . The first chemical synthesis of these four-disulfides (4-DS) insulin analog was achieved through sequential disulfide bond formation that included an iodine oxidation step. An iodine-free synthesis of this challenging target suggests that the methodology may prove useful in the synthesis of other peptides with multiple disulfides, especially those with methionine and tryptophan. ## Synthesis of relaxin. The synthesis of relaxin (Supplementary Figures The insulin-extended A-chain S1 and B-chain S2 synthesis incorporated Fmoc-Cys(Trt)-OH at A10 and B4 but otherwise were identical to the previously presented insulin protocol, and they were, respectively, achieved in yields of 24% and 15% (Supplementary Figures 35 and 36). The ligated linear precursor S3 was folded without modification of the insulin protocol, and the single-chain, 4-DS analog S4 was obtained in 40% yield (Supplementary Figure 34 and 37). The excision step was achieved in 5 h to yield the pure 4-DS insulin analog 3 in 64% yield (Supplementary Figure 38). This peptide as assessed by LC-MS and in vitro potency was indistinguishable from the same insulin analog as prepared by orthogonal disulfide bond formation 22 , and only slightly less potent than native insulin (Supplementary Figure 10). The single-chain form of the 4-DS analog S4 was sizably less potent than the two-chain form, demonstrating the deleterious impact of an N-terminal constraint on bioactivity, but not on the ability to form native disulfides with a linker of appropriate length. Insulin with a comparable crosslink at the N-termini of A-and B-chains was suppressed Four disulfide human insulin, 3 L B19(Pen)-insulin analog, 5 in bio-potency to nearly the same extent as observed in the 4-DS analog 14 (Supplementary Figure 10). Synthesis of penicillamine-containing insulin analogs. The synthesis of the A7(Pen) A-chain S5 and B19(Pen) B-chain S8 (Supplementary Figure 6 and 7) employed the same protocol as employed for insulin, except for Fmoc-Pen(Trt)-OH at A7 or B19. The chain assembly yields following purification were 22% for the A-chain analog S5, and 14% for B-chain S8 (Supplementary Figures 40 and 45). The oxime ligation of S5 to 12 and S8 to 9 was conducted as previously achieved for native insulin, and the ligated purified synthetic intermediates S6 and S9 were obtained in respective yields of 55% and 50% (Supplementary Figures 41 and 46). The subsequent folding of S6 and S9 was without protocol modification as reported for native sequence and was complete with comparable efficiency in 12 h to provide S7 at 20% yield, and S10 at 19% (Supplementary Figures 39, 44, 42, and 47). The cleavage of the DKP-peg-bis linker was achieved in 9 h (pH 7.0, 56 °C), to provide analogs 4 and 5 in total yields of 30% and 28% (Supplementary Figures 43 and 48). The native disulfide pattern was implied by single LC-peaks in the Glu-C peptide mapping (Supplementary Figure 10, Supplementary Table 1), which was definitively confirmed for the Pen-A7 analog by comparison to disulfide isomers prepared by orthogonal synthesis (Supplementary Figures 60 and 61). The in vitro bioactivity of these novel insulin analogs was assessed and observed to be reduced to varying degrees relative to native hormone (Table 1, Supplementary Figure 11). The other four single-site, penicillamine insulin analogs (A6, A11, A20, and B7) were chemically synthesized using a linear desDI single-chain precursor without issue, (Supplementary Figures 55-59) 15 . The A7 and B19 proved to be synthetically accessible only by the Ntermini ligation approach we describe in this manuscript. The bioactivity of the penicillamine analog at A20 was least affected in a relative sense, especially when compared to the analogs at the other inter-chain cysteines (A7, B7, and B19). Interestingly, the placement of the gem-dimethyl substituent at A11 was approximately 100-fold more disabling than at A6, the other partnering residue in the single intra-chain disulfide. Synthesis of Lys-Pro insulin. Lys-Pro insulin represents the first hormone analog produced by rDNA-technology approved for human use 23 . The inversion of the natural dipeptide to Lys-Pro eliminates trypsin-like proteolysis. Consequently, this analog should be equally accessible by an enzyme-based approach as a synthesis that is DKP mediated. To prove this point and assess the relative efficiency in the removal of the auxiliary N,N-crosslink, insulin 6 was synthesized (Supplementary Figure 8) as described for native sequence, but with replacement of the DKPsusceptible dipeptide with a Gly-Lys dipeptide. Peptide chain synthesis, oxime ligation and disulfide formation in insulin 6 were achieved as with native hormone in 46% yield (Supplementary Figures 49-52). The single-chain S14 was converted to the twochain form 6 by digestion (Supplementary Figure 54), in Tris buffer pH 8 for 1 h. The Lys-Pro insulin was obtained after purification by RP-HPLC in 66% yield (Supplementary Figure 53). The yield of 6 as produced by enzyme was 30% from purified A-chain, which is identical to the yield of 1 obtained by DKP-mediated chemical cleavage. ## Discussion We report a general synthetic route to insulin-related peptides with likely application to the broader family of disulfide rich, twochain peptides. This straightforward method demonstrates the use of a non-native N-N linkage that is compatible with automated SPPS. The use of identical N-terminal A-and B-chain extensions and conventional ligation streamlines the assembly of the heterodimer, followed by single-step excision of the auxiliary tether. Insulin and relaxin, which have historically constituted Human insulin, 1 Single chain insulin, 14 difficult synthetic targets, were produced by this procedure within a few days, in high yield. Notably, an initial attempt to synthesize insulin through an N-N linkage without an N-terminal extension was reported to be unsuccessful 16 . In our experience, the folding efficiency was dramatically enhanced by incorporation of the OEG-based N-terminal extensions. The central, enabling element of this approach is the reversible N-terminal crosslinking of the A-and B-chains to enable intramolecular native disulfide bond formation. The efficiency is highlighted in the synthesis of relaxin from A-and B-chains employing only a final chromatographic purification step in a 38% yield (Table 2). The use of OEG-extended linkers was found to improve handling of the individual peptide chains, the ligated intermediate, and to enhance the subsequent formation of native disulfides. These conditions were applicable to the native hormones and translated to a synthetic target that had previously required orthogonal stepwise synthesis, a four-disulfide containing insulin analog 22 . The successful syntheses of two individual penicillamine substituted insulin analogs, that we could not prepare by native folding using a bio-mimetically linked insulin precursor 15 , demonstrate a unique virtue to this synthetic approach. The analogs complete an otherwise full set of selective penicillamine substitutions for each of the native cysteines (Table 1). There was no direct relationship in the difficulty of synthesis relative to bioactivity, as the B19 analog was of intermediate potency to the full set while A7 was least potent. We envision the application of this approach beyond the insulin/relaxin super family. The methodology is compatible with peptides produced by any method where the linker can be semisynthetically conjugated to a selective amine, preferably the Nterminus 24 . The linker can be further optimized to enhance the biophysical properties of synthetic intermediates. The synthetic approach is not limited to oxime linkage and could conceivably utilize other linkage chemistries. A sagacious aspect of the reported syntheses is the use of DKP formation, an adverse reaction in peptide synthesis 25 as controlling element in the removal of the auxiliary crosslink 17,18 . Further refinements in the propensity to cyclize will broaden the ability to accelerate or delay reversal of the crosslink. As exemplified in the synthesis of Lys-Pro insulin, the synthetic strategy is compatible with selective proteolysis. Notably, the synthetic yields in use of the chemical and enzymatic cleavage were comparable, attesting to the productivity of the former. nd not determined a Purified A-and B-chains b Purified A-B dimers following ligation and starting with purified A-and B-chains c Purified two-chain peptides following DKP-mediated cleavage of the purified folded single-chain A-B dimer d Total yield of purified peptides following ligation, folding, and DKP-mediated cleavage, starting with purified A-and B-chains e Purified single-chain disulfide-bonded A-B dimers following ligation and folding, starting with purified A-and B-chains f Purified two-chain peptide following ligation, folding, and DKP-mediated cleavage, starting with purified A-and B-chains g Purified peptides following folding, starting with purified ligated A-B dimer

chemsum

{"title": "Synthesis of disulfide-rich heterodimeric peptides through an auxiliary N, N-crosslink", "journal": "Nature Communications Chemistry"}

random_forest_refinement_of_the_kecsa2_knowledge-based_scoring_function_for_protein_decoy_detection

## Abstract: Knowledge-based potentials generally perform better than physics-based scoring functions in detecting the native structure from a collection of decoy protein structures.Through the use of a reference state, the pure interactions between atom/residue pairs can be obtained through the removal of contributions from ideal-gas state potentials. However, it is a challenge for conventional knowledge-based potentials to assign different importance factors to different atom/residue pairs. In this work, via the use of the 'comparison' concept, Random Forest (RF) models were successfully generated using unbalanced data sets that assign different importance factors to atom pair potentials to enhance their ability to identify native proteins from decoy proteins. Individual and combined data sets consisting of twelve decoy sets were used to test the performance of the RF models. We find that RF models increase the recognition of native structures without affecting their ability to identify the best decoy structures. We also created models using scrambled atom types, which create physically unrealistic probability functions, in order to test the ability of the RF algorithm to create useful models based on inputted scrambled probability functions. From this test we find that we are unable to create models that are of similar quality relative to the unscrambled probability functions. Next we created uniform probability functions where the peak positions as the same as the original, but each interaction has the same peak height. Using these uniform potentials we were able to recover models as good as the ones using the full potentials suggesting all that is important in these models are the experimental peak positions. ## Introduction According to the "thermodynamic hypothesis". 1 the native protein in its preferred milieu should adopt a structure that has the lowest Gibbs free energy. Hence, accurate energy functions are needed to solve the protein folding, the protein structure prediction, and protein design problems. Extant scoring functions can be classified into three broad categories, physics-based, knowledge-based, and machine learning based scoring functions. Physicsbased scoring functions typically employ a classical force field that represents the protein at the atomic level through the use of energy terms that represent bond, angle, torsion, van der Waals, and electrostatics interactions using relatively simple terms. Alternatively, knowledge-based scoring functions extract radial distributions of atom/residue pairs from a protein structure database and use different statistical analysis to gain "pure" interactions between atom/residue pairs. 2, Machine learning based scoring functions utilize different machine learning/deep learning algorithms and a large variety of information from protein structures. For example, SVMQA combines different statistical potentials, secondary structure information, and surface area as the descriptors for protein structure prediction. 53 Currently, knowledge-based scoring functions have been more successful than physics-based potentials in protein structure prediction. 2 Knowledge-based scoring function can be classified into coarse-grained residue 17, or atomic level 2, potentials due to different descriptions of the interactions present. The main challenge of building up a knowledge-based potential is how to set up an appropriate reference state. In 1985, Miyazawa and Jernigan introduced the 'Random-mixing approximation', 26 which states that, in the reference state, amino acid and solvent molecules would be uniformly distributed throughout the volume due to the absence of interactions. Many different kinds of reference states have been constructed since then. For instance, the freely-jointed chain (FJQ) model 41,42 was applied to construct a random-walk reference state. 39 Knowledge-based potentials are very sensitive to the completeness of the structural database used to describe the potential and given the current status of available structural information the contribution or presence of each atom/residue pair combination is not equal for every interaction. Thus, different importance factors need to be applied to each pair wise interactions in a knowledge-based potential to reflect this deficiency. In principle, the reference state removes the contribution from an ideal-gas state potential, but the process of accurately assigning different importance factors to each atom/residue pair remains challenging. In order to obtain an accurate potential, and estimate of the relative importance of each atom pair is needed. In this work, we focus on using the Random forest (RF) machine learning (ML) model to refine a knowledge-based potential by designating different importance factors to each atom pair in a given potential. Here, the knowledge-based and empirical combined potential, KECSA2, with 167 atom types, 2001 torsion potentials, and 14028 non-bonded atom pairs is used in the RF refinement. Our goal is to construct a ML model that can accurately differentiate the native structure from decoys using a potential refined by RF optimization. Firstly, a 'comparison' concept is used to change an unbalanced database to a balanced one for the construction of the RF refinement models. Secondly, the RF refinement models are applied both on individual and combined decoy sets from twelve commonly used decoy sets. As a result of this optimization all of the RF refined models recognize native structure more accurately than conventional potentials. The performance improvement realized by our RF refinement protocol can be applied to the refinement of potentials other than KECSA2. Finally, the importance of KECSA2 was examined through the use of scrambled and uniform probability functions. The comparison result suggests that only the peak positions are important in the construction of RF models, and the RF models can be used to optimize the peak heights for different atom pairs. ## Building up the probability list for the proteins If all independent pair wise probabilities with different magnitudes in a n-body system are known, the probability of the whole n-particle system can be obtained as: where pn is the probability of the n-particle system, cij is the scaling factor, which can be evaluated using the random forest model, of pair wise probability pij, i and j represent two different particles. Using a knowledge-based potential with pair wise independent interactions, the independent pair wise probabilities for the bond, angle, torsion, and nonbonding terms can be obtained. If the protein structure is treated as a n-particle system, the probability is: pprotein is the protein structure probability, c and p represent the scaling factor and the probability of atom pair  and , the subscripts ij, kl, mn, and pq correspond to bond, angle, torsion, and non-bonded atom pairs, respectively. In this work, we make two further assumptions: (i) ∏ 𝑝 𝑏𝑜𝑛𝑑 𝑏𝑜𝑛𝑑 and ∏ 𝑝 𝑎𝑛𝑔𝑙𝑒 𝑎𝑛𝑔𝑙𝑒 are similar for native and all decoys and, hence, are treated as a constant C; (ii) the probabilities for the torsion and nonbond atom pairs are independent since a reference state is used to remove contributions from the ideal-gas state. With these assumptions, the probability of a n-atom protein can be written as: Taking the logarithm on both sides of equation 3 we get: where xmn and xpq are the logarithm of cmn and cpq, respectively. A detailed potential database, KECSA2, was utilized to obtain pmn and ppq. Below we use O-MET-CG-MET as an example for what is involved in calculating the pair wise probability of a given protein. From KECSA2, the probability versus distance function, shown as a red curve in Figure 1, can be found. If the distance between O-MET-CG-MET in the protein is 4.5 , we first obtain the corresponding probabilities for the distances from 4 to 5 with an interval of 0.005 . Next, we take the logarithm of the average of the 201 probabilities obtained in the previous step, and use it to represent the probability at distance 4.5 . Using equation 5, the probability for each atom pair present in the protein can be obtained; for the same atom pairs, the probabilities were summed yielding the final probability. In this way, the probability list for each protein examined can be generated. ). The descriptor sets are defined as the 'Descriptor vector' in Figure 2. Next, the descriptor vector of the native minus the vector of each decoy are classified as class '1', which means 'more stable than' since the native structure is always more stable than the decoys; the descriptor vector of each decoy minus the vector of the native is defined as class '0', which represents 'less stable than'. The resultant descriptors are described as the 'final descriptor vector' in Figure 2. In this way, equal members of class '0' and class '1' can be generated, which is an ideal situation for classification. Hence, a RF model can be obtained based on using those two classes. Through the use of this classification system a RF model can be generated where the relative probabilities of two proteins with the same sequence can be compared. A final descriptor vector can be generated using the descriptor vector of the first protein minus the second's. Then the RF model can be used to predict the class for that final descriptor vector. If the prediction from the RF model is '1', it means the first protein is 'more stable than' the second one, and if the prediction is '0', the first protein is 'less stable than' the second one. ## Building up the ranking list for a decoy set Constructing a RF model that can accurately differentiate native and decoy structures is not enough. For a native recognition blind test, in order to identify the native structure, a ranking of all structures should be generated. Thus, the RF model needs to be used to obtain the ranking list for a decoy set. Figure 3 gives the protocol used to obtain the ranking of a decoy set with n structures. First, the probability descriptor of each protein structure can be built using the KECSA2 database. Second, a table for each structure was obtained from the probability descriptor of the individual protein structure minus the probability vectors of all the other structures. Then, the RF model is used to predict the class of each column in all tables; in other words, the RF model is used to 'compare' two structures. Finally, a row with length n-1 can be generated for each structure. The value of each column in the resultant row is either '0' or '1', which represents the comparison result of each structure with all other structures. The sum of the resultant row is defined as a 'score', which indicates if the corresponding structure is more stable than the "score" amount of decoys. In this way, the score of each structure can be generated, thereby, creating a ranking list. ## Decoy sets The decoy sets we used include the multiple decoy sets from the Decoys 'R' Us collection (http://compbio.buffalo.edu/dd/download.shtml), which include the 4state_reduced, fisa, fisa_casp3, hg_structal, ig_structal, ig_structal_hires, lattice_ssfit, lmds, and lmds_v2 decoy sets. The MOULDER decoy set was downloaded from https://salilab.org/decoys/; the I-TASSER decoy set-II was obtained from https://zhanglab.ccmb.med.umich.edu/decoys/decoy2.html; and the ROSETTA all-atom decoy set from https://zenodo.org/record/48780#.WvtCA63MzLF. Our RF model was compared to the following potentials designed for decoy detection: KECSA2, GOAP, 2 DFIRE, 40 dDFIRE, 37,43 and RWplus. 39 The programs for these methods were downloaded from the corresponding author's website. ## Protein structure preparation All protein structures (including both native and decoys) were converted into their biological oligomerization state and prepared with the Protein Preparation Wizard, 44 which adds missing atoms, optimizes the H-bond network, and performs energy minimization to clean up the structures for subsequent calculations. The decoy sets can be found here https://github.com/JunPei000/protein_folding-decoy-set. ## The KECSA2 potential KECSA is a potential specifically designed for protein and protein/ligand systems that was developed in our group. 45 A detailed atom type definition was used in this potential; in other words, every atom type represents a specific atom in the twenty naturally occurring amino acids. For instance, 'CA_ALA' corresponds to the alpha carbon in alanine. A detailed description of out atom typing scheme can be found here https://github.com/JunPei000/protein_folding-decoy-set. ## Machine learning and validation The sklearn.ensemble.RandomForestClassifier function from Scikit-learn was used to create the proposed classification model. 46 For each decoy set, ten iterations described above were run creating ten sets of hyperparameters. Hyperparameters with the highest frequency are shown in Table s1. ## (ii) Overall decoy sets training In order to avoid duplicate protein system that existed both in the training and test set, different decoy sets were combined based on different proteins. For instance, 2cro is a decoy set that is present in both the 4state_reduced and fisa decoy sets; herein, the two 2cro sets were combined together. In total, there were 235 different decoy sets. In overall decoy set training, the top 100 to top 500 most important features were used instead of using all 16029 descriptor elements, based on an importance feature analysis generated from the combined decoy sets. In order to obtain the overall importance feature analysis, ten importance feature analyses on the decoy subsets were generated. Specifically, for each importance feature analysis on a decoy subset, a grid search was done on 20% of the data from the combined data set, then an importance feature analysis of the best parameter set was generated. This process was repeated ten times in order to cover the whole data set, and the ten importance feature analyses were combined to obtain the overall result. Based on this analysis, the top 100, 200, 300, 400, and 500 most important features were selected for use in the RF model. Again ten iterations were run to find the best hyperparameter set for the combined data set. Where AB represents the torsion atom pair, rAB is the distance between atom A and B, rmax_AB is the corresponding distance of the lowest potential/highest probability point for atom pair AB. For nonbond interactions, equation 8 was used to generate the uniform probability function of different atom pairs based only on the rmax values from the KECSA2 nonbond interactions. Where AB is a nonbond atom pair, rAB represents the distance between atom A and B, E1 and E2 are parameters for repulsion and attraction interactions, respectively. It is known that the probability function achieves a maximum at rmax_AB; hence, the derivative of equation 8 should be equal to 0. By setting the maximum value to a constant, the parameters E1 and E2 can be obtained using equations 9 and 10. In this way, if the constant in equation 10 is the same for all nonbonded interactions, probability functions with the same heights but different peak positions can be constructed. In the end, descriptor vectors and final descriptor vectors can be generated based on these uniform probability functions. ## Model accuracy (individual decoy set training) The most important characteristic of the resultant scoring function is its ability to differentiate the native structure from decoys. Table 1 shows the accuracy for both the RF models and traditional scoring functions. Since ten cycles of independent training and testing were performed for each decoy set, the highest, lowest, and averaged accuracy were used to represent the general performance of RF models on that specific decoy set. In this way, the performance of the RF model can be better interpreted. In general, the RF model shows higher accuracies than all traditional methods for all of the decoy sets. For some decoy sets, like fisa, ig_structal, lmds_v2, and rosetta, RF models significantly improved the averaged accuracy to nearly 1.000, and the lowest accuracy values are still higher than the best accuracies of the other scoring functions. For the other decoy sets, such as 4state_reudeced, fisa_casp3, hg_structal, ig_structal_hires, I-TASSER, lattice_ssfit, lmds, and MOULDER, the averaged accuracies from the RF models are similar to the best accuracies of traditional scoring functions, while the lowest accuracies of the RF models are similar to the accuracies of other methods. Overall, the RF models show better performance. ## Ranking of the native structure Although the accuracies of RF models are higher than other methods, we still wanted to further validate their performance. Other than accuracy, another important criteria for judging a model/scoring function is whether the model/scoring function identifies the native structure as having the lowest rank. Hence, native structure ranking from the different methods were also compared. Table 2 shows the rankings of the native structures from several different models. The highest, lowest, and averaged rankings are shown to assess the performance of RF models. For the decoy sets fisa, ig_structal, lmds, lmds_v2, and ROSETTA, the RF models substantially improves native structure ranking over the other models. In the remain decoy sets, the averaged rankings of native structures are similar to the best performance of the other scoring functions. It can be concluded that, in general, the RF model shows a better performance in ranking the native structure over other methods we tested. ## RMSD and TMscore of the first selected decoy structure Although the ability to recognize the native structure as the most stable structure is a crucial characteristic of a good model/potential. For a model/potential to be useful for guiding conformation sampling, it should have a good correlation with structural quality. The RMSD and TM-score were used as two criteria for assessing the quality of each decoy structure. RMSD is the root mean squared deviation of all C pairs of the decoy to the native structure. TM-score 47 gives a large distance a small weight and makes the magnitude of TM-score more sensitive to the topology. Table 3 and Table 4 summarize the results of best model selection of different methods. a) RF models were trained on different decoy sets. Table 3 shows the 1 st decoy's RMSD of RF models and against a range of available scoring functions. The RMSD values of available methods are generally within the range of lowest and highest RMSD values of the RF models for each decoy set. This means the performance of those traditional scoring functions are within the confident range of our RF models. Table 4 shows the 1 st decoy's TM-score for the RF model and against several models; these results are similar to what we observed for the RMSD analysis. In each decoy set, the 1 st decoy's TM-score is within the range of the lowest and the highest TM-score from the RF models. Considering that the independent training and testing process was done ten times, the range of lowest to highest RMSD/TM-score values show the confidence range of RF models for each decoy set. In general, the RMSD/TM-score performance of available models are within the confidence range of the RF models, and the averaged values are similar in RF models and models we tested against. In other words, the performance of the RF models against a range of models, when it comes to selecting the best decoy structure, were similar. ## Feature importance analysis for the overall decoy set It is important to understand whether the better performances of RF models are due to overfitting because a large number of descriptors were used. To this end, a feature importance analysis was performed and is shown in Figure s1. Based on the analysis, new RF models using only the top 500 features were constructed using the previous procedure. Table 5 shows the comparison of the accuracies between using only the top 500 and all 16029 features. In general, the highest, lowest, and averaged accuracy values of the RF models using the top 500 features for each decoy set are similar to the corresponding values of the RF models using all features. Table s2 shows the comparison of the native structure's ranking between RF models with the top 500 and all features. The native rankings of RF models with 500 features are similar to those rankings for the RF models using the full feature set except for decoy set fisa_casp3. For fisa_casp3, the native ranking increases from 1.00 to 36.40 due to the loss of the remaining 15529 features. Among the 15529 features, they might contain important information to differentiate the native from decoys in that particular decoy set. But in general, the performance of RF models with 500 features for native structure ranking is still similar to the RF models with all features. Table s3 and Table s4 show the results of best decoy selection for RF models with the top 500 and whole feature set. Similar to comparisons of accuracies and native rankings, for the best decoy selection, the performance of the RF models using the top 500 features is still similar to the RF models with all features. Hence, we conclude that the better performance of RF models with all features is not simply due to overfitting and the current method is robust even in the face of potentially non-essential features. ## Comparison of overall RF model with traditional scoring functions Besides creating RF models for each individual decoy set, combined RF models using all decoy sets were also constructed. This examines the situation where in a study one might generate decoys using one method and then score them with another. There were 291 individual systems across the 12 decoy sets that were combined finally, yielding 235 different protein systems (several proteins overlapped amongst the decoy sets). In these studies, 80% of the combined data set was used as the training data to build the RF models instead of choosing several specific decoy sets (like 4state_redueced, fisa, etc.). This was done to insure that the training and testing data set covered the same feature space and had the same distributionthis is known as an independent and identical distribution (IID). 48 The feature space and distribution of decoys from different decoy sets are different because different models were used to generate those structures. Table 6 shows the result of comparing the overall performance of RF models with a number of available potentials. Due to the large number of descriptors, it is impossible to obtain RF models using the entire 16029 feature set. Based on the importance analysis discussed previously, instead of using all features, top 100, 200, 300, 400, and 500 features were used to build up the overall RF models on the combined decoy sets. First, all RF models with different importance features provide higher averaged accuracy values than other traditional scoring functions. Clearly, the accuracies of the RF models outperform the other conventional methods. Second, the highest rankings of the native structure from RF models are smaller than the rankings of other methods, and all of the averaged rankings of the RF models were ~10 or less, which means the RF models can identify the native structure within the top ten structures. Hence, the RF models outperform other methods on this task. ## Importance of potential Based on the previous discussion, it is clear that the RF models with KECSA2 perform the best in accuracy and ranking both on individual and overall decoy sets. This which means those two atoms are most stable when they form a hydrogen bond at that distance. However, after scrambling, the peak position might change to 4.51 (peak position of CA-GLY and C-THR), which no longer represents a hydrogen bond. Thus, the scrambled probability function suggests that the atom pair O-PRO and N-ALA is most stable when they do not form a hydrogen bond. It is clear that the scrambled probability functions are unphysical. We expect that it would be unlikely that the RF model, as employed herein, could correct these deficiencies so, the performance of the scrambled probability function is expected to be worse than original KECSA2. Second, the uniform probability functions (or potential functions) were built for the top 100 and 500 atom pairs to test if the KECSA2 probability peak heights (or well-depths) are important in RF models. The uniform probability functions have the same peak positions found in KECSA2, but with same heights. By doing so, the interaction strength 'bias' of different atom types from KECSA2 can be eliminated via use of uniform probability functions. If the KECSA2 probability peak heights (or interaction potentials) are significant, the performance of uniform potential should be worse than KECSA2. The comparison of the result between the original KECSA2 potential, scrambled potential, and uniform potential are shown in Table 7. From the comparison between the original KECSA2 and the scrambled potentials, we find the accuracy of the models decreased ~0.15, which gives a clear signal that the full KECSA2 potential (well depth and energy minimum) plays a role in RF models. The comparison between the uniform and original KECSA2 potential gives an evidence of how important the rmax component of the KECSA2 potential is in building an effective model. For the RF models with the top 100 features, the averaged, highest, and lowest accuracies based on the original KECSA2 potential are slightly higher than the corresponding accuracies from the RF models based on uniform potentials. However, if the number of features is increased to 500, the averaged, highest, and lowest accuracies from the RF models based on the original KECSA2 are similar to the uniform potentials. This provides strong evidence that only peak positions in the probability functions are critical in building up RF models for native protein structure detection. More importantly, the result also implies that RF models can be used to tune the height of peaks in probability functions (or the depth of potential functions) only with the information of peak positions in protein structures. ## Conclusion In this work, we utilized a 'comparison' concept to construct RF models on an unbalanced data set. With these RF models, the knowledge-based potential, KECSA2, was refined via assignment of different importance factors to different atom pairs present in the scoring function. The performance of the resultant RF models were assessed with individual and combined decoy sets and compared with the results from conventional models. We find that the RF models perform better in accuracy and native ranking and have similar performance in the RMSD and TM-score tests. In other words, the RF models improved the effectiveness of finding native structures from a set of decoys, without compromising their ability to find the best decoy structures. This RF model based refinement not only can be used to improve the performance of KECSA2, but it can also be applied to other atom/residue pair based potentials. More importantly, we find that only peak positions in probability functions play a significant role in constructing the RF models. This result implies that, with peak position information, RF models can be created to construct probability functions (or potential functions) by tuning the height of peaks in those functions based on native and decoy protein structures.

chemsum

{"title": "Random Forest Refinement of the KECSA2 Knowledge-based Scoring Function for Protein Decoy Detection", "journal": "ChemRxiv"}

generating_ampicillin-level_antimicrobial_peptides_with_activity-aware_generative_adversarial_networ

## Abstract: Antimicrobial peptides are a potential solution to the threat of multidrug-resistant bacterial pathogens. Recently, deep generative models including generative adversarial networks (GANs) have been shown to be capable of designing new antimicrobial peptides. Intuitively, a GAN controls the probability distribution of generated sequences to cover active peptides as much as possible. This paper presents a peptide-specialized model called PepGAN that takes the balance between covering active peptides and dodging non-active peptides. As a result, PepGAN has superior statistical fidelity with respect to physicochemical descriptors including charge, hydrophobicity and weight.Top six peptides were synthesized and one of them was confirmed to be highly antimicrobial. The minimum inhibitory concentration was 3.1µg/mL, indicating that the peptide is twice as strong as ampicillin. ## Introduction Antibiotic resistance is a serious and immediate threat against humanity, as currently available antibiotics become increasingly obsolete. The annual deaths due to antimicrobial resistance are expected to exceed 10 million by 2050. 1 Antimicrobial peptides (AMPs) are a possible solution to this problem. 2 They are considered as less prone to resistance, because microbes have been exposed to natural AMPs for millions of years, but widespread resistance against them has not been reported. Human experts have been developing successful AMPs by using simple hydrophilic/hydrophobic repeats, 3 or modifying an existing AMP. 4 Given the huge peptide space, however, it is very likely that numerous AMPs are yet to be found. Deep generative models 5 are one of the viable ways to boost the speed of AMP discovery, and several studies have been reported so far. Purely computational studies employing recurrent neural networks (RNN), 6 variational auto encoders (VAE) 7 and GAN 8 showed promising results in statistical terms, but experimental validation is yet to be done. Nagarajan et al. 3 were the first to show that a recurrent neural network can generate AMPs that works in vitro. They identified two peptides with minimum inhibitory concentration (MIC) 4µg/mL against E.coli. The potency of these peptides is at ampicilin-level, because their MIC is comparable to that of ampicilin (6.25µg/mL), a widely used antibiotic. Their neural network model has two parts. First, a recurrent neural network (i.e., generator) trained with known antimicrobial peptides generates a large number of peptides. Next, a classifier neural network trained with peptide-MIC pairs ranks the generated sequences, and top-ranked peptides are subject to experimental validation. Drawbacks of the model by Nagarajan et al. 3 are as follows. 1) LSTM is an obsolete model that is often outperformed by GANs. 9 2) The generator is trained only with positive examples (i.e., AMPs), despite the fact that a plenty of negative examples (i.e., non-AMPs) are available. Aiming to solve the drawbacks, we develop a specialized model called Pep-GAN by engineering LeakGAN, one of the state-of-the-art sequence generators. 10 PepGAN enhances the performance of LeakGAN with the help of activity predictor that is trained separately with positive and negative examples together (Figure 1). Another challenge in deep-learning-based AMP design is how to incorporate physicochemical properties such as charge, hydrophobicity, normalized van der Waals volume and polarity. Deep learning models are essentially a language model and it is not clear how to incorporate such information. To this aim, Nagarajan et al. 3 included several filtering steps in the model. Instead of complicating our model further, we simply chose to rerank PepGAN-generated peptides with an external AMP prediction tool (i.e., CAMP server 11 ) trained with various physicochemical features. As a result of our experimental validation, the MIC of the best peptide was as low as 3.1µg/mL, i.e., twice as strong as ampicilin. We made a python library of PepGAN publicly available at https://github.com/tucs7/PepGAN to contribute in the developing open-source ecosystem of peptide design. The generator samples a number of sequences stochastically. The reward for the sequences is evaluated by the discriminator, and transmitted back to the generator to update the parameters. In normal GANs, the reward function represents fidelity, i.e., how the sequences are similar to AMPs. In PepGAN, an activity predictor (shown in red) is incorporated in reward computation. Finally, top peptides are subject to experimental validation. ## Generative Model In various tasks including scheduling and maze solving, reinforcement learning has been used to generate a sequence of actions that maximizes a reward function. 12 A reward function represents the quality of a generated sequence. In traditional settings, it is given a priori and stays unchanged during training. When optimizing multiple reward functions at once, a linear combination of them is used in many cases. 13,14 Recently, Yu et al. 15 introduced SeqGAN that employs a machine-learned reward function for generating texts that resembles real sentences. A deep neural network called discriminator is trained to discriminate generated ones against real ones, and the training loss is adopted as the reward function. High reward implies high statistical fidelity: generated sequences are statistically indistinguishable from real ones. Later, SeqGAN is extended to LeakGAN 10 by introducing the ideas of hierarchical reinforcement learning. 16 In LeakGAN, the reward function for a sequence Y is designated as the output of the discriminator D(Y ), i.e., the probability of Y being real. The reward function of our model, PepGAN, is described as where F (Y ) is a separately-trained activity predictor, and λ denotes the mixing constant. The activity predictor has a GRU (Gated Recurrent Unit) 17 with 256 hidden variables. Given a sequence Y , it computes a hidden vector at each position. The hidden vector is fed to a one-layer dense neural network to yield a partial score at each position. ## Statistical Fidelity As training examples, we collect sequences not longer than 52 amino acids from the following databases: APD, 18 CAMP, 19 LAMP 20 and DBAASP. 21 Redundant sequences are removed via multiple sequence alignment with cut-off ratio 0.35. The final dataset contains 16648 positive sequences (i.e., AMPs) and 5583 negative sequences (i.e., non-AMPs). The activity predictor is first trained with all sequences and later the rest of PepGAN is trained only with positive sequences. PepGAN is used with three different parameter settings λ = 0, 0.5 and 1. Notice that LeakGAN corresponds to the case λ = 1. For each setting, 10000 peptide sequences are generated. We investigate the statistical fidelity of generated sequences from multiple viewpoints. The generated sequences are regarded as high-quality, if their statistics match well with those of the positive sequence set. First, we investigate the following physicochemical properties: length, molar weight, charge, charge density, isoelectric point, aromaticity, global hydrophobicity and hydrophobic moment. ModlAMP package 6 was used to compute these properties. Obtained statistics are summarized in Table 1. With respect to seven in eight properties, PepGAN with the activity predictor (λ = 0 and 0.5) were better than LeakGAN (λ = 1). This result shows that the activity predictor has a favorable impact in statistical fidelity. In the following experiments, λ = 0.5 is adopted, because it achieved the best result here. It is reported that samples generated by GANs tend to lose diversity due to mode collapse. 9 To check if mode collapse happened or not, the diversity of PepGAN-generated sequences is measured as follows. For each sequence, a BLEU score between that and all the other sequences is computed. The diversity score called self-BLEU is then computed as the average of all the BLEU scores. Table 3 shows self-BLEU scores for the positive sequence set (i.e., AMPs) and the generated sequences sets of PepGAN variations. In all cases, generated sequences were as diverse as the positive set, and mode collapse did not happen. Table 3: Self-BLEU scores based on k-grams (k = 2, 3, 4, 5) for three variants of PepGAN (λ = 0, 0.5, 1) and the positive sequence set (AMPs). AMPs λ = 0 λ = 0.5 λ = 1 Self-BLEU-2 0.965 0.969 0.970 0.970 Self-BLEU-3 0.802 0.835 0.842 0.846 Self-BLEU-4 0.550 0.592 0.608 0.621 Self-BLEU-5 0.393 0.372 0.381 0.405 ## Experimental Validation Fo experimental validation, generated peptides are prioritized according to the AMP likehood computed by the CAMP server. 11 Top six peptides are shown in Table 4. In addition, the worst four sequences are chosen as negative controls. Figure 2 shows helical wheel plots of these peptides 22 together with their hydrophobic moments. 23 Cell penetrating peptides tend to have a high relative abundance of positively charged amino acids, and contain an alternating pattern of polar and hydrophobic amino acids (i.e., amphiphilicity). 2 Our top peptides are observed as highly cationic and amphiphilic, because they contain a large number of positively charged amino acids and no negatively charged ones, and their hydrophobic moments are high (0.58±0.048). In comparison, negative controls are neither cationic nor amphiphilic. The potency of the peptides are evaluated based on minimum inhibitory concentration (MIC) against E.coli. MIC is determined as the minimum concentration of an antimicrobial at which the growth of a target microbe is suppressed. We found that as many as five out of six AMP peptides exhibited effective antimicrobial activity. Among them, AMP4 exhibited the best antimicrobial performance, 3.1 µg/mL, which is better than a well-known antimicrobial, ampicillin (6.25 µg/mL). The high production ratio, 5/6, and the sufficiently low MIC of AMP4 validate PepGAN's ability to generate industry-level peptides. In contrast, all four negative control peptides did not exhibit effective antimicrobial activity. ## Conclusion We presented PepGAN, a generative model for designing peptides, and demonstrated its statistical and in vitro success in AMP design. AMP-specific tricks are intentionally left out of our python library. Thus, our library can directly be applicable in development of other kinds of peptides such as drug-delivery peptides 24 and anti-cancer peptides. 25 To achieve our goal of boosting the speed of peptide development, experimental researchers, who are not necessarily familiar with machine learning, should be able to use computational tools such as PepGAN. Although we made our code public, we have not reached this level of utility. Open-source ecosystems in machine translation and computer vision are well-developed to the point that non-experts can use them without difficulty. In future work, we continue to develop PepGAN with an aim to make it a core of the emerging ecosystem of peptide design tools. ## Peptide Synthesis We synthesized all peptides on rink amide resin (ProTide, CEM corporation, NC, USA) using an automated microwave peptide synthesizer (Liberty Blue, CEM corporation). Each peptides were cleaved from resin and purified by reversed-phase HPLC using a C18 column (COSMOSIL 5C18-AR-II, Nacalai tesque, Japan) at 25 • C for 60 min with a linear gradient of acetonitrile in water containing 0.1% trifluoroacetic acid (v/v) at a flow rate of 1 mL/min. We confirmed the purified peptides using a MALDI-TOF MS (microflex, Bruker, MA, USA). ## MIC determination The MICs of the peptides were determined using the micro-dilution test with some modifications. We used em E.coli TOP10 (Thermo scientific, MA, USA) in the late-log phase for this test. The colony forming unit of E.coli was determined using optical density at 600nm. For each peptide we prepared 11 wells containing 150 µL of 5×10 5 CFU/ml of E.coli and the series of concentration of each peptide from 10 to 0.01 µg/ml (2-fold dilution for 10 times). We also prepared 1 well containing only 150 µL of 5 × 10 5 CFU/ml of E.coli without the peptide. Similarly, we prepared 11 wells for ampicillin in the same well-plate. We incubate the plates for 26 hours at 37 • C for the growth of E.coli. To read optical density for each well, we set the plates in a plate reader (EnSpire 2300, Perkin Elmer). After shaking the plate for 10 s at 300 rpm in double-orbital motion (diameter 1mm), we measured the optical density of each solutions (600 nm).

chemsum

{"title": "Generating ampicillin-level antimicrobial peptides with activity-aware generative adversarial networks", "journal": "ChemRxiv"}

[ag]2[b12cl12]_as_a_catalyst_in_phicl2_mediated_chlorination

## Abstract: The weakly coordinating [B12Cl12] 2originates from a family of carboranes typically reserved for application in coordination chemistry. Here, we show its readily accessible Ag(I) salt, [Ag]2[B12Cl12], can be used as a catalyst in the PhICl2 mediated chlorination of arenes, alkenes, and alkynes. The promising activity displayed by [Ag]2[B12Cl12] over a variety of commercially available Ag(I) sources merits its incorporation to the toolkit of commonly screened silver catalysts in synthesis. PhICl2, the first reported λ 3 -iodane compound, is a versatile oxidant, primarily acting as a chlorinating agent representing a convenient substitute for Cl2. Cl2 is a highly corrosive, toxic gas, which in addition to being hazardous, is challenging to deliver in a stoichiometric fashion. Conversely, PhICl2 is an easily weighed solid which is readily accessible from PhI, HCl, and H2O2, can be used without the need for rigorously anhydrous conditions, and has been used widely in the oxidation of organic and inorganic compounds. PhICl2, which can also be generated from a combination of PhI and Cl2, is not without limitation. It is necessarily a weaker oxidizing agent than the Cl2 it replaces and is unreactive towards many substrates. Activation of PhICl2 can be accomplished using Lewis acids with a handful of reports over the years, including by stoichiometric AgBF4 and SbCl5 in chlorination of norbornene derivatives, and by catalytic AlCl3 in the replacement of diazo groups with chlorines. Lewis acids such as BF3 have also been shown to increase the activity of the related oxidant PhI(OAc)2. Numerous groups over the years have used TMS-OTf to generate purported PhI(OTf)2 from PhI(OAc)2 as a stronger oxidant, however this has recently been shown to actually be PhI(OTf)(OAc). A recent paper by Nagib described the activation of PhI(OAc)2 using either HCl or acid chloride, or of PhICl2 using acetic anhydride, in each case giving a mixed PhI(OAc)(Cl) species capable of chlorinating the C-H bonds of a variety of (hetero)arenes in a few hours at 50 °C. Lupton employed the same concept a decade earlier using excess pyridinium chloride as the chloride source in concert with PhI(OAc)2 to chlorinate α,β-unsaturated carbonyls and alkenes (Scheme 1). Scheme 1. General classes of reported halogenation reaction using λ 3 -iodanes. In this report we show that abstraction of chloride from PhICl2 using catalytic amounts of silver salts of the weakly coordinating anion [B12Cl12] 2increases the activity of PhICl2 such that substrates unreactive or poorly reactive to PhICl2 can be rapidly chlorinated at room temperature. Our initial goal in this study was generation of the [Ph-I] 2+ dication, likely a highly reactive species. To achieve this we aimed to generate the [Ph-I][B12Cl12] salt, using the weakly coordinating and highly robust nature of the [B12Cl12] 2dianion to allow for an isolable or at least observable species. To this end, PhICl2 was reacted with stoichiometric [Ag]2[B12Cl12] in CHCl3. A 1 H-NMR spectrum of an aliquot of the reaction mixture revealed the presence of several species. Notably, similar reactivity was observed in the presence of catalytic (10 mol%) We have previously observed electrophilic aromatic substitution processes in reactions with electron poor λ 3 -iodane species, and therefore surmised that residual PhI generated from the decomposition of PhICl2 was undergoing electrophilic aromatic chlorination. [Ag]2[B12Cl12] was essential for the reaction to suggestive of an "iodonium" type mechanism, in which Ag(I) abstracts chloride from PhICl2 resulting in a in an [PhICl] + species, which is presumably stabilised by the weakly coordinating [B12Cl12] 2anion (Scheme 3). As discussed, attempts to isolate [PhICl] + or similar were unsuccessful. Scheme 3. Proposed [Ag]2[B12Cl12] mediated "iodonium" mechanism for electrophilic aromatic chlorination of iodobenzene. Encouraged by this preliminary reactivity, we decided to explore the efficacy of a number of other Ag(I) sources as mediators of electrophilic aromatic chlorination with PhICl2 using the electron rich arene, anisole (1), as an exemplar substrate (Table 1). Only minor conversion to 4-chloroanisole (2) was observed in the absence of any catalyst (entry 1). As expected, addition of 5 mol% [Ag]2[B12Cl12] resulted in substantially greater reactivity, which was further improved to a conversion of 67% upon doubling catalyst loading to 10 mol% (entries 2 & 3). Pleasingly, conversion was rapid (20 minutes) and occurred readily at room temperature. Changing solvent to acetonitrile (entry 4), or altering the nature of the counteranion and cation (entries 5 & 6) were all detrimental. AgCl, potentially generated in quantities up to 20 mol% during the catalytic cycle with [Ag]2[B12Cl12], was also investigated as a potential catalyst, and resulted little conversion at 20% loading (entry 7). [Ag]2[B12Cl12] is synthesized from relatively inexpensive and non-toxic precursors, primarily NaBH4, I2, and SO2Cl2. The Ag(I) salt is most conveniently obtained by metathesis using AgNO3 from a Cs salt. Unlike a number of related carborane reagents, which require the use of highly toxic and expensive reagents (i.e. B10H14), and are extremely time consuming to make, synthesis of [Ag]2[B12Cl12] is relatively straightforward, and can be delivered on a decagram scale in a few days in a typically equipped academic laboratory. Nonetheless, it was considered prudent to investigate several commercially available Ag(I) sources as alternative activators of PhICl2. Some conversion was observed in all cases, however activity was on the whole worse than [Ag]2[B12Cl12]. Reactions with AgOTf, AgBF4, and AgSbF6 (traditionally considered weakly coordinating), displayed similar levels of activity only when stoichiometric quantities of Ag(I) were used (entries 9, 11, & 13, respectively). Lower loadings resulted in activity comparable to the use of no Ag(I) at all. AgNO3 was the worst activator investigated, resulting in only minimal conversion even when deployed stoichiometrically (entry 14). The optimum conditions (i.e. entry 3) were then applied to a range of substituted arenes in order to investigate the scope of the system. For each substrate, a control reaction (% yield in brackets) was also performed in the absence of [Ag]2[B12Cl12] to probe its innate propensity to react with PhICl2 (Table 2). Anisole translated well to scale, with 4-chloroanisole (2) isolated as the sole isomer in 55% yield. 2,6-Dimethyl phenol was also isolated exclusively as the 4-chloro isomer (3), and performed much better in the presence of catalyst (i.e. 76% vs. 30%). Selectivity and yields were lower for unsubstituted phenol, with the 2,4-dichloro isomer (4) isolated in higher proportions in the presence of catalyst. Incorporation of an electron withdrawing group to the phenol ring was well tolerated, as demonstrated by ethyl salicylate, which was chlorinated in the 4-position relative to the hydroxyl group (6). Notably, this reaction did not proceed at all in the absence of catalyst. Napthalen-1-ol proved problematic, with a range of isomers (7-10) isolated in either event, although the 4-chloro isomer (10) was the major in each case. Rerunning the reaction at 0 ºC gave similar results and did not afford any improvement in selectivity. Using a phenol in which the 4position was blocked gave a mixture of 2-chloro isomers (i.e. 11 and 12). Interestingly, moving to propiophenone, an electron deficient arene, resulted exclusively in chlorination alpha to the ketone (13 and 14), with the aromatic ring left untouched. Introduction of an electron donating substituent marked a complete reversal in chemoselectivity (15). Neither substrate showed reactivity in a control reaction. 3,4,5-Trimethoxybenzoic acid, the most electron rich arene in the series, was the only member to display superior reactivity in the absence of catalyst, giving chloride 16 in 84% yield in a control experiment, but only 64% in the presence of [Ag]2[B12Cl12]. The reason for this remains unclear, but may be explained, in part, by the propensity of residual PhI (generated as a byproduct of successful SEAr) to undergo chlorination (as depicted in Scheme 2), thereby reducing the amount of PhICl2 available for productive pathways. The conditions were also successful in delivering chlorinated oxazolidinone 18 as a single isomer, as confirmed by HSQC and subsequently X-ray crystallography. Compound 18 is a structural analogue of the commercially available antibiotic Linezolid, and highlights the utility of this approach in late stage chlorination, an attractive strategy in drug design. Finally, heteroarenes were investigated, and unfortunately proved to be a limitation. Quinoline was not amenable to chlorination (19). 4-Dimethylaminopyridine, which is contrast is electron rich and activated towards SEAr, we have previously found is readily chlorinated without added Ag(I). Pyridine gave a mixture of species for which only pyridnium chloride could be identified. Given related methods (i.e. Nagib and Lupton) have both capitalised on PhI(OAc)(Cl), an active intermediate capable of delivering a single chlorine atom, and recent reports of the enantioselective dichlorination of alkenes, we speculated whether our methodology would be capable of activating PhICl2 to formally deliver a unit of molecular Cl2. To this end, the chlorination of several alkenes/alkynes was investigated (Table 3). Gratifyingly, this approach proved fruitful. Styrene delivered 1,2-dichloro styrene (20), albeit in modest yield. Methyl cinnamate was also readily chlorinated, giving the corresponding dichlorides (21 and 22) in a combined yield of 56% and a 2:1 d.r. in favour of the anti-isomer. Minor amounts of the elimination product, methyl β-chlorocinnamate (23), were also isolated. Diphenylacetylene gave the corresponding trans-dichloride, 24, as well as minor amounts of compound 25, presumably arising as a result of nucleophilic attack of residual PhI to the less hindered side of the transient vinyl cation. The structure of both compounds were confirmed by X-ray crystallography, with the trans-dichloride having been previously reported. In all examples, the presence of [Ag]2[B12Cl12] was essential, and reactions were completely chemoselective for exocyclic π-bonds over arenes. The electron poor dimethyl acetylenedicarboxylate (DMAD), was not tolerated under these conditions, under which no chlorinated adducts (26) were observed. In summary, we have demonstrated that catalytic [Ag]2[B12Cl12] can activate PhICl2 to act as a source of Cl + in the electrophilic aromatic substitution of arenes, and also to deliver a full equivalent of Cl2 in the chlorination of alkenes and alkynes. The reactions discussed herein likely proceed through the intermediacy of [PhICl] + via an "iodonium" mechanism, as opposed to a radical cation mechanism observed by others in related systems, and thereby present an attractive complimentary reactivity manifold. Further evidence for this comes from the fact that electron rich arenes outperformed their electron poor counterparts, and that chlorination was generally selective for positions on which the greatest delocalisation of partial negative charge would be expected. Whilst innate reactivity was observed with some arenes, in all but one substrate surveyed, [Ag]2[B12Cl12] resulted in enhanced reactivity. Presence of the Ag(I) salt was essential for the chlorination of alkenes and alkynes. Current usage of the [B12Cl12] 2dianion is largely limited to the inorganic community, where it enjoys a position amongst several related carborane reagents which act as superacids, an unparalleled source of strong electrophiles, and can be used in the isolation and X-ray crystallography of exotic carbocations. It is our hope that in demonstrating the superior activity of [Ag]2[B12Cl12] over several commonly used silver salts as a source of Ag(I), other practitioners will be encouraged to further investigate its application in related areas of organic synthesis.

chemsum

{"title": "[Ag]2[B12Cl12] as a Catalyst in PhICl2 Mediated Chlorination", "journal": "ChemRxiv"}

chiroptical_inversion_of_a_planar_chiral_redox-switchable_rotaxane

## Abstract: A tetrathiafulvalene (TTF)-containing crown ether macrocycle with C s symmetry was designed to implement planar chirality into a redox-active [2]rotaxane. The directionality of the macrocycle atom sequence together with the non-symmetric axle renders the corresponding [2]rotaxane mechanically planar chiral. Enantiomeric separation of the [2]rotaxane was achieved by chiral HPLC. The electrochemical propertiescaused by the reversible oxidation of the TTFare similar to a non-chiral control. Reversible inversion of the main band in the ECD spectra for the individual enantiomers was observed after oxidation. Experimental evidence, conformational analysis and DFT calculations of the neutral and doubly oxidised species indicate that mainly electronic effects of the oxidation are responsible for the chiroptical switching. This is the first electrochemically switchable rotaxane with a reversible inversion of the main ECD band. ## Introduction Evidenced by the homochirality in our biosphere, chirality is a fundamental principle, which governs the molecular recognition and activity of virtually all biomolecules. Therefore, gaining control over the preferred isomer of a molecule or an assembly by carefully designing a molecular system is a worthwhile endeavour. The term "chiroptical switch" has been used by Canary to refer to molecules, which are capable of "changes in their interaction with polarized light". 4 Potential applications are information processing, data storage and sensing. In this context, the ground breaking work of Feringa and co-workers on overcrowded alkenes, which act as light triggered chiroptical switches was awarded with the Nobel Prize in chemistry 2016 "for the design and synthesis of molecular machines" 8 and underlines the general interest in this topic. Mechanically interlocked molecules (MIMs) consist of parts that can move relative to each other guided by intramolecular forces. Therefore, we envisioned them to be ideal candidates for chiroptical switches in which coconformational or even confgurational changes in the MIM occur. An achiral wheel with directionality in its atom sequence forms a chiral rotaxane, when threaded onto a directional Scheme 1 (a) Reversible one-electron oxidations of the TTF moiety, (b) reversible oxidation of a directional crown ether wheel bearing a TTF unit, (c) chiroptical switching of the planar chiral rotaxane enantiomers. axle (Scheme 1). In 1997, Vögtle et al. reported on the frst resolution of a racemate of such mechanically planar chiral rotaxanes. 13 Chiral rotaxanes may be chiral from inclusion of classical stereogenic elements or by virtue of being mechanically planar chiral. Since then, several examples followed, in which the mechanically interlocked structure was used to induce directionality in polymers, for sensing, and to act as an enantioselective catalyst. 32 Today, sophisticated synthetic protocols allow an efficient enantioselective synthesis. For example, Goldup and co-workers 33,34 described elegant protocols to synthesise planar chiral enantiopure rotaxanes using readily available chiral auxiliaries. However, switchable planar chiral rotaxanes remain rare. So far, the modulation of chirality relies on heat, 21 the choice of solvent, anion exchange, 35 or pH. 36 Recently, we described redox-switchable rotaxanes, in which the wheels are decorated with tetrathiafulvalenes (TTF). TTF can be reversibly oxidised to the TTFc + and TTF 2+ states (Scheme 1). Large-amplitude motion and co-conformational changes in (oligo)rotaxanes were triggered by redox chemistry. 38, Apart from rotaxanes, TTF derivatives with covalently bound chiral substituents exhibited a chiroptical response to a change of their redox-state. Hence, our switchable rotaxanes display ideal optoelectronic properties since they are air stable in their neutral and oxidised form and show a clear-cut optical output, 37 which is even visible by the naked eye. In this paper, we report the synthesis, characterisation and optical resolution of a new mechanically planar chiral tristable rotaxane based on the 24-crown-8/secondary ammonium binding motif. 58 The rotaxane consists of the directional wheel dTTFC8 (Scheme 2), which is derived from a C 2v -symmetric TTF-decorated crown ether TTFC8 (Scheme 2) published by our group recently. 38 ECD measurements show reversible chiroptical switching, which can be explained mainly by electronic changes. The measurements are supported by quantum chemical calculations, which were also used to determine the absolute confguration. To the best of our knowledge, this is the frst example of a chiroptical switch with a complete sign reversal of the main band in the ECD spectra based on electronic changes in a mechanically bound assembly. ## Synthesis and characterisation The prerequisite for rotaxane formation is a sufficiently high binding constant between the crown ether and the ammonium axle. ITC experiments revealed an association constant of K a ¼ (3.6 AE 0.3) 10 5 M 1 and a 1 : 1 stoichiometry for pseudorotaxane formation from dTTFC8 and axle A1 (Scheme 2). The binding constant is very similar to that of our previous non-directional TTF-decorated wheel TTFC8 (K a ¼ (4.4 AE 0.4) 10 5 M 1 , for thermodynamic parameters see ESI, † Section 4), 38 which indicates the positional change of the TTF unit not to signifcantly affect the binding properties of the wheel. As for the non-chiral rotaxane 1, rotaxane formation was achieved with nitrile-oxide stopper St1 using a catalyst-free end-capping protocol established by Takata and co-workers 59 yielding a racemic mixture of rotaxane (rac)-2 (73%). The nonionic version (rac)-2Ac (95%) was obtained through N-acylation with Ac 2 O 60 (Scheme 2). The 1 H NMR spectra of (rac)-2 and (rac)-2Ac (Fig. 1) reveal a diastereotopic splitting of the macrocycle's methylene protons as well as of the axle methylene protons H h . 37,38 The splitting of both macrocycle and axle protons is characteristic for the formation of a chiral, yet racemic rotaxane. Isoxazole formation during stopper attachment leads to a strong downfeld shift of 3.88 ppm for proton H i . In (rac)-2, the S-methyl protons on dTTFC8 split into two singlets of the same intensity. Comparable rotaxanes also showed this behaviour on the same position. 27,28 HR-ESI mass and tandem MS experiments support the interlocked architecture (Fig. S1 †). For non-ionic (rac)-2Ac, the shift of H i (Dd ¼ +0.28 ppm) and H h (Dd ¼ +0.76 ppm) relative to (rac)-2 suggests that the wheel translates towards the isoxazole moiety in the absence of attractive interactions with the ammonium ion. Two sets of signals are observed due to the cis-trans isomerism of the amide bond in (rac)-2Ac. Variable temperature NMR experiments (Fig. S3 †) in DMSO-d 6 reveal the same barrier (DG ‡ ¼ 74 AE 2 kJ mol 1 ) for amide cis-trans isomerisation as observed for a similar acetylated rotaxane. ## Optoelectronic properties Photometric titrations of (rac)-2 and (rac)-2Ac with Fe(ClO 4 ) 3 (Fig. 2a and b) show similar bands for the three redox states (TTF, TTF + c and TTF 2+ ) of both rotaxanes with distinct isosbestic points. These fndings are consistent with structurally related rotaxanes featuring a non-directional TTF-decorated wheel. 38 Cyclovoltammetric (CV) experiments were conducted with dTTFC8, (rac)-2 and (rac)-2Ac in dichloromethane (Fig. 2c). The potentials for (rac)-2 (116 mV and 407 mV) are considerably higher for both oxidation steps as compared to dTTFC8 (64 mV and 362 mV). Both oxidations are thus energetically disfavoured because of the charge repulsion between the TTF cation radical as well as the TTF dication and the ammonium station. In case of (rac)-2Ac (18 mV and 392 mV) the frst oxidation is more easily accomplished in comparison to the free macrocycle and the second oxidation is disfavoured. We attribute this behaviour to a stabilising interaction with the isoxazole moiety on the axle for the frst oxidation. For the second oxidation, the limited accessibility of the TTF 2+ by counterions caused by the steric demand of the axle needs to be taken into account. Again these trends were already observed for the non-directional macrocycle and rotaxanes thereof. 38 The reversibility of the redox-waves of (rac)-2 and (rac)-2Ac strongly indicated that the interlocked structures remain intact during the redox switching, however it is reasonable to assume conformational changes to occur due to charge repulsion and charge stabilisation. The data does not show any signifcant change in the electrochemical properties by introducing directionality into the TTF decorated wheel. ## Enantiomer separation on chiral HPLC and CD spectroscopy The two enantiomers of (rac)-2Ac could be separated using HPLC with a CHIRALPAK® IA stationary phase. The optical purity was determined (>99% ee; Fig. 3a) and mirror-image CD spectra were obtained for the neutral enantiomers with bands at 242 nm and 325 nm (Fig. 3b). We assigned the absolute confguration based on the computational results (see below). The oxidised species 2Acc + and 2Ac 2+ show bands at the same wavelengths. While no sign inversion occurs at 325 nm, the band at 242 nm exhibits a sign inversion during the frst and a signifcant intensity increase during the second oxidation step. To exclude decomposition to be responsible for the switching, 2Ac 2+ was reduced back to the neutral state using Zn dust and then showed the initial CD spectrum again (Fig. 3d and e dashed lines). Surprisingly, no other CD signals are observed at a higher wavelength, although the change in UV/Vis absorption is most pronounced at 460 nm and 844 nm for the radical cation and at 703 nm for the dication (Fig. 2b). The reason for the sign change remains ambiguous. In fact, conformational changes were observed to induce transitions in CD spectra of non-interlocked TTF derivatives with centrochiral elements earlier. 53,64 Other examples show varying intensities 54 or shifts of the maxima 56 upon oxidation of the TTF attached. Nevertheless, no TTF derivative is reported that shows a sign reversal in the maximum of an ECD spectrum without a shift in the wavelength. Apart from TTF derivatives, chiroptical switching via a redox process can be achieved with catechol, 65 viologen, 66 and tetraarylethylene 67 building blocks. Intense switching with a sign reversal was also observed for a viologentype dicationic helquat. 68 Chiral inversion can also be achieved with metal ion complexation 69,70 acid-base-71 and photoswitching. 6,68,72 Computational results To investigate whether the redox-induced sign inversion at 242 nm in the ECD spectra of (R mp )-2Ac is due to a change in its electronic properties or to a (co-)conformational change, density functional theory (DFT) calculations were performed at the TPSS-D3(BJ) and uB97X-D3 (ref. 76) levels. Conformational analyses reveal the structure depicted in Fig. 4 (left) to be the most stable one for (R mp )-2Ac. It is at least 18 kJ mol 1 more favourable than any other possible conformation found by theory (see Table S2 †). For (R mp )-2Ac 2+ , there are two conformations relatively close in electronic energy: Conformer A (Fig. 4 middle) and B (Fig. 4 right) with a flipped naphthalene unit, ca. 9 kJ mol 1 more stable than A. This conformational change is explained by the oxidation of (R mp )-2Ac occuring fairly localised at the TTF unit. 77 The emerging charge of the oxidised TTF moiety is then stabilised by the naphthalene that moves into close proximity of the TTF 2+ . Additionally, an atoms-inmolecules (AIM) analysis suggests that the electrostatic attraction between the naphthalene and TTF moieties outweighs all other non-covalent interactions for (R mp )-2Ac 2+ , while the maximisation of non-covalent interactions (C-H/p and p-pstacking) is the most important factor in the neutral state (see ESI † for details). The simulated CD spectra in Fig. 4 were obtained using simplifed time-dependent DFT 78 at the uB97X-D3 level. The spectrum of (R mp )-2Ac shows a deviation of around 40-50 nm, while that of (R mp )-2Ac 2+ is off by less than 20 nm compared to experiment. The experimentally detected sign inversion at 242 nm is reproduced well by the calculations. The conformational change of (R mp )-2Ac upon oxidation, however, hardly influences the shape of the CD spectra as both conformations yield very similar CD spectra in the region between 230 and 400 nm. Therefore, we exclude the conformational change as the prime origin of the sign inversion. To rationalise the optical behaviour of (R mp )-2Ac and (R mp )-2Ac 2+ , we examined its valence electronic structure, which is, as expected, dominated by orbitals localised at the TTF moiety (see Fig. S20 †). Analysing the electronic transitions in the spectral region between 230 and 400 nm reveals that practically every excitation involves the TTF unit to some extent. While many transitions are of local nature, i.e., between orbitals in close proximity, quite a few display a chargetransfer-like behaviour (insets Fig. 4 and S21 †). For neutral (R mp )-2Ac, the vast majority of these transitions can be described by advancing an electron from an orbital centred at the TTF core, usually the HOMO, into an orbital located in another part of the rotaxane (e.g. the dimethoxy-phenyl moiety). For (R mp )-2Ac 2+ , the corresponding transitions progress from some orbital in the molecule into an orbital localised at the TTF moiety, usually the LUMO or LUMO+1. This induces differently oriented magnetic dipole transition moments leading to different signs in the CD spectrum. Hence, we conclude that the sign inversion in the CD spectra upon oxidation can be exclusively attributed to the change of the electronic structure. ## Conclusions In conclusion, electrochemically switchable crown ether/ ammonium rotaxanes bearing a directional wheel are reported. The wheel features a redox-switchable TTF unit. The directionality had no observable impact on the electrochemical and optical properties of the racemic mixtures determined by UV/Vis spectroscopy and CV measurements. Instead, the pure enantiomers of the acetylated non-ionic derivatives display a redox-induced reversible inversion of the sign in the ECD spectrum without a change of absolute confguration. The mechanism and the absolute confguration of this chiroptical switch has been examined by computational methods. While co-conformational changes have hardly any impact on the ECD spectra, the changes in electronic structure induced by oxidation play a pivotal role. These results underline the impact of the mechanical bond, which allows the construction of intriguing switchable chemical assemblies with unexpected properties. This is the frst in class example of a redoxcontrolled chiroptical switch with a complete sign reversal based on a mechanically planar chiral rotaxane. In the future, these properties could be employed in materials science to construct novel optoelectronic building blocks.

chemsum

{"title": "Chiroptical inversion of a planar chiral redox-switchable rotaxane", "journal": "Royal Society of Chemistry (RSC)"}

an_aqueous_electrolyte_of_the_widest_potential_window_and_its_superior_capability_for_capacitors

## Abstract: A saturated aqueous solution of sodium perchlorate (SSPAS) was found to be electrochemically superior, because the potential window is remarkably wide to be approximately 3.2 V in terms of a cyclic voltammetry. Such a wide potential window has never been reported in any aqueous solutions, and this finding would be of historical significance for aqueous electrolyte to overcome its weak point that the potential window is narrow. In proof of this fact, the capability of SSPAS was examined for the electrolyte of capacitors. Galvanostatic charge-discharge measurements showed that a graphitebased capacitor containing SSPAS as an electrolyte was stable within 5% deviation for the 10,000 times repetition at the operating voltage of 3.2 V without generating any gas. The SSPAS worked also as a functional electrolyte in the presence of an activated carbon and metal oxides in order to increase an energy density. Indeed, in an asymmetric capacitor containing MnO 2 and Fe 3 O 4 mixtures in the positive and negative electrodes, respectively, the energy density enlarged to be 36.3 Whkg −1 , which belongs to the largest value in capacitors. Similar electrochemical behaviour was also confirmed in saturated aqueous solutions of other alkali and alkaline earth metal perchlorate salts. It has been generally understood that aqueous electrolytes have many advantages compared with non-aqueous solvents with respect to electrochemical behaviour as well as economic and environmental impacts. However, there exists serious disadvantage in aqueous electrolytes that water is easily electrolyzed to generate gases. This is attributed to the narrow electrochemical potential windows of aqueous solutions. As a matter of fact the thermodynamic potential window of water is known to be 1.23 V. For the study on capacitors the width of potential window is essentially important, since the energy storage in electric double-layer capacitors is proportional to the square of applied voltage 1 . Therefore, the use of aqueous electrolytes in capacitors should be accompanied by the extension of the potential window. Numerous attempts have been reported to extend the potential window of water. In 5 M (M = mol dm −3 ) LiNO 3 aqueous solutions 2 , the potential windows were determined to be 2.3 V(− 0.55~1.75 V) by a constant current (50 μ A cm −2 ) method. The potential window of 2.0 V was reported in a 0.1 M KCl unbuffered aqueous electrolyte by using nanostructured platinum electrodes, where the change in local acidity at the electrodes contributed to the expansion of the potential window. A gel electrolyte 3 consisting of the mixture of polyvinyl alcohol and KOH aqueous solution gave the potential window of 2.0 V. Much attention has been focused on electrode materials of high overpotentials for oxygen and/or hydrogen evolution. An attempt was made for supercapacitors (we will call just capacitor hereafter) consisting of the composites of carbon nanotubes with MnO 2 in the positive electrode/active carbon in the negative electrode in KNO 3 aqueous solution gave 2.0 V for the maximum charging voltage 4 . Other attempts to improve the operation voltage of aqueous electrolytes were reported by using asymmetric and symmetric capacitors, i.e. MnO 2 /carbon asymmetric capacitors , and a carbon/carbon symmetric capacitor 9 . In the case of MnO 2 /graphene capacitor in 1 M Na 2 SO 4 aqueous solution 5 , a cell voltage was 2.0 V giving a large energy density of 25.2 Whkg −1 . In asymmetric aqueous capacitors, MnO 2 / nanoporous activated carbon 6 and MnO 2 /high purity carbon nanotubes 7 , the operation voltages were 1.5 V, and 2 V, respectively. In a symmetric carbon/carbon capacitor consisting of a homogeneous mixture of 80% activated carbon, 10% of acetylene black and 10% of binder 9 , the operation voltage was 1.6 V. In a review paper 1 , data are summarized on a variety of capacitors using various aqueous electrolytes and various electrodes. However, the potential window listed was 1.8 V at the maximum. As a result from earlier studies , the potential window of aqueous electrolytes is limited at around 2 V even though by using any specific electrode materials. Although the expansion of the water window to 2 V is a great advance in aqueous electrolytes, it is still not enough to replace non-aqueous electrolytes with aqueous electrolytes. The aim of the present study is to expand the potential window of aqueous electrolytes based on the view point of solution chemistry. At first, a question arises why water is easily electrolyzed. There are two major impacts to contribute the electrolysis of water. One is the acid equilibrium of water and the other is the hydrogen bond between water molecules. It is well known that electrolysis of water is sensitive in the acidity. However, to our knowledge, no one has paid attention electrochemically with respect to the hydrogen bond of water. The hydrogen bond is originated between a hydrogen and a neighboring oxygen (or other electronegative atoms) at the distance shorter than 0.3 nm. There is a trade-off relationship between the covalent O-H bond and the hydrogen bond 11 , that is, weaker the hydrogen bond, stronger the O-H bond. This leads to a conclusion that the O-H bond of water should be strengthened if a water molecule is isolated from others because of the weakened hydrogen bond. As a matter of fact, the O-H bond strength in isolated gaseous water is even stronger than the O-H bonds of methanol and ethanol, i.e. 497 kJmol −1 in water is compared with 436 and 437 kJmol −1 of methanol and ethanol 12 . The addition of ions into water disturbs the hydrogen bond network, where the electronegative oxygen atom of water is attracted to positive ions and the electron poor hydrogen atom is attracted to negative ions. In the case of saturated sodium perchlorate aqueous solution (SSPAS), since sodium perchlorate is extremely soluble in water (the solubility is 219.6 g in 100 g water at 25 °C), only 3.3 water molecules exist per one sodium perchlorate molecule. Under such a condition the hydrogen bond could be destroyed almost completely due to the lack of neighboring water molecules and also the strong hydration to Na + and ClO 4 − . The acid equilibrium could be also limited under the condition that the content of free water is too low for H + to form hydronium ion, H 3 O + . Note that H + has considerably large hydration energy among mono-valent ions, and hence only the hydrated species is able to exist in aqueous solutions. We do not know in detail, but it is true that the evaporation rate of SSPAS is extremely slow owing to the strong hydration. This unusual phenomenon is known to be a theoretical base that water may exist in Mars 13 . Similar behaviour was also reported in a saturated magnesium perchlorate aqueous solution 13 . This suggests that the saturated magnesium perchlorate aqueous solution (SMPAS) would possibly be a superior electrolyte. Consequently, it is not surprising that the potential windows of SSPAS is extended to the region of non-aqueous solvents because of the enhanced strength of O-H bond due to the loss of hydrogen bond. In the present paper, we will report a definite evidence for the widest potential window of SSPAS and demonstrate its superior capability as the electrolyte for capacitors using only low cost standard materials for this purpose. ## Results Cyclic voltammetry. Cyclic voltammetry (CV) is most commonly used to investigate the electrochemical properties of electrolytes and electrode materials . The CV measurements for SSPAS were performed together with typical acidic (1 M H 2 SO 4 ) and basic (1 M NaOH) aqueous solutions by a BAS CV-50W using a glassy carbon for a working electrode, a platinum counter electrode and an Ag/AgCl reference electrode under the argon atmosphere. Measurements were made separately for the positive scan to 2.3 V and for the negative scan to − 2.0 V at the scan rate from 30 mVs −1 to 200 mVs −1 at 25 °C. The cyclic voltammograms are exhibited in Fig. 1. Because of large over potential of the glassy carbon used for the working electrode, the potential windows are relatively wide. Even though, the potential window of SSPAS is remarkably large compared with those of typical acidic and basic solutions. The potential window of SSPAS was determined to be approximately 3.2 V from the cyclic voltammogram. ## Capability of SSPAS as an electrolyte in graphite-based capacitors. A proposed scheme of the electrical double-layer capacitor for SSPAS is illustrated in Fig. 2 under the charging, where Na + and ClO 4 − might be strongly attracted to negative and positive electrodes respectively because of the weakened shield due to limited number of water molecules. The capability of SSPAS as the electrolyte for capacitors was examined first by a simple symmetrical graphite-based capacitor consisting of the following mixture: 80% graphite, 10% acetylene black and 10% carbon felt (we call hereafter a graphite acetylene black mixture or simply a graphite mixture abbreviated as GA). Then GA was made to form a thin film under the pressure as described later, where SSPAS well penetrated through thus made GA film compared with a pure graphite film. Galvanostatic charge-discharge experiments were performed under the following conditions: cut-off voltage 3.2 V, current density 15 mA cm −2 using a cation exchange membrane (CEM) as a separator, where voltage was plotted against capacity (mAh). For testing the reliability of SSPAS, the charge and discharge measurements were repeated 10,000 times (Fig. 3a). As seen in Fig. 3a, except the first charging plot, the galvanostatic cycles are consistent within 5% deviations for 10,000 cycles. The average energy density was 0.45 Whkg −1 (based on the total mass of active materials) and 76% of the charge and discharge efficiency and the time spent for one cycle was 12 s. The detailed gas analysis was carried out after the charge-discharge measurements by the gas chromatogram using a Shimadzu GC-2014 under the helium gas flow at the rate of 25 ml min −1 , where the whole cell was vacuumed before its opening for the analysis of inside gases. Hydrogen gas was not detected and oxygen was detected, but within the background. The results definitely indicate that SSPAS behaves well as the electrolyte at the operation voltage of 3.2 V. The addition of activated carbon (AC) to both negative and positive electrodes forming a symmetrical capacitor increased the energy density remarkably as shown in Fig. 3b-e. Since the complete water free activated carbon was not easily obtained, the charge-discharge curves were affected by a small amount of water contained, the curves were unstable above 3 V. Therefore we chose the cut off-voltage at 3 V only in the capacitors consisting of the activated carbon mixtures. In Fig. 3b, cation exchange membrane (CEM) was used as the separator, while in Fig. 3c and d, a membrane filter (MF) and a filter paper (FP) were also used as separators, to compare with the 13 and number of free water molecules would not be enough to form a hydrogen bond. result of using CEM. The result indicates that there exists little difference in the shapes of charge-discharge curves, and in the values of the energy density, that is 7.3~8.3 Whkg −1 , by the use of these separators. The times spent for one cycle (Ts) were all similar in Fig. 3b,c and e to be about 300 s. The increased addition of AC increased the energy density to be 18.7 Whkg −1 at 40% AC (Fig. 3e). ## Effect of metal oxides for the graphite-based capacitor. It has been known that metal oxides contribute to store energy in capacitors. Among a variety metal oxides, we have examined the capability of SSPAS for graphite-based capacitors containing naturally abundant metal oxides such as Fe 2 O 3 and Fe 3 O 4 , V 2 O 3 and V 2 O 5 , and MnO 2 . of AC with a CEM, (c) with a MF and (d) with a FP as the separator (e). Charge and discharge cycles of a symmetric graphite capacitor containing 40% of AC with a MF as the separator. Other specific conditions are described in Table 1. First of all, we tried to examine the individual contributions in negative and positive electrodes especially for an asymmetric capacitor by using a reference Ag/AgCl electrode. We chose the electrode consisting of graphite mixture as a working electrode and the electrode containing 30% Fe 2 O 3 as a counter electrode. Experiments were carried out under the N 2 gas bubbling at 25 °C. The galvanostatic charge-discharge cycles plotted against time(s) are exhibited in Fig. 4-1, where (a) refers to the potential in the counter electrode, (b) the potential in the working electrode and (c) refers to the cell voltage. It can be seen in this figure that the plots (a) and (b) were almost symmetrical to the reference. This means that the electron absorption in the negative electrode and loss in the positive electrode takes place simultaneously. The cell voltage is the sum of the absolute values of (a) and (b). In Fig. 4-2, the galvanostatic charge-discharge cycles, the cell voltages versus capacity (mAh), are shown for various symmetric and asymmetric capacitors containing metal oxides, Fe 2 O 3 , Fe 3 O 4 , V 2 O 3 , V 2 O 5 and MnO 2 . As seen in these figures, the charge-discharge curves are different in shape from those of usual capacitors, particularly in the symmetric capacitors containing iron oxides, where the curves looked like those in typical rechargeable batteries having plateaus. The addition of metal oxides brought remarkable gain in the energy density. This may owe to the redox effects by metal oxides. It should be noted in symmetric capacitors (4-2b) and (4-2d) containing Fe 2 O 3 and Fe 3 O 4 , respectively that the charge and discharge curves moved toward one direction according to the repetitions. It may be expected that a redox reaction takes place between Fe(II) and Fe(III). Similarly, redox reactions are also expected in the addition of V 2 O 3 and V 2 O 5 , because vanadium possesses four oxidation states from V(II) to V(V). MnO 2 is used in positive electrodes in (g) and (h) because of its electron emitting in nature. The results and specific conditions are summarized in Table 1. Although sodium perchlorate is most soluble in any perchlorate salts, there exist other metal perchlorate salts of high solubility in water. Particularly, saturated lithium perchlorate aqueous solution behaves most similarly to SSPAS as exhibited in charge and discharge curves of a symmetric capacitor consisting of graphite mixture (Fig. 5a). However, our interest has been focused on more naturally abundant metal perchlorate salts, Mg(ClO 4 ) 2 , Ca(ClO 4 ) 2 , Ba(ClO 4 ) 2 and Al(ClO 4 ) 3 . Galvanostatic cycles of symmetric capacitors containing these saturated aqueous solutions as electrolyte are shown in Fig. 5b-e. It is interesting to see in the discharge curves in (b) Mg(ClO 4 ) 2 and (e) Al(ClO 4 ) 3 that the curves are not linear, being larger in the capacity and hence having larger energy densities compared with SSPAS. This might owe to the partial reductions of Mg 2+ and Al 3+ to Mg and Al respectively during charging like as rechargeable batteries. ## Discussion In Fig. 1, from the CVs of dilute aqueous solutions, the decomposition voltage shifts towards positive direction from basic (1 M NaOH) to acidic (1 M H 2 SO 4 ) solutions as usual. The CV of SSPAS indicates that SSPAS is oxidized at more positive voltage than the decomposition voltage of 1MH 2 SO 4 to be about + 1.6 V and reduced at more negative voltage than the decomposition voltage of 1 M NaOH to be about − 1.6 V. This leads to the potential window of SSPAS is approximately 3.2 V. The potential window thus determined above 3 V is largest being ever reported for any aqueous solutions. It should be noted that the CV curve for SSPAS is nearly symmetrical. Since the concentration of H 3 O + and OH − would be very low because of the lack of free water to hydrate in SSPAS, the electrolysis could be caused by the direct decomposition of the OH bond of water, and hence the potential barrier is expected to be the same in the oxidation and in the reduction. This reflects the nearly symmetrical CV curve of SSPAS. The large potential window of SSPAS could be resulted from the weakened hydrogen bond of water. In order to evaluate the hydrogen bond, NMR measurements were performed, since the hydrogen bond of water is related to the chemical shift of 1 H NMR signal. In an extreme case, under the supercritical condition, where water approaches to gaseous form and hence the hydrogen bond should be weakened, the chemical shift of 1 H NMR signal of water moves to the higher magnetic field compared with those measured under normal conditions 20 . The 1 H NMR of water was measured in SSPAS and in 1 M NaClO 4 aqueous solution, respectively, and the chemical shifts were determined to be 3.69 ppm in SSPAS and 4.76 ppm in 1 M NaClO 4 aqueous solution. This upfield shift in SSPAS is attributed to the weakened hydrogen bond in SSPAS despite the downfield contribution due to strong hydration towards Na + , though the quantitative contribution by the hydration is not known. Furthermore, the line-width at the half height in SSPAS was 0.0269 ppm, which was much smaller than 0.0483 ppm in 1 M NaClO 4 aqueous solution. This narrowing of the line-width in SSPAS can be explained by the weakened dipole-dipole coupling between the neighboring proton spin through the hydrogen bond 21 . The result in the CV measurement for SSPAS showing the large potential window was consistent with the stable galvanostatic charge-discharge performances at the high operation voltage in Fig. 3. In Fig. 3a, the galvanostatic charge-discharge cycles for the symmetric graphite-based capacitor were stable within 5% deviations for 10,000 cycles and the time spent for one cycle was 12 s. In this figure, the charging curves deviate from the linearity at the high applied voltage. We estimate in storing energy that the electrical double-layer process as illustrated in Fig. 2 could be combined by an additional redox process owing to an electron adsorption and an emission in the negative and positive electrodes, respectively as written bellow. x x The deviations from the linearity in galvanostatic curves become larger in the presence of activated carbon (AC) as seen in Fig. 3b,c and d, where the discharge curves also deviate from the linearity. The energy densities increased more than the ten times by the addition of AC compared with the capacitor consisting of only graphite-acetylene black mixture (Fig. 3a). In Fig. 3b,c and d, the galvanostatic curves are similar with the similar energy densities in spite of using different separators, CEM, MF and FP, respectively. The time spent for one charge-discharge cycle (Ts) were about 300 s in all cases. These results indicate that the mobility of ions through separators is not important. As a matter of fact, Na + and ClO 4 − are able to pass through MF and FP, while only Na + can pass through CEM. The increased addition of AC gives a considerable gain in the energy density as shown in Fig. 3e, i.e. the symmetric capacitor consisting of 40% activated carbon yielded the energy density of 18 Whkg −1 . This value is quite large in carbon-based capacitors 14,21,22 , despite of using standard inexpensive carbon materials. It should be noted that the cycle time (Ts) also increased by the addition of AC from 12 s of the graphite-based capacitor to 750 s of the capacitor containing 40% AC. These results suggest that the storing energy does not proceed simply through an electric double-layer process, but proceeds through composite processes. On the assumption that the cycle time (Ts) is caused by the diffusion controlled mechanism for the simple graphite capacitor, the enlarged Ts due to the addition of AC would be attributed to a slower additional process. The latter slower process is estimated to be an electrochemical process such as an electron adsorption-emission reaction. As exhibited in Fig. 4-2, the galvanostatic curves of the capacitors containing metal oxides are different in shapes from those of the carbon-based capacitors in Fig. 3. The discharge curves of the asymmetric capacitors containing Fe 2 O 3 , Fe 3 O 4 , V 2 O 3 or V 2 O 5 in the negative electrodes are all similar, i.e., slowly decrease until 1.7 V in the case of Fe 2 O 3 and Fe 3 O 4 , and until 2.0 V in the case of V 2 O 3 and V 2 O 5 , then decrease relatively fast to zero. Consequently, the energy density is larger in the later cases. As seen in Fig. 4-2b (containing Fe 2 O 3 ) and Fig. 4-2d (containing Fe 3 O 4 ), the charge and discharge curves of the symmetric capacity look no longer like capacitor, instead rechargeable batteries having plateaus at the voltage above 1.0 V. A number of studies have be concerned with respect to the role of a variety of metal oxides, such as RuO 2 , MnO 2 , Fe 3 O 4 , IrO 2 and V 2 O 5 for capacitors . Especially, RuO 2 has been studied most extensively because of its conductivity and capability of fast reversible electron transfer between multiple oxidation states within 1.2 V as written below 15 . where 0 ≤ x ≤ 2 Since the acid equilibrium is unknown under the present condition in SSPAS, the equilibrium (3) might be preferably written as below. In the case of vanadium oxides V 2 O n (where n is 3 or 5), the mechanism could be similar as written below. where x is not necessarily integer. The existence of equilibrium (4) was supported by the fact that the negative electrode became basic (pH = 11.3 measure after the dilution by water) after the full charge of capacitor. We do not deny the possibility of redox reactions of vanadium itself, because vanadium possesses four oxidation states from 2 to 5 within 1 V. On the other hand, in the positive electrode, the GA releases electron to form a positively charged form as described by the equation ( 2) or an equilibrium analogous to (4) as written below. x 2 The equation (2′ ) would be more favorable than (2), because the positive electrode is acidic (pH = 2.31 measured after dilution by water) after the full charge. XRD measurements were carried out to examine the structural change in graphite mixture (GA) by the full charge of the capacitor, i.e., GA containing 10% V 2 O 3 in the negative electrode and only GA in the positive electrode. As seen in Fig. 6-1, the diffraction pattern of the negative electrode exhibits a sharp peak at 2θ = 26°(a), which is characteristic of graphite 28 . On the other hand, the same peak at the positive electrode (b) is broadened after the charge. This indicates that the graphite in the positive electrode tends to lose a distinct structure to be amorphous according to the release of electron, but not to form graphite oxide 29 . A similar XRD result was also confirmed in the presence RuO 2 . An XPS measurement was carried out for samples produced in the same way as described in Fig. 6-1 under the same condition by a ULVAC PHI 5000 VersaProbe III. The spectra are exhibited in Fig. 6-2 together with that of a graphite sample before use, which does not contain perchlorate. The peak corresponding to the graphite is assigned at 284.4 eV 30 appearing at the middle. The spectra of the samples after use are observed at the edge of the large oxygen 1 s peak of perchlorate. The peak of the negative electrode (red line) shifts slightly to the lower energy and the peak of the positive electrode shifts slightly to the higher energy. The shifts are within 0.5 eV, which is too small to prospect any major change in 2p orbital of graphite during charge and discharge. As a matter of fact, peaks of graphite oxide were reported to be larger than 286 eV 30 . In conclusion, the result of XPS is in agreement with that of XRD, where major difference in chemical bond would not take place during charge and discharge. As described above, the role of vanadium and iron oxides in the negative electrode is to adsorb electron during the charge. On the other hand, as seen in Fig. 4-2g, MnO 2 was effective in the positive electrode. This means that MnO 2 emits electron during the charge as reported in an earlier paper 5 . On the basis of the above assumption for the adsorption and for the emission of electron by metal oxides, we made capacitors containing 30% of MnO 2 in the positive electrode. The charge and discharge cycles are exhibited in Fig. 4-2g and h, where GA with 20% AC and 30% Fe 3 O 4 was involved in the negative electrode, respectively, and the energy densities and the time spent for one cycle (Ts) were 16.9 Whkg −1 and 24.1Whkg −1 , 306 s and 442 s, respectively. The result indicates that Fe 3 O 4 is more effective than AC for larger energy density in the capacitors containing MnO 2 in the positive electrode. A capacitor containing 60% of Fe 3 O 4 and MnO 2 in the negative and in the positive electrode gave the energy density of 36.3 Whkg −1 and Ts was 700 s. The energy density is largest of the present series of experiments. In earlier papers 5,7,14 , galvanostatic charge-discharge measurements for capacitors containing manganese oxide in positive electrodes were performed in aqueous solutions. The results are summarized as follows. a: MnO 2 /graphene in 1 M Na 2 SO 4 , operation voltage 2 V, current 5 mAcm −2 and Ts 900 s 5 , b: MnO 2 /AC in 2 M KNO 3 , operation voltage 2.2 V, current 100 mAg −1 and Ts 1200 s 7 , c: α MnO 2 /graphene in 6 M KOH, operation voltage 2 V, 1 Ag −1 and Ts 1000 s 14 . In these earlier studies, Ts was around 1000 s, though the conditions were all different. A quantitative comparison for these results involving the present study is difficult, since the materials used, the electrolytes and the galvanostatic conditions are all different. Even though, the charge-discharge rates are all similar from 700 s to 1200 s. This fact indicates that the rate of storing energy would be determined by (2) XPS spectra for carbon 1 s of graphite. The sample were taken in the same way as in Fig. 6(1) under the same condition. Because of the large 1 s signal of perchlorate oxygen, the S/N ratio of the sample signals are lower compared with the standard peak of original graphite (green). The red peak corresponds to the negative electrode and blue one positive electrode, respectively. electrochemical reactions instead of a diffusion controlled mechanism, and that SSPAS behaves well as equal as other dilute aqueous solutions. It is interesting to see the charge and the discharge curbs for the symmetric capacitor containing Fe 2 O 3 (Fig. 4-2b) and Fe 3 O 4 (Fig. 4-2d), because they look like as rechargeable batteries having plateaus in the discharge curves, and furthermore the curves shift to one direction. We do not know in detail at moment, but it may be estimated that a redox reaction takes place between Fe(II) and Fe(III) during the charge and the discharge keeping the same oxide structures. The electron exchange reaction between Fe 2+ and Fe 3+ has been long known or it is the history of electron exchange reaction 31 . At moment, the detailed analysis is not possible, but it is important for understanding the charge-discharge mechanism in metal containing capacitors. It has been found that perchlorate salts other than NaClO 4 , such as LiClO 4 , Mg(ClO 4 ) 2 , Ca(ClO 4 ) 2 , Ba(ClO 4 ) 2 and Al(ClO 4 ) 3 , are very soluble in water and that their saturated aqueous solutions works as excellent electrolytes. Galvanostatic cycles of symmetric graphite-based capacitors containing saturated aqueous solutions of these salts are exhibited in Fig. 5 with their specific conditions in Table 2. The saturated lithium perchlorate aqueous solution is quite similar to SSPAS (Fig. 5a). However, we are more interested in naturally abundant other perchlorate salts. Particularly, the saturated Mg(ClO 4 ) 2 aqueous solution (SMPAS) is the superior electrolyte as well as SSPAS. As seen in Fig. 5b and Table 2, and the current density is 40 mAcm −2 , which is largest at the present study, and hence raises the power density as large as 472 Wkg −1 . The discharge curve deviates from a linear line being slow down at below 1 V and such a discharge curve increases the energy density to be 1.2 Whkg −1 , which is more than double compared with 0.45 Whkg −1 in SSPAS. The time spent for one cycle is 10 s, which is shortest in the present study. A similar deviation from the linearity in the discharge curve was observed in the case of Al(ClO 4 ) 3 . Since the standard redox potentials for Mg 2+ /Mg and Al 3+ /Al are − 1.66 V and − 2.36 V, respectively, the partial reductions of Mg 2+ and Al 3+ to metals could not be ruled out under the present conditions even though in consideration of strong hydrations toward Mg 2+ and Al 3+ . Recently, an aqueous electrolyte of using an eutectic hydrate melt which consists of a mixture of two organic lithium salts, that is Li(TFSI) 0.7 (BETI) 0.3 •2H 2 O, where TFSI is (bis(trifluoromethylsurphonyl)imide and BETI bis(pentauluoroethylsulphonyl)imide was reported for the use of Li ion battery 32 . The paper represents that this hydrate melt has a potential window over 3 V exhibiting an excellent capability for the Li ion battery. At moment, it is difficult to compare the difference in the above hydrate melt and saturated aqueous solutions of perchlorate salts as the electrolytes. A need exists to use batteries or capacitors under extremely low temperatures. The eutectic temperature of SSPAS is − 37 °C and that of SMPAS is even lower at − 67 °C13 . Considering these data, the capacitors of using SSPAS or SMPAS would be functionally operative at very low temperatures. ## Conclusion Finally we conclude the present study. The most important finding was to successfully expand the potential window of the aqueous electrolyte first over 3 V by the use of the saturated sodium perchlorate aqueous solution (SSPAS). The electrolyte was demonstrated to behave well for graphite-based capacitors with respect to the stability for charge-discharge repetitions and the enlarged energy densities by the addition of activated carbon and metal oxides. It was also found that the other perchlorate electrolytes, particularly, the saturated Mg(ClO 4 ) 2 aqueous solution (SMPAS) is very feasible for the superior electrolyte. On the basis of the present results, together with the safety and economy impacts, SSPAS or SMPAS could replace non-aqueous electrolytes in commercial capacitors in the near future. ## Methods A graphite capacitor was produced by the following procedure. A mixtures of carbon powder, i.e., graphite (J-SP-α of Nippon Graphite Industries Ltd.) 80%, acetylene black (Denka) 10%, carbon felt (TOYOBO) 10%, was pressed at 1 kgcm −2 by a hydraulic machinery to make a thin film. Thus made films were wetted by SSPAS and used for both negative and positive electrodes. These were assembled with a separator to make a capacitor using a Hohsen battery unit, where glassy carbons (Tokai carbon) are used in both current collectors. As separators, non-conductive water permeable sheets such as a cation exchange membrane (NEOSEPTA CIMS), a filter paper (ADVANTEC 5B) and a membrane filter (Millipore JVWP) and a PPS fiber (TORAY Torcon) were able to be used. Galvanostatic measurements were carried out by using an instrument of Bio-Logic VSP at temperature 30 °C. A Rigaku MiniFlexII was used for XRD measurements and a JNM-ECX400 P for NMR measurements. Safety test. It has been known that dried sodium perchlorate has potential danger of explosion in the presence of organic compounds. Therefore, we made a safety test as follows: the mixture of sodium perchlorate and graphite containing activated carbon or vanadium pentoxide was heated at 200 °C and examined the change of the mixture in increasing temperature. As a result, nothing was happened by the heating and we confirmed the safety of the capacitors.

chemsum

{"title": "An aqueous electrolyte of the widest potential window and its superior capability for capacitors", "journal": "Scientific Reports - Nature"}

patterned_dried_blood_spot_cards_for_improved_sampling_of_whole_blood

## Abstract: Dried blood spot (DBS) cards perform many functions for sampling blood that is intended for subsequent laboratory analysis, which include: (i) obviating the need for a phlebotomist by using fingersticks, (ii) enhancing the stability of analytes at ambient or elevated environmental conditions, and (iii) simplifying transportation of samples without a cold chain. However, a significant drawback of standard DBS cards is the potential for sampling bias due to unrestricted filling caused by the hematocrit of blood, which often limits quantitative or reproducible measurements. Alternative microsampling technologies have minimized or eliminated this bias by restricting blood distribution, but these approaches deviate from clinical protocols and present a barrier to broad adoption. Herein, we describe a patterned dried blood spot (pDBS) card that uses wax barriers to control the flow and distribution of blood and provide enhanced sampling by minimizing the hematocrit effect. Patterned cards reproducibly fill four replicate extraction zones independent of the hematocrit. We demonstrate a 3-fold improvement in accuracy for the quantitation of hemoglobin using pDBS cards compared to unpatterned cards. Patterned cards also facilitate the near quantitative recovery (ca. 95%) of sodium with no evidence of a statistically significant difference between dried and liquid blood samples. Similarly, recovery of select amino acids was conserved in comparison to a recent report with improved inter-card precision. We anticipate that this approach presents a viable method for preparing and storing samples of blood in limited resource settings while maintaining current clinical protocols for processing and analyzing dried blood spots. ## Introduction Blood is a complex matrix, comprising cellular and liquid fractions, that contains a wealth of diagnostically relevant biomarkers, which are inclusive of the cells themselves (e.g., neutrophil count), DNA/RNA (e.g., endogenous or from pathogens), and myriad solutes in plasma (e.g., proteins, metabolites, free amino acids). For these reasons, blood is often thought of as the ideal specimen for evaluating the health status of a patient. Obtaining liquid blood samples in centralized facilities or even local clinics is routine practice. In these settings, a trained phlebotomist will collect milliliter volumes of blood by venipuncture, which can either be immediately processed and tested in the laboratory or stored for future analysis within a defined period of time dependent on storage temperature. 1,2 However, these same practices face unique challenges at the point-of-care or in resource-limited settings. Specifically, storage and transportation of liquid blood are complicated by unreliable modes of transportation and inadequate access to cold-chain storage. These limitations often require liquid samples to be discarded due to substantial degradation or significant changes to critical hematological indices. In contrast to liquid samples, storing blood in a porous matrix, such as chromatography paper, enhances analyte stability at ambient or even elevated temperatures. 3,4 Dried blood spot (DBS) cards additionally offer simplified sampling using fingersticks and reliable transportation of dried blood by mail, thus circumventing the need for cold chain storage. 5,6 Traditional DBS cards, such as the Whatman 903 Protein Saver card, are a simple construction of a single sheet of thick cellulose cardstock affixed to an envelope for sample identification and handling. 7 Circles are printed onto the surface of the paper using a thin layer of toner to provide guidance for sample application. Fingerstick volumes of blood (e.g., 50-100 µL per spot) are applied to the card and allowed to dry for a minimum of four hours at ambient conditions (and ideally overnight), rendering the card non-biohazardous, before they are sealed and shipped through the mail for laboratory analysis. 8 Self-sampling low volumes of blood without the need for cold chain storage could broadly expand access to basic health information by providing direct-to-consumer testing, facilitate critical population screening, and biobanking efforts. 9 Although traditional DBS cards offer simple operation, require low volumes of blood, and can be collected outside of the clinic, they are severely limited by usability associated with unrestricted sample application zones. This user error can result in non-uniform or smeared blood spots, which will ultimately impact the quality of subsequent laboratory analysis and represents a considerable barrier for ubiquitous use of traditional DBS cards. 9,10 Beyond usability, traditional DBS cards do not account for differences in hematocrit values (Hct)-the ratio of packed red blood cells (RBCs) to total blood sample volume. The normal range of hematocrit spans 36-50% and is affected by variables such as race, sex, age, hydration, and overall health status. 11 Currently, the hematocrit value must be known prior to analysis for accurate quantitation of analytes using DBS cards. Whether caused by filling imprecision or hematocrit, the uniformity of how cells and liquid plasma are distributed throughout the paper cardstock has a substantial impact on the overall utility of a DBS card. The hematocrit effect has been extensively reviewed as the main obstacle to overcome for quantitative analysis using traditional DBS cards. Since the hematocrit value represents the ratio of cellular matter to liquid plasma, blood samples with a high hematocrit value (e.g., 55%) will be more viscous than samples with a low hematocrit value (e.g., 30%). Variation in viscosity results in variable sample flow and distribution through the paper, which negatively impacts the reproducibility of sample volumes obtained from a single, fixed punch extraction from DBS. Uncontrolled saturation or spreading of blood through the DBS paper can also result in heterogeneous distribution of analytes throughout the area of the resulting DBS (i.e., volcano effect). 19 Because analytes are typically eluted from DBS via a fixed punch, any variation in sample volume and distribution will manifest in downstream clinical measurements causing a lack of precision (i.e., intra-spot agreement) or accuracy (i.e., agreement with liquid sample). Many methods for minimizing the hematocrit effect in traditional DBS cards have previously been reported. Two distinct approaches stand out: (i) whole spot analysis 23,24 and (ii) assay specific calibrants stored within the paper. 25 Both present viable options for minimizing the hematocrit effect by quantitative removal of the entire blood spot (dependent on application of accurate sample volume) or inclusion of an internal standard at a known concentration to estimate extraction efficiency. However, both methods are limited by the number of tests that can be conducted from a single DBS spot. In each format, samples can only be used to perform a single test due to the complete destruction of the entire dried spot or analyte-specific internal standard. Alternatively, three-dimensional blood spheroids eliminate chromatographic effects observed in traditional DBS and reduce the volume of blood required per spot by utilizing functionalized hydrophobic paper. 24 This approach has successfully demonstrated increased stability of enzymes and labile organic compounds. Recently, DBS technologies that operate independent of the hematocrit by constricting sample volume have also been described. The ADX Test Card by Accel Diagnostics utilizes a microfluidic network and magnetic beads to collect, distribute, and analyze blood. 26 The HemaSpot HF comprises pre-cut paper wedges contained within a plastic housing, which hold a finite volume of sample. 27 Similarly, the HemaPEN 28 and Capitainer 29 integrate multiple fixed-volume capillary tubes to standardize the volume of blood applied to a porous matrix. While these devices provide enhanced control over application of sample volume, they do not conform with current clinical collection or automated punching and elution protocols. In order to improve the utility of DBS cards with an intent for widespread use, current clinical protocols for sample collection and subsequent analysis should be maintained. Therefore, innovation should build upon the major benefits of traditional DBS technology (i.e., single layer of cardstock). An attractive approach for enhanced sampling is controlling the flow of blood samples in the cardstock with hydrophobic wax barriers. 30 Defining specific areas for (i) sample addition, (ii) distribution, and (iii) storage by wax patterning presents a method for addressing the limitations of current DBS technologies without creating additional clinical barriers. Ideally, blood sampling would be performed via a self-administered fingerstick, simple collection onto a solid matrix, drying, and delivery to a laboratory for testing without significant degradation of the sample at ambient conditions. Herein, we describe the creation of patterned dried blood spot (pDBS) cards to address the limitations of traditional DBS cards directly related to the hematocrit effect. Patterning traditional DBS cardstock with hydrophobic wax barriers regulates sample application, distribution, and volume control while operating independently of the hematocrit over a broad range of clinical values (20-60%). A user simply needs to apply a volume of blood to the center of the card and the sample will automatically distribute to four replicate punch zones. Providing more spots for analysis while also maintaining reproducible spreading across physiological hematocrit values can (i) increase the number of technical replicates or (ii) increase the number of clinical assays performed from one sample of whole blood without concern for significant punch-to-punch variation. We first investigated the capacity of pDBS cards for quantitative sampling by estimating the volume of blood contained in a standard 6-mm paper punch and reported minimal variation even when the sample input deviates from the World Health Organization (WHO) recommended volume of 75 µL. 5 Next, we demonstrated enhanced usability and spot uniformity independent of the hematocrit for samples collected with pDBS cards compared to traditional, unpatterned cardstock. We highlighted a broad class of analytes to showcase this approach including the quantitation of hemoglobin by UV-vis spectrophotometry, sodium by inductively coupled plasma atomic emission spectroscopy (ICP-AES), and specific amino acids by high-performance liquid chromatography (HPLC). pDBS cards permit enhanced sampling of small volumes of blood that can be generated from a fingerstick and represent a reproducible method capable of performing multiple tests without requiring multiple sample collections or altering established laboratory workflows. We anticipate the quantitative nature of this self-sampling method of blood collection will empower patients by providing critical, accurate diagnostic information at home or in lowincome economies without impacting existing clinical procedures. ## Experimental Design Card Design and Fabrication pDBS cards comprise a single layer of cardstock impregnated with wax to form three distinct features: (i) sample addition zone, (ii) lateral distribution channels, and (iii) four replicate, collection punch zones (Figure 1A). We designed our cards to accommodate a sample input volume of 75 µL and output punch diameter of 6-mm in accordance with the WHO recommended specifications for DBS sampling. The design features (e.g., lateral channels) and geometries were informed by our previous experience with whole blood in paper for measuring the hematocrit. 31,32 Whole blood is transported from the sample addition zone along the lateral channels via capillary action and fills four replicate collection punch zones at the end of each channel. Extending the lateral channels past the collection punch zones allowed complete saturation of the punch zone for more accurate sampling compared to traditional DBS cards. Wax printing is typically performed by direct deposition of wax onto relatively thin (≤ 250 µm), smooth papers followed by application of heat to allow the wax to coat the paper fibers. 33 For papers > 250 µm thick, standard printing practices cannot deposit sufficient wax to form complete hydrophobic barriers (Figure S1A). 34 Incomplete barriers resulted in uncontrolled sample flow and represent a challenge for patterning DBS papers. Alternative methods for patterning thick materials with photoresist or paraffin have been reported previously. 35 However, to maintain the numerous benefits of wax printing, we utilized a double-sided wax transfer method 36 to successfully pattern papers commonly used for traditional DBS cards (e.g., Whatman CF-12, Ahlstrom 226, Munktell TFN) (Figure S1B). First, we printed the top and bottom designs onto laminate sheets using a Xerox ColorQube 8580 wax printer. Next, we aligned a sheet of chromatography paper with the top and bottom designs using a custom acrylic alignment jig. Finally, we used a Promo Heat CS-15 T-shirt press (45 seconds at 280 °C) to transfer the wax from the laminate sheets to the paper to form hydrophobic barriers through the full thickness of the paper. Patterning each side with a unique design allowed partial coating of the cellulose fibers through approximately half the thickness of the paper to reduce the void volume of the sample addition and lateral distribution channels in pDBS cards (Figure 1B). This process provided an added benefit of minimizing sample input volume while maximizing sample collection volume from the punch zones. After addition of whole blood, we dried pDBS cards under ambient conditions in a biosafety cabinet (ca. 16 hours), whereby they can be used immediately or sealed in a foil pouch with silica desiccant packets and a humidity indicator card for long-term storage. All data presented herein were collected using pDBS cards fabricated from TFN grade cardstock. We chose to demonstrate the utility of our cards for sampling a range of analytes (e.g., hemoglobin, sodium, and select amino acids) and technique groups (e.g., UV-vis spectrophotometry, ICP-AES, and HPLC). ## Effects of Evaporation on the Quantitation of Hemoglobin Evaporation at ambient conditions is the driving force for drying samples of blood in DBS cards. Sealing-or partially sealing-sections of our pDBS cards influenced the location and extent of evaporation. Additionally, altering the bottom design of the pDBS card can affect evaporation by controlling the amount of unpatterned card area that is exposed to the environment. We iteratively added or removed a layer of laminate to the top and bottom sides of the pDBS card and evaluated the effects of evaporation on the quantitation of hemoglobin using a modification of the standard Drabkin's assay (Figure S2). The bottom design either (i) excluded (designs A and B) or (ii) included (designs C and D) the lateral distribution channels. Evaluating these design features across a range of hematocrit is critical for understanding the effects of evaporation since these samples have varying volumes of liquid plasma (e.g., 52.5 µL of plasma in 75 µL of 30% hematocrit blood vs. 37.5 µL of plasma in 75 µL of 50% hematocrit blood). Excluding the lateral channels and sample addition zone in the bottom design reduced the total void volume of the unpatterned area and eliminated evaporation from the bottom side of the lateral channels and sample addition zone. Reducing the void volume improved reproducibility for card filling. Further covering the lateral channels on the top side of the pDBS card minimized evaporation along the channel and effectively concentrated the blood sample in the collection punch zones. Preconcentration of blood in the collection punch zones resulted in higher percent deviation for the quantitation of hemoglobin (Table S1), which we expect is due to the volume dependency of the Drabkin's assay. 37 Both designs B and C had comparable performance even though design B included no laminate covering the unpatterned area and design C was completely laminated (except at the sample addition zone). We chose to move forward with design B for two reasons: (i) it yielded the lowest percent error for both 30% and 50% hematocrit samples and (ii) reduced the number of laminate layers necessary which simplified the manufacturing and operational processes. ## Estimation of Sample Volume in 6-mm Paper Punch After finalizing the form factor of our pDBS card and minimizing the effect of evaporation through unique bottom patterning, we measured the volume of a dried sample contained in an individual 6-mm paper punch in order to correlate the concentration of an analyte to the total sample of blood. Accurate comparison of liquid reference samples to our pDBS card is dependent on the sample volume contained within a punch. This type of measurement has been accomplished using a variety of methods including ion suppression by liquid chromatographytandem mass spectrometry 20 and electrical conductivity of DBS extract by a ring disk electrode. 10 We utilized the volume dependency of the Drabkin's assay to estimate the output sample volume in our pDBS card. 38 First, we constructed a series of calibration curves (Figure S3A) using liquid hemoglobin standards with varied sample input volumes (3-11 µL) to establish a relationship between linear slope of the calibration curve and sample volume (Figure S3B). Then, we 10 calibrated our pDBS cards with hemoglobin standards and estimated the sample volume contained in a 6-mm paper punch using the resultant linear relationship and slope of the calibration curve in our pDBS card (Figure S3C). All hemoglobin samples reproducibly filled the pDBS cards (Figure S3D). We estimated that each 6-mm paper punch contained 10.3 ± 0.4 µL of whole blood, representing a total output sample volume of approximately 41.2 µL from an input volume of 75 µL blood. The low variation (< 5%) observed in the sample volume contained in a paper punch indicated consistent sample distribution in pDBS cards. Deviating from the recommended sample input volume of 75 µL can negatively impact the quantitation of analytes such as hemoglobin. To simulate under-and overfilling, we applied a range of sample volumes 60-90 µL in 5 µL increments at a single hematocrit (Figure S4A). Our pDBS cards reproducibly filled four replicate punch zones with a sample volume ≥ 65 µL (Figure S4B). The average deviation for replicate cards with sample input varying ± 15 µL was only 12.0% compared to the liquid reference sample. This result provided confidence that slight variations in the sample input volume (e.g., from direct addition of a fingerstick rather than sample addition by volumetric pipette) will not substantially impact quantitative results if volumetric sample application is unavailable at the site of collection. ## pDBS Cards Fill Independent of Hematocrit Value We aimed to further evaluate the effect of sample input on quantitation of hemoglobin by surveying the physiological range of hematocrit values (20-60%). We anticipated that controlling the total area of the pDBS card through patterning would minimize the negative effects of variable sample spreading caused by the hematocrit. Direct comparison of pDBS cards and unpatterned TFN clearly demonstrated how the hematocrit influenced the results of standard assays such as the quantitation of hemoglobin (Table 1). Patterned cards yielded ≤ 7% error across the full range of hematocrit values, while unpatterned cards yielded 3-fold higher percent error at low hematocrit (21% error at 20% hematocrit, Table 1). Inter-and intra-card variation (i.e., spot-to-spot variation) were consistent between both card types (Table S2), which suggested the deviation in the quantitation of hemoglobin can largely be attributed to uncontrolled sample spreading of blood in unpatterned cards. Agreement between pDBS cards (Figure 2A) and unpatterned TFN (Figure 2B) with the reference liquid blood is represented by Bland-Altman plots. 39 The observed bias was reduced in pDBS cards (-0.7 g/dL) compared to TFN (-1.0 g/dL). Similarly, the limit of agreement was narrower for pDBS (2.2 g/dL) than TFN (3.0 g/dL) in comparison to the reference method. Patterned cards reproducibly filled four replicate collection punch zones (6-mm diameter) across the full range of hematocrit values (Figure 3A). Since pDBS cards filled independently of the hematocrit, four replicate punches can always be collected for analysis and enable more tests to be performed from a single card. In stark contrast, the diameter of the blood spot in unpatterned TFN decreased with increasing hematocrit (20-60% hematocrit) (Figure 3B). A direct consequence of the decreased blood spot diameter in unpatterned DBS is one less technical replicate punch of dried blood under idealized conditions (Figure 3C). ## Quantitation of Sodium by ICP-AES Blood sodium levels are routinely measured as part of a basic metabolic panel that often includes additional electrolytes such as calcium, chloride, and potassium. Accurate quantitation of sodium is critical for controlling blood pressure and evaluating proper nerve and muscle function. 40 Additionally, because sodium is found both intra-and extracellularly, it represented an attractive analyte class to further evaluate the quantitative capabilities of pDBS cards. The concentration of sodium in blood samples obtained from pDBS cards (1715 ± 21 ppm) was nearly identical to the concentration in the reference liquid sample (1810 ± 24 ppm), suggesting that there is no apparent loss or evaporative concentration of sodium to the TFN paper (Figure 4). A two-tailed Student's t-test yielded a p-value of 0.26, providing no evidence of a statistically significant difference in sodium concentration between the dried and liquid blood samples. The clinical reference range for sodium in blood is 135-145 mEq/L. 41 Both the dried and liquid blood samples fell below the expected range with 74.6 and 78.7 mEq/L sodium, respectively. Both samples were prepared using nitric acid digestion, which included multiple liquid handling and quantitative transfer steps, which could account for the low observed concentrations. While the range of concentrations of sodium extracted from pDBS punches (683.9 ppm) was more dispersed than those from liquid samples of blood (392.4 ppm), the standard deviation was slightly less. Comparison of variances (F-test) yielded a p-value of 0.27, indicating no significant difference between the variance of the data sets. Therefore, the precision of pDBS card microsampling could be amenable to use of calibration standards for quantitative results. ## Quantitation of Amino Acids by HPLC Amino acid analysis via DBS sampling is commonly used for the detection of various inborn errors of amino acid metabolism including phenylketonuria (PKU) in newborns. Efforts to streamline and improve the quantitation of amino acids from DBS have been extensively reported. For demonstrative purposes, we selected three representative hydrophobic amino acids (e.g., tryptophan, leucine, and proline) and one basic-or positively charged-amino acid (e.g., lysine) for analysis. Recovery of each amino acid from pDBS cards was determined by the ratio of extracted analyte concentration (µM) and liquid reference concentration (µM) as analyzed by HPLC. Two distinct sample groups (e.g., 20% and 40% hematocrit) were selected to represent (i) a high liquid-to-cell ratio-which can be prone to underestimating analytes of interest-and (ii) the average hematocrit obtained from our panel of healthy donors, respectively. Each amino acid yielded excellent recovery for both blood sample groups (Table 2). While most samples fell in the range of 82-93% recovery, two samples yielded higher concentrations when extracted from pDBS cards compared to liquid reference samples (proline 115%, lysine 102%). Resultant loss and variability in analyte recovery may be attributed to the number of liquid handing steps required to extract, process, and derivatize samples prior to analysis by HPLC. However, all reported values in Table 2 are in agreement with other reports where recovery of amino acids ranged from 84.2 ± 22.2-96.0 ± 12.0%. 45 Additionally, the evaluation of interassay precision (i.e., card-to-card 13 comparison) demonstrated a coefficient of variation (%CV) for tryptophan of 0.8-5.2%, leucine 2.6-6.7%, proline 1.0-5.5%, and lysine 5.4-5.7% (Table S3). These %CV values are considerably improved in comparison to recently reported %CV for amino acid analysis by traditional DBS sampling using similar methods (e.g., %CV for leucine 8.3-15.3%). 44 Successful quantitation and improved interassay precision of select amino acids by HPLC supported the enhanced sampling capabilities of pDBS cards. ## Conclusions We aimed to develop a device that can improve the sampling of whole blood at the pointof-care while maintaining current clinical protocols for DBS analysis. Our approach comprised wax-patterned DBS cardstock to restrict the flow and distribution of whole blood with four defined extraction zones. Controlling the flow of blood in the pDBS card allowed reproducible filling across the full range of hematocrit values and reduced the sampling bias for pDBS cards compared to unpatterned TFN cardstock. Specifically, the accuracy for the quantitation of hemoglobin with low hematocrit (20%) was improved by 3-fold using pDBS cards. Sampling was further improved by spatially defining extraction zones, which consistently produced four replicate 6-mm diameter punches from a single application of blood (75 µL), independent of the hematocrit value. We designed these cards to accommodate direct application of fingerstick volumes of blood and modeled ideal conditions by dispensing blood using a volumetric pipette. The highly controlled nature of this method of sample dispensing may be reflective of the conserved inter-and intracard variations reported for both pDBS and traditional DBS cards. We anticipate that the patterned features of pDBS cards will maintain uniform filling and address the reported challenges associated with applying fingersticks to DBS at the point-of-care. Surveying common DBS analytical techniques such as ICP-AES and HPLC indicated good agreement with liquid reference samples for the quantitation of sodium and select amino acids, respectively. Additionally, we were able to process and analyze samples of whole blood without changing recommended handling procedures for DBS cards (i.e., amenable to automated punching machines). Standardizing the sample output from pDBS cards could expand the number of tests performed from a single sample collection or permit increased numbers of technical replicates compared to traditional unpatterned DBS cards. Beyond the classes of analytes and techniques demonstrated in this manuscript, quantitative DBS sampling has the potential for myriad applications related to molecular amplification (e.g., screening for viral diseases), nutritional evaluations, immunologic studies, pharmacokinetics, therapeutic drug monitoring, and genetic testing. 46 Since pDBS cards are exposed to ambient conditions during sample application, spreading, and drying, we expect performance may vary under certain environmental conditions at the time of collection, as similarly experienced with traditional DBS cards. For example, sample spreading may be reduced due to extremely dry conditions (relative humidity below 10%) or high temperatures, which could cause excessive evaporation. However, this effect is commonplace for DBS technologies and is not identified as a major obstacle for ubiquitous use. 6 While the pDBS card presented here was used for sampling whole blood, we anticipate that we could expand on this approach to collect and store additional sample types such as saliva, tears, or blood plasma to provide enhanced sampling and quantitative analysis in a workflow that connects the point-ofcare to a clinical laboratory infrastructure. ## Conflicts of Interest The authors declare no conflicts of interest.

chemsum

{"title": "Patterned Dried Blood Spot Cards for Improved Sampling of Whole Blood", "journal": "ChemRxiv"}

an_enzyme-free_molecular_catalytic_device:_dynamically_self-assembled_dna_dendrimers_for_<i>in_situ<

## Abstract: DNA has become a promising material to construct high-order structures and molecular devices owing to its sequence programmability. Herein, a DNA machine based on branched catalytic hairpin assembly (bCHA) is introduced for dynamic self-assembly of DNA dendrimers. For this system, a Y-shaped hairpin trimer tethered with three kinds of hairpins (H1, H2 and H3) is constructed. The introduction of an initiator (I) triggers a cascade of CHA reactions among hairpin trimers, leading to the formation of DNA dendrimers.Through labeling fluorophore/quencher pairs in the hairpin trimers, this catalytic DNA machine is applied as a versatile amplification platform to analyze nucleic acids using microRNA-155 (miR-155) as a model analyte. Benefiting from the "diffusion effect", the proposed bCHA achieves a greatly improved sensitivity in comparison with traditional CHA. This catalytic amplifier exhibits high sensitivity toward miR-155 detection with a dynamic range from 2.5 nM to 500 nM and demonstrates excellent selectivity to distinguish the single-base mismatched sequence from the perfectly complementary one, which is further applied to detect low-abundance miR-155 spiked in complex matrices with minimal interference. This method is further applied for in situ imaging of miR-155 in different live cells. The bCHA reaction can be specifically triggered by intracellular miR-155, achieving monitoring of the dynamic miRNA expression and distribution. Overall, our proposed enzyme-free dynamic DNA self-assembly strategy provides a versatile approach for the development of DNA nanotechnology in biosensing and bioimaging, and monitoring the cellular miRNA-related biological events. ## Introduction Deoxyribonucleic acids (DNAs) have been considered as promising building blocks to fabricate a variety of nanostructures and devices owing to their sequence programmability and specifc recognition properties. 1 The applications of DNAs as engineering materials provide a solid foundation for the development of DNA nanotechnology and materials science. 2 DNA nanotechnology can be mainly divided into two categories: structural DNA nanotechnology and dynamic DNA nanotechnology, in which DNA strands are employed to program the spatial and temporal distribution of matter. 3 Based on 'bottom-up' engineering approaches, structural DNA nanotechnology has realized the fabrication of 2D and 3D DNA assemblies with various sizes and spatial structures, such as DNA lattices, 4,5 DNA origami, 6 DNA tetrahedron structures, DNA nanotubes 10,11 and so on. Unlike structural DNA nanotechnology, dynamic DNA nanotechnology lays emphasis on the non-equilibrium dynamics, in which the formation of DNA nanostructures results from successive dynamic assembly of DNA motifs. 3 That is, the nanostructures formed by dynamic DNA nanotechnology can present mechanical operation processes. To date, a variety of dynamic DNA devices have been constructed, which are propelled by DNAzymes, 12 specifc DNA-modifying enzymes 13,14 and toehold-mediated strand displacement (TMSD) reactions. The TMSD reaction, a concept pioneered by Yurke et al., occurs when an invading strand displaces a target strand on a double-stranded complex with the help of a single-stranded sequence (termed a toehold). 18 By means of the toehold sequence, the DNA invasion reaction can be accelerated and realize the kinetically controlled self-assembly of DNA. 3,19 Thus, the TMSD reaction has been used as powerful tool to program enzyme-free DNA circuits and DNA nanomachines. 20,21 To date, synthetic molecular machines have spurred substantial research efforts in the felds of biosensing for detecting a variety of targets such as miRNA, proteins and so on, and bioimaging for accurate miRNA imaging in living cells. 34,35 Catalytic hairpin assembly (CHA) can be considered as one of the most prominent TMSD reactions. In CHA, DNA hairpins are kinetically trapped and exist metastably. Upon the introduction of an initiator, successive assembly steps are triggered among DNA hairpins to form branched junctions via cascades of strand displacement reactions, accompanied by a disassembly step in which the initiator is displaced to catalyze another CHA process. 17 Since one initiator can be continuously replaced and recycled to trigger a new reaction, CHA has been applied as a powerful enzyme-free signal transducer for isothermal amplifcation analysis of a wide range of targets from nucleic acids, small molecules to proteins and even cancer cells. Very recently, CHA has been further applied to construct diverse DNA structures and molecular machines. 27 MicroRNAs (miRNAs) are a group of small noncoding RNA molecules that play important roles in a series of cellular processes such as cell differentiation, proliferation, apoptosis and so on. More and more research has demonstrated that the cellular dysregulated expression of miRNAs is related to the genesis of many cancers. Therefore, it is necessary to sensitively detect miRNAs for early diagnosis of cancers and cellular level research. Currently, benefting from the advantages of CHA, some CHA-based detection methods have been established for quantitative detection and analysis of miRNA expression. However, effective signal amplifcation of miR-NAs is still greatly needed due to their low expression level in cells. Herein, we construct an enzyme-free catalytic device based on bCHA for dynamic self-assembly of DNA dendrimers. This strategy involves the formation of a Y-shaped hairpin trimer which contains a Y-scaffold and three kinds of hairpins. The assembly into explicit DNA dendrimers is initiated and mediated via a series of CHA processes among the hairpin trimers. By using the target-triggered mechanism, we demonstrate the versatile applications of the proposed bCHA for signal ampli-fcation biosensing and in situ imaging of miR-155 in live cells with high sensitivity and selectivity. For all we know, this is the frst time an intracellular CHA process is reported for in situ monitoring the dynamic expression and distribution of miRNA in live cells. Compared with other molecular devices, especially traditional CHA, the major advantage of our proposed bCHA is the signifcantly improved amplifcation efficiency which benefted from the "diffusion effect". In bCHA, three hairpins are localized on a Y-scaffold DNA, which thus provides an addressable substrate for those released initiators, achieving greater control for cascade hairpin assembly. ## Principle of bCHA The proposed bCHA strategy consists of a Y-scaffold DNA and three kinds of hairpins (HP1, HP2 and HP3). The sequences and the secondary structures of the oligonucleotides used in this work are listed in Table S1 and Fig. S1, † respectively. The assembly pathways are shown in Fig. 1a. The Y-scaffold DNA is composed of three single-stranded DNAs (ssDNAs), Y1, Y2 and Y3, in which each strand contains three segments: a long "sticky end" to interact with its complementary sequence on the corresponding hairpin and two parts to hybridize with the other two strands. On the basis of Watson-Crick base pairs, the Yshaped hairpin trimer is formed via the hybridization of the "sticky ends" between the Y-scaffold DNA and hairpins (Fig. 1a, reaction 1). The formed hairpin trimer can metastably coexist in solution, and acts as a building block to self-assemble DNA dendrimers upon addition of the initiator to activate the dynamic self-assembly process of bCHA. During the bCHA, the initiator hybridizes and docks to the exposed toehold domain a of H1 and subsequently displaces the parts of x, b, and y from the duplex of the stem in H1 via a TMSD reaction, opening the loop of H1 (Fig. 1a, reaction 2). The opened loop makes the branch migration irreversible. Meanwhile, the newly exposed domain b* of H1 hybridizes with domain b of H2, which is the component of a proximal hairpin trimer in solution, to operate the branch migration, resulting in the opening of the loop of H2 and the formation of the hybrid of two three-arm branched hairpin structures (Fig. 1a, reaction 3). Another proximal hairpin trimer then invades the hybrid because the toehold domain c of H3 can dock to the newly exposed domain c* of the hybrid. And then a disassembly step is performed, in which H3 displaces I from the complex to catalyze another dynamic selfassembly process (Fig. 1a, reaction 4). One CHA reaction is completed as above. Theoretically, this branching chain growth would continue to form DNA dendrimers after a cascade of CHA processes among the hairpin trimers (Fig. 1a, reaction 5), which is thus termed branched catalytic hairpin assembly (bCHA). The morphology of the DNA dendrimers assembled via bCHA was characterized by microscope imaging (Fig. 1b). In the absence of an initiator, no assembly event of hairpin trimers occurs. In contrast, a large number of two-dimensional hyperbranched DNA structures appear upon adding the initiator into the hairpin trimers. The size distribution of the hairpin trimers before and after being incubated with the initiator was further measured by DLS (Fig. 1c). The size of hairpin trimers is determined to be 45 AE 15 nm, whereas, it signifcantly increases to 4.5 AE 2.5 mm upon addition of the initiator, which are consistent with the observation in microscopy images. Therefore, the above results fully demonstrate the successful dynamic self-assembly of DNA dendrimers via bCHA as anticipated. Furthermore, the zeta potential of the DNA dendrimers was measured to be 8.35 AE 3.44 mV, which can be attributed to the negatively charged phosphate backbones of the packed DNA strands constituting the DNA dendrimers (Fig. S2 †). ## Native polyacrylamide gel electrophoresis (PAGE) As a proof-of-concept experiment, the assembly of Y-scaffold DNA was verifed by native PAGE (Fig. 2a). The bands in lanes 1, 2 and 3 correspond to Y1, Y2 and Y3, respectively, which migrate faster than the assembled Y-scaffold DNA (lane 7). The bands representing Y1 + Y2, Y1 + Y3, and Y2 + Y3 (lanes 4, 5, and 6) are in between. These results indicate the effectivity of the assembly process, and Y-scaffold DNA has been formed as designed. Subsequently, we compared the assembly pathways of traditional CHA and the proposed bCHA by native PAGE. For the traditional CHA, three-arm DNA junctions are formed through self-assembly of H1, H2 and H3 triggered by the initiator, while in bCHA the DNA dendrimers are generated using hairpin trimers as building blocks. As shown in Fig. 2b, lanes 10-12 are H1, H2 and H3, respectively, while lane 9 corresponds to the mixture of H1, H2 and H3 without an initiator. These bands are almost at the same migration rate. In the presence of the initiator, three-arm DNA junctions are formed (lane 8), which migrate slower than hairpin species. The formation of Yscaffold DNA in lane 13 is almost the same as that shown in Fig. 2a. When the Y-scaffold DNA is mixed with H1, H2 and H3 at a ratio of 1 : 1 : 1 : 1, the hybrid of the Y-shaped hairpin trimer is formed in lane 14. With the addition of the initiator, the band corresponding to the hairpin trimer disappears and the assembly products with much lower mobility are observed in lane 15, demonstrating the formation of DNA dendrimers with high molecular weight via the bCHA. ## Amplication biosensing of microRNA Since the initiator can be displaced to continuously trigger a new cycle of assembly, we further employed the bCHA strategy as a catalytic amplifer for sensitive and selective detection of microRNA. It has been reported that miR-155 plays an important role in various physiological and pathological processes, and can be considered as an important biomarker for diagnosis, staging, progression, and prognosis of cancers. 40,41 To demonstrate the feasibility of the bCHA for amplifed detection of miR-155, H1 is modifed with a fluorophore (FAM) at the 3 0 terminus and a quencher (Dabcyl) at the appropriate position (Fig. 3a). In the absence of miR-155, the efficient quenching effect resulted from the fact that the closely positioned Dabcyl quencher makes the fluorescent emission from the FAM minimal. In contrast, the introduction of the miR-155 initiates dynamic self-assembly among hairpin trimers, resulting in the opening of H1 to remove the quencher from the fluorophore. Thus, the signifcantly amplifed fluorescent signal output can be monitored for highly sensitive detection of miR-155. The real-time fluorescence triggered by different concentrations of miR-155 was measured to investigate the growth kinetics of the proposed bCHA (Fig. 3b). At the beginning of the reaction, the annealed hairpins on trimers are in a closed structure so that the fluorophore and the quencher are brought into close proximity, resulting in an "off" fluorescence signal. This phenomenon is verifed by monitoring the fluorescence intensity of the control sample consisting of only H1. For the blank sample without miR-155, the fluorescence intensity should be almost the same as the control since no DNA dendrimer is generated and the fluorophore separate from the quencher. However, a slightly increased fluorescence is observed, which can be attributed to the system leakage induced by few imperfectly annealed hairpin species. Thankfully, the fluorescence intensity of the blank sample remains at a low level which has a negligible influence on the reaction, demonstrating that the Y-shaped hairpin trimers can metastably coexist in the absence of an initiator. Upon the introduction of target miR-155, the self-assembly process of bCHA is activated, resulting in the formation of hyperbranched DNA structures. In this case, the fluorescence of H1 is recovered with an enlarged spatial distance between the fluorophore and the quencher, resulting in the "fluorescence-on" mode. The increased fluorescence intensity is in direct proportion to the concentration of miR-155. In particular, the emission intensity rapidly reaches a plateau when the concentrations of miR-155 are at 250 nM and 500 nM, indicating a large proportion of reactants can be depleted when the initiator is at high concentrations. The bCHA protocol has revealed high amplifcation efficiency for quantitative analysis of miR-155. As low as 2.5 nM miR-155 can be well distinguished from the blank. This sensitivity is satisfactory and comparable with other molecular devices. 43,44 To further assess the quantitative behavior of the proposed method, we choose 20 000 a.u. as the threshold, at which all samples can be taken into consideration. A log-linear trend between the initial concentration of miR-155 and the time to reach a fluorescence threshold of 20 000 a.u. is shown as Fig. 3c. This relationship is presented as log 10 [time] ¼ 7.63 1.1 log 10 [C] with a correlation coefficient (R 2 ) of 0.99. Furthermore, fve parallel measurements are performed to study the repeatability of this strategy by adding 50 nM miR-155 to trigger the bCHA. The relative standard deviation (RSD) of 3.01% indicates the acceptable repeatability of the proposed catalytic amplifcation platform for biosensing of miRNAs. Moreover, the specifcity of the proposed bCHA strategy for miRNA analysis was confrmed by using one-base mismatched, three-base mismatched, and perfectly complementary miR-155 as the analytes, respectively (Fig. 3d). Under the same reaction conditions, this method could completely distinguish mismatched miR-155, even one-base mutant, showing high sequence specifcity toward the target miRNA due to the precise hybridization properties of TMSD reactions. To assess the reliability of this assay in real samples, the standard addition method was implemented by detecting miR-155 with different concentrations (5 nM, 25 nM, and 50 nM) spiked in 100-fold diluted healthy human serum. As shown in Table S2, † the recoveries of miR-155 range from 95.6% to 108.0% with the RSD of 2.8-3.3%. The results indicate that the bCHA strategy has a high ability to prevent interference for miRNA analysis, which thus is available for the development of diagnostic systems in clinical applications. ## Amplication efficiency of bCHA To demonstrate the amplifcation efficiency of bCHA, control experiments were carried out using free DNA hairpins instead of hairpin trimers to perform the traditional CHA, in which the miR-155 triggered a cascade of assembly steps with H1, H2 and H3 to form a three-arm DNA junction (Fig. S3a †). The comparison of real-time fluorescence intensities between CHA and bCHA triggered by different concentrations of miR-155 could be observed (Fig. S3b †). In the absence of an initiator, the fluorescence intensities of both systems remain at low levels. The slightly higher fluorescence of bCHA can be attributed to the system leakage induced by the mutual interference among multiple DNA sequences. The corresponding fluorescence intensity at the reaction time of 360 min is presented in Fig. 3e and S3c. †The fluorescence intensities of bCHA can be obviously distinguished from those of CHA upon the addition of miR-155 with the same concentrations (25 nM and 50 nM). In particular for high concentration of the initiator (50 nM), the signal amplifcation capability of bCHA was calculated as (F bCHA(50 nM) F bCHA(0 nM) )/(F CHA(50 nM) F CHA(0 nM) ), showing that the bCHA had about a 1.68 times higher fluorescence signal than the traditional CHA. Thus, the proposed bCHA system provides signifcantly improved amplifcation efficiency for the detection of initiators (target miR-155 detection in this assay). We can attribute this phenomenon to the "diffusion effect". 42 Briefly, in a traditional CHA system the released initiators during reaction may interact with other hairpins present in solution. Nevertheless, the released initiators from bCHA are much more likely to trigger the nearby hairpin molecules attached on the Yscaffold DNA for cascade hairpin assembly, which thus achieves greater control over directing reaction pathways. ## In situ imaging of miR-155 in live cells It has been reported that miR-155 is overexpressed in many types of cancer cells, especially breast cancer. 41 Thus, MCF-7 cells (human breast adenocarcinoma cell line) were selected as the model to investigate the optimal time and monitor the dynamics of the bCHA in live cells (Fig. 4a). We frst test the stability of hairpin trimers, in which free H1 and hairpin trimers were treated with 100-fold diluted healthy human serum for 6 h, respectively. The result shows that the fluorescence recovery from H1 in hairpin trimers was much lower than that of free H1 (Fig. S4 †), demonstrating that the structure of the hairpin trimer could protect hairpins from nuclease degradation to a certain degree. Subsequently, the transfection times required for bCHA activation by miR-155 in MCF-7 cells were investigated (Fig. 4b and S5 †). On prolonging the transfection time, the fluorescence intensities increase gradually, and robust fluorescence signals produced by intracellular miR-155 can be readily observed at 4 h. 3D confocal fluorescence imaging further confrmed that the hairpin trimers have successfully entered MCF-7 cells and have been activated by miR-155 in the cytoplasm (Fig. 4c). We further designed a mutated H1 (six-base mismatched H1), named mH1, to the mechanism of bCHA in living cells. Almost no fluorescence was observed since the mH1 cannot hybridize with miR-155 to form DNA dendrimers, demonstrating that the observed fluorescence was indeed produced by target-induced assembly of hairpin trimers among H1, H2 and H3 (Fig. S6 †). To demonstrate the amplifcation efficiency of the proposed bCHA for monitoring intracellular miR-155, MCF-7 cells were transfected with only H1, the mixture of hairpin monomers (H1, H2 and H3) and hairpin trimers to perform direct fluorescence in situ hybridization (FISH) which involved the use of a singlestranded DNA probe modifed with FAM and Dabcyl, traditional CHA and bCHA, respectively. (Fig. 5a and b). When MCF-7 cells were transfected with only H1, since H1 was opened by miR-155 via TMSD reactions with a ratio of 1 : 1, a weak fluorescence signal was observed. However, for the transfection of hairpin monomers H1, H2 and H3 together, the traditional CHA reaction occurred by intracellular miR-155, showing a clear fluorescence signal due to the isothermal target recycling amplifcation of CHA. Interestingly, in comparison with traditional CHA, thanks to the "diffusion effect" discussed above, the strongest fluorescence signal could be observed when MCF-7 cells were treated with hairpin trimers, demonstrating the feasibility of bCHA for in situ imaging of intracellular miRNAs with enhanced signals. Flow cytometry analysis was further carried out (Fig. 5c and S7 †). Compared with MCF-7 cells treated with either H1 or three hairpin monomers (H1, H2 and H3), the hairpin trimer-transfected MCF-7 cells showed a signifcantly enhanced fluorescence signal, which was well consistent with the confocal microscopy results. To further demonstrate the effectivity of bCHA, control experiments were performed in which MCF-7 cells were pretreated with miR-155 mimics and the miR-155 inhibitor, respectively, and MCF-7 cells without treatment acted as the control group (Fig. 6). As anticipated, no fluorescence readout was observed by transfecting the anti-miRNA antisense inhibitor oligonucleotide because it can bring down the content of intracellular miR-155. In contrast, MCF-7 cells treated with miR-155 mimics to imitate the high expression of miRNA-155 showed much more intense green fluorescence than the untreated cells. It was clearly demonstrated that the fluorescence readout of the bCHA system was closely related to the miR-155 concentration in living cells, and the signal increased with increasing miR-155 concentration. To investigate the capacity of the bCHA strategy to evaluate intracellular miRNA expression levels in different live cells, four types of cells, including normal human hepatocytes (L-02), human cervical cancer cells (HeLa), and breast cancer cells (MCF-7 and MDA-MB-231) were respectively transfected with hairpin trimers with identical concentrations (200 nM) at 37 C for 4 h (Fig. 7). From Fig. 7, it can be seen that the expressions of miR-155 in different cell lines are various. Because the expression level of miR-155 in normal L-02 cells is very low, 45 the fluorescence signal is emerged hardly in L-02 cells. In contrast, the fluorescence signals can be readily observed in cancer cells since miR-155 is overexpressed in cancerous processes. 46 In addition, MCF-7 and MDA-MB-231 cells show stronger fluorescence intensities than HeLa cells, which is in accordance with previous reports that the expression level of miR-155 is higher in MCF-7 and MDA-MB-231 cells than in HeLa cells. 47,48 Moreover, the bCHA system can be applied as a general platform for the detection of other miRNAs, such as miR-21 (Fig. S8 †). It has been proven that miR-21 is an oncogene which is overexpressed in many cancers, such as breast cancer, lung cancer and so on. 49,50 Compared with the fluorescence signal induced by miR-155 in MCF-7 cells under the same conditions (CLSM imaging of MCF-7 cells in Fig. 7), a stronger fluorescence signal could be observed because the expression level of miR-21 is higher than that of miR-155 in MCF -7 cells. 51,52 In addition, traditional CHA and direct FISH are also performed in which bCHA exhibits a stronger fluorescence signal than CHA and direct FISH, demonstrating the proposed bCHA strategy can be used as a general platform for detecting various miRNA targets with high sensitivity and selectivity. ## Conclusions In summary, we have successfully demonstrated an enzyme-free DNA catalytic device of bCHA for dynamic self-assembly of DNA dendrimers. In comparison with traditional CHA, the signal amplifcation efficiency of bCHA is signifcantly improved by means of the "diffusion effect", achieving sensitive and selective in vitro detection of miRNAs and in situ imaging of miRNAs in live cells. Given the versatility of DNA, this dynamic DNA catalytic device can be readily used for amplifed detection of a wide range of analytes through combining with aptamer recognition, such as proteins, small molecules and even tumor cells. Therefore, the proposed bCHA strategy holds great potential not only in the construction of complex DNA structures but also as a versatile amplifcation platform in the felds of biosensing and bioimaging. ## Native polyacrylamide gel electrophoresis (PAGE) The reaction pathways of the catalytic DNA device were confrmed using 8% acrylamide gel. Firstly, 2.7 mL of 30% acrylamide/bis-acrylamide gel solution (29 : 1), 1 mL of 10 TAE buffer, 90 mL of 10% ammonium persulfate (APS), 10 mL of N,N,N 0 ,N 0 -tetramethylethylenediamine (TEMED) and 6.2 mL of double-distilled deionized ultrapure water were mixed to prepare the hydrogel. After polymerization for 30 min at room temperature, the gel was soaked in 1 TAE buffer (pH 8.0). Subsequently, 12 mL of each sample was mixed with 2 mL of 10 loading buffer and added to the resulting 8% native polyacrylamide gel for electrophoresis. The PAGE was run at the voltage of 170 V for 5 min and 110 V for about 45 min. Then, the gels were stained with diluted 4S Red Plus (Shanghai Sangon Biotech, China) for 40 min at room temperature. Finally, the images of the stained gels were recorded using a Tanon 2500R gel imaging system (Shanghai, China). ## Morphology characterization For microscope imaging, 2 mL of each annealed H1, H2 and H3 (10 mM for each) were added to 6 mL of the as-prepared Yscaffold DNA, resulting in the formation of hairpin trimers after reaction at 25 C for 1 h. The hairpin trimers as the reaction precursor were further mixed with an isometric volume of the initiator (0.1 mM), followed by incubation at 25 C for 1.5 h. The resultant products were washed by centrifugation at 10 000 rpm for 3 min with double-distilled deionized ultrapure water. The washing step was repeated three times to remove salt ions in solution. Then, 10 mL of the sample was dropped onto the glass slides. Finally, the images were taken on a microscope (Leica DM4 P, Germany) in transmission mode. ## Dynamic light scattering (DLS) and zeta potential The size distribution and zeta potential of the hairpin trimers and DNA dendrimers were measured on a Zetasizer Nano Series ZEN3700 (Malvern Instruments, UK). The DNA dendrimers were prepared according to the method mentioned above, which were diluted to 700 mL with double-distilled deionized ultrapure water for measurement. Fluorescence monitoring 2 mL of the initiator with different concentrations was rapidly mixed with 12 mL of hairpin trimers (the preparation method was the same as mentioned above) and 26 mL of TE buffer. The real-time fluorescence intensity was recorded immediately using a LineGene 4800 Real-Time detection system (Hangzhou, China) with intervals of 30 s. And the reaction was performed at 37 C for 6 h. Cell culture MCF-7 cells, HeLa cells, L-02 cells and MDA-MB-231 cells were all cultured in DMEM supplemented with 10% fetal calf serum, penicillin (100 mg mL ), and streptomycin (100 mg mL 1 ) in a humidifed and 5% CO 2 incubator at 37 C. PBS buffer (0.01 M, pH 7.4) was used to wash cells. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "An enzyme-free molecular catalytic device: dynamically self-assembled DNA dendrimers for <i>in situ</i> imaging of microRNAs in live cells", "journal": "Royal Society of Chemistry (RSC)"}

controlling_the_rotation_modes_of_hematite_nanospindles_by_dynamic_magnetic_fields

## Abstract: The magnetic field-induced actuation of colloidal nanoparticles has enabled tremendous recent progress towards microrobots, suitable for a variety of applications including targeted drug delivery, environmental remediation or minimally invasive surgery.Further size reduction to the nanoscale requires enhanced control of orientation and locomotion to overcome dominating viscous properties.Here we demonstrate how the coherent precession of nanoscale hematite spindles can be controlled via dynamic magnetic field. Using time-resolved Small-Angle Scattering and optical transmission measurements, we reveal a clear frequency-dependent variation of orientation and rotation of an entire ensemble of hematite nanospindles. Our findings 1 are in line with the different motion mechanisms observed for much larger, micron sized elongated particles near surfaces. The different dynamic rotation modes promise hematite nanospindles as a suitable model system towards field-induced locomotion in nanoscale magnetic robots. ## Introduction Field-driven actuation on magnetic particles builds the foundation of various intriguing applications, including self-propelling particles in active matter, 1,2 mixers in microfluidics, 3 and viscosity probes in nanorheology, 4 but also aspects based on structure formation, such as field-induced self-organization and smart fluids. 8,9 Biological microorganisms such as magnetotactic bacteria are inspiring model systems for magnetic microswimmers 10 operating at low Reynolds numbers, but also large enough so that their movement and dynamics can be tracked by optical techniques. The development of artificial magnetically-driven micro-and nanorobots has received tremendous attention in the last few years, 2,11,12 leading to advanced applications including magnetic two-arm nanoswimmers 13 or the combination of magnetic hyperthermia and magnetically driven propulsion for local pollutant remediation. 14 Achieving a remotely controlled, sustained translational motion for active Brownian particles in a viscous fluid is a challenging endeavour. This is typically approached by advanced synthesis of complex, chiral structures, 15 such as helical structures for magnetically actuated propulsion in dynamic, rotating or oscillating magnetic fields. 16 Very recently, a rich variety of motion mechanisms has been achieved using achiral, elongated objects with perpendicular magnetization. Even random shapes based on nanoparticle aggregates have been shown to succeed as fast magnetic micropropellers. 20 All these particles propel when stirred by a relatively weak rotating uniform magnetic field (of the order of a few mT). The motion mechanism is tunable from tumbling through precessing to rolling motion via the frequency of the rotating magnetic field, 19 ultimately fading into the step-out behavior associated with declining propulsion velocity at highest frequencies. 21 These examples emphasize the great potential of dynamic magnetic fields for the controlled locomotion of mesoscopic magnetic particles in a viscous medium and the design of magnetoresponsive soft matter. If the particle size is reduced to the nanoscale, thermal effects dominate, and stochastic motion of the particles due to random collisions (i.e., Brownian motion) complicate a controlled steering of magnetic nanoparticles in viscous fluids. 22 One of the rare examples of successfully synthesized propellers with dimensions below 1 µm was demonstrated with carboncoated aggregates of nanoparticles involving a post-synthesis selection process. 23 Hematite nanospindles represent a peculiar case of anisometric nanoparticles that promise to progress magnetohydrodynamics to an even smaller length scale. Hematite nanospindles are achiral, elongated nanoparticles (i.e., particles with at least two dimensions between 1-100 nm) that intrinsically exhibit a magnetization perpendicular to the long axis. This geometry is similar to the microwires with artificially engineered perpendicular magnetization 17,19 and therefore promising for locomotion of nanoobjects in dynamic magnetic fields. As the particle moments align parallel with an applied magnetic field, the hematite nanospindles orient with their long spindle axis perpendicular to a static field direction. 24,25 Within this plane perpendicular to the field the spindle orientation fluctuates randomly. There are two different characteristic relaxation frequencies associated with (1) a rotation of the spindle along the long rotational axis and (2) a rotation along the minor axes. Here we demonstrate that by application of a rotating magnetic field in a suitable frequency range, the dynamic reorientation of hematite nanospindles can be forced into a synchronized spinning. Using time-resolved small-angle X-ray scattering (SAXS), we reveal how the orientation of an ensemble of water-dispersed hematite nanospindles can be tuned between a coherent precession and a collinear alignment by suppression of one of the two distinct rotation directions. The time-resolution of stroboscopic SAXS provides information on the dynamic particle orientation while applying alternating or rotating magnetic field. Our results are supported with complimentary optical transmission measurements and agree with theoretical estimates of the rotational diffusion constant. The observations demonstrate that hematite nanospindle hold out the key features for future application as nanopropellers. ## Results and Discussion To achieve well-separated relaxation time scales, a large aspect ratio of the hematite spindles is required. We therefore synthesized hematite nanospindles according to the hydrolysis method developed by Matijević and co-workers that allows to tune the aspect ratio. 26 TEM analysis reveals an average diameter of 51.5 nm and a particle length of 374 nm along with normal size distributions in the range of 13-20%, corresponding to an aspect ratio of around 7.3 (Figure S1). Small-angle X-ray scattering (SAXS) of the colloidal suspension of nanospindles in water, performed at the instrument ID13 (ESRF), confirms the average particle diameter of 49.9 nm with an intermediate lognormal size distribution of 9.6% (Figure S5). The quasi-static magnetization measurement of the colloidal suspension further confirms the weakly ferromagnetic behavior of hematite above the Morin transition by the pseudo-superparamagnetic field dependence typically observed for magnetically blocked nanoparticles in dispersion (Figure S2). Refinement of the Langevin behavior reveals an integral nanoparticle moment of 7.643 •10 −19 J/T, corresponding to 82414(705) µ B . Along with the morphological particle volume of 5.19 •10 −22 m 3 as determined from TEM and SAXS analysis, a spontaneous magnetization of 1473 A/m is derived. The obtained spontaneous magnetization represents only 74% of the bulk value for hematite of ∼ 2000 A/m, 27 in agreement with earlier reports. 24,28 From this follows that the particles are not interacting and hence can freely rotate in suspension when applying external magnetic fields. The anisotropy of the relaxation frequencies is confirmed by depolarized dynamic light scattering (DDLS, Fig. S3) and AC magnetic susceptometry (ACMS, Fig. S4). The rotational diffusion constant D R = 149(7) s −1 derived from DDLS analysis agrees well with the theoretical estimate of 179 s −1 for an ellipsoid of revolution of the same dimensions 29,30 and corresponds to a characteristic frequency of ν ⊥ = D R /π = 47(2) Hz. In contrast, ACMS reveals a characteristic frequency of ν ACMS = 278 Hz. The strong difference between these characteristic frequencies highlights the different mechanisms probed by both techniques. Whereas DDLS is sensitive to the orientational diffusion of the particles in dilute dispersion, and hence can only sense the rotation around the minor spindle axis, ACMS probes the fieldinduced orientation of the spindle magnetization, including contributions of both rotation around the minor and principal axes. The characteristic frequency for rotation around the principal spindle axis is therefore not unambiguously accessible from ACMS, and will lie well beyond ν ACMS = 278 Hz. The significant difference between the characteristic frequencies for rotation around principal and minor spindle axes is an important prerequisite for a tunable motion control: we expect that for a magnetic field rotating with a frequency between both characteristic frequencies, a transition between coherent precession and synchronized spinning of nanospindles in dispersion is achievable. 31 Small-angle scattering probes nanoscale inhomogeneities with suitable spatial resolution to address the orientation distribution of a nanoparticle ensemble, 25,32,33 whereas timeresolved approaches give versatile opportunities to study in-situ nanoparticle ensemble dynamics. Stroboscopic SAXS provides the necessary time resolution to monitor changes in the orientation distribution near the characteristic frequencies. We therefore performed time-resolved SAXS at the microfocus beamline ID13 (ESRF) to monitor the dynamic fielddependent orientation of hematite spindles driven by a custom-made set of Helmholtz coils. The two Helmholtz coils can each generate alternating magnetic fields, and together a magnetic field rotating in the horizontal plane when both applied fields have the same amplitude and frequency with a 90°phase shift (Fig. 1a). The magnetic field period was divided into 20 frames by synchronizing the coil setup with the detector to obtain the time-resolved scattering patterns as shown in Figure 1f. The Maxipix detector allowed to measure stroboscopically up to 300 Hz. To emphasize the scattering anisotropy caused by the particle alignment in the applied field, all the following scattering images will be the difference pattern with respect to the isotropic scattering pattern in zero field (see Fig. S6). At low frequencies of 25 Hz, where the moments of the entire nanospindle ensemble follow the applied field, a periodic variation of the anisotropic scattering intensities is observed for both alternating and rotating magnetic fields (Fig. 2, full period shown in Fig. S7). In alternating magnetic field, the time-resolved variation of scattering intensities indicates a dynamic interplay of order and disorder. Whereas the SAXS intensity is isotropic at times near the zero-field condition (1/2 π and 3/2 π), corresponding to the random nanoparticle orientation in the absence of a magnetic field, a clear scattering anisotropy is observed at maximum field of 10 mT (0 and π). This scattering anisotropy results from the orientation of the spindles perpendicular to the inducing field, where the degree of alignment corresponds well to the Langevin parameter of 1.85, estimated for the integral nanospindle moment in the maximum applied field of 10 mT. We confirm this by computing the expected scattering pattern for a spindle ensemble using the spindle dimensions, integral particle moment, and average applied field for each frame according to the Boltzmann statistics of the particle distribution. 25 As shown in Fig. 2, the computed scattering pattern is in excellent agreement with the measured pattern. In case of a rotating magnetic field of 10 mT, the easy magnetic axes of the nanospindles will maintain their orientation towards the applied field and the nanoparticle ensemble may fulfill a complete turn within one period of the applied field as long as thermal motion and fluid friction are subsidiary effects. We identify this behavior in low frequency rotating magnetic field as coherent precession: the spindle ensemble fulfills precession around the field normal with a coherent phase behavior, albeit different precession angles. The difference between the two cases of alternating and rotating magnetic fields becomes very clear at the 1/2 π and 3/2 π time frames. In these time frames the applied rotating field is oriented parallel to the X-ray beam, resulting in isotropic scattering patterns. However, the difference of the scattering intensities against the zero-field state is negative. The computed scattering pattern (Fig. 2) confirms that the nanospindle ensemble is oriented parallel to the detector plane, whereas for the alternating field these time frames correspond to an isotropic, disordered ensemble at nearly zero field and hence vanishing difference scattering intensities. The evolution of difference scattering patterns with increasing frequency for both the alternating and rotating field (Fig. 3) reveals how the time-dependent fluctuation of the scattering intensities disappears, while the scattering anisotropy remains. This is a clear signal of a transition from the dynamic particle reorientation observed at low frequency towards a confined particle arrangement. A more quantitative picture is established by analysis of the time-and frequency-dependent scattering anisotropy of the two field configurations (Fig. 4a and 4b), derived as the difference between scattering intensity in horizontal and vertical directions (see SI). Whereas in the low frequency case (25 Hz) a clear time-resolved variation between maximum and vanishing scattering anisotropies is evident, the increasing magnetic field frequency is accompanied with a significant phase lag of the scattering anisotropy. This is a strong indication that there is a dissipative process acting such that the nanospindles cannot follow the dynamic magnetic field anymore at elevated frequencies. The time-dependent amplitude in scattering anisotropy decreases, corresponding to a more and more static orientation of the ensemble of nanospindles. However, there is a significant scattering anisotropy even at the highest investigated stroboscopic frequency of 300 Hz at all times. This indicates that the average orientation of a significant portion of the nanospindle ensemble is not isotropic. For higher frequencies beyond 300 Hz, only time-averaged scattering anisotropies, corresponding to the dotted lines in Fig. 4a and b, are accessible from time-averaged SAXS data. Over the complete frequency range, these illustrate the transitions between different types of collective motion (Fig. 4c). A maximal scattering anisotropy occurs for 150-200 Hz, i.e. in between the characteristic frequencies for rotation around the minor axis (47 Hz as determined from DDLS) and principal axis (beyond 278 Hz as estimated from ACMS and DDLS). In this frequency range, rotation around the principal axis is still allowed while rotation around the minor axis is inhibited. Beyond the characteristic frequencies for rotation around both principal and minor axes, the spindles do not follow the field variation anymore. Consequently, the scattering anisotropy decreases continuously, indicating an increasingly isotropic orientation of the nanoparticle ensemble. Despite the similar behavior of the scattering anisotropy with dynamic field frequency, a different spindle orientation distribution for alternating and rotating magnetic fields at intermediate frequency is inferred from the scattering intensities shown in Fig. 3b. The differential scattering patterns correspond well to those computed for orientation distributions with the long spindle axis either confined to the plane perpendicular to the alternating magnetic field or to the direction perpendicular to the rotating field plane (Fig. 3b). In case of the alternating field, the time-independent but anisotropic scattering intensity indicates that the spindle long axes stabilize permanently in the plane perpendicular to the field direction. We understand this such that beyond the characteristic frequency for rotation around the minor axis, the average field-induced angular moment becomes larger than the disordering Brownian momentum. In effect, the spindles become confined in the plane perpendicular to the field and fulfill random rotation on this 2D plane. 37 In the rotating case, on the other hand, a particle experiences a constant torque produced by the rotating magnetic field that increases with frequency and eventually becomes significantly larger than Brownian random fluctuation within the 2D plane. A self-stabilizing rotational motion is favored as the viscous friction is reduced in a rotation around the long axis as compared to precessing motion involving rotation around the short axis. As a result, the spindle long axis stabilizes permanently in the direction perpendicular to the rotating field plane such that the magnetic moments can follow the rotating field on the shortest path of least action. We understand this as a synchronized spinning: the ensemble of collinearly aligned nanospindles rotates synchronously around their major spindle axes. The field-induced, frequency-dependent reorientation variation observed by time-resolved SAXS is strongly supported by optical transmission measurements (Fig. 5). The optical transmission of linearly polarized laser light through a dilute suspension of elongated nanoparticles depends directly on the relative orientation of laser light polarization and principal nanoparticle axis. For hematite nanospindles in a static magnetic field, an increase in optical transmission with applied static field parallel to the polarization direction and a decrease in optical transmission for a magnetic field applied perpendicular to the polarization direction is consequently observed (Fig. 5a). The time-resolved optical transmission recorded in rotating magnetic field of 25 Hz (Fig. 5b) oscillates exactly between the maximum and minimum static transmission, confirming that the entire nanoparticle ensemble follows the applied field. Similarly, in alternating magnetic field of 25 Hz (Fig. 5c) the timeresolved optical transmission oscillates between the transmission extrema and 1, indicating considerable orientation of the nanospindles in the maximum applied field and isotropic orientation in zero field. With increasing frequency of both rotating and alternating magnetic field (Fig. 5d), the full orientational order of the static or nearly-static case is not reached anymore. This is indicated by the decrease of the time-resolved amplitudes as well as by the center of the relative optical transmission approaching 1. However, the time-averaged anisotropy remains high, so that the full rotation amplitudes move above the isotropic case (I/I 0 = 1, grey line in Fig. 5d), indicating that a new, ordered state is achieved. The baseline of the rotation amplitude above 1 indicates that in the characteristic frequency range (with a maximum at 200-250 Hz), a significant portion of the spindles must be aligned perpendicular to the polarization direction at all times, which can only be the case if the spindles are aligned perpendicular to the rotating field plane. Likewise, the oscillation amplitudes in the alternating field do not reach the fully isotropic state at elevated frequencies, supporting the anisotropic nanospindle orientation observed with SAXS. To summarize the response of the nanospindle ensemble in rotating fields: at low frequencies below the characteristic relaxation frequencies, the nanospindles follow the magnetic field and remain quasi-statically oriented with their major axis perpendicular to the inducing magnetic field, corresponding to coherent precession around the field normal. For intermediate magnetic field frequencies between the two relaxation time scales, rotation around the minor particle axis becomes suppressed, resulting in a decreasing precession angle. In consequence, the particles are driven into a collinearly aligned orientation of the major particle axis with synchronized spinning around the normal to the rotation plane. At high frequencies beyond the characteristic frequencies of rotation around both axes, the spindle ensemble disorients toward isotropic disorder. ## Conclusion We have elucidated the frequency-dependent reorientation behavior of hematite nanospindles in dynamic magnetic fields. The study emphasizes the potential of dynamic fields to control the rotation modes of shape anisotropic colloidal magnetic nanoparticles with perpendicular magnetic anisotropy. Stroboscopic SAXS resolves signatures of different types of motion with a clear enhancement of the particle orientation in the intermediate frequency range between the characteristic relaxation time scales. The orientation behavior strikingly differs between alternating or rotating magnetic field. Time-resolved SAXS is a valuable tool to investigate this type of dynamic self-organization in-situ towards nearly collinear alignment of nanospindles perpendicular to the rotating magnetic field. The peculiar behavior in rotating magnetic field from coherent precession to synchronized spinning is understood similar to the frequency dependent variation of motion mechanisms of individual microwires near a surface boundary, ranging from tumbling and precessing to a rolling motion with increasing frequency. 19 With further increasing frequency, the effective field that aligns the particle upwards reduces due to the increasing phase lag of the spindle magnetization towards the rotating magnetic field. For dynamic frequencies well above the characteristic frequencies for rotation around the major axis, the synchronized rotational motion of the spindle ensemble ceases. This corresponds well to the step-out behavior observed as a decay in propulsion velocity of helical objects in rotating field with increasing frequency. 21,38 The characteristic frequencies for rotational diffusion of hematite spindles are adjustable by variation of the spindle length and aspect ratio through synthetic considerations. This will enable direct tuning of the frequency range needed to control the different rotation modes. The ability to control the dynamic reorientation of large ensembles (in the order of 10 10 ) nanoscale magnetic particles establishes an important step towards field-driven actuation and locomotion. With this prerequisite, oriented locomotion of a swarm of nanoparticles may become accessible using complex magnetic field geometries, such as combined rotating and static magnetic fields. 39

chemsum

{"title": "Controlling the rotation modes of hematite nanospindles by dynamic magnetic fields", "journal": "ChemRxiv"}

dialumenes_–_aryl_<i>vs.</i>_silyl_stabilisation_for_small_molecule_activation_and_catalysis

## Abstract: Main group multiple bonds have proven their ability to act as transition metal mimics in the last few decades.However, catalytic application of these species is still in its infancy. Herein we report the second neutral NHC-stabilised dialumene species by use of a supporting aryl ligand (3). Different to the trans-planar silyl-substituted dialumene (3 Si ), compound 3 features a trans-bent and twisted geometry. The differences between the two dialumenes are explored computationally (using B3LYP-D3/6-311G(d)) as well as experimentally. A high influence of the ligand's steric demand on the structural motif is revealed, giving rise to enhanced reactivity of 3 enabled by a higher flexibility in addition to different polarisation of the aluminium centres. As such, facile activation of dihydrogen is now achievable. The influence of ligand choice is further implicated in two different catalytic reactions; not only is the aryl-stabilised dialumene more catalytically active but the resulting product distributions also differ, thus indicating the likelihood of alternate mechanisms simply through a change of supporting ligand. ## Introduction The ability to isolate and stabilise complexes containing metalmetal bonds is of fundamental interest, providing both experimental and theoretical insights into the intrinsic nature of the metal centre. 1 Since the discovery that the so-called 'double bond rule' could be broken in the beginning of the last quarter of the 20th century, 2-5 efforts within main group chemistry have strived towards isolating a plethora of both homo-and heteromain group element multiply bonded compounds, which have been the subject of numerous reviews. 6,7 Aside from curiosity, one of the driving forces behind this research area is the ability to use main group multiple bonds as transition metal mimics. This is possible due to similarly energetically accessible frontier molecular orbitals. Thus, reduction of small molecules, such as dihydrogen, under ambient conditions by sustainable main group metal centres is achievable. 11 Whilst the ability to mimic transition metals is now possible in regard to oxidative addition reactions, main group elements still fall short in terms of catalytic activity due to the resulting stability of the higher oxidation state complexes, i.e. the frst step in a redox based catalytic cycle. In order to truly compete with transition metals that are currently employed in industry, the ability to influence the stability, and thus reactivity, of main group metal centres is paramount. One method of influencing stability is through choice of stabilising ligand. If you consider disilenes, the choice of silyl, aryl and nitrogen-based ligands has been shown to influence the structural parameters around the double bond, 12,13 with silyl groups tending towards transplanar geometries 13 and aryl groups promoting trans-bent character. It was not until the use of an N-heterocyclic imine (NHI) based ligand, which results in a highly trans-bent and twisted geometry, that dihydrogen activation was achieved. 14 An electropositive silyl supporting ligand was used to stabilise the frst neutral aluminium-aluminium double bond, namely dialumene. 15 DFT calculations found the HOMO to consist of a p-bond formed from almost pure Al p-orbitals and as such a planar geometry was observed. As predicted, the dialumene behaved as a transition metal mimic towards a variety of small molecules, as well as enabling catalytic reduction of CO 2 . 16 Prior to the isolation of the frst neutral dialumene, several compounds with Al-Al bond orders greater than 1 were isolated. 17 These can be classed as radical monoanionic species, one electron p-bonded compounds, a dianionic complex and masked dialumenes. The stick with latter, reported independently by Power 18 and Tokitoh, 19 proposed the intermediacy of aryl-stabilised dialumenes, with the masked species being a result of [2 + 4]-cycloaddition reaction due to the use of aromatic solvent. This was additionally accounted for through a series of [2 + 2]-cycloaddition reactions with internal alkynes. Tokitoh further showed that the benzene derived masked species was capable of activating dihydrogen; 20 however, upon switching to an anthracene derived masked species no reactivity towards dihydrogen was observed. On descending group 13, heavier digallenes and dithallenes have been isolated which show notable trans-bent character and have been known to dissociate to their corresponding monomers in hydrocarbon solutions. However, digallanes have been shown to react as the double bonded species, rather than the monomer with regards to cycloadditions of unsaturated C-C bonds and even dihydrogen activation. Motivated by our group's previous efforts in dialumene chemistry, we targeted the isolation of a neutral aryl-stabilised dialumene to compare the intrinsic nature of the aluminiumaluminium double bond through the influence of ligand stabilisation. Whilst silyl and aryl groups have been routinely used in main group multiple bond chemistry, no direct comparisons of their influence on multiple bonds as reactive species have been drawn. As such, we proposed a systematic study of both dialumenes towards activation of a range of small molecules and their use in catalysis, with the aim of providing experimental and theoretical insight into the influence of these ligand classes on main group multiple bond reactivity. ## Synthesis of aryl-stabilised dialumene Following on from the successful isolation of the frst neutral dialumene, we focused our attention on expanding the scope of this class of compounds towards aryl stabilised systems. As such, we targeted the use of the Tipp ligand (Tipp ¼ 2,4,6-tri-isopropylphenyl) for the stabilisation of a new dialumene. In keeping with the previous dialumene, the choice of Nheterocyclic carbene (NHC) remained the same, I i Pr 2 Me 2 (I i Pr 2 Me 2 ¼ 1,3-di-iso-propyl-4,5-dimethyl-imidazolin-2ylidene). Direct reaction of I i Pr 2 Me 2 AlH 3 and LiTipp at 78 C resulted in formation of the monosubstituted aluminium dihydride complex I i Pr 2 Me 2 Al(Tipp)H 2 (1) (Scheme 1) in good yield (66%, 27 Al: d 112.9 ppm). 29 The identity of compound 1 was confrmed upon inspection of the 1 H NMR spectrum wherein three resonances for the iso-propyl groups were identifed in a 2 : 2 : 1 ratio (NHC : o-Tipp : p-Tipp iso-propyl signals) as well as a characteristic broad signal for the Al-H 2 protons ( 1 H: d 5.11 ppm). Additionally, a sharp IR stretching band at 1711 cm 1 (Al-H) was observed in the IR spectrum. Conversion of 1 towards formation of I i Pr 2 Me 2 Al(Tipp)I 2 (2) could be achieved through reaction with BI 3 $dms (dms ¼ dimethyl sulfde) or with a small excess of methyl iodide, with the latter resulting in higher and cleaner conversion; moreover, the concomitant formation of methane allows for facile reaction monitoring. Loss of signals relating to Al-H were observed in both the 1 H NMR and IR spectra, and further characterisation by single crystal XRD confrmed the identity of 2 (Fig. S54 †). Compound 2 is structurally analogous to the corresponding silyl supported complex (2 Si ) with Al-C NHC bond lengths essentially the same (2: 2.0645(18) ; 2 Si : 2.0673(17) ), indicating the dative nature of the NHC ligand. This is additionally confrmed on comparison with the Al-C Tipp bond length (1.9887(19) ), which is smaller than the sum of the covalent radii (R Al-C ¼ 2.01 ). 30 Following the analogous synthetic protocol to the silyl dialumene, compound 2 was stirred vigorously with KC 8 at room temperature (Scheme 1). Through monitoring the reaction by 1 H NMR, it was found that this reaction requires 72 hours rather than the 24 hours required for the previous case. Compound 3 was isolated as a black solid, and in contrast to the silyl stabilised dialumene (3 Si ), 3 is highly soluble in a broad range of aromatic, alkyl and ethereal solvents. Both dialumenes are stable in the solid state in an inert atmosphere for prolonged periods; however, they decompose in solution after 24 hours. The 1 H NMR spectrum of 3 shows a large broad signal at room temperature ($7.0-5.5 ppm) which resolves into distinct signals at 228 K for the iso-propyl groups, indicating a degree of rotational fluxionality in this system (Fig. S11 †). Single crystals were grown from a concentrated n-hexane solution at 5 C and revealed a trans-bent and twisted geometry of the aryl stabilised dialumene (compound 3, Fig. 1) (q ¼ 17.25 , 23.70 , s ¼ 12.06 ), which contrasts with the trans-planar geometry observed previously. Furthermore, in 3 Si the NHC groups were found to be parallel to each other, whilst in 3 they are found to be almost perpendicular (85 ). The change from a planar to a trans-bent and twisted geometry has also been observed in disilene chemistry on switching between aryl and silyl-based ligands. 12,13 The Al-Al bond length is 2.4039(8) which is fractionally longer than that in the previous dialumene (3 Si : 2.3943(16) ). Another notable difference between the two systems lies in the Al-C NHC bond length (3: 2.0596(16), 2.0422(17); 3 Si : 2.073(3) ). The shorter bond length in the case of aryl stabilisation likely indicates a decrease in dative character and thus an increase in the covalent nature of the Al-C NHC bond (which is also supported by the calculated bond dissociation energy, see below). ## Computational discussion of aryl-stabilised dialumene To gain a deeper insight into the differences between these two classes of dialumenes, we performed density functional theory (DFT) calculations at the B3LYP-D3/6-311G(d) level of theory (for detailed information see the ESI †). The optimised geometry of 3 is in good agreement with experimental values, with the addition of the dispersion required to account for the trans-bent and twisted geometry. For comparison, all calculated values for 3 Si including dispersion are given in the ESI. † Analysis of the frontier orbitals of 3 revealed similar features to 3 Si , in which the HOMO1 and HOMO contain the Al-Al sand p-bonds, respectively, as well as the LUMO representing the Al-C NHC p bond (Fig. 2). The main difference to 3 Si (see Fig. S57 † for the corresponding orbitals) is the loss of uniform arrangement of the HOMO on the two aluminium centres as well as additional p-incorporation of the NHC present in 3. We attribute this to the different orientation of the NHCs in 3, enabling overlap with the p-orbital of the carbene carbon atom, which is experimentally observed as a shortened Al-C NHC bond in the SC-XRD structure and further evidenced by an increased Gibbs free energy of bond dissociation of 26.0 kcal mol 1 (cf. 3 Si ¼ 16.9 kcal mol 1 ). The conjugation towards the Al-C NHC p-bond in 3 is also observed on inspection of the monomers (for further details see ESI Fig. S62 †), which also gives rise to the decreased HOMO-LUMO gap in 3 compared to 3 Si (3 ¼ 1.86 eV, 3 Si ¼ 2.24 eV) based on decreased overlap of the monomers being possible. TD-DFT calculations corroborated the experimental UV/vis spectrum of 3. This showed an intense absorption band at 833 nm (3 ¼ 6273 L mol 1 cm 1 ) (calc. value 794 nm), which can be assigned to the HOMO to LUMO transition and is responsible for the highly coloured compound. 31 Natural bond orbital (NBO) analysis provided electronic insight into the nature of the Al-Al double bond. The Al-Al p-bond of 3 bears reduced pcharacter compared to 3 Si . Moreover, this NBO orbital has a lower occupancy based on partial population of the p*orbitals of the C-N bonds of the NHC, rationalising the increased interaction of the NHC for 3. This is based on its different orientation, which we mainly attribute to the reduced steric demand of the ligand. Furthermore, this is reflected in the decreased Wiberg bond index (WBI) of the Al-Al bond of 1.67 to 1.53 going from silyl to aryl (Fig. 3), yet still indicating a high degree of multiple bonding character in both systems. Analysis of NPA charges clearly reflects the silyl effect (Fig. 3): the aluminium centres in 3 Si bear a nearly neutral charge of +0.08, while in 3 they account for +0.48/+0.49. We attribute this to the silyl substituents, with their strong s-donating properties, possessing a more effective orbital overlap with the aluminium centres in the s(Al-Si) bonds. This also becomes evident upon examination of the results of NBO analysis for the Al-C Tipp /Al-Si bonds: the Al-C Tipp bonds are highly polarised (17% Al, 83% C Tipp ) compared to Al-Si bonds in 3 Si (36% Al, 64% Si). This difference is also rationalised upon comparison of Al-C and Al-Si Pauling electronegativities (Dc Al-C 0.94; Dc Al-Si 0.29), thus resulting in less polarised Al-Si bonds. To elucidate the effect of sterics around the aluminium centre, we initially compared the steric demand of the Tipp and Si t Bu 2 Me ligands, which revealed similar percentages of buried volume (% V bur ) of 29.9% (3) and 30.7% (3 Si ) (Fig. S58 and S59 †). However, the shape and thus distribution of kinetic stabilisation vary. Further calculation and comparison of reduced model systems were performed with the I i Pr 2 Me 2 carbene replaced by IMe 4 ( IMe4 3 and IMe4 3 Si , Fig. S60 †). 32 For IMe4 3 Si the most stable isomer exhibits a strongly transbent and twisted geometry (q ¼ 42.1 , 30.4 , s ¼ 11.7 ) with the NHC planes orientated almost parallel and a substantially elongated Al-Al bond length of 2.43 . In the Tipp-substituted IMe4 3 the trans-bent character is decreased compared to 3, accompanied by a small increase of the Al-Al bond length due to further rotation of the NHC planes towards the Al-Al plane, enabling more effective p-interaction of NHC and the AlAl moiety. This also becomes more apparent upon examination of the corresponding frontier orbitals with IMe4 3. This features enhanced delocalisation of the HOMO onto the NHC moiety, as a consequence of further rotation of the NHC planes towards the Al-Al bond (angle between NHC planes: 49 ). In contrast, the HOMO in IMe4 3 Si exhibits contributions from the silyl groups, as conjugation towards the NHC p-system is not possible due to the different orientation. Moreover, the HOMO-LUMO gap decreases for aryl and increases for the silyl case, attributed to the decreased/increased trans-bent geometry. The smallest possible model systems, by reducing I i Pr 2 Me 2 to IH 4 as well as Tipp/Si t Bu 2 Me to phenyl/TMS, were calculated and yield comparable results: S 3 and S 3 Si both possess transbent but no twisted confguration. In S 3 Si a slightly shorter Al-Al bond length and a decreased trans-bent angle (21.2 vs. 31.8 for S 3) are observed; however, in each case the NHC planes are rotated towards the Al]Al bond (see Fig. S61 †). Hence, transbent structures are obtained for the aryl and the silyl substituted dialumenes bearing minimal steric effects. Thus, it is clearly demonstrated that the steric effects of both NHC and the ligands govern the binding motif of dialumenes. The shape of the ligand influences the interaction with the NHC: either only a weak and purely s-donating type of interaction, as observed in planar 3 Si , or more flexible coordination of the NHC, with the porbital of the C carbene able to form a slipped p-bond with the AlAl core, as observed in 3. The trans-bent and twisted structure obtained for 3 is therefore a result of the difference in steric demand of the Tipp ligand compared to Si t Bu 2 Me. From the different aspects, steric as well as electronic, discussed above we thus conclude that the structural difference between 3 and 3 Si is caused by the different steric demand of the ligands. From the electronic point of view the change of silyl to Tipp ligand in the dialumene changes the orbital situation at the central AlAl core, which is accompanied by a reduced HOMO-LUMO gap. Moreover, the polarisation of the aluminium centres is different, based on differences in electronegativity between C and Si. We thus anticipate differences in reactivity with respect to activation of strong bonds, such as those in small molecules, as well as an increased accessibility towards a bigger range of reagent molecules of 3, based on the increased flexibility of this system. ## Reactivity of dialumenes Further differences between these two systems were sought experimentally. Firstly, reactivity towards a series of C-C multiple bonds was examined (Scheme 2). In the case of ethylene, compound 3 underwent formal [2 + 2]-cycloaddition to yield the dialuminacyclobutane compound 4, akin to the reactivity observed with 3 Si . 15 Upon reaction with 1 equivalent of phenylacetylene, clean formation of a single species by NMR spectroscopy was noted to occur to form compound 5. This is in contrast to the reactivity of 3 Si where both [2 + 2]-cycloaddition (5 Si ) and C-H activation were observed. Varying the number of equivalents of phenyl acetylene (2 : 1 and 3 : 1 with respect to 3) did not result in further incorporation of phenyl acetylene into the complex even at elevated temperatures. However, compound 5 was found to decompose in solution to yield styrene (see ESI Fig. S37 and S38 †). Monitoring a C 6 D 6 solution of 5 showed that this occurs intramolecularly, with the additional protons required to make styrene likely the result of C-H activation. In further support of an intramolecular C-H activation, addition of a hydrogen source, e.g. dihydrogen, phenyl silane, pinacol borane or amine borane, failed to provide any notable increase in the rate of styrene formation. Unfortunately, attempts to identify the fate of the resulting aluminium containing species were unsuccessful. It is proposed that initially [2 + 2]-cycloaddition occurs to form compound 5, followed by C-H activation of the iso-propyl groups of the Tipp ligand, as this was not observed with the analogous silyl complex. Intramolecular C-H activation of the Mes* ligand (Mes* ¼ 2,4,6-tri-tert-butylphenyl) has been previously observed by thermolysis of (Mes*) 2 AlH, 33 Further reactivity towards C-C multiple bonds was trialed with diphenylacetylene (PhCCPh). Addition of 1 eq. of PhCCPh to 3 Si failed to cause a reaction, and after prolonged heating only decomposition of 3 Si was observed. In contrast, PhCCPh was observed to react readily with 3, notably through the instant colour change from the dark black solution of 3 to a yellow solution of compound 6. This difference in reactivity was surprising, considering that both 3 and 3 Si reacted cleanly with both non-polar (ethylene to form 4 and 4 Si ) and polar (phenylacetylene to form 5 and 5 Si ) C-C multiple bonds. This difference in reactivity is thought to be a direct result of the choice of stabilising ligand. The flexibility of the Tipp ligand, due to the rotational iso-propyl groups, makes the central AlAl core more accessible for reactant molecules, thus enabling reactivity with more sterically demanding reagents. Moreover, the positive NPA charges of 3 make it more electrophilic in comparison to 3 Si , thus implying higher reactivity towards nucleophilic C-C multiple bonds. Compounds 4 and 6 were crystallised from concentrated pentane solutions at 30 C. The XRD structures revealed addition of the C-C multiple bonds to the dialumene, resulting in the formation of 4-membered rings (Fig. 4). Loss of double bond character from the dialumene was confrmed due to elongation of the Al-Al bond (4 2.6035(13), 6 2.5918(6) vs. Extension of this work towards C-N triple bonds focused on the use of 2,6-dimethylphenylisocyanide (XylNC). Previously, Tokitoh and co-workers had shown that reaction of their masked dialumene species resulted in homocoupling of isocyanides. 35 Reactions of varying equivalents of XylNC to 3 Si all resulted in an ill-defned mixture of species; unfortunately, attempts to separate species through fractional crystallisation failed. In contrast, reaction of 2 eq. of XylNC with 3 resulted in a clear colour change from black to red and produced a well-defned but complex 1 H NMR spectrum of compound 7 (Scheme 3). This complex contains bridging CNXyl units due to the observed downfeld signal in the 13 C NMR spectrum at d 303.4 ppm. This was similar to the previously observed bridging carbonyl fragment observed with 3 Si , in the rearrangement of CO 2 (d 276.0 ppm) 16 and the bridging isocyanide intermediate reported by Tokitoh (d 294.7 ppm). 35 Single crystals of 7 were grown from a 2 : 1 (toluene : hexane) mixture at 5 C, revealing a butterfly confguration with two m-CNXyl units (Fig. 5). The central Al-C Xyl -Al-C Xyl core in compound 7 is puckered (34. 4) ), which is in line with average C]N bond lengths. Additionally, the change in angle around the nitrogen in XylNC from linear to bent (126.3 (2) ) confrms reduction of the C-N triple bond. This butterfly confguration has been previously observed with transition metal complexes; however, they all contain a M-M bond, and those without M-M bonds contain a planar central ring. To confrm the nature of the bonding within compound 7 DFT studies were performed again, at the B3LYP-D3/6-311G(d) level of theory. The optimised structure is in good agreement with the one obtained experimentally by SC-XRD, including the calculated C]N IR stretching frequencies (experimental: 1545 cm 1 vs. calculated: 1568 cm 1 ). Orbital analysis (Fig. 6) ## Small molecule activation Further reactivity differences were sought through investigation towards small molecules (Scheme 4). Previously, reaction of 3 Si towards carbon dioxide (CO 2 ) resulted in an initial CO 2 fxation complex. 16 This subsequently underwent C-O cleavage reaction, in the absence of additional CO 2 through rearrangement to a bridging carbonyl complex, whilst in the presence of CO 2 , formation of a carbonate species with elimination of CO was observed. On reaction of 3 with CO 2 immediate loss of the black colour and formation of a colourless solution was observed. On inspection of the 13 C NMR spectrum, the presence of CO (d 184.4 ppm) and CO 3 (d 159.12 ppm) was observed, indicating the formation of the carbonate complex compound 9. In contrast to 3 Si , attempts to isolate the CO 2 fxation product were unsuccessful as it rapidly converted to compound 9. Use of the labile I i Pr 2 Me 2 -CO 2 species allowed for the formal [2 + 2]cycloaddition product (compound 8) to be observed due to its characteristic 13 C resonance at d 207.7 ppm (8 Si d 209.9 ppm). However, this reaction resulted in multiple species as well as compound 9, owing to the higher reactivity of the Tipp dialumene (compound 3), thus indicating that formation of 9 proceeds through the CO 2 fxation species in a similar manner to 3 Si . Reaction of 3 with O 2 resulted in the expected dioxo product, compound 10, same as the previously reported reaction of 3 Si . In a similar manner to 10 Si , compound 10 can also be reacted with CO 2 resulting in carbonate complex 9. In a notable difference to the silyl supported reactivity, addition of N 2 O to compound 3 resulted in a dark red solution at room temperature (3 Si yielded colourless compound 10 Si ). This red solution was observed to slowly fade to colourless over a few hours and the formation of compound 10 was confrmed by 1 H NMR spectroscopy. Use of 1 eq. of an oxygen donor reagent, namely N-methylmorpholine-N-oxide, with 3 allowed for clean isolation of the red species, compound 11. Compound 11 is stable in the solid state for up to two months in a glovebox freezer; however, at room temperature and in solution, further oxidation to compound 10 occurs. Whilst 1 H NMR showed similar environments for both 10 and 11 (Fig. S39 †), compound 11 is intensely coloured (UV/vis ¼ 512 nm, 3 ¼ 1155 L mol 1 cm 1 ) whilst 10 is colourless. As such compound 11 was tentatively assigned as a bridging aluminium(II) mono-oxide species, rather than a terminal aluminium(III) mono-oxide complex. Unfortunately, SC-XRD analysis did not provide clear structural parameters for the mono-oxide species due to superposition with compound 10 (Fig. S55 †). To provide further insight, calculations were also performed on compounds 10 and 11. The optimised structure of 10 is symmetric relating to the Al-O bonds, as previously observed for the analogous silyl compound (10 Si ). 16 TD-DFT calculations revealed the highest transition at 259 nm, in line with the experimental colourless appearance. In contrast, compound 11 bears substantial p-electron density between the two aluminium centres in the HOMO as depicted in Fig. 7, reminiscent of the disilaoxirane reported by our group. 53 The LUMO represents the unoccupied p(Al-C NHC ) bond. TD-DFT calculations verifed the experimentally observed red colour (UV-vis 512 nm) with good accordance (calc. 519 nm), assigned to the HOMO to LUMO transitions of 11. Extension of this small molecule reactivity towards dihydrogen was investigated. Firstly, a J-Young NMR tube containing a purple solution of 3 Si was freeze-pump-thaw degassed and then backflled with approximately 1 atm of H 2 . After 24 hours at room temperature no reaction was noted to have occurred; increasing the temperature to 60 C and regular monitoring only resulted in the observed decomposition of 3 Si . Repetition of this reaction with the aryl stabilised dialumene, compound 3, also resulted in no reaction at room temperature. After 16 h at 50 C, however, the black colour of 3 had faded to a dark brown/yellow solution (Scheme 5). On inspection of the 1 H NMR spectrum, no Al-H signals could be observed owing to the quadrupolar nature of the Al centre. Additionally, three distinct iso-propyl signals similar to that observed for compound 1 were identifed. These, however, were not identical and therefore complete hydrogenation and cleavage of the Al-Al bond can be ruled out. Thus, it is likely that compound 12 consists of either terminal or bridging hydrides. IR spectroscopy was utilised to differentiate between the two likely structures; two broad but distinct peaks at 1593 and 1634 cm 1 were observed. Compounds containing no Al-Al bond but both terminal and bridging hydrides are found at approximately 1880 cm 1 and 1350 cm 1 , respectively, 20,54 whilst terminal hydrides within complexes containing Al-Al bonds are found within 1680-1835 cm 1 , 55 thus pointing more in the direction of a terminal hydride with Al-Al bonds. Additionally, for the previous terminal hydride in the related silyl system, from C-H activation of phenyl acetylene, this Al-H was found at 1666 cm 1 . 15 Unfortunately, attempts to grow crystals suitable for SC-XRD were unsuccessful. Therefore, additional insight for this structure was sought computationally. Different possible isomers of product 12 were calculated, including bridging, terminal, and combinations of both as well as different rotational isomers (H: cis or trans; NHC and Tipp ligands: cis or trans). The two lowest lying isomers were found to possess terminal hydrides in the cis and trans confgurations (Fig. 8). The Al-H stretching frequencies were calculated to be 1634 and 1676 cm 1 , respectively, which is in good agreement with the experimentally obtained values (for detailed information see the ESI †). It is, therefore, suggested that the activation of hydrogen by 3 results in both the cis and trans isomers of compound 12. ## Catalysis Further comparisons between the two dialumenes examined their use in catalytic applications. Two archetypal catalytic reactions (hydroboration and dehydrocoupling) were studied due to their prevalence in main group catalysis, as well as the implication of metal-hydrides in facilitating turnover. 56,57 With the ability to form dialuminium-hydrides in the case of 3 and not for 3 Si , differences in activity and mechanistic pathways are anticipated. ## CO 2 hydroboration Previously, 3 Si was found to selectively catalyse the reduction of CO 2 to a formic acid equivalent (product A, Scheme 6) with pinacol borane (HBpin). 16 Whilst this reaction does proceed at room temperature, it required up to 1 week and 10 mol% of 3 Si (Table 1, entry 1); use of higher temperatures allowed for reduced catalyst loadings and decreased reaction times (Table 1, entry 2). As 3 has so far shown increased reactivity, a trial reaction with 5 mol% of 3 towards hydroboration of CO 2 at room temperature was carried out (Table 1, entry 3). On regular monitoring through 1 H and 11 B NMR spectroscopy the consumption of HBpin was noted to occur along with the formation of new B-O containing species. The corresponding 1 H NMR spectrum showed the presence of further reduced species (A-D, Scheme 6), indicating that 3 is not only more catalytically active, but also proceeds through an alternate mechanism due to the presence of B-C. Again, through use of a higher temperature (60 C), the catalyst loading could be decreased down to 1 mol% (Table 1, entry 5). This resulted in the same consumption of HBpin after 24 h (at RT) as with 5 mol% (Table 1, entry 4); however, the resulting product distribution differs, with higher temperatures favouring the formation of the triply reduced methanol equivalent (product C, Scheme 6). The formation of more highly reduced species indicates a likely change in the mechanism. Previously, 3 Si showed no reactivity towards HBpin and as such a non-hydridic mechanism based on the initial formation of 9 Si was proposed. From computational analysis, coordination of HBpin and subsequent reduction of the exocyclic carbonyl of 9 Si was found to be rate determining. Turnover was achieved through coordination/ insertion of an additional CO 2 on the opposite side of the Al/Al plane resulting in formation of an 8-membered ring which collapses to reform 9 Si with release of the formic acid equivalent. This mechanism also further supports the observed selectivity towards product A. 16 In this instance, use of 3 results in the formation of products B-D in notable amounts; therefore, an alternative mechanism for the hydroboration of CO 2 is highly likely. As such, 3 was reacted with 1 eq. of HBpin; the 11 B NMR spectrum showed complete consumption of HBpin and formation of a new upfeld doublet at d 2.57 ppm (J HB ¼ 112.18 Hz). The same signal and coupling were observed from the reaction of HBpin and I i Pr 2 Me 2 ; therefore it is proposed that the stoichiometric reaction of 3 and HBpin results in NHC abstraction. Notably, this does not result in H-B bond cleavage and formation of an Al(H)-Al(Bpin) type species, which was observed with diborene 58 and disilyne 59 chemistry. Addition of CO 2 to this I i Pr 2 Me 2 -HBpin adduct did not result in formation of any reduced CO 2 species, or any reaction after 24 h at room temperature; therefore it is unlikely that this is the catalytically active species. It is of note that NHCs have been shown to catalyse hydroboration (with HBpin) of carbonyl compounds, in acetonitrile. 60 Whilst these experimental observations preclude defnitive mechanistic analysis, it is proposed that the aryl stabilised dialumene (3) acts as a pre-catalyst, with CO 2 hydroboration occurring through an initial hydroalumination of CO 2 and subsequent Al-O/B-H s-bond metathesis, in line with other previously reported main group hydroelementation of CO 2 mechanisms. Amine borane dehydrogenation Main group catalysts (largely group 1, 2, and 13) have also been shown to be viable dehydrocoupling catalysts. 56,57,71,72 These reactions largely proceed through formation of M-H and M-E Scheme 6 Catalytic hydroboration of CO 2 mediated by dialumene (3). 2, entry 5). As aluminium-hydrides have been used in amine borane catalysis previously, 73 and to rule out complete hydrogenation and Al-Al bond cleavage during the reaction, compound 1 was tested for dehydrocoupling activity. After 24 h at room temperature (Table 2, entry 6) no conversion of Me 2 NHBH 3 was observed; on increasing the temperature to 60 C (Table 2, entry 7), some formation of H 2 and products A-D was observed. Due to the increased temperature and prolonged reaction times required for compound 1, it is proposed that the retention of an Al-Al bond accounts for the increased catalytic activity. Comparable to hydroboration reactions, the aryl stabilised dialumene (3) was found to be more catalytically active than the silyl-stabilised counterpart (3 Si ). Mechanistically speaking, amine-borane dehydrocoupling reactions generally occur through formation of M-H and M-E bonds. 57 As such, it has been shown that formation of Al-H bonds is more accessible from 3 compared to 3 Si , therefore accounting for the difference in catalytic activity. Both reactions show initial formation of a catalytic equivalent of HB(NMe 2 ) 2 (product B, Scheme 7) which then remains constant throughout the catalysis. It has previously been shown by Wright and co-workers that formation of B is the result of the formation of the catalytically active Al-H containing species from Al(NR 3 ) 3 (R ¼ Me, iPr). 74 Furthermore, Braunschweig and co-workers recently showed that Me 2 NHBH 3 can be used to isolate hydrogenated diborenes; thus, analogous reactivity is anticipated. 75 However, in our hands, stoichiometric reactions of 3 and Me 2 NHBH 3 resulted in a mixture of species, whilst reaction with a higher number of equivalents of Me 2 NHBH 3 resulted in the dehydrocoupling products. We therefore conclude that the dialumene acts as a pre-catalyst in this reaction and the active catalyst is generated in situ. ## Conclusions In conclusion, we have shown the fundamental differences between aryl and silyl supporting ligands for the stabilisation of dialumenes and their subsequent influence on reactivity. The increased flexibility of the trans-bent and twisted structure for the aryl dialumene (3) enables reactivity with more sterically demanding substrates, and in addition is now able to activate dihydrogen. Further differences are observed in the catalytic ability of the two dialumenes, with the latter exhibiting higher activity. This is likely due to different mechanisms in the catalytic cycle and the ability of the aryl dialumene to stabilise a metal-hydride intermediate in contrast to the silyl ligand. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Dialumenes \u2013 aryl <i>vs.</i> silyl stabilisation for small molecule activation and catalysis", "journal": "Royal Society of Chemistry (RSC)"}

co2_and_water_activation_on_ceria_nanocluster_modified_tio2_rutile_(110)

## Abstract: Surface modification of TiO2 with metal oxide nanoclusters is a strategy for the development of new photocatalyst materials. We have studied modification of the (110) surface of rutile TiO2 with ceria nanoclusters using density functional theory corrected for on-site Coulomb interactions (DFT+U). We focus on the impact of surface modification on key properties governing the performance of photocatalysts, including light absorption, photoexcited charge carrier separation, reducibility and surface reactivity. Our results show that adsorption of the CeO2 nanoclusters, with compositions Ce5O10 and Ce6O12, is favourable at the rutile (110) surface and that the nanocluster-surface composites favour non-stoichiometry in the adsorbed ceria so that reduced Ce ions will be present in the ground state. The presence of reduced Ce ions and low coordinated O sites in the nanocluster lead to the emergence of energy states in the energy gap of the TiO2 host, which potentially enhance the visible light response. We show, through an examination of oxygen vacancy formation, that the composite systems are reducible with moderate energy costs. Photoexcited electrons and holes localize on Ce and O sites of the supported nanoclusters. The interaction of CO2 and H2O is favourable at multiple sites of the reduced CeOx-TiO2 composite surfaces. CO2 adsorbs and activates, while H2O spontaneously dissociates at oxygen vacancy sites. ## INTRODUCTION Since the seminal paper by Fujishima and Honda in 1972, 1 titanium dioxide, TiO2, has remained at the forefront of photocatalysis research due to its abundance, low-cost, nontoxicity and robustness under operating conditions. The large bandgap (>3 eV) means that TiO2 is UV active so that extending light absorption to the visible range is necessary to maximize solar energy absorption for large-scale implementation of photocatalytic technologies. Strategies to induce visible light absorption in TiO2 have included substitutional cation or anion doping at Ti or O sites respectively and co-doping, where multiple dopants are incorporated in the TiO2 host. Doping introduces impurity energy states into the band gap of the TiO2 host, facilitating electronic transitions that have energies in the visible range. However, localized defect states have been shown to act as recombination centres, impeding carrier migration and reducing photocatalytic activity 9,12,13 and practical issues with reproducibility, solubility and stability persist with doped metal oxides. Extending the absorption edge to longer wavelengths should not be the sole research focus as efficient separation of photoexcited charge carriers and molecular activation are essential to the performance of photocatalysts. The development of the dye sensitized solar cell (DSSC) 21 can inspire similar strategies in photocatalysis; dyes simultaneously promote visible light absorption and charge separation but do suffer from degradation. 13 Noble-metal loading of TiO2 has been reported to improve photocatalytic efficiency in the UV and visible through plasmon resonance in the metal. However, the use of precious metals such as Ag, Au and Pt drives up costs. Surface modification of TiO2 with dispersed metal oxide nanoclusters of non-precious metals has been investigated experimentally via chemisorption-calcination cycle (CCC) 27,28 and atomic layer deposition (ALD). 29 These studies reported both band gap reduction and enhanced visible light photocatalytic activity for FeOx-modified TiO2. Photoluminescence spectroscopy revealed that the modification suppressed carrier recombination 27 and the observed red shift was due to cluster derived states above the valence band maximum (VBM) as identified by X-ray photoelectron spectroscopy (XPS) and confirmed with density functional theory (DFT) simulations. 27,28,30 DFT has been used in combination with experiment to examine these and similar systems. 28, Our previous work 28, 31, 34-40, 43, 44 has indicated the potential for metal oxide modifiers to induce a bandgap reduction over bare TiO2. In addition, an enhanced separation of photoexcited electrons and holes has been reported to result from nanocluster surface modification of TiO2. 34,36,37,40,45 We have further highlighted the role of low-coordinated nanocluster metal and/or oxygen sites in trapping and separating charge carriers 34,36,37,40 and this suggests that modification of TiO2 promotes electron and hole separation. Initial work on CO2 activation at metal oxide modified TiO2 has been presented. 31 In this paper, we present a DFT study of TiO2 rutile (110) modified with sub-nm nanoclusters of CeO2, with specific compositions Ce5O10 and Ce6O12. CeO2 is an interesting modifier as Ce 4f states are crucial in optical properties, reducibility and reactivity. 46,47 In particular, the facile conversion between Ce 4+ and Ce 3+ oxidation states has important implications for catalytic performance and metal/CeOx/TiO2 composites with Ce 3+ cations have displayed enhanced activity for the water gas shift (WGS) reaction. 42, We focus in particular on the impact of nanocluster modification on (1) the interfacial atomic structure, (2) the valence or conduction band edges of rutile (110), (3) charge localization after excitation, (4) the reducibility of the composite system and (5) the interaction of feedstock molecules, such as H2O and CO2, with the reduced CeOx-TiO2 composite. Reducibility and interactions with molecules are crucial for catalysis. A more reducible catalyst could use a combination of photocatalytic and thermal catalytic effects. 51 Electrons and holes can be produced by light absorption and thermally produced defects, such as oxygen vacancies, can provide sites for the adsorption and activation of feedstock species. Transition metal oxides have been widely studied as catalysts for the oxygen evolution reaction (OER). 52,53 Nørskov and colleagues studied water oxidation at the rutile TiO2 (110) surface using DFT 54 and found that the most difficult step was the dissociation of a H2O molecule at a vacancy site to form an adsorbed hydroxyl group (OH*). Other DFT studies have examined water adsorption at extended surfaces of both rutile TiO2 (110) 55 and CeO2 (111) 56,57 and found that molecular adsorption of water is more favourable than dissociative adsorption, albeit marginally so for the latter system (0.01-0.03 eV depending on exchange-correlation functional and separated by an energy barrier of 0.1 eV). A recent study compared water adsorption at the (101) (Eads = 0.89 eV) and (001) (Eads = 0.29 eV) surfaces of anatase TiO2. 58 The authors attributed the greater activity of the (001) surface for water oxidation to hydrogen bonds between H2O and terminal hydroxyls which facilitate rapid hole transfer. In general, the adsorption configuration of water molecules at TiO2 surfaces depends on the crystal form, termination, stoichiometry and degree of coverage. 33,59 However, an important first step, consistent across mechanisms describing water oxidation, is the dissociative adsorption of H2O at the catalyst surface. 60-62 DFT+U studies examining ceria nanoclusters, of composition Ce2O3, supported on rutile (110) found that water dissociation is exothermic (Eads = -0.7 eV) with a small energy barrier (0.04 eV). 42,48 In addition, a number of studies have highlighted the role played by dissociated H2O, in particular surface hydroxyls, in trapping holes at catalyst surfaces, 58,61, which is important for subsequent steps in the water oxidation reaction. The CO2 reduction reaction (CO2RR) competes with the hydrogen evolution reaction (HER) in the presence of water and in general H2O adsorption at catalyst surfaces is preferential to CO2 adsorption. 66 It is therefore necessary to promote selectivity for the desired reactions. The adsorption of CO2 molecules at titania surfaces has been studied using DFT and a crucial role has been ascribed to oxygen vacancies in the activation process. 67,68 The presence of excess electrons and holes was shown to affect adsorption and activation of CO2 at rutile (110), with implications for binding energies, structure and reactivity of adsorbed CO2; both bent CO2anion (excess electrons) and CO2 + cation (excess holes) configurations were identified with energy barriers of 1.12 eV and 0.75 eV respectively. 69 Yang and colleagues showed that subnm Pt clusters at the anatase (101) surface enhanced CO2 activation through provision of additional adsorption sites at the edge of the Pt cluster, reporting increased adsorption strength and, in some instances, the spontaneous formation of a CO2anion due to accumulation of negative charge on the C atom. The authors also reported transfer of electron density from the cluster to the TiO2 substrate which facilitated adsorption at surface sites away from the supported Pt octamer. 70 Bismuth pyrochlore oxides have been shown to have a high CO2 chemisorption capacity as identified through FTIR and attributed to the Bi2O3 motif with a Bi 3+ lone pair. 71 Cu2O has emerged as a potential candidate for CO2RR due to its favourable band gap position and width 72 and studies have focussed on the interaction of CO2 molecules with various cuprous oxide surfaces and terminations. Wu and colleagues studied the adsorption of CO2 at the Cu2O (111): O terminated surface with oxygen vacancies, 75 finding that oxygen vacancies have a negative impact on the interaction energy; the most stable adsorption configuration at the perfect surface was more favourable by 0.15 eV than adsorption at the O vacancy surface. However, the authors also reported the formation of a stable CO2 δradical anion species upon adsorption at the O vacancy surface, but this has an adsorption energy of ~0 eV. Another study examined the adsorption of CO2 at Cu2O (111) using Hybrid DFT 77 and found adsorption only in non-activated form. Physisorption of a linear CO2 molecule is favoured over adsorption in activated form at Cu2O (111), 73,76 however strong chemisorption, with energy gains of as much as 1.76 eV, was reported for CO2 adsorption at the Cu-O terminated (110) surface. 73 Activation, with bending and elongation of bonds, upon exothermic (Eads = -0.96 eV) adsorption of CO2 at the (011) surface of CuO has been reported. 74 The conversion of CO2 to methanol on Cu2O nanolayers and clusters 72 was studied using Hybrid DFT with water as the source of H atoms for hydrogenation steps. Other theoretical studies have been conducted into reaction pathways involving the hydrogenation of CO and CO2 at a variety of catalytic surfaces, including Cu/CeO2 and Cu/CeO2/TiO2, 46 Cu/ZnO/Al2O3 78 and copper surfaces. 79,80 Enhanced photoreduction of CO2 with H2O vapour has been reported for dispersed CeO2 on anatase TiO2, prepared using a one-pot hydrothermal method; 47 the role of Ce 3+ in visible light absorption, photogenerated charge separation and strengthening CO2surface bonding was highlighted. The adsorption and activation of the CO2 molecule at the catalyst surface is an important first step in subsequent reactions. In the following we will present results of DFT studies of Ceria nanocluster modified rutile (110). Our clusters have compositions Ce5O10 and Ce6O12 and compliment earlier work on Ce2O3 reduced nanoclusters supported on rutile (110). 42,48,50 These nanoclusters allow us to examine composition effects on stability, band gaps, charge localization and reducibility. We show that adsorption of ceria-nanoclusters at the rutile (110) surface is favourable and that the nanocluster-surface composites favour non-stoichiometry so that reduced Ce ions will be present in the ground state. This off-stoichiometry leads to the emergence of Ce-derived occupied states in the band gap and low coordinated oxygen sites in the supported nanoclusters which contribute to the density of states (DOS) at the TiO2-derived valence band maximum (VBM), potentially increasing the visible light response. Results using a photoexcitation model show that the excited electron and hole prefer to localize onto the supported CeO2 nanocluster. We will also present results which indicate that the interaction of CO2 and H2O is favourable at multiple sites of the reduced CeOx-TiO2 composite surfaces. CO2 adsorbs and activates, forming a bent complex with elongated C-O distances. Finally, H2O spontaneously dissociates at oxygen vacancy sites to generate surface bound hydroxyls. ## METHODOLOGY All calculations were performed using periodic plane wave density functional theory (DFT) as implemented in the VASP5.2 code. 81,82 The valence electrons are described with a plane wave basis set with an energy cut-off of 396 eV. Projector augmented wave (PAW) potentials account for the core-valence electron interaction, 83,84 with 4 valence electrons for Ti, 6 for O, 12 for Ce, 4 for C and 1 for H. The exchange-correlation functional is approximated by the Perdew-Wang (PW91) functional. 85 Γ-point sampling is used for the (2 × 4) surface expansion of the rutile (110) surface with a supercell consisting of 18 monolayers (6 neutral trilayers) and a vacuum gap of 20 . The computed bulk lattice constants of rutile TiO2 are = = 4.64 and = 2.97 . The convergence criteria for the energy and forces are 10 eV and 10 eV respectively. All calculations are spin polarized. To consistently describe the partially filled Ti3d and Ce4f states a Hubbard U correction is applied. 86,87 We use values of U(Ti) = 4.5 eV and U(Ce) = 5.0 eV with these values chosen from previous work on CeO2 and TiO2. 37,38,40,48,88,89 For calculations involving the model excited state and valence band hole formation we apply an additional +U correction to the O2p state with U(O) = 5.5 eV. Previous work has highlighted the necessity for such a correction in obtaining a correctly localized oxygen hole state in metal oxides. 33, 40 The ceria nanoclusters were adsorbed in different adsorption configurations at the rutile (110) surface which, for the purposes of this work, is free of the point defects and surface hydroxyls which can be present on real surfaces. 90,91 The adsorption energies are computed using: where , and are the energies of the adsorbate-surface composite system, the bare rutile (110) surface and the gas phase nanocluster respectively. Once a stable adsorption configuration was identified we examined the energies associated with oxygen vacancy formation in the adsorbed CeO2 nanocluster. One oxygen ion is removed from the adsorbed CeO2 cluster and the vacancy formation energy is calculated as: where the first and third terms on the right hand side of the equation are the total energy of the cluster-surface composite with and without an oxygen vacancy and the energy is referenced to half the total energy for molecular O2. The calculation was performed for each oxygen site of the supported nanoclusters (see Table S1 Supporting Information) to determine the most stable non-stoichiometric composite. Once this is identified, we remove a second and third oxygen atom, as required, to describe situations in which multiple oxygen vacancies are present in the nanocluster (vide infra). The oxidation states were determined through Bader charge analysis 92 and computed spin magnetizations and the corresponding values are quoted in the following sections. We also investigated the adsorption and activation of H2O and CO2 at the CeO2-modified rutile (110) composites, taking into particular account the presence of oxygen vacancies in CeO2. The adsorption energies for the molecules adsorbed at the nanocluster are calculated as: where , and refer to the energies of the molecule and modified surface in interaction, the modified surface and the gas phase molecule (H2O or CO2) respectively. We model photoexcitation by imposing a triplet electronic state on the system. 93 This promotes an electron to the conduction band, with a corresponding hole in the valence band, and enables an evaluation of the energetics and charge localization associated with photoexcitation. The following energies are computed:  The ground state energy of the system, yielding .  A single point energy calculation at the ground state geometry with the triplet state imposed, yielding .  An ionic relaxation of the triplet electronic state which gives . From the results of these calculations we compute: 1. The singlet-triplet vertical excitation energy: This is the difference in energy between the ground (singlet) state and the imposed triplet state at the singlet geometry and corresponds to the simple VB-CB energy gap from the computed density of states. ## The singlet-triplet excitation energy: = − . This is the difference in energy between the relaxed triplet state and the relaxed singlet state and gives a crude approximation of the excitation energy. 3. The triplet relaxation (carrier trapping) energy: This difference in energy between the unrelaxed and relaxed triplet states is the energy gained when the electron and hole are trapped at their metal and oxygen sites upon structural relaxation. This energy relates to the stability of the trapped electron and hole. These quantities are summarized schematically in Figure 1. ## Stoichiometric CeO2-modified TiO2 structures We focus on ceria nanoclusters of two compositions, namely Ce5O10 and Ce6O12, and we first examine the stoichiometric nanocluster adsorption energies and structures shown in Figures 2(a) and 2(d). The adsorption energies were computed using Eq. 1 and are -4.75 eV for Ce5O10 and -2.47 eV for Ce6O12 adsorption on rutile (110). The negative adsorption energies show that the interaction between the nanocluster and the surface is favourable, with the magnitude of the energy indicating the strength of the interaction. The larger adsorption energy for the Ce5O10 nanocluster is reflected in the larger number of surface-to-nanocluster bonds; 7 compared with 6 for the adsorbed Ce6O12 nanocluster. Another reason for the smaller adsorption energy of the Ce6O12 cluster may be due to the presence of under-coordinated terminal oxygen ions and this idea will be expounded upon in the next section. From the adsorption energies we expect the nanoclusters to be stable at the surface without desorbing or migrating over the surface to form aggregates. Henceforth, the composites will be denoted as Ce5Ox-rutile-( 110) and Ce6Ox-rutile- (110), where the subscript x will vary according to the stoichiometry. The interfacial bonding between the nanocluster and the surface results in an appreciable distortion of the local atomic structure at the surface. Where a bridging surface oxygen is bound to a nanocluster cation the Ti-O bond involving this oxygen elongated by up to 10% compared with a typical unmodified bond length of 1.88 . Surface titanium atoms that bind with oxygen in the nanoclusters migrate out of the surface plane towards the cluster by as much as 0.92 , lengthening the subsurface Ti-O distance. ## Reduction of CeO2-rutile by oxygen vacancy formation From our relaxed stoichiometric nanocluster-surface composites we remove oxygen ions from CeO2 and examine the energies involved in the formation of these vacancy sites. Previous work on small CeO2 structures on rutile (110) has shown that these prefer to be reduced, with loss of oxygen in the ground state, giving composition Ce2O3. 46,48 It is not known if a similar composition would be found for larger but still sub-nm ceria clusters. The oxygen vacancy formation energies are important as their stability determines the stoichiometry of the composite. If the composite is then reduced, the formation energy can be a further important factor in determining if feedstock species will interact with the CeOx-rutile composites. If the energy cost to form a reducing vacancy is low, the system favours non-stoichiometry and fixation and activation of molecular species, via a redox or Mars van Krevelen process, may not occur and no reactions can take place. On the other hand, while large vacancy formation energies can promote reoxidation via feedstock reduction, these require a large initial energy input and may also result in too strong interaction with molecular species, leading to poisoning of the surface. Table 1 presents the computed oxygen vacancy formation energies in each supported ceria nanocluster. The most stable oxygen vacancy in Ce5O10-rutile- (110), which results in formation of Ce5O9-rutile- (110), has a small cost of 0.18 eV. A formation energy of this magnitude suggests that an off-stoichiometric ground state will be present. For a second oxygen vacancy, giving a composition Ce5O8-rutile- (110), the most stable vacancy site has an energy cost of 1.48 eV, relative to Ce5O9-rutile- (110). Thus the second oxygen vacancy is the reducing oxygen vacancy and this has a moderate cost. For the Ce6O12-rutile composite, the first two oxygen vacancies have negative formation energies, of -0.46 eV and -0.16 eV, which means that the ground state is highly offstoichiometric as the vacancies will form spontaneously at = 0 K. The ground state therefore has the composition Ce6O10-rutile- (110). This instability of the stoichiometric Ce6O12 nanocluster adsorbed on the rutile (110) surface sheds light on its small adsorption energy relative to the Ce5O10 nanocluster. Computing the adsorption energy of an off-stoichiometric Ce6O10 nanocluster on rutile (110), yields a value of -3.45 eV, suggesting a stronger interaction at the surface and enhanced stability. The energy cost required to produce the third oxygen vacancy in the most stable cluster site of the larger nanocluster-surface composite is +0.30 eV. This is a moderate cost and we consider the Ce6O9-rutile-( 110) composite as being in a reduced state. Thus, for the CeO2-rutile composites, with rutile modified by a sub-nm ceria nanocluster, we expect a highly non-stoichiometric system with multiple potential activation sites at moderate temperatures, consistent with the work of Graciani et al. In these non-stoichiometric nanocluster-surface composites we expect to find two electrons released for each neutral oxygen vacancy and the spin density plots are used to determine the location of the electrons after relaxation. The spin density plots for the ground and reduced states of CeOx-rutile are presented in Figure 3 and show that electron localization occurs at Ce atoms in each nanocluster, which results in the formation of reduced Ce 3+ cations. Ce 3+ form in preference to Ti 3+ cations and this has also been seen in DFT+U studies of Ce-doped TiO2 and some surfaces. 42,50 For the non-stoichiometric ground state structures, the smaller nanocluster has a Ce5O9 configuration and two Ce atoms are reduced as shown in Figure 3 These results are further confirmed through Bader charge analysis, the results of which are included in 2. The spin polarized projected electronic density of states (PEDOS) for the stoichiometric, offstoichiometric ground state and reduced nanocluster-surface composites described above are presented in Figure 4. Panels (a) and (d) of Figure 4 show the stoichiometric configurations where the most obvious feature is the presence of states due to cluster oxygens at the top of the valence band for the Ce6O12 nanocluster. These states are due to the singly coordinated oxygen ions described previously. However, the nanocluster-derived oxygen 2p states above the TiO2 VB persist even after removing these oxygen sites. In panels (b), (c), (e) and (f), which correspond to the off-stoichiometric CeOx-rutile composites, we can see that the presence of reduced Ce 3+ cations in the nanoclusters after formation of oxygen vacancies introduces states into the TiO2-derived band gap. These states arise due to the singly occupied 4f 1 orbital configuration of the reduced Ce 3+ cations. The modification of rutile with CeOx nanoclusters will result in a red shift of the TiO2 adsorption edge; this is due to a combination of 2p states of low coordinated O sites of the cluster pushing the VBM to higher energy and the emergence of mid-gap states associated with reduced Ce 3+ ions in the off-stoichiometric composites. Insets in the top panels show the mid-gap Ce-derived states in the range -0.5 eV -2.0 eV. ## Modelling charge separation upon photoexcitation. We apply the photoexcited model to the ground state CeOx-TiO2 systems which consist of the off-stoichiometric composites, Ce5O9-rutile-( 110) and Ce6O10-rutile- (110). Table 3 presents the computed vertical and singlet triplet energies and the electron-hole localization (relaxation) energies, as discussed in Section 2. Firstly, we note the underestimation of the bandgap inherent in approximate DFT is present in our DFT+U computational set-up. The +U corrections used herein are chosen to consistently describe the localization of electrons and holes rather than to reproduce the bandgap of bulk TiO2, which is not advised. This underestimation is clear in the computed values for E vertical and E excite which are clearly smaller than the experimental values. However, what is important for our study is the change in these quantities with modification of the rutile (110) surface. We note that E excite is always smaller than E vertical and the simple valence-conduction band energy difference, as the former energy includes the ionic relaxations and polaron formation in response to "exciting" the electron which then lowers the energy of the triplet electronic state. Comparison of these computed energies across different structures yields useful qualitative information about the effect of surface modification. In particular, a reduction in E excite for a composite structure relative to the unmodified metal oxide will correspond to a red shift in light absorption for the surface modified system. The energies presented in Table 3 show that modification of the ( 110) surface of rutile TiO2 with nanoclusters of CeO2 leads to a red shift in light absorption whether we use the vertical or excitation energies. This effect is stronger for the larger nanocluster, consistent with the PEDOS. Relaxation energies of 0.8 eV upon charge localization in each heterostructure indicate high stability of the photogenerated electron-hole pairs. We can also examine the localization of the electron-hole pair through analysis of computed Bader charges, spin magnetizations and excess spin density plots. 2, and spin magnetizations of 0.97 µB were computed for these sites. the hole localizes at a singly coordinated terminal oxygen site; the Ce-O distance increases from 1.9 in the ground state to 2.3 after excitation. Hole localization is accompanied by a change in the computed Bader charge of the oxygen by 0.4 electrons, from 7.1 to 6.7 electrons, in each case. For the Ce5O9-rutile-( 110) composite there is some spreading of the hole to neighbouring two-fold coordinated oxygen sites of the nanocluster but this spreading is typical for this DFT+U set-up and is accompanied by changes of < 0.1 electrons in the computed Bader charges, so that we can conclude the hole predominantly localizes on one oxygen site in the nanocluster. This is confirmed by a computed spin magnetization of 0.73 µB for the oxygen hole on Ce5O9-rutile-( 110) and compares with a value of 0.78 µB for the singly terminated oxygen site at which the hole localizes in Ce6O10-rutile- (110). For the CeOx-rutile-( 110) composites we can see that both the electron and hole localize on the nanocluster modifiers, which may have consequences for recombination. However, in looking at Figure 5 we see that the spatial separation of the charges is maximal, given that both electrons and holes localize at nanocluster sites. In addition, the large relaxation or trapping energies act to impede migration of the charges and thus the impact on recombination should be minor. We also note that our photoexcited model, which involves the imposition of a triplet state to induce a transition from the VB to the CB, precludes transitions from the highest occupied, Ce 4fderived states of the off-stoichiometric ground states (see Figures 4(b) and 4(e)). Such transitions would amount to electron hopping between Ce sites of the nanocluster with no change in electronic configuration after "excitation". Rather, our model with a triplet electronic state (in addition to the unpaired electrons on reduced Ce 3+ ) will induce transitions from OC 2p-derived states, which sit at the top of the titania-derived VB, to the unoccupied Ce 4f states. ## CO2 adsorption at the reduced nanocluster-surface composites. With the motivation that O vacancies in reduced metal oxides can act as sites for the adsorption and activation of CO2, as found in other studies, 75, we have examined the adsorption of CO2 at various sites on our reduced Ce5O8-and Ce6O9-rutile-( 110)); the computed adsorption energies for the most stable configurations are -1.36 eV on Ce5O8-rutile-(110) and -1.09 eV on Ce6O9-rutile- (110). The relaxed geometries for these configurations are shown in Figure 6. The Supporting Information shows additional adsorption structures and energies for the CO2-CeOx-TiO2 interaction. The most stable CO2 adsorption sites have negative adsorption energies and the magnitudes of these adsorption energies are indicative of a strong exothermic interaction between CO2 and the oxides. Our results follow the trend that CO2 interaction is stronger at vacancy sites with a higher formation energy, as previously reported. 31 A DFT+U study of CO2 activation on CeO2 (110) found that the most stable O vacancy had a formation energy of +1.65 eV; 101 the authors reported that CO2 adsorption at this site, with a bent geometry and Eads = -1.22 eV, was the most stable adsorption configuration. A DFT+U study of CO2 reduction on CeO2 (111) also found that interaction was strongest at the O-defective surface; 102 the O vacancy formation energy was +2.78 eV and CO2 adsorbed in a bent geometry with Eads = -1.12 eV. While the adsorption energies are comparable across these studies, the vacancy formation energy is not sufficient in predicting the strength of interaction of adsorbed CO2. Previous work on Ce3O6rutile-( 110) 31 found that this composite was reducible with an O vacancy formation energy of +0.31 eV; CO2 was calculated to adsorb exothermically with Eads = -0.20 eV. This compares with an adsorption energy of -1.09 eV for CO2 interacting at the reduced Ce6O9-rutile- (110) composite of the present work in which oxygen vacancies are produced with similar energy costs (see Table 1). This would suggest that, in the case of ceria nanocluster modifiers, the degree of non-stoichiometry, which is related to the number of reduced Ce cations in the cluster, may play a role in stabilizing adsorbed surface species. Future work, involving larger ceria nanocluster modifiers, can shed further light on the nature of the trade-off between interaction strength and reducibility. This means that the interaction leads to an activated, chemisorbed CO2 species and the formation of a carbonate species can be ruled out. The Ce-O bonds established with the adsorbed species are comparable in length to those within the nanocluster which are typically in the range of 2.2-2.6 . The calculated Bader charges show that charge transfer is qualitatively consistent across the various adsorption sites. Between 0.1 and 0.2 electrons are transferred from the CO2-derived O atoms to the cluster while the cluster oxygen site with which the CO2 interacts gains between 0.4 and 0.6 electrons through the interaction. ## H2O adsorption at the reduced nanocluster surface composites. We also examined how water interacts at the vacancy sites in the reduced CeOx-rutile (110) composites. We compute the adsorption energies of H2O interacting at a range of sites of the reduced Ce5O8-and Ce6O9-rutile-(110) surfaces using Eq. 2. We find that adsorption of water is favourable at multiple sites on the CeOx nanocluster. Adsorption energies for the most stable adsorption configurations of water are -1.8 eV on Ce5O8-rutile-( 110) and -0.9 eV on Ce6O9rutile- (110); the corresponding geometries displayed in Figure 7. The Supporting Information shows additional adsorption sites and energies for the H2O-CeOx-TiO2 interaction. In the interaction of H2O at the reduced composites, starting from an initial water adsorption in molecular form, the most stable adsorption mode is that in which the water molecule dissociates spontaneously upon relaxation. This dissociation involves the transfer of an H atom to an O site of the supported nanocluster and the hydroxyl from the water molecule bridges two cluster Ce sites. On the larger CeOx nanocluster, the moderate adsorption energy means that hydroxyls should not be overstabilized and could be active in catalysis. Thus, water dissociation and activation can be promoted on these ceria-rutile composites. Despite this distortion of the larger nanocluster upon H2O adsorption, the interaction is strong and favourable as shown by an adsorption energy of -0.9 eV. Similarly to the smaller nanocluster, there is a redistribution of charge with water oxygen transferring 0.3 electrons to the nanocluster and this charge is donated to the nanocluster oxygen that binds with hydrogen from water. These results compare with studies of water dissociation at Ce2O3-TiO2. 48,106 In these studies the authors followed the energy pathway from water adsorbed in molecular form to dissociation, finding that the dissociation process was exothermic (-0.70 eV) with a small energy barrier of 0.04 eV. We found that dissociation of molecular water occurred spontaneously, suggesting that the size of the supported CeOx nanocluster and the number of Ce 3+ sites play a role in the ability of the composite to dissociate water. While the ability of metal oxides to dissociate H2O is well established, the mechanism which promotes dissociation upon adsorption remains of interest. A number of studies have looked at CeO2 surfaces as model systems for the study of water dissociation. Defects, step edges and terraces in surfaces play a role as such features provide low-coordinated adsorption sites. CeO2 (111) with O vacancies and Ce 3+ ions shows a preference for dissociative water adsorption, relative to the pristine surface, where there is little energetic difference between adsorption in molecular and dissociated form. 111 surface promotes dissociation of H2O over molecular adsorption. 113 For the reduced Ce6O9rutile-(110) composite, the Ce-O distances, at the sites of H2O adsorption (see Figures 7(c) and (d)), are longer by ~1% relative to typical distances (~2.37 ) in the pristine CeO2 (111) surface. This suggests that tensile strain may indeed contribute to promoting the dissociation of water. In Ce5O8-rutile-( 110), Ce-O distances are shorter (~2.2 ), due to the lower coordination of the cluster O sites, and elongate after the dissociative adsorption of H2O. However, the Ce-Ce distance prior to water adsorption is 4.2 , which is considerably longer than neighbouring Ce-Ce distances (~3.9 ) in CeO2 (111). After the dissociative adsorption of water, this Ce-Ce distance decreases to 3.6 , further indicating that tensile strain may play a role in driving the dissociation. Figure 8 shows the PEDOS of the H2O molecule and reduced CeOx-rutile-( 110) composites in the non-interacting case (H2O + surface) and after dissociative adsorption (H2O-surface). For the non-interacting systems the molecule and surface are relaxed in the same unit cell with sufficient spatial separation such that they do not interact. In the non-interacting cases (left panels of Figure 8), the water-derived OW 2p states are well defined peaks at energies of -2.9 eV and -0.8 eV (Figure 8 Despite these differences, some trends are consistent across both composites. In both cases the interaction increases the gap between the occupied Ce 4f-derived states and the CBM of the TiO2 host (see insets of panels in Figure 8); i.e. the occupied Ce 3+ states are pushed to lower energy after interaction. In addition, integrating the OC and OW-derived DOS lying above the TiO2 VBM in both the non-interacting and interacting cases shows that after interaction the occupied states are driven to lower energies. For both composites the number of states lying above the TiO2 VBM is reduced by 2 in the interacting cases relative to the non-interacting systems; this suggests that passivation of high lying O 2p states is a factor driving the interaction of water with the reduced CeOx-rutile-(110) composite surfaces. ## CONCLUSIONS We have studied the (110) surface of rutile TiO2 modified with ceria nanoclusters of compositions Ce5O10 and Ce6O12 using first principles DFT+U analysis. Our results show that the ground state of the nanocluster-surface composites is off-stoichiometric with one or more oxygen vacancies forming spontaneously or at very low energy cost, so that under typical experimental conditions, there will be oxygen vacancies present. The consequence of this is that Ce 3+ ions will be present in the nanoclusters in their ground state (with no Ti 3+ species) and this leads to the emergence of occupied Ce 4f-derived states in the TiO2-derived bandgap. Together with CeOx-derived O 2p states above the TiO2 VB, this may induce a red shift in light absorption making these systems visible light active. It is low-coordinated oxygen atoms in the supported nanoclusters that contribute to these new states above the valence band edge. In our model of the photoexcited state, in which an electron is promoted from the O 2p derived valence band, we found that electron and hole localization occur at Ce and low-coordinated oxygen sites on the supported nanocluster respectively. The consequence of this for charge recombination may not be detrimental as the electron-hole pair has a large trapping energy of 0.8 eV so this can reduce the migration of charges over the nanocluster. Verification of both the predicted red shift and the charge recombination effects would be welcome. In terms of activity, the CeO2-rutile composites are more reducible compared to the unmodified rutile (110) surface and moderate energy inputs are required to produce multiple oxygen vacancies. Electrons released after forming the oxygen vacancies localize on Ce sites in the supported nanoclusters. We have examined the interaction of oxygen vacancies in the reduced composites with CO2 and H2O to determine how CeOx-rutile-(110) activates these molecules. We find that CO2 adsorption is favourable at multiple sites on the nanocluster-modified surface, with exothermic adsorption energies up to 1.36 eV. This strong adsorption is accompanied by a distortion from the linear gas phase CO2 geometry, in which the molecule bends, with O-C-O angles of 125⁰-128⁰, and the C-O bonds in CO2 show an elongation of ~0.10 . In combination with some transfer of charge between the adsorbed species and the nanocluster, this suggests the formation of activated CO2 which is the crucial first step in the transformation of CO2 to more useful molecules. We welcome attempts to study these materials for their ability to activate and convert CO2, although the actual conversion process may be limited by other factors. Nevertheless, the finding that these ceria-modified TiO2 systems can activate CO2 is a promising first step. Finally, the interaction of H2O at the ceria-modified rutile composites was investigated. This is important for a number of reactions, such as water gas shift or water oxidation and one of the limiting steps in these reactions is water dissociation which usually has an energy cost and an activation barrier. On both reduced ceria-TiO2 systems, water adsorption is exothermic and favourable and, importantly, this leads to spontaneous dissociation of water to form surface bound hydroxyls. The results of this paper show that ceria-modified rutile TiO2 composites can (1) have reduced Ce 3+ cations, (2) show red shift in light absorption, (3) adsorb and activate carbon dioxide and (4) adsorb and activate water. This makes these composites interesting materials for the activation and conversion of CO2 and water.

chemsum

{"title": "CO2 and Water Activation on Ceria Nanocluster Modified TiO2 Rutile (110)", "journal": "ChemRxiv"}

probing_protein_shelf_lives_from_inverse_mean_first_passage_times

## Abstract: Protein aggregation is investigated theoretically via protein turnover, misfolding, aggregation and degradation. The Mean First Passage Time (MFPT) of aggregation is evaluated within the framework of Chemical Master Equation (CME) and pseudo first order kinetics with appropriate boundary conditions. The rate constants of aggregation of different proteins are calculated from the inverse MFPT, which show an excellent match with the experimentally reported rate constants and those extracted from the ThT/ThS fluorescence data. Protein aggregation is found to be practically independent of the number of contacts and the critical number of misfolded contacts. The age of appearance of aggregation-related diseases is obtained from the survival probability and the MFPT results, which matches with those reported in the literature. The calculated survival probability is in good agreement with the only available clinical data for Parkinson's disease. Most proteins have been evolved to spontaneously fold to their native states, which determine their functional specificity and diversity. 1,2 Any phenotypic or genotypic variations may induce abnormal amino acid modifications and cause protein misfolding. Misfolded proteins disrupt normal cellular functions and may be potentially toxic. 6 The spontaneous self-assembly of misfolded proteins often lead to the formation of aggregates, which are associated with a wide variety of debilitating disorders like Alzheimer's, Parkinson's, Creutzfeldt-Jakob's, Huntington's, Amyotrophic lateral sclerosis (ALS) and dementia. 2, The Protein Quality Control (PQC) system present in the cell manages these misfolded proteins and helps them to either refold back to their respective native conformations via chaperones or degrades them to amino acids and eventually replaces them with their newly synthesized replicas. 10,11 This phenomenon known as protein turnover, is a highly specific and precisely regulated process that involves a constant renewal of the functional proteins by allowing the damaged or non-functional ones to be eliminated from the cell. 10 The underlying link among protein folding, misfolding, aggregation and degradation equilibria implies that a change in any one of these components would directly/indirectly affect the others. 12 External factors like aging, genetic mutation, oxidative stress, pH and temperature results in the failure of the protein turnover process and leads to the formation of aggregates/fibrils. 13,14 These aggregates are typically highly organized hydrogen-bonded structures that are more stable compared to the native protein, 6,7 kinetically-trapped in the lowest free energy state. Thus once formed such aggregates are extremely stable for long time periods and acts as a nucleus for further propagation. This work analyzes the folding outcome of a protein through protein turnover followed by misfolding, aggregation and degradation. The rate of formation of proteins from the amino acids follows a zero-order kinetics, 15,16 which is an input for the subsequent P ⇀ ↽ M equilibrium, that is governed by the time evolution of the misfolded contacts. The Chemical Master Equation (CME) for this equilibria is derived from the splitting probabilities of the misfolded contacts at a particular time instant. The misfolded proteins self-associate to form aggregates as described by a first order differential equation. The Mean First Passage Time (MFPT) required for the protein to form aggregates from the misfolded proteins is calculated from both CME and the first order differential equation under appropriate boundary conditions. The rate constants of aggregation of different disease causing proteins are evaluated from the inverse MFPT, which show an excellent match with the experimentally reported rate constants and those extracted from the ThT/ThS fluorescence data. The age of appearance of these diseases are directly evaluated from the MFPT and the survival probability results, which agrees well with those reported in literature. The survival probability result is in good agreement with the only available clinical data for Parkinson's disease. begins from a pool of amino acids via protein turnover. 24 The synthesized native proteins may misfold and the misfolded proteins subsequently self assemble to form aggregates. 8,10,12 Both misfolded proteins and aggregates may degrade to by-products, which is eliminated from the system. 10 The misfolded state represents the ensemble of misfolded proteins, where each one is characterized by a critical number of misfolded contacts, q M C . For a given protein, all chains in the native conformational ensemble are assumed to be of equal lengths with equal number of contacts that are in equilibrium with the misfolded state. The number of proteins, n, present at time t may be calculated assuming zero order kinetics 15,16 with the rate constant k n . The solution of this rate equation is n = k n t. The total number of contacts present at time t is given by: q(t) = n P n = n P k n t, where n P is the number of contacts present in each protein. The number of misfolded contacts present in the native protein at time t + ∆t is q M (t). The protein acquires a misfolded conformation M at time t M when the number of misfolded contacts reaches a critical value, q M C . The rate of increase/decrease of a misfolded contact at an infinitesimal time interval, ∆t, may be given by rate(q M (t)→q where, k pm denotes the rate constant for the conversion of a native contact into a misfolded one, while k mp is the rate constant for the backward reaction. The transition probabilities for the gain and loss of a misfolded contact at time ∆t are represented as W (q M (t) + 1, q M (t))∆t and W (q M (t) − 1, q M (t))∆t respectively. Thus the probability to remain in a given misfolded state with q M (t) misfolded contacts at time ∆t is 25 Thus the probability, P (M, t M | q M , t) to acquire the misfolded conformation, M , at time t M may be expressed as a difference equation. 25,26 P (M, The Chemical Master Equation (CME) 25, for the native conformational ensemble may be obtained from Eqn (1) in the limit ∆t→0 as The probability P (M, t M | q M , t) follows the reflecting boundary condition for the number of misfolded contacts, q M (t) < q M C . The MFPT may be obtained from Eqn (2) as (refer to the Supporting Information (SI)) The equation 25,29 holds true for all values of q M (t) ranging from 1 to q M C . Since τ (0) = 0 and τ (q M C + 1) is not required, this equation may be solved to obtain the MFPT, τ M , of the misfolded proteins in terms of the Gauss hypergeometric function 2 F 1 (α, β; γ; z) (refer to SI). where r = k mp /k pm < 1. 30 The generalized equation of MFPT is simplified using the integral identity as 26 Degradation of the misfolded proteins follow first order kinetics. 15,16,31 The rate equation for degradation may be defined in terms of the evolution of q M (t) with time as: where k d is the rate constant for the degradation of misfolded proteins calculated from the half-life 31 of a protein as, k d = 0.693/t 1/2 . This first order differential equation may be solved as Protein aggregation may be viewed as the self-assembly of misfolded proteins. 17,32 The rate equation for aggregation followed by degradation of the aggregates is given by (refer to where, n As and n M are the number of aggregates and the number of misfolded proteins present in an aggregate respectively. The ratio R is defined as R = n As /n M . k agg denotes the pseudo first order aggregation rate constant, whereas k da is the degradation rate constant of the aggregates. Eqn (7) may be solved by using absorbing boundary condition defined by n As ; t = τ agg ; absorbing boundary condition The solution of Eqn ( 7) is given for n As as The time required for the aggregation of misfolded proteins, τ agg , may be obtained by rearranging Eqn (8) as where the ratio of rate constants, K is defined as K = k da /k agg . Thus the MFPT of aggregation may be expressed as 26,33 and half-lives for a specified value of the rate constant k da = 10 −3 k agg . Inset figure depicts the MFPT at initial times. with time followed by the formation of aggregates. The MFPT of each protein reaches a plateau with time marking the age of appearance of the aggregation-related diseases. The MFPT of the selected proteins are calculated from Eqn (10) using the respective values of the reported rate constants 26,33 and half-lives as listed in Table 1. The MFPT of the aggregate remains constant for fixed values of R for a given n As . This affirms that protein aggregation is independent of the number of aggregates, n As for fixed values of k n , k pm , k mp , k agg , t 1/2 and R. To the best of our knowledge there are no reported literature values of the rate constants or half-lives for the degradation of aggregates. The rate of degradation of these aggregates is much slower compared to the rate of their formation, as the aggregated proteins are very stable. 6,7 For the given range of K = 10 −1 − 10 −5 , the values of R are tuned to match the MFPT with the age of appearance of aggregation-related diseases as given in S2 of SI. The rate constant of aggregation is proportional to the inverse MFPT, which may be calculated as where C is the proportionality constant equal to R. Thus, Eqn ( 11) is recast as Table 1 displays the values of MFPT of the selected proteins by varying K and R. Table 1 also shows a comparison between the calculated and experimental values of the rate constants of aggregation of these proteins. The calculated values of k agg show an excellent match with the rate constants extracted from the ThT/ThS fluorescence data (refer to Figures S1(a), (b) and (c) of SI) and those obtained from experiments. Protein aggregation is found to be practically independent of the number of contacts (n P ) and the critical number of misfolded contacts (q M C ) (refer to SI). The MFPT is independent of the rate constant, k mp for fixed values of k n , k pm and k agg . 26 The survival probability is calculated by assuming that the distribution of proteins in the conformational ensemble is Gaussian 26 at time t. The average number of proteins at an infinitesimal time interval ∆t may be estimated as The survival probability of the proteins is given by where σ 2 is the variance of the Gaussian distribution. Table 1: MFPT of the selected proteins (calculated from Eqn (10)) and a comparison of the experimentally obtained rate constants of aggregation, k agg with those calculated from our theory for k mp 26 = 10 −12 s −1 . Proteins Diseases Figure 3(a) shows the survival probability of the selected aggregation-prone proteins for reported values of the rate constants 26,33 and half-lives. All proteins are initially present in their respective native states. Thus, the survival probability of these proteins shows a maximum that remains constant upto a threshold time, after which it exhibits a slow decrease with time due to the initiation of misfolding. The survival probability decreases monotonically with time and reaches zero after a long time, marking the formation of aggregates. The zero value of the survival probability corresponds to the age of appearance of the aggregation-related diseases. Figure 3(b) displays a comparison of the survival probability obtained from our theory with the only available clinical data of Killinger et al. 23 for Parkinson's disease. The calculated survival probability is in good agreement with this clinical data. 23 Table 2 provides a comparison of the age of appearance of the aggregationrelated diseases from our results of MFPT and survival probability with the respective values reported in the literature. In this work, protein aggregation is investigated theoretically via protein turnover, misfolding, aggregation and degradation. The rate of formation of proteins in turnover follows a zero-order kinetics, which is used in the P ⇀ ↽ M equilibrium, that is governed by time evolution of the misfolded contacts. The Chemical Master Equation for this equilibria is derived from the splitting probabilities of the misfolded contacts at a particular instant of Corresponding Author: *E-mail: pbiswas@chemistry.du.ac.in

chemsum

{"title": "Probing Protein Shelf Lives from Inverse Mean First Passage Times", "journal": "ChemRxiv"}

from_machine_learning_to_transfer_learning_in_laser-induced_breakdown_spectroscopy_analysis_of_rocks

## Abstract: surface asperity on the hydrogen emission line has been investigated 11 . Our recently published work 12 observed and analyzed the performance of a machine learning-based model 13 , trained with a set of pressed rock powder pellets for total alkali-silica (TAS) classification 14 of rocks in their natural state. A significant degradation of the model prediction performance compared to the prediction for pellet samples has been observed. Such degradation prevents the models trained with laboratory standards from reliable predictions with LIBS spectra acquired on raw rock samples, a situation that can lead to misinterpretations for in situ LIBS analysis of rocks on Mars, since we are not yet able to bring materials back from Mars.In order to search a solution for the issue raised, this work introduced transfer learning in LIBS spectral data treatment to more specifically overcome the physical matrix effect. Transfer learning is considered in machine learning when the knowledge gained while solving one problem is required to be applied to a different but related problem 15 . Its necessity comes from the fact that a major assumption in machine learning data processing is that the training and the model-targeted samples to be analyzed should share the same feature space and have the same distribution 16 . It is unfortunately not the case for the application scenario that we consider. Moreover, transfer learning has recently emerged as a new learning framework to address the problem of insufficient training data in an application (target domain) with the help of the knowledge learnt from a related application having the capability to get sufficient training data (source domain) 17 . Such strategy fits well the requirement of LIBS analysis of rocks on Mars, where sufficient laboratory standards can be prepared as the source domain, whereas real Mars rock samples are not yet available as the target domain. Simulation of their chemical as well as physical properties by terrestrial materials, whether natural or artificial, appears therefore a suitable solution. According to the specific contents of the "knowledge" to be transferred, we can distinguish feature-representation-transfer, where parts of relevant features respectively from the both source and target domains are merged and selected for their low sensitivity to the difference between the two domains, to form a common set of features contributing to the training of a transfer learning model 18,19 . Instance-transfer is another specificity of transfer learning where data of the samples from the both source and target domains participate in the model training, with a conditional testing on the relevance of each sample from the source domain for its effectiveness in improving the performance of the model in a cross-validation process with the data from the target domain 18,19 . A weight is then applied to each source domain sample participating the training, according to its contribution in improving the performance of the model for predicting with target domain data. We note that algorithms belonging to transfer learning, low rank alignment of manifolds or feature-based transfer learning for example, have been used respectively for calibration transfers between different LIBS instruments 20 or metallic samples with different temperatures 21 .More specifically, in our experiment, on the basis of the LIBS spectra acquired from a set of laboratory standard samples in the form of pressed powder pellet, machine learning-based multivariate models were trained, validated and then used to predict the concentrations of major oxides necessary for TAS classification of rocks, SiO 2 , Na 2 O and K 2 O, with LIBS spectra acquired from natural rocks. The purpose was first to observe the physical matrix effect due to the difference in surface states between pressed powder pellets and rocks. Since for a LIBS measurement, such difference can be in particular due to the surface hardness, heterogeneity or roughness of a rock, the rock was thus measured in its raw state and with a polished surface, in such way that the different contributions to the physical matrix effect can be investigated separately. Transfer learning-based models were trained with the implementation of feature-representation-transfer and instance-transfer to effectively correct the physical matrix effect in the concentration prediction for rocks in their raw state or prepared with a polished surface, allowing their satisfactory TAS classifications. The correct TAS classification rate increases from 25% for polished rocks and 33.3% for raw rocks with a machine learning model, to 83.3% with a transfer learning model for the both types of rock samples. Although the matrix effects correspond to a general issue in LIBS, they become accentuated in the case of rock analysis for Mars exploration, because of the large variation of rock compositions leading to the chemical matrix effect, and the difference in surface physical properties between laboratory standards (in pressed powder pellet, glass or ceramic) used to establish calibration models and natural rocks encountered on Mars, leading to the physical matrix effect. The chemical matrix effect has been tackled in the ChemCam project with large sets of laboratory standards offering a good representation of various compositions of Mars rocks. The present work more specifically deals with the physical matrix effect which is still lacking a satisfactory solution. The approach consists in introducing transfer learning in LIBS data treatment. For the specific application of total alkali-silica (TAS) classification of rocks (either with a polished surface or in the raw state), the results show a significant improvement in the ability to predict of pellet-based models when trained together with suitable information from rocks in a procedure of transfer learning. The correct TAS classification rate increases from 25% for polished rocks and 33.3% for raw rocks with a machine learning model, to 83.3% with a transfer learning model for both types of rock samples. It is generally considered that the matrix effects, both the chemical 1 and the physical 2 matrix effects, represent a critical issue in analysis with laser-induced breakdown spectroscopy (LIBS) for either qualitative classification or quantitative determination 3 . Suitable solutions with respect to such consideration become paramount for applications as important as LIBS analysis of rocks in Mars explorations 4 . The scientific goals, searching for present and past water activities and the traces of the life as well as studying the habitability of Mars , rely, at least partially, on the reliability and the accuracy of the analytical data that one can extract from the LIBS spectra recorded by LIBS instruments embarked on Mars rovers 8 . The diversity of chemical compositions of Mars rocks has been studied in previous missions, the absence of real sample from Mars, except meteorites, requires a large number of laboratory rock standard samples in order to cover the expected chemical variety of Mars rocks. It was the purpose of the sets of laboratory standard rock samples prepared and used by the ChemCam team for training and validation of the Mars LIBS spectral data processing models. The number of the involved samples was first 69 9 , and was further increased to 408 in order to offer a more complete representation of the chemical and mineral compositions of Mars rocks 10 . It is important to point out that all the above mentioned laboratory rock standards were prepared in the forms of pressed powder pellet, glass, or ceramic to minimize the heterogeneity and the surface roughness of the samples in the scale of LIBS observations of typically several hundred μm. Such sample preparation leads to obvious differences in surface physical properties between laboratory standards and real rocks analyzed by LIBS instruments on Mars. From these differences, changes in the spectra can rise (physical matrix effect) which can impact the analytical results. With this concern, the effects of sample ## OPEN School of Physics and Astronomy, Shanghai Jiao Tong University, Shanghai 200240, People's Republic of China. * email: jin.yu@sjtu.edu.cn ## Samples, experimental setup and measurement protocol Samples. In this work, 20 natural terrestrial rocks were used as samples for LIBS analysis. The rocks were first washed using alcohol and distilled water before any further treatment. All the rocks were prepared in 3 different forms. Raw rocks: LIBS measurements took place on the natural surface of each rock; Polished rocks: LIBS measurements took place on a polished flat surface of each rock (prepared using a 300-mesh sandpaper); Pellets: a part of each rock was crushed and ground into a powder by a laboratory mill and then sieved by a 300-mesh screen (grain size < 50 μm). A binder (microcrystalline cellulose powder) with a similar particle size was mixed into the rock powder at a weight ratio of 20%. One gram of the obtained mixture powder was pressed under a pressure of 850 MPa for 30 min to form a pellet of 15 mm diameter and 2 mm thickness. The composition, with especially the concentrations of major oxides, SiO 2 , Na 2 O, K 2 O, of each rock was determined with X-ray fluorescence spectroscopy (XRF) performed on the pellets with large enough analyzed area to get their bulk composition. The detailed compositions and the geological names of the rocks are presented in the section "Methods" (Table 4), which allows presenting the rocks in a TAS diagram as shown in Fig. 1. The short notations of the 15 fields (surrounded by circles) are according to Reference 22 In order to better simulate an application where the samples to be characterized are not available as in the case of Mars exploration, we first isolated 2 samples (S2 and S5) for model validation, they were excluded from the model training processes. Although these 2 samples were randomly selected as typical isolated validation samples, later, for the model performance evaluation, we will involve other pairs of validation samples excluded from the model training process in order to obtain average model performances independent on the choice of validation samples. The corresponding pellets of these 2 samples were used to validate the machine learning model trained using the rest 18 pellets, while the rock forms of these 2 samples joined the rock validation samples in the validation of the transfer learning models, without counterpart pellet in the training sample set. Among the 18 remaining samples, 8 rocks were selected as training samples (S3, S7, S8, S11, S13, S14, S18 and S19). They joined the 18 pellet samples in the training process of the transfer learning models. The rest 10 rocks, together with the above 2 isolated rocks, ensured the validation of the transfer models. In Fig. 1, the 2 isolated samples are shown in green stars, the 10 additional validation samples in blue dots, and the 8 training rocks samples in red crosses. ## Experimental setup. A detailed description of the used experimental setup can be found elsewhere 12 . Briefly as shown in Fig. 2, a Q-switched Nd:YAG laser operated at a wavelength of 1064 nm with a pulse duration of 7 ns and a repetition of rate of 10 Hz, was used to ablate the samples with a pulse energy of 8 mJ. A lens of 50 mm focal length focused laser pulses about 0.86 mm below the surface of a sample. The diameter of the www.nature.com/scientificreports/ laser spot on the sample surface was estimated to 150 μm, leading to a laser fluence on the sample surface of about 45 J/cm 2 , or an irradiance of about 6.5 GW/cm 2 . The emission from a generated plasma was collected by a combination of two quartz lenses with a same focal length of 75 mm into an optical fiber of 50 μm core diameter. The output of the fiber was connected to the entrance of an echelle spectrometer equipped with an ICCD camera (Mechelle 5000 and iStar, Andor Technology) which provided a wide spectral range from 230 to 900 nm with spectral resolution power of 5000. The ICCD camera was triggered by laser pulses and set with a delay and a gate width of respectively 500 ns and 2000 ns. A lateral CCD camera (not shown in the figure) allowed capturing time-integrated plasma images as shown in the inset of Fig. 2. Samples were mounted on a 3D translation stage allowing recording replicate spectra on a sample surface with an ablation crater matrix, while keeping a constant distance between the focusing lens and the sample surface (approximately for a raw rock). We can see in Table 1 that the mean intensities exhibit different values for different sample forms even though the same sample were concerned, which correspond to biases introduced by the physical matrix effect. In addition, for a given emission line, the RSD value generally increases from the pellet to the corresponding polished and raw rocks. This observation indicates that starting from an initial spectral intensity fluctuation of a pellet sample, the fluctuation increases for the corresponding polished and raw rocks due to heterogeneity and then surface asperity. The most important information from the table is that the RSDs calculated over all the 3 sample forms are significantly larger than the values calculated for a given sample form. This means that the physical matrix effect due to change of sample form represents the dominant variability of spectral intensity, much more important than usual fluctuations observed when measuring a heterogeneous material such as a rock, confirming the above observation of the biases on the mean intensities. ## Data treatment method The general data treatment flowchart used in this work is shown in Fig. 3. It was respectively applied to pairs of sample types, pellets/polished rocks and pellets/raw rocks. Several steps can be distinguished: data pretreatment, feature selection, machine learning (ML) and transfer learning (TL) model trainings, as well as model validation. Such flowchart allowed a comparative study between the performances of a machine learning (ML) model and those of a transfer learning (TL) model. As mentioned above, for the machine learning model, the 18 training pellets were used as the training samples set. The resulted model was validated by the 2 isolated pellets as well as the 12 validation rocks including the 2 isolated rocks without counterpart pellet in the training sample set. For the transfer learning models, the training sample set was composed by the 18 training pellets and the 8 training rocks. The resulted models were validated by the 12 validation rocks including the 2 isolated rocks without counterpart pellet in the training sample set. ## Data pretreatment. The pretreatment consisted in the following operations. (i) Average in order to reduce experimental fluctuations and the influence of sample heterogeneity: For each sample, the 50 raw replicate spectra on each sample were averaged in a procedure where an averaged spectrum was calculated with a first group of randomly selected 30 spectra. The remaining 20 spectra then replaced one by one, a spectrum in the first group, each time the new group of 30 spectra was averaged to generate 20 other average spectra. 21 average spectra were generated for each sample. (ii) Baseline correction: an average spectrum was decomposed into a set of cubic splines of undecimated wavelet scales, the local minima were found, then the spline function was interpolated through the different minima to construct the spectral baseline which was removed 23 . (iii) Normalization: baseline-corrected average spectra were normalized with their respective total intensity (the area under the spectrum). (iv) Standardization: standard normal variate (SNV) transformation was respectively applied to the normalized and baseline-corrected average spectra of the training set of the pellet samples (18 × 21 = 378 spectra) and the training set of the rock samples (8 × 21 = 168 spectra). Within a given sample set, for each channel in a spectrum (22,161 channels in total), the variation range of the intensity value over all the samples was transformed into a range with a mean value equal to 0 and a standard derivation (SD) equal to 1. The parameters determined in the standardization of the training sets of the pellet and rock samples (the means and the SDs) were respectively applied to the 2 isolated pellet samples (2 × 21 = 42 spectra) and the validation rock samples (12 × 21 = 252 spectra) by assuming a same statistical distribution of the data for all the pellets or rock samples. The ensemble of the above operations generated pretreated spectra. Table 1. Mean value, standard derivation (SD) and relative standard deviation (RSD) of the intensities of the Si I 251.6 nm, Na I 589.0 nm, and K I 766.5 nm lines, recorded from sample S1 in the 3 the forms of pressed pellet, polished and raw rocks, as well as calculated over all the replicate spectra of the 3 sample forms. www.nature.com/scientificreports/ Spectral feature selection. SelectKBest (SKB) algorithm was respectively applied to the pretreated spectra of the training pellet and training rock sample sets, and successively for the 3 concerned oxides. Within a sample set, for each spectral channel, covariance was calculated between the channel intensity and the concentration of the concerned compound in the corresponding sample, over all the spectra of the sample set. A score was then calculated as a function of the covariance according to the definition given in Reference 13 . A ranking index, ρ i,j , was thus associated to each spectral channel according to its obtained score, with 2 indexes (i, j) and a value varying from 1 to 22,161, which ranks the channels from the lowest score to the highest one. Such procedure was applied to the 2 sample sets ( i = 1 : training pellets, i = 2 : training rocks) and the 3 concerned oxides ( j = 1 : SiO 2 , j = 2 : Na 2 O, j = 3 : K 2 O). A feature selection procedure identified 100 highest ranked spectral channels respectively for each of the 3 oxides in each of the 2 training sample sets. Pearson's correlation coefficient 27 related to the above mentioned covariance was calculated for the 6 groups of 100 selected features. The results showed that all the selected features had a Pearson's coefficient larger than 0.75. As we can see in the Fig. 3, the 3 groups of 100 features selected for the 3 oxides for the training pellet sample set were directly used to respectively train the calibration models for the 3 oxides base on a back-propagation neural network (BPNN). The training algorithm that involved stochastic gradient descent (SGD) and mini-batch stochastic gradient descent (MSGD) optimization iterations, as well as cross-validations with randomly generated statistically equivalent data configurations, has been presented in detail in Reference 13 . For transfer learning model training, and according to the principle of feature-representation-transfer discussed above, an ensemble of common selected features was identified between the training pellet and the training rock sample sets, by calculating a total ranking index ρ j = ρ 1,j + ρ 2,j . One hundred highest ranked features according to the value of ρ j from the highest one to the lowest one, were retained as the common selected features, respectively for the 3 oxides. These groups of features were then fed into the transfer learning model training algorithm. The results of feature selection for Na 2 O for the pair of sample types pellet/raw rock, are shown in Fig. 4. Similar behaviors can be observed in the feature selections for the other 2 oxides and with the 2 pairs for the sample types pellet/raw rock and pellet/polished rock, the corresponding results are shown in the section "Methods" in Figs. 10 and 11. In Figs. 4a, we can see that for the training pellets, the spectral channels with high SKB scores are clearly concentrated around the several Na emission lines: Na I 330.24 nm and 330.30 nm lines, Na I 588.99 nm and 589.59 nm lines (with 2 groups of ghost lines around 572.1 nm and 606.9 nm), Na I 818.33 nm and 819.48 nm lines. For the training raw rocks in Fig. 4b, the selected features are distributed also among other channels with a significant decrease of the scores for all the important features. This means that the physical matrix effect perturbs the inherent correlation between the emission line intensities of an element and its concentration in the material, and reduces therefore the importance of the line intensities in the concentration determination. At the same time, other spectral channels, such as those around 275 nm and between 410 and 460 nm, get relatively higher scores. This means that they become important in the determination of elemental concentration when using a model based on the training set of the rock samples. These features, representative of the rock samples, www.nature.com/scientificreports/ are thus included in the common selected features for transfer learning model training. Figure 4c shows in red dots, the total ranking index of the 100 common selected features for Na 2 O. These features are indicated in a typical spectrum in Fig. 4d in red dots. We can see that, beside the features related to the Na emission lines, some features important for the rock samples are included. A more detailed peak identification using the NIST database 28 , shows the contributions from Fe II 268.475 nm and 275.57 nm lines, Si II 385.366 nm and 385.602 nm lines, and the probable contributions from K I 404.414 nm and 404.721 nm lines, Ca I 409.85 nm lines, and Si II 412.807 nm and 413.089 nm lines. A selected feature around 461 nm cannot have easy interpretation. In the insets of Fig. 4d, 2 parts of the spectrum are enlarged. The inset around 589 nm shows the sodium D lines together with the selected features in red dots. We can see that the selected features are located in the side parts of the line profiles, whereas the central parts of the lines are not retained by the feature selection algorithm. This might be due to self-absorption of the strong resonant Na D lines, which affects much more the central part of the spectral lines. It would also be the reason for the higher scores observed for the weaker Na emission lines around 330 nm. This observation would show the ability of the feature selection process to reduce the influence of self-absorption by selecting the most suitable features inside of a line profile. The second inset in Fig. 4d shows an enlarged part of the spectrum around 820 nm, where we can see the selected features related to the Na I 819.5 nm line in red dots. Due to the spectral interference with the N I 820.0 nm line, only the short wavelength part of the spectral profile around 820 nm is included in the selected features, showing the effectiveness of the feature selection to avoid the influence of spectral interference. ## Transfer learning-based calibration model training. A transfer learning model training algorithm was developed in this work on the basis of that used for machine learning model training presented in detail in our previous publication 13 and used for various application scenarios 12, Data formatting. According to the above discussed principles of feature-representation-transfer and instance transfer in transfer learning, spectra from the 18 training pellet samples (the source domain) and those from the 8 training rock samples (the target domain), with their 100 common selected features, participated in the training process. These spectra were organized in a given data configuration where the replicate spectra for each sample were arranged in an arbitrarily given order. The effectiveness of each training pellet was tested within an iteration loop where the RETs with and without the spectra from the pellet were compared in order to decide the exclusion or the definitive inclusion of the pellet in the final transfer learning model training sample set. It was why the ensemble of pretreated replicate spectra associated to one of the 18 training pellets was indexed with k that went from 1 to 18 (Fig. 5a). The 8 training rock samples contributed to the transfer learning model training and in particular, were used in a cross-validation process during the optimization of the neural network. It was why the corresponding spectra were first organized in different data configurations where each configuration j corresponded to a certain arrangement of pretreated replicate spectra for each training rock (Fig. 5a). The data configurations were all statistically equivalent since the order of a replicate spectrum of a sample was a dummy www.nature.com/scientificreports/ index. The number of different data configurations were limited to 3 in this work because more configurations did not bring further improvement of the model as tested in the experiment. For a given configuration j, the pretreated replicate spectra of each sample were further organized into 5 groups containing respectively 4, 4, 4, 4 and 5 spectra, respectively. A new index i was introduced to designate ensemble of the groups of pretreated replicate spectra of all the training rocks as shown in Fig. 5a. In the model training process, the index i went from 1 to 5 within an iteration loop of cross-validation, indicating each time the validation ensemble of the groups of pretreated replicate spectra. Model training by optimization. A 3-layer back-propagation neural network (BPNN) similar to that used in Reference 13 was employed in this work for the transfer learning model. The network was composed of an input layer of 100 neurons corresponding to the 100 common selected features of each input spectrum; a hidden layer 5 neurons and an output layer with a single neuron corresponding to the targeted compound concentration. The function of the network was therefore to map an input spectrum (a vector of 100 dimensions) to a scalar which can be considered as the module of a vector in a hyperspace of 100 dimensions. The accuracy of the mapping was improved during the training process through different iteration loops under the supervision of the targeted concentration and using the model performance indication parameters specified above. As shown in Fig. 5b, 3 hierarchized iteration loops, i, j, k , among them i, j are doubled loops for a given k ( ±k ) surrounding the BPNN optimization loop performing the supervised optimization of the model. -A doubled inner loop for i = 1 to 5: for the double cases of a given sample k among the pellet samples being excluded ( −k ) or included ( +k ) in the training data set, and a given data configuration j of the rock spectra, the network was optimized within a cross-validation process where the model was trained using 4 ensemble of groups of replicate spectra, of for example, i = 2, 3, 4, 5 with respectively 4, 4, 4 and 5 spectra for each sample. The resulted REC(ij − k) and REC(ij + k) were calculated for the respectively optimize d models for test (ij − k) and (ij + k) . These models were then tested using the rest ensemble of groups of replicate spectra, i = 1 for instance, generating RET(j − k) and RET(j + k) , together with the optimized models for test (j − k) and (j + k). -A doubled intermediate loop for j = 1 to 3: in this loop, the above discussed loop i was executed with 3 independent training rock data configurations for the 2 cases of a given sample k among the pellets being excluded from or included in the training data set. The model was further optimized. Corresponding calculation of RET resulted in RET(−k) and RET(+k). -An outer loop for k = 1 to 18: in this loop the above discussed loop i and loop j were executed for each of the 18 training pellet samples successively assigned as the pellet k. For a given pellet k, RET(−k) and RET(+k) were compared. If an improvement was observed with the sample, it was kept in the final training sample set, otherwise it was rejected. This loop generated a model for test (k) for each considered pellet sample with the corresponding RET(k) . The optimization process finally generated a model for validation with a minimized RET and RMSET. Model validation. The resulted transfer learning model was validated by the pretreated spectra from the 12 validation rock samples including the 2 isolated rocks without counterpart pellet in the training sample set, with the identified features according to the common selected features between the training pellets and the training rocks. The parameters assessing the performance of the model for prediction, REP, RMSEP and RSD were calculated. These parameters indicate the performance of the model when used for predictions with LIBS spectra from rock samples, including unseen rocks, simulating thus a real application scenario. Remark that some of the training pellets, counterparts of the validation rocks and initially included in the model training sample set, were later rejected by the model training process (see Table 5 in the section "Methods") and did not participate to the final model optimization process. Such configuration of validation provided the assessments of the transfer learning model in the both cases of rocks with counterpart pellets more or less seen during the model training and rocks totally unknown by the model. ## Results and discussion Analytical performances with the machine learning model. In order to emphasize the improvement with transfer learning, we first present the results obtained with the machine learning models trained using the 18 training pellet samples and validated using the 2 isolated pellets and the 12 validation rocks respectively for the 3 concerned oxides, SiO 2 , Na 2 O and K 2 O. Such double validations allowed the correction of chemical matrix effect being explicitly checked with independent pellets before the check of physical matrix effect with rock samples. The training method described in Reference 13 was implemented in this work to train a neural network. The training procedure was similar to the inner (loop i) and the intermediate (loop j) iteration loops used in the transfer learning model training (Fig. 5b) with a similar neural network structure. As shown in Fig. 3, the input variables were the 100 selected features in a pretreated spectrum of a training pellet sample for the training, and the 100 identified features in a pretreated spectrum of an isolated pellet sample or a validation rock sample for the validation. www.nature.com/scientificreports/ nals are plotted in the figures as a reference for the models. The extracted parameters for assessment of model performances are presented in Table 2 according to the definitions provided above. Although in Fig. 6, the results are presented with a given typical pair of isolated validation samples (S2 and S6), in Table 2 the validation performances are calculated as average values of those obtained with 6 different pairs of isolated validation samples (S2 and S5; S1 and S6; S4 and S12; S15 and S16; S10 and S17; S9 and S20), which ensures the independence of these performances on the choice of validation samples. For validation with rocks, we make the distinction between the 2 isolated rocks and the 10 rocks with counterpart pellets. In Fig. 6 and Table 2, we can see that the machine learning models trained with the training pellet samples present good calibration performances in terms of the usual assessment parameters including r 2 , LOD , REC , RET , and RMSE . In addition, the validation with the 2 isolated pellet samples also show satisfactory results. This indicates an effective correction of the chemical matrix effect with machine learning regression, as we also observed in our previous works 12,13 . Whereas, we can remark an obvious degradation of the performance when the models were tested using the validation rock samples, in terms of REP , RMSEP and RSD due to the influence of the physical matrix effect. In Fig. 6, the 2 isolated rocks do not show a particularly "bad" behavior with respective to the other validation rocks with counterpart pellets in the training sample set, which would indicate the fact that the absence of bulk chemistry of a rock for the model training does not particularly further influence its prediction by the model. This remark is confirmed by Table 2. Moreover, Fig. 6 shows that the use of a model trained with pellet samples for prediction with the spectra from rock samples can lead to bias, with a shift of the linear regression of the validation data with respect to that of the training data, as well as variance, with a rotation of the linear regression of the validation data with respect to that of the training data. We can also remark that the model performance degradation observed with polished rock samples is in general, further aggravated for raw rock samples, as also indicated by Table 2 where we can see increased average REP and RMSEP when one passes from polished rocks to raw rocks. This means that the physical matrix effect arises due to different surface hardness and heterogeneity of a polished rock with respect to its corresponding pressed powder pellet. Surface roughness of a raw rock introduces additional perturbations leading to in general, larger prediction uncertainties. A detailed look on the validation performances in Table 2 however shows that the influence due to surface roughness (raw rocks) remains smaller than that due to surface hardness and heterogeneity (polished rocks), which contributes to the largest part of the physical matrix effect. As a consequence of the influence of the physical matrix effect, the TAS classification of the validation rock samples with the pellet machine learning models led to an unsatisfactory result as shown in Fig. 7. In this figure, the reference positions in the TAS diagram of the validation rock sample determined by their compositions measured using XRF (as shown in Fig. 1 and Table 4) are indicated with solid green stars for the 2 isolated rocks, and solid blue circular points for the rest of the validation rocks. The position predicted by the pellet machine learning models for the same sample is represented by a cross of the same color with error bars. More precisely, the Analytical performances with the transfer learning model. Calibration models resulting from transfer learning are shown in Fig. 8 with a similar format as those resulting from machine learning presented in Fig. 6, in order to review the improvements by comparison. The extracted parameters for assessment of the model performances are presented in Table 3 with validation performances calculated as average values of those obtained with the 6 different pairs of isolated validation samples. In Fig. In Fig. 8, we do not remark particular behavior for the 2 isolated rocks with respect to the other validation rocks as in Fig. 6. In Table 3, we can see that although the transfer learning models present in general, slightly lower calibration performances in terms of r 2 , LOD , REC , RET and RMSE compared to the machine learning models, the prediction performance for polished and raw rock samples are much improved, especially for REP and RMSEP . This means that the participation of the 8 rock samples in the training data set together with the retained pellet samples with common selected features, effectively takes into account the physical matrix effect and reinforces the robustness of the models for prediction for rock samples, including isolated rocks totally unknown by the models. We remark in particular, the prediction performances for both polished and raw rocks are simultaneously improved, showing the effectiveness of the transfer learning models in the correction of physical matrix effects of different origins. The calibration models shown in Fig. 8 were used to present the validation rock samples in a TAS diagram. The obtained results are shown in Fig. 9a for polished rock samples and Fig. 9b for raw rock samples using the same symbols as in Fig. 7. We can see a much improved result conforming the good performances of the transfer learning models shown in Fig. 8 and Table 3. A detailed counting shows 10 correctly classified validation samples for the both polished and raw rocks, including the 2 isolated rocks. Only two samples were classified into a wrong field (S12 and S15 for polished rocks, S4 and S6 for raw rocks). The rate of correct classification can thus be determined to be 83.3% in the both cases. These results show the effectiveness of the developed method to www.nature.com/scientificreports/ correct the physical matrix effect. Confirming the observation in Fig. 8, no particular behavior can be observed for the 2 isolated rocks in the both cases of polished and raw rocks with respect to the other validation rocks. ## Conclusions In this work, within a given application of classification of rocks using the TAS diagram, we have introduced transfer learning in LIBS spectral data treatment to improve the performance of the models trained using laboratory standard samples in the form of pressed powder pellet, when used for prediction with LIBS spectra acquired www.nature.com/scientificreports/ from natural rocks with a polished surface or in a raw state. Such scenario corresponds to the important application of analysis of rocks with LIBS on Mars, although the used experimental configuration compared to the current rovers on Mars remains still quite different, concerning the ambient gas, the laser excitation, as well as the spectrum detection. The purpose was therefore to work on a general method that can be later implemented according to specific experimental conditions into particular applications. More precisely, feature-representationtransfer and instance-transfer as the two important processes of transfer learning were implemented in the LIBS spectral data treatment. The performances of the resulted transfer learning models were compared with those of the machine learning models. Significant improvements have been realized for prediction with LIBS spectra acquired on polished and raw rock samples for the 3 concerned compounds involved in the TAS classification, SiO 2 , Na 2 O and K 2 O. The rate of correct TAS classification has been improved from 25% for polished rocks and 33.3% for raw rocks with the machine learning models to 83.3% for the both types of rock samples with the transfer learning models. The obtained results therefore demonstrate the effectiveness of transfer learning to overcome the physical matrix effect due to the change of sample physical state in LIBS analyses. There are still steps forward to realize in research and development to apply the method developed in this work to Mars explorations with LIBS. Such steps should involve a larger set of samples, with the possibility to isolate more rock samples for the independent validation of the transfer learning models, although the results shown in this work do not reveal obviously different behavior of the isolated validation rocks with respect to the other validation rocks that can have a counterpart pellet in the model training set. This would indicate a dominant physical matrix effect in the given configuration of study. It is also to be taken into account the experimental conditions, including the measurement environment (ambient gas and its pressure), the used laser parameters and the spectrum detection, in order to reduce the dissimilarities between a laboratory simulation experiment and the in situ LIBS measurements on Mars to a strict minimal related to the lack of complete knowledge about a real sample to be analyzed on Mars. Beyond analysis of rocks with LIBS in Mars explorations, our findings in this work can also have more general interests in the development of LIBS technique for various applications involving sets of samples with different surface physical properties.

chemsum

{"title": "From machine learning to transfer learning in laser-induced breakdown spectroscopy analysis of rocks for Mars exploration", "journal": "Scientific Reports - Nature"}

catalytic_decarboxylative_radical_sulfonylation

## Abstract: Sulfones are not only important structural motifs in pharmaceuticals and agrochemicals but also versatile intermediates in organic synthesis. However, C(sp 3 )-sulfonyl bond formations remain underdeveloped. In this issue of Chem, Li and co-workers demonstrate that the merger of photo-organocatalysis and copper-catalysis enables the decarboxylative radical sulfonylation with organosulfinates at room temperature under redox-neutral conditions. The method leads to the improved synthesis of anti-prostate cancer drug bicalutamide and should find more important applications in drug discovery. ## INTRODUCTION Sulfones are ubiquitous in nature. They possess a wide variety of biological activities and thus serve as important structural motifs in pharmaceuticals and agrochemicals. For example, bicalutamide (CASODEX, AstraZeneca's blockbuster drug) is an orally active, nonsteroidal anti-androgen for the treatment of prostate cancer (Figure 1). 1,2 Certinib (Zykadia, Novartis) is a new drug approved by FDA (U.S. Food and Drug Administration) in 2014 to treat anaplastic lymphoma kinase (ALK)-positive non-small cell lung cancer. 3 Another example is oral drug apremilast (Otezla, Celgene), the only phosphodiesterase 4 (PDE4) inhibitor approved by FDA to treat active psoriatic arthritis and plaque psoriasis with an annual sales of over 1.2 billion US dollars. 4 In the meantime, sulfones are also versatile synthetic intermediates in organic chemistry. They serve as the key building blocks or reagents in many chemical transformations such as Julia-Lythgoe olefination and fluoroalkylation. 8 As a consequence, the synthesis of sulfones has received a considerable attention and significant progress has been achieved in recent years. However, recent advances focus mainly on C(sp 2 )-sulfonyl bond formations such as aromatic or vinylic sulfonylation, whereas C(sp 3 )-sulfonyl bond formations remain underdeveloped. Conventional C(sp 3 )-sulfonylation methods include (1) nucleophilic sulfonylation of electrophiles such as alkyl halides, epoxides, or Michael acceptors with organosulfinates or thiosulfonates under basic conditions 14 (Scheme 1A) and (2) electrophilic sulfonylation of sulfonic acid derivatives such as sulfonate esters or sulfonyl chlorides with organometallic reagents (Scheme 1B). However, the former suffers from the competing O-alkylation (to give sulfinate esters), while the latter often leads to the corresponding sulfoxides. 9 The recently developed sulfonyl radical addition to unsaturated bonds such as alkenes The Bigger Picture Sulfones are not only important structural motifs in pharmaceuticals and agrochemicals but also versatile synthetic intermediates in organic chemistry. Despite the significant progress in the synthesis of sulfones in recent years, C(sp 3 )sulfonyl bond formations remain underdeveloped. In particular, there have been no reports to date of general methods for the sulfonylation of alkyl radicals. In this article, we introduce the copper-catalyzed cross coupling of sulfinates with alkyl radicals generated via photoredoxcatalyzed decarboxylation of redox-active esters derived from aliphatic carboxylic acids. This unprecedented protocol exhibits broad substrate scope and wide functional group compatibility, allowing the late-stage sulfonylation of complex molecules. The synthetic utility of the method is further demonstrated by the improved synthesis of anti-prostate cancer drug bicalutamide. provides a powerful means for C(sp 3 )-sulfonyl bond formations (Scheme 1C). This method can be extended to three-component condensation involving the fixation of SO 2 . 12,13 Nevertheless, the strategy is mainly limited to arenesulfonyl radicals (or aryl radicals plus SO 2 ), while examples of alkyl radical addition to SO 2 (e.g., Reed reaction 17 ) are rare due to the fast desulfonylation of alkanesulfonyl radicals. In fact, there have been no reports to date of general methods for the sulfonylation of alkyl radicals. Given that alkyl radicals are common intermediates easily generated from various types of organic compounds, it is certainly highly desirable to develop efficient and general methods for the sulfonylation of alkyl radicals (Scheme 1D). In particular, the decarboxylative sulfonylation of aliphatic carboxylic acids should be an important transformation useful in organic synthesis. To meet the challenge, we propose the concept of ''RSO 2group-transfer from Cu(II)-SO 2 R to alkyl radicals,'' in accordance with our previous ideas of Cu(II)-assisted CF 3 -group-transfer or F-atom-transfer. 21 Specifically, we target the copper-catalyzed cross coupling of sulfinates with alkyl radicals generated via photoredox-catalyzed decarboxylation of redox-active esters derived from aliphatic carboxylic acids. ## RESULTS AND DISCUSSION Carboxylic acids and sulfinates 44 are both attractive raw materials for chemical synthesis due to their ready availability, high stability, and low cost. Decarboxylative cross coupling of aliphatic carboxylic acids with sulfinates should therefore be an ideal method for sulfone synthesis. Given that sulfinates are much easier to be oxidized than the corresponding carboxylic acids, direct oxidative decarboxylative coupling is not feasible. Reductive decarboxylation of redox-active esters of carboxylic acids might provide the solution. We then commenced our investigations with the redox-active ester of 4-phenylbutanoic acid (1a) and sodium 4-methylbenzenesulfinate (2a) as the model substrates. The redox-active ester was easily prepared by condensation of 4-phenylbutanoic acid with N-hydroxyphthalimide (NHPI). After extensive screening of reaction conditions (see Table S1 for details), we were pleased to find that, with 2 mol % 1,2,3,5tetrakis(carbazol-9-yl)-4,6-dicyanobenzene (4CzIPN) 45,46 and 20 mol % Cu(OTf) 2 as catalysts and 2 equiv of dibutyl hydrogen phosphate as the additive, the visiblelight-induced reaction of ester 1a and sulfinate 2a at room temperature (RT) for 12 h afforded the desired sulfone 3a in 95% yield (Table 1, entry 1). When the reaction was performed in the absence of (BuO) 2 P(O)OH, the yield dropped to 65% (Table 1, entry 2). Switching the additive to trifluoroacetic acid also led to a high product yield (Table 1, entry 3), while weak acids such as acetic acid showed no improvement (Table 1, entry 4). Interestingly, the addition of 2,2 0 -bipyridine (20 mol %) significantly inhibited the reaction (Table 1, entry 5). Control experiments revealed that photocatalyst 4CzIPN, copper catalyst, and visible light were all essential for the transformation (Table 1, entries 6-8). In addition, the redox-active ester could be prepared in situ without purification, providing the sulfonylation product in 85% yield (Table 1, entry 9). It is worth mentioning that the use of free 4-methylbenzenesulfinic acid in place of sodium 4-methylbenzenesulfinate resulted in a very low (7%) yield of 3a. However, the combination of 4-methylbenzenesulfinic acid with a weak base such as NaHCO 3 or CF 3 CO 2 Na increased the product yield to 51% and 89%, respectively (see Table S2). These experiments suggest that (BuO) 2 P(O)OH or trifluoroacetic acid might serve as a buffer to adjust the pH value of reaction solution. The use of (BuO) 2 P(O)OH or trifluoroacetic acid as the additive proved to be even more critical in the decarboxylative sulfonylation of secondary alkyl acids, resulting in a sharp increase in product yield by inhibiting the generation of the alkene byproduct (see Table S2). While the detailed mechanism remains unclear for the remarkable acid effect, it might be possible that the presence of an acid decreases the basicity of sulfinate and thus retards a-deprotonation of alkyl radicals (to give alkene radical anions and hence alkenes after electron transfer). Note that the CH groups adjacent to a carbon radical center are quite acidic as pointed out by Studer and co-workers. 47,48 With the optimized conditions in hand, we examined the scope of the method. As shown in Scheme 2, various NHPI esters derived from primary and secondary alkyl acids underwent smooth decarboxylative sulfonylation with sulfinate 2a to provide the corresponding sulfones 3a-3v in good to excellent yields. The presence of a wide range of functional groups was tolerated by the process. For example, terminal alkenes, alkynes, alkyl or aryl bromides, aldehydes, ketones, esters, amides, sunfonamides, ethers, and unprotected indoles all proved to be compatible with the reaction. Protected a-amino acids were also suitable substrates, as evidenced by the synthesis of sulfone 3m. The reaction could be operated in gram scale without the loss of efficiency. The method was also applicable to NHPI esters derived from tertiary alkyl acids, albeit in a lower efficiency (e.g., 3w) presumably because of steric hindrance. The protocol has also shown a broad scope in terms of organosulfinates, as demonstrated in Scheme 3. A number of arenesulfinates with either electron-withdrawing or electron-donating substituents on the aromatic ring all underwent sulfonylation reactions furnishing the corresponding sulfones 4a-4d in satisfactory yields. Moreover, the protocol was also applicable to alkanesulfinates. Primary, secondary, and tertiary Article alkanesulfinates were all suitable partners in the cross coupling, as exemplified by the efficient synthesis of sulfones 4e-4k. An excellent chemoselectivity was observed in the reaction of 1-allylcyclopropanesulfinate furnishing sulfones 4j and 4k in which the allyl group remained intact. Nevertheless, the sulfonylation with pyridine-3-sulfinate afforded sulfone 4l in a low yield due to the competing Minisci alkylation, and no desired product could be observed in the sulfonylation with trifluoromethanesulfinate. The results also indicated that sulfonyl radicals were unlikely to be involved in the reaction given the fast desulfonylation of alkanesulfonyl radicals. The above results clearly demonstrate the broad substrate scope and wide functional group compatibility of the method. In addition, the reactions were run at room temperature under redox-neutral conditions free from external reducing or oxidizing agents. These characteristics enabled the late-stage modification of complex natural products or drug molecules (Scheme 4). For example, steroids such as dehydrocholic acid or chenodeoxycholic acid were readily converted to the corresponding sulfones 5a or 5b in high to excellent yield. Drug molecules such as isoxepac, chlorambucil, mycophenolic acid, gibberellic acid, or indometacin were all transformed into the corresponding sulfones highly efficiently. The protocol was also applicable to carbohydrate-containing acids, as evidenced by the synthesis of 5f in 91% yield. To further demonstrate the synthetic utility of the sulfonylation method, we chose the anti-prostate cancer drug bicalutamide as the target molecule. Bicalutamide and its analogs were achieved by multi-step synthesis starting from the commercially Article available and inexpensive (S)-malic acid (6). However, three steps were required for the installation of the sulfonyl group in bicalutamide consisting of (1) Barton decarboxylative bromination of acid 7 with CBrCl 3 , (2) nucleophilic substitution of the resulting bromide with 4-fluorobenzenethiol, and (3) subsequent oxidation with mCPBA (meta-chloroperoxybenzoic acid). With the newly developed decarboxylative sulfonylation method, the three steps could be shortened into one (Scheme 5). Specifically, the condensation of acid 7 with NHPI and DIC (diisopropylcarbodiimide) produced the corresponding NHPI ester, which, without purification, was subjected to the treatment with p-fluorobenzenesulfinate under the optimized Article conditions to provide sulfone 8 in 60% yield. The following hydrolysis of 8 with KOH in aqueous methanol at room temperature afforded cleanly compound 9, which, without purification, was subjected to the condensation with 4-cyano-3-trifluoromethylaniline according to the literature procedure, 49 furnishing (R)-bicalutamide in 91% yield based on 8. The new synthetic route offers a more concise and efficient entry to bicalutamide and avoids the use of toxic CBrCl 3 , odorous p-fluorobenzenethiol, and dangerous mCPBA. To gain further insight into the sulfonylation, mechanistic studies were carried out (Scheme 6). The reaction of the NHPI ester derived from cyclopropylacetic acid (10) under the optimized conditions produced exclusively the ring-opening sulfonylation product 3i in 88% yield. When the NHPI ester of hept-6-enoic acid (11) was subjected to the treatment with sulfinate 2a under the optimized conditions, the cyclized product 3e was obtained in 50% yield along with the uncyclized product 12 isolated in 25% yield. These radical clock experiments unambiguously demonstrated the intermediacy of alkyl radicals. Additionally, radical trapping experiment with 1,1-diphenylethylene also suggested the involvement of alkyl radicals rather than sulfonyl radicals (see Scheme S1). Furthermore, copper(II) p-toluenesulfinate prepared from Cu(OH) 2 and sodium p-toluenesulfinate was treated with an equimolar amount of triethylborane (as the ethyl radical precursor) in acetonitrile at room temperature under aerobic conditions, and ethylsulfone 13 was achieved in 81% yield. This experiment provided a solid evidence for the proposed mechanism of Cu(II)-assisted RSO Article elimination. Given that uncomplexed copper(II) sulfinate is a mild oxidant rather than a reducing agent and the presence of 2,2 0 -bipyridine inhibits the sulfonylation, the direct RSO 2 group transfer seems more likely the case. More mechanistic studies are certainly required to provide a detailed understanding on the mechanism. The ratio of cyclized product 3e versus uncyclized product 12 was 2:1 in the radical clock experiment shown in Scheme 6. The rate constant for the cyclization of hex-5-en-1-yl radical at 20 C is known to be approximately 2.3 3 10 5 s 1 , as determined by Ingold and co-workers. 50 By assuming that the active intermediate species responsible for sulfonylation is of the same concentration as the catalyst Cu(OTf) 2 (0.02 M) and remains constant throughout the reaction, the rate constant for the toluenesulfonyl group transfer from Cu(II)-SO 2 Tol to a primary alkyl radical can Article then be calculated to be around 2.3 3 10 7 M 1 s 1 , which is much larger than the rate constant for primary alkyl radical addition to an unactivated monosubstituted alkene (10 3 $ 10 4 M 1 s 1 ) 51 or to benzene (3.8 3 10 2 M 1 s 1 ). 52 This well explains the remarkable chemoselectivity in the reaction of 1-allylcyclopropanesulfinate (to give 4j and 4k). Furthermore, it may also account for the low efficiency in the reaction of pyridine-3-sulfinate (to give 4l), given that rate constants for radical addition to protonated 4-cyanopyridine (Minisci alkylation) range from 8.9 3 10 5 M 1 s 1 for n-butyl radical to 6.3 3 10 7 M 1 s 1 for t-butyl radical. 52 Thus, the rate constant for RSO 2 group transfer determined above offers a quantitative view on the reaction mechanism and sets the stage for the rational design of new synthetic methodology based on radical sulfonylation. ## Conclusions In conclusion, the merger of photo-organocatalysis and copper catalysis enables the successful development of a practical protocol for the sulfonylation of alkyl radicals generated from aliphatic carboxylic acids. As the procedure is operationally simple, catalytic in copper, broad in scope, free from external reducing or oxidizing agents, tolerant of sensitive functional groups, and utilizes cheap and stable sulfinates, the method should find widespread applications in sulfone synthesis. It is also conceivable that more new methods of radical sulfonylation will be developed given that alkyl radicals can be generated by many other ways (such as hydrogen atom abstraction, olefin radical addition, etc.). Furthermore, the proposed mechanism of Cu(II)-assisted RSO 2 group transfer should stimulate further research toward the cross coupling of alkyl radicals with nucleophiles other than sulfinates. The research in this direction is currently underway in our laboratory. ## SUPPLEMENTAL INFORMATION Supplemental Information can be found online at https://doi.org/10.1016/j.chempr. 2020.02.003.

chemsum

{"title": "Catalytic Decarboxylative Radical Sulfonylation", "journal": "Chem Cell"}

downstream_processing_of_isochrysis_galbana:_a_step_towards_microalgal_biorefinery

## Abstract: An algae-based biorefinery relies on the efficient use of algae biomass through its fractionation of several valuable/bioactive compounds that can be used in industry. If this biorefinery includes green platforms as downstream processing technologies able to fulfill the requirements of green chemistry, it will end-up with sustainable processes. In the present study, a downstream processing platform has been developed to extract bioactive compounds from the microalga Isochrysis galbana using various pressurized green solvents. Extractions were performed in four sequential steps using (1) supercritical CO 2 (ScCO 2 ), (2) ScCO 2 /ethanol (Gas Expanded Liquid, GXL), (3) pure ethanol, and (4) pure water as solvents, respectively.The residue of the extraction step was used as the raw material for the next extraction. Optimization of the ScCO 2 extraction was performed by factorial design in order to maximize carotenoid extraction.During the second step, different percentages of ethanol were evaluated (15%, 45% and 75%) in order to maximize the extraction yield of fucoxanthin, the main carotenoid present in this alga; the extraction of polar lipids was also an aim. The third and fourth steps were performed with the objective of recovering fractions with high antioxidant activity, eventually rich in carbohydrates and proteins. The green downstream platform developed in this study produced different extracts with potential for application in the food, pharmaceutical and cosmetic industries. Therefore, a good approach for complete revalorization of the microalgae biomass is proposed, by using processes complying with the green chemistry principles. † Electronic supplementary information (ESI) available. See ## Introduction A biorefinery involves biomass conversion processes and equipment to produce fuel, power, and added-value chemicals from organic materials 1 such as renewable resources or microalgae. Microalgae are among the most promising raw materials for the sustainable supply of commodities and the use of algae. 2,3 They use light energy, residual nutrients and carbon dioxide (that can be obtained from flue gas) with higher photosynthetic efficiency than plants for the production of biomass. 4 Moreover, these organisms may be grown on nonarable land, thus, not competing with food needs for biofuel production. Microalgae biomass is an excellent source of oils (including high amounts of long chain polyunsaturated fatty acids (LC PUFAs)), proteins, polysaccharides (such as starch, xylans, pectins, glucans, extracellular polysaccharides (EPS)) and other high-added value compounds such as carotenoids, pigments, antioxidants, sterols and minerals. The potential for the production of these different components may even be tuned by setting particular growing conditions. Therefore, the microalgae-based biorefinery concept relies on the complete process chain ranging from optimization of biomass production to the development of a platform able to generate a wide range of products, from bulk chemicals, food supply ( proteins, fibres), bioactive compounds, and oils with respect to its use as a biofuel. Isochrysis galbana is a small marine flagellate (Phylum: Haptophyta) widely used in aquaculture as a PUFA-rich microalga. 5 It is commercially produced as feed for the early larval stages of mollusks, fish, and crustaceans. In fact, I. galbana cells produce antibacterial substances, which increase the toxicity of free fatty acids such as eicosapentanoic acid (EPA) to several pathogens, without the use of chemicals that might harm organisms under culture conditions or the environment. 6 Besides polyunsaturated fatty acids, I. galbana is a valu-able source of proteins, carbohydrates and photosynthetic pigments such as chlorophyll a and fucoxanthin. 7 Fucoxanthin, a major carotenoid present in the chloroplasts of brown seaweeds, contributes to more than 10% of the estimated total production of carotenoids in nature. Although fucoxanthin is clearly a valuable pigment with various health benefits, its use has been limited due to the low extraction efficiency from marine materials and the difficulty to synthesize it. In this respect, algae, such as I. galbana, can be considered as a potential source of fucoxanthin. 8 In order to fully develop the microalgae-based biorefinery concept, new aspects related to technologies for extraction, isolation and fractionation of the biomass into multiple products (lipids, proteins, polysaccharides, bioactives, etc.) should be studied. Also, steps into integrated approaches for multiproduct biorefinery should be taken into account to improve the efficiency and minimize the energy and resource consumption, 9 especially when green chemistry principles and sustainability issues are to be considered. Traditionally, extraction of lipophilic compounds from algae, such as carotenoids and lipids, has been performed by means of toxic organic solvents like hexane. Nowadays there is a demand for fast, selective, efficient and greener processes able to provide extractions with high yields; besides, the costs associated have to be reduced, for instance, by minimizing the removal of solvent residues. High-pressure extractions such as supercritical fluid extraction (SFE) and pressurized liquid extraction (PLE) using GRAS (generally recognized as safe) solvents such as CO 2 , ethanol or water, have emerged as promising alternatives to face these challenges. 10 This was the subject of a specifically devoted workshop on Supercritical Fluids and Energy that was conducted in Brazil in December 2013, 11 with the idea of assessing the potential of supercritical (pressurized fluids in general) technologies in the fields of energy, materials science, process technology, green chemistry and sustainable technologies. SFE offers a fast extraction rate, high selectivity and is an ecofriendly technology with minimal or no use of organic solvents, although the low polarity of supercritical CO 2 (ScCO 2 ) limits its applications. ScCO 2 has been reported as an interesting approach for the extraction of lipids with antimicrobial activity from the microalgae Chaetoceros muelleri, 12 n-3 fatty acids from the seaweed Hypnea charoides, 1 lutein and β-carotene from Scenedesmus almeriensis 13 and fucoxanthin from the seaweed Undaria pinnatifida 14 and Sargassum muticum, 15 among others. In this latter application, the addition of ethanol as a co-solvent improved the yield of fucoxanthin in both algal species. 15,16 Ethanol is often used as a modifier or a co-solvent of ScCO 2 in order to overcome the CO 2 limitations towards the extraction of medium polarity bioactive compounds. For instance, CO 2 modified with ethanol has been applied for the extraction of astaxanthin from Haematococcus pluvialis 17 and various pigments from Spirulina platensis. 18 The use of a co-solvent at a higher concentration allows working in the region of gasexpanded liquids (GXLs), 19 which is a promising intermediate between PLE and SFE for the extraction of medium or highpolarity compounds. Carbon dioxide expanded ethanol (CXE) has been recently used to obtain astaxanthin enriched extracts from H. pluvialis. 20 Pressurized liquid extraction has demonstrated an interesting potential for extracting bioactive compounds from macroand microalgae. 10,20 This extraction technique allows obtaining higher yields than those achieved by conventional extraction techniques, in a shorter time and with less solvent consumption. 10 PLE using ethanol has been reported for the extraction of carotenoids from Neochloris oleoabundans, 22 Dunaliella salina 2 and Chlorella ellipsoidea. 3 In addition, 90% ethanol was used for the extraction of fucoxanthin from Eisenia bicyclis 23 and the mixture of ethanol/limonene (1 : 1, v/v) has been proposed as a green approach for PLE extraction of lipids from microalgae. 9 In the present study, we propose an integrated sequential extraction process based on the use of green compressed fluids, in increasing order of polarity, for the fractionation of bioactive compounds from the microalga I. galbana, as an approach to develop a microalgae biorefinery procedure. 21 The developed process comprises the sequential extraction with ScCO 2 , CO 2expanded ethanol, PLE using ethanol and subcritical water extraction. Finally, different tools are employed for the chemical and functional characterization of the obtained fractions. ## Chemicals and samples HPLC-grade methyl tert-butyl ether (MTBE), methanol, acetone, and ethanol were from VWR (Leuven, Belgium). Sea sand (0.25-0.30 mm diameter) and potassium persulfate were from Panreac. Butylated hydroxytoluene (BHT), formic acid (LC-MS grade), triethylamine (99.5%) and standards of β-carotene, fucoxanthin, chlorophyll a (from Anacystis nidulans algae), ABTS (2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)diammonium salt), D-methionine and Trolox (6-hydroxy-2,5,7,8tetramethylchromane-2-carboxylic acid) were obtained from Sigma-Aldrich (St Louis, MO, USA). The water used was Milli-Q water (Millipore, Billerica, MA, USA). Dichloromethane, chloroform, hexane, methanol, isooctane, and isopropanol were HPLC-grade and purchased from LabScan (Gliwice, Poland). Freeze-dried samples of I. galbana (T-ISO) were obtained from Fitoplancton Marino S.A. (Cadiz, Spain), and stored under dry and dark conditions until further use. I. galbana was grown in outdoor vertical 400 L reactors. Air containing 2% CO 2 is injected into the reactors, while natural light-dark cycles and ambient temperature are used (10-11 h of light, temperatures ranging from 10 to 22 °C). These reactors are inoculated with cultures grown in growth chambers under the standard conditions of Fitoplancton Marino S.A. ## 2.2.1. High pressure extraction processes. All highpressure extractions were carried out in a Speed Helix super-critical fluid extractor from Applied Separations (Allentown, PA, USA). This equipment can be used to perform both SFE (with or without a co-solvent) and PLE. For each extraction, 10 g samples of I. galbana were mixed with 30 g of washed sea sand into a 300 mL basket sandwiched between filter paper. The basket was placed into the high-pressure stainless-steel extraction cell. The CO 2 pneumatic pump pressurizes the CO 2 to the required set value. In the experiments with CO 2expanded ethanol, ethanol was fed by using a liquid pump set at the required volumetric flow rate, and the solvent mixture in the feed tubing was preheated to the extraction temperature. In all experiments, a constant flow rate (5 L min −1 , CO 2 gas) of premier quality CO 2 (Carburos Metálicos, Madrid, Spain) was adjusted at the exit of the extraction cell using a CO 2 gas flow meter. CO 2 extracts were collected in a Falcon tube, while the rest of the extracts were collected in glass bottles. Extractions were performed in four sequential steps using (1) supercritical CO 2 (ScCO 2 ), (2) ScCO 2 /ethanol (CXE), (3) pure ethanol (PLE), and (4) pure water (PLE) as solvents, respectively. The different extraction steps were selected in increasing order of polarity (ScCO 2 < CXE < ethanol < water), to exhaust the microalgae biomass of extractable compounds, fractionating its components in order to give valuable isolated fractions. Step 1: ScCO 2 extraction conditions were optimized using a response surface methodology (RSM) to reveal the functional relationship between the extraction responses (extract yield, total carotenoids and total chlorophylls of extracts) and independent variables (extraction pressure and extraction temperature). A three-level factorial design (3 2 ) was used. The studied factors were pressure (100-300 bar) and temperature (40-60 °C). To determine the extraction time of this step, a kinetic study was performed at the central point of the experimental design (200 bar, 50 °C), collecting the extract every 20 min and calculating the percentage of the extractable material. The parameters of the model were estimated by multiple linear regression using the Statgraphics Centurion XVI software (Statpoint Technologies, Warrenton, Virginia, USA), which allows both the creation and the analysis of experimental designs. Step 2: The second step involved a carbon dioxide expanded ethanol (CXE) extraction in order to increase the polarity of the extracted fraction. This step was carried out in the residual biomass from the first step. The pressure was set at 70 bar, while the temperature was maintained at 50 °C to match the optimum temperature used in the first step in order to avoid unnecessary heating or cooling of the system and thus, minimizing operational costs. Three different percentages of ethanol were tested, 15%, 45% and 75%; the extraction time selected was 1 h. The extraction in the center point (45% EtOH) was performed in triplicate for the precision study. Step 3: The residue from the previous extractions was extracted again using PLE at 100 bar and 80 °C for 30 min, using pure ethanol as an extracting solvent. Step 4: In the fourth and last step, PLE was employed using water as a solvent under the same extraction conditions employed in step 3 (100 bar and 80 °C for 30 min). All the collection recipients were protected from light and 0.1% (w/v) BHT was added to the extracts. Finally, the solvent (ethanolic extracts) was evaporated in a rotary evaporator (Buchi, Flawil, Switzerland) or the samples were freeze-dried (water extracts). The extracts were stored at −80 °C to prevent degradation until analysis. ## Conventional extraction method. Conventional acetone extraction was performed (in triplicate) to determine the total extractable compounds in I. galbana using the method of Reyes et al. 20 Briefly, 200 mg of lyophilized algae were mixed with 20 mL acetone containing 0.1% (w/v) BHT in a 50 mL Falcon tube and the mixture was shaken for 24 h in an orbital shaker (DOS-20L, Elmi Ltd, Riga, Latvia) at 250 rpm in the dark. Following the extraction, the exhausted substrate was precipitated out in a refrigerated centrifuge (Sorvall Evolution RC, Thermo Electron, Asherville, NC, USA) operating at 11 952g at 4 °C for 10 min. The supernatant was collected, and the solvent was removed using a stream of N 2 . Dry acetone extracts were weighed and stored at −20 °C. ## Total carotenoid and chlorophyll determination A spectrophotometric method was used to determine the total carotenoid and total chlorophyll concentration, based on their characteristic absorbance. Extracts from steps 1 and 4 were dissolved in methanol at a concentration of 0.1 mg mL −1 , while extracts of steps 2 and 3 were dissolved in methanol at a concentration of 0.05 mg mL −1 . Absorbance of these solutions was recorded at two specific wavelengths, 470 and 665 nm, for carotenoids and chlorophylls, respectively. External standard calibration curves of fucoxanthin (0.5-10 µg mL −1 ) and chlorophyll a (0.5-7.5 µg mL −1 ) were used to calculate the total carotenoid and chlorophyll content. Total carotenoids were expressed as mg carotenoids per g extract, by interpolating the absorbance of the extract at 470 nm in the calibration curve of fucoxanthin. Total chlorophylls were expressed as mg chlorophyll per g extract, by interpolating the absorbance of the extract at 665 nm in the calibration curve of chlorophyll a. ## Analysis of carotenoids and chlorophylls by HPLC-DAD The carotenoid and chlorophyll profile of I. galbana extracts was determined by HPLC-DAD (diode-array detector) according to a method previously described for N. oleoabundans by Castro-Puyana et al. 22 HPLC analyses of the extracts were conducted using an Agilent 1100 series liquid chromatograph (Santa Clara, CA, USA) equipped with a diode-array detector, and using a YMC-C 30 reversed-phase column (250 mm × 4.6 mm inner diameter, 5 μm particle size; YMC Europe, Schermbeck, Germany) and a pre-column YMC-C 30 (10 mm × 4 mm i.d., 5 μm). The mobile phase was a mixture of methanol-MTBE-water (90 : 7 : 3 v/v/v) (solvent A) and methanol-MTBE (10 : 90 v/v) (solvent B) eluted according to the following gradient: 0 min, 0% B; 20 min, 30% B; 35 min, 50% B; 45 min, 80% B; 50 min, 100% B; 60 min, 100% B; 62 min, 0% B. The flow rate was 0.8 mL min −1 while the injection volume was 10 μL. The detection was performed at 280, 450 and 660 nm, although spectra from 240 to 770 nm were recorded using the DAD ( peak width >0.1 min (2 s) and slit 4 nm). The instrument was controlled by LC ChemStation 3D Software Rev. B.04.03 (Agilent Technologies, Santa Clara, CA, USA). Extracts were dissolved in solvent A prior to HPLC analysis at a concentration of 1 mg mL −1 for the extract of steps 2 and 3; the extracts from the first (ScCO 2 ) and fourth steps were analyzed at 10 mg mL −1 (and filtered through 0.45 µm nylon filters). For the calibration curve, twelve different concentrations of fucoxanthin in ethanol, ranging from 0.97 × 10 −4 to 0.2 mg mL −1 , were analyzed using the LC-DAD instrument. ## Identification of carotenoids by HPLC-APCI-MS/MS LC-MS characterization of I. galbana extracts was performed according to the method previously described by Castro-Puyana et al. 22 An Agilent (Santa Clara, CA, USA) 1200 liquid chromatograph equipped with a diode-array detector was directly coupled to an ion trap mass spectrometer (Agilent ion trap 6320) via an atmospheric pressure chemical ionization (APCI) interface. The HPLC conditions employed for performing the analysis were the same as those described in the previous section. MS analysis was conducted with APCI in positive ionization mode using the following parameters: capillary voltage, −3.5 kV; drying temperature, 350 °C; vaporizer temperature, 400 °C; drying gas flow rate, 5 L min −1 ; corona current (which sets the discharge amperage for the APCI source), 4000 nA; nebulizer gas pressure, 60 psi. Full scan was acquired in the range from m/z 150 to 1300. Automatic MS/MS analysis was also performed, fragmenting the two highest precursor ions (10 000 counts threshold; 1 V Fragmentor amplitude). ## Analysis of lipid class compositions by HPLC-evaporative light scattering detection Separation of lipid classes was done using the method described by Castro-Gómez et al. 24 The analysis was performed using an HPLC system (model 1260; Agilent Technologies Inc.) coupled with an evaporative light scattering detector (SEDEX 85 model; Sedere SAS, Alfortville Cedex, France) using prefiltered compressed air as the nebulizing gas at a pressure of 3.5 bar at 60 °C; the gain was set at 3. Two columns were used in series (250 × 4.5 mm Zorbax Rx-SIL column with 5 μm particle diameter; Agilent Technologies Inc.) and a precolumn with the same packing was used. Before analysis, samples were dissolved in CH 2 Cl 2 (5 mg mL −1 ) and 50 μL was injected. The autosampler temperature was maintained at 4 °C, while the column temperature was set at 40 °C. Solvent mixtures and gradients are detailed in ref. 24. ## Antioxidant capacity assay The TEAC (Trolox equivalent antioxidant capacity) value was determined using the method described by Re et al. 25 with some modifications. The ABTS •+ (2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)diammonium salt) radical was produced by reacting 7 mM ABTS and 2.45 mM potassium persulfate in the dark at room temperature for 16 h. The aqueous ABTS •+ solution was diluted with 5 mM sodium phosphate buffer pH 7.4 to an absorbance of 0.7 (±0.02) at 734 nm. Ten microliters of the sample (5 different concentrations) and 1 mL of the ABTS •+ solution were mixed in an Eppendorf vial and 300 μL of the mixture was transferred into a 96-well microplate. The absorbance was measured at 734 nm every 5 min for 45 min in a microplate spectrophotometer reader (Synergy HT, BioTek). "Trolox" (6-hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid) was used as the reference standard and the results are expressed as TEAC values (mmol Trolox equivalents per g sample). These values are obtained from five different concentrations of each sample tested in the assay giving a linear response between 20 and 80% of the blank absorbance. All analyses were performed in triplicate. ## Protein analysis of PLE extracts Protein analysis was performed according to the Dumas method 26 by using a FlashEA 1112 nitrogen analyzer (Thermo Fisher Scientific, Waltham, MA, USA). Ten milligrams of the dry extract were weighed in a cup of tin and tightly pelleted and subsequently analyzed. A calibration curve of D-methionine was used within the range 1-20 mg. A N-to-protein conversion factor of 4.68 was used to calculate total protein from total nitrogen. The N-to-protein conversion factor was obtained by determination of the amino acid composition of I. galbana according to ref. 27. Analyses were performed in duplicate. ## Sugar composition analysis of PLE extracts The hydrolysis of algae extracts was performed according to Saeman et al. 28 75 mg of the extract was hydrolyzed for 1 h in 72% (w/w) H 2 SO 4 at 30 °C and subsequently water was added giving 1 M H 2 SO 4 and the mixture was incubated for 3 h at 100 °C. After hydrolysis the samples were cooled in ice and then centrifuged (3000g, 15 min, at room temperature). The supernatant of each sample was used for analysis of the sugar composition. The neutral sugar composition was determined according to de Keijzer et al. 29 by high performance anion exchange chromatography (HPAEC) using an ICS-3000 ion chromatography HPLC system equipped with a CarboPac PA-1 column (2 × 250 mm) in combination with a CarboPac PA guard column (2 × 25 mm) and a pulsed electrochemical detector in pulsed amperometric detection mode (Dionex, Sunnyvale, USA). A flow rate of 0.3 mL min −1 was used and the column was equilibrated with 17 mM NaOH. Elution was performed in two steps: 0-0.5 min, 17-0 mM NaOH and 0.5-35 min, 0-35 mM NaOH in 0-350 mM sodium acetate. Detection of the monomers was possible after the post column addition of 0.5 M sodium hydroxide (0.2 mL min −1 ). Before analysis samples were diluted (1 : 3) in water and to a 1 mL sample, 2.5 µL 0.1% (w/v) bromophenol blue in ethanol was added. To adjust the pH, solid barium carbonate was added until a clear magenta color was obtained. Subsequently, the solution was filtered using a 0.45 µm PTFE filter. Fucose was used as an internal standard in the case where fucose was not present in the sample. Analysis was performed in duplicate. ## Results and discussion The strategy has been selected considering the compounds of interest that can be found in I. galbana (such as lipids, proteins, carbohydrates and carotenoids, mainly fucoxanthin and its isomers), the need for re-extracting the residual biomass from the previous extraction step, and the use of green solvents with increasing polarity. Experimental conditions of the different extraction steps were either optimized or selected according to the previous results obtained in our laboratory for the extraction of similar compounds in other microalgae samples. Moreover, minimization of operational and energy costs was also considered in the integrated process, thus minimizing heating/ cooling operations and collection or treatment of the microalgae biomass. ## Optimization of supercritical CO 2 extraction of Isochrysis galbana (step 1) As mentioned above, SFE using CO 2 as a solvent is considered a green process for the extraction of non-polar compounds from natural sources. 30,31 With the objective of maximizing the extraction of the less polar fraction of I. galbana biomass, supercritical CO 2 extraction conditions were optimized using a three-level factorial design (3 2 ). Extraction time was selected after performing a kinetic study under the central conditions (200 bar, 50 °C) measuring the percentage of the extractable material vs. extraction time by collecting samples every 20 minutes (data not shown). An extraction time of 60 min was selected as the most appropriate since after that time the amount of extracted material did not increase. Table 1 shows the experimental design employed, together with the results of the different response variables measured, i.e. extraction yield and total carotenoids and total chlorophyll content. As shown in Table 1, extraction yields ranged from 0.31 to 5.00% while the carotenoid content can be as high as 16.15 mg per g extract at a pressure of 300 bar and medium temperature (50 °C). This is in agreement with the previous results obtained for the extraction of carotenoids from other microalgae such as D. salina. 32,33 After performing the ANOVA (evaluation of the experimental design with Statgraphics Centurion XVI software) for each of the responses (data not shown), the statistical model was fitted and optimized. Considering that the goal of the first step was to maximize the yield and carotenoid content, while minimizing chlorophylls, a desirability function was selected for meeting these goals and giving to all responses the same weight. As shown in Fig. 1, this function provided an optimum of 299 bar and 51 °C to increase the extraction yield and carotenoid content while minimizing the chlorophyll content. The optimization desirability was equal to 0.66, while the values predicted by the model under the optimum extraction conditions were 4.41% for extraction yield, carotenoid content of 16.4 mg carotenoids per g extract and 4.3 mg chlorophylls per g extract for total chlorophylls. Experiments under the optimum conditions provided experimental values close to that predicted by the statistical model (Table 1, experiment 300.50). ## Design of the conditions of sequential extraction of Isochrysis galbana (steps 2-4) Following this first step, three sequential extractions were studied in order to further fractionate the biomass achieving extracts with different compositions. The second step was selected to increase the polarity of the solvent mixture while taking advantage of the intermediate conditions, such as those provided by GXLs that allow working at lower pressures than those of SFE and using smaller volumes of solvents (compared to PLE). This approach has already been successfully applied to the extraction of astaxanthin from H. pluvialis microalgae. 20 Thus, for the second step, a pressure of 70 bar was selected, which is lower than the CO 2 critical pressure (73.8 bar). The temperature was fixed at the optimum value of the first step (50 °C) in order to minimize energy consumption due to heating or cooling of the system. Three different percentages of ethanol, corresponding to low (15%), medium (45%) and high (75%) levels were tested to fully study the possible advantages offered by this intermediate process. Steps 3 and 4 were performed under PLE conditions, using ethanol and water, respectively, which implies an increasing order of polarity. At this point, different bioactive compounds were sought such as polar lipids, proteins and carbohydrates. Moreover, the final objective was to extract all the valuable components contained in the microalgae biomass attaining different fractions and minimizing the leftovers. The extraction values selected included a pressure of 100 bar and a temperature of 80 °C. These values were maintained relatively low in order to avoid degradation of compounds. The scheme of the overall extraction process, along with the target compounds expected in each step is depicted in Fig. 2. 2 and 3. These are corresponding to the pigments detected in steps 1 and 2, ScCO 2 extraction and CXE extraction using 45% ethanol, respectively. ScCO 2 extracted mainly carotenoids from I. galbana (see Table 2). Fucoxanthin isomers ( peaks 4-7) and diadinoxanthin derivatives ( peaks 11-13) could be tentatively assigned due to their UV and MS/MS spectra. Besides, pheophytin a′ ( peak 23) was tentatively identified in the extract in agreement with its [M + H] + ion. Other carotenoids also present in the extract could not be positively identified due to the lack of enough ionization efficiency. Chromatographic profiles are shown in Fig. S1 (ESI, † step 1). Since the percentage of ethanol in the CXE step did not affect the chromatographic profile, the HPLC-DAD chromatogram obtained for 45% ethanol in CO 2 has been used to illustrate the identification of carotenoids and chlorophylls in the second step of the sequential extraction (see Fig. S1, step 2, ESI †). Fucoxanthin was again the main compound present in the extracts, but several chlorophylls and chlorophyll derivatives were also detected (see Table 3). The protonated molecule [M + H] + was not observed for any of the fucoxanthin isomers. Interestingly, E-and 13(′)Z-fucoxanthin isomers showed the same parent ions, corresponding to the dehydrated molecule ([M + H-H 2 O] + ) and a fragment corresponding to a loss of 78 Da consistent with the sequential losses of the C-3 carbomethoxy group (acetic acid) and a water molecule. MS/MS analyses of these ions exhibited a loss of 92 Da that could be attributed to the loss of toluene from the polyene chain. Fucoxanthin metabolite fucoxanthinol (Table 3, peak 3) was tentatively identified by its protonated molecule. MS/MS analysis of fucoxanthinol led to dehydration of the molecule. Diadinoxanthin ( peak 11) was also identified in the extracts by the presence of its typical ions at m/z 583.6 ([M + H] + ) and m/z 565.6 ([M + H-H 2 O] + ). The same MS spectrum was obtained for peaks 12 and 13, but for these peaks, a hypso- chromic shift of 15-20 nm was observed in all UV maxima. Therefore, these compounds can be tentatively identified as 5,8-epoxy derivatives of diadinoxanthin, according to Crupi et al. 34 Chlorophyll a and its epimer chlorophyll a′ ( peaks 15 and 17) lost the phytyl group (C 20 H 39 ) 35 and showed the same fragment, m/z 615.5, which corresponds to the chlorophyllides a and a′, respectively. Besides, the loss of the phythyl group (C 20 H 39 ) can also be used for the identification of pheophytins a and a′ ( peaks 22 and 23). 36 The identification of chlorophyll a in the extract was confirmed by using a commercial standard, and thus peak 10 was assigned to chlorophyll a′. The same elution order was considered for pheophytins a and a′. Several chlorophyll c pigments were tentatively identified in the extracts, although no information could be obtained from the MS in this case. Nevertheless, they were grouped in chlorophyll c 1 -like ( peak 24) and chlorophyll c 2 -like ( peaks 26 and 28) compounds, on the basis of their UV-VIS spectra, since the band ratios (II/III) and the position of maxima are different. The ratios of band II (at ∼630 nm) to band III (at ∼580 nm) intensities are >1 for Chl c 1 -like chromophores, ≈1 for Chl c 2like chromophores and <1 for Chl c 3 -like chromophores. 37 3.3.2. Quantification of total carotenoids, total chlorophylls and fucoxanthin in the different extracts obtained (steps 1-4). Fig. S1 (ESI †) shows the chromatographic profile obtained for the analysis of pigments (carotenoids and chlorophylls) in the extracts obtained for the four different extraction steps. In general, the concentration of carotenoids (mainly fucoxanthin) in the second extraction step is higher compared to the first step, although the amount of chlorophylls (marked with an asterisk) is also higher. The main compounds determined correspond to carotenoids, most-notably fucoxanthin isomers, E-fucoxanthin being the most abundant compound by far. It is interesting to note that different bioactivities have been assigned to the fucoxanthin isomers, as 13Z and 13′Z isomers, which exert higher antiproliferative effects in various cancer cell lines, compared to the E isomer. 38 For this reason, the quantification of each isomer should be of interest. Bearing this in mind, the different fucoxanthin isomers were quantified for the different experimental conditions, in order to evaluate the difference (if any) in selectivity achieved under the different extraction conditions. The vast majority of total fucoxanthin is formed by E-fucoxanthin, while the amount of the other isomers remains very low under the different extraction conditions, except for extractions at 300 bar and 60 °C (data not shown). Under these con-ditions, the sum of 13(′)Z isomer concentration is higher, although still extremely low compared to E-isomers, which could be due to an increase in their solubility under these extraction conditions. In general, the highest extraction of fucoxanthin occurred at 300 bar and 50 °C, over the experimental range that was explored. Table 4 shows the quantification of fucoxanthin isomers, the total carotenoid amount and the total chlorophyll content of the extracts obtained after each step of the sequential integrated process. The highest total chlorophyll content (expressed as chlorophyll a) was found in the CXE extract obtained using 15% ethanol, while the highest content of total carotenoids (expressed as fucoxanthin) was obtained in the CXE extract containing 75% ethanol. In any case, total carotenoids and chlorophylls extracted with carbon dioxide expanded ethanol were higher than total carotenoids and chlorophylls extracted with acetone (146.58 vs. 57.19 mg per g extract and 96.56 vs. 44.48 mg per g extract, respectively, for carotenoids and chlorophylls). On the other hand, the highest content of E-fucoxanthin was found in the CXE extract containing 45% ethanol (40.69 ± 2.28 mg per g extract), and is comparable to the concentration of E-fucoxanthin obtained with acetone conventional solid-liquid extraction (44.60 ± 2.68 mg per g extract). Regarding Z isomers, the sum of 13Z + 13′Z isomers, as well as the amount of 9(′)Z isomers, is higher in acetone extracts, compared to CXE extracts. The content of fucoxanthinol, however, is comparable between acetone and CXE extracts. On the other hand, pooling both ethanol containing extracts (steps 2 and 3), the content of fucoxanthin isomers surpasses acetone extractions, thus validating the use of this new type of green technology for extraction of high value-added compounds. It is worth mentioning that the content of E-fucoxanthin in any of the CXE extracts (36-43 mg g −1 ) was higher than that previously reported for I. galbana using acetone extraction 39 and for Isochrysis sp., using conventional extraction with methanol. 34 3.3.3. Lipid profile at the different steps of the integrated process. The method employed for the analysis of all sequential extracts allows, not only the separation of lipid classes, but also further separation of polar lipidsas phospholipidsin the same run. An example of the chromatograms obtained for each sequential step is shown in Fig. 3, where it is clearly shown that different lipid profiles were achieved for each extraction step. Chromatograms have been divided in three segments in order to facilitate the discussion of the results. In the first segment, eluted triacylglycerides (TAGs); medium polar lipids as mono-(MAGs) and diacylglycerides (DAGs) eluted in the second segment, together with free fatty acids (FFAs), carotenoids and chlorophylls; finally, polar lipids eluted in the third segment of the chromatogram. In the first step of the sequential process, corresponding to ScCO 2 extraction, TAGs were mainly extracted, while polar lipids are not detected at all. In the second (CXE) and third (PLE with 100% ethanol) steps, a similar profile is observed: medium polar compounds and polar lipids were extracted, with a small residue of triacylglycerides. Finally, as expected, lipids were not found in the water extracts obtained in the last step. protein), thus confirming that the percentage of ethanol used in the previous step (CXE) did not affect the extraction, although a slightly higher amount of total protein is observed for CXE-75% ethanol compared to the others. These results are displayed in Fig. 4. As can be seen, the subsequent water PLE extracts showed approximately double the amount of total protein (14-18% (w/w)) than the PLE-ethanol extracts. The sugar composition of ethanol and water PLE extracts was similar. Detailed results are shown in Table 5. Fucose, glucuronic acid, galacturonic acid, N-acetylglucosamine, N-acetylgalactosamine, glucosamine and galactosamine not detected. Xylose (only present in CXE75-water) and mannose were found only in water extracts. A slightly higher amount of total sugars can be observed in the extracts obtained after CXE-75% ethanol compared to the extracts obtained after 15% and 45% ethanol. In any case, the total amount of sugars did not exceed 10% of the extract weight (see Fig. 4). Galactose is the main sugar in ethanol extracts, ranging from 5.69 to 6.68% of dry weight. In water extracts, galactose is present in a smaller amount (1.44-2.83% dry weight), while glucose is the main sugar found (3.91-4.11% dry weight). The results corresponding to the antioxidant capacity assay (expressed as TEAC, mmol of Trolox per g sample), are shown in Table 4. As can be seen, ethanol extracts contained twice the activity as water extracts. This observation cannot be directly related to the total content of sugars, which was similar in both water and ethanol extracts. However, a different composition of sugars in ethanol and water extracts can be expected. Since ethanol is commonly used to precipitate polymeric sugars, monomers or oligomers may be preferably present in ethanol extracts, while oligomeric and polymeric sugars can be expected in water extracts. The total content of protein was lower in ethanol extracts, but proteins extracted in ethanol can be different from proteins present in water extracts, and therefore the activity can be different, too. On the other hand, the amount of fucoxanthin and total carotenoids in ethanol extracts is more than two times higher than the concentration of carotenoids in water extracts. Consequently, despite the fact that there is no linear relationship between the carotenoid content and antioxidant activity, data seem to indicate that higher antioxidant activity in ethanol extracts might be related to the fucoxanthin and fucoxanthin isomer content; Zhang et al. 40 and Sachindra et al. 41 previously confirmed the potent antioxidant activity of these compounds by using different methods. ## Concluding remarks A downstream processing platform is described for the first time to extract bioactive compounds from the microalga I. galbana using GRASgenerally recognized as safesolvents and pressurized technologies. Extractions were performed in four sequential steps using (1) supercritical CO 2 (ScCO 2 ), (2) ScCO 2 /ethanol (Gas Expanded Liquids, GXLs), (3) PLE with pure ethanol, and (4) PLE with pure water as solvents, considering the residue of the previous extraction step as the raw material for extraction. The results obtained showed that the extraction process was partially selective according to the polarity of the solvent/mixture of solvents used. ScCO 2 extracts were rich in triacylglycerides and showed less carotenoid and chlorophyll contents than ethanolic extracts. The main carotenoid identified was fucoxanthin which was found in highest amount in CXE extracts obtained with 45% ethanol. Steps 3 and 4 provide with extracts enriched in proteins and carbohydrates. Further studies should be carried out to determine more in depth the composition of the obtained extracts and their relationships with the antioxidant activity. Also, from our point of view, a scaling up to the industrial level of the process will be of interest.

chemsum

{"title": "Downstream processing of Isochrysis galbana: a step towards microalgal biorefinery", "journal": "Royal Society of Chemistry (RSC)"}

effects_of_π-extension_on_pyrrole_hemithioindigo_photoswitches

## Abstract: The most red-shifted hemithioindigo photoswitches have been identified through systematic introduction of aryl units to a parent pyrrole hemithioindigo photoswitch. Increasing the size of the 5'-aryl substituent is ineffective at producing further redshifted chromophores. A second generation of 3',5'-diarylated photoswitches which possess increased tunability is reported. Experimental and computational evidence indicates the 4' position is electronically isolated from the bulk of the conjugated system. ## Introduction: Visible-light activated small molecule photoswitches comprise a class of molecular machines that are the subject of a great deal of interest in fields such as drug delivery, data storage, and photomechanical polymers. 1 Seminal advances in this field were accordingly recognized with the Nobel Prize in Chemistry in 2016. 2 A particular subclass of these compounds, E/Z-type photoswitches, are of particular interest for controlling biological systems due to the large geometric change conferred by double-bond isomerization. 3 Hemithioindigo photoswitches have recently received considerable attention in part due to their longer wavelength absorption relative to the more well explored azobenzenes. 4 Our laboratory has recently reported a new class of E/Ztype photoswitches which are designed to possess a key intramolecular hydrogen-bonding interaction in only one of the two isomeric states (Figure 1). 5 As a result of this interaction, these pyrrole hemithioindigo (PHTI) photoswitches can undergo quantitative photoisomerizations using visible light for both isomerization reactions. While the photoswitches we reported undergo quantitative isomerization using blue light for ZàE photoisomerization and red light for EàZ photoisomerization, it would be ideal to induce both isomerizations using light in the infrared window for subsurface drug delivery applications. 6 The amino-substituted photoswitches described in our previous report undergo isomerization at longer wavelengths than their oxygen-substituted analogs, however, they also undergo photobleaching upon repeated irradiation. Therefore, we decided not to test even more electron rich substrates due to photobleaching concerns. Rather, we chose to explore the effects of creating more conjugated π-systems as a strategy towards longer-wavelength photoswitches. 7 Upon reinvestigation of the previously described synthesis of arylpyrrole HTIs, it was found that the condensation of appropriate pyrrole-2-carboxaldehydes with benzothiophen-3-one was adequately catalyzed by piperidine, rather than using stoichiometric DBU. This obviated the need for removal of the DBU-water adduct and provided improved yields and purity (Figure 2). For example, while 2a was previously obtained in 44% using DBU as base, 2b-h were each obtained in yields not lower than 76%. If was found that replacement of the 5'-phenyl group (2a) with a 5'-(1-naphthyl) moiety (2b) provided a photoswitch with shorter wavelength absorption maxima. 4 The 2-naphthyl isomer 2c, proved superior, affording longer absorption maxima than 2a or 2b. Anthracyl derivatives 2d and 2e were prepared, and a similar trend was observed, with the longer end-to-end 2-anthracyl PHTI 2e absorbing at longer wavelength than the 9-anthracyl analog 2d (Z: 474 nm à 507 nm; E: 525 nm à 556 nm). This came at the cost of a reduced bathochromic shift and poor photostationary state selectivity. Further extension of the π-system with a pendant 1-pyrenyl moiety (2f) provided only a minimal redshift compared to the 5'-phenyl PHTI. ## Results and Discussion Considering the drawbacks of such polycyclic arenes, namely poor solubility and step-intensive routes for tuning polycyclic aromatic hydrocarbons via substituent effects, we decided against further exploration of these avenues of π-extension. At this point instead of employing larger fused aromatics, we turned our attention to exploring biphenyl type moieties. In our previous report, we had previously observed that the introduction of an aromatic group at the 5'-position not only resulted in a redshift, but also an augmented bathochromic shift relative to a photoswitch without substitution on the pyrrole. It had been hypothesized that this increased bathochromic shift was the result of the increased change in geometry of the longer π-system. Therefore, 2g was synthesized to see if this effect would manifest itself further. However, 2g was found to possess a redshift of only 4-6 nm of either isomer was observed relative to the 5'-phenyl PHTI (2a). Introduction of an alkynyl linker (2h) led to a further small redshift, potentially due to diminished out-of-plane distortion of the two arenes. Unfortunately, these substantial increases in end-to-end distance change upon isomerization did not translate to a substantial increase in the bathochromic shift of the two isomers. Although some redshifting was observed with these compounds, the solubility and synthetic challenges associated with the fused arenes are still present, albeit to a somewhat lesser degree. Therefore, we hypothesized that instead of extending the 5'-substituent, installation of an additional arene moiety on the pyrrole could induce the desired redshift. Synthesis of the 4,5-diaryl photoswitches 4a-b was accomplished by double Suzuki-Miyaura coupling of 4,5dibromopyrrole-2-carboxaldehyde and then aldol condensation with benzothiophen-3-one. Diphenyl photoswitch 4a (Figure 3), however, showed no discernible redshifting over its monoarylated congener 2a. Installation of electron-donating methoxy substituents at the para position of both arenes also proved inferior to the single 5'-p-methoxy photoswitch. Steric distortion between the 4',5'-diaryl groups may lead to these poor properties, thus, a library of 3',5'-diaryl photoswitches was synthesized. The precursor diarylpyrrole-2carboxaldehydes are conveniently synthesized from a chalcone starting material, enabling the synthesis of differentially substituted diarylpyrrole photoswitches with complete regioselectivity. 8 Fortuitously, as shown in Figure 4, it was found that introduction of a 3'-aryl group (3a) provided a 10 nm redshift when compared to 2a (Z: 491 nm à 501 nm; E: 550 nm à 560 nm). In addition to this redshift, introduction of a methoxy to the 3'-arene (3b) substituent provides a 4 nm redshift, roughly half as large of the 10 nm shift induced by addition of a methoxy group to the 5'-arene (3c). Remarkably, these substituent effects appear to be additive, with bis-p-methoxy photoswitch 3d displaying a redshift of approximately 15 nm relative to the unsubstituted 3a. Like the previously disclosed first-generation 5'-aryl photoswitches, these compounds undergo highly selective photoisomerization in both directions using visible light. These photoswitches are also considerably more soluble in organic solvents. This increase in solubility is sufficient for convenient observation and quantitation of the photoisomerization via 1 H NMR (Figure 5). As described above, our initial hypothesis was that out-of-plane distortions between the pendant aryl groups in 4a and 4b lead to diminished conjugation such that two non-planar aryl groups provided similar redshifting to a single, more co-planar π-extension. Separating these groups as in the series 3a-d allowed their effects to be additive. However, when monoarylated photoswitches 5 and 6 were synthesized, we were surprised to observe that even in the absence of an out-of-plane distortion, 4'-phenyl switch 5 was markedly less redshifted than its 5'-phenyl isomer 2a (Figure 6). Photoswitch 6 was slightly redshifted relative to 5, although still substantially blueshifted relative to 2a. With these observations in mind, we turned to DFT calculations for insight into the origin of these electronic effects. Structures for photoswitches 5, 6, 2a, 4a, and 3a, were optimized at the B3LYP/6-31G* level of theory. While the out-of-plane distortions in 4a were apparent, another striking trend was observed. Regardless of the presence of other arenes, the coefficients of the LUMO on the 4' aryl groups were minimal. This indicates that electronic insulation of the 4' position is responsible for the poor performance of photoswitches possessing aryl groups at this position. The extent of the LUMO on 3' phenyl groups was diminished but non-zero, in line with the reduced redshifts observed by introduction of electron-donating substituents at this position (3b, 3d). However, for PHTIs, TD-DFT calculations generally underestimate the absorbance maxima by around 60-70 nm. ## Conclusion In conclusion, we have mapped the effects of π-extension on the pyrrole moiety of pyrrole hemithioindigos. Previously reported 5'-arylated pyrrole hemithioindigos are a uniquely selective class of visible-light photoswitches. Substitution of the pyrrole unit with an aryl group leads to increasing conjugation in the following order: 4' < 3' < 5'. Experimental and computational evidence suggests the 4' position is mostly electronically isolated from the system. Combining 3' and 5' aryl substitution results in a system with improved solubility that maintains the high selectivity of the first generation while providing more opportunities for tuning and derivatization. Importantly, the computational results described herein demonstrate that this system is amenable to redshift predictions based on DFT calculations. ## General Experimental Procedure All reactions were carried out under an inert nitrogen atmosphere with dry solvents under anhydrous conditions unless otherwise stated. All reactions were capped with a rubber septum, or Teflon-coated silicon microwave cap unless otherwise stated. Stainless steel cannula or syringe were used to transfer solvent, and air-and moisture sensitive liquid reagents. Reactions were monitored by thin-layer chromatography (TLC) and carried out on 0.25 mm Merck silica gel plates (60F-254) using UV light as the visualizing agent and potassium permanganate, an acidic solution of p-anisaldehyde, a solution of 2,4-dinitrophenylhydrazine, or vanillin as developing agents. Flash column chromatography employed SiliaFlash ® P60 (40-60 µm, 230-400 mesh) silica gel purchased from SiliCycle Inc. ## Materials All reaction solvents were purified using a Seca solvent purification system by Glass Contour, except for n-butanol. Anhydrous n-butanol was purchased from Sigma-Aldrich and degassed by bubbling N 2 through the solvent while sonicating for 30 minutes. Pd(OAc) 2 was purchased from Strem Chemical Inc. XPhos (CAS: 564483-18-7) was purchased from Oakwood Chemical. All other reagents were used as received without further purification, unless otherwise stated. ## Instrumentation All new compounds were characterized by means of 1 H-NMR, 13 C-NMR, FTIR (thin film), and HR-MS. Copies of the 1 H-and 13 C-NMR spectra can be found at the end of each experimental procedure. NMR spectra were recorded using a Varian 400 MHz NMR spectrometer, Varian 500 MHz NMR spectrometer, or a Varian 600 MHz NMR spectrometer. All 1 H-NMR data are reported in δ units, parts per million (ppm), and were calibrated relative to the signals for residual dichloromethane in CD 2 Cl 2 (5.32 ppm), residual chloroform (7.26 ppm) in deuterochloroform (CDCl 3 ), or residual DMSO (2.50 ppm) in DMSO-d 6 . All 13 C-NMR data are reported in ppm relative to CD 2 Cl 2 (54.0 ppm), CDCl 3 (77.16 ppm), or DMSO-d 6 (39.52) and were obtained with 1 H decoupling unless otherwise stated. The following abbreviations or combinations thereof were used to explain the multiplicities: s = singlet, d = doublet, t = triplet, q = quartet, br = broad, m = multiplet, and a = apparent. All IR spectra were taken on an FT-IR/Raman Thermo Nicolet 6700. High resolution mass spectra (HR-MS) were recorded on a Bruker microTOF mass spectrometer using ESI-TOF (electrospray ionization-time of flight). All UV/Vis spectra were taken on a Cary 3E Spectrophotometer using quartz cuvettes purchased from Starna Cells (P/N: 29-Q-10) and spectrometric or HPLC grade solvents. Photoirradiation was carried out with LEDs purchased from Mouser Electronics (405, 460, 523, 567, 590, 623, 660, and 740 nm) or Roithner-LaserTechnik GmbH (420, 490, 505, 690, and 720 nm). For photostationary state determination, samples were irradiated in borosilicate glass HPLC vials purchased from Thermo Scientific (C4000-1). For part numbers of individual LEDs and detailed description of the irradiation setup, see the accompanying Supplementary Information. HPLC quantitation of photostationary state composition was performed using an Agilent 1260 HPLC with a Chiralpak IA column ((250 x 4.6mm, 5µM particle size). ## General Procedure for Suzuki coupling with solid aryl bromides To a flame-dried 5 mL microwave vial flask equipped with a magnetic stir bar was added N-Bocpyrrole-2-boronic acid (316 mg, 1.5 mmol, 1.5 equiv), Pd(OAc) 2 (4.5mg, 0.02 mmol, 0.02 equiv), XPhos (19.0 mg, 0.04 mmol, 0.04 equiv), K 3 PO 4 (424 g, 2.0 mmol, 2.0 equiv), and aryl bromide (1.0 mmol, 1.0 equiv). The flask was evacuated and backfilled with nitrogen three times, and then the gas line adapter was quickly replaced with a rubber septum and a balloon of nitrogen. To the flask was added degassed (by sonication for 30 minutes with N 2 sparging), anhydrous n-butanol [2 mL (0.5 M in ArBr)]. The heterogeneous reaction mixture was stirred vigorously for 2 hours and then poured into ethyl acetate (~20 mL). This mixture was filtered through a pad of silica and concentrated under reduced pressure by rotary evaporation. This product was used without further purification. To the flask containing the crude product of the crude Suzuki coupling product was added an ovendried magnetic stir bar. The flask was evacuated and backfilled with nitrogen three times, and then the gas line adapter was quickly replaced with a rubber septum and a balloon of nitrogen. To the flask was added anhydrous THF (10 mL) and then NaOMe (5 wt. % in MeOH; 342 µL, 1.5 mmol, 1.5 equiv). The reaction mixture was stirred until no more starting material was observed by TLC (30 -60 minutes). The reaction was quenched by the addition of sat. aq. NH 4 Cl (10 mL). The organic layer was separated, and the aqueous layer was extracted with Et 2 O (3 x 10 mL). The combined organic layers were then washed water (1 x 10 mL), brine (1 x 10 mL) and dried over anhydrous Na 2 SO 4. . The combined organic layers were then filtered and concentrated under reduced pressure by rotary evaporation. This product was used without further purification. The crude product from the deprotection was dried by azeotropic distillation with benzene under reduced pressure by rotary evaporation three times. This flask was then evacuated and backfilled with nitrogen, and then the gas line adapter was quickly replaced with a rubber septum and a balloon of nitrogen. To this flask was added anhydrous CH 2 Cl 2 (3 mL) and the flask was sonicated to dissolve the solid. To a flame-dried 25mL round bottomed flask equipped with a magnetic stir bar was added anhydrous CH 2 Cl 2 (1 mL), POCl 3 (98 µL, 1.05 mmol, 1.05 equiv), and anhydrous DMF (92µL, 1.2 mmol, 1.2 equiv). After stirring for 5 minutes, the solution of 2-arylpyrrole in CH 2 Cl 2 was transferred by syringe to this flask. Upon addition of the arylpyrrole, the solution immediately became brightly colored (color depending on substrate). After stirring for 8-16 hours, the reaction was concentrated first under a stream of N 2 and then under reduced pressure. When no solvent remained, the flask was backfilled with N 2 and then the gas line adapter was quickly replaced with a rubber septum and a balloon of nitrogen. To this flask was added THF (2 mL) and 3M NaOH (aq) (2 mL). This biphasic mixture was stirred vigorously for 30-60 minutes. To the flask was added H 2 O (5 mL) and Et 2 O (5 mL). The organic layer was separated, and the aqueous layer was extracted with Et 2 O (3 × 10 mL). The combined organic layers were washed with water (1 × 20 mL) and brine (1 × 20 mL). The combined organic layers were then dried over anhydrous Na 2 SO 4, filtered and concentrated under reduced pressure by rotary evaporation. Purification by flash column chromatography on silica gel afforded the corresponding aldehydes. ## General Procedure for Suzuki coupling with liquid aryl bromides To a flame-dried 5 mL microwave vial equipped with a magnetic stir bar was added N-Boc-pyrrole-2-boronic acid (316 mg, 1.5 mmol, 1.5 equiv), Pd(OAc) 2 (4.5mg, 0.02 mmol, 0.02 equiv), XPhos (19.0 mg, 0.04 mmol, 0.04 equiv), and K 3 PO 4 (424 g, 2.0 mmol, 2.0 equiv),. The flask was evacuated and backfilled with nitrogen three times, and then the gas line adapter was quickly replaced with a rubber septum and a balloon of nitrogen. To the flask was added degassed (by sonication for 30 minutes with N 2 sparging), anhydrous n-butanol [2 mL (0.5 M in ArBr)] and aryl bromide (1.0 mmol, 1.0 equiv). The heterogeneous reaction mixture was stirred vigorously for 2 hours at room temperature and then poured into ethyl acetate (~20 mL). This mixture was filtered through a pad of silica and concentrated under reduced pressure by rotary evaporation. This product was used without further purification as described in the general procedure above. ## General Procedure for synthesis of ArylPyrrole-HTI photoswitiches To a 5 mL flame-dried microwave flask was added benzo [b]thiophen-3(2H)-one (0.24 mmol, 0.12 equiv) and 5-aryl-2-formylpyrrole (0.2 mmol, 0.1 equiv). The flask was capped with an aluminum-PTFE crimp cap, sealed, and evacuated and backfilled with nitrogen three times. To the flask was then added anhydrous toluene (2 mL, 0.1 M in aldehyde) and piperidine (10 µL, 0.1 mmol, 0.5 equiv). The flask was transferred to a pre-warmed oil bath set to 111 ˚C and stirred for two hours. After two hours the flask was removed from the oil bath and cooled to room temperature and then to 0 ˚C in a water-ice bath. To the flask was added hexanes (5 mL) and the flask was allowed to sit for an addition 10-30 minutes. The mixture was the filtered, and the precipitate was then triturated with hexanes to until the filtrate ran clear to provide the pure product as a red, blue, or purple solid depending on the substrate. ## Precursor Synthesis Benzo [b]thiophen-3(2H)-one, 4,5-bis(4-methoxyphenyl)-pyrrole-2-carboxalehyde, and 4phenyl-pyrrole-2-carboxaldehyde, and 3-phenyl-pyrrole-2-carboxaldehyde were synthesized according to published procedures. ## 5-(naphthalen-1-yl)-1H-pyrrole-2-carbaldehyde (SI-1): Prepared according to general procedure. Purification by flash column chromatography on silica gel (hexanes/Et 2 O = 100:0 to 80:20) afforded the title compound (158 mg, 71%) as a pink solid. Rf: 0.32 (hexanes:Et 2 O = 1:1) 1 H NMR (600 MHz, CDCl 3 ): δ 9.59 (s, 1H), 9.44 (br s, 1H), 8.20 -8.15 (m, 1H), 7.95 -7.89 (m, 2H), 7.60 -7.49 (m, 4H), 7.13 (dd, J = 3.9, 2.5 Hz, 1H), 6.64 (dd, J = 3.9, 2. 5 To a flame-dried 20 mL microwave vial equipped with a magnetic stir bar was added N-4,5-dibromopyrrole-2-carboxaldehyde (121 mg, 0.48 mmol, 1.0 equiv), phenylboronic acid (292 mg, 2.4 mmol, 5.0 equiv), Pd(PPh 3 ) 4 (56 mg, 0.048 mmol, 0.10 equiv), and Na 2 CO 3 (305 mg, 2.88 mmol, 6.0 equiv). The flask was evacuated and backfilled with nitrogen three times, and sealed with an aluminum crimp cap. To the flask was added water (4.2 mL) and dioxane (4.2 mL, final concentration ~0.6M in pyrrole). The reaction apparatus was then transferred into a pre-heated oil bath and stirred at 80˚C for two hours. After two hours the reaction was cooled to room temperature, poured into H 2 O (20 mL), and extracted with EtOAc (3 × 20 mL). The combined organic layers were washed with brine (50 mL), dried over anhydrous sodium sulfate, filtered, and concentrated under reduced pressure by rotary evaporation to provide a crude brown solid. Purification by flash column chromatography on silica gel (Et 2 O/hexanes = 0/10 to 2/8) afforded SI-8 (107.2 mg, 73%) as a tan solid. Rf: 0.37 (hexanes:Et 2 O = 1:1) To a flame-dried 5 mL microwave vial equipped with a magnetic stir bar was added anhydrous CH 2 Cl 2 (1 mL), POCl 3 (98 µL, 1.05 mmol, 1.05 equiv), and anhydrous DMF (92µL, 1.2 mmol, 1.2 equiv). After stirring for 5 minutes, a solution of 2,4-diphenylpyrrole (219 mg, 1.0 mmol 1.0 equiv) in CH 2 Cl 2 (3 mL) was transferred by syringe to this flask. After stirring for 16 hours, the reaction was concentrated first under a stream of nitrogen and then under reduced pressure. When no solvent remained, the flask was backfilled with nitrogen and then the gas line adapter was quickly replaced with a rubber septum and a balloon of nitrogen. To this flask was added THF (2 mL) and 3M NaOH (aq) (2 ml). This biphasic mixture was stirred vigorously for 30-60 minutes. To the flask was added H 2 O (5 mL) and Et 2 O (5 mL). The organic layer was separated, and the aqueous layer was extracted with Et 2 O (3 × 50 mL). The combined organic layers were washed with water (1 × 100 mL) and brine (1 × 100 mL). The combined organic layers were then dried over anhydrous Na 2 SO 4 , filtered and concentrated under reduced pressure by rotary evaporation. Purification by flash column chromatography on silica gel (hexanes/Et 2 O = 100:0 to 70:30) afforded the title compound (132 mg, 53%) as a pink solid. Rf: 0.39 (hexanes:Et 2 O = 1:1) 1 H NMR (600 MHz, CDCl 3 ): δ 9.65 (s, 1H), 9.49 (br s, 1H), 7.65 (d, J = 8.4 Hz, 2H), 7.55 (d, J = 8.2 Hz, 2H), 7.47 (at, J = 7.6 Hz, 4H), 7.43 -7.35 (m, 2H), 6.74 (dd, J = 2.9, 1. 2 Aldehyde SI-10 was prepared from 2.0 mmol 3-(4-methoxyphenyl)-4-nitro-1phenylbutan-1-one according to O'Shea's procedure. Without purification, this crude diarrylpyrrole was then formylated as per the procedure for compounds SI-1-SI-7. Purification by flash column chromatography on silica gel (hexanes/Et 2 O = 100:0 to 70:30) afforded the title compound (164 mg, 29.5% over three steps) as a pink solid. Rf: 0.25 (hexanes:Et2O = 1:1) 1 H NMR (600 MHz, CDCl 3 ): δ 9.62 (d, J = 1.1 Hz, 1H), 9.39 (s, 1H), 7.63 (d, J = 7.9 Hz, 2H), 7.52 -7.43 (m, 4H), 7.43 -7.34 (m, 1H), 7.04 -6.92 (m, 2H), 6.69 (dd, J = 2.8, 1.1 Hz, 1H), 3.87 (s, 3H). 13 C NMR (151 MHz, CDCl 3 ): δ 179. Aldehyde SI-11 was prepared from 1.0 mmol 1-(4-methoxyphenyl)-4-nitro-3phenylbutan-1-one according to O'Shea's procedure. Without purification, this crude diarrylpyrrole was then formylated as per the procedure for compound SI-1-SI-7. Purification by flash column chromatography on silica gel (hexanes/Et 2 O = 100:0 to 70:30) afforded the title compound (68 mg, 24.5% over three steps) as a pink solid. Rf: 0.25 (hexanes:Et 2 O = 1:1) 1 H NMR (600 MHz, CDCl 3 ): δ 9.61 (s, 1H), 9.43 (br s, 1H), 7.65 -7.57 (m, 2H), 7.55 (dd, J = 6.9, 1. 5 Aldehyde SI-12 was prepared from 2.0 mmol 3-(4-methoxyphenyl)-4-nitro-1phenylbutan-1-one according to O'Shea's procedure.* Without purification, this crude diarrylpyrrole was then formylated as per the procedure for compound SI-1-SI-7. Purification by flash column chromatography on silica gel (hexanes/Et2O = 100:0 to 50:50) afforded the title compound (269.6 mg, 17% over four steps) as a purple solid. *Note: In a variation from O'shea's procedure, the Paal-Knorr cyclization step was conducted at 50˚C for 30 minutes to minimize degradation, and then immediately formylated. Rf: 0.11 (hexanes:Et 2 O = 1:1) 1 H NMR (600 MHz, CDCl 3 ): δ 9.58 (s, 1H), 9.35 (s, 1H), 7.57 (d, J = 8. 7 ## Photoswitch 2e: Prepared according to general procedure. Obtained as a red solid (62.0 mg, 77%). 1 H NMR 1 H NMR (600 MHz, DMSO-d 6 ) δ 12.46 (br s, 1H), 8.57 (s, 2H), 8.47 (s, 1H), 8.17 (d, J = 8.9 Hz, 1H), 8.13 (d, J = 7.3 Hz, 1H), 8.10 -8.07 (m, 1H), 8.01 (s, 1H), 7.95 (d, J = 8.9 Hz, 1H), 7.84 (d, J = 7.6 Hz, 1H), 7.81 (d, J = 7.9 Hz, 1H), 7.71 (t, J = 7.5 Hz, 1H), 7.57 -7.47 (m, 2H), 7.41 (t, J = 7.4 Hz, 1H), 7.20 -7.18 (m, 1H), 7.00 -6.87 (m, 1H). 13 C NMR (126 MHz, DMSO-d 6 @ 120˚C): 186.9, 144. ## Photoswitch 3a: Prepared according to general procedure (0.25 mmol scale of aldehyde). Obtained as a red/purple solid (77.0 mg, 73 %). 1 H NMR (500 MHz, CD 2 Cl 2 ): δ 8.03 (d, J = 7.9 Hz, 1H), 7.90 (d, J = 7.7 Hz, 2H), 7.54 (ddt, J = 20.5, 14.7, 7.3 Hz, 8H), 7.42 (q, J = 7.1 Hz, 2H), 7.35 (d, J = 3. 4 ## Photoswitch 3d: Prepared according to general procedure. Obtained as a purple solid (67.5 mg, 77%).

chemsum

{"title": "Effects of \u03c0-Extension on Pyrrole Hemithioindigo Photoswitches", "journal": "ChemRxiv"}

thorium–ligand_multiple_bonds_via_reductive_deprotection_of_a_trityl_group

## Abstract: Reaction of [Th(I)(NR 2 ) 3 ] (R ¼ SiMe 3 ) (2) with KECPh 3 (E ¼ O, S) affords the thorium chalcogenates, [Th(ECPh 3 )(NR 2 ) 3 ] (3, E ¼ O; 4, E ¼ S), in moderate yields. Reductive deprotection of the trityl group from 3 and 4 by reaction with KC 8 , in the presence of 18-crown-6, affords the thorium oxo complex, [K(18crown-6)][Th(O)(NR 2 ) 3 ] (6), and the thorium sulphide complex, [K(18-crown-6)][Th(S)(NR 2 ) 3 ] (7),respectively. The natural bond orbital and quantum theory of atoms-in-molecules approaches are employed to explore the metal-ligand bonding in 6 and 7 and their uranium analogues, and in particular the relative roles of the actinide 5f and 6d orbitals. ## Introduction The study of actinide-ligand multiple bonds has intensifed in recent years due to the need to understand the extent of both forbital participation and covalency in actinide-ligand bonding. In this regard, the past ten years have seen considerable progress in the synthesis of oxo, imido, carbene, and nitrido complexes of uranium. More recently, several terminal phosphinidene 36,37 and chalcogenido (S, Se, Te) complexes of uranium have also been isolated, demonstrating that this chemistry can be extended to the heavier main group elements. Despite these advancements, multiply-bonded complexes of the other actinides remain rare. 43 Only one thorium terminal oxo complex is known, namely, [h 5 -1,2,4-t Bu 3 C 5 H 2 ] 2 -Th(O)(dmap) (dmap ¼ 4-dimethylaminopyridine), which was recently reported by Zi and co-workers. 44 In addition, a handful of terminal imido complexes have been isolated, 45 including [Cp* 2 Th(NAr)(THF)] (Ar ¼ 2,6-dimethylphenyl), which was reported by Eisen and co-workers in 1996. 46 A few thorium carbene complexes are also known, but in each example the carbene ligand is incorporated into a chelating ligand, which kinetically stabilizes the Th]C bond. Also of note, terminal thorium sulphides have been invoked as reaction intermediates, 44 but have not been isolated. This paucity of examples can be rationalized by the higher energy of the thorium 5f orbitals, relative to uranium, which likely weakens metal-ligand p-bonding. 50 However, this hypothesis requires further verifcation, highlighting the need for new complexes that feature thorium-ligand multiple bonds. Recently, we reported that selective removal of the trityl protecting group from the U(IV) alkoxide, [U(OCPh 3 )(NR 2 ) 3 ] (R ¼ SiMe 3 ), allowed for the isolation of the oxo complex, [K(18crown-6)][U(O)(NR 2 ) 3 ]. 41 Signifcantly, the uranium centre does not undergo a net oxidation state change during the transformation. Inspired by this result, we endeavoured to synthesize the analogous thorium oxo complex, and its sulphido congener, using this deprotection protocol. Thorium was chosen for this study, in part, to address the scarcity of multiply-bonded complexes of the other actinides, but also because Th 4+ is effectively redox inactive, which makes the traditional synthetic routes to multiple bonds (such as oxidative atom transfer) more challenging. Herein, we describe the synthesis of a thorium sulphide and a thorium oxo, along with an analysis of their electronic structures by density functional theory. ## Results and discussion Reaction of ThCl 4 (DME) 2 with 3 equiv. of NaNR 2 (R ¼ SiMe 3 ) in THF affords colourless crystals of [Th(Cl)(NR 2 ) 3 ] (1) in 56% yield, upon crystallization from Et 2 O/hexanes. This material was previously prepared by Bradley 51 and Andersen; 52 however, it was never structurally characterized. Crystals of complex 1 suitable for X-ray crystallographic analysis were grown from a concentrated diethyl ether (Et 2 O) solution stored at 25 C for 24 h. Determination of the solid-state structure revealed the anticipated pseudotetrahedral geometry about the thorium centre (see ESI † for full details). In addition, this material has a melting point of 208-210 C, nearly identical to that reported by Andersen and co-workers. 52 Interestingly, crystallization of the reaction mixture from THF/pentane resulted in isolation of the "ate" complex, [Na(THF) 4.5 ][Th(Cl) 2 (NR 2 ) 3 ], as determined by Xray crystallography (see the ESI †). However, this material can readily be converted into 1 upon extraction into, and recrystallization from, Et 2 O. (1) Subsequent reaction of complex 1 with 12 equiv. of Me 3 SiI, in diethyl ether, affords [Th(I)(NR 2 ) 3 ] (2) as a white powder in 95% yield (eqn (1)). A similar procedure was recently used to prepare the related cerium iodide complex, [Ce(I)(NR 2 ) 3 ]. 53 Crystals of 2 suitable for X-ray diffraction analysis were grown from a concentrated diethyl ether solution stored at 25 C for 24 h. Complex 2 crystallizes in the hexagonal setting of the rhombohedral space group R3c, and its solid state molecular structure is shown in Fig. S19. † Complex 2 is isostructural with its chloride analogue 1. Its Th-N distance (2.299(4) ) is identical to that of 1, while the Th-I bond (3.052(1) ) is longer than the Th-Cl bond of 1 (2.647(1) ), consistent with the larger single bond covalent radius of I (1.33 ) vs. Cl (0.99 ). 54 The 1 H and 13 C NMR spectra of 2 each exhibit a single resonance, at 0.45 ppm and 5.13 pm, respectively, assignable to the methyl groups of the silylamide ligands (Fig. S5 and S6 †). We previously reported the synthesis of a U(IV) alkoxide complex, [U(OCPh 3 )(NR 2 ) 3 ], via reaction of KOCPh 3 and [U(I)(NR 2 ) 3 ], 41 and with 2 in hand, we endeavoured to synthesise the analogous thorium alkoxide. Thus, addition of 1 equiv. of KOCPh 3 to a cold (25 C) suspension of 2 in toluene affords a colourless solution, concomitant with the deposition of fne white powder. A colourless oil is obtained upon work-up, and storage of this oil at 25 C for 24 h affords [Th(OCPh 3 )(NR 2 ) 3 ] (3) as a colourless crystalline solid in 33% yield (eqn (2)). Similarly, reaction of complex 2 with 1 equiv. of KSCPh 3 , in toluene, results in the formation of [Th(SCPh 3 )(NR 2 ) 3 ] (4) in 57% yield, after crystallization from hexanes (eqn (2)). (2) We were unable to obtain X-ray quality crystals of 3; however, complex 4 was amenable to an X-ray diffraction analysis. This material crystallizes in the triclinic space group P 1, and features a pseudotetrahedral geometry about the thorium centre (Fig. 1). The Th-S bond length in 4 (2.704(1) ) is similar to those of other thorium thiolate complexes (ca. 2.74). 55,56 In addition, the Th-S-C angle (136.72 (1) ) is rather small, suggesting that there is minimal 3p p-donation from S to Th. Other thorium thiolates also feature similarly acute Th-S-C angles. 55,56 The 1 H NMR spectrum of 3 exhibits a singlet at 0.39 ppm, in benzene-d 6 , assignable to the methyl groups of the silylamide ligands. In addition, it features resonances at 7.09, 7.18, and 7.39 ppm, in a 3 : 6 : 6 ratio, respectively, corresponding to the p-, m-, and o-aryl protons of the trityl-alkoxide ligand (Fig. S7 †), consistent with the proposed formulation. Not surprisingly, the 1 H NMR spectrum of 4, in benzene-d 6 , is almost identical to that of 3, and also features resonances assignable to three silylamide ligands and one trityl moiety (Fig. S9 †). Interestingly, the 1 H NMR spectrum of the trityl-alkoxide reaction mixture exhibits resonances due to a second, minor Th-containing product. This was subsequently identifed to be the bis(alkoxide) complex, [Th(OCPh 3 ) 2 (NR 2 ) 2 ] (5), which is likely formed by reaction of 3 with another equivalent of KOCPh 3 . The 1 H NMR spectrum of 5 features a sharp singlet at 0.26 ppm, in benzene-d 6 , which is assignable to the methyl groups of the silylamide ligands (Fig. S11 †). This resonance is slightly upfeld from that observed for complex 3, which allows 5 to be distinguished from that complex. Complex 5 was also characterized by X-ray crystallography (see ESI †). Interestingly, there is no evidence for the formation of the analogous uranium complex in the reaction of KOCPh 3 with [U(I)(NR 2 ) 3 ], 41 consistent with the reduced ionicity of the U-N bond vs. the Th-N bond (see also below), which increases the barrier for ligand scrambling in uranium. Complex 1 can also be used as a precursor to 3, but in this case even greater amounts of complex 5 are formed during the reaction. Prompted by our aforementioned success at selectively cleaving the C-O bond in [U(OCPh 3 )(NR 2 ) 3 ] to afford a uranium oxo complex, 41 we explored the reductive cleavage of the C-E (E ¼ O, S) bonds in complexes 3 and 4. Gratifyingly, reduction of 3 with 2 equiv. of KC 8 , in the presence of 18-crown-6, in THF, results in formation of a vibrant red solution, indicative of the presence of [CPh 3 ] . 41,57 Extraction of the reaction mixture into diethyl ether, followed by fltration, permits removal of the [K(18-crown-6)(THF) 2 ][CPh 3 ] by-product, which is insoluble in this solvent. Work-up of the fltrate affords the thorium oxo complex, [K(18-crown-6)][Th(O)(NR 2 ) 3 ], (6) as colourless blocks in 23% yield (eqn (3)). Similarly, reaction of 4 with 2 equiv. of KC 8 , in the presence of 18-crown-6, in THF, results in the formation of the thorium sulphide, [K(18-crown-6)] [Th(S)(NR 2 ) 3 ] (7), which can be isolated as colourless needles in 62% yield after a similar work-up (eqn (3)). The 1 H NMR spectra of 6 and 7, in benzene-d 6 , both feature two sharp resonances (6 : 0.64 and 3.09 ppm; 7 : 0.74 and 3.17 ppm) in a 54 : 24 ratio, assignable to the methyl groups of the silylamide ligands and the methylene groups of the 18-crown-6 moiety, respectively (Fig. S13 and S15 †), consistent with their proposed formulations. Unfortunately, the Th]E vibrational modes in 6 and 7 could not be defnitively identifed by either IR or Raman spectroscopies. (3) Complex 6 crystallizes in the orthorhombic spacegroup Pbca, as a diethyl ether solvate, 6$0.5Et 2 O, while complex 7 crystallizes in the triclinic spacegroup P 1, with two independent molecules in the asymmetric unit. Their solid state molecular structures are shown in Fig. 2, and selected bond lengths and angles can be found in Table 1. Both complexes feature pseudotetrahedral geometries about their metal centres, along with dative interactions between the chalcogenido ligands and the K + ion of the [K(18-crown-6)] moiety. The Th-O bond length (1.983(7) ) in 6 is slightly longer than the Th-O distance in the other structurally characterized thorium oxo (Th-O ¼ 1.929(4) ), 44 but is signifcantly shorter than a typical Th-O single bond (ca. 2.20 ), suggestive of multiple bond character within the Th-O interaction. Interestingly, the Th-O distance in 6 is 0.09 longer than the analogous distance in [K(18-crown-6)][U(O)(NR 2 ) 3 ] (1.890(5) ), 41 a difference that is greater than the difference in the 4+ ionic radii of these two metals (0.05 ). 64 The Th-S bond lengths in 7 (2.519(1) and 2.513(1) ) are signifcantly shorter than a typical Th-S single bond (ca. 2.74 ), 44,55,56,65 and are again suggestive of multiple bond character within the Th-S interaction. In addition, the Th-S distances in 7 are 0.07 longer than the analogous distances in [K(18-crown-6)]-[U(S)(NR 2 ) 3 ] (2.4463(6) and 2.4513(6) ), 41 which is in-line with the anticipated difference based on ionic radii considerations alone. 64 In order to gain further insight into the electronic structure and bonding of 6 and 7, as well as the uranium analogues 6-U and 7-U, we turned to quantum chemistry in the form of density functional theory (DFT). We began by optimising the geometries of the four target molecules using the PBE functional; selected bond lengths and angles are given in Table 1. For complexes 6 and 6-U the agreement between experiment and theory is very good, with differences in bond length of no more than 0.04 . DFT predicts both molecules to be almost linear along the M-O-K vector, 179.9 and 176.2 for 6 and 6-U respectively, in reasonable agreement with the experimental angles of 167.5(4) and 170.0(3) , respectively. In contrast, 7 and 7-U have two molecules in the asymmetric unit, with very different M-S-K angles. The PBE optimised structures agree very well with the experimental data for the molecules with the smaller M-S-K angles; the deviation from the experimental angles is only ca. 1.5 . In addition, a constrained geometry optimisation of the Th-S-K angle in 7, from the optimised angle of 150.4 , yields converged geometries up to Th-S-K ¼ 170.4 , at which point the molecule is only 2.6 kJ mol 1 less stable than in the fully optimised structure. Given this shallow bending potential, we wondered if the differences between the two molecules in the asymmetric units of 7 and 7-U might arise from dispersion forces, and hence re-optimised all four targets with these included via the Grimme D3 corrections. The data for these structures are collected in Table 1 and show that, with the exception of a slight shortening of the O-K distance, there is almost no difference between the PBE and PBE + D3 structures for 6 and 6-U. By contrast, the inclusion of dispersion corrections signifcantly modifes the geometries of 7 and 7-U, most notably the M-S-K angle, which increases by ca. 30 to linear in both cases, and the E-K distances which, in agreement with experiment, shorten by almost 0.1 between the bent and linear structures. For the latter, calculation predicts the M-E bond length reduction on going from Th to U to be ca. 0.06 in both the oxo and sulphido cases, essentially the same as the difference in ionic radius between Th 4+ and U 4+ , hence underestimating by ca. 0.03 the experimentally determined M-O bond length reduction on going from 6 to 6-U. We have analysed the electronic structures of all four targets using the Natural Bond Orbital (NBO) and Quantum Theory of Atoms-in-Molecules (QTAIM) approaches and, in order to allow for better comparison, decided to focus on the linear forms of 7 and 7-U, i.e. the electronic structures have been analysed at the PBE + D3 geometries for all four molecules. Complexes 6 and 7 are, of course, closed shell species and hence there is no net spin density for these systems; for 6-U and 7-U, however, NBO fnds net spin densities of 2.092 and 2.085 respectively, as expected for U(IV). In all four cases, NBO fnds the M-E interaction to be a triple bond; the s + 2p Th-O natural localised molecular orbitals (NLMOs) in 6 are shown in Fig. 3, and the compositions of the p NLMOs are collected in Table 2 for all four targets. In all cases the orbitals are largely chalcogenbased, a little more so for thorium than uranium. There is clearly more metal involvement in these orbitals in the sulphur systems than the oxygen, and while this is predominantly dbased for thorium there is an almost equal contribution of d and f in 6-U and 7-U. NBO fnds the M-N interactions to have double bond character. Three dimensional representations of one set of Th-N NLMOs in 6 are shown in Fig. 4, and the averaged compositions of the s and p character orbitals are collected in Table 3 for all four targets. As with the M-E bonding, these NLMOs are all strongly polarized toward the nitrogen. There is slightly more uranium contribution than thorium in analogous NLMOs. For the s orbitals, the metal contributions are signifcantly more dbased than f (more so for thorium than uranium), while for the p component there is much more even metal d/f content, with a little more f than d for the uranium NLMOs and vice versa for thorium. The deviations of the actinide natural atomic orbital populations (Natural Population Analysis (NPA)) from their formal values are given in Table 4. Typically, deviations from formal M-E 1.983 (7) populations are taken as a measure of covalency, and such an approach is valid for the early actinides. Table 4 shows that the 7s and 7p orbitals are little involved in bonding. The 6d orbitals have larger deviations from the formal population than the 5f; these are very similar for the two sulphur compounds (1.49 and 1.50), and reduced for the two oxygen compounds, with slightly more 6d in the uranium system than the thorium (1.17 vs. 1.12). A similar situation is found for the 5f populations; the deviations of the sulphur compounds are very similar for thorium and uranium and larger than for the oxygen compounds, for which the uranium 5f population is a little larger than the thorium 5f. In summary, and in agreement with the analysis of the NLMO compositions, these data suggest greater covalency in the sulphur than the oxygen compounds, greater 6d covalency than 5f and, for the latter orbitals, slightly larger covalency in uranium than thorium. Table 5 presents the calculated atomic partial charges, using the QTAIM and NPA approaches. While the absolute values differ between methods, the trends are the same and suggest strongly polar M-E and M-N bonding. Taking the difference in charge between the metal and the surrounding atoms as a measure of ionicity, the data indicate that the bonding in the thorium compounds is more ionic than the uranium, and that bonding in the oxygen systems is more ionic than the sulphur, in agreement with the compositions of the NLMOs, which are more thorium localized than uranium, and more oxygen localized than sulphur. We have pioneered the use of the QTAIM in the study of actinide covalency 2,4 and bond strength, 66,67 and Table 6 collects selected bond critical point (BCP) electron (r) and energy (H) densities and ellipticities (3), and delocalisation indices (d(A, B) -QTAIM measures of bond order). The ellipticity data reinforce the NBO results, indicating cylindrical (or, for 6-U, near cylindrical) triple-bond symmetry for the M-O interactions, and signifcantly non-cylindrical double-bond symmetry for M-N. 68 The M-O BCP electron densities for 6 and 6-U are very large for actinide bonds, bordering the 0.2 au covalency threshold, and the M-N BCP r data are typical. 67,69 For both M-O and M-N, the BCP data are larger in an absolute sense in 6-U vs. 6. This is also true of the delocalisation indices, reinforcing the NBO conclusion of greater covalency in 6-U vs. 6. This is also the case for 7 vs. 7-U; the M-S and M-N QTAIM metrics are all larger in an absolute sense in the uranium system. The M-E r and H and, to a lesser extent, d(A,B) are signifcantly smaller in the sulphur compounds than the oxygen. We have previously cautioned, however, in the context of Th/U-S/Se ## View Article Online bonding, 70 against the interpretation of such reductions in terms of reduced covalency. The QTAIM covalency metrics show very strong dependence on bond length, and we believe that the very signifcant (>0.5 ) difference between M-S and M-O is the dominant factor here. ## Conclusions We have demonstrated the synthesis of oxo and sulphide complexes of thorium via reductive removal of the trityl protecting group. This work further demonstrates the generality of the reduction deprotection methodology, suggesting that this method will be broadly applicable towards the synthesis of multiple bonds in other metal systems, including lanthanides and transition metals, and we are currently exploring this possibility. Quantum chemical analysis (NBO and QTAIM) of the bonding in the thorium systems, and analogous uranium oxo and sulphido molecules, indicates that the M-E interactions are s + 2p triple bonds that are strongly polarised toward the chalcogen, while the M-N bonds (also largely ligand-based) have double bond character. For both the M-E and M-N bonds, there is greater metal-ligand orbital mixing (which, in the early part of the actinide series, we are comfortable describing as covalency) in the sulphur than the oxygen compounds. The Thligand bonds are found to be more ionic than the uranium analogues. Finally, the 6d orbitals play a larger role in the Th-E and Th-N bonds than do the 5f, while the latter are more involved in the uranium-ligand bonding. ## Experimental General All reactions and subsequent manipulations were performed under anaerobic and anhydrous conditions under an atmosphere of nitrogen. Hexanes, Et 2 O, THF, and toluene were dried using a Vacuum Atmospheres DRI-SOLV Solvent Purifcation system and stored over 3 sieves for 24 h prior to use. Benzened 6 was dried over 3 molecular sieves for 24 h prior to use. ThCl 4 (DME) 2 was synthesized according to the previously reported procedure. 71 All other reagents were purchased from commercial suppliers and used as received. NMR spectra were recorded on a Varian UNITY INOVA 400, a Varian UNITY INOVA 500 spectrometer, a Varian UNITY INOVA 600 MHz spectrometer, or an Agilent Technologies 400-MR DD2 400 MHz Spectrometer. 1 H and 13 C{ 1 H} NMR spectra were referenced to external SiMe 4 using the residual protio solvent peaks as internal standards. IR spectra were recorded on a Nicolet 6700 FT-IR spectrometer. Elemental analyses were performed by the Micro-Analytical Facility at the University of California, Berkeley. ## [Th(Cl)(NR 2 ) 3 ] (1) To a colourless, cold (25 C), solution of ThCl 4 (DME) 2 (385.7 mg, 0.70 mmol), in THF (4 mL) was added a cold (25 C) solution of NaN(SiMe 3 ) 2 (381.6 mg, 2.08 mmol) in THF (4 mL). This mixture was allowed to stir for 18 h, whereupon the solvent was removed in vacuo to afford a colourless solid. This solid was triturated with hexanes (3 4 mL) to yield a colourless powder. The resulting powder was extracted with diethyl ether (10 mL) and fltered through a Celite column supported on glass wool (0.5 cm 3 cm). The cloudy fltrate was again fltered through a Celite column supported on glass wool (0.5 cm 3 cm) to give a clear colourless fltrate. The volume of this fltrate was reduced in vacuo to 4 mL and layered with hexanes (5 mL). Storage of this mixture at 25 C for 24 h resulted in the deposition of colourless crystals, which were isolated by decanting off the supernatant (167 mg, 32%). The supernatant was then dried in vacuo to afford a colourless solid. This solid was then extracted with diethyl ether (5 mL) and fltered through a Celite column supported on glass wool (0.5 cm 3 cm) to afford a colourless fltrate. The volume of this fltrate was reduced to 2 mL in vacuo and layered with hexanes (4 mL). Storage of this mixture at 25 C for 24 h resulted in the deposition of an additional batch of colourless crystals, which were isolated by decanting off the supernatant. Total yield: 294.2 mg, 56%. Crystals suitable for Xray crystallographic analysis were grown from a concentrated Et 2 O solution stored at 25 C for 24 h. Melting point: 208-210 C (lit. value ¼ 210-212 C). [ To a stirring suspension of [Th(Cl)(NR 2 ) 3 ] (1) (852.3 mg, 1.14 mmol) in hexanes (8 mL) was added TMSI (2 mL, 14.05 mmol). This mixture was allowed to stir for 96 h, whereupon the solvent was removed in vacuo to afford a white solid. The solid was triturated with pentane (2 3 mL) to yield a white powder (908.2 mg, 95%). Crystals suitable for X-ray crystallographic analysis were grown from a concentrated CH 2 Cl 2 solution stored at 25 C for 24 h. Anal. calcd for C 18 H 54 IN To a colourless, stirring suspension of 2 (231.4 mg, 0.28 mmol) in toluene (4 mL) was added a cold (25 C) solution of KOCPh 3 (84.7 mg, 0.28 mmol) in toluene (4 mL), in two portions over the course of 1 h. This mixture was allowed to stir for another hour, resulting in the deposition of a fne white powder. An aliquot (0.25 mL) of the reaction mixture was taken, the solvent was removed in vacuo, and a 1 H NMR spectrum in benzene-d 6 was recorded. This spectrum indicated the presence of starting material, complex 3, and a small amount complex 5. The amount of remaining starting material was estimated from relative area of its silylamide resonance, whereupon an additional portion of KOCPh 3 (13.4 mg, 0.045 mmol) was added to the reaction mixture. After 1 h of stirring, this mixture was fltered through a Celite column supported on glass wool (0.5 cm 3 cm) to afford a colourless fltrate. The solvent was then removed in vacuo to yield a colourless oil. Storage of this oil at 25 C for 24 h resulted in the formation of crystals within the matrix of the oil. The crystalline material was isolated by decanting off the remaining oil and then washed with cold (25 C) pentane (2 mL). This material consisted mostly of complex 5 and was discarded. The oil and the pentane washings were combined and the solvent was removed in vacuo to yield a colourless oil. Storage of this oil at 25 C for 24 h resulted in the deposition of colourless crystals, which were isolated by decanting off the remaining oil. 88.0 mg, 33%. Anal [Th(SCPh 3 )(NR 2 ) 3 ] (4) To a stirring suspension of KSCPh 3 (51.4 mg, 0.16 mmol) in toluene (5 mL) was added 2 (137.4 mg, 0.16 mmol). This solution was allowed to stir for 1 h, whereupon the solvent was removed in vacuo. The resulting white solid was extracted with hexanes (10 mL) and fltered through a Celite column supported on glass wool (0.5 cm 3 cm), to provide a colourless fltrate. The volume of the fltrate was reduced to 3 mL in vacuo. Storage of this solution for 48 h resulted in the deposition of colourless crystals, which were isolated by decanting off the supernatant (92. 3 To a colourless, cold (25 C), stirring solution of 3 (189.9 mg, 0.20 mmol) in THF (3 mL) was added KC 8 (56.1 mg, 0.42 mmol), which immediately yielded a dark red mixture. After 2 min, a cold (25 C), colourless solution of 18-crown-6 (104.3 mg, 0.39 mmol) in THF (3 mL) was added to this mixture. The solution was allowed to stir for 30 min, whereupon it was fltered through a Celite column supported on glass wool (0.5 cm 3 cm) to provide a vibrant red fltrate. The fltrate was dried in vacuo to provide a red solid that was triturated with diethyl ether (3 3 mL). The resulting red powder was extracted with diethyl ether (5 mL) and fltered through a Celite column supported on glass wool (0.5 cm 3 cm) to afford a large plug of bright red solid and a pale orange-red fltrate. The volume of the fltrate was reduced to 1 mL in vacuo. Storage of this solution at 25 C for 24 h resulted in the deposition of colourless crystals, which were isolated by decanting off the supernatant (47.0 mg, 23% (18-crown-6). IR (KBr pellet, cm 1 ): 599 (m), 665 (m), 677 (m), 724 (w), 755 (m), 770 (m), 832 (s), 867 (s), 966 (s), 986 (s), 1116 (s), 1182 (w), 1243 (s), 1285 (w), 1353 (m), 1455 (w), 1474 (w). Raman (neat solid, cm 1 ): 389 (w), 615 (s), 678 (m). [K(18-crown-6)][Th(S)(NR 2 ) 3 ] (7) To a colourless, cold (25 C), stirring solution of 4 (144.7 mg, 0.15 mmol) in THF (3 mL) was added KC 8 (41.2 mg, 0.30 mmol), which immediately yielded a dark red mixture. After 2 min, a cold (25 C), colourless solution of 18-crown-6 (76.5 mg, 0.29 mmol) in THF (3 mL) was added to this mixture. This solution was allowed to stir for 15 min, whereupon it was fltered through a Celite column supported on glass wool (0.5 cm 3 cm) to provide a vibrant red fltrate. The fltrate was dried in vacuo to provide a red solid that was triturated with diethyl ether (8 mL). The resulting red powder was extracted with diethyl ether (8 mL) and fltered through a Celite column supported on glass wool (0.5 cm 3 cm) to afford a large plug of bright red solid and a pale orange-red fltrate. The volume of the fltrate was reduced to 2 mL in vacuo. Storage of this solution at 25 C for 24 h resulted in the deposition of colourless crystals, which were isolated by decanting off the supernatant (48.7 mg, 32%). Subsequent concentration of the mother liquor and storage at 25 C for 24 h resulted in the deposition of additional crystals. (18-crown-6). IR (KBr pellet, cm 1 ): 605 (m), 664 (m), 685 (w), 699 (w), 785 (sh), 771 (m), 842 (s), 882 (sh), 936 (s), 963 (s), 1108 (s), 1182 (m), 1252 (s), 1285 (w), 1352 (m), 1455 (w), 1474 (w). Raman (neat solid, cm 1 ): 385 (w), 578 (s), 630 (s), 682 (s), 843 (m), 883 (m), 1014 (s). ## X-ray crystallography Data for 1, [Na(THF) 4.5 ][Th(Cl) 2 (NR 2 ) 3 ], 2, 4-7 were collected on a Bruker KAPPA APEX II diffractometer equipped with an APEX II CCD detector using a TRIUMPH monochromator with a Mo Ka X-ray source (a ¼ 0.71073 ). The crystals were mounted on a cryoloop under Paratone-N oil, and all data were collected at 100(2) K using an Oxford nitrogen gas cryostream. Data were collected using u scans with 0.5 frame widths. Frame exposures of 2 s were used for parameter determination were conducted using the SMART program. 72 Integration of the data frames and fnal cell parameter refnement were performed using SAINT software. 73 Absorption correction of the data was carried out using the multi-scan method SADABS. 74 Subsequent calculations were carried out using SHELXTL. 75 Structure determination was done using direct or Patterson methods and difference Fourier techniques. All hydrogen atom positions were idealized, and rode on the atom of attachment. Structure solution, refnement, graphics, and creation of publication materials were performed using SHELXTL. 75 Further crystallographic details can be found in Tables S1 and S2. † For [Na(THF) 4.5 ][Th(Cl) 2 (NR 2 ) 3 ], one sodium atom and its coordinated THF molecules exhibited positional disorder and were modelled over two positions in a 50 : 50 ratio. The C-C and C-O bond were constrained to 1.5 and 1.4 , respectively, using the DFIX command. In addition, the diethyl ether solvate of 6 exhibited positional disorder; one of the carbon atoms of this molecule was modelled over two positions in a 50 : 50 ratio. The anisotropic parameters of the disordered carbon atoms were constrained using the EADP command. Hydrogen atoms were not added to disordered carbon atoms. ## Computational details Density functional theory calculations were carried out using the PBE functional, 76,77 as implemented in the Gaussian 09 Rev. D.01 quantum chemistry code. 78 Dispersion corrections (D3) due to Grimme et al. 79 were included, as discussed in the main text. (14s 13p 10d 8f)/[10s 9p 5d 4f] segmented valence basis sets with Stuttgart-Bonn variety relativistic pseudopotentials were used for Th and U. 80 For the geometry optimisations, the 6-31G** basis sets were used for all other atoms. The ultrafne integration grid was employed in all calculations, as were the SCF convergence criteria. The default RMS force geometry convergence criterion was relaxed to 0.000667 au using IOP 1/7; the maximum force at each converged geometry is given in the ESI. † The electronic structures at the PBE + D3 geometries were recalculated using improved basis sets for the ligands; 6-311G** for O, S, N, K; 6-31G** for C and H. Natural bond orbital calculations were performed using the NBO6 code, interfaced with Gaussian. 81 QTAIM analyses were performed using the AIMAll program package, 82 with wfx fles generated in Gaussian used as input.

chemsum

{"title": "Thorium\u2013ligand multiple bonds via reductive deprotection of a trityl group", "journal": "Royal Society of Chemistry (RSC)"}

efficient_photoredox_conversion_of_alcohol_to_aldehyde_and_h₂_by_heterointerface_engineer

## Abstract: Controllable and precise design of bimetal-or multimetal-semiconductor nanostructures with efficient light absorption, charge separation and utilization is strongly desired for photoredox catalysis applications in solar energy conversion. Taking advantage of Au nanorods, Pt nanoparticles, and CdS as the plasmonic metal, nonplasmonic co-catalyst and semiconductor respectively, we report a steerable approach to engineer the heterointerface of bimetal-semiconductor hybrids. We show that the ingredient composition and spatial distribution between the bimetal and semiconductor significantly influence the redox catalytic activity. CdS deposited anisotropic Pt-tipped Au nanorods, which feature improved light absorption, structure-enhanced electric field distribution and spatially regulated multichannel charge transfer, show distinctly higher photoactivity than blank CdS and other metal-CdS hybrids for simultaneous H 2 and value-added aldehyde production from one redox cycle. ## Introduction Harvesting solar energy to drive artifcial photosynthesis for the production of value-added chemicals and renewable energy is a promising strategy to solve the growing worldwide energy crisis. However, the solar-to-chemical conversion efficiency of semiconductors is often limited by their fnite light absorption and/or slow charge separation rates and reaction kinetics. Integrating different functional materials into a single hybrid structure with precise design holds great promise for constructing efficient composite photocatalysts owing to the synergistic properties induced by the interactions between these components in an integrative ensemble. 5, Bimetallic nanostructures, coupling a surface plasmon resonance (SPR, Scheme S1 †) functionality with an efficient cocatalytic effect, could be an ideal candidate to simultaneously modulate the photoabsorption and steer the multichannel charge separation/transfer and reaction kinetics of semiconductors. The design of effective bimetal-semiconductor composite photocatalysts requires the rational understanding of the structural design principle, because the randomly hybridizing counterparts would often shield the SPR intensity and local electric feld of the plasmonic metal or weaken the net photoabsorption of the semiconductor. In addition, recent research has predominantly focused on charge separation and transfer mainly occurring at the plasmonic metal domains (i.e., SPR-induced sensitization effect), and the quantum efficiency over such systems is still relatively low in contrast to those by semiconductor photoexcitation. To obtain high photocatalytic activity, the structural design of special composite catalysts to synergistically utilize the SPR enhancement mechanism and band-gap photoexcitation of semiconductors is desirable. 28,29 However, thus far, it is still unclear how the composition and structural arrangement of plasmonic-nonplasmonic bimetal hybrids comprehensively affect their interactions with the semiconductor and the subsequent band alignment, charge transfer dynamics, local electric feld distribution and redox catalysis performance. Herein, we report a controllable way to design bimetalsemiconductor hybrids with different heterointerfaces for progressively improved redox catalysis conversion of alcohol to H 2 and aldehyde under visible-near-infrared (Vis-NIR) light irradiation. Au nanorods (NRs) and Pt nanoparticles (NPs) are promising metallic nanocrystal platforms for fabricating metalsemiconductor hybrids due to their tunable longitudinal-SPR absorption in the near-infrared (NIR) region and high efficiency for catalytic proton reduction reaction, respectively. 13, In this regard, we chose Au NRs and Pt NPs to form the plasmonic bimetal component. CdS, a well-known visible (Vis) light responsive photocatalyst with a direct band-gap of around 2.4 eV, 8, is chosen as the semiconductor component. Integrating the co-catalytic factor with semiconductor photoexcitation and SPR resonance modes in different optical response regions by the rational assembly of nonplasmonic Pt NPs and semiconductor CdS on the surface of Au NRs can provide the spatial transfer multichannel for electrons and boost the local electric feld, promoting the generation and migration of electron-hole charge carriers. As a result, the CdS deposited anisotropic Pt-tipped Au NRs (Au-Pt@CdS), which features the multiple metal-semiconductor and metal-metal heterojunctions, exhibits distinctly higher photoactivity than blank CdS and other metal-CdS hybrid counterparts for photocatalytic conversion of alcohol to aldehyde and H 2 by simultaneous utilization of photogenerated holes and electrons in one redox cycle. Resorting to electric feld simulations, transient absorption spectroscopy and multiple control experiments, it is demonstrated that the redox photoactivity is highly dependent on the ingredient composition and spatial distribution of the metal components because they play a signifcant role in affecting multiple electron transfer pathways and local electric feld enhancement. ## Results and discussion The fabrication process is based on progressively seed-mediated growth, as illustrated in Scheme 1. The stronger interaction between the cetyltrimethylammonium bromide (CTAB) surfacecapping molecules and the side facets of Au NRs results in the higher packed density of CTAB at the sides than that at the ends of Au NRs. 24,25 By taking advantage of this anisotropic surface structure, Pt NPs were frstly selectively deposited on the tips of Au NR seeds for the preparation of anisotropic Pt-tipped Au NRs (Au-Pt). The spatial distribution of Pt on the Au NR surface depends on the surface-capping molecules. Pt-covered Au NRs (Au@Pt) can be synthesized by decreasing the concentration of CTAB molecules. The Au-based NRs were subsequently used as seeds for the assembly of semiconductor CdS via a facile and controllable refluxing process. It should be noticed that our synthesis reaction is performed at a low temperature (85 C) and in the absence of metal ion sources (e.g., Ag + or Cu 2+ ) as a "bridge" between Au NRs and CdS to achieve the heterogrowth, 39,40 which can prevent the Au NRs from undesired reshaping in the wet-chemistry growth process. The samples obtained in each growth stage were characterized using scanning electron microscopy (SEM) and transmission electron microscopy (TEM) analysis. The average diameter and length of the original Au NRs are 29.0 and 92.5 nm, respectively (Fig. 1a and S1 †). After the growth of Pt on the tips of Au NRs, the smooth ends of Au NRs become gibbous, while the surface at the side of Au NRs remains almost unchanged (Fig. 1b and S2 †). The average diameter at the middle of Au-Pt NRs is 29.1 nm, and the average length is increased to 114.0 nm. These results suggest that Pt tends to anisotropically deposit at the two ends of Au NRs. Fig. 1c and S3 † show the morphology of Au@Pt with an average diameter of 33.2 nm and length of 107.9 nm, which are enclosed by the Pt shell. The morphology information of CdS deposited Au NRs (Au@CdS), CdS deposited Pt-tipped Au NRs (Au-Pt@CdS) and CdS deposited Pt-covered Au NRs (Au@Pt@CdS) is disclosed in Fig. 1d-f and S4. † No noticeable structural deformation of Aubased seeds can be discerned after depositing CdS nanoshells, which should beneft from the mild reaction process at a low temperature. For Au@CdS, we can clearly see from Fig. 1d and S4a-c † that the Au NR core with hemispherical ends are coated with a loose and porous CdS semiconductor layer in thicknesses of 15-20 nm. TEM images of Au-Pt@CdS (Fig. 1e) and Au@Pt@CdS (Fig. 1f) indicate their similar overall shape to Au@CdS. However, the bimetal cores in Au-Pt@CdS and Au@Pt@CdS respectively show convex (Fig. S4d-f †) and bar-like (Fig. S4g-i †) shapes. Elemental mapping analysis of Au-Pt@CdS (Fig. 1g) discloses that the Pt is located at the ends of Au NRs, further confrming the tip-coated morphology of the Au-Pt core in Au-Pt@CdS. Elemental mapping study of Au@Pt@CdS indicates that the layers of Pt and CdS shells are both isotropically deposited on the entire surface of Au NRs Scheme 1 Schematic flowchart illustrating the controllable preparation of CdS deposited Pt-tipped Au NRs (Au-Pt@CdS), CdS deposited Au NRs (Au@CdS) and CdS deposited Pt-covered Au NRs (Au@Pt@CdS). Fig. 1 SEM images of (a) Au NRs, (b) Au-Pt, and (c) Au@Pt, scale bar, 200 nm. The insets of (a-c) show TEM images of the corresponding samples, scale bar, 20 nm. TEM images of (d) Au@CdS, (e) Au-Pt@CdS and (f) Au@Pt@CdS, scale bar, 200 nm. Elemental mapping results of (g) Au-Pt@CdS and (h) Au@Pt@CdS. (Fig. 1h). The morphology of blank CdS was also analyzed (Fig. S5 †), suggesting that blank CdS is composed of tiny NPs. X-ray diffraction (XRD) patterns of blank CdS and metal-CdS hybrids are shown in Fig. 2a, in which the diffraction peaks of hexagonal phase CdS (JCPDS, no. 41-1049) and cubic phase Au (JCPDS, no. 65-8601) can be well identifed. The characteristic diffraction peaks of Pt are hardly observed due to their relatively low content in the composites. 41 Raman spectra, as displayed in Fig. 2b, show that blank CdS exhibits Raman signals at 299 cm 1 , 597 cm 1 and 890 cm 1 , which are identifed as characteristic peaks of the longitudinal optical (LO) phonons of the CdS phase. 42 In comparison with CdS, all of these peaks are signifcantly intensifed for metal-CdS hybrids, suggesting an enhanced light-matter interaction due to the strong electric feld enhancement near the surface of plasmonic Au-based NRs. 42 The evolution of optical absorption spectra along with the structural variation of the samples is displayed in Fig. 2c. The raw Au NRs have two SPR absorption bands centered at 510 and 800 nm, which are respectively assigned to transverse-SPR (T-SPR) and longitudinal-SPR (L-SPR) photoabsorption (inset of Fig. 2c). After Pt is deposited on Au NRs, the L-SPR band red-shifts to 833 and 865 nm for Au@Pt and Au-Pt, respectively. In addition, the T-SPR band of Au@Pt slightly red-shifts to 527 nm, while almost no change occurs for that of Au-Pt, which is probably due to the selective deposition of Pt onto the tips of Au NRs for the Au-Pt sample. 24 Both T-SPR and L-SPR bands are further red-shifted after the semiconductor CdS is grown on Au@Pt and Au-Pt, resulting from the increase of the surrounding dielectric constant. 31,43 The blank CdS and metal-CdS hybrids disclose a similar optical absorption threshold located at 520 nm, which corresponds to the optical band-gap of about 2.4 eV in the pristine CdS semiconductor. However, the absorption of metal-CdS hybrids is higher than that of blank CdS throughout the whole Vis to NIR region due to the contribution from the SPR absorption of Au-based NRs. 36,43 We tested the samples as dual-function photocatalysts for the conversion of benzyl alcohol (BA) to H 2 and benzaldehyde (BAD) under Vis-NIR light (l > 420 nm, Fig. S6 †) irradiation. We initially studied the effects of the proportion of different components on the photoactivity and determined the sample with optimal photoactivity, as shown in Fig. S7a-c. † The direct photoactivity comparison with the H 2 and BAD production rates over the samples with the optimal proportion of components is shown in Fig. 2d. Notably, the molar ratio of the reduction product (H 2 ) and oxidation product (BAD) is calculated to be ca. 1.0, suggesting a stoichiometric dehydrogenation reaction. The fne control of the heterointerface of metal-semiconductor composites can result in progressively optimal photoactivity. Blank Au-Pt and CdS show very low activity for H 2 and BAD production due to the fast recombination of electron-hole pairs. 11,12,20,31 For binary Au@CdS, the H 2 evolution rate is enhanced to 53.2 mmol h 1 . After coupling the co-catalyst Pt with Au@CdS, the H 2 evolution rates over ternary Au@Pt@CdS and Au-Pt@CdS reach 96.6 mmol h 1 and 153.0 mmol h 1 , respectively, which are about 13.4 and 21.2 times as high as that over blank CdS (7.2 mmol h 1 ). Notably, the anisotropic Au-Pt@CdS exhibits higher H 2 and BAD production rates than Au@CdS and Au@Pt@CdS, implying that the photoactivity enhancement of bimetal-semiconductor composites is not only dependent on the ingredient composition, but also affected by the spatial distribution of metal components. For comparison, we also investigated the photoactivity of CdS supported Au-Pt NRs (Au-Pt/CdS, Fig. S7d †) and the result shows that the photoactivity of such supported Au-Pt/CdS is much lower than that of Au-Pt@CdS, which manifests that the core-shell confguration can strengthen the interfacial contact between the metal and semiconductor CdS, thereby facilitating the charge carrier separation and transfer across the interfacial domain and consequently boosting the photoactivity. The photoactivity enhancement and general applicability of Au-Pt@CdS were further investigated by conversion of other aromatic alcohols (R-PhCH 2 OH, R ¼ OCH 3 , CH 3 , Cl and OH) with different substituent groups. By taking unsubstituted BA (Fig. 2d) as the reference, the results (Fig. S8a-d †) indicate that the H 2 and aldehyde production rates are enhanced by electrondonating substituents (R ¼ OCH 3 and CH 3 ) and retarded by electron-withdrawing groups (R ¼ Cl and OH) for all photocatalysts. Au-Pt@CdS displays higher photoactivity than blank CdS and other metal-CdS hybrids for conversion of these alcohols. In addition, the photocatalytic performance of Au-Pt@CdS is either comparable or superior to other similar coupled reaction systems for simultaneous production of H 2 and value-added chemicals (Table S1 †). Fig. 3a shows the long-term photocatalytic performance of Au-Pt@CdS toward the conversion of BA under Vis-NIR light irradiation. After 5 h of irradiation, the conversion of BA is about 80.1% and the selectivity for BAD reaches 94.0%. The H 2 and BAD production rates display no obvious decrease during four successive runs within 20 h. In addition, the morphology, composition and crystalline structure of Au-Pt@CdS after recycling photocatalytic reactions are similar to those of the fresh sample (Fig. S9 †). These results indicate the good stability of Au-Pt@CdS under our photocatalytic reaction conditions. To understand the roles respectively played by the semiconductor and metal components of Au-Pt@CdS in the photoactivity enhancement, we performed the wavelength-dependent experiment. As shown in Fig. 3b, the action spectrum of apparent quantum yield efficiency (AQY) is in agreement with the bandgap absorption of the CdS component. This result reveals that the photocatalytic conversion of BA to produce BAD and H 2 proceeds through light absorption by the semiconductor CdS component instead of SPR excitation of Au-Pt bimetal. 24,44 Therefore, it is crucial to explore what is the specifc role of plasmonic Au and nonplasmonic Pt in improving the photoactivity of semiconductor CdS, and what kind of contribution of the SPR excitation is associated with bimetallic Au-Pt in influencing the separation of charge carriers photogenerated from CdS in Au-Pt@CdS. Since electric feld distribution and intensity are related to the SPR enhancement, 25,45,46 we simulated the electric feld (|E|/ |E 0 |) distribution around different metal-CdS hybrids, aiming to ascertain the region of electric feld enhancement upon SPR excitation and its possible contribution to the photoactivity enhancement. As shown in Fig. 3c, differing from strong electric feld distribution at both ends of binary Au@CdS, the ternary Au@Pt@CdS sample exhibits weak electric feld enhancement. When the continuous Pt shell encapping the entire surface of Au NRs in Au@Pt@CdS is replaced by discrete Pt selectively tipped onto the two ends of Au NRs in Au-Pt@CdS, the electric feld enhancement around the surface of Au-Pt@CdS is greatly boosted. The weakest electric feld enhancement of Au@Pt@CdS could be ascribed to the plasmon-induced electron transfer from Au to Pt and the serious plasmon damping effect of the continuous Pt shell. 36 According to the simulation result, the maximum values of electric feld enhancement at the outside surface of CdS for Au@CdS, Au-Pt@CdS and Au@Pt@CdS are about 15, 13 and 5, respectively. The enhanced electric feld can improve the light harvesting ability of the catalyst, and increase the generation and separation rate of electron-hole pairs near the surface of the semiconductor material. However, the trend in electric feld enhancement (i.e., Au@CdS > Au-Pt@CdS [ Au@Pt@CdS) is inconsistent with that of photoactivity (i.e., Au-Pt@CdS > Au@Pt@CdS > Au@CdS). This indicates that in addition to the effect of electric feld enhancement factor on the photoactivity of metal-semiconductor hybrids, other factors should be simultaneously considered to account for the difference in photoactivity enhancement. We then investigated charge transfer dynamics in CdS and metal-CdS hybrids by using ultrafast transient absorption (TA) spectroscopy. Upon 370 nm excitation, a negative absorption peak at about 500 nm corresponding to the ground state bleach (GSB) signal of the CdS component is formed due to the band-flling by photoinduced electrons and holes (Fig. 3d and S10 †). The recovery kinetics of this GSB reflects the charge dissipating processes by recombination and/or transfer (Fig. 3e). In blank CdS, the recovery kinetics originates from the recombination of charge carriers and the possible charge trapping by the defects. 15,48 In the presence of metal nanostructures, the kinetics become signifcantly faster. We attribute this observation to the photoinduced electron transfer from CdS to the metal ingredient, as reported in similar semiconductor-metal composites. 11,29,49 In order to quantitatively determine the electron transfer rate, the recovery kinetics were ftted by a biexponential function and the electron transfer rates were estimated by comparing the charge transfer rate constants (k ct , see eqn (S1)-(S3) and Table S2 †). 50 In comparison with blank CdS, the charge transfer rate over binary Au@CdS is increased, which indicates an electron transfer channel from CdS to Au because the work function of Au (5.10 eV) is larger than that of CdS (4.20 eV). 51,52 After the Pt NPs (with a work function of 5.40 eV) are deposited on Au NRs, the k ct values of both Au@Pt@CdS and Au-Pt@CdS are further enhanced. The highest k ct value for Au-Pt@CdS signifes that the Au-Pt bimetal nanostructure is more effective than Au@Pt in facilitating the photoexcited charge transfer in the photocatalytic system, due to the additional interfacial electron transfer channels from CdS to Au and to Pt guaranteed by the direct contact and matched band alignment among the three components (Fig. S11 †). 51,52 Fig. 3 To study the effect of SPR excitation on photoactivity enhancement, we performed a wavelength-control experiment over Au-Pt@CdS under selective photoexcitation of the plasmonic Au-Pt component (Fig. S12a †). 31,53 It is seen from Fig. S12b † that the Au-Pt@CdS hybrids show only trace photoactivity under SPR excitation alone, indicating that the effect of hot electron injection from Au-Pt to CdS on the photoactivity is negligible. When we used the photon energy at 570 nm to excite the T-SPR of Au-Pt@CdS in the TA measurement, the sample shows no signal corresponding to excitation and relaxation kinetics (Fig. S12c †), which further confrms the absence of the hot electron transfer process. Thus, the SPR enhancement mechanisms should be interpreted in terms of the electric feld effect. 45 In addition, wavelength-control experiments under selective photoexcitation of CdS or simultaneous photoexcitation of CdS and Au-Pt (Fig. S13a †) disclose that the Au-Pt@CdS exhibits enhanced photoactivity as compared to bare CdS under both irradiation conditions, and the photoactivity enhancement is more prominent under simultaneous photoexcitation of CdS and Au-Pt components (Fig. S13b †). We also prepared nonplasmonic Pt-loaded CdS (Pt-CdS) through the well-reported photodeposition method. 44,54 The photocatalytic activity of nonplasmonic Pt-CdS is much lower than that of plasmonic Au-Pt@CdS under Vis-NIR light irradiation (Fig. S13c †). The above joint results signify that the promoted photoactivity of Au-Pt@CdS is induced by the synergistic coupling of the band-gap photoexcitation of semiconductor CdS with the electric feld enhancement and the electron sink effect of Au-Pt bimetal, as depicted in Fig. 4a. 55,56 Specifcally, when the Au-Pt@CdS ternary heterostructure is illuminated by Vis-NIR light (l > 420 nm), the CdS component can harvest short-wavelength visible light (420 nm < l < 520 nm), yielding electron-hole charge carriers (e -h + ). Meanwhile, the SPR excitation of plasmonic Au-Pt bimetal by longer-wavelength light irradiation provides extra electric feld enhancement to improve the photoabsorption, and promote the generation and separation of electron-hole pairs in CdS. 45,46,57 Subsequently, the photogenerated electrons traverse from the conduction band (CB) of CdS to the anisotropic Au-Pt bimetal component. The protons (H + ) diffuse through the loose and porous CdS thin layer and react with electrons to generate H 2 . 35,58,59 The holes retained in the valence band (VB) of CdS can selectively oxidize alcohols to aldehydes, which is the overall activity-limiting step in the redox cycle. 44,60 The photogenerated electrons and holes are spatially separated, signifcantly reducing the recombination probability, which leads to greatly improved photoactivity of Au-Pt@CdS. The investigation of the photoelectrochemical (PEC) process can further reveal the picture of electron transfer and the underlying photoactivity enhancement mechanism. As shown in Fig. 4b, transient photocurrent responses of different samples under chopped Vis-NIR light illumination indicate that the construction of bimetal-semiconductor hybrids is conducive to enhancing the photocurrent of semiconductor CdS, which follows the same order as the photoactivity: Au-Pt@CdS > Au@Pt@CdS > Au@CdS > CdS. This result indicates a more effective separation of electron-hole pairs over Au-Pt@CdS than other electrode samples. 9,35,61,62 Au-Pt@CdS and Au@Pt@CdS consist of the same hetero-components, but the anisotropic Au-Pt@CdS exhibits higher photocurrent and photoactivity than isotropic Au@Pt@CdS under Vis-NIR light illumination. This suggests that the inappropriate arrangement of the plasmonic-nonplasmonic bimetal heterostructure in Au@Pt@CdS leads to the suppression of electric feld enhancement and electron sink effect, as respectively proved by the results of electric feld simulation (Fig. 3c) and TA characterization (Fig. 3e). 24,25 Electrochemical impedance spectroscopy (EIS, Fig. 4c) and Mott-Schottky (Fig. 4d) analysis were conducted under Vis-NIR light illumination. The arc at the middle frequencies of the Nyquist plot in Fig. 4c is characteristic of charge transportation resistance. 63 Apparently, the diameter of the arc for the Au-Pt@CdS electrode is much smaller than that of CdS, Au@CdS and Au@Pt@CdS, indicating that the resistance of interfacial charge transportation is signifcantly decreased over the Au-Pt@CdS electrode. The Mott-Schottky analysis can be used to provide fundamental insights into the charge carrier density (N D ), which is obtained from the following equation: 63,64 Fig. 4 where C is the capacitance, e is the elementary electronic charge, 3 0 is the permittivity in vacuum, and 3 is the dielectric constant, specifcally 8.99 for CdS. From the slope in the plot of 1/C 2 vs. V in Fig. 4d, N D can be determined by the above equation. A summary of N D values is given in Table S3. † The plot of the Au-Pt@CdS electrode shows the smallest slope value among the samples, indicating a much higher N D than that of CdS, Au@CdS and Au@Pt@CdS. The higher N D signifes lower resistance, faster charge transfer rate and consequently enhanced PEC and photocatalytic performance. In addition, all plots show positive slopes, indicating the n-type semiconductor characteristic of CdS. 63,66 The formed multiple Schottky barrier between the n-type CdS semiconductor and Au-Pt bimetal would prevent the electrons the metal drifting back to the semiconductor. 67 Based on the above results, it can be concluded that Au-Pt can effectively facilitate the generation and transfer of charge carriers from CdS, due to the boosted local electric feld and the enriched spatial charge transfer channels. Fig. 4e displays the linear sweep voltammetry (LSV) curves of Au-Pt@CdS, Au@Pt@CdS, Au@CdS and CdS electrodes without light irradiation, from which it is seen that the addition of nonplasmonic Pt into Au@CdS can obviously enhance the cathodic current density and decrease the overpotential of H 2 production, thus further promoting catalytic efficiency. 31,48 ## Conclusions In summary, we have accomplished a controllable design of bimetal-semiconductor hybrids for optimizing the redox photocatalysis performance of alcohol conversion to H 2 and valueadded aldehyde for the frst time. semiconductor hybrids, as well as the underlying SPR-coupled multichannel electron transfer mechanism illustrated here, could be instructive for further rational design of plasmonic bimetal-or multimetal-semiconductor hybrids toward efficient redox catalysis. ## Synthesis of Au NRs The Au NRs were prepared using a seed-mediated growth method. 68 The seed solution for Au NR growth was prepared as follows: 5 mL of 0.5 mM HAuCl 4 was mixed with 5 mL of 0.2 M CTAB solution in a 20 mL scintillation vial. 0.6 mL of fresh 0.01 M NaBH 4 was diluted to 1 mL with water and was then injected into the above solution under vigorous stirring (1200 rpm). The solution color changed from yellow to brownish yellow and the stirring was stopped after 2 min. The seed solution was aged at room temperature for 2 h before use. Subsequently, 1.4 g of CTAB and 0.25 g of NaOL were dissolved in 50 mL of warm water ($50 C) in a 100 mL Erlenmeyer flask. The solution was allowed to cool down to 30 C and 3.6 mL of 4 mM AgNO 3 solution was added. The mixture was kept undisturbed at 30 C for 15 min after which 50 mL of 1 mM HAuCl 4 solution was added. The solution became colorless after 90 min of stirring (700 rpm) and 0.3 mL of HCl (37 wt% in water, 12.1 M) was then introduced. After another 15 min of slow stirring at 400 rpm, 0.25 mL of 0.064 M AA was added and the solution was vigorously stirred for 30 s. Finally, 0.4 mL of seed solution was injected into the growth solution. The resultant mixture was stirred for 30 s and left undisturbed at 30 C for 12 h for NR growth. ## Preparation of Au-Pt Au-Pt was prepared by using Au NRs as the template. 24 Briefly, 10 mL of the prepared Au NRs was separated from excess CTAB by centrifugation twice (at 10 000 rpm) and redispersed in 0. added into 10 mL of as-made Au NR suspension. 100 mL of 0.01 M H 2 PtCl 6 $6H 2 O and subsequently 0.08 mL of 0.01 M HCl were added to the reaction mixture. The mixture was left undisturbed for 24 h at 30 C. Pt-covered Au NRs (Au@Pt) were prepared by using Au NRs in 0.03 M CTAB solution as the template following the same procedure. ## Preparation of Au-Pt@CdS For the synthesis of Au-Pt@CdS, 10 mL of the as-prepared Au-Pt seeds and 10 mL of aqueous Gly solution (0.2 M) were mixed in a 50 mL vial, and subsequently 300 mL aqueous NaOH solution (2 M) was added to the reaction mixture. The mixture was kept at 30 C for 30 min without stirring. Next, 200 mL of Cd(Ac) 2 (0.1 M) and TAA (0.1 M) solution were injected into the above solution drop by drop. The reaction was allowed to proceed at 85 C under vigorous stirring for 2 h. The suspensions were centrifuged at 12 000 rpm for 10 min and washed with ethanol three times, and then the precipitates were vacuum dried at 40 C. Au@CdS, Au@Pt@CdS and blank CdS were also synthesized by the above method, except that the corresponding seeds or blank CTAB (0.1 M) aqueous solution was used instead of Au-Pt, respectively, in the synthesis process. The Au-Pt/CdS sample was prepared by mixing the blank CdS with an appropriate amount of Au-Pt in solution with vigorous stirring for 24 h followed by the wash and dry process similar to Au-Pt@CdS. ## Material characterization The crystal phase properties of the samples were analysed with a Bruker D8 Advance X-ray diffractometer (XRD) using Ni-fltered Cu Ka radiation at 40 kV and 40 mA in the 2q range from 20 to 80 with a scan rate of 0.02 per second. The optical properties of the samples were characterized using a Cary-5000 ultraviolet-visible-near infrared diffuse reflectance spectrophotometer (DRS, Varian Co.). The morphology and elemental distribution of the samples were analysed by feld-emission scanning electron microscopy (FESEM) on a FEI Nova NANO-SEM 230 spectrophotometer and transmission electron microscopy (TEM), high-resolution TEM (HRTEM) and elemental mapping analysis using a JEOL 2100F instrument at an accelerating voltage of 200 kV. Raman spectroscopy was performed on a Renishaw inVia Raman System 1000 with a 532 nm Nd:YAG excitation source at room temperature. X-ray photoelectron spectroscopy (XPS) measurements were performed using a Thermo Scientifc ESCA Lab250 spectrometer that consists of monochromatic Al Ka as the X-ray source, a hemispherical analyser, and a sample stage with multiaxial adjustability to obtain the composition on the surface of the samples. All of the binding energies were calibrated using the C 1s peak of the surface adventitious carbon at 284.6 eV. The elemental concentration analysis was performed using an inductively coupled plasma emission spectroscopy instrument (ICP, PerkinElmer Optima 8000). The carrier dynamics were measured by using femtosecond transient absorption spectroscopy (Time-Tech Spectra, femtoTA-100). Part of the 800 nm output pulse from the amplifer was used to pump a TOPAS Optical Parametric Amplifer (OPA) which generates the 370 nm pump beam. The pump pulses were chopped by a synchronized chopper at 500 Hz and the absorbance change was calculated with two adjacent probe pulses (pump-blocked and pumpunblocked), and the pump power was approximately 1 mW. The samples were dispersed in water for all pump-probe characterizations performed under ambient conditions. The photoelectrochemical analysis was carried out in a conventional three-electrode cell using a Pt plate and an Ag/AgCl electrode as the counter electrode and reference electrode, respectively. The electrolyte was 0.2 M Na 2 SO 4 aqueous solution containing 0.1 mM BA. The working electrode was prepared on indium-tin oxide (ITO) glass that was cleaned by sonication in ethanol for 30 min and dried at 353 K. The boundary of ITO glass was protected using Scotch tape. 5 mg of the sample was dispersed in 0.5 mL of N,N-dimethylformamide (DMF, supplied by Sinopharm Chemical Reagent Co., Ltd) by sonication to get a slurry. The slurry (20 mL) was spread onto pre-treated ITO glass. After air drying, the working electrode was further dried at 393 K for 2 h to improve adhesion. Then, the Scotch tape was unstuck, and the uncoated part of the electrode was isolated with epoxy resin. The exposed area of the working electrode was 0.25 cm 2 . The light irradiation source was a 300 W Xe arc lamp system equipped with a flter to cut off light of wavelength l < 420 nm. The cathodic polarization curves were obtained using the linear sweep voltammetry (LSV) technique with a scan rate of 2 mV s 1 . The electrochemical impedance spectroscopy (EIS) experiments were conducted on an AUTOLAB M204 workstation. ## FDTD simulations The electric feld simulations were carried out by using a fnitedifference time-domain (FDTD) software package (Lumerical Solutions, Inc.). The refractive indices of CdS, BTF, Au and Pt were taken from previous reported data. 19,23,69 A total-feld/ scattered feld (TFSF) source was used as an incident feld into the simulation region, which coincided with the wavelength of our experimental irradiation source. Perfectly matched layer (PML) boundary conditions were used in our simulations. The Au NRs are cylinders with round ends and the size was taken to match their average values. The radius of the cylinder is 14.5 nm and radii of the ends are 8 nm. The total length of the Au NRs is 92.5 nm. The shell thickness of CdS is 20 nm. For the model of Au-Pt@CdS, elliptical Pt spheres with diameters of 14 nm were located at both ends of the Au NRs with a total length of 98 nm. For Au@Pt@CdS, the Pt shell is a cylinder with a radius of 2.1 nm and the total length of the Au NRs is 107.9 nm. ## Photoactivity The experiments were performed referring to the previously reported literature with some modifcations. 44,70 In a typical photocatalytic reaction, 2 mL catalyst suspension (1 mg mL 1 in BTF) containing 0.5 M alcohol was placed in a quartz reactor (25 mL) equipped with a magnetic stir bar. The reaction suspension was sonicated for 2 min at room temperature and purged with Ar gas for 30 min. The reactor was then tightly sealed and then stirred for 30 minutes in the dark to achieve an adsorptiondesorption equilibrium. A 300 W Xe arc lamp (PLS-SXE 300, Beijing Perfect light Co., Ltd) with a flter to cut off light of wavelength l < 420 nm was used as the irradiation source. The light intensity was fxed at 800 mW cm 2 . The photon flux of incident light was measured using an Optical Power/Energy Meter (Newport 842-PE). The irradiation area is 3.8 cm 2 . During photocatalysis, the suspension was continuously stirred to ensure uniform irradiation. The evolved gases were analysed using a gas chromatograph (Shimadzu GC-2014C, MS-5 A column, Ar carrier) equipped with a thermal conductivity detector (TCD). Products in solution were quantifed using an Agilent Gas Chromatograph (GC-7820) with a flame-ionization detector (FID) and identifed by gas chromatography-mass spectrometry (GC-MS, Agilent Technologies, GC6890N, MS 5973). The recycling test for photocatalytic conversion of BA over Au-Pt@CdS was performed as follows. After 5 h reaction under Vis-NIR light irradiation, the suspension was centrifuged and mixed with 2 mL BTF containing 0.5 M BA for continuous test. The conversion of BA and selectivity for BAD were calculated with the following equations: 71 here C 0 is the initial concentration of BA; C BA and C BAD are the concentrations of the residual BA and the corresponding BAD at a certain time after the catalytic reaction, respectively. The apparent quantum efficiency (AQE) was calculated from the ratio of twice the number of H 2 molecules to the number of incident photons by using the following expression: 22,31 AQE ¼ 2 number of H 2 molecules number of incident photons 100% where M is the molar amount of H 2 molecules, N A is the Avogadro constant, h is the Planck constant, c is the speed of light, S is the irradiation area, P is the intensity of the irradiation, t is the photoreaction time and l is the wavelength of the monochromatic light. The power density of different monochromatic lights was fxed at 30.0 mW cm 2 . ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Efficient photoredox conversion of alcohol to aldehyde and H₂ by heterointerface engineering of bimetal\u2013semiconductor hybrids", "journal": "Royal Society of Chemistry (RSC)"}

examining_sterically_demanding_lysine_analogs_for_histone_lysine_methyltransferase_catalysis

## Abstract: Methylation of lysine residues in histone proteins is catalyzed by S-adenosylmethionine (SAM)dependent histone lysine methyltransferases (KMts), a genuinely important class of epigenetic enzymes of biomedical interest. Here we report synthetic, mass spectrometric, nMR spectroscopic and quantum mechanical/molecular mechanical (QM/MM) molecular dynamics studies on KMtcatalyzed methylation of histone peptides that contain lysine and its sterically demanding analogs. our synergistic experimental and computational work demonstrates that human KMts have a capacity to catalyze methylation of slightly bulkier lysine analogs, but lack the activity for analogs that possess larger aromatic side chains. overall, this study provides an important chemical insight into molecular requirements that contribute to efficient KMT catalysis and expands the substrate scope of KMTcatalyzed methylation reactions.Posttranslational modifications on histone proteins regulate the structure and function of human chromatin 1-3 . Well-established examples include lysine acetylation, which is linked with the transcriptionally active region of human genome, and lysine methylation, which is associated with gene activation and suppression, depending on the histone sequence and methylation state 4,5 . Histone lysine methylation is catalyzed by S-adenosylmethionine (SAM)-dependent histone lysine methyltransferases (KMTs), and can lead to a formation of monomethyllysine (Kme), dimethyllysine (Kme2) and trimethyllysine (Kme3) 6,7 . It is generally believed that the methylation state depends on the constitution of the KMT active site (Fig. 1a) 8 . With the exception of DOT1L, all members of KMT family possess the SET (Su(var)3-9, Enhancer-of-zeste and Trithorax) domain [9][10][11] . Structural analyses of KMTs complexed with histone peptide/methylated peptide and S-adenosylhomocysteine product (SAH) revealed that the lysine side chain occupies a narrow, hydrophobic channel, typically comprised of side chains of several tyrosine and phenylalanine residues (Fig. 1b) 8 . The positioning of the lysine's N ε amino group towards the electrophilic methyl group of the SAM cosubstrate results in an efficient methyl transfer via S N 2 reaction 12,13 .Recent examinations of lysine analogs as substrates for human histone lysine methyltransferases revealed that KMTs possess a high degree of specificity for lysine residues. Enzymatic assays employing MALDI-TOF MS verified that human KMTs preferentially catalyze methylation of lysine residues with L-stereochemistry over D-stereochemistry 14 . Combined experimental and computational studies on histone peptides that bear lysine analogs of different chain length revealed that lysine exhibits an optimal chain length for KMT-catalyzed methylation 15 , and that the enzymatic methylation is limited to N-nucleophiles 16 . Members of KMTs were also found to catalyze methylation of the cysteine-derived γ-thialysine on intact histones and histone peptides 17,18 . Substrate capturing studies using the genetically encoding photo-lysine showed that slightly bulkier γ-diaza-lysine undergoes efficient SETD7-catalyzed methylation in cells 19 . In addition to the essential role of the lysine's side chain, its main chain also plays an important role in productive KMT catalysis 20 . Despite these recent findings that shed light on basic understanding of KMT catalysis, a broader scope of lysine analogs as substrates for KMTs has not been explored yet. Here we report enzymatic evaluations of sterically demanding lysine analogs as substrates for human KMTs employing MALDI-TOF MS assays, NMR spectroscopic analyses, and quantum mechanical/ molecular mechanical (QM/MM) molecular dynamics and free energy studies. The lysine's side chain is comprised of four hydrophobic methylene groups and the terminal nucleophilic N ε amino group. The zig-zag orientation of the flexible C-C bonds might enable a proper orientation of the lysine's side chain in a narrow hydrophobic pocket of KMTs, leading to efficient KMT catalysis. It remains to be established whether this narrow lysine-binding pocket can accommodate larger moieties that resemble lysine. The objective of this work is to explore whether KMTs do have a capacity to catalyze methylation of bulkier lysine analogs present on histone peptides. We selected six sterically demanding lysine analogs: (i) cyclopropyllysine (K CP ), which bears an additional methylene group adjacent to the N ε amino group; (ii) benzylamine (K ba ), an analog with a larger but highly nucleophilic side chain; (iii) meta-aminophenylalanine (F 3a ), a significantly larger aromatic lysine analog that possesses the terminal N ε amino group with a weaker nucleophilic character; (iv) para-aminophenylalanine (F 4a ), another aniline derivative with less nucleophilic N ε amino group; (v) pyridylalanine (A P ), which possesses a nucleophilic pyridine functionality; and (vi) tyrosine (Y), an electron-rich aromatic system with a potential to undergo O-or C-methylation (Fig. 1c). ## Results and Discussion Fmoc-and Boc-protected cyclopropyllysine (Fmoc-K CP (Boc)-OH, 1) was synthesized in nine steps using a modification of the reported procedure (Fig. 2) 21 . To install an alcohol on the side chain, perbenzylation of L-glutamic acid 2 produced a tetra-substituted compound that underwent selective reduction of the side chain ester in the presence of DIBAL-H to afford the intermediate 3. Swern oxidation was applied to give the amino aldehyde, which reacted directly with t-butyl diethylphosphonoacetate via Horner-Wadsworth-Emmons reaction to produce the α,β-unsaturated t-butyl ester 4. 1 H NMR data confirmed that 4 exists as the E-isomer. Diazomethane was then generated in situ and distilled directly into a solution of 4 containing catalytic amounts of palladium(II) acetate, to yield the α,β-cyclopropyl t-butyl ester, which was selectively hydrolyzed with TFA to yield compound 5. The 13 C NMR spectrum of 5 revealed an ~1:1 "doubling" of many of the signals into small doublets. This finding was indicative of either diastereomeric cyclopropylation whereby the methylene is added above or below the alkene plane in roughly equal percentages or that the compound had some form of hindered rotation that resulted in two identical molecules with nearly identical conformations. The latter is unlikely as there are no stereotypical bond-types that form rotamers, and nonspecific cyclopropylation is the more reasonable explanation as two different diastereomers are formed due to the chiral C α of the backbone. The Boc-protected cyclopropyl amine 6 was then produced through a Curtius rearrangement after refluxing in t-butanol. Subsequently, deprotection of the benzyl group on amine/carboxylate with Pd/C under hydrogen atmosphere, followed by the Fmoc-protection of the free amine afforded Fmoc-K CP (Boc)-OH 1. The presence of 13 C "doubling" was also present in the final building block 1. High-resolution 2D 1 H-13 C HSQC-TOCSY spectra were able to produce 1 H spectra of each of the two diastereomers from the projection of the cross-peaks. These 1 Hs exhibited very small but noticeable differences in chemical shift (Supplementary Fig. 1). Furthermore, exploring the effect of temperature on the lineshape of the 1 H signals from 25 °C to 50 °C revealed no significant effects, further supporting the explanation of diastereomers versus rotamers for the observed spectral doubling (Supplementary Fig. 2). The other five Fmoc-protected lysine analogs, i.e. Fmoc-K ba (Boc)-OH, Fmoc-F 3a (Boc)-OH), Fmoc-F 4a (Boc)-OH, Fmoc-A P -OH, and Fmoc-Tyr( t Bu)-OH, are commercially available. All sterically demanding lysine analogs were incorporated into histone peptides using solid-phase peptide synthesis; H4K20 analogs (GGAKRHRKVLRDNIQ), H3K4 analogs (sequence ARTKQTARKSTGGKA), and H3K9 analogs (sequence ARTKQTARKSTGGKA) were synthesized. All histone peptides were purified by preparative HPLC, and the purity of synthetic histone peptides bearing lysine analogs was confirmed by analytical HPLC and ESI-MS analyses (Supplementary Tables 1-2 and Supplementary Figs. 3-13). We examined histone peptides bearing lysine and its sterically demanding analogs as potential substrates for human KMTs employing MALDI-TOF MS assays. Enzymatic assays with SETD8 (2 µM) and SAM cosubstrate (200 µM) showed different degrees of methylation of H4K20 peptides (100 µM) after 1 hour at 37 °C. While natural sequence H4K20 underwent quantitative monomethylation, cyclopropyl-containing H4K CP 20 peptide appeared to be monomethylated to a comparatively lesser extent (60% of H4K CP 20me) (Fig. 3a). Under the same conditions, none of the other five aromatic lysine analogs were observed to be methylated within limits of detection in the presence of SETD8 (Fig. 3). Increased amounts of SETD8/SAM and prolonged incubation at 37 °C resulted in almost complete formation of the monomethylated H4K CP 20me product (Supplementary Fig. 14), but still did not lead to appearance of detectable amounts of the monomethylated products of the remaining five lysine analogs (Supplementary Figs. 15-19). As expected, control experiments in the absence of SETD8 or SAM verified that monomethylation of H4K CP 20 is SETD8-catalyzed and also requires the presence of SAM cosubstrate (Supplementary Figs. 20-21). MALDI-TOF analyses of SETD7-catalyzed methylation of H3K4 peptides showed that none of histone peptides that contain sterically demanding lysine analogs was methylated within limits of detection (only traces of H3K CP 4 were observed); SETD7 in the presence of SAM indeed catalyzed the formation of monomethylated H3K4 with a natural sequence (Supplementary Fig. 22). At high concentration of SETD7 (10 µM) and SAM (1 mM) and longer incubation (3 hours), an increased amount of the monomethylated H3K CP 4me product was observed (Supplementary Fig. 23). Despite being monomethyltransferases, SETD7 and SETD8 appear to have somewhat different abilities to accept substrates other than lysine. In line with our work on γ-thialysine 18 , SETD8 seems to have a slightly broader substrate scope than SETD7, possibly due to subtle differences of the active sites (e.g. positioning of Y273 in SETD8 and Y305 in SETD7). Our recent investigations demonstrated that, in contrast to monomethyltransferases SETD8 and SETD7, H3K9 trimethyltransferases G9a and GLP appear to exhibit a somewhat broader substrate scope for the enzymatic methylation reaction. Enzymatic studies of natural and unnatural H3K9 peptide sequences (100 µM) in the presence of G9a/GLP (2 µM) and SAM (500 µM) at 37 °C showed that both enzymes do have a potential to catalyze methylation of H3K CP 9, minor methylation of H3K ba 9 (traces detected), whereas we did not observe any www.nature.com/scientificreports www.nature.com/scientificreports/ methylated products with other four bulkier lysine analogs within limits of detection (Fig. 3b and Supplementary Fig. 24). H3K CP 9 underwent predominant GLP-catalyzed dimethylation (75%), while monomethylated (10%) and trimethylated (10%) products were also observed after 1 hour under standard conditions; longer incubation times led to slightly increased amounts of H3K CP 9me3 (Supplementary Fig. 25). Under the same conditions, H3K CP 9me2 (60%) and H3K CP 9me3 (40%) were formed in the presence of G9a after 1 hour, whereas equal amounts of both methylated products were found after 3 hours at 37 °C (Supplementary Fig. 26). Notably, increased amounts of GLP (10 µM) and SAM (1 mM) afforded almost exclusive formation of H3K CP 9me3 and significant (55%) monomethylation of H3K ba 9 after 5 hours at 37 °C, whereas other sterically demanding lysine analogs were still not methylated within detection limits (Supplementary Fig. 27). Control experiments in the absence of G9a/GLP or SAM additionally confirmed that both the enzyme and the cosubstrate are required for methylation on H3K CP 9 to occur (Supplementary Figs. 28-29). To establish the substrate efficiency of lysine-and K CP -containing histone peptides, we carried out enzyme kinetics analysis, employing the MALDI-TOF MS assays 22 . Both enzymes preferentially catalyze methylation of natural histone sequences, however, bulkier K CP -containing peptides still underwent favorable kinetics profiles (Table 1 and Supplementary Fig. 30). The lower substrate efficiencies for H4K CP 20 and H3K CP 9 compared to natural sequences were a result of higher K M values, implying a less favorable association of bulkier K CP in a narrow binding pocket of KMTs. Next, we carried out competition studies between histone peptides that bear lysine and its analogs. In the presence of SETD8, SAM and equimolar amounts of H4K20 and H4K CP 20, we observed the formation of both monomethylated products, albeit a comparatively larger degree of monomethylation of H4K20 was found (70% of H4K20me, 40% of H4K CP 20me). This result implies that H4K20 and H4K CP 20 do compete for binding with SETD8, and that H4K20 possesses a somewhat higher binding affinity, which presumably leads to being a better substrate for SETD8. It is also possible that subtle differences in sterics and electronics of H4K CP 20 when compared to H4K20 do contribute to observed differences in the degree of methylation in the competition experiment. In line with observations that sterically demanding lysine analogs do not undergo SETD8-catalyzed methylation, we found that they also do not significantly inhibit monomethylation of H4K20 (Supplementary Fig. 31). These results are in agreement with inhibition and binding studies of related aromatic lysine analogs that exhibited limited ability to associate with SETD8 23 . Similarly, we observed that H3K CP 9 competes with H3K9 for G9a-catalyzed methylation, however, other bulkier lysine analogs do not significantly inhibit G9a-catalyzed methylation of H3K9 (Supplementary Fig. 32). We then moved on to investigate in more detail whether the histone peptides bearing unnatural lysine analogs that are not substrates for methyltransferase catalysis, have an ability to inhibit KMT-catalyzed methylation of H3K4 and H3K9. Inhibition studies were carried out employing MALDI-TOF MS assays . Initially, all unnatural histone peptides were screened for inhibition at 100 µM (Fig. 4). For H3K4 analogs it was found that all peptides have a very limited ability (IC 50 > 100 µM) to inhibit SETD7's methyltransferase activity, at most 11% inhibition was observed at 100 µM of H3F 4a 4. From the peptides bearing unnatural lysine analogs at position 9, we were pleased to find that H3F 3a 9 showed significant inhibition against G9a (IC 50 = 14.8 µM) and GLP (IC 50 = 26.0 µM), whereas other histone peptides showed a limited inhibition activity (Fig. 4 and Supplementary Figs. 33-34). For inhibition of GLP by H3K CP 9, we found that IC 50 ≈ 100 µM, whereas for the other analogs we observed IC 50 > 100 µM for both GLP and G9a. Having shown that H3K CP 9 acts as a substrate for GLP, 1D and 2D NMR spectra were acquired to further elucidate the chemical structure of the methylated H3K CP 9 product (Fig. 5). To characterize the methylated H3K CP 9 product of GLP-catalyzed reaction, 1 H NMR and 1 H- 13 C HSQC (Heteronuclear Single Quantum Coherence) spectra of the H3K CP 9 peptide were recorded prior to enzymatic reaction (Supplementary Fig. 35). We verified by NMR spectroscopy that GLP-catalyzed methylation of lysine residue in the H3K9 peptide gives indicative signals in the 1 H NMR spectrum, as also previously examined (Fig. 5a) 15,27 . The appearance of a triplet at 2.62 ppm was assigned to the SAH-CH 2 γ, a characteristic coproduct signal that appears during the methylation reaction of lysine residues by KMTs. In addition, a new resonance at 3.03 ppm indicated the formation of the trimethylated species of lysine residue at position 9. These data were also supported by 1 H- 13 C HSQC analysis (Fig. 5h). GLP-catalyzed methylation of histone peptides that bear unnatural lysine analogs was also examined by NMR spectroscopy (Fig. 5). As shown in Fig. 5b, 1 H NMR data of H3K CP 9 in the presence of SAM and GLP after 1 h at 37 °C showed new resonance peaks of the dimethylated product (H3K CP 9me2) at 2.73 ppm and the trimethylated product (H3K CP 9me3) at 2.99 ppm. A triplet of SAH-CH 2 γ was also observed at 2.62 ppm. A conversion of the cyclopropyllysine residue at position 9 to di-and trimethylated products was additionally confirmed by multiplicity-edited HSQC. The resonance at 2.73 ppm in the 1 H NMR spectrum is in a correlation with ( 13 C: 43.1 ppm) and represents the dimethylated product, whereas the resonance at 2.99 ppm is in a correlation with ( 13 C: 52.5 ppm) and represents the trimethylated product (Fig. 5i). The methylene protons of the attached cyclopropyl were unable to be observed due to very low concentration, however, chemical shift changes and the addition of new resonances for the cyclopropyl methylene indicate a transformation in the vicinity of the cyclopropyl group. Control reactions with H3K9 and H3K CP 9 in the absence of GLP showed no formation of methylated products and SAH, again demonstrating that methylation reactions are GLP-catalyzed (Supplementary Figs. 36-37). After showing that the H3K CP 9 peptide is dimethylated and trimethylated in the presence of GLP and SAM by NMR, we tested whether GLP catalyzed methylation of H3K ba 9, H3F 3a 9 , H3F 4a 9, H3A P 9 and H3Y9 peptides, and whether GLP mediated the conversion of SAM to SAH. In line with results from MALDI-TOF MS assays, a lack of new characteristic resonances, namely a triplet at 2.62 ppm (SAH-CH 2 γ) and a singlet in the range of 2.5-3.1 ppm (NMe, NMe 2 or NMe 3 ), indicates that these sterically demanding lysine analogs were not methylated in the presence of GLP (Fig. 5c-g and Supplementary Figs. 38-41). To gain additional insight into KMT-catalyzed methylation of bulkier lysine analogs, we carried out quantum mechanical/molecular mechanical (QM/MM) molecular dynamics and free energy studies on SETD8 and GLP in complex with K CP and F 3a . The free-energy profiles for the monomethylation reactions in SETD8 involving H4K20, two diastereoisomers of K CP (see the structure inserted in Fig. 6b and Supplementary Fig. 42) and F 3a are plotted in Fig. 6a. The free energy barriers for the methyl transfers obtained here are 20.0 and 19.3 kcal mol −1 for the two diastereoisomers of K CP , respectively, that are quite similar to the barrier when H4K20 was used as ). The active site structures of the reactant complexes for the methylations (Fig. 6b and Supplementary Fig. 42) show that the lone pair of electrons on N ζ of K CP can be aligned with the transferable methyl group even with the constrains of the three-membered rings. The free-energy profile for the methylation reaction in SETD8 involving F 3a shows that the free energy barrier becomes much higher (25.1 kcal mol −1 ), suggesting that the methylation reaction could not occur with this sterically demanding lysine analog even if this molecule was able to bind to the active site (Fig. 6c,d). The active site structure demonstrates that the transferable methyl group from SAM could not be aligned with the lone pair of electrons on N ζ for the methyl transfer to F 3a . In fact, the N ζ H 2 group is expected to be a part of the conjugated system containing the benzene ring, and one of the hydrogen atoms on N ζ (rather than the lone pair of electrons) would point to the transferable methyl group. Indeed, the distribution map on the right shows that the angle (θ) between the direction of electron lone pair on N ζ and the C M -S bond is between 45 and 120 degrees. In order to have the methylation reaction to occur, the N ζ H 2 group needs to undergo some rotations so that the lone pair of electrons can be aligned with the methyl group. Figure 6d shows that this is the case near the transition state where the N ζ H 2 group has undergone rotations with the lone pair of electrons pointing to the transferable methyl group. The free energy profiles for the first, second and third methylation reactions in GLP involving K CP are given in Fig. 7a. As evident from Fig. 7a, all the free energy barriers are rather low and similar (~18-19 kcal mol −1 ), suggesting that GLP is a trimethyltransferase for K CP , in agreement with the experiments. Figure 7b shows that for the reactant complex of the first methyl transfer, the transferable methyl group from SAM can be aligned with the lone pair of electrons on N ζ . By contrast, for the reactant complex of the third methyl transfer the transferable methyl group from SAM cannot be well aligned with the lone pair of electrons on N ζ (Fig. 7c). Nevertheless, the free energy barrier is rather low as well for the third methyl transfer to K CP (18.4 kcal mol −1 ), indicating that the www.nature.com/scientificreports www.nature.com/scientificreports/ methylation can still occur. The structure near the transition state for the third methyl transfer is plotted in Fig. 7d (and Supplementary Fig. 43). It is of interest to note that there seems to be some additional transition state stabilization through the interactions involving one of the methyl groups and Y1124. Such interactions may lower the free energy barrier, leading to the third methyl transfer. A similar explanation has been used to understand the substrate/product specificities of Suv4-20h2 28 . The free energy profile for the first methyl transfer to F 3a in GLP shows that the free energy barrier for the methyl transfer is quite high (23.6 kcal mol −1 ), suggesting that GLP cannot catalyze the methylation reaction for F 3a , as already verified experimentally (Supplementary Fig. 44). Similar to the case involving SETD8, the active site structure shows that the transferable methyl group from SAM cannot be aligned with the lone pair of electrons on N ζ in GLP for the methyl transfer to F 3a (Supplementary Fig. 45). ## conclusion Overall, our combined synthetic, enzymatic and computational studies, which examine histone peptides that contain sterically demanding lysine analogs, reveal that human histone lysine methyltransferases exhibit a limited ability to catalyze methylation of bulky lysine analogs. Although members of human KMTs do have an ability to catalyze methylation of cyclopropyl-containing lysine (K CP ) and to a lesser extent benzylamine-containing glycine (K ba ), they cannot methylate significantly bulkier and less nucleophilic aminophenylalanine, pyridine and tyrosine residues. Despite the biomedical importance of members of KMT family of enzymes, basic molecular requirements for efficient KMT catalysis are only partially understood. Our work provides an important insight into chemical aspects of KMT catalysis by highlighting that human KMTs can accommodate and catalyze methylation of lysine analogs that possess a slightly larger side chain (e.g. K CP ). Furthermore, we showed that the H3F 3a 9 peptide has an ability to inhibit G9a and GLP methyltransferase activity. This peptide may serve as a starting point for the development of more potent peptide-based inhibitors of G9a and GLP. Along with recent work that has demonstrated that KMTs accept chemically diverse SAM analogs as cosubstrates , our study shows that KMTs also possess an ability to catalyze methylation of substrates that mimic lysine. It is envisioned that similar approaches that rely on modern experimental and computational tools will advance our fundamental understanding of epigenetic processes that play essential roles in human health and disease. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods expression and purification of KMts. Proteins expression and purification were performed as described 15 . Briefly, the four human proteins (SETD8, SETD7, G9a and GLP) were expressed in E. coli BL21 (DE3)pLysS-rosetta cells in TB growth medium supplemented with Kanamycin and chloramphenicol. Cells were grown at 37 °C until an OD 600 of 0.5-0.6. The temperature was then reduced to 16 °C and isopropyl β-D-1-thiogalacttopyranoside (IPTG) was added. Cells were then harvested and lysed by sonication. Purification of the N-terminally his6-tagged KMTs was carried out using Ni-NTA affinity chromatography. Further purification was carried out using size-exclusion chromatography (SEC) using a Superdex-75 preparative grade column on an AKTA system. Protein was separated by SDS-PAGE on a 4-15% gradient polyacrylamide gel (Bio-Rad) and the concentrations were determined using the Nanodrop DeNovix DS-11 spectrophotometer. Histone peptides synthesis. The peptides, carboxylated at their C termini for SETD8, G9a and GLP, were synthesized manually using a cartridge (6 mL, 20 µm, Screening Devices B.V., The Netherlands). Amino acids residues protected with acid labile moieties employing fluorenylmethyloxycarbonyl (Fmoc) chemistry. Deprotected peptide H4K20 and its unnatural bulkier lysine derivatives for SETD8 substrate examination were prepared possessing the residues (GGAKRHRK 20 VLRDNIQ). Deprotected peptide H3K4 and its unnatural bulkier lysine derivatives for SETD7 substrate examination were prepared possessing the residues (ARTK 4 QTARKSTGGKA). Deprotected peptide H3K9 and its lysine analogs for G9a and GLP were prepared bearing the residues (ARTKQTARK 9 STGGKA). From a loading batch 0.5 mmol/g, a capacity of 0.21 mmol (100 mg) per each synthesis was employed to obtain the required sequence. All standard amino acids (3.0 equivalents) were coupled using HOBt (3.6 equivalents) and DIPCDI (3.3 equivalents) in dimethylformamide (DMF) for 1 h at room temperature. In case of cyclopropylamine peptide substrate, (1.5 equivalents) of the protected unnatural amino acid was used for the coupling. Fmoc deprotection was performed using 20% piperidine in DMF for 30 min. Modified amino acid residues at position 20 of H4 and positions 9 and 4 of H3 coupled with elongated time overnight to ensure efficient coupling. The Fmoc deprotection and the coupling of the residues were monitored using Kaiser test on few resin-beads. Coupling of the amino acids and Fmoc-deprotection were performed by rolling on a rotating-mixer RM-5 (CAT Zipperer, Staufen, Germany). After the final Fmoc removal, peptides were cleaved from the resin using a 2.5% triisopropylsilane (TIS) and 2.5% water in 95% trifluoroacetic acid (TFA). The peptides were precipitated in cold diethyl ether (−20 °C) and purified via preparative HPLC. The yields of SPPS were estimated as isolated yields, in which the molecular weights of individual peptides were calculated as TFA salts at Lys and Arg positions. The peptides were purified by RP-HPLC on a Phenomenex Gemini-NX C18 column and their purities were assessed using analytical HPLC. Bruker instrument in the reflectron positive mode. For regular methyltransferase standard conditions experiment which carried out in 30 µL total volume, the mixture contains peptide (100 µM), SAM (200 µM), SETD8 or SETD7 (2 µM), in assay buffer 50 mM Tris at optimal pH 8.0. In case of G9a and GLP, similar conditions were used, except (500 µM) of SAM was added to the reaction mixture. Samples were incubated in an Eppendorf vial 1.5 mL using thermomixer for 1 h at 37 °C. A 5 µL aliquot of the solution was mixed with 5 µL of MeOH, after which 5 µL of this mixture was mixed with 5 µL of α-cyano-4-hydroxycinamic acid matrix (CHCA, 5 mg/mL in 125:125 µL acetonitrile/water). The spots were placed on a stainless steel MALDI plate (MS 96 target ground steel BC of Bruker, Germany). The mass corresponding to one monomethylation observed as +14 Da, dimethylation observed as +28 Da and trimethylation observed as +42 Da. Data from a set of 100 laser shots (3×) were accumulated to give an acceptable spectrum. The enzymatic activity was determined by taking the peak areas of each methylation state, including all isotopes and adducts, and was annotated using FlexAnalysis software (Bruker Daltonics, Germany). None-enzyme and none-SAM controls experiments were carried out to ensure that the conditions of MS assay did not affect the noticeable methylation states. Methylated peptide substrates were repeated five times and the unmethylated substrates were triplicated. Sequences of the examined peptides are given in (Supplementary Table 1). inhibition studies. A mixture of histone peptide (0-100 µM final conc.) and SETD7, G9a or GLP (100 nM final conc.) was preincubated for 5 minutes at 37 °C in 18 µL of 50 mM glycine pH 8.8 containing 2.5% glycerol as assay buffer. Then 2 µL of a pre-mixture of SAM (20 µM final conc.) and 21-mer H3 histone peptide (residues 1-21, 5 µM final conc.) was added to afford a final assay mixture (20 µL) and the enzymatic reaction was incubated for an additional 30 minutes at 37 °C before quenching with 20 μL of MeOH. The quenched reaction (1 μL) was mixed with a solution of saturated α-cyano-4-hydroxycinnamic acid (5 μL) and spotted on the MALDI plate for crystallisation. The enzymatic activity was determined by taking the peak areas of each methylation state (including all isotopes and adducts) and is expressed relative to a control reaction in the absence of unnatural histone peptides. Each inhibition experiment was carried out in replicate. nMR assays. NMR enzymatic experiments for methyl transferase activities were performed with G9a. Incubations by an Eppendorf vials using thermomixer were carried out in 50 mM Tris-d 11 .HCl (pD 8.0) and 37 °C for 1 h. The samples (300 µL) typically contained G9a (8 μM) and SAM (2 mM), and H3K9 peptide (400 μM) or any of its sterically demanding analogs H3K cp 9/H3F 3a 9/H3F 4a 9/H3A p 9/H3Y9 peptide. After 1 h, the sample diluted to 550 μL and measured by 1 H NMR at 298 K. Controls were run in parallel at the same time. NMR spectra were acquired using a Bruker Avance III 500 MHz NMR spectrometer equipped with a Prodigy BB cryoprobe. The probe temperature was at 298 K in all instances. The 1D 1 H spectra were acquired in manual mode, whereas subsequent 2D experiments were acquired in full automation mode. Analysis parameters for 1 H NMR acquisition were: numbers of scans (NS) 256, relaxation delay 4 seconds, and spectral width (SW) 10 ppm. All the 1D experiments were performed with suppression of residual water signal by presaturation during the relaxation delay using presaturation (pulse program zgpr). Analysis parameters for 2D HSQC acquisition were: NS is 32, relaxation delay 1.5 seconds, acquired size 512, spectral width (SW) for 1 H was 11 ppm and 13 C was 160 ppm. When processing HSQC, additional measures such as a t1 noise reduction produced cleaner spectra. Spectral resolution for HSQC was enhanced by apodization. NMR data were processed using MestreNova software (version 10.0.2). All the spectra were phase and baseline corrected. QM/MD methods. To understand the experimental observations, the QM/MM free energy (potential of mean force) and MD simulations were undertaken for SETD8 and GLP to calculate the free energy profiles of the methyl transfers from SAM to some of the unnatural amino acid residues. Three-membered and six-membered rings were introduced into lysine sidechain using the CHARMM program 32 . The QM part of the systems included a portion (-CH 2 -CH 2 -S + (Me) -CH 2 -) of SAM and the lysine analog chains, and the rest of the system was described by MM. To separate the QM and MM parts, the link-atom approach 33 was applied; a modified TIP3P water model 34 was used for the solvent. The QM/MM simulations were based on the stochastic boundary molecular dynamics method 35 , which partitions the system into a reaction zone and a reservoir region. The reaction zone was further divided into a reaction region and a buffer region. The radius r for reaction region was 20 with the buffer region extended over 20 ≤ r ≤ 22 . The N ζ atom of the lysine analogs was used the reference center for partitioning the system. The final systems for the QM/MM simulations had around 5300 atoms (including roughly 900-1000 water molecules). For the QM atoms, the DFTB3 method 35 was used. This semi-empirical approach has been used on a number of systems previously with reasonable results obtained 36 . The PARAM27 of all-hydrogen CHARMM potential function 37 was adopted here for the MM atoms. The reactant complexes of the methylation were generated based on the crystal structures of the enzyme complexes (SETD8: PDB ID = 2BQZ; GLP: PDB ID = 3HNA); SAM was generated by adding a methyl group to SAH. The methyl lysine was changed to lysine by removing the methyl group manually. The two three-membered rings and one six-membered ring were introduced to the lysine sidechain to generate the three lysine analogs with steric constrains. The stochastic boundary systems were first optimized based on the steepest descent (SD) and adopted-basis Newton-Raphson (ABNR) methods and then gradually heated from 50.0 to 298.15 K in 50 ps. The time step used for integration of the equation of motion was 1-fs, and for every 50 fs the coordinates were saved for analyses. 1.5 ns QM/MM MD simulations were performed for each of the reactant complexes 28,38 . To determine the changes of the free energy (potential of mean force) as a function of the reaction coordinate for the methyl transfer in SETD8 and GLP, respectively, the umbrella sampling method 39 along with the Weighted Histogram Analysis Method (WHAM) 40

chemsum

{"title": "Examining sterically demanding lysine analogs for histone lysine methyltransferase catalysis", "journal": "Scientific Reports - Nature"}

isolated_neutral_[4]helicene_radical_provides_insight_into_consecutive_two-photon_excitation_photoca

## Abstract: Direct activation of strong bonds in readily available, benchtop substrates offer a straightforward simplification, albeit in most cases existing catalytic systems are limited to unlock such activation. In recent years, a surge of in-situ generated organic radicals that can act as potent photoinduced electron transfer (PET) agents have proved to be a powerful manifold for the activation of remarkably stable bonds. Herein we document the use of N,N′-di-n-propyl-1,13-dimethoxyquinacridine ( n Pr-DMQA • ), an isolated and stable neutral helicene radical, as a highly photoreducing species. This isolable doublet state open shell radical offers a unique opportunity to shed light on the mechanism behind PET reactions of organic radicals. Experimental and spectroscopic studies revealed that this doublet radical has a long lifetime of 4.6 ± 0.2 ns, an estimated excited state oxidation potential of -3.31 V vs SCE, and can undergoes PET with organic substrates. The strongly photoreducing nature of the n Pr-DMQA • was experimentally confirmed by the demonstration of photo activation of electron rich aryl bromides and chlorides. We further demonstrated that n Pr-DMQA • can be photochemically generated from its cation analog ( n Pr-DMQA + ) allowing catalytic functionalization of aryl halide via a consecutive photoexcitation mechanism (ConPET). Dehalogenation, photo-Arbuzov, photo-borylation and C-C bond formation reactions with aryl chlorides and bromides are reported herein, as well as the α-arylation of carbonyl using cyclic ketones. The latter transformation exhibits the facile synthesis of α-arylated cyclic ketones as critical feedstock chemical for diverse useful molecules, especially in the biomedical enterprises. ## INTRODUCTION Over the past decade, photoredox catalysis has received a fast and growing interest from the world of synthetic chemistry. 1 By combining visible light with a photocatalyst (PC), a large variety of efficient and selective transformations have been achieved under mild conditions, during which the excited state PC is involved in a single electron transfer (SET) with a substrate or a co-catalyst. However, SET with conventional photocatalyst is usually limited to reduction potential down to -2.0 V vs. Saturated Calomel electrode (SCE). 2 In recent years, several elegant reports have shown that open shell doublet radicals, generated in-situ, either electrochemically 3 or photochemically, can act as potent photoreducing agents. 4 For example, a two-photon excitation process, commonly purported as consecutive photoelectron transfer (conPET) pathway, is proposed as the photochemical generation pathway of potent photoreducing organic radicals (Figure 1a, left). 5 In that process, the excited state of a close shell single photocatalyst PC* (neutral or cationic) is generated upon the first excitation. Then, a sacrificial electron donor can act as a reductant and participate in SET with PC* to generate PC • radicals (anionic or neutral). A second successive photoexcitation generates the radical excited state PC • * that can act as super photoreducing agent (E1/2 red * = -2.3 to -3.4 V vs SCE) (Figure 1a). 6 This concept of two-photon excitation process has been reported with numerous notable photocatalysts such as PDI, 7 DCA, 8 anthraquinone, 9 Rhodamine 6G, 10 benzo[ghi]perylene (BPI), 11 4-DPAIPN, 12 3-CzEPAIPN, 13 Mes-Acr, 14 and Deazaflavin 15 (Figure 1b). As a common benchmark reaction, photoredox C(sp 2 )-X bond activation in aryl bromides and chlorides, Birch reduction, and sulfonamide cleavage showcased the extreme photoreducing ability of radical photocatalysts in most cases. Despite these convincing reports, an intriguing aspect of photoactive open shell doublet radicals generated in-situ is that alternative mechanisms cannot be completely ruled out. For example, Leonori et. al. recently reported that α-aminoalkyl radical, generated via SET between alkyl amines and a close shell excited PC*, can initiate halogen atom transfer (XAT) (Figure 1a, right). 16,17 Therefore, the concept of super photoreducing radicals in conPET systems is weakened due to the common usage of amine as the sacrificial electron donor. Furthermore, a recent study by Nocera and colleagues questioned the viability for radicals generated in-situ during conPET or electrophotocatalysis to act as efficient photocatalysts, due to their short-lived excited state which should hamper their participation in bimolecular reactions with substrates. They concluded that, instead, a close shell singlet species, such as a Meisenheimer complex or side product impurities, formed from the reactive open shell doublet radical can act as the super reducing photoreagent (Figure 1a, top). 18 To date, stable and isolatable radicals able to undergo photoinduced electron transfer during an organic transformation remain elusive. Therefore, the synthesis and isolation of such photoactive organic radicals are of great interest in order to shed light on the ability for open shell doublet species to act as photoreducing agents. As part of our ongoing interest in the photochemical properties of the helical carbenium system, we recently reported the use of N,N′-di-n-propyl-1,13-dimethoxyquinacridinium ( n Pr-DMQA + ) tetrafluoroborate 19 as an organic photoredox catalyst for photoreductions and photooxidations under red light (λmax = 640 nm). 20 Several fundamental organic transformations involving either oxidative quenching or reductive quenching pathways have been demonstrated. During these studies, we identified the neutral helicene radical ( n Pr-DMQA • ) as a possible radical intermediate in the reductive quenching photocycle. In our contemporary studies, we reported the chemical synthesis, isolation, and characterization of n Pr-DMQA • as part of a family of neutral quinolinoacridine radicals from their quinolinoacridinium cation analogs. 21,22 Our studies showed that these radicals are highly persistent in their solid form as well as in solution for several months under inert conditions, and reversibly oxidize back to the cation upon exposure to air. We now report that the stable helicene radical n Pr-DMQA • , first observed by Larsen et. al. 22 and isolated by our group, 21 is a highly photoactive species with strong absorptions of light in the visible region. The excited state oxidation potential of this helicene radical has been estimated to be -3.31 V vs SCE which makes it one of the strongest photoreductants. This radical can be made on gram scale, isolated, purified and stored in a glovebox, which allowed us to investigate its photophysical and photochemical properties, as well as its ability to act as a strong photoreducing agent without questioning the involvement of impurities or side products. Furthermore, the closed shell cation counterpart, n Pr-DMQA + , is photoactive in red light, while n Pr-DMQA • is not, which offers a unique opportunity to probe the mechanism for the photoactivation of aryl bromides and chlorides under blue light excitation. ## RESULTS AND DISCUSSION Photophysical properties of n Pr-DMQA • radical. We recently reported that the chemical reduction of n Pr-DMQA + allows the synthesis and isolation of the stable double state organic radical n Pr-DMQA • . 21 Interestingly this radical was found to possess strong absorption of light in the visible region (391 nm, 440 nm, 467 nm, and 557 nm) and exhibits emission maxima at 593 nm (Figure 2a, and Figure S1 -S3). The life-time of the excited state was determined using time correlated single photon counting (TCSPC, see Supporting information). Excitation of n Pr-DMQA • resulted in a strong emission band characteristic of n Pr-DMQA • * (λ𝑚𝑎𝑥 𝑒𝑚 = 593 nm) whose average lifetime (τ) was measured to be 4.6 ± 0.2 ns (Figure 2b, and Figure S5 -S9). 23 Interestingly, the fluorescence lifetime of the open shell doublet radical was found to be longer than its singlet cation analog (5.7 ns for n Pr-DMQA + *). 24 A multi nanosecond scale excited state lifetime is suitable for bimolecular electron transfer suggesting that the doublet neutral radical n Pr-DMQA • could act as an effective PET agent. 25 As previously oberved, 20d, 21 the cyclic voltammogram of the n Pr-DMQA scaffold in acetonitrile revealed the presence of two reversible events, E1/2 ( n Pr-DMQA ++• / n Pr-DMQA + ) = + 1.27 V and E1/2 ( n Pr-DMQA + / n Pr-DMQA • ) = -0.85 V vs SCE when recorded between -1.5 V and + 1.5 V vs. SCE (Figure 2c, i) trace). However, when the potential window is extended to reach -2.0 V vs. SCE, an irreversible event at E1/2 ( n Pr-DMQA • / n Pr-DMQA -) = -1.84 V vs SCE is observed, which then triggers the generation of an intermediate with an irreversible event at + 0.67 V vs SCE (Figure 2c, ii) trace). Laursen et. al. have previously assigned this intermediate as the neutral close shell singlet n Pr-DMQA-H, which forms by reaction between acetonitrile and the highly reactive n Pr-DMQA -. Based on the electrochemical and photophysical properties of n Pr-DMQA • (E1/2 ox (C + /C • )= -0.85 V vs SCE, E1/2 red (C • /C -) = -1.84 V vs SCE, , and excitation energy E0,0 = 2.15 eV at λex =557 nm and E0,0 = 2.46 eV at λex = 440 nm, the excited state redox potentials of this helicene radical are estimated to be E*1/2 ox (C + / C •* )= -3.03 V vs SCE (λex = 557 nm), E*1/2 ox (C + / C •* ) = -3.31 V vs SCE (λex = 440 nm), and E*1/2 red (C •* /C -) = + 0.45 V vs SCE (see supplementary information). As a result, n Pr-DMQA • can be described as a mild photooxidant and one of the most potent photoreductant. Considering recent reports suggesting that in-situ generated close-shell singlets can be involved during photochemical transformations, 18 we turned our focus on n Pr-DMQA-H. Following Lacour et. al. synthetic protocol, 26 we have synthesized and studied the electro-and photophysical properties of this neutral close shell singlet (see supporting information). The absorption spectroscopy reveals a photo inactive species that possesses an absorption at 316 nm and no emission. The cyclic voltammogram of n Pr-DMQA-H (Figure 2c, iii) traces) reveals an irreversible oxidation event at +0.67 V vs SCE, followed by a reversible oxidation at + 1.27 V vs SCE which was assigned to the n Pr-DMQA ++• / n Pr -DMQA + redox couple. Importantly, no event is observed at negative potential during the first cycle (solid trace), however the reversible event associated to n Pr-DMQA + / n Pr-DMQA • appears after the second cycle (doted trace) suggesting that n Pr-DMQA + is electrochemically generated from the oxidation of n Pr-DMQA-H. Consistent with these electrochemical data and the previous observation by Lacour, UV-Vis spectroscopy monitoring of the colorless n Pr-DMQA-H revealed the slow formation of the green n Pr-DMQA + upon exposure to air in acetonitrile (see Figure 2d, left). 27 Interestingly, n Pr-DMQA-H was stable in acetonitrile in absence of oxygen and in the dark, but undergoes photoinduced homolysis to form n Pr-DMQA • when irradiated with a 440 nm LED. This conversion was also monitored by UV-Vis spectroscopy (Figure S12). These results support that even if formed during photocatalysis (vide-infra), the closed shell single n Pr-DMQA-H is expected to convert back to n Pr-DMQA • via photolysis and/or to n Pr-DMQA + under oxidative conditions, undermining its involvement as a possible photoinduced electron transfer agent. S15 and S17). However, under 440 nm irradiation we observed the disappearance of the radical n Pr-DMQA • absorption band and the appearance of the cation n Pr-DMQA + absorption bands (Figure 3a). These observations suggest that the excited state n Pr-DMQA • * can undergo an oxidative quenching process via SET with aryl halide to generate an aryl radical anion and n Pr-DMQA + . The photoredox properties of n Pr-DMQA • discussed above also suggest that in presence of a suitable donor (E1/2 > +0.45 V vs SCE), reductive quenching of n Pr-DMQA • * can be observed resulting in the formation of the highly reactive n Pr-DMQAwhich will then rapidly convert to the stable singlet closed shell n Pr-DMQA-H (vide-supra). UV-Vis spectroscopy monitoring of a solution of n Pr-DMQA • and a large excess of pyrrolidine under 440 nm LED irradiation revealed the rapid and quantitative formation of n Pr-DMQA-H (Figure 3b). A large excess of amine was required to drive the reaction toward the formation of n Pr-DMQA-H and overcome the reverse photolysis of n Pr-DMQA-H back to n Pr-DMQA • . To further probe the viability of a photoexcited direct SET between n Pr-DMQA • and aryl halides we performed a stoichiometric photo-Arbuzov reaction using different light sources (440 nm and 640 nm), different DMQA species ( n Pr-DMQA • , n Pr-DMQA + , n Pr-DMQA-H), and different aryl bromide substrates with a range of reduction potentials (4-bromo benzonitrile 2a, E1/2 red = -1.94 V vs SCE; and 4-bromo anisole 2b, E1/2 red = -2.90 V vs SCE) (Figure 3c, and supporting information). Equimolar amounts of aryl bromide and DMQA species were irradiated for 16 h with visible light in presence of 3.0 equivalent of the aryl radical trapping agent, triethyl phosphite P(OEt)3. Using DMQA + , under either 440 nm or 640 nm irradiation, did not produced any aryl activated product consistent with the mild photoreducing potential of this singlet cation species (See Figure 3c, entry 1 -2). On the other hand, when the doublet neutral radical n Pr-DMQA • was irradiated under 440 nm both aryl halides were activated, with a higher yield for the less electron rich bromo benzonitrile (3a, <90% yield) and 3b, 50% yield, Figure 3c, entry 3). Furthermore, the red color of the reaction mixture, characteristic of the helicene radical, turned to the blue/green color of the helicenium ion supporting the previously observed reversibility between radical and cationic form of the DMQA system. Importantly, no product formation for the electron rich 4-bromo anisole was observed when red light irradiation (640 nm) was used (Figure 3c, entry 4), consistent with the weak absorption of n Pr-DMQA • at wavelength higher than 600 nm (Figure 1c). However, full conversion was detected under red light with the more easily reducible 4-bromo benzonitrile suggesting that an alternative pathway to PET maybe involved with this easily activated substrate (Figure 3c, entry 4). Finally, we noted a similar outcome when using n Pr-DMQA-H (Figure 3c, entry 5 -6), which supports the formation of n Pr-DMQA • under light. These results substantiated our hypothesis that the doublet neutral helicene radical n Pr-DMQA • * can undergo rapid PET with substrates such as electron rich aryl bromide under 440 nm light. To further support these observations and the ability for n Pr-DMQA • to undergo single electron transfer with aryl halides, we performed transient absorption measurements using a home-built apparatus with broadband detection (see supporting information). Transient absorption data was collected for the neutral radical in acetonitrile and in an acetonitrile/4-bromo anisole (3.0 equiv.) solution. Figure 3d (i) presents the transient absorption data for the n Pr-DMQA • neutral radical in acetonitrile. The region corresponding to the maximum of the negative ground state bleach signal has been omitted due to pump scatter. Stimulated emission signal expected appear at the maximum of the fluorescence emission is not evident as it is overlapped with the strong excited state absorption signal. Two excited state absorptions that appear as positive signals with maxima at 523 nm and 588 nm are present in the data on either side of the ground state bleach. Both excited state absorption contributions fully decay by 48 ps as shown in Figure 3d (i). At longer delay times, the only remaining signal arises from the recovery of the ground state bleach, which corresponds to the 4.6 ± 0.2 ns excited state lifetime measured by TCSPC. Figure 3d (ii) presents the transient absorption for the n Pr-DMQA • neutral radical in acetonitrile in the presence of an aryl halide. In comparison to the n Pr-DMQA • alone in acetonitrile, the addition of the aryl halide produces significant changes in the transient absorption data. Both samples have similar nanosecond ground state recoveries (negative signal centered at 558 nm) consistent with the measured fluorescent lifetime, with different dynamics evident for the positive going contributions on either side of the ground state bleach. The positive signal features in the neutral radical in acetonitrile data decay within tens of picoseconds. In contrast, the signal for the neutral radical in the presence of an aryl halide has positive features that persist for nanoseconds. The maxima of the lower energy positive feature in the data with the aryl halide is peaked at 615 nm rather than 588 nm for the acetonitrile-only solution. This indicates the in-situ generation of the cation n Pr-DMQA + photoproduct whose S0 to S1 transition is peaked at 616 nm. The persistence of the cation photoproduct signal is consistent with the previously measured excited state lifetime of 5.7 ns. Photocatalytic activity of n Pr-DMQA radical and mechanism. Inspired by the stoichiometric photo-Arbuzov experiment and our previous reports on the photoactivity of n Pr-DMQA + , we questioned if n Pr-DMQA • can be photochemically generated from n Pr-DMQA + in presence of a light source and an electron donor, allowing the use of n Pr-DMQA • as potent photoreducing agent for catalytic transformations. Consistent with our previous observation, the successful generation of the helicene radical n Pr-DMQA • was detected in an EPR experiment, as well as using UV-Vis spectroscopy, in presence of three different electron donors under both blue and red-light sources (Figure 4a, and Figure S24 -S28). Next, we attempted the catalytic reductive dehalogenation of 4-bromobenzonitrile (E1/2 = -1.94 V and BDE = 80 kcal/mol) 28 and 4-bromoanisole (E1/2 = -2.90 V and BDE > 85 kcal/mol) in presence of pyrrolidine with n Pr-DMQA + or n Pr-DMQA • as a photocatalyst under either blue or red light (Figure 4b, and supporting information). In case of 4-bromo anisole, both PCs induced a dehalogenation reaction under blue light with similar yields. The near identical yield for both PCs supports a mechanism during which n Pr-DMQA + and n Pr-DMQA • are involved in the photocatalytic cycle (Figure 4b). However, no product was detected with either PCs when low energy light was employed, consistent with the fact that n Pr-DMQA • has little to no absorption in red light, and that n Pr-DMQA +* is not reductive enough to undergo PET with aryl halides. Similarly, with the electron poor bromobenzonitrile, both PCs afforded full conversion under blue light (Figure 4b). However, unlike with anisole, some conversions were detected with both n Pr-DMQA + and n Pr-DMQA • under low energy red light. We recently, reported that photochemically generated ammonium radical can initiate XAT mechanism with electron poor substrate such as 4-bromobenzonitrile but not with 4-bromoanisole. In the present system, pyrrolidine acts as an electron donor and forms pyrrolidinium radical during the photoreduction of n Pr-DMQA + to n Pr-DMQA • or n Pr-DMQA • to n Pr-DMQA -. The lower yield observed in red light compared to blue light for bromobenzonitrile supports that single electron transfer between n Pr-DMQA •* and aryl halides is the most effective mechanistic pathway. Together, these observations suggest that for electron poor substrate such as bromobenzonitrile both SET and XAT mechanism are accessible pathways in blue light while XAT is the only viable mechanism under red light. On the other hand, for electron rich species such as 4-bromo anisole, SET from n Pr-DMQA •* is the only sustainable mechanism. Interestingly, at the end of the catalytic transformations under blue light, the reaction mixtures were a pale almost colorless solution which rapidly turn to the green color of n Pr-DMQA + when exposed to air. This observation suggests that at high conversion (excess amine and low concentration of aryl halide) n Pr-DMQA-H generated from n Pr-DMQA •* build up in solution and can be considered as an off-cycle catalytic resting state. Based on these experimental results and previous reports, we proposed a plausible conPET pathway for the photoactivation of electron rich aryl halides using n Pr-DMQA + as photocatalyst under blue light (Figure 4c). First the photoexcitation of n Pr-DMQA + results in an excited cationic n Pr-DMQA + * under blue light irradiation. A single electron transfer (SET) from the sacrificial electron donor D (pyrrolidine, DIPEA or enamine) generates the neutral helicene radical n Pr-DMQA • . As the radical species is photoactive in the blue light region the second photoexcitation leads to the excited state helicene radical, n Pr-DMQA • *. Now the electrochemical potential of this excited neutral radical (E1/2 red < -3.31 V vs SCE in CH3CN, τ = 4.6 ± 0.2 ns) is strong enough to reduce aryl chlorides and bromides via SET closing the catalytic cycle by regenerating the cationic helicenium n Pr-DMQA + . Finally, after reduction the aryl halide radical anion fragmentate to the corresponding aryl radical which can abstract a hydrogen to form the dehalogenated product, or be further coupled with suitable substrates (phosphites, boranes, pyrroles or carbonyls). In a non-productive pathway, the excited state helicene radical n Pr-DMQA • * can also generate n Pr-DMQAvia SET with pyrrolidine, followed by the generation of n Pr-DMQA-H which can fragmentate back to n Pr-DMQA • . ## Photocatalytic dehalogenation of aryl halides. With the photo-excited behavior of n Pr-DMQA • and a probable mechanism for consecutive photoelectron transfer elucidated, we sought to use this neutral helicene radical as a photoreducing catalyst for the functionalization of aryl halides. At first, reductive dehalogenation of aryl halides was performed as the benchmark reaction to evaluate the extent of the extremely potent reducing behavior of the helicene radical. Dehalogenation of 4-bromo anisole (E1/2 red = -2.9 V vs SCE) was optimized using n Pr-DMQA-BF4 as the photocatalyst under a 440 nm light source. Pyrrolidine was identified as an efficient sacrificial electron donor for the generation of n Pr-DMQA • from n Pr-DMQA + , in situ. After optimization (see Table S1), we observed that 5.0 mol% n Pr-DMQA-BF4 as the photocatalyst in presence of 440 nm blue light and 3.0 equiv. of pyrrolidine furnished the desired hydrodebrominated product (anisole) in 94% yield within 16 h. Following the identification of optimal reaction conditions, the competency of this strong radical photoreductant has been demonstrated by the reduction of a wide range of electronically diverse aryl chlorides and bromides (Table 1). Successful dehalogenation of a variety of electron rich and electron poor aryl chlorides with reduction potential of -2.9 V vs SCE and lower have been demonstrated in good to excellent yields (4a-4f, 81-99 % yield). We then focused on diverse electron rich aryl bromides (4g-4l, 90-98 % yield) along with a myriad of functional group containing substrates (4m-4s, 71-99 % yield) for excellent hydrodebromination reaction. Polyaromatic substrates and heteroaromatic bromides were also found to be efficient substrates for reductive dehalogenation (4t-4x, 80-96 % yield). Furthermore, the bisreduction of polyhalogenated arenes gave the corresponding bis-dehalogenated products in moderate yield (4y and 4z, 73 % and 78 % yield, respectively). In general, aryl halides bearing nitrile (4c, 4o), ester (4d, 4n), trifluoromethyl (4e, 4s), ketone (4m), free acid (4p), free hydroxy (4q) and free amine (4r) groups were well tolerated and found to be excellent substrate for this neutral helicene radical catalyzed photoredox dehalogenation method. ## Table 1: Substrate scope for the photoredox reductive dehalogenation of aryl halides enabled by neutral helicene radical. Photocatalytic functionalization of aryl halides. After successful examination of hydrodehalogenation reaction, we decided to extend this radical photocatalysis to other arene-functionalization processes (Table 2). Aryl phosphonates are important structural motifs found in many pharmaceutically active molecules 29 and easily accessed by photo-Arbuzov reaction using triethyl phosphite, P(OEt)3. 30 Addition of 3.0 equiv of P(OEt)3 to the optimal reaction condition of hydrodehalogenation reaction furnished aryl phosphonates in high isolated yields (Table 2 and S2). Using ConPET enable by n Pr-DMQA • allow us to expand scope of reactions towards less reactive aryl bromides or aryl chlorides bearing very negative reduction potentials. Electron rich aryl halides with substituents in ortho, meta and para position demonstrated excellent reaction yields (7a-7c, 81-91 % yield). Aryl bromides and chlorides with electron withdrawing groups also showed good conversion to the desired product (7d-7f, 76-89 % yield). In addition to that, polyaromatic substrate (7g, 90% yield) and heteroaromatic substrates furnished the corresponding aryl phosphonate in excellent yields (7h-7j, 84-92 % yield). Following the photo-Arbuzov reaction, borylation reaction for synthesis of aryl borate which is considered as important coupling partner in late-stage derivatization, 31 was examined with this established ConPET using n Pr-DMQA • . The addition of dipinacol diborane to optimized reductive dehalogenation condition ended up with the formation of aryl borate in good yields (7k-7l, 79-86 % yield). Furthermore, we explored the photoredox generation of aryl radicals from aryl halides for C-C bondforming reactions with arene such as N-methyl pyrrole. The desired arylated products were isolated in good yields in presence of 3.0 equiv. of N-methyl pyrrole (7m-7n, 88-90 % yield). During these transformations, dehalogenation was detected as side reaction. The use of DIPEA (Di-isopropylethylene diamine) as electron donor instead of pyrrolidine was found to give more selective result (see supporting information). ## Table 2: Scope for the reductive functionalization of aryl halides under photoredox condition catalyzed by neutral helicene radical. Photocatalytic α-arylation of cyclic ketones. The α-arylated cyclic ketone scaffolds are medicinally significant as well as critical building blocks for numerous pharmaceutical agents and bioactive natural products. 32 Unsurprisingly, transition-metal catalyzed α-arylation of carbonyls are well established reactions in classic organic chemistry. Typically, expensive transition metal catalysts and ligands, harsh reaction condition or multistep protocols are employed in the synthesis of medicinally relevant α-arylated cyclic ketones. 33 Nonetheless, considering the medicinal impact of such metal-mediated methods for the synthesis of pharmaceutical drug molecules, a completely organic catalyst based direct α-arylation reaction is in high demand. 34 Recently, our group reported photoredox α-arylation of carbonyl compounds using an electron rich acridinium as photocatalyst. However, scope of aryl halide was limited mainly to aryl iodides with moderate to good yields due to the lower excited state potential of our acridinium photocatalyst. 35 Using the present DMQA system as conPET photocatalyst for α-arylation of carbonyls would allow to expand the reaction scope to electron rich aryl bromides. In our contemporary report, 35 we demonstrated that in-situ enamine formation via condensation reaction between the carbonyl and the amine moieties was required. We further demonstrated that the cyclic amine, pyrrolidine, provided the highest yield due to the rapid enamine formation, and that both the pyrrolidine and its enamine analog acted as sacrificial electron donor. The EPR analysis of n Pr-DMQA-BF4 in presence of enamine confirms the formation of n Pr-DMQA • in a photoexcited state. Furthermore, using a readily available enamine (1-pyrrolidino-1-cyclohexene) as sacrificial electron donor instead of pyrrolidine, under optimized condition of reductive dehalogenation, resulted in the formation α-arylated of cyclohexanone, 85% yield. (See supporting information). Literature studies showed that enamine radical cations mainly exhibit the C-center free radical property due its high spin population, 36 which can lead to radical recombination pathway for generation of alpha-substituted product. Hence, to develop the helicene radical catalyzed α-arylation of carbonyls we employed cyclohexanone and pyrrolidine with aryl bromides. Optimizations under catalytic condition showed that 4bromoethyl benzoate (1.0 equiv.) in presence of 5 mol% of n Pr-DMQA-BF4 and combination of cyclohexanone with pyrrolidine (3.0 equiv. each) furnished the desired alpha-arylated product in 86 % yield (See Table S4). With these optimal conditions for the alpha-arylation of cyclic ketones in hand, we examined the scope with respect to the aryl bromide component. As shown in Table 3, para-substituted bromoarenes containing functional groups like ester, carbonyl and free acid generated the corresponding alpha-arylated product in very good yields (8a-8d, 84-88% yield). Medicinally significant functional groups like nitrile and trifluoromethyl substituted arylbromides were also evaluated. Gratifyingly, substitution in para or meta position did not affect the product yield neither in nitrile (8e and 8f, 86 and 82 % yield) nor in trifluoromethyl containing substrates (8g and 8h, 75 and 76% yield). Fluorinated bromobenzene and unsubstituted bromobenzene were also well accommodated, furnishing alpha-arylated adducts in good yield (8i and 8j, 81 and 80% yield). In addition, polyaromatic and heteroaromatic bromoarene like 9bromophenanthrene and bromoindole were also found to be excellent substrate for this transformation (8k and 8l, 74 and 76% yield). Finally, as a demonstration that this method can be extended to the installation of multiple chiral centers containing arenes, alpha-arylation can be accomplished using this new protocol to provide the corresponding alpha-arylated cyclic ketone in excellent efficiencies (8m and 8n, 82 and 80% yield). Similar to aryl bromides, different cyclic ketones were also evaluated with respect to the optimal conditions. As shown in Table 3, a series of differentially substituted cyclohexanone-derived substrates were readily coupled with an aryl radical. It is of note that incorporation of both alkyl and aryl substituents at positions 4 of the cyclohexanone ring is well-tolerated (8o-8s, 78−84% yield). As expected, the presence of single substituent at the 4-position in the cyclohexanone ring induced higher levels of diastereoselectivity in product. Disubstituted cyclohexanones at the 4-position of the ring also successfully transformed to the corresponding alpha-arylated derivatives in good yield (8t and 8u, 86 and 82 % yield). Spirocyclic cyclohexanone derivative and heteroatom containing cyclic ketone were also well tolerated in optimal reaction condition (8v and 8w, 80 and 78 % yield). Interestingly, cyclopentanone was also found to be an efficient ketone substrate for this transformation with an excellent yield (8x, 85% yield). Table 3: Helicene radical catalyzed photoredox ɑ-arylation of cyclic ketones using aryl bromides. a Reactions were run on 0.2 mmol scale. Isolated yields are reported. See Supplementary Information for details. ## CONCLUSIONS While several reports have supported the involvement of openshell doublet radicals as potent photoreducing species, the isolation of a stable photoactive radical that can allow an extensive mechanistic study of photoinduced electron transfer during an organic transformation has remained elusive. Herein, we have reported that N,N′-di-n-propyl-1,13-dimethoxyquinacridine ( n Pr-DMQA • ), a stable and isolable open shell doublet radical, is a photoactive neutral helicene radical. The facile synthesis and isolation of this helicene radical asl allowed us to investigate its photophysical properties, photochemical reactivities and photocatalytic abilities. First, we reported that n Pr-DMQA • possess strong absorption of light in the visible region (391nm, 440 nm, 467 nm, and 557 nm), exhibits emission maxima at 593 nm, and an excited state lifetime of 4.6 ± 0.2 ns. The photophysical and electrochemical properties of n Pr-DMQA • suggest that this radical possesses an estimated excited state oxidation potential of -3.31 V vs SCE (E1/2(C + /C • *)) and excited state reduction potential of + 0.45 V vs SCE (E1/2(C • */ C -)). Monitoring by UV-Visible spectroscopy the irradiation of n Pr-DMQA • in acetonitrile at 440 nm in presence of electron acceptor (aryl halide) or electron donor (amine) revealed that in both cases photoinduced electron transfer occurred leading to the formation of the cationic n Pr-DMQA + and the anionic n Pr-DMQArespectively. The anionic n Pr-DMQAwas found to rapidly convert to the photo-inactive closed shell singlet n Pr-DMQA-H. These observations were further supported by transient absorption spectroscopy which showed that n Pr-DMQA* • undergoes efficient single electron transfer with an aryl halides electron acceptor -. We then probed the photoreducing ability of n Pr-DMQA • using stoichiometric photo-Arbuzov reaction with an electron poor (4-bromo benzonitrile) and an electron rich (4-bromo anisole) aryl bromide, under both 440 and 640 nm. The results obtained, coupled to the control experiments using n Pr-DMQA + and n Pr-DMQA-H, substantiate that the helicene radical is a potent photoreducing agent under 440 nm irradiation. Similar results were obtained for the catalytic dehydrogenation of the same aryl bromides. Experimental and spectroscopic studies suggest that the neutral helicene radical act as a strongly reducing species and is photochemically regenerated from the cationic helicenium analog, implying that a consecutive photoexcitation mechanism is the most viable mechanistic pathway. The strongly photoreducing nature of a neutral helicene radical n Pr-DMQA • was further used for the well-studied photo-dehalogenation, photo-Arbuzov, photo-borylation and C-C bond formation reactions. Additionally, this catalytic system was used in synthesis alpha-arylated cyclic ketones which is considered as an important building block for medicinal chemistry. In summary, we believe that the conPET process enabled by this neutral helicene radical, together with its operational simplicity and sustainability, will help to understand radical photoredox catalysis as well as considered as alternative way for streamline the synthesis of complex functionality in both academia and industry.

chemsum

{"title": "Isolated Neutral [4]Helicene Radical Provides Insight into Consecutive Two-Photon Excitation Photocatalysis", "journal": "ChemRxiv"}

three-dimensional_large-pore_covalent_organic_framework_with_stp_topology

## Abstract: Three-dimensional (3D) covalent organic frameworks (COFs) are excellent porous crystalline polymers for numerous applications, but their building units and topological nets have been limited. Herein we report the first 3D large-pore COF with stp topology constructed with a 6-connected triptycene-based monomer. The new COF (termed JUC-564) has high surface area (up to 3300 m 2 g -1 ), the largest pore (43 Å) among 3D COFs, and record-breaking low density in crystalline materials (0.108 g cm -3 ). The large pore size of JUC-564 is confirmed by the incorporation of a large protein. This study expands the structural varieties of 3D COFs as well as their applications for adsorption and separation of large biological molecules. Covalent organic frameworks (COFs), a remarkable class of organic porous crystalline materials with high surface areas and promising stabilities, have attracted wide interests in varied fields including gas adsorption and separation, catalysis, optoelectronics, and some others. Over the past decade, most researches have been focused on two-dimensional (2D) COFs with eclipsed AA stacking modes. Three-dimensional (3D) COFs are considered as ideal platforms for abundant uses because of their interconnected channels, superior surface areas, and fully exposed active sites. However, only few topologies are available for 3D COFs so far, such as ctn, bor, dia, and pts, and almost all of them are based on tetrahedral building blocks, which have extremely limited the structural diversities of 3D COFs. 5 Interestingly, Wang, Feng and co-workers synthesized the first 3D anionic COFs with rra topology, CD-COFs, in which each boron atom is joined to four γ-cyclodextrin struts, and each γ-cyclodextrin is connected to eight boron atoms. 37 Thomas, Roeser and coworkers have recently demonstrated a novel 3D anionic silicate COF adopting a two-fold interpenetrated srs-c topology by reticulating dianionic hexacoordinate [SiO6] 2nodes with 3connected triphenylene building blocks. 38 In principle, the employment of new building units, such as 6-connected monomer with D3h geometry, can establish novel architectures in 3D COFs; however, its realization has remained an enormous challenge. Herein, we for the first time reported a 3D triptycene-based COF with large pores and stp topology. This novel COF, termed JUC-564 (JUC = Jilin University China), was constructed from a stereoscopic 6-connected triptycene-based building unit, 2,3,6,7,14,15-hexa(4′-formylphenyl)triptycene (HFPTP). As a result, JUC-564 showed high surface area (> 3300 m 2 g -1 ), the largest pore (43 ) among 3D COFs, and the lowest density among all crystalline materials (0.108 g cm -3 ). Moreover, due to the presence of large pores, JUC-564 showed a favorable adsorption of a large protein with suitable dimensions. Structural identification is one of major roadblocks for developing 3D frameworks with new topologies. Different from other crystalline porous materials, such as aluminosilicate molecular sieves 39 and metal-organic frameworks (MOFs), 40,41 single crystals are not common in COFs and their crystal structures are mostly obtained through powder X-ray diffraction (PXRD) patterns along with structural simulation. Usually, more than one possible topology is available for combinations of multiple building block geometries. After investigating RCSR database carefully, we fortunately found that only definite stp topology is available for [6 + 4] (3D-D3h + 2D-D2h) nets (Scheme 1), facilitating the structural determination of the target products. 42 To implement this strategy, we firstly designed a 6-connected 3D-D3h building block, HFPTP, based on a triptycene moiety with a link angle of 60º (Scheme 1a). Condensation of HFPTP and a synergistic 4-connected 2D-D2h monomer with a link angle of 120º (1,3,6,8-tetra(4-aminophenyl)pyrene, TAPPy, Scheme 1b) leads to an expanded [6 + 4] connected network (JUC-564, Scheme 1c and 1d). To the best of our knowledge, JUC-564 represents the first COF with a 6-connected 3D-D3h building block and a stp net. The synthesis of JUC-564 was carried out through traditional solvothermal approach by suspending HFPTP and TAPPy in a mixed solvent of mesitylene and dioxane with the presence of 6 M acetic acid followed by heating at 120 ºC for 3 days. Complementary methods have been employed for detailed structural determination and characterization. Scanning electron microscopy (SEM, Figure S1) and transmission electron microscopy (TEM, Figure S2) images revealed isometric microcrystals. Fourier transform infrared (FT-IR) spectrum exhibited a new adsorption corresponding to the characteristic of the C=N bond at 1628 cm −1 . The concomitant reducing of the C=O stretching (1700 cm −1 for HFTPT) and N-H stretching (3312 cm −1 for TATPy) confirmed the transformation of aldehyde and amine groups (Figure S3). The solid-state 13 C cross-polarization magicangle-spinning (CP/MAS) NMR spectroscopy further verified the presence of imine groups by the peak at 157 ppm (Figure S4). High thermal stability (∼450 °C) was observed by thermogravimetric analysis (TGA, Figure S5). The crystal structure was resolved by PXRD measurements in conjunction with structural simulations (Figure 1). After a geometrical energy minimization of JUC-564 by the Materials Studio software package on the basis of stp net, 43 the simulated PXRD pattern was in good agreement with the experimental one. Furthermore, the full profile pattern matching (Pawley) refinement was conducted based on experimental peaks at 1.93, 3.34, 3.87, 5.47, 6.96, and 9.46° corresponding to (100), ( 110 2). Notably, benefiting from its highly void framework and light constitutional elements, JUC-564 has a calculated density of 0.108 g cm -3 , which is the lowest reported for any crystalline material known to date, such as MOFs (0.22 g cm -3 for MOF-200, 44 0.195 g cm -3 for IRMOF-74-XI, 45 and 0.124 g cm -3 for NU-1301 46 ) and COFs (0.19 g cm -3 for JUC-518, 20 0.17 g cm -3 for COF-108, 27 and 0.13 g cm -3 for DBA-3D-COF 1 47 ). To investigate the porosity of JUC-564, gas sorption study of N2 was conducted at 77 K. As shown in Figure 3a, JUC-564 exhibited typical reversible type IV isotherms, which is one of the main characteristics of mesoporous materials. The surface area was calculated to be 3383 m 2 g -1 using the Brunauer-Emmett-Teller (BET) model (Figure S6). Pore size distribution calculated by nonlocal density functional theory (NLDFT) illustrated two kinds of pores with sizes of 15 and 41 (Figure 3b), which are in good agreement with those of the proposed structure (14 and 43 ). Remarkably, the largest pore size of JUC-564 (43 ) is far superior to that of other reported 3D COFs (Table S1), such as 13.5 for COF-102, 27 13.6 for DL-COF-1, 33 15.4 for JUC-518, 20 and 28 for DBA-3D-COF 1. 47 Furthermore, its BET surface area (3383 m 2 g −1 ) is much higher than that of other 3D imine-based COFs (Figure 4 and Table S2), such as 1360 m 2 g −1 for COF-300, 48 1513 m 2 g −1 for JUC-508, 15 2020 m 2 g −1 for LZU-111, 31 and 3023 m 2 g −1 for JUC-552. To further define the structure and large channels of JUC-564, incorporation of large biomolecules with suitable dimensions as probes was explored (Figures S7-12). The uptake ability of JUC-564 for myoglobin (Mb, about 21 × 35 × 44 ) 49 was confirmed by UV-vis spectrum, which proves the existence of the wide channel (43 ) in JUC-564. For comparison, no observable adsorption of Mb in the microporous COF-320 took place due to its smaller pore size (~12 ). In summary, we have developed a large-pore 3D COF with novel stp topology utilizing a rare 6-connected D3h node based on triptycene. JUC-564 exhibited interconnected channel systems with high surface areas (3383 m 2 g -1 ), ultra-large channels (up to 43 ), record-breaking low density (0.108 g cm -3 ), and positive uptake of a large protein molecule. This work not only opens a door to enrich 3D structures of COFs but also promotes new applications of 3D COFs in adsorption of large biological molecules.

chemsum

{"title": "Three-Dimensional Large-Pore Covalent Organic Framework with stp Topology", "journal": "ChemRxiv"}

heterocycle-derived_β-s-enals_as_bifunctional_linchpins_for_the_catalytic_synthesis_of_saturated_het

## Abstract: We demonstrate how heterocycle-derived β-S-enals can be employed as bifunctional substrates in a cascade of two rhodium-catalysed C-C bond forming reactions to deliver substituted heterocyclic products. A single rhodium-catalyst, generated in situ from a commercial salt and ligand combination, is used to promote both an initial alkene or alkyne hydroacylation reaction, and then a Suzuki-type cross-coupling, resulting in a three-component assembly of the targeted heterocycles. Substrates based on N-, Oand S-heterocycles are included, as are a range of alkenes, alkynes and boronic acid derivatives. Scheme 1 Saturated and partially saturated heterocycles in pharmaceuticals and natural products. Scheme 2 Heterocycle-derived β-S-enals as building blocks towards saturated heterocycles. † Electronic supplementary information (ESI) available: Experimental details and supporting characterisation data. See Due to the favourable physiochemical properties often associated with their incorporation into candidate structures, saturated, or partially saturated, heterocycles are becoming increasingly targeted in drug discovery programs. 1 Their presence in biologically active molecules has significant precedent, and Scheme 1 shows several examples of N-, O-and S-heterocycles embedded in pharmaceuticals and natural products used in a variety of applications. 2 The N-based congeners such as pyrrolidines, piperidines, and tropanes are the most commonly encountered structures. 3 In order to access saturated and partially saturated heterocycles decorated with a variety of substituents, we conceived an approach based on a common class of bifunctional building blocks that could be elaborated using cascade catalytic reactions. The key building blocks that we settled on were heterocycle-derived β-S-enals (1, Scheme 2). Variants of 1 featuring N, O, and S-atoms are all accessible from the parent ketones using established methods. 4 With the key building blocks available we speculated that a single rhodium-catalyst could mediate an initial alkene or alkyne hydroacylation reaction, 5 and then a Suzuki-type cross-coupling, to convert enals 1 into difunctionalised products 2 in a single step, joining together three separate components. A variety of methods could then be used to convert enones 2 into the fully saturated heterocycles. The design of the β-S-enals allows the S-atom to function as the directing atom for the initial chelation-controlled hydroacylation reaction, and then for the O-atom of the resul-tant enone to direct the Suzuki-type coupling. The use of aldehydes with β-S-directing groups in hydroacylation reactions is well established. In addition, our laboratory, 10 and others, 11 has recently reported on the required aryl methyl sulfide Suzuki chemistry, including a cascade reaction on a benzenederived substrate. We began by evaluating the basic reaction sequence using pyran-derived enal 1a, tert-butyl acetylene and p-tolyl boronic acid as the reaction partners (Scheme 3). Earlier precedent, 10,12 and initial investigations, 13 suggested that a Rh(I) catalyst incorporating the small-bite-angle bis-phosphine ligand dcpm should be able to mediate both of the key C-C bond forming reactions. When a catalyst of this type was generated in situ and applied to the targeted transformation, good yields of enone 2a were obtained. Achieving a short reaction time for the initial hydroacylation reaction was key to obtaining a high yield for the final product. Accordingly, warming to 55 °C resulted in the hydroacylation reaction being complete in only 5 minutes, with the following Suzuki-type coupling then requiring 5 hours. It was pleasing to note that the β-S-enal substrate appeared to allow significantly faster reactions than the previously explored fully aromatic system. With a catalyst and appropriate reaction conditions for the cascade process in hand, we next explored the scope of the alkynes and alkenes that could be employed in the hydroacylation step. Given the importance of N-heterocycles in medicinal chemistry, we chose the tetrahydropyridine substrate, 1b, in combination with tolyl boronic acid, as a suitable platform to evaluate the chemistry (Scheme 4). As can be seen the use of tert-butyl acetylene was successful, delivering the three-component product, enone 2b, in high yield. A terminal alkyne substituted with a benzyl ether (2c), and an internal alkyne (2d) were both successful substrates. Less sterically demanding alkyne substrates resulted in mixtures of linear and branched regioisomers in the hydroacylation step, resulting in lower overall yields. For example, phenyl acetylene (2e) and hex-5ynenitrile (2f ) delivered 4 : 1 and 5 : 1 mixtures of linear : branched isomers, respectively. Terminal alkenes could be employed in the desired cascade process, however, their lower reactivity in the hydroacylation step, relative to alkynes, necessitated the use of five equivalents to achieve suitably fast reactions. Using these conditions with octene delivered the desired three-component coupled product (2g) in 84% yield. Terminal alkenes substituted with bromo (2h), phenyl (2i) and free hydroxyl groups (2j) were also successfully employed. Disubstituted alkenes were unreactive in the described system. 14 Tetrahydropyridine enal 1b was then used to evaluate the scope of the boronic acid coupling partner (Scheme 5). Initially employing tert-butyl acetylene as the hydroacylation coupling partner, a range of electronically varied aryl boronic acids could be readily employed (2k-2p), including halogen substituents. Although substitution at the meta-position was possible (2n), attempts to employ ortho-substituted boronic acids were unsuccessful. A heterocyclic boronic acid, in the form of 3-thienyl, and also an alkenyl boronic acid performed well (2q and 2r). Using octene as the hydroacylation coupling partner allowed a similar range of boronic acids to be successfully incorporated into the cascade process (2s-2x). One of the goals of the present chemistry was to show that a variety of different heterocycles could be accessed using a single strategy. Accordingly, in Scheme 6 we demonstrate the successful use of N-, O-, and S-based heterocyclic building Scheme 3 Establishing reaction conditions for the hydroacylation-Suzuki cascade sequence to prepare dihydropyran 2a. Scheme 4 Substrate scope of the alkyne and alkene component in the cascade synthesis of tetrahydropyridines 2. Reaction conditions: 1b (1.0 equiv.), alkyne (1.5 equiv.) or alkene (5.0 equiv.), [Rh(nbd) 2 ]BF 4 (5 mol%), dcpm (5 mol%), acetone, 55 °C, 5-10 min; then p-tolyl boronic acid (1.5 equiv.), Ag 2 CO 3 (1.0 equiv.), 55 °C, 5 h. Isolated yields of the major isomer. a Measured by 1 H NMR spectrometry on the crude reaction mixture. blocks. For O-based heterocycles, a dihydropyran and a 2Hchromene-based substrate were both combined successfully with alkene and alkyne coupling partners and aryl boronic acids (2a, 3a-3d). S-heteocycles were represented by a 2H-thiochromene-based substrate, which was employed without incident (3e, 3f ). Unfortunately, it was not possible to prepare the dihydrothiopyran substrate due to stability issues. Finally, in addition to tetrahydropyridine substrate 1b already described, we were able to prepare and exploit a tropane-based substrate (1e), allowing access to alkyne and alkene coupled products in good yields (3g and 3h). All of the scoping experiments described in Schemes 4-6, were performed on a relatively small scale (0.2 mmol of enal), and as such 5 mol% of catalyst was employed due to ease of use. However, for larger preparative scale reactions, it was possible to lower the catalyst loading. Scheme 7 shows the use of tetrahydropyridine substrate 1b, and tropane-derived sub-strate 1e, used in alkene hydroacylation initiated cascades, employing just 3 mol% of catalyst, to deliver gram scale quantities of coupled products (2g and 3i) in excellent yields. Finally, as an illustration of synthetic potential of the enone products obtained from the developed cascade processes, we have shown that tetrahydropyridine-derived product 2g undergoes high-yielding detosylation and alkene reduction in a single-step, providing piperidine 4 (Scheme 8). This one-step Scheme 5 Substrate scope of the boronic acid component in the cascade synthesis of tetrahydropyridines 2. Reaction conditions: 1b (1.0 equiv.), alkyne (1.5 equiv.) or alkene (5.0 equiv.), [Rh(nbd) 2 ]BF 4 (5 mol%), dcpm (5 mol%), acetone, 55 °C, 5-10 min; then boronic acid (1.5 equiv.), Ag 2 CO 3 (1.0 equiv.), 55 °C, 5 h. Isolated yields. Scheme 6 Variation of the heterocyclic enal component in the preparation of di-coupled products 3. Reaction conditions: 1 (1.0 equiv.), alkyne (1.5 equiv.) or alkene (5.0 equiv.), [Rh(nbd) 2 ]BF 4 (5 mol%), dcpm (5 mol%), acetone, 55 °C, 5-10 min; then boronic acid (1.5 equiv.), Ag 2 CO 3 (1.0 equiv.), 55 °C, 5 h. Isolated yields. Scheme 7 Preparative scale synthesis of products 2g and 3i. transformation was achieved using magnesium metal in methanol under sonication conditons. 15 In conclusion, we have shown that heterocycle-derived β-Senals are efficient substrates for rhodium-catalysed hydroacylation-Suzuki type coupling cascade processes. Both alkyne and alkene hydroacylation reactions can be used as the initial C-C bond-forming event, and a variety of boronic acids can be employed as substrates in the Suzuki-type coupling. The products are obtained in good to excellent yields, and show potential as precursors to access biologically relevant compounds.

chemsum

{"title": "Heterocycle-derived \u03b2-S-enals as bifunctional linchpins for the catalytic synthesis of saturated heterocycles", "journal": "Royal Society of Chemistry (RSC)"}

the_extraordinary_richness_of_the_reaction_between_diazomethane_and_tetracyanoethylene:_can_computat

## Abstract: In the quest of the structure of the intermediate between D 1 -and D 2 -pyrazolines, the reactivity of these molecules tetrasubstituted by cyano groups in adjacent positions (3,3,4,4 or 4,4,5,5) has been explored in their neutral and protonated forms. Many reactions reported in the literature for pyrazolines have been studied and quantified (energies and transition states). Thirty-three structures of pyrazolines, their open-ring counterparts and their complexes are described. Acid-base equilibria, rotations, electrocyclic reactions and sigmatropic transpositions are reported. ## Introduction The reaction of diazomethane 2 with tetracyanoethylene 1, both very common compounds, has only been studied two times. In 1962, Bastu ´s and Castells reported the reactions of Fig. 1 (their numbering of formulae is different). 1 They indicated that 3 can be explosive and this probably prevented other authors from repeating its preparation. D 1 -pyrazoline 3 (4,5-dihydro-5H-pyrazole-3,3,4,4-tetracarbonitrile) was isolated and it spontaneously evolved nitrogen to yield 1,1,2,2-tetracyanocyclopropane 4 that was already known having been prepared by other methods. 2 Compound 3 was washed with benzene to eliminate all traces of 1 and was slowly dissolved in dry ether to yield a compound to which structure 5 (2,2,3,3-tetracyano-1,5-diaza-bicyclo[2.1.0]pentane) was assigned. Compound 5 when treated for about 90 min with a 5% solution of 1 in dry ether afforded D 2 -pyrazoline 6. The isomerization of 6 to 5 was performed using wet ether or dry ether containing traces of hydrogen chloride. Both substances can be kept for several weeks without alteration. According to Bastu ´s and Castells, the 6 to 5 isomerization involves the pyrazolinium cation 6bH + . 1 The role of TCNE (1) in the 5 -6 isomerization was assigned to a 1 : 5 complex. Huisgen et al. repeated the reaction. 3,4 They cited Banu ´s and Castells but they isolated only 4 and 6. Leaving aside 4 (also 6bH + was not characterized), we have summarized in Table 1 all the available information on the compounds in Fig. 1. Calatroni and Gandolfi reported a series of reactions that are related to the work of Banu ´s and Castells (Fig. 2). 5 D 1 -pyrazoline I reacted in two ways with TCNE (1) to afford a charge-transfer complex II and adduct III that was not isolated nor identified. They assumed that the reaction I " III is fast and reversible. Besides, TCNE promotes the isomerization I -IV in agreement with Castells' results. A reaction product V was postulated corresponding to the reaction of IV with TCNE. Calatroni and Gandolfi indicated in their paper that they wanted to determine the crystal structure of II (yellow crystals) but they probably failed because no structure like II was reported in the CSD. 6 The only one that bears resemblance to II is FEJDUT (VI), see Fig. We decided to study theoretically structures 3, 5 and 6 and their corresponding protonated cations (Fig. 4) as well as some concerted reactions related to Woodward-Hoffmann rules. ## Computational details The geometries of pyrazolines (Pz) and pyrazolinium cations (PzH + ) have been fully optimized using the functional B3LYP 11 and the 6-311++G(d,p) basis set 12 in the spin restricted formalism as implemented in the Gaussian 16 package (the coordinates of all the optimized geometries are gathered in the ESI †). 13 The minimum energy and transition state structures of all compounds were characterized using frequency analysis. The solvent effects have been evaluated by re-optimizing the structures at the B3LYP/6-311++G(d,p) level and using the self-consistency reaction field (SCRF) method 14 based on the polarized continuum model (PCM) of Tomasi and co-workers 15 in ethanol, diethylether and benzene as solvents using the standard parameters provided by the Gaussian-09 program. For the infrared spectra, the frequencies have been scaled by a factor of 0.9679. 16 Absolute chemical shieldings have been calculated with the GIAO approximation 17 and then transformed into chemical shifts using empirical equations. 18 The static intrinsic reaction coordinates (IRCs) 19 were analyzed in two cases. pyrazoles. 20 For the reaction 1 + 2 -3, D 1 -pyrazoline 3 lies 17.1 kJ mol 1 higher than the potential surface minimum, the D 2 -pyrazoline 6 (Table 2), and the barrier with regard to 3 is 73.8 kJ mol 1 . For the reaction between diazomethane and different olefins, Ess and Houk calculated barriers between 57 and 70 kJ mol 1 . 21 Calculated IR, the NMR properties of compound 5 are reported in Table 3. ## Results and discussion This compound has two conformations depending on the position of the NH bond, towards or out of the ring (Fig. 5). Compounds 5 lie 165.8 (5a) and 151.0 (5b) kJ mol 1 over the potential surface minimum and D 2 -pyrazoline 6 (Table 2). Calculated IR and NMR properties of compound 3 are reported in Table 3. D 2 -pyrazoline 6 is the most stable of the three isomers (Table 2). The 3 -6 tautomerization involves a 1,3 CH to NH prototropy. A direct proton transfer has a high barrier that in similar fivemembered rings prevents this mechanism, while the following one involves assistance by solvent molecules (one or two) such as water or alcohols with a low barrier. 22 The calculate TS between 3 and 7 is 289.7 kJ mol 1 and between 7 and 8, it is 277.8 kJ mol 1 . The experimental IR spectrum (KBr) and the calculated bands (gas phase) agree exceptionally well assuming that there is an intercept. The corresponding linear regression line is: Exp. = (126 AE 37) + (0.93 AE 0.1) Calc., n = 7, R 2 = 0.999, RMS residual = 32 cm 1 (1) We have removed Castells' bands at 1592 and 1595 cm 1 assigned to NH bending because according to eqn (1) they should appear at 1269 and 1417 cm 1 . On the other hand, if the 1592 cm 1 band is assigned to NQNQC of 7, the agreement is good (fitted value 1544 cm 1 ). In conclusion, on IR grounds, structure 5b should be rejected while structure 7 is acceptable. The available 13 C NMR data of compound 6 3 agree with the calculated values according to the linear regression, Exp. = (173.5 AE 3.4) (0.91 AE 0.04) Calc., n = 7, R 2 = 0.991, RMS residual = 2.9 ppm (2) The most different signals of the three compounds are the 15 N chemical shifts of the ring nitrogen atoms: +91.3 and +105.4 ppm for 3; 236.8 and 29.5 for 6; 285.8 and 297.0 for 5b. ## 3.4. A first attempt to find a more stable structure for intermediate 5 We have considered that the intermediate instead of being diaziridine 5b could be zwitterion 7 (Fig. 6). Compound 7 is much more stable than 5b, 88.3 vs. 151.0 kJ mol 1 , but still too high for allowing the 6 -5 backward reaction. Solvent effects (amongst them those used by Castells 1 ) calculated using the PCM model decrease the difference by a small amount: benzene, 79.7; diethylether, 75,4; and ethanol, 70.4 kJ mol 1 . Calculation of the IR spectrum of compound 7 leads to scaled bands at 3419 cm 1 (nNH) and 1523 cm 1 [(NQNQC) AE ]. Using eqn (1), these values in KBr should be 3319 and 1520 cm 1 . Castells et al. did not report a CQN band but 3319 cm 1 is consistent with the experimental value of 3295 cm 1 . ## Acid-base equilibria, rotations, electrocyclic reactions and sigmatropic transpositions Two reviews provide an overview of the electrocyclic reactions and sigmatropic reactions involving pyrazolines. 23,24 We have already reported the 1 + 2 -3 reaction with a TS of 73.8 kJ mol 1 with regard to 3 (IRC, Fig. 7). Other reactions that we have studied theoretically are reported in Fig. 8 and 11 together with acidbase equilibria (protonation) and with rotations about single bonds. Note that the loss of dinitrogen to afford 4 is a very exergonic reaction. Regarding Fig. 8, as we have already commented, the proposed structure of intermediate 7 lies 88.3 kJ mol 1 higher than 6 (71.2 kJ mol 1 above 3). A 1,2 (C to N) transfer of hydrogen has a barrier of 289.7 kJ mol 1 ; the subsequent 1,2 (C to N) transfer of hydrogen to afford 6 has a barrier of 289.7 kJ mol 1 . These very high barriers do not correspond to real pathways because proton transfers assisted by solvent molecules have much lower barriers (see previous discussion). Ring opening of 7 to 8a should occur thermally in a disrotatory way. 23,24 However, the geometry of 8a (Fig. 9) allows only a conrotatory mechanism with a barrier of 71.9 kJ mol 1 with regard to 7. A rotation about the single NC bond affords the more stable 8b while prototropy results in the less stable azine 8c. Note that according to the Woodward-Hoffmann rules, 8-10 a conrotatory mechanism is thermally forbidden. We then considered that the complex 1 : 5 (Fig. 1) could be the cycloaddition product of 7 (an azomethine imine) with 1 but the resulting 9 is a bicyclic compound very destabilized by the eight cyano groups (Fig. 8). D 2 -Pyrazoline (4,5-dihydro-1H-pyrazole-4,4,5,5-tetracarbonitrile) 6 can result from the reaction of tetracyanoethylene 1 with diazen-1-ium-1-ylidenemethanamide (10), a 1,3-dipole with octet stabilization (a nitrile imine). The cycloreversion barrier of 166 kJ mol 1 is much higher than that of diazomethane. Isomerization 6 -12a [(2,2,3,3)-tetracyanocyclopropyldiazene, 67.8 kJ mol 1 ] corresponds to a -sigmatropic transposition that has been reported in other compounds. 23 In particular, Rosenkranz and Schmid described the photochemical transformation of 5-phenyl-D 2 -pyrazolines into compounds similar to 12a (1-methylazo-2-phenyl-cyclopropanes); on thermal treatment, these compounds are reconverted into the corresponding pyrazolines. Compound 12b, an isomer of 12a, also possible from a similar mechanism, lies much higher in energy. Loss of HCN from 6 affords 15. According to Rodrı ´guez Mora ´n, when they carried out the reaction of 1 + 2 to afford 6, 3 release of hydrogen cyanide was observed. 27 The last reaction we have studied is the cycloaddition of diazomethane 2 on D 2 -pyrazoline 6. It is known that D 2 -pyrazolines react with 2 to afford 1,2-diazabicyclo[3.1.0]hexanes, related to 14. 28 These compounds should result from the loss of dinitrogen of either 13a or 13b. According to the literature, the cycloaddition of diazomethane on imines (or Schiff bases) results in 1,2,3-triazolines related to 13a. 29 In our case, the reverse addition leads to a more stable compound, 1,3,4-triazoline 13b instead of 1,2,3-triazoline 13a; this is probably related to the fourth N atom, i.e. our compounds are hydrazones, not imines. The loss of dinitrogen to form 14 is strongly favored (151.4 kJ mol 1 ). We then decided to carry out calculations parallel to those of Calatroni and Gandolfi. 5 They are reported in Fig. 10. Compound 16 is the complex of 1 and 3 (compared with the 1 : 5 complex of Fig. 1 and with complex II of Fig. 3); it is located 7.1 kJ mol 1 above 6 and at 10.0 kJ mol 1 from 3. The zwitterion 17 (compare with III of Fig. 3) is not stable and reverts to 16 both in the gas phase and in ethanol (PCM); this sheds doubt on the hypothetical III structure. 5 Structure 19 lies 66.0 kJ mol 1 above 6, also an indication that structure V in Fig. 3 was probably never formed. Charge-transfer complex 18a is not stable and evolves to the hydrogen-bonded complex 18b; it is located at 20.1 kJ mol 1 from 6. The stabilization results from the hydrogen-bond and from a CN/CN stacking between both molecules. It is also interesting to study the reactivity of the conjugated acids of structures of Fig. 8 because Castells et al. indicated that 5/6 equilibration was acid catalyzed (Fig. 11). 1 Usually, D 2 -pyrazolines protonate on N1 (type a) 30 but probably due to the presence of the four cyano groups, in this case, the protonation takes place preferably on N2 (6bH + ), which corresponds to the hypothesis of Castells et al.. However, the most important finding is that 5H + (even the most stable 5bH + , Table 2, 154.3 kJ mol 1 ) has an energy that makes it impossible to isomerize 6H + into 5bH + . The cations have relative energies similar to those of the neutral molecules, 6 o 3 { 5. Although the synthesis of D 2 -pyrazolines protonated on N1, 6aH + , from 1 and 10H + , is unknown, both the stability and the barrier made it a feasible possibility. One of the reactions of D 2 -pyrazolines protonated on N1 is to open into compounds related to 20b; this has been proven experimentally. 30b However, in the present case, the optimization led to a quaternary azetinium (2,3-dihydroazetium) cation 20a, still too high in energy. The loss of HCN from 6bH + generates a protonated isopyrazole 15H + (1H-isopyrazolium) that is slightly less stable. Protonation of the cyano groups leads to cations 6cH + (CN at position 4) and 6dH + (CN at position 5) of similar energies to 6aH + , i.e. the CN groups are similar to the amino group of a pyrazoline. This is a surprising result because the PA of CH 3 CN is 787 kJ mol 1 and that of CH 3 NH 2 is 899 kJ mol 1 , 31 but the role of the four CN groups and the D 2 -pyrazoline structure may modify the proton affinity; it is known that cyano derivatives can be superbases. 32 From the most stable of these cations, loss of protonated hydrogen cyanide results in the formation of 15 with an energy of 59.9 kJ mol 1 . ## Conclusions Castells intermediate 5 (Fig. 1) 1 is characterized by IR in KBr by a stretching NH at 3295 cm 1 and a bending NH at 1592 cm 1 as well as by the absence of a CQN band although these last bands appear in D 2 -pyrazolines at 1555-1570, 33 1557, 34 1560, 35 1580-1592, 36 1600 37 and 1618-1622 cm 1 . 38 Therefore, the compound has a NH group, i.e., it is not a D 1 -pyrazoline, but the presence or absence of a CQN band is dubious. The reactivity of intermediate assumed to be 5 is that it never affords 3 but it is in equilibrium with 6. Light and TCNE isomerize 5 to 6; water, HCl and time isomerize 6 to 5. 1 If pure 6 can be transformed into 5, this latter must be an isomer not containing TCNE. Of all the compounds that we have studied, which is the best candidate? Our calculations of IR spectra and its high energy show that 5 can be definitely excluded. That the intermediate could be a salt was possible when HCl was used but not with wet ether. If, by an experimental error, TCNE still remained in the reaction medium, 16 is a good candidate due to its energy and absence of barrier. Finally, the best candidates because they are consistent with the IR data are 7, as already discussed, and 8b. The linear equation relating both variables is: The problem is that the available information is too scarce, there is not even the complete IR spectrum of 5 let alone the 1 H NMR data. This added to the non-reproducibility of the preparation of 5 3,27 made it impossible to ascertain its structure. Charge transfer complexes are colored (yellow/red) but no indication of the color is reported. 1 On the other hand, a plethora of structures and reactions surrounding the chemistry of tetracyanoethylene (1) has been explored covering different aspects of the chemistry of pyrazolines 39 that would prove useful in related studies because several of them were not experimentally known.

chemsum

{"title": "The extraordinary richness of the reaction between diazomethane and tetracyanoethylene: can computational calculations shed light on old papers?", "journal": "Royal Society of Chemistry (RSC)"}

reduction_in_water_pollution_in_yamuna_river_due_to_lockdown_under_covid-19_pandemic

## Abstract: The epidemic of Novel COVID-19 was reported in India in January 2020 and increased day by day due to the movement of people from abroad to India and then to the different parts of the country. The COVID-19 has been declared as pandemic because of its high transmission rate and coved more than 2010 countries of the world. Under this scenario when there is no medicine for its treatment, the only solution to this problem is to break the chain of transmission and restrict the count of infected people. To contain a coronavirus (COVID-19) outbreak, the Government of India announced the nationwide lockdown with effect from the midnight of 24 th March 2020 followed by the extension of the lockdown periods and presently it is in its 4 th phase. The various provisions were made under lockdown for closing the industries, transportation, etc. except the essential services. It has been very interesting to note that the behavioural changes in nature are highly positive and atmosphere, hydrosphere, and biosphere are rejuvenating and it gives an appearance that the earth is under lockdown for its repairing work. Under this natural recovery, we tried to look at the improvement in the water quality of the Yamuna River in Delhi, which has been one of the burst polluted rivers. To study this river, the concentrations of pH, EC, DO, BOD, and COD have been measured which showed a reduction by 1-10%, 33-66%, 51%, 45-90%, and 33-82% respectively during the lockdown phase in comparison to the pre-lockdown phase. The Nizamuddin Bridge, Okhla U/s, Najafgarh Drain and Shahdara Drain were the major hotspots responsible for the deterioration of the water quality of Yamuna River while passing by Delhi region. Five major locations of Yamuna River have been analysed in this paper that showed a very impressive recovery of the water quality during the lockdown phase as compared to the pre-lockdown status of water quality. ## Introduction The origin of the deadly pandemics coronavirus (COVID-19) has been in December 2019 from the City of Wuhan, China (Travaglio et al., 2020;Raibhandari et al., 2020;Chauhan and Singh, 2020) and spread to the almost entire globe. The source of COVID-19 is reported from the novel coronavirus (SARS-CoV-2) and would have been produced from other mammals (Travaglio et al., 2020). The World Health Organisation (WHO) in his report updated on 23 May 2020 at 05:30 GMT said that there are confirmed cases of infection by COVID-19 to 5,206,614 people while 337,736 has lost their life from 216 countries (WHO Report, 2020 dated 24 May 2020). Recognizing the rate of spread of this virus with the personal contacts, various countries have imposed complete lockdown in order to maintain forced social distancing and break the chain of the spread of coronavirus. Still if there are some urgent requirements of movement of people, they were asked to under quarantine for 14 days considering the appearance of the symptoms that take about 14 days. The Govt. of India, taking note of the activities adopted by the COVID-19 affected countries, first requested the countrymen to be at home for the entire day and Prime Minister of India gave this a name as Janata (People's) Curfew, which was observed on 22 nd March 2020. On this day of Janta Curfew all flights, trains, bus services, industrial and commercial activities were closed. After its success, an absolute lockdown was imposed on 25 th March 2020 for 21 days to break the chain of COVID-19 (Long, 2020). Further, the lockdown has been extended in phases 2, 3, and 4 till May 31, 2020 to control the spread of infection through a complete halt on the movement. The total lockdown has elevated pandemonium among people but helped in reducing the pace of spreading the virus among society. However, under this lockdown period, Nature started to respond very positively and started giving several signals of improvement to natural parameters of the atmosphere, hydrosphere, and biosphere. It appears that the earth is rejuvenating under the lockdown period and it's a closure for the repairing of earth. With the understanding of this natural recovery, we tried to look at the water quality status and improvement, if any, for the Yamuna River in Delhi, which has been famous for its high pollution level in Delhi. The concentrations of pH, EC, DO, BOD and COD have been measured at various hot spots for the pollutions on the bank of river Yamuna. The water quality parameters were compared between the lockdown phase and the pre-lockdown phase. Five major locations of Yamuna River have been analysed in this paper that showed a very impressive recovery of the water quality during the lockdown phase as compared to the pre-lockdown status of water quality. This showed that Nature is flourishing during the coronavirus pandemic followed by the lockdown in the larger part of the world forcing the closure of the sources of anthropogenic pollution. Yamuna river is one of the highly polluted rivers in India, especially in Delhi (Upadhyay et al., 2011) where the recent observations reflected that the water pollution has reduced across the Yamuna River channel during the lockdown phase. Hence to quantify the status of water pollution in one of the highly polluted locations of Yamuna river, we have carried out an analysis of pH, Conductivity (EC), DO, BOD and COD at various locations of the Yamuna River, where complete lockdown has been imposed. ## Study Area and Methodology The Yamuna River is the second-largest and longest tributary of Ganga which enters Delhi at village Palla. It traverses 22 km to Wazirabad barrage where entire water is impounded to meet the drinking water requirement of Delhi. River Yamuna ceases to exist downstream of Wazirabad Barrage in most of the periods of the year and receives its flow from the Najafgarh drain at Wazirabad downstream. No major fresh water is allowed to flow downstream of Wazirabad barrage except during the monsoon season (Fig. 1). As the river traverses further downstream the flow is blocked by a barrage at Okhla. There are a total of 23 drains discharging wastewater in the river Yamuna. Out of 23, a total of 16 drains are discharging wastewater in river Yamuna between Wazirabad downstream to Okhla upstream and 04 drains meet the Yamuna in downstream of Okhla Barrage and 03 remaining drains discharge their wastewater further down at Agra Canal and Gurgaon Canal. There are 05 drains having interception and diversion provision of sewage to the nearby STPs for ensuring further treatment. During the year 2019, the total flow of wastewater was estimated as 3026.24 MLD and BOD load was estimated as 0.10-61.44 TPD (CPCB, 2020). The water quality data of various pollutants were collected from the Central Pollution Control Board (CPCB). The data were analyzed for the Yamuna at 5 monitoring stations i.e. Palla, Nizamuddin Bridge, Okhla (U/S), Najafgarh Drain and Shahdara Drain for the period from 1 st March to 7 April 2020 which have been studied in two phases as pre-lockdown phase (11-23 March, 2020) and lockdown phase (24 March -7 April, 2020). ## Effect of lockdown on Yamuna River water quality The nationwide lockdown has come in effect since the midnight of March 24 because of the COVID-19 pandemic. Under the lockdown, the major sectors responsible for water pollution like industries, power plants, construction activities, transportation, etc. were put on halt. The academic institutions and hospitality services were also adjourned. Under these circumstances, the improvement in water quality was noticed in the river system of the country. Delhi, which is the hub of air pollution and being counted as number one for most of the time has resulted in a noticeable improvement in the water quality of Yamuna during the lockdown period of the country. Scattered rains in Delhi on 27 th March and during March 28-29, 2020 further helped in improving the water quality of Yamuna River during the lockdown phase. This result has been substantiated while analyzing the data of water pollution and water quality before and after the imposition of lockdown. ## pH level in Yamuna River, Delhi The pH of the Yamuna River observed alkaline in nature which varies from 7.1 to 8.7 with a mean value of 7.6 during the pre-lockdown phase (Fig. 2) while it has been observed between 7.1-7.4 in Najafgarh and Shahdara drain during the pre-lockdown phase. However, pH varies from 7.1 to 7.8 with a mean value of 7.3 in the Yamuna during the lockdown phase. The highest pH (8.7) was recorded at village Palla (entry point of Yamuna in Delhi) and lowest (7.1) at Shahdara drain during the pre-lockdown phase. During the lockdown phase, a slight reduction in pH has been observed due to the reduction of industrial activities, the nonfunctioning of essential commercial units, and prevailing weather conditions. The maximum reduction (10%) of pH has been observed at Village Palla during the lockdown phase. The concentration of pH was also correlated with the primary water quality criteria for a bathing water and designated best usable water quality criteria of India (https://cpcb.nic.in/wqm/Primary_Water_Quality_Criteria.pdf). These exercises helped in understanding that the concentrations were greater than the threshold limit of pH (6.5-8.5) daily at the village Palla which is vulnerable to the health problem. During the pre-lockdown phase, the pH levels were lower than the threshold limit (6.5-8.5) except at village Palla while it became much lower during the lockdown phase at all locations. The pH drives most of the chemical and biological changes in water. It acts as the driving force in controlling species distributions in aquatic habitats. The varying pH values provides space to different species to flourish within however the optimum pH range is 6.5-8.0 for most of the aquatic organisms. The variability of pH outside this range physiologically put stress on numerous species and may affect decreased reproduction and growth, attack of disease, or even death. Hence beyond the optimum value of pH can adversely affect the biological diversity in water bodies. ## Conductivity level in Yamuna River, Delhi In the Yamuna, conductivity varies from 688 to 2485 µS/cm with a mean value of 1526 µS/cm during the pre-lockdown phase (Fig. 3) while it observed between 273-1657 µS/cm during the lockdown phase. The highest conductivity (2485 µS/cm) was recorded at Shadra Drain and lowest (688 µS/cm) at village Palla during the pre-lockdown phase. During the lockdown phase, a slight reduction in conductivity has been observed due to the reduction of industrial activities, the nonfunctioning of essential commercial units, and prevailing weather conditions. The maximum reduction (66%) of conductivity has been observed at the Nizamuddin bridge followed by village Palla (59%), Okhla U/s (43%), and Nazafgarh/Shadra drain (33%) during the lockdown phase (Table 1). Discharges into the streams are capable of changing the conductivity depending on their makeup. A failing sewage system raises the conductivity because of the higher presence of chloride, phosphate, and nitrate. It may be noted that 16 drains are discharging wastewater in river Yamuna which are influencing the conductivity of the Yamuna River. ## Dissolved Oxygen level in Yamuna River, Delhi Dissolved oxygen (DO) is one of the most important indicators of water quality on which the survival of aquatic life depends. When DO becomes too low, fish and other aquatic organisms cannot survive. The data for DO was not available at Nizamuddin Bridge and Okhla U/s location during the pre-lockdown phase while it was 17.01 mg/l at village Palla in the same period (Fig. 4). However, DO vary from 1.2 to 8.3 mg/l with a mean value of 3.9 mg/l in the Delhi region of Yamuna during the lockdown phase. During the lockdown phase, improvement in DO has been observed at both the Nizamuddin Bridge and Okhla U/s due to the reduction of industrial activities and rainfall in Delhi. It may be noted that DO was not detected at both Nizamuddin Bridge and Okhla U/s during the pre-lockdown phase due discharge of huge amount of industrial and domestic wastewater. The comparative analysis is given in Table 1. The concentration of DO was also correlated with the Primary Water Quality Criteria for bathing water and designated best use water quality criteria of India. The DO levels were lower than the threshold limit (5 mg/l) except at village Palla during both pre-lockdown and lockdown phase at all locations. Low DO affects most biological processes in water and responsible for lower biological diversity in water bodies. ## Biological Oxygen Demand level in Yamuna River, Delhi Biological Oxygen Demand (BOD) is one of the most important indicators of water quality. BOD directly affects the amount of dissolved oxygen in water bodies. The greater demand for BOD more rapidly depletes the oxygen in the water bodies making lesser availability of oxygen for higher forms of aquatic life. The consequences of the high BOD are similar to the effect of less oxygen availability putting aquatic life under stress, suffocation and could be lethal. The major sources of increase of BOD in the Yamuna river include dead plants and animals; animal manure; industrial/domestic effluents, wastewater treatment plants, failing septic systems; and urban stormwater runoff. BOD vary from 7.9 to 163 mg/l with a mean value of 66.58 mg/l during the pre-lockdown phase (Fig. 5) while it observed between 2-89 mg/l during the lockdown phase. The highest BOD (163 mg/l) was recorded at Shahdara Drain and lowest (7.9 mg/l) at village Palla during the prelockdown phase. However, improvement in BOD (i.e. the reduced demand) has been observed at all locations in the lockdown phase due to the reduction of industrial activities and prevailing weather conditions. The maximum reduction (90%) of the BOD level has been observed at the Nizamuddin Bridge during the lockdown phase followed by Okhla U/s (77%), village Palla (75%), Shahdara drain (45%) and Najafgarh drain (29%). The comparative analysis is given in Table 1. The concentration of BOD was also correlated with the primary water quality criteria for bathing water and designated best use water quality criteria of India. The BOD levels were much higher than the threshold limit (3 mg/l) at all locations during the pre-lockdown phase. A similar trend was also observed during the lockdown phase except for village Palla. Higher BOD affects most biological processes in water and can ultimately lead to reduced biological diversity in streams. ## Chemical Oxygen Demand level in Yamuna River, Delhi Chemical oxygen demand (COD) is an indicator of contamination that shows the amount of dissolved matter in water susceptible to being oxidized. COD is responsible for the reduction of DO in water bodies. Higher concentration of COD is responsible for quick deterioration of oxygen in water bodies and oxygen availability for higher forms of aquatic life. The major sources that increases the COD in the Yamuna River are industrial/domestic effluents, wastewater treatment plants, failing septic systems; and urban stormwater runoff. COD varies from 28 to 574 mg/l with a mean value of 211.6 mg/l during the pre-lockdown phase (Fig. 6) while it observed between 6 to 383 mg/l during the lockdown phase. The highest COD (574 mg/l) was recorded at Shahdara drain and lowest (28 mg/l) at village Palla during the prelockdown phase. However, improvement in COD has been observed at all locations in the lockdown phase due to the reduction of industrial activities, rainfall, and prevailing weather conditions. The maximum reduction (82%) of the COD level has been observed at the Nizamuddin Bridge during the lockdown phase followed by Okhla U/s (81%), village Palla (79%), Najafgarh drain (45%) and Shahdara drain (33%). The comparative analysis is given in Table 1. ## Major Pollution hotspots in the Yamuna During the pre-lockdown period at Nizamuddin Bridge, the results showed pH (7.3), EC (1369 μs/cm), BOD (57 mg/L), DO (not detected), and COD (90 mg/L) whereas in the lockdown period pH (7.2), EC (460 μs/cm), BOD (5.6 mg/L), DO (2.4 mg/L) and COD (16 mg/L) were observed and not complying to the primary water quality criteria for outdoor bathing w.r.t analyzed parameters of DO and BOD which can be attributed to the contribution from mainly 14 drains discharging both treated and untreated sewage, no industrial effluent discharges from the industrial areas or no other human activities such as bathing, throwing of worship materials or solid waste and freshwater discharges from U/s of river Yamuna. A similar trend was also at Okhla U/s where the analysis showed pH (7.2), EC (861 μs/cm), BOD (27 mg/L), DO (not detected), and COD (95 mg/L) during pre-lockdown phase whereas pH (7.1), EC (488 μs/cm), BOD (6.1 mg/L), DO (1.2 mg/L) and COD (18 mg/L) were observed during the lockdown phase and not complying to the primary water quality criteria for outdoor bathing w.r.t analyzed parameters such as DO and BOD which can be attributed to contribution only from two drains carrying both treated or untreated sewage, no industrial effluent discharges and there is a river Yamuna stretch of about 7.5 km (after Nizamuddin Bridge) and might be helping in selfpurification of river Yamuna. The Najafgarh Drain discharges 1938 MLD of wastewater into river Yamuna. During the pre-lockdown period at Najafgarh Drain, the analysis showed pH (7.3), SS (152 mg/L), BOD (78 mg/L), COD (271 mg/L) whereas the analysis showed pH (7.3), EC (1501 μs/cm), BOD (55 mg/L), and COD (150 mg/L) during the lockdown period. While at Shahdara Drain, the results showed pH (7.1), BOD (163 mg/L), COD (574 mg/L) during prelockdown period whereas pH (7.2), EC (1657 μs/cm), BOD (89 mg/L) and COD (303 mg/L) were observed in lockdown phase. The comparative analysis is given in Table 1. Betterment in the water parameters of the Yamuna in Delhi during the lockdown phase is due to no contribution of effluent from all the 23 sources. ## Conclusion During the lockdown period, there has been a general improvement in water quality in the Yamuna River as a result of the restrictions imposed during the lockdown and due to no contribution of effluent from all the 23 sources. The concentrations of pH, EC, DO, BOD and COD showed 1-10%, 33-66%, 51%, 45-90% and 33-82% of reduction in pH, EC, DO, BOD and COD concentrations, respectively during the lockdown phase, as compared to the pre-lockdown phase in Yamuna River. The Nizamuddin Bridge, Okhla U/s, Najafgarh Drain and Shahdara Drain are the major hotspots of effluent in the Yamuna River catchment area in Delhi. The Covid-19 lockdown situation in almost the entire world has shown the importance of nature in our day to day life and gave a true picture of the overexploitation of the natural resources and proved that we are responsible for the degradation of nature and putting risk to our wellbeing as well. This lockdown showed that the solution for natures' cleanliness lies in our hands goes through the path of preservation of natural resources and sustainable development. The cleanliness observed in the river Yamuna during the lockdown is much better than the several efforts and actions for Yamuna cleaning where a huge amount of the money was invested but the results were never at the satisfaction and status of the revival of Yamuna at the pre pollution level was not achieved. There are several issues due to the lockdown at the front of social and economic wellbeing which cannot be appreciated at all but some positive lessons related to nature gave us a way forward for restraining from the natural calamities if care for nature is established with honesty.

chemsum

{"title": "Reduction in Water Pollution in Yamuna River due to lockdown under COVID-19 Pandemic", "journal": "ChemRxiv"}

enhancement_of_hydrogen_evolution_reaction_kinetics_in_alkaline_media_by_fast_galvanic_displacement_

## Abstract: Energy-efficient hydrogen production is one of the key factors for advancing the hydrogen-based economy. Alkaline water electrolysis is the main route for the production of high-purity hydrogen, but further improvements of hydrogen evolution reaction (HER) catalysts are still needed. Industrial alkaline electrolysis relies on Ni-based catalysts, and here we describe a drastic improvement of HER activity of Ni in alkaline media using several model catalysts for HER obtained upon nickel surface modification in aqueous solution of rhodium salts, when a spontaneous deposition of rhodium takes place based on the chemical displacement reaction 3Ni + 2Rh 3+ = 3Ni 2+ + 2Rh. In the case of smooth Ni-poly electrodes, HER activity surpasses the activity of Pt-poly already after 30 s of exchange with Rh. SEM analysis showed that Rh is uniformly distributed, while surface roughness changes within 10%, agreeing with electrochemical measurements. Furthermore, XPS analysis has shown effective incorporation of Rh in the surface, while DFT calculations suggest that hydrogen binding is significantly weakened on the Rh-modified Ni surfaces. Such tuning of the hydrogen binding energy is seen as the main factor governing HER activity improvements. The same galvanic displacement protocols were employed for nickel foam electrodes and electrodeposited Ni on Ti mesh. In both cases, somewhat longer Rh exchange times are needed to obtain superior activities than for the smooth Ni surface, but up to 10 min. HER overpotential corresponding to −10 mA cm −2 for nickel foam and electrodeposited Ni electrodes, after modification with Rh, amounted to only −0.07 and −0.09 V, respectively. Thus, it is suggested that a fast spontaneous displacement of Ni with Rh could effectively boost HER in alkaline media with minor cost penalties compared to energy saving in the electrolysis process. ## Introduction Faced with the global energy crisis, we strive to reduce fossil fuel consumption and increase the usage of renewable energy sources . The electrochemical production of hydrogen via hydrogen evolution reaction (HER) is a promising way for obtaining clean and renewable energy . In alkaline solutions the HER mechanism involves water dissociation and subsequent hydrogen reduction and adsorption (H 2 O + e −  H ads + OH − ), in the Volmer step, which is followed by, either the Heyrovsky step (H ads + H 2 O + e −  H 2 + OH − ) or by the Tafel step (2H ads H 2 ) . Recent improvements in the cost and performance of electrochemical water splitting technologies point towards a more economically viable future for their application at an industrial scale . Currently, the two prevalent methods for obtaining hydrogen are proton-exchange membrane (PEM) electrolysis and alkaline water electrolysis (AWE) . The hydrogen production efficiency of PEM electrolyzers is higher but requires Pt-based catalysts . In the case of AWE, the key goal is high catalytic activity and stability under intermittent polarization in alkaline media . It is well known that the HER rate in alkaline media is inferior to that in acidic ones because of the additional water dissociation step , precedes the discharge step, but also due to the blockage of active sites by adsorbed hydroxyl ions, but also partly due to hydroxyl ion adsorption resulting in the blockage of active sites . On the other hand, HER in alkaline media leads to less environmental pollution and equipment corrosion . Nickel and nickel-based electrocatalysts have been thoroughly investigated because of nickel's excellent stability in alkaline media . However, it has been shown that the catalytic activity of Ni largely depends on its morphology and active surface area . Therefore, multiple design strategies have been developed to increase the mediocre activity of clean Ni. Alloying Pt with Ni gives rise to two effects that increase the overall activity of the material. The first is the -ligand effect‖, referring to the change in electronic properties of both metals . The other is the -lattice strain effect,‖ which represents the change in Pt-Pt interatomic distance, resulting in a shift of the Pt d-band center . A large number of non-PGM (PGMplatinum group metal) Ni alloys, such as binary and tertiary Ni-Mo alloys , NiCu , Raney Ni and NiCo , exhibit high activity. Surface modification of Pt with Ni(OH) 2 is a prime example of a bifunctional electrocatalyst for alkaline HER. As suggested, Ni(OH) 2 promotes water dissociation, and the hydrogen intermediates subsequently adsorb and combine on the Pt surface , relying on tailoring catalytically active interfaces. Another approach for improving Ni performance is enhancing interfacial processes related to fast removing adsorbed hydrogen, like interfacing with reduced graphene oxide, which stimulated Hads spillover and improved HER activity while maintaining excellent stability . Despite numerous approaches to boost the HER activity of Ni, which work well in practice, a fundamental understanding of the key interactions in catalytic systems is of utmost importance. It is well known that rhodium exhibits high catalytic activity towards both reduction and oxidation reactions . Although several Rh-containing materials have been used in electrocatalysis of HER , only a small fraction of these was tested in alkaline media . An example is a Ni 89 Rh 11 alloy which has been experimentally shown to have excellent activity , which can largely be attributed to the bifunctionality of the surface. Furthermore, it was discussed that the Ni/Rh phases play the roles of H-adsorption/desorption sites, while the Ni(OH) 2 and Rh 2 O 3 phases catalyze water dissociation, but no direct pieces of evidence were provided for such claims . Moreover, Ni-Rh catalysts were also used for other catalytic reactions , showing the versatility of this catalytic system. However, the downside of Rh-containing catalysts is the high price of Rh, caused by its scarcity. With this in mind, it is important to optimize the design of novel Rh-containing electrocatalysts, taking advantage of the high catalytic activity of Rh while maintaining its low concentrations. Finally, it is important to note that the application of Ni/Rh electrocatalysts is potentially highly favorable, since not only does Rh enhance the HER performance and the overall water splitting , but, most likely, there is an underlying synergistic effect of Ni and Rh which could go beyond HER catalysis. This study is aimed to investigate the effects of surface modification of Ni using galvanic displacement with Rh on the HER activity in alkaline media. Using the mentioned approach, only the surface of an electrocatalyst is modified with low amounts of Rh , which is highly beneficial for maintaining the low cost of such prepared catalysts while significantly increasing catalytic activity. The Rh-exchanged Ni electrodes, starting with smooth Ni surfaces to nickel foam and Ni electrodeposits, were investigated by cyclic voltammetry for different duration of Ni immersion in rhodium salt solutions. In addition, the effects of electrode oxidation pretreatment were investigated to discuss the HER mechanism on modified electrodes. Apart from electrochemical measurements, the effects of the Rh-exchange process on the surface properties of Ni electrodes were also characterized by scanning electron microscopy with energy dispersive X-ray analysis (SEM/EDX) and X-ray photoelectron spectroscopy (XPS). Moreover, Density Functional Theory (DFT) calculations were used to rationalize the enhancement of HER activity upon surface modifications with Rh. We note that this work is focused on the origins of HER activity enhancement on various Ni surfaces upon spontaneous displacement with Rh, while the stability of investigated catalysts is not addressed. ## Surface Rh exchange and electrochemical measurements Electrochemical measurements were done using IVIUM Vetex.One or Gamry Interface 1011e potentiostats in one compartment three-electrode electrochemical cell with Saturated Calomel Electrode (SCE) as a reference electrode and a 3×3 cm Ni foam (Goodfellow) as a counter electrode. As the electrolytic solution, 1 mol dm −3 KOH solution (Sigma Aldrich) prepared with ultrapure deionized water was used in all experiments. All the measurements were done at room temperature. In this work, potentials are referred to SCE, and to calculate overpotentials for HER, potentials are converted to the Reversible Hydrogen Electrode (RHE) scale as E RHE = E SCE + 0.244 V + 0.059 V × pH (pH = 13.45). Electrolyte resistance was corrected using hardware settings, but only up to 80 % of the resistance value, determined using single-point impedance measurement at −1 V vs. SCE. If a higher percentage of iR drop was corrected, the measurements were unstable. HER measurements were done using cyclic voltammetry at a sweep rate of 10 mV s −1 . Before the potential sweep, the electrode was held −1 V vs. SCE until current dropped below 1 μA cm −2 . No anodic excursions of the working electrode were allowed before the HER measurements unless explicitly stated otherwise. ## Working electrodes preparation The first set of measurements relates to smooth Ni-poly rotating disk electrodes (RDE). The Ni disk had a diameter of 3 mm and was inserted in a Teflon cylinder with a 10 mm diameter. The disk was polished to mirror-finish using alumina powder and then sonicated for 1 minute, washed in deionized water, diluted HCl, and again in deionized water. Then it was transferred to the electrolytic cell, and HER measurements were done. Ni/Rh displacement experiments were done by immersing Ni-poly RDE into 0.1 mol dm −3 RhCl 3 solution in 0.1 mol dm −3 HClO 4 for a given amount of time (for the measurements with Ni-poly RDE up to 45 s). After the exchange, the electrode was rinsed in deionized water, 1 mol dm −3 KOH solution, covered with the droplet of the same solution, and quickly transferred into the electrochemical cell. The transfer to the cell typically took under 30 s. During the HER measurements, the electrode was rotated at 1800 rpm to remove any bubbles formed on the surface, which could block the electrode surface. The second set of the exchange and HER measurements was done using commercial nickel foam (NF) and electrodeposited Ni on Ti mesh (Special Metals, India) (Ni-dep). Before HER and Ni/Rh displacement experiments, NF (5×5 mm was exposed to the electrolyte) was cleaned in HCl, acetone, and deionized water. Exchange with Rh was done in the same way as for the Ni-poly disk. Then, HER measurements were done as described above. Ni-dep was produced on Ti mesh, previously cleaned in HCl, acetone, and deionized water. Ni deposition was done on Ti mesh with dimensions of 3×3 mm, from the deposition bath containing 0.2 mol dm −3 H 3 BO 3 , 0.5 mol dm −3 NH 4 Cl, and 0.125 mol dm −3 NiSO 4 . The deposition was done in a 2-electrode electrochemical cell, using Ni foam as a counter electrode, under potentiostatic conditions (2.4 V) for 90 s. Under these conditions, the typical deposition current is around 50 mA. Once the deposition was done, Ni-dep electrodes were thoroughly washed with deionized water and transferred to the electrochemical cell or Rh-exchange solution, the same one as described above. We note that here we report currents normalized per geometric surface area. For the measurements with the Ni-poly disk, this is straightforward. For the measurements with NF, we used a geometrical cross-section multiplied by 2. The same is done for Ni-dep electrodes, but we excluded the surface of voids between the Ti wires forming the mesh. It is done because the used mesh is rather sparse, and using a finer mesh would give higher currents within the same geometrical area (3×3 mm). As a benchmark, we used a platinized Pt-poly disk (3 mm in diameter), prepared by potentiodynamic cycling of smooth Pt-poly disk in H 2 PtCl 6 solution, as described in ref. . Using H UPD peaks, the roughness factor (RF) of the used platinized Pt-poly disk was evaluated to 7.9. In addition, the activity of the used Pt-disk was also checked in acidic media (0.5 mol dm −3 H 2 SO 4 solution, Figure S1, Supplementary Information) using the same HER measurement protocol as in alkaline solution, confirming that a highly active Pt surface is obtained. ## Characterization Morphology analysis and chemical composition were probed using SEM-EDX with Phenom ProX Scanning Electron Microscope (Phenom, Netherlands). SEM characterization was done using an acceleration voltage of 10 kV, while the chemical composition was probed with the acceleration voltage of 15 kV. XPS was used to assess the chemical speciation on the samples surfaces after the exchange with Rh. Samples were analyzed using SPECS Systems with XP50M X-ray source for Focus 500 and PHOIBOS 100/150 analyzer. AlKα source (1486.74 eV) at a 12.5 kV and 32 mA was used for this study. Survey spectra (0-1000 eV binding energy) were recorded with a constant pass energy of 40 eV, step size 0.5 eV and dwell time of 0.2s in the FAT mode. High-resolution spectra of Ni 2p and Rh 3d peaks were recorded with a constant pass energy of 20 eV, step size of 0.1 eV and dwell time of 2s in the FAT mode. Spectra were obtained at a pressure of 9×10 −9 mbar. SPECS FG15/40 electron flood gun was used for charge neutralization to minimize the effects of charging at the samples. All the peak positions were referenced to C1s at 284.8 eV. Spectra were collected by SpecsLab data analysis software supplied by the manufacturer and analyzed with a commercial CasaXPS software package. ## DFT calculations The first-principle DFT calculations were performed using the Vienna ab initio simulation code (VASP) . The Generalized Gradient Approximation (GGA) in the parametrization by Perdew, Burk and Ernzerhof combined with the projector augmented wave (PAW) method was used . Cut-off energy of 600 eV and Gaussian smearing with a width of σ = 0.025 eV for the occupation of the electronic levels were used. A Monkhorst-Pack Γ-centered 14×14×1 k-point mesh was used. We modelled Ni(001) and Ni(111) surfaces using the corresponding p(2×2) cells of given surfaces, with 10 and 9 layers slabs, respectively. Rh was inserted in the surface layer (Rh surf ) or the sub-surface layer (Rh sub ), and the top 6 layers of a given surface were allowed to relax fully. Knowing that on Ni HER proceeds under the conditions of high surface coverage , the surfaces were completely saturated by H ads (1 monolayer of hydrogen, 4 H ads per cell), and the average hydrogen binding energy (E H,b ) was calculated as: where E SURF+H , E SURF , and 4E H stand for the total energy of the surface with a monolayer of hydrogen, the total energy of the clean surface and the total energy of an isolated hydrogen atom. To convert E H,b to the Gibbs free energy for H ads formation (ΔG H,ads ), the procedure of Nørskov et al. was used (ΔG ads (H) = E H,b + 1/2E H2 + 0.24 eV, E H2the total energy of an isolated H 2 molecule). ## Rh depositionsmooth Ni-poly surface The clean, polished Rh disk measurements showed rather poor HER activity with η 10 more negative than −0.4 V and the corresponding Tafel slope of −140 mV dec −1 (Figure 1). Even though the Ni-poly disk was polished to a mirror finish, the roughness factor was somewhat higher than 1. By plotting the currents measured at −0.85 vs. SCE as the function of the potential scan rate and using the value of specific capacitance of 20 μF cm −2 (for metallic surfaces ), the roughness factor was determined to be 2.8. According to the benchmarking by McCrory et al. , η 10 for the Ni surface with roughness factor 20±10 was approximately −0.25 V. For the exchange with Rh up to 20 s, we could not see any significant changes in recorded cyclic voltammograms (Figure S2, Supplementary Information), but obtained a significant increase of HER activity. However, for 30 s of exchange with Rh and more, cyclic voltammetry of Rh-exchanged Ni-poly disk shows H UPD regions characteristic for Rh and practically unchanged regions corresponding to Ni 2+ /Ni 3+ oxidation (Figure 1). For 15, 30, and 45 s of exchange with Rh, η 10 amounted to −168, −120, and −110 mV, without significant change of the Tafel slope . These results align with the unchanged surface roughness, evidenced by 3D surface reconstruction using SEM, and a low percentage of Rh on the surface for short Rh-exchange times, with a very uniform distribution. These issues will be discussed later on. We have used platinized Pt-poly as a reference to Rh-modified Ni-poly surfaces. In ref. η 10 for Pt was reported to amount (-100±20) mV for smooth Pt-poly surface (RF 6±2), and ~ −40 mV for Pt-poly with RF of 90±20. Our Pt disk has η 10 of -110 mV, which closely agrees with the results of McCrory et al. . Given that the RF values of Ni-poly do not change for more than 10% upon the exchange with Rh (Figures S2 and S3, Supplementary Information), and that the Rh-exchanged electrodes have measured HER current densities (normalized per geometric surface area) close to that of Pt (Figure 1), surface-specific activities of Rh-exchanged Ni-poly actually exceed the activities of Pt (Figure 2). A direct comparison with the results of Nguyen et al. , who investigated HER activity of NiRh 3 alloy nanosponges in alkaline media, can be made. With the catalyst loading of 169 μg cm −2 , the authors reported η 10 of −107 mV for NiRh 3 alloy and −119 mV for commercial Pt/C catalyst . The previous improvements of HER activity were achieved using similar approaches, like spontaneous deposition of Pd and Rh on smooth Pt-poly (RF 2) , where Rh-modified Pt-poly showed higher HER activity than Pd-modified Pt-poly, without significant change of the roughness factor (2.1 for Rh-modified surface, and 2.08 for Pd-modified surface). Described catalytic surfaces should also be compared to pure Rh, and recently Rh films on Ni foam and Ti mesh, prepared by aerosol-assisted chemical vapor deposition (AACVD) technique, were investigated as HER catalysts in alkaline media . It was found that Rh film on NF had η 10 of −127 mV, while the one prepared on Ni mesh had the corresponding overvoltage of −67 mV . As there are many studies of HER catalysts in alkaline media, it would be possible to extend the comparison with literature data to a large number of reports. However, here we refer to the review of Bouzek et al. and Rh 3+ and metallic Rh (from Ni 3d high-resolution spectra) . The results shown are for 10 s exchange with Rh, in which case neither electrochemistry nor SEM shows the formation of welldefined Rh islands, but it is obvious that there are certain parts of the surface containing Rh aggregates. However, after 30 s exchange with Rh, there are clear indications that Rh islands are formed, and the effect is more prominent for higher exchange times. For example, after 45 s of exchange with Rh, the characteristic H UPD region is clearly seen (Figure 1), which is the feature of the Rh phase . This observation unambiguously confirms that longer exchange times lead to the formation of metallic Rh islands on the Ni-poly surface. ## HER mechanism on Rh-modified Ni surfaces In order to rationalize enhanced HER activity of Rh-exchanged Ni surface, we turned to the analysis of H ads energetics on Rh-modified Ni surfaces. We studied Ni(111) and Ni(001) using DFT calculations, and to model modification by Rh, we added Rh in the surface or subsurface layer of these Ni(hkl), effectively modelling the surface of subsurface alloys. We find that surface incorporation of Rh is energetically more favored than subsurface inclusion of Rh on both Ni surfaces, in agreement with previous reports . In the case of Ni(111), surface incorporation of Rh is more favored by −0.20 eV, while this difference for Ni(001) is −0.11 eV. Considering that H ads energetics was shown to give the volcano-type relationship with HER activities in alkaline media , just like in acidic media , we used the methodology of Nørskov et al. to establish Tafel analysis performed on HER polarization curves of Rh-exchanged smooth Ni-poly shows Tafel slopes that are close but usually larger than −120 mV dec −1 (absolute values). Deviation from the theoretical value of −120 mV dec −1 can be ascribed to incomplete correction for the iR drop (see Section 2.1) and the problems with removing H 2 bubbles (despite electrode rotation during the measurements). Thus, although there it is quite inconvincible to determine the mechanism of HER solely based on the Tafel slope, there is enough reliable literature data to conclude that Heyrovski reaction is the rate-determining step (RDS) on Ni-poly surface, at least at higher HER overpotentials where we performed Tafel analysis. The same reference claims that, at lower HER overpotentials, the Tafel reaction is the RDS . However, more recent findings suggest that surface modifications by Ni-oxy-hydroxides enhance HER kinetics through H 2 O dissociation where H ads ends up on the metal surface. At the same time, OH ads is governed by the oxy-hydroxide phase (and ultimately released back to the solution) . In the original interpretation, this means that the rate of Volmer reaction is increased, allowing fast recombination of H ads and formation of molecular H 2 via the Tafel step. rate is increased, one can also expect that HER proceeds through with the Heyrovsky reaction as the RDS will be enhanced. Actually, the Ni surface binds H ads strongly, so there is no reason to expect that the Volmer reaction (formation of H ads ) will be RDS, in agreement with ref. . Nevertheless, the HER activity significantly decreases if the Rh-exchanged surface is exposed to the same oxidation protocol (Figure 4). Rh surface binds H ads weaker than Ni surface, but, still, ΔG ads (H) is negative (Figure 4). However, the Rh|Ni interface conversion to Rh|Ni-oxy-hydroxide interface hinders HER. As Rh-exchanged Ni surface binds H ads weaker than clean Ni surface, there is no reason to expect that Ni-oxy-hydroxide surface will bind OH ads weaker than on pure Ni surface, Brønsted−Evans−Polanyi relations for water dissociation suggest lower H 2 O dissociation rate. Thus, slower Volmer and Heyrovsky reactions compared to pure Ni surface. However, on pure Ni-poly Tafel reaction is the RDS at low overpotentials due to strong binding of H ads , but this is not the case for Rh-exchanged surface, and the surface is effectively cleaned from H ads by fast recombination/desorption and the formation of molecular H 2 . However, we also observed an additional effect of surface oxidation. ## Rh exchange on expanded Ni surfaces While metallic Rh is extremely expensive , significant improvements of HER activity of Rh-exchanged Ni, surpassing the activity of Pt (Figure 3), suggest this approach could be effectively Considering tremendous HER activity improvement, this is a small cost compared to the energy savings achievable using described surface modification of Ni. Thus, we further explore the effects of Rh exchange on the HER activity of extended Ni surface. First, we investigate the effect of Rh exchange on commercial Ni foam. The results show that the effects of time of exchange by Rh are not as pronounced as in the case of smooth Ni surface (Figure 6), and the most pronounced effects are seen after several minutes of exchange by Rh. For clean NF η 10 was found to be −0.26 V, and for 2 and 10 min exchange with Rh it reduces to −0.17, and −0.07 V. EDX analysis showed a monotonous increase of Rh surface concentration, without significant alteration of mesh morphology caused by dissolution in acidic Rh-containing solution (Figure 6). Nevertheless, for higher Rh exchange times, we clearly observed the formation of Rh phase on NF surface (Figure S4, Supplementary Information), which we consider mainly responsible for HER activity, as metallic Rh is much more active for HER than Ni. Also, we must note that it was impossible to perform electrode rotation to remove the H 2 bubbles formed on the surface in the case of NF. Thus the measured activity of Rh-exchanged Ni foam is slightly affected by H 2 bubbles blocking the surface. Still, the activities are better than those reported for pure Rh films prepared on NF using AACVD , while the modification procedure is greatly simplified. We also note that the improvements of the NF HER activity were previously reported for the case of spontaneous deposition of Ru and Pd . However, the activities reported in that work are much lower than those reported here, not only for the modified NF but also for the clean one. For this reason, we also investigated long Rh exchange time, 5 minutes, and obtained a further improvement of HER activity. In this case, η 10 amounted to −0.09 V, and the Tafel slope was −122±5 mV dec −1 . However, after SEM and EDX analysis, it was observed that the surface is significantly eroded (Figure S7, Supplementary Information), while EDX showed ~20 at.% Rh, in line with clear H UPD peaks after 5 min of exchange with Rh (Figure 7). This finding is in line with recorded cyclic voltammograms of Ni-dep electrodes before and after Rh exchange, showing progressive loss of voltammetric response corresponding to Ni 2+ /Ni 3+ transition with the extension of Rh exchange time (Figure 7). Such erosion of Ni-dep is likely due to fast corrosion of a highly developed Ni surface, and it was not observed for NF or smooth Ni-poly. We ascribe this to the fact that Ni-poly has a low concentration of highly uncoordinated Ni sites, prone to corrosion, unlike Nidep. Moreover, the surface of NF is also rather smooth, much smoother than that of Ni-dep (Figure 6, insets, and Figure 7). We believe this is why no pronounced dissolution and erosion of NF is seen. ## Conclusions Fast galvanic exchange of Ni with Rh, using concentrated acidic Rh solution, leads to tremendous HER activity improvements in alkaline media. In the case of a smooth Ni-poly surface, HER activity surpasses Pt-poly for Rh exchange times of 30 seconds. Our results suggest that there is no significant change in the surface roughness of Rh-exchanged electrodes, and when surface-specific activities of Rh-exchanged smooth Ni-poly are compared to that of Pt-poly, the HER activity of the modified Ni-poly (30 s of exchange) is roughly 2.5 times higher than the activity of Pt-poly. DFT calculations suggest that Rh-modified Ni surface interacts weaker with H ads , making the formation of

chemsum

{"title": "Enhancement of hydrogen evolution reaction kinetics in alkaline media by fast galvanic displacement of nickel with rhodiumfrom smooth surfaces to electrodeposited nickel foams", "journal": "ChemRxiv"}

peptide_late-stage_c(sp³)–h_arylation_by_native_asparagine_assistance_without_exogenous_d

## Abstract: There is a strong demand for novel native peptide motifs for post-synthetic modifications of peptides without pre-installation and subsequent removal of directing groups. Herein, we report an efficient method for peptide late-stage C(sp 3 )-H arylations assisted by the unmodified side chain of asparagine (Asn) without any exogenous directing group. Thereby, site-selective arylations of C(sp 3 )-H bonds at the N-terminus of di-, tri-, and tetrapeptides have been achieved. Likewise, we have constructed a key building block for accessing agouti-related protein (AGRP) active loop analogues in a concise manner. ## Introduction Peptides are increasingly important drug candidates, which are largely employed to treat metabolic disorders, cancer, allergy, and immune and cardiovascular diseases. 1 They also represent key tools that modulate biological events mediated by proteinprotein interactions (PPIs). 2 Native peptides usually suffer from poor pharmacological features due to lack of structural diversity or enzymatic degradation, 3 but chemically modifed nonnatural peptides could feature higher binding affinities to the target, as well as improved pharmacokinetics, stability, and cell permeability. 4 The late-stage modifcation represents an effective strategy to obtain structurally diverse peptides and peptidomimetics. Thus, late-stage modifcation methods of peptides have been achieved in terms of arylations, 5 alkylations, 6 and cycloadditions. 7 Over the past few years, C-H activation has been recognized as an atom-and step-economical pathway towards molecular syntheses, 8 with remarkable applications in materials science, 9 the agrochemical industry, 10 and drug discovery, 11 among others. 12 To the best of our knowledge, studies on latestage functionalizations of peptides via C(sp 3 )-H activation have been scarcely reported. In this context, Yu 13 successfully implemented C(sp 3 )-H activation of peptides using native N,Oor N,N-bidentate coordination without external auxiliary (Fig. 1a). On a different note, Noisier/Albericio 14 reported the synthesis of a novel class of stapled peptides. Likewise, research studies of post-synthetic modifcation of peptides through C(sp 3 )-H activation by installing exogenous auxiliary assistance have been pursued. In 2017, Ackermann 15 developed a strategy of triazole (Tzl)-assisted C(sp 3 )-H arylations of peptides. In the same year, Chen 16 described 8-aminoquinoline (AQ)-directed C(sp 3 )-H arylation to generate cyclophane-braced peptide macrocycles. Recently, Shi 17 established a palladium-catalyzed site selective g-C(sp 3 )-H silylation and d-C(sp 3 )-H alkylation of amino acids and peptides utilizing picolinamide (PA) auxiliary (Fig. 1b). The installation and subsequent removal of DGs often implies additional and non-trivial steps. Considering the atom-and step-economy of late-stage modifcation of peptides, we intended to utilize the natural amino acid embedded in the peptide backbone for chelation assistance. To our knowledge, C(sp 3 )-H functionalizations of peptides assisted by the unmodifed side chain of a natural amino acid has not been accomplished thus far. Asn is a natural amino acid with a side chain bearing a primary amide and could potentially be exploited as a directing group. This prompts us to survey whether the side chain and backbone of Asn could coordinate with palladium, leading to a bidentate coordinated palladium complex. Therefore, we introduce Asn as an internal bidentate DG to accomplish C(sp 3 )-H activation of peptides. Simultaneously, Asn is a common residue contained in many bioactive peptides, which display a range of biological activities, such as antioxidant activity, 18 blocking the neprilysin activity, 19 and inhibiting ACE activity. 20 Remarkably, Phe-Asn is an essential sequence that exists in some bioactive peptides, for example novel ACE inhibitory peptides, 21 anticancer peptides, 22 and AGRP. 23 Inspired by the signifcant work by Ackermann et al., 15,24 we provide a useful strategy employing Asn as an internal directing group for C(sp 3 )-H functionalization of peptides. The unmod-ifed side chain of Asn combined with the backbone was utilized as the N,N-bidentate coordination via 5,6-fused bicyclic palladacycles (Fig. 1c) to perform the late-stage peptide C(sp 3 )-H arylation. The complex has facilitated the inert C(sp 3 )-H bond arylation in peptides. Thereby, arylated di-, tri-, and tetrapeptides containing Asn have been assembled. The salient features of our approach comprise (a) C(sp 3 )-H activation of peptides assisted by a natural amino acid which circumvent the preinstallation and removal of DGs; (b) the frst unmodifed side chain of the natural amino acid as the endogenous auxiliary assistance applied in C(sp 3 )-H activation; and (c) discovery of native bidentate assistance through less-strained 5,6-fused bicyclic palladacycles. 25 ## Optimization of reaction conditions To validate our hypothesis, we initiated our studies by exploring reaction conditions for the palladium(II)-catalyzed primary C(sp 3 )-H arylation of N-phthaloyl protected dipeptide 1 with 3-iodotoluene (Tables 1 and S1 in the ESI †). Initial optimization revealed DCE to be the best solvent of choice (Table S1, † entries 1-5), with KF being identifed as the optimal additive (entries 1-3). By replacing Pd(OAc) 2 by PdCl 2 as the catalyst the yield of product 2a was excitingly increased to 67% when the amount of AgOAc and KF was increased to 2.5 equivalent (entry 4). Notably, the reaction failed to proceed using AgOTf as the additive (entry 5), while Cu(OAc) 2 gave a dramatically decreased yield (entry 6). Encouraged by the good efficiency of PdCl 2 , other palladium catalysts were further investigated. Gratifyingly, Pd(MeCN) 2 Cl 2 was found to slightly improve the yield of peptide 2a to 72% (entries 7-9). It is noteworthy that other metal catalysts, based on ruthenium, rhodium or cobalt, were ineffective (entries 10-12). The control experiment verifed the essential role of the palladium catalyst (Table S1, † entry 20). ## Substrate scope With the optimal reaction conditions in hand, the substrate scope of a range of aryl iodides was investigated, and the results are summarized in Scheme 1. Both substrates with electrondonating (Me-, MeO-, and t-Bu-) and electron-withdrawing (F-, Cl-, Br-, CF 3 -, and CO 2 Me-) groups reacted smoothly and afforded the desired products in good yields. Pleasingly, biphenyl and naphthyl moieties were also tolerated, leading to the corresponding products (2m and 2n) in 63% and 64% yields. The reaction performed with good chemo-selectivity. Encouraged by the success of the arylation of dipeptides, we next investigated the feasibility of applying this approach to the arylation of tripeptides and tetrapeptides (Scheme 2). Using tripeptide 3a as the substrate, through minor adjustment of the reaction conditions (Table S2 in the ESI †), we were pleased to fnd that the arylation of 3a with 1-iodo-4-methoxybenzene 4a could deliver the expected product 5aa in 61% isolated yield. Then, the scope of substrates was evaluated under the optimized reaction conditions. Satisfyingly, a wide range of aliphatic amino acids, including Leu, Ala, Val, and Lys, at the Cterminus of the tripeptides are compatible with these conditions. In addition, aryl iodides bearing electron-donating as well as electron-withdrawing substituents were tolerated, affording products 5aa-5ej. Given the feasibility of the tripeptide arylation, we expanded the peptide substrates to structurally complex tetrapeptides. The arylation products of tetrapeptides 6fa-6gh could be obtained in moderate yields (50-58%). Phecontaining tetrapeptide 3h could also be arylated albeit with lower yields (6hg-6hk, 25-36%). While considerable progress has been made in 3 )-H activation, 26 our strategy enabled position-selective arylation of Ala assisted by N,N-bidentate coordination of the Asn in tri-and tetra-peptides. To further demonstrate that the reaction coordination site is the primary amide of Asn, the control reaction and competition reaction were investigated under the standard conditions (Scheme 3). First, we removed the Asn side chain of dipeptide 1 and replaced it with a methyl group, while retaining the tertbutyl ester of the dipeptide. Therefore, N-phthaloyl protected dipeptide 7 was independently prepared, and subjected to the optimized reaction conditions. It failed to afford arylated products of arylation of C(sp 3 )-H bonds at the N-terminus (Scheme 3a). Since tripeptides or tetrapeptides both contain Asn bidentate and backbone amide bidentate, it is important to analyze the key role of Asn bidentate in promoting C(sp 3 )-H functionalization. For the competition experiment between tripeptides 3c and 8a, product ratio of approximately (5cg : 9a ¼ 6 : 1) (Scheme 3b) was obtained. ## Mechanistic investigation Additionally, we probed the catalyst mode of action by means of computational studies at the PW6B95-D4/def2-TZVP+SMD (DCE)//PBE0-D3BJ/def2-SVP level of theory (Fig. 2). 27 A detailed analysis between the C-H activation and reductive elimination elementary steps provided support for the C-H activation to be the rate-determining step with an activation energy of 19.6 kcal mol 1 , with oxidative addition being energetically more favorable by only 1 kcal mol 1 . An alternative pathway Scheme 1 Scope of ArI C(sp 3 )-H arylation of dipeptides. Reaction conditions: 1 (0.2 mmol), ArI (2.0 equiv.), Pd(MeCN) 2 Cl 2 (10 mol%), AgOAc (2.5 equiv.), KF (2.5 equiv.), DCE (2.0 mL), 130 C in air, 12 h, yields of isolated products. Scheme 2 Scope of C(sp 3 )-H arylation of tripeptides and tetrapeptides. a Reaction conditions: 3 (0.2 mmol), ArI (2.5 equiv.), PdCl 2 (15 mol%), AgOAc (3.0 equiv.), KF (2.0 equiv.), DCE (3.0 mL), 110 C in air, 12 h, yields of isolated products. b AgOAc (2.5 equiv.). c KF (1.0 equiv.) where the NH 2 of the terminal amide is deprotonated was also taken into consideration (Fig. S1, see the ESI †). The latter was shown to be overall energetically disfavored, with reductive elimination as the rate-determining step with a high energy barrier of 30.9 kcal mol 1 . These studies provide strong support for the palladium-catalyzed C(sp 3 )-H arylation to occur through a Pd(II/IV) pathway where the NH of the internal, instead of the terminal amide is deprotonated. Based on previous reports on palladium-catalyzed amidedirected C-H bond activation and computational studies, we propose a plausible catalytic cycle to be initiated by a facile organometallic C-H activation (Scheme 4). Initially, the palladium catalyst coordinates covalently with the deprotonated NH of the internal amide generating a bidentate coordinated palladium(II) complex A. Subsequently, complex A undergoes slow C(sp 3 )-H bond cleavage to form the 5,6-fused bicyclic palladium complex B. The oxidative addition of the aryl iodide to B affords palladium(IV) intermediate C, which then undergoes reductive elimination followed by protonation leading to the formation of the corresponding arylated product. The silver salt is proposed to accelerate the rate of the oxidative addition or the reductive elimination, while likewise acting as a halide scavenger. 8i,24a,28 Agouti-related protein (AGRP) is a potent orexigenic peptide that antagonizes the melanocortin-3 and melanocortin-4 receptors (MC3R and MC4R). 29 This protein has been physiologically implicated in regulating food uptake, body weight control, and energy homeostasis. 30 In attempts to improve the antagonist activity and selectivity of AGRP active loop, previous studies have applied a substitution strategy to prepare AGRP active loop analogues. 31 The results have indicated that some substitutions of amino acid could increase potency of AGRP. However, the synthesis of AGRP loop analogues requires the introduction of modifed unnatural amino acids. Some unnatural amino acids are expensive and difficult to synthesize, such as L-4,4 0 -biphenylalanine (Bip) and 3-(2-naphthyl)-L-alanine (Nal(2 0 )). Through C-H activation, the functional group could be installed directly into native peptides, such an approach is highly efficient, step-and atom-economical. Thus, we attempted to apply our strategy to synthesize new AGRP loop analogues. The arylation products 6 through deprotection of phthaloyl (Phth) gave NH 2 -free tetrapeptides 10 (details see the ESI †). Tetrapeptides 10a and 10b subsequently were coupled with Cbz-DPro-Pro-Arg(Pbf)-Phe-OH to obtain linear octapeptides, which were cyclized to access AGRP loop analogues. AGRP loop analogues 11a and 11b were obtained through this strategy (see the ESI † synthesis of AGRP loop analogues); the introduction of a bromide atom in 11b potentially enables further latestage derivatization of this peptide (Scheme 5). ## Conclusion In conclusion, we have developed an efficient strategy for palladium(II/IV)-catalyzed late-stage C(sp 3 )-H arylations of peptides using unprecedented internal Asn. The protocol avoids the additional requirement for installation and removal of exogenous directing groups. Importantly, our approach has provided a novel synthetic route to access the key building block for the synthesis of AGRP loop analogues. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Peptide late-stage C(sp³)\u2013H arylation by native asparagine assistance without exogenous directing groups", "journal": "Royal Society of Chemistry (RSC)"}

3d_printed_multifunctional_peek_bone_scaffold_for_multimodal_treatment_of_osteosarcoma_and_osteomyel

## Abstract: In this work, we developed the first 3D PEEK based bone scaffold with multi-functions targeting challenging bone diseases such as osteosarcoma and osteomyelitis. 3D printed PEEK/graphene nanocomposite scaffold was deposited with drug laden (antibiotics and/or anti-cancer drugs) hydroxyapatite coating. The graphene nanosheets within the scaffold served as effective photothermal agents that endowed the scaffold with on-demand photothermal conversion function under NIR laser irradiation. The bioactive hydroxyapatite coating significantly boosted the stem cell proliferation in vitro and promoted the new bone growth in vivo. The presence of antibiotics and anti-cancer drugs enabled eradication of drug resistant bacteria as well as ablation of osteosarcoma cancer cells, the treatment efficacy of which can be further enhanced by the on-demand laser induced heating. The promising results demonstrate the strong potential of our multi-functional scaffold in applications such as bone defect repair as well as multimodal treatment of osteosarcoma and osteomyelitis. ## Introduction Osteosarcoma (OS) is the most common type of primary bone cancer and it usually occurs in children and adolescents aged 10-19 years . It is mostly found in the long bones of the lower extremity and metastasis is diagnosed in 20% of patients . Osteomyelitis (OM) on the other hand, is a bacterial induced bone infection which also targets a considerable number of patients, particularly children, elderly, and patients with comorbidities (e.g. diabetics). The disease can lead to serious functional impairment, long-lasting disability, permanent handicap and even life threatening conditions. Both bone diseases are challenging clinical conditions, and the treatment protocol usually involves resection of the diseased bone tissue, followed by use of anti-cancer or antibiotic drugs. The treatment would inevitably result in large bone defect beyond the bone's self-healing ability (critical size bone defect), and the patients often suffer from recurrence of the diseases due to the residual cancer cell or bacteria. Multimodal therapy, such as chemo-photothermal therapy (chemo-PTT) , photothermal/photodynamic therapy (PTT/PDT) , PDT/chemotherapy therapy etc, are emerging strategies for the treatment of osteosarcoma and osteomyelitis. For instance, in chemo-PTT of bone cancer, cancer drugs can be delivered to the tumorous tissue in the targeted area, and the near-infrared (NIR) laser induced heating can enhance the sensitivity of tumor cells towards chemotherapy, leading to reduced drug dosage and improved treatment efficacy . Combined use of laser heating and antibiotics is also known to enhance bacterial eradication at the infection site , which can be potentially used for osteomyelitis treatment. To date, chemo-PTT is mainly achieved through local delivery of nanomaterials such as CuFeSe 2 , Fe-CaSiO 3 , graphene oxide , etc, which act as drug delivery vehicles and photothermal (PT) conversion agents. Some researchers also attempted to introduce PT conversion agents and drugs into hydrogels (such as pNIPAAm-co-pAAm) or 3D porous ceramic (such as β-TCP) or polymer (such as chitosan) scaffolds for bone repair and chemo-PTT therapy. However, the above mentioned materials cannot provide sufficient mechanical properties, which would impede their applications, particularly in load-bearing bone repair . In the past decade, polyetheretherketone (PEEK) has attracted increased attention in the biomedical field. PEEK has outstanding properties such as biocompatibility , X-ray/thermal/chemical stability , and mechanical properties (elastic modulus) similar to that of the human bone . Being a FDA approved biomaterial, PEEK has been successfully deployed in applications such as artificial knee joints , spine fusion , skull , and orthopedic implants , etc. Its thermoplastic nature also allows it to be 3D printed into tissue scaffold with be-spoke geometry . One major limitation of PEEK is its lack of bioactivity, which impedes its application for bone regeneration . To address this issue, various strategies (such as surface coating or bulk reinforcement with bioactive agents) have been adopted to enhance the bioactivity of PEEK implants, a detailed review of which can be found in . In the present study, we created a highly functional 3D printed PEEK/graphene composite scaffold with drug laden bioactive coating. The graphene nanosheets within the scaffold act as strong photothermal conversion agent, while the loaded drugs (such as antibiotics stearyltrimethylammonium chloride (STAC), and/or cancer drug cisplatin (DDP)) enabled effective cancer and/or bacteria eradication with the aid of NIR laser induced heating. The scaffold has been sucessfully demonstrated for bone regeneration, as well as multimodal treatment (chemo-PTT) of cancer ablation and bacterial eradication. We believe our multi-functional bone scaffold can potentially serve as a new platform for the management of challenging bone diseases such as osteosarcoma and osteomyelisis. wt% G were dispersed in ethanol and sonicated at 25 for 30 min. After filtration ℃ and drying, the power mixture was spun into filaments for further 3D printing (Funmat HT, Intamsys, Shanghai, China) of PEEK/G composite (PG) scaffold. 3D printed scaffolds (100 mm × 100 mm × 4 mm) with simple cubic lattice structure and different pore size (200 μm and 500 μm, respectively) were prepared. The PG scaffold were then plasma treated to modify the surface with oxygen rich functional groups, which can facilitate the subsequent electrophoretic deposition process. 3D printed pristine PEEK (P) scaffold was also prepared for comparison. ## Preparation of HA coated PG scaffold Hydroxyapatite (HA) was synthesized by hydrothermal method according to established procedures . Positively charged stearyltrimethylammonium chloride (STAC) (Aladdin, China) was loaded onto HA particles through physisorption and uniformly dispersed STAC/HA suspension was used for subsequent electrophoretic coating deposition. Electrophoretic deposited STAC/HA coating was deposited onto the 3D PEEK scaffold following under a DC voltage of 50 V for 60 min with a carbon rod being the anode and PG scaffold being the cathode. The coated scaffold was named as PGH. The coated scaffold was then immersed in sodium chloride solution of cisplatin (DDP, 1 mg/mL) following for anti-cancer drug loading, and the final drug laden scaffold is named as PGHD . ## Materials characterization Mechanical testing coupons (10 mm × 10 mm × 4 mm) and electrical conductivity testing coupons (100 mm × 100 mm × 1 mm) were produced from hot pressed P and PG sheet. Compressive tests were performed using universal mechanical testing machine (MTS, model E45, USA) at a speed of 1 mm/min, following GB/T 1041-2008/ISO 604:2002. Four-point probes method (RTS-8, Tianjin Nuleixinda Technology Co., Ltd., China) was used for electrical conductivity measurements. X-ray photoelectron spectroscopy analysis (XPS, XSAM800, Kratos, England) was performed to confirm the elemental state of PG before and after plasma treatment. Zeta potential (Zetasizer Nano ZS, Malvern, England) was used to determine the surface charge of the STAC-HA particles. Scanning electron microscope (SEM, JSM-7500F, JEOL, Japan) and EDS (JSM-7500F, JEOL, Japan) were used to analyze the morphology and elemental information of HA and the coating deposited on PGH. The photothermal conversion effect of all samples was analyzed in air and in phosphate buffered solution (PBS), respectively. A 808 nm NIR laser (0-2 W/cm 2 , Richeng Science and Technology Development Co. LTD, Shanxi, China) was used at different laser power densities (0.05 W/cm 2 , 0.15 W/cm 2 , 0.30 W/cm 2 ) and the temperature of the scaffold was monitored in real time using an infrared thermal imaging system (TiS20+, Fluke, USA). The temperature data were analyzed using FLUKE software. For drug release analysis, PGHD (10 mm ×10 mm × 4 mm) was immersed in 5 ml deionized water and placed in a shaking incubator under 37 . ℃ The cumulative release of DDP (Pt element) was detected by Inductively Coupled Plasma Emission Spectroscopy (ICP-OES, AXIS Ultra DLD, Kratos, UK) at different time intervals (1 h, 3 h, 10 h and 24 h). XPS was performed to detect the state of Pt element. ## Antibacterial testing Gram-negative Escherichia coli (E. coli, ATCC25922) and gram-positive Methicillin-resistant Staphylococcus aureus (MRSA, ATCC29213) were used to evaluate the antibacterial properties of the P, PG, PGH and PGHD scaffolds. Ten-fold dilution method was used to quantitatively measure the bactericidal rate (BR) defined by Eq.1 . Briefly, bacterial strain was incubated in culture medium for 24 h and the subculture from the second passage was used as the pre-made bacterial fluid. 50 μL pre-made bacterial fluid was drop-casted onto scaffold samples and incubated for 2 h followed by topping of 4 mL physiological saline. The mixture of bacterial fluid and physiological saline extracted from each sample was subsequently diluted 10 4 times. Finally, 50 μL of each diluted fluid sample was inoculated on nutrient Luria-Bertani agar plate. The number of bacterial colonies up to 30~300 CFU (colonyforming unit) was counted. Where n 0 is the number of colonies in the control group, n is the number of colonies in the experimental group. To investigate the effect of photothermal conversion on bacteria eradication, a separate set of scaffold samples were irradiated by NIR laser for 10 min before the ten-fold dilution method was applied. The bacterial fluid irradiated by NIR laser was used as the control (named NIR only). ## Tumor ablation experiments MG-63 cells were seeded in a 48-well plate at a density of 1.0 ×10 were measured every two days using a vernier caliper. The tumor volume (V t ) was calculated following: Where L is the tumor length, W is the tumor width, V s is the scaffold volume (5 mm 3 ). Relative tumor volume is defined as Where V t0 = V t (day 0) -V s . The body weight of the mice was also recorded every two days. Whole-body fluorescent imaging was performed on day 0 and day 10, respectively. On day 11, the mice were sacrificed and the tumors were harvested along with the heart, liver, spleen, lungs and kidney to evaluate the potential side effect of the scaffolds and/or the laser treatment. The tissue and organs were infiltrated with 4% paraformaldehyde, embedded in paraffin, and finally stained with hematoxylin and eosin (H&E). ## Cytocompatibility and Bone Regeneration Mouse MC3T3-E1 pre-osteoblasts ( 10 To investigate the proliferation of cells on the scaffolds, 10 4 MC3T3-E1 cells were seeded in 48-well plates containing scaffolds at 37 with 5% CO ℃ 2 . The cell proliferation on different scaffolds was assessed on day 1, 3, 5, and 7 by cell counting kit (CCK-8; Dojindo, Japan). Briefly, the culture medium was removed and 10% CCK-8-containing medium was added to each well in dark. After 2 h of incubation, 100 μL CCK-8 solution was transferred to a 96-well plate and examined by Multiscan Spectrum (Synergy Mx, Biotek, USA) at 450 nm. To study the scaffolds biocompatibility, MC3T3-E1 cells (1.0 ×10 4 cells/well) were seeded in a 48-well plate and the scaffolds were gently placed into the plate 24 h later. After an additional 24 h of incubation, the scaffolds were removed and the cells in the plate were stained with PI and Calcein-AM, respectively. Finally, cells were observed under a fluorescence microscope (Olympus IX83). Mouse MC3T3-E1 pre-osteoblasts were seeded in 48-well plate at a density of 10 4 cells/well with α -MEM (Hyclone) supplemented with 10% FBS (Gibco) and 1% Penicillin-Streptomycin-L-glutamine (Hyclone). 48 h after incubation, the cells reached confluence and the growth medium was removed and replaced with osteogenic medium comprising growth media supplemented with dexamethasone (100 nM; Sigma), L-ascorbic acid (50 μg/mL; Sigma), and β-glycerophosphate (10 mM; Sigma). The scaffolds were gently placed into the plate and the medium was replaced every 2 days. The expressions of osteogenesis-related genes were quantitatively analyzed by real-time reverse-transcriptase polymerase chain reaction (real-time PCR) on day 7. The total RNA was extracted using Trizol reagent (Invitrogen, USA) and the complementary DNA (cDNA) was obtained by synthesizing DNA from 1 μg of total RNA via reverse-transcription using an iScript cDNA Synthesis Kit (Bio-Rad) following the manufacturer's instructions. Real-time PCR was performed with EvaGreen Dye (Bio-Rad) using RT-PCR instrument (CFX Connect, Bio-Rad). The forward and reverse primers for different genes were listed in Table S1. Cycle threshold (Ct) values were used to determine fold differences according to the ΔΔCt method. β-actin was used as an internal reference to normalize the data. On day 10, two representative osteogenic proteins (OPN and OCN) were evaluated by immunofluorescence. The cells co-cultured with different scaffolds were fixed in 4% paraformaldehyde for 10 min, permeabilized with 0.25% Triton X-100 for 5 min, and incubated with 1% bovine serum albumin (BSA) for 1h at room temperature. Then the cells were incubated with primary antibodies in 1% BSA overnight at 4 (mouse anti-OPN monoclonal antibody, 1:50, Novus; rabbit anti-℃ OCN polyclonal antibody, 1:100, Absin). The cells were then incubated in dark with 1:500 dilution of secondary antibodies in 1% BSA for 1h at room temperature (Alexa Fluor ® 488-conjugated anti-mouse IgG and Alexa Fluor ® 555-conjugated anti-rabbit IgG, Cell Signaling). Finally, the cytoskeleton and nuclei were stained with phalloidin (Phalloidin-iFluor™ 555 Conjugate, AAT Bioquest; AbFluor™ 488-Phalloidin, Abbkin) and DAPI (Beijing Solarbio Science and Technology, China), respectively. All staining were followed by rinsing with PBS three times. Representative images were obtained using a fluorescence microscope (Olympus IX83). Alkaline phosphate (ALP) activity of MC3T3-E1 pre-osteoblasts with different scaffolds was quantified using ALP activity assay kit (Beijing Solarbio Science and Technology, China) on day 7. Briefly, cells were removed from the scaffold surfaces using Triton X-100 (1% v/v) and then centrifuged (12,000 rpm at 4 ) for 30 min to ℃ remove all cell debris. The supernatant was mixed with p-nitrophenyl phosphate (Songon) at 37 for 60 min. 4-Nitrophend NaOH was added to the cell supernatant ℃ and the ALP activity was examined by Multiscan Spectrum (Synergy Mx, Biotek, USA) at 510 nm. For standardization, the total protein content was calculated by a bicinchoninic acid (BCA) protein assay kit (Beyotime). The ALP activity was ultimately expressed as the total protein content (μM/mg). For ALP staining, the cells were fixed with 4% paraformaldehyde for 15 min, followed by staining using azo-coupling ALP color development kit (Beijing Solarbio Science and Technology, China) for 20 min in dark. Finally, cells were imaged using a fluorescence microscope (Olympus IX83). The calcium nodules formed by the MC3T3-E1 cells co-cultured with different scaffolds were stained by Alizarin Red S (Al, Sigma, USA) on day 21. Specifically, cells were fixed in 4% paraformaldehyde for 15 min. The cells were washed thrice with distilled water and stained with Al for 20 min. Afterwards, the excess Al was thoroughly removed with distilled water and the deposited calcium was imaged. For quantitative analysis, the stained samples were dissolved with 10% cetylpyridinium chloride and analyzed by a Multiscan Spectrum (Synergy Mx, Biotek, USA) at 600 nm. ## Bone regeneration in vivo All surgical procedures were approved by the Animal Ethics Committee of West China Hospital of Sichuan University, China. Sixteen male Sprague-Dawley (SD) rats (8 weeks old, 200 ± 20 g) were chosen to construct distal femoral condyle defect models for the evaluation of bone osseointegration in vivo. The rats were randomly split into four groups: P, PG, PGH and PGHD, and were anaesthetized by isoflurane. A hole (3 mm in diameter and 7 mm in depth) was made using a dental drill and the scaffold sample was press fit into the hole. 8 weeks post-surgery, the rats were sacrificed and their femurs were used for micro-CT scanning and histological staining. The harvested femoral specimen were fixed by 4% buffered paraformaldehyde, dehydrated using graded ethanol solution (70%, 85%, 95% and 100%), and embedded in polymethylmethacrylate without decalcification. The specimens were then microtomed into 5 μm thick slices along the cross-sectional surface, stained by methyl blue and basic fuchsin and observed under optical microscopy (Olympus IX83). The area of new bone tissue was calculated by the Image J software. Fluorescence observation was performed using CLSM (Nikon, Japan). The excitation/emission wavelengths used to observe the chelating fluorochromes were 543/620 nm and 488/520 nm for alizarin red S (AL, red) and calcein (CA, green), respectively. ## Statistical analysis Statistical analysis was performed using SPSS software (version 22.0; IBM Corp., Armonk, NY, USA), and the statistical significance was analyzed using oneway analysis of variance (ANOVA) followed by the LSD post hoc test. The relationships between different parameters were performed using Pearson's correlation test. Statistical significance was considered for p < 0.05, while high significance was p < 0.01. ## Results and discussion Fig 1a show the compressive moduli data of P and PG scaffolds with different pore size. Both mechanical properties decrease with increasing scaffold pore size. The scaffolds with 500 µm pores demonstrated compressive moduli comparable to that of human cancellous bone (<3 GPa) , and are therefore selected for further investigation. The electrical conductivity of PG is 2.75×10 -3 S/cm (Fig 1b), which is twelve order of magnitude higher than that of pristine P (~10 -15 S/cm), and is comparable to semiconducting material. ## Cancer ablation in vitro The CLSM images (Fig 6a ) showed that without NIR irradiation, MG-63 cells on P, P+, PG, and PGH spread well with strong pseudopodia attachment on the surfaces in all directions. The cells also showed bright cell nuclei and actin in all groups. For PGHD, the presence of cancer drug has led to marked change of MG-63 cell morphology, i.e., round shape with no clear orientation, indicating inhibition of the cell growth. Upon NIR irradiation (0.30 W/cm 2 for 10 min), MG-63 cells on all samples displayed round shape with no pseudopodia and featured thinning of the actin layer around the cell nucleus. The quantitative CCK-8 results of MG-63 (Fig 6b ) showed that in the absence of NIR irradiation (NIR-), only PGHD demonstrated significantly reduced MG-63 viability (59.8%) due to the presence of anti-cancer drug. All other groups (P, PG, PGH and control) showed no obvious anti-cancer effect (cell viability ~100%). With NIR irradiation, the number of live cells declined drastically in PG+, PGH+ and PGHD+, while the cell viabilities of P, P+, PG, and PGH were similar to that of the control group. Under 0.30 W/cm 2 NIR irradiation, the scaffolds (e.g. PG+, PGH+, and PGHD+) could reach the 45 within 100 s in immersed condition and remain ℃ stabilized for the remainder of the treatment time. Literature suggests that the temperature in the range of 45~50 can result in rapid necrosis of tumor cells ℃ (cancer ablation) because of their lower heat tolerance result can result in DNA and protein denaturation . It is worth noting that the cell viability of PGHD+ further decreased to 24.9%, significantly lower than that of PG+ and PGH+ (45.7% and 45.4%, respectively). The results of live/dead staining (Fig 6c ) were in accordance with the CCK-8 results., confirming PGHD+ is the most effective group in cancer ablation due to the combined effect of cancer drug release and hyperthermia introduced by PPT. ## In vitro and in vivo osteogenic differentiation The bone regeneration is essential for bone defect repair. Osteogenic differentiation of the scaffolds was investigated both in vitro and in vivo and related markers were quantified by real-time PCR (Fig 10a). ALP, OPN and OCN are the corresponding makers representing different stages of osteogenic differentiation (ALP: early stage; OPN: secondary stage; OCN: late stage.) . Col-I is one of the key components of extra cellular matrix deposition . After culturing for 7 days, the gene expressions of ALP, OPN, OCN and Col1α1 in PG, PGH and PGHD were significantly upregulated as compared to P, where PGH and PGHD had similar performance. The osteo-related protein expressions of different scaffolds were also evaluated by immunofluorescence (Fig 10b and 10c). The highest OPN and OCN fluorescence intensity was observed in PGH and PGHD, followed by PG and then P. The trend was consistent with what was observed from the real-time PCR results. The above in vitro osteogenic differentiation data revealed that the boneregeneration capability of different groups follows the order: PGH and PGHD>PG>P. The best osteogenesis ability demonstrated by PGH and PGHD groups can be attributed to the presence of nanostructured HA coating, which promoted the serum protein adsorption, cell adhesion, attachment and and proliferation . The calcium and phosphate ions released from HA can also upregulate the expression of osteogenic genes and activate secretion and extracellular matrix mineralization . The presence of G in PEEK of PG group also demonstrated improved the scaffold osteogenic differentiation due to the osteoinductive capability of G . Clinical translation of biomedical materials/products replies on rigorous in vitro and in vivo studies. Well-designed correlation analysis may help to identify the key in vitro parameters that can be used to predict the bone regeneration capacity in vivo. This could subsequently reduce the economic and time-cost for in vivo studies . In this study, Pearson's correlation test was used to analyze the relationship between the biocompatibility parameters (CCK8-1d, 3d, 5d and 7d), osteogenesis-related genes ## Conclusion In this study, we design and developed a 3D printed hybrid PEEK/graphene scaffolds with drug laden bioactive hydroxyl apatite coating. The scaffold has tailored porous structure / mechanical properties, and has combined drug release/photothermal therapeutic function. Results show that the scaffold with loaded antibiotic and/or anticancer drugs and on-demand photothermal conversion effect can achieve near total eradication of Escherichia coli (E. coli) and Methicillin-resistant Staphylococcus aureus (MRSA), as well as effective ablation of osteosarcoma cancer cells. The coated scaffold also exhibited strong bone regeneration ability, demonstrating its strong potential in applications such as bone defect repair as well as multimodal management of osteosarcoma and osteomyelitis.

chemsum

{"title": "3D printed multifunctional PEEK bone scaffold for multimodal treatment of osteosarcoma and osteomyelitis", "journal": "ChemRxiv"}

time-resolved_proteomic_analysis_of_quorum_sensing_in_vibrio_harveyi

## Abstract: Bacteria use a process of chemical communication called quorum sensing to assess their population density and to change their behavior in response to fluctuations in the cell number and species composition of the community. In this work, we identified the quorum-sensing-regulated proteome in the model organism Vibrio harveyi by bio-orthogonal non-canonical amino acid tagging (BONCAT). BONCAT enables measurement of proteome dynamics with temporal resolution on the order of minutes. We deployed BONCAT to characterize the time-dependent transition of V. harveyi from individual-to group-behaviors.We identified 176 quorum-sensing-regulated proteins at early, intermediate, and late stages of the transition, and we mapped the temporal changes in quorum-sensing proteins controlled by both transcriptional and post-transcriptional mechanisms. Analysis of the identified proteins revealed 86 known and 90 new quorum-sensing-regulated proteins with diverse functions, including transcription factors, chemotaxis proteins, transport proteins, and proteins involved in iron homeostasis. ## Introduction Bacteria assess their cell numbers and the species complexity of the community of neighboring cells using a chemical communication process called quorum sensing. Quorum sensing relies on the production, release, accumulation and group-wide detection of signal molecules called autoinducers. Quorum sensing controls genes underpinning collective behaviors including bioluminescence, secretion of virulence factors, and bioflm formation. The model quorum-sensing bacterium Vibrio harveyi integrates population-density information encoded in three autoinducers AI-1, CAI-1, and AI-2, which function as intraspecies, intragenus, and interspecies communication signals, respectively. V. harveyi detects the three autoinducers using the cognate membrane-bound receptors LuxN, CqsS, and LuxPQ, respectively. At low cell density (LCD), autoinducer concentrations are low, and the unliganded receptors act as kinases, funneling phosphate to the phosphorelay protein LuxU. 10 LuxU transfers the phosphoryl group to the response regulator protein LuxO, which activates transcription of genes encoding fve homologous quorum regulatory small RNAs (qrr sRNAs). 11,12 The Qrr sRNAs post-transcriptionally activate production of the transcription factor AphA and repress production of the transcription factor LuxR. AphA and LuxR are the two master quorum-sensing regulators that promote global changes in gene expression in response to population density changes. At high cell density (HCD), autoinducer binding to the cognate receptors switches the receptors from kinases to phosphatases, removing phosphate from LuxU and, indirectly, from LuxO. Dephosphorylated LuxO is inactive so transcription of the qrr sRNA genes ceases. This event results in production of LuxR and repression of AphA. 12 Thus, the circuitry ensures that AphA is made at LCD, and it controls the regulon required for life as an individual, whereas LuxR is made at HCD, and it directs the program underpinning collective behaviors. Previous microarray studies examined the transcriptomic response during quorum-sensing transitions. That work showed that AphA and LuxR control over 150 and 600 genes, respectively and $70 of these genes are regulated by both transcription factors. 15 Both AphA and LuxR act as activators and as repressors, and thus the precise pattern of quorumsensing target gene expression is exquisitely sensitive to fluctuating levels of AphA and LuxR as cells transition between LCD and HCD modes. Developing a comparable understanding of the quorum-sensing-controlled proteome requires measurement of dynamic changes in protein abundance throughout the transition from individual to collective behavior. In this work, we used the bio-orthogonal non-canonical amino acid tagging (BONCAT) method to track the proteomewide quorum-sensing response in V. harveyi with temporal precision. BONCAT enabled us to identify 176 proteins that are regulated during the transition from individual to collective behavior; 90 of these proteins are in addition to those identifed in earlier studies. We show that a broad range of protein functional groups, including those involved in metabolism, transport, and virulence, change during the transition to group behavior. We demonstrate how particular temporal patterns of protein production are linked to particular tiers of the regulatory cascade by comparing the proteomic profles of the regulon controlled by the post-transcriptional Qrr sRNAs to the regulon controlled by the transcriptional regulator LuxR. Using this approach, we, for example, determined that the V. harveyi type VI secretion system is LuxR-regulated. ## Results The BONCAT method was developed to provide time-resolved analyses of the cellular proteome. 16,17 In a BONCAT experiment, the non-canonical amino acid L-azidohomoalanine (Aha; Fig. S1a †) is provided to cells and, subsequently, incorporated into proteins in competition with methionine. 18 Aha-labeled proteins are chemically distinct from the remainder of the protein pool and thus, labeled proteins can be selectively conjugated to affinity tags for enrichment and mass spectrometry analysis (Fig. 1a). Because Aha can be introduced into cells in a well-defned pulse, BONCAT offers excellent temporal resolution and high sensitivity to changes in protein synthesis in response to biological stimuli. 19 Our goal was to identify time-dependent changes in protein production associated with quorum sensing. We chose to monitor the transition from individual to group behavior in V. harveyi because the core transcriptional regulon is wellestablished, providing a solid foundation for comparisons between transcriptional and translational outputs. 15 To experimentally manipulate the transition from LCD to HCD, we used V. harveyi strain TL25 in which the genes encoding the autoinducer receptors for CAI-1 (cqsS) and AI-2 (luxPQ) and the AI-1 synthase (luxM) have been deleted. 15 Thus, V. harveyi TL25 responds exclusively to exogenously supplied AI-1, which enables precise control over the activation of quorum sensing. The hallmark phenotypic response controlled by quorum sensing in V. harveyi is bioluminescence, which is activated by LuxR during the transition from LCD to HCD. 20 Thus, we reasoned that light production could serve as a proxy for activation of quorum sensing. 20 Upon treatment of a culture of V. harveyi TL25 with AI-1, bioluminescence increases sharply after 30 min and plateaus at a level 400-fold higher than the preaddition level after approximately 90 min (Fig. 1b). Detection of Aha incorporation in V. harveyi cultures by in-gel fluorescence showed that BONCAT experiments could be performed in this system with a temporal resolution of ten minutes (Fig. 1c). Using the bioluminescence profle as a guide, we combined two techniques, BONCAT and stable isotope labeling with amino acids in cell culture (SILAC), to monitor both increases and decreases in protein synthesis in ten-minute intervals between 0 and 90 min following addition of AI-1 (Fig. 1b and S1c and d †). 19,21 V. harveyi cultures that were not treated with AI-1 served as references for relative quantifcation. As expected, the production of the luciferase subunits LuxA and LuxB tracked with the bioluminescence profle in cultures treated with AI-1 (Fig. 1b). We detect LuxB at 30 min, slightly before we can detect LuxA. The LuxB measurement is coincident with the frst increase in bioluminescence. Between 40 and 50 min, bioluminescence and LuxA and LuxB levels exhibited sharp increases, after which, both continued to climb at slower rates. Between 60 and 90 min, the production rates of LuxA and LuxB remained nearly constant while bioluminescence continued to increase. LuxA and LuxB increased about 8-fold total in response to autoinducer supplementation. This result highlights the fact that BONCAT measures protein synthesis rates during individual time intervals (not total protein abundance), whereas bioluminescence output reports on the total accumulated LuxAB activity. LuxA and LuxB are encoded by the lux operon, which also encodes LuxC, an acyl-CoA reductase, LuxD, an acyl transferase, and LuxE, a long-chain fatty-acid ligase. LuxCDE synthesize the substrate required by the LuxAB luciferase enzyme. All fve proteins exhibited large, concurrent increases in translation at 50 min (Fig. 1d). The increase in bioluminescence precedes production of LuxCDE, which suggests some basal level of luciferase substrate is present. The coincidence of the production of LuxA and LuxB with the onset of bioluminescence, and the simultaneous up-regulation of all of the proteins in the lux operon validate the BONCAT technique as a reliable method for time-resolved analysis of the quorum-sensing response. ## Detection of quorum-sensing regulators At the core of the quorum-sensing circuit are the transcriptional regulators LuxO, AphA, and LuxR, which drive quorum-sensing transitions. Expression of luxO, aphA, and luxR are themselves controlled by multiple regulatory feedback loops. 13,15, To assess the consequences of addition of AI-1 to V. harveyi TL25 on these core regulators, we monitored both mRNA and protein synthesis using qRT-PCR and BONCAT, respectively. LuxO, AphA, and LuxR all showed rapid changes in protein production within 20 min of AI-1 treatment (Fig. 2). AphA and LuxR reached near-maximal differences in translation at the 30 min point; AphA protein production decreased 4-fold and LuxR protein production increased 16-fold. The mRNA levels of aphA and luxR tracked with those of AphA and LuxR protein changes, with the exception that luxR mRNA decreased in abundance between 60 and 90 min while the protein level remained constant. LuxO protein exhibited a consistent 2-fold increase in abundance throughout the time-course, whereas the corresponding mRNA levels slightly decreased. This pattern is consistent with the recent fnding that the Qrr sRNAs control luxO mRNA through a sequestration mechanism such that the Qrr sRNAs repress LuxO protein production while not signifcantly altering mRNA abundance. 25 ## Quorum sensing causes global changes in protein synthesis Using the above protocol for induction of quorum sensing in V. harveyi TL25, we next examined the quorum-sensingcontrolled proteome using BONCAT to monitor protein synthesis in ten-minute time intervals immediately following addition of AI-1. We collected a total of 700 174 MS/MS spectra and identifed 9238 peptides and 1564 unique protein groups (Fig. S3a and b, dataset S1 †). Proteins were identifed with an average of 6 peptides (median ¼ 4); 88% of proteins were iden-tifed by 2 or more peptides (Fig. S3c †). Relative protein abundances at each time point were calculated with an average of 49 unique quantifcations (median ¼ 17) (Fig. S3d †). By comparing evidence counts, MS-MS counts, and MS intensities of Met and Aha-containing peptides, we estimated the extent of replacement of Met by Aha to be roughly 15% (Table S1 †). Proteins with differences greater than 1.5-fold with false discovery rateadjusted p-values less than 0.05 were considered signifcant. Induction of quorum sensing altered production of 176 proteins (Fig. 3a). Unsupervised hierarchical clustering partitioned the regulated proteins into 10 groups based on their temporal production profles (Fig. 3a and b). Proteins from the lux operon clustered closely (group F), and LuxR and AphA, which exhibited distinct production profles, were assigned to very small clusters. Several clusters showed differences in protein production at early time points (groups D, E, I), whereas other clusters changed more abruptly at the 50 min time point (groups B, D, F, H) (Fig. 3b). Differences in protein production between AI-1-treated and control cultures were modest within the frst 20 min, with only 7 and 19 signifcant protein changes at 0-10 min and 10-20 min, respectively. The number of autoinducer-regulated proteins increased with time after induction, with 42-119 proteins altered between 40-90 min after AI-1 treatment (Fig. 3c and d). 90 of the AI-1-regulated proteins are newly associated with quorum sensing in V. harveyi (Fig. 3e, Table 2). In total, our analysis identifed 278 proteins that are members of the previously established aphA, luxR, or quorumsensing regulons. 15 Interestingly, only 86 of these proteins exhibited signifcant up-or down-regulation by BONCAT (Fig. S4 †). ## Bioinformatic analysis reveals regulation of functionally related protein groups To identify major shifts in protein production in response to induction of quorum sensing, we used principal component analysis (PCA) to simplify the dataset by reducing the dimensionality from 9 time points to 2 principal components. Weighting vectors showing the contribution of each time point to the principal components highlighted three distinct proteomic states: (1) an early period in which few proteins changed (10-30 min), (2) a transitional period that included rapid changes in protein production (40-50 min), and (3) a late period in which many proteins exhibited large differences in translation (60-90 min) (Fig. 4a, Table S2 †). As confrmation of these states, proteins with principal component coordinates near the 1 st , 2 nd , and 3 rd sets of vectors exhibited time-course production profles with punctuated changes at early, middle, and late stages (Fig. 4b). Gene ontology analysis identifed 13 protein groups regulated by quorum sensing (Fig. 4c and S5 †). Several of these groups were involved in transport, including iron, oligopeptide, and dicarboxylic acid transport. A set of 50 proteins with functional annotations for transporter activity was the largest of enriched ontology groups. Other groups of biological processes included bioluminescence, type VI secretion, siderophore synthesis, thiamine metabolism, and chemotaxis. To identify groups of functionally related proteins with similar patterns of protein production, we mapped protein interactions from the STRING database onto the PCA plot and scanned for protein networks that localized via their principal components (Fig. 4d). Consistent with our gene ontology analysis, we identifed interacting protein groups associated with regulation of bioluminescence, type VI secretion, chemotaxis, iron homeostasis, oligopeptide transport, and thiamine metabolism in the quorum-sensing response (Fig. 4d). For example, regarding peptide transport, synthesis of the substrate binding protein of the oligopeptide permease complex, OppA, decreased two-fold between 50-90 min. 26 Also, a large group of proteins ( 16) involved in iron transport exhibited decreased production profles late in the experiment, and a group of ironregulatory proteins (6) increased in levels. With respect to chemotaxis, we observed both increases and decreases in protein levels: homologs of methyl-accepting chemotaxis proteins and the CheA and CheY signaling proteins decreased, whereas putative methyl-accepting chemotaxis proteins increased in abundance. Taken together, these results suggest an overall quorum-sensing-driven remodeling of iron homeostasis and chemotactic behavior. ## Dening the temporal order of protein regulation in response to quorum sensing The Qrr sRNAs play a central role in dictating the transition between LCD and HCD states by controlling expression of the quorum-sensing transcriptional regulators, AphA, LuxR, and LuxO (Fig. 5a). 22,23 The Qrr sRNAs directly regulate 16 additional targets outside of the quorum-sensing cascade with functions in virulence, metabolism, polysaccharide export, and chemotaxis. 27 The direct Qrr targets constitute the set of "frstresponse" genes and also trigger the later, broader changes in downstream gene expression. With respect to the second wave of quorum-sensing gene expression changes, LuxR plays the major role. Therefore, we compared the temporal patterns of regulation of proteins known to be direct targets of either the Qrr sRNAs or LuxR. 27,28 We detected regulation of production of seven proteins known to be encoded by Qrr-regulated genes, all of which exhibited signifcant differences in expression within 20 minutes of AI-1 treatment (Fig. 5b, Table S3 †). Conversely, 20 of the 21 LuxR-regulated proteins identifed by BONCAT showed differences in production only after at least 30 minutes of AI-1 induction. Thus, the differences in timing between Qrr-and LuxR-regulated genes reflect the underlying structure of the quorum-sensing circuitry. We investigated the protein production profles of the newly identifed proteins to pinpoint additional candidates for regulation by the Qrr sRNAs. We found 19 additional proteins that are regulated within 20 minutes of AI-1 treatment, suggesting that the corresponding mRNAs may be targeted by the Qrr sRNAs (Table 1). The candidates include two putative chemotaxis proteins, the serine protease inhibitor ecotin, the type III secretion protein chaperone SycT, a chitinase, and several other proteins involved in metabolism. Strikingly, the mRNA and protein production of VIBHAR_02788 (a predicted chemotaxis protein) increased 4-and 12-fold, respectively, within the frst 10 minutes after AI-1 treatment, suggesting that VIBHAR_02788 is a good candidate for posttranscriptional regulation by the Qrr sRNAs (Fig. 5c). The mechanisms that control production of quorum-sensingregulated proteins undoubtedly become more complex as the response progresses. We identifed proteins that were regulated at all stages (early (0-20 min), intermediate (20-60 min), and late (60-90 min)) following AI-1 treatment (Fig. 5d, dataset S1 †). Differences in the timing of quorum-sensing-regulated proteins suggest that additional regulatory components or mechanisms orchestrate the transition from individual to group behavior. For example, direct LuxR targets were regulated in both the intermediate and late phases, despite the fact that LuxR reaches its peak production at 30 min (Fig. 2a). This result suggests that accumulation of LuxR or additional transcriptional regulators contribute to control of LuxR-regulated genes. ## Quorum sensing regulates type VI secretion proteins in V. harveyi Components of the type VI secretion system (TSSS) were among the proteins most strongly up-regulated in response to AI-1 treatment (Fig. 6a). Identifed TSSS proteins included the haemolysin co-regulated effector protein (Hcp; VIBHAR_05871), and two additional proteins whose homologs have been implicated in TSSS regulation and Hcp secretion (VIBHAR_05854 and VIB-HAR_05858). 29,30 TSSS proteins exhibited a coordinated increase in production at 50 min, a profle similar to that of LuxCDABE. In V. harveyi, the TSSS homologs are encoded by fve putative operons: VIBHAR_05855-05851, VIBHAR_05856-05858, VIBHAR_05865-05859, VIBHAR_05871-05866, and VIB-HAR_05872-05873 (Fig. S6a †). Analysis of the mRNA levels of the operons confrmed the increase in expression of TSSS components between 50 to 60 min after AI-induction; timing consistent with second-tier regulation (Fig. 6b). Previous microarray data comparing wild-type, DluxR, DaphA, and DluxR DaphA V. harveyi strains showed that TSSS gene expression was reduced in DluxR strains, but expression was not altered in the DaphA strain, providing evidence that expression of TSSS genes is LuxR-dependent and AphA-independent (Fig. S6b †). 13 Consistent with this notion, ChIP-seq data identifed a LuxR binding site in the bi-directional promoter region of VIB-HAR_05855-05856. 28 Using electrophoretic mobility shift assays, we confrmed the presence of this LuxR binding site and determined that LuxR binds to two additional promoter regions in the TSSS locus (Fig. S6c †). This result shows that, unlike Vibrio cholerae which deploys the Qrr sRNAs to posttranscriptionally regulate TSSS, V. harveyi uses LuxR to control TSSS production. 31 This fnding suggests that although both organisms have TSSS under quorum-sensing control, they employ different regulatory strategies to achieve distinct timing of TSSS protein production. ## Discussion and conclusions Global transcriptomic studies of V. harveyi have uncovered a continuum of changes in gene expression during the transition from LCD to HCD. As V. harveyi responds to changes in concentrations of autoinducers, shifts in the levels of the regulatory components AphA, LuxR, and the Qrr sRNAs occur, which in turn alter the expression of the downstream genes in the quorumsensing regulon. Here we used the BONCAT method to measure changes in the quorum-sensing-regulated proteome during the transition from LCD to HCD, with a time-resolution of 10 min. We found correlated changes in production of the LuxCDABE enzymes and in the intensity of bioluminescence produced by the culture, and we observed regulation of the core regulatory components AphA, LuxR, and LuxO. Notably, the increase in LuxO upon induction of quorum sensing occurred at the level of the protein, but not the mRNA, consistent with the hypothesis that the luxO mRNA is regulated by sequestration by the Qrr sRNAs. 25 The time resolution of the BONCAT method allowed us to identify proteins whose rates of synthesis were altered during the early, intermediate, and late stages of the LCD to HCD transition. The proteins found to be regulated within the frst 20 min of autoinducer treatment included seven of the 20 known Qrr sRNA targets along with 19 other proteins not previously associated with Qrr regulation. No known Qrr targets were regulated at later times. In contrast, changes in the known LuxR targets occurred between 30 and 90 min following induction. Notably, proteins in the TSSS were up-regulated between 40 and 50 min following autoinducer treatment, suggesting LuxR regulation of type VI secretion in V. harveyi; this conclusion was confrmed by electrophoretic mobility shift assays. Several LuxR-regulated genes exhibited changes in protein production only very late in the BONCAT experiment, which suggests either that they are responsive to accumulating LuxR levels, that they are regulated by another transcription factor downstream of LuxR, or that they are co-regulated by other factors. We found quorum-sensing-dependent changes in 176 proteins that span a broad range of functional groups, including those related to iron homeostasis, molecular transport, metabolism, and chemotaxis. Ninety of these proteins are newly associated with quorum sensing in V. harveyi, and expand what is known about the roles that quorum sensing plays in these processes. 13,32 The remaining 86 proteins are members of the previously established quorum-sensing, AphA, and/or LuxR regulons. Interestingly, nearly 200 other proteins from these regulons were identifed by BONCAT but were not signifcantly up-or down-regulated. For example, the quorum-sensing regulon, which was defned by differences in gene expression between a mutant V. harveyi strain locked at LCD and a strain locked at HCD, contains 365 regulated genes as determined by microarray analysis. 15 We quantifed protein expression levels of 127 (35%) of these genes, 45 (35%) of which were signifcantly regulated. The differences between the genetic and proteomic results may arise, at least in part, from differences in regulation at the levels of mRNA and protein, or from differences in the growth media used in the two experiments (rich (LM) medium in the genetic study vs. minimal (AB) medium here). 13,15 Furthermore, we would not expect the rapid addition of saturating amounts of AI-1 to a V. harveyi culture to reproduce precisely the effects of genetically locking the strain into either the LCD or the HCD state. Determining how environmental conditions affect the quorum-sensing response will be important to the development of a full understanding of bacterial communication in complex natural environments. The BONCAT method has allowed us to identify a diverse set of proteins that respond to the induction of quorum sensing in V. harveyi. The method facilitates monitoring of changes in protein synthesis on a time scale of minutes, and enables correlation of those changes with the underlying temporal pattern of regulation of the quorum-sensing response. The approach described here should prove useful in studies of a wide variety of time-dependent cellular processes. ## Cell culture For each set of experiments, overnight cultures of V. harveyi strain TL25 (DluxM DluxPQ DcqsS) was used to inoculate 625 mL of AB minimal medium containing 18 amino acids (-Met, -Lys) at an OD 600 of 0.003. 15 The culture was divided into six 100 mL aliquots. Three aliquots were supplemented with "light" Lys and three were supplemented with "heavy" Lys (U-13 C 6 U-15 N 2 L-lysine, Cambridge Isotope Laboratories). When the aliquoted cultures reached an OD 600 of 0.1 ($5 doublings), two "heavy" cultures (replicates 1 and 2) and one "light" culture (replicate 3) were treated with AI-1 at a fnal concentration of 10 mM ('AI-1 added'); the other three cultures were left untreated ('no AI-1 added'). At the specifed time intervals, Aha was pulsed into all six cultures at a fnal concentration of 1 mM. After 10 min of Aha treatment, protein synthesis was halted by the addition of 100 mg mL 1 chloramphenicol (Sigma). Cells were pelleted, frozen at 80 C, and stored for downstream processing. Aha was synthesized as described previously. 33 Cultures were grown at 30 C in a shaking incubator at 250 rpm. ## Molecular methods To measure changes in gene expression following induction of quorum sensing in V. harveyi TL25, cultures were grown as described above, divided in half, and AI-1 was added to one of the aliquots. Samples were collected every 10 min and RNA was isolated as described previously. 13 cDNA synthesis and qRT-PCR were performed as described previously. 22 The levels of gene expression were normalized to the internal standard hfq using either the DDC T method or the standard curve method. At least two replicates were collected for each sample ('AI-1 added' or 'no AI-1 added'). The graphs show the average of those measurements and are calculated as 'AI-1 added' divided by 'no AI-1 added'. Electrophoretic mobility shift assays were performed as previously described. 15 PCR products were generated using oligonucleotides (Integrated DNA Technologies) listed in Table S4. † BONCAT Cells were lysed by heating in 1% SDS in PBS at 90 C for 10 min and lysates were cleared by centrifugation. Protein concentrations were determined with the BCA protein quantitation kit (Thermo Scientifc), and paired 'light' and 'heavy' cultures were mixed at equal quantities of total protein. Azide-alkyne click chemistry was performed as described in Hong et al. with a 0.1 mM alkyne-DADPS tag and allowed to proceed for 4 h at room temperature (Fig. S1e †). 34 The DADPS tag was synthesized as described previously. 35 Proteins were concentrated by acetone precipitation and solubilized in 2% SDS in PBS. Solutions were diluted to 0.15% SDS in PBS, and tagged proteins were captured by incubating with streptavidin UltraLink resin (Thermo Scien-tifc) for 30 min at room temperature. Resin was washed with 35 column volumes of 1% SDS in PBS and 10 column volumes of 0.1% SDS in ddH 2 O. The DADPS tag was cleaved by incubating the resin in 5% formic acid in 0.1% SDS in ddH 2 O for 1 h. Columns were washed with 5 column volumes of 0.1% SDS in H 2 O, during which proteins remained bound, and proteins were subsequently eluted in 15 column volumes of 1% SDS in PBS. Protein enrichment was confrmed by SDS-PAGE, and eluted proteins were concentrated on 3 kDa MWCO spin flters (Amicon). ## In-gel digestion Concentrated proteins were separated on precast 4-12% polyacrylamide gels (Life Technologies) and visualized with colloidal blue stain (Life Technologies). Lanes were cut into 8 slices and proteins were destained, reduced, alkylated, digested with LysC (Mako), and extracted as described in Bagert et al. 19 Extracted peptides were desalted with custom-packed C 18 columns as described in Rappsilber et al., lyophilized, and resuspended in 0.1% formic acid (Sigma). 36 Liquid chromatography-mass spectrometric analyses Liquid chromatography-mass spectrometry and data analyses were carried out on an EASY-nLC-orbitrap mass spectrometer (Thermo Fisher Scientifc, Bremen, Germany) as previously described with the following modifcations. 37 For the EASY-nLC II system, solvent A consisted of 97.8% H 2 O, 2% ACN, and 0.2% formic acid and solvent B consisted of 19.8% H 2 O, 80% ACN, and 0.2% formic acid. For the LC-MS/MS experiments, samples were loaded at a flow rate of 500 nL min 1 onto a 16 cm analytical HPLC column (75 mm ID) packed in-house with ReproSil-Pur C 18 AQ 3 mm resin (120 pore size, Dr Maisch, Ammerbuch, Germany). The column was enclosed in a column heater operating at 30 C. After ca. 20 min of loading time, the peptides were separated with a 60 min gradient at a flow rate of 350 nL min 1 . The gradient was as follows: 0-30% solvent B (50 min), 30-100% B (1 min), and 100% B (8 min). The orbitrap was operated in data-dependent acquisition mode to alternate automatically between a full scan (m/z ¼ 300-1700) in the orbitrap and subsequent 10 CID MS/MS scans in the linear ion trap. CID was performed with helium as collision gas at a normalized collision energy of 35% and 30 ms of activation time. ## Protein quantication and ratio statistics Thermo RAW fles were processed with MaxQuant (v. 1.4.1.2) using default parameters and LysC/P as the enzyme. Peptide and protein false discovery rates were fxed at 1% using a targetdecoy approach. Additional variable modifcations for Met were Aha (4.9863), L-2,4-diaminobutanoate (30.9768), a product of Aha reduction, alkyne-DADPS (+835.4300), and 5-hexyn-1-ol (+93.0868), a product of alkyne-DADPS cleavage. Multiplicity was set to 2, and light and heavy (+8.0142) lysine labels were specifed for all experiments. Aha and 5-hexyn-1-ol modifcations were included in protein quantifcation. We required protein quantifcations to be calculated with at least two evidences for each set of experiments. Both pooled variances and bootstrap statistical methods were employed as previously described to estimate the individual protein ratio standard errors. 19,38 First, pooled estimates of peptide variation were calculated separately for peptides with well-characterized ratios and those based on requantifcation in MaxQuant. Second, standard errors of the overall protein ratios were calculated by generating 1000 bootstrap iterations. These iterations were generated by resampling the replicates and peptides and adding a small amount of random variation to each measurement based on the pooled variance estimates. Once the bootstrapped samples were generated for each protein, the standard error of the protein ratio was calculated from the standard deviation of the bootstrapped iterations. Using the standard error, proteins with ratios signifcantly different from 1 : 1 were identifed using a Z-test and p-values were adjusted to account for multiple hypothesis testing using the Benjamini and Hochberg method. 39 ## Bioinformatic analysis Hierarchical clustering was performed with R (v. 3.1.1) using Ward's method. 40 Confdence intervals (95 th percentile) for cluster time-series data were calculated by a bootstrapping approach using the tsplot function from the Python (v. 2.7) module seaborn (v. 0.4.0). Singular value decomposition was computed for PCA with the Python module matplotlib.mlab (v. 1.4.0). Gene ontology analysis was performed using a combination of GO terms and KEGG orthology and module terms. Group scores were defned as the mean of protein distances from the origin of the PCA biplot (PC1 vs. PC2). Statistical cutoffs (p-value < 0.05) were calculated using a bootstrapping approach that calculates scores for 100 000 groups randomly selected from the total pool of quantifed proteins. Cutoffs were calculated individually for each group size (n ¼ 4, 5, etc.) and groups with fewer than 4 members were excluded. Version 9.1 of the STRING database was used for identifying protein interactions, and interacting networks were identifed by manual inspection. 41

chemsum

{"title": "Time-resolved proteomic analysis of quorum sensing in Vibrio harveyi", "journal": "Royal Society of Chemistry (RSC)"}

a_novel_application_of_generation_model_in_foreseeing_'future'_reactions

## Abstract: Deep learning is widely used in chemistry and can rival human chemists in certain scenarios. Inspired by molecule generation in new drug discovery, we present a deep learning-based reaction generation approach to perform reaction generation with the Trans-VAE model in this study. To comprehend how exploratory and innovative the model is in reaction generation, we constructed the dataset by time-split. We applied the Michael addition reaction as the generation vehicle and took the reactions reported before a certain date as the training set and explored whether the model could generate reactions that were reported after the date. We took 2010 and 2015 as the time points for the splitting of the Michael addition reaction respectively. Among the generated reactions, 911 and 487 reactions were applied in the experiments after the respective split time points, accounting for 12.75% and 16.29% of all reported reactions after each time point. The generated results were in line with expectations and additionally generated a large quantity of new chemically feasible Michael addition reactions, which also demonstrated the learnability of the Trans-VAE model for reaction rules. Our research provides a reference for future novel reaction discovery using deep learning. ## Introduction Organic synthesis is one of the challenging processes in drug discovery, and the exploration of new organic reactions has always been a key stumbling block in the development of synthetic organic chemistry. 1,2 New reactions enrich synthetic routes in the chemistry and materials field. Conventionally, the majority of new reactions have been discovered by the "chemical intuition" of scientists, which was a complex task requiring sufficient luck. For instance, the product of the Diels-Alder reaction was known to chemists as early as 1906, but it was not until 1950 that the reaction was applied to total synthesis experiments. 3,4 The long and intricate progress of discovering new reactions intervenes the promotion of drug discovery. 5,6 Over the past few years, artificial intelligence (AI) technology has provided a number of important applications in various aspects of chemistry and has brought disruptive effects. Reaching or even surpassing human-level capability at combining chemical reactions with AI remains a new challenge with broad feasible applications. The exploration of AI in chemical reactions primarily involved reaction prediction, 16,17 retrosynthesis analysis, 18,19 reaction condition optimization, 20 and reaction classification, 21 etc. In principle, reaction prediction can be realized by extracting the rules of various chemical reactions, and then directly deriving the products related to reactants. The current mainstream methods usually treat reaction prediction tasks as similarity transformation of molecular graphs or text translation, and the corresponding models are graph convolutional neural network (GCN) and sequence-to-sequence models. 22,23 The performance of text-based reaction prediction has been significantly improved since the release of google company's transformer model, which is entirely based on the attention mechanism. The Molecular Transformer proposed by Schwaller et al, in which the molecules involved in a reaction are all represented as the Simplified Molecular Input Line Entry System (SMILES), was a state-of-the-art SMILES-based sequenceto-sequence model that could reach a 90.4% top-1 accuracy on the USPTO_MIT Data Set with separated reagents. 24 In addition to innovations in the model structure, many strategies can assist AI in better comprehending chemical reactions, including data augmentation 25 and transfer learning, 26 which have shown satisfactory functions in tackling low chemical data regimes. However, it is arduous to discover new reactions by automatically extracting rules from known chemical reactions. Inspired by molecular generation, which refers to the generation of undiscovered active or target molecules by extracting the characteristics from a set of molecules known to have specific biological activities, recent studies have turned their attention to generation models and put forward de novo reaction generation. In the task of molecular generation, many active molecules aiming at a specific target have been successfully generated, and some molecules are already in the clinical research stage, which provides a great reference for finding new reactions through generation models. The reactions generated by a reliable generation model can not only guide future chemical research, but also provide a wealth of reaction data to drive deep learning 2 / 10 models. However, a chemical reaction that implies the chemical transformations between reactants and products is a more intricate object for a computer than a pure chemical compound that contains only SMILES rules and information on structural properties. The first attempt at reaction generation was presented by Bort et al 27 . They constructed Bidirectional Long Short-Term Memory (LSTM) layers and trained on the database called USPTO. All reactions were modified from the original SMILES in the form of corresponding Condensed Graph Reaction (SMILES/CGR). Via visualizing latent variables as Generative Topographic Mapping (GTM), they located the position of Suzuki reaction, and found some reactions have particular structural motifs that were not present in the training data. But the new reactions generated in their work have not been proved by follow-up experiments. In subsequent studies by Wang et al, 28 the type of data set used for reaction generation was restricted to Heck reactions. And the Transformer XL, which is a fully attention-based model and more suitable for long sequences, was applied in their study. The result analysis proved that the generated reaction conforms to the Heck reaction rule, and the model also had a favorable grasp of deeper chemical knowledge such as site selectivity. They further selected some reactions for laboratory synthesis to verify the reliability of the generated reactions. Is there a simple and efficient way to test the reliability of the generation model and the novelty of generated reactions? We devised a scheme where the model is trained with chemical reactions published in journals prior to a certain time point to test if the model could produce reactions after that point in time. The schematic diagram of the method is shown in Figure 1. In this study, we used the classical Michael addition, which is a representative reaction for carbon chain growth, carbon ring formation, and heteroatom introduction in organic synthesis, as a reaction generation vehicle. And the Trans-VAE 29 model where both encoder and decoder are built with transformer is applied to accommodate the long sequence generation of the reaction. We imported the Michael addition reactions before a certain date into the model as the training input, and part of the reactions generated by our model have been verified by chemists in the literature after the date. The result proved the superiority of the model in certain aspects of reaction generation. More importantly, some generated reactions were brand-new Michael addition reactions that have never shown up in the literature and are valuable for confirming chemical feasibility. With the Michael addition reaction being successfully generated and supported by literature, it not only provides us with a simple and effective way to verify the chemical level generation model, but also sets the stage for generating new types of future reactions in our next work. ## Results and discussion The primary purpose of reaction generation is to generate reactions that can be used in future research. And secondly, it can also be used to expand the data volume of small data set reactions to break through the data volume bottleneck of deep learning technology in the field of chemistry. Obviously, it is more difficult to generate new reactions that meet the demand of researchers in the process of model generation. Take 2010 as the split time point, we utilized a total of 3,218 reactions to training the model and then generated 32,979 new Michael addition reactions, 911 of which were reported in the literature after 2010 and validated experimentally. Similarly, we divided the data with 2015 as the split date, and fed the 6,962 training reaction data before 2015 into the model, and finally generated 81,377 new reactions, 487 of which were applied in the literature after 2015. As listed in Table 1, we observe that generated reactions reported after 2010 accounted for 12.75% of all reactions reported after that date, and when 2015 is the split time point, the ratio is 16.29%. We also depict the variation of the ratio with the progress of reaction generation in supplementary Fig. S2, indicating that the model-generated reactions are very reliable and also provides a guarantee for the application of the rest of the new reactions to chemical research in the future. We randomly selected some model-generated Michael addition reactions, which were reported after 2015 in Figure 2, as well as some brand-new examples of Michael addition reactions. As shown in Figure 2, (a)-(c) are model-generated reactions that have been applied in practical studies, (d)-(f) are completely new reactions. These examples are consistent with the reaction characteristic rule of Michael addition reaction. On basis of the reaction generated from the dataset with 2015 as split-date, we would evaluate the quality of the model generation in terms of the distribution and similarity of the generated reactions to the training reactions and the chemical properties. Because a complete reaction includes reactants and products and the chemical rule between them, it is necessary to compare the component relationships between the training set and the generated set. As listed in Table 2, we counted the types of Michael acceptors and donors and products in the training set and generated set. To visually represent the distribution between the generated set and the training set, we used the t-SNE 30 (t-distributed stochastic neighbor embedding) method to visualize the molecular Morgan fingerprints 31 and further verified the validity of the generated molecules. t-SNE is a dimensionality reduction visualization technique that creates a dimensionality-reduced feature space where similar data points 3 / 10 in a high-dimensional space map to similar distances in a low-dimensional space, and their distributions also remain similar. Morgan molecular fingerprints is a circular topological fingerprint, which obtained by adapting the standard Morgan algorithm. In contrast to MACCS which depends on predefined molecular features to be matched, Morgan molecular fingerprints is a systematic exploration of atomic types and molecular connectivity by searching all the substructures in the compound for a given step by a search algorithm. Figure 3A, 3B, and 3C show the t-SNE plots of the Michael donor, acceptor, and product in the generated set with the Morgan fingerprints of the corresponding reactants in the training set, respectively. It can be seen from the plots that the training set molecules overlap well with the corresponding generated set, which indicates that both the reactant and product molecules generated by the model varied around the training set with a certain novelty and also fit the distribution of the training set. Shifting the gaze to the overall reaction level, the process of combining the corresponding reactants and product molecules into a reaction means that the model must learn the Michael addition reaction rule. Despite that the Michael addition reaction is one of the most widely used catalytic carbon-carbon bond forming tools in organic synthesis, its rule is complicated for the Trans-VAE model. To further demonstrate that the reactions generated by the model belong to Michael addition reactions, we utilized TMAP to visualize the reaction fingerprint (rxnfp) of the reactions. The reaction fingerprint is derived from the reaction representation learned by the Bidirectional Encoder Representations from Transformers (BERT) model, which used unsupervised learning to construct a reaction space in a large database consisting of unannotated chemical reaction SMILES, and fine-tuned on limited labeled data to construct an accurate reaction classifier. 32 TMAP is a method to visualize the highdimensional space as a tree diagram. 33 As shown in Figure 3E, TMAP connects reactions in the generated reactions(10,000 reactions randomly selected from the generated set) and training dataset based on rxnfp similarity, with each reaction represented as a point in the tree diagram. In addition, the USPTO-50K which contains ten major classes of chemical reactions was downloaded and curated by Liu et al 34 were used to form the backbone of chemical space. Furthermore, we used UMAP 35 to reduce the dimensionality of the rxnfp to validate the distribution of the training set and generated set (Figure 3D). It turns out that the model grasps and reproduces the reaction rules in the training set relatively satisfactorily. Michael addition reaction, which can effectively build various carbon-carbon or carbon-hetero bonds, has good compatibility with various functional groups. It can also be applied to the preparation of complex compounds and has very important practical value. To explore in detail whether our model fully understand the Michael addition reaction, we perform an in-depth analysis of the generated Michael addition reaction set. Firstly, we divide the Michael addition reaction into intermolecular and intramolecular reactions. If a molecule contains both donor and acceptor functional groups, intramolecular reactions may occur to constitute carbon rings or heterocycles. As listed in Table3, there are 6,707 intermolecular reactions and 255 intramolecular reactions in training dataset. As for generated reactions, intermolecular reaction accounts for 99.6%, which is consistent with the distribution of intermolecular reactions in training set. As shown in supplementary Fig. S3, we listed several representative examples of intermolecular reactions and intramolecular reactions from training and generated datasets. Because the Michael addition reaction is reversible and the thermodynamically most stable product usually predominates. Five-and six-membered rings are usually more stable due to the lower ring tension Our model accurately captures this feature and the intramolecular reaction of Michael addition is mainly used for the synthesis of more stable fiveor six-membered rings. Besides alkene Michael acceptors, electron-deficient alkynes conjugated with electron-withdrawing groups can also be used as Michael acceptors, although alkynes are less reactive than alkene Michael acceptors. Table 4 shows the distribution of Michael acceptors types, mainly divided into alkene acceptors and alkyne acceptors. We select several alkyne Michael addition reactions from the training and generated sets and displayed them in supplementary Fig. S4. In the Michael addition reaction, when the enolate as the Michael donor is formed from a simple carbonyl compound under the action of a base, an important feature of the reaction is that the stereo structure of the product is closely related to that of the enolate. As listed in supplementary Fig. S5A, there are two examples from training dataset, where the Z-enolate creates anti-product and E-enolate creates syn-product. It is confirmed that the model has perceived this rule, and the generated reactions follow it satisfactorily as depicted in supplementary Fig. S5A. In the case of simple carbonyl compounds with asymmetric Michael donors, the acceptor reacts mainly with α-carbon atoms having more substituents, depending on the stability of the intermediate enol. In general, the more electron-donating substituents on the double bond, the more stable the enol is and the more Michael addition reaction is promoted. A typical instance is shown in supplementary Fig. S5B. For stable carbanion conjugated to multiple heteroatoms, reactions with the acceptor typically yield 1-4 addition products. Most of these heteroatom-containing stable groups are easy to leave and can be considered as conjugated auxiliary groups. We also list some instances in supplementary Fig. S6 where we can see that the carbon atom in the middle of the two carbonyl 4 / 10 groups is more acidic than the carbon atom on the other side of the carbonyl group, so it is more likely to be deprotonated by a base to form a carbanion. The reactions generated by our model also fit this signature. The molecular structure of Michael acceptor includes an electron-withdrawing group and an unsaturated system. Almost all alkene compounds substituted with electron-withdrawing group can be utilized as Michael acceptors as exhibited in supplementary Fig. S7. However, if the acceptor molecule contains two or more electron-withdrawing groups at the same time, the regioselectivity of the reaction is usually controlled by the relatively active group. supplementary Fig. S7A2 is a generated reaction example that conforms to this rule. We can see that the Michael donor in supplementary Fig. S7A2 has two electronabsorbing groups, nitro, and cyano. Since nitro is more electron-absorbing, the carbon atom to which nitro is attached is more likely to lose a proton to form a carbanion. The model learns this principle during training and reflects it in the generated reaction. It is worth mentioning that, in addition to carbon nucleophiles, some heteroatom groups can also be used as donors for the Michael addition reaction due to their nucleophilic properties. For example, alkylamines or arylamines are widely used as Michael donors. The reaction has promising chemical selectivity and generally does not generate imine by-products. We have further added reaction data for the heteroatom Michael addition reaction to the data with 2015 as the split date. After retraining the model on the data, we observe whether the model still grasps the reaction Michael addition under the obfuscation of heteroatoms. In this study, we mainly consider heteroatoms such as N, S, O. Table 5 is the classification and proportion of heteroatom nucleophiles. It could be seen that the generated carbon nucleophiles, nitrogen nucleophiles, oxygen nucleophiles, and sulfur nucleophiles occupied most of the generated reactants, which is similar to the distribution of these four reactants in the training dataset. We present examples of Michael additions involving heteroatoms in the training and generated sets in supplementary Fig. S8, respectively. These results are exciting as it proves that our Trans-VAE model is sufficiently expressive to produce the correct reactions. ## Conclusion In this work, we applied the Trans-VAE model for the reaction generation task. To explore whether the model can 'break the time limit' and generate Michael addition reactions that have been applied in reports after a certain date, we simulate this scenario by dividing the dataset by 'time-split'. Thanks to the transformer-based encoder and decoder architecture, the model can capture both SMILES rules and Michael addition reaction feature information in the long sequence of reactions. We used 2010 as the split time point, and trained the model on the reactions before 2010. The results showed that the model generated reactions that were applied after 2010, accounting for 12.75% of all published reactions after that date. providing initial evidence of the reliability of the Trans-VAE model in reactions generation. To demonstrate the effectiveness rather than haphazardness of our model, we conducted another experiment with 2015 as the split time point, and the rate was 16.29%. We then further inspected whether the model mastered the rules of Michael addition reactions by analyzing the generated Michael addition reactions in terms of their chemical characteristics. The final analysis shows that the model captures reaction characteristics consistent with the now discovered chemical laws of Michael addition reactions, indicating the reliability of applying deep learning models to reaction generation, and laying the foundation for our subsequent exploration of the vast chemical space using deep learning models and the discovery of the generation of completely new types of chemical reactions. ## Methods Dataset. The reaction generation model was trained on SMILES files containing only Michael addition reactions that were extracted from the "Reaxys" database based on a search of reaction templates and/or reaction names (all entries using the "Michael addition reaction" phrase). The extracted Excel files were subjected to a pre-processing process with a series of python scripts to obtain a high-quality dataset that met the requirements for new reactions generation. In this step, reactions where the SMILES string was invalid or the reactants and products were identical were removed from the file, and the remaining reactions were canonized using RDkit 36 so that the same compound was represented by the same SMILES. Finally, the non-compliant reactions were filtered based on the Michael addition reaction template using RDkit's Python script. As for the time point of the reaction, it was considered that the same reaction may be reported in the literature at different times, we took the time when it was first reported and deleted the rest of the same reactions to obtain a dataset containing 12,322 Michael addition reactions. Taking 2015 as the split line, the reactions before 2015 were divided into training and validation sets (9:1), while those after this time were used as a reference for whether the model could generate 'future' reactions. ## / 10 Model. With the rapid development of natural language processing (NLP) models, the text-based format has been widely used in previous works, such as chemical reaction prediction and ADMET prediction. We represented the reactions in the form of Simplified Molecular Input Line Entry System (SMILES) strings 37 , in which every character corresponds to an atom or chemical bond. The regularized expression which is arbitrarily extensible with reaction information is adopted as the rule for SMILES word segmentation. Based on this rule, a chemical atom with multiple characters and special environments (e.g 'Cl', '[O-]') is treated as one token instead of being divided into multiple tokens, which is more in line with chemical specifications. 38 The fundamental architecture of Trans-VAE consists of an encoder and a decoder, and the workflow of the model is displayed in Figure 1B. The encoder maps the discrete SMILES to a dense latent representation and transforms it into a continuous fixed-dimensional vector, while the decoder attempts to convert the vector in the latent space back into input with the smallest possible error. By adding noise to the encoded SMILES, molecules would have corresponding probability distribution in the latent space rather than individual points, and the decoder also learns to discover more robust representations from latent points. The training process intends to minimize the reconstruction loss between original SMILES and generative SMILES, while satisfying the probability distribution of the generated data is similar to that of the training data. Consider the fact that the SMILES representation of the reaction has increased by two to three times compared to the molecule, which calls for the model to have salient performance for long sequences. Therefore, we applied the VAE model proposed by Dollar et al 30 which implements the transformer as the encoder and decoder. Compared with the recent commonly built Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) as encoder and decoder, it replaces the recurrence with the attention head entirely, which reduces the spreading path of information in the model and helps to grasp the longrange syntactic dependence in reaction SMILES. However, the highly parallelizable transformer also brings a lot of trainable parameters. To avoid parameter explosion, a convolution bottleneck is employed to stochastically compressed the encoder output and then feed it to the decoder. The purpose of our work is to explore whether the Trans-VAE model can generate reactions reported afterward based on reactions before 2015, that is pay more attention to the exploration. Therefore, we give priority to sampling within the highentropy dimension, where all of the meaningful structural information is involved. Different from random sampling, highentropy sampling is better able to explore chemical phase space and obtain novel reactions, albeit may reduce valid SMILES.

chemsum

{"title": "A novel application of generation model in foreseeing 'future' reactions", "journal": "ChemRxiv"}

molecular_recognition_of_<i>n</i>-acetyltryptophan_enantiomers_by_β-cyclodextrin

## Abstract: The enantioselectivity of β-cyclodextrin (β-CD) towards L-and D-N-acetyltryptophan (NAcTrp) has been studied in aqueous solution and the crystalline state. NMR studies in solution show that β-CD forms complexes of very similar but not identical geometry with both L-and D-NAcTrp and exhibits stronger binding with L-NAcTrp. In the crystalline state, only β-CD-L-NAcTrp crystallizes readily from aqueous solutions as a dimeric complex (two hosts enclosing two guest molecules). In contrast, crystals of the complex β-CD-D-NAcTrp were never obtained, although numerous conditions were tried. In aqueous solution, the orientation of the guest in both complexes is different than in the β-CD-L-NAcTrp complex in the crystal. Overall, the study shows that subtle differences observed between the β-CD-L,D-NAcTrp complexes in aqueous solution are magnified at the onset of crystallization, as a consequence of accumulation of many soft host-guest interactions and of the imposed crystallographic order, thus resulting in very dissimilar propensity of each enantiomer to produce crystals with β-CD. ## Introduction Cyclodextrins (CDs) are cyclic, water-soluble carbohydrates with a rather non-polar cavity that can host a variety of organic molecules (guests) and form inclusion complexes . The guest molecules may be completely or partly enclosed inside the cavity depending on their size and the CD macrocycle's dimensions. The host-guest interactions established in the cavity are of van der Waals type, whereas between parts of the guest extending out of the cavity and the host's hydroxy groups are H-bonding interactions and/or of electrostatic nature. CDs have been studied and used for the enhancement of solubility, bioavailability and stability of drugs . Moreover, being oligomers of α-D-glucopyranose, CDs possess an intrinsic chirality, thus they form diastereomeric inclusion complexes with enantiomeric pairs and frequently they exhibit enantioselectivity in aqueous solution or they can co-precipitate with only one enantiomer (enantioseparation). The separation of enantiomers via cyclodextrin inclusion is particularly important in the case of guests of pharmaceutical interest, since enantiomerically pure drugs are crucial for the pharmaceutical industry . It has been proven difficult so far to explain and to predict the recognition abilities of specific CDs towards enantiomers, especially in solution. An interesting attempt is a thermodynamic study in aqueous solution with microcalorimetry of a large number (43) and variety of chiral organic compounds with β-CD at room temperature. It was shown that properties and interactions important for chiral recognition include (i) weak non-bonding interactions rather than polar, (ii) nonsymmetrical non-polar penetrating guests and (iii) large distance of the chiral center from charged/hydrophilic groups. Moreover, trends in enantioselectivity do not follow trends in association constants, i.e., the association constants for the β-CD complexes of both enantiomers of N-acetyltyrosine, N-acetylphenylalanine and N-acetyltryptophan are in decreasing order, whereas their enantioselectivity (ratio of the binding constants, K, of the L-to the D-enantiomer) shows an increasing order (1.04, 1.1 and 1.34, respectively). X-ray crystallography, on the other hand, can improve our understanding of chiral recognition by CDs at the atomic level by providing insight into the interactions and the fit of the guest in the cavity, taking into account that crystal lattice forces may introduce additional and more stringent parameters for the enantiodiscrimination . However, the crystallographic structures of diastereomeric complexes of CDs with chiral guest molecules in the literature are scarce. For β-CD with fenoprofen , a partial chiral resolution of the racemic mixture occurs, since the obtained crystals contain discrete β-CD dimers enclosing (R)-or (S)-enantiomers in a S/R ratio = 3:1. The enantiomers adopt different orientations in the β-CD dimers and preference of the (S) complex is dictated both by stronger H-bonding of the carboxyl group, as well as more favorable methyl-phenyl interactions inside the cavity. In contrast, no discrimination is shown by β-CD for (R)-and (S)flurbiprofen , since the crystals grown from the racemic mixture have both enantiomers enclosed (as a head-to-head dimer) in a β-CD dimer. In the case of substituted CDs, 2,3,6tri-O-methyl-α-CD discriminates between (R)-and (S)-mandelic acid as it forms very different crystals from a racemic mixture. The same host crystallizes exclusively with (R)-(−)-1,7dioxaspiro [5.5]undecane, the Dacus Oleae pheromone, from an aqueous solution of the racemic mixture (enantioseparation) also exhibiting high enantioselectivity in solution. Likewise, heptakis-(2,3,6-tri-O-methyl)-β-CD displays high enantio-selectivity in solution towards (S)-(+)-1,7-dioxaspiro [5.5]undecane and under certain conditions it co-crystallizes only with the (S)-enantiomer . Induced host-guest fit, made possible by the macrocyclic flexibility of the permethylated CDs plays a crucial role in their capacity for chiral discrimination. Chiral recognition of amino acids and their derivatives by CDs has been also tested using phase-solubility diagrams , NMR spectroscopy and electrochemical methods , as well as by X-ray crystallography . Detailed structures of β-CD with L-and D-N-acetylphenylalanine (NAcPhe) grown separately has shown that although the two complexes are isomorphous (same space group, very similar unit cell dimensions and same packing of β-CD dimers) there are differences regarding the positioning of the guest molecules, the D-enantiomer being ordered, whereas the L-enantiomer extensively disordered. This disparity seems to be determined by subtle hydrophobic differences and H-bonding interactions among guests themselves and with the host and co-crystallized water molecules in the lattice. Additional structures of β-CD with different L-phenylalanine derivatives confirm the above general result. In the present study, we report on the inclusion of the Land D-enantiomers of N-acetyltryptophan (NAcTrp) in β-CD (Scheme 1) in an effort to contribute to the study of chiral recognition of amino acid derivatives by CDs in the crystalline state and in solution. The guest NAcTrp has been selected because of its large aromatic side chain with appropriate dimensions to fit tightly in the β-CD cavity thus expected to have restricted mobility and limited disorder. Indicative of the interest and possible applications of the CD use in chiral selectivity/discrimination of tryptophan are studies in aqueous solution , in electrochemistry for sensor development , as components of solid phases in chromatography , or in capillary electrophoresis . ## NMR studies In deuterium oxide (D 2 O), each of the NAcTrp enantiomers induced significant chemical shift displacements (shielding) in the 1 H NMR signals of the β-CD cavity protons, namely H3 (near the wider, secondary side) and H5, H6,6' (at the narrower, primary side), signifying cavity inclusion of each enantiomer (Scheme 1). When a racemic mixture of NAcTrp was added to a β-CD solution no differentiation in the signals was observed due to in situ formation of diastereomers, except for a very small splitting of the methyl signal of the N-acetyl group. No differentiation was detected in the 13 C NMR spectrum either. In order to determine the stoichiometry of the complexes continuous variation (Job) plots were drafted. For β-CD protons only the cavity signals due to H5, H3 and H6,6' showed significant shifts upon complexation (Supporting Information File 1, Figure S1). The inflection point of the graphs at 0.5 indicates a 1:1 stoichiometry for both enantiomers. The tryptophan protons were affected differently upon complexation (Supporting Information File 1, Figure S2), i.e., the graphs due to shifts of the indole's benzene ring protons (H3, H4, H5 and H6) indicate a 1:1 host/guest stoichiometry, whereas those of the indole moiety (H8) and of the aliphatic protons (H9,9', H10, H12), with an inflection point at ≈0.3, suggest a host/guest ratio close to 2:1. This behavior reveals the existence of two different complexation modes, one involving the indole phenyl ring with one host only and the aliphatic chain with two host molecules. The fact that the second mode takes place mainly when there is an excess of host concentration indicates that the inclusion of the indole moiety is the predominant mode of interaction. Moreover, it is observed that the magnitude of the shifts of the L-enantiomer are always larger and the slopes of the Job plots steeper than those of the D-enantiomer, suggesting stronger binding of β-CD with L-than with D-NAcTrp. 2D ROESY spectra of each enantiomer with β-CD at a 1:1 mole ratio in D 2 O were obtained under identical conditions (temperature, concentration, acquisition parameters). Strong intermolecular dipolar interactions were observed between indole protons (H3, H4, H5, H6, H8) and the β-CD cavity protons (H5, H6,6', H3) in both enantiomeric guests, confirming full inclusion of the Trp side chain. To facilitate the comparison of the two in situ formed diastereomers and to visualize the small differences (Supporting Information File 1, Figure S3) in dipolar through space intermolecular interactions in each case, 3D correlation maps were employed. They were displayed carefully so as to ensure the same intensity for the reference intramolecular correlations between NAcTrp H9,9' with H6 (average distance ≈3.5 ) and with H8 (average distance ≈4.0 ) in each of the enantiomers (Figure 1a), enlarging the points of difference in the magnified maps (Figure 1b). Thus (i) guest-H6,H3,H5-host-H5,H66',H3 interactions are very similar in both enantiomers with guest-H5-host-H3 clearly weaker than the others, and guest-H6/host-H6,6' stronger in L-than in D-, suggesting that guest-H6,H3 are embedded inside the cavity, guest-H5 is closer to the narrow β-CD rim and L-H6 is closer to it than D-H6. (ii) Guest-H8-host-H3 interactions are equally strong in both enantiomers, stronger than the guest-H8/host-H5,H6,6' ones, which in turn are stronger in L-than in D-. Moreover, interactions between guest-Me12-host-H3 are strong for both enantiomers (Supporting Information File 1, Figure S3b), suggesting that the N-acetyl group is in both cases at the wide secondary opening of β-CD, and L-H8, is closer to the primary opening than D-H8 suggesting a difference in tilting. (iii) Guest-H4-host-H5 interactions are similar in both enantiomers but this of guest-H4-host-H6,6' is considerably stronger in D-than in L-, while guest-H4-host-H3 interactions are practically absent for both enantiomers, implying that L-H4 is extended further out of the primary side than D-H4. (iv) Guest-H9,9' and H10 show weak interactions with host-H3 thus they reside mostly closer to the wide opening of the host. The above interactions detected by NMR suggest that in the aqueous environment the inclusion modes in each diastereomeric complex are very similar but non-identical. D-H4 is located near the primary side of the host, while L-H4 is completely outside (scarcely communicates with the cavity). On the other hand H8 (at a ≈7 distance from H4) is at the secondary side in both enantiomers, slightly closer to H5 of the host only in the L-enantiomer. These interactions suggest a common binding model, with the indole part included in the direction H4 to H8 from primary to secondary opening and with the L-enantiomer having its H4 end exposed and its NAc group at the secondary side in contact with CD-H3. A different degree of tilting with respect to the β-CD axis to accommodate the hydrophobic NAc group in the cavity is inferred by the NMR data in each case, thus explaining the small differences observed in solution. However, as the Job plots suggested, the aliphatic part is influenced by a second host molecule presumably via its secondary side. This implies that in solution, host-guest association is possible through additional orientations and stoichiometry, thus the presence of alternative arrangements in low percentage cannot be excluded. ## X-ray crystallography studies In the crystalline state, the structure of the inclusion complex of L-NAcTrp in β-CD comprises dimers. The asymmetric unit of the complex contains two crystallographically independent β-CD hosts (A and B) forming a dimer (Figure 2), in which two guest molecules of L-NAcTrp are enclosed in a head-to-head fashion (host:guest ratio, 1:1). The pair of L-NAcTrp molecules inside the dimer are found in orientational disorder, i.e., the guest exhibits a major orientation, molecules C and D (occupancy 65%), and a co-existing minor orientation (molecules E and F, occupancy 35%) in a statistical fashion. The dimers pack along the axis a at an angle of 19° thus forming a broken channel (Intermediate packing) . The mean distance of the centers of mass of two consecutive β-CD dimers is 5.78 . Co-crystallized with each dimer, 21.45 water molecules are found distributed over 36 sites. The water molecules form the usual water networks of H-bonds, one linking the primary and the other the secondary hydroxy groups , many of them stabilizing the crystal lattice (structural water molecules). The glucopyranose residues (in 4 C 1 chair) of both A and B β-CD have a rather undistorted conformation (Supporting Information File 1, Table S1) (angles between the glycosidic oxygen atoms O-4n similar to these of the regular heptagon, 128.57°, deviations of the O-4n atoms from their mean plane, close to zero). The tilt of the mean glucopyranose planes towards their 7-fold axis are small and close to their average values (7.1 and 7.7°, respectively). As in all β-CD dimeric complexes , the macrocycles' conformation is stabilized by hydrogen bonds connecting (i) intramolecularly, the O-3n and O-2(n+1) atoms of neighboring glucopyranose units (mean 2.73 and 2.75 for A and B, respectively, 2.78 in hydrated β-CD) and (ii) intermolecularly, the O-3nA and O-(8−n)B atoms of monomers A and B, respectively (range of distances 2.7-2.8 , Supporting Information File 1, Table S2). At the primary side, only β-CD molecule B exhibits disorder of the C-Ο63Β bond in two conformations, the major (−)-gauche C-Ο63Βa (occupancy 78%) pointing outward and the minor (+)-gauche C-Ο63Βb (22%) pointing towards the interior of the cavity, the latter interacting with guests C and D of neighboring dimers (Figure 3a). The aromatic moieties of both guest orientations maintain the same relative position with the host, their planes interacting in a π•••π fashion (Figure 2 and Figure 3) (dihedral angle between D, F) in cavity B are close to the secondary hydroxy level, apparently in order to optimize the π•••π interactions between the indole planes (Figure 2 and Figure 3, Table 1). The above suggest a tight fit of the guest inside the cavity. On the other hand, the aliphatic part of NAcTrp, positioned in the space between dimers, exhibits more freedom: the carboxylic and acetylamino groups of guests D and F inside β-CD B are close and parallel, whereas in β-CD monomer A the acetylamino moiety of the major guest C is close to the carboxyl group of minor guest E, their respective carboxyl and acetylamino groups pointing to opposite directions (Figure 2 and Figure 3). These differences maximize the strong interactions between major guests C and D (Figure 3, Table 1). Numerous trials to crystallize the inclusion complex of β-CD with D-NAcTrp have failed to give anything but hydrated β-CD crystals , as described in detail in the experimental section, however, some crystals were grown after hydrothermal treatment of the solution (65 °C for duration of 6 days) . The structure of the latter could not be solved by isomorphous replacement (using the coordinates of β-CD-glutaric acid complex , that is isomorphous to hydrated β-CD . This was an indication that the structure should be quite different from hydrated β-CD. However, no guest could be located during the refinement and the present structure (henceforth "β-CD-D-Table 1: H-bond distances of β-CD-L-NAcTrp complex: (1) between guest molecules themselves and with the host (2) with water molecules, (3) between structural water molecules and the host. NAcTrp") was refined as a β-CD-water complex (Table 2). "β-CD-D-NAcTrp" exhibits the "herringbone" packing of the β-CD monomers (Figure 4) as the hydrated β-CD structures reported so far [29, , as well as several monomeric β-CD complexes . The conformation of the β-CD macrocycle (Supporting Information File 1, Table S3) is similar to the monomeric β-CD structures , but more distorted than in the dimeric β-CD-L-NAcTrp complex: The glucopyranose residues adopt the regular 4 C 1 chair conformation, but the angles be-tween them deviate from the angle of the regular heptagon and the tilt of their average planes towards the 7-fold β-CD axis varies between 5.0 and 25.8°. At the primary side, two hydroxy groups (O61 and O65) point towards the interior of the cavity and two exhibit two-way disorder of the C-Ο63 and C-Ο67 bonds. Comparison of the "β-CD-D-NAcTrp" structure to this of hydrated β-CD pinpoints the difficulty of solving the struc- ture as an isomorph. It can be seen (after the appropriate transformation of coordinates due to different origin and axes; Supporting Information File 1, Figures S4 and S5) that the hydrated β-CD macrocycle does not superpose exactly in the lattice of "β-CD-D-NAcTrp", which may render the two structures not quite isomorphous. It is worth noting that many of the hydrated β-CD structures [29, , as well as several monomeric β-CD complexes are determined in lattices with different origin or interchanged crystallographic axes or even inverse coordinates (Supporting Information File 1, Figure S4). Further, by superposition of one glucopyranose unit of "β-CD-D-NAcTrp" to the equivalent unit of hydrated β-CD the difference in coordinates of the two structures is more apparent (Supporting Information File 1, Figure S6). In contrast, the same kind of superposition applied to monomeric structures mentioned above shows that they superpose completely on hydrated β-CD. Although the NMR results have shown that β-CD forms complexes with both L-and D-NAcTrp in aqueous solution at room temperature, it was not possible to crystallize the β-CD-D-NAcTrp complex. In contrast, the β-CD complexes of both enantiomers of N-acetylphenylalanine (NAcPhe) have been determined and they are isomorphous with β-CD-L-NAcTrp. Although the isomorphous complexes of L-NAcPhe and D-NAcPhe exhibit identical packing of the β-CD dimers, the relative stability of the guest molecules enclosed in them is controlled by subtle changes in the guest positioning. L-NAcPhe is highly disordered even at 20 K probably due to very weak nonpolar and polar interactions, whereas D-NAcPhe is highly ordered, although the non-polar interactions between the phenyl moieties are also weak. Its stability is gained by the N-acetyl group of one D-NAcPhe guest, which rotates and "hides" inside the dimer cavity (probably because of unfavourable exposure to the aqueous environment). Similarly, β-CD-L-NAcTrp is also more stable than β-CD-L-NAcPhe due to the larger side chain of the guest. L-NAcPhe is shorter than in L-NAcTrp, which has two consequences for the stability of the complex (a) no strong π•••π interactions at 3.5 can be established in the middle of the β-CD dimer as in L-NAcTrp (Figure 5); (b) the aliphatic moieties of β-CD-L-NAcPhe protruding from the primary sides between dimers do not interact directly or even indirectly via β-CD hydroxy groups along the channels, as in the L-NAcTrp complex. Modeling the possibility of formation of a dimer β-CD-D-NAcTrp complex by energy minimization of the interactions of D-NAcTrp inside the β-CD dimer (as determined in the β-CD-L-NAcTrp structure) revealed a complex similar to β-CD-L-NAcTrp (Supporting Information File 1, Figure S7). The positioning of the D-indole groups is very similar to these of the L-enantiomer (closest distance 3.46 between the aromatic planes). The approaching aliphatic moieties between two β-CD dimers along the channel could be stabilized possibly by an inward pointing hydroxy group Ο63Βb of β-CD (assuming that the β-CD host remains unchanged), which H-bonds to the carboxylic oxygen atom of the D guest and the acetyl O1 atom of the C guest, however, the acetyl methyl group of C is exposed to the water environment. "Hiding" of the latter group inside the cavity, as in the case of the β-CD-D-NAcPhe complex, is not possible due to the bulkier indole group of D-NAcTrp that fills the cavity. This unfavorable environment might be a factor that forbids the formation of a β-CD-D-NAcTrp dimer structure. The difficulty in crystallizing the β-CD-D-NAcTrp may arise from a higher free energy barrier of crystal nucleation compared to other competing processes in solution at room temperature, but under the higher temperature and pressure conditions of the hydrothermal cell the presence of D-NAcTrp or of the complex β-CD-D-NAcTrp may influence the initial crystal nuclei which eventually lead to the grown crystals and differentiates them slight from hydrated β-CD. It is worth noting that hydrothermal treatment in crystallization trials has yielded uncommon structures such as, novel packing of β-CD-ethanol crystals during trials to crystallize the β-CD-N-(1adamantyl)salicylaldimine complex in ethanol, novel association of β-CD monomers in structures of β-CD complexes, e.g., with 4-pyridinealdazine , polyethylene glycol or adamantane . ## Conclusion This work has been focused on the ability of β-CD to discriminate between the enantiomers of N-acetyltryptophan. NMR studies in aqueous solution show that both enantiomers form similar, but not identical complexes with β-CD. L-NAcTrp induces larger shifts of β-CD cavity protons, suggesting stronger binding. For both enantiomers the prevailing complexation mode involves insertion in the cavity with the N-acetyl group in the secondary side and the indole moiety exiting the primary side, more exposed in L-than in D-NAcTrp. The tendency of the N-acetyl group to hide in the cavity is considered as the major cause for the differences between the two complexes that also results in somewhat folded NAcTrp structures, compared to the conformation observed in the crystal. In addition, both complexes are in contact with a second β-CD molecule suggesting presence of higher stoichiometries and possibility of different inclusion modes at low concentration. Overall, the orientation of both enantiomeric guests with respect to the macrocycle in the solution structures is opposite to the orientation of L-NAcTrp in ther crystal. On the other hand, only the complex β-CD-L-NAcTrp crystallizes readily forming a dimeric complex (two host and two guest molecules) packed in broken channels, isomorphous to the known β-CD complexes of the NAcPhe enantiomers. Numerous crystallization trials failed to produce crystals of the β-CD-D-NAcTrp complex yielding only hydrated β-CD crystals. The fact that β-CD-D-NAcTrp could not be crystallized in dimers as the β-CD-L-NAcTrp might be due to destabilization of the interface between dimers, because of exposure of the acetyl group to the water environment of the exterior and the inability to "hide" in the cavity, due to the bulky indole group occupying it. Trials to employ more energetic crystallization conditions resulted in crystals of a slightly different structure than hydrated β-CD crystals. The disagreement between solution and crystal structure in terms of complex formation and orientation/conformation of the guest indicates that the lattice forces and organization in the crystal prevail by far over the soft host-guest contacts established in solution and determines the final orientation of the guest inside the host and the formation of the crystals per se. ## Experimental Materials and methods N-Acetyl-L-tryptophan (L-NAcTrp), N-acetyl-D-tryptophan (D-NAcTrp) and β-CD were obtained from Sigma-Aldrich. Deuterium oxide was a product of Deutero GmbH. ## NMR spectroscopy The spectra were carried out on a 500 MHz Bruker Avance instrument at 300 K using a BBI probe, the library pulse sequences and 300 ms mixing time for the 2D ROESY runs. The compounds were dissolved in unbuffered D 2 O. The data was processed with Topspin. ## X-ray crystallography Crystallisation of β-CD-L-NAcTrp. In an aqueous solution of β-CD (6 mM) an equimolar quantity of L-NAcTrp was added and stirred for an hour until the solution became clear, which indicated formation of a complex. Then the solution was placed in an incubator at 23 °C, where by slow evaporation of the solvent, single crystals appropriate for X-ray data collection were obtained. The crystals had a diamond shape and a slightly pink color. Crystallisation trials of β-CD-D-NAcTrp. Trials to crystallize the complex of β-CD with D-NAcTrp under various conditions, including the above, did not result to single crystals of the complex. D-NAcTrp in the presence of β-CD (6 mΜ) at 50-60 °C, required a small quantity of ethanol in order to obtain a clear solution, from which crystals of hydrated native β-CD precipitated. This was proved from data collection from several crystals and structure determination based on isomorphous replacement using the coordinates of the β-CD-glutaric acid complex , which is isomorphous to hydrated β-CD . Use of racemic mixtures of NAcTrp produced also native β-CD crystals. However, use of a hydrothermal cell , in which β-CD (0.050 mM) and D-NAcTrp (0.025 mM) were placed in 2 mL of water and left at 65 °C for 5-7 days, produced crystals that could not be refined by isomorphous replacement using the coordinates of hydrated β-CD or other isomorphous crystals, as above. Structure determination. Low temperature X-ray data were collected at synchrotron radiation light sources. A single crystal, covered with a drop of paraffin oil, was mounted on a hair fiber loop and was instantly frozen to 100 K. Crystal data and analysis details are given in Table 2. β-CD-L-NAcTrp. Data of the β-CD-L-NAcTrp complex were collected at the beamline X13 of EMBL at DESY, Hamburg, by the oscillation method using a CCD of 165 mm radius detector. The DENZO and SCALEPACK software were used for data processing and scaling, respectively. The unit cell parameters and their esds were determined by the least square method from the high resolution frames of the collected data. The structure was solved by the isomorphous replacement method using the host coordinates of the β-CD-1,12-dodecanodioic acid complex . The structure solution and the refinement were carried out with the SHELXL97 program . The coordinates of the guest and solvent atoms were determined by successive cycles of difference maps and refinement. The non-hydrogen β-CD atoms and the oxygen atoms of the co-crystallized water molecules were treated anisotropically. Hydrogen atoms were placed at idealized positions and refined by the riding model (UH = 1.25 UC). The refinement of the structure, by full matrix least squares, converged to R1 = 0.0609, wR2 = 0.1663 and Goodness-of-fit = 1.076, for Fo > 4σ(Fo). Refinement details appear in CCDC 1531988. The structures were rendered in PyMOL . "β-CD-D-NAcTrp". Diffraction data were collected at the X06DA beamline, Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland. The XDS software package was used to reduce data and determine the unit cell parameters and space group, which were the same as hydrated β-CD. Trials to use isomorphous replacement (using the coordinates of β-CD-glutaric acid complex , which is isomorphous to hydrated β-CD , to refine the structure was unsuccessful (vide supra). The structure was solved finally by molecular replacement methods using the β-CD-glutaric acid complex coordinates. The refinement was carried out with the same strategy as in β-CD-L-NAcTrp complex. Early in the refine-ment numerous peaks appeared mainly at the primary hydroxy side of the cavity. Some were at bonding distances with each other, but by introducing the strongest of them as water molecules into the refinement did not result in a model of the guest (Table 2). Refinement details appear in CCDC 1531987. The structures were rendered in PyMOL . ## Molecular modeling The molecular models of D-NAcTrp complexes were based (a) on the geometry of the major orientation of β-CD-L-NAcTrp by changing the chirality of the Cα atom and (b) on β-CD non-hydrogen atoms of the corresponding lattice. To relieve steric clashes, restrained energy minimization of D-NAcTrp have been performed, while non-hydrogen atoms of β-CD are kept fixed in space. The XLEaP module of the AMBER 16 suite was used and the GAFF parameters were applied to the β-CD molecules with AM1-BCC atomic charges using the Antechamber module , while the ff99SB parameters were employed for NAcTrp. Restraint energy minimizations in implicit solvent were performed for 1,000 steps using a pairwise generalized Born model , while all β-CD nonhydrogen atoms were kept fixed in space using harmonic restraints of 10 kcal/mol 2 . For the "β-CD-D-NAcTrp" complex, the indole moiety was placed inside the β-CD cavity with the aliphatic part protruding from its primary side towards the empty space formed by three neighboring β-CD monomers of the lattice (Figure 4b), whereas for the β-CD-D-NAcTrp dimer model the crystallographic coordinates of the β-CD-L-NAcTrp dimer were employed after changing the chirality of the L-NAcTrp Cα atom only to generate the D-NAcTrp guest molecule.

chemsum

{"title": "Molecular recognition of <i>N</i>-acetyltryptophan enantiomers by \u03b2-cyclodextrin", "journal": "Beilstein"}

synthesis_of_tunable_porosity_of_fluorine-enriched_porous_organic_polymer_materials_with_excellent_c

## Abstract: We herein report the construction of four the novel fluorine-enriched conjugated microporous polymers (FCMP-600@1-4), which have permanent porous structures and plenty of fluorine atoms in the skeletons as effective sorption sites. Among them, FCMP-600@4 shows considerable adsorption capacity of CO 2 of 5.35 mmol g −1 at 273 K, and 4.18 mmol g −1 at 298 K, which is higher than the reported values for most porous polymers. In addition, FCMP-600@1-4 display high selectivity of CO 2 /N 2 and high CH 4 uptakes.Today, world climate change and environmental problems have become increasingly prominent, so that people have to face the impact of excessive carbon dioxide of atmosphere on humanity, such as the global warming and acid rain. People are eager to find a solution to reduce the concentration of carbon dioxide in the atmosphere, while limiting its emissions, but also studying the ability to capture and storage of new materials. For this purpose, porous organic polymers (POPs) are emerged as the times required, which is a new kind of porous materials with large specific surface area and permanent pore structure. Because of its low density, large specific surface area, adjustable size, and high porosity, as well as a great potential in gas storage, separation, heterogeneous catalysis and other aspects 1,2 . POPs has become one of the hotspots in the recent years and rapid development. People have studied a series of POPs, in addition to traditional zeolites 3 and activated carbons 4 , including polymers of intrinsic microporosity (PIMs) 5 , hypercross-linked polymers (HCPs) 6 , conjugated microporous polymers (CMPs) 7 , and covalent organic frameworks (COFs) 8 . Compared with inorganic microporous materials and metal organic frameworks (MOFs), the synthesis of POPs has just started. But the organic synthesis of chemistry and polymer chemistry have been provided a wide range of development space for the synthesis of such materials. Therefore, from scientific research and practical application, design and synthesis of POPs with good adsorption property of carbon dioxide are of great significance.Among them, CMPs have attracted a high degree of concern in the recent years due to the excellent capture performance of CMPs for carbon dioxide 1,7-13 . CMPs are synthesized via metal-catalyzed cross-coupling chemistry to form cross-linked network. It is a subclass of POPs with conjugated structure, precise adjustment of micropore, large specific surface area and high stability, and the introduction of functional groups in the pore skeleton can effectively improve the capture capacity of carbon dioxide.In particular, the existence of nitrogen atoms in the porous skeleton, the aromatic heterocyclic network, the introduction of ions and so on are all beneficial to improving the adsorption of carbon dioxide on the materials in reported research studies [9][10][11][12] . In order to improve the adsorption properties of carbon dioxide on the polymers, in this paper, fluorine-enriched monomer 4,4′-dibromooctafluorobiphenyl (DBFB) and comonomer containing acetylene bond were selected to synthesize a series of structural tunable CMPs by Sonogashira-higihara reaction under Pd(0) catalysis (Fig. BET surface areas of the polymers increase from 755 to 901 m 2 g −1 , and the total pore volumes of the polymers hole also vary from 0.4242 to 0.6654 cm 3 g −1 . Particularly, FCMP-600@1-4 exhibit the outstanding adsorption of carbon dioxide and methane. ## Synthesis and characterization. All of the polymer networks were synthesized by palladium(0)-catalyzed Sonogashira-higihara reaction of 4,4′-dibromooctafluorobiphenyl (DBFB) and comonomers containing acetylene moities. All the reactions were carried out at a fixed reaction temperature and reaction time (120 °C/48 h). The general synthetic routes toward FCMP@1-4 polymers are shown in Fig. 1. The insoluble polymers were filtered and washed with water, tetrahydrofunan, chloroform, and methanol, respectively, in order to remove the inorganic salts, organic monomers, residual catalyst, and oligomers. Then the pyrolysis reactions of the FCMP@1-4 were carried out on quartz tubes in an electric furnace under a argon atmosphere. The FCMP@1, 2, 3, and 4 samples were heated from the room temperature to 400 °C, 600 °C, and 800 °C with a heating rate of 3 °C/min, then pyrolyzed at 400 °C, 600 °C, and 800 °C for 2 h in argon gas (400 sccm), respectively. Then, we investigated the CO 2 adsorption capacity of these samples at 273 K and 298 K, respectively. We found these precursors at 600 °C displayed the best results compared to precursors at 400 °C and 800 °C. Therefore, we selected samples processed under 600 °C condition to be carefully investigated. The pyrolysis reactions at 600 °C in argon gas were denoted to FCMP-600@1, FCMP-600@2, FCMP-600@3, and FCMP-600@4. Our aim is to explore the effect of structure and connecting position of linker on pore properties of the resulting porous polymers. All of these polymers are insoluble in common organic solvents because of their highly cross linked structures. Formation of FCMP@1-4 was confirmed by the FT-IR analysis. The disappearance of C-Br bonds in spectra of FCMP@1-4 compared with monomer 4,4′-dibromooctafluorobiphenyl demonstrated the success of phenyl-acetylene coupling (ESI, Figure S1). The four infrared spectra of the polymers are basically similar and demonstrate two main adsorption regions: a first absorption band in the 650-1250 cm −1 region, which is assigned as to benzene ring skeleton vibration; while the second peak close to 2900 cm −1 , corresponding to -C-H stretching of benzene ring. In addition, a relatively weak peak at approximate 2202 cm −1 , which referred to -C≡C-stretching of alkynyl moiety of FCMP@1-4, which was further proved that the polymers were synthesized successfully. Elemental analysis indicated that the carbon and hydrogen contents of FCMP@1-4 were close to the theoretical values of an ideal network with a high degree of polycondensation. X-ray diffraction (XRD) showed the amorphous nature of the resulting FCMP@1-4 (ESI, Figure S2a) and FCMP-600@1-4 (ESI, Figure S2b). Transmission electron microscopy (TEM) analyses also showed the amorphous texture of FCMP@1-4 (ESI, Figure S3(a-d)) and FCMP-600@1-4 (ESI, Figure S3(e-h)) materials. Field-emission scanning electron microscopy (FE-SEM; ESI, Figure S4(a-d)) was utilized to investigate the morphology of FCMP@1-4 polymers. The results of FE-SEM show that FCMP@1-4 are irregular sphere shape with particle size 100~300 nm, while FCMP-600@1-4 are irregular lumps with nanometre dimensions (ESI, Figure S4(e-h)). Furthermore, X-ray photoelectron spectroscopy (XPS) results display fluorine elements still exist in FCMP-600@1-4 after pyrolysis (ESI, Figure S5). The surface areas and porous properties of FCMP@1-4 and FCMP-600@1-4 were analyzed by nitrogen sorption analysis at 77.3 K. As shown in Figure S6(a-d), except for FCMP@4, the isotherms of FCMP@1, 2 and 3 showed rapid nitrogen adsorption at low pressure. The Brunauer-Emmett-Teller (BET) surface areas of FCMP@1, 2, 3 and 4 were calculated to be 551, 636, 692, and 88 m 2 g −1 , respectively. The total pore volumes were 0.3865, 0.6983, 0.4074 and 0.1180 cm 3 g −1 , respectively (ESI, Table S1). Compared to FCMP@1-3, FCMP@4 has a significantly low surface area and pore volume. This could be caused by the strong π-π stacking effect between the molecules tetrakis(4-ethynylphenyl)ethene, which lead to formation of planar sheet-like rather than three-dimensional structure 14 . Besides that, the pore size distributions of FCMP@1-4 are very broad (Figure S6(e)). The porosity data of the polymers are summarized in Table S1. In order to overcome this, a successive cross-linking pathway was utilized to improve the BET surface area of the porous polymers. The obtained porous materials FCMP-600@1-4 displayed high surface areas via template-free pyrolysis of FCMPs precursors at 600 °C. The BET surface areas were obtained to be 755, 780, 807 and 901 m 2 g −1 and the total pore volumes were 0.4242, 0.6654, 0.4033 and 0.4331 cm 3 g −1 (micropore volumes calculated from the nitrogen isotherms at P/P 0 = 0.0500 are 0.1951, 0.4502, 0.1636 and 0.1998 cm 3 g −1 ) for FCMP-600@1, FCMP-600@2, FCMP-600@3 and FCMP-600@4, respectively. These results indicated that the surface area and pore volume could be indeed increased by using the pyrolysis of POPs without any templates. As shown in Fig. 2a, FCMP-600@1-4 materials show type I isotherms featured by a sharp uptake at the low-pressure region between P/P 0 = 1 × 10 −5 to 1 × 10 −2 , reflecting the presence of micropores. Distinctly, FCMP-600@1 and 2 possess obvious hysteresis extending to low pressure between the adsorption and desorption isotherms, while FCMP-600@3 displays a relatively tiny hysteresis, which is partly attributed to the swelling in a flexible polymer network, as well as mesopore contribution . Compared with FCMP-600@1, 2, and 3, FCMP-600@4 exhibits a negligible hysteresis loop in the whole pressure range, suggesting that this polymer possesses a very rigid molecular structure. The increase in nitrogen sorption at a high relative pressure for FCMP-600@1-4 may arise from the interparticulate porosity associated with the mesopores of the samples. The pore size distributions were calculated from the nonlocal density functional theory (NLDFT) using the model of carbon as an adsorbent, and the main micropore size peaked at 1.05, 1.74, 0.78, and 0.84 nm for FCMP-600@1, 2, 3, and 4, respectively (Fig. 2b). ## Discussion Gas uptake capacity and separation. The CO 2 adsorption capacities of FCMP-600@1-4 under 273 K and 298 K were also measured (Fig. 3), which displayed linear trend at both 273 K and 298 K, respectively. At 273 K and 1.05 bar, the CO 2 capture uptakes of FCMP-600@1, 2, 3, and 4 are 88, 68, 73, and 119 cm 3 g −1 (5.35 mmol g −1 ), respectively (Fig. 3a). The adsorbance also can reach 65, 49, 61 and 93 cm 3 g −1 (4.18 mmol g −1 ) for FCMP-600@1, 2, 3, and 4 at 298 K (Fig. 3b). FCMP-600@1-4 can enhance the CO 2 uptake by 3.3-, 2.3-, 2.7-, and 4.2-fold than those of the corresponding precursor FCMP@1, 2, 3, and 4, respectively (ESI, Figure S7 and Table S1). Among them, FCMP-600@4 dispalys the highest CO 2 capture capacity than those of other three polymers at both 273 K and 298 K, which could be attributed to the narrower micropore size and higher micropore surface area of FCMP-600@4. This value is a little lower than that of recently reported P-PCz (S BET = 1647 m 2 g −1 , 5.57 mmol g −1 ) 9 , FCTF −1 -600 (S BET = 1535 m 2 g −1 , 5.53 mmol g −1 ) 15 , and PPF −1 (S BET = 1740 m 2 g −1 , 6.12 mmol g −1 ) 16 , but can (a) Nitrogen adsorption/desorption isotherms measured at 77.3 K for FCMP-600@1-4, the isotherms of FCMP-600@1-3 are shifted vertically by 300, 200 and 100 cm 3 g −1 for better visibility, respectively; (b) pore size distributions calculated using density functional theory (DFT) method, for clarity, the curves of FCMP-600@1-3 are shifted vertically by 3, 2 and 1 cm 3 g −1 , respectively. compete with the best performing POP-based adsorbents like BILP-4 (S BET = 1135 m 2 g −1 , 5.34 mmol g −1 ) 17 , ALP −1 (S BET = 1235 m 2 g −1 , 5.37 mmol g −1 ) 18 . In particularly, FCMP-600@4 (4.18 mmol g −1 ) also exhibits an excellent CO 2 capacity at 298 K, which is higher than the reported values for most porous polymers at 273 K, such as CPOP-9 (S BET = 2440 m 2 g −1 , 4.14 mmol g −1 ) 19 , CPOP-8 (S BET = 1610 m 2 g −1 , 3.75 mmol g −1 ) 18 , and fl-CTF400 (S BET = 2862 m 2 g −1 , 4.13 mmol g −1 ) 20 . Compared to FCMP-600@1 and 3, FCMP-600@2 has a broader micropore size distribution. Therefore, FCMP-600@2 exhibits lower CO 2 capture capacity than those of FCMP-600@1 and 3 at the same conditions, although FCMP-600@1, 2 and 3 show the similar BET surface areas. The isosteric heat (Q st ) of adsorption CO 2 was estimated from adsorption data collected under 273 K and 298 K through the Clausius-Clapeyron equation. At zero coverage, the Q st of FCMP-600@1, 2, 3, and 4 are 23.4, 19.9, 17.3, and 21.4 kJ mol −1 , respectively (Fig. 3c). The Q st are lower than the values reported for imine-linked organic polymers, CTFs, diimide polymers, and so on 9, . The relatively high CO 2 uptake and binding by FCMP-600@1-4 are most likely due to favorable interactions of the polarizable CO 2 molecules through hydrogen bonding and/or dipole quadrupole interactions that utilize the proton-free fluorine sites of phenyl rings . In light of high CO 2 capture capacities, high surface areas, fluorine-enriched skeletons, and small pore sizes for FCMP-600@1-4, it is reasonable to study the selective uptake of FCMP-600@1-4 for small gases (CO 2 , CH 4 , and N 2 ) to evaluate their potential use in gas separation. The methane isotherms depicted in Fig. 3d are fully reversible and the uptakes of FCMP-600@1, 2, 3, and 4 reach 36, 27, 30, and 53 cm 3 g −1 at 273 K and 1.0 bar, respectively (ESI, Figure S8 and Table S1). The CH 4 uptakes of FCMP@1-4 were 4.8-6.6 cm 3 g −1 at the same conditions (ESI, Figure S9 and Table S1). Apparently, FCMP-600@1-4 are higher 4.5 −1 1 times than those of the FCMP@1-4 precursors. This result implyed that the FCMP-600@1-4 can efficiently capture CH 4 due to high microsurface area and micropore volume 26,27 . The selectivities of FCMP-600@1-4 toward CO 2 over CH 4 and N 2 were investigated by collecting pure component physisorption isotherms at 273 K (ESI, Figure S8), and then which were predicted from the experimental pure component isotherms using the ideal adsorbed solution theory (IAST). At zero coverage, the high CO 2 /N 2 selectivity was recorded for FCMP-600@1-4 (109-77 at 273 K) (ESI, Figure S10a). Moreover, FCMP-600@1-4 show a moderate level CH 4 /N 2 selectivities: 8-11 (273 K) (ESI, Figure S10b). ## Iodine capture. In the recent years, the capture of iodine using porous materials has attracted considerable interest. Most interestingly, we found the fluorine-enriched polymers were highly efficient for the iodine adsorption. The absorption of solid iodine was conducted by exposing the samples to nonradioactive iodine vapor in a sealed vessel at 350 K and ambient pressure, which was the typical fuel reprocessing condition. Gravimetric measurement was performed at different time intervals during the iodine loading (Fig. 4a). Except for FCMP-600@2, the maximum iodine uptakes of other three porous materials were reached quickly saturated in the first 4 h. As the synthesized polymers, FCMP-600@2 has the maximum value for iodine uptake reached up to 141 wt.%, followed by FCMP-600@4 (111 wt.%), FCMP-600@1 (108 wt.%), and FCMP-600@3 (90 wt.%). The thermogravimetric analysis (TGA) of the I 2 -loaded FCMP-600@2 and 4 polymers reveal a significant weight loss from 90 to 300 °C (ESI, Figure S11), the calculated iodine mass loss were 152 and 105 wt.% for FCMP-600@2 and FCMP-600@4, respectively, which was close to the saturated adsorption value. Additionally, the FCMP-600@1-4 are capable of capture iodine in solution. When the FCMP-600@2 (30 mg) in iodine/hexane solution (4 mg mL −1 , 3 mL), the dark purple solution gradually faded to light purple (Fig. 4b and Figure S12). The UV/Vis absorption intensity of the samples was decreased with the prolonged action time (ESI, Figure S13). It can be observed from the adsorption kinetics of iodine at room temperature that the adsorption process was affected by the contact time (Fig. 4c). In the initial stage, the adsorption capacity increased quickly with the prolonged contact time, and then slow down to equilibrium after about 10 h. The removal efficiencies of polymers achieved for the solution are 81.2-92.4%. The adsorption kinetics of iodine for FCMP-600@1-4 were analyzed through the frequently used pseudo-first-order and pseudo-second-order models were adopted 22 . Results show that the adsorption data fits well in pseudo-second-order kinetic model with good linear correlation coefficient (R 2 ) values of 0.9961, 0.9962, 0.9962 and 0.9962 for iodine solution of FCMP-600@1, 2, 3 and 4, respectively (ESI, Table S2, Figures S14 and S15). This confirmed that the iodine adsorption process in this work was governed by the pseudo-second-order kinetics. The XPS spectrum of fluorine-enriched polymers indicated that the coexistence of elemental iodine and triiodide ion, which suggested a hybrid of physisorption and chemisorption (ESI, Figure S16). Furthermore, it is very easy to remove or release the trapped iodine molecules of the samples via immersion of the iodine-loaded sample in ethanol. When the I 2 -FCMP-600@1-4 were immersed in ethanol, the colour of the solvent were changed from colourless to dark brown (ESI, Figure S17), indicating that the iodine guests were released from the solid. The four samples were recycled easily for at least five times without significant loss of iodine uptake (ESI, Figure S18). The saturated iodine adsorption capacities of FCMP-600@1-4 can be determined from the adsorption isotherms (ESI, Table S3, Figures S19 and S20). Two different adsorption stages were observed from the plot of the equilibrium concentration versus the quantities of the adsorbed iodine at equilibrium. At the equilibrium uptake increases linearly with the increase of iodine solution concentration at low concentration. Then, the adsorption reached its maximum value and the adsorption process turned to be independent on the concentration. The simulation results revealed that the iodine adsorption of samples could be well described using Langmuir adsorption isotherm (ESI, Figures S19 and S20), suggesting a monolayer adsorption behavior for iodine molecule on the surface of polymers. From the sorption kinetics, the maximum capacities for iodine uptake reached up to around 550, 729, 520, and 539 mg g −1 for FCMP-600@1, 2, 3, and 4, respectively. ## Conclusion In summary, four the novel fluorine-enriched porous materials were successfully designed and synthesized. The properties of FCMP-600@1-4 were well investigated and discussed. The BET surface areas of FCMP-600@1-4 can be tuned by changing the geometry and size of comonomer. FCMP-600@1-4 have the BET specific surface areas of 755-901 m 2 g −1 as well as permanent microporosity, and the abundant fluorine atoms in the skeleton endow the materials with high CO 2 /N 2 (109-77) and CH 4 /N 2 (8 −1 1) selectivities. At 273 K and 1.05 bar, FCMP-600@4 exhibits the highest CO 2 uptake of 119 cm 3 g −1 , and CH 4 uptake of 53 cm 3 g −1 , and the rest materials are in the range of 68-88 cm 3 g −1 for CO 2 . Meanwhile, FCMP-600@1-4 show good adsorption capacities of 90-141 wt.% toward iodine vapor. We hope this type of fluorine-doped absorbent can be effective for gas storage and will bring new application possibilities. ## Methods Synthesis of FCMP@1. 4,4′-Dibromooctafluorobiphenyl (114 mg, 0.25 mmol) and 1,3-diethynylbenzene (47 mg, 0.375 mmol) were put into a 50 mL two-necked round-bottom flask, then the flask exchanged 3 cycles under vacuum/N 2 . Then added to 2 mL DMF and 2 mL triethylamine (Et 3 N), the flask was further degassed by the freeze-pump-thaw for 3 times. When the solution had reached reaction temperature, a slurry of tetrakis(triphenylphosphine)palladium(0) (19.9 mg, 0.017 mmol) in the 1 mL DMF and copper(I) iodine (3.1 mg, 0.017 mmol) in the 1 mL Et 3 N was added, and the reaction was stirred at 120 °C for 48 h under nitrogen atmosphere. The solid product was collected by filtration and washed well with THF, methanol, acetone, and water for 4 times, respectively. Further purification of the polymer was carried out by Soxhlet extraction with methanol, and THF for 24 h, respectively, to give FCMP@1 as yellow powder (88.7% yield). Elemental Analysis (%) C 69.24, H 1.55. Found: C 66.88, H 1.16. Synthesis of FCMP@2. 4,4′-Dibromooctafluorobiphenyl (114 mg, 0.25 mmol) and 1,3,5-triethynylbenzene (45 mg, 0.25 mmol) were put into a 50 mL two-necked round-bottom flask, then the flask exchanged 3 cycles under vacuum/N 2 . Then added to 2 mL DMF and 2 mL triethylamine (Et 3 N), the flask was further degassed by the freeze-pump-thaw for 3 times, purged with N 2 . When the solution had reached reaction temperature, a slurry of tetrakis(triphenylphosphine)palladium(0) (27.7 mg, 0.024 mmol) in the 1 mL DMF and copper(I) iodine (5.7 mg, 0.032 mmol) in the 1 mL Et 3 N was added, and the reaction was stirred at 120 °C for 48 h under nitrogen atmosphere. The solid product was collected by filtration and washed well with THF, methanol, acetone, and water for 4 times, respectively. Further purification of the polymer was carried out by Soxhlet extraction with methanol, and THF for 24 h, respectively, to give FCMP@2 as brown solid (93.6% yield). Elemental Analysis (%) Calcd. (Actual value for an infinite 2D polymer) C 70.86, H 1.11. Found: C 68.17, H 0.95. Synthesis of FCMP@3. 4,4′ -Dibromo o c taf luorobipheny l (114 mg, 0.25 mmol) and tetrakis(4-ethynylphenyl)methane (78 mg, 0.188 mmol) were put into a 50 mL two-necked round-bottom flask, then the flask exchanged 3 cycles under vacuum/N 2 . Then added to 2 mL DMF and 2 mL triethylamine (Et 3 N), the flask was further degassed by the freeze-pump-thaw for 3 times. When the solution had reached reaction temperature, a slurry of tetrakis(triphenylphosphine)palladium(0) (17.9 mg, 0.015 mmol) in the 1 mL DMF and copper(I) iodine (3.7 mg, 0.02 mmol) in the 1 mL Et 3 N was added, and the reaction was stirred at 120 °C for 48 h under nitrogen atmosphere. The solid product was collected by filtration and washed well with THF, methanol, acetone, and water for 4 times, respectively. Further purification of the polymer was carried out by Soxhlet extraction with methanol, and THF for 24 h, respectively, to give FCMP@3 as yellowish-brown powder (94.3% yield). Elemental Analysis (%) C 82.44, H 3.07. Found: C 78.28, H 3.43. Synthesis of FCMP@4. 4,4′-Dibromooctafluorobiphenyl (114 mg, 0.25 mmol) and 1,1,2,2-tetrakis(4-ethynylphenyl)ethene (71.5 mg, 0.188 mmol) were put into a 50 mL two-necked round-bottom flask, then the flask exchanged 3 cycles under vacuum/N 2 . Then added to 2 mL DMF and 2 mL triethylamine (Et 3 N), the flask was further degassed by the freeze-pump-thaw for 3 times. When the solution had reached reaction temperature, a slurry of tetrakis(triphenylphosphine)palladium(0) (17 mg, 0.015 mmol) in the 1 mL DMF and copper(I) iodine (2.7 mg, 0.015 mmol) in the 1 mL Et 3 N was added, and the reaction was stirred at 120 °C for 48 h under nitrogen atmosphere. The solid product was collected by filtration and washed well with THF, methanol, acetone, and water for 4 times, respectively. Further purification of the polymer was carried out by Soxhlet extraction with methanol, and THF for 24 h, respectively, to give FCMP@4 as pale yellow powder (88.7% yield). Elemental Analysis (%) C 82.83, H 3.01. Found: C 75.79, H 2.12. Synthesis of FCMP-600@1-4. The pyrolysis reactions of the FCMP@1-4 were carried out on quartz tubes in an electric furnace under argon atmosphere. The FCMP@1, 2, 3, and 4 samples were heated from the room temperature to 600 °C with a heating rate of 3 °C/min, then pyrolyzed at 600 °C for 2 h in argon gas (400 sccm), respectively. The pyrolysis reactions at 600 °C in argon gas were denoted to FCMP-600@1, FCMP-600@-2, FCMP-600@-3, and FCMP-600@4, respectively.

chemsum

{"title": "Synthesis of tunable porosity of fluorine-enriched porous organic polymer materials with excellent CO2, CH4 and iodine adsorption", "journal": "Scientific Reports - Nature"}

copper-catalyzed_cuaac/intramolecular_c–h_arylation_sequence:_synthesis_of_annulated_1,2,3-triazoles

## Abstract: Step-economical syntheses of annulated 1,2,3-triazoles were accomplished through copper-catalyzed intramolecular direct arylations in sustainable one-pot reactions. Thus, catalyzed cascade reactions involving [3 + 2]-azide-alkyne cycloadditions (CuAAC) and C-H bond functionalizations provided direct access to fully substituted 1,2,3-triazoles with excellent chemo-and regioselectivities. Likewise, the optimized catalytic system proved applicable to the direct preparation of 1,2-diarylated azoles through a one-pot C-H/N-H arylation reaction. ## Introduction Transition-metal-catalyzed C-H bond functionalizations are increasingly viable tools for step-economical syntheses of various valuable bioactive compounds , which avoid the preparation and use of preactivated substrates . This streamlining of organic synthesis has predominantly been accomplished with palladium , rhodium or ruthenium complexes . However, less expensive nickel, cobalt, iron or copper catalysts bear great potential for the development of economically attractive transformations . In this context, we previously reported on the use of costeffective copper(I) catalysts for direct arylations of 1,2,3-triazoles. Thus, we showed that intermolecular copper-catalyzed C-H bond functionalizations could be combined with the Huisgen copper(I)-catalyzed [3 + 2]-azide-alkyne cycloaddition (CuAAC) , while C-H bond arylations of 1,2,3-triazoles were previously only accomplished with more expensive palladium or ruthenium catalysts. Notably, this strategy allowed for the atom-economical synthesis of fully substituted 1,2,3-triazoles in a highly regioselective fashion . While the research groups of Rutjes as well as Sharpless elegantly devised alternative approaches exploiting 1-haloalkynes , we became interested in exploring a single inexpensive copper catalyst for onepot reaction sequences comprising a 1,3-dipolar cycloaddition along with an intramolecular C-H bond arylation; in particular, because of the notable biological activities exerted by fully substituted 1,2,3-triazoles . As a consequence, we wish to present herein novel cascade reactions, in which cost-effective copper(I) compounds serve as the catalyst for two mechanistically distinct transformations for the synthesis of fully substituted annulated 1,2,3-triazoles as well as for twofold N-H/C-H bond arylations. Notable features of our strategy include (i) the development of a chemoselective C-H arylationbased three-component reaction, as well as (ii) the use of inex-pensive CuI for the formation of up to one C-C and three C-N bonds in a site-selective fashion (Scheme 1). ## Results and Discussion We initiated our studies by exploring reaction conditions for the key copper-catalyzed intramolecular direct C-H bond arylation, employing substrate 3a (Table 1). Notably, the envisioned C-H bond functionalization occurred readily with the aryl iodide 3a when catalytic amounts of CuI were used, even at a reaction temperature as low as 60 °C, with optimal yields being obtained With optimized reaction conditions for the intramolecular direct arylation in hand, we tested the possibility of its implementation in a sequential synthesis of 1,4-dihydrochromeno [3,4d] triazole (4b, Scheme 2). We were delighted to observe that the desired reaction sequence consisting of a coppercatalyzed 1,3-dipolar cycloaddition and an intramolecular C-H bond arylation converted alkyne 1a to the desired product 4b with high catalytic efficacy. Subsequently, we explored the extension of this approach to the development of a chemoselective three-component one-pot reaction. Thus, we found that alkyl bromides 2 could be directly employed as user-friendly substrates for the in situ formation of the corresponding organic azides (Scheme 3). Notably, the catalytic system proved broadly applicable, and a variety of organic electrophiles 2, thereby, delivered differently decorated Importantly, performing the one-pot reaction in a sequential fashion was not found to be mandatory. Indeed, our strategy turned out to be viable in a nonsequential manner by directly employing equimolar amounts of the three substrates. Hence, inexpensive CuI allowed the direct assembly of aryl iodides 1, alkyl bromides 2 and NaN 3 with excellent chemo-and regioselectivities (Scheme 4). Thereby, a variety of annulated 1,2,3triazoles 4 were obtained, featuring six-or seven-membered rings as key structural motifs. It is particularly noteworthy that the copper-catalyzed transformation enabled the formation of one C-C and three C-N bonds in a chemoselective manner, and thereby provided atom-and step-economical access to annulated carbo-as well as O-and N-heterocycles. Finally, we found that the catalytic system also proved to be applicable to the one-pot copper-catalyzed direct arylation of various azoles 5 through N-H/C-H bond cleavages with aryl iodides 6 as the organic electrophiles (Scheme 5). ## Conclusion In summary, we have reported on the use of inexpensive copper(I) complexes for step-and atom-economical sequential catalytic transformations involving direct C-H bond arylations. Thus, CuI enabled the synthesis of fully substituted 1,2,3-tria-zoles through cascade reactions consisting of copper(I)catalyzed [3 + 2]-azide-alkyne cycloadditions (CuAAC) and intramolecular C-H bond arylations. Notably, the optimized copper catalyst accelerated two mechanistically distinct transformations, which set the stage for the formation of up to one C-C and three C-N bonds in a chemo-and regioselective fashion, and also allowed for twofold C-H/N-H bond arylations on various azoles. ## Experimental General information Catalytic reactions were carried out under an inert atmosphere of nitrogen using predried glassware. All chemicals were used as received without further purification unless otherwise specified. DMF was dried over CaH 2 . Alkynes 1 and triazoles 3 were synthesized according to previously described methods. CuI (99.999%) was purchased from ABCR with the following specifications: Ag <3 ppm, Ca = 2 ppm, Fe = 1 ppm, Mg <1 ppm, Zn <1 ppm. Yields refer to isolated compounds, estimated to be >95 % pure, as determined by 1 H NMR. Thin- General procedure for the synthesis of triazoles 4 NaN 3 (1.05 equiv), CuI (10 mol %), LiOt-Bu (2.00 equiv), alkyne 1 (1.00 equiv) and alkyl bromide 2 (1.00 equiv) were dissolved in DMF (3.0 mL) and stirred at 80 °C for 20 h. Then, H 2 O (50 mL) was added at ambient temperature, and the resulting mixture was extracted with EtOAc (3 × 50 mL). The combined organic layers were washed with saturated aq NH 4 Cl (50 mL), H 2 O (50 mL) and brine (50 mL), dried over Na 2 SO 4 , filtered and concentrated in vacuo. The remaining residue was purified by column chromatography on silica gel (n-hexane/ EtOAc).

chemsum

{"title": "Copper-catalyzed CuAAC/intramolecular C\u2013H arylation sequence: Synthesis of annulated 1,2,3-triazoles", "journal": "Beilstein"}

disorder_and_defects_are_not_intrinsic_to_boron_carbide

## Abstract: A unique combination of useful properties in boron-carbide, such as extreme hardness, excellent fracture toughness, a low density, a high melting point, thermoelectricity, semi-conducting behavior, catalytic activity and a remarkably good chemical stability, makes it an ideal material for a wide range of technological applications. Explaining these properties in terms of chemical bonding has remained a major challenge in boron chemistry. Here we report the synthesis of fully ordered, stoichiometric boron-carbide B 13 C 2 by high-pressure-high-temperature techniques. Our experimental electron-density study using high-resolution single-crystal synchrotron X-ray diffraction data conclusively demonstrates that disorder and defects are not intrinsic to boron carbide, contrary to what was hitherto supposed. A detailed analysis of the electron density distribution reveals charge transfer between structural units in B 13 C 2 and a new type of electron-deficient bond with formally unpaired electrons on the C-B-C group in B 13 C 2 . Unprecedented bonding features contribute to the fundamental chemistry and materials science of boron compounds that is of great interest for understanding structure-property relationships and development of novel functional materials.Boron carbide is one of the hardest substances, surpassed only by diamond and boron nitride 1 . The high mechanical and thermal stability, low density and low costs of fabrication have made boron carbide the prime choice in a series of technological applications [1][2][3][4][5][6][7] . Boron carbide preserves the same structure for a range of compositions, and details of this crystal structure have been discussed in terms of chemical disorder of boron and carbon atoms as well as the presence of vacancies 1,[8][9][10][11] . Electronic-structure calculations suggest that the properties of boron carbide depend on the stoichiometry and the details of the disorder 2,7,12,13 .Experimentally, chemical bonding can be accessed through single-crystal x-ray diffraction. Reliable information on the distribution of the electron density in the unit cell can be obtained only for good-quality single crystals with minimal structural disorder 14 . Synthesis of defect-free material is the most challenging task in boron carbide chemistry. We have succeeded in growing small single crystals of the stoichiometric composition B 13 C 2 by high-pressure-high-temperature techniques (see Methods). The material is transparent with a dark red or maroon color, indicating an insulator or a large-band-gap semiconductor. This is in agreement with some literature data 15 , but it is inconsistent with the relatively high electrical conductivity reported for boron carbide 1 . To the best or our knowledge, dark red transparent boron carbide has not been reported before.A multipole (MP) model has been obtained for the crystal structure of B 13 C 2 by refinement against accurately measured intensities of Bragg reflections (see Methods and Supplementary Information Section S1) 14 . The excellent fit to the diffraction data with R 1 = 0.0197 provides strong evidence for the stoichiometry of B 13 C 2 , in agreement with the composition obtained by Energy-dispersive x-ray (EDX) analysis (see Methods). The excellent fit furthermore indicates the absence of disorder: B 13 C 2 is composed of B 12 icosahedral clusters and CBC linear chains (Fig. 1 and Supplementary Information Section S2). Lattice parameters and values of atomic displacement parameters (ADPs) fall within a range previously assigned to the composition B 12 C 3 1,8-10 . The possibility of different compositions was investigated by additional MP refinements with small amounts of carbon at the B P site, corresponding to B 12 + x C 3 − x stoichiometries with x = 0.44 and x = − 0.11, respectively (see Supplementary Information Section S1 for details). Both models gave a slightly worse fit to the diffraction data than the B 13 C 2 model. More importantly, the number of valence electrons of C at the B P site refined to zero, thus showing that the MP refinement has effectively removed carbon from the B P site, providing further support for the ordered stoichiometric character of the investigated crystal. Interestingly, a refinement of the independent atom model (IAM) including the site occupancy factors of C at the B P and B E sites resulted in 19% occupancy of the B P site by carbon (x = − 0.11). Contrary to the MP model (R 1 = 0.0197), the IAM with disorder (R 1 = 0.0287) leads to only a small improvement of the fit to the data (Table S4). These results suggest that the charge transfer towards B P in the MP model is mimicked in the disordered IAM by a fractional occupancy of the B P site by C. Discrepancies between the present values of the lattice parameters and ADPs and those reported in the literature 1, for the same composition may be the result of different degrees of disorder and defects between different samples. The single MP refinement 16 reported previously for B 13 C 2 gave a much worse fit to their XRD data (R 1 = 0.0440), which questions the reliability of that model. The single MP refinement 17 for B 12 C 3 also led to a substantially worse fit to their XRD data (R 1 = 0.0250) than we have obtained for our model against the present XRD data (R 1 = 0.0197). Thus, a highly precise MP refinement refutes recent less accurate diffraction studies 13 and theoretical electronic-structure calculations 2,12 , where a disorderly replacement by carbon of a certain fraction of the boron atoms of the B 12 clusters was considered as absolutely essential for the stability of B 13 C 2 . The MP model extends the independent atom model (IAM) of spherical atoms by parameters describing the reorganization of electron density due to chemical bonding. Previous electron-density studies on boron carbide 18,19 have been restricted to a discussion of the qualitative features of the electron densities. Quantitative information about chemical bonding can be extracted from the static electron density of the MP model through its topological properties according to Bader's quantum theory of atoms in molecules (QTAIM) 14,20 . Critical points (CPs) are defined as the positions where the gradient of the electron density is zero [∇ρ (r) = 0] 20 . They are classified according to the number of positive eigenvalues of the Hessian matrix of second derivatives as local maxima (3 positive eigenvalues), bond critical points BCPs (2), ring critical points RCPs (1) and local minima (0 positive eigenvalues) 20 . All atomic positions of the present MP model can be identified with local maxima in the static electron density, while additional local maxima do not exist. BCPs and RCPs have been found between the atoms of the B 12 cluster in a similar pattern as for α-boron 21 , and with comparable values for the electron densities and Laplacians (Table 1). Together, these features indicate similar bonding by molecular-type orbitals on the B 12 clusters in B 13 C 2 and α-boron 21 . According to Wade's rule 22 , this bonding involves 26 of the 36 valence electrons of the twelve boron atoms of this closo-cluster, thus leaving for each boron atom one orbital but only 5/6 electrons for exo-cluster bonding 21,23 . The crystal structure of B 13 C 2 comprises four crystallographically independent atoms. CBC chains contain the carbon atom and a boron atom denoted as B C ; the B 12 cluster is made of six polar and six equatorial atoms, denoted as B P and B E , respectively (Fig. 1). According to the QTAIM 20 , bonding between a pair of atoms exists, if the electron density possesses a BCP between those atoms. For B 13 C 2 , we have found BCPs between pairs of B P atoms from neighboring clusters. The distance B P -B P is slightly larger and the magnitudes of the electron density, ρ BCP , and Laplacian, ∇ 2 ρ BCP , are slightly smaller than those of the corresponding inter-cluster bonds in α-boron 21 and γ -boron 24 (Table 1). The high value of ρ BCP together with a negative value of ∇ 2 ρ BCP of large magnitude indicate a strong covalent interaction between these atoms 20 . The similarities with bonding in α-boron 21 (Table 1) allow this bond to be classified as a 2-electron-2-center (2e2c) bond. Further evidence for this interpretation comes from the QTAIM theory, which assigns a charge to each atom by integration of the electron density over the atomic basins. A charge of − 0.21 electrons has been obtained by integrating the experimental static electron density over the atomic basin of B P (Table 2). This value is in good agreement with electron counting. With 5/6 electrons per boron atom available for exo-cluster bonding, a formal charge of − 0.17 is obtained for B P involved in a 2e2c B P -B P exo-cluster bond. Bond-critical points are also found between a B E atom and the closest C atom. Large magnitudes of ρ BCP and the negative Laplacian ∇ 2 ρ BCP indicate a strong covalent interaction and a 2e2c C-B E bond. An equal split of these electrons between C and B E again gives a formal charge of − 0.17 for B E , and it would result in a (B 12 ) 2− group 2 However, carbon is more electronegative than boron and should attract most of the bonding electrons. Indeed, the integration of the electron density over the atomic basins leads to a positive atom B E and a strongly negative C atom (Table 2). A detailed analysis of the electron density shows that the positive charge of B E is the result of a strong polar-covalent character of the C-B E bond, with the BCP much closer to B E than to C (Fig. 2; Table 1), but with a large value of ρ BCP as opposed to an expected small value for ionic bonding 20 . With the interpretation of B P -B P and C-B E bonds as 2e2c bonds, only three electrons are left for the two C-B C bonds of the CBC group (see Supplementary Material Section S3). These bonds can therefore be described as a three-electron-three-center (3e3c) bond or as resonance between two equivalent combinations of one 2e2c and one 1e2c bond (Fig. 3). The large values of the electron density along the bond path (Fig. 2a) correlate with the short bond length, which is explained by the internal pressure on the CBC group 2 . Large magnitudes of ρ BCP and ∇ 2 ρ BCP indicate a covalent interaction. The electron deficient character of this bond is in complete agreement with the ionic charge of + 2.30 of B C . The latter value is the result of the extremely small volume of the atomic basin of this atom, which demonstrates that the internal pressure has squeezed out most of the electrons of B C , reminiscent of the effect of pressure on the electrons in lithium metal 25 . A 3e3c C-B C -C bond contains one unpaired electron per formula unit B 13 C 2 . Experimentally, unpaired spins have been observed at much lower concentrations in boron carbides of different compositions 2,4,5,26,27 . One explanation lies in chemical disorder and vacancies, which are necessarily present for other compositions than stoichiometric B 13 C 2 , and which reduce the number of unpaired spins. On the other hand, the itinerant character of the electron states or localization as bipolarons may be in agreement with low concentrations of unpaired spins 2,5,12 . The presence of an unsaturated bond on the CBC chains should result in a high chemical reactivity of this bond. However, we have found that B C is extremely small (Table 2) and well shielded from the outside by C atoms and bulky B 12 clusters. Steric effects hindering access to reactive sites is known to stabilize radicals 28,29 . High temperatures can overcome these barriers, and a high reactivity at elevated temperatures towards oxidizing agents has been described for boron carbide 30 . Recently, amorphisation 6,31 of boron carbide B 12 C 3 has been explained on the basis of the presence of carbon atoms at a small fraction of the B P sites 32 . Stoichiometric B 13 C 2 is a form of boron carbide that lacks this detrimental property of technical boron carbide with compositions on the carbon-rich side of B 13 C 2 . In summary, we have synthesized stoichiometric boron carbide B 13 C 2 , which is free of intrinsic disorder, and is built of B 12 icosahedral clusters and C-B C -C chains. Unlike band-structure calculations 2,12 on fully ordered B 13 C 2 , the ordered stoichiometric compound is an insulator or large band-gap semiconductor. An experimental electron-density study by X-ray diffraction conclusively determines that B 13 C 2 is an electron-precise material. The electron-deficient character is explained by B C being stripped of two of its valence electrons and the existence of a unique, electron deficient 3e3c bond on the C-B C -C chains. The low chemical reactivity follows from the extremely small volume of B C . Table 2. Atomic basins (volume V Basin ) and ionic charges for the four crystallographically independent atoms in B 13 C 2 along with their multiplicity in the unit cell. ## Methods summary Crystal growth. Single crystals of boron-carbide were grown at high pressures of 8.5-9 GPa and high temperatures of 1873-2073 K using a 1200-ton (Sumitomo) multi-anvil hydraulic press at the Bayerisches Geoinstitut. Energy-dispersive x-ray (EDX) analysis has been employed to determine the composition as B 6.51 (12) C, in agreement with stoichiometric B 13 C 2 . The presence of other elements could be excluded. ## X-ray diffraction experiment for/and electron density analysis. A single crystal of boron-carbide of dimensions 0.09 × 0.08 × 0.05 mm 3 was chosen for an x-ray diffraction experiment with synchrotron radiation at beamline F1 of Hasylab, DESY in Hamburg, Germany. The sample was kept at a temperature of 100 K, while a complete data set of accurate intensities was measured for Bragg reflections up to sin(θ )/λ = 1.116 −1 . The diffraction data were integrated using the computer program EVAL 33 . Structure refinements have been performed with the software XD2006 34 . A topological analysis of the static electron density has been performed by the modules TOPXD and XDPROP of the computer program XD2006. Two-dimensional density maps have been generated by the module XDGRAPH. See the Supplementary Information for details on procedures and the MP model.

chemsum

{"title": "Disorder and defects are not intrinsic to boron carbide", "journal": "Scientific Reports - Nature"}

hybrid_coordination-network-engineering_for_bridging_cascaded_channels_to_activate_long_persistent_p

## Abstract: We present a novel "Top-down" strategy to design the long phosphorescent phosphors in the second biological transparency window via energy transfer. Inherence in this approach to material design involves an ingenious engineering for hybridizing the coordination networks of hosts, tailoring the topochemical configuration of dopants, and bridging a cascaded tunnel for transferring the persistent energy from traps, to sensitizers and then to acceptors. Another significance of this endeavour is to highlight a rational scheme for functionally important hosts and dopants, Cr/Nd co-doped Zn 1−x Ca x Ga 2 O 4 solid solutions. Such solid-solution is employed as an optimized host to take advantage of its characteristic trap site level to establish an electron reservoir and network parameters for the precipitation of activators Nd 3+ and Cr 3+ . The results reveal that the strategy employed here has the great potential, as well as opens new opportunities for future new-wavelength, NIR phosphorescent phosphors fabrication with many potential multifunctional bio-imaging applications.There is an increasing interest in the use of long persistent phosphorescence in the biologically transparent window to drive the photonic bioprobe for tracing the cancer cells 1 . Long phosphorescent phosphors (LPPs) can help avoiding the challenging requirement of high-intensity illumination during the signal collection, which often leads to decreased signal-to-noise ratio and photon-induced deterioration of analytes 2 . This emerging research trend, which incorporates various fields of materials science, biology, chemistry, engineering, physics and pharmaceuticals, follows two main directions: operation waveband and persistent duration, with many relevant crossing points in between 3,4 . As we know, there are two biologically transparent windows: first one at 650-950 nm and second one at 1000-1350 nm 5 ; Near-infrared (NIR) light in the first transparency window can penetrate biological tissues such as skin and blood more efficiently than visible light 6 , yet the second region has even lower absorption and scattering therefore offers more efficient tissue penetration 7 . However, the main researches about the operational waveband of NIR LPPs mainly focus on the short wavelength region, i.e. first NIR window.In addition to altering the emission center and tailoring the crystal field surrounding the activator, another useful strategy to extend the operational waveband, is to transfer the persistent energy of sensitizers to acceptors 8 . In fact, although the afterglow properties are predominantly controlled by the active traps, more subtle effects, such as topochemical coordination-configuration of dopant ions, can also have a profound role to the spectroscopic features of LPPs, which has long been recognized as a significant issue lying at the heart of doping chemistry and photoluminescent theory 9 . Considering the advanced engineering of cascaded tunnel of energy transfer (traps → activator(A) → activator(B)) and going into the details of it, one has at one's disposal several decades worth of well-established principles in the coincident matching of macroscopical and microscopic features in spectroscopy, coordination chemistry and network connectivity relating to activators and hosts 10,11 . Traditionally, materials scientists view such network-engineering design accessed via active impurities with a practical eye intent on describing integral architectures in terms of ion types, valency and radius, local coordination geometries, as well as their concomitant implications for electronegativity and chemical bonding 12 . However, due to the complex attribute of topological network, there are still remaining grand challenges: to gain better modulation for the local coordination configuration of dopants, to understand the principle linking the indispensable transfer channel of independent individual, and to realize true predictability to the arrangements of traps and dopants (sensitizers, activators, or co-dopants) in coordinated network. In this work we present a new "Top-down" approach to design and synthesize the long phosphorescent phosphors in the second biological transparent window. The material design approach employed here involves an ingenious engineering for hybridizing the coordination networks of hosts, tailoring the topochemical configuration of dopants, and bridging cascaded channels for transferring the persistent energy from traps, to sensitizers, to acceptors. We present a closed energy transfer channel from Cr 3+ to Nd 3+ in ZnGa 2 O 4 phosphor and invalid electronic reservoir in CaGa 2 O 4 phosphor, respectively. Persistent energy-transfer could occur in Zn 1-x Ca x-Ga 2 O 4 solid-solution because two dopants were successfully locked in a cage via the efficient crystal packing at an appropriate distance, in addition to the preservation of native electron traps. The hybrid network topologies and structural motifs, thus far will be outlined with particular emphasis on how specific route of energy transfer can be prepared via premeditatedly designing a material system. Such design strategy will notably open a vista of potential avenues for the design of new optical functional materials for the future. ## Results and Discussion Our strategy was inspired by the fundamental spectroscopic theory of energy transfer and local intercalation reaction in inorganic polycrystals (Fig. 1) 13 . In our view, a typical long phosphorescent phosphor (MSI, in Fig. 1a) features a prominent electron reservoir (C, in Fig. 1) with the distinct ability of storing and releasing the captured electrons, as well as a notable photon-emitter (A, in Fig. 1) with higher quantum efficiency under the condition of accurately matching lattice-coordination network and atomic radius 14 . A pre-established electronic transfer channel (AG, traps (C) → activator (A), in Fig. 1) ensures the long persistent phosphorescence. However, the topological network does not provide an opportunity for another activator (B, in Fig. 1) to embed itself into the suitable lattice site. Such structural constraint thus, closes the possible channel of energy transfer (ET, in Fig. 1) between (A) and (B), leading to the luminescent and phosphorescent quenching. Fortunately, the existing chemical and spectroscopic knowledge offer a far-sighted technique to select another material system (MSII, in Fig. 1b), which allows a synchronous precipitation of activator (A) and (B), as well as engineers a theoretically existent energy-transfer channel (activator (A) → activator (B)). But to our surprise, this scheme misses the necessary electron reservoir so as to completely decrease the probability of electrons trapping-detrapping (Fig. 1b). The use of solid-solution complexes to engineer predictable, multi-dimensional infinite networks has received ever-increasing attention in the area of chemistry and materials science 15 . Solid-solutions have already proven their superiorities in the areas of optical, optoelectronic, electrical and magnetic properties than the single component 16 a state-of-the-art Li x FePO 4 solid-solution technology, opening the door for lithium ion batteries to take their place in large-scale applications 17 . In addition, a series of solid-solution, such as AgGa 1−x Al x O 2 , Zn 1−x Cu x S, (SrTiO 3 ) 1−x (LaTiO 2 N) x , also have been developed and used as the advanced photocatalysts to enhance the photocatalytic activity of a given semiconductor photocatalyst 18 . Therefore, solid-solution highlights hybrid coordination network of host, and is expected to open up a possibility in the visualization of the structural and functional binding process of traps and all activators into an independent system 19 . By rationally deploying an indirect intercalation complex comprised by polyhedron ligands of materials (MSI) and (MSII), hybrid coordination-network of novel solid-solution (MSIII, in Fig. 1c) is engineered to steady the activators (A) and (B), modulate the topochemical configuration of activators, and realize the cascaded energy transfers, traps(C) → activator(A) → activator(B) (Fig. 1c). Such novel structural motif is anticipated to adopt a disturbance to native unit cell and bridging a predictable periodic coordination network. To validate research idea, a typical NIR long phosphorescent phosphor ZnGa 2 O 4 : Cr was pursued as preferential material system, which has been proven capable of supporting high defect densities, thought to be associated primarily with Zn vacancies (V Zn ) and O vacancies (V O ), as well as some antisite deficiencies (Zn Ga ) 20 . Making use of its defect capacity, ZnGa 2 O 4 : Cr has been demonstrated as a NIR photo-emitter with surprisingly long persistent phosphorescence in first NIR window (Supplementary Fig. S1). Here, we target the operating waveband in the second NIR window by transferring the persistent energy of Cr 3+ to Nd 3+ in Cr/Nd-codoped ZnGa 2 O 4 LPPs. Nd 3+ ion is chosen as the emission center in order to take advantage of the appropriate energy level characteristic, i.e. NIR-absorption (680, 750 and 800 nm) and NIR emission (1064 nm) 21 . The various sharp transitions of Nd 3+ [ 4 I 9/2 → 4 F 5/2 ], [ 4 I 9/2 → 4 F 7/2 ], [ 4 I 9/2 → 4 F 9/2 ], just overlap the electron transition from metastable state ( 4 T 2 ) to ground state ( 4 A 2 ) of Cr 3+ , allowing the potential energy transfer from Cr to Nd 22 . However, no any NIR phosphorescence in the second NIR window can be observed in Cr/Nd-codoped ZnGa 2 O 4 LPPs (Fig. 2a). In fact, the desired phosphorescence is still absent in Nd 3+ singly doped ZnGa 2 O 4 phosphor after ceasing the excitation (Supplementary Fig. S2). It is notable that the diffuse reflection spectrum consists of the characteristic transition bands centered at 530, 588, 688, 748 and 808 nm, respectively, corresponding to Nd 3+ f-f transition, in Nd 3+ doped ZnGa 2 O 4 phosphor (Supplementary Fig. S3) 21 ; yet under the excitation at 748 nm, emission peak at 1064 nm, attributed to Nd 3+ [ 4 F 3/2 → 4 I 11/2 ] transition is not identifiable (Supplementary Fig. S4). This attractive optical quenching-phenomenon of luminescence and phosphorescence may be not concerned with the trap distribution, but the microcosmic network architecture. In ZnGa 2 O 4 , a majority of [Ga VI ] cations occupy octahedral sites, whereas all of the [Zn IV ] cations occupy tetrahedral sites 23 . As a preliminary conjecture, Cr 3+ has proven its strong ability to substitute for Ga 3+ in distorted octahedral coordination, whereas Nd 3+ cannot be effectively introduced into this specific network configuration (inset of Fig. 2a). In order to identify this possibility of ion doping, we focus on the intricate topochemical coordination geometry of Cr and Nd ions in zinc gallate spinel. The elucidation is performed in detail by a combination of XRD data and 71 Ga solid state nuclear magnetic resonance (NMR) studies. XRD patterns of ZnGa 2 O 4 : xCr (x = 0.5%, 5%, 10% and 20%) and ZnGa 2 O 4 : xNd (x = 0.5%, 5%, 10% and 20%) phosphors were measured and shown in Fig. 2b, Supplementary Fig. S5, S6. The peaks in XRD patterns of all Cr-doped samples are well indexed to pure ZnGa 2 O 4 spinel structure (JCPDS 86-0848). In stark contrast, the higher doping content (up to 5%) of Nd ion gives rise to an impure phase, NdGaO 3 (JCPDS, 70-3810) in Nd-doped samples. Another interesting phenomenon, i.e. XRD dominated peak (PI in Fig. 2b) shifting towards to higher 2θ value with the increment of Cr content, reveals a small linear variation in ZnGa 2 O 4 unit cell lattice parameter with Cr 3+ substitution, whereas no any shift of same peak is observed in Nd-doped samples, further ensuring the distinct phase splitting (Fig. 2c). In addition, a decline of the peak intensity in Fig. 2c also is present. Nevertheless, the causes of this decline may be different and rooted from either the substitution or the phase splitting. NMR allows the observation of specific quantum mechanical and magnetic properties of atomic nucleus, as well as provides the detailed information about the structure, dynamics, reaction state, and chemical environment of molecules 24 . Many scientific techniques exploit NMR phenomena to cover the interplay between the ligands and geometric centers, as well as study the topological network motif in crystals, microcrystalline powders, or anisotropic solutions, etc 25 . 71 Ga solid-state NMR is famous for the permission of quantitative analyses to different Ga 3+ central coordination state in inorganic solids 26 . Figure 2d,e shows the systematical physical investigations of Ga coordination geometry in Cr and Nd singly doped ZnGa 2 O 4 , respectively. For the undoped ZnGa 2 O 4 samples, 71 Ga NMR spectra exhibit two well-resolved resonances. The relative higher intensive signal at about 31 ppm is characteristic of sixfold coordinated Ga atoms, and the other weaker one ~at 170 ppm corresponds to Ga atoms in the tetrahedral sites of the spinel structure 26 . It is necessary to mention that with increasing Cr content (from 0.5% to 10%), 71 Ga NMR spectra present a significant broadening of spectral lines (Fig. 2d). In prominent contrast, scarcely any distinct influence on NMR spectral lines can be found by varying the Nd doping content in solid NMR spectra of ZnGa 2 O 4 : xNd (x = 0.5% and 10%) phosphors (Fig. 2e). The clear separation of NMR chemical shift at ~31 ppm between the two samples implies the precipitation of Cr 3+ into the octahedral lattice site and the excludability of local configuration to Nd 3+ ions. The NMR results are in accordance with XRD data, offering a powerful structural evidence to explain the interesting phenomena of phase splitting and luminescence quenching. Actually, rare-earth elements generally form complexes which have high coordination numbers (CNs) and weak metal-ligand bonds, because of their large ionic radii and relatively low oxidation states 27 . Typically transition-metal and main-group elements have coordination numbers 2-6, while rare-earth metals have CNs > 6 28 . The resulting coordination polyhedra include trigonal prisms (CN = 6) or its variation by stepwise capping of the prism face up to CN = 9, in addition to square antiprisms (CN = 8); Coordination number 3 is realized only under extreme conditions 28 . Therefore, to supply an ideal dwelling for Nd 3+ , a suitable material system should be proposed. Alkaline-earth metals have large ionic radii and various coordination-numbers 3-8 in different hosts, which ensure the selection of alkaline-earth gallates 29 . CaGa 2 O 4 has a similar spinel crystal structure with ZnGa 2 O 4 . In CaGa 2 O 4 , [Ca VI ] cations occupy octahedral sites 29 . This configuration thus features a path of easy doping ion precipitation into the octahedral [Ca VI ] under the condition of matching geometrical lattice and atomic radius, which occurs with rare earth ion, Nd. As expected, Fig. 3a exhibits the characteristic transitions of Nd 3+ in Nd singly doped CaGa 2 O 4 phosphor. However, the idealistic and aspirational long persistent phosphorescence is still absence in Cr singly, Nd singly and Cr/Nd doped CaGa 2 O 4 phosphors, respectively (Supplementary Fig. S7). A possible cause of this problem is due to the lack of effective traps (Fig. 3b). In sharp contrast to Cr/Nd codoping ZnGa 2 O 4 , photoluminescence excitation (PLE) spectrum monitored at 1064 nm of Cr/Nd codoping CaGa 2 O 4 sample consists of two specific excitation bands centered at ~410 and ~620 nm, in addition to Nd 3+ characteristic f-f transitions (Fig. 3c), indicating an energy transfer from Cr 3+ to Nd 3+ . Obviously, the strong one is attributed to the Cr 3+ [ 4 A 2 → 4 T 1 ], while the weak one corresponds to Cr 3+ [ 4 A 2 → 4 T 2 ] 30 . Further verification of energy transfer between Cr 3+ and Nd 3+ is supplied by emission spectrum and decay curve monitored at 1064 nm under the excitation wavelength at 410 nm (Fig. 3c and Supplementary Fig. S8). A possible channel of energy transfer from Cr 3+ to Nd 3+ is Cr 3+ [ 4 T 2 → 4 A 2 ]: Nd 3+ [ 4 I 9/2 → 4 F 5/2 ], [ 4 I 9/2 → 4 F 7/2 ], or [ 4 I 9/2 → 4 F 9/2 ], depending on the overlap between Cr 3+ emission band and Nd 3+ absorption band (Fig. 3d) 31 . As discussed above, due to the similar atomic radius and geometric configurations, Nd ions can easily precipitate on Ca lattice site in CaGa 2 O 4 , enabling the distinct photoluminescence (PL). To probe the lattice configuration and substitution progress in CaGa 2 O 4 , we performed XRD and solid state NMR experiments. X-ray diffraction pattern first confirms the crystallization of Nd-doped calcium gallate (Fig. 3e). In contrast to Nd-doped ZnGa 2 O 4 , all Nd-doped CaGa 2 O 4 samples can be indexed as standard phase CaGa 2 O 4 (JCPDS 16-0593). There is no any apparent observation of phase splitting from XRD data, even under a higher doping content of Nd 3+ , firmly supporting the rational inclusion of Nd 3+ into an inert matrix, CaGa 2 O 4 . This result is also supported by 71 Ga solid state NMR spectra. In contrast to ZnGa 2 O 4 host, the undoped CaGa 2 O 4 sample has a dominant chemical shift at 170 ppm (Fig. 3f). With increasing dopants content, CaGa 2 O 4 : Nd also has the same effect of NMR resonances' line broadening and the linear increase of NMR resonances integrated intensity, strongly suggesting the successful substitution in substantial amounts of Nd into Ca lattice site. Seemingly, as the individual backbone, MGa 2 O 4 (Zn and Ca) polymorph is chosen as the prototypical coordination network for its respective ability to engineer the functionally independent tunnel, traps(C) → activator(A), or activator(A) → activator(B), used to transfer the required energy. The only regret is the fundamentally missing connection of traps(C) → activator(A) → activator(B) in a separate material system. To address this issue, we anticipate a novel solid-solution Zn 1-x Ca x Ga 2 O 4 to bridge a new channel for transferring the persistent energy from traps to desired ions, based on the cautious consideration for crystal structure, ion valency and chemical bond relating to hosts and dopants. The desired NIR phosphorescence at 1064 nm is finally present in the afterglow spectra of Zn 1-x Ca x Ga 2 O 4 (x = 0.1, 0.3, 0.4 and 0.5) solid-solution (Fig. 4a). Significantly, we also observe a strong dependence (i.e. rising first followed by a decline) of phosphorescent peak intensity and decay dynamics on Ca concentration in Fig. 4a. We attribute this special spectral change of Nd 3+ to the successful persistent energy transfer from Cr 3+ to Nd 3+ , which is supported by the meticulous spectral studies of Nd 3+ in an optimal Zn 0.6 Ca 0.4 Ga 2 O 4 : 0.5Cr/0.5Nd solid-solution: PLE band at 410 nm should be assigned to Cr 3+ transition [ 4 A 2 − 4 T 2 ], while a distinct NIR PL peak at 1064 nm is observed under the excitation at 410 and 600 nm (Fig. 4a and Supplementary Fig. S9). The additional support for the formation of an unrestricted energy tunnel, traps → Cr 3+ → Nd 3+ , is the analysis of kinetic processes in Z 0.6 C 0.4 GO: 0.5%Cr/xNd (x = 0, 0.5%, 1% and 2%) samples (Fig. 4b). PL decay dynamics study of Cr 3+ shows a notable shortening in decay lifetime from 7.8 (Z 0.6 C 0.4 GO− 0.5Cr), to 7.59 (Z 0.6 C 0.4 GO− 0.5Cr0.5Nd), to 7.32 ms (Z 0.6 C 0.4 GO− 0.5Cr2Nd), giving clear evidence of successfully simultaneous precipitation of two activators into the corresponding lattice along with the effective energy transfer from Cr 3+ to Nd 3+ . It should be noted that, to the best of our knowledge, this type of NIR long-persistence phosphorescence has not been previously reported to occur in hybrid coordination networks by engineering cascaded energy transfer channels. Such substantial progress is strongly influenced by two key attributes; one is trap distribution and another is network architecture. Apparently, the variation of trap distribution may be not a crucial factor in exploring the nature of transfer channel, because the indispensable electron reservoir is still steadily embedded in all the Zn 1-x Ca x Ga 2 O 4 solid-solutions (Fig. 4c). To probe the evolution of topological network-dependent topochemical coordination, the systematic characterization, such as, XRD, solid NMR, EDX mapping and Raman spectra should be conducted 32 . XRD peaks in Z 1-x C x GO (x = 0.1, 0.4, 0.5 and 0.7) samples indicate their ZnGa 2 O 4 spinel solid-solution nature, while the superimposed peaks in samples Z 0.5 C 0.5 GO and Z 0.3 C 0.7 GO can be well indexed by the diffraction peaks of ZnGa 2 O 4 and CaGa 2 O 4 (Fig. 4d, and Supplementary Fig. S10). EDX mapping analysis reveals the solid-solutions have uniform distribution of Ca elements in all of the spinel solid-solutions (Supplementary Fig. S11, S12 and Fig. 4e). EDX experimental composition approximating the theoretical value supports the successful inclusion of Ca elements in spinel crystals (Supplementary Table S2). ## 71 Ga NMR spectra have provided some insights into the coordination variation of Ga center in ZnGa 2 O 4 and CaGa 2 O 4 phosphors, due to the incorporation of Cr and Nd. it is also expected to manifest its ability in resolving the question of topochemical configuration's evolution process, as the addition of Ca element. As shown in Fig. 4f, with increasing Ca content (0, 0.2, 0.4 and 0.7), two resonances at 170 and 31 ppm in 71 Ga NMR spectra increasingly present the linear broadening. In these solid solutions, Zn-O and Ga-O tetrahedron could suppress the intrusion of Ca element due to the mismatch of coordination configuration. In fact, to steady Ca ion, parts of Ga-O octahedron must reorient to form the new polyhedron network Ca-O octahedron along with the transformation from Ga-O octahedron to Ga-O tetrahedron due to the decrease of Zn-O tetrahedron. In detail, for the samples Z 1-x C x GO (x = 0, 0.2 ,0.4), the motion of local hybrid coordination-networks evolution include: (1) the precipitation of Ca on the lattice site of octahedron Ga, giving rise to the broadening of NMR resonance at 164 ppm; (2) the conversion from Ga-O octahedron to Ga-O tetrahedron, resulting in the enhancement of NMR resonance at 65 ppm. This interesting redeployment of network configuration thus permits the modification of topochemical state of dopants, as well as opens the possibility of bridging cascaded channels to transfer the persistent energy. To further validate the research idea aiming at the network configuration, Raman spectra of the fabricated samples also can be selected as the pertinent tool to further analyze the evolution of network architecture (Fig. 4g and Supplementary Fig. S13) 33 . In stark contrast to samples ZGO-0.5Nd and ZGO-5Nd, normalized Raman spectra of samples CGO-0.5Nd and CGO-5Nd do not exhibit the notable Raman peak shift and variation of Raman peak intensity, indicating a strong constraint of topological network to the migration of Nd ions in CaGa 2 O 4 . In fact, only two distinct Raman bands at ~1358 and 1434 cm −1 are present in the Raman spectrum of CGO-0.5Nd, while the Raman spectrum of ZGO-0.5Nd includes three identifiable Raman peaks at ~1341, 1389 and 1425 cm −1 . Thus, Raman spectra of Z 1-x C x GO (x = 0.1, 0.4 and 0.7) solid-solutions consequentially show a unit number decrease of Raman peaks with the increment of Ca content (inset of Fig. 4g). The variation of middle peak at 1401 cm −1 as a function of Ca doping content ensures the strong signature of the hybrid network structure, which is in accordance with the XRD and solid-state NMR data. In summary, we report a principle of bridging cascaded energy transfer channels to activate long persistent phosphorescence in the second biological window and fabrication of novel near-infrared phosphorescent phosphor Cr/Nd codoped Zn 1-x Ca x Ga 2 O 4 solid-solutions. Structural studies offer the powerfully fundamental evidences to explain the closed energy transfer channel from Cr 3+ to Nd 3+ in ZnGa 2 O 4 phosphor and invalidation of electronic reservoir in CaGa 2 O 4 phosphor. We believe that the ingenious solid-solution technology featuring the superiority of engineering a hybrid coordination-network opens new paths for advanced dynamic management of activation energy and gives the inspiration to design future new-wavelength, NIR phosphorescent phosphors by energy transfer. ## Methods Materials. 4N pure CaCO 3 , Ga 2 O 3 , ZnO, Nd 2 O 3 and Cr 2 O 3 were selected as the raw materials. Preparation of ZnGa 2 O 4 : xCr/yNd. Phosphors with molar compositions of ZnGa 2 O 4 : xCr/yNd (x = 0, 0.5%, 5%, 10%, 20%; y = 0, 0.5%, 5%, 10%, 20%), (Supplementary Table S1) were prepared by the solid state reaction method. The reaction included a two-step thermal treatment (i.e., initial calcination at 800 °C for 5 h, secondary calcination at 1350 °C for 3 h). Preparation of CaGa 2 O 4 : xCr/yNd. Phosphors with molar compositions of CaGa 2 O 4 : xCr/yNd (x = 0, 0.5%; y = 0, 0.5%, 5%, 10%), (Supplementary Table S1) were prepared by the solid state reaction method. The reaction included a two-step thermal treatment (i.e., initial calcination at 800 °C for 5 h, secondary calcination at 1200 °C for 3 h). Preparation of Zn 1-x Ca x Ga 2 O 4 : 0.5Cr/yNd. Phosphors with molar compositions of Zn 1-x Ca x Ga 2 O 4 : 0.5Cr/yNd (y = 0, 0.5%, 1%, 2%; x = 0.1, 0.2, 0.3, 0.4, 0.5, 0.7), (Supplementary Table S1) were prepared by the solid state reaction method. The reaction included a two-step thermal treatment (i.e., initial calcination at 800 °C for 5 h, secondary calcination at 1350, 1350, 1300, 1300, 1270, 1250 °C for 3 h as a function of x, respectively). Characterization. The prepared materials were analyzed by X-ray diffraction (Cu/Kα ) to confirm the sole crystalline phase. Room-temperature photoluminescence (PL), photoluminescence excitation (PLE) spectra, afterglow spectra and decay curves were measured with a high-resolution spectrofluorometer (UK, Edinburgh Instruments, FLS920) equipped with a 500 W Xenon lamp as an excitation source, with a Hamamatsu R928P visible photomultiplier (PMT) (250-850 nm) and a liquid nitrogen-cooled Hamamatsu R5509-72 NIR PMT as the detectors. TL glow curves and TL excitation (TLE) spectra were measured with a FJ-427A TL meter (China, Beijing) to characterize defect properties. Unless otherwise mentioned, the samples were pre-annealed at 600 K before testing, and some measurements were taken after pre-irradiating the samples for 10 min by using a xenon lamp. EDX images are characterized by a field emission scanning electron microscopy (FE-SEM), Nova NanoSEM 430. 71 Ga Hahn echo NMR experiments were performed on Bruker Avance III spectrometers operating at magnetic fields of 111.4 T corresponding to 71 Ga Larmor frequencies of 152.54 MHz) using Bruker 2.5 mm triple and double resonance probe heads. The 90 0 degree pulse length is 1.25 μ m with a recycle delay of 8s. 71 Ga chemical shifts were referenced relative to a 1.0 M aqueous solution of Ga(NO 3 ) 3 . All 71 Ga spectra were fitted using the Dmfit software. Raman spectra were collected with a Renishaw inVia Raman microscope irradiated by a visible laser at 532 nm.

chemsum

{"title": "Hybrid coordination-network-engineering for bridging cascaded channels to activate long persistent phosphorescence in the second biological window", "journal": "Scientific Reports - Nature"}

electronic_differentiation_competes_with_transition_state_sensitivity_in_palladium-catalyzed_allylic

## Abstract: Electronic differentiations in Pd-catalyzed allylic substitutions are assessed computationally from transition structure models with electronically modified phospha-benzene-pyridine ligands. Although donor/acceptor substitutions at P and N ligand sites were expected to increase the site selectivity, i.e. the preference for "trans to P" attack at the allylic intermediate, acceptor/acceptor substitution yields the highest selectivity. Energetic and geometrical analyses of transition structures show that the sensitivity for electronic differentiation is crucial for this site selectivity. Early transition structures with acceptor substituted ligands give rise to more intensive Pd-allyl interactions, which transfer electronic P,N differentiation of the ligand more efficiently to the allyl termini and hence yield higher site selectivities. ## Introduction Palladium-catalyzed allylic substitutions allow very selective and mild allylations of C-,N-and O-nucleophiles. The selectivity derives from steric and electronic properties of substrate and catalyst structures. "Side arm guidance" of nucleophiles with multifunctional phosphinoferrocenes or "chiral pockets" in C 2 -symmetric diphosphanes based on 2-(diphenyl-phosphino)benzoic acid amides were applied especially successfully. Chiral P,N-ligands (e.g. phosphinooxazolines, phox) provide in addition to steric control the possibility for "electronic differentiation", originating from the trans-influence of different donor atoms. Nucleophiles (e.g. dimethylmalonate) normally favour addition to the "trans to phosphorus" position at the Pd-η 3 -allylic inter-mediate (Scheme 1). This "trans to P" rule is supported by X-ray and computational analyses of Pd-η 3 -allylic intermediates, which exhibit longer and hence weaker Pd-C allyl bonds trans to P (i.e. the stronger π-acceptor vs. N) and hence are more susceptible to nucleophilic attack (Scheme 1). This electronic differentiation contributes to the high selectivity in Pd-catalyzed asymmetric allylic substitutions and provides also an explanation for α-memory effects. Computational model systems for P,N-ligands, i.e. PH 3 and para-substituted pyridines, have shown that cis-trans differentiations, i.e. the electronic site selectivity, of nucleophilic additions to Pd-η 3 -allylic intermediates is highest for electron poor pyridine ligands. Scheme 1: Electronic and steric differentiations provide the basis for the high selectivity of P,N-ligands in Pd-catalyzed allylic substitutions. Effects are studied with P-N-model ligands with para-substituted, coplanar phosphabenzene and pyridine moieties. ## Table 1: Activation (E a ) and reaction energies (E r ) reflecting electronic differentiations in transition structures (ΔE a cis-trans ) and Pd-ene products relative to Pd-allyl and NH 3 reactands (pb = phosphabenzene; py = pyridine moieties) [a] To further explore origins of site selectivities based on electronic differentiations in Pd-catalyzed allylic substitutions, we here employ a more advanced model system with phosphabenzene, and pyridine moieties for the crucial step of Pd-catalyzed allylic substitutions. Both P-and N-coordination sites are tuned electronically with para-substituents to reveal energetic and geometrical effects on cis-vs. trans-additions of nucleophiles to the Pd-η 3 -allylic intermediates (Scheme 1). ## Results and Discussion Electron donating or withdrawing groups (i. e. X, Y = HNMe, H, NO 2 ) in para-positions of phosphabenzene (X) and pyridine (Y) units tune electronic characteristics of P,N-ligand models in Pd-catalyzed allylic substitutions (Scheme 1). The phosphabenzene and pyridine moieties are linked via C ar -C ar bonds and a methylene bridge retains planarity and limits conformational flexibility. NHMe rather than higher substituted NMe 2 was employed as donor group, to retain lp-aryl conjugation. Ammonia serves as model nucleophile and attacks the Pd-η 3allylic intermediate cis or trans to phosphorus. This cis vs. trans site selectivity is employed as measure for electronic differentiation induced by the ligand system (Scheme 2). The lowest activation energies (E a , Table 1) for ammonia addition to the Pd-η3-allylic intermediate are apparent for strong electron withdrawing para-substituted phosphabenzene and pyridine units, i.e. X, Y = NO 2 (Figure 1 and Figure 2, E a trans = 2.19, E a cis = 2.52 kcal mol -1 , Table 1). The highest activation energies result from electron donating amino groups X, Y = NHMe (Figure 3 and Figure 4, E a trans = 10.67, E a cis = 10.47 kcal mol -1 , Table 1, Scheme 2). Such electronic tunings of the ligands strongly affect the reactivity and give rise to increased or decreased electrophilicity of Pd-allyl intermediates. The reaction energies (E r ) for ammonia addition to the Pd-η3allylic intermediate show a similar preference: Pd-ene-adduct formation is favoured most for X, Y = NO 2 (E r trans = 0.29, E r cis Scheme 2: Activation (ΔE a ) and reaction (ΔE r ) energies (kcal mol -1 ), computed for the P,N-ligand model with tuneable electronic differentiation. = -0.25 kcal mol -1 ) and becomes most unfavourable (i.e. endothermic) for X, Y = NHMe (E r trans = 10.98, E r cis = 10.33 kcal mol -1 , Table 1, Scheme 2). This points to a more π-donating character of the ene product relative to the allyl-cation reactant. In agreement with the "trans to phosphorus" rule, attack of ammonia is preferred for most X, Y combinations trans to P, due to the stronger π*/σ* acidity at P in phosphabenzene relative to N in pyridine (Table 1). Surprisingly however, this electronic site selectivity, as it is measured from relative energies of the transition structures (ΔE a TS ), is not largest for different X, Y donor-acceptor combinations (Figure 5, Figure 6, Figure 7 and Figure 8), but is highest for X and Y = NO 2 (ΔE a TS = 0.33 kcal mol -1 , Table 1). Likewise, the smallest electronic site "trans to P" selectivity is not found for X, Y donoracceptor combinations, but for strong donating X and Y = NHMe. Here, the selectivity is so low, that it even inverts to "cis to P" (ΔE a TS = -0.20 kcal mol -1 , Table 1). For each phosphabenzene moiety with X = H or NHMe or NO 2 , the "trans to P" site selectivity ΔE a TS increases for pyridine substituents Y in the order NHMe < H < NO 2 (Figure 9, Table 1). Hence, there is apparently an additional effect, which controls the site selectivity ΔE a TS besides the electronic donor vs. acceptor properties of different ligand atoms, i.e. P vs. N. Via this effect; electron withdrawing groups (e.g. NO 2 ) give rise to the highest site-selectivities. NO 2 -substituted ligands give rise to earlier transition structures with longer (forming) H 3 N-C α bonds (Table 2, Figure 1 to Figure 8), e.g. trans-TS with X = Y = NO 2 : H 3 N-C α = 2.04 (Figure 1). In contrast, amino-donor substitution leads to later transition structures with shorter H 3 N-C α distances, e.g. trans-TS with X = Y = NHMe: H 3 N-C α = 1.866 (Figure 3). This agrees with the more electrophilic properties of cationic Pd-allyl intermediates induced by electron withdrawing ligands. These positions on the reaction coordinate indeed correspond to the site selectivity of the transition structures, i.e. ΔE a TS : earlier transition structures have higher, later transition structures exhibit lower "trans to P" selectivities (Figure 10). The distance between Pd and the allylic systems decreases from early (allyl cation like) to late (ene like) positions on the reaction coordinate. A closer, more intense Pd-C α contact (e.g. 2.674 , Figure 2, Table 2) stronger delivers electronic differentiation of the ligand, and hence "trans to P" selectivity. Hence, higher electronic site selectivity closely corresponds to intense Pd-allyl interactions with short Pd-C α distances (Figure 11). Apparently, the positions on the reaction coordinate influence the site selectivity even stronger than the electronic differentiation between P and N ligand atoms: No substitution (X = Y = H) gives rise to even higher ΔE a TS than more pronounced electronic differentiations with X, Y = NO 2 or NHMe (Figure 11), due to higher TS-sensitivity originating from closer Pd-allyl contact. ## Conclusion In Pd-catalyzed allylic substitutions, the electronic site selectivity, i.e. the preference for "trans to P" addition, is affected by the intrinsic electronic differentiation of the ligand atoms, e.g. P vs. N. However, the sensitivity for this electronic differentiation depends on the intensity of the Pd-allyl interaction. A close Pd-allyl distance in an early, allyl cation like transition structure delivers the electronic differentiation of the ligand system more efficiently to the allylic termini (C α ) than a more distant Pd-allyl (more ene like) unit of a late transition structure. Electron withdrawing (e.g. NO 2 ) substituents in the ligand system generate earlier transition structures with more intense Pd-allyl interactions and higher sensitivity for electronic differentiations. Hence, both intrinsic electronic differentiation in the ligand and high TS-sensitivity appear to be crucial for high site-selectivity in Pd-catalyzed allylic substitutions. ## Computational details All structures were fully optimized and characterized by frequency computations as minima or transition structures using Gaussian 03 with standard basis sets and the B3LYP hybrid-DFT method. Zero point energies and thermochemical analysis were scaled by 0.9806.

chemsum

{"title": "Electronic differentiation competes with transition state sensitivity in palladium-catalyzed allylic substitutions", "journal": "Beilstein"}

complex_loop_dynamics_underpin_activity,_specificity_and_evolvability_in_the_(βα)8_barrel_enzymes_of

## Abstract: Enzymes are conformationally dynamic, and their dynamical properties play an important role in regulating their specificity and evolvability. In this context, substantial attention has been paid to the role of ligand-gated conformational changes in enzyme catalysis; however, such studies have focused on tremendously proficient enzymes such as triosephosphate isomerase and orotidine 5'-monophosphate decarboxylase, where the rapid (μs timescale) motion of a single loop dominates the transition between catalytically inactive and active conformations. In contrast, the (βα)8-barrels of tryptophan and histidine biosynthesis, such as the specialist isomerase enzymes HisA and TrpF, and the bifunctional isomerase PriA, are decorated by multiple long loops that undergo conformational transitions on the ms (or slower) timescale. Studying the interdependent motions of multiple slow loops, and their role in catalysis, poses a significant computational challenge. This work combines conventional and enhanced molecular dynamics simulations with empirical valence bond simulations to provide rich detail of the conformational behavior of the catalytic loops in HisA, PriA and TrpF, and the role of their plasticity in facilitating bifunctionality in PriA and evolved HisA variants. In addition, we demonstrate that, similar to other enzymes activated by ligand-gated conformational changes, loops 3 and 4 of HisA and PriA act as gripper loops, facilitating the isomerization of the large bulky substrate ProFAR, albeit now on much slower timescales. This hints at convergent evolution on these different (βα)8-barrel scaffolds.Finally, our work highlights the potential of engineering loop dynamics as a powerful tool to artificially manipulate the diverse catalytic repertoire of TIM-barrel proteins. ## Introduction Enzymes are dynamic systems able to explore many different conformations, and these dynamical properties are clearly connected to their biological function. Examples of this include allosteric regulation and product release, 1 as well as the role of conformational selection in enzyme catalysis, 2-7 promiscuity 4 and evolution. In addition, such conformational dynamics can, in principle, be engineered in a targeted fashion to allow enzymes to acquire new catalytic functions and/or physiochemical properties. 8,9, Understanding how enzymes manipulate and modulate conformational dynamics during both natural and directed evolution is an important step in this direction. In particular, understanding the dynamical behavior of decorating loops that cover enzyme active sites is important as such loops can regulate substrate selectivity, the evolution of new activities, and potentially also turnover rates. 3, As such, targeting the dynamics of such active site loops is attractive from an engineering perspective, 16,33 and therefore there is substantial interest in understanding loop dynamics and its impact on selectivity and catalysis. In this context, there have been extensive studies of a wide-range of enzymes, such as triosephosphate isomerase (TPI), 34,35 orotidine 5'-monophosphate decarboxylase, (OMPDC) 36 glycerol 3-phosphate dehydrogenase (GPDH), 37,38 1-deoxy-D-xylulose-5-phosphate reductoisomerase, 39,40 and β-phosphoglucomutase, 41 which have been demonstrated to be activated by ligand-gated conformational changes. Specifically, interactions between a key "gripper" loop decorating the active site and the non-reactive phosphodianion groups of the substrates of these enzymes trigger substantial conformational changes in the gripper loop, facilitating energetically unfavorable transitions from catalytically inactive open to catalytically active closed conformations, and these conformational transitions are central to the catalytic activities and high proficiencies of these enzymes. 22,31 It is noteworthy that several of the aforementioned enzymes have TIM-barrel folds. This fold comprises eight repeated (βα)-units, and most if not all TIM-barrel proteins possess decorating loops, 18,42 the conformational diversity of which likely plays an important role in regulating specificity and function. 25,28 These flexible loops can vary in length, 18 but are typically used to bind substrate, and to sequester the active site from solvent by closing over the active site, and it has been suggested that the active site geometries of these enzymes are shaped by the residues of these loops. 43 However, these well-characterized examples of proteins activated by ligand-gated conformational changes all focus on the roles and importance of single loops, such as gripper loop 6 in TPI. Studying the ligand-gated motion of a single loop can already pose substantial challenges; 44 systems with multiple active site loops undergoing substantial conformational changes are even more complex, and therefore unsurprisingly understudied in the literature. We have sought to address this gap in knowledge by studying active site loop dynamics in the (βα)8-barrels of tryptophan and histidine biosynthesis. The isomerase enzymes HisA, TrpF and PriA are model systems for the evolution of specificity and activity. 28, As shown in Figure 1, HisA catalyzes isomerization of the aminoaldose N'-[(5'-phosphoribosyl)-formimino]-5aminoimidazole-4-carboxamide-ribonucleotide (ProFAR) into the aminoketose Nʹ-[(5ʹphosphoribulosyl)-formimino]-5-aminoimidazole-4-carboxamide-ribonucleotide (PRFAR). TrpF catalyses the same Amadori rearrangement on N-(5ʹ-phosphoribosyl)anthranilate (PRA), producing 1-(2-carboxy-phenylamino)-1'-deoxyribulose-5'-phosphate (CdRP). This rearrangement proceeds via a Schiff acid-base mechanism, that utilizes aspartate (and in the case of TrpF) cysteine residues as acid-base pairs. 49 Interestingly, many actinobacteria lack the trpF gene, instead possessing a gene for a bifunctional isomerase, PriA. 50 The PriA from Mycobacterium tuberculosis (MtPriA) has been particularly well characterized, and has kcat/KM values of ~10 4 M -1 s -1 for HisA activity and ~10 5 -10 6 M -1 s -1 for TrpF activity. 25,51 Not only are there naturally occurring bifunctional enzymes, but promiscuous TrpF activity has been detected on both ancestral and extant specialist HisA enzymes, with kcat values ranging from 10 -4 to 10 -2 s -1 . 52 HisA has also been converted into TrpF by directed evolution, 53 and in serial passaging experiments. 46 In the latter study, laboratory evolution of the ProFAR-specific HisA (lacking TrpF activity) from Salmonella enterica over 3,000 generations yielded an extensive suite of mutations in the S. enterica HisA (SeHisA) that resulted in specialist HisA enzymes, specialist TrpF enzymes and PriA-like bifunctional enzymes. green, yellow and dark red on each structure, respectively. Loop 1 in TmTrpF is short (four residues), which is why no corresponding loop is annotated on this panel. Note that for clarity, N7D and A176D reversions were applied to the structure of SeHisA in complex with ProFAR (these reversions were also applied in our simulations, as described in the Methodology section). (D) The proposed mechanism for the Amadori rearrangement leading to the isomerization of substrates ProFAR and PRA by the different enzymes. 49 As shown in Figure 1, HisA and PriA are decorated by three long catalytic loops, loops 1, 5, and 6 (or two analogous loops in the case of TrpF, which has lost most of loop 1). 25,28 Of these loops, loop 5 carries key residues that are important for substrate binding, loop 6 carries the catalytically important aspartic acid side chain, and a number of mutations important for interaction with substrate PRA have been observed at position 15 of loop 1. 28,45,54 The mutants generated by Näsvall et al. 46 were the subject of detailed structural and biochemical analyses. 28 As with PriA, 25 this analysis of the evolved bifunctional SeHisA variants indicated that the bifunctionality is driven by competition between not just the substrates ProFAR and PRA, but also between structurally distinct conformations of loops 1 and 5, in particular 28 (Figure S1). Furthermore, although the isomerization of both ProFAR and PRA proceeds through the same Amadori rearrangement (Figure 1), ProFAR (the native substrate of HisA) is a much larger molecule, including the presence of a second phosphate group. This group forms hydrogen bonding interactions with the side chains of R83 from loop 3 and S103 from loop 4 of SeHisA, and with the corresponding side chains of R85 and T105 in MtPriA (Figure 2). Although neither of these residues are on the primary mobile loops of either enzyme (Figure 1), nevertheless, the interactions with the second phosphate group of ProFAR are very similar to analogous interactions in other enzymes activated by ligand-gated conformational changes, suggesting that loops 3 and 4 in SeHisA and MtPriA may similarly act as "gripper loops" allowing these enzymes to attain relevant catalytically active conformations for the isomerization of the larger substrate. This effect would clearly not be present when the smaller substrate, PRA, is bound to the active site as this substrate lacks a non-reactive phosphodianion group to interact with these loops. the active site tryptophan that forms stacking interactions with the larger substrate ProFAR in SeHisA and MtPriA (W145 in both enzymes), as well as the key gripper residues that interact with the distal phosphate group of the larger substrate/product in SeHisA and MtPriA (R83 and S103 in SeHisA and R85 and T105 in MtPriA). Key hydrogen bonding interactions are also highlighted, using the distances () found in the corresponding crystal structures. Note that for clarity, N7D and A176D reversions were applied in SeHisA in complex with substrate ProFAR (and this reversion was also applied in our simulations, as outlined in the Methodology section). In the present work, therefore, we combine long-timescale conventional molecular dynamics, enhanced sampling and empirical valence bond (EVB) simulations to present a comprehensive computational study of a number of wild-type and variant forms of SeHisA, MtPriA and TrpF from Thermotoga maritima (TmTrpF). All variants studied in this work, and the corresponding structures used, are summarized in Table S1. We chose these systems because, in all three cases, there are high-quality structural data in unliganded and ligand-bound forms. For SeHisA, we chose 8 to study the unliganded and substrate (ProFAR) bound forms, 54 as well as key SeHisA variants from ref. 28 that were selected based on their specificity patterns (specialists vs. generalists, Table S2). MtPriA and TmTrpF were similarly selected on the basis of high-quality structural data of each enzyme in both unliganded and product (PRFAR) or product analogue (rCdRP) bound forms respectively, as summarized in Table S1. Prior simulation studies of TPI by both us and others have indicated that the large ligand-gated conformational change of the gripper loop 6 is correlated with smaller conformational motions in other decorating loops on the active site. 44,56 We made similar observations when studying loop motions in PTP1B. 32 However, as these simulations indicate, even studying the motion of one large dominating conformational change is computationally non-trivial, and the current systems involve the interdependent conformational rearrangements of multiple loops simultaneously. Our current simulations of HisA, PriA and TrpF (1) provide rich detail of the conformational behavior of the catalytic loops in the different systems, and (2) provide insight into the link between conformational dynamics, catalytic activity and functional evolution in the different enzymes, in particular the role of loops 1 and 5 in regulating PriA and HisA's activity and selectivity, as well as the gripper loops 3 and 4 in driving ligand-gated conformational changes in these enzymes. 28, This, in turn, is significant, because in recent years, there has been substantial (and increasing) interest in exploiting techniques such as loop grafting and related approaches in order to engineer flexible loops in enzymes as a means of controlling their catalytic activity. 16,33 Our data provide clear evidence that this is likely to be a powerful strategy for artificially manipulating the diverse catalytic repertoire 47 of TIM-barrel proteins. ## Methodology Methodological details are presented here in brief. Full details of all simulations and any nonconventional parameters used in our simulations are provided in the Supporting Information. ## System Preparation for Conventional and Enhanced Sampling Molecular Dynamics Simulations. Simulations were performed on wild-type SeHisA, MtPriA and TmTrpF, as well as relevant enzyme variants, in both their unliganded forms and in complex with various ligands (substrates ProFAR and PRA and, in the case of the enhanced sampling simulations, products PRFAR and CdRP). A total of fourteen crystal structures were used to generate starting points for these simulations, and all simulations performed as well as associated structures used are summarized in Table S1. Where present, the D7N, D11N and D176A substitutions were reverted to wild-type using the Dunbrack 2010 Rotamer Library, 57 as implemented in UCSF Chimera, v. 1.14. 58 Missing regions in the catalytic loops were reconstructed using Modeller v. 9.23. 59 The catalytic aspartic acid side chain in the active site of each enzyme (D176 in HisA, D175 in PriA and D126 in TrpF) was kept protonated in line with the mechanism shown in Figure 1. All other residues (except H50 in PriA, which was doubly protonated) were kept in their default protonation states at physiological pH determined by use of PROPKA 3.1, 60 and visual inspection. The substrates ProFAR and PRA were manually placed into the relevant active sites in the same conformation as found in the structure of the HisA wild-type enzyme in complex with ProFAR and in the case of PRA (Figure 1D), was placed by manual overlay of the reactive part of PRA with the reactive part of ProFAR, and with the carboxylate group of PRA keeping key interactions with active site residues. Partial charges for ligands ProFAR, PRA, PRFAR and CdRP, were calculated using the standard restrained electrostatic potential (RESP) protocol using Antechamber v. 17.3, 61 and based on the vacuum electrostatic potential calculated at the HF/6-31G(d) level of theory, using Gaussian 09 Rev. E.01. 62 All other simulation parameters were described using the general Amber force field 2 (GAFF2) 63 (see Tables S3 to S6). Finally, to keep the substrate stably bound in the enzyme active sites, weak distance restraints were applied to protein-substrate distances, as described in Table S7. ## Conventional Molecular Dynamics Simulations Conventional MD simulations were performed using the CUDA version of the PMEMD module of the AMBER 16 simulation package. 64 The protein, ligands and solvent were described using the ff14SB force field, 65 the General AMBER Force Field 2 (GAFF2), 63 and the TIP3P water model, 66 respectively. Following initial minimization and equilibration, each system (summarized in Table S1) was subjected to 10 x 500 ns of molecular dynamics simulations controlled by the Langevin thermostat with a collision frequency of 2 ps -1 , 67 and the Berendsen barostat with a 1 ps coupling constant. 68 This led to a cumulative 5 μs of production simulations per system, and a cumulative total of 70 μs of conventional MD simulations over all systems studied (Table S1). ## Enhanced Sampling Molecular Dynamics Simulations Steered molecular dynamics simulation (sMD) were performed using GROMACS 2018.4 in order to pull products PRFAR and CdRP out of the active site of the SeHisA(dup13-15/D10G/G102A/Q24L) variant, as described in the Supporting Information. The system preparation was performed as for the conventional MD simulations, and using the same force fields and water models as the conventional MD simulations. Following initial minimization and equilibration, 10 x 50 ns production MD simulations were performed on each system. The first 5 ns of production MD were unrestrained, after which an external force with a force constant of 10 kcal mol -1 -2 was applied to pull the product out of the active site. This external force was then released for the last 5 ns of the MD simulation run. ## Empirical Valence Bond Simulations Following our prior success in using the empirical valence bond (EVB) approach 69 to study a wide range of analogous ring opening reactions, such as lactone 70,71 and epoxide 72,73 hydrolysis, we extended this approach to study the enzyme-catalyzed opening of the ribose ring of substrates ProFAR and PRA (Figure 1), as catalyzed by wild-type and variant forms of HisA, PriA, and TrpF. Our focus for our EVB simulations was specifically on the ribose ring-opening reaction (the first step of the mechanism shown in Figure 1D), as motivated in the Results and Discussion, and described using the valence bond states shown in Figure 3. Simulations were performed on wild-type SeHisA, MtPriA and TmTrpF as well as selected variants, as described in the Supporting Information. All simulations were performed using the Q6 simulation package, 74,75 using the OPLA-AA force field. 76 All EVB parameters necessary to reproduce our work, as well as a detailed description of the computational methodology and subsequent simulation analysis can be found in the Supporting Information, with the full parameters used in our simulations updated to Zenodo (DOI: 10.5281/zenodo.5893598). Each system was simulated using 30 individual replicas, with each replica first equilibrated for 20 ns and the endpoint of that equilibration being used as the starting point for propagating an EVB trajectory. Each EVB free energy perturbation/umbrella sampling (EVB-FEP/US) 69 ## Analysis of Conventional and Enhanced Sampling Molecular Dynamics Simulations Unless stated otherwise, all analysis of all conventional and enhanced sampling molecular dynamics simulations was performed using CPPTRAJ. 77 Hydrogen bonds were defined as formed if the donor−acceptor distance was ≤3.0 and the donor-hydrogen-acceptor angle was within 180 ± 45°. Principal component analysis (PCA) was performed by first RMS fitting to a whole protein Ca carbon atoms and then performing PCA on the Cα carbon atoms of loops 1, 5 and 6, as well as loop 1 for HisA loop 1 elongated systems analysis. Other analyses were performed as described in the Supporting Information. 13 ## Active Site Plasticity and Substrate Binding in the Different Enzymes The TIM-barrel structures of SeHisA and MtPriA are similar, with an RMSD of 1.09 between them (comparing the Ca atoms in PDB IDs 3ZS4 55 and 5A5W 54,55 ). TmTrpF, in contrast, is a smaller enzyme with 40 fewer residues in the sequence than MtPriA and SeHisA. We performed 10 x 500 ns conventional molecular dynamics simulations of unliganded HisA, TrpF and PriA, and calculated the average and standard deviations of the active site volumes of each enzyme using the MDpocket 78 tool, which is provided as part of the fpocket 79 suite of pocket detection programs, as described in the Supporting Information. The resulting calculated volumes are shown in Table S8. We obtained average volumes of 745.4 ± 129.6 3 , 1033.8 ± 217.0 3 and 1173.5 ± 158.0 3 , for the active sites of TrpF, PriA and HisA, respectively during our simulations. From this, it can be seen that the TrpF active site is more compact than that of PriA and HisA, which have successively larger active site volumes, with more "flexible" pockets than TrpF (using the standard deviation on the volume as a proxy for this flexibility). For comparison, the substrates ProFAR and PRA have volumes of 829 and 559 3 , respectively, calculated using Alexander Balaeff's Mol_Volume program Version 1.0, with default radii of 1.7 and a probe sphere of 0.5. This confirms the structural data indicating that the active site pocket of TrpF is too compact to accommodate the much larger substrate, ProFAR, leading to the selectivity of this enzyme towards PRA. 80 Furthermore, the PriA active site is the most flexible of the three, in line with structural data 25 that indicates that PriA is capable of significantly rearranging its active site (in particular loop 5 conformation) when accommodating the different substrates ProFAR and PriA (Figure S1). Structures and mutagenesis experiments have identified two key active site side chains in HisA and PriA, which are important for binding of the substrate ProFAR. 25,28,54,81 These are W145, which forms a stabilizing stacking interaction with the substrate, and R83 (R85), which interacts with the second phosphodianion group of the substrate (Figure 2). To further explore the conformational diversity of these key tryptophan and arginine residues in HisA and PriA, respectively, we examined the joint dihedral angle distribution of the side chains of these residues in simulations of unliganded HisA and PriA, as well as HisA and PriA in complex with both From these data it can be seen that both the tryptophan and arginine side chains are highly conformationally flexible in the unliganded enzymes. However, while the binding of the substrate ProFAR to HisA restricts the conformational space of the arginine side chain on the "gripper" loop 3 to a catalytically competent position that helps stabilize the bound substrate, in PriA, the R143 side chain is only 4.8 from the "gripper" residue R85 (distance between the two side chain carbon atoms, based on PDB ID: 3ZS4 55 ). This in turn creates electrostatic repulsion between the two arginine side chains, thus destabilizing loop 5 as well as the interaction between the substrate ProFAR and the R85 side chain (Figure S3). In contrast, when PRA, that lacks the second phosphodianion group, is bound to the PriA active site, the R83/R85 side chains increase their conformational flexibility again, sampling more or less the same conformational space as in the case of the unliganded enzyme (Figure S2). Therefore, the interaction of these residues with the larger substrate ProFAR is likely playing an important role in the ability of these enzymes to bind and isomerize this compound. In the case of W145, this side chain slightly increases its conformation flexibility when the smaller substrate PRA is bound to HisA (note that PRA was placed manually in the active site by overlay with substrate ProFAR, as described in the Supporting Information). However, in the case of PriA, both W145 the and R143 side chains are conformationally restricted to a catalytically competent position due to a rearrangement of loop 5 upon PRA binding, that swaps the position between these two residues compared to when ProFAR is bound to PriA, preventing the electrostatic repulsion between the R85 and R143 side chains that is observed when ProFAR (PRFAR) is bound to the active site (Figure S1). 25 As noted previously, the R83/R85 side chain is one of a number of key residues on the "gripper" loop that interact with the distal phosphodianion group of the larger substrate ProFAR, contributing to the stabilization of the substrate in HisA active site. Hence, in PriA, the loop 5 rearrangement required for ProFAR substrate binding 25 prevents R85 from gripping the second phosphodianion group and thus showing clear preference for the isomerization of PRA substrate. We note the similarity of these gripper interactions to corresponding interactions in enzymes such as TIM, OMPDC and GPDH, where interactions with the non-reactive phosphodianion group of the substrate drives a ligand-gated conformational change. This in turn stabilizes otherwise energetically unfavorable but catalytically important closed conformations of key catalytic loops over the respective active sites of these enzymes. In the case of the current systems, the interaction between the HisA gripper residues and the remote phosphodianion group of ProFAR appears to be similarly important for maintaining the closed conformation of loop 5, and when this interaction is lost, as in PriA, we see corresponding opening of loop 5 (Figure S3). This supports the likelihood that HisA and PriA are also activated by ligand-gated conformational changes, albeit with more complex loop dynamics (due to the involvement of not one but three highly mobile and long catalytic loops) than in other previously characterized systems. ## Conformational Dynamics of Key Catalytic Loops of HisA, PriA, and TrpF To further explore the impact of the binding of the two substrates on loop dynamics, we extended our simulations to also included simulations of TrpF in both its unliganded form, and in complex with PRA (Table S1). We then performed Principal Component Analysis (PCA) to characterize the motion of the key catalytic loops during our simulations in each of the individual systems, similarly to prior analysis we have performed on triosephosphate isomerase, 44 except in our prior work substantial conformational changes take place in only one and not two (TmTrpF) or three (SeHisA; MtPriA) distinct loops. The PCA analysis was performed on the mass-weighted Cartesian coordinates of each enzyme compared to the coordinates of the corresponding closed state, allowing us to explore the variation of the conformations of these loops in coordinate space. We subsequently projected the free energies for each enzyme along the most dominant motions, PC1 and PC2, from simulations of each of HisA, PriA and TrpF in their unliganded forms as well as in complex with substrates ProFAR and PRA, respectively (here, the smaller substrate PRA was artificially placed in the HisA active site by manual overlay with the reactive part of ProFAR, as outlined in the Supporting Information). This allowed us to compare the free energy surfaces defined by these two principal components both between the different enzymes, and the effect of ligand binding on these surfaces. The resulting data are shown in Figure 6. Note that, as shown in Figure 5, these projected free energy surfaces show the combined motion of all key catalytic loops along each principal component. In the unliganded forms of all three enzymes, the catalytic loops can explore a range of "wideopen" conformations and are overall highly conformationally diverse (Figure 5). This observation is consistent with our prior simulation studies on both triosephosphate isomerase, 44 and the protein tyrosine phosphatases PTP1B and YopH, 32 as well as chimeric forms of these enzymes. 83 However, and consistently with structural data, the binding of ProFAR to HisA fully restricts the conformational sampling of all three loops (Figure 6B). In the case of PriA, the binding of both ProFAR and PRA also stabilizes the closed conformation of the three active site loops, but still allows for some conformational flexibility in these loops (Figure 6, based on both the topologies of the projected free energy surfaces, and the corresponding energies). In sharp contrast, in the liganded form of TrpF, our MD simulations show that PRA is not stable in the active site due to the flexibility of loop 6, which explores transitions towards open conformations, similarly to the unliganded system (Figures 5E, F and 6G, H). In the crystal structure of the unliganded enzyme (PDB ID: 1NSJ 55,82 ), this loop is present in a closed position but with missing density, whereas in our simulations of both the liganded and unliganded form of the enzyme, loop 6 samples open conformations, suggesting that it is not the correct closed state to stabilize the substrate PRA in the active site. We note that the structure used for these simulations (PDB ID 1LBM 49,55 ) was solved in complex with the product analog rCdRP. Our simulations suggest that the loop conformations observed in this structure are a conformational state on the trajectory to product release, rather than an ideal conformational state for stabilizing the Michaelis complex. In addition, in HisA and PriA, we observe the formation of a stacking interaction between the substrate ProFAR and the side chain of W145 in our conventional MD simulations (Figure S4), with an average distance of 3.9 ± 0.3 and an angle g = 11.7 ± 5.5° between the center of mass of the imidazole ring of ProFAR and the indole ring of the W145 side chain during our simulations. Our results confirm the role for W145 that was proposed previously, based on experimentally determined structures. 54 This is furthermore consistent with prior structural analysis that indicates that HisA activity is abolished in SeHisA(dup13-15), because the extended conformation of loop 1 blocks this side chain from interacting with ProFAR. 28 In contrast, in the case of PriA, and again in agreement with prior structural analysis, 25 we observe two possible conformations of loop 5, depending on which substrate is bound to the active site. That is, when ProFAR is bound to the active site, we sample a conformation similar to that observed in wild-type HisA (Figure S1A) with a similar stacking interaction between ProFAR and W145 (Figure S4), however, the loop 5 rearrangement required to optimize the stacking position of W145 with ProFAR creates transient electrostatic repulsion between loop 5 and the rest of the enzyme, making this loop more conformationally dynamic, which we observe in our analysis in the form of an increased standard deviation in the distance and angle of the corresponding stacking interaction (d = 4.1 ± 0.6 , g = 19.2 ± 13.9°) (Figures S3 and S4). The greater plasticity of this interaction, in turn, decreases substrate stability in the active site (the ProFAR RMSF increases from 18.4 in our simulations of wild-type HisA to 22.2 in our simulations of wild-type PriA), and thus the corresponding ProFAR isomerization activity of PriA. For comparison, in our simulations of PriA in complex with substrate PRA, we sample an active site conformation in which the R143 side chain (loop 5 residue) forms salt bridges with the side chains of with D130 and D175, and the arginine acts as a "shield" dampening the electrostatic repulsion between the D130 side chain and the anthranilate carboxylate group of PRA. This interaction also stabilizes the catalytic aspartic (D175), placing it in an optimal position for catalysis (Figure S5 and Table S9), as shown in previous studies 25 . This PriA conformation is similar to the "TrpF-active" conformation observed in the SeHisA(dup13-15/D10G/Q24L/G102A) crystal structure 28 (PDB ID: 5AB3 28,55 , Figure S6, with manual placement of PRA in the active site), where the arginine is close to residue D129. While we observe this conformation in our PriA simulations, we do not observe the formation of a corresponding interaction in our simulations of wild-type HisA in complex with PRA, the negative charge repulsion between PRA and the D129 side chain destabilizes the position of the substrate in the active site (Figure S7), as well as the stability of the loop 6 carrying the key catalytic aspartic acid side chain (Figure S7). We do, however, observe a similar interaction with the R169 side chain in the SeHisA(L169R) variant, with interactions with D129 and, in this case, a salt bridge interaction with the anthranilate carboxylate group of PRA (Figure S7 and Table S9), consistent with experimental work that demonstrated that the introduction of the L169R substitution in HisA induces TrpF activity (Table S2). 46 Finally, in the case of TrpF in complex with PRA, we observe a salt-bridge interaction between E184 and the side chain of R36 on loop 3 (Figure S6 and Table S9), and we can see that as for HisA and PriA, that R36 is again acting as a "shield" avoiding possible negative repulsion interactions between the substrate and the negatively charged side chain. However, we do not observe clear interactions between the R36 side chain and the substrate PRA (Figure S6 and Table S9, with the fraction of simulation time in which this interaction is observed being <0.1). Overall, we observed that this arginine plays an essential role in the introduction of TrpF activity, by shielding electrostatic repulsion between negatively charged side chains and the anthranilate carboxylate group of PRA. This is in agreement with experiments where introducing an arginine or removing the negative residue introduce TrpF activity in HisA systems. 46 ## Conformational Dynamics of Loop 1 in HisA and PriA and Its Impact on Selectivity While loops 5 and 6 of HisA and PriA have been clearly identified as being important for binding and catalysis (loop 6 carries the catalytic aspartic acid side chain, Figure 1), 25,28,54 the precise catalytic role of loop 1 remains unclear, although extending the conformation of loop 1 through duplication of residues 13-15 (HisA(dup13-15)) plays an important role in the acquisition of bifunctionality 28,84 in a real-time evolution experiment on HisA, 46 and substitutions at position 15 on this loop appear to be important for facilitating the TrpF activity of this enzyme. 28,54 Therefore, we also performed simulations of the SeHisA(dup13-15/D10G) variant, as described in the Methodology section. SeHisA(dup13-15/D10G) is a bifunctional enzyme that can catalyze the isomerization of both ProFAR and PRA with modest catalytic efficiencies, 28 and the corresponding crystal structure (PDB ID: 5AC7 28,55 ) shows the enzyme in a 'PRA-active' conformation with loops 1 and 6 in a closed state, and loop 5 in an open state. When initiating simulations of the unliganded SeHisA(dup13-15/D10G) variant starting from these loop 1 closed conformations, we did not observe any opening of loop 1. This is in contrast to our simulations of the wild-type enzyme, where we sampled open conformations of this loop when we started from the unliganded closed conformation observed in PDB ID: 5A5W, 54 removing the ProFAR substrate. This provides evidence, in addition to the discussion in ref. 28 , that the elongation of loop 1 heavily stabilizes this PRA-active conformation. This is further supported by examining the root mean square fluctuations of all Cα atoms in our simulations of these two enzymes (Figure 7), where we observe that loop 1 is more flexible than either of loops 5 or 6 in simulations of the wild-type enzyme, but has reduced flexibility in simulations of the SeHisA(dup13-15/D10G) variant. S2). Interestingly, while the elongation of loop 1 through the duplication of residues 13-15 (VVR) appears to be essential for the change of specificity towards the isomerization of PRA (facilitated by the presence of a new stabilizing arginine side chain close to the active sites, Figure S1, 28 simply the loop duplication by itself is not enough to induce bifunctionality. That is, while the duplication elongates loop 1, it also rigidifies it, such that SeHisA(dup13-15) does show some ability to isomerize PRA (k cat > 0.15 s -1 ), but at the expense of losing all ability to isomerize the larger substrate ProFAR (no detectable activity). 28 Therefore, the duplication by itself simply leads to a switch in activity from a modestly efficient isomerase towards ProFAR (k cat 7.8 s -1 ) towards a less efficient isomerase with activity towards PRA. Critical to the bifunctionality is the inclusion of an additional substitution, present in both variants studied above, namely D10G. This substitution increases the flexibility of the elongated loop 1 (Figure 8), allowing for the loop to take on wide-open configurations which in turn facilitate the entry and binding of ProFAR to the active site (Figure 8B, wide-open conformation). Our conventional MD simulations (10 x 500 ns per system) are overall short, taking in particular into account the slow turnover numbers of these enzymes (that suggest loop motions on the ms to s timescale). 28 However, our observations that the D10G substitution leads to increased flexibility of loop 1 are in good agreement with prior NMR relaxation dispersion experiments. 28 These detected μs to ms motions at 14 backbone 15 N positions in the SeHisA(dup13-15/D10G) variant, compared to only three positions for the SeHisA(dup13-15) variant, and with two resonances that are unique to SeHisA(dup13-15/D10G). As a result, adding this substitution is sufficient to convert SeHisA(dup13-15/D10G) back to a bifunctional enzyme, through exploitation of conformational dynamics, with k cat of 0.09 s -1 for the isomerization of PRA, and 0.05 s -1 for the isomerization of ProFAR (Table S2). 28 To explore this further, we therefore examined simulations of four loop elongated variants, specifically: dup13-15 (PDB ID: 5G2I 28 ), dup13-15/D10G (PDB ID: 5AC7, 28,55 ), dup13-15/D10G/G102A (PDB ID: 5AC8 28 ), and dup13-15/D10G/G102A/Q24L/V15[b]M (PDB ID: 5G1Y 28 ). The first and last of these variants are only active towards the isomerization of PRA, whereas the middle two variants are bifunctional towards both PRA and ProFAR (Table S2). In all cases, when performing conventional MD simulations starting from the closed conformation of loop 1, this loop is very stable, and remains closed over our simulation timescales (Table S1). We From analysis of our MD simulations (Table 1), we clearly see how all variants carrying the D10G substitution are able to populate all three of closed, open and wide-open conformations. However, the relative populations of these states depends strongly on enzyme variant: already, the D10G substitution by itself appears to be sufficient to cause a conformational shift towards a closed conformation, and the SeHisA(dup13-15/D10G/Q102A/Q24L/V15[b]M) variant, which shows the highest TrpF activity of all variants studied in ref. 28 (Table S2), also shows the most significant population shift towards sampling a closed conformation of loop 1. This is consistent with structural analysis, 28 which indicated that the Q24L substitution is important because it introduces a new stabilizing interaction with V15b, as well as an even better interaction with V15[b]M, such that the Q24L interaction is just as important as the VVR duplication for the adaptive benefit of the V15[b]M substitution to be realized. Finally, the wide-open conformation is also only rarely sampled in the SeHisA(dup13-15) variant, which does not carry the D10G substitution. We hypothesize that in the case of these variants, this is due to the presence of a Gly-Gly dyad in the hinge of loop 1, which provides the loop with enough flexibility to explore these wide-open conformations. Clearly, dup (13-15), as well as the inclusion of additional substitutions, is significantly impacting the conformational space sampled by this loop, shifting towards a loopclosed conformation of loop 1 that is favorable for TrpF activity. ## Empirical Valence Bond Simulations of the Enzyme-Catalyzed Ribose Ring Opening Step in the Isomerization of Substrates ProFAR and PRA To further probe the role of loop 1 in HisA and PriA, we have complemented our conventional and enhanced sampling molecular dynamics simulations with empirical valence bond (EVB) 69 simulations of the initial ribose ring opening step in the Amadori reaction of substrates ProFAR and PRA (Figure 1), as catalyzed by wild-type and variant forms of the two enzymes (see the Methodology section for details of the simulation and parameterization procedures). We note that the calculated activation free energies for this step are not trivial to directly compare with the experimental turnover numbers: no kinetic data exists on the rates of the individual chemical steps. Furthermore, the observed kcat values for these systems are extremely low -on the order of 1 s -1 (or lower) for both the HisA and TrpF reactions catalyzed by HisA and its variants. 28 However, E. coli TrpF has a kcat value of 30-40 s -1 , with the rate-limiting step being a spontaneous keto-enol tautomerization step that occurs off the enzyme, after the ring-opening step. 86 When also taking into account the potential involvement of loop dynamics in determining the turnover rates, as is the case in other enzymes with catalytically important conformational changes such as protein tyrosine phosphatases, 26,29,32,83 this means that the experimental turnover numbers for the isomerization of ProFAR and PRA by the enzymes of interest do not correspond to a chemical step occurring in the enzyme active site. However, as the Amadori reaction that occurs between the enzyme-catalyzed ring-opening reaction and the non-enzymatic tautomerization is likely to be fast (even the uncatalyzed reaction occurs spontaneously at 25 °C85 ), the ring-opening reaction is likely the slowest enzyme-catalyzed chemical step in the catalytic cycle. This is supported by QM/MM studies of the mechanism of the HisA-catalyzed reaction, 84 although this work does not take into account that chemistry is not rate-limiting here. However, even if the experimentally measured turnover numbers (kcat) do not directly correspond to this step, they do produce a lower limit for the rate of this step (and thus an upper limit for the corresponding activation free energy for the rate-limiting step of the enzymatic reaction). Thus, comparing the calculated activation free energies for the ring-opening in different variants, and in different conformational states of loop 1, can still provide insight into the impact of loop dynamics and key amino acid substitutions on the rate of the slowest enzyme-catalyzed step. Taking these limitations into account, the resulting experimental and calculated activation energies are shown in Tables S10 and S11 and S10 and S11). The experimental activation free energies (∆G ‡ exp) were derived from the kinetic data presented in refs. 25,28 . Structures were selected based on clustering analysis using the hierarchical agglomerative algorithm, as implemented in CPPTRAJ. 77 Note that the annotated catalytic distances are average values over 6000 snapshots extracted for each state from our EVB trajectories (from 30 x individual 200 ps EVB mapping windows per stationary point/system). For a full list of reacting distances across all variants, see Tables S12 and 13. As can be seen from these data, our calculations reproduce the experimental trend in activation free energies derived from the turnover numbers relatively well for both substrates ProFAR (Table S11) and PRA (Table S12), reproducing these values to within 2.0 kcal mol -1 for either substrate. This was unexpected, considering the experimental turnover number does not correspond to a chemical step in the enzyme, as discussed above, and indicates that the observed changes in activity nevertheless do have a chemical component. While both wild-type PriA and HisA are capable of catalyzing isomerization of the substrate ProFAR, only PriA can catalyze the isomerization of the smaller substrate PRA. As already seen 33 from the PriA crystal structures with both products bound, loop 5 can be rearranged either to accommodate one substrate or the other, displaying two slightly different conformations of the loop (Figure S1). 25 These are a "knot-like" pro-ProFAR conformation of loop 5, with W145 pointing "in" towards the substrate ProFAR (PDB ID: 3ZS4 55 ), and a pro-PRA β-hairpin conformation of loop 5, with R143 pointing "in" towards the substrate PRA (PDB ID: 2Y85 25,55 ), extrapolating the substrate positioning from the position of the analogous product PRFAR and product analog rCdRP in the respective conformations. Here, we used PriA in its pro-ProFAR conformation as a reference state to calibrate our EVB simulations of the initial ring-opening of ProFAR. The resulting EVB parameters were then used unchanged in all relevant systems. Based on these parameters, we obtain an activation free energy of 17.5 ± 0.6 kcal mol -1 for the analogous reaction catalyzed by wild-type HisA, which is only 1.1 kcal mol -1 higher than the experimental value (derived from kcat) of 16.4 kcal mol -1 . To further validate our PriA/HisA-ProFAR results, we performed single amino acid substitutions in each enzyme (R19A and D130A in PriA and S202A, D129N and D10G in HisA), and performed EVB simulations on these variants. In most of the systems we obtain activation free energies ~1 kcal mol -1 higher than the corresponding experimental values, as for wild-type HisA. However, for D130A in PriA and S202A in HisA, we underestimate the activation free energies by 1.5 -2.0 kcal mol -1 in comparison with the experimental value, suggesting that the experimental effect is due to either a change in substrate positioning or loop conformation or dynamics, which we are unable to capture in our simulations when simply starting from the wild-type crystal structure and manually truncating these residues. In the case of the substrate PRA, we again used the reaction catalyzed by wild-type MtPriA as our EVB reference state, this time with the loop 5 in its pro-PRA conformation. We note that as 34 shown in Table S2, wild-type SeHisA does not show TrpF activity, whereas variants in which loop 1 is extended through dup13-15 do. 28 In the SeHisA(dup13-15) variant, we obtained an activation free energy of 19.9 ± 1.2 kcal mol -1 , which is within 1.3 kcal mol -1 of the experimental value (derived from kcat) of 18.6 kcal mol -1 (noting again that the rate-limiting step for the enzymecatalyzed reaction occurs off the enzyme, 86 so this value is only a proxy for the barrier for the ringopening). We then extended our EVB calculations to model PRA ring-opening as catalyzed by a set of variants of MtPriA and SeHisA(dup13-15) (Figure 9 and Table S11). In the case of the SeHisA(dup13-15) variants, all the variants yielded results within reasonable agreement (~1 kcal mol -1 ) with experimental values. We note as an aside that we also performed simulations on wildtype HisA for PRA substrate (which is not active toward this substrate) and obtained an activation free energy of 19.5 ± 0.6 kcal mol -1 , very similar to the one obtained for SeHisA(dup13- 15), suggesting that in theory the wild-type enzyme could catalyze this reaction if all loops are in the correct conformation and the substrate is optimally positioned, and that the experimental lack of activity is not due to a high barrier to the chemical reaction catalyzed by this enzyme. Tying in with this, as described in the Supporting Information, our PRA simulations are initated from an idealized position of this substrate in the active site, based on overlay with the position of the larger substrate ProFAR. However, the stability of this ProFAR conformation in the active site is facilitated by interaction with the gripper loop 3, whereas PRA lacks the distal phosphodianion group of ProFAR and is thus not able to make this interaction (Figure 3). This suggests that if PRA could be gripped properly (and thus optimally aligned), turnover could in turn happen. In the case of MtPriA we modeled three single amino acid substitutions (R19A, D130A and R143A) and extended our EVB simulations to model the effect of these substitutions (Table S11, Figure 9), in order to specifically capture the impact of the loss of electrostatic contribution of each truncated side chain on the activation free energy. In the case of the R19A variants, we obtain excellent agreement with the experimental value. However, in the case of the D130A variant, our model significantly under-estimates the activation free energy difference compared to experiment, again suggesting that the experimental effect is rather related to a change in substrate positioning or loop dynamics, that is not captured in our EVB simulations. In the case of the R143A variant, we obtain an activation free energy 1.9 kcal mol -1 lower than the wild-type enzyme. We note that, experimentally, this substitution has been shown to significantly impair the isomerization activity of MtPriA towards PRA (kcat/KM reduced from 1.7 x 10 5 M -1 s -1 to 6.0 x 10 3 M -1 s -1 on introduction of this substitution 25 ). However, it is unclear if this is an effect on kcat, KM or both, and it is plausible that the loss of activity is due to structural effects that prohibit productive substrate binding, that are not captured in our simulations. 38 This latter issue would be similar to our observations from a recent study of an analogous system activated by a ligand-gated conformational change, glycerol-3-phosphate dehydrogenase (GPDH). 38 In this system a substantial loss of activity upon truncation of a key catalytic arginine to alanine could only be explained by structural rearrangements (predominantly blocking of the closure of a key catalytic loop), that were observed upon crystallization of this variant. In contrast, this loss of activity could not be captured simply by performing a truncation of this side chain on the wild-type enzyme and considering only electrostatic effects which only accounted for a smaller part of the observed change in activity. S11 and S12). Due to the lack of relevant crystal structures, we used structures with loop 1 in an open conformation extracted from our conventional 36 MD simulations as starting points for the EVB simulations. As can be seen from these data, for the catalyzed isomerization of PRA, when the reaction was modelled with loop 1 in an open conformation, we obtain much higher activation free energies for ring-opening than when modeling the reaction from a loop-closed conformation, due to a combination of the loss of key interactions between loop 1 and the substrate, and also extra solvent-exposure of the active site when this loop opens up. However, we observe no impact on the activation free energy when modelling the catalyzed isomerization of ProFAR, staring with loop 1 in open conformation. Therefore, while the catalytic importance of loops 5 and 6 is well-established, 25,28,54 our EVB calculations show a clear role also for correct closure of loop 1 for PRA isomerization, with full closure of the loop into a catalytically competent conformation being essential for efficient isomerization of PRA. ## Overview and Conclusions In the present work, we use a combination of conventional and enhanced sampling molecular dynamics simulations, as well as empirical valence bond calculations, to explore the role of loop dynamics in dictating the selectivity and evolvability of the evolutionarily important model enzymes, HisA and TrpF, 28, which selectively catalyze the isomerization of substrates ProFAR and PRA, respectively (Figure 1), as well as the bifunctional isomerase PriA, which catalyzes both reactions in bacteria such as M. tuberculosis. 87 The roles of loop dynamics and ligand-gated conformational changes in TIM-barrel proteins and proteins from other folds have been a topic of substantial research interest (e.g. refs. 18, , among many others). However, what makes the current enzymes stand out from these prior studies is the importance of not one but two (TrpF) or even three (HisA and PriA) long, mobile loops (Figure 1), the specific conformations of which have been suggested to play an important role in facilitating the selectivity of PriA and evolved HisA variants. 25,28,54 Thus, these enzymes undergo more complex loop dynamics than the aforementioned systems. In addition, prior enzymes that have been characterized as being activated by ligand-gated conformational changes, such as TPI and OMPDC, are extremely proficient enzymes. In contrast, all enzymes studied here are relatively inefficient (Table S2), 25,28,80 , with turnover numbers of ~10 s -1 or less, 28,52,101 despite catalyzing a reaction that is intrinsically very fast. 85 Related to this, while loop motions in highly proficient TIM-barrel enzymes such as TPI are relatively fast (on the μs timescale 96 ), motions of up to the ms timescale have been detected in the evolved HisA variants, 28 and thus loop motions are likely to be (at least partially) rate-limiting in these enzymes. At the simplest level, our simulations show, in agreement with structural data, 25 that the enzymes TrpF, PriA and HisA have increasingly large (in terms of active site volume) and "breathable" active sites (Table S8), allowing for the accommodation of substrate ProFAR by HisA and PriA unlike PRA-specific TrpF, the active site of which is clearly too small to accommodate the larger substrate. 25 More importantly, HisA and PriA both possess "gripper residues" interacting with the non-reactive phosphodianion group of ProFAR (R83 and S103 in HisA, R85 and T105 in PriA, Figure 2) that are very similar to analogous interactions in other enzymes activated by ligand-gated conformational changes, such as TPI. 34,35 Of note, however, is that the HisA/PriA "gripper" residues are contributed from the less mobile loops 3 and 4, unlike the primary gripper loop, loop 6 in TPI, which undergoes a substantial conformational change upon ligand binding. 22 Other TIM-barrel proteins such as OMPDC possess analogous gripper loops to TPI loop 6, 22 showing evidence for convergent evolution on these different (βα)8-barrel scaffolds. Our simulations show that while the gripper interaction is stable in HisA throughout the simulations, in PriA, there is electrostatic repulsion between R85 and an additional active site arginine, R143, which causes instabilities in the catalytic loops (in particular loop 5, Figure S3), as well as in the substrate positioning in the active site, such that the larger ProFAR is bound less stably in the PriA active site than in the HisA active site. This is in effect a ligand-gated effect, where interaction with the non-reactive phosphodianion (which is not present in the smaller substrate, PRA) facilitates the stability of catalytically important loop 5. Thus, the underlying principles driving loop stability are similar to those of other enzymes that are activated by ligandgated changes. Following from this, PCA analysis on our simulations shows that the active site loops in these enzymes are not rigid, but can sample a range of wide-open conformations with transitions between them, with their conformational flexibility being stabilized by ligand binding (although less so in the bifunctional PriA than the ProFAR-specific SeHisA). In contrast, in TrpF, which binds a smaller substrate and lacks the gripper, the active site loops remain dynamic, in particular loop 6 (Figure 1), which samples a range of open conformations even with substrate PRA bound to the active site. Related to this, all HisA variants from the real-time evolution experiment 46 studied here also sample a range of open and wide-open conformations. In this context, however, the single D10G substitution on loop 1 appears to be sufficient by itself to increase the population of the closed conformation sampled during our simulations (Table 1), and the highest proportion of closed conformation is observed in simulations of the SeHisA(dup13-15/D10G/Q102A/Q24L/V15[b]M) variant, which has the highest TrpF activity 28 (Table S2). Furthermore, pulling simulations where we pull products PRFAR and CdRP out of the SeHisA(dup13-15/D10G/G102A/Q24L) active site (the only variant from the real-time evolution experiment 28 with a PRFAR-bound crystallographic structure) show significant conformational changes in both loops 1 and 6 when pulling PRFAR out of the active site, whereas when pulling CdRP out of the active site, loop 1 is much more stable and the main requirement is for loop 6 to open. This suggests that loop 1 dynamics are more important for binding of ProFAR and subsequent release of PRFAR, than for the smaller substrate PRA and its product CdRP, whereas loop 1 dynamics appears to be critical to catalysis (Table S11). In addition, the substantial rearrangements that we observe for both compounds suggests that a slow (potentially rate-limiting) product release step is the reason for the low turnover numbers observed for an otherwise facile reaction, further providing evidence that turnover rates are being regulated by loop dynamics. In addition, in contrast to HisA, which undergoes conformational changes of loop 1 during the real-time evolution experiment that changes its selectivity from ProFAR-specific to PRAspecific, 28,46 the bifunctional enzyme PriA is already able to rearrange its active site in its wildtype form, to accommodate the different substrates through alternation between Pro-ProFAR and Pro-PRA conformations of loop 5. 25 These conformational changes both reduce repulsion between the two active site arginine side chains that in turn destabilize loop 5 dynamics (Figure S3), as well as disrupting the stacking interaction between the W145 side chain and the substrate ProFAR (Figure S4). This rationalizes the preference of this enzyme towards PRA rather than ProFAR, despite its similarities with the HisA active site, and we note also that loss of the stacking interaction between W145 and the substrate ProFAR was also presented as one aspect of the gain of PRA-isomerization activity in the SeHisA(dup13-15) variant from the real time evolution experiment. 28,46 Finally, we performed EVB simulations of the first ring-opening step of ProFAR and (where relevant) PRA isomerization (Figure 1) by wild-type HisA, PriA and variants. As described above, the actual rate of the chemical step in these enzymes is unknown, since the rate-limiting step is likely to occur off the enzyme. 86 However, of the steps that occur in the enzyme active site, this is the step that is likely to be the slowest, and therefore it is of interest how the substitutions affect the rate of the ring-opening reaction. Here, we see that our calculated activation free energies for the ring-opening reaction trend well with the differences in activation free energies derived using the measured turnover numbers as an upper limit for this value, suggesting there is both a chemical and a dynamical component to the observed changes in activity upon substitution and/or duplication of key residues. Furthermore, in order to reproduce the relevant PriA activation free energies, it was necessary to start from different structures of loop 5, following earlier structural analysis that demonstrate the loop can exist in either a knot-like pro-PRA or beta-hairpin pro-ProFAR conformation (Figure S1), depending on what product is bound to the active site. 25 Clearly, the ease with which this rearrangement can occur will also impact the selectivity of this enzyme. Also, EVB simulations of wild-type SeHisA, MtPriA and the SeHisA(dup13-15/D10G/G102A/Q24L/V15[b]M) variant with loop 1 in an open conformation all yield substantially higher energies for PRA isomerization, whereas ProFAR isomerization (by wild-type HisA) seems to be unaffected (Tables S10 and 11). This further emphasizes the importance of the correct closure of loop 1 for isomerization of the smaller substrate PRA. This is in contrast to substrate binding, where the conformational plasticity of loop 1 appears to be far more important for facilitating correct binding of ProFAR than PRA (Figure S8). Taken together, these observations highlight the critical role of multiple decorating loops in HisA, PriA and TrpF in facilitating catalysis. These enzymes stand out from prior systems that have been demonstrated to be activated by ligand-gated conformational changes due to a number of factors. First, we have shown the inter-dependent motion of three long loops (or two in TrpF), none of which dominates and each of which is capable of undergoing substantial conformational changes to facilitate the turnover of different substrates. Second, unlike prior systems which show substantial rate accelerations compared to the uncatalyzed reactions with comparatively rapid loop motions, in these enzymes, the catalyzed reaction is already intrinsically fast 85 whereas loop-motion is slow and appears to be controlling the reaction rate. It has been argued that a PriA-like gene product could have been the common evolutionary ancestor for both HisA and TrpF. 53,87,102 Ancestral sequence reconstruction has also been used to suggest that ancient HisA precursors were likely bifunctional, and that this bifunctionality persisted over at least a two-billion-year time span. 52 However, as shown in Figure 5, HisA and PriA exploit loops 1, 5 and 6 to facilitate activity, whereas TrpF lacks an analog for loop 1, and isomerizes PRA harnessing just two catalytic loops, 3 and 6. Our results suggest that an evolutionary trajectory from a PriA-like ancestor to an extant TrpF would be surprisingly complex. Loop 1 would need to be truncated (and not extended, as when SeHisA was artificially evolved into a TrpF 28 ) and the inter-dependency of this third loop would need to be lost, raising the question of what evolutionary path would take a PriA-like precursor to TrpF, while completely abolishing this loop. Despite the novel aspects of the systems studied here, a key similarity with prior systems is the generalizability of ligand-gated conformational changes across a wide range of systems, in particular TIM-barrel proteins, 34 which tend to possess flexible loops decorating their active sites. The conservation of such ligand-gated conformational changes-albeit triggered in different loops-suggests that these decorating loops evolve independently of the barrel providing a starting point for the emergence and divergence of new enzyme activities. 31,47 In addition, TrpF, for example, has been shown to be highly tolerant of variations in loop 6 sequence, such that grafting sequences from related enzymes such as TrpA, HisA and PriA onto the TrpF scaffold did not abolish activity. 103 This is significant considering the high evolvability of this scaffold, 18 and the wide range of chemistry it supports, 47 which makes it very desirable as a starting point for protein engineering efforts. In addition, it could be argued that the real-time evolution experiment that bestowed PRA-isomerization activity to HisA 46 effectively performed "natural" loop-engineering by altering the conformations of key active site loops. 28 Our work suggests therefore that, more broadly, loop grafting and engineering is a powerful tool for generating novel enzymes with tailored activities and specificities, even in complex systems with multiple highly mobile and interdependent catalytic loops. ## Supporting Information Full details of computational methodology, additional structural and simulation analysis. Representative EVB input files, starting structures, parameter files, and Q6 topology files have been uploaded to Zenodo, DOI: 10.5281/zenodo.5893598.

chemsum

{"title": "Complex Loop Dynamics Underpin Activity, Specificity and Evolvability in the (\u03b2\u03b1)8 Barrel Enzymes of Histidine and Tryptophan Biosynthesis", "journal": "ChemRxiv"}

superconcentrated_electrolytes_widens_insertion_electrochemistry_to_soluble_layered_halides

## Abstract: Insertion compounds provide the fundamental basis of today's commercialized Li-ion batteries. Throughout history, intense research has focus on the design of stellar electrodes mainly relying on layered oxides or sulfides, and leaving aside the corresponding halides because of solubility issues. This is no longer true. In this work, we show for the first time the feasibility to reversibly intercalate electrochemically Li + into VX3 compounds (X = Cl, Br, I) via the use of superconcentrated electrolytes, (5 M LiFSI in dimethyl carbonate), hence opening access to a novel family of LixVX3 phases. Moreover, through an electrolyte engineering approach we unambiguously prove that the positive attribute of superconcentrated electrolytes against solubility of inorganic compounds is rooted in a thermodynamic rather than a kinetic effect. The mechanism and corresponding impact of our findings enrich the fundamental understanding of superconcentrated electrolytes and constitute a crucial step in the design of novel insertion compounds with tunable properties for a wide range of applications beyond Li-ion batteries. ## Introduction Redox chemistry provides the fundamental basis for numerous energy-related electrochemical devices, among which Li-ion batteries (LIB) have become the premier energy storage technology for portable electronics and vehicle electrification. Throughout its history, LIB research has witnessed a frenetic race for designing new intercalation compounds, so that most of the crystallographic families with open framework have been investigated and though chances to discover new promising phases are slim. Nevertheless, some compounds sharing similar structures with those of archetypal LIBs cathode materials have never been envisioned as potential host materials. For instance, while transition metals oxides and sulfides have been widely studied, members of the vanadium tri-halide family VX3 (X = Cl, Br or I) were never investigated as battery intercalation compounds despite sharing similar structure with the iconic Li-ion cathode layered materials TiS2 and LiCoO2. Yet, recent studies highlight the interesting physical properties of theses vanadium halide phases. 1,2 For instance their magnetic and electronic structures can be finely tuned by playing with the interlayer coupling between the Van der Walls gap. 3 Such structural modifications are easily achieved through cation intercalation, 4 reinforcing our motivation for testing these materials as Li intercalation hosts. However, the absence of reports on layered halides intercalation compounds is partially explained by their high solubility in polar solvents, as highlighted by their use as vanadium precursors in redox-flow batteries. 5 The delicate balance between materials and liquid electrolytes was already one of the main hurdles that delayed the commercialization of secondary lithium-ion batteries relying on insertion reactions. For instance, the physical and chemical integrity of carbonaceous electrodes were found altered upon lithium intercalation in polypropylene carbonate (PC) based electrolytes initially employed. 6 Such difficulty was later on addressed by tuning the lithium solvation shell with the use of ethylene carbonate (EC), suppressing the co-intercalation of PC and forming a stable passivating layer on the carbon anode, 7 or more recently by using superconcentrated electrolytes. In turn, history teaches us that an electrolyte engineering approach can be used to unlock lithium intercalation in phases hitherto believed to be too unstable for the application. Besides graphite intercalation, superconcentrated electrolytes have also shown promising functionalities to solve many fundamental drawbacks relative to the use of carbonate based electrolytes, such as incompatibility with Li-metal, instability at high voltages, flammability or dissolution of transition metals from the cathode. Hence, the use of superconcentrated electrolytes was identified as an attractive direction to explore the reversible intercalation of Li + into halide-based compounds such as the members of the VX3 family. Herein we demonstrate the feasibility, via the use of superconcentrated electrolytes, to unlock the reversible electrochemical intercalation of lithium in vanadium halides previously considered as transition metal salts too soluble to be used in Li batteries. As a proof of concept, we studied the electrochemical behavior of three vanadium-based halides: VCl3, VBr3 and VI3. Combining electrochemical measurements with structural (synchrotron X-ray and neutron diffraction) and solubility measurements, we successfully demonstrate that nearly one Li + per formula unit can be reversibly intercalated into these materials, hence opening the door to new intercalation chemistries going beyond our current knowledge regarding oxides, sulfides or polyanionic compounds. ## Results Intercalation of lithium into vanadium halides enabled thanks to superconcentrated electrolytes While VCl3 is commercially available, VI3 and VBr3 were grown in evacuated quartz sealed tubes by reacting elemental vanadium, iodine and bromine in nearly stoichiometric conditions (see methods). Their structure was found to match those reported in the literature, where edge-shared VX6 octahedra form honeycomb layers stacked along the c direction with a AB sequence (O1 type structure 19 ) (Figure 1. a, Supplementary Figure 1, Supplementary Tables 1 and 2). These materials were tested by assembling battery half-cells to investigate their ability to electrochemical insert lithium cations (Li + ), using state-of-the-art LIB electrolyte For these three compounds (Figure 1, b), a relatively small discharge capacity is obtained and the process is clearly irreversible as no capacity is observed during the subsequent charge. Moreover, large amount of vanadium traces are observed after discharge on the negative electrode (Supplementary Figure 2), pointing to the dissolution of the cathode material and an irreversible solution process as the origin for the limited discharge capacity. Similar poor performances were encountered using ionic liquid electrolytes (Supplementary Figure 3). Hence, there is a need to develop a new strategy to overcome these dissolution issues. The first attempt consists in replacing the liquid electrolyte by a solid-state electrolyte in which no dissolution can occur. Effectively, when composite cathodes are prepared and tested in an all-solid-state battery configuration (see methods), the electrochemistry is greatly improved compared to that in LP30, thus hinting towards the electrochemical intercalation of Li + into VX3 compounds (Figure 1. c). However, in such configuration the complex interplay between ionic conductivity of the solid electrolyte, chemical compatibility between components and electrode microstructure requires a separate optimization of electrode formulation and testing conditions for each compound. This is indeed exemplified in the chemical incompatibility found between VBr3 or VI3 and argyrodite solid-state electrolyte, evidenced as a clear color change upon mixing. This observation, combined with the difficulty of assembling ASSB and characterizing active materials in such configuration, renders critical the search for an adequate liquid electrolyte in which halides would be stable. To select the correct electrolyte, reconsidering the physico-chemical properties of classical aprotic Li-ion batteries electrolytes proves to be insightful. Commercial LP30 electrolyte is made from a mixture of carbonate solvents (EC and DMC) in which a salt (LiPF6) is dissolved. While EC is a polar component essential to ensure both ion-pairs dissociation and the formation of a stable solid electrolyte interphase (SEI) at the negative electrode, its high dielectric constant is presumably responsible for the VX3 dissolution. Hence, not only the chemical composition of the electrolyte must be modified to prevent vanadium cations and/or halides solubility but also its solvation properties. Inspired by the recent observation of the non-miscibility of lithium halides salts with lithium imide salts, 20,21 we then investigated the electrochemical behavior of vanadium halides in superconcentrated electrolytes. As shown in Figure 1 d, when LP30 is substituted with a 5 M lithium bis(fluorosulfonyl)imide (LiFSI) in DMC electrolyte (denoted LiFSI 5 M hereafter), the electrochemical behavior of the three different phases is drastically even at C-rates as low as C/40, indicating that superconcentrated electrolytes can tackle the solubility issue observed in LP30. Indeed, albeit some initial irreversible capacity being observed in the first cycle, these results suggest that almost 1 Li + per formula unit can be reversibly inserted into these hosts. Furthermore, no vanadium was found at the negative electrode at the end of discharge (Supplementary Figure 2). Overall, switching from regular diluted carbonate-based electrolytes to a superconcentrated electrolyte appears as a key to unlock the electrochemical intercalation of Li + into halide layered compounds. To grasp further insights into the behavior upon cycling of these halides, their structural evolution was monitored by operando synchrotron X-ray diffraction (SXRD). For LixVI3 (Figure 2 a), two successive biphasic processes can be distinguished upon discharge, consistent with the two plateaus previously observed (Figure 1 d). For 0 < x < 0.6, the pristine phase disappears at the expense of an intermediate phase which is then replaced by a fully lithiated phase. On charge, the process is found reversible with the discharged phase being converted back to the pristine one; similar reversible bi-phasic intercalation processes are observed for both VCl3 and VBr3 (Supplementary Figures 4 and 5). Having proved the full structural reversibility of the insertion process and the absence of vanadium dissolution, we believe that part of the observed irreversible capacity is mostly nested in minute amounts of amorphous impurities of chemically absorbed I or Br in our starting materials that were made by gas phase reactions. The structure of every intermediate and fully lithiated phases were then determined by Rietveld refinement. This analysis reveals that the O1type layered structure (𝑅3 space group) of the pristine is preserved for every intermediate phase, with the sole evolution of the lattice parameters (Supplementary Figure 6 and Supplementary Tables 3 and 4). Unlike for the intermediate phases, the structure of the fully discharged phases is dependent on the nature of the anion. Hence, while for VI3 the fully discharged phase possesses the O1 structure, for VCl3 and VBr3 a distortion to an O3 layered structure (𝑅3 𝑚 space group) is observed (Supplementary Figures 7 and 8, Supplementary Tables 5,6 and 7), in agreement with a previous report on a VCl3 lithiated phase prepared by a solid-state route. 22 Such a subtle Li-driven structural difference depending upon the nature of the halide may simply be rooted in their size, following the ionic radii Cl < Br < I. Finally, and to no surprise for layered compounds, it was confirmed by neutron powder diffraction experiments that lithium cations sit in the interlayer of the phases (Supplementary Figure 7). To gain a deeper understanding of the kinetics of the redox process taking place during intercalation, galvanostatic intermittent titration technique (GITT) was performed. Interestingly, the quasi-equilibrium path is almost identical for the three halides. Two discharge plateaus are observed, only differing in their potentials which correlate nicely with the ligand electronegativity (the more electronegative the halide, the higher the potential, Figure 3 a,b). This result suggests that the electrochemical trace is solely related to the redox potential of the V(III)/V(II) couple. To validate this charge compensation mechanism, operando X-ray absorption spectroscopy was performed on VCl3 (Figure 3 c). During the discharge, a shift of the V K-edge position to lower energy (Figure 3 c) reveals a shift from V(III) (pristine) to V(II) (Figure 3 d). The overall evolution during the discharge can be fully described using three principal components: the pristine phase, the end of discharge phase and an intermediate phase (Supplementary Figures 9, 10 and 11), confirming the existence of an intermediate phase as previously observed during operando synchrotron XRD. ## Rationale for the decreased solubility of halides in superconcentrated electrolytes In summary, we directly proved that superconcentrated electrolytes can be used to explore new intercalation compounds and synthesize novel phases for chemistries previously disregarded as highly soluble in liquid electrolytes. To rationalize this effect, the solubility of VCl3 was measured at different concentrations of LiFSI as supporting salt in DMC by inductively coupled plasma mass spectrometry (ICP-MS). The results, shown in Figure 4 a, reveal an initial increase of the vanadium solubility as a function of the LiFSI concentration followed by a drastic decrease when LiFSI concentration is greater than 1 M, reaching values as low as few mM for the LiFSI 5 M superconcentrated electrolyte. Furthermore, vanadium chloride powder was mixed with 5 M LiFSI in DMC solutions for 3 days at 55°C and 85°C, before allowing the solutions to rest at room temperature and monitoring the amount of dissolved vanadium. Since the measured concentrations were extremely close to the ones obtained at room temperature (cV(RT) = 6.3 mM, cV(55°C) = 7.6 mM and cV(85°C) = 5.5 mM), any scattering of the results due to a kinetic hindering of the powder dissolution can be discarded to explain the aforementioned trend. Moreover, similar results were obtained for VI3 and VBr3 (Supplementary Figure 12). Thus, this peculiar bell shape reflects a thermodynamicallydriven phenomenon. To derive the equilibrium law that governs the solubility of vanadium at different supporting salt concentrations, we first need to understand the nature of the dissolution reaction by identifying the chemical environment of the vanadium cations dissolved in solution. Visually, every 1 M LiFSI solution saturated with VX3 exhibits a pronounced coloration, as evidenced by the presence of absorption peaks in their UV-vis spectra (Figure 4 b). Interestingly, the wavelengths of the absorption peaks are halidedependent, suggesting the formation of a vanadium-halide complex in solution. Moreover, the initial increase of the solubility with LiFSI concentration (at concentrations below 1 M) advocates for the participation of the salt anion in the formation of this complex. Such observations can be rationalized by the formation of adducts in solution in the form of [VX3-nFSIp] n-p , through a chelation or a ligand exchange mechanism, as proposed below: where n and p are integers with 0 ⩽ n ⩽ 3 and p > 0 (n = 0 for the chelation mechanism). Moreover, the solubility of LiCl is very low in pure DMC (1.3 mM). 23 Hence, since the VCl3 solubility in 1 M LiFSI electrolyte is around 100 times greater than this value, Clions must be generated from the VCl3 dissolution and should undergo an almost complete re-precipitation: This second step, which can hardly be observed due to the very limited amount of LiCl precipitated (see discussion in the SI), can drastically shift the first equilibrium towards the formation of the product, as expressed by the equation below: At the thermodynamic equilibrium, the Guldberg and Waage law of mass action gives: Thus, considering the activity of a solute as the product of its concentration c and its activity coefficient , divided by the standard concentration (c° = 1 mol L -1 ), the solubility s of the VX3 salt can be expressed by the following equation: Since the solubility of LiBr (sLiBr) and LiI (sLiI) are expected to be greater than the one of LiCl (e.g. sLiBr = 4 mM in DMC at 25°C, 3 times that of LiCl) 23 , one would expect K°2(Cl) > K°2(Br) > K°2(I), which is consistent with the trend observed for the solubilities of the VX3 compounds that follows VCl3 > VBr3 > VI3 (Figure 4 c). For a chelation mechanism, K°2 = 1 and the solubilities of the different vanadium halides are expected to vary less than for the ligand-exchange mechanism. Hence, our experimental observations point towards dissolution through a ligand-exchange mechanism (see SI for a deeper discussion). Neglecting a variation of the activity coefficients, hypothesis which usually holds for low salt concentrations, 24 the VX3 solubility should thus increase with the LiFSI concentration, in agreement with our experimental observations made for concentrations < 1 M (Figure 4 a). However, such explanation does not hold anymore at high ionic strength, regime in which the activity coefficients are known to largely deviate from unity. 24 First, as expressed in equation ( 5), a large decrease of 𝛾 and 𝛾 could potentially explain the lowering of the VX3 solubility. Nevertheless, as suggested in a recent theoretical work, the activity coefficients of single ions are more likely to be greater at high salt concentrations. 24 Hence, the decreased solubility is rather explained by a large increase of 𝛾 ( ) which, owing to the interaction of the vanadium adduct with surrounding solvent molecules, would require developing a refined model to accurately account for the variation of its activity coefficient with the ionic strength. We should further emphasize that independently on the dissolution mechanism (ligand-exchange or chelation), the VX3 solubilities are expected to vary in a similar manner depending on the LiFSI concentration (Figure 4 d). Our study provides the rationalization for the observed bell shape behavior: the initial increase of the solubility originates from the increased concentration of FSI anions, while the decrease observed in the concentrated regime arises from a large deviation from unity of the transition-metal complex activity coefficient value (Figure 4e). More importantly application-wise, the solubility of the fully discharged LiVCl3 phase shows similar bell shape (Supplementary Figure 13), albeit the solubility is found lower than for VCl3 which ensures a good resistance to dissolution for the material even upon reduction. However, we note that upon long cycling a severe capacity decay is observed and can be attributed to vanadium dissolution accumulated over time (Supplementary Figure 14). Obviously, future work ranges in better tuning the delicate electrolyte-solvent balance to improve the capacity retention both at RT and 55°C. In conclusion, these findings confirm that the fundamental framework developed in this work captures every stage of lithiation and can be transposed to the study of other intercalation compounds. Overall, we established the use of superconcentrated electrolyte as a platform to discover new families of materials for Li + intercalation. Driven by the broad interest for their physical properties, we (reversibly) intercalated Li + into layered vanadium halides to form new layered phases, the impact of which goes beyond the sole development of intercalation electrodes for secondary batteries. Indeed, preliminary nuclear magnetic resonance (NMR) results (Supplementary Figure 15) suggest a fast exchange of Li + cations between different crystallographic sites, which does not come as a total surprise owing to the ongoing interests for halides as solid state ionic conductors. Moreover, these layered halides have been widely investigated for their promising magnetic properties, and preliminary magnetic susceptibility measurements show that Li + intercalation can be used to tune their magnetic structure. Aside from unlocking the synthesis of new layered structures, our study also lays the fundamental background to comprehend a so far ill-understood effect related to the use of superconcentrated electrolytes, i.e. the decreased solubility of active material. Indeed, we demonstrate that the low solubility of transition metal compounds in superconcentrated electrolytes originates from a shift of the solubility equilibrium, i.e. from a thermodynamic effect rather than a kinetics one, which also applies to current collectors whose stability in presence of corrosive anions (e.g. FSI, TFSI) was observed to increase in superconcentrated electrolytes. In general, our study highlights the critical need for solid-state chemists to include knowledge about the physical chemistry of liquids when developing novel intercalation compounds. Evidently, the future success of such explorative and collaborative work will rely on answering fundamental questions regarding equilibrium and ion activities in this novel class of electrolytes. ## Synthesis VBr3 and VI3 were grown in evacuated quartz sealed tubes by reacting elementary vanadium (vanadium powder, -325 mesh 99.5% -Alfa Aesar), iodine (99+%, Alfa Aesar) and purified bromine (Sigma-Aldrich) with a slight excess of halide. Because bromine is a liquid, the tube was place in a liquid nitrogen bath to freeze bromine before being evacuated and quickly flame-sealed. The tubes were placed in a tubular oven, the extremity containing the reactants being heated at 400°C for VBr3 and 450°C for VI3 while the other extremity was placed at almost room temperature, as reported elsewhere. 1,2 After a 72h synthesis, large crystals were collected at the cold extremity of the tube, placed in a Schlenk tube and further heated at 200°C for few hours under vacuum (10 -2 mbar) to eliminate the excess bromine or iodine traces. VCl3 (97%, Sigma-Aldrich) was used as received. Lithiated phases were obtained by adding the VX3 phases in a 3 times excess of nbutyllithium solution (2.5 M in hexane, Sigma Aldrich) and stirring the suspension for at least 1 hour. The solution was further centrifuged and the collected powder was washed 3 times with hexane before drying under vacuum in the glovebox antechamber. The Li3PS4 solid electrolyte was obtained in our laboratory via a THF-mediated route proposed by Liu et al. 25 and with a room temperature ionic conductivity of RT= 0.21 mS/cm. On the other hand, commercial Li6PS5Cl electrolyte was used (NEI), having an ionic conductivity of RT= 3.8 mS/cm. ## Liquid electrolytes All the electrochemical experiments were carried out in an Ar-filled glovebox. All the materials tested were mixed with conductive carbon super-P in an active-material/carbon ratio of 7/3. For all the experiments using a liquid electrolyte (LP30 1 M LiPF6 dissolved in 1:1 v/v ethylene carbonate/dimethyl carbonate, Solvionic or LiFSI 5 M in dimethyl carbonate, Solvionic) the as prepared composite were tested in a coin-cell configuration versus a Li metal negative electrode separated by one Whatman glass fiber separator, soaked with ~ 100 L of electrolyte. The mass of composite loaded in the coin cells was comprised between 3 and 5 mg. The cells were cycled on VMP or MPG potentiostats from BioLogic at room temperature, except for the GITT for which the cells we placed in a 25°C oven. ## Solid-state batteries Both anode and cathode composites were prepared by hand grinding the components in the proportions mentioned below with mortar and pestle within an Ar-filled glovebox. The electrochemical testing of the VX3 compounds (X = Cl, Br, I) in all-solid state configuration was conducted in a three-pieced homemade cell consisting in two stainless steel pistons which are inserted into a PEI body (Supplementary Figure 16). The cell is closed by means of six axial screws which also provide the pressure required for correct operation. Additionally, a ferrule-cone pair is also integrated in each piston making the setup airtight. For the battery assembling 35 mg of Li6PS5Cl electrolyte were firstly loaded into the cell and cold pressed at 200 kg/cm 2 for 1 minute. Next, 10 to 12 mg/cm 2 of catholyte (VX3:Li3PS4:VGCF, 65:30:5 wt.%) were evenly spread onto one side of the compressed solid electrolyte and a pressure of 1000 kg/cm 2 was applied during 1 minute. Lastly, 35 mg of anode composite (Li0.8In: Li6PS5Cl, 60:40 wt.%) were added onto the opposite side and the whole stack was densified at 4000 kg/cm 2 for 15 Min. The fully assembled cell was later closed with a torque key applying 2.3 Nm torque in each screw, which yields an internal pressure of ~1000kg/cm 2 . Galvanostatic cycling of the cells was carried out at room temperature, in the voltage range at C/10 using the VMP3 electrochemical workstation by Bio-Logic Science Instruments SAS. ## Materials characterization Ex Situ synchrotron X-ray diffraction (SXRD) patterns were collected on the BL04-MSPD beamline of the ALBA synchrotron (Barcelona area, Spain) at wavelength λ = 0.41378 using the Position Sensitive Detector MYTHEN. Powder samples were filled in 0.6 mm diameter borosilicate capillaries inside an Ar-atmosphere glovebox and subsequently flame-sealed. Operando measurements were carried out in transmission mode using dedicated coin cells 26 assembled under argon filled glovebox and mounted on an ALBA designed 4 samples changer. Constant wavelength (λ = 1.622 ) neutron powder diffraction (NPD) data were collected for LiVX3 (X= Cl, Br, I) at room temperature using the ECHIDNA high angular resolution powder diffractometer installed at the OPAL research reactor (Lucas Heights, Australia). 27 To prevent reaction of the samples with ambient atmosphere, the samples were loaded into 9 mm diameter cylindrical vanadium cans in Ar-filled glove box and sealed with In seals. All diffraction patterns were refined using the FullProf program. The scanning electron microscopy images were measured on a FEI Magellan scanning electron microscope equipped with an Energy dispersive X-ray spectroscopy (EDX) Oxford Instrument detector. EDX measurements were carried out using an acceleration voltage of 10 kV. The Li metal anode samples were collected from cycled cells, washed with dimethyl carbonate and sealed in an air tight container. The transfer from the container to the microscope vacuum chamber were realized rapidly (~ 10 s) to minimize air exposure. ## X-Ray Absorption Spectroscopy Synchrotron X-ray absorption spectroscopy was performed at the vanadium K-edge at the ROCK beamline of the SOLEIL synchrotron facility (Saint-Aubin, France). 28 The Si(111) quick-XAS monochromator with an oscillation frequency of 2 Hz was employed to select the incident photons energy. The spectra were collected in transmission using three gas ionization chambers in series as detectors. A vanadium metal foil was placed between the second and the third ionization chambers to ensure the energy calibration. An average of 900 scans per spectrum (corresponding to 15 minutes of acquisition time for one merged XAS spectrum) was recorded to ensure the reproducibility and to increase the signal-to-noise ratio. The VCl2 and VCl3 reference samples were prepared by mixing uniformly the active material with carbon, then pressed into pellets of 10 mm in diameter. For the operando measurements, a self-standing electrode was prepared by mixing VCl3 active material with carbon black and polytetrafluoroethylene (PTFE) in the ratio of 20:70:10 (by wt.%). The electrode was placed in the in situ electrochemical cell 29 and cycled against metallic Li using a Celgard membrane as separator and ~100 L of 5 M LiFSI in dimethyl carbonate (DMC) as electrolyte. Then, the in situ cell was discharged from the OCV (3.025 V) to 2.5 V vs Li + /Li at a C-rate of C/20 (corresponding to 1 mole of Li + inserted in 20 hours), and a XAS spectrum was recorded every 15 minutes. The details for the principal component analysis are given in the SI. ## Solubility measurements To ensure that the concentrations of vanadium measured in the electrolytes match the thermodynamic limit of solubility for VX3 materials, a large excess of powder (~100-500 mg, depending on the material) was added in 1 mL of organic electrolyte and stirred for 3 days at room temperature in a Ar-filled glovebox. The as prepared solutions were centrifuged (6 000 rpm, 1 hour), and the supernatant was collected and further filtered on 0.2 m pore size PTFE syringe filter in an Ar-filled glovebox. Outside of the glovebox, the as prepared solutions were diluted 1 000 or 10 000 times in a 2w% HNO3 (prepared from HNO3 99.999% metal basis Alfa Aesar and Mili-Q ultrapure water) to reach final vanadium concentrations below 1 ppm. The vanadium concentration in the as prepared aqueous solutions were measured by ICP-MS (Nexion 2000 Perkin Elmer) using a calibration curve obtained by diluting a vanadium standard solution for ICP-MS (TraceCERT, 1 mg/L V in nitric acid, Sigma Aldrich).

chemsum

{"title": "Superconcentrated electrolytes widens insertion electrochemistry to soluble layered halides", "journal": "ChemRxiv"}

aspergiloid_i,_an_unprecedented_spirolactone_norditerpenoid_from_the_plant-derived_endophytic_fungus

## Abstract: An unusual C 18 norditerpenoid, aspergiloid I (1), was isolated from the culture broth of Aspergillus sp. YXf3, an endophytic fungus derived from Ginkgo biloba. Its structure was unambiguously established by analysis of HRMS-ESI and spectroscopic data, and the absolute configuration was determined by low-temperature (100 K) single crystal X-ray diffraction with Cu Kα radiation. This compound is structurally characterized by a new carbon skeleton with an unprecedented 6/5/6 tricyclic ring system bearing an α,βunsaturated spirolactone moiety in ring B, and represents a new subclass of norditerpenoid, the skeleton of which is named aspergilane. The hypothetical biosynthetic pathway for 1 was also proposed. The cytotoxic, antimicrobial, anti-oxidant and enzyme inhibitory activities of 1 were evaluated. ## Introduction Plant-derived fungi, which have drawn considerable attention from natural product chemists, have been proved to be a rich source of bioactive natural compounds . Recently, a wide variety of biologically active and structurally unique metabolites were isolated from these types of microorganisms , demonstrating their promise as a source of novel and/or bioac-tive natural products. Our previous chemical investigation of the bioactive secondary metabolites produced by the endophytic Aspergillus sp. YXf3 associated with Ginkgo biloba led to the isolation of new p-terphenyls and novel types of diterpenoids including pimarane-type diterpenoids (sphaeropsidins A and B, aspergiloids D and E), a cleistanthane-type diterpenoid (aspergiloid C), and norcleistanthane-type diterpenoids (asergiloids A, B, and F-H), many of which were reported from this microorganism for the first time . Interestingly, sphaeropsidins A and B were also discovered from both Aspergillus chevalieri and phytopathogenic fungus Sphaeropsis sapinea, displaying anti-gram-positive bacterial, antiviral, antiprotozoal and phytotoxic activity . We further focused on the fractionation containing the minor terpenoid constituents with characteristic signals for terminal vinyl group detected by 1 H NMR from the liquid fermentation broth of Aspergillus sp. YXf3 and isolated a novel norditerpenoid, namely, aspergiloid I (1) (Figure 1). Herein, we report the production, isolation, structure characterization, and biological activity of 1, a rare spirolactone metabolite with a novel carbon skeleton. ). These data show that 1 has two double bonds and one carbonyl which require three degrees of unsaturation, thus, 1 must also contain three rings. ## NMR data ( The gross structure of 1 was initially deduced by comprehensive analysis of its 1D and 2D NMR data. The 13 C NMR and HSQC spectra of 1 allowed all protons to be assigned to their respective carbons. The 1 H, 1 H three-bond couplings from H-1 to H-3 observed in the COSY experiment established a spin system from C-1 to C-3 (Figure 2). The COSY correlation between H-8 and H-9 revealed C-8 to C-9 connectivity. A terminal vinyl moiety H-14/H 2 -15 was also confirmed by 1 The structure of 1 was further confirmed by a low-temperature (100 K) single-crystal X-ray diffraction experiment, which is shown in Figure 3. As compound 1 has a relatively high percentage of oxygen, it shows enough anomalous dispersion of Cu Kα radiation and allows to determinate the absolute stereochemistry with the Hooft parameter 0.17 (15) for 992 Bijvoet pairs by single-crystal X-ray diffraction experiment . Therefore, the absolute configurations of the chiral centers in ## Experimental General experimental procedures The melting point was measured on a Beijing Taike X-5 stage apparatus and reported without correction. The optical rotation was recorded using a Rudolph Autopol III polarimeter. The UV spectrum was obtained on a Hitachi U-3000 spectrophotometer. The CD spectrum was measured on a JASCO J-810 spectrometer, and the IR spectrum (KBr) was obtained on a Nexus 870 FTIR spectrometer. NMR data were acquired using a Bruker AVANCE III-500 NMR spectrometer at 500 MHz for 1 H NMR and 125 MHz for 13 C NMR. The chemical shifts were given in δ (ppm) and referenced to the solvent signal (DMSO-d 6 , δ H 2.50, δ C 39.5; CDCl 3 , δ H 7.26, δ C 77.1) as the internal standard, and coupling constants (J) are reported in Hz. The high resolution mass measurement was conducted on an Agilent 6210 TOF LC-MS spectrometer. Silica gel (200-300 mesh; Qingdao Marine Chemical Factory, Qingdao, China) and Sephadex LH-20 gel (Pharmacia Biotech, Sweden) were used for column chromatography (CC). Semipreparative HPLC was conducted on a Waters ODS (250 × 4.6 mm, 5 μm) on a Hitachi HPLC system consisting of a L-7110 pump (Hitachi) and a L-7400 UV-vis detector (Hitachi). All other chemicals used in this study were of analytical grade. ## Fungal material, cultivation, extraction and isolation The fungal strain Aspergillus sp. YXf3 was isolated by one of the authors (Z.K.G.) from a healthy leaf of Ginkgo biloba collected in the campus of Nanjing University (Nanjing, P. R. China), in October 2008 . The strain was cultured on MEA (consisting of 20 g/L malt extract, 20 g/L sucrose, 1 g/L peptone, 20 g/L agar and deionized water) at 28 °C for 5 days. ## X-ray crystallographic analysis of 1 Colorless crystal of 1 was obtained by crystallizing from a solution of 2 mL methanol with two drops of distilled water. The single crystal X-ray diffraction data were collected at 100 K with Cu Kα radiation (λ = 1.54178 ) on a Bruker APEX DUO CCD diffractometer, equipped with an Oxford Cryostream 700+ cooler. Structures were solved using the program SHELXS-97 , and refined anisotropically by full-matrix least-squares on F 2 using SHELXL-97. The absolute configurations were determined by computation of the Hooft parameter , in all cases yielding a probability of 1.000 that the reported configuration is correct. Crystal data: C

chemsum

{"title": "Aspergiloid I, an unprecedented spirolactone norditerpenoid from the plant-derived endophytic fungus <i>Aspergillus</i> sp. YXf3", "journal": "Beilstein"}

a_rechargeable_molecular_solar_thermal_system_below_0_°c

## Abstract: An optimal temperature is crucial for a broad range of applications, from chemical transformations, electronics, and human comfort, to energy production and our whole planet. Photochemical molecular thermal energy storage systems coupled with phase change behavior (MOST-PCMs) offer unique opportunities to capture energy and regulate temperature. Here, we demonstrate how a series of visiblelight-responsive azopyrazoles couple MOST and PCMs to provide energy capture and release below 0 C. The system is charged by blue light at À1 C, and discharges energy in the form of heat under green light irradiation. High energy density (0.25 MJ kg À1 ) is realized through co-harvesting visible-light energy and thermal energy from the environment through phase transitions. Coatings on glass with photo-controlled transparency are prepared as a demonstration of thermal regulation. The temperature difference between the coatings and the ice cold surroundings is up to 22.7 C during the discharging process. This study illustrates molecular design principles that pave the way for MOST-PCMs that can store natural sunlight energy and ambient heat over a wide temperature range. ## Introduction Thermal management is crucial in our modern society, regardless of whether we are considering chemical transformations, electronics, human comfort, energy production or our whole planet. Thermal management materials based on specifc heat capacity or phase change are seeing increased use in applications such as electronics, and domestic and industrial heat management. Phase change materials (PCMs) are a broad class of materials whose latent heat during a phase transition from solidto-liquid can be used for energy storage applications. Latent heat storage offers a signifcant advantage if the application involves temperature cycles close to the melting point since in those cases, the corresponding storage density of sensible thermal storage is small. In building applications, phase change materials made from paraffins, salt hydrates, fatty acids or ice can be used as central heat sinks 6 and also in floors, windows or walls. 7,8 Common to all "traditional" thermal energy storage materials is that they operate via heat transfer, both in energy input and energy output. 9 This leads to design challenges and scaling factors that restrict practical performance and implementation. Molecular solar thermal (MOST) systems have been recognized as a promising avenue to harvest and store thermal energy. In the charging process, a stable isomer of a photochromic molecule absorbs photon energy and is converted into a high-energy metastable isomer, thereby storing solar energy in chemical bonds. The MOST system is discharged when the metastable isomer switches back to the stable isomer by external stimuli, with the release of stored energy in the form of heat. While the MOST system shares some properties with PCMs, the process of energy storage and release in the MOST system is controlled by photons and molecular thermodynamics, 16,17 whereas in PCMs it is controlled by heat transfer. Recently, combining the functions of MOST and PCMs into a single component material (MOST-PCM) has been utilized to add storage capacity to the MOST system since the charging of the system is not only happening via solar irradiation but also by taking energy directly from the environment. This dual input leads to an increased energy density by almost 100%. 19 Another attractive feature is added to PCMs; since the solidifcation of the cis liquid is not happening spontaneously, the phase change is locked by the photochemical system. This feature dramatically extends the functionality of the MOST-PCM combination since the phase change is controlled by external stimuli and no insulation is needed to hold the latent heat. However, a severe limitation of MOST-PCMs based on azomolecules studied until now is their inability to be charged and discharged in the solid state in cold environments, especially below 0 C, because of the high melting point (T m ) of cisisomers. This is a critical condition since many applications such as thermo regulated fabrics, 22 or functional coatings will need to be able to function at that temperature. 23 Generally, the trans-cis photoisomerization of azo molecules requires a large free volume 24,25 and can only occur in the surface layers of transcrystals, thus preventing the charging process in the neat solid state. If the ambient temperature exceeds the cis-isomer T m , the generated cis-isomer melts into a liquid and exposes new transcrystal surfaces, and fnally the trans-crystals are entirely transformed into cis-liquids. But most reported T m values of cisisomers are in the range of 20-200 C, 26 which means that their photoisomerization from trans-crystals to cis-liquid cannot occur at low ambient temperatures. On the other hand, although some cis-isomers can maintain liquid states below 0 C due to their supercooling behavior to achieve discharging at low temperature, the charging process is still at room temperature (27 C), limiting the versatility of the system. 19,20 Another challenge for the MOST-PCM is that the charging process generally needs UV light irradiation, since UV light causes damage to materials and the human body, and comprises a small fraction (4.5%) of the total solar spectrum, 27 resulting in the low utilization efficiency of solar energy. To date, only one study has reported the utilization of ortho-functionalized azobenzene derivatives to store both visible light energy and room temperature ambient heat. However, this system could not be charged below 0 C, and the energy density was in the 0.07-0.15 MJ kg 1 range. 28 ortho-Substitution can increase the energy of trans-isomers 29 or decrease the energy of cis-isomers, 30 so that the DH iso of this type of azo molecule decreases to only 6-25 kJ mol 1 (0.01-0.05 MJ kg 1 ). 28 Therefore, a reversibly charging/discharging and visible-light-energy storage MOST-PCM working at lowtemperature remains to be explored. Here, we report new arylazopyrazoles as MOST-PCMs, which are rechargeable below 0 C by visible light, as illustrated in Fig. 1. In addition, by co-harvesting the visible light energy and low-temperature ambient heat, an energy density of 0.25 MJ kg 1 is achieved, which is an increase of 67% over previous comparable systems. 28 Furthermore, the cis-isomer has a halflife of 22 days at 0 C, demonstrating its stable energy storage capacity. The combination of high energy density, storage time, and the fact that the system can be charged at low temperatures provides the opportunity to explore the function of the material in a new type of optically regulated MOST-PCM window as a proof-of-concept study, designed to illustrate the function of the MOST-PCM system in a coating. Glass coated with arylazopyrazoles is prepared as a miniature energy storage window. One point to highlight is that the novel windows can be charged and discharged at 1 C by 400 nm blue light and 532 nm green light, respectively. During the discharging process, the surface temperature of the window can reach from 1 C up to 21.7 C (a temperature increase of 22.7 C), corresponding to a thermal power output of 256.2 W m 2 during a continuous period of 60 s. Charging and discharging energy at low temperatures has potential implications for functional clothing, 31 advanced sunglasses, deicing 23 and home heating 32 under ice-cold conditions, thereby increasing thermal comfort and reducing the energy consumption of conventional heating. ## Synthesis and UV-Vis absorption spectra of bidirectional visible-light-driven photoswitches The fundamental principle to realizing low-temperature working and visible-light energy storing MOST-PCM systems is to design photochromic molecules that drive phase transition with visible light at low temperature. The introduction of a 4thiomethyl group and changing the bridging positions between the azo group and pyrazole ring was expected to achieve bidirectional visible light switching through extending the pconjugation. Accordingly, three 4-methylthioarylazopyrazoles (S3, S4, and S5) were designed and their synthetic routes are shown in Scheme 1. In previous studies, 33 arylazopyrazoles were usually prepared by the Mills reaction of nitrosobenzene analogs and aminopyrazoles, but in this method it was difficult to synthesize arylazopyrazoles with electron-donating groups because the electron-donating groups would destabilize nitrosobenzene analogs. 34 Hence, we frst prepared 3(5)-nitroso-1H-pyrazole, which was then coupled with aniline analogs to give S3(5)-H. S3 and S5 were subsequently produced in one pot by N-methylation at two selectable positions, beneftting from the proton transfer and tautomerism of 1H-pyrazole. S4 was formed via diazo-coupling and cyclization reactions. Their photoisomerization behaviors were studied using UV-Vis absorption spectra in an acetonitrile solution. As shown in Fig. 2b-d, all trans-isomers exhibited single and intense absorption bands in the 350-400 nm region (3 max ¼ 25-32 10 3 M 1 cm 1 ) due to p-p* transition. Compared to the reported 4methoxyarylazopyrazole (O4, 342 nm), 35 the p-p* l max of S3 and S4 were red-shifted to about 360 nm, and further to 385 nm for S5 (Table S1 †). The trans/cis relative absorption of S3, S4 and S5 at 400 nm was strong (Table S1 †), and hence it was possible to realize trans-cis isomerization using visible light. The photostationary states (PSSs) at different wavelengths (from 365 to 532 nm) were studied (Fig. S1 †), and the isomeric compositions are presented in Table S2. † As a result, 400 nm blue light induced a near-quantitative yield (>95%) of trans-cis isomerization for S5, while only $85% for S3 and S4. The higher transcis photoconversion of S5 was attributed to its p-p* l max closer to 400 nm and stronger trans/cis relative absorption at l ¼ 400 nm. Exciting the tail of n-p* bands of three cis-isomers using green light (532 nm) resulted in cis-trans isomerization. The high overlap between the n-p* band of cis-S4 and the longwavelength absorption band of trans-S4 led to a relatively low cis-trans conversion (85%). In contrast to cis-S4, cis-S3 and cis-S5 exhibited n-p* transitions red-shifted to 25 nm and 36 nm (Table S1 †), respectively, which led to partial separation of the n-p* bands of the cis and trans isomers, thereby inducing high (91% S3) to near-quantitative (>95% S5) isomerization. ## Theoretical analysis of bidirectional visible-light-driven photoswitches The geometries of O4, S3, S4, and S5 and their electronic transition characteristics were calculated using density functional theory (DFT) modelling 36 to evaluate the relationship between molecular structures and photophysical properties (see Section 3 in the ESI †). All trans-isomers exhibited a planar structure with a C-N-N-C dihedral angle of 180.0 (Fig. 2e and S5 †), which resulted in symmetry-forbidden n-p* transitions (S 0 / S 1 ) with negligible oscillator strength (f ¼ 0.00, Table S4 †). Compared to trans-O4, the 4-methylthioarylazopyrazole series showed more effective extension of the p-conjugated system due to the increase of p-p conjugation by introducing the 4-SMe group, as indicated by their frontier molecular orbitals (Fig. S7-S10 †). Consequently, the energy gap between the HOMO (p orbital) and LUMO (p* orbital) was smaller for the 4methylthioarylazopyrazole series (3.714, 3.776, and 3.542 eV for S3, S4, and S5, respectively, as depicted in Fig. 2f), leading to a bathochromic shift of their p-p* transition (S 0 / S 2 ). The increased red-shift of S5 was caused by a "complete" conjugation pathway between 5-pyrazole and the azo group that further expands the p-conjugation of the system. 40,41 cis-S4 showed a nearly T-shaped conformation with a C-C-N-N dihedral angle of 82.8 (Fig. 2e and S5 †), resulting in a weak n-p* transition with f of only 0.0048. In contrast, cis-S3 and cis-S5 were found to disfavor the T-shaped conformation owing to the presence of the ortho nitrogen atom (Fig. 2e and S5 †). Therefore, their n-p* absorbance bands were remarkably enhanced (f ¼ 0.1068 and 0.1096 for cis-S3 and cis-S5, respectively, Table S4 †), and nearly quantitative conversions from cis to trans isomers were achieved. ## Design of low-temperature charging/discharging MOST-PCMs Thus, all the above data indicate that S5 can be used as a bidirectional visible-light-driven photoswitch, which provided nearquantitative trans-cis and cis-trans photoconversions in acetonitrile solution. However, pristine S5 was not able to store both visible-light energy and phase transition latent heat at low temperature since the photoisomerization from trans-crystals to cis-liquid was inhibited even at room temperature. To charge and discharge energy below 0 C, two additional principles are considered: (i) the T m of the cis-isomer should be below 0 C, thus forming amorphous cis-liquids to store photon energy and ambient heat; (ii) the trans-isomer should have a T m and crystallization point (T cry ) much higher than 0 C, thus forming trans-crystals to release energy. In order to adjust the T m of the trans and cis isomers, we varied the length of linear alkyl chains with or without a vinylic end group on the thioalkyl group of S5, denoted as An-S5 and Bn-S5 (Fig. 3a and Scheme S1 †). Studying the phase behaviour of the systems using differential scanning calorimetry (DSC), we fnd that the trans-isomers with longer and intermediate alkyl chain lengths (n ¼ 6-12 for An-S5, and n ¼ 7, 9, and 11 for Bn-S5) showed T m (40-60 C) and T cry above 0 C (10-45 C, Tables S5 and S6 †). The cis-isomers with intermediate alkyl chain lengths had the lowest T m (5 C for A6-S5 and 1 C for B7-S5), as shown in Fig. 3b, c and S13, S14. † Hence, B7-S5 could undergo a reversible visible-light-triggered trans-crystal 4 cis-liquid transition in cold environments thanks to its low cis-isomer T m (below 0 C) and high transisomer T cry (34 C). This property illustrates, to the best of our knowledge, the frst example of a functional low-temperature visible-light controlled energy storage MOST-PCM system. ## Charging/discharging below 0 C The charging and discharging processes of B7-S5 at 1 C are discussed as follows. As shown in Fig. 3d and Video S1, † after 400 nm light (40 mW cm 2 ) irradiation, trans-B7-S5 in the orange crystal state lost birefringence and melted into red liquids, which then switched back to a crystal state by 532 nm light (110 mW cm 2 ) irradiation. UV-Vis spectroscopy was used to record the photoisomerization yields during irradiation of a neat sample of B7-S5 with a mass of 2 mg and a thickness of about 50 mm. During the charging process (Fig. 3e), the trans-cis isomerization of B7-S5 in a neat state proceeded easily and produced a high yield of photoisomerization (93%) at 40 min, slightly lower than that in dilute solutions (>95%). During the discharging process (Fig. 3f), the trans-isomer content of the sample increased exponentially, and a near-quantitative (95%) cis-trans isomerization was achieved within 2 min of irradiation. X-ray diffraction (XRD) analyses were carried out to further understand and verify the reversible trans-crystal 4 cis-liquid behavior of B7-S5 (Fig. 3g). The sharp peaks at 2q of 5-35 before irradiation corresponded to the regular stack of azo molecules in trans-crystals. After exposure to 400 nm light (40 mW cm 2 ), the peaks disappeared, indicating that cis-B7-S5 had an amorphous structure. Subsequently, XRD patterns were recovered when the sample was irradiated with 532 nm light (110 mW cm 2 ). The X-ray crystal structure of trans-B7-S5 reveals an antiparallel packing, and several weak contacts such as alkylalkyl, alkyl-phenyl, and alkyl-pyrazolyl dominating the intermolecular interactions (Fig. 3e). Presumably, the absence of strong intermolecular interactions in trans-B7-S5 offered enough flexibility to the system, which was benefcial for the charging process. 42,43 ## Energy storage time and energy density The cis-isomers of azo molecules can thermally relax into transisomers in the dark spontaneously, which determines the energy storage stability of cis-B7-S5. The change of absorbance value at p-p* l max (385 nm) as a function of time was measured in acetonitrile solution between 25 and 40 C (Fig. 4a and S15 †). The thermal isomerization rate constants, k cis/trans (Table S8 †), were calculated based on the frst-order reaction kinetics at 25 C, 30 C, 35 C, and 40 C, respectively. Based on the Arrhenius equation, cis-B7-S5 was found to have a half-life t 1/2 of 22.4 days at 0 C, indicating its stable thermal energy storage capacity at low temperatures. Thermal cis / trans kinetics was also studied in neat states (Fig. S16 and Table S9 †), and cis-B7-S5 liquid still had a t 1/2 as long as 6.3 days at 0 C. It is ubiquitous to observe a lower t 1/2 of photoswitches in condensed states than in solution. Cis-isomers tend to adopt fully relaxed geometries in solution, whereas in condensed states different geometries may be preferred due to intermolecular interactions, characterized by lower isomerization barriers. 44 Additionally, B7-S5 demonstrated excellent photon-harvesting ability with a quantum yield F trans/cis of 0.39 AE 0.01 for photoisomerization in acetonitrile solution (Fig. 4b and S17 †), similar to other reported azopyrazoles compounds. 19,41,45 To evaluate the energy density of the MOST-PCM, the isomerization enthalpy DH iso of the thermally induced cis-liquid to trans-liquid reversion reaction and crystallization enthalpy DH cry of the trans-liquid to trans-crystal transition were measured by DSC. As shown in Fig. 4c, the cis-B7-S5 liquid revealed a broad exothermic peak over 60-120 C during the thermally activated cis-trans isomerization, and the integrated area under the peak represented a DH iso of 0.14 MJ kg 1 (44 kJ mol 1 ). This result was consistent with our calculations (49 kJ mol 1 , Table S7 †) based on DFT and similar to the pristine azobenzene (<50 kJ mol 1 ). Furthermore, the DSC cooling curve displayed a sharp exothermic peak at around 33 C with a DH cry of 0.11 MJ kg 1 (35 kJ mol 1 ), which was due to the trans-liquid to trans-crystal transition. Therefore, the total thermal energy density of the MOST-PCM was 0.25 MJ kg 1 (79 kJ mol 1 ). According to F trans/cis and DH iso , the solar efficiency h was estimated to be up to 1.3% (see Section 6 in the ESI †), which was one of the highest values reported for azo-based MOST systems (0.2-1.3%). 19,28 Rechargeable energy storage coatings To illustrate the practical applications of B7-S5, trans-B7-S5 and diethoxydimethylsilane modifed chain-like silica were co-dissolved in ethanol and isopropanol (1 : 1, v/v) solution, and then the mixture was drop-cast onto a glass substrate to form rechargeable coatings, as shown in Fig. 5a. The modifed chainlike silica formed a transparent porous network structure 46 on glass substrates to prevent the leakage of cis-liquid (Fig. S18 †). Various patterns were created on the same rechargeable glass substrate through selectively writing/erasing processes (Fig. 5b). First, a mask was placed on the glass and irradiated with 400 nm blue light. The exposed area of the glass was transformed from the initial opacity to semi-transparency, and the color changed from orange to red. As a result, the target pattern was written on the glass substrate. Subsequently, the erasing process could happen when the substrate without a mask was exposed to 400 nm blue light irradiation, resulting in a globally semi-transparent glass. And then, this glass substrate could be further patterned via selectively irradiating with 532 nm green light. Finally, the glass recovered to its opaque state after exposure to 532 nm green light without a mask. The transmittance spectra of the rechargeable glass under different light irradiations were recorded (Fig. 5c). The glass sheet in a discharged state had a low transmittance of less than 10% in the wavelength range of 500-800 nm, and after exposure to 400 nm blue light, the glass sheet is charged with the transmittance up to $70% in the wavelength range of 650-800 nm. In addition, the rechargeable glass showed good durability under alternating blue light and green light irradiations (Fig. 5d). Thanks to the high transmittance of the charged state glass sheet and the low-temperature phase transition of B7-S5, the rechargeable glass had the potential for use as a photochromic solar thermal energy storage window in daily life, especially in cold winters. As a proof-of-principle study, the rechargeable glass sheets were installed on a miniature house model as windows (size 10 by 12 mm and coating thickness $400 mm). The house model was placed in 5 C surroundings to ensure that the surface temperature of the windows was around 1 C. As shown in Fig. 5e, upon exposure to 400 nm blue light (40 mW cm 2 ), the window stored visible light energy and low-temperature ambient heat while transforming from opacity to semitransparency. Then, by triggering it with 532 nm green light (110 mW cm 2 ), the stored energy was rapidly released on demand as high-temperature heat. A high-resolution infrared thermal imaging camera was used to track the temperature changes of the window when exposed to 400 nm (40 mW cm 2 ) and 532 nm light (110 mW cm 2 ) (Fig. 5f, S19 and Videos S2-S5 †). During the 400 nm light irradiation (charging process), the window exhibited a temperature difference of about 3 C above the ambient temperature, indicating a weak photothermal effect. The charged window reached 21.7 C at 60 s during the 532 nm light irradiation (discharging process), about 22.7 C higher than the cold surroundings. A control experiment of irradiating the discharged window with 532 nm light showed a much lower temperature change (9.3 C), which means that the temperature change between the charged window and environment was mainly due to cis-trans isomerization. Assuming that the cis-trans isomerization was fully completed at approximately 60 s, the corresponding thermal power output was estimated to be 256.2 W m 2 . Such high-temperature heat release also means that the B7-S5 molecules on the surface of the window can act as a photon-driven molecular heat pump, upgrading thermal energy from low to high temperature. Furthermore, the optically controlled heat release makes it possible to reach about an order of magnitude higher temperature gradients than is possible with traditional MOST window coating concepts. 32 These solar thermal energy storage coatings show unprecedented performances, including visible-light trigger/storage, high energy density, and recyclable ice-cold charging/discharging, thus holding great promise for future energy management systems. ## Conclusions In conclusion, we have successfully designed a series of bidirectional visible-light switching azo molecules and applied them as MOST-PCMs for storing and releasing solar energy below 0 C. The molecular design strategies are summarized as follows: (i) the 4-thioalkyl substituent on azo molecules shifts the p-p* absorption bands to long-wavelength, enabling bidirectional visible light photoisomerization; (ii) replacing one phenyl ring on the azo molecules with a pyrazole ring increases the half-life of the metastable cis-isomer; (iii) varying the length of thioalkyl chains changes the intermolecular forces, which could adjust the T m of both trans and cis isomers. Eventually, reversible visiblelight-triggered trans-crystal 4 cis-liquid transitions are achieved below 0 C. Accordingly, the azo molecules can simultaneously store visible-light energy and low-temperature ambient heat to achieve a high energy density (0.25 MJ kg 1 ). Moreover, a rechargeable coating is prepared by drop-coating a solution containing azo photoswitches and modifed chain-like silica on the glass surface. The coating shows potential as energy storage windows due to optical transmittance in the charged state, but it is also clear that more work is needed to increase the optical transmittance of the material. Future studies could focus on red-shifting l max of azo molecules to the near-infrared region to fabricate efficient semitransparent energy storage windows. Other possible application areas are functional coatings and fabrics with controllable heat release functions. We note that the structure-property relations derived from the chemical design provide a blue-print for how to design future MOST-PCM systems with tailored temperature functions and optimised optical properties. We envision that this work can open an avenue for the design of advanced MOST-PCM systems that store natural sunlight and ambient heat over a wide temperature range. ## Conflicts of interest The authors declare no competing interests.

chemsum

{"title": "A rechargeable molecular solar thermal system below 0 \u00b0C", "journal": "Royal Society of Chemistry (RSC)"}

ascorbate_oxidation_by_iron,_copper_and_reactive_oxygen_species:_review,_model_development,_and_deri

## Abstract: Ascorbic acid is among the most abundant antioxidants in the lung, where it likely plays a key role in the mechanism by which particulate air pollution initiates a biological response. Because ascorbic acid is a highly redox active species, it engages in a far more complex web of reactions than a typical organic molecule, reacting with oxidants such as the hydroxyl radical as well as redox-active transition metals such as iron and copper. The literature provides a solid outline for this chemistry, but there are large disagreements about mechanisms, stoichiometries and reaction rates, particularly for the transition metal reactions. Here we synthesize the literature, develop a chemical kinetics model, and use seven sets of laboratory measurements to constrain mechanisms for the iron and copper reactions and derive key rate constants. We find that micromolar concentrations of iron(III) and copper(II) are more important sinks for ascorbic acid (both AH 2 and AH − ) than reactive oxygen species. The iron and copper reactions are catalytic rather than redox reactions, and have unit stoichiometries: Fe(III)/ Cu(II) + AH 2 /AH − + O 2 → Fe(III)/Cu(II) + H 2 O 2 + products. Rate constants are 5.7 × 10 4 and 4.7 × 10 4 M −2 s −1 for Fe(III) + AH 2 /AH − and 7.7 × 10 4 and 2.8 × 10 6 M −2 s −1 for Cu(II) + AH 2 /AH − , respectively. Ascorbic acid is of great interest in food, where it is both an essential vitamin and a natural preservative. Ascorbic acid is also vital for plants. It not only plays a role in photosynthesis, cell growth and signal transduction, but also helps defend from oxidative stress as the most abundant water-soluble antioxidant in plants . Because of its importance for food and in plants, food chemists and botanists have performed the vast majority of studies of ascorbic acid oxidation chemistry . In mammalian systems, ascorbic acid is a common and important molecule with roles in metabolic function, oxidative stress responses and immune system maintenance taking place in epithelial lung lining fluid and other areas in the body 8,9 . In an air pollution context, inhaled particulate matter, a highly complex and variable mixture of inorganic and organic compounds, encounters the lung lining fluid containing substantial concentrations of ascorbic acid. Growing evidence indicates that transition metals in inhaled particles are particularly active components capable of inducing a wide range of negative health effects including myocardial infarction, adverse birth outcomes and respiratory illnesses . A leading hypothesis for how airborne particles induce health effects is via oxidative stress, and redox-active transition metals such as iron and copper have been heavily implicated in the ability of particles to generate reactive oxygen species and therefore potentially contribute to aerosol toxicity 13,14 . For example, soluble iron and copper in synthetic lung fluid correlate with the formation of reactive oxygen species OH ⋅ and H 2 O 2 15, 16 . Because ascorbic acid is a key antioxidant in lung lining fluid, ascorbic acid consumption is one of the assays used by atmospheric chemists to quantify aerosol oxidative potential 17 ; aerosol oxidative potential is proving to be better at predicting adverse health outcomes than particle mass 18 . Ascorbate/ ascorbic acid consumption has been observed to be positively correlated with total iron and copper concentrations in ambient aerosol 14 . Ascorbic acid can have both pro-and anti-oxidant roles, and it reacts with reactive oxygen species (ROS) and transition metals. Ascorbic acid can be readily oxidized by undergoing a one-or two-electron transfer, terminating the free radical-mediated chain reactions in foods and tissue, reducing lipid peroxidation and deterioration of foods 6 . The autoxidation of ascorbic acid by oxygen in the presence of transition metals, especially cupric (Cu(II)) and ferric (Fe(III)) ions accounts for the majority of loss of this ascorbic acid activity in food. Despite its role as an efficient antioxidant, ascorbic acid can also accelerate oxidative deterioration of flavor and color in food through Fenton-type radical reactions 6,19 . This pro-oxidant effect occurs when transition metal ions are present, and the level of available ascorbic acid is relatively low and not sufficient to scavenge the radicals formed by Fenton-type reactions. In both food and physiological conditions, the key loss pathways for ascorbic acid are via ROS and transition metals, especially Fe(III) and Cu(II). However, for the transition metal reactions, the stoichiometries, mechanisms and rate constants are all very uncertain. Further, while the ROS ascorbic acid reactions are reasonably well understood from a mechanistic standpoint, the range of rate constants in the literature for these reactions spans about a factor of 15 . Here we develop a model in the Kinetics Preprocessor (KPP) 26 environment based on available ascorbic acid chemistry with ROS and free iron and copper chemistry from the literature. The model is validated against measurements of the formation of dehydroascorbic acid (DHA), the main oxidation product of ascorbic acid in the presence of micromolar concentrations of Fe(II), Fe(III) and Cu(II) at pH 2.8 and 7.0. The measurements at pH 2.8 were made to develop an online measurement of ascorbic acid consumption by ambient particulate matter 17 , and allow us to probe the chemistry of ascorbic acid, AH 2 . Additional measurements were made at pH 7.0 to probe the reactions of the deprotonated form, AH − . We also use measurements of ascorbate loss and/or OH ⋅ formation from Lin and Yu 27 and Charrier and Anastasio 15 at ~ pH 7 to further constrain the model. We then use the model to constrain the mechanisms and derive rate constants for the catalytic reactions of Fe(III) and Cu(II) with both AH 2 and AH − in the presence of oxygen. ## Ascorbic acid chemistry review Here we use 'ascorbic acid' to mean the sum of the protonated form, AH 2 and the deprotonated form AH − , and the chemical formulas to indicate the individual species. pH dependence. Ascorbic acid reacts with several species of ROS, as well as the oxidized forms of several transition metals (Fig. 1). As ascorbic acid (AH 2 ) can readily lose a proton to form the ascorbate anion (AH − ), (pK a,1 = 4.1; pK a,2 = 11.8) both AH 2 and AH − play roles in chemistry at low and neutral pHs. Typically, AH 2 and Reactions with hydroxyl and hydroperoxyl radicals and superoxide. OH ⋅ reactions with both AH 2 and AH − (Table 1, R4 and R7) appear to proceed at close to diffusion-controlled collision rates. The rate constants for these reactions fall in the ranges (4.5-7.9) × 10 9 M −1 s −1 at pH 1-1.5 and (1-11) × 10 9 M −1 s −1 at pH 7-11 respectively (Supplementary Table S1). We adopt the rate constant of 7.9 × 10 9 M −1 s −1 from Redpath and Willson 21 for the oxidation of AH 2 by OH ⋅ and 1.1 × 10 10 M −1 s −1 from Buettner and Schafer 22 for AH − in our study, because the deprotonated ascorbic acid tends to react more rapidly than the protonated form. Both AH 2 and AH − readily undergo one-electron oxidation by superoxide (O 2 ## .− ), hydroperoxyl radical (HO 2 ## ⋅ ) and hydroxyl radical (OH ⋅ ) to form the ascorbate radical (A .− ) (R4-9, Table 1). The pK a of AH . is sufficiently low that the protonated radical (AH . ) can be ignored. The unpaired electron of A .− residing in the π-system makes A .− relatively unreactive 22 , however A .− can form DHA via disproportionation (R12,13, Table 25 and Cabelli and Bielski 24 . The experimental data disagree by a factor of 1.5-15, although they have the same shape, with a maximum in the observed rate at about pH 4.5. Because both AH 2 and the hydroperoxyl radical have similar pK a s (4.1 and 4.8 respectively), in the pH range 2-7, the contributions of R6 and R8 are difficult to separate, while at low and high pH R5 and R9 dominate, respectively. Because Nadezhdin and Dunford 25 neglected the AH 2 reactions (R5 and R6) in their discussion and their data span a smaller pH range, we use values based on the data and analysis by Cabelli and Bielski 24 . Cabelli and Bielski 24 conclude that it is not possible to deconvolute k 6 and k 8 , but the sum (0.356 k 6 + k 8 ) can be said to be equal to 1.22 × 10 7 M −1 s −1 . Because we use updated pK a s for ascorbic acid and HO 2 ⋅ (Tables 1 and 2) we adjust 24 , d Van der Zee and Van den Broek 36 , e Bielski et al. 37 , f Vislisel et al. 57 , g Dewhirst and Fry 47 , h Parsons et al. 45 , z Estimate numbers. 12d,e 2A 1). The rate constants for reactions of A ⋅− and HO 2 ⋅ and O 2 ⋅− were measured by Cabelli and Bielski 24 using radiolysis; using pH to select HO 2 ⋅ (pH = 1-3) or O 2 ⋅− (pH = 7.8-8), and were determined to be 5.0 × 10 9 M −1 s −1 and 2.6 × 10 8 M −1 s −1 , respectively. ## ROS reactions Autooxidation. Ascorbic acid is an excellent electron-donor antioxidant. The relatively low reduction potential of ascorbate (0.19 V for DHA/AH − at pH 3.5) should allow it to be readily oxidized by molecular oxygen 5 . However, while this redox reaction is thermodynamically favorable, it is spin forbidden; molecular oxygen is a triplet with two unpaired electrons, while ascorbate is in the ground state 5 . The only ascorbate species that is capable of true autoxidation, determined after treating the solutions with Chelex resin to remove trace metals appears to be the ascorbate dianion (A 2− ) + O 2 ## 9 . Because there is little A 2− at pHs below ~ 10 (Table 1 R2), the autoxidation rate for ascorbate (all forms) is slow, ~ 6 × 10 -7 s −1 at pH 7 5 . We verified this as part of our measurements (not shown). ## Transition metal reactions. Catalytic or redox? Ascorbate reactions with iron and copper are central to metal-mediated antioxidant chemistry, and it is clear that ascorbate reacts overwhelmingly with the oxidized forms of the metals (Fe(III) and Cu(II)). The early studies of this reaction uniformly interpreted their data with catalytic mechanisms such as: ) for Cu 2+ , CuOH + and CuSO 4 70 (q) for CuCl + and CuCl 2 78p Other reactions specific to a subset of experiments , CuCl + and CuCl 2 . These forward and back reactions are written separately in the KPP input file. ∆ An upper limit is used for these reactions. i Gonzalez et al. 40 , j Miller et al. 58 , k Pham and Waite 51 , l De Laat and Le 38 , m Herrmann et al. 43 , n Stumm and Morgan 50 , o Powell et al. 42 , p Deguillaume et al. 41 , q Wang et al. 59 , r Lee et al. 60 , s Pham et al. 61 , t Pham et al. 62 , u Goldberg et al. 63 , v Wu et al. 64 , w Skogareva et al. 65 www.nature.com/scientificreports/ Following this, Buettner 32 reported rate constants for the bimolecular reactions of Fe(III) and Fe(III)/EDTA and Cu(II) with AH − at pH 7 in oxygenated solution, and suggested it was a catalytic reaction, although the rate constants they reported did not include an oxygen dependence. Later, Buettner and Jurkiewicz 19 instead described it as a redox reaction: and suggested a somewhat higher rate constant for the Fe(III)/EDTA complex. Subsequent modeling studies adopted the redox reaction 33,34 . Overall, however, the ascorbate mechanistic literature does not support a significant role for the redox reaction. Most or all studies point to the catalytic reaction instead; this includes the original source of the rate constant used for the redox reaction in Buettner 32 , and the mechanistic studies described below. We also test the redox and catalytic mechanisms with our model ("Ascorbate oxidation via the catalytic, redox or OH ⋅ /HO 2 ⋅ /O 2 ⋅− Pathways"). Transition metal ascorbic acid reaction mechanism. Three detailed mechanisms have been proposed to describe the oxidation process of ascorbic acid by iron and copper . All of them begin with the oxidized form (Fe(III) or Cu(II)), consume oxygen, and produce A .− or DHA plus a reduced form of oxygen (HO 2 . , O 2 .− or H 2 O 2 ). In the proposed mechanisms, metal, ascorbic acid and oxygen molecules form a complex, with the metal ion serving as a bridge that transfers one or two electrons from ascorbic acid to oxygen and maintains its valence. In Scheme A, proposed by Khan and Martell 31 , a ternary metal-ascorbate-oxygen complex forms in which one electron is transferred from AH 2 or AH − through metal ion to oxygen (Scheme 1). Scheme A Khan and Martell 31 . Subsequently, Jameson and Blackburn 28 and Jameson and Blackburn 29 suggested a mechanism that involves an initial two-electron transfer to oxygen and the formation of Cu(III) intermediates (Supplementary Scheme B, Supplementary Eqs. ( 7)-( 14)). Shtamm et al. 30 then proposed a two-electron transfer mechanism involving Cu(I)-Cu(II) redox couple (Supplementary Scheme C). Although there is no agreement on the step by step oxidation state of metal ions in the catalytic cycle, there is some evidence showing that the reducibility of metal ion was necessary for it to be an active catalyst. Khan and Martell 31 tested VO 2+ , Mn 2+ , Co 2+ , Ni 2+ and Zn 2+ , and of these only VO 2+ was able to catalyze the oxidation of ascorbic acid. Further it is not clear if A .− is an intermediate of the oxidation of ascorbic acid or if ascorbic acid is directly oxidized to DHA 30 . The disproportionation reaction for A .− is now well established (R12 and 13, Table 1) and is sufficiently rapid to not be rate limiting in this mechanism. This difference can have a moderate effect on the fitted rate constants; for iron of more consequence is the amount of OH . that is produced. The stoichiometry for the metal ion-catalysed oxidation reactions of ascorbic acid by oxygen is also debated. Khan and Martell 31 measured iron and copper-catalyzed oxidation of ascorbic acid by oxygen at pH 2-5.5 and reported the reaction was first-order in ascorbic acid, metal and oxygen. Jameson and Blackburn 28 investigated the copper-catalyzed oxidation of ascorbic acid by oxygen in 0.1 M potassium nitrate at pH 2-3.5 and found a first-order dependence on copper and ascorbate (AH − ) and half-order on oxygen. The same rate law (for AH − ) was investigated by Shtamm et al. 30 at pH 2.7-4. Consistent with the rate law derived by Jameson and Blackburn 28 the rate observed by Shtamm et al. 30 was inversely related to the pH, indicating the rate law only applies to ascorbate. Moreover, Jameson and Blackburn 29 suggest that the stoichiometry can change depending on the nature and concentration of electrolytes, finding evidence that high concentrations of chloride ions (0.1 M) shifted the dependence from 1st order on AH − to half order on total ascorbic acid (AH 2 + AH − ). Transition metal rate constants. Measured and estimated reaction rate constants for AH 2 and AH − with Fe(III) and Cu(II) in the literature are summarized in Table 3. The literature is divided into values for catalytic reactions, including cases for which it is possible to re-calculate a value given for the redox reaction as a catalytic reaction, and values for the redox reactions. The various catalytic reaction stoichiometries as well as pH dependencies and other caveats are also shown in Table 3. ## Reactions of the radical anion. The main product of the AH 2 and AH − reactions is the radical anion, A .− (Fig. 1), which disproportionates to form AH − and DHA or AH 2 and DHA at low pH: 2A •− H + ⇔ AH 2 /AH − + DHA (R12 and R13, Table 1). There is a wide range of values for the equilibrium constant K 12 in the literature; Foerster et al. 35 found pH dependent values of 1.6-7.9 × 10 14 M -1 at pH 4-6.4 (recalculated for the form of the equilibrium constant above) and Buettner and Schafer 22 reported 5 × 10 14 M -1 at pH 7.4. More recently Van der Zee and Van den Broek 36 found a value of 1.7 × 10 16 M -1 at pH 7.4 using ESR to monitor A .− and improved calibration techniques. We adopt the value from Van der Zee and Van den Broek 36 for K 12 and calculate the equilibrium www.nature.com/scientificreports/ constant for R13 using K −1 13 = K −1 12 K AH 2 , where K AH 2 is the first dissociation constant of ascorbic acid. We include both reactions in the model due to the highly pH-dependent A .− decay rates. For the forward reactions 2A •− + H + → AH − + DHA (R12) and 2A •− + 2H + → AH 2 + DHA (R13), we use 7 × 10 4 and 8 × 10 7 M −1 s −1 based on the radiolysis study by Bielski et al. 37 . The rate constants are chosen from the two plateaus in the pH dependent rate constants they derived 37 . ## Other model uncertainties. For the model presented here, some of the chemistry is well established, including much of the ROS chemistry, acid-base equilibria, inorganic iron chemistry, and probe and buffer chemistry. There are several general sources of error and uncertainty for the set of reactions in Tables 1 and 2, in addition to the specific uncertainties described above. These include errors in the rate constants, which range from a few percent to a factor of ten or more. In some cases, reaction stoichiometries and product distributions are also uncertain. Measurement data was often collected in solutions containing other solutes that may impact reaction rates and/or reaction mechanisms, but how and at what concentrations other solutes effect the rates is not known. Temperature differences also introduce uncertainties when data are collected at different temperatures. Further, usually only a small number of the uncertain reactions are important for a given set of experimental conditions. Some of our validation data was collected at 37 °C to mimic physiological conditions. Unfortunately, almost all literature data available for the set of reactions used here were reported for room temperature, and temperature dependencies were not available. Because temperature dependence is reaction specific, and can even have different signs, we have only adjusted the small number of rate coefficients for which there is temperature dependence. Gas solubility is also temperature dependent; we use a dissolved oxygen content corresponding to the temperature of each experiment. ## Methods Model description and extraction of rate coefficients. The model (Tables 1 and 2) includes reactions that describe the chemistry of reactive oxygen species, iron, ascorbate, sulfate and chloride, and a few reactions specific to the detection of DHA or OH ⋅ corresponding to the experimental datasets used to validate the model. The model builds on previous models describing aqueous OH ⋅ production kinetics in the presence of iron and sulfate . Additional reactions describing the copper 41,42 and chlorine 43 chemistry have also been updated. The ascorbic acid mechanism (Fig. 1) is built into the model based on the detailed review of the available literature (described above). The final form of the ascorbic acid-metal reaction is similar to Scheme A; however, we have also explored many other forms and stoichiometries of the reactants (below). The chemical kinetics mechanisms were solved using the Kinetics Pre-Processor (KPP) 2.2.3 26 with the gfortran compiler and the Rosenbrock solver. For measurements using two reaction coils in series with different conditions (below), the model was run separately for each set of conditions, and the output of the first segment was used as an input for the second. To solve for the rate constants of two unknown reactions, such as the two reactions needed for copper (R16 and R17) we employ a two-dimensional binary search algorithm, specifically, we vary the rate constants for the two key reactions on an 11 × 11 grid field. For each grid (a combination of two rate constants), we run the model for each Fe(III) or Cu(II) concentration for which we have a measurement and calculate a mean squared error. The MSE of each unit square is obtained by averaging the MSEs of the nearest four grid points. After one cycle, we arrive at a unit square centered by a minimum MSE and we then divide this square into a new 11 × 11 grid field. The range containing the minimum MSE is narrowed as this process is repeated, and after 4 times the rate constant combination with minimum MSE is determined to be the best fit. For cases where we need to fit more than two rate constants, the grid search method is not efficient enough, thus, we use a coordinate search method instead. We start from randomly chosen initial rate constants for these reactions, along with an initial search range. Each time we vary one rate constant within the search range while keeping the other rate constants fixed, calculate corresponding MSEs, and update the rate constant with the one that produces minimum MSE. This process was applied to each reaction in turn and when this sees no improvement, we reduce the search range in order to continuously decrease the MSE. Finally, when the search range exceeds our required precision, the optimization stops. Validation data. We measured the oxidation of ascorbic acid by Fe(II), (III) and Cu(II) by quantifying the oxidation product dehydroascorbic acid (DHA), as described in detail in Campbell et al. 17 . Briefly, DHA is reacted with o-phenylenediamine (oPDA) to produce a highly fluorescent product 3-(1,2-dihydroxyethyl)fluoro [3,4-b]quinoxaline-1-one (DFQ) with unit yield. DFQ is then quantified via fluorescence spectroscopy. OH ⋅ formation from ascorbate reactions with Fe(II) and Cu(II) at around pH 7 were reported by Charrier and Anastasio 15 (2.8 µM OH ⋅ from 1 µM Fe(II) and 14 µM from 1 µM Cu at 24 h) and Lin and Yu 27 (0.3 µM OH ⋅ from 1 µM Fe(II) at 3.8 h and 8.8 µM from 0.3 µM Cu at 6.3 h); Lin and Yu 27 also reported ascorbate consumption. While the measurements are difficult to compare due to measurement differences and potential non-linear dependencies on both concentration and reaction time, the results appear to be in good agreement for Fe(II) and weak agreement for Cu(II). Reagents and chemical preparation. All chemicals were obtained from Sigma-Aldrich. Ascorbic acid (≥ 99.0%), Chelex 100 sodium form, 0.1 M HCl solution, 0.1 M NaOH solution, CuSO 4 (≥ 99.0%), FeSO 4 (≥ 99%), Fe 2 (SO 4 ) 3 (≥ 98%), o-phenylenediamine (≥ 99.5%), DHA (≥ 96%), HEPES (≥ 99.5%) were used as received. A 200 µM solution of ascorbic acid was prepared in Chelex-resin treated MilliQ water (resistivity ≥ 18.2 MΩ cm −1 ), to ensure as low as possible trace metal concentrations and minimize background DHA formation. The Vol:.( 1234567890 ## Online measurements of Fe(II), Fe(III) and Cu(II). Measurements of ascorbic acid oxidation by iron and copper were conducted in an online instrument described in detail in Campbell et al. 17 . Briefly, a flow of 1.1 mL/min of 200 µM ascorbic acid is added to an equivalent 1.1 mL/min flow of either Cu(II)SO 4 , Fe(II)SO 4 or Fe(III) 2 (SO 4 ) 3 . The reaction mixture was then incubated in reaction coil-1 (Supplementary Table S2) housed in ethylene glycol for 20 min at 37 °C. After passing through the reaction coil, a solution containing 46 mM oPDA in 0.1 M HCl was added at 1.1 mL/min and mixed with the ascorbic acid/metal reaction mixture for 10 min (at pH 2.8) at room temperature in reaction coil-2 (Supplementary Table S2). DHA formed by the oxidation of ascorbic acid/ ascorbate reacted rapidly with oPDA to form the highly fluorescent compound DFQ. The reaction mixture containing DFQ then passes through a fluorescence detection cell (details in Campbell et al. 17 ). The extent of ascorbic acid oxidation is then expressed in terms of µM DHA using a DHA calibration curve 17 . A summary of reaction conditions and dilution ratios are presented in Supplementary Table S2. While water typically contains low levels of H 2 O 2 (not measured here but generally below 10 nM 44 ) in the absence of transition metals, there are no pathways to form radicals either from O 2 or H 2 O 2 . Consistent with this, in the absence of added metals, DHA formation in the reaction coils was below detection limits. ## Results and discussion DHA loss pathways. Observations of the stability of DHA indicate it decreases with increasing pH; in pH 2-4, aqueous DHA solutions are stable for days, while at neutral pH, the half-life of DHA is around 20 min 45,46 . Our model includes three degradation pathways of DHA: reactions with the hydroxyl radical, H 2 O 2 and hydrolysis to produce 2,3-diketogulonic acid. For the reaction with hydroxyl radical, we estimate a rate constant of 1 × 10 10 M −1 s −1 , the diffusion limit. The hydrolysis rate constant is thought to be negligible at low pH but reaches (5.3-5.8) × 10 -4 s −1 at neutral pH 46,47 . DHA also reacts with H 2 O 2 with an estimated rate constant of 4.2 × 10 -2 M −1 s −1 and oxalyl l-threonate, cyclic oxalyl L-threonate and oxalate as the main products 45 . Calculated DHA degradation from these three pathways is negligible at low pH. At neutral pH, DHA degradation reaches 29-39% of total DHA formation in the first reaction coil for copper and 28% for iron, with hydrolysis as the main pathway. Although the reaction of DHA and oPDA is thought to be fast, with a reaction time of about 14 s, in the first (20 min) reaction coil where oPDA is absent, DHA degradation is quite significant. The evidence in favor of the catalytic reaction for iron is as follows. First, the redox pathway fails to produce enough DHA especially at pH 2.8. This is both because the redox reaction converts Fe(III) to Fe(II), and the system can only slowly reoxidize the Fe(II), so Fe(III) can consume only a limited amount of ascorbic acid. Also, the redox reaction seems to produce A ⋅− , which only produces DHA with 50% efficiency (R12, 13); for the catalytic reaction production of A ⋅− vs. direct formation of DHA is less of a settled question. Second, both the catalytic and redox reactions make more cumulative OH ⋅ from the Fe(III)-ascorbic acid reactions than observed by Lin and Yu 27 and Charrier and Anastasio 15 (discussed more in "Comparison with OH • formation data"), but the catalytic reaction is reasonably close to the observations ("Comparison with OH • formation data") while the redox reaction vastly overshoots. The catalytic reaction directly generates the OH ⋅ precursor H 2 O 2 , but it does not produce the Fe(II) needed for the Fenton reaction to convert H 2 O 2 to OH ⋅ as does the redox reaction. ## Ascorbate oxidation via the catalytic, redox or OH The evidence in favor of the catalytic reaction for copper is as follows. In experiments where the extent of the reaction is high, as for Cu(II) at pH 7 (Fig. 3), DHA formation eventually stops increasing. The catalytic reaction is able to reproduce the general asymptotic shape of the DHA formation dependence on Cu(II) (Fig. 3) while the redox mechanism predicts a linear relationship (this is also observed for Fe(II), but the iron phenomenon likely depends more on autoxidation of Fe(II) to Fe(III) rather than the Fe(III) reacting with ascorbate). The reason the catalytic reaction produces an asymptotic behavior is that oxygen is consumed in the closed reaction tubes, limiting reactions 14-17 (Table 1). Additionally, when we include both the redox and catalytic pathways in the model using the coordinate search algorithm for pH 2.8 Fe(III) and pH 7 Cu(II), the optimization steps in the direction where rate constants for the redox reactions of AH 2 + Fe(III) and AH − + Cu(II) continuously decrease; for the other DHA data the model cannot differentiate between the catalytic and redox reactions. OH ⋅ production from the copper via both pathways falls between the two divergent observational results 15,27 . Further support for the catalytic mechanism comes from Jameson and Blackburn 28 , who reported that although some degree of charge transfer occurred within the copper-ascorbate complex, the complete one-electron redox reaction did not happen when there was no oxygen. ROS reactions are only important in the experimental systems if OH ⋅ , HO 2 . and/or O 2 ⋅− concentrations are high, such as for conditions associated with experiments with added H 2 O 2 , or when Fe(II) is the dominant form of iron, for example. Were more DHA formation to be attributed to ROS chemistry, much higher concentrations of OH ⋅ would be needed, a situation clearly not supported by the OH ⋅ measurements by both Charrier and www.nature.com/scientificreports/ Anastasio 15 and Lin and Yu 27 . Our result indicates that for both Fe(III) and Cu(II), the catalytic pathway is always dominant compared to the ROS pathway, by 4-6 and 1-3 orders of magnitude for Fe(III) and Cu(II) respectively. However, in the Fe(II) case, the contribution of the catalytic pathway and ROS pathways are more comparable; the ratio of catalytic to ROS pathways decreases from 11 at 2.5 µM Fe(II) to 1.4 at 200 µM Fe(II). This is because Fe(II) produces ROS via reduction of molecular oxygen and the Fenton reaction, pathways not available to Fe(III). While the oxidized form of the Fe and Cu might be expected to dominate ascorbate consumption in many situations, the contributions of OH ⋅ , HO 2 ⋅ and/or O 2 ⋅− may be significant under some conditions. Given the rate constants (Table 1), OH ⋅ , HO 2 ⋅ /O 2 ⋅− need to be ~ 10 -9 , 10 -4 × [Fe(III)] or 10 -8 , 10 -3 × [Cu(II)], respectively to account for around half of ascorbate loss. OH ⋅ , HO 2 ⋅ in liquid phases in equilibrium with gas phase concentrations of 10 6 and 10 7 molec/cm 3 result in liquid phase concentrations of ~ 10 -3 and 10 -2 nM, respectively 48 . However, 1 and 2. Because of the additional uncertainties and lack of relevance of the high Fe(II) concentrations, only experimental data for Fe(II) concentration within 200 µM at pH 2.8 is shown in the figure . Table 3. Summary of rate constants for reactions of iron and copper with ascorbic acid. *The reaction appears to be catalytic, but these papers assumed a redox reaction; the rate has been recalculated here assuming an oxygen dependence. www.nature.com/scientificreports/ even with modest concentrations of organics, the radicals will be rapidly depleted away from the interface 48 . Consistent with this, model calculations for lung lining fluid estimated concentrations of 10 -10 -10 -7 nM for OH ⋅ , 10 -6 -10 -4 nM for HO 2 ⋅ and 10 -3 -10 -1 nM for O 2 ## ⋅− , with the highest concentrations associated with extremely high PM concentrations 33 . In comparison a typical concentration for iron and copper is around several micromolar in the bronchoalveolar lavage and can be even higher when exposed to highly polluted environments, and thus should usually be the dominant sink for ascorbic acid and ascorbate 49 . Catalytic reaction rate constants. Fe(III). In the Fe(III)/AH 2 system, the model is very sensitive to the catalytic reactions AH 2 (R14) or AH − (R15) or both, depending on pH. We minimize the sum of MSEs for the two iron data sets (Fe(III) at pH 2.8 and 7) to derive rate constants for the catalytic reactions R14 and R15. Bestfit third-order rate constants [Fe(III)] [AH 2 /AH − ][O 2 ] for R14, R15 are 5.7 × 10 4 and 4.7 × 10 4 M −2 s −1 , respectively; these values produce good agreement with the DHA formation data for both Fe(III) and the Fe(II) over the concentration range as shown in Fig. 2. The AH − result is in good agreement (within 35%) with 3.5 × 10 4 M −2 s −1 from Buettner 32 , assuming a first order oxygen dependence in their study (Table 3). Our values for AH 2 and AH − are significantly lower than 4.0 × 10 5 and 2.4 × 10 7 M −2 s −1 reported by Khan and Martell 31 and the AHvalue after recalculation for the catalytic pathway from Lakey et al. 33 of 2.3 × 10 5 M −2 s −1 , Table 3. ## Fe(II) + O 2 . Because the asymptotic behavior observed for very high Fe(II) at pH 2.8 is far outside the relevant range for environmental samples, we include only Fe(II) data up to 200 µM. The model is reasonably successful at reproducing DHA formation from Fe(II) at pH 2.8 at lower Fe(II) concentrations (Fig. 2), although it overpredicts DHA formation at high Fe(II) (> 100 µM). Ascorbate is only oxidized by OH ⋅ , HO 2 ⋅ , O 2 − and Fe(III), so in the presence of Fe(II) ascorbate oxidation should be controlled by the production of Fe(III) or ROS. Both of these pathways are initiated by the reduction of O 2 via R45 and enhanced by H 2 O 2 production from Fe(III) + ascorbic acid (R14, 15). The rate constant for R45, Fe(II) + O 2 is pH sensitive. Stumm and Morgan 50 suggest that when pH > 5 the rate increases with pH, with a second-order dependence on OH − concentration. At low pH (1-4), this rate constant is ~ 10 -5 M −1 s −1 and independent of pH. This rate constant has been found to increase by a factor of 10 for a 15 °C temperature increase 50 . Therefore, for pH 2.8 we use a rate constant of 10 -5 M −1 s −1 at room temperature and 10 -4 M −1 s −1 at 37 °C. For pH 7.0, a k 45 of 0.39 M −1 s −1 was suggested by Pham and Waite 51 which we adjusted upward to 3.9 M −1 s −1 at 37 °C. The Fe(II) measurements (Fig. 2) show that initial AH 2 consumption/DHA production by Fe(II) is around one sixth of that of Fe(III), however, this likely results partly from conversion of Fe(II) to Fe(III) in the stock solution. Although the Fe(II) stock solution was made fresh daily, in the few hours required for the measurements Fe(II) slowly oxidizes to Fe(III). Rate constants for R45 at room temperature are ~ 0.11 M −1 s −1 for pH 6.0 and 0.17 M −1 s −1 for pH 6.5 51 . The pH of our 1 mM FeSO 4 stock solution is ~ 6.3, so about 13%/h of Fe(II) is oxidized to Fe(III). Because the degree of oxidation of the stock solution was not the same for each experiment, this may explain the larger error bars for Fe(II) (Fig. 2). The modelling results for Fe(II) shown here are based on the assumption that 10% of Fe(II) was oxidized to Fe(III) before each experiment. Copper catalytic reaction rate constants. Compared to iron, there are more studies of the copper reactions with ascorbic acid, and more disagreements (Table 3). Here, the rate laws from Jameson and Blackburn 28 , Jameson and Blackburn 29 , Shtamm et al. 30 and Khan and Martell 31 are all tested. For simplicity, we use overall reactions for the catalytic pathway; the products of the catalytic reaction are H 2 O 2 and either DHA or A •− , depending on the charge balance of the equations. Ascorbic acid oxidation by Cu(II) is more efficient than by Fe(III) (Figs. 2 and 3). For the 1:1:1 stoichiometry first proposed by Khan and Martell 31 , the third-order rate constants [Cu(II)][AH 2 /AH − ][O 2 ] for R16 and R17 that provide the best fit of the data are 7.7 × 10 4 and 2.8 × 10 6 M −2 s −1 , respectively (Fig. 3). The Cu(II) ion catalyzes the oxidation of ascorbic acid more efficiently at neutral pH than acidic pH; this is reflected in the much higher value for k 17 than k 16 . In the pH 7.0 measurements, the theoretical maximum DHA concentration in the second reaction coil is 100 μM, but similar to the observation for Fe(II), DHA formation reached a much lower maximum of about 45 μM at 2.5 µM Cu(II). The model matches this behavior well for Cu(II), due to a combination of the depletion of oxygen in the solution which suppresses DHA production at high Cu(II) concentrations, and the hydrolysis of DHA, which is significant at high pH. The fitted rate constants for AH 2 and AH − are lower than 3.8 × 10 5 and 6.0 × 10 7 M −2 s −1 in Khan and Martell 31 . Using a first order oxygen dependence, (redox) rate constants for Cu(II) and AH − in Buettner 32 and Lakey et al. 33 can be converted into (catalytic) third-order rate constants k([Cu(II)][AH − ][O 2 ]) of 3.1 × 10 6 and 3.0 × 10 6 M −2 s −1 respectively, in good agreement with the rate constant we derived. Both the iron and copper data suggest our experimental data are in better agreement with measurement of Buettner 32 , while there might be an overestimation of the rate constants in Khan and Martell 31 . If we instead use the rate law [Cu(II)][AH − ][O 2 ] 1/2 suggested by Jameson and Blackburn 28 and Shtamm et al. 30 with a 2.5 order rate constant we find a best-fit value for AH − of 1.2 × 10 5 M −3/2 s −1 . This value is about 1.5 orders of magnitude larger than 4.3 × 10 3 M −3/2 s −1 from Jameson and Blackburn 28 and 5.2 × 10 3 M −3/2 s −1 from Shtamm et al. 30 . Some or all of the difference could be due to the difference in temperature; as the earlier measurements were made at ~ 25 °C and ours were mostly at 37 °C or from the low concentrations of chloride ions 5 included in our experiments. Nevertheless, this 1:1:0.5 rate law predicts higher DHA formation at high Cu(II) than observed in the pH 7.0 data. This is partly because the reaction's half-order oxygen dependence means O 2 is less depleted and is less able to limit DHA formation. The 1:1:1 stoichiometry of Cu(II), AH 2 /AH − and oxygen fits the shape of DHA formation curve better and arrives at a smaller MSE. However, the room temperature DHA hydrolysis rate used in the model may also underestimate DHA consumption. Jameson and Blackburn 29 suggested a [Cu(II)][AH 2 + AH − ] 1/2 [O 2 ] 1/2 (1:0.5:0.5) rate law for solutions containing 0.1 M chloride ions. Our experiment includes chloride ions as follows: for pH 2.8 experiments the pH in the first reaction coil was adjusted with HCl resulting in 1.6 × 10 -3 M Cl − , and the oPDA mixed into the 2nd reaction coil is prepared in 0.1 M HCl increasing the Cl − concentration in the 2nd coil by 0.03 and 0.05 M for pH 2.8 and 7 experiments respectively, but DHA forms in the first reaction coil, so this is less important. This stoichiometry provides a much worse fit for our data than the other rate laws, implying our concentrations of Cl − are too low to significantly alter the Cu(II)-ascorbic acid reaction. Comparison with OH ⋅ formation data. To further test the reactions we derived with data from the literature, we include benzoate (the OH ⋅ probe) and phosphate buffer reactions (R87-95) and calculate OH ⋅ production for the conditions used by Charrier and Anastasio 15 and Lin and Yu 27 . Although both of these studies only investigated Fe(II), our model includes reactions that oxidize this species to Fe(III), and thus we can use the data to constrain the products of the Fe(III) + AH 2 /AH − reaction. With A ⋅− + HO 2 ⋅ as products, the model produces OH ⋅ that exceeds the observations from both Charrier and Anastasio 15 and Lin and Yu 27 by 2 orders of magnitude, suggesting the products are H 2 O 2 + DHA. The H 2 O 2 + DHA combination of products produces 4 µM OH ⋅ formation from 1 µM Fe(II) after 24 h, in reasonable agreement with the Charrier and Anastasio 15 measurement of 2.8 µM. A similar calculation for the Lin and Yu 27 conditions overshoots the reported OH, resulting in a concentration that is about a factor of three higher, and somewhat overestimates consumption of ascorbate (the modeled concentration at 3.8 h is about 90% of the measured value). The discrepancy may be due to several factors, including the temperature difference, errors in various rate constants or the observational data, or incorrect assumptions about the products of the Fenton reaction; there is some evidence in the literature that the Fenton reaction can produce either OH ⋅ or Fe(IV), with a pH dependent product distribution. Fe(IV) production may be favored around neutral pH . Our model only considers OH ⋅ as a product and could thus overestimate its formation. For copper, the model results fall between the OH ⋅ formation measurements from Charrier and Anastasio 15 and Lin and Yu 27 , which are widely divergent from each other, at about 2.6 times the former and 27% of the latter. Although both the Lin and Yu 27 data and our model agree at nearly 100% ascorbic acid loss at ~ 6 h, ascorbic acid concentration predicted by the model decreases with time nonlinearly, different from the linear trend in the experiment. Several other aspects of Cu(I) and Cu(II)-ROS chemistry are uncertain; some additional discussion of the gaps can be found in the SI. ## Implications and conclusions In recent years, there has been widespread application of acellular assays to measure particle-bound ROS and aerosol oxidative potential (OP), with measurements spanning large spatial, temporal and chemical spaces 11,18,55 . OP is proving to be better at predicting adverse health impacts than particle mass 11,18 . The complex interplay between OP assays, including the ascorbic acid assay, and redox-active PM components has widely been demonstrated 14,18,56 . Detailed understanding of assay responses is crucial to fully elucidate both the role of chemical composition on aerosol OP and to directly probe the role of aerosol OP in particle toxicity. Ultimately, this information should be translatable into policy that specifically targets the components in PM that drive OP (and

chemsum

{"title": "Ascorbate oxidation by iron, copper and reactive oxygen species: review, model development, and derivation of key rate constants", "journal": "Scientific Reports - Nature"}

identification_of_drugs_targeting_multiple_viral_and_human_proteins_using_computational_analysis_for

## Abstract: The SARS-CoV2 is a highly contagious pathogen that causes a respiratory disease named COVID-19. The COVID-19 was declared a pandemic by the WHO on 11th March 2020. It has affected about 5.38 million people globally (identified cases as on 24th May 2020), with an average lethality of ~3%. Unfortunately, there is no standard cure for the disease, although some drugs are under clinical trial. Thus, there is an urgent need of drugs for the treatment of COVID-19. The molecularly targeted therapies have proven their utility in various diseases such as HIV, SARS, and HCV. Therefore, a lot of efforts are being directed towards the identification of molecules that can be helpful in the management of COVID-19.In the current studies, we have used state of the art bioinformatics techniques to screen the FDA approved drugs against thirteen SARS-CoV2 proteins in order to identify drugs for quick repurposing. The strategy was to identify potential drugs that can target multiple viral proteins simultaneously. Our strategy originates from the fact that individual viral proteins play specific role in multiple aspects of viral lifecycle such as attachment, entry, replication, morphogenesis and egress and targeting them simultaneously will have better inhibitory effect.Additionally, we analyzed if the identified molecules can also affect the host proteins whose expression is differentially modulated during SARS-CoV2 infection. The differentially expressed genes (DEGs) were identified using analysis of NCBI-GEO data (GEO-ID: GSE-147507). A pathway and protein-protein interaction network analysis of the identified DEGs led to the identification of network hubs that may play important roles in SARS-CoV2 infection. Therefore, targeting such genes may also be a beneficial strategy to curb disease manifestation. We have identified 29 molecules that can bind to various SARS-CoV2 and human host proteins. We hope that this study will help researchers in the identification and repurposing of multipotent drugs, simultaneously targeting the several viral and host proteins, for the treatment of COVID-19. ## Introduction: Novel zoonotic viruses with potential for rapid spread and significant pathology pose a grave threat to humans. During the last few decades many epidemics of viral diseases have occurred such as Ebola, Zika, Nipah, Avian influenza (H7N9), HIN1, Severe Acute Respiratory Syndrome Coronavirus 1 (SARS-CoV1), and Middle East Respiratory Syndrome Coronavirus (MERS-CoV)(1) (2). In the end of 2019, mysterious pneumonia cases begin to emerge in China's Wuhan city. A novel coronavirus, which was later renamed as severe acute respiratory syndrome coronavirus 2 (SARS-CoV2) was found to be the causative organism and the disease was termed "coronavirus diseases-19" (COVID-19) (3). COVID-19 is the world's worst pandemic and has so far affected about 5 million people globally (identified cases as on 19th May 2020), with average lethality of ~3%. Infection with SARS-CoV2 results in acute respiratory distress syndrome (ARDS) leading to lung injury, respiratory distress and lethality. Elderly patients and those with comorbidities have been reported to be at risk of higher mortality. The SARS-CoV2, SARS-CoV1 and MERS-CoV belongs to the family of Coronaviridae and β-coronavirus genus (4). While bats are considered to be the origin of SARS-CoV1 and SARS-CoV2, the intermediate host that led to human transmission of SARS-CoV2 is still unknown. Sequence analysis reveals that SARS-CoV2 is similar to coronavirus identified in Malayan pangolins (Manis javanica) (5). The SARS-CoV2 genome is 29.8 -29.9kb positive-sense single stranded RNA with 5′-cap and 3′-poly-A tail. Its genome is organised into two segments that encode non-structural (Nsp) and structural proteins. The first segment is directly translated by ribosomal frameshifting into polyprotein 1a (486 kDa) or 1ab (790 kDa) (ORF1a, ORF1ab) which results in generation of non-structural proteins and formation of replication-transcription complex (RTC) (1,6). Discontinuous transcription of the viral genome results in formation of subgenomic RNAs (sgRNAs) containing common 5′-and 3′leader and terminal sequences which serve as the template for subgenomic mRNA production (6). The ORF1a/1ab covers the two-thirds of the whole genomic length and encodes for the 16 non-structural proteins (Nsp1-16), which play critical role in various viral processes. The second segment at the 3′-terminus of the genome encodes the four main structural proteins: spike (S), membrane (M), envelope (E), and nucleocapsid (N) proteins (6). The life cycle of SARS-CoV2 starts with its entry into the host cell through receptor mediated endocytosis initiated by the binding of its Spike protein to the ACE2 receptor. Subsequently, uncoating of the virus particle releases the genome, which is translated to generate replication-transcription complex proteins. The viral RTC complex then generates full length negative sense RNA which is subsequently transcribed into full length genome. The viral genome and structural proteins are assembled into virions near the ER and Golgi interface and are transported out of the cell through vesicles by the process of exocytosis (7). The detailed understanding of the clinical manifestations and the underlying molecular mechanisms that drive disease pathogenesis are still unclear. There is no standard cure for the disease and currently the therapeutic regimen involves symptomatic treatment and previously approved drugs against other viral infections and diseases. Worldwide efforts to develop vaccines and drug against SARS-CoV2 are ongoing. Based on the similarity and information available from other coronaviruses, repurposing of approved drugs is among the best and rapid strategies to identify potential drug candidates (8). In this context, the computational techniques can quickly identify novel molecules that target viral proteins to suggest candidates for repurposing. Hence, during the COVID pandemic a lot of studies have been reported using a variety of such strategies (9). The in-silico studies have identified many drugs that can target viral proteins viz. RNAdependent RNA polymerase (RdRp), Spike, Membrane, 3CL pro and human proteins such as angiotensin converting enzyme 2 (ACE2) which serves as receptor for SARS-CoV2. Among them zanamivir, indinavir, saquinavir, lopinavir, and remdesivir are notable (10,11). There are many drugs such as baricitinib (12), lopinavir (10), ritonavir (13), remdesivir (14,15), hydroxychloroquine (16,17), arbidol (18) etc., that are currently under trial to treat SARS-CoV-2 infection. However, only a few studies have reported targeting more than one viral protein with a single molecule or using combination therapy. In this study we attempted to identify molecules that can simultaneously bind to multiple proteins of the SARS-CoV2. The strategy to target multiple proteins originates from the fact that individual viral proteins play specific role in multiple aspects of viral lifecycle such as attachment, entry, replication, morphogenesis and egress. Single molecules that can potentially target many viral proteins can perturb viral lifecycle at multiple points and thereby can be highly efficient in curbing SARS-CoV2 infection. In addition, the molecules that simultaneously target multiple viral proteins will have a higher barrier towards emergence of resistant mutants. In this work, we have used the 3D-structures of the SARS-CoV2 proteins to identify FDA approved drugs that can bind to these proteins using bioinformatics methods. The FDA approved drugs were chosen so that they can be quickly repurposed for treating COVID19. Additionally, we also analyzed if the identified molecules can affect the host proteins that get differentially expressed as a result of SARS-CoV2 infection. These molecules can be used as modulators of both the SARS-CoV2 and human proteins. ## Methods: Protein structure modelling: The SARS-CoV2 proteins for which there is no crystal structure reported were modelled using Modeller v9.22 (19) (homology modeling) or obtained from I-TASSER server (threading) (20). The modelling template for each protein was identified by performing Delta-BLAST against the PDB database. Proteins were modelled using either single or multiple templates based on the query coverage. The final homology modeling was performed using the modeller9. 22. Further, the model stereochemistry and other structural parameters were assessed using standalone PROCHECK tool. The proteins for which suitable templates were not found were obtained from I-TASSER server (https://zhanglab.ccmb.med.umich.edu/I-TASSER/). ## Molecular docking of FDA approved drugs in SARS-CoV2 proteins: The structures of the FDA approved drugs were obtained from the DrugBank (https://www.drugbank.ca/) repository. Ligands were prepared by Schrödinger LigPrep wizard ligands using the default parameters. The active site of the modelled proteins were identified using either of the two methods 1) the ligand binding pocket, if the co-crystal structures are available or 2) the active site was predicted using sitemap algorithm in Schrodinger v9.3 molecular modelling software (21). The proteins with active site pocket volume of <150 3 were removed as smaller pockets may not be amenable to docking. Finally, 13 proteins were selected for docking. The molecular docking was performed using the Glide module of Schrodinger molecular modelling software (www.schrodinger.com/glide). The molecules showing a docking score of -8.5 (roughly corresponding to 1 µm) (22) or better were selected for further analysis. ## The differential gene expression (DEGs) analysis: The differentially expressed genes were identified by analysing the data from NCBI GEO (GEO ID: GSE-147507) that contains data on cell lines infected with various virus including SARS-CoV-2. The above dataset also included an RNA sequencing study done using lung tissue of two normal and one COVID-19 patients. The DEGs were identified using limmavoom with the criteria of |log2FC≥1| and p-value≤0.01. The DEGs were further studied for their involvement in various pathways, processes and diseases using Ingenuity Pathway Analysis (IPA). ## Protein-protein interaction network analysis: The identified DEGs were mapped for their interactions with other human proteins using HIPPIE v2.2 which contains 14855 proteins and 411430 interactions. The reported proteinprotein interactions with a minimum score of 0.63 (medium confidence, 2 nd quartile) (23) were used for creation of the network using Cytoscape v3.7.2. The largest interconnected component was extracted and connectivity of individual nodes (degree) were calculated to assess their importance. The calculations were performed on the high performance Linux cluster. The flowchart of the methodology is presented in Fig. 1. The differentially expressed genes were identified using analysis of GEO data. Analysis was done to identify important host proteins, enriched pathways, and processes, etc. The DEGs were searched against databases to identify their modulators. ## Results and discussion: As stated earlier a total of 13 viral proteins (Table 1) were selected for molecular docking. The computational analysis of ligands binding to various proteins is a powerful method to quickly identify potential molecules for further analysis. These methods have been successfully used in various studies (24). In the first stage, the molecules were docked into the SARS-CoV2 proteins using Glide module of Schrodinger (www.schrodinger.com/glide) in standard precision (SP) mode. The redocking was done to ensure the appropriate selection of top hits. The molecules were then ranked using Glide score as implied in Schrodinger. We adapted the following notion for our drug repurposing analysis: 1. drugs that can inhibit viral entry into host cell by perturbing the function of surface glycoproteins like the spike, membrane and envelope protein. Preventing the function of other non-structural proteins that play accessory role in viral processes such as Nsp2, Nsp4 and Nsp10. 4. drugs that can also affect differentially expressed host proteins in COVID-19 along with the viral proteins. ## Molecules docking to SARS-CoV2 Structural proteins: The hallmark feature of coronaviruses is their transmembrane spike (S) glycoprotein as this protein is reason for its name "Corona" in Latin meaning, "Crown". The spike protein exists as homo-trimers. Each monomer is about 180kDa and has two distinct subunits S1 and S2. While octreotide, and lapatinib bind to spike protein with appreciable affinity (Fig. 2). Other groups have also predicted the binding of posaconazole to spike protein which further substantiates our analysis (27). Posaconazole is an antifungal agent used in the prevention of invasive fungal infections and is also shown to inhibit the entry of Chikungunya virus (28) and replication of Zika and Dengue viruses by binding to oxysterol-binding protein (sterol transporter) (29). Octreotide is a long-acting somatostatin analogue used for treatment of gastrointestinal tract bleeding, hepatocellular carcinoma and hemorrhage associated with Cytomegalovirus induced colitis (30) (31). Mefloquine is an antimalarial drug used in chloroquine resistant malaria. Nebivolol is an antihypertensive molecule with a very good safety profile in subjects with obstructive respiratory comorbidities (32) and can be an important drug to consider in SARS like diseases. The docking score of -8.5 indicates that nebivolol binds to spike protein with good affinity (Fig. 2). antibodies against N protein suggesting its role in eliciting humoral immune response (37,38). Our study predicts that ribavirin, vasopressin, octreotide, and capreomycin bind to N protein (Fig. 4). Of these, capreomycin, a polypeptide (isolated from Streptomyces capreolus) is used in the treatment of multidrug resistant tuberculosis. Previous studies report α-ketoamides, lopinavir and ritonavir as inhibitor of 3CL pro (45,46). Indinavir is shown to inhibit HIV protease by blocking its active site and leads to immature virus particle formation, however high doses have been linked to lipodystrophy syndrome (47). Naldemedine, is a μ-opioid receptor antagonist used for the treatment of opioid-induced constipation (48). Our study also predicts that tenofovir, nebivolol, ribavirin, nilotinib, lanreotide, ibrutinib, mefloquine, lopinavir, desmopressin, pasireotide, and methotrexate are among top molecules binding to the protease PL pro . An interesting observation is the identification of folic acid as a high affinity ligand of PL pro (Fig. 7). its activity (49). Our analysis shows that lanreotide, methotrexate, octreotide, cangrelor, and pibrentasvir bind to the helicase with high affinity (Fig. 8). Pibrentasvir, is a HCV NS5A inhibitor effective against all HCV genotypes (50). Methotrexate acts as an antimetabolite and thus used as an antineoplastic drug. It is also anti-inflammatory and used in treatment of inflammatory diseases like rheumatoid arthritis. It decreases the de novo synthesis of purines and pyrimidines and forms dimers with thymidylate synthase (TS), hence also used as antiparasitic drug. Methotrexate is also shown to effectively reduce replication of Zika and Dengue viruses (51). The most vital enzyme responsible for the replication/transcription of the viral genome is the RNA-dependent RNA polymerase (RdRp) also known as Nsp12. The primer for RdRp RNA synthesis is synthesized by Nsp8 (52). Our analysis shows that cobicistat, capreomycin, pibrentasvir, elbasvir, indinavir and remdesivir among others can bind with RdRp (Fig. 9). Cobicistat is known to inhibit the cytochrome-mediated metabolism of HIV protease and was approved in 2012 by FDA as pharmacoenhancer for HIV treatment (53). Other groups have also predicted that cobicistat and capreomycin can inhibit SARS-CoV2 protease (54) (55). Pibrentasvir and elbasvir are HCV NS5A inhibitors and indinavir is potent HIV protease inhibitor (56). The molecules we identified to bind to RdRp can serve as potential alternatives to remdesivir. The Nsp15 is EndoRNase with endoribonuclease activity. It cleaves the 5′ and 3′ of uridylate residues in RNA by forming 2′-3′cyclic phosphodiester. Its mechanism is similar to that of RNase A, RNAse T1 and XendoU (57). Its NendoU activity can interfere with the host's innate immune response and masks the exposure of viral dsRNA to host dsRNA sensors (58). In our analysis, drugs such as octreotide, desmopressin, macimorelin, and simeprevir were found to target Nsp15 (Fig. 10). Nsp14 is the 3'-5'exonuclease that plays a role in proofreading mechanism (59). Nsp14 contains four conserved DE-D-D acidic and a zinc-finger (ZnF) domain (60). Our molecular docking predicted that cangrelor, venetoclax, pimozide, nilotinib, droperidol, nebivolol, indacaterol, ezetimibe, simeprevir, siponimod, lapatinib, elagolix bind to Nsp14 (Fig. 11). Pimozide, a calmodulin inhibitor is shown to inhibit Chikungunya virus secretion (61). Moreover, it binds to the envelope protein of HCV and inhibits infection with many HCV genotypes (62). Ezetimibe is shown to inhibit formation of capsid-associated relaxed circular DNA of Hepatitis B Virus (HBV) (63) and is also shown to inhibit Dengue infection by interfering in formation of replication complex (64). Indacaterol is the β2-adrenoceptor agonist and used in the treatment of chronic obstructive pulmonary disease (COPD) since it induces bronchodilation effect (65). It is a promising candidate for therapeutics against SARS-CoV2 due to its ability to regulate genes involved in suppressing proinflammatory cytokine production and attenuation of airway hyper-responsiveness (66). However, dose and treatment schedule needs to be evaluated due to its counter effect on the expression of RNase L which is vital for antiviral response. respectively (41,81). Lanreotide was the only molecule that showed appreciable binding affinity with Nsp10 (Fig. 14). ## Drugs targeting multiple SARS-CoV2 proteins Cangrelor: is a P2Y12/13 inhibitor and is used as antiplatelet drug. The abnormal blood clotting is an increasingly recognized complication of COVID-19 and adverse prognosis. Therefore, it has been suggested to use thrombolytics/antiplatelet agents in the early stages of infection (82,83). Importantly in our screening Cangrelor was found to bind to multiple SARS-CoV2 proteins. Among the important targets are the main protease, spike protein, exonuclease, and helicase. Interestingly the target of Cangrelor in humans is P2Y12/13 which is also found to be differentially expressed as a result of SARS-CoV2 infection (Supplementary table 2). Additionally, there are no reported drug interactions between investigational COVID-19 therapies and Cangrelor (97). Therefore, this drug can be important for repurposing to treat the COVID-19. However, it was found to bind with many SARS-CoV2 proteins therefore it has to be seen whether it has a privileged scaffold or it is a promiscuous binder. The short plasma half-life upon intravenous administration may limit its efficacy against SARS-CoV2. Nilotinib: is a potent tyrosine kinase inhibitor and is used as an anticancer drug. It is found to bind to multiple SARS-CoV2 proteins. It was intriguing to see it binding with two nonstructural proteins (Nsp2 and Nsp14) with high affinity. A literature survey showed that it has been reported to exhibit antiviral activity against Human Cytomegalovirus (99). The mechanism is not very clear; however, it has been surmised that it may disrupt the productive replication of the virus. The SARS-CoV2 coronavirus depends on Abl2 kinase activity to fuse and enter into the cells. The kinase inhibitor imatinib is already under clinical trials for COVID-19 (https://clinicaltrials.gov/ct2/show/NCT04357613). Thus, other kinase inhibitors that are binding to viral proteins can also be potential candidates for repurposing. Lapatinib: is another potent kinase inhibitor, similar to nilotinib, that is showing good binding with the SARS-CoV2 proteins (S-protein, Nsp4 and Nsp13). Lapatinib is a HER2 (ligandindependent receptor tyrosine kinase) inhibitor used in the treatment of HER2-positive breast cancers (84). Interestingly, HER2 inhibition is shown to activate TBK1 through cGAS-STING pathway which plays a crucial role in anti-viral innate immune signaling (80). Hence, lapatinib can be effective in perturbing SARS-CoV2 replication as well as upregulating anti-viral signaling. Lancreotide: a long-acting analog of the drug somatostatin widely used in the treatment of Graves' ophthalmopathy, Acromegaly and Endocrine tumors ( 85),( 86),( 87), (88), is found to have high binding affinity to multiple SARS-CoV2 proteins such as PL pro (Nsp3), Nsp10, 13 and 16. Octreotide: another somatostatin analog similar to lancreotide. It is used in the treatment of diarrhoea, pancreatic neuroendocrine tumors and massive hemorrhage caused by cytomegalovirus colitis (89),( 90), (40). This drug also binds to the multiple proteins of SARS Cov2 virus which includes the structural proteins S and N and the helicase, endonuclease and methyl transferase. Study suggested, several anticancer drug have potential target of SARS-Cov2 as repurposing drugs (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7239789/). Capreomycin: a polypeptide isolated from Streptomyces capreolus used in the treatment of multidrug resistant tuberculosis and co-infected (HIV) is also found to bind with 2'OMT, nuclease and RdRp proteins. However, this compound has serious nephrotoxicity and ototoxicity, which has to be taken in into account (39). Pibrentasvir: the hepatitis C virus NS5A inhibitor was found to bind with helicase and RdRp proteins. Indinavir: HIV protease inhibitor and major component of highly active antiretroviral therapy (HAART) for treatment of HIV/AIDS was found to bind RdRp and 3CL pro protease with good affinity. Venetoclax: is a BCL-2 (antiapoptotic protein) inhibitor is used for the treatment of chronic lymphocytic leukemia (CLL) or small lymphocytic lymphoma (SLL) (93). It is found to bind with Nsp16 and exonuclease. Alatrofloxacin a fluroquinolone antibiotic that targets the DNA gyrase enzyme was found to bind with Nsp2 and Nsp4 proteins with good affinity. A heatmap (Fig. 15) was generated using the docking scores to summarize the binding of important drugs to multiple proteins. The detailed list drugs and their docking scores is given in supplementary table 1. As stated earlier, the differentially expressed genes were identified by analyzing the data from NCBI GEO (GEO ID: GSE-147507) using limma-voom with the criteria of |log2FC≥1| and p-value≤0.01. The differentially expressed genes were then analyzed using IPA and Cytoscape. Details are given in Supplementary Table 2. A protein-protein-interaction network analysis was done using Cytoscape to identify the network hubs based on their interactions with other proteins using Degree centrality. The giant component was extracted from the network with 1446 nodes containing 1770 interactions. Top 5% of the proteins (total 72) were selected as hubs based on the degree centrality for further analysis. A search for the DEGs (136) at drug-gene interaction database (DGIdb) resulted in the identification of 352 drugs. (Supplementary table 3). Further, we wanted to know if some of them can also target viral proteins. An intersection with drugs identified early revealed 29 drugs that can target both viral and human proteins. It is worth to mention that among the human proteins many of them (e.g. IL2RG, DAPP1, CCL7, IFIT1, MMP8, FOS) are hubs i.e. very important proteins in the generated protein-protein-interaction network. Therefore, the drugs with multi-targeting ability against these proteins as well as SARS-CoV2 proteins can have a significant therapeutic utility for COVID-19 which is a novelty of this study (Supplementary Table 4). The analysis indicated that some of proteins upregulated during SARS-CoV2 infection are also targeted by the drugs identified to bind to viral proteins in our analysis. TUBB48 (Tubulin Beta 4A Class IVa) is upregulated during SARS-CoV2 infection. Previous study suggests that druginduced microtubule depolymerization results in reduction of infectious virus particle release due to defect in spike protein incorporation into the virions. Paclitaxel, which targets the N protein, is a cytoskeletal drug which stabilizes the microtubule polymer formation and protects it from disassembly (94). The transcription complex Activator protein 1 (AP1) is composed of homo/hetero dimers of Fos, Jun, CREB and others ATFs. The studies on the SARS-CoV1 infection in the Vero and Huh7 cell shows that nucleocapsid protein is the potent activator of (AP-1) (95). Interestingly, asthmatic patients show higher expression of c-fos in their epithelial cells. It is also observed that TNF-α induced ROS and intracellular glutathione depletion in the airway epithelial cells induces the production of AP-1 and leads to the pulmonary fibrosis (96,97). Our analysis suggests that paclitaxel and bromocriptine which dock with nucleocapsid and Nsp4 proteins can also effectively bind to c-Fos and thereby would be beneficial in inhibiting the SARS-CoV2 as well as in alleviating lung injury observed in COVID19. The transcriptome analysis revealed that S100/calgranulin is upregulated during SARS-CoV2 infection. Calgranulin is polypeptide released by the activated inflammatory cells such as leukocytes, PBMC phagocytes and lymphocytes and is accumulated at the sites of chronic inflammation. It is the ligand for RAGE receptors and is the major initiator of cascading events in inflammation amplification (98). Interestingly, the differential gene expression analysis of PBMCs from SARS patients shows the higher expression of this calgranulin families (97) and this protein is found in higher quantity in the Bronchoalveloar Lavage Fluid (BALF) and sputum of patients with inflamed lungs, COPD, and ARDS (99). Our analysis suggests that the anti-inflammatory agent methotrexate which has high affinity to the Nsp13 protein of SARS-CoV2 also shows appreciable binding to calgranulin and can thereby be useful to curtail systemic inflammation in lungs observed during COVID19 in addition to its inhibitory effect on SARS-CoV2. Transcriptomic analysis also suggests increased expression of endogenous prolactin, which leads to prolactin induced STAT5 activation and its pathways. Prolactin has a dual role in human physiology functioning as a hormone (secreted from anterior pituitary gland) and cytokine (secreted by immune cells). It causes anti-apoptotic effect and induces proliferation in immune cells in response to antigens leading to increased production of immunoglobulin, cytokines, and autoantibodies (100). We envisage that prolactin may be one of the significant player in trigger of cytokine storm implicated in COVID19. Interestingly, our study suggests that zidovudine which target the O'-methyl transferase (Nsp16) can also bind to prolactin and can be of high significance in management of COVID19 due to dual ability to affect Nsp16 and prolactin. ## Conclusions The overall goal of this study was to identify molecules that can dock with multiple SARS-CoV2 proteins that play vital role(s) in the viral lifecycle. Our study predicted several promising drug candidates with high binding affinity towards many of SARS-CoV2 proteins. These drugs will be very effective than drugs that target single viral proteins due to their ability which targets the exonuclease is also a promising agent due to its ability to regulate genes involved in suppressing pro-inflammatory cytokine production and attenuation of airway hyper-responsiveness (71). We also observed that the molecules that bind to Nsp4, 13, 16 and N proteins can also bind to human proteins that play a pivotal role in disease pathogenesis by promoting inflammatory signalling leading to cytokine storm thereby suggesting that the molecules such as a paclitaxel, methotrexate, and zidovudine can have a dual beneficial effect in the management of COVID19. Overall our study predicts promising agents with potential to inhibit crucial viral processes, upregulate anti-viral host response and alleviate severe lung disease condition thereby providing attractive avenues for design of potential and multipronged therapeutic strategies against COVID 19.

chemsum

{"title": "Identification of drugs targeting multiple viral and human proteins using computational analysis for repurposing against COVID-19", "journal": "ChemRxiv"}

catalytic_oxidative_desulfurization_of_a_4,6-dmdbt_containing_model_fuel_by_metal-free_activated_car

## Abstract: Commercial micro/mesoporous activated carbons were utilized as metal-free catalysts for the desulfurization of a model fuel, i.e. 4,6-dimethyldibenzothiophene (4,6-DMDBT) in hexadecane under ambient conditions. Both adsorption and catalytic oxidation were investigated as means of 4,6-DMDBT removal.The effect of chemical modification/oxidation of the carbon surface via treatment with two different acids (HNO 3 or H 2 SO 4 ) aiming to introduce additional functional groups was also investigated. The catalysts were characterized by FT-IR spectroscopy, N 2 porosimetry, potentiometric titration, Boehm titration, and SEM-EDX, while adsorption and catalytic oxidation activity towards sufloxides and sulfones were assessed by GC-MS and UV-Vis analysis. The surface chemistry of the carbons, expressed by the density of the acidic functional groups, was found to be the most critical parameter with regard to adsorption or to catalytic oxidative performance. The surface modification of carbons by oxidation had a positive impact on the catalytic oxidation activity, leading to a 100% conversion of 4,6-DMDBT towards the corresponding sulfoxide and sulfone, compared to 67% with the parent non-oxidized carbon. Reusability tests showed that the oxidation activity of the carbons can be maintained for at least 5 cycles. † Electronic supplementary information (ESI) available. See ## Introduction Nowadays, great interest has been focused on the mitigation of sulfur compounds in fuels in order to comply with stringent regulations, as they cause environmental problems and human health issues. According to the Environmental Protection Agency of USA (EPA), the admissible sulfur concentration in diesel fuel has been 15 ppm since 2006. 1,4 The European Parliament and the Council of the European Union have established in 2009 an even lower concentration of sulfur in fuels, 10 ppm. 5 The industrial process used for decades to remove sulfur from fuels is hydrodesulfurization (HDS), a catalytic process in which organic sulfur compounds are converted to hydrogen sulfide and sulfur-free hydrocarbons by reaction with hydrogen over CoMo/Al 2 O 3 or NiMo/Al 2 O 3 catalysts. 6,7 An important drawback of HDS is the required high hydrogen pressure, 4 leading to a very costly process, especially when deep desulfurization is aimed. 8,9 Furthermore, HDS is not effective in the removal of sulfur heterocyclic hydrocarbons such as dibenzothiophene (DBT) and its derivatives, especially 4,6-dimethyldibenzothiophene (4,6-DMDBT), which is not so reactive due to steric hindrance effects. 4, Within this context, the development of more efficient and greener/sustainable desulfurization methods which have the potential to produce extremely ultra-low-sulfur fuels, so as to replace or complement the HDS process, is of great importance. Adsorption, 14,15 biodesulfurization, 16,17 extraction by ionic-liquids, 18 photocatalytic oxidation 19 and oxidative desulfurization (ODS) 20,21 are some new processes that have been introduced for efficient fuel desulfurization. Among these approaches, oxidative desulfurization constitutes a promising method, as it is simple and of higher efficiency compared to HDS. 22,23 In addition, 4,6-DMDBT is expected to exhibit higher reactivity in ODS. 24,25 The ODS method involves the oxidation of the sulfur-containing compounds, followed by the extraction/removal of the oxidized products from the fuel due to their polarity. In the ODS process, sulfur-containing compounds are oxidized using selective oxidants such as nitric acid, 26 nitrogen oxides, 27 organic hydroperoxides, 28 hydrogen peroxide 24,29,30 or/and ozone 31 in the presence of a catalyst to produce sulfone compounds which can be extracted preferably due to their increased polarity. 32 The most commonly used oxidant is hydrogen peroxide (H 2 O 2 ), due to the fact that it is environ-mentally friendly, it has low cost and it is commercially available. 29 Various homogeneous and heterogeneous catalysts like polyoxometalates (POMs), 33 mono-or bimetallic alloys, 34 organic acids, 35 ionic liquids, 18,36 multi-walled carbon nanotubes 37 and activated carbons 38 have been shown to be active in the desulfurization reactions. In particular, the latter ones have been previously utilized as supports for active phases like metal oxide, metal or bimetallic nanoparticles, as for example in the deposition of molybdenum cobalt nanocatalysts on a carbon support which led to 34% increment of dibenzothiophene hydrodesulfurization from a model oil, compared to pure MoCo. 44 Even though carbons are often used as catalyst supports, they can also be utilized as catalysts on their own, because of their physicochemical properties, such as the rich surface chemistry due to the presence of oxygen containing functional groups, the large micro/mesopore surface area, and their relatively stable structure and morphology at high temperatures and/or various liquid reaction media. One of the most known reactions catalyzed by carbonaceous materials, the decomposition of hydrogen peroxide, is highly dependent on the nature of the surface functional groups. Chemical reactions on carbon's surfaces usually follow a free radical mechanism. 48,49 The produced radicals are stabilized on the surface of the carbon so that they can act as adsorption/oxidation sites of sulfur compounds. 48,50 Despite the wide application of carbons in environment related reactions and processes, limited investigation has been devoted to the use of activated carbons as metal-free catalysts in desulfurization processes and especially for ODS of 4,6-DMDBT. In most cases, the carbons were studied as adsorbents for the removal of smaller DBT or other derivatives. 22,50 Considering that the production of carbons with different morphology, texture and surface properties can be achieved by utilizing the most abundant and renewable source, biomass, the use of biomass derived carbocatalysts will be of great importance towards a sustainable future. In the present study, we examined five activated carbons, with or without prior treatment with acids, for the oxidation of 4,6-DMDBT within the process of fuel desulfurization. Emphasis was given to the investigation of the role of carbon's textural characteristics and surface chemistry features towards maximization of desulfurization activity. ## Materials and reagents Five commercial activated carbons, obtained from Cabot Norit activated carbon and CPL activated carbons, were chosen to be tested as carbocatalysts: (i) Norit SX-PLUS, (ii) Norit SAE-SUPER, (iii) Norit D-10, (iv) Norit SAE-2, and (v) CPL. 4,6-DMDBT (4,6-dimethyldibenzothiophene), hexadecane, commercially available 30% wt% H 2 O 2 , methanol, HNO 3 , and H 2 SO 4 99.999% purity were purchased from Sigma Aldrich. ## Modification of the carbon catalyst The SX PLUS carbon was chemically treated either with HNO 3 or H 2 SO 4 , targeting to modify its surface chemistry. 2.2.1. Oxidation with HNO 3 . For the preparation of the HNO 3 oxidized activated carbon sample, 10 g of the activated carbon SX PLUS was oxidized in a 70% HNO 3 solution (100 mL) for 4 hours under vigorous stirring at room temperature. The excess of acid and the possibly formed soluble products upon the oxidation process were removed by filtration and extensive washing of the carbon sample in a Soxhlet apparatus, until constant pH. 4 The obtained material was oven dried at 60 °C for 24 h. This carbon sample was named SX PLUS N-ox. 2.2.2. Oxidation with H 2 SO 4 . The H 2 SO 4 modified activated carbons were prepared by oxidizing 10 g of the activated carbon with concentrated H 2 SO 4 (100 mL) at 60 °C for 4 hours under stirring. The oxidized carbons were recovered by filtration, washed thoroughly in a Soxhlet apparatus until constant pH and dried at 60 °C for 24 h. 55 This sample is referred to as SX PLUS S-ox. ## Materials characterization methods 2.3.1. Textural/porosity characterization of activated carbons. The textural characterization of the activated carbons was carried out by measuring the N 2 adsorption/desorption isotherms with an AS1Win (Quantachrome Instruments, FL, USA) porosimeter. In a typical measurement, 0.05 g of the activated carbon were initially outgassed under vacuum at 150 °C overnight, followed by determination of the N 2 adsorption and desorption isotherms at −196 °C. The BET surface area was calculated from the isotherm data using the Brunauer, Emmett and Teller (BET) equation, while the pore size distribution curves were estimated using the DFT method. 56,57 2.3.2. pH measurement. The pH of the activated carbons provides information about the acidity and basicity of their surface. For the pH measurement, 0.4 g of the carbon was added to 20 ml of deionized water and the suspension was kept under magnetic stirring at room temperature for about 24 hours to achieve equilibrium. The pH of the solution was then measured using a CRISON basic-20 pH meter. 2.3.3. Boehm titration. The oxygen containing groups located on the surface of the activated carbons were determined by Boehm titration. This method relies on the different acidity of the surface groups, where each group is neutralized with a different reagent of similar activity. 58 In particular, sodium bicarbonate (NaHCO 3 ) neutralizes only carboxyls belonging to the strong acidic groups on the carbon surface. Sodium carbonate (Na 2 CO 3 ) neutralizes carboxyls and lactones and sodium hydroxide (NaOH) neutralizes carboxyls, lactones and phenols belonging to the weaker acidic groups. The determination of the basic surface groups was performed with hydrochloric acid solution, HCl. 59 In a typical measurement, 1 g of sample was placed in 50 ml of 0.05 N solution of each base (NaOH, NaHCO 3 , Na 2 CO 3 ) or the acid (HCl) in sealed conical flasks under stirring for 24 h at ambient temperature. Then, the solution was filtered, and the specific volume of the filtrate was titrated. Thus, the excess base or acid in the solution was neutralized with HCl or NaOH, respectively. The numbers of total surface acidic and basic sites, as well as the content of carboxylic, phenol and lactone groups, were thus determined. 2.3.4. Potentiometric titration. Potentiometric titration measurements were carried out with a Mettler Toledo T50 automatic titrator. In a typical measurement, 0.1 g of activated carbon was placed in a conical flask with 50 mL KNO 3 solution (0.1 mol L −1 ) under stirring for 24 h, at 25 °C. The solution was then titrated with NaOH solution (0.1 mol L −1 ) under a N 2 atmosphere over a wide pH range. The total surface charge, Q (mmol g −1 ), was calculated as a function of pH from the following equation: 4 where C A and C B (mol L −1 ) are the acid (HCl) and base (NaOH) concentrations (mol L −1 ), respectively, [H + ] and [OH − ] are the equilibrium concentrations of these ions (mol L −1 ) and W is the concentration of the solid (g L −1 ). 2.3.5. Point of zero charge. A certain volume of 0.01 M NaCl solution was placed in titration vessels at constant temperature (25 °C) and 0.05 g of activated carbon was added to each vessel. The pH initial of the dispersions was adjusted to values between 3 and 9 (3, 5, 7, 9) and the suspensions were allowed to equilibrate, under stirring, for 48 h. The final pH was measured and was plotted for each dispersion against the initial pH. The pH at which the curve crossed the line pH initial = pH final was taken as the point of zero charge (PZC). 60 2.3.6. Fourier transform-infrared spectroscopy (FTIR). Fourier transform-infrared (FTIR) spectra (10 scans per measurement) were recorded on a PerkinElmer 2000-FTIR spectrophotometer (Dresden, Germany) in the wavenumber range of 4000-450 cm −1 , by applying the KBr-pellet technique. 2.3.7. Energy-dispersive X-ray spectroscopy (EDX). Elemental analysis was performed by EDX using a scanning electron microscope, model Zeiss Supra 55VP, Jena, Germany. The hermetically sealed conical flasks were placed in a shaking bath for 48 h at constant temperature. After the equilibrium time (evaluated from kinetic studies), the remaining concentration of 4,6-DMDBT was determined by using a UV-vis spectrophotometer (Hitachi U2000) at a wavelength of 313 nm using a quartz cuvette. The removal percentage (R%) of 4,6-DMDBT was calculated by the equation: where C 0 and C e ( ppmw of sulfur) are the initial and the equilibrium concentrations, respectively. 61 Kinetic experiments were performed by dispersing 0.01 g L −1 of carbon in 20 mL of 4,6-DMDBT solution in hexadecane (C 0 = 20 ppmw of S), under stirring at 60 °C for different intervals of time. 62 For each separate experiment, filtration through a 50 μm pore size membrane was performed and the 4,6-DMDBT concentration was measured by using a UV spectrophotometer as mentioned above. The experimental results were fitted by Lagergren's pseudofirst order kinetic model, 63 given by the equation lnðq e q t Þ ¼ ln q e k 1 t where q t and q e are the amounts of 4,6-DMDBT (mg g −1 ) adsorbed at time t and at equilibrium, respectively and k 1 is the rate constant of the pseudo-first order adsorption process (min −1 ), as well as by the pseudo-second order kinetic model, 64 given by the following equation in linear form, where k 2 , the rate constant (g mg −1 min −1 ): 1 q e t 2.4.2. Catalytic oxidation of 4,6-DMDBT/oxidative desulfurization. The model 4,6-DMDBT solution in hexadecane was of 20 ppmw concentration in sulfur. In a 50 ml round-bottom flask with a reflux condenser, 0.025 g of the carbon catalyst were weighed, and then 10 ml of the model solution and 1 ml H 2 O 2 (30 wt%) were added. The catalytic reactions were conducted at 60 °C for 24 h, under mechanical stirring. After the reaction, the mixture was filtered and the residual amount of 4,6-DMDBT in the solution was determined using UV-Vis spectroscopy at λ = 313 nm. The liquid products were also analyzed by GC-MS on a 7890A/5975C system by Agilent (electron energy 70 eV, helium flow rate: 0.7 cc min −1 , Column: HP-5MS 30 m × 0.25 mm ID × 0.25 μm). Identification of mass spectra peaks was performed by the use of the scientific library NIST11s. Extraction solutions of the spent samples were also analyzed with the same instrument. ## Characterization of activated carbon The porosity characteristics and surface chemical features of the different activated carbons were initially determined and correlated with the adsorption and catalytic oxidation of 4,6-DMDBT. The N 2 adsorption-desorption isotherms of all the carbon samples (Fig. s1a †) showed a bimodal type of shape based on IUPAC classification. 65 At low relative pressures, there is a clear typical type I adsorption isotherm for all parent commercial carbons, revealing the abundance of micropores. At intermediate and higher relative pressures, the adsorption isotherms tend to adopt a type IV shape with the corresponding hysteresis loop, which was more pronounced for the CPL and SAE SUPER carbons, due to capillary condensation in narrow slit-shaped mesopores. 66,67 The pore size distribution curves (Fig. s1b †) revealed that there is a relatively narrow distribution in the low micropore range, i.e. <1 nm, and a broader distribution for pores >1 nm, except for the case of CPL which showed a relatively narrow distribution between 1 and 2 nm. The BET surface area values and porosity parameters for the activated carbons are presented in Table s1. † It is seen that CPL (1702 m 2 g −1 ) presented the highest surface area, followed by SX PLUS (1280 m 2 g −1 ), SAE SUPER (1182 m 2 g −1 ), and SAE2 (892 m 2 g −1 ), while D10 showed a relatively low surface area (515 m 2 g −1 ). The total pore volumes (Table s1 †) were found to follow the same trend as the SSA BET . The values of the microand meso-pore volume are shown in Fig. 1a (also listed in Table s1 †). CPL revealed also the highest ratio of mesopore to micropore or total pore volume, while SX PLUS the lowest. Considering that the molecular size of 4,6-DMDBT is 0.59 × 0.89 nm, 4 it could be expected that micropores of ≤1 nm, as well as bigger micropores and small mesopores (up to ca. 2-3 nm), would play an important role regarding the adsorption and oxidation activity of the carbons. The surface chemistry characteristics of the activated carbons are reported in Table s2. † From the surface pH measurements, it can be seen that D10, SAE SUPER, and SAE 2 activated carbons possess a relatively basic surface, SX PLUS almost neutral, while CPL exhibits a more acidic nature. The results from the point of zero charge ( pzc) determination (Fig. s2 †) are in-line with the proton binding curve potentiometric titration results (Fig. s3 †). The most interesting observation can be derived from the ratio of acidic to basic groups, with CPL having only acidic groups, followed by SX PLUS (0.70), while SAE SUPER has the lowest ratio (0.08). The calculated density of acidic surface functional groups (d ASFG ) and of the basic (d BSFG ) per surface area is presented in Fig. 1b. The Boehm titration results revealed mainly two kinds of oxygen functional groups on carbon's surface, i.e. lactonic and phenolic, while no carboxyl groups were detected. The amount of each type of group (mmol g −1 ) as well as their density per specific surface area unit (μmol m −2 ) is collected in Table s2. † SX PLUS shows the highest number of lactones, while SAE2 the highest amount of phenol groups. ## Desulfurizationadsorption and catalytic oxidation results for the commercial carbons The 4,6 DMDBT (20 ppmw of sulfur) adsorption results in hexadecane as a solvent in the dark, at 60 °C, without H 2 O 2 are presented in Fig. 2a. SX PLUS showed by far the best performance, achieving 60% removal while the lowest adsorptive capability was observed for the carbon with the lowest surface area and pore volume, D10. On the other hand, the activated carbon CPL with the highest BET surface area exhibited a moderate adsorption efficiency. Furthermore, SX PLUS and SAE SUPER have a similar surface area, but the former showed substantially higher adsorption of 4,6-DMDBT. It can thus be concluded that the BET surface area of the carbons is not the sole critical parameter that would define the adsorption performance. If one considers the correlation between the size of 4,6-DMDBT and the size of the pores, it appears that the carbons with a higher abundance of smaller micropores, i.e. with a size of ≤1 nm, such as SX-PLUS, SAE-2 and SAE SUPER, are more efficient possibly due to increased confinement effects due to size similarity. With regard to the effect of the functional groups and correlation with the surface pH, SX PLUS and CPL provided a neutral/slightly acidic pH, compared to SAE-2 and SAE SUPER which were basic, and furthermore, they exhibited the highest ratio of acidic to basic surface groups, i.e. 1.0 for CPL and 0.7 for SX PLUS. In order to further investigate the possible combined effects of the surface area and acidic sites, the density of the acidic surface functional groups, per surface area unit, was estimated and it was plotted against the removal efficiently (Fig. 2b). As can be seen in Fig. 2b, the maximum adsorption capacity increases with the increment of the density of acidic surface functional groups (d ASFG ) up to an optimum value followed by a decrement upon further increase of d ASFG . This fact can be assigned to an enhanced steric effect when the active surface functional groups are close to each other, as well as to a blockage of the pores' entrance by adsorbed molecules, hindering the penetration of other molecules towards the active adsorption sites at the interior of the pores. This behavior leads to the suggestion that the density and dispersion of the active sites may be an important parameter for the adsorption efficiency of the carbons. However, when comparing the properties and performance of SX PLUS and CPL, it can be seen that the former contains a high portion of lactones while phenolic groups are predominant in the latter (Fig. 1c). This may also lead to the assumption that the more acidic lactones compared to phenols are more prone to interact with 4,5-DMDBT, considering also the relatively (i.e. compared to pyrrole for example) moderate Lewis basicity strength and enhanced aromatic stabilization of DMDBT. Overall, it can be suggested that both the type of acidic functional group and their density are key factors towards increasing the adsorption efficiency of the carbons. Moreover, the Langmuir 68 and Freundlich 69 isotherms were fitted to the experimental results, as presented in Fig. s4, † and models' parameters are presented in Table s3. † Both models exhibit relatively good fitting indicating the initial monolayer formation of sorbed 4,6-DMDBT as well as the formation of secondary layers, due to the presence of meso/macroporosity as well as due to the inhomogeneity of the adsorption sites. The models' parameters are presented in Table s3, † from where it is seen that SX PLUS presented the highest adsorption capacity between all carbons. The potential of fuel desulfurization via catalytic oxidation was investigated by the use of H 2 O 2 as the oxidation agent with the parent commercial activated carbons as catalysts (Fig. 2a ## Effect of oxidation reaction parameters The effect of different parameters i.e. catalyst's dose, reaction temperature, and H 2 O 2 relative amount, on the catalytic oxidation was also studied, using SX PLUS being identified above as the most efficient. As mentioned above, no conversion of 4,6-DMDBT was observed in the presence of H 2 O 2 without using carbons at 60 °C, showing that the activated carbon actually acts as a catalyst in the reaction. 70 The 4,6-DMDBT removal efficiency with different amounts of catalysts is shown in Fig. 3a. It is observed that by increasing the amount of carbocatalyst, the removal extent is enhanced. More than 90% removal is achieved by a relatively low amount of carbon (100 mg). 25 mg can be assumed as the optimum one, since the removal extent by a further increase of the carbon amount is minimal. The effect of temperature was also investigated by carrying out catalytic experiments at three different temperatures, 25, 60 and 90 °C and the results are shown in Fig. 3b. From 25 to 60 °C, a clear improvement in the removal efficiency can be observed from ∼39% to ∼76%. This can be linked to a variety of factors, like activation of the acidic surface functional groups and increase of the thermodynamics of the system, resulting in enhanced mass transfer (diffusion inside the pores towards the active sites) and decrement of adsorption/ decomposition activation energy. In contrast, a further temperature rise to 90 °C had a negative effect, decreasing the removal performance to ∼47%. This can be assigned to an increase of the hydrogen peroxide decomposition rate, which may also lead to undesirable secondary species other than the targeted hydroxyl radicals. Besides, oxidation of useful components in the fuel may also occur at elevated temperatures. This result is in agreement with other research results reported in the literature for the same desulfurization via adsorption/ oxidation methods. 30, In view of these results, the optimum reaction temperature was set at 60 °C. The results of the effect of the amount of H 2 O 2 on the removal of 4,6-DMDBT are shown in Fig. 3c. It is observed that by increasing the volume ratio, and therefore the molar ratio of H 2 O 2 , the removal percentage of 4,6-DMDBT is enhanced during the catalytic oxidation. The maximum removal was for 1 mL of H 2 O 2 solution (25 mg SX PLUS, 20 ppmw of sulfur PLUS) and the results are consistent with the literature. 70 An important conclusion was gained for the test in which only one 1 mL of water was used (no H 2 O 2 addition); the removal efficiency was limited to ∼44%, a value even lower than the maximum adsorptive removal without hydrogen peroxide (adsorption tests, Fig. 2a). This is strong evidence that the presence of water has a negative impact on the removal of 4,6-DMDBT from hexadecane (or a real fuel) by blocking/hindering the active adsorptive/reactive centers. The above experiments (referring to the results of Fig. 3) were also carried out in the dark in order to explore if the presence of ambient light has an effect of the adsorption/oxidation capability. The catalytic oxidation removal under ambient light exposure was 10% higher than that in the dark. This can be linked either to the fact that the light plays a positive role in the reactive and/or adsorptive sites or that light power can enhance the formation of reactive oxygen species during the decomposition of H 2 O 2 on the carbon's surface. These two aspects will be discussed herein after in more detail. ## Reusability cycles To investigate the possibility of regeneration and reusability of the catalyst, the carbocatalyst SX PLUS was washed with water and methanol (dried for 2 h at 60 °C) after the oxidation experiments, in order to remove any product or reactant from its surface. From the results presented in Fig. 4, a small decrease of 4.5% in the removal efficiency occurred after the first cycle of reuse, while after 5 cycles the removal efficiency loss was not higher than 10.8%. It can be concluded that the catalytically active sites are maintained during reaction/regeneration. It is worth mentioning that the removal extent (∼63%) even after 5 cycles is higher than the maximum adsorptive capability of the rest of the commercial carbons tested (Fig. 2a). ## Adsorptive/catalytic performance of oxidized carbons Taking into account the high adsorption and catalytic oxidation efficiency of the parent commercial SX PLUS carbon, this carbon was further used to investigate the effect of chemical modification of its surface by increasing the density of the acidic surface functional groups. To this end, two different counterparts of modified SX PLUS were prepared. The first one was obtained after treatment with HNO 3 , herein-after referring to as SX PLUS-Nox, while for the second one H 2 SO 4 was used, with the prepared sample named SX PLUS-Sox. The catalytic oxidation results of the HNO 3 and H 2 SO 4 oxidized counterparts, as well as of the adsorption data (included for the sake of comparison), are presented in Fig. 5. A significant increase of the removal efficiency via catalytic oxidation can be observed for both treated SX PLUS samples, reaching 100% removal, being an additional indication of the positive effect of the surface acidity on the catalytic oxidation by activated carbons as metal-free catalysts. A small positive effect can also be identified in the adsorption removal, verifying further the enhanced interaction between 4,6-DMDBT and the surface acidic sites of carbon. ## Characterization of oxidized carbons The porosity characteristics of the parent and oxidized SX PLUS carbons are shown in Table 1. Both the BET surface area and total pore volume were decreased upon oxidation. The decrement of the S BET was slightly more pronounced after treatment with HNO 3 (−27%) rather than after H 2 SO 4 treatment (−22%). Similar trends of decrease upon oxidation were found for the pore volumes. It is worth pointing out that although the oxidized counterparts showed inferior porosity characteristics in comparison with the parent SX PLUS, their adsorptive performance was superior, a fact that confirms a more crucial role of carbon's surface chemistry against porosity. The surface pH as well as the amount of acidic and basic surface groups of the SX PLUS carbons is shown in Fig. 6, while the densities per surface area unit can be seen in Table s4. † The general observation is that the amount of the acidic surface groups was increased upon oxidation and the amount of the basic surface groups was decreased, thus resulting in a decrease of surface pH. The increment of the acidic groups was substantially more pronounced in the case of SX PLUS-Nox, while the decrement of the basic groups was more pronounced for SX PLUS-Sox. Boehm titration experiments (results not shown) revealed that the phenol groups were totally eliminated, lactones were increased, and carboxylic acids were formed by the oxidative treatment. With regard to the correlation between adsorptive/catalytic performance and the surface acidity, the above derived volcano-type trend between adsorption of 4,6-DMDBT and the density of surface acid groups (Fig. 2b) was found valid also for the parent SX PLUS and its oxidized counterparts, since SX PLUS-Nox showed a significantly higher d ASFG but a slightly lower adsorption performance compared to SX PLUS-Sox. On the other hand, in the case of catalytic oxidation performance, both treated carbons reached the maximum removal efficiency, as shown in Fig. 5, at least for the experimental conditions used in this study. ## Effect of contact timekinetics of the 4,6-DMDBT removal In order to examine the effect of the adsorption/oxidation reaction time on the removal efficiency of the carbon catalysts, the remaining sulfur in the solution, expressed as C/C 0 , was plotted against the contact time and the derived curves are presented in Fig. 7. A different 4,6-DMDBT removal rate can be observed between the adsorption and the catalytic oxidation process. In the adsorption experiments, 4,6-DMDBT removal Table 1 Porosity characteristics of SX PLUS and its counterparts oxidized by HNO 3 and H 2 SO 4 (in parenthesis the % differences for the oxidized samples compared to SX PLUS) ## Activated carbons S BET , m 2 g −1 V tot , cm 3 g −1 V mic , cm 3 g −1 V mes , cm 3 increased rapidly in the first 5-10 min followed by a gradual equilibration and a plateau after ca. 50 min, for both the parent SX PLUS and the two oxidized counterparts. The pseudo-first and the pseudo-second order kinetic models in their linear form were applied for the fitting of the experimental results. The linear fitting curves and the corresponding kinetic parameters are shown in Fig. s5 and Table s5, † respectively. The pseudo-second kinetic model found to fit better the adsorptive removal results, as concluded by the R 2 values presented in Table s5, † indicative of a physical adsorption onto the carbon surface, results consistent with findings by other researchers. 73,74 The kinetic experiments of the 4,6-DMDBT catalytic oxidation showed that 100% removal was achieved for the SX PLUS-Nox and SX PLUS-Sox carbons within 10 and 16 min, respectively. Interestingly, more than half of the 4,6-DMDBT was converted within less than 2 min. Since two separate slopes were observed (a steep one for 0 to 2 min and one for 2 to 10 or 16 min), it can be suggested that initially the catalytic oxidation/removal occurs on the large internal surface of the micropores where the majority of the active sites exist, and is being enhanced by the favorable confinement effects due to size similarity, as discussed above. As the micropores are filled by the first reactant molecules, in combination with the decrease of the concentration and abundance of H 2 O 2 , the reaction takes place at the meso/macropore and external surface of the carbons at lower rates. This theory can be supported from the fact that the catalytic oxidative removal follows both the pseudo-first and the pseudo-second order kinetics (Table s5 †), although with a slightly better fitting indicative of an oxidation process. 3,30,70,75 The importance of the oxidation treatment and enrichment of the carbon's surface with acidic groups can be also concluded from the fact that in the case of pristine SX PLUS, 4,6- DMDBT was not completely eliminated (Fig. 7). Taking also into account that the adsorptive performance between the parent and the oxidized SX PLUS carbons is more or less similar, then it can be further suggested that the formed acidic groups, i.e. carboxyls and lactones, have a higher impact on the oxidation mechanism of 4,6-DMDBT than on just increasing the chemical interaction/sorption. ## Characterization of the carbons after reaction In order to examine the surface chemistry alterations and the possibility of strongly sorbed reaction products on the carbon catalysts' surface, FTIR spectra analysis was conducted (Fig. 8a-c). The comparative study of the parent and oxidized SX PLUS samples, as well as of the corresponding carbons after exposure to 4,6-DMDBT, shed light on the chemical interactions between 4,6 DMDBT and the catalysts' surface functional groups and on the related oxidation mechanism. In all spectra, the bands presented at about 1610-1630 cm −1 can be attributed to C-C stretching vibrations of the aromatic rings. The bands at 1100-1150 cm −1 can be attributed to CvO and O-H bonds of alcoholic, phenolic and carboxyl oxygen groups while the peak at about 1700-1725 cm −1 can be attributed to CvO stretching vibrations of the carboxyl groups. For the modified activated carbons, (SX PLUS-Nox and SX PLUS-Sox) prior to the exposure to 4,6-DMDBT in hexadecane, an increase in the intensity of the bands at 1720 and 1100 cm −1 was presented, representing carboxylic acids and lactones, due to their oxidation with HNO 3 or H 2 SO 4 acids. The bands at 1330-1370 cm −1 , which appeared on the spectra for the carbons after catalytic oxidation tests and after heating to 280 °C to remove the solvent, can be attributed to the sulfur compounds formed on the carbon surface, i.e. to sulfoxides, sulfones or sulfonic acid. The sulfones are specified by absorption at 1300-1350 cm −1 and 1120-1160 cm −1 . The bands at 1166, 1076 and 1020 cm −1 are characteristic of the molecular vibrations of C-S bonds. The increase in intensity of the bands at 1730 cm −1 may be due to hydrogen bond interactions of the 4,6-DMDBT oxidation products with carboxyl groups. The FTIR results indicated that 4,6-DMDBT was oxidized onto the carbon surface to the corresponding sulfones, sulfoxides or/and sulfonic acids as well as that these products were strongly adsorbed on the carbon's surface. Besides, EDX analysis, presented in Fig. 8d and e, supported further the FTIR results and proved the presence of sulfur on the surface of the carbocatalysts' after the catalytic experiment. The slight decrease of oxidation/removal performance observed in the reusability tests can be linked to the formation and strong retention of these sulfur compounds which were not totally removed/desorbed by washing from the active sites. ## UV-Vis and GC-MS analysis of the liquid products and extracts from the used catalysts At the end of the catalytic oxidation tests, the hexadecane solution was spectroscopically (UV-vis) analyzed, while the oxidation products accumulated on the carbons' surface were extracted with methanol for analysis. Methanol was selected because it is a polar solvent and could dissolve the formed sulfones and sulfoxides, compounds with enhanced polarity. The UV-Vis spectra of the liquid phase as well as the spectrum of the methanol extract after a catalytic oxidation experiment with SX PLUS-Nox as a catalyst are shown in Fig. s6. † In the spectra of the reaction product, the intensity decrement of the bands attributed to 4,6 DMDBT was obvious (Fig. s6a †), while in the spectrum of the methanol extract (Fig. s6b †), new peaks appeared corresponding to 4,6-DMDBT derived sulfoxides and/ or sulfones. 76 From the spectra it can be concluded that the oxidation products were accumulated/strongly sorbed on the carbon surface after the oxidation of 4,6-DMDBT. The liquid phases were also analyzed by GC-MS. From the results presented in Fig. 9a, the decrease of the peak's intensity corresponding to 4,6-DMDBT (at a retention time of 31.996 min) by the use of the parent activated carbon SX PLUS and the disappearance of this peak for SX PLUS-Nox and SX PLUS-Sox verify the UV-Vis analysis results and the obtained high or 100% removal of 4,6-DMDBT (Fig. 5 and 7). GC-MS analysis of the extracts of the used carbon with another polar solvent (acetonitrile) also showed the presence of 4,6-DMDBT products, such as sulfone and sulfoxide, as can be seen in Fig. 9b. ## Mechanistic insights From the above results it can be suggested that the aromatic ring of 4,6-DMDBT interacts with the carbon surface through the π-π stacking and/or through the donor acceptor mechanism, and these interactions are responsible for the high adsorptive capability of the studied activated carbons. The activity of the carbons as oxidation catalysts is associated with the concentration of oxygen surface functional groups, i.e. lactones or/and carbonyl groups. 50 These surface groups can also be responsible for the strong retention of the formed oxidation products, like sulfoxide and sulfones. The most crucial and still not clearly elucidated aspect is the catalytic role of the carbon surface. The major and most well-established pathway is that the above functional groups on the carbon's surface catalyze the decomposition of H 2 O 2 to free radicals such as • OH or • OOH, which are powerful oxidants for the oxidation of sulfur compounds to their corresponding sulfoxide and/or sulfones. These radicals can also be responsible for the formation of active molecular oxygen anions that also can act as an oxidant of 4,6-DMDBT. And all these reactions occur inside the micropores of carbon that act as "microreactors". In order to conclude if the catalytic capability can be linked to the formation of free radicals, methanol was used as a scavenger (Fig. 10). In the presence of methanol, the conversion/ removal percentage of 4,6-DMDBT was decreased for all carbons. In the case of SX PLUS-Sox, the decrement of the removal was 40% by the addition of 10% methanol per volume of the solvent. This removal efficiency is even lower from the measured adsorptive performance of this sample, a fact that suggests the blockage of both adsorption and catalytically active sites. This can be assigned to the water moieties, which remain adsorbed on the surface of the carbon by polar forces or/and hydrogen bonds, forming a film that hinders the interaction with 4,6-DMDBT (Fig. 11a). This is in perfect agreement with the above presented results when pure water was added, leading to a decrement of the oxidation/removal efficiency. The negative effect upon methanol addition may be attributed to the reaction of methanol with the free radicals, which are generated by the decomposition of hydrogen peroxide on the surface of the activated carbon, 50,77,78 Since hydrogen peroxide can undergo a fast decomposition in the presence of activated carbon, the most crucial aspect is the simultaneous presence of 4,6-DMDBT and H 2 O 2 inside the pores in order for all the consecutive reactions to occur (Fig. 11a). The presence of water in a high amount can decrease the reaction kinetics by blocking/occupying the adsorption/reaction sites. The oxidation of 4,6-DMDBT undergoes through a redox cycle of carbon with the participation of superoxide anionic radicals and H 2 O 2 (Fig. 11b). Analogous redox cycle based catalytic oxidation pathways were reported for metal oxides. 79 In general, peroxide has a triple role. The first one is to activate carbon's surface (Fig. 11c), the second to act as a pool of molecular oxygen (Fig. 11d), and the third to participate in the formation of superoxide anionic radicals (Fig. 11e). In more recent publications, free radical species ( • OH/ • OOH) have been postulated as the main intermediates in the reaction mechanism, whose formation would take place through an electron-transfer reaction similar to the Fenton mechanism, with AC and AC + as the reduced and oxidized catalyst states. The recombination of free radical species ( • OH/ • OOH) in the liquid phase and/or onto the activated carbon surface will produce water and oxygen according to reactions in Fig. 11. When 4,6-DMDBT is adsorbed on the surface of activated carbon, in which also H 2 O 2 is decomposed into radicals H + and HOO t , the latter reacts with O 2 to superoxide, which further oxidizes 4,6-DMDBT to sulfoxides and/or sulfones. It is also feasible that the free • OH or • OOH radicals attack directly and oxidize 4,6-DMDBT. ## Conclusions In the current work, we presented the potential use of activated carbons not only as adsorbents of sulfur compounds but also as metal-free oxidation catalysts for the desulfurization of fuels via oxidation of 4,6-DMDBT to the corresponding sulfones and sulfoxides in a reusable manner and under ambient conditions. The comparison of five commercial porous carbons revealed that the removal capability depends predomi-nately on the density of acidic surface functional groups, while the adsorption is governed by the formation of donor-acceptor complexes between the adsorbent and the adsorbate. Further chemical modification of the carbon's surface via treatment/ oxidation with HNO 3 or H 2 SO 4 was also investigated. In both cases, the basic surface groups were decreased, the phenolics were eliminated, while the lactones were increased, and carboxyl groups were formed. The modification with the former acid had a more pronounced effect on the formation of acidic groups and induced faster catalytic oxidation/removal kinetics, while treatment with the latter acid led to a significant decrease of the basic sites and to the highest adsorptive capability. Although the oxidation of carbons had a relatively moderate negative effect on porosity characteristics, the oxidized counterparts showed a 100% desulfurization capability, a value being 33% higher compared to the performance of the parent non-modified commercial carbon, confirming the predominant role of the surface chemistry. In addition, it was shown that the abundance of micropores with a size of ≤1 nm is beneficial for the adsorptive/catalytic reactivity due to confinement effects considering the similar size of 4,6-DMDBT. The reaction of 4,6-DMDBT in the presence of H 2 O 2 led to the formation of the corresponding sulfoxide and sulfone, as a result of the formation of superoxide and free • OH and • OOH radicals upon the decomposition of H 2 O 2 on the carbocatalyst surface. These results can boost the research activities and exploitation of activated porous carbons, without the use of any external metals with redox activity, as highly efficient green catalysts in oxidative deep desulfurization of fuels or other related processes.

chemsum

{"title": "Catalytic oxidative desulfurization of a 4,6-DMDBT containing model fuel by metal-free activated carbons: the key role of surface chemistry", "journal": "Royal Society of Chemistry (RSC)"}

a_fundamental_catalytic_difference_between_zinc_and_manganese_dependent_enzymes_revealed_in_a_bacter

## Abstract: The catalytic mechanism of the cyclic amidohydrolase isatin hydrolase depends on a catalytically active manganese in the substrate-binding pocket. The Mn 2+ ion is bound by a motif also present in other metal dependent hydrolases like the bacterial kynurenine formamidase. The crystal structures of the isatin hydrolases from Labrenzia aggregata and Ralstonia solanacearum combined with activity assays allow for the identification of key determinants specific for the reaction mechanism. Active site residues central to the hydrolytic mechanism include a novel catalytic triad Asp-His-His supported by structural comparison and hybrid quantum mechanics/classical mechanics simulations. A hydrolytic mechanism for a Mn 2+ dependent amidohydrolases that disfavour Zn 2+ as the primary catalytically active site metal proposed here is supported by these likely cases of convergent evolution. The work illustrates a fundamental difference in the substrate-binding mode between Mn 2+ dependent isatin hydrolase like enzymes in comparison with the vast number of Zn 2+ dependent enzymes.The catalytic activity of metallohydrolase enzymes strongly depends on the identity of the metal and on the nature of its binding site. Zn 2+ bound hydrolytic enzymes, in particular metallo-β-lactamases 1 , probably constitute the most extensively studied cases for such a group. Depending on the enzymatic class, hydrolytic enzymes with both a mononuclear or dinuclear Zn 2+ sites have been identified as widely distributed throughout all kingdoms of life 2 . The catalytic mechanism for the mononuclear and dinuclear Zn 2+ sites have been debated for years [3][4][5] , and only recently the collection of accumulated knowledge allowed the proposition of a plausible mechanism [6][7][8][9] . Despite Zn 2+ being the most widely distributed hydrolytic cation, it has been shown that Zn 2+ also possesses an inhibitory potential for certain metallohydrolases as, for example, in carboxypeptidase A 10 .The specialized metallo-enzyme amidohydrolase superfamily (AHS) is able to carry out a diverse range of chemical reactions including the degradation of metabolic precursors like pyrimidine (dihydropyrimidinases) and hydantoin, (hydantoinases) 11 . The AHS members are often categorised as both Zn 2+ and Mn 2+ dependent hydrolases. However, amidohydrolases involved with purines degradation in both bacteria and plants are mainly Mn 2+ dependent allantoinases 12,13 , even though Zn 2+ dependent allantoinase activity has been reported 14 .The compound isatin is a metabolic precursors and was originally identified as an intermediate product of the indole-3-acetic acid (IAA) degradation pathway in Bradyrhizobium diazoefficiens 15 . Hydrolytic activity of isatin hydrolase (IH) was first demonstrated in various symbiotic rhizobial species 16 . Recent bioinformatics studies indicates that IH are found widespread in bacteria including pathogens specific to the human gut and plants, e.g., B. enterica and R. solanacearum 17 . Isatin hydrolase A from Labrenzia aggregata (LaIHA) and R. solanacearum (RsIHA) share 59% sequence identity while the two putative orthologues LaIHA and LaIHB in L. aggregata share 51% sequence identity. Both LaIHA and LaIHB contain the central metal binding motif HxG[T/A]HxDxPxH in each protomer of the catalytically active dimer. Structural characterization of LaIHB revealed a novel fold, which later was also attributed to other amidohydrolases such as bacterial kynurenine formamidase B (KynB) (EC 3.5.1.9) 17 . The fold is generally described as an α/β hydrolase with a central scaffold resembling the swivelling β/α/β fold 18 , while the majority of AHS members contains a monomeric (β/α) 8 -TIM like-barrel structural fold 19 . A notable difference between LaIHB and kynurenine formamidase from Bacillus anthracis (BaKynB) is the presence of a mononuclear manganese-binding site in LaIHB, while the structure of KynB accommodates a bi-nuclear zinc site 20 . LaIHB shows a strict specificity for Mn 2+ over Zn 2+ , as reported by kinetic measurements 17 , and rationalised by computer simulations 21 . Here we present two crystal structures of LaIHA, one bound to the hydrolytic product isatinate, and one bound to benzyl benzoate, as well as the apo structure of homologous RsIHA. Together with results from enzyme kinetic analysis, the resolved structures are used to describe the molecular basis for the Mn 2+ -dependent mechanism of hydrolysis, showing that IHA is a true orthologue of IHB. For the first time, the mechanism of a Mn 2+ dependent hydrolysis reaction is characterised in detail by computer simulations using multiscale quantum mechanics/molecular mechanics (QM/MM) 22 and ab initio molecular dynamics (MD) simulations 23 . The simulations were crucial to unravelling the essential catalytic role of the conserved residues His79 and His207 for the formation of the product. Based on the identification of mechanistically important residues presented in this work, including the Mn 2+ binding residues in the LaIHA active site, we further present evidence that the Mn 2+ dependent allantoate amidohydrolase (AAH) (E.C. 3.5.3.9) displays a striking similarity in the active site geometry despite the differences in the overall protein fold. As this likely represents a case of convergent evolution, we propose that the generalised catalytic mechanism described here for the IHs may also apply to the AAH family and other similar Mn 2+ dependent hydrolases. ## Materials and Methods Cloning, expression and purification of LaIHA and RsIHA. Chemicals were purchased from Sigma-Aldrich (Norway) unless stated otherwise. The open reading frame encoding LaIHA (UniProtKB: A0P0F0) was amplified by PCR from a boiled colony of L. aggregata IAM12614. DNA fragments were isolated and digested with BamHI and EcoRI and cloned into the T7-RNA polymerase dependent E. coli expression plasmid pT7H6 24 . The 6xHis containing LaIHA was expressed in E. coli BL21 AI cells (Invitrogen) in 2xTYE-medium for 4 hours at 37 °C. Cells were harvested by centrifugation, re-suspended, and lysed by sonication in a buffer containing 50 mM Tris-HCl pH 8.0 and 0.5 M NaCl supplemented with a protease inhibitor tablet (Roche cOmplete, EDTA-free). Insoluble material was removed by centrifugation (4000 g, 10 min.). The soluble protein extract was batch-adsorbed onto 25 mL Ni-NTA agarose resin (Qiagen) per litre of original culture and loaded into liquid chromatography columns. The protein loaded Ni-NTA columns were washed with >20 column volumes (CV) of equilibration buffer (50 mM Tris-HCl pH 8.0, 500 mM NaCl, 50 mM imidazole). The protein was eluted from the resin with elution buffer (50 mM Tris-HCl pH 8.0, 500 mM NaCl, 10 mM Na 2 EDTA). Fractions containing the recombinant proteins were pooled, and the buffer changed into a low salt buffer (10 mM Tris-HCl pH 8.0, 100 mM NaCl, and 1 mM Na 2 EDTA) on a Sephadex G-25 column (GE Healthcare). Before crystallisation and kinetic experiments, LaIHA was further treated with 10 mM EDTA and dialysed (500 fold dilution) into a stability buffer (5 mM bis-tris pH 7.0, 100 mM NaCl and 1 mM DTT). The protein solutions were centrifuged at 180,000 g for 10 minutes at 4 °C, and the supernatant could hereafter be stored in a stable condition at 4 °C for at least 12 months. The RsIHA (UniProtKB: Q8XYC3) was ordered from Genscript and inserted into the expression vector pET-M11. The six-histidine containing construct was expressed in E. coli BL21 gold cells, in LB-medium (containing 50 µg/mL kanamycin) overnight at 18 °C. Cells were harvested by centrifugation, re-suspended, and lysed by a bead-beater (Biospec) in a lysis buffer (50 mM Tris-HCl pH 8.0 and 0.1 M NaCl), protease inhibitor tablet (Roche cOmplete, EDTA-free). Insoluble material was removed by centrifugation, and the lysate was supplemented with 20 mM imidazole before loaded onto a pre-equilibrated HisTrap HP (GE). The column was washed with 10 CV with equilibration buffer (50 mM Tris-HCl pH 8.0, 0.1 M NaCl, and 20 mM imidazole) and eluted with elution buffer (50 mM Tris-HCl pH 8.0, 0.1 M NaCl, and 300 mM imidazole). The pooled fractions were supplemented with a molar 1:50 ratio of tobacco etch virus (TEV) protease and dialysed overnight against dialysis buffer (50 mM Tris-HCl pH 8.0, 100 mM NaCl). The RsIHA was again loaded to the HisTrap HP column and the flow-through was collected. RsIHA was concentrated to 20 mg/mL and loaded onto a Superdex 200 10/300 GL size exclusion column (GE) pre-equilibrated in SEC buffer (50 mM Tris-HCl pH 8.0, 100 mM NaCl). Pure RsIHA were pooled, concentrated, and stored at −80 °C. RsIHA was treated like LaIHA before crystallisation and kinetic experiments. Crystallization of LaIHA and RsIHA ligand complexes. The pre-treatment of LaIHA and RsIHA with an excess of EDTA and dialysis was mainly carried out to remove any potential trace of metals, leaving an estimated EDTA concentration at 0.02 mM. Prior to crystallization 1 mM MnCl 2 was added. LaIHA and RsIHA were crystallized at 22 °C using a protein concentration of 15 mg/mL in 5 mM Bis-Tris pH 7.0, 100 mM NaCl, and 1.0 mM MnCl 2 . LaIHA was crystallised by vapour diffusion using sitting drops 1 + 1 μL (LaIHA: Reservoir) over 100 μL reservoir solution. The best diffracting crystals were obtained in 20% glycerol and 20% PEG 1500. The LaIHA:isatinate structure was obtained by soaking the crystals for approximately 1 min with 500 µM isatin before flash-freezing the crystals in liquid nitrogen (LN 2 ). This condition also served as cryo-protection. RsIHA was also crystallised with vapour diffusion using hanging drops, 1 + 1 μL (RsIHA: Reservoir) over 400 μL reservoir solution sealed with immersion oil (56822, Sigma). The best diffracting crystals of RsIHA grew in 1 M succinate, 100 mM HEPES pH 6.5. Cryo-protection was achieved by adding 1 μL of 60% w/v PEG400 suspended in reservoir solution to the drop. The crystals were mounted and flash-cooled immediately after. Crystals of LaIHA:benzyl benzoate was obtained in 28% PEG 1500 and 1 mM MnCl 2 with a hanging drop setup using 250 μL reservoir and 1 + 1 μL drops. With a protein concentration of 20 mg/mL used. The hanging drop wells were sealed with immersion oil. Crystals appeared within one week. Cryo-protection was obtained by adding glycerol to the reservoir solution to a final concentration of ~12%, which after 12 hours would allow additional vapor diffusion to take place before mounting and flash-cooling in LN 2 ## 25 . Data collection and structure determination. All datasets were collected at 100 K. For LaIHA:Benzyl benzoate a dataset extending to 1.50 resolution was collected at a wavelength of 1.0004 . The space group was determined to be P1, and a collection strategy was calculated by iMOSFLM 26 . The data were processed with XDS 27 . A test set of 5% was used for R free calculation 28 . Phasing was performed by molecular replacement (MR) using LaIHB (PDB ID: 4J0N) as a search model for Phaser 29 . Initial model building and refinement were performed in Phenix 30 . The final model was built using Coot 31 with small ligand models and restraints produced using eLBOW 30 . For LaIHA:Isatinate, a diffraction dataset was collected at beamline P13 at PETRA III (Hamburg, Germany) 32 . The space group was determined and would only allow a P1 lattice. A complete dataset was collected at 1.79 resolution with and oscillation range of Ω = 0.01° and processed with XDS. The structure was solved with MR using a LaIHA: Benzyl benzoate monomer as the search model and performed with Phaser in Phenix. Diffraction data on RsIHA was collected at beamline ID30B at ESRF (Grenoble, France). The space group was determined and allowed P 3 2 2 1, a complete dataset with at Ω = 0.05° over 126° was collected. The data extended to 2.65 and was processed with XDS. The structure was solved with MR using LaIHA:Benzyl benzoate monomer as the search model in Phenix. Model building and refinement was performed using coot 33 and phenix. refine within the phenix package 30 . NCS restraints were applied throughout the refinement, however released in the final round (Number of molecules in the asymmetric of each crystals structure have been added in the Supplementary Table S1) 34 . All structure validations were performed using MolProbity 30 . Structure analysis and figure preparation were done using PyMOL Molecular Graphics System, Version 1.5.0.3, Schrödinger, LLC. The finalised model and structure factors were deposited to Protein Data Bank (PDB) and given the PDB accession codes (LaIHA:Benzyl benzoate PDB ID 5NNA; RsIHA PDB ID 5NMP; LaIHA:isatinate PDB ID 5NNB). ## Enzymatic assay. The measurement of the activity, and its metal dependence, were performed as described by Bjerregaard-Andersen et al. 17 . The method is based on an increased absorption of the product isatinate at 368 nm. This can be followed using a standard spectrophotometer. Measurements were performed using a JASCO V-630 spectrophotometer with a cuvette light path of 1.0 cm. An extinction coefficient of 4.5 × 10 3 cm −1 mol −1 L was used for isatinate. The data were fitted with the Michaelis-Menten equation assuming one binding site using the software GraphPad Prism 6. ## Sequence analysis. Structural homologues of LaIHA and E. coli allantoate amidohydrolase (EcAAH) were identified using the DaliLite v. 3 35 , followed by inspection in PyMoL. The sequences of the identified homologues were then imported in Jalview 36 using the fetch from PDB option and combined with sequences of LaIHA, RsIHA, and Rhodococcus rhodochrous HpoH (RrHpoH). Sequences with the following UniProt accession codes (in parentheses) were used: LaIHA (A0P0F0), LaIHB (A0NLY7), RsIHA (Q8XYC3), EcKynB (B4E9I9), PaKynB (Q9I234), BaKynB (Q81PP9), BsAHD (P84132), MjAHD (Q58193), RrHpoH (B5MAD9). For AAH alignment, the following accession codes were used: EcAAH (P77425), AtAAH (Q8VXY9), BvHYD (A4JQA0), BmHYD (A0A0H3KRF1), BcHYD (B4EHA1), GeRAC (Q53389), SkSYN (Q96W94). Sequence alignments were performed using the Clustal algorithm with default settings in Jalview. Phylogenetic trees were calculated in Jalview as average distance using identity percentage and exported to TreeDyn 198.3 37 for visualisation. The sequences were included in alignment as supplied by the PDB. ## Molecular modelling. The system consisting of LaIHA, isatin, 47331 water molecules, and 19 Na + ions was placed in a 105 × 126 × 118 3 simulation box. The initial geometry of the protein was taken from the X-ray structure of LaIHA:isatinate resolved by us and described in this text. Isatin was introduced in the system by molecular replacement, fitting the aromatic moiety of isatinate. The Mn 2+ ion, isatin, the catalytic water, the side chains of the ligating amino acids (His69, His73, Asp75, Gln219), and of the neighbouring His79, His207, Asp193 residues were treated at a quantum-mechanical level (QM) using density functional theory with the Becke and Lee-Yang-Parr (BLYP) approximations for the exchange-correlation functional 38,39 . The Kohn-Sham orbitals for the valence electrons were expanded over a DZVP Gaussian basis set 40 and an auxiliary plane-wave basis set with a cut-off of 240 Ry. The core electrons were integrated out using Goedecker-Teter-Hutter pseudopotentials 41 . The remainder of the system was described at the molecular-mechanics (MM) level using the Amberff14SB force field 42 . In a first step, the system was equilibrated by full classical molecular dynamics (MD) simulations in the NpT ensemble (T = 300 K, p = 1 atm), constraining the QM region to its initial positions, and using a time-step of 1.5 fs. Thereafter, QM/MM Born-Oppenheimer MD simulations were performed at 300 K within the NVT ensemble using the Nosé-Hoover thermostat and a time-step of 0.25 fs. The simulations were run using the CP2K package (https://www.cp2k.org/) . The mechanism of hydrolysis was sampled by coupling QM/MM MD to metadynamics simulations 49,50 using collective variables defined as differences between coordination numbers (Visualized in Fig. S1, in the supporting information): where CN is the coordination number between two atoms. Each CN is defined as: where d ij is the distance between atoms i, j, and d 0 , p, and q are free parameters. In our simulations, d 0 was set to 1.6 for CV1, and 1.1 for CV2 and CV3. Values of p = 12 and q = 14 were used for all the three CVs. The d0, p, q values where chosen so that the profile of the sigmoidal function (equation 1) did not change significantly during a standard thermal oscillation of the bond, while it signalled the formation or breaking of a chemical bond. Gaussian hills with a height of 2 kcal/mol were spanned every 100 step (i.e., every 25 fs). A Gaussian spread of 0.15 was used for all three collective variables. The free energy profile for the reaction was obtained as the sum of the Gaussians added during the whole metadynamics simulation. The three CVs used in this study were chosen to take into account both the nucleophilic addition/elimination at the carbonyl group and the proton rearrangements necessary to yield the product. Runs using any combination of only two of these CVs did not produce any reactive profile, indicating that all the three CVs are necessary for the determination of the reaction coordinate. With this setup, the system reacted after 1.1 ps of metadynamics simulations, the run was continued until a total time of 1.85 ps to reach convergence in the free energy profile. We assumed convergence of the profile after observation of a second re-crossing of the barrier (Supplementary Fig. S2). ## Results IHA contains a mononuclear manganese binding site involving an additionally conserved glutamine. The structures of LaIHA:benzyl benzoate and LaIHA:isatinate were determined at 1.50 and 1.79 resolution, respectively, while the one of RsIHA was determined at 2.65 resolution. Data collection and model refinement statistics are collected as described in Supplementary Table S1. Both LaIHA structures are highly similar to the one of LaIHB with a root mean square deviation (rmsd) value of ~0.7 computed on all main chain atoms. Both LaIHB and LaIHA are dimers with two exchanged regions -a small N-terminal α-helical exchange and a larger β-hairpin exchange (Fig. 1a). A hallmark of this fold is the contribution of two conserved residues of one monomer in the formation of the substrate-binding pocket located in the other monomer. For the IH's these two residues are two tryptophans (Fig. 1b). The highly conserved Mn 2+ binding site is located at the bottom of the substrate-binding pocket. The Mn 2+ is found in an octahedral complex coordinated by His69, His73, Asp75, and two water molecules (Fig. 1c). Asp75 forms a bidentate contact in LaIHA:Benzyl benzoate and a mono-dentate electrostatic contact in the LaIHA:isatinate structures. The mono or bi-dentate coordination results in two or one water molecules as additional ligands, respectively, preserving the octahedral geometry. The side chain carbonyl of Gln219 in LaIHA (Ala224 in LaIHB) is found in alternative conformations, one of which coordinates Mn 2+ . The non-binding conformation allows additional water to approach to Mn 2+ (W1025 A/B , Fig. 1c). Sequence analysis reveals that Gln219 is part of a motif, conserved as either GLAS (e.g. in LaIHB) or GLQC (e.g. in LaIHA) (Figs 1 and 2a). The short residues motifs AS and QC are conserved as sequence pairs, which could indicate a functional dependency. Interestingly, the residues S or C are found to be the key residue involved in regulating the proton flow through the water channel. A mutation of the serine residue to a cysteine to form a GLAC motif caused a gain in activity in LaIHB 17 . The GLQC/GLAS sequence, combined with the conserved manganese binding site appears to be a signature motif for isatin hydrolase activity. In KynB, the comparable position holds a conserved I(L/I)E motif, which is also found in the two uncharacterized structures of amidohydrolases (AHD) from Bacillus stearothermophilus (BsAHD) and Methanococcus jannaschii (MjAHD) (Fig. 2a, Full sequence alignment shown in Supplementary Fig. S3). The motif is absent in RrHpoH, which has confirmed cyclase activity. Whether the I(L/I) E motif and perhaps, in particular, the glutamate is indicative of activity on linear amides or a binuclear metal site remains uncertain. Activity measurements confirm that both LaIHA and RsIHA have isatin hydrolase activity (Fig. 1d,f). The determination of the catalytic parameters assuming Michaelis-Menten kinetics yields K m values for LaIHA and RsIHA of 16 μM and 10 μM, respectively. Structural superposition revealed identical metal binding sites in the two homologues. Thus, a metal dependence analysis was only performed on LaIHA. LaIHA shows the highest activity in the presence of Mn 2+ confirming that this enzyme is Mn 2+ dependent. However, the relative activity of 15.1 ± 2.2% is also observed in the presence of Cd 2+ . Notably, LaIHA could not be activated in the presence of both Zn 2+ and Cu 2+ (Fig. 1e). ## QM/MM calculations reveal the importance of an Asp-His dyad in the proton abstraction. In LaIHA, both the carbonyl group of the substrate and the catalytic water coordinate to the Mn 2+ ion. Like in other hydrolytic enzymes, the nucleophilic water is activated by a pre-organized base, which can efficiently extract a proton to form hydroxide (OH − ). In LaIHA, His207 acts as such base. The positively charged protonated His207-H + is stabilised by the presence of a hydrogen-bonded partner, i.e., Asp193. The water-His207-Asp193 H-bond network closely resembles the Ser-His-Asp triad common to serine proteases. His79, also H-bonded to the catalytic water, completes a complex His79-water-His207-Asp193 proton-shuttle system (Fig. 2b,d). Water deprotonation in LaIHA is extremely efficient and occurs spontaneously during the QM/MM simulations at room temperature 21 . Activation of the catalytic water is favoured by its isolation from the water channel that The manganese is found in octahedral coordination similarly to that described in 17 . Gln219 resides in a double conformation and only partially coordinates to the manganese. Also, W1025 is found in a double conformation (denoted A and B in Fig. 1). Note that Asp75 is coordinating bidentate in LaIHA:isatinate while monodentate in LaIHA: benzyl benzoate. connects the substrate cavity to the exterior. Confinement of the catalytic water has been reported as an important structural requirement in other hydrolytic (metallo)-enzymes . Figure 3 depicts the molecular details of the hydrolytic reaction. The nucleophilic attack of OH − to the carbonyl of isatin leads to the formation of a tetrahedral intermediate. Our model predicts an activation barrier of a few (i.e., less than 6) kcal mol −1 . Thus, as in other metallohydrolases (i.e., zinc-β-lactamases), this step is not rate-limiting for the reaction. The tetrahedral intermediate is only metastable and it can easily recombine into the reactant state. In fact, the following separation of the scissile C-N bond constitutes the rate-limiting step of the reaction with a barrier height of ~12 kcal mol −1 . Our data report that C-N separation occurs through a dissociative pathway. The transition state is stabilised by formation of two hydrogen-bonds (H-bonds), the first between the nucleophilic OH − and the Nδ atom of His79, and the second between the isatin N atom and the Nε-H moiety of His207 (Fig. 3, Supplementary movie). The final product formation of isatinate is accompanied by synchronous proton exchange in both of these two H-bonds. Our simulations indicate that the proton transfers occur after the transition state has been reached and the C-N bond is broken. Kinetic measurements in a deuterated environment did not reveal any observable kinetic isotope effect on the reaction, providing an additional indication that proton transfers are not involved in the rate-limiting step of the enzymatic process (data not shown). The preference of a dissociative pathway over an associative or concerted mechanism may be a consequence of the fact that the N-atom in isatin acts as a poor base. In general, the π conjugation between the lone pair of N and the aromatic ring present in aryl-amines decreases the pKa value by few pH units. As a result, we observe that formation of a stable H-bond with His207, and consequently the proton transfer, occurs only upon a major elongation of the isatin C-N bond, which is associated to increased localization of the electron density at the N atom. Additional stabilization of the moiety into the product is obtained by the deprotonation of the nucleophile OH − group by His79 to produce a carboxylate group chelating the Mn 2+ as observed in the crystal structure. The isatinate conformer produced by hydrolysis is not the most stable one. In fact, the α-carbonyl can isomerise via a bicycle motion, similar to that described in excited state isomerisation of retinal in rhodopsin 56 . This isomerised product found in our simulations is in agreement with the binding geometry of the product, isatinate, described in the crystal structure (Fig. 1b). In particular, the final binding state of isatinate shows bidentate coordination to the manganese ion, and similar stacking interactions with the indole rings of Trp59, Trp61, Trp80 and Tyr204 (Fig. 1b). The strong exothermic isomerisation step (−10 kcal mol −1 ) yields a structure of isatinate with a strong intra-molecular H-bond between the amino and the α-carbonyl groups. This possible intermediate most likely constitutes the driving force of the reaction and traps the isatinate as a ligand to Mn 2+ in its product state before it is replaced by water and released from the binding pocket (Fig. 3). ## Figure 3. Reaction mechanism describing the hydrolysis of isatin. The reaction is visualised from the initial nucleophilic attack of OH − to the carbonyl of isatin until the final substrate isatinate is formed. The final step involves isatinate isomerization, promoted by the strong intra-molecular H-bond between its amino and α-carbonyl groups, as well as the reformation of the hydrogen bond between Asp193 and His207. Free energy differences were estimated from metadynamics simulations. Isatin hydrolase and kynurenine formamidase active-site conservation indicates a similar mechanism. The recently determined crystal structure of kynurenine formamidase from Bacillus anthracis (BaKynB) contains a binuclear Zn 2+ site and thus is part of the EC 3.5 class that has a similar overall protein fold when compared with LaIHA (sequence identity 24%). Figure 2d shows the superposition of the metal binding sites for LaIHA:isatinate and BaKynB (PDB ID: 4COG). The structural architecture of the metal binding sites is highly conserved between both enzymes (Fig. 2d). The most striking difference is the position of Gln219 in LaIHA, which is occupied by glutamate (Glu172) in BaKynB. This residue is conserved for all verified KynB enzymes. In BaKynB, the His60 (equivalent to His79 in LaIHA) is proposed to be necessary for the hydrolysis reaction to take place, by accepting a proton from the bi-coordinated activated water. In the mechanism for the isatin hydrolysis presented here, both His79 and His207 are involved in the catalytic mechanism (Supplementary Movie M1). In BaKynB, the Zn 2+ in position (I) Zn I 2+ is octahedrally coordinated and bound in an equivalent position as Mn 2+ in LaIHA. In a homologous structure of KynB from Burkholderia cenopacia (BcKynB), this position is occupied by Cd 2+ . The Zn 2+ in position (II), Zn II 2+ is also found in an octahedral conformation. Convergent evolution of the active-site geometry in LaIHA and EcAAH. Amidohydrolases with known manganese dependency where identified through literature searches and examined manually for structural similarity. This approach revealed two candidates that, upon close inspection, shared metal binding sites residues and residues identified as catalytically important. The allantoate amidohydrolases (AAHs) are part of the purine degradation pathway and hydrolyse the allantoate to ureidoglycolate, carbon dioxide, and ammonium in the process, the allantoin pathway is found in both plants 13 and bacteria 57 (Fig. 2a,c). The identified structures were manually superposed with the active site of representatives from the AAH's, (Fig. 2e). The structure for E. coli AAH (EcAAH, PDB ID: 4PXD) was used as input for the DALI server 35 . The found hits were visually inspected for structural homology and included in alignment to confirm residue conservation (Fig. 2a). ## Discussion The Baltic sea bacteria Labrenzia aggregata genome includes two homologous open reading frames for isatin hydrolases A and B that share 51% sequence identity (SI). The more distant homolog from Ralstonia solanacearum (RsIHA) shares 59% SI with LaIHA and 51% SI with LaIHB. Unlike the whole sequence, the active site and substrate binding pocket are highly conserved among these proteins. Even though the key substrate binding pocket residues originally identified in the structure of LaIHB -Phe63, Trp65 and Trp84 and Phe209 17 do not share equivalent positions in the sequences of LaIHA or RsIHA, their aromatic side chains do adopt an identical structural conformation in the respective binding pockets, thus indicating functional conservation (Fig. 1b). Compared to LaIHB, the Mn 2+ binding sites in both RsIHA and LaIHA include Gln219, also responsible for increased metal specificity 21 . The crystal structure of BaKynB from Bacillus anthracis (PDB ID: 4COG) presents a glutamate (Glu173, Fig. 2d) adopting an equivalent conformation as that of Gln219 in LaIHA. In conjunction with the conserved His161 and Asp56 (BaKynB numbering), it forms a secondary Zn 2+ site (Zn II ) binding site in the presented binuclear metal binding (Fig. 2d). The remaining metal-coordinating residues are identical to both RsIHA and LaIHA. It cannot be completely ruled out that, in certain conditions, Gln219 would allow the formation of a secondary metal-binding site in LaIHA. This is however unlikely, due to the different Lewis-acid/ base properties of an amide group, compared to those of a carboxylate. The structure of LaIHB, which features an alanine residue in the position of Gln219, does not have this option and thus is expected to be strictly mononuclear, with the highest hydrolytic activity measured in the presence of Mn 2+ , as also observed for LaIHA presented in this manuscript. In the case of the amidohydrolase diaminopimelate desuccinylase (dapE), it was found that the compound L-captopril only binds to the Zn 2+ bound form of the enzyme and not to the physiologically relevant Mn 2+ bound form, thus stressing the importance of identifying the relevant metal 58 . Different metallo-β-lactamases can bind and be catalytically active in the presence of both one and two Zn 2+ metals in the binding site. In both cases, the reaction mechanisms were characterized by computational modelling in the past (1 Zinc: B1 BcII from Bacillus cereus and B2 CphA from Aeromonas hydrophila, 2 Zinc: B1 CcrA from Bacteroides fragilis) , the mono and binuclear reaction scheme for Zn 2+ binding beta-lactamases has recently been revised in Lisa MN et al. 6 . The mechanism of hydrolysis in LaIHA shares similarities with those found in single-Zn 2+ enzymes. In particular, the reaction proceeds through a two-step mechanism, where the nucleophilic attack on the amide carbonyl of isatin is followed by the breaking of the C-N bond. On the contrary, in CcrA from B. fragilis, the reaction follows a concerted mechanism where the nucleophilic attack of the hydroxyl takes place simultaneously with the opening of the β-lactam ring. Common to all these enzymes is that the rate-limiting second step of the reaction is constituted by the dissociation of the C-N bond, and that the pathway to the product is facilitated by proton transfer on the N atom that usually occurs at the transition state or immediately after. The proton-shuttling pathways are nonetheless different in the Zn-bound enzymes and IH. The lone pair of the amidic nitrogen of isatin is π-conjugated with both the amidic carbonyl and the aromatic phenyl ring present in the molecular structure. As a consequence of the resulting electronic delocalisation, its basicity remains low even after the initial attack of the OH − and the subsequent formation of the tetrahedral intermediate. In fact, the evolution from the TS to the product requires both the pre-organised His207-Asp193 dyad shuttling one proton on the N atom and the stabilisation of C-N heterolytic breakage by a secondary proton transfer from the carboxylic group to His79. As a consequence, again differently from metallo-β-lactamases, the product formation in LaIHA leaves the protein in a higher protonation state than the initial complex. The deprotonation of the active site after the release of the product likely occurs through the water channel originally described for LaIHB 17 . Our in silico studies yield further support to the necessity for proton removal, in order to recover the catalyst, and allow further enzymatic cycles. As the calculations produced a slightly endothermic profile (+2 kcal), the excess energy released by deprotonation of the protein may also constitute the driving force for the enzymatic process. Though, we stress that due to high computational costs, the free energy profile was obtained with a resolution that is of the same order as the reaction free energy balance. The sequence alignment and structural superposition confirm that not only the metal coordinating residues (Fig. 2a-c, dots), but also the residues involved in the proton transfer (Fig. 2a-c, asterisk) are highly conserved. Only the position corresponding to LaIHA residue Asp193 (Fig. 2a) alternates between aspartate and glutamate, thus showing anyway functional conservation. The conservation suggests functional importance and, supported by the structural superposition and mechanistic elucidation, points to a common catalytic role. This is particularly interesting in the light of the sequence variation within the IH-like and AAH-like folds as illustrated in Fig. 2b,c,e. Here, the phylogenetic tree of proteins belonging to the respective folds indicates a large sequence variation, perhaps best exemplified by the significant divergence of EcAAH compared to Arabidopsis thaliana AAH (AtAAH), i.e., prokaryotic and eukaryotic. However, both the overall fold, displaying domain-swapped dimers, and the active site geometry are highly conserved (Fig. 2c-e). It remains to be established whether the catalytic mechanism exemplified in LaIHA by our structural and in silico studies is specific to the IH-like enzymes, or could represent a broader case of convergent evolution across a range of AHS metallo-hydrolases. In the case of the AAH homologues, the Mn 2+ binding site is structurally completely conserved, including the binding residues in the Mn I site (His69, His73 and Asp75 for LaIHA), and the proton transfers triad (His79, His203 and Asp193) in LaIHA compared to Glu127, His382 and His226 in EcAAH, with the peculiar difference that in EcAAH the Mn 2+ occupies site II (See comparison in Fig. 2c,e). The IH, KynB and AAH families are not the only ones to have divergent metal site constellation, despite similar metal binding residues, the ureidoglycolate amidohydrolase from Arabidopsis thaliana (AtUAH) has a binuclear Mn 2+ site 62 (PDB ID 4PXB), while the putative hydantoinases (PDB ID 5I4M, 5THW and 4WJB) and β-alanine synthase from Saccharomyces kluyveri 63 (PDB ID 2VL1) have binuclear Zn 2+ sites, the L-N-carbamoylase from Geobacillus stearothermophilus has a mononuclear Co 2+ site 64 (PDB ID 3N5F). The model for hydrolysis presented here would only allow a Mn 2+ in site I, and a Zn 2+ would act as an inhibitor. In particular, according to previous simulations, Zn 2+ is not able to bind both isatin and the water, disrupting the preorganization of the active site and preventing the initial nucleophilic attack 21 . Thus, it remains an open question whether the Mn 2+ in site II also act as an inhibitor or merely adopt an alternative position in the structure of EcAAH due to the crystal conditions. It is a reoccurring problem for metal dependent hydrolases that the majority of the information is collected from in vitro experiments and that conflicting conclusions are reached when it comes to metal specificity. As in the case of KynB were a binuclear Zn 2+ is described in the structure 20 and Zn 2+ dependent activity is presented, however it still remains an open question whether this is the physiological metal, as in other systems such as Ralstonia metallidurans a 15 to 20-fold activation of kynurenine formamidase was detected after treatment with Mn 2+ 65 . The major difference between the Michaelis complexes in LaIHA/B and zinc-β-lactamases is constituted by the binding mode of the substrate. β-lactam substrates, bind in the proximity of the metal(s), but their amide moiety do not participate directly to their coordination sphere(s) as is the case for the reacting cyclic amide in the IH. It is carboxylate groups of the carpapenem family, that later during the catalysis interact directly with the Zn 2+ ion as proposed is for several beta lactamases including the more recent New Delhi metallo-β-lactamases 9,66,67 . In LaIHA it is the oxygen of the electrophilic amide carbonyl on isatin that is directly coordinated to the catalytic Mn 2+ ion (Fig. 3). The enzyme kinetics analyses, accompanying structural determination of LaIHA and RsIHA, present striking evidence that RsIHA, LaIHA, and LaIHB are true orthologues enzymes. These data support the claim in Bjerregaard-Andersen et al. 17 that isatin metabolism is present in bacteria native to the human gut and strengthen the hypothesis that the molecule isatin may be a signalling molecule that links the gut-brain axis under conditions such as stress 68,69 .

chemsum

{"title": "A fundamental catalytic difference between zinc and manganese dependent enzymes revealed in a bacterial isatin hydrolase", "journal": "Scientific Reports - Nature"}

from_molecular_adsorption_to_decomposition_of_methanol_on_various_zno_facets:_a_periodic_dft_study

## Abstract: Methanol is an interesting and important molecule to study because of its potential to replace existing fuels. It is also a prominent hydrogen source which can be used to generate hydrogen in-situ. ZnO is widely used as catalyst in synthesis of methanol from CO 2 at industrial scale. In this work, we demonstrate that the same catalyst could be used for MeOH decomposition. We have carried out a systematic study of interaction of methanol with various flat and stepped facets of ZnO by employing Density Functional Theory (DFT). Two flat [(1010) and (1120)] and two stepped [(1013) and (1122)] facets are investigated in detail for methanol adsorption. Chemisorption of MeOH with varying strength is common to all four facets. Most importantly spontaneous dissociation of O-H bond of methanol is observed on all facets except (1120). Our DFT calculations reveal that molecular adsorption is favored on flat facets, while dissociation is favored on step facets. Also, (1010) facet undergoes substantial reconstruction upon MeOH adsorption. Activation of C-H bond along with strengthening of C-O bond on ZnO facets suggest partial oxidation of methanol. With our DFT investigations, we dig deeper into the underlying electronic structure of various facets of ZnO and provide rationale for the observed facet dependent interaction of ZnO with MeOH. ## Introduction Methanol is one of the most important chemical for industrial reactions and a primary feed-stock for energy production. Due to easier transportation and compatibility with the existing infrastructure, methanol attracts considerable attention. Methanol is also a promising hydrogen source because of its high hydrogen content. However, use of methanol as a source of hydrogen requires breaking of its O-H and C-H bonds with substantial bond dissociation energies, viz. 96.1 kcal/mol and 104.6 kcal/mol respectively. Over the past two decades, extensive studies of activation and decomposition of methanol on various metal surfaces, metal alloys, metal clusters, metal oxides, mixed metal oxides, and zeolites have been carried out. In general metal oxides turn out to be better catalyst for activation of methanol due to the presence of oxygen on the surface, which acts as an active site. ZnO is considered as a very active catalyst for many reactions because of its mixed covalent and ionic bonding. [ Zn terminated ZnO(0001) surfaces using high-resolution electron energy loss spectroscopy (HREELS) in conjunction with temperature programmed desorption (TPD). They found that for all three ZnO surfaces, methanol adsorb dissociatively at room temperature which leads to the formation of hydroxyl and methoxy species. Upon heating to higher temperatures (370K and 440K), the dissociated and intact methanol species on ZnO(1010) predominantly undergo molecular desorption releasing CH 3 OH. While on both polar surfaces, thermal decomposition of CH 3 OH occurs to produce CH 2 O, H 2 , CO, CO 2 , and H 2 O at temperatures higher than 500K. Although ZnO is used extensively as a catalyst in many reactions, its potential is not truly realized. To the best of our knowledge only non-polar (1010) and polar (0001) facets of ZnO have been studied for methanol activation. XRD pattern shows that (1120), (1013), and (1122) are also prominent facets. These facets are hardly studied for methanol activation. In the present work, we have systematically studied the interaction of methanol with various flat (1010), (1120) and stepped surfaces (1013), (1122) by employing periodic DFT. We report not only molecular adsorption and activation of O-H bond of methanol on these facets but also spontaneous dissociation of its O-H bond leading to formation of methoxy species. The quenched C-O bond-length in methanol along with partial double bond type character indicates onset of oxidation of methanol. Finally, we demonstrate various possibilities regarding interaction of MeOH with ZnO and bring out the rationale behind the reactivity in terms of electronic structure of these facets. ## Computational Details All the calculations are carried out within the Kohn-Sham formalism of Density Functional Theory. Projector Augmented Wave potential is used, with Perdew Burke Ernzerhof (PBE) approximation for the exchangecorrelation and generalized gradient approximation, as implemented in planewave, pseudopotential based code, Vienna Ab initio Simulation Package (VASP ). The bulk unit cell is taken from the materials project. The bulk lattice parameters upon optimization are a = 3.28 and c = 5.30 demonstrate excellent agreement with the experimentally measured (a = 3.24 , c = 5.20 ) lattice parameters. Two flat facets, (1010) and (1120) of ZnO are modeled as slabs by cleaving a surface with 3x3 periodicity in x and y direction with 4 layers using Quantumwise-VNL-2017.1. Two stepped facets, (1013) and (1122) are also cleaved by taking 3x1 and 2x2 periodicity respectively in the x and y direction with 6 layers. In every model, bottom layer is fixed and rest all layers and adsorbate are fully relaxed. Van der Waals corrections are applied to account for dynamic correlations between fluctuating charge distribution by employing Grimme method (DFT-D2). It is observed that 20 of vacuum is sufficient to avoid interaction between adjacent images of planes along the z-direction. Geometry optimization is carried out with a force cutoff of 0.01 eV/ on the unfixed atoms and the total energies are converged below 10 −4 eV for each SCF cycle. A Monkhorst-Pack grid of 3x2x1 for (1010) and 3x3x1 for (1120) slabs is used. For both stepped surfaces, Monkhorst-Pack grid of 2x2x1 is used. The difference in energies is less than 4meV/atom for every system upon refining the K mesh further. Entire surface is scanned by placing MeOH molecule at all available unique sites. To compare the interaction of methanol at these sites, interaction energy is calculated using the formula: E int = E system -(E surf ace + E molecule ) where E system is energy of the system when MeOH is placed on the surface, E surf ace is energy of the bare surface and E molecule is energy of the MeOH molecule. To understand the electronic structure of these facets, total Density of States (tDOS ) are calculated with denser k-mesh using LOBSTER. Mulliken charges are computed for all the atoms on the surface. ## Results and Discussion Bulk ZnO crystallizes in the hexagonal wurtzite structure consisting of hexagonal Zn and O planes stacked alternately. Both oxygen and zinc atoms are coordinated by four zinc and oxygen atoms respectively. Polar ((0001) and ( 0001)) and non-polar ((1010), ( 1011), (1120)), (1013), and (1122)) facets have prominent peaks in XRD. In this work, we have studied the interaction of methanol with two flat ((1010), (1120)) and two stepped ((1013), ( 1122 Another flat facet that we studied is (1120). This is a highly symmetric facet with less number of inequivalent sites on the surface as shown in Fig. SI4-(a). ## Positions Interestingly, when MeOH is placed on any sites except 4 th , upon optimization it gets chemisorbed at one specific site as schematically represented in Fig. ## SI4-(b) . The orientation of methanol on this site is shown in Fig. SI4-(c). E ads for methanol at this site is -1. ## Conflicts of interest "There are no conflicts to declare".

chemsum

{"title": "From molecular adsorption to decomposition of methanol on various ZnO facets: A Periodic DFT study", "journal": "ChemRxiv"}

fret_pumping_of_rhodamine-based_probe_in_light-harvesting_nanoparticles_for_highly_sensitive_detecti

## Abstract: In this work we presented novel strategy for increasing the performance of popular fluorescent probes on the basis of rhodamine-lactam platform. This strategy is based on the incorporation of probe molecules into the light-harvesting nanoparticles to pump modulated optical signal by Förster resonant energy transfer. Using the commercially available Cu 2+ probe as a reference chemical, we have developed an efficient approach to significantly improve its sensing performance. Within obtained nanoparticles coumarin-30 nanoantenna absorbs excitation light and pumps incorporated sensing molecules providing bright fluorescence to a small number of emitters, while changing the probe-analyte equilibrium from liquid-liquid to solidliquid significantly increased the apparent association constant, which together provided a ~100-fold decrease in the detection limit. The developed nanoprobe allows highly sensitive detection of Cu 2+ ions in aqueous media without organic cosolvents usually required for dissolution of the probe, and demonstrate compatibility with inexpensive fluorometers and the ability to detect low concentrations with the naked eye. ## Introduction In recent years, colorimetric and fluorescent probes have become an effective analytical tool due to unique capability for sensitive monitoring of metal ions , anions , reactive oxygen species and biomolecules . Among the various chromophore scaffolds, lactam derivatives of rhodamines represent one of the most developed platform for the design of selective colorimetric and fluorescent probes due to their high extinction coefficients, high quantum yields, and excellent chemical and photostability . Typically, such probes are rhodamine dyes modified with a selective receptor capable of interacting with the analyte. The sensing mechanism is based on the structural change of a molecule from a spirocyclic lactam to an open-ring amide . Without analyte the probe exists in a colorless non-emissive spirocyclic form, while interaction of ligand group with specific molecule or ion results in ring opening and appearance of fluorescence and pink coloration. Obviously, the sensitivity of analyte determination depends on the brightness of the probe (absorptivity x fluorescence quantum yield), as well as the equilibrium constant of probe-analyte interaction. The molar absorptivity of rhodamine 6G is approximately 10 ^5, and in a standard cuvette with an optical path of 10 mm, a change in [probe-analyte] complex concentration of 10 nM will correspond to change in optical density of ~0.001, which is equal to the measurement error of a typical spectrophotometer. The binding constants of real probes are in the range 10 3 -10 5 , and the typical working range of analyte detection is ~1μm-0.1mM. Of course, fluorescence measurements are more accurate than colorimetric, but they do not provide a significant increase in sensitivity, since they have exactly the same fundamental limitations. Modern approaches for ultrasensitive chemical sensing on the basis of surface-enhanced Raman scattering or surface-enhanced fluorescence , allow determination of significantly lower analyte concentrations, but require powerful and expensive experimental equipment, both for measurements and for fabrication of nanostructured enhancing substrates. It is clear that to improve the performance of fluorescent probes, it is necessary to significantly enhance the response signal by pumping the probe with a much brighter luminescent nanoparticle via Förster resonance energy transfer (FRET), however, nanoparticles are generally not efficient FRET donors because their sizes are beyond the FRET radius (1-10 nm), and in case of semiconductor nanocrystal quantum dots, 10-50 acceptor molecules are required to ensure efficient FRET . New possibilities appeared with recently introduced light-harvesting FRET nanoparticles on the basis of cationic dyes separated by bulky hydrophobic counterions that prevent dye self-quenching . In these NPs a short inter-fluorophore distance controlled by the counterion enable ultrafast dye-dye excitation energy migration on a femtosecond time scale through the whole particle within the fluorescence lifetime until it reaches a donor close to the acceptor leading to FRET. Therefore, the energy can be transferred beyond Förster radius from multiple donors to a single acceptor, providing a basis for signal amplification . Inspired by recent works on the topic demonstrating the capabilities of this method, we aimed to enhance the response signal of popular rhodamine-based probes. Using the commercially available Cu 2+ probe as a reference chemical, we have developed an efficient approach to significantly improve its sensing performance, which requires chemical modification of the probe and incorporation it into lightharvesting FRET nanoparticles to amplify its fluorescence signal. Within these particles coumarin 30 nanoantenna absorbs excitation light and pumps incorporated sensing molecules providing bright fluorescence to a small number of emitters, while changing the probe-analyte equilibrium from liquid-liquid to solidliquid significantly increased the apparent association constant, which together provided a ~100-fold decrease in the detection limit. The developed nanoprobe allows highly sensitive detection of Cu 2+ ions in aqueous media without organic cosolvents usually required for dissolution of the probe, and demonstrate compatibility with inexpensive fluorometers and the ability to detect low concentrations with the naked eye. ## Chemicals and instruments Rodamine 6G (99%), Coumarin 30 (99%), 4-(dimethylamino)benzaldehyde (99%), hexane (95%), chloroform (99%), ethyl acetate (99.8%), N 2 H 4 *H 2 O (98%), sodium tetrakis [3,5-bis(trifluoromethyl)phenyl]borate, sodium tetrakis[3,5-bis(1,1,1,3,3,3hexafluoro-2-methoxy-2-propyl)phenyl]borate trihydrate, silica gel (100/200 μm) were purchased from Sigma-Aldrich and used as received. All other reagents were of analytical grade and used without purification. All aqueous solutions were prepared using deionized water. The Fourier transform infrared radiation (FT-IR) spectra of the compounds in the range 400-4000 cm -1 were recorded using a Perkin Elmer Spectrum 100BX II spectrometer in KBr pellets. 1 H, 13 C NMR spectra were performed on a Bruker Avance 400 with the frequency of proton resonance 400 MHz using CDCl 3 as the solvent and tetramethylsiliane as the internal reference. UV-VIS and fluorescence measurements were performed on Shimadzu UV-2600 spectrophotometer and Shimadzu RF-6000 spectrofluorophotometer using 1 cm path length cuvettes at room temperature. The size and electrokinetic potential of fluorescent particles were determined using ZetaSizer Nano ZS analyzer (Malvern Instruments Ltd.) The pH measurements were carried out using a Sartorius Professional Meter PP-50. ## Synthesis and characterization of d98 Rhodamine 6G hydrazide was prepared as described in . Yield 80%. 1 H NMR (400 MHz, CDCl 3 , ppm, δ): 7.96 (m, 1H), 7.45 (m, 2H), 7.06 (m, 1H), 6.39 (s, 2H), 6.26 (s, 2H), 3.58 (s, 2H), 3.54 (br.s, 2H), 3.22 (q, 4H), 1.92 (s, 6H), 1.32 (t, 6H); Elemental Analysis data: Calc. C, 72.87; H, 6.59; N, 13.07; Expt. C, 72.97; H, 6.66; N, 12.89. d98. A 300 mg (7E-4 mol) of rhodamine 6G hydrazide and 220 mg (1.4E-3 mol) of 4-(dimethylamino)benzaldehyde were dissolved in 15 ml of ethanol (95%). The mixture was refluxed for 6 h. The solvent was removed under reduced pressure and the crude product was purified by flash column chromatography on a silica gel with hexane/ethylacetate (v/v = 2/1) as the eluent to afford the product as a crystal powder (290 mg, 74%). 1 pH range (2-10), however fluorescence of acidic solutions was higher than that of basic ones. Since the work was focused on improving the performance of the rhodamine-lactam platform, all further experiments were performed at pH = 5, the maximally acidic environment in which the response of such probes is not crossinfluenced by hydrogen ions . In the search for the optimal C30/counterion ratio, which ensures the stability of resulting colloidal solution and a high FLQY, we varied the excess of the counterion at a fixed dye concentration. Fig. 2(a, c, d) shows that with a sufficient excess of corresponding f-TPB, the absorbance of the resulting solution is significantly higher than that of coumarin aqueous solution of the same concentration, while the absorption band is red-shifted (Fig. 2a, solid curves). This spectral shift clearly indicates protonation of coumarin and redistribution of electronic density, which was also confirmed by acidification of the coumarin with hydrochloric acid, where a similar redshift was observed (Fig. 2a, dashed curves). In the case of an equimolar ratio or a slight excess of the counterion (less than 5), the resulting solutions are much less stable and partially precipitate within a few hours, while at higher counterion concentration the solutions remained stable for at least several weeks. DLS measurements (Fig. 2(c, d), blue curves) revealed that the size of the formed particles also changes with an increase in counterion excess. In the case of F12, the size sequentially decreases from 140 to 30 nm at a 20-fold excess and remains unchanged (Fig. 2c, blue curve), while in the case of F6 it gradually increases from 60 to 100 nm (Fig 2d, blue curve). This difference can be explained by difference in solubility of corresponding anion precursors. Thus, NaF12 is soluble in water, the resulting NPs mainly consist of the insoluble salt C30/F12, while the excess of anions is adsorbed on solid surface, forming the first negatively charged layer of the electric double layer (more anions stabilize smaller particles). The solubility of NaF6 in water is much lower and it coprecipitate with C30/F6 forming large mixed particles. It is noteworthy that in both cases a higher excess of counterions provides the dye a higher FLQY (Fig. 2(c, d)): at pH=5, the maximum quantum yields were ~ 40% and 20% for F12 and F6, respectively, whereas in an aqueous solution without counterions, FLQY of coumarin was 0.08 at pH=7 and 0.04 at pH=5 (for comparison, in EtOH, FLQY=0.8). A similar impressive fluorescence light up has been noted for rhodamine B, a cationic dye, and was attributed to unique hydrophobic environment within the nanoparticles , however, for a neutral dye that is quenched upon protonation even in good solvent (Fig. S2), this is the firsttime report. We then obtained FRET nanoparticles with the fixed C30/counterion ratios (C30/F6=1/50, C30/F12=1/20) and a variable amount of the model acceptor (R6G) in order to estimate the donor/acceptor ratio, ensuring a noticeable energy transfer within nanoparticles. As shown in Fig. 3(a, b, d, e FRET efficiency in F12-and F6-based NPs, respectively. It should also be noted that with an increase in acceptor loading, FLQY of obtained NPs slightly increases (Fig. 3f) and reaches maximum at a ratio of R6G/C30 = 1/10. Thus, taking into account d98 has 2 binding sites -hydrazone and aninoaromatic fragments (marked with red circles in Fig. 4). The hydrazone fragment has affinity for both Cu 2+ and H + , while the aminoaromatic one only for H + . In an acidic medium, both sites are occupied and d98 is colored and fluorescent, while in the presence of Cu 2+ only the hydrazone fragment is occupied, and d98 is colored and non-fluorescent. In a moderately acidic medium, the copper ion displaces both protons from the molecule, and thus its concentration can be determined from the change in the fluorescence intensity of the solution at constant absorption (i.e., by decrease in FLQY). To test whether a similar effect will take place inside the nanoparticles, lightharvesting nanoprobe were synthesized with an optimized ratio of precursors the same. To confirm that observed quenching is a result of d98-Cu 2+ interaction, we analyzed the stoichiometry of binding event according to Job`s method. Fig. 5c clearly demonstrate 1:1 binding stoichiometry between d98 and Cu 2+ both in the case of d98 solution and in the case of NPs containing d98. All these data reliably confirm the successful implementation of the key idea of the work: to increase the performance of the probe by pumping it with multiple donor molecules; however, the achieved increase in performance turned out to be much higher than the optical pumping factor. We believe that such behavior is associated with a change in the type of chemical equilibrium. Fluorescent probes are usually poorly soluble in water and require organic co-solvent, which provides the probe solubility and reversible probe-complex equilibrium in homogeneous solution. In our case, solid nanoparticles act as a cation exchanger; thus, the observed response is the result of a heterogeneous sorption process with much higher equilibrium constant. F6-and F12-based NPs provide 10-and 100-fold better sensitivity than d98 in solution. ## Conclusions To conclude, we have presented novel strategy for increasing the performance of very popular fluorescent rhodamine probes, which is based on the incorporation of probe molecules into the light-harvesting nanoparticles to pump optical signal by Förster resonant energy transfer. According to published works, for the formation of bright FRET particles (without aggregation caused quenching), cationic dyes should be associated with bulky hydrophobic anions, however, organic cationic dyes emit in the yellow-NIR spectral range, which means that the blue-green absorbing dyes can be used only as donors, but not as acceptors. We have shown that neutral dyes containing amino groups act as cationic dyes and can also be used in the ion association method for preparation of fluorescent nanoparticles to pump blue-green absorbing cationic dyes. Finally, on the basis of commercially available Cu 2+ probe we have developed FRET-nanobrobe with sub-nM detection limit which demonstrate compatibility with inexpensive fluorometers and the ability to detect low concentrations of Cu 2+ with the naked eye.

chemsum

{"title": "FRET Pumping of Rhodamine-Based Probe in Light-Harvesting Nanoparticles for Highly Sensitive Detection of Cu 2+", "journal": "ChemRxiv"}

proline_provides_site-specific_flexibility_for_in_vivo_collagen

## Abstract: Fibrillar collagens have mechanical and biological roles, providing tissues with both tensile strength and cell binding sites which allow molecular interactions with cell-surface receptors such as integrins. A key question is: how do collagens allow tissue flexibility whilst maintaining well-defined ligand binding sites? Here we show that proline residues in collagen glycine-proline-hydroxyproline (Gly-Pro-Hyp) triplets provide local conformational flexibility, which in turn confers well-defined, low energy molecular compression-extension and bending, by employing two-dimensional 13 C-13 C correlation NMR spectroscopy on 13 C-labelled intact ex vivo bone and in vitro osteoblast extracellular matrix. We also find that the positions of Gly-Pro-Hyp triplets are highly conserved between animal species, and are spatially clustered in the currently-accepted model of molecular ordering in collagen type I fibrils. We propose that the Gly-Pro-Hyp triplets in fibrillar collagens provide fibril "expansion joints" to maintain molecular ordering within the fibril, thereby preserving the structural integrity of ligand binding sites.The dominant protein components of the extracellular matrix are ordered fibrillar collagens. These collagens must provide well-defined binding sites for many matrix proteins and cell-adhesion receptors, exemplified here by integrins. These same collagen fibrils also constitute the main mechanical component of the extracellular matrix, constantly subjected to local forces from adherent cells, which induce local collagen molecular movements likely to disrupt collagen-ligand bindings. How the collagen molecular and fibrillar structures are able to fulfil at first sight contradictory ligand binding and mechanical roles is an important question for both biology and materials scientists developing biomimetic implant materials.Fibrillar collagens are triple-helical proteins and, with the exception of their short N-and C-terminal telopeptides, consist entirely of G-X-Y (or Gly-Xaa-Yaa) triplet repeats where X is most commonly the cyclic imino acid, proline, and Y, hydroxyproline (O/Hyp), a post-translational modification of proline [1][2][3][4] . The hydroxylation of a Yaa position P, occurring immediately after synthesis of the collagen precursors, causes the resulting O ring to strongly favour the exo conformation (ring Cγ pointing away from the residue C=O group), which in turn pre-organises the peptide chain structure at the O residues towards the polyproline II helix that is required in each strand of the collagen triple helix; thus, GPO triplets in collagens are widely viewed as being essential for triple helix folding and stabilizing the triple helix structure 5,6 .The X-position P rings (P X ) in GPO triplets of short model collagen peptides occupy metastable structures, and endo and exo ring conformations (Fig. 1) are almost equally favoured at biologically-relevant temperatures. Endo and exo conformations have significantly different backbone geometry, and, moreover, flipping between them occurs on a nanosecond timescale 7 . Proline rings in purified collagen preparations are also dynamic on a nanosecond timescale [8][9][10][11][12][13][14] ; in native hard and soft tissues, both backbone dynamics 15,16 and proline flipping dynamics 17 are still retained. Taken together, these data suggest that native collagen GPO triplets are actually flexible, rather than rigid, structural regions. The high GPO triplet abundance in native collagen sequences led us to speculate that GPO P X flexibility could be key to allowing collagen to simultaneously provide essential mechanical properties and structurally well-defined protein binding sites for their biological roles. ## Results and Discussion Proline(X) rings are flexible in biologically-derived samples. A significant source of P X flexibility arises from ring flipping between endo and exo conformations. Solid-state NMR spectroscopy provides an accurate method for assessing the distribution of collagen proline ring endo-exo conformations in intact tissues. In previous work, we utilised the 13 C chemical shift of proline Cγ to determine the distribution of proline P X ring endo-exo conformations in model collagen peptides 7 . There, we deduced that the collagen GPO P X 13 C shift is about 24 ppm for the endo conformation, and 25.7 ppm for the exo conformation. Proline rings in crystalline proline 18 and in model collagen peptides 7 rapidly flip between endo and exo conformations at room temperature and above, while by measuring dipolar coupling order parameters in native collagen, proline rings have been shown to undergo greater angular fluctuations 17 compared to glycine and hydroxyproline 13 . Rapid P X ring flipping means that the observed 13 Cγ isotropic shift is a population-weighted average of the endo and exo conformation 13 Cγ chemical shifts, allowing the endo:exo population ratios for P X rings to be determined from the population-weighted average shift 7 . In our previous work, we have demonstrated the presence of rapid endo-exo flips for proline rings in model collagen-like peptides containing the GPO/POG motif at room and physiological temperatures, and our simulations indicate that these puckering motions are coupled to the backbone motion of the triple helix. In the present work, we look for evidence of similarity in proline conformation dynamics between model peptides and more biologically-relevant samples to extend our findings to native collagen fibrils in the extracellular matrix. Resolving the GPO proline 13 Cγ signals for model collagen peptides where specific 13 C labelling is readily achieved is straightforward. However, resolving this signal in intact tissue samples is not as easy; at natural 13 C abundance, the 13 C NMR spectrum of an intact collagenous tissue contains signals from all amino acid residues, including collagen GPY triplets, where Y is any residue not hydroxyproline (O), as well as the GPO proline signals of interest. Here we use 13 C, 15 N-enriched mouse bone and in vitro sheep osteoblast matrix, both generated as described previously 19,20 , as models for fibrillar collagen type I in intact tissues. 1D NMR spectra of all samples reported in this manuscript can be found in Supplementary Fig. S2. Isotopic enrichment is necessary to use two-dimensional 13 C NMR spectroscopy for resolving the proline 13 Cγ chemical shift distributions in these intact tissues through 13 C- 13 C double-quantum-single quantum (DQ-SQ) correlation spectra. In this type of spectrum, pairs of strongly dipolar-coupled (i.e. spatially close) 13 C nuclei give signals in the double-quantum spectrum at the sum of the chemical shifts for the two 13 C nuclei. These DQ signals are correlated in the other spectral dimension with the normal (single) quantum 13 C NMR spectrum for the respective pair of 13 C nuclei. Horizontal slices through the DQ-SQ correlation spectra thus yield one-dimensional 13 C spectra of pairs of spatially close carbon nuclei. 2D 13 C DQ-SQ spectra from three 13 C, 15 N-labelled collagen/ collagen-like samples are shown in Fig. 2 for comparison: mouse bone, cultured osteoblast matrix and a model triple-helical collagen peptide, ((GPO) 5 (G*P*O)(GPO) 5 ) 3 , where * indicates a U- 13 C, 15 N-labelled residue (in each and every chain of the triple helix). As Fig. 2 shows, the connectivity of bonded carbons in the proline ring can be traced out in the DQ-SQ spectrum. By comparison with the signal connectivities in the model peptide spectrum, we can make straightforward assignment of collagen proline P X signals in the 13 C, 15 N-labelled mouse bone and in vitro matrix spectra. We then need to separate 13 C signals from P X in general GPY triplets, where Y ≠ O (233 per collagen triple helix) from those in GPO triplets (103 per triple helix) in the mouse bone collagen DQ-SQ spectrum. We utilise the fact the 13 C signals from P X in GPO triplets are subject to the proline effect 21,22 , which manifests as a 1-2 ppm reduction in 13 C chemical shift for Cα, Cβ and C′ for residues that precede an imino acid, as in the case of P X in GPO triplets, where P X precedes Hyp/O, but not for P X in GPY triplets. The GPO P X Cα-Cβ correlation in the bone DQ-SQ spectrum can be distinguished from the corresponding GPY P X correlation using this criterion, as indicated in Fig. 2. From the GPO P X Cα-Cβ correlation, the GPO P X Cβ-Cγ correlations are found, as shown in Fig. 2 and Fig. 3, from which the mean 13 Cγ chemical shift for the bone GPO P X rings is determined as 24.8 ppm, with partial overlap with the main population of GPY centred at 25.3 ppm. We note that the strong similarity of the proline signal chemical shift distribution between the mouse bone and in vitro matrix samples for all proline signals is a clear indication that the collagen structures in the two samples are similar, unlike the peptide proline distribution, which clearly only overlaps with a subset of that found in bone. As discussed above, this 13 Cγ chemical shift represents a population-weighted average between the 13 Cγ chemical shifts expected for the endo (23.8 ppm) and exo (25.7 ppm) proline ring conformations 7 . The mean 13 Cγ chemical shift of 24.8 ppm for bone collagen GPO P X , while only a small change compared to the 13 Cγ of 25.3 ppm for the (GPO) 11 model collagen peptide 7 , indicates the mean population distribution for collagen GPO P X in the intact tissues is slightly more skewed towards endo compared to the collagen model peptide. Nonetheless, this observed 13 Cγ chemical shift (24.8 ppm) is still distinct from a "pure" endo chemical shift (23.8 ppm), indicating that the proline rings in GPO triplets in bone do not strongly favour the thermodynamically more stable endo conformation , and instead show a distribution between endo and exo conformations, with fast exchange over the NMR time scale, i.e. they are flexible. To confirm our assignment, we note that the intensity of the signals arising from GPO triplets compared to the GPY (where Y ≠ O) should match the expected ratio in the sequence for mouse collagen type I. Figure 4 shows 1D slices taken from the SQ-DQ experiment in Fig. 2, at the C′-Cα, Cα-Cβ, and Cβ-Cγ DQ (sum) frequencies. The GPY:GPO triplet ratio in the sequence is 233:103, which means we expect to observe a GPY signal that is approximately 2.3 times more intense than that of GPO (assuming similar linewidths, if not then the integral of the signals should match this ratio). From Fig. 4, we can see that this is indeed true. The effect is clearer in the case of the osteoblast matrix cultured in vitro than the bone spectrum, as the bone spectrum is labelled in other amino acids apart from proline and glycine, giving rise to overlapping signals that can skew the observed ratio. We expected that GPO P X would occupy a much larger range of conformations for fibrils in intact tissues compared to model collagen-like triple helical peptides. However, this is not the case, as evidenced by the NMR lineshapes for different samples shown in Fig. 4. The relative line widths are most easily assessed for the model peptide and in vitro osteoblast matrix, because the Pro 13 C signals are not obscured by signals from other labelled amino acids as they are for the bone sample. There is some overlap between GPO and GPY proline 13 C signals for the osteoblast ECM sample, particularly for C′, which needs to be taken into account when comparing the Pro 13 C lineshapes (GPY C′ sites give a shoulder on the high frequency side of the GPO C′ signal). Nevertheless, it is clear that the GPO proline Cα and Cβ lineshapes are highly similar between model peptide and in vitro matrix collagen. The Cα lineshapes are directly comparable. In the Cβ spectrum, there is some additional small signal to low frequency for the in vitro matrix sample, but the majority signal has a similar linewidth to that for the model peptide. The lineshapes allow us to directly compare the range of conformations for GPO P X between collagen in an intact tissue and the model collagen peptide. The single-quantum NMR lineshapes are in effect the sum of 13 C DQ-SQ correlation spectra of mouse calvarial bone (green), in vitro osteoblast extracellular matrix (blue) and model collagen peptide (orange). All spectra were obtained at 10 kHz MAS and 297 K, and only the proline carbon regions are shown in this figure (full spectra can be found in Supplementary Fig. S3). The peptide and osteoblast ECM samples are isotopically enriched specifically in glycine and proline residues. In the bone sample, ~20% of essential amino acids and glycine are U-13 C, 15 N labelled. The GPO proline signals are labelled in the figure and the 13 C- 13 C correlations between them are traced (purple line) for the model peptide spectrum, starting with the amide carbon at around 172 ppm. The same trace is overlaid on the bone and osteoblast matrix spectra to show how we arrived at the assignment of the collagen GPO proline signals in those spectra, with a slight difference in that the Cγ signal in the spectra obtained from biologically-derived samples is at 24.8 ppm rather than 25.3 ppm in the peptide spectrum (pink line). The GPY (where Y ≠ O) connectivity is shown to be separate from that of GPO (grey line). The arrows on the bone spectrum indicate the two sets of signals corresponding to Cβ-Cγ, where the difference between the GPO (purple) and GPY (grey) signals can be most clearly observed. The black dotted line in each spectrum indicates the DQ-SQ diagonal. individual lineshapes from each GPO P X residue in the sample, each having an isotropic shift determined by the local molecular geometry for each individual P X . These individual lineshapes are not resolved, but the width of their sum signal, i.e. the observed signal in the SQ dimension of the DQ-SQ spectra, retains information on the range of GPO P X isotropic chemical shifts and thus on the distribution of time-averaged P X ring conformations. Apaft from the distribution of proline ring geometries, the NMR linewidths can be affected by by other factors: T 2 (transverse) relaxation, homonuclear 13 C- 13 C dipolar coupling and molecular motion all contribute to the homogeneous linewidth. Hydration and sample temperature also play a role. While we have carried out our experiments with a fairly low level of hydration in order to capture a maximum range of conformations (lyophilization generally traps conformations and increases heterogeneous broadening), and maintained them at similar experimental temperatures, we cannot resolve the different contributions to the 13 C lineshapes. However, correspondence of the overall 13 C lineshapes implies some similarity in the distribution of proline residue conformations, molecular motion and T 2 relaxation time (which itself depends on molecular motion). 3 is shown with a green arrow on the bone GPO Cβ-Cγ slice. Vertical scaling was normalised on the larger signal within a GPO/GPY pair, maintaining the intensity difference between GPO and GPY to scale, but the vertical scale between pairs of GPO/GPY and between different samples are not shown to a unified scale. All spectra are presented on the same horizontal scale, enabling comparison of the chemical shift and also the linewidth at half height. We further confirmed similarity in the distribution of GPO geometries between the in vitro matrix collagen and the collagen model peptide by following the build-up of G-P 13 C correlation signal intensity in a series of 2D 13 C-13 C proton-driven spin diffusion (PDSD) correlation NMR experiments in which the mixing time for 13 C to 13 C magnetization transfer is varied (see Supplementary Fig. S4). In summary, we see that P X in GPO exhibits a mix of endo and exo conformations, in a distribution that is similar to what we see in the model peptide. In our previous work 7 , we have shown that collagen-like model peptides exhibit a dynamic equilibrium of endo and exo conformations, with the equilibrium positioned at a ratio of near 50:50 at biologically-relevant temperatures. With the experimental data summarised thus far, we propose that the same distribution of endo and exo conformations (with very little bias towards endo) is observed in biological samples. The Pro(X) rings in GPO triplets have greater local conformational flexibility than in GPY sequences. From the NMR data presented above, we determined that P X in GPO in native collagen samples has a distribution over endo and exo conformations. However, it is also clear that P X in GPY (where Y ≠ O) exhibits a 13 Cγ chemical shift that is also greater than that in a pure endo case. In previous work 7 , we used the potential energy landscape approach 26 to demonstrate the inherent flexibility or frustration of helix parameters, which is coupled to P X endo-exo flips in GPO sequences. We use the same approach here to analyse whether the flexibility of P X rings in GPO triplets differs in any way from GPY triplets, where Y is a conformationally less constrained amino acid. Specifically, we predict the structural behaviour of P X rings in a GPA triplet compared with those in a GPO triplet. A GPA triplet was selected since it is the most common collagen type I GPY triplet in which Y is an amino rather than an imino acid. All the calculations involve geometry optimisation to characterise local minima and the transition states and pathways that connect them on the potential energy landscape. By applying appropriate structural perturbations to a ground state conformation, this procedure allows for sets of energy minima to be computed that span particular regions of conformational space. For the present purposes we are interested in the subspace of minimum energy conformations that can be adopted by perturbing the backbone dihedral angles of a single GPY triplet while it is embedded within in a larger collagen triple helix. Flexibility at a residue comes from the presence of a range of energetically-accessible conformations for that residue. The relative number of distinct minima found in each subspace and the corresponding range of dihedral angles are taken as a measure of the likely flexibility. If many alternative conformations exist, separated by thermally accessible barriers on the experimental time scale, then we predict greater flexibility. The structural perturbations to the ground state structure were carefully designed to uniformly sample the space of backbone dihedral angles and proline endo and exo conformations for the sixth triplet in the trailing chain in both (POG) 12 and (PAG) 12 superstructures. A single perturbation consisted of choosing a pair of atoms separated by a linear set of covalent bonds, and rotating as a rigid body all the intervening atoms by an angle randomly selected within a maximum amplitude. When such rotations occur between two backbone atoms the local ring geometry and endo/exo structures are conserved, but the backbone dihedrals, in which the atom pair are involved, change. Rotations applied to the atoms between the Cβ and Cδ atoms in the proline rings can force a subsequent relaxation into either endo or exo conformations, depending on the angle of rotation, which also modifies the backbone dihedrals of the proline and neighbouring residues. Further details of the particular groups considered can be found in the Methods. By repeating our structural perturbations for each triplet, an ensemble of feasible conformations, corresponding to local potential energy minima, were generated. This analysis yielded only four minima in the case of (PAG) 12 , and 64 minima in the case of (POG) 12 . Rather than presenting the whole database of structures, which would be unwieldy, we illustrate the backbone dihedral angles of two neighbouring P X and Y residues in the centre of the sequence for each structure in a Ramachandran plot (Fig. 5). For ease of viewing, each residue (proline and alanine, or proline and hydroxyproline) is presented on different panels of the figure. In the case of P X in (POG) 12 , we have further divided the plot into two, based on whether the hydroxyproline in the structure has an endo or an exo pucker. Thus, each point in each Ramachandran plot represents a minimum in the potential energy landscape, i.e. an alternative accessible conformation. For P X rings in the GPA triplet, all the perturbed minima in the case of P X endo pucker have nearly identical dihedral angles, resulting effectively in just two possible backbone conformations, one associated with endo pucker at P X , and the other with exo pucker. There are therefore, two well-defined, accessible backbone conformations at GPA triplets. To deform the GPA triplet, or in other words, to access GPA P X backbone structures with significantly different φ/ψ angles to these endo or exo minima, would require large energy perturbations, corresponding to thermally inaccessible barriers at biologically relevant temperatures. In striking contrast, the P X ring in a GPO triplet can move between local energy minima that exhibit a range of φ/ψ angles for both endo and exo P X conformations. The diversity of conformations accessed via local structural perturbations for GPO P X rings depends on whether the neighbouring Hyp is in the endo or exo conformation. The combination of endo and exo conformations for proline and hydroxyproline in GPO gives rise to four clusters, as illustrated in Fig. 5(d,e), where P X (and Hyp) endo-exo ring conformations are accompanied by systematic shifts in backbone conformation corresponding to extension-compression of the helical chain. Combined with the NMR results presented in the first section, we conclude that the GPO proline rings in native collagen proteins exhibit flexibility in the form of endo-exo pucker, just as previously observed in model peptides. Our computational results further demonstrate that these endo-exo pucker motions are not limited locally at the proline ring, but are coupled to the backbone, and can lead to overall extension and compression of the triple helix at GPO triplets. In the context of heterotrimeric collagens and fibrils formed from collagen triple helices, it is likely that these extension and compression motions will lead to bending of the triple helix and the fibril. ## GPO Pro(X) flexibility may be necessary for the biomechanical function of collagen. We next asked what is the role of such flexibility at GPO triplets, and specifically whether it could impact on fibril as well as molecular flexibility. To determine the location of the evolutionarily conserved GPO triplets in the fibril, we carried out sequence alignment across diverse species for collagen type I and generated a consensus sequence for each chain. Using the database of collagen type I sequences, we also calculated the conservation of each GPO triplet. Full results of the calculations for collagen type I are presented in Supplementary Table S1 and Fig. S5. From our sequence analysis, we find that there are 33 highly conserved GPO sites across collagen α1(I) chains out of 43 in the consensus sequence, and 20 highly conserved GPO triplets across collagen α2(I) chains out of 33 in the consensus sequence. If we consider only mammalian sequences, these values increase to 41 out of 43 GPO triplets in the α1(I) chain, and 24 out of 33 in the α2(I) chain being defined as highly conserved at the same conservation threshold (75%). The distribution of GPO sites in the experimentally-determined model 27 of collagen type I fibrils and also the consensus sequence is shown in Fig. 6. For ease of comparison, the consensus sequence for the collagen type I α1 Figure 5. The energy landscape simulation results presented as Ramachandran plots. For each different, accessible conformational structure, the dihedral angles of two consecutive residues, proline and the residue subsequent to proline, are plotted. These backbone conformations arise from perturbing a single Pro residue in (PAG) 12 and (POG) 12 whilst maintaining the overall ring conformation as either endo or exo. For (PAG) 12 , only four accessible structures were found, three of which were very similar in backbone dihedral angles in the current scaling, and therefore overlap nearly perfectly. (a) shows the backbone dihedral angles for the alanine residue of these four structures and (b) shows the backbone dihedral angles for the proline residues, showing a clear separation in backbone dihedral angles for endo and exo ring conformations. For (POG) 12 , 64 accessible structures were found. (d) shows the backbone dihedral angles for the hydroxyproline for all 64 structures, split into two populations according to the conformation of the hydroxyproline ring. The preceding proline ring conformations were plotted for the case where hydroxyproline is exo (e) and endo (f). Panel (c) illustrates the dihedral angle change in terms of overall backbone conformation using structures generated for a GGG tripeptide (glycine was used for clarity) of constant dihedral angle using Avogadro 1.1.1 54 , ranging from a fully extended backbone conformation (180°, 180°) to a much more compressed coiled conformation (−20°, 120°). It is clear that as φ and ψ increase, the peptide backbone is increasingly extended, while as the dihedral angles tend towards lower values, the backbone is increasingly compressed. Although the dihedral angle changes presented in the simulation represent a level of change that is smaller than that shown in the middle two peptides in (c) (a length difference of under 5%), the absolute extent of expansion/compression will scale with the length of the peptide. As previously reported 7 , the idealized 7/2 (tighter) and 10/3 (looser) triple helices both have dihedral angles that more closely match endo P X in (b,e and f). and α2 chains is arranged by D-period. The consensus sequence for type I collagen is provided as a larger PDF image in the Supplementary data. From Fig. 6, it is clear that there is a spatial correlation of GPO sequences; far from being randomly scattered over the fibril structure, these sequences occur primarily as banded clusters across the fibril structure that persist across different D-periods, exemplified in the XRD-derived collagen type I structural model 28 in Fig. 6a and the sequence representation in Fig. 6b. The distribution of GPO triplets varies within the fibril structure. Collagen fibrils have two zones: the overlap zone, where all five D periods overlap, and the hole zone, where the short D5 period leads to a gap or hole in the fibrillar structure. There are clearly more GPO triplets present in the overlap zone, and fewer in the hole/gap zone. The banding of GPO triplets is also more obvious in the overlap zone. Bearing in mind the conformational flexibility of the GPO triplet structure, these clusters of GPO triplets in the fibril represent regions where the fibril structure itself can rapidly extend, compress or bend, and so serve as flexible joints for both the molecular and fibrillar structure, without significantly increasing the energy of the system. Hence, we propose that these GPO triplets act as "expansion joints", especially when viewed in the context of their distribution over the fibril. While Y-position hydroxyproline is known to increase the thermal stability of a collagen triple helix 29 , and to promote PPII formation in the backbone 5,30 , the secondary effect of placing the preceding proline into a conformationally frustrated state (i.e. with many pucker conformations of similar energy) 7 provides local flexibility at these conserved locations of the triple helix and therefore also of the fibril. Such local flexibility can be important for the overall integrity of the fibril, thus explaining the observation that thermal stability is associated with total imino acid content (proline and hydroxyproline) and not hydroxyproline alone 31 . We consider three important implications arising from our proposal of GPO sites as expansion joints in collagen fibrils: firstly, the role of of GPO triplets in collagen binding by considering the distribution of GPO relative to known binding sites on collagen proteins; secondly, the possibility that asymmetrical distribution GPO triplets may aid larger angle bending and functional displacement of the C terminus in the D5 period; and thirdly, the consequences of known mutations of conserved GPO sites. ## Role of GPO triplets in binding of collagen fibrils to other ECM components. Different types of binding are likely to exert different mechanical forces on the collagen fibril. The integrin α2β1 domain-peptide co-crystal (PDB 1DZI) 32 describes a bend in the model collagen peptide upon ligation, suggesting that molecular distortion may be necessary for optimal interaction of native collagen with the integrin. Such molecular distortion Figure 6. The distribution of GPO sites within a type I collagen fibril, from diffraction data (a) and in the consensus sequence (b). In (a), the collagen fibrillar structure (from PDB 3HR2 27 ) is modelled from X-ray diffraction data on rat tail tendon 28 with the Cα atoms in GPO triplets represented as red spheres. In (b) the consensus sequence based on the most frequent amino acid at each position is shown. The three chains of collagen are staggered by one residue with respect to each other, based on a previous study on the VWF A3 domain binding 55 . The D period arrangement shows a good match to the experimentally-derived fibril structure model above. Highly conserved (75%+) GPO triplets are highlighted in red; others in pink. High affinity integrin binding sites are in blue and the DDR/ VWF binding site in green. upon binding to the integrin would affect the integrity of the collagen fibril, and if replicated many times by the multiple collagen-integrin interactions presented by a cell, significant disorder of the collagen fibril would result. We note that the high affinity integrin binding sites, shown in Fig. 6, all occur in close proximity and on the N-terminal side of conserved GPO triplets. On the other hand, the main DDR and VWF binding site, also occurring close to multiple strongly conserved GPO triplets, is found on their C-terminal side. We hypothesise that the clusters of GPO triplets serve to protect the structural integrity of the adjacent sites when the fibrillar structure is subjected to external forces. The multiple highly conserved GPO triplets close to the DDR/VWF binding site in the fibril (as seen in the D-period arrangement of the consensus sequence in Fig. 6b) may provide additional controlled flexibility for this binding site. We speculate that local structural distortion is necessary to maintain the organization of collagen molecules within the fibrillar structure when ligands bind, and that clusters of GPO triplets across the fibril in close proximity to the collagen-ligand binding site provide controlled, reversible local distortion of the structure. ## Asymmetrical distribution of GPO triplets may facilitate functional bending of the triple helix. The distribution of GPO triplets between the three chains of the collagen triple helix can affect its accessible overall motion range. Where a GPO sequence occurs at the same locus of all three chains, proline ring endo-exo flips will allow extension-compression in all three chains, permitting concerted movement at this position. The stagger between the three chains dictates that extension-compression is never restricted to the axis of the triple helix, but will always include a bending component. For collagen type I, a heterotrimer containing two α1 chains and one α2 chain, there are many cases where GPO triplets do not occur at the same locus in both the α2 and α1 chains. Chain extension-compression can then only occur asymmetrically, bending the triple helix in a specific direction, through an angular range defined by the change in chain backbone structure upon proline endo-exo ring flip, typically a change of around 15° in proline dihedral angle. Interestingly, the majority of the GPO triplets in the fibril hole zone only occur in either the α1 or α2 chains. The X-ray fibre diffraction-derived model of the collagen fibril structure 27 shows the collagen triple helices twisting around one another through the hole zone. If this molecular twisting were to occur via homogeneously flexible regions in the collagen triple helices, molecular ordering would be compromised, as there would be no control over which direction (or by how much) a collagen triple helix can bend. However, an appropriate distribution of GPO triplets within the hole zone could readily allow well-defined bending in a specific direction, which would allow the collagen triple helices to twist around one another without the possibility of molecular disordering. Such considerations, exemplified here by collagen I, will not apply to the homotrimeric fibrillar collagens, II, III, XXIV and XXVII, but will be relevant to the heterotrimeric collagens V and XI. The (GPO) 5 and (GPO) 4 sequences at the C termini of the α1 chains and α2 chains, respectively, are well-conserved in the fibrillar collagens. The triple helix extension-compression ability that these sequences confer would allow the C-terminus some flexibility in its spatial location, thus our hypothesis that the GPO sequences play the role of "expansion joints". All other GPO triplets in D5 that precede the (GPO) 5 sequence occur only in the α1 chains, so that D5 has controlled freedom to bend at these three points. It has previously been hypothesized that the C-terminal telopeptide protects underlying ligand binding sites in the collagen fibril molecular structural arrangement 33 ; the (GPO) 5 sequences and D5 period that precede it in primary sequence may assist in such a function, by controlling the possible displacements of the C-terminal, but at the same time allowing it to be displaced, thus allowing ligand binding. ## GPO clusters can tolerate pathological mutations in collagen type I. While the GPO positions are evolutionarily conserved, we assessed the known mutations at these sites that lead to human disease. Using the NCBI ClinVar database 34 , we assessed missense mutations in collagen, and the overlap of these mutations with locations of conserved GPO triplets, to understand the functional and likely pathological consequences. We are aware that this type of analysis is prone to survivor bias. With this caveat, we will offer a few interpretations of our results. As far as the collagen type I triple helical region is concerned, there are 95 unique single base mutations for the α1(I) chain, and 110 for the α2(I) chain. For α1(I), we found 33 highly conserved GPO sites in the consensus sequence, of which 10 correspond to the reported mutation sites (human variants). Conserved GPOs account for 9.8% of the total helix sequence, and 10.5% of their locations overlap with possible mutation sites. For α2(I) we found that there are 9 GPO locations out of 20 highly conserved that overlap with reported mutation sites. 5.9% of the chain consists of conserved GPO triplets, and 8.2% of the 110 mutations overlap with conserved GPOs. By inspecting the sites where mutations overlap with bands or clusters of conserved GPOs, we note that 16 mutations (in both chains α1(I) and α2(I)) occur within a GPO band/cluster, while six mutations occur in relatively isolated GPOs. However, given the fact that many more GPO triplets occur in clusters within the triple helix or across the fibril, and relatively few are isolated, it is difficult to draw firm conclusions from this observation. The diseases associated with the reported mutations appear to vary by chain. Most mutations in α1(I) are associated with osteogenesis imperfecta, and in α2(I) with Ehlers Danlos syndrome. These results show that although mutations occurring in GPO triplets in collagen type I can be tolerated, in the sense that it is possible to survive past embryonic developmental stages, they lead to pathology in most cases. A more detailed analysis including more collagen types will be required to separate the effects of GPO (or lack thereof) on biomechanical properties of the fibril and its effects on biosynthesis and embryonic development. ## Methods Peptide synthesis and purification. ((GPO) 5 (G*P*O)(GPO) 5 ) 3 was synthesized as previously reported 7 . Briefly, the peptide was synthesized (0.1 mmol scale) as C-terminal amides on a TentaGel R RAM resin (loading of 0.19 mmol/g, Rapp Polymere) following the standard Fmoc-based solid-phase peptide synthesis strategy on a microwave-assisted automated peptide synthesizer (Liberty ™ , CEM), and purified by reverse phase HPLC. Pure peptides were characterised by matrix-assisted laser desorption and ionization-time of flight (MALDI-TOF) mass spectrometry (Supplementary Fig. S1). Labelling of mouse tissue. Our feeding and euthanasia methods were unregulated procedures under the UK Animals (Scientific Procedures) Act 1986 and therefore were not subjected to formal ethics review. Our methods complied with the review processes of the University Biomedical Service of the University of Cambridge, which is overseen by the Animal Welfare Ethical Review Body of the University of Cambridge. All experiments were performed in accordance with the UK Animals (Scientific Procedures) Act 1986. The procedure is based on our previous work 19 . Briefly, in vivo labelling was achieved with 2000 g of a gel diet (modified Classic A03 Geldiet, SAFE, Augy, France) to minimize in-cage spillage, comprising 50 g 13 C, 15 N-labelled Celtone powder (Cambridge Isotope Laboratories, Andover, MA, U. S. A.), 67 g fish hydrolysate, 410 g protein free diet (mainly corn starch), 72.9% water, 2.1% preservatives and texture additives, which was packaged in 100 g packs and irradiated at 10 kGy. Three young adult female C57Bl/6 mice, housed together, were fed ad libitum for ca. 3 weeks until the labelled diet was consumed, humanely euthanized using a Schedule 1 method and tissues harvested. Bone tissues from all animals were examined by solid-state NMR to ensure we were not assigning small biological variations between animals, and did not show major variation. ## Isolation of fetal sheep osteoblasts. Fetal sheep osteoblasts were isolated from a fetus removed from an 18 weeks pregnant sheep sacrificed for an unrelated study. Femurs were removed from the fetus. After washing several times with 1% trigene (Medichem International), the femur was stripped of muscle and non-osseous tissue to expose the bone which was sectioned into small longitudinal pieces and washed with 70% ethanol followed by repeated washings with Minimum Essential Medium (MEM; Invitrogen) to remove all traces of ethanol. Bone strips were then transferred to Dulbecco's Modified Eagle Medium (DMEM; Invitrogen) containing bacterial collagenase A (0.5 mg/mL) and dispase II (3 mg/mL) both from Roche Diagnostics. A total of 100 mL of enzyme-media mixture was used for bone sections taken from 3 limbs. Bone strips were incubated at 37 °C in a shaking water bath for 3 hours to release osteoblasts into the medium. After incubation the cell suspension was transferred to a fresh tube and the bone sections were rinsed in DMEM with 20% fetal calf serum (FCS; Invitrogen) to stop the enzymatic digestion. Rinse medium and cell suspension were pooled and passed through a 40 μm mesh filter (Appleton Woods). The cell suspension was then centrifuged at 1000 g for 5 min at room temperature to pellet the cells. The pellet was resuspended in DMEM complete medium and transferred to two T-175 cm 3 culture flasks (Nunc) and placed in a 37 °C CO 2 incubator. When the cultures were almost confluent, cells were detached with 10 ml of 0.25% trypsin containing 1 mM EDTA (SigmaAldrich) and incubated for 5 min at room temperature. The flasks were tapped at the end of incubation period to completely dislodge the cells from the flask. Trypsin was neutralized by adding 15 mL of DMEM complete media to the culture flask. The cell suspension was centrifuged in a 50 mL tube (Greiner) at 1200 rpm for 5 min and resuspended in 10 mL of DMEM. The cells were transferred into T-175 cm 3 culture flasks and were expanded to passage 3 for subsequent experiments. Basal Medium Eagle (BME) complete medium was prepared by adding 10% FCS, 30 μg/mL L-ascorbic acid 2-phosphate (Sigma), 10 mL/L L-glutamine-penicillin-streptomycin (200 mM L-glutamine, 10,000 units/ml penicillin, and 10 mg/ml streptomycin in 0.9% sodium chloride; Sigma). DMEM complete medium was prepared by adding 10% FCS, 30 μg/mL L-ascorbic acid 2-phosphate, and 10 mL/L L-glutamine-penicillin-streptomycin. All supplements were filter sterilized (0.22 μm filter, Appleton Woods) before addition. Culturing osteoblasts with labelled compounds. The procedure is based on our previous work 19 . Briefly, osteoblasts were cultured to confluence in T-175 flasks containing 25 mL BME complete medium. Labelled (U-13 C 5 , 15 N) proline (Cambridge Isotope Laboratories) and (U-13 C 2 , 15 N) glycine (Cambridge Isotope Laboratories) were added to a final concentration of 46 mg/L and 30 mg/L respectively after filter sterilization (0.22 μm filter). The cultures were incubated at 37 °C in a humidified atmosphere of 95% air and 5% CO 2 . The culture medium with isotope labelled supplements was renewed every two days until the cells and matrix began to detach from the culture flask, by which time enough ECM had formed for solid-state NMR. Refinement of the cell culture method to produce ECM that was a similar as possible to the native mouse bone tissue as judged by their respective 2D solid-state NMR spectra included adjusting the cell culture medium, the frequency with which the medium was changed, the concentration of ascorbic acid and the manner in which the cultures were handled (so as to produce minimal shear forces on the cells during changes of medium produced the optimal ECM). Samples from more than 20 batches using the final optimized protocol were prepared using isotope-enriched amino acids and all characterized by solid-state NMR to ensure reproducibility of results. Harvesting ECM from cell culture. The matrix was harvested after nine days of culture, when the cells produce a dense matrix which started to peel off the surface of the tissue culture flask. The medium was removed and the cells were washed with 20 mL 1 x phosphate buffered saline. The flask was placed in a freezer at −80 °C for 24 hours and the cells were lysed by thawing the flasks at room temperature for 30 minutes. The debris produced by cell lysis was removed by repeated washes with PBS. The decellularized ECM was dislodged by gently swirling the flask in the presence of 20 mL PBS. The matrix collected in PBS was transferred to a fresh 50 mL tube and centrifuged at 1200 rpm for 5 min at room temperature. The supernatant was poured off and the ECM dehydrated in an oven at 37 °C overnight. The samples were stored at −20 °C until NMR analysis. ECM of mouse tissue was used directly in solid-state NMR experiments without extraction, purification, or excessive processing. Rotor packing for solid-state NMR experiments. All samples were placed into Kel-F inserts for Bruker 4 mm rotors prior to being placed into a full-length (17 mm), normal wall thickness (1 mm) Bruker 4 mm rotor. The inserts provide the advantage of restricting the sample length to the region of the coil in the NMR probe that has optimal RF homogeneity. The peptide ((GPO) 5 (G*P*O)(GPO) 5 ) 3 and the ECM obtained from cell culture were lyophilised prior to packing into the Kel-F insert. The dry mass of the samples used for solid-state NMR experiments were 20.0 mg of pure peptide and 14.2 mg of ECM. The mouse calvaria bone was not subject to any drying or dehydration procedures, and was simply broken into clips without cryomilling prior to packing into the Kel-F insert. We expect that this will lead to the bone sample being slightly more hydrated than the peptide or the cell culture ECM samples, though no discernible amounts of water was observed visually after the experiment, nor was a significant reduction in sample mass observed. 12.6 mg of mouse bone was used for solid-state NMR experiments. Solid-state NMR Spectroscopy. All solid-state NMR spectra were recorded on a Bruker Avance I NMR spectrometer with a 9.4 T superconducting magnet, operating at 400 MHz ¹H, 100 MHz ¹³C and 40 MHz 15 N frequencies. For all experiments, magic angle spinning rate was set at 10 kHz, and the sample temperature is 297 K, unless otherwise specified. 13 C cross-polarisation ( 13 C CP) experiments: the standard cross polarisation sequence in the Bruker pulse programme library was used: ¹H 90° pulse length 2.5 μs, contact time 2.5 ms, with a ramped pulse on ¹H. During acquisition, SPINAL-64 35 decoupling at 100 kHz was applied on ¹H. 2D 13 C-13 C double quantum (DQ)-single quantum (SQ) correlation NMR experiment: initial cross polarisation parameters were the same as in 13 C CP experiments. At 10 kHz MAS, 70 kHz POST-C7 pulse sequence 36 was applied on 13 C to excite double quantum coherence in 0.4 ms. Magnetisation was returned to zero quantum by another 0.4 ms of POST-C7 sequence. During DQ evolution and reconversion, 100 kHz Lee-Goldburg decoupling was applied on ¹H. During acquisition 100 kHz SPINAL-64 decoupling was applied on ¹H. The pulse sequence used was an adapted version of the Avance I large sweep width POST-C7 experiment in the Bruker library. For in vitro ECM 128-256 scans and for the heavy mouse bone samples 432 scans per t 1 -slice were recorded. 2D 13 C-13 C proton-driven spin diffusion (PDSD) correlation NMR experiment: initial cross polarisation parameters were also the same as in 13 C CP experiments. At 10 kHz MAS, the magnetisation was allowed to evolve at single-quantum coherence during the incremental delay, and returned to zero quantum coherence by a 13 C 90° pulse with a length of 3.8 μs. ¹H decoupling was switched off during the mixing period to allow transfer of 13 C magnetisation via dipolar coupling and spin diffusion 37 , with a 13 C 90° readout pulse at the end of the mixing period. During both the incremental delay and acquisition periods, SPINAL-64 decoupling was applied at 100 kHz. The pulse sequence used was an adapted version of the Avance I CP spin diffusion experiment in the Bruker library. Mixing periods of between 5 and 200 ms were recorded. For in vitro ECM 64-400 scans (depending on the amount of sample) and for the heavy mouse bone samples 256 scans per t 1 -delay were recorded. Analysis of energy landscapes. The calculations employed a standard atomistic force field, namely Amber9 with implicit solvent (igb = 2), and the FF99SB parameter set 38 . In the computational potential energy landscape approach we employ geometry optimisation techniques to calculate local minima and the transition states and pathways that connect them, to construct a kinetic transition network 26 . The low energy region of the landscape was first sampled using basin-hopping global optimisation 39,40 . These minima were then connected using double-ended transition state searches, which identify new minima and pathways, progressively augmenting the database. Observable properties were extracted using standard methods of statistical mechanics and unimolecular rate theory, and the underlying database was refined using additional connection attempts until the quantities of interest appeared to have converged. Details of these procedures can be found in recent reviews 41,42 The simulations conducted here, which seek to explain the results of the ssNMR data relating to flexibility of P X prolines adjacent to hydroxyproline residues, required parameters for the Amber9 force field obtained from experimentally derived properties of hydroxyproline with post-translational modifications 43 . The initial starting points for (POG) 12 were derived from PDB 1V7H 44,45 ; those for (PAG) 12 were generated from (POG) 12 by replacing the hydroxyprolines with alanine. For each of the two peptide sequences, (POG) 12 and (PAG) 12 , 10000 geometry perturbations were applied to the equivalent PYG group (the 6th group of the trailing chain) in the central region of each collagen triple helix. The perturbed structures were then relaxed to local minima. The 10000 resulting minima were filtered to identify the unique structures, which correspond to the space of accessible conformations for the backbone in the given PGY triplets. The unique minima were then connected in further transition state searches, which helps to ensure that all the relevant structures in the accessible configuration space were located. We ensured uniform exploration of the backbone dihedral space by randomly rotating rigid groups of atoms between carefully chosen atom pairs. For the (PAG) 12 system the pairs were: the Cβ to Cδ axis of the proline tip, the Cα-Cα axis of the peptide bond between the proline and alanine, the N-C axis of the alanine residue, and the Cα to Cα axis of the peptide bond that trailed the alanine residue. This combination of rotation groups ensures that psi and phi dihedrals of the proline, the psi and phi dihedrals of the alanine, the phi angle of the trailing glycine, and the psi angle of the preceding glycine residue are sampled. In addition, the full range of endo and exo conformations of the P X ring are encountered. For the (POG) 12 system, the corresponding groups were chosen, except that the C-N rotation of the alanine was replaced by a Cβ-Cδ rotation of the hydroxyproline tip, again inducing exploration of the endo/exo states of the O Y residue and corresponding backbone dihedrals. The uniform sampling of the full set of backbone dihedral angles can generate local minima that include peptide bond rotations and stereochemical rotations, both of which do not correspond to a biologically-feasible collagen triple helix structure. For our analysis, we are only interested in dihedral angle distributions that can be found in a collagen triple helix 46,47 . Therefore, only those structures where all the perturbed residues relax into the range of phi = −50° to −90° and psi = 130° to 170° were accepted. Sequence alignment and analysis. For our analysis of the conservation of the GPO positions, we built multiple sequence alignments across diverse species for collagen α1(I) and α2(I) sequences. For each alpha chain, a set of orthologous sequences was acquired using the National Centre for Biotechnology Information (NCBI) Protein Database resources (https://www.ncbi.nlm.nih.gov/protein) 48 . The NCBI Reference Sequence Database (RefSeq) 49 was used as the main resource for protein sequence data. A small proportion of the data was obtained from GenBank and the EMBL databases 50,51 . The initial sets of sequences identified from these databases were filtered to remove duplicates (including isoforms), leaving a single representative sequence for each species. Each set of sequences was aligned using the software program MUSCLE (http://www.drive5.com/muscle/) 52,53 . The final sequence sets used can be found in Supplementary Table S2-S4. Conservation of each GPO triplet was determined by calculating the probability of the occurrence of the GPO triplet across all aligned sequences. The amino acids with the highest frequency at each alignment position was used to generate a consensus sequence. To analyze the effect of missense mutations on GPO sites, we used the consensus sequences generated in the same procedure as described above. We mapped mutation sites reported for human variants into the consensus sequences and assessed how they align with conserved GPO positions on the collagen type I primary sequence. The National Centre for Biotechnology Information (NCBI) ClinVar database (https://www.ncbi.nlm.nih.gov/ clinvar/) was used as a mutation data resource 34 . The search for mutation sites across procollagen alpha chains was performed using the relevant gene names and the following criteria: molecular consequence: missense mutation; variation type: single nucleotide; review status: at least one star. The GPO triplet is counted as being present at a mutation site if one position of any of its residues (G, P or O) overlaps with a reported mutation site. The analysis of correlation between mutation sites and conserved GPO locations was carried out for the major fibrillar alpha chains (type I, II and III) and for collagen type V alpha chains.

chemsum

{"title": "Proline provides site-specific flexibility for in vivo collagen", "journal": "Scientific Reports - Nature"}

ovalbumin_epitope_siinfekl_self-assembles_into_a_supramolecular_hydrogel

## Abstract: Here we show that the well-known ovalbumin epitope sIINFeKL that is routinely used to stimulate ovalbumin-specific T cells and to test new vaccine adjuvants can form a stable hydrogel. We investigate properties of this hydrogel by a range of spectroscopic and imaging techniques demonstrating that the hydrogel is stabilized by self-assembly of the peptide into nanofibres via stacking of β-sheets. As peptide hydrogels are known to stimulate an immune response as adjuvants, the immunoactive properties of the sIINFeKL peptide may also originate from its propensity to self-assemble into a hydrogel. This finding requires a re-evaluation of this epitope in adjuvant testing.Fibrillar nanostructures can be assembled by a number of natural and designed peptides, which usually possess distinct biophysical properties, such as propensities for β-sheet folds. Nanofibrils can also form as a result of protein or peptide misfolding, which is associated, for example, with pathogeneses of neurodegenerative disorders such as Parkinson's and Alzheimer's diseases 1 . In biological settings, peptide nanofibrils can have important functional roles, the extracellular amyloid fibrils of E. coli, for example, are used for cell propulsion 2 . More recently, it has been possible to tune and employ the fibril-forming properties of peptides in the design and manufacture of functional nanomaterials that can be used in disease treatment and prevention 3,4 . Self-assembled peptides in this case possess several advantages, which include multi-valency, defined synthetic composition, tuned specificity, and ease of further functionalisations 5,6 . These advantages have allowed peptide nanofibrils, as well as the associated hydrogels, to be successfully used as scaffolds in regenerative medicine, cell culture matrices and vehicles for drug delivery 7,8 . Certain peptides that constitute fragments of full-length proteins have been described to form fibrils and hydrogels. These include mouse laminin a-1 9 , human transthyretin 10 and human troponin C 11 . Peptide fragments of each of these proteins assemble into nanofibrils networks in vitro, with further assembly leading to formation of stiff hydrogels.A new and exciting application of fibrillar peptide assemblies is in adjuvanting of subunit vaccines 12 . Although subunit vaccines show remarkable promise in treatment and prevention of deadly diseases, they suffer from poor immunogenicity, which substantially limits their efficacies 13 . Adjuvanting or enhancing the immunogenicities of subunit vaccines is therefore a significant challenge in biomedical research. A range of materials is currently under development as vaccine adjuvants and among them are hydrogels consisting of peptide nanofibrils. Peptides in this case can be conjugated to specific antigens or epitopes, often themselves peptides, and subsequently used in targeted immunisations [14][15][16][17] . The disease antigens and individual peptide epitopes are identified through an analysis of a given immune response against an immunogen, which lead to the individual peptide fragments that directly interact with the cells of the immune system 18 . Some of these peptide epitopes are capable of eliciting an immune response in the absence of the parent pathogen but such activity is heavily dependent on the addition of proper adjuvants 19 .During the development of adjuvants, well-known and well-characterised peptide antigens are used to evaluate the adjuvant efficacies. Ovalbumin (OVA) has been historically a popular source of such antigens, since OVA can induce both humoral and cellular immune responses based on well-characterised peptide epitopes 20,21 . The OVA 257-264 octapeptide was one of the first OVA epitopes to be characterised, it has an amino acid sequence SIINFEKL, which is recognised by cytotoxic T lymphocytes 18 . Immunisation with the adjuvanted SIINFEKL peptide induces long-lasting CD8+ T cell immunity in mice 22 . Here we report on the properties of the OVA epitope SIINFEKL to self-assemble into fibrillar nanostructures that lead to formation of a peptide hydrogel. The properties of this peptide assembly have been analysed by use of rheology, electron microscopy, small angle X-ray scattering (SAXS), circular dichroism (CD) and infrared (IR) spectroscopies. The molecular analysis of the peptide fold has been carried out with peptide nuclear magnetic resonance (NMR) measurements. It is demonstrated that SIINFEKL forms fibrillar assemblies similar to other peptide hydrogels. The immunoactive properties of this peptide can therefore be related to its self-assembling nature. ## Results and Discussion The OVA 257-264 octapeptide SIINFEKL was prepared via standard Fmoc-SPPS and purified via HPLC (Fig. SI-1). Gel formation was observable first during precipitation of this peptide in diethyl ether following cleavage with trifluoroacetic acid (TFA). The gel forming property was confirmed by preparing a 1% (w/v) solution of the purified peptide in Millipore water, where the gel formed immediately after dissolving the peptide (Fig. 1a). A hydrogel also forms with 0.5% peptide in water following overnight incubation. A typical peptide concentration used to prepare peptide-based vaccines is 8 mM and the case of SIINFEKL the concentration of 8 mM would amount to 7.7 mg/ml or 0.77% (w/v) solution, which is well within the gelation conditions. The mechanical properties of the SIINFEKL hydrogel at these two concentrations were evaluated by use of rheology after incubation at room temperature for 24 h. Constant sheer strain (γ = 1%) and angular frequency (ο = 6 rad/s) were applied in time sweep experiments. Both gels exhibited properties typical of peptide hydrogels, where elasticities (storage modulus G') are higher compared to the corresponding viscosities (loss modulus G") (Fig. 1a,b) 23 . In the case of the 1% gel, G' was more than 6 times stronger than G", which is a characteristic feature of peptide hydrogels 6 . The value of G' also indicates moderate mechanical stability. The micro-and nanoscale properties of the hydrogel material were studied by electron microscopy and small angle X-ray scattering (SAXS). The presence of fibrillar networks with fibrils on the micrometer range and widths of less than 25 nm is apparent in micrographs of the 1% hydrogel (Fig. 2). SAXS allows for measurement of nanoscale changes in density of a given sample and it is therefore a common technique in evaluation of nanoscale properties of different materials 24 . SAXS measurements of the SIINFEKL hydrogel showed an intensity increase towards small q-values (scattering vector) and a constant intensity towards higher q-values (Fig. SI-2). The data was normalised to time and beam intensity. A decrease of the total X-ray intensity was visible for each subsequent measurement of the sample. This could be caused by increased fibrillisation of the sample, which would decrease the number of scattering bodies. The radius of the fibrils was evaluated from Guinier theory, modified for long rod-like objects with circular cross-section 24,25 . Calculating a weighted mean of the three measurements resulted in a mean radius of the cylindrical www.nature.com/scientificreports www.nature.com/scientificreports/ objects of 12.3 ± 1.2 nm. The calculated radius correlates well with the diameter of the narrowest long peptide fibrils observed by TEM (22 nm, Fig. 2). The peptide conformation within the hydrogel was evaluated by CD and IR spectroscopies. Due to their specific optical activity, secondary structure of different biomolecules can be analysed by use of CD 26 . The corresponding spectra of the peptide hydrogel diluted in water show a combined profile with properties of both a random coil (slope at 200-210 nm, Fig. 3a) and a β-sheet (slope at 210-230 nm). The CD curve was fitted using a β-structure selection algorithm, which predicted 55% of the sample to be in a random coil conformation, 20% anti-parallel β-sheets, and 19% β-turns (fitted curve, Fig. 3a, RMSD = 0.0738) 26 . In conjunction with CD, FTIR is a technique that is often used for evaluation of biomolecular secondary structure. The FTIR spectrum of the SIINFEKL hydrogel was measured following chloride ion exchange in D 2 O. Although broad, a signal spectrum at 1632 cm-1, can be observed (Fig. 3b), which can also be attributed to the amide I band of anti-parallel β-sheets 27 . The precise secondary arrangement of the individual amino acids can be obtained by use of NMR 28 . NMR spectra of the SIINFEKL hydrogel were recorded from the 5% hydrogel in 1:5 D 2 O/H 2 O. Although hydrogel formation was visible, the NMR spectra did not display any line broadening (Fig. SI-6). By use of TOCSY, it was possible to assign the proton shifts corresponding to the amide, αand most of the β-hydrogens of the individual amino acids as well as the αand β-carbons. Most of the identified backbone NMR shifts fell within the range of the random coil conformation with the exception of αand β-carbons (Fig. 3c,d, respectively) of the isoleucine residues at positions 2 and 3. These shift differences from the random coil structure indicate the β-sheet arrangement for the two isoleucing residues 28 . In order to evaluate the role of the N-terminal isoleucine residues on the gelation properties of the peptide, an analogue with alanine substitutions was prepared (peptide sequence: SAANFEKL). When subjected to conditions, at which SIINFEKL peptide undewent gelation, the solution of the new analogue remained liquid even after a 72 hour incubation at room temperature (Fig. SI-3). Another analogue was prepared, this time with replacement www.nature.com/scientificreports www.nature.com/scientificreports/ of the C-terminal hydrophilic residues by alanines (peptide sequence: SIINFAAL). This analogue proved challenging to dissolve in aqueous buffers due to its increased hydrophobicity, which made it impossible to evaluate the hydrogel forming properties of this peptide. The propensity of SIINFEKL to form hydrogels has been previously briefly described by Lowenheim and co-workers 29 , who showed that the peptide conjugate can serve a scaffold for neurite outgrowth. However, no analysis of the propertied of the SIINFEKL hydrogel beyond scaffolding was carried out. In this work we show that the morphology of SIINFEKL fibrils is similar to that of other peptide hydrogels, as it consists of a dense network of apparently flexible fibrils but with a rather large diameter of 20-25 nm as opposed to 2-5 nm usually observed for peptide fibrils, as for example for the troponin C peptide fragment VEQLTEEQKNEFKAAFDIFVLGA 11 . TEM micrographs show that the individual SIINFEKL nanofibres further assemble into micro-scale fibres with distinct branched structures. The size and shape of the nanofibres were confirmed by SAXS measurements. Assembly into fibrils is often facilitated by a distinct fold adapted by a portion of the monomer peptide, where β-sheets are by far the most common motifs 6 . Under our experimental conditions the SIINFEKL peptide also adopts a partial β-sheet fold, which is in contrast to the native fold of this peptide within the context of crystalline ovalbumin, where the termini of SIINFEKL are involved in two separate α-helices, the center of the peptide is in a random coil, and no portion of the peptide forms a β-sheet 30 . Our NMR data indicates that only the two N-terminal isoleucine residues of the peptide are not in a random coil conformation. Due to their side-chain hydrophobicity, isoleucine residues are known to facilitate self-assembly of short peptides, as is the case for the peptide IIIK that forms stable nanofibers in aqueous solution via β-sheet formation combined with molecular amphiphilicity 31 . The continued assembly of IIIK leads to formation of a soft peptide hydrogel. The fibrillar structure of SIINFEKL assemblies is also likely to be facilitated via such amphilicity, as the N-terminal half of the peptide contains two hydrophobic residues (IleIle), whereas the C-terminal half contains two adjacent hydrophilic amino acids (GluLys). ## Conclusion As the peptide SIINFEKL is a very well-characterised epitope, it is often used in proof-of-principle studies aimed at demonstrating the efficacies of new adjuvant systems. It has been employed in evaluation of, among others, monophosphoryl lipid A 32 , bacterial membrane vesicles 33 and HMGB1 peptide 34 . SIINFEKL has also been used to study the immune response to the Bacillus Calmette-Guérin vaccine, which is commonly used against tuberculosis 35 . It is now, however, well-known that peptide hydrogels consisting of nanofibres can themselves act as adjuvants 14,36 , and therefore, our results should caution the use of the SIINFEKL peptide during evaluation of adjuvant efficacies because its ability to form fibrils and a hydrogel. ## Methods peptide synthesis and gel formation. Solid phase peptide synthesis (SPPS) was done using fluorenylmethoxycarbonyl (Fmoc) chemistry on 100 µmol scale. Peptides were deprotected and cleaved in a mixture of trifluoroacetic acid, triisopropylsilane and water (38: 1: 1) for 2 hours at room temperature. The peptide was precipitated with diethyl ether and centrifuged. After washing twice with ether, the precipitated peptide was dissolved in water and lyophilised. The peptide was purified by RP-HPLC using a semi-preparative Kromasil C18 column. Analysis of the purified peptide SIINFEKL (38 mg, 39% yield), was carried out using an analytical Kromasil C4 column with a gradient of 5% buffer B (acetonitrile + 0.08% TFA) in buffer A (water + 0.1% TFA) to 65% B in A over 20 min at 1 mL/min and UV measurement at 214 nm (Fig. SI-1). Peptides SAANFEKL (Fig. SI-4) and SIINFAKL (Fig. SI-5) were synthesised and analysed in a similar manner. Hydrogel formation was monitored visually as a function of pH, peptide and buffer concentration (Table SI-2). In the standard gelation procedure, water was added to the peptide and the resulting mixture mixed until the sample assumed a homogeneous consistency. Gelation was apparent minutes after dissolving, with final pH 4.2. The pH was then adjusted to 7 by adding aliquots of an appropriate NaOH aqueous solution and the volume was adjusted to give the target final concentration. The solutions were allowed to stand at room temperature overnight. Similar experiments were carried out at a fixed peptide concentration of 1 wt% and samples were incubated at different temperatures. ## Rheology. The rheological measurements were performed on a stress-controlled rheometer (TA Instruments HR2) fitted with a 50 mm diameter plate geometry, with a gap of 0.2 mm. The sample was allowed to anneal at 25 °C for 1 hr prior to time-sweep experiments in linear regime for both strain and frequency as discussed in the manuscript. ## Microscopy. For scanning electron microscopy (SEM), peptide hydrogel was applied to a Thermanox ™ coverslip and air dried. The coverslips were sputter coated with gold in high vacuum (Bal-Tec SCD 005). SEM images were recorded with Zeiss SEM Supra 55 VP operating at 20 kV. For transmission electron microscopy (TEM), peptide hydrogel was applied onto carbon coated copper grid and subsequently viewed with Philips CM200 at 200 kV. TEM images were acquired with OriusTM SC600 Gatan CCD camera. Circular Dichroism. CD spectra were recorded on peptide solutions at 0.5% by weight in water using a 1 mm quartz cuvette. Solutions had been incubated at room temperature for a minimum of 24 h. Scans were performed in 1 nm increments with 3 s scans at 20 °C and averaged over 5 scans. FtIR. FTIR-ATR was recorded on peptide solutions at 1% by weight in D2O using Agilent Cary 630 with a single bounce diamond ATR-cell and potassium-bromide optics. The spectral resolution was set to 2 cm -1 , with zero filling factor 2, which resulted in a formal resolution of 0.46892 cm -1 . (2019) 9:2696 | https://doi.org/10.1038/s41598-019-39148-8 www.nature.com/scientificreports www.nature.com/scientificreports/ sAXs. SAXS measurements were performed using X-rays from a Nanostar (Bruker AXS) system, operating at λ = 0.1542 nm (CuKα-radiation) and equipped with a two-dimensional detector (Våntec 2000). X-ray patterns were radially integrated to obtain the scattering intensity of the peptide in dependence on the scattering vector q = (4π/λ) sinθ, with 2θ being the scattering vector. The samples were filled into capillaries and the solvent sample was subtracted as background from the peptide sample dissolved in the solvent. Sample measurement time was 3, 6 and 12 hours. NMR. NMR spectra were acquired with a 700 MHz Bruker Avance III HD NMR spectrometer on peptide sample dissolved in 20% D 2 O in H 2 O at 5 mg/ml. For H-H Total Correlation Spectroscopy (TOCSY) 16 transients were collected using 16 dummy scans with spectral width of 10 ppm in both dimensions.

chemsum

{"title": "Ovalbumin Epitope SIINFEKL Self-Assembles into a Supramolecular Hydrogel", "journal": "Scientific Reports - Nature"}

improved_photodecarboxylation_properties_in_zinc_photocages_constructed_using_m-nitrophenylacetic_ac

## Abstract: The methoxy-and fluoro-derivatives of meta-nitrophenylacetic acid (mNPA) chromophores undergo photodecarboxylation with comparable quantum yields () to unsubstituted mNPA, but uncage at red-shifted excitation wavelengths. This observation prompted us to investigate DPAdeCageOMe (2-[bis(pyridin-2ylmethyl)amino]-2-(4-methoxy-3-nitrophenyl)acetic acid) and DPAdeCageF (2-[bis(pyridin-2-ylmethyl)amino]-2-(4-fluoro-3nitrophenyl)acetic acid) as Zn 2+ photocages. DPAdeCageOMe has a high  and exhibits other photophysical properties comparable to XDPAdeCage ({bis[(2-pyridyl)methyl]amino}(9-oxo-2-xanthenyl) acetic acid), the best perforiming Zn 2+ photocage reported to date.Since the synthesis of DPAdeCageOMe is more straightforward than XDPACage, the new photocage will be a highly competitive tool for biological applications. We are developing decarboxylation reactions of metanitrophenylacetatic acid (mNPA) and xanthone (XAN) groups to design photocages that release Zn 2+ upon irradiation with light. 1,2 Such photocaged complexes block the biological activity of metal ions until the Zn 2+ release is initiated by exposure to light of a specific wavelength. 3 We recently described DPAdeCage (1) and XDPAdeCage (2), two Zn 2+ photocages that incorporate a dipyridyl amine (DPA) chelating group to selectively bind Zn 2+ over other common loosely bound metal ions found in cells (Error! Reference source not found.). 2 Irradiation at or near the max initiates a photoreaction that results in the loss of the carboxylic acid functional group as CO2, which shifts the Zn 2+ binding equilibrium toward unbound metal ion (Scheme 1). The transformation from a 4-coordinate to a 3coordinate chelator leads to a 10 5 -fold reduction in binding affinity (Kd). For biological applications, a rapid synthesis to produce photocages in high yields would be ideal. The photophysics of XAN chromophores are superior to those of mNPA groups, but the low-yielding synthesis of the XAN scaffolding limits the pace of photocage development. To overcome these limitations, we functionalized several mNPA chromophores to develop DPAdeCage derivatives with improved photophysical properties. In addition to those improvements, the added functional groups facilitate selective nitration at the 3-position of the aromatic ring simplifying the purification procedure for the mNPA chromophore that will facilitate the rapid generation of photocages. Designing effective photocages requires avoiding short wavelength excitation (max), which can induce cellular photodamage, while still delivering sufficient energy to break chemical bonds. Compared to our mNPA-based DPAdeCage, the max of XAN-derived XDPAdeCage is red-shifted by over 80 nm (Error! Reference source not found.). Furthermore, upon coordination to Zn 2+ the photoreactivity of DPAdeCage is significantly reduced. The XAN chromophore offers significant photophysical advantages that include maintaining a high  upon Zn 2+ binding. Accessing the XAN chromophore necessary to make XDPAdeCage requires a 2-step synthesis with a maximum 20% yield. By comparison, the nitration step to synthesize the mNPA chromophore has a much higher yield, increasing the total material obtained (~70% yield of mNPA). The difference in reaction time, 24 h and 1 h respectively, also impacts the rate of photocage synthesis. We hypothesize that the relative ease of making mNPA derivatives compared to the XAN chromophore, as well as the numerous readily available starting materials, will facilitate the rapid production of mNPAbased photocages for numerous applications. Furthermore, if functionalization of the mNPA group could shift the abs sufficiently, we can prepare and screen Zn 2+ photocages for biological applications more quickly as well as expand the toolbox of tunable chromophores. Measurements obtained in 40 mM HEPES buffer (pH 7.5, 100 mM KCl). Quantum yields were obtained in solutions containing 30% methanol to maintain photoproduct solubility. a. Hammet constants indicating the relative electron withdrawing or donating properties of each R group. b. 3e lacks a red-shifted absorbance band; = 3264 M -1 cm -1 at 300, 356 M -1 cm -1 at350. In previous studies on ortho-nitrobenzyl photocages (oNB), introducing electron donating groups (EDGs) or electron withdrawing groups (EWG) produced photocages with different rates of photolysis, quantum yields and excitation wavelengths. Tsien, Kaplan, and others reported functionalization of the oNB ring with methoxy groups shifted the max to longer wavelengths, 7,8 suggesting we might benefit from a similar approach. Modification of the nitrobenzyl ring, however, can also impact the electronics of the aromatic ring and therefore the nature of the photoreaction. 10,12 Unlike oNBs that undergo a Norrish type II photoreaction, 13,14 the mechanism for photodecarboxylation of mNPA chromophores involves an electron transfer from the nitro group to the methylene bridge of the phenylacetate to initiate the photodecarboxylation. This suggests increasing the electron-density around the aromatic ring would impede electron-transfer to the methylene bridge, which could reduce photolysis quantum yields (). Thus, we predicted EDGs would exhibit reduced photoactivity compared to mNPAs functionalized with EWGs. By adjusting the electronic-structure of the mNPA chromophore we hoped to access a DPAdeCagederivative that will maintain a high  upon coordination to Zn 2+ . A series of mNPA chromophores were prepared with various EDGs and EWGs at the 4-position of the aromatic ring (Error! Reference source not found.). We hypothesized that functional groups with heteroatoms would extend the conjugated system and red-shift the max. Functionalization resulted in the appearance of a new red-shifted absorbance feature that was not observed in the parent mNPA 3e (Figure 2). The red-shifted feature appears as a shoulder of the strong UV band centred at circa 270 nm in most of the compounds, but the band shifts sufficiently in the methoxy-(3f) and hydroxy-derivatives (3g) to appear as a distinct new absorbance feature. While all the derivatives possess a weaker extinction coefficient () of the large absorbance band below 300 nm compared to 3e, the compounds also absorb more strongly at wavelengths greater than 320 nm (Error! Reference source not found.). The photolysis quantum yields () were determined under simulated physiological conditions using LC/MS to monitor the loss of 3a-3i. Despite the red-shifted band, the photoactivity of the acetyl (3a), chloro (3b), and amide-functionalized (3d) compounds is diminished relative to 3e. The only EWGcontaining derivative that exhibits a  similar to 3e is the fluoroderivative 3c. Although 3b has a similar chloro group, the observed photoreactivity is diminished relative to 3c. Notably the absorptivity of 3c is stronger than 3b across the entire spectrum and likely contributes to the observed difference in (Error! Reference source not found.). While 3c exhibits photoactivity comparable to XDPAdeCage, the max is blueshifted by over 30 nm. The mass peaks observed in the LC/MS spectrum of 3a-f after photolysis indicate tolyl and aldehyde derivatives analogous to those previously observed with mNPA-based photocages predominate the photoproducts (Scheme 1). 1,2 Compounds containing EDGs (3f-3i) exhibit a bathochromic shift in max compared to the parent compound 3e; however only 3f retained photodecarboxylation activity. The hydroxy compound 3g exhibits a higher (2989 M -1 , cm -1 ) and a more red-shifted absorbance (max = 427 nm) but lacks photoactivity compared to the corresponding methoxy derivative (3f). Excited state proton transfer does not appear to be responsible for the inactivity of 3g as the compound remains photochemically inert under strongly basic conditions (pH 14). Although 3f contains an electron-donating methoxy group, the photodecarboxylation and red-shifted max is the most comparable to XDPAdeCage. While the methoxy functional group is electron-donating by resonance, inductive effects provide the most plausible explanation for the anomalous behaviour in the EDG series. The model methoxy-and fluoro-compounds 3f and 3c exhibited the most comparable photophysical properties to XDPAdeCage, therefore we chose these chromophores to build new DPAdeCage derivatives (Error! Reference source not found.). Our initial attempts to prepare these photocages via our established route were unsuccessful due to oxidative decomposition during the final nitration reaction. Alternatively, starting with 3f and 3c provided 4 (DPAdeCageOMe) and 5 (DPAdeCageF) in 11.6% (5 steps) and 11.8% (4 steps) respectively. The only significant difference between the two synthetic pathways is the final ester hydrolysis. DPAdeCageF required acidic conditions at elevated temperature, while the basic conditions used to prepare DPAdeCage successfully provided DPAdeCageOMe. A combination of product decomposition in the harsher deprotection reaction as well as the more difficult isolation of DPAdeCageF from acidic solution, reduced the overall yield of the final product. A comparison of the synthetic pathways required to develop the new mNPA photocages and XDPAdeCage reveals both DPAdeCageOMe and DPAdeCageF can be more rapidly prepared and in greater amounts than XDPAdeCage. The photophysical properties of DPAdeCageOMe and DPAdeCageF are nearly identical to the parent compounds 3f and 3c under aqueous conditions (Error! Reference source not found.). Notably, the max of both DPAdeCageOMe and DPAdeCageF are red-shifted relative to DPAdeCage. The methoxy-functionalized DPAdeCageOMe (342 nm) has a max nearly identical to XDPAdeCage (347 nm). The addition of Zn 2+ to the photocages results in a slight hypsochromic shift of max, but relatively little change in absorptivity (Table 1). An analysis of the LC/MS spectrum of DPAdeCageOMe and DPAdeCageF following irradiation, indicate successful photodecarboxylation following irradiation. In the absence of Zn 2+ , both compounds appear to successfully release the carboxylate group based on the change in mass. In DPAdeCage, 2 a decrease in  is observed upon coordination to Zn 2+ and DPAdeCageF exhibits an even more dramatic decrease upon Zn 2+ coordination. Only extended irradiation times (>30 min) of [Zn(DPAdeCageF)] + resulted in evidence of photolysis, which severely limits the potential applications of the photocage. The addition of Zn 2+ does not appear to affect the resulting photoreaction of DPAdeCageOMe; however, DPAdeCageF forms a wider slate of photoproducts in the presence of Zn 2+ after long irradiation. Some higher mass, emissive photoproducts suggest that Zn 2+ facilitates the formation of coupled DPAdeCageF products TD-DFT calculations were performed to probe if the electronic structure contributed to the observed difference in photoreaction of [Zn(DPAdeCageOMe)] + and [Zn(DPAdeCageF)] + . The structure of DPAdeCageOMe, DPAdeCageF, [Zn(DPAdeCageOMe)] + and [Zn(DPAdeCageF)] + were optimized using DFT and the frontier molecular orbitals contributing to the electronic transitions found in the TD-DFT calculations were visualized. In all cases, the lowest energy excitation involves excitation from the HOMO to the LUMO. The LUMOs are localized around the nitro group and phenyl ring for all compounds, consistent with previous studies of mNPA chromophores (Error! Reference source not found.). 17 We believe that promoting an electron into a LUMO localized on the nitro group is necessary for photolysis to occur, suggesting [Zn(DPAdeCageF)] + should be capable of undergoing photodecarboxylation based on the calculated electronic structure. A closer examination of the observed oscillator strengths (f) representing the molar absorptivity of Zn(DPAdeCageOMe) + (f = 0.04) and Zn(DPAdeCageF) + (f = 0.01) reveal a nearly 4-fold decrease in the oscillator strength of the lowest energy transition for Zn(DPAdeCageF) + ,which may explain a decrease in observed photolysis, although other factors probably contribute to the difference given the large drop in  observed experimentally. Through a study of model mNPA chromophores, we were able to identify derivatives with comparable photophysical properties compared to the decarboxylation reaction observed with XAN groups. Specifically, fluoro-and methoxy-groups introduce a distint red-shifted absorption band; however, only the methoxy derivative retains sufficient photodecarboxylation activity when integrated into a Zn 2+ photocage. We were pleased to see DPAdeCageOMe maintains a high  when coordinated to Zn 2+ and has nearly identical photocaging properties to XDPAdeCage. A methoxy group in the 4-position also provides a potential site for modification of future photocages through an ether linkage that should not impact Zn 2+ binding significantly and will be the subject of future investigations. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Improved photodecarboxylation properties in zinc photocages constructed using m-nitrophenylacetic acid variants", "journal": "ChemRxiv"}

multivalent_and_multifunctional_polysaccharide-based_particles_for_controlled_receptor_recognition

## Abstract: Polysaccharides represent a versatile class of building blocks that are used in macromolecular design. By choosing the appropriate saccharide block, various physico-chemical and biological properties can be introduced both at the level of the polymer chains and the resulting self-assembled nanostructures. Here, we synthetized amphiphilic diblock copolymers combining a hydrophobic and helical poly(γbenzyl-L-glutamate) PBLG and two polysaccharides, namely hyaluronic acid (HA) and laminarin (LAM). The copolymers could self-assemble to form particles in water by nanoprecipitation. In addition, hybrid particles containing both HA and LAM in different ratios were obtained by co-nanoprecipitation of the two copolymers. By controlling the self-assembly process, five particle samples with different morphologies and compositions were developed. The interaction between the particles and biologically relevant proteins for HA and LAM, namely CD44 and Dectin-1 respectively, was evaluated by surface plasmon resonance (SPR). We demonstrated that the particle-protein interaction could be modulated by the particle structure and composition. It is therefore suggested that this method based on nanoprecipitation is a practical and versatile way to obtain particles with controllable interactions with proteins, hence with the appropriate biological properties for biomedical applications such as drug delivery.Polysaccharides represent an important class of polymers for the design of functional biomaterials, especially towards biomedical 1-4 and cosmetic applications [5][6][7] . Compared to synthetic polymers, they provide better biocompatibility, biodegradability, and also diverse bioactivities depending on their structures [8][9][10][11][12] . Furthermore, as bio-sourced polymers, polysaccharides are fully renewable ingredients, which make them particularly relevant in the context of green chemistry [13][14][15] . In addition, their production can be completed in an eco-friendly way with minimal environmental impact, which perfectly matches the criteria of sustainable development for industry.Among all the available polysaccharides, hyaluronic acid (HA), a non-sulfated glycosaminoglycan (GAG), is composed of alternative units of D-glucuronic acid and N-acetyl glucosamine linked by β-1,3 and β-1,4 glycosidic bonds. HA is one of the main components of the extracellular matrix (ECM), and as such, is highly abundant in the ECM-rich tissues such as synovial fluid, vitreous body of eyes or dermis 16,17 . With a highly hydrophilic structure, HA plays important roles for water retention in skin and eyes 18,19 . In addition, the molecular weight of native HA being large (>1000KDa), the prepared solutions are generally very viscous with a shear-thinning character 20 . These properties make HA the unique space-filling material in tissues to maintain their morphology and homeostasis 21 . This is also the reason why HA is widely used in the dermal fillers for plastic surgery 5 .In addition to the functions related to its physicochemical properties, HA has numerous biological effects, especially via its specific interaction with HA binding proteins (hyaladherins) 22 , such as CD44, RHAMM, Stabilin-2, LYVE-1 and aggrecan. The interaction between HA and CD44 has been widely explored, mainly because CD44 is a glycoprotein expressed at the surface of most cell types including skin keratinocytes 23 and fibroblasts 24 , and is involved in a number of signal transduction pathways 25 . By interacting with CD44, HA exhibits activities in biological processes such as cell proliferation/migration 26 , wound healing 27 and tissue regeneration 28 . Compared to other hyaladherins, CD44 attracts researchers' high interest due to its involvement in cancer 29 . Overexpression of CD44 can be observed in cancer cells including breast, pancreas, gastric, prostate, ovarian and colon [30][31][32][33][34][35] , making CD44 + a biomarker of cancer cells 36 . Targeting overexpressed CD44 in cancer is becoming one of the important strategies in cancer therapy 37 . As the main ligand of CD44, HA is now increasingly used in the nanomaterial design for drug delivery 38 . It has been demonstrated that the interaction of HA with targeted proteins and the relevant effects depends on its molecular weight. Indeed, native HA with high molecular weight in ECM contributes to skin integrity and prevents inflammation and angiogenesis, whereas fragmented HA is generated and involved in a range of immunological processes during the tissue injury and inflammation . LAM is another important oligosaccharide that may present relevant biological activity. Indeed, as a β-D-glucan, LAM is a ligand of dectin-1 42 , a pattern-recognition receptor of immune system involved in immune response initiation during fungal infection 43,44 . By interacting with dectin-1, LAM can be used to modulate immune system and bring biological activities including immunostimulatory and antitumorous effects 45,46 . In this study, particles were designed by self-assembly of polysaccharide-b-polypeptide block copolymers. Based on the synthesis strategy previously established in our group 47,48 , two copolymers, based on HA and LAM, were obtained. We further developed a nanoprecipitation process to obtain a range of particles with different morphologies and compositions, including particles containing only HA or LAM and hybrid particles containing both of them in different ratios. The interaction with CD44 was measured by using surface plasmon resonance (SPR). We investigated the interaction of the particles with CD44 in comparison with linear HA of different molecular weights and CD44. We demonstrated that our particles exhibit an enhanced interaction with CD44 compared to linear HA, and that the particle-CD44 interactions could be modulated by changing the ratio of HA in the hybrid particle. Furthermore, a practical method was developed based on SPR to observe the dual functionality of the hybrid particles and their ability to associate simultaneously with CD44 and Dectin-1. Altogether, our experiments demonstrated that our unique copolymer design and self-assembly approach allowed both for an increase of ligand efficacy and dual functionality of particles. ## Result and Discussion Polysaccharide-b-polypeptide block copolymer synthesis. A three-step synthesis based on copper-catalyzed Huisgen cycloaddition "click chemistry", previously reported by our group, was developed to obtain the polysaccharide-b-polypeptide di-block copolymer 47 . A poly(γ-benzyl-L-glutamate) with azide functional group (PBLG-N 3 ) was obtained by the ring-opening polymerization of the corresponding N-carboxyanhydride (NCA) initiated by 3-azido-1-propanamine primary amine (Fig. 1a). This controlled and living polymerization process led to the formation of PBLG-N 3 with a degree of polymerization DP = 30 (corresponding to 6.6 kDa) and low dispersity (Đ = 1.1). Propargyl functionality was introduced to the reducing end of both HA (5 kDa) and LAM (5 kDa) by reductive amination with propargylamine (Fig. 1b). Besides HA 49 and LAM 50 , other polysaccharides such as xylan, pullulan, dextran or galactan can also be functionalized in the same way . The two building Chemical modifications of polysaccharides are often introduced on the lateral functional groups on their backbones such as -OH and -COOH. However, the modification ratio can only be controlled statistically and consequently brings additional variability to the resulting product. Furthermore, the biological activities of polysaccharides can be significantly altered by such side chain modifications 56,57 . We can notice that the structure of HA and LAM in our block copolymer is nearly intact compared to its natural structure. Each polysaccharide chain is modified once on the reducing end and the full conversion can be confirmed by 1 H-NMR. In this way, the chemical structure of the copolymer can be precisely controlled. The resulting nanostructures based on these copolymers can also maintain the properties of HA and LAM to associate with their target biological receptors, respectively CD44 and dectin-1, as shown later in the SPR study. Amphiphilic copolymer self-assembly and particle formation. The two synthesized amphiphilic copolymers HA-b-PBLG and LAM-b-PBLG were self-assembled in water by using a nanoprecipitation approach, adapted from previous reports 47,58 . First dissolved in DMSO, a good solvent for both blocks, the copolymer solution was then diluted with water. The self-assembly was driven by the insolubility of the hydrophobic PBLG segments in water, that formed the particle core during solvent-displacement. Covalently attached to the PBLG in the copolymer structure, the hydrophilic polysaccharide moieties remained at the particle surface. In this way, we obtained stable self-assembled particles in water. The particle morphology could be modulated by process parameters such as mixing rate and water phase composition 59 . In our study, two process protocols, named fast nanoprecipitation and slow nanoprecipitation, have been optimized, as detailed in the Method section. Three monofunctional particle samples were thus obtained: HA-b-PBLG-30nm, HA-b-PBLG-150nm, LAM-b-PBLG-30nm as listed in Fig. 2. The two copolymers have similar molecular structures, with the same ratio between the hydrophilic polysaccharide block and the hydrophobic polypeptide block. The resulting particles obtained by fast nanoprecipitation had consistently very similar sizes between 30 nm and 50 nm. In this study, we have identified the experimental conditions to co-nanoprecipitate HA-b-PBLG and LAM-b-PBLG copolymers together from their mix solution in DMSO during a fast nanoprecipitation process. Indeed, in these conditions, the particles were rapidly formed and instantly "frozen" during the solvent displacement. Such an out-of-equilibrium process favors the homogeneous mixture of the two copolymers in the particles, since the potential phase separation between the copolymers is minimized. The resulting hybrid particles were stable in water, containing both HA and LAM moieties. By changing the initial concentration of each copolymer from the DMSO solution, the composition of the particle can be continuously modulated. In the present study, after extensive dialysis against pure water in order to remove any trace of DMSO, particles with the same size range (around 40 nm), with relatively low polydispersity (PDI < 0.2) and negative zeta-potential (ξ ranging from −38 to −29 mV), but with composition of 90 wt% and 50 wt% of laminarin were obtained, and named Hybrid-nano 90%LAM and Hybrid-nano 50%LAM respectively (Fig. 2). All the relevant characteristics of the developed particles that will be further studied are summarized in Fig. 2. ## Interaction of HA free chains and HA-based particles with CD44 observed by SPR. Previously reported studies exhibited the capability of HA conjugates and particles functionalized with HA to recognize CD44 receptors . In our group, polymer vesicles based on HA-b-PBLG have been proven efficient to target the overexpressed CD44 in cancer cells and deliver the loaded actives to limit cancer progression . In the present study, the interaction between CD44 and HA-based compounds (HA of different molecular weights and the HA-containing particles) was investigated in detail by surface plasmon resonance (SPR). All the analyses were performed in non-saturated conditions, meaning that the CD44 receptors were never fully occupied by the sample. In the block copolymer, the molecular weight of the polysaccharide moiety was 5 kDa and that of polypeptide is 6.6 kDa. As a result, the copolymer and the particles formed by the copolymer, contained only 43 wt% polysaccharide, which is the active moiety to interact with the receptors. In a SPR analysis, the binding signal represents the entire mass of compounds attached to the bioreceptor-functionalized surface 68 . In case of particle-receptor interaction study by SPR, the polypeptide (PBLG) blocks, forming the particle cores, contributed to the SPR signal once the particle was attached to the surface, even they were not involved in the interaction with the receptors. In order to subtract the contribution of PBLG, the SPR signal of the particles was normalized by multiplying by 0.43, the polysaccharide weight ratio in the particle as shown in Fig. 3a. The resulting normalized signal revealed uniquely the quantity of bioactive polysaccharide involved in the interaction. However, the mass percentage of the particle samples (23ppm) in the tests was larger than that of HA samples (10ppm), so that all the samples contained the same quantity of active moiety for the interaction, which was the polysaccharide (Fig. 3b). It is important to notice that all the comparisons in this study were performed in these "double normalized conditions", meaning that the samples that we used contained the same concentration of HA, and we compare the quantity of normalized SPR signals revealing the quantity of HA associated with the surface by CD44. As shown in Fig. 3b, the interaction between HA and CD44 increases with the molecular weight of HA, observed by enhanced SPR binding signals. The SPR signal of a high molecular weight HA (1000 kDa) was almost 5-fold stronger than that of a small molecular weight HA (5 kDa). These observations are consistent with previous studies from Dan Peer's group, also obtained by SPR 69 , and that of Ritchter and coll. through quartz crystal microbalance (QCM-D) analysis 70 . To synthesize the HA-b-PBLG copolymer, a low molecular weight HA (5 kDa) was used and consequently the resulting particles were covered with the same units of HA-5kDa. In SPR analysis (Fig. 3b), the particle sample showed a much stronger interaction with CD44 compared to a HA-5kDa sample with the same HA concentration. Even with slower kinetics, the binding signal of the particle sample was even higher than that of the highest molecular weight HA samples (100 kDa and 1000 kDa). With the same HA-b-PBLG copolymer, the particle interaction with CD44 could be modulated by its morphology, and therefore by the formulation process. Indeed, as observed in Fig. 3c, the SPR signal of larger HA-particles (150 nm) was significantly higher than that of those of smaller diameters (30 nm). These observations are consistent with recent in vitro experiments performed on different lung cancer cell lines expressing different level of CD44 71 . Richter's lab also recently observed the multivalent interaction between CD44 and high molecular weight HA by a multivalent interaction model 72 . They especially demonstrated by single molecule force spectroscopy that CD44/HA bonds have a high tensile strength despite their low affinity, and that multiple bonds along an HA chain rupture independently under load. Our experiments suggest that this interaction enhancement with the HA particle performed with the same principle. As schematically illustrated in Fig. 3d, a free HA-5kDa moiety in the solution can probably associate with only one or a very low number of CD44, whereas the HA particles were able to bind simultaneously a larger number of CD44. These multivalent interactions strengthen the interfacial force, and hence the binding signal in SPR analysis. This was also the reason why we were unable to measure the association and the dissociation constants. Indeed, to obtain these constants by curve fitting algorithms, it is essential to know the precise interaction model (1:1, 1:2 etc.) between the receptors and the analytes in the sample, which is very challenging with any particle systems. Interaction modulation by controlling the particle composition. The interaction of the three HA-containing particles with CD44 was further compared with different HA content toward LAM (meaning HA-b-PBLG-30nm, Hybrid-nano 50%LAM and Hybrid-nano 90%LAM). All of them presented significant interaction signals with CD44, but the binding level was reduced with the decreasing ratio of HA in the particle as shown in Fig. 4. The three different particle samples were obtained by the same fast nanoprecipitation with similar sizes (Fig. 4), so that they were able to get in contact with a similar quantity of CD44 on the SPR surface. However, LAM was unable to associate with CD44. By introducing LAM on the particle surface, the density of HA on each particle was reduced, and so was the multivalency degree of the interaction with CD44 as well. As a consequence, the interaction strength with CD44 can be weakened by this dilution and modulated by the 'dilution' factor, which explains the SPR signal reduction in Fig. 4. The dual functionality of the hybrid particle confirmed by SPR. The hybrid particles were obtained by co-nanoprecipitating the both HA-b-PBLG and LAM-b-PBLG copolymers from a common solution. By formulating polysaccharide moieties in a nanomaterial, it is possible to modulate the bioactivities of the resulting structure such as biological process regulation and inflammatory responses 73 . However, it was important to confirm that the two copolymers self-assemble together in our process and the resulting particle sample was not a mixture of the monofunctional particles. For this purpose, a two-step assay based on SPR was designed to check the coexistence of HA and LAM on the hybrid particle surface, hence its dual functionality. As illustrated in Fig. 5a, the CD44 surface in this assay can interact with the HA moieties and capture the hybrid particle in the first step. A short natural dissociation step was applied to confirm the stable interaction between the sample and the CD44 surface. Then, a second injection was directly performed with dectin-1, a specific biological receptor of LAM 44 . Dectin-1 can associate with LAM units on the hybrid particle but not with HA, and form the sandwich-like structure shown in Fig. 5a. The adsorption of dectin-1 at the surface of the particles can generate a positive binding signal during the injection. As shown in Fig. 5b,c, both hybrid particle samples gave positive binding signals as expected during the two injection steps. Meanwhile as a negative control, the same assay was applied to a mixture of monofunctional particles based on HA and LAM (Fig. 5d). The CD44 surface captured the HA particles and gave a binding signal during the first injection. However, no binding signal was observed during the second injection of Dectin-1, since no LAM was present on the surface of the HA particles. These observations really proved that (i) only particles with HA can bind to CD44 and that there is no unspecific adsorption of LAM based particles, and (ii) the LAM polysaccharide chains present at the surface of hybrid particles are still available for interaction with Dectin-1. By increasing the ratio of LAM on the particle surface, the interaction signal with CD44 decreased whereas that with dectin-1 increased. Through this assay, the dual functionality of the hybrid particles to interact simultaneously with CD44 and dectin-1 is confirmed and thus the concept of interaction modulation by changing the particle composition. In summary, we have reported a versatile strategy to design particles with tunable interaction with targeting proteins by the combination of different amphiphilic copolymers in a nanoprecipitation process. Using two different polysaccharide-b-polypeptide copolymers, namely HA-b-PBLG and LAM-b-PBLG, a broad range of particles was obtained by changing the formulation process parameters. The control of the nanoprecipitation and co-nanoprecipitation processes allows the design of particles with controlled sizes and compositions in an accurate and reproducible manner. As demonstrated using SPR as a fast and efficient screening method, the interaction of the HA-particle with CD44 was far stronger than that of linear HA on its surface, which can be explained by a multivalent interaction between HA segments on the particle surface and CD44. The "multivalent-degree" can be controlled by changing the particle morphology and composition, hence leading to the modulation of the interaction strength. The combination of the two copolymers results in the formation of hybrid particles with the functionalities from both HA and LAM, as confirmed by an original method based on SPR analysis. All these observations and experimental results strongly suggest that such a simple co-nanoprecipitation process can be a versatile and practical way to design multifunctional particles with tunable biological activities. ## Materials Hyaluronic acid (HA) sodium salt of different molecular weights (research grade, HA-5kDa, HA-20kDa, HA-100kDa, HA-1000kDa) was purchased from LifeCore Biomedical (Cheska, MN, USA). Laminarin from Laminaria Digitata and all the chemicals used in the copolymers synthesis were purchased from Sigma Aldrich (St. Louis, MO, USA). Recombinant human CD44 and dectin-1 were purchased from R&D system (Minneapolis, MN, USA). CM5 chip and all the solvents and reagents for SPR analysis including the running buffer, N-hydroxysuccinimide (NHS), ethyl-3(3-dimethylamino)propylcarbodiimide (EDC), 1 M ethanolamine solution pH 8.5, the regeneration solution were purchased from GE Healthcare (Uppsala, Sweden) and used as suggested. Ultrafiltration discs were purchased from EMD Millipore (Billerica, MA, USA). ## Measurements 1 H-NMR spectra were obtained with a Bruker Avance 400 MHz spectrometer (Rheinstetten, Germany). Dynamic light scattering was measured Malvern Zetasizer NANO ZS (Worcestershire, UK). Surface plasmon resonance analysis was performed with Biacore T200 (Uppsala, Sweden). IR spectra were recorded on Perkin Elmer Spectrum One FT-IR (Shelton, CA, USA). ## Methods Synthesis of HA-b-PBLG and LAM-b-PBLG. HA-b-PBLG was prepared as reported elsewhere 47 , and LAM-b-PBLG was prepared using the same synthetic approach. The polysaccharide blocks have a 5 kDa molecular weight and the same PBLG of 6.6 kDa was used for the click reactions. The copolymers were analyzed by Particle preparation by nanoprecipitation-induced self-assembly. As mentioned in Fig. 2, the experimental conditions of two processes, fast and slow nanoprecipitation, have been optimized as reported below: Fast nanoprecipitation: 9 ml PBS buffer (10 mM, pH = 7.4, 154 mM ionic strength) was heated to 50 °C and stirred at 500 rpm by a magnetic rotor. 1 ml copolymer solution in DMSO (1 wt%), previously heated to 50 °C, was added dropwise to the PBS buffer. The resulting solution was further stirred at 50 °C for 30 min then cooled down to room temperature. DMSO was removed by ultrafiltration with the PBS buffer against a MWCO = 100 kDa filter. The particle size was measured by DLS with a scattering angle at 173°. Slow nanoprecipitation: 1 ml copolymer solution in DMSO (1 wt%) was maintained at 60 °C and stirred at 500 rpm. 9 ml PBS buffer (10 mM, pH = 7.4, 154 mM ionic strength) was added dropwise for 400 seconds by a syringe pump. The resulting solution was further stirred at 60 °C for 30 min then cooled down to room temperature. DMSO was removed by ultrafiltration with the PBS buffer against a MWCO = 100 kDa filter. The particle size was measured by DLS with a scattering angle at 173°. HA-b-PBLG-30nm was obtained by using a 1 wt% solution of HA-b-PBLG in DMSO with fast nanoprecipitation. HA-b-PBLG-150nm was obtained by using a 1 wt% solution of HA-b-PBLG in DMSO with controlled nanoprecipitation. LAM-b-PBLG-30nm was obtained by using a 1 wt% DMSO solution of LAM-b-PBLG in DMSO with fast nanoprecipitation. Hybrid-nano 90%LAM was obtained by using a DMSO solution containing 0.9 wt% LAM-b-PBLG and 0.1 wt% HA-b-PBLG with fast nanoprecipitation. Hybrid-nano 50%LAM was obtained by using a DMSO solution containing 0.5 wt% LAM-b-PBLG and 0.5 wt% HA-b-PBLG with fast nanoprecipitation. SPR analysis: surface functionalization and interaction analysis. CM5, a carboxymethylated dextran sensor chip from GE Healthcare, was used for SPR analysis. Recombinant human CD44 was immobilized on the CM5 chip by using a standard amine coupling protocol of Biacore T200. Briefly, the sensor chip flow cell was activated by an EDC/NHS mixture for 420 seconds. CD44 was dissolved in 10 mM acetate buffer pH 4 at 10 µg/ ml then injected to the activated surface for 300 seconds. About 4000RU of CD44 was immobilized on the chip during the injection. The remaining activated positions on the chip were then deactivated by 1 M ethanolamine pH 8.5. A blank flow cell was prepared as a reference by the same protocol without CD44 injection. HBS-EP+ buffer, proposed by GE Healthcare, was chosen as the running buffer to perform all SPR analysis in this study. All the sample injection was performed at a rate of 30 µl/min. For the study in Fig. 4, the solutions of HA of different molecular weights (5 kDa, 20 kDa, 100 kDa, 1000 kDa) was prepared at 10 ppm in the running buffer, whereas those of HA-b-PBLG-30nm and HA-b-PBLG-150nm were prepared at 23ppm so that all the samples contain the same quantity of HA for comparison. For the same reason, HA-b-PBLG-30nm was prepared at 100ppm, Hybrid nano-50%LAM was prepared at 200ppm and Hybrid nano-90%LAM was prepared at 1000ppm for the study in Fig. 4. The particle samples were injected at 100ppm and dectin-1 solution was added at 10 µg/ml in the study shown in Fig. 5. The responses on the blank flow cell were systematically subtracted from the signal obtained on the CD44 coated flow cell to remove the contribution of unspecific interaction independent from CD44. The surface was regenerated by adding 50 mM NaOH for 30 seconds after each analysis. The full removal of the analyte attached to CD44 was confirmed by the baseline level after the regeneration.

chemsum

{"title": "Multivalent and multifunctional polysaccharide-based particles for controlled receptor recognition", "journal": "Scientific Reports - Nature"}

molecular_recognition_using_receptor-free_nanomechanical_infrared_spectroscopy_based_on_a_quantum_ca

## Abstract: Speciation of complex mixtures of trace explosives presents a formidable challenge for sensors that rely on chemoselective interfaces due to the unspecific nature of weak intermolecular interactions. Nanomechanical infrared (IR) spectroscopy provides higher selectivity in molecular detection without using chemoselective interfaces by measuring the photothermal effect of adsorbed molecules on a thermally sensitive microcantilever. In addition, unlike conventional IR spectroscopy, the detection sensitivity is drastically enhanced by increasing the IR laser power, since the photothermal signal comes from the absorption of IR photons and nonradiative decay processes. By using a broadly tunable quantum cascade laser for the resonant excitation of molecules, we increased the detection sensitivity by one order of magnitude compared to the use of a conventional IR monochromator. Here, we demonstrate the successful speciation and quantification of picogram levels of ternary mixtures of similar explosives (trinitrotoluene (TNT), cyclotrimethylene trinitramine (RDX), and pentaerythritol tetranitrate (PETN)) using nanomechanical IR spectroscopy. ## D etection, speciation, and quantification of extremely small concentrations of explosive vapors with high selectivity and sensitivity have immediate applications in many areas such as national security, forensics, and humanitarian demining 1 . Microfabricated chemical sensors are being actively investigated as potential sensor platforms capable of mass deployment. Despite having many advantages such as miniature size, high sensitivity, low-cost, and low power consumption, the microfabricated sensors suffer from poor chemical selectivity. Miniature sensors rely on adsorption-induced changes in physical variables such as adsorbed mass, surface stress, refractive index, resistance, capacitance, and temperature which are sensitive indicators of molecular binding. Since the changes in physical properties due to molecular adsorption are not chemically specific, sensors are usually modified with receptors (chemical interfaces) which can provide selectivity. Immobilized chemoselective interfaces, however, can only provide partial selectivity due to the unspecific nature of chemical binding; especially those based on weak intermolecular interactions such as hydrogen bonding. The interference from other chemical vapors which cause unacceptable levels of false positives is a major challenge for all sensors based on analyte interactions with immobilized chemical interfaces. Therefore, achieving chemical selectivity in a mixture of unknown chemical vapors is extremely difficult without resorting to molecular separation. Even approaches based on sensor arrays immobilized with unique chemical interfaces and subsequent analysis of array response using pattern recognition algorithms fail when it comes to ternary mixtures . Adding to the challenge is the low vapor pressure of explosives, which severely limits the number of molecules reaching the sensor surface in an acceptable detection time which requires extremely high sensitivity. In addition, explosive vapors can be present with a variety of vapors in the sensing environment and this can interfere with the selectivity of detection. Therefore, an urgent need exists for developing techniques which are selective, sensitive and quantitative for different mixtures of explosive vapors. In order to be effective, the sensor must have the ability to differentiate between the explosive molecules and other similar compounds. For surface adsorption-based sensors, interfering chemicals which have a high vapor pressure, such as volatile organic compounds, do not cause interference since they do not adsorb well on the surface at room temperature 7 . However other materials with very low vapor pressure, such as different types of explosives, can cause challenges in selective detection and quantification. Unlike sensing paradigms based on immobilized chemoselective interfaces on sensor surfaces, spectroscopic techniques based on unique molecular vibrational transitions in the mid infrared (IR) ''molecular fingerprint'' regime, where many molecules display characteristic vibrational peaks free from overtone, are highly selective 8 . Spectroscopic signal from a mixture of molecules follows the superposition principle, unless there are intermolecular interactions. This is different from sensing based on partially selective chemoselective interfaces, which fails in mixtures due to the lack of orthogonality in sensor responses. Although mid-IR absorption spectroscopy based chemical sensing offers high selectivity, it lacks sensitivity when used for detection of surface adsorbed chemicals. Photothermal cantilever deflection spectroscopy (PCDS), which combines the extreme thermal sensitivity of a bi-material microcantilever with the high selectivity of mid-IR spectroscopy, is capable of obtaining molecular signatures of trace amounts of adsorbed molecules on the cantilever surface . In the PCDS technique, the target molecules are first allowed to adsorb on a bi-material cantilever. During resonant excitation of target molecules using IR light, the bi-material cantilever undergoes deflection, the amplitude of cantilever deflection as a function of IR wavelength resembles the infrared absorption spectra of the adsorbed molecules. Unlike conventional IR absorption spectroscopy, in which a small intensity change is collected by cryogenically cooled mid-IR detectors in a large background with inherent laser source noise, PCDS, an ''action spectroscopy'', measures the photothermal effect of a small number of adsorbed molecules with a high photon flux taking full advantage of the high brightness of a quantum cascade laser (QCL) light source. Therefore, the PCDS signal strength scales with the intensity of incident photons while the noise in PCDS mainly comes from thermomechanical noise of a microcantilever. Using a high power tunable QCL, we have been able to increase the sensitivity of detection by one order of magnitude. This, therefore, lays the foundation for the enhancement of the sensitivity of detection by increasing the intensity of the excitation source. In this report, we demonstrate the successful implementation of PCDS for selective detection and quantification of the ternary mixtures of similar explosive molecules (trinitrotoluene (TNT), cyclotrimethylene trinitramine (RDX), and pentaerythritol tetranitrate (PETN)), which represent the three most commonly found explosive classes (nitro-aromatics, nitramines, and nitrate esters) with tens of picogram resolution in ambient condition. In addition, we improve the limit of recognition (LOR), the maximum recognizable mixture composition range, by an order of magnitude (,3251) with a 100% recognition rate, by calibrating and analyzing PCDS spectra compared to the LOR of immobilized chemoselective interfaces-based multi-transducer array microsensors. In contrast to an immobilized chemoselective interfaces-based microsensor array, which has a relatively short operation-life due to the degradation of coating in ambient condition, resulting in the loss of selectivity and capability of the quantitative detection, the PCDS technique utilizes a very robust single microstructure without any chemoselective interfaces. Significant field applications potential is demonstrated by the selective detection and quantification of ternary mixtures of explosive molecules in ambient condition. ## Results The PCDS setup used in this study is shown in Fig. 1a. Inherently, the PCDS provides two orthogonal signals in a single transducer platform. The nanomechanical IR spectrum, a differential plot of the amplitude of the cantilever deflection as a function of impinging IR wavelength with and without target molecules, represents molecular signatures of the target molecules adsorbed on the cantilever surface while the resonance frequency change of the microcantilever gives real time information of adsorbed mass (Fig. 1b). The bi-material microcantilever serves as an extremely sensitive thermal sensor as well as a microresonator for the detection, speciation, and quantification of ternary mixtures of the explosive molecules. When IR photons are absorbed by the explosive molecules on the cantilever surface, the explosive molecules undergo transitions from the n50 ground vibrational states to the n51 excited states (Fig. 1c). The high frequency low density normal modes (represented on a sharp Morse potential) is coupled to the high n excited states of other low frequency high density normal modes (one of which represented on a flat Morse potential) via intramolecular vibrational energy redistribution. Eventually, the energy is released to the phonon bath of the bi-material cantilever surface through multiple steps of vibrational energy relaxation. These nonradiative decay processes result in heating up the bi-material cantilever, generating the deflection of the cantilever. Within certain dynamic ranges, the normalized peak amplitudes of the nanomechanical IR spectrum can be utilized to estimate the relative mass ratio of each target molecule in a mixture since the IR spectrum of a mixture is a linear superposition of individual spectra. Fig. 2 shows the normalized nanomechanical IR absorption spectra of TNT (black), RDX (green), PETN (red), 15151 mixture of these three explosives (blue), a weighted linear superposition of individual explosive spectra (sky blue), and the Fourier transform infrared (FTIR) spectrum of the ternary mixture (orange). The PCDS spectra of the individual explosives were taken separately as references and agreed quite well with our previous report 15 . These spectra were acquired with a monochromatic IR source and normalized by the adsorbed mass of explosive molecules for calibration purposes. The adsorbed mass of TNT, RDX, PETN, and the ternary mixture on the cantilever was 6.78 ng, 6.44 ng, 6.99 ng, and 9.8 ng respectively, as calculated using Eq. 2. The relative mass ratio of the ternary mixture (TNT5RDX5PETN) adsorbed on the cantilever surface was estimated to be 0.3650.150.54 from the mathematical fitting of PCDS spectrum of the ternary mixture with normalized PCDS spectra of individual explosive molecules. Therefore, the actual mass of TNT, RDX, and PETN on the cantilever surface was determined to be 3.53 ng, 0.98 ng, and 5.29 ng respectively. Even though the same volume for each explosive solution with same concentration was mixed, the actual mass of each explosive adsorbed on the cantilever surface was quite different due to different molecular affinity to the silicon oxide surface. Since their vapor pressures are different, desorption rates from the surface are also different 7 . Several distinct peaks and shoulders appeared in the ternary mixture spectrum since the mixture spectrum is a linear superposition of individual spectra (note that spectral resolution of our IR monochromator in this range is approximately 0.12 mm). The peaks at 6.06, 6.38, and 6.49 mm are due to the asymmetric stretching of the NO 2 (nitro) group bonds while the peaks at 7.27, 7.46, and 7.82 mm are from the symmetric stretching of the same group bonds. Comparing these peaks with those of individual TNT, RDX, and PETN spectra, it is apparent that the peaks at 6.49 and 7.46 mm are from TNT, the peaks at 6.38 and 7.27 mm are from RDX, and the peaks at 6.06 and 7.82 mm are from PETN molecules . It is interesting to note that the prominent RDX peak at 7.57 mm is not clear in the PCDS spectrum and the prominent PETN peak at 7.82 mm is missing in the FTIR spectrum due to the difference in the relative mass ratio between RDX and PETN. Other than these two differences, all of the characteristic peaks are separate and distinct, matching closely with each other. This demonstrates the capability of PCDS to distinguish between these closely related explosive molecular species. Since the peaks and shoulders are highly distinguishable, we can surmise that PCDS can detect differences between such closely related molecular species and anticipate that PCDS can distinguish between other interfering compounds and target molecules while sensing in ''real world'' environments. We have improved the limit of detection (LOD) and explored the LOR of this PCDS setup by employing a tunable QCL as a powerful IR source. Fig. 3a presents the normalized peak amplitude of RDX at 7.57 mm (green squares) and the standard deviation of spectrum noise in a non-absorbing region (violet circles) as a function of the incident laser power. The inset shows the magnified view of the standard deviation of noise. The straight lines are the linear fit of the normalized peak amplitudes and the standard deviation of noise. Although the noise increased when increasing the incident laser power, the PCDS signal enhancement dominated and consequently signal-to-noise ratio (SNR) increased in our tested power range. Fig. 3b shows normalized PCDS spectra of TNT (black), RDX (green), and PETN (red) acquired with the same cantilever using the maximum power of QCL in our tested range with a 5 nm spectral resolution. The peaks between 7.1 and 8.0 mm are from the symmetric stretching vibration of the NO 2 (nitro) group bonds with carbon (C-NO 2 ) in TNT (7.46 mm); nitrogen (N-NO 2 ) in RDX (7.57 mm); and oxygen (O-NO 2 ) in PETN (7.82 mm), respectively. The peak amplitudes at 7.46, 7.57, and 7.82 mm for TNT (black squares), RDX (green triangles), and PETN (red circles) were plotted as a function of adsorbed mass of each explosive molecule in Fig. 3c to explore the LOD of the PCDS setup. The straight lines are the linear fit of the peak amplitudes for TNT (black), RDX (green), and PETN (red) respectively. It was estimated that the limit of detection for TNT, RDX, and PETN is 39 pg, 28 pg, and 79 pg, respectively with an SNR of 3. The concept of a LOR was introduced more than a decade ago and is well established as an additional criterion for evaluating the performance of a vapor sensor array 21 . Originally, LOR was defined as the maximum recognizable mixture composition range which can be reliably determined from the response pattern of a sensor array. This is especially important in the speciation of vapor mixture components. To estimate the LOR of our PCDS setup, 15 ternary mixtures of standard explosives samples were prepared using varying volume ratios of each standard sample solution. Fig. 4a shows the normalized PCDS spectra of ternary mixtures (TNT5RDX5PETN) of explosives with the volume ratio of the TNT sample solution increasing from 15151 to 155151. The intensity of a peak is directly proportional to the adsorbed mass. The actual masses of TNT, RDX, and PETN on the cantilever surface were determined with the total adsorbed mass from the resonance frequency shift measurements and the estimated relative mass ratios obtained from the mathematical fitting of the PCDS spectra of the ternary mixtures with linear superpositions of normalized PCDS spectra of individual explosive molecules. A representative fitting result is provided in Supplementary Information. In a similar manner, the normalized PCDS spectra of ternary mixtures of explosives with increasing RDX and PETN concentrations (Fig. 4b and 4c) were analyzed and the relative mass ratios were plotted with respect to the relative mass ratio of increasing explosives as shown in Fig. 4d. It was estimated that the LOR based on the LOD of our PCDS setup for TNT to RDX is 3251 and RDX to TNT is 3151 with an SNR of 3 which ensures there are no false positives or negatives. The LOR for TNT to PETN is 2351 and PETN to TNT is 3251; the LOR for RDX to PETN is 3051 and PETN to RDX is 2651 with the same SNR. The results demonstrate that the PCDS technique using a very robust single microcantilever transducer without any chemical interfaces overcomes the LOR of immobilized chemoselective interfacesbased multi-transducer array microsensors up to an order of magnitude 6 and achieves room temperature reversibility without leading to unacceptable levels of false positives or negatives. ## Discussion The relationship between adsorbed mass and resonant frequency shift is given by 22 Dm~m where Dm is the mass of the adsorbate and m 0 is the mass of the clean cantilever. E 0 , t 0 , and f 0 are the initial values of the Young's modulus, thickness and resonance frequency of the cantilever, respectively. DE, Dt, and Df are the changes in the Young's modulus, thickness, and resonance frequency of the cantilever, respectively. In this study, if the changes in thickness and Young's modulus are negligible and explosive molecules are considered uniformly adsorbed on the surface, then Eq. 1 can be simplified to: Using this equation, the adsorbed mass of explosives on the cantilever is determined. In PCDS, the amplitude of cantilever deflection depends on the impinging power of IR, the IR absorption mode and the amount of adsorbed molecules as well as the thermal sensitivity of the cantilever 15 . Although the PCDS signal can be further increased by enhancing the thermal sensitivity of the microcantilever as well as by increasing the impinging power of IR, we should take thermomechanical noise, a dominant noise source for PCDS, into account in order to evaluate SNR which determines the limit of detection 23,24 . The root mean square amplitude of cantilever deflection from thermomechanical noise at well below the resonance frequency is given by 25 where k B is the Boltzmann constant, T is the absolute temperature, B is the measurement bandwidth, Q is the quality factor, k is the spring constant, and v 0 is the angular resonance frequency of the cantilever. Although the thermomechanical noise increases when increasing the incident IR laser power due to heating effect, the maximum temperature rise in the cantilever is linearly proportional to the incident laser power as detailed in Supplementary Information. Therefore, thermomechanical noise grows slower than the PCDS signal shown in Fig. 3a and consequently SNR increases. The LOR for the PCDS setup also can be further extended by using broadly tunable QCLs ranging from 5.5 mm to 8.0 mm which covers the asymmetric and symmetric stretching vibrations of the NO 2 (nitro) group bonds in explosive molecules. The more characteristic peaks of analytes we have, the wider the range of recognizable mixture composition we can expect to achieve. With the advent of miniature IR sources, it is possible to decrease the size of the device into a handheld one. Therefore, significant field application potential is anticipated by miniaturizing a broadly tunable IR source and readout electronics which are integrated into a microcantilever sensor system. We have observed that, although the positions of the absorption peaks in the PCDS spectra agree closely with those in conventional IR spectra, the relative intensities of the observed PCDS peaks do not 15 . We reasoned that, unlike conventional IR absorption spectroscopy which measures the molecular IR absorption following the Beer-Lambert law, PCDS signals come from the nonradiative decayinduced thermal variation of the cantilever as a result of molecular IR absorption and multi-step coupling of the vibrational excited states of surface adsorbed molecules to the phonon bath of the cantilever through intramolecular vibrational energy redistribution and vibrational energy relaxation. Therefore, the differences between the relative peak intensities of PCDS and conventional IR spectra could be used for quantifying the vibrational dynamics of the adsorbed molecules and the energy transfer processes on the surface. ## Methods Chemicals. Three standard explosive samples (TNT, RDX, and PETN) were purchased from AccuStandard, Inc. (New Haven, CT) and used without further purification. As indicated by the manufacturer, the standard concentration of each explosive is 1 mg/mL. Preparation of a microcantilever. Rectangular silicon cantilevers (CSC12-E) were obtained from MikroMasch USA (San Jose, CA). The dimension of each cantilever was 350 mm in length, 35 mm in width, and 1 mm in thickness. The microcantilevers were cleaned by rinsing with acetone, ethanol, and a UV ozone treatment then coated with 10 nm of chromium (adhesion layer) followed by 200 nm of gold using an e-beam evaporator. The individual explosive and ternary mixtures (by volume of standard sample solution) of explosive molecules (TNT, RDX, and PETN) were deposited on a microcantilever sequentially using micro glass capillaries and the cantilever was completely regenerated by UV ozone cleaning following each measurement. The PCDS experimental setup. For the PCDS experiments, the explosives-deposited cantilever was mounted on a stainless steel cantilever holder which was attached to the head unit of MultiMode atomic force microscope (AFM) (Bruker, Santa Barbara, CA). The deflection and resonance frequency of the microcantilever were measured using the optical beam deflection method with a laser diode and a position sensitive detector. The IR radiation from the monochromator (Foxboro Miran 1A-CVF) was mechanically chopped at 80 Hz and focused on the cantilever. The IR wavelength was scanned from 2.5 mm to 14.5 mm (4000 cm 21 to 690 cm 21 in wavenumber) and has a resolution of 0.05 mm at 3 mm, 0.12 mm at 6 mm, and 0.25 mm at 11 mm according to the manufacturer. The 200 kHz pulsed IR radiation with 10% duty cycle from the QCL (Daylight Solutions, UT-8) was electrically burst at 80 Hz using a function generator DS345 (Stanford Research Systems, Sunnyvale, CA) and directed to the cantilever. The laser power was measured with a FieldMax II laser power meter (Coherent Inc., Santa Clara, CA). The IR wavelength was scanned from 7.1 mm to 8.3 mm (1408 cm 21 to 1204 cm 21 in wavenumber) with a spectral resolution of 5 nm. The nanomechanical IR spectra were taken using a SR850 lock-in amplifier (Stanford Research Systems, Sunnyvale, CA) and the resonance frequencies of the microcantilever were measured with a SR760 spectrum analyzer (Stanford Research Systems, Sunnyvale, CA). FTIR microscope. The ternary mixture of explosive molecules on the microcantilever chip was characterized using a standard FTIR technique as a reference. Two (2) microliters of 15151 mixture solution was drop-cast onto the microcantilever chip, and the FTIR spectra were obtained using an FTIR microscope (Nicolet Continumm FTIR microscope) in reflection mode. The number of registered scans was 200 with resolution of 4 cm 21 .

chemsum

{"title": "Molecular recognition using receptor-free nanomechanical infrared spectroscopy based on a quantum cascade laser", "journal": "Scientific Reports - Nature"}

ultrafast_realization_of_ionic_liquids_with_excellent_co_2_absorption:_a_trinity_study_of_machine_le

## Abstract: Efficient CO2 capture is indispensable for achieving a carbon-neutral society while maintaining a high quality of life. Since the discovery that ionic liquids (ILs) can absorb CO2, various solvents composed of molecular ions have been developed and their CO2 solubility has been studied. However, it is challenging to optimize these materials to realize targeted properties as the number of candidate ion combinations for designing novel ILs is of the order of 10 18 . In this study, electronicstructure informatics was applied as an interdisciplinary approach to quantum chemistry calculations, and combined with machine learning to search 402,114 IL candidates to identify those with better CO2 solubility than known materials. Guided by the machine-learning results, trihexyl(tetradecyl)phosphonium perfluorooctanesulfonate was synthesized and it was experimentally confirmed that this IL has higher CO2 solubility than trihexyl(tetradecyl)phosphonium bis(trifluoromethanesulfonyl)amide, which is the previous best IL for CO2 absorption. The method developed in this study could be transferable to gas-absorbing liquids in general, such as deep eutectic solvents (hydrogen-bonded mixed organic solvents in a broad sense), which also have numerous practical applications. Therefore, we believe that our method for developing functional liquids will significantly contribute to the development of a carbon-neutral society. ## INTRODUCTION Developing CO2-absorbing materials with better performance is a critical step toward achieving efficient CCUS (carbon dioxide capture, utilization, and storage), which has been proposed as one strategy to address global warming. 1 While there are several types of CO2 absorbers, the intensive study of ionic liquids (ILs: room temperature molten salts) would give basic science contributing to environmental engineering toward the development of a safe and sustainable society. ILs are highly stable (low volatility and highly heat resistant) and can be chemically designed with a wide variety of ion combinations. ILs have been applied to various gas-fixing technologies, such as temperature/pressure swing, chemical/physical absorption, and cryogenics/membrane-based separation. 6,7 A major limitation of the development of ILs for applications is the combinatorial explosion of the ion species, i.e., there are 10 18 ILs in theory. 2 Even when ILs are limited to physical absorbents (those not leading the CO2 chemisorption), the optimization of the CO2 absorption capacity is experimentally challenging. The synthesis of ILs and characterization of their physical properties are more labor intensive than for electrically neutral molecular liquids due to their high viscosities. Despite the vast number of candidates, the record for the maximum amount of CO2 physically absorbed by an IL has not been broken for over a decade. 8 To date, performing a series of targeted syntheses followed by high-precision experimental measurements have been necessary to obtain practical CO2 absorbers. Machine learning has proven to be an excellent approach for such combinatorial problems. 9-13 However, to date, machine learning has not been applied to identify better ILs for CO2 absorption in a comprehensive manner including molecular design, synthesis, and CO2 solubility observations. One of the simple reasons is that the chemical structure of ionic species cannot be uniquely defined due to charge delocalization. In addition, slight differences in the electronic states of ionic species, which can have critical effects on the properties of ILs, are challenging to describe using fingerprints. Considering this background, this study focused on a materials informatics approach based on quantum chemistry calculations to construct a sizeable electronic structure database for ILs (Figure 1). This study aimed to develop a method to search for better CO2-absorbing ILs with high accuracy and a much shorter development time by using quantum-chemical features and machinelearning techniques (electronic structure informatics). The prediction model is proposed for quick identification of ILs with better CO2 absorption properties which is applicable for large group of materials. To prove the effectiveness of the method, the synthesis of two promising candidates and the measuring of their CO2 solubilities are performed. ## METHODS First, a theoretical screening was performed based on quantum chemistry calculations and machine learning. Following our previous studies, 14,15 6,991 stable ion structures (6,933 cations and 58 anions in Figure 2 and Table 1) were explored by density functional theory calculations at the BP/TZVP-D3 level. Using the surface-charge-density distributions (σ-profiles) and the optimal structures, the geometric and electronic features given in Table 2 were calculated for all of the ions. Then, a set of Henry's law constant (! !" ! ) values at 298.15 K was evaluated by COSMO-RS theory 25 for 20,000 ILs (randomly selected from the total 402,114 candidates). Half of this set were used to train a machine learning model to predict ! !" ! for the other IL candidates in this study. Using a cycle of feature selection, model creation, and performance evaluation, the essential molecular (ion) features for the CO2 absorption problem were systematically selected (wrapper method). 26 For model creation, the Gaussian process regression method with the ARDMatern 5/2 kernel 27 and 5-fold cross-validation were applied to the standardized data. The performance of the created model was evaluated with 10,000 sets of test data from the value of the coefficient of determination (R 2 ), the root mean squared error (RMSE), and the mean absolute error (MAE). Using the well-trained model, superior ILs for CO2 absorption were predicted within a minute. The quantum chemistry and statistical thermodynamics calculations and machine learning were performed with TURBOMOLE 7.0. 28 , COSMOtherm C30_1705 29 , and MATLAB 30 packages. ## Class Structures Fluorineinorganic Nonfluorineinorganic ## Fluorineorganic Nonfluorineorganic respectively. These ILs were synthesized according to methods in the literature. 31,32 The density (ρ), viscosity (η), and CO2 solubility (3 +, ! ) under atmospheric to high pressure conditions were measured for the ILs using a vibrating tube densimeter (Anton Paar, DMA 5000M), a rotating-cylinder viscometer (Anton Paar, Stabinger SVM 3000), and a magnetic suspension balance (Rubotherm GmbH) (see Supporting Information and references ). 4 +, ! was evaluated for the ILs as the solubility gradients ( -. -/ "# ! ) in the region of 3 +, ! < 0.1. Finally, the Gibbs energy (DabsG ∞ ), enthalpy (DabsH ∞ ), and entropy (DabsS ∞ ) of CO2 absorption for the ILs were experimentally obtained using the following thermodynamic relations (1)-( 3) as a function of ln 4 +, ! , 40 where DabsH ∞ and DabsS ∞ are closely related to the solute-solvent interaction strength and the free volume size of the solvent. ## RESULTS AND DISCUSSION The accuracy of the machine-learning was confirmed by the learning curves (Figure 3). It was shown that highly accurate HCO2 predictions (R 2 > 0.90, RMSE< 1.5 MPa, MAE< 0.13 MPa) were achieved when 12 features were applied (Figure 3(a)). The wrapper method clarified that the first-and second-most important features were geometric V and electronic M2. This is considered reasonable because CO2 molecules are physically absorbed in the polar voids that consist of ions in the ILs. It was also confirmed that the 10,000 sets of training data (2% of the candidates) were sufficient for obtaining an accurate model (Figure 3(b)). S2 and S4). Both the mole fraction and molarity-scaled CO2 solubilities were also measured (Figure 4 and Table S6). The solubilities proportional to pressure means that the ILs absorb CO2 by a physisorption mechanism. The CO2 solubility increased in the anion order of PF6 − < TFSA − < PFOS − . The large numbers of fluorine atoms and S=O bonds in PFOS − can develop strong interactions between the anion and CO2, which leads to the largest anion-CO2 contact probability for PFOS − (PFOS − : 0.318, TFSA − : 0.232, PF6 − : 0.151) in the COSMO-RS results. The absorption Gibbs energy of [P66614][PFOS] was the most stable among the three ILs due to the large negative enthalpy (Table 3). In this study, we focused on the predictivity of the CO2 solubility only by assuming that all cation/anion pairs would produce liquids instead of solids. Nevertheless, the fact that we have

chemsum

{"title": "Ultrafast Realization of Ionic Liquids with Excellent CO 2 Absorption: A trinity study of machine learning, synthesis, and precision measurement", "journal": "ChemRxiv"}

two-photon_aie_bio-probe_with_large_stokes_shift_for_specific_imaging_of_lipid_droplets

## Abstract: Lipid droplets are dynamic organelles involved in various physiological processes and their detection is thus of high importance to biomedical research. Recent reports show that AIE probes for lipid droplet imaging have the superior advantages of high brightness, large Stokes shift and excellent photostability compared to commercial dyes but suffer from the problem of having a short excitation wavelength. In this work, an AIE probe, namely TPA-BI, was rationally designed and easily prepared from triphenylamine and imidazolone building blocks for the two-photon imaging of lipid droplets. TPA-BI exhibited TICT+AIE features with a large Stokes shift of up to 202 nm and a large two-photon absorption cross-section of up to 213 GM.TPA-BI was more suitable for two-photon imaging of the lipid droplets with the merits of a higher 3D resolution, lesser photobleaching, a reduced autofluorescence and deeper penetration in tissue slices than a commercial probe based on BODIPY 493/503, providing a promising imaging tool for lipid droplet tracking and analysis in biomedical research and clinical diagnosis. ## Introduction Lipid droplets (LDs) that contain mainly diverse neutral lipids such as triacylglycerol and cholesteryl ester are widely found in adipocytes, hepatocytes and the adrenal cortex. For many years, LDs have been regarded as inert reservoirs of neutral lipids for energy storage. However, recent results show that LDs are considered to be dynamic organelles and associated with the storage and metabolism of lipids, signal transduction, apoptosis and so on. 1 The abnormalities of LDs are generally related to some important diseases. 1,2 For example, LDs are found to be critical for the proliferation of the hepatitis C virus, 3 infection of which will lead to chronic hepatitis, liver cirrhosis and hepatocellular carcinoma. Thus, the localization and analysis of LDs are highly important for biomedical research and clinical diagnosis. Techniques based on fluorescent materials are emerging as powerful and popular tools for biomedical studies both in vitro and in vivo. 4 They exhibit excellent performances in applications such as localizing subcellular organelles, and monitoring the physiological changes of pH, temperature, viscosity, ions, proteins, and so on with the superior advantages of high resolution and sensitivity, easy operation and low cost. Recently, fluorescent probes for the localization of LDs have been developed. Two commercial dyes, namely Nile Red and BODIPY 493/ 503, are widely used but they show some fatal drawbacks, such as strong backgrounds and small Stokes shifts. 5 Worse still, these conventional organic fluorophores unavoidably face a problem of aggregation-caused quenching (ACQ), where their fluorescence is quenched at high concentrations due to the formation of detrimental species such as excimers and exciplexes by strong p-p stacking. 6 The ACQ effect has largely confned their working concentration to a very low nanomolar level, leading to quick photo-bleaching for bioimaging. For many years, we and other groups have worked on the development of molecules with aggregation-induced emission (AIE) characteristics that are the exact opposite of the ACQ fluorophores. The restriction of intramolecular motion (RIM) has been proposed as the mechanism for the AIE effect. 7 AIE luminogens (AIEgens) are weakly emissive in solutions due to the deactivation of the excited states by active intramolecular motions. However, such motions are suppressed in the aggregate state, thus enabling them to emit intensely upon excitation. AIEgens have found promising biomedical applications due to their superior merits of large Stokes shifts, high brightness, good biocompatibility, excellent photostability, etc. 8 Therefore, the development of LD-specifc AIE bioprobes could provide a promising approach to solving the problem observed in commercial dyes. Indeed, in our previous work, LD-specifc AIE bioprobes, such as TPE-AmAl, FAS, DPAS and TPE-AC (Chart 1), show better performances in terms of brightness, specifcity and photostability than their commercial counterparts in both fxed and living cell imaging. Meanwhile, these AIE-based bioprobes can be easily synthesized and have good cell permeability. 9 However, most of the LD-specifc AIE bioprobes developed so far bear either UV excitation or shortwavelength emission, which is harmful to living cells and suffers from limited penetration depth to tissue and serious auto-fluorescence from biosamples. 4,10 Although TPE-AC exhibited a fascinating NIR emission (705 nm), 11 the excitation wavelength was merely 450 nm, which was not long enough to reach the optical window for optimal tissue penetration (750-950 nm). 10b Thus, LD-specifc AIE bioprobes with excitation wavelengths in the red and near-infrared (NIR) regions will solve these problems and are in urgent demand. Many efforts have been devoted to designing new LD-specifc AIE bioprobes with red and NIR excitations. Unfortunately, such a task is not easy in terms of tedious synthesis and low emission efficiency of the resulting molecules. Recently, two-photon fluorescence microscopy (2PM) has become popular in biomedical diagnosis and therapy, due to its advantages of a longer-wavelength excitation, lower auto-fluorescence, higher 3D resolution and less photobleaching. 12 Luminescent materials with two-photon excitation are crucially determined by a two-photon absorption (2PA) cross section (d 2PA ). Materials bearing higher d 2PA will show stronger twophoton excited fluorescence (TPEF) and a less deleterious thermal effect from the strong laser pulse. 13 Therefore, the design of AIE bioprobes with two-photon excitations can provide an easier way to realize red and NIR excitations. Benzylidene imidazolone (BI), the analogue of the chromophore of green fluorescent protein, has been wildly-studied due to its facile synthesis and excellent biocompatibility. 14 Recently, many of its derivatives have been designed and found to be AIEactive. 15 Compared to TPE, BI possesses a more rigid structure with a less twisted conformation and would be an ideal building block for 2PA materials. 16 However, to the best of our knowledge, BI-based 2PA materials have been rarely reported. Herein, we attempt to integrate the merits of AIE and 2PA into BI. On the other hand, triphenylamine (TPA) is a popular design unit for 2PA 10b,17 and a well-known strong electron donor. The decoration of BI with TPA is thus expected to give a luminogen with a high d 2PA and a longer-wavelength emission. The structural design of the molecule, abbreviated to TPA-BI, is shown in Scheme 1. Indeed, TPA-BI possessed a large d 2PA and exhibited strong TPEF. TPA-BI can specifcally stain lipid droplets in both fxed and live cells with a large Stokes shift and a superior twophoton imaging performance. ## Results and discussion Synthesis TPA-BI was readily synthesized in a good yield by the Suzuki coupling of (Z)-5-(4-bromobenzylidene)-2-methyl-3-propyl-3,5dihydro-4H-imidazol-4-one ( 2) and (4-(diphenylamino)phenyl) boronic acid (3) (Scheme 1). Detailed experimental procedures are provided in the Electronic Supplementary Information (ESI †). The structure of TPA-BI was fully characterized and confrmed by NMR and high-resolution mass spectroscopies (ESI, Fig. S1-S3 †). ## Solvatochromism and twisted intramolecular charge-transfer Molecules with donor (D)-p-acceptor (A) structures are characterized by a prominent solvatochromic effect, where their photophysical properties change by varying the solvent polarity. Hence, the absorption and photoluminescence (PL) spectra of TPA-BI in solvents with different polarities were investigated and the results are shown in Fig. 1 and S4. † In Fig. 1A, under UV light irradiation, the emission colour of the TPA-BI solution could be fnely tuned from blue to red when the solvent changed from n-hexane to acetonitrile, nearly covering the full visible spectrum. The emission maximum varied gradually from 447 nm to 619 nm (Fig. 1B). Simultaneously, a pronounced decrease in the emission intensity was observed. On the contrary, the absorption of TPA-BI exhibited little change on changing the solvent polarity (Fig. S4 †). The absorption maximum of TPA-BI only changed from 400 nm to 414 nm by increasing the solvent polarity with an extinction coefficient of $34 000 M 1 cm 1 . All of these results indicate that the photophysical properties of TPA-BI are strongly dependent on the solvent polarity, which is ascribed to the twisted intramolecular charge transfer (TICT) effect from the electron-donating TPA unit to the electron-accepting imidazolone functionality. A large Stokes shift of up to 212 nm was realized, largely avoiding the overlap of the absorption spectrum and emission spectrum. This property is highly demanded for fluorescence probes as it prevents the self-absorption or "inner-flter" effect to increase the signal to noise ratio for fluorescence imaging. To evaluate the effect of the solvents on the PL of TPA-BI, the change in the PL maximum with the solvent polarity parameter (E T (30)) 18 is plotted in Fig. 1C and summarized in Tables S1 and S2. † A linear line with a correlation coefficient of R 2 ¼ 0.992 and a large slope of 11.8 was obtained, indicating the remarkable solvatochromism of TPA-BI. The solvatochromic properties of TPA-BI were also confrmed by the dependence of the fluorescence transition energy on the solvent orientation polarizability (Df 0 ) according to the revised Lippert-Mataga equation for TICT molecules (Table S1 and Fig. S5 †). Both results indicate that TPA-BI shows strong solvatochromism resulting from the TICT effect. The TICT effect of TPA-BI can be interpreted by density functional theory (DFT) calculations (Fig. S6 †). The photoexcitation from the S 0 to S 1 state of TPA-BI involves a substantial intramolecular charge transfer (ICT) from TPA to the imidazolone unit. Since the donor and acceptor are linked via a freely rotatable single bond, the activation of the ICT process is likely accompanied by a signifcant molecular geometry change and the formation of a TICT state. The TICT state will be largely stabilized and populated in solvents with higher polarity, resulting in a red-shift in the emission band. The TICT effect is responsible for the solvatochromism of TPA-BI and the increase in the Stokes shift from non-polar to polar solvents. The decrease in the PL intensity in a polar solvent should be attributed to the rapid consumption of the energy of the TICT state through non-radiative relaxation pathways. 19 ## Aggregation-induced emission Besides solvatochromism, TPA-BI also shows an aggregationinduced emission (AIE) phenomenon. As shown in Fig. 2, with an increase in the water fraction from 0 to 40% in the dimethylsulfoxide (DMSO)/water mixture, the emission of TPA-BI decreased, accompanied with a slight red-shift in the PL spectrum. This is due to the enhancement of the TICT effect in the presence of the more polar solvent of water in the surrounding environment. Upon further increasing the water fraction from 40% to 70%, an abrupt increase in the emission intensity ($100-fold) was observed along with a blue shift in the PL maximum from 615 nm to 555 nm. To have a more accurate evaluation of the AIE characteristics, we have measured and plotted the quantum yields of TPA-BI in mixtures with different water fractions by an integrating sphere. 20 The plot shows a similar trend with I/I 0 (Fig. S7A †). The fluorescence quantum efficiency of TPA-BI in a 70% aqueous mixture was 22%, which was appreciably high for an orange emitter. Due to its poor solubility in water, in solution with high water fractions, aggregates of TPA-BI would be formed. This greatly restricts the intramolecular motion and activates the AIE process. The domination of the AIE effect over the TICT effect results in an increase in the PL intensity. 9a,11 Surprisingly, the emission became weaker and was slightly red-shifted again when the water fraction increased from 70% to 90%. This may be attributed to (1) the crystallization-induced emission feature of TPA-BI and ( 2) the effect of the aggregate size. 21 TPA-BI may form crystalline aggregates at low water fractions. At water fractions above 70%, the fast aggregation of the TPA-BI molecules will form less emissive, redder amorphous species with smaller sizes, as confrmed by the DLS results (Fig. S7B †). The emission of small-sized aggregates may be more vulnerable to being affected by the surrounding solvent environment, leading to an emission drop and red-shift at high water fractions. ## Two-photon excited uorescence TPA-BI possesses a conjugated structure with strong electron donating and withdrawing groups and thus it is expected to exhibit strong 2PA. The 2PA of TPA-BI was studied using a TPEF technique with a femtosecond pulsed laser source, and the relative TPEF intensity in different solvents was measured using Rhodamine 6G and fluorescein as the standards. 22 The measured wavelength was varied from 720 to 920 nm at an interval of 40 nm and the d 2PA values were obtained. The results are summarized in Fig. 3 and Table S2. † In THF, the maximum d 2PA value (159 GM) was obtained at 840 nm. In various solvents, the highest d 2PA was obtained in diethyl ether and was equal to 213 GM, which was much higher than those of most fluorescent proteins (usually < 100 GM, only 39 GM for EGFP), 23 synthetic BI derivatives (<40 GM), 15c and BODIPY dyes (82-128 GM). 24 Thus, TPA-BI may serve as a good two-photon imaging probe to living cells. Apart from 2PA, the TPEF of TPA-BI under different laser powers was also studied. The plot of the fluorescence intensity against the excitation laser power gave a linear line with a slope of 1.911, confrming the occurrence of two photon absorption (Fig. S8 †). 19c When excited by laser light at 840 nm, TPA-BI emitted intense PL at 447-619 nm in solvents with different polarities, suggesting the TICT feature even under the condition of two-photon excitation (Table S2 †). The spectral patterns resemble the one-photon ones, revealing the same excited state for the radiative decay processes (Fig. 3B). The TPEF cross sections (d 2PEF ) are crucial parameters for biomedical imaging and are provided in Table S2. † The high d 2PEF values in different solvents suggest that TPA-BI possesses a promising potential application in the biomedical feld. ## One-photon LD imaging To explore the application of TPA-BI in living cell imaging, its cytotoxicity was frstly evaluated using 3-(4,5-dimethyl-2thiazolyl)-2,5-diphenyltetrazolium bromide (MTT) assay under different dye concentrations. As suggested in Fig. S9, † no signifcant variation in the cell viability was observed even when a high dye concentration of 20 mM was used. This indicates that TPA-BI shows almost no cytotoxicity to living cells and possesses a good cell biocompatibility. Cell imaging experiments were then carried out by incubating HeLa cells with 1 mM of TPA-BI for 15 min followed by examination under a fluorescence microscope at an excitation wavelength of 400-440 nm. As shown in Fig. 4A, the lipophilic TPA-BI was prone to accumulating in the hydrophobic spherical LDs with bright greenish-blue emission due to the "like-like" interactions. Compared with BODIPY 493/503, a commercial probe for LD imaging, the images stained by TPA-BI showed a lower background signal, thanks to its AIE feature. Colocalization of TPA-BI and BODIPY 493/503 was performed and the same patterns were obtained solely by TPA-BI or BODIPY 493/ 503 with good overlap, demonstrating a good specifcity of TPA-BI to LDs (Fig. 4C-E and S10 †). Besides a high LD specifcity, TPA-BI also showed an excellent resistance to photo-bleaching. More than 80% of its fluorescence signal was retained even when it was continuously irradiated by laser light for 50 scans (Fig. S11 †). Such a high photostability is comparable to that of BODIPY 493/503. 25 TPA-BI can also be utilized in LD imaging in other cells lines, such as HepG-2 and A549, and in fxed cells (Fig. S12 †). In addition, a negligible emission color change was observed with the increase of the dye concentration, oleic acid concentration or incubation time of oleic acid (Fig. S13 and S14F †). However, more and larger lipid droplets were observed after increasing the concentration or incubation time of oleic acid, and the fluorescence intensity of the whole cell was increased (Fig. S13 and S14A-E †). The statistical results were further confrmed by flow cytometry using BODIPY 493/503 and TPA-BI for staining (Fig. S15 †), suggesting that TPA-BI can be practically applied in the quantitative analysis of LDs by flow cytometry. All of these results demonstrate that TPA-BI indeed acts as a superior probe for LD imaging and analysis bearing wide applications in biomedical research and clinical diagnosis. Why does TPA-BI exhibit greenish-blue emission in LDs? To understand this, we measured its fluorescence using a confocal microscope in the mode of wavelength scanning. In Fig. 4F, the fluorescence spectrum exhibits a peak at 495 nm. The peak value reflects the value of E T (30) of the environment and suggests a low polarity inside the LDs. This is understandable as the LDs are surrounded by a phospholipid monolayer and consist of various neutral lipids such as triacylglycerol and cholesteryl ester. 1a To further verify our claim, we carried out the analogue experiment outside cells using the major components in LDs such as 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and trioleate glycerol (TAG). Without DMPC and TAG, the aggregates of TPA-BI in phosphate buffered saline (PBS) solution emitted orange coloured light at 570 nm, while the emission colour and intensity blue-shifted and increased slightly upon the addition of DMPC only. Further addition of TAG resulted in an abrupt increase in the emission intensity and a peak maximum (Fig. 4G). This should be ascribed to the TICT effect of TPA-BI since TAG is more hydrophobic and less polar than DMPC. Under two-photon excitation, the fluorescence intensity of TPA-BI increased more than 10-fold upon the addition of TAG, and the extent of this was higher than that achieved by one-photon excitation (Fig. S16 †). This suggests a larger signal to noise ratio for LD imaging by two-photon excitation to allow better contrast. Thus, two-photon excitation clearly outperforms in LD imaging. ## Two-photon LD imaging As discussed above, TPA-BI shows a large d 2PA of up to 213 GM and its response to the TAG/LDs is largely enhanced under twophoton excitation. To evaluate whether TPA-BI is suitable for two-photon imaging of the LDs, we compared its performance with commercial BODIPY 493/593. As seen in Fig. 5, sufficient signals were obtained for both TPA-BI and BODIPY 493/503 under one-photon excitation. While clear images of the LDs stained by TPA-BI were still observed at an excitation wavelength of 840 nm, almost no signal was obtained for the LDs stained by BODIPY 493/503 under the same conditions. Similar results were obtained even when the excitation wavelength was changed to 900 nm and 980 nm (Fig. S17 †). Thus, TPA-BI is more suitable for two-photon imaging as it can be excited readily by laser light with a low power, thus avoiding photothermal damage to living cells caused by high laser power. Several experiments were then conducted to demonstrate the superior advantages of two-photon microscopy (2PM) over onephoton microscopy (1PM) which are better 3D resolution, lesser photobleaching and autofluorescence and deeper penetration depth. As shown in Fig. 6A, clustered LDs in HeLa cells were observed with a blurred background by 1PM. The blurred background is believed to be caused by the fluorescence of the LDs below and above the focus plane. This problem was solved by 2PM due to the intrinsic sectioning property of 2PM. While a small layer of fluorophores was excited at the focus plane in 2PM, all of the fluorophores were excited in the light pathway in 1PM. Thus, fewer fluorophores were photobleached in 2PM during prolonged observation. To prove this, an experiment was carried at a low concentration of TPA-BI (1 mM) to enable the occurrence of photobleaching. As shown in Fig. 6B, while almost 100% of the signal intensity was retained in 2PM, only half was retained in 1PM. Autofluorescence is a well-known difficult problem in tissue slices, which often leads to a low image contrast and is even detrimental to dyes with low emission intensity. Intense auto-fluorescence was observed in the fxed liver tissue slice by 1PM, which was largely reduced by 2PM (Fig. 7A and B). After staining with TPA-BI, clear spherical spots with intense fluorescence were observed with a much lower background than with 1PM (Fig. 7C and D). Due to the lesser absorption and scattering of the near-infrared light in the tissue, 26 the longer excitation light (840 nm) in 2PM is believed to have a deeper penetration depth than that of one-photon excitation (442 nm). The fluorescent signal of the spherical spot could be detected at a z depth of 45 mm (Fig. 8). Compared to our previous LD-specifc AIE bioprobes, 9,11 TPA-BI not only exhibits the merits of AIE probes in 1PM but also performs well in 2PM with a large d 2PA and NIR excitation, exhibiting higher 3D resolution, lower photobleaching rate, reduced auto-fluorescence and low damage to living cells. This makes TPA-BI suitable for LD imaging both in cells and tissue slices with two-photon excitation, providing another tool for tissue slice-based disease diagnosis of lipid droplets. ## Conclusion In this work, an AIE probe (TPA-BI) for LD imaging was rationally designed and synthesized. Due to its D-p-A structure, TPA-BI exhibited solvatochromism with a high sensitivity to environmental polarity. TPA-BI exhibited both TICT and AIE features, showing a large Stokes shift of up to 202 nm and a large 2PA cross section of up to 213 GM. TPA-BI demonstrated good cell biocompatibility, high brightness, low background, high selectivity and excellent photostability. The lipid droplet imaging in TPA-BI was applicable for various live cell lines and fxed cells. It also allowed LD analysis by flow cytometry. Compared to commercial BODIPY dyes, TPA-BI was more suitable for two-photon imaging of LDs with the merits of higher 3D resolution, lesser photobleaching and autofluorescence and deeper penetration in tissue, providing a promising imaging tool for LD tracking and analysis in biomedical research and clinical diagnosis. Due to its high sensitivity to polarity and good 2PA cross section, TPA-BI can be further utilized to detect the localized polarity of samples with two-photon excitation in a mixed bulk sample, such as for indicating the phase separation in polymer blends. Because of its synthetic accessibility, further modifcation of TPA-BI for imaging of other cell organelles or bio-sensing is under investigation in our laboratories.

chemsum

{"title": "Two-photon AIE bio-probe with large Stokes shift for specific imaging of lipid droplets", "journal": "Royal Society of Chemistry (RSC)"}

thienoisoindigo-based_semiconductor_nanowires_assembled_with_2-bromobenzaldehyde_via_both_halogen_an

## Abstract: We fabricated nanowires of a conjugated oligomer and applied them to organic field-effect transistors (OFETs). The supramolecular assemblies of a thienoisoindigo-based small molecular organic semiconductor (TIIG-Bz) were prepared by co-precipitation with 2-bromobenzaldehyde (2-BBA) via a combination of halogen bonding (XB) between the bromide in 2-BBA and electron-donor groups in TIIG-Bz, and chalcogen bonding (CB) between the aldehyde in 2-BBA and sulfur in TIIG-Bz. It was found that 2-BBA could be incorporated into the conjugated planes of TIIG-Bz via XB and CB pairs, thereby increasing the π − π stacking area between the conjugated planes. As a result, the driving force for onedimensional growth of the supramolecular assemblies via π − π stacking was significantly enhanced. TIIG-Bz/2-BBA nanowires were used to fabricate OFETs, showing significantly enhanced charge transfer mobility compared to OFETs based on pure TIIG-Bz thin films and nanowires, which demonstrates the benefit of nanowire fabrication using 2-BBA. Organic transistors have been developed over the last decades as next-generation electronic devices owing to their versatility for flexible devices and facile fabrication process compared to rigid inorganic semiconductors 1,2 . In particular, one-dimensional (1D) organic nano/micro-wire transistors have attracted significant attention due to their excellent electronic properties, including high charge-carrier mobility in specific directions . Various strategies have been used to prepare organic semiconducting wires, including an ink-jet method 7,8 , electro-spinning 9 , hard and soft template tools , self-assembly in solution and physical vapor transport 20,21 . The excellent charge-carrier mobilities were obtained for self-assembled single crystals of organic semiconductors, because their closely assembled structures by intermolecular interactions enable electrons to move efficiently through a conjugated backbone plane 16,22 . Organic semiconductors are prone to be crystallized due to the presence of π − π stacking of the conjugated main backbones, which provides the driving force for 1D growth. In general, the self-assembly of single crystalline based on nano/micro-wires of the organic semiconductors shows the strong dependence of the solubility on recrystallization temperature 17,19,23 and solvent type 18,21,24 , which is usually determined by the molecular structure of the organic semiconductors 16,25,26 . Organic semiconductors usually require flexible alkyl side chains to facilitate solution processing 27 . However, the alkyl side chains increase the tendency for lateral packing of the organic semiconductors and disturb 1D growth in the direction of the π − π stacking of conjugated planes during precipitation or recrystallization, resulting in spherical morphologies due to hydrophobic interactions in the lateral direction of the conjugated planes 28 . Therefore, enhancing the driving force for π − π stacking so that π − π interactions between conjugated planes dominate lateral hydrophobic interactions between alkyl chains is critical for preparing nano/micro-organic semiconductor wires. Halogen bonding (XB) is a strong and tunable form of non-covalent bonding between halogen atoms and negative sites, with bonding strengths of 10-150 kJ/mol 29 . Since the XB originates from the attraction between the partially positive charged region of halogen atoms, the σ hole, and its counterpart in electron-rich donor atoms, it has been widely used in supra-molecular assembly and crystal engineering of self-assembled structures . The halogen atoms have dipolar charges; the π hole is perpendicularly negative, while the σ hole is horizontally positive. These charges simultaneously attract electron-rich functional groups, including lone-pair electrons in the same plane, and electron-poor groups in the vertical direction 36 . Furthermore, the XB should benefit for electronic device applications because it does not involve acidic or basic groups that can trap charges and disturb charge transfer, which is clearly different to hydrogen bonding 37 . Hence, the XB is considered an important intermolecular interaction for co-crystallization with conjugated oligomers or polymers for organic electronic devices . Chalcogen bonding (CB) is a newly identified type of weak non-covalent interaction, recently described by using various modern characterization techniques, such as nuclear magnetic resonance, X-ray photoelectron spectroscopy (XPS), and X-ray diffraction (XRD). Despite its weak interaction, the CB can play a dominant role in crystal design as it has a controllable binding strength of a few to hundreds of kJ/mol 49 . The CB involves highly polarizable and electronegative atoms (i.e., chalcogen atoms (O, S, Se, Te) which have σ holes and π holes similar to halogen atoms. In this way, the σ hole, the positive part of chalcogen atoms at the opposite side of the covalent bond, can combine with the electron donor and bind via electrostatic interaction 50 . In most chalcogen-containing organic semiconductors, the CB plays an important role in facilitating charge transport by the improving electronic delocalization of their backbones 41,50 . Although it has been proven that CB or XB is a useful interaction for molecular assembly, most approaches using CB or XB for transistor devices based on organic semiconductors have focused on intramolecular interactions in planar conjugated planes and consequently high carrier mobility . Semiconducting oligomers or polymers bearing chalcogen or halogen atoms can have enhanced intramolecular electron delocalization and narrow optical band gaps due to the planar structure of their backbones. For example, the thienoisoindigo (TIIG) unit has been used as a promising building block to construct high-mobility organic semiconductors for organic field-effect transistors (OFETs). Advantages of this material include strong π − π stacking with a large overlapping area induced by a combination of the planar backbone structure and intramolecular CB, where improved electronic delocalization is derived from the quinoidal structure 51,52 . At present, it is hard to find molecular assemblies of organic semiconductors utilizing intermolecular XB and CB. In this study, we fabricated nanowire assemblies of a TIIG-based organic semiconductor, (E)-2,2′-diphenyl-4,4′-bis(2-ethylhexyl)-[6,6′-bithieno [3,2-b]pyrrolylidene]-5,5′(4 H,4 H′)-dione, TIIG-ethylhexyl benzene (TIIG-Bz), which was able to simultaneously engage in XB and CB interactions with 2-bromobenzealdehyde (2-BBA), and demonstrated their use for organic field-effect transistor (OFET) applications. We examined the TIIG-Bz/2-BBA nanowire assemblies using XPS, powder XRD (PXRD), grazing incidence XRD (GIXD), and XPS, and investigated the influence of 2-BBA on nanowire assemblies and charge transfer mobility of OFETs. ## Results and Discussion Nanowires of TIIG-Bz and supra-molecular nanowire assemblies of TIIG-Bz and 2-BBA were fabricated using the bisolvent phase transfer method 28,32,53 , shown schematically in Fig. 1. This method is a popular fabrication process for nano-wire assemblies because it optimally accommodates molecules with different structures and sizes. In the process, a solution with organic semiconductors dissolved in a good solvent is placed in a vial, then a poor solvent is carefully poured onto the solution. Despite quite good miscibility between the poor and good solvents, their density difference kept them from immediately mixing with each other. Instead, at the interface of the two miscible solvents, conjugated organic molecules dissolved in the good solvent slowly self-assemble as the two solvents mix and the solubility of the organic semiconductor decreases. Consequently, self-assembled nanowires are formed at the interface 28,32,53 . In this study, for supra-molecular assemblies, TIIG-Bz and 2-BBA were dissolved in chloroform (d = 1.4459 g/cm 3 at 20 °C), a good solvent for both chemicals, and co-precipitated at the interface between chloroform and methanol (d = 0.7914 g/cm 3 at 20 °C) upon the inter-diffusion of chloroform and methanol. As shown in Fig. 2(a), nanowires of only TIIG-Bz with a thickness ~1-2 μm (60-200 μm in length) were formed via this bisolvent phase transfer process. The crystals of TIIG-Bz molecules preferentially grow in one direction, even without the use of 2-BBA, due to their molecular rigidity and planarity that enhance intermolecular π − π interactions upon exposure to methanol when dissolved in chloroform. However, it should be noted that some TIIG molecules were aggregated in a bundle of wires and formed much thicker wire groups or particles (Fig. S1), indicating inhomogeneous self-assembly in both planar and vertical direction of conjugated plane due to competition between π − π stacking of conjugated planes and lateral packing of ethylhexyl chains. When TIIG-Bz was assembled with 2-BBA via bi-phase separation, supra-molecular nanowires were successfully assembled with a higher aspect ratio, thickness ~800 nm, and length ~180 µm, than the TIIG-Bz nanowires assembled without 2-BBA, as shown in Fig. 2(b,c). During fabrication of TIIG-Bz and 2-BBA, both chloroform and methanol are good solvents for 2-BBA and we used the same concentration of 2-BBA in both the chloroform and methanol phases to prevent interphase diffusive mass transfer of 2-BBA due to concentration differences. Thus, the diffusive mass transfer of TIIG-Bz from the chloroform phase into the methanol phase was induced by a TIIG-Bz concentration gradient and enabled the formation of supra-molecular assemblies at the interface between the chloroform and methanol phases with intermixing of the solvents, followed by the decreased solubility of TIIG-Bz. The optical properties of TIIG-Bz dissolved in chloroform and its nanowires dispersed in methanol were examined by UV-Vis-NIR spectroscopy. When TIIG-Bz was dissolved in the chloroform solution, the maximum absorption band appeared at 616 nm with a shoulder at around 650 nm (Fig. 3). In comparison, the absorption spectra of TIIG-Bz assemblies with and without 2-BBA dispersed in methanol showed increasing absorption intensity in the NIR region of 700-1000 nm. The maximum absorption peaks of the TIIG-Bz solution, TIIG-Bz-only, and TIIG-Bz/2-BBA nanowires were all observed at ~616 nm. However, the shoulder band around 700 nm in the spectrum of the TIIG-Bz solution was significantly red-shifted to near 800 nm for the assembled intermolecular structure as shown in Fig. 3(a). In addition, the absorption intensity of the shoulder band in the TIIG-Bz/2-BBA assembly spectra for wires were strongly enhanced compared to those of the pristine TIIG-Bz samples. Such absorbance enhancement of the TIIG-Bz/2-BBA in the NIR region of the spectra was clearly shown even after excluding any scattering effect by measuring the spectra with an integrating sphere. The higher absorbance of the TIIG-Bz wires assembled with 2-BBA in the NIR region was consistent with our recent results of density functional theory calculations for conjugated polymer nanoparticles 54,55 . Such a red shift in the absorption spectra originates from enhanced intermolecular π − π interactions with energy level adjustments, and has been widely observed in thin films for organic optoelectronic devices or nanomaterials of conjugated polymers . Furthermore, this enhancement in NIR absorption clearly indicated improved intermolecular π − π interactions in the direction of nanowire growth, which was significantly enhanced with increasing nanowire length, showing better aggregation. Molecular assembly structures of TIIG-Bz and TIIG-Bz/2-BBA nanowires characterized by PXRD showed clearly different diffraction patterns depending on the incorporation of 2-BBA into TIIG-Bz assemblies. The 2D PXRD pattern of TIIG-Bz nanomaterials (right inset of Fig. 4(a)) shows ring patterns in the small-angle range and short dashes in the wide-angle range up to 2θ = 70° that were symmetric to the vertical, central line as the measurements were made over 360°. These short dashes are a typical feature of single crystals. On the other hand, the TIIG-Bz/2-BBA nanowires showed powder-like features, i.e., peaks with identical intensities along the Debye ring (right inset of Fig. 4(b)). Reflecting these characteristics of the 2D PXRD patterns, the circular averaged 1D profile of the TIIG-Bz/2-BBA nanowires (Fig. 4(b)) showed weakened peaks in the wide-angle range above 2θ = 30°, while that of TIIG-Bz nanomaterials showed peaks up to 2θ = 70° (Fig. 4(a)). It should be noted that identification of all crystallographic peaks in our PXRD data was difficult as both the TIIG-Bz nanomaterials and TIIG-Bz/2-BBA nanowires may not have been single-phase. As shown in the SEM image of Figure S1, TIIG-Bz-only nanomaterials were a mixture of wires, needles, and particles. In addition, the TIIG-Bz/2-BBA nanowires may have been a mixture of assemblies of TIIG-Bz, TIIG-BZ and one 2-BBA, and TIIG-Bz and two 2-BBA molecules due to inhomogeneous 2-BBA distribution. Detailed peak assignment is being performed by preparing single crystals or supramolecular assemblies with homogeneous morphology (this study will be published separately). However, at present, it is obvious that a new structure is developed upon the addition of 2-BBA into the TIIG-Bz assembly, as shown in the left inset of Fig. 4. In the inset 1D PXRD pattern (below 2θ = 10°), TIIG-Bz nanomaterials showed five diffraction peaks at 2θ = 5.563°, 5.933°, 7.121°, 7.734°, and 9.023°. In comparison, new diffraction peaks (indicated by the arrows) were obvious in the 1D PXRD pattern of TIIG-Bz/2-BBA nanowires, where six diffraction peaks appeared at 2θ = 5.563°, 5.716°, 5.933°, 7.146°, 7.261°, and 9.023° (Figure S2). The diffraction peaks of both TIIG-Bz and TIIG-Bz/2-BBA assemblies were assigned to a triclinic lattice structure (Figure S2 S1) increased with the addition of 2-BBA into the TIIG-Bz assembly. The PXRD results indicated that 2-BBA molecules in TIIG-Bz/2-BBA nanowires limit the growth of TIIG-Bz-only crystals and make the crystal structure of the original TIIG-Bz assembly a bit larger. As shown in the SEM image in Fig. 3(b) and (c), it seems that the addition of 2-BBA is advantageous for 1D nanowire growth of TIIG-Bz assembly. The existence of both XB and CB in TIIG-Bz/2-BBA nanowires was confirmed by analyzing XPS spectra, as shown in Fig. 5. The changes of the binding energies in the XPS spectra for each atom and compound are summarized in Table 1. XB and CB are intermolecular interactions between an electron donor and an electron accepter that partially give and take electron pairs. Bromine or chalcogen atoms that have the electron-deficient σ hole can act as the electron accepter in this interaction; these elements can be bound to electron-rich oxygen in carbonyl groups and partially receive electrons via XB and CB. As a result, the electrons of these elements are increased and the binding energies of each electron become weaker. The decrease in the binding energy of the bromine atom in 2-BBA was observed upon assembly of 2-BBA with TIIG-Bz, as shown in Fig. 5(a). The binding energy of bromine 3d 5/2 in 2-BBA was 70.86 eV and that of the TIIG-Bz/2-BBA nanowires decreased by 1.60 eV to 69.26 eV, which is a characteristic of XB 62,63 . In addition, the binding energies of the chalcogen atom, sulfur in TIIG-Bz, 2p 3/2 , and 2p 1/2 decreased from 164.02 and 165.20 to 163.94 and 165.12 eV by 0.08 eV, respectively, as shown in Fig. 5b. The larger change in binding energies of bromine compared to sulfur was due to the stronger binding of XB than CB as the σ hole of the bromine atom is more positive than that of sulfur; this is consistent with a recent report related to the XB mechanism 62 . On the other hand, the binding energies of N 1 s in both TIIG-Bz and TIIG-Bz/2-BBA nanowires were the same (399 eV), as shown in Fig. 5(c). This result is explained as follows. It was shown previously that the electron donor of XB and CB is usually a non-covalent electron pair of nitrogen 30,64,65 . However, in TIIG-Bz, the non-covalent electrons of nitrogen are delocalized in the thienoisoindigo moieties and are oriented perpendicular to the conjugated plane. In addition, ethyl hexyl chains sterically hinder the close association of the nitrogen atom with 2-BBA. Hence, the nitrogen in TIIG-Bz is unable to bind with electron acceptors via XB or CB. Instead of nitrogen, electron-rich oxygen in the aromatic carbonyl of TIIG-Bz can contribute to XB or CB as electron donors. However, it was difficult to confirm the effect of the addition of 2-BBA in the TIIG-Bz composite using the O 1 s binding energy spectrum due to the presence of multiple external peaks, including methanol or water (~534 eV), carbon dioxide (~532 eV), and the Si substrate (~532 eV). Figure 5(d) shows peaks at 530.76 eV that were attributed to contributions from O 1 s electrons in the aromatic carbonyls of TIIG-Bz. The strongest peak for both TIIG-Bz and TIIG-Bz/2-BBA nanowires was at 532.14 eV and was assigned to the Si-O peak from the SiO2 wafer substrates. Noticeably, the binding energies of O 1 s electrons in TIIG-Bz increased by 0.08 eV (from 530.76 eV to 530.84 eV) where their intensities significantly decreased, indicating electron donation from O 1 s electrons to XB or CB. Meanwhile, O 1 s electrons in the aliphatic carbonyl group of 2-BBA should have a binding energy at 532.5 eV 66,67 . However, their binding energy increased by 0.46 eV to 532.96 eV when 2-BBA was assembled with TIIG-Bz (Fig. 5(d)). One of the most plausible explanations for these peak changes in the O 1 s spectra is that oxygen in the aromatic carbonyl of TIIG-Bz interacts with the σ hole in bromine of 2-BBA via XB, while the other oxygen in the aliphatic carbonyl of 2-BBA reacts with the σ hole in sulfur of TIIG-Bz via CB. Thus, XPS characterization suggested that one TIIG-Bz and two 2-BBA molecules were associated with each other via both XB and CB in supramolecular assemblies of nanowires, forming slipped one-dimensional stacks 68 , as schematically illustrated in Fig. 1. The configuration of one TIIG-Bz and two 2-BBA molecules in supramolecular assemblies (as shown in Fig. 1) was indirectly verified by measuring charge transfer efficiencies of OFETs based on thin films of TIIG-Bz and 2-BBA mixtures with various mixing ratios from 1:1 to 1:5. The devices had a BG/TC structure. Figure 5(a) and (b) show the output and transfer curves of an OFET based on a pristine TIIG-Bz thin film with a low hole mobility of 2.34 × 10 −4 cm 2 /Vs. In comparison, the output and transfer curves of a thin film OFET fabricated with a 1:2 molar mixing ratio of TIIG-Bz to 2-BBA (Fig. 6(c) and (d)) exhibited the highest hole mobility of 6.39 × 10 −4 cm 2 / Vs (about three times higher when we measured the hole mobilities of the OFETs with the different mixing ratios (Fig. 6(e) and Table 2; average values from seven devices); representative transfer and output curves are shown Fig. S3. This optimal molar mixing ratio was used to fabricate thin films of TIIG-Bz and 2-BBA, which clearly indicated that the TIIG-Bz/2-BBA film was the most well-ordered and the structure with one TIIG-Bz and two 2-BBA molecules was optimum for charge carrier transfer as TIIG-Bz has two positions available for interaction with 2-BBA via XB and CB. It should be noted that the assembly structure of the pristine TIIG-Bz film was not significantly changed by mixing it with 2-BBA for spin coating the film. GIXD profiles of the pristine TIIG-Bz film and TIIG-Bz/2-BBA with a 1:2 mixing ratio (Fig. S4) showed identical diffraction peak positions, indicating no significant differences in the lattice structures. However, the profile of the TIIG-Bz/2-BBA film showed stronger ordering in the assembly structure than that of the pristine TIIG-Bz film. These results suggested that the spin-coating time was not sufficient to produce a new lattice structure by the assembly of 2-BBA with TIIG-Bz, while the addition of 2-BBA enhanced structural ordering of TIIG-Bz assembly. The enhanced ordering at the optimal ratio of 1:2 strongly supports the hypothesis that two 2-BBA molecules interacted with one TIIG-Bz molecule. We propose that simultaneous XB and CB bonding between TIIG-Bz and 2-BBA increased the overall area of the conjugated plane, resulting in enhancement of the intermolecular transfer of charge carriers via the strongly ordered molecular structure. In the case of higher 2-BBA contents (i.e., the 1:3, 1:4, and 1:5 ratios), it also seemed that unbounded 2-BBA acted as an impurity on the TIIG-Bz/2-BBA complex and interfered with the formation of well-ordered complexes, resulting in reduced charge transfer mobility. The applicability of supramolecular nanowire assemblies based on TIIG-Bz and 2-BBA was demonstrated by fabricating organic nanowire FETs. We fabricated OFETs using both nascent TIIG-Bz nanowires and TIIG-Bz/2-BBA nanowires, and investigated the intrinsic charge transport properties. We used the BG/TC device structure for facile fabrication, as schematically illustrated in Fig. 7(a) and shown by the SEM image in Fig. 7(b). The basic transistor parameters for the devices are summarized in Table 2. The OFET based on the TIIG-Bz nanowires showed a hole mobility of 3.50 × 10 −3 cm 2 /Vs, about one order of magnitude higher than OFETs based on the pristine TIIG-Bz thin films (2.34 × 10 −4 cm 2 /Vs), as shown in the representative transfer curve of the TIIG-Bz nanowire OFET device (Fig. 7(c)). The nanowires in the OFET contributed to increasing device performance by improving hole transport mobility via the nanowires compared to the thin-film-based OFET. The improved mobility was due to one-directional assembly causing the holes to move directly to the cathode, as is widely accepted 3,4,9 . The OFET based on TIIG-Bz/2-BBA nanowires showed an even higher hole mobility (9.34 × 10 −3 cm 2 /Vs), three times higher than that of the OFET based on TIIG-Bz nanowires, as shown in Fig. 7(d) and Table 2. The highest hole transfer mobility was 0.01146 cm 2 It should be noted that there was excess 2-BBA in the chloroform solution, ten times more than the molar content of TIIG-Bz in the preparation of TIIG-Bz/2-BBA nanowires. However, the excess 2-BBA in the nanowire preparation process did not deteriorate the device characteristics, but rather, significantly improved the mobility. This suggests that the excess 2-BBA available for dissolution in the good solvents (chloroform and methanol) during bisolvent phase transfer process was excluded from the formation of TIIG-Bz/2-BBA composite nanowires without interfering with the formation of the well-ordered supra-molecular assembly. Overall, it was confirmed that the OFET based on TIIG-Bz/2-BBA nanowires assembled via XB and CB had significantly enhanced hole mobility compared to OFETs based on thin films and nanowires of TIIG-Bz; this was due to the supramolecular structure with preferred 1D orientation resulting in efficient charge carriers transfer. It should be mentioned that the charge transfer mobilities reported here were lower than or similar to those of TIIG-Bz-based OTFTs employing a different device structure, such as top gate/bottom contact and different dielectric materials 51 . However, these preliminary results offer ## Conclusion We demonstrated a supramolecular nanowire assembly of a TIIG-based organic semiconductor with 2-BBA and their FET device characteristics. The analysis of XPS data strongly supported the insertion of two 2-BBA molecules into the conjugated plane of one TIIG-Bz molecule via simultaneous XB and CB between TIIG-Bz and 2-BBA. This supramolecular arrangement was further supported by XRD analysis and the fact that the highest hole mobility of OFETs based on thin films was observed for a molar ratio of TIIG-Bz to 2-BBA of 1:2. When nanowires of TIIG-Bz were assembled via a bisolvent phase transfer method, an increased conjugated plane area and the insertion of 2-BBA enhanced intermolecular π − π interactions. The enhanced π − π interactions is more dominant than lateral packing by hydrophobic interactions of alkyl chains, promoting 1D crystal growth and narrow nanowire width. It was clear that the increased conjugated area of the TIIG-Bz/2-BBA supramolecular assembly allowed efficient hole transport in the crystallization direction, although the lattice structure of the original TIIG-Bz assembly become larger with incorporation of 2-BBA. The formation of TIIG-2BBA supramolecular structure suggests a new method simultaneously using XB and CB to control the self-assembled morphology and crystalline feature of organic semiconductors with increasing intermolecular interaction, improving crystallization, and achieving high charge transport mobility in OFETs. ## Methods Materials. TIIG-Bz (MW = 650.94 g/mol) was synthesized using a Pd-catalyzed Suzuki coupling reaction between (E)-2,2′-dibromo-4,4′-bis(2-ethylhexyl)-[6,6′-bithieno [3,2-b]pyrrolylidene]-5,5′(4 H,4 H′)-dione and phenyl boronic acid. Full details of the synthesis and characterization were published in our previous work 52 . 2-BBA (MW = 185 g/mol) were purchased from Alfa Aesar (United Kingdom) and used as received. Chloroform and methanol were obtained from the Dae-Jung Reagent Chemical Company (South Korea). Preparation of TIIG-Bz and 2-BBA nanowires. TIIG-Bz and 2-BBA nanowires were prepared via a bisolvent phase transfer method, as schematically illustrated in Fig. 1, following a published procedure 28,32,53 . TIIG-Bz (1.95 mg; 3 μmol) and 2-BBA (5.55 mg; 30 μmol, 10 times the molar content of TIIG-Bz) were individually dissolved in chloroform (5 mL) in two separate vials. We placed 0.5 mL of each solution in a 70 mL vial and sonicated it for 10 min to ensure complete mixing. Then, 30 mL of a methanolic 2-BBA solution with the same 2-BBA concentration as the mixed chloroform solution (0.555 mg/mL CHCl 3 ) was slowly and carefully added to the chloroform mixture containing TIIG-Bz and 2-BBA to maintain separated methanol (top) and chloroform (bottom) phases. The bi-solvent mixture was set aside for 16 h and nanowires of TIIG-Bz and 2-BBA self-assembled at the interface during diffusive mixing of the two solvents due to low solubility of TIIG-Bz. The prepared nanorods were filtered through a 0.2 μm membrane and re-dispersed in methanol to remove any residual 2-BBA. Device fabrication. The organic nanowire FETs were fabricated by spreading 0.5 mL of the nanowire methanol solution over an n-doped Si/SiO 2 substrate. Shadow masks yielding a channel width of 1500 μm and channel length 100 μm were overlaid on the nanowires in the vertical direction of the channel and gold electrodes with a thickness of 140 nm were deposited to complete the bottom-gate top-contact (BG/TC) devices. For comparison, BG/TC OFETs based on thin films of only TIIG-Bz and TIIG-Bz/2-BBA mixtures were fabricated by spin casting (1500 rpm, 30 s, ramp-up speed: 0.1 sec) a 2.5 mg/mL TIIG-Bz chlorobenzene solution and TIIG-Bz/2-BBA chlorobenzene solutions with different mixing ratios onto Si/SiO 2 substrates. For a fixed amount of TIIG-Bz (2.5 mg/mL), five different 2-BBA contents were mixed at molar ratios from 1:1, 1:2, 1:3, 1:4 and 1:5 (0.71, 1.42, 2.13, 2.84, and 3.55 mg/mL of 2-BBA, respectively). The TIIG-Bz films were then annealed at 80 °C for 20 min, followed by deposition of 70-nm-thick gold electrodes via thermal evaporation through a metal shadow mask, yielding a channel length and width of 150 μm and 1500 μm, respectively. The field-effect mobility was determined in the saturation regime using the following relationship. where I DS is the saturation drain current, L is the channel length, W is the nanowire width, C is the capacitance (~35 nF/cm 2 ) of the SiO 2 dielectric (100 nm), V G is the gate bias, and V th is the threshold voltage. The device performance was evaluated in air using an HP4156A Precision Semiconductor Parameter Analyzer (Agilent, U.S.A) . Characterization. The optical properties of TIIG-Bz and its nanowire assemblies with 2-BBA were examined with an ultraviolet-visible (UV-Vis) spectrometer (V-670, JASCO, USA). The UV-vis-NIR absorption spectra were also obtained using a JASCO V-670 UV-vis-NIR spectrophotometer equipped with a 60 mm BaSO 4 -coated integrating sphere, to prevent absorption spectra distortion due to light scattering by the nanowires. The measurements of UV-vis-NIR absorption spectra were carried out in a 1-cm path-length quartz optical cell. The polymer nanowire solution was dispersed in methanol (spectroscopy-grade), and ultrasonication was applied for 10-20 secs before the absorption measurements. A field-emission scanning electron microscope (FE-SEM; sigma, Carl Zeiss, USA) and an atomic force microscope (AFM; XE-100, PSIA, South Korea) were used to observe the morphologies of the nanostructures. Structures of the TIIG-Bz assemblies with 2-BBA were analyzed in detail using PXRD and GIXD experiments at a synchrotron facility (PLS-II 6D UNIST-PAL beamlines at Pohang Accelerator Laboratory, South Korea). For PXRD measurements, dry samples of TIIG-Bz and TIIG-Bz/2-BBA assemblies were placed into polyimide capillary tubes and continuously rotated during the measurements. For GIXD measurements, methanol dispersions of the assemblies were dropped onto slices of silicon wafers and dried. The molecular spacing and packing orientation relative to the substrates were characterized using GIXD profiles. XPS was carried out using an AXIS Ultra DLD instrument (Kratos, U.K.) in an advanced in-situ surface analysis system (AISAS; KBSI, Korea) operating at a base pressure of 1.6 × 10 −10 mbar at 300 K with a monochromatic Al Kα line at 1486.69 eV. The samples were prepared on a Si wafer substrate, or a Si wafer substrate coated with a 140-nm-thick gold layer, which were electrically grounded to the sample stage. The binding energy scales were calibrated by the C 1 s core level position at 284.8 eV as an internal reference and the Fermi edge of a gold standard. Survey and narrow spectrum scans were obtained with analyzer pass energies of 160 and 40 eV, respectively at 150 W. In order to separate the chemical bonding states in the spectra, the spectral line shape was simulated using Casa XPS software using a Shirley background and a GL (30) line shape (70% Gaussian, 30% Lorentzian).

chemsum

{"title": "Thienoisoindigo-Based Semiconductor Nanowires Assembled with 2-Bromobenzaldehyde via Both Halogen and Chalcogen Bonding", "journal": "Scientific Reports - Nature"}

following_particle-particle_mixing_in_atmospheric_secondary_organic_aerosols_by_using_isotopically_l

## Abstract: Atmospheric fine particles contain thousands of organic compounds. Natural compounds from trees are often terpenes, consisting of multiple isoprene units, which when oxidized produce hundreds of poorly understood product compounds, many of which have extremely low vapor pressures and partition to particles. The interactions of these compounds control many particle properties, but it is difficult to distinguish them from each other. By synthesizing isotopically labeled terpenes, we were able to follow the interactions of individual particles with precision.C. Mixing SVOCs gas particle exchange through evaporation and condensation! ## INTRODUCTION Probing changes in the chemical composition of suspended nanoparticles upon chemical exchange that may arise from mixing different types of aerosols is a major challenge. This difficulty is caused by the need to somehow mark, ideally without the use of external labels, the starting aerosol populations such that their constituents, and exchanges among them, can be uniquely identified in individual particles. Although deuterium-labeling offers, in principle, the opportunity to probe chemical exchange among aerosol particles formed from organic compounds with a reasonably low level of invasiveness, doing so has been hampered by a lack of isotopologs other than those that are commercially available. Here, we have overcome this hurdle by using unlabeled and deuterium-labeled precursors to generate and characterize secondary organic aerosol (SOA), a class of aerosols made from the chemical oxidation and reaction of vapors. SOA is a class of atmospheric constituents that are among the least understood components in the climate system. 1,2 The particles, which have diameters of a few hundred nanometers or less, are suspended in air and act as individual nano-containers, among which we monitored exchange by using single-particle mass spectrometry with D:H readout. This approach is advantageous because SOA mass spectra from different terpenes are quite similar. However, by synthesizing isotopically labeled precursor molecules, we were able to generate SOA populations whose mass spectra are clearly distinguishable. ## The Bigger Picture The exchange of constituents between distinct types of aerosols is relevant to many processes important to atmospheric chemistry, combustion, bio-threat detection, and consumer-product formulations. However, because of the high similarity of aerosol mass spectrometer signals, it is difficult to distinguish between different aerosol populations and to track constituent exchange. We have overcome this hurdle by synthesizing deuterium-labeled terpenes as precursors for secondary organic aerosols and studying mixing driven by semivolatile vapor exchange with particles formed from other unlabeled terpenes as well as toluene. We found that particles from isoprene and a-pinene ozonolysis absorbed vapors rapidly. Particles from limonene ozonolysis showed slower exchange, and particles from b-caryophyllene ozonolysis showed limited exchange. Our results show that molecular exchange among particles from terpene oxidation becomes slower and less extensive as the precursor carbon number increases. As shown below, our work provides detailed compositional information needed to address the critical issue of how different particles and products from different precursors interact with one another. Interaction and potentially chemical exchange may be limited by volatility, 3,4 non-ideality, 5 or diffusion, depending on the particle viscosity, which in turn may vary with relative humidity (RH) and temperature. 9,10 We therefore probed how particle-particle interactions, and chemical exchange among particles via semi-volatile partitioning, are modulated by conditions of varying RH and temperature. We used isotopic labeling of SOA by synthesizing selectively deuterated a-pinene, the most common biogenically emitted terpene over the boreal forest ecosystem, 11 to explore the interactions of aerosols derived from the oxidation of various terpene precursors emitted in the natural environment. Yet, although biogenic volatile organic compound (VOC) emissions dominate global budgets, biogenic SOA formation is also strongly correlated with the presence of anthropogenic pollutants. 12,13 This strong correlation indicates that an as yet undetermined chemical or physical mechanism couples biogenic precursors and anthropogenic perturbations to atmospheric chemistry. Therefore, we also prepared SOA from commercially available deuterated toluene, a well-known marker for combustion because of its anthropogenic activity, to study interactions among and the evolution of SOA populations derived from anthropogenic and biogenic emissions. This model, although idealized, is to explore possible consequences of mixing aerosol-containing outflow from urban or industrial centers with pristine background air. 14 Although many studies rely on rheology to infer diffusivity within SOA particles, 15,16 we directly investigated mixing driven by semi-volatile vapor exchange between different SOA populations. The timescales in our mixing experiments were on the order of several hours, comparable with atmospheric timescales of diurnal cycling and meteorological changes. In the experiments, we formed separate populations from two precursors and characterized, by using H:D readout, the chemical composition of individual particles with highly sensitive single-particle mass spectrometers before and after intermixing those two populations. We observed the populations for hours in a large reaction chamber and varied RH and temperature to explore potential physical and chemical changes to the particles under conditions typical of those in the atmosphere. The key science questions that this work addresses are (1) whether SOA from different sources is readily miscible, (2) whether particles formed from the oxidation of various terpenes show significant limitations to diffusion and thus equilibration, and what fraction of the SOA formed from various terpenes is effectively semi-volatile by showing mobility between particles on a timescale of less than several hours. As mentioned above, the SOA precursors we used were deuterated a-pinene prepared in house as well as commercially available deuterated toluene, along with several non-labeled commercially available terpenes. The SOA systems in our study are very common model systems used to mimic SOA formation in chamber experiments. Parameterizations of those experiments inform organic-aerosol representations in chemical transport models. Furthermore, they are also the model systems that have received the most attention in recent research into glassy SOA. ## RESULTS AND DISCUSSION We provide full experimental details in the Experimental Procedures. In brief, we formed two aerosol particle populations in separate smog chambers and then brought them into contact to measure the composition of the two populations by using two single-particle aerosol mass spectrometers. We left one population (the chamber population) where it was formed, and we transferred the second population (the probe population) into that chamber, as shown in Figure 1. Because the chamber population was largely unperturbed, these experiments most directly targeted the behavior of the probe particles. We conducted a series of mixing experiments involving SOA formed from at least one terpene to explore volatility and diffusion limitations for a sequence of terpene-derived SOA varieties. To explore interactions among different terpenes, we synthesized isotopologs of a-pinene, as described below. In Table 1, we list the precursors, SOA mass concentration (in mg m 3 ), and oxygen-to-carbon ratio (O:C) of the chamber population and the probe population immediately before we brought the two populations into contact for all mixing experiments. To track evolution of composition of the particles, we used two single-particle mass spectrometers along with other particle and gas monitors. One of the mass spectrometers permits a new method, event-triggering aerosol mass spectrometry (ET-AMS), which enables particle detection at a much higher rate (up to 200 events per second) than our previous light-scattering single-particle method (LSSP-AMS); this greatly improved the time resolution of the mixing experiments. We started with an experiment in which we investigated mixing between SOA from limonene and SOA from deuterated toluene (99.5% purity; Cambridge Isotope Laboratories) as an example to demonstrate the process of a typical mixing experiment. Then, we systematically explored SOA from terpene precursors with increasing carbon number. Two SOA populations are generated separately as shown in (A) by either ozonolysis (terpenes) or photo-oxidation (toluene). The ''chamber'' population is formed by condensing SOA onto preexisting ammonium sulfate seed particles. The ''probe'' population is formed by nucleation followed by condensation. After both populations stabilize, the probe population is transferred into the main chamber (B). Any semi-volatile organic compounds (SVOCs) produced from SOA formation will partition between the gas and particle phases on the basis of their effective activity coefficients, whereas low-volatile organic compounds (LVOCs) will remain in the particle phase (C). Single-particle mass spectrometers and other instruments collect real-time data of particle and gas concentration and composition. In Table 2, we summarize the mixing experiments and their essential outcomes. Subsequently we discuss each experiment in detail. ## Mixing of SOA from Limonene and Toluene Oxidation We began with an experiment on SOA produced from ozonolysis of limonene (limonene SOA) as the probe population and SOA produced from photo-oxidation of D 8 -toluene (D 8 -toluene SOA) as the chamber population shown in Figure 2. This is a natural extension of previous experiments addressing interactions of a-pinene SOA and D 8 -toluene SOA 7,8 but with a terpene precursor that forms more highly oxidized SOA. The high time resolution of ET-AMS reveals the dynamics of vapor uptake over shorter timescales than experiments using LSSP-AMS (which we present in Figure S4). This experiment serves as a model for all experiments presented in this paper, and all other experiments follow a similar sequence of events. In Figure 2, we show time sequences of chamber RH and temperature (upper panel), time sequences of single-particle data (middle panel), and normalized ion signal intensity from both populations for different periods of the experiment (lower panel). For the middle panel, we plot the single-particle data on the y axis according to a simple cosine similarity score with respect to the mass spectrum for pure limonene SOA. We also set the symbol color according to the sum of two mass fragments, f(C 2 H 3 O + ) + f(CHO + ), which are major fragments of the limonene SOA and almost absent in D 8 -toluene SOA. The black and brown curves in the middle panel of Figure 2 show the organic mass fraction of limonene SOA in each population. We show the time series of the bulk mass measurement from AMS in Figure S3. At t = 0, we transferred some limonene probe SOA to the main chamber, which we had pre-filled with the D 8 -toluene SOA chamber population. Two hours after contact, we added water vapor to the chamber, increasing the RH to 30%. At t = 3.5 hr, we raised the chamber temperature from 22 C to 35 C. We know from prior experiments as well as calculations that particle coagulation is negligible on the fewhour timescale of these experiments for the number concentrations we used; thus, the only way the two particle populations can gain material from the opposite SOA types is via vapor condensation. The high time resolution of the ET-AMS data allows us to observe that the mass fraction of D 8 -toluene SOA in the limonene SOA probe particles jumped to 0.15 almost immediately after contact, but also that the mass fraction continued to rise by roughly 0.05 hr 1 for the next 3 hr. This gradual change in composition proceeded unabated when we humidified the chamber but settled to a steady-state value before we increased the chamber temperature. We plot the normalized ion signal intensity of each population from different periods of the experiment in the lower panel of Figure 2, showing signal associated with limonene-SOA in gold and signal associated with D 8 -toluene SOA in blue. The representative periods are (1) before mixing, (2) after mixing under dry conditions, and (3) after humidification to 30% RH. We neglected the ion signal at m/z 44 in the ET-AMS spectra (shown in gray) because of a large interference from gas-phase CO 2 . During period 1, the two populations were cleanly (and physically) separated, but during periods 2 and 3, the blue (chamber) material invaded the probe particles, causing a large increase in CHO + and C 2 H 3 O + in the probe population. The yellow (probe) material also invaded the chamber population, although to a much lesser extent. The condensation of D 8 -toluene SOA onto the limonene-SOA probe particles confirms that the chamber aerosol contained semi-volatile organics, consistent with our expectations for the D 8 -toluene SOA. 7,8 However, the particles themselves resisted uptake of semi-volatile compounds (even between particles formed from deuterated and non-deuterated toluene oxidation) until RH a 30%. 8 In this study, on the few-hour timescale of these experiments, ''semi-volatile'' in practice means constituents with saturation concentrations C* >1 mg m 3 . To the extent that the D 8 -toluene SOA forms an ideal solution with the limonene SOA, the 30% mass fraction of semi-volatile material that eventually transfers to the limonene SOA probe particles is consistent with the mass fraction of semi-volatile D 8 -toluene SOA being 0.3. The multiple time constants in the data suggest that several phenomena could contribute to the overall behavior; the slower uptake could be rate limited by diffusion into the limonene SOA probe particles. The temperature dependence of these mixing results is potentially complex. Assuming that these mixtures are close to ideal, heating will enhance mixing by encouraging vapor exchange, resulting in an equilibrium state consisting of a single homogeneous population; this happens when populations consisting of docosane isotopologs are heated above the docosane melting point. 7 However, in practice there are several competing effects. Increasing the chamber temperature by 15 C will increase the saturation concentration of any given species by roughly one order of magnitude, 4,17 driving down the gas-phase saturation ratio of ''semi-volatile'' species and causing them to evaporate from all particles while at the same time causing less volatile constituents in the particles to evaporate and exchange between the populations. Heat will also soften the particles and drive activity coefficients toward 1, which tends to anneal the aerosol toward a uniform composition. However, heating will also decrease the RH and thus the water content of the particles. For this first system, overall we observed that heating caused a slight decrease in the exchanged mass fractions of each population, suggesting that evaporation of semi-volatile species has the most influence. The RH and temperature for this experiment are plotted in the upper panel. In the middle panel, each particle is plotted according to their similarity score to the aggregated mass spectrum of limonene SOA. The symbol-connected lines show the mass fraction of the limonene SOA in both populations, also denoted as the extent of mixing, for every 3-min duration. Each particle is color coded by f(C 2 H 3 O + ) + f(CHO + ), which are two major peaks in the limonene SOA mass spectrum (lower panel). The brackets in the middle panel indicate the time periods corresponding to the aggregated mass spectra from both populations plotted in the lower panels. Before t = 0, the two populations had distinct mass spectra. After limonene SOA was transferred into the chamber at t = 0, it lost C 2 H 3 O + and CHO + as a result of evaporation. At the same time, it progressively took up vapors from D 8 -toluene SOA, indicated by the C 2 D 3 O + and CDO + mass spectra signals that are negligible before mixing. Two hours later, we raised the chamber RH to 30% from 10%, and limonene SOA became roughly 30% D 8 -toluene SOA. At the end of the experiment, the chamber temperature increased to 34 C, evaporating more volatile vapors from the condensed phase, and reduced the extent of mixing. The ion signal at m/z 44 was neglected (shown in gray) because of large interference from gas-phase CO 2 + . We generally focused on uptake of chamber SOA into the probe population. Quantitative assessment of probe aerosol vapor uptake into the chamber particles was complicated by the added effects of dilution as well as uncertain vapor losses during probe aerosol transfer, but this interaction is nonetheless informative. In this case, the D 8 -toluene SOA chamber population also showed some uptake of vapors from the limonene SOA but to a lower extent, with little sign of an RH or temperature effect. We recently showed that toluene SOA particles have diffusion limitations that will delay equilibration only for RH < 20%, 8 and so the relatively small uptake into these D 8 -toluene-SOA chamber particles at higher RH suggests that the limonene-SOA probe aerosol has a small semi-volatile content. Overall, our results indicate that constituents from limonene SOA and toluene SOA form a relatively ideal solution, but they also suggest that the mixing timescale of limonene SOA is on the order of 3-4 hr. There is relatively little evidence that increasing the RH from 10% to 30% changes the behavior of the limonene-SOA particles. However, the limonene SOA particles behave differently from a-pinene SOA particles, showing multiple timescales where the a-pinene SOA particles take up semi-volatile D 8 -toluene products in a rapid step. 7,8 Mixing of SOA from Isoprene Having shown that we can use these mixing experiments to explore semi-volatile vapor uptake by SOA from terpenes, we now systematically explore the mixing behaviors of SOA from C 5 to C 15 terpenes. We started with SOA from ozonolysis of isoprene (C 5 H 8 ), which contributes up to 30% of global particulate organic matter. 18 Isoprene is the lightest terpene and thus should produce the most volatile and diffusive SOA with relatively low viscosity. It is thought that the majority of SOA derived from isoprene consists of oligomers, for example, from the reactive uptake of monomers such as isoprene epoxydiol (IEPOX) and isoprene hydroperoxide (ISOPOOH). 19,20 So it is unclear whether this will cause any different behavior in the isoprene SOA. As shown in Figure 3, we transferred SOA probe particles formed via isoprene ozonolysis into a chamber containing SOA derived from D-toluene photo-oxidation. We held the chamber at a constant 22 C and 10% RH for the entire experiment. As the data show, the SOA probe particles from isoprene ozonolysis absorbed semivolatile vapors from the chamber SOA derived from D 8 -toluene almost instantly upon entering the chamber and reached a mass fraction of 25%-30%. For the remainder of the experiment, the semi-volatile species from D 8 -toluene oxidation that had diffused into the isoprene SOA probe particles gradually re-evaporated, presumably as a result of vapor loss to the chamber walls, slightly reducing their mass fraction in the isoprene SOA. We observed a similar behavior for carefully prepared two-component mixtures of oleic acid (semi-volatile with C = 6 mg m 3 ) and squalane (nearly non-volatile with 0.01 ( C ( 0.1 mg m 3 ). 21 Figure S5 shows histograms of f 30 (CDO + ) + f 46 (C 2 D 3 O + ) in the isoprene SOA probe particles for t < 0 (before contact) and for t = 0.5, 1.5, and 2.5 hr after contact. As with the bulk composition, f 30 (CDO + ) + f 46 (C 2 D 3 O + ), originally from D-toluene SOA, slightly decreased with time after contact in isoprene SOA, indicating again that the semi-volatile organics from the D-toluene SOA gradually reevaporate after the initial mixing. The diffusion of semi-volatile organics into the isoprene SOA particles appeared faster than diffusion into SOA particles from limonene, possibly because of the lower molecular weight of the constituents in the condensed phase. Previous studies have also shown evidence that even when the RH is low, mass-transfer limitations are unlikely in SOA produced from isoprene and OH radicals. 22,23 This indicates that SOA from isoprene is probably liquid-like with low viscosity. ## Mixing of SOA from Monoterpenes Our central objective was to explore cross-mixing among different types of terpene SOA. Without isotopic labeling, this would be difficult for different terpene precursors because of the similarity of the mass spectra and completely impossible for the crucial experiments involving mixing of SOA from chemically identical precursors, where entropic mixing is the only potential driving force. By synthesizing D 6 -a-pinene and D 3 -a-pinene as described in the Experimental Procedures, we generated two terpene SOA populations with distinguishable mass spectra. In this way, we could test whether these different SOA systems interacted with each other and explore the potential for diffusion and other limitations to mass transfer. In Figure 4A, we show LSSP-AMS data from a cross-mixing experiment between H-a-pinene SOA (H-pinene SOA, probe population) and D 6 -a-pinene SOA (D-pinene SOA, chamber population). Given that the D-H substitutions were the only chemical difference in the two precursors, we had every reason to expect the two populations to form an ideal solution. As the brown aggregated curve shows, the probe particles took up vapors from the chamber population relatively rapidly over the first hour. After that, the mixing remained more or less constant even when we raised the RH to 20% and then to 40%. The slightly higher chamber fraction at 40% RH could possibly have been caused by water-induced chemistry or waterinduced partitioning of semi-volatile organic compounds (SVOCs). 24 However, the data do not suggest that there were any substantial diffusion limitations within the particles but rather that approximately 25% of the a-pinene SOA was functionally For the remaining 2.75 hr, the SVOCs that were absorbed by isoprene gradually re-evaporated and were lost to the chamber wall, leading to a slightly increased fraction of isoprene SOA in the probe population. D 8 -toluene SOA remained essentially pure. semi-volatile under these conditions. The chamber particles likewise showed a corresponding uptake of probe SOA, indicating that they absorbed a combination of vapors from the probe chamber and vapors from probe-particle evaporation. The relatively rapid partial equilibration in both directions is consistent with previous studies where a-pinene SOA particles did not show significant diffusion limitations even at low RH. 7,8,22 At the end of the experiment, we increased the chamber temperature to 34 C, which forced some semi-volatile species to partition to the gas-phase, decreasing the fraction of materials originally from chamber population in the probe particles. In Figure 4B, we show a mixing experiment for a-pinene SOA and limonene SOA using LSSP-AMS measurements. Here, we produced D-pinene SOA by using a mixture of D 6 -a-pinene and D 3 -a-pinene with a volume ratio of 8:5. Similar to the experiment shown in Figure 2, the vapor uptake into the limonene SOA particles took place progressively, although it accelerated when we increased the RH to 30% at t = 3 hr. The mixing of a-pinene SOA into the limonene SOA particles was slower than the mixing of a-pinene SOA into a-pinene SOA particles. A consistent explanation is that the relatively more oxidized and thus more polar products from limonene ozonolysis are somewhat more viscous and therefore show some diffusion limitations. In addition, the uptake of limonene SOA into the deuterated a-pinene SOA chamber particles was significantly less than the corresponding uptake of unlabeled a-pinene SOA vapors, which is consistent with the higher oxidation state of limonene SOA constituents resulting in lower volatility than a-pinene SOA. A B Mixing of SOA from a Sesquiterpene Lastly, we used SOA derived from b-caryophyllene as the probe population and mixed it with SOA from a terpene (D 6 -a-pinene) and SOA from toluene. b-Caryophyllene, a sesquiterpene with a molecular formula of C 15 H 24 , also has two double bonds and a higher molecular weight than monoterpenes. It has very high SOA mass yields, although with a somewhat lower overall carbon oxidation state than limonene SOA. 25,26 The SOA from b-caryophyllene is also significantly less hygroscopic. The hygroscopicity parameter k for b-caryophyllene SOA is about 100 times smaller than k for a-pinene SOA at 1% supersaturation (k is a linear single parameter of aerosol hygroscopicity and is proportional to the inverse of water activity, a w ). 27 In Figure 5A, we show LSSP-AMS data from an experiment in which we exposed b-caryophyllene SOA probe particles to D-a-pinene chamber SOA at 7% RH, followed by RH steps to 20% and then to 75%. The temperature stayed at 22 C throughout the experiment. When the chamber was at 7% RH, b-caryophyllene SOA took up a very small amount (z5%) of D 8 -toluene SOA vapor. When we increased the RH to 20%, the vapor uptake increased slightly. We subsequently increased the RH to 75% and waited for 2 hr. The mass exchange remained no more than 10%. Similarly in Figure 5B, the extent of mixing between SOA from b-caryophyllene and SOA from D 8 -toluene was much less than the mixing between isoprene or monoterpene SOA and D 8 -toluene SOA populations. After 2 hr at 50% RH, the extent of mixing was still less than 10%. In addition, the chamber population showed a small but non-zero uptake of vapors from the b-caryophyllene SOA. We know from previous experiments that the chamber SOA in both cases shown in Figure 5 contains semi-volatile organics with activities between 0.2 and 0.3, so we would expect significant uptake into an ideal solution with no diffusion limitations. It is possible that the reduced uptake of semi-volatile constituents into the b-caryophyllene probe particles is because they are highly viscous. However, a significant diffusion limitation at 75% RH seems unlikely, although the hygroscopic growth for longifolene SOA (another sesquiterpene with a single exo double bond) is very low at 75% RH. 28 Another possible explanation is that constituents in b-caryophyllene SOA do not form an ideal solution with constituents in toluene SOA and a-pinene SOA. ## Conclusion By preparing SAO from isotopically labeled (deuterated) and unlabeled a-pinene, we explored the transfer of molecules among nanoparticles suspended in air. This would have been impossible without the isotopically labeled precursor because of the similarities among the aerosol mass spectrometer signals from terpene SOA samples. In addition, the efficiency of our single-particle method allows us to resolve composition changes in multiple populations with high time resolution after they are brought into contact during mixing experiments. Four different phenomena can drive or limit uptake to or exchange between particles, and the isotopologs are critical to deconvolving the effects. First, exchange can be limited simply because the constituents are barely volatile, but non-volatile vapors will condense under almost all circumstances. Second, immiscibility (non-ideality) can limit exchange even of semi-volatile constituents. Third, exchange and uptake can be driven in part by condensed-phase reactions, especially involving at least one relatively volatile gas-phase species. Fourth and finally, in-particle diffusion limitations (viscous particles) can inhibit both semi-volatile and reactive uptake by sustaining a strong activity gradient between the particle surface and its interior. Aerosols generated from isotopologs under identical conditions should form ideal mixtures with each other and reach reactive equilibrium as well. Thus, exchange is limited only by low volatility or high viscosity. Because water plasticizes viscous particles, 29 RH ramps resolve this final ambiguity. Oxidation of the precursors investigated here generated particles composed principally of low-volatility constituents (C % 1 mg m 3 ; p % 10 5 Pa). However, up to one-third of the particle composition consisted of more volatile constituents that readily diffused into the SOA particles under most conditions. Unlike SOA from toluene, which resists vapor exchange when dry, 7,8 there are no significant diffusive limitations in particles derived from isoprene and monoterpene oxidation, although SOA derived from limonene shows signs of some diffusion limitations when dry. Under the conditions used here, when semi-volatile uptake was uninhibited and nearly ideal, probe SOA particles took up roughly 20% by mass vapors from a-pinene oxidation and roughly 30% by mass vapors from toluene oxidation, consistent with earlier experiments. 7,8 Only particles formed from oxidation of b-caryophyllene showed less uptake, even at 75% RH. The high time resolution of our single-particle measurements shows that the particles derived from limonene and b-caryophyllene take up vapors from deuterated toluene slowly, i.e., over the course of 2-3 hr, until they reach near equilibrium. Yet, we also found that the limiting mass fraction is roughly 0.3 in limonene SOA and less than 0.1 in b-caryophyllene SOA. Our results suggest that under most circumstances in the planetary boundary layer, semivolatile vapors are sufficiently diffusive in SOA particles for equilibration to occur on relevant atmospheric timescales and that particles of biogenic and anthropogenic origin will mix with each other quickly. However, the fresh SOA we studied here had a lower O:C than aged ambient SOA. 30 Future experiments should explore particles with a wider O:C range, as well as the mixing of SOA formed from higher-molecular-weight intermediate VOCs, which are also important precursor molecules for atmospheric SOA. 31,32 EXPERIMENTAL PROCEDURES ## Synthesis of Deuterated a-Pinene The two isotopically labeled a-pinene compounds used in this study were synthesized with previously published methods. 33 Both ()-a-pinene-10,10,10-d 3 and ()-a-pinene-9,9,9,10,10,10-d 6 were accessed from the common intermediate, nopinone, which was synthesized through the ozonolysis of b-pinene (Sigma-Aldrich). ()-a-Pinene-10,10,10-d 3 and ()-a-pinene-9,9,9,10,10,10-d 6 can be accessed in two and eight synthetic steps, respectively, from the nopinone intermediate. Both routes enabled access to 100 mg quantities of each isotopically labeled a-pinene compound. Compound purity and percentage deuteration based on nuclear magnetic resonance (NMR) integration was determined to be 98% and R99%, respectively, for both compounds. That is more than sufficient to ensure that the resulting SOA particles are chemically defined by the desired compound and that the resulting particle mass spectra are easily separated. In the Supplemental Information, we provide 1 H NMR spectra for the isotopically labeled synthesized compounds and unlabeled a-pinene. ## Two-Chamber Setup We exposed particles derived from terpene oxidation to semi-volatile vapors containing oxidation products from other SOA sources (either terpenes or toluene) and investigated the vapor uptake over several hours at different RH. We formed SOA from terpenes by dark ozonolysis and from toluene by photo-oxidation under high NO x conditions typical of a polluted urban plume. In the ozonolysis chamber, the ozone concentration in the experiments ranged from 200 to 500 ppb. In the photo-oxidation chamber, the NO x concentration was about 500 ppb. We focused on the mixing behavior of SOA formed via oxidation of terpenes. Figure 1 shows the setup of the mixing experiments. 7,8 In each mixing experiment, we generated two SOA populations separately in two containers, a 10 m 3 chamber and a nearby 7 m 3 chamber (Figure 1A). The population in the 10 m 3 chamber (the ''chamber'' population) formed via condensation on pre-existing ammonium sulfate seeds. The seed particles were generated by atomizing 1 g/L ammonium sulfate solution, and the aerosol flow was subsequently dried in a diffusion dryer to well below the efflorescence RH. The dry seed particles in the chamber had a modal diameter of 90-100 nm. The population in the 7 m 3 chamber (the ''probe'' population) formed through nucleation. The mass signals from the non-volatile ammonium sulfate seeds in the particles assisted us in separating the two populations when analyzing the data. A single sampling line connected both chambers to an instrument suite via a three-way valve. We used one or two aerosol mass spectrometers (Aerodyne Research) to measure the chemical composition both of the ensemble and of single particles. We also measured particles and conditions in the chamber by using a scanning mobility particle sizer (SMPS, TSI 3081), an ozone monitor, a NO x monitor, and a temperature and RH sensor. After both aerosol populations stabilized, meaning that changes in mass concentration were dominated by particle wall loss instead of new mass formation, we transferred a portion of the probe population into the 10 m 3 chamber by using two Dekati dilutors with output flow rates of about 40 lpm (Figure 1B). The mixing timescale of the main chamber was roughly 5 min. We observed the changing composition of individual particles within each aerosol population by using the single-particle mass spectrometers (Figure 1C). This allowed us to probe SOA mixing driven by SVOC exchange. SOA particles in our experiments had mobility diameters between 200 nm and 700 nm, and we used atmospherically relevant mass concentrations as shown in Table 1. In all experiments, we formed one population by oxidizing an isotopically labeled precursor, generating particles with a unique mass spectrum. This included experiments with essentially identical SOA populations formed from the same precursor (i.e., a-pinene and isotopically labeled a-pinene). Analysis of these SOA populations is critical for establishing baseline aerosol properties under conditions where the SOA is completely miscible and where there is no reason to expect ongoing chemical reactions between reactive products, which may occur when chemically distinct populations are mixed. Our principal isotopically labeled precursors were synthesized D 6 -a-pinene and D 3 -a-pinene. 33 We also used commercially prepared D 8 -toluene (Cambridge Isotope Laboratories, D > 99.5%). We provide structures of deuterated a-pinene in Figure S1 and the NMR spectra of non-labeled a-pinene, D 3 -a-pinene and D 6 -a-pinene in Figure S2. Because of isotopic labeling, major fragments in the unlabeled SOA mass spectrum shifted from m/z 29 (CHO + ) and m/z 43 (C 2 H 3 O + ) to m/z 30 (CDO + ) and m/z 46 (C 2 D 3 O + ) in the labeled SOA mass spectrum. ## Single-Particle Mass Spectrometry We used two single-particle mass spectrometers in this study. Both are aerosol mass spectrometers built by Aerodyne Research. One is a light-scattering single-particle mass spectrometer, whose function has been described in detail elsewhere. 34,35 The other is an event-triggering aerosol mass spectrometer with a high-speed data acquisition card capable of triggering data acquisition on the basis of a signal pulse from an individual particle. In both instruments, particles enter a particle time-offlight region via an aerodynamic lens that focuses a beam of sub-micrometer particles onto a mechanical chopper with two 1% opening slits operating at 145 Hz. Under typical conditions, at most one particle is in the particle time-of-flight chamber during any given chopper cycle. At the end of the particle time-of-flight region, particles strike a 600 C tungsten vaporizer, and the resulting vapors are ionized by 70 eV electron ionization. Ions are then extracted and analyzed by a time-of-flight mass spectrometer (Tofwerk) with a mass resolving power (m/Dm) of 2,100 at 200 amu and high ion transmission efficiency. 36 In LSSP-AMS, a 405 nm laser intersects the particle beam at a right angle 16 cm downstream of the chopper, and the flight time between the chopper and the laser determines the particle aerodynamic size. Particles larger than 250 nm generate a sufficiently large scattering signal to trigger data acquisition for single-particle analysis. The final data product after processing by the Aerodyne software (Sparrow) contains the unit mass resolution mass spectrum of each particle. Mass spectra in all chopper cycles with a light pulse are downloaded and saved. Unwanted mass spectra are discarded in a post-processing step. These include chopper cycles containing no mass spectrum because the particle bounced off the vaporizer as well as chopper cycles containing mass spectra from more than one particle; truly coincident pulses are exceedingly rare at the particle concentrations we used. In ET-AMS, the mass spectrometer runs continuously, and firmware on a fast data acquisition card evaluates an ion signal threshold within user-defined regions of interests (ROIs). Each ROI corresponds to a mass-to-charge range and its ion signal thresholds. Users also specify logical ''AND'' or ''OR'' filters to combine the ROI into a single Boolean trigger. If an event passes the threshold setting, its spectrum is downloaded. In addition to the mass spectrum at the triggering time, two pretrigger spectra and five post-trigger spectra are also downloaded, ensuring that we capture the entire particle and help us establish baseline values. Particle aerodynamic size is calculated from the flight time between the chopper and the ionizer. The data are processed by Tofware developed by Tofwerk. For ET-AMS measurements, because of the fast data acquisition card and the ROI setting, ''filtering'' to select signals is conducted online. The advantages of ET-AMS are that (1) there are less data per particle because fewer spectra per particle are recorded and (2) no time is wasted downloading unwanted data. Data-acquisition efficiency is thus much higher in ET-AMS than in LSSP (as can be seen in the figures). Furthermore, because ET-AMS does not rely on scattered light to trigger data acquisition, it is not limited to particles larger than 250 nm. Particles smaller than 100 nm can be detected as long as they generate a sufficient ion signal. However, ET-AMS has an intrinsic chemical bias (or selectivity) in particle collection because it downloads only data satisfying the specified ROI thresholds, whereas LSSP obtains data on the basis of a physical light-scattering signal. In our study, we set three different ROIs (specified for each experiment described below) in order to minimize ET-AMS sampling bias. Quantitative Single-Particle Analysis We used the experiment presented in Figure 2 as a model for the rest of the experiments. In the middle panel of Figure 2, we plotted the single-particle data on the y axis according to the cosine similarity score with respect to the mass spectrum for pure limonene SOA. Electron ionization in the mass spectrometers produced only a few ions (at least six) per particle, resulting in single-particle spectra that are not statistically meaningful. However, our objective for this first stage of the data analysis was to separate the particles into two groups: the chamber and probe populations. This is similar to a clustering analysis except that it involves forced external mixture generating two clusters with known general properties. This task is simplified when one population also has a unique non-volatile seed, as is the case with ammonium sulfate in the D 8 -toluene SOA here. Once we separated the particles into two populations, we then aggregated the spectra by adding the individual-particle spectra within every 3-min interval and only afterward dividing by the total ion signal to obtain a normalized spectrum for each population. We quantitatively analyzed the aggregated mass spectra for their composition (specifically the relative content of the two kinds of pure SOA) via a simple linear regression to reference mass spectra from the pure populations. We assumed that the (normalized) aggregated mass spectrum of mixed particles at any given time, f p , can be described as a linear combination of the (normalized) mass spectra for the two pure particle populations, f H and f D (in each case this applies to the organic fraction of the mass spectrum and not any inorganic seeds), with contributions a H and (1 a H ): An identical equation applies for the chamber population, f c (p for probe, c for chamber). The black and brown curves in the middle panel of Figure 2 show the results of this composition analysis by plotting the organic mass fraction of limonene SOA in each population. ## SUPPLEMENTAL INFORMATION Supplemental Information includes five figures and can be found with this article online at https://doi.org/10.1016/j.chempr.2017.12.008.

chemsum

{"title": "Following Particle-Particle Mixing in Atmospheric Secondary Organic Aerosols by Using Isotopically Labeled Terpenes", "journal": "Chem Cell"}

overall_water_splitting_by_pt/g-c₃n₄photocatalysts_without_using_sacrificial_a

## Abstract: We report the direct splitting of pure water by light-excited graphitic carbon nitride (g-C 3 N 4 ) modified with Pt, PtO x , and CoO x as redox cocatalysts, while pure g-C 3 N 4 is virtually inactive for overall water splitting by photocatalysis. The novelty is in the selective creation of both H 2 and O 2 cocatalysts on surface active sites of g-C 3 N 4 via photodeposition triggering the splitting of water for the simultaneous evolution of H 2 and O 2 gases in a stoichiometric ratio of 2 : 1, irradiated with light, without using any sacrificial reagents. The photocatalyst was stable for 510 hours of reaction. Using photocatalysts to produce hydrogen sustainably by water splitting is the "holy grail" in modern science. Over the past 40 years, inorganic semiconductors, such as metal oxides and metal (oxy)nitrides, have been utilized as photocatalysts for hydrogen production. However, direct water splitting in a wireless powder photocatalytic system to produce gaseous hydrogen and oxygen has not yet been achieved using conjugated polymers (CPs). These materials have already shown great promise for use in organic electronics and photovoltaic devices, such as solar cells, light-emitting diodes, and feld-effect transistors, due to their good processability and tuneable electronic structures. The key challenge to using pristine CPs for direct water splitting is the insufficient hopping charge transport of the chains (usually below 10 4 cm 2 V 1 s 1 ) and a poor stability in water and under light irradiation. 12 Increasing the structural dimensions of the CPs (e.g., from 1D chains to 2D architectures) is desirable because the hole mobility is greatly increased (up to 0.1 cm 2 V 1 s 1 ) by the remarkably reduced binding energies of the Frenkel-type excitons and the robust stability of the 2D extended p-conjugated units. 14 However, further progress in direct water splitting by CPs will rely on breakthroughs in combining stable CP light transducers with suitable redox cocatalysts (usually noble metals) to promote charge separation and to reduce charge build-up on the polymer surface to prevent photocorrosion. Indeed, the promise of this type of system has been demonstrated by the successful development of 2D graphitic carbon nitride (g-C 3 N 4 ) polymer and metal-based redox cocatalyst systems for CO 2 reduction, organic synthesis and water half-splitting reactions using sacrifcial reagents. In contrast, it is difficult to achieve overall water splitting without using sacrifcial reagents because it depends not only on a rational chemical synthesis to tune the textural properties of the polymer but also on a rational design of the composite to control the reaction kinetics on the polymer surface. Photocatalytic water splitting by a prototypical g-C 3 N 4 polymer was shown to be thermodynamically possible because the C 2p and N 2p orbital bands straddle the water splitting redox potentials, but pure g-C 3 N 4 is typically limited by sluggish kinetics in photocatalyzing overall water splitting due to a lack of surface redox active sites. By optimizing the g-C 3 N 4 bulk and morphological properties and employing suitable redox cocatalysts (e.g., Pt for H 2 evolution and Co(OH) 2 for oxygen evolution), activities for the water half-splitting reactions (water reduction and oxidation) can be dramatically increased. Therefore, if the appropriate water redox cocatalysts are simultaneously deposited on g-C 3 N 4 , pure water splitting to produce gaseous hydrogen and oxygen could be achieved. However, the rough deposition of cocatalysts by traditional chemical reduction (e.g., H 2 and NaBH 4 ) cannot fully amplify the activity. Besides, the densely stacked graphitic layer also causes trouble for charge separation and migration due to a long bulk diffusion distance, resulting in a low photocatalytic quantum efficiency. 15 It is advisable to reduce the diffusion distance by rational synthesis of a g-C 3 N 4 nanosheet together with suitable cocatalyst modifcation to achieve water splitting. Up to now, direct water splitting photocatayzed by g-C 3 N 4 CPs in the absence of sacrifcial reagents has never been realized and still remains a signifcant basic science challenge. Here, we demonstrate that light-excited g-C 3 N 4 CPs can induce a one-step water splitting reaction via a four-electron pathway to generate gaseous H 2 and O 2 in a stoichiometric molar ratio of 2 : 1 when their morphology is modifed and the reaction kinetics are improved by modifcation with Pt, PtO x , and CoO x via photodeposition. The optimal g-C 3 N 4 -based nanocomposite had a turnover number of 3.1 moles of H 2 and O 2 per mole of g-C 3 N 4 photocatalyst for the overall water splitting reaction. The nanocomposite was stable in water and under light irradiation. The g-C 3 N 4 polymers used for photocatalytic water splitting were typically prepared by thermally polymerizing urea into heptazine units at 550 C which pack together like graphitic crystals. This structure was confrmed by X-ray diffraction (XRD), Fourier transform infrared (FT-IR) spectroscopy, and Raman spectroscopy (Fig. S1 †). 15, The g-C 3 N 4 optical properties measured by UV-vis diffuse reflection spectroscopy (DRS) were characteristic of a semiconductor; g-C 3 N 4 had an optical absorption edge at 442 nm due to the excitation of electrons from its valence band to its conduction band (Fig. 1a). The conduction band minimum (CBM) and valence band maximum (VBM) of the g-C 3 N 4 semiconductor were determined to be 1.31 V and 1.49 V (vs. NHE, pH ¼ 7), respectively, from electrochemical Mott-Schottky plots (Fig. 1b and c), where an estimated flat potential was directly used as the conduction band potential. Density functional theory (DFT) calculations revealed that the band gap was 2.56 eV with the CBM and VBM located at 1.0137 and 1.5505 V (vs. NHE, pH ¼ 7), respectively, which enables g-C 3 N 4 to act as a redox shuttle for the water splitting reaction (Fig. S2 †). This calculated band gap is consistent with the experimental data and further demonstrates that in theory, g-C 3 N 4 could be used to split water. First, the effects of g-C 3 N 4 morphology on the photocatalytic activity were investigated. We prepared three types of g-C 3 N 4 using dicyandiamide (DCDA), ammonium thiocyanate (ATC) and urea as precursors. The results showed that after in situ photo-deposition with Pt, the urea-derived g-C 3 N 4 exhibited signifcant photocatalytic activity for the overall water splitting reaction, while the other samples were inactive for overall water splitting (Table S1 †). It should be noted that all pure g-C 3 N 4 polymers showed no activity for overall water splitting in the absence of cocatalysts, implying that surface kinetic control using Pt species was indispensable to achieve overall water splitting by g-C 3 N 4 based photocatalysts. N 2 sorption measurements revealed that the DCDA-and ATC-derived g-C 3 N 4 samples had smaller surface areas than the urea-derived samples (ca. 10 m 2 g 1 vs. 61 m 2 g 1 ). However, mpg-C 3 N 4 with a surface area of ca. 67 m 2 g 1 also exhibited no water splitting activity. This indicated that surface area was not the major factor in controlling the water splitting activity and the splitting of water on densely stacked g-C 3 N 4 polymers was indeed very difficult to achieve. To better understand the real mechanism of water splitting on the soft surface of the CPs, we characterized the morphology of the above different polymers. TEM images of DCDA-and ATC-derived g-C 3 N 4 and mpg-C 3 N 4 samples revealed densely stacked polymer layers, which were very different from the silk-like thin nanosheets of the urea-derived one (Fig. S3 †). The fast evolution of O in the form of CO 2 or CO could accelerate the deamination rate. Thus, the texture, morphology and electronic properties of the CNU samples were optimized, and contributed to creating the active Pt/g-C 3 N 4 photocatalysts for overall water splitting. Evidently, accelerated charge separation and migration on the nanosheets can be obtained in comparison with densely stacked graphitic layers, which is elucidated well by the corresponding large decrease of PL emission intensity (Fig. S4 †). The nanosheet structure can also be certi-fed by an AFM experiment. As shown in Fig. 2a, the thickness of the nanosheet is determined as $2 nm. One can now easily conclude that the ultrathin 2D geometry of urea-derived g-C 3 N 4 is crucial for achieving overall water splitting as demonstrated by the fact that g-C 3 N 4 samples prepared from urea at different temperatures all have remarkable water splitting activities (Fig. S5 †) due to their similar thin nanosheet structures (Fig. S6 †). The CNU samples prepared at 550 C showed optimum activities. This is because when the temperature is lower than 550 C, the heptazine cycle doesn't completely form, while partial decomposition occurs when the temperature is higher than 550 C. Both of these two aspects may generate inactive CNU samples. DCDA-and ATC-derived g-C 3 N 4 and mpg-C 3 N 4 samples revealed densely stacked polymer layers, and the deposition rate of Pt nanoparticles on the surface of the polymer was very slow, in the absence of organic sacrifcial agents to react with the holes. Optimization of the deposition technique of Pt is needed to enhance the overall water splitting activities of this bulky g-C 3 N 4 . We then investigated the effect of cocatalyst loading techniques on the photocatalytic water splitting activity. Three different cocatalyst loading techniques, in situ photodeposition, and H 2 and NaBH 4 reduction, were developed to decorate the g-C 3 N 4 nanosheets. As shown in Fig. S7, † evident water splitting activity was generated for photodeposition of Pt on the surface of the g-C 3 N 4 nanosheets, while only very slow H 2 and no O 2 evolution were found for both H 2 and NaBH 4 reduction modi-fed ones. In the frst case, when g-C 3 N 4 was irradiated with light, photoexcited charge carriers were generated and then immediately migrated to the surface of the g-C 3 N 4 nanosheets without recombination. The surface adsorbed Pt 4+ was then reduced in situ by the excited electrons and deposited on the active sites, which can efficiently promote the water splitting. For the other investigated techniques, Pt 4+ was reduced by H 2 or NaBH 4 and then randomly deposited on the surface, resulting in poor activities. The selective photodeposition of Pt on thin g-C 3 N 4 nanosheets resulted in a uniform dispersion of ultrafne Pt nanoparticles ($1-2 nm) with a (111) crystal lattice spacing of $0.23 nm (Fig. 2b and c). The homogeneous deposition of Pt can be further proved by STEM imaging (Fig. 2d). However, serious particle accumulation occurred when the Pt cocatalysts were deposited by H 2 and NaBH 4 reduction (Fig. S8 †), which was the major hindrance which led to decreased water splitting activity. We also investigated the chemical composition and valence state of the Pt species. As shown in Fig. 3a and b, electron energy loss spectroscopy (EELS) and XRD analysis confrmed the existence of a Pt (111) plane. 38 Besides, no evident structure variation occurred after modifcation with the Pt cocatalysts, implying a robust stability of the g-C 3 N 4 CPs. Three pairs of XPS peaks corresponding to Pt 0 , Pt 2+ , and Pt 4+ with binding energy at 72.13, 74.26 and 78.17 eV, respectively, were measured (Fig. 3c). Pt 0 was effective for H 2 evolution while PtO x were able to promote O 2 evolution. 42 However, two pairs of XPS peaks were deconvoluted for a NaBH 4 reduction modifed one (Fig. S9 †), indicating the complete reduction of Pt 4+ into Pt 2+ and Pt 0 . To confrm that PtO x were active for the promotion of a water oxidation reaction, we evaluated the photocatalytic water oxidation activities of the as-prepared PtO x /g-C 3 N 4 . As shown in Fig. S10, † this material showed enhanced activity for water oxidation in comparison with the pure one, emphasizing the positive role of PtO x in improving the water oxidation rate. In addition, the water splitting rates and evolved H 2 /O 2 gas ratio (Fig. S11 †) could be fnely tuned by simply adjusting the total loading from 0.2 to 5 wt% due to the change of the ratio of Pt and PtO x intensities (Fig. S12 and Table S2 †) and the alteration of particle size (Fig. S13 †). The creation of metal/polymer surface junctions promotes the interfacial redox reaction which can be confrmed by a rapidly decreased PL intensity (Fig. 3d). The optimum activity was achieved at a Pt loading content of 3 wt%. When Pt or PtO x were singly deposited on the g-C 3 N 4 nanosheets, the sample exhibited very poor activity in both cases, which once again highlighted that the simultaneous creation of both H 2 and O 2 evolution cocatalysts on the active sites was indeed essential for triggering the overall splitting of water. The g-C 3 N 4 nanosheets modifed by other noble metals (e.g., Rh, Ru, or Au) via in situ photodeposition all just showed trace H 2 and no O 2 evolution (Fig. S14 †), implying the importance of Pt for water splitting. The pH value and amount of polymer powders used for water splitting were also optimized (Fig. S15 and S16 †). The optimum water splitting rate was obtained for samples prepared by photodepositing 3 wt% Pt on 0.2 g of g-C 3 N 4 nanosheets under neutral conditions. We then evaluated their stability for long term reaction. As shown in Fig. S17, † the optimized Pt/g-C 3 N 4 showed good water splitting stabilities under both UV and visible light irradiation for 580 hours of continuous reaction. It should be noted that N 2 gas was evolved along with H 2 and O 2 at the initial stage of the reaction. This arises from the self-oxidation of the surface un-condensed amino groups (-NH) by excited holes. As the reaction proceeded, after 80 hours almost no N 2 evolution was observed, suggesting a complete consumption of the -NH groups. When the Xe lamp was turned off, the amounts of the evolved gases quickly diminished in just four hours (Fig. S18 †), indicating a fast occurrence of the backward reaction of water splitting on the Pt species (H 2 and O 2 recombination for water formation). Thus, to further enhance the overall water splitting activity of the system, an efficient restraint of the backward reaction via rational structural design of the cocatalysts (e.g., core/shell nanostructure) should be considered. The addition of cobalt species for in situ formation of cobaltbased cocatalysts can also sufficiently promote the water oxidation selectivity and efficiency of metal-free semiconductors such as g-C 3 N 4 and h-BCN. As expected, the simultaneous evolution of H 2 and O 2 gases in a stoichiometric ratio of 2 : 1 by Pt-Co/g-C 3 N 4 under UV (l > 300 nm) (12.2 and 6.3 mmol h 1 ) (Fig. 4a) and visible light irradiation (l > 420 nm) (1.2 and 0.6 mmol h 1 ) (Fig. 4b) was signifcantly enhanced after 1 wt% CoO x were further modifed for use as O 2 evolution cocatalysts, which can be determined by XPS analysis (Fig. S19 †). The slightly decreased activity in each run of reaction may be attributed to the stacked samples on the inner side of the reactor (Fig. S20 †). Furthermore, no obvious deactivation was observed after 510 hours of reaction (Fig. S21 †), demonstrating the robust resistance of the composites to water and light corrosion at the soft interface. The total amount of gaseous H 2 and O 2 collected reached $6.2 mmol, which corresponded to turnover numbers (TON) of 3.1 and 111.3 based on g-C 3 N 4 and Pt, respectively. The apparent quantum yield (AQY) for the overall water splitting reaction was calculated to be 0.3% at 405 nm (Fig. S22 †) and was monitored by an on-line gas chromatograph (Fig. S23 †). This is lower than the value of 2.5% of (Ga 1x Zn x ) (N 1x O x ) inorganic photocatalysts. However, it is a remarkable frst observation that photocatalytic overall water splitting can occur on the surface of an organic/polymer semiconductor via a 4-electron pathway. Optimization of the system to further improve the efficiency is ongoing in our lab. ## Conclusions The discovery of Pt/g-C 3 N 4 CPs that can split pure water without the use of sacrifcial reagents establishes a new chemical paradigm for exploiting clean, renewable solar energy using organic semiconductor light-energy transducers. Ongoing efforts are focused on modifying the electronic and textural structures of g-C 3 N 4 CPs and coupling them to low-cost kinetic promoters to facilitate photocatalytic cascade processes for water splitting and CO 2 fxation that are relevant to sustainable energy production via artifcial photosynthesis.

chemsum

{"title": "Overall water splitting by Pt/g-C₃N₄photocatalysts without using sacrificial agents", "journal": "Royal Society of Chemistry (RSC)"}

ultrafast_charge_separation_dynamics_in_opaque,_operational_dye-sensitized_solar_cells_revealed_by_f

## Abstract: Efficient dye-sensitized solar cells are based on highly diffusive mesoscopic layers that render these devices opaque and unsuitable for ultrafast transient absorption spectroscopy measurements in transmission mode. We developed a novel sub-200 femtosecond time-resolved diffuse reflectance spectroscopy scheme combined with potentiostatic control to study various solar cells in fully operational condition. We studied performance optimized devices based on liquid redox electrolytes and opaque TiO 2 films, as well as other morphologies, such as TiO 2 fibers and nanotubes. Charge injection from the Z907 dye in all TiO 2 morphologies was observed to take place in the sub-200 fs time scale. The kinetics of electron-hole back recombination has features in the picosecond to nanosecond time scale. This observation is significantly different from what was reported in the literature where the electron-hole back recombination for transparent films of small particles is generally accepted to occur on a longer time scale of microseconds. The kinetics of the ultrafast electron injection remained unchanged for voltages between +500 mV and -690 mV, where the injection yield eventually drops steeply. The primary charge separation in Y123 organic dye based devices was clearly slower occurring in two picoseconds and no kinetic component on the shorter femtosecond time scale was recorded.Dye-sensitized solar cells (DSCs) are promising candidates for solar energy conversion applications. These devices do not rely on rare or expensive materials, so they could be more cost-effective than cells based on silicon and thin-film technologies. Recently, DSCs device efficiency has reached a maximum power conversion efficiency of over 12% using donor-bridge -acceptor (D-π -A) zinc porphyrin dye in combination with a cobalt-based redox mediator 1 .The performance of DSCs is based on kinetics competition between the electron injection from the sensitizer to an electron collecting material, usually TiO 2 , regeneration of the oxidized dye with redox electrolyte and unwanted back reactions of injected electrons recombining with oxidized dye molecules or oxidized species of redox electrolyte 2 . A challenge in this field is that the kinetics of charge carriers may be altered in complete devices showing top performances. A deep understanding of many parameters controlling the overall performance is crucial for achieving improvements in performance. Despite numerous studies, there is still a debate on the electron injection time scale for the optimized solar cells and according to the proposed "kinetics redundancy", the optimized solar cells might not have ultrafast electron injection kinetics 3 . Existing studies are mainly based on the classical pump-probe transient absorption spectroscopy, which is widely used to measure the kinetics of electron injection processes. Since optical transparency is required, to perform transient absorption studies, only model systems based on a transparent TiO 2 thin film sensitized with various dyes and semiconductors in different environments (solid samples or in solution) were investigated so far . It should, however, be noted that the most efficient liquid-based solar cell devices are not transparent. Indeed, these devices are based on a double layer of TiO 2 film, which contains a scattering layer made of 400 nm TiO 2 particles deposited on top of a mesoporous transparent layer 8 . The resulting light transmittance of the cell is less than 15% in the visible region and, hence, conventional transient absorption spectroscopy in transmission mode cannot be applied in this case. Despite the importance of the subject, the kinetics of electron injection in actual optimized, opaque dye-sensitized solar cell devices under working conditions has not so far been reported. We aim here to investigate the dynamics of charge carriers directly in fully functional devices, using potential control and state-of-the-art pump-probe diffuse reflectance spectroscopy. Despite the great potential of the latter technique, only a few studies can be found in literature investigating its implementation and application. We aim to demonstrate that diffuse reflectance spectroscopy is of great value for time-resolved analysis of photophysical processes in opaque or highly absorbing materials. Time-resolved diffuse reflectance spectroscopy was first reported by Wilkinson et al. 9 in microsecond time regime in 1981 followed by Bowman et al. 10 and Asahi et al. 11,12 . The technique was utilized on scattering systems like powders of organic microcrystals, and by Furube et al. 13 on DSCs under open circuit condition. Here we have developed an ultrafast time-resolved pump-probe diffuse reflectance spectrometer with a sub-200 femtoseconds time-resolution. This required application of novel optical design for the collection of diffuse reflected light. In addition and for the first time, we combined the femtosecond time-resolved diffuse reflectance laser spectroscopy with potential control and photovoltage measurements. Furthermore, the technique has enabled us to investigate the charge separation kinetics in DSCs based on photoanodes of other TiO 2 film morphologies. For example, we studied samples of anodized nanotubes on Ti foil 14,15 and nanostructured fibers 16,17 . These samples have exhibited promising behavior in cell performance but are not optically transparent and are not suitable for investigation with pump-probe transmission based transient absorption technique. Our studies reveal that the charge separation dynamics in Ru-based dye in the complete device is ultrafast and is indeed affected by the morphology of the TiO 2 film. We observed an early charge recombination in scattering TiO 2 particles, TiO 2 fibers and anodized TiO 2 nanotubes. These recombinations had different amplitudes and were not previously reported for small particles and are rationalized in terms of different electron mobility and trapping states in different TiO 2 films. Under an applied voltage bias condition from + 500 mV up to -690 mV, the kinetics of the electron injection from the dye excited-state into the oxide remains ultrafast. However, the injection yield decreases at the bias point of -690 mV. In contrast to Ru-complex based dye, the organic D-π-A dye Y123 exhibited slower charge injection kinetics. The excited-state lifetime of Y123 dye is measured to be 50 ps. The time constant of the electron injection process is measured being 1.1 ps. While this classifies as ultrafast, it is about one order of magnitude slower than for Ru-based dyes, which was measured to have features in femtosecond time scale. ## Results and Discussions Figure 1a shows the schematics of the standard optimized high-performance liquid solar cell. In the conventional DSC scheme, the mesoporous layer is made of 20 nm-diameter interconnected TiO 2 particles. Although this structure offers a large surface area for dye adsorption, Rayleigh scattering with this size of TiO 2 particles is small, resulting in high transparency of the dye-sensitized film in a broad spectral region. A significant amount of light (70% in the near infrared region) is transmitted without interacting with dye molecules in the cell. The working electrode applied in highly efficient devices is based on a TiO 2 double layer film 18 , sensitized with dye molecules on top of a TiCl 4 -treated conductive glass. The structure of these samples is shown in Fig. 1a. The first layer is a transparent mesoporous anatase TiO 2 film, consisting of interconnected spherical nanoparticles (20 nm). Another layer made of 400 nm-diameter TiO 2 particles is deposited on top of the transparent layer. Figure 1b shows the total transmittance, total reflectance and total absorptance of the Z907 dye-sensitized TiO 2 double layer film based DSC photoanode. The 400 nm particles act as light scattering centers enhancing light absorption by increasing the light pathway within the film. Consequently, the total transmittance of the cell in the visible and near-infrared region is less than 15% as it can be seen in Fig. 1b. This suggests that the diffuse reflectance spectroscopy is the only versatile optical laser spectroscopy technique capable of studying such devices. The Kubelka-Munk function, F(R) spectra is derived from diffuse reflectance of the film according to equation (3), presented in the method section. The F(R) spectrum is compared with the absorptance spectrum of the opaque photoanode in Fig. 1b. As it is seen, the Kubelka-Munk spectrum follows the shape of the absorptance curve, and the similarity in both spectra is observed. The peak around 520 nm corresponds to the Z907 dye ground state absorption that serves as an absorbing medium. The shoulder at 380 nm corresponds to the absorption of TiO 2 substrate that serves as the scattering media in Kubelka-Munk theory. Two types of liquid electrolyte-based devices were selected for the present study. The first type of DSC is prepared with a Ru-complex sensitizer (Z907) in combination with an iodide/ triiodide based redox electrolyte in 3-methoxypropionitrile solvent. This combination was reported to result in highly stable devices when subjected to light and thermal stress during long-term aging 19,20 . The second type of cell is based on the organic D-π -A, sensitizer Y123 and a cobalt complex-based redox electrolyte. This type of device yielded a power conversion efficiency of over 9% 21,22 and over 12% in combination with a porphyrin dye 1 . The thickness of both TiO 2 layers affects both photocurrent and photovoltage, which were optimized in earlier studies in terms of final power conversion efficiency 8 . ## Femtosecond diffuse reflectance spectroscopy on operational DSC device based on Z907 sensitizer. In order to unravel the electron injection dynamics in dye-sensitized opaque solar cells, we utilized pump-probe diffuse reflectance spectroscopy. We applied this technique to the study of Z907 dye-sensitized TiO 2 films, which are from the same family of N719 and N3 Ru-based dyes. According to earlier transient absorption studies on N719 (cis-bis (isothiocyanate)bis(4,4′ -dicarboxylic-2,2′ -bipyridyl) ruthenium(II)) and N3 (cis-di(thiocyanate)bis(2,2′ -bipyridyl-4,4′ -dicarboxylic acid)ruthenium(II)) dye-sensitized TiO 2 films, the transient absorption spectrum around 800 nm is assigned to the oxidized dye molecules 13, and the absorption spectrum around 1200 nm is attributed to absorption by conduction band electrons 13,26,27 . The kinetics of electron injection in the complete device is studied by monitoring the evolution of oxidized dye molecules and photo-injected electrons in TiO 2 . The oxidized dye molecules are monitored at the characteristic absorption onset in the visible wavelength region at 670 nm or in the NIR region at 840 nm, and photo-injected electrons are monitored at 1200 nm. Figure 2a,b compare the early and later time evolution of absorptance of oxidized dye molecules anchored on three different TiO 2 films in the presence of MPN solvent. The samples are excited at 600 nm. The time delayed diffuse reflected probe beam is measured at 840 nm. Transient absorptance change is extracted from the measured transient diffuse reflectance change given by equation (2), depicted in the method section. In Fig. 2a,b, it is seen that in the presence of MPN solvent, the kinetics of electron injection in double layer film resembles that of transparent film made of small TiO 2 particles. All samples have an instrument response-limited transient absorptance onset within 200 fs and a slow rise of the signal with a time constant of 1.1 ps. The first ultrafast response limited rise of the signal corresponds to the ultrafast electron injection from dye to TiO 2 conduction band. Ultrafast fluorescence studies performed by Chergui and coworkers 28 on N719 sensitized TiO 2 small particles films revealed that the electron injection occurs with a time constant of 10 fs for non-thermalized levels of the dye and 120 fs from the thermalized level. The second slow rise component up to 5 ps is observed in all three TiO 2 films was also previously reported in measurements of transparent films by different research groups 29,30 . Studies by Wenger et al. 29 assigned this feature to the presence of dye aggregates in the films, which have a larger distance for electron injection and, therefore, less electronic coupling for electron transfer process. Sundström et al. proposed another description in terms of a two-state mechanism. They assigned the fast and slow components to the injection from the singlet and the triplet excited-states of the ruthenium complex, respectively 30 . Figure 2c illustrates the long time kinetics of the evolution of oxidized dye molecules up to 500 ps after excitation in the working cell based on TiO 2 double layer film. Remarkably, the kinetics features a decay component in several hundred picoseconds after excitation. The trace is fitted by an exponential decay function with a time constant of 4 ns. It should be noted that all measurements shown in Fig. 2 are performed at very low excitation intensities. For measurements shown in Fig. 2, the excitation energy of each pulse at the sample is 200 nJ. The repetition rate is 0.5 kHz. The beam diameter is close to 500 μm. Under these conditions, the excitation irradiance is 102 μJ cm −2 . With having the repetition rate of 0.5 kHz, the excitation power is 51 mW cm −2 , which is equivalent to 50% of the sun irradiance power at AM 1.5 condition. In addition, the number of photons normalized to the volume of TiO 2 within the irradiation area is about 2.15 × 10 11 . This is significantly smaller than the number of adsorbed dye 3) depicted in the method section. The total transmittance of the cell in the visible and infrared region is less than 15%. The Kubelka-Munk function spectrum follows the shape of the absorptance spectrum. molecules on the microscopic surface of TiO 2 film, within the irradiated area which is 9.8 × 10 16 (assuming one monolayer of adsorbed dye molecules). Hence, our observations are not due to some non-linear effects. The excitation intensity dependence of the observed kinetics is also presented in Supplementary information, Figure S4. All traces measured at low excitation intensities can be fitted by a single exponential function. This early decay kinetics is again observed when the evolution of oxidized dye molecules is monitored in the visible wavelength region at 670 nm (Supplementary Figure S5). Moreover, the same depleting kinetics can be observed (Fig. 3b) when photo-injected electrons are monitored at 1200 nm in the presence of the electrolyte. Therefore, we assign the early decay kinetics observed for the complete opaque DSC in the presence of MPN solvent, to the early back recombination of photo-injected electrons with oxidized dye molecules. The kinetics of back recombination is even more strongly accelerated in the presence of the electrolyte. Traces in Fig. 3a compare the measurements in the presence (black markers) and absence of redox electrolyte (red markers) on double layer TiO 2 film. In the vicinity of the redox electrolyte, the formation of the signal is again ultrafast, and the sub-200 fs ultrafast component is still present. As it can be observed, 26% of the signal of the oxidized dye molecule decays with a time constant of τ 1 = 9 ps. The same fast decay kinetics are also present in the samples dipped in the redox-inactive ionic liquid, 3-methyl-1-ethylimidazolium bis (trifluoromethane) sulfonimide (EMITFSI), (see Fig. 4d). Therefore, the observed kinetics cannot be assigned to processes like reductive quenching of the excited-state of the dye molecules by redox electrolyte. The accelerated charge recombination in the presence of both redox active and redox inactive electrolyte in some picoseconds after excitation is rationalized by electric fields induced by the charges in the electrolyte and charge screening effects. The local electric field induced by ions present in the redox electrolyte or ionic liquid at the surface is accelerating the charge recombination between electron-hole geminate pair after initial charge separation. Figure 3b provides more evidence, as photo-injected electrons in TiO 2 are directly monitored at 1200 nm. The trace is fitted with two exponential decay functions. The rate constant of the fast decay component of photoelectrons measured at 1200 nm is 1.1 × 10 11 s −1 and for the slower decay kinetic is 5.917 × 10 8 s −1 , giving a lifetime of 8.9 ps and 1.7 ns, respectively. Interestingly, these time constants are consistent with the kinetics fit values of measurements at 840 nm monitoring oxidized dye molecules. Taken together, the observed decay kinetics at 840 nm and the mirror kinetics at 1200 nm are due to an early back recombination of photo-injected electrons with the oxidized dye molecules anchored on the surface of TiO 2 particles. Therefore, a significant observation in our measurements of the DSC devices based on scattering particles and all other TiO 2 morphologies (as it is discussed below), is that the kinetics of electron back recombination with oxidized dye molecule has features in the picosecond to nanosecond time scale. This observation is remarkably different to what is normally stated in the literature for model systems of transparent films made of small particles. The small particles based films are the only morphology that has been studied to date. In earlier studies of small TiO 2 particles sensitized with Ru dyes based on transient absorption spectroscopy at visible or NIR region or by 2D-IR spectroscopy 23, , such electron back recombination was not observed and the oxidized dye molecule was accepted to be stable until much longer time scale of microseconds. We observe that the kinetics is accelerated in full device relative to what is so far accepted for small particles. In a complementary study, presented in Supplementary Table S1, the photovoltaic, optical and structural characteristics of the transparent and double layer based devices are depicted. The value of photocurrent normalized to the light absorptance of the scattering layer is about 30% less than that for transparent layer based device. This indicates that in big particles some fraction of photoelectrons are lost and is consistent with the early back recombination of electrons and oxidized dye molecules observed in laser spectroscopy measurements. In addition, the pump-probe diffuse reflectance technique has also enabled us to investigate the electron injection profiles in many other DSC devices. These studies include measurements of different opaque nanostructured TiO 2 films such as TiO 2 nanofibers 16 and TiO 2 nanotubes. TiO 2 nanotubes are prepared by anodization of Ti foil as the substrate 14 . Due to opacity, these samples could never be studied by transmission based transient absorption technique. Figure 4 shows the dynamics of electron injection on Z907 dye-sensitized TiO 2 films of different morphologies monitored at 840 nm. Panel a shows the kinetics of charge separation in Z907 dye-sensitized standard double layer based DSC device in the presence and absence of redox electrolyte. Panel b depicts the pump-probe diffuse reflectance measurements on Z907 dye-sensitized anodized TiO 2 nanotube film on Ti foil. The same studies are performed on dye-sensitized TiO 2 nanostructured fibers. Measurements of fibers in the presence of cobalt-based electrolyte (CO II /CO III ) and also redox-inactive electrolyte (EMITFSI), are shown in panel c and d, respectively. It is interesting that in all measurements the electron injection is still in the ultrafast regime and recombination features with different amplitudes comparable to that of double layer film is present. In standard full DSC based on double layer film 30% of the signal decays in 20 ps after excitation, this compares to only 16% for DSCs based on anodized nanotubes. The observed kinetics in the presence of redox-inactive ionic liquid suggests that the decay kinetics cannot be due to the reductive quenching of the dye excited-state by redox electrolyte. The observed difference in early back recombination of photoelectrons in the TiO 2 films of nanoparticles and anodized nanotubes is rationalized in terms of morphological parameters of the two films such as trap state distribution. We hypothesize that photo-injected electrons may get trapped in the TiO 2 particle surface states where they form geminate pairs with holes (oxidized dye molecules) and result in the observed early fast recombination in different TiO 2 morphologies with different amplitudes. It should be noted that, our control studies (Figure S3) shows that at low excitation intensities of 0.25 sun irradiance at AM1.5, the excited dyes inject more than 3.8 × 10 16 cm −3 electrons into the TiO 2 , which is much smaller than the trap state density (10 18 cm −3 to 10 20 cm −3 ) The observed kinetics in the vicinity of electrolyte measured at 1200 nm monitoring the kinetics of photo-injected electrons in TiO 2 . The solid lines correspond the fit to the result by exponential function. The time constant of exponential fit to the measurements at 840 nm is τ 1 = 9.22 ps and for trace measured at 1200 nm is τ 1 = 8.9 ps. The pump wavelength is 530 nm and pulse intensity is 200 nJ. reported for TiO 2 films 14 . One should also consider the energetic distribution of these traps. For instance, the trap state distribution in the anodized nanotubes is measured using macroscopic techniques like charge extraction experiments by Hagfeldt et al. 14 . The nanotubular electrodes have a trap state distribution significantly different from nanoparticulate electrodes. Nanotubes possess relatively deeper traps with a characteristic energy of the exponential distribution more than twice than that of nanoparticulate electrodes. Throughout time-resolved terahertz studies, Schmuttenmaer et al. 34 have claimed that the low mobility in polycrystalline TiO 2 nanotubes is not only due to scattering from grain boundaries or disorders as is in other nanomaterials but instead results from a single sharp resonance from exciton-like trap states. These observations are in good agreement with our spectroscopy studies. Indeed, electrons can get more localized in energetically deeper traps in nanotubes films and, therefore, less early back recombination with oxidized dye molecule is observed in these films in comparison with nanoparticles based films. Our observations show direct evidence of an ultrafast electron injection occurring in a complete Ru-dye based DSC device. This has not been confirmed so far for DSCs having all components like scattering layer, electrolyte, conductive glass, etc. Charge injection from the amphiphilic Ru II (bipyridyl) Z907 dye in all different TiO 2 morphologies was observed to take place in the sub-200 fs time scale. We observed that the kinetics of charge separation is indeed influenced by the morphological parameters of the TiO 2 substrate. The photo-injected electron in TiO 2 fibers, nanotubes, and 400 nm particles shows prompt back recombination kinetics with oxidized dye molecules in the picoseconds-nanoseconds time scale after excitation. Electronic parameters like density of trap states and energetic of trapped electrons, i.e. how deep electrons are trapped, and consequently the mobility of electrons, might play a vital role in the behavior of photo-injected electrons after initial interfacial charge separation. We have combined the diffuse reflectance spectroscopy with potentiometric techniques to monitor the electron injection process in DSC standard device under working condition. Figure 5a shows the photovoltaic characteristics of the Z907 sensitizer and iodide based redox electrolyte device. The photovoltaic parameters of the device at full sunlight are; short circuit current density (Jsc) of 15.5 mA/cm 2 , open circuit photovoltage (Voc) of 698 mV, fill factor (FF) of 0.71 and power conversion efficiency (PCE) of 7.6%. Figure 5b presents the typical transient absorptance of the cell in short circuit condition and under different bias voltages of + 500 mV, − 500 mV (close to max power point) and − 690 mV. Transient absorptance traces are not easily distinguishable. All traces in this figure can be fitted by exponential function with close fitting parameters; a component with a lifetime of 50-70 ps, and the flat behavior until hundred picoseconds, which is shown in inset. By raising the bias to − 690 mV close to open circuit condition, the amplitude of the observed signal is almost half of the others while the kinetics remain similar. In other words, as the amplitude of the pump-probe signal is proportional to the number of oxidized dye molecules, with increasing the applied bias voltage the quantum yield of electron injection is reduced. The energy level diagram, which contains the energy level of the conduction band of TiO 2 (CB), HOMO and LUMO level of Z907 dye and iodide-based redox electrolyte (Eredox) and trap state distribution are illustrated in Fig. 5c. Increasing the forward bias voltage would raise the quasi-Fermi level position by filling up the trap states to some extent in TiO 2 , which is highlighted in Fig. 5c. As it is observed by shifting the quasi-Fermi level toward the LUMO level of the dye, the energy difference gets noticeably smaller. In example the energy difference at the bias level of − 600 mV is 75% of the energy difference at the bias level of − 500 mV and further reduces to 50% at a higher bias level of − 700 mV. It should be noted that at the bias level of − 690 mV, the amount of voltage in the films is less than this value due to the dark current. The voltage drop in the cell at a voltage bias of − 700 mV is estimated as 90 mV. By taking account of this voltage drop, the respective amount of the cell photocurrent measured at bias voltages up to − 500 mV and at − 690 mV is consistent with the respective amplitude of the pump-probe signals. In our control studies (Supplementary Figures S6 and S7), the photovoltage induced by each laser pulse is estimated to be only of μV order when the cell is biased in the conventional voltages of hundreds of mV. This indicates that our pump-probe measurements under potentiostatic condition can be considered as a perturbation technique. Electron injection in DSC based on D-π-A organic sensitizer and cobalt electrolyte. One of the most significant advances in design of light-harvesting materials is the so-called donor-conjugated linker-acceptor (D-π -A) organic dyes. In comparison with Ru-based dyes, organic dyes have higher molar extinction coefficient and can be readily designed for a desired absorption spectrum 1, . These molecular structures look attractive in terms of electron donor-acceptor interactions 35 . In order to understand the electron injection dynamics in these systems, femtosecond diffuse reflectance spectroscopy is applied to DSC devices containing Y123 and a cobalt complex based redox electrolyte. Three different morphologies of the photoanode are used, and the results are compared in Fig. 6. In these measurements, the pump beam wavelength is 600 nm to excite the dye, and the probe beam is 840 nm. Upon laser excitation, an ultrafast formation of the signal happens in 200 fs, followed by a fast decay of the signal to 50% of its amplitude in 2 ps. After two picoseconds, the signal reaches a plateau in all the 3 morphologies of the TiO 2 layers. Unlike Ru-based dyes, no obvious difference is observed in the kinetics for measurements in the presence of MPN solvent and redox electrolyte and the medium has no influence on the observed kinetics (measurements in the presence of MPN solvent are provided in Supplementary information Figure S9). The slow growth component of the signal in hundred picoseconds time scale seen in the films made of scattering particles or the double layer film, is assigned to the electron injection from dye aggregates. This component is removed from the signal when the sample is immersed in acetonitrile solvent for several hours (Supplementary Figure S10). TiO 2 films prepared with scattering particles might have enough space to accommodate aggregated dye molecules within the pores, which are loosely in contact with the TiO 2 layer. This makes a larger distance for electron transfer between the dye and TiO 2 . In order to study the mechanism of the very fast decay of the signal in 2 ps, we performed the transient broadband absorption measurements. We compared the broadband transient absorption of Y123-sensitized TiO 2 film with that of Y123 dye in solution as a reference sample where interfacial electron transfer process is deactivated. By comparison of the recorded spectra of the dye-sensitized films with the dye in solution we can clearly resolve the spectral absorption contribution of the excited-state and oxidized state of the Y123 dye. The transient absorption spectra of the samples measured at NIR region is also shown in Supplementary Figure S11. The ground state optical absorption spectrum of the dye is provided in Supplementary Figure S8. For the dye in solution (Fig. 7a), the negative peak at the characteristic ground state absorption of the dye around 520 nm, is attributed to the ground state bleaching of the dye formed upon photo-excitation. In Fig. 7a, a positive transient absorption is observed in the wavelength region from 630 nm up to 700 nm. This transition is assigned to excited-state absorption of dye molecule. The dye excited-state relaxation time constant is 52 ps and has a mirror-like kinetics to ground state relaxation. The transient absorbance spectrum of the dye-sensitized TiO 2 films is presented in Fig. 7b. In this sample, the ground state bleaching is observed at 580 nm, which is around 60 nm shifted with respect to the steady-state absorption onset of the dye. This red-shift in the transient absorption spectrum is an evidence of a Stark-shift of the absorption spectrum of the dye molecule. The Stark-shift is explained by the shift of ground state absorption of the dye molecule due to the local electric field induced by the electric dipole of the neighbor dye molecules. This effect was also previously reported for the same family of D-π -A dyes 36,37 . For the dye-sensitized TiO 2 film the positive absorption feature is extended over the 630 nm wavelength regions. This positive feature is now assigned to a contribution of both of the excited-state absorption of the dye and the absorption by oxidized dye molecule formed upon injection of electrons into TiO 2 conduction band. Figure 8 compares the kinetics of a Y123 sensitized TiO 2 film probed in two different wavelength regions of 690 nm and 740 nm. According to Fig. 7a, at 690 nm the excited-state of Y123 has absorption while at 740 nm the excited-state does not absorb. Therefore, the blue trace in Fig. 8 recorded at 740 nm, represents a pure monitoring of the kinetics of the electron injection process. The formation of the signal, which reflects the electron injection time, is occurring in picosecond time scale and is fitted with an exponential growth function with a time constant of 1.1 ps. As a result, we observe that in the Y123 D-π -A dye, the charge injection kinetics is not as fast as in Ru-based dyes, as reported by Chergui and co-workers 28 . The difference in the electron injection time in the Y123 with Ru-based dyes should be due to the D-π -A structure of this dye and its coupling with the TiO 2 film. In the Y123 dye, the electron donor part is the triphenylamine unit, and the acceptor orbitals are located on the cyanoacrylate group. Cyclopentadithienophene (CPDT) is working as a π -bridge between the donor and acceptor parts for the conjugation of electrons. This bridge helps increasing the dipole moment and enhancement in the molar extinction coefficient of the dye molecule 38 . Our results suggest that the relaxation of the excited-state is fast with a time constant of 52 ps. This process competes with electron injection into the lower lying conduction band of TiO 2 . Moreover, non-adiabatic charge transfer from a molecular electronic excited state into a continuum of acceptor levels constituted by the conduction band of a semiconductor can be described by Fermi's golden rule 2 . Due to the very high density of acceptor levels, the nuclear factor in the equation tends to a constant value. As a consequence, the thermodynamics of the process, the temperature, and the reorganization energy are not expected to affect the injection dynamics. The electronic coupling between the donor and the acceptor (electronic coupling matrix element squared |H] 2 ) is then likely to control in a large extend the electron transfer rate. As |H| 2 depends exponentially upon the charge transfer distance, the adsorption geometry and the electronic coupling between the dye's HOMO and the empty d 4 orbital manifold of Ti IV sites on which the dye is anchored must be determining the electron injection time. In summary, our experimental technique allowed us to reveal the charge separation dynamics in a complete opaque solar cell device under applied bias voltage. Also, the interfacial charge separation in different dye-sensitized opaque TiO 2 nanostructured interfaces was determined. We showed that this technique could be a powerful and sensitive tool for measurements of opaque and highly absorbing materials. Coupling diffuse reflectance spectroscopy with potentiometric characterization tools gives a unique possibility to study charge carriers in devices under real operational conditions. In Ru-complex based solar cells, the kinetics of electron injection is confirmed to be ultrafast and is not affected by the bias voltage. In standard opaque and other TiO 2 morphologies based devices, after ultrafast electron injection, an early recombination of photo-injected electrons with oxidized dye molecules is observed with features in the picosecond to nanosecond time scale. The respective amplitude of this recombination process is influenced by morphological parameter i.e. trap states in TiO 2 films. We found that the ultrafast electron injection kinetics is not influenced by trap state filling upon increasing forward bias up to − 500 mV. By applying a higher voltage close to open circuit conditions, and shifting the Fermi level of TiO 2 closer to the dye excited-state level, the electron injection is less efficient but the injection kinetics is still ultrafast. For a DSC with a Y123 organic dye, the dynamic of the excited-state of the dye and the kinetics of electron injection process significantly differ from Ru-base dyes. The excited-state relaxation in the Y123 dye molecule competes with electron injection into TiO 2 . In contrast to Ru-complex based dyes, which show ultrafast electron injection in femtoseconds time scale, the electron injection for the Y123 dye is precisely monitored to occur within 2 ps after excitation of the dye. Finally, the technique proposed here will be an excellent tool to be implemented for studies on highly absorbing materials, such as the new emerging perovskite based devices and can open up a new avenue of characterization research. ## Methods Pump-probe femtosecond diffuse reflectance spectrometer. In principle, the configuration of the diffuse reflectance spectroscopy is similar to the traditional transient absorption in transmission mode. The differences are: firstly in the probe beam geometry to collect the diffuse reflected light, which carries the information of transient species, secondly the sample structure and thirdly the optical model in data treatment. Here, a new optical scheme for collection of diffuse reflected light is designed which gives a unique time-resolution of sub-200 fs (Supplementary Figure S1). In this configuration, diffuse scattered light from the sample is collected, collimated, and focused onto the detector with two coupled off-axis 90° parabolic mirrors. Having this configuration, the large solid angle of light collection results in improved signal to noise ratio. The other advantage of using parabolic mirrors over lenses is that no further dispersion is introduced to the pulses; therefore, a better time-resolution is expected. Indeed, in this configuration, the time-resolution is limited by the time broadening of the beam in the diffusive sample. This time broadening is typically measured as about 30 fs in our samples 39 . For the transient absorption measurements, the pump beam at a defined wavelength is produced using a twostage non-collinearly phase-matched optical parametric amplifier (NOPA). It is modulated using a synchronized chopper at a frequency of 0.5 kHz, which is half the repetition frequency of the laser. It is focused onto the sample at an angle of about 60° from normal. The pump beam has a diameter of 500 μm at the surface of the sample and typical energy of about 100-200 nJ/pulse. The probe beam is provided by a second NOPA having less energy than the pump on the sample to avoid multiple excitations and is focused having the spot size of around half of that of the pump. The polarization between the pump and probe beams is at the magic angle (54.7°). The transient response of the sample is measured by collecting the diffuse reflected pulses of the probe. The light scattered by the sample is focused onto the detector (photodiode: Nirvana detector, New Focus, model 2007). The signal of the detector is amplified by use of a lock-in amplifier. Lock-in parameters are set as integration time 1s, dwell time 4s, time constant 1s for measurements. A power supply (Weir) is used to apply a fixed bias voltage on the solar cell for diffuse reflectance measurements on the cell under voltage bias condition. The bias voltage between the two electrodes is changed from − 690 mV to + 500 mV and time-resolved diffuse reflectance of the device is measured at each applied bias. In these measurements, the pump and probe beam are irradiated to the cell from the backside (photoanode side), similar to the photovoltaic measurements. All the rest experimental details of experiments are similar to that previously explained. Time-resolution of diffuse reflectance setup and linearity tests. The time-resolution of the setup is defined by using optical Kerr gating technique. In this technique pump and probe beams are focused and spatially overlapped on a non-linear media (SF10 crystal or glass substrate). The cross-correlation of pump and probe beam on a Kerr-media is measured at an angle of 45° between the polarization of pump and probe. We performed the cross-correlation experiment in the diffuse reflectance mode. As the Kerr-media once the SF10 crystal and another time the scattering sample with cover glass is used. The reason is to compare the broadening of the cross-correlation peak when the diffuse reflectance is measured on these types of substrates. The time-resolution of the setup is sub-200 fs. It should be noted that the time-resolution of this technique is limited by the time broadening of the beam in the sample due to the scattering effect. This time broadening is measured to be 120 fs for our samples. This is in contrast with the transmission based transient absorption technique in which the time-resolution is solely determined by the pulse duration of pump and probe and their cross-correlation. The setup has an unprecedented sensitivity as it enables measurements of transparent, non-reflective samples (see green curve in Fig. 6) with a reasonable signal to noise ratio. In order to check the linearity of measurements, the intensity of the pump beam is changed over a broad range of energy from 0.047 μJ/pulse to a high intensity of 0.950 μJ/pulse. The diffuse reflectance of dye-sensitized sample at a fixed time delay (i.e. 50 ps) is measured. Both absorptance and Kubelka-Munk formalism show a very good linear fit to the measurements over the whole intensities of excitations (Supplementary Figures S2 and S3). Data acquisition and treatment. In the configuration of reflectance measurements, the transient absorptance ( )  change can be measured and corrected by determining the absolute amount of diffuse reflected light with and without pump beam. This is practically achieved by chopping the pump pulse at half repetition frequency of the laser. The time-resolved diffuse reflectance of samples is measured by varying the delay time between the pump and probe pulses. Therefore, transient absorptance is displayed as: In case of opaque samples that the optical transmittance of sample is negligible, absorptance change (  ∆ ) can be deduced only from reflectance change, as: Where T is the intensity of the transmitted light and R and R 0 represent the intensity of the diffuse reflectance of probe pulse with and without excitation, respectively. The linearity of absorptance change upon excitation intensity is tested in our control studies, over a wide range of excitation intensities. Another theory describing the optical behavior for a tightly packed isotropic absorbing and scattering medium is Kubelka-Munk 40,41 , in which the Kubelka-Munk function relates the measurable so-called diffuse reflectance of the sample to the ratio of the absorption coefficient (K) and scattering coefficient (S). In the case of diluted medium, K is linearly dependent on concentration of absorbing species (c), in the same way, the Lambert-Beer Law is also valid in solutions; equation (3): ln (10) (3) For quantitative simulation, use of the Kubelka-Munk function is essential, however, treating the transient reflectance of the samples with both equations of (2) and (3), does not change the kinetics. Broadband transient absorption setup. The pump− probe technique uses a compact CPA-2001, 1 kHz, Ti: Sapphire-amplified femtosecond laser (Clark-MXR), with a pulse width of about 120 fs at a wavelength of 775 nm. The pumping beam is generated using an NOPA tuned to 600 nm to generate pulses of approximately 8 μJ, that are then compressed in an SF10-glass prism pair down to duration of less than 60 fs (at FWHM). At the sample, the excitation pulse energy is decreased to a few hundred nJ. The probe beam is a white light continuum generated in a sapphire plate and splits before the sample into signals and reference beams in order to account for intensity fluctuations. Both beams were recorded shot by shot with a pair of 163 mm spectrographs (Andor Technology, SR163) equipped with 512 × 58 pixels back-thinned CCD cameras (Hamamatsu S07030-0906). The polarization of pump and probe pulses was set at a magic angle. ## Solar cell fabrication. Dye-sensitized solar cells were fabricated using a double-layered photoanode made of mesoporous TiO 2 film. A transparent, 9 μm-thick layer of 20 nm particles was screen-printed onto an FTO glass plate (NSG-10, Nippon Sheet Glass). Subsequently, a 5 μm-thick layer of scattering particles (400 nm diameter) was deposited by screen-printing. The surface area of TiO 2 film was 1 cm 2 . The TiO 2 film was sintered up to 500 °C by a stepwise heating program. Prior and after TiO 2 deposition a TiCl 4 treatment was performed on the samples. The BET surface area of the mesoporous transparent film and scattering film were 85 m 2 g −1 and 27 m 2 g −1 . The values of the two films porosity were 70% and 65% for transparent film and scattering film respectively. Prior to dye loading, photoanodes were sintered again at 480 °C for 30 minutes. Afterward, substrate was cooled down to 80 °C and immersed in the dye solutions for overnight. After rinsing with the acetonitrile, the stained substrates were sealed with pieces of thermally platinized electrode. The platinized electrode was made using a solution of H 2 PtCl 6 on FTO glass (TEC15, Pilkington), and served as a counter electrode. The working and counter electrodes were separated by 25 μm-thick hot melt ring (Surlyn, DuPont) and sealed by heating. The electrolytes were introduced to the cells via pre-drilled holes in the counter electrodes. ## Photovoltaic characterization. The setup used for standard photovoltaic characterization (J-V curve) consisted of a 450 W Xenon lamp (Oriel), whose spectral output was matched in the region of 350-750 nm with the aid of a Schott K113 Tempax sunlight filter (Präzisions Glas & Optik GmbH), and a source meter (Keithley 2400) to apply potential bias and measure the photocurrent. A set of metal mesh filters was used to adjust the light intensity to a desired level. A black metal mask defined the cell active area to be 0.158 cm 2 .

chemsum

{"title": "Ultrafast charge separation dynamics in opaque, operational dye-sensitized solar cells revealed by femtosecond diffuse reflectance spectroscopy", "journal": "Scientific Reports - Nature"}

on_the_use_of_dft+u_to_describe_the_electronic_structure_of_tio2_nanoparticles:_(tio2)35_as_a_case_s

## Abstract: One of the main drawbacks in the density functional theory (DFT) formalism is the underestimation of the energy gaps in semiconducting materials. The combination of DFT with an explicit treatment of electronic correlation with a Hubbard-like model, known as DFT+U method, has been extensively applied to open up the energy gap in materials. Here, we introduce a systematic study where the selection of U parameter is analyzed considering two different basis sets: plane-waves (PWs) and numerical atomic orbitals (NAOs), together with different implementations for including U, to investigate the structural and electronic properties of a well-defined bipyramidal (TiO2)35 nanoparticle (NP). This study reveals, as expected, that a certain U value can reproduce the experimental value for the energy gap. However, there is a high dependence on the choice of basis set and, and on the +U parameter employed. The present study shows that the linear combination of the NAO basis functions, as implemented in FHI-aims, requires a lower U value than the simplified rotationally invariant approaches as 2 implemented in VASP. Therefore, the transferability of U values between codes is unfeasible and not recommended, demanding initial benchmark studies for the property of interest as a reference to determine the appropriate value of U. I. INTRODUCTIONTitanium dioxide, TiO2, nanoparticles involving a mixture of anatase and rutile polymorphs, in particular in the commercialized Degussa P25 form, constitute the most studied photocatalytic material and a model system for the mechanisms involved in photocatalysis. [1][2][3][4] The performance of TiO2 largely depends on its optical, electronic, structural, morphological and surface properties, 5-7 and one of the key properties of TiO2, especially in the anatase polymorph, is the formation of photogenerated charge carriers (holes and electrons), activated by the absorption of ultraviolet (UV) light. Indeed, the need for UV radiation constitutes one of the major bottlenecks towards developing efficient TiO2 photocatalysts that can work under sunlight as only ~5% of the incident solar spectrum corresponds to UV light. Hence, a major challenge in the development of competitive TiO2-based photocatalysts is reducing the energy gap to the visible (VIS) region. 8 In principle, the properties of TiO2 can be modulated by designing nanoparticles (NPs) with different sizes, shapes, crystallinities, and surface facets. 9-12 . However, to determine the relationship between structural and electronic properties of TiO2 nanoparticles, experimentally, is not a simple task. Alternatively, computational techniques provide a feasible, accurate, and unbiased approach to study such correlations and, consequently, can contribute to build connections between experiment and theory. 13 Density functional theory (DFT) 14,15 has been widely used to study the properties of different types of materials with high accuracy in the prediction of crystal structures and reasonable description of electronic structure features at a moderate computational cost 16 and with a well-established reproducibility. 17 Unfortunately, energy gaps computed using the popular local density approximation (LDA) and the generalized gradient approximation (GGA) are consistently underestimated by 30-100% 18,19 . The error arises from the inherent lack of derivative discontinuity and the delocalization error. To overcome the drawbacks of LDA and GGA for estimating this electronic property, hybrid functionals, which include a part of the nonlocal Fock exchange, have been proposed and widely employed. 18,23,24 Depending on the type of basis set, the use of hybrid functionals can represent a significant increase in the cost of the calculations. Inspired by the Hubbard Hamiltonian, 25 Anisimov et al. 26 proposed to avoid the computational load inherent to hybrid functionals by implementing an empirical onsite Hubbard (U) correction to a selected atomic energy level, within standard DFT. The resulting method is often referred to as DFT+U, an unfortunate term as DFT is an exact theory. DFT+U has been broadly used, especially after the contribution of Dudarev et al. 27 and is particularly useful in the description of the partially filled d-states of the transition metals -in the case of TiO2, the U-correction is applied to the Ti 3d orbital. 28,29 The DFT+U method combines the high efficiency of standard DFT with an explicit, albeit approximate and empirical treatment of electron on-site correlation, and constitutes one of the simplest approaches to describe the ground state of strongly correlated systems. 30 However, the choice of the appropriate U parameter value for each compound constitutes a challenge. This obstacle can be solved through (i) a linear response, fully consistent method, 31 or (ii) alternative routes based on comparison with experimental results for some physical property of interest such as magnetic moment, energy gap, redox potentials or reaction enthalpies. For instance, the latter strategy has been employed in the study of electron transport in rutile phase, 35,36 of reduced forms of TiO2, 37,38 and ultrathin films of the rutile phase. 39 Nevertheless, the selection of the U parameter is not straightforward. Moreover, the choice of the appropriate form of the projector functions inherent to the method is also a concern, 40 especially after the work of Kick et al. 41 who recently implemented DFT+U with a numerical atomic orbital basis set. The authors showed that the value for U depends on the choice of projector function, which in turn depends on the type of basis set (atomic orbitals or plane waves) used. The aim of the current study is to evaluate the effect of the basis sets in the selection of the U value necessary to describe the electronic structure of semiconducting nanoparticles, taking a previously investigated, well-defined (TiO2)35 bipyramidal NP as a case study. 42 ## II. MODELS AND COMPUTATIONAL METHODS The well-defined bipyramidal stoichiometric (TiO2)35 anatase NP, which fulfills the requirement of a Wulff construction, 43 and was used in previous studies, 42 is selected for the present study (Fig. 1). This nanoparticle exposes the most favorable (101) facets only, as found in experiments. 7 Furthermore, its ~2 nm size is also appropriate to rationalize experimental results reported for TiO2 anatase NPs. 44 The calculations reported here have been carried out using two widely used codes, namely the Vienna Ab Initio Simulation Package (VASP) 45,46 and the Fritz Haber Institute ab initio molecular simulations (FHI-aims). 47 In both cases, the Perdew-Wang (PW91) exchangecorrelation functional 48 is used and spin-polarization is accounted for explicitly, although the final results do not exhibit any spin-polarization. The partially filled Ti3d states were consistently described by applying the Hubbard U correction 26 under the simplified rotationally invariant approach introduced by Dudarev et al. 27 In the following, we will refer to the resulting approach as PW91+U, which is more appropriate. The calculations carried out with VASP employ a plane waves (PWs) basis set with a kinetic energy cut-off of 396 eV. To account for the effect of inner electrons on the valence density we implement the projector augmented wave (PAW) method of Bloch 49 as implemented by Kresse and Joubert, 50 with 12 and 6 valence electrons for Ti and O atoms, respectively. The (TiO2)35 NP is included in a 20  20  40 supercell to give a vacuum gap of 11 in the x-and y-directions and 20 in the z-direction. Γ-point sampling is used and the convergence criteria for the energy and forces are 10 -4 eV and 0.02 eV/ -2 , respectively. On the other hand, the calculations carried out by the FHI-aims code include all electrons (AEs) and account for relativistic effects through the so-called zero-order regular approximation (ZORA) 51,52 proposed earlier by Chang et al. 53 A tier-1 light grid numerical atom-centered orbital (NAO) basis set has been used with a quality comparable to that of a TZVP Gaussian Type Orbital basis set for TiO2. 42 Here, for the implementation of Hubbard U correction, the projection functions for Ti3d states are introduced as an explicit linear combination of the NAO basis functions with the double-counting correction in the fully localized limit (FLL); see details in Ref. 41. The convergence threshold for the energy and is 10 -4 eV. Note that, hereinafter, the notation of PW and NAO is used to refer to the calculations performed with VASP and FHI-aims, respectively. ## III. RESULTS AND DISCUSSION To provide a sound reference for the study, we first discuss the energy gap of fully relaxed anatase and rutile bulk phases as predicted from spin polarized DFT calculations with the PW91 GGA type density functional and using either PW or NAO basis sets. To avoid problems arising from a difference in the quality of the basis sets we increase the kinetic energy cutoff for the PW to 550 eV and used a more extended NAO basis set of tier-2 tight quality. For rutile, the PW/NAO calculated band gap is 1.70/1.91 eV whereas for anatase the PW and NAO calculated band gaps coincide and amount to 2.10 eV. The difference in the rutile phase must be attributed to small differences in the optimized structure arising from the different treatment of the core electrons. In any case, the PW and NAO calculations for bulk rutile and anatase lead essentially to the same results with a deviation of at most 0.2 eV in the band gap. Clearly, these calculated energy gaps are underestimated with respect to the experimental values, which are 3.0 and 3.2 eV, for rutile and anatase phases, respectively. Hybrid functionals with an ad-hoc amount of non-local Fock exchange are known to provide a better estimate, as discussed for instance by Ko et al. 57 , while DFT+U can be tuned to recover the experimental band gap, but usually at the cost of a poorer description of other materials properties. Next we focus on the representative (TiO2)35 anatase NP depicted in Fig. 1. The atomic structure of this NP has been obtained from a geometry optimization using both VASP and FHI-aims computational packages and PW91+U. However, to perform a rigorous comparison of the effect of U when using PW or NAO basis sets we consider four different situations which are as follows: The structure is optimized in FHI-aims with PW91 (U=0) and single-point calculations are run with both FHI-aims and VASP at each U value, U=0-10 eV; (ii) The structure is optimized in VASP with PW91 (U=0) and single-point calculations are run with both FHI-aims and VASP at each U value, U=0-10 eV; (iii) The structure is fully optimized in both FHI-aims and VASP at each U=0-10 eV. ## (iv) Each structure obtained by FHI-aims (VASP) in (iii) is submitted to a single point calculation in VASP (FHI-aims) at the same U-value. The first and second sets of calculations allow one to investigate differences in the description of the electronic structure that are not due to a difference in the atomic structure but to the different type of basis set and the implementation of the +U term. 41 The third set of calculations provides information about differences in the final optimized structure, and the effect of this optimization on the energy gap. Finally, the fourth set of calculations shows to what extent the fully relaxed atomic structure impacts on the electronic structure. In each of these data sets we can compare the results of the different set-ups by a linear fit of the data. ## A. Structure Analysis We start the discussion by analyzing the structural properties of the (TiO2)35 NP focusing mainly on its length and width (Fig. 2). The PW91 (U=0) fully optimized structures of the (TiO2)35 NP predicted by VASP and FHI-aims are almost the same. In both cases, the nanoparticle length, which is taken from the terminal atoms located in the apical region (see Fig. 1), is 19.61 . For the width of the NP, FHI-aims predicts a width that is 0.02 larger than VASP. Hence, in the absence of U, both types of basis set lead to the same structure, as expected. 17 Therefore, any difference in the PW91+U structure predicted by the two types of basis sets (codes) has to be attributed to differences in the implementation of U. Regarding the atomic structure, the main effect of U is to slightly increase nanoparticle length (Fig. 2a). The tendency is consistent, regardless of the basis set, up to U = 5 eV. When U is larger than 5 eV the lengths predicted by VASP and FHI-aims follow different trends. The analysis of the nanoparticle width presents some interesting features (Figs. 2b and 2c). Here the effect of U is different depending on whether the calculation is carried out with a PW or NAO basis set. When using NAO, the optimized NP width drops almost linearly with U up to U = 7 eV, whereas when using PW, the dependence with U is very small, almost negligible. We note that, when using PW, the trends are very stable along the interval of U. However, this is not the case when NAO basis set are employed, and the regular trend is broken at U = 7 eV. Note also that the breaking of the trend at U > 7 eV for the calculations with NAOs indicates that this value is too large to correctly describe correlation effects as it has an exceedingly large influence on the properties of the nanoparticle and induces unreasonable structural changes. Similar observations on the effect of U on the phase stability of TiO2 have been reported. 33 It is assumed that the large effect of U on the atomic structure predicted by the calculations using the NAO basis set arise from the more localized character of the atomic NAO Hubbard projectors as implemented in FHI-aims. 41 ## B. Energy Gap Analysis The analysis of the energy gap of the (TiO2) 35 anatase NP provides further interesting comparisons. The Kohn-Sham energy gaps, computed in the set-ups described in scenarios (i) and (ii), above, are shown in Fig. 3 and Table I. This data corresponds to two structures, each optimized with the respective codes, FHI-aims and VASP, at the PW91 (U=0) level. We begin by comparing the results of the single-point PW calculations performed on the FHI-aims (green) and VASP (blue) relaxed structures. At each U-value, the difference in computed energy gap between the two structures is negligible; in this case, the PW basis set implementation of +U is not sensitive to the geometry at which the electronic structure is computed. This result contrasts with the NAO data: for each U-value, NAO calculations predict a larger energy gap for the FHI-aims structure, relative to the VASP structure. The energy gaps computed from single point NAO calculations over the FHI-aims relaxed structure (red) are positively offset by ~0.5 eV with respect to those values computed over the VASP relaxed structure (black). The change in the energy gap with increasing U is consistent, regardless of the atomic structure, as revealed by the slopes (a-values) of the red and black trendlines, presented in Table I; i.e. the 0.5 eV offset is maintained over the range of U-values. This result is interesting because, as discussed, both FHI-aims and VASP predict similar structures, vis length and width, at the PW91 (U=0) level. However, small differences in the atomic structures yield appreciable differences in the energy gaps computed with the NAO basis set, while no differences were shown with the PW basis set. This highlights that, to avoid misunderstanding interpretations in the analysis of the electronic properties, structural relaxation is crucial when using NAO basis set. It appears that the impact of U is greater with NAO, related to the localized projector functions. 41 It is also interesting to compare NAO and PW results when these calculations are performed on the same starting structure. For the FHI-aims relaxed structure, the energy gaps predicted by NAO (red) and PW (green) calculations are in agreement for small U-values, but the differences in the predicted gaps increase with increasing U. This is reflected in the slopes (a-values) of the trendlines fitted to the NAO (red) and PW (green) data, which are 0.103 and 0.075, respectively (see Table I). In this case, the energy gap varies to a greater extent in the NAO calculations, which consistently predict larger gaps with respect to the PW calculations. Conversely, for the VASP relaxed structure, the energy gaps predicted by NAO (black) and PW (blue) differ over the entire range of considered U-values. For U = 0 eV, the PW-computed energy gap is larger than that computed with NAO by ~0.5 eV, but this difference decreases with increasing U, in accordance with the larger slope for the NAO data (0.106), with respect to that of the PW data (0.080). These results suggest that the differences observed in the computed Kohn-Sham energy gaps are not attributable to differences in the atomic structure, but rather to differences in the implementation of DFT+U for the NAO or PW basis set. Finally, we note that each of the computational set-ups, with the exception of NAO calculations on the VASP relaxed structure (black), predict similar energy gaps of ~2.5 eV for U = 0 eV. For these three set-ups, the differences in the computed energy gaps are reasonable, i.e. within 0.15 eV, for U-values up to 4 eV. For U > 4 eV, the NAO basis set promotes a larger energy gap with respect to the PW basis set. The data obtained from the calculations described in scenarios (iii) and (iv), above, are presented in Fig. 4 and Table II. We first look at the computed energy gaps for the structures optimized at each U-value in FHI-aims (red) and VASP (blue). The energy gaps computed with the NAO basis set increase from 2.5 eV to 3.8 eV as U increases from 0 eV to 10 eV. This monotonic increase with U is expected and is corroborated in the trendline data, shown in Table II. Interestingly, the opposite trend is observed for the energy gaps computed for the structures that were fully relaxed at each U with the PW basis set: in this case, the energy gaps decrease monotonically with increasing U. As seen in our discussion of Figure 3, increasing the U-value in a PW calculation on a fixed structure yields a larger energy gap. Thus, here we must attribute the decrease in the energy gaps to effects arising from the structural optimization at each U. This result is surprising, not only because it is unexpected, but also because the changes in the PW-computed atomic structures over the range of U-values are modest (see Fig. 2), yet the impact on the electronic structure is significant, with states in the gap attributed to the presence of the low coordinated O atoms; see density of states plots in Fig. 5. In fact, for the VASP-relaxed PW91 (U=0) structure, a single-point PW calculation with U = 4 eV yields an energy gap of 2.76 eV whereas for the fully relaxed structure the energy gap is 2.35 eV. In other words, the emergence of the gap states occurs at lower U values in the PW calculations. This is clearly seen in the density of states plots in Fig. 5 corresponding to the VASP and FHIaims calculations for U = 2 and 6 eV, respectively. Performing a single-point PW calculation on the FHI-aims relaxed structures at each U-value produces the energy gaps represented with the green data points in Figure 4. Here we see that the data points agree with those computed with the NAO basis set (red) within 0.1 eV up to U = 4 eV, after which the differences increase. This is in agreement with the trendline data listed in Table II; the slopes for the NAO (red) and PW (green) basis sets are 0.136 and 0.095, respectively. Importantly, single-point PW calculations on the FHI-aims relaxed structures, at each U, predict an increase in energy gap with increasing U. This further confirms that the decreasing trend in energy gaps for the VASP-relaxed structures arises from structural effects. The energy gaps computed with single-point NAO calculations on the VASP-relaxed structures, at each U, are shown with the black data points in Fig. 4. An outlier in this data is the energy gap computed for U = 0 eV, which is 2.02 eV. This value has been checked and the presence of an error in the calculation can be ruled out. Interestingly, for U = 1-10 eV, the computed energy gaps are consistently ~2.5-2.6 eV and this data shows no discernible increasing or decreasing trend. As seen in our discussion of single-point NAO calculations on both the FHI-aims and VASP PW91 (U=0) relaxed structures, the predicted energy gaps increase monotonically with increasing U. Once again, this suggests that subtleties in the structural optimization within the PW implementation of DFT+U, probably linked to the low coordinated O atoms at the NP edge, produce these effects in the electronic structure. For the NAO calculations, consistent with the linear trends for the red data reported in the legends of Figures 3 and 4, the relaxation at each U value has a negligible effect, as expected, on the fitting offset with respect to the calculation at the PW91 (U=0) structure. However, the fully relaxed calculations result in changes in the fitting slope. Thus, the opening of the energy gap is more pronounced for the fully optimized structures when employing the NAO basis. This latter situation, where the NP structure is fully relaxed at each U in each code, is the most reasonable scenario to analyze the different behavior observed between basis sets because artifacts due to the use of a structure not optimized within the method/basis set are ruled out. First of all, the energy gaps between the PW and NAO basis set are shifted by 0.25 eV (see Fig. 3), which can be attributed to a different treatment of the effect of the core electrons and also relativistic effects. 58,59 The former are included explicitly in the calculations with the NAO basis set, whereas they are included through a frozen orbital type approach through the PAW in the calculations with the PW basis. Similarly, the relativistic effects are included explicitly at the ZORA level with the NOA basis and implicitly through the PAW description of the core electrons in the PW calculations. In principle, the most accurate results are obtained from the all-electron basis set implemented in FHI-aims. The most relevant results are found in the variation of the energy gap in response to increasing U. These are depicted in Figure 4 and the trends (Table II) reflected in the linear fittings with slopes of 0.136 and -0.028 for NAO and PW basis set, respectively. This result clearly shows the effect of U on the resulting energy gap does not only depend on the numerical value of this parameter but also on the projection of the Kohn-Sham states to determine the occupation numbers that enter the +U correction and the structural optimization, which, in turn, depend on the basis set used. Thus, the +U part of the exchange-correlation potential severely depends on the DFT code, as already shown by Kick et al. for some systems. 41 To clarify this issue, we comment on how results from the PW91+U approaches used In summary, the U value fitted to reproduce an experimental or hybrid functional calculated value using a given DFT code cannot be transferred to another code as it depends on the basis set used and, on the method employed to define the corresponding projectors. Thus, for each materials system and DFT code, one should recompute suitable values for U through making initial benchmarks. ## IV. CONCLUSIONS The effect of the DFT+U method on the structural and electronic properties of the (TiO2)35 NP is systematically investigated by two different basis sets, namely, plane-waves (PWs) and numerical atomic orbitals (NAOs), along with different approaches for the implementation of U value. In the absence of U, PW and NAO calculations report the same structure and, consequently, the structural variations observed by its inclusion are due to the different implementation of U based on a simplified rotationally invariant approach and a linear combination of the NAO basis functions, respectively. Interestingly, the analysis of the energy gap reveals that a certain U value can reproduce the experimental value, however, it depends on the basis set and on the employed U parameter. Therefore, the transferability of U values between codes is not to be recommended and requires initial benchmarks for the property of interest as a reference to find the appropriate value. This study clearly shows that the DFT+U implementation in a localized basis set code such as FHI-aims entails much lower values of U to reproduce results obtained with a plane wave basis set code such as VASP. II. FIG. 5 Projected electronic density of states (PEDOS) of the full relaxed (TiO2)35 NP using PW and NAO basis sets for U=0, 2, 4, and 6 eV.

chemsum

{"title": "On the use of DFT+U to describe the electronic structure of TiO2 nanoparticles: (TiO2)35 as a case study", "journal": "ChemRxiv"}

energetic_basis_and_design_of_enzyme_function_demonstrated_using_gfp,_an_excited-state_enzyme

## Abstract: The last decades have witnessed an explosion of de novo protein designs with a remarkable range of scaffolds. It remains challenging, however, to design catalytic functions that are competitive with naturally occurring counterparts as well as biomimetic or non-biological catalysts. Although directed evolution often offers efficient solutions, the fitness landscape remains opaque. Green fluorescent protein (GFP), which has revolutionized biological imaging and assays, is one of the most re-designed proteins.While not an enzyme in the conventional sense, GFPs feature competing excited-state decay pathways with the same steric and electrostatic origins as conventional groundstate catalysts, and they exert exquisite control over multiple reaction outcomes through the same principles. Thus, GFP is an "excited-state enzyme". Herein we show that rationally designed mutants and hybrids that contain environmental mutations and substituted chromophores provide the basis for a quantitative model and prediction that describes the influence of sterics and electrostatics on excited-state catalysis of GFPs.As both perturbations can selectively bias photoisomerization pathways, GFPs with fluorescence quantum yields (FQYs) and photoswitching characteristics 1-4 tailored for specific applications could be predicted and then demonstrated. The underlying energetic landscape, readily accessible via spectroscopy for GFPs, offers an important missing link in the design of protein function that is generalizable to catalyst design. ## INTRODUCTION Numerous methods have been employed in developing GFPs with desired behaviors , including directed evolution and high-throughput screening of mutant libraries that optimize brightness. Machine learning has afforded redder and brighter GFPs 10,11 , and de novo protein design has reduced the size of GFP 12 . Unfortunately, the former lacks physical insight, and the latter does not factor in structure-FQY relationships, leading to a FQY (~ 2%) substantially below those of GFPs derived from Aequorea victoria (avGFP; FQY ~ 80%). Only through further substantial screening and chromophore modification were brighter versions (FQY ~ 23%) obtained 13 . Photoswitching, the ability to toggle between strongly and weakly fluorescent states through irradiation 18,19 , is another useful function that facilitates super-resolution imaging and optogenetic applications 20,21 . One of the most common photoswitching mechanisms is photoisomerization (Figure 1A), an excited-state bond-rotation pathway that competes with fluorescence emission. Due to this competition, selecting for an efficient photoswitchable protein is difficult via high-throughput screens; past efforts have relied on naturally occurring photoisomerizable GFPs as starting points 14 and/or painstaking combinations of rational design and screening . A physical framework capturing the protein environmental factors that control the FQY and photoisomerization in GFPs is necessary to guide more efficient designs, and this is intimately related to the challenge of catalyst design. Potential energy surface (PES) for the GFP chromophore along the isomerization coordinate. After excitation from the cis ground state (indigo arrow), the chromophore can either fluoresce (kfl) or decay by isomerization through excited-state barrier crossing (kiso) and conical intersections (trajectory not shown) or by other nonradiative pathways (kother) back to the ground state. Isomerization can either occur about the phenolate bond (P bond; kP, phenolate ring flip) or the imidazolinone bond (I bond; kI, cis-trans isomerization), with opposite directions of electron flow. The relative barrier heights (EP and EI) depend on steric and electrostatic factors of the environment around the chromophore 22 , catalyzing one pathway over the other. (B) The driving force of the chromophore 𝛥𝜈̅ is defined as the relative energy between the P (left) and the I (right) resonance forms in a given environment. In all proteins studied in this work, the P form is consistently lower in energy 23 , defined as a positive driving force. (C) Marcus-Hush model explaining shifts in transition energy 𝜈̅ 𝑎𝑏𝑠 depending on the electrostatic influence of the protein environment on the chromophore's ground and excited states 23 . (D) The chromophore and its local environment within Dronpa2. R66, S142, and T159 are the residues mutated in this work, while tyrosine analogues in place of Y63 are used to introduce substituents into the phenolate ring of the chromophore 22 . In earlier work, we discovered that the FQY of the anionic GFP chromophore embedded in the fixed native protein environments of Dronpa2 or superfolder GFP can be modulated through the introduction of electron-donating and -withdrawing substituents 22 . The FQY exhibits a peaked trend when correlated with transition energy (Figure 2A; now converted into driving force, vide infra); the shift in transition energy reflects the extent of electronic perturbation conferred by the substituents. This observation reveals two competing nonradiative photoisomerization pathways (Figure 1A), with the probability of each influenced by the electrostatic interaction between the protein environment and the electron flow within the chromophore during photoisomerization 24,25 . Because the twisting about the two exocyclic bonds (the P and I bonds) in the excited state is associated with opposite electron flow directions (Figure 1A), electrostatics can cause bond-selective photoisomerization of the chromophore, complementing the more commonly argued role of steric hinderance in suppressing chromophore (photo)isomerization 3,26,27 . The relative barrier heights EP and EI determine the outcome, and control of these barrier heights is analogous to conventional concepts in catalysis. To quantify this electrostatic perturbation, we use the driving force Δ𝜈̅ (Figure 1B) 23,28 , which is the relative energy between the P and I resonance forms of the chromophore. Δ𝜈̅ is obtained from the observed transition energy (absorption peak maximum) 𝜈̅ 𝑎𝑏𝑠 through the Marcus-Hush treatment 23,28 : where V0 (= 9530 cm -1 23 ) is the electronic coupling between the two resonance forms. With respect to the wild-type environment or chromophore, any decrease or increase in Δ𝜈̅ caused by modifications results in a red or blue shift, respectively (Figure 1C). The driving force can be perturbed through either direct modification of the chromophore or through changes in the protein environment, so it can serve as an ideal quantity to reflect the electron distribution of the chromophore 23 , unify both sources of perturbations 29 , and connect to the underlying theme of electrostatic catalysis. ## RESULTS AND DISCUSSION 2.1. Tuning Electrostatics with Mutants and Hybrids. Figure 1D shows the chromophore environment of Dronpa2, which exhibits a balance between emission and photoisomerization. To isolate the electrostatic effects, residues immediately surrounding the chromophore were replaced with amino acids that minimized differences in size. The S142A mutation causes a red shift by destabilizing the P form through removal of a hydrogen bond to the phenolate oxygen (Figures 1B, 1C, S1A, and S2A). The blue-shifted R66M mutant results from I-form destabilization via the removal of the favorable electrostatic interaction between the arginine and the imidazolinone oxygen (Figures 1B, 1C, S1A, and S2B). Within an isosteric T159 mutant series (T159M, T159Q, T159E), T159M is the most red-shifted (by 15 nm compared to wild type), while increasing polarity and/or charge causes a blue shift in T159Q/E; the glutamine and glutamate in T159Q and T159E mutants, respectively, replace S142 as the primary hydrogen bonding partner to the phenolate oxygen and preferentially stabilize the P form (Figures S1A and S2C-S2F). We next measured the FQYs (Table S1) and plotted them against the corresponding driving forces (eq 1) to determine the electrostatic effect on photoisomerization (Figure 2A). S142A and R66M have a decreased FQY along with strong red-and blue-shifted peak maxima, respectively, recapitulating the peaked trend for chromophore variants (Figure 2B). In contrast, the isosteric T159 mutant series displays a linear correlation with peak maximum, rendering Dronpa (T159M) an outlier of the trend. We attribute this to an increased steric effect for the isosteric series in conjunction with the electrostatic mechanism (vide infra). Nevertheless, we still find that the FQY can be tuned electrostatically through environmental mutations. To circumvent the confounding steric effect, we created hybrids by introducing substituted chromophores into environmental mutants. We first chose one red-shifted (S142A) and one blue-shifted (T159E) mutant with the wild-type Dronpa2 chromophore. We then introduced electron-donating or -withdrawing chromophore substituents to the P ring, which would be predicted to either respectively enhance or compensate for the electronic effect of the mutant with respect to wild-type properties. For example, as the S142A mutation destabilizes the P form, an "enhancing" chromophore modification would be electron-donating and push the electronic properties of the chromophore (driving force and FQY) even further from wild type. A "compensating" modification with an electronwithdrawing group would stabilize the P form, countering the mutational effect and creating a more wild-type-like chromophore (Figure 2C). Note that the same substituent can act as enhancing or compensating in different environmental contexts according to electrostatic FQY tuning. For the hybrids, we can quantitatively predict the optimal substituent, within the range available 22 , to pair with a given mutant based on driving force additivity (Table 1). Each point mutant has a driving force, to which a fixed value is added or subtracted based on the chromophore substituent, obtained from the difference between the driving force of Dronpa2 with a natural and substituted chromophore 23 . For the compensating hybrids, the optimal substituents to bring the driving force of S142A and T159E close to wild type are 2,3-F2 and 3-OCH3, respectively. For the enhancing hybrids, we chose substituents with low steric bulk but that still provide a large perturbation to the driving force: S142A/3-CH3 and T159E/2,3-F2. The observed absorption peak maximum for each hybrid agrees well with the predictions (Table 2; Figures S1B and S1C): incorporation of electrondonating and -withdrawing substituents leads to the predicted red and blue shift, respectively. Figure 2D shows the correlation between FQY and driving force for the Dronpa2 hybrids. Both enhancing hybrids (S142A/3-CH3 and T159E/2,3-F2) have a decreased FQY, pushing the values further from wild type as anticipated from electrostatic FQY tuning. Remarkably, both compensating hybrids (S142A/2,3-F2 and T159E/3-OCH3) have an increased FQY compared to the respective mutant with the unsubstituted chromophore, bringing the values closer to the wild-type value. This observation implies that the electronic effect of the chromophore substituent successfully compensates for the electrostatic perturbation caused by the environmental mutation. Either the chromophore substituents (2,3-F2 or 3-OCH3) or the environmental mutations (S142A or T159E) alone each cause a decrease in FQY compared to the wild-type Dronpa2, so the observation of an increased FQY in these compensating hybrids suggests cooperativity ("reciprocal sign epistasis") 8,30 between deleterious perturbations that cannot otherwise be explained without electrostatic FQY tuning. The FQY φfl is the ratio between the intrinsic spontaneous emission rate kfl and the total excited-state decay rate constants 31 (Figure 1A): where kiso and kother denote the total rate constant for excited-state isomerization and other nonradiative pathways, respectively; 𝜏 is the fluorescence lifetime. We can then dissect the temperature, electrostatic, and steric dependence of each term to understand how the chromophore's FQY is influenced by its environment. kfl is minimally tunable through electrostatics as evidenced by the nearly constant transition dipole moment across different GFP mutants 23,32 ; steric effects are irrelevant since emission is a Franck-Condon process. The only way the protein environment can tune the FQY is through modulating the competing nonradiative decay pathways. kfl is estimated to be (3.5 ns) -1 33 , so any nonradiative process much slower than this value cannot tune FQYs. kother arises from both direct internal conversion and intersystem crossing, but the latter is much less competitive than other excited-state processes 34 . Accordingly, we can approximate kother with a single rate constant from direct internal conversion kIC due to vibrational wavefunction overlap between the ground and excited electronic states, which is relatively temperature insensitive (see Section S6 of ref. 22 and Section S11 of ref. 30). To obtain kIC, we examine a GFP mutant series in which the threonine at position 203 is replaced with aromatic side chains that π-π stack with the chromophore P ring and can be varied in electron richness. The corresponding FQYs are nearly constant around 77% despite the modified electrostatic interaction (Figure S3 and Table S2). Steric hinderance by the aromatic ring overwhelms electrostatics and renders kiso uncompetitive; the remaining 23% of excited-state decay can be ascribed to internal conversion; kIC is (12 ns) -1 and imposes an upper limit for GFP's FQY of approximately 80% 35 , close to that of avGFP. Extensive mutational studies also demonstrate that avGFP is indeed located at the local maximum of the fitness landscape for brightness 8 . Any approach that slows excited-state isomerization down to tens of nanoseconds is sufficient to maximize FQY. In contrast with other processes, excited-state isomerization requires crossing over an energy barrier along with significant electronic and nuclear motion (Figure 1A), so the isomerization rate kiso is almost solely responsible for the temperature, electrostatic, and steric dependence of FQY 22 . The associated barriers are typically > 3 kcal/mol for GFPs 22 , and the corresponding rate constants are comparable with kfl (ns timescale). The rapid intramolecular vibrational energy redistribution (ps timescale) 31,39 right after excitation renders the system thermally equilibrated before emission and isomerization, so the assumption for Arrhenius behavior, also common for ground-state catalysis, is met for isomerization. A pre-exponential factor A and an energy barrier E can thus be assigned for each isomerization pathway: where kiso is then approximated with a single Arrhenius expression when we measure the excited-state energy barrier E of Dronpa2 variants using the temperature dependence of their fluorescence lifetimes 22 . If AP and AI are close in value, A should be close to both AP and AI, and the measured excited-state barrier height E can well approximate the lesser of the two barriers, EP or EI (Figure 1A). A is 10 3 -10 5 ns -1 22 , agreeing well with the value estimated from transition state theory ( 𝑘𝑇 ℎ ~ 10 13 s -1 ). This suggests that when the excitedstate barrier exceeds 9 kcal/mol (i.e., kiso being 1% of kfl at 300 K), as for the π-π stacking GFP mutants (Figure S3), no further increase in FQY can be seen as it reaches the upper limit. We now replot the excited-state barriers from Dronpa2 variants (Figure 3B in ref. 22) against the corresponding driving forces to better understand the electrostatic effect (Figure 3A). Linear fits to the electron-donating and -withdrawing substituent data exhibit slopes of +0.6 and -0.7, reflecting the electrostatic sensitivity of EP and EI, respectively. These slopes are about equal in magnitude (~ 0.65 within experimental errors) and opposite in sign; the signs agree well with a model treating the chromophore as an allylic anion 22 . Analogous to electrostatic enzyme catalysis 40,41 , this electrostatic sensitivity originates from chromophore charge redistribution during photoisomerization interacting with the protein environment (Figure 1A), effectively an excited-state enzyme that selectively catalyzes either P-or I-bond rotation. We expect these slopes in Figure 3A to be directly transferable to different environments around the chromophore, since the driving force is the only parameter responsible for the electrostatic sensitivity of the entire PES 22 : 𝐸 𝑃 = 0.65Δ𝜈̅ + 𝐶 𝑃 and 𝐸 𝐼 = −0.65Δ𝜈̅ + 𝐶 𝐼 (4) where the steric effects, including the intrinsic barrier to bond isomerization in the absence of any external steric constraint, can be separated out in terms of empirical constants CP and CI (y-intercepts of red and blue lines in Figure 3B, respectively). We can then rewrite eqs 2 and 3 to explicitly show the electrostatic and steric dependence of the FQY: Two factors mediate excited-state pathway selection: sterics, which acts upon large scale nuclear motion of two rings during isomerization, and electrostatics, which interacts with electronic redistribution during isomerization (or driving force). The electrostatic influence of the red fluorescent protein environment on the corresponding chromophore's FQY is also extensively discussed by a recent paper 42 , while our physical model treats electrostatics differently and explicitly incorporates the steric component (see Section S2 in Supporting Information). According to eq 5, FQY is a nonlinear function of Δ𝜈̅ , and thus the linear additivity of driving force does not translate to an additivity of FQY, as observed from the compensating hybrids (Figure 2D and Table 2). Cooperativity between mutations, a phenomenon that renders protein design and even directed evolution challenging 29,43 , could similarly be partly explained by a nonlinear function (i.e., FQY) encoding two (or more) pathways dependent on an additive underlying parameter (i.e., driving force) 23 . Steric effects CP and CI serve as an alternative tuning mechanism for the excited-state barriers EP and EI, preventing the FQY from being completely tied to color via electrostatics, as is the case for other photophysical properties 23 . If CP equals CI, there should be no preference for either isomerization pathway when Δ𝜈̅ = 0, corresponding to a maximum FQY (eq 5; Figure 3B, case 1). Since Δ𝜈̅ = 0 also corresponds to the reddest possible absorption (eq 1), a combination of these two equations would suggest that the redder the chromophore, the higher the FQY by varying Δ𝜈̅ . However, we observe an apex in the trend that is not centered at Δ𝜈̅ = 0 (Figure 3A), suggesting that CP is not identical to CI. Intuitively, the volume-demanding I twist experiences more steric hinderance than the P twist within the protein environment since the I ring is covalently anchored. With eq 4, we can explain the apex position in the FQY (or excited-state barrier) vs driving force plot (Figure 3B). The sign of the driving force is defined positive when the P form is more stable than the I form, which is the case for all proteins studied so far 23 (Figure 1B). With zero differential sterics from the protein environment (CP = CI; dashed lines) and zero driving force, the negative charge of the anionic chromophore is maximally delocalized and both exocyclic bonds are equally probable to twist upon excitation. This corresponds to the largest possible barrier when CP = CI, and the apex is located at Δ𝜈̅ = 0 (Figure 3B, case 1). When the driving force becomes positive (right side of Figure 3B), electron density is reduced at the I bond (i.e., more single-bond character) upon excitation, and the I twist becomes more favorable 44 (Figure 3B, case 2). If the I ring is anchored inside the protein, CI becomes larger than CP (yellow arrows and solid lines in Figure 3B). Consequently, the apex shifts along the x-axis and lies at a positive driving force, as observed in Figure 3A, and it also increases along the y-axis due to the resulting constriction on bond rotation (Figure 3B, case 3). At that apex, the driving force from electrostatic influences matches the apex shift caused by differential steric interactions. However, when the steric effects are large enough to render kiso uncompetitive with kfl (Figure S3), the maximally allowed FQY is reached, and the apex for FQY cannot be detected. Note that the driving force at the apex is determined from the differential sterics (CI -CP), while the barrier heights are affected by the absolute sterics (CI or CP), so it is possible to have an apex location at zero driving force when steric hinderance to the P twist is comparable with I ring anchoring (Figure 3C). ## Applications, Generalizations, and Implications for Design. This model allows us to quantitatively evaluate the contributions of sterics and electrostatics to excited-state catalysis. From Figure 3A, wild-type Dronpa2 sits at the apex among all Dronpa2 variants. As its FQY (~ 50%) is far from the maximally allowed 80%, this implies that the corresponding driving force (23.6 kcal/mol) offsets the differential sterics, so we can estimate the differential sterics as 31 kcal/mol (23.6 × 2 × 0.65, Figures 3C and 3D). For superfolder GFP, the apex (the monochlorinated variant, Figure 2A) lies at a driving force of 19.9 kcal/mol and approaches the FQY limit of 80% 22 . The corresponding differential sterics is 26 kcal/mol (= 19.9 × 2 × 0.65). Combined with the fact that GFP has a higher apex FQY than Dronpa2, we can infer that the overall steric contribution should be higher for GFP than Dronpa2, but the differential sterics is also 5 kcal/mol smaller (= 31 -26) for GFP, leading to an apex located at a smaller driving force than Dronpa2 (Figure 3D). This is explained by a tighter β-barrel for GFP compared to Dronpa2, resulting in a more sterically hindered P twist (Figure 4A). Moreover, since the unmodified chromophore in Therefore, both steric and electrostatic (to a lesser extent) effects work together in the GFP barrel to promote chromophore fluorescence, while Dronpa2 exhibits a higher photoisomerization efficiency (Figure 5A). For the Dronpa2 T159 isosteric series, the lengthened side chain creates more steric bulk to P twist and shifts the apex to a smaller driving force and higher FQY (Figure 2B), explaining why T159M appears as an outlier to the peaked trend. This analysis can also explain why the de novo designed mFAPs (Figure 4B) failed to recapitulate avGFP's high FQYs (Figure 4C) 12 and more generally how an understanding of the energy landscape can provide guidance for the design of functional proteins. Original mFAPs utilize the same difluorinated chromophore as the RNA mimic Spinach (Figure 4D) 45 to encourage chromophore deprotonation, but fluorines lower the I-twist barrier as electron-withdrawing substituents 22 . In Spinach, π-π stacking with Gquadruplexes effectively inhibits isomerization (Figure 4D) 45,46 , leading to a FQY of 72%. In mFAPs, however, the chromophore is neither anchored to the protein as in avGFP (Figure 4C) nor motionally restricted. M27W is present in mFAP1 and mFAP2 to interact with the I ring via a hydrogen bond (Figure 4B), but this interaction is not sufficient to restore the maximal FQY. To further increase the FQYs, this analysis suggests the addition or removal of fluorines from the chromophore's I or P rings, respectively, and the introduction of aromatic amino acids near the chromophore's P ring to encourage π-π stacking interactions. In fact, the newly installed -CF3 group on the I ring and L104H likely explains the much-improved FQY (23%) of chromophore-bound mFAP10 13 . ## CONCLUSIONS GFP is both green and fluorescent, while the free GFP chromophore in water is neither, so it is tempting to ascribe this drastic change in properties to the protein environment. However, the chromophore's ability to be green and fluorescent is already encoded in its PESs (i.e., energy landscape), and these properties can also be elicited using non-protein environments 3,27 . An analogous example is the relationship between an enzyme and its substrate. The availability of different reaction pathways and the potential for pathway selection, existing for numerous ground-state and excited-state enzymes , are already inscribed in the PES(s) of the chromophore/substrate, illustrated by diverse examples in Figure 5. The protein environment can only stabilize the transition state of one particular pathway over another that is otherwise suppressed; it cannot than create new reactions. Therefore, to rationally design enzymes that are superior at catalyzing a reaction, it is important to sample a wide range of perturbations to substrates (or chromophores capable of structural change) and the environment's steric or electrostatic influences on the energetics of non-productive yet competitive pathways rather than only those that exhibit more desirable phenotypes 51 . Only when those less desirable cases are understood can we mechanistically deduce why the more productive pathway is not taken, guiding future design efforts to optimize the desired function. Dronpa2 (Figure 3D). (B) Y(M210)F mutant (purple) of Rhodobacter sphaeroides photosynthetic reaction center reveals that tyrosine at M210, which stabilizes the first intermediate, is in part responsible for the unidirectional excited-state electron transfer of wild type (orange) 52,53 . (C) Wild-type Fe(II)/2-oxoglutarate (2OG)-dependent halogenases (orange) chlorinate their substrates, but their intrinsic hydroxylating power can be unleashed upon mutation (purple) 54,55 . The default (blue) and the side pathways (red for all and green for panel A) are shown on the right and left for each panel, respectively. Energies are not drawn to scale.

chemsum

{"title": "Energetic basis and design of enzyme function demonstrated using GFP, an excited-state enzyme", "journal": "ChemRxiv"}

modular_microfluidic_system_for_on-chip_extraction,_preconcentration_and_detection_of_the_cytokine_b

## Abstract: The cytokine interleukin 6 (IL-6) is involved in the pathogenesis of different inflammatory diseases, including cancer, and its monitoring could help diagnosis, prognosis of relapse-free survival and recurrence. Here, we report an innovative microfluidic approach that uses the fluidization of magnetic beads to specifically extract, preconcentrate and fluorescently detect IL-6 directly on-chip. We assess how the physical properties of the beads can be tuned to improve assay performance by enhancing mass transport, reduce non-specific binding and multiply the detection signal threefold by transitioning between packed and fluidization states. With the integration of a full ELISA protocol in a single microfluidic chamber, we show a twofold reduction in LOD compared to conventional methods along with a large dynamic range (10 pg/mL to 2 ng/mL). We additionally demonstrate its application to IL-6 detection in undiluted serum samples. Biomarkers are considered as objective quantizers of biological processes and particularly pathophysiological processes; they can be used for patient diagnosis or prognosis as well as to monitor disease progression or patient response to treatment. Biomarkers provide guidance through the development of new medicines and are pivotal to decipher molecular or cellular mechanisms involved in pathologies. The increased interest for biomarkers has been accompanied by the emergence of a wide range of bioanalytical developments such as mass spectrometry or high throughput screening. Among the different biomarkers (cellular, molecular, vesicular), proteins have significantly demonstrated their potential and many of them are analysed and quantified for clinical diagnosis of diseases from asthma and allergies 4,5 through infections 6 and cancer 7 . Cytokines are small proteins involved in cell signalling often used as indicators for disease monitoring 8 such as in tumor progression 9 , liver diseases 10,11 or hepatic inflammations and fibrosis 12 . In particular, interleukin 6 (IL-6) is involved in the response of the human immune system to infection and cellular injury 13,14 , being secreted by T cells and macrophages into the serum in case of acute and chronic inflammation. Recently, it has been suggested that coronaviruses may activate dysregulated host immune responses. Exploratory studies have suggested that interleukin-6 (IL-6) levels are elevated in cases of complicated COVID-19 15,16 . Thus, a quantitative analysis of cytokines in bodily fluids, and IL-6 in particular, can benefit the monitoring of a wide range of diseases. The current standard methods to detect and analyse cytokines are immunoassays, typically in the form of ELISA, microarrays and bead-based assays 17 . While immunoassays can be highly specific and sensitive, cytokine detection by batchwise immunoassay remains challenging due to their very low concentrations in biological samples down to sub pico or femto-molar concentration . The potential benefits of microfluidics are multi-fold: decrease analysis time, improve bioassays sensitivity, reduce sample and reagent volumes, decrease costs and miniaturize and integrate complex protocols. Impressive results of IL-6 detection in microfluidic systems have already been published relying on glass capillary 21 , modified controlled-pore glass packet 22 or carbon nanotube forests 23 . But while several microfluidic systems have already OPEN 1 Laboratoire Physico Chimie Curie, Institut Curie, PSL Research University, CNRS UMR168 Paris, France. 2 Institut Pierre-Gilles de Gennes, Paris, France. 3 LAAS-CNRS, Université de Toulouse, CNRS, INSA, 31400 Toulouse, France. 4 Université Paris-Saclay, CNRS, Institut Galien Paris-Saclay, 92296 Châtenay-Malabry, France. * email: lucile.alexandre@curie.fr; stephanie.descroix@curie.fr shown their efficiency for cytokine detection and quantification, there is still a need for new technologies and methods that can tackle the challenge of precise detection in complex matrices at low cost. Here we approach this challenge by improving current immunoassay-based protocols in an integrated bead-based microfluidic format. Immunoassays have been widely implemented in microfluidic , initially as a miniaturization of conventional microtiter plate ELISA, the antibody being grafted at the surface of the microchannel 28,29 . To further improve the specific surface of interaction and consequently the surface to volume ratio, solid supports have been inserted in microdevices starting with mechanical trapping of micrometric polystyrene beads functionalized with antibodies 30 . The interest of using microbeads as solid support was exemplified by Teste et al. demonstrating theoretically and experimentally that the kinetics of target analyte capture is improved by using micro-and nano-magnetic particles compared to standard microtiter plates 31,32 . Since then, other strategies have been investigated leveraging electrokinetic and magnetic forces , in particular in the form of droplet immunoassays combined with magnetics beads . Previously, we developed the microfluidic magnetic fluidized bed, a beads-based microfluidic technology based on a homogeneous suspension of magnetic beads inside a microfluidic chamber 40 . A balance of drag and magnetic forces on the beads results in physical properties similar to those of a macroscale fluidized bed. The resulting high surface to volume ratio, constant mixing and compatibility with commercial and functionable beads make it attractive for bioanalysis integration. The porosity of the bed of beads plays a key role in the efficiency of the system, as it could affect the sample residence time and diffusion distances to the solid phase. This was demonstrated in a wide range of applications: bacteria analysis in raw samples 41,42 , detection of histone modifications 43 and as a miniaturized platform for extracorporeal circulation 44 . However, sensitive protein detection requires relatively complex multi-step protocols, challenging to integrate in a single device. Here we leverage the fluidized bed as a tool able to perform all the steps of an ELISA protocol for IL-6 detection: specific extraction, preconcentration, enzymatic binding and detection in a single microfluidic chip. We optimize and evaluate the performance of each step and study how the unique physical features of the mobile solid phase can be tuned to improve assay performance. We do this by adjusting the porosity of the system and the arrangement of the beads according to the molecular diffusion constant and the characteristics of the enzymatic reaction. Finally, we compare the results of our optimized system to the performance of current standard protocols for IL-6 quantitation. ## Materials and methods Reagents and chemicals. The washing buffer was prepared with Tris HCl (200 mM, Sigma Aldrich), Bovine Serum Albumin "BSA" (1%, Sigma Aldrich) and Tween 20 (0.1%, Sigma Aldrich). The pH was adjusted at 7.5. The washing buffer was stored at − 4 °C. The enzymatic substrate MUP (Methylumbelliferyl phosphate, Thermofisher Scientific) was dissolved in the washing buffer at 10 mM, pH adjusted at 8.0. The substrate was kept at − 20 °C. Tosylactivated beads (DynabeadsTM M-280 Tosylactivated, Thermofisher Scientific) were grafted following the Dynabeads datasheet with anti-human IL6 antibodies (Thermofisher Scientific from the kit CHC1263): they were shaken and incubated with anti-IL6 antibodies, Tris buffer and ammonium sulfate (3 M) at 37 °C for 18 h. Then, the beads were washed and resuspended in PBS with BSA 0.1% at a final concentration of 5 mg/mL. The detection antibodies and enzyme were provided by this same kit. Chip fabrication. The microfluidic chip was described in previous publications 40,45 . It consists of an elbow channel leading to a diamond-shape chamber, with an opening angle of 13°. The height of the chamber and channel was set at 50 µm, and the total volume of the PDMS chamber was 0.6 µL. Microfluidic chips were designed using a micro-milled mold. These molds were machined in brass pieces of 5 cm × 5 cm. The designs were a positive replica of the chip. The chips were fabricated by pouring polydimethylsiloxane (PDMS, Sylgard184, Dow Corning) into the molds (concentration 1:10) and were bonded by oxygen plasma. A surface treatment of PDMA-AGE 0.5% 46 was incubated inside the chip chamber for 2 h then rinsed with distilled water and dried with compressed air. ## Microfluidic setup. The liquid flow was produced by a pressurization of the sample reservoir using a pressure controller (MFCSTM, Fluigent) allowing to reach a range of pressures from 0.1 mbar to 1 bar, translated in flowrates between 0.1 and 3 µL/min. The outlet of the chip was connected to a flowrate controller (Flowunit S, Fluigent), which allowed precise flowrate measurements and feedback control on the pressure based on the Maesflo software (Fluigent). Peek tubing (Tube Peek 1/32" × 0.25 mm, Cil Cluzeau Info Labo) was used to connect the microfluidic chip to the other elements of the experimental set up. A tube (12 cm peek tubing of 0.063 mm diameter) was positioned at the entrance of the chip to increase the hydrodynamic resistance of the device. A permanent magnet made of NdFeB 1. 47 T (N52, size 20 mm × 20 mm × 30 mm, magnetization direction through the thickness, by ChenYang technologies) was aligned with the chamber axis at a 1.5 mm distance from the chip inlet. Operating conditions. Manual in-tube labelling ELISA. The sample (50 µL) and the detection antibodies (50 µL, 0.5 µg/mL) were incubated off-chip at room temperature for 50 min with continuous shaking. This mix (50 µL) was flowed inside the chip chamber containing functionalized magnetic beads at a 1 µL/min flowrate. The beads were then rinsed with the washing buffer for 30 min at 2 µL/min inside the chip prior to fluorescent detection. of a homogenous suspension of magnetic beads in a microfluidic chamber. The drag force applied on the beads is due to the flow of liquid in the microchannel and is balanced by a magnetic force created by a permanent magnet 40 . Beads are in free suspension in the liquid phase, a configuration that allows to avoid clogging issues. The microfluidic magnetic fluidized bed has shown interesting features regarding biomarkers analysis as it has already demonstrated its efficiency for nucleic acid analysis 47,48 , but so far it has not been applied to protein biomarker analysis in real samples. Here, we put forward the potential of this technology by combining both preconcentration in a dynamic configuration and on-beads detection. First, a high throughput extraction and preconcentration steps of the analyte are performed directly on the magnetic beads, then the detection of the target biomarker by sandwich ELISA is performed on chip and on beads in a very small volume with properties that can be tuned using the fluidized bed properties to achieve optimal performances. ## Off-chip optimization of bead-based assay. Bead grafting optimization. To implement a microfluidic fluidized bed-based bioassay for IL-6 detection, we first optimized the bead grafting with capture antibodies for IL-6 capture and preconcentration. In this approach, the magnetic beads in the fluidized bed have a pivotal role as they ensure an efficient extraction of the target in a continuous flow while being used in the second step as solid support to perform the sandwich ELISA. In order to optimize the bioassay performance, we first compared two widely used grafting strategies, based on covalently-grafted capture antibody using either tosyl-activated surface or carboxylic acid modified surfaces (tosyl-activated Dynabeads® and MyOne Carboxylic M-270, respectively) 45 . These grafting strategies have been first compared in tube, using 20 µg of capture antibody per mg of beads for the tosyl-activated beads, and 4 µg of antibody per mg of beads for the MyOne Carboxylic beads as advised by the supplier, the capture antibody being here an anti-human IL6 antibody. To compare the different bead-grafting strategies, sandwich immunoassays were performed in the presence or absence of IL-6 to evaluate the specific to non-specific signal. Data shown in Table S1 demonstrated that the tosyl-activated Dynabeads® allow achieving the highest signal to noise ratio with both positive signal higher and negative signal lower than with carboxylic beads. The tosyl-activated Dynabeads™ as solid phase were thus selected as solid support for the IL-6 bioassay development. Buffer optimization. Further optimizations were performed in tube conditions to determine the best parameters to be integrated on chip. In particular, the choice of washing buffer is very important to limit non-specific adsorption; the performances have been evaluated regarding the immunoassay specificity with different buffer compositions. To do so, washing buffers, with pH ranging from 7 to 8 with different ionic strength and ions/ counter ions nature were evaluated: NaHCO 3 (pH = 8, 100 mM), Tris EDTA (pH = 8, 100 mM), Tris HCl (pH = 8, 200 mM), and PBS (pH = 7, 150 mM). Our studies demonstrate a superiority of Tris-HCl buffer over the other ones (data not shown). In parallel, we have also investigated how the presence of BSA and Tween-20 affects the assay performances, both being well known to limit non-specific adsorption on beads as well as on PDMS. The presence of both additives indeed greatly improves the assay performances (Table S2). Finally, we selected as Vol:.( 1234567890 45 . We have thus compared both approaches in the microfluidic fluidized bed. In particular, we have evaluated if the formation of the target-detection antibody complex before its injection on chip could affect its extraction and consequently the bioassay performance. We investigated if the formation of the target-detection antibody complex prior on-chip injection could decrease the diffusion constant of the complex and potentially limits its capture in the continuous flow extraction within the fluidized bed. For the off-chip immune complex formation, the sample, detection antibody and the enzyme were incubated together prior to their on-chip injection (Figure S1.I). The enzyme (Alkaline Phosphatase) was conjugated with the detection antibody through a streptavidin/biotin binding simultaneously with the complex IL-6/detection antibody formation in solution. Magnetic beads were functionalized as previously described (Figure S1.II). The immuno-complex was then injected in the fluidized bed to be captured on the magnetic beads (Figure S1.III). After a washing step to remove the excess of detection antibody and enzyme, the enzymatic substrate was injected within the bed to perform the detection step (Figure S1.IV). In the case of the sequential injection, only the conjugation of the detection antibody by the enzyme was performed in tube prior to on-chip injection. Magnetic beads were functionalized as previously described (Fig. 1I). Then, as in a conventional ELISA in microtiter plate, each step was performed sequentially by injecting in the fluidized bed of each solution as follows: the sample was first injected within the fluidized bed (Fig. 1II), a washing step was performed, conjugated detection antibody was injected inside the chip (Fig. 1III), a washing step was repeated, and then the enzymatic substrate was finally injected to perform the detection (Fig. 1IV). A series of on-chip experiments were conducted to compare these two approaches. Our results showed that a first pre-incubation seemed to slightly enhance the raw signal of detection of the antibody-enzyme complex. These results are in good agreement with previous studies 49,50 . In contrary, the nonspecific signal was significantly lower for sequential injection and the signal to noise ratio was two times higher for sequential injection compared to manual in-tube labelling (Table 1). The fluidization and continuous injection through the suspension of beads allows to reduce non-specific binding to the magnetic beads compared to in tube incubation. As our final goal www.nature.com/scientificreports/ is to inject complex matrices such as serum within the fluidized bed, the reduction of non-specific interaction needs to be prioritized. In addition, the sequential injection had the advantages of simplified automation. Thus, we selected the sequential injection mode for all subsequent experiments, for which the microfluidic fluidized bed features can be optimized to reach lower limits of detection. Optimization of the detection step. The immunoassay format being selected, the detection antibody concentration was next optimized as a compromise between sensitivity and specificity. As previously described, the detection antibody used is an anti-human IL-6 biotinylated antibody conjugated with a streptavidin Alkaline Phosphatase enzyme. The 1X concentration corresponds to a detection antibody concentration of 0.5 µg/mL. The concentration of the antibody-enzyme complex was thus varied between 0.5 µg/mL (1 X) and 25 µg/mL (50 X). As shown in Fig. 2A, the intensity of the signal obtained in the presence of IL-6 at 5 ng/mL can be slightly increased when increasing the detection antibody concentration. However, this goes hand in hand with a significant increase of the non-specific signal and a decrease of the signal to noise ratio (Fig. 2B). The optimal concentration of detection antibody was set at 0.5 µg/mL, condition for which the higher signal to noise ratio was reached while reducing the cost per assay. Table 1. Influence of the process of injection on the specific and non-specific signal. The experiments are performed with a sample of IL-6 at 10 ng/mL (for the specific signal) or a buffer solution mL (for the nonspecific signal) as described in the Material and Methods. parameters of the bead-based immunoassay optimized, we next leveraged the fluidized bed format to improve assay performance. The microfluidic magnetic fluidized bed design has been optimized to reach a high homogeneity of bead distribution within the microchamber 40 , but it has also been shown that the bead bed porosity can be tuned at will within the chip. A change of flowrate induces a change of the drag force applied to the beads. The balance between drag and magnetic forces is modified so that bed of beads expands as the flowrate increases. ## Mode of injection The influence of the flow rate on the on-chip immunoassay has thus been investigated as a change in porosity can impact not only the analysis time (from 25 to 100 min for 50 μL of sample) but also the diffusion distance, as well as the residence time of the target biomolecule within the bed of particles. From previous work 40 , we know that the magnetic beads in microfluidic fluidized beds tend to self-organize in cylindrical clusters of diameter d c ≈12 μm due to bead-bead interactions. In a simple 1D approximation, we can consider the bed porosity ε being defined by the distance between these clusters d s and their size d c : For an efficient capture, the analyte needs to be able to reach, by diffusion, a magnetic bead before leaving the fluidized bed due to the flow-driven convection. If we call t d the time to reach a surface of capture by diffusion and t c the time to cross the magnetic bed by convection, we need to ensure that t d t c ≪ 1 , so that the antibodyantigen interaction can occur effectively within the residence time. The time t d needed to travel the distance d = d s /2 allowing an analyte to reach the closer cluster by diffusion can be estimated with Einstein's relation: 8D , where D is the diffusion constant of the analyte. On the other hand, the residence time of the analyte within the bed can be approximately evaluated as t c = HL 2 tan α 2 /Q , where L is the bed's length, α the aperture angle of the chamber and H the chamber height. Hence, the ratio between both times is: Considering a diffusion constant of cytokine IL-6 D = 8.5 10 -8 cm 2 .s −151 and the aperture angle of the fluidized bed being α = 35° = 0.61 rad, we can estimate the ratio of times based on published measures of the bed length L and porosity ε 40 at flowrates of 0.5, 1 and 2 μL/min (Table 2). While this remains an approximation, note that the ratio t d t c approaches 1 for a flowrate of 1 μL/min, already significantly above 1 for a flowrate of 2 μL/min (Table 2). Hence, we would expect our system to efficiently promote the interactions between the analyte and the surface of capture of the beads up to a maximum flowrate of ~ 1 μL/min. We thus experimentally investigated how the flowrate of the sample injection impacts the specific signal intensity at the outlet of the chip (Fig. 2C). Our experiments showed that a shorter residence along with larger distance to the particles can significantly affect the immunoassay performances, the fluorescence intensity decreasing as the flowrate increases, in agreement with our model previously described. As a compromise between the assay sensitivity and the analysis time, the flowrate was set at 1 µL/min, allowing to reach the higher signal to noise ratio (Fig. 2D). In those conditions, the signal intensity is decreased by 20% compared to a slower flowrate but the analysis time is divided by two and the background noise is decreased by 40%. Furthermore, the correlation between the volume of the sample and the intensity of the detected signal was investigated and we were able to show a high correlation (Figure S2), showing the versatility of our device towards the volume of sample. Finally, the choice of the fluidized bed to integrate IL-6 immunoassay was also motivated by its unique modularity to improve the immunoassay performance. As fluidization occurs when passing liquid through a packed bed of particles at a sufficient velocity to compensate magnetic forces, two regimes can be achieved with this device: below a threshold flowrate, the beads are in close contact and organized as a packed bed of particles, while above this flowrate interparticle distance increases resulting in higher porosity and improved fluid/ solid contact in a fluidized bed regime. These two regimes (packed bed and fluidized bed) can then coexist in our system as a function of the flow rate applied and have been compared here to improve the last step of the immunoassay: the enzymatic reaction. The enzymatic reaction taking place within the fluidized bed offers the possibility to consider two detection modes with potentially improved performances. Indeed, after injection of the enzymatic substrate in the bed, the fluorescent signal can be detected either continuously (in flow method) or by sequentially changing the flowrate above and below the threshold (stop-and-go method). In the continuous in-flow detection approach, the enzymatic substrate was flowed continuously inside the chamber at 1 µL/min for 6 min (Fig. 3A). The bed is, at this flow rate, in a fluidized bed regime; the enzymatic product generated at the surface of the beads is thus continuously flowed through the bed to reach the detection area. The signal has a shape of an asymptotic curve (Fig. 3B). The value of interest is the height of the plateau, proportional to the concentration of IL-6 in the initial sample. Evolution of the ratio t d /t c between the time of diffusion between the capture beads clusters and the residence time inside the fluidized bed due to the flowrate Q of the liquid percolating the bed. www.nature.com/scientificreports/ In the stop-and-go method, a given volume of enzymatic substrate (0.7 µL) is first flowed through the bed at 0.4 µL/min. The pressure is then decreased so that the substrate and the magnetic beads can be incubated in a packed bed regime for 10 min decreasing drastically the diffusion distances between the beads and the enzymatic substrate (Fig. 3C). After 10 min of incubation, the pressure is increased with a retro-controlled program so that the solution is flowed towards the area of detection at 1 µL/min (Fig. 3D). A fluorescence peak is thus obtained (Fig. 3E) as the quantity of product obtained during the 10-min incubation being a finite quantity. The opening of the bed after the 10 min of incubation is a critical step which could affect the shape and dimensions of the peak. By this process, a high quantity of fluorescent product can be accumulated inside the bed of beads before reaching the detection area while switching on the flow rate. With this approach, we aimed at increasing the IL-6 immunoassay sensitivity. The differences in the shapes of the recorded signals are related to the physical properties of the fluidized bed such as the porosity of bead assembly. As shown in Table 3, the signals of three quantification methods of the fluorescent signal were compared to choose the most accurate one. A higher signal was recorded when working with the stop-and-go mode, as expected. Interestingly, the coefficient of variation was smaller when using the peak area measurement rather than the peak height whereas the mean specific signal was higher. It allowed us to reach a signal to noise ratio almost as high as the one of in-flow mode, but with a mean specific signal more than 3 times higher. Finally, despite quite similar performances in terms of signal to noise ratio, we selected the stop-and-go mode rather than the in-flow one in order to reach higher specific signal to lower the limit of detection. Our analytical model showed that an increase of the flowrate above 1 μL/min would limit the antigen-antibody interaction, not allowing improvements when working with the continuous mode. A solution to circumvent this issue lies in the addition of the incubation steps. This choice left more freedom for further optimization if needed: the sensitivity could be increased by optimizing either the injected volume of enzymatic substrate or the incubation time of the stop-and-go mode. Evaluation of the performance of the system for IL-6 detection. To further evaluate the performances of the fluidized bed-based ELISA in terms of dynamic range and sensitivity, we established a calibration S3). A 50 µL sample was flowed through the chip at 1 µL/min then the detection antibody (0.5 µg/mL) was injected at room temperature for 50 min at the same flowrate. A quite large dynamic range was obtained with linear response from 10 pg/mL to 2 ng/mL, as shown in Fig. 4B. This dynamic range of our approach is competitive compared to those reported in the literature 52,53 or to commercial immunoassays 54 usually going from few pg/mL to ng/mL. Negative controls were performed with non-spiked buffer. We next evaluated the sensitivity of our newly developed immunoassay on the basis that significant signal is three standard deviation above the negative control. We obtained a limit of detection (LOD) at 6 pg/mL, more than two times lower than the one of the standard beads immunoassay (LOD is 15.6 pg/mL from manufacter's data). We finally applied the on-chip method to complex samples and performed a calibration curve with IL-6 spiked in fetal bovine serum (FBS) to validate our integrated approach towards complex sample matrix analysis (Fig. 4C). This was performed on a reduced range of concentration, closer to clinical sample conditions. The results show a strong similarity with the results achieved in Tris-HCl, with a linear response in the whole dynamic range and an LOD of 60 pg/mL. Due to its tunable porosity, the microfluidic magnetic fluidized bed is well suited to work with complex matrices. However, as expected due to the high protein content of serum, it is associated with a slightly decreased sensitivity. We assume that some screening effect due to the high protein content of the serum extraction could affect IL-6 specific capture and consequently the assay sensitivity. Altogether our results showed that the tunable properties of a magnetic and microfluidic fluidized bed allows to integrate an automated sequence of on-chip extraction and detection of Il-6. The fluidization regime can be used to limit the non-specific interactions and avoid clogging when working with complex matrices whereas the packed state could be used to enhance the detection step. The control of these two regimes (packed and fluidized states) allowed us to reach a relevant limit of detection in the tens of pg/mL, compatible with, for instance, the requirements of IL-6 detection in patient serum in sepsis . Based on this first proof of concept, the modulable device may be further adapted for the detection of other cytokines. Table 3. Comparison between continuous and 'stop-and-go' methods for experiments performed with a sample of IL-6 at 10 ng/mL (for the specific signal) or a buffer solution mL (for the non-specific signal) as described in the Material and Methods. ## Conclusion Microfluidic systems have demonstrated their potential to enhance bioassay performance and integration, particularly in the case of immunoassays. However, fully integrated multi-step protocols combining analyte extraction, preconcentration and detection in a single module are still challenging, particularly when addressing high sensitivity and/or compatibility with complex matrices. We believe the microfluidic fluidized bed-based approach presented here provides a new way to tackle high-performance immunoassays in fully automated protocols. Its versatility is based on the addition of new variables: the tunable porosity and arrangement of the magnetic beads. In particular, the extraction conditions can be tuned as a function of the diffusion constant of the analyte, while the enzymatic step can further be modified to improve the assay performances. This new technology is also compatible with a large range of biomarker concentrations as well as with sample volumes ranging from one to a few hundred µL. We demonstrate here that it allows a fast detection of the cytokine IL-6 with a large dynamic range (10 pg/ mL to 2 ng/mL), in less than 2 h, with an LOD in the picomolar range. The sensitivity achieved in this first proof of concept is applicable for instance to severe sepsis infection, where IL-6 levels can go up to a few hundreds of pg/mL in human serum samples. A LOD of 6 pg/mL for IL-6 spiked in buffer solution was achieved. This value is lower than the LOD achieved in experiments performed with the same reagents in on-bench conditions, and better overlaps with clinical ranges 19,56 . Moreover, we demonstrated its compatibility with challenging matrices of high protein content and no dilution while still ensuring a clinically-relevant sensitivity. We believe that this capability to enhance the performance of conventional assays while fully integrating complex sequential protocols make this approach a promising tool for future biomarker detection and quantification applications.

chemsum

{"title": "Modular microfluidic system for on-chip extraction, preconcentration and detection of the cytokine biomarker IL-6 in biofluid", "journal": "Scientific Reports - Nature"}

efficiency_gains_for_thermally_coupled_solar_hydrogen_production_in_extreme_cold

## Abstract: Hydrogen produced from water using solar energy constitutes a sustainable alternative to fossil fuels, but solar hydrogen is not yet economically competitive. A major question is whether the approach of coupling photovoltaics via the electricity grid to electrolysis is preferential to higher levels of device integration in 'artificial leaf' designs. Here, we scrutinise the effects of thermally coupled solar water splitting on device efficiencies and catalyst footprint for sub-freezing ambient temperatures of -20 • C. These conditions are found for a significant fraction of the year in many world regions. Using a combination of electrochemical experiments and modelling, we demonstrate that thermal coupling broadens the operating window and significantly reduces the required catalyst loading when compared to electrolysis decoupled from photovoltaics. Efficiency benefits differ qualitatively for double-and triple junction solar absorbers, which has implications for the general design of outdoor-located photoelectochemical devices. Similar to high-efficiency photovoltaics that reached technological maturity in space, application cases in polar or alpine climates could support the scale-up of solar hydrogen at the global scale. Fossil fuels constitute a versatile and large fraction of our energy sources, but contribute significantly to anthropogenic warming . Solar-driven water splitting (aka electrolysis) produces hydrogen, an alternative energy carrier free of greenhouse gases, sustainably and without the limitations of wind power and biomass . The main obstacle preventing large-scale implementation are currently the comparatively high production costs. However, logistics for fossil fuels can also be expensive and environmentally hazardous, in particular for the year-round energy supply of remote research bases such as the Neumayer Station in Antarctica or Paranal observatory in Chile. Therefore, local hydrogen production can become both economically and environmentally favourable, but is challenging in the often cold environments . The currently most mature approach for solar-driven hydrogen production is to supply polymer electrolyte membrane electrolysers with electricity from photovoltaic (PV) solar cells via the grid [2, . The complete separation of light absorption and electrolysis does, however, come with electrical and thermal losses. Firstly, an additional DC-to-DC converter is required . Secondly, internal thermalisation, i.e. de-excitation of charge-carriers to the band edges under the release of phonons, reduces the extractable energy of excited electronhole pairs. In addition, transmission losses of low-energy photons with energies below the bandgap of the light-absorbing semiconductor limit the current . The heat generated by these two latter loss mechanisms has the potential to benefit catalysis. Therefore, a thermally tightly coupled water-splitting device -with the light absorber immersed into the electrolyte or not -represents a highly attractive concept. Such a design uses the waste heat of the absorbers to decrease the internal device electrical resistance and reduce the requirements for the catalysts, while simultaneously cooling the absorbers and hence boosting their efficiency . Furthermore, a design that allows safe product separation without a degradation-and failure-prone membrane also reduces costs and increases operating life . Recently, the beneficial effects of thermally coupled water splitting at ambient temperatures were demonstrated . Solar water-splitting research in general has just recently started to consider the influence of colder ambient temperatures on device operation in the temperate zones . So far, however, the impact of very low temperatures over extended times as found in high-latitudes, high-altitudes or winters in the temperate zones on operation have not been considered. In this work, we investigate a route of expanding the thermal window that makes solar water splitting feasible down to an outdoor temperature of -20 • C. Climate data analysis shows that many world regions at high latitudes or altitudes could benefit from our considerations, prominent examples are Antarctica or the Himalaya region. We use a numerical device model to explore the influence of such low temperatures on the solar-to-hydrogen (STH) efficiency of solar water splitting devices and quantify the beneficial effects of thermal coupling and a suitable device insulation. These predictions are scrutinised under idealised laboratory conditions and the first solar water splitting device operating at -20 • C is demonstrated. We can also show that these benefits differ qualitatively for dual-and triple-junction solar absorbers, which has significant implications for the general design of outdoor-located photoelectrochemical devices. Finally, we discuss the energy supply of high-latitude and highaltitude remote research stations as a first potentially economic competitive implementation of our considerations. Current large-scale technologies for water splitting operate at temperatures between 50 and 1000 • C . Meanwhile, laboratory studies for solar water splitting typically employ ambient temperatures of about 20 • C. For small-to medium-scale, distributed hydrogen production, however, the impact of outdoor temperatures on device operation must be considered. While the volumetric density of dissipated heat of electrolysers in the MW-range are often large enough to require cooling, this changes for smaller plants that are more effectively cooled by outdoor temperatures. In the limit of small-scale applications, such as the powering of weather stations, process temperatures will, without external heating, very closely follow ambient conditions. Yet the mean annual temperatures of a considerable part of the world is below the freezing point of water (Fig. 1a). Low electrolyte temperatures lead to losses from higher catalysis and ion transport overpotentials, but also cause issues for (near-)neutral electrolytes, frequently used for solar water splitting , that do not depress the freezing point of water sufficiently. Hydrogen production would then cease and the volumetric expansion of freezing water can damage the reactor. The energy harvesting potential for conventional solar hydrogen production can be evaluated from Fig. 1c, in which we show the annual cumulative available solar energy for days with mean temperatures above the freezing point of water. In colder regions, however, temperatures remain in a temperature envelope between -20 • C and 0 • C for a considerable fraction of the year (Fig. 1b). This would create the need for energy-intensive temperature stabilisation of the device. A distributed energy system that can operate with a maximum degree of autonomy would, however, be even more important in these regions, since they are typically more sparsely populated and fuel supply is associated with great expense and effort. Since the long-term storage of hydrogen in gas bottles at very low temperatures is not a challenge, it is a predestined energy carrier for these extreme climate conditions. Extending the ambient operating conditions for efficient, small-scale solar water splitting to the above-defined temperature envelope is feasible through the use of electrolytes with a lowfreezing point (e.g. 30 wt% H 2 SO 4 ). Efficiency losses due to the low operating temperatures can be compensated by tight thermal coupling and device insulation as discussed below in detail. This would facilitate additional energy harvesting by solar hydrogen production in some regions considerably as depicted in Fig. 1d. Based on our considerations, solar hydrogen production could benefit in parts of China, Mongolia, the Himalayas, Russia, the Alpine region, Greenland, the Andean mountains, USA, Canada, and Antarctica. More than half of the world population is currently living in areas, where temperatures are in this envelope for at least 30 days per year. The influence of low temperatures on the solar-to-hydrogen conversion efficiency of a solar water-splitting device, as sketched in Fig. 2a, is characterised by two contrary effects. Firstly, there is the lower catalytic performance and higher ohmic losses of the electrochemical component, and secondly, the increased solar-cell efficiency, as indicated in Fig. 2b. Which one prevails depends on a number of device parameters, such as the ohmic cell resistance and the temperature coefficient of the solar-cell open-circuit voltage (V OC ). To understand these effects in detail and explore thermal coupling to compensate possible efficiency losses, we developed an open-source Python-based model combining solar-cell parameters, electrochemistry, and thermal fluxes. In short, the model predicts the STH efficiency based on the temperature-dependent current-voltage (IV)-characteristics of the solar cell and catalysts by computing the operating temperature in an iterative, self-consistent cycle for a quasi-steady-state condition. This means that the absorbed luminous power equals the sum of the power used to split water at thermoneutral conditions plus the power dissipated by radiation and convection (see Fig. 2a). The following calculations are based on a device consisting of high-efficiency double-or triple-junction III-V solar cells together with Pt-and IrO x -catalysts for hydrogen/oxygen evolution reaction (HER/OER), respectively. 30 wt% H 2 SO 4 with a freezing point of −35 • C was used in the model as the electrolyte. Note that one important parameter determining the device temperature is the area ratio A housing /A solar-cell , influencing the heat dissipation from the housing by radiation and convection (see SI Fig. 1). This ratio was fixed to a practical value of 2.5 in the following considerations, representing the case of small-to medium-scale applications. A description of the full model and a list of all input parameters can be found in the supplementary information. Note that a low-cost, but less-efficient alkaline device design could in principle also be realised using Si solar cells, NiFeO x (OER catalyst), NiMo (HER catalyst), and 18 wt% NaOH as an electrolyte. Similar to III-V solar cells for space applications , costs of the absorber are, however, probably not a major issue for niche-applications in remote world regions, where the approach competes with the supply of conventional fuels that is associated with great expense and effort. As soon as the technology is established, costs can benefit from scale-up as well as from emerging low-cost, high-efficiency approaches that would extend the commercial use case beyond just niche. The same applies to the catalysts. For initial applications, the A catalyst /A solar-cells -ratio by YaSoFo as a function of the A catalyst /A solar cells -ratio, for the decoupled, thermally coupled, and coupled as well as insulated case using a double junction (c-e) and a triple junction (f-h) solar cell. d,g, Increase in absolute STH efficiency caused by thermal coupling in comparison to the non-coupled device. e,h, Efficiency gain from insulation of the electrochemical compartment referring to the coupled configuration. could be adjusted to ensure operating potential below the the maximum power point (MPP) for a high-efficiency triple-junction solar cell. However, for the long-term goal of producing hydrogen on a terrawatt (TW) scale costs -but even more so materials availability -will play an important role. There, the use of low-performance, abundant catalysts (or low noble-catalyst loadings) is highly preferable and any additional voltage losses should be avoided as discussed below in detail. Therefore, the parameter ranges of A catalyst /A solar-cells in the model were chosen to cover operating potentials close to the MPP. Figs. 2c,f show the outdoor temperature-dependent STH efficiencies as a function of the A catalyst /A solar-cells -ratio for a thermally decoupled device based on a double-junction (V OC = 1.88 V) and triple-junction (V OC = 2.7 V) solar cell, respectively. For the double junction, the STH efficiency initially increases, but then decreases with decreasing outdoor temperature implying that the device only benefits from improved solar-cell performance for moderate temperature decreases. Consequently, considering only a narrow temperature window could lead to the unsubstantiated generalisation of a steadily increasing or decreasing efficiency with dropping temperature. Note that this effect is very sensitive to the V OC , its temperature coefficient, and the ohmic resistance of the cell. A more negative V OC temperature coefficient and a lower V OC , for example, result in less or almost no decrease of the efficiency at very low temperature, while a higher ohmic cell resistance increases this effect (and vice versa, see SI Figs. 3 & 4). The modelling of the triple junction reveals a constant STH efficiency decrease with decreasing outdoor temperature, an effect that is more stable with respect to parameter variation. In this case, the lower performance of the electrochemical compartment always prevails. This can explained by the higher ohmic potential loss caused by the higher catalyst current densities in the triple junction device. The absolute efficiency gain for a thermally coupled device design is illustrated in Figs. 2d,g. As expected, thermal coupling increases the absolute STH efficiency by up to 6% at low outdoor temperatures for configurations where the operating point of the device would otherwise drop below the MPP. To further increase the efficiency, we explored the influence of thermally insulating the device housing. In the model, aluminium foil was employed to reduce radiation losses and polystyrene was used to minimise further heat dissipation. The results are shown in Figs. 2e,h as an absolute efficiency change referring to the thermally coupled design. For the double junction, the insulation has a detrimental effect on parameter regions in which the efficiency increases with decreasing outdoor temperature, while it has a positive effect in regions where the efficiency decreases with decreasing outdoor temperature (also see SI Fig. 4). For the triple junction, where the efficiency steadily decreases with decreasing outdoor temperature, the insulation is always beneficial. SI figure 3 shows, in a similar manner as Fig. 2c-h, the functional dependence with respect to the geometry-corrected distance, which is a measure for the electrode spacing, taking into account the device geometry. Again, we observe a qualitative difference for double-and triple junctions. This emphasises that these parameters span a multi-dimensional space and therefore must be carefully considered when designing a double-junction device intended to operate at low temperatures. Only considering a subset in the parameter space is probably the reason for apparently contradictory observations in the literature . The insulation design of the electrochemical compartments could be further improved with the ongoing development of macroscopic thermal rectifiers that allow heat to transfer preferentially in one direction . Such a rectifier would offer the possibility of creating and maintaining a thermal gradient between the electrochemical compartment and the solar cells, especially for short-term intermittency of the irradiance. This can then increase the efficiency boost for the device. Apart from conventional absorber materials like Si or III-V semiconductors, transition metal oxides are also feasible absorbers for photoelectrochemical water splitting . Charge carriers in many of these oxides form small polarons resulting in a small drift mobility implying that the carrier transport can be enhanced via thermalactivating . This suggests that insulating an oxide-based device might not only increase the catalytic performance and lower the ohmic losses, but would be also beneficial for the photoabsorber. The illumination was turned on at t = 0. e, IV curves of the solar-cell array and the Pt-and IrO x -catalysts in a 2-electrode configuration of the respective configurations in thermal equilibrium, i.e. after the 3h-measurements. To scrutinise our predictions under idealised laboratory conditions, we built a watersplitting device based on commercial triple junction GaInP/GaInAs/Ge solar cells and commercial Pt-and IrO x -catalysts. The electrodes were separated via a wedge for buoyancydriven product separation as can be seen in SI Fig. 5. Note that this membrane-less concept is a relatively novel approach and there are ongoing efforts to investigate the product crossover as a function of the cell geometry . The A catalyst /A solar-cells -ratio was set to 0.34 and the total device resistance (including the multimeter) was 2.2 Ω at 21 as predicted in our model, resulting in the increase of the operating current and a decrease of the operating potential. For the insulated configuration, it can be clearly seen in Fig. 3e that the additional increase of the catalyst performance (∆V=0.18 V at 0.179 A) in comparison to the coupled configuration is not offset by the additional loss of V OC (∆V OC =0.08 V) caused by the higher device temperature. The energy supply for research stations in high-latitude regions such as Antarctica represents an ideal test application for our considerations. There are ongoing efforts to shift the power supply away from the use of fossil fuels towards renewable energy systems , also for reasons of contamination due to spillage events. Here, hydrogen was indeed already proposed as a future energy carrier , and initial practical experience with an indoor, wind-powered electrolyser was gained . While the overall impressions and results were positive, the complexity of the system caused some technical issues and hence relatively high maintenance efforts . Here, a device that operates outdoors with a maximum degree of autonomy could be highly advantageous and represent the first economically competitive case for thermally coupled solar water splitting. Moderate light concentration could, in principle, reduce the costs for hydrogen production. However, this depends on the location. In near-polar regions, diffuse irradiation that cannot be concentrated can prevail the direct radiation in the solar spectrum . Beyond Antarctica, many research stations across the globe are situated in other remote locations at high-latitude and/or high-altitude (Fig. 1a). In almost all of the 100 stations, we considered, the mean annual temperature, T avg,y , remains below the freezing point of water for large parts of the year (Fig. 4). The developments we report here will allow to expand the thermal window that makes solar hydrogen production feasible for most stations, except for in the Antarctic interior. A realistic hydrogen fuel-cell efficiency is 65% (lower heating value) . Then, powering a Raspberry Pi computing device ( For the long-term goal of producing hydrogen on a TW scale, where materials abundance for both light absorber and catlaysts will play an important role as discussed above, the increase of performance by thermal coupling and insulation offers benefits in three major areas: Firstly, it can increase the solar-to-hydrogen efficiency by shifting the operating point of a given device to higher current densities or using solar cells with lower bandgaps . Furthermore, the amount of catalyst loading can be decreased or less active, yet more abundant catalyst materials become feasible. From Fig. 3e, it can be estimated that 21% and 35% of the catalyst loading or activity for the coupled and coupled/insulated case, respectively, could be saved in our test device to achieve the same operation current as in the decoupled configuration. Therefore, thermal coupling also increases the efficiency with respect to the use of catalysts in the device. Note that these numbers depend on the ohmic cell resistance, device configuration, the V OC , and its temperature coefficient as discussed above. Finally, the reduced overpotentials from catalysis and ion transport offer the opportunity to use emerging solar cell material configurations such as III-V/Si tandem configurations, where the challenges of internal interfaces reduce the effective photovoltage . These considerations are highly relevant for the design and commercialisation of highly efficient thermally coupled solar water splitting, which could use III-V/Si or Perovskite/Si 5 and Ref. . With the expansion of the thermal window down to a threshold temperature of T 0 = 253 K (blue shading), efficient, distributed solar water splitting is feasible for most research stations with low module areas. In contrast, for a threshold temperature of T 0 = 273 K, (green shading) only some high-altitude and Arctic sites could benefit. tandem cells in a wired design or with the absorber fully immersed in the electrolyte. Some scientific and engineering challenges remain. For example, overall efficiency would benefit from a device operating at high pressures to eliminate the need for hydrogen compressors. Gaseous products need to be safely separated. The device needs to be demonstrably stable for years, not hours. Nevertheless, we lay the foundation for a thermally tightly coupled water-splitting device that can produce hydrogen under extreme climatic conditions with a maximum degree of autonomy. Our simple concept of insulating photoelectrochemical reactors with low-cost materials can boost the efficiency -both in terms of production rate and catalyst loading -of devices suffering from ohmic or catalysis losses at low temperatures. We clearly demonstrate that the approach of highly integrated solar water splitting, where the absorber is immersed in the electrolyte , benefits efficiency, though it is technologically challenging. A hybrid approach is the one used in this work, where the solar cells are not immersed, but thermally coupled, providing technological maturity at the expense of thermal coupling efficiency. Our work offers a pathway for transition to a fossil-fuel-free energy supply in high-latitude and high-altitude areas, and also opens opportunities for decentralised hydrogen production in world regions with less extreme, but still cold outdoor temperatures. This can become a considerable contribution towards the decarbonisation of the energy sector at the global scale. ## Author Contributions The project was developed by MM and KR, the experimental design by MK and MM. Code for the photoelectrochemical model was implemented by MK and MM, the experiments conducted by MK, the climate data analysis done by KR. The authors jointly wrote the manuscript. ## Conflicts of interest Electronic Supplementary Information: Efficiency gains for thermally coupled solar hydrogen production in extreme cold Moritz Kölbach, 1, 2 Kira Rehfeld, 3 and Matthias M. May 2 1 Institute for Solar Fuels, Helmholtz-Zentrum Berlin, Germany We used the latest version of the ERA5 reanalysis . To assess the energy harvesting potential for solar-water splitting, we require high-resolution information on temperature and insolation. Solar radiation output in ERA5 has been favourably evaluated globally and at high latitudes . Surface temperature characteristics at high latitudes are also well captured . Climate and solar input were analysed based on global fields of surface temperature (variable 'tas') and surface downward shortwave radiation (variable 'ssrd') at daily/less than 31 km resolution. A day (defined as 24-hour period) is considered suitable for solar-water-splitting with a device if its average (24-hour) temperature is within the temperature envelope of the device. The temperature thresholds we consider are (a) above the freezing point of water (273 K) for conventional electrolysis, (b) above 253 K, the minimal temperature we have evaluated our device for, and (c) above the freezing point of 30 wt% sulfuric acid, 238 K, which can be seen as the lowest temperature limit of the device. The added solar water-splitting potential in Jm −2 , or the energy that we estimate could have been harvested in 2019 under the expected limitations of the proposed device, is calculated as a weighted sum Q t = 365 i=1 p(t) i q i , with the weight for day, i, set to unity if its mean temperature is above the temperature threshold T 0 : p i = 1 if t i ≥ T 0 and to zero if it is not, The harvesting potential of storable renewable energy for high-latitude and highaltitude research stations We assess the opportunity for local production of renewable, storable energy at 100 high-latitude and high-altitude research stations and field camps. Location details for active stations were compiled based on publicly available information and publications . Supplementary Table 5 provides coordinates, names, operators, and links as well as the estimated solar water-splitting energy harvesting potential for the year 2019. Temperature and radiation for the station locations was extracted from ERA5 reanalysis output using bilinear interpolation. Comparing the theoretical efficiency of a triple-junction solar cell SI 2 for AM1.5G with the efficiency for a spectrum to be expected in Antarctica , we find the limit to drop by 6% relative. Therefore, we estimate a relative error for the solar hydrogen harvesting potential in Fig. 4 as 10%. ## Device model description Thermal flux for a thermally coupled device In our model for the thermally coupled device design adapted from Min et al. , we assume that the temperature of the electrolyte is equal to the operating temperature of the solar cells, denoted as the device temperature (T device ). Ignoring heat transfer by educt and product flux, the device reaches its quasi-steady temperature when the absorbed luminous power (P in ) equals the sum of the fraction of the luminous power used to split water at thermoneutral conditions (f •P in ) plus the power dissipated by radiation (q rad ) and convection (q conv ): The absorbed luminous power can be described using the following equation, where T optical-train is the transmissivity of the optical train, A solar-cells is the area of the solar cells, and I 0 is the power density of the incident radiation. The fraction f of the absorbed luminous power that is used to split water at thermoneutral conditions can be expressed as where j op is the temperature-dependent operation current density, C is the light concentration factor, and E th is thermoneutral voltage for water splitting of 1.48 V. Note that the influence of the temperature on E th is very small and is neglected here. The power dissipated through radiation can be described as follows: Here, σ is the Stephan-Boltzmann constant, ε solarcells is the surface emissivity, A housing is the area of the device housing, ε housing is the surface emissivity of the housing, and T out is the outdoor temperature. The power dissipated through convection can be expressed as where h solar cells -air is the convective transfer coefficient, and U t is the overall heat transfer coefficient. The latter can be used to implement effects of insulation with a thickness of l ins and a thermal conductivity k ins around the device housing and is given by: Thermal flux for a decoupled device For the decoupled device design, we estimate the operating temperature of the solar cell by assuming that an area of twice the solar cell area is available for heat dissipation and that the electrolyte temperature is equal to the outdoor temperature. This implies that the sun and the operating potential E op higher than the thermoneutral voltage do not heat up the electrolyte. The fraction f decoupled of the absorbed luminous power that is used to split water can then be expressed as: ## Electrochemistry The following part describes the temperature-dependent IV-characteristics for a 2-electrode water splitting setup based on the single OER and HER catalyst characteristics in an aqueous electrolyte. The temperature-dependent standard potential of the OER is given by: Here, E 0,OER, T ref is the standard potential at the reference temperature T ref , and E 0,OER, T coeff is the temperature coefficient. By definition, the standard potential of the HER is zero: The Nernst-potentials of the OER and HER can be expressed as follows: where z is the number of electrons involved in the reaction, F is the Faraday constant, R is the universal gas constant, and pH is the pH value of the aqueous electrolyte. The thermodynamic potential of the water splitting reaction is then defined as the difference between the Nernst-potential of the HER and OER: The OER current as a function of the applied potential E can be modelled using the anodic branch of the Butler-Volmer equation: where α a • z is the anodic charge transfer coefficient multiplied with the number of electrons involved that can be extracted from the tafel slope, and j 0,OER is the temperaturedependent exchange current density, which can be calculated from the exchange current density j 0,OER, T ref at the reference temperature T ref and the activation energy E a,OER as follows: Note that mass transport limitations are neglected in this model for simplicity. The HER current as a function of the applied potential can be modelled analogously: In the model, the OER and HER current are merged to the (ohmic loss-free) 2-electrode overall water splitting current j catalysts (E, T device ). Subsequently, the ohmic loss is calculated from SI 5 the distance of the electrodes d, the geometry correction factor G corr and the temperaturedependent conductivity of the aqueous electrolyte κ electrolyte as follows: The linear correlation of the electrolyte conductivity with temperature is given by: Solar cell 57 Our model can perform the calculation of the temperature-dependent solar cell IVcharacteristics using the detailed balance limit or experimental parameters from the manufacturer's datasheet . Here we used the latter. The temperature and intensity dependence of the short-circuit current density I sc , the open circuit potential E oc can be approximated as: where influence of shunt and series resistances, the temperature-dependent IV-characteristics of the solar cell can be approximated by : STH efficiency ## 58 To obtain the operating current density j op , the model matches the IV-characteristics of the solar cell and the 2-electrode catalyst current for overall water splitting based on the SI 6 A catalyst / A solar cells -ratio: Since the device temperature depends on the operating current density (and vice versa, see equation 3), our model iteratively solves equations 1 to 22 until convergence is reached. Finally, the STH efficiency is calculated as follows, assuming a Faradaic efficiency of unity: If not otherwise stated, the input parameters listed in SI Table 1,2,3 and 4 are used. Device design 15 GaInP/GaInAs/Ge solar cells (CPV 3C44, Azur Space) with an photoactive size of 10 x 10 mm 2 were back-contacted on a 55 x 40 mm 2 polished copper sheet with a thickness of 0.35 mm using electrically conductive epoxy adhesive (Elecolit 323, Panacol) with the help of a laser-cut placing mask. The front contacts were connected in parallel by wire bonding. The copper plate was attached to a macro-cuvette (402.000-OG, Hellma Analytics) with a thin thermally conductive double-sided adhesive foil (WLFT 404, Fischer Elektronik) to ensure good thermal coupling. The outer dimensions of the cuvette were 55 x 40 x 23.6 mm 3 with a wall thickness of 2 mm. A 5 x 55 mm 2 black polyvinyl sheet was used to mask the busbar for the front-contact resulting in a total illuminated area of 19.25 cm 2 . The positive and negative terminal of the solar cell array was wired to the OER-catalyst (IrO x on a titanium mesh with 12 g Ir/m 2 , Metakem) and HER-catalyst (platinised titanium mesh with 50 g Pt/m 2 , Metakem), respectively. Both catalysts were 3.0 x 1.0 cm 2 , which translates into a surface area of 5.1 cm 2 when accounting for the surface factor of 1.7 of the mesh. The catalysts were positioned parallel to each other separated by 1 cm. A PTFE wedge in parallel to the catalysts at a separation of 0.75 cm to the cuvette bottom was used for product separation. The distance between the lower edge of the catalysts and the cuvette bottom was 1.9 cm. The cuvette was filled with 22 ml of 30 wt% H 2 SO 4 (pure, Carl Roth) as an electrolyte. For experiments with insulation, the cell was covered in aluminium foil (emissivity of 0.2) and inserted into a 2 cm thick foamed polystyrene housing (thermal conductivity of 0.03 Wm −1 K −1 ). Photographs of the cell are shown in SI Figure 5. ## Device characterisation To perform experiments at low temperatures, the device was put into a freezer (LGUex1500, Liebherr) and illuminated through an 8 x 8 cm 2 hole in the door that was covered with a 0.5 cm thick quartz plate. As an illumination source, a WACOM Class AAA solar simulator (WXS-100S-L2H, AM1.5G, 1000 W/cm 2 ) was used. The overall power, as measured within the fridge, was set to 1000 W/m 2 using a S401C powermeter (Thorlabs). Such a single-point calibration for a multi-junction cell is, in principle, prone to errors due to the spectral mismatch between solar simulator and the AM1.5G standard spectrum. In our case of an almost perfectly current-matched triple junction, this does, however, not lead to an over-but rather an underestimation of the overall efficiency, as follows. Optimising the power ratio of the xenon and halogen lamps while keeping the total power at 1000 W/m 2 , resulted in a current density of 13.15 instead of the 15.4 as reported in the datasheet. Raising the current to the photocurrent from the datasheet of a reference cell without correcting the total power by an infrared filter, as is sometimes done in the literature , might artificially increase the overall efficiency due to higher thermal power available for heating the electrolyte, and is therefore avoided here. The quartz plate (without anti-reflection coating) as our optical train was inserted into the light path after light-power calibration. For the efficiency calculation, the total illuminated area of 19.25 cm 2 instead of the 15 cm 2 photoactive area was used. of the solar cell array were recorded using a Keysight 2400 source meter with a scan rate of 100 mV/s. For reference measurements in the thermally non-coupled device configuration, the solar cell array on the empty cuvette was illuminated, while the electrochemical compartment was placed in a second cuvette in the freezer located out of the direct beam path. ## G corr Geometry correction factor: ratio of the resistance of a restricted cell with no obstacle in between the electrodes (e.g. wedge for product separation) and the real cell resistance during operation 0.3 -

chemsum

{"title": "Efficiency gains for thermally coupled solar hydrogen production in extreme cold", "journal": "ChemRxiv"}

molecular_characterization_of_sequence-driven_peptide_glycation

## Abstract: Peptide glycation is an important, yet poorly understood reaction not only found in food but also in biological systems. The enormous heterogeneity of peptides and the complexity of glycation reactions impeded large-scale analysis of peptide derived glycation products and to understand both the contributing factors and how this affects the biological activity of peptides. Analyzing time-resolved Amadori product formation, we here explored site-specific glycation for 264 peptides. Intensity profiling together with in-depth computational sequence deconvolution resolved differences in peptide glycation based on microheterogeneity and revealed particularly reactive peptide collectives. These peptides feature potentially important sequence patterns that appear in several established bio-and sensory-active peptides from independent sources, which suggests that our approach serves system-wide applicability. We generated a pattern peptide map and propose that in peptide glycation the herein identified molecular checkpoints can be used as indication of sequence reactivity. Glycation presents a ubiquitous non-enzymatic post-translational modification 1,2 , which is formed by the reaction of amino compounds and reducing sugars. It refers to a complex reaction network and produces a multitude of heterogeneous reaction products, also known as Maillard reaction products (MRPs) or advanced glycation end products (AGEs) 3,4 . Glycation is a multifactorial reaction, which depends on the nature of the precursors and the reaction conditions, including time and concentration . The Maillard reaction (MR) is one of the most common and essential reactions in food processing and determines color, flavor and taste of food. Further, its reaction products are known to affect human health 9,10 , and contribute to various pathologies, such as diabetes 11,12 . Here, glycation leads to molecular and cellular changes in a series of complicated events. Hyperglycemia drives glycation of lipids and proteins and development of vascular lesions by AGE engagement of the receptor for AGE (RAGE) in cells of the vessel wall . Due to their broad relevance, thorough understanding of glycation reactions is indispensable. As insights into the MR of amino acids continue to emerge, new models are needed to improve the understanding of peptide and protein glycation 17 . Many previous studies analyzing the health effects of glycation products and peptides point to their miscellaneous bioactivities and their potential as nutraceuticals and functional food ingredients . Glycation induced alterations in the bioactivity and improvement of sensory attributes have been described for various peptide mixtures . However, the specific peptides related to these changes remain largely uncharacterized and, even more important, the behavior of peptides in glycation reactions has barely been systematically analyzed and thus is unaccounted. Only a limited number of studies on peptide reactivity in the MR have been conducted and focused on synthetic peptides 26,33,34 or peptide derived MRPs in specific foods . These approaches have revealed the relevance of both peptide length and composition in the MR and the importance of peptide glycation in various fields, not only including diverse food matrices but also biological systems and disease progression 10 . The information describing the general determinant factors of peptide glycation, however, remains elusive. Therefore, particular sequences and, thus classes of proteins that have preference to undergo glycation reactions and the final consequences, such as loss and gain of bioactivities, cannot be determined. Especially required are model systems for large-scale MRP characterization that enable general understanding of the influence of the amino acid composition, sequence and peptide length on glycation product formation. ## OPEN 1 Chair of Analytical Food Chemistry, Technical University Munich, Maximus-von-Imhof-Forum 2, 85354 Freising, Germany. 2 Research Unit Analytical BioGeoChemistry (BGC), Helmholtz Zentrum München, Ingolstädter Landstrasse 1, 85764 Neuherberg, Germany. 3 The Waltham Pet Science Institute, Mars Petcare UK, Waltham-on-the-Wolds, Leicestershire LE14 4RT, UK. * email: michelle.berger@tum.de; schmitt-kopplin@ helmholtz-muenchen. de To acquire a better understanding of site-specific peptide glycation the analysis of the Amadori product (AP), a relatively stable intermediate of the MR, and consecutive downstream reaction products MRPs is particularly well-suited. High-resolution mass spectrometry (MS) is a fast and highly sensitive method, which enables detection and identification of both early and advanced products of the MR 39,40 . Information on net chemical transformations and precursor reactivity in such systems can be gained by non-targeted experiments and generation of mass difference networks 41,42 . Collision-induced dissociation (CID) after electrospray ionization (ESI) MS has been successfully applied for peptide derived AP analysis . However, non-targeted large-scale analysis and interpretation of peptide derived MRPs is limited. Only a few studies have gained insight into peptide reactivity and the influence of sequence microheterogeneity 33,34, . This includes glycation based on accessibility of the N-terminus and catalytic effects in some short-chain peptides. Here, we report that the combination of highresolution ESI quadrupole time of flight (QTOF) MS, bioinformatics and multivariate statistics enables a deep and molecular-level investigation of complex peptide systems. Using this combinatorial method for large-scale AP analysis, we characterized the reaction behavior of 264 casein-derived peptides in the MR and used this data to gain insight into sequence-dependent differences in AP formation profiles and, thus, peptide reactivity. Furthermore, we discovered potentially relevant glycation-patterns and demonstrate system-wide applicability of this study to various food-peptide sources by in silico sequence mapping. Database search serves as a reference for investigation of bioactive and sensory-active peptide reaction behavior. This approach may be amendable to practically any type of glycation system, and it allows exploration at various levels of information, from the influence of the peptide composition to the role of specific sequence-patterns in peptide glycation. ## Results A time-resolved analysis of peptide dependent Amadori product formation. To study the reaction behavior of peptides in glycation, we heated complex model systems containing glucose (2.7-54 mg/mL) and tryptone at 95 °C for 2, 4, 6 and 10 h, respectively. Compared to an in silico tryptic protein digest, tryptone provides approximately four times more free peptides with increased diversity (Fig. 1a). A vast number of non-enzymatic cleavage sites generates many diverse peptides (Supplementary Fig. 1) with partially overlapping amino acid sequences. This enables characterization of site-specific microheterogeneity and, ultimately, identification of specific sequence patterns that promote glycation. Apart from that, C-and N-terminal amino acids in tryptone peptides comprise a much greater diversity than enzymatic digests can cause, which becomes apparent by comparison with an in silico tryptic casein digest (Fig. 1b and Supplementary Fig. 2a). Unlike tryptone, there is a bias toward lysine-(Lys) and arginine-(Arg) containing peptides for tryptic casein digestion (Supplementary Fig. 2b) and poor enzymatic cleavage of certain protein regions (Supplementary Fig. 3) may lead to the preclusion of particularly relevant peptides. Even alternative proteases or sequential digestion would only provide a marginally increased total number and diversity of peptides compared to trypsin 49 . With the applied method, we could nearly completely cover the casein protein sequences (Fig. 1c). With this extensive dataset in hand we first explored the concentration-and time-dependent reaction behavior of this diverse pool of peptides based on the formation of the corresponding APs. The AP is a relatively stable intermediate of the early MR 50 , which is formed via condensation between the amino compound and the reducing sugar, and subsequent rearrangement 11 . The number of hexose residues coupled to an amino acid or peptide was estimated by a mass increase of 162.0528 Da per attached monosaccharide (C 6 H 12 O 6 -H 2 O). Tandem MS was applied to obtain structural information, which allowed confirmation of the AP chemical structure of 47 amino compounds (Supplementary Table 1). Significant correlations (p-value ≤ 0.05) were observed across normalized AP intensity profiles, and APs clustered by the influence of sugar concentration but also reaction time (Fig. 2a). Interestingly, APs clustered that were formed from peptides with comparable amino acid sequences. Sequence similarity is highlighted by the suspended numbers indicating sequence groups, e.g. HLPLP and LHLPLP (Cluster 2, sequence group 16). Certain APs found in Cluster 1 and 2 seemed to form isomers, which caused them to also appear in Cluster 3. In Fig. 2b, the representative normalized mean intensity profiles for the three clusters are shown, and Supplementary Fig. 4 provides the individual normalized intensity curves for each AP, separately. The intensities of APs in Cluster 1 and 2 reached their maximum after two hours for all glucose concentrations and decreased with further reaction time (Fig. 2b). In contrast, Cluster 3 contains APs, which increased with time and reached maximum intensity after either six or ten hours of incubation. Further, enrichment of AP levels at different glucose concentrations was observed. The highest intensity of Cluster 1-APs was either detected at 5.4 or 27 mg/mL of glucose, whereas for APs in Cluster 2 (27-54 mg/mL) and 3 (54 mg/mL) higher sugar concentrations were required to reach their maximum. Figure 2c compares AP formation between different peptide lengths, demonstrating that nearly all amino acids and dipeptides were found in Cluster 3, whereas larger peptides www.nature.com/scientificreports/ did not show uniform normalized AP intensity profiles. Moreover, the percentage observed as an AP for each peptide length is displayed, but no general correlation between peptide reactivity and sequence length could be observed. Note, this calculation is based on the total number of peptides per length, so one length may appear to be glycated to a greater extent if a minor number of peptides was identified with this length. A higher proportion of APs derived from dipeptides compared to tripeptides resembles observations from previous studies 33,43 , which suggested decreasing reactivity with increasing peptide length. Here, we confirm and extend these observations by investigating a larger range of peptide chain lengths. Interestingly, longer-chain peptide sequences were not generally associated with reduced reactivity in early glycation reactions, e.g. when comparing penta-and hexapeptides or nona-and decapeptides. Even with these observations, it is difficult to comment on the influence of peptide length on glycation on a universal level. All APs significantly increased after 2 h (t-test, p-value ≤ 0.05); however, different glucose levels were required (Supplementary Table 1). Most (25) of the identified APs significantly increased at a glucose concentration of 2.7 mg/mL, while other peptides required higher glucose levels. Interestingly, different observations were made for highly similar peptides. For example, VPQLEIVPN required a glucose concentration of 27 mg/mL to increase, significantly, whereas for KVPQLEIVPN only 5.4 mg/mL of glucose were needed. Analogous behavior was observed for the APs of VAPFPE (27 mg/mL) and VAPFPEV (5.4 mg/mL). Shedding light onto the role of peptide composition in glycation. AP analysis uniquely facilitates characterization of site-specific glycation, and our dataset provides insight into the highly complex and yet largely unknown reaction behavior of peptides in glycation. Previous studies explored peptide derived MRPs in particular foods providing limited information from a global prospect . Others have investigated the reactivity of highly specific synthetic peptides depending on factors explored herein to some degree, such as peptide length (discussed above) and amino acid composition 33,34 . Here, we aspired to approach these research questions from a general level using a large reservoir of peptides and APs. Mapping glycated peptides onto casein proteins revealed that AP formation was observed for only 14% of α-S2-and 9% of κ-casein peptides (Supplementary Fig. 5), but APs related to α-S1-casein (29%) and ß-casein (45%) were detected to a larger extent (Fig. 3a). Analyzing the amino acid sequence of α-S2-and κ-casein peptide APs, we identified protein-specific peptides such as FLPYP (F 55 -P 59 of κ-casein) . The ability to profile glycosites at this scale provides opportunities to determine the relative susceptibility of peptide collectives with similar amino acid sequences to the early MR. Especially reactive peptide classes are captured here, as the protein heatmaps show the frequency that sequences co-occurred on glycated peptides . AP formation of peptides sharing certain amino acid sequences appeared to be favored, e.g. including peptides that originated from N 73 -V 92 and V 170 -V 173 of β-casein. To examine the influence of the amino acid composition on peptide reactivity in the early phase of glycation, we calculated the contribution of each amino acid to AP formation, given as a percentage of the total observations in tryptone peptides. Figure 3b shows amino acids, e.g. glutamic acid (Glu) and leucine (Leu), that appeared equally in peptide APs from α-S1-(top) and β-casein (bottom). Other amino acids, such as proline (Pro) and histidine (His) showed considerable variations in their contribution to AP formation (Supplementary Fig. 6) depending on the source protein, meaning the overall peptide sequence, which fits with the known role of the microenvironment of amino acids in glycation. Importantly, a previous report described varying in vivo reactivity of lysine depending on its position in the albumin sequence and, thus, its neighboring amino acids 46 . Further, investigation of short-chain peptide model systems showed that AP formation is considerably influenced by the immediate chemical environment, hence, adjacent amino acids side chains in the peptide sequence 26,33 , which overall indicates that we may also have identified reactivity-sequence interrelation for peptide structures. We visualized the median percentage of each amino acid in APs to explore the effect of amino acid composition and microheterogeneity over all four casein proteins (Fig. 3c). Most amino acids showed a wide distribution of the values, again demonstrating that the type of amino acids that contribute to AP forming peptides can vary based on their immediate chemical environment. This presents a promising starting point to explore for sequencespecific glycation. To dive into this intriguing facet of peptide glycation, we examined the location of amino acids relative to the reactive peptide N-termini . This was based on De Kok's hypothesis that the side chain carboxylic group of Glu catalyzes glycation of primary amino groups 33 , and on the suggestion of Lhiang Zhili and co-workers that Leu and isoleucine (Ile) promote AP formation 26 . As short-chain peptides were investigated in these studies, they describe the influence of directly neighboring amino acid side chains on N-terminal peptide glycation. Hence, we reasoned that our dataset could allow to explore the impact of both the N-terminal amino acid and the adjacent amino acid side chain across a large number of highly diverse peptide species. To detect amino acid overrepresentation at the mentioned positions, we generated sequence logos by comparing sequences between peptide APs and non-glycated peptides (Fig. 4a and Supplementary Fig. 7a). Here, amino acids enriched at certain positions in AP forming peptides are illustrated (relative abundance glycated − relative abundance non-glycated > 0). This analysis indicated preference for valine (Val), Ile and Leu at the first two positions of the amino acid sequence for glycated peptides (Fig. 4a). Indeed, the percentage of N-terminal Val was considerably increased for AP forming peptides compared to peptides, for which the corresponding AP could not be identified (Fig. 4b, top). An illustration of the absolute amino acid abundance can be found in Supplementary Fig. 7b. Interestingly, we further found substantially higher relative frequencies for Ile, Leu and Val next to the N-terminus in peptides also observed as an AP (Fig. 4b, bottom), echoing the result from the sequence logo. To account for preferred glycation of peptides with Ile, Leu or Val at the second sequence position, their summed relative frequency depending on AP detectability is shown in Fig. 4c. This result complies with previous findings that hypothesized that Ile and Leu promote N-terminal glycation based on pronounced hydrophobicity 51 (Ile 1.80; Leu 1.70) and polarizability 52,53 www.nature.com/scientificreports/ www.nature.com/scientificreports/ to Ile and Leu and was previously shown to exert a similar effect on the reactivity of the peptide N-terminus 26 . Furthermore, the percentage of aspartic acid (Asp), methionine (Met), phenylalanine (Phe) and Pro next to the N-terminus was considerably decreased in glycated peptides (Fig. 4b, top). We also observed that glycation of peptides was disfavored for Pro at the first two sequence positions (Fig. 4b and Supplementary Fig. 7a). A recent study on Pro containing dipeptides (Gly-Pro, Pro-Gly) suggested that its secondary amine may hinder Schiff base formation 54 . Nevertheless, we found that proline was frequently observed at the third and fifth sequence position of glycated peptides (Fig. 4a), thus raising the possibility of its involvement in increased peptide reactivity towards early glycation. This is supported by an increased relative abundance of proline at the same locations relative to the glycation site in AP forming peptides compared to peptides, for which the corresponding AP was not detected (Supplementary Fig. 7c and d). Capturing relevant sequence patterns in peptide glycation. Large-scale peptide derived AP analysis enables us to identify potentially relevant glycation-patterns, and our dataset can provide an initial glimpse into this intriguing aspect of glycation. While others have explored the influence of the amino acid sequence based on di-and tripeptide glucose model systems 26,33 , we can now comment on trends across 264 peptides originating from four proteins. In enzymatic glycosylation the importance of the N-X-S/T sequon and the negative effect of Pro in X position has been shown, which may result from conformational changes 55 . However, relevant structural motifs in peptide glycation have not been identified. In this detailed analysis, we identified small regions of identical subsequences in casein proteins and, thus, the thereof arising peptides (length = 2, 3, and 4) using the amino acid one letter code (Fig. 5a, top). We mapped (non-) glycated peptides onto proteins (Fig. 5a, bottom right part) and across each other (Fig. 5a, bottom left part). This provided information on the degree of co-occurrence for sequence patterns on glycated peptides with a different overall amino acid sequence. This protocol allows to detect relevant glycation patterns that are anticipated to be important factors in determining the preference for early peptide glycation. Analysis of common sequence patterns in casein proteins revealed several sequence overlays. Figure 5b provides information on the degree of sequence similarity, from which it is evident that the number of shared sequences varied for different pattern lengths and pairs of proteins. No shared tetra-sequences were found for κ-casein. To explore sequence patterns of maximum length, tri-patterns were chosen for further investigation. Figure 5c captures the total frequency of tri-sequence patterns in the casein protein sequences and which percentage of these differentially located subsequences was covered by peptide APs. This analysis allows to identify how different patterns contribute to glycation of peptides with shared subsequences but different overall amino acid composition as they originate from different casein protein regions. Of the three P-E-V sequons in casein proteins (P 44 -V 46 of α-S1-casein and P 105 -V 107 of ß-casein, Fig. 3a; P 171 -V 173 of κ-casein, Supplementary Fig. 5), two appear as subsequences in peptide APs, as well as the alternated V-E-P (V 131 -P 133 of ß-casein, Fig. 3a). Several interesting cases where substructures highly similar to P-E-V contribute to AP formation are highlighted (P-E-L, V-L-N, V-P-N, and V-P-Q; Fig. 5c). These subsequences all share amino acids with a low dissimilarity score (D), which is based on 134 categories of activity and structure 56 , and are partially rearranged (Supplementary Table 2). For example, P-E-V and P-E-L only differ by a single amino acid with strong physicochemical similarity (D(Val, Leu) = 9), whereas in case of P-E-V and V-P-Q (D(Glu, Gln) = 14) also the sequence order was changed. By comparison, peptides that contain I-V-E, which shows more pronounced sequence variations (D(Pro, Ile) = 24 and sequence rearrangement), do not participate in AP formation. All of these patterns, which show strong contribution to AP formation, either comprise Glu (carboxylic group), glutamine (acid amide group) or asparagine (acid amide group). A catalytic effect of the carboxylic group on AP formation was previously hypothesized 33 , which resembles the here found promoting effect of Glu-containing sequence patterns on the early MR. P-Y-P and P-F-P contain amino acids with comparable properties. The substructures P-I-P, P-L-P and P-V-P feature pronounced physicochemical similarities as well 56 . All these patterns showed a high co-occurrence on peptide APs relative to their total abundance (displayed as percent in Fig. 5c), which indicates their contribution to glycation independent of the overall peptide composition, thus their origin in the source protein . In contrast, P-N-P (P 198 -V 200 of α-S1-casein) did not contribute to an AP (Fig. 3a) and shows pronounced differences in its amino acid characteristics. A peptide-sequence pattern plot in Fig. 5d maps relevant sequence patterns to different (glyco-) peptides for which they could be observed. These peptides vary in their overall composition and peptide length. The map indicates, which patterns contribute to glycation on a peptide-level, and other peptide properties that considerably affect their reaction behavior. This analysis reveals several interesting trends, such as pronounced discrepancies in glycation of peptides with the same pattern and, perhaps most striking differences in AP formation of strongly related peptides. Small variations in the peptide amino acid sequence may cause (VVPPFLQPEV vs. VVVPPFLQPEV; YPFPGPI vs. VYPFPGPI) or not cause (VAPFPE vs. VAPFPEV; VYPFPGPI vs. VYPFPGPIN) differentiated behavior in the early MR. General correlation was not observed between AP formation and peptide physicochemical properties, expressed as hydrophobicity according to their retention time. System-wide analysis of bioactive and sensory-active peptide glycation enabled by in silico sequence mapping. The complexity of glycation represents a great challenge for the identification of glycation patterns that are associated with the gain or loss of bioactivity and glycation induced changes in sensory attributes. A combination of bioinformatics and database search enabled to study the established sensory and bioactivities of peptides in our dataset and to evaluate their behavior in glycation. We matched 60 peptides (Supplementary Table 3) with reported bioactivities (Fig. 6a), which were included in databases compiled from literature 57,58 . While 36 peptides were exclusively found in milk, 24 peptides appeared in a variety of other food sources as well (Fig. 6b). We also found that these peptides of diverse chain lengths (Fig. 6c) cover various bioactivity categories (Supplementary Fig. 8), suggesting that tryptone models may facilitate inter-disciplinary investigation of peptide glycation. In our study, for 62% of the bioactive peptides the corresponding AP was detected (33% confirmed by MS/MS tandem experiment; Fig. 6d). Hence, approximately 34% of the 47 peptides, for which the AP was identified via MS2 fragmentation, were previously reported to be bioactive. Furthermore, we identified 25 peptides with sensory attributes 59 . Figure 6e illustrates the prevalence of AP detection for sensory-active peptides and the level of AP identification (MS and MS2). Approximately 24% of all sensoryactive peptides and 19% of the bitter peptides were observed as the corresponding AP on MS2 level. As noted by Shiyuan Dong and co-workers 32 , bitterness of MRPs was decreased compared to original casein peptides, and further reduced with heating time and glucose concentration. Xiaohong Lan et al. previously published that bitter soybean peptides below 1000 Da decreased 28.49% after reaction with xylose at 120 °C60 . In contrast to the reported experiment, digesting casein with trypsin would produce large peptides, through protein cleavage after arginine and lysine, which would lead to the loss of highly interesting bioactive peptides (Supplementary Fig. 9) and would not allow comprehensive investigation of their reaction behavior (Supplementary Fig. 10 and Supplementary Fig. 11). For example, using a tryptic casein digest, the analysis of the opioid peptides, e.g. YPF-PGPI and YPVEPF, for which N-glycation is known to have major consequences on the bioactivity of the parent www.nature.com/scientificreports/ peptide 61,62 , would not be possible. By comparison, in our experiment the large and diverse peptide spectrum of tryptone enabled us to widely predict bioactive peptide glycation. ## Convergence of tryptone peptides and peptides with established activities into common sequences. Given that similarities in the amino acid sequence and sequence patterns may determine the reaction behavior of peptides, we reasoned our approach could provide insight into peptide reactivity from a systems level. To test this hypothesis, we searched peptides from our dataset as a pattern of bioactive and sensory-active peptide sequences reported in databases by substring matching . First, we examined the num- www.nature.com/scientificreports/ ber of matches observed. In total, tryptone peptides were successfully mapped to 1172 unique bioactive peptide species (Supplementary Fig. 12). Even though caseins are major proteins in milk, a considerable number of common sequences between tryptone peptides and a plethora of peptides from other sources was found (Fig. 7a). Overall, we achieved 3046 sequence overlays with 675 bioactive milk peptides, 1350 overlays with 510 peptides from other sources, and 599 overlays with 174 sensory-active peptides (Supplementary Fig. 13). Importantly, not only small sequence commonalities were found as indicated by the length of the tryptone peptides mapped (Fig. 7a). Tryptone amino acid sequences, up to a length of thirteen amino acids, occurred on bioactive peptides from entirely different origins. Even larger sequences (n = 14) co-occurred in peptides with established sensory activity. The heatmap (Fig. 7a) visualizes classes of peptides delineated by the number of overlays per peptide length, showing increased sequence similarities for sensory-active peptides and fish. Or, in case of potato peptides we found a lower proportion of common di-and tri-sequences compared to other sources and no matches for larger peptides (n > 3) were observed. We note that calculating co-occurring sequences is straightforward and may provide information about glycation susceptibility of specific peptide classes from various protein origins. Supplementary Fig. 14 demonstrates that a broad variety of bioactivities is covered by the database peptides, to which tryptone peptides were successfully mapped. Other trends arise, such as the presence of a relatively high abundance of antihypertensive peptides. Note, this is expected because of the inordinate number of antihypertensive peptides in the databases used. The same number of dipeptides was successfully mapped to bioactive peptides independent of the source (Fig. 7b), and the number of tripeptides found as a subsequence was comparable as well. For larger tryptone peptides, approximately half of the peptides matched for milk were found to overlay with peptides, which were (also) found in other food sources. A striking feature of this analysis are the percentages of peptides in our dataset found as a subsequence (Fig. 7c). Considering the 264 peptides that could be mapped to a domain, 82% existed in bioactive peptides exclusively from milk and 47% in those from (milk and) other sources. Furthermore, 39% of tryptone peptides were successfully mapped to sensory-active peptides (Fig. 7c). Figure 7d depicts the percentage of matched tryptone peptides, for which the corresponding AP was identified. Up to 22% of the shorter substring peptides (n ≤ 6 amino acids) were detected as an AP by tandem MS experiments, while the majority of larger peptides appeared to be glycated. In total, 93% of the peptides detected as an AP are substructures of bioactive peptides. This represents a considerably larger proportion compared to other peptide groups, e.g. sensory-active peptides (Fig. 7d). ## Discussion In all, here we present a straightforward approach to refine evaluation of peptide derived APs by using the power of high-resolution MS in combination with multivariate statistics and bioinformatics to access large-scale information about peptide reactivity in the MR and the influence of both reaction time and sugar concentration. Investigation of glucose-tryptone model systems enabled the most in-depth profiling of peptide APs to date. By comparison with a in silico tryptic casein-digest, we demonstrated considerable advantages of tryptone models, such as a notably larger coverage of (bioactive) peptides from various food sources. This strategy is amenable to virtually any type of MR model system or reactivity study with known reaction intermediates. Finally, the reaction behavior of 264 casein derived peptides was characterized by AP analysis from a single type of model system, which demonstrates that new models must be developed to unravel the glycation reaction network in its full complexity. Clearly, large-scale studies are needed to explore peptide glycation and its importance particularly in food but also biological systems and, thus, health. A caveat of practically any MS-based experiment is that detectability can be affected by the type of ionization, analyte concentration as well as sample complexity. Thus, there may be a bias toward specific peptides and APs to consider in this dataset. Furthermore, Fig. 1c shows a high frequency of tryptone peptides from certain protein regions , which may arise from its production process and above-mentioned detectability issues. Our data interpretation, however, reflects on ubiquitous observations in the overall dataset, and not on specific peptide species . Despite this, we achieve in-depth characterization of a large reservoir of peptides and provide thorough information on peptide properties influencing AP formation. We relied on normalized AP intensity profiles for reaction behavior investigation, meaning there are limitations in the stability of this early reaction intermediate to consider. Greifenhagen et al. has noted pronounced susceptibility of the N-terminally acetylated Amadori peptide Ac-Ala-Lys-Ala-Ser-Ala-Ser-Phe-Leu-NH 2 toward oxidative degradation in aqueous model systems 63 . Loss of the Amadori compound of the endogenous opioid pentapeptide leucine-enkephaline (Tyr-Gly-Gly-Phe-Leu) was also noticed by Jakas and Horvat 64 , however, markedly slower degradation behavior was observed in this study. Different reaction conditions, such as concentration of catalytically active phosphate buffer and temperature, tend to affect AP stability, but also the amino acid sequence of the peptides. Even with this, all peptide APs detected in this dataset remained above the limit of detection at all time points. While other studies have investigated glycation using a limited number of highly specific synthetic peptides, we could simultaneously study the reaction behavior of a large pool of casein peptides. We can also see similar trends to previous studies, such as the influence of both reaction time and sugar concentration 65,66 . In contrast, we can provide detailed information on how APs derived from specific peptide sequences are affected. We show that upon reaction time the bulk of peptides differentiated into three clusters, reaching maximum AP intensity at different time points and, thus that AP formation rates likely depend on the peptide structure. Peptides forming APs that peaked at early reaction time points and low glucose concentrations may represent more reactive precursors in glycation. A consecutive decrease in relative AP peak intensity may be attributed to further rearrangement and oxidative cleave reactions yielding heterogeneous AGEs 67 . Further, we show that there is no general correlation between peptide length and reactivity. More pronounced susceptibility of dipeptides compared to tripeptides toward glycation was seen in early studies of glycine (Gly) peptide model systems (GlyGly > GlyGlyGly) and in more miscellaneous synthetic peptide studies 26,33,34 . Similarly, we found higher proportions of glycated dipeptides www.nature.com/scientificreports/ than tripeptides. Although our observations are in congruence with previous studies, we are the first to investigate the impact of peptide length at this scale providing a new perspective on its influence on reactivity. We also show that there is not a general correlation between amino acid content and susceptibility towards the initial phase of glycation reactions. This suggests a strong contribution of other factors such as the amino acid sequence, thus, the amino acid microenvironment. In peptide glycation, reaction behavior has been proposed to be driven by the amino acid sequence. An important role of the amino acid adjacent to the N-terminus has been suggested based on short-chain peptide model systems 26,33,34 . Here, we noticed strong preference of glycation for valine-starting peptides and noted more pronounced AP formation of peptides, which contain Ile, Leu and Val positioned next to the N-terminal amino acid. Further, we observed prevalence of Met, Phe, and especially Pro at the second sequence position in peptides, for which the corresponding AP could not be identified, which indicates that steric hindrance or conformational changes may prevent N-terminal glycation. Congruent observations were made for the N-X-S/T sequon, where Pro in the X position causes pronounced changes in conformation and, thus prevention of enzymatic glycosylation 55 . We show that neighboring Glu, for example, may not always exert a catalytic effect on N-terminal glycation as a result of its carboxylic side chain. Interestingly, APs from peptides containing Asp adjacent to the N-terminus were not observed, even though its structure is closely related to Glu. Depending on the amino acid sequence and the peptide N-terminus, differentiated effects of the neighboring amino acid may be observed 33 , and here, we can cover a broad range of diverse peptide properties. Such thorough and integrated characterization of peptide APs depending on the reaction conditions is necessary for a complete understanding of peptide glycation and its impact on food and biological systems. We further found two location sites near the N-terminus with increased relative abundance of Pro in AP forming peptides, namely the third and fifth position of the amino acid sequence. We identified peptide collectives particularly prone to early glycation reactions by mapping APs to casein sequences and across each other. Furthermore, we leverage these reactive peptide species to provide information on potentially important tri-sequence patterns and propose that glycation patterns among many other factors promote peptide glycation, which is indicated by strong connectivity in glycation susceptibility and presence of specific sequence patterns. Even though N-terminal proline may inhibit Schiff base formation 54 , here, we established several proline-rich sequence patterns, which considerably triggered AP formation. Furthermore, Glu containing sequence patterns, such as P-E-V, may exert a catalytic effect towards early MR 33 . Depending on whether glycation is desired or not, peptides may be chosen, accordingly (discussed below). Overall, we provide a comprehensive set of molecular checkpoints for peptide reactivity estimation towards glycation. Finally, we established system-wide applicability of tryptone model systems by mapping tryptone peptides to a plethora of bioactive and sensory-active peptides from various food sources. Depending on whether glycation of peptides is desired or not, we suggest that the amino acid sequences may be chosen, accordingly. For development of functional foods, health benefits must be preserved, thus for most bioactive peptides AP formation needs to be obviated. For example, for opioid peptides, the requirement of free N-terminal tyrosine was demonstrated and the loss of antihypertensive properties of casein peptides as a result of glycation was revealed in various model systems 29,30,68 . Conversely, if increased antioxidant potency 26,69 must be achieved, we suggest that the peptide species may be capitalized that are more prone to AP formation. The increased antioxidative properties of MRPs compared to their corresponding casein peptides has previously been determined 27,28,30 , and consequently targeted peptide glycation may enable to enhance the health benefits of peptides. Superior antioxidative properties have been established for MRPs derived from small peptides compared to larger peptide species 69 , which makes tryptone particularly suitable for identification of potential peptide candidates. In total, we observed APs from 47 amino compounds. For 34% of them, bioactivity was previously established and 93% were identified as a substructure of bioactive peptides, which suggests that bioactive peptides are particularly prone to glycation. For sensory-active peptides, others have observed reduced bitterness of MRPs compared to heated casein peptides alone, while antioxidative properties where increased 26 . Similar findings were reported for peptides from other sources 60,70,71 . Even more benefits for food production can thus be provided by choosing appropriate peptide candidates, such as enhanced sensory attributes offoods. To assess desired flavor improvement 71,72 , we reasoned that selection of peptides susceptible to early glycation may be promising. As bitter peptides cannot be employed above their bitter flavor threshold 72 , increased bioactivity accompanied concurrently by decreased bitterness may be desirable for the production of functional foods to improve health and enhance customer acceptance 23 . Taken together, our dataset allows to select suitable peptide candidates, given (1) a checklist for estimation of their reaction behavior in early glycation reactions according to the amino acid at the N-terminus, the adjacent sequence position and presence of relevant sequence patterns, and (2) screening for established sensory attributes and bioactivity. Future studies are required to investigate a wider range of peptides from different proteins and, thus a broader variety of amino acid sequences to gain more global information on the relevance of amino acid composition and sequence patterns. Model systems prepared from highly specific synthetic peptides have provided valuable findings in previous studies, which suggests that targeted investigation of peptides, in particular with potentially relevant sequence patterns, may be promising for identification of peptides especially prone to peptide glycation. These strategies also present an opportunity for determination of peptides less susceptible toward glycation reactions. Investigation of changes in sensory attributes and bioactivity as a result of glycation will be a worthwhile endeavor in future experiments to gain the necessary refined information for systematic use of specific peptides and their glycation products as functional food ingredients. Chemical peptide structures were confirmed by peptide mapping in GeneData Expressionist Refiner MS 13.0 (GeneData GmbH, Basel, Switzerland) with an absolute m/z tolerance of 0.005 and 0.1 for the precursor and product ions, respectively, unspecific enzyme cleavages, a minimum peptide length of 1 AA, and no fixed or variable modifications. The fragmentation type was set to ESI CID/HCD. The peptide mapping was performed using a text file containing four AA sequences in FASTA format of bovine milk caseins including α-S1-, α-S2-, β-, and κ-casein. Top-down sequencing annotations for each of the four casein proteins were exported from Refiner MS module providing a list of the identified peptides along with their positions in the protein sequence they were successfully mapped to. Further processing was performed in R software (version 3.5.2). Amadori product precursor mass was calculated by a mass increase of 162.0528 Da. Amadori product precursor signals were computationally assigned by an algorithm within a mass tolerance of ± 10 ppm. Putatively assigned Amadori products with available MS2 spectra from our data were clustered according to their similarity in normalized intensity profiles using Pearson correlation. Product ion annotation was automatically performed in R software by in silico fragmentation 43,73 and manually validated. Monoisotopic mass tolerance was set to ± 0.005 Da for product ions. To separate false positive assignments, we excluded signals with a poor fit of the MS/MS spectrum to the in silico predicted fragments and maximum intensity in the tryptone control samples heated for 10 h. GeneData Expressionist Refiner MS 13.0 (GeneData GmbH, Basel, Switzerland) peptide mapping activity provides a consolidated score, which describes the average fit for each peptide across all MS2 spectra available. Consolidated scores of all peptides, for which the corresponding Amadori product (MS2 level) could be verified, were computed. The minimum consolidated score per peptide length was chosen as a threshold for peptide identification. ## Statistical analysis. Pearson correlation coefficients were calculated in R software between intensity values for putatively assigned MS2 Amadori products (n ≥ 3, p ≤ 0.05). For this analysis, relative intensity values were used. Relative intensity values were calculated by normalizing intensity values to the maximum intensity value across all time points and for each Amadori product, respectively. Hierarchical clustering to provide the domain ordering was done using R software. Amadori product wise distances were calculated based on these correlations using the as.dist() function followed by hierarchical clustering using the hclust() function. To assess the importance of small sequence variations, pairwise two-sided t-tests were performed. Intensity values in model systems and control samples were compared. Significantly increased Amadori products (p ≤ 0.05) and relevant reaction conditions are reported in Supplementary Table 1. Multiple testing correction was performed using the Benjamini-Hochberg procedure. Sequence grouping. The computational analysis of sequence groups was performed with the peptide single letter code using R software. To identify peptides with common sequences the grepl() and match() base functions were applied. Based on derived sequence commonalities, we assigned all peptides to sequence groups (Supplementary Table 1). Amino acids were not assigned to sequence groups. Sequence groups, for which no Amadori product was detected, were excluded in Supplementary Table 1. Database search. Bioactive peptide database search was carried out using the Milk Bioactive Peptide Database (March 13, 2020) 57 and the BioPepDB database (March 13, 2020) 58 . Sensory peptide database search was

chemsum

{"title": "Molecular characterization of sequence-driven peptide glycation", "journal": "Scientific Reports - Nature"}

a_unified_strategy_to_access_trans-syn-fused_drimane_meroterpenoids:_chemoenzymatic_total_syntheses_

## Abstract: Trans-syn-fused drimane meroterpenoids are unique natural products that arise from contra-thermodynamic polycyclizations of their polyene precursors. Herein we report the first total syntheses of four trans-syn-fused drimane meroterpenoids, namely polysin, N-acetyl-polyveoline, chrodrimanin C and verruculide A in 7-18 steps from sclareolide. The trans-syn-fused drimane unit is accessed through an efficient acid-mediated C9 epimerization of sclareolide. Subsequent applications of enzymatic C-H oxidation and contemporary annulation methodologies install the requisite C3 hydroxyl group and enable rapid generation of structural complexity to provide concise access to these natural products. Meroterpenoids are a highly diverse class of natural products that arise in nature from hybrid terpenoid/non-terpenoid biosynthetic pathways. 1 One highly prevalent motif in many meroterpenoids is the C3-oxidized drimane substructure, which forms the central core of many families including the 3,5-dimethylorsellinic acid-derived fungal meroterpenoids 2 and the sesquiterpenyl indoles. 3 Biosynthetically, this motif is produced through the cyclization of a linear polyisoprene-derived epoxide precursor by various terpene cyclases. Compelling literature evidence 3,4,5 has suggested that these enzymes are capable of pre-organizing their respective substrates in specific conformations to generate products with unique ring topologies and stereoconfigurations, which ultimately contribute to the immense structural diversity of the meroterpenoids. Among the possible polycyclization product topologies, the alltrans configuration is the most favored thermodynamically as it allows the fused cyclohexane rings to adopt an all-chair conformation. However, there exists a subset of drimane-containing meroterpenoids that possess alternative ring fusions (Figure 1A), such as the trans-syn-cis-fusion found in polysin 6 (1) and N-acetyl-polyveoline 7 (2) and the trans-syn-trans-fusion found in the chrodrimanins 8,9 (e.g., 3 and 4). Access to thermodynamically disfavored ring fusions in these meroterpenoids is made possible by the ability of the respective cyclases to generate the less stable, boat-like transition state during their reactions (Figure 1B). Such conformational requirements have proven to be prohibitive for synthetic recapitulation as attempts to effect biomimetic cyclizations to prepare trans-synfused drimanes have been met with limited to no success. 10,11 A synthetic approach towards polyveoline featuring an indoleterminated polyene cyclization failed to overcome the innate thermodynamic preference of the substrate and resulted in exclusive formation of the undesired all-trans product. 10 Though the use of substrates with alternative olefin placement or geometry has garnered some success, 12,13 these approaches have resulted in either sub-optimal diastereoselectivity or low yields for the desired products. 2. A. Synthetic strategy to access polysin, N-acetyl-polyveoline and the chrodrimanins from 9, which could be obtained via C9 epimerization of sclareolide (7). B. Screening of P450BM3 variants in our collection for the C3 hydroxylation of 9. C. Screening of P450BM3 variants in our collection for the C3 hydroxylation of 12. See Supporting Information for the identities of the variants tested. Reaction conditions for enzymatic hydroxylation were: 9 or 12 (5.0 mM), NADP + (1.0 mM), NaHPO3 (100 mM), clarified lysate of E. coli BL21(DE3) expressing the appropriate P450BM3 variant and Opt13 (suspension in 50 mM kPi (pH 8.0) and pre-lysis at an optical density of 30, measured at a wavelength of 600 nm) for 20 h at 20 °C. *additional regioisomers were detected in the product mixture. In parallel with the above efforts, several groups have sought to harness the power of terpene cyclases to biocatalytically access trans-syn-fused terpenoids. Since van Tamelen's landmark study on a cyclase from rat liver, 14 several reports have demonstrated the feasibility of this approach. Virgil and coworkers were able to use an unnatural oxidosqualene derivative in an enzymatic polycyclization to access the isomalabaricane tricyclic core 15 and more recently, a collaborative work by the Porco and Abe laboratories 5 showcased the utility of several fungal cyclases in constructing unnatural meroterpenoids with unusual ring fusions from synthetic substrates. These demonstrations notwithstanding, the approach suffers from low material throughput arising from the inefficiency of the enzymatic reaction and the difficulty in obtaining large quantities of the membrane-bound enzymes. In the context of target-oriented chemical synthesis, only three total syntheses of trans-syn-fused drimane terpenoids have been reported thus far (Figure 1C). Two of these syntheses 16,17 pertain the brasilicardin natural products and involved lengthy synthetic sequences to generate the key trans-syn-trans-fused tricyclic intermediates. More recently, a landmark synthesis of the isomalabaricanes (e.g., stellettin E, 5) by the Sarlah group has enabled initial structure-activity relationship studies on the cytotoxicity of the scaffold. 18,19 To complement the aforementioned approaches, we sought to develop an alternative strategy to collectively prepare trans-syn-fused drimane meroterpenoids through the use of a chiral pool approach. This report discloses the development of a unified strategy to access both trans-syn-cisand trans-syn-trans-fused drimane meroterpenoids from sclareolide that culminates in the first total syntheses of polysin, N-acetyl-polyveoline, chrodrimanin C and verruculide A. To the best of our knowledge, this is the first reported de novo constructions of trans-syn-cis-fused perhydrobenz[e]indene and trans-syn-trans-fused dodecahydro-1Hbenzo[f]chromene frameworks. This work was made possible by the use of an underexplored epimerization reaction on sclareolide, 20 which was combined with enzymatic C-H oxidations and efficient ring annulations to complete the divergent syntheses. We anticipate that the strategy delineated herein will find a broad range of applications in the preparation of other drimane terpenoids with unusual ring fusions. As noted above, our synthetic strategy was predicated upon the ability of sclareolide (7) to undergo facile epimerization at its C8 and C9 positions. Prior report from Ohloff 20 (Figure 2A, inset) showed that treatment of 7 with mineral acid at room temperature could readily afford the C8-epimerized product 12 (8). Alternatively, the C9-epi product (9) was observed as the major product at elevated temperature, likely via elimination to the corresponding C8-C9 olefin, followed by re-protonation from the b-face at C8 and quenching of the C9 carbocation from the a-face by the pendant carboxylic acid. In our hands, this transformation could be routinely conducted on multigram scale with 95% yield. With the C9 stereocenter established, access to polysin and N-acetyl-polyveoline could be accomplished through selective C-H oxidation at C3 and the appendage of a pendant indole unit (Figure 2A). We envisioned introducing the former through enzymatic C-H hydroxylation 21,22 and the latter through ring synthesis by leveraging the C12 carbonyl as a chemical handle. Adaptation of this idea to access the chrodrimanin series would necessitate the invention of a synthetic sequence to construct the C-ring pyran while also inverting the stereochemistry at C8. While the general pyran structure could be prepared via a one-carbon homologation, the stereoinversion at C8 was expected to be non-trivial as it would result in an A/B/C-ring connectivity that forces the B-ring to adopt the energetically-unfavored twist-boat conformation. Nevertheless, if this transformation could be realized, an efficient synthesis of chrodrimanin C would ensue through subsequent use of the C-ring lactone as a chemical handle in an aromatic annulation sequence. Finally, enzymatic conversion of 3 to 4 through a series of in vitro reactions has previously been reported by Matsuda, Abe and co-workers. 23 Scheme 1. Chemoenzymatic total synthesis of polysin (1) and N-acetyl-polyveoline (2) via enzymatic C-H oxidation of 9. In light of our previous work in the synthesis of a-pyrone meroterpenoids from 7, a route involving enzymatic C3 oxidation of 7 with variants of P450BM3, 22 followed by epimerization at C9 was initially considered. However, preliminary forays into this route showed that the C3 alcohol is incompatible with strong acids, even in its protected form. As a workaround, we decided to investigate the feasibility of performing enzymatic C-H oxidation on lactones 9 and 12, which was prepared in seven steps from 9 (vide infra). Despite the high structural similarities of 9 and 12 to 7, it is widely accepted that even minor alterations in substrate structure could result in dramatic changes in reactivity in enzymatic transformations. Gratifyingly, initial screening of a subset of our P450BM3 library revealed a few variants with C3 hydroxylation activity on 9 (Figure 2B). Variant KSA15, 24 previously developed by Reetz and co-workers for steroid hydroxylation, showed the highest conversion (54%) among all the library members tested. Following an analogous screening with lactone 12, variant MERO1 L75A, previously developed in our laboratory for the synthesis of oxidized meroditerpenoids, 22 was identified to be the optimal enzyme to hydroxylate 12 at C3 (Figure 2C). While variant KSA15 provided higher conversion (95%) in its reaction with 12, additional product regioisomers could be detected. Thus, we elected to perform subsequent C-H oxidation scale-up with MERO1 L75A. With the above results in hand, we set our sights establishing a concise access to polysin and N-acetyl-polyveoline (Scheme 1). Preparative scale enzymatic hydroxylation of 9 provided alcohol 13 with 67% yield, which was subjected to a Smith-modified Madelung indole synthesis 25 to provide a mixture of two adducts, 14 and 15. Treatment of this mixture with p-toluenesulfonic acid (PTSA) effected complete formation of the indole nucleus with concomitant dehydration of the tertiary alcohol at C8. In light of its potential incompatibility with acidic conditions needed for the subsequent cyclization step, oxidation of the C3 alcohol at this stage was deemed strategic and was accomplished using Albright-Goldman protocols. 26 Following an extensive screening of Lewis acids (see Supporting Information Table S4), we arrived at the use of MK-10 under microwave heating to generate a mixture of Friedel-Crafts adducts. As the C-cyclized product was observed to be unstable, an in-situ capping approach with Ac2O was devised to deliver a mixture of enol ethers 17 and 18 in 40% and 25% yields respectively under telescoped procedure. Routine saponification of 18 completed the synthesis of polysin (1) in seven steps from 7. Conversely, 17 was subjected to hydrogenation in the presence of palladium on carbon, followed by a diastereoselective reduction with K-selectride to complete the synthesis of N-acetyl-polyveoline (2) in eight steps from sclareolide (7). S5). Unfortunately, no marked increase in diastereoselectivity was observed in all conditions tried and under the best set of conditions, a diastereomeric ratio of 1:1 at C8 was obtained. At this stage, the desired tertiary alcohol diastereomer 22 was saponified and converted to trans-syn-trans-fused lactone 12 through the use of Yamaguchi's reagent. To improve material throughput, the unwanted diastereomer 21 could be recycled into the sequence by simple methyl ester formation to regenerate 20 along with its olefin regioisomer. After three cycles, a combined 60% isolated yield of 22 could be achieved. Enzymatic hydroxylation of 12 with P450BM3 variant MERO1 L75A was next conducted on preparative scale to provide alcohol 11 in 82% yield. The structure of this compound was verified by Xray diffraction analysis, which prominently revealed the twistboat configuration of the B-ring. Drawing inspiration from the syntheses of arene-containing terpenoids by Li and co-workers, 28,29 we sought to construct the central arene ring of the chrodrimanins through a 6p electrocyclization of the corresponding triene precursor. Toward this goal, the C3 alcohol of 11 was temporarily protected as the trimethylsilyl (TMS) ether and the C-ring lactone was converted to the corresponding vinyl triflate (compound 23). Sonogashira coupling of 23 with alkyne 24, synthesized in 6 steps from (R)-methyl 3-hydroxybutanoate (see Supporting Information), delivered dienyne 25 in 67% yield over 3 steps from alcohol 11. With the goal of introducing a suitable functional handle for subsequent phenol formation, an alkyne hydrosilylation approach was pursued. Previous work by Ferreira and co-workers 30 showed that the regioselectivity of alkyne hydrosilylation is predominantly dictated by electronic effects whereby hydride delivery would take place at the sp carbon that is further away from electron-withdrawing group. Indeed, treatment of alkyne 25 with Et3SiH in the presence of catalytic Pt(DVDS) successfully provided the desired hydrosilylation product 26 as a single regioisomer. ## While indole hydrogenation typically requires high H2 Following precedent by Li, Nicolaou and co-workers, 31 a 6p electrocyclization/aromatization sequence could be effected to generate arene 27. In agreement with their work, the use of CuOTf as a Lewis acid promoter was found to improve the yield of the reaction (73% isolated yield) while also effecting a concomitant hydrolysis of the TMS ether at C3. While oxidation of the C3 alcohol to the corresponding ketone proceeded uneventfully, attempts to effect a Fleming-Tamao oxidation 32 to convert 27 to the corresponding phenol were met with failure. Similar outcomes were obtained when alternative silanes at C4' were tested in the reaction. Earlier iterations of the route featuring a late-stage sp 2 C-H oxidation at C4' using Ru catalysis 33 or peroxide-based reagents 34,35 also failed to deliver the desired product. As a workaround, silane 27 was first subjected to desilylative iodination with NIS to provide 28. Following screening of several reported conditions for haloarene hydroxylation, access to chrodrimanin C (3) could be realized through the use of Cu(acac)2 and N,N'-bis(4-hydroxyl-2,6-dimethylphenyl)oxalamide (BHMPO) on 28. 36 This method, initially reported by Ma and co-workers, proved superior to Pd-based hydroxylation methods 37,38 and with slight modifications to the originally reported conditions, the desired phenol product could be obtained in 83% isolated yield. A-ring desaturation of 3 proceeded uneventfully under standard Saegusa conditions to deliver verruculide A (29) in 82% yield. Interestingly, TMS ether formation at the phenolic OH was not observed in this reaction, likely due to the presence of an intramolecular hydrogen bonding with the neighboring lactone. Overall, this sequence provided a 16-step synthesis of chrodrimanin C (3) and a 18-step synthesis of verruculide A (29) from sclareolide (7), respectively. As noted earlier, the biosynthetic pathway towards the chrodrimanins was recently elucidated by Matsuda, Abe and co-workers 23 and we anticipate that future work involving incorporation of some of the enzymes from the pathway would allow for a rapid chemoenzymatic diversification of the scaffold to provide a wider range of synthetic chrodrimanins. This work reports the development of a chiral-pool-based strategy for the asymmetric synthesis of trans-syn-fused drimane meroterpenoids. Two enabling features in the synthesis are the strategic use of an acid-mediated C9-epimerization of sclareolide to generate the general trans-syn-fused architecture of these natural products and the ability to perform regioselective C-H oxidations on different key synthetic intermediates at their C3 position by relying on a small pool of P450BM3 biocatalysts. By combining these features with contemporary annulation methodologies, the first total syntheses of polysin, N-acetyl-polyveoline, chrodrimanin C and verruculide A could be realized. The route disclosed herein lays the foundation for future synthetic access to other unusually-cyclized meroterpenoids and their unnatural derivatives to facilitate a more thorough investigation into their pharmacology.

chemsum

{"title": "A Unified Strategy to Access Trans-Syn-Fused Drimane Meroterpenoids: Chemoenzymatic Total Syntheses of Polysin, N-Acetyl-Polyveoline and the Chrodrimanins", "journal": "ChemRxiv"}

a_model_for_optical_gain_in_colloidal_nanoplatelets

## Abstract: Cadmium chalcogenide nanoplatelets (NPLs) and their heterostructures have been reported to have low gain thresholds and large gain coefficients, showing great potential for lasing applications. However, the further improvement of the optical gain properties of NPLs is hindered by a lack of models that can account for their optical gain characteristics and predict their dependence on the properties (such as lateral size, concentration, and/or optical density). Herein, we report a systematic study of optical gain (OG) in 4monolayer thick CdSe NPLs by both transient absorption spectroscopy study of colloidal solutions and amplified spontaneous emission (ASE) measurement of thin films. We showed that comparing samples with the same optical density at the excitation, the OG threshold is not dependent of the NPL lateral area, while the saturation gain amplitude is dependent on the NPL lateral area when comparing samples with the same optical density at the excitation wavelength. Both the OG and ASE thresholds increase with the optical density at the excitation wavelength for samples of the same NPL thickness and lateral area. We proposed an OG model for NPLs that can successfully account for the observed lateral area and optical density dependences. The model reveals that OG originates from stimulated emission from the bi-exciton states and the OG threshold is reached when the average number of excitons per NPL is about half the occupation of the band-edge exciton states. The model can also rationalize the much lower OG thresholds in the NPLs compared to QDs. This work provides a microscopic understanding of the dependence of the OG properties on the morphology of the colloidal nanocrystals and important guidance for the rational optimization of the lasing performance of NPLs and other 1-and 2-dimensional nanocrystals. ## Introduction Cadmium chalcogenide nanoplatelets (NPLs), CdX (X ¼ Se, S, Te), and their heterostructures have shown many novel properties, such as large absorption cross-sections, uniform 1D quantum confnement, long biexciton Auger lifetimes, and giant oscillator strength. These materials have attracted intense interest for lasing applications due to the reported large gain coefficients and low optical gain (OG) threshold. For example, the reported threshold of the amplifed spontaneous emission (ASE) of the CdSe NPLs is as low as 6 mJ cm 2 , 15 which is over an order of magnitude lower than that in cadmium chalcogenide quantum dots (QDs) or QD heterostructures. 26,27 Although OG models for QDs are well understood, it is unclear whether they are applicable to 1D nanorods (NRs) and 2D NPLs because of the fundamental differences in their exciton properties. In 0D QDs, the excitons are confned in all three dimensions, whereas in 2D NPLs and 1D NRs, the excitons are free to move in the plane and along the long axis, respectively, which increases the degree of degeneracy of the band edge exciton states and may alter their gain mechanism. So far, there is not an OG model for 2D NPLs or 1D NRs, and the reasons for the superior OG properties in NPLs remain unclear. In addition, many other interesting differences between the NPLs and QDs may also contribute to their different OG properties. For example, because of the atomically precise thickness, the NPLs have uniform quantum confnement energy and narrow exciton transition linewidth, which should reduce the overlap between the stimulated absorption (loss) and emission (gain) transitions. 3 It has also been argued that the exciton transition oscillator strength in NPLs may be enhanced by the coherent delocalization of the exciton center of mass in the lateral direction, which should affect the stimulated emission cross section. 3 The biexciton Auger recombination lifetimes in the NPLs are much longer than QDs 6,17 and have been shown to increase linearly with their lateral size. It has been proposed that the low OG thresholds in NPLs can be attributed to their longer Auger lifetime. 17,19 These observations would suggest that one of the key differences between NPLs and QDs is the possibility of tuning their OG performance through their lateral size. Olutas et al. have reported that the ASE threshold of CdSe NPLs increases with their lateral area. 18 However, She et al. reported a lateral areaindependent ASE threshold of the same materials. 19 These contradictory observations and a lack of understanding of the OG mechanisms in NPLs suggests the need for a systematic study and a model for optical gain in these materials. Herein, we report a systematic study of the dependence of OG on the lateral area and optical density at pump wavelength in 4 monolayer (ML) CdSe NPLs. We investigate the OG characteristics by femtosecond transient absorption (TA) spectroscopy of the colloidal NPL samples and ASE measurements of the NPL flms at room temperature. We show that the OG thresholds are independent of the lateral area of the NPLs, whereas the saturation OG amplitude increases linearly with the area when comparing samples of the same optical density at the excitation wavelength. For NPLs of the same size, their OG and ASE thresholds increase with the optical density at the excitation wavelength. We propose a biexciton gain model that can satisfactorily account for the experimental observations in the NPLs and explain the origin for their much lower OG thresholds compared to QDs. We believe that this model should also be applicable to other 2D nanosheets and 1D nanorods. ## Sample characterization 4 ML CdSe NPLs (with 5 Cd layers, 4 Se layers, and a thickness of $1.8 nm) were synthesized according to reported procedures with slight modifcations. 3 The lateral size of the NPLs was tuned by changing the synthesis temperature and reaction time. The detailed synthesis procedures are described in the ESI. † The NPL samples with different lateral sizes are named NPLa to NPLd with increasing lateral size. The same batch of samples have been used in a previous study of the lateral size dependence of the biexciton Auger lifetime in NPLs. 28 S1 †). The absorption spectra of NPLa to NPLd (solid lines, Fig. 1b) show A ($512 nm) and B ($480 nm) exciton peaks that correspond to the electron-heavy hole (e-hh) and electron-light hole (e-lh) transitions, respectively. All the NPL samples of different lateral sizes have the same A and B exciton transition energy. The static photoluminescence (PL) spectrum of NPLb (blue dashed line in Fig. 1b) shows a sharp band edge (e-hh) emission peak at $518 nm with a full width at half maximum (FWHM) of $38 meV. The PL spectra of NPLa to NPLd are compared in Fig. S3, † showing that the band edge emission is independent of the lateral size. ## Lateral area independent optical gain threshold To determine the optical gain threshold, we carried out a pump fluence dependent TA spectroscopic study of NPLa to NPLd in hexane at room temperature. In a TA measurement, the optical density of the samples under illumination is given by DA(l,t) + A 0 (l), where DA(l,t) is the pump-induced absorbance change shown in the transient absorption spectra and A 0 (l) is the static absorbance prior to excitation. Thus optical gain is achieved when DA(l,t) + A 0 (l) < 0. Because the gain threshold is dependent on the optical density of the sample (see below), to enable the comparison of the NPL samples of different lateral areas, their optical density at pump wavelength (400 nm) was controlled to the same value to ensure the same number of absorbed photons (Fig. 1b). The TA spectra of NPLc at the lowest pump fluence (3 mJ cm 2 , Fig. S4a †) show long-lived bleach signals of A ($512 nm) and B ($480 nm) excitons. According to our previous work on CdSe NPLs, both A and B exciton bleaches can be attributed to state-flling on the frst electron level in the conduction band (CB), and the contribution of the hole state-flling in the valence band (VB) is negligible due to degeneracy and strong mixing between the denser hole levels in the VB, 14,30,31 similar to cadmium chalcogenide quantum dots and nanorods. Fig. S4b † shows the TA spectra of NPLc at the highest pump fluence (629 mJ cm 2 ) when the bleach amplitudes of the A and B exciton states at an early delay time have saturated. Compared to those at low pump fluence, these spectra show an additional broad negative peak, DA(l,t) < 0, at energy lower than the A exciton ($520-560 nm), which can be attributed to the optical gain (OG) signal, 19 similar to that reported in CdSe quantum dots (QDs). 27 The gain spectra (DA(l,t) A 0 (l)) at 3-4 ps of NPLc at different pump fluences are shown in Fig. 2a and an expanded view of the gain spectra (Fig. 2a inset) shows a broad OG peak with a maximum at $528 nm. The kinetics of the gain signal of NPLc at the OG peak wavelength ($528 nm) and with different pump fluences are compared in Fig. 2b. All the kinetics show a negative signal around time zero (<1 ps), which reflects a red-shifted exciton absorption caused by exciton-exciton interaction. 14,20, After 1 ps, the OG amplitude remains negative, indicating no OG, for pump fluences below 27 mJ cm 2 . The OG amplitude increases with increasing pump fluence and becomes positive, indicating As shown in Fig. 2b and S4, † the OG amplitude of all the samples reaches the highest value at a delay time of 3-4 ps, after which the OG signals decay due to multiple exciton Auger recombination. 28 A plot of the maximum OG amplitudes (at 3-4 ps) as a function of the pump fluence (Fig. S7b †) shows that for all the NPL samples, the OG reaches saturation at high pump fluences, but the saturation OG value increases linearly with lateral area (Fig. 2c). To facilitate comparison of the gain threshold, we have scaled the OG of different samples to the same saturation amplitude and plotted the normalized OG as a function of the pump fluence in Fig. 2d. The comparison shows that the normalized OG of all the NPL samples exhibits the same dependence on the pump fluence: a linear increase of OG with pump fluence between 15-150 mJ cm 2 and reaching saturation between 150-500 mJ cm 2 . As shown in the inset of Fig. 2d, the intercept of the OG amplitude on the x-axis yields the same OG threshold of 54.6 AE 1.8 mJ cm 2 for all the four samples under our experimental conditions (optical density of 0.31 AE 0.01 at 400 nm pump), independent of their lateral area. ## Optical density dependent OG threshold To investigate how the OG threshold changes with sample optical density, we carried out TA study of the NPLc samples in hexane solution with different concentrations, named NPLc1 to NPLc4 in the order of increasing NPL concentration (NPLc3 is the sample used in Fig. 2). The absorption spectra of NPLc1 to NPLc4 (Fig. 3a) show that the optical density at 400 nm increases from 0.12 to 0.49 from sample NPLc1 to NPLc4. These samples were investigated using the same pump fluence dependent TA measurement and analysis method described above. Their OG kinetics as a function of the pump fluence are shown in Fig. S4. † Their peak OG amplitudes at 3-4 ps and $528 nm are plotted as a function of the pump fluence in Fig. 3b. The intercept of these data on the x-axis yields OG thresholds of 43.0 AE 1.6, 52.5 AE 1.7, 54.6 AE 1.8, and 63.5 AE 2.2 mJ cm 2 for NPLc1 to NPLc4, respectively. As shown in Fig. 3c, these OG threshold values increase with the optical density at pump wavelength (400 nm). Fig. 3d shows that the saturation OG amplitude increases linearly with optical density at 400 nm, indicating more gain at saturation if more photons are absorbed. Similar optical density dependent ASE thresholds using NPLc flms prepared by spin-coating of NPLc solutions with different concentrations on a glass substrate were also observed (Fig. S5 †). ## Model of the optical gain threshold To explain the experimental results described above, we propose a model for OG in NPLs. The details of this OG model can be found in the ESI † and only the key aspects are summarized here. This model is an extension of the previous gain model proposed for QDs, 26 which, because of the confnement in all three dimensions, can only accommodate two band edge excitons. In this model, we assume that because of the large (unconfned) lateral dimension of NPLs, the number (N s ) of band edge (or A) excitons can exceed 2, increasing the complexity of the number of transitions associated with single and multiple band edge exciton states, as shown in Fig. 4 (for an example of N s ¼ 4). This assumption is based on our previous observations of NPLs, 20 and the 2D hydrogen-like exciton model in 2D structures. 35,36 On the basis of the redshift of OG and ASE from NPL emission, it has been proposed that OG or ASE in NPLs can be attributed to stimulated emission from band edge bi-exciton states, similar to QDs. 27 Therefore, our model only considers band edge exciton states with 0, 1, . N s band edge excitons, which are labeled as 0, X, XX, . states, respectively, and their population probabilities are indicated by N i (i ¼ 0 to N s ). Each exciton state (i) can undergo stimulated absorption (upward arrows in Fig. 4) or emission (downward arrows in Fig. 4) with partial cross-sections per NPL of A i (i ¼ 0 to N s 1) and A * i (i ¼ 1 to N s ), respectively, given by eqn (1) and (2): i from 0 to N s 1; (1) In eqn ( 1) and ( 2), h is the Planck's constant, n r is the refractive index, and c is the speed of light. A T is the transition strength of the band edge excitons (e-hh) per NPL, which is proportional to the NPL lateral area, A QW . 2g i and 2g * i are the full width at half maximum of the absorption and emission spectra of the N i species, respectively. E i and E * i are the stimulated absorption and emission peak energy for the N i species, respectively. We set both g 0 and g * 1 to $19 meV for both single band edge exciton absorption and emission according to Fig. 1b. We assume both g i1 and g * i (i from 2 to N s ) to be the same as the broad OG spectra ($50 meV) shown in Fig. 2a. E 0 and E * 1 are 2.42 eV and 2.39 eV, respectively, according to the A exciton absorption (512 nm) and emission (518 nm) wavelengths in Fig. 1b. The energy of bi-exciton absorption (E 1 ) and emission ðE * 2 Þ is assumed to be the optical gain energy, E OG , which is set to the OG peak value (2.35 eV, 528 nm) according to Fig. 2a. The inhomogeneous distribution of E OG is ignored due to the uniform 1D quantum confnement of the NPLs. The shift (from E OG ) of the transition energies for the tri and higher exciton states is assumed to be much smaller than the transition line width: This assumption is based on the broad transition width for the tri and higher exciton states and Coulomb screening of the multiexcitons reported in other 2D materials, 37 although these values have not been observed directly in our NPLs. The absorption coefficient of the NPL ensemble at OG energy is: where N en is the number of NPLs in the ensemble, which is proportional to the NPL molar concentration, C m . The population probability of the NPL species (N i ) is assumed to follow Poisson distribution: P n ðmÞ ¼ m n e m n! , which represents the possibility of fnding NPLs with n excitons when the average number of excitons per NPL is m. The optical gain threshold is achieved when a(E OG ) ¼ 0. Solving eqn (3) numerically under this condition leads to m th (N s ), the average number of excitons per NPL at the OG threshold, of 0.49 (AE0.01) N s (see Fig. S6 and Table S2 †). The result suggests that OG is reached when about half of the band edge exciton states are occupied. Under this condition, the gain (emission from excited states) equals the loss (absorption from ground states). Within the limits of QDs (N s ¼ 2), m th $ 1, which is consistent with previous fndings on QDs. 26 Because m is proportional to the pump fluence (I) and the optical density at pump wavelength following Beer's law, the m th value can be converted to threshold pump fluence, i.e. the OG threshold (I th ), according to eqn (4). In eqn (4), OD ¼ 3zA QW C m L, hn is the pump photon energy (3.1 eV), N A is Avogadro's constant, A X ¼ A QW /N s , 3 is the molar absorption coefficient per unit NPL volume, z is the NPL thickness ($1.8 nm), L is the light path of the cuvette (1 mm) and m th /N s ¼ 0.49 AE 0.01. The details of the derivation can be found in the ESI. † According to eqn (4), when comparing NPL samples of the same thickness, their OG thresholds are independent of the NPL lateral area as long as the optical densities at pump wavelength are the same. This prediction is consistent with the experimental result shown in Fig. 2d. Moreover, the observed OD dependent OG and ASE thresholds can be well ftted by eqn (4), as shown in Fig. 3c and S5f †, respectively, providing further support for our OG model. At the limit of large m, the optical gain reached saturation with the gain amplitude given by eqn (5). This predicts that the saturation gain amplitude increases linearly with both the lateral area (proportional to N s ) and optical density (OD) of the NPL (Fig. S7a †), both of which are consistent with the experimental fndings as shown in Fig. 2c and 3d, respectively. Finally, our model (eqn (3)) also predicts how OG increases with the pump fluence. The observed OG amplitude as a function of m can be reasonably well ftted by our model (Fig. S7h †), although the simulated OG saturates at a lower value of m compared to the experimental results for NPLs with large N s (N s > 3). The origin of this deviation is not well understood, but it indicates that some loss factors are not fully accounted for in our model. This is likely due to the lack of consideration of a transition width distribution from higher exciton states in our model, which have not been experimentally observed. There have been two contradicting reports on whether the ASE threshold depends on the NPL lateral area. 18,19 In ref. 19, the optical density at pump wavelength of different NPL samples was controlled to similar values, and the lateral areaindependent ASE thresholds were observed, 19 which is consistent with our experimental results and OG model. In ref. 18, the lateral area dependent ASE threshold was observed, but it is unclear whether the optical density at pump wavelength for samples of different NPL areas was controlled to the same values. 18 Our result suggests that optical gain is achieved when the average number of excitons per NPL is close to half (0.49) of the band edge exciton states, which is similar to the OG requirement in QDs. Despite this similarity, the optical gain thresholds in QDs have been reported to be more than an order of magnitude higher than those in NPLs. 15,26,27 According to our model, the lower OG threshold of the NPLs can be attributed to the following reasons. First, the intrinsic absorption cross section of NPLs, i.e. the absorption coefficient per unit volume (3), is larger than that of QDs, which according to eqn (4) leads to lower OG threshold. Recently, Achtstein et al. have reported that the intrinsic absorption coefficient of CdSe NPLs is over 3 folds larger than that in CdSe QDs due to the larger aspect ratio of NPLs. 38 Second, the ratio of biexciton binding energy ($40 meV) and transition linewidth ($38 meV), ðE * 1 E * 2 Þ=2g * 1 , in NPLs ($1.0) is larger than that in QDs (<0.3) whose biexciton binding energy is <30 meV and transition linewidth is $100 meV. 26 This reduces the overlap between the gain and loss transitions, decreasing the OG threshold in NPLs. The latter can be attributed to the atomically precise uniform thickness of NPLs, which reduces the inhomogeneous broadening of the exciton transition energy. Such sharp transitions are difficult to achieve in QDs because of the broad size distribution and large inhomogeneous distribution of transition energies. In addition, the symmetry of the NPLs dictates that both the electric feld of the exciton and the dipole moments lie within the lateral plane, 39 which may account for the observed large Stark effect induced shift of transition energy between the bi-exciton and single exciton states. ## Conclusions In summary, we have systematically studied the dependence of the OG properties of CdSe NPLs on their lateral area and the optical density at pump wavelength using TA spectroscopy and ASE measurements. We show that the OG threshold is lateral area independent when comparing samples of the same optical density at the excitation wavelength, although the saturation OG amplitude increases with the lateral area. Furthermore, for samples of the same NPL size, the OG and ASE threshold increases with their optical density at pump wavelength. To account for these observations, we proposed an optical gain model for 2D CdSe NPLs. This model assumes that the number of band edge excitons scales with the NPL lateral area (and can exceed 2) and optical gain results from the stimulated emission from biexciton states. Our model successfully explains the experimental observations. The model also reveals that OG is achieved when the average number of excitons reaches $49% of the band edge exciton states. This OG requirement is similar to that in QDs, despite the observed OG threshold in NPLs being an order of magnitude smaller than that in QDs. According to our model, the lower OG threshold of NPLs can be attributed to their unique 2D morphology, which leads to a larger intrinsic absorption coefficient, narrower transition linewidth, and larger shift between the bi-and single-exciton state. This work provides not only important insights on how the crystal morphology affects the OG properties of the colloidal nanocrystals, but also guidance on the rational improvement of the OG and ASE in NPL materials for lasing applications. Finally, we believe that this OG model should be applicable to other 2D and 1D nanocrystals. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "A model for optical gain in colloidal nanoplatelets", "journal": "Royal Society of Chemistry (RSC)"}

cycloparaphenylene-phenalenyl_radical_and_its_dimeric_double_nanohoop

## Abstract: We report the first example of a neutral spin-delocalized carbon-nanoring radical, achieved by integration of an open-shell graphene fragment phenalenyl into cycloparaphenylene (CPP). We show that spin distribution in this hydrocarbon partially extends from the phenalenyl onto the CPP segment as an interplay of steric and electronic effects. The resulting geometry is reminiscent of a diamond ring, with pseudo-perpendicular arrangement of the radial and the planar π-surface. Remarkably, this geometry gives rise to a steric effect that governs a highly selective dimerization pathway, yielding a giant double nanohoop. Its π-framework made of 158 sp 2 -carbon atoms was unambiguously elucidated by single-crystal X-ray diffraction, which revealed a three-segment CPP-peropyrene-CPP structure. This nanocarbon shows a fluorescence profile characteristic of peropyrene, regardless of which segment gets excited. These results in conjunction with DFT suggest that adjustment of the size of the CPP segments in this double nanohoop could deliver true donor-acceptor systems.Open-shell nanographenes [1] characterized by delocalized spin density are synthetic targets of fundamental interest to chemists. They are investigated for their amphoteric redox properties, [2] magnetism [ 3 ] and conductivity, [ 4 ] and potential applications in spintronics [ 5 ] and quantum computing. [6] The prototype of such class of compounds is phenalenyl (Phen; Figure 1a), a spin-1/2 odd-alternant neutral hydrocarbon and the smallest triangular graphene fragment. [ 7 ] Its characterization by electron paramagnetic resonance (EPR) spectroscopy in 1950s confirmed the delocalized nature of its open-shell electronic structure, [ 8 ] where one of the 13 π-electrons is unpaired and evenly distributed between the six α-positions of the triangular core (Figure 1a). In solution, phenalenyl exists in equilibrium with its σ-dimer, which in the presence of oxygen undergoes a reaction cascade to yield peropyrene as the final product. The reactivity of phenalenyl can be controlled by steric effects of substituents or electronically by π-extension that affects spin distribution, which can lead to a selective reactivity or suppress it. The known π-extended spin-1/2 derivatives of phenalenyl are restricted to planar and helical examples. [13a,14] To explore new structural platforms, we turned our attention to a cylindrical cycloparaphenylene (CPP) framework, which offers unique opportunities to control spin distribution and reactivity by a combination of steric and electronic effects. CPPs are structural models and synthetic templates for armchair carbon nanotubes. On account of their curved cyclic π-conjugation and hollow interior, they are attractive synthetic targets in material science and supramolecular chemistry. Their curved geometry weakens the overlap of p orbitals, which gives rise to a partial quinoidal character that increases with decreasing diameter. Quinoidal structures favor delocalization of unpaired electrons, as it does not involve the loss of aromatic stabilization energy, which has been observed in radical cations and anions of CPPs but not in neutral species. Changing the linkage mode of one phenylene ring from para to meta results in a geometry, where the meta-phenylene ring lies within the CPP plane perpendicularly to the neighboring para-phenylene rings, which disrupts their π-communication. To utilize the potential of CPPs as a platform to control spin distribution and reactivity, we began to explore hybrid nanocarbon systems, comprising open-shell nanographenes integrated within the CPP nanorings. Herein, we present the prototype of such systems, a CPP-based neutral radical CPP-Phen (Figure 1b), in which one phenylene ring of the CPP loop is replaced by a phenalenyl unit connected in a pseudo-meta-fashion via two α-positions (α3 and α6). This linkage mode results in a geometry reminiscent of a diamond ring, where the phenalenyl unit lies in the CPP plane, that is, almost perpendicularly to the connecting phenylene rings. Consequently, (1) the majority of spin density resides on the phenalenyl unit and only a part of it extends onto the CPP segment, and (2) the dimerization pathway through the α3-α6-positions of the phenalenyl unit is suppressed. This steric effect leads to a highly selective dimerization via α1-and α2-postions, yielding a double-nanohoop peropyrene (CPP-PP; Figure 1c) as the exclusive dimeric product. Notably, CPP-PP represents a fully π-conjugated hydrocarbon framework made of 158 sp 2 -carbon atoms as a novel member of CPP architectures that feature multiple nanoring-cycles. The pivotal intermediate in our synthetic strategy was nanoring 1 (Scheme 1), formed by one dihydrophenalenone unit and 11 phenylene rings. It was synthesized by means of the Suzuki-Miyaura cross-coupling macrocyclization of dihydrophenalenone 2 and C-shaped linker 3. Precursor 2 was prepared in multiple steps starting from trisubstituted naphthalene 4, which first underwent a selective Kumada cross-coupling with 4-(trimethylsilyl)phenylmagnesium bromide to afford 5 in 76% yield. The subsequent Heck coupling and the reduction of the C-C double bond followed by hydrolysis afforded 6 in 71% yield. In the final step, ICl-promoted iodination, acyl chloride formation and Friedel-Crafts acylation provided 2 in 65% yield. The structure of this key intermediate was determined by single-crystal X-ray diffraction (SC-XRD; Figure S2). The directing angle of the corner linker plays a crucial role in the construction of the cyclic scaffold in the CPP chemistry. After screening various candidates, C-shaped linker 3 was found ideal to unite with 2 and afford nanoring 1 in 15% yield upon reductive aromatization. The structure of 1 was unequivocally validated by SC-XRD. The inspection of the solid-state structure (Figure S3) revealed that the dihydrophenalenone unit is non-coplanar with the neighboring phenylene rings, with the dihedral angles (64° and 67°) significantly larger than those between any two neighboring phenylene rings of the CPP segment (28° on average). Finally, 1 was subjected to the reduction and dehydration to afford CPP-1H-phenalene intermediate, the direct precursor of CPP-Phen. Upon the addition of p-chloranil in toluene at room temperature under a nitrogen atmosphere, the pale-yellow solution of the hydroprecursor rapidly changes color to yellow, then green, and after some time, a dark brown-orange mixture is observed. This observation suggests the "decomposition" pathway of phenalenyl to peropyrene, which involves σ-dimer formation as shown in Scheme 1. Kubo and co-workers identified [10b] all intermediates of this reaction sequence (Scheme S3), which involves three oxidation steps starting from hydroprecursor (-6H). This means that at least three equivalents of p-chloranil had to be used for the full conversion of CPP-1H-phenalene to CPP-PP. The use of less than three equivalents of p-chloranil results in a mixture of all intermediates -these conditions were used for the EPR spectroscopic characterization of CPP-Phen. The dimeric product CPP-PP was separated and its formation was confirmed with MALDI-TOF MS, which showed ionized species with an m/z value of 2226.91 (Figure S1). CPP-PP exhibited a poor solubility in common organic solvents due to its large and rigid structure. It is noteworthy that the total isolated yield of CPP-PP starting from 1 is 32% (~80% per step), surpassing that of peropyrene from a predimerized dihydrophenalenone precursor (<22%). [10b] This facile π-skeletal expansion from CPP-Phen to CPP-PP via an oxidative dimerization cascade represents a novel synthetic tactic to achieve double nanohoop architectures. It is distinct from the conventional synthesis of lemniscate CPPs, where the bridging building block is constructed prior to the macrocyclization. In addition, to better understand the photophysical properties of CPP-PP (vide infra), two segments of it, namely, CPP and PP (Scheme 1b), were synthesized and their structures were fully characterized including SC-XRD (see the Supporting Information). The paramagnetic nature of CPP-Phen was probed by EPR spectroscopy. A diluted toluene solution of CPP-1H-phenalene and p-chloranil (1:1, ~10 −3 M) gave a well-resolved 16-peakmultiplet EPR spectrum at 307 K (Figure 2d) with a g value of 2.0039, which is typical for delocalized spin-1/2 hydrocarbon radicals. The measured EPR spectrum was simulated as a quartet of a pentet (qp) using proton hyperfine coupling constants (hcc) of 6.10 G (four α-protons) and 1.95 G (three β-protons). This splitting pattern is in an excellent agreement with the results of DFT calculations, which reveal that the majority of spin density is localized on the phenalenyl unit (Figure 2b,c). The simulated hcc values agree well with the calculated ones (Figure 2a, ∼6.5 G for α-and ∼2.4 G for β-protons), which are slightly higher. Based on DFT, the protons of the neighboring two phenylene rings possess dramatically smaller hcc values (0.27-0.58 G), which are not resolved in the experimental spectrum. This spin distribution can be rationalized by a pseudoperpendicular arrangement between the phenalenyl unit and the neighboring phenylene rings, which is favored for units with a meta-linkage (angle of 120°). The steric preference for perpendicular arrangement acts against the electronic preference for a coplanar arrangement, which is ideal for spin-delocalization onto the phenylene rings. This "clash of forces" is supported by the calculated dihedral angles (~64° on average, Figure S10), which are lower compared to CPP-PP (~80° on average, Figure 3a), and by comparing CPP-Phen to its strain-free linear analog (Figures S5 and S12), which displays a higher degree of spin density on the phenylene rings. Interestingly, in the lowest-energy geometry, the spin density in both systems is not delocalized evenly on both sides. Single crystals of CPP-PP were obtained by slow vapor diffusion of acetonitrile into a CHCl3 solution at room temperature. The XRD analysis unambiguously confirmed the fully π-conjugated framework of CPP-PP with a C2 symmetry, where the peropyrene segment serves as a rigid bridge between two CPP segments. The dihedral angles at the four linkages are ~80° on average (Figure 3a, green), implying an unprecedented perpendicular alignment of three π-conjugated segments (radial-planar-radial). The overall length reaches 4.20 nm and the two cavities have an oval shape with a short axis of 1.52 nm. The distance between the protons of the peropyrene segment and the opposite propoxylated phenylene rings is 1.45 nm, which implies that the cavities of CPP-PP might be suited for the encapsulation of fullerenes such as C60 or C70. No evidence of binding C60 or C70, however, was observed, which suggests that the cavities are not an ideal fit. The herringbone packing structure is similar to that of most CPP molecules [30a] that arrange in staggered layers (Figure 3c). The photophysical properties of CPP, PP and CPP-PP were investigated in CHCl3 solutions (Figure 4). CPP showed only one intense absorption band at 342 nm with an extinction coefficient (ε) of 1.49 × 10 5 M −1 cm −1 , which is nearly identical to that of parent CPP. According to timedependent (TD)-DFT calculations, the HOMO→LUMO (S0→S1) transition is weakly allowed with f = 0.283, and the major contributions for the observed band are HOMO−1→LUMO (S0→S2, f = 2.30) as well as HOMO−2→LUMO and HOMO→LUMO+1 (S0→S3, f = 3.26) transitions (Figure S17). PP exhibited three main absorption bands: the most intensive one at 345 nm with ε = 0.77 × 10 5 M −1 cm −1 and two less intensive bands at 447 and 477 nm, which are characteristic of the peropyrene backbone (HOMO→LUMO (S0→S1, f = 1.47) transition, Figure S18). CPP-PP showed an absorption spectrum that represents a superposition of those of CPP and PP, except a minor shift (Figure 4a). The dominant absorption band at 338 nm (ε = 2.24 × 10 5 M −1 cm −1 ) is primarily derived from the absorption of two CPP segments, and it can be ascribed to multiple HOMO−n→LUMO+m transitions (n = 0-7, m = 0-11; S0→S5, f = 4.46; S0→S7, f = 4.43; Figure S19), including CPP-localized (e.g., HOMO−3→LUMO+2) and charge-transfer transitions (e.g., HOMO→LUMO+5). In addition, the two typical absorption bands of the peropyrene segment were observed at 434 nm and 463 nm (HOMO→LUMO (S0→S1, f = 2.15) peropyrene-localized transition), blue-shifted by 14 nm compared to those of PP. This blue shift is presumably due to the weaker conjugation between the peropyrene and CPP segments, as indicated by the significantly larger dihedral angles at the four linkages (avg. 80°, Figure 3a) compare to those of PP (avg. 57°, Figure S4). The profile of the fluorescence spectrum of CPP-PP is almost identical to that of PP, except a blue shift of 18 nm, without any contribution from CPP (Figure 4). For a comparison, a 1:1 mixture of CPP and PP shows a fluorescence profile, which represents the superposition of the individual spectra of both components. The fluorescence quantum yields for CPP, PP and CPP-PP in CHCl3 are 0.63, 0.76 and 0.75, respectively (excitation of 350 nm, Figure S7). These results indicate an efficient internal conversion from the Sn (n > 1) states to the S1 state in CPP-PP. To conclude, we synthesized the first example of a neutral spin-delocalized carbon-nanoring radical and demonstrated that the radially π-conjugated CPP framework is a unique structural platform to control spin distribution and reactivity of spin-delocalized systems such as phenalenyl. By means of SC-XRD, EPR and UV-Vis spectroscopy, and DFT calculations, we provided the first insight into the interplay of steric and electronic effects that govern spin distribution in this new type of open-shell nanocarbon hybrids. Our results lay the ground-work for future research towards the understanding of spin-delocalization through a radially π-conjugated backbone. For example, systems with a smaller ring size of the CPP segment are expected to favor a quinoidal over a benzenoid structure and thus lead to a more extended spin-delocalization. These concepts can be applied to understand through-space or guest-mediated spin interactions in multi-spin-unit and hostguest CPP systems, respectively. The highly selective dimerization of the nanoring radical to the double nanohoop represents a novel synthetic approach towards CPP architectures with unusual arrangements of π-surfaces such as the perpendicularly alternating "CPP-peropyrene-CPP" array mode. Our photophysical and DFT studies suggest that decreasing the size of the CPP segments in our double nanohoop, and thus lowering the HOMO-LUMO gap, could deliver true donor-acceptor systems. We believe that investigation of other members of this novel family of non-planar openshell hydrocarbon radicals will have implications in the fields of material science, spintronics and supramolecular chemistry. This work is ongoing in our laboratory.

chemsum

{"title": "Cycloparaphenylene-Phenalenyl Radical and Its Dimeric Double Nanohoop", "journal": "ChemRxiv"}

regioselective_simmons–smith-type_cyclopropanations_of_polyalkenes_enabled_by_transition_metal_catal

## Abstract: A [ iÀPr PDI]CoBr 2 complex (PDI ¼ pyridine-diimine) catalyzes Simmons-Smith-type reductive cyclopropanation reactions using CH 2 Br 2 in combination with Zn. In contrast to its non-catalytic variant, the cobalt-catalyzed cyclopropanation is capable of discriminating between alkenes of similar electronic properties based on their substitution patterns: monosubstituted > 1,1-disubstituted > (Z)-1,2disubstituted > (E)-1,2-disubstituted > trisubstituted. This property enables synthetically useful yields to be achieved for the monocyclopropanation of polyalkene substrates, including terpene derivatives and conjugated 1,3-dienes. Mechanistic studies implicate a carbenoid species containing both Co and Zn as the catalytically relevant methylene transfer agent. ## Introduction Cyclopropanes are common structural elements in synthetic and natural biologically active compounds. 1 The Simmons-Smith cyclopropanation reaction was frst reported over half a century ago but remains today one of the most useful methods for converting an alkene into a cyclopropane. 2 As compared to diazomethane, which is shock sensitive and must be prepared from complex precursors, CH 2 I 2 is both stable and readily available, making it an attractive methylene source. Additionally, the stereospecifcity of the Simmons-Smith reaction allows diastereomeric relationships in cyclopropanes to be established with a high degree of predictability. Several advances have addressed many of the limitations of the initial Simmons-Smith protocol. For example, Et 2 Zn can be used in the place of Zn to more reliably and quantitatively generate the active carbenoid reagent. 3 Acidic additives, such as CF 3 CO 2 H 4 and substituted phenols, 5 have been found to accelerate the cyclopropanation of challenging substrates. Finally, Zn carbenoids bearing dialkylphosphate anions 6 or bipyridine ligands 7 are sufficiently stable to be stored in solution at low temperatures (Fig. 1). Despite the many notable contributions in Zn carbenoid chemistry, a persistent limitation of Simmons-Smith-type cyclopropanations is their poor selectivity when attempting to discriminate between multiple alkenes of similar electronic properties. For example, the terpene natural product limonene possesses a 1,1-disubstituted and a trisubstituted alkene. Friedrich reported that, under a variety of Zn carbenoid conditions, the two alkenes are cyclopropanated with similar rates, resulting in mixtures of monocyclopropanated (up to a 5 : 1 ratio of regioisomers) and dicyclopropanated products. 8 This issue is exacerbated by the challenge associated with separating the two monocyclopropane regioisomers, which only differ in the position of a non-polar CH 2 group. In general, synthetically useful regioselectivities in Simmons-Smith reactions are only observed for substrates containing directing groups. 9 In principle, catalysis may provide an avenue to address selectivity challenges in Simmons-Smith-type cyclopropanations; however, unlike diazoalkane transfer reactions, which are catalyzed by a broad range of transition metal complexes, 9b,10 there has been comparatively little progress toward the development of robust catalytic strategies for reductive cyclopropanations. 11 Lewis acids in substoichiometric loadings have been observed to accelerate the Simmons-Smith reaction, but in many cases, this rate effect is restricted to allylic alcohol substrates. 12,13 Recently, our group described an alternative approach to catalyzing reductive cyclopropanation reactions using a transition metal complex that is capable of activating the dihaloalkane reagent by C-X oxidative addition. A dinickel catalyst was shown to promote methylene 14 and vinylidene 15 transfer using CH 2 Cl 2 and 1,1-dichloroalkenes in combination with Zn as a stoichiometric reductant. Here, we describe a mononuclear [PDI]Co (PDI ¼ pyridine-diimine) catalyst 16 that imparts a high degree of steric selectivity in the cyclopropanation of polyalkene substrates. Mechanistic studies suggest that the key intermediate responsible for methylene transfer is a heterobimetallic conjugate of Co and Zn. ## Results and discussion 4-Vinyl-1-cyclohexene contains a terminal and an internal alkene of minimal electronic differentiation and thus provided a suitable model substrate to initiate our studies (Table 1). 8,17 Under standard CH 2 I 2 /Et 2 Zn conditions (entry 1), there is a modest preference for cyclopropanation of the more electronrich disubstituted alkene (rr ¼ 1 : 6.7) with increasing amounts of competing dicyclopropanation being observed at higher conversions (entries 2 and 3). Other modifcations to the conditions, including the use of a Brønsted acid 4 (entry 4) or a Lewis acid additive 12b,18 (entry 5), did not yield any improvements in selectivity. Likewise, an Al carbenoid generated using CH 2 I 2 and AlEt 3 afforded a similar preference for cyclopropanation of the endocyclic alkene (entry 6). 19 In a survey of transition metal catalysts, the [ iPr PDI]CoBr 2 complex 1 was identifed as a highly regioselective catalyst for the cyclopropanation of 4-vinyl-1-cyclohexene, targeting the less hindered terminal alkene (Table 2). CH 2 Br 2 and Zn alone do not afford any background levels of cyclopropanation (entry 1); however, the addition of 6 mol% [ iPr PDI]CoBr 2 (1) provided monocyclopropane 3 (81% yield) with a >50 : 1 rr and <1% of the dicyclopropane product (entry 5). The steric profle of the catalyst appears to be critically important for yield. For example, the mesityl-(entry 6) and phenyl-substituted variants (entry 7) of the ligand provided only 58% and 4% yield respectively under the same reaction conditions. Related N-donor ligands similarly afforded low levels of conversion (entries 8-12) as did the use of other frst-row transition metals, including Fe (entry 14) and Ni (entry 15), in the place of Co. In order to defne the selectivity properties of catalyst 1, we next conducted competition experiments using alkenes bearing different patterns of substitution (Fig. 2). Reactions were carried out using an equimolar amount of each alkene and run to full conversion of the limiting CH 2 Br 2 reagent (1.0 equiv.). Monosubstituted alkenes are the most reactive class of substrates using 1 but are not adequately differentiated from 1,1-disubstituted alkenes (3 : 1). By contrast, terminal alkenes are signifcantly more reactive than internal alkenes, providing synthetically useful selectivities ($31 : 1). Furthermore, a model Z-alkene was cyclopropanated in preference to its E-alkene congener in a 33 : 1 ratio. Using catalyst 1, trisubstituted alkenes are poorly reactive, and no conversion is observed for tetrasubstituted alkenes. The synthetic applications of the catalytic regioselective cyclopropanation were examined using the terpene natural products and derivatives shown in Fig. 3. In all cases, the selectivity properties follow the trends established in the competition experiments. Substrates containing ether or free alcohol functionalities (e.g., 7, 10, and 11) exhibit a strong directing group effect under classical Simmons-Smith conditions; however, catalyst 1 overrides this preference and targets the less hindered alkene. Additionally, the presence of electron-defcient a,b-unsaturated carbonyl systems (e.g., 9, 13, and 14) do not perturb the expected steric selectivity. Vinylcyclopropanes are a valuable class of synthetic intermediates that engage in catalytic strain-induced ring-opening reactions. 20 The monocyclopropanation of a diene represents an attractive approach to their synthesis but would require a catalyst that is capable of imparting a high degree of regioselectivity and avoiding secondary additions to form dicyclopropane products. 21 These challenges are addressed for a variety of diene classes using catalyst 1 (Fig. 4). Over the substrates that we have examined, the selectivities for cyclopropanation of the terminal over the internal double bond of the diene system are uniformly high. Additionally, the catalyst is tolerant of vinyl bromide (15) and vinyl boronate (23) functional groups, which are commonly used in cross-coupling reactions. Like the non-catalytic Simmons-Smith reaction, 2c the cyclopropanation using 1 is stereospecifc within the limit of detection, implying a mechanism in which the two C-C s-bonds are either formed in a concerted fashion or by a stepwise process that does not allow for single bond rotation. For example, cyclopropanation of the Z-alkene 24 affords the cis-disubstituted cyclopropane 25 in 95% yield as a single diastereomer (Fig. 5a). Furthermore, the vinylcyclopropane substrates 26 and 28, commonly used as tests for cyclopropylcarbinyl radical intermediates, react without ring-opening to afford products 27 and 29 (Fig. 5b). a Reaction conditions: 4-vinylcyclohexene (0.14 mmol), THF (1.0 mL), 24 h, 22 C. Yields and ratios of regioisomers were determined by GC analysis against an internal standard. Under standard catalytic conditions, the reaction mixtures using 1 adopt a deep violet color, which persists until complete consumption of the alkene. The UV-vis spectrum of the catalytic mixture at partial conversion is consistent with a Co(I) resting state (Fig. 6a). The authentic [ iPr PDI]CoBr complex (30) can be prepared by stirring the [ iPr PDI]CoBr 2 complex 1 over excess Zn metal. 22 Cyclic voltammetry data (Fig. 6b) indicates an E 1/2 for the Co(II)/Co(I) redox couple of 1.00 V vs. Fc/Fc + . The large peak-to-peak separation (0.96 V in 0.3 M [n-Bu 4 N] [PF 6 ]/THF) is characteristic of a slow bromide dissociation step following electron transfer. The second Co(I)/Co(0) reduction event is signifcantly more cathodic at 1.93 V and is inaccessible using Zn. In order to decouple the cyclopropanation steps of the mechanism from catalyst turnover, we conducted stoichiometric reactions with the isolated [ iPr PDI]CoBr complex in the absence of Zn (Fig. 6c). The reaction of 30 with 4-vinylcyclohexene and CH 2 Br 2 generates the [ iPr PDI]CoBr 2 complex 1 within 24 h at room temperature but forms cyclopropanated products in a relatively low combined yield of 26%, which is not commensurate with the efficiency of the catalytic process. Furthermore, the regioselectivity is only 3 : 1, whereas the catalytic cyclopropanation achieves a >50 : 1 selectivity for this substrate. When the same stoichiometric reaction is conducted in the presence of ZnBr 2 , the yield and selectivity of the catalytic process is fully restored. The Co-containing product (31) of the stoichiometric reaction in the presence of ZnBr 2 is green, which is notably distinct from the tan color of the [ iPr PDI]CoBr 2 complex 1. This green species is NMR silent but may be crystallized from saturated solutions in Et 2 O to afford 31 (Fig. 6f). The solid-state structure reveals the expected [ iPr PDI]CoBr 2 fragment in a distorted square pyramidal geometry (s 5 ¼ 0.36) with a Zn(THF/Et 2 O)Br 2 Lewis acid coordinated to one of the Br ligands. This interaction induces an asymmetry in the structure, causing the Co-Br1 distance (2.557(1) ) to be elongated relative to the Co-Br2 distance (2.358(2) ). Collectively, these studies suggest that both Co and Zn are present in the reactive carbenoid intermediate, and that ZnBr 2 may interact with the [ iPr PDI]Co complex through Lewis acidbase interactions. There is a notable similarity between the observed Co/Zn effect and previous studies of Lewis acid acceleration in the Simmons-Smith cyclopropanation. For example, Zn carbenoid reactions are known to be accelerated by the presence of ZnX 2 , 12c which is generated as a byproduct of the reaction. DFT calculations conducted by Nakamura have suggested that the origin of this rate acceleration may be due to the accessibility of a fve-membered ring transition state, which requires the presence of an additional Zn equivalent to function as a halide shuttle. 23 ## Conclusions In summary, transition metal catalysis provides a pathway to accessing unique selectivity in reductive carbenoid transfer reactions. A [ iPr PDI]CoBr 2 complex functions as a robust catalyst for Simmons-Smith type cyclopropanation using a CH 2 Br 2 /Zn reagent mixture. This system exhibits the highest regioselectivities that have been observed in reductive cyclopropanations based solely on the steric properties of the alkene substrate. Accordingly, a range of terpenes and conjugated dienes may be converted to a single monocyclopropanated product. Ongoing studies are directed at exploring the applications of transition metal catalysts to other classes of carbenoid transfer reactions.

chemsum

{"title": "Regioselective Simmons\u2013Smith-type cyclopropanations of polyalkenes enabled by transition metal catalysis", "journal": "Royal Society of Chemistry (RSC)"}

homochiral_nanotubes_from_heterochiral_lipid_mixtures:_a_shorter_alkyl_chain_dominated_chiral_self-a

## Abstract: It is an important topic to achieve homochirality both at a molecular and supramolecular level. While it has long been regarded that "majority rule" guides the homochiral self-assembly from an enantiomer mixture, it still remains a big challenge to manipulate the global homochirality in a complex system containing chiral species that are not enantiomers. Here, we demonstrate a new example wherein homochiral nanotubes self-assembled from a mixture of heterochiral lipids that deviated from the "majority rule". We have found that when two heterochiral lipids with mirror headgroups but a 2-methylene discrepancy in alkyl chain length are mixed, homochiral nanotubes are always formed regardless of their mixing ratio. Remarkably, the helicity of the nanotube is exclusively controlled by the molecular chirality of the lipids with shorter alkyl chains, i.e., the chiral self-assembly was dominated by the lipid with the shorter alkyl chain. MD simulation reveals that the match of both the alkyl chain length and hydrogen-bonding between two kinds of lipids plays an important role in the assembly. This work provides a new insight into the supramolecular chirality of complex systems containing multi chiral species. ## Introduction Homochirality in living organisms, i.e. almost all of the amino acids and sugars are L-and D-enantiomers, respectively, is one of the most mysterious phenomena and has attracted long-term interest in biology, chemistry, physics and material science. Such molecular homochirality in the biological system requires the design of drug molecules as a single enantiomer, 13 which is suggested to be related to the different interactions between proteins and enantiomers of drug molecules. 14,15 Thus, the homochirality issue is extended to a supramolecular level such that the stereochemical communication or chiral-chiral interaction between various chiral species becomes vitally important. So far, two important rules on stereochemical communication, the "majority rule" and "sergeant-andsoldiers rule", 27,31, have been well-established with respect to covalent and non-covalent bonding of chiral polymers or supramolecular assemblies. Generally, the "majority rule" is related to two chiral molecules with mirrored confguration and states that the global chirality of the system is always determined by the chirality of the excess enantiomeric species. The "sergeants-and-soldiers rule" deals with the interaction between chiral sergeants and achiral soldiers and states that the chirality of the whole system follows the chirality of the sergeant. However, there is still a big challenge to manipulate the interaction or communication between different chiral species in complex systems, 3, such as chiral lipids with different chain lengths in a biological membrane, where the chiral species are not necessarily in exact mirror confgurations. Here, we designed a series of enantiomeric glutamide lipids with various alkyl chain lengths and investigated their selfassembly behaviours (Fig. 1). Absolutely mirrored heterochiral lipid mixtures are found to follow the "majority rule", i.e. the majority enantiomers control the global chirality of the system and the racemate is often achiral. However, when two heterochiral lipids with mirror headgroups but a 2-methylene discrepancy in alkyl chain length were mixed, a homochiral composite nanotube was always obtained. Remarkably, the helical sense was not determined by the majority component but by the lipids with the shorter alkyl chain no matter how small the amount of that lipid. This phenomenon deviates from the reported stereochemical communication rules and has never been reported before. It demonstrates that a small variation in molecular structure also plays an important role in stereochemical communication apart from intrinsic molecular chirality. By combining various experimental characterization methods and theoretical molecular dynamics (MD) simulation, the mechanism of this unprecedented phenomenon is disclosed. ## Results and discussion Lipid molecule design and synthesis N,N 0 -bis(alkyl)-D/L-glutamic diamide lipids, with enantiomerically pure glutamic acid as the polar headgroup and double hydrophobic nonpolar alkyl tails, were designed to mimic natural amphiphilic chiral lipids with different chain lengths (Fig. 1). The lipid molecules were synthesized by two simple steps, as previously reported: 55 the tert-butoxycarbonyl (Boc)protected D/L-glutamic acid was frstly connected to two equimolar alkyl amines, then the Boc group was eliminated to free the polar amine headgroup. ## Self-assembly of heterochiral lipid mixtures The self-assembly of the lipids was all performed in ethanol medium through a heat-and-cooling gelation process. Briefly, lipids or their mixtures were dispersed into ethanol at room temperature and then heated to a transparent solution. After the solution was cooled down to room temperature, the gel was formed. All the lipids as well as their mixtures could form white opaque gels and self-assembled into well-defned nanostructures upon gelation (see Experimental section for details). Fig. 1 Self-assembly of chiral lipids. (a) Enantiomerically pure D-and L-lipids form M-and P-helices, respectively. (b) Mixing of racemates follows the "majority-rule". However, mixing of two heterochiral lipids with mirror chiral head groups but a 2-methylene discrepancy in alkyl chain length leads to the homochiral composite nanotube, whose helical sense is exclusively determined by the molecular chirality of the lipid with the shorter alkyl chain regardless of their mixing ratios. Characterization of the self-assembled nanostructures from heterochiral lipid mixtures Fig. 2 shows the representative morphologies of the nanostructures, and three important features can be found. First, all the pure enantiomers formed chiral nanotubes with helicity following their molecular chirality, regardless of their chain length, i.e., D-and L-lipids produced left-and righthanded nanotubes, respectively (Fig. 2a and b). Second, when two opposite enantiomeric lipids with absolute mirror-confguration (n ¼ m) such as 20D/20L, 18D/18L, and 16D/16L were mixed, they obeyed the "majority rule", i.e., the helicity was determined by the excess enantiomeric lipid. In particular, a planar nanosheet without any chirality was formed for an equimolar mixture (Fig. 2a and b). Third, when two pseudo-enantiomeric heterochiral lipids, i.e., with opposite chiral head groups and a 2-methylene discrepancy in chain lengths, such as the combinations of 20L/18D, 20D/18L, 18D/16L, 18L/16D, 16L/14D, and 16D/14L, were mixed, helical nanotubes were exclusively formed at various mixing ratios, even for equimolar mixtures (Fig. 2c-e, S1 and S2 †). In this case (n-m ¼ 2 system), the "majority rule" is no longer operative. In order to elucidate these new observations, various characterization methods, such as XRD, FTIR spectroscopy, CD spectroscopy and DSC thermal analysis, were carried out. Hereafter, the self-assembly of the 18D/18L and 18D/16L systems will be studied as an example. FTIR spectra are powerful in discriminating molecular interactions. As shown in the FT-IR spectra (Fig. S3 †), all the nanostructures showed obvious H-bonded vibrations from N-H, amide I and amide II. However, their precise vibrations are different for the different lipid mixtures. The N-H, amide I and amide II bands at 3326, 1636, and 1531 cm 1 for the 18D (18L) nanotube shifted to 3302, 1633, and 1544 cm 1 for the 18D/18L nanosheet, indicating that the 18D/18L nanosheet has stronger hydrogen bonding interactions than that of either the 18D or 18L nanotube 55 (Fig. 3a and S3, Table S1 †). This was further confrmed by DSC thermogram analysis of the 18D/18L nanostructures (Fig. 3b and S4 †), where the phase transition temperature (T m ) is ca. 121 C regardless of the mixing ratio, indicating the miscible nature of the 18D and 18L lipids. 56 This means that the nanoscale chirality is counterbalanced at a molecular level. 55 Consequently, the helical torsion force in the racemate bilayer is decreased, which is evidenced by the dspacing expansion of the racemate bilayer (4.85 nm, equimolar 18D/18L) compared to the enantiomerically pure bilayers (4.23 nm, 18D) (Fig. 3c). Therefore, achiral planar nanosheets are produced for 18D/18L at an equimolar ratio. In contrast, the FTIR spectra showed scarcely any change of the hydrogen bonding interaction in all 18D/16L combinations compared to the 18D or 16L nanotubes (Fig. 3a and S3, Table S2 †). It seems that the helical torsion force in the heterochiral bilayer of 18D/16L is unaffected. Therefore, the nanotube rather than the planar sheet formed for all heterochiral 18D/16L combinations. However, only one T m peak was found in the DSC thermograms of the 18D/16L nanotubes (Fig. S4 †) and the plot of T m value to mixing ratio is a U-shaped curve (Fig. 2b), with a T m value of 115 C for equimolar 18D/16L lower than those of either 18D (121 C) or 16L (119 C), suggesting the mutual diluent effect and co-self-assembly 30 of 18D and 16L. 56 Moreover, the XRD patterns (Fig. 3c) showed a single bilayer (4.08 nm) just between that of 18D (4.23 nm) and 16L (3.92 nm), further suggesting the co-assembly of all 18D/16L combinations. It should be noted that if two respectively self-assembled nanotubes were mixed, we can observe two sets of peaks (Fig. 3c). Therefore, we can conclude that when 18D and 16L were mixed, they tended to co-assemble rather than self-sort. ## Supramolecular chirality of the composite nanotubes from heterochiral lipid mixtures Given that two opposite chiral lipids are involved in the n-m ¼ 2 system, the supramolecular and nanoscale chirality of the composite nanotubes is alluring. High-resolution SEM images (Fig. 4a-f) show that the composite nanotube is chiral at the nanoscale. Moreover, the chirality is exclusively one handed, which is always consistent with that of the nanotubes formed from the shorter lipids alone. Specifcally, the composite 18D/ 16L nanotubes are always right-handed (Fig. 4b-f) like the 16L nanotube and the 18L/16D nanotubes are left-handed (Fig. 4a and S5 †) like the 16D nanotube. Obviously, the helicity of the composite nanotubes from heterochiral lipids is basically determined by the molecular chirality of the shorter lipids. The helicity of the nanotubes was further investigated by CD spectroscopy. Since these lipid molecules do not possess any chromophore, an achiral dye, meso-tetra(4-sulfonatophenyl) porphyrin (TPPS), was used as a probe to reflect the helicity of the nanotube through aggregation on the surface of the nanotubes (Fig. 4g, h and S6 †). UV/Vis spectra displayed two strong bands at 493 and 708 nm, indicating induced Jaggregation of TPPS at the surface of all nanotubes. 57 The CD spectra of the D-lipids displayed two strong Cotton effects at 495() and 486(+) with a crossover at 490 nm, and 430() and 415 (+) with a crossover at 422 nm, while the L-lipids showed mirrored Cotton effects to those of the D-lipids, which reflected the chiral packing manner of the lipids at the surface of the nanotubes, i.e. an M-helix for D-lipids and P-helix for L-lipids. Both 18D/18L and 16D/16L were CD silent, indicating achiral packing at the surface of the planar nanosheets. On the other hand, 18L/16D and 18D/16L displayed strong negative and positive Cotton effects, respectively. Once 16L was involved in the system, 18D/16L exclusively showed positive CD signals regardless of the molar ratios of 16L to 18D (Fig. 4h). The CD results are well consistent with the SEM observations, indicating that the heterochiral lipid nanotubes are globally homochiral and that the helicity is essentially determined by the lipids with the shorter alkyl chain. Theoretical analysis and molecular dynamics (MD) simulation To further disclose the unprecedented phenomenon and deeply understand the chiral self-assembly process, theoretical analysis was carried out via MD simulation. According to the previous theoretical studies, 64,65 the handedness of aggregates is dependent on the molecular orientation, which is actually the orientation of amide groups in the lipid molecules here. Besides, the alkyl chain should be matched to maintain the bilayer. Therefore, we mainly focus on the alkyl chain length match and the orientation of amide groups to analyze the handedness of the heterochiral lipid bilayer. There are two amide groups in both 16L and 18D. The a-amide and amino groups can form an intramolecular hydrogen bond, which induces the a-amide to produce an orientation, while the direction of the g-amide is uncertain. The interaction of the oppositely chiral headgroups leads to a ca. 90 difference in the directors (d) of the a-amide groups in 16L and 18D (Fig. 5a, b and S8 †). The theoretical studies by Selinger et al. showed that rotating the tilt direction by 90 should change the curvature direction by 90 , giving a handedness reversal. 64 Therefore, the different chirality of 16L and 18D bilayers can be easily understood. Planar 16L (18D) bilayer aggregation has two different stacking manners (Fig. S9 †). However, the aggregation with C2 symmetry (essentially the orientations of a-amides on two sides) where the rotation axis lies along the bilayer aggregation direction will lead to damage of the bilayer structure after MD simulation. Only the pre-assembly aggregation with C2 symmetry where the rotation axis lies perpendicular to the bilayer plane can result in chiral bilayer structures. For obtaining the chiral structures of the pure 16L and 18D systems, we built planar bilayer aggregates containing two layers and a total of 120 molecules with a 3.6 d-space for MD simulations. After the equilibriums were reached, we sampled one snapshot per 1 ps and extracted the average confgurations during 5.5-6 ns for the pure 16L and 18D systems. It was found that the 16L molecules form a P-helix bilayer structure (Fig. 5a), while the 18D molecules form an M-helix (Fig. 5b). For the 16L/16D mixture with a 1 : 1 ratio, the lengths of alkyl tails perfectly match with each other and the intermolecular hydrogen bonds can form between a-amide groups. Moreover, the a-amide orientations of 16L and 16D are perpendicular to each other, fnally resulting in the achiral nanosheet structure. However, in the 16L/18D mixture, a conformation rearrangement on the molecular structure of 18D happened due to the existence of 16L. As presented in Fig. 5c, when the a-amide in 16L was connected to the g-amide in 18D, and the g-amide in 16L was connected to the a-amide in 18D, the length of the alkyl chains between the two molecules could be perfectly matched. In this situation, the orientation of the a-amide in 18D was lost, while the orientation of the g-amide in 18D was induced and it pointed in the same direction as that of the a-amide in 16L. Hence, for further study on the 16L/18D aggregate by MD simulation, we built a pre-assembly bilayer with a planar structure containing two layers and a total of 120 molecules (16L/18D ¼ 1/1). As with the pure systems, the orientations of the amides on both sides of 16L/18D should keep C2 symmetry where the rotation axis is perpendicular to the bilayer plane. After the equilibrium was reached, we also sampled one snapshot per 1 ps and extracted the average confguration during 5.5-6 ns for the 16L/18D system. It was found that a P-helix was achieved for the 16L/18D aggregate. The MD simulation is consistent with the experimental results, and well explains the unprecedented phenomenon. ## Conclusions In summary, the self-assembly behaviors of two heterochiral lipids and their mixtures were systematically investigated (Fig. 6). For individual chiral lipid self-assembly, the intramolecular hydrogen bond between the a-amide and amino groups induces the a-amide to produce an orientation, and the oppositely chiral headgroups cause a ca. 90 difference in the directors of the a-amide groups. Consequently, L-lipids always form P-helical nanotubes and D-lipids form M-helical nanotubes. For the absolutely mirrored heterochiral lipid mixtures (n ¼ m system), the lengths of the alkyl tails can perfectly match with each other and intermolecular hydrogen bonds can form between a-amide groups. In the composite bilayer, the a-amide orientations of the L-lipids and D-lipids are perpendicular to each other, fnally resulting in the achiral nanosheet structure. For the two heterochiral lipids with mirror headgroups but a 2-methylene discrepancy in alkyl chain length (n-m ¼ 2 system), under alkyl chain communication, the conformation of the longer lipids is rearranged in order to match the shorter lipids (Fig. 6). Consequently, the a-amide of the short lipids was connected to the g-amide of the longer lipids, and the g-amide of the shorter lipids was connected to the a-amide of the longer lipids. In this situation, the alkyl chain length between the two lipids could be perfectly matched. Moreover, the orientation of the g-amide of the longer lipids was induced and pointed in the same direction as that of the a-amide of the shorter lipids, while the orientation of the a-amide of the longer lipids was lost. Finally, the alkyl chain packing, hydrogen-bonding connection and orientation of the two lipids were perfectly matched. Thus, globally homochiral nanotubes are produced and the helicity of the heterochiral lipid nanotube is exclusively determined by the As for 18D/16L (c), the shorter lipid 16L induced a conformation rearrangement of the longer lipid 18D, leading to the disappearance of the orientation of the a-amide and an induced orientation of the g-amide in 18D, which is the same as that of the aamide in 16L. In this case, the alkyl chains between the two lipids are also perfectly matched. Therefore, a P-helical bilayer was achieved for the 18D/16L heterochiral lipid mixture like that of pure 16L. Fig. 6 "Induced conformation rearrangement" mechanism of homochiral nanotube from heterochiral lipids. Alkyl chain communication between heterochiral lipids induced the conformation of the longer lipids to rearrange in the presence of the shorter lipids. Consequently, the orientation of the g-amide of the longer lipids was induced and pointed in the same direction as that of the a-amide of the shorter lipids, while the orientation of the a-amide was lost. Finally, the alkyl chain packing, hydrogen-bonding connection and orientation of the two lipids were perfectly matched. Thus, globally homochiral nanotubes were produced and the helicity was exclusively determined by the molecular chirality of the shorter lipids. This journal is © The Royal Society of Chemistry 2019 Chem. Sci., 2019, 10, 3873-3880 | 3877 molecular chirality of the shorter lipids. The "induced conformation rearrangement" mechanism well interpreted the formation of the homochiral nanotube from heterochiral lipid mixtures regardless of the mixing ratio. The present contribution sheds new light on the understanding of homochirality at a supramolecular and nanoscale level in complex lipid systems and provides new guidance in exploring homochiral materials in complex supramolecular systems. 3, Experimental Self-assembly procedure For the self-assembly of pure lipids: the lipid solids (3 10 5 mol) were put into a seal-capped vial with 1 mL of ethanol added (0.03 M). Then, the sample vial was heated up to 75 C for a while to make a clear solution and subsequently allowed to cool down to room temperature naturally (25 C, cooling rate was about 10 C min 1 ). White gels were obtained, which were fully aged for 12 hours under ambient conditions before being measured. For the self-assembly of mixed lipids: the required amount of D-and L-lipids was mixed at a specifc proportion in one sample vial and 1 mL of ethanol was added (the total concentration was kept at 0.03 M). Then, the sample was treated using the above procedure. Characterization SEM and TEM. The fully aged gel was transferred from a sample vial to single-crystal silica wafers with a thin flm of Pt coating for SEM observation and to carbon-coated Cu grids stained with 2% uranyl acetate (wt%, aqueous, about 2 min) for TEM observation; images were taken using a Hitachi S-4300 or S-4800 FE-SEM (15 kV) and a JEM-2010 (200 kV), respectively. XRD. The quartz-plate-sustained xerogel flms of selfassembled lipids or lipid mixtures were used for XRD measurements on a Rigaku D/Max-2500 X-ray diffractometer (Japan) with Cu/Ka radiation (l ¼ 1.5406 , 40 kV, 200 mA). For t18D/t16L, two respectively pre-self-assembled nanotubes of 18D lipid and 16L lipid were vacuum-dried, and the solids were mixed and ground with an agate mortar and pestle for fully mixing. FTIR spectroscopy. KBr pellets of vacuum-dried xerogels were prepared for Fourier-transform infrared (FTIR) spectral measurements on a Bruker Tensor 27 FTIR spectrometer (resolution: 4 cm 1 ). DSC. The vacuum-dried self-assembled solids (3-5 mg) of pure and mixed lipids were recorded on a METTLER TOLEDO DSC882e to obtain DSC thermograms in a nitrogen atmosphere at a heating rate of 5 C min 1 from 35 to 135 C. For thermal analysis of the mechanical mixture of 18D and 16L nanotubes (18D ¼ 16L), the dried solids of respectively pre-self-assembled 18D nanotubes (2.17 mg) and 16L nanotubes (1.98 mg) were directly added into the sample pan and measurements were performed under the same conditions as described above. UV/Vis and CD. A 10 3 M aqueous solution of TPPS (tetrakis(4-sulfonatonphenyl)porphine, Dojindo Laboratories) was prepared and divided into several aliquots in which the asprepared lipid gels were added. The mixtures were gently shaken for a while and settled overnight under ambient conditions for full absorption of TPPS on the surface of the nanostructures. The excess TPPS in the aqueous solutions was removed using a centrifuge (Anke TGL-16C, Shanghai) at 6000 rpm for 5 min. The green sediments were dispersed in water and re-centrifuged several times until the supernatant liquid was colourless. After that, the sediments were dispersed into 3 mL of aqueous hydrochloric acid (0.1 M) and then centrifuged to remove residual acid. Finally, the sediments were re-dispersed into methanol for UV/Vis and CD spectral measurement on a JASCO UV-550 and J-815 CD spectrophotometer, respectively. MD simulation. The pre-assembly aggregates of bilayers were solvated in H 2 O boxes with sufficient capacity by the PACKMOL program. Then, MD of solution systems was performed within the NPT ensemble (constant number of atoms, pressure, and temperature) in GROMACS-4.6.7. A Berendsen thermostat with a time-step of 1 fs was employed to regulate the temperature at 298 K. All simulations were carried out for 6 ns to achieve a fully relaxed confguration by using the General Amber Force-Field (GAFF). ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Homochiral nanotubes from heterochiral lipid mixtures: a shorter alkyl chain dominated chiral self-assembly", "journal": "Royal Society of Chemistry (RSC)"}

gas_adsorption_selectivity_in_topologically_disordered_metal-organic_frameworks

## Abstract: Disordered metal-organic frameworks are emerging as an attractive class of functional materials, however their applications in gas storage and separation have yet to be fully explored. Here, we investigate gas adsorption in the topologically disordered Fe-BTC framework and its crystalline counterpart, MIL-100. Despite their similar chemistry and local structure, they exhibit very different sorption behaviour towards a range of industrial gases, noble gases and hydrocarbons. Virial analysis reveals that Fe-BTC has enhanced interaction strength with guest molecules compared to MIL-100. Most notably, we observe striking discrimination between the adsorption of C3H6 and C3H8 in Fe-BTC, with over a twofold increase in the amount of C3H6 being adsorbed than C3H8. Thermodynamic selectivity towards a range of industrially relevant binary mixtures is probed using ideal adsorbed solution theory (IAST). Together, this suggests the disordered material may possess powerful separation capabilities that are rare even amongst crystalline frameworks. ## Introduction Metal-organic frameworks (MOFs) are hybrid materials known for their chemical and structural diversity. 1 With surface areas reported in excess of 7,000 m 2 g −1 , they are emerging as promising candidates for gas storage. 2,3 Furthermore, their pore architectures, tuneable between 3 and 100 in diameter, are well suited for various separation processes, such as industrial gases and hydrocarbons. MOFs are capable of gas separation through thermodynamic, kinetic, molecular sieving, and even quantum effects. 11 Separation depends on many properties of the framework; surface area, pore size, surface chemistry, the presence of functional groups, unsaturated metal sites, extra-framework ions and water molecules. 5 The kinetic diameter, polarizability and permanent polarity of the guest also play an important role, complicating the prediction of porous material's separation capabilities prior to measuring adsorption isotherms. The separation of industrial gases, specifically the capture of CO2 from CO2/N2 or CO2/CH4 mixtures, is increasing in importance given the environmental impact of the rising global greenhouse gas emissions. 12 Another key area is the petrochemical industry, which relies on the separation of hydrocarbons for use as fuels and chemical feedstocks for the production of polymers. 7 Separation of hydrocarbons of the same chain length is not easily achieved given their similar physical properties. For example, the boiling points of C2H4 and C2H6 differ by 15 K, while for C3H6 and C3H8 the difference is only 5 K. 13 Coupled with a molecular size difference of less than 0.4 , C3H6/C3H8 separation is a very challenging and extremely energy-intensive process. 7 Very few crystalline MOFs have been found to successfully discriminate between the two gases. Developing MOFs for these applications requires consideration of not only the selectivity itself but also the chemical stability, recyclability, handling stability and capacity of the material, which are all important for practical use within an industrial setting. 14,15 MOFs are particularly attractive candidates due to the intricate ways in which their pore architecture and surface chemistry can be modified to tune the interactions with guest molecules. 16 MIL-100 is a crystalline MOF, comprising Fe(III) centres and 1,3,5-benzenetricarboxylate linkers, arranged into oxo-centred trimer motifs that further assemble into hybrid tetrahedra [Figs. 1a & b]. 17 Its hierarchical structure has a specific surface area in excess of 2,000 m 2 g −1 and contains two types of mesopore with internal diameters of 25 and 29 , accessible through 5.5 and 8.6 apertures, respectively [Fig. 1c]. The crystallographic unit cell has a cell volume in excess of 390,000 3 [Fig. 1d]. Activation of MIL-100 at temperatures below 150 °C generates coordinatively unsaturated Fe(III) sites, while temperatures above 150 °C induce partial reduction to form unsaturated Fe(II) sites, which may lead to enhanced interactions with certain adsorbates. 18 These unsaturated metal sites are effective in aiding the separation of light hydrocarbons, with interaction strengths following the order C2H2 > C2H4 > C2H6 > CH4. 19 A similar increase in interaction strength was observed between MIL-100 and C3H6 compared to C3H8 upon generation of unsaturated Fe(II) sites through activation, suggesting the potential for C3H6/C3H8 separation using materials activated at temperatures above 150 °C. 20 MIL-100 has also found applications in water harvesting and as a Friedel-Crafts reaction catalyst. 21 Amorphous MOFs share the local structure of their crystalline counterparts but lack any long-range order. 22,23 Typically, they are obtained via the collapse of a crystalline material through the application of heat or pressure. Structural disorder within MOFs has been demonstrated to improve conductivity, mechanical response, and mass transfer capacity compared to crystalline frameworks. 24 Notable examples of amorphous MOFs are found within the zeolitic imidazolate framework (ZIF) family, such as the hybrid glass agZIF-62 [Zn(Im)2(bIm)2-x; Im -imidazolate, bIm -benzimidazolate; ag denoting the glass state], which is obtained through the meltquenching of crystalline ZIF-62. 25 Recently, mechanical amorphisation of ZIF-8 [Zn(mIm)2; mIm -2-methylimidazolate], to form amZIF-8 (am denoting amorphisation by ball milling), was found to enhance the adsorption selectivity of C3H6 over C3H8. 26,27 Despite this, investigations into disordered MOFs for applications in gas storage and separation are of comparatively low prevalence with respect to their crystalline counterparts, particularly in the area of C3H6/C3H8 separation. Fe-BTC, known commercially as Basolite® F300, has the same chemical composition as MIL-100 and is topologically disordered, lacking long-range order [Fig. 1e]. 28,29 X-ray absorption near edge structure analysis of Fe-BTC revealed the presence of octahedrally coordinated Fe(III) ions at ambient temperature. 30 Furthermore, analysis of the extended X-ray absorption fine structure region revealed Fe-BTC possessed the same trimer unit structure present in MIL-100. Synchrotron X-ray pair distribution function analysis revealed the presence of mixed hierarchical local structure within Fe-BTC, confirming the trimer unit's existence and the presence of tetrahedral assemblies as observed in MIL-100. 29 This was used to produce the first atomic-scale model of Fe-BTC. Upon activation at 120 °C, removal of water molecules coordinated to the Fe(III) ions in the trimer unit leads to the formation of coordinatively unsaturated metal sites. 30 This causes a lowering of the octahedral symmetry but does not induce the structural rearrangement of the framework. Upon activation at the slightly higher temperature of 150 °C, a small proportion of unsaturated Fe(II) sites were detected using NO-probed infrared spectroscopy in Fe-BTC and MIL-100. 31 At 250 °C the proportion of Fe(II) sites was greater in MIL-100 than Fe-BTC; beyond this temperature, Fe-BTC began to show signs of decomposition. Fe-BTC has been studied for its catalytic ability and outperforms MIL-100 in Lewis acid catalysis. 32 This has been attributed to its unsaturated metal sites and additional Brønsted acid sites, which are likely to influence its interactions with guest molecules. 31,33 The morphological and porous nature of Fe-BTC is highly synthesis dependent. Often obtained via a sol-gel route, the Fe-BTC gel can subsequently be dried (i) through exchange with supercritical CO2 (sCO2) to obtain hierarchically porous aerogels, (ii) at room temperature for several days to afford xerogels, or (iii) at higher temperatures overnight to produce powdered samples. 28 The use of sCO2 exchange avoids destructive capillary forces, which cause the collapse of the hierarchically porous architecture. Aerogel samples of Fe-BTC possessed a total pore volume of 5.62 cm 3 g −1 and a single point Brunauer-Emmett-Teller (BET) surface area of 1,618 m 2 g −1 . Xerogel samples had a BET surface area of around 800 m 2 g −1 that could be increased to 1,182 m 2 g −1 with a total pore volume of 0.71 cm 3 g −1 through the ageing of the gel before drying. The powdered sample of Fe-BTC was least porous but was easiest to prepare. Quenched solid density functional theory analysis of the N2 adsorption isotherm revealed that the xerogel had a broad pore size distribution in both the micro-and mesopore range up to 4.5 nm with maxima at 1.3 and 3.0 nm. The aerogel contained an even broader distribution in the whole range of micro-and mesopores with the same maximum at 1.3 nm. 28 More recent investigations, using the harsher drying conditions employed in this study, gave rise to powdered Fe-BTC samples that were essentially non-porous to nitrogen at 77 K, yet still retained the same local structure as MIL-100. 29 Basolite® F300, whose exact synthetic route remains undisclosed, has a BET surface area in the range 1,300 to 1,600 m 2 g −1 as reported by the manufacturers. Independent experimental measurements have reported the BET surface area to be around 685 to 840 m 2 g −1 , with a total pore volume of 0.29 cm 3 g −1 and pore size of 2.2 nm. 31,34 Computational modelling has suggested that the degree of tetrahedral assembly in Fe-BTC materials influences the porosity, with accessible surface area increasing with the proportion of tetrahedral assemblies. 29 One atomic-scale model of Fe-BTC, for example, contained appreciable internal porosity while remaining non-porous to a nitrogen-sized probe. This model demonstrated the successful percolation of a 3.19 diameter spherical probe through the structure and revealed a maximum spherical cavity size of 9.15 . This analysis was performed Please do not adjust margins Please do not adjust margins from a purely geometric perspective and did not consider dynamics within the real material. This suggests that Fe-BTC may have adsorption capacity for other gases given the dynamic nature of the material under realistic conditions. Given the existing promise of MIL-100 in gas sorption and separation, the absence of long-range order in Fe-BTC means it is well placed to explore how topological disorder affects sorption and selectivity in MOFs. The area of C3H6/C3H8 separation is particularly interesting, given the limited number of crystalline MOFs reported with this behaviour. Ultimately, we aim to demonstrate the importance of disorder as a tool to augment and enhance properties of MOFs in the field of gas sorption and separation. ## Materials All chemicals were obtained from commercial suppliers and used as received. Iron (III) nitrate nonahydrate (99.95%), 1,3,5-benzenetricarboxylic acid (95%), methanol (99.8%), ethanol (99.8%), ammonium fluoride (99.99%), sodium hydroxide pellets (98%) and iron (II) chloride tetrahydrate (99.99%) were all purchased from Sigma Aldrich. Ultrahigh purity gases were used as received from BOC Gases. ## MIL-100 MIL-100, Fe3(OH)(H2O)2O[(C6H3)(CO2)3]2.nH2O, was synthesised following the procedure in Ref. 35. 1,3,5-benzenetricarboxylic acid (1.676 g), dissolved in 1 M aqueous sodium hydroxide (23.72 g), was added dropwise to a solution of iron (II) chloride tetrahydrate (2.260 g) dissolved separately in water (97.2 mL). The green suspension was left to stir at room temperature for 24 hours. The product was recovered by centrifugation, washed thoroughly with ethanol (3×20 mL), and dried overnight at 60 °C. The orange powder was purified following Ref. 36. Briefly, the powder was dispersed and heated for 3 hours in each water (700 mL at 70 °C), ethanol (700 mL at 65 °C) and 38 mM aqueous ammonium fluoride solution (700 mL at 70 °C). The powder was recovered between each stage by centrifugation. The final product was dried overnight at 60 °C. ## Fe-BTC Fe-BTC was synthesised following the procedure in Ref. 29. Both iron (III) nitrate nonahydrate (2.599 g) and 1,3,5benzentricarboxylic acid (1.177 g) were dissolved in methanol (20 mL each). The two solutions were combined at room temperature and left to stir for 24 hours, forming a viscous orange solution. This was washed with ethanol (3×20 mL) before drying overnight at 60 °C. The powder was then purified as described above and left to dry overnight at 60 °C. ## Powder X-ray Diffraction Powder X-ray diffraction data were collected at room temperature using a Bruker D8 diffractometer using Cu Kα1 (λ = 1.5406 ) radiation and a LynxEye position-sensitive detector with Bragg-Brentano parafocusing geometry. Samples of finely ground powder were dispersed onto low-background silicon substrates and loaded onto the rotating stage of the diffractometer. Data were collected over the angular range 2° < 2θ < 50°. Pawley refinements were carried out using TOPAS Academic (V6) software. 37 The unit cell parameters were refined against those previously reported for MIL-100. 17 A modified Thompson-Cox-Hasting pseudo-Voigt peak shape and simple axial divergence correction were employed. ## Gas Sorption Gas adsorption isotherms and kinetic profiles were measured using a Quantachrome iQ2 instrument. Prior to measurement, samples were degassed at 150 °C for 12 hours. Sample masses were measured using degassed samples after the sample tube was backfilled with N2. Sample temperatures were accurately equilibrated at 273 K and 293 K using a temperature-controlled water bath and at 77 K using a Dewar filled with liquid N2. Under these conditions, MIL-100 and Fe-BTC are expected to contain coordinatively unsaturated Fe(III) sites with little to no partial reduction to Fe(II) occurring. See Supplementary Methods for details on the surface area, non-local density functional theory, virial, and ideal adsorbed solution theory analyses. ## Structural Characterisation and Nitrogen Adsorption Samples of MIL-100 and Fe-BTC were prepared following previously reported procedures (see Experimental Methods for details). Powder X-ray diffraction measurements confirmed the crystalline nature and phase purity of MIL-100 [Figs. S1]. The diffraction pattern for Fe-BTC did not contain sharp Bragg scattering. Instead, very broad regions of weak scattering were observed, consistent with its lack of Please do not adjust margins Please do not adjust margins long-range order yet possession of trimer-based local structure similar to that present in MIL-100. 29 The N2 adsorption isotherm for MIL-100 at 77 K displayed the expected intermediate Type I and IV behaviour indicative of the presence of both micro-and mesopores with a secondary uptake at approximately 0.12 P/P0, a signature of the dual-pore architecture [Fig 3]. 17 The BET surface area was 1,465 m 2 g −1 , with a maximal uptake of 479.4 cm 3 g −1 [Table S2]. Fe-BTC, however, exhibited an almost negligible maximal uptake of 42.1 cm 3 g −1 and hence the BET surface area (68 m 2 g −1 ) cannot readily be regarded as reliable. ## & S1, & Table Non-local density functional theory can be used to extract the pore size distribution from an adsorption isotherm [Figs. S2 & S3]. This approach can be informative, despite the limitations discussed in the Supplementary Methods. The pore size distribution for MIL-100 had maxima at 9.5 and 16.9 , while Fe-BTC possessed maxima at 9.1 and 17.3 [Figs. S4 & S5]. The pores in Fe-BTC were present in a significantly lower amount than in MIL-100, hence the lower porosity of Fe-BTC. Notably, the contribution of the larger pore cavity was appreciably less in Fe-BTC than in MIL-100, suggesting the porous interior of Fe-BTC is largely comprised of the smaller pore structure. This pore cavity has a smaller crystallographic window aperture of 5.5 (c.f. 8.6 aperture in the larger pore). It is the ordered network structure of MIL-100 that facilitates uptake and diffusion of guest molecules through the pores and connected apertures. In contrast, the topologically disordered nature of Fe-BTC disrupts the accessibility of the pores and results in a decrease in accessible surface area towards N2 at this temperature. This has also been observed in ZIFs and their melt-quenched glass counterparts. 38 These results suggest that the porous interior of Fe-BTC is inaccessible to N2 at 77 K. Measurement of the N2 isotherms at 273 K revealed that Fe-BTC is not a dense, non-porous material [Fig. S6]. At 77 K there is an 11-fold decrease in the maximal uptake of N2 in Fe-BTC compared to MIL-100. Whereas at 273 K this is greatly reduced to a factor of only 1.4. This is symptomatic of activated diffusion, where N2 molecules cannot successfully diffuse through Fe-BTC at 77 K but can overcome this energetic barrier at higher temperatures. 39 Motivated by this, we further investigated the porous nature of these two materials by measuring a series of gas sorption isotherms (H2, CO2, Xe, Ar, CH4, C2H4, C2H6, C3H6 and C3H8) at 77, 273 or 293 K up to 100 kPa [see Table S3 for analyte properties]. ## Hydrogen and Carbon Dioxide Pure isotherms of H2 were measured at 77 K for MIL-100 and Fe-BTC [Fig. 4]. The initial uptake was similar in both materials up to 20 kPa, exhibiting rapid adsorption kinetics. 40 The H2 isotherms deviate at higher pressure where the larger surface area of MIL-100 allows for greater uptake of H2. At the highest pressure, the maximal uptake of MIL-100 was 110.7 cm 3 g −1 , and 74.7 cm 3 g −1 for Fe-BTC, though neither isotherm reached saturation at the pressures studied here [see Table 1 for maximal uptakes]. Hence the extent of adsorption correlates with the guest molecule binding strength to the framework between 0 to 100 kPa. Gas sorption of CO2, with its highly polar bonding, is a simple way to probe a MOF's textural properties [Fig. 4]. MIL-100 displayed a maximal uptake of 64.4 cm 3 g −1 at 273 K, while Fe-BTC adsorbed 35.4 cm 3 g −1 in comparison. This further demonstrates that Fe-BTC is indeed capable of adsorption and is not dense, nor does it possess a collapsed porous interior. They both exhibited mild hysteresis upon desorption, indicating the adsorption and desorption branches are not in equilibrium. The BET surface areas derived from these isotherms were 127 and 77 m 2 g −1 for MIL-100 and Fe-BTC, respectively [Table S4]. 41 These values represent the lower limit of accessible surface area, however they are nonetheless useful, comparative values given the limited utility of the N2 isotherms at 77 K. These results are further diagnostic of activated diffusion occurring in Fe-BTC with the higher temperature of 273 K enabling diffusion of CO2 through the structure. Hence, the ratio of maximal uptakes for CO2 is more comparable to N2 at 273 K than 77 K. Please do not adjust margins Please do not adjust margins 2]. 18 This is consistent with stronger interactions occurring in Fe-BTC due to confinement of the CO2 molecules within the smaller pore cavity compared to MIL-100. Qst decreases at higher loadings of CO2 in both materials. ## Noble Gases Pure isotherms for Xe and Ar were measured at 273 K for MIL-100 and Fe-BTC [Fig. 5]. The Xe isotherm for MIL-100 was almost linear in the pressure range studied, with a maximal uptake of 43.1 cm 3 g −1 , comparable to previous reports. 42 Fe-BTC had a maximal Xe uptake of 20.1 cm 3 g −1 . The Xe isotherm for Fe-BTC began to plateau at higher pressures, indicating it was approaching saturation. Such behaviour is consistent with reduced pore space available in Fe-BTC compared to MIL-100. Fe-BTC displayed large hysteresis upon desorption, suggesting restricted diffusion of Xe through the structure, compared to the more accessible pores of MIL-100. Both materials exhibited similar sorption of Ar, with nearlinear adsorption and no hysteresis upon desorption. The maximal adsorption to MIL-100 was 4.4 cm 3 g −1 whilst Fe-BTC adsorbed 4.0 cm 3 g −1 ; these values are comparable to that of ZIF-8 and HKUST-1. 43 MIL-100 and Fe-BTC's maximal Ar uptakes are considerably lower than for Xe, which exhibits a greater adsorption strength due to its higher boiling point and results in the hysteresis upon desorption for Fe-BTC. The smaller size of Ar enables easier diffusion through the structure leading to no hysteresis being observed. ## Hydrocarbons Gas sorption isotherms of five short-chain hydrocarbons (CH4, C2H4, C2H6, C3H6 and C3H8) were collected at 273 K [Fig. 6 & 7]. In MIL-100, maximal uptakes at 100 kPa were primarily dictated by the hydrocarbon chain length (increasing C1 < C2 < C3) due to the higher boiling points of the longer chain molecules resulting in stronger adsorbate-adsorbent interactions. Conversely, Fe-BTC did not follow this trend, instead exhibiting some interesting adsorption behaviour. In both MIL-100 and Fe-BTC, CH4 exhibited the lowest maximal uptakes, adsorbing 10.8 and 10.6 cm 3 g −1 , respectively [Fig. 6a & b]. Neither material exhibited hysteresis and both isotherms were near-linear, together indicating that the pores were far from saturation. An additional isotherm of CH4 was collected at 293 K for MIL-100 and Fe-BTC [Figs. S12 & S13]. In both materials the maximal uptake of CH4 was reduced at the higher temperature. Virial analysis revealed Qst values at nearzero coverage of 11. S6]. Qst values for CH4 remained almost constant as the adsorbate loading increased. In MIL-100, C2H4 exhibited steeper initial uptake than C2H6 up to 50 kPa [Fig. 6a]. As pressures increased, the C2H4 and C2H6 isotherms intersect and at 100 kPa the maximal uptake of C2H6 (70.7 cm 3 g −1 ) was marginally higher than C2H4 (63.0 cm 3 g −1 ). Neither C2H4 nor C2H6 displayed hysteresis upon desorption in MIL-100. This inability to discriminate between C2H4 and C2H6 is very common due to the similar physical properties of the two gases. 7 In Fe-BTC, C2H4 also showed steeper initial uptake than C2H6 at low pressure [Fig. 6b]. The maximal uptakes of C2H4 and C2H6 were 32.5 and 28.7 cm 3 g −1 , respectively. Upon desorption, Fe-BTC exhibited hysteresis with C2H6 but not C2H4. The steeper ## Analyte Temperature (K) MIL-100 (cm 3 g −1 ) Fe-BTC (cm Please do not adjust margins Please do not adjust margins initial adsorption of C2H4 in both materials occurs due to unsaturated metal sites that possess a greater affinity towards the unsaturated hydrocarbons at low pressure. C2H6 is only slightly larger than C2H4 (c.a. 0.2 ), and both have similar polarizability. Together this results in very similar maximal uptakes of C2H4 and C2H6 in both materials. However, the slightly larger size of C2H6 results in diffusion limitations through the disordered Fe-BTC structure. Hence, appreciable hysteresis is observed upon desorption of C2H6 in Fe-BTC, but not in the more accessible MIL-100 pore network. The most interesting observations were made with the adsorption of C3H6 and C3H8. In MIL-100, C3H6 exhibited steeper initial uptake than C3H8 at low pressure [Fig. 7a]. At the highest pressure, the maximal uptakes of C3H6 (147.4 cm 3 g −1 ) and C3H8 (141.7 cm 3 g −1 ) were almost identical, and neither isotherm demonstrated hysteresis upon desorption, similar to previous reports. 44 Again, this inability to discriminate between C3H6 and C3H8 is very common amongst most MOFs. 7 Strikingly, Fe-BTC demonstrated very different sorption behaviour. C3H6 showed significantly steeper initial uptake than C3H8, suggesting very different interactions upon initial adsorption; the maximal uptakes of C3H6 and C3H8 were 44.0 and 18.8 cm 3 g −1 , respectively [Fig. 7b]. Unambiguous discrimination exists between the two gases, with over twice the amount of C3H6 adsorbed than C3H8 in Fe-BTC. This level of discrimination is uncommon in the crystalline MOF domain and remains even rarer amongst disordered materials. Both gases exhibited hysteresis upon desorption. Similar to C2H4, the steeper initial adsorption of C3H6 in both materials is due to the increased interaction between the unsaturated metal sites and the unsaturated molecules. The marginally larger size and very similar polarizability of C3H8 compared to C3H6 has little effect in MIL-100, with near-identical maximal uptakes of the two. However, the disordered structure of Fe-BTC, predominantly containing the smaller pore cavity with a window aperture of only 5.5 , results in a pronounced magnification of the different physical properties of C3H6 and C3H8. This enables a significantly higher maximal uptake of C3H6, which can more freely occupy the pore space of Fe-BTC, at 100 kPa. We hypothesize that the larger kinetic diameter of C3H8 is comparable to the window aperture size of the small pore, which leads to substantial diffusion restrictions and hysteresis upon desorption. Motivated by these results, additional C3H6 isotherms were collected at 293 K [Figs. S7]. Further highlighting the high affinity of Fe-BTC for C3H6. We measured the kinetics of the uptake of C3H6 by MIL-100 and Fe-BTC [Fig. 8]. In MIL-100, we observed rapid adsorption kinetics with saturation being obtained in under three minutes. This is due to the large pores and apertures enabling unhindered diffusion. Fe-BTC exhibited significantly slower uptake of C3H6 compared to MIL-100, reaching saturation over a time of around 75 minutes. The slower diffusion of C3H6 in Fe-BTC is consistent with our structural model of this material, which contains tortuous diffusion pathways with bottlenecks and apertures that are comparable to the kinetic diameter of C3H6. We speculate that there is potential for the kinetic discrimination of C3H6 and C3H8 by Fe-BTC based on the additional kinetic barriers that are encountered by the latter. ## Thermodynamic Gas Selectivity The topological disorder in Fe-BTC has a large impact on its adsorption properties. Given the technical challenges associated with measuring adsorption isotherms of gaseous mixtures, ideal adsorbed solution theory (IAST) was employed to quantify the thermodynamic gas selectivities in MIL-100 and Fe-BTC. 45,46 Specifically, the selectivities towards 50:50 binary mixtures of CO2/N2, CH4/N2 and CO2/CH4 were investigated at 273 K. These mixtures are industrially relevant for the processing of flue gases, biogas purification, and natural gas purification. 45 Initially, each pure-component adsorption isotherm was fitted to one of four models before being used to calculate the thermodynamic selectivities of MIL-100 and Fe-BTC [See Supplementary Methods & Fig. S22 & Table S8]. In MIL-100, the CO2/N2 selectivity was 55.7 at 1 kPa, which increased by a factor of 2.5 to 139.7 at 100 kPa, while Fe-BTC had an initial selectivity of 67.8 that increased by a factor of almost 2.6 to 174.8 at 100 kPa [Fig. 9a]. This favourable adsorption of CO2 over N2 relates to the activated diffusion phenomenon discussed previously and the reduced impact of the molecular sieving effect experienced by CO2. Our results for MIL-100 are approximately four times higher than that previously reported; however, this is likely due to the lower experimental temperatures used here. 45 The adsorption of CH4 and N2 is equally competitive, and the resulting selectivity was an order of magnitude smaller than for CO2/N2 [Fig. 9b]. The CH4/N2 selectivity for MIL-100 was 3.3 at 1 kPa and 2.5 at 100 kPa, while Fe-BTC was 7.2 at 1 kPa and 5.0 at 100 kPa. Neither material exhibited pressure dependence. The CO2/CH4 selectivity for MIL-100 was 21.8 at 1 kPa and 17.4 at 100 kPa, consistent with previous reports [Fig. 9c]. 45 Fe-BTC had a CO2/CH4 selectivity of 12.8 at 1 kPa and 11.7 at 100 kPa. Again, neither material showed significant pressure dependence. The slightly enhanced selectivity in MIL-100 is consistent with the larger, relative Qst value derived for CO2 with respect to CH4 for MIL-100 compared to Fe-BTC, which results from the large quadrupole moment of CO2. Due to the adsorption and desorption branches of Fe-BTC's C3H6 isotherm not being at equilibrium, we avoided performing IAST analysis on these data to examine its thermodynamic selectivity towards C3H6/C3H8 mixtures. ## Conclusions As perhaps anticipated, the giant pore architecture and high porosity of crystalline MIL-100 mean that it exhibits a higher maximal uptake for many gases compared to its disordered counterpart, Fe-BTC. However, the absence of long-range order in the disordered material effects higher affinity towards certain gases, such as CO2 and CH4. It also leads to the emergence of highly sought after C3H6/C3H8 discrimination capabilities and highlights the prospective utility of disordered MOFs in the field of gas sorption. Please do not adjust margins Please do not adjust margins Critically, our study has established that while the ordered nature of MIL-100 facilitates accessibility to the larger internal pore network, the presence of substantial structural disorder, as in Fe-BTC, may impart powerful separation abilities on the framework. The balance between order and disorder in MOFs as a route to augment their sorption properties is an appealing avenue for future investigation. Two potential routes to tune the interplay between selectivity and capacity in these materials include (i) the progressive incorporation of defects into MIL-100 to introduce disorder or (ii) adjusting the synthesis of Fe-BTC to retain a greater degree of porosity. The number of disordered, functional MOF structures is rapidly growing; simultaneously, computational methods further accelerate the discovery of new disordered materials, 23 and databases set up to store the few characterised amorphous MOF configurations. 47 In the future, we anticipate the curation of disordered MOF structural databases, much like the crystalline CoRE MOF database, that can be screened for desirable properties. 48 Ultimately, this will enable us to survey the broad synthetic landscape of disordered MOFs in the search for high-performance materials in applications such as hydrocarbon separation. While we have used the MIL-100 and Fe-BTC pairing here as an important demonstration that structural disorder can enhance the gas sorption properties of MOFs, we are certain that this behaviour extends beyond these two specific materials. The use of structural disorder within MOFs as a general tool to enhance gas storage and separation abilities remains widely underappreciated, which we believe hinders the realisation of the full potential of this fascinating class of materials. ## Author Contributions A.F.S. and T.D.B. designed the project. A.F.S. synthetised the samples and collected the powder X-ray diffraction data. C.W.A., L.K.M. and S.J.L collected the gas sorption data. A.F.S. analysed the data with the help of S.G.T. A.F.S. wrote the manuscript with the input of all authors.

chemsum

{"title": "Gas Adsorption Selectivity in Topologically Disordered Metal-Organic Frameworks", "journal": "ChemRxiv"}