Datasets:

nxt-g3n

/

papers

like 0

that passed through all 64 deceleration stages. These mol- ecules entered the decelerator with a velocity of 271.5 m/sand exit the decelerator with a velocity of 91.8 m/s. Theearlier peak originates from molecules that were already oneperiod further inside the decelerator at the start of the timesequence. These molecules also entered the decelerator witha velocity of 271.5 m/s, but since they missed the last two deceleration stages they exit the decelerator with a slightlyhigher velocity of 102 m/s. The peak appearing at a latertime in the TOF distribution originates from molecules en-tering the decelerator with a lower initial velocity of 268 m/s,catching up with the time sequence one period later.Throughout the decelerator they trail the central group ofmolecules by 11 mm, exiting the decelerator with the samefinal velocity of 91.8 m/s. Due to transverse focusing of thebeam, the signal intensity typically increases with a factor of14 when the decelerator is switched on; the signal of thedecelerated beam is 1.4 times as large as the signal of theoriginal beam. It should be noted that the signal of the origi-nal beam is the signal to which all hyperfine structure levelsof the upper inversion component contribute whereas only2/3 of the levels ~48 of the 72 hyperfine levels, including m degeneracy !contribute to the signal when the decelerator is switched on. The gray curves show the result of a 3D MonteCarlo simulation of the experiment using the calculated elec- tric field and the Stark effect of 14ND3as input. The Gauss- ian velocity distribution that has been used as input for theMonte Carlo simulation has been adjusted to match the ob-served TOF profile of the original beam. With this input, thesimulations are seen to quantitatively reproduce the TOFprofiles recorded with the decelerator on. In Fig. 8 the density of 14ND3molecules 24.5 mm behind the decelerator is shown as a function of time after the startof the time sequence.TheTOF profiles are all recorded using a time sequence corresponding to a phase angle f0570°. Lower final velocities are obtained by starting from lowerinitial velocities; final velocities of 91.8 m/s, 73.5 m/s, 57.9m/s, 36 m/s, and 15 m/s are obtained by starting from initialvelocities of 271.5 m/s, 265.9 m/s, 261.9 m/s, 258 m/s, and256 m/s, respectively. The gray curves show the result of aone-dimensional Monte Carlo simulation of the TOF pro-files. For each TOF profile, the results from the Monte Carlosimulation are shown as dots in a phase-space diagram, to-gether with some trajectories calculated using the model pre-sented in Sec. II A. The dots represent the position and ve-locity of individual molecules along the molecular beam axisat the exit of the decelerator, relative to those of the synchro-nous molecule. In the lower graph, the initial velocity andposition at the entrance of the decelerator of the moleculesthat are decelerated from 256 m/s to 15 m/s are shown aswell. The final kinetic energies of the bunches of moleculesfor which the TOF profiles are shown are separated by ~mul- tiples of !0.8 cm 21.The phase-space distributions can there- fore be interpreted as the phase-space distribution inside thedecelerator at the 56th, 59th, 61st, 63rd, and 64th stage, for abeam that is decelerated from an initial velocity of 256 m/sto a final velocity of 15 m/s. It is seen that the shape of thephase-space distribution of the beam at low final velocitiesstarts to deviate from the model. The trajectories of the mol-ecules in phase space are no longer closed and the phase- space area changes ‘‘from a fishlike to a golf-club-likeshape.’’ This is known to occur also in charged particle ac-celerators when the acceleration rate is high ~see, for in- stance, Ref. @49#,p .2 2 2 !.The phase-space distribution of the beam at vz515 m/s resembles a tilted ellipse, which cov- ers the same area in phase space as the phase fish originatingfrom the model. Each data point in the measurements shown in Fig. 8 represents the density of ammonia molecules at the laser fo-cus at a certain time. This signal decreases with lower for-ward velocity due to spreading of the package both in longi-tudinal and transverse directions. To eliminate the effect of FIG. 8. TOF profiles for14ND3molecules recorded 24.5 mm behind the decelerator as a function of time after the start of thetime sequence. The TOF profiles are all recorded using a time se-quence for a phase angle f0570°, but starting from different initial velocities. The gray curves show the result of a one-dimensionalMonte Carlo simulation of the TOF profiles. For each TOF profile,the results from the Monte Carlo simulation are shown as dots in aphase-space diagram, together with some trajectories calculated us-ing the model presented in Sec. II A. The full scale of the phase-space diagrams is 22 mm to 2 mm for the zaxis and 28 m/s to 8 m/s for the vzaxis.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416 053416-8

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0008

the spreading in the longitudinal direction, the signal is inte- grated over time. The time integrated signal is proportionalto the number of molecules passing through the laser focus,divided by their velocity; slow molecules contribute more tothis signal than fast molecules. In Fig. 9, the time-integratedsignal times the forward velocity is shown as a function ofthe forward velocity. The measurements are corrected for theintensity of the molecular beam at the initial velocity. Underthe assumption that the transverse temperature is indepen-dent of the longitudinal velocity ~see Sec. II B !, the observed decrease in these measurements is modeled. The solid curvesshown in Fig. 9 are the result of a simulation for transversevelocity distributions with different widths.The observed de-crease is best described by a transverse velocity distributionwith a width of around 5 m/s ~FWHM !in good agreement with the 3D Monte Carlo simulations. The measurements arevery sensitive to the average velocity of the molecular beamthat is used for correction. The transverse velocity distribu-tion that is found can therefore be considered as a consis-tency check only. However, the fact that the measured pointsfollow the calculated curves ~irrespectively of the average velocity that is taken !indicates that the transverse velocity distribution is indeed independent of the final velocity of thedecelerated beam. From the simulations the emittance of the decelerated beam is found to be @1.1 mm 36.5 m/s #3@1m m 35 m/s #2, where the position and velocity spread indicate the FWHM of the fitted Gaussian distributions. These veloc-ity distributions correspond to a translational temperature of18 mK and 11 mK in the longitudinal and transverse direc-tion, respectively. The ion signal is proportional to the den-sity at the center of the trap. Typically, 125 ions/pulse aredetected for the decelerated beam with a velocity of 91.8m/s.As both the strength of the transition used for detection,and the exact size of the laser focus is not known, it is notpossible to calculate a reliable number for the absolute den-sity. However, a rough estimate can be made. The diffraction-limited waist of the laser focus is 35 mm. Using 10 mJ/pulse of laser radiation the transition is seen to satu-rate. It is assumed that all molecules within a cylinder with adiameter of two times the waist of the laser focus and a length of 2 mm ~determined by the holes in the end cap of the trap !are ionized. Assuming a unit ion detection effi- ciency, a density of 2 310 7molecules/cm3results. This is the density 24.5 mm behind the decelerator, the density justbehind the decelerator is approximately five times higher.The total number of molecules in the decelerated bunch is 10 5. These numbers can be assumed to be a lower limit. By combining the density in the beam with the emittance of thebeam, the phase-space density is found. The definition for the phase-space density used here is n 0L3where Lis the thermal de Broglie wavelength, L5A2p\2/mkTandn0is the peak density @50#. The phase-space density for the decel- erated beam is then found to be 4 310212. Similarly, the phase-space density of the ammonia molecules in uJ,MK& 5u1,21&state in the source region of the beam is calculated to be 2 310211. This is the phase-space density at the peak velocity of the beam, the space density at the velocity of271.5 m/s selected by the decelerator, will be slightly lower.Therefore, the phase-space density in the decelerator de-creases by about a factor of 4, attributed to the fact that thehexapole has not been used during these experiments. V. SIMULTANEOUS DECELERATION OFAMMONIA ISOTOPOMERS The deceleration process as employed here depends on the Stark shift that the molecules experience in the appliedelectric fields. In the model used to describe this process, aspresented in Sec. II A, the dependence on the exact shape ofthe electric field along the molecular beam axis is removed,and the motion is determined only by the difference in Starkenergy before and after switching of the field. One can won-der, therefore, if it would be possible to simultaneously de-celerate bunches of molecules having different masses and/ordifferent Stark shifts. Let us consider the situation wheremolecules with a mass mand experiencing a Stark shift W(z) are decelerated using a time sequence for a phase angle f0. The question now is whether for molecules with a dif- ferent mass m8and/or a different Stark shift W8(z) a syn- chronous molecule exists for this time sequence, i.e., whether a molecule with a phase angle f08that is constant throughout the decelerator exists. If this synchronous mol- ecule with mass m8exists, it will, by definition, traverse a distanceLin the time interval between subsequent switching times of the time sequence, just as the synchronous moleculewith mass mdoes. Therefore, the synchronous molecules with mass mandm 8will have the same average velocity and the same change in average velocity per deceleration stage.This implies that DK 8~f08! m85DK~f0! m. ~6! A synchronous molecule with mass m8can therefore be found, provided that @DK8(f08)#max>(m8/m)DK(f0), which is always fulfilled when the time sequence used is calculated for the molecule that can be least well decelerated. FIG. 9. Time-integrated signal of the slow molecules exiting the decelerator times their forward velocity as a function of their for-ward velocity, using the TOF profiles shown in Fig. 8. The mea-surements are normalized to the signal at 91.8 m/s.The solid curvesshow the expected dependence of the number of molecules passingthrough the laser focus on the forward velocity using transversevelocity distributions with different widths, as indicated.DECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416 053416-9

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0009

Obviously, molecules with different masses and/or different Stark shifts will have different values of the phase angle f0 and will thus have a different acceptance. In Fig. 10, DK(f0)/mis shown as a function of the phase angle for14NH3,15ND3, and14ND3molecules. The hori- zontal line shows DK(f0)/m50.032 cm21/amu corre- sponding to a phase angle of f0570°, 56°, and 53° for 14NH3,15ND3, and14ND3, respectively. In Fig. 11, TOF profiles recorded 24.5 mm behind the decelerator as a function of time after the start of the timesequence are shown for three different isotopomers of am-monia. The measurements are performed using a time se- quence that has been calculated for 14NH3at a phase angle f0570°. The various ammonia isotopomers are decelerated from 256 m/s to 96 m/s. It is evident from the measurements that the time integrated signal is larger for14ND3and15ND3 than for14NH3. This is the combined effect of a larger spa- tial extent together with a larger longitudinal velocity spreadof the decelerated beam, as indicated in the phase-space dia- gram in the inset. The small ( ;0.5 mm) shift in the average position of the decelerated beam as seen in the phase-space diagram results in a small ( ;5 ms) increase in the overall time of flight for the deuterated ammonia molecules. The peculiar TOF profile observed for14ND3and15ND3reflects the phase-angle dependence of the transverse acceptance.Molecules contributing to the signal in the wings of the TOFprofile have experienced large excursions in phase anglewhile traversing the decelerator. These molecules have,therefore, experienced widely varying transverse focusingforces ~see Fig. 4 !. As explained in Sec. II B, this leads to a reduced transverse acceptance for these molecules. The re-sults of 3D Monte Carlo simulations, shown underneath theexperimental TOF profiles, are seen to reproduce the obser- vations. The TOF profile for 14NH3is more symmetric, and effects due to transverse focusing are almost absent; the ob-served TOF profile can be reproduced equally well in a 1DMonte Carlo simulation. Experiments have been performed on other ammonia iso- topomers like 14NHD2and14NDH2as well. For these iso- topomers the molecular symmetry is reduced, allowing thesemolecules to cool down to the absolute ground state. Nolow-field-seeking states were found to be populated in thebeam. VI. ELECTROSTATIC TRAP Electrostatic trapping of atoms and molecules was first considered by Wing @51#. The trap used here is very similar to the trap proposed in that original work. A cut through thetrap along the zaxis is shown in Fig. 12.The geometry of the trap is the same as that of a Paul trap for charged particles@52#. However, instead of using rf fields as used for trapping charged particles, static electric fields are used for trappingneutral molecules. The trap consists of two end caps and aring electrode. The inner radius Rof the ring electrode is 5 mm. The end cap half-spacing is R/ A2. In the end caps 2 mm diameter holes are made to enable the molecular beamto pass through. In the ring electrode 2 mm diameter holesare made to allow for laser detection of the trapped mol-ecules at the center of the trap. The electric field in the traphas a quadrupole symmetry and is given by E5V ~x21y214z2!1/2/R2, ~7! withVthe voltage difference applied between the ring elec- trode and the end caps and with the ( x,y,z) coordinates rela- tive to the center of the trap. The absolute value of the elec-tric field in the center is zero, and larger elsewhere. In weak electric fields the Stark shift of 14ND3and15ND3is qua- FIG. 10. The kinetic energy lost per stage divided by the mass for14NH3,15ND3, and14ND3molecules in the uJ,MK&5u1,21& state, as a function of the phase angle f0.The horizontal line shows DK(f0)/m50.032 cm21/amu. FIG. 11. Time-of-flight profiles for14NH3,15ND3, and14ND3 molecules recorded 24.5 mm behind the decelerator as a function of time after the start of the time sequence. All TOF profiles are re-corded using the time sequence calculated for 14NH3at a phase angle f0570°. The gray curves show the results of 3D Monte Carlo simulations of the deceleration process. The area in phasespace at the exit of the decelerator is indicated in the inset for 14NH3~inner curve !,15ND3~middle curve !, and14ND3~outer curve !.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416 053416-10

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0010

dratic and the restoring force close to the center is linear in the displacement. In higher electric fields the Stark shift be-comes linear, and the force becomes constant. With a voltagedifference of 12 kV between the ring electrode and the endcaps the maximum electric field of a closed contour of equalelectric field is approximately 20 kV/cm. The maximumelectric field that occurs on the electrodes, which limits thevoltage difference that can be applied, is about three times as high. For ND 3the trap depth is about 0.24 cm21,o r3 5 0 mK, as can be seen in the lower curve of Fig. 12. Near thecenter of the trap the motion is harmonic with an oscillation frequency V z/2p52Vx/2p52Vy/2p52 kHz. The maximum velocity of14ND3molecules, which can be confined in the trap with these settings, is approximately 17 m/s. In Fig. 13 lines of equal energy for14ND3molecules are shown in phase space, representing the acceptance of thetrap along the molecular beam direction. The phase-spacediagrams of the trap in the xandydirection are the same as that in the zdirection, but with a two-times larger spatial acceptance in xandy. In order to load the molecules into the trap without loss in phase-space density, the emittance of thebeam must be matched to the acceptance of the trap. This is done in two steps. First, the molecules are focused both inthe transverse and longitudinal directions onto the entranceof the trap. Second, the molecules are decelerated upon en-tering the trap by applying the voltages asymmetrically to thetrap electrodes. The average velocity of the beam is therebyreduced to zero at the center of the trap. By using a phase angle of f0570° to decelerate14ND3 molecules, the emittance of the beam exiting the decelerator is@1.1 mm 36.5 m/s #3@1m m 35 m/s #2. The absolute velocity spread of the beam remains constant throughout thedecelerator. Therefore, by decelerating the beam the relativevelocity spread and the divergence of the beam increase. Toload the trap, the beam is decelerated down to 15 m/s. Overthe 24.5 mm flight path from the decelerator to the trap, thedensity will decrease by more than two orders of magnitude.In order to focus the beam both in the longitudinal and trans-verse directions, a ring electrode with an inner diameter of 4mm is mounted in front of the entrance end cap of the trap.By applying a voltage difference between this ring electrodeand the entrance end cap, an electric field is generated with aminimum on the molecular beam axis, focusing molecules inlow-field-seeking states in the transverse direction. Unlike ina hexapole, this minimum electric field will be nonzero. Thisnonzero field on the molecular beam axis can be used tofocus the molecules in the longitudinal direction ~‘‘bunch- ing’’!. In Fig. 14 a cut through the trap with the ring electrode in front is shown. Contours of equal electric field are drawn foran applied voltage difference of 10 kVbetween the ring elec-trode and the entrance end cap of the trap. In the lower part of Fig. 14 the potential energy of 14ND3molecules in the uJ,K&5u1,1&low-field-seeking state along the molecular beam axis is shown. For the transverse and longitudinal fo-cusing of the beam, the electric field distribution near the exitof the ring electrode, particularly the rising edge of the field in thez5216 mm to 211 mm region, is important. In this FIG. 12. Configuration of the electrostatic trap with the voltages as applied during trapping. In the trap, lines of equal electric fieldstrength are shown. The lower curve shows the potential energyalong the molecular beam axis ( zaxis!for 14ND3molecules in the uJK&5u11&low-field-seeking state. FIG. 13. Phase-space diagram for14ND3molecules in the uJK&5u11&low-field-seeking state. Lines of equal energy, repre- senting the acceptance of the trap along the zaxis, are shown. FIG. 14. A cut through the trap with the ring electrode, with voltages as applied for bunching and loading of the trap. The solidcurves are lines of equal electric field. In the lower part of the figurethe potential energy along the molecular beam axis for 14ND3in the uJ,K&5u1,1&low-field-seeking state is shown.DECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416 053416-11

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0011

region the potential energy along the molecular beam axis is nearly quadratically dependent on z. The molecules will ex- perience a linear force F52k(z2z0) withz0the position where the quadratic potential starts ( z05216 mm) and k the force constant. This force can be written in terms of the positionzsof the synchronous molecule as F52k(z2zs) 2k(zs2z0). The force is thereby separated in a restoring force towards the synchronous molecules and a force thatlowers the velocity of the synchronous molecule. The restor-ing force leads to a rotation of the longitudinal phase-spacedistribution of the beam around the position of the synchro-nous molecule in phase space. In order for the molecules torotate over the same angle in phase space, the potential mustbe switched on ~and off !when all molecules are in the qua- dratic part of the potential. In our case, the range over whichthe potential is quadratic allows only the central part of thebeam to be focused. When the potential is on during theappropriate time interval, this central part of the beam willrefocus in free flight from the buncher to the trap, forming animage of the beam exiting the decelerator onto the entranceof the trap. The magnification of the image is determined bythe ratio of the time of flight from the decelerator to the ringelectrode to the time of flight from the ring electrode to thetrap. VII. TRAP-LOADING EXPERIMENTS In the upper part of Fig. 15, the TOF profiles for14ND3 molecules recorded in the trap are shown for four differenttimes, indicated by arrows, at which the electric field of the buncher is switched off. The time intervals DTgiven in the figure are the intervals between the time at which the syn-chronous molecule exits the decelerator ~2.830 ms after start of the time sequence !and the times at which the buncher is switched off. The lower curve shows the TOF profile whenthe buncher is not used. These measurements reflect the spa-tial distribution of the beam inside the trap, which is consid-erably narrowed by using the buncher. When the electricfield of the buncher is switched off at later times, the mol-ecules are decelerated more and are seen to arrive later at thecenter of the trap. In addition, the signal is increased due totransverse focusing of the beam, which is optimal at latertimes. The gray curves underneath the measurements showthe results of 3D Monte Carlo simulations of the experiment.The simulations are seen to qualitatively reproduce both thelongitudinal and transverse focusing of the beam. Since theouter shape of the entrance end cap electrode, and thus theelectric field in front of the electrostatic trap, is rather poorlydefined in our case, a better agreement would not be ex-pected. In the lower part of Fig. 15, the evolution of thebeam in phase space is shown for the situation where the buncher is switched off 800 ms after the synchronous mol- ecule exits the decelerator. In this case the central part of thebeam is spatially bunched at the entrance of the trap. Thebunched part of the beam occupies a phase-space area of @2m m 34 m/s #in the forward direction, and has an aver- age velocity of 12 m/s. In the transverse direction the imageis less well-defined and is in practice determined by the sizeof the hole in the entrance end cap of the trap.In order to bring the average velocity of the beam down to zero, the voltages to the trap electrodes are applied asym-metrically, as shown in Fig. 14. With a voltage of 10 kV onthe entrance end cap and 12 kV on the ring electrode, theelectric field will be small at the entrance of the trap and willincrease towards the center of the trap. The potential energy of 14ND3molecules in the uJK&5u11&low-field-seeking state along the molecular beam axis is shown in the lowerpart of Fig. 14. The synchronous molecule entering the trap with a velocity of 12 m/s ( E kin50.12 cm21) will come to a standstill at the center of the trap. The potential energy isnearly quadratically dependent on the position along the mo-lecular beam axis. Therefore, in the time that the synchro-nous molecule is slowed down to zero the whole beam is rotated over an angle of 1 2pin phase space. This is identical FIG. 15. Time-of-flight profiles for14ND3molecules recorded at the center of the trap for four different times, indicated by arrows, atwhich the buncher is switched off. The time intervals DTbetween the time at which the synchronous molecule exits the decelerator~2.830 ms on the horizontal axis !and the times at which the buncher is switched off, are given.The lower curve shows the time-of-flight profile of the beam when the buncher is not used. The graycurves show the results of 3D Monte Carlo simulations of the ex-periment. In the lower part of the figure the phase-space diagramsof the beam are shown for different times after exiting the decel-erator, together with the potential energy W(z)~right axis !along the molecular beam axis for 14ND3in the uJ,K&5u1,1&low-field- seeking state.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416 053416-12

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0012

to keeping the original phase-space distribution and simply interchanging the velocity and position axes, with the appro-priate scaling. In Fig. 16 the TOF profiles for 14ND3molecules are re- corded at the center of the trap with the buncher switched off 800ms after the synchronous molecule has exited the decel- erator. The lower curve is recorded without deceleration ofthe molecules inside the trap, and is the same as the curveshown in Fig. 15. In the upper curve, the molecules are de-celerated upon entering the trap. When the voltages on thetrap electrodes are kept on, the trap will act as a concavemirror for the ammonia molecules. The upper curve showsthe density of ammonia molecules at the center of the trapwhile the molecules are reflected from the electrostatic mir-ror. The increased signal of the upper curve relative to thelower one is due to bunching and transverse focusing whenthe electric field is on. The gray curves show the results of3D Monte Carlo simulations of the experiment. In the lowerpart of Fig. 16 the phase-space diagrams of the beam areshown for different times after exiting the decelerator, to- gether with the potential energy W(z)~right axis !along themolecular beam axis for 14ND3in the uJ,K&5u1,1&low- field-seeking state. The phase-space diagram at 1.830 msshows the beam when the synchronous molecule, enteringthe trap with 12 m/s, has come to a standstill. The central part of the beam occupies a phase-space area of @0.8mm •10m/s #in the forward direction. This phase-space distribu- tion is shown enlarged in Fig. 17, with the longitudinal ac-ceptance diagram of the trap as an overlay. VIII. TRAPPING EXPERIMENTS In order to trap the molecules the voltage on the entrance end cap is switched off once the synchronous molecule has come to a standstill. In Fig. 18 the density of14ND3mol- ecules at the center of the trap is shown as a function of time.The trap is switched on 1.830 ms after the synchronous mol-ecule exits the decelerator, i.e., 4.660 ms after the start of thetime sequence. Shortly after the trap is switched on the den-sity at the center of the trap is seen to oscillate. These oscil-lations have faded away after several milliseconds and asteady signal from the trapped ammonia molecules remains.The inset in the figure shows the ammonia density on alonger time scale. The ammonia signal intensity is seen to exponentially decay with a 1/ etime of 0.3 s, although at FIG. 16. Time-of-flight profiles for14ND3molecules recorded at the center of the trap with the buncher switched off at DT 5800ms. The upper curve is recorded with voltages applied asymmetrically to the trap electrodes. The lower curve is recordedwithout deceleration of the molecules upon entering the trap. Thegray curves show the results of 3D Monte Carlo simulations of theexperiment. In the lower part of the figure the phase-space diagramsof the beam are shown, together with the potential energy W(z) ~right axis !along the molecular beam axis for 14ND3in the uJ,K& 5u1,1&low-field-seeking state. FIG. 17. The phase-space distribution of the beam at DT 51830 ms at the position of the trap together with the acceptance diagram of the trap. FIG. 18. Density of14ND3molecules recorded at the center of the trap as a function of time.The time at which the trap is switchedon is indicated with an arrow. The inset shows the density of 14ND3 molecules on a longer time scale, and an exponential fit to the data~solid line !w i t ha1 /edecay time of 0.3 s.DECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416 053416-13

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0013

short times the decay appears to be slightly faster. The ob- served trap loss is attributed to collisions with backgroundgas in the vacuum chamber; the background pressure of typi- cally 2 310 28Torr is consistent with the observed trap loss rate. The oscillations that are observed shortly after the trap is switched on, indicate that the emittance of the beam is notperfectly matched to the acceptance of the trap. Two situa-tions can be distinguished: ~i!the emittance and the accep- tance are not centered around the same point in phase space,i.e., the average position of the beam is not at the center ofthe trap and/or the average velocity of the beam is nonzerowhen the trap is switched on, and ~ii!the emittance of the beam and the acceptance of the trap do not have the sameshape in phase space, i.e., the ratio of the spatial distributionto the velocity distribution of the beam is different from thatof the acceptance diagram of the trap. These two situationslead to different types of oscillations. In the first situation,the whole beam will oscillate back and forth in the trap. Inthe second situation, the beam will perform a breathing mo-tion in the trap. The oscillations will decay due to the anhar-monicity of the trap potential. In order to experimentally study these oscillations, the density of 14ND3molecules is measured at two positions on the molecular beam axis, symmetrically located around thecenter of the trap. In the observed oscillation pattern at thesetwo positions, the two types of oscillation will appear in adifferent manner.The oscillation of the whole beam back andforth in the trap results in an oscillation at the longitudinaltrap frequency of 2 kHz; the measurements at the two posi- tions will have a phase difference of 180°. The breathing motion in the trap results in an oscillation at twice the lon-gitudinal trap frequency, with the same phase at both loca-tions. In Fig. 19 the density of 14ND3molecules in the trap is shown as a function of time. Pairs of measurements are takenunder identical experimental conditions but with the detec- tion laser focused at z510.35 mm and z520.35 mm. The middle set of measurements are recorded when the trapis switched on at 1.830 ms after the synchronous moleculehas exited the decelerator. This is the time at which, accord-ing to the calculations, the synchronous molecule has cometo a standstill at the center of the trap ~see Fig. 17 !. The upper and lower measurements are recorded when the trap is switched on 80 ms later and earlier, respectively. Although oscillations can be recognized in all three sets of measure-ments, it is evident that these oscillations are least pro-nounced when the trap is switched on at 1.830 ms, whichthus experimentally appears to be the right time indeed. Inthis case the signal intensity away from the center of the trapis slightly less than in the other measurements, as expected.Under these conditions, the signal intensity at the center ofthe trap, shown in Fig. 18, is optimal. When the trap isswitched on too late, as for the upper curves, the moleculesare already reflected by the mirror and the signal intensity isminimum ~maximum !downstream ~upstream !from the cen- ter of the trap. In the upper curve taken at z510.35 mm, the oscillation back and forth in the trap with a period ofapproximately 0.5 ms is clearly visible. Although still vis-ible, the oscillation back and forth is less pronounced in the curve recorded at z520.35 mm. This indicates that also a breathing motion is present, which can be recognized as suchat twice the longitudinal trap frequency. The measurements that are taken when the trap is switched on 80 ms too early, are similar to the ones that are taken when the trap is switched on too late; the measurements taken at z5 20.35 mm ( z510.35 mm) when the trap is switched on too early have to be compared to the measurements taken at z510.35 mm ( z520.35 mm) when the trap is switched on too late. In the measurements taken at 1.830 ms the os-cillation back and forth in the trap is largely absent. The onlyoscillation expected to be present in this case is due to thebreathing motion in the trap. Although frequencies aroundtwice the trap frequency are observed in both curves, thecurves are less similar than expected.This is probably causedby the anisotropic phase-space distribution of the beam whenthe trap is switched on, as seen in Fig. 17. The transverse oscillations back and forth and the trans- verse breathing motion in the trap are expected at 1 kHz and2 kHz, respectively. In order to study these oscillations mea-surements need to be performed with the laser beam focusedsymmetrically above and below the molecular beam axis.Such measurements are more complicated in the current set-up, as the detection efficiency is strongly position dependentoff axis, and have not been performed. In the measurementsshown in Fig. 19 it is not clear if transverse oscillations arepresent at all. However, it might be that the remaining 2 kHz FIG. 19. Density of14ND3molecules recorded in the trap as a function of time, at two different positions on the zaxis. The three sets of measurements show the dependence of the oscillation pat-tern on the time at which the trap is switched on.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416 053416-14

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0014

oscillation in the measurement shown in Fig. 18, taken at the center of the trap when the trap is switched on at the righttime, is due to the transverse breathing motion in the trap. The longitudinal and transverse distributions generally do not have the same temperature, even when the phase-space distribution of the decelerated beam perfectly matches theacceptance of the trap. In the absence of collisions, redistri-bution of molecules in the trap will lead to a change in tem-perature due to a coupling of the longitudinal and transversemotion. This can be an alternative explanation for the ob-served faster trap loss during the first 20 ms of trapping, asseen in Fig. 18. In order to determine the temperature of the trapped am- monia molecules, the spatial distribution of the molecules inthe trap is measured along the zaxis. This is done by scan- ning the position of the focused detection laser through the 2mm diameter holes in the ring electrode of the trap. For thismeasurement the trap is switched on at 1.830 ms. The detec-tion laser is fired after the molecules have been trapped for50 ms, long compared to the oscillation periods in the trap.Since the density in the trap is low, no collisions betweentrapped molecules will take place during this time interval.Strictly speaking, temperature is not defined in this case; it isused here as a measure of the average energy of the mol- ecules in the trap. Knowing the trapping potential for 14ND3 molecules in the uJ,K&5u1,1&low-field-seeking state, the spatial and velocity distributions can be calculated assuminga certain temperature @50#. To make sure that ion detection is not disturbed by residual electric fields, the detection laser is actually fired 20 ms after the voltages on the trap electrodes are switched off. During this 20 ms time interval the mo- lecular cloud will slightly expand, which is included in thesimulations. In Fig. 20 the density of trapped 14ND3molecules is shown as a function of the position along the zaxis. The detection efficiency is constant over the range in zcovered in these measurements. Each data point is averaged over tentrapping cycles. The error bar indicates the statistical spreadin the measurements. The solid curves are the results of aMonte Carlo simulation, assuming a thermal distribution ofmolecules in the trapping potential. Detection of molecules in a 70- mm-diameter, 2-mm-long cylinder is assumed. The measurements are best described using a thermal distributionwith a temperature of 25 mK. In principle, similar measure-ments could be performed to determine the spatial distribu-tion in the transverse direction as well. As discussed, thesemeasurements are more complicated and have not been per-formed. The maximum number of ammonia ions detected per laser pulse is approximately 45. Assuming a unit ionization effi-ciency of the ammonia molecules in a 2-mm-long, 70- mm-diameter volume of the laser focus and with a unit ion detection efficiency, a lower limit for the peak density of trapped14ND3ofn051.33107molecules/cm3results. The total number of molecules in the trap is then approximately 104—a factor of 10 less than the number of molecules that exited the decelerator. Together with the measured tempera-ture of 25 mK, corresponding to a thermal de Broglie wave- length of L52.5 nm, this yields a phase-space density n 0L3of 2310213. This can be compared to the phase-space den- sity of the original beam, which is calculated to be 2310 211. For the trapping experiment the beam is decelerated starting from an initial velocity of 256 m/s, where the phase-space density is about a factor of 2 lower than at the peakvelocity. The overall loss in phase-space density from thebeam to the trap is therefore approximately 50.As discussedearlier, there is a factor of 4 loss in phase-space density fromthe matching of the molecular beam onto the decelerator.Theremaining factor of 12 loss results from the ~nonideal !load- ing of the decelerated beam into the trap. In Fig. 21 the ion signal resulting from the trapped sample of 14ND3molecules is shown together with the signal ob- tained at the center of the trap for ammonia molecules in theoriginal molecular beam at the central velocity of 285 m/s,and in the decelerated beam at a velocity of 92 m/s. Identical measurements have been performed on 15ND3 using the same time sequence as in the experiment on 14ND3.Although it has been detailed in Sec. V that the same time sequence can be used for deceleration of both the iso-topomers, it is not a prioriclear that this holds for bunching and trap loading as well. One might expect that in the present setup an optimum is found for 15ND3at different settings FIG. 20. Density of trapped14ND3molecules recorded as a function of the position along the zaxis. The solid curves show the results of a Monte Carlo simulation for thermal distributions withdifferent temperatures. FIG. 21. The density of14ND3molecules recorded at the center of the trap for the original molecular beam at the central velocity of285 m/s, for the decelerated beam at a velocity of 92 m/s and for thetrapped molecules. Each data point is averaged over 20 laser shots.DECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416 053416-15

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0015

than used for14ND3. This has not been investigated further. The measurements for the spatial extent of15ND3in the trap is shown in Fig. 22. The measurements are well reproducedwhen a temperature of 25 mK is assumed. Within the experi-mental accuracy the absolute signal is the same for both theisotopomers of ammonia. Both isotopomers have also beentrapped simultaneously at a factor of 2 lower peak densityfor each species. In Fig. 23, (2 11)-REMPI spectra of 14ND3are shown, recorded via intermediate states of different vibrational sym-metry. The spectra are recorded under conditions close tosaturation, and the width of the lines is determined by powerbroadening. All spectra are recorded on the same intensityscale, and are averaged over 40 shots. In the upper panel, the n2854 vibrational level in the B˜1E9state is used as interme- diate. Via this level, only molecules in the lower inversion doublet levels ~vibrational ssymmetry !in the electronic and vibrational ground state are detected. Measurements areshown for the molecular beam ~upper trace !and for trapped molecules ~lower trace !. TheJ Krotational quantum numbers of the ground-state levels are indicated in the figure @53#.I n the lower panel, the n2855 vibrational level is used as inter- mediate, only allowing detection of molecules in the upper inversion doublet levels ~vibrational asymmetry !. In the beam, both sandacomponents of the 0 0,10, and 11levels are observed. The 1 0and 11level are located 10 cm21and 8 cm21above the rotational ground-state level, respectively. The 1 1level is the only para level that is populated in the beam, indicating a rotational temperaturebelow 3 K. The spectra recorded on the trapped moleculesonly show the five allowed transitions originating from the upper inversion doublet level of the 1 1state. IX. CONCLUSIONS AND FUTURE PROSPECTS In this paper a method to decelerate and trap polar mol- ecules using time varying electric fields is demonstrated. It isshown that the high phase-space density of molecules that ispresent in the moving frame of a pulsed molecular beam can,in principle, be transferred to the laboratory frame withoutloss. Different molecules and isotopomers can be trapped simultaneously. On the order of 10 4,14ND3molecules are trapped at a density of over 107molecules/cm3and at a temperature of 25 mK. The corresponding phase-space den- sity in the trap is 2 310213, 50 times less than the initial phase-space density in the beam. This loss can be reducedwith a better matching of the decelerated beam to the trap. Ultimately, the phase-space density in the trap is limited by the phase-space density that can be obtained in the beam.In a pulsed supersonic expansion, densities of 10 13molecules/cm3at a translational temperature o f1Ka r e routinely obtained @45#. Phase-space densities on the order of 1028appear feasible. Compared to this, the measured phase- space density in our beam is rather low, apparently due to thesmall pumping capacity in the current setup. Furthermore,Xe is used as a carrier gas to start with a low velocity of theinitial beam. Due to its large polarizability, Xe readily formsclusters and is therefore not considered to be the ideal carriergas. The use of Kr orAr as a carrier gas may lead to a moreintense initial beam, requiring a longer decelerator, however. To increase the number of trapped molecules the area in phase space that is imaged from the beam source to the trapshould be increased. Currently, the area in phase space thatcan be imaged is limited by the buncher. By using a buncherconsisting out of multiple ring electrodes and a hexapolefocuser, this can be largely improved. The acceptance of the FIG. 22. Density of trapped15ND3molecules as a function of the position along the zaxis. The solid curves show the results of a Monte Carlo simulation for thermal distributions with different tem-peratures. FIG. 23. (2 11)-REMPI spectra recorded via the B˜1E9,n28 54~upper panel !andn2855~lower panel !state in14ND3. In each panel, the upper curves are from measurements on the molecularbeam, whereas the lower curves are recorded on the trapped mol-ecules.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416 053416-16

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0016

decelerator can be increased by using higher electric fields or by using a larger spacing between the electrodes, at the samevalue of the electric field. By using more deceleration stages,the decelerator can be operated at a lower phase angle. Op-timization of the overall deceleration and trapping processwill be performed via feedback control optimization of thetime sequence, using genetic learning algorithms. For agiven experimental geometry and molecule, this optimization needs to be performed only once. Another possibility to increase the number of trapped molecules would be to accumulate molecules from succes-sive deceleration cycles in the trap. An increase of phase-space density via this accumulation is only possible in spe-cific cases, e.g., for the NH radical, where laser-inducedspontaneous decay provides a unidirectional path to reload atrap@54#. It is always possible to increase the total number of molecules while keeping the phase-space density constant.This process is known as phase-space stacking in acceleratorphysics @49#. With the scheme outlined in this article, any polar mol- ecule can be decelerated and trapped provided that it has asufficiently large positive Stark shift in experimentally at-tainable electric fields. In Tabl e I a selection of polar mol- ecules suited for deceleration and trapping is given. The ac-celeration experienced by the molecules in the beam dependson the ratio of their ~positive !Stark shift to their mass. This is given as a ‘‘figure of merit’’ in Table I, scaled relative to 14ND3. Unless noted otherwise the Stark shift has been cal- culated at an electric field of 200 kV/cm. The number ofstages that would be required to bring the molecules to astandstill in the present setup, i.e., with electric fields on theaxis of the decelerator of 90 kV/cm, is given. For this, themolecules are assumed to have an initial velocity determinedby the Xe carrier gas at a temperature determined by thevapor pressure of the molecules. For stable molecules, thetemperature at which the vapor pressure is 0.1 atm is taken;for radicals a room-temperature expansion is assumed. Mostof the molecules that have been listed have been used exten-sively in hexapole focusing experiments, in which case areference to these experiments is given. Both the production~in a molecular beam !and the detection methods for the molecules listed in Table I are well known. For radicals suchas OH, production of intense pulsed beams with densities above 10 11molecules/cm3per quantum state have been re- ported @70#. The use of electrostatic fields to trap molecules, offers many possibilities. Electric fields can be made in a widevariety of shapes, and can be ramped and switched rapidly.Molecules can be trapped around a charged wire @71–74 #. With a hexapole bent into a torus, a storage ring for polarmolecules can be constructed @36,75 #. ac traps can be con- structed to trap molecules in high-field-seeking states@76,77 #. It should be possible to produce microcircuits to manipulate and trap molecules on a chip. Electrostatic trap-ping fields can easily be superimposed with magnetic traps @54#, optical traps @22#, or high- Qoptical cavities @78#. Ultrahigh resolution spectroscopy will greatly benefit from the increased interaction time and the reduced thermalbroadening offered by cold molecules. Molecules are used inthe search for violation of time reversal symmetry @79#and in the study of weak interactions in chiral molecules @80#.B y accuratelymeasuringtheratioofthetransitionfrequenciesofvarious types of optical transitions in molecules, the tempo-ral variation of the fine-structure constant might be deter-mined in a laboratory experiment @81#. The use of cold mol- ecules is expected to increase the precision of these studiesby orders of magnitude. The long storage time in traps makes it possible to study the lifetimes of vibrationally or electronically excited meta-stable states. Molecules can be prepared in these states eitherin the trap or prior to deceleration. Closed rovibrational tran-sitions might actually be used to cool the trapped moleculesfurther via laser cooling. The study of ultracold collisions and reactions forms an interesting research topic.At low temperatures the associatedde Broglie wavelength of the molecules becomes of the samesize as the collision complexes, dramatically changing theinteractions at these temperatures @2#. Besides studying col- lisions among trapped species, interesting effects are pre-dicted for collisions at nonzero collision energies. Calcula-tions show that elastic and inelastic cross-sections showsharp resonances at low collision energies, increasing by twoorders of magnitude at these energies @82#. Measuring the width and position of these resonances can be performed bysending a velocity tuneable molecular beam through thesample of trapped molecules loaded from a previous cycle.Scattering resonances occur, for instance, when collidingmolecules begin to rotate, leaving them with insufficienttranslational kinetic energy to overcome their van der Waalsattraction. The formation of long-lived, transiently bound,molecular complexes might enable reactions to occur viatunneling through reaction barriers, opening up novel routesfor low-temperature chemistry. With a sufficiently high density of trapped molecules, the regime of evaporative cooling might be reached @83#. The success of evaporative cooling depends on the ratio of theelastic to inelastic collision rate of the trapped molecules.This ratio depends critically on molecular properties such asthe magnitude of the dipole moment and of the zero-fieldsplitting @84#. As the trapping scheme presented here can be used for a large number of different molecules, one can hopeto find a suitable candidate molecule. Alternatively, polarmolecules can be trapped in their lowest-energy quantumstate in an ac trap, by which inelastic collisions can beavoided. In this case additional heating due to the rf fieldswill take place. It is expected, however, that under certainconditions evaporative cooling will dominate over this heat-ing@85#. Another viable method to cool the molecules further might be sympathetic cooling with laser-cooled atoms. Re- cently, Bose-Einstein condensation of 41K atoms has been achieved via sympathetic cooling with a87Rb gas @86#. When further cooling of the trapped molecules proves to be feasible, collective effects can be studied in thesesamples. If a molecular Bose-Einstein condensate is formed,there is the possibility to study its behavior for molecules inselected vibrational and ~end-over-end !rotational states. Ar- guably the most interesting aspect of molecules compared toDECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416 053416-17

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0017

atoms is that many molecules have a permanent electric di- pole moment. The anisotropic interaction between the di-poles will give rise to rich and new physics in the cold di-polar gases @87#.The mean-field interparticle interaction, and hence the occurrence of molecular Bose-Einstein condensa-tion, will depend strongly on, and can therefore be manipu-lated by, the trapping geometry @88#. This offers new possi- bilities for controlling and engineering macroscopic quantumstates. It has been predicted that cold dipolar gases of fermi- onsareexcellentcandidatesforachievingthesuperfluidtran-sition @89#. At ultracold temperatures dipolar molecules are expected to form crystalline structures, similar to the forma-tion of Wigner crystals in ion traps @90#.Another challenging prospect is that trapped polar molecules might be used forquantum computation; in this design, the qubits would be theelectric dipole moments of ultracold molecules, orientedTABLE I. A selection of polar molecules suited for deceleration and trapping experiments, with their relevant properties. Molecule State Number Stark shift at Shift/massNumber of stages Dipole of hyperfine 200 kV/cm ~relative required in moment levelsa(cm21)bto ND3) present ~D! setup CH@56,55 # uX2P1/2,J51/2,MV521/4& 4 1.54 1.03 90 ma51.46 uX2P3/2,J53/2,MV529/4& 4 1.88 1.50 71 CF@57# uX2P1/2,J51/2,MV521/4& 4 0.44 0.13 533 ma50.65 uX2P1/2,J53/2,MV523/4& 4 0.32 0.09 845 CH2F2@58# uJtM&5u2220& 1.52~184 kV/cm !0.25 168 mb51.96 CH3F@44,59 # uJKM&5u100& 0.54~128 kV/cm !0.13 217 ma51.86 uJKM&5u16171& 1.05~118 kV/cm !0.27 99 CO@48# ua3PV51,J51,MV521& 2 1.71 0.53 89 ma51.37 ua3PV52,J52,MV524& 2 2.87 0.89 63 H2CO@61,58,60 # uJtM&5u111& 6 1.44 ~144 kV/cm !0.42 130 ma52.34 D2CO@61# uJtM&5u111& 6 1.11 ~112 kV/cm !0.30 155 ma52.34 H2O uJtM&5u111& 6 0.45 0.22 1081 mb51.82 D2O uJtM&5u111& 6 0.72 0.31 667 mb51.85 HDO uJtM&5u111& 12 0.73 0.36 558 ma50.66 mb51.73 HCN u(v1,v2l,v3),J,M&5u(0,00,0),1,0 & 6 0.92 ~144 kV/cm !0.29 177 ma53.01 u(v1,v2l,v3),J,Ml&5u(0,11,0),1, 21& 12 1.80 ~136 kV/cm !0.59 79 u(v1,v2l,v3),J,Ml&5u(0,11,0),2, 21& 12 2.58 0.83 150 u(v1,v2l,v3),J,Ml&5u(0,11,0),2, 22& 12 1.66 ~200 kV/cm !0.53 109 LiH uX1S1,J51,M50& 8 3.11 3.42 45 ma55.88 LiD uX1S1,J51,M50& 12 2.67 2.62 34 NH ua1D,J52,M52& 12 3.34 1.94 48 ma51.49 14NH3@62,63 # uJKM&5u16171& 12 2.11 1.09 79 mc51.47 15NH3 uJKM&5u16171& 8 2.11 1.03 84 mc51.47 14ND3@62# uJKM&5u16171& 48 2.29 1.00 65 mc51.50 15ND3 uJKM&5u16171& 32 2.29 0.95 68 mc51.50 NO@59,64 # uX2P1/2,J51/2,MV521/4& 6 0.17 0.05 1179 ma50.16 N2O@59,65 # u(v1,v2l,v3),J,Ml&5u(0,11,0),1, 21& 18 0.26 0.05 1041 ma50.17 OCS @63,66 # u(v1,v2l,v3),J,M&5u(0,00,0),1,0 & 1 0.13 ~82 kV/cm !0.02 1172 ma50.72 u(v1,v2l,v3),J,M&5u(0,00,0),2,0 & 1 0.43 0.06 1407 u(v1,v2l,v3),J,Ml&5u(0,11,0),1, 21& 2 0.25 ~84 kV/cm !0.04 647 ma50.70 u(v1,v2l,v3),J,Ml&5u(0,11,0),2, 21& 2 0.56 0.08 759 u(v1,v2l,v3),J,Ml&5u(0,11,0),2, 22& 2 0.24 ~120 kV/cm !0.04 779 OH@67,68 # uX2P3/2,J53/2,MV529/4& 4 3.22 1.66 56 ma51.67 OD@68# uX2P3/2,J53/2,MV529/4& 6 3.22 1.57 58 ma51.65 SH@57,69 # uX2P3/2,J53/2,MV529/4& 4 1.51 0.40 227 ma50.76 SD uX2P3/2,J53/2,MV529/4& 6 1.49 0.38 235 ma50.76 SO2@60,62 # uJtM&5u100& 1 1.47 0.21 343 mb51.59 aIncluding mdegeneracy. bWhen the Stark shift reaches a maximum at lower values of the electric field, this maximum shift is given, together with the value of the electric field ~in brackets !.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416 053416-18

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0018

along or against an external electric field, with coupling be- tween bits via the electric dipole-dipole interaction @91#. ACKNOWLEDGMENTS This work is part of the research program of the Stichting voor Fundamenteel Onderzoek der Materie ~FOM !, which isfinancially supported by the Nederlandse Organisatie voor Wetenschappelijk Onderzoek ~NWO !. The research of R.T.J. has been made possible by the support from the Royal Neth-erlands Academy of Arts and Sciences. We acknowledge theexpert technical assistance of A.J.A. van Roij. We thankP.H.M. Smeets, B. Sartakov, and G. Berden for support in theexperiments and for helpful discussions. @1#J.M. Doyle and B. Friedrich, Nature ~London !401, 749 ~1999!. @2#D. Herschbach, Rev. Mod. Phys. 71, S411 ~1999!. @3#C.J. Williams and P.S. Julienne, Science 287, 986 ~2000!. @4#J.T. Bahns, P.L. Gould, andW.C. Stwalley,Adv.At., Mol., Opt. Phys.42, 171 ~2000!. @5#B.G. Levi, Phys. Today 53~19!,4 6~2000!. @6#H.R. Thorsheim, J. Weiner, and P.S. Julienne, Phys. Rev. Lett. 58, 2420 ~1987!. @7#Y.B. Band and P.S. Julienne, Phys. Rev. A 51, R4317 ~1995!. @8#A.P. Mosk, M.W. Reynolds, T.W. Hijmans, and J.T.M. Wal- raven, Phys. Rev. Lett. 82, 307 ~1999!. @9#N. Herschbach, P.J.J. Tol, W. Vassen, W. Hogervorst, G. Woes- tenenk, J.W. Thomsen, P. van der Straten, and A. Niehaus,Phys. Rev. Lett. 84, 1874 ~2000!. @10#W.I. McAlexander, E.R.I.Abraham, N.W.M. Ritchie, C.J. Wil- liams, H.T.C. Stoof, and R.G. Hulet, Phys. Rev. A 51, R871 ~1995!. @11#P.D. Lett, K. Helmerson, W.D. Phillips, L.P. Ratliff, S.L. Rolston, and M.E. Wagshul, Phys. Rev. Lett. 71, 2200 ~1993!. @12#H. Wang, P.L. Gould, and W.C. Stwalley, Phys. Rev. A 53, R1216 ~1996!. @13#A.N. Nikolov, E.E. Eyler, X.T. Wang, J. Li, H. Wang, W.C. Stwalley, and P.L. Gould, Phys. Rev. Lett. 82, 703 ~1999!. @14#G. Zinner,T. Binnewies, F. Riehle, and E.Tiemann, Phys. Rev. Lett.85, 2292 ~2000!. @15#J.D. Miller, R.A. Cline, and D.J. Heinzen, Phys. Rev. Lett. 71, 2204 ~1993!. @16#C. Gabbanini, A. Fioretti, A. Lucchesini, S. Gozzini, and M. Mazzoni, Phys. Rev. Lett. 84, 2814 ~2000!. @17#A. Fioretti, D. Comparat,A. Crubellier, O. Dulieu, F. Masnou- Seeuws, and P. Pillet, Phys. Rev. Lett. 80, 4402 ~1998!. @18#R. Wynar, R.S. Freeland, D.J. Han, C. Ryu, and D.J. Heinzen, Science287, 1016 ~2000!. @19#J.M. Gerton, D. Strekalov, I. Prodan, and R.G. Hulet, Nature ~London !408, 692 ~2000!. @20#J.P. Shaffer,W. Chalupczak, and N.P. Bigelow, Phys. Rev. Lett. 82, 1124 ~1999!. @21#U. Schlo¨der, C. Silber, and Z. Zimmermann, Appl. Phys. B: Lasers Opt. 73, 801 ~2001!. @22#T. Takekoshi, B.M. Patterson, and R.J. Knize, Phys. Rev. Lett. 81, 5105 ~1998!. @23#N. Vanhaecke, W. de Souza Melo, B. Laburthe Tolra, D. Com- parat, and P. Pillet, Phys. Rev. Lett. ~to be published !. @24#J.D. Weinstein, R. deCarvalho, T. Guillet, B. Friedrich, and J.M. Doyle, Nature ~London !395, 148 ~1998!. @25#R. deCarvalho, J.M. Doyle, B. Friedrich, T. Guillet, J. Kim, D. Patterson, and J.D. Weinstein, Eur. Phys. J. D 7, 289 ~1999!.@26#J. G. King and J. R. Zacharias, Quarterly Progress Report, Research Laboratory of Electronics, Massachusetts Institute ofTechnology. No. 48, 1958 ~unpublished !. @27#R. Golub, Ph.D. thesis, Massachusetts Institute of Technology, 1967. @28#D. Auerbach, E.E.A. Bromberg, and L. Wharton, J. Chem. Phys.45, 2160 ~1966!. @29#R. Wolfgang, Sci. Am. 219~4!,4 4~1968!. @30#E.E.A. Bromberg, Ph.D. thesis, University of Chicago, 1972. @31#H.L. Bethlem, G. Berden, and G. Meijer, Phys. Rev. Lett. 83, 1558 ~1999!. @32#J.A. Maddi, T.P. Dinneen, and H. Gould, Phys. Rev. A 60, 3882 ~1999!. @33#H.L. Bethlem, G. Berden, A.J.A. van Roij, F.M.H. Crompvo- ets, and G. Meijer, Phys. Rev. Lett. 84, 5744 ~2000!. @34#A.J. Lichtenberg, in Phase-Space Dynamics of Particles ~Wiley, New York, 1969 !. @35#H.L. Bethlem, G. Berden, F.M.H. Crompvoets, R.T. Jongma, A.J.A. van Roij, and G. Meijer, Nature ~London !406, 491 ~2000!. @36#F.M.H. Crompvoets, H.L. Bethlem, R.T. Jongma, and G. Meijer, Nature ~London !411, 174 ~2001!. @37#V. I. Veksler, J. Phys. ~USSR !9, 153 ~1945!. @38#E.M. McMillan, Phys. Rev. 68, 143 ~1945!. @39#D.A. Edwards and M.J. Syphers, in An Introduction to the Physics of High EnergyAccelerators ~Wiley, NewYork, 1993 !. @40#D.A. Dahl, Simion 3D Version 6.0 ~Idaho National Engineer- ing Laboratory, Idaho Falls, 1995 !. @41#W.H. Wing, Prog. Quantum Electron. 8, 181 ~1984!. @42#H.L. Bethlem, A.J.A. van Roij, R.T. Jongma, and G. Meijer, Phys. Rev. Lett. 88, 133003 ~2002!. @43#J. van Veldhoven, R.T. Jongma, B. Sartakov, W.A. Bongers, and G. Meijer ~unpublished !. @44#P.W. Harland, W.-P. Hu, C. Vallance, and P.R. Brooks, Phys. Rev. A60, 3138 ~1999!. @45#D.R. Miller, in Atomic and Molecular Beam Methods , edited by G. Scoles ~Oxford University Press, New York, 1988 !, Vol. I. @46#R.V.Latham,in HighVoltageVacuumInsulation:ThePhysical Basis ~Academic Press, London, 1981 !. @47#J. Reuss, in Atomic and Molecular Beam Methods ~Ref.@45#!. @48#R.T. Jongma, Th. Rasing, and G. Meijer, J. Chem. Phys. 102, 1925 ~1995!. @49#S.Y. Lee, in Accelerator Physics ~World Scientific, Singapore, 1999!. @50#O.J. Luiten, M.W. Reynolds, and J.T.M. Walraven, Phys. Rev. A53, 381 ~1996!. @51#W.H. Wing, Phys. Rev. Lett. 45, 631 ~1980!.DECELERATION AND TRAPPING OF AMMONIA USING. . . PHYSICAL REVIEW A 65053416 053416-19

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0019

@52#W. Paul, Rev. Mod. Phys. 62, 531 ~1990!. @53#M.N.R. Ashfold, R.N. Dixon, N. Little, R.J. Stickland, and C.M. Western, J. Chem. Phys. 89, 1754 ~1988!. @54#S.Y.T. van de Meerakker, R.T. Jongma, H.L. Bethlem, and G. Meijer, Phys. Rev. A 64, 041401 ~R!~2001!. @55#M.A. Weibel, T.D. Hain, and T.J. Curtiss, J. Chem. Phys. 108, 3134 ~1998!. @56#M. Takezaki, H. Ohoyama, T. Kasai, and K. Kuwata, Laser Chem.15,1 1 3 ~1995!. @57#M.A. Weibel, T.D. Hain, and T.J. Curtiss, J. Vac. Sci. Technol. A15, 2238 ~1997!. @58#T.D. Hain, R.M. Moision, and T.J. Curtiss, J. Chem. Phys. 111, 6797 ~1999!. @59#D.H. Parker, H. Jalink, and S. Stolte, J. Phys. Chem. 91, 5427 ~1987!. @60#E.M. Jones and P.R. Brooks, J. Chem. Phys. 53,5 5~1970!. @61#N.F. van Hulst, J.J. ter Meulen, and A. Dymanus, J. Chem. Phys.86, 1407 ~1987!. @62#N.I. Butkovskaya, M.N. Larichev, I.O. Leipunskii, I.I. Moro- zov, and V.L. Tal’rose, Chem. Phys. 12, 267 ~1976!. @63#S.R. Gandhi, Q.-X. Xu, T.J. Curtiss, and R.B. Bernstein, J. Phys. Chem. 91, 5437 ~1987!. @64#J.J. van Leuken, F.H.W. van Amerom, J. Bulthuis, J.G. Snijders, and S. Stolte, J. Phys. Chem. 99,1 55 7 3 ~1995!. @65#H. Jalink, F. Harren, D. van den Ende, and S. Stolte, Chem. Phys.108, 391 ~1986!. @66#S.R. Gandhi and R.B. Bernstein, J. Chem. Phys. 87, 6457 ~1987!. @67#K. Schreel, J. Schleipen, A. Eppink, and J.J. ter Meulen, J. Chem. Phys. 99, 8713 ~1993!. @68#T.D. Hain, M.A. Weibel, K.M. Backstrand, and T.J. Curtiss, J. Phys. Chem. A 101, 7674 ~1997!. @69#T. Kasai, K. Ohashi, H. Ohoyama, and K. Kuwata, Chem. Phys. Lett. 127, 581 ~1986!.@70#M.C. van Beek and J.J. ter Meulen, Chem. Phys. Lett. 337, 237 ~2001!. @71#S.K. Sekatskiı ˘, Pis’ma Zh. Eksp. Teor. Fiz. 62, 900 ~1995! @JETP Lett. 62, 916 ~1995!#. @72#H.J. Loesch, Chem. Phys. 207, 427 ~1996!. @73#R.T. Jongma, G. von Helden, G. Berden, and G. Meijer, Chem. Phys. Lett. 270, 304 ~1997!. @74#H.J. Loesch and B. Scheel, Phys. Rev. Lett. 85, 2709 ~2000!. @75#D.P. Katz, J. Chem. Phys. 107, 8491 ~1997!. @76#F. Shimizu and M. Morinaga, Jpn. J. Appl. Phys., Part 2 31, L1721 ~1992!. @77#E. Peik, Eur. Phys. J. D 6, 179 ~1999!. @78#V. Vuletic´and S. Chu, Phys. Rev. Lett. 84, 3787 ~2000!. @79#E.A. Hinds, Phys. Scr. T70,3 4~1997!. @80#Ch. Daussy, T. Marrel, A. Amy-Klein, C.T. Nguyen, Ch.J. Borde´, and Ch. Chardonnet. Phys. Rev. Lett. 83, 1554 ~1999!. @81#J.K. Webb, M.T. Murphy, V.V. Flambaum, V.A. Dzuba, J.D. Barrow, C.W. Churchill, J.X. Prochaska, and A.M. Wolfe,Phys. Rev. Lett. 87, 091301 ~2001!. @82#N. Balakrishnan,A. Dalgarno, and R.C. Forrey, J. Chem. Phys. 113, 621 ~2000!. @83#W. Ketterle and N.J. van Druten,Adv.At., Mol., Opt. Phys. 37, 181~1996!. @84#J.L. Bohn, Phys. Rev. A 63, 052714 ~2001!. @85#E.A. Cornell, C. Monroe, and C.E. Wieman, Phys. Rev. Lett. 67, 2439 ~1991!. @86#G. Modugno, G. Ferrari, G. Roati, R.J. Brecha,A. Simoni, and M. Inguscio, Science 294, 1320 ~2001!. @87#M. Baranov, Ł. Dobrek, K. Go ´ral, L. Santos, and M. Lewen- stein, e-print cond-mat/0201100. @88#L. Santos, G.V. Shlyapnikov, P. Zoller, and M. Lewenstein, Phys. Rev. Lett. 85, 1791 ~2000!. @89#M.A. Baranov, M.S. Mar’enko, Val.S. Rychkov, and G.V. Shlyapnikov, e-print cond-mat/0109437. @90#G.V. Shlyapnikov ~private communications !. @91#D. DeMille, Phys. Rev. Lett. 88, 067901 ~2002!.HENDRICK L. BETHLEM et al. PHYSICAL REVIEW A 65053416 053416-20

6a29bdbbc33cbdac22141ee1826fcd55640e89b8

page_0020

Targeted Protein Degradation by Small Molecules Daniel P. Bondeson and Craig M. Crews Department of Molecular, Cellular, and Developmental Biology, Department of Chemistry, and Department of Pharmacology, Yale University, New Haven, Connecticut 06511 Abstract Protein homeostasis networks are highly regulated systems responsible for maintaining the health and productivity of cells. Whereas therapeutics have been developed to disrupt protein homeostasis, more recently identified techniques have been used to repurpose homeostatic networks to effect degradation of disease-relevant proteins. Here, we review recent advances in the use of small molecules to degrade proteins in a selective manner. First, we highlight all-small- molecule techniques with direct clinical application. Second, we describe techniques that may find broader acceptance in the biomedical research community that require little or no synthetic chemistry. In addition to serving as innovative research tools, these new approaches to control intracellular protein levels offer the potential to develop novel therapeutics targeting proteins that are not currently pharmaceutically vulnerable. Keywords protein degradation; ubiquitin proteasome system; chemical knockdown; PROTACs; IMiDs INTRODUCTION: THE LIMITS OF INHIBITORS AND GENETIC KNOCKDOWNS Much of biomedical research consists of interpreting data gleaned from the functional inhibition of proteins. Chemical inhibitors, genetic knockdown and knockout models, and mutagenesis screens have elucidated many of the complex signaling networks in biology. Furthermore, protein inhibition is a mainstay of drug development and the many beneficial therapeutics that improve human health. Despite this, modern technologies for perturbing protein function clearly have their limitations. Chemical inhibitors, typically small molecules that bind to an enzyme or receptor active site, function by locking their target protein in a state that prevents it from performing its function. This mode of action has two limitations. First, these inhibitors are active only while bound to the protein of interest (POI), necessitating high levels of compound that may elicit unwanted off-target effects ( 1). Second, these inhibitors must be able to bind to an active site or allosteric site of a protein to inhibit the protein’s function. DISCLOSURE STATEMENT C.M.C. is a founder, consultant, and shareholder at Arvinas, LLC. HHS Public Access Author manuscript Annu Rev Pharmacol Toxicol . Author manuscript; available in PMC 2017 September 06. Published in final edited form as: Annu Rev Pharmacol Toxicol . 2017 January 06; 57: 107–123. doi:10.1146/annurev- pharmtox-010715-103507. Author Manuscript Author Manuscript Author Manuscript Author Manuscript

cfa9da281941a329299bc535ce3bb85c2ee70727

page_0000

These requirements exclude the approximately 80% of the human proteome classified as undruggable ( 2). Although these limitations hamper chemical inhibitors, knockdown strategies provide an efficacious alternative. RNA interference (RNAi), antisense oligonucleotides, and gene- editing techniques, most notably CRISPR/Cas9, provide means for circumventing these issues. Acting at the genetic level, these techniques lack the limitations regarding a target’s druggability. Initial excitement for rapidly developing a genetic knockdown for a target has subsided owing to nucleic acid delivery issues, but efforts are still being made using nanomaterials and other packaging technologies to develop these technologies into therapeutics ( 3). Additionally, these techniques are typically irreversible (limiting their utility as research tools) and often contain unwanted off-target effects (thus narrowing the therapeutic index and obfuscating interpretation of data). Here, we review technologies with the temporal control and pharmaceutical properties of small molecules capable of knocking down POI levels. These chemical knockdown strategies provide therapeutic options that extend beyond traditional druggable space but still share similar pharmacokinetic properties with typical small-molecule drugs. When and How Are Proteins Normally Degraded? After translation, polypeptides are processed, folded, chaperoned, and modified before attaining their final functional state. Additionally, because many proteins are required only transiently, cells possess elegant systems for the removal of unwanted or damaged proteins. The largest protein disposal system is the ubiquitin proteasome system, which comprises nearly 1% of cellular mass. This system consists of a cascade of enzymes that activate, conjugate, and ligate the 76–amino acid protein ubiquitin onto a lysine residue of a protein. Because ubiquitin contains several lysine residues itself, a chain of polyubiquitin can be assembled: Chains coupled to different ubiquitin lysine residues act as signaling motifs for various processes. Most apropos, certain linkages send the protein to the 26S proteasome, a large protease complex that recognizes, unfolds, and degrades ubiquitinated proteins. The molecular determinants of which proteins are targeted for ubiquitination are defined by a class of enzymes known as E3 ubiquitin ligases. In this review, we highlight studies using a particular E3 ligase family, the cullin ring E3 ligases (CRLs); these are large scaffolding complexes in which one portion recruits the substrate and brings it into close proximity to the reactive E2 ubiquitin for ubiquitination. Several excellent reviews have been written on this subject ( 4–6). Another protein disposal system is autophagy. In this process, the protein to be removed is sent to the lysosome, an organelle that contains an acidic environment and up to 50 different kinds of enzymes focused on degrading and processing these substrates ( 7). Although ubiquitination can signal autophagy, more commonly, the regulated autophagy system employs specific chaperones to bring targeted proteins to the lysosome ( 8).Bondeson and Crews Page 2 Annu Rev Pharmacol Toxicol . Author manuscript; available in PMC 2017 September 06. Author Manuscript Author Manuscript Author Manuscript Author Manuscript

cfa9da281941a329299bc535ce3bb85c2ee70727

page_0001

Outline This review is divided into two parts. In the first, we review those chemical knockdown strategies that have the most direct clinical application. These small molecule–based approaches are capable of degrading target proteins without requiring any genetic manipulation. These include selective estrogen downregulators, immunomodulatory drugs (IMiDs) or cereblon binding molecules, proteolysis targeting chimeras (PROTACs) against a variety of targets, and hydrophobic tags for androgen receptor (AR) and Her3. In the second half, more generalized strategies for manipulating levels of any targeted protein are presented. These techniques require the fusion of a second protein to the POI, which extends the technique into broader applications: for example, when no ligand is available for a POI. Here, auxin-inducible degrons, PROTACs that target modular fusion proteins, and several other techniques are reviewed. SELECTIVE ESTROGEN RECEPTOR DOWNREGULATORS Selective estrogen receptor downregulators (SERDs) were among the first class of compounds identified that have the added benefit of inducing degradation of their target protein. Estrogen receptor α (ERα) is a well-known oncogenic driver for metastatic breast cancer ( 9). Although ER α modulators have been in the clinic since tamoxifen was first approved by the US Food and Drug Administration (FDA) in 1977, spurious ER α activation in various tissues led to a need for pure antiestrogens. Fulvestrant (ICI 182,780 or Faslodex™) was first described in the early 1990s as a pure antagonist capable of overcoming these partial agonistic issues. Its therapeutic mechanism was soon attributed to its ability to decrease intracellular ER α levels ( 10, 11), but despite approval by the FDA in 2002, fulvestrant suffers from poor bioavailability and is administered by monthly intramuscular injection. Given these issues, new SERD compounds have been developed recently, with several entering clinical trials. The most advanced compound is ARN-810 (Figure 1 a), developed by Seragon Pharmaceuticals. With improved oral bioavailability, this compound shows promising preclinical results in mice ( 12–14) and is now in Phase I clinical trials. Other compounds with oral bioavailability and high potency have also been described by AstraZeneca ( 15–19), Pfizer ( 20), and other research groups ( 21). SERDs may be oldest application of induced protein degradation, but the mechanism by which ER degradation is achieved is not well understood. It is thought that upon binding, a SERD induces conformational changes of the protein, exposing novel hydrophobic motifs that can be recognized by chaperones and trigger degradation ( 22, 23). Only one SERD- ERα crystal structure has been solved, but the mechanism by which that compound induces ERα degradation is not shared with other SERD compounds. Because the ligand-ER α structures are thought to differ, clear routes to rationally design SERDs are thus not obvious. Indeed, even in the most recent reports, some groups note surprising differences in degradation with minor structural changes to their SERDs ( 16). However, high-throughput screening assays have been developed that identify compounds that either unveil the relevant hydrophobic surfaces ( 23) or assess intracellular levels of ER α rapidly ( 12, 15).Bondeson and Crews Page 3 Annu Rev Pharmacol Toxicol . Author manuscript; available in PMC 2017 September 06. Author Manuscript Author Manuscript Author Manuscript Author Manuscript

cfa9da281941a329299bc535ce3bb85c2ee70727

page_0002

IMMUNOMODULATORY DRUGS The phthalimide class of compounds known as IMiDs has a storied history. First discovered and described in the 1950s and 1960s as a sedative, thalidomide was discovered to be a potent teratogen. However, since then, it and its close analogs have been repurposed as potent anticancer agents ( 24). Given the excitement regarding thalidomide’s anticancer activity, particularly for multiple myeloma, medicinal chemistry efforts were undertaken to develop related compounds possessing these beneficial activities. Thalidomide itself is approved for newly diagnosed multiple myeloma, and three other thalidomide analogs are of particular interest. First, lenalidomide (Revlimid, CC-5013, Figure 1 b) is approved for the treatment of relapsed multiple myeloma, myelodysplastic syndrome (MDS, particularly those genotyped to a chromosomal 5 truncation) ( 25), and mantle cell lymphoma. It is also in Phase III trials to expand its approval within these diseases as well as for the treatment of acute myeloid leukemia and chronic lymphoblastic leukemia. Second, pomalidomide (Pomalyst, CC-4047) has also been approved for relapsed multiple myeloma ( 26). And third, a more recently described compound, CC-122, also shows activity as a pleiotropic pathway modifier and is in Phase I trials for multiple myeloma, diffuse large B cell lymphoma, chronic lymphoblastic leukemia, and several solid tumors ( 27). How these compounds have such potent activity has only recently been elucidated. In 2010, a major step toward understanding IMiD action was made upon identification of cereblon as a major target of thalidomide teratogenicity. Using a chemoproteomic probe of immobilized thalidomide, the authors identified cereblon as a substrate adapter for the CRL4a ubiquitin ligase complex, whose auto-ubiquitination is inhibited by thalidomide. Importantly, mutations in cereblon that block thalidomide binding also inhibit models of thalidomide teratogenicity in chicken and zebrafish ( 28). Another major breakthrough in the understanding of IMiD mechanisms was demonstrated in three elegant studies ( 29–31) that used different profiling techniques to identify the transcription factors Ikaros and Aiolos as proteins whose ubiquitination is increased upon IMiD treatment. Whereas IMiDs decrease cereblon auto-ubiquitination, they also increase Ikaros ubiquitination. Cells expressing an IMiD-resistant Ikaros mutant (Q147H) are resistant to IMiD-induced cytotoxicity ( 29, 30). More recently, casein kinase 1 α (CK1α) was similarly identified as a protein selectively degraded by lenalidomide ( 27, 32). Although these results collectively implicated cereblon and Ikaros/CK1 α as the target of IMiD action, the mechanism of this cytotoxicity is unclear. Ikaros family members activate the IRF4 locus, a gene to which myeloma cells are addicted ( 33, 34), although not all IMiD- sensitive cell lines have decreased IRF4 levels, indicating IRF4 might not be necessary for this cytotoxicity ( 30). In del(5q) MDS, haploinsufficiency of CK1 α leads to hyperproliferation, whereas homozygous loss leads to apoptosis, a finding that helps explain lenalidomide sensitivity in CK1 α (32, 35). From a biochemical standpoint, recent studies have also led to an understanding of how IMiDs recruit these new substrates to cereblon. Crystal structures of the cereblon-IMiD complex ( 36, 37) and the ternary complex between cereblon, lenalidomide, and CK1 α (38) have confirmed that the IMiD glutarimide moiety binds to a hydrophobic cavity in cereblon, Bondeson and Crews Page 4 Annu Rev Pharmacol Toxicol . Author manuscript; available in PMC 2017 September 06. Author Manuscript Author Manuscript Author Manuscript Author Manuscript

cfa9da281941a329299bc535ce3bb85c2ee70727

page_0003

whereas the phthalimide ring is free to form contacts with the substrate. The phthalimide ring, in combination with local residues from cereblon, creates a surface that binds to a remarkably small beta hairpin loop on CK1 α. Furthermore, this hairpin shares structural, but not sequence, homology to Ikaros, providing mechanistic data for the selectivity of the different IMiD compounds for their respective targets. In conclusion, IMiDs are an intriguing class of compounds with surprising mechanisms of action. Because minor differences among family members affect substrate binding, it will be interesting to identify other proteins that may be targeted by related compounds. In addition, given the recently discovered structural information on CK1 α recruitment, will it be possible to design IMiD-like compounds that recruit particular hairpin motifs? PROTACS Although IMiDs and SERDs have found clinical success, the applicability of the system is currently limited. For example, rationally designing a thalidomide analog to target a specific protein for degradation would be difficult given the small structural determinant on the potential substrate that would be challenging to predict and exploit. Previous Generations of PROTACs For the past 15 years, our lab has developed the PROTAC technology, which lacks these limitations and is able to induce selective protein degradation without the need for genetic manipulation. PROTACs are heterobifunctional molecules that have discrete binding moieties for the substrate of interest and for an E3 ligase connected by a chemical linker. The first PROTAC, developed in collaboration with the Deshaies group at CalTech ( 39), consisted of the natural product ovalicin and a peptidic ligand for the CRL1 F-box protein β- TRCP. This initial PROTAC demonstrated ternary complex (substrate–PROTAC–E3 ligase) formation, ubiquitination activity, and limited degradation of its target protein in Xenopus extracts ( 40). Since this first publication, our group and others have published approximately 30 papers validating this technology. These studies have explored both the limitations and potential of the PROTAC technology, and several key lessons have been learned. First, different E3 ligases can be hijacked by PROTACs for selective protein degradation. β-TRCP, MDM2 ( 41), CIAP ( 42), and von Hippel–Lindau (VHL) ( 43) have all been employed for induced protein ubiquitination using a heterobifunctional dimer approach. Although they are not technically PROTACs, other bifunctional peptides have been used to direct POIs to the lysosome for degradation ( 44). Second, small molecules have been employed for either binding moiety. The MDM2 inhibitor Nutlin ( 41) or the IAP ligand bestatin ( 45–47) have both been used in PROTACs to engage their cognate E3 ligases. Likewise, small-molecules have also been used as substrate- targeting ligands [e.g., small-molecule agonists of the retinoic acid receptor ( 42), fumagillin and ovalicin for methionyl aminopeptidase 2 ( 48)]. Third, and disappointingly, these compounds have been very limited in their potency. Most of these early-generation compounds are, at best, active in the low-micromolar range with only partial degradation of the POI. Because these compounds are large and charged (or at Bondeson and Crews Page 5 Annu Rev Pharmacol Toxicol . Author manuscript; available in PMC 2017 September 06. Author Manuscript Author Manuscript Author Manuscript Author Manuscript

cfa9da281941a329299bc535ce3bb85c2ee70727

page_0004

least highly hydrophilic), cell permeability is a key contributor to this lack of potency, although the low affinity of these peptides for their targets is also likely a contributing factor. Another issue, which is only now being appreciated, is the role that proper linker geometry has in PROTAC potency. This is discussed in more detail below. From a technological standpoint, four papers published in May and June of 2015 made significant advances toward the therapeutic application of PROTACs. Next-Generation PROTACs To develop potent PROTACs, high-affinity small-molecule E3 ligase ligands had to be developed. The E3 ubiquitin ligase CRL2VHL is responsible for the regulated ubiquitination of hypoxia inducible factor 1 α (HIF1α). This interaction is very specific: A specific hydroxylation event on a single proline residue is sufficient to mediate the VHL-HIF1 α interaction ( 49). Given this concise molecular determinant for binding, our lab sought to develop a small-molecule VHL ligand for use in PROTACs based on the hydroxyproline residue. Using a combination of in silico and fragment-based screening, an initial VHL ligand with low micromolar affinity was further developed into a high-affinity ligand with a Kd of 180 nM ( 50, 51). With this VHL ligand in hand, three different classes of VHL-targeting PROTACs were made to target the bromodomain-containing protein 4 (BRD4), the receptor interacting serine/threonine protein kinase 2 (RIPK2), and the nuclear hormone receptor estrogen- related receptor α (ERRα) (52, 53). BRD4 is a reader protein of epigenetic marks, and although it is not mutated in cancers, inhibition of BRD4 has been shown to decrease expression of the oncogene CMYC , leading to selective killing of cancers addicted to c-Myc (54, 55). RIPK2 is strongly implicated in autoimmune diseases such as Crohn’s disease as well as cancer, and it functions through a combination of enzymatic and scaffolding roles (56, 57). ERRα is known as a master regulator of metabolic homeostasis ( 58) and some cancers ( 59). Based on the importance of these targets, VHL-based PROTACs were synthesized targeting RIPK2 (PROTAC_RIPK2; Figure 1 c), ERRα (PROTAC_ERR α), and BRD4 (MZ1) to VHL. Each of these ligands was chosen for their high affinity and selectivity, as well as for their known protein-ligand structural data. In designing a PROTAC, a key decision is the attachment point of the linker; solvent-exposed surfaces of the ligand are necessary. For RIPK2, the amino-quinoline-based kinase inhibitor ligand bound with near 100-fold selectivity over RIPK3, and unpublished structural data indicated a solvent-exposed region that could be modified. Although several ERR α ligands are known, one compound has nearly 100-fold selectivity over the closely related ERR γ, a Kd of approximately 40 nM, and the appropriate structural data for the design of the linker attachment point ( 60). Similarly, the choice of OTX015 as the BRD4 ligand was due to selectivity for BRD2/3/4, high potency, and a known solvent-exposed region for linker attachment ( 61). Linker attachment at a solvent-exposed region on the target protein ligand is critical, but the optimal linker length and composition is more difficult to discern, as discussed below. As these PROTACs demonstrate well, no one linker is optimal for every target protein: ERR α was degraded with a 6-atom linker, RIPK2 with a 14-atom linker, and BRD4 with a 10-atom linker. In each Bondeson and Crews Page 6 Annu Rev Pharmacol Toxicol . Author manuscript; available in PMC 2017 September 06. Author Manuscript Author Manuscript Author Manuscript Author Manuscript

cfa9da281941a329299bc535ce3bb85c2ee70727

page_0005

case, near-complete removal of the protein was achieved at PROTAC concentrations as low as 5 nM ( 53). Whereas the VHL ligand on which these PROTACs are based was generated through a traditional structure-based drug design approach, other ligands with selectivity and high affinity for an E3 ligase were discovered recently: IMiDs. The available structural data for these IMiDs ( 36, 37) was used to generate the PROTACs ARV-825 ( 62) and dBet1 ( 63), which target BRD4 for ubiquitination by the E3 ligase CRL4aCRBN. ARV-825 and dBet1 are based on the BRD4-selective inhibitors OTX-15 and JQ1, respectively ( 55), coupled to an IMiD phthalimide. Linker-wise, the two compounds differ significantly; ARV-825 contains a 14-atom PEG linker, whereas dBet1 has a 7-atom, primarily alkyl linker. These linker differences may be responsible for the significant differences in intracellular potencies between these two PROTAC molecules: Greater than 90% BRD4 degradation is observed at 1 nM and 0.5 μM for ARV-825 and dBet1, respectively. On Choosing PROTAC Warheads and Linkers Although the differences between these compounds are certainly interesting, hard and fast principles for effective PROTAC design remain elusive. To aid in our understanding of so- called linkerology, our lab recently explored the effects of different variables on PROTAC efficacy and target protein selectively, using two E3 ligands, three targeting ligands, and four different linkers spanning a diversity of chemical space ( 64). The goal was the design of a PROTAC capable of degrading the oncoprotein BCR/Abl, with the hope that BCR/Abl degradation would possibly eliminate kinase-independent functions of the protein ( 65, 66). This study of BCR/Abl PROTAC development yielded several interesting conclusions regarding the effects of the targeting warhead, E3 ligase, and linker. First, the choice of protein-targeting ligand has a large influence on PROTAC selectivity and degradation activity. The Abl tyrosine kinase inhibitor imatinib was unable to degrade Abl or BCR/Abl when incorporated into a PROTAC, whereas bosutinib-based PROTACs gave the most profound degradation of Abl, and dasatinib-based PROTACs were best at degrading BCR/ Abl. These differences might arise from differences in affinity, as imatinib has a much lower affinity than do the other compounds. However, because bosutinib and dasatinib bind with similar affinities, perhaps subtle structural changes in BCR/Abl change the efficiency of ubiquitin transfer. More studies are required to illustrate conclusively why this might be. Second, the PROTAC linker may influence cell permeability more than target-protein degradation; although linker changes did not influence whether BCR/Abl or Abl was degraded, they did affect the efficiency with which the substrate was degraded. Third, the E3 ligase being recruited can also significantly influence the PROTAC’s ability to degrade different substrates. In this study, only cereblon was able to degrade both BCR/Abl and Abl, whereas VHL was only able to degrade Abl efficiently. Several hypotheses could explain this rather curious result. For example, owing to steric differences between BCR/Abl (approximately 210,000 Da) versus c-Abl (110,000 Da), perhaps only Abl can fit into VHL complexes. Alternatively, whereas VHL seems to have the larger endogenous substrate Bondeson and Crews Page 7 Annu Rev Pharmacol Toxicol . Author manuscript; available in PMC 2017 September 06. Author Manuscript Author Manuscript Author Manuscript Author Manuscript

cfa9da281941a329299bc535ce3bb85c2ee70727

page_0006

(HIF1α = 95,000 Da versus MEIS2 = approximately 50,000 Da), perhaps CRL2VHL has a less flexible cullin domain than does CRL4aCRBN (67). The need for proper presentation of lysines on the target protein to the recruited E3 ligase may also play a key role. This hypothesis of ubiquitination zones for selecting particular lysine residues on the substrate has been reviewed recently ( 68). HYDROPHOBIC TAGGING Given the clinical success of fulvestrant, which mediates ER α degradation by exposing a hydrophobic patch on the surface of the protein, we hypothesized that a ligand for a POI could be functionalized similarly into a hydrophobic tag to mimic a partially unfolded state. In this way, a hydrophobically tagged protein would be recognized by the same cellular quality control that recognizes and discards terminally misfolded or unfolded proteins. Although we have published this strategy in model systems before, it has been employed recently to degrade endogenous proteins, [i.e., the pseudokinase Her3 ( 69) and the AR ( 70)] without the need for fusion protein genetic engineering. A similar system based on appending a large Boc 3Arg motif to a POI ligand has also been reported to induce selective degradation ( 71). Recently, however, it was shown that the Boc 3Arg motif inhibits global translation by blocking the mammalian target of rapamycin complex 1 pathway ( 72). How this data reconciles with the purported degradation caused by this ligand is unclear. Her3 Degradation Although Her3 is a member of the ErbB family of receptor tyrosine kinases, sharing high sequence similarity with the archetypical epidermal growth factor receptor, it differs from family members in that it lacks key catalytic residues in its active site, leading to loss of detectable kinase activity. Thus, researchers have proposed that Her3 functions primarily as a pseudokinase (i.e., a scaffolding protein rather than an active kinase). Given the difficulty of pharmacological targeting of pseudokinases using current small-molecule approaches, these proteins make attractive targets for strategies based on induced protein degradation. As a first step, a Her3 ligand was identified by screening a library of ATP-competitive compounds in a competitive time-resolved fluorescence energy transfer assay, and its potency was improved via the addition of an acrylate, thus generating a first-in-class, highly selective covalent ligand to Her3 ( 69, 73). Subsequent coupling of a hydrophobic adamantyl moiety produced a Her3 degrader compound that abrogated almost all Her3-dependent signaling in cultured tumor cells. Androgen Receptor Degradation Hydrophobic tagging technology has also been employed to degrade the AR, a well-defined oncology target ( 74). Like ER α, the AR drives the growth of many hormone-responsive tumors, particularly in prostate cancer, which is responsible for the second-highest cancer- related mortality rate for men in the western world. Although aromatase inhibitors and AR antagonists have been largely successful in treating early stages of prostate cancer, resistance develops in ways similar to ER α-targeting selective estrogen receptor modulators. By appending the alkylfluoryl chain of fulvestrant onto dihydrotestosterone, the first selective androgen receptor downregulator (SARD) compound was discovered ( 75). Other Bondeson and Crews Page 8 Annu Rev Pharmacol Toxicol . Author manuscript; available in PMC 2017 September 06. Author Manuscript Author Manuscript Author Manuscript Author Manuscript

cfa9da281941a329299bc535ce3bb85c2ee70727

page_0007