Dataset Viewer
id
stringlengths 30
36
| source
stringclasses 1
value | format
stringclasses 1
value | text
stringlengths 5
878k
|
|---|---|---|---|
no-problem/0001/cond-mat0001338.html | ar5iv | text | # The magnetic neutron scattering resonance of high-𝑇_c superconductors in external magnetic fields: an SO(5) study
\[
## Abstract
The magnetic resonance at 41 meV observed in neutron scattering studies of YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7</sub> holds a key position in the understanding of high-$`T_\mathrm{c}`$ superconductivity. Within the SO(5) model for superconductivity and antiferromagnetism, we have calculated the effect of an applied magnetic field on the neutron scattering cross-section of the magnetic resonance. In the presence of Abrikosov vortices, the neutron scattering cross-section shows clear signatures of not only the fluctuations in the superconducting order parameter $`\psi `$, but also the modulation of the phase of $`\psi `$ due to vortices. In reciprocal space we find that i) the scattering amplitude is zero at $`(\pi /a,\pi /a)`$, ii) the resonance peak is split into a ring with radius $`\pi /d`$ centered at $`(\pi /a,\pi /a)`$, $`d`$ being the vortex lattice constant, and consequently, iii) the splitting $`\pi /d`$ scales with the magnetic field as $`\sqrt{B}`$.
\]
Soon after the discovery of high-$`T_c`$ superconductivity in the doped cuprate compounds, its intimate relation to antiferromagnetism was realized. A key discovery in the unraveling of this relationship was the observation of the so called 41 meV magnetic resonance later also denoted the $`\pi `$ resonance. In inelastic neutron scattering experiments on YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7</sub> at temperatures below $`T_\mathrm{c}90\mathrm{K}`$, Rossat-Mignod et al. found a sharp peak at $`\mathrm{}\omega 41\mathrm{meV}`$ and $`𝐪=(\pi /a,\pi /a)`$, $`a`$ being the lattice constant of the square lattice in the copper-oxide planes. Later its antiferromagnetic origin was confirmed by Mook et al. in a polarized neutron scattering experiment and subsequently Fong et al. found that the magnetic scattering appears only in the superconducting state. Recently, Fong *et al.* have also observed the $`\pi `$ resonance in Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8+δ</sub>, which means that it is a general feature of high-$`T_c`$ superconductors and not a phenomenon restricted to YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7</sub>. This gives strong experimental evidence for the $`\pi `$ resonance being related to antiferromagnetic fluctuations within the superconducting state. Conversely, it may be noted that angular-resolved photoemission spectroscopy has shown how the single-particle gap within the antiferromagnetic state inherits the $`d`$-wave modulation of the superconducting state.
A number of different models have been proposed to explain the $`\pi `$ resonance. In particular, Zhang was inspired by the existence of antiferromagnetic fluctuations in the superconducting state to suggest a unified SO(5) theory of antiferromagnetism and $`d`$-wave superconductivity in the high-$`T_\mathrm{c}`$ superconductors. It is of great interest to extend the different theoretical explanations to make predictions for the behavior of the $`\pi `$ resonance *e.g.* in an applied magnetic field. An experimental test of such predictions will put important constraints on theoretical explanations of the $`\pi `$ resonance in particular and of high-$`T_c`$ superconductivity in general. In this paper we treat the $`\pi `$ resonance in the presence of an applied magnetic field within the SO(5) model.
Zhang proposed that the cuprates at low temperatures can be understood as a competition between $`d`$-wave superconductivity and antiferromagnetism of a system which at higher temperatures possesses SO(5) symmetry. The SO(5) symmetry group is the minimal group that contains both the gauge group U(1) \[$`=`$SO(2)\] which is broken in the superconducting state, and the spin rotation group SO(3) which is broken in the antiferromagnetic state. Furthermore, the SO(5) group also contains rotations of the superspin between the antiferromagnetic sector and the superconducting sector. The relevant order parameter is a real vector $`𝐧=(n_1,n_2,n_3,n_4,n_5)`$ in a five dimensional superspin space with a length which is fixed ($`\left|𝐧\right|^2=1`$) at low temperatures. This order parameter is related to the complex superconducting order parameter, $`\psi `$, and the antiferromagnetic order parameter, $`𝐦`$, in each copper-oxide plane as follows: $`\psi =fe^{i\varphi }=n_1+in_5`$ and $`𝐦=(n_2,n_3,n_4)`$. Zhang argued how in terms of the five dimensional superspin space one can construct an effective Lagrangian $`(𝐧)`$ describing the low energy physics of the $`t`$-$`J`$ limit of the Hubbard model.
Two comments are appropriate here. Firstly, we note that relaxing the constraint $`\left|𝐧\right|^2=1`$ in the bulk superconducting state will introduce high energy modes, but these can safely be ignored at low temperatures. Moreover, they do not alter the topology of vortices in the order parameter, which is our main concern. Secondly, one may worry that results obtained from a pure SO(5) model deviate substantially from those obtained from the recently developed, physically more correct projected SO(5) theory . However, the two models are only significantly different close to half filling, and our study concerns AF-modes in the bulk superconductor in a weak magnetic field, a state which although endowed with the topology of vortices is far from half filling. For simplicity, we thus restrict the calculations in this paper to the original form of the SO(5) theory.
In the superconducting state the SO(5) symmetry is spontaneously broken which leads to a “high” energy collective mode where the approximate SO(5) symmetry allows for rotations of $`𝐧`$ between the superconducting and the antiferromagnetic phases. These rotations have an energy cost $`\mathrm{}\omega _\pi `$ corresponding to the $`\pi `$ resonance and fluctuations in $`𝐧`$ will thus give rise to a neutron scattering peak at $`\mathrm{}\omega _\pi `$ which, through the antiferromagnetic part of the superspin, is located at $`𝐪=𝐐`$, where $`𝐐=(\pi /a,\pi /a)`$ is the antiferromagnetic ordering vector. The uniform superconducting state ($`f=1`$) can be characterized by a superspin $`𝐧=(f\mathrm{cos}\varphi ,0,0,0,f\mathrm{sin}\varphi )`$, and the $`\pi `$ mode is a fluctuation $`\delta 𝐧(t)(0,0,0,fe^{i\omega _\pi t},0)`$ around the static solution, where $`\widehat{𝐳}`$ has been chosen as an arbitrary direction for $`\delta 𝐦`$. In this case with $`f=1`$ we have $`\delta 𝐦e^{i\omega _\pi t}`$, i.e. a sharp peak at $`\omega =\omega _\pi `$ and $`𝐪=𝐐`$.
In the presence of an applied magnetic field, the superconductor will be penetrated by flux quanta, each forming a vortex with a flux $`h/2e`$ by which the complex superconducting order parameter $`\psi `$ acquires a phase shift of $`2\pi `$ when moving around the vortex. In YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7</sub> the vortices arrange themselves in a triangular vortex lattice with an area of the hexagonal unit cell given by $`𝒜=h/2eB`$ and consequently a lattice constant given by $`d=3^{1/4}\sqrt{h/eB}`$. In the work by Arovas et al., Bruus et al., and Alama et al. the problem of Abrikosov vortices was studied within the SO(5) model of Zhang. In the center of a vortex core, the superconducting part of the order parameter is forced to zero. This leaves two possibilities: i) either the vortex core is in a metallic normal state (as it is the case in conventional superconductors) corresponding to a vanishing superspin or ii) the superspin remains intact but is rotated from the superconducting sector into the antiferromagnetic sector. The prediction of the possibility of antiferromagnetically ordered insulating vortex cores is thus quite novel and allows for a direct experimental test of the SO(5) theory. However, the antiferromagnetic ordering of vortices is according to our knowledge still to be confirmed experimentally. In this paper we report a different consequence of the SO(5) theory in neutron scattering experiments; we consider the $`\pi `$ mode in the presence of vortices and show that the peak at $`𝐪=𝐐`$ splits into a ring with a radius $`\pi /d`$ centered at $`𝐪=𝐐`$ where it has zero amplitude. Consequently the splitting scales with magnetic field $`B`$ as $`\pi /d\sqrt{B}`$.
We start by considering just one vortex, then generalize the result to a vortex lattice. To make our calculations quantitative, we consider YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7</sub> for which $`a=3.8\mathrm{\AA }`$, $`\kappa 84`$, and $`\xi 16\mathrm{\AA }`$ for the lattice constant, the Ginzburg–Landau parameter, and the coherence length, respectively. The order parameter can be written in the form
$$𝐧(𝐫)=(f(r)\mathrm{cos}\varphi _𝐫,0,m(r),0,f(r)\mathrm{sin}\varphi _𝐫),$$
(1)
where $`\varphi _𝐫=\mathrm{arg}(𝐫)`$. The isotropy of the antiferromagnetic subspace allows us to choose $`𝐦`$ to lie in the $`y`$-direction without loss of generality. Static numerical solutions for $`f(r)`$ and thereby also $`m(r)`$ in the presence of a vortex are derived as described in Refs. . Due to the high value of $`\kappa `$ the absolute value $`f`$ of the superconducting order parameter $`\psi `$ increases from zero at the center of the vortex ($`r=0`$) to its bulk value ($`f=1`$) at a distance of the order $`\xi `$ from the center. The antiferromagnetic order parameter follows from $`f`$ since $`m=\sqrt{1f^2}`$.
For the $`\pi `$ mode in the presence of a vortex, Bruus et al. found that the fluctuation of the superspin is
$$\delta 𝐧(𝐫,t)=(0,0,0,\delta \theta f(r)\mathrm{cos}\varphi _𝐫e^{i\omega _\pi t},0),$$
(2)
where the small angle $`\delta \theta `$ by which $`𝐧`$ rotates into the antiferromagnetic sector is undetermined. Since the excitation depends on $`f`$ and not on $`m`$ it is a de-localized excitation with zero amplitude at the center of the vortices and in terms of energy it actually corresponds to an energy at the bottom edge of the continuum of an effective potential associated to the vortices.
For an isotropic spin space, the magnetic scattering cross-section for neutrons is proportional to the dynamic structure factor, which is the Fourier transform of the spin-spin correlation function (see e.g. Ref. ),
$$𝒮(𝐪,\omega )=_{\mathrm{}}^{\mathrm{}}dte^{i\omega t}\underset{\mathrm{𝐑𝐑}^{}}{}e^{i𝐪(𝐑𝐑^{})}\widehat{𝐒}_𝐑(t)\widehat{𝐒}_𝐑^{}(0).$$
(3)
To make a connection to the SO(5) calculations we make the semiclassical approximation $`<\widehat{𝐒}_𝐑(t)\widehat{𝐒}_𝐑^{}(0)><\widehat{𝐒}_𝐑(t)><\widehat{𝐒}_𝐑^{}(0)>`$ so that
$`𝒮(𝐪,\omega )`$ $``$ $`{\displaystyle _{\mathrm{}}^{\mathrm{}}}dte^{i\omega t}{\displaystyle \underset{𝐑,𝐑^{}}{}}e^{i\left(𝐪+𝐐\right)(𝐑𝐑^{})}`$ (4)
$`\times 𝐦(𝐑,t)𝐦(𝐑^{},0),`$ (5)
where $`𝐦(𝐑,t)=e^{i𝐐𝐑}𝐒_𝐑(t)`$ is the antiferromagnetic order parameter which enters the superspin $`𝐧`$.
With a superspin given by $`𝐧(𝐫,t)=𝐧(𝐫)+\delta 𝐧(𝐫,t)`$ the dynamical structure factor has two components — an elastic and an inelastic. The elastic component
$$𝒮_{\mathrm{el}}(𝐪,\omega )=\left|\underset{𝐑}{}e^{i(𝐪+𝐐)𝐑}m(R)\right|^22\pi \delta (\omega ),$$
(6)
is located at $`𝐪=𝐐`$ and has a width $`\pi /\xi `$. In elastic neutron scattering experiments the observation of this peak would directly prove the antiferromagnetical ordering in vortex cores.
The inelastic contribution is
$`𝒮_{\mathrm{in}}(𝐪,\omega )`$ $`=`$ $`\left(\delta \theta \right)^2\left|{\displaystyle \underset{𝐑}{}}e^{i(𝐪+𝐐)𝐑}f(R)\mathrm{cos}\varphi _𝐑\right|^2`$ (7)
$`\times 2\pi \delta (\omega \omega _\pi ).`$ (8)
For $`𝐪=𝐐`$ the phase factor $`e^{i(𝐪+𝐐)𝐑}`$ vanishes, and the cosine factor makes the different terms in the summation cancel pairwise so that $`𝒮_{\mathrm{in}}(𝐐,\omega _\pi )=0`$. The presence of a single vortex moves the intensity away from $`𝐪=𝐐`$ and a ring-shaped peak with radius $`\delta q\pi /L`$ centered at $`𝐪=𝐐`$ is formed, $`L\sqrt{A}`$ being the size of the sample. In the semiclassical approximation the zero amplitude at $`𝐪=𝐐`$ is a topological feature, which is independent of the detailed radial form $`f(r)`$ of the vortex. This robustness relies on the identification of the $`\pi `$ mode as being proportional to the superconducting order-parameter (including its phase). Quantum fluctuations may add some amplitude at $`𝐪=𝐐`$, but such an analysis beyond leading order is outside the scope of this work.
It is interesting to see how this result compares to predictions based on the BCS theory. The neutron scattering cross-section is given by the spin susceptibility, which for a homogeneous (vortex free) superconductor has been calculated via the BCS-Lindhard function. Here we briefly consider how the BCS coherence factor $`[u_kv_{k+q}v_ku_{k+q}]^2`$ appearing in the Lindhard function is modified by the presence of vortices. In a semiclassical approximation the spatial variation of the superconducting phase $`\varphi (𝐫)`$ leads to a coherence factor of the form $`[u_k(𝐫_1)e^{i\varphi (𝐫_1)/2}v_{k+q}(𝐫_2)e^{i\varphi (𝐫_2)/2}v_k(𝐫_1)e^{i\varphi (𝐫_1)/2}u_{k+q}(𝐫_2)e^{i\varphi (𝐫_2)/2}]^2`$. Therefore in contrast to Eq. (8) the superconducting phase does not separate in the two spatial positions, and consequently the spatial average in general is not zero at $`𝐪=𝐐`$. It thus appears that the above mentioned ring-shaped peak in the dynamic structure factor is special for the SO(5) model.
We now generalize the single-vortex SO(5)-result to the case of a vortex lattice. For non-overlapping vortices we construct the full superconducting order parameter by
$$\stackrel{~}{\psi }(𝐫)=\stackrel{~}{f}(𝐫)e^{i\stackrel{~}{\varphi }(𝐫)}=_j\psi (𝐫𝐫_j),$$
(9)
where the $`𝐫_j`$ denote the positions of the vortices. The function $`\stackrel{~}{f}(𝐫)=_jf(𝐫𝐫_j)`$ is $`1`$ except for close to the vortices where it dips to zero. Also the phase $`\stackrel{~}{\varphi }(𝐫)=_j\mathrm{arg}(𝐫𝐫_j)`$ has by construction the periodicity of the vortex lattice (modulo $`2\pi `$) and the contour integral $`_Cd𝐥\mathbf{}\stackrel{~}{\varphi }(𝐫)`$ equals $`2\pi n`$ where $`n`$ is the number of vortices enclosed by the contour $`C`$. In the limit of non-overlapping vortices we can capture the main physics by considering the single vortex solution within a unit cell of the vortex lattice. We comment on the inclusion of the entire vortex lattice further on, but for now we restrict the summation in Eq. (8) to lattice sites $`𝐑`$ inside the vortex lattice unit cell. In Fig. 1 we show the result for a magnetic field $`B=10\mathrm{T}`$. As seen, the presence of vortices moves the intensity away from $`𝐪=𝐐`$ and a ring shaped peak with radius $`\delta q`$ centered at $`𝐪=𝐐`$ is formed. We note that the only relevant length scale available is the vortex lattice constant $`d`$ and consequently we expect that $`\delta q=\pi /d`$. Since $`d=3^{1/4}\sqrt{h/eB}`$ we consequently expect that $`\delta q=3^{1/4}\pi \sqrt{eB/h}0.008\times (\pi /a)\sqrt{B/[\mathrm{T}]}`$. Had we included all the vortex lattice unit cells in our analysis, the structure factor of the hexagonal vortex lattice would have led to a breaking of the ring in Fig. 1 into six sub-peaks sitting on top of the ring. In a real experiment these sub-peaks could easily be smeared back into a ring-shaped scattering peak if either the vortex lattice were slightly imperfect or if the resolution of the spectrometer were too low. To describe the main effect of the SO(5) theory we therefore continue to use the single unit cell approximation.
In Fig. 2 we show the splitting as a function of the magnetic field and indeed we find the expected scaling with a pre-factor confirming that the splitting is given by $`\delta q=\pi /d`$. The full width half maximum of the ring is given by $`\mathrm{\Gamma }3.1\times \delta q=3.1\times \pi /d`$.
In Fig. 3 we show the amplitude of the ring as a function of magnetic field. The amplitude approximately decreases as $`1/B`$ with the magnetic field, but with a small deviation. This deviation makes the $`𝐪`$-integrated intensity, which is proportional to the amplitude times $`(\delta q)^2`$, decrease as $`I(B)/I(0)10.004\times B/[\mathrm{T}]`$ which reflects that the area occupied by vortices increases linearly with $`B`$ and consequently the superconducting region decreases linearly with $`B`$. In fact, the reduction is given by $`𝒜^12\pi rdrm^2(r)0.004\times B/[\mathrm{T}]`$, where the integral gives the effective area of the vortex. The reduction in integrated intensity should be relatively easy to observe experimentally, but is not a unique feature of the SO(5) model. Thus, while it will aid to prove that the $`\pi `$ resonance only resides in the superconducting phase, it will not clearly distinguish between different theories.
In order to discuss the experimental possibilities for testing our predictions, we note that the original observation of the zero-field $`\pi `$ resonance was an experimental achievement and hence that the experiment proposed here constitutes a great challenge. However, since the first observation of the $`\pi `$ resonance in 1991, the field of neutron scattering has developed considerably. To observe the ring-like shape (see inset of Fig. 1) of the excitation would require a resolution better than $`\pi /d`$ along two directions in reciprocal space, which seems unachievable with current spectrometers. However, the overall width of the ring can in fact be measured with good resolution along just one direction in the reciprocal plane. Scans along this direction (as in Fig. 1) could then reveal a broadening of $`3.1\times \pi /d`$. With a sufficiently optimized spectrometer we believe this to be possible, and the reward is a stringent test of a quantitative prediction of the SO(5) theory. We note that Bourges et al. have investigated the $`\pi `$ resonance in a magnetic field of $`B=11.5\mathrm{T}`$ and report a broadening in energy, but do not report data on the $`𝐪`$-shape.
In conclusion we have found that within the SO(5) model, the $`\pi `$ resonance splits into a ring centered at $`𝐪=(\pi /a,\pi /a)`$ in the presence of a magnetic field. The ring has the radius $`\pi /d`$ and full width half maximum of about $`3.1\times \pi /d`$, where $`d`$ is the vortex lattice constant. Consequently the splitting is found to scale with the magnetic field as $`B^{1/2}`$. We emphasize that the amplitude of the $`\pi `$ resonance is zero at $`𝐪=(\pi /a,\pi /a)`$ in the presence of a magnetic field.
We acknowledge useful discussions with J. Jensen, N. H. Andersen, A.-P. Jauho and D. F. McMorrow. H.M.R. is supported by the Danish Research Academy and H.B. by the Danish Natural Science Research Council through Ole Rømer Grant No. 9600548. |
no-problem/0001/cond-mat0001128.html | ar5iv | text | # Roughening and preroughening transitions in crystal surfaces with double-height steps
## Abstract
We investigate phase transitions in a solid-on-solid model where double-height steps as well as single-height steps are allowed. Without the double-height steps, repulsive interactions between up-up or down-down step pairs give rise to a disordered flat phase. When the double-height steps are allowed, two single-height steps can merge into a double-height step (step doubling). We find that the step doubling reduces repulsive interaction strength between single-height steps and that the disordered flat phase is suppressed. As a control parameter a step doubling energy is introduced, which is assigned to each step doubling vertex. From transfer matrix type finite-size-scaling studies of interface free energies, we obtain the phase diagram in the parameter space of the step energy, the interaction energy, and the step doubling energy.
Much attention has been paid to the phase transitions in crystal surfaces since they show rich critical phenomena. The interplay between roughening and reconstruction results in interesting phases, such as a disordered flat (DOF) phase, as well as flat and rough phases . In the DOF phase the surface is filled with macroscopic amount of steps which are disordered positionally but have up-down order. Several solid-on-solid (SOS) type models have been studied, which reveals that the DOF phase is stabilized by the repulsive step-step interactions or by specific topological properties of surfaces, e.g., Si(001) .
The SOS type model studies have been done in cases where the nearest-neighbor (NN) height difference, $`\mathrm{\Delta }h`$, is restricted to be equal to or less than 1 in units of the lattice constant. However, in real crystals there also appear steps with $`|\mathrm{\Delta }h|>1`$. For example, double-height steps on W(430) become more favorable than single-height steps at high temperatures since they have lower kink energy . In this paper we investigate the phase transitions in crystal surfaces in the presence of the double-height steps with $`|\mathrm{\Delta }h|=2`$, especially focusing on the stability of the DOF phase. We study a generalized version of the restricted solid-on-solid (RSOS) model on a square lattice with the Hamiltonian given in Eq. (2). We study the model under the periodic and anti-periodic boundary conditions, from which various interface free energies are defined. The interface free energy is calculated from numerical diagonalizations of the transfer matrix, and the phase diagram is obtained by analyzing their finite-size-scaling (FSS) properties.
In the RSOS model the surface is described by integer-valued heights $`h_𝐫`$ at each site $`𝐫=(n,m)`$ on a square lattice. (The lattice constant in the $`z`$ direction is set to 1.) Only the single-height step (S step) with $`|\mathrm{\Delta }h|=1`$ is allowed. It was found that the RSOS model with NN and next-nearest-neighbor (NNN) interactions between height displays the DOF phase when the NNN coupling strength is large enough . The NNN coupling accounts for the repulsive interactions between parallel (up-up or down-down) step pairs. Parallel step pairs cost more energy than anti-parallel (up-down or down-up) step pairs.
The double-height step (D step) is incorporated into the RSOS model by relaxing the restriction on the NN height difference to $`|\mathrm{\Delta }h|=0,1,2`$. We only consider quadratic NN and NNN interactions between heights since they are sufficient to describe the key feature of the phase transitions. The total Hamiltonian is written as
$$H_0=K\underset{𝐫,𝐫^{}}{}(h_𝐫h_𝐫^{})^2+L\underset{(𝐫,𝐫^{\prime \prime })}{}(h_𝐫h_{𝐫^{\prime \prime }})^2$$
(1)
where $``$ and $`()`$ denote the pair of NN and NNN sites. With this Hamiltonian, a D step costs more energy than two separate S steps by an amount of $`2K+4K`$ per unit length. Even though the D steps are energetically unfavorable, we will show that their effect is not negligible. We also consider a step-doubling energy $`E_D`$ to study the effect of the step doubling. It is assigned to each vertex where two S steps merge into a D step (see Fig. 1). The electronic state at step edges may be different from that at a flat surface, which contributes to the step energy. When two S steps merge into a D step, the electronic state near the vertex may be changed. The change leads to an additional energy cost, which is reflected by $`E_D`$. When $`E_D`$ is positive (negative), it suppresses (enhances) the step doubling. The Hamiltonian including $`H_0`$ and the step-doubling energy is then given by
$$H=H_0+E_DN_D$$
(2)
where $`N_D`$ is the total number of step-doubling vertices. (For a notational convenience the energy is measured in unit of $`k_BT`$.) The model with the Hamiltonian Eq. (2) with $`E_D=0`$ and with the restriction $`|\mathrm{\Delta }h|=0,1`$ will be referred to as the RSOS3 model, and the model with the Hamiltonian Eq. (2) and with $`|\mathrm{\Delta }h|=0,1,2`$ will be referred to as the RSOS5 model.
In a continuum description phase transitions in crystal surfaces are described by the sine-Gordon model
$$H=d^2𝐫\left[\frac{1}{2}K_G(\varphi )^2\underset{q=1}{\overset{\mathrm{}}{}}u_q\mathrm{cos}(2q\pi \varphi )\right],$$
(3)
where $`\varphi (𝐫)(\mathrm{},\mathrm{})`$ is a real-valued local average height field, $`K_G`$ the stiffness constant, and $`u_q`$ the fugacity of $`q`$-charge . In the renormalization group sense $`u_1`$ is irrelevant at high temperatures where the model renormalizes to the Gaussian model with a renormalized stiffness constant $`K_G<\frac{\pi }{2}`$ describing the rough phase. As temperature decreases, $`u_1`$ becomes relevant at a roughening transition temperature. There appear two kinds of low temperature phases depending on the sign of $`u_1`$: For positive $`u_1`$ the Hamiltonian favors an integer average height and hence the surface is flat. For a negative $`u_1`$ it favors a half-integer average height. Since the microscopic height is integer-valued, the surface can take the half-integer average height by forming steps with up-down order, i.e., the surface is in the DOF phase. As temperature decreases further, the sign of $`u_1`$ changes and the surface falls into the flat phase. At the roughening transition between the rough phase and the flat or DOF phase, the renormalized stiffness constant takes the universal value of $`\frac{\pi }{2}`$. The flat and DOF phases are separated by the preroughening transition characterized by $`u_1=0`$ .
The phase boundaries can be obtained using FSS properties of the interface free energies. Consider the model on a finite $`N\times M`$ square lattice rotated by $`45^{}`$ under the various boundary conditions (BC’s): The periodic BC, $`h(n+N,m)=h(n,m)+a`$ with integer $`a`$, and the anti-periodic BC, $`h(n+N,m)=h(n,m)+a(\text{mod }2)`$ with $`a=0\text{ and }1`$. They will be referred to as $`(\pm ,a)`$ BC’s (the upper (lower) sign for the (anti-)periodic BC’s). The free energy is obtained from the largest eigenvalue of the transfer matrix. Detailed description of the transfer matrix set-up can be found in Ref. . The boundary conditions except for the $`(+,0)`$ BC induce a frustration in the surface. The interface free energy $`\eta _\kappa `$ is defined as the excess free energy per unit length under the $`\kappa `$ BC with $`\kappa =(\pm ,a)`$ from that under the $`(+,0)`$ BC:
$$\eta _\kappa =\frac{1}{M}\mathrm{ln}\frac{Z_\kappa }{Z_{(+,0)}}$$
(4)
with $`Z_\kappa `$ the partition function satisfying the $`\kappa `$-BC.
The interface free energies have characteristic FSS properties in each phase. In the rough phase they show the universal $`1/N`$ scaling in the semi-infinite limit $`M\mathrm{}`$ as
$`\eta _{(+,a)}`$ $`=`$ $`{\displaystyle \frac{\zeta }{2}}{\displaystyle \frac{K_Ga^2}{N}}+o\left({\displaystyle \frac{1}{N}}\right)`$ (5)
$`\eta _{(,a)}`$ $`=`$ $`{\displaystyle \frac{\pi \zeta }{4N}}+o\left({\displaystyle \frac{1}{N}}\right),`$ (6)
where $`K_G\frac{\pi }{2}`$ is the renormalized stiffness constant of the Gaussian model and $`\zeta `$ is the aspect ratio of the lattice constants in the horizontal and vertical directions . In the flat phase $`\eta _{(+,a)}`$ and $`\eta _{(,1)}`$ are finite because at least one step is induced under the $`(+,a)`$ and $`(,1)`$ BC’s, while $`\eta _{(,0)}`$ is exponentially small in $`N`$ since the $`(,0)`$ BC may not induce any steps . In the DOF phase the $`(,1)`$ BC does not induce any frustration in the step up-down order, but the $`(+,a)`$ and $`(,0)`$ BC’s do. So $`\eta _{(,1)}`$ is exponentially small in $`N`$, and $`\eta _{(+,a)}`$ and $`\eta _{(,0)}`$ are finite . From these FSS properties the roughening points can be estimated from
$$\eta _{(+,1)}=\frac{\pi \zeta }{4N},$$
(7)
where the universal value of $`K_G=\frac{\pi }{2}`$ at the roughening transition is used in Eq. (5). The preroughening points between the flat and the DOF phase can be estimated from the crossing behaviors of $`N\eta _{(,0)}`$ or $`N\eta _{(,1)}`$, which converges to zero in one phase and diverges to infinity in the other phase as $`N`$ grows.
The estimation of transition points using the interface free energies suffers from slow convergence due to corrections to the scaling. They may smooth out the crossing behaviors of $`N\eta _{(,0)}`$ and $`N\eta _{(,1)}`$ at the preroughening transitions for small $`N`$. But one can safely cancel out leading corrections to scaling by taking the ratio or the difference of them, which can be seen as follows. Consider the lattice version of the continuum model in Eq. (3). It is obvious, using the transformation $`\varphi \varphi 1/2`$, that the model under the $`(,0)`$ BC is the same as that under the $`(,1)`$ BC with $`u_q`$ replaced by $`u_q`$ for odd $`q`$. It yields the relation
$$\eta _{(,0)}(u_1,u_2,u_3,\mathrm{})=\eta _{(,1)}(u_1,u_2,u_3,\mathrm{}).$$
(8)
So if one neglects all higher order contributions from $`u_{q3}`$, the location of $`u_1=0`$ is found from the condition $`\eta _{(,0)}\eta _{(,1)}=0`$ or $`R=1`$ with
$$R\frac{\eta _{(,0)}}{\eta _{(,1)}}.$$
(9)
It is not influenced by correction-to-scalings from $`u_2`$. Therefore the relation $`R=1`$ can be used to get the $`u_1=0`$ point more accurately. One can easily see that $`R>1`$ for negative $`u_1`$ and $`R<1`$ for positive $`u_1`$. It approaches 1 in the rough phase and at the preroughening transition points, diverges in the DOF phase, and vanishes in the flat phase as $`N\mathrm{}`$.
In the RSOS3 model the exact point with $`u_1=0`$ is known along the line $`L=0`$ . It is called the self-dual point and is located at $`K=K_{SD}=\mathrm{ln}[\frac{1}{2}(\sqrt{5}+1)]`$. From numerical studies of the RSOS3 model transfer matrix, we could obtain the exact value of $`K_{SD}`$ with error less than $`10^{12}`$ by solving $`R=1`$ even with small system size $`N=4`$, which indicates that $`R`$ is a useful quantity to determine the preroughening transition points accurately. It will be used in the analysis of the RSOS5 model.
We first consider the RSOS5 model in a special case of $`E_D=0`$ and compare its phase diagram with that of the RSOS3 model to have insight into the role of the D step. At low temperatures the D step is unfavorable due to larger free energy cost than the S step. So the nature of the low temperature phase in the RSOS5 model is not different from that in the RSOS3 model, i.e., the flat phase. At high temperatures, the surface is in the rough phase in the RSOS3 model. Since the rough phase is critical and there is no characteristic length scale, there will be no difference between S and D steps. So the RSOS5 model will also have the rough phase as a high temperature phase. There is significant difference in intermediate temperature range, where the repulsive step interactions stabilize the DOF phase in the RSOS3 model. Without the D steps the parallel steps have less meandering entropy than anti-parallel ones. It is energetically unfavorable for parallel steps to approach each other closer than the interaction range while anti-parallel steps can approach each other at will . However, if one allows the D step, two parallel S steps can approach each other and form a D step without the interaction energy cost. Provided that the energy cost of the D step is not too high, the presence of the D step reduces repulsive interaction strength effectively and enhances the meandering entropy of parallel steps. Then it will suppress the DOF phase.
To see such effects quantitatively, we calculate the ratio $`R`$ for the RSOS3 model and the RSOS5 model with $`E_D=0`$ along a line $`L=5K`$ (see Fig. 2). The strip width for the transfer matrix is $`N=4,6,8`$, and $`10`$ for the RSOS3 model and $`N=4,6`$, and $`8`$ for the RSOS5 model. The RSOS3 model displays the roughening and the preroughening transitions along the line $`L=5K`$, which is manifest in Fig. 2(a). There are three regions where $`N`$ dependence of $`R`$ is distinct with each other. The surface is in the rough phase with negative $`u_1`$ in the small $`L`$ (high temperature) region, where $`R`$ approaches $`1`$ from above. And the surface is in the DOF (flat) phase for the intermediate (large) $`L`$ region, where $`R`$ grows (vanishes). The roughening and preroughening transition points are estimated from Eq. (7) and $`R=1`$ with $`R`$ in Eq. (9), respectively, which are represented by broken vertical lines.
The situation changes qualitatively in the RSOS5 model. As can be seen in Fig. 2(b), $`R`$ is always less than 1, and there are only two regions with distinct $`N`$ dependence of $`R`$. In the small $`L`$ region $`R`$ approach $`1`$ from below, and in the large $`L`$ region $`R`$ vanishes as $`N`$ increases. They correspond to the rough phase with positive $`u_1`$ and the flat phase, respectively. The roughening transition point is estimated from Eq. (7) and represented by the broken vertical line. It shows that the DOF phase is suppressed in the presence of the D step. We have also checked that $`R`$ is always less than 1 ($`u_1>0`$) and the DOF phase does not appear at any values of $`K`$ and $`L`$ in the RSOS5 model with $`E_D=0`$.
We can argue the reason why the DOF phase disappear in the presence of the D step as follows. Consider two parallel S steps merging at a vertex. If the D step is not allowed, the possible vertex configuration is as shown in Fig. 3(a) and the energy cost for such configuration is $`2K+4L`$. On the other hand, if the D steps is allowed, the step doubling may occur in two ways as shown in Fig. 3(b) with the energy cost $`3K+5L`$. Though the step doubling costs more energy ($`K+L`$), entropic contribution of the step doubling ($`\mathrm{ln}2`$) may lower the free energy of parallel steps below than the value without the step doubling. Our numerical results above show that the step doubling suppresses the DOF phase entirely in the $`E_D=0`$ case. In our model a D step costs more energy than two separate S steps. The two energy scales may be comparable to each other in a more realistic model, where the suppression effect will be stronger.
From the above arguments, one finds that the step doubling plays an important role in phase transitions. So we introduce a new term $`E_DN_D`$ in Eq. (2) with the step-doubling energy $`E_D`$ and study the phase diagram in the parameter space $`(K,L,E_D)`$. When $`E_D<0.0`$ ($`>0.0`$), the step doubling is favored (suppressed). One can easily expect that the DOF phase does not appear for negative $`E_D`$.
For positive $`E_D`$ the step doubling is suppressed and the effect of the step interaction becomes important. So we expect there appears the DOF phase in the positive $`E_D`$ side of the parameter space. In Fig. 4 we show the ratio $`R`$ for $`e^{E_D}=0.2`$ and along the line $`L=5K`$. Though the convergence is not good, compared with Fig. 2(a), one can identify three regions as the rough, DOF, and flat phases from the $`N`$ dependence of $`R`$. The roughening point between the rough phase and the DOF phase is estimated using Eq. (7), and the preroughening point using $`R=1`$ for $`N=8`$. They are denoted by broken vertical lines.
We obtain the phase diagram in the whole parameter space using the conditions $`\eta _{(+,1)}=\frac{\pi \zeta }{4N}`$ for the roughening transition boundary and $`R=1`$ for the preroughening transition boundary. It is obtained for strip width $`N=4,6`$, and $`8`$. Since the maximum $`N`$ we can handle is small, the convergence of the phase boundary is poor especially as one approaches $`e^{E_D}=0`$. But there is no qualitative change in shape. So we only present the phase diagram obtained from $`N=8`$ in Fig. 5. The region under the surface represented by broken lines corresponds to the rough phase. The DOF phase is bounded by the surfaces of broken lines and solid lines. The region above the surfaces corresponds to the flat phase. One should notice that there is a critical value of $`E_D`$, approximately $`0.071`$, smaller than which the DOF phase does not appear.
In summary, we have studied the phase transitions in the RSOS5 model with the Hamiltonian in Eq. (2) with D steps as well as S steps. We have found that the D step, which has not been considered in previous works, plays an important role in phase transitions in crystal surfaces. The presence of the D step reduces the strength of the repulsive interaction between parallel steps through the step doubling, and hence suppresses the DOF phase. We also found that the step-doubling energy is an important quantity which characterizes a surface upon the roughening.
I would like to thank D. Kim and M. den Nijs for helpful discussions. I wish to acknowledge the financial support of Korea Research Foundation made in the program year 1997. This work is also supported by the KOSEF through the SRC program of SNU-CTP. |
no-problem/0001/cond-mat0001396.html | ar5iv | text | # Stability of Trions in Strongly Spin Polarized Two-Dimensional Electron Gases
## Abstract
Low-temperature magneto-photoluminescence studies of negatively charged excitons ($`X_s^{}`$ trions) are reported for n-type modulation-doped ZnSe/Zn(Cd,Mn)Se quantum wells over a wide range of Fermi energy and spin-splitting. The magnetic composition is chosen such that these magnetic two-dimensional electron gases (2DEGs) are highly spin-polarized even at low magnetic fields, throughout the entire range of electron densities studied ($`5\times 10^{10}`$ to $`6.5\times 10^{11}`$ cm<sup>-2</sup>). This spin polarization has a pronounced effect on the formation and energy of $`X_s^{}`$, with the striking result that the trion ionization energy (the energy separating $`X_s^{}`$ from the neutral exciton) follows the temperature- and magnetic field-tunable Fermi energy. The large Zeeman energy destabilizes $`X_s^{}`$ at the $`\nu =1`$ quantum limit, beyond which a new PL peak appears and persists to 60 Tesla, suggesting the formation of spin-triplet charged excitons.
Magnetic two-dimensional electron gases (2DEGs) represent a relatively new class of semiconductor quantum structure in which an electron gas is made to interact strongly with embedded magnetic moments. Typically, magnetic 2DEGs (and 2D hole gases) are realized in modulation-doped II-VI diluted magnetic semiconductor quantum wells in which paramagnetic spins (Mn<sup>2+</sup>, $`S=\frac{5}{2}`$) interact with the confined electrons via a strong $`J_{sd}`$ exchange interaction. This interaction leads to an enhanced spin splitting of the electron Landau levels which follows the Brillouin-like Mn<sup>2+</sup> magnetization, saturating in the range 10-20 meV by a few Tesla. Since the spin splitting can greatly exceed both the cyclotron ($``$1 meV/T) and Fermi energies, these magnetic 2DEGs consist largely of spin-polarized Landau levels, and serve as interesting templates for studies of quantum transport in the absence of spin gaps. In addition, it has been recognized that this interplay between the cyclotron, Zeeman and Fermi energies may also be exploited in magneto-optical experiments to gain insights into the rich spectrum of optical excitations found in 2DEGs. The aim of this paper is to use strongly spin-polarized magnetic 2DEGs, containing a wide range of electron densities, to shed new light on the spin-dependent properties of negatively charged excitons (or trions).
Predicted in 1958 by Lampert and first observed by Kheng in 1993, the singlet state of the negatively charged exciton (the $`X_s^{}`$ trion) consists of a spin-up and spin-down electron bound to a single hole. The energy required to remove one of these electrons (leaving behind a neutral exciton $`X^0`$) is the $`X_s^{}`$ ionization energy $`\mathrm{\Delta }E_X`$, usually defined as the energy between $`X_s^{}`$ and $`X^0`$ features in optical studies. $`\mathrm{\Delta }E_X`$ is small; typically only $``$1 meV, $``$3 meV, and $``$6 meV in GaAs-, CdTe-, and ZnSe-based 2DEGs respectively. The spin-singlet nature of the two electrons in $`X_s^{}`$ suggests that $`\mathrm{\Delta }E_X`$ – and hence trion stability – should be sensitive to the Zeeman energy and spin-polarization of the 2DEG. Here, we explicitly study highly spin-polarized magnetic 2DEGs to establish empirical correlations between Zeeman energy and trion stability over a broad range of carrier densities. In particular, magneto-photoluminescence (PL) measurements demonstrate the striking result that $`\mathrm{\Delta }E_X`$ follows the energy of the Fermi surface, which can be tuned independently from the Landau levels via the strong Zeeman dependence on temperature and applied field. The role of the Fermi and Zeeman energies in determining $`\mathrm{\Delta }E_X`$ is studied for all carrier densities, and qualitative agreement with numerical calculations is found. The giant spin-splitting in these systems is found to reduce $`\mathrm{\Delta }E_X`$, eventually driving a rapid suppression of $`X_s^{}`$ by the $`\nu =1`$ quantum limit, beyond which the formation of a new peak in the PL (which persists to 60T) may signify the formation of spin-triplet charged excitons.
These experiments are performed at the National High Magnetic Field Laboratory, in the generator-driven 60 Tesla Long-Pulse magnet and a 40T capacitor-driven magnet (with 2000 ms and 500 ms pulse duration, respectively), as well as a 20T superconducting magnet. Light is coupled to and from the samples via single optical fibers (200$`\mu m`$ or 600$`\mu m`$ diameter), and excitation power is kept below 200$`\mu W`$. Thin-film circular polarizers between the fiber and sample permit polarization-sensitive PL studies. In the pulsed magnet experiments, a high-speed CCD camera acquires complete optical spectra every 1.5 ms, enabling reconstruction of the entire spectra vs. field dependence in a single magnet shot. The magnetic 2DEG samples are MBE-grown n-type modulation-doped 105$`\AA `$ wide single quantum wells into which Mn<sup>2+</sup> are “digitally” introduced in the form of equally-spaced fractional monolayers of MnSe. Specifically, the quantum wells are paramagnetic digital alloys of (Zn<sub>1-x</sub>Cd<sub>x</sub>Se)<sub>m-f</sub>(MnSe)<sub>f</sub> with x= 0.1 to 0.2, m=5 and f=1/8 or 1/16 effective monolayer thickness. The electron densities, determined from Shubnikov-deHaas (SdH) oscillations in transport, range between $`5\times 10^{10}`$ and $`6.5\times 10^{11}`$ cm<sup>-2</sup>. All samples show a large spin splitting at 1.5 K, with “effective” g-factors in the range $`70<g_e^{eff}(H0)<100`$.
Figure 1a shows the evolution of the PL spectra in a magnetic 2DEG with relatively low carrier density $`1.24\times 10^{11}`$ cm<sup>-2</sup> and $`g_{eff}=73`$ at 1.5K. This sample has a mobility of 14000 cm<sup>2</sup>/Vs and exhibits clear SdH oscillations in transport. At $`H=0`$, the data show a strong PL peak at 2.74 eV with a small satellite $``$6 meV higher in energy. With applied field, the peaks shift rapidly to lower energy in the $`\sigma ^+`$ polarization due to the large Zeeman energy (the $`\sigma ^{}`$ emission disappears completely at low fields in all the magnetic 2DEGs, much like their undoped counterparts). By 1 T, the satellite develops into a clear peak of comparable amplitude, and as will be verified in Fig. 2, we assign the high- and low-energy PL features to $`X^0`$ and $`X_s^{}`$. At $`\nu =1`$ (5.5 T), the smooth evolution of the PL spectra changes abruptly as the $`X_s^{}`$ resonance collapses and a strong, single PL peak emerges at an energy between that of $`X^0`$ and $`X_s^{}`$, as shown. This new PL feature persists to 60 T. Fig. 1b shows the energies of the PL peaks (the data are fit to Gaussians), where the discontinuity at $`\nu =1`$ is clearly seen. The $`X_s^{}`$ ionization energy $`\mathrm{\Delta }E_X`$ decreases and oscillates with magnetic field (inset, Fig 1b). Anticipating Figs. 3 and 4, we note that $`\mathrm{\Delta }E_X`$ qualitatively mimics the Fermi energy in this low-density magnetic 2DEG (plotted in Fig. 1a, inset).
Owing to the giant spin splitting in this sample, the “ordinary” Landau level (LL) fan diagram for non-magnetic 2DEGs (with Landau levels evenly spaced by $`\mathrm{}\omega _c`$, and spin splitting $`\mathrm{}\omega _c`$) is replaced by that shown in the inset of Fig. 1a. The LLs are simply calculated as
$$\epsilon _{l,s}=\mathrm{}\omega _c(l+\frac{1}{2})+sE_ZB_{5/2}(5g_{Mn}\mu _BH/2k_BT^{})$$
(1)
where $`l`$ is the orbital angular momentum index and $`s`$ is the electron spin ($`\pm \frac{1}{2}`$). Here, $`\mathrm{}\omega _c`$ =0.83 meV/T is the electron cyclotron energy, and the second term is the Zeeman energy: $`B_{5/2}`$ is the Brillouin function describing the magnetization of the $`S=\frac{5}{2}`$ Mn<sup>2+</sup> moments, $`E_Z`$ is the saturation value of the electron splitting, $`g_{Mn}`$=2.0, and $`T^{}`$ is an empirical “effective temperature” which best fits the low-field energy shifts. We ignore the much smaller contribution to the Zeeman energy arising from the bare electron g-factor. At low fields, the spin-down LLs (solid lines) are Zeeman-shifted well below the spin-up LLs (dotted lines), leading to a highly spin-polarized electron gas - e.g., by 1T, over 95% of the electrons are oriented spin-down in this sample. The Fermi energy $`\epsilon _F`$ (thick line) is calculated numerically by inverting the integral
$$N_e=_{\mathrm{}}^{\mathrm{}}g[\epsilon ,B,T]f[\epsilon ,\epsilon _F,T]𝑑\epsilon .$$
(2)
Here, $`N_e`$ is the known electron density, $`f[\epsilon ,\epsilon _F,T]`$ is the Fermi-Dirac distribution and $`g[\epsilon ,B,T]`$ is the density of states, taken to be the sum of Lorentzian LLs of width $`\mathrm{\Gamma }=\mathrm{}/2\tau _s`$ centered at the energies $`\epsilon _{l,s}`$ given in Eq.1. The electron scattering time $`\tau _s`$ is obtained from analyzing SdH oscillations, or alternatively from the measured mobility.
Typically, identification of $`X^0`$ and $`X_s^{}`$ relies on their polarization properties in reflection or absorption\- measurements which directly probe the available density of states. However in these magnetic 2DEGs, the huge Zeeman splitting and the relatively broad spectral linewidths (resulting from the high Mn<sup>2+</sup> concentration) complicate these standard analyses. While reflectivity studies in these samples do confirm the presence of two bound states at zero field (as expected for $`X^0`$ and $`X_s^{}`$), we rely on spin-polarized PL excitation measurements to verify the peaks in finite field, shown in Fig. 2. At fixed field and temperature, we record the PL while tuning the energy and helicity of the excitation laser (a frequency-doubled cw Ti:Sapphire laser). Since the PL is entirely $`\sigma ^+`$ polarized, it must arise from the recombination of a spin-down ($`m_s=\frac{1}{2}`$) electron with a $`m_j=\frac{3}{2}`$ valence hole (see diagram, Fig. 2c). If that $`m_s=\frac{1}{2}`$ electron is part of an $`X_s^{}`$ trion, emission will occur at the $`X_s^{}`$ energy. Thus the probability of forming $`X_s^{}`$ is related to the number of spin-up ($`m_s=+\frac{1}{2}`$) electrons present in the system. By specifically injecting spin-up electrons at the $`\sigma ^{}`$ resonance, we do indeed observe an enhancement of the $`X_s^{}`$ intensity (Fig. 2a). In contrast, injecting spin-down electrons with $`\sigma ^+`$ light can (and does) only favor the $`X^0`$ intensity (Fig. 2b). The amplitude ratio, I($`X_s^{}`$)/I($`X^0`$), is plotted in Fig. 2c, where the effects of pumping spin-up and spin-down electrons are more easily seen. Of related interest, no difference in this ratio is observed when exciting above the ZnSe barriers (2.8 eV) - evidence that the injected spin is scrambled when the electrons spill into the well from the barrier regions.
With the aid of the diagram in Fig. 2c, the evolution of the PL spectra in Fig. 1 may be interpreted as follows: $`X_s^{}`$ and $`X^0`$ are competing channels for exciton formation, with $`X_s^{}`$ dominating at zero field. With small applied field, the large spin-splitting drives a rapid depopulation of the spin-up electron bands, reducing the probability of $`X_s^{}`$ formation and thus increasing $`X^0`$ formation, as observed. With increasing field and Zeeman energy, $`X_s^{}`$ continues to form until it is no longer energetically favorable to bind a spin-up electron – in this case, evidently, at $`\nu =1`$ when the Fermi energy falls to the lowest LL. The PL peak which forms at $`\nu =1`$ (and persists to 60T), with an energy between that of $`X_s^{}`$ and $`X^0`$, represents formation of a stable new ground state. A likely candidate is the spin-triplet state of the negatively charged exciton ($`X_t^{}`$), wherein both bound electrons are oriented spin-down. The $`X_t^{}`$ trion, predicted to become the ground state in nonmagnetic 2DEGs at sufficiently high magnetic field, may also form stably in highly spin-polarized magnetic 2DEGs due to Zeeman energy considerations, although no theoretical description of these effects exists at present.
We turn now to results from high-density samples. Fig. 3 shows PL spectra and energy shifts observed in a high-density magnetic 2DEG ($`n_e=4.3\times 10^{11}`$ cm<sup>-2</sup>, mobility=2700 cm<sup>2</sup>/Vs, and $`g_e^{eff}(H0)=95`$ at 1.5K). These data are characteristic of that obtained in samples with $`n_e`$ up to $`6.5\times 10^{11}`$ cm<sup>-2</sup>, the highest density studied. Again, we observe a dominant PL peak at $`H=0`$ which shifts rapidly down in energy with applied field. However, in contrast with the low-density 2DEGs, the high-energy satellite peak does not appear until 2 Tesla (at 1.5K). This satellite grows to a peak of comparable amplitude by 12 Tesla, and exhibits similar sensitivity to the energy and helicity of the pump laser as seen in Fig 2; therefore we again assign these features to $`X_s^{}`$ and $`X^0`$. At $`\nu =1`$ (17 Tesla), these resonances collapse and are again replaced by a strong emission at an intermediate energy which persists to 60T. The energy of the observed PL peaks at 1.5K, 4K, and 10K are plotted in Fig. 3b, along with $`\mathrm{\Delta }E_X`$ (inset). Several features are notable. First, the $`X^0`$ peak only becomes visible at a particular spin splitting – not field – in support of the assertion that $`X^0`$ forms readily only when the spin-up electron subbands depopulate to a particular degree. In addition, the collapse of the $`X^0`$ and $`X_s^{}`$ peaks occurs at $`\nu =1`$ independent of temperature, again indicating that the drop of the Fermi energy to the lowest LL destabilizes $`X_s^{}`$. Finally, $`\mathrm{\Delta }E_X`$ again follows the calculated Fermi energy in this sample, exhibiting oscillations in phase with the Fermi edge.
This latter behavior is unexpected but appears to be true in all of our samples. In contrast with studies in nonmagnetic 2DEGs, these data clearly demonstrate the relevance of both the Zeeman energy and the Fermi energy in determining the trion ionization energy $`\mathrm{\Delta }E_X`$. In Figure 4 we explicitly study this behavior and reveal the surprising result that $`\mathrm{\Delta }E_X`$ closely follows the energy of the Fermi surface regardless of electron density, temperature, and applied field. Fig. 4a shows the measured field dependence of $`\mathrm{\Delta }E_X`$ in six magnetic 2DEGs with electron densities from $`n_e5\times 10^{10}`$ to $`2.5\times 10^{11}`$ cm<sup>-2</sup>. The data are plotted from the field at which distinct $`X^0`$ and $`X_s^{}`$ PL peaks first appear, until the collapse of the PL spectra. $`\mathrm{\Delta }E_X`$ is seen to decrease rapidly with field at the lowest densities, but remain roughly constant and exhibit weak oscillations at high densities. Further, a rough extrapolation (dotted lines) reveals that $`\mathrm{\Delta }E_X`$ at zero field increases from $``$7meV to 10meV with carrier density. Aside from a $``$7meV difference in overall magnitude, these features are qualitatively reproduced by the numerical computation of the Fermi energy in these samples, plotted in the lower graph. It is natural to associate 7 meV with the “bare” ($`n_e0`$) $`X_s^{}`$ binding energy, in reasonable agreement with earlier studies in low-density, nonmagnetic ZnSe-based 2DEGs. Thus, at least at zero field, $`\mathrm{\Delta }E_X`$ reflects the “bare” $`X_s^{}`$ binding energy plus the Fermi energy, in agreement with a recent viewpoint wherein the ionization process requires removing one electron from $`X_s^{}`$ to the top of the Fermi sea.
In nonzero field, the Zeeman energy reduces the $`X_s^{}`$ ionization energy. The explicit temperature dependence of $`\mathrm{\Delta }E_X`$ in the low-density magnetic 2DEG is particularly telling (Fig. 4b): Here, the small Fermi energy should play a minimal role ($`\epsilon _F`$1.5meV $``$ 9meV total spin splitting), and the data should directly reveal the $`X_s^{}`$ ionization energy. At different temperatures, $`\mathrm{\Delta }E_X`$ decreases from its zero-field value of $``$7.5meV at a rate which depends on the Brillouin-like spin splitting. In this sample, the 2DEG is almost immediately completely spin-polarized - no gas of “spin-up” electrons remains – and thus the drop in $`\mathrm{\Delta }E_X`$ must reflect the influence of the Zeeman energy. Physically, the energy of the spin-up electron in $`X_s^{}`$ increases with spin splitting, becoming more weakly bound, reducing $`\mathrm{\Delta }E_X`$ by roughly half of the total Zeeman splitting until the $`X_s^{}`$ destabilizes. Within this scenario, however, the rolloff in the slope of the data towards zero field is puzzling, possibly indicating that the energy between the Fermi edge and the spin-up subbands (rather than the Zeeman energy itself) may be the relevant parameter, as the calculated Fermi energy shows precisely the same behavior. No theoretical framework for this behavior exists at present. Alternatively, Fig 4c shows typical data from the high electron density sample where the Fermi energy (7.7meV) is comparable to the total spin splitting (12.6meV). Here, the measured $`\mathrm{\Delta }E_X`$ clearly follows the oscillations of the calculated Fermi energy, with no clear indication of the role played by the Zeeman energy. We pose these questions for future theoretical models for $`X_s^{}`$ formation, which must necessarily include the Zeeman energy and the influence of a finite Fermi energy.
In conclusion, we have presented a systematic study of charged exciton formation in strongly magnetic 2DEGs, wherein the giant spin splitting dominates the cyclotron energy and the electron gas is highly spin-polarized. The trion ionization energy $`\mathrm{\Delta }E_X`$ tracks the energy of the Fermi edge regardless of electron density, temperature or applied field, highlighting the important roles played by both the Fermi- and Zeeman energies. With increasing electron density, the data suggest that $`\mathrm{\Delta }E_X`$ – at least at zero magnetic field – reflects the “bare” $`X_s^{}`$ ionization energy of $``$7 meV plus the Fermi energy. Studies in low density samples show that the “bare” $`X_s^{}`$ binding energy is reduced by an amount proportional to the Zeeman energy, and in high density samples $`\mathrm{\Delta }E_X`$ follows the oscillations of the Fermi surface as it moves between Landau levels. Quantitative interpretation of these data must await a more complete theory of $`X_s^{}`$ formation in electron gases. This work is supported by the NHMFL and NSF-DMR 9701072 and 9701484. |
no-problem/0001/nucl-th0001014.html | ar5iv | text | # Vector Mesons in Medium and Dileptons in Heavy-Ion Collisions
## 1 Introduction
The investigation of hadron properties inside atomic nuclei constitutes one of the traditional research objectives in nuclear physics. However, in terms of the underlying theory of strong interactions QCD) even the description of the nuclear ground state remains elusive so far. Valuable insights can be expected from a careful study of transition regimes between hadronic and quark-gluon degrees of freedom. E.g., in electron-nucleus scattering experiments the corresponding control variable is the momentum transfer, whereas heavy-ion reactions, performed over a wide range of collision energies, aim at compressing and/or heating normal nuclear matter to witness potential phase transitions into a Quark-Gluon Plasma (QGP).
Among the key properties of the low-energy sector of strong interactions is the (approximate) chiral symmetry of the QCD Lagrangian and its spontaneous breaking in the vacuum. This is evident from such important phenomena as the build-up of a chiral condensate and constituent quark mass ($`M_q0.4`$ GeV), or the large mass splitting of $`\mathrm{\Delta }M0.5`$ GeV between ’chiral partners’ in the hadron spectrum (such as $`\pi (140)`$-$`\sigma (4001200)`$, $`\rho (770)`$-$`a_1(1260)`$ or $`N(940)`$-$`N^{}(1535)`$). It also indicates that medium modifications of hadron properties can be viewed as precursors of chiral symmetry restoration.
In this talk the focus will be on the vector (V) and axialvector (A) channels. The former is special in that it directly couples to the electromagnetic current (i.e., real and virtual photons) at which point it becomes ’immune’ to (strong) final state interactions thus providing direct experimental access to in-medium properties of vector mesons, e.g., through photoabsorption/-production on nuclei, or dilepton ($`e^+e^{}`$, $`\mu ^+\mu ^{}`$) spectra in heavy-ion reactions. The key issue is then to relate the medium effects to mechanisms of chiral restoration. This necessitates the simultaneous consideration of the axialvector channel, which, however, largely has to rely on theoretical analyses.
This talk is structured as follows: Sect. 2 is devoted to vector-meson properties in nuclear matter, Sect. 3 contains applications to heavy-ion reactions and Sect. 4 finishes with conclusions. A more complete discussion of the presented topics can be found in a recent review .
## 2 (Axial-) Vector Mesons in Cold Nuclear Matter
### 2.1 Correlators and Duality Threshold
The general quantity that is common to most theoretical approaches is the current-current correlation function which in the (axial-) vector channel is defined by
$$\mathrm{\Pi }_{V,A}^{\mu \nu }(q)=id^4xe^{iqx}0|𝒯j_{V,A}^\mu (x)j_{V,A}^\nu (0)|0.$$
(1)
For simplicity we will concentrate on the (prevailing) isospin $`I=1`$ (isovector) projections
$$j_{I=1}^\mu =\frac{1}{2}(\overline{u}\mathrm{\Gamma }^\mu u\overline{d}\mathrm{\Gamma }^\mu d)\mathrm{with}\mathrm{\Gamma }_V^\mu =\gamma ^\mu ,\mathrm{\Gamma }_A^\mu =\gamma _5\gamma ^\mu .$$
(2)
At sufficiently high invariant mass both correlators can be described by their (identical) perturbative forms which read (up to $`\alpha _S`$ corrections)
$$\mathrm{Im}\mathrm{\Pi }_{V,I=1}^{\mu \nu }=\mathrm{Im}\mathrm{\Pi }_{A,I=1}^{\mu \nu }=(g^{\mu \nu }\frac{q^\mu q^\nu }{M^2})\frac{M^2}{12}\frac{N_c}{2},MM_{dual}$$
(3)
($`M^2=q_0^2\stackrel{}{q}^2`$). At low invariant masses the vector correlator is accurately saturated by the (hadronic) $`\rho `$ spectral function within the Vector Dominance Model (VDM), i.e.,
$`\mathrm{Im}\mathrm{\Pi }_{V,I=1}^{\mu \nu }`$ $`=`$ $`{\displaystyle \frac{(m_\rho ^{(0)})^4}{g_\rho ^2}}\mathrm{Im}D_\rho ^{\mu \nu },MM_{dual}`$ (4)
$`\mathrm{Im}\mathrm{\Pi }_{A,I=1}^{\mu \nu }`$ $`=`$ $`{\displaystyle \frac{(m_{a_1}^{(0)})^4}{g_{a_1}^2}}\mathrm{Im}D_{a_1}^{\mu \nu }f_\pi ^2\pi \delta (M^2m_\pi ^2)q^\mu q^\nu ,MM_{dual}`$ (5)
with a similar relation involving the $`a_1`$ meson in the axialvector channel. The spontaneous breaking of chiral symmetry (SBCS) manifests itself in both the difference of the $`a_1`$ and $`\rho `$ spectral functions as well as the additional pionic piece in $`\mathrm{\Pi }_A`$ (notice that $`f_\pi `$ is another order parameter of SBCS). In vacuum the transition from the hadronic to the partonic regime (’duality threshold’) is characterized by the onset of perturbative QCD around $`M_{dual}1.5`$ GeV. In the medium, chiral restoration requires the degeneration of $`V`$\- and $`A`$-correlators over the entire mass range.
### 2.2 Model-Independent Results: V-A Mixing and Sum Rules
In a dilute gas the prevailing medium effect can be computed via low-density expansions. Using soft pion theorems and current algebra Krippa extended an earlier finite-temperature analysis to the finite-density case to obtain
$`\mathrm{\Pi }_V^{\mu \nu }(q)`$ $`=`$ $`(1\xi )\mathrm{\Pi }_V^{\mu \nu }(q)+\xi \mathrm{\Pi }_A^{\mu \nu }(q)`$
$`\mathrm{\Pi }_A^{\mu \nu }(q)`$ $`=`$ $`(1\xi )\mathrm{\Pi }_A^{\mu \nu }(q)+\xi \mathrm{\Pi }_V^{\mu \nu }(q)^{}`$ (6)
i.e., the leading density effect is a mere ’mixing’ of the vacuum correlators $`\mathrm{\Pi }^{\mu \nu }`$. The ’mixing’ parameter
$$\xi \frac{4\varrho _N\overline{\sigma }_{\pi N}}{3f_\pi ^2m_\pi ^2}$$
(7)
($`\varrho _N`$: nucleon density) is determined by the ’long-range’ part of the $`\pi N`$ sigma term,
$$\overline{\sigma }_{\pi N}=4\pi ^3m_\pi ^2N|\pi ^2|N20\mathrm{MeV}.$$
(8)
Chanfray et al. pointed out that $`\overline{\sigma }_{\pi N}`$ is in fact governed by the well-known nucleon- and delta-hole excitations in the pion cloud of the $`\rho `$ (or $`a_1`$) meson which have been thoroughly studied within hadronic models to be discussed in the following section. A naive extrapolation of eq. (7) to the chiral restoration point where $`\xi =1/2`$, gives $`\varrho _c2.5\varrho _0`$, which is not unreasonable. Nonetheless, as we will see below, realistic models exhibit substantial medium modifications beyond the mixing effect.
Similar in spirit, i.e., combining low-density expansions with chiral constraints, is the so-called master formula approach applied in ref. : chiral Ward identities including the effects of explicit breaking are used to express medium corrections to the correlators through empirically inferred $`\pi N`$, $`\rho N`$ (or $`\gamma N`$, etc.) scattering amplitudes times density. Resummations to all orders in density cannot be performed either in this framework.
Model independent relations which are in principle valid to all orders in density are provided by sum rules. Although typically of little predictive power, their evaluation in model calculations can give valuable insights. One example are the well-known QCD sum rules which have been used to analyze vector-meson spectral functions in refs. . It has been found, e.g., that the generic decrease of the quark- and gluon-condensates on the right-hand-side (r.h.s.) is compatible with the phenomenological (left-hand) side if either (i) the vector meson masses decrease (together with small resonance widths), or, (b) both width and mass increase (as found in most microscopic models).
Another example of sum rules are the ones derived by Weinberg , being generalized to the in-medium case in ref. . The first Weinberg sum rule, e.g., connects the pion decay constant to the integrated difference between the $`V`$\- and $`A`$-correlators:
$$f_\pi ^2=\underset{0}{\overset{\mathrm{}}{}}\frac{dq_0^2}{\pi (q_0^2q^2)}\left[\mathrm{Im}\mathrm{\Pi }_V(q_0,q)\mathrm{Im}\mathrm{\Pi }_A(q_0,q)\right]$$
(9)
for arbitrary three-momentum $`q`$ (here, the pionic piece has been explicitly separated out from $`\mathrm{\Pi }_A`$). We will come back to this relation below.
### 2.3 Hadronic Models and Experimental Constraints
Among the most spectacular predictions for the behavior of vector mesons in medium is the Brown-Rho Scaling hypothesis . By imposing QCD scale invariance on a chiral effective Lagrangian at finite density and applying a mean-field approximation it was conjectured that all light hadron masses (with the exception of the symmetry-protected Goldstone bosons) drop with increasing density following an approximately universal scaling law. The scaling also encompasses the pion decay constant (as well as an appropriate power of the quark condensate) and therefore establishes a direct link to chiral symmetry restoration being realized through the vanishing of all light hadron masses.
More conservative approaches reside on many-body techniques to calculate selfenergy contributions to the vector-meson propagators $`D_V`$ arising from interactions with surrounding matter particles (nucleons). They are computed from gauge-invariant (vector-current conserving) as well as chirally symmetric Lagrangians. The $`\rho `$ propagator, e.g., takes the form
$$D_\rho ^{L,T}(q_0,q;\varrho _N)=\left[M^2(m_\rho ^{(0)})^2\mathrm{\Sigma }_{\rho \pi \pi }^{L,T}(q_0,q;\varrho _N)\mathrm{\Sigma }_{\rho BN}^{L,T}(q_0,q;\varrho _N)\right]^1$$
(10)
for both transverse and longitudinal polarization states (which in matter, where Lorentz-invariance is lost, differ for $`q>0`$). $`\mathrm{\Sigma }_{\rho \pi \pi }`$ encodes the medium modifications in the pion cloud (through $`NN^1`$ and $`\mathrm{\Delta }N^1`$ bubbles, so-called ’pisobars’) , and $`\mathrm{\Sigma }_{\rho BN}`$ stems from direct ’rhosobar’ excitations of either $`S`$-wave ($`N(1520)N^1`$, $`\mathrm{\Delta }(1700)N^1`$, $`\mathrm{}`$) or $`P`$-wave type ($`\mathrm{\Delta }N^1`$, $`N(1720)N^1`$, $`\mathrm{}`$. The parameters of the interaction vertices (coupling constants and form factor cutoffs) can be estimated from free decay branching ratios of the involved resonances or more comprehensive scattering data (e.g., $`\pi N\rho N`$ , or $`\gamma N`$ absorption) which determine the low-density properties of the spectral functions. Additional finite-density constraints can be obtained from the analysis of photoabsorption data on nuclei. Invoking the VDM, the total photoabsorption cross section can be readily related to the imaginary part of the in-medium vector-meson selfenergy in the zero mass limit (needed for the coupling to real photons). An example of such a calculation is displayed in Fig. 2 where a reasonable fit to existing data on various nuclei has been achieved.
The low-density limit (represented by the long-dashed line in Fig. 2) cannot reproduce the disappearance of especially the $`N`$(1520) as seen in the data. A selfconsistent calculation to all orders in density , however, generates sufficiently large in-medium widths, on the order of $`\mathrm{\Gamma }_{N(1520)}^{med}`$ 200-300 MeV (resulting in the full line).
Fig. 2 shows the final result for the $`\rho `$ spectral function which has been subjected to the aforementioned constraints. The apparent strong broadening is consistent with other calculations . Similar features, albeit less pronounced, emerge within analogous treatments for $`\omega `$ and $`\varphi `$ mesons .
Let us now return to the question what these findings might imply for chiral restoration. In a recent work by Kim et al. an effective chiral Lagrangian including $`a_1`$-meson degrees of freedom has been constructed. Medium modifications of the latter are introduced by an ’$`a_1`$-sobar’ through $`N(1900)N^1`$ excitations to represent the chiral partner of the $`N(1520)N^1`$ state. Pertinent (schematic) two-level models have been employed for both the $`\rho `$ and $`a_1`$ spectral densities which, in turn, have been inserted into the Weinberg sum rule, eq. (9) (supplemented by perturbative high energy continua).
The resulting density-dependence of the pion decay constant, displayed in Fig. 3, exhibits an appreciable decrease of $``$ 30% at $`\varrho _N=\varrho _0`$, which bears some sensitivity on the assumed branching ratio of the $`N(1900)Na_1`$ decay (or $`N(1900)Na_1`$ coupling constant). However, the mechanism is likely to be robust: due to the low-lying $`\rho `$-$`N(1520)N^1`$ and $`a_1`$-$`N(1900)N^1`$ excitations, accompanied by a broadening of the elementary resonance peaks, the $`\rho `$ and $`a_1`$ spectral densities increasingly overlap, thus reducing $`f_\pi `$.
## 3 Electromagnetic Observables in Heavy-Ion Reactions
In central collisions of heavy nuclei at (ultra-) relativistic energies (ranging from $`p_{lab}`$=1-200 AGeV in current experiments to $`\sqrt{s}`$=0.2-10 ATeV at RHIC and LHC) hot and dense hadronic matter is created over extended time periods of about 20 fm/c. Local thermal equilibrium is probably reached within the first fm/c, after which the ’fireball’ expands and cools until the strong interactions cease (’thermal freezeout’) and the particles stream freely to the detector. Electromagnetic radiation (real and virtual photons) is continuously emitted as it decouples from the strongly interacting matter at the point of creation.
The thermal production rate of $`e^+e^{}`$ pairs per unit 4-volume can be expressed through the electromagnetic current correlation function (summed over all isospin states $`I`$=0,1),
$$\frac{dN_{ee}^{th}}{d^4xd^4q}=\frac{\alpha ^2}{\pi ^3M^2}f^B(q_0;T)\frac{1}{3}(\mathrm{Im}\mathrm{\Pi }_{em}^L+2\mathrm{I}\mathrm{m}\mathrm{\Pi }_{em}^T)$$
(11)
($`f^B`$: Bose distribution function; a similar expression holds for photons with $`M0`$). Fig. 4 shows that the medium effects in the $`\rho `$ propagator (including interactions with nucleons as well as thermal pions, kaons, etc.) induce a substantial reshaping of the emission rate (full lines) as compared to free $`\pi \pi `$ annihilation (dashed line) already at rather moderate temperatures and densities (left panel). In fact, under conditions close to the expected phase boundary (right panel) the $`\rho `$ resonance is completely ’melted’ and the hadronic dilepton production rate is very reminiscent to the one from a perturbative Quark-Gluon Plasma (dashed-dotted lines in Fig. 4) down to rather low invariant masses of $``$ 0.5 GeV ($`\alpha _S`$ corrections to the partonic rate might improve the agreement at still lower masses). It has been suggested to interpret this as a lowering of the in-medium quark-hadron duality threshold as a consequence of the approach towards chiral restoration.
The total thermal yield in a heavy-ion reaction is obtained by a space-time integration of eq. (11) over the density-temperature profile for a given collision system, modeled, e.g., within transport or hydrodynamic simulations. At CERN-SpS energies (160–200 AGeV) this ’thermal’ component is dominant over (or at least competitive with) final state hadron decays (at low $`M`$) and hard initial processes such as Drell-Yan annihilation (at high $`M`$) in the invariant mass range $`M`$ 0.2-2 GeV. A consistent description of the measured data is possible once hadronic many-body effects are included , cf. Figs.6 and 6. However, at this point also the dropping mass scenario is compatible with the data (cf. dashed curve in Fig. 6).
Optimistically one may conclude that strongly interacting matter close to the hadron-QGP phase boundary has been observed at the CERN-SpS. Other observables such as hadro-chemistry or $`J/\mathrm{\Psi }`$ suppression also support this scenario. Nonetheless, further data are essential to substantiate the present status and resolve the open questions.
## 4 Conclusions
This talk has focused on medium modifications of vector mesons in connection with chiral symmetry restoration in hot/dense matter. In accordance with a variety of empirical information hadronic spectral functions are characterized by the appearance of low-lying excitations as well as a broadening of the resonance structures. A schematic treatment of the $`a_1`$ meson on similar footings shows that these features encode an approach towards chiral restoration in nuclear matter as signaled by the decrease of the pion decay constant when evaluating the first Weinberg sum rule.
The application of these model calculations to electromagnetic observables as measured in recent heavy-ion experiments at the CERN-SpS leads to a reasonable description of the data from 0 to 2 GeV in invariant mass. The structureless in-medium hadronic dilepton production rates resemble perturbative $`q\overline{q}`$ annihilation in the vicinity of the expected phase boundary indicating that chiral restoration might be realized through a reduction of the quark-hadron duality threshold which in vacuum is located around 1.5 GeV. It would also corroborate the interrelation between temperature/density and momentum transfer in the transition from hadronic to partonic degrees of freedom.
In the near future further dilepton data will be taken by the PHENIX experiment at RHIC (advancing to a new energy frontier) as well as the precision experiment HADES at GSI. Thus electromagnetic observables can be expected to continue the progress in our understanding of strong interaction physics.
Acknowledgments
It is a pleasure to thank G.E. Brown, E.V. Shuryak and H. Sorge for collaboration and many fruitful discussions. |
no-problem/0001/physics0001047.html | ar5iv | text | # Path Integral Monte Carlo Calculation of the Deuterium Hugoniot
## I Introduction
Recent laser shock wave experiments on pre-compressed liquid deuterium have produced an unexpected equation of state for pressures up to 3.4 Mbar. It was found that deuterium has a significantly higher compressibility than predicted by the semi-empirical equation of state based on plasma many-body theory and lower pressure shock data (see SESAME model ). These experiments have triggered theoretical efforts to understand the state of compressed hydrogen in this range of density and temperature, made difficult because the experiments are in regime where strong correlations and a significant degree of electron degeneracy are present. At this high density, it is problematic even to define the basic units such as molecules, atoms, free deuterons and electrons. Conductivity measurements as well as theoretical estimates suggest that in the experiment, a state of significant but not complete metalization was reached.
A variety of simulation techniques and analytical models have been advanced to describe hydrogen in this particular regime. There are ab initio methods such as restricted path integral Monte Carlo simulations (PIMC) and density functional theory molecular dynamics (DFT-MD) . Further there are models that minimize an approximate free energy function constructed from known theoretical limits with respect to the chemical composition, which work very well in certain regimes. The most widely used include .
We present new results from PIMC simulations. What emerges is a relative consensus of theoretical calculations. First, we performed a finite size and time step study using a parallelized PIMC code that allowed simulation of systems with $`N_P=64`$ pairs of electrons and deuterons and more importantly to decrease the time step from $`\tau ^1=10^6\mathrm{K}`$ to $`\tau ^1=810^6\mathrm{K}`$. More importantly, we studied the effect of the nodal restriction on the hugoniot.
## II Restricted path integrals
The density matrix of a quantum system at temperature $`k_BT=1/\beta `$ can be written as a integral over all paths $`𝐑_t`$,
$$\rho (𝐑_0,𝐑_\beta ;\beta )=\frac{1}{N!}\underset{𝒫}{}(\pm 1)^𝒫\underset{𝐑_0𝒫𝐑_\beta }{}𝑑𝐑_te^{S[𝐑_t]}.$$
(1)
$`𝐑_t`$ stands for the entire paths of $`N`$ particles in $`3`$ dimensional space $`𝐑_t=(𝐫_{1t},\mathrm{},𝐫_{Nt})`$ beginning at $`𝐑_0`$ and connecting to $`𝒫𝐑_\beta `$. $`𝒫`$ labels the permutation of the particles. The upper sign corresponds to a system of bosons and the lower one to fermions. For non-relativistic particles interacting with a potential $`V(𝐑)`$, the action of the path $`S[𝐑_t]`$ is given by,
$$S[𝐑_t]=_0^\beta 𝑑t\left[\frac{m}{2}\left|\frac{d𝐑(t)}{\mathrm{}dt}\right|^2+V(𝐑(t))\right]+\text{const}.$$
(2)
One can estimate quantum mechanical expectation values using Monte Carlo simulations with a finite number of imaginary time slices $`M`$ corresponding to a time step $`\tau =\beta /M`$.
For fermionic systems the integration is complicated due to the cancellation of positive and negative contributions to the integral, (the fermion sign problem). It can be shown that the efficiency of the straightforward implementation scales like $`e^{2\beta Nf}`$, where $`f`$ is the free energy difference per particle of a corresponding fermionic and bosonic system . In , it has been shown that one can evaluate the path integral by restricting the path to only specific positive contributions. One introduces a reference point $`𝐑^{}`$ on the path that specifies the nodes of the density matrix, $`\rho (𝐑,𝐑^{},t)=0`$. A node-avoiding path for $`0<t\beta `$ neither touches nor crosses a node: $`\rho (𝐑(t),𝐑^{},t)0`$. By restricting the integral to node-avoiding paths,
$`\rho _F(𝐑_\beta ,𝐑^{};\beta )`$ $`=`$ (3)
$`{\displaystyle 𝑑𝐑_0}`$ $`\rho _F`$ $`(𝐑_0,𝐑^{};0){\displaystyle \underset{𝐑_0𝐑_\beta \mathrm{{\rm Y}}(𝐑^{})}{}}𝑑𝐑_te^{S[𝐑_t]},`$ (4)
($`\mathrm{{\rm Y}}(𝐑^{})`$ denotes the restriction) the contributions are positive and therefore PIMC represents, in principle, a solution to the sign problem. The method is exact if the exact fermionic density matrix is used for the restriction. However, the exact density matrix is only known in a few cases. In practice, applications have approximated the fermionic density matrix, by a determinant of single particle density matrices,
$$\rho (𝐑,𝐑^{};\beta )=\left|\begin{array}{ccc}\rho _1(𝐫_1,𝐫_1^{};\beta )& \mathrm{}& \rho _1(𝐫_N,𝐫_1^{};\beta )\\ \mathrm{}& \mathrm{}& \mathrm{}\\ \rho _1(𝐫_1,𝐫_N^{};\beta )& \mathrm{}& \rho _1(𝐫_N,𝐫_N^{};\beta )\end{array}\right|.$$
(5)
This approach has been extensively applied using the free particle nodes,
$$\rho _1(𝐫,𝐫^{},\beta )=(4\pi \lambda \beta )^{3/2}\text{exp}\left\{(𝐫𝐫^{})^2/4\lambda \beta \right\}$$
(6)
with $`\lambda =\mathrm{}^2/2m`$, including applications to dense hydrogen . It can be shown that for temperatures larger than the Fermi energy the interacting nodal surface approaches the free particle (FP) nodal surface. In addition, in the limit of low density, exchange effects are negligible, the nodal constraint has a small effect on the path and therefore its precise shape is not important. The FP nodes also become exact in the limit of high density when kinetic effects dominate over the interaction potential. However, for the densities and temperatures under consideration, interactions could have a significant effect on the fermionic density matrix.
To gain some quantitative estimate of the possible effect of the nodal restriction on the thermodynamic properties, it is necessary to try an alternative. In addition to FP nodes, we used a restriction taken from a variational density matrix (VDM) that already includes interactions and atomic and molecular bound states.
The VDM is a variational solution of the Bloch equation. Assume a trial density matrix with parameters $`q_i`$ that depend on imaginary time $`\beta `$ and $`𝐑^{}`$,
$$\rho (𝐑,𝐑^{};\beta )=\rho (𝐑,q_1,\mathrm{},q_m).$$
(7)
By minimizing the integral:
$$𝑑𝐑\left(\frac{\rho (𝐑,𝐑^{};\beta )}{\beta }+\rho (𝐑,𝐑^{};\beta )\right)^2=0,$$
(8)
one determines equations for the dynamics of the parameters in imaginary time:
$$\frac{1}{2}\frac{H}{\stackrel{}{q}}+\stackrel{}{𝒩}\dot{\stackrel{}{q}}=0\text{where}H\rho \rho 𝑑𝐑.$$
(9)
The normalization matrix is:
$`𝒩_{ij}`$ $`=`$ $`\underset{q^{}q}{lim}{\displaystyle \frac{^{\mathrm{\hspace{0.17em}2}}}{q_iq_j^{}}}\left[{\displaystyle 𝑑𝐑\rho (𝐑,\stackrel{}{q};\beta )\rho (𝐑,\stackrel{}{q}^{};\beta )}\right].`$ (10)
We assume the density matrix is a Slater determinant of single particle Gaussian functions
$$\rho _1(𝐫,𝐫^{},\beta )=(\pi w)^{3/2}\text{exp}\left\{(𝐫𝐦)^2/w+d\right\}$$
(11)
where the variational parameters are the mean $`𝐦`$, squared width $`w`$ and amplitude $`d`$. The differential equation for this ansatz are given in . The initial conditions at $`\beta 0`$ are $`w=2\beta `$, $`𝐦=𝐫^{}`$ and $`d=0`$ in order to regain the correct FP limit. It follows from Eq. 8 that at low temperature, the VDM goes to the lowest energy wave function within the variational basis. For an isolated atom or molecule this will be a bound state, in contrast to the delocalized state of the FP density matrix. A further discussion of the VDM properties is given in . Note that this discussion concerns only the nodal restriction. In performing the PIMC simulation, the complete potential between the interacting charges is taken into account as discussed in detail in .
Simulations with VDM nodes lead to lower internal energies than those with FP nodes as shown in Fig. 1. Since the free energy $`F`$ is the integral of the internal energy over temperature, one can conclude that VDM nodes yield to a smaller $`F`$ and hence, are the more appropriate nodal surface.
For the two densities considered here, the state of deuterium goes from a plasma of strongly interacting but un-bound deuterons and electrons at high $`T`$ to a regime at low $`T`$, which is characterized by a significant electronic degeneracy and bound states. Also at decreasing $`T`$, one finds an increasing number of electrons involved in long permutation cycles. Additionally for $`T\mathrm{15\hspace{0.17em}625}\mathrm{K}`$, molecular formation is observed. Comparing FP and VDM nodes, one finds that VDM predicts a higher molecular fraction and fewer permutations hinting to more localized electrons.
## III Shock Hugoniot
The recent experiments measured the shock velocity, propagating through a sample of pre-compressed liquid deuterium characterized by an initial state, ($`E_0`$$`V_0`$$`p_0`$) with $`T=19.6\mathrm{K}`$ and $`\rho _0=0.171\mathrm{g}/\mathrm{cm}^3`$. Assuming an ideal planar shock front, the variables of the shocked material ($`E`$$`V`$$`p`$) satisfy the hugoniot relation ,
$$H=EE_0+\frac{1}{2}(VV_0)(p+p_0)=0.$$
(12)
We set $`E_0`$ to its exact value of $`15.886\mathrm{eV}`$ per atom and $`p_0=0`$. Using the simulation results for $`p`$ and $`E`$, we calculate $`H(T,\rho )`$ and then interpolate $`H`$ linearly at constant $`T`$ between the two densities corresponding to $`r_s=1.86`$ and $`2`$ to obtain a point on the hugoniot in the $`(p,\rho )`$ plane. (Results at $`r_s=1.93`$ confirm the function is linear within the statistical errors). The PIMC data for $`p`$, $`E`$, and the hugoniot are given in Tab. I.
In Fig. 2, we compare the effects of different approximations made in the PIMC simulations such as time step $`\tau `$, number of pairs $`N_P`$ and the type of nodal restriction. For pressures above 3 Mbar, all these approximations have a very small effect. The reason is that PIMC simulation become increasingly accurate as temperature increases. The first noticeable difference occurs at $`p2.7\mathrm{Mbar}`$, which corresponds to $`T=\mathrm{62\hspace{0.17em}500}\mathrm{K}`$. At lower pressures, the differences become more and more pronounced. We have performed simulations with free particle nodes and $`N_P=32`$ for three different values of $`\tau `$. Using a smaller time step makes the simulations computationally more demanding and it shifts the hugoniot curves to lower densities. These differences come mainly from enforcing the nodal surfaces more accurately, which seems to be more relevant than the simultaneous improvements in the accuracy of the action $`S`$, that is the time step is constrained more by the Fermi statistics than it is by the potential energy. We improved the efficiency of the algorithm by using a smaller time step $`\tau _F`$ for evaluating the Fermi action than the time step $`\tau _B`$ used for the potential action. Unless specified otherwise, we used $`\tau _F=\tau _B=\tau `$. At even lower pressures not shown in Fig. 2, all of the hugoniot curves with FP nodes turn around and go to low densities as expected.
As a next step, we replaced the FP nodes by VDM nodes. Those results show that the form of the nodes has a significant effect for $`p`$ below 2 Mbar. Using a smaller $`\tau `$ also shifts the curve to slightly lower densities. In the region where atoms and molecules are forming, it is plausible that VDM nodes are more accurate than free nodes because they can describe those states . We also show a hugoniot derived on the basis of the VDM alone (dashed line). These results are quite reasonable considering the approximations (Hartree-Fock) made in that calculation. Therefore, we consider the PIMC simulation with the smallest time step using VDM nodes ($``$) to be our most reliable hugoniot. Going to bigger system sizes $`N_P=64`$ and using FP nodes also shows a shift towards lower densities.
Fig. 3 compares the Hugoniot from Laser shock wave experiments with PIMC simulation (VDM nodes, $`\tau ^1=210^6\mathrm{K}`$) and several theoretical approaches: SESAME model by Kerley (thin solid line), linear mixing model by Ross (dashed line) , DFT-MD by Lenosky et al. (dash-dotted line), Padé approximation in the chemical picture (PACH) by Ebeling et al. (dotted line), and the work by Saumon et al. (thin dash-dotted line).
The differences of the various PIMC curves in Fig. 2 are small compared to the deviation from the experimental results . There, an increased compressibility with a maximum value of $`6\pm 1`$ was found while PIMC predicts $`4.3\pm 0.1`$, only slightly higher than that given by the SESAME model. Only for $`p>2.5\mathrm{Mbar}`$, does our hugoniot lie within experimental errorbars. In this regime, the deviations in the PIMC and PACH hugoniot are relatively small, less than $`0.05\mathrm{gcm}^3`$ in density. In the high pressure limit, the hugoniot goes to the FP limit of 4-fold compression. This trend is also present in the experimental findings. For pressures below 1 Mbar, the PIMC hugoniot goes back lower densities and shows the expected tendency towards the experimental values from earlier gas gun work and lowest data points from . For these low pressures, differences between PIMC and DFT-MD are also relatively small.
## IV Conclusions
We reported results from PIMC simulations and performed a finite size and time step study. Special emphasis was put on improving the fermion nodes where we presented the first PIMC results with variational instead of FP nodes. We find a slightly increased compressibility of $`4.3\pm 0.1`$ compared to the SESAME model but we cannot reproduce the experimental findings of values of about $`6\pm 1`$. Further theoretical and experimental work will be needed to resolve this discrepancy.
###### Acknowledgements.
The authors would like to thank W. Magro for the collaboration concerning the parallel PIMC simulations and E.L. Pollock for the contributions to the VDM method. This work was supported by the CSAR program and the Department of Physics at the University of Illinois. We used the computational facilities at the National Center for Supercomputing Applications and Lawrence Livermore National Laboratories. |
no-problem/0001/math0001153.html | ar5iv | text | # Introduction
## Introduction
Let $`B`$ be an ideal in a polynomial ring $`R=k[X_1,\mathrm{},X_n]`$ in $`n`$ variables over a field $`k`$. The local cohomology of $`R`$ at $`B`$ is defined by
$$H_B^i(R)=\mathrm{lim}\mathrm{Ext}_R^i(R/B^d,R).$$
In general, this limit is not well behaved: the natural maps
$$\mathrm{Ext}_R^i(R/B^d,R)H_B^i(R)$$
are not injective and it is difficult to understand how their images converge to $`H_B^i(R)`$ (see Eisenbud, Mustaţǎ and Stillman for a discussion of related problems).
However, in the case when $`B`$ is a monomial ideal we will see that the situation is especially nice if instead of the sequence $`\{B^d\}_{d1}`$ we consider the cofinal sequence of ideals $`\{B_0^{[d]}\}_{d1}`$, consisting of the “Frobenius powers” of the ideal $`B_0=\mathrm{radical}(B)`$. They are defined as follows: if $`m_1,\mathrm{},m_r`$ are monomial generators of $`B_0`$, then
$$B_0^{[d]}=(m_1^d,\mathrm{},m_r^d).$$
Our first main result is that the natural map
$$\mathrm{Ext}_R^i(R/B_0^{[d]},R)H_B^i(R)$$
is an isomorphism onto the submodule of $`H_B^i(R)`$ of elements of multidegree $`\alpha `$, with $`\alpha _jd`$ for all $`j`$.
The second main result gives a filtration of $`\mathrm{Ext}_R^i(R/B,R)`$ for a squarefree monomial ideal $`B`$. For $`\alpha \{0,1\}^n`$, let $`\mathrm{supp}(\alpha )=\{j|\alpha _j=1\}`$ and $`P_\alpha =(X_j|j\mathrm{supp}(\alpha ))`$.
We describe a canonical filtration of
$$\mathrm{Ext}_R^i(R/B,R):0=M_0\mathrm{}M_n=\mathrm{Ext}_R^i(R/B,R)$$
such that for every $`l`$,
$$M_l/M_{l1}\underset{|\alpha |=l}{}(R/P_\alpha (\alpha ))^{\beta _{li,\alpha }(B^{})}.$$
The numbers $`\beta _{li,\alpha }(B^{})`$ are the Betti numbers of $`B^{}`$, the Alexander dual ideal of $`B`$ (see section 3 below for the related definitions). For an interpretation of this filtration in terms of Betti diagrams, see Remark 1 after Theorem 3.3, below.
In a slightly weaker form, this result has been conjectured by David Eisenbud.
Let’s see this filtration for a simple example: $`R=k[a,b,c,d]`$, $`B=(ab,cd)`$ and $`i=2`$. Since $`B`$ is a complete intersection, we get $`\mathrm{Ext}_R^2(R/B,R)R/B(1,1,1,1)`$. Our filtration is $`M_0=M_1=0`$, $`M_2=R\overline{a}\overline{c}+R\overline{a}\overline{d}+R\overline{b}\overline{c}+R\overline{b}\overline{d}`$, $`M_3=R\overline{a}+R\overline{b}+R\overline{c}+R\overline{d}`$ and $`M_4=\mathrm{Ext}_R^2(R/B,R)`$.
From the description of $`\mathrm{Ext}_R^2(R/B,R)`$ it follows that
$$M_2/M_1=M_2R/(b,d)(0,1,0,1)R/(b,c)(0,1,1,0)$$
$$R/(a,d)(1,0,0,1)R/(a,c)(1,0,1,0),$$
$$M_3/M_2R/(b,c,d)(0,1,1,1)R/(a,c,d)(1,0,1,1)$$
$$R/(a,b,d)(1,1,0,1)R/(a,b,c)(1,1,1,0),$$
$$M_4/M_3R/(a,b,c,d)(1,1,1,1).$$
On the other hand, $`B^{}=(bd,bc,ad,ac)`$. If $`F_{}`$ is the minimal multigraded resolution of $`B^{}`$, then
$$F_0=R(0,1,0,1)R(0,1,1,0)R(1,0,0,1)R(1,0,1,0),$$
$$F_1=R(0,1,1,1)R(1,0,1,1)R(1,1,0,1)R(1,1,1,0),$$
$$F_2=R(1,1,1,1).$$
We see that for each $`\alpha \{0,1\}^4`$ such that $`R(\alpha )`$ appears in $`F_{l2}`$, there is a corresponding summand $`R/P_\alpha (\alpha )`$ in $`M_l/M_{l1}`$.
In order to prove this result about the filtration of $`\mathrm{Ext}_R^i(R/B,R)`$ we will study the multigraded components of this module and how an element of the form $`X_jR`$ acts on these components. As we have seen, it is enough to study the same problem for $`H_B^i(R)`$.
We give two descriptions for the degree $`\alpha `$ part of $`H_B^i(R)`$, as simplicial cohomology groups of certain simplicial complexes depending only on $`B`$ and the signs of the components of $`\alpha `$. The first complex is on the set of minimal generators of $`B`$ and the second one is a full subcomplex of the simplicial complex associated to $`B^{}`$ via the Stanley-Reisner correspondence. The module structure on $`H_B^i(R)`$ is described by the maps induced in cohomology by inclusion of simplicial complexes.
As a first consequence of these results and using also a formula of Hochster , we obtain an isomorphism
$$\mathrm{Ext}_R^i(R/B,R)_\alpha \mathrm{Tor}_{|\alpha |i}^R(B^{},k)_\alpha ,$$
for every $`\alpha \{0,1\}^n`$.
This result is equivalent to the fact that in our filtration the numbers are as stated above. This isomorphism has been obtained also by Yanagawa . It can be considerd as a strong form of the inequality of Bayer, Charalambous and Popescu between the Betti numbers of $`B`$ and those of $`B^{}`$. As shown in that paper, this implies that $`B`$ and $`B^{}`$ have the same extremal Betti numbers, extending results of Eagon and Reiner and Terai .
As a final application of our analysis of the graded pieces of $`\mathrm{Ext}_R^i(R/B,R)`$, we give a topological description for the associated primes of $`\mathrm{Ext}_R^i(R/B,R)`$. In the terminology of Vasconcelos , these are the homological associated primes of $`R/B`$. In particular, we characterize the minimal associated primes of $`\mathrm{Ext}_R^i(R/B,R)`$ using only the Betti numbers of $`B^{}`$.
We mention here the recent work of Terai on the Hilbert function of the modules $`H_B^i(R)`$. It is easy to see that using the results in our paper one can deduce Terai’s formula for this Hilbert function.
The problem of effectively computing the local cohomology modules with respect to an arbitrary ideal is quite difficult since these modules are not finitely generated. The general approach is to use the $`D`$-module structure for the local cohomology (see, for example, Walther ). However, in the special case of monomial ideals our results show that it is possible to make this computation with elementary methods.
Our main motivation for studying local cohomology at monomial ideals comes from the applications in the context of toric varieties. Via the homogeneous coordinate ring, the cohomology of sheaves on such a variety can be expressed as local cohomology of modules at the “irrelevant ideal”, which is a squarefree monomial ideal. For a method of computing the cohomology of sheaves on toric varieties in this way, see Eisenbud, Mustaţǎ and Stillman . For applications to vanishing theorems on toric varieties and related results, see Mustaţǎ .
The main reference for the definitions and the results that we use is Eisenbud . For the basic facts about the cohomology of simplicial complexes, see Munkres . Cohomology of simplicial complexes is always taken to be reduced cohomology. Notice also that we make a distinction between the empty complex which contains just the empty set (which has nontrivial cohomology in degree $`1`$) and the void complex which doesn’t contain any set (whose cohomology is trivial in any degree).
This work has been done in connection with a joint project with David Eisenbud and Mike Stillman. We would like to thank them for their constant encouragement and for generously sharing their insight with us. We are also grateful to Josep Alvarez Montaner for pointing out a mistake in an earlier version of this paper.
## §1. Local cohomology as a union of Ext modules
Let $`BR=k[X_1,\mathrm{},X_n]`$ be a squarefree monomial ideal. All the modules which appear are $`𝐙^n`$-graded. We partially order the elements of $`𝐙^n`$ by setting $`\alpha \beta `$ iff $`\alpha _j\beta _j`$, for all $`j`$.
###### Theorem 1.1
For each $`i`$ and $`d`$, the natural map
$$\mathrm{Ext}_R^i(R/B^{[d]},R)H_B^i(R)$$
is an isomorphism onto the submodule of $`H_B^i(R)`$ of elements of degree $`(d,\mathrm{},d)`$.
Proof. We will compute $`\mathrm{Ext}_R^i(R/B^{[d]},R)`$ using the Taylor resolution $`F_{}^d`$ of $`R/B^{[d]}`$ ( see Eisenbud , exercise 17.11). The inclusion $`B^{[d+1]}B^{[d]}`$, $`d1`$ induces a morphism of complexes $`\varphi ^d:F_{}^{d+1}F_{}^d`$. The assertions in the theorem are consequences of the more precise lemma below.
###### Lemma 1.2
If $`(\varphi ^d)^{}:(F_{}^d)^{}(F_{}^{d+1})^{}`$ is the dual $`Hom_R(\varphi ^d,R)`$ of the above map, then in a multidegree $`\alpha 𝐙^n`$ we have:
(a) If $`\alpha (d,\mathrm{},d)`$, then $`(\varphi ^d)_\alpha ^{}`$ is an isomorphism of complexes.
(b) If $`\alpha _j<d`$ for some $`j`$, $`1jn`$, then $`(F_{}^d)_\alpha ^{}=0`$, so $`(\varphi ^d)_\alpha ^{}`$ is the zero map.
Proof of the lemma. Let $`m_1,\mathrm{},m_r`$ be monomial minimal generators of $`B`$. For any subset $`I`$ of $`\{1,\mathrm{},r\}`$ we set
$$m_I=\mathrm{LCM}\{m_iiI\}.$$
As each $`m_I`$ is square-free, $`degm_I𝐙^n`$ is a vector of ones and zeros.
Recall from Eisenbud that $`F_{}^d`$ is a free $`R`$-module with basis $`\{f_I^d|I\{1,\mathrm{},r\}\}`$, where $`deg(f_I^d)=ddeg(m_I)`$. Therefore, the degree $`\alpha `$ part of $`(F_{}^d)^{}`$ has a vector space basis consisting of elements of the form $`ne_I^d`$ where $`nR`$ is a monomial, $`e_I^d=(f_I^d)^{}`$ has degree equal to $`ddeg(m_I)`$, and $`deg(n)ddeg(m_I)=\alpha .`$
Part (b) of the Lemma follows at once. For part (a), note that $`(\varphi ^k)^{}:(F_{}^d)^{}(F_{}^{d+1})^{}`$ takes $`e_I^d`$ to $`m_Ie_I^{d+1}`$. The vector $`deg(e_I^{d+1})=(d+1)deg(m_I)`$ has entry $`(d+1)`$ wherever $`deg(m_I)`$ has entry 1, so any element $`ne_I^{d+1}`$ of degree $`\alpha (d,d,\mathrm{},d)`$ must have $`n`$ divisible by $`m_I`$. It is thus of the form $`(\varphi ^d)^{}(x)`$ for the unique element $`x=(n/m_I)e_I^d`$, as required.
## §2. Local cohomology as simplicial cohomology
To describe $`H_B^i(R)`$ in a multidegree $`\alpha 𝐙^n`$, we will use two simplicial complexes associated with $`B`$ and $`\alpha `$. We will assume that $`B(0)`$.
By computing local cohomology using the Taylor complex we will express $`H_B^i(R)_\alpha `$ as the simplicial cohomology of a complex on the set of minimal generators of $`B`$. We will interpret this later as the cohomology of an other complex, this time on the potentially smaller set $`\{1,\mathrm{},n\}`$. This one is a full subcomplex of the complex associated to the dual ideal $`B^{}`$ via the Stanley-Reisner correspondence. In fact, this is the complex used in the computation of the Betti numbers of $`B^{}`$ (see the next section for the definitions). We will use this result to derive the relation between $`\mathrm{Ext}(R/B,R)`$ and $`\mathrm{Tor}^R(B^{},k)`$ in Corolary 3.1 below.
Let $`m_1,\mathrm{},m_r`$ be the minimal monomial generators of $`B`$. As above, for $`J\{1,\mathrm{},r\}`$, $`m_J`$ will denote $`\mathrm{LCM}(m_j;jJ)`$.
For $`i\{1,\mathrm{},n\}`$, we define
$$T_i:=\{J\{1,\mathrm{},r\}|X_im_J\}.$$
For every subset $`I\{1,\mathrm{},n\}`$, we define $`T_I=_{iI}T_i`$. When $`I=\mathrm{}`$, we take $`T_I`$ to be the void complex. It is clear that each $`T_i`$ is a simplicial complex on the set $`\{1,\mathrm{},r\}`$, and therefore so is $`T_I`$.
For $`\alpha 𝐙^n`$, we take $`I_\alpha =\{i|\alpha _i1\}\{1,\mathrm{},n\}`$. Note that the complex $`T_{I_\alpha }`$ depends only on the signs of the components of $`\alpha `$ (and, of course, on $`B`$).
If $`e_1,\mathrm{},e_n`$ is the canonical basis of $`𝐙^n`$ and $`\alpha ^{}=\alpha +e_l`$, we have obviously $`I_\alpha ^{}I_\alpha `$, with equality iff $`\alpha _l1`$. Therefore, $`T_{I_\alpha ^{}}`$ is a subcomplex of $`T_{I_\alpha }`$.
###### Theorem 2.1
(a) With the above notation, we have
$$H_B^i(R)_\alpha H^{i2}(T_{I_\alpha };k).$$
(b) Via the isomorphisms given in (a), the multiplication by $`X_l`$:
$$\nu _{X_l}:H_B^i(R)_\alpha H_B^i(R)_\alpha ^{}$$
corresponds to the morphism:
$$H^{i2}(T_{I_\alpha };k)H^{i2}(T_{I_\alpha ^{}};k),$$
induced in cohomology by the inclusion $`T_{I_\alpha ^{}}T_{I_\alpha }`$. In particular, if $`\alpha _l1`$, then $`\nu _{X_l}`$ is an isomorphism.
Proof. We have seen in Lemma 1.2 that
$$\mathrm{Ext}_R^i(R/B^{[d]},R)_\alpha H_B^i(R)_\alpha $$
if $`\alpha (d,\mathrm{},d)`$. We fix such a $`d`$. With the notations in Lemma 1.2 , we have seen that the degree $`\alpha `$ part of $`(F_{}^d)^{}`$ has a vector space basis consisting of elements of the form $`ne_J^d`$, where $`nR`$ is a monomial and $`deg(n)ddeg(m_J)=\alpha `$. Therefore, the basis of $`(F_p^d)_\alpha ^{}`$ is indexed by those $`J\{1,\mathrm{},r\}`$ with $`|J|=p`$ and $`\alpha +ddeg(m_J)(0,\mathrm{},0)`$. Because $`\alpha _j1`$ iff $`jI_\alpha `$ and $`\alpha (d,\mathrm{},d)`$, the above inequality is equivalent to $`X_j|m_J`$ for every $`jI_\alpha `$ i.e. to $`JT_{I_\alpha }`$.
Let $`G^{}`$ be the cochain complex computing the relative cohomology of the pair $`(D,T_{I_\alpha })`$ with coefficients in $`k`$, where $`D`$ is the full simplicial complex on the set $`\{1,\mathrm{},r\}`$.
If $`I_\alpha \mathrm{}`$, then the degree $`\alpha `$ part of $`(F_p^d)^{}`$ is equal to $`G^{p1}`$ for every $`p`$. Moreover, the maps are the same and therefore we get $`H_B^i(R)_\alpha H^{i1}(D,T_{I_\alpha };k)`$. Since $`D`$ is contractible, the long exact sequence in cohomology of the pair $`(D,T_{I_\alpha })`$ yields $`H_B^i(R)_\alpha H^{i2}(T_{I_\alpha };k)`$.
If $`I_\alpha =\mathrm{}`$, then $`(F_{}^d)^{}`$ in degree $`\alpha `$ is up to a shift the complex computing the reduced cohomology of $`D`$ with coefficients in $`k`$. Since $`D`$ is contractible, we get $`H_B^i(R)_\alpha =0=H^{i2}(T_{I_\alpha };k)`$, which completes the proof of part (a).
For part (b), we may suppose that $`I_\alpha ^{}(0)`$. With the above notations, $`\nu _{X_l}`$ is induced by the map $`\varphi _l:(F_p^d)_\alpha ^{}(F_p^d)_\alpha ^{}^{}`$, given by $`\varphi _l(ne_J^d)=X_lne_J^d`$.
If $`G^{}`$ is constructed as above, but for $`\alpha ^{}`$ instead of $`\alpha `$, then via the isomorphisms:
$$(F_p^d)_\alpha ^{}G^{p1},$$
$$(F_p^d)_\alpha ^{}^{}G^{p1},$$
the map $`\varphi _l`$ corresponds to the canonical projection $`G^{p1}G^{p1}`$, which concludes the proof of part (b).
###### Remark
The last assertion in Theorem 2.1(b), that $`\nu _{X_l}`$ is an isomorphism if $`\alpha _l1`$ has been obtained also in Yanagawa .
The next corollary describes $`H_B^i(R)_\alpha `$ as the cohomology of a simplicial complex with vertex set $`\{1,\mathrm{},n\}`$.
We first introduce the complex $`\mathrm{\Delta }`$ defined by:
$$\mathrm{\Delta }:=\{F\{1,\mathrm{},n\}|\underset{jF}{}X_jB\}.$$
In fact , by the Stanley-Reisner correspondence between square-free monomial ideals and simplicial complexes (see Bruns and Herzog ), $`\mathrm{\Delta }`$ corresponds to $`B^{}`$.
For any subset $`I\{1,\mathrm{},n\}`$, we define $`\mathrm{\Delta }_I`$ to be the full simplicial subcomplex of $`\mathrm{\Delta }`$ supported on $`I`$:
$$\mathrm{\Delta }_I:=\{F\{1,\mathrm{},n\}|F\mathrm{\Delta },FI\}.$$
When $`I=\mathrm{}`$, we take $`\mathrm{\Delta }_I`$ to be the void complex. It is clear that if $`II^{}`$, then $`\mathrm{\Delta }_I^{}`$ is a subcomplex of $`\mathrm{\Delta }_I`$. This is the case if $`\alpha ^{}=\alpha +e_l`$, $`I=I_\alpha `$ and $`I^{}=I_\alpha ^{}`$.
###### Corollary 2.2
(a) With the above notation, for any $`\alpha 𝐙^n`$
$$H_B^i(R)_\alpha H^{i2}(\mathrm{\Delta }_{I_\alpha };k).$$
(b) Via the isomorphisms given by (a), the multiplication map $`\nu _{X_l}`$ corresponds to the morphism:
$$H^{i2}(\mathrm{\Delta }_{I_\alpha };k)H^{i2}(\mathrm{\Delta }_{I_\alpha ^{}};k),$$
induced in cohomology by the inclusion $`\mathrm{\Delta }_{I_\alpha ^{}}\mathrm{\Delta }_{I_\alpha }`$.
Proof. Using the notation in Theorem 2.1, if $`I_\alpha \mathrm{}`$, then $`T_{I_\alpha }=_{iI_\alpha }T_i`$.
If $`i_1,\mathrm{},i_kI_\alpha `$ and $`_{1pk}T_{i_p}\mathrm{}`$, then
$$\underset{1pk}{}T_{i_p}=\{J\{1,\mathrm{},r\}|X_{i_p}m_J,1pk\}$$
is the full simplicial complex on those $`j`$ with $`X_{i_p}m_j`$, for every $`p`$, $`1pk`$. Therefore it is contractible.
This shows that we can compute the cohomology of $`T_I`$ as the cohomology of the nerve $`𝒩`$ of the cover $`T_I=_{iI}T_i`$ (see Godement ). But by definition, $`\{i_1,\mathrm{},i_k\}I`$ is a simplex in $`𝒩`$ iff $`_{1pk}T_{i_p}\mathrm{}`$ iff there is $`j`$ such that $`X_{i_p}m_j`$ for every p, $`1pk`$. This shows that $`𝒩=\mathrm{\Delta }_I`$ and we get that $`H_B^i(R)_\alpha H^{i2}(\mathrm{\Delta }_I;k)`$ when $`I\mathrm{}`$.
When $`I=\mathrm{}`$, $`H_B^i(R)_\alpha =0`$ by theorem 2.1 and also $`H^{i2}(\mathrm{\Delta }_I;k)=0`$ (the reduced cohomology of the void simplicial complex is zero).
Part (b) follows immediately from part (b) in Theorem 2.1 and the fact that the isomorphism between the cohomology of a space and that of the nerve of a cover as above is functorial.
###### Remark
The same type of arguments as in the proofs of Theorem 2.1 and of Corollary 2.2 can be used to give a topological description for $`\mathrm{Ext}_B^i(R/B,R)_\alpha `$, for a possibly non-reduced nonzero monomial ideal $`B`$. Namely, for $`\alpha 𝐙^n`$, we define the simplicial complex $`\mathrm{\Delta }_\alpha `$ on $`\{1,\mathrm{},n\}`$ by $`J\mathrm{\Delta }_\alpha `$ iff there is a monomial $`m`$ in $`B`$ such that $`deg(X^\alpha m)_j<0`$ for $`jJ`$. We make the convention that $`\mathrm{\Delta }`$ is the void complex iff $`\alpha 0`$. Then
$$\mathrm{Ext}_R^i(R/B,R)_\alpha H^{i2}(\mathrm{\Delta }_\alpha ;k).$$
Moreover, we can describe these $`k`$-vector spaces using a more geometric object. If we view $`B𝐙^n𝐑^n`$, let $`P_\alpha `$ be the subspace of $`\mathrm{R}^n`$ supported on $`B`$, translated by $`\alpha `$, minus the first quadrant. More precisely,
$$P_\alpha =\{x𝐑^n|x\alpha m,\mathrm{for}\mathrm{some}mB\}𝐑_+^n.$$
Then, using a similar argument to the one in the proof of corollary 1.4, one can show that
$$\mathrm{Ext}_R^i(R/B,R)_\alpha H^{i2}(P_\alpha ;k),$$
where the right-hand side is the reduced singular cohomology group. Here we have to make the convention that for $`\alpha 0`$, $`P_\alpha `$ is the “void topological space”, with trivial reduced cohomology (as oposed to the empty topological space which has nonzero reduced cohomology in degree $`1`$).
We leave the details of the proof to the interested reader.
## §3. The filtration on the Ext modules
The Alexander dual of a reduced monomial ideal $`B`$ is defined by
$$B^{}=(X^F|F\{1,\mathrm{}n\},X^{F^c}B),$$
where $`F^c:=\{1,\mathrm{},n\}F`$ (see Bayer, Charalambous and Popescu for interpretation in terms of Alexander duality ). Note that $`(B^{})^{}=B`$.
We will derive first a relation between $`\mathrm{Ext}_R(R/B,R)`$ and $`\mathrm{Tor}^R(B^{},k)`$. This can be seen as a stronger form of the inequality in Bayer, Charalambous and Popescu between the Betti numbers of $`B`$ and $`B^{}`$.
For $`\alpha 𝐙^n`$, we will denote $`|\alpha |=_i\alpha _i`$.
###### Corollary 3.1
Let $`BR=k[X_1,\mathrm{},X_n]`$ be a reduced monomial ideal and $`\alpha 𝐙^n`$ a multidegree. If $`\alpha \{0,1\}^n`$, then $`\mathrm{Tor}_i^R(B^{},k)_\alpha =0`$, and if $`\alpha \{0,1\}^n`$, then
$$\mathrm{Tor}_i^R(B^{},k)_\alpha \mathrm{Ext}_R^{|\alpha |i}(R/B,R)_\alpha .$$
Proof. We will use Hochster’s formula for the Betti numbers of reduced monomial ideals (see, for example, Hochster or Bayer, Charalambous and Popescu ). It says that if $`\alpha \{0,1\}^n`$, then $`\mathrm{Tor}_i^R(B^{},k)_\alpha =0`$ and if $`\alpha \{0,1\}^n`$, then
$$\mathrm{Tor}_i^R(B^{},k)_\alpha H^{|\alpha |i2}(\mathrm{\Delta }_I;k),$$
where $`I`$ is the support of $`\alpha `$.
Obviously, we may suppose that $`B(0)`$. If $`\alpha \{0,1\}^n`$, then corollary 2.2 gives
$$H^{|\alpha |i2}(\mathrm{\Delta }_I;k)H_B^{|\alpha |i}(R)_\alpha $$
and theorem 1.1 gives
$$H_B^{|\alpha |i}(R)_\alpha \mathrm{Ext}_R^{|\alpha |i}(R/B,R)_\alpha .$$
Putting together these isomorphisms, we get the assertion of the corollary.
We recall that the multigraded Betti numbers of $`B`$ are defined by
$$\beta _{i,\alpha }(B):=\mathrm{dim}_\mathrm{k}\mathrm{Tor}_\mathrm{i}^\mathrm{R}(\mathrm{B},\mathrm{k})_\alpha .$$
Equivalently, if $`F_{}`$ is a multigraded minimal resolution of $`B`$, then
$$F_i\underset{\alpha 𝐙^n}{}R(\alpha )^{\beta _{i,\alpha }(B)}.$$
One says that $`(i,\alpha )`$ is extremal (or that $`\beta _{i,\alpha }`$ is extremal) if $`\beta _{j,\alpha ^{}}(B)=0`$ for all $`ji`$ and $`\alpha ^{}>\alpha `$ such that $`|\alpha ^{}||\alpha |ji`$.
###### Remark
Using Theorems 1.1, 2.1(b) and Corollary 3.1 one can give a formula for the Hilbert function of $`H_B^i(R)`$ using the Betti numbers of $`B^{}`$. This formula is equivalent to the one which appears in Terai .
As a consequence of the above corollary, we obtain the inequality between the Betti numbers of $`B`$ and $`B^{}`$ from Bayer, Charalambous and Popescu . It implies the equality of extremal Betti numbers from that paper, in particular the equality $`\mathrm{reg}B=\mathrm{pd}(R/B^{})`$ from Terai .
###### Corollary 3.2
If $`BR`$ is a reduced monomial ideal, then
$$\beta _{i,\alpha }(B)\underset{\alpha \alpha ^{}}{}\beta _{|\alpha |i1,\alpha ^{}}(B^{}),$$
for every $`i0`$ and every $`\alpha \{0,1\}^n`$. If $`\beta _{|\alpha |i1,\alpha }(B^{})`$ is extremal, then so is $`\beta _{i,\alpha }(B)`$ and
$$\beta _{i,\alpha }(B)=\beta _{|\alpha |i1,\alpha }(B^{}).$$
Proof. Since $`\beta _{i,\alpha }(B)=\mathrm{dim}_\mathrm{k}\mathrm{Tor}_\mathrm{i}^\mathrm{R}(\mathrm{B},\mathrm{k})_\alpha `$, by the previous corollary we get
$$\beta _{i,\alpha }(B)=\mathrm{dim}_\mathrm{k}\mathrm{Ext}_\mathrm{R}^{|\alpha |\mathrm{i}}(\mathrm{R}/\mathrm{B}^{},\mathrm{R})_\alpha =\mathrm{dim}_\mathrm{k}H^{|\alpha |\mathrm{i}}(\mathrm{Hom}(\mathrm{F}_{},\mathrm{R}))_\alpha ,$$
where $`F_{}`$ is the minimal free resolution of $`R/B^{}`$.
Since $`F_{|\alpha |i}=_{\alpha ^{}𝐙^n}R(\alpha ^{})^{\beta _{|\alpha |i1,\alpha ^{}}(B^{})}`$, we get
$$\beta _{i,\alpha }(B)\underset{\alpha ^{}𝐙^n}{}\beta _{|\alpha |i1,\alpha ^{}}(B^{})\mathrm{dim}_\mathrm{k}(\mathrm{R}(\alpha ^{})_\alpha )=\underset{\alpha \alpha ^{}}{}\beta _{|\alpha |\mathrm{i}1,\alpha ^{}}(\mathrm{B}^{}).$$
If $`\beta _{|\alpha |i1,\alpha }(B^{})`$ is extremal, the above inequality becomes $`\beta _{i,\alpha }(B)\beta _{|\alpha |i1,\alpha }(B^{})`$. Applying the same inequality for $`ji`$ and $`\alpha ^{}>\alpha `$ such that $`|\alpha ^{}||\alpha |ji`$ and the fact that $`\beta _{|\alpha |i1,\alpha }(B^{})`$ is extremal, we get that $`\beta _{i,\alpha }(B)`$ is extremal.
Applying the previous inequality with $`B`$ replaced by $`B^{}`$, we obtain $`\beta _{|\alpha |i1,\alpha }(B^{})\beta _{i,\alpha }(B)`$, which concludes the proof.
We fix some notations for the remaining of this section. Let $`[n]=\{0,1\}^n`$ and $`[n]_l=\{\alpha [n]||\alpha |=l\}`$, for every $`l`$, $`0ln`$. For $`\alpha [n]`$, let $`\mathrm{supp}(\alpha )=\{j|\alpha _j=1\}`$ and $`P_\alpha =(X_j|j\mathrm{supp}(\alpha ))`$. The ideals $`P_\alpha `$, $`\alpha [n]`$ are exactly the monomial prime ideals of $`R`$.
The following theorem gives the canonical filtration of $`\mathrm{Ext}_R^i(R/B,R)`$ announced in the Introduction.
###### Theorem 3.3
Let $`BR`$ be a squarefree monomial ideal. For each $`l`$, $`0ln`$, let $`M_l`$ be the submodule of $`\mathrm{Ext}_R^i(R/B,R)`$ generated by all $`\mathrm{Ext}_R^i(R/B,R)_\alpha `$, for $`\alpha [n]`$, $`|\alpha |l`$. Then $`M_0=0`$, $`M_n=\mathrm{Ext}_R^i(R/B,R)`$ and for every $`l`$, $`0ln`$,
$$M_l/M_{l1}\underset{\alpha [n]_l}{}(R/P_\alpha (\alpha ))^{\beta _{li,\alpha }(B^{})}.$$
Proof. Clearly we may suppose $`B0`$. The fact that $`M_0=0`$ follows from Corollary 2.2(a).
Let’s see first that $`M_n=\mathrm{Ext}_R^i(R/B,R)`$. For this it is enough to prove that all the minimal monomial generators of $`\mathrm{Ext}_R^i(R/B,R)`$ are in degrees $`\alpha `$, $`\alpha [n]`$.
Indeed, if $`\alpha _j1`$ for some $`j`$, then the multiplication by $`X_j`$ defines an isomorphism
$$\mathrm{Ext}_R^i(R/B,R)_{\alpha e_j}\mathrm{Ext}_R^i(R/B,R)_\alpha $$
by Corollary 2.2(b) and Theorem 1.1. In particular, there are no minimal generators in degree $`\alpha `$.
On the other hand, by Theorem 1.1, $`\mathrm{Ext}_R^i(R/B,R)_\alpha =0`$ if $`\alpha _j2`$, for some $`j`$. Therefore we have $`M_n=\mathrm{Ext}_R^i(R/B,R)`$.
Suppose now that we have homogeneous elements $`f_1,\mathrm{},f_r`$ with $`deg(f_q)[n]_l^{}`$, $`l^{}l`$, for every $`q`$, $`1qr`$. We suppose that they are linearly independent over $`k`$ and that their linear span contains $`\mathrm{Ext}_R^i(R/B,R)_\alpha `$, for every $`\alpha [n]_l^{}`$, $`l^{}l1`$. We will suppose also that $`deg(f_r)=\alpha `$, $`|\alpha |=l`$. If $`T:=_{1qr1}Rf_q`$, let $`\overline{f}_r`$ be the image of $`f_r`$ in $`M_l/T`$.
Claim. With the above notations, $`\mathrm{Ann}_R(\overline{f}_r)=P_\alpha `$.
Let $`F=\mathrm{supp}(\alpha )`$. If $`jF`$, then $`deg(X_jf_r)=(\alpha e_j)`$, $`\alpha e_j[n]`$. By our assumption, it follows that $`X_jf_rT`$, so that $`P_\alpha \mathrm{Ann}_R(\overline{f}_r)`$.
Conversely, consider now $`m=X_j^{m_j}\mathrm{Ann}\overline{f}_r`$ and suppose that $`m(X_j|jF)`$. We can suppose that $`m`$ has minimal degree. Let $`j`$ be such that $`m_j1`$. Then $`jF`$ and therefore $`m_j\alpha _j=m_j1`$. Since $`mf_rT`$, we can write
$$mf_r=\underset{q<r}{}c_qn_qf_q,$$
where $`n_q`$ are monomials and $`c_qk`$. Since $`deg(f_q)0`$ for every $`q`$, in the above equality we may assume that $`X_j|n_q`$ for every $`q`$ such that $`c_q0`$. But by Corollary 2.2(b) and Theorem 1.1, the multiplication by $`X_j`$ is an isomorphism:
$$\mathrm{Ext}_R^i(R/B,R)_{\alpha +degme_j}\mathrm{Ext}_R^i(R/B,R)_{\alpha +degm}.$$
Therefore $`m/X_j\mathrm{Ann}\overline{f}_r`$, in contradiction with the minimality of $`m`$. We get $`\mathrm{Ann}\overline{f}_r=(X_j|jF)`$, which completes the proof of the claim.
The first consequence is that for every nonzero $`fM_l`$, $`deg(f)=\alpha `$, $`\alpha [n]_l`$, if $`\overline{f}`$ is the image of $`f`$ in $`M_l/M_{l1}`$, then $`\mathrm{Ann}_R(\overline{f})=P_\alpha `$, so that $`R\overline{f}R/P_\alpha (\alpha )`$.
Let’s consider now a homogeneous basis $`f_1,\mathrm{},f_N`$ of $`_{\alpha [n]_l}\mathrm{Ext}_R^i(R/B,R)_\alpha `$. By Corollary 3.1,
$$\mathrm{dim}_\mathrm{k}\mathrm{Ext}_\mathrm{R}^\mathrm{i}(\mathrm{R}/\mathrm{B},\mathrm{R})_\alpha =\beta _{\mathrm{l}\mathrm{i},\alpha }(\mathrm{B}^{}).$$
Therefore, to complete the proof of the theorem, it is enough to show that
$$M_l/M_{l1}\underset{1jN}{}R\overline{f}_j.$$
Here $`\overline{f}_j`$ denotes the image of $`f_j`$ in $`M_l/M_{l1}`$.
Since $`M_l=M_{l1}+_{1jN}Rf_j`$, we have only to show that if $`_{1jN}n_jf_jM_{l1}`$, then $`n_jf_jM_{l1}`$ for every $`j`$, $`1jN`$.
Let $`\{g_1,\mathrm{},g_N^{}\}`$ be the union of homogeneous bases for $`\mathrm{Ext}_R^i(R/B,R)_\alpha `$, for $`\alpha [n]_l^{}`$, $`l^{}l1`$.
Let’s fix some $`j`$, with $`1jN`$. If $`deg(f_j)=\alpha `$, by applying the above claim to $`f_j`$, as part of $`\{f_p|\mathrm{\hspace{0.17em}1}pN\}\{g_p^{}|\mathrm{\hspace{0.17em}1}p^{}N^{}\}`$, we get that $`n_jP_\alpha `$. But we have already seen that $`P_\alpha f_jM_{l1}`$ and therefore the proof is complete.
###### Remark 1
We can interpret the statement of Theorem 3.3 using the multigraded Betti diagram of $`B^{}`$. This is the diagram having at the intersection of the $`i^{\mathrm{th}}`$ row with the $`j^{\mathrm{th}}`$ column the Betti numbers $`\beta _{j,\alpha }(B^{})`$, for $`\alpha 𝐙^n`$, $`|\alpha |=i+j`$.
For each $`i`$ and $`j`$ we form a module corresponding to $`(i,j)`$:
$$E_{i,j}=\underset{\alpha [n]_{i+j}}{}(R/P_\alpha (\alpha ))^{\beta _{j,\alpha }(B^{})}.$$
Theorem 3.3 gives a filtration of $`\mathrm{Ext}_R^i(R/B,R)`$ having as quotients the modules constructed above corresponding to the $`i^{\mathrm{th}}`$ row: $`E_{i,j}`$, $`j𝐙`$.
Notice that by definition, $`\mathrm{Tor}_i^R(B^{},k)`$ is obtained by a “dual” procedure applied to the $`i^{\mathrm{th}}`$ column (in this case the extensions being trivial). Indeed, if for $`(j,i)`$ we put
$$E_{j,i}^{}=\underset{\alpha [n]_{i+j}}{}k(\alpha )^{\beta _{i,\alpha }(B^{})},$$
then $`\mathrm{Tor}_i^R(B^{},k)_{j𝐙}E_{j,i}^{}`$.
###### Remark 2
Using Theorem 3.3 one can compute the Hilbert series of $`\mathrm{Ext}_R^i(R/B,R)`$ in terms of the Betti numbers of $`B^{}`$. Using local duality, one can derive the fomula, due to Hochster , for the Hilbert series of the local cohomology modules $`H_{\underset{¯}{m}}^{ni}(R/B)`$, where $`\underset{¯}{m}=(X_1,\mathrm{},X_n)`$ (see also Bruns and Herzog , Theorem 5.3.8).
We describe now the set of homological associated primes of $`R/B`$ i.e. the set
$$_{i0}\mathrm{Ass}(\mathrm{Ext}_R^i(R/B,R))$$
(see Vasconcelos ). Since the module $`\mathrm{Ext}_R^i(R/B,R)`$ is $`𝐙^n`$-graded, its associated primes are of the form $`P_\alpha `$, for some $`\alpha [n]`$. In fact, Theorem 3.3 shows that
$$\mathrm{Ass}(\mathrm{Ext}_R^i(R/B,R))\{P_\alpha |\beta _{|\alpha |i,\alpha }(B^{})0\}.$$
The next result gives the necessary and sufficient condition for a prime ideal $`P_\alpha `$ to be in $`\mathrm{Ass}(\mathrm{Ext}_R^i(R/B,R))`$. In particular, we get the characterization of the minimal associated primes of this module using only the Betti numbers of $`B^{}`$.
###### Theorem 3.4
Let $`BR`$ be a nonzero square-free monomial ideal and $`\alpha [n]`$. Let $`F=\mathrm{supp}(\alpha )`$.
(a) The ideal $`P_\alpha `$ belongs to $`\mathrm{Ass}(\mathrm{Ext}_R^i(R/B,R))`$ iff
$$\underset{jF}{}Ker(H^{i2}(\mathrm{\Delta }_F;k)H^{i2}(\mathrm{\Delta }_{Fj};k))0.$$
(b) The ideal $`P_\alpha `$ is a minimal prime in $`\mathrm{Ass}(\mathrm{Ext}_R^i(R/B,R))`$ iff
$$\beta _{|\alpha |i,\alpha }(B^{})0$$
and
$$\beta _{|\alpha ^{}|i,\alpha ^{}}(B^{})=0,$$
for every $`\alpha ^{}[n]`$, $`\alpha ^{}\alpha `$, $`\alpha ^{}\alpha `$.
Proof. By Corollary 2.2, the condition in (a) is equivalent to the existence of $`u\mathrm{Ext}_R^i(R/B,R)_\alpha `$, $`u0`$ such that $`X_ju=0`$ for every $`jF`$. Since $`\alpha _j=0`$ for $`jF`$, Corollary 2.2(b) and Theorem 1.1 imply that for every monomial $`m`$, $`mP_\alpha `$, the multiplication by $`m`$ is injective on $`\mathrm{Ext}_R^i(R/B,R)_\alpha `$.
Therefore, in the above situation we have $`\mathrm{Ann}_R(u)=P_\alpha `$, so that $`P_\alpha `$ is an element of $`\mathrm{Ass}(\mathrm{Ext}_R^i(R/B,R))`$.
Conversely, suppose that $`P_\alpha \mathrm{Ass}(Ext_R^i(R/B,R))`$. Since $`P_\alpha `$ and $`\mathrm{Ext}_R^i(R/B,R)`$ are $`𝐙^n`$-graded, this is equivalent to the existence of $`u\mathrm{Ext}_R^i(R/B,R)_\alpha ^{}`$, for some $`\alpha ^{}𝐙^n`$, such that $`P_\alpha =\mathrm{Ann}_R(u)`$. To complete the proof of part (a), it is enough to show that we can take $`\alpha ^{}=\alpha `$.
By Theorem 1.1, $`\alpha ^{}(1,\mathrm{},1)`$. Since $`X_ju=0`$ for $`jF`$, multiplication by $`X_j`$ on $`\mathrm{Ext}_R^i(R/B,R)_\alpha `$ is not injective so that by Corollary 2.2(b), we must have $`\alpha _j^{}=1`$ for $`jF`$.
Let’s consider some $`jF`$. If $`\alpha _j^{}1`$, by Corollary 2.2(b) there is $`u^{}\mathrm{Ext}_R^i(R/B,R)_{\alpha ^{\prime \prime }}`$, $`\alpha ^{\prime \prime }=\alpha ^{}\alpha _j^{}e_j`$ such that $`X_j^{\alpha _j^{}}u^{}=u`$ and $`\mathrm{Ann}_R(u^{})=\mathrm{Ann}_R(u)=P_\alpha `$. Therefore, we may suppose that $`\alpha _j^{}0`$.
If $`\alpha _j^{}=1`$, since $`X_j\mathrm{Ann}_R(u)`$, which is prime, we have $`\mathrm{Ann}_R(X_ju)=\mathrm{Ann}_R(u)=P_\alpha `$. This shows that we may suppose $`\alpha _j^{}=0`$ for every $`jF`$, so that $`\alpha ^{}=\alpha `$.
The sufficiency of the condition in part (b) follows directly from part (a) and Corollary 3.1. For the converse, it is enough to notice that if for some $`G\{1,\mathrm{},n\}`$, there is $`0uH^{i2}(\mathrm{\Delta }_G;k)`$, then there is $`HG`$ such that $`X^Hu`$ corresponds to a nonzero element in $`_{jGH}Ker(H^{i2}(\mathrm{\Delta }_{GH};k)H^{i2}(\mathrm{\Delta }_{G(Hj)};k))`$.
###### Example 1
Let $`R=k[a,b,c,d]`$ and $`B=(ab,bc,cd,ad,ac)`$. Then $`\mathrm{\Delta }`$ is the simplicial complex:
Theorem 3.4(a) gives easily that
$$\mathrm{Ass}(\mathrm{Ext}_R^3(R/B,R))=\{(a,b,d),(b,c,d)\}.$$
###### Example 2
In general, it is not sufficient for $`\beta _{|\alpha |i,\alpha }(B^{})`$ to be nonzero in order to have $`P_\alpha \mathrm{Ass}(\mathrm{Ext}_R^i(R/B,R))`$.
Let’s consider $`R=k[a,b,c]`$ and $`B=(a,bc)`$. Then $`\mathrm{\Delta }`$ is the simplicial complex:
Using Theorem 3.4(a), we get:
$$\mathrm{Ass}(\mathrm{Ext}_R^2(R/B,R))=\{(a,b),(a,c)\},$$
while
$$\{F|\beta _{|\alpha _F|2,\alpha _F}(B^{})0\}=\{\{a,b,c\},\{a,b\},\{a,c\}\}.$$
References
D.Bayer, H.Charalambous, and S.Popescu (1998). Extremal Betti numbers and applications to monomial ideals, preprint.
W.Bruns and J.Herzog (1993). Cohen Macaulay rings, Cambridge Univ. Press.
J.Eagon and V.Reiner (1996). Resolutions of Stanley-Reisner rings and Alexander duality, preprint.
D.Eisenbud (1995). Commutative Algebra with a View Toward Algebraic Geometry, Springer.
D.Eisenbud, M.Mustaţǎ and M.Stillman (1998). Cohomology on toric varieties and local cohomology with monomial support, preprint.
R.Godement (1958). Topologie algebrique et theorie des faisceaux, Paris, Herman.
M.Hochster (1977). Cohen-Macaulay rings, combinatorics and simplicial complexes, in Ring theory II, B.R.McDonald, R.A.Morris (eds), Lecture Notes in Pure and Appl. Math., 26, M.Dekker.
J.R.Munkres (1984). Elements of algebraic topology, Benjamin/Cummings, Menlo Park CA.
M.Mustaţǎ (1999). Vanishing theorems on toric varieties, in preparation.
N.Terai (1997). Generalization of Eagon-Reiner theorem and h-vectors of graded rings, preprint.
N.Terai (1998). Local cohomology with respect to monomial ideals, in preparation.
W.V.Vasconcelos (1998). Computational methods in commutative algebra and algebraic geometry, Algorithms and Computation in Mathematics, vol.2, Springer–Verlag.
U.Walther (1999). Algorithmic computation of local cohomology modules and the cohomological dimension of algebraic varieties, Journal for Pure and Applied Algebra, to appear.
K.Yanagawa (1998). Alexander duality for Stanley-Reisner rings and squarefree $`N^n`$ graded modules, preprint.
Author Adress:
Mircea Mustata
Department of Mathematics, Univ. of California, Berkeley; Berkeley CA 94720
mustata@math.berkeley.edu Institute of Mathematics of the Romanian Academy, Calea Grivitei 21, Bucharest, Romania
mustata@stoilow.imar.ro |
no-problem/0001/hep-ex0001053.html | ar5iv | text | # A Prototype RICH Detector Using Multi-Anode Photo Multiplier Tubes and Hybrid Photo-Diodes
## 1 Introduction
This paper reports results from a prototype Ring Imaging Cherenkov (RICH) counter and compares the performance of Multi- Anode Photomultiplier tubes (MAPMT) and two types of Hybrid Photo-diode Detectors (HPD) for detecting the Cherenkov photons. The experimental arrangement represents a prototype of the downstream RICH detector of the LHCb experiment at CERN.
The LHCb experiment will make precision measurements of CP asymmetries in B decays. Particle identification by the RICH detectors is an important tool and an essential component of LHCb. For example, separating pions and kaons using the RICH suppresses backgrounds coming from $`B_d^0K^+\pi ^{}`$, $`B_s^0K^+\pi ^{}`$ and $`B_s^0K^+K^{}`$ when selecting $`B_d^0\pi ^+\pi ^{}`$ decays, and backgrounds coming from $`B_sD_s^\pm \pi ^{}`$ when selecting the $`B_sD_s^\pm K^{}`$ decay mode.
LHCb has two RICH detectors. Together they cover polar angles from 10 to 330 mrad. The upstream detector, RICH1, uses aerogel and $`C_4F_{10}`$ radiators to identify particles with momenta from 1 to 65 GeV/c. The downstream detector, RICH2, has 180 cm of $`CF_4`$ radiator and identifies particles with momenta up to 150 GeV/c. It uses a spherical focusing mirror with a radius of curvature of 820 cm which is tilted by 370 mrad to bring the image out of the acceptance of the spectrometer. A flat mirror then reflects this image onto the photodetector plane. For tracks with $`\beta 1`$, RICH2 is expected to detect about 30 photoelectrons .
The LHCb collaboration intends to use arrays of photodetectors with a sensitive granularity of $`2.5\mathrm{mm}\times 2.5\mathrm{mm}`$ covering an area of $`2.9\mathrm{m}^2`$ with a total of 340,000 channels, to detect the Cherenkov photons in both RICH detectors. These photodetectors are expected to cover an active area of at least 70% of the detector plane. Current commercially available devices<sup>1</sup><sup>1</sup>1 Commercial HPD devices from Delft Electronische Producten (DEP), The Netherlands, Commercial MAPMT devices from Hamamatsu Photonics, Japan. have inadequate coverage of the active area and their performance at LHC speeds remains to be proven. The beam tests described here used prototypes of three of the new photodetector designs that have been proposed for LHCb.
The results from the LHCb RICH1 prototype detector tests carried out during 1997 are reported in an accompanying publication . The data used in this paper were collected during the summer and autumn of 1998 at the CERN SPS facility. The main goals of these RICH2 prototype studies are:
* To test the performance of the $`CF_4`$ radiator, using the full-scale optical layout of RICH2,
* To test the performance of the photodetectors using the RICH2 geometry by measuring the Cherenkov angle resolution and photoelectron yields.
Section 2 of this paper describes the main features of the test beam setup. Section 3 describes the simulation of the experiment and is followed by a discussion of the photoelectron yields and Cherenkov angle resolution measurements for each of the photodetectors. Finally a summary is given in Section 6, with plans for future work.
## 2 Experimental Setup
The setup included scintillators and a silicon telescope which defined and measured the direction of incident charged particles, a radiator for the production of Cherenkov photons, a mirror for focusing these photons, photodetectors and the data acquisition system. A brief description of these components is given below, and a more complete description of the experimental setup can be found in . The photodetectors were mounted on a plate customised for particular detector configurations. A schematic diagram of the setup is shown in Figure 1.
### 2.1 Beam line
The experimental setup was mounted in the CERN X7 beam line. The beam was tuned to provide negative particles (mainly pions) with momenta between 10 and 120 GeV/c. The precision of the beam momentum for a given setting ($`\delta `$p/p) was better than 1%. Readout of the detectors was triggered by the passage of beam particles which produced time-correlated signals from two pairs of scintillation counters placed 8 metres apart along the beam line. The beam size was $`20\times 20\mathrm{m}\mathrm{m}^2`$ as defined by the smaller of these counters.
### 2.2 Beam Trajectory Measurement
The input beam direction and position were measured using a silicon telescope consisting of three planes of pixel detectors. Each of these planes has a $`22\times 22`$ array of silicon pixels with dimensions $`1.3\mathrm{mm}\times 1.3\mathrm{mm}`$. Two of the planes were placed upstream of the radiator and the third one downstream of the mirror. The first and third planes were separated by 8 metres.
Using the silicon telescope, the beam divergence was measured to be typically 0.3 mrad and 0.1 mrad in the horizontal and vertical planes respectively.
### 2.3 The RICH Detector
During different data-taking periods, air and $`CF_4`$ were used as radiators. The pressure and temperature of these radiators were monitored for correcting the refractive index . The gas circulation system which provided the $`CF_4`$ is described below.
During the $`CF_4`$ runs, data were taken at various pressures ranging from 865 mbar to 1015 mbar and at different temperatures between $`20^0C`$ and $`30^0C`$. The refractive index of $`CF_4`$ as a function of wavelength at STP using the parametrization in is plotted in Figure 2.
As shown in the schematic diagram in Figure 3, the prototype Cherenkov vessel was connected into the gas circulation system, which was supplied by $`CF_4`$ gas <sup>2</sup><sup>2</sup>2 as supplied by CERN stores: reference SCEM 60.56.10.100.7 at high pressure. A molecular sieve (13X pore size) was included in the circuit to remove water vapour. The system used a microprocessor interface <sup>3</sup><sup>3</sup>3Siemens S595U to set and stabilise the required gas pressure and to monitor and record pressure, temperature and concentrations of water vapour and oxygen throughout the data taking. The absolute pressure of the $`CF_4`$ in the Cherenkov vessel was maintained to within 1 mbar of the required value using electromagnetic valves which controlled the gas input flow and the output flow to the vent. Throughout the data taking the oxygen concentration was below 0.1$`\%`$ and the water vapour concentration was below 100 ppm by volume.
The Cherenkov photons emitted were reflected by a mirror of focal length 4003 mm which was tilted with respect to the beam axis by 314 mrad, similar to the optical layout of the LHCb RICH2. Using micrometer screws, the angle of tilt of the mirror was adjusted to reflect photons on different regions of the photodetector plane which was located 4003 mm from the mirror. The reflectivity of the mirror, measured as a function of the wavelength, is shown in Figure 4.
The important characteristics of the three different designs of photodetectors tested are briefly summarised as follows:
* The 61-pixel Hybrid Photo-Diode (HPD) is manufactured by DEP and has an S20 (trialkali) photocathode deposited on a quartz window. The quantum efficiency of a typical HPD measured by DEP, is plotted in Figure 5 as a function of the incoming photon wavelength. Photoelectrons are accelerated through a 12 kV potential over 12 mm onto a 61-pixel silicon detector. The image on the photocathode is magnified by 1.06 on the silicon detector surface. This device gives an approximate gain of 3000. The pixels are hexagonally close packed and measure 2 mm between their parallel edges. The signal is read out by a Viking VA2 ASIC.
* The 2048-pixel HPD is manufactured in collaboration with DEP. It has electrostatic cross-focusing by which the image on the photocathode is demagnified by a factor of four at the anode. The operating voltage of this HPD is 20 kV. The anode has a silicon detector, which provides an approximate gain of 5000, with an array of 2048 silicon pixels bump bonded to an LHC1 binary readout ASIC. Details of this device and its readout can be found in .
Using the measurements made by DEP, the quantum efficiency of the S20 photocathode used on the 2048-pixel HPD is plotted in Figure 5 as a function of the photon wavelength. This tube has an active input window diameter of 40 mm and the silicon pixels are rectangles of size 0.05 mm $`\times `$ 0.5 mm. It represents a half-scale prototype of a final tube which will have an 80 mm diameter input window and 1024 square pixels with 0.5 mm side.
* The 64-channel Multi-Anode PMT (MAPMT) is manufactured by Hamamatsu. It has a bialkali photocathode deposited on a borosilicate-glass window and 64 square anodes mounted in an 8 $`\times `$ 8 array with a pitch of 2.3 mm. The photoelectrons are multiplied using a 12-stage dynode chain resulting in an approximate overall gain of $`10^6`$ when operated at 900 V. From the measurements made by Hamamatsu, the quantum efficiency of a typical MAPMT as a function of the wavelength is shown in Figure 5.
During some runs, pyrex filters were placed in front of the photodetectors in order to limit the transmission to longer wavelengths where the refractive index of the radiators is almost constant. In Figure 6 the transmission of pyrex as a function of photon wavelength is plotted.
### 2.4 Experimental Configurations
The detector configurations used are summarised in Table 1. In configuration 1, seven 61-pixel HPDs and one MAPMT were placed on a ring of radius 113 mm on the detector plate. In configurations 2 and 3, a 2048-pixel HPD and three 61-pixel HPDs were placed on a ring of radius 90 mm on the detector plate. In addition to these configurations, the different radiator, beam and photodetector conditions used for the various runs are shown in Table 2.
### 2.5 Data Acquisition System
The 61-pixel HPDs and the MAPMT use analogue readout whereas the 2048-pixel HPD uses binary readout. A detailed description of their respective data acquisition systems can be found in and .
For the analogue readout system, the mean and width of the pedestal distributions for each pixel were calculated using dedicated pedestal runs, interleaved between data runs triggered with beam. Some data were also taken using light emitted from a pulsed Light Emitting Diode (LED) for detailed studies of the photoelectron spectra. Zero suppression was not used on analogue data from the photodetectors.
A pixel threshold map was established on the 2048-pixel HPD using an LED . For this, the high voltage applied on the tube was varied, and the voltage for each channel to become active was recorded. This threshold map was used to identify pixels with too low a threshold, which were then masked. It was also used to identify pixels with too high a threshold and hence insensitive to photoelectrons. A histogram of the threshold map is shown in Figure 7 where the pixels which were masked or insensitive (26$`\%`$) are indicated by the entries in the first bin. For this device, the noise ($`\sigma _N`$) of the readout electronics is 160 electrons (0.6 kV Silicon equivalent) and the distribution of the silicon pixel thresholds has an rms width of 1.6 kV.
In Figure 8 an online display, integrating all events in a run, with seven 61-pixel HPDs and an MAPMT in configuration 1 is shown. Part of the Cherenkov ring falls on the photodetectors and is clearly visible.
## 3 Simulation of RICH2 prototype
The RICH2 prototype configurations are simulated to allow detailed comparisons of expected performance with that found in data. The simulation program generates photons uniformly in energy and with the corresponding Cherenkov angle. The trajectories of these photons, and the photoelectrons they produce, are simulated using the beam divergence, beam composition and the optical characteristics of the various components of the RICH detector shown in Figures 2 to 6. The air radiator is simulated using a gas mixture consisting of 80% Nitrogen and 20% Oxygen.
The program also simulates the response of the various photodetectors. Since the 2048-pixel HPD used binary readout, to study its response the program simulates the threshold map ( Figure 7) used for this readout. The simulation of the response of the silicon detector of this HPD is described in Section 4.1
## 4 Estimates of Photoelectron yield
The average number of photoelectrons detected per event in a photodetector defines the photoelectron yield for that detector. This is determined for the configurations 1 and 2 indicated in Table 1. Since the 61-pixel HPD and the MAPMT use analogue readout, the distinction between signal and background depends upon the threshold above the pedestal peak assigned to the measured photoelectron spectrum. To get the true photoelectron yield at a given threshold, estimates are made for the level of background present and for the amount of signal loss that occurs as a result of applying the threshold cut, specified in terms of the width ($`\sigma `$) of the pedestal spectrum.
In the two types of HPDs, there is 18$`\%`$ probability at normal incidence, for electrons to backscatter at the silicon surface, causing some loss of signal. In the 61-pixel HPD, the backscattered electrons can “bounce“ off the silicon surface more than once, whereas in the 2048-pixel HPD the electric field is such that they do not return to the silicon detector. Passage through the dead layers of the silicon wafer can also cause a small amount of signal loss in the HPDs. Since the 2048-pixel HPD uses binary readout, its photoelectron yield depends mainly upon the threshold map of the readout system.
From the estimate of the photoelectron yield ($`N_{pe}`$) of a photodetector, the figure of merit ($`N_0`$) is calculated using:
$`N_0=N_{pe}/(ϵ_AL\mathrm{sin}^2\theta _c)`$ where $`ϵ_A`$ is the fraction of the Cherenkov ring covered by the photodetector, L is the length of the radiator and $`\theta _c`$ is the mean Cherenkov angle measured using the method described in Section 5.
### 4.1 Photoelectron yield for the 2048-pixel HPD
The response of the silicon detector of this HPD is simulated as follows:
Each photoelectron is accelerated through a potential of 20 kV towards the silicon surface. The probability for backscattering at the silicon surface is 18 $`\%`$. During the backscattering process, only a fraction of the 20 keV energy is released in the silicon detector. For an energy release varying from 5 to 20 keV, the energy loss in the dead layer of the silicon ranges from 5 to 1.2 keV as described in and references therein. A readout channel is expected to fire only when the charge signal generated in the silicon detector exceeds the corresponding pixel threshold by at least 4 times the electronic noise.
A flat background of 0.01 photoelectrons per event is observed in the real data on the detector surface from beam related sources such as photons and photoelectrons reflected in random directions from different surfaces in the prototype. This is also incorporated into the simulation. The resultant photoelectron yield from the simulation in the presence of a pyrex filter is shown in Figure 9(a), and in the absence of any filter is shown in Figure 9(b).
The systematic error in the photoelectron yield is evaluated from the simulation by varying the parameters which are listed below. The result of these variations are tabulated in Table 3.
* Quantum efficiency of the phototube: The quantum efficiency of the 2048-pixel HPD is found to be approximately half that of the 61-pixel HPD. The simulation is repeated by replacing the quantum efficiency of the 2048-pixel HPD with those from the 61-pixel HPD, scaled down by a factor of two.
* Amount of photon absorption in oxygen: The simulation is repeated with and without activating the photon absorption although this is significant only for wavelengths below 195 nm.
* Wavelength cutoff of the photocathode: To account for any variation in the active wavelength range among different versions of the photocathodes, the simulation is repeated by varying lower cutoff between 190 nm and 200 nm, and the upper cutoff between 600 nm and 900 nm.
* Backscattering probability at the silicon surface: The simulation is repeated by varying the backscattering probability between 16$`\%`$ and 20 $`\%`$.
The simulated photoelectron yield per detector in the case without any filter is 0.46 $`\pm `$ 0.07, whereas in real data the yield is 0.49 (Figure 9 (b)). The simulated yield per detector, for the case with the pyrex filter, is 0.18 $`\pm `$ 0.02 and the corresponding yield in real data is 0.15 (Figure 9 (a)). Using these yields, the figure of merit is estimated to be 97 $`\pm `$ 16 $`cm^1`$ in the case without any filter and 30 $`\pm `$ 5 $`cm^1`$ in the case with the pyrex filter. For the case without any filter, an independent determination of the figure of merit for the same tube, agrees with the present estimate.
### 4.2 Photoelectron yield for the 61-pixel HPD
Figure 10 shows a typical photoelectron spectrum obtained from a single pixel in a 61-pixel HPD. The peaks corresponding to the pedestal and signal can be clearly seen. In similar distributions obtained for each of the pixels, the background contamination in the photoelectron yield and the amount of signal lost are estimated as a function of the threshold cut using two different analysis methods. One of these methods is described below and the other one is described in Section 4.3 where similar estimates are made for the MAPMT.
The signal loss is estimated using data where the signals were provided by photons from an LED as only these runs have adequate statistics for this purpose. The signal loss is considered to have a Gaussian component and a backscattering component which are described below.
An example of the spectra for each detector pixel in LED data is shown in Figure 11. It can be divided mainly into three parts identified as distributions for the pedestal, one photoelectron and two photoelectrons, in addition to two underlying distributions corresponding to the backscattering contributions to the single and double photoelectron spectra. In order to estimate these backscattering contributions, a backscattering probability of 18$`\%`$ is assumed. The energy distribution of the backscattered electrons is made by convoluting the distribution of the energy fraction of the backscattered electrons for 10 keV electrons incident on aluminium, obtained from , with a Gaussian that has the same width as that of the pedestal spectrum in LED data.
The adc spectrum in LED data is fitted with a function that modelled the spectrum as a sum of three Gaussians with contributions from two backscattering components. The three Gaussians correspond to the distributions of pedestal, one photoelectron and two photoelectrons. The result of the fit is superimposed over the adc spectrum in Figure 11. The widths of the Gaussians for the photoelectrons are then corrected to account for the slight difference in the widths of the pedestal observed in LED data and Cherenkov photon data.
In the region below the threshold cut, the sum of the area which is under the one photoelectron Gaussian and the corresponding backscattering component is then taken as the sum of the Gaussian and backscattering components of the signal loss.
This procedure is repeated using a different LED run and varying the backscattering probability between 16$`\%`$ and 20$`\%`$. The resultant variations obtained in the signal loss estimate are taken as contributions to systematic error from this method.
At the threshold cut of 3$`\sigma `$, the Gaussian component of the signal loss is 0.9$`\%`$ whereas the backscattering component is 11.2$`\%`$.
The background remaining in the Cherenkov photoelectron spectrum after a given threshold cut is considered to have a Gaussian component due to electronic noise, and a non-Gaussian component induced by detector noise and photons from extraneous sources. For the first component, a single Gaussian is fit to the pedestal part of this spectrum. The area under this fit spectrum above the threshold cut is then taken as the Gaussian component of the background. This procedure was repeated changing the upper range of the Gaussian fit from 1.2$`\sigma `$ to 2$`\sigma `$ and the resultant variation in the background estimate is taken as a contribution to the systematic error.
In order to evaluate the second component, data from pedestal runs are used. The fraction of the spectrum above the threshold cut, after removing the fit single Gaussian to the pedestal spectrum, is taken as the non-Gaussian component. The variation in this estimate obtained using different pedestal runs is taken as a contribution to the systematic error.
After correcting the distribution of the number of photoelectrons in each pixel for background and signal loss, their spatial distribution on the silicon surface is fitted with a function which assumes the Cherenkov angle distribution to be a Gaussian. A residual flat background observed in this fit is considered as beam related background and is subtracted from the photoelectron signal. The fit is repeated by varying the parameters of the function and the resultant variations in the background estimate is taken as a contribution to the systematic error.
The results obtained for the photoelectron multiplicities after correcting for background and signal loss using the above method are reported below. These are in agreement with the results obtained from the alternative method described in the next section.
In these estimates, the statistical error is found to be negligible compared to the overall systematic error which is obtained by adding the various contributions in quadrature. The contributions to the systematic error are shown in Table 4. In Table 5, the corrected photoelectron yields for the data with pyrex filter and with no filter are shown along with the corresponding expectations from simulation. The yields from data and simulation agree.
As a systematic check, the stability of the corrected photoelectron yields obtained by varying the threshold cut from 2$`\sigma `$ to 5$`\sigma `$ for the data with pyrex filter, is shown in Table 6. The small variation seen in the yields between 3$`\sigma `$ and 4$`\sigma `$ is quoted as a systematic error contribution in Table 4. The fact that the corrected photoelectron yields estimated are independent of the threshold cut and that the two analysis methods yield similar results give confidence in the results shown in Table 5.
Using the yield estimates in Table 5, the figure of merit is estimated to be 89 $`\pm `$ 8 $`cm^1`$ in the case with pyrex filter and 258 $`\pm `$ 24 $`cm^1`$ in the case without any filter.
### 4.3 Photoelectron yield for MAPMT
Figure 12 shows a typical pulse height distribution for a pixel in the MAPMT in beam triggered runs. The photoelectron signal and pedestal peaks can be clearly distinguished. The amount of signal lost and the amount of background contamination to the photoelectron yield are estimated using the method described below.
This method also uses data where the photons from an LED provided signals to the MAPMT. A Gaussian is fit to the pedestal part of the pulse height distribution. The contribution of the pedestal is removed, and in the remaining spectrum that part below the threshold cut is taken to be the signal loss. The contributions to the systematic error in this estimate are listed below:
* The change in signal loss obtained by swapping the width of the pedestal in Cherenkov photon data with that from LED data, is taken as a contribution to the systematic error.
* In the Cherenkov photon data and LED data, the ranges of the fits to the pedestals are varied and any resultant change in the signal loss is taken as the contribution to the systematic error.
In order to estimate the background level, data from a special run are used where the pressure in the $`CF_4`$ radiator was reduced such that the Cherenkov ring passed through a different set of pixels than in the other runs. In these data, the photoelectron yield is estimated after applying the threshold cut to the spectrum from the pixels which are selected to be off the Cherenkov ring. Assuming a uniform background across the MAPMT, this yield is taken as the background contribution. This procedure is repeated by varying the set of pixels which are selected for this estimate and the resultant change in the background estimate is taken as contribution to the systematic error.
These estimates for the background level and signal loss are repeated for different threshold cuts in the spectra with the results given in Table 7. The photoelectron yields resulting from these estimates are independent of the threshold cuts applied. The systematic error in this measurement is estimated in the same way as for the 61-pixel HPD described in the previous section. Above a threshold cut of 3 $`\sigma `$, the yield after the corrections is estimated to be 0.48 $`\pm `$ 0.03. The corresponding expectation from simulation is 0.52. The discrepancy between data and simulation is attributed to the uncertainty in the knowledge of the quantum efficiency of the particular MAPMT used in these tests. Using this yield estimate, the figure of merit is estimated to be 155 $`\pm `$ 13 $`cm^1`$.
## 5 Resolution of the Reconstructed Cherenkov Angle
As described in , the reconstruction of the Cherenkov angle requires the coordinates of the hit on the photodetector, the centre of curvature of the mirror and the photon emission point (E) which is assumed to be the middle point of the track in the radiator. The point (M) where the photons are reflected off the mirror, is reconstructed using the fact that it lies in the plane defined by the aforementioned three points. The reconstructed Cherenkov angle is the angle between the beam direction and the line joining E and M.
Figures 13(a),(b) show the Cherenkov angle distribution obtained using air radiator and 100 GeV/c pions for the 2048-pixel HPD and a 61-pixel HPD which were diametrically opposite to each other on the detector plate in configuration 2 with pyrex filter. The 2048-pixel HPD has a better resolution than the 61-pixel HPD since the pixel granularity is 0.2 mm for the former and 2 mm for the latter. Figure 13(c) shows the Cherenkov angle distribution obtained using $`CF_4`$ radiator and 120 GeV/c pions for an MAPMT with 2.3 mm pixel granularity in configuration 1.
### 5.1 Sources of Uncertainty in the Cherenkov Angle Measurement
* Chromatic Error: This is due to the variation of refractive index of the radiator with wavelength and is largest in the UV region. Use of pyrex filters reduces this contribution.
* Emission point uncertainty: This comes from the fact that the mirror is tilted with respect to the beam axis and that the emission point is assumed to be in the middle of the radiator, regardless of the true but unknown point of emission.
* Pixel size of Photodetector.
* Measurement of beam trajectory: This contribution comes from the granularity of the pixels in the silicon detectors which are used to measure the direction of the incident beam particle.
* Alignment: This contribution comes from residual misalignments between the silicon telescope, the mirror and the photodetectors.
In Table 8 the resolutions from each of the above components are tabulated for each of the three photodetectors in typical configurations. In each case, the overall simulated resolution is in good agreement with that measured in the beam triggered data.
In configuration 1 with seven HPDs it was possible to perform a detailed investigation of the Cherenkov angle resolution. Figure 14(a) shows the resolution measured in data and from simulation for each of the seven 61-pixel HPDs in this configuration. Agreement is seen between data and simulation in all cases. Each HPD in this figure was located at a different azimuth on the detector plate and hence has a different emission point uncertainty. Hence the overall resolution for different HPDs are different. Figure 14(b) shows the same resolutions, for the data using the pyrex filter, which reduces the contribution from chromatic error.
The expectation from the LHCb Technical proposal is to have a resolution of 0.35 mrad which is already achieved for the MAPMT, the 2048-pixel HPD and some of the HPDs shown in Figure 14.
### 5.2 Multiphoton Resolution
The mean value of the Cherenkov angle from all the photoelectron hits in each event is calculated for the data from the seven 61-pixel HPDs in configuration 1 without pyrex filters. The width of this distribution versus the number of photoelectrons detected per trigger is plotted in Figure 15. For a perfectly aligned system, the width is expected to be inversely proportional to the square root of the number of photoelectrons as indicated by the curve. The disagreement between data and simulation is compatible with the residual misalignment in the system which is of the order of 0.1 mrad.
### 5.3 Particle Identification
Figure 16 shows the Cherenkov angle distribution for the 2048-pixel HPD without pyrex filter in configuration 3 where the beam used was a mixture of pions and electrons at 10.4 GeV/c. Good separation is obtained between the two particle types. Figure 17 shows the plot of the the mean Cherenkov angle calculated from the hits in the 61-pixel HPDs without pyrex filter in configuration 1, where the beam was a mixture of kaons and pions, approximately in the ratio 1:9, at 50 GeV/c. Peaks corresponding to the two charged particle types can be seen in this figure.
## 6 Summary and Outlook for the Future
The goals set for the RICH2 prototype tests have largely been accomplished. The performance of the $`CF_4`$ radiator and the optical layout of the RICH2 detector have been tested. Photoelectron yields from the prototype HPDs and MAPMTs have been measured and found to agree with simulations. A Cherenkov angle precision of 0.35 mrad as assumed in the LHCb technical proposal has been demonstrated with all three photodetectors.
Improvements in the integrated quantum efficiency of both HPDs and MAPMTs are expected in future devices. The LHCb RICH detector will require photodetectors with higher active to total area than those tested here. HPDs with 80% active area and a lens system for MAPMTs are currently being developed. These will be tested with LHC compatible readout (25 ns shaping time) during 1999-2000.
## 7 Acknowledgements
This work has benefited greatly from the technical support provided by our colleagues at the institutes participating in this project. In particular the mirror reflectivity and the pyrex transmission were measured by A. Braem. The radiator vessel extensions were manufactured by D. Clark and I. Clark. The printed circuits for the MAPMT were designed and assembled by S. Greenwood. The silicon telescope was provided by E. Chesi and J. Seguniot. We also received valuable advice and assistance from our colleagues in the LHCb collaboration, in particular from R. Forty, O. Ullaland and T. Ypsilantis.
Finally, we gratefully acknowledge the CERN PS division for providing the test beam facilities and the UK Particle Physics and Astronomy Research Council for the financial support. |
no-problem/0001/quant-ph0001012.html | ar5iv | text | # Evidence for a dynamic origin of charge
## I The nature of charge
Since the discovery of the electron by J. J. Thomson the concept of electric charge has remained nearly unchanged. Apart from Lorentz’ extended electron , or Abraham’s electromagnetic electron , the charge of an electron remained a point like entity, in one way or another related to electron mass . In atomic nuclei we think of charge as a smeared out region of space, which is structured by the elementary constituents of nuclear particles, the quarks .
The first major modification in this picture occurred only in the last decades, when experiments on the quantum hall effect suggested the existence of ”fractional charge” of electrons. Although this effect has later been explained on the basis of standard theory , its implications are worth a more thorough analysis. Because it cannot be excluded that the same feature, fractional or even continuous charge, will show up in other experiments, especially since experimental practice more and more focuses on the properties of single particles. And in this case the conventional picture, which is based on discrete and unchangeable charge of particles, may soon prove too narrow a frame of reference. It seems therefore justified, at this point, to analyze the very nature of charge itself. A nature, which would reveal itself as an answer to the question: What is charge?
It must be noted, in this respect, that the picture of continuous charge, in classical theories, is due to the omission of the atomic structure of matter. In any modern sense, continuous charge can only be recovered by considering dynamic processes within the very particles themselves.
With this problem in mind, we reanalyze the fundamental equations of intrinsic particle properties . The consequences of this analysis are developed in two directions. First, we determine the interface between mechanic and electromagnetic properties of matter, where we find that only one fundamental constant describes it: Planck’s constant $`\mathrm{}`$. And second, we compute the fields of interaction within a hydrogen atom, where we detect oscillations of the proton density of mass as their source. Finally, the implications of our results in view of unifying gravity and quantum theory are discussed and a new model of gravity waves derived, which is open to experimental tests.
## II The origin of dynamic charge
The intrinsic vector field $`𝐄(𝐫,t)`$, the momentum density $`𝐩(𝐫,t)`$, and the scalar field $`\varphi (𝐫,t)`$ of a particle are described by (see , Eq. (18)):
$`𝐄(𝐫,t)={\displaystyle \frac{1}{\overline{\sigma }}}\varphi (𝐫,t)+{\displaystyle \frac{1}{\overline{\sigma }}}{\displaystyle \frac{}{t}}𝐩(𝐫,t)`$ (1)
Here $`\overline{\sigma }`$ is a dimensional constant introduced for reasons of consistency. Rewriting the equation with the help of the definitions:
$`\beta :={\displaystyle \frac{1}{\overline{\sigma }}}\beta \varphi (𝐫,t):=\varphi (𝐫,t)`$ (2)
we obtain the classical equation for the electric field, where in place of a vector potential $`𝐀(𝐫,t)`$ we have the momentum density $`𝐩(𝐫,t)`$. This similarity, as already noticed, bears on the Lorentz gauge as an expression of the energy principle ( Eqs. (26) - (28)).
$`𝐄(𝐫,t)=\varphi (𝐫,t)+\beta {\displaystyle \frac{}{t}}𝐩(𝐫,t)`$ (3)
Note that $`\beta `$ describes the interface between dynamic and electromagnetic properties of the particle. Taking the gradient of (3) and using the continuity equation for $`𝐩(𝐫,t)`$:
$`𝐩(𝐫,t)+{\displaystyle \frac{}{t}}\rho (𝐫,t)=0`$ (4)
where $`\rho (𝐫,t)`$ is the density of mass, we get the Poisson equation with an additional term. And if we include the source equation for the electric field $`𝐄(𝐫,t)`$:
$`𝐄(𝐫,t)=\sigma (𝐫,t),`$ (5)
$`\sigma (𝐫,t)`$ being the density of charge, $`ϵ`$ set to 1 for convenience, we end up with the modified Poisson equation:
$`\mathrm{\Delta }\varphi (𝐫,t)=\underset{staticcharge}{\underset{}{\sigma (𝐫,t)}}\underset{dynamiccharge}{\underset{}{\beta {\displaystyle \frac{^2}{t^2}}\rho (𝐫,t)}}`$ (6)
The first term in (6) is the classical term in electrostatics. The second term does not have a classical analogue, it is an essentially novel source of the scalar field $`\varphi `$, its novelty is due to the fact, that no dynamic interpretation of the vector potential $`𝐀(𝐫,t)`$ exists, whereas, in the current framework, $`𝐩(𝐫,t)`$ has a dynamic meaning: that of momentum density.
To appreciate the importance of the new term, think of an aggregation of mass in a state of oscillation. In this case the second derivative of $`\rho `$ is a periodic function, which is, by virtue of Eq. (6), equal to periodic charge. Then this dynamic charge gives rise to a periodic scalar field $`\varphi `$. This field appears as a field of charge in periodic oscillations: hence its name, dynamic charge. It should be noted that dynamic charge is essentially different from a classical dipole: in that case the field can appear zero (cancellation of opposing effects), whereas in case of dynamic charge it is zero. Even, as shall be seen presently, for monopole oscillations.
## III Oscillations of a proton
We demonstrate the implications of Eq. (6) on an easy example: the radial oscillations of a proton. The treatment is confined to monopole oscillations, although the results can easily be generalized to any multipole. Let a proton’s radius be a function of time, so that $`r_p=r_p(t)`$ will be:
$`r_p(t)=R_p+d\mathrm{sin}\omega _Ht`$ (7)
Here $`R_p`$ is the original radius, $`d`$ the oscillation amplitude, and $`\omega _H`$ its frequency. Then the volume of the proton $`V_p`$ and, consequently, its density of mass $`\rho _p`$ depend on time. In first order approximation we get:
$`\rho _p(t)`$ $`=`$ $`{\displaystyle \frac{3M_p}{4\pi }}\left(R_p+d\mathrm{sin}\omega _Ht\right)^3`$ (8)
$`\rho _p(t)`$ $``$ $`\rho _0\left(1x\mathrm{sin}\omega _Ht\right)x:={\displaystyle \frac{3d}{R_p}}`$ (9)
The Poisson equation for the dynamic contribution to proton charge then reads:
$`\mathrm{\Delta }\varphi (𝐫,t)=\beta x\rho _0\omega _H^2\mathrm{sin}\omega _Ht`$ (10)
Integrating over the volume of the proton we find for the dynamic charge of the oscillating proton the expression:
$`q_D(t)={\displaystyle _{V_p}}d^3r\beta x\rho _0\omega _H^2\mathrm{sin}\omega _Ht=\beta xM_p\omega _H^2\mathrm{sin}\omega _Ht`$ (11)
This charge gives rise to a periodic field within the hydrogen atom, as already analyzed in some detail and in a slightly different context . We shall turn to the calculation of a hydrogen’s fields of interaction in the following sections. But in order to fully appreciate the meaning of the dynamic aspect it is necessary to digress at this point and to turn to the discussion of electromagnetic units.
## IV Natural electromagnetic units
By virtue of the Poisson equation (6) dynamic charge must be dimensionally equal to static charge, which for a proton is + e. But since it is, in the current framework, based on dynamic variables, the choice of $`\beta `$ also defines the interface between dynamic and electromagnetic units. From (11) we get, dimensionally:
$`[e]=[\beta ][M_p\omega _H^2][\beta ]=\left[{\displaystyle \frac{e}{M_p\omega _H^2}}\right]`$ (12)
The unit of $`\beta `$ is therefore, in SI units:
$`[\beta ]=C{\displaystyle \frac{s^2}{kg}}=C{\displaystyle \frac{m^2}{J}}[SI]`$ (13)
We define now the natural system of electromagnetic units by setting $`\beta `$ equal to 1. Thus:
$`[\beta ]:=1[C]={\displaystyle \frac{J}{m^2}}`$ (14)
The unit of charge C is then energy per unit area of a surface. Why, it could be asked, should this definition make sense? Because, would be the answer, it is the only suitable definition, if electrostatic interactions are accomplished by photons.
Suppose a $`\delta ^3(𝐫𝐫^{})`$ like region around $`𝐫^{}`$ is the origin of photons interacting with another $`\delta ^3(𝐫𝐫^{\prime \prime })`$ like region around $`𝐫^{\prime \prime }`$. Then $`𝐫^{}`$ is the location of charge. Due to the geometry of the problem the interaction energy will decrease with the square of $`|𝐫^{}𝐫^{\prime \prime }|`$. What remains constant, and thus characterizes the charge at $`𝐫^{}`$, is only the interaction energy per surface unit. Thus the definition, which applies to all $`r^2`$ like interactions, also, in principle, to gravity.
Returning to the question of natural units, we find that all the other electromagnetic units follow straightforward from the fundamental equations . They are displayed in Table I.
If we analyze the units in Lorentz’ force equation, we observe, at first glance, an inconsistency.
$`𝐅_L=q\left(𝐄+𝐮\times 𝐁\right)`$ (15)
The unit on the left, Newton, is not equal to the unit on the right. As a first step to solve the problem we include the dielectric constant $`ϵ^1`$ in the equation, since this is the conventional definition of the electric field $`𝐄`$. Then we have:
$`[𝐅_L]={\displaystyle \frac{Nm}{m^2}}\left({\displaystyle \frac{m^4}{N}}{\displaystyle \frac{N}{m^3}}+{\displaystyle \frac{m}{s}}{\displaystyle \frac{Ns}{m^4}}\right)=N+N{\displaystyle \frac{N}{m^4}}`$ (16)
Interestingly, now the second term, which describes the magnetic forces, is wrong in the same manner, the first term was before we included the dielectric units. It seems thus, that the dimensional problem can be solved by a constant $`\eta `$, which is dimensionally equal to $`ϵ`$, and by rewriting the force equation (15) in the following manner:
$`𝐅_L={\displaystyle \frac{q}{\eta }}\left(𝐄+𝐮\times 𝐁\right)`$ (17)
$`[\eta ]=Nm^4=Cm^3=[\sigma ]`$ (18)
The modification of (15) has an implicit meaning, which is worth being emphasized. It is common knowledge in special relativity, that electric and magnetic fields are only different aspects of a situation. They are part of a common field tensor $`F_{\mu \nu }`$ and transform into each other by Lorentz transformations. From this point of view the treatment of electric and magnetic fields in the SI, where we end up with two different constants ($`ϵ,\mu `$), seems to go against the requirement of simplicity. On the other hand, the approach in quantum field theory, where one employs in general only a dimensionless constant at the interface to electrodynamics, the finestructure constant $`\alpha `$, is over the mark. Because the information, whether we deal with the electromagnetic or the mechanic aspect of a situation, is lost. The natural system, although not completely free of difficulties, as seen further down, seems a suitable compromise. Different aspects of the intrinsic properties, and which are generally electromagnetic, are not distinguished, no scaling is necessary between $`𝐩,𝐄`$ and $`𝐁`$. The only constant necessary is at the interface to mechanic properties, which is $`\eta `$. This also holds for the fields of radiation, which we can describe by:
$`\varphi _{Rad}(𝐫,t)={\displaystyle \frac{1}{8\pi \eta }}\left(𝐄^2+c^2𝐁^2\right)`$ (19)
Note that in the natural system the usage, or the omission, of $`\eta `$ ultimately determines, whether a variable is to be interpreted as an electromagnetic or a mechanic property. Forces and energies are mechanic, whereas momentum density is not. The numerical value of $`\eta `$ has to be determined by explicit calculations. This will be done in the next sections. We conclude this section by comparing the natural system of electromagnetic units to existing systems.
From an analysis of the Maxwell equations one finds three dimensional constants $`k_1,k_2,k_3`$, and a dimensionless one, $`\alpha `$, which acquire different values in different systems (see Table II).
Judging by the number of dimensional constants it seems that the natural system is most similar to the electrostatic system of units. However, since we have defined a separate interface to mechanic properties, it is free of the usual nuisance of fractional exponents without a clear physical meaning. The other difference is that c, in the esu, is a constant, whereas it only signifies the velocity of a particle in the natural system. For photons this velocity equals c, but for electrons it is generally much smaller. We note in passing that all the fundamental relations for the intrinsic fields remain valid. Also the conventional relations for the forces of interaction and the radiation energy remain functionally the same. Only the numerical values will be different.
Comparing with existing systems we note three distinct advantages: (i) The system reflects the dynamic origin of fields, and it is based on only three fundamental units: m, kg, s. A separate definition of the current is therefore obsolete. (ii) There is a clear cut interface between mechanics (forces, energies), and electrodynamics (fields of motion). (iii) The system provides a common framework for macroscopic and microscopic processes.
## V Interactions in hydrogen
Returning to proton oscillations let us first restate the main differences between a free electron and an electron in a hydrogen atom : (i) The frequency of the hydrogen system is constant $`\omega _H`$, as is the frequency of the electron wave. It is thought to arise from the oscillation properties of a proton. (ii) Due to this feature the wave equation of momentum density $`𝐩(𝐫,t)`$ is not homogeneous, but inhomogeneous:
$`\mathrm{\Delta }𝐩(𝐫,t){\displaystyle \frac{1}{u^2}}{\displaystyle \frac{^2}{t^2}}𝐩(𝐫,t)=𝐟(t)\delta ^3(𝐫)`$ (20)
for a proton at $`𝐫=0`$ of the coordinate system. The source term is related to nuclear oscillations. We do not solve (20) directly, but use the energy principle to simplify the problem. From a free electron it is known that the total intrinsic energy density, the sum of a kinetic component $`\varphi _K`$ and a field component $`\varphi _{EM}`$ is a constant of motion :
$`\varphi _K(𝐫)+\varphi _{EM}(𝐫)=\rho _0u^2`$ (21)
where $`u`$ is the velocity of the electron and $`\rho _0`$ its density amplitude. We adopt this notion of energy conservation also for the hydrogen electron, we only modify it to account for the spherical setup:
$`\varphi _K(𝐫)+\varphi _{EM}(𝐫)={\displaystyle \frac{\rho _0}{r^2}}u^2`$ (22)
The radial velocity of the electron has discrete levels. Due to the boundary values problem at the atomic radius, it depends on the principal quantum number $`n`$. From the treatment of hydrogen we recall for $`u_n`$ and $`\rho _0`$ the results :
$`u_n={\displaystyle \frac{\omega _HR_H}{2\pi n}}\rho _0={\displaystyle \frac{M_e}{2\pi R_H}}`$ (23)
where $`R_H`$ is the radius of the hydrogen atom and $`M_e`$ the mass of an electron. Since $`\rho _0`$ includes the kinetic as well as the field components of electron ”mass”, e.g. in Eq. (22), we can define a momentum density $`𝐩_0(𝐫,t)`$, which equally includes both components. As the velocity $`u_n=u_n(t)`$ of the electron wave in hydrogen is periodic:
$`𝐮_n(t)=u_n\mathrm{cos}\omega _Ht𝐞^r`$ (24)
the momentum density $`𝐩_0(𝐫,t)`$ is given by:
$`𝐩_0(𝐫,t)={\displaystyle \frac{\rho _0u_n}{r^2}}\mathrm{cos}\omega _Ht𝐞^r`$ (25)
The combination of kinetic and field components in the variables has a physical background: it bears on the result that photons change both components of an electron wave . With these definitions we can use the relation between the electric field and the change of momentum, although now this equation refers to both components:
$`𝐄_0(𝐫,t)={\displaystyle \frac{}{t}}𝐩_0(𝐫,t)={\displaystyle \frac{\rho _0u_n}{r^2}}\omega _H\mathrm{sin}\omega _Ht𝐞^r`$ (26)
Note that charge, by definition, is included in the electric field itself. Integrating the dynamic charge of a proton from Eq. (11) and accounting for flow conservation in our spherical setup, the field of a proton will be:
$`𝐄_0(𝐫,t)={\displaystyle \frac{q_D}{r^2}}={\displaystyle \frac{M_p\omega _H^2}{r^2}}x\mathrm{sin}\omega _Ht𝐞^r`$ (27)
Apart from a phase factor the two expressions must be equal. Recalling the values of $`u_n`$ and $`\rho _0`$ from (23), the amplitude $`x`$ of proton oscillation can be computed. We obtain:
$`x={\displaystyle \frac{3d}{R_p}}={\displaystyle \frac{M_e}{(2\pi )^2M_p}}{\displaystyle \frac{1}{n}}`$ (28)
In the highest state of excitation, which for the dynamic model is $`n=1`$, the amplitude is less than $`10^5`$ times the proton radius: Oscillations are therefore comparatively small. This result indicates that the scale of energies within the proton is much higher than within the electron, say. The result is therefore well in keeping with existing nuclear models. For higher $`n`$, and thus lower excitation energy, the amplitude becomes smaller and vanishes for $`n\mathrm{}`$.
It is helpful to consider the different energy components within the hydrogen atom at a single state, say $`n=1`$, to understand, how the electron is actually bound to the proton. The energy of the electron consists of two components.
$`\varphi _K(𝐫,t)={\displaystyle \frac{\rho _0u_1^2}{r^2}}\mathrm{sin}^2k_1r\mathrm{cos}^2\omega _Ht`$ (29)
is the kinetic component of electron energy ($`k_1`$ is now the wavevector of the wave). As in the free case, the kinetic component is accompanied by an intrinsic field, which accounts for the energy principle (i.e. the requirement, that total energy density at a given point is a constant of motion). Thus:
$`\varphi _{EM}(𝐫,t)={\displaystyle \frac{\rho _0u_1^2}{r^2}}\mathrm{cos}^2k_1r\mathrm{cos}^2\omega _Ht`$ (30)
is the field component. The two components together make up for the energy of the electron. Integrating over the volume of the atom and a single period $`\tau `$ of the oscillation, we obtain:
$`W_{el}`$ $`=`$ $`{\displaystyle \frac{1}{\tau }}{\displaystyle _0^\tau }𝑑t{\displaystyle _{V_H}}d^3r\left(\varphi _K(𝐫,t)+\varphi _{EM}(𝐫,t)\right)`$ (31)
$`=`$ $`{\displaystyle \frac{1}{2}}M_eu_1^2`$ (32)
This is the energy of the electron in the hydrogen atom. $`W_{el}`$ is equal to 13.6 eV. The binding energy of the electron is the energy difference between a free electron of velocity $`u_1`$ and an electron in a hydrogen atom at the same velocity. Since the energy of the free electron $`W_{free}`$ is:
$`W_{free}=\mathrm{}\omega _H=M_eu_1^2`$ (33)
the energy difference $`\mathrm{}W`$ or the binding energy comes to:
$`\mathrm{}W=W_{free}W_{el}={\displaystyle \frac{1}{2}}M_eu_1^2`$ (34)
This value is also equal to 13.6 eV. It is, furthermore, the energy contained in the photon field $`\varphi _{Rad}(𝐫,t)`$ of the proton’s radiation
$`W_{Rad}`$ $`=`$ $`\mathrm{}W={\displaystyle \frac{1}{\tau }}{\displaystyle _0^\tau }𝑑t{\displaystyle _{V_H}}d^3r\varphi _{Rad}(𝐫,t)`$ (35)
$`=`$ $`{\displaystyle \frac{1}{2}}M_eu_1^2`$ (36)
This energy has to be gained by the electron in order to be freed from its bond, it is the ionization energy of hydrogen. However, in the dynamic picture the electron is not thought to move as a point particle in the static field of a central proton charge, the electron is, in this model, a dynamic and oscillating structure, which emits and absorbs energy constantly via the photon field of the central proton. In a very limited sense, the picture is still a statistical one, since the computation of energies involves the average over a full period.
## VI The meaning of $`\eta `$
The last problem, we have to solve, is the determination of $`\eta `$, the coupling constant between electromagnetic and mechanic variables. To this end we compute the energy of the radiation field $`W_{Rad}`$, using Eqs. (19), (27), and (28). From (19) and (27) we obtain:
$`\varphi _{Rad}(r,t)={\displaystyle \frac{1}{8\pi \eta }}𝐄^2={\displaystyle \frac{1}{8\pi \eta }}{\displaystyle \frac{M_p^2\omega _H^4}{r^4}}x^2\mathrm{sin}^2\omega _Ht`$ (37)
Integrating over one period and the volume of the atom this gives:
$`W_{Rad}`$ $`=`$ $`{\displaystyle \frac{1}{\tau }}{\displaystyle _0^\tau }𝑑t{\displaystyle _{R_p}^{R_H}}4\pi r^2𝑑r\varphi _{Rad}(𝐫,t)`$ (38)
$``$ $`{\displaystyle \frac{1}{4\eta }}{\displaystyle \frac{M_p^2\omega _H^4x^2}{R_p}}`$ (39)
provided $`R_p`$, the radius of the proton is much smaller than the radius of the atom. With the help of (28), and remembering that $`W_{Rad}`$ for $`n=1`$ equals half the electron’s free energy $`\mathrm{}\omega _H`$, this finally leads to:
$`W_{Rad}={\displaystyle \frac{1}{4\eta }}{\displaystyle \frac{M_p^2\omega _H^4x^2}{R_p}}={\displaystyle \frac{1}{2}}\mathrm{}\omega _H`$ (40)
$`\eta ={\displaystyle \frac{M_e^2\nu _H^3}{2hR_p}}={\displaystyle \frac{1.78\times 10^{20}}{R_p}}`$ (41)
since the frequency $`\nu _H`$ of the hydrogen atom equals $`6.57\times 10^{15}`$ Hz. Then $`\eta `$ can be calculated in terms of the proton radius $`R_p`$. This radius has to be inferred from experimental data, the currently most likely parametrization being :
$`{\displaystyle \frac{\rho _p(r)}{\rho _{p,0}}}={\displaystyle \frac{1}{1+e^{(r1.07)/0.55}}}`$ (42)
radii in fm. If the radius of a proton is defined as the radius, where the density $`\rho _{p,0}`$ has decreased to $`\rho _{p,0}/e`$, with e the Euler number, then the value is between 1.3 and 1.4 fm. Computing $`4\pi `$ the inverse of $`\eta `$, we get, numerically:
$`{\displaystyle \frac{4\pi }{\eta }}`$ $`=`$ $`0.92\times 10^{34}(R_p=1.3fm)`$ (43)
$`=`$ $`0.99\times 10^{34}(R_p=1.4fm)`$ (44)
$`=`$ $`1.06\times 10^{34}(R_p=1.5fm)`$ (45)
Numerically, this value is equal to the numerical value of Planck’s constant $`\mathrm{}`$ :
$`\mathrm{}_{UIP}=1.0546\times 10^{34}`$ (46)
Given the conceptual difference in computing the radius the agreement seems remarkable. Note that this is a genuine derivation of $`\mathrm{}`$, because nuclear forces and radii fall completely outside the scope of the theory in its present form. If measurements of $`R_p`$ were any different, then we would be faced, at this point, with a meaningless numerical value. Reversing the argument it can be said, that the correct value - or rather the meaningful value - is a strong argument for the correctness of our theoretical assumptions. Since these assumptions involve to a greater or lesser extent the whole theory of matter waves developed so far, we devote the rest of this section mainly to a critical analysis of this result and shall show the most striking physical implications only at the end.
Starting with the approximations involved, we note (i) a first order approximation in d, and (ii) an approximation in the integration. Since $`d10^5R_p`$, and $`R_p10^5R_H`$, both errors are negligible. In view of the standard experimental error margins, also the deviation of a few percent, depending on how we define the proton radius, seems acceptable. On the positive side, there are two plausibility arguments, indicating that we deal not only with a numerical coincidence: (i) The Planck constant describes the interface between frequency and energy in all fundamental experiments. Since we started with a frequency (= proton oscillations), and calculated an energy, it must have, at some point, entered the calculation. The only unknown quantity in the calculation was $`\eta `$: therefore it should contain $`\mathrm{}`$. (ii) What we have in fact developed with this model of hydrogen, is in spirit very close to the harmonic oscillator in quantum theory; the rest energy term is related to the energy of our photon field. In order to be compatible with quantum theory, the energy must contain $`\mathrm{}`$. Again, the only variable, which could contain it, is $`\eta `$.
It can also be asked, why electromagnetic variables are multiplied by $`\mathrm{}`$ to give the energy of radiation. Especially, since the finestructure constant contains a division by $`\mathrm{}`$:
$`\alpha ={\displaystyle \frac{e^2}{\mathrm{}c}}`$ (47)
To answer this question, consider a variable in electrodynamics $`A_{ED}`$ and its correlating variable $`A_M`$ in mechanics. Then the transition from $`A_{ED}`$ to $`A_M`$ is described by a transformation $`T`$, so that:
$`A_M=T_{M,ED}A_{ED}`$ (48)
Since the inverse transformation must exist and the variables are assumed to be unique, the transformation is unitary:
$`T_{M,ED}T_{ED,M}=T_{M,ED}T_{M,ED}^1=1`$ (49)
In our case the primary variables are the electromagnetic ones: $`𝐩,𝐄,𝐁`$. And the transformation involves a multiplication by $`\mathrm{}`$.
$`A_M=\mathrm{}A_{ED}`$ (50)
The fundamental units m, kg, s are, in this system, the natural system, tied to the electromagnetic variables. In quantum theory, on the other hand, the fundamental variables are Newtonian. Then the transformation between electromagnetic and mechanic variables involves the inverse transformation.
$`A_M(QM)=\mathrm{}^1A_{ED}(QM)`$ (51)
If we consider, in addition, that charge has been included in the definition of $`𝐄`$, the transformation, in conventional units and in quantum theory should read:
$`A_M(QM)={\displaystyle \frac{e^2}{\mathrm{}}}A_{ED}(QM)`$ (52)
which is the finestructure constant multiplied by $`c`$. And $`c`$ is, generally, only a matter of convention. Therefore we think, the conclusion, that $`4\pi /\eta `$ really is $`\mathrm{}`$, is a reasonable and safe one. But in this case Planck’s constant has not much bearing on a different scale of measurement, as is often invoked, when there is talk about $`\mathrm{}0`$ in the macroscopic scale. The constant bears on two fundamental aspects of matter itself. As we see it, $`\mathrm{}`$ describes the interface between electromagnetic and mechanic variables of matter. It is therefore even more fundamental than currently assumed. For the following, let us redefine the symbol $`\mathrm{}`$ by:
$`\mathrm{}:=1.0546\times 10^{34}[N^1m^4]`$ (53)
Then we can rewrite the equations for $`𝐅`$, the Lorentz force, for $`𝐋`$, angular momentum related to this force, and $`\varphi _{Rad}`$, the radiation energy density of a photon in a very suggestive form:
$`𝐅=\mathrm{}q\left({\displaystyle \frac{𝐄}{4\pi }}+𝐮\times {\displaystyle \frac{𝐁}{4\pi }}\right)`$ (54)
$`𝐋=\mathrm{}q𝐫\times \left({\displaystyle \frac{𝐄}{4\pi }}+𝐮\times {\displaystyle \frac{𝐁}{4\pi }}\right)`$ (55)
$`\varphi _{Rad}={\displaystyle \frac{\mathrm{}}{2}}\left[\left({\displaystyle \frac{𝐄}{4\pi }}\right)^2+c^2\left({\displaystyle \frac{𝐁}{4\pi }}\right)^2\right]`$ (56)
Every calculation of mechanic properties involves a multiplication by $`\mathrm{}`$. Since $`\mathrm{}`$ is a scaling constant, the term ”quantization”, commonly used in this context, is misleading. Furthermore, it is completely irrelevant, whether we compute an integral property (the force in (54)), or a density ($`\varphi _{Rad}`$ in (56), a force density can also be obtained by replacing charge q by a density value). From the interaction fields within a hydrogen atom, e.g. Eq. (37), it is clear that the field varies locally and temporary and can reach any value between zero and its maximum. Although it is described by:
$`\varphi _{Rad}(r,t)={\displaystyle \frac{\mathrm{}}{2}}{\displaystyle \frac{M_p^2\omega _H^4}{r^4}}x^2\mathrm{sin}^2\omega _Ht`$ (57)
it is not ”quantized”. Neither would be the forces based on the field $`𝐄_0`$, or the angular momenta. Although, in both cases, they are proportional to $`\mathrm{}`$. What is, in a sense, discontinuous, is the mass contained in the shell of the atom. But this mass depends, as does the amplitude of $`\varphi _{Rad}(r,t)`$, on the mass of the atomic nucleus. So the only discontinuity left on the fundamental level, is the mass of atomic nuclei. That the energy spectrum of atoms is discrete, is a trivial observation in view of boundary conditions and finite radii. To sum up the argument: There are no quantum jumps.
All our calculations so far focus on single atoms. To get the values of mechanic variables in SI units used in macrophysics, we have to include the scaling between the atomic domain and the domain of everyday measurements. Without proof, we assume this value to be $`N_A`$, Avogadro’s number. The scale can be made plausible from solid state physics, where statistics on the properties of single electrons generally involve a number of $`N_A`$ particles in a volume of unit dimensions . And a dimensionless constant does not show up in any dimensional analysis.
## VII Solar gravity fields
We conclude this paper, which seems to open a new perspective on a number of very fundamental problems, by a brief discussion. The first issue concerns the nature of gravity. From the given treatment it is possible to conclude, that there is maybe no fundamental difference between electrostatic and gravitational interaction. Both seem to be transmitted with the velocity of light, both obey a $`r^2`$ law, both are related to the existence of mass, whether its static or its dynamic features. So the conjecture, that also gravity is transmitted by a ”photon”, has it least some basis. But here the similarities end. Because of the vast differences in the coupling $`ϵG10^{22}`$ one must assume a very different frequency scale. From Eq. (27):
$`|𝐄|={\displaystyle \frac{q_D}{r^2}}{\displaystyle \frac{M\omega ^2}{r^2}}`$ (58)
it can be inferred that the frequency scale for gravity and electrostatics would differ by about $`10^{11}`$.
$`\omega _G10^{11}\omega _E`$ (59)
Here $`\omega _E`$ is the characteristic electromagnetic frequency, $`\omega _G`$ its gravitational counterpart. The hypothesis can in principle be tested. If we assume that $`\nu _G`$, the frequency of gravity radiation, is about $`10^{11}`$ times the frequency of proton oscillation, we get:
$`\nu _G1100kHz`$ (60)
If therefore electromagnetic fields of this frequency range exist in space, we would attribute these fields to solar gravity. To estimate the intensity of these, hypothetical, waves, we use Eq. (26):
$`𝐆_S(𝐫,t)={\displaystyle \frac{}{t}}𝐩_E(𝐫,t)`$ (61)
Here $`𝐆_S`$ is the solar gravity field. The momentum density and its derivative can be inferred from centrifugal acceleration.
$`{\displaystyle \frac{}{t}}𝐩_E(𝐫,t)=\rho _Ea_C𝐞^r`$ (62)
$`\rho _E={\displaystyle \frac{3M_E}{4\pi R_E^3}}a_C=\omega _E^2R_O`$ (63)
where $`R_O`$ is the earth’s orbital radius and where we have assumed isotropic distribution of terrestrial mass. Then Eq. (37) leads to:
$`\varphi _G(r=R_O)={\displaystyle \frac{\mathrm{}}{2}}\left({\displaystyle \frac{G_S}{4\pi }}\right)^2={\displaystyle \frac{\mathrm{}}{2}}\left({\displaystyle \frac{3M_ER_O}{4R_E^3\tau _E^2}}\right)^2`$ (64)
Note the occurrence of Planck’s constant also in this equation, although all masses and distances are astronomical. The intensity of the field, if calculated from (64), is very small. To give it in common measures, we compute the flow of gravitational energy through a surface element at the earth’s position. In SI units we get:
$`J_G(R_O)=\varphi _G(R_O)N_Ac70mW/m^2`$ (65)
Compared to radiation in the near visible range - the solar radiation amounts to over 300 Watt/m<sup>2</sup> \- the value seems rather small. But considering, that also radiation in the visible range could have an impact on terrestrial motion, the intensity of the gravity waves could be, in fact, much higher.
## VIII Is there static charge?
In the conventional models a particle’s charge is not only discrete, but has also a defined sign. Although anti-particles are thought to exist, the charge of protons is positive, the charge of electrons negative. Dynamic charge is neither discrete, nor does it possess a defined sign. Depending on the exact moment, the charge of a proton:
$`q_p=M_p\omega _H^2x\mathrm{sin}\omega _Ht`$ (66)
either has a positive or a negative value, which determines the direction of the energy flow within the hydrogen atom. The difference between electrons and protons in this model is mainly due to their density of mass.
Related to this feature is a shift of focus within the dynamic model of atoms. Although the states of the atom are described by quantum numbers ($`n`$ for the principal state, $`l`$ and $`m`$ if multipoles are included), these numbers refer primarily to nuclear states of oscillation. States of the atom’s electron are merely a reaction to them. Therefore the properties of an atom, in the dynamic model, refer to properties of the atomic nucleus. How this model bears on chemical properties, remains to be seen.
The last issue is a consequence of our treatment of the hydrogen atom. In this case the main features, the energy spectrum as well as ionization energy and the energy of emitted photons can be explained from dynamic charge alone. There is, in contrast to the conventional treatment, no necessity to invoke static potentials. It will also have been noted that in natural units and based on dynamic processes interactions are generally free of any notion of ”charge” in its proper sense. So does that mean, it could be asked, that there is no charge? Based on the current evidence and considering the situation in high energy physics, this is definitely too bold a statement. Considering, though, that the notion of a fixed ”elementary charge” lies at the heart of all current accounts of these experiments, the degree of theoretical freedom in the dynamic picture is incomparably higher. So that still, after a few years of development, we might end up with a tentative answer: maybe not. And in this case, the question about the true nature of charge will have been answered. It is dynamic in nature, we would then say.
## IX Conclusions
In this paper we presented evidence for the existence of a dynamic component of charge. It derives, as shown, from the variation of a particle’s density of mass. A new system of electromagnetic units, the natural system, has been developed, which bears on these dynamic sources. We have given a fully deterministic treatment of hydrogen, where we used our theoretical model to determine the fundamental scaling constant between electromagnetic and mechanic variables. The constant, we found, is $`\mathrm{}`$, Planck’s constant. The constant thus has no bearing on any length scale, as frequently thought. And finally we have discussed these results in view of unifying gravity and quantum theory. The intensity of the postulated solar gravity waves seems sufficiently high, so that these waves, in the low frequency range of the electromagnetic spectrum, can in principle be detected.
## Acknowledgements
I’d like to thank Jaime Keller. In our discussions I realized, for the first time, that the most efficient theory of electrodynamics might be one without electrons. |
no-problem/0001/astro-ph0001472.html | ar5iv | text | "# Faint Infrared Flares from the Microquasar GRS 1915+105\n\n## 1 Introduction\n\nAs the archetypal(...TRUNCATED) |
no-problem/0001/math0001034.html | ar5iv | text | "# 1 Introduction\n\n## 1 Introduction\n\nUniversal twists connecting (affine) quantum groups to (el(...TRUNCATED) |
End of preview. Expand
in Data Studio
Marin Markdownified Ar5iv
Markdownified Ar5iv transforms academic papers from arXiv into clean, structured Markdown format consisting of 2.74B tokens across two splits. This dataset preserves th content while making it accessible for language model training on academic text.
| Value | |
|---|---|
| Tokens | 2 742 463 924 |
| Primary source | https://sigmathling.kwarc.info/resources/ar5iv-dataset-2024/ |
| File format | JSONL |
| License | C-UDA-1.0 (mirrors upstream Ar5iv licenses) |
Processing and Cleaning Pipeline
Our conversion pipeline combines several sophisticated techniques to transform raw Wikipedia HTML into high-quality Markdown:
HTML Preprocessing: We start with the Ar5iv dump in Extended DOLMA format, which provides HTML representations of academic papers with metadata.
Structural Cleanup
- The abstract is transformed into a proper section heading for consistent document structure
- LaTeX equations are carefully preserved using inline ($...$) and display ($$...$$) notation
- Code blocks and listings maintain proper formatting with appropriate line breaks
Noise Reduction:
- Author information is removed
- Title page elements are streamlined to avoid redundancy
- The Ar5iv footer is removed to eliminate conversion metadata
- Figure captions are removed to focus on the main content
- Bibliography sections, footnotes, and citation links are removed
Formatting Cleanup:
- List items are cleaned to prevent duplicate numbering patterns (e.g., "1. 1.")
- Content before the first main section (typically metadata) is removed
- Equation tables are converted to inline elements for better rendering
DOM Simplification: We employ a custom-enhanced version of Resiliparse that preserves semantic HTML structure. Rather than flattening to plain text, we retain important elements like headings, paragraphs, lists while removing scripts, tracking code, and boilerplate.
Markdown Conversion: Our custom Markdownify implementation transforms the simplified DOM into clean Markdown. The final output stores each article as a JSON object containing the Markdown text and essential metadata.
Dataset Variants
The Markdownified Ar5iv dataset comes in two variants:
- Ar5iv No Problem (2.74B tokens): Papers that were converted without significant issues or warnings during the HTML generation process. This subset represents the cleanest and most reliable papers.
- Ar5iv Warning (19.6B tokens): Papers that generated warnings during conversion from LaTeX to HTML. While still valuable, these may contain occasional formatting artifacts.
Usage Example
from datasets import load_dataset
ds = load_dataset(
"marin-community/ar5iv-no-problem-markdown",
split="train",
streaming=True
)
for article in ds.take(3):
print(article["text"])
Citation
If you use this dataset in your research, please cite both the original Wikipedia contributors and our work:
@misc{markdownified_ar5iv_2024,
title = {Markdownified Ar5iv},
author = {The Marin Community},
year = {2024},
url = {https://huggingface.co/datasets/marin-community/ar5iv-no-problem-markdown}
}
License
All content inherits Ar5iv's licensing: C-UDA-1.0. Our conversion tools and pipeline are released under Apache 2.0.
Acknowledgement
We extend our gratitude to:
- Arxiv Labs and KWARC for their work on the Ar5iv dataset
- Janek Bevendorff for the Resiliparse project
- Matthew Dapena-Tretter for Markdownify
- Downloads last month
- 262