cond-mat0401066/epc.tex
1: \documentclass[prb,twocolumn,showpacs,showkeys]{revtex4}
2: \usepackage{graphicx,epsfig}
3: \usepackage{bm}
4: \usepackage{dcolumn}
5: 
6: %Some definitions are here 
7: 
8: \def\bea{\begin{eqnarray}}
9: \def\eea{\end{eqnarray}}
10: \def\be{\begin{equation}}
11: \def\ee{\end{equation}}
12: \def\pp{\partial}
13: 
14: \begin{document}
15: 
16: \title{The electron-phonon coupling is large for localized states }
17: 
18: \author{Raymond Atta-Fynn}
19: \email{attafynn@phy.ohiou.edu}
20: \author{Parthapratim Biswas}
21: \email{biswas@phy.ohiou.edu}
22: \author{D.~A.~Drabold}
23: \email{drabold@ohio.edu}
24: 
25: \affiliation{Department of Physics and Astronomy, Ohio University, Athens, OH 45701}
26: 
27: \keywords{amorphous silicon, electron localization, electron-phonon coupling}
28: 
29: \pacs{71.23.Cq, 71.15.Mb, 71.23.An}
30: 
31: \begin{abstract}
32: From density functional calculations, we show that localized states stemming 
33: from defects or topological disorder exhibit an anomalously large electron-phonon 
34: coupling. We provide a simple analysis to explain the observation and perform a 
35: detailed study on an interesting system: amorphous silicon. We compute first 
36: principles deformation potentials  (by computing the sensitivity of specific 
37: electronic eigenstates to individual classical normal modes of vibration). We 
38: also probe thermal fluctuations in electronic eigenvalues by first principles 
39: thermal simulation. We find a strong correlation between a static property of the 
40: network [localization, as gauged by inverse participation ratio (IPR)] and a 
41: dynamical property (the amplitude of thermal fluctuations of electron energy 
42: eigenvalues) for localized electron  states.  In particular, both the electron-phonon 
43: coupling and the variance of energy eigenvalues are proportional to the IPR 
44: of the localized state. We compare the results for amorphous Si to photoemission 
45: experiments. While the computations are carried out for silicon, very similar 
46: effects have been seen in other systems with disorder.  
47: \end{abstract}
48: 
49: \maketitle
50: \section{INTRODUCTION}
51: 
52: Electron states of finite spatial extent, so called ``localized" states are ubiquitous 
53: in nature. Localized molecular orbitals occur in molecules and in solid state 
54: systems with disorder (this could take the form of defects, such as a dangling 
55: bond state or more subtly from topological and or chemical disorder in amorphous 
56: materials or glasses). Such localized states are also known in polymers. Surface 
57: states, as another form of defect, also are often spatially compact. The study of 
58: localization in its own right has been an important and active subfield of 
59: condensed matter theory since the fifties.
60: 
61: Another physical quantity of key importance is the electron-phonon (e-p) coupling, 
62: the interaction term connecting the electronic and lattice systems. Perhaps most 
63: spectacularly, the e-p coupling is the origin of superconductivity as expressed 
64: in BCS theory\cite{Bardeen}.\,\,Phillips\cite{Phillips} has shown that large e-p couplings 
65: in the cuprate superconductors can lead to a successful model of high T$_c$ 
66: superconductivity\cite{Bednorz} within the framework of conventional BCS 
67: superconductivity. The e-p coupling is also the mediator of all light-induced 
68: structural changes in materials. In amorphous silicon the greatest outstanding 
69: problem of the material, the Staebler-Wronski effect\cite{Staebler}, depends 
70: critically upon the electron-lattice interaction. A zoo of analogous effects is 
71: studied in glasses; perhaps the most important example is reversible photo-amorphization 
72: and photo-crystallization used in the GeSbTe phase-change materials used in 
73: current writable CD and DVD technology.
74: 
75: Previous thermal simulations with Bohn-Oppenheimer dynamics have indicated 
76: that there exists a large electron-phonon coupling for the localized 
77: states in the band tails and in the optical gap~\cite{Li,Drabold1}. 
78: Earlier works on chalcogenide glasses by Cobb and Drabold~\cite{Cobb} have 
79: emphasized a strong correlation between the thermal fluctuations as gauged 
80: by root mean square (RMS) variation in the LDA eigenvalues and wave function 
81: localization of a gap or tail state (measured by inverse participation 
82: ratio\cite{Drabold2}, a simple measure of localization). Drabold and 
83: Fedders\cite{Drabold3} have also shown that localized eigenvectors may 
84: fluctuate dramatically even at room temperature. Recently, Li and 
85: Drabold relaxed the adiabatic (Born-Oppenheimer) approximation to 
86: track the time-development of electron packets scattered by lattice vibrations\cite{Li}.
87: In this paper, we examine the electron phonon coupling and provide a heuristic 
88: analysis of the e-p coupling for localized electron states. We explore the e-p 
89: coupling in some detail for a particular model system (amorphous silicon) which 
90: provides us with a convenient variety of localized, partly localized ``bandtail" 
91: and extended states. We compute  deformation potential (which measures the response 
92: of a selected electron state to a particular phonon), and also track thermally-induced 
93: fluctuations of electronic eigenvalues. We find that localized states always exhibit 
94: a large e-p coupling. Our computations are carried out using a first principles 
95: molecular dynamics code {\sc Siesta}, and the eigenvalues and states that we 
96: study are from the Kohn-Sham equations with a rich local orbital basis. A rationale 
97: for the study of the Kohn-Sham states is given elsewhere\cite{Attafynn}. We 
98: emphasize that the results that we give are qualitatively general -- not just 
99: an artifact of studying a disordered phase of silicon (we have, for example, 
100: seen exactly the same effects in various binary glasses which exhibit very 
101: different topological and chemical disorder).
102: \section{THEORY}
103: 
104: To establish a connection between electron-phonon coupling and 
105: wave function localization for the electrons, we consider an 
106: electronic eigenvalue $\lambda_n$ near the band gap of a-Si. The 
107: sensitivity of $\lambda_n$ due to an arbitrary small displacement 
108: of an atom (possibly thermally induced) can be estimated using the 
109: Hellmann-Feynman theorem~\cite{Feynman},
110: %
111: \[
112: \frac{\pp \lambda_n}{\pp \bf R_{\alpha}} = \langle 
113: \psi_n|{\frac{\pp \bf H}{\pp \bf R_{\alpha}}}|\psi_n \rangle.
114: \] 
115: %
116: Here we have assumed that the basis functions are fixed and 
117: $|\psi_n \rangle$ are the eigenvectors of the Hamiltonian {\bf H}. 
118: For small lattice distortion $\{\delta\bf R_\alpha\}$, the 
119: corresponding change in $\delta\lambda_n$ is, 
120: %
121: \be
122: \label{first}
123: \delta\lambda_n \approx \sum_{\alpha = 1}^{3N}\,\langle \psi_n|\frac{\partial {\bf H}}{\partial {\bf 
124: R_{\alpha}}}|\psi_n \rangle\,\delta\bf R_\alpha
125: \ee
126: %
127: where $\rm N$ is the total number of atoms in the model. If the displacement 
128: $\delta\bf R_{\alpha}(\rm t)$  arises from classical vibrations, one can 
129: write~\cite{Note1}, 
130: %
131: \be
132: \label{second}
133: \delta \bf R_{\alpha}(\rm t) = \sum_{\omega = 1} ^{3N} \rm A(\rm T,\omega)\,\cos(\omega\,t 
134: + \phi_\omega)\,\chi_{\alpha}(\omega),
135: \ee
136: %
137: where $\rm \omega$ indexes the normal mode frequencies, $\rm A(\rm T,\rm \omega)$ is 
138: the temperature dependent amplitude of the mode with frequency $\rm \omega$,  
139: $\phi_\omega$ is an arbitrary phase, $ \rm \chi_{\rm \alpha}(\rm \omega)$ is a normal mode 
140: with frequency  $\omega$ and vibrational displacement index $\alpha$. Using
141: the temperature dependent squared amplitude
142: $ \rm A^2 = \rm 3k_BT/\rm M \rm \omega^2$, 
143: the trajectory (long time) average of $\rm \delta \rm\lambda_{n}^2$ 
144: can be written (using Eq.\,\ref{first} and \ref{second}) as, 
145: 
146: \be
147: \label{eq-1}
148: \langle {\delta \lambda_{n}^2}\rangle  = \lim_{\tau \rightarrow \infty}\frac{1}{\tau}\int_0^\tau dt \, \delta 
149: \lambda^2_n(t) \approx \rm \left(\frac{3k_BT}{2M}\right) \sum_{\omega=1}^{3N}\,\frac{\Xi_{n}^2(\omega)}{\omega^2},
150: \ee
151: where the electron-phonon coupling $\rm \Xi_{n}(\omega)$ is given by,
152: 
153: \be
154: \label{eq-2}
155: \Xi_{n}(\omega) = \sum_{\alpha = 1}^{3N} \langle \psi_n|\frac{\partial{\bf H}}{\partial{\bf R_{\alpha}}}|\psi_n
156: \rangle\,\chi_{\alpha}(\omega).
157: \ee
158: %
159: One can infer from Eq.\,\ref{eq-1} that thermally induced fluctuation
160: in the energy eigenvalues is a consequence of electron-phonon 
161: coupling.  Note that for a given electronic eigenvalue, the 
162: contribution to the coupling comes from the entire vibrational spectrum 
163: involving all the atoms in the systems. Since the normalized eigenstate
164: can be written as $|\psi_n\rangle = \sum_{i}a_{ni}\,|i\rangle$ 
165: where $|i\rangle$ are the basis orbitals, it follows from Eq.\,\ref{eq-2} that, 
166: 
167: \begin{widetext}
168: \bea
169: \label{eq-3}
170: \Xi^2_{n}(\omega) &=& \sum_{\alpha,\beta,i,j,k,l}\,
171: a^{*}_{ni}a_{nj}a^{*}_{nk}a_{nl}\langle i|\frac{\partial \bf H}{\partial \bf R_{\alpha}}|j 
172: \rangle\,\langle k|\frac{\partial \bf H}{\partial \bf R_{\beta}}|l\rangle \, \chi_{\alpha}(\omega) 
173: \, \chi_{\beta}(\omega) \nonumber \\ 
174: &=& \sum_{i,\alpha}|a_{ni}|^4\, [\langle i|\frac{\partial \bf H}
175: {\partial \bf R_{\alpha}}|i \rangle]^2  \, \chi_{\alpha}^2(\omega) + 
176: \sum_{ijkl\alpha \beta}^{\prime} a^{*}_{ni}a_{nj}a^{*}_{nk}a_{nl}\langle i|\frac{\partial \bf H}
177: {\partial \bf R_{\alpha}}|j \rangle\,\langle
178:  k|\frac{\partial \bf H}{\partial \bf R_{\beta}}|l\rangle 
179: \, \chi_{\alpha}(\omega) \,\chi_{\beta}(\omega)
180: \eea
181: \end{widetext}
182: 
183: The first term in the second line of Eq.\,\ref{eq-3} is positive definite (diagonal) 
184: while the second one, the off-diagonal term (indicated by the prime), is not of a single sign. 
185: In the event that only a few $a_{ni}$ dominate (the case for localized states), then the 
186: leading contribution to the electron-phonon coupling
187: originates largely from the diagonal term. The addition of a large number 
188: of terms of mixed sign and small magnitude leads to cancellations in the off-diagonal term leaving behind a small contribution to electron-phonon 
189: coupling. By comparing to direct calculations with the full Eq.\,\ref{eq-3}, we show 
190: that dropping the second term appears to be reasonable for well-localized electron 
191: states.  The approximate ``diagonal" electron-phonon coupling can be written as, 
192: 
193: \bea
194: \label{eq-4}
195: \Xi^2_{n}(\omega) & \approx & \sum_{\alpha  = 1}^{3N}\sum_{i=1}^{N_b} \,|a_{ni}|^4\,
196: [\langle i|\frac{\partial{\bf H}}{\partial{\bf R_{\alpha}}}|i \rangle]^2 
197: \chi_{\alpha}^2(\omega) \nonumber \\ 
198: &  =  & \sum_{\alpha = 1}^{3N} \sum_{i=1}^{N_b}\, {q^2_{ni}}\, 
199: [\langle i|\frac{\partial{\bf  H}}{\partial{\bf R_{\alpha}}}|i \rangle]^2  
200: \chi_{\alpha}^2(\omega)
201: \eea
202: 
203: where N$_b$ is the number of basis orbitals and $q_{ni} = |a_{n,i}|^2$ is 
204: the charge sitting on the $i$th orbital for a given normalized eigenstate 
205: $|\psi_n \rangle$. The degree of wave function localization can be measured
206: by defining inverse participation ratio $ \rm {\cal I}$ for the eigenstates
207: $|\psi_n \rangle$,
208: %
209: \be
210: \label{eq-5}
211: \rm {\cal I}(n) = \sum_{i=1}^{\rm N_{b}}\,q^2_{ni}.
212: \ee
213: %
214: Equation \ref{eq-4} leads to an approximate but analytic connection between 
215: ${\cal I}$ and electron-phonon coupling. Since $\rm {\cal I}$ is large for 
216: localized states, one expects $\Xi_{n}(\omega)$ (and therefore $\langle 
217: \delta \lambda^2_n \rangle$) to be large for a localized state. If we further
218: assume that $\gamma^2(\omega,i)= \sum_{\alpha = 1}^{3N}
219: [\langle i|\partial{\bf  H}/\partial{\bf R_{\alpha}}|i \rangle]^2  
220: \chi_{\alpha}^2(\omega)$ is weakly dependent upon site/orbital index $i$, 
221: then 
222: 
223: \be
224: \label{third} 
225: \Xi^2_{n}(\omega)~\sim {\cal I}_n \times f(\omega), 
226: \ee
227: 
228: where $f(\omega)$ is defined from $\gamma^2$ and Eq.\,\ref{eq-4}. In 
229: this ``separable" approximation, it is also the case that $\langle \delta \lambda^2_n\rangle \propto {\cal I}_n$.  
230: 
231: \begin{figure}
232: \includegraphics[width=3.5in, height=3.5in]{fig1}
233: \caption{\label{fig1}
234: Electron-phonon coupling surface plot for a 216-atom model 
235: of \emph{a}-Si. The absolute value of electron-phonon coupling 
236: $|\Xi|$ (cf. Eq.\,\ref{eq-2}) is plotted as a function of phonon 
237: frequency $\omega$ and energy eigenvalues near the gap. 
238: The largest value of $|\Xi|$ in the plot corresponds to the 
239: eigenvalue for HOMO, which is the most localized state in the 
240: spectrum.}
241: \end{figure}
242: 
243: 
244: \section{Method}
245: 
246: The model of \emph{a}-Si we have used in our calculations was
247: generated by Barkema and Mousseau~\cite{Barkema} using an improved
248: version of the Wooten, Winer and Weaire (WWW) algorithm~\cite{Wooten}.
249: The details of the construction was reported in Ref.\,\onlinecite{Barkema}. 
250: The model consists of 216 atoms of Si packed inside a cubic box of length 16.282{\AA}
251: and has two 3-fold coordinated atoms. The average bond angle is 
252: $109.5^{\circ}$ with a root mean square deviation of $11.0^{\circ}$.
253: The density functional calculations were performed within the local density
254: approximation (LDA) using the first principles code {\sc Siesta}
255: \cite{Siesta1,Siesta2,Siesta3}. We have used a non self-consistent version of
256: density functional theory based on the linearization of the Kohn-Sham
257: equation by Harris functional approximation~\cite{Harris} along with the
258: parameterization of Perdew and Zunger~\cite{Perdew} for the exchange-correlation
259: functional. 
260: %Norm conserving Troullier-Martins pseudopotentials~\cite{Troullier}
261: %factorized in the Kleinman-Bylander form~\cite{Kleinman} were used to
262: %remove the core electrons. 
263: The choice of an appropriate basis is found be very important and has been discussed at length 
264: in a recent communication~\cite{Attafynn}. While the minimal basis
265: consisting of one $s$ and three $p$ electrons can adequately describe the
266: electronic structure of amorphous silicon in general, there is some 
267: concern about the applicability of these minimal basis in describing deeply 
268: localized and low lying excited states in the conduction bands accurately. 
269: We have therefore employed a larger single-$\zeta$ basis with 
270: polarization (d) orbitals (SZP)~\cite{Sankey, Note2} in the present work. 
271: Throughout the calculation we have used only the $\Gamma$ point to 
272: sample the Brillouin zone.
273: 
274: \begin{figure}
275: \includegraphics[width=3.0in, height=3.5in]{fig2}
276: \caption{\label{fig2} 
277: (Color online) Mean square fluctuations of electronic eigenvalues 
278: versus inverse participation ratio plot at different temperature. 
279: The fluctuations at temperature 150K and 300K are found to be 
280: linearly correlated with the participation ratio for the corresponding 
281: eigenstates as predicted in section II. The correlation coefficient ($r$)  
282: for different temperature is indicated in the plot. 
283: }
284: \end{figure}
285: 
286: Starting with a fully 
287: relaxed configuration, we construct the dynamical matrix elements by successively displacing 
288: each atom in the supercell along three orthogonal directions (x, y and z) 
289: by 0.01{\AA} and computing the forces for each configuration. 
290: Within the harmonic approximation, the spring constant associated with 
291: each atom and direction can be written as a second derivative of the total energy 
292: with respect to the displacement of the atom in that direction. We have checked 
293: the convergence of both the matrix elements by using a different set of values for 
294: atomic displacement and used a value of {0.01{\AA}} in our calculations. 
295: $\partial \lambda/\partial {\bf R}_\alpha$ was obtained by finite differences from the 
296: dynamical matrix calculations.  The calculation does not need any extra effort beyond that of forming the
297: dynamical matrix. 
298: 
299: To explore the validity of our analysis and to elucidate
300: the connection between the localization (IPR) ($\rm {\cal I}$)
301: of electronic eigenstates and fluctuation of the conjugate eigenvalues, we performed
302: thermal MD simulations at constant temperatures using a No\'se-Hoover thermostat. The 
303: simulations were performed at
304: temperatures 150K, 300K, 500K and 700K with a time step of 2.5fs for
305: a total period of 2.5ps. For a given temperature, the mean square fluctuations 
306: were computed by tracking the eigenvalues at each time step and averaging over 
307: the total time of simulation excepting the first few hundred time steps  to 
308: ensure equilibration. 
309: %
310: \begin{figure}
311: \includegraphics[width=3.5in, height=3.7in]{fig3}
312: \caption{\label{fig3} 
313: The average electronic density of band tails states for
314: four different temperature T = 150K, 300K, 500K and 700K. 
315: Note that the conduction band tails (right) near the Fermi level (which is between the two tails) 
316: are more sensitive to thermal disorder than the valence band tails (left) 
317: providing a qualitative agreement with experimental result in 
318: Ref.\,25.  
319: }
320: \end{figure}
321: %
322: The mean square fluctuation ($\rm {\cal R}$) for an energy 
323: eigenvalue $\rm \lambda_n(t)$ is defined as : 
324: \be
325: \label{eq-6}
326:  \rm {\cal R}_n = \langle (\lambda_n(t) - \langle \lambda_n \rangle)^2
327: \rangle,
328: \ee
329: where $\langle \,\rangle$ denotes the average over time.
330: We study the fluctuations of $\{\rm \lambda_n(\rm t)\}$ by 
331: plotting against time at a given temperature and compare it 
332: with the $\rm {\cal I}$ obtained for the corresponding 
333: eigenvalues. For illustrations of such adiabatic evolution of Kohn-Sham
334: eigenvalues, see Ref.\,~\onlinecite{Drabold3}.
335: 
336: \section{Results}
337: 
338: % Report e-p coupling obtained without approximation 
339: 
340: In Fig.\,\ref{fig1}, we have plotted the electron-phonon coupling 
341: for the states near the band gap obtained directly from Eq.\,{\ref{eq-2}. 
342: It is clear from the figure that the e-p coupling is large only in 
343: the vicinity of conduction and valence band tails. The largest e-p 
344: coupling in the plot corresponds to the highest occupied molecular 
345: orbital (HOMO) in the optical frequency regime around 415 cm$^{-1}$. 
346: The lowest occupied molecular orbital (LUMO) also has a large feature 
347: around the same frequency. 
348: A Mulliken charge analysis and inverse participation ratio 
349: calculation of the electronic eigenfunctions have shown that 
350: both these two states -- the HOMO and LUMO are highly localized 
351: and are centered around the dangling bonds present in the model. 
352: On moving further from the band tails in either direction 
353: along the energy axis, the e-p coupling drops quickly and the 
354: surface becomes featureless for a given eigenvalue.  This behavior of e-p coupling 
355: can be understood from the arguments presented in section II where we have 
356: shown that the e-p coupling for localized states is directly proportional 
357: to the inverse participation ratio. 
358: For a localized state, therefore, the large value of electron-phonon 
359: coupling can be attributed to the large value of inverse participation ratio associated 
360: with that state. 
361: Since HOMO and LUMO are the two most localized states in the spectrum, the e-p 
362: coupling is large for these states and as we move toward the tail states, 
363: the coupling decreases. It is important to note that the plot in the 
364: Fig.\,\ref{fig1} has been obtained from Eq.\,\ref{eq-2} without making any 
365: approximation and is exact inasmuch as the matrix elements obtained from 
366: the density functional Hamiltonian are correct. 
367: This observation supports our assumptions that the dominant contribution 
368: to e-p coupling comes from the diagonal term in Eq.\,\ref{eq-3} and 
369: that $\gamma^2(\omega,i)$ is weakly dependent upon site/orbital index $i$ and also indicates 
370: from direct simulation there exists a linear relationship between mean 
371: square fluctuation of electronic eigenvalues and the corresponding 
372: inverse participation ratio for localized states. 
373: 
374: %%% On the correlation between MS and IPR
375: 
376: In order to justify our arguments further presented in section II, we 
377: now give a look at the mean square fluctuation of energy eigenvalues. 
378: As outlined in section III, we have computed the mean square fluctuations 
379: at four different temperature (150K, 300K, 500K and 700K) from MD runs 
380: over a period of 2.5ps and plotted in the Fig.\,\ref{fig2}. The fluctuation
381: obtained this way provides a dynamical characteristic of the band tails 
382: states and is compared with a static property, the inverse participation 
383: ratio of the same states. 
384: A simple linear fit reveals a strong correlation between the eigenvalue 
385: fluctuation and the corresponding inverse participation ratio for the states. 
386: The correlation is found to be as high as $\approx$ 0.95 for T=150K and 
387: 300K and falls to $\approx$ 0.8 at high temperature. The value of the 
388: correlation coefficient for different temperature is indicated in the 
389: Fig.\,\ref{fig2}. Once again, we see that the result is in accordance with our prediction in 
390: section II and provides a simple physical picture for having a large 
391: electron-phonon coupling for the localized states. 
392: 
393: %%% On thermal broadening on EDOS and Aljishi et al. experiment works.
394: 
395: In Fig.\,\ref{fig3}, we have plotted the time averaged electronic density 
396: of states for four different temperature in order to study the effect of 
397: thermal disorder on the tail states. It is quite clear from the figure 
398: that the effect of thermal broadening is quite significant on both sides of 
399: the gap. Photoelectron spectroscopic studies on \emph{a}-Si:H by Aljishi et 
400: al.\,\cite{Aljishi} have shown that the conduction tail is indeed more 
401: susceptible to thermal disorder than the valence tail. The temperature 
402: dependence can be conveniently expressed by introducing a characteristic 
403: energy $\rm E_0$ and fitting the electronic density of states to 
404: $ \rm \rho(E) \approx \exp(|E-E_f|/E_o(T))$. Aljishi et al.\,expressed 
405: the temperature dependence of the tail states by the slope of the 
406: $\rm E_o(T)$ vs. T plot and obtained a smaller value for the conduction 
407: band tail. We have observed a qualitative agreement of our results with 
408: experiment. The key observation that one should note from Fig.\,{\ref{fig3} is 
409: the following: the shape of the tail in the conduction band rapidly changes 
410: as the temperature rises from 150K to 700K. The corresponding change in the 
411: valence tail for the same range of energy (0.4eV) is however much less and is rather 
412: smooth compared to the conduction tail. Since the localized defect states 
413: (coming from the two dangling bonds) have been removed before plotting, this 
414: observation qualitatively suggests that the conduction tail states are more 
415: susceptible to lattice motion. It is tempting to attempt to estimate decay 
416: parameters for a direct comparison to experiment\cite{Aljishi}, but the sparse
417: sampling of tail states for this 216-atom model makes this a dangerous exercise. The
418: basic features do appear to be represented in our study, however.
419: 
420: 
421: \section{CONCLUSION}
422: Using accurate methods and a reasonable model of a-Si, we 
423: showed that there is 1) a large e-p coupling for localized 
424: states, 2) a significant correlation between thermal fluctuation 
425: of electron energy eigenvalues conjugate to localized states 
426: and the IPR of the model at T=0, 3) We find qualitative 
427: agreement with photoemission experiments\cite{Aljishi}, 4) we 
428: provide a simple
429: analytic argument for the origin of these effects. Identical 
430: experience with models of other amorphous materials has convinced us
431: that the results are correct in at least a qualitative 
432: way for binary glasses and amorphous materials, and perhaps other 
433: systems beside.
434: \begin{acknowledgments}
435: We thank the National Science Foundation for support under 
436: grants  DMR-0205858 and DMR-0310933. We thank Dr. J. C. Phillips
437: for helpful conversations and pointing out the larger significance
438: of the results. We thank Normand Mousseau 
439: for sending us the model of amorphous silicon used in this 
440: calculation. 
441: 
442: \end{acknowledgments}
443: 
444: \begin{thebibliography}{99}
445: 
446: \bibitem{Bardeen}
447: J.~Bardeen, L.~N.~Cooper, J.~R.~Schrieffer, Phys.~Rev. {\bf 108}, 1175 (1957).
448: 
449: \bibitem{Phillips} J. C. Phillips, Phys. Rev. Lett. 
450: {\bf 72} 3863 (1994); ibid. {\bf 59} 1856 (1987).
451: 
452: \bibitem{Bednorz}
453: J.~G.~Bednorz and K.~A.~Mueller, Z.~Phys.~B {\bf 64}, 189 (1986).
454: 
455: \bibitem{Staebler}
456: D.~Staebler and C.~R.~Wronski, Appl.~Phys.~Lett. {\bf 31}, 292 (1977).
457: 
458: \bibitem{Drabold1}
459: D.~A.~Drabold and Jun~Li, {\it Amorphous and heterogeneous silicon 
460: based films}, Materials Research Society Proceedings. Vol.~{\bf 715}, 
461: 2002.
462: 
463: \bibitem{Li} Jun Li and D. A. Drabold, Phys. Rev. B {\bf 68} 033103 (2003).
464: 
465: \bibitem{Cobb}
466: M. Cobb and D. A. Drabold, Phys.~Rev.~B 56, 3054 (1997).
467: 
468: \bibitem{Drabold2} The IPR ({\cal I}) is in fact a useful but {\it ad hoc} measure of localization 
469: (a more fundamental gauge is the information entropy). From a practical point of 
470: view however, the IPR usually gives a result qualitatively similar to the entropy.  
471: See D. A. Drabold, P. Biswas, T. DeNyago and R. Atta-Fynn, in {\it Non-crystalline 
472: Materials for Optoelectronics}, M. Popescu and G. Lucovsky Ed., INOE, Bucharest, 2004; 
473: cond-mat/0312607.
474: 
475: \bibitem{Drabold3}
476: D.~A.~Drabold, P.~A.~Fedders, Stefan~Klemm and O.~F.~Sankey, 
477: Phys.~Rev.~Lett. {\bf 67}, 2179 (1991); 
478: D.~A.~Drabold and P.~A.~Fedders, 
479: Phys.~Rev.~B {\bf 60}, R721, (1999); D.~A.~Drabold, J.~Non-Crys.~Sol. 
480: {\bf 266}, 211 (2000).
481: 
482: \bibitem{Attafynn} 
483: R. Atta-Fynn, P. Biswas, P. Ordej\'on and D. A. Drabold, Phys. Rev. B (In press); cond-mat/032333
484: 
485: \bibitem{Feynman}
486: R.~P.~Feynman, Phys.~Rev.~{\bf 56}, 340 (1939)
487: 
488: \bibitem{Note1} 
489: Here we have only considered the $\Gamma$ point in our calculations 
490: which corresponds to ${\vec k}$=0. 
491: 
492: 
493: \bibitem{Barkema}
494: G.~T.~Barkema and N.~Mousseau, Phys.~Rev.~B {\bf 62}, 4985 (2000).
495: 
496: \bibitem{Wooten}
497: F.~Wooten, K.~Winer, and D.~Weaire, Phys.~Rev.~Lett. {\bf 54}, 1392 (1985).
498: 
499: 
500: \bibitem{Siesta1}
501: P.~Ordej\'on, E.~Artacho and J.~M.~Soler,
502: Phys. Rev. B {\bf 53}, 10441, (1996).
503: 
504: \bibitem{Siesta2}
505: D. S\'anchez-Portal, P. Ordej\'on, E. Artacho and J. M. Soler,
506: Int. J. Quantum Chem. {\bf 65}, 453 (1997)
507: 
508: \bibitem{Siesta3}
509: J.~M.~Soler, E.~Artacho, J.~D.~Gale, A.~Garc{\'{\i}}a,
510: J.~Junquera, P.~Ordej{\'o}n,
511: and D.~Sanchez-Portal, J.~Phys.~Cond.~Matter. {\bf 14}, 2745 (2002).
512: 
513: \bibitem{Harris}
514: J.~Harris, Phys.~Rev.~B {\bf 31}, 1770 (1985).
515: 
516: \bibitem{Perdew}
517: J.~P.~Perdew and A.~Zunger, Phys.~Rev.~B {\bf 23}, 5048 (1981).
518: 
519: \bibitem{Sankey}
520: O.~F.~Sankey and D.~J.~Niklewski, Phys.~Rev.~B {\bf 40}, 3979 (1989).
521: 
522: \bibitem{Note2}
523: For amorphous silicon single-$\zeta$ polarized basis consists of 
524: one \emph{s} and three \emph{p} and five \emph{d} valence electrons.
525: For details on the generation of the basis orbitals, see 
526: E. Artacho, D. S\'anchez-Portal, P. Ordej\'on, A. Garc\'{\i}a, 
527: and J. M. Soler, phys. stat. sol. (b) {\bf 215}, 809 (1999).
528: 
529: 
530: %\bibitem{Louie}
531: %M.~S.~Hybertsen and S.~G.~Louie, Phy.~Rev.~B {\bf 34}, 5390 (1986).
532: 
533: \bibitem{Aljishi}
534: S.~Aljishi, J.~D. Cohen, S.~Jin and L.~Ley, Phys. Rev. Lett. 
535: {\bf 64}, 2811 (1990)
536: 
537: \end{thebibliography}
538: \end{document}
539: 
540: