1:
2: %\documentclass[preprint,aps]{revtex4}
3: %\documentclass[aps,prl,preprint,groupedaddress]{revtex4}
4: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
5:
6:
7: %\newcommand{\mwcd}[1]{\marginpar{\protect\scriptsize \protect\setlength
8: % {\baselineskip}{8pt} to editor- MS revision: #1}}
9: %\setlength {\marginparwidth}{7.0cm}
10:
11: \usepackage{graphicx}% Include figure files
12:
13: \begin{document}
14:
15: \title
16: {The static and dynamic conductivity of warm dense Aluminum and Gold
17: calculated within a density functional approach.
18: }
19:
20: %
21: \author
22: {
23: M.W.C. Dharma-wardana}
24: %
25: \affiliation{
26: Institute of Microstructural Sciences,
27: National Research Council of Canada, Ottawa,Canada. K1A 0R6
28: }
29: \email[Email address:\ ]{chandre.dharma-wardana@nrc.ca}
30: %\date{13-feb-2002}
31: \date{\today}
32: %
33:
34: \begin{abstract}
35: The static resistivity of dense Al and Au plsmas are
36: calculated where all the needed inputs are obtained
37: from density functional theory (DFT). This is used
38: as input for a study
39: of the dynamic conductivity.
40: These calculations
41: involve a self-consistent determination of (i)
42: the equation of state (EOS) and the ionization balance,
43: (ii) evaluation of the
44: ion-ion, and ion-electron pair-distribution functions.
45: (iii) Determination
46: of the scattering amplitudes, and finally the conductivity.
47: We present data for
48: for the static resistivity of
49: Al for compressions 0.1-2.0, and in the
50: the temperature range
51: $T$ = 0.1 - 10 eV.
52: Results for Au in the same
53: temperature range and for compressions 0.1-1.0 is also given.
54: In determining the dynamic conductivity for a range of frequencies
55: consistent with standard laser probes, a knowledge of the
56: electronic eigenstates and occupancies of Al- or Au plasma
57: become necessary. They are calculated using a neutral-pseudoatom model.
58: We examine a number of first-principles approaches
59: to the optical conductivity, including
60: many-body perturbation theory,
61: molecular-dynamics evaluations, and simplified
62: time-dependent DFT. The modification to the
63: Drude conductivity that arises from the presence of
64: shallow bound states in typical Al-plasmas is examined and
65: numerical results are given at the level of the Fermi Golden rule
66: and an approximate form of time-dependent DFT.
67: \end{abstract}
68: \pacs{PACS Numbers: 52.25.Mq, 05.30.Fk, 71.45.Gm}
69: %\vspace{0.5in}
70: %\hspace{0.5in} see file /usr/people/chandre/tekstuff/xc2d/ms1.tex
71: %
72: \maketitle
73: %
74: \section{Introduction}
75: The static conductivity of matter in the plasma state can
76: be calculated with some confidence, at least for a number of
77: ``simple'' plasmas, using an entirely first-principles approach
78: \cite{per87,eos95}. Comparisons with experiments are
79: available for a wide range of conditions,
80: at least for Al plasmas
81: \cite{benageR}.
82: Another transport property of great interest
83: is the frequency-dependent
84: conductivity $\sigma(\omega)$.
85: Optical probes provide a convenient diagnostic tool for this
86: plasma conductivity, and common
87: laser-probe wavelengths, e.g., 300-800 nm, become the window
88: of interest. In fact, many experiments on laser-plasma interactions
89: related to inertial fusion have concentrated on frequency tripled
90: (with the fundamental $\omega$ at 1$\mu$m from glass lasers)
91: light, $3\omega$, 351nm, while there is also current interest
92: in the $2\omega$ (527 nm) regime as an option for
93: indirect-drive ignition\cite{niemann}.
94: Normally, if the probe frequency is less than the
95: plasma frequency, the probe photons fail to enter the material
96: and only weakly ionized systems can be accessed. However, the
97: recent development of extremely thin ("nanoscale")
98: plasma-slab techniques\cite{ng}, known as idealized-slab plasmas\cite{ngcdw}
99: has made it possible to study dense plasmas with low-frqeuency probes,
100: both in transmission and in reflectivity.
101:
102: Density-functional calculations are of two types.
103: The first type depends heavily on quantum MD
104: (QMD) simulations, e.g, Car and Parrinello
105: treat {\it only the electrons} via Kohn-Sham theory,
106: while the
107: ion subsystem, explicitly represented by a convenient number $N$
108: (of the order of $\sim$ 32-256) of ions, is made to evolve in time
109: and an average over millions of configurations is taken. In full
110: Quantum Monte Carlo (QMC) simulations\cite{ceperley81,kwon}, even
111: Kohn-Sham theory is not
112: used and hence full QMC is not practical for typical plasma problems.
113: The second type of DFT is typified by
114: our approach where both {\it electrons and ions} are treated by DFT, so
115: that both subsystems are described by two
116: coupled ``single-electron'' and ``single-ion'', Hartree-like
117: Kohn-Sham
118: equations. The many-body effects (i.e, many-electron and many-ion
119: effects) are included through the exchange-correlation or correlation
120: potentials. This allows an enormous simplification
121: in the numerical work involved.
122: These equations may be further reduced to extremely simpilified
123: Thomas-Fermi approaches yielding approximate results for
124: plasma properties which are within an order of magnitude of the more
125: refined results, even in unfavourable cases\cite{leemore}.
126:
127: The general optical conductivity problem
128: is essentially the same as that of opacity calculations for
129: warm dense matter\cite{gri}.
130: However, here we approach it from the
131: static-conductivity ($\omega=0$) regime, and take account of
132: free-free processes as well as
133: shallow bound states which may exist.
134: The
135: static-conductivity provides an
136: evaluation of the dynamic conductivity near $\omega$ = 0
137: via the basic Drude formula which assumes
138: a constant relaxation time $\tau$, taken to be the {\it static} collision
139: time $\tau(0)$. Hence our task can be stated as follows:
140: \begin{enumerate}
141: \item
142: {Calculation of the static resistivity. This involves
143: the following steps:}
144: \begin{enumerate}
145: \item
146: {Calculate the Kohn-Sham atom in an electron gas of density $n$.}
147: \item
148: {Use the Kohn-Sham results to form pseudopotentials $V_{ie}$ and
149: scattering cross sections at the given density $\rho$ and
150: temperature $T$.}
151: \item
152: {Use the $V_{ie}$ to form pair potentials and pair-distribution
153: functions. Here the Kohn-Sham equation for the ion-subsystem
154: can be approximated by some form of the
155: hyper-netted-chain equation (HNC) with bridge terms.}
156: \item
157: {Calculate the EOS, ionization balance etc., to obtain
158: effective ionic charges $\bar{Z}$, electron density
159: $n=\rho/\bar{Z}$etc., and
160: self-consistently, repeating from item (a).}
161: \item
162: {Calculate the static resistivity and the static
163: relaxation time $\tau(0)$
164: using, e.g, using a Ziman-type
165: formula valid for strong coupling and finite $T$.}
166: \end{enumerate}
167: \item
168: {Use the energy-level structure of the Kohn-Sham atom
169: for the $\rho$
170: and
171: $T$ regimes of interest to set up bound-bound, or bound-free
172: processes which fall within the range of frequency $\omega$ considered.
173: Corrections may be necessary since the Kohn-Sham theory does not
174: provide good excitation energies.
175: }
176:
177: \item
178: {Construct the dynamic conductivity $\sigma^0(\omega)$.
179: This yields results equivalent to the
180: Fermi golden rule.
181: }
182: \item
183: { Extend the calculation of $\sigma(\omega)$
184: using time-dependent density-functional
185: theory\cite{gri}.Include the coupling of
186: electronic transitions to ion-dynamics.}
187: \end{enumerate}
188:
189: While some aspects of this program can be carried out,
190: it is fair to say that a complete, consistent theory
191: of the transport properties of such many-body systems as posed by warm dense
192: matter, or indeed, even the static properties as embodied
193: in the equation of state {\it for some regimes of density and temperature}
194: are open to debate, even for well studied systems like
195: aluminum\cite{eosb}, and hydrogen\cite{hyd}.
196: The objective of the present study is to calculate the dynamic conductivity
197: of Al and Au plasmas for a number of compressions and densities where
198: the Drude theory may need correction due to the presence of
199: shallow bound states. Such boundstates ionize when the compression is
200: changed, and produce distinctive changes in transport properties.
201: Sharply rising static resistivities
202: under a change of compression is a common feature of some of the theoretical
203: calculations shown here. However, the establishment of a genuine phase
204: transition requires more care\cite{eos95}. The $\sigma(\omega)$ of expanded
205: liquid metals and plasmas show\cite{bhat} effects arising from clustering
206: and excitonic effects, as the metal-insulator transition is approached.
207: These excitonic effects are not important in dense systems.
208: % linearization of time-dependent density functional
209: %theory is used to include dynamic screening from electrons, and
210: %coupling to ion-dynamics.
211: While Al has been an object of extensive study,
212: recent experiments in the warm dense matter regime
213: have focused on gold targets\cite{japmore,aung}. Here we present
214: numerical results for Al and Au for
215: several compressions and temperatures.
216: %
217: \section{static resistivity}
218: %
219: We use atomic units (Hartree=1 a.u., the Bohr
220: radius $a_0$ = 1, with $|e|=\hbar=m_e=1$).
221: The atomic unit of resistivity, given by $\hbar a_0/e^2$
222: has a value of 21.74 $\mu\Omega$cm.
223: If the equilibriation of the electron distribution
224: perturbed by the applied electric field
225: is governed by a relaxation time $\tau$, the conductivity
226: $\sigma$ is given as:
227: \begin{equation}
228: \sigma=\frac{\omega_p^2}{4\pi}\,\tau
229: \end{equation}
230: A mean free path $l_{mfp}$ = $<v>/\tau$,
231: where $<v>$ is some characteristic mean velocity, is often
232: introduced. If electrons
233: were classical {\it point} particles,
234: then it is evident that $l_{mpf}$ cannot be
235: smaller than the mean separation between collision centers. This is
236: sometimes called the Joffe-Regel-Mott rule, and holds well in
237: many semiconductors. However, electrons
238: are quantum particles (wave packets)
239: with an extension of the order of the thermal de Broglie
240: length.
241: Further, $\tau$ depends on the electron momentum $k$. Thus
242: there are examples where
243: $l_{mpf}$, obtained by some averaging process, is in fact smaller than
244: some estimated ``mean-ion separation''.
245: Although the concept of the mean free path is implicit in the
246: Boltzmann-equation approach, this is not necessary
247: in the Dyson equation which replaces
248: the Boltzmann equation in the quantum case.
249: If the mean free path is large, the ``particle picture'' of the electron
250: applies, while if this is comparable to the lattice parameter,
251: then we are in the diffraction limit and the wave picture
252: must be used. The Dyson equation, valid at both extremes, describes the
253: one-particle propagator which is closely related to the distribution function
254: appearing in the Boltzmann equation.
255: The derivation of a transport coefficient
256: from the Dyson equation leads us to the current-current correlation function.
257: This is closely related to various two-body distributions and collision
258: kernels found even in classical kinetic models.
259: Unfortunately, although formal expressions
260: can be written down, their evaluation using Green's functions or related
261: methods becomes impractical, especially if bound states are present.
262: In effect, Green's-function methods can be pushed
263: to, say, second order in the screened interactions. Such
264: an approach is sufficient if there are no boundstates associated with
265: the scattering potential. Attempts to go further
266: rapidly
267: become intractable and useless.
268:
269: Our approach is to use a variety of techniques
270: and replace the ion-electron interactions by
271: suitably constructed pseudopotentials, or use scattering cross sections
272: calculated from phase shifts. This requires a
273: fairly sophisticated non-perturbative description of the
274: ion-ion and ion-electron correlations in the plasma.
275: %
276: \subsection{Description of the plasma using the Kohn-Sham equations}
277: %
278: We begin with the bare nuclei and construct the
279: electronic and ionic structure of the plasma.
280: The interaction of the nucleus with the electron fluid is a highly nonlinear
281: process and attempting to treat it using perturbation theory
282: is unfruitful. Hence we use the
283: Kohn-Sham technique, and construct the non-linear charge density
284: around the nucleus. The nucleus, together with its
285: charge cloud of bound
286: states and continuum-electron states
287: constitute a neutral object. This neutral object is called
288: the {\it neutral pseudo-atom} (NPA), following the
289: usage of e.g., Ziman and Dagens\cite{dagens}.
290: Thus an important result of the
291: Kohn-Sham procedure is the charge-density ``pileup'', $n(r)$,
292: around the nucleus that essentially screens the nucleus.
293: A part of this arises from the free electrons and is denoted by $n_f(r)$.
294: This $n(r)$ and $n_f(r)$ depend on the mean density ${\bar n}$ of the electron fluid,
295: temperature $T$,
296: and the nuclear charge $Z$.
297: The Kohn-Sham procedure leading to the
298: NPA provides the phase shifts suffered by the continuum electrons when
299: they scatter from the nuclei. These are used
300: for constructing the scattering cross sections
301: (or pseudopotentials) which describe the electron-ion interaction.
302: The pseudopotential has an effective ionic charge $\bar{Z}$ and
303: it behaves as $-\bar{Z}/r$ for large $r$. The rapid oscillations of the
304: potential near the nucleus is replaced by a weak, smooth core region
305: as the ``valence'' electrons do not really penetrate the ion-core.
306: The pseudpotentials used in many of the solid-state or molecular
307: code packages\cite{codes}
308: have the necessary transferability and could be quite useful. However,
309: they assume that the pseudopotentials
310: would be used within a Schrodinger or Kohn-Sham type
311: procedure rather than in a linear-response scheme, and hence
312: they cannot be
313: directly used within a Ziman-type formula.
314:
315: Sometimes, instead of using the all-electron Kohn-Sham equation or
316: using a suitably constructed pseudopotential,
317: the electron-ion interaction is
318: replaced by a Yukawa-type interaction (effectively, a Debye-screened
319: interaction). The electronic structure is calculated for such a
320: static-screened nucleus. This procedure is not justifiable since
321: the energies of bound-state electrons correspond to very high frequencies
322: at which there is no screening. Further, the orthogonality of the
323: continuum eigenstates and the bound states ensures that there is very little
324: penetration of the free electrons into the bound-electron region.
325: Consequently, there is very little screening of the inner bound states,
326: where as the Yukawa potential screens the inner bound states as well.
327: Thus
328: calculations using a Debye-like potential in warm dense matter
329: is likely to be
330: incorrect irrespective of
331: whether the Born approximation, or a T-matrix, etc., were used.
332:
333: In the case of Al and also Au, it is
334: possible to construct, in many situations, a soft
335: pseudopotential $V_{ie}(q)$
336: which is
337: weak in the sense that it is possible to recreate the non-linear
338: electron-density ``pileup'' $n_f(r)$, obtained via the Kohn-Sham equation, by
339: within {\it linear} response theory.
340: That is, we {\t define} the $V_{ie}(q)$ such that
341: \begin{equation}
342: n_f(r)=-V_{ie}(q)\chi(q)
343: \end{equation}
344: Here $\chi(q)$ is the electron linear-response function.
345: Unlike the transferable pseuodpotentials used
346: in {it ab initio} packages, this $V_{ie}(q)$ is specific to the
347: chosen $Z$, $\bar{n}$, $T$, and
348: the atomic number $Z$. It is often convenient to write the
349: pseudopotential in the form
350: \begin{equation}
351: V_{ie}(q)=\bar{Z}M_qv(q), \; v_q=4\pi/q^2
352: \end{equation}
353: where $v_q$ is the bare Coulomb potential and $M_q$ is a
354: form factor. Only a local pseudopotential is used, and this is
355: quite adequate for an analysis of the experimental data
356: currently available.
357: Since the pseudopotential is weak by construction, the ion-ion
358: pair potential can be taken to be
359: \begin{equation}
360: U_{ii}(q)=4\pi\bar{Z}^2/q^2 + |V_{ie}(q)|^2\chi(q)
361: \end{equation}
362: Given $U_{ii}(q)$,
363: the ion-ion
364: distribution function $g_{ii}(r)$ and the structure factor $S(k)$
365: can be calculated using the
366: Hyper-netted-chain (HNC) equation or its extension where a bridge
367: term is included. We note (see below) that the HNC equation
368: (or its extension) is in fact the Kohn-Sham equation for classical
369: particles ( here, Al$^{z+}$ or Au$^{z+}$ ions ) for certain choices of
370: the ion-ion correlation
371: potential (there is no exchange potential because the ions are classical
372: particles).
373:
374: To summarize, by using the Kohn-Sham procedures, we have thus obtained
375: the ion-electron pseudopotential $V_{ie}(q)$, the structure factor
376: $S_{ii}(q)$, and the charge density $n_{ie}(r)$, and an effective ionic
377: charge $\bar{Z}$ which enters into the pseudopotential. The
378: phase shifts have been used to construct a
379: scattering cross section\cite{per87} which
380: may be used instead of the pseudopotential.
381: Hence we have a completely self-consistent procedure for obtaining
382: all the relevant quantities starting from the nuclear charge of the element.
383: The numerical codes for carrying out these procedures
384: are available via the internet, to any interested
385: researcher\cite{web}.
386:
387: \subsection{Is this a ``one-center'' approach ?}
388: To answer this question, we
389: consider the density-functional theory of a two component system
390: consisting of electrons (density profile $n(r)$ ), and ions, with a
391: density profile $\rho(r)$, with respect to an ion positioned at the
392: origin\cite{ilciacco}.
393: Then the Hohenberg-Kohn-Mermin theorem states that the free energy
394: $F[n(r),\rho(r)]$ is a functional of the density distributions such that:
395: \begin{eqnarray}
396: \partial F[n(r),n(r)]/\partial n(r)&=&0\\
397: \partial F[n(r),\rho(r)]/\partial \rho(r)&=&0.
398: \end{eqnarray}
399: The first of these equations leads to an effective
400: {\it single-electron} equation,
401: viz., the Kohn-Sham equation
402: where the effective potential contains an exchange-correlation potential
403: which takes account of many-electron effects. The second equation also leads
404: to a Kohn-Sham equation which is a classical equation for a {\it single ion}.
405: This also contains
406: an ion-ion correlation potential
407: which brings in the effects of the multi-centered
408: system. It can be seen that this classical Kohn-Sham
409: equation reduces to the
410: HNC equation for a certain choice of the ion-correlation potential.
411: Further, for ``simple'' metallic plasmas like Al where the pseuopotential is
412: weak, this
413: scheme relates closely to pair-potential based
414: liquid-metal theory. Such a simplification does not hold,
415: for e.g., for hydrogen
416: plasmas\cite{hyd}. Since ion-ion, ion-electron and electron-electron
417: correlations are included in the theory, it is {\it not} a
418: single-ion model of the
419: plasma. It is firmly rooted in a many-electron, many-ion DFT approach
420: which does {\it not} invoke the Born-Oppenheimer approximation.
421: Since the theory
422: is explicitly based on distribution functions, it is manifestly non-local
423: and can be
424: easily implemented to be free of electron self-interaction errors.
425: Our approach may be contrasted with the Car-Parrinello (CP)
426: approach where DFT
427: is
428: used only for the electrons, while the ions are explicitly and individually
429: treated by classical molecular dynamics.
430: CP avoids the need for an ion-correlation potential, but demands
431: a much larger computational effort. Some of these efforts, based on MD
432: methods (e.g, Ref.~\cite{kwon,surh}) have confirmed results obtained
433: by the numerically simpler methods that we have used.
434: %
435: \subsection{Extended Ziman formula for strongly-coupled
436: electrons and ions.}
437: The Ziman formula is an application of the Boltzmann equation to
438: liquid metals.
439: It was extended to finite
440: temperatures by a number of authors.\cite{geoffry}
441: The crux of the problem is the evaluation of the
442: collision rate. Compared to some methods well known in plasma theory
443: (e.g, Lenard-Balescu) the collision rate is easily evaluated using the
444: ``Fermi golden rule''. In this section we briefly
445: recapitulate Ziman theory
446: in the language of the Fermi golden tule.
447: In the relaxation-time approximation we
448: assume that the perturbed Fermi distribution $f(k)$ for electrons
449: with momentum $\vec{k}$ relaxes towards the
450: equilibrium distribution $f_0(k)$ according to the equation:
451: \begin{equation}
452: \label{tau}
453: -\frac{\partial f}{\partial t}\big|_{col}=\frac{f(k)-f_0(k)}{\tau(k)}
454: \end{equation}
455: Considering an electron scattered elastically from state $\vec{k}$
456: to state $\vec{k'}$, with $|k|=|k'|$, $\epsilon_k=\epsilon_{k'}$, the net
457: scattering rate is the difference of the
458: two processes $(\vec{k}\to \vec{k'}) - (\vec{k'}\to \vec{k})$.
459: The initial and final densities of states for the $\vec{k}\to \vec{k'}$
460: process is $f(k)$ and $[1-f(k')]$. Hence the Fermi Golden rule gives
461: \begin{eqnarray*}
462: R(k\to k') &=& (2\pi/\hbar)\sum|T_{kk'}|^2
463: \delta(\epsilon_k-\epsilon_{k'})f(k)(1-f(k')\\
464: R(k'\to k) &=& \mbox{ permutation of $k$ with $k'$ etc.,}
465: \end{eqnarray*}
466: Since $\epsilon_k=\epsilon_{k'}$, this involves only an angular
467: integration and $|k|=|k'|$. Since the energy is not changed, the
468: static resistivity
469: arises {\it purely from momentum randomization}.
470: The change in momentum is
471: $q^2=2k(1-cos(\theta))$.
472: Here $\theta$ is the angle between $\vec{k'}$ and $\vec{k}$.
473: This $(1-cos(\theta))$ term does not
474: appear in the usual relaxation time which is the time between scattering
475: events. Using these rates in Eq.~\ref{tau},
476: we obtain a result
477: for the {\it inverse} of the relaxation time.
478: \begin{equation}
479: \frac{1}{\tau(k)}=2\pi\sum\delta(\epsilon_k-\epsilon_k')|T_{kk'}|^2
480: (1-cos(\theta))
481: \end{equation}
482: Here the sum merely indicates the integration over $\theta$.
483: The ``$T$-matrix'' appearing here describes the scattering of an electron
484: by the whole ion-distribution (i.e, not just one ion).
485: Given a set of ions at instantaneous positions
486: $\vec{R_I}$, then the interaction of an electron at $\vec{r}$
487: with the whole distribution is of the form:
488: \begin{equation}
489: V(r)=\sum_I V_{ie}(\vec{r}-\vec{R}_i)
490: \end{equation}
491: The matrix element between the initial state
492: $\vec{k}$ and the final state $\vec{k'}$,
493: with $\vec{q}$ = $\vec{k'} - \vec{k}$ is:
494: \begin{equation}
495: V(q)=\sum_I \bar{Z}M_q v_q e^{iq\cdot\vec{R_I}}.
496: \end{equation}
497: Note that
498: \begin{eqnarray}
499: \rho_q&=&\sum_I exp(i\vec{q}\cdot\vec{R_I})\\
500: <\rho_q\rho_{-q}>&=&N_iS_{ii}(q).
501: \end{eqnarray}
502: Thus the ion-ion structure factor $S(q)$ and the single-ion
503: scattering cross section (or the pseudopotential)
504: from a single ion combine to give the full scattering $T$-matrix.
505: The dependence on the
506: structure factor becomes negligible for $T> 10$ eV. The
507: individual scattering cross section can be replaced by a single-center
508: $T$-matrix (to be denoted by $t_{kk'}$) obtained from the phase shifts of the
509: NPA calculation for a single nucleus. The $S(q)$ is also obtained from the
510: pair-potential constructed
511: from the same NPA calculation. In Ref.~\cite{per87} we showed how
512: to avoid the calculation of the $S(q)$ by directly computing the
513: scattering cross section from the {\it whole ion distribution}. That is,
514: $T_{kk'}$ s not factored into single-center $t_{kk'}$ and the associated
515: structure factor. Such an approach is needed
516: for {\it strongly interacting systems} where such a
517: factorization may not be valid. In this context we
518: note that the resistivities calculated by us using the "single-center" model
519: (ie., $t_{kk'}S(q))$, and the full ion-distribution model ( $T_{kk'}$
520: for strong-scattering) for H-plasmas were independently confirmed by
521: the quantum Monte Carlo simulations of Kwon et al\cite{kwon}.
522: However, if the pseudopotential is weak, the $S(k)$ and the single-center
523: $t(k)$ may be used.
524:
525: Given the inverse relaxation time $1/\tau(k)$ for an electron of
526: momentum $k$, or equivalently, $1/\tau(\epsilon)$ for the
527: energy, $\epsilon = k^2/2$, we need to average this
528: over all electron energies to
529: obtain a resistivity or a conductivity. The averaging used in the Ziman
530: formula leads to a {\it resistivity}, while a direct application
531: of the Boltzmann equation would lead to a conductivity.
532: Thus,
533: \begin{equation}
534: \sigma = \frac{\omega^2_p}{4\pi}<\tau(\epsilon)>, \;\;
535: R =\frac{4\pi}{\omega^2_p}<1/\tau(\epsilon)>
536: \end{equation}
537:
538: The Boltzmann equation shows that the averaging relevant
539: to the conductivity calculation is
540: such that
541: \begin{equation}
542: \sigma=\frac{\omega^2_p}{4\pi}\frac{2}{3n_e}
543: \int \frac{d \epsilon}{\pi^2}(\surd{2}
544: \epsilon^{3/2})\frac{-\partial f_0(\epsilon)}{\partial \epsilon}\tau(\epsilon)
545: \end{equation}
546: On the other hand, the averaging over the $1/\tau(\epsilon)$ used in the
547: extended Ziman formula for the resistivity is somewhat different.
548: \begin{equation}
549: < 1/\tau> = -\int_0^\infty d\epsilon g(\epsilon)
550: \frac{\partial f(\epsilon)}{\partial \epsilon} \tau(\epsilon)^{-1}
551: \end{equation}
552: Here $g(\epsilon)$ is a density-of states factor which is unity for
553: free non-interacting electrons. It should be constructed from the
554: phase shifts in a
555: strongly scattering environment.
556:
557: The initial assumption, Eq.~\ref{tau},
558: was that the modified
559: distribution was defined via a relaxation time. A more complete
560: approach is to represent the modified part $\delta f(k)$ as a
561: series expansion in a set of suitably constructed orthogonal polynomials,
562: and obtain a variational solution. The set of polynomials appropriate
563: for degenerate (and partially degenerate) electrons has been discussed
564: by Allen.\cite{allenpoly}. In the classical limit, such polynomials are
565: the well known Sonine polynomials. The usual relaxation-type approach is
566: equivalent to a single-polynomial solution. This is
567: adequate for dense Al-plasmas, and for the range $0 < T < 10$ eV
568: studied here. However, this is probably not so for Al at 1/4 of the
569: normal density, or for lower densities. We have less experience
570: with Au-plasmas to assess the quality of the resistivities for Au
571: obtaind here.
572: %
573: \subsection{Numerical results in the static limit.}
574: The static resistivities for Al calculated using the above
575: methods (and some extensions of it)
576: have been compared with experiment by Benage et al.\cite{benageR}
577: Given the uncertainties in the experiment and the various approximations
578: in the theory, the agreement is quite good.
579: However, there are a number of difficulties in the calculation.
580: If we consider an Al-plasma at a compression $\kappa$=0.25, at $T> 2.5$ eV
581: the Al-ion has the bound shells $1s, 2s, 2p$, and $3s$. The $3s$ level
582: becomes increasingly shallow as the temperature is {\it reduced}. Below
583: approximately $T \sim 2.5$ eV, the $3s$ level begins to ``evaporate''
584: and becomes
585: free, i.e, the $3s$ bound electrons ionize.
586: Even at 2.5 eV, the occupation of the $3s$ level is $\sim$ 0.744.
587: The ionization state $\bar{Z}$ increases from
588: $\sim 1.667$ to 3. The model where we have a single average $\bar{Z}$ is
589: clearly not satisfactory in such a region. The plasma contains
590: an equilibrium mixture of several ionization states $\bar{Z_i}$ with
591: concentrations $x_i$ such that
592: \begin{equation}
593: \bar{Z}=\Sigma_i x_i\bar{Z_i}
594: \end{equation}
595: For such situations (and indeed in general), the concentrations $x_i$
596: have to be determined from the minimum of the total free energy. If
597: $n$ species of ions are possible, then $n$ different
598: Kohn-Sham equations have to be solved, and $n(n+1)/2$ ion-ion
599: distribution functions have to be determined, and the total
600: free energy has to be calculated as a function of $n$ and $x_i$.
601: The minimum property of the free energy yields the equilibrium
602: plasma conditions
603: from which the resistivity is calculated, using the $n$ scattering
604: cross sections and the $n(n+1)/2$ structure factors.
605: We reported such a calculation for $Al$
606: in Ref.~\cite{eos95}. Our experience is that, even with a multi-ion
607: fluid model, the self-consistent equations
608: may fail to converge
609: since the $3s$ level (or the $2p$ level)
610: affects the iterative procedure. Majumdar and Kohn have
611: shown that physical properties of a system should be continuous across
612: the region where an electronic state moves from being a bound state
613: to a continuum state\cite{majum}. Hence we may calculate the
614: resistivity in two adjacent regions separated by a non-convergent
615: region, and smoothly join the calculated resistivity
616: across the ``difficult'' region. In Table~\ref{K-S} we show the Kohn-Sham
617: eigenvalues of the bound-electron states for Al ions in
618: various
619: plasmas. Although the Kohn-Sham eigenvalues do not exactly
620: correspond to the excitation energies, they provide an initial
621: estimate which can be improved using the methods of time-dependent
622: density functional theory,
623: or using self-energy calculations\cite{dyson-mott}.
624: $\,$
625: %
626: \begin{table}
627: \caption
628: { Resistivity (in $\mu$ohm$\,$cm) of Al plsamsa as a function of the
629: compression $\kappa$ and the temperature in eV, calculated within the
630: approach described in the text.
631: }
632: \begin{ruledtabular}
633: \begin{tabular}{lcccccc}
634:
635: T$\downarrow \;\;\; \kappa\to$& 0.1 & 0.25 & 0.5 & 1.0 & 2.0 & 4.0 \\
636: \hline\\
637: 0.10 &1620 & 801 & 77 & 28 & 38 & 48 \\
638: 0.25 &1582 & 812 & 84.1 & 30.5 & 37.4 & 47.0 \\
639: 0.50 &1527 & 847 & 92.1 & 34.4 & 35.8 & 45.6 \\
640: 0.75 &1447 & 859 & 99.7 & 36.7 & 34.5 & 45.0 \\
641: 1.00 &1340 & 846 & 107 & 38.7 & 34.5 & 44.8 \\
642: 1.75 &1063 & 707 & 122 & 44.4 & 34.2 & 44.6 \\
643: 2.50 &844 & 597 & 135 & 50.2 & 35.4 & 44.7 \\
644: 3.75 &1196 & 448 & 148 & 60.4 & 39.5 & 46.0 \\
645: 5.00 &916 & 434 & 160 & 70.8 & 45.1 & 48.3 \\
646: 7.50 &646 & 380 & 180 & 90.6 & 57.7 & 54.7 \\
647: 10.0 &544 & 379 & 194 & 108. & 70.1 & 61.8 \\
648: \end{tabular}
649: \end{ruledtabular}
650: \label{restab}
651: \end{table}
652: %
653: %\vspace{0.25 in}
654: %
655: \begin{figure}
656: \includegraphics*[width=6.5 cm, height=8.0 cm]{alres3.ps}
657: \caption
658: {Resistivity of warm dense Aluminum as a function of
659: the compression for several temperatures
660: }
661: \label{fig1}
662: \end{figure}
663: %
664: %
665: \begin{table}
666: \caption
667: {
668: Kohn-Sham energy-level structure for several $Al$-plasmas within the
669: NPA
670: averag-configuration model }
671: \begin{ruledtabular}
672: \begin{tabular}{ccccccc}
673:
674: Level &$\kappa=4$& $\kappa=2$& $\kappa=1$&$\kappa=0.5$&$\kappa=0.25$&$\kappa=0.1$\\
675: \hline\\
676:
677: &T=10 &&&&&\\
678: $\bar{Z}$& 3.043 & 3.0166 & 3.0164 & 3.0194 & 2.6060& 2.2427 \\
679: 2s &-5.7016 &-6.3615 &-6.9213 &-7.4089 &-7.9601&-8.5825 \\
680: 2p &-2.9311 &-3.5922 &-4.1529 &-4.6411 &-5.1928&-5.8159 \\
681: 3s & -- & -- & -- & -- &-0.1871&-0.6168 \\
682: 3p & -- & -- & -- & -- & -- &-0.1643 \\
683: &&&&&&\\
684: &T=2.5 &&&&&\\
685: $\bar{Z}$& 3.0427 & 3.0051 & 3.0003& 3.0000 & 1.6643& 1.6240 \\
686: 2s &-5.5247 &-6.1819 &-6.6655&-7.0257 &-7.4435& -7.6634 \\
687: 2p &-2.7838 &-3.4422 &-3.9268&-4.2875 &-4.7055&-4.9258 \\
688: 3s & -- & -- & -- & -- &-0.1178&-0.2758 \\
689: &&&&&&\\
690: &T=0.1 &&&&&\\
691: $\bar{Z}$& 3.0437 & 3.0053 & 3.0003& 3.0000 &does not &does not \\
692: 2s &-5.4903 &-6.1444 &-6.6282&-6.9811 &converge &converge \\
693: 2p &-2.7462 &-3.4008 &-3.8853&-4.2393 &due to &due to \\
694: 3s & -- & -- & -- & & 3s level &3s level \\
695: \end{tabular}
696: \end{ruledtabular}
697: \label{K-S}
698: \end{table}
699: %
700: As the density decreases and the temperature increases,
701: we inevitably pass through regions of $T$, $\kappa$ where
702: the problem of bound states which hover near ionization
703: becomes important. We return to this question
704: in discussing the dynamic conductivity of Al-plasma
705: at $\kappa$ = 0.25,
706: and 0.1.
707: %
708: \begin{figure}
709: \includegraphics*[width=6.5 cm, height=8.0 cm]{aures.ps}
710: \caption
711: {Resistivity of warm dense gold as a function of
712: the compression for several temperatures
713: }
714: \label{fig1au}
715: \end{figure}
716: %
717: Table I shows that, for $\kappa$ = 0.1 the resistivity essentially
718: decreases with $T$, while for $\kappa$ = 0.25 the resistivity gradually
719: increases in value, goes through a plateau-like region,
720: and then begins to decrease with temperature.
721: The same behaviour holds for the other higher compressions, but the
722: table given here does not go high enough in temperature
723: (for the higher densities)
724: to show the
725: plateau effect and the onset of the decrease of $R$. Some authors
726: have interpretted the plateau in the resistivity as indicating a situation
727: where the mean-free path has become as small as possible ( as in the
728: Joffe-Regel rule). We {\it disagree} with this explanation of the
729: existence of a resistivity plateau in these cases in terms of a
730: saturation of the mean-free path. Quantitative agreement is
731: provided with a very {\it different picture}
732: of what happens in the plasma. As the
733: material is heated, its resistivity increases just as in a metal, due to
734: the increased availability of a strip of states (of width $k_BT$)
735: at the Fermi surface
736: for the ``phonon-like'' scattering to take place.
737: However, as $k_BT$ increases, the number of
738: current carrying electrons also increases, compensating the
739: resistivity increase.
740: During this process the chemical potential $\mu$ of the plasma electrons
741: begins to decrease, and
742: a temperature $T_{\mu=0}$ is reached when
743: $\mu$ passes through zero and towards negative values. Then the
744: Fermi sphere is gets broken down and from
745: then on, {\it all} the electrons,
746: and not just those near the Fermi surface,
747: begin to conduct. The plasma is essentially classical.
748: The resistivity {\it decreases}
749: as the temperature increases.
750: We are in fact in the Spitzer-like regime. The plateau defines
751: the transition to the Spitzer-like regime.
752: The Fermi energy of the $\kappa$ = 0.1 case is small and
753: it is already behaving like a classical plasma.
754: In reality the picture is
755: more complicated since the Fermi surface changes not only
756: because of the temperature,
757: but also because the ionization of bound electrons changes
758: the value of $\bar{Z}$.
759: This pushes the value of
760: $T_{\mu=0}$, and the onset of the plateau to higher temperatures
761: than in a model
762: with constant $\bar{Z}$. It is easy to allow for this in a calculation of
763: $T_{\mu=0}$ and confirm that the resistivity $R(T,\kappa)$
764: given in our
765: table (and in Ref.\cite{milsch}) is consistent with this picture.
766: %
767: \subsection{Contribution to the resistivity from electron-electron scattering.}
768: Discussions of the electrical resistivity of
769: plasmas sometimes contain
770: allusions to the e-e contribution to the electrical resistivity.
771: However, the
772: electron-current operator $\vec{j}=(e/m)\vec{k}$ commutes with the
773: electron-electron interaction
774: Hamiltonian $H_{ee}$.
775: \begin{equation}
776: \vec{j}H_{ee}-H_{ee}\vec{j}=0
777: \end{equation}
778: This shows that the current is {\it conserved}
779: under the e-e interaction.
780: Hence electron-electron interactions {\it cannot} contribute to
781: the resistivity arising from the electron current.
782: However, the e-e interaction has an indirect
783: effect since it screens the electron-ion pseudopotential
784: $V_{e-i}(q)$. That is, electron-ion vertices must occur
785: in all diagrams which contribute to the resistivity.
786: In perturbation-theory
787: approaches to the conductivity or resistivity,
788: it is quite easy to get a contribution
789: to $R$ from e-e scattering alone, if the theory is carried
790: out only to, say, second order.
791: This means, if an all-oder calculation were done, the higher
792: order corrections would exactly
793: cancel the low-oder result for the e-e scattering contribution.
794: Electron-electron interactions contribute to the resistivity of
795: solids where the periodic potential generates {\it Umkapp} scattering.
796: But this is not the case in plasmas if they can be considered
797: uniform to within a length scale significantly
798: larger than the mean free path. Classical transport calculations
799: for systems with gradients (i.e, no translational invariance)
800: is well known\cite{matt}. The quantum calculation is also
801: well known in transport across heterostructures\cite{lanbut}.
802: %
803: \section{Dynamic conductivity.}
804: If we apply a field $\vec{E}\, cos(\omega t)$ to the system,
805: say
806: using a light probe of frequency $\omega$, then the polarization of the
807: medium is described by the polarization function $\Pi(q,\omega)$ which is
808: directly connected with the transverse dielectric function
809: $\varepsilon(\omega)$ and the dynamic conductivity $\sigma(\omega)$. The
810: wavevector $q$ of the photon is nearly zero
811: and may usually be omitted.
812: The real part $\sigma_r$=$\Re\sigma(\omega)$ at $\omega\to 0$ reduces to the
813: static conductivity $\sigma$ that was already discussed.
814: \begin{eqnarray}
815: \varepsilon(\omega)&=&1-\frac{\omega^2_p}{\omega^2}-
816: \frac{4\pi}{\omega^2}\Pi(\omega) \\
817: \sigma(\omega)&=&i\frac{\omega^2_p}{4\pi\omega^2}
818: +i\frac{\Pi(\omega)}{\omega}
819: \label{sigmapi}
820: \end{eqnarray}
821: If the effect of interactions and bound states is small,
822: the conductivity of "free" electrons driven by the field
823: $E_0exp(i\omega t)$, and damped by scatterng at ion centers
824: is well approximated
825: by the Drude model. It uses a relaxation time $\tau_0$, (or a
826: damping parameter $\gamma_0$)
827: independent of the frequency. In partially ionized systems, or when there are
828: interband effects in solids, it is necessary to include bound-free
829: and bound-bond contributions to $\sigma(\omega)$. A useful practical form is
830: the extension of the Drude model where a model dielectric function is used.
831: Thus,
832: \begin{eqnarray}
833: \label{modeleps}
834: \varepsilon(\omega)&=&\varepsilon_r+i\varepsilon_i \\
835: &=&1-\frac{\omega^2_{p0}}{\omega(\omega+i\gamma_0)}
836: -\sum^{\lambda_{m}}_{\lambda=1}\frac{\omega^2_{p\lambda}}
837: {\omega(\omega+i\gamma_{\lambda})-\epsilon_{\lambda}}\\
838: \sigma_r(\omega)&=&\frac{\omega\varepsilon_i(\omega)}{4\pi}
839: \end{eqnarray}
840: Here the pure free-electron Drude term uses the damping parameter
841: $\gamma_0$ and a plasma frequency $\omega_{p0}$, while other processes
842: are modeled by a finite set of oscillators
843: parametrized by $\gamma_{\lambda}$ and
844: $\omega_{p\lambda}$. This is a form useful for fitting
845: experimental data since reflection and transmission
846: experiments could be used to
847: extract best fit values of $\gamma_{\lambda}$,$\omega_{p\lambda}$,
848: subject to the sum rule
849: (f-sum rule):
850: \begin{equation}
851: \int_0^\infty d\omega \sigma_r(\omega)=
852: \frac{1}{4}\sum^{\lambda_m}_{\lambda=0}\omega_{p\lambda}=\pi n_{tot} e^2/m_e
853: \end{equation}
854: Here $n_{tot}$ involves the free electrons
855: (given by the average charge $\bar{Z}$)
856: as well as the electrons occupying the localized
857: atomic states participating in the
858: optical transitions. Unlike in liquid-metal studies, these are not readily
859: available for warm dense partially degenerate plasmas. Further, the electron
860: populations in the bound and free states are linked by the Fermi distribution
861: associated with the electron temperature and the chemical potential.
862: Hence this LTE
863: (local thermodynmaic equilibrium) type constraint also should be imposed in fitting
864: the experimental data to the model dielectric function.
865: Thus the most fruitful approach
866: is to use the values of the mean ionization $\bar{Z}$, occupation numbers,
867: electron chemical potential
868: $\mu_e$, $\gamma_{\lambda}$, $\epsilon_{\lambda}$ etc.,
869: provided from the density-functional NPA calculation
870: in setting up the fitting process.
871: The simple Drude equation with a {\it fixed} value of $\gamma_0$ satisfies the
872: f-sum rule.
873: It can be shown that if we define $\Pi(\omega)$ in terms of the
874: non-interacting polarizability $\Pi^0(\omega)$, by the
875: form:
876: \begin{equation}
877: \Pi(\omega)=\Pi^0(\omega)
878: /[1+G(\omega)\Pi^0(\omega)]
879: \end{equation}
880: where $G(\omega)$ is called a ``local-field'' correction, then the sum rule
881: is satisfied if it holds for $\Pi^0(\omega)$. Thus, if
882: $\Pi(\omega)$ can be calculated, then we can
883: calculate the dynamic conductivity from it.
884:
885: The effect of corrections to the simple Drude term
886: arising from interband terms in
887: solid Aluminum, were examined many years ago by Dresselhaus, Harrison, and
888: by Ashcroft and Sturm.\cite{ash-sturm}. They showed that there are
889: band to band transitions near $\sim$.05 and 0.15 of
890: the Fermi energy $E_F$ (which is
891: $\sim 12$ eV. for normal density $Al$). The optical conductivity of
892: normal-density liquid Al
893: slightly above the melting point has been measured by Miller\cite{miller} and
894: shows a simple Drude form. A discussion of liquid metal data including
895: Cu, Ag and Au is found in Faber\cite{faber}.
896: The deviations from the Drude form seen in solid-Al arise from the
897: splitting of degenerate bands due to the crystal potential, and to
898: normal interband transitions. Benedict et al. have argued that the
899: thermal broadening of the electron self-energy is by itself sufficient to
900: ``wash out'' these solid-state effects, even if the ion lattice remained
901: intact\cite{benedict}, as may be the case in short-pulse laser
902: generated plasmas. The effect of electron-electron as well as ion-electron
903: interactions on the electron self-energy at finite-$T$ was also discussed by
904: us for the hydrogenic case\cite{dyson-mott}
905:
906: When we consider
907: warm dense $Al$-plasmas, the Kohn-Sham eigenvalues
908: can be used to assess if the
909: driving field $\omega$ could excite bond-bound or bound-free
910: processes. As seen from Table~\ref{K-S}, the $2s$ levels for the
911: compressed systems ($\kappa$ =0.1, 0.5, 1, 2, 4) occur
912: between -7.4 au to -5.5 au.,
913: while the $2p$ level ranges from -4.6 au. to -2.7 au. Hence, the Drude
914: formula would be reasonable for these systems and for standard
915: optical probes. The situation is quite different for
916: the $\kappa$ =0.25 case. Here
917: the $2p$ level ranges from $\sim$ -4.5 au. to -5.5 au., but the
918: $3s$ level is very close to the ionization threshold. At $T$ = 8 eV, the
919: $3s$ level is at -0.165 au, and rises to -0.105 au. near $T$= 3 eV, and
920: then completely disappears for $T$ below $\sim$ ~2.5 eV.
921: Since a 300-400 nm optical probes corresponds to about 4-3 eV,
922: it is clear that
923: such probes would show deviations from the Drude conductivity for the
924: $\kappa$ =0.25 case. When we go to low temperatures, the $\bar{z}=3$
925: ionization is very stable, while the $3s$ level creates the presence of
926: $\bar{z}=1,2$
927: ionization states. In the case of $\kappa$ = 0.1, we have {\it two}
928: shallow states , viz., $3s$ and $3p$. The conductivity derived from
929: the model dielectric function, i.e., Eq.~\ref{modeleps} would be relevant to
930: the representation of experimental data for such systems. In the following
931: we look at first-principles calculations.
932: %
933: \subsection{Perturbational approach to the dynamic conductivity.}
934: In a theory of $\sigma(\omega)$
935: we are in effect looking for the
936: polarization function $\Pi(\omega)$.
937: If the potential
938: $V_{ie}(q)$ is weak,
939: (no bound states) diagrammatic methods can be used. However,
940: the second-order expression in the screened interaction is essentially the
941: only one that is tractable for realistic potentials.
942: A number of
943: such quasi-second order results exist in the literature, derived using various methods.
944: An old result, due to Hopfield, holds if the structure factor is
945: essentially unity\cite{hopfield}.
946: \begin{equation}
947: \Re\sigma(\omega)=-\frac{\omega_p^2}{(4\pi\omega)^2}
948: \int \frac{d\vec{q}}{(2\pi)^3}q^2q_z|V_{ie}(q)|^2\Im[\frac{1}{\varepsilon(q,\omega)}]
949: \end{equation}
950: A more complete result, including the contribution from the
951: dynamic structure factor of the ions can be written down, using
952: an approach similar to that given by Mahan for phonons.\cite{mahan}
953: R\u{o}pke et al. have also discussed second-order expressions
954: within the Zubarev approach which is most suitable for short-ranged
955: potentials free of bound states.\cite{ropke} The approach
956: used by Mahan is more interesting and may be used for
957: long-range potentials.
958: In fact, even in the two-temperature case where the ions are at a
959: temperature $T_i$, while the electrons are at a temperature $T_e$,
960: it is easy to show using the Keldysh technique
961: (assuming that the probe
962: frequency $\omega$ does not overlap with core transitions)
963: that the frequency dependent collisional relaxation time $\tau(\omega)$
964: is\cite{milsch},
965: \begin{eqnarray}
966: \label{cond}
967: \lefteqn{\tau^{-1}(\omega)=-(\omega^2_p\omega)^{-1}\int\frac{q^2dqM^2_q}{(2\pi)^3}
968: q_z^2\int\frac{d\nu}{2\pi}}\cdot \\
969: & &\Im[\chi_{ii}(q,\nu)]\Im[\chi_{ee}(q,\nu+\omega)]
970: \Delta N(T_e,T_i,\nu,\omega) \nonumber
971: \end{eqnarray}
972: where we have set
973: \begin{equation}
974: \Delta N(T_e,T_i,\nu,\omega)=[N\{\beta_e(\nu+\omega)\}
975: -N\{\beta_i(\nu)\}].
976: \end{equation}
977: Here $N\{\beta_i(\nu)\}$ is a Bose factor giving the occupation number
978: of density-fluctuation modes at the temperature $T_i$ and energy $\nu$.
979: For the equilibrium situation we simply set $T_i=T_e$.
980: Then, as $\omega\to 0$, this equation can be shown to reduce to the
981: inverse collision time used in the Ziman formula.
982: In Eq.~\ref{cond} the electron response $\chi_{ee}$
983: and the ion-response $\chi_{ii}$
984: mediate the energy and momentum exchange between the two
985: subsystems. The
986: imaginary parts of the response functions are related to the
987: dynamic structure factors by a relation of the form:
988: \begin{equation}
989: S(k,\omega)=-\frac{1}
990: {2\pi}\coth\big[\frac{\omega}{2T}\big]\Im\chi(k,\omega)
991: \end{equation}
992: In the case of the electrons, $\chi(k,\omega)$ can be written down in
993: terms of the Lindhard function $\chi^0(k,\omega)$ and the local-field
994: correction $G(k,\omega)$. In the limit where the electrons becomes
995: classical, the Lindhard function simply reduces to the Vaslov
996: function.
997: The dynamic structure factor of the ions can also be modeled
998: in a similar fashion.\cite{elr} Clearly, if the
999: probe frequency $\omega$ is smaller than the electron plasma frequency,
1000: then we may make a static approximation for $G(k,\omega)$.
1001: Since normal-density Al (and even Au, depending on the compression)
1002: have high plasma frequencies, this is
1003: a good approximation for dense Al or Au. However,
1004: the static $G(k,\omega=0)$ used should be such
1005: that at least the compressibility sum rule is satisfied.
1006: Also, we can introduce an $\omega$ dependent model
1007: $G(k,\omega)$ with $G(k,\omega)$= $G(k,\omega=0)$ for $\omega<\omega_P$,
1008: and $G(k,\omega)$ fitted to the high-frequency moment sum rules
1009: for $\omega>\omega_P$. In practice it is not known how to
1010: satisfy all the
1011: sum rules.
1012: The static pair-distribution function $g(r)$
1013: recovered from
1014: the
1015: imaginary part of such a response function should also agree with the
1016: known $g(r)$ of the electrons at that density. The successful calculation
1017: of the $g(r)$, $S(k)$, and the $G(k)$ in a consistent manner for
1018: electrons at strong-coupling and arbitrary temperatures (i.e, arbitrary
1019: degeneracies) was presented
1020: recently in Ref.~\cite{chncT}
1021:
1022: Although an extension of the above equation to take account
1023: of bound-free and bound-bound electron processes can be written down
1024: ``by hand'', it is not easy to provide a systematic development.
1025: More fruitful approaches are to use density-functional molecular-dynamics
1026: simulations or
1027: time-dependent
1028: density functional theory. We consider these below.
1029: %
1030: \subsection{Conductivity via the Kubo-Greenwood formula and molecular dynamics.}
1031: Another approach to the conductivity $\sigma(\omega)$ is to
1032: use molecular dynamics to develop the ionic-liquid structure,
1033: while retaining
1034: a DFT approach only for the electronic structure\cite{silv1,alavi,gillan}.
1035: The
1036: ion subsystem is modeled with, say, typically
1037: 32-256 atoms in a simulation box
1038: of volume $V_b$ which is periodically repeated.
1039: An externally constructed pseudopotential is used and an ionization model is
1040: {\it assumed}. The
1041: required number of electrons based on the ionization model
1042: is placed in the box. The ions are held at some fixed ionic
1043: configuration \{$R_i$\} and the Kohn-Sham electronic
1044: wavefunctions $\psi_f(r,\{R_i\})$ and energies $\epsilon_f$ are computed.
1045: Given the size of the system, it is not practical to do more
1046: than a few k-points; usually only one $k$-point, e.g.,
1047: the
1048: $\Gamma$ point is computed. The conductivity for the given configuration,
1049: and for the selected $k$ point with weight $W(k,\{R_i\})$
1050: is estimated using the second-order
1051: Fermi golden rule formula.
1052: \begin{eqnarray*}
1053: \sigma(\omega,\{R_i\},k)&=&\frac{2\pi}{3V_b\omega}
1054: \sum_{fg}|W(R_i, k)<\psi_f|\vec{p}|\psi_g>|^2\\
1055: & &(n_f(k)-n_g(k))
1056: \delta(\epsilon_f(k)-\epsilon_g(k)-\omega)
1057: \end{eqnarray*}
1058: This is in fact the Kubo-Greenwood (KG) procedure (sampled with
1059: one k-point) that one may use for
1060: a crystalline
1061: solid. A Kohn-Sham form is used instead of the
1062: many-electron eigenstates. The occupation factors $n_f(k)$ are also
1063: the one-particle occupations at the appropriate temperature.
1064: The energies used are the LDA eigenvalues without corrections.
1065: Even a crystalline
1066: structure has its phonon modes and their effect is ignored in the
1067: energy denominators.
1068: Similarly, the self-energy effects from the electron-electron
1069: processes are also ignored, although they can be quite
1070: large\cite{dyson-mott,benedict}.
1071: Only the {\it umklapp} processes associated with the reciprocal vectors of
1072: the simulation box contribute to the static conductivity.
1073: This is because the
1074: wavefunctions are eigenstates of the ``crystalline'' structure \{$R_i$\}. The
1075: actual conductivity has to be obtained by taking a configuration average which
1076: requires the Helhmoltz free energies for all crystal configurations \{$R_i$\}.
1077: In effect, the method uses simple LDA-DFT for the electrons,
1078: but abstains from using
1079: DFT for the ions and carries out a detailed MD evaluation
1080: of the liquid structure.
1081: We believe that this is unnecessary, especially for systems where the
1082: spherical symmetry of the plasma is a statistically reasonable assumption.
1083: The work of Kown et al.\cite{kwon} on strongly-coupled H-plasmas using
1084: these methods, and their comparison with our work is an example of this.
1085: In Fig.~\ref{figgr} we compare the
1086: ion-ion pair-distribution functions obtained from our NPA+
1087: HNC+bridge type procedures with available simulations
1088: from Silvestrelli et al\cite{silv1}, and from Levesque et al\cite{lwr}.
1089: %
1090: \begin{figure}
1091: \includegraphics*[width=8.5 cm, height=11.0 cm]{figgr.ps}
1092: \caption
1093: {Ion-ion pair-distribution functions of Al from this work compared with
1094: those obtained by Silvestri\cite{silv1}. The inset shows the comparison at
1095: normal density and at the melting point. Here no data from
1096: ref.\cite{silv1} is availble, but we use accurate molecular-dynamics
1097: simulations (Levesque et al\cite{lwr}) for comparison.
1098: }
1099: \label{figgr}
1100: \end{figure}
1101: %
1102: A simulation box of 125 atoms implies only 5 atoms per dimension and hence
1103: even the central atom feels only two atomic shells around it.
1104: The calculation of the $\omega\to 0$ limit needed to obtain a static
1105: conductivity is also
1106: quite difficult, and one approach is to
1107: {\it assume the validity of the Drude form} and
1108: fit a free-electron Drude form to achieve this.
1109: The Kubo-Greenwod type MD procedure would nevertheless
1110: provide useful complementary results for
1111: comparison with our two-component DFT approaches, and would be of
1112: much interest in studying low-temperature systems with a tendency
1113: to covalent bonding and clustering.
1114: However,
1115: these methods bring in a number of approximations of their own and
1116: need cautious reconsideration. In this context, an interesting test
1117: would be to examine expanded liquid-Hg\cite{bhat}
1118: using the Kubo-Greenwood MD approach.
1119: %
1120: \begin{figure}
1121: \includegraphics*[width=8.5cm, height=11.0cm]{kp25.ps}
1122: \caption
1123: {Dynamic conductivity (au.) of Al at a compression
1124: of 0.25 at $T$ =2.5 eV and 10 eV. The Drude conductivity
1125: is modified by the presence of the $3s$ level. Time-dependent
1126: DFT calculation greatly weakens the effect.
1127: }
1128: \label{kp25}
1129: \end{figure}
1130: %
1131: \subsection{ Dynamic conductivity from
1132: time-dependent density functional theory.}
1133: %
1134: Time dependent DFT provides a convenient approach
1135: to the calculation of the dynamic conductivity of interacting systems.
1136: The TDFT formulation relevant to dense plasmas\cite{gri}
1137: includes dynamic screening and coupling to ion-dynamics
1138: in a computationally convenient, self-consistent manner.
1139: The main consequence of the TDFT formulation is to replace, say,
1140: the dipole matix elements $<i|\vec{r}|j>$ between states $i,j$ by a
1141: dynamic form $<i|\vec{r}(\omega)|j>$ where the modification
1142: of the driving field by the response of the system is taken into
1143: account in a self-consistent manner.
1144: Consider a weak external field $\vec{E}_{ext}(t)$ = $\vec{E}cos(\omega t)$.
1145: This corresponds to an external potential:
1146: \begin{equation}
1147: U_{ext}(\vec{r},t)= e\vec{r}\cdot E_{ext}(t)
1148: \end{equation}
1149: The dipole form of the interaction is used since one of the objectives is to
1150: include the corrections arising from the
1151: presence of bound states. However, the dipole form of the matrix elements
1152: can be easily replaced by the momentum or acceleration formulation
1153: when needed. We assume that the
1154: electric field is directed along the z-direction,
1155: and suppress vector notation for the
1156: field unless needed.
1157: The external potential induces an electron density fluctuation $\delta n(r)$
1158: which in turn generates corrections to the Coulomb and exchange-correlation
1159: potentials. Since the linear absorption coefficient (or optical conductivity)
1160: is the object of our study, $\delta n(r)$ etc., can be written in terms of the
1161: electron response function $\chi^0(r,r'|\omega)$ which
1162: can be approxmiately constructed\cite{gri} from the
1163: Kohn-Sham eigenstates of the plasma. Then we have
1164: \begin{eqnarray}
1165: \label{uind}
1166: U(\vec{r},\omega)&=& U_{ext}(\vec{r},\omega)+V^c_{ind}+V^{xc}_{ind}\\
1167: V^c_{ind}&=&\int d\vec{r'}
1168: \frac{\delta n(\vec{r},\omega)}{|\vec{r}-\vec{r'}|}\\
1169: V^{xc}_{ind}&=&\left[\frac{\partial V_{xc}(\vec{r},\omega)}
1170: {\partial n(\vec{r},\omega)}\right]\delta n(\vec{r},\omega)\\
1171: \delta n(\vec{r},\omega)&=&
1172: \int d\vec{r'}\chi(\vec{r},\vec{r'}|\omega)U(\vec{r'},\omega)
1173: \end{eqnarray}
1174: %
1175: Here $V^{xc}_{ind}(r)$ is calculated from the
1176: gradient $\partial V_{xc}/\partial n$
1177: evaluated at the density $n_0(r)$ in the unperturbed neutral pseudo atom.
1178: Since the form of
1179: the time-dependent exchange-correlation potential $V_{xc}(r,\omega)$
1180: is still not established, most implementations
1181: use the
1182: static $V_{xc}[n(r)]$
1183: of ordinary density functional theory.
1184: The above set of equations have to be solved self consistently to obtain the
1185: total perturbing potential $U(\vec{r},\omega)$. In effect, the dipole operator,
1186: or equivalently, the momentum operator of the scattering
1187: electron is replaced by
1188: a space and time dependent quantity which enters into the polarizability.
1189: Given the spherical symmetry of the system, the total perturbing
1190: potential has the form
1191: \begin{equation}
1192: U(\vec{r},\omega)=-(1/2)EU_{\omega}(r)Y^0_1(\vec{r}/r)
1193: \end{equation}
1194: Here $Y^0_1$ is a spherical harmonic.
1195: If we ignored these induced fields, the conductivity of
1196: the system can be written as:
1197: \begin{equation}
1198: \Re\,\sigma^0(\omega)= \frac{e^2}{a_0\hbar}
1199: \Im\,\sum_{\nu,\nu'}\frac{\pi N_i \omega}{3}
1200: \frac{|<\nu|r|\nu'>|^2(f_\nu-f_{nu'})}
1201: {\omega+\epsilon_\nu-\epsilon_\nu'+i\delta}
1202: \end{equation}
1203: Here all the quantities on the right of the summation are in atomic units.
1204: The $e^2/(a_0\hbar)$ factor is the atomic unit of
1205: conductivity.
1206: $N_i$ is the number of ions per unit volume,
1207: and we neglect the effect of ion-ion
1208: correlations in the bound-free and bound-bound processes.
1209: (It can be shown that these contribute mainly to the width
1210: of the transition by broadening the levels)
1211: Also when $\nu,\nu'$ =
1212: $k,l,m$ and $k',l'm'$ for free-free transitions,
1213: the dipole matrix element is replaced by the
1214: momentum form, i.e, $|<\nu|\vec{r}|\nu'>|^2$ =
1215: $|<\nu|\vec{\nabla}|\nu'>|^2/\omega^2$.
1216:
1217: The conductivity expression
1218: $\sigma^0(\omega)$ is known to be particularly inadequate when treating
1219: b-f and b-b processes, unless the initial bound state is a deep
1220: lying (e.g, 1s) state\cite{kedge}. Hence we need to include the
1221: induced fields
1222: in treating under-dense plasmas where
1223: there is a bound state close to ionization.
1224: This involves the use of the total perturbing
1225: potential, $U(\vec{r},\omega)$, rather than the external
1226: potential $\vec{r}\cdot\vec{E}$ in constructing the
1227: matrix elements contained in the conductivity expression:
1228: \begin{eqnarray}
1229: \lefteqn{\Re\,\sigma(\omega)=\frac{ N_i\pi \omega}{3}\Im\sum_{\nu,\nu'}}\\
1230: & &|<\nu|U_{\omega}(r)Y^0_1(\vec{r}/r)|\nu'>|^2
1231: *(f_\nu-f_{nu'})
1232: \delta(\omega+\epsilon_{\nu}-k^2/2) \nonumber
1233: \end{eqnarray}
1234: The effect of level broadening can be included in the above
1235: expression by replacing the delta function $\delta(\omega+\epsilon_{\nu}-k^2/2)$
1236: by a form containing the self-energy corrections to the single-particle levels,
1237: as in Grimaldi et al, where self-energies were calculated. However, for
1238: our present purpose, we use the $\tau$ used in the Drude calculation
1239: to provide a broadening parameter.
1240: In fact, we have shown in Ref.~\cite{dyson-mott} that the self-energy
1241: contribution of ion-electron scattering to level broadening is
1242: identical to that given by the Ziman formula.
1243: Thus, setting $\gamma=1/\tau$, we
1244: replace the $\delta$-function in the above equation with a Lorentzian.
1245: $$\delta(\omega+\epsilon_{\nu}-k^2/2)\to \frac{\gamma/\pi}
1246: {(\omega+\epsilon_{\nu}-k^2/2)^2+\gamma^2}$$
1247: %
1248: \subsection{Some numerical results.}
1249: In this section we present results for the dynamic conductivity of
1250: some Al-plasmas to illustrate the effect of the scattering events where
1251: shallow bound states modify the Drude-like conductivity. These
1252: ``bound states'' are really Kohn-Sham eigenvalues and hence their
1253: values may need improvement by evaluating the corrections
1254: using a Dyson equation.\cite{dyson-mott}
1255: The first step of the calculation is the solution of the NPA model to
1256: obtain the Kohn-Sham basis set $\{\phi(r)_\nu\}$ for
1257: the given electron density
1258: and temperature. The codes necessary for the NPA calculation,
1259: the resistivity calculation, as
1260: well as the plasma conditions (EOS) that go into the resistivity
1261: or conductivity calculations may be accessed via the internet
1262: at our website\cite{web}.
1263: %
1264: \begin{figure}
1265: \includegraphics*[width=8.5cm, height=11.0cm]{kp1.ps}
1266: \caption
1267: {
1268: Dynamic conductivity (au.) of Al at a compression
1269: of 0.1 at $T$ =2.5 eV and 10 eV. The Drude conductivity
1270: is modified by the presence of shallow $3s$ at $T$ = 2.5 eV, and
1271: also the $3p$ level at 10 eV. The effect is reduced in the
1272: TDFT calculation.
1273: }
1274: \label{kp1}
1275: \end{figure}
1276: %
1277: %\begin{figure}
1278: %\caption
1279: %{
1280: %The dynamic conductivity of solid Aluminum at $T$ = 300K, experiment
1281: %and theory. The modification of the Drude form is due to
1282: %interband processes. The conductivity $\sigma$ is in units of
1283: %1.014 x 10$^{15}$ sec$^{-1}$. A scattering rate of
1284: %$\tau$ = 0.6 x 10$^{-14}$ is used.
1285: %}
1286: %\label{fig4}
1287: %\end{figure}
1288: %
1289: Some results for the shallow bound states present in
1290: $Al$-plasmas at T=2.5 eV and 10 eV, for compressions $\kappa=0.25$, and 0.1
1291: are given in table II. In the T=10, $\kappa=0.1$ case we have two shallow
1292: bound states, viz., $3s$ and $3p$.
1293: In Table~\ref{occ} their occupation numbers and the interacting
1294: chemical potential needed for calculating the Fermi factors are given.
1295: In calculating the matrix element $|<nlm|r|kl'm'>|^2$, the
1296: correct density of states $\aleph(\epsilon)$
1297: for the continuum state $|kl'm'>$, normalized
1298: in a sphere of radius $R$, with $R\to\infty$ should be
1299: included.
1300: That is,
1301: \begin{eqnarray*}
1302: k_nR-\pi l/2&=&n\pi +\delta_{kl}\\
1303: \aleph(\epsilon)&=&\frac{\partial n}{\partial k}
1304: \frac{\partial k}{\partial \epsilon}
1305: \end{eqnarray*}
1306: The calculations are very simple if (a)
1307: we ignore the Zangwill-Soven type TDFT effects arising from the
1308: reaction of the system which act to modify the external field.
1309: (b) if we ignore the phase shifts $\delta_{kl}$,
1310: and replace the boundstates by hydrogen-like states with the correct
1311: $Z=\bar{Z}$.
1312: The final results depend on the level broadening parameter
1313: $\gamma$ used in the calculation.
1314:
1315: In fig.\ref{kp25} and Fig.\ref{kp1} we show the conductivity
1316: arising from
1317: the presence of the $3s$ and $3p$ levels, as well as the Drude term, and the
1318: modification when TDFT is included in an approximate manner.
1319: These TDFT calculations should be regarded as highly provisional
1320: results only; in fact, the treatment of shallow boundstates using
1321: DFT is itself questionable since the Kohn-Sham eigenvalues
1322: are known to be a poor approximation to the actual excitation spectrum.
1323: %
1324: \subsubsection{bound-bound processes.}
1325: Clearly, the presence of a partially occupied $3s$ and a $3p$ state
1326: with an energy separation of the order of 0.46 au. would lead to a
1327: contribution from b-b processes at around 12 eV. This contribution is easily
1328: included in the calculation under the simplifying assumptions that we
1329: noted before. However, this is a relatively sharp ``line'' resonance
1330: which is expected to undergoes
1331: significant modification when the time-dependent effects are taken in to
1332: account. We have not included it in
1333: our figures.
1334: %
1335: \begin{table}
1336: \caption{details regarding the n=3 shallow states in
1337: underdense $Al$ at T=10 eV and 2.5 eV. Full occupation is when Occ
1338: is unity}
1339: \begin{ruledtabular}
1340: \begin{tabular}{cccccc}
1341: $\kappa$&0.25 & 0.25&0.1& 0.1 \\
1342: $ T$ &2.5 eV & 10 eV &2.5 eV & 10 eV\\
1343: \hline\\
1344: $\mu$ &0.0393&-0.5538&-0.6652 &-0.9589\\
1345: Occ(3s)&0.7444&,0.2217& 0.6850 &0.1455 \\
1346: Occ(3p)& -- & -- & -- &0.0843\\
1347: \end{tabular}
1348: \end{ruledtabular}
1349: \label{occ}
1350: \end{table}
1351: %
1352: \section{Conclusion.}
1353: The first-principles calculation of the
1354: dynamic conductivity of warm dense matter may be conveniently
1355: carried out within the
1356: framework of multi-component density functional theory.
1357: The static calculation
1358: (NPA etc) provides the Kohn-Sham basis set, phase shifts, pseudopotentials
1359: for constructing ion-ion pair potentials and structure factors. These
1360: immediately provide results for the static conductivity. Further, the
1361: energy-levels and occupation numbers obtained from the Kohn-Sham
1362: NPA solution can be the starting approximation for a time-dependent
1363: density functional calculation of the optical conductivity. This
1364: proceeds in much the same way as the optical absorption cross section.
1365: Details of such a time-dependent calculation, based on the
1366: method of Zangwill and Soven, may be found in
1367: Grimaldi, Lecourte and Dharma-wardana.\cite{gri}
1368:
1369:
1370: %\begin{references}
1371: \begin{thebibliography}{99}
1372: %
1373: \bibitem{per87}
1374: F. Perrot and M.W.C. Dharma-wardana, Phys. Rev. A {\bf 36},238 (1987)
1375: \bibitem{eos95}
1376: F. Perrot and M.W.C. Dharma-wardana,Phys. Rev. E. 52, 2920 (1995);
1377: Int. J. Thermophys. {\bf 20}, 1299 (1999)
1378: \bibitem{benageR}
1379: J. F. Benage, W. R. Shanahan and M. S. Murillo, Phys. Rev. Lett.
1380: 83, 2953 (1999)
1381: \bibitem{niemann}
1382: C. Niemann, L. Divol, D. H. Froula, G. Gregori, O. Jones, R. K. Kirkwood,
1383: A. J. McKinnon, N. B. Meezan, J. D. Moody, C. Sorce, L. J. Suter, R. Bahr,
1384: W. Seka, and S. H. Glenzer, Phys. Rev. Lett., {\bf 94}, 85005 (2005)
1385: \bibitem{ng}
1386: A. Forsman, A. Ng, G. Chiu, and R. M. More, Phys. Rev. E 58, R1248-R1251 (1998)
1387: \bibitem{ngcdw}
1388: A. Ng, T. Ao, F. Perrot, M. W. C. Dharma-wardana
1389: and M. E. Foord, http://arxiv.org/physics-0505070 ;
1390: to appear in Lasers and Particle Beams.
1391: \bibitem{ceperley81}
1392: D. M. Ceperley, {\em Recent Progress in Many-body Theories}, Ed. J. B.
1393: Zabolitsky (springer, Berlin 1981), p262
1394: \bibitem{kwon}
1395: I. Kwon, L. Collins, J. Kress and N. Troullier, Phys. Rev. E. {\bf 54},
1396: 2844 (1996)
1397: \bibitem{leemore}
1398: Y. T. Lee and R. M. More, Phys. Fluids, {\bf 27}, 1273 (1984)
1399: \bibitem{gri}
1400: F. Grimaldi, A. Grimaldi-Lecourte, and M.W.C. Dharma-wardana, Phys. Rev. A
1401: {\bf 32}, 1063 (1985)
1402: \bibitem{eosb}
1403: F. Perrot, M.W.C. Dharma-wardana, and John Benage, Phys. Rev. E 65, 046414
1404: (2002)
1405: \bibitem{hyd}
1406: M. W. C. Dharma-wardana and Fran\c{c}ois Perrot, Phys. Rev.B (2002)
1407: \bibitem{bhat}
1408: R. N. Bhatt and T. M. rice, Phys. Rev. B, {\bf 20}, 466 (1979)
1409: \bibitem{japmore}
1410: H. Yoneda, H. Morikami, K.-i Ueda, and
1411: R. M.More, Phys. Rev. Lett., {\bf 91}, 75004 (2003)
1412: \bibitem{aung}
1413: Andrew Ng, private communication.
1414: \bibitem{dagens}
1415: L. J. Dagens, Phys. (Paris) {\bf 34}, 879 (1973)
1416: \bibitem{codes}
1417: VASP, www.b-initio Simulation Package,
1418: AB-INIT, www
1419: GAUSSIAN, www
1420: \bibitem{web}
1421: http://babylon.phy.nrc.ca/ims/qp/chandre/
1422: \bibitem{ilciacco}
1423: Density Functional Theory, Ed. E. K. U. Gross and Dreizler
1424: \bibitem{dyson-mott}
1425: F. Perrot and M. W. C. Dharma-wardana, Phys. Rev. A 29, 1378 (1984)
1426: \bibitem{surh}
1427: Michael Surh, T. W. Barbee III, and L. H. Yang, Phys. Rev. Lett.
1428: {\bf 86}, 5958 (2001)
1429: \bibitem{geoffry}
1430: R. Evans, B. L. Gyorffy, N. Szabo and J. M. Ziman, in
1431: {\it Properties of liquid metals}, Ed. S. Takeuchi (Wiley, New York, 1973)
1432: \
1433: \bibitem{allenpoly}
1434: W. W. Schulz and P. B. Allen, Phys. Rev. B {\bf 52}, 7994 (1995)
1435:
1436: \bibitem{majum}
1437: C.Majumdar and W. Kohn, Phys. Rev. 138, A1617 (1965)
1438: \bibitem{milsch}
1439: M.W.C. Dharma-wardana and F. Perrot, Phys. Let., A {\bf 163}, 223 (1992);
1440: M.W.C. Dharma-wardana in {\it Laser Interactions with Atoms, Solids, and
1441: Plasmas}, Edited by R.M. More (Plenum, New York, 1994), p311
1442: \bibitem{matt}
1443: S. Ethier and J.-P. Matte, Phys. Plasmas {\bf 8}, 1650 (2001), and references
1444: there-in.
1445: \bibitem{lanbut}
1446: M. B\u{u}ttiker, A. Pr\`{e}tre, and H. Thomas,
1447: Phys. Rev. Lett. {\bf 70}, 4114 (1993) and references there-in.
1448: \bibitem{ash-sturm}
1449: N. W. Ashcroft and K. Sturm, Phys. Rev. B, {\bf 3} 1898 (1971);
1450: G Dresselhaus, M. S. Dresselhaus, and D. Beaglehole, as reported
1451: in Ashcroft and Sturm.
1452: \bibitem{miller}
1453: J. C. Miller, Phil. Mag. {\bf 20}, 1115 (1969)
1454: \bibitem{faber}
1455: T. E. Faber, {\it An introduction to the theory of liquid metals},
1456: (Cambridge University press,
1457: Cambridge,1972).
1458: \bibitem{benedict}
1459: L. Benedict, C. D. Spataru, and S. G. Louie,
1460: Phys. Rev. B {\bf 66}, 085116 (2002)
1461: \bibitem{hopfield}
1462: J. J. Hopfield, Phy. Rev. A {\bf 139}, 419 (1965)
1463: \bibitem{mahan}
1464: G. D. Mahan, J. Phys. Chem. Solids {\bf 31}, 1477 (1970) gives a calculation
1465: for phonon scattering.
1466: \bibitem{ropke}
1467: G. R\"{o}pke, R. Redmer, A. Wierling and H. Reinholz, Phys. Rev E
1468: {\bf 60} R2484 (1999)
1469: \bibitem{elr}
1470: M.W.C. Dharma-wardana and F. Perrot, Phys. Rev. E {\bf 58}, 3705 (1998)
1471: and Erratum, {\bf 63}, 069901 (2001)
1472: \bibitem{chncT}
1473: F. Perrot and M.W.C. Dharma-wardana, Phys. Rev. B15 {\bf 62}, 14766 (2000)
1474: \bibitem{silv1}
1475: P. L. Silvestrelli, Phys. Reb B {\bf 60}, 16382 (1999);
1476: P. L. Silvestrelli, A. Alavi, and M. Parrinello,
1477: Phys. Rev. B {\bf 55}, 15515 (1997)
1478: \bibitem{alavi}
1479: A. Alavi, M. Parrinello, and D. Fenkel, {\it Science}, {\bf 269},
1480: 1252 (1995)
1481: \bibitem{lwr}
1482: D. Levesque, J. J. Weis, and J. Reatto, Phys. Rev. Lett. {\bf 54}, 451 (1985);
1483: M. W. C. Dharma-wardana and G. C. Aers, {\it Ibid}, {\bf 56}, 1211 (1986)
1484: \bibitem{gillan}
1485: Gillan et al, see the phonon website
1486: %\bibitem{vigulrich}
1487: %See E. K. U. Gross, J. F. Dobson and Petersilka, in
1488: %{\it Density Functional Theory II}, vol 181 of {\it Topics in Current
1489: %Chemistry} (Springer, Berlin, 1996); G. Vignale and W. Kohn,
1490: %Phys. Rev. Lett., {\bf 77}, 2037 (1996)
1491: \bibitem{zw}
1492: A. Zangwill and P. Soven, Phys. Rev. A {\bf 21},1561 (1980)
1493: \bibitem{kedge}
1494: F. Perrot and M.W.C. Dharma-wardana, Phys. Rev. Lett.
1495: {\bf 71} , 797 (1993)
1496: %
1497:
1498: \end{thebibliography}
1499:
1500:
1501:
1502:
1503:
1504:
1505: \end{document}
1506:
1507:
1508:
1509: %\bibitem{gross}
1510: %N. Iwamoto and E. K. U. Gross, Phys. Rev. B {\bf 50}, 3003 (1987)
1511: % and references
1512: % therein.
1513: %
1514: %W. E. Pickett and J. Q. Broughton {\bf 48}, 14859 (1993)
1515: %\bibitem{vBH}
1516: %U. von Barth and L. Hedin, J. Phys. C {\bf 5}, 1629 (1972)
1517: %
1518: %\bibitem{kittel}
1519: %C. Kittel, {\it Quantum theory of Solids} p 94, (Wiley, New York) 1987
1520: %
1521: %\bibitem{verlet69}
1522: %D. Levesque and L. Verlet, Phys. Rev. {\bf 182}, 307 (1969)
1523: %
1524: %\bibitem{sibert92}
1525: % L.E. Gonzalez, D.J. Gonzalez and M. Sibert, Phys. Rev. A {\bf 45}, 3803 (1992).
1526: %
1527: %\bibitem{kuk}
1528: % Other models of pair-interactions which include xc-effects as well
1529: %have been given by C. A. Kukkonen and A. W. Overhauser, Phys. Rev. B
1530: %{\bf 20}, 550 (1979), A. H. MacDonald, Can J. Phys.{\bf 60}, 710 (1982)
1531: %and e.g., in ref.\cite{richardson}.
1532: %
1533: %\bibitem{kushida}
1534: %T. Kushida, J. C. Murphy, and M. Hanabusa, Phys. Rev. {\bf B 13}, 5136 (1976)
1535: %
1536:
1537: %\bibitem{dwp82}
1538: % M.W.C. Dharma-wardana and F.
1539: %Perrot, Phys. Rev. A {\bf 26}, 2096 (1982)
1540: %
1541: %\bibitem{callaway}
1542: % H. Iyetomi and S. Ichimaru, Phys. Rev. A {\bf 34}, 433 (1986);
1543: %D.G. Kanhere, P.V.
1544: %Panat, A.K. Rajagopal and J. Callaway, Phys. Rev. A {\bf 33}, 490 (1986);
1545: % F. Perrot
1546: %and M.W.C. Dharma-wardana, Phys. Rev. A {\bf 30}, 2619 (1985?)
1547: %\bibitem{Ichimaru}
1548: %S. Ichimaru and S. Tanaka, Phys. Rev. A {\bf 32}, 1790 (1985)
1549:
1550:
1551:
1552:
1553:
1554:
1555: \begin{table}
1556: \caption
1557: { Resitivity (in $\mu$ohm$\,$cm) as a function of the {\em electron}
1558: compression $\kappa_e$ and the temperature in eV.
1559: }
1560: \begin{ruledtabular}
1561: \begin{tabular}{ccccccccc}
1562:
1563: T(eV) & $n_0/32$ & $n_0/16$ & $n_0/8$ & $n_0/4$ & $n_0/2 $ & $n_0$ & $2n_0 $ 4n_0$ \\
1564: \hline\\
1565: 0.04 & 975 & 330 & 109 & 37.0 & 6.20 & 3.63 & 1.29 & \\
1566: 0.06 &1462 & 497 & 191 & 66.6 & 20.7 & 5.45 & 1.93 & \\
1567: 0.08 &1949 & 663 & 274 & 97.2 & 36.6 & 7.27 & 2.57 & \\
1568: 0.10 &2437 & 828 & 359 & 126. & 52.4 & 9.08 & 3.22 & \\
1569: 0.25 &4090 & 1796& 713 & 240. & 83.1 & 24.2 & 8.04 & \\
1570: 0.50 &3854 & 1953& 809 & 307. & 96.3 & 27.7 & 35.7 & \\
1571: 0.75 &3416 & 1962& 860 & 358. & 113. & 31.4 & 35.7 & \\
1572: 1.00 &3040 & 1912& 874 & 386. & 130. & 35.3 & 35.7 & \\
1573: 1.75 &2021 & 1518& 858 & 433. & 170. & 43.3 & 35.8 & \\
1574: 2.50 &1685 & 1158& 786 & 436. & 197. & 50.2 & 35.8 & \\
1575: 3.75 &1224 & 918 & 670 & 419. & 226. & 60.2 & 43.0 & \\
1576: 5.00 &946 & 772 & 593 & 401. & 243. & 70.8 & 47.9 & \\
1577: 7.50 &707 & 629 & 501 & 375. & 267. & 89.1 & 64.2 & \\
1578: 10.0 &633 & 534 & 452 & 357. & 279. & 108. & 76.1 & \\
1579: 20.0 &538 & 466 & 351 & 312. & 255. & 155. & 104. & \\
1580: 30.0 &487 & 385 & 328 & 258. & 220. & 169. & 118. & \\
1581: 40.0 &420 & 252 & 280 & 238. & 187. & 153. & 120. & \\
1582: 60.0 &343 & 331 & 220 & 190. & 160. & 130. & 107. & \\
1583: 80.0 &301 & 215 & 180 & 157. & 133. & 113. & 93.1 & \\
1584: 100. &262 & 189 & 150 & 132 & 114 & 97.5 & 81.0 & \\
1585: \end{tabular}
1586: \end{ruledtabular}
1587: \label{restab}
1588: \end{table}
1589:
1590:
1591:
1592:
1593: