1: \documentstyle[aps,multicol,psfig,amsfonts]{revtex}
2: \newcommand{\ds}{\rule[-1.5mm]{0mm}{4.5mm}\displaystyle}
3: \renewcommand{\bar}[1]{\overline{#1}}
4: \begin{document}
5: \newcommand{\be}{\begin{equation}}
6: \newcommand{\ee}{\end{equation}}
7: \newcommand{\br}{\bbox{r}}
8: \newcommand{\brp}{\bbox{r}^\prime}
9: \newcommand{\rd}{{\rm d}}
10: \newcommand{\re}{{\rm e}}
11: \newcommand{\ri}{{\rm i}}
12: \newcommand{\au}{\mbox{\,[a.u.]}}
13: \newcommand{\wc}{\omega_{\rm c}}
14: \renewcommand{\wp}{\omega_{\rm p}}
15: \newcommand{\rs}{r_{\rm s}}
16: \newcommand{\ks}{k_{\rm s}}
17: \newcommand{\qs}{q_{\rm s}}
18: \bibliographystyle{prsty}
19: \widetext
20: \title{
21: Collective versus single-particle effects
22: in the optical spectra of
23: finite electronic quantum systems}
24: \author{M. Santer and B. Mehlig}
25: \address{Theoretical Quantum Dynamics, University
26: of Freiburg, Hermann-Herder-Str. 3, Freiburg}
27: \date{\today}
28: \maketitle{ }
29: \begin{abstract}
30: We study optical spectra of finite electronic quantum
31: systems at frequencies smaller than the
32: plasma frequency
33: using a quasi-classical approach.
34: This approach
35: includes collective effects and enables
36: us to analyse how the nature of the (single-particle) electron dynamics
37: influences the optical spectra in finite electronic
38: quantum systems. We derive an analytical expression for the low-frequency
39: absorption coefficient of electro-magnetic radiation
40: in a finite quantum system with ballistic
41: electron dynamics and specular reflection
42: at the boundaries: a two-dimensional electron gas confined to
43: a strip of width $a$ (the approach can be applied to systems of any shape and
44: electron dynamics -- diffusive or ballistic, regular or irregular motion).
45: By comparing with results of numerical computations using the random-phase
46: approximation we show that our analytical
47: approach provides a qualitative and
48: quantitative understanding of the optical spectrum.
49: \end{abstract}
50: \pacs{03.65.Sq,73.20.Dx,73.23.-b,0545.Mt}
51: \begin{multicols}{2}
52: Optical spectra of finite metallic systems
53: have been intensively investigated for
54: almost a century. Early approaches
55: such as Mie's \cite{mie08} are of classical nature.
56: In \cite{mie08} the absorption of electro-magnetic radiation
57: by conducting spheres is determined.
58: It is shown that the absorption spectrum
59: exhibits a resonance at $\omega_{\rm p}/\sqrt{3}$
60: (where $\omega_{\rm p}$ is the bulk plasma frequency)
61: due to collective motion of the charge carriers.
62:
63: In the last decades there has been a substantial
64: amount of work on the nature of such collective resonances in
65: metallic clusters \cite{eka99}, nuclei \cite{ber94}, thin films \cite{fer58,kan61,new70},
66: small metal particles \cite{carr}, and dimensionally
67: reduced quantum systems \cite{all83,fet86,heit92} investigating,
68: in particular, quantum-mechanical
69: effects. In most of these cases,
70: the electron dynamics is {\em ballistic} (the mean free path $\ell$
71: is larger than the system size $a$).
72: The majority of theoretical papers
73: on the Mie resonance of finite metallic
74: systems use the so-called random phase approximation (RPA),
75: a self-consistent, quantum-mechanical approach
76: incorporating collective effects.
77: The nature of the Mie resonance in finite
78: electronic quantum systems is well understood,
79: both qualitatively and quantitatively
80: (for a recent review see for instance \cite{eka99}).
81:
82: The emergence of the field of Mesoscopic Physics
83: has fueled an increased
84: interest in the electronic and optical properties
85: of finite, disordered quantum systems (with {\em diffusive} electron
86: dynamics, $\ell \ll a$) in external fields.
87: In this context, attention has largely focussed on
88: \mbox{(quasi-)static} properties, $\omega \simeq \Delta/\hbar$,
89: where $\Delta$ is the mean level spacing of the system
90: in question (which is generally smaller than $\omega_{\rm p}$
91: by many orders of magnitude). Mesoscopic fluctuations
92: of the static polarisability and the capacitance,
93: for instance, were characterised in \cite{efe96},
94: and the electron density itself in \cite{agam}.
95: In all of these cases, collective effects
96: must be taken into account in order to adequately
97: deal with the screening of the external field.
98: In the static limit, a Thomas-Fermi (TF) {\em ansatz} is appropriate
99: \cite{efe96,agam}.
100:
101: Much less is known about optical spectra
102: in the low-frequency region $\Delta/\hbar < \omega \ll \omega_{\rm p}$,
103: despite the fact that this regime is of particular interest:
104: one expects that the spectra strongly depend
105: on the nature of the (single-particle) electron
106: dynamics: In {\em ballistic} systems, for example,
107: it was argued \cite{aus93} that optical spectra should
108: exhibit {\em resonances near multiples
109: of $\wc = \pi v_{\rm F}/a$}
110: (see also \cite{meh97e}).
111: Usually $\Delta/\hbar \ll \wc \ll \wp$ .
112: According to \cite{aus93}, these resonances
113: should overlap to give a {\em linear
114: frequency dependence} of the absorption
115: coefficient for $\omega \gg \wc$ in two-dimensional systems
116: (classical and local electro-magnetic theories predict a quadratic dependence
117: independent of dimension). These
118: are striking and unexpected results. They were,
119: however, obtained within a TF approximation
120: which is valid in the {\em static} limit.
121: It must be examined to which extent {\em dynamic} screening effects may modify
122: the results \cite{note3}.
123:
124: Furthermore, in systems with {\em ballistic chaotic} dynamics,
125: the optical spectra should, at least within a single-particle picture,
126: reflect universal energy-level correlations found in classically
127: chaotic quantum systems\cite{boh}. This issue
128: was addressed in the
129: seminal paper by Gorkov and Eliashberg \cite{gor65},
130: investigating the polarisability of metallic
131: particles with disordered walls.
132: However, the treatment neglects
133: screening effects all together \cite{ric73,siv87,bla96}.
134: How does {\em dynamical} screening modify this picture?
135:
136: Numerical calculations (based, for instance on the RPA)
137: are ill-suited to answer these questions: They provide little
138: qualitative insight
139: in the low-frequency region and, more importantly, it is necessary
140: to consider small systems or to make use of symmetries in order to
141: make the numerical computations feasible.
142: Disordered and (asymmetric) chaotic systems are
143: very difficult to deal with.
144: In order to understand how the classical single-particle
145: dynamics is reflected in optical spectra
146: of finite quantum systems, it is thus greatly desirable to
147: have an analytical theory incorporating collective effects.
148:
149: The main result of this paper
150: is an analytical expression for
151: the absorption properties
152: for a two-dimensional electron gas confined to a strip
153: of width $a$ [eqs.~(\ref{eq:absresult}) and (\ref{eq:result})].
154: This example exhibits all features of more complicated geometries.
155: However, in the case discussed here, numerical RPA calculations are
156: feasible and allow us to discuss the accuracy of
157: the analytical approach.
158:
159: In the following it is assumed
160: that $\Delta/\hbar < \omega \ll \wp,E_{\rm F}/\hbar$
161: where $E_{\rm F}$ is the Fermi energy. It is furthermore
162: assumed that $\lambda,\lambda_{\rm s} \gg a \gg\lambda_{\rm F}$
163: where $\lambda$ is the wave length of the external
164: radiation, $\lambda_{\rm s}$ is the skin depth
165: and $\lambda_{\rm F}$ is the Fermi wave-length.
166:
167: We consider a closed metallic quantum system
168: placed in an external electric field
169: $\bbox{E}_{\rm ext}$ [with a time dependence
170: $\propto \exp(\ri\omega t)$]. If the wave-length is
171: much larger than the system size $a$, the external field is
172: approximately constant
173: throughout the system and
174: (neglecting retardation effects) can be
175: written as the gradient of an electric potential
176: $\varphi_{\rm ext}(\br) = E_0 x \hat{\bf e}_x$
177: [$\br = (x,y,z)$ is a three-dimensional coordinate vector].
178: Within the RPA,
179: the electronic response of the system to $\bbox{E}_{\rm ext}$
180: is calculated by solving a set of self-consistent
181: equations for the effective electrical potential
182: $\varphi(\br) = \varphi_{\rm ext}(\br) + \delta\varphi(\br)$
183: [$\delta\varphi(\br)$ is the potential
184: due to the induced charge density $\delta\varrho(\br)$]
185: \begin{eqnarray}
186: \label{eq:rpa}
187: \delta\varphi(\br) &=& \int\rd\br^\prime G(\br,\brp)\delta\varrho(\brp)\,,\\
188: \delta\varrho(\br) &=& -\int\rd\br^\prime \Pi_0(\br,\br^\prime;\omega)\,
189: \varphi(\br)
190: \nonumber
191: \end{eqnarray}
192: with the boundary condition that $\delta\varphi(\br)$ vanishes
193: as $|\br| \rightarrow \infty$. $G(\br,\brp)$ is the
194: Green function of the Laplace equation
195: $\Delta G(\br,\brp) = -\epsilon_0^{-1} \delta(\br-\brp)$,
196: $\epsilon_0$ is the dielectric constant.
197: $\Pi_0(\br,\br^\prime;\omega)$
198: is the non-local polarisability
199: \begin{eqnarray}
200: \nonumber
201: \Pi_0(\br,\br^\prime;\omega)
202: &=& -2 e^2\sum_{\alpha,\beta}
203: \frac{f(\varepsilon_\alpha)-f(\varepsilon_\beta)}{%
204: \varepsilon_\alpha\!-\!\varepsilon_\beta\!-\!\hbar\omega\!+\!\ri\gamma}
205: \\
206: &&\hspace*{1cm}\times\psi_\alpha^\ast(\brp)\psi_\beta(\brp)
207: \psi_\beta^\ast(\br)\psi_\alpha(\br)\,,
208: \end{eqnarray}
209: $e$ is the electron charge, $\varepsilon_\alpha$ and $\psi_\alpha(\br)$
210: are the single-particle eigenvalues and eigenfunctions
211: of the undisturbed system, they are usually calculated
212: with in a Hartree-Fock or a local-density approximation.
213: $f(\varepsilon) = \Theta(E_{\rm F}-\varepsilon)$,
214: and $\gamma>0$ is smaller than $\Delta$.
215: Within the RPA, the absorption coefficient
216: (proportional to the energy dissipation per unit time)
217: may be written as \cite{ber94}
218: \be
219: \label{eq:alpha1}
220: \alpha(\omega) = \frac{\hbar\omega}{2E_0^2} \mbox{Im}\, d(\omega)
221: \ee
222: where
223: \be
224: d(\omega) =
225: \int\rd\br\rd\brp \delta\varrho^\ast(\br)
226: \Pi_0(\br,\br^\prime;\omega)\varphi(\brp)
227: \ee
228: is the (complex) dipole moment and
229: the asterisk denotes complex conjugation.
230:
231: In the following we derive an explicit analytical
232: expression for the absorption
233: coefficient $\alpha(\omega)$ of a
234: finite electronic quantum system
235: in an external electric field, valid
236: in the frequency range
237: $\Delta/\hbar < \omega \ll \omega_{\rm p}$.
238: We make use of the approximations suggested
239: in \cite{wilk00}.
240: First, according to Fermi's golden rule,
241: the absorption coefficient is
242: \be
243: \label{eq:golden}
244: \label{eq:alpha2}
245: \alpha(\omega) \simeq
246: \frac{\pi\hbar^2 e^2\omega^2}{2\Delta^2 E_0^2}
247: \Big|\langle \psi_\alpha |\varphi| \psi_\beta\rangle
248: \Big|^2_{\mbox{}\hspace*{-2mm}\stackrel{\scriptstyle \varepsilon_\alpha \simeq E_{\rm F}}{\mbox{}\hspace*{3mm}\varepsilon_\alpha\!-\varepsilon_\beta\simeq \omega}}\,.
249: \ee
250: Second, the matrix elements of $\varphi$ are evaluated within
251: a semi-classical approximation
252: \cite{meh97e,wilk87,eck92,meh95,meh00a}.
253: Third, $\varphi$ itself is
254: determined within a quasi-classical approximation:
255: according to (\ref{eq:rpa}),
256: the effective electric potential is given by
257: (in symbolic notation)
258: \begin{eqnarray}
259: \label{eq:varphi}
260: \varphi &=& -\Pi_0 ^{-1}\delta\varrho\,.
261: \end{eqnarray}
262: In order to determine $\delta\varrho$,
263: eqs. (\ref{eq:rpa}) are usually
264: solved numerically, using a
265: real-space discretisation \cite{ber90} or
266: by expanding in a suitable
267: basis set.
268: An approximate
269: analytical solution may be obtained
270: by noting that for $\omega \ll \omega_{\rm p}$,
271: $||G\Pi_0||\gg 1$. In other words,
272: $\delta\varrho$ is well approximated
273: by the classical charge density $\delta\varrho_{\rm cl}$
274: of the metallic system subjected to
275: an external potential $\varphi_{\rm ext}$,
276: \be
277: \label{eq:rhocl}
278: \label{eq:class}
279: \Delta\varphi_{\rm cl} = -\delta\varrho_{\rm cl}/\epsilon_0
280: \ee
281: with $\varphi_{\rm cl}(\br) \rightarrow \varphi_{\rm ext}(\br)$
282: as $|\br|\rightarrow\infty$ and $\varphi_{\rm cl} = 0$
283: within the system (in the classical limit, the external field
284: is thus screened out completely).
285: Eq. (\ref{eq:rhocl}) may be solved
286: for $\delta\varrho_{\rm cl}$ using standard methods
287: \cite{lan84}.
288: Fourth, $\Pi_0$ is determined within a quasi-classical
289: approximation \cite{kir75,wilk00}
290: \be
291: \label{eq:quasicl}
292: \Pi_0(\br,\brp;\omega) = e^2\nu_d \left[ \delta(\br-\brp) + \ri \omega
293: P^{(d)}(\br,\brp;\omega)\right]
294: \ee
295: where $\nu_d$ is the density of states per unit volume
296: in $d$ dimensions
297: and $P^{(d)}(\br,\brp;\omega)$ is the Fourier transform
298: of the classical propagator $P^{(d)}(\br,\brp;t)$.
299: In ballistic systems it is written as a sum over classical
300: paths $p$ from $\br$ to $\brp$
301: \be
302: P^{(d)}(\br,\brp;\omega) =\!\!\!\!\!\!\! \sum_{{\rm cl. paths}\,p}
303: \left|\mbox{det}\left[\frac{\partial (\brp)}{\partial (\tau_p,\bbox{n}_p)}
304: \right]\right|^{-1}\!\!\!\exp(\ri \omega \tau_p)\,.
305: \label{eq:prop}
306: \ee
307: Here $\tau_p$ is the time taken from $\br$ to $\brp$
308: along the path $p$, and $\bbox{n}_p$ is a unit vector
309: describing the direction of the initial velocity.
310: For diffusive systems see e.g. \cite{meh97a}.
311:
312: In the following we show
313: by comparison with quantum-mechanical RPA calculations
314: that (\ref{eq:golden}-\ref{eq:prop}) provide a qualitative
315: and quantitative description of absorption
316: in the frequency range $\Delta/\hbar < \omega \ll \omega_{\rm p}$.
317:
318: {\em Two-dimensional strip.}
319: We consider a two-dimensional electron gas
320: confined to a strip in the $x$-$y$-plane surrounded by vacuum \cite{note},
321: subject to a time-dependent
322: electric field $\bbox{E}_{\rm ext}$
323: directed along the negative $x$-axis [compare fig.~\ref{fig:1}(a)].
324: The width of the strip (along the $x$-axis) is $a$, its
325: length $L$ (along the $y$-axis),
326: with $L \gg a$. Within the system, the electrons move ballistically,
327: and they are specularly reflected at the
328: boundaries at $x = \pm a/2$.
329:
330: We write $\delta\varrho(\br) = \delta\sigma(x,y)\,\,\delta(z)$.
331: For $L \gg a$, the surface-charge density $\delta\sigma$ depends
332: on $x$ only and the resolvent $G$ is written as
333: \be
334: G =
335: \frac{1}{\epsilon_0}
336: \int\!\frac{\rd q}{2\pi}\, \frac{1}{2 |q|}\, {\rm e}^{\ri q (x-x^\prime)}
337: {\rm e}^{- |q| |z-z^\prime|}\,.
338: \ee
339: With
340: \be
341: \frac{1}{L}\int\rd y \rd y^\prime\,\Pi_0(\br;\brp;\omega)
342: = \Lambda_0(x,x^\prime;\omega)\,\delta(z)\,\delta(z^\prime)\,,
343: \ee
344: the RPA-equations (\ref{eq:rpa}) are reduced to a set of
345: one-dimensional
346: equations for $\varphi(x,z\!=\!0)$
347: with the kernel $\Lambda_0(x,x^\prime;\omega)$.
348: We model the confinement in the $x$-direction by introducing
349: hard-wall boundary conditions. This is adequate in the
350: range of parameters considered below and simplifies
351: the quasi-classical analysis.
352: We solve the resulting self-consistent equations
353: numerically using a real-space discretisation and obtain
354: the absorption coefficient from (\ref{eq:alpha1}).
355:
356: The corresponding quasi-classical approximation
357: for $\alpha(\omega)$ is obtained
358: as described above:
359: the classical surface-charge density
360: is determined
361: by solving (\ref{eq:class}) in elliptic cylinder coordinates:
362: $\delta\sigma_{\rm cl}(x) = 2\epsilon_0 E_0 (a^2/4-x^2)^{-1/2}
363: \Theta\left(|x| -a/2\right)$.
364: The corresponding classical
365: field lines are shown in fig.~\ref{fig:1}(a).
366: According to eqs.~(\ref{eq:quasicl},\ref{eq:prop}) the one-dimensional kernel
367: $\Lambda_0(x,x^\prime;\omega)$ is given by
368: a sum over classical paths as shown in fig.~\ref{fig:1}(b)
369: which may be summed by Poisson summation.
370: Using (\ref{eq:varphi}) one obtains for
371: the effective electric potential
372: \begin{eqnarray}
373: \label{eq:phi1}
374: \varphi(x) &=&
375: \sum_ {\mu > 0} \varphi_\mu \cos\left[\mu\pi\left(x/a+1/2\right)\right]\,,\\
376: \varphi_\mu &=&\frac{2\pi\epsilon_0 E_0}{e^2 \nu_2}
377: \frac{\sqrt{\omega^2\!\!-\!\wc^2\mu^2}}
378: {\omega\!-\!\!\sqrt{\omega^2\!\!-\!\omega_{\rm c}^2\mu^2}}
379: \sin\Big(\frac{\mu\pi}{2}\Big)J_1\Big(\frac{\mu\pi}{2}\Big)\,.
380: \nonumber
381: \end{eqnarray}
382: Eq. (\ref{eq:phi1}) has an intuitive interpretation:
383: neglecting a correction term $\varphi_{\rm bdy}(x)$ which
384: is small except for $x$ in a boundary layer of width
385: $\delta x\propto v_{\rm F}/\omega$, eq. (\ref{eq:phi1}) can be written as
386: a sum over two terms
387: \be
388: \label{eq:phi2}
389: \varphi \simeq \varphi_{\rm stat} + \varphi_{\rm dyn}
390: \ee
391: where $\varphi_{\rm stat}(x,z\!=\!0)
392: = (\epsilon_0\,q_{\rm s})^{-1} \delta\sigma(x)$
393: is the (linearised) TF potential ($q_s = e^2\nu_2/\epsilon_0$
394: is the two-dimensional TF screening vector) and
395: $\varphi_{\rm dyn}$ is a dynamical contribution
396: corresponding to a current building up the
397: screening charges. It obeys
398: $\partial^2 \varphi_{\rm dyn}/\partial x^2
399: = - (m_e \omega^2/e^2) \,\, \delta\sigma/\sigma_0
400: $
401: where $m_e$ is the electron mass and $\sigma_0$ is the areal charge
402: density of the electrons.
403:
404: Fig.~\ref{fig:2}(a) shows $\varphi(x)$
405: according to (\ref{eq:phi1}) and (\ref{eq:phi2})
406: compared with the results of a numerical RPA
407: calculation. One observes excellent agreement
408: (and $\varphi_{\rm bdy}$ is small except at the boundary).
409: Our results show that for larger frequencies
410: $(\omega > \omega_{\rm c})$,
411: the dynamical potential $\varphi_{\rm dyn}$ makes a significant contribution
412: to $\varphi$ and dynamical screening effects cannot be neglected.
413: For the absorption coefficient we obtain using
414: eqs. (\ref{eq:golden}) and (\ref{eq:phi1})
415: \be
416: \label{eq:absresult}
417: \alpha(\omega) \simeq \frac{\pi^2\hbar\epsilon_0^2a}{e^2\nu_2}\frac{\mbox{}\,\omega^2}{\mbox{}\,\wc^2}
418: \sum_{\mu> [\omega/\wc]}^{\rm odd} \frac{\sqrt{\mu^2\wc^2-\omega^2}}{\mu^2} J_1^2(\mu\pi/2)
419: \ee
420: with limiting forms
421: \be
422: \label{eq:result}
423: \alpha(\omega) =
424: \left\{
425: \begin{array}{ll}
426: \pi^2 C\hbar\epsilon_0^2 a \omega^2/(e^2 \nu_2 \wc)& \mbox{for $\omega \ll\omega_{\rm c}$}\,,\\
427: \pi\hbar \epsilon_0^2 a\omega/(4e^2\nu_2) & \mbox{for $\omega \gg\wc$}
428: \end{array}
429: \right .
430: \ee
431: and $C\simeq 0.12$. In fig.~~\ref{fig:2}(b),
432: we show quantum-mechanical RPA results in comparison
433: with eqs. (\ref{eq:absresult}) and (\ref{eq:result})
434: and observe excellent agreement.
435: We observe
436: prominent resonances in the absorption coefficient
437: near odd multiples of $\wc$,
438: due to single-particle cyclotron orbits
439: (electrons moving in phase with the external field).
440: This establishes that the single-particle
441: resonances conjectured in \cite{aus93}
442: exist within the RPA.
443: Their positions, strengths and shapes
444: are very well described by (\ref{eq:absresult}).
445: Eq. (\ref{eq:result}) and the inset of fig.~\ref{fig:2}(b) show that
446: in the limit of large frequencies ($\omega \gg \omega_{\rm c}$),
447: the absorption coefficient is linear
448: in $\omega$. In the opposite limit ($\omega \ll\wc$)
449: where the TF approach is adequate,
450: it is quadratic.
451:
452: We conclude that the quasi-classical approximation described
453: above, for the parameters considered here,
454: provides a quantitative description of
455: the optical properties.
456:
457: {\em Three-dimensional thin film.}
458: To conclude we discuss $\varphi$ for
459: a thin film \cite{note2} of width $a$ in the
460: $y$-$z$-plane subject to an external potential
461: $\varphi_{\rm ext}(\br) = E_0 x \hat{\bf e}_x$.
462: The classical charge density is concentrated
463: at the boundary,
464: $\delta\varrho_{\rm cl}(x) = \pm\delta\sigma\,\delta(x\pm a/2)$.
465: The corresponding static and dynamical contributions
466: to $\varphi(x)$ are singular;
467: we thus use the TF charge density \cite{new70} instead
468: of $\delta\varrho_{\rm cl}$
469: (appropriate in the limit of small $\ks$
470: corresponding to high electron densities):
471: $\delta\varrho_{\rm TF}(x) = \ks \epsilon_0 E_0 \sinh(\ks x)/\cosh(\ks a/2)$.
472: Here $k_{\rm s}^2 = e^2\nu_3/\epsilon_0$ is the three-dimensional
473: TF screening vector.
474: The RPA equations are easily solved
475: within a real-space discretisation. Our results for $\varphi(x)$ are shown in
476: fig.~\ref{fig:3}, and compared to results of the analytical approach
477: using eqs. (\ref{eq:golden}-\ref{eq:prop}).
478:
479: We have also calculated the absorption coeffcient
480: for $\Delta/\hbar < \omega \ll \omega_p$ within the RPA.
481: The analytical approach must be used with caution
482: in the case of the film since it requires that
483: $\varphi$ be smooth on the scale of $\lambda_{\rm F}$.
484: It turns out that $\alpha(\omega)$ is, to a good approximation, quadratic in $\omega$
485: as opposed to the two-dimensional case. As in $d=2$ dimensions we observe
486: resonances near odd multiples of $\wc$ (not shown).
487: \begin{thebibliography}{xxx}
488: \bibitem{mie08} G. Mie, Ann. Phys. {\bf 25} (1908) 377.
489: \bibitem{eka99} {\em Metal clusters}, W. Ekardt, ed.,
490: Wiley, Chichester (1999).
491: \bibitem{ber94} {\em Oscillations in finite quantum systems},
492: G. F. Bertsch and R. A. Broglia,
493: CUP, Cambridge (1994).
494: \bibitem{fer58} R. A. Ferrell, Phys. Rev. {\bf 111} (1958) 1214.
495: \bibitem{kan61} H. Kanazawa, Prog. Theor. Phys. {\bf 26} (1961) 851.
496: \bibitem{new70} D. M. Newns, Phys. Rev. B {\bf 1} (1970) 3304.
497: \bibitem{carr} G. L. Carr and S. Perkowitz,
498: in: Infrared and Millimetre Waves {\bf 13} (1985) 169.
499: \bibitem{all83} S. J. Allen, H. L. St\"ormer and J. C. M. Hwang,
500: Phys. Rev. B {\bf 28} (1983) 4725.
501: \bibitem{fet86} A. L. Fetter, Phys. Rev. B {\bf 33} (1986) 5221.
502: \bibitem{heit92} D. Heitmann, in: {\em Physics of Nanostructures},
503: Proceedings of the 38th Scottish Universities Summer School
504: in Physics, St. Andrews (1992).
505: \bibitem{efe96} K. B. Efetov, Phys. Rev. Lett. {\bf 76} (1996) 1908;
506: Y. Noat, B. Reulet, and H. Bouchiat,
507: Europhys. Lett. {\bf 36} (1996) 701;
508: Ya. M. Blanter and A. D. Mirlin, Phys. Rev. B {\bf 57} (1998)
509: 4566.
510: \bibitem{agam} O. Agam, Phys. Rev. B {\bf 54} (1996) 2607.
511: \bibitem{aus93} M. Wilkinson and E. J. Austin,
512: J. Phys. Cond. Mat. {\bf 6} (1994) 4153; {\em ibid.} {\bf 5} (1993) 8461.
513: \bibitem{meh97e}B. Mehlig and K. Richter,
514: Phys. Rev. Lett. {\bf 80} (1998) 1936.
515: \bibitem{note3}
516: In diffusive systems the TF approximation fails
517: at $\omega > \omega_{\rm D} = D/a^2$, $D$ is the diffusion constant.
518: Surprisingly, when taking into account dynamical screening effects,
519: $\alpha(\omega)$ is identical above and
520: below $\omega_{\rm D}$ and follows the classical
521: law, see U. Sivan and Y. Imry,
522: Phys. Rev. B {\bf 35} (1987) 6074 and ref. \cite{meh97a}.
523: \bibitem{boh} O. Bohigas, in {\em Chaos and quantum physics},
524: eds: M.-J. Giannoni, A. Voros and J. Zinn-Justin,
525: North-Holland (1994).
526: \bibitem{gor65} L. P. Gorkov and G. M. Eliashberg,
527: J. Exptl. Theoret. Phys. {\bf 48} (1965) 940.
528: \bibitem{ric73} M. J. Rice, W. R. Schneider and S. Str\"assler,
529: Phys. Rev. B {\bf 8} (1973) 474;
530: R. P. Devaty and A. J. Sievers,
531: Phys. Rev. B{\bf 22} (1980) 2123.
532: \bibitem{siv87} U. Sivan and Y. Imry,
533: Phys. Rev. B {\bf 35} (1987) 6074.
534: \bibitem{bla96} Ya. M. Blanter and A. D. Mirlin,
535: Phys. Rev. B {\bf 53} (1996) 12601.
536: \bibitem{wilk00} M. Wilkinson and B. Mehlig,
537: J. Phys. Cond. Mat. {\bf } (2000), in press.
538: \bibitem{wilk87} M. Wilkinson, J. Phys. A: Math. Gen. {\bf 20} (1987) 2415.
539: \bibitem{eck92} B. Eckhardt, S. Fishman, K. M{\"u}ller and D. Wintgen,
540: Phys. Rev. A {\bf 45} (1992) 3531
541: \bibitem{meh95} B. Mehlig, D. Boos{\'e} and K. M{\"u}ller,
542: Phys. Rev. Lett. {\bf 75} (1995) 57.
543: \bibitem{meh00a} B. Mehlig and M. Wilkinson,
544: Ann. Phys. (2000), in press.
545: \bibitem{ber90} G. Bertsch,
546: Comp. Phys. Comm. {\bf 60} (1990) 247.
547: \bibitem{lan84} L. D. Landau, E. M. Lifshitz and L. P.
548: Pitaevskii, {\em Electrodynamics of Continuous Media},
549: Pergamon Press (1984).
550: \bibitem{kir75} D. A. Kirzhnitz, Yu. E. Lozovik, and G. V. Shpatakovskaya,
551: Usp. Fiz. Nauk {\bf 117} (1975) 3.
552: \bibitem{meh97a} B. Mehlig and M. Wilkinson,
553: J. Phys. Cond. Mat. {\bf 9} (1997) 3277.
554: \bibitem{note} The influence of dielectrics above and below
555: the strip, or of grounded electrodes are easily
556: incorporated, see for instance \cite{fet86}.
557: \bibitem{note2} Collective effects and the nature of the plasma resonances
558: in thin films were discussed in \cite{fer58} and, within
559: the RPA, in \cite{kan61,new70}.
560:
561:
562: \end{thebibliography}
563: \narrowtext
564: \begin{figure}
565: \centerline{\psfig{file=field.eps,width=3.0cm}\hspace*{0.5cm}
566: \psfig{file=paths.eps,width=2.7cm}}
567: \vspace*{5mm}
568: \caption{\label{fig:1}
569: (a) Electric field lines (in the $x$-$z$-plane)
570: for an infinitely long metallic strip of width $a$
571: (in the $x$-$y$-plane, oriented along the $y$-axis)
572: placed in a constant external electric field
573: $\bbox{E}_{\rm ext} = -E_0 \hat{\bf e}_x$ .
574: (b) Classical paths from $x$ to $x^\prime$
575: contributing to $\Lambda_0(x,x^\prime;\omega)$.}
576: \end{figure}
577:
578: \begin{figure}
579: \centerline{\psfig{file=Fig2a.eps,width=4.2cm,clip=}
580: \hspace*{0.1cm}
581: \psfig{file=Fig2b.eps,width=3.9cm,clip=}}
582: \vspace*{0.2cm}
583: \caption{\label{fig:2}
584: Left: $\varphi(x)$ for
585: a strip of width $a=10^3 \au$ with $\rs = 1$:
586: quantum-mechanical results ($\bullet$),
587: analytical results according
588: to eq. (\ref{eq:phi1})
589: (\protect\rule[0.5ex]{0.75cm}{0.15mm})
590: and eq. (\ref{eq:phi2})
591: ($- - -$). The inset shows the correction term $\varphi_{\rm bdy}(x)$.
592: Right: Shows $\protect\mbox{Im}\, d(\omega)$
593: (for $a=10^4 \au$ and $\rs = 1$) as a function
594: of $\omega$: RPA result
595: (\protect\rule[0.5ex]{0.75cm}{0.15mm})
596: and eq.~(\protect\ref{eq:absresult}) ($- - -$).
597: The inset shows $\alpha(\omega)$ as a function
598: of $\omega$: RPA result (\protect\rule[0.5ex]{0.75cm}{0.15mm})
599: and eq.~(\protect\ref{eq:result}) ($- - -$).
600: As usual
601: $\rs=r_0/a_0$ where $a_0$ is the Bohr radius and
602: $r_0$ is the length scale defined in terms of
603: area per electron. }
604: \end{figure}
605:
606: \begin{figure}
607: \centerline{\psfig{file=Fig3.eps,width=4.5cm}}
608: \hspace*{0.5cm}
609: \caption{\label{fig:3} Shows $\varphi(x)$ for
610: a thin film of width $a=50\au$ with $\rs = 1$:
611: RPA results ($\bullet$),
612: analytical results according
613: to eq. (\ref{eq:varphi})
614: (\protect\rule[0.5ex]{0.75cm}{0.15mm})
615: and using $\varphi\simeq\varphi_{\rm stat} + \varphi_{\rm dyn}$
616: ($- - -$).}
617: \end{figure}
618: \end{multicols}
619: \end{document}
620: