1: % ****** Start of file apssamp.tex ******

2: %

3: %   This file is part of the APS files in the REVTeX 3.1 distribution.

4: %   Version 3.1 of REVTeX, September 1996.

5: %

6: %   Copyright (c) 1992 The American Physical Society.

7: %

8: %   See the REVTeX 3.1 README file for restrictions and more information.

9: %

10: %

11: %

12: % \documentstyle[preprint,eqsecnum,aps]{revtex}

13: \documentstyle[aps,prl,twocolumn,floats,graphicx]{revtex}

14: \def\btt#1{{\tt$\backslash$#1}}

15: \def\BibTeX{\rm B{\sc ib}\TeX}

16: \begin{document}

17: \draft \preprint{}

18: \title{Optical conductivity of a superfluid density wave}

19: \wideabs{

20: \author{Joseph Orenstein}

21: \address{Materials Sciences Division, Lawrence Berkeley National Laboratory and Physics Department, University

22: of California, Berkeley, California 94720}

23: \date{\today}

24: \maketitle

25: \begin{abstract}

26: We present a calculation of the low frequency optical conductivity

27: of a superconductor in the presence of quenched inhomogeneity in

28: both the superfluid and normal fluid densities.  We find that

29: inhomogeneity in the superfluid density displaces spectral weight

30: from the condensate to a frequency range that depends critically

31: on the spatial correlation of normal and superfluid density

32: fluctuations.

33: \end{abstract}

34: \pacs{74.20.-z, 74.20.-g, 74.20.De}}

35: 

36: Measurements of the optical conductivity, $\sigma(\omega)$, in the

37: microwave and terahertz regimes have proved to be a valuable probe

38: of quasiparticle dynamics in cuprate superconductors. Because even

39: relatively poor samples are in the clean limit where the

40: scattering rate is smaller than the gap frequency, or

41: $1/\tau\ll\Delta$, the spectral weight of the dissipation due to

42: quasiparticle scattering is far greater than that associated with

43: pair creation. As a consequence the quasiparticle Drude peak can

44: be clearly resolved, and its width can yield the scattering rate.

45: Important clues as to the nature of the scattering can be obtained

46: by comparing the rates measured by microwave conductivity with

47: those measured by angle-resolved photoemission (ARPES).

48: 

49: Extracting $1/\tau$ from the microwave conductivity of a cuprate

50: requires $\sigma(\omega)$ to be consistent with the two-fluid

51: model for a clean, d-wave superconductor \cite{bonn}.  In this

52: model the total conductivity contains two contributions: a

53: $\delta$-function at $\omega=0$ due to the condensate and a

54: Drude-like peak due to quasiparticle scattering. The spectral

55: weights of these components are the super and normal fluid

56: densities, respectively. As the temperature $T$ tends to zero the

57: normal fluid density, $\rho_n$, is expected to decrease linearly

58: with $T$. The superfluid density, $\rho_s$ is expected to increase

59: at the same rate, so that the total density remains constant. This

60: simple behavior permits $\rho_{n,s}$ and $1/\tau$ to be determined

61: in the superconducting state.

62: 

63: The two-fluid model provides an excellent description of the

64: microwave properties of $YBa_{2}Cu_{3}O_{7-\delta}$ (YBCO) single

65: crystals \cite{hosseini}. The scattering rates inferred from the

66: two-fluid analysis are remarkably small, approaching $\sim$ 10 GHz

67: as $T\rightarrow0$. On the other hand, analysis of ARPES yields

68: low temperature scattering rates that are larger by a factor of at

69: least 100 \cite{valla}.  One possible explanation is that the

70: single particle self-energy measured by ARPES is very different

71: from the momentum relaxation rate that determines the

72: conductivity.  However, an alternative explanation is that the

73: ARPES data are collected almost exclusively from studies of

74: $Bi_2Sr_2CaCu_2O_{8+\delta}$ (BSCCO) rather than YBCO.

75: 

76: It would seem that the microwave data on BSCCO would hold the

77: resolution to this question. However, analysis of such data has

78: been problematic because the spectra are not consistent with the

79: two-fluid model described previously. In fact the success of the

80: two-fluid description is unique to YBCO single crystals - all

81: other cuprates that have been measured show distinctly different

82: properties \cite{sflee,jacobs,pronin,waldram}. In these materials

83: the spectral weight of the Drude peak extrapolates to a large

84: residual value as $T\rightarrow0$, suggesting that a substantial

85: fraction of the normal fluid fails to condense.

86: 

87: Recently, the frequency dependence of the conductivity was

88: reported in optimally doped BSCCO \cite{corson}. The measurements

89: were performed using coherent time-domain spectroscopy, a

90: technique which can determine $\sigma(\omega)$ up to frequencies

91: $\omega/2\pi\sim$1 THz. The width of the Drude peak at low

92: temperature was determined to be $\sim$300 GHz. Integration of the

93: real part of the conductivity over the experimental range showed

94: that the uncondensed portion is at least 30$\%$ of the condensate

95: spectral weight.  A study of BSCCO for several carrier

96: concentrations showed that this percentage increases with doping,

97: reaching about 60$\%$ in an overdoped sample \cite{phd}.

98: 

99: The two-fluid description of $\sigma(\omega, T)$ in such samples

100: is not internally consistent. Fitting the spectra requires

101: $\rho_n$ to vary with frequency and actually increase with

102: decreasing $T$ for $\omega \lesssim $300GHz. On the other hand,

103: the quasiparticle conductivity could be described by physically

104: reasonable parameters if Re $\sigma$ has a third component: a peak

105: centered at $\omega$=0, whose spectral weight grows in proportion

106: to $\rho_s$ as $T$ decreases.

107: 

108: The proportionality to $\rho_s$ suggests that this anomalous part

109: of $\sigma$ is related to current fluctuations of the superfluid.

110: However, in a homogeneous superconductor the spectral weight of

111: such fluctuations decreases exponentially with $T$, which is

112: opposite to the growth of spectral weight that was observed.

113: Recently, Barabash \textit{et al.} \cite{bs} showed that a

114: component of Re $\sigma$ can increase as $T\rightarrow0$ if

115: $\rho_s$ is spatially inhomogeneous.  They considered a Josephson

116: junction network with a quenched variation $\delta J$ about the

117: mean coupling $\bar{J}$. ($J$ is the superfluid density defined on

118: a lattice rather than a continuum). They showed that the global

119: phase stiffness $J$ equals $\bar{J}-\langle \delta

120: J^2\rangle/\bar{J}$. $J$ is less than $\bar{J}$ because the phase

121: varies more rapidly in regions where the stiffness is below the

122: average value.

123: 

124: The result stated above is important to the optical conductivity

125: because the spectral weight of the condensate $\delta$-function is

126: proportional to $J$, whereas the total spectral weight of

127: condensate current fluctuations is proportional to $\bar{J}$. Thus

128: inhomogeneity of the superfluid density must displace spectral

129: weight $\sim \langle \delta J^2\rangle/\bar{J}$ from the

130: $\delta$-function to nonzero frequency. If $\delta J$ varies in

131: proportion to $\bar{J}$ then the displaced spectral weight will

132: track that of the condensate, in agreement with terahertz

133: experiments \cite{corson}.

134: 

135: The existence of strong spatial inhomogeneity in BSCCO has been

136: demonstrated by scanning tunneling microscopy (STM) measurements

137: \cite{pan}. The local density of states (LDOS) of BSCCO varies

138: randomly in space, with spatial fluctuations that have a minimum

139: wavelength of $\sim$50 $\AA$.  The variations in LDOS suggest

140: quenched inhomogeneity in local carrier concentration, $x$, and

141: therefore in the local $\rho_s$.

142: 

143: To determine whether quenched inhomogeneity in $\rho_s$ is

144: responsible for the anomaly in $\sigma(\omega)$, we need to

145: consider the spectrum of the spectral weight removed from the

146: condensate. In a system with no normal component the spectral

147: weight would be expected to appear at the natural oscillation

148: frequency of the condensate, which is the Josephson plasma

149: frequency, $\omega_s$. In optimal cuprates

150: $\omega_s/2\pi$$\sim$200 THz, whereas the anomalous dissipation is

151: found below $\sim$1 THz. Inhomogeneity can only explain the data

152: if the presence of a normal fluid in addition to the superfluid

153: causes the displaced spectral weight to appear at frequencies much

154: smaller than $\omega_s$.

155: 

156: In this paper we calculate the change in $\sigma(\omega)$ due to

157: inhomogeneity in a superconductor in which normal and superfluid

158: coexist. This is the first calculation in which the spectral

159: weight of the conductivity is consistent with the theorem of Ref.

160: \cite{bs}.  We find that $\sigma(\omega)$ is extremely sensitive

161: to correlations of the normal and superfluid density variations in

162: the the medium. The spectrum depends crucially on whether the

163: correlation is positive (such that regions of large $\rho_s$ have

164: large $\rho_n$) or negative.  If the correlation is positive, the

165: displaced spectral weight indeed shifts to the plasma frequency.

166: However, when the fluctuations of normal and superfluid density

167: are anticorrelated, the spectral weight appears at low

168: frequencies, in agreement with microwave and terahertz

169: measurements on cuprate superconductors.

170: 

171: To treat the conductivity in the presence of inhomogeneity we

172: apply the extended two-fluid phenomenology developed by Pethick

173: and Smith \cite{pethick}, and Kadin and Goldman \cite{kadin}. This

174: approach successfully describes quenched inhomogeneity at the

175: normal-superconductor interface and fluctuating inhomogeneity, as

176: in the Carlson-Goldman oscillations \cite{cg}. In the extended

177: two-fluid model the superfluid is accelerated by gradients of the

178: chemical potential as well as electric fields, that is,

179: 

180: \begin{equation}

181: \dot{\vec{J_s}}=\rho_s(\vec E-\nabla\mu_s). \label{eq:first}

182: \end{equation}

183: The chemical potential has the subscript $s$ because in a

184: superconductor $\mu$ is the energy per electron required to add a

185: pair to the condensate.  The corresponding equation for the normal

186: fluid current requires solving the Boltzmann equation for the

187: quasiparticle distribution function. However for frequencies less

188: than $1/\tau$, the distribution function is the equilibrium

189: distribution shifted by the 'quasiparticle chemical potential' or

190: $\mu_n$.  If the normal fluid is in local equilibrium with the

191: condensate $\mu_n=\mu_s$, which differs from "global" equilibrium

192: where $\mu_n$=0. In the low frequency regime the constitutive

193: relation for the normal fluid has the simple form:

194: 

195: \begin{equation}

196: \dot{\vec J_n}=\rho_n(\vec E-\nabla\mu_n)-\vec J_n/\tau.

197: \label{eq:second}

198: \end{equation}

199: 

200: A closed system of equations requires continuity relations. The

201: total charge of the superconductor separates naturally into a

202: normal component, $Q_n$ that depends on both the coherence factors

203: and the distribution function,

204: 

205: \begin{equation}

206: Q_{n}\equiv \sum_{k} q_k f_k \label{eq:fourth}

207: \end{equation}

208: where $q_k^2\equiv u_k^2-v_k^2$, and a superfluid component that

209: depends only on coherence factors,

210: 

211: \begin{equation}

212: Q_{s}\equiv \sum_{k} 2ev_k^2=2eN_{F}\mu_s. \label{eq:fifth}

213: \end{equation}

214: Under conditions for which $\mu_n$ can be defined, the normal

215: fluid charge is given by,

216: 

217: \begin{equation}

218: Q_{n}=2N_{F}\lambda(\mu_s-\mu_n),\label{eq:fourth}

219: \end{equation}

220: where $N_F$ is the normal state density of states at the Fermi

221: level. The parameter $\lambda$ relates the normal fluid charge to

222: the shift of $\mu_n$ away from local equilibrium.

223: 

224: In a superconducting medium the normal and superfluid charge are

225: not separately conserved. Interconversion of normal and

226: superconducting charge occurs through two types of processes.  In

227: the first process, $Q_n$ changes as quasiparticles recombine or

228: scatter. In the second, the quasiparticle charge changes even if

229: $f_k$ remains constant. In this process, $Q_n$ varies because the

230: quasiparticle excitation spectrum, and consequently the effective

231: charge, adjusts to the local value of $\mu_s$. Continuity

232: equations that include both types of exchange between the two

233: fluids are:

234: 

235: \begin{equation}

236: \dot{Q}_{n,s}+\nabla\cdot \vec

237: J_{n,s}=(-,+)({Q_n\over\tau_Q}-\lambda\dot{Q}_s),

238: \label{eq:fourth}

239: \end{equation}

240: where $\tau_Q$ is the rate of conversion of normal charge into

241: superfluid charge due to scattering and recombination processes.

242: The above system of equations is closed by $\nabla\cdot \vec

243: E={Q/\epsilon_0}$.

244: 

245: To see how quenched inhomogeneity affects the optical

246: conductivity, we consider the simplest possible model: a

247: one-dimensional sinusoidal variation in normal and superfluid

248: density.  We assume that the densities of the two components vary

249: as

250: $\rho_{s,n}(x)=\langle\rho_{s,n}\rangle+Re\{\rho_{sq,nq}e^{iqx}\}$,

251: where $\langle\rho\rangle$ is the average density and $\rho_q$ is

252: the amplitude of the inhomogeneity at wavevector $q$. Because the

253: medium is inhomogeneous, a uniform applied field generates a field

254: at $q$. Solving the extended two-fluid equations to lowest order

255: in $\rho_{nq,sq}$, we obtain,

256: 

257: \begin{equation}

258: {E_q\over E_0}=-{\rho_{sq}+\rho_{nq}F\over\omega_s^2

259: -\omega^2(1-\lambda)+(\omega_n^2-i\omega\lambda/\tau)F},

260: \label{eq:fourth}

261: \end{equation}

262: where,

263: 

264: \begin{equation}

265: F\equiv{\omega (\omega+i/\tau_Q^*)-v_s^2q^2/(1-\lambda)\over

266: (\omega+i/\tau)(\omega+i/\tau_Q^*)-v_n^2q^2/\lambda},

267: \label{eq:fourth}

268: \end{equation}

269: $\omega_{s,n}^2\equiv\rho_{s,n}/\epsilon_0+v_{s,n}^2q^2$,

270: $v_{s,n}^2\equiv\rho_{s,n}/2N_Fe^2$, and

271: $\tau_Q^*\equiv\tau_Q(1-\lambda)$.

272: 

273: The uniform current density in response to these fields is given

274: by $J_0=\sigma_0E_0+\sigma_qE_{-q}$, where $\sigma_0$ is the

275: uniform two-fluid conductivity, and $\sigma_q$ is the conductivity

276: that varies with wavevector $q$. In the two fluid model these are

277: given by,

278: 

279: \begin{equation}

280: \sigma_{0,q}={i\rho_{s,sq}\over\omega}+{i\rho_{n,nq}\over\omega+i/\tau}.

281: \label{eq:fourth}

282: \end{equation}

283: The effective conductivity of the medium,

284: $\sigma=\sigma_1+i\sigma_2$, is the ratio of the uniform current

285: density to the uniform field, so that,

286: $\sigma=\sigma_0+\sigma_qE_{-q}/E_0$. The second term in this

287: equation is the change in the optical conductivity due to

288: inhomogeneity, or $\Delta\sigma$.

289: 

290: The extra term in the conductivity is particularly simple if

291: $\rho_n=0$, in which case,

292: 

293: \begin{equation}

294: \Delta\sigma_2=-{\rho_{sq}\over\omega}{\rho_{sq}\over\omega_s^2-\omega^2}.

295: \label{eq:fourth}

296: \end{equation}

297: The inhomogeneity in the superfluid density indeed removes

298: spectral weight $\rho_{sq}^2/\rho_s$ from the condensate

299: $\delta$-function, in agreement with the results of Ref.

300: \cite{bs}. In the absence of a normal fluid component the spectral

301: weight reappears in a $\delta$-function at the Josephson plasma

302: frequency.

303: 

304: We next assume that $\rho_n\neq 0$ (even at $T=0$), and describe

305: how this assumption affects $\Delta\sigma(\omega)$.  We focus on

306: the behavior of $\Delta\sigma$ when the normal fluid density

307: fluctuations are either perfectly correlated or anticorrelated

308: with those of the superfluid density. We take for two-fluid

309: parameters values that are suggested by the terahertz and

310: microwave experiments: $(\rho_s/\epsilon_0)^{1/2}$=1000 THz,

311: $(\rho_n/\epsilon_0)^{1/2}$=800 THz, and $\tau^{-1}=\tau_Q^{-1}$=3

312: THz. The existence of normal fluid at $T=0$ implies a large

313: density of states at the Fermi level, $N_0$. While the origin of

314: $N_0$ is outside the scope of this paper, one may speculate that

315: it is closely related to the strong inhomogeneity observed by STM.

316: If we make the reasonable approximation of neglecting BCS

317: coherence factors for the quasiparticle states introduced by

318: disorder, then $\lambda=N_0/N_F$.

319: 

320: \begin{figure}[h]

321:      \includegraphics[width=3.0in]{SDWFig1.eps}

322: \caption{$\Delta\sigma_1$ as a function of frequency

323: ($\omega/2\pi$) for anticorrelated variations in $\rho_s$ and

324: $\rho_n$.  Spectral weight decreases for increasing $v_sq\tau$:

325: 0.25, 0.5, 1.0, and 2.0. The same curves are shown in a normalized

326: plot in the inset.} \label{fig:First}

327: \end{figure}

328: 

329: We begin with the case where the density fluctuations are

330: perfectly anticorrelated, so that $\rho_{sq}=-\rho_{nq}$ and the

331: total fluid density is uniform throughout the medium. Fig. 1 shows

332: $\Delta\sigma_1(\omega)$ with $(\rho_{sq}/\epsilon_0)^{1/2}$=100

333: THz, for several values of $v_sq\tau$. $\Delta\sigma_1(\omega)$ is

334: positive and centered at $\omega=0$ rather than $\omega_s$. The

335: spectra depend strongly on $v_sq\tau$. For $v_sq\tau\ll$1,

336: $\Delta\sigma_1$ has a Drude-like spectrum, whose width is $\sim

337: 1/\tau$. As $v_sq\tau$ increases beyond unity, the spectral weight

338: drops and a peak near the Carlson-Goldman frequency $v_sq$ appears

339: in the spectrum.

340: 

341: The key issue is the fraction of the displaced condensate spectral

342: weight that appears in the low-frequency peak, as opposed to

343: $\omega\sim\omega_s$. In Fig. 2 we compare the reduction in

344: condensate weight with the increase in dissipation at low

345: frequency. The change in condensate spectral weight was evaluated

346: from the $\lim_{\omega\rightarrow0}(\pi/2)\omega

347: \Delta\sigma_1(\omega)$. The low frequency spectral weight was

348: obtained by numerically computing the integral of $\Delta\sigma_1$

349: with respect to $\omega$ from 0 to 100 THz. Fig. 2 shows these two

350: quantities, normalized to $\rho_{sq}^2/\rho_s$, as a function of

351: $v_sq\tau$. They are equal in magnitude but opposite in sign,

352: which shows that \textit{all} of the spectral weight removed from

353: the condensate appears at low frequency and none appears at

354: $\omega_s$. Moreover, the decrease of condensate spectral weight

355: coincides exactly with the prediction of Barabash \textit{et al.}

356: \cite{bs} as $v_sq\tau\rightarrow0$, but vanishes for

357: $v_sq\tau\gg$1.

358: 

359: \begin{figure}[h]

360:      \includegraphics[width=3.0in]{SDWFig2.eps}

361: \caption{Comparison of the spectral weight displaced from the

362: condensate by the quenched variation in $\rho_s$ and the spectral

363: weight that appears in $\Delta\sigma_1$ at low frequency.}

364: \label{fig:Second}

365: \end{figure}

366: 

367: The surprising results presented above are a straightforward

368: consequence of the anticorrelation of the density fluctuations.

369: There is no dissipation near $\omega_s$ or

370: $\omega_p\equiv[(\rho_s+\rho_n)/\epsilon_0]^{1/2}$ because the

371: inhomogeneous medium is \textit{dynamically} homogeneous at high

372: frequencies.  For $\omega\sim\omega_p$ the fractional difference

373: in the conductivity of the superfluid rich and superfluid poor

374: regions is only $\sim 1/(\omega_p\tau)^2$. The anomalous

375: dissipation appears instead at low frequency where the

376: conductivity is strongly inhomogeneous. $\sigma$ is smaller in the

377: regions that are superfluid poor and normal fluid rich. $E$ will

378: be larger in such regions, which is precisely equivalent to more

379: rapid phase variation in regions of low stiffness. Thus the

380: increase in dissipation arises ultimately from an amplification of

381: the electric field in regions with greater than average $\rho_n$.

382: Finally, the dynamical inhomogeneity disappears when $\omega\gg

383: v_sq$ because the normal and superfluid response again become

384: indistinguishable in this regime.

385: 

386: \begin{figure}[h]

387:      \includegraphics[width=3.0in]{SDWFig3.eps}

388: \caption{$\Delta\sigma_1$ vs. frequency for the same parameters as

389: Fig. 1, except variations in $\rho_s$ and $\rho_n$ are positively

390: correlated.} \label{fig:Third}

391: \end{figure}

392: 

393: 

394: We are now prepared to understand the sharp difference in

395: $\Delta\sigma$ if the super and normal fluid density waves are in

396: phase.  Fig. 3 shows $Re\{\Delta\sigma(\omega)\}$ calculated with

397: identical parameters as in Fig. 1, except that $\rho_s=\rho_n$.

398: The change in conductivity is negative, and the spectra are nearly

399: independent of $q$ in contrast to the strongly $q$ dependent and

400: positive change generated when $\rho_s=-\rho_n$. The reduction in

401: conductivity is exactly as expected from the previous arguments:

402: now regions of large $E$ coincide with normal fluid poor regions

403: and the dissipation is attenuated.  The spectra are nearly

404: independent of $q$ because the dynamical inhomogeneity does not

405: tend to zero when the normal and superfluid response become

406: indistinguishable. The spectral weight is displaced equally from

407: the condensate and the normal fluid, and weight

408: 2$\rho_{sq}^2/\rho_s$ shifts to $\omega_p$.

409: 

410: We conclude with a qualitative comparison of the calculation and

411: the low frequency $\sigma(\omega)$ of the cuprates. Despite an

412: oversimplified representation of quenched disorder, the model

413: provides a natural explanation for the most surprising feature of

414: the data, the dramatic increase of the anomalous spectral weight

415: as $x$ increases from under to overdoped values. The calculation

416: shows that this can occur if the spatial correlation of $\rho_s$

417: and $\rho_n$ depends strongly on $x$. Such a change in the nature

418: of the correlations upon crossing optimal doping $x_{opt}$ would

419: indeed be expected, based on very general considerations.  For a

420: sample with $x<x_{opt}$, regions of lower than average $\rho_s$

421: will be strongly underdoped patches. These patches will be

422: insulators, or at least regions with small $\rho_n$. On the other

423: hand, the regions of low $\rho_s$ in a sample with $x>x_{opt}$

424: will be strongly overdoped, \textit{i.e}. regions of large

425: $\rho_n$.  Thus the correlation varies systematically from

426: positive to negative with increasing $x$, leading to a dramatic

427: increase in the low frequency spectral weight.

428: 

429: We acknowledge valuable discussions with D.-H. Lee, J.C. Davis, S.

430: Kivelson, A.J. Millis, and P.A. Lee.  This work was supported by

431: NSF-9870258 and DOE-DE-AC03-76SF00098.

432: \begin{references}

433: \bibitem{bonn} D.A. Bonn, Phys. Rev. B {\bf 47}, 11314 (1993).

434: \bibitem{hosseini} A. Hosseini {\em et al.}, Phys. Rev. B. {\bf 60}, 1349 (1999).

435: \bibitem{valla} T. Valla {\em et al.}, Science {\bf 285}, 2110 (1999).

436: \bibitem{sflee} S.-F. Lee {\em et al.}, Physical Review Letters {\bf 77}, 735 (1996).

437: \bibitem{jacobs} T. Jacobs {\em et al.}, Phys. Rev. Lett. {\bf 75}, 4516 (1995).

438: \bibitem{pronin} A.V. Pronin {\em et al.}, JETP Letters {\bf 68} 6432(1998).

439: \bibitem{waldram} J.R. Waldram {\em et al.}, Phys. Rev. B {\bf 59}, 1528 (1999).

440: \bibitem{corson} J. Corson {\em et al.}, Phys. Rev. Lett. {\bf 85}, 2569 (2000).

441: \bibitem{phd} J. Corson {\em et al.}, Physica B {\bf 280}, 212

442: (2000).

443: \bibitem{bs} S. Barabash {\em et al.} Phys. Rev. B {\bf 61} 14924 (2000).

444: \bibitem{pan} S. H. Pan {\em et al.} Nature {\bf 413} 282 (2001).

445: \bibitem{pethick} C.J. Pethick and H. Smith, Ann. of Phys.

446: {\bf 119}, 133 (1979).

447: \bibitem{kadin} A.M. Kadin and A.M. Goldman, in Nonequilibrium Superconductivity, ed. by D.N.

448: Langenberg and A.I. Larkin, pp. 253-324 (North-Holland, Amsterdam,

449: 1986).

450: \bibitem{cg} R.V. Carlson and A.M. Goldman, Phys. Rev. Lett. {\bf

451: 34} 67 (1976).

452: \end{references}

453: \end{document}

454: 

455: 

456: % ****** End of file apssamp.tex ******

457: