1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

2: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

3: 

4: % Some other (several out of many) possibilities

5: %\documentclass[preprint,aps]{revtex4}

6: %\documentclass[preprint,aps,draft]{revtex4}

7: %\documentclass[prb]{revtex4}% Physical Review B

8: \usepackage{graphicx}% Include figure files

9: \usepackage{dcolumn}% Align table columns on decimal point

10: \usepackage{bm}% bold math

11: 

12: 

13: \begin{document}

14: 

15: 

16: \title{The Low-Frequency Character of the Thermal Correction to the Casimir Force between

17: Metallic Films\\

18: LA-UR 02-4696}

19: 

20: \author{J.R. Torgerson and S.K. Lamoreaux}

21: 

22: \affiliation{University of California,Los Alamos National

23: Laboratory,Physics Division P-23, M.S. H803, Los Alamos, NM 87545}

24: 

25: \date{today}

26: 

27: 

28: \begin{abstract}

29: 

30: The frequency spectrum of the finite temperature correction to the

31: Casimir force can be determined by use of the Lifshitz formalism

32: for metallic plates of finite conductivity. We show that the

33: correction for the $TE$ electromagnetic modes is dominated by

34: frequencies so low that the plates cannot be modelled as ideal

35: dielectrics.  We also address issues relating to the behavior of

36: electromagnetic fields at the surfaces and within metallic

37: conductors, and calculate the surface modes using appropriate

38: low-frequency metallic boundary conditions. Our result brings the

39: thermal correction into agreement with experimental results that

40: were previously obtained.  We suggest a series of measurements

41: that will test the veracity of our analysis.

42: 

43: \end{abstract}

44: \pacs{12. 20. Ds, 41. 20. -q, 12. 20. Fv}

45: \maketitle

46: 

47: \section{Introduction}

48: 

49: A recent paper \cite{1}, in which finite conductivity and

50: temperature corrections to the Casimir force between metal plates

51: are simulatenously considered, suggests a large thermal correction

52: to the force at distances greater than about 1 $\mu$m.  This

53: correction deviates significantly from experimental results

54: \cite{2,decca1} and previous theoretical work, and has attracted

55: considerable interest. The principal conclusion in \cite{1}

56: leading to this discrepancy is that the $TE$ electromagnetic mode

57: (${\bf E}$ parallel to the surface) does not contribute to the

58: force at finite temperature. Arguments against the analysis given

59: in \cite{1} have been numerous \cite{3,4,5,6} but the arguments

60: have not been universally accepted \cite{7,8}.

61: 

62: A careful numerical analysis of the problem leads us and others to

63: conclude that the results presented in \cite{1} are mathematically

64: correct.  As we show here, this analysis does not accurately

65: represent the experimental arrangement used in \cite{2}. The

66: aspect of the problem that has not been considered in detail is

67: the appropriateness of a dielectric model of the metallic plates

68: at low frequencies, which, as we will show, are most relevant for

69: the thermal correction. The first purpose of this note is to

70: expand on our previous work \cite{9} and to point out that the

71: proper boundary conditions for conductors have not yet been

72: directly applied to this problem, and to show that the

73: experimental result \cite{2} can be fully explained by this

74: application.

75: 

76: The second purpose of this note is to contrast the points of view

77: put forward in \cite{1} and \cite{15}.  Use of the surface

78: impedance to calculate the waveguide modes, as was done in

79: \cite{15} allows description of the Casimir force by a single

80: analytic function in the complex $\omega$ plane \cite{12}.

81: Treating metals as dielectrics, as was done in \cite{1}, leads to

82: the requirement that different boundary conditions must be used

83: when the skin depth of the electromagnetic field is smaller than

84: the electron mean free path in the metal. Therefore, with the

85: dielectric treatment, the Casimir force cannot be described by a

86: single analytic function so the techniques used in \cite{1} are

87: not applicable to the problem.

88: 

89: Finally, we suggest that the analysis in \cite{1} is applicable to

90: insulating dielectrics, and possibly to materials such as

91: intrinsic or lightly dope Ge or Si where the skin depth is longer

92: than the electron mean free path.  Measurements with a dielectric

93: such as Diamond would provide an excellent test of the theory and

94: allow the possibility to discharge the surface by use of

95: ultraviolet light.  The ultimate purposes of this note are to call

96: for further theoretical studies and experimental measurements as

97: suggested here.

98: 

99: \section{Spectrum of the $TE$ Mode Thermal Correction of the Casimir Force}

100: 

101: 

102: Following Ford \cite{10}, the spectrum of the Casimir force is

103: given by Eqs. (2.3) and (2.4) of Lifshitz' seminal paper

104: \cite{11}. We note that

105: \begin{equation}

106: {1\over 2}\coth {\hbar\omega\over 2kT}={1\over 2} + {1\over

107: \exp(\hbar\omega/ kT)-1}={1\over 2} + g(\omega)

108: \end{equation}

109: and we only include the second term on the right-hand side in the

110: determination of the spectrum of the thermal correction.  From Eq.

111: (2.4) of \cite{11}, the spectrum of the $TE$ mode excitation

112: between parallel plates can be described by

113: \begin{eqnarray}\label{lif}

114: \left[{\hbar\over \pi^2 c^3}\right]F_\omega&=&\left[{\hbar\over

115: \pi^2 c^3}\right]\omega^3 g(\omega)\cr &\times &{\rm Re}\int_C

116: p^2dp

117:  \left[{(s+p)^2\over(s-p)^2}e^{-2ip\omega a/c}-1\right]^{-1},

118: \end{eqnarray}

119: \begin{equation}\label{s}

120: s=\sqrt{\epsilon(\omega)-1+p^2}

121: \end{equation}

122: where $a$ is the plate separation, and we have assumed that the

123: plates are made of the same material with vacuum between them. The

124: integration path $C$ can be separated into $C_1$ for  $p=1$ to 0,

125: which describes the effect of plane waves, and $C_2$ with pure

126: imaginary values $p=i0$ to $i\infty$ for exponentially damped

127: (evanescent) waves.

128: 

129: In anticipation that the effect is a low-frequency phenomenon, we

130: use the parameters for Au in \cite{1} for ${\rm Im}\

131: \epsilon=\epsilon_2$  and employ the Kramers-Kronig relations to

132: determine ${\rm Re}\ \epsilon=\epsilon_1$. We find for frequencies

133: $\omega<10^{14}$ s$^{-1}$ that, to good approximation,

134: \begin{equation}\label{epsilon}

135: \epsilon_1={-1.48\times 10^{4}\over 1+(\omega/\omega_0)^2};\ \

136: \epsilon_2={1.8\times 10^{18}\over \omega(1+(\omega/\omega_0)^2)}

137: \end{equation}

138: with $\omega_0=3.3\times 10^{13}\ {\rm s}^{-1}$.

139: 

140: In \cite{1}, a net deviation from the zero-temperature value of

141: the Casimir force is predicted to be about 25\% for a plate

142: separation of $1\ \mu$m at 300 K.  The experimental results

143: reported in \cite{2} had their greatest sensitivity around 1

144: $\mu$m, and disagree significantly with the results in \cite{1}.

145: As a comparison, we numerically integrate Eq. {\ref{lif}) for

146: $a=1\ \mu$m and $T=300$ K, using Eq. (\ref{epsilon}) for the

147: permittivity. The results are shown in Fig. 1, where we have

148: separated the results from the two integration paths.   In Fig.

149: 1a, it can be seen that there is no significant deviation from the

150: perfectly conducting case. On the other hand, the contribution

151: from evanescent waves, shown in Fig. 1b, is large and the

152: integrated value is in good agreement with the result given in

153: \cite{1}.

154: 

155: We see immediately that the main contributions of the $TE$-mode

156: finite conductivity correction are around $\omega=10^{10}-10^{13}$

157: s$^{-1}$.  This behavior is due to an approximately quadratic

158: increase with $\omega$ of the $C_2$ integral and a suppression

159: beginning at $\omega=kT/\hbar=4\times 10^{13}$ s$^{-1}$ due to

160: $g(\omega)$. This is a low frequency range and we can question

161: certain assumptions in \cite{1} and in the Lifshitz calculation,

162: among others, in regard to theoretical predictions relevant to the

163: experimental arrangement in \cite{2}.

164: 

165: 

166: \section{Low Frequency Limit and Field Behavior in Metallic Materials}

167: 

168: 

169: When the depth of penetration of the electromagnetic field into a

170: metal,

171: \begin{equation}

172: \delta=c/\sqrt{2\pi\mu\sigma\omega}

173: \end{equation}

174: where $\sigma$ is the conductivity and $\mu$ is the permeability

175: (for Au and Cu, $\sigma\approx 3\times 10^{17}\ {\rm s}^{-1}$,

176: $\mu=1$), becomes of the same order as the mean free path of the

177: conduction electrons, it is no longer possible to describe the

178: field in terms of a dielectric permeability \cite{12,13}. This

179: occurs for optical frequencies $\omega\approx 5\times 10^{13}$

180: s$^{-1}$ for metals such as Au and Cu where the mean free path, at

181: 300 K, is about $3\times 10^{-6}$ cm \cite{14} (p. 259). At

182: frequencies above $10^{14}$ s$^{-1}$ the permeability description

183: again becomes valid because on absorbing a photon, a conduction

184: electron acquires a large kinetic energy and has a shortened mean

185: free path. However, in the interaction of a field with a material

186: surface, $\bf E$ and $\bf H$ can always be related linearly

187: through the surface impedance (which relates the electric field at

188: the surface to a surface current hence magnetic field); this

189: approach has been used in calculation of the Casimir force

190: \cite{15}. A related correction arises from the the plasmon

191: interaction with the surface which becomes significant near the

192: plasma frequency of the metal, and has been estimated as nearly

193: 10\%\space \cite{16} for sub-$\mu$m plate separations.

194: 

195: The proper boundary conditions for a conducting plane have been

196: discussed by Boyer \cite{17}. He points out that when (using here

197: the notation of \cite{1}) $\omega\ll\eta^2\rho/4\pi$, where $\rho$

198: is the resistivity and $\eta$ is the dissipation, the usual

199: dielectric boundary conditions are not applicable. For Au, using

200: the parameters in \cite{1}, this limit is met for

201: $\omega\ll4\times 10^{14}\ {\rm s}^{-1}$. This corresponds to an

202: optical wavelength of 5\,$\mu$m, which implies that for plate

203: separations significantly larger than this, and of course for

204: $\omega\rightarrow 0$, the plates must be treated as good

205: conductors.

206: 

207: The boundary conditions for a conducting surface are discussed in

208: \cite{18} (Sec. 8.1). At low frequencies (e.g., where the

209: displacement current can be neglected), a tangential electric

210: field at the surface of a conductor will induce a current ${\bf

211: j}_\|=\sigma {\bf E}_\|$, where $\sigma$ is the conductivity. The

212: presence of the surface current leads to a discontinuity in the

213: normal derivative of ${\bf H}_\|$, hence a discontinuity in the

214: normal derivative of ${\bf E}_\|$, at the boundary of a conducting

215: surface.  These boundary conditions are quite different from the

216: dielectric case where the fields and their derivatives are assumed

217: continuous.

218: 

219: These boundary conditions are applicable when the skin depth of

220: the electromagnetic field is much smaller that the characteristic

221: wavelength of the field.  The wavelengths that contribute most to

222: the Casimir force correspond to wavevector $k\approx 1/4a$, {\it

223: independent of frequency}, by numerical determination.  When

224: $k<\sqrt{2}/\delta$, the boundary conditions are applicable. This

225: is well-satisfied over the entire frequency range of the finite

226: temperature effect for the conditions of the experiment \cite{2};

227: when $\omega>10^{11}\ {\rm s^{-1}}$ in which case $\delta< 0.7\

228: \mu$m and the relationship, for $a=1\ \mu$m

229: \begin{equation}

230: {1\over 4 a}= 2.5\times 10^3< {\sqrt{2}\over \delta}=1.4\times

231: 10^4\ {\rm cm}

232: \end{equation}

233: at the lowest frequency of interest.  In this frequency range,

234: specifying the $k$ vector in the material as a boundary condition

235: is not warranted.

236: 

237: This can also be understood by noting that the propagation of

238: electromagnetic field in a conductive material is described by the

239: diffusion equation.  If we imaging a spatially periodic varying

240: field on the surface of the material as $\exp(ikz)$, the

241: variations propagate into the material, damped exponentially as

242: $\exp(-(k^2+2/\delta^2)z)$ into the material, where

243: $\sqrt{2}/\delta$ is interpreted as the diffusion length.  We

244: therefore see that over the frequency range of interest, the

245: conducting boundary conditions are appropriate.  In this limit,

246: the displacement current is small compared to the real current,

247: ${\bf j}=\sigma {\bf E}$, for good conductors of interest here.

248: 

249: \section{Electromagnetic Modes between Metallic Plates}

250: 

251: We are interested in modes between two conducting plates separated

252: by a distance $a$. In the limit that the plates are thin films of

253: thickness $\delta$, the skin depth, we can assume that the plates

254: are infinitely thick and the problem is considerably simplified.

255: This is valid for the experiment \cite{2} where the Cu/Au metallic

256: film was  1 $\mu$m. Essentially all of the $TE$ mode thermal

257: correction comes in the $10^{11}$ and $10^{13}$ s$^{-1}$ range as

258: shown in Fig. 1, so $0.07<\delta<0.7\ \mu$m.

259: 

260: Taking the $\hat z$ axis as perpendicular to the plates, and the

261: mode propagation direction along $\hat x$, for the case of $TE$

262: modes (also referred to as $H$ or magnetic modes), $E_x=0$. The

263: plates surfaces are located at $z=0$ and $z=a$.  For a perfect

264: conductor, $\partial H_z/\partial z=0$ at the conducting surfaces.

265: A finite conductivity makes this derivative non-zero, and can be

266: estimated from the small electric field $E_y$ that exists at the

267: surface of the plate, (see \cite{18}, Sec.  8.1 and Eq. (8.6)),

268: %

269: \begin{equation}

270: \vec E_\|=\hat y E_y=\sqrt{\omega\over 8\pi \sigma}(1-i)\hat

271: n\times\vec H_\|.

272: \end{equation}

273: %

274: where $\vec H_\|= \hat x H_x$ and it is assumed that the

275: displacement current in the metal plate can be neglected

276: ($\sigma>>\omega$), and that the inverse of the mode wavenumber is

277: less than $\delta$. $E_y$ and $H_x$ are related through Maxwell's

278: equation $\vec\nabla\times\vec H=\partial \vec E/c\partial t$.

279: Assuming a time dependence of $e^{-i\omega t}$, and vacuum between

280: the plates,

281: %

282: \begin{equation}\label{diffeq}

283: {\partial H_x\over \partial z}=\pm {i\omega\over c} E_y.

284: \end{equation}

285: where $\pm$ indicates sign of $\hat n$ at $z=0$ and $z=a$

286: respectively.  The boundary conditions at the surfaces are thus

287: \begin{equation}

288: {\partial\over\partial z}H_x=\pm i\sqrt{\omega\over 8 \pi

289: \sigma}\left({\omega\over c}\right) (1-i)H_x\equiv \pm \alpha H_x.

290: \end{equation}

291: %

292: %

293: Solutions of the form $H_x(z)=Ae^{Kz}+Be^{-Kz}$, where

294: $K^2=k^2-\omega^2/c^2$ and $k$ is the transverse wavenumber, can

295: be constructed for the space between the conducting plates. The

296: eigenvalues $K$ can be determined by the requirement that Eq.

297: (\ref{diffeq}) be satisfied at $z=0$ and $z=a$. With the usual

298: substitution $\omega=i\xi$, the eigenvalues $K$ are then given by

299: (see \cite{19}, Sec. 7.2)

300: %

301: \begin{equation}\label{gte}

302: G_{TE}(\xi)\equiv{(\alpha + K)^2\over (\alpha-K)^2}e^{2Ka}-1=0

303: \end{equation}

304: %

305: and the force can be calculated by the techniques outlined in

306: \cite{19}, Sec. 7.3.

307: 

308: This result can be recast in the notation of the Lifshitz

309: formalism, and the spectrum of the thermal correction can be

310: calculated as before. Noting that $K=i\omega p/c$,

311: \begin{equation}

312: F_\omega=\omega^3g(\omega)\int_C p^2dp\left[{(\alpha+i\omega

313: p/c)^2\over(\alpha-i\omega p/c)^2}e^{-2i\omega p/c}-1\right]^{-1}.

314: \end{equation}

315: Results of a numerical integration are shown in Fig. 2, where it

316: can be seen by comparison with Fig. 1 that the metallic plate

317: boundary condition does not show a significant contribution from

318: the $C_2$ integral of the $TE$ mode thermal correction and is

319: therefore similar to that for the ``perfect conductor'' boundary

320: condition. This reconciles the discrepancy between the prediction

321: in \cite {1} and the experimental results reported in \cite{2}.

322: Note that the function Eq. (\ref{gte}) in only applicable where

323: the skin depth is small compared to the mode wavelength.  Our

324: result is in agreement, in its range of applicability, with the

325: analysis presented in \cite{15} which is valid for all

326: frequencies.  In this note, we essentially determined the surface

327: impedance from the bulk properties, which is possible in over the

328: frequency range of interest.

329: 

330: \section{Conclusion}

331: 

332: The problem of calculating the $TE$ mode contribution to the

333: Casimir force has been previously treated with the  ``Schwinger

334: prescription'' \cite{20} of setting the dielectric constant to

335: infinity before setting $\omega=0$. This prescription has become

336: controversial \cite{21}, a term that can be used to describe the

337: entire history of the theory of the temperature correction.

338: However, there is no doubt that the issues brought up in \cite{1}

339: are important.

340: 

341: The purpose of our calculation is to take a different approach and

342: to study the low-frequency behavior of the correction in order to

343: understand its character. We have shown that the finite

344: temperature correction in \cite{1} is a low-frequency phenomenon.

345: The frequency is sufficiently low so that treating the plates as

346: bulk dielectrics is not valid.  By use of a more realistic

347: description of the field interaction with the plates we show that

348: the modes between metallic plates of finite conductivity produce a

349: finite temperature correction in close agreement with the

350: perfectly conducting case.  The principal difference between our

351: result and the previous work is that we have allowed for the

352: possibility that the derivatives of the fields at the conducting

353: boundary are discontinuous.  This possibility exists because the

354: fields produce currents in the conducting plates that are

355: discontinuous across the boundary between the vacuum and the

356: conductor.  Although it is tempting to model the finite

357: conductivity as a modification to the dielectric permittivity,

358: such a model fails when the mean free path of the conduction

359: electrons exceeds the penetration depth of the electromagnetic

360: field, and thus fails for frequencies of interest for the thermal

361: correction to the $TE$ electromagnetic mode.

362: 

363: We have shown that the conducting boundary conditions that are

364: applicable for frequencies where the $TE$ mode thermal correction

365: has its significant contribution lead to a net increase of the

366: $TE$ mode force, and is of the same magnitude as the perfectly

367: conducting case.  This result is in agreement with the

368: experimental results reported in \cite{2}.  However, additional

369: and improved experiments with large plate separations (greater

370: than 2 $\mu$m) with both conducting and dielectric plates would

371: provide the definitive test.  A particularly tempting dielectric

372: would be diamond which offers both theoretical and experimental

373: benefits: its dielectric properties can be calculated from

374: first-principles, and stray surface charges can be eliminated by

375: exposing it to ultraviolet light, making it photoconductive.  A

376: semiconductor such as lightly-doped Germanium would also provide a

377: useful test of this theory.  Ge with resistivity $40\ \Omega$cm is

378: readily available and would have a skin depth about 1,000 times

379: that of Cu or Au.  Additional high-accuracy measurements at

380: long-range with Au or Cu are also important for testing the

381: theory.  We are presently constructing a new torsion pendulum

382: system that will be able to measure the Casimir force with 1\%

383: accuracy at distances greater than 2 $\mu$m, at a fixed

384: temperature of 300 K.  We hope that this note will spur further

385: theoretical work on the questions and basic analysis presented

386: here.

387: 

388: \bibliography{L2}

389: \bibliographystyle{plain}

390: 

391: 

392: 

393: \begin{figure*}

394: \includegraphics{fig1L.eps}

395: \caption{The net finite-temperature contribution to the Casimir

396: force is determined by $F=(\hbar/\pi c^3)\int_0^\infty F_\omega

397: d\omega$ and is attractive when $F>0$. a: The two curves represent

398: the $C_1$ path for perfectly conducting plates (dashed curve) and

399: for plates with permittivity given by Eq. (\ref{epsilon})(solid

400: curve). The net force force for the latter is $0.95$ times the

401: perfectly conducting case. b: For a perfect conductor, the $C_2$

402: integral is zero. The net contribution from the $C_2$ path is

403: $-169$ times the perfectly conducting contribution from the $C_1$

404: path, and its addition to the $TE$ mode zero-point contribution

405: reduces the net $TE$ mode force to nearly zero, which is the

406: result obtained in \cite{1}. All are for $a=1\ \mu$m, $T=300$ K.}

407: \end{figure*}

408: 

409: \begin{figure*}

410: \includegraphics{fig2L.eps}

411: \caption{Numerical results for $F_\omega$ using the finite

412: conductivity boundary conditions.  The integrated force for the

413: $C_2$ path contribution is $1.47$ times greater than the $C_1$

414: integration, and the total net force for both paths is $1.75$

415: times greater than the perfectly conducting case. Treatment of the

416: plates as conducting metals fails above $\omega =10^{14}\ {\rm

417: s}^{-1}$. All are for $a=1\ \mu$m, $T=300$ K.}

418: \end{figure*}

419: 

420: \end{document}

421: