1: \documentclass[12pt]{iopart}

2: \begin{document}

3: 

4: \input{psfig} 

5: 

6: \letter{Divergence of Dipole Sums and the

7:   Nature of Non-Lorentzian Exponentially Narrow Resonances in

8:   One-Dimensional Periodic Arrays of Nanospheres}

9:  

10: \author{Vadim A. Markel

11: \footnote[3]{e-mail: vmarkel@mail.med.upenn.edu}} 

12: 

13: \address{Departments of Radiology and Bioengineering, \\

14:   University of Pennsylvania, Philadelphia, PA 19104}

15: 

16: \begin{abstract}

17:   Origin and properties of non-Lorentzian spectral lines in linear

18:   chains of nanospheres are discussed. The lines are shown to be

19:   super-exponentially narrow with the characteristic width

20:   $\propto\exp[-C(h/a)^3]$ where $C$ is a numerical constant, $h$ the

21:   spacing between the nanospheres in the chain and $a$ the sphere

22:   radius. The fine structure of these spectral lines is also

23:   investigated.

24: \end{abstract}

25: 

26: \pacs{78.67.-n}

27: \submitto{\JPB}

28: %\maketitle

29: \vspace{2cm}

30: 

31: One-dimensional periodic chains (ODPC) of metallic nanospheres have

32: attracted significant recent attention due to their unusual optical

33: properties.  Although the general theoretical framework for analyzing

34: electromagnetic interactions in ODPC has been built a decade

35: ago~\cite{markel_93_1}, the recent dramatic advances in

36: nanofabrication have reinvigorated the interest in ODPC, which, in

37: turn, has led to several new results of high experimental relevancy.

38: In particular, radiatively non-decaying surface plasmons (SPs) in ODPC

39: with possible applications to building novel lasers were discussed in

40: Ref.~\cite{burin_04_1}; unusual shifts of plasmon resonance

41: frequencies were found in Ref.~\cite{zhao_03_1} and a dramatic narrowing of

42: SP spectral lines was found in Ref.~\cite{zou_04_1,zou_04_2} in finite

43: chains of moderate length. In this letter I show that two of these

44: phenomena (unusual shifts and narrowing of SP spectral lines) are

45: directly related to a logarithmic divergence of dipole sums

46: (electromagnetic eigenvalues) - a theoretical interpretation that has

47: not been given so far. The SPs that can be excited as a result of this

48: divergence posses highly unusual properties. In particular, the

49: resonance line-shapes are essentially non-Lorentzian and are

50: characterized by a vanishing integral weight. This is in a sharp

51: contrast to spectral line broadening or narrowing due to change in the

52: decay rate. Another consequence which has not been previously noticed

53: is that each narrow resonance is paired with an even more narrow

54: spectral hole. An interesting new property discussed below is that the

55: narrow collective SP resonances can be excited in ODPCs even when the

56: distance between neighboring spheres is much larger than the sphere

57: diameters. However, the resonances become so narrow in this case that,

58: unless one takes special care, it is extremely unlikely to notice them

59: in any numerical or experimental investigation.

60: 

61: \begin{figure}

62: %\psfig{file=figure1.ps,bbllx=0bp,bblly=300bp,bburx=596bp,bbury=842bp,clip=t}

63: \centerline{\psfig{file=sketch1.ps,width=8.5cm,bbllx=0bp,bblly=375bp,bburx=600bp,bbury=675bp,clip=t}}

64: \caption{Sketch of the ODPC excitation by a linearly polarized plane

65:   wave.}

66: \label{fig:sketch}

67: \end{figure}

68: 

69: We start with a brief summary of the underlying physics. The geometry

70: of a ODPC excited by a plane wave of the form ${\bf E}_0\exp(i{\bf

71:   k}\cdot{\bf r})$ is illustrated in Fig.~\ref{fig:sketch}. Here the

72: polarization vector ${\bf E}_0$ is perpendicular to the chain, the

73: incidence angle is denoted by $\theta$, lattice spacing by $h$ and the

74: sphere radius by $a$. Each sphere is assumed to be polarizable and

75: characterized by the dipole polarizability $\alpha$. We work in the

76: approximation introduced by Doyle~\cite{doyle_89_1} in which each

77: sphere is treated as an elementary dipole located at its center but is

78: characterized by non-quasistatic polarizability $\alpha$ which is

79: calculated from the coefficient $a_1$ of the Mie

80: theory~\cite{bohren_book_83}:

81: 

82: \begin{equation}

83: \label{alpha}

84: \alpha = {{3i} \over {2k^3}} {{m\psi_1(mka)\psi_1^{\prime}(ka) -

85:     \psi_1(ka)\psi_1^{\prime}(mka)} \over {m\psi_1(mka)

86:     \xi_1^{\prime}(ka) - \xi_1(ka)\psi_1^{\prime}(mka)}} \ ,

87: \end{equation}

88: 

89: \noindent

90: where $\psi_1(x)$ and $\xi_1(x)$ are the Riccati-Bessel functions,

91: $m=\sqrt{\epsilon}$ is the complex refractive index of the spheres and

92: $k=\omega/c=2\pi/\lambda$ is the wave number of the incident wave. The

93: above approximation allows one to include spheres which are not small

94: compared to the incident wavelength $\lambda$ while staying within the

95: purely dipole theory. The higher multipole interactions of the

96: spheres, as well as the input of higher multipoles to the optical

97: cross sections, are ignored in this approximation. Note that the

98: polarizability $\alpha$ defined by (\ref{alpha}) is the {\em exact}

99: dipole polarizability with respect to excitation by a plane wave, but

100: not by secondary waves scattered by spheres in the chain.  However,

101: the dipole approximation was shown to be very accurate when $h$ is the

102: order of or larger than $2a$ (which is the case discussed below) by

103: direct comparison with a converged T-matrix solution~\cite{zou_04_1}.

104: In general, it is known that short-range multipole interactions of

105: orders higher than the first (dipole) do not play a significant role

106: for transverse electromagnetic excitations of finite or infinite

107: linear arrays of interacting spheres even when the spheres are in

108: close proximity~\cite{ruppin_89_1,mazets_00_1}.  Physically, one can

109: argue that the short-range interaction is not important for transverse

110: excitations because it does not lead to an electric current along the

111: chain (in a sharp contrast to the longitudinal excitations which are

112: not discussed in the letter).

113: 

114: \begin{figure}

115: %\psfig{file=figure2.ps,bbllx=0bp,bblly=300bp,bburx=596bp,bbury=842bp,clip=t}

116: \centerline{\input{fig2a.tex}}

117: \centerline{\input{fig2b.tex}}

118: \caption{(a) Function ${\rm Re}S(x)$ for $\theta=0$.  (b) ${\rm

119:     Re}(a^3/\alpha)$ calculated for silver for different values of

120:   $a$; the curves $a={\rm 5nm}$ and $a={\rm 10nm}$ correspond to the

121:   quasi-static limit and are indistinguishable.}

122: \label{fig:x_delta}

123: \end{figure}

124: 

125: In the approximation formulated above, each sphere is characterized by

126: a dipole moment which, in the case of geometry shown in

127: Fig.~\ref{fig:sketch}, is collinear with the polarization vector

128: ${\bf E}_0$ and has the amplitude $d_n$, $n$ being the index which

129: labels spheres in the chain. The amplitudes $d_n$ are coupled to the

130: external field and to each other by the coupled-dipole

131: equation~\cite{markel_93_1}

132: 

133: \begin{equation}

134: \label{CDE}

135: d_n = \alpha\left[E_0\exp(ikn\sin\theta) + \sum_{n^{\prime}\neq n}

136:   W_{n-n^{\prime}}(kh) d_{n^{\prime}} \right] \ .

137: \end{equation}

138: 

139: \noindent

140: Here $W_{n-n^{\prime}}$ is the dipole interaction term given by

141: 

142: \begin{equation}

143: W_n(x) = k^3\left( {x^{-1} \over{\vert n\vert}} +

144:   {ix^{-2} \over{\vert n \vert^2}} - {x^{-3} \over{\vert

145:       n \vert^3}} \right) \exp(ix\vert n \vert) \ .

146: \label{W_nn}

147: \end{equation}

148: 

149: \noindent

150: The coupled-dipole equation (\ref{CDE}) is easily solved to yield 

151: 

152: \begin{equation}

153: \label{d_n}

154: d_n = {{ a^3E_0\exp(ikhn\sin\theta)} \over {a^3/\alpha - (ka)^3 S(kh)}}

155: \ ,

156: \end{equation}

157: 

158: \noindent

159: where the dimensionless quantity $(ka)^3 S(kh)$ is the dipole sum that

160: characterizes excitation of the SP with the wave number

161: $q=k\sin\theta$. The function $S(x)$ is given by

162: 

163: \begin{equation}

164: S(x) = 2 \sum_{n>0} \left({ 1 \over {xn}} + {i \over {(xn)^2}} - {1

165:     \over {(xn)^3}} \right)

166: \exp(inx)\cos(nx\sin\theta)  \ .

167: \label{S_def}

168: \end{equation}

169: 

170: \noindent

171: It can be seen that the above series diverges logarithmically when $(1 \pm

172: \sin\theta)kh = 2\pi l$, $l$ being an integer.  It is convenient to

173: separate the sum into two parts: $S(x)=S_1(x) + S_2(x)$ where $S_1$ is

174: given by

175: 

176: \begin{equation}

177: \label{S_1_def}

178: S_1(x) = 2\sum_{n>0} {{\cos(nx)\cos(nx\sin\theta)} \over {nx}}

179: =-{1 \over {2x}}\ln\left[4(\cos x - \cos(x\sin\theta))^2\right] 

180: \end{equation}

181: 

182: \noindent

183: and diverges when $\cos x = \cos(x\sin\theta)$ while $S_2(x)$ is the

184: reminder of series (\ref{S_def}) and converges for all values of

185: parameters.  For simplicity, we will assume everywhere below normal

186: incidence ($\theta=0$) which was also the case considered in

187: Ref.~\cite{zou_04_1}. Then the divergence takes place when $h=\lambda

188: l$.

189: 

190: The specific extinction cross section per one sphere is given

191: by~\cite{markel_93_1}

192: 

193: \begin{equation}

194: \label{sigma_def}

195: \sigma_e = {\rm Im} {{4\pi k a^3} \over {a^3/\alpha - (ka)^3 S(kh)}} \

196: .

197: \end{equation}

198: 

199: \noindent

200: Optical resonances occur when the real part of the denominator in the

201: above expression vanishes. In Fig.~\ref{fig:x_delta}(a) we plot ${\rm

202:   Re}S(x)$ for $\theta=0$. The sharp peaks in the plot correspond to

203: the points where ${\rm Re}S$ diverges. In Fig.~\ref{fig:x_delta}(b) we

204: also plot ${\rm Re}(a^3/\alpha)$ calculated according to (\ref{alpha})

205: for silver.  Interpolated experimental dielectric function

206: from~\cite{palik_book_v1_85} was used in calculations. It can be seen

207: that in an isolated sphere, the SP (Frohlich) resonance takes place in

208: the interval ${\rm 350 nm}<\lambda < {\rm 380 nm}$, depending on the

209: value of $a$. The resonant wavelength is obtained from ${\rm

210:   Re}(a^3/\alpha)=0$. Above the Frohlich resonance (at smaller

211: wavelengths), the spectral variable ${\rm Re}(a^3/\alpha)$ becomes

212: negative. (Here we ignore the region $\lambda<{\rm 320 nm}$ where no

213: resonance excitation can take place due to the strong interband

214: absorption.) Therefore, in order to excite a SP in an interacting ODPC

215: in this spectral region, the variable ${\rm Re}S$ must be also

216: negative. As can be seen from Fig.~\ref{fig:x_delta}(a), this happens

217: for sufficiently small values of the parameter $x=kh$ and corresponds

218: to the conventional blue shift of the transverse electromagnetic

219: oscillations, which is well known and can be described by the

220: quasistatic interaction~\cite{markel_92_1}.  However, below the

221: Frohlich resonance (at larger wavelength), the spectral parameter

222: ${\rm Re}(a^3/\alpha)$ is positive. Therefore, in order to excite

223: an SP in this spectral region, the variable ${\rm Re}S$ must be also

224: positive.  Obviously, this requirement is fulfilled near the points of

225: divergence of ${\rm Re}S$. Thus, if $h/\lambda$ is close to an

226: integer, the transverse collective oscillations of the chain are

227: shifted to the red from the Frohlich wavelength, contrary to the usual

228: case. Quite remarkably, the collective resonance can take place even

229: if $a<<h$ and $(ka)^3<<1$. Indeed, no matter how small $ka$ is, the

230: resonant condition can always be satisfied sufficiently close to the

231: point $\lambda = h/l$.  Below, we focus on the first of these

232: resonances, which corresponds to $\lambda \approx h$ and can be

233: experimentally observed in metal ODPCs in the visible and IR spectral

234: regions below the Frohlich frequency of an isolated sphere.

235: 

236: \begin{figure}

237: \centerline{\input{fig3a.tex}}

238: \centerline{\input{fig3b.tex}}

239: \centerline{\input{fig3c.tex}}

240: \centerline{\input{fig3d.tex}}

241: \caption{(a) Extinction efficiency $Q_e$ for $h=500{\rm nm}$ and

242: $a=50{\rm nm}$. (b) Illustration of cancellation of the real part of

243: the denominator in (\ref{sigma_def}). (c) Spectral dependence of the

244: imaginary part of the denominator. (d) The narrow resonance shown in

245: the panel (a) fully resolved.}

246: \label{fig:spectra}

247: \end{figure}

248: 

249: In Fig.~\ref{fig:spectra}(a) we plot the dimensionless extinction

250: efficiency $Q_e=\sigma_e/4\pi k a^3$ for the following values of

251: parameters: $a=50{\rm nm}$ and $h=500{\rm nm}$. The sharp resonance

252: corresponding to $\lambda\approx h$ is clearly visible. The curve is

253: very close to the one shown in Fig.~1(a) of Ref.~\cite{zou_04_1}. In

254: Fig.~\ref{fig:spectra}(b) the origin of the sharp resonance is

255: illustrated. Namely, we see that in a very narrow spectral interval

256: near $\lambda=500{\rm nm}$ real parts of the two terms in the

257: denominator of (\ref{sigma_def}) cancel each other. From the analysis

258: of Fig.~\ref{fig:spectra}(b), it is obvious that there should be, in

259: fact, two closely spaced narrow resonances separated by a spectral

260: hole. However, the first of the two resonances is suppressed due to

261: radiative losses. The imaginary part of the denominator of

262: (\ref{sigma_def}) is plotted in Fig.~\ref{fig:spectra}(c). As was

263: pointed out in Ref.~\cite{markel_93_1}, the imaginary part of $S(x)$

264: experiences a jump at $x=2\pi$. However, no exact cancellation of the

265: imaginary part of the denominator can take place.  For the geometry

266: considered here, it can be shown that ${\rm Im}[(ka)^3 S(kh)] >

267: -5(ka)^3/12$ while ${\rm Im}(a^3/\alpha) < -2(ka)^3/3$, so that the

268: imaginary part of the denominator is $\leq - (ka)^3/4$, with the

269: equality taking place for nonabsorbing materials with ${\rm

270:   Im}\epsilon=0$.  Finally, In Fig.~\ref{fig:spectra}(d) the sharp

271: resonance seen in Fig.~\ref{fig:spectra}(a) is completely resolved.

272: The narrow spectral hole located exactly at $\lambda=h$ can be also

273: seen in this figure.

274: 

275: Let us estimate the width and amplitude of the narrow resonances

276: occurring due to the divergence of $S(x)$. We define the width of the

277: resonance as the distance from the resonance wavelength $\lambda_r$,

278: determined from the condition ${\rm Re}[a^3/\alpha(\lambda_r) - (2\pi

279: a/\lambda_r)^3 S(2\pi h/\lambda_r)]=0$ to the center of the spectral

280: hole at $\lambda=h$; thus, $\Delta\lambda=\vert \lambda_r -

281: h\vert$. We estimate $\lambda_r$ assuming that the dominant

282: contribution to ${\rm Re}S$ comes from the logarithmically diverging

283: term (\ref{S_1_def}). We also assume that $\Delta\lambda<<2\pi$ and

284: expand the argument of the cosine in (\ref{S_1_def}) near the point

285: $x=kh=2\pi$, which leads to the following estimate:

286: 

287: \begin{equation}

288: \label{Delta_lambda}

289: \Delta\lambda \approx {h \over 2\pi} \exp\left[- {{C} \over

290:     {2(2\pi)^2}} \left({h \over a}\right)^3 \right] \ ,

291: \end{equation}

292: 

293: \noindent

294: where $C={\rm Re}[a^3/\alpha(\lambda=h)]$ is a constant of the order

295: of unity. For example, for $h=500{\rm nm}$, $C\approx 0.5$. Using

296: $a=50{\rm nm}$, we obtain from (\ref{Delta_lambda}) $\Delta\lambda

297: \approx 0.14{\rm nm}$ in agreement with Fig.~\ref{fig:spectra}(d).

298: Thus, the width of the resonance is completely determined by the

299: geometrical factors (the ratio $h/a$) and is not in any way controlled

300: by relaxation. The latter, however, influences amplitude of the

301: resonance. Indeed, the maximum value of $Q_e$ in the peak is given by

302: $1/{\rm Im}[a^3/\alpha - (ka)^3S(kh)]$. For the geometry considered

303: here, it can be verified that this value can not be greater than

304: $(h/\pi a)^2$, which is the limit for nonabsorbing material. However,

305: for strongly absorbing materials amplitude of the resonance can become

306: negligibly small.

307: 

308: Since the amplitude of the narrow resonances does not increase when

309: the width decreases super-exponentially, it is impossible to

310: effectively excite these resonances by a near-field probe. For

311: example, consider the case when a single sphere (say, $n=0$) is

312: excited by a near-field microscope tip of small aperture. Then the

313: coupled-dipole equation for the amplitudes $d_n$ can be solved by

314: Fourier transformation:

315: 

316: \begin{equation}

317: \label{d_n_Fur}

318: d_n = \int_{-\pi/h}^{\pi/h} {{a^3E_0\exp(iqhn)} \over {a^3/\alpha -

319:     (ka)^3 \tilde{S}(kh,qh)}} {h dq \over 2\pi} \ ,

320: \end{equation}

321: 

322: \noindent

323: where $\tilde{S}(kh,qh)$ is given by (\ref{S_def}) in which

324: $x\sin\theta$ in the argument of cosine must be formally substituted

325: by $qh$ and the variable $x$ in the reminder of the formula

326: substituted by $kh$. Function $\tilde{S}(kh,qh)$ diverges

327: logarithmically when $\cos(kh)=\cos(qh)$; in particular, if $kh=2\pi$

328: as in the examples considered above, the only point of divergence

329: within the integration interval is $q=0$. Similarly to resonances in

330: extinction spectra, this resonance is super-exponentially narrow in the

331: SP wavenumber $q$, and its input into the above integral is

332: negligible. We emphasize again that the resonances discussed here are

333: essentially non-Lorentzian, and the conditions for applicability of

334: the quasiparticle pole approximation, which under normal circumstances

335: would properly describe coupling of the near-field probe to SPs, are

336: severely violated.

337: 

338: The narrow resonance in Fig.~\ref{fig:spectra}(d) was obtained for

339: $h=500{\rm nm}$ and $a=50{\rm nm}$. We now show that the narrow

340: resonances also exist for smaller values of $a$ and larger ratios

341: $h/a$. To this end, we plot the extinction efficiency $Q_e$ for

342: $h=500{\rm nm}$ and $a=45{\rm nm}$ [Fig~\ref{fig:narrow}(a)] and

343: $a=40{\rm nm}$ [Fig~\ref{fig:narrow}(b)]. The dielectric function does

344: not vary noticeably over the narrow spectral range shown in

345: Fig.~\ref{fig:narrow} and was therefore taken to be constant,

346: $\epsilon=-8.5+0.76i$, which corresponds to the experimental value at

347: $\lambda=h=500{\rm nm}$ given in Ref.~\cite{palik_book_v1_85}. The

348: narrow non-Lorentzian resonances are well manifested in

349: Fig.~\ref{fig:narrow}. The central spectral hole is resolved to some

350: degree in Fig.~\ref{fig:narrow}(a) but is shown only as a vertical

351: line in Fig.~\ref{fig:narrow}(b).  Obviously, it is impossible to

352: resolve the spectral holes completely, since they do not have the

353: Lorentzian structure and are non-differentiable at the point

354: $\lambda=h$.  Note that the resonances shown in this figure are much

355: more symmetrical with respect to the point $\lambda=500{\rm nm}$ than

356: the one shown in Fig.~\ref{fig:spectra}(d). This is due to the fact

357: that with increasing the ratio $h/a$, the influence of radiative

358: losses on the shape of resonance lines decreases. For even larger

359: values of $h/a$, the resonances quickly become extremely narrow but do

360: not disappear completely, at least in chains of sufficient length.

361: 

362: It is interesting to consider the possibility of narrow resonances in

363: situations when the dipole approximation is not applicable, i.e., for

364: spheres in close proximity. It can be shown that the resonances

365: discussed in this letter do not disappear or get broadened when the

366: full multipole interaction is taken into account. Furthermore, the

367: spherical shape of the particles is also not fundamental because the

368: phenomenon discussed here originates due to long-range interaction in

369: ODPC while the higher multipole interaction is short-range. This

370: conclusion is in agreement with the numerical study of extinction

371: spectra of periodic chains of cylindrical disks~\cite{zou_04_2} which

372: were shown to have sharp resonances similar to those found in chains

373: of spheres.  Generalization to two-dimensional arrays of particles is

374: also possible. Two-dimensional periodically modulated structures have

375: also attracted significant recent attention, with the possible

376: application including random lasers~\cite{burin_04_1}, development of

377: novel chemical and biological sensors~\cite{zhao_03_1} and the study

378: of anomalous optical transmission through metal

379: films~\cite{darmanyan_04_1}.

380: 

381: To conclude this letter, we discuss several factors that contribute to

382: broadening of the spectral lines discussed above. The most important

383: factor is the finite length of a chain, since the divergence of the

384: dipole sums is logarithmic. As was mentioned above, narrow resonances

385: very close to those in infinite chains were obtained

386: in Ref.~\cite{zou_04_1} for $h/a=10$ and only 50 spheres; however,

387: observing more narrow resonances with $h/a>10$ will require a

388: substantially larger number of spheres. One possible solution to this

389: problem is to place the ODPC into a circular optical fiber.  Disorder

390: is another important factor. Numerical simulations in finite chains

391: (400 particles)~\cite{zou_04_2} revealed that random uncorrelated

392: displacements of particles with the amplitude of $\sim 0.1h$ do not

393: noticeably change the resonance lineshape. This is an expected result

394: for short-range disorder, i.e., the disorder with the correlation

395: length of one or few lattice spacings. However, disorder with

396: long-range correlations can result is much stronger changes in the

397: resonance lineshapes. Further, the account of nonlocality of the

398: dielectric response will not alter the nature of positive interference

399: (synchronism) which results in the logarithmic divergences, and is not

400: expected to broaden the narrow spectral lines.  The two physical

401: phenomena whose effects on the spectral lines discussed here are

402: difficult to predict are the nonlinearity of the optical response

403: (e.g., Kerr-type third-order nonlinearity) and quantum effects. These

404: effects will be the subject of future work.

405: 

406: \begin{figure}

407: \centerline{\input{fig4a.tex}}

408: \centerline{\input{fig4b.tex}}

409: \caption{(a) Narrow resonances for $h=500{\rm nm}$ and $a=45{\rm nm}$

410:   (a) and $a=40{\rm nm}$ (b).}

411: \label{fig:narrow}

412: \end{figure}

413: 

414: \section*{References}

415: \begin{thebibliography}{10}

416:   

417: \bibitem{markel_93_1} Markel V A 1993 {\em J. Mod. Opt.} {\bf 40}

418:   2281.

419:     

420: \bibitem{burin_04_1} Burin A L {\em et al} 2004 {\em J. Opt. Soc.  Am.

421:     B} {\bf 21} 121.

422:   

423: \bibitem{zhao_03_1} Zhao L L, Kelly K L and Schatz G C 2003 {\em J.

424:     Phys. Chem.} {\bf 107} 7343.

425:   

426: \bibitem{zou_04_1} Zou S, Janel N and Schatz G C 2004 {\em J. Chem.

427:     Phys.} {\bf 120} 10871.

428: 

429: \bibitem{zou_04_2}

430: Zou S and Schatz G C 2004 {\em J. Chem. Phys.} {\bf 121}  12606.

431: 

432: \bibitem{doyle_89_1}

433: Doyle W T 1989 {\em Phys. Rev. B} {\bf 39} 9852.

434: 

435: \bibitem{bohren_book_83}

436: Bohren C F and Huffman D R {\em Absorption and Scattering of Light by

437:   Small Particles} (John Wiley \& Sons, New York, 1983).

438: 

439: \bibitem{ruppin_89_1} Ruppin R 1989 {\em J. Phys. Soc. Japan} {\bf 58}

440:   1446 (1989).

441: 

442: \bibitem{mazets_00_1}

443: Mazets I E 2000 {\em Technical Phys.} {\bf 45} 8.

444: 

445: \bibitem{palik_book_v1_85} {\em Handbook of Optical Constants of

446:     Solids}, vol. 1, ed. by Palik E D (Academic Press, New York,

447:   1985).

448:   

449: \bibitem{markel_92_1} Markel V A 1992 {\em J. Mod. Opt.} {\bf 39}

450:   853.

451:   

452: \bibitem{darmanyan_04_1} Darmanyan S A, Neviere M and Zayats A V 2004

453:   {\em Phys. Rev. B} {\bf 70} 075103.

454: 

455: \end{thebibliography}

456: %\bibliographystyle{prsty}

457: %\bibliography{abbrevplain,article,book}

458: 

459: \end{document}

460: