1: %User: cond-mat/0210171, Password: vez7t

2: %LX8100

3: %Second version was done on Jan. 12, 2003.

4: 

5: \documentclass[twocolumn,showpacs,preprintnumbers,prb]{revtex4}

6: 

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

8: 

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

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

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

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

13: \newcommand{\BE}{\begin{equation}}

14: \newcommand{\EE}{\end{equation}}

15: \newcommand{\BA}{\begin{eqnarray}}

16: \newcommand{\EA}{\end{eqnarray}}

17: \textheight 9.75in

18: %\textwidth 6.7in \oddsidemargin -0.25in

19: %\topmargin -0.5in

20: 

21: \usepackage{graphicx}% Include figure files

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

23: \usepackage{bm}% bold math

24: %\preprint{cond-mat/0210171}

25: \begin{document}

26: 

27: \title{Localization of electromagnetic waves in two-dimensional random dielectric systems}

28: 

29: \author{Bikash C. Gupta} \author{Yu-Yu Chen} \author{Zhen

30: Ye}\email{zhen@phy.ncu.edu.tw} \affiliation{Wave Phenomena

31: Laboratory, Department of Physics, National Central University,

32: Chungli, Taiwan 32054}

33: 

34: %\date{February 2, 2002}

35: \date{January 14, 2003}

36: 

37: \begin{abstract}

38: 

39: We rigorously calculate the propagation and scattering of

40: electromagnetic waves by rectangular and random arrays of

41: dielectric cylinders in a uniform medium. For regular arrays, the

42: band structures are computed and complete bandgaps are discovered.

43: For random arrays, the phenomenon of wave localization is

44: investigated and compared in two scenarios: (1) wave propagating

45: through the array of cylinders; this is the scenario which has

46: been commonly considered in the literature, and (2) wave

47: transmitted from a source located inside the ensemble. We show

48: that within complete band gaps, results from the two scenarios are

49: similar. Outside the gaps, however, there is a distinct

50: fundamental difference, that is, waves can be blocked from

51: propagation by disorders in the first scenario, but such an

52: inhibition may not lead to inhibition or wave localization in the

53: second scenario. The study suggests that the traditional method

54: may be ambiguous in discerning localization effects.

55: 

56: \end{abstract}

57: 

58: \pacs{42.25.Hz, 41.90.1e, 71.55.Jv} \maketitle

59: 

60: Wave localization is a peculiar property of random media that

61: completely block wave propagation due to multiple scattering, thus

62: inducing a surprising phase transition, for example, in optical or

63: acoustic transparency or electrical conductivity. When localized,

64: waves remain confined in space until dissipated.

65: 

66: More than two decades have passed since the phenomenon of wave

67: localization was explored for propagation of electromagnetic (EM)

68: waves in random media. During this period, a great body of

69: literature has been generated\cite{reviews}. And the interest in

70: the subject continues to grow even further in recent

71: years\cite{weak,Marian,Sigalas,Wiersma,AAA,AAC,debate,Emile}.

72: Despite the efforts, however, some important problems still remain

73: unsolved.

74: 

75: The first issue is with the way in which the localization effect

76: is investigated. To date, claims of localization have been based

77: on observations of the exponential decay of waves as they

78: propagate {\it through} disordered media\cite{AAC}. That is, in

79: most previous experimental or theoretical studies, the apparatus

80: was set up in such a manner that waves were transmitted at one end

81: of a scattering ensemble, then the scattered waves were recorded

82: on the other end to measure the transmission through the sample.

83: The results were then compared with the theory to infer the

84: localization effect. In this method, it is quite plausible that

85: other effects such as reflection and deflection due to the

86: presence of boundaries may also attenuate waves, resulting in a

87: similar decay in transmission and obscuring the data

88: interpretation. Therefore it is desirable to look for a unique

89: feature which can differentiate localization from other effects;

90: the inability to discriminate the localization effect from other

91: effects has caused significant debate in the

92: literature\cite{debate}.

93: 

94: Second, although it has been suggested a while ago that the

95: regions of localized states coincide with the positions of the

96: gaps, the relation between localization in random media and

97: bandgaps of the corresponding regular systems is still

98: inconclusive\cite{Sigalas,Datta}.

99: 

100: Third, it has been the prevailing view over the past twenty years

101: that all EM waves are localized in two dimensions (2D) for any

102: given amount of disorder, following the scaling analysis of

103: electronic systems\cite{gang4}. Recently, there is an intensive

104: debate on this view from new experiments\cite{Pudalov,Shashkin2},

105: as reviewed in \cite{EA}. Since localization in electronic and EM

106: systems has the same physical origin, it is therefore imperative

107: to re-look at the view that all EM waves are always localized in

108: 2D random systems. This task may be difficult, due to the obvious

109: limitation of the finite sample size for either numerical or

110: experimental workers, but at least one may examine whether the

111: phenomenon of localization has been explored in a proper way in

112: the past.

113: 

114: With this Letter, we wish to shed new light on these questions.

115: Here, we present a rigorous study of EM wave scattering and

116: propagation in media containing many dielectric cylinders. The

117: approach is based upon the self-consistent theory of multiple

118: scattering\cite{Twersky} and has been used previously to study

119: acoustic localization in liquid media\cite{Emile} and acoustic

120: attenuation by rigid cylinders in air\cite{Chen}. Wave propagation

121: is expressed by a set of coupled exact equations and is solved

122: rigorously. We show that wave localization can be achieved in

123: ranges of frequencies, coincident with yet wider than the complete

124: bandgap. For the phenomenon of wave localization, we compare two

125: scenarios by analogy with the acoustic case\cite{Scaling3}: (1)

126: the traditional setup of probing localization both numerically and

127: experimnentally, as stated in, e.~g. Ref.~\cite{Sigalas,AAC}, that

128: is, wave propagating through the array of cylinders, and (2) wave

129: transmitted from a source located inside the ensemble. We show

130: that within complete band gaps, results from the two scenarios are

131: similar, whereas there is a distinct qualitative difference

132: outside the gap. Moreover, when localized, not only are waves

133: confined near the transmitting source but a unique collective

134: phenomenon emerges, illustrated by a phase diagram in analogy to

135: the acoustic system\cite{Emile}.

136: 

137: The system considered here is similar to what has been presented

138: in \cite{Sigalas}. Assume that $N$ uniform dielectric cylinders of

139: radius $a$ are placed in parallel in a uniform medium,

140: perpendicular to the $x-y$ plane. The arrangement can be either

141: random or regular. For brevity, we only consider the case of the

142: E-polarization, i.~e. the E-field parallel to the z-direction. The

143: qualitative features for both E- and H-polarizations are similar.

144: The scattering and propagation of EM waves can be solved by using

145: the exact formulation of Twersky\cite{Twersky}. While the details

146: can be found in \cite{Chen}, here we brief the main procedures. A

147: unit pulsating line source transmitting monochromatic waves is

148: placed at a certain position. The scattered wave from each

149: cylinder is a response to the total incident wave, which is

150: composed of the direct contribution from the source and the

151: multiply scattered waves from each of the other cylinders. The

152: response function of a single cylinder is readily obtained in the

153: form of the partial waves by invoking the usual boundary

154: conditions across the cylinder surface. The total wave ($E$) at

155: any space point is the sum of the direct wave ($E_0$) from the

156: transmitting source and the scattered wave from all the cylinders.

157: The normalized field is defined as $T \equiv E/E_0$; thus the

158: trivial geometrical spreading effect is eliminated.

159: 

160: \input{epsf}

161: \begin{center}

162: \begin{figure}[h]

163: %\vspace{10pt}

164: \epsfxsize=2.25in\epsffile{figure1.eps}

165: \caption{Left panel: The band structures computed by the plane

166: wave expansion method. Right panel: Here is shown the normalized

167: transmission $\log_{10} |T|^2$ versus frequency; the solid line

168: refers to the result from the [10] direction propagation, and the

169: dotted line to that from the [11] direction propagation lines.}

170: \label{fig1}

171: \end{figure}

172: \end{center}

173: 

174: In line with \cite{Sigalas}, the following parameters are used in

175: the computation. The ratio of the dielectric constant between the

176: cylinders and the hosting medium is 10; the dielectric constant of

177: the medium is taken as one. The filling factor $\beta$, the

178: fraction of area occupied by the cylinders per unit area, is 0.28.

179: The radius $a$ of the cylinders is 0.38 cm. The lattice constant

180: $d$ of the corresponding square lattice array of the cylinders is

181: thus about 1.28 cm ($d = a\sqrt{\pi/\beta}$). For convenience, we

182: scale all lengths by the lattice constant $d$. The computation is

183: continued until the convergence is reached.

184: 

185: First, in Fig.~\ref{fig1} we show the band structure of the

186: corresponding square lattice arrangement of the cylinders,

187: obtained by the plane wave method. The wave transmission in two

188: symmetric directions is also shown. Two complete bandgap regions

189: are identified and are consistent with the highly attenuated

190: regions in the transmission computation. These results are also

191: consistent with that in Fig.~4 of \cite{Sigalas}, thereby

192: verifying our numerical scheme.

193: 

194: To investigate wave localization, two situations are considered

195: and compared: (1) wave propagating through the array of cylinders,

196: labeled hereafter as the `Outside' situation that imitates the

197: traditional experimental\cite{Wiersma} and theoretical

198: setups\cite{Sigalas,AAA}, and (2) wave transmitted from a source

199: located inside the ensemble, labeled hereafter as the `Inside'

200: situation. Both cases are illustrated by Fig.~\ref{fig2}. For the

201: `Outside' case, all cylinders are randomly or regularly placed

202: within a rectangular area with length $L$ and width $W$. The

203: transmitter and receiver are located at some distance from the two

204: opposite sides of the scattering area. For the `Inside' situation,

205: all cylinders are placed either completely randomly or regularly

206: within a circle of radius $L$ with the source located at the

207: center and the receiver located outside the scattering cloud.

208: 

209: \input{epsf}

210: \begin{center}

211: \begin{figure}[hbt]

212: %\vspace{10pt}

213: \epsfxsize=2.25in\epsffile{figure2.eps} \caption{(a) The `Outside'

214: case: Electromagnetic propagation through a cloud of dielectric

215: cylinders. (b) The `Inside' case: Electromagnetic transmission

216: from a line source located inside the array of dielectric

217: cylinders.} \label{fig2}

218: \end{figure}

219: \end{center}

220: 

221: The frequency response of the averaged logarithmic transmission is

222: presented in Fig.~\ref{fig3} for both `Inside' and `Outside'

223: scenarios. Here we see that the disorder somewhat tends to enhance

224: transmission within the bandgaps for both scenarios, while

225: obviously reduces the transmission for all frequencies outside the

226: gaps in the `Outside' case. For the `Inside' situation, however,

227: the reduction for regions outside the gaps is not generally

228: obvious, and is only seen near the gap edges. It has been

229: suggested in the literature that the transmission reduction in the

230: `Outside' scenario indicates wave localization. For example, the

231: authors in \cite{Sigalas,AAA} computed the transmission in the

232: context of the `Outside' scenario, and subsequently obtained the

233: localization length for all frequencies. We find that this

234: approach towards localization may not be appropriate. The reasons

235: follow.

236: 

237: If the transmission reduction in the `Outside' scenario is only

238: caused by the localization effect, it will be implied that the

239: random system only supports localized states. Then waves will not

240: be allowed to propagate not only through but also inside the

241: system. Therefore we would expect the transmission to follow an

242: exponential decay with increasing sample size for both `Inside'

243: and `Outside' setups.

244: 

245: \input{epsf}

246: \begin{center}

247: \begin{figure}[hbt]

248: %\vspace{10pt}

249: \epsfxsize=2.0in\epsffile{figure3.eps} \caption{Transmission

250: versus frequency for both random and regular arrays of cylinders:

251: (a) the `Outside' case with W = 6 and L = 10; and (b) the `Inside'

252: case with L = 10. Please refer to Fig.~\ref{fig2} and the text for

253: the explanation about the `Outside' and `Inside' cases.}

254: \label{fig3}

255: \end{figure}

256: \end{center}

257: 

258: Fig.~\ref{fig4} presents the results for the random ensemble

259: averaged transmission and its fluctuation as a function of the

260: sample size at two frequencies. The sample size is varied by

261: adjusting the number of the cylinders. For the `Outside' case, we

262: have done the following to remove the effect of the width $W$.

263: With a fixed sample size (i.~e. the length $L$), we plot the

264: transmission versus width. We find that the transmission is very

265: nicely saturating to a certain value in an exponential manner. We

266: have done for several lengths, and obtained the corresponding

267: saturated value for each length. Then we plot these values versus

268: sample lengths. As an example, the results for 8.64 GHz are shown

269: in Fig.~\ref{fig4} (e) and (f). For 6.54 GHz, the localization is

270: strong, the width effect diminishes very quickly when the width

271: increases. Here the plot for 6.54 GHz has width 26 in the

272: `Outside' scenario. Note that the width should not be started at a

273: value too close to zero; otherwise the variance will be too large,

274: making the results unstable. The average has been taken for 500

275: configuration to ensure the stability.

276: 

277: A few important features are discovered. For the frequency of 6.54

278: GHz (within the first gap), the transmission decays exponentially

279: with the sample size for both `Outside' and `Inside' situations

280: with almost the same slop of -1.35, suggesting that at this

281: frequency, waves are localized. And inside the localization

282: regime, the absolute value of the transmission fluctuation is

283: small, as expected from an earlier work \cite{Emile}. Here we see

284: that within the localization regime, wave localization can be

285: indeed observed in both `Outside' and `Inside' scenarios.

286: 

287: For 8.64 GHz (between the first and the second gaps), the

288: `Outside' and `Inside' scenarios differ significantly. For the

289: `Outside' case the transmission decreases exponentially with a

290: slop of -0.0612 along the path. If this exponential decay is

291: caused by localization, then we should also observe the

292: exponential decay for the same sample size ($L$) in the `Inside'

293: scenario. The result in the center panel of Fig.~\ref{fig4}

294: clearly does not support this point of view. Instead,

295: Fig.~\ref{fig4} tends to indicate that waves are not yet localized

296: at 8.64 GHz in the `Inside' scenario. The fact that the

297: exponential decay only occurs in one scenario but not in the other

298: for the same sample size ($L$) is itself intriguing and important.

299: Therefore we may conclude that the `Outside' scenario is

300: inappropriate in isolating the localization effect, and it would

301: be a mistake to interpret the exponential decay or transmission

302: reduction shown in the `Outside' situation as a conclusive

303: indication of wave localization. Furthermore, as at this

304: frequency, waves are not yet localized in the `Inside' case and

305: they have a weaker exponential decay in the `Outside' case, the

306: transmission will be more sensitive to the arrangement of the

307: cylinders. Therefore the fluctuation at this frequency is stronger

308: than that at 6.54 GHz. However, the ratio between the fluctuation

309: and the transmission at 8.64 GHz can be smaller than that at 6.54

310: GHz.

311: 

312: \input{epsf}

313: \begin{center}

314: \begin{figure}[hbt]

315: %\vspace{10pt}

316: \epsfxsize=3in\epsffile{figure4.eps} \caption{The averaged

317: logarithmic transmission and its fluctuation versus the sample

318: size for two frequencies: one is within the first bandgap and the

319: other is above the first but below the second gap. The left and

320: center panels refer to the `Outside' and `Inside' cases

321: respectively. The estimated slops for the transmission are

322: indicated in the figure. The right panel shows the effect of width

323: $W$ and the plot of the transmission versus length $L$ at the

324: extrapolated infinite width (see the text).} \label{fig4}

325: \end{figure}

326: \end{center}

327: 

328: To this end, a few notes are appropriate. We have also examined

329: other frequencies in general and two in particular: one is within

330: the second gap and the other is above the second gap. The results

331: are very similar to that shown in Fig.~\ref{fig4}. For brevity, we

332: will not show the results here. From Fig.~\ref{fig3}, the fact

333: that the transmission reduction occurs not only within but also

334: outside the gaps (at areas around the edges of the gaps) indicates

335: that the localized regions are coincident with the complete

336: bandgaps, and these regions seem wider than the gaps. Our results

337: show that although the disorders may block waves from propagation

338: {\it through} the medium, but they may not yet localize the waves

339: inside a 2D system.

340: 

341: Now we discuss a unique feature of EM wave localization. The

342: energy flow of EM waves is $\vec{J} \sim \vec{E}\times\vec{H}$. By

343: invoking the Maxwell equations to relate the electrical and

344: magnetic fields, we can derive that the time averaged energy flow

345: is $<\vec{J}>_t \equiv \frac{1}{T}\int_0^T dt \vec{J} \sim

346: |\vec{E}|^2\nabla\theta,$ where the electrical field is written as

347: $\vec{E} = \vec{e}_E |\vec{E}|e^{i\theta}$, with $\vec{e}_E$

348: denoting the direction, $|\vec{E}|$ and $\theta$ being the

349: amplitude and the phase respectively. It is clear that when

350: $\theta$ is constant, at least by spatial domains, while

351: $|\vec{E}| \neq 0$, the flow would come to a stop and the energy

352: will be localized or stored in the space. We assign a unit phase

353: vector, $\vec{u} = \cos\theta_i\vec{e}_x + \sin\theta_i\vec{e}_y$

354: to the oscillation phase $\theta_i$ of the dipoles. Here

355: $\vec{e}_x$ and $\vec{e}_y$ are unit vectors in the $x$ and $y$

356: directions respectively. These phase vectors are represented by a

357: phase diagram in the $x-y$ plane.

358: 

359: \input{epsf}

360: \begin{center}

361: \begin{figure}[h]

362: %\vspace{10pt}

363: \epsfxsize=2.25in\epsffile{figure5.eps} \caption{The phase diagram

364: and spatial distribution of electromagnetic energy for two

365: frequencies for one random configuration. Left panel: the phase

366: diagram for the phase vectors defined in the text; here the phase

367: of the direct field $E_0$ is set to zero. Right panel: the energy

368: spatial distribution.} \label{fig5}

369: \end{figure}

370: \end{center}

371: 

372: In Fig.~\ref{fig5}, the two-dimensional spatial distribution of EM

373: energy ($\sim |E/E_0|^2$) and the phase vectors of the E-field are

374: plotted for the two frequencies discussed in Fig.~\ref{fig4}. The

375: phase vectors are located randomly in the $x-y$ plane but to avoid

376: the positions of the cylinders. The `Inside' scenario is

377: considered. Here we clearly see that for 6.54 GHz, the energy is

378: mainly confined near the source, consistent with Fig.~\ref{fig4}.

379: The phase vectors are orderly oriented. These fully comply with

380: the above general discussion. Therefore at this frequency, EM wave

381: is indeed localized. When we increasingly add an imaginary part to

382: the dielectric constant, the ordered orientation of the phase

383: vectors will disappear, confirming that the phase coherence is a

384: unique feature of EM wave localization. We note from

385: Fig.~\ref{fig5} that near the sample boundary, the phase vectors

386: start to point to different directions. This is because the

387: numerical simulation is carried out for a finite sample size. For

388: a finite system, the energy can leak out at the boundary,

389: resulting in disappearance of the phase coherence. When enlarging

390: the sample size, we observe that the area showing the perfect

391: phase coherence will increase. At 8.64 GHz, however, there is no

392: ordering in the phase vectors $\vec{u}(\vec{r})$. The phase

393: vectors point to various directions. The energy distribution is

394: extended in the $x-y$ plane, and no EM wave localization appears,

395: in agreement with what has been described for Fig.~\ref{fig4}.

396: 

397: In summary, we have examined some fundamental problems of EM wave

398: localization in 2D. Although it may be still hard to conclude that

399: extended waves are possible in 2D random media, as limited by the

400: finite sample size, the present results do indicate that the

401: traditional method may be unable to isolate the localization

402: effect. It is also shown that the localization region is related

403: to and seems to be wider than the complete bandgaps. When

404: localized, not only are waves confined near the transmitting

405: source but a unique collective phenomenon emerges.

406: 

407: This work is supported by the National Science Council of Republic

408: of China.

409: 

410: \begin{thebibliography}{99}

411: 

412: \bibitem{reviews} For a review, please refer to S. John, Phys.

413: Today {\bf 40}, 32 (1991); {\it Scattering and localization of

414: classical waves in random media}, edited by P. Sheng (World

415: Scientific, Singapore, 1990); A. Lagendijk and B. A. van Tiggelen,

416: Phys. Rep. {\bf 270}, 143 (1996).

417: 

418: \bibitem{weak} Y. Kuga, A. Ishimaru, J. Opt. Soc. Am. A1, 831 (1984);

419: M. P. van Albada, A. Lagendijk, Phys. Rev. Lett. 55, 2692 (1985);

420: P. E. Wolf, G. Maret, Phys. Rev. Lett. 55, 2696 (1985); A. Tourin,

421: et al., Phys. Rev. Lett. 79, 3637 (1997); M. Torres, J. P.

422: Adrados, F. R. Montero de Espinosa, Nature {\bf 398}, 114 (1999).

423: 

424: \bibitem{Marian} M. Rusek and A. Orlowski, Phys. Rev. E{\bf 51},

425: R2763 (1995).

426: 

427: \bibitem{Sigalas} M. M. Sigalas, C. M. Soukoulis, C.-T. Chan, and

428: D. Turner, Phys. Rev. B{\bf 53}, 8340 (1996).

429: 

430: \bibitem{Wiersma} D. S. Wiersma, P. Bartolini, A. Lagendijk, and R.

431: Roghini, Nature {\bf 390}, 671 (1997).

432: 

433: \bibitem{AAA} A. A. Asatryan, et al. Phys. Rev. B{\bf 57}, 13535

434: (1998).

435: 

436: \bibitem{AAC} A. A. Chabanov, M. Stoytchev, and A. Z. Genack, Nature {\bf

437: 404}, 850 (2000).

438: 

439: \bibitem{debate} F. Scheffold, R. Lenke, R. Tweer, and G. Maret,

440: Nature {\bf 398}, 206 (1999);  D. Wiersma, et al. Nature, {\bf

441: 398}, 207 (1999).

442: 

443: \bibitem{Emile} E. Hoskinson and Z. Ye, Phys. Rev. Lett. {\bf 83}, 2734

444: (1999).

445: % Z. Ye and E. Hoskinson, Appl. Phys. Lett., {\bf 77}, 4428 (2000).

446: 

447: \bibitem{Datta} C. M. Soukoulis, S. Datta, and E. N. Economou,

448: Phys. Rev. B{\bf 49}, 3800 (1994).

449: 

450: \bibitem{gang4} E. Abrahams, P. W. Anderson, D. C. Licciardello, and

451: T. V. Ramakrishnan, Phys. Rev. Lett. {\bf 42}, 673 (1979).

452: 

453: \bibitem{Pudalov} V. M. Pudalov, M. D'Iorio, S. V. Kravchenko, and

454: J. W. Campbell, Phys. Rev. Lett. {\bf 70}, 1866 (1993).

455: 

456: %\bibitem{Shashkin} A. A. Shashkin, G. V. Kravchenko, and V. T.

457: %Dolgopolov, JETP Lett. {\bf 58}, 220 (1993).

458: 

459: \bibitem{Shashkin2} A. A. Shashkin,  V. T. Dolgopolov and G. V.

460: Kravchenko, Phys. Rev. B{\bf 49}, 14486 (1994).

461: 

462: \bibitem{EA} E. Abrahams, S. V. Kravchenko, and M. P. Sarachik, Rev.

463: Mod. Phys. {\bf 73}, 251 (2001).

464: 

465: \bibitem{Twersky} V. Twersky, J. Acoust. Soc. Am. {\bf 24}, 42

466: (1951).

467: 

468: \bibitem{Chen} Y.-Y. Chen and Z. Ye, Phys. Rev. Lett. {\bf 87},

469: 184301 (2001).

470: 

471: \bibitem{Scaling3} Y.-Y. Chen and Z. Ye, Phys. Rev. E{\bf 65},

472: 056612 (2002).

473: 

474: 

475: \end{thebibliography}

476: 

477: \end{document}

478: 

479: \newpage

480: 

481: \begin{center}

482: 

483: {\bf \large Correspondence from the Editor and the referees}

484: \end{center}

485: 

486: \section{Editor's correspondence (received: Saturday, January 04, 2003 4:15 AM}

487: 

488: \begin{center}

489: {\bf The following comments are exactly copied from the original

490: correspondence from the editor.}

491: \end{center}

492: 

493: \subsection{Letter from the Editor}

494: 

495: Re: LX8100\\

496:     Localization of electromagnetic waves in two-dimensional random

497:     dielectric systems\\

498:     by Bikash C. Gupta, Yu-Yu Chen, and Zhen Ye

499: 

500: Dr. Zhen Ye Wave Phenomena Lab. Dept. of Physics Natl. Central

501: Univ. Chungli, TAIWAN 32054

502: 

503: 

504: Dear Dr. Ye,

505: 

506: The above manuscript has been reviewed by our referees.

507: 

508: A critique drawn from the reports is enclosed.  On this basis, we

509: judge that the paper is not appropriate for Physical Review

510: Letters, but might be suitable for publication in another journal,

511: possibly with some revision.  Therefore, we recommend that you

512: submit your manuscript elsewhere.

513: 

514: 

515: Yours sincerely,

516: 

517: Orla Feeney\\ Assistant Editor Physical Review Letters \\

518: Email:

519: prl@aps.org Fax: 631-591-4141 http://prl.aps.org/

520: 

521: 

522: \subsection{Report of Referee A -- LX8100/Gupta}

523: 

524: The authors performed numerical simulation of wave transport in 2D

525: random dielectric systems. I do not think their simulation results

526: presented in this paper justify their conclusions.

527: 

528: (i) One main conclusion of this paper is that the waves are not

529: always localized in 2D system (in contrast to the previous

530: assertion as they said). It is well known that the previous

531: assertion is for infinitely large system, namely, light is always

532: localized in infinitely large 2D system. But the 2D system that

533: the authors calculated has finite size, actually its size is quite

534: small. I estimated the localization length l from the slop in Fig.

535: 4. At 8.64 GHz (outside the gap), the slope of 0.1648 gives l

536: $\sim$ 1/0.16 $\sim$ 6. In the Outside case, the system length L =

537: 6 or 10. So the system length is comparable to the localization

538: length. That is why the wave is not localized: the system is too

539: small! If the authors increase the system s length L and width W

540: to over 100, they should observe exponential decay in the

541: transmission.

542: 

543: (ii) Another claim in this paper is that the traditional method is

544: unable to isolate the localization effect. This claim comes from

545: authors calculation of transmission through the array of cylinders

546: (the Outside scenario). But in the traditional method, the

547: transmission is calculated through a slab of infinite width. In

548: authors calculation, the width of their system is short: W = 6.

549: Since W is comparable to the local length l at 8.64 GHz (outside

550: the gap), light may transmit the finite system through the surface

551: modes. This would give error to the transmission calculation. The

552: authors should increase W to be much larger than the localization

553: length and redo their calculation before claiming the traditional

554: method does not work.

555: 

556: Some minor points:

557: 

558: (i) In the second paragraph on page 3,  For brevity, yet without

559: losing generality, we only consider the case of the

560: E-polarization. It is well known that it is much harder to

561: localize light wave than electron wave because the former is

562: vector field. By considering only the E-polarization, the authors

563: made the approximation that the wave is scalar field. So their

564: calculation loses the generality for light localization.

565: 

566: (ii) The top label in Fig. 4 is incorrect. (a) should be Inside,

567: (b) should be Outside.

568: 

569: (iii) Since wave is more localized at 6.54 GHz (inside the gap)

570: than at 8.64 GHz (outside the gap), the fluctuation of

571: transmission at 6.53 GHz should be larger than that at 8.64 GHz.

572: But in Fig. 4, the variance of transmission at 6.54 GHz is smaller

573: than that at 8.64 GHz. Why?

574: 

575: (iv) The authors should specify how many random configurations

576: they calculated to extract the average and variance of the

577: transmission.

578: 

579: \subsection{Report of Referee B -- LX8100/Gupta}

580: 

581: The authors of this paper make two very strong claims: 1. that

582: they have proven the absence of wave localization in 2d, and 2.

583: that identifying localization with the exponential decay of the

584: transmission through a disordered sample is wrong. If these claims

585: were true thay would constitute a significant change in our

586: understanding of the localization phenomenon. However, I did not

587: find author's arguments convincing enough to substantiate such

588: revolutionary claims. The main shortcoming of the author's

589: approach, in my opinion, is the lack of analysis of the effects

590: due to the finite size of the system in the direction

591: perpendicular to the propagation direction (in the outside

592: scenario, using the author's terminology). The localization length

593: for the frequency inside the propagation band is about 6, i.e. is

594: the same as the transvers size of the system. This is not a 2d

595: situation, and the slope of the dependence of the average

596: logarithm of transmission versus L found by the authors in this

597: case has nothing to do with the real localization length.

598: Therefore, the comparison of this length with the localization

599: length in the inside scenario does not make too much sense. The

600: absence of the localization, which the authors found for the

601: inside scenario (which looks more like a real 2d situation) most

602: likely means that the real localization length is larger than the

603: system size. Thus, I conclude that the authors's claims are not

604: substantiated, and the paper, therefore, cannot be published in

605: its current form.

606: 

607: There is also a minor remark, which I would like to add. The

608: authors refer to observations of the 2-d metal-insulator

609: transition to corraborate their claim of the absence of

610: localization in 2d. The existence of such a transition, however,

611: is still under intensive debate, and the authors should've at

612: least mentioned both sides of this discussion. Moreover, it

613: appears that a consensus is being formed in the community that so

614: called 2d MIT is not a real localization-delocalization

615: transition, and can be explained within a more traditional

616: framework. I think, therefore, that the authors should exercise

617: caution using this argument.

618: 

619: \subsection{Replies}

620: 

621: \subsubsection{List of changes}

622: 

623: \begin{enumerate}

624: 

625: \item We have changed Fig.~4 to include the effects of the width

626: in the `Outside' scenario.

627: 

628: \item Figs.~3 and 5 have been re-plotted to look better.

629: 

630: \item We have modified the paper on the discussion about the

631: absence of wave localization. And we have more specified the point

632: of the paper, that is, to show that there is a fundamental and

633: qualitative difference between the two scenarios of probing

634: localization.

635: 

636: \item We have modified the reference to the MIT in 2D.

637: 

638: \item We have deleted the word generality.

639: 

640: \item We have added information about the configuration, and fluctuation.

641: 

642: \item Some other mistakes are corrected.

643: 

644: \item We have removed a few unnecessary references.

645: 

646: \end{enumerate}

647: 

648: \subsubsection{General reply}

649: 

650: We thank both referees for the careful comments. As said by the

651: referees, the main points suggested in the first version were (1)

652: there could be extended waves in 2D random media; (2) the

653: traditional way of probing localization may not be quite proper.

654: 

655: As for point (1), we agree with both referees that it may be hard

656: to make a strong claim. This is simply because of the inevitable

657: problem of finite system size. This problem not only concerns any

658: numerical simulation, but also experiments. If it is the faith

659: that waves are always localized in 2D, then any disagreement with

660: this faith could be easily attributed to the finite sample size

661: not only for simulations, but for experiments. We think that both

662: referees can agree with us on this point. Therefore, it may not be

663: so proper for us to state that there are extended waves in 2D

664: disordered media. But at least, we can say that we do not observe

665: localization which can be expected from the traditional way of

666: probing localization. We have modified the paper so not to

667: strongly state that there is absence of wave localization in 2D.

668: 

669: The above discussion leads us to point (2). What went wrong, if we

670: do not observe localization as expected? We may state the

671: following three possibilities:

672: 

673: \begin{enumerate}

674: 

675: \item ``Our numerical codes are wrong"

676: 

677: This is very unlikely. We have used our codes to reproduce many

678: experimental data, and some theoretical simulations. Some of these

679: results can be found in the literature (e.~g. Phys. Rev. E64,

680: 036626 (2001), Phys. Rev. Lett. 87, 184301 (2001), and a paper to

681: appear in PRE).

682: 

683: \item ``The previous concept of localization is wrong"

684: 

685: This is also unlikely. The concept of localization has been

686: summarized clearly by Lee and Ramakrishnan in their excellent

687: review (on Page 288 of Rev. Mod. Phys. 57, 287 (1985)). In fact,

688: we did show localization features in Fig.~5, as well as in other

689: papers (Phys. Rev. Lett. {\bf 83}, 2734 (1999), Appl. Phys. Lett.,

690: {\bf 77}, 4428 (2000)). These results comply with the editorial

691: description of localization on the covering page of Nature of the

692: December issue of 1997, and the relevant contents and a paper

693: therein (Wiersma, et al., Nature, {\bf 373}, 203 (1997)).

694: 

695: \item ``The traditional way of probing localization may not be quite

696: proper, thus causing improper investigations of localization in

697: the past"

698: 

699: This is what has been suggested in the present paper. And we

700: believe that this has been evidenced in the paper. We have

701: compared two scenarios in probing localization. One is the

702: `Inside' and the other is the `Outside' scenario. They are

703: illustrated by Fig.~\ref{fig2}. The `Outside' scenario, as stated

704: clearly in Chabanov et al. (Nature {\bf 404}, 850 (2000)), is the

705: traditional way of exploring the phenomenon of localization.

706: 

707: At this stage, we can state the following: So far, all our

708: simulations show the {\it qualitative} difference between the

709: results from the two scenarios.

710: 

711: The results are: (1) for all frequencies we have simulated so far,

712: there is an exponential decay for the transmission across the

713: random sample, i.~e. in the `Outside' scenario; (2) in the

714: `Inside' case, the exponential decay appears for some frequencies,

715: but not for others at which there is the exponential decay in the

716: `Outside' scenario; (3) When the exponential decay appears in both

717: cases, the decaying slop is nearly the same. These features are

718: exemplified by Fig.~\ref{fig4} for two frequencies 6.54 GHz and

719: 8.64 GHz. We are still working on for more frequencies.

720: 

721: However, both referees challenge our results by raising the

722: concern about the finite width of a slab of a sample, which is

723: unavoidable in both numerical and experimental studies. This is a

724: very good point, and has indeed been included in our consideration

725: and simulation. Regretfully, we did not include the result in our

726: first version because we were worried about the length limitation

727: of PRL. In fact, we have done the following.

728: 

729: With a fixed sample size (i.~e. the length L), we plot the

730: transmission versus width. We find that when the width is large

731: enough, the transmission will saturate to a certain value - we

732: have fitted the data by two different ways, both yielding nearly

733: the same value. We have done for several lengths, and obtained the

734: corresponding saturated value for each length. Then we plot these

735: values versus sample length. The exponential decay observed for

736: the `Outside' case remains, indicating that the exponential decay

737: is not caused by the finite sample width. However, the finite

738: width does affect the value of the decaying slop. In our revision,

739: we have added new figures and some discussion on this point. Here

740: we emphasize that we are more concerned with the {\bf qualitative}

741: difference between the two ways of probing localization. This, we

742: believe, is the most important part in the fundamental research.

743: We think that the referees would also agree with us on this point.

744: 

745: There are other results about the transmission versus the band

746: structures. Due to the space limitation, we will not show here.

747: These results to be reported elsewhere do not change the main

748: conclusions of the paper.

749: 

750: \end{enumerate}

751: 

752: \subsubsection{Reply to referee A}

753: 

754: \begin{enumerate}

755: 

756: \item Reply to point (i)

757: 

758: \begin{enumerate}

759: 

760: \item We agree that the previous assertion is for an infinite

761: system. Unfortunately, an infinite system is not possible for

762: either experimental and numerical workers. As we discussed in the

763: general reply, the finite size can always be a concern. However,

764: as long as the sample size is equal or larger than the

765: localization length, the exponential decay expected for

766: localization should appear, no matter how small the system is;

767: this point can also be inferred from the localization theory for

768: 2D (e.~g. Ref~[1], i.~e. the diffusion vanishes when the size is

769: equal to the localization length). Here we have shown the

770: transmission for two frequencies for two scenarios discussed in

771: the general reply.

772: 

773: If the traditional way of probing localization is proper, the

774: localization length can be estimated from the observed exponential

775: decay.  Then as long as the sample size is larger than this

776: localization length, the exponential decay should appear. But

777: clearly, this is not supported by our results. Explicitly, at 8.64

778: GHz, there is a residual exponential decay for the extrapolated

779: infinite width of sample (shown nicely by the new Fig.~4~(f)).

780: From the slop, we estimate the decay length is about 16. If this

781: length were the localization length, as expected from the previous

782: perception, then we would also expect to see the exponential decay

783: in the `Inside' scenario as long as the sample size is larger than

784: this length. Unfortunately, we do not see, as evidenced by the

785: result shown by the flat curve in Fig.~4~(c). In any event, the

786: fact that the exponential decay only occurs in one scenario but

787: not in the other for the same sample size ($L$) is itself

788: intriguing and an important finding. It is subtle, but that is

789: what we get honestly (our codes are open for any one who wants to

790: check). These discussions and those in the general reply should

791: ease the concern of finite size effects.

792: 

793: \item The referee further wrote ``I estimated the localization

794: length l from the slop in Fig. 4. At 8.64 GHz (outside the gap),

795: the slope of 0.1648 gives l $\sim$ 1/0.16 $\sim$ 6. In the Outside

796: case, the system length L = 6 or 10. So the system length is

797: comparable to the localization length. That is why the wave is not

798: localized: the system is too small! If the authors increase the

799: system s length L and width W to over 100, they should observe

800: exponential decay in the transmission."

801: 

802: We feel that there could be some confusion here. The exponential

803: decay at 8.64 GHz is only observed for the outside case. The decay

804: slope depends on the width. As said above, when extrapolating to

805: an infinite width, the slop becomes three fold smaller. The point

806: in our paper is:  If the exponential decay in the outside case is

807: caused by localization, then we should also observe the

808: exponential decay for the {\it same} sample size ($L$) in the

809: `Inside' scenario. The result in the center panel of

810: Fig.~\ref{fig4} clearly does not support this point of view.

811: 

812: It is also worth noting that the exponential decay length in the

813: outside case at 8.64 is estimated as about 16. Fig.~4(a) shows

814: that the exponential decay behavior already appears when the

815: sample length ($L$) is smaller than this length, indicating that

816: the sample needs not to be too much larger than the decay length

817: to observe the exponential decay or localization. The same

818: conclusion can be drawn from the case with 6.54 GHz.

819: 

820: \end{enumerate}

821: 

822: \item Reply to point (ii)

823: 

824: This point raised by the referee again concerns the finite width

825: effect. As we have discussed above, although the quantitative

826: value of the decaying slop in the `Outside' scenario is indeed

827: affected by the width, the qualitative difference between the

828: `Inside' and `Outside' cases is not caused by the finite width. We

829: have added new discussion and new figures to support our

830: observation.

831: 

832: \item Reply to minor points of referee A

833: 

834: \begin{enumerate} \item About the brevity yet without losing

835: generality

836: 

837: We agree that it hard to localize light, because of the two

838: polarizations. What we said about the generality is as follows. We

839: have done both E and H polarizations and found the same

840: qualitative conclusions. To avoid possible ambiguities, we have

841: modified the wording.

842: 

843: \item About the label in Fig.~4

844: 

845: There was a mistake in the caption of Fig.~4, and another mistake

846: in citing the frequency from Fig.~4. We have corrected these.

847: 

848: \item About the fluctuation

849: 

850: We think that normally the stronger is the localization, the

851: weaker the fluctuation. At 8.64 GHz, we think that the wave is not

852: yet localized. Therefore the wave can travel around the sample and

853: become sensitive to the configuration. Therefore it has a stronger

854: fluctuation feature. On the other hand, the sample size for the

855: 6.54 GHz is much larger than the localization length, the

856: possibility for either total intensity or the diffusive intensity

857: to escape is small, so we observe smaller fluctuation at this

858: frequency. However, we did not show the ratio between the

859: fluctuation and averaged transmission. In fact, it is true that

860: the ratio at 6.54 GHz can be larger than that at 8.64 GHz. Perhaps

861: this is what the referee have meant. We have modified the paper

862: accordingly.

863: 

864: \item About the configuration.

865: 

866: We have done simulations from 50 to 500 configurations. We found

867: that for most cases, 50 would be enough. But the results shown in

868: Fig.~4 are taken from 500 configuration.

869: 

870: \end{enumerate}

871: 

872: \end{enumerate}

873: 

874: \subsubsection{Reply to referee B}

875: 

876: \begin{enumerate}

877: 

878: \item Referee B wrote ``The authors of this paper make two very strong

879: claims: 1. that they have proven the absence of wave localization

880: in 2d, and 2. that identifying localization with the exponential

881: decay of the transmission through a disordered sample is wrong. If

882: these claims were true they would constitute a significant change

883: in our understanding of the localization phenomenon."

884: 

885: We agree that it may be too strong to say that we have proven

886: ``the absence of wave localization in 2d". But our results

887: evidently supports the second point by the referee, that is,

888: identifying localization with the exponential decay of the

889: transmission through a disordered sample in the traditional way,

890: is not quite proper, if not wrong.

891: 

892: We appreciate the referee's saying ``If these claims were true

893: they would constitute a significant change in our understanding of

894: the localization phenomenon." We would like to say that while the

895: first claim is hard to prove (prove might not be a good word), but

896: at least the second claim is true. Meanwhile, we do not think our

897: results will change the understanding of the concept of

898: localization, but we hope that our results can stimulate some new

899: proper or unambiguous ways in probing localization. This is very

900: important. We have revised the paper accordingly.

901: 

902: \item The referee wrote ``The localization length

903: for the frequency inside the propagation band is about 6, i.e. is

904: the same as the transvers size of the system. This is not a 2d

905: situation, and the slope of the dependence of the average

906: logarithm of transmission versus L found by the authors in this

907: case has nothing to do with the real localization length.

908: Therefore, the comparison of this length with the localization

909: length in the inside scenario does not make too much sense. The

910: absence of the localization, which the authors found for the

911: inside scenario (which looks more like a real 2d situation) most

912: likely means that the real localization length is larger than the

913: system size. Thus, I conclude that the authors's claims are not

914: substantiated, and the paper, therefore, cannot be published in

915: its current form."

916: 

917: What the referee raised here concerns the finite width. As we

918: discussed in the general reply, possible finite size effect has

919: always topped the priority of our consideration. Referring to the

920: discussion in the general reply section, here we just stress that

921: we have indeed included the finite width in our consideration and

922: simulation. We have re-plotted Fig.~4, which explicitly shows how

923: we exclude the possible effects of finite width on reaching our

924: conclusion that there is a fundamental and qualitative difference

925: between the two scenarios of probing localization, discussed

926: earlier. The point is:  If the exponential decay in the outside

927: case is caused by localization, then we should also observe the

928: exponential decay for the {\it same} sample size ($L$) in the

929: `Inside' scenario. The result in the center panel of

930: Fig.~\ref{fig4} clearly does not support this point of view. In

931: the revision, we have specified this more clearly.

932: 

933: We appreciate that the referee agrees that the `Inside' case is

934: more like a real 2D situation. It is this scenario that should be

935: considered for measuring localization effects.

936: 

937: \item About the minor point

938: 

939: We are aware that there is an intensive debate on the MIT in 2D.

940: We have modified our reference to this.

941: 

942: \end{enumerate}

943: 

944: \end{document}

945: