1: \documentclass[runningheads,fleqn]{svmult}
2:
3: \usepackage{amssymb} % to have \lesssim symbol
4:
5: \usepackage{makeidx} % allows index generation
6: \usepackage{graphicx} % standard LaTeX graphics tool
7: % for including eps-figure files
8: %\usepackage{subeqnar} % subnumbers individual equations
9: % within an array
10: \usepackage{multicol} % used for the two-column index
11: \usepackage{physmult} % flushleft layout of captions etc.
12: \makeindex % used for the subject index
13: % please use the style sprmidx.sty with
14: % your makeindex program
15:
16: %%upright Greek letters (example below: upright "mu")
17: \newcommand{\greeksym}[1]{{\usefont{U}{psy}{m}{n}#1}}
18: \newcommand{\umu}{\mbox{\greeksym{m}}}
19: \newcommand{\udelta}{\mbox{\greeksym{d}}}
20: \newcommand{\uDelta}{\mbox{\greeksym{D}}}
21: \newcommand{\uPi}{\mbox{\greeksym{P}}}
22: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
23: \begin{document}
24:
25: \title*{Self-Trapping of Light \\
26: and Nonlinear Localized Modes \\
27: in 2D Photonic Crystals and Waveguides}
28:
29: \toctitle{Self-Trapping of Light and Nonlinear Localized Modes
30: \protect\newline in 2D Photonic Crystals and Waveguides}
31:
32: \titlerunning{Nonlinear Localized Modes
33: in Photonic Crystals}
34:
35: \author{Serge F. Mingaleev $^{1,2}$
36: \and Yuri S. Kivshar $^{1}$}
37:
38: \institute{$^1$ Nonlinear Physics Group, Research School of
39: Physical Sciences and Engineering, \\
40: The Australian National University, Canberra ACT 0200,
41: Australia \\
42: $^2$ Bogolyubov Institute for Theoretical Physics,
43: Kiev 03143, Ukraine}
44:
45: \authorrunning{Serge F. Mingaleev and Yuri S. Kivshar}
46: \maketitle
47:
48: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
49: \section{Introduction}
50: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
51:
52: Photonic crystals are usually viewed as an optical analog of
53: semiconductors that modify the properties of light similar
54: to a microscopic atomic lattice that creates a semiconductor
55: band-gap for electrons \cite{book}. It is therefore believed
56: that by replacing relatively slow electrons with photons as
57: the carriers of information, the speed and band-width of
58: advanced communication systems will be dramatically
59: increased, thus revolutionizing the telecommunication
60: industry. Recent fabrication of photonic crystals with a
61: band gap at optical wavelengths from 1.35 $\umu$m to
62: 1.95 $\umu$m makes this promise very realistic
63: \cite{opt}.
64:
65: To employ the high-technology potential of photonic crystals, it is
66: crucially important to achieve a dynamical tunability of their
67: band gap \cite{john2}. This idea can be realized by changing
68: the light intensity in the so-called {\em nonlinear photonic
69: crystals}, having a periodic modulation of the nonlinear
70: refractive index \cite{berger}. Exploration of {\em nonlinear
71: properties} of photonic band-gap (PBG) materials is an important
72: direction of research that opens new applications of photonic crystals
73: for all-optical signal processing and switching, allowing an
74: effective way to create tunable band-gap structures operating
75: entirely with light.
76:
77: One of the important physical concepts associated with
78: nonlinearity is {\em the energy self-trapping and
79: localization}. In the linear physics, the idea of
80: localization is always associated with disorder that breaks
81: translational invariance. However, during the recent years
82: it was demonstrated that localization can occur in the
83: absence of any disorder and solely due to nonlinearity in
84: the form of {\em intrinsic localized modes} \cite{review}.
85: A rigorous proof of the existence of time-periodic,
86: spatially localized solutions describing such nonlinear
87: modes has been presented for a broad class of Hamiltonian
88: coupled-oscillator nonlinear lattices \cite{mak}, but
89: approximate analytical solutions can also be found in many
90: other cases, demonstrating a generality of the concept of
91: {\em nonlinear localized modes}.
92:
93: Nonlinear localized modes can be easily identified in
94: numerical molecular-dynamics simulations in many different
95: physical models (see, e.g., Ref. \cite{review} for a
96: review), but only very recently the {\em first experimental
97: observations} of spatially localized nonlinear modes have
98: been reported in mixed-valence transition metal
99: complexes \cite{bishop}, quasi-one-dimensional
100: antiferromagnetic chains \cite{sievers}, and arrays of
101: Josephson junctions \cite{JJ}.
102: Importantly, very similar types of spatially localized
103: nonlinear modes have been experimentally observed in
104: {\em macroscopic} mechanical \cite{zolo} and guided-wave
105: optical \cite{silb} systems.
106:
107: Recent experimental observations of nonlinear localized modes, as
108: well as numerous theoretical results, indicate that
109: nonlinearity-induced localization and spatially localized modes can be
110: expected in physical systems of very different nature. From the
111: viewpoint of possible practical applications, self-localized states in
112: optics seem to be the most promising ones; they can lead to different
113: types of nonlinear all-optical switching devices where light
114: manipulates and controls light itself by varying the input
115: intensity. As a result, the study of {\em nonlinear
116: localized modes in photonic structures} is expected to bring
117: a variety of realistic applications of intrinsic localized
118: modes.
119:
120: One of the promising fields where the concept of nonlinear
121: localized modes may find practical applications is the
122: physics of {\em photonic
123: crystals} [or photonic band gap (PBG) materials] --
124: periodic dielectric structures that produce
125: many of the same phenomena for photons as the crystalline
126: atomic potential does for electrons \cite{book}.
127: Three-dimensional (3D) photonic crystals for
128: visible light have been successfully fabricated only within
129: the past year or two, and presently many research groups
130: are working on creating tunable band-gap switches and
131: transistors operating entirely with light. The most
132: recent idea is to employ nonlinear properties of band-gap
133: materials, thus creating {\em nonlinear photonic crystals}
134: including those where nonlinear susceptibility is periodic
135: as well \cite{berger,sukh}.
136:
137: Nonlinear photonic crystals (or photonic crystals with
138: embedded nonlinear impurities) create an ideal environment
139: for the generation and observation of nonlinear localized
140: photonic modes. In particular, the existence of
141: such modes for the frequencies in the photonic band gaps has
142: been predicted \cite{john} for 2D and 3D photonic crystals
143: with Kerr nonlinearity. Nonlinear localized modes can be also
144: excited at nonlinear interfaces with quadratic nonlinearity
145: \cite{sukh2}, or along dielectric waveguide structures
146: possessing a nonlinear Kerr-type response \cite{mcgurn}.
147:
148: In this Chapter, we study self-trapping of light and
149: nonlinear localized modes in nonlinear photonic crystals
150: and photonic crystal waveguides. For simplicity, we
151: consider the case of a 2D photonic crystal with embedded
152: nonlinear rods (impurities) and demonstrate that the
153: effective interaction in such a
154: structure is nonlocal, so
155: that the nonlinear effects can be described by a nontrivial
156: generalization of the nonlinear lattice models that include
157: the long-range coupling and nonlocal nonlinearity. We
158: describe several different types of nonlinear guided-wave
159: states in photonic crystal waveguides and analyse their
160: properties \cite{mingaleev}.
161: Also, we predict the existence of {\em stable} nonlinear
162: localized modes (highly localized modes analogous to gap
163: solitons in the continuum limit) in the reduced-symmetry
164: nonlinear photonic crystals \cite{mingaleev2}.
165:
166:
167: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
168: \section{Basic Equations}
169: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
170:
171: Let us consider a 2D photonic crystal created by a periodic
172: lattice of parallel, infinitely long dielectric rods in
173: air (see Fig. \ref{fig:band-r0.18}).
174: We assume that the rods are parallel to the
175: $x_3$ axis, so that the system is characterized by the
176: dielectric constant $\varepsilon(\vec{x})=\varepsilon(x_1,
177: x_2)$. As is well known \cite{book}, the photonic
178: crystals of this
179: type can possess a complete band gap for the $E$-polarized
180: (with the electric field $\vec{E} \, || \, \vec{x}_3$)
181: light propagating in the $(x_1, x_2)$-plane.
182: The evolution of such a light is governed by the scalar
183: wave equation
184: \begin{equation}
185: \nabla^2 E(\vec{x}, t) - \frac{1}{c^2} \, \partial_t^2
186: \left[ \varepsilon(\vec{x}) E \right] = 0 \; ,
187: \label{sys:eq-E-t}
188: \end{equation}
189: where
190: $\nabla^2 \equiv \partial_{x_1}^2 + \partial_{x_2}^2$
191: and $E$ is the $x_3$ component of $\vec{E}$.
192: Taking the electric field in the form
193: %%\begin{displaymath}
194: $E(\vec{x}, t) = \E^{- \I \omega t} \, E(\vec{x}, t \,|\,
195: \omega) \; ,$
196: %%\end{displaymath}
197: where $E(\vec{x}, t \,|\, \omega)$ is a slowly varying
198: envelope, i.e.
199: $\partial^2_t E(\vec{x}, t \,|\, \omega) \ll \omega
200: \partial_t E(\vec{x}, t \,|\, \omega)$,
201: Eq. (\ref{sys:eq-E-t}) reduces to
202: \begin{equation}
203: \left[ \nabla^2 + \varepsilon(\vec{x})
204: \left( \frac{\omega}{c} \right)^2 \right]
205: E(\vec{x}, t \,|\, \omega) \simeq -2 \, \I \,
206: \varepsilon(\vec{x}) \frac{\omega}{c^2} \,
207: \frac{\partial E}{\partial t} \; .
208: \label{sys:eq-E-omega-t}
209: \end{equation}
210: In the stationary case, i.e. when the r.h.s. of Eq.
211: (\ref{sys:eq-E-omega-t}) vanishes,
212: this equation describes an eigenvalue problem which can be
213: solved, e.g. by the plane waves method
214: \cite{Maradudin:1993:PBGL}, in the case of a perfect
215: photonic crystal, for which the dielectric constant
216: $\varepsilon(\vec{x}) \equiv \varepsilon_{p}(\vec{x})$
217: is a periodic function defined as
218: \begin{equation}
219: \varepsilon_{p}(\vec{x}+\vec{s}_{ij}) =
220: \varepsilon_{p}(\vec{x}) \; ,
221: \label{sys:eps-pc}
222: \end{equation}
223: where $i$ and $j$ are arbitrary integers, and
224: \begin{equation}
225: \vec{s}_{ij} = i \, \vec{a}_1 + j \,
226: \vec{a}_2
227: \label{sys:s-ij}
228: \end{equation}
229: is a linear combination of the lattice vectors
230: $\vec{a}_1$ and $\vec{a}_2$.
231:
232: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
233: %\vspace{10mm}
234: \begin{figure}[t]
235: \begin{minipage}{70mm}
236: \includegraphics[width=70mm,angle=0,clip]{mk-fig1a.eps}
237: \end{minipage}
238: %%\hspace*{0mm}
239: \begin{minipage}{35mm}
240: \includegraphics[width=35mm,angle=270,clip]{mk-fig1b.eps}
241: \end{minipage}
242: \vspace{3mm}
243: \caption{The band-gap structure of the photonic crystal
244: consisting
245: of a square lattice of dielectric rods with $r_0=0.18a$ and
246: $\varepsilon_0=11.56$ (the band gaps are shaded).
247: The top center inset shows a cross-sectional
248: view of the 2D photonic crystal depicted in the right inset.
249: The bottom center inset shows the corresponding Brillouin
250: zone, with the irreducible zone shaded.}
251: \label{fig:band-r0.18}
252: \end{figure}
253: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
254:
255: For definiteness, we consider the 2D photonic crystal
256: earlier analyses (in the linear limit)
257: in Refs. \cite{Mekis:1996:PRL,Mekis:1998:PRB}. That is,
258: we assume that cylindrical rods with radius
259: $r_0=0.18a$ and dielectric constant $\varepsilon_0=11.56$
260: form a square lattice with the distance $a$ between two
261: neighboring rods, so that
262: $\vec{a}_1=a \vec{x}_1$ and $\vec{a}_2=a \vec{x}_2$.
263: The frequency band structure for this type of 2D photonic
264: crystal is shown in Fig. \ref{fig:band-r0.18} where, using
265: the notations of the solid-state physics, the wave dispersion
266: is mapped onto the Brillouin zone of the so-called
267: {\em reciprocal lattice} that faces are known as $\Gamma$,
268: $M$, and $X$. As follows from Fig. \ref{fig:band-r0.18},
269: there exists a large (38\%) band gap
270: that extends from the lower cut-off frequency,
271: $\omega=0.302 \times 2\pi c/a$, to the upper band-gap
272: frequency, $\omega=0.443 \times 2\pi c/a$.
273: If the frequency of a low-intensity light falls into the
274: band gap, the light cannot propagate through the photonic
275: crystal and is reflected.
276:
277: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
278: \section{Defect Modes: The Green Function
279: Approach}
280: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
281:
282: One of the most intriguing properties of photonic band
283: gap crystals is the emergence of exponentially localized modes
284: that may appear within the photonic band gaps when a defect
285: is embedded into an otherwise perfect photonic crystal.
286: The simplest way to create a defect in a 2D photonic
287: crystal is to introduce an additional defect rod with
288: the radius $r_d$ and the dielectric constant
289: $\varepsilon_{d}(\vec{x})$.
290: In this case, the dielectric constant
291: $\varepsilon(\vec{x})$ can be presented as a sum of periodic
292: and defect-induced terms, i.e.
293: \begin{displaymath}
294: \varepsilon(\vec{x})=\varepsilon_{p}(\vec{x})+
295: \varepsilon_{d}(\vec{x}) \; ,
296: %\label{sys:eps}
297: \end{displaymath}
298: and, therefore, Eq. (\ref{sys:eq-E-omega-t}) takes the form
299: \begin{eqnarray}
300: \left[ \nabla^2 + \left( \frac{\omega}{c} \right)^2
301: \varepsilon_{p}(\vec{x}) \right] E(\vec{x},
302: t \,|\, \omega) = - \hat{\cal L} E(\vec{x}, t
303: \,|\, \omega) \; ,
304: \label{sys:eq-E-omega-t2}
305: \end{eqnarray}
306: where the operator
307: \begin{eqnarray}
308: \hat{\cal L} = \left( \frac{\omega}{c} \right)^2
309: \varepsilon_{d}(\vec{x})
310: + 2 \, \I \, \varepsilon(\vec{x})
311: \frac{\omega}{c^2} \frac{\partial}{\partial t}
312: \end{eqnarray}
313: is introduced for convenience.
314: Equation (\ref{sys:eq-E-omega-t2}) can also be written in
315: the equivalent integral form
316: \begin{equation}
317: E(\vec{x}, t \,|\, \omega) =
318: \int \D^2\vec{y} \,\,\, G(\vec{x},
319: \vec{y} \,|\, \omega) \, \hat{\cal L} \,
320: E(\vec{y}, t \,|\, \omega) \; ,
321: \label{sys:eq-green-int}
322: \end{equation}
323: where $G(\vec{x}, \vec{y} \,|\, \omega)$ is the Green
324: function defined, in a standard way, as a
325: solution of the equation
326: \begin{equation}
327: \left[ \nabla^2 + \left( \frac{\omega}{c} \right)^2
328: \varepsilon_{p}(\vec{x}) \right]
329: G(\vec{x}, \vec{y} \,|\, \omega) = -
330: \delta(\vec{x}-\vec{y}) \; .
331: \label{sys:eq-green-omega}
332: \end{equation}
333: General properties of the Green function of a perfect 2D
334: photonic crystal are described in more details in Ref.
335: \cite{Maradudin:1993:PBGL}. Here, we notice
336: that the Green function is {\em symmetric}, i.e.
337: \begin{displaymath}
338: G(\vec{x}, \vec{y} \,|\, \omega) =
339: G(\vec{y}, \vec{x} \,|\, \omega)
340: %\label{sys:green-symm}
341: \end{displaymath}
342: and {\em periodic}, i.e.
343: \begin{displaymath}
344: G(\vec{x} + \vec{s}_{ij}, \vec{y} + \vec{s}_{ij} \,|\,
345: \omega) = G(\vec{x}, \vec{y} \,|\, \omega) \; ,
346: %\label{sys:green-period}
347: \end{displaymath}
348: where $\vec{s}_{ij}$ is defined by Eq. (\ref{sys:s-ij}).
349:
350: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
351: %\vspace{10mm}
352: \begin{figure}[t]
353: \begin{minipage}{40mm}
354: \centerline{{\large\bf (a)}}
355: \includegraphics[width=40mm,angle=0,clip]{mk-fig2a.eps}
356: \end{minipage} \hspace*{10mm}
357: \begin{minipage}{70mm}
358: \centerline{{\large\bf (b)}}
359: \includegraphics[width=70mm,angle=0,clip]{mk-fig2b.eps}
360: \end{minipage}
361: \vspace{3mm}
362: \caption{{\bf (a)} Electric
363: field structure of a linear localized mode
364: supported by a single defect rod with radius
365: $r_d=0.1a$ and $\varepsilon_d=11.56$ in a square-lattice
366: photonic crystal with $r_0=0.18a$ and
367: $\varepsilon_0=11.56$.
368: The rod positions are indicated by circles and
369: the amplitude of the electric field is indicated by color.
370: {\bf (b)}
371: Frequency of the defect mode as a function of the
372: radius $r_d$: calculated precisely from
373: Eq. (\ref{sys:eq-green-int}) ({\em full line with circles})
374: and approximately from Eq. (\ref{sys:eq-defect})
375: ({\em dashed line with triangles}).}
376: \label{fig:defect}
377: \end{figure}
378: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
379:
380: The Green function can be calculated by means of the
381: Fourier transform
382: \begin{equation}
383: G(\vec{x}, \vec{y} \,|\, \omega) =
384: \int_{-\infty}^{\infty} \D t \,\,\,
385: e^{\I \omega t} G(\vec{x}, \vec{y}, t)
386: \end{equation}
387: applied to the time-dependent Green function governed by
388: the equation
389: \begin{equation}
390: \left[ \nabla^2 - \varepsilon_{p}(\vec{x}) \, \partial_t^2
391: \right] G(\vec{x},\vec{y}, t) = - \delta(t) \,
392: \delta(\vec{x}-\vec{y}) \; ,
393: \end{equation}
394: which has been solved by the finite-difference time-domain
395: method \cite{Ward:1998:PRB}.
396:
397: Now that we have calculated the Green function, we can
398: figure out the defect states solving Eq. (\ref{sys:eq-green-int})
399: directly. For example, Fig.~\ref{fig:defect}(a)
400: shows a defect mode created by introducing a single
401: defect rod with the radius $r_d=0.1a$ and dielectric constant
402: $\varepsilon_d=11.56$ into the 2D photonic crystal shown
403: in Fig.~\ref{fig:band-r0.18}.
404: Although direct numerical solution of the integral
405: equation (\ref{sys:eq-green-int}) remains possible even
406: in the case of a few defect rods, it becomes severely limited
407: by the current computer facilities as soon as we increase
408: the number of the defect rods and start investigation of
409: the line defects (waveguides) and their branches.
410: Thus, looking for new approximate numerical techniques which
411: could combine reasonable accuracy, flexibility, and power
412: to forecast new effects is an issue of the key importance.
413:
414: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
415: \section{Effective discrete equations}
416: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
417:
418:
419: Studying the electric field distribution of the defect
420: mode in Fig. \ref{fig:defect}(a), one can suggest that
421: a reasonably accurate approximation should be provided by
422: the assumption that the electric field inside the defect rod
423: remains constant. Indeed, let us assume that nonlinear defect rods
424: embedded into a photonic crystal are located at the points
425: $\vec{x}_m$, where $m$ is the index (or a combination of
426: two indices in the case of a two-dimensional array of
427: defect rods) introduced for explicit numbering of the
428: defect rods.
429: In this case, the correction to the dielectric constant is
430: \begin{eqnarray}
431: \varepsilon_d(\vec{x}) = \left\{\varepsilon_{d}^{(0)} +
432: |E(\vec{x}, t \,|\, \omega)|^2\right\}
433: \sum_m \theta (\vec{x}-\vec{x}_m) \; ,
434: \label{sys:delta-eps}
435: \end{eqnarray}
436: where
437: \begin{equation}
438: \theta (\vec{x}) = \left\{
439: \begin{array}{c}
440: 1 \; , \quad \mbox{for} \quad |\vec{x}| \leq r_d \; , \\
441: 0 \; , \quad \mbox{for} \quad |\vec{x}| > r_d \; .
442: \end{array}
443: \right.
444: \label{sys:theta}
445: \end{equation}
446: The second term in Eq. (\ref{sys:delta-eps}) takes into
447: account a contribution due to the Kerr nonlinearity
448: (we assume that the electric field is scaled with the
449: nonlinear susceptibility, $\chi^{(3)}$).
450: Assuming, as we discussed above, that the electric field
451: $E(\vec{x}, t \,|\, \omega)$ inside the defect rods is
452: almost constant, one can derive, by substituting
453: Eq. (\ref{sys:delta-eps}) into Eq. (\ref{sys:eq-green-int})
454: and averaging over the cross-section of the rods
455: \cite{mcgurn},
456: an approximate {\em discrete nonlinear equation}
457: \begin{eqnarray}
458: i \sigma \frac{\partial}{\partial t} E_{n} - E_n
459: + \sum_m J_{n-m}(\omega) (\varepsilon_{d}^{(0)} +
460: |E_m|^2) E_m = 0 \; ,
461: \label{sys:eq-E-disc}
462: \end{eqnarray}
463: for the amplitudes of the electric field
464: $E_n(t \,|\, \omega) \equiv E(\vec{x}_n, t \,|\, \omega)$
465: inside the defect rods. The parameter $\sigma$ and the
466: coupling constants
467: \begin{equation}
468: J_{n}(\omega) = \left( \frac{\omega}{c} \right)^2
469: \int_{r_d} d^2 \vec{y} \,\,\,
470: G(\vec{x}_0, \vec{x}_n + \vec{y} \,|\, \omega )
471: \label{sys:Jn}
472: \end{equation}
473: are determined in this case by the Green function
474: $G(\vec{x}, \vec{y} \,|\, \omega)$
475: of the perfect photonic crystal.
476:
477: To check the accuracy of the approximation provided
478: by Eq. (\ref{sys:eq-E-disc}), we solved it in the
479: linear limit for the case of a single defect rod.
480: In this case Eq.~(\ref{sys:eq-E-disc})
481: is reduced to the equation
482: \begin{equation}
483: J_{0}(\omega_{d})= 1/\varepsilon_{d}^{(0)} \; ,
484: \label{sys:eq-defect}
485: \end{equation}
486: from which one can obtain an estimation for
487: the frequency $\omega_{d}$ of the localized defect mode.
488: As is seen from Fig. \ref{fig:defect}(b), the mode frequency
489: calculated in the framework of this approximation is in a good
490: agreement with that calculated directly from Eq.
491: (\ref{sys:eq-green-int}), provided the defect radius
492: $r_d$ is small enough. Even for $r_d=0.15a$ an error
493: introduced by the approximation does not exceed 5\%.
494: It lends a support to the
495: validity of Eq. (\ref{sys:eq-E-disc}) allowing
496: us to use it hereafter for studying nonlinear localized
497: modes.
498:
499: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
500: \section{Nonlinear Waveguides in 2D Photonic Crystals}
501: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
502:
503: One of the most promising applications of the PBG structures
504: is a possibility to create a novel type of optical waveguides.
505: In conventional waveguides such as
506: optical fibers, light is confined by {\em total internal
507: reflection} due a difference in the refractive indices of
508: the waveguide core and cladding.
509: One of the weaknesses of such waveguides is that
510: creating of bends is difficult. Unless the radius of
511: the bend is large compared to the wavelength, much of the
512: light will be lost. This is a serious obstacle for creating
513: ``integrated optical circuits'', since the space
514: required for large-radius bends is unavailable.
515:
516: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
517: %\vspace{10mm}
518: \begin{figure}[t]
519: \centerline{
520: \includegraphics[width=70mm,angle=0,clip]{mk-fig3.eps}}
521: \caption{Electric field of the linear guiding mode
522: in a waveguide created by an array of the defect
523: rods. The rod positions are indicated by circles and
524: the amplitude of the electric field is indicated by color.}
525: \label{fig:wg}
526: \end{figure}
527: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
528:
529: The waveguides based on
530: the PBG materials employ {\em a different physical
531: mechanism}: the light is
532: guided by a line of coupled defects which possess a
533: localized defect mode with frequency inside the band gap.
534: The simplest photonic-crystal waveguide can be created by
535: a straight line of defect rods, as shown in Fig.
536: \ref{fig:wg}. Instead of a single localized state of
537: an isolated defect, a waveguide supports propagating states
538: (guided modes) with the frequencies in a narrow band
539: located inside the band gap of a perfect crystal.
540: Such guided modes have a periodical profile along the
541: waveguide, and they decay exponentially in the transverse
542: direction, see Fig. \ref{fig:wg}.
543: That is, photonic crystal waveguides operate in a manner
544: similar to
545: resonant cavities, and the light with guiding frequencies
546: is forbidden from propagating in the bulk.
547: Because of this, when a bend is created in a
548: photonic crystal waveguide,
549: the light remains trapped and the only possible problem
550: is that of reflection. However, as was predicted numerically
551: \cite{Mekis:1996:PRL,Mekis:1998:PRB}
552: and then demonstrated in microwave \cite{Lin:1998:SCI}
553: and optical \cite{Tokushima:2000:APL} experiments,
554: it is still possible to get very high transmission
555: efficiency for nearly all frequencies inside the gap.
556:
557: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
558: %%\vspace{10mm}
559: \begin{figure}[t]
560: \centerline{
561: \includegraphics[width=80mm,angle=0,clip]{mk-fig4.eps}}
562: %\vspace{0mm}
563: \caption{The Green function $G(\vec{x}_0, \vec{x}_0 +
564: \vec{y} \,|\, \omega)$ for the photonic crystal shown in
565: Fig.~\protect\ref{fig:band-r0.18}
566: ($\vec{x}_0=\vec{a}_1/2$ and
567: $\omega=0.33 \times 2\pi c/a$).}
568: \label{fig:green-r0.18}
569: \end{figure}
570: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
571:
572: To employ the high-technology potential of photonic crystal
573: waveguides, it
574: is crucially important to achieve a tunability of their
575: transmission properties. Nowadays, several
576: approaches have been suggested for this purpose.
577: For instance, it has been recently demonstrated both
578: numerically \cite{Cheng:1999:PRB}
579: and in microwave experiments \cite{Jin:1999-sep:APL},
580: that transmission spectrum of straight and sharply bent
581: waveguides
582: in {\em quasiperiodic photonic crystals} is rather rich in
583: structure and only some frequencies get near perfect
584: transmission. Another possibility is creation of the
585: {\em channel drop system} on the bases of two parallel
586: waveguides coupled by the point defects between them.
587: It has been shown \cite{Fan:1998-feb:PRL}
588: that high-Q frequency selective complete transfer can occur
589: between such waveguides by creating resonant defect states
590: of different symmetry and by forcing an accidental
591: degeneracy between them.
592:
593: However, being frequency selective, the above mentioned
594: approaches do not possess {\em dynamical tunability} of
595: the transmission properties. The latter idea can be realized
596: by changing the light intensity in the so-called
597: {\em nonlinear photonic crystal waveguides} \cite{mingaleev},
598: created by inserting an additional row of rods made
599: from a Kerr-type nonlinear material characterized by the
600: third-order nonlinear susceptibility $\chi^{(3)}$
601: and the linear dielectric constant $\varepsilon_d^{(0)}$.
602: For definiteness, we assume that
603: $\varepsilon_d^{(0)} = \varepsilon_0 = 11.56$ and
604: that the nonlinear defect rods are
605: embedded into the photonic crystal along a selected
606: direction $\vec{s}_{ij}$, so that they are located at the points
607: $\vec{x}_m = \vec{x}_0 + m \, \vec{s}_{ij}$.
608: As we show below, changing the radius $r_d$ of these defect
609: rods and their location $\vec{x}_0$ in the crystal, one
610: can create nonlinear waveguides with quite different properties.
611:
612: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
613: %\vspace{15mm}
614: \begin{figure}[t]
615: \begin{minipage}{55mm}
616: \centerline{{\large\bf (a)}} \vspace{1mm}
617: \includegraphics[width=55mm,angle=0,clip]{mk-fig5a.eps}
618: \end{minipage} \hspace*{3mm}
619: \begin{minipage}{55mm}
620: \centerline{{\large\bf (b)}} \vspace{1mm}
621: \includegraphics[width=58mm,angle=0,clip]{mk-fig5b.eps}
622: \end{minipage}
623: \vspace{3mm}
624: \caption{{\bf (a)}
625: Dispersion relation for the photonic crystal waveguide
626: shown in the inset ($\varepsilon_0=\varepsilon_d=11.56$,
627: $r_0=0.18a$, $r_d=0.10a$). The grey areas are the projected
628: band structure of the perfect 2D photonic crystal.
629: The frequencies at the indicated points are:
630: $\omega_A=0.378 \times 2\pi c/a$ and
631: $\omega_B=0.412 \times 2\pi c/a$.
632: {\bf (b)} Mode power $Q(\omega)$ of the nonlinear mode
633: excited in the corresponding photonic crystal waveguide.
634: The right inset gives the dependence $J_n(\omega)$
635: calculated at $\omega=0.37 \times 2\pi c/a$.
636: The left inset presents the profile of the corresponding
637: nonlinear localized mode.}
638: \label{fig:x2-0.10}
639: \end{figure}
640: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
641:
642: As we have already discussed \cite{mingaleev,mingaleev2},
643: the Green function $G(\vec{x}, \vec{y} \,|\, \omega)$
644: and, consequently, the coupling coefficients
645: $J_{m}(\omega)$ are usually highly long-ranged functions.
646: This can be seen directly from Fig. \ref{fig:green-r0.18}
647: that shows a typical spatial profile of the Green function.
648: As a consequence, the coupling coefficients $J_n(\omega)$ calculated
649: from Eq. (\ref{sys:Jn}) decrease slowly with the site number
650: $n$. For $\vec{s}_{01}$ and $\vec{s}_{10}$ directions, the
651: coupling coefficients can be approximated by an exponential
652: function as follows
653: \begin{displaymath}
654: |J_n(\omega)| \approx \left\{
655: \begin{array}{lcc}
656: J_0(\omega) \; , & \mbox{for} & n=0 \; , \\
657: J_{*}(\omega) \, e^{-\alpha(\omega) |n|} \; ,
658: & \mbox{for} & |n| \geq 1 \; ,
659: \end{array}
660: \right.
661: %\label{sys:Jn-exp}
662: \end{displaymath}
663: where the characteristic decay rate $\alpha(\omega)$ can
664: be as small as $0.85$, depending on the values of
665: $\omega$, $\vec{x}_0$, and $r_d$, and
666: it can be even smaller for other types of photonic
667: crystals (for instance, for the photonic crystal used in
668: Fig. \ref{fig:norm} we find
669: $J_{m} \sim \, (-1)^m \exp(-0.66 m)$ for $m \geq 2$).
670: By this means,
671: Eq. (\ref{sys:eq-E-disc}) is a nontrivial
672: long-range generalization of a 2D discrete
673: nonlinear Schr\"odinger (NLS)
674: equation extensively studied during the last
675: decade for different applications \cite{DNLS}.
676: It allows us to draw an analogy between the problem
677: under consideration and a class of the NLS equations
678: that describe nonlinear excitations in quasi-one-dimensional molecular
679: chains with long-range (e.g. dipole-dipole) interaction
680: between the particles and local on-site nonlinearities
681: \cite{Gaididei:1997:PRE,Johansson:1998:PRE}.
682: For such systems, it was shown
683: that the effect of nonlocal interparticle interaction
684: brings some new features to the properties of nonlinear localized modes
685: (in particular, bistability in their spectrum).
686: Moreover,
687: for our model the coupling coefficients $J_n(\omega)$ can be
688: either unstaggered and monotonically decaying, i.e.
689: $J_n(\omega)=|J_n(\omega)|$, or
690: staggered and oscillating from site to site, i.e.
691: $J_n(\omega)=(-1)^{n}|J_n(\omega)|$.
692: We therefore
693: expect that effective nonlocality in both linear and
694: nonlinear terms of Eq. (\ref{sys:eq-E-disc}) may also
695: bring similar new features into the properties of nonlinear
696: localized modes excited in the photonic crystal
697: waveguides.
698:
699: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
700: \subsection{Staggered and unstaggered localized modes}
701: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
702:
703: As can be seen from the structure of the Green function
704: presented in Fig. \ref{fig:green-r0.18}, the case of
705: monotonically varying coefficients $J_n(\omega)$ can
706: occur for the waveguide oriented in the $\vec{s}_{01}$
707: direction with $\vec{x}_0=\vec{a}_1/2$. In this case, the
708: frequency of a linear guided mode, that can be excited in
709: such a waveguide, takes a minimum value at $k=0$ [see
710: Fig. \ref{fig:x2-0.10}(a)], and the corresponding
711: nonlinear mode is expected to be unstaggered.
712:
713: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
714: %\vspace{15mm}
715: \begin{figure}[t]
716: \begin{minipage}{55mm}
717: \centerline{{\large\bf (a)}} \vspace{1mm}
718: \includegraphics[width=55mm,angle=0,clip]{mk-fig6a.eps}
719: \end{minipage} \hspace*{3mm}
720: \begin{minipage}{55mm}
721: \centerline{{\large\bf (b)}} \vspace{1mm}
722: \includegraphics[width=58mm,angle=0,clip]{mk-fig6b.eps}
723: \end{minipage}
724: \vspace{3mm}
725: \caption{{\bf (a)}
726: Dispersion relation for the photonic crystal waveguide
727: shown in the inset ($\varepsilon_0=\varepsilon_d=11.56$,
728: $r_0=0.18a$, $r_d=0.10a$). The grey areas are the projected
729: band structure of the perfect 2D photonic crystal.
730: The frequencies at the indicated points are:
731: $\omega_A=0.346 \times 2\pi c/a$ and
732: $\omega_B=0.440 \times 2\pi c/a$.
733: {\bf (b)} Mode power $Q(\omega)$ of the nonlinear mode
734: excited in the corresponding photonic crystal waveguide.
735: Two cases are presented: the case of nonlinear rods in
736: a linear photonic crystal ({\em solid line}) and the case
737: of a completely nonlinear photonic crystal ({\em dashed
738: line}). The right inset shows the behavior of the
739: coupling coefficients $J_n(\omega)$ for
740: $n \geq 1$ ($J_0=0.045$) at $\omega=0.33 \times 2\pi c/a$.
741: The left inset shows the profile of the
742: nonlinear mode.}
743: \label{fig:x1-0.10}
744: \end{figure}
745: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
746:
747: We have solved Eq. (\ref{sys:eq-E-disc}) numerically and found
748: that nonlinearity can lead to the existence of {\em guided
749: modes localized in both directions}, i.e. in
750: the direction perpendicular to the waveguide, due to the
751: guiding properties of a channel waveguide created by
752: defect rods, and in the direction of the waveguide,
753: due to the nonlinearity-induced self-trapping effect.
754: Such nonlinear modes exist with the frequencies below the
755: frequency of the linear guided mode of the waveguide, i.e.
756: below the frequency $\omega_A$ in Fig. \ref{fig:x2-0.10}(a),
757: and are indeed unstaggered, with the bell-shaped profile
758: along the waveguide direction shown in the left inset of
759: Fig. \ref{fig:x2-0.10}(b).
760:
761: The 2D nonlinear modes localized in both dimensions can be
762: characterized by the mode power which we define, by
763: analogy with the NLS equation, as
764: \begin{equation}
765: Q = \sum_n |E_n|^2.
766: \label{sys:norm}
767: \end{equation}
768: This power is closely related to the energy of
769: the electric field in the 2D photonic crystal accumulated
770: in the nonlinear mode. In Fig. \ref{fig:x2-0.10}(b) we
771: plot the dependence of $Q$ on frequency, for the
772: waveguide geometry shown in Fig. \ref{fig:x2-0.10}(a).
773:
774: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
775: %\vspace{15mm}
776: \begin{figure}[t]
777: \begin{minipage}{55mm}
778: \centerline{{\large\bf (a)}} \vspace{1mm}
779: \includegraphics[width=55mm,angle=0,clip]{mk-fig7a.eps}
780: \end{minipage} \hspace*{3mm}
781: \begin{minipage}{55mm}
782: \centerline{{\large\bf (b)}} \vspace{1mm}
783: \includegraphics[width=58mm,angle=0,clip]{mk-fig7b.eps}
784: \end{minipage}
785: \vspace{3mm}
786: \caption{{\bf (a)}
787: Dispersion relation for the photonic crystal waveguide
788: shown in the inset ($\varepsilon_0=\varepsilon_d=11.56$,
789: $r_0=0.18a$, $r_d=0.10a$). The grey areas are the projected
790: band structure of the perfect 2D photonic crystal.
791: The frequencies at the indicated points are:
792: $\omega_A=0.352 \times 2\pi c/a$, $\omega_B=0.371 \times
793: 2\pi c/a$, and $\omega_C=0.376 \times 2\pi c/a$ (at $k=0.217
794: \times 2\pi/a$).
795: {\bf (b)} Mode power $Q(\omega)$ of the nonlinear mode
796: excited in the corresponding photonic crystal waveguide.
797: The right inset shows the behavior of the
798: coupling coefficients $J_n(\omega)$ for
799: $n \geq 1$ ($J_0=0.068$) at $\omega=0.345 \times 2\pi c/a$.
800: The left inset shows the profile of the corresponding
801: nonlinear mode.}
802: \label{fig:x12-0.10}
803: \end{figure}
804: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
805:
806: As can be seen from the Green function shown in
807: Fig. \ref{fig:green-r0.18}, the case of staggered
808: coupling coefficients $J_n(\omega)$ can be obtained for
809: the waveguide oriented in the $\vec{s}_{10}$ direction
810: with $\vec{x}_0=\vec{a}_1/2$.
811: In this case, the frequency dependence of the linear
812: guided mode of the waveguide takes the minimum at
813: $k=\pi/a$ [see Fig. \ref{fig:x1-0.10}(a)].
814: Accordingly, the nonlinear guided mode localized along the
815: direction of the waveguide is expected to exist with the
816: frequency below the lowest frequency $\omega_A$ of the
817: linear guided mode, with a staggered profile.
818: The longitudinal profile of such a 2D
819: nonlinear localized mode is shown in the left inset in
820: Fig. \ref{fig:x1-0.10}(b), together with the dependence of the
821: mode power $Q$ on the frequency (solid curve), which in
822: this case is again monotonic.
823:
824: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
825: %\vspace{10mm}
826: \begin{figure}
827: \begin{minipage}{63mm}
828: \centerline{{\large\bf (a)}}
829: %\centerline{
830: \includegraphics[width=57mm,angle=0,clip]{mk-fig8a.eps}
831: \end{minipage}
832: \hspace*{1mm}
833: \begin{minipage}{60mm}
834: \centerline{{\large\bf (b)}}
835: %\centerline{
836: \includegraphics[width=53mm,angle=0,clip]{mk-fig8b.eps}
837: \end{minipage}
838: \begin{minipage}{110mm}
839: \vspace{3mm}
840: \centerline{
841: \includegraphics[width=90mm,angle=0,clip]{mk-fig8c.eps}}
842: \end{minipage}
843: \vspace{3mm}
844: \caption{Examples of the {\bf (a)} symmetric and
845: {\bf (b)} antisymmetric localized modes.
846: The rod positions are indicated by circles and the amplitude
847: of the electric field is indicated by color [red, for
848: positive values, and blue, for negative values];
849: {\bf (c)} Power $Q$ vs. frequency dependencies
850: calculated for two modes of different symmetry in the
851: photonic crystal waveguide shown in Fig.
852: \protect\ref{fig:x12-0.10}.}
853: \label{fig:antisymm}
854: \end{figure}
855: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
856:
857: The results presented above are obtained for linear photonic
858: crystals with nonlinear waveguides created by a row of defect
859: rods. However, we have carried out the same analysis for the
860: general case of {\em a nonlinear photonic crystal} that is
861: created by rods of different size but made of the same
862: nonlinear material.
863: Importantly, we have found relatively small difference in all the
864: results presented above provided nonlinearity is
865: weak. In particular, for the photonic crystal waveguide
866: shown in Fig. \ref{fig:x1-0.10}(a), the results for linear
867: and nonlinear photonic crystals are very close. Indeed, for
868: the mode power $Q$ the results corresponding to a nonlinear
869: photonic crystal are shown in Fig. \ref{fig:x1-0.10}(b) by a
870: dashed curve, and for $Q<20$ this curve almost coincides
871: with the solid curve corresponding to the case of a
872: nonlinear waveguide embedded into a 2D linear photonic
873: crystal.
874:
875: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
876: \subsection{Instability of nonlinear localized
877: modes}
878: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
879:
880: Let us now consider the waveguide created by a row of
881: defect rods which are located at the points
882: $\vec{x}_0=(\vec{a}_1+\vec{a}_2)/2$,
883: along a straight line in either the $\vec{s}_{10}$ or
884: $\vec{s}_{01}$ directions. The results for this case are
885: presented in Figs.
886: \ref{fig:x12-0.10}--\ref{fig:antisymm}.
887: The coupling coefficients $J_n$ are described by a slowly
888: decaying staggered function of the site number $n$, so that
889: the effective interaction decays on the scale larger than
890: in the two cases considered above.
891:
892: It is remarkable that, similar to the NLS models with
893: long-range dispersive interactions
894: \cite{Gaididei:1997:PRE,Johansson:1998:PRE},
895: we find a {\em non-monotonic} behavior of the mode power
896: $Q(\omega)$ for this type of nonlinear photonic crystal
897: waveguides: specifically, $Q(\omega)$ {\em increases} in
898: the frequency interval $0.344 < (\omega a/ 2\pi c) < 0.347$
899: [shaded in Fig.~\ref{fig:antisymm}(c)]. One can expect
900: that, similar to the results earlier obtained for the
901: nonlocal NLS models
902: \cite{Gaididei:1997:PRE,Johansson:1998:PRE},
903: the nonlinear localized modes in this interval
904: are unstable and will eventually decay or
905: transform into the modes of higher or lower frequency
906: \cite{citeNLS}. What counts is that there is an interval
907: of mode power in which {\em two stable nonlinear localized
908: modes of different widths do coexist}. Since the mode
909: power is closely related to the mode energy, one can
910: expect that the mode energy is also non-monotonic function
911: of $\omega$. Such a phenomenon is known as
912: {\em bistability}, and in the problem under consideration
913: it occurs as a direct manifestation of the nonlocality of
914: the effective (linear and nonlinear) interaction between
915: the defect rod sites.
916:
917: Being interested in the mobility of the nonlinear localized
918: modes we have investigated, in addition to the symmetric modes
919: shown in the left inset in Fig. \ref{fig:x12-0.10}(b) and
920: in Fig. \ref{fig:antisymm}(a),
921: also the {\em antisymmetric localized modes} shown in
922: Fig. \ref{fig:antisymm}(b).
923: Our calculations show that the power $Q(\omega)$ of the
924: antisymmetric modes always (for all values of $\omega$
925: and all types of waveguides) exceeds that for symmetric
926: ones [see, e.g., Fig. \ref{fig:antisymm}(c)].
927: Thus, antisymmetric modes are expected to be unstable and
928: they should transform into a lower-energy symmetric
929: modes.
930:
931: In fact, the difference between the power of antisymmetric and
932: symmetric modes determines the Peierls-Nabarro
933: barrier which should be overtaken for realizing
934: the mobility of a nonlinear localized mode. One can see in Fig.
935: \ref{fig:antisymm}(c) that the Peierls-Nabarro barrier is
936: negligible for $0.347 < (\omega a/ 2\pi c) < 0.352$ and thus
937: such localized modes should be mobile. However, the
938: Peierls-Nabarro barrier becomes sufficiently large for
939: highly localized modes with
940: $\omega < 0.344 \times 2 \pi c/a$ and, as a consequence,
941: such modes should be immobile.
942: Hence, the bistability phenomenon in the photonic crystal
943: waveguides of the type depicted in Figs.
944: \ref{fig:x12-0.10}--\ref{fig:antisymm}
945: opens up fresh opportunities \cite{Johansson:1998:PRE}
946: for {\em switching} between immobile localized modes
947: (used for the energy storage) and mobile localized modes
948: (used for the energy transport).
949:
950: The foregoing discussions on the mode mobility, based
951: on the qualitative picture of the Peierls-Nabarro barrier,
952: have been established for the discrete {\em one-dimensional} arrays.
953: It is clear that the {\em two-dimensional} geometry of photonic
954: crystals under consideration
955: will bring new features into this picture. However,
956: all these issues are still open and would require a further
957: analysis.
958:
959:
960: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
961: \section{Self-Trapping of Light in a Reduced-Symmetry
962: 2D Nonlinear Photonic Crystal}
963: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
964:
965: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
966: %\vspace{10mm}
967: \begin{figure}[t]
968: \begin{minipage}{70mm}
969: \includegraphics[width=70mm,angle=0,clip]{mk-fig9a.eps}
970: \end{minipage}
971: %%\hspace*{12mm}
972: \begin{minipage}{35mm}
973: \includegraphics[width=35mm,angle=270,clip]{mk-fig9b.eps}
974: \end{minipage}
975: %\vspace{3mm}
976: \caption{Band-gap structure of the reduced-symmetry photonic
977: crystal with $r_0 = 0.1a$, $r_d = 0.05a$, and
978: $\varepsilon = 11.4$ for both types of rods.
979: Full lines are calculated by the MIT Photonic-Bands
980: program \protect\cite{mpb-prog} whereas dashed line is
981: found from the effective discrete model.
982: The top center inset shows a cross-sectional
983: view of the 2D photonic crystal depicted in the right
984: inset. The bottom center inset shows the corresponding
985: Brillouin zone.}
986: \label{fig:band}
987: \end{figure}
988: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
989:
990: A low-intensity light cannot propagate through a photonic
991: crystal if the light frequency falls into a band gap.
992: However, it has been recently suggested \cite{john} that in
993: the case of a 2D periodic medium with a Kerr-type nonlinear
994: material, high-intensity light with frequency inside the
995: gap can propagate in the form of {\em finite
996: energy solitary waves} -- {\em 2D gap solitons}.
997: These solitary waves were found to be {\em stable}
998: \cite{john}, but the conclusion was based on the
999: coupled-mode equations valid for a {\em weak modulation} of
1000: the dielectric constant $\varepsilon(\vec{x})$.
1001: However, in real photonic crystals the
1002: modulation of $\varepsilon(\vec{x})$ is {\em comparable to
1003: its average value}.
1004: Thus, the results of Ref. \cite{john} have a limited applicability to
1005: the properties of localized modes in {\em realistic photonic crystals}.
1006:
1007: More specifically, the coupled-mode equations are valid if
1008: and only if the band gap $\Delta$ is vanishingly small,
1009: i. e. $\Delta \sim A^2$ where $A$ is an effective amplitude
1010: of the mode, that is a small
1011: parameter in the multi-scale asymptotic expansions \cite{kiv}.
1012: If we apply this
1013: model to describe nonlinear modes in a wider gap (see, e.g.,
1014: discussions in Ref. \cite{kiv}), we obtain a 2D nonlinear
1015: Schr{\"o}dinger (NLS) equation
1016: known to possess {\em no stable localized solutions}.
1017: Moreover, the 2D localized modes described by the coupled-mode equations
1018: are expected to possess
1019: {\em an oscillatory instability} recently discovered
1020: for a broad class of coupled-mode Thirring-like
1021: equations \cite{dima}. Thus, it is clear that, if nonlinear
1022: localized modes do exist in realistic PBG materials, their
1023: stability should be associated with {\em different physical
1024: mechanisms} not accounted for by simplified continuum
1025: models.
1026:
1027: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1028: %%\vspace{0mm}
1029: \begin{figure}
1030: \centerline{
1031: \includegraphics[width=80mm,angle=0,clip]{mk-fig10.eps}}
1032: \vspace{3mm}
1033: \caption{Coupling coefficients $J_{n,m}(\omega)$ for
1034: the photonic crystal depicted in Fig.
1035: \protect\ref{fig:band} (the contribution of the
1036: coefficient $J_{0,0}=0.039$ is not shown).
1037: The frequency $\omega=0.4456$
1038: falls into the first band gap.}
1039: \label{fig:Jnm}
1040: \end{figure}
1041: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1042:
1043: In this Section we follow Ref. \cite{mingaleev2} and
1044: study the properties of nonlinear
1045: localized modes in a 2D photonic crystal composed of
1046: {\em two types of circular rods}: the rods of radius $r_0$ made
1047: from a linear dielectric material and placed at the corners
1048: of a square lattice with the lattice spacing $a$, and the
1049: rods of radius $r_d$ made from a nonlinear dielectric
1050: material and placed at the center of each unit cell (see
1051: right inset in Fig. \ref{fig:band}).
1052: Recently, such {\em photonic
1053: crystals of reduced symmetry} have attracted considerable
1054: interest because of their ability to possess
1055: {\em larger absolute band gaps} \cite{symmetry}.
1056: The band-gap structure of the
1057: reduced-symmetry photonic crystal is shown in
1058: Fig. \ref{fig:band}.
1059: As is seen, it possesses two band gaps, first of
1060: which extends from $\omega=0.426 \times 2 \pi c/a$ to
1061: $\omega=0.453 \times 2 \pi c/a$.
1062:
1063: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1064: %\vspace{5mm}
1065: \begin{figure}
1066: \begin{minipage}{50mm}
1067: \includegraphics[width=40mm,angle=0,clip]{mk-fig11a.eps}
1068: \end{minipage} \hspace*{-5mm}
1069: \begin{minipage}{50mm}
1070: \includegraphics[width=70mm,angle=0,clip]{mk-fig11b.eps}
1071: \end{minipage}
1072: %\vspace{5mm}
1073: \caption{Top (left) and 3D (right) views of a
1074: nonlinear localized mode in the first band gap of
1075: 2D photonic crystal depicted in
1076: Fig. \protect\ref{fig:band}.}
1077: \label{fig:mode}
1078: \end{figure}
1079: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1080:
1081: The reduced-symmetry ``diatomic'' photonic crystal shown in
1082: Fig. \ref{fig:band} can be considered as
1083: a square lattice of the ``nonlinear defect rods'' of small
1084: radius $r_d$ ($r_d < r_0$) embedded into the ordinary
1085: single-rod photonic crystal formed by a square lattice of
1086: rods of larger radius $r_0$ in air.
1087: The positions of the defect rods can then be described by the
1088: vectors $\vec{x}_{n,m} = n \, \vec{a}_{1} + m \,
1089: \vec{a}_{2}$, where $\vec{a}_1$ and $\vec{a}_2$
1090: are the primitive lattice vectors
1091: of the 2D photonic crystal. Here, in contrast to the
1092: photonic crystal waveguides discussed in the previous
1093: section, the nonlinear defect rods are characterized by
1094: two integer indices, $n$ and $m$. However, it is
1095: straightforward to extend Eq. (\ref{sys:eq-E-disc})
1096: and write an approximate 2D discrete nonlinear equation
1097: \begin{eqnarray}
1098: \label{sys:eq-E-disc:2D}
1099: i \sigma \frac{\partial}{\partial t} E_{n,m}
1100: - E_{n,m} + \sum_{k,l} J_{n-k, \, m-l}(\omega)
1101: (\varepsilon_{d}^{(0)} + |E_{k,l}|^2) E_{k,l}
1102: = 0\; ,
1103: \end{eqnarray}
1104: for the amplitudes of
1105: the electric field $E_{n,m}(t \,|\, \omega) \equiv
1106: E(\vec{x}_{n,m}, t \,|\, \omega)$ inside the
1107: defect rods.
1108: We have checked the accuracy of
1109: the approximation provided by Eq. (\ref{sys:eq-E-disc:2D})
1110: solving it in the linear limit, in order to find the band-gap structure
1111: associated with linear stationary mode.
1112: Since the coupling coefficients $J_{n,m}(\omega)$ in the
1113: photonic crystal depicted in Fig. \ref{fig:band}
1114: are highly long-ranged functions
1115: (see Fig. \ref{fig:Jnm}), one should take into account the
1116: interaction between at least 10 neighbors to reach accurate
1117: results.
1118: As is seen from Fig. \ref{fig:band}, in this case
1119: the frequencies of the linear modes
1120: (depicted by a dashed line, with a minimum at
1121: $\omega=0.446 \times 2\pi c/a$) calculated from Eq. (\ref{sys:eq-E-disc:2D})
1122: are in a good agreement with those calculated directly from Eq.
1123: (\ref{sys:eq-E-omega-t}). It lends a support to the validity
1124: of Eq. (\ref{sys:eq-E-disc:2D}) and allows us to use it for
1125: studying nonlinear properties.
1126:
1127: Stationary nonlinear modes described by Eq.
1128: (\ref{sys:eq-E-disc:2D})
1129: are found numerically by the Newton-Raphson
1130: iteration scheme. We reveal the existence of {\em a continuous family of
1131: such modes}, and a typical example [smoothed by continuous optimization for
1132: Eq. (\ref{sys:eq-E-omega-t2})] of nonlinear localized mode
1133: is shown in Fig. \ref{fig:mode}.
1134: In Fig. \ref{fig:norm}, we plot the dependence
1135: of the mode power
1136: \begin{equation}
1137: Q(\omega) = \sum_{n,m} |E_{n,m}|^2 \; ,
1138: \label{sys:norm:2D}
1139: \end{equation}
1140: on the frequency $\omega$ for
1141: the photonic crystal shown in Fig. \ref{fig:band}.
1142: As we have already discussed,
1143: this dependence represents a very important characteristic of
1144: nonlinear localized modes which allows to determine
1145: their stability by means of the Vakhitov-Kolokolov
1146: stability criterion: $dQ/d\omega>0$ for unstable modes
1147: (this criterion has been extended \cite{Laedke}
1148: to 2D NLS models).
1149:
1150: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1151: %%\vspace{0mm}
1152: \begin{figure}[t]
1153: \centerline{\hbox{
1154: \includegraphics[width=90mm,angle=0,clip]{mk-fig12.eps}}}
1155: %\vspace{3mm}
1156: \caption{Power $Q$ vs. frequency $\omega$ for
1157: the 2D nonlinear localized modes in the photonic crystal
1158: of Fig. \protect\ref{fig:band} with two different
1159: $\epsilon_d^{(0)}$. Solid lines -- stable modes,
1160: dashed lines -- unstable modes. Insets show typical profiles
1161: of stable modes, and an enlarged part of the power
1162: dependence. Grey areas show the lower and upper bands of
1163: delocalized modes surrounding the band gap.}
1164: \label{fig:norm}
1165: \end{figure}
1166: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1167:
1168: As is well known \cite{Laedke,Stabil},
1169: in the 2D discrete cubic NLS equation, only high-amplitude
1170: localized modes are stable, whereas no stable modes exist in
1171: the continuum limit. For our model, the high-amplitude modes
1172: are also stable (see inset in Fig. \ref{fig:norm}),
1173: but they are not accessible under realistic conditions:
1174: To excite such modes one should increase the refractive index
1175: at the mode center in more than 2 times.
1176: Thus, for realistic conditions and relatively small values of
1177: $\chi^{(3)}$, only low-amplitude localized modes become a
1178: subject of much interest since they can be excited in
1179: experiment. However, such modes in unbounded 2D NLS models
1180: are always unstable and either collapse or spread out
1181: \cite{DNLS}.
1182: In fact, they can be stabilized by some external forces
1183: (e.g., due to interactions with boundaries or
1184: disorder \cite{disorder}), but in this case the excitations
1185: are pinned and cannot be used for energy or signal transfer.
1186:
1187: Here we reveal that, in a sharp contrast to the
1188: 2D discrete NLS models discussed earlier in various
1189: applications, the low-amplitude localized modes
1190: of Eq. (\ref{sys:eq-E-disc:2D}) can be stabilized
1191: due to {\em nonlinear long-range dispersion} inherent to
1192: the photonic crystals. It should be emphasized that such
1193: stabilization does not occur in the models with only {\em linear long-range}
1194: dispersion \cite{DNLS}.
1195: In order to gain a better insight into the stabilization
1196: mechanism, we have carried out the studies of Eq.
1197: (\ref{sys:eq-E-disc}) for the exponentially decaying
1198: coupling coefficients $J_{n,m}$. Our results show that
1199: the most important factor which determines stability
1200: of the low-amplitude localized modes is a ratio of the
1201: coefficients
1202: at the local nonlinearity ($\sim J_{0,0}$) and the
1203: nonlinear dispersion ($\sim J_{0,1}$). If the coupling
1204: coefficients $J_{n,m}$ decrease with the distances $n$ and $m$
1205: rapidly, the low-amplitude modes of
1206: Eq. (\ref{sys:eq-E-disc:2D})
1207: with $\epsilon_d^{(0)}=11.4$ are essentially stable
1208: for $J_{0,0}/J_{0,1} \leq 13$. However, this estimation
1209: is usually lowered because the stabilization is favored
1210: by the presence of long-range interactions.
1211:
1212: It should be mentioned that the stabilization of
1213: low-amplitude 2D localized modes is not inherent to all types
1214: of nonlinear photonic crystals. On the contrary, the photonic
1215: crystals must be {\em carefully designed} to support
1216: {\em stable low-amplitude nonlinear modes}. For example, in the
1217: photonic crystal considered above such modes are stable
1218: at least for $11 < \epsilon_d^{(0)} < 12$, however they become
1219: unstable for $\epsilon_d^{(0)} \geq 12$
1220: (see Fig. \ref{fig:norm}).
1221: The stability of these modes can also be controlled by
1222: varying $r_d$, $r_0$, or $\epsilon_0$.
1223:
1224:
1225: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1226: \section{Concluding Remarks}
1227: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1228:
1229: Exploration of nonlinear properties of PBG materials
1230: may open new important application of photonic crystals
1231: for all-optical signal processing and switching,
1232: allowing an effective way to create tunable band-gap
1233: structures operating entirely with light. Nonlinear
1234: photonic crystals,
1235: and nonlinear waveguides created in the photonic structures
1236: with
1237: a periodically modulated dielectric constant, create an ideal
1238: environment for the generation and observation of nonlinear
1239: localized modes.
1240:
1241: As follows from our results, nonlinear localized modes can
1242: be excited in photonic crystal
1243: waveguides of different geometry. For several geometries
1244: of 2D
1245: waveguides, we have demonstrated that such modes
1246: are described by a new type of nonlinear lattice models
1247: that include
1248: long-range interaction and effectively nonlocal nonlinear
1249: response.
1250: It is expected that the general features of nonlinear guided modes
1251: described here will be preserved in other types of photonic crystal
1252: waveguides. Additionally, similar types of nonlinear localized modes are
1253: expected in photonic crystal fibers \cite{russell} consisting of a
1254: periodic air-hole lattice that runs along the length of the fiber,
1255: provided the fiber core is made of a highly nonlinear material
1256: (see, e.g., Ref. \cite{egg}).
1257:
1258: Experimental observation of nonlinear photonic localized modes would
1259: require not only the use of photonic materials with a relatively large
1260: nonlinear refractive index (such as AlGaAs waveguide PBG
1261: structures \cite{algas} or polymer PBG crystals \cite{jap}, but also
1262: a control of the group-velocity dispersion and band-gap parameters.
1263: The latter can be achieved by employing the surface coupling
1264: technique \cite{ast} that is able to provide coupling to specific
1265: points of the dispersion curve, opening up a very straightforward way
1266: to access nonlinear effects.
1267:
1268:
1269: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1270: \section*{Acknowledgments}
1271: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1272:
1273: The authors are indebted to O. Bang, K. Busch,
1274: P.L. Christiansen, Yu.B. Gaididei, S. John, A. McGurn,
1275: C. Soukoulis, and A.A. Sukhorukov for encouraging
1276: discussions, and R.A. Sammut for collaboration at
1277: the initial stage of this project.
1278: The work has been partially supported by the Large Grant
1279: Scheme of the Australian Research Council and the
1280: Performance and Planning Foundation grant of the
1281: Institute of Advanced Studies.
1282:
1283: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1284: \begin{thebibliography}{99}
1285: \addcontentsline{toc}{section}{References}
1286: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1287:
1288: \bibitem{book} J. D. Joannoupoulos, R. B. Meade, and
1289: J. N. Winn, {\em Photonic Crystals: Molding the Flow of Light}
1290: (Princeton University Press, Princeton N.J., 1995).
1291:
1292: \bibitem{opt} J. G. Fleming and S.-Y. Lin,
1293: Opt. Lett. {\bf 24}, 49 (1999).
1294:
1295: \bibitem{john2} See, e.g., K. Busch and S. John,
1296: Phys. Rev. Lett. {\bf 83}, 967 (1999),
1297: and discussions therein.
1298:
1299: \bibitem{berger} V. Berger, Phys. Rev. Lett. {\bf 81}, 4136 (1998);
1300: and also the recent experiment: N.G.R. Broderick, G.W. Ross, H.L. Offerhaus, D.J. Richardson, and D.C. Hanna, Phys. Rev. Lett. {\bf 84}, 4345 (2000); S. Saltiel and Yu.S. Kivshar, Opt. Lett. {\bf 25}, 1204 (2000).
1301:
1302:
1303: \bibitem{review} See, e.g., S. Flach and C.R. Willis, Phys. Rep. {\bf 295}, 181
1304: (1998); O.M. Braun and Yu.S. Kivshar, Phys. Rep. {\bf 306}, 1 (1998), Chap. 6.
1305:
1306: \bibitem{mak} R.S. MacKay and S. Aubry, Nonlinearity {\bf 7}, 1623 (1994);
1307: see also S. Aubry, Physica D {\bf 103}, 201 (1997).
1308:
1309: \bibitem{bishop} B.I. Swanson, J.A. Brozik, S.P. Love, G.F. Strouse, A.P.
1310: Shreve, A.R. Bishop, W.Z. Wang, and M.I. Salkola, Phys. Rev. Lett. {\bf
1311: 82}, 3288 (1999).
1312:
1313: \bibitem{sievers} U.T. Schwarz, L.Q. English, and A.J. Sievers, Phys. Rev.
1314: Lett. {\bf 83}, 223 (1999).
1315:
1316: \bibitem{JJ} E. Trias, J.J. Mazo, and T.P. Orlando, Phys. Rev. Lett. {\bf
1317: 84}, 741 (2000); P. Binder, D. Abraimov, A.V. Ustinov, S. Flach, and Y.
1318: Zolotaryuk, Phys. Rev. Lett. {\bf 84}, 745 (2000).
1319:
1320: \bibitem{zolo} F.M. Russel, Y. Zolotaryuk, and J.C. Eilbeck, Phys. Rev. B
1321: {\bf 55}, 6304 (1997).
1322:
1323: \bibitem{silb} H.S. Eisenberg, Y. Silberberg, R. Marandotti, A.R. Boyd, and
1324: J.S. Aitchison, Phys. Rev. Lett. {\bf 81}, 3383 (1998).
1325:
1326:
1327: \bibitem{sukh} A.A. Sukhorukov, Yu.S. Kivshar, O. Bang, J. Martorell, J.
1328: Trull, and R. Vilaseca, Optics and Photonics News {\bf 10} (12) 34 (1999).
1329:
1330: \bibitem{john} S. John and N. Ak\"ozbek,
1331: Phys. Rev. Lett. {\bf 71}, 1168 (1993);
1332: Phys. Rev. E {\bf 57}, 2287 (1998).
1333:
1334: \bibitem{sukh2} A.A. Sukhorukov, Yu.S. Kivshar, and O. Bang, Phys. Rev. E
1335: {\bf 60}, R41 (1999).
1336:
1337: \bibitem{mcgurn}
1338: A.~R. McGurn, Phys. Lett. A {\bf 251}, 322 (1999);
1339: Phys. Lett. A {\bf 260}, 314 (1999).
1340:
1341: \bibitem{mingaleev}
1342: S.~F. Mingaleev, Yu.~S. Kivshar, and R.~A. Sammut,
1343: Phys. Rev. E {\bf 62}, 5777 (2000).
1344:
1345: \bibitem{mingaleev2}
1346: S.~F. Mingaleev and Yu.~S. Kivshar,
1347: Phys. Rev. Lett. {\bf 86} (June 2001), in print
1348: [arXiv:cond-mat/0102066].
1349:
1350: \bibitem{Maradudin:1993:PBGL}
1351: A.~A. Maradudin and A.~R. McGurn, in {\em Photonic Band Gaps and
1352: Localization}, {\em NATO ASI Series B: Physics}, Vol. 308,
1353: Ed. C.~M. Soukoulis (Plenum Press, New York, 1993), p. 247.
1354:
1355: \bibitem{Mekis:1996:PRL}
1356: A. Mekis, J.C. Chen, I. Kurland, S. Fan, P.R. Villeneuve, and
1357: J.D. Joannopoulos, Phys. Rev. Lett. {\bf 77}, 3787 (1996).
1358:
1359: \bibitem{Mekis:1998:PRB}
1360: A. Mekis, S. Fan, and J.~D. Joannopoulos, Phys. Rev. B {\bf 58}, 4809 (1998).
1361: \bibitem{Ward:1998:PRB}
1362: A.~J. Ward and J.~B. Pendry, Phys. Rev. B {\bf 58}, 7252 (1998).
1363:
1364: \bibitem{Lin:1998:SCI}
1365: S.-Y. Lin, E. Chow, V. Hietala, P.R. Villeneuve, and
1366: J.D. Joannopoulos, Science {\bf 282}, 274 (1998).
1367:
1368: \bibitem{Tokushima:2000:APL}
1369: M. Tokushima, H. Kosaka, A. Tomita, and H. Yamada,
1370: Appl. Phys. Lett. {\bf 76}, 952 (2000).
1371:
1372: \bibitem{Cheng:1999:PRB}
1373: S.~S.~M. Cheng, L.~M. Li, C.~T. Chan, and Z.~Q. Zhang,
1374: Phys. Rev. B {\bf 59}, 4091 (1999).
1375:
1376: \bibitem{Jin:1999-sep:APL}
1377: C. Jin, B. Cheng, B. Mau, Z. Li, D. Zhnag, S. Ban, B. Sun,
1378: Appl. Phys. Lett. {\bf 75}, 1848 (1999).
1379:
1380: \bibitem{Fan:1998-feb:PRL}
1381: S. Fan, P.~R. Villeneuve, J.~D. Joannopoulos, and
1382: H.~A. Haus, Phys. Rev. Lett. {\bf 80}, 960 (1998).
1383:
1384: \bibitem{Gaididei:1997:PRE}
1385: Y.~B. Gaididei, S.~F. Mingaleev, P.~L. Christiansen, and K.~{\O}. Rasmussen,
1386: Phys. Rev. E {\bf 55}, 6141 (1997).
1387:
1388: \bibitem{Johansson:1998:PRE}
1389: M. Johansson, Y.~B. Gaididei, P.~L. Christiansen, and K.~{\O}.
1390: Rasmussen, Phys. Rev. E {\bf 57}, 4739 (1998).
1391:
1392: \bibitem{citeNLS} See, e.g., the examples for the continuous
1393: generalised NLS models, D.~E. Pelinovsky, V.~V. Afanasjev, and
1394: Yu.~S. Kivshar, Phys. Rev. E {\bf 53}, 1940 (1996).
1395:
1396: \bibitem{symmetry} C. M. Anderson and K. P. Giapis,
1397: Phys. Rev. Lett. {\bf 77}, 2949 (1996);
1398: Phys. Rev. B {\bf 56}, 7313 (1997).
1399:
1400: \bibitem{mpb-prog} S. G. Johnson,
1401: http://ab-initio.mit.edu/mpb/
1402:
1403: \bibitem{kiv} Yu. S. Kivshar, O. A. Chubykalo, O. V. Usatenko, and
1404: D. V. Grinyoff, Int. J. Mod. Phys. B {\bf 9}, 2963 (1995).
1405:
1406: \bibitem{dima}
1407: I. V. Barashenkov, D. E. Pelinovsky, and E. V. Zemlyanaya,
1408: Phys. Rev. Lett. {\bf 80}, 5117 (1998);
1409: A. De Rossi, C. Conti, and S. Trillo,
1410: Phys. Rev. Lett. {\bf 81}, 85 (1998).
1411:
1412: \bibitem{DNLS} See, e.g.,
1413: V.~K. Mezentsev, S.~L. Musher, I.~V. Ryzhenkova,
1414: and S.~K. Turitsyn, JETP Lett. {\bf 60}, 829 (1994);
1415: S. Flach, K. Kladko, and R.~S. MacKay,
1416: Phys. Rev. Lett. {\bf 78}, 1207 (1997);
1417: P.~L. Christiansen {\em et al.},
1418: Phys. Rev. B {\bf 57}, 11303 (1998).
1419:
1420: \bibitem{Laedke}
1421: E.~W.~Laedke {\em et al.}, JETP Lett. {\bf 62}, 677 (1995).
1422:
1423: \bibitem{Stabil}
1424: E.~W.~Laedke, K. H. Spatschek, S. K. Turitsyn, and
1425: V. K. Mezentsev, Phys. Rev. E {\bf 52}, 5549 (1995);
1426: Yu. B. Gaididei, P. L. Christiansen, K. {\O}. Rasmussen, and
1427: M. Johansson, Phys. Rev. B {\bf 55}, R13365 (1997).
1428:
1429: \bibitem{disorder}
1430: Yu. B. Gaididei, D. Hendriksen, P. L. Christiansen, and
1431: K. {\O}. Rasmussen, Phys. Rev. B
1432: {\bf 58}, 3075 (1998).
1433:
1434: \bibitem{russell} T.A. Birks, J.C. Knight, and P.St.J. Russell, Opt. Lett.
1435: {\bf 22}, 961 (1997).
1436:
1437: \bibitem{egg}
1438: B.J. Eggleton, P.S. Westbrook, R.S. Windeler, S. Sp\"alter,
1439: and T.A. Strasser, Opt. Lett. {\bf 24}, 1460 (1999).
1440:
1441: \bibitem{algas} P. Millar {\em et al.},
1442: Opt. Lett. {\bf 24}, 685 (1999); A.A. Helmy {\em et al.}, Opt. Lett. {\bf 25}, 1370 (2000).
1443:
1444: \bibitem{jap} S. Shoji and S. Kawata,
1445: Appl. Phys. Lett. {\bf 76}, 2668 (2000).
1446:
1447: \bibitem{ast} V. N. Astratov {\em et al.},
1448: Phys. Rev. B {\bf 60}, R16255 (1999).
1449:
1450: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1451: \end{thebibliography}
1452: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1453: \end{document}
1454: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1455: