nlin0011043/prl01.tex
1: %\documentstyle[aps,prl,psfig,preprint]{revtex}
2: %\documentstyle[aps,aps10,psfig,prl,multicol]{revtex} 
3:  
4: 
5: 
6: \documentstyle[multicol,aps,psfig,prl]{revtex}
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8: %TCIDATA{OutputFilter=LATEX.DLL}
9: %TCIDATA{Created=Sat Dec 18 18:15:56 1999}
10: %TCIDATA{LastRevised=Mon Nov 20 23:47:59 2000}
11: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
12: %TCIDATA{Language=American English}
13: %TCIDATA{CSTFile=revtex.cst}
14: 
15: \begin{document}
16: \title{Pulse confinement in optical fibers with random
17: dispersion}
18: \author{M. Chertkov, I. Gabitov, J. Moeser}
19: \address{Theoretical Division, LANL, Los Alamos, NM
20: 87545}
21: \date{\today}
22: 
23: \maketitle
24: 
25: \begin{abstract}
26: Short range  correlated uniform  noise in the  dispersion coefficient,
27: inherent  in many  types of  optical fibers,  broadens  and eventually
28: destroys  all  initially   ultra-short  pulses.   However,  under  the
29: constraint that the integral of the random component of the dispersion
30: coefficient   is   set   to   zero,   or   pinned,   periodically   or
31: quasi-periodically  along   the  fiber,   the  nature  of   the  pulse
32: propagation  changes dramatically.   For the  case that  randomness is
33: added  to  constant   positive  dispersion,  the  pinning  restriction
34: significantly reduces pulse broadening.  If the randomness is added to
35: piecewise constant  periodic dispersion, the pinning  may even provide
36: probability  distributions of  pulse parameters  that  are numerically
37: indistinguishable  from the  statistically steady  case.   The pinning
38: method  can be  used to  both  manufacture better  fibers and  upgrade
39: existing fiber links.
40: \end{abstract}
41: 
42: \pacs{42.81.Dp, 05.45.-a, 46.65.+g}
43: 
44: %\footnote{
45: %Theoretical Division, LANL, Los Alamos, NM 87545 . 
46: %$^\#$To whom correspondence should be addressed.E-mail:
47: %chertkov@lanl.gov . 
48: %$^1$Landau Institute of Theoretical Physics,
49: %117334 Moscow, Russia. $^2$Division of Applied Mathematics, Brown
50: %University, Providence, RI 02806.}$^{,\#}$, 
51: 
52: \begin{multicols}{2}
53: 
54: 
55: \bigskip  The  effect  of   random  perturbations  in  optical  fibers
56: increasingly  attracts attention  as  the demand  for  the quality  of
57: transmission  grows  daily~  \cite{Recent  reviews}.   The  impact  of
58: randomness on signal transmission in  a single mode fiber is negative;
59: it   causes  degradation   of   the  signal   and  lowers   transition
60: capabilities~\cite{Multimode}.       In      particular,     amplifier
61: noise~\cite{Gordon-Hauss},  and  noise   in  the  fiber  birefringence
62: (double  refraction)~\cite{PMD} lead  to  random shifts  in the  pulse
63: position (timing jitter) and  to pulse broadening, respectively.  Both
64: effects eventually  cause a  destruction of bit-patterns  and eventual
65: increase  of the  Bit-Error-Rate (BER),  the most  important parameter
66: describing performance in fiber communications systems~\cite{97A}.
67: 
68: In the  present paper,  we consider the  effect of  random dispersion,
69: which  is, for  ultrashort  pulses, potentially  as  dangerous as  the
70: aforementioned effects.   We, however, propose a  way to significantly
71: reduce the pulse deterioration and eventually reduce the BER caused by
72: the  noise  in  dispersion  by  using passive  (independent  of  pulse
73: properties)  periodic control  of  the accumulated  dispersion of  the
74: fiber  link. Furthermore,  the method  may even  provide statistically
75: steady propagation of the pulse along the fiber.
76: 
77: Chromatic dispersion  is an important  characteristic of a  medium and
78: can significantly degrade the  integrity of wave packets. In practice,
79: chromatic dispersion  is not uniformly distributed  and often exhibits
80: random  variations  in  space  and  time.  On  the  other  hand,  wave
81: propagation through  the medium is  usually much faster  than temporal
82: variations  of  the  chromatic  dispersion.  Therefore,  these  random
83: variations  can be  treated as  ``spatial'' multiplicative  noise that
84: does not change in time. This multiplicative noise is conservative and
85: the  wave  energy  remains  constant during  propagation  through  the
86: medium.   Recently,  high precision  measurements  of fiber  chromatic
87: dispersion as  function of a fiber  length experimentally demonstrated
88: the significance of the dispersion randomness ~\cite{96MMN,98GM}.
89: 
90: The overall  chromatic dispersion in  an optical fiber comes  from two
91: sources.  The first source is  the medium itself. The second source is
92: the specific geometry of the waveguide profile. Material dispersion in
93: the  optical  fiber  is   a  relatively  stable  parameter,  uniformly
94: distributed  along the  fiber.  However,  waveguide dispersion  is not
95: nearly as  stable.  Existing technology does not  yet provide accurate
96: control of  the waveguide geometry of modern  fibers, where dependence
97: of dispersion on wavelength is complex. As a result, the magnitudes of
98: random variations of fiber chromatic dispersion are typically the same
99: as,  or   in  some   cases  even  greater   than  that  of   the  mean
100: dispersion~\cite{96MMN,98GM}.
101: 
102: In the short-wavelength regime,  a universal description of the signal
103: envelope in the reference frame  moving with the packet group velocity
104: is  given by  the  nonlinear Schr\"{o}dinger  equation  (NLS) for  the
105: complex scalar field, $\psi (z;t)$, see for example \cite{97A},
106: \begin{equation}
107: -i\partial _{z}\psi =d(z)\partial _{t}^{2}\psi +2\psi ^{2}\bar{\psi}.
108: \label{disp}
109: \end{equation}
110: The    equation    is    written    in    the    dimensionless    form
111: \cite{Param}. Variations in the  medium enter this description through
112: the
113: dispersion coefficient $%
114: d(z)=d_{\det }(z)+\xi (z)$, which is decomposed into its deterministic
115: part, $d_{\det  }(z)$, and a random  part, $\xi (z)$. Here  $z$ is the
116: position along  the fiber and $t$  is the retarded  time.  The initial
117: profile $\psi  (0;t)$ is localized  in $t$. We consider  two different
118: models  of deterministic  dispersion, both  of which  are  standard in
119: fiber  optic  communications.   Model  (A)  is the  case  of  constant
120: dispersion, $d_{\det }=d_{0}$. In  the absence of noise ($\xi (z)=0$),
121: $\psi _{0}(z;t)=a\exp \left[
122: izd_{0}/b^{2}\right] sech%
123: \left[  t/b\right]  $,  where  $a^{2}b^{2}=d_{0}$  ($a$  is  the  peak
124: amplitude, $b$  is the pulse width),  is an exact  soliton solution of
125: (\ref{disp}). The  existence of the soliton \cite{72ZS}  is the result
126: of a dynamic equilibrium  between dispersion and nonlinearity: the two
127: spatial
128: scales, nonlinearity $%
129: z_{NL}=1/a^{2}$,  and dispersion $z_{d}=b^{2}/d_{0}$,  coincide. Model
130: (B)  is the case  of dispersion  management (DM),  $d_{\det }=d_{0}\pm
131: d_{DM}$ \cite {80LKC}. Here dispersion is piecewise constant: positive
132: and   negative    spans   alternate   with    period   $z_{DM}$,   and
133: $0<d_{0}<d_{DM}$.  There is no exact  solution for the pure (no noise)
134: Model (B), but theoretical  evidence, confirmed by extensive numerical
135: studies  and  experimental  results,  indicates  the  existence  of  a
136: breathing  solution   (DM  soliton)  with  a   nearly  Gaussian  shape
137: \cite{96GT,96SKDBB,99TSSM,99GGDBMGHFG}. The localized solution here is
138: again due  to the  interplay of dispersion  and nonlinearity.   In the
139: presence  of a  periodic  dispersion map,  however,  the (DM)  soliton
140: acquires  an important  characteristic, quadratic  phase  (chirp).  In
141: contrast to conventional soliton  solutions, DM solitons can exist for
142: zero (or even negative) values of average dispersion.
143: %~\cite{Kodama,Doran}. 
144: 
145: Approximate scale  characteristics of the dispersion  noise present in
146: real     fibers     can     be     extracted     from     experimental
147: results~\cite{96MMN,98GM}. These results  show that the smallest scale
148: of noticeable  change in the  dispersion value is  approximately $\sim
149: 1-2  km$. For  constant  dispersion fibers  (model  A), the  amplifier
150: spacing is $\sim  50-60 km$, and for dispersion  managed fibers (model
151: B), the  period of a typical  dispersion map is also  $\sim 50-60 km$.
152: These scales  are much longer  than that of the  dispersion variation,
153: justifying the  idealized consideration of  delta-correlated noise for
154: both  models.  Previously,  the  stability of  initial  pulses in  the
155: presence  of the  short-range-correlated Gaussian  uniform  noise $\xi
156: _{u}$ with zero mean,
157: %\begin{equation}
158: $\left\langle  \xi  _{u}(z_{1})\xi  _{u}(z_{2})\right\rangle  =D\delta
159: \left(
160: z_{1}-z_{2}\right)$, %\label{delta}\end{equation}
161: was studied  for both  models (A) and  (B).  For model  (A), adiabatic
162: theory, valid if the noise  is weak, shows dynamical broadening of the
163: pulse and its eventual destruction~\cite{98ACS}.  The pulse-broadening
164: effects of  $\xi _{u}$  on model (B)  were studied numerically  and by
165: means of a variational approach~\cite{98ACS,00AB}.
166: 
167: The problem  can also be addressed  in the limit of  strong noise.  On
168: short scales  the nonlinearity is weak and  propagation is essentially
169: linear : only  the phase of the pulse is changed  by a rapidly varying
170: dispersion,  while its  frequency  spectrum is  not.   In this  weakly
171: nonlinear case  the following change of variables  is suggested, $\psi
172: \left(  z;t\right)  =\int_{-\infty }^{\infty  }d\omega  \exp \left[  -
173: i\left(
174: \omega t+\omega ^{2} %
175: \left[   \int_{0}^{z}\left(  d(z^{\prime   })-d_{0}\right)  dz^{\prime
176: }\right] \right) \right] \psi  _{\omega }(z),$ where $\psi_\omega (z)$
177: describes the pulse evolution on  longer scales.  The equation for the
178: noise average of the  slow filed $ \varphi _{\omega}=\left\langle \psi
179: _{\omega }(z)\right\rangle $ derived from (\ref{disp}) is
180: \begin{eqnarray} & &
181: \left(  -i\partial  _{z}\!+\!id_{0}\omega ^{2}\right)\varphi  _{\omega
182: }\!=\!2\int  d\omega_{1,2,3}   \delta  \left(  \omega  _{1}\!+\!\omega
183: _{2}\!-\!\omega  _{3}\!-\!\omega   \right)  \nonumber  \\   &&  \times
184: \exp\left[  -i\Delta   \int\limits_{0}^{z}dz^{\prime  }\left(  d_{\det
185: }-d_{0}\right)\right]\    e^{-\Delta   ^{2}Dz_{\ast    }/2}\   \varphi
186: _{1}\varphi _{2}\bar{\varphi}_{3}.  \label{kernel}
187: \end{eqnarray}
188: Here,   $\Delta    \equiv   \omega   _{1}^{2}+\omega   _{2}^{2}-\omega
189: _{3}^{2}-\omega  ^{2}$, and  the second  exponential on  the  right of
190: (\ref{kernel}), which we will call the kernel, is the average of $\exp
191: \left[ -i\Delta \int_{0}^{z}\xi \left( z^{\prime }\right) dz^{\prime
192: }\right] $. $z_{\ast}=z$%
193: ,   thereby   setting  the   correlation   scale,  $z_{\xi   }=[\Delta
194: ^{2}D]^{-1}\sim   \lbrack   \omega   ^{4}D]^{-1}\sim  b^{4}/D$.    The
195: approximation  is  justified,  i.e.    the  nonlinearity  is  weak  if
196: $z_{\xi}\ll  z_{NL}$.  The exponential  decay of  the kernel  with $z$
197: disrupts the balance between nonlinearity and dispersion necessary for
198: steady pulse propagation.
199: 
200: The grim analytical conclusion is that the natural noise in dispersion
201: leads to destruction of initially  localized signal in the extremes of
202: both weak
203: %($z_{\xi}\gg z_{NL}$) 
204: and strong
205: %($z_{\xi}\ll z_{NL}$)
206: noise.   Numerical study  of  the intermediate  range  shows the  same
207: effect.
208: 
209: The natural  question is that of  the existence of  {\em an artificial
210: constraint}  which  may  reduce   or  completely  prevent  this  pulse
211: broadening. We  demonstrate that such a constraint  does indeed exist,
212: and can be readily implemented in real fibers. All that is required is
213: that {\em the accumulated  dispersion, $\int_{0}^{z}dy\xi (y)$, is set
214: to zero, or pinned,  either periodically or quasi-periodically} with a
215: period of the  order or less than $z_{\xi  }$.  The resulting Gaussian
216: nonuniform noise, $\xi _{n}$
217: with zero mean is described by %\begin{equation}
218: $\left\langle \xi _{n}(y)\xi _{n}(z)\right\rangle
219: =D\left( \delta (z-y)-%
220: \frac{1}{l_{j+1}-l_{j}}\right) $ if
221: %\label{deltazeroInt}\end{equation}
222: $y$ and $z$ belong to the  same segment bounded by an adjacent pair of
223: pinning points.
224: %i.e.,
225: %$l_{j}<y,z<l_{j+1}$, where $j$ is the
226: %pinning points label. 
227: Otherwise there are no correlations.
228: %, i.e. $\left\langle \xi _{n}(y)\xi _{n}(z)\right\rangle =0$. 
229: 
230: 
231: Consider the  effect of nonuniform  noise, $\xi=\xi_n$, in  the weakly
232: nonlinear    case.    $z_\ast$    in   (\ref{kernel})    is   $z\left(
233: l_{j+1}-z\right)  /\left( l_{j+1}-l_{j}\right)  $  with $z\in  \lbrack
234: l_{j};l_{j+1}]$.  Thus  the decay of  the kernel in  (\ref{kernel}) is
235: replaced  by oscillatory  (or  quasi-oscillatory, when  the period  is
236: fluctuating)  behavior. In the  nonuniform case,  additional averaging
237: over  a  single  pinning  leg  reduces  the  $z$-dependent  kernel  of
238: (\ref{kernel})  to  $\exp  \left[  -\Delta  ^{2}Dl/4\right]  \sqrt{\pi
239: /\left[    \Delta   ^{2}Dl\right]    }\mbox{Erfi}\left[   \sqrt{\Delta
240: ^{2}Dl}/4\right] $.  In Model  (B), independent averaging of the first
241: exponential  on the  right-hand side  of (\ref{kernel})  over $z_{DM}$
242: (see
243: also~\cite{96GT,96SKDBB}%
244: ) replaces it by $2\sin \left[ \Delta z_{DM}/4\right]
245: /\left[ \Delta z_{DM}%
246: \right] $. The  overall result of the averaging  procedure in the case
247: of $\xi =\xi _{n}$ is the emergence of a nonlinear term which
248: does not vanish as $%
249: z\rightarrow \infty $, in contrast  to the case $\xi =\xi _{u}$, where
250: the nonlinearity dies away with $z\rightarrow \infty $.
251: One also concludes that $%
252: z_\xi$ sets the critical pinning period: the pinning
253: is efficient only if $%
254: l=l_{i+1}-l_i\lesssim z_\xi$. It is possible to show that the averaged
255: equation  for $\xi  =\xi  _{n}$ has  a steady  solution~\cite{00CGMb}.
256: Notice also  that the averaged kernel decays  exponentially as $\Delta
257: ^{2}\rightarrow \infty $. This is much faster than the algebraic decay
258: obtained  from  averaging  over  the  dispersion map  period  in  pure
259: dispersion   management.     Therefore,   the   model   of~\cite{99ZM}
260: corresponding to  a very narrow  type of nonlinear kernel  (in $\omega
261: $-space) is  better suited for the  case considered here  than for the
262: case of  pure DM for which  it was originally  proposed.  The averaged
263: equation approach  is {\it  \'{a} priori} applicable  for a  large but
264: finite $z$,  so the emergence of  a steady localized  solution for the
265: averaged equation does not mean  that the original problem possesses a
266: steady state.  In the  direct numerical simulation of (\ref{disp}), to
267: which we now  switch our attention, the effect  is seen as essentially
268: limiting the pulse broadening.
269: 
270: We perform  numerical investigations of  both models (A) and  (B) with
271: $\xi =\xi  _{u}$ and  $\xi=\xi_{n}$ in intermediate  case, $z_{NL}\sim
272: z_\xi$.
273: %The initial ($z=0$) pulse is the exact soliton for model (A) and it is
274: %a Gaussian pulse close to the respective DM soliton for model (B). 
275: Fourier split-step scheme with
276: %to solve models (A) and (B)
277: %numerically for the two cases, $\xi =\xi _{u}$ and $\xi =\xi _{n}$. 
278: $2^{13}$ temporal Fourier modes and periodic conditions imposed on the
279: boundaries of the domain $t\in [-180,180]$ is implemented.
280: The spatial step is $%
281: z_{step}=0.01$ and  the numerical  convergence was checked  by varying
282: the  size of the  periodic box  and number  of the  Fourier harmonics.
283: Parameters for the  initial signal were chosen to  be $d_{0}=1$, $a=1$
284: in
285: model (A), and $%
286: d_{0}=0.15$,                 $d_{DM}=0.1$,                 $z_{DM}=1$,
287: $|\psi(0;t)|=0.79\exp(-t^2/2.6)$ in model (B).  The setup in model (B)
288: is  borrowed  from~\cite{99TSSM}  and  corresponds  to  experimentally
289: available   DM  fibers.   Gaussian   zero-mean  noise   correlated  at
290: $z_{n}=0.1$ with amplitude $d_{n}=1$
291: models the $\delta $%
292: -correlated uniform noise with $D=d_{n}^{2}z_{n}=0.1$.  The nonuniform
293: noise  is  constructed  from   the  uniform  noise  by  the  following
294: subtraction at every pinning leg: $\xi _{n}(z)=\xi
295: _{u}(z)-\frac{1}{l_{j+1}-l_{j}}%
296: \int_{l_{j}}^{l_{j+1}}dy\xi (y)$. Pinning  strategies of two types are
297: considered: strictly periodic,  $l_{j+1}-l_{j}=l$, where $l$ is fixed;
298: and  quasi-periodic, $l_{j+1}-l_{j}=l(1+\eta  )$, where  $\eta $  is a
299: random number  uniformly distributed between $\pm  1/2$.  The averaged
300: (or otherwise strict) pinning period  for the nonuniform case is taken
301: to be $1  $, $5$ or $10$. The simulation runs  until the pulse arrives
302: at  $z=95$.  Statistics  were collected  for $10^2-10^{3}$,  and  in a
303: special case $10^{4}$, realizations.
304: 
305: 
306: \begin{figure}[tbp]
307: \psfig{file=figa.eps,width=8cm,height=5.5cm}
308: %\hskip1cm
309: \psfig{file=figb.eps,width=8cm,height=5.5cm}
310: %\caption{Averaged pulse-width as a function of $z$ measured at the
311: %half-of-the-peak-amplitude position. Fig1A and 
312: %Fig1B correspond to the model A and B respectively. }
313: %\label{Fig1}
314: \end{figure}
315: 
316: The averaged pulse-width  (full width at half maximum  amplitude) as a
317: function  of $z$  is shown  in Figure  1, see  also  \cite{Movie}. The
318: capital letter  subscript of  the figures corresponds  to the  type of
319: model, (A) or (B). Solid black, red, green, and blue represent uniform
320: noise,   and   nonuniform  noise   with   pinning  period   $l=1,5,10$
321: respectively.  The  quasi-periodic curves are  dashed and of  the same
322: color as the respective periodic ones.
323: 
324: For Model(A), all types  of nonuniform noise demonstrate a significant
325: reduction  in the  rate of  pulse  broadening when  compared with  the
326: uniform  case.  The  individual configurations  that  degrade (through
327: pulse splitting,  etc.) in the  uniform case maintain  pulse integrity
328: when each  type of nonuniform compensation is  applied. The dependence
329: on the  pinning period is monotonic:  the peak amplitude  of the pulse
330: decays faster as the  pinning period increases. The difference between
331: periodic and respective quasi-periodic cases is minor, with a slightly
332: better  confinement   observed  for  the   quasi-periodic  case.   The
333: destruction of the pulse is also accompanied by emission of continuous
334: radiation by the  soliton. The radiation is clearly  seen in the movie
335: made for individual runs~\cite{Movie}.  Once the radiation reaches the
336: boundaries of  the box, it reflects  and starts to  interfere with the
337: still localized  solution.  The latter shows  up in the  change of the
338: averaged-width   behavior  at  larger   distances,  $z\sim   20$.   In
339: principle, one could introduce absorbing boundary conditions to enable
340: longer-time simulations. However,  these conditions are artificial, as
341: in  real  data streams  pulses  are  not  isolated. Each  pulse  emits
342: continuous  radiation  which  eventually  interacts  with  neighboring
343: pulses.   In this paper  we purposefully  study self-interaction  of a
344: pulse and its radiation via reflecting boundary conditions, which is a
345: simple model for the behavior of a real bit stream. Our results reveal
346: that after  $z\sim 20$, the  rate of pulse disintegration  is reduced,
347: suggesting that in  a real bit pattern, pulse  disintegration might be
348: prevented by radiation absorption from neighboring pulses.
349: 
350: The  effect  of nonuniform  noise  is more  dramatic  in  the case  of
351: Model(B). For  nonuniform compensation with the  averaged period $l=1$
352: (and also  less) one observes  a tendency toward  statistically steady
353: behavior: the  average pulse  width does not  decay (in contrast  to a
354: decay in  the uniform case),  and the PDF  of the pulse width  (and of
355: other  variables   characterizing  the  pulse   propagation,  such  as
356: amplitude)  does not  change shape  with $z$~\cite{D5}.   There  is no
357: visible emission of  radiation by the localized solution  for any case
358: of Model (B). We have  also checked that temporal and spatial averages
359: (for example for the PDF of the pulse width) coincide.  The dependence
360: on  the  type  of  compensation  (for $l>1$)  is  monotonic,  and  the
361: difference between the  random and quasi-random cases is  minor, as in
362: the case of Model(A).
363: 
364: Notice that the tremendous reduction in the pulse decay is achieved by
365: very minor changes in dispersion (see \cite{Movie}, for the comparable
366: plots of the dispersion profiles).
367: 
368: Our  analysis  gives  practical  recommendations for  improving  fiber
369: system  performance  that  is   limited  by  randomness  in  chromatic
370: dispersion.   The  limitation  originates  from  accumulation  of  the
371: integral dispersion.  The distance between naturally occurring nearest
372: zeros of the accumulated dispersion grows with the fiber length ($\sim
373: \sqrt{z}$  as  it is  shown  in  \cite{00CGMb}).   This growth  causes
374: asymptotic decay  of the nonlinearity  with $z$ and,  therefore, pulse
375: degradation.   We have  shown that  the  signal can  be stabilized  by
376: periodic   or  quasi-periodic  pinning   of  the   accumulated  random
377: dispersion \cite{different}.  This can  be achieved by first measuring
378: the mismatch between the nominal  dispersion of the fiber line and the
379: actual accumulated dispersion, and  second, inserting a small fiber to
380: compensate this mismatch \cite{00MMGNMGV}.
381: 
382: 
383: We are grateful to G.D. Doolen, G. Falkovich, I.
384: Fatkullin, J. Hesthaven, I.
385: Kolokolov, P. Lushnikov, P. Mamyshev,  F.G. Omenetto,
386: H. Rose, T. Schaefer,
387: Z. Toroczkai, and S. Tretiak  for constructive
388: comments. M.C. wishes to
389: acknowledge the support of  J.R. Oppenheimer
390: fellowship and the hospitality
391: of ITP Santa  Barbara, where part of this work has
392: been done. This work was 
393: supported in part (I.G.) by DOE Contract
394: W-7-405-ENG-36 and by the  DOE
395: Program in Applied Mathematical Sciences, KJ-01-01.
396: 
397: \begin{references}
398: \bibitem{Recent reviews}  G. Steinmeyer, D.H. Sutter,
399: L. Gallmannn, N. 
400: Matuschek, U. Keller, Science {\bf 286}, 1507 (1999);
401: G.A. Thomas,  A.A.
402: Ackerman, P.R. Prunchal, S.L. Cooper, Physics Today
403: {\bf 53,  9}, 30 (2000); 
404: %;G.A. Thomas, B.I. Shraiman, P.F. Glodis, M. Stephen,
405: %Nature {\bf 404}, 262 (2000).
406: 
407: \bibitem{Multimode}  Notice, however, that in the
408: multi-channel  situation
409: (which will not be discussed here) the opposite to the
410:  standard expectation
411: is also possible : namely the presence of random  inter-modes
412: (channels) scattering
413: can be used here to improve the  total information
414: capacity of a multi-mode
415: optical fiber. See, A.F.  Garito, J. Wang, R. Gao,
416: Science {\bf 281}, 962
417: (1998); H. Stuart,  Science {\bf 289}, 281 (2000), and
418: also discussion of
419: similar ideas  in the field of wireless
420: communications: G. J. Foschini, M.J.
421: Gans,  Wireless Personal Commun. {\bf 6}, 311 (1998);
422: A.L. Mousstakas,  H.U.
423: Baranger, L. Balents, A.M. Sengupta, S.H. Simon,
424: Science {\bf \ 287}, 287
425: (2000).
426: 
427: \bibitem{Gordon-Hauss}  J.N. Elgin, Phys. Lett. {\bf
428: 110A}, 441 (1985); 
429: Phys. Rev. E {\bf 47}, 4331 (1993);J.P. Gordon, H.A.
430: Haus, Opt.  Lett. {\bf %
431: 11}, 665 (1986); G. Falkovich, I. Kolokolov, V.
432: Lebedev,  S. Turitsyn,
433: preprint 04/2000, 
434: http://xyz.lanl.gov/abs/nlin.CD/0004001.
435: 
436: \bibitem{PMD}  C.D. Poole, Opt. Lett. {\bf 13}, 687
437: (1988); {\bf 14},  523
438: (1989); C.D. Poole, J.H. Winters, and J.A. Nagel, Opt.
439: Lett. {\bf 16}, 372
440: (1991); N. Gisin, Opt. Comm. {\bf 86}, 371-373 (1991);
441:  P.K. Wai,
442: C.R.Menyak, H.H. Chen, Opt. Lett. {\bf 16}, 1231
443: (1991).
444: 
445: \bibitem{97A}  G.P. Agrawal, Fiber-optic communication
446: systems, New  York,
447: Wiley, 1997.
448: 
449: \bibitem{96MMN}  L.F. Mollenauer, P.V. Mamyshev, M.J.
450: Neubelt, 
451: %''Method for
452: %facile and accurate measurement of optical fiber dispersion maps'', 
453: Opt. Lett. {\bf 21}, 1724 (1996).
454: 
455: \bibitem{98GM}  J. Gripp, L.F. Mollenauer, 
456: %Enhanced range far OTDR-like dispersion map measurements.
457: Opt. Lett. {\bf 23}, 1603 (1998).
458: 
459: \bibitem{Param}  All parameters are presented here in
460: dimensionless units
461: which transform to real-world fiber units according to
462: the following rules.
463: The envelope of the electric field is in the form
464: $\psi =E/\sqrt{P_{0}}$,
465: where $P_{0}$ is the peak pulse power. The propagation
466: variable is $%
467: z=x(\alpha P_{0}/2)$, where $x$ is distance along the
468: fiber and $\alpha $ is
469: the Kerr nonlinearity coefficient. The Kerr
470: coefficient can be expressed in
471: terms of other fiber parameters, $\alpha =2\pi
472: n_{2}/(\lambda S_{eff})$,
473: where $n_{2}$ is the nonlinear component of fiber
474: refractive index, $\lambda 
475: $ is operating wavelength, and $S_{eff}$ is an
476: effective core area of the
477: fiber. The spatial coordinate is $t=\tau /\tau _{0}$,
478: where $\tau $ is in
479: the reference frame of the group velocity and $\tau
480: _{0}$ is the
481: characteristic pulse width. The dispersion coefficient
482: is $d=2\beta
483: _{2}/(\alpha P_{0}\tau _{0}^{2})$, where $\beta _{2}$
484: is the second order
485: dispersion parameter. The typical parameters for
486: dispersion shifted fiber
487: are: $\lambda =1550$nm , $\tau _{0}=7.01$ps, $x=50$km,
488: $P_{0}=$ $4$mW, $%
489: \beta _{2}$ $=2$ ps$^{2}/$km, $\alpha =10$
490: Wt$^{-1}$km$^{-1}$.
491: 
492: \bibitem{72ZS}  V. E. Zakharov, A.B. Shabat, Sov.
493: Phys. JETP {\bf 34},  62
494: (1972).
495: 
496: \bibitem{80LKC}  C. Lin, H. Kogelnik, L.G. Cohen, Opt.
497: Lett. {\bf 5},  476
498: (1980).
499: 
500: 
501: 
502: %\bibitem{00MMGNMGHV}  L.F. Mollenauer, P.V. Mamyshev,
503: %J. Gripp,  M.J.
504: %Neubelt, N. Mamysheva, L. Gruner-Nielsen, T. Veng, 
505: % Demonstration of massive wavelength-division
506: %multiplexing over
507: % transoceanic distances by use of dispersion-managed
508: %solitons, 
509: %Opt. Lett.,{\bf 25}, 704, (2000).
510: 
511: \bibitem{96GT}  I. Gabitov, S.K. Turitsyn, Opt. Lett.
512: {\bf 21}, 327  (1996);
513: JETP Lett. {\bf 63}, 861 (1996).
514: 
515: \bibitem{96SKDBB}  N. Smith, F. M. Knox, N.J. Doran,
516: K.J. Blow, I.  Bennion, 
517: %''Enhanced power solitons in optical fiber transmission line'',
518: Electron. Lett. {\bf 32}, 54 (1996).
519: 
520: \bibitem{99TSSM}  S.K. Turitsyn, T. Schafer, K.H.
521: Spatschek, V.K. 
522: Mezentsev, 
523: %''Path-averaged chirped optical soliton in dispersion-managed fiber
524: %communication lines'', 
525: Opt. Comm. {\bf 163}, 122 (1999).
526: 
527: \bibitem{99GGDBMGHFG}  D.Le. Guen, et al.,
528: Post Deadline Paper, OFC/IOOC'99, San
529: Diego, CA, USA.
530: 
531: \bibitem{98ACS}  F. Kh. Abdullaev, J.G. Caputo, M.P.
532: Sorensen, in New 
533: trends in Optical Soliton Transmission Systems, A.
534: Hasegawa, Eds.  (Kluwer,
535: Dordrecht, 1998); F.Kh. Abdullaev, J. C. Bronski and
536: G.  Papanicolaou,
537: Physica {\bf D}135, 369 (2000).
538: 
539: \bibitem{00AB}  F. Kh. Abdullaev, B. B. Baizakov, 
540: %'' Disintegration of asoliton in a disperion-managed
541: %optical 
542: % communication line with random parameters'', 
543: Opt. Lett. {\bf 25}, 93 (2000).
544: 
545: \bibitem{00CGMb}  M. Chertkov, I. Gabitov, J. Moeser,
546: Z. Toroczkai, 
547: unpublished.
548: 
549: \bibitem{99ZM}  V.E. Zakharov, S.V. Manakov, JETP
550: Lett. {\bf 70}, 578 
551: (1999).
552: 
553: \bibitem{Movie}   Movies of single realization
554: dynamics in $z$, comparative
555: plot of the dispersion profile with and without
556: compensation, and more
557: figures characterizing statistics of the pulse
558: propagation (in the various
559: cases considered) are available at
560: http:/cnls.lanl.gov/\symbol{126}chertkov/Fiber.
561: 
562: \bibitem{D5}  Notice that for the case with the same
563: pinning period $l=1$
564: but  greater $D=2.5$, a minor, but still observable
565: degradation of pulse 
566: occurred. This is consistent with the statement
567: concerning the  efficiency
568: of the pinning made in the text: the greater $D$, the 
569: lower the critical
570: pinning period.
571: 
572: \bibitem{different}  Real fiber spans, usually each of
573: length $0.5-2$ $z_{NL}
574: $, are not homogeneous, as they are often produced at
575: different  times by
576: different companies.
577: 
578: \bibitem{00MMGNMGV}  After the work was completed we
579: learned about the 
580: paper of L.F.Mollenauer, et al. [ Opt. Lett. {\bf 25},
581: 704 (2000)],  where
582: long haul transmission experiments on fibers
583: built from  spans of
584: different types are described. It was shown there that
585:  periodic
586: compensation of the overall dispersion to the fixed 
587: residual value
588: (achieved via insertion of an extra span) optimizes 
589: propagation of pulses.
590: These results are consistent with, and also  give an
591: experimental support
592: to, the present proposed pinning  method.
593: \end{references}
594: 
595: \end{multicols}
596: \end{document}
597: 
598: