nlin0309009/cgl.tex
1: \documentclass{article}
2: \usepackage[dvips]{graphicx,psfrag}
3: 
4: \begin{document}
5: \title{
6:   Exact Phase Solutions  of
7:   Nonlinear Oscillators on Two-dimensional Lattice
8:   }
9: \author{
10:   Tsunehiro Yokoi, Hiroyasu Yamada, and Kazuhiro Nozaki
11:   }
12: \date{}
13: \maketitle
14: 
15: \begin{abstract}
16:   We present various exact solutions of a discrete
17:   complex Ginzburg-Landau (CGL) equation on a 
18: plane lattice, which describe target patterns 
19: and spiral patterns and derive their stability criteria.
20: We also obtain similar  solutions to a system of 
21: van der Pol's oscillators.
22: \end{abstract}
23: 
24: \section{Introduction}
25: Near the onset of the Hopf bifurcation in continuous media, a complex 
26: amplitude of
27: distributed oscillators is described by a normal form equation called  
28: the complex Ginzburg-Landau (CGL) equation. Some exact solutions of the 1+1 
29: dimensional
30: CGL equation are known  as  hole, solitary-wave and shock solutions 
31: \cite{nozaki-bekki}. In the 2+1 dimensional space, numerical studies 
32: showed existence of interesting solutions describing target patterns 
33: and spiral patterns  \cite{kramer1,kramer2}.
34: However,  explicit
35: and  exact solutions representing target patterns or spiral patterns 
36: have not  been constructed yet.
37: 
38:   Here  we consider a system of  oscillators which undergo the Hopf 
39: bifurcation on a two dimensional lattice instead  of continuous media. 
40: Each oscillator is governed by the normal form equation of the Hopf 
41: bifurcation \cite{gholms} and couples diffusively and dispersively with 
42: its nearest-neighbor  oscillators so that the system becomes a 
43: spatially discretized  version of the 2+1 dimensional CGL (discrete 
44: CGL) equation\cite{mori}.
45: Such a discrete  CGL equation has also been studied on 
46: a network in connection with \cite{strogatz,yamada}. In this paper,
47:   we show that the discrete  CGL equation has exact solutions 
48: describing various
49:   phase patterns consisting of target patterns and spiral patterns. Each 
50: spiral pattern solution needs  a defect in its origin of spiral, where 
51:   oscillators are absent.
52:   This defect may corresponds to the singularity of  the origin in a 
53: spiral solution
54:   in  continuous media. In the other hand, target pattern solutions do 
55: not have any defects.
56:   We derive  the stability condition of the exact solutions to 
57: modulational perturbations
58:   and conduct some numerical simulations to confirm the stability 
59: criterion. Similar phase pattern solutions are constructed to a system 
60: of van der Pol's oscillators.
61: 
62:  %%%%%%%
63: \section{Nonlinear Oscillators on Two-dimensional Lattice }
64: %%%%%%%%
65: Let us consider the following system of normal form oscillators to the 
66: Hopf bifurcation
67: on a two-dimensional lattice.
68: \begin{equation}
69:   \frac{dW_n}{dt} = -(\mu_r +i \mu_i)|W_n|^2 W_n
70:    +(\nu_r+ i \nu_i)W_n
71:    +(\lambda_r +i \lambda_i)\sum_{j \sim  n} ^4 (W_j - W_n) \label{eq:1}
72:   \end{equation}
73: where
74: $W_n $ are complex amplitudes of oscillators and
75: $ \mu _r,\ \mu_i,\ \nu_r,\ \nu_i,\ \lambda_r,\ \lambda _i $
76: are real constants; $ \mu _r$ 
77: and $ \nu _r$ are assumed to be positive (the  subcritical bifurcation) and
78: % $W_n \in {\it C}, \mu _r,\mu _i,\nu_r,\nu_i,\lambda_r ,\lambda _i \in {\it R}$
79:   $\sum^4 _{j \sim n}$ denotes summation over nearest-neighbor 
80: oscillators
81:   at site $n$. Through this summation term, each oscillator has 
82: diffusive ($\lambda _r$) and dispersive  ($\lambda _i$) interactions 
83: with its nearest-neighbors. Here, we call
84:    Eq.(\ref{eq:1})
85:   a discrete CGL equation.
86: 
87: By changing variables
88: $W_n  \rightarrow W_n \exp[-i \nu _i t]$ and $t \rightarrow \mu_r t$,
89: and replacing coefficients as
90: $p_1 = \mu _i /\mu_r $, $ p_2 = \nu_r/\mu_r  $,
91: $p_3 = \lambda_r/ \mu_r$ and $p_4 = \lambda_i/\mu_r$,
92: Eq.(\ref{eq:1}) is transformed into
93: \begin{equation}
94:   \frac{dW_n}{dt} = -(1 +i p_1)|W_n|^2 W_n
95:    +p_2 W_n+(p_3 +i p_4)\sum^4_{j \sim  n}(W_j - W_n) ,\label{eq:2}
96: \end{equation}
97: or
98: \begin{equation}
99:   \frac{dW_n}{dt} = -(1 +i p_1)|W_n|^2 W_n
100:    +(p_2 -4(p_3 +i p_4))W_n+(p_3 +i p_4)\sum^4_{j \sim  n}W_j 
101: ,\label{eq:3}
102: \end{equation}
103: where $p_2>0$.
104: We use Eq.(\ref{eq:2}) or Eq.(\ref{eq:3}) as a basic system in this 
105: paper.
106: 
107: %%%%%%%%%
108: \section{Exact Solutions}
109: %%%%%%%%%%
110: In this section, we derive some exact solutions of the discrete CGL 
111: equation (\ref{eq:3}),
112: which represent a target pattern, a spiral pattern and various mixtures 
113: of those patterns.
114: A crucial step to find exact solutions is to assume
115: \begin{equation}
116:    \sum^4_{j \sim n} W_j =0
117:    \ \mbox{(for all sites}\  n \mbox{)}.\label{eq:4}
118: \end{equation}
119: Then, Eq. (\ref{eq:3}) takes the same equation for all $n$:
120: \begin{equation}
121:    \frac{dW_n}{dt}=-(1 +i p_1)|W_n|^2 W_n +(p_2-4 (p_3 +i p_4)) W_n 
122: .\label{eq:5}
123: \end{equation}
124: We choose an equilibrium  solution of Eq.(\ref {eq:5}).
125: \begin{equation}
126:    W_n =A e^{i(\omega t +\phi(n))} ,\label{eq:6}
127: \end{equation}
128: where $A ,\ \omega $ are real constants satisfying
129: \begin{eqnarray}
130:   A^2 &=& p_2-4p_3 , \\
131:   \omega &=& -p_1 (p_2 -4 p_3)-4 p_4 ,
132: \end{eqnarray}
133: and a functional form of the phase $\phi(n)$ with respect to $n$ is 
134: determined later.
135: Because of $A^2 >0$,  we need
136: \begin{equation}
137: p_2-4p_3>0. \label{cond1}
138: \end{equation}
139: Since $p_2>0$, the condition (\ref {cond1}) is automatically satisfied 
140: for
141: $p_3 \le 0$ (negative diffusion or diffusionless case).
142: The solution (\ref {eq:6}) has the same amplitude $A$ at  all cites but
143: its phase could vary from site to site. The variation of the phase must 
144: be determined
145: so that Eq.(\ref {eq:4}) is satisfied.
146: \begin{equation}
147:    \sum^4_{j \sim n} e^{i \phi(j)}=0
148:    \ \mbox{(for all sites}\  n \mbox{)}.\label{eq:7}
149: \end{equation}
150: We consider a case that there are only two values for $\phi (n)$,
151: that is, $\phi(n)=0$ or $\pi$. At every site, we set phase values of two oscillators  
152:   among four nearest-neighbor oscillators as $\phi(n)=0$, while the 
153: other two phases are chosen to be $\pi$.
154: Then, the condition (\ref {eq:7}) is fulfilled at all sites, that is,
155: \begin {eqnarray}
156:    \sum^4_{j \sim n} e^{i \phi(j)} &=&
157:    e^{i \cdot 0} +e^{i \cdot 0} +e^{i \cdot \pi} +e^{i \cdot 
158: \pi}\nonumber \\
159:   &=&0 \ \mbox{(for all sites}\  n \mbox{)} .\label {exact1}
160: \end{eqnarray}
161: This choice of phase values is quite simple but produces a variety of 
162: phase patterns
163: such as  stripe patterns and target patterns as exact solutions.
164: We can construct not only a single target pattern but also a combined 
165: pattern of
166: some target patterns as shown in Fig.1. Although we  present only a 
167: combined pattern of
168: four target patterns,  combined patterns of more or less target 
169: patterns are also
170: possible. We can construct  various combined patterns of  target 
171: patterns  indefinitely if the system size becomes larger.
172: 
173: When the lattice has a defect, that is, an oscillator is absent at a 
174: site, we should  modify the above discussion  for four oscillators 
175: around a defect.
176: To four oscillators around a defect, the coupling terms in 
177: Eq.(\ref{eq:2}) are changed to
178: \begin{eqnarray}
179:    (p_3 +i p_4) \sum_{j \sim  n} ^3 ( W_j - W_n )
180:    &=& (p_3 +i p_4)( \sum_{j \sim n} ^3 W_j -3 W_n) \nonumber \\
181:    &=& (p_3 +i p_4)\left\{ (W_n + \sum_{j \sim n} ^3 W_j)-4 W_n 
182: \right\},
183: \end{eqnarray}
184:   where $n$ indicates one of the four oscillators and the summation term 
185: in Eq.(\ref{eq:3}) becomes
186: \begin{equation}
187:    (p_3 +i p_4)(W_n + \sum_{j \sim n} ^3 W_j )  .
188: \end{equation}
189: Thus, Eq.(\ref{eq:7})  is modified for each $n$-th oscillator  around a 
190: defect as
191: \begin{equation}
192:      e^{i \phi(n)}+\sum_{j \sim n} ^3 e^{i \phi(j)}=0 .\label{eq:7m}
193: \end{equation}
194: Comparing Eq.(\ref{eq:7m}) to Eq.(\ref{eq:7}), the term 
195: $\exp (i \phi(n))$  in (\ref{eq:7m})
196: appears replacing the term which  comes from an oscillator at a defect 
197: site.
198: Therefore, for the $n$-th oscillator around a defect, the defect 
199: behaves as a mirror,
200: that is, the defect acts as if it is  an oscillator with the same 
201: phase as the $n$-th phase.
202: Owing to this flexible property of a defect, we can construct spiral 
203: patterns with $2^k$ arms around $k \times k$ combined defects. For 
204: example,  we obtain a spiral pattern with two arms around a single 
205: defect ($k=1$) and a spiral pattern with four arms around four combined 
206: defects ($k=2$) as shown in Fig.2 and Fig.4, respectively. Two separated defects produce
207: two spiral patterns with the same or opposite rotation directions as 
208: shown in Fig.3.
209: We can also construct various  patterns in which spirals and targets 
210: are concurrent and show one of the simplest examples  in Fig.2. On a 
211: large enough lattice, we could construct virtually an indefinite number 
212: of such spiral and target patterns.
213: 
214: 
215: \begin{figure}[htbp]
216:   \begin{tabular}{cc}
217:    \begin{minipage}{0.5\hsize}
218:     \begin{center}
219:      \includegraphics[width=3.5cm]{pic1-1.eps}
220: 
221:     \end{center}
222:    \end{minipage}
223:    \begin{minipage}{0.5\hsize}
224:     \begin{center}
225:      \includegraphics[width=3.5cm]{pic1-2.eps}
226:    \end{center}
227:   \end{minipage}
228: \end{tabular}
229: \caption{A target pattern.
230:    Black points denote $\phi (n) =0$, while
231:    gray points denote $\phi (n) =\pi$.}
232: \label{fig:1}
233: \end{figure}
234: 
235: \begin{figure}[htbp]
236:   \begin{minipage}{0.5\hsize}
237:     \begin{center}
238:      \includegraphics[width=3.5cm]{pic2-1.eps}
239: 
240:     \end{center}
241:   \end{minipage}
242:    \begin{minipage}{0.5\hsize}
243:     \begin{center}
244:      \includegraphics[width=3.5cm]{pic2-2.eps}
245: 
246:     \end{center}
247:   \end{minipage}
248:       \caption{A spiral pattern with a defect and
249:       a combined pattern of a spiral and a target.}
250:      \label{fig:2}
251: \end{figure}
252: 
253: \begin{figure}[htbp]
254:   \begin{tabular}{cc}
255:    \begin{minipage}{0.5\hsize}
256:     \begin{center}
257:      \includegraphics[width=3.5cm]{pic3-1.eps}
258:     \end{center}
259:    \end{minipage}
260:    \begin{minipage}{0.5\hsize}
261:     \begin{center}
262:      \includegraphics[width=3.5cm]{pic3-2.eps}
263:     \end{center}
264:    \end{minipage}
265: \end{tabular}
266:      \caption{ Two spiral patterns  with  the opposite and the same
267:      spiral rotations, respectively.}
268:      \label{fig:3}
269:    \end{figure}
270:    
271: \begin{figure}[htbp]
272:   \begin{tabular}{cc}
273:    \begin{minipage}{0.5\hsize}
274:     \begin{center}
275:      \includegraphics[width=3.5cm]{pic4.eps}
276:     \end{center}
277:      \caption{A four-arm spiral pattern  and a two-arm spiral pattern}
278:    \label{fig:3.5}
279:   \end{minipage}
280: \end{tabular}
281: 
282: \end{figure}   
283: 
284: 
285: 
286: 
287: %%%%%%%%%
288: \section{Stability of Solutions}
289: %%%%%%%%%%
290: Let us  investigate a linear stability of the exact solution:
291: \begin{eqnarray}
292: &&  W_n ^0 (t) =
293:    A e^{ i(\omega t +\phi(n))}\ (\phi(n)=0\ \mbox{or}\  \pi), \nonumber\\
294:    &&  \sum^4_{j \sim n} e^{i \phi(j)}=0
295:    \ \mbox{(for all site}\  n \mbox{)}.\nonumber
296: \end{eqnarray}
297: Substituting a perturbed solution $W_n =W_n ^0 +\delta W_n$
298: into Eq.(\ref{eq:3}), we have
299: \begin{eqnarray}
300:   \frac{d}{dt}\delta W_n&=& i \omega \delta W_n-(1+i p_1)A^2 \delta W_n 
301: \nonumber\\
302: &&  -(1+i p_1)A^2 e^{2 i(\omega t+\phi(n))} \delta W_n ^*
303:    +(p_3 +i p_4) \sum^4_{j \sim n}\delta W_j. \label{eq:8}
304: \end{eqnarray}
305: As $\phi (n)=0$ or $\pi$, the  coefficient of the third term in the
306: RHS of Eq.(\ref {eq:8}) becomes
307: $-(1+i p_1)A^2 \exp[i 2 \omega t]$, which does not depend on
308: a site number $n$. Therefore, all coefficients of the linearized 
309: equation
310: (\ref {eq:8}) are independent of $n$, which greatly simplifies  
311:  stability analysis.
312: Changing the variable as
313: $\delta W_n \rightarrow \delta W_n \exp[-i \omega t]$,
314: Eq.(\ref{eq:8}) is transformed into
315: \begin{equation}
316:   \frac{d}{dt}\delta W_n= -(1+i p_1)A^2(\delta W_n + \delta W_n ^*)
317:    +(p_3 +i p_4) \sum^4_{j \sim n}\delta W_j. \label{eq:9}
318: \end{equation}
319: We consider plane-wave perturbations
320: \begin{equation}
321:    \delta W_n=a(t)e^{i \vec{k} \cdot \vec{x}}
322:    =a(t) e^{i (k_x x+k_y y)} ,\label{eq:10}
323: \end{equation}
324: where $\vec{x}=(x,y)$ indicates a lattice point and $\vec{k}=(k_x,k_y)$ 
325: is a wave-number vector.
326: Then, the summation term of Eq.(\ref{eq:9}) reads
327: \begin{eqnarray}
328:   \sum^4_{j \sim n} \delta W_j &=&
329:    a e^{\vec{k} \cdot \vec{x}}
330:    (e^{i k_x}+e^{i k_y}+e^{-i k_x}+e^{-i k_y}) \nonumber \\
331:   &=& a e^{\vec{k} \cdot \vec{x}}(2 \cos k_x + 2 \cos k_y) \nonumber \\
332:   &=&2 \delta W_n K, \nonumber
333: \end{eqnarray}
334:   where $K = \cos k_x +\cos k_y$.
335: Hence, Eq.(\ref{eq:9})  reduces to
336: \begin{eqnarray}
337:   \frac{d}{dt}\delta W_n  &=& -(1+i p_1)A^2(\delta W_n + \delta W_n ^*)
338:    +2 (p_3 +i p_4) K \delta W_n \nonumber \\
339:   &=&  (2 (p_3 +i p_4) K-(1+i p_1)A^2)\delta W_n
340:    -(1+i p_1)A^2 \delta W_n ^* ,
341:    \end{eqnarray}
342: or  in terms of the real and imaginary parts of $\delta W_n$
343: \begin{eqnarray}
344:   \frac{d}{dt} \left(
345:                 \begin{array}{c} \Re [\delta W_n]\\
346:                 \Im [\delta W_n]\end{array}
347:                \right)=2
348:   \left(
349:    \begin{array}{cc} p_3 K-A^2 & - p_4 K  \\
350:     p_4 K -p_1 A^2 & p_3 K\end{array}
351:   \right)
352:   \left(
353:    \begin{array}{c} \Re [\delta W_n]\\
354:    \Im [\delta W_n]\end{array}
355:   \right).\label{eq:11}
356: \end{eqnarray}
357: Eigenvalues of Eq.(\ref {eq:11}) are
358: \begin{equation}
359:   2 p_3 K-A^2 \pm \sqrt{-4  p_4 ^2 K^2
360:    +4 p_1 p_4 A^2 K+A^4}.
361: \end{equation}
362: We denote an eigenvalue as $\alpha_+$ which has a larger real part
363: than the other.
364: \begin{equation}
365:   \alpha_+ =2 p_3 K-A^2 +
366:    \Re \left[p
367:        \sqrt{-4  p_4 ^2 K^2
368:        +4 p_1  p_4 A^2 K+A^4}
369:       \right]. \label{eq:14}
370: \end{equation}
371: If $\alpha_+ \leq 0$ for any $K$ in $|K|\leq 2$,
372: the various phase solutions  are stable to a perturbation (\ref{eq:10}).
373: For stability, $\alpha_+$ must take  local maximum at K=0 since 
374:   $\alpha_+ =0$ at $K=0$
375: (see Fig.\ref{fig:5}, Fig.\ref{fig:6}).
376: The derivative of $\alpha_+$ at $K=0$ is given as
377: \begin{eqnarray}
378: \left.\frac{\partial \alpha_+}{\partial K} \right| _{K=0}
379: &=& 2 \left[
380:   p_3  -p_4 \frac{2 p_4 K- p_1 A^2}{
381:               \sqrt{-4 p_4 ^2 K^2 +4 p_1 p_4 K+A^4 }
382:               }
383:          \right]_{K=0} \nonumber \\
384:   &=& 2 (p_3+ p_1 p_4). \nonumber
385: \end{eqnarray}
386: A condition  $p_3+p_1 p_4=0$ is necessary for stability. 
387: When $p_3+p_1 p_4=0$, we have
388: \begin{eqnarray}
389: \left.\alpha _+ \right| _{K= \frac{A^2}{2 p_3}}
390: &=& \Re \left[
391:   \sqrt{- \left(\frac{p_4}{p_3} \right)^2
392:     + 2 \frac{p_1 p_4}{p_3} +1} \right] A^2\nonumber \\
393:   &=& \Re \left[ \sqrt{- \left(\frac{p_4}{p_3} \right)^2 -1} \right] A^2 \nonumber\\
394:   &=&0,\nonumber
395: \end{eqnarray}
396: which shows that  $K=A^2 /2 p_3 $ is also a vanishing point of $\alpha _+$, where its derivative with respect to $K$ is finite.
397: For stability, we need $$ |A^2 /2 p_3|\geq 2$$ or 
398: \begin{equation}
399: A^2=p_2 -4 p_3  \geq 4 |p_3|,  \label{cond-sta}
400: \end{equation}
401: Since $p_2>0$ is supposed, the condition (\ref {cond-sta})
402: reads
403: $ p_3 <0$  or $p_2>8p_3>0$.
404: 
405: 
406: Thus, we obtain
407: the following conditions for our  solutions to be stable.
408: \begin{eqnarray}
409: &&  p_3+p_1 p_4=0 \label{eq:15} \\ 
410: &&p_3 <0\  \mbox{or}\  p_2\geq 8p_3>0.\nonumber
411:  % A^2=p_2 -4 p_3 & \geq & 4 |p_3| \nonumber
412: \end{eqnarray}
413: 
414: \begin{figure}[htbp]
415:   \begin{tabular}{cc}
416:    \begin{minipage}{0.45\hsize}
417:     \begin{center}
418:      \psfrag{a}[][][0.6]{$ \alpha _+$}
419:      \psfrag{K}[][][0.6]{K}
420:      \includegraphics[width=5.0cm]{pic5-1.eps}
421:      \caption{$\alpha_+ (K)$ vs. $K$ for stable parameters
422:        ($p_1 =-0.5$, $p_2 =4.0$, $p_3=0.5$, $p_4 =1.0$).}
423:      \label{fig:5}
424:     \end{center}
425:   \end{minipage}
426:   \begin{minipage}{0.1\hsize}
427:     \end{minipage}
428:    \begin{minipage}{0.45\hsize}
429:     \begin{center}
430:      \psfrag{a}[][][0.6]{$ \alpha _+$}
431:      \psfrag{K}[][][0.6]{K}
432:      \includegraphics[width=5cm]{pic5.eps}
433:      \caption{$\alpha_+ (K)$ vs. $K$  for unstable parameters
434:      ($p_1 =1.3$, $p_2 =3.0$, $p_3=0.25$, $p_4 =0.8$).}
435:      \label{fig:6}
436:     \end{center}
437:    \end{minipage}
438:   \end{tabular}
439: \end{figure}
440: 
441: 
442: %%%%%%%%%%%%%%%%%%%%%%%
443: \section{Numerical analysis of the system}
444: %%%%%%%%%%
445: First, we we conduct numerical analysis of the system with random 
446: initial perturbations in order to confirm validity of the stability 
447: condition (\ref{eq:15}).
448:   We set a boundary condition which enables
449:   our exact solutions  to persist. It is supposed
450:   that, outside a boundary,  there are oscillators whose phases are
451: different from those just inside  the boundary by $\pi$. This boundary 
452: condition is called a vanishing boundary condition in this paper.
453: For values of parameters satisfying the stability condition 
454: (\ref{eq:15}),
455: both target patterns and spiral patterns are found to be
456: stable even for   random perturbations.
457: For example, in a stable case that  $p_1 =-0.5$, $p_2 =4.0$, $p_3=0.5$ 
458: and $p_4 =1.0$, a single target pattern and the pattern with two 
459: spirals are  stable even for max $20 \%$ random perturbations
460: as shown in Fig.\ref{fig:7}.
461: 
462: \begin{figure}[htbp]
463:    \begin{minipage}{0.5\hsize}
464:      \begin{center}
465:        \includegraphics[width=5.0cm]{pic7-1.eps}
466:      \end{center}
467:    \end{minipage}
468:      \begin{minipage}{0.5\hsize}
469:        \begin{center}
470:          \includegraphics[width=5.0cm]{pic7-2.eps}
471:        \end{center}
472:      \end{minipage}
473:      \caption{Initial 20\% random perturbations
474:        are added to  exact solutions in a
475:          stable case that  $p_1 =-0.5,p_2 =4.0,p_3=0.5,p_4 =1.0$.
476:          Length of an arrow indicates $|W_n |$
477:          and its direction  means $\arg W_n$ .}
478:        \label{fig:7}
479: \end{figure}
480: 
481: \begin{figure}[htbp]
482:    \begin{tabular}{cc}
483:      \begin{minipage}{0.5\hsize}
484:        \begin{center}
485:          \includegraphics[width=5.0cm]{pic8-1.eps}
486:        \end{center}
487:      \end{minipage}
488:      \begin{minipage}{0.5\hsize}
489:        \begin{center}
490:          \psfrag{amp}[][][0.6]{$|W|^2$}
491:          \psfrag{time}[][][0.6]{time}
492:          \includegraphics[width=5.0cm]{pic8-2.eps}
493:        \end{center}
494:      \end{minipage}
495:    \end{tabular}
496:    \caption{ A target pattern of the numerical solution at time $t=50$ for the 
497:      initial condition given in Fig.\ref{fig:7} and $|W_n| ^2$ vs. $t$ for all
498:      $n$, which shows stability of the target pattern.  }
499:   \label{fig:8}
500: \end{figure}
501:  
502: \begin{figure}[htbp]
503:    \begin{tabular}{cc}
504:         \begin{minipage}{0.5\hsize}
505:        \begin{center}
506:          \includegraphics[width=5.0cm]{pic9-1.eps}
507:        \end{center}
508:      \end{minipage}
509:      \begin{minipage}{0.5\hsize}
510:        \begin{center}
511:          \psfrag{amp}[][][0.6]{$|W|^2$}
512:          \psfrag{time}[][][0.6]{time}
513:          \includegraphics[width=5.0cm]{pic9-2.eps}
514:        \end{center}
515:      \end{minipage}
516:    \end{tabular}
517:    \caption{A combined spiral pattern of the numerical solution at time $t=50$ for the 
518:      initial condition given in Fig.\ref{fig:7} and $|W_n| ^2$ vs. $t$ for all
519:      $n$, which shows stability of the combined spiral pattern.
520:    }
521:    \label{fig:9}
522: \end{figure}
523: 
524: 
525: Next, we show that a single target pattern can be reached
526: from some initial conditions far from the exact solution.
527: For example, setting initial values as  $W_n=1/2,\ \phi(n)=\pi/4$ for 
528: all $n$ in the case $p_1 =0.5$, $p_2 =3.0$, $p_3 =0.25$ and  $p_4 
529: =-0.5$, we eventually have a target pattern, where $|W_n|=2$, shown in  
530: Fig.\ref{fig:13}. We also find
531:   other initial conditions far from the exact solution that lead to
532: the single target pattern  but  those do not give enough informations 
533: for a basin structure of the target patterns.
534: \begin{figure}[htbp]
535:    \begin{tabular}{cc}
536:      \begin{minipage}{0.5\hsize}
537:        \begin{center}
538:          \includegraphics[width=5.0cm]{pic10-1.eps}
539:        \end{center}
540:      \end{minipage}
541:      \begin{minipage}{0.5\hsize}
542:        \begin{center}
543:          \includegraphics[width=5.0cm]{pic10-2.eps}
544:        \end{center}
545:      \end{minipage}
546:    \end{tabular}
547:    \caption{An initial condition  $|W|=0.5$
548:      and  phase$= \pi /4$ for all oscillators and
549:      a target pattern at time $t=150$ developed from
550:      the initial condition.
551:        }
552:    \label{fig:13}
553: \end{figure}
554: 
555: %%%%%%%%%
556: \section{Van der Pol's Oscillators}
557: %%%%%%%%%%
558: Similar exact target and/or spiral solutions can be also constructed for
559: other diffusively and dispersively coupled systems of oscillators
560: which undergo the  subcritical Hopf bifurcation.
561: For example, we consider a system of van der Pol's oscillators
562: which are coupled with each other diffusively and dispersively on a
563: plane lattice.
564: \begin{eqnarray}
565:   \frac{d}{dt}x_n  &=&y_n+ax_n- \frac{a}{3}x_n^3+
566:   D_x \sum^4_{j \sim n}(x_j-x_n) \nonumber \\
567:   \frac{d}{dt}y_n  &=&-x_n+
568:   D_y \sum^4_{j \sim n}(x_j-x_n), \label{van-der}
569:    \end{eqnarray}
570: where $a$ is a positive constant and $D_x, D_y$ are diffusive
571: and dispersive constants respectively.
572: Let us seek  phase solutions of Eq.(\ref {van-der})
573: similar to those discussed earlier.
574: Assuming
575: \begin{equation}
576:   \sum^4_{j \sim n}x_j =0, \label{van-rel}
577:   \end{equation}
578: Eqs.(\ref {van-der}) are reduced to a single  van der Pol's equation
579: \begin{equation}
580: \frac{d^2}{dt^2}X_n+(X_n^2-1)\frac{d}{dt}X_n+X_n=0, \label{svan-der}
581:   \end{equation}
582: where $\sqrt{1+4D_y}t \rightarrow  t $,   $x_n=(1-4D_x/a)^{1/2} X_n$
583: and
584: \begin{equation}
585: (1+4D_y)>0, \quad (1-4D_x/a)>0 \label{van-cond}
586: \end{equation}
587: are assumed.
588: If the van der Pol's attractor of Eq.(\ref {svan-der}) is written as
589: $x_n(t)=A(t)$, its opposite-phase function $x_n(t)=-A(t)$ is also
590: the attractor of Eq.(\ref {svan-der}).
591: By choosing  $x_n(t)$ as $A(t)$ or $-A(t)$ for all $n$ so that
592:   $x_n(t)$ are satisfy the relation  (\ref {van-rel}),
593:   we can construct various exact solutions whose phase patterns
594:   are exactly the same as those in the discrete CGL equation.
595:   However, the stability analysis will be much more difficult than
596:   the previous case.
597: %%%%%%%%%
598: \section{Conclusions}
599: %%%%%%%%%%
600: We present various exact solutions of a discrete CGL equation on a 
601: plane lattice, which describe target patterns 
602: and spiral patterns and derive their stability criteria.
603: Exact solutions obtained here represent not only a single target pattern and  a 
604: single spiral pattern but also patterns consisting  
605: of targets and spirals. In the center of a spiral pattern, at least one 
606: defect is necessary and  a number of spiral's arms depends on a 
607: number of defects. For example, a defect produces a two-arm spiral 
608: and four defects in the center generate a four-arm spiral and so on.
609: For a system of the van der Pol's oscillators, we can also 
610: construct various exact solutions whose phase patterns
611:   are exactly the same as those in the discrete CGL equation.
612: pp
613: %%%%%%%%%%%%%%%%%%%%%%%%%%
614: \section* {Acknowledgement}
615: %%%%%%%%%%%%%%%%%%%%%%%%%%
616: One of the authors (K.N.) has been, in part, supported by a 
617: Grant-in-Aid for
618: Scientific Research
619: (C) 13640402 from the Japan Society for the Promotion of Science.
620: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
621: 
622: 
623: 
624: %%%%%%%%%%%%%%%%%%%%%%%%%%%
625: \begin{thebibliography}{99}
626: %%%%%%%%%%%%%%%%%%%%%%%%%%%
627: \bibitem{nozaki-bekki} K.Nozaki and N.Bekki, J. Phys. Soc. Jpn.
628: {\bf 53} 
629: 53 (1984), 1581.
630: \bibitem{kramer1}
631:   I.S. Aranson, L. Aranson, L. Kramer, and A. Weber,
632:   Phys. Rev. A 46, R2992 (1992)
633:   
634: \bibitem{kramer2}
635:   Igor S. Aranson and Lorenz Kramer, Rev. Mod. Phys. 74, 99 (2002) 
636: 
637: \bibitem{gholms} J. Guckenheimer and P. Holmes,
638: {\it Nonlinear Oscillations,Dynamical Systems and Bifurcations of 
639: Vector Fields}, Springer-Verlag (1983).
640: 
641: \bibitem{mori}
642:    S. Mori: M. Thesis, Faculty of Science,
643:    Ochanomizu University, Otsuka, Bunkyo-ku, Tokyo, 1996.
644: 
645: \bibitem{yamada} 
646:   H. Yamada and T. Nakagaki,
647:  talk at Traffic and Granular Flow '01, October 15-71, 2001, Nagoya University, Japan. 
648: 
649:    
650: \bibitem{strogatz}
651:   S. H. Strogatz, Nature 410 (2001), 268. 
652: 
653: 
654: 
655: 
656: 
657:   
658: 
659: %%%%%%%%%%%%%%%%%%%%%%
660: \end{thebibliography}
661: %%%%%%%%%%%%%%%%%%%%%%
662: 
663: \end{document}
664: 
665: