1:
2: \documentclass[final]{siamltex}
3: \usepackage{amssymb}
4:
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: \usepackage{epsfig}
7: \usepackage{graphicx}
8:
9: %TCIDATA{OutputFilter=Latex.dll}
10: %TCIDATA{LastRevised=Fri Oct 13 10:08:41 2006}
11: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
12:
13: \newcommand{\pe}{\psi}
14: \def\d{\delta}
15: \def\ds{\displaystyle}
16: \def\e{{\epsilon}}
17: \def\eb{\bar{\eta}}
18: \def\enorm#1{\|#1\|_2}
19: \def\Fp{F^\prime}
20: \def\fishpack{{FISHPACK}}
21: \def\fortran{{FORTRAN}}
22: \def\gmres{{GMRES}}
23: \def\gmresm{{\rm GMRES($m$)}}
24: \def\Kc{{\cal K}}
25: \def\norm#1{\|#1\|}
26: \def\wb{{\bar w}}
27: \def\zb{{\bar z}}
28: \def\bfE{\mbox{\boldmath$E$}}
29: \def\bfG{\mbox{\boldmath$G$}}
30: %\input{tcilatex}
31:
32: \begin{document}
33:
34: \title{Mathematical modeling of pattern formation in sub- and supperdiffusive reaction-diffusion systems}
35: \author{
36: Vasyl Gafiychuk\thanks{Institute of Applied Problem of Mechanics and Mathematics, National Academy
37: of Sciences of Ukraine, Naukova Street 3 B, Lviv, Ukraine 79053(\texttt{viva@iapmm.lviv.ua})}\and
38: Bohdan Datsko\thanks{Institute of Applied Problem of Mechanics and Mathematics, National Academy
39: of Sciences of Ukraine, Naukova Street 3 B, Lviv, Ukraine 79053(\texttt{viva@iapmm.lviv.ua})}\and
40: Vitaliy Meleshko\thanks{Institute of Applied Problem of Mechanics and Mathematics, National Academy
41: of Sciences of Ukraine, Naukova Street 3 B, Lviv, Ukraine 79053(\texttt{viva@iapmm.lviv.ua})}
42: }
43: \maketitle
44:
45: \begin{abstract}
46: This paper is concerned with analysis of coupled fractional
47: reaction-diffusion equations. It provides analytical comparison
48: for the fractional and regular reaction-diffusion systems. As an
49: example, the reaction-diffusion model with cubic nonlinearity and
50: Brusselator model are considered. The detailed linear stability
51: analysis of the system with cubic nonlinearity is provided. It is
52: shown that by combining the fractional derivatives index with the
53: ratio of characteristic times, it is possible to find the marginal
54: value of the index where the oscillatory instability arises.
55: Computer simulation and analytical methods are used to analyze
56: possible solutions for a linearized system. A computer simulation
57: of the corresponding nonlinear fractional ordinary differential
58: equations is presented. It is shown that the increase of the
59: fractional derivative index leads to the periodic solutions which
60: become stochastic at the index approaching the value of 2. It is
61: established by computer simulation that there exists a set of
62: stable spatio-temporal structures of the one-dimensional system
63: under the Neumann and periodic boundary condition. The
64: characteristic features of these solutions consist in the
65: transformation of the steady state dissipative structures to
66: homogeneous oscillations or space temporary structures at a
67: certain value of fractional index.
68: \end{abstract}
69:
70: \begin{keywords}
71: reaction-diffusion system, fractional differential equations,
72: oscillations, dissipative structures, pattern formation,
73: spatio-temporal structures
74: \end{keywords}
75:
76: \begin{AMS}
77: 37N30, 65P40, 37N25, 35K50, 35K45, 34A34, 34C28, 65P30
78: \end{AMS}
79:
80: \pagestyle{myheadings} \thispagestyle{plain}
81: \markboth{V.
82: GAFIYCHUK, B. DATSKO and V. MELESHKO}{V.
83: GAFIYCHUK, B. DATSKO and V. MELESHKO}
84:
85: \section{Introduction}
86:
87: Reaction-diffusion (RD) systems are inherent in many branches of
88: physics, chemistry, biology, ecology etc. The review of the theory
89: and applications of reaction-diffusion systems one can find in
90: books and numerous articles (See, for
91: example\cite{pr,ch,m90,KO,dk89,dk03,lg,gl,sbg,mo02,g99}). The
92: popularity of the RD system is driven by the underlying richness
93: of the nonlinear phenomena, which include stationary and
94: spatio-temporary dissipative pattern formation, oscillations,
95: different types of chemical waves, excitability, bistability etc.
96: The mechanism of the formation of such type of nonlinear phenomena
97: and the conditions of their emergence have been extensively
98: studied during the last couple decades. Although the mathematical
99: theory of such type of phenomena has not been developed yet due to
100: the essential nonlinearity of these systems, from the viewpoint of
101: the applied and experimental mathematics, the pattern of possible
102: phenomena in RD system is more or less understandable.
103:
104: In the recent years, there has been a great deal of interest in
105: fractional reaction-diffusion (FRD) systems \cite{hw,hw1,he02,
106: add,add1,gd,GDI05,ar1,zw,vb} which from one side exhibit
107: self-organization phenomena and from the other side introduce a
108: new parameter to these systems, which is a fractional derivative
109: index, and it gives a greater degree of freedom for diversity of
110: self-organization phenomena. At the same time, the process of
111: analyzing such FRD systems is much complicated from the analytical
112: and numerical point of view.
113:
114: In this article, we consider two coupled reaction-diffusion systems:
115:
116: The first one is the classical system
117:
118: \begin{equation}
119: \tau _{1}\frac{\partial n_{1}(x,t)}{\partial t}=l^{2}\nabla
120: ^{2}n_{1}(x,t)-W(n_{1},n_{2},\mathcal{A}), \label{1}
121: \end{equation}
122: \begin{equation}
123: \tau _{2}\frac{\partial n_{2}(x,t)}{\partial t}=L^{2}\nabla
124: ^{2}n_{2}(x,t)-Q(n_{1},n_{2},\mathcal{A}), \label{2}
125: \end{equation}
126: where $x,t\in $ $\Bbb{R}$; $\nabla ^{2}=\frac{\partial ^{2}}{\partial x^{2}}%
127: ;n_{1}(x,t)$, $n_{2}(x,t)\in $ $\Bbb{R}$ -- two variables, $W$ ,$Q\in $ $%
128: \Bbb{R}$ are the nonlinear sources of the system modeling their production
129: rates, $\tau _{1},\tau _{2},l,L,\in $ $\Bbb{R}$ -- characteristic times and
130: lengths of the system, $\mathcal{A}\in $ $\Bbb{R}$ -- is an external
131: parameter
132:
133: And the other model is the fractional RD system
134: \begin{equation}
135: \tau _{1}\frac{\partial ^{\alpha }n_{1}(x,t)}{\partial t^{\alpha }}%
136: =l^{2}\nabla ^{2}n_{1}(x,t)-W(n_{1},n_{2},\mathcal{A}), \label{3}
137: \end{equation}
138: \begin{equation}
139: \tau _{2}\frac{\partial ^{\alpha }n_{2}(x,t)}{\partial t^{\alpha }}%
140: =L^{2}\nabla ^{2}n_{2}(x,t)-Q(n_{1},n_{2},\mathcal{A}) \label{4}
141: \end{equation}
142: with the same parameters and fractional derivatives $\frac{\partial ^{\alpha
143: }n(x,t)}{\partial t^{\alpha }}$ on the left hand side of equations (\ref{3}%
144: ),(\ref{4}) instead of standard time derivatives, which are the Caputo
145: fractional derivatives in time of the order $0<\alpha <2$ and are
146: represented as \cite{skm,po}
147:
148: \[
149: \frac{\partial ^{\alpha }}{\partial t^{\alpha }}n(t):=\frac{1}{\Gamma
150: (m-\alpha )}\int\limits_{0}^{t}\frac{n^{(m)}(\tau )}{(t-\tau )^{\alpha +1-m}}%
151: d\tau ,\mbox{}\;m-1<\alpha <m,m\in N.
152: \]
153:
154: The article is devoted to the second problem and the first one we need for
155: comparing the obtained results with classical one.
156:
157: Equations (\ref{3}),(\ref{4}) at $\alpha =1$ correspond to standard RD
158: system described by equations (\ref{1}),(\ref{2}). At $\alpha <1,$ they
159: describe anomalous sub-diffusion and at $\alpha >1$~ - anomalous
160: superdiffusion .
161:
162: In this paper, we always assume that the following conditions are fulfilled
163: on the boundaries 0;$l_{x}$:
164:
165: (i) Neumann: \
166: \begin{equation}
167: dn_{i}/dx|_{x=0}=dn_{i}/dx|_{x=l_{x}}=0 \label{bc1}
168: \end{equation}
169:
170: (ii) Periodic:
171: \begin{equation}
172: n_{i}(t,0)=n_{i}(t,l_{x}),\;dn_{i}/dx|_{x=0}=dn_{i}/dx|_{x=l_{x}},
173: \label{bc2}
174: \end{equation}
175: where $i=1,2.$
176:
177: \section{Linear stability analysis}
178:
179: Stability of the steady-state constant solutions of the system (\ref{3}),(%
180: \ref{4}) correspond to homogeneous equilibrium states
181: \begin{equation}
182: W(n_{1},n_{2},\mathcal{A})=0,\;Q(n_{1},n_{2},\mathcal{A})=0
183: \label{e}
184: \end{equation}
185: can be analyzed by linearization of the system nearby this solution. In this
186: case the system (\ref{3})(\ref{4}) can be transformed to linear system
187: \begin{equation}
188: \frac{\partial ^{\alpha }\mathbf{u}(x,t)}{\partial t^{\alpha }}=\widehat{A}%
189: \mathbf{u}(x,t), \label{lin}
190: \end{equation}
191: where $\mathbf{u}(x,t)=\left(
192: \begin{array}{c}
193: \triangle n_{1}(x,t) \\
194: \triangle n_{2}(x,t)
195: \end{array}
196: \right) ,$ $\widehat{A}=\left(
197: \begin{array}{cc}
198: (l^{2}\nabla ^{2}-a_{11})/\tau _{1} & -a_{12}/\tau _{1} \\
199: -a_{21}/\tau _{2} & (L^{2}\nabla ^{2}-a_{22})\tau _{2}
200: \end{array}
201: \right) ,$ $a_{11}=W_{n_{1}}^{\prime },$ $a_{12}=W_{n_{2}}^{\prime },$ $%
202: a_{21}=Q_{n_{1}}^{\prime },$ $a_{22}=Q_{n_{2}}^{\prime }$ (all derivatives
203: are taken at homogeneous equilibrium states (condition (\ref{e})). By
204: substituting the solution $\mathbf{u}(x,t)=\left(
205: \begin{array}{c}
206: \triangle n_{1}(t) \\
207: \triangle n_{2}(t)
208: \end{array}
209: \right) \cos kx,k=\frac{\pi }{l_{x}}j,j=1,2...$ into FRD system (\ref{3}),(%
210: \ref{4}) we can get the system of linear ordinary differential equations (%
211: \ref{lin}) with the matrix $A$ determined by the operator $\widehat{A}$, the
212: stability conditions of which are given by eigenvalues of this matrix.
213:
214: Let us analyze the stability of the solution (\ref{e}) of the linear system
215: with integer derivatives and find the conditions of this instability (See,
216: for example:\cite{ch,m90,KO,dk89}). We repeat this process in order to
217: compare the results obtained with the results of the fractional RD system
218: considered in this article. In this case, by searching for the solution of
219: the linear system in the form $\mathbf{u}(x,t)=\left(
220: \begin{array}{c}
221: \triangle n_{1} \\
222: \triangle n_{2}
223: \end{array}
224: \right) \exp (\lambda t)\cos kx$, we get a homogeneous system of linear
225: algebraic equations for constants $\triangle n_{1}, \triangle n_{2} $. The
226: solubility of this system leads to the characteristic equation
227:
228: \begin{equation}
229: \det |\lambda I-A|=0, \label{aa}
230: \end{equation}
231: where
232: \begin{equation}
233: A=-\left(
234: \begin{array}{cc}
235: (l^{2}k^{2}+a_{11})/\tau _{1} & a_{12}/\tau _{1} \\
236: a_{21}/\tau _{2} & (L^{2}k^{2}+a_{22})/\tau _{2}
237: \end{array}
238: \right), \label{a}
239: \end{equation}
240: $I$ \ is the identity matrix. As a result, the characteristic equation takes
241: on a form of a simple quadratic equation $\lambda ^{2}-trA\lambda +\det A=0.$%
242: \
243:
244: The linear boundary value problem for RD system (\ref{1}),(\ref{2}) is
245: unstable according to inhomogeneous wave vectors $k\neq 0$ if ($trA<0,\det
246: A<0$)
247: \begin{equation}
248: a_{11}<-[(l/L)^{2}a_{22}+2(l/L)(a_{22}a_{11}-a_{12}a_{21})^{1/2}],
249: \label{tu}
250: \end{equation}
251: \[
252: \tau _{2}(a_{11}+k^{2}l^{2})+\tau
253: _{1}(a_{22}+k^{2}L^{2})>0,\,a_{22}a_{11}-a_{12}a_{21}>0
254: \]
255: (Turing Bifurcation) and according to homogenous ($k=0$) fluctuations (Hopf
256: Bifurcation) if ($trA>0,\det A>0$)
257: \begin{equation}
258: \tau _{2}a_{11}+\tau _{1}a_{22}<0,a_{22}a_{11}-a_{12}a_{21}>0. \label{h}
259: \end{equation}
260:
261: For analyzing the equations (\ref{3}),(\ref{4}) let us also consider a linear
262: system obtained near the homogeneous state (\ref{e}). As a result, the
263: simple linear transformation can convert this linear system to a diagonal
264: form
265:
266: \begin{equation}
267: \frac{d^{\alpha }\mathbf{\eta (}t\mathbf{)}}{dt^{\alpha
268: }}=C\mathbf{\eta }(t) \label{f},
269: \end{equation}
270: where $C$ is a diagonal matrix of $A$: $\ C=P^{-1}AP=\left(
271: \begin{array}{cc}
272: \lambda _{1} & 0 \\
273: 0 & \lambda _{2}
274: \end{array}
275: \right) ,$ eigenvalues $\lambda _{1,2}$ are determined by the same
276: characteristic equation (\ref{aa}) with matrix (\ref{a}), $\lambda _{1,2}=%
277: \frac{1}{2}(trA\pm \sqrt{tr^{2}A-4\det A}),\ \mathbf{\eta }(t)=P^{-1}\left(
278: \begin{array}{c}
279: \triangle n_{1}(t) \\
280: \triangle n_{2}(t)
281: \end{array}
282: \right) ,$ $P$ \ is the matrix of eigenvectors of matrix $A$.
283:
284: \begin{figure}[tbp]
285: \begin{center}
286: \begin{tabular}{cc}
287: \includegraphics[width=0.35\textwidth]{fig21a.eps} & %
288: \includegraphics[width=0.47\textwidth]{fig21b.eps} \\
289: (a) & (b)
290: \end{tabular}
291: \end{center}
292: \caption{Schematic view of the marginal curve describing fixed points for
293: two-dimensional vector field -- (a) the marginal value of $\protect\alpha$ --
294: (b).}
295: \label{rys1}
296: \end{figure}
297:
298: In this case, the solution of the vector equation (\ref{f}) is given by
299: Mittag-Leffler functions\bigskip\ \cite{skm,po,os06,a,tz}
300:
301: \begin{equation}
302: \triangle n_{i}(t)=\sum\limits_{k=0}^{\infty }\frac{(\lambda _{i}t^{\alpha
303: })^{k}}{\Gamma (k\alpha +1)}\triangle n_{i}(0)=E_{\alpha }(\lambda
304: _{i}t^{\alpha })\triangle n_{i}(0), \quad i=1,2. \label{sol}
305: \end{equation}
306: Using the result obtained in the papers \cite{GDI05,mi}, we can
307: conclude that if for any of the roots
308: \begin{equation}
309: |arg(\lambda _{i})|<\alpha \pi /2 \label{100}
310: \end{equation}
311: the solution has an increasing function component then the
312: system is asymptotically unstable.
313:
314: Analyzing the roots of the characteristic equations, we can see that at $%
315: 4\det A-tr^{2}A>0$ eignevalues are complex and can be represented as
316: \[
317: \lambda _{1,2}=\frac{1}{2}(trA\pm i\sqrt{4\det A-tr^{2}A})\equiv
318: \lambda _{Re}\pm i\lambda _{Im}.
319: \]
320: The roots $\lambda _{1,2}$ are complex inside the parabola (Figure \ref{rys1}%
321: (a)) and the fixed points are the spiral source ($trA>0$)\ \ or spiral sinks (%
322: $trA<0$). The plot of the marginal value $\alpha :\alpha =\alpha _{0}=$ $%
323: \frac{2}{\pi }|arg(\lambda _{i})|$ which follows from the conditions (\ref
324: {100}) is given by\ the formula
325:
326: \begin{equation}
327: \alpha _{0}=\left\{
328: \begin{array}{cc}
329: \frac{2}{\pi }\arctan \sqrt{4\det A/tr^{2}A-1}, & trA\geq 0, \\
330: 2-\frac{2}{\pi }\arctan \sqrt{4\det A/tr^{2}A-1}, & trA\leq 0
331: \end{array}
332: \right. \label{al}
333: \end{equation}
334: and is presented in the Figure \ref{rys1}(a) below the parabola in the
335: coordinate system $(trA,\alpha ).$
336:
337: Let us analyze the system solution with the help of the Figure
338: \ref{rys1}(a). Consider the parameters which keep the system
339: inside the parabola. It is a well-known fact, that at $\alpha =1$\
340: the domain on the righthand side of the parabola ($trA>0$) is
341: unstable with the existing limit circle, while the
342: domain on the left hand side ($trA<0$) is stable. By crossing the axis $%
343: trA=0 $ the Hopf bifurcation conditions become true. In the general case of
344: \ $\alpha :0<\alpha < 2$\ \, for every point inside the parabola there
345: exists a marginal value of $\alpha _{0}$ where the system changes its
346: stability. The value of $\alpha $ is a certain bifurcation parameter which
347: switches the stable and unstable state of the system. At lower $\alpha :$ $%
348: \alpha <\alpha _{0}=$ $\frac{2}{\pi }|arg(\lambda _{i})|$ the system has
349: oscillatory modes but they are stable. Increasing the value of $\alpha
350: >\alpha _{0}=$ $\frac{2}{\pi }|arg(\lambda _{i})|$ leads to instability. As
351: a result, the domain below the curve\ $\alpha _{0}$, as a function of $trA$%
352: \, is stable and the domain above the curve is unstable.
353:
354: The plot of the roots, describing the mechanism of the system
355: instability, can be understood from the Figure \ref{rys1}(b) where
356: the case $\alpha_{0}>1$ is described. In fact, having complex
357: number $\lambda_{i}$ with ${\it Re} \lambda_{i}<0$\ at $\alpha \rightarrow 2$
358: it is always possible to satisfy the condition
359: $|arg(\lambda_{i})|<\alpha \pi /2$, and the system becomes
360: unstable according to homogeneous oscillations (Figure
361: \ref{rys1}(b)). The smaller is the value of $trA$, the easier it is
362: to fulfill the instability conditions.
363:
364: In contrast to this case, a complex values of $\lambda _{i},$
365: with$\ {\it Re} \lambda _{i}>0$ lead to the system instability for regular system with $%
366: \alpha=1 .$ However fractional derivatives with $\alpha <1$ will
367: stabilize the system if $\alpha <\alpha _{0}=$ $\frac{2}{\pi
368: }|arg(\lambda _{i})|.$ This
369: makes it possible to conclude that fractional derivative equations with $%
370: \alpha <1$\ are more stable that their integer twinges.
371:
372: \section{Solutions of the coupled fractional ordinary differential equations
373: (FODEs)}
374:
375: Our particular interest here is the analysis of the specific
376: non-linear system of FRD equations. We consider two very
377: well-known examples. The first one is the RD system with cubical
378: nonlinearity \cite{m90,KO,lg} which probably is the simplest one
379: used in RD systems modeling
380: \begin{equation}
381: W(n_{1},n_{2})=-n_{1}+n_{1}^{3}/3+n_{2},\quad Q=n_{2}-\beta n_{1}-\mathcal{A}%
382: . \label{nlin2}
383: \end{equation}
384: The second example is known as Brusselator model \cite{pr} and it describes
385: certain chemical reaction-diffusion processes with a pair of variables whose
386: concentrations are controlled by nonlinearities
387: \begin{equation}
388: W(n_{1},n_{2})=-\mathcal{A}+(\beta +1)n_{1}-n_{1}^{2}n_{2},\quad
389: Q(n_{1},n_{2})=- \beta n_{1}+n_{1}^{2}n_{2}, \label{nlin1}
390: \end{equation}
391:
392: Let us first consider the coupled fractional ordinary differential
393: equations (FODEs) with nonlinearities (\ref{nlin2}) and analyze
394: the stability conditions for such systems. The plot of isoclines
395: for the system (\ref {nlin2}) is represented on Figure \ref{rys2}(a).
396: In this case, for$\ $
397: homogeneous solution, which can be determined from the system of equations $%
398: W=Q=0$, is the solution of cubic algebraic equation\ \ ($\beta -1)\overline{n%
399: }_{1}+\overline{n}_{1}^{3}/3+\mathcal{A}=0.$ Simple calculation makes it
400: possible to write the expressions required for analysis $A=-\left(
401: \begin{array}{cc}
402: (-1+\overline{n}_{1}^{2})/\tau _{1} & \quad 1/\tau _{1} \\
403: -\beta /\tau _{2} & \quad 1/\tau _{2}
404: \end{array}
405: \right) ,$ $trA=(1-\overline{n}_{1}^{2})/\tau _{1}-1/\tau _{2},$ $\quad \det
406: A= ( (\beta -1)+\overline{n}_{1}^{2})/\tau _{1}\tau _{2} .$
407:
408: \begin{figure}[tbp]
409: \begin{center}
410: \begin{tabular}{cc}
411: \includegraphics[width=0.9\textwidth]{fig31.eps} &
412: \end{tabular}
413: \end{center}
414: \caption{Null isoclines -- (a), Instability domains (shaded regions) in coordinates $(\overline{n}%
415: _{1},\protect\tau _{1}/\protect\tau _{2})$ (Dependence of
416: $\protect\tau _{1}/\protect\tau _{2}$ on $\overline{n}_{1}$) -- (b), Dependence of $%
417: \protect\alpha _{0}$ on $\overline{n}_{1}^{2}$ -- (c), Dependence of $\protect%
418: \tau _{1}/\protect\tau _{2}$ on $\protect\alpha _{0}$ (the domains
419: in coordinates $(\overline{n}_{1}$, $\protect\tau
420: _{1}/\protect\tau _{2}$) correspond to domains represented on
421: figure (b) -- (d). } \label{rys2}
422: \end{figure}
423:
424: It is easy to see that if the value of $\tau _{1}/\tau _{2}$, in
425: certain cases, is smaller than $1$, the instability conditions (
426: $trA>0$)\ lead to Hopf bifurcation for regular system ($\alpha
427: =1$) \cite{pr,ch,m90,KO,dk89}. In this case, the plot of the
428: domain, where instability exists, is shown on the Figure
429: \ref{rys2}(b).
430:
431: % plural .... solution ...they
432: The linear analysis of the system for $\alpha =1$ shows that, if
433: $\tau _{1}/\tau _{2}>1$, the solution corresponds to the
434: intersections of two isoclines, and it is stable. The marginal
435: curve, separating stability and instability domains, is given on
436: Figure \ref{rys2}(b). The smaller is the ratio of $\tau _{1}/\tau
437: _{2}$, the wider is the instability region.
438: Formally, at $\tau _{1}/\tau _{2}\rightarrow 0$, the instability region in $%
439: \overline{n}_{1}$\ coinsides with the interval ($-1,1$) where the
440: null isocline $W(n_{1},n_{2})=0$ \ has its increasing part. The
441: maximum value of the curve $\tau _{1}/\tau _{2}(n_1)$ corresponds
442: to the value $\tau _{1}/\tau _{2}=1$ where the system is neutrally
443: stable. These results are very widely known in the theory of
444: nonlinear dynamical systems \cite{pr,ch,m90,KO,dk89}.
445:
446: In the FODEs the conditions of the instability change (\ref{100}), and we
447: have to analyze the real and the imaginary part of the existing complex
448: eigenvalues, especially the equation: $4\det A-tr^{2}A=4((\beta -1)+%
449: \overline{n}_{1}^{2})/\tau _{1}\tau _{2}-\left( (1-\overline{n}%
450: _{1}^{2})/\tau _{1}-1/\tau _{2}\right) ^{2}>0.$ In fact, with the
451: complex eigenvalues, it is possible to find out the corresponding
452: value of $\alpha $ where the condition (\ref{100}) is true. We
453: show that this interval is not correlated with the increasing part
454: of the null isocline of the system. Indeed, omitting simple
455: calculation, we can write an equation for marginal
456: values of $\overline{n}_{1}$: $\overline{n}_{1}^{4}-2(1+%
457: \frac{\tau _{1}}{\tau _{2}})\overline{n}_{1}^{2}+\frac{\tau _{1}^{2}}{\tau
458: _{2}^{2}}-2\frac{\tau _{1}}{\tau _{2}}(2\beta -1)+1=0,$ and expression $%
459: \overline{n}_{1}^{2}=1+\frac{\tau _{1}}{\tau _{2}}\pm 2\sqrt{\beta \frac{%
460: \tau _{1}}{\tau _{2}}}$ estimates the maximum and minimum values of $%
461: \overline{n}_{1}$ where the system can be unstable at certain value of $%
462: \alpha =\alpha _{0}$. For example, examine the domain of the FODEs where
463: the eigenvalues are complex for fixed value $\tau _{1}/\tau _{2},$ for
464: example, consider $\tau _{1}=\tau _{2}=1$ and $\beta =2.$ In this case, $%
465: trA=-\overline{n}_{1}^{2},\det A=1+\overline{n}_{1}^{2},4\det A-tr^{2}A=4+4%
466: \overline{n}_{1}^{2}-\overline{n}_{1}^{4}>0,$ which immediately leads to the
467: region of existing of complex roots $-\sqrt{2+2\sqrt{2}}\leq \overline{n}%
468: _{1}\leq \sqrt{2+2\sqrt{2}}$ and we can conclude that the instability region, due to the
469: fractional order of the derivatives, can be much wider
470: than the same region for ($\alpha =1)$. The dependance of the value of $%
471: \alpha $ on the value $\overline{n}_{1}^{2}$ (\ref{al})\ is given
472: on Figure \ref{rys2}(c).
473:
474: In this case the plot is obtained at $\tau _{1}/\tau _{2}=1$ and
475: that is why the marginal instability curve is determined for
476: $\alpha >1$ (for $\alpha <1$ the system is stable). On this figure
477: the domain below the curve corresponds to stability \ and above it
478: to instability conditions.
479:
480: Similar analysis can be provided for $trA>0$ where $(\tau
481: _{1}/\tau _{2})$ is smaller than unity. In this case, the
482: instability conditions are also not correlated with the increasing
483: part of the null isocline $W_{1}(n_{1},n_{2}).$ In this case the
484: plot of $\alpha $ will be start not from 1 but from certain value
485: smaller than unity. However, it is much better for understanding
486: to get marginal curve $\alpha _{0}$ as a function of $\tau
487: _{1}/\tau _{2}$.
488:
489:
490: Let us analyze the dependence of $\alpha _{0}$ on the parameter
491: $\tau _{1}/\tau _{2}$ where the system changes its stability. As
492: it is easy to see from Figure
493: \ref{rys2}(a),(b)\ that the easiest way to reach instability domain is realized at $%
494: \mathcal{A}=0$\ when two isoclines are intersected themselves at
495: the point $(0,0) $ and this corresponds to maximum of the curve on
496: Figure \ref{rys2}(b)\ . Let us consider the dependence of this\
497: marginal value of $\alpha _{0}$ on the parameter $\tau _{1}/\tau
498: _{2}$ at $\mathcal{A}=0$ and determine the plot of this maximum \
499: as a functions of $\alpha _{0}$ \ on $\tau _{1}/\tau _{2}.$\ Such
500: curve for the given model is represented in the Figure
501: \ref{rys2}(d). This curve obtained from (\ref{al}) corresponds to
502: the dependence of $\alpha $ on $trA$ (Figure \ref{rys1}(a)). Below
503: this curve the system is unstable, and above it - it is stable. We
504: may therefore focus our attention on the
505: general form of this curve. At sufficiently small value $\tau _{1}/\tau _{2}$%
506: \ oscillatory instability is valid even for small $\alpha <1$. In contrast,
507: at $\alpha >1$, the instability conditions could have place even for those
508: cases when $\tau _{1}/\tau _{2}$ is sufficiently large. This means that
509: fractional differential equations, by corresponding combination of the
510: parameter $\tau _{1}/\tau _{2}$, can be stable or unstable practically in
511: all the region $1<\alpha <2.$
512:
513:
514: It should be noted that, even if the eigenvalues are not complex
515: ($\lambda _{Im}=0$), the systems with fractional derivatives can
516: poses oscillatory damping oscillations. Such situation takes
517: place when $4\det A-tr^{2}A<0,$ $trA<0,$ $\det A>0$ and two
518: eigenvalues are real and less than zero. In this case, at
519: $1<\alpha <2$\ steady state solutions of the system are stable and
520: any perturbations are damping. Such system was considered, for
521: example, in the article \cite{zse05},\ where an analytical
522: solution for fractional oscillator is obtained. Namely with this
523: case we start analyzing possible solutions.
524:
525: Several examples of linear FDOEs were solved analytically by Adomnian
526: decomposition method, as well as numerically. The obtained solutions were
527: compared with the analytical solutions obtained by Mittag-Leffler functions (%
528: \ref{sol}) and the results of the work \cite{zse05}.
529:
530: Three different solutions are plotted on the same Figure
531: \ref{rys3}. The results of the computer simulation of linear
532: ordinary differential equations for stable system - a) and
533: unstable system - b) of one variable $n_{1}$ are given on the time
534: domain interval $[0,30]$. We can see that in the stable domain
535: $\alpha <\alpha _{0}$ the oscillations are damping, and at $\alpha
536: >\alpha _{0}$ they grow exponentially. So, the ordinary
537: differential equations (space derivatives are equal to zero) of
538: the system have two different modes. The solution is either
539: asymptotically stable or unstable.
540:
541: It is a statement, that FODEs are at least as stable as their integer order
542: counterparts \cite{po}. At the same time, we have established here that the
543: dynamics of the FODEs can be much more complicated than that of the
544: equations with integer order \cite{El96,mi,wz}. Our task here is to clarify
545: these statements by finding out not only the conditions of the bifurcation
546: but also the real time dynamics of FODEs.
547:
548: The developed technique of Adomian decomposition \cite{a94,bbi04,jd06,db04}
549: is a powerful method for solving the system of ordinary differential
550: equations. The effectiveness of this method was demonstrated for solving
551: linear fractional differential equations, and we used this method here to
552: test the computer program and to compare these analytical results with
553: Mittag-Leffler functions.
554:
555: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
556: First, we consider the case when the system is always asymptotically stable
557: at $\alpha <\alpha _{0}=2$, for example $\alpha=1.7$ (Figure \ref{rys3}(a)).
558: Three solutions obtained by these different methods practically do not differ from
559: each other and they show oscillatory damping time behavior (the model described in \cite{zse05}).
560: Analytical results obtained by Adomian decomposition is given by the formula
561: \begin{equation}
562: n_{1}=\sum_{k=0}^{\infty}(-1)^{k}\frac{t^{k\alpha}}{\Gamma (k\alpha +1)},
563: \end{equation}
564: which is plot on the Figure \ref{rys3}(a) by empty circled curve
565: (we truncate the series at k=60). At small time interval (for
566: example $t<7$) this series can be represented by next expansion
567: \begin{eqnarray*}
568: n_{1} &=& 1-0.6473808267\cdot t^{1.7}+0.9865725648\cdot 10^{-1}\cdot
569: t^{3.4}-0.7019911214\cdot 10^{-2}\cdot t^{5.1}
570: \\&&+0.296127716\cdot 10^{-3}\cdot t^{6.8}-0.838275933\cdot 10^{-5}\cdot
571: t^{8.5}+ 0.171848173\cdot 10^{-6}\cdot t^{10.2}
572: \\&&-0.2686528893\cdot 10^{-8}\cdot t^{11.9}+ 0.3324725330\cdot 10^{-10}\cdot t^{13.6}
573: \\&&-0.3350625811\cdot 10^{-12}\cdot t^{15.3}+0.2811457252\cdot 10^{-14}\cdot t^{17}.
574: \end{eqnarray*}
575: Despite the considered system, in this particular case, cannot be
576: unstable (Figure \ref{rys3}(a)) because the values of $\lambda $
577: are real and negative $|arg(\lambda _{1})|=\pi >\alpha \pi /2$, in
578: the limit at $\alpha =2$ we have a regular linear oscillator
579: \cite{zse05}, analytical solution of which we get from Adomian
580: decomposition in the form of power series. The terms of this
581: series completely coincides with the expansion of linear
582: oscillator solution $cos(x)$.
583:
584:
585: It should be noted that all damped oscillations solutions of the
586: linear oscillator of fractional order have similar plots.
587:
588: Quite different is the dynamics of FODEs for $\alpha >\alpha _{0}$ when
589: oscillatory mode becomes unstable and leads to increasing oscillations at $%
590: \alpha >\alpha _{0}$ (Figure \ref{rys3}(b)). In this case, the
591: analytical solution obtained by Adomian decomposition for
592: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
593: $\alpha =1.3$ looks like
594: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
595: \begin{eqnarray*}
596: n_{1} &=&0.67-0.1716397314\cdot t^{1.3}-0.05663459553\cdot
597: t^{2.6}+0.008906834532\cdot t^{3.9} \\
598: &&+0.00006991997714\cdot t^{5.2}-0.00004692163249\cdot
599: t^{6.5}+0.1292316598\cdot 10^{-5}\cdot t^{7.8} \\
600: &&+0.5298004324\cdot 10^{-7}\cdot t^{9.1}-0.2781417477\cdot 10^{-8}\cdot t^{10.4}
601: \\
602: &&+0.3895105489\cdot 10^{-11}\cdot t^{11.7}+0.1811701048\cdot 10^{-11}\cdot
603: t^{13}+...
604: \end{eqnarray*}
605:
606: The result of taking into account 50 terms in Adomian
607: decomposition expansion makes it possible to represent the
608: solution in the time interval till $t=30$ is presented on the
609: Figure \ref{rys3}(b). As a matter of fact, such representation has
610: rather theoretical sense because our system is essentially
611: nonlinear and does not allow this type of solutions. For the
612: initial stage dynamics or for the linear system, Adomian
613: decomposition method is very effective. The application of Adomian
614: decomposition to the nonlinear FDOEs is not successful. Taking
615: into account 10 terms makes it possible to find out the solution
616: on time interval $t=6-7$ and at higher values of $t$ the
617: discrepancy between the numerical solution and the one, obtained
618: by Adomian decomposition, increases rapidly.
619: \begin{figure}[tbp]
620: \begin{center}
621: \begin{tabular}{cc}
622: \includegraphics[width=0.49\textwidth]{fig32a.eps} & %
623: \includegraphics[width=0.5\textwidth]{fig32b.eps} \\
624: (a) & (b)
625: \end{tabular}
626: \end{center}
627: \caption{The damped oscillator solution for $\frac{\partial^\alpha
628: u}{\partial t^\alpha}=-u \quad at \quad \protect\alpha=1.7, \quad
629: u(0)=1$ -- (a), and the increasing time domain oscillations of the
630: linearized system (\ref{3}),(\ref{4}) for $n_1$ and
631: $\protect\alpha =1.3, \mathcal{A}=-0.1, \beta=1, \tau _{1}=\tau
632: _{2}=1, n_1(0)=(-3\mathcal{A})^{\frac{1}{3}},
633: n_2(0)=(-3\mathcal{A})^{\frac{1}{3}}+\mathcal{A}$ -- (b). Small
634: filled circled line -- Mittag-Leffler solution, empty circled
635: line -- Adomnian decomposition solution, solid gray line --
636: numerical solution } \label{rys3}
637: \end{figure}
638:
639: To demonstrate the nontrivial properties of FDOEs, here we consider the
640: nonlinear dynamics of the above mentioned nonlinear fractional differential
641: equations.
642:
643: \begin{figure}[tbp]
644: \begin{center}
645: \begin{tabular}{cc}
646: \includegraphics[width=0.35\textwidth]{fig33a.eps} & %
647: \includegraphics[width=0.35\textwidth]{fig33b.eps} \\
648: (a) & (b)
649: \end{tabular}
650: \begin{tabular}{cc}
651: \includegraphics[width=0.35\textwidth]{fig33c.eps} & %
652: \includegraphics[width=0.35\textwidth]{fig33d.eps} \\
653: (c) & (d)
654: \end{tabular}
655: \begin{tabular}{cc}
656: \includegraphics[width=0.35\textwidth]{fig33e.eps} & %
657: \includegraphics[width=0.35\textwidth]{fig33f.eps} \\
658: (e) & (f)
659: \end{tabular}
660: \end{center}
661: \caption{Two dimensional phase portrait (a)-(e) and time domain oscillations
662: corresponding to plot (e) - (f) of the system (\ref{3} ),(\ref{4}) with
663: nonlinearities (\ref{nlin2}) for $\mathcal{A}=-0.1, \protect\beta=1, \protect%
664: \tau _{1}=\protect\tau _{2}=1, l=L=0$. \quad (a) -- $\protect\alpha$=1.3,
665: (b) -- $\protect\alpha $=1.7, (c) -- $\protect\alpha$=1.8, (d) -- $\protect%
666: \alpha$=1.90, (e) -- $\protect\alpha$=2.0, \quad time domain
667: oscillations \quad (f) -- $\protect\alpha$=2.00} \label{rys4}
668: \end{figure}
669:
670: \begin{figure}[tbp]
671: \begin{center}
672: \begin{tabular}{cc}
673: \includegraphics[width=0.35\textwidth]{fig34a.eps} & %
674: \includegraphics[width=0.35\textwidth]{fig34b.eps} \\
675: (a) & (b)
676: \end{tabular}
677: \begin{tabular}{cc}
678: \includegraphics[width=0.35\textwidth]{fig34c.eps} & %
679: \includegraphics[width=0.35\textwidth]{fig34d.eps} \\
680: (c) & (d)
681: \end{tabular}
682: \end{center}
683: \caption{Two dimensional phase portrait (a)-(c) and time domain oscillations
684: corresponding to plot (c) - (d) of the system (\ref{3} ),(\ref{4}) with
685: nonlinearities (\ref{nlin1}) for $\mathcal{A}=2, \protect\beta=2, \protect%
686: \tau _{1}=\protect\tau _{2}=1, l=L=0$. \quad (a) --
687: $\protect\alpha$=1.5, (b) -- $\protect\alpha $=1.6, (c) --
688: $\protect\alpha$=1.7, \quad time domain oscillations \quad (f) --
689: $\protect\alpha$=1.7} \label{rys5}
690:
691: \end{figure}
692:
693: For $\alpha >\alpha _{0}$ small perturbation of the steady state
694: solution, due to the memory inherent in fractional derivatives
695: survive in the process of evolution and grow in amplitude while
696: nonlinear terms of the system (\ref {nlin2}) restrict the value of
697: these oscillations. In this case, time dependence of the variables
698: corresponds to the oscillatory solution. (The phase portrait and
699: isoclines are presented on Figure \ref{rys4}(a)-(e)).
700:
701: Analyzing the phase trajectory of the FODEs, we can see that the
702: amplitude of the oscillations increases with increasing $\alpha.$
703: At $\alpha $ approaching 2, the oscillations become more
704: complicated and at $\alpha =2 $ they look more quasichaotic. The
705: time dependence of this solution is represented in Figure
706: \ref{rys4}(f).
707:
708: Brusselator system with nonlinearities (\ref{nlin1}) has quite similar
709: behavior. Calculating, for example, at $\mathcal{A}=2, \beta=2, \tau
710: _{1}=\tau _{2}=1$ the value of $\alpha_{0}$ we find that $\alpha_{0}=1.54$.
711:
712: The phase portrait and isoclines for (\ref{nlin1}) are presented
713: in Figure \ref{rys5}
714: (a)-(c). At $\alpha \lesssim 1.5$, we obtain steady state solution and at $%
715: \alpha \gtrsim 1.5$ - the steady state oscillation. The increase
716: of the value $\alpha $ leads to the complication of the phase
717: paths, and the two-dimensional phase portrait looks much more
718: complicated (Figure \ref{rys5}(b)-(c)). The attractor of the system of
719: the two coupled nonlinear differential equations gets the features
720: of strange attractor and at $\alpha \rightarrow 2 $ it corresponds
721: to the attractor of the fourth order differential equations
722: determined by the nonlinearities (\ref{nlin1}).
723:
724: \section{Computer simulation of pattern formation}
725:
726: This section contains a discussion of the results of the numerical
727: study of the initial value problem of the system
728: (\ref{3})(\ref{4}). The system with corresponding initial and
729: boundary conditions was integrated numerically using the explicit
730: and implicit schemes with respect to time and centered difference
731: approximation for spatial derivatives. The fractional derivatives
732: were approximated using two different schemes on the basis
733: Riemann-Liouville definition : $L1$-scheme for $0 \leq \alpha <1$,
734: $L2$-scheme for $1 \leq \alpha < 2$ (see below and \cite{oldh}),
735: as well as the scheme on the basis of Grunwald-Letnikov definition
736: for $0 < \alpha < 1$ and $1 < \alpha < 2$ \cite{po} . In other
737: words, for the system of $n$ fractional RD equations
738: \begin{equation} \tau_j \frac{^C\partial^{\alpha_j}
739: u_j(x,t)}{\partial t^{\alpha_j}}=d_j\frac{\partial^2
740: u_j(x,t)}{\partial x^2}+f_j(u_1,...,u_n), \qquad j=\overline{1,n},
741: \end{equation}
742: where $\tau_j, d_j, f_j $ -- certain parameters and nonlinearities of
743: the RD system correspondingly, we used the next numerical schemes:
744: \newline
745: \bigskip
746: {\bf L1-scheme}
747: $$
748: \delta_j u_{j,i}^{k}-\frac{d_j}{(\Delta
749: x)^2}(u_{j,i-1}^{k}-2u_{j,i}^{k}+u_{j,i+1}^{k})-
750: f_j(u_{1,i}^{k},...,u_{n,i}^{k})=
751: $$
752: $$
753: =-\delta_j \left(u_{j,i}^{0}w_k^{(\alpha_j)}+
754: \sum\limits_{l=1}^{k-1}u_{j,i}^{l}\beta_{k-l+1}^{(\alpha_j)}\right)+\tau_j\frac{(k
755: \Delta t)^{-\alpha_j}}{\Gamma(1-\alpha_j)}u_{j,i}^{0}
756: $$
757: $$
758: \delta_j=\frac{\tau_j (\Delta t)^{-\alpha_j}}{\Gamma(2-\alpha_j)},
759: \quad
760: w_k^{(\alpha_j)}=\frac{1-\alpha_j}{k^{\alpha_j}}-k^{1-\alpha_j}-(k-1)^{1-\alpha_j},
761: $$
762: $$
763: \beta_{s}^{(\alpha_j)}=s^{1-\alpha_j}-2(s-1)^{1-\alpha_j}+(s-2)^{1-\alpha_j},
764: \quad s=\overline{2,k};
765: $$
766: \bigskip
767: {\bf L2-scheme}
768: $$
769: \delta_j u_{j,i}^{k+1}- \frac{d_j}{\Delta
770: x^2}(u_{j,i-1}^{k+1}-2u_{j,i}^{k+1}+u_{j,i+1}^{k+1})-
771: f_j(u_{1,i}^{k+1},...,u_{n,i}^{k+1})=
772: $$
773: $$
774: =-\delta_j [ u_{j,i}^{0}w_{0,k}^{\alpha_j} +
775: u_{j,i}^{1}w_{1,k}^{\alpha_j}+\sum\limits_{l=2}^{k-1}u_{j,i}^{l}\beta_{k-l+2}^{(\alpha_j)}
776: + u_{j,i}^{k}(2^{2-\alpha_j}-3) ] + \tau_j \sum \limits_{p=0}^{m}
777: \frac{(k \Delta t)^{p-\alpha_j}} {\Gamma(p-\alpha_j+1)}
778: \frac{\partial^p}{\partial t^p} u_{j,i}^{0},
779: $$
780: $$
781: \delta_j=\frac{\tau_j (\Delta t)^{-\alpha_j}}{\Gamma(3-\alpha_j)},
782: \quad
783: w_{0,k}^{\alpha_j}=\frac{(1-\alpha_j)(2-\alpha_j)}{k^{\alpha_j}} -
784: \frac{(2-\alpha_j)}{k^{\alpha_j-1}} +
785: k^{2-\alpha_j}-(k-1)^{2-\alpha_j},
786: $$
787: $$
788: w_{1,k}^{\alpha_j}=\frac{(2-\alpha_j)}{k^{\alpha_j-1}}-2k^{2-\alpha_j}+3(k-1)^{2-\alpha_j}-(k-2)^{2-\alpha_j},
789: $$
790: $$
791: \beta_{s}^{(\alpha_j)}=s^{2-\alpha_j}-3(s-1)^{2-\alpha_j}+3(s-2)^{2-\alpha_j}-(s-3)^{2-\alpha_j},
792: \quad s=\overline{3,k}
793: $$
794: \bigskip
795: \newline and {\bf G-L scheme}
796: $$
797: u_{j,i}^{k}-\frac{d_j (\Delta t)^{\alpha_j}}{\tau_j (\Delta x)^2}
798: \left(u_{j,i-1}^{k}-2u_{j,i}^{k}+u_{j,i+1}^{k}
799: \right)-\frac{(\Delta
800: t)^{\alpha_j}}{\tau_j}f_j(u_{1,i}^{k},...,u_{n,i}^{k})=
801: $$
802: $$
803: =(\Delta t)^{\alpha_j}\sum \limits_{p=0}^{m}\frac{(k \Delta
804: t)^{p-\alpha_j}}{\Gamma(p-\alpha_j+1)}\frac{\partial^p}{\partial
805: t^p} u_{j,i}^{0}-\sum \limits_{l=1}^{k}
806: c_{l}^{(\alpha_j)}u_{j,i}^{k-l},
807: $$
808: $$
809: c_{0}^{(\alpha_j)}=1, \qquad
810: c_{l}^{(\alpha_j)}=c_{l-1}^{(\alpha_j)}\left(1-\frac{1+\alpha_j}{l}\right),
811: \qquad l=1,2,...
812: $$
813: where $u_{j,i}^{k} \equiv u_j(x_i,t_k) \equiv u_j(i \Delta x, k
814: \Delta t), \quad m=[\alpha]$.
815:
816: The applied numerical schemes are implicit, and for each time layer they are
817: presented as the system of algebraic equations solved by Newton-Raphson
818: technique. Such approach makes it possible to get the system of equations
819: with band Jacobian for each node and to use the sweep method for the
820: solution of linear algebraic equations. Calculating the values of the
821: spatial derivatives and corresponding nonlinear terms on the previous layer,
822: we obtained explicit schemes for integration. Despite the fact that these
823: algorithms are quite simple, they are very sensitive and require small steps
824: of integration, and they often do not allow to find numerical results. In
825: contrast, the implicit schemes, in certain sense, are similar to the
826: implicit Euler's method, and they have shown very good behavior at the
827: modeling of fractional reaction-diffusion systems for different step size of
828: integration, as well as for nonlinear function and the power function of
829: fractional index. Moreover, by modeling according to this algorithm system (%
830: \ref{1}),(\ref{2}), we have observed that these results fully match the
831: results obtained prior.
832:
833: It should be noted that the definition of the fractional derivative in
834: Grunwald-Letnikow form is equivalent to the one in Riemann-Liouville method,
835: but for numerical calculations it is much more flexible.
836:
837: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
838:
839: We have considered here the kinetics of formation of dissipative
840: structures for different values of $\alpha $. These results are
841: presented on Figures \ref{rys6} and \ref{rys7}.
842:
843: \begin{figure}[tbp]
844: \begin{center}
845: \begin{tabular}{cc}
846: \includegraphics[width=0.5%
847: \textwidth]{fig41a.eps} & %
848: \includegraphics[width=0.5%
849: \textwidth]{fig41d.eps} \\
850: (a) & (d)
851: \end{tabular}
852: \begin{tabular}{cc}
853: \includegraphics[width=0.5%
854: \textwidth]{fig41b} & %
855: \includegraphics[width=0.5%
856: \textwidth]{fig41e.eps} \\
857: (b) & (e)
858: \end{tabular}
859: \begin{tabular}{cc}
860: \includegraphics[width=0.5%
861: \textwidth]{fig41c.eps} & %
862: \includegraphics[width=0.5%
863: \textwidth]{fig41f.eps} \\
864: (c) & (f)
865: \end{tabular}
866: \end{center}
867: \caption{Numerical solution of the fractional reaction-diffusion equations (%
868: \ref{3} ),(\ref{4}) with nonlinearities (\ref{nlin2}).Dynamics of variable $%
869: n_1$ on the time interval (0,50) for $l_x=8, \mathcal{A}=-0.1, \protect\beta%
870: =1, \protect\tau _{1}= \protect\tau _{2}=1, \quad l^{2}=0.05, L^{2}=1$; (a) --
871: $\protect\alpha$=0.1, (b) -- $\protect\alpha$=0.5, (c) -- $\protect\alpha$=0.99,
872: (d) -- $\protect\alpha$=1.5, (e) -- $\protect\alpha$= 1.6, (f) -- $\protect\alpha$%
873: =1.8} \label{rys6}
874:
875: \end{figure}
876:
877: The simulations were carried out for a one-dimensional system on an
878: equidistant grid with spatial step $h$ changing from = 0.1 to 0.01 and time
879: step $\Delta t$ changing from 0.001 to 0.1. We used imposed Neuman (\ref
880: {bc1})\ or periodic boundary conditions (\ref{bc2}). As the initial
881: condition, we used the uniform state which was superposed with a small
882: spatially inhomogeneous perturbation.
883:
884: \begin{figure}[tbp]
885: \begin{center}
886: \begin{tabular}{cc}
887: \includegraphics[width=0.5%
888: \textwidth]{fig42a.eps} & %
889: \includegraphics[width=0.5%
890: \textwidth]{fig42d.eps} \\
891: (a) & (d)
892: \end{tabular}
893: \begin{tabular}{cc}
894: \includegraphics[width=0.5%
895: \textwidth]{fig42b.eps} & %
896: \includegraphics[width=0.5%
897: \textwidth]{fig42e.eps} \\
898: (b) & (e)
899: \end{tabular}
900: \begin{tabular}{cc}
901: \includegraphics[width=0.5%
902: \textwidth]{fig42c.eps} & %
903: \includegraphics[width=0.5%
904: \textwidth]{fig42f.eps} \\
905: (c) & (f)
906: \end{tabular}
907: \end{center}
908: \caption{Numerical solution of the fractional reaction-diffusion equations (%
909: \ref{3} ),(\ref{4}) with nonlinearities (\ref{nlin1}).Dynamics of variable $%
910: n_1$ on the time interval (0,20) for $l_x=10, \mathcal{A}=2, \protect\beta%
911: =2, \protect\tau _{1}= \protect\tau _{2}=1,\quad l^{2}=0.1, L^{2}=10 $; (a) --
912: $\protect\alpha$=0.1, (b) -- $\protect\alpha$=0.5, (c) -- $\protect\alpha$=0.99,
913: (d) -- $\protect\alpha$=1.5, (e) -- $\protect\alpha$= 1.6, (f) -- $\protect\alpha$%
914: =1.8} \label{rys7}
915:
916: \end{figure}
917:
918: The systems have rich dynamics, including steady state dissipative
919: structures, homogeneous and nonhomogeneous oscillations, and spatiotemporal
920: patterns. In this paper, we focus mainly on the study of general properties
921: of the solutions depending on the value of $\alpha$.
922:
923: As discussed in Section 2, there are two different regions in
924: parameter $\mathcal{A}$, where the system can be stable or unstable. In the
925: case of $\alpha =1$ the steady state solutions in the form of
926: nonhomogeneous dissipative structures are inherent to unstable
927: region $\overline{n}_{1}\in (-1,1)$. Figures \ref{rys6}(a)-(c) show
928: the steady state dissipative structure formation and Figures
929: \ref{rys6} (d)-(f) present the spatio-temporal evolution of
930: dissipative structures, which eventually leads to homogeneous
931: oscillations.
932:
933: On the Figure \ref{rys6}(a)-(d), the value $\alpha $ increases from \ $0.1$
934: to $1.5 $ and on this whole interval the structures are in steady
935: state. This is due to the case $\alpha <\alpha _{0}$, the
936: oscillatory perturbations are damping, and we can see that small
937: oscillations are at the transition period $n_{1}.$ With increasing
938: $\alpha $, the steady state structures change to the
939: spatio-temporary behavior (Figure \ref{rys6}(e)-(f)).
940:
941: The emergence of homogeneous oscillations, which destroy pattern
942: formation (Figure \ref{rys6}(e),(f))\ \, has deep physical meaning. The
943: matter is that the stationary dissipative structures consist of
944: smooth and sharp regions of variable $n_{1}$, and the smooth shape
945: of $n_{2}$. The linear system analysis shows that the homogeneous
946: distribution of the variables is
947: unstable according to oscillatory perturbations inside the wide interval of $%
948: \overline{n}_{1},$ which is much wider then interval ($-1,1$). At the same
949: time, smooth distributions at the maximum and minimum values of $n_{1}$ are $%
950: \pm \sqrt{3}$ correspondingly. In the first approximation, these smooth
951: regions of the dissipative structures resemble homogeneous ones and are
952: located inside the instability regions. As a result, the unstable
953: fluctuations lead to homogeneous oscillations, and the dissipative
954: structures destroy themselves. We can conclude that oscillatory modes in
955: such type FODEs have a much wider attraction region than the corresponding
956: region of the dissipative structures.
957:
958: For a wide range of the parameters ${\alpha }$, the numerical
959: solutions of the Brusselator problem show similar behavior (Figure
960: \ref{rys7}(a)-(d)). The stationary solutions emerge practically in
961: the same way. At small ${\alpha }$, we see aperiodic formation of
962: the structures, and approaching ${\alpha }_{0}$, the
963: damping oscillations of the dissipative structures arise. At ${\alpha }%
964: _{0}=1.7$ certain non-stationary structures arise (Figure
965: \ref{rys7}(e)). In this case, the dissipative structures look
966: quite similar to those we obtained for regular system
967: \cite{gk91,gd98}. The increase of $\alpha $ leads to a larger
968: amplitude of pulsation. All these patterns are robust with respect
969: to small initial perturbations. The further increase of $\alpha $
970: leads to spatially temporary chaos (Figure \ref{rys7}(f)).
971:
972: In the contrast to previous case, such nonhomogeneous behavior is stable and
973: does not lead to homogeneous oscillations. The matter is that in Brusselator
974: model, the dissipative structures are much grater in amplitude and do not
975: have smooth distribution at the top.
976:
977: It should be noted that the pulsation phenomena of the dissipative
978: structures is closely related to the oscillation solutions of the
979: ODE (Figures \ref{rys4}, \ref{rys5}). Moreover, the fractional
980: derivative of the first variable has the most impact on the
981: oscillations emergence. It can be obtained by performing a
982: simulation where the first variable is a fractional derivative and
983: the second one is an integer. It should be emphasized that the
984: distribution of $n_{2}$, within the solution, only shows a small
985: deviation from the stationary value (that is why this variable is
986: not represented in the figures).
987:
988: \section{Conclusion}
989:
990: In this article we developed a linear theory of instability of reaction
991: diffusion system with fractional derivatives. The introduced new parameter
992: -- marginal value ${\alpha }_{0}$ plays the role of bifurcation parameter.
993: If the fractional derivative index $\alpha$ is smaller than ${\alpha }_{0}$,
994: the system of FODEs is stable and has oscillatory damping solutions. At $%
995: \alpha >{\alpha }_{0}$, the FODEs becomes unstable, and we obtain
996: oscillatory or even more complex - quasi chaotic solutions. In addition, the
997: stable and unstable domains of the system were investigated.
998:
999: By the computer simulation of the fractional reaction-diffusion systems we
1000: provided evidence that pattern formation in the fractional case, at $\alpha $
1001: less than a certain value, is practicably the same as in the regular case
1002: scenario $\alpha =1$. At $\alpha>{\alpha }_{0}$, the kinetics of formation
1003: becomes oscillatory. At $\alpha ={\alpha }_{0}$, the oscillatory mode arises
1004: and can lead to homogeneous or nonhomogeneous oscillations. In the last case
1005: scenario, depending on the parameters of the medium, we can see a rich
1006: variety of spatiotemporal behavior.
1007:
1008: \begin{thebibliography}{99}
1009: \bibitem{pr} \textsc{G. Nicolis, I. Prigogine}. Self-organization in
1010: non-equilibrium systems. Wiley, New York. 1977.
1011:
1012: \bibitem{ch} \textsc{M. C. Cross and P. S. Hohenberg.} \emph{Pattern
1013: formation outside of equilibrium}, Rev. Modern Phys., 65 (1993), pp.
1014: 851--1112.
1015:
1016: \bibitem{m90} \textsc{A. S. Mikhailov.} \emph{Foundations of Synergetics},
1017: Springer-Verlag, Berlin, 1990.
1018:
1019: \bibitem{KO} \textsc{B.S. Kerner, V.V. Osipov.} \emph{Autosolitons} Kluwer,
1020: Dordrecht, 1994.
1021:
1022: \bibitem{dk89} \textsc{J. D. Dockery and J. P. Keener} \emph{Diffusive
1023: effects on dispersion in excitable media}, SIAM J. Appl. Math., 49 (1989),
1024: pp. 539--566.
1025:
1026: \bibitem{dk03} \textsc{A. Doelman and T. J. Kaper.} \emph{Semistrong pulse
1027: interactions in a class of coupled reaction-diffusion equations}
1028: SIAM J. Applied dynamical systems Vol. 2, No. 1, (2003), pp.
1029: 53--96.
1030:
1031: \bibitem{lg} \textsc{A.Lubashevskii, V.V.Gafiychuk.} \emph{The projection
1032: dynamics of highly dissipative system}. Phys. Rev. E. vol.50, No.1,
1033: (1994), pp.171--181.
1034:
1035: \bibitem{gl} \textsc{V.V. Gafiychuk, I.A. Lubashevskii.} \emph{Variational
1036: representation of the projection dynamics and random motion of highly
1037: dissipative systems.} J. Math. Phys. v.36. \#10, (1995), pp. 5735-5752.
1038:
1039: \bibitem{sbg} \textsc{P.Schutz, M.Bode, V.V.Gafiychuk.} \emph{Transition
1040: from stationary to travelling localized patterns in two-dimensional
1041: reaction-diffusion system. }Phys. Rev. E. v.52., N4, (1995), pp. 4465-4473.
1042:
1043: \bibitem{mo02} \textsc{C. B. Muratov and V. V. Osipov.} \emph{Stability of
1044: the static spike autosolitons in the Gray-scott model} Siam J. Appl. Math.
1045: Vol. 62, no. 5, (2002), pp. 1463--1487.
1046:
1047: \bibitem{g99} \textsc{M. Golubitsky, D. Luss and S.H. Strogatz}. \emph{Pattern Formation in
1048: Continuous and Coupled Systems}, IMA Volumes in Mathematics and
1049: its Applications 115 , Springer, New York, (1999).
1050:
1051: \bibitem{hw} \textsc{B.I. Henry, T.A.M. Langlands and S.L. Wearne } \emph{%
1052: Turing pattern formation in fractional activator-inhibitor
1053: systems}. Phys. Rev. E., 72, \# 026101, (2005).
1054:
1055: \bibitem{hw1} \textsc{B.I. Henry, S.L. Wearne.} \emph{Fractional
1056: reaction-diffusion.} Physica A 276, (2000), pp. 448--455.
1057:
1058: \bibitem{he02} \textsc{B.I. Henry and S.L. Wearne.} \emph{Existence of
1059: turing instabilities in a two-species fractional
1060: reaction-diffusion system}. Siam J. Appl. Math. Vol. 62, No. 3,
1061: (2002), pp. 870--887.
1062:
1063: \bibitem{add} \textsc{M. O. Vlad and J. Ross.} \emph{Systematic derivation
1064: of reaction-diffusion equations with distributed delays and relations to
1065: fractional reaction-diffusion equations and hyperbolic transport equations:
1066: Application to the theory of Neolithic transition}. Phys. Rev. E 66, 061908
1067: (2002).
1068:
1069: \bibitem{add1} \textsc{K. Seki, M. Wojcik, and M. Tachiya.} \emph{%
1070: Fractional reaction-diffusion equation}. J. Chem. Phys. 119, 2165, (2003).
1071:
1072: \bibitem{gd} \textsc{V. Gafiychuk, B. Datsko.} \emph{Pattern formation in a
1073: fractional reaction-diffusion system}. Physica A: Statistical
1074: Mechanics and its Applications. Vol. 365, (2006), pp. 300--306.
1075:
1076: \bibitem{GDI05} \textsc{V.V.Gafiychuk, B.Y.Datsko,
1077: Yu.Yu.Izmajlova}. Analysis of the dissipative structures in
1078: reaction-diffusion systems with fractional derivatives. Math.
1079: metody ta phys.-mech. polia. V.49, \#4, 2006, pp. 109-116 (in
1080: Ukrainian)
1081:
1082: \bibitem{ar1} \textsc{R.K. Saxena, A.M. Mathai and H.J. Haubold.}
1083: Fractional reaction-diffusion equations. arXiv:math.CA/0604473 v1 21 Apr
1084: 2006.
1085:
1086: \bibitem{zw} \textsc{G.M. Zaslavsky and H. Weitzner} Some Applications of
1087: Fractional Equations. E-print nlin.CD/0212024, (2002).
1088:
1089: \bibitem{vb} \textsc{C. Varea and R. A. Barrio.} Travelling Turing patterns
1090: with anomalous diffusion. J. Phys.: Condens. Matter 16, (2004), pp. 5081--5090.
1091:
1092: \bibitem{tz} \textsc{\ V. E Tarasov and G. M Zaslavsky.} Nonholonomic
1093: constraints with fractional derivatives. J. Phys. A: Math. Gen. 39
1094: (2006), pp. 9797--9815.
1095:
1096: \bibitem{skm} \textsc{S.G. Samko, A. A. Kilbas, and O. I. Marichev,}\emph{\
1097: Fractional Integrals and Derivatives: Theory and Applications,} Gordon and
1098: Breach, Newark, N. J., 1993.
1099:
1100: \bibitem{po} \textsc{I. Podlubny}. \emph{Fractional Differential Equations}%
1101: . Academic Press, 1999.
1102:
1103: \bibitem{os06} \textsc{Z. M. Odibat, N. T. Shawagfeh} \emph{Generalized
1104: Taylor's formula. Applied Mathematics and Computation} xxx (2006) xxx-xxx
1105: (available online at www.sciencedirect.com)
1106:
1107: \bibitem{a} \textsc{R Yu, H. Zhang,} \emph{New function of Mittag-Leffler
1108: type and its application in the fractional diffusion-wave equation,} Chaos,
1109: Solitons and Fractals 30 (2006), pp. 946--955.
1110:
1111: \bibitem{a94} \textsc{G. Adomian}. Solving Frontier Problems of Physics:
1112: The Decomposition Method, Kluwer Academic Publishers, Dordrecht, 1994.
1113:
1114: \bibitem{bbi04} \textsc{\ J. Biazar, E. Babolian, R. Islam} Solution of the
1115: system of ordinary differential equations by Adomian decomposition method,
1116: Appl. Math. Comput. 147 (3) (2004), pp. 713--719.
1117:
1118: \bibitem{jd06} \textsc{H. Jafari, V. Daftardar-Gejji}. \emph{Solving a
1119: system of nonlinear fractional differential equations using Adomian
1120: decomposition}. Journal of Computational and Applied Mathematics 196 (2006), pp. 644 - 651
1121:
1122: \bibitem{db04} \textsc{V. Daftardar-Gejji, A. Babakhani}. Analysis of a
1123: system of fractional differential equations, J. Math. Anal. Appl.
1124: 293 (2004), pp. 511--522.
1125:
1126: \bibitem{zse05} \textsc{G.M. Zaslavsky, A.A. Stanislavsky, M. Edelman}
1127: Chaotic and Pseudochaotic Attractors of Perturbed Fractional Oscillator,
1128: arXiv:nlin.CD/0508018.
1129:
1130: \bibitem{mi} \textsc{D. Matignon,} \emph{Stability results for fractional
1131: differential equations with applications to control processing, }%
1132: Computational Eng. in Sys. Appl., Vol. 2, Lille, France 963, 1996.
1133:
1134: \bibitem{El96} \textsc{A.M.A. El-Sayed.} \emph{Fractional
1135: differential-difference equations,} J. Fract. Calc. 10 (1996), pp. 101--106.
1136:
1137: \bibitem{wz} \textsc{H.Weitzner, G.M.Zaslavsky}. \emph{Some applications of
1138: fractional equations}. Communications in Nonlinear Science and Numerical
1139: Simulation 8 (2003), pp. 273--281.
1140:
1141: \bibitem{oldh} \textsc{K.D.Oldham and J.Spanier} \emph{The Fractional
1142: Calculus: Theory and Applications of Differentiation and Integration to
1143: Arbitrary Order, Vol.111 of Mathematics in Science and Engineering }
1144: Academic Press, New York, 1974.
1145:
1146: \bibitem{gk91} \textsc{V.V.Gafiychuk, B.S.Kerner, V.V.Osipov,
1147: T.M.Scherbatchenko}. Formation of pulsating thermal-diffusion autosolitons
1148: and turbulence in a nonequilibrium electron-hole plasma. Sov. Phys. Sem.
1149: (USA). v.25, \#11, (1991) (Translation of Fiz.Tekh. Poluprovodn (USSR).
1150: v.25, No.11, (1991), pp. 1696--1702).
1151:
1152: \bibitem{gd98} \textsc{V.V.Gafiychuk, A.V.Demchuk}. Analysis of the
1153: dissipative structures in Gierer-Meinhardt model. Matematicheskie metody i
1154: physico-mechanicheskie polia. V.40, \#2, 1997, pp.48-53 (in Ukrainian,
1155: English translation in Journal of Mathematical Sciences, v.88, \#4, 1998).
1156:
1157: \end{thebibliography}
1158:
1159: \end{document}
1160: