You are on page 1of 10

May 3, 2012 14:27 WSPC/S0218-1274 1250087

International Journal of Bifurcation and Chaos, Vol. 22, No. 4 (2012) 1250087 (10 pages)
c World Scientic Publishing Company
DOI: 10.1142/S0218127412500873
PATTERN FORMATION IN FRACTIONAL
REACTIONDIFFUSION SYSTEMS WITH
MULTIPLE HOMOGENEOUS STATES
BOHDAN DATSKO

Institute of Applied Problems of Mechanics and Mathematics,


NASU, Naukova 3b, Lviv, 79061, Ukraine
b datsko@yahoo.com
YURY LUCHKO
Beuth Technical University of Applied Sciences,
Berlin, Germany
luchko@beuth-hochschule.de
VASYL GAFIYCHUK
NASA Ames Research Center, Moett Field, USA
vagaf@yahoo.com
Received March 31, 2011; Revised June 17, 2011
This paper is devoted to the investigation of self-organization phenomena in time-fractional
reactiondiusion systems with multiple homogeneous states. It is shown that the fractional
reactiondiusion systems possess some new properties compared to the systems with derivatives
of integer orders. In particular, some complex spatio-temporal solutions that cannot be found in
the standard reactiondiusion systems are identied. The simulation results are presented for
the case of a incommensurate time-fractional reactiondiusion system with a cubic nonlinearity.
Keywords: Bifurcation; pattern formation; fractional derivative; time-fractional reaction
diusion equation.
1. Introduction
Within the last few decades, the investigation
of reactiondiusion systems (RDS) could essen-
tially extend the understanding of nonlinear phe-
nomena to many complex systems. On the basis
of mathematical modeling in terms of reaction
diusion equations, several nonlinear eects in
physical, biological, and chemical systems were
described and explained (see e.g. [Nicolis & Pri-
gogine, 1977; Cross & Hohenberg, 1993; Kerner &
Osipov, 1994]). It is well known that the dynamics
of a standard reactiondiusion system (SRDS) is
based on its local characteristics and can be clas-
sied into bistable, excitable, and HopfTuring
unstable systems. Within these three categories,
both dissipative pattern formation and auto-waves
can be observed. In this paper, a generalization
of the SRDSs the so called time-fractional
reactiondiusion systems are considered. We
shall restrict ourselves to the bistable systems and
show that their local dynamics strongly depends on
the orders of the fractional derivatives and is essen-
tially dierent from one observed in SRDS.
A bistable system is characterized by the prop-
erty that it possesses exactly two stable steady

Author for correspondence


1250087-1
May 3, 2012 14:27 WSPC/S0218-1274 1250087
B. Datsko et al.
states. Usually both steady states are separated by
a saddle point. If the system is located in one of
the stable states, small perturbations do not cause
any changes in the system state. But when pertur-
bations become suciently big, a transition from
one steady state to another one can happen. In
some cases, the transition can trigger a traveling
front or transition waves that propagate through
the medium.
Whereas an integer order derivative of a func-
tion depends only on the local behavior of the func-
tion, a fractional one takes into account all values
of the function and thus can be used in the mod-
eling of processes with memory. Applications of
the fractional dierential equations in modeling of
anomalous and other complex processes and their
analysis are attracting even more attention of both
physicists and mathematicians (see e.g. [Podlubny,
1999; Metzler & Klafter, 2000; Kilbas et al., 2006;
Agrawal et al., 2007; Uchaikin, 2008; Monje et al.,
2010; Caponetto et al., 2010; Ortigueira, 2011]). In
particular, the fractional reactiondiusion systems
(FRDS) seem to be a useful technique for modeling
of self-organization phenomena that possess some
kind of their own memory, too. The memory eects
of the fractional dierentiation operators result in
changing of dynamics of characteristic types of
spatially-non-homogeneous solutions to FRDS and
lead to some new classes of nonlinear solutions.
Moreover, in contrast to the SRDS with integer
derivatives, properties of solutions to FRDS depend
both on the orders of the fractional derivatives and
on the ratio between them.
In a number of papers (see e.g. [Henry et al.,
2005; Gaychuk & Datsko, 2006; Langlands et al.,
2007; Nec & Nepomnyashchy, 2007; Gaychuk &
Datsko, 2008; Gaychuk et al., 2009; Gaychuk &
Datsko, 2010; Datsko & Gaychuk, 2011]), prop-
erties of pattern formation in time-fractional RDS
with one homogeneous state were analyzed. This
paper is devoted to the investigation of self-
organization phenomena in FRDS with multiple
homogeneous states. Similar to the case of the
FRDS with one homogeneous state, the solutions
to the FRDS with multiple homogeneous states are
shown to possess some new properties compared to
the SRDS with derivatives of integer order.
2. Mathematical Model
Our starting point is the following one-dimensional
time-fractional reactiondiusion system (see
[Kerner & Osipov, 1994; Gaychuk & Datsko,
2008])

1
n
1
(x, t)
t

1
= l
2
1

2
n
1
(x, t)
x
2
+ W(n
1
, n
2
, A), (1)

2
n
2
(x, t)
t

2
= l
2
2

2
n
2
(x, t)
x
2
+ Q(n
1
, n
2
, A) (2)
on the interval ]0, L
x
[ subject to the Neumann
boundary conditions
n
i
x

x=0
=
n
i
x

x=Lx
= 0, i = 1, 2
and to certain initial conditions. In Eqs. (1)
and (2), n
1
(x, t) means the activator and n
2
(x, t)
the inhibitor variable, W(n
1
, n
2
, A), Q(n
1
, n
2
, A)
are smooth reaction kinetics functions,
1
,
2
, l
1
,
l
2
are the characteristic times and lengths of the
system, respectively, and A is an external param-
eter. Fractional derivatives d

i
n
i
/dt

i
, i = 1, 2 on
the left-hand side of (1) and (2) are understood in
the Caputo sense (see e.g. [Podlubny, 1999; Kilbas
et al., 2006; Uchaikin, 2008]):
d

dt

n(t)
1
(s )
_
t
0
n
(s)
()
(t )
+1s
d,
s 1 < < s, s N. (3)
When the ratio
1
/
2
of orders of the fractional
derivatives in (1) and (2) is a rational number, the
systems (1) and (2) can be transformed to an equiv-
alent system of the fractional dierential equations
of the same order (
1
= p,
2
= q, p, q N)
[Gaychuk & Datsko, 2008]:

n
1
(x, t)
t

= n
3
,

n
2
(x, t)
t

= n
p+2
, (4)

n
i
(x, t)
t

= n
i+1
, i [3, p] [p + 2, q + p 1]
(5)

n
p+1
(x, t)
t

= l
2
1

2
n
1
(x, t)
x
2
+ W(n
1
, n
2
, A), (6)

n
p+l
(x, t)
t

= l
2
2

2
n
2
(x, t)
x
2
+ Q(n
1
, n
2
, A). (7)
1250087-2
May 3, 2012 14:27 WSPC/S0218-1274 1250087
Pattern Formation in Fractional ReactionDiusion Systems
Using the presentation (4)(7), the character-
istic equation for the FRDS (1) and (2) can be
explicitly written down for any relation between
derivative orders [Gaychuk & Datsko, 2008]. It
gives the key to the analysis of main properties
of incommensurate system depending on the base
parameters.
It should be noted that the introduction of
time-fractional derivatives into SRDS essentially
extends the family of reactiondiusion systems.
The FRDS provide a possibility of transitions
between the parabolic, elliptic and hyperbolic types
of partial dierential equations. In other words, the
systems (1) and (2) can be considered to be a gen-
eralized reactiondiusion system that describes all
combinations of diusion and oscillation processes.
Analysis of the basic properties of such systems and
their solutions thus is of considerable interest even
in the case of nonlinearities of a simple type.
3. Stability Analysis of Standard
and Fractional Reaction Diusion
Systems
To demonstrate the properties of the FRDS, let us
consider a model with the polynomial nonlinearity
W(n
1
, n
2
) = n
1
an
3
1
n
2
(8)
of the third order for the activator variable and the
linear function
Q(n
1
, n
2
) = n
2
+ bn
1
+A (9)
for the inhibitor variable. This model is taken just
for illustration aims and the obtained results can
be generalized for dierent types of systems with
N-shaped null-clines (see e.g. [Kerner & Osipov,
1994]).
The homogeneous solutions n
1
and n
2
of the
Eqs. (1) and (2) are obtained from the system of
equations
W(n
1
, n
2
) = 0, Q(n
1
, n
2
) = 0 (10)
that can have one or three solutions depending
on the system parameters. The equations can be
graphically represented in the phase space in form
of two null-clines n
2
= n
2
(n
1
). These null-clines are
crossed at a point P = (n
1
, n
2
) when the condition
n
1
an
3
1
bn
1
A = 0 holds true (see Fig. 1). In this
case, the values of n
1
are determined by the values
of the external parameters a, b, and A. Thus we can
consider the parameter n
1
as the main parameter
for the system analysis [Gaychuk & Datsko, 2010].
In the case under consideration, the null-clines can
have either one intersection point (b > 1) or three
ones (b < 1) depending on the value of b and A.
In this paper, the focus is mainly on a fractional
bistable system [Fig. 1(b)] in case it is unstable in
all three steady states. In particular, the values of
the system parameters that lead to instability are
determined.
Let us rst consider stability properties of the
SRDS with respect to homogeneous perturbations
in the bistable null-clines intersections. The local
dynamics of the systems (1) and (2) on the (n
1
, n
2
)
phase plane is governed by a N-shaped null-cline for
n
1
and a linear n
2
-null-cline [Figs. 1(a) and 1(b)].
A homogeneous stationary state n
s
= (n
s
1
, n
s
2
)
T
is determined as a solution of system (10). The
slope of the null-cline W(n
1
, n
2
) = 0 is given by
dn
2
/dn
1
|
W
= W

n
1
(n
1
, n
2
)/W

n
2
(n
1
, n
2
) and that
(a) (b)
Fig. 1. Typical intersections of null-clines W = n
1
an
3
1
n
2
= 0, and Q = bn
1
n
2
+ A = 0 for (a) mono-stable and
(b) bistable system.
1250087-3
May 3, 2012 14:27 WSPC/S0218-1274 1250087
B. Datsko et al.
for the null-cline Q(n
1
, n
2
) = 0 by dn
2
/dn
1
|
Q
=
Q

n
1
(n
1
, n
2
)/Q

n
2
(n
1
, n
2
). A bistable system is
characterized by the property that at the central
intersection point n
0
= (n
0
1
, n
2
0
)
T
on the unstable
(increasing) part of the null-cline W(n
1
, n
2
) = 0
the condition dn
2
/dn
1
|
W
> dn
2
/dn
1
|
Q
holds true
whereas on the stable (decreasing) parts of the null-
cline W(n
1
, n
2
) = 0 at the upper intersection point
n
1
= (n
1
1
, n
1
2
)
T
and the lower intersection point n
2
=
(n
2
1
, n
2
2
)
T
the inequality dn
2
/dn
1
|
W
< dn
2
/dn
1
|
Q
is
satised [Fig. 1(b)]. As a result, in this case the
system is unstable at the central intersection point
n
0
= (n
0
1
, n
0
2
)
T
and stable at two other intersection
points if they are located behind the increasing part
of the null-cline W(n
1
, n
2
) = 0.
Now let us start with the stability analy-
sis of the FRDS. The stability of the homoge-
neous steady-state solutions of the time-fractional
reactiondiusion systems (1) and (2) for
1
=

2
= with 0 < < 2 can be analyzed by the
linearization of the right-hand sides of the system
near these solutions [Gaychuk et al., 2008]. The
linearization of (1) and (2) leads to a system of
ODEs of fractional order with the right-hand side
matrix given by
F(k) =
_
_
_
_
_
_
a
11
k
2
l
2
1

1
a
12

1
a
21

2
a
22
k
2
l
2
2

2
_
_
_
_
_
_
, (11)
where a
11
= W

n
1
, a
12
= W

n
2
, a
21
= Q

n
1
, a
22
= Q

n
2
and k = (/L
x
)j, j = 0, 1, 2, . . . .
The eigenvalues
1,2
of the matrix F(k) are
determined by the formula

1,2
=
1
2
(tr F
_
tr
2
F 4 det F) (12)
with tr F(k) = (a
11
k
2
l
2
1
)/
1
+(a
22
k
2
l
2
2
)/
2
and
det F(k) = ((a
11
k
2
l
2
1
)(a
22
k
2
l
2
2
) a
12
a
21
)/
1

2
.
For every point inside of the parabola det F =
tr
2
F/4, the roots
1,2
are complex and we can intro-
duce marginal values of fractional derivative order
[Gaychuk et al., 2008]:

0
=
_

_
2

arctan
_
4 det F
tr
2
F
1, tr F > 0,
2
2

arctan
_
4 det F
tr
2
F
1, tr F < 0.
(13)
The derivative order
0
for systems (1) and (2) is
an additional bifurcation parameter which changes
the stability of homogeneous stationary solutions.
For the values of from the interval ]0,
0
[, the
system possesses some oscillatory modes, but they
are stable. When becomes larger than
0
, the
systems (1) and (2) shows oscillatory instability.
For the coecients a
ij
of the Jacoby matrix for
a bistable system, the inequalities a
11
> 0, a
12
< 0,
a
22
< 0, a
21
> 0 hold true in the case of an unsta-
ble steady-state solution whereas the inequalities
a
11
< 0, a
12
< 0, a
22
< 0, a
21
> 0 have to
be satised for a stable steady-state solution. For
homogeneous perturbations (k = 0) we get in this
case det F(0) = a
11
a
22
a
12
a
21
< 0 for an unstable
point and det F(0) > 0 for a stable one. In turn,
the trace of the system tr F = (a
11
)/
1
+ (a
22
)/
2
is greater than zero for an unstable point under the
condition
1

2
and for a stable point it is always
less than zero. As a result, we can conclude that the
steady-state solutions at outer intersection points of
the null-clines of the standard systems (1) and (2)
are always stable since for them, the inequalities
tr F < 0 and det F > 0 hold true.
For FRDS (1) and (2) we can thus determine
the critical value
0
of the bifurcation parameter ,
such that for >
0
the steady-state solutions of
the system became unstable. When
1

2
and a
11
is close to zero, the value of tr F is suciently small
so that the instability of the system can be easily
realized for > 1, since tr F(0) is a very small value
in this case. Results of numerical calculation of
0
for dierent values of parameter
2
are presented in
Fig. 2. The top plot of Fig. 2(a) presents both null-
clines and their intersection points that correspond
to the steady-state solutions. On the bottom plot of
Fig. 2(a),
0
as a function of the outer intersection
points of the null-clines is given. This plot shows
that for
2
with a big enough value, the instabil-
ity region on n
1
is small. Making the value of
2
smaller increases the instability region on n
1
and
increases the order of the fractional derivative in
the systems (1) and (2) that leads to the instability.
Dependence of
0
on
2
and the upper intersection
point n
1
= (n
1
1
, n
1
2
)
T
of the null-clines is presented
in Fig. 2(b) [the dot lines correspond to the solid
lines in Fig. 2(a)].
Under obtained oscillation conditions, a limit
cycle in the vicinity of the upper unstable point
emerges. When the amplitude of the oscillation
becomes greater than |n
1
n
0
| a transition wave
1250087-4
May 3, 2012 14:27 WSPC/S0218-1274 1250087
Pattern Formation in Fractional ReactionDiusion Systems
(a) (b)
Fig. 2. (a) The null-clines W = n
1
an
3
1
n
2
= 0, Q = bn
1
n
2
+ A = 0, a = 1/3, b = 1/3, A = 0 in coordinates n
1
, n
2
and dependence of
0
on n
1
for
1
= 1 and dierent values of
2
:
2
= 0.5 (blue),
2
= 1 (gray),
2
= 2 (black),
2
= 10 (red
lines, (b) dependence of
0
on (n
1
1
,
2
/
1
).
appears in the way similar to what is observed
in the SRDS. The dierence to the SRDS is that
whereas in the SRDS this motion has to be forced
by an external perturbation, the transition wave
in the FRDS can arise spontaneously without any
external inuence. At the lower intersection point
n
2
= (n
2
1
, n
2
2
)
T
of the null-clines, a similar dynam-
ics is observed. This point is unstable with respect
to the homogeneous oscillations, too, that forces
the limit cycle to go over into a transition wave
that transits the system into the upper intersection
point. As a result, a complex spatio-temporal non-
linear dynamics for a broad range of system param-
eters of the FRDS is expected.
If k = 0, the instability conditions for the inter-
section points of the null-clines are more strict com-
pared to the case k = 0 since the trace of the
matrix (11) becomes less. For the outer intersection
points the inequalities
a
22
k
2
l
2
1
a
11
k
2
l
2
2
+ k
4
l
2
1
l
2
2
+a
11
a
22
a
21
a
21
> 0
have to be satised to make the steady-state
solution unstable. If a
22
< 0, a
11
< 0, the for-
mulae (13) became more strict since some nonho-
mogeneous structures may arise for larger values of

0
. In this case, the homogenous oscillations have
predetermined spatio-temporal dynamics.
In the case
1
=
2
, 0 <
1
,
2
< 2, the
spectrum of the nonlinear time-fractional reaction
diusion systems (1) and (2) is described by the
characteristic equation [Gaychuk & Datsko, 2008]
()
p+q
+ (1)
q1
a
22
(k)

2
()
p
+(1)
p+1
a
11
(k)

1
()
q
+ (1)
p+q
det F(k)
= 0 (14)
with a
11
(k) = a
11
k
2
l
2
1
, a
22
(k) = a
22
k
2
l
2
2
.
Based on the representation (4)(7) of the sys-
tems (1) and (2), the eigenvalue spectrum and the
instability domains of homogeneous solutions to the
systems (1) and (2) can be obtained by numeri-
cal simulations for any parameters and derivatives
orders
1
,
2
. If = det F is a small enough num-
ber, the roots of Eq. (14) with the absolute value
close to zero can be found by the perturbation
technique:

i
=
_

a
22
+ O(
2
)
_
1/p
, if
1
<
2
, i 1, p,

i
=
_

a
11
+ O(
2
)
_
1/q
, if
1
>
2
, i 1, q.
(15)
For the eigenvalues given by (15), the conditions
under that they are greater than zero and thus
the instability conditions can be determined in
explicit form. For the middle intersection point of
the null-clines the inequalities det F <0, a
11
>0
and a
22
<0 hold true and thus the system is
unstable at this point. For the outer intersection
points of the null-clines we get det F >0, a
11
<0,
a
22
<0 and the instability is because of the rst
1250087-5
May 3, 2012 14:27 WSPC/S0218-1274 1250087
B. Datsko et al.
expression in (15). The homogeneous state of the
time-fractional RDS (1) and (2) can thus always
become unstable. In this case, the system can show
a complex auto-wave dynamics. This situation takes
place when the null-clines of the system become
nearly tangential to each other.
4. Numerical Simulations for the
Bistable Fractional
ReactionDiusion Model
In this section, some results of computer simula-
tions for the time-fractional model (1), (2) with the
nonlinear source term (8) for the activator variable
and the linear source term (9) for the inhibitor vari-
able are presented. The FRDS (1), (2) along with
the corresponding initial and boundary conditions
was integrated numerically using an implicit dif-
ference scheme for the time-fractional derivatives
and a centered dierence approximation for the
spatial derivatives. The Caputo time-derivatives for
0 < < 2 were approximated using the Gr unwald
Letnikov denition of the fractional derivatives (see
e.g. [Podlubny, 1999; Uchaikin, 2008]). It is well
known (see e.g. [Podlubny, 1999; Weilbeer, 2005])
that the Caputo and the Gr unwaldLetnikov frac-
tional derivatives for a function u C
s
[0, ], s =
[] are connected to each other by the following
relation
C
0
D

u() =
GL
0
D

u()

i=0
()
i
(i + 1)

i
u(0
+
). (16)
The Gr unwaldLetnikov fractional derivative can
be approximated on the interval ]0, [ by the nite
sum (see [Podlubny, 1999; Uchaikin, 2008])
GL
0
D

u()
[/t]

j=0
c
()
j
u( jt),
where c
()
j
= (t)

(1)
j
(

j
) are the Gr unwald
Letnikov coecients that can be computed by using
a recursive scheme. By discretization of the system
of m time-fractional RD equations

j
C

j
u
j
(x, t)
t

j
= d
j

2
u
j
(x, t)
x
2
+ f
j
(u
1
, . . . , u
m
),
j = 1, m,
where
j
, d
j
, f
j
are certain parameters and nonlin-
earities of the FRDS, correspondingly, the numeri-
cal scheme can be represented in the form
u
k
j,i

d
j
(t)

j
(x)
2
(u
k
j,i1
2u
k
j,i
+ u
k
j,i+1
)

(t)

j
f
j
(u
k
1,i
, . . . , u
k
m,i
)
= (t)

j
s

p=0
(kt)
p
j
(p
j
+ 1)

p
t
p
u
0
j,i

l=1
c
(
j
)
l
u
kl
j,i
,
where u
k
j,i
u
j
(x
i
, t
k
) u
j
(ix, kt), s = [],
c
(
j
)
0
= 1, c
(
j
)
l
= c
(
j
)
l1
(1
1+
j
l
), l = 1, 2, . . . .
Because the applied numerical scheme is an implicit
one, for each time layer it takes the form of a sys-
tem of algebraic equations that were solved by the
NewtonRaphson technique. This approach makes
it possible to get a system of equations with a band
Jacobian for each node and to use the sweep method
for the solution of the obtained system of linear
algebraic equations. It is worthy to note that even
if the explicit schemas are simpler compared to the
implicit ones, they are very sensitive with respect
to the step value that as a rule leads to numerical
instability. The implicit scheme used for numerical
simulations presented in this paper is in a certain
sense similar to the implicit Eulers method. For
the time-fractional reactiondiusion systems, very
good numerical stability behavior is shown both for
dierent step sizes of integration and for diverse
nonlinearities. It should be noted that other implicit
numerical methods based on the fractional formula-
tion of the classical Adams method (for more details
see e.g. [Diethelm et al., 2005; Diethelm et al., 2006;
Tadjeran et al., 2006; Li & Tao, 2009; Weilbeer,
2005]) show good behavior for numerical investiga-
tion of nonlinear FRDS, too. The reasons why we
decided to employ the numerical methods presented
above were the nonlocal character of the fractional
derivatives and exibility of the scheme based on
the Gr unwaldLetnikov denition of the fractional
derivative. Even if our approach is probably more
costly in terms of computational time and storage
requirements compared to other known numerical
methods for the dierential equations of fractional
order, it turned out to be quite eective at least for
1250087-6
May 3, 2012 14:27 WSPC/S0218-1274 1250087
Pattern Formation in Fractional ReactionDiusion Systems
systems of one-dimensional time-fractional dieren-
tial equations dealt with in this paper.
For computer simulation, the most interest-
ing case of the systems (1) and (2), namely, when

1
/
2
1 and l
1
/l
2
1 was considered. Under
these conditions, a very complex spatio-temporal
dynamics is realized in the standard RDS, because
in this case both the Hopf and the Turing types
of instability appear simultaneously [Kerner &
Osipov, 1994]. Intuitively, these conditions should
lead to complex dynamics in fractional RDS, too.
Moreover, in the case of FRDS, the relation between
the orders of the fractional derivatives can signi-
cantly change a spectrum of the linearized system
and lead therefore to a much more complex system
dynamics compared to the case of SRDS.
The instability of the time-fractional sys-
tems (1) and (2) is mainly determined by the roots
of the characteristic equation (14). Consider, for
example, a typical eigenvalue spectrum under the
conditions
1
/
2
1 and l
1
/l
2
1 for the orders

1
,
2
of the fractional derivatives that satisfy the
relations
1
=
2
and
1
= 2
2
, respectively
(see Fig. 3). In the case of mono-stable system
at
1
=
2
the spectrum of eigenvalues does not
allow a sharp separation of eigenvalues for dier-
ent wave numbers [Fig. 3(a)]. In the central part of
the plot Fig. 3(a), some approximately equal eigen-
values of the linearized system for dierent wave
numbers k = 0, 1, 2 can be seen. An eigenvalue
for a mode with the shortest wave number is the
largest one in magnitude and therefore conditions
for the Turing bifurcation become true. From the
other side, because of the preferable development
of oscillations caused by a Hopf bifurcation, in the
area of complex eigenvalues the Turing pattern for-
mation cannot be dominant. Therefore, practically
in the whole region of system instability the spatio-
temporal oscillations of large amplitude, which is
limited by decreasing parts of the null-clines of the
systems (1) and (2), are observed.
Variations in the ratio of the fractional deriva-
tive orders qualitatively change the spectrum of
the linearized system as well as the type of sys-
tem solutions. Eigenvalues of the linearized system
for the case
1
= 2
2
are presented in Fig. 3(c).
(a) (b)
(c) (d)
Fig. 3. (a), (c) Imaginary (black, gray) and real (red, blue) parts of eigenvalues in cases of one ( = 2) and (b), (d) three
intersection points ( = 0.8) as function of n
1
for k = 0 (solid lines), k = 1 (dot lines), k = 2 (dash lines) for the parameters
values
1
= 0.1,
2
= 1.0, l
2
1
= 0.05, l
2
2
= 1.0. In (a), (b)
1
=
2
, in (c), (d)
1
= 2
2
.
1250087-7
May 3, 2012 14:27 WSPC/S0218-1274 1250087
B. Datsko et al.
Domination of the activator derivative over the
inhibitor one leads to the domination of the modes
k = 0 compared to the one for k = 0. In its
turn, this leads to domination of spatially inho-
mogeneous solutions over the homogeneous oscilla-
tions. Because the eigenvalues for k = 1 and k = 2
are very close to each other, some small perturba-
tions of the homogeneous solutions corresponding to
these wave numbers can form stationary dissipative
structures with a period that corresponds to these
initial perturbations. Making the value =
1
=
2
2
larger leads to an increase of inuence of the
ratio between imaginary and real parts of eigenval-
ues with long wave modes and to a transition from
the space-inhomogeneous solutions to the homo-
geneous oscillations. If the inuence of dierent
modes is nearly of the same order, complex spatio-
temporal oscillations can arise in the time-fractional
systems even at
2
<
1
< 1 [see Figs. 3(c)
and 4(a)].
Decreasing the parameter b in (9) changes over
the system from a mono-stable state to a bistable
one and leads to the situation when the eigenval-
ues for dierent wave numbers k start to overlap
[Figs. 3(b) and 3(d)]. Similar to the case
1
=
2
,
for
1
= 2
2
three unstable stationary homoge-
neous solutions of the time-fractional RDS can be
observed. But whereas in the case =
1
=
2
the
outer intersection points become unstable for > 1,
in the case
1
>
2
they can be unstable even for

1
< 1 and
2
< 1 [see Figs. 3(b) and 3(d)].
In a standard bistable system, two out of three
stationary states are stable when the parameter A
varies in a certain interval. This is a typical situa-
tion for many nonlinear systems. Like in the case
of the mono-stable systems, the bistable systems
can show auto-waves of a complex type under the
condition Q

n
1
> 0, Q

n
1
W

n
2
< 0. However, evo-
lution of spatio-temporal solutions of the bistable
systems has a number of peculiarities compared
(a) (b)
(c) (d)
Fig. 4. Pattern formation scenarios in system with (a) one- and (b)(d) three intersection points of the null-clines. Results
of computer simulation for the parameters values (a) l
2
1
= 0.1, l
2
2
= 1,
1
= 0.1,
2
= 1,
1
= 1.0,
2
= 0.5, = 2.0, A = 0.4;
(b) l
2
1
= 0.05, l
2
2
= 1,
1
= 0.1,
2
= 1 and
1
= 1.0,
2
= 0.5, = 0.85, A = 0.05; (c)
1
= 1.5,
2
= 0.75, = 0.85, A = 0.1
and (d)
1
= 1.5,
2
= 0.75, = 0.95, A = 0.15.
1250087-8
May 3, 2012 14:27 WSPC/S0218-1274 1250087
Pattern Formation in Fractional ReactionDiusion Systems
to those of the mono-stable systems. In particular,
auto-waves of a complex type that switch a system
from one stable state to another one can be excited
in the bistable systems. Type and velocity of these
auto-waves depend on the corresponding intersec-
tion points of the null-clines. The order of fractional
derivatives and the ratio between them can quali-
tatively change properties of systems with multi-
ple steady states (Fig. 3). In particular, increasing
the order of fractional derivatives and ratio between
them can lead to a situation when all three steady
states become unstable. The plots from Figs. 3(c)
and 3(d) demonstrate that unlike the situation
presented in Figs. 3(a) and 3(b) the eigenvalues
with positive real parts belong to the interval that
exceeds the interval (1, 1) on which the null-cline
W = 0 is an increasing function n
2
= n
2
(n
1
). As
a result, all three points of the null-clines intersec-
tions are unstable even for
1
< 1 and
2
< 1.
Because the spectrum of eigenvalues does not con-
tain a sharp separation of eigenvalues for dierent
wave numbers k, a bistable FRDS can demonstrate
a very complex nonlinear dynamics including com-
plex transition waves and spatio-temporal chaos
[Figs. 4(b)4(d)]. Moreover, under some conditions,
dierent stationary solutions to the fractional RDS
can be unstable with respect to the dierent bifur-
cation types that cannot ever be realized within the
standard RDS.
The situation becomes even more complicated
when the parameter b tends to 1 from left (b < 1)
that corresponds to the case when both null-clines
have similar slopes. In this case, in the middle
intersection point of the null-clines the modes for
the wave numbers k = 0 have higher increment
compared to the modes with k = 0 and at the
outer intersection points the modes with k = 0 are
unstable in contrast to the inhomogeneous modes.
Because of interaction of these modes, some com-
plex spatio-temporal patterns appear in the system
[Figs. 4(c) and 4(d)]. In Fig. 4(b), the transition
waves in a bistable system can be seen. Increas-
ing the fractional derivative order as well as chang-
ing the value of the intersection points can lead to
the complicated spatio-temporal patterns including
chaotic dynamics [see Fig. 4(d)].
5. Conclusions
In this paper, the spatio-temporal structures
observed in a bistable time-fractional reaction
diusion system have been investigated. It was
shown that the characteristic time scales and the
orders of the fractional derivatives play an impor-
tant role for the instability conditions of the time-
fractional reactiondiusion systems (1) and (2).
The character of bifurcations and stability of the
spatio-temporal patterns of this system were dis-
cussed, too. The nonlinear dynamics of the system
was determined by the method of computer sim-
ulation in the case of three intersection points of
the null-clines. Dierent kinds of complex spatio-
temporal patterns that combine oscillation and
quasi-chaotic features were observed for the bistable
system. Increasing the orders of the fractional
derivatives in the systems (1) and (2) and the
inhibitor time lead to even more complex spatio-
temporal patterns.
Acknowledgments
The rst named author is thankful for the sup-
port by the German Academic Exchange Service
(DAAD, grant No. A-11-05319).
References
Agrawal, O. P., Tenreiro Machado, J. A. & Sabatier,
J. [2007] Advances in Fractional Calculus: Theoretical
Developments and Applications in Physics and Engi-
neering (Elsevier, Dordrecht).
Caponetto, R., Dongola, G., Fortuna, L. & Petras, I.
[2010] Fractional Order Systems: Modeling and Con-
trol Applications (World Scientic, Singapore).
Cross, M. C. & Hohenberg, P. C. [1993] Pattern for-
mation outside of equilibrium, Rev. Mod. Phys. 65,
8511112.
Datsko, B. & Gaychuk, V. [2011] Chaotic dynamics
in Bonhoervan der Pol fractional reactiondiusion
system, Sign. Process. 91, 452460.
Diethelm, K., Ford, N. J., Freed, A. D. & Luchko, Yu.
[2005] Algorithms for the fractional calculus: A selec-
tion of numerical methods, Comput. Meth. Appl.
Mech. Engrg. 194, 743773.
Diethelm, K., Ford, N. J., Ford, J. M. & Weilbeer, M.
[2006] Pitfalls in fast numerical solvers for fractional
dierential equations, J. Comp. Appl. Math. 186,
482503.
Gaychuk, V. & Datsko, B. [2006] Pattern formation
in a fractional reactiondiusion system, Physica A
365, 300306.
Gaychuk, V. & Datsko, B. [2008] Spatiotemporal pat-
tern formation in fractional reactiondiusion sys-
tems with indices of dierent order, Phys. Rev. E
77, 066210-19.
1250087-9
May 3, 2012 14:27 WSPC/S0218-1274 1250087
B. Datsko et al.
Gaychuk, V., Datsko, B. & Meleshko, V. [2008] Mathe-
matical modeling of time fractional reactiondiusion
systems, J. Comp. Appl. Math. 372, 215225.
Gaychuk, V., Datsko, B., Meleshko, V. & Blackmore, D.
[2009] Analysis of the solutions of coupled nonlinear
fractional reactiondiusion equations, Chaos Solit.
Fract. 41, 10951104.
Gaychuk, V. & Datsko B. [2010] Mathematical model-
ing of dierent types of instabilities in time fractional
reactiondiusion systems, Comp. Math. Appl. 59,
11011107.
Henry, B. I., Langlands, T. A. M. & Wearne, S. L. [2005]
Turing pattern formation in fractional activator-
inhibitor systems, Phys. Rev. E 72, 026101.
Kerner, B. S. & Osipov, V. V. [1994] Autosolitons
(Kluwer, Dordrecht).
Kilbas, A. A., Srivastava, H. M. & Trujillo, J. J. [2006]
Theory and Applications of Fractional Dierential
Equations (Elsevier, Netherlands).
Langlands, T. A. M., Henry, B. I. & Wearne, S. L. [2007]
Turing pattern formation with fractional diusion
and fractional reactions, J. Phys.: Condens. Matt.
19, 065115.
Li, C. P. & Tao, C. X. [2009] On the fractional Adams
method, Comp. Math. Appl. 58, 15731588.
Metzler, R. & Klafter, J. [2000] The random walks
guide to anomalous diusion: A fractional dynamics
approach, Phys. Rep. 339, 177.
Monje, C. A., Chen, Y. Q., Vinagre, B. M., Xue,
D. & Feliu, D. [2010] Fractional-order Systems and
Controls: Fundamentals and Applications (Springer,
London).
Nec, Y. & Nepomnyashchy, A. A. [2007] Turing insta-
bility in sub-diusive reactiondiusion systems, J.
Phys. A: Math. Theor. 40, 1468714702.
Nicolis, G. & Prigogine, I. [1997] Self-Organization in
Non-Equilibrium Systems (Wiley, NY)
Ortigueira, M. D. [2011] Fractional Calculus for Scien-
tists and Engineers (Springer, Dordrecht).
Podlubny, I. [1999] Fractional Dierential Equations
(Academic Press, San Diego).
Tadjeran, C., Meerschaert, M. M. & Scheer, H. P.
[2006] A second-order accurate numerical approxi-
mation for the fractional diusion equation, J. Com-
put. Phys. 213, 205213.
Uchaikin, V. [2008] Fractional Calculus (Artishok,
Ulyanovsk).
Weilbeer, M. [2005] Ecient Numerical Methods for
Fractional Dierential Equations and their Analyti-
cal Background (Dissertation, Technischen Universi-
tat Braunschweig).
1250087-10

You might also like