Commun. Theor. Phys. (Beijing, China) 45 (2006) pp.

761764
c International Academic Publishers Vol. 45, No. 4, April 15, 2006
Turing Patterns in a Reaction-Diusion System

WU Yan-Ning, WANG Ping-Jian, HOU Chun-Ju, LIU Chang-Song, and ZHU Zhen-Gang
Key Laboratory of Material Physics, Institute of Solid State Physics, the Chinese Academy of Sciences, P.O. Box 1129,
Hefei 230031, China
(Received July 6, 2005; Revised August 31, 2005)
Abstract We have further investigated Turing patterns in a reaction-diusion system by theoretical analysis and
numerical simulations. Simple Turing patterns and complex superlattice structures are observed. We nd that the shape
and type of Turing patterns depend on dynamical parameters and external periodic forcing, and is independent of eective
diusivity rate in the LengyelEpstein model. Our numerical results provide additional insight into understanding the
mechanism of development of Turing patterns and predicting new pattern formations.
PACS numbers: 82.40.Ck
Key words: Turing pattern, LengyelEpstein (LE) model
1 Introduction
In 1952, Alan Turing predicted that the interplay of
chemical reaction and diusion could destabilize a ho-
mogeneous stationary state and result in the formation
of temporally stationary, spatially inhomogeneous struc-
tures in his celebrated paper entitled the chemical ba-
sis of morphogenesis.
[1]
The necessary condition for the
pure Turing instability is that the diusion coecient of
the inhibitor should exceed that of activator.
[2]
Such
circumstances are rare in inorganic chemical systems. Tur-
ing structure remained experimentally elusive until 1990,
when De Kepper and his colleagues, working with an open
unstirred gel reactor, observed the rst experimental evi-
dence for Turing patterns in the chlorite-iodide-malonic
(CIMA) reaction.
[3]
Since diusion coecients of small
molecules are usually close to each other in solutions, the
Turing structures observed in CIMA system seem to be
aided by immobilization of iodine species through starch
that is present to act as an indicator.
[4]
A ve-variable model describing CIMA system was
reduced
[5]
by Lengyel and Epstein to a two-variable model
named as LengyelEpstein (LE) model. This model has
been widely used to investigate Turing patterns from nu-
merical simulations and experiments.
[611]
Here we fur-
ther explore the LE model by theoretical analysis and
numerical simulations. Families of complex structures,
consisting of combinations of basic hexagons, squares and
stripes, are found in the modied LE model by external
periodic forcing. We try to explain mechanism of devel-
opment of some pattern formations.
2 LE Model
To study Turing patterns in the CIMA reaction-
diusion system, we rstly employ the LE model. The
LE model is a two-species model with one equation for an
activator [I

] and another for inhibitor [ClO

2
]. In dimen-
sionless variables the model takes the form
[6,7]
u
t
= a u 4
uv
1 + u
2
+ u,
v
t
=

u
uv
1 + u
2

+ dv

,
where u and v are dimensionless concentrations of [I

]
and [ClO

2
], respectively; d is the ratio of their diusiv-
ities in aqueous, generally close to one; and a and b are
dimensionless parameters. The eective diusivity rate
becomes larger due to introduction of starch.
Fig. 1 Bifurcation diagram of the LengyelEpstein (LE)
model. b
T
is Turing bifurcation, while b
H
is Hopf bifur-
cation with xed a = 12.
Because the homogeneous system has only one steady
state, (u
ss
= a/5, v
ss
= 1 + (a
2
/25)), to homogeneous
perturbation, bifurcation diagram of the LE model can
be obtained by the linear stability analysis. In Fig. 1, we
show two lines where the homogeneous stationary state
will lose stability. The b
T
line represents the locus of the
bifurcation to a Turing structure given by
[6,7]
b
T
=
13a
2
4

10a

25 + a
2
+ 125
5a
.

The project supported by National Natural Science Foundation of China under Grant No. 10374089 and the Knowledge Innovation
Program of the Chinese Academy of Sciences under Grant No. KJCX2-SW-W17
762 WU Yan-Ning, WANG Ping-Jian, HOU Chun-Ju, LIU Chang-Song, and ZHU Zhen-Gang Vol. 45
The Turing bifurcation point does not depend on the con-
centration of the complex agent because simultaneously
renormalizes the reaction term and the diusion term in
the reaction-diusion equation. The boundary separating
the homogeneous steady state and homogeneous oscilla-
tion is the locus of Hopf bifurcation labeled by
[6,7]
b
H
=
3a
2
125
5a
.
The domain above Turing and Hopf bifurcation lines is
the region of the steady state (ss). Both Turing and Hopf
instability may occur when b decreases. For low values of
, precisely for < 12.934, the Hopf bifurcation always
precedes the Turing bifurcation. Above the value, the Tur-
ing bifurcation occurs rstly. In the pure Hopf instability
domain, the system rstly becomes bulk oscillation, and
then evolves into phase waves.
The reaction-diusion equations are solved numeri-
cally in two spatial dimensions using a nite dierence
approximation for the spatial derivatives and an explicit
Euler method for the time integration. The rst variable
is displayed in space-time plots, and gray levels quantify
its value; white corresponds to the maximum value and
black to the minimum. The time step is set as 0.0025.
The frame size is 256 256 space units. Zero ux or peri-
odic boundary condition is employed, and random initial
condition is used in our simulations. For zero ux bound-
ary condition, we have
u(0)
r
=
u(l)
r
=
v(0)
r
=
v(l)
r
= 0 .
For periodic boundary condition, we employ
u(0) = u(l) = v(0) = v(l) = 0 .
In our simulations, we take b as control parameter.
Figure 2 shows that near the boundary of Turing instabil-
ity, stable hexagons are observed [Fig. 2(a)]. With the dis-
tance to Turing bifurcation increased, hexagonal patterns
become unstable, and then convert to stripes [Fig. 2(b)].
Figure 2(c) shows that as b further decreases, stripes un-
dergo the Echhaus instability, resulting in labyrinthine
stripes. As b continually decreases, the system goes into
the domain where Turing instability and Hopf instability
interact in (b, ) phase space; labyrinthine stripes are bro-
ken into a stationary array of clear spots [Fig. 2(d)]. If we
take as variable and the other the same as those used in
Fig. 2, we can obtain pattern formations similar to Fig. 2.
The eective diusion rate seems not to determine the
type of Turing patterns in the LE model.
Fig. 2 Turing patterns in the LE model. In our simulations
a = 12, d = 1 and = 20. (a) Hexagons, b = 0.37; (b) Co-
existence of hexagons and stripes, b = 0.33; (c) Labyrinthine
stripes, b = 0.3; (d) Mixture of stripes and spots, b = 0.02.
Fig. 3 Superlattice square Turing patterns in the modied
LE model. (a) The (simulating) on-o square mask, period
n = 9. In our simulations a = 12, d = 1, = 40 and b = 0.12
(b); 0.1 (c); 0.08 (d).
3 Modied LE Model
Turing patterns observed in the CIMA reaction system provide the rst evidence that diusion induced instability
is likely to be a major mechanism for pattern formation. However, the CIMA reaction-diusion system only forms
automatically simple patterns such as hexagon and stripes. In order to gain complex patterns consisting of combinations
of basic hexagons, squares and stripes, the reaction-diusion system should be applied external periodic forcing.
No. 4 Turing Patterns in a Reaction-Diusion System 763
The CIMA reaction used in some experiments has revealed a sensitivity of this reaction to visible light, and open
the possibility of controlling Turing patterns by constant or periodic illumination.
[8]
When the light is periodically
switched on and o, the pattern suppression is observed in the corresponding well-stirred system.
[8]
When some masks
with the transmittance of masks determined by the sums of sinusoidal function are placed between the light source and
the reactor, families of complex superlattice structures were found in the photosensitive reaction-diusion system.
[9]
Light is often used to study eects of external perturbation on the dynamics of nonlinear reaction-diusion system.
We employ the two-variable model modied to include the eect of illumination mask (w). The modied Lengyel-
Epstein (LE) model is given by the equations
[810]
u
t
= a u 4
uv
1 + u
2
w + u,
v
t
=

u
uv
1 + u
2

+ w + dv

,
where u and v are the dimensionless concentrations of [I

] and [ClO

2
], respectively; a, b, d, and are the same as
mentioned above; w is dimensionless rate of the photochemical reaction, which is proportional to the light intensity.
Superlattices are highly symmetric, sometimes strikingly intricate structures composed of two or more simple com-
ponent patterns, such as hexagons and squares.
[10]
Figure 3 shows examples of superlattice Turing patterns composed
of squares. In our simulations, as parameter b decreases, some spots in some distorted circulars disappear, four new
squares are formed; as b continues to drop down, the other spots gradually eliminate. With varying in the interval
from 15 to 100, we can gain patterns analogous to those shown in Fig. 3, no signicant change takes place. Figure 4
shows superlattice hexagonal Turing patterns induced by external periodic forcing through the on-o hexagonal mask
in our numerical simulations. With parameter b decreased, Turing patterns are less stable. When > 15, we repeat
our simulations and obtain similar patterns, no dierences occur. Some ndings of our simulations are also in good
qualitative agreement with experiments shown in Ref. [9]. Most importantly, our simulations suggest that the stability
of superlattice structures depends on dynamical parameters and external periodic forcing.
Fig. 4 Superlattice hexagonal Turing patterns in the modi-
ed LE model. The rst column shows the (simulating) on-o
hexagonal masks, and the second column displays our simu-
lation results. In our simulations a = 12, b = 0.32, d = 1,
= 30. Period of masks n = 7 (a), 18 (b).
Fig. 5 The modulated stripes and parallel stripes in the
modied LE model. (a) The (simulating) stripes mask, period
n = 12. In our simulations, a = 12, d = 1, = 40, b = 0.02
(b), 0.12 (c), 0.33 (d).
Figure 5 displays patterns found for the stripes mask. Here most typical pattern formations are modulated stripes.
The modulated stripes in Fig. 5(c) are unstable if the period of mask n is too small; spots eventually evolve into the
whole system. Figure 5(d) shows stable parallel stripes. When we perform another set of simulations with dierent
, we can obtain patterns identical to that of Fig. 5. Only one dierence in Figs. 5(b) 5(d) is distance to Turing
bifurcation point in (b, ) phase plane. One plausible explanation is the interaction of Turing, Hopf instability and
external periodic forcing for the modulated stripes. Figure 6 shows oscillating Turing patterns produced in the system,
where Turing mode is subcritical and Hopf mode is supercritical, when this system is subjected to external periodic
764 WU Yan-Ning, WANG Ping-Jian, HOU Chun-Ju, LIU Chang-Song, and ZHU Zhen-Gang Vol. 45
forcing. One full period of oscillating Turing patterns is given in Figs. 6(b) 6(f), oscillating period of which is 2
time units. If we repeat our numerical experiments in dierent parameter b when < 12, we can observe analogous
oscillating patterns. However, in the Brusselator model dierent patterns were obtained with diusion coecient
D
v
varying.
[12,13]
The shape and type of Turing patterns seem to be independent of eective diusion rate in the
(modied) LE model. There is no signicant dierence between simulations with zero ux and period boundary
conditions as mentioned in Ref. [9].
Fig. 6 Oscillatory Turing patterns in the modied LE model. In our simulations a = 12, d = 1, b = 0.42 and = 11.
(a) The (simulating) sinusoidal squares mask, period n = 5. (b) (f) One full period of the oscillatory Turing patterns.
4 Conclusions
In real reaction-diusion systems, three-dimensional eect and imperfection of the system may have great inuence
on pattern formations; hence the (modied) LE model is an ideal model to explore mechanism of patterns. Although
it is dierence of diusion coecients that leads to Turing pattern formations, however, eective diusion rate has
less inuence on shape and type of Turing patterns in the (modied) LE model. In order to obtain complex Turing
patterns, we should chance appropriate dynamical parameters, type, and period of external periodic forcing. Interaction
of Turing, Hopf instability and inuence of external periodic forcing are responsible for the observed patterns. We
think our simulation results can provide additional insight into predicting new Turing patterns and understanding the
corresponding behavior of CIMA reaction-diusion system by external periodic forcing from temporal and spatial.
References
[1] A.M. Turing, Philos. Trans. R. Soc. London Ser. B 327
(1952) 37.
[2] L. Segel and J.L. Jackson, J. Theor. Biol. 37 (1972) 545.
[3] V. Castets, E. Dulos, J. Boissonade, and P. De Kepper,
Phys. Rev. Lett. 64 (1990) 2953.
[4] Arkady Rovinsky and Michael Menzinger, Phys. Rev. A
46 (1992) 6315.
[5] I. Lengyel and I.R. Epstein, Science 251 (1991) 650.
[6] T.K. Callahan and E. Knobloch, Physica D 132 (1999)
339.
[7] B. Rudovics, E. Barillot, P.W. Davies, E. Dulos, J. Bois-
sonade, and P. De Kepper, J. Phys. Chem. A 103 (1999)
1790.
[8] Milos Dolnik, Anatol M. Zhabotinsky, and Irving R. Ep-
stein, Phys. Rev. E 63 (2001) 026101.
[9] Igal Berenstein, Lingfa Yang, Milos Dolnik, Anatol M.
Zhabotinsky, and Irving R. Epstein, Phys. Rev. Lett. 91
(2003) 058302.
[10] Attila K. Horvath, Milos Dolnik, Alberto P. Munuzuri,
Anatol M. Zhabotinsky, and Irving R. Epstein, Phys. Rev.
Lett. 83 (1999) 2950.
[11] Stephen L. Judd and Mary Silber, Physica D 136 (2000)
45.
[12] Liu Yu-Fang, Wu Yan-Ning, Xu Hou-Ju, and Sun Jin-
Feng, Commun. Theor. Phys. (Beijing, China) 42 (2004)
637.
[13] Lingfa Yang, Anatol M. Zhabotinsky, and Irving R. Ep-
stein, Phys. Rev. Lett. 92 (2004) 198303.