Journal of Mathematics and Statistics 3 (3): 88-92, 2007

ISSN 1549-3644
© 2007 Science Publications

Asymptotic Behavior for Solution of Reaction-Diffusion Systems

1
M. Yahi and 2K. Saoudi
1
Scientific and Technical Research Center in Welding and Testing, 16 000. Algeria
2
Department of Mathematics, Khenchela University, 40 000. Algeria

Abstract: The existence, uniqueness, and asymptotic behavior of the solution of a balanced two-
component reaction-diffusion system have been investigated. It was shown that a global and
unique solution existed and its second component can be estimated using the Lyapunov
Functional. It was, also, demonstrated that each component of the solution converged, at infinity,
to a constant which can be found in terms of the reacting function and the initial data. The results
of the current research can be used in several areas of applied mathematics, especially when the
system equations originate from mathematical models of real systems such as in Biology,
Chemistry, Population Dynamics, and other disciplines.

Keywords: Reaction-Diffusion systems, Lyapunov Functional, Asymptotic Behavior, Global

Existence.

INTRODUCTION verifies
 f (0, s ) = 0 ∀s ≥ 0 (1.1)
This article deals with some aspects of the diagonal 
  ln(1 + f (r , s )) 
lim  ≤ α *∀r ≥ 0
reaction-diffusion system ( S1 )  s →∞  s 

 ∂u where
 ∂ t − d 1 ∆ u = g (u ,v ) in ]0, +∞[×Ω
8d1d 2 . (1.2)

 ∂ v − d ∆ v = f (u ,v )
α* =
 ∂ t 2 in ]0, +∞[×Ω n || u0 ||∞ (d1 − d 2 ) 2
 ∂u ∂v
 = =0 in ]0, +∞[×∂Ω
 ∂η ∂η Existence of a local solution in time: In the Banach
 u (0, x ) = u 0 ,v (0, x ) = v 0 in Ω.

 space C (Ω) we define the norm
After stating local existence and uniqueness, we see || u ||∞ = sup x ∈Ω | u (x ) | . The product space
some criteria for existence in the large (theorem 3.1
and 3.2) and we state the principal result (theorem X = C (Ω) × C (Ω) endowed with the norm
4.1) for the asymptotic behavior of the unique
|| (u ,v ) ||X =|| u ||∞ + || v ||∞ becomes a Banach
solution of ( S1 ) . This generalized results of [3],
space. If w (t ) = (u (t ),v (t )) , then we can convert
[4] and [5].
Notations and Assumptions: Therein, the system (S1 ) to an abstract first-order system in
∆ = ∑ k =1 ∂ denotes the Laplacian operator in the
n 2 the Banach space X of the form
k
 d
Euclidian coordinates, Ω is a domain (bounded  [ w(t )] = ∏ w(t ) + F [ w(t ) ] , t > 0
n
 dt
connected open set) of \ with smooth boundary  w(0) = w0 = (u0 , v0 ) ∈ X
∂Ω , ∂/∂η is the outer normal on ∂Ω , d 1 and
d 2 ∈ \∗+ . Two of the most fundamental assumptions [ ]
With F w(t ) = ( − f (u , v), f (u , v )) i.e. F is a
locally lipschitz function on the space X. The operator
associated with ( S1 ) are preservation of positivity
and conservation of total mass. So, we suppose that ∏ : D(∆) × D(∆) → X is defined by the matrix
the balance condition is satisfied which meant
f (u ,v ) + g (u ,v ) = 0 and the reacting term f is,
also, supposed to be a nonnegative and nonlinear
 d1∆ 0 
 
continuously differentiable function in \
2
+ which  0 d1∆ 
Corresponding Author: Yahi Mustefa, Scientific & Technical Research Center in Welding & Testing, 16000. Algeria
88
 J. Math. & Stat. 3 (3): 88-92, 2007

where under condition α = 0 , where ϕ is a continuously

∗

 ∂u 
D(∆) = u ∈ C (Ω) : ∆u ∈ C (Ω), = 0 . differentiable nonnegative function not polynomially
 ∂η  bounded but is required to grow less than
If d > 0 , then the operator d∆ generates an exponentially.
analytic semigroup of compact operators on the space All methods cited above (invariants regions,
maximum principle, comparisons ...) have some
C (Ω) . If S1 (t ) and S 2 (t ) are the analytic limitations. Recall that the boundedness of the first
semigroups generated, respectively, by d1∆ and component u of the solution is guaranteed by the
maximum principle which assert, in one of its
d 2 ∆ , on the space C (Ω) , then the operator ∏ version, that: If the component u (t , x ) verifies
generates an analytic semigroup of compact operators
 ∂u
on the space X which is given  ∂t − d1∆u ≤ 0 in ]0, Tmax [×Ω

by S (t ) = 1S (t ) 0  The existence of a classical  ∂u
 .  ≤ 0 in ]0, Tmax [×∂Ω
 0 S2 (t )   ∂η
local solution of (2.1) in a maximal interval [0, Tmax [
with initial value problem follows from standard Then u (x , t ) ≤ max u 0 (x ) =|| u 0 ||∞ .
x ∈Ω
arguments (cf. Henry D. [2], Lorenzi L. and all.
[7] and Pazy A. [9] ). We discuss now the boundedness of the second
component v of the solution. A Lyapunov functional
Existence of a global solution in-time: It may be L for the R-D system (S1 ) , is a continuously
surprising that question concerning the global + +
differentiable nonnegative function L : \ → \
existence of solution to the system ( S1 ) remain open
defined by L (t ) = L (u (t ,∗),v (t ,∗)) such that
until recently. If d 1 = d 2 : Nonnegative solutions of •
∂L [u (t ,∗),v (t ,∗)]
( S1 ) exist globally in-time. In fact, if we add the two L (t ) = ∂t
≤ 0 for all solution
equations of ( S1 ) we get the heat equation: [u (t ,∗),v (t ,∗)] of the system (S1 ) .
∂ Our main result, in this section, is:
(u + v ) − d 1∆ (u + v ) = 0. Then, by the use of Theorem 3.1: Suppose that (u (t ,∗);v (t ,∗)) is a
∂t
the balanced condition together with the maximum solution of the system (S1 ) , then
∞
principle to obtain the L -estimation of u and v
t → L (t ) = ∫ [M − u (t , x )]−γ exp[ β .v (t , x )]dx
since || (u + v )(t ) ||∞,Ω ≤|| u 0 + v 0 ||∞,Ω . So, global Ω

existence of solution holds. If d 1 ≠ d 2 : The situation is a decreasing functional in the interval [0,T ∗ [ iff
is quite delicate. With the homogeneous Neumann
[1] the constants β ,γ ≥ 0 are such that
boundary conditions, Alikakos N. D. established
4d1 .d 2
the global existence via the embedding theorem of βM <γ < (3.1)
Sobolev and L∞ -bounds of solutions for positive (d1 − d 2 ) 2
initial data when f (u ,v ) = −uv β and Where M verify || u 0 ||∞ < M .
1 < β < n+n 2 . Masuda K. [8] Showed that the global
Proof: Differentiating the functional L as a function
solutions of ( S1 ) exist for β > 1 and converge to
of t and using the two equations of the system (S1 )
constant when t → +∞ . If the nonlinearity f is we obtain
polynomially bounded, then it is known that the •
∂
solutions to ( S1 ) exist and remain uniformly bounded L (t ) = ∫ω ∂t ([M − u (t , x )]−γ exp[ β .v (t , x )]) dx = I + J .
where
in the L∞ (Ω) -norm. Analogous result have been
I = ∫ [γ d 1 (M − u )−γ −1e βv ∆u + β d 2 (M − u )−γ e βv ∆v ]dx
[3] ω
also established by Haraux A. and Youkana A. ,
and generalized their result to handle the J = ∫ [ β (M − u )−γ − γ (M − u )−γ −1 ].f .e βv dx .
ω

nonlinearities of the function f (u ,v ) = u ϕ (v ) To compute I , we use a Green formula to obtain

89
 J. Math. & Stat. 3 (3): 88-92, 2007

I = − ∫ Q (∇u ,∇v )(M − u ) −γ − 2 e βv dx (b 2 − 4ac ) γ (d 1d 2 − γ (d 1 − d 2 ) 2 )

> = m1 > 0.
Ω
8c 2d 2 β 2 M 2
Where
Q (∇u , ∇v ) = γ d 1 (γ + 1) | ∇u |2 + βγ (d 1 + d 2 )( M − u )∇u ∇v + β 2d 2 (M − u ) 2 | ∇v |2 . Then
I ≤ − ∫ [m1 | ∇u |2 + m 2 | ∇v |2 ](M − u )−γ −1e βv dx < 0.
Ω
We deduce that Q is a quadratic form in ∇u and To show that J ≤ 0 , use γ > βM and the
∇v i.e. Q (x , y ) = a11x 2 + 2a12 xy + a22 y 2 . If maximum principle to obtain M −u ≤ M .
I ≤ 0 then Q will be a nonnegative quadratic form Consequently

∫
− ( γ +1)
all principal determinants of its symmetric matrix J ≤ −(γ − β M ).M f (u ,v ).e βv dx .
Ω

βγ (d 1 + d 2 )(M − u ) 

d 1γ (γ + 1) 
− ( γ +1)

 2

 Now, let C = (γ − β M ).M be a nonnegative
 
AQ = constant, then J ≤ 0 . Finally, we have showed
 
 
 
 βγ (d 1 + d 2 )(M − u ) 

 β d 2 (M − u )
2 2 
 •

2 
 
L (t ) ≤ −∫ Ω
[m1 | ∇u |2 + m 2 | ∇v |2 ](M − u )−γ −1e βv dx − C ∫ f (u ,v ).e βv dx ≤ 0.
Ω
associated to Q will be nonnegative. The first
principal determinant of AQ is d 1γ (γ + 1) ≥ 0 , the DISCUSSION

second is | AQ | which is nonnegative if We have established that L (t ) ≤ k for some

4d 1d 2 .If J ≤0 constant k and
γ≤ then
(d 1 − d 2 ) 2 hence (M − u ) e
−γ βv ∞∗
∈ L ([0,T [, L (Ω)) . Since
1

β ( M − u ) −γ − γ ( M − u ) −γ −1 ≤ 0, (M − u ) −γ ≥ M −γ
we deduce that
i.e. β ( M − u ) − γ ≤ 0. Since we have 0 < u < M βv ∞
e ∈ L ([0, T [, L (Ω)) ∗ 1
(3.2)
for all u then we obtain γ > β M . Thus, we have From Smoller J. [10], it is obvious that the region E
established that if I ≤ 0 and J ≤ 0 then (3.1) is is invariant for the system ( S1 ). This ensures the
satisfied. Now suppose that (3.1) is satisfied. First of
stability for solution to be nonnegative for all time.
all, recall that the Young inequality for a quadratic
form, for
∗
a ∈ \ , b and c ∈ \ , has the form ∗ Since the reaction f is continuous in \ 2+ and by the
+ −
maximum principle, 0 < u < M , it follow that for
(b2 −4ac)  x2 y2 
ax + bxy + cy ≤ −  4c + 4 a  , r ∈ [0, M ] we can choose α ∈ \ + to have,
2 2
for all
2
 
x, y ∈ \ . To show that I ≤ 0 , we observe that from (1.1), lim  log(1 + f ( r , s ))  < α < α ∗ . Then,
s →+∞  
all
 s
I = ∫ [a | ∇ |2 b ∇u .∇v + c | ∇v |2 ](M − u ) −γ − 2 e βv dx γ ∈ \ + so that f (r , s ) ≤ γ e α s , for all
there is
Ω

Where
s ≥ 0 and r ∈ [0, M ] . Now, from α < α ∗ , we
a = −d 1γ (γ + 1) , n 4d 1d 2
have α< . Choose p ∈ ` +∗
b = − βγ (d 1 + d 2 )(M − u ) and 2 || u 0 ||∞ (d 1 − d 2 ) 2
such that p > n / 2 and sufficiently close to n / 2
c = − β 2d 2 (M − u ) 2 .
4d 1d 2
By the Young inequality such that pα < . Put β = pα
(b 2 − 4ac )  | ∇v |2 | ∇u |2  || u 0 ||∞ (d 1 − d 2 ) 2
a | ∇ |2 +b ∇u .∇v + c | ∇v |2 ≤ −  + .
2  4a 4c  4d 1d 2
to obtain β || u 0 ||∞ < . Now, from (3.2)
We have d 1d 2 − γ (d 1 − d 2 ) 2 > 0 since γ > γ/ 4 . (d 1 − d 2 ) 2
Then we obtain
and since β = pα and f (u ,v ) ≤ γ e , we
αv
(b − 4ac ) d 1d 2 − γ (d 1 − d 2 )
2 2
> = m 2 > 0. deduce that:
8a 2d 1 (γ + 1) The p -norm || f (u ,v ) || p is uniformly estimated on
In the same way, since 0 < u < M for all u
(Maximum Principle), we obtain:
[0,T ∗ [ for p > n / 2 .

90
 J. Math. & Stat. 3 (3): 88-92, 2007

[2] 2 2
Following Henry D. , one has ∂u ∂v
Theorem 3.2: If f ∈ C ∞ (\ 2+ ) is a nonnegative
∫K ∂t dxdt < +∞ and ∫K ∂t dxdt < +∞.
continuously differentiable function such that Proof: Let F = [0, +∞[×∂Ω . Multiply the first
f (0, s ) = 0 for all s ≥ 0 and satisfying the ∂u
equation of the system ( S1 ) by and integrate on
condition (1.1) then all solutions of ( S1 ) with initial ∂t
data in the region E , are global in time and K , then apply the Green formula
uniformly bounded in [0, +∞[×Ω.
2
∂u  ∂u  ∂u ∂u
∫
K ∂t
dxdt + d 1 ∫ ∇ 
K  ∇udxdt − d 1 ∫F
 ∂t  ∂t
∇udxdt = − ∫
K ∂t
f (u ,v )dxdt .

Asymptotic Behavior for solution of Reaction- Then

diffusion system: This section deals with behavior 2
∂u d ∂u
for solutions when t → +∞ of the system ( S1 ) . ∫ dxdt + 1  ∫ | ∇u |2 dx − ∫ | ∇u 0 |2 dx  = − ∫ f (u ,v )dxdt
K ∂t 2 Ω Ω  K ∂t

With the help of boundedness of paths in the Banach

space X = C (Ω) × C (Ω) , the asymptotic behavior and hence
2
for solutions can be analyzed by application of the ∂u d ∂u
[6] ∫ K ∂t
dxdt ≤ 1 ∫ | ∇u 0 |2 dx + ∫
2 Ω K ∂t
f (u ,v ) dxdt .
“Invariance Principle” and a result on paths
precompacties which asserts that: “if u and v are
solutions of ( S1 ) and if f ∈ L ([0, +∞),C (Ω))
∞ | k |2 + | h |2
Now, apply the inequality ≥| kh | , and
∞ [4] 2
then u and v ∈ L ([0, +∞),C (Ω)) ” .
∂u
put k = and h = f (u ,v ) , one gets
The Main result, of this section, is the following: ∂t
Theorem 4.1: Let (u ,v ) a global solution of the ∂u
2

system ( S1 ). There is a nonnegative ∫ K ∂t

dxdt ≤ d 1 ∫ | ∇u 0 |2 dx + ∫ | f (u ,v ) |2 dxdt .
Ω K

constants c1 and c2 such that Since u 0 ∈ L∞ (Ω) and f ∈ L∞ ([0,T ∗ ), L∞ (Ω))

t →∞ t →∞ 2
|| u − c1 ||∞ → 0 and || v − c 2 ||∞ → 0. ∂u
we deduce ∫
K ∂t
dxdt < ∞ and in the same way,

Furthermore, we have f (c1 , c 2 ) = 0 and ∂v

2

1 we deduce that ∫ ∂t
dxdt < ∞. This prove the
| Ω | ∫Ω
K
c1 + c 2 = (u 0 + v 0 )dx .
lemma.
To establish this result, we need the followings two According to the lemma 1, put
lemmas. 2
∂u
Lemma 1: Let ϕ ∈ L1 (\ + ) and ϕ (t ) ≥ 0 for all ϕ : t 6 ϕ (t ) = ∫ dx to obtain, by lemma 2,
∂t Ω
t ≥ 0 . Then lim ϕ (t ) = 0.
t →+∞ ∂u (t , x ) ∂v (t , x )
Proof: Suppose that we have lim ϕ (t ) = c for
lim = lim = 0. (4.1)
t →+∞
t →+∞ ∂t t →+∞ ∂t
Now, we prove the main result of theorem 4.1. With
some constant c . Then for all ε > 0 there is tε the help of the two precedent lemmas.
such that for t > tε we have ϕ (t ) − c ≥ −ε . Put
Proof of theorem 4.1: The paths {u (t ), t ≥ 0} and
c c
ε= , to obtain ϕ (t ) ≥ and {v (t ), t ≥ 0} associated to the system ( S1 ) are
2 2
precompact in the Banach space C(Ω) × C(Ω)
+∞ +∞ +∞ c [4]
.
∫0
ϕ (t )dt ≥ ∫ ϕ (t )dt ≥ ∫
tε 0 2
dt = +∞ i .e . ϕ ∉ L1 (\ + ).
Pick a sequence (t n ) n ∈ which tend to infinity when
This contradicts the hypothesis.
Lemma 2: Let K = [0, +∞[×Ω and (u ,v ) be a t → +∞ such that lim u (t n ,∗) = u 1 and
t →+∞
global solution of ( S1 ), then
91
 J. Math. & Stat. 3 (3): 88-92, 2007

lim v (t n ,∗) = v 1 in C(Ω) . Since the function f 1 1

t →+∞
|Ω| ∫Ω
u (t , x)dx and
| Ω | ∫Ω
v(t , x)dx converge
is continuous, we obtain
in time and from (4.2) we deduce
lim f (u (t n , x )),v ((t n , x )) = f (u1 ,v 1 ) 1 1
t →+∞
| Ω | ∫Ω | Ω | ∫Ω
c1 + c2 = (u (t , x ) + v (t , x )) dx = (u0 + v0 ) dx.
in C(Ω) . By (4.1), the system ( S1 ) becomes at the
Conclusion. The results of this study have given rise
infinity, a system (S∞ ) of the form to an open similar problem of the existence,
uniqueness and asymptotic behavior of the solution of
general balanced n-diagonal (or n-tridiagonal)
− d1∆u = − f (u , v) in Ω reaction-diffusion systems, which can be thought of
 − d ∆v = f (u, v) Ω
 2 in as further research.

 ∂u ∂v Acknowledgment. The author would like to thank the
= =0 in ∂Ω.
 ∂η ∂η anonymous referee for his valuable suggestions.
Multiply the first equation of (S∞ ) by u1 and REFERENCES
integrate on Ω we obtain d1 ∫ u1.∆u1dx ≤ 0. The p
Ω
1. Alikakos N. D., 1979. L -bounds for solutions
Green Formula gives us d1 ∫ | ∇u1 | dx = 0 , then
2
of reaction diffusion equations. Comm. Partial
Ω
Differential Equations, 4: 827-868.
∇u1 = 0 and hence u1 = c1 where c1 is a constant. 2. Henry D., 1981. Geometry theory of semilinear
parabolic equations. Lect. Notes in Math.,
On the other hand, the function v1 verifies the
Springer-Verlag, New York, 840.
homogenuous Neumann Problem 3. Haraux A., and Youkana A., 1988. On a result
of K. Masuda concerning Reaction-Diffusion
 ∆v1 = 0 in Ω Equations. Tôhoku Math. J., 40: 159-163.
 4. Kirane M., and Haraux A., 1983. Estimation
 ∂v1
 ∂η = 0 in ∂Ω. C 1 pour des problèmes paraboliques semi-
 lineaires. Annal. Fac. des Sc. Toulouse, 5: 265-
It’s solution is v1 = c2 where c2 is a constant. Since 280.
5. Kouachi S., and Youkana A., 2001. Global
2
the function f is continuous on \ + , at infinity, we existence for a class of reaction-diffusion
systems. Bull. of Polish Academy of Sciences,
obtain f (c1 , c2 ) = 0 . The sequence (tn ) n∈` is 49 No 3: 303-308
arbitrary, then || u − c1 ||∞ → 0 and || v − c2 ||∞ → 0 6. La Salle J. P., 1976. Stability theory and
when t → +∞ . Since the inclusion invariance principle. Dynamical systems, An
International Symposium. Academic Press,
C (Ω) → L (Ω) is continuous we deduce that
1
New-York, Vol. I: 211-222
7. Lorenzi L., Lunardi A., Metafune G. and Diego
P., 2005. Analytic Semigroups and Reaction-
∫Ω
(u (t , x) + v(t , x)) dx → | Ω | (c1 + c2 ) Diffusion Problems. Internet Seminar: 1-108.
(4.2) 8. Masuda K., 1983. On Global existence and
Asymptotic behavior of solutions of reaction-
The convergence in L (Ω) is obvious since the first
1
diffusion equations. Hokkaido Math. J., 12:
and the second equation of ( S1 ) yields 360-370
9. Pazy A., 1983. Semigroups of Linear Operators
∂ ∂ and Applications to Partial Differential
∂t ∫Ω
u (t , x)dx ≤ 0 and
∂t ∫Ω
v(t , x)dx ≥ 0 then
Equations. Springer-Verlag, New York.
10. Smoller J., 1983. Shock Waves and Reaction-
∫ u (t , x)dx is a decreasing function in t and
Ω Diffusion Equations. Springer-Verlag, Berlin.

∫Ω
v (t , x)dx is an increasing function in t . But u
is nonnegative and v is bounded, then

92