ROBUST EXPONENTIAL ATTRACTORS FOR MEMORY RELAXATION
OF PATTERN FORMATION EQUATIONS
WANG Yuwei, LIU Yongfeng & MA Qiaozhen*
College of Mathematics and Information Science, Northwest Normal University, Lanzhou Gansu 730070, China. *Email: maqzh@nwnu.edu.cn.
ABSTRACT
In this paper, we prove the existence of the robust exponential attractors for memory relaxation of pattern formation
equations in the phase-space
0
,
H
and we improve results in [3].
Key words: Pattern formation equations; Robust exponential attractors.
1. INTRODUCTION
Let
3 3
[
l l
, ]
¡
be a bounded domain with smooth boundary ,we consider the pattern formation equations2
( )
0
t
u
u
u
u
g u
(1.1) subject to the Dirichlet boundary condition
( ) |
0
u t
,
t
0
.
We assume
2
,
g
C
¡
with
g
0
0,
such that2
( )
(1
).
g r
c
r
r
¡
,
(1.2)
1
( )
lim inf
r
g r
r
, (1.3)
where
1is the first eigenvalue of
with Dirichlet boundary condition.According to [6] the long-term dynamics between the abstract first order evolution equation
u
t
Au
B u
0
(1.4) and the memory relaxation of (1.4)(1.5)
are close when
is sufficiently small. Hereu
u t
: [0
,
]
H
, and be a Hilbert space
:
A D A
H
H
is a strictly positive self-adjoint linear operator with compact inverse
1 2
:
(
)
B D A
H
H
is a nonlinear operator. As detailed in [12] there is a complete equivalence between (1.5) and the following system
(1.6)
In this paper, we set
2
1,
A
with the domain
4
2
0D A
H
H
and
0
[
]
0
t
u
s
Au t
s
B u t
s
ds
0
0,
[ ].
tt
u
s
s ds
T
Au
B u
,
B u
g u
then the memory relaxation of equation(1.1) is equivalent to the following evolution system
0
1 2
0,
.
tt
u
s
s ds
T
Au
A u
u
g u
(1.7)The pattern formation equation portrays the chemical reaction and the flame combustion phenomenon (see [1.2] for detail). Existence of the global attractors for the equation (1.1) has been studied by A. ION in [3]. The global attractor is compact, fully invariant and attractive for the semigroup, but may attractor trajectories at a slow rate. Conversely, an exponential attractor is a compact set, positively invariant under the action of the semigroup, which attracts exponentially fast trajectories departing from the bounded sets. Thus exponential attractors are expected to be more robust than global attractors under perturbations. In this paper, we study the existence of the robust exponential attractor for (1.7) in the phase-space
H
0.
It is worth noting that the exponential attractors attract bounded subsets of the whole phase-space. The main results of this paper is Theorem 3.8.2. PRELIMINARIES
Let
,
and
denote the inner product and the norm inH
L
2, respectively. Forr
¡
,
the hierarchy of compactly nested Hilbert spaces1 ( ) / 2
2
(
r)
r
V
D A
with the standard inner products1 1
( ) ( )
2 2
2 2
1
,
2 r 1,
2.
r r
V
u u
A
u A
u
Notice that
H
V
12.
Identifying H with its dual spaceH
,
there holds
V
r
V
r 1.
Denoting
¡
(0, ),
we assume that
W
1,1
¡
is nonnegative ,and that the exponential decay condition
s
s
0,
fora e
. .
s
¡
(2.1)holds for some
0.
Furthermore ,for
0,1 ,
we set
s
1
2
s
,
then we have
0
1
,
s ds
0
s
s ds
1.
(2.2) Then, we introduce the weighted Hilbert spacesM
r
L
2
¡
;
V
r1
,
endowed with the inner products
11
,
2 Mr 0s
1s
,
2s
Vrds
.
We make use of the infinitesimal generator of the right-translation semigroup on
M
0,
the linear OperatorT
s
s
(
s
being the distributional derivative with respect tos
) with domain
0 0
:
s,
0
0 .
D T
M
M
On account of (2.1) , there holds (see [6,12])0 0
2
,
,
2
M M
T
D T
.
(2.3) For
0,1
,we define the product Banach spaces,
0,
0.
r r
r
r
V
M
if
H
V
if
(2.4).
=
)
,
(
,0),
(
=
u
u
Q
u
L
We recall the compact and dense injections in [10]
D A
(
s2)
↪D A
(
r2),
s
r
,
(2.5)and the continuous embedding
2
(
s)
D A
↪L
6 3 2 s,
[0, )
3
2
s
. (2.6)If
M
r,
then from
1.7
1, we have the following inequality1
2
1
2.
r r
t V M
u
(2.7)For
i
0,1
andR
0
we set
{
:
r}.
i r
H
B R
z
H
z
R
Definition 2.1 [6] A family
{ }
0,1 of compact subsets ofH
0 is said to be a robust family of exponential attractors if the following conditions holds
(i) Each
is positively invariant forS t
.
(ii) There exist
0
and a positive increasing function M such that, for everyR
0
there holds
0
0
,
t.
H
dist
S t B
R
M R e
(iii) The fractal dimension of
inH
0 is uniformly bounded with respect to
.(iv) There exists a continuous increasing function Θ : [0; 1] → [0;∞) with Θ(0) = 0 such that
0
( ,
0)
( ).
sym H
dist
L
where dist and
dist
symdenote the nonsymmetric and symmetric Hausdorff distance between sets respect- tively. For a fixed
, (i) − (iii) are nothing but the usual conditions defining an exponential attractor (but observe that, contrary to the original definition see [9] for detail) we require the attraction property on the whole phase-space), while condition (iv) characterizes the robustness property.Given
z
( ,
u
0
0)
H
0(
z
u
0 if
0),
we denote the weak solution at time t to (1.7) with initial dataz
as
( ( ),
),
0,
( ),
0.
t
u t
if
S t z
u t
if
(2.8)Conditions on B. We assume that
B V
:
0
V
1.
For every R ≥ 0 ,there exists C = C(R) ≥ 0 such that
1 01
1 2 1 2
sup
,
i V
V V
u R
B u
B u
C u
u
P P
P
P
(2.9)
0 1sup
.
V
V
u R
B u
C
P P
(2.10)
Assumption (S). For every
[0,1]
system (1.7) generates a strongly continuous semigroupS t
on the phase-spaceH
0 Moreover, for every given R ≥ 0 ,there exists K = K(R) ≥ 0 such that
1
2 0 1 2 0,
Kt
H H
S t z
S t z
Ke
z
z
(2.11) Wheneverz
i H0R
.
Theorem 2.1 [6] Let the following assumptions hold.
( 1)
H
fori
0,1,
there existsR
i
0
such thatB R
i
i is an absorbing set forS t
onH
i,
uniformly with respect to
.Namely, given any R ≥ 0, there exists a timet
i
0
,depending only on R ,such that
i i
i,
S t
R
R
t
t
i.
Moreover, for every R ≥ 0 ,there existsC
i
C R
i
0,
such that
sup
i,
i H
i H
z R
S
t z
C
P P
t
0.
(
H
2
) There existR
2
0
such that
1
R
2 satisfies
0
0 1
0
,
2.
t H
dist
S t
R
R
Me
for some
M
0
and
0
.(
H
3
)Given any R > 0 there exist
0,1
and
,
: 0,
¡
(possibly depending on R) with
1
2
t
for t large enough such that for every
z z
1, 2
1
r
,
the mapS t
admits the decomposition
1
2
1,
2
1,
2
,
S t z
S t z
L t
z z
N
t
z z
Where
L t
andN
t
satisfy
1,
2
H0
1 2 H0,
L t
z z
t z
z
N
t
z z
1,
2
H
t z
1
z
2 H0.
Besides
t
¤
t N
t
z z
1,
2
fulfils the Cauchy problem0
,
0,
0,
t
T
t
for some
w
satisfying, for all T > 0,
V1
1 2 H0,
w t
t z
z
t
(0, ).
T
Then,there exists a family
of robust exponential attractors.Remark 2.2 Condition (H3) with
1
is actually implied by the stronger condition (H4) For every R ≥ 0, there exists C = C(R) ≥ 0 ,such that
0 0 11 2 1 2
sup
.
i V
V V
u R
B u
B u
C u
u
P P
Lemma 2.3[4] Let X be a Banach space ,and let
¢
C
¡
,
X
Let E : X
¡
be a function such that
sup
tE z t
m
¡
and
E z
0
M
,for some m,M ≥ 0 and every z ∈
¢
. In addition, assume that for every z
¢
the function ta
E z t
is continuously differentiable and satisfies the differential inequality
2,
Xd
E z t
z t
k
dt
For some
0
andk
0
both independent of z
¢
then
2
sup
:
2
,
X X
E z t
E
k
t
m
M
.
k
Lemma 2.4 [5] Let
be an absolutely continuous positive function on¡
which satisfies for some
0
the differential inequality
2
( ) ( )
,
d
t
t
g t
t
h t
dt
for almost everyt
¡
,
where g and h are functions on¡
,
such that
1
1
,
t
g y dy
m
t
t
0,
for some
m
1
0
and
0,1 ,
and
1
2 0
sup
t,
t t
h y dy
m
for some
m
2
0.
Then
t
0
e
t
,
t
¡
,
for some
m
1,
1
and
m e
2 / 1
e
.
Lemma 2.5 [7] Let
1,
2,
3be subsets of X such that
11
,
2 1,
v t X
dist
S t
L e
22
,
3 2,
v t X
dist
S t
L e
for some
v v
1,
2
0
andL L
1,
2
0
. Assume also that for allz z
1,
2
t0S t K
( )
j(
j
1, 2,3)
there holds0
1 2 0 1 2
( )
( )
v tX X
S t z
S t z
L e
z
z
for some
v
0
0
and someL
0
0.
Then it follows that
1,
3
,
vt X
dist
S t
Le
where 1 20 1 2
v v
v
v
v
v
andL
L L
0 1
L
2.
3. ROBUST EXPONENTIAL ATTRACTORS IN
H
0We need to verify conditions on B, assumptions (S)
,
and (H1),
(H2),
(H3) in theorem2.1.
In fact,
B[u] = g(u) is easily seen to fulfil conditions(2.9),(2.10)
and (H4). By means of the Galerkin schemeadapted to systems with memory (see [8] for detail)
,
one can show that,
for every
0,1 ,
system(1.7)
generates a strongly continuous semigroupS t
on the phase-spaceH
0.
Lemma 3.1 For every given R ≥ 0 ,there exist K = K(R) ≥ 0 such that
0 0
1 2 1 2
( )
( )
H Kt H,
S t z
S t z
Ke
z
z
(3.1)Whenever
z
i H0R
.
Proof : Set
u
%
u
1u
2, ( )
s
1( )
s
2( ),
s
then we get
).
(
)
(
~
~
~
~
=
~
0,
=
d
)
(
~
)
(
~
2 1
1/2 0
u
g
u
g
u
u
A
u
A
T
s
s
s
u
t t
(3.2)
Multiplying
(3.2)
1byAu
%
in 1V
and(3.2)
2by n% in 0,
M
we obtain.
~
),
(
)
(
2
~
,
~
2
~
,
~
2
=
~
,
~
2
)
~
~
(
d
d
0 2 1
0 0
1/2 0
2 0 2
0
M M
M M
M
P
P
P
P
u
T
A
u
u
g
u
g
u
t
VBecause of
(1.2),(2.3)
and(2.5),
we have the following estimates),
~
~
(
d
~
)
(
d
~
)
(
~
,
~
2
21 20 200 2 1 1/2 0 0 1/2
MM
P
P
P
P
P
P
P
P
V V
V
s
s
s
c
u
u
A
s
u
A
2
~
,
~
(
~
~
20),
2 0 0
M M
P
P
P
P
Vu
c
u
1 1 2 1 1 2 1~
)
(
)
(
~
),
(
)
(
2
V V
V
c
g
u
g
u
u
g
u
g
P
P
P
P
)
~
)
(
)
(
(
1
2 21
21
V Vu
g
u
g
c
P
P
P
P
),
~
~
(
212 1
V Vu
c
P
P
P
P
2
(
)
(
),
~
(
~
~
20),
2 0 0 2 1
M M
P
P
P
P
Vu
c
u
g
u
g
Hence).
~
~
(
)
~
~
(
d
d
2 0 2 0 2 0 2 0
M MP
P
P
P
P
P
P
P
V
V
c
u
u
t
(3.4) By Gronwall lemma ,we complete the proof.Lemma 3.2 Condition (H1) holds for i = 0.
Proof : For any given
z
( , )
u v
H
0,
we consider the functional0 1 1 1 0
2
2 2 2
( )
H V2
( ),1
V2
,
MV
E z
z
u
u
G u
u
,where G(r) =
0
( )
,
r
g
d
for some
0
which will be chosen small enough so that the following estimates hold. From(1.3)
and(2.2),
there exists
(0,1)
such that1 0
2
( ),
V(1
)
Vg u u
u
c
,
(3.5)1 0
2
2
G u
( ),1
V
(1
)
u
V
c
,
(3.6)
0 0 0
2
1
22
,
.
4
2
M
u
V M
(3.7)The last two inequalities, together with
(1.2),
yield0 0
2 3
( )
(1
).
2
Hc
E z
c
z
H
(3.8)We now fix
z
H
0 withz
H0R
,
and we denote( ( ),
)
( ) .
t
u t
S t z
Multiplying(1.7)
1by1
2
( )
Au
A u
u
g u
inV
1and(1.7)
2by
inM
0,
we obtain1 1 1 0 0
1 1
2
,
,
,
, ( )
,
,
0.
t V t t V t V t M M
V
u Au
u A u
u u
u g u
u
T
Hence ,we get
1
0 0 1 1 1
2
2 2 2 2 2
0
(
V M V2
(( ),1
V)
( )
( )
V0
V
d
u
u
u
G u
u
s
s
ds
dt
Besides, recalling
(2.2)
,
and multiplying(1.7)
2byu
inM
0,
it is straightforward to obtain1
0 0 0 0 1 1
2
2 2 2
1
1
1
1
,
M t,
M,
M V V( ),
V.
V
d
u
u
T
u
u
u
u
g u u
dt
(3.10)Therefore,
E
E S t z
(
( ) )
satisfies1 1 0
2 2 2 2
1
0
( )
V2
V2
22
Vd
E
s
ds
u
u
u
dt
1 1
1 2
0 0
2
( ),
V2
( )
, ( )
V2
( ) ( )
Vg u u
s u
s
ds
s
s
(3.11)Using
(3.5) (3.7),
we obtain0 1
2 2
0
( (
( ) ))
2
V( )
( )
V
d
E S t z
u
s
s
ds
dt
1 1
2 2
1
0 2
2
2
2
( )
, ( )
V V V
u
u
s u
s
ds
1 2 0
2
( ) ( )
.
V
s
s
c
(3.12)Because of
(1.7)
1and(2.7),
the terms on the right-hand side can be controlled as1 0
2 2 2
1 2
2
2
2
,
3
V V V
u
u
u
1 0 1
2 2
0 0
1
2
( )
, ( )
( )
( )
,
3
V2
VV
s u
s
ds
u
s
s
ds
c
0 1
2
2 0
2
( ) ( )
.
4
MV
s
s
c
By the force of
(2.1),
and choose
small enough we get0 0
2 2
(
( ) )
( )
.
8
H M
d
E S
t z
S t z
c
dt
(3.13)In view of
(3.8),
from lemma2.3,
there existst
0
t R
0( )
such that0 0
2
(
( ) )
sup{ ( ) :
},
H HE S t z
E
c
t
t
0.
Using
(3.8),
and subsequently integrating(3.13)
on(0, ),
t
0 we meet the claim.Remark 3.3 In view of
(2.7),
and integrating(3.13)
with
0
on¡
,
we find the following Integral estimate1 2
0
0 t
( )
V,
u y dy
(3.14) for some
0 0( )
R
0.
Lemma 3.4 Under the above assumptions ,there exist
t
0
0
such that1 2
1
2
,
V
Proof : Multiply (1.1) with Au in
V
12, we have 21 2 1 1
2 2 2
1
( ),
,
.
2
d
A u
Au
A u
g u Au
A u Au
dt
(3.15)we have the following controls
2
2 1 2
2
( ),
( )
(
( )
)
( )
,
4
Au
g u Au
Au g u dx
Au
g u dx
g u dx
P
P
2 2
1 1 1 1
2
,
2 2 2.
4
Au
A u Au
A u Au dx
Au A u
A u
Hence we have
2 2
1 2
1
( )
,
2
d
A u
g u dx
dt
(3.16)because of
(1.2),
we get2 6
( )
(1
)
g u dx
C
u
dx
C
C
66
.
Lu
We set
A
~
=
subject to the Dirichlet boundary condition then we have),
~
(
=
)
(
)
~
(
=
)
(
)
(
=
)
~
(
3 2 01 3
2 0 2
0
H
A
L
A
H
A
D
D
D
Lemma 3.5 There exist
0
and an increasing function
such that, ifz
HR
for some
0.1 ,
then0
( )
t(
)
H H
S t z
M e
z
And
1
1
2 2
( )
1 (
)
t
t V
u y
dy
t
For some
M
( )
R
0
and
( )
R
0.
Proof :For
0
small enough we introduce the energy functional1 1
2 2 2
1 2
( )
( )
H V2
( ),
V2
,
M.
E t
S t z
u
u
g u u
u
c
Choose the constant
c
appearing inE t
( )
large enough it is apparent that2 2
1
( )
( )
2
( )
2
S t z
M
E t
S t z
H
c
. (3.17)Multiplying
(1.7)
1 by1
2
( )
Au
A u
u
g u
inV
1, 2
inM
and(1.7)
2 by
inM
, 2
in,
M
this yields1
2 2
0
( )
( )
( )
V2
Vd
E t
s
s
ds
u
dt
1 2 2
2
Vu
1 12
2
u
V
2
g u u
( ),
V1
1 2 0
2
( ) ,
t V2
( ) ( )
Vg u u u
s
s
1 0
2
( )
, ( )
.
V
s u
s
ds
Arguing as in the previous proof and taking into account lemma
3.2,
we have1 1
2
0 0
2
( ) ( )
2
( )
, ( )
V V
s
s
s u
s
ds
1
2 2
0
1
2
( )
( )
.
2
M
s
s
Vc
(3.19)Hence ,we obtain
1 2 0
1
( )
( )
( )
( )
2
Vd
E t
E t
s
s
ds
dt
1
2 2
2
g u u u
( ) ,
t V2
M2
u
,
Mc
.
(3.20)Moreover, using the embedding
1 2 2
(
)
V
D A
↪
3
1
( )
L
. we have the controls3 3
1 1 2 3
2 1 2
2
( ) ,
t V(1
L)
t LL
g u u u
c
u
u
A
u
1 1
( )
2 2
(1
V)
t Vc
u
u
A
u
P
P
1
(1
V)
t V Vc
u
u
u
P P P P
(3.21) Exploiting(2.1)
and0
2 2
2
u
,
M Mc
,
provided that
is small enough we end up with0 1 1
2
8
M t V t Vd
E
E
c
c u
c u
E
dt
P
P
P
P
. (3.22)Because of
(3.14),
we apply lemma2.4
to get( 2)
( )
(0)
t.
E t
cE
e
c
Finally, integrating the differential inequality
(3.20)
with
0
on( , ),
t
and using(2.7)
, we complete the proof.Remark 3.6 In particular, for
1,
settingR
1
2 (
R
0)
andC
1
M
1
(
C
0),
we see that condition (H1) issatisfied for i = 1 as well.
Remark 3.7 From the estimates of (3.19) we know that the sobolev embedding is the maximal but how to control a higher growth condition than (1.2) is open.
In order to prove (H2) analogously to what observed in [11] a function
g
C R
2( )
such that(1.2),(1.3)
admits a decompositiong
g
0
g
1 satisfying0
( )
(1
),
g r
c
r
g
0
(0)
0,
g r r
0( )
0,
g r
1
( )
c
.
r
¡
(3.23)Next, for any fixed
z
B R
0(
0),
we make the decomposition( )
( )
( )
S t z
D t z
K t z
0
1 2
0
( ) ( )
0,
( ),
(0)
t
t
v
s
s ds
T
Av
A v
v
g v
D
z
z
(3.24)
and
0 1
2
0
( ) ( )
0,
( )
( ),
(0)
0
t
t
w
s
s ds
T
Aw
A w
w
g u
g v
K
z
(3.25)
By a slight modification of lemma 3.2, it can be shown that there exist
M
0
0
and
0
0
such that 00 0
( )
H t,
D t z
M e
(3.26)Lemma 3.8 Given
[0, ],
3
4
ifz
HR
,
there exists
( )
R
0
such that1 4 4
( )
H,
K t
t
0.
Proof : From the previous results we know
K t z
( )
H0c
.
For
0
small enough We introduce the energyfunctional
4 3 1 4
1 4 4 1 4 3 4 4
4 4 4
2 2 2
0
( )
( )
2
( )
( ),
V2
,
M.
H V V
E t
K t z
w
w
g u
g v w
w
c
P P
P P
Choose theconstant c appearing in
E t
( )
large enough it is apparent that1 4 1 4
4 4
2 2
1
( )
( )
2
( )
.
2
H HK t z
E t
K t z
c
P
P
(3.27)Arguing as in the previous proof we obtain
4 3
1 4 4
4 2
0 0
( )
( )
2 (
( )
( )) ,
4
t VM
d
E t
E t
g u
g v u w
dt
4 3 4 3
4 4
1 0
2
g u u w
( ) ,
t V 2
g v w w
( )
t,
V c
.
(3.28)Due to
(3.21),(2.5)
and1
( )
2
1 2
(
)
V
D A
↪
3
2
( ),
L
we get3 4 3
4 2 12
4
4 1 4 1
0 0
2 (
( )
( )) ,
t V L t LL
g u
g v u w
c w
u
A
w
P
P
1
3 3 4
8 8
t V
c A w
u
A
w
P
P
P
P
1 3 4
2 8
t V
c u
A
w
1 2 1 4
4
t V
c u
w
P P
c u
t V1E
,
on account of
(3.21),(2.5),
we obtain4 3 3
12 2
4
4 1 4 1 1
2
( ) ,
t V t LL
g v u w
c u
A
w
P
P
1 3 4
2 8
t V
c u
A
w
P
P
c u
t V1
c u
P
tP
V1E
,because of
(2.5),(2.7),(3.21),
this yields4 3 3
12 12
4 2
3 4 4 1
4 1 0
2
( )
t,
V L tL L
g v w w
c v
w
A
w
P
P
P
P
0 1 4 1 4
4 4
t V
V V
c v
w
w
P
P
P P
1 4 0 4
2
.
4
VM
c v
E
P P
Hence
d
E t
( )
E t
( )
hE
h
c
,
dt
where
h
c v t
( )
V0
c u
P P
t V1.
Due to (3.24) lemma 3.4 and the inequality, we get 34
( )
[1 (
) ],
th y dy
c
t
For some
c
0.
According to lemma 2.4 ,we meet the claim.
Lemma 3.9 Condition (H2) holds.
Proof : Successive applications of lemma 3.6 for
0, , , ,
1 1 3
4 2 4
together with the exponential decay(3.24),
we construct a sequence of five balls starting from
B R
0(
0),
each exponentially attracted by the next one. Thus on account of the continuous dependence estimate(3.1),
the transitivity of the exponential attraction lemma 2.5 entails the desired property (H2).Our main theorem reads as follows.
Theorem 3.10 Assume that
g
C
2( ),
¡
and satisfies(1.2),(1.3)
withg
(0)
0.
Then there exist robust exponential attractors{ }
H
0for the semigroup of operatorsS t
( )
generated by system(1.7).
As a straightforward consequence we have the following corollary.
Corollary 3.11 The semigroup
S t
( )
acting on the phase-spaceH
0possesses a connected global attractor.
REFERENCES
[1]. H. Haken Advanced Synergetirs Instability hierarchies of self-organizing systems and devices Springer-Verlag. Berlin (1983);
[2]. A. C. Newell T. Passot J. Lega Order Parameter Equations for Patterns Annual Review of Fluid Mechanics. 25 (1993);
[3]. A. Ion A. Georgescu On the existence and on the fractal and hausdorff dimensions of some global attractor Nonlinear Analysis TMA 8 (1997):5527-5532.
[4]. V. Belleri V. Pata Attractors for semilinear strongly dampedwave equation on
R
3 Discrete Contin. Dynam. Systems 7 (2001):719-735.[5]. M. Grasselli V. Pata Asymptotic behavior of a parabolic-hyperbolic system Commun. Pure Appl. Anal. 38 (2004):49-81.
[6]. S. Gatti M. Grasselli A. Miranville and V. Pata Memory relaxation of first order evolution equations Nonlinearity.18 (2005):1859-1883.
[7]. Fabrie P. Galusinski C. Miranville A. Zelik S. Uniform exponential attractors for a singularly perturbed damped wave equation. Discrete Contin. Dynam. Systems 10 (2004):211-238.
[8]. Grasselli M and Pata V. Existence of a universal attractor for a fully hyperbolic phase-field system. J. Evol. Eqns 4 (2004):27-51.
[9]. A.EdenC.FoiasB.NicolaenkoR.Temam Exponential Attractors for Dissipative Evolution Equations New York:Masson.Paris:Wiely(1994).
[10]. R.Temam Infinite-dimensional dynamical systems in mechanics and physic Springer New York(1997). [11]. Arrieta J. Carvalho A.N. Hale J.K. A damped hyperbolic equation with critical exponent Comm. Partial
Differ. Eqs.17 (1992):841-866.