This paper considers the existence of uniform attractors for a non-autonomous thermoviscoelastic equation with strong damping in a bounded domain Ω⊆Rn(n≥1) by establishing the uniformly asymptotic compactness of the semi-process generated by the global solutions.
In this paper we investigate the existence of uniform attractors for a nonlinear non-autonomous thermoviscoelastic equation with strong damping
| u t | ρ u t t − Δ u − Δ u t t + ∫ 0 + ∞ g ( s ) Δ u ( t − s ) d s − Δ u t + ∇ θ = σ ( x , t ) , x ∈ Ω , t > τ , (1.1)
θ t − Δ θ + div u t = f ( x , t ) , x ∈ Ω , t > τ , (1.2)
θ ( x , t ) = u ( x , t ) = 0 , on ∂ Ω × [ τ , + ∞ ) , (1.3)
u ( x , τ ) = u 0 τ ( x ) , u t ( x , τ ) = u 1 τ ( x ) , u ( x , t ) = u τ ( x , t ) , θ ( x , τ ) = θ 0 t ( x ) , x ∈ Ω , (1.4)
where Ω ⊆ ℝ n ( n = 1 , 2 ) is a bounded domain with smooth boundary ∂ Ω , u and θ are displacement and temperature difference, respectively. u τ ( x , t ) (the past history of u) is a given datum which has to be known for all t ≤ τ , the function g represents the kernel of a memory, σ = σ ( x , t ) , f = f ( x , t ) are non-autonomous terms, called symbols, and ρ is a real number such that
1 < ρ ≤ 2 n − 2 if n ≥ 3 ; ρ > 1 if n = 1 , 2. (1.5)
Now let us recall the related results on nonlinear one-dimensional thermoviscoelasticity. Dafermos [
Our problem is derived from the form
f ( u t ) u t t − Δ u − Δ u t t = 0 , (1.6)
which has several modeling features. The aim of this paper is to extend the decay results in [
Let us recall some results concerning viscoelastic wave equations. In [
u t t − Δ u + ∫ 0 t g ( t − τ ) Δ u = | u | γ + 1 , (1.7)
he proved that the energy decays similarly with that of g. In [
| u t | ρ u t t − Δ u − Δ u t t + g ∗ Δ u + | u | p u = 0 (1.8)
with the same boundary and initial conditions as (1.7), the author proved that, for a class of kernels g which is singular at zero, the exponential decay rate of the solution energy. Later, Han and Wang [
| u t | ρ u t t − Δ u − Δ u t t + g ∗ Δ u + | u t | m u t = 0 , (1.9)
with Dirichlet boundary condition, where ρ > 0 , m > 0 are constants, they proved the energy decay for the viscoelastic equation with nonlinear damping. Then Park and Park [
| u t | ρ u t t − Δ u t t − Δ u + g ∗ Δ u + h ( u t ) = 0 , (10)
with the Dirichlet boundary condition, where ρ > 0 is a constant. In [
| u t | ρ u t t − Δ u − Δ u t t + g ∗ Δ u − γ Δ u t = 0 (1.11)
where g ∗ Δ u = ∫ 0 t g ( t − s ) Δ u ( s ) d s . They established a global existence result for γ ≥ 0 and an exponential decay of energy for γ > 0 , and studied the interaction within the | u t | ρ u t t and the memory term g ∗ Δ u . Messaoudi and Tatar [
| u t | ρ u t t − Δ u − Δ u t t + g ∗ Δ u = b | u | p − 2 u , (1.12)
with Dirichlet boundary condition, where γ ≥ 0 , ρ , b > 0 , p > 2 are constants. In the case b = 0 in (1.12), Messaoudi and Tatar [
| u t | ρ u t t − Δ u − Δ u t t + ∫ 0 + ∞ μ ( s ) Δ u ( t − s ) d s + f ( u ) = h ( x ) ,
and proved the global existence, uniqueness and exponential stability, and the global attractor was also established, but they did not establish the uniform attractors for non-autonomous equation. Then, Qin et al. [
| u t | ρ u t t − Δ u − Δ u t t + ∫ 0 + ∞ g ( s ) Δ u ( t − s ) d s + u t = σ ( x , t ) , x ∈ Ω , t > τ ,
Moreover, we would like to mention some results in [
For problem (1.1)-(1.4) with σ ( x , t ) = 0 , when ∫ 0 + ∞ g ( s ) Δ u ( t − s ) d s was replaced by g ∗ Δ u , Han and Wang [
Motivated by [
η = η t ( x , s ) = u ( x , t ) − u ( x , t − s ) , t ≥ τ , ( x , s ) ∈ Ω × ℝ + . (1.13)
A direct computation yields
η t t ( x , s ) = − η s t ( x , s ) + u t ( x , t ) , t ≥ τ , ( x , s ) ∈ Ω × ℝ + , (1.14)
and we can take as initial condition ( t = τ )
η τ ( x , s ) = u 0 τ ( x ) − u 0 τ ( x , τ − s ) , ( x , s ) ∈ Ω × ℝ + . (1.15)
Thus, the original memory term can be written as
∫ 0 + ∞ g ( s ) Δ u ( t − s ) d s = ∫ 0 + ∞ g ( s ) d s ⋅ Δ u − ∫ 0 + ∞ g ( s ) Δ η t ( s ) d s , (1.16)
and we get a new system
| u t | ρ u t t − ( 1 − ∫ 0 + ∞ g ( s ) d s ) Δ u − Δ u t t − ∫ 0 + ∞ g ( s ) Δ η t ( s ) d s − Δ u t + ∇ θ = σ ( x , t ) , (1.17)
θ t − Δ θ + div u t = f ( x , t ) (1.18)
η t t + η s t = u t , (1.19)
with the boundary conditions
u = 0 on ∂ Ω × ℝ + , η t = 0 on ∂ Ω × ℝ + × ℝ + , (1.20)
and initial conditions
u ( x , τ ) = u 0 τ ( x ) , u t ( x , τ ) = u 1 τ ( x ) , η t ( x , 0 ) = 0 , η τ ( x , s ) = u 0 τ ( x ) − u ( x , τ − s ) . (1.21)
The rest of our paper is organized as follows. In Section 2, we give some preparations for our consideration and our main result. The statements and the proofs of our main results will be given in Section 3 and Section 4, respectively.
For convenience, we denote the norm and scalar product in L 2 ( Ω ) by ‖ ⋅ ‖ and ( ⋅ , ⋅ ) , respectively. C 1 denotes a general positive constant, which may be different in different estimates.
We assume the memory kernel g : ℝ + → ℝ + is a bounded C 1 function such that
g ( s ) < + ∞ , l = 1 − ∫ 0 + ∞ g ( s ) d s > 0 (2.1)
and suppose that there exists a positive constant ξ 2 verifying
g ′ ( t ) ≤ − ξ 2 g ( t ) , ∀ t ≥ 0, (2.2)
In order to consider the relative displacement η as a new variable, one introduces the weighted L2-space
M = L g 2 ( ℝ + ; H 0 1 ( Ω ) ) = { u : ℝ + → H 0 1 ( Ω ) | ∫ 0 + ∞ g ( s ) ‖ ∇ u ( s ) ‖ 2 d s < + ∞ } ,
which is a Hilbert space equipped with inner product and norm
( u , v ) M = ∫ 0 + ∞ g ( s ) ( ∫ Ω ∇ u ( s ) ∇ v ( s ) d x ) d s and ‖ u ‖ M 2 = ∫ 0 + ∞ g ( s ) ‖ ∇ u ( s ) ‖ 2 d s ,
respectively.
Let
H = H 0 1 ( Ω ) × H 0 1 ( Ω ) × L 2 ( Ω ) × M . (2.3)
Define the generalized energy of problem (1.17)-(1.21)
F ( t ) = 1 ρ + 2 ‖ u t ( t ) ‖ ρ + 2 ρ + 2 + l 2 ‖ ∇ u ( t ) ‖ 2 + 1 2 ‖ ∇ u t ( t ) ‖ 2 + 1 2 ‖ θ ‖ 2 + 1 2 ‖ η t ‖ M 2 . (2.4)
To present our main result, we need the following global existence and uniqueness results.
Theorem 2.1. Let ( u 0 τ , u 1 τ , θ 0 τ , η τ ) ∈ H ( ∀ τ ∈ ℝ + ) , ℝ τ = [ τ , + ∞ ) , and any fixed σ , f ∈ E 1 . Assume (2.1) and (2.2) hold. Then problem (1.17)-(1.21) admits a unique global solution ( u , u t , θ , η t ) ∈ C ( [ 0, T ] , H ) such that
u ∈ L ∞ ( ℝ τ , H 0 1 ( Ω ) ) , u t ∈ L ∞ ( ℝ τ , H 0 1 ( Ω ) ) , u t t ∈ L 2 ( ℝ τ , H 0 1 ( Ω ) ) , (2.5)
θ ∈ L ∞ ( ℝ τ , H 0 1 ( Ω ) ) , η t ∈ L ∞ ( ℝ τ , M ) . (2.6)
We now define the symbol space for (1.17)-(1.21).
Let
G = ( σ , f ,0 ) ∈ E 1 ≡ L 2 ( ℝ + , ( L 2 ( Ω ) ) 3 ) . (2.7)
Observe the following important fact: The properly defined (uniform) attractor A of problem (1.17)-(1.21) with the symbol G 0 must be simultaneously the attractor of each problem (1.17)-(1.21) with the symbol G ( t ) ∈ H + ( G 0 ) , which is called the hull of G 0 and defined as
Σ = H + ( G 0 ) = [ G 0 ( t + h ) | h ∈ ℝ + ] E 1 (2.8)
where [ ⋅ ] E 1 denotes the closure in Banach space E 1 .
We note that
G 0 ∈ E 1 ⊆ E ^ 1 = L l o c 2 ( ℝ + , ( L 2 ( Ω ) ) 3 ) .
where G 0 is a translation compact function in E ^ 1 in the weak topology, which means that G 0 is compact in E ^ 1 . We consider the Banach space L l o c p ( ℝ + , E 1 ) of functions μ ( s ) , s ∈ ℝ + with values in a Banach space E 1 that are locally p-power integrable in the Bochner sense. In particular, for any time interval [ t 1 , t 2 ] ⊆ ℝ + ,
∫ t 1 t 2 ‖ μ ( s ) ‖ E 1 p d s < + ∞ .
Let μ ( s ) ∈ L l o c p ( ℝ + , E 1 ) , consider the quantity
η μ ( h ) = sup t ∈ ℝ + ∫ t t + h ‖ μ ( s ) ‖ E 1 p d s .
Lemma 2.1. Let Σ defined as before and G 0 ∈ E 1 , then
1) G 0 is a translation compact in E ^ 1 and any G ∈ Σ = H + ( G 0 ) is also a translation compact in E ^ 1 , moreover, H + ( G ) ⊆ H + ( G 0 ) ;
2) The set H + ( G 0 ) is bounded in L 2 ( ℝ + , L 2 ( Ω ) ) such that
η G ( h ) ≤ η G 0 ( h ) < + ∞ , for all G ∈ Σ .
Proof. See, e.g., Chepyzhov and Vishik [
Lemma 2.2. For every τ ∈ ℝ , every non-negative locally summable function ϕ 0 on R τ ≡ [ τ , + ∞ ) and every ν > 0 , we have
sup t ≥ τ ∫ τ t ϕ 0 ( s ) e − ν ( t − s ) d s ≤ 1 1 − e − ν sup t ≥ τ ∫ t t + 1 ϕ 0 ( s ) d s
for a.a. t ≥ τ .
Proof. See, e.g., Chepyzhov, Pata and Vishik [
Similar to Theorem 2.1, we have the following existence and uniqueness result.
Theorem 2.2. Let Σ = H + ( G 0 ) = [ G 0 ( t + h ) | h ∈ ℝ + ] E 1 , where G 0 ∈ E 1 is an arbitrary but fixed symbol function. Assume (2.1) and (2.2) hold. Then for any G ∈ Σ and for any ( u 0 τ , u 1 τ , θ 0 τ , η τ ) ∈ H ( ∀ τ ∈ ℝ + ) , problem (1.17)-(1.21) admits a unique global solution ( u , u t , θ , η t ) ∈ H , which generates a unique semi-process { U G ( t , τ ) } , ( t ≥ τ ∈ ℝ + , G ∈ Σ ) on H of a two-parameter family of operators such that for any t ≥ τ , τ ∈ ℝ + , ℝ τ = [ τ , + ∞ ) ,
U G ( t , τ ) ( u 0 τ , u 1 τ , θ , η τ ) = ( u , u t , θ , η t ) ∈ H , (2.9)
u ∈ L ∞ ( ℝ τ , H 0 1 ( Ω ) ) , u t ∈ L ∞ ( ℝ τ , H 0 1 ( Ω ) ) , u t t ∈ L 2 ( ℝ τ , H 0 1 ( Ω ) ) ,
θ ∈ L ∞ ( ℝ τ , H 0 1 ( Ω ) ) , η t ∈ L ∞ ( ℝ τ , M ) . (2.10)
Our main result reads as follows.
Theorem 2.3. Assume that G ∈ E 1 and Σ is defined by (2.8), then the family of processes { U G , f ( t , τ ) } ( G ∈ Σ , t ≥ τ , τ ∈ ℝ + ) corresponding to (1.17)-(1.21) has a uniformly (w.r.t. G ∈ Σ ) compact attractor A Σ .
The global existence of solutions is the same as in [
We consider two symbols σ 1 , f 1 and σ 2 , f 2 and the corresponding solutions ( u , θ 1 , η t ) and ( v , θ 2 , ξ t ) of problem (1.17)-(1.21) with initial data ( u 0 τ , u 1 τ , θ 10 , η τ ) and ( v 0 τ , v 1 τ , θ 20 , ξ τ ) respectively. Let ω ( t ) = u ( t ) − v ( t ) , p ( t ) = θ 1 ( t ) − θ 2 ( t ) , ζ t ( x , s ) = η t ( x , s ) − ξ t ( x , s ) .
Then ( ω , p , ζ t ) verifies
| u t | ρ ω t t + v t t ( | u t | ρ − | v t | ρ ) − l Δ ω − Δ ω t t − ∫ 0 + ∞ g ( s ) Δ ζ ( s ) d s − Δ ω t + ∇ p = σ 1 − σ 2 , x ∈ Ω , t > τ , (3.1)
p t − Δ p + div ω = f 1 − f 2 , (3.2)
ζ t t + ζ s t = ω t , (3.3)
with Dirichlet boundary conditions and initial conditions
ω ( x , τ ) = ω 0 τ , ω t ( x , τ ) = ω 1 τ , p ( x , τ ) = p 1 τ , ζ τ = η τ − ξ τ . (3.4)
The corresponding energy for (3.1)-(3.3) is defined
E ω , p ( t ) = 1 2 ∫ Ω | u t | ρ ω t 2 d x + l 2 ‖ ∇ ω ‖ 2 + 1 2 ‖ ∇ ω t ‖ 2 + 1 2 ‖ θ ‖ 2 + 1 2 ‖ ζ t ‖ M 2 . (3.5)
It is easy to see that
( ζ s t , ζ t ) M = 1 2 ∫ Ω ( ∫ 0 + ∞ g ( s ) d d s | ∇ ζ t ( s ) | 2 d s ) d x = − 1 2 ∫ Ω ( ∫ 0 + ∞ g ′ ( s ) | ∇ ζ t ( s ) | 2 d s ) d x .
Noting that x → | x | ρ is differentiable since ρ > 1 . Then
1 2 d d t ∫ Ω | u t | ρ ω t 2 d x = ∫ Ω | u t | ρ ω t t ω t d x + ρ 2 ∫ Ω | u t | ρ − 2 u t u t t ω t 2 d x ,
and clearly
d d t E ω , p ( t ) = − ‖ ∇ ω t ‖ 2 + 1 2 ∫ 0 + ∞ g ′ ( s ) ‖ ∇ ζ t ( s ) ‖ 2 d s + ∫ Ω ( σ 1 − σ 2 ) ω t d x + ∫ Ω ( f 1 − f 2 ) θ d x + ρ 2 ∫ Ω | u t | ρ − 1 u t t ω t 2 d x − ∫ Ω v t t ω t ( | u t | ρ − | v t | ρ ) d x . (3.6)
To simplify notations, let us say that the norm of the initial data is bounded by some R > 0 . Then given T > τ we use C R T to denote several positive constants which depend on R and T.
By Young’s inequality and the interpolation inequalities, we derive
| ∫ Ω ( σ 1 − σ 2 ) ω t d x | ≤ ‖ σ 1 − σ 2 ‖ ‖ ω t ‖ ≤ ‖ σ 1 − σ 2 ‖ 2 + C R T E ω ( t ) , (3.7)
| ∫ Ω ( f 1 − f 2 ) θ d x | ≤ ‖ f 1 − f 2 ‖ 2 + C R T E ω , p ( t ) , (3.8)
ρ 2 | ∫ Ω | u t | ρ − 1 u t t ω t 2 d x | ≤ ρ 2 ‖ u t ‖ 2 ( ρ + 1 ) ρ − 1 ‖ u t t ‖ ‖ ω t ‖ 2 ( ρ + 1 ) 2 ≤ C R T ‖ ∇ u t t ‖ ‖ ∇ ω t ‖ 2 , (3.9)
| − ∫ Ω v t t ( | u t | ρ − | v t | ρ ) ω t d x | ≤ C 1 ∫ Ω | v t t | ( | u t | ρ − 1 + | v t | ρ − 1 ) ω t 2 d x ≤ C 1 ‖ v t t ‖ ( ‖ u t ‖ 2 ( ρ + 1 ) ρ − 1 + ‖ v t ‖ 2 ( ρ + 1 ) ρ − 1 ) ‖ ω t ‖ 2 ( ρ + 1 ) 2 ≤ C 1 ‖ ∇ v t t ‖ ‖ ∇ ω t ‖ 2 ,
which, together with (3.6)-(3.9), yields for some C 1 > 0 large
d d t E ω , p ( t ) ≤ ‖ σ 1 − σ 2 ‖ 2 + ‖ f 1 − f 2 ‖ 2 + C 1 ( 1 + ‖ ∇ u t t ‖ + ‖ ∇ v t t ‖ ) E ω , p ( t ) . (3.10)
Integrating (3.10) from τ to t and using Hölder’s inequality, we have
E ω , p ( t ) ≤ E ω , p ( τ ) + ∫ τ t ‖ σ 1 ( s ) − σ 2 ( s ) ‖ 2 d s + ∫ τ t ‖ f 1 ( s ) − f 2 ( s ) ‖ 2 d s
+ C 1 ∫ τ t ( 1 + ‖ ∇ u t t ‖ + ‖ ∇ v t t ‖ ) E ω , p ( s ) d s ≤ E ω , p ( τ ) + ∫ τ T ‖ σ 1 ( s ) − σ 2 ( s ) ‖ 2 d s + ∫ τ t ‖ f 1 ( s ) − f 2 ( s ) ‖ 2 d s + C 1 ( ∫ τ t ( 1 + ‖ ∇ u t t ‖ + ‖ ∇ v t t ‖ ) 2 d s ) 1 2 ( ∫ τ t E ω , p 2 ( s ) d s ) 1 2 . (3.11)
Noting that
∫ τ T ( 1 + ‖ ∇ u t t ‖ + ‖ ∇ v t t ‖ ) 2 d s ≤ C R T ,
then we get for any t ∈ [ τ , T ]
E ω , p 2 ( t ) ≤ 2 ( E ω , p ( τ ) + ∫ τ T ‖ σ 1 ( s ) − σ 2 ( s ) ‖ 2 d s + ∫ τ T ‖ f 1 ( s ) − f 2 ( s ) ‖ 2 d s ) 2 + C R T ∫ τ t E ω 2 ( s ) d s . (3.12)
Applying Gronwall’s inequality, we see that
E ω , p ( t ) ≤ 2 ( E ω ( τ ) + ∫ τ T ‖ σ 1 ( s ) − σ 2 ( s ) ‖ 2 d s + ∫ τ T ‖ f 1 ( s ) − f 2 ( s ) ‖ 2 d s ) exp ( C R T 2 T ) , ∀ t ∈ [ τ , T ] . (3.13)
Using ∫ Ω | u t | ρ | ω t | 2 d x ≤ ‖ u t ‖ ρ + 2 ρ ‖ ω t ‖ ρ + 2 2 ≤ C R T ‖ ∇ ω t ‖ 2 , we know that E ω , p ( t ) is equivalent to the norm of u , θ in H and we get
E ω , p ( τ ) ≤ C R T ‖ ( ω 0 τ , ω 1 τ , p 0 τ , ζ τ ) ‖ H 2 ,
which, together with (3.13), gives for all τ ≤ t ≤ T
‖ u ( t ) − v ( t ) ‖ H 0 1 2 + ‖ u t ( t ) − v t ( t ) ‖ H 0 1 2 + ‖ η t − ξ t ‖ M 2 ≤ C R T ( ‖ u 0 τ − v 0 τ ‖ H 0 1 2 + ‖ u 1 τ − v 1 τ ‖ H 0 1 2 + ‖ η τ − ξ τ ‖ M 2 + ‖ σ 1 − σ 2 ‖ L 2 ( τ , T ; L 2 ( Ω ) ) 2 ) .
This shows that solutions of (1.17)-(1.21) depend continuously on the initial data. We complete the proof of Theorem 2.1.
In this section, we shall establish the existence of uniform attractors for system (1.17)-(1.21). To this end, we shall introduce some basic conceptions and basic lemmas. For more results concerning uniform attractors, we can refer to [
Let X be a Banach space, and Σ ^ be a parameter set. The operators { U G ( t , τ ) } ( t ≥ τ , τ ∈ ℝ + , G ∈ Σ ^ ) are said to be a family of processes in X with symbol space Σ ^ if for any G ∈ Σ ^ ,
U G ( t , s ) U G ( s , τ ) = U G ( t , τ ) , ∀ t ≥ s ≥ τ , τ ∈ ℝ + , (4.1)
U G ( τ , τ ) = I d ( identity ) , ∀ τ ∈ ℝ + . (4.2)
Let { T ( s ) } be the translation semigroup on Σ ^ , we say that a family of processes { U G ( t , τ ) } ( t ≥ τ , τ ∈ ℝ + , G ∈ Σ ^ ) satisfies the translation identity if
U G ( t + s , τ + s ) = U T ( s ) G ( t , τ ) , ∀ G ∈ Σ ^ , t ≥ τ , τ , s ∈ ℝ + , (4.3)
T ( s ) Σ ^ = Σ ^ , ∀ s ∈ ℝ + . (4.4)
By B ( X ) we denote the collection of the bounded sets of X, and ℝ τ = [ τ , + ∞ ) , τ ∈ ℝ + .
Definition 4.1. A bounded set B 0 ∈ B ( X ) is said to be a bounded uniformly (w.r.t G ∈ Σ ^ ) absorbing set for { U G ( t , τ ) } ( G ∈ Σ ^ , t ≥ τ , τ ∈ ℝ + ) if for any τ ∈ ℝ + and B ∈ B ( X ) , there exists a time T 0 = T 0 ( B , τ ) ≥ τ such that
∪ G ∈ Σ ^ U G ( t , τ ) B ⊆ B 0 , (4.5)
for all t ≥ T 0 .
In the following, as usual, (w.r.t) will represent “with respect to”.
Definition 4.2. The family of semi-processes { U σ ( t , τ ) } ( t ≥ τ , τ ∈ ℝ + , σ ∈ Σ ^ ) is said to be asymptotically compact in X if { U σ ( t , τ ) ( u 0 τ ( n ) , u 1 τ ( n ) , θ 0 τ ( n ) , η τ ( n ) ) } is precompact in X, whenever ( u 0 τ ( n ) , u 1 τ ( n ) , θ 0 τ ( n ) , η τ ( n ) ) is bounded in X, G ( n ) ⊂ Σ ^ , and t n ∈ ℝ τ , t n → + ∞ as n → + ∞ .
Definition 4.3. A set A ⊆ X is said to be uniformly (w.r.t G ∈ Σ ^ ) attracting for the family of semi-processes { U G ( t , τ ) } ( t ≥ τ , τ ∈ ℝ + , G ∈ Σ ^ ) if for any fixed τ ∈ ℝ + and any B ∈ B ( X ) ,
lim t → + ∞ ( sup d i s t ( U G ( t , τ ) B , A ) ) = 0, (4.6)
here d i s t ( ⋅ , ⋅ ) stands for the usual Hausdorff semidistance between two sets in X. In particular, a closed uniformly attracting set A Σ ^ is said to be the uniform (w.r.t G ∈ Σ ^ ) attractor of the family of the semi-process
{ U G ( t , τ ) } ( t ≥ τ , τ ∈ ℝ + , G ∈ Σ ^ )
if it is contained in any closed uniformly attracting set (minimality property).
Definition 4.4. Let X be a Banach space and B be a bounded subset of X , Σ ^ be a symbol (or parameter) space. We call a function ϕ ( ⋅ , ⋅ ; ⋅ , ⋅ ) , defined on ( X × X ) × ( Σ ^ × Σ ^ ) to be a contractive function on B × B if for any sequence { x n } n = 1 ∞ ⊆ B and any { μ n } ⊆ Σ ^ , there is a subsequence { x n k } k = 1 ∞ ⊂ { x n } n = 1 ∞ and { μ n k } k = 1 ∞ ⊂ { μ n } n = 1 ∞ such that
lim k → ∞ lim l → ∞ ϕ ( x n k , x n l ; μ n k , μ n l ) = 0. (4.7)
We denote the set of all contractive functions on B × B by C o n t r ( B , Σ ^ ) .
Lemma 4.1. Let { U G ( t , τ ) } ( t ≥ τ , τ ∈ ℝ + , G ∈ Σ ^ ) be a family of semi-processes satisfying the translation identities (4.3) and (4.4) on Banach space X and has a bounded uniformly (w.r.t G ∈ Σ ^ ) absorbing set B 0 ⊆ X . Moreover, assuming that for any ε > 0 , there exist T = T ( B 0 , ε ) > 0 and ϕ T ∈ C o n t r ( B 0 , Σ ^ ) such that
‖ U G 1 ( T , 0 ) x − U G 2 ( T , 0 ) y ‖ ≤ ε + ϕ T ( x , y ; G 1 , G 2 ) , ∀ G ∈ Σ ^ , t ≥ τ , τ ∈ ℝ + . (4.8)
Then { U G ( t , τ ) } ( t ≥ τ , τ ∈ ℝ + , G ∈ Σ ^ ) is uniformly (w.r.t G ∈ Σ ^ ) asymptotically compact in X.
Proof. This lemma is a version for semi-processes of a result by Khanmamedov [
Next, we will divide into two subsections to prove Theorem 2.3.
In this subsection we shall establish the family of processes { U G ( t , τ ) } has a bounded uniformly absorbing set given in the following theorem.
Theorem 4.1. Assume that G ∈ E 1 and Σ is defined by (2.7), then the family of processes { U G ( t , τ ) } ( G ∈ Σ , t ≥ τ , τ ∈ ℝ + ) corresponding to (1.17)-(1.21) has a bounded uniformly (w.r.t. G ∈ Σ ) absorbing set B in H .
Proof. We define
F ( t ) = 1 ρ + 2 ‖ u t ‖ ρ + 2 ρ + 2 + l 2 ‖ ∇ u ‖ 2 + 1 2 ‖ ∇ u t ‖ 2 + 1 2 ‖ θ ‖ 2 + 1 2 ‖ η t ‖ M 2 . (4.9)
Using Young’s inequality, Poincaré’s inequality, we arrive at
F ′ ( t ) = − ‖ ∇ u t ‖ 2 − ( η s t , η t ) M + ( σ , u t ) + ( f , θ ) ≤ − 1 2 ‖ u t ‖ 2 − 1 2 ‖ ∇ θ ‖ 2 + 1 2 ∫ 0 + ∞ g ′ ( s ) ‖ ∇ η t ( s ) ‖ 2 d s + 1 2 ε ( ‖ σ ‖ 2 + ‖ f ‖ 2 ) . (4.10)
Let
F 1 ( t ) = F ( t ) + 1 2 ∫ t + ∞ ( ‖ σ ( s ) ‖ 2 + ‖ f ( s ) ‖ 2 ) d s , for all t ≥ τ . (4.11)
Then (4.11) gives F ′ 1 ( t ) ≤ 0 , whence from (4.9), for t ≥ τ > 0
F ( t ) ≤ F 1 ( t ) ≤ F 1 ( τ ) = F ( τ ) + 1 2 ∫ τ + ∞ ( ‖ σ ( s ) ‖ 2 + ‖ f ( s ) ‖ 2 ) d s = F ( τ ) + 1 2 ( ‖ σ ‖ L 2 ( ℝ τ , L 2 ( Ω ) ) 2 + ‖ f ‖ L 2 ( ℝ τ , L 2 ( Ω ) ) 2 ) , (4.12)
l 2 ‖ ∇ u ‖ 2 + 1 2 ‖ ∇ u t ‖ 2 + l 2 ‖ θ ‖ 2 + 1 2 ‖ η t ‖ M 2 ≤ F ( t ) ≤ F 1 ( t ) ≤ F 1 ( τ ) . (4.13)
Now we define
Φ ( t ) = 1 ρ + 1 ∫ Ω | u t | ρ u t u d x + ∫ Ω ∇ u t ⋅ ∇ u d x . (4.14)
From (1.17), integration by parts and Young’s inequality, we derive for any ε ∈ ( 0,1 ) ,
Φ ′ ( t ) = ( l Δ u + ∫ 0 + ∞ g ( s ) Δ η t ( s ) d s + Δ u t − ∇ θ + σ , u ) + 1 ρ + 1 ( | u t | ρ u t , u t ) − ( Δ u t , u t ) = − l ‖ ∇ u ‖ 2 − ( ∇ u , ∫ 0 + ∞ g ( s ) ∇ η t ( s ) d s ) − ( ∇ θ , u ) + ( Δ u t , u ) + 1 ρ + 1 ‖ u t ‖ ρ + 2 ρ + 2 + ‖ ∇ u t ‖ 2 + ( σ , u ) . (4.15)
Using Young’s inequality, Hölder’s inequality and Poincaré’s inequality, we deduce
| ( Δ u t , u ) | ≤ ε ‖ ∇ u ‖ 2 + 1 4 ε ‖ ∇ u t ‖ 2 , (4.16)
| − ( ∇ θ , u ) | ≤ ε ‖ u ‖ 2 + 1 4 ε ‖ ∇ θ ‖ 2 ≤ ε λ 2 ‖ ∇ u ‖ 2 + 1 4 ε ‖ ∇ θ ‖ 2 , (4.17)
| − ( ∇ u , ∫ 0 + ∞ g ( s ) ∇ η t ( s ) d s ) | ≤ ε ‖ ∇ u ‖ 2 + 1 4 ε ∫ Ω ( ∫ 0 + ∞ g ( s ) ∇ η t ( s ) d s ) 2 d x ≤ ε ‖ ∇ u ‖ 2 + 1 − l 4 ε ∫ 0 + ∞ g ( s ) ‖ ∇ η t ( s ) ‖ 2 d s , (4.18)
| ( σ , u ) | ≤ ε ‖ u ‖ 2 + 1 4 ε ‖ σ ‖ 2 ≤ ε λ 2 ‖ ∇ u ‖ 2 + 1 4 ε ‖ σ ‖ 2 , (4.19)
hereinafter we use λ to represent the Poincaré constant.
From the expression of F ( t ) , we get
‖ ∇ u ‖ 2 = 2 l F ( t ) − 2 l ( ρ + 2 ) ‖ u t ‖ ρ + 2 ρ + 2 − 1 l ‖ ∇ u t ‖ 2 − 1 l ‖ θ ‖ 2 − 1 l ‖ η t ‖ M 2 ,
which, together with (4.15)-(4.19), yields
Φ ′ ( t ) ≤ − 1 2 ( l − 2 ε λ 2 − 2 ε ) ‖ ∇ u ‖ 2 − 1 2 ( l − 2 ε λ 2 − 2 ε ) ( 2 l F ( t ) − 2 l ( ρ + 2 ) ‖ u t ‖ ρ + 2 ρ + 2 − 1 l ‖ ∇ u t ‖ 2 − 1 l ‖ θ ‖ 2 − 1 l ‖ η t ‖ M 2 ) + 1 − l 4 ε ‖ η t ‖ M 2 + 1 ρ + 1 ‖ u t ‖ ρ + 2 ρ + 2 + ( 1 + 1 4 ε ) ‖ ∇ u t ‖ 2 + 1 4 ε ( ‖ σ ‖ 2 + ‖ ∇ θ ‖ 2 )
≤ − 1 2 ( l − 2 ε λ 2 − 2 ε ) ‖ ∇ u ‖ 2 − l − 2 ε λ 2 − 2 ε l F ( t ) + ( l − 2 ε λ 2 − 2 ε l ( ρ + 2 ) + 1 ρ + 1 ) ‖ u t ‖ ρ + 2 ρ + 2 + [ 1 + 1 4 ε + 1 2 ( l − 2 ε λ 2 − 2 ε ) ] ‖ ∇ u t ‖ 2 + ( 1 2 l ( l − 2 ε λ 2 − 2 ε ) + 1 − l 4 ε ) ‖ η t ‖ M 2 + 1 4 ε ‖ σ ‖ 2 + ( λ l + 1 4 ε ) ‖ ∇ θ ‖ 2 (4.20)
Noting that ‖ ∇ u t ‖ 2 ≤ 2 F ( t ) ≤ 2 F 1 ( τ ) and the embedding theorem H 1 ( Ω ) ↣ L 2 ( ρ + 1 ) ( Ω ) , we have for any ε ∈ ( 0,1 ) ,
( l − 2 ε λ 2 − 2 ε l ( ρ + 2 ) + 1 ρ + 1 ) ‖ u t ‖ ρ + 2 ρ + 2 ≤ C 1 ‖ u t ‖ ‖ u t ‖ 2 ( ρ + 1 ) ρ + 1 ≤ C 1 ‖ ∇ u t ‖ ρ + 1 ‖ u t ‖
≤ ε ‖ ∇ u t ‖ 2 ( ρ + 1 ) + C ε ‖ u t ‖ 2 ≤ 2 ρ + 1 ε F 1 ρ ( τ ) F ( t ) + C ε ‖ u t ‖ 2 ,
which, together with (4.20) and Poincaré’s inequality, gives
Φ ′ ( t ) ≤ − 1 2 ( l − 2 ε λ 2 − 2 ε ) ‖ ∇ u ‖ 2 − ( l − 2 ε λ 2 − ε l − 2 ρ + 1 ε F 1 ρ ( τ ) ) F ( t ) + ( 1 + 1 4 ε C ε λ 2 + l − 2 ε λ 2 − 2 ε 2 l + 1 4 ε ) ‖ ∇ u t ‖ 2 + ( l − 2 ε λ 2 − 2 ε 2 l + 1 − l 4 ε ) ‖ η t ‖ M 2 + 1 4 ε ‖ σ ‖ 2 + ( λ l + 1 4 ε ) ‖ ∇ θ ‖ 2 . (4.21)
Now we take ε ∈ ( 0,1 ) so small that
l − 2 ε λ 2 − 2 ε ≥ l 2 , l − 2 ε λ 2 − 2 ε l − 2 ρ + 1 ε F 1 ρ ( τ ) ≥ 1 4 . (4.22)
Hence from (4.21)-(4.22), it follow
Φ ′ ( t ) ≤ − l 4 ‖ ∇ u ‖ 2 − 1 4 F ( t ) + C 1 ‖ η t ‖ M 2 + C 1 ( ‖ ∇ u t ‖ 2 + ‖ ∇ θ ‖ 2 ) + 1 4 ε ‖ σ ‖ 2 . (4.23)
We define the functional
Ψ ( t ) = ∫ Ω ( Δ u t − 1 ρ + 1 | u t | ρ u t ) ∫ 0 + ∞ g ( s ) η t ( s ) d s d x . (4.24)
It follows from (1.17) that
Ψ ′ ( t ) = ∫ Ω ( − l Δ u − ∫ 0 + ∞ g ( s ) Δ η t ( s ) d s + u t + ∇ θ − σ ) ∫ 0 + ∞ g ( s ) η t ( s ) d s + ∫ Ω ( Δ u t − 1 ρ + 1 | u t | ρ u t ) ∫ 0 + ∞ g ( s ) η t t ( s ) d s : = I 1 + I 2 . (4.25)
From Young’s inequality, Hölder’s inequality and Poincaré’s inequality, we derive for any δ ∈ ( 0,1 ) ,
| ∫ Ω ( − l Δ u ) ∫ 0 + ∞ g ( s ) η t ( s ) d s | ≤ δ ‖ ∇ u ‖ 2 + l 2 ( 1 − l ) 4 δ ‖ η t ‖ M 2 , (4.26)
| ∫ Ω − ( ∫ 0 + ∞ g ( s ) Δ η t ( s ) d s ) 2 d s | ≤ ( 1 − l ) ‖ η t ‖ M 2 , (4.27)
| ∫ Ω − Δ u t ⋅ ∫ 0 + ∞ g ( s ) η t ( s ) d s | ≤ δ ‖ ∇ u t ‖ 2 + 1 − l 4 δ ‖ η t ‖ M 2 , (4.28)
| ∫ Ω ∇ θ ⋅ ∫ 0 + ∞ g ( s ) η t ( s ) d s | ≤ δ ‖ ∇ θ ‖ 2 + λ 2 ( 1 − l ) 4 δ ‖ η t ‖ M 2 , (4.29)
| ∫ Ω ( − σ ) ∫ 0 + ∞ g ( s ) η t ( s ) d s | ≤ 1 2 ‖ σ ‖ 2 + 1 − l 2 ‖ η t ‖ M 2 ,
which, together with (4.26)-(4.29), gives
I 1 ≤ δ ( ‖ ∇ u t ‖ 2 + ‖ ∇ u ‖ 2 + ‖ ∇ θ ‖ 2 ) + 1 2 ‖ σ ‖ 2 + ( 3 ( 1 − l ) 2 + ( 1 + λ 2 + l 2 ) ( 1 − l ) 4 δ ) ‖ η t ‖ M 2 . (4.30)
Noting that
∫ 0 + ∞ g ( s ) η t t ( s ) d s = − ∫ 0 + ∞ g ( s ) η s t ( s ) d s + ∫ 0 + ∞ g ( s ) u t ( t ) d s = ∫ 0 + ∞ g ′ ( s ) η t ( s ) d s + ( 1 − l ) u t ,
then we have
I 2 = − ( 1 − l ) ‖ ∇ u t ‖ 2 − 1 − l ρ + 1 ‖ u t ‖ ρ + 2 ρ + 2 + ∫ 0 + ∞ g ′ ( s ) ∫ Ω Δ u t ( t ) η t ( s ) d x d s + 1 ρ + 1 ∫ Ω | u t | ρ u t ∫ 0 + ∞ g ′ ( s ) η t ( s ) d s d x . (4.31)
By Young’s inequality, we derive
∫ 0 + ∞ g ′ ( s ) ∫ Ω Δ u t ( t ) η t ( s ) d x d s ≤ − ∫ 0 + ∞ g ′ ( s ) ‖ ∇ u t ( t ) ‖ ‖ ∇ η t ( s ) ‖ d s ≤ 1 − l 4 ‖ ∇ u t ( t ) ‖ 2 − 1 1 − l ∫ 0 + ∞ g ′ ( s ) ‖ ∇ η t ( s ) ‖ 2 d s , (4.32)
and for any ε > 0
1 ρ + 1 ∫ Ω | u t | ρ u t ∫ 0 + ∞ g ′ ( s ) η t ( s ) d s d x ≤ ε ‖ ∇ u t ‖ 2 − C ε ∫ 0 + ∞ g ′ ( s ) ‖ ∇ η t ( s ) ‖ 2 d s ,
which, together with (4.30)-(4.32) and taking ε > 0 small enough, yields
I 2 ≤ − 1 − l 2 ‖ ∇ u t ‖ 2 − 1 − l ρ + 1 ‖ u t ‖ ρ + 2 ρ + 2 − C 1 ∫ 0 + ∞ g ′ ( s ) ‖ ∇ η t ( s ) ‖ 2 d s . (4.33)
Inserting (4.30) and (4.33) into (4.25), we arrive at
Ψ ′ ( t ) ≤ − 1 − l 4 ‖ ∇ u t ‖ 2 + δ ( ‖ ∇ u ‖ 2 + ‖ ∇ θ ‖ 2 ) + C 1 ‖ η t ‖ M 2 + 1 2 ‖ σ ‖ 2 − C 1 ∫ 0 + ∞ g ′ ( s ) ‖ ∇ η t ( s ) ‖ 2 d s − 1 − l ρ + 1 ‖ u t ‖ ρ + 2 ρ + 2 . (4.34)
Set
H ( t ) = M F ( t ) + ε Φ ( t ) + Ψ ( t ) , (4.35)
where M and ε are positive constants.
Then it follows from (4.10), (4.23), (4.34) and (2.2) that
H ′ ( t ) ≤ M 2 ‖ u t ‖ 2 − ( l ε 4 − δ ) ‖ ∇ u ‖ 2 − ε 4 F ( t ) − ( 1 − l 4 − C 1 ε + M 2 ) ‖ ∇ u t ‖ 2 − ( M 2 − δ − C 1 ε ) ‖ ∇ θ ‖ 2 + ( M 2 − C 1 − C 1 ξ 2 ) ∫ 0 + ∞ g ′ ( s ) ‖ ∇ η t ( s ) ‖ 2 d s + ( M 2 ε + ε 4 ε + 1 2 ) ‖ σ ‖ 2 + M 2 ε ‖ f ‖ 2 − 1 − l ρ + 1 ‖ u t ‖ ρ + 2 ρ + 2 . (4.36)
Now we claim that there exist two constants β 1 , β 2 > 0 such that
β 1 F ( t ) ≤ H ( t ) ≤ β 2 F ( t ) , t ≥ 0. (4.37)
For any t ≥ τ , we take ε so small that
M 2 + 1 − l 4 − C 1 ε > 0. (4.38)
For fixed ε , we choose δ small enough and M so large that
M 2 + δ − C 1 ε > 0 , l ε 4 − δ > 0 , M 2 − C 1 − C 1 ξ 2 > 0.
Then there exist a constant γ > 0 such that
H ′ ( t ) ≤ − γ F ( t ) + C 1 ( ‖ σ ‖ 2 + ‖ f ‖ 2 ) , (4.39)
which, together with (4.37), gives
H ′ ( t ) ≤ − γ β 2 H ( t ) + C 1 ( ‖ σ ‖ 2 + ‖ f ‖ 2 ) . (4.40)
Integrating (4.40) over [ τ , t ] with respect to t and using Lemmas 2.2-2.3, we obtain
H ( t ) ≤ H ( τ ) e − γ β 2 ( t − τ ) + C 1 ∫ τ t e − γ β 2 ( t − s ) ( ‖ σ ( s ) ‖ 2 + ‖ f ( s ) ‖ 2 ) d s ≤ C B 0 e − γ β 2 ( t − τ ) + C 1 1 1 − e − γ β 2 sup t ≥ τ ∫ t t + 1 ( ‖ σ ( s ) ‖ 2 + ‖ f ( s ) ‖ 2 ) d s ≤ C B 0 e − γ β 2 ( t − τ ) + C 1 1 1 − e − γ β 2 η G ( 1 ) . (4.41)
Now for any bounded set B 0 ⊆ H , for any ( u 0 τ , u 1 τ , θ 0 τ , η τ ) ∈ B 0 , there exists a constant C B 0 > 0 such that F ( τ ) ≤ C B 0 ≤ C 1 . Taking
R 0 2 = 2 ( 2 C 1 η G 0 ( 1 ) 1 − e − γ β 2 + 1 ) ,
t 0 = τ − ( γ β 2 ) − 1 log ( C 1 η G 0 ( 1 ) + 1 C B 0 ( 1 − e − γ β 2 ) ) ,
then for any t ≥ t 0 ≥ 1 , we have
H ( t ) ≤ C B 0 e − γ β 2 ( t − τ ) + C 1 1 1 − e − γ β 2 η G 0 ( 1 ) ≤ R 0 2 2 ,
which gives
‖ ( u , u t , θ , η t ) ‖ H ≤ 2 H ( t ) = R 0 2 ,
i.e.,
B = B ( 0 , R 0 ) = { ( u , u t , θ , η t ) ∈ H : ‖ ( u , u t , θ , η t ) ‖ H 2 ≤ R 0 2 } ⊆ H
is a uniform absorbing ball for any G ∈ E 1 . The proof is now complete.
In this subsection, we will prove the uniformly (w.r.t. G ∈ Σ ) asymptotic compactness in H , which is given in the following theorem.
Theorem 4.2. Assume that G ∈ E 1 and Σ is defined by (2.8), then the family of processes { U G ( t , τ ) } ( G ∈ Σ , t ≥ τ , τ ∈ ℝ + ) corresponding to (1.17)-(1.21) is uniformly (w.r.t. G ∈ Σ ) asymptotically compact in H .
Proof. For any ( u 0 i τ , u 1 i τ , θ 0 i τ , η i τ ) ∈ B , i = 1 , 2 . We consider two symbols σ 1 , f 1 and σ 2 , f 2 and the corresponding solutions u 1 , θ 1 and u 2 , θ 2 of problem (1.17)-(1.21) with initial data ( u 0 i τ , u 1 i τ , θ 0 i τ , η i τ ) , i = 1 , 2 , respectively. Let ω ( t ) = u 1 ( t ) − u 2 ( t ) , p ( t ) = θ 1 ( t ) − θ 2 ( t ) , ζ t ( x , s ) = η 1 t ( x , s ) − η 2 t ( x , s ) .
Then ( ω , ζ t ) verifies
| u 1 t | ρ ω t t + u 2 t t ( | u 1 t | ρ − | u 2 t | ρ ) − l Δ ω − Δ ω t t − ∫ 0 + ∞ g ( s ) Δ ζ ( s ) d s − Δ ω t + ∇ p = σ 1 − σ 2 , x ∈ Ω , t > τ , (4.42)
p t − Δ p + div ω = f 1 − f 2 (4.43)
ζ t t + ζ s t = ω t , (4.44)
with Dirichlet boundary conditions and initial conditions
ω ( x , τ ) = ω 0 τ , ω t ( x , τ ) = ω 1 τ , p ( x , τ ) = p 0 τ , ζ τ = η 1 τ − η 2 τ . (4.45)
The corresponding energy for (4.42)-(4.45) is defined
E ω , p ( t ) = 1 2 ∫ Ω | u 1 t | ρ ω t 2 d x + l 2 ‖ ∇ ω ‖ 2 + 1 2 ‖ ∇ ω t ‖ 2 + 1 2 ‖ θ ‖ 2 + 1 2 ‖ ζ t ‖ M 2 . (4.46)
Clearly,
d d t E ω , p ( t ) = − ‖ ∇ ω t ‖ 2 + 1 2 ∫ 0 + ∞ g ′ ( s ) ‖ ∇ ζ t ( s ) ‖ 2 d s + ∫ Ω ( σ 1 − σ 2 ) ω t d x + ∫ Ω ( f 1 − f 2 ) p d x + ρ 2 ∫ Ω | u 1 t | ρ − 1 u 1 t t ω t 2 d x − ∫ Ω u 2 t t ω t ( | u 1 t | ρ − | u 2 t | ρ ) d x . (4.47)
Using Hölder’s inequality, Poincaré’s inequality and Theorem 4.1, we derive
| ∫ Ω ( σ 1 − σ 2 ) ω t d x | ≤ ‖ σ 1 − σ 2 ‖ ‖ ω t ‖ , (4.48)
| ∫ Ω ( f 1 − f 2 ) p d x | ≤ ‖ f 1 − f 2 ‖ ‖ p ‖ , (4.49)
ρ 2 | ∫ Ω | u 1 t | ρ − 1 u 1 t t ω t 2 d x | ≤ ρ 2 ‖ u 1 t ‖ 2 ( ρ + 1 ) ρ − 1 ‖ u 1 t t ‖ ‖ ω t ‖ 2 ( ρ + 1 ) ‖ ω t ‖ ≤ C B ‖ ∇ u 1 t ‖ ρ − 1 ‖ ∇ ω t ‖ ‖ ∇ u 1 t t ‖ ‖ ω t ‖ ≤ C B ‖ ω t ‖ ‖ ∇ u 1 t t ‖ , (4.50)
| − ∫ Ω u 2 t t ( | u 1 t | ρ − | u 2 t | ρ ) ω t d x | ≤ C 1 ‖ u 2 t t ‖ 2 ( ρ + 1 ) ( ‖ u 1 t ‖ 2 ( ρ + 1 ) ρ + ‖ u 2 t ‖ 2 ( ρ + 1 ) ρ ) ‖ ω t ‖ ≤ C 1 ‖ ∇ u 2 t t ‖ ( ‖ ∇ u 1 t ‖ ρ + ‖ ∇ u 2 t ‖ ρ ) ‖ ω t ‖ ≤ C B ‖ ∇ u 2 t t ‖ ‖ ω t ‖ ,
which, combined with (4.47)-(4.50), yields
d d t E ω , p ( t ) ≤ − ‖ ω t ‖ 2 − ‖ ∇ p ‖ 2 + 1 2 ∫ 0 + ∞ g ′ ( s ) ‖ ∇ ζ t ( s ) ‖ 2 d s + ‖ ω t ‖ ‖ σ 1 − σ 2 ‖ + ‖ f 1 − f 2 ‖ ‖ p ‖ + C B 0 ( ‖ ∇ u 1 t t ‖ + ‖ ∇ u 2 t t ‖ ) ‖ ω t ‖ . (4.51)
We define
Φ ω , p ( t ) = ∫ Ω | u 1 t | ρ ω t ω d x + ∫ Ω ∇ ω t ⋅ ∇ ω d x + 1 2 ∫ Ω ( ∇ ω ) 2 d x . (4.52)
It is very easy to verify
| Φ ω , p ( t ) | ≤ 1 2 ( ‖ ∇ ω ‖ 2 + ‖ ∇ ω t ‖ 2 ) + ‖ u 1 t ‖ 2 ( ρ + 1 ) ρ ‖ ω t ‖ 2 ( ρ + 1 ) ‖ ω ‖ + 1 2 ‖ ω ‖ 2 ≤ C B ( ‖ ∇ ω ‖ 2 + ‖ ∇ ω t ‖ 2 ) ≤ C B E ω , p ( t ) . (4.53)
Taking the derivative of Φ ω ( t ) , it follows from (4.42)-(4.43) that
Φ ′ ω , p ( t ) = − ∫ Ω u 2 t t ( | u 1 t | ρ − | u 2 t | ρ ) ω d x − l ‖ ∇ ω ‖ 2 − ∫ Ω ∇ ω ∫ 0 + ∞ g ( s ) ∇ ζ t ( s ) d s d x + ∫ Ω ∇ p ω d x + ∫ Ω ( σ 1 − σ 2 ) ω d x + ∫ Ω ( ρ | u 1 t | ρ − 1 u 1 t t ω t ω + | u 1 t | ρ ω t 2 ) d x + ‖ ∇ ω t ‖ 2 = ∑ i = 1 5 A i − l ‖ ∇ ω ‖ 2 + ‖ ∇ ω t ‖ 2 . (4.54)
Applying Hölder’s inequality, Young’s inequality, Poinceré’s inequality and Theorem 4.1, we get
| A 1 | ≤ C 1 ‖ u 2 t t ‖ 2 ( ρ + 1 ) ( ‖ u 1 t ‖ 2 ( ρ + 1 ) ρ − 1 + ‖ u 2 t ‖ 2 ( ρ + 1 ) ρ − 1 ) ‖ ω t ‖ 2 ( ρ + 1 ) ‖ ω ‖ ≤ C 1 ‖ ∇ u 2 t t ‖ ( ‖ ∇ u 1 t ‖ ρ − 1 + ‖ ∇ u 2 t ‖ ρ − 1 ) ‖ ∇ ω t ‖ ‖ ω ‖ ≤ C B ‖ ∇ u 2 t t ‖ ‖ ω ‖ , (4.55)
| A 2 | ≤ ‖ ∇ ω ‖ ‖ ∫ 0 + ∞ g ( s ) ∇ ζ t ( s ) d s ‖ ≤ ε ‖ ∇ ω ‖ 2 + 1 − l 4 ε ‖ ζ t ‖ 2 , ∀ ε ∈ ( 0,1 ) , (4.56)
| A 3 | ≤ ‖ σ 1 − σ 2 ‖ ‖ ω ‖ , (4.57)
| A 4 | ≤ ‖ ∇ p ‖ ‖ ω ‖ , (4.58)
| A 5 | ≤ C 1 ‖ u 1 t t ‖ 2 ( ρ + 1 ) ‖ u 1 t ‖ 2 ( ρ + 1 ) ρ − 1 ‖ ω t ‖ 2 ( ρ + 1 ) ‖ ω ‖ + C 1 ‖ u 1 t ‖ ρ + 2 ρ ‖ ω t ‖ ρ + 2 2 ≤ C B ‖ ∇ u 1 t t ‖ ‖ ω ‖ + C B ‖ ∇ ω t ‖ 2 . (4.59)
By virtue of (4.46), we have
‖ ∇ ω ‖ 2 = 2 l E ω , p ( t ) − 1 l ∫ Ω | u 1 t | ρ ω t 2 d x − 1 l ‖ ∇ ω t ‖ 2 − 1 l ‖ ∇ p ‖ 2 − 1 l ‖ ζ t ‖ M 2 . (4.60)
Then from (4.54)-(4.59), we can conclude
Φ ′ ω , p ( t ) ≤ − l − ε 2 ‖ ∇ ω ‖ 2 − l − ε 2 ( 2 l E ω , p ( t ) − 1 l ∫ Ω | u 1 t | ρ ω t 2 d x − 1 l ‖ ∇ ω t 2 ‖ − 1 l ‖ ∇ p ‖ 2 − 1 l ‖ ζ t ‖ M 2 ) + C B ‖ ∇ ω t ‖ 2 + C B ‖ ω ‖ ( ‖ ∇ u 1 t t ‖ + ‖ ∇ u 2 t t ‖ + ‖ ∇ p ‖ ) + C 1 ‖ σ 1 − σ 2 ‖ ‖ ω ‖ + 1 − l 4 ε ‖ ζ t ‖ M 2 ≤ − l − ε 2 ‖ ∇ ω ‖ 2 − ( l − ε ) E ω ( t ) + C ε ‖ ∇ ω t ‖ 2 + C ε ‖ ∇ p ‖ 2 + C ε ‖ ζ t ‖ M 2 + C 1 ‖ σ 1 − σ 2 ‖ ‖ ω ‖ + C B ‖ ω ‖ ( ‖ ∇ u 1 t t ‖ + ‖ ∇ u 2 t t ‖ + ‖ ∇ p ‖ ) . (4.61)
Now we define
Ψ ω , p ( t ) = ∫ Ω ( Δ ω t − | u 1 t | ρ ω t ) ( ∫ 0 + ∞ g ( s ) ζ t ( s ) d s ) d x . (4.62)
From (4.42)-(4.43) and integration by parts, we derive
Ψ ′ ω , p ( t ) = ∫ Ω u 2 t t ( | u 1 t | ρ − | u 2 t | ρ ) ∫ 0 + ∞ g ( s ) ζ t ( s ) d s d x + l ∫ Ω ∇ ω ∫ 0 + ∞ g ( s ) ∇ ζ t ( s ) d s d x + ∫ Ω ( ∫ 0 + ∞ g ( s ) ∇ ζ t ( s ) d s ) 2 d x − ∫ Ω Δ ω t ∫ 0 + ∞ g ( s ) ζ t ( s ) d s d x + ∫ Ω ∇ p ∫ 0 + ∞ g ( s ) ζ t ( s ) d s d x − ∫ Ω ( σ 1 − σ 2 ) ∫ 0 + ∞ g ( s ) ζ t ( s ) d s d x − ρ ∫ Ω | u 1 t | ρ − 1 u 1 t t ω t ∫ 0 + ∞ g ( s ) ζ t ( s ) d s d x + ∫ Ω Δ ω t ∫ 0 + ∞ g ( s ) ζ t t ( s ) d s d x − ∫ Ω | u 1 t | ρ ω t ∫ 0 + ∞ g ( s ) ζ t t ( s ) d s d x = ∑ i = 1 9 B i . (63)
Using Hölder’s inequality, Poincaré’s inequality and Theorem 4.1, we derive for any δ ∈ ( 0,1 ) ,
B 1 ≤ ‖ u 2 t t ‖ 2 ( ρ + 1 ) ( ‖ u 1 t ‖ 2 ( ρ + 1 ) ρ − 1 + ‖ u 2 t ‖ 2 ( ρ + 1 ) ρ − 1 ) ‖ ω t ‖ ‖ ∫ 0 + ∞ g ( s ) ζ t ( s ) d s ‖ 2 ( ρ + 1 ) ≤ C B ‖ ∇ u 2 t t ‖ ‖ ω t ‖ ‖ ∫ 0 + ∞ g ( s ) ∇ ζ t ( s ) d s ‖ ≤ C B ( 1 − l ) 1 2 ‖ ∇ u 2 t t ‖ ‖ ω t ‖ ‖ ζ t ‖ M ≤ C B ( 1 − l ) 1 2 ‖ ∇ u 2 t t ‖ ‖ ω t ‖ ( ‖ η 1 t ‖ M + ‖ η 2 t ‖ M ) ≤ C ′ B ‖ ∇ u 2 t t ‖ ‖ ω t ‖ , (4.64)
B 2 ≤ δ ‖ ∇ ω ‖ 2 + ( 1 − l ) l 2 4 δ ‖ ζ t ‖ M 2 (4.65)
B 3 ≤ ( 1 − l ) ‖ ζ t ‖ M 2 , (4.66)
B 4 ≤ δ ‖ ∇ ω t ‖ 2 + λ 2 4 δ ( 1 − l ) ‖ ζ t ‖ M 2 , (4.67)
B 5 ≤ δ ‖ ∇ p ‖ 2 + ( 1 − l ) l 2 4 δ ‖ ζ t ‖ M 2 (4.68)
B 6 ≤ λ ( 1 − l ) 1 2 ‖ σ 1 − σ 2 ‖ ‖ ζ t ‖ M ≤ C 1 ‖ σ 1 − σ 2 ‖ 2 + C 1 ‖ ζ t ‖ M 2 (4.69)
B 7 ≤ ‖ u 1 t ‖ 2 ( ρ + 1 ) ρ − 1 ‖ u 1 t t ‖ 2 ( ρ + 1 ) ‖ ω t ‖ ‖ ∫ 0 + ∞ g ( s ) ζ t ( s ) d s ‖ 2 ( ρ + 1 ) ≤ C B ‖ ∇ u 1 t t ‖ ‖ ω t ‖ ‖ ∫ 0 + ∞ g ( s ) ∇ ζ t ( s ) d s ‖ ≤ C B ( 1 − l ) 1 2 ‖ ∇ u 1 t t ‖ ‖ ω t ‖ ‖ ζ t ‖ M ≤ C ′ B ‖ ∇ u 1 t t ‖ ‖ ω t ‖ . (4.70)
Noting that
∫ 0 + ∞ g ( s ) ζ t t ( s ) d s = ∫ 0 + ∞ g ( s ) d s ⋅ ω t − ∫ 0 + ∞ g ( s ) ζ s t ( s ) d s = ( 1 − l ) ω t + ∫ 0 + ∞ g ′ ( s ) ζ t ( s ) d s ,
then we see that
B 8 ≤ − ( 1 − l ) ‖ ∇ ω t ‖ 2 − ∫ 0 + ∞ g ′ ( s ) ‖ ∇ ω t ( t ) ‖ ‖ ∇ ζ t ( s ) ‖ d s ≤ − 1 − l 2 ‖ ∇ ω t ‖ 2 − 1 2 ( 1 − l ) ∫ 0 + ∞ g ′ ( s ) ‖ ∇ ζ t ( s ) ‖ 2 d s , (4.71)
B 9 ≤ − ( 1 − l ) ∫ Ω | u t | ρ ω t 2 d x + ‖ u 1 t ‖ 2 ( ρ + 1 ) ρ ‖ ω t ‖ 2 ( ρ + 1 ) ‖ ∫ 0 + ∞ g ′ ( s ) ζ t ( s ) d s ‖ ≤ − ( 1 − l ) ∫ Ω | u t | ρ ω t 2 d x + δ ‖ ∇ ω t ‖ 2 − C 1 λ 2 g ( 0 ) 4 δ ∫ 0 + ∞ g ′ ( s ) ‖ ∇ ζ t ( s ) ‖ 2 d s . (4.72)
Plugging (4.64)-(4.72) into (4.63), we get
Ψ ′ ω , p ( t ) ≤ C ′ B ( ‖ ∇ u 1 t t ‖ + ‖ ∇ u 2 t t ‖ ) ‖ ω t ‖ + C 1 ‖ ζ t ‖ M 2 + δ ( ‖ ω t ‖ 2 + ‖ ∇ p ‖ 2 ) + C 1 ‖ σ 1 − σ 2 ‖ 2 − ( 1 − l 2 − 2 δ ) ‖ ∇ ω t ‖ 2 − C 1 ∫ 0 + ∞ g ′ ( s ) ‖ ∇ ζ t ( s ) ‖ 2 d s − ( 1 − l ) ∫ Ω | u 1 t | ρ ω t 2 d x . (4.73)
On the other hand, we can get
| Ψ ω , p ( t ) | ≤ ( 1 − l ) 1 2 ‖ ∇ ω t ‖ ‖ ζ t ‖ M + ‖ u 1 t ‖ ρ + 2 ρ ‖ ω t ‖ ρ + 2 ‖ ∫ 0 + ∞ g ( s ) ζ t ( s ) d s ‖ ≤ ( 1 − l ) 1 2 ‖ ∇ ω t ‖ ‖ ζ t ‖ M + C 1 ( 1 − l ) 1 2 λ ‖ ∇ u 1 t ‖ ρ ‖ ∇ ω t ‖ ‖ ζ t ‖ M ≤ C B ( ‖ ∇ ω t ‖ 2 + ‖ ζ t ‖ M 2 ) ≤ C B E ω , p ( t ) . (4.74)
Define
G ω , p ( t ) = M E ω , p ( t ) + ε Φ ω , p ( t ) + Ψ ω , p ( t ) , (4.75)
which, together with (4.53) and (4.74), yields
( M − C B ε − C B ) E ω , p ( t ) ≤ G ω , p ( t ) ≤ ( M + C B ε + C B ) E ω , p ( t ) . (4.76)
Now we take ε > 0 so small and M so large that
M 2 E ω ( t ) ≤ G ω ( t ) ≤ 2 M E ω ( t ) . (4.77)
Then for any t ≥ τ , we have
G ′ ω , p ( t ) ≤ − ( l − ε ) ε E ω , p ( t ) − ( M 2 ξ 2 − C 1 ξ 2 − C 1 ) ‖ ζ t ‖ M 2
− ( M 2 − C ε − δ ) ‖ ∇ p ‖ 2 − ( l − ε 2 − δ ) ‖ ∇ ω ‖ 2 − ( M + 1 − l 2 − 2 δ − C ε ε ) ‖ ∇ ω t ‖ 2 + C ( B , M , ε ) ( ‖ ∇ u 1 t t ‖ + ‖ ∇ u 2 t t ‖ + ‖ ∇ p ‖ ) ( ‖ ω t ‖ + ‖ ω ‖ ) + C ( B , M , ε ) ( ‖ ω t ‖ + ‖ ω ‖ ) ‖ σ 1 − σ 2 ‖ + C 1 ( ‖ σ 1 − σ 2 ‖ 2 + ‖ f 1 − f 2 ‖ 2 ) . (4.78)
Now we take δ > 0 and ε > 0 so small that
M + 1 − l 2 − 2 δ − C ε ε > 0 , l − ε 2 − δ > 0 , M 2 − C ε − δ > 0.
For fixed ε and δ , we choose M so large that
M 2 ξ 2 − C 1 ξ 2 − C 1 > 0.
Then there exist some constant β > 0 such that
G ′ ω , p ( t ) ≤ − β E ω , p ( t ) + C 1 ( ‖ ∇ u 1 t t ‖ + ‖ ∇ u 2 t t ‖ + ‖ ∇ p ‖ ) ( ‖ ω t ‖ + ‖ ω ‖ ) + C 1 ( ‖ σ 1 − σ 2 ‖ 2 + ‖ f 1 − f 2 ‖ 2 ) + C 1 ‖ σ 1 − σ 2 ‖ ( ‖ ω t ‖ + ‖ ω ‖ ) ≤ − β 2 M G ω , p ( t ) + C 1 ( ‖ ∇ u 1 t t ‖ + ‖ ∇ u 2 t t ‖ ) ( ‖ ω t ‖ + ‖ ω ‖ ) + C 1 ( ‖ σ 1 − σ 2 ‖ 2 + ‖ f 1 − f 2 ‖ 2 ) + C 1 ‖ σ 1 − σ 2 ‖ ( ‖ ω t ‖ + ‖ ω ‖ ) . (4.79)
Integrating (4.79) over ( τ , t ) with respect to t, we derive
G ω , p ( t ) ≤ G ω , p ( τ ) e − β 2 M ( t − τ ) + C 1 ∫ τ t e − β 2 M ( t − s ) ( ‖ σ 1 − σ 2 ‖ 2 + ‖ f 1 − f 2 ‖ 2 ) d s + C 1 ∫ τ t e − β 2 M ( t − s ) ( ‖ ∇ u 1 t t ‖ + ‖ ∇ u 2 t t ‖ + ‖ p ‖ ) ( ‖ ω t ‖ + ‖ ω ‖ ) d s + C 1 ∫ τ t e − β 2 M ( t − s ) ‖ σ 1 − σ 2 ‖ ( ‖ ω t ‖ + ‖ ω ‖ ) d s ≤ G ω , p ( τ ) e − β 2 M ( t − τ ) + C 1 ( ∫ τ t ( ‖ ω t ‖ 2 + ‖ ω ‖ 2 ) d s ) 1 2 + C 1 ∫ τ t ‖ σ 1 − σ 2 ‖ 2 d s + C 1 ( ∫ τ t ( ‖ σ 1 − σ 2 ‖ 2 + ‖ f 1 − f 2 ‖ 2 ) d s ) 1 2 ( ∫ τ t ( ‖ ω t ‖ 2 + ‖ ω ‖ 2 ) d s ) 1 2 . (4.80)
For any fixed ε ∈ ( 0,1 ) , we choose T > τ so large that
G ω , p ( τ ) e − β 2 M ( T − τ ) ≤ ε ,
which, together with (4.77) and (4.80), gives
E ω , p ( t ) ≤ ε + C 1 ( ∫ τ t ( ‖ ω t ‖ 2 + ‖ ω ‖ 2 ) d s ) 1 2 + C 1 ∫ τ t ( ‖ σ 1 − σ 2 ‖ 2 + ‖ f 1 − f 2 ‖ 2 ) d s + C 1 ( ∫ τ t ‖ σ 1 − σ 2 ‖ 2 d s ) 1 2 ( ∫ τ t ( ‖ ω t ‖ 2 + ‖ ω ‖ 2 ) d s ) 1 2 (4.81)
Let
ϕ T ( ( u 01 τ , u 11 τ , θ 01 τ , η 1 τ ) , ( u 02 τ , u 12 τ , θ 02 τ , η 2 τ ) ; G 1 , G 2 ) = ∫ τ t ∫ Ω ( σ 1 − σ 2 ) ω t d x d s + ∫ τ t ∫ Ω ( f 1 − f 2 ) p d x d s (4.82)
Then
E ω , p ( t ) ≤ ε + ϕ T ( ( u 01 τ , u 11 τ , θ 01 τ , η 1 τ ) , ( u 02 τ , u 12 τ , θ 02 τ , η 2 τ ) ; G 1 , G 2 ) . (4.83)
It suffices to show ϕ T ( ⋅ , ⋅ , ⋅ , ⋅ ) ∈ C o n t r ( B , Σ ) for each fixed T > τ . From the proof of existence theorem, we can deduce that for any fixed T > τ , and the bound B depends on T,
∪ G ∈ Σ ∪ t ∈ [ τ , T ] U G ( t , τ ) B (4.84)
is bounded in H .
Let ( u n , u n t , θ n , η n t ) be the solutions corresponding to initial data ( u 0 n τ , u 1 n τ , θ 0 n τ , η n τ ) ∈ B with respect to symbol G n ∈ Σ , n = 1 , 2 , ⋯ . Then from (4.84), we get
u n → u ⋆ -weakly in L ∞ ( 0, T ; H 0 1 ( Ω ) ) , (4.85)
u n t → u t ⋆ -weakly in L ∞ ( 0, T ; H 0 1 ( Ω ) ) , (4.86)
θ n → θ ⋆ -weakly in L ∞ ( 0, T ; H 0 1 ( Ω ) ) . (4.87)
Taking u 1 = u n , u 2 = u m , θ 1 = θ n , θ 2 = θ m , σ 1 = σ n , f 1 = f n , f 2 = f m , σ 2 = σ m , noting that compact embedding H 0 1 ( Ω ) ↣ L 2 ( Ω ) , passing to a subsequence if necessary, we have
u n and u n t converge strongly in C ( [ τ , T ] ; L 2 ( Ω ) ) .
Therefore we get
∫ τ T ‖ u n t − u m t ‖ 2 d s → 0 , as m , n → + ∞ , (4.88)
∫ τ T ‖ u n − u m ‖ 2 d s → 0, as m , n → + ∞ , (4.89)
∫ τ T ‖ θ n − θ m ‖ 2 d s → 0, as m , n → + ∞ . (4.90)
On the other hand, by σ m , σ n , f m , f n ∈ Σ , we see that
∫ τ T ‖ σ n − σ m ‖ 2 d s → 0, as m , n → + ∞ , (4.91)
∫ τ T ‖ f n − f m ‖ 2 d s → 0, as m , n → + ∞ . (4.92)
Hence it follows from (4.88)-(4.92)
ϕ T ( ( u n , u n t , θ n , η n t ) , ( u m , u m t , θ m , η m t ) ; G n , G m ) → 0 as m , n → + ∞ , (4.93)
that is, ϕ T ∈ C o n t r ( B , Σ ) .
Therefore by Lemma 3.1, the semigroup { U G ( t , τ ) } ( t ≥ τ > 0 , G ∈ Σ ) is uniformly asymptotically compact and the proof is now complete.
Proof of Theorem 2.3. Combining Theorems 4.1-4.2, we can complete the proof of Theorem 2.3.
Shanghai Polytechnical University and the key discipline Applied Mathematics of Shanghai Polytechnic University with contract number XXKPY1604.
The author declares no conflicts of interest regarding the publication of this paper.
Ma, Z.Y. (2018) Uniform Attractors for a Non-Autonomous Thermoviscoelastic Equation with Strong Damping. Journal of Applied Mathematics and Physics, 6, 2475-2497. https://doi.org/10.4236/jamp.2018.612209