This work is licensed under the Creative Commons Attribution 4.0 International License.
Introduction
We consider the following nonlinear evolution equation
\left\{ {\begin{array}{*{20}{c}}
{{u_{tt}} + {{( - \Delta )}^m}u + {{( - \Delta )}^m}{u_t} + {{( - \Delta )}^m}{u_{tt}} + g(x,u) = f(x),}&{(x,t) \in \Omega \times (0,\infty ),}\\
{u(x,0) = {u_0}(x),{u_t}(x,0) = {u_1}(x),}&{x \in \Omega ,}\\
{\frac{{{\partial ^i}u(x,t)}}{{{\partial ^i}{v^i}}} = 0,i = 1,2, \ldots ,m - 1,}&{(x,t) \in \partial \Omega \times [0,\infty ),}
\end{array}} \right.
where in a bounded domain Ω ⊂ Rn with smooth boundary ∂Ω, the assumption on f, g, u0 and u1 will be made below.
When m = 1, the equation (1.1) is following form
{u_{tt}} - \Delta u - \Delta {u_t} - \Delta {u_{tt}} + g(x,u) = f(x).
Chen and Wang [19] proved the existence of global attractor for the problem (1.2). Lately, Xie and Zhong in [8] studied the existence of global attractor of solution for the problem (1.1) with f = 0. Also, there are some authors studied the existence and nonexistence, asymptotic behavior of global solution for (1.2) (see [2, 3, 4, 5, 6, 7] for more details ). Nakao and Yang in [9] showed the global attractor of the Kirchhoff type wave equation.
In this paper, we improve our result by adopting and modifying the method of [19], we studied more general form of the equation.
This paper is organized as follows: In section 2, we give some assumptions and state the main results. In section 3, we prove the global existence of solution using the Faedo-Galerkin method. Also, we write some important estimates for the solution. In section 4, the existence of the global attractor is proved. In Section 5, the proof of decay property for solution is showed.
Preliminaries and main results
We write the Sobolev space Hk (Ω) = Wk,2 (Ω),
H_0^k(\Omega ) = W_0^{k,2}(\Omega )
. Furthermore, we show by (.,.) the inner product of L2 (Ω), by ‖.‖p the norm of Lp (Ω), p ≥ 1 and by ‖.‖E the norm of any other Banach space E. As usual, we give u(t) instead of u(x,t), and u′ (t) for ut (t) and so on.
Later, we assume H1 = ‖h1‖2, H2 = max {‖h2‖2, ‖h2‖n/2}.
Clearly, the function g(x,u) = a(x)|u|α−1u − b(x)|u|β−1u(1 ≤ β < α) supplies (2.1) and (2.2) for some a(x), b(x).
Next, we show the definition and lemmas relating to the global attractor, (see [9, 11, 12]).
Definition 1
Suppose that E is Banach space and {S (t)}t≥0a semigroup on E. A set A ⊂ E is said a (E,E)−global attractor if and only iff
A is never changing (invariant), namely, S (t)A = A for whole t ≥ 0;
A is compact in E;
A is a bounded set in E and absorbs all bounded subset B in E relating with E topology, that is, for whichever bounded subset B ⊂ E,
dis{t_E}(S(t)B,A*) = \mathop {\sup \inf }\limits_{y \in {B^{x \in A*}}} ||S(t)y - x|{|_E} \to 0\,as\,t \to \infty .
Lemma 2
Assume E is Banach space and {S (t)}t ≥0is a semigroup of continuous operators on E. Then, there exists (E,E)−global attractor A if the following conditions are supplied:
There exists a bounded absorbing set B0in E, that is, for whichever bounded subset B ⊂ E, there is a T = T (B) such that S (t)B ⊂ B0for any t ≥ T.
{S (t)}t≥0as asymptotically compact in E, that is, for any bounded sequence {yn} in E and tn→ ∞ as n → ∞,
\left\{ {S\left( {t_n } \right)y_n } \right\}_{n = 1}^\inftyhas a convergent subsequence relating to E topology.
Show the solution in Theorem 1 by S (t)(u0,u1) = (u(t),ut (t)). We are now in a position to prove some continuity of S (t) relating to the initial data (u0,u1), which will be needed for the proof of the existence of global attractor.
Theorem 4
Suppose whole conditions in Theorem 3. Assume S (t)(u0k,u1k) and S (t)(u0,u1) are the solutions of the problem (1.1) with the initial data (u0k,u1k) and (u0,u1). If (u0k,u1k) → (u0,u1) in X as k → ∞, then S (t)(u0k,u1k) → S (t)(u0,u1) in X as k → ∞.
Theorem 4 denotes that the semigroup S (t) : X → X is continuous on X.
Theorem 5
Assume every assumptions in Theorem 3 be provided. Then, the semigroup {S (t)}t≥0related with the solution of the problem (1.1) accepts a (X,X)−global attractor A.
For the decay property of solution u(t) for the problem (1.1), we get
Theorem 6
Suppose u is a weak solution in Theorem 3 with f = 0 and g(x,u) = g(u). Besides, suppose 0 ≤ 2G(u) ≤ ug(u). Then, for whichever q > 0, there is{C_1} = {C_1}\left( {{{\left\| {{P^{\frac{1}{2}}}{u_0}} \right\|}_2},{{\left\| {{P^{\frac{1}{2}}}{u_1}} \right\|}_2}} \right)such thatE(t) = \frac{1}{2}\left( {\left\| {u(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_t}(t)} \right\|_2^2} \right) + \int_\Omega {G(} u(t))dx \le {C_1}{(1 + t)^{ - 1/q}}.
The Proof of Theorem 3
In this section, we suppose that all assumptions in Theorem 3 are supplied. Firstly, we establish the global existence of a solution to problem (1.1) with Fadeo-Galerkin method as in [16, 17].
Assume ωj (x) (j = 1,2,...) is the complete set of properly normalized eigenfunctions for the operator (−Δ)m in
H_0^m(\Omega )
. Then, the family {ω1,ω2...,ωk,...} holds an orthogonal basis for both
H_0^m(\Omega )
and L2 (Ω), see [16, 17]. For each positive integer k, show Vk = span{ω1,ω2...,ωk,...}. We search for an approximation solution uk (t) to the problem (1.1) in the form
{u_k}(t) = \sum\limits_{j = 1}^k {{d_{jk}}(t){\omega _j}}
where djk (t) are the solution of the nonlinear ordinary differential equation (ODE) system in the variant t:
(u_k^{\prime\prime},{\omega _j}) - (P{u_k},{\omega _j}) - (Pu_k^\prime,{\omega _j}) - (Pu_k^{\prime\prime},{\omega _j}) + (g,{\omega _j}) = (f,{\omega _j}),j = 1,2, \ldots ,k,
with the initial conditions
{d_{jk}}(0) = ({u_{0k}},{\omega _j}),d_{jk}^\prime(0) = ({u_{1k}},{\omega _j})
where u0k and u1k are chosen in Vk so that
{u_{0k}} \to {u_0},{u_{1k}} \to {u_1}\,{\rm{in}}\,{H^{2m}}(\Omega ) \cap H_0^m(\Omega )\,{\rm{as}}\,k\, \to \infty .
Here (.,.) shows the inner product in L2 (Ω). Then, Sobolev imbedding theorem means that ∃c0> 0, such that
\left\| {{u_k}(0)} \right\|_{H_0^m}^2 \le {c_0}\left\| {{P^{\frac{1}{2}}}{u_0}} \right\|_2^2,\left\| {u_k^\prime(0)} \right\|_{H_0^m}^2 \le {c_0}\left\| {{P^{\frac{1}{2}}}{u_1}} \right\|_2^2\forall k = 1,2, \ldots ,
and (3.1) shows that for any v ∈ Vk,
\left( {u_k^{\prime\prime},v} \right) - \left( {P{u_k},v} \right) - \left( {Pu_k^\prime,v} \right) - \left( {Pu_k^{\prime\prime},v} \right) + (g,v) = (f,v),\forall v \in {V_k}.
We know, the system (3.1) and (3.2) accept a unique solution uk (t) on the interval [0,T ] for any T > 0. Such a solution can be expanded to the overall interval [0,∞). We show by Ci (i = 1,2,...) the constants that are independent of k and t ≥ 0, by C0 the constant depending on k1, k2 in (A2) and Sobolev imbedding constant c0 in (3.4). These constants may be different from line to line.
Multiplying (3.1) by d′jk (t) and summing the resulting equations over j, we obtain
E_1^\prime(t) + \left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2 = 0,\forall t \ge 0
where
{E_1}(t) = \frac{1}{2}\left( {\left\| {u_k^\prime(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_k}(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2} \right) + \int_\Omega {G(x,{u_k}(t))dx - } \int_\Omega {f(x){u_k}(t)dx.}
Also, multiplying (3.1) by djk (t), we get
E_2^1(t) + \left\| {{P^{\frac{1}{2}}}{u_k}(t)} \right\|_2^2 + \int_\Omega {g(x,{u_k}){u_k}(t)dx = \left\| {u_k^\prime(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2 + } \int_\Omega {f(x){u_k}(t)dx}
where
{E_2}(t) = \frac{1}{2}\left\| {{P^{\frac{1}{2}}}{u_k}(t)} \right\|_2^2 + \int_\Omega {{u_k}(t)u_k^1(t)dx + } \int_\Omega {{P^{\frac{1}{2}}}{u_k}(t){P^{\frac{1}{2}}}u_k^\prime(t)dx.}
If we take sufficient large k1> 0 and use the assumption (A2), we get
\psi _k^\prime(t) + {\lambda _1}{\psi _k}(t) \le {C_0}({F^2} + H_1^{2m}),{\psi _k}(t) = {k_1}{E_1}(t) + {E_2}(t)
with some positive λ 1, relating to the indicated constants in (A2).
Also, we differentiate (3.1) with respect to t and get
\left( {u_k^{\prime\prime\prime},{\omega _j}} \right) - \left( {Pu_k^\prime,{\omega _j}} \right) - \left( {Pu_k^{\prime\prime},{\omega _j}} \right) - \left( {Pu_k^{\prime\prime\prime},{\omega _j}} \right) + \left( {{g_u}u_k^\prime,{\omega _j}} \right) = 0,j = 1,2, \ldots ,k.
Multiplying (3.14) by
d_{jk}^{\prime\prime}(t)
and summing the resulting equations over j, we obtain
E_3^\prime(t) + \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(t)} \right\|_2^2 + \int_\Omega {{g_u}u_k^\primeu_k^{\prime\prime}dx = 0}
with
{E_3}(t) = \frac{1}{2}\left( {\left\| {u_k^{\prime\prime}(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(t)} \right\|_2^2} \right) \le {C_0}{\left( {\left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2} \right)_2},t \ge 0
in which the Sobolev embedding theorem has been used.
Furthermore, the growth condition (2.2) and the Hölder inequality mean that
\int_\Omega {\left| {{g_u}u_k^\primeu_k^{\prime\prime}} \right|dx \le } {k_2}\int_\Omega {\left( {\left| {{h_2}} \right|\left| {u_k^\prime} \right| + {{\left| {{u_k}} \right|}^{\alpha - 1}}\left| {u_k^\prime} \right|\left| {u_k^{\prime\prime}} \right|} \right)dx \le {C_0}} \left( {{{\left\| {{h_2}} \right\|}_{n/2}} + \left\| {{P^{\frac{1}{2}}}{u_m}} \right\|_2^{({\rm{\alpha }} - 1)}} \right){\left\| {{P^{\frac{1}{2}}}u_k^\prime} \right\|_2}{\left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}} \right\|_2}.
Therefore, we get
\int_\Omega {\left| {{g_u}u_k^\primeu_k^{\prime\prime}} \right|dx \le } \frac{1}{2}\left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(t)} \right\|_2^2 + {C_0}\left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2\left( {\left\| {{P^{\frac{1}{2}}}{u_k}(t)} \right\|_2^{2({\rm{\alpha }} - 1)} + H_2^{2m}} \right)
and
E_3^\prime(t) + \frac{1}{2}\left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(t)} \right\|_2^2 \le {C_0}\left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2\left( {\left\| {{P^{\frac{1}{2}}}{u_k}(t)} \right\|_2^{2({\rm{\alpha }} - 1)} + H_2^{2m}} \right).
Then, the applications of the estimates (3.13) and (3.15)–(3.18) give that ∃λ 1 ≥ λ 2 > 0, depending on C0, such that
E_3^\prime(t) + {\lambda _2}{E_3}(t) \le {C_0}\left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2\left( {1 + \left\| {{P^{\frac{1}{2}}}{u_k}(t)} \right\|_2^{2({\rm{\alpha }} - 1)} + H_2^{2m}} \right) \le {C_3}{e^{ - {\lambda _1}t}} + {C_4}.
Here, assume
{C_3} = {C_3}\left( {{{\left\| {{P^{\frac{1}{2}}}{u_0}} \right\|}_2},{{\left\| {{P^{\frac{1}{2}}}{u_1}} \right\|}_2},F,{H_1},{H_2}} \right)
, C4 = C4 (,F,H1,H2). Then (3.19) means that
{E_2}(t) \le {E_3}(0){e^{ - {\lambda _2}t}} + {C_3}{e^{ - {\lambda _2}t}} + \lambda _2^{ - 1}{c_4},t \ge 0.
We show that E3 (0) is uniformly bounded for k under the conditions in Theorem 3 now. It follows by (3.1) that
\left( {u_k^{\prime\prime}(t) - P{u_k}(t) - Pu_k^\prime(t) - Pu_k^{\prime\prime}(t),u_k^{\prime\prime}(t)} \right) = \left( {f,u_k^\prime(t)} \right) - \left( {g,u_k^{\prime\prime}(t)} \right).
Especially, suppose t = 0, we get
\begin{array}{*{20}{l}}
{\left\| {u_k^{\prime\prime}(0)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0)} \right\|_2^2 + \int_\Omega {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0).\left( {{P^{\frac{1}{2}}}{u_k}(0) + {P^{\frac{1}{2}}}u_k^\prime(0)} \right)} dx}\\
{ = \int_\Omega {f(x)u_k^{\prime\prime}(0)dx - \int_\Omega {\left( {g,{u_k}(0)} \right)} u_k^{\prime\prime}(0)dx.} }
\end{array}
By Young inequality with ɛ,
\begin{array}{*{20}{c}}
{\int_\Omega {\left| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0).{P^{\frac{1}{2}}}{u_k}(0)} \right|dx \le \varepsilon \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0)} \right\|_2^2 + {C_\varepsilon }\left\| {{P^{\frac{1}{2}}}{u_k}(0)} \right\|_2^2,} }\\
{\int_\Omega {\left| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0).{P^{\frac{1}{2}}}u_k^\prime(0)} \right|dx \le \varepsilon \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0)} \right\|_2^2 + {C_\varepsilon }\left\| {{P^{\frac{1}{2}}}u_k^\prime(0)} \right\|_2^2,} }\\
{\int_\Omega {\left| {g\left( {x,{u_k}(0)} \right)u_k^{\prime\prime}(0)dx} \right| \le {{\left\| {u_k^{\prime\prime}(0)} \right\|}_{\frac{{2n}}{{n - 2}}}}{{\left\| g \right\|}_{{\mu _1}}} \le \varepsilon \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0)} \right\|_2^2 + {C_\varepsilon }\left\| g \right\|_{{\mu _1}}^2,} }
\end{array}
and
\int_\Omega {\left| {f(x)u_k^{\prime\prime}(0)} \right|dx \le } \varepsilon \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0)} \right\|_2^2 + {C_\varepsilon }\left\| f \right\|_2^2
with μ1 = 2n/(n + 2). Since μ1α = 2nα/(n + 2) ≤ 2n/(n − 2), we obtain by (2.2) that
\int_\Omega {{{\left| g \right|}^{{\mu _1}}}dx \le {C_0}} \int_\Omega {\left( {{{\left| {{u_k}(0)} \right|}^{{\mu _1}{\rm{\alpha }}}} + {{\left| {{h_2}} \right|}^{{\mu _1}}}} \right)dx \le {C_0}\left( {\left\| {{P^{\frac{1}{2}}}{u_0}} \right\|_2^{{\mu _1}{\rm{\alpha }}} + \left\| {{h_2}} \right\|_2^{{\mu _1}}} \right).}
Suppose 0 < ɛ ≤ 1/6. Then, from (3.22) to (3.24) that
\begin{array}{*{20}{l}}
{{E_3}(0)}&{ \le \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(0)} \right\|_2^2 + \left\| {u_k^{\prime\prime}(0)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^\prime(0)} \right\|_2^2}\\
{}&{ \le {C_0}\left( {\left\| {{P^{\frac{1}{2}}}u_k^\prime(0)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_k}(0)} \right\|_2^2 + {F^2} + \left\| g \right\|_{{\mu _1}}^2} \right)}\\
{}&{ \le {C_0}\left( {\left\| {{P^{\frac{1}{2}}}{u_1}} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_0}} \right\|_2^2 + {F^2} + \left\| {{P^{\frac{1}{2}}}{u_0}} \right\|_2^{2{\rm{\alpha }}} + \left\| {{h_2}} \right\|_2^2} \right) \equiv {C_3}.}
\end{array}
Therefore, the inequality (3.20) shows
\left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^\prime(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u_k^{\prime\prime}(t)} \right\|_2^2 \le {C_3}{e^{ - {\lambda _2}t}} + \lambda _2^{ - 1}{C_4},t \ge 0
and the estimates (3.13) and (3.26) give that
\left\{ {\begin{array}{*{20}{l}}
{\left\{ {{u_k}(t)} \right\}{\rm{is}}\,{\rm{bounded}}\,{\rm{in}}\,{L^\infty }([0,\infty );H_0^m(\Omega )),}\\
{\left\{ {u_k^\prime(t)} \right\}{\rm{is}}\,{\rm{bounded}}\,{\rm{in}}\,{L^\infty }([0,\infty );H_0^m(\Omega )),}\\
{\left\{ {u_k^{\prime\prime}(t)} \right\}{\rm{is}}\,{\rm{bounded}}\,{\rm{in}}\,{L^\infty }([0,\infty );H_0^m(\Omega )).}
\end{array}} \right.
So, there exists a subsequences in {uk} (still showed by {uk}) such that
\left\{ {\begin{array}{*{20}{l}}
{{u_k} \to u\,{\rm{weakly}}\,{\rm{star}}\,{\rm{in}}\,{L^\infty }([0,\infty );H_0^m(\Omega )),}\\
{u_k^\prime \to {u^\prime}\,{\rm{weakly}}\,{\rm{star}}\,{\rm{in}}\,{L^\infty }([0,\infty );{L^2}(\Omega )),}\\
{u_k^{\prime\prime} \to {u^{\prime\prime}}\,{\rm{weakly}}\,{\rm{star}}\,{\rm{in}}\,{L^2}([0,\infty );H_0^m(\Omega )),.}
\end{array}} \right.
From applying the fact that L∞ ([0,∞);
\left. {H_0^m\left( \Omega \right)} \right) \hookrightarrow {L^2}\left( {|0,\infty } \right)\
;
\left. {H_0^m \left( \Omega \right)} \right)
) and the Lions-Aubin compactness Lemma in [20], we obtain from (3.27) and (3.28) that
{u_k} \to u,u_k^\prime \to \user1{strongly}\,\user1{in}\,{L^2}\left( {[0,\infty );{L^2}\left( \Omega \right)} \right)
and then uk→ u a.e in Ω× [0,∞).
Using the growth condition (2.2), for any T > 0, the integral
\int_0^T {\int_\Omega {{{\left| {g\left( {x,{u_k}} \right)\left( {x,t} \right)} \right|}^{\frac{{\alpha + 1}}{\alpha }}}dxdt} }
is bounded. Accordingly, by Lemma 2 in Chap. 1 [17], we conclude
g\left( {x,{u_k}} \right) \to g\left( {x,u} \right){\rm{weakly}}\,{\rm{in}}\,{L^{\frac{{\alpha + 1}}{\alpha }}}\left( {\left[ {0,T} \right];{L^{\frac{{\alpha + 1}}{\alpha }}}\left( \Omega \right)} \right)
with these convergences, by using the limit in the approximate equation (3.5), we get
\left( {u\prime\prime\left( t \right),v} \right) - \left( {Pu,v} \right) - \left( {Pu\prime,v} \right) - \left( {Pu\prime\prime,v} \right) + \left( {g\left( {x,u} \right),v} \right) = f\left( {f,v} \right),\forall v \in H_0^m\left( \Omega \right),
So, u(t) is a weak solution of (1.1) and supplies (2.5) and (2.6), and the proof of existence for the solution u(t) of (1.1) is completed.
We derive the estimates for ‖Pu(t)‖2 and ‖Put (t)‖2 now. Also, we write u instead of uk for convenience and view the estimates for u as a limit of uk. Supposing v = −Pu in (3.31), we obtain
E_4^\prime\left( t \right) + \left\| {Pu\left( t \right)} \right\|_2^2 \le \left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2^2 + \left\| {P{u_t}\left( t \right)} \right\|_2^2 + {C_0}\left( {{F^2} + \left\| g \right\|_2^2} \right)
with some C0> 0 and
{E_4}\left( t \right) = \frac{1}{2}\left\| {Pu\left( t \right)} \right\|_2^2 + \int_\Omega {{P^{\frac{1}{2}}}{u_t}} \left( t \right){P^{\frac{1}{2}}}u\left( t \right)dx + \int_\Omega {P{u_t}} \left( t \right)Pu\left( t \right)dx.
Also, assuming v = −Put in (3.31), we get
\begin{array}{*{20}{l}}
{\int_\Omega {P{u_t}\left( { - {u_{tt}} + Pu + P{u_{tt}}} \right)dx} + \left\| {P{u_t}} \right\|_2^2}&{ = \int_\Omega {g{P^{\frac{1}{2}}}{u_t}dx - } \int_\Omega {fP{u_t}dx} }\\
{}&{ \le \frac{1}{2}\left\| {P{u_t}} \right\|_2^2 + {C_0}\left( {{F^2} + \left\| g \right\|_2^2} \right).}
\end{array}
This means that
E_5^\prime + \frac{1}{2}\left\| {P{u_t}\left( t \right)} \right\|_2^2 \le {C_0}\left( {{F^2} + \left\| g \right\|_2^2} \right)
with
{E_5}\left( t \right) = \frac{1}{2}\left( {\left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2^2 + \left\| {P{u_t}\left( t \right)} \right\|_2^2 + \left\| {Pu\left( t \right)} \right\|_2^2} \right).
We note that
\left\| u \right\|_{2\alpha }^{2\alpha } \le {C_0}\left\| {{P^{\frac{1}{2}}}u} \right\|_2^{2\alpha \theta } + \left\| {Pu} \right\|_2^{2\alpha \left( {1 - \theta } \right)} \le \eta \left\| {Pu} \right\|_2^2 + {C_\eta }\left\| {{P^{\frac{1}{2}}}u} \right\|_2^{2\beta }
with small η > 0 and 2αθ = (n − 2)α − n < 2, β = α (1 − θ )/(1 − αθ ) > 0. Then, (3.37) shows
\left\| g \right\|_2^2 \le {C_0}\left( {\left\| u \right\|_{2\alpha }^{2\alpha } + H_2^{2m}} \right) \le \eta \left\| {Pu} \right\|_2^2{C_\eta }\left\| {{P^{\frac{1}{2}}}u} \right\|_2^{2\beta } + {C_0}H_2^{2m}.
Then, by (2.5), (3.35) and (3.38) that
\begin{array}{*{20}{l}}
{E_5^\prime\left( t \right) + \frac{1}{2}\left\| {Pu\left( t \right)} \right\|_2^2}&{ \le \left\| {Pu\left( t \right)} \right\|_2^2 + {C_\eta }\left\| {{P^{\frac{1}{2}}}u\left( t \right)} \right\|_2^{2\beta } + {C_0}\left( {{F^2} + H_2^{2m}} \right)}\\
{}&{ \le {C_1}{e^{ - {\lambda _1}\beta t}} + \eta \left\| {Pu\left( t \right)} \right\|_2^2 + {C_2}.}
\end{array}
Assume φ (t) = k1E5 (t) + E4 (t). We get from (3.32) and (3.39) that
{\phi ^\prime}\left( t \right) + \frac{{{k_1} - 1}}{2}\left\| {P{u_t}\left( t \right)} \right\|_2^2 + \left( {1 - \left( {1 + {k_1}/2} \right)\eta } \right)\left\| {Pu\left( t \right)} \right\|_2^2 \le {C_1}{e^{ - {\lambda _1}\beta t}} + {C_2}.
Suppose k1≥ 3 and η is small that 1 − η (1 + k1/2) ≥4/5. Then, (3.40) shows
{\phi ^\prime}\left( t \right) + \left\| {P{u_t}} \right\|_2^2 + \frac{1}{2}\left\| {Pu\left( t \right)} \right\|_2^2 \le {C_1}{e^{ - {\lambda _1}\beta t}} + {C_2}.
We note that
{E_4}\left( t \right) \le \frac{3}{5}\left\| {Pu} \right\|_2^2 + 3\left\| {P{u_t}\left( t \right)} \right\|_2^2 + \frac{1}{2}\left( {\left\| {{P^{\frac{1}{2}}}u} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_t}} \right\|_2^2} \right)
and
\begin{array}{*{20}{l}}
{\varphi \left( t \right)}&{ \le \left( {\frac{3}{5} + \frac{{{k_1}}}{2}} \right)\left\| {Pu} \right\|_2^2 + \left( {3 + \frac{{{k_1}}}{2}} \right)\left\| {P{u_t}} \right\|_2^2 + \frac{1}{2}\left( {\left\| {{P^{\frac{1}{2}}}u} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_t}} \right\|_2^2} \right)}\\
{}&{ \le {C_0}\left( {\left\| {Pu\left( t \right)} \right\|_2^2 + \left\| {P{u_t}\left( t \right)} \right\|_2^2 + {C_1}{e^{ - {\lambda _1}\beta t}} + {C_2}} \right).}
\end{array}
Also (3.41) and (3.43) give that ∃λ 1β ≥ λ 3 > 0, depending on C0, such that
\varphi \prime\left( t \right) + {\lambda _3}\varphi \left( t \right) \le {C_1}{e^{ - {\lambda _1}\beta t}} + {C_2},t \ge 0.
So,
\varphi \left( t \right) \le \varphi \left( 0 \right){e^{ - {\lambda _3}t}} + {C_1}{e^{ - {\lambda _3}t}} + {C_2}{\lambda _3}^{ - 1},t \ge 0.
Otherwise, we get
\begin{array}{*{20}{l}}
{\varphi \left( t \right)}&{ = {k_1}{E_4}\left( t \right) + {E_3}\left( t \right) \ge \frac{{{k_1}}}{2}\left( {\left\| {{P^{\frac{1}{2}}}{u_t}} \right\|_2^2 + \left\| {P{u_t}} \right\|_2^2 + \left\| {Pu} \right\|_2^2} \right)}\\
{}&{\,\,\,\,\, - \frac{1}{2}\left( {\left\| {{P^{\frac{1}{2}}}{u_t}} \right\|_2^2 + \left\| {P{u_t}} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u} \right\|_2^2} \right)}\\
{}&{ \ge \frac{{{k_1} - 1}}{2}\left\| {{P^{\frac{1}{2}}}{u_t}} \right\|_2^2 + \left( {\frac{{{k_1}}}{2} - 1} \right)\left\| {P{u_t}} \right\|_2^2 + \frac{{{k_1} - {c_0}}}{2}\left\| {Pu} \right\|_2^2}\\
{}&{ \ge \left\| {{P^{\frac{1}{2}}}{u_t}} \right\|_2^2 + \left\| {P{u_t}} \right\|_2^2 + \left\| {Pu} \right\|_2^2,}
\end{array}
where the facts k1≥ {4,2 + c0} and Sobolev imbedding theorem (see [17])
\left\| {{P^{\frac{1}{2}}}u} \right\|_2^2 \le {c_0}\left\| {Pu} \right\|_2^2\,\,\,\,\forall u \in {H^{2m}}\left( \Omega \right) \cap H_0^m\left( \Omega \right)
have been used. So, by the estimates (3.45) and (3.46) that
\left\| {{P^{\frac{1}{2}}}{u_t}} \right\|_2^2 + \left\| {Pu\left( t \right)} \right\|_2^2 + \left\| {P{u_t}\left( t \right)} \right\|_2^2 \le {C_5}{e^{ - {\lambda _3}t}} + {C_4}\lambda _3^{ - 1},t \ge 0
with C4 = C4 (F,H1,H2), C5 = C5 (‖Pu0‖2, ‖Pu1‖2 ,F,H1,H2).
To establish the uniqueness, we suppose that u(t) and v(t) are two solutions of (1.1), which supply the estimates (2.5)–(2.7) and u(0) = v(0), u′ (0) = v′ (0). Taking U (t) = ut (t), V (t) = vt (t) and W (t) = U (t) − V (t), then we see from (1.1) that
{W_t} - PW - P{W_t} - P\left( {u - v} \right) = g\left( {x,v} \right) - g\left( {x,u} \right),\,\,x \in \Omega ,\,\,t \ge 0.
Multiplying (3.48) by W, we obtain
\begin{array}{l}
\frac{1}{2}\frac{d}{{dt}}\left( {\left\| {W\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}W\left( t \right)} \right\|_2^2} \right) + \left\| {{P^{\frac{1}{2}}}W\left( t \right)} \right\|_2^2 + \int_\Omega {{P^{\frac{1}{2}}}\left( {u - v} \right){P^{\frac{1}{2}}}Wdx} \\
= \int_\Omega {\left( {g\left( {x,v} \right) - g\left( {x,u} \right)} \right)Wdx}
\end{array}
and
\begin{array}{l}
\left\| {W\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}W\left( t \right)} \right\|_2^2 + 2\int_0^t {\left\| {{P^{\frac{1}{2}}}W\left( s \right)} \right\|_2^2ds} + 2\int_0^t {\int_\Omega {{P^{\frac{1}{2}}}\left( {u\left( s \right) - v\left( s \right)} \right){P^{\frac{1}{2}}}W\left( s \right)dxds} } \\
= 2\int_0^t {\int_\Omega {\left( {g\left( {x,v\left( s \right)} \right) - g\left( {x,u\left( s \right)} \right)} \right)W\left( s \right)dxds.} }
\end{array}
Since
\left| {{P^{\frac{1}{2}}}\left( {u\left( s \right) - v\left( s \right)} \right)} \right| \le \int_0^s {\left| {{P^{\frac{1}{2}}}\left( {{u_\tau }\left( \tau \right) - {v_\tau }\left( \tau \right)} \right)} \right|d\tau } = \int_0^s {\left| {{P^{\frac{1}{2}}}W\left( \tau \right)} \right|d\tau }
then
{\left\| {{P^{\frac{1}{2}}}\left( {u\left( s \right) - v\left( s \right)} \right)} \right\|_2} \le {s^{1/2}}{\left( {\int_0^s {\left\| {{P^{\frac{1}{2}}}W\left( \tau \right)} \right\|_2^2d\tau } } \right)^{1/2}}
and
\begin{array}{*{20}{l}}
{\int_0^t {\int_\Omega {\left| {{P^{\frac{1}{2}}}\left( {u\left( s \right) - v\left( s \right)} \right){P^{\frac{1}{2}}}W\left( s \right)} \right|dxds} } }&{ \le \int_0^t {\int_\Omega {\int_0^s {\left| {{P^{\frac{1}{2}}}W\left( s \right)} \right|} \left| {{P^{\frac{1}{2}}}W\left( \tau \right)} \right|dxd\tau ds} } }\\
{}&{ \le \int_0^t {\int_0^s {{{\left\| {{P^{\frac{1}{2}}}W\left( s \right)} \right\|}_2}} {{\left\| {{P^{\frac{1}{2}}}W\left( \tau \right)} \right\|}_2}d\tau } ds}\\
{}&{ \le t\int_0^t {\left\| {{P^{\frac{1}{2}}}W\left( s \right)} \right\|_2^2ds.} }
\end{array}
Now, taking Uɛ (s) = ɛu(s) + (1 − ɛ)v(s), 0 ≤ ɛ ≤ 1, we get
\begin{array}{*{20}{l}}
G&{ = \int_0^t {\int_\Omega {\left| {g\left( {x,u\left( s \right)} \right) - g\left( {x,v\left( s \right)} \right)} \right|\left| {W\left( s \right)} \right|dxds = \int_0^t {\int_\Omega {\left| {\int_0^1 {\frac{d}{{d\varepsilon }}g\left( {x,{U_\varepsilon }} \right)d\varepsilon } } \right|\left| {W\left( s \right)} \right|dxds} } } } }\\
{}&{ \le \int_0^t {\int_\Omega {\int_0^1 {\left| {{g_u}\left( {x,{U_\varepsilon }} \right)} \right|\left| {u\left( s \right) - v\left( s \right)} \right|\left| {W\left( s \right)} \right|d\varepsilon dxds} } } }\\
{}&{ \le {k_2}\int_0^t {\int_\Omega {\left( {{{\left| u \right|}^{\alpha - 1}} + {{\left| v \right|}^{\alpha - 1}} + {h_2}\left( x \right)} \right)\left| {u\left( s \right) - v\left( s \right)} \right|\left| {W\left( s \right)} \right|dxds} } }\\
{}&{ \le {c_0}\int_0^t {\left( {\left\| {u\left( s \right)} \right\|_{{\sigma _1}}^{{\sigma _1}} + \left\| {v\left( s \right)} \right\|_{{\sigma _1}}^{{\sigma _1}} + \left\| {{h_2}\left( s \right)} \right\|_{{\sigma _2}}^{{\sigma _2}}} \right){{\left\| {{P^{\frac{1}{2}}}\left( {u\left( s \right) - v\left( s \right)} \right)} \right\|}_2}{{\left\| {{P^{\frac{1}{2}}}W\left( s \right)} \right\|}_2}ds} }
\end{array}
where σ1 = n(α − 1)/2 ≤2n/(n − 2), σ2 = n/2.
From (2.5) and Sobolev imbedding theorem, there is C3> 0 such that
\left\| {u\left( s \right)} \right\|_{{\sigma _1}}^{{\sigma _1}} + \left\| {v\left( s \right)} \right\|_{{\sigma _1}}^{{\sigma _1}} + \left\| {{h_2}} \right\|_{{\sigma _2}}^{{\sigma _2}} \le {C_0}\left( {\left\| {{P^{\frac{1}{2}}}u\left( s \right)} \right\|_2^{{\sigma _1}} + \left\| {{P^{\frac{1}{2}}}v\left( s \right)} \right\|_2^{{\sigma _1}} + \left\| {{h_2}} \right\|_{{\sigma _2}}^{{\sigma _2}}} \right) \le {C_3}\forall s \ge 0.
Then,
G \le {C_3}\int_0^t {{s^{1/2}}} {\left( {\int_0^s {\left\| {{P^{\frac{1}{2}}}W\left( \tau \right)} \right\|_2^2d\tau } } \right)^{1/2}}{\left\| {{P^{\frac{1}{2}}}W\left( s \right)} \right\|_2}ds \le {C_3}t\int_0^t {\left\| {{P^{\frac{1}{2}}}W\left( \tau \right)} \right\|_2^2d\tau } .
Then, the estimates (3.50)–(3.52) indicate that
\left\| {W\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}W\left( t \right)} \right\|_2^2 + 2\int_0^t {\left\| {{P^{\frac{1}{2}}}W\left( s \right)} \right\|_2^2 \le \left( {{C_3} + 1} \right)t\int_0^t {\left\| {{P^{\frac{1}{2}}}W\left( s \right)} \right\|_2^2ds} .}
The integral inequality (3.53) represents that there exists T1> 0, such that W (t) = 0 in [0,T1]. As a result, u(t) − v(t) = u(0) − v(0) = 0 in [0,T1].
Then, we conduce that u(t) = v(t) on [T1,2T1], [2T1,3T1],..., and u(t) = v(t) on [0,∞). This shows the proof of uniqueness.
Now, we establish u ∈ C ([0,∞);
\left. {H_0^m \left( \Omega \right)} \right)
). Assume t > s ≥ 0. Then,
\begin{array}{*{20}{l}}
{\left\| {{P^{\frac{1}{2}}}\left( {u\left( t \right)} \right) - u\left( s \right)} \right\|_2^2}&{ = {{\int_\Omega {\left| {\int_s^t {{P^{\frac{1}{2}}}{u_\tau }\left( \tau \right)d\tau } } \right|} }^2}dx}\\
{}&{ \le \left( {t - s} \right)\int_s^t {\left\| {{P^{\frac{1}{2}}}{u_\tau }\left( \tau \right)d\tau } \right\|_2^2d\tau } \to 0\,{\rm{as}}\,t \to s.}
\end{array}
This indicates u(t) ∈ C ([0,∞);
\left. {H_0^m \left( \Omega \right)} \right)
. Also, we get
\begin{array}{*{20}{l}}
{\left\| {P\left( {u\left( t \right) - u\left( s \right)} \right)} \right\|_2^2}&{ = \int_\Omega {{{\left| {\int_s^t {P{u_\tau }\left( \tau \right)d\tau } } \right|}^2}dx} }\\
{}&{ \le \left( {t - s} \right)\int_s^t {\left\| {P{u_\tau }\left( \tau \right)} \right\|_2^2d\tau } \to 0\,{\rm{as}}\,t \to s.}
\end{array}
and u(t) ∈ C [0,∞);
H^{2m} \left( \Omega \right) \cap \left. {H_0^m \left( \Omega \right)} \right)
.
Moreover, we get
\left\| {{P^{\frac{1}{2}}}\left( {{u_t}\left( t \right) - {u_t}} \right)\left( s \right)} \right\|_2^2 \le \left( {t - s} \right)\int_s^t {\left\| {{P^{\frac{1}{2}}}{u_{tt}}\left( \tau \right)} \right\|_2^2d\tau } \to 0\,{\rm{as}}\,t \to s.
This shows that u(t) ∈ C1 ([0,∞);
{H_0^m }
) and the proof of Theorem 3 is completed.
Global attractor for the problem (1)
By Theorem 3, we see that the solution operatör S (t)(u0,u1) = (u(t),ut (t)), t ≥ 0 of the problem (1.1) creates a semigroup on
X = \left( {H^{2m} \left( \Omega \right) \cap H_0^m \left( \Omega \right)} \right) \times \left( {H^{2m} \left( \Omega \right) \cap H_0^m \left( \Omega \right)} \right)
, which supplies these properties:
S (t) : X → X for all t ≥ 0;
S (t + s) = S (t)S (s) for t,s ≥ 0;
S (t)(u0,u1) → S (s)(u0,u1) in X as t → s for any (u0,u1) ∈ X.
For establishing the existence of the (X,X)-global attractor for the problem (1.1), firstly, we show the continuity of S (t) relating to the initial data (u0,u1).
The proof of Theorem 4
Suppose uk (t), u(t) is corresponding solution of the problem (1.1) with the initial data (u0k,u1k) and (u0,u1) respectively, k = 1,2,....
Since (u0k,u1k) → (u0,u1) in X, {(u0k,u1k)} is bounded in X. Set wk (t) = uk (t) − u(t). Then, wk holds
\left\{ {\begin{array}{*{20}{c}}
{w_k^{\prime\prime} - P{w_k} - Pw_k^\prime - Pw_k^{\prime\prime} = g\left( {x,u} \right) - g\left( {x,{u_k}} \right) = {G_k},\left( {x,t} \right) \in \Omega \times \left( {0,\infty } \right),}\\
{{w_k}\left( {x,0} \right) = {u_{0k}}\left( x \right) - {u_0}\left( x \right),w_k^\prime\left( {x,0} \right) = {u_{1k}}\left( x \right) - {u_1}\left( x \right),x \in \Omega ,}\\
{{w_k}\left( {x,t} \right) = 0,\left( {x,t} \right) \in \partial \Omega \times [0,\infty ).}
\end{array}} \right.
Multiplying the equation in (4.1) by w′k,−Pwk and −Pw′k, we get
\frac{1}{2}\frac{d}{{dt}}\left( {\left\| {w_k^\prime} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{w_k}} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime} \right\|_2^2} \right) + \left( {1 - \eta } \right)\left\| {{P^{\frac{1}{2}}}w_k^\prime} \right\|_2^2 \le {C_\eta }\left\| {{G_k}} \right\|_2^2
and
\begin{array}{*{20}{l}}
{\,\,\frac{d}{{dt}}\left( {\frac{1}{2}\left\| {P{w_k}} \right\|_2^2 + \int_\Omega {\left( {Pw_k^\primeP{w_k} + {P^{\frac{1}{2}}}{w_k}{P^{\frac{1}{2}}}w_k^\prime} \right)dx} } \right) + \left( {1 - \eta } \right)\left\| {P{w_k}} \right\|_2^2}\\
{ \le \left\| {Pw_k^\prime} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime} \right\|_2^2 + {C_n}\left\| {{G_k}} \right\|_2^2 \le {c_0}\left\| {Pw_k^\prime} \right\|_2^2 + {C_\eta }\left\| {{G_k}} \right\|_2^2}
\end{array}
and
\frac{1}{2}\frac{d}{{dt}}\left( {\left\| {{P^{\frac{1}{2}}}w_k^\prime} \right\|_2^2 + \left\| {P{w_k}} \right\|_2^2 + \left\| {Pw_k^\prime} \right\|_2^2} \right) + \left( {1 - \eta } \right)\left\| {Pw_k^\prime} \right\|_2^2 \le {C_\eta }\left\| {{G_k}} \right\|_2^2
with small η > 0. Then, by (4.2) and (4.4) we obtain
y_k^\prime\left( t \right) + \left( {k\left( {1 - \eta } \right) - {c_0}} \right)\left\| {Pw_k^\prime} \right\|_2^2 + \left( {1 - \eta } \right)\left\| {{P^{\frac{1}{2}}}w_k^\prime} \right\|_2^2 + \left( {1 - \eta } \right)\left\| {P{w_k}\left( t \right)} \right\|_2^2 \le k{C_\eta }\left\| {{G_k}} \right\|_2^2
where
\begin{array}{*{20}{l}}
{{y_k}(t)}&{ = \frac{{{k_1} + 1}}{2}\left( {\left\| {P{w_k}(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime(t)} \right\|_2^2} \right)}\\
{}&{\,\,\,\,\, + \frac{1}{2}\left( {\left\| {{P^{\frac{1}{2}}}{w_k}(t)} \right\|_2^2 + \left\| {w_k^\prime(t)} \right\|_2^2} \right) + \frac{{{k_1}}}{2}\left\| {{P^{\frac{1}{2}}}w_k^\prime(t)} \right\|_2^2}\\
{}&{\,\,\,\,\, + \int_\Omega {\left( {Pw_k^\prime(t)P{w_k}(t) + {P^{\frac{1}{2}}}{w_k}(t){P^{\frac{1}{2}}}w_k^\prime(t)} \right)dx} }\\
{}&{\, \le \frac{{{k_1} + 2}}{2}\left( {\left\| {P{w_k}(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime(t)} \right\|_2^2} \right)}\\
{}&{\,\,\,\,\, + \frac{{{k_1} + 1}}{2}\left\| {Pw_k^\prime(t)} \right\|_2^2 + \left\| {w_k^\prime(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime(t)} \right\|_2^2}\\
{}&{ \le {C_0}\left( {\left\| {Pw_k^\prime(t)} \right\|_2^2 + \left\| {P{w_k}(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime(t)} \right\|_2^2} \right)}
\end{array}
By taking k1≥ 3
\begin{array}{*{20}{l}}
{{y_k}(t)}&{\, = \frac{{{k_1} + 1}}{2}\left( {\left\| {P{w_k}(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime(t)} \right\|_2^2} \right)}\\
{}&{ + \frac{1}{2}\left( {\left\| {{P^{\frac{1}{2}}}{w_k}(t)} \right\|_2^2 + \left\| {w_k^\prime(t)} \right\|_2^2} \right) + \frac{{{k_1}}}{2}\left\| {{P^{\frac{1}{2}}}w_k^\prime(t)} \right\|_2^2}\\
{}&{ - \frac{1}{2}\left( {\left\| {Pw_k^\prime} \right\|_2^2 + \left\| {Pw_k^\prime} \right\|_2^2} \right) - \frac{1}{2}\left( {\left\| {{P^{\frac{1}{2}}}w_k^\prime} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime} \right\|_2^2} \right)}\\
{}&{ \ge \left\| {Pw(t)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime(t)} \right\|_2^2 + \left\| {Pw_k^\prime(t)} \right\|_2^2,\,\,t \ge 0.}
\end{array}
Otherwise, we obtain from assumption (A2),
\begin{array}{*{20}{l}}
{\left\| {{G_k}} \right\|_2^2}&{ = \int_\Omega {{{\left| {g\left( {x,{u_k}} \right) - g\left( {x,u} \right)} \right|}^2}dx} = \int_\Omega {g_u^2w_k^2dx} }\\
{}&{ \le {c_0}\int_\Omega {\left( {{{\left| {{u_k}} \right|}^{2\left( {\alpha - 1} \right)}} + {{\left| u \right|}^{2\left( {\alpha - 1} \right)}} + h_2^2} \right)w_k^2dx.} }
\end{array}
The application of Sobolev imbedding theorem and the estimate (2.7) gives
\int_\Omega {{{\left| {{u_k}} \right|}^{2\left( {\alpha - 1} \right)}}} w_k^2dx \le \left\| {{w_k}} \right\|_{2{\mu _2}}^2\left\| {{u_k}} \right\|_{2\left( {\alpha - 1} \right){\mu _3}}^{2\left( {\alpha - 1} \right)} \le {C_3}\left\| {{w_k}} \right\|_{2{\mu _2}}^2 \le {C_3}\left\| {{w_k}} \right\|_2^2
with μ2 = n/(n − 4)+ and μ3 = μ2/(μ2 − 1). Similarly,
\int_\Omega {{{\left| {{u_k}} \right|}^{2\left( {\alpha - 1} \right)}}} w_k^2dx \le \left\| {{w_k}} \right\|_{2{\mu _2}}^2\left\| u \right\|_{2\left( {\alpha - 1} \right){\mu _3}}^{2\left( {\alpha - 1} \right)} \le {C_3}\left\| {{w_k}} \right\|_{2{\mu _2}}^2 \le {C_3}\left\| {P{w_k}} \right\|_2^2
and
\int_\Omega {h_2^2} w_k^2dx \le \left\| {{w_k}} \right\|_{2{\mu _2}}^2\left\| {{h_2}} \right\|_{N/2}^2 \le {C_3}\left\| {{w_k}} \right\|_{2{\mu _2}}^2 \le {C_3}\left\| {P{w_k}} \right\|_2^2.
Then, we get from (4.5) to (4.11) that λ 4 > 0, such that
y_k^\prime\left( t \right) + {\lambda _4}{y_k}\left( t \right) \le {C_3}\left\| {{G_k}} \right\|_2^2 \le {C_3}\left\| {{w_k}} \right\|_{2{\mu _2}}^2 \le {C_3}\left\| {P{w_k}} \right\|_{2{\mu _2}}^2 \le {C_3}{y_k}\left( t \right)
where C3 is as in (2.6), independent of k. The differential inequality (4.12) means
{y_k}\left( t \right) \le {y_k}\left( 0 \right){e^{\left( {{C_3} - {\lambda _4}} \right)t}},t \ge 0.
Then, from (4.6) and (4.7), we obtain
{y_k}\left( 0 \right) \le {C_0}\left( {\left\| {{P^{\frac{1}{2}}}\left( {{u_{1k}} - {u_1}} \right)} \right\|_2^2 + \left\| {P\left( {{u_{0k}} - {u_0}} \right)} \right\|_2^2 + \left\| {P\left( {{u_{1k}} - {u_1}} \right)} \right\|_2^2} \right) \to 0\,{\rm{as}}\,k \to \infty
and
\left\| {P{w_k}\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_k^\prime} \right\|_2^2 + \left\| {Pw_k^\prime} \right\|_2^2 \le {y_k}\left( t \right) \le {y_k}\left( 0 \right){e^{\left( {{C_3} - {\lambda _4}} \right)t}} \to 0\,\,{\rm{as}}\,k \to \infty .
This indicates that S (t) : X → X is continuous. Now we show that {S (t)}t≥0 is asymptotically compact in X from the method in [9].
Assume {(u0k,u1k)} is a bounded sequence and {uk (t)} be the corresponding solutions of the problem (1.1) in C [0,∞);
H^{2m} \left( \Omega \right) \cap \left. {H_0^m \left( \Omega \right)} \right)
. We suppose tk→ ∞ as k → ∞. For any T > 0, assume tn,tk> T. Then, the application of (4.12) to wkn (t) = un (t + tn − T ) − uk (t + tn − T ), we get
{Y_{kn}}\left( t \right) \le {Y_{kn}}\left( 0 \right){e^{ - {\lambda _4}t}} + {C_3}\int_0^t {{e^{ - {\lambda _4}\left( {t - s} \right)}}} \left\| {{w_{kn}}\left( s \right)} \right\|_{2{\mu _2}}^2ds,t \ge 0
with
{Y_{kn}}\left( t \right) = \left\| {P{w_{kn}}\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}w_{kn}^\prime\left( t \right)} \right\|_2^2 + \left\| {Pw_{kn}^\prime\left( t \right)} \right\|_2^2.
Especially, we take t = T and obtain
\begin{array}{l}
\left\| {P\left( {{u_{ & n}}\left( {{t_n}} \right)} \right) - {u_k}\left( {{t_k}} \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}\left( {\user1{u}_n^\prime\left( {{t_n}} \right) - u_k^\prime\left( {{t_k}} \right)} \right)} \right\|_2^2 + \left\| {P\left( {u_n^\prime\left( {{t_n}} \right) - u_k^\prime\left( {{t_k}} \right)} \right)} \right\|_2^2\\
\le {Y_{kn}}\left( 0 \right){e^{ - {\lambda _4}T}} + {C_3}\mathop {\sup }\limits_{0 \le s \le T} \left\| {{u_k}\left( {{t_k} - T + s} \right) - {u_n}\left( {{t_n} - T + s} \right)} \right\|_{2{\mu _2}}^2.
\end{array}
Since the embedding
\left( {{H^{2m}}\left( \Omega \right) \cap H_0^m\left( \Omega \right)} \right) \hookrightarrow {L^{2{\mu _2}}}\left( \Omega \right)\
is compact, we can remove a subsequence {ukk1(tkk1 − T + s)} which converges in L2μ2 (Ω). Therefore, for any ɛ > 0, firstly we fix T > 0, such that
{Y_{kn}}\left( 0 \right){e^{ - {\lambda _4}T}} < \frac{\varepsilon }{2}.
Supposing n0> 0 and k1, j > n0, we get
{C_3}\mathop {\sup }\limits_{0 \le s \le T} \left\| {{u_{{k_{{k_1}}}}}\left( {{t_{{k_{{k_1}}}}} - T + s} \right) - {u_{{k_j}}}\left( {{t_{{k_j}}} - T + s} \right)} \right\|_{2{\mu _2}}^2 < \frac{\varepsilon }{2}.
Then, it follows by (4.18) to (4.20) that {ukk1(tkk1)} is a Cauchy sequence in X and we finalize that {S (t)}t≥0 is asyptotically compact on X and now Theorem 4 is established.
Proof of Theorem 5
From Lemma 2, it is sufficient to indicate that there exists a continuous operator semigroup {S (t)} on X such that S (t)(u0,u1) = (u(t),ut (t)) for each t ≥ 0. By the estimates (2.7), we accomplish that
{\beta _0} = \left\{ {\left( {u,v} \right) \in X\left\| {{P^{\frac{1}{2}}}v} \right\|_2^2 + \left\| {Pu} \right\|_2^2 + \left\| {Pv} \right\|_2^2 \le {C_4}} \right\}
is an absorbing set of {S (t)}t≥0 and for any (u0,u1) ∈ X,
\user1{dis}{\user1{t}_X}\left( {S\left( t \right)\left( {{u_0},{u_{ & 1}}} \right),{\beta _0}} \right) \le {C_5}{e^{ - {\lambda _3}t}},t \ge 0
where the constants C4, C5 are in (2.7). By Theorem 2, S (t) : X → X is continuous and asymptotically compact on X. From a general theory (see [1, 11]), we conclude that S (t) admits a global attractor A on X defined by
A = \omega \left( {{\beta _0}} \right) = \mathop \cap \limits_{\tau \ge 0} {\left[ {\mathop { \cup S\left( t \right)}\limits_{t \ge \tau } {\beta _0}} \right]_X}
where [D]X is the closure of the set D in X. Then we prove the Theorem 5.
Decay property of solution for (1)
In this section, we search the decay property of solution to (1.1) with f ≣ 0. Firstly, we present a well-known Lemma that will be needed.
Assume E : [0,∞) → [0,∞) is a non-increasing function and suppose that there are constants q ≥ 0 and γ > 0 such that\int_S^\infty {{E^{q + 1}}} \left( t \right)dt \le {\gamma ^{ - 1}}E{\left( 0 \right)^q}E\left( s \right),\forall S \ge 0.Then, we getE\left( t \right) \le E\left( 0 \right){\left( {\frac{{1 + q}}{{1 + q\gamma t}}} \right)^{1/q}}\forall t \ge 0\,ifq > 0andE\left( t \right) \le E\left( 0 \right){e^{1 - \gamma t}}\forall t \ge 0ifq = 0.
Proof of Theorem 7
Suppose u(t) is a weak solution in Theorem 3 with f = 0. Show
E\left( t \right) = \frac{1}{2}\left( {\left\| {u\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2^2} \right) + \int_\Omega {G\left( {u\left( t \right)} \right)dx,t \ge 0.}
Then, we obtain by (1.1) that
E\prime\left( t \right) + {\left\| {{P^{\frac{1}{2}}}{u_t}} \right\|^2} = 0,\,\,\,\forall \ge 0.
This indicates that E (t) is non-increasing in [0,∞).
Multiplying the equation in (1.1) by Eq (t)u(t), q > 0, we obtain
\int_S^T {{E^q}} \int_\Omega {u\left( {{u_{tt}} - Pu - P{u_{tt}} + g\left( u \right)} \right)dxdt} = 0,\,\,\forall T > S \ge 0.
We note that
\begin{array}{l}
\left. {\int_S^T {{E^q}\left( t \right)\left( {u,{u_{tt}}} \right)} dt = {E^q}\left( t \right)\left( {u,{u_t}} \right)} \right|_S^T - \int_S^T {\left( {qE{{\left( t \right)}^{q - 1}}E\prime\left( t \right)\left( {u,{u_t}} \right) + {E^q}\left( t \right)\left\| {{u_t}\left( t \right)} \right\|_2^2} \right)} \\
- \int_S^T {{E^q}\left( t \right)\left( {u,Pu} \right)dt} = \int_S^T {{E^q}\left( t \right)\left\| {{P^{\frac{1}{2}}}u} \right\|_2^2} \\
- \int_S^T {{E^q}} \left( t \right)\left( {u,P{u_t}} \right)dt = \int_S^T {{E^q}\left( t \right)\left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}{u_t}} \right)dt}
\end{array}
and
\begin{array}{*{20}{l}}
{ - \int_S^T {{E^q}\left( t \right)\left( {u,P{u_{tt}}} \right)dt} }&{ = - \int_S^T {\left( {qE{{\left( t \right)}^{q - 1}}E\prime\left( t \right)\left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}{u_t}} \right) + {E^q}\left( t \right)\left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2^2} \right)dt} }\\
{}&{\,\,\,\,\,\left. { + {E^q}\left( t \right)\left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}{u_t}} \right)} \right|_S^T.}
\end{array}
Then, we get by (5.6) that
\begin{array}{*{20}{l}}
{2\int_S^T {{E^{q + 1}}\left( t \right)dt = } }&{ - {E^q}\left( t \right)\left[ {\left. {\left( {u,{u_t}} \right) + \left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}{u_t}} \right)} \right|_S^T} \right]}\\
{}&{ + q\int_S^T {E{{\left( t \right)}^{q - 1}}E\prime\left( t \right)\left[ {\left( {u,{u_t}} \right) + \left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}ut} \right)} \right]} dt}\\
{}&{ + 2\int_S^T {{E^q}\left( t \right)\left( {\left\| {{u_t}\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2^2} \right)dt} }\\
{}&{ + \int_S^T {{E^q}\left( t \right)\left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}{u_t}} \right)dt} + \int_S^T {{E^q}} \left( t \right)\left( {2G\left( u \right) - ug\left( u \right)} \right)dt.}
\end{array}
Since G(u) ≥ 0, E (t) ≥ 0. Moreover, we get the following estimates from (5.5):
{\left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2} \le {\left( { - E\prime\left( t \right)} \right)^{1/2}},\,\,\,\,\,\left\| {{P^{\frac{1}{2}}}u\left( t \right)} \right\|_2^2 \le 2{\left( {E\left( t \right)} \right)^{1/2}},{\left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2} \le 2{\left( {E\left( t \right)} \right)^{1/2}},\forall \ge 0,\left| {{E^q}\left( t \right)\left( {\left( {u,{u_t}} \right) + \left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}{u_t}} \right)} \right)} \right| \le {C_0}{E^q}\left( t \right){\left\| {{P^{\frac{1}{2}}}u} \right\|_2}{\left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2} \le {C_0}{E^{q + 1}}\left( t \right),\begin{array}{l}
\int_S^T {\left| {E{{\left( t \right)}^{q - 1}}E\prime\left( t \right)} \right|\left[ {\left( {u,{u_t}} \right) + \left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}{u_t}} \right)} \right]} dt\\
\le {C_0}\int_S^T {E{{\left( t \right)}^{q - 1}}} \left| {E\prime\left( t \right)} \right|{\left\| {{P^{\frac{1}{2}}}u} \right\|_2}{\left\| {{P^{\frac{1}{2}}}{u_t}} \right\|_2}dt \le {C_0}{E^{q + 1}}\left( S \right),
\end{array}2\int_S^T {{E^q}} \left( t \right)\left( {\left\| {{u_t}\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2^2} \right)dt \le {C_0}\int_S^T {{E^q}} \left( t \right){\left( { - E\prime\left( t \right)} \right)^{1/2}} \le {C_0}{E^{q + 1}}\left( s \right),\begin{array}{*{20}{l}}
{\int_S^T {{E^q}\left( t \right)\left( {{P^{\frac{1}{2}}}u,{P^{\frac{1}{2}}}{u_t}} \right)dt} }&{ \le \int_S^T {{E^q}\left( t \right){{\left\| {{P^{\frac{1}{2}}}u} \right\|}_2}{{\left\| {{P^{\frac{1}{2}}}{u_t}} \right\|}_2}} }\\
{}&{ \le \int_S^T {{E^{q + 1}}\left( t \right)dt} + {C_1}{E^{q + 1}}\left( S \right).}
\end{array}
Then we obtain from (5.8) to (5.12) that
\int_S^T {{E^{q + 1}}\left( t \right)dt} \le {C_0}{E^{q + 1}}\left( S \right) \le {C_0}{E^q}\left( 0 \right)E\left( S \right) \equiv {\gamma ^{ - 1}}{E^q}\left( 0 \right)E\left( S \right).
From Lemma 10, we get
\begin{array}{*{20}{l}}
{E\left( t \right)}&{ = \frac{1}{2}\left( {\left\| {u\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}u\left( t \right)} \right\|_2^2 + \left\| {{P^{\frac{1}{2}}}{u_t}\left( t \right)} \right\|_2^2} \right) + \int_\Omega {G\left( {u\left( t \right)} \right)dx} }\\
{}&{ \le E\left( 0 \right){{\left( {\frac{{1 + q}}{{1 + q\gamma t}}} \right)}^{1/q}} \le {C_1}{{\left( {1 + t} \right)}^{ - 1/q}}.}
\end{array}
This is the estimates (2.8) and the proof of Theorem 7 is completed.
Conclusion 8.In this paper, we obtained the global attractor and the asymptotic behavior of global solution for the higher-order evolution equation with damping term. This improves and extends many results in the literature such as (Xie and Zhong (2007);
Chen et al. (2011)).