
We present a comprehensive investigation of the long-term dynamics generated by a semilinear wave equation with time-dependent coefficients and quintic nonlinearity on a bounded domain subject to Dirichlet boundary conditions. By employing rescaling techniques for time and utilizing the Strichartz estimates applicable to bounded domains, we initially study the global well-posedness of the Shatah–Struwe (S–S) solutions. Subsequently, we establish the existence of a uniform weak global attractor consisting of points on complete bounded trajectories through an approach based on evolutionary systems. Finally, we prove that this uniformly weak attractor is indeed strong by means of a backward asymptotic a priori estimate and the so-called energy method. Moreover, the smoothness of the obtained attractor is also shown with the help of a decomposition technique.
Citation: Feng Zhou, Hongfang Li, Kaixuan Zhu, Xin Li. Dynamics of a damped quintic wave equation with time-dependent coefficients[J]. AIMS Mathematics, 2024, 9(9): 24677-24698. doi: 10.3934/math.20241202
[1] | Tingting Liu, Tasneem Mustafa Hussain Sharfi, Qiaozhen Ma . Time-dependent asymptotic behavior of the solution for evolution equation with linear memory. AIMS Mathematics, 2023, 8(7): 16208-16227. doi: 10.3934/math.2023829 |
[2] | Sen Ming, Xiaodong Wang, Xiongmei Fan, Xiao Wu . Blow-up of solutions for coupled wave equations with damping terms and derivative nonlinearities. AIMS Mathematics, 2024, 9(10): 26854-26876. doi: 10.3934/math.20241307 |
[3] | Fatima Siham Djeradi, Fares Yazid, Svetlin G. Georgiev, Zayd Hajjej, Khaled Zennir . On the time decay for a thermoelastic laminated beam with microtemperature effects, nonlinear weight, and nonlinear time-varying delay. AIMS Mathematics, 2023, 8(11): 26096-26114. doi: 10.3934/math.20231330 |
[4] | Mustafa Inc, Hadi Rezazadeh, Javad Vahidi, Mostafa Eslami, Mehmet Ali Akinlar, Muhammad Nasir Ali, Yu-Ming Chu . New solitary wave solutions for the conformable Klein-Gordon equation with quantic nonlinearity. AIMS Mathematics, 2020, 5(6): 6972-6984. doi: 10.3934/math.2020447 |
[5] | Adel M. Al-Mahdi . Long-time behavior for a nonlinear Timoshenko system: Thermal damping versus weak damping of variable-exponents type. AIMS Mathematics, 2023, 8(12): 29577-29603. doi: 10.3934/math.20231515 |
[6] | Sobajima Motohiro, Wakasugi Yuta . Remarks on an elliptic problem arising in weighted energy estimates for wave equations with space-dependent damping term in an exterior domain. AIMS Mathematics, 2017, 2(1): 1-15. doi: 10.3934/Math.2017.1.1 |
[7] | Jing Wang, Qiaozhen Ma, Wenxue Zhou . Attractor of the nonclassical diffusion equation with memory on time- dependent space. AIMS Mathematics, 2023, 8(6): 14820-14841. doi: 10.3934/math.2023757 |
[8] | Jian-Gen Liu, Jian Zhang . A new approximate method to the time fractional damped Burger equation. AIMS Mathematics, 2023, 8(6): 13317-13324. doi: 10.3934/math.2023674 |
[9] | Noorah Mshary . Exploration of nonlinear traveling wave phenomena in quintic conformable Benney-Lin equation within a liquid film. AIMS Mathematics, 2024, 9(5): 11051-11075. doi: 10.3934/math.2024542 |
[10] | Abdulghani R. Alharbi . Traveling-wave and numerical solutions to nonlinear evolution equations via modern computational techniques. AIMS Mathematics, 2024, 9(1): 1323-1345. doi: 10.3934/math.2024065 |
We present a comprehensive investigation of the long-term dynamics generated by a semilinear wave equation with time-dependent coefficients and quintic nonlinearity on a bounded domain subject to Dirichlet boundary conditions. By employing rescaling techniques for time and utilizing the Strichartz estimates applicable to bounded domains, we initially study the global well-posedness of the Shatah–Struwe (S–S) solutions. Subsequently, we establish the existence of a uniform weak global attractor consisting of points on complete bounded trajectories through an approach based on evolutionary systems. Finally, we prove that this uniformly weak attractor is indeed strong by means of a backward asymptotic a priori estimate and the so-called energy method. Moreover, the smoothness of the obtained attractor is also shown with the help of a decomposition technique.
In this paper, we are concerned with the following semilinear damped wave model
{∂t(α(t)∂tu)+β(t)Au+γ(t)∂tu+β(t)g(u)=β(t)f(x),u|∂Ω=0,u(x,τ)=u0τ,∂tu(x,τ)=u1τ. | (1.1) |
Here, Ω⊂R3 is a bounded domain with a smooth boundary, A=−Δ, g(u) is a given source term, and the coefficients α(t), β(t), and γ(t) all depend on time. From now on, we assume the external force f∈L2(Ω).
The semilinear wave equations with time-dependent coefficients have been investigated quite extensively by several authors in recent years, with particular regard to its long-term behavior. For example, wave equations with time-dependent speed of propagation were investigated by Conti et al. [12,13], and very recently this model was generalized to the hyperbolic equations with time-dependent memory kernel in [15]; Uesaka et al. [19] made some significant progress in the oscillation property of semilinear wave equations with time-dependent coefficients, and subsequently the dynamics of this equation were studied by Aragão et al. in [1], including the continuity of pullback attractors.
It is worth observing that the nonlinear term g satisfies g(u)∼u|u|q−1 with 1≤q≤3 in the aforementioned papers. Thus, it seemed natural to extend these results to the sup-cubic case. The case of sup-cubic growth rate is a bit more complicated since the uniqueness of energy-weak solutions is unknown as q>3, see, e.g., [3,20]. In order to overcome these difficulties, the authors in [4] studied the semilinear oscillon equation with the growth index 3<q<5 by using parabolic approximations governed by the fractional powers of the wave operator. Another effective way to deal with sup-cubic nonlinearity is using S–S solutions, which have more delicate space-time integrability, such as u∈L4loc(R,L12(Ω)). Very recently, based on the recent extension of Strichartz estimates for the bounded domains, the Eq (1.1) in the case α=β=1 with a sign changing damping and sub-quintic nonlinearity (0<q<5) was discussed in detail in [5].
In this paper, motivated by the studies in [1,5,14], we consider the problem of the existence of a uniform global attractor for Eq (1.1) with quintic nonlinearity in the natural energy phase space H10(Ω)×L2(Ω). The difficulties with this problem mainly stem from the following aspects:
● How do you generalize the usual Strichartz estimates to wave equations with time-dependent coefficients? In the case of the whole space Ω=R3, the Strichartz type estimates for variable coefficient wave equations have been studied by many authors, one can refer to [16] and the references cited therein. However, when Ω⊂R3 is a general bounded domain, as far as we known, the corresponding results are still lacking.
● How to establish the asymptotic compactness of the system generated by S–S solutions of Eq (1.1) with quintic nonlinearity? In the sub-quintic case, one can establish the so-called energy-to-Strichartz (ETS) estimate (2.2), and based on the ETS estimate, one can obtain the well-posedness, dissipativity, asymptotic compactness, and existence of attractors in way that is similar to the classical cubic case. In contrast to this, in the quintic case, the ETS estimate is only proved in the case when Ω=R3 or Ω=T3 with periodic boundary conditions. Since the ETS estimate for a general domain is still an open problem, it is impossible to deduce the asymptotic compactness by giving any control of the Strichartz norm in terms of the initial data, and the control of this norm may be a priori lost when passing to the limit t→∞, and the attractor may contain solutions that are less regular than the S–S ones, for which we may not have the energy equality.
In this paper, in order to circumvent the difficulties mentioned above, we present a new scheme to study the dynamics of the wave equations (1.1), and summarize the main method in Figure 1 for clarity.
(1) By rescaling techniques for time, we can reduce the Eq (1.1) to an equation of simple form (2.3), and then the well-posedness as well as the energy dissipativity of S–S solutions can be proved by the usual method.
(2) We apply a newly developed framework named evolutionary systems (see [7]) for studying the asymptotic dynamics of S–S solutions, and obtain the existence and structure of the uniform weak global attractor Aw. Since the evolutionary systems E (3.3) generated by S–S solutions may not be closed with respect to weak topology on the phase space, we follow an interesting technique initiated by Cheskidov and Lu in [10], which is based on taking a closure of the evolution systems ˉE (3.4). In what follows, our main task is to show that E((−∞,∞))=ˉE((−∞,∞)) via a newly developed approach presented in [20].
(3) Using the energy method developed in [3] and remembering the backward regularity of complete trajectories contained in E((−∞,∞)), we can prove that the uniform weak global attractor Aw constructed in Step 2 is in fact a strongly compact strong global attractor As. Moreover, we obtain a bounded uniform attractor AΣs⊂E1 by a decomposition technique.
We impose the following standing assumptions on the nonlinear damping and coefficients:
Assumption 1.1. (G) g∈C2(R) with g(0)=0 and satisfies
|g″(s)|≤Cg(1+|s|q−2),g′(s)≥−κ1+δ|s|q−1, | (1.2) |
g(s)s−4G(s)≥−κ2,G(s)≥κ3|s|q+1−κ4,∀s∈R. | (1.3) |
Here 3≤q≤5, G(s)=∫s0g(τ)dτ, κi (i=1,2,3,4), δ, and Cg are given positive constants.
(COEF) α,β,γ∈C2b(R) satisfying
α0≤α(t)≤α1,β0≤β(t)≤β1,γ0≤γ(t)≤γ1, | (1.4) |
|α′(t)|≤α2,|β′(t)|≤β2,γ′(t)≤γ2,∀t∈R, | (1.5) |
where αi, βi and γi (i=0,1,2) are all positive constants, and 2γ0β0>α2β1+α1β2.
The outline of our paper is given below. In Section 2, the property of the S–S solutions of Eq (1.1) is discussed in Theorem 2.2. In Section 3, the existence and structure of the uniform weak global attractor are studied in Theorem 3.7, and the backward asymptotic regularity of complete trajectories contained in ˉE((−∞,∞)) are proved in Theorem 3.11. Finally, the existence and regularity of the uniformly strong global attractor is established in Theorems 4.3 and 4.4.
Let ‖⋅‖ and ⟨⋅,⋅⟩ be the usual norm and inner product in L2(Ω). For convenience, we denote Hs=D(As2), Es=Hs+1×Hs, s∈R. Then, H0=L2(Ω), H1=H10(Ω), H2=H2(Ω)∩H10(Ω), and H−1 is the dual space to H10(Ω). In particular, we denote E:=E0=H10(Ω)×L2(Ω).
Definition 2.1. A function u(t) is a weak solution of Eq (1.1) iff ξu:=(u,∂tu)∈L∞(τ,T;E) and Eq (1.1) is satisfied in the sense of distribution, i.e.,
−∫Tτ⟨α(t)∂tu,∂tϕ⟩dt+∫Tτβ(t)⟨∇u⋅∇ϕ,1⟩dt+∫Tτγ(t)⟨∂tu,ϕ⟩dt+∫Tτβ(t)⟨g(u),ϕ⟩dt=∫Tτβ(t)⟨f,ϕ⟩dt |
for any ϕ∈C∞0((τ,T)×Ω). A weak solution is a Shatah–Struwe (S–S) solution of Eq (1.1) on the interval [τ,T] iff u∈L4(τ,T;L12(Ω)).
Theorem 2.2. Under Assumption 1.1, then for every ξu(τ)=(u0τ,u1τ)∈E, the Eq (1.1) admits a unique global S–S solution u(t) with the estimate
‖ξu(t)‖E≤e−ϖ(t−τ)Q(‖ξu(τ)‖E)+Q(‖f‖2),∀t≥τ, | (2.1) |
where the positive constant ϖ and the monotone increasing function Q are independent of u, t and τ. In the sub-quintic case, we have, in addition, the estimate
‖ξu(t)‖E+‖u‖L4(t,t+1;L12(Ω))≤e−ϖ(t−τ)Q(‖ξu(τ)‖E)+Q(‖f‖2),∀t≥τ. | (2.2) |
Moreover, if ξu(τ)∈E1, then Eq. (1.1) admits a unique global strong solution and estimate (2.1) also holds.
Proof. (Sketch) Using the change of variable s=ϕ(t)=∫t0√β(ω)α(ω)dω and chain rule (see [1, Section 3] for more details), we can rewrite the Eq (1.1) as follows:
{∂2sv−Δv+η(s)∂sv+g(v)=f(x),v|∂Ω=0,v(x,μ)=u0τ,∂tv(x,μ)=√α(τ)β(τ)u1τ, | (2.3) |
where u(x,t)=u(x,ϕ−1(s))=v(x,s), η(s)=(√αβ)s+γ√αβ and μ=ϕ(τ). The local well-posedness of Eq (2.3) can be verified by using the Galerkin method and Strichartz estimate, and the global existence and regularity of the S–S solution can be proved by Morawetz–Pohozhaev identity and a prior estimate; see [17] for further details. Analyzing the term η(s) and recalling the assumption (COEF), we have
η0≤η(s)=(√αβ)s+γ√αβ=αtβ+αβt+2βγ2β√αβ≤η1,∀s≥μ. | (2.4) |
Here, we can choose η0=2β0γ0−α2β1−α1β22β1√α1β1 and η1=2β1γ1+α2β1+α1β22β0√α0β0. Taking the multiplier ∂sv+εv (ε>0 is small enough) in Eq (2.3), and applying dissipative assumptions (1.3) and (1.4) and Gronwall's inequality, we can obtain the estimate (2.1) for v, e.g., see [5]. In the sub-quintic case, we can also obtain an estimate (2.2) for v by using the standard bootstrapping method in [17]. Finally, in view of
‖ξu(t)‖E≅‖ξv(ϕ(t))‖Eand‖u‖L4(τ,T;L12(Ω))≅‖v‖L4(ϕ(τ),ϕ(T);L12(Ω)), | (2.5) |
then the theorem is completed.
Here we recall some basic ideas and results from the abstract theory of evolutionary systems; see [7,8,9,10] for details. Let (X,ds(⋅,⋅)) be a metric space endowed with a metric ds, which will be referred to as a strong metric. Let dw(⋅,⋅) be another metric on X satisfying the following conditions:
(1) X is dw-compact.
(2) If ds(un,vn)→0 as n→∞ for some un, vn∈X, then dw(un,vn)→0.
Due to the property 2, dw(⋅,⋅) and ds(⋅,⋅) will be referred to as weak metric, and strong metric respectively. Let C([a,b];X∙), where ∙=s or w, be the space of d∙-continuous X-valued functions on [s,t] endowed with the metric
dC([a,b];X∙)(u,v):=supt∈[a,b]d∙(u(t),v(t)). |
Let also C([a,∞);X∙) be the space of d∙-continuous X-valued functions on [a,∞) endowed with the metric
dC([a,∞);X∙)(u,v):=∑K∈N12KdC([a,a+K];X∙)(u,v)1+dC([a,a+K];X∙)(u,v). | (3.1) |
To define an evolutionary system, first let
T:={I:I=[T,∞)⊂R,or I=(−∞,∞)}, |
and for each I∈T, let F(I) denote the set of all X-valued functions on I.
Definition 3.1. A map E that associates with each I∈T a subset E(I)⊂F(I) will be called an evolutionary system if the following conditions are satisfied:
(1) E([0,∞))≠∅.
(2) E(I+s)={u(⋅):u(⋅−s)∈E(I)} for all s∈R.
(3) {u(⋅)∣I2:u(⋅)∈E(I1)}⊂E(I2) for all pairs I1, I2⊂T, such that I2⊂I1.
(4) E((−∞,∞))={u(⋅):u(⋅)∣[T,∞)∈E([T,∞)),∀T∈R}.
We will refer to E(I) as the set of all trajectories on the time interval I. Let P(X) be the set of all subsets of X. For every t≥0, define a map
R(t):P(X)→P(X),R(t)A:={u(t):u(0)∈A,u∈E([0,∞))},A⊂X. |
Definition 3.2. A set Aw⊂X is a dw-global attractor of E if Aw is a minimal set that is
(1) dw-closed;
(2) dw-attracting: for any B⊂X and ϵ>0, there exists t0, such that
R(t)B⊂Bw(Aw,ϵ):={u:infx∈Awdw(u,x)<ϵ},∀t≥t0. |
Definition 3.3. The ω∙-limit set (∙=s,w) of a set A⊂X is
ω∙(A):=⋂T≥0¯⋃t≥TR(t)A∙. |
In order to extend the notion of invariance from a semiflow to an evolutionary system, we will need the following mapping:
˜R(t)A:={u(t):u(0)∈A,u∈E((−∞,∞))},A⊂X,t∈R. |
Definition 3.4. A set A⊂X is positively invariant if
˜R(t)A⊂A,∀t≥0. |
A is invariant if
˜R(t)A=A,∀t≥0. |
A is quasi-invariant if, for every a∈A, there exists a complete trajectory u∈ E((−∞,∞)) with u(0)=a and u(t)∈A for all t∈R.
As shown in [7,10], a semiflow {S(t)} or a family of a processes {Uσ(t,τ)}, σ∈Σ, defines an evolutionary system. In order to investigate the existence and structure of Aw, we use a new method initiated by Cheskidov and Lu in [10] by taking a closure of the evolutionary system E. Let
ˉE([τ,∞)):=¯E([τ,∞))C([τ,∞);Xw),∀τ∈R. |
Obviously, ˉE is also an evolutionary system. We call ˉE the closure of the evolutionary system E, and add the top-script − to the corresponding notations. Below is an important assumption that we will impose on E in some cases.
⋄ A1 E([0,∞)) is pre-compact in C([0,∞);Xw).
Theorem 3.5. [10] Let E be an evolutionary system. Then the weak global attractor Aw exists. Furthermore, assume that E satisfies A1. Let ˉE be the closure of E. Then
(1) Aw=ωw(X)=ˉωw(X)=ˉAw={u0∈X:u0=u(0) for someu∈ˉE((−∞,∞))}.
(2) Aw is the maximal invariant and maximal quasi-invariant set w.r.t. ˉE.
(3) (Weak uniform tracking property) For any ϵ>0, there exists t0, such that for any t∗>t0, every trajectory u∈E([0,∞)) satisfies
dC([t∗,∞);Xw)(u,v)≤ϵ, |
for some complete trajectory v∈ˉE((−∞,∞)).
Let Uσ(t,τ):E→E, t≥τ be the S–S solution operator of Eq (1.1), where σ=(α,β,γ)∈Σ:=[Thσ,h∈R]C2b(R), (Thσ)(⋅):=σ(⋅+h), then ThΣ=Σ. We construct the skew product flow by
S(t)(ξ,σ):=(Uσ(t,0)ξ,Ttσ),t≥0, | (3.2) |
where (ξ,σ)∈E=E×Σ with norm ‖(ξ,σ)‖E=(‖ξ‖2E+‖σ‖2C2b(R))12. Then {S(t)}t≥0 forms a semigroup. Now define an evolutionary systems (ES) on E by
E([0,∞)):={(ξu(⋅),σ(⋅)):(ξu(t),σ(t))=S(t)(ξ,σ),ξu(t)∈X,σ∈Σ,∀t≥0}, | (3.3) |
where X:={ξu∈E:‖ξu‖2E≤2Q(‖f‖2)}. Let
ˉE([0,∞)):=¯E([0,∞))C([0,∞);Xw), | (3.4) |
where Xw=Xw×C2b(R) and the metric on C([0,∞);X∙) defined in the same manner as (3.1).
Lemma 3.6. Suppose σ is translation compact in C2b(R), and let ξun=(un,∂tun) be a sequence of S–S solutions of Eq (1.1) with symbols σn such that (ξun(t),σn(t))∈X for all t≥t0. Then
ξunis bounded inL∞([t0,T];E),∂tξunis bounded inL∞([t0,T];E−1),∀T>t0. | (3.5) |
Moreover, there exists a subsequence nj such that σnj converges in C2b(R) to some σ∈Σ and ξunj converges to some ξu in C([t0,T];Ew), i.e., (ξunj,ϕ)→(ξu,ϕ) uniformly on [t0,T] as nj→∞ for all ϕ∈E.
Proof. Applying Theorem 2.2 and remembering that ξun are the S–S solutions of Eq (2.3), thus we obtain (3.5). Now applying the Alaoglu compactness theorem to extract a subsequence ξunj which w∗-converges to some function ξu∈L∞([t0,T];E), i.e.,
ξunj⇀ξu weakly-∗ in L∞([t0,T];E). | (3.6) |
Recalling the following compact embedding
{(u,∂tu)∈L∞([t0,T];E)}∩{∂2tu∈L∞([t0,T];H−1(Ω))}⊂⊂{(u,∂tu)∈C([t0,T];H1−ι(Ω)×H−ι(Ω))} |
for some 0<ι≤1, we deduce that the weak-∗ convergence (3.6) implies the strong convergence ξunj→ξu in C([t0,T];Ew)}. The proof is completed.
Theorem 3.7. Under Assumption 1.1 assume that σ is translation compact in C2b(R). Then the uniform weak global attractor Aw for ES E defined by (3.3) exists. In addition, E satisfies A1 and Aw:={(ξu0,σ0):(ξu0,σ0)=(ξu(0),σ(0)) for someΞ=(ξu,σ)∈ˉE((−∞,∞))}. Moreover, ∀ϵ>0 there is some t0:=t0(ϵ) such that for any t∗>t0 and every trajectory Ξ∈E([0,+∞)) satisfies dC([0,∞):Xw)(Ξ,Ξ∗)<ϵ for some complete trajectory Ξ∗∈ˉE((−∞,∞)).
Proof. The existence of the attractor Aw can be established by using Theorem 3.5 directly. Let Ξn be a sequence in E([0,∞)). Using Lemma 3.6, we extract a subsequence (still denoting by Ξn) that converges to some Ξ1∈C([0,1];Xw) as n→∞. Passing to a subsequence and still denote ξun once more, we obtain that Ξn→Ξ2∈C([0,2];Xw) as n→∞ for some Ξ2∈C([0,2];Xw) with Ξ1=Ξ2 on [0,1]. Continuing this diagonalization process, we get a subsequence Ξnj converges to Ξ∈C([0,∞);Xw), and A1 is proven. The other statement contained in the above theorem can be proved by applying Theorem 3.5 again.
Theorem 3.8. Under Assumption 1.1 assume that σ is translation compact in C2b(R). Then the complete trajectory Ξ=(ξu,σ)∈ˉE((−∞,∞)) iff there exists a sequence of times tn→−∞ and a sequence of S–S solutions ξun(t) of Eq (1.1):
{∂t(αn(t)∂tun)−βn(t)Δun+γn(t)∂tun+βn(t)g(un)=βn(t)f(x),ξun(tn)=ξ0n∈X,t≥tn, | (3.7) |
such that (ξun,σn)⇀(ξu,σ) in C([−T,∞);Xw) for any T>0, where σn=(αn,βn,γn)∈Σ.
Proof. Let Ξ=(ξu,σ)∈ˉE((−∞,∞)) and denote Ξn=Ξ|[tn,∞)∈ˉE([tn,∞)), where tn→−∞ as n→∞. Obviously Ξn⇀Ξ in C([−T,∞);Xw), ∀T>0. Since Ξn∈ˉE([tn,∞)), then there exists a sequence {Ξ(k)n}∞k=1∈E([tn,∞)) such that Ξ(k)n⇀Ξn in C([tn,∞);Xw) as k→∞. By a standard diagonalization process, we obtain that there exists a sequence Ξ(n)n (denoted by Ξn=(ξun,σn)) such that Ξn⇀Ξ in C([−T,∞);Xw) for any T>0. Recalling the definition of E and Ξ, we know that ξun is the S–S solution of Eq (1.1).
Conversely, let Ξn=(ξun,σn)∈E([tn,∞)) and Ξn⇀Ξ in C([−T,∞);Xw), ∀T>0. So {Ξn|[−T,∞):Ξn∈E([tn,∞)}⊂E([−T,∞)) converges to Ξ|[−T,∞)∈C([−T,∞);Xw). Thus Ξ∈ˉE([−T,∞)) for any T>0. By definition, this implies Ξ∈ˉE((−∞,∞)).
Remark 3.9. Since every S–S solution ξun can be obtained as a limit of Galerkin approximations (see [14,17] for more detail), then for any Ξ=(ξu,σ)∈ˉE((−∞,∞)), we can extract a sequence ξ(k)uk by using a standard diagonalization process again such that ξ(k)uk⇀ξu in C([−T,∞);Xw) for any T>0, and u(k)k=∑kl=1dkl(t)el satisfies
{∂t(αk(t)∂tu(k)k)−βk(t)Δu(k)k+γk(t)∂tu(k)k+βk(t)Pkg(u(k)k)=βk(t)Pkf(x),ξ(k)uk(tk)=Pkξuk(tk),t≥tk, | (3.8) |
where tk→−∞ as k→∞, {ek}∞i=1 be the orthonormal system of eigenvectors of the Laplacian −Δ with Dirichlet boundary conditions and Pk is the projector from L2(Ω) to Ek:=span{e1,e2,⋯,ek}.
Corollary 3.10. Let the assumptions of Theorem 3.8 be satisfied. Then, for any Ξ=(ξu,σ)∈ˉE((−∞,∞)), we have
∫∞−∞‖∂tu(r)‖2dr≤Q(‖f‖2),∂tu∈Cb(R,H−ι)andlimt→±∞‖∂tu(t)‖H−ι=0 | (3.9) |
for any 0<ι≤1, where Q(⋅) is a monotone increasing function.
Proof. Let Ξ=(ξu,σ)∈ˉE((−∞,∞)), taking the multiplier ∂tv in (2.3) and combining Theorem 2.2 and Remark 3.9 to deduce that ∫∞−∞‖∂tu(r)‖2dr≤Q(‖f‖2). In order to prove convergence in (3.9), we note that ˉE((−∞,∞)) is bounded in (Cb(R;E)∩C1b(R;E−1))×C2b(R) and σ is translation compact in C2b(R), then the convergence is a standard corollary of dissipative integral in (3.9) and the compact embedding [Cb(R;E)∩C1b(R;E−1)]⊂⊂Cloc(R;H1−ι×H−ι) for every 0<ι≤1.
The following Theorem 3.11 discusses the backward smoothing property of the complete trajectory included in ˉE((−∞,∞)), and the proof is similar to the ones given in [20, Theorem 2.1], and for this reason we give a sketch of the main steps of the proof for the reader's convenience.
Theorem 3.11. Under Assumption 1.1 and assume that σ is translation compact in C2b(R), then for every complete trajectory Ξ=(ξu,σ)∈ˉE((−∞,∞)), there exists a time T=T(u,σ) such that ξu∈Cb((−∞,T];E1) and ‖ξu‖Cb((−∞,T];E1)≤Q(‖f‖2,‖σ‖C2b(R)).
Proof. We divide the proof into several steps.
Step 1. Rewrite Eq (2.3) as follows:
∂2sv−Δv+η(s)∂sv+L(−Δ)−1v+g(v)=h(s):=L(−Δ)−1v+f(x). |
From the definition of h and applying Theorem 2.2, we have ‖h(s)‖2≤Q(‖f‖2) and ∫S+1S‖∂sh(s)‖2H2ds=∫S+1S‖∂sv(s)‖2ds≤Q(‖f‖2). Using Corollary 3.10, we infer that
∂th∈Cb(R;H2−ι),lims→−∞‖∂sh(s)‖H2−ι=0,∀0<ι≤1. | (3.10) |
Step 2. Applying Lemma 2.2 in [20], we know that for sufficiently large L (depending on the coefficients in Assumption 1.1), the parabolic equation
∂sw−Δw+g(w)+L(−Δ)−1w=h(s),s∈R | (3.11) |
possesses a unique solution w(s) in the class Cb(R;H2) with the following estimates:
‖w(s)‖H2≤Q(‖f‖2),∂sw∈Cb(R;H2),∂2sw∈L2([S,S+1];H1),∀S∈R, | (3.12) |
and the following convergence
limS→−∞{‖∂sw(S)‖H2+‖∂2sw‖L2([S,S+1];H1)}=0. | (3.13) |
Step 3. For a sufficiently large L, there exists a time S=S(v,L,σ) such that the problem
∂2sz−Δz+η(s)∂sz+L(−Δ)−1z+g(z)=h(s),s≤S | (3.14) |
possesses a unique regular bounded backward solution ξz∈E1, which satisfies
‖∂sz(s)‖H2+‖z(s)‖H2≤Q(‖f‖2,‖σ‖C2b),s≤S and lims→−∞‖∂sz(s)‖L∞(Ω)=0. | (3.15) |
To see this, let z=w+Z, where w satisfying Eq (3.11), then Z satisfies
∂2sZ−ΔZ+η(s)∂sZ+L(−Δ)−1Z+g(w+Z)−g(w)=Fw(s):=−∂2sw−(η(s)−1)∂sw. | (3.16) |
We can apply the implicit function theorem in order to solve Eq (3.16) in the space
ΦS:=Cb((−∞,S],E1)), | (3.17) |
where S is small enough. Applying Step 2, we have
Fw∈L2([s,s+1],H1)∀s∈R and limS→−∞‖Fw‖L2([S,S+1],H1)=0. |
Now, we intend to verify that the variation equation at Z=0
∂2sZ−ΔZ+η(s)∂sZ+L(−Δ)−1Z+g′(w)Z=H(s),s≤S | (3.18) |
is uniquely solvable for every H∈L2loc((−∞,S],H1) such that
‖H‖L2b((−∞,S],H1):=sups∈(−∞,S−1)‖H‖L2((s,s+1],H1)<∞ |
if S is small enough. Firstly, taking the multiplier ∂sZ+εZ in (3.18) yields
ddtEZ+QZ=2⟨H,∂sZ+εZ⟩+⟨g″(w)∂sw,Z2⟩, | (3.19) |
where
EZ=‖∂sZ‖2+‖Z‖2H1+⟨g′(w)Z,Z⟩+2ε⟨∂sZ,Z⟩+L‖Z‖2H−1,QZ=2(η−ε)‖∂sZ‖2+2ε‖Z‖2H1+2εη⟨∂sZ,Z⟩+2εL‖Z‖2H−1+2ε⟨g′(w)Z,Z⟩. |
Choosing L≥4C2κ21, we obtain
−2⟨g′(w)Z,Z⟩≤2κ1‖Z‖2≤2Cκ1‖Z‖H1‖Z‖H−1≤12(‖Z‖2H1+4C2κ21‖Z‖2H−1)≤12(‖Z‖2H1+L‖Z‖2H−1). | (3.20) |
Combining (3.19) and (3.20) and recalling (3.12), there exists a sufficiently small parameter ε>0 such that
C‖ξZ(s)‖2E≤EZ(s)≤C‖ξZ(s)‖2E |
and
ddsEZ(s)+εEZ(s)≤C‖H(s)‖2+⟨g″(w)∂sw,Z2⟩−α4‖Z‖2H1. |
Using (3.12), (3.13), and embedding H2(Ω)⊂L∞(Ω), we have
ddsEZ(s)+εEZ(s)≤C‖H(s)‖2, if s≤S and Sis small enough. |
Applying Gronwall's inequality, we deduce
‖∂sZ(s)‖2+‖Z(s)‖2H1≤C∫s−∞e−ε(s−r)‖H(r)‖2dr,s≤S. | (3.21) |
Thus, the solution to (3.18) is unique. Secondly, taking the multiplier −Δ(∂sZ+εZ) in (3.18), interpreting g′(w)Z as an external force, and using (3.21) yields
‖∂sZ(s)‖2H1+‖Z(s)‖2H2≤C∫s−∞e−ε(s−r)‖H(r)‖2H1dr≤C‖H‖2L2b((−∞,S],H1),s≤S. | (3.22) |
Thus, the Eq (3.18) is uniquely solvable in space (3.17) if S is small enough. Now, applying the implicit function theorem for Eq (3.16), for a sufficiently small S∈R, there exists a unique solution ξZ∈ΦS of problem (3.16) satisfying
‖∂sZ(s)‖2H1+‖Z(s)‖2H2≤Q(‖f‖2,‖σ‖C2b),s≤S and lims→−∞‖∂sZ(s)‖H1=0. | (3.23) |
Combining the estimates of w in Step 2 and (3.23) we have
‖∂sz(s)‖2H1+‖z(s)‖2H2≤Q(‖f‖2,‖σ‖C2b),s≤S(L,v,σ) and lims→−∞‖∂sz(s)‖H1=0. | (3.24) |
Finally, differentiate equation (3.14) and set ∂sz=ζ, we have
∂2sζ−Δζ+η(s)∂sζ+L(−Δ)−1ζ=hz(s):=h′(s)−g′(z)∂sz−η′(s)∂sz. | (3.25) |
Recalling (3.10) and (3.24), we obtain
lims→−∞‖hz(s)‖H1=0. | (3.26) |
Similar to (3.22), we have
‖∂sζ(s)‖2H1+‖ζ(s)‖2H2≤C∫s−∞e−ε(s−r)‖hz(r)‖2dr,s≤S. | (3.27) |
Since H2⊂C(¯Ω), then (3.26) and (3.27) imply (3.15).
Step 4. We need to prove z=v, for s≤S. Applying Remark 3.9, there exists a sequence of Galerkin approximations such that ξ(k)vk⇀ξv in C([−T,∞);Xw) for any T>0, and
∂2sv(k)k−Δv(k)k+ηk(s)∂sv(k)k+L(−Δ)−1v(k)k+Pkg(v(k)k)=hk(s):=L(−Δ)−1v(k)k+Pkf,s≥sk. | (3.28) |
Here sk=∫tk0√β(ω)α(ω)dω, ηk∈C1b(R), so there exits a subsequence (still denote) ηk→η in C1b(R). Now let zk(s)=Pkz(s), s≤S. According to Step 3, the solution ξz(s) is bounded in E1 when s≤S, and consequently
limk→∞‖ξzk−ξz‖Cb((−∞,s],E)=0,limk→∞‖ξzk−ξz‖Cb((−∞,S]×Ω)=0. | (3.29) |
Here we used the fact that H2⊂⊂C(¯Ω) again, and the convergence of the Fourier series is uniform on compact sets. Now denote V(s):=v(s)−z(s) and Vk(s):=v(k)k(s)−zk(s), recalling (3.28) yields the following equation:
∂2sVk−ΔVk+ηk(s)∂sVk+L(−Δ)−1Vk+Pk[g(Vk+zk)−g(zk)]=Ak, | (3.30) |
where Ak(s)=[η(s)−ηk(s)]∂szk(s)+Pk[g(z(s))−g(zk(s))]. Taking the inner product between (3.30) and ∂sVk+εVk, we have
ddsEVk(s)+εEVk(s)=⟨Ak(s),2∂sVk(s)+εVk(s)⟩−QVk(s)+GVk(s), | (3.31) |
where
EVk=‖∂sVk‖2+‖Vk‖2H1+2ε⟨∂sVk,Vk⟩+L‖Vk‖2H−1+2⟨G(Vk+zk)−G(zk)−g(zk)Vk,1⟩,QVk=2εηk(s)⟨∂sVk,Vk⟩+εL‖Vk‖2H−1+ε‖Vk‖2H1−2ε2⟨∂sVk,Vk⟩+(2ηk(s)−3ε)‖∂sVk‖2,GVk=2ε[⟨G(Vk+zk)−G(zk)−g(zk)Vk,1⟩−⟨g(Vk+zk)−g(zk),Vk⟩]++2⟨g(Vk+zk)−g(zk)−g′(zk)Vk,∂tzk⟩. |
Using the fact that
G(v+w)−G(v)−g(v)w−(g(v+w)−g(v))w≤κ12|w|2,|g(v+w)−g(v)−g′(v)w|≤C|w|2(1+|v|q−2+|w|q−2), |
we have
GVk≤κ1ε2‖Vk‖2+C‖∂szk‖L∞(Ω)⟨|Vk|2(1+|zk|q−2+|Vk|q−2),1⟩. |
Combining (3.15), (3.29), and (3.31), we can infer that there exists a time S′≤S such that, for sufficiently large k, we have
ddsEVk(s)+εEVk(s)≤C‖Ak(s)‖2,∀s≤S′. |
Applying Gronwall's inequality, we obtain
EVk(s)≤EVk(sk)e−ε(s−sk)+C∫sske−ε(s−r)‖Ak(r)‖2dr, |
where constants C and ε are independent of k. Indeed, we have
‖∂sVk(s)‖2+‖Vk(s)‖2H1≤C(1+‖ξv(k)k(sk)‖2E+‖ξzk(sk)‖2E)e−ε(s−sk)++C∫sske−ε(s−r)‖Ak(r)‖2dr. | (3.32) |
Now, passing to the limit k→∞ in (3.32), we obtain ‖ξV(s)‖E=0 for all s≤S′. The proof of Theorem 3.11 is now complete.
Corollary 3.12. Let the assumptions of Theorem 3.11 be satisfied. Then the uniform weak global attractor Aw for ES E is in a more regular space: Aw⊂E1×C2b(R).
Proof. Let Ξu=(ξu,σ) be the complete trajectory of Eq (1.1), then Ξv:=(ξv,η) be the corresponding complete trajectory of Eq (2.3). Recall that there exists time S0, such that ξv(s)∈E1 for all s≤S0. Due to Theorem 2.2, there is an extension ˉv for s≥S0 such that ˉv(s)=v(s) for s≤S0 and ˉv(s) is a S–S solution of Eq (2.3) for all s∈R. Indeed, we can conclude that ξˉv(s)∈E1 for all s∈R. We are now ready to prove that ξˉv(s)=ξv(s) for all s∈R. Since Ξv:=(ξv,η)∈ˉE((−∞,∞)), and applying Remark 3.9, we get
∂2sv(k)k−Δv(k)k+ηk(s)∂sv(k)k+Pkg(v(k)k)=Pkf,ξ(k)vk(sk)=Pkξvk(sk), | (3.33) |
where s≥sk and limk→∞sk=−∞. Obviously, ˉvk=Pkˉv satisfying
∂2sˉvk−Δˉvk+η(s)∂sˉvk+Pkg(ˉv)=Pkf,ξˉvk(sk)=Pkξˉv(sk). | (3.34) |
Denote W=v−ˉv, Wk=v(k)k−ˉvk, then combining (3.33) and (3.34) we know that Wk satisfies
∂2sWk−ΔWk+ηk(s)∂sWk+Pk[g(v(k)k)−g(ˉvk)]+L(−Δ)−1Wk=Bk, | (3.35) |
where
Bk=Pk[g(ˉv)−g(ˉvk)]+[η−ηk]∂sˉvk+L(−Δ)−1Wk. |
Taking the multiplier ∂sWk+εWk in (3.35) yields
ddsEWk(s)+εEWk(s)=Λk(s):=2⟨Bk(s),∂sWk(s)+εWk(s)⟩−QWk(s)+GWk(s), | (3.36) |
where
EWk=‖∂sWk‖2+‖Wk‖2H1+2ε⟨∂sWk,Wk⟩+L‖Wk‖2H−1+2⟨G(Wk+ˉvk)−G(ˉvk)−g(ˉvk)Wk,1⟩,QWk=(2εηk(s)−2ε2)⟨∂sWk,Wk⟩+εL‖Wk‖2H−1+ε‖Wk‖2H1+(2ηk(s)−3ε)‖∂sWk‖2,GWk=2ε[⟨G(Wk+ˉvk)−G(ˉvk)−g(ˉvk)Wk,1⟩−⟨g(Wk+ˉvk)−g(ˉvk),Wk⟩]++2⟨g(Wk+ˉvk)−g(ˉvk)−g′(ˉvk)Wk,∂tˉvk⟩]. |
Choosing L large enough and applying [20, Proposition 2.1], we discover that the right-hand side of (3.36) satisfies \Lambda_{k}(s)\leq C\|B_{k}(s)\|^{2} and C is independent of k . Invoking Gronwall's inequality, we have the estimate
\begin{align} \mathcal{E}_{W_{k}}(s)\leq C(\mathcal{E}_{W_{k}}(s_{k}))e^{-\varepsilon(s-s_{k})}+C\int_{s_{k}}^{s}e^{-\varepsilon(s-r)}\|B_{k}(r)\|^{2}dr,\; \forall s\geq S_{0}. \end{align} | (3.37) |
Passing to the limit k\rightarrow \infty and employing the convergence B_{k}\rightarrow L(-\Delta)^{-1}W strongly in \mathcal{C}((-\infty, S_{0}], \mathcal{H}^{0}) , \|B_{k}\|_{\mathcal{C}((-\infty, S_{0}], \mathcal{H}^{0})}\leq C ( C is independent of k ), \mathcal{E}_{W_{k}} is equivalent to \|\xi_{W_{k}}\|_{\mathscr{E}}^{2} and the fact that \bar{v}(s) = v(s) for s\leq S_{0} , we have
\begin{align} \|\xi_{v}(s)-\xi_{\bar{v}}(s)\|_{\mathscr{E}}^{2} \leq CL^{2}\int_{S_{0}}^{s}e^{-\varepsilon(s-r)}\|(-\Delta)^{-1}(v(r)-\bar{v}(r)\|^{2}dr,\; \forall s\geq S_{0}. \end{align} | (3.38) |
Invoking again Gronwall's inequality to relation (3.38) and noting v(S_{0}) = \bar{v}(S_{0}) , we derive that v(s) = \bar{v}(s) for all s\in\mathbb{R} .
Remark 3.13. The proof of Corollary 3.12 indicates that for any \Xi = (\xi_{u}, \sigma)\in\bar{\mathfrak{E}}((-\infty, \infty)) , then \xi_{u} is the S–S solution of Eq (1.1), i.e., \bar{\mathfrak{E}}((-\infty, \infty)) = \mathfrak{E}((-\infty, \infty)) . Moreover, we have \xi_{u}(t)\in\mathscr{E}^{1} \text{ for all }\; t\in\mathbb{R}.
We introduce some definitions; see [2,6,11] for more details.
Definition 4.1. Let S(t) be a semigroup acting on a Banach space \mathcal{Y} . A set \mathscr{A}_{s} \subset\mathcal{Y} is a (strong) global attractor of S(t) if
(1) The set \mathscr{A}_{s} is compact in \mathcal{Y} ;
(2) The set \mathscr{A}_{s} is strictly invariant: S(t) \mathscr{A}_{s} = \mathscr{A}_{s} ;
(3) It is an attracting set for the semigroup S(t) , i.e., for any bounded set B \subset\mathcal{Y} ,
\begin{align*} dist_{\mathcal{Y}}(S(t)B,\mathscr{A}_{s}): = \sup\limits_{x\in B}\inf\limits_{y\in\mathscr{A}_{s}}\| S(t)x-y\|_{\mathcal{Y}}\rightarrow0,\quad\text{as }\;t\rightarrow \infty. \end{align*} |
Definition 4.2. A set A \subset\mathcal{Y} is said to be uniformly (w.r.t. \sigma\in\Sigma ) attracting for the family of processes \left\{U_\sigma(t, \tau)\right\}, \sigma \in \Sigma , if for any fixed \tau \in \mathbb{R} and every bounded set B\subset\mathcal{Y}
\begin{align*} \lim _{t \rightarrow +\infty}\left(\sup\limits_{\sigma \in \Sigma} dist_{\mathcal{Y}}\left(U_\sigma(t,\tau)B,A\right)\right) = 0. \end{align*} |
A closed, uniformly attracting set \mathscr{A}_{s}^{\Sigma} is said to be the uniform (w.r.t. \sigma \in \Sigma ) attractor of the family of processes \left\{U_\sigma(t, \tau)\right\}, \sigma \in \Sigma , if it is contained in any closed uniformly attracting set (minimality property).
The kernel \mathcal{K}_{\sigma} consists of all bounded complete trajectories of the process U_{\sigma}(t, \tau) , i.e.,
\begin{align*} \mathcal{K}_{\sigma} = \left\{u(\cdot) \mid\|u(t)\|_\mathcal{Y} \leqslant C_u, U_{\sigma}(t, \tau) u(\tau) = u(t)\quad \forall t \geq\tau,\; \tau\in\mathbb{R}\right\}, \end{align*} |
and \mathcal{K}_{\sigma}(s) denotes the kernel section at a time moment s\in\mathbb{R} :
\begin{align*} \mathcal{K}_{\sigma}(s) = \{u(s) \mid u(\cdot) \in \mathcal{K}_{\sigma}\}, \quad \mathcal{K}_{\sigma}(s) \subset\mathcal{Y}. \end{align*} |
Theorem 4.3. Under Assumption 1.1 assume that \sigma is translation compact in \mathcal{C}_{b}^{2}(\mathbb{R}) . Then the semigroup \mathbb{S} defined in (3.2) possesses a strong global attractor \mathscr{A}_{s} in \mathscr{E}^{1}\times\mathcal{C}_{b}^{2}(\mathbb{R}) , which coincides with the uniform weak attractor \mathscr{A}_{w} in Theorem 3.7 and satisfies the following properties:
(i) \Pi_1 \mathscr{A}_{s} = \mathscr{A}_{s}^{\Sigma} is the uniform (w.r.t. \sigma \in \Sigma ) attractor of the family of processes \left\{U_\sigma(t, \tau)\right\} , \sigma\in\Sigma and \Pi_{1} is projector from \mathscr{E} \times \Sigma onto \mathscr{E} ;
(ii) The uniform attractor satisfies \mathscr{A}_{s}^{\Sigma} = \bigcup\limits_{\sigma \in \Sigma} \mathcal{K}_\sigma(0) , where \mathcal{K}_\sigma(0) is the section at t = 0 of the kernel \mathcal{K}_\sigma of the process \left\{U_\sigma(t, \tau)\right\} with the symbol \sigma \in \Sigma .
Proof. In order to apply [6, Theorem Ⅳ.5.1] and [18, Theorem 3.4], we have to check that the processes \{U_{\sigma}(t, \tau)\}_{\sigma\in\Sigma, t\geq\tau} corresponding to the S–S solutions of Eq (1.1) be (\mathscr{E}\times\Sigma, \mathscr{E}) -continuous and uniformly asymptotically compact.
Firstly, consider two S–S solutions \xi_{u_{i}} of Eq (1.1) with symbols \sigma_{i} and with initial values \xi_{u_{i\tau}} , then correspondingly, \xi_{v_{i}} are S–S solutions of Eq (2.3) with symbols \eta_{i} , i = 1, 2 . Then \xi_{w} = \xi_{v_{1}}-\xi_{v_{2}} satisfies the equation
\begin{align} \partial_{s}^{2}w-\Delta w+\eta_{1}\partial_{s}v_{1}-\eta_{2}\partial_{s}v_{2}+g(v_{1})-g(v_{2}) = 0. \end{align} | (4.1) |
Taking the scalar product of (4.1) with \partial_{s}w , we obtain
\begin{align} \frac{d}{ds}\|\xi_{w}(s)\|_{\mathscr{E}}^{2}+(\eta_{1}+\eta_{2})\|\partial_{s}w\|^{2} = 2\langle g(v_{2})-g(v_{1}), \partial_{s}w\rangle+(\eta_{1}-\eta_{2})\langle\partial_{s}v_{1}+\partial_{s}v_{2},\partial_{s}w\rangle. \end{align} | (4.2) |
Thanks to Assumption 1.1 (G) and applying the elementary inequality, we have
\begin{align} 2|\langle g(v_{1})-g(v_{2}),\partial_{s}w\rangle|&\leq C_{g}q(s)\|\xi_{w}\|_{\mathscr{E}}^{2}, \end{align} | (4.3) |
\begin{align} |(\eta_{1}(s)-\eta_{2}(s))\langle\partial_{s}v_{1}+\partial_{s}v_{2}, \partial_{s}w\rangle|&\leq\varepsilon\|\partial_{s}w\|^{2} +C_{\varepsilon}|\eta_{1}(s)-\eta_{2}(s)|^{2}. \end{align} | (4.4) |
Where q(s) = (1+\|v_{1}(s)\|_{L^{12}(\Omega)}^{4}+\|v_{2}(s)\|_{L^{12}(\Omega)}^{4}) . Applying (4.3) and (4.4) in (4.2) and employing Gronwall's inequality, we obtain that
\begin{align} \|\xi_{w}(s)\|_{\mathscr{E}}^{2} \leq e^{C_{g}\int_{\mu}^{s}q(r)dr} (\|\xi_{w}(\mu)\|_{\mathscr{E}}^{2}+C_{\varepsilon}(s-\mu)\|\eta_{1}-\eta_{2}\|^{2}_{\mathcal{C}_{b}^{1}(\mathbb{R})}). \end{align} | (4.5) |
Then the (\mathscr{E}\times\Sigma, \mathscr{E}) -continuity follows in a standard way from the energy inequality (4.5).
Secondly, we intend to verify the uniform asymptotic compactness of the processes \{\widetilde{U}_{\eta}(s, \mu)\}_{\eta\in\widetilde{\Sigma}, s\geq\mu} corresponding to the S–S solutions of Eq (2.3), where \widetilde{\Sigma} = [T_{h}(\eta), h\in\mathbb{R}]_{\mathcal{C}_{b}^{1}(\mathbb{R})} . Let \{\eta_{n}\}\subset\widetilde{\Sigma} , \{-\mu_{n}\}\subset(-\infty, 0] , \mu_{n}\rightarrow \infty as n\rightarrow \infty . \xi_{\mu_{n}} belongs to a bounded subset in \mathscr{E} . Since \sigma is translation compact in \mathcal{C}_{b}^{2}(\mathbb{R}) , without loss of generality, we may assume that \eta_{n}\rightarrow \eta (n\rightarrow \infty) in \mathcal{C}_{b}^{1}(\mathbb{R}) , and \xi_{\mu_{n}}\rightharpoonup\xi_{\mu} weakly in \mathscr{E} as n\rightarrow \infty . Denote \xi_{v_{n}}(s) = \widetilde{U}_{\eta_{n}}(s, \mu_{n})\xi_{\mu_{n}} the corresponding solutions, then v_{n} solves
\begin{align} \partial_{s}^{2}v_{n}-\Delta v_{n}+\eta_{n}(s)\partial_{s}v_{n}+g(v_{n}) = f, \quad s\geq\mu_{n}\text{ and }\xi_{v_{n}}(\mu_{n}) = \xi_{\mu_{n}}. \end{align} | (4.6) |
Taking the multiplier \partial_{s}v_{n}+\varepsilon v_{n} with 0 < \varepsilon\ll1 in Eq (4.6), we derive the following energy type identity:
\begin{align} \frac{d}{ds}\mathcal{E}_{v_{n}}(s)+\varrho\mathcal{E}_{v_{n}}(s)+\mathcal{Q}_{v_{n}}(s)+\mathcal{G}_{v_{n}}(s)+ \mathcal{F}_{v_{n}}(s) = 0, \end{align} | (4.7) |
where
\begin{align*} \mathcal{E}_{v_{n}}& = \|\partial_{s}v_{n}\|^{2}+\|v_{n}\|_{\mathcal{H}^{1}}^{2}+2\varepsilon\langle \partial_{s}v_{n},v_{n}\rangle+2\langle G(v_{n}),1\rangle-2\langle f,v_{n}\rangle, \\\mathcal{Q}_{v_{n}}& = (2\eta_{n}-2\varepsilon-\varrho)\|\partial_{s}v_{n}\|^{2}+ (2\varepsilon-\varrho)\|v_{n}\|_{\mathcal{H}^{1}}^{2}+2\varepsilon(\eta_{n}-\varrho)\langle\partial_{s}v_{n},v_{n}\rangle, \\\mathcal{G}_{v_{n}}& = 2[\varepsilon\langle g(v_{n}),v_{n}\rangle-\varrho\langle G(v_{n}),1\rangle], \quad\mathcal{F}_{v_{n}} = 2(\varrho-\varepsilon)\langle f,v_{n}\rangle. \end{align*} |
Now, integrate Eq (4.7) with respect to s\in[-\mu_{n}, 0] to deduce that
\begin{align} \mathcal{E}_{v_{n}}(0)+\int_{-\mu_{n}}^{0}e^{\varrho r}(\mathcal{Q}_{v_{n}}(r)+\mathcal{G}_{v_{n}}(r)+\mathcal{F}_{v_{n}}(r))dr = \mathcal{E}_{v_{n}}(0)e^{-\varrho\mu_{n}}. \end{align} | (4.8) |
Our intention now is to pass to the limit n\rightarrow \infty in equality (4.8). To do this, we remind that \xi_{v_{n}} is uniformly bounded in \mathcal{C}((-\mu_{n}, \infty], \mathscr{E}) and \eta_{n}\rightarrow \eta (n\rightarrow \infty) in \mathcal{C}_{b}^{1}(\mathbb{R}) , then we get that
\begin{align*} \Xi_{n} = (\xi_{v_{n}},\eta_{n})\rightharpoonup\Xi = (\xi_{v},\eta), \quad\text{in }\;\mathcal{C}(\mathbb{R}, \mathbb{X}_{w}) \end{align*} |
and \Xi = (\xi_{v}, \eta)\in\bar{\mathfrak{E}}((-\infty, \infty)) = \mathfrak{E}((-\infty, \infty)) by recalling Corollary 3.12. In addition, we also know that \xi_{v} is an S–S solution and \xi_{v_{n}}(0)\rightharpoonup\xi_{v}(0) weakly in \mathscr{E} . Applying the compact embedding \mathcal{C}_{loc}((-\infty, 0], \mathscr{E})\subset\subset\mathcal{C}_{loc}((-\infty, 0], \mathcal{H}) , we can get that v_{n}\rightarrow v strongly in \mathcal{C}_{loc}((-\infty, 0], \mathcal{H}) , including almost everywhere. On the other hand, from the assumption (1.3), we can choose \varrho = \frac{\varepsilon}{4} , which guarantees that \mathcal{G}_{v_{n}}(s)\geq -\kappa_{2}|\Omega| and choose 0 < \varepsilon\leq\varepsilon_{0} small enough such that the quadratic form \mathcal{Q}_{v_{n}} is positive definite and satisfying C_{1}\|\xi_{v}\|_{\mathscr{E}}^{2}\leq\mathcal{Q}_{v}\leq C_{2}\|\xi_{v}\|_{\mathscr{E}}^{2} . Now, using the Fatou lemma, we conclude that
\begin{align} 0 = \liminf\limits_{n\rightarrow \infty}\Big(&\mathcal{E}_{v_{n}}(0) +\int_{-\mu_{n}}^{0}e^{\varrho r}(\mathcal{Q}_{v_{n}}(r)+\mathcal{G}_{v_{n}}(r)+\mathcal{F}_{v_{n}}(r))dr\Big) \\\geq& \mathcal{E}_{v}(0)+\int_{-\infty}^{0}e^{\varrho r}(\mathcal{Q}_{v}(r)+\mathcal{G}_{v}(r)+\mathcal{F}_{v}(r))dr. \end{align} | (4.9) |
According to Theorem 3.11, v is an S–S solution with more regularity in \mathscr{E}^{1} , and obviously v satisfies the energy equality. Then, by repeating the derivation of (4.8), for solution v , we obtain the energy equality
\begin{align} \mathcal{E}_{v}(0)+\int_{-\infty}^{0}e^{\varrho r}(\mathcal{Q}_{v}(r)+\mathcal{G}_{v}(r)+\mathcal{F}_{v}(r))dr = 0. \end{align} | (4.10) |
Combining (4.9) and (4.10), we ascertain
\begin{align} \liminf\limits_{n\rightarrow \infty}\mathcal{E}_{v_{n}}(0) = \mathcal{E}_{v}(0). \end{align} | (4.11) |
Applying the Fatou lemma and weak lower semi-continuous of the norm, we find that
\begin{align} \liminf\limits_{n\rightarrow \infty}\langle G(v_{n}(0)),1\rangle\geq\langle G(v(0)),1\rangle,\quad \liminf\limits_{n\rightarrow \infty}\|\xi_{v_{n}}(0)\|_{\mathscr{E}}^{2}\geq\|\xi_{v}(0)\|_{\mathscr{E}}^{2}. \end{align} | (4.12) |
Obviously, energy equality is true only when inequalities (4.12) are also equalities. Recalling \xi_{v_{n}}(0)\rightharpoonup\xi_{v}(0) , we obtain U_{\eta_{n}}(0, \mu_{n})\xi_{\mu_{n}} = \xi_{v_{n}}(0)\rightarrow \xi_{v}(0) strongly in \mathscr{E} . Finally, we get the uniformly asymptotically compact of the processes \{U_{\sigma}(t, \tau)\}_{\sigma\in\Sigma, t\geq\tau} and the theorem is proved.
Theorem 4.4. Under Assumption 1.1 assume that \sigma is translation compact in \mathcal{C}_{b}^{2}(\mathbb{R}) . Then the global attractor \mathscr{A}_{s} of the solution semigroup \mathbb{S}(t) defined in (3.2) is a bounded set in \mathscr{E}^{1}\times\mathcal{C}_{b}^{2}(\mathbb{R}) .
Proof. For any initial data (\xi_{u_{0}}, \sigma_{0})\in\mathscr{A}_{s} , we will prove that
\begin{align} \|\xi_{u}(t)\|_{\mathscr{E}^{1}}^{2}\leq e^{-\frac{\lambda}{8}t}\mathcal{Q}(\|\xi_{u_{0}}\|_{\mathscr{E}^{1}}^{2}+\|f\|^{2})+ C_{\|f\|^{2},\mathscr{A}_{s}},\quad\forall t\geq0, \end{align} | (4.13) |
where the positive constant \lambda and monotone increasing function \mathcal{Q}(\cdot) are independent of u and t . We proceed in three steps:
Step 1. Claim #1: Denote the restriction of the trajectory in \mathfrak{E}((-\infty, \infty)) to the time interval t\in[0, 1] as
\begin{align*} \mathcal{F}: = \{u|_{t\in[0,1]},\Xi = (\xi_{u},\sigma)\in\mathfrak{E}((-\infty,\infty))\}. \end{align*} |
Then \mathcal{F} is a compact set of L^{4}(0, 1;L^{12}(\Omega)) :
\begin{align} \mathcal{F}\subset\subset L^{4}(0,1;L^{12}(\Omega)). \end{align} | (4.14) |
Proof of claim. First note that the attractor \mathscr{A}_{s} is compact in \mathscr{E}\times\mathcal{C}_{b}^{2}(\mathbb{R}) , then there exists T = T(\mathscr{A}_{s}) > 0 such that
\begin{align*} \|u\|_{L^{4}(0,T;L^{12}(\Omega))}\leq C \end{align*} |
for any S–S solution u(t) with \xi_{u}(0)\in\Pi_{1}\mathscr{A}_{s} , where C may depend on \mathscr{A}_{s} , but is independent of u . Indeed, we obtain
\begin{align} \|u\|_{L^{4}(0,T;L^{12}(\Omega))}\leq C,\qquad\forall \; \Xi = (\xi_{u},\sigma)\in\mathfrak{E}((-\infty,\infty)). \end{align} | (4.15) |
Using \mathfrak{E}((-\infty, \infty)) is invariant with respect to time shifts, for any \Xi = (\xi_{u}, \sigma)\in\mathfrak{E}((-\infty, \infty)) we have
\begin{align} \sup\limits_{t\in\mathbb{R}}\|u\|_{L^{4}(t,t+1;L^{12}(\Omega))}\leq C_{0} \end{align} | (4.16) |
for some positive constant C_{0} , independent of u . Now let \xi_{v_{i}} are two S–S solutions of Eq (2.3) with symbols \eta_{i} , i = 1, 2 , and denote \xi_{w} = \xi_{v_{1}}-\xi_{v_{2}} , then recalling (4.5) and applying (4.15) or (4.16) to find that
\begin{align} \|\xi_{w}(s)\|_{\mathscr{E}}^{2} \leq Ce^{Ks} (\|\xi_{w}(0)\|_{\mathscr{E}}^{2}+C\|\eta_{1}-\eta_{2}\|^{2}_{\mathcal{C}_{b}^{1}(\mathbb{R})}),\quad\forall s\in[0,1], \end{align} | (4.17) |
where the constants C and K are independent of \xi_{v_{i}}(0) , i = 1, 2 . Then, applying (1.2) and (4.17), we have
\begin{align} \|g(v_{1})-g(v_{2})\|_{L^{1}(0,1;L^{2}(\Omega))} &\leq C\int_{0}^{1}(1+\|v_{1}\|_{L^{12}(\Omega)}^{4}+\|v_{2}\|_{L^{12}(\Omega)}^{4})\|v_{1}-v_{2}\|_{L^{6}(\Omega)}ds \\&\leq C(\|\xi_{w}(0)\|_{\mathscr{E}}+C\|\eta_{1}-\eta_{2}\|_{\mathcal{C}_{b}^{1}(\mathbb{R})}). \end{align} | (4.18) |
Applying Strichartz estimates for Eq (4.1) and recalling (4.17) and (4.18) gives us
\begin{align} \|v_{1}-v_{2}\|_{L^{4}(0,1;L^{12}(\Omega))} \leq C(\|\xi_{w}(0)\|_{\mathscr{E}}+\|\eta_{1}-\eta_{2}\|_{\mathcal{C}_{b}^{1}(\mathbb{R})}). \end{align} | (4.19) |
Assertion (4.14) is now a consequence of (4.19) and the general fact that \mathscr{A}_{s} is compact in \mathscr{E}\times\mathcal{C}_{b}^{2}(\mathbb{R}) .
Step 2. Claim #2: For any \varepsilon > 0 and any \Xi = (\xi_{u}, \sigma)\in\mathfrak{E}((-\infty, \infty)) , we can split the solution into two parts u = \bar{u}+\tilde{u} , where
\begin{align} \sup\limits_{t\geq0}\|\tilde{u}\|_{L^{4}(t,t+1;L^{12}(\Omega))}\leq\varepsilon\quad\text{and}\quad \|\bar{u}(t)\|_{L^{\infty}(\mathbb{R}^{+};\mathcal{H}^{2})}\leq C_{\varepsilon}. \end{align} | (4.20) |
Here, the constant C_{\varepsilon} depends on \varepsilon , but is independent of u .
Proof of claim. In fact, \forall\varepsilon > 0 , there exists a finite \varepsilon -net \{(y_{i}, \sigma_{i})\}_{i = 1}^{m}\subset\mathcal{D} satisfying
\begin{align} \mathcal{D}\subset\bigcup\limits_{1\leq i\leq m} B_{L^4(0,1 ; L^{12}(\Omega))\times\mathcal{C}_{b}^{2}(\mathbb{R})}\left((y_{i},\sigma_{i}), \frac{\varepsilon}{4}\right), \end{align} | (4.21) |
where \mathcal{D}: = \{(u, \sigma)|_{t\in[0, 1]}, \Xi = (\xi_{u}, \sigma)\in\mathfrak{E}((-\infty, \infty))\} and B_X\left(x_0, r\right) denotes the r -ball centered on x_0 in the space X . By (4.16), we also have
\begin{align*} \sup _{1 \leq i \leq m}\left\|y_{i}\right\|_{L^4(0,1 ; L^{12}(\Omega))} \leq C_{0}. \end{align*} |
Then approximate y_i by a smoother function \tilde{y}_{i} such that
\begin{align} \|\tilde{y}_i-y_i\|_{L^4(0,1 ; L^{12}(\Omega))} \leq \frac{\varepsilon}{4} \quad \text { and } \quad\|\tilde{y}_i\|_{\mathcal{C}(0,1 ;\mathcal{H}^2)} \leq C_{\varepsilon}, \end{align} | (4.22) |
where i = 1, 2, \cdots, m and the constant C_{\varepsilon} is independent of y_i . Combining (4.21) and (4.22), we otain
\begin{align} \mathcal{D}\subset\bigcup\limits_{1\leq i\leq m} B_{L^4(0,1 ; L^{12}(\Omega))\times\mathcal{C}_{b}^{2}(\mathbb{R})}\left((\tilde{y}_{i},\sigma_{i}), \frac{\varepsilon}{2}\right). \end{align} | (4.23) |
For every \Xi = (\xi_{u}, \sigma)\in\mathfrak{E}((-\infty, \infty)) , we observe that
\begin{align} (u,\sigma)|_{t\in[n,n+1]} = \mathbb{S}(n)(u,\sigma)|_{t\in[0,1]}\in\mathcal{D}. \end{align} | (4.24) |
So in view of (4.23) and (4.24), there exists (\tilde{y}_{i_{n}}, \sigma_{i_{n}}) such that
\begin{align} (u,\sigma)|_{t\in[n,n+1]}\in B_{L^4(0,1 ; L^{12}(\Omega))\times\mathcal{C}_{b}^{2}(\mathbb{R})}\left((\tilde{y}_{i_{n}},\sigma_{i_{n}}), \frac{\varepsilon}{2}\right). \end{align} | (4.25) |
Define the function \tilde{u}(t) as
\begin{align*} \bar{u}(t) = \tilde{y}_{i_{n}}(t-n),\quad\text{ if }\;t\in[n,n+1),\; \forall n\in\mathbb{N}, \end{align*} |
and the function \tilde{u}(t) = u(t)-\bar{u}(t) . Then
\begin{align} \|\bar{u}(t)\|_{L^{\infty}(\mathbb{R}^{+};\mathcal{H}^{2})} \leq\sup\limits_{n\in\mathbb{N}}\|\tilde{y}_{i_{n}}\|_{\mathcal{C}(n,n+1;\mathcal{H}^{2})} \leq C_{\varepsilon}. \end{align} | (4.26) |
For any t\geq0 , [t, t+1]\subset[n, n+2) for some n , then combining (4.24) and (4.25) leads us to the estimate
\begin{align} \|\tilde{u}\|_{L^{4}(t,t+1;L^{12}(\Omega))} &\leq\|(u-\bar{u},\sigma-\sigma_{i_{n}})\|_{L^{4}(n,n+1;L^{12}(\Omega))\times\mathcal{C}_{b}^{2}(\mathbb{R})} \\&\quad\qquad+\|(u-\bar{u},\sigma-\sigma_{i_{n+1}})\|_{L^{4}(n+1,n+2;L^{12}(\Omega))\times\mathcal{C}_{b}^{2}(\mathbb{R})} \\& = \|\mathbb{S}(n)(u,\sigma)-(\tilde{y}_{i_{n}},\sigma_{i_{n}})\|_{L^{4}(0,1;L^{12}(\Omega))\times\mathcal{C}_{b}^{2}(\mathbb{R})} \\&\qquad\quad+\|\mathbb{S}(n+1)(u,\sigma)-(\tilde{y}_{i_{n+1}},\sigma_{i_{n+1}})\|_{L^{4}(0,1;L^{12}(\Omega))\times\mathcal{C}_{b}^{2}(\mathbb{R})} \leq\varepsilon. \end{align} | (4.27) |
Hence, (4.26) and (4.27) imply the stated assertion (4.20) easily.
Step 3. The following estimates will be deduced by a formal argument, which can be justified by using Faedo–Galerkin method. Differentiating Eq (2.3) and setting \theta(s) = \partial_{s}v , we see that
\begin{align} \partial_{s}^{2}\theta-\Delta \theta+\eta(s)\partial_{s}\theta+\eta'(s)\theta+g'(v)\theta = 0 \end{align} | (4.28) |
with the initial condition
\begin{align} \xi_{\theta}(0) = (\partial_{s}v(0),\partial_{s}^{2}v(0)) = (v_{1},\Delta v_{0}-g(v_{0})-\eta(0)v_{1}+f)\in\mathscr{E}. \end{align} | (4.29) |
Taking the multiplier \partial_{s}\theta+\lambda\theta in (4.28), we can discover
\begin{align*} \frac{d}{ds}\mathcal{E}_{\theta}(s) +\mathcal{Q}_{\theta}(s)+\mathcal{G}_{\theta}(s) = 0, \end{align*} |
where
\begin{align*} \mathcal{E}_{\theta}(s)& = \|\partial_{s}\theta\|^{2}+\|\nabla\theta\|^{2}+\lambda\langle\partial_{s}\theta,\theta\rangle, \\\mathcal{Q}_{\theta}(s)& = (2\eta(s)-\lambda)\|\partial_{s}\theta\|^{2}+\lambda\|\nabla\theta\|^{2} +(2\eta'(s)+\lambda\eta(s))\langle\theta,\partial_{s}\theta\rangle +\lambda\eta'(s)\|\theta\|^{2}, \\\mathcal{G}_{\theta}(s)& = \lambda\langle g'(v),\theta^{2}\rangle+2\langle g'(v)\theta,\partial_{s}\theta\rangle. \end{align*} |
Choosing \lambda small enough such that
\begin{align} \mathcal{E}_{\theta}\cong\|\xi_{\theta}\|_{\mathscr{E}}^{2},\quad\text{and}\quad \frac{d}{ds}\mathcal{E}_{\theta}(s)+\frac{\lambda}{2}\mathcal{E}_{\theta}(s) \leq C_{\lambda,\|\eta\|_{\mathcal{C}_{b}^{1}(\mathbb{R})}}\|\theta(s)\|^{2}-2\langle g'(v)\theta,\partial_{s}\theta\rangle. \end{align} | (4.30) |
We employ the decomposition (4.20) and then let
\begin{align*} \bar{v}(x,s) = \bar{u}(x,\phi^{-1}(s)),\quad \tilde{v}(x,s) = \tilde{u}(x,\phi^{-1}(s)), \end{align*} |
to discover
\begin{align} |&\langle g'(v)\theta,\partial_{s}\theta\rangle| \leq|\langle (g'(\tilde{v}+\bar{v})-g'(\bar{v}))\theta,\partial_{s}\theta\rangle|+ |\langle g'(\bar{v})\theta,\partial_{s}\theta\rangle| \\\leq &C\langle(1+|\bar{v}|^{3}+|\tilde{v}|^{3})|\tilde{v}|,|\theta||\partial_{s}\theta|\rangle +\|g'(\bar{v})\|_{L^{\infty}}\|\theta\|\|\partial_{s}\theta\| \\\leq &C(1+\|\tilde{v}\|_{L^{12}(\Omega)}^{3}+\|\bar{v}\|_{L^{12}(\Omega)}^{3})\|\tilde{v}\|_{L^{12}(\Omega)}\|\theta\|_{L^{6}(\Omega)}\|\partial_{s}\theta\| +C(1+\|\bar{v}\|_{\mathcal{H}^{2}}^{4})\|\partial_{s}v\|\|\partial_{s}\theta\| \\\leq&C(1+\|\tilde{v}\|_{L^{12}(\Omega)}^{3}+\|\bar{v}\|_{L^{12}(\Omega)}^{3})\|\tilde{v}\|_{L^{12}(\Omega)} \|\xi_{\theta}\|_{\mathscr{E}}^{2}+ \frac{\lambda}{4}\|\partial_{s}\theta\|^{2}+C_{\lambda,\mathscr{A}_{s},\|\eta\|_{\mathcal{C}_{b}^{1}(\mathbb{R})}}\|\partial_{s}v\|^{2} \\\leq&l_{\varepsilon}(s)\|\xi_{\theta}\|_{\mathscr{E}}^{2}+C_{\lambda,\mathscr{A}_{s},\|\eta\|_{\mathcal{C}_{b}^{1}(\mathbb{R})} }\|\partial_{s}v\|^{2} +\frac{\lambda}{4}\|\partial_{s}\theta\|^{2}, \end{align} | (4.31) |
where l_{\varepsilon}(s) = C(1+\|\tilde{v}\|_{L^{12}(\Omega)}^{3}+\|\bar{v}\|_{L^{12}(\Omega)}^{3})\|\tilde{v}\|_{L^{12}(\Omega)} . Owing to (4.16) and (4.20), we conclude
\begin{align} &\int_{s}^{s+1}l_{\varepsilon}(r)dr\leq C\left(\int_{s}^{s+1}(1+\|\bar{v}\|_{L^{12}(\Omega)}^{3}+\|\tilde{v}\|_{L^{12}(\Omega)}^{3})^{\frac{4}{3}}dr\right)^{\frac{3}{4}} \left(\int_{s}^{s+1}\|\tilde{v}\|_{L^{12}(\Omega)}^{4}dr\right)^{\frac{1}{4}} \\\leq &C\left(1+\|\tilde{v}\|_{L^{4}(s,s+1;L^{12}(\Omega))}^{3}+\|\bar{v}\|_{L^{4}(s,s+1;L^{12}(\Omega))}^{3}\right) \|\tilde{v}\|_{L^{4}(s,s+1;L^{12}(\Omega))} \leq C \varepsilon \end{align} | (4.32) |
for some positive constant C independent of \varepsilon . Combining now (4.30) and (4.31) and employing Gronwall's inequality, we deduce
\begin{align} \|\xi_{\theta}(s)\|_{\mathscr{E}}^{2} \leq e^{-\int_{0}^{s}(\frac{\lambda}{4}-l_{\varepsilon}(r))dr}\mathcal{Q}(\|\xi_{\theta}(0)\|_{\mathscr{E}}^{2}) +C\int_{0}^{s}e^{-\int_{0}^{r}(\frac{\lambda}{4}-l_{\varepsilon}(\mu))d\mu}\|\xi_{v}(r)\|_{\mathscr{E}}^{2}dr \end{align} | (4.33) |
for some monotone function \mathcal{Q}(\cdot) and positive constant \lambda , which are independent of \varepsilon and v . We estimate using (4.32) and (4.33):
\begin{align} \|\xi_{\theta}(s)\|_{\mathscr{E}}^{2} \leq e^{-\frac{\lambda}{8}s}\mathcal{Q}(\|\xi_{\theta}(0)\|_{\mathscr{E}}^{2}) +C\|\xi_{v}\|_{\mathcal{C}(\mathbb{R}_{+};\mathscr{E})}^{2} \leq e^{-\frac{\lambda}{8}s}\mathcal{Q}(\|\xi_{\theta}(0)\|_{\mathscr{E}}^{2}) +C_{\|f\|^{2},\mathscr{A}_{s}}. \end{align} | (4.34) |
Recalling now (4.29), we see that in fact
\begin{align*} \|\xi_{\theta}(0)\|_{\mathscr{E}}^{2} \leq C(\|\xi_{v}(0)\|_{\mathscr{E}^{1}}^{2}+\|f\|^{2}). \end{align*} |
Inserting this estimate into (4.34), we discover that
\begin{align} \|\xi_{\theta}(s)\|_{\mathscr{E}}^{2} \leq e^{-\frac{\lambda}{8}s}\mathcal{Q}(\|\xi_{v}(0)\|_{\mathscr{E}^{1}}^{2}+\|f\|^{2}) +C_{\|f\|^{2},\mathscr{A}_{s}}. \end{align} | (4.35) |
Recalling (2.3) and employing (1.2), we deduce that
\begin{align} \|v(s)\|_{\mathcal{H}^{2}}^{2}\leq C(\|f\|^{2}+\|\xi_{\theta}(s)\|^{2}). \end{align} | (4.36) |
Combining (4.35) and (4.36) and remembering (2.5), we derive the estimate (4.13). The estimate, together with the invariance of \mathscr{A}_{s} , completes the proof.
Corollary 4.5. Under the assumptions of Theorem 4.4, the family of processes \{U_{\sigma}(t, \tau)\} , \sigma\in\Sigma corresponding to Eq (1.1) has a compact uniform ( w.r.t. \sigma\in\Sigma ) attractor \mathscr{A}_{s}^{\Sigma} which is bounded in the phase space \mathscr{E}^{1} .
Indeed, applying Theorems 4.3 and 4.4, we can state the result on the boundedness of the strong uniform attractor \mathscr{A}_{s}^{\Sigma} in \mathscr{E}^{1} .
We have investigated the dynamical behavior of a wave equation with time-dependent coefficients and quintic nonlinearity on a bounded domain, and established results on the existence and smoothness of a uniform attractor \mathscr{A}_{s}^{\Sigma} in natural energy spaces \mathscr{E} .
There is still much work that needs to be done in this field. For example, the continuity of pullback attractors for Eq (1.1) with cubic nonlinearity g was studied by Aragão et al. in [1]. Hence, a natural question is: Is it possible to otain similar results for Eq (1.1) when the non-linearity g is assumed to have a sub-quintic or quintic rate? As we have already mentioned, the key difficulty in this problem is establishing the so-called ETS estimate (2.2). Nevertheless, we conjecture it is true, at least in the sub-quintic case. On the other hand, up to the moment, we do not know how to establish an ETS estimate in the quintic case, and this can be regarded as an open problem.
Feng Zhou: Conceptualisation, writing-original draft, formal analysis, project administration, writing-review and editing; Hongfang Li: Supervision and editiong; Kaixuan Zhu: Editing and project administration; Xin Li: Review and editing. All authors have read and approved the final version of the manuscript for publication.
The authors declare that they have not used Artificial Intelligence (AI) tools in the creation of this article.
The authors would like to thank the anonymous referees for their interesting comments on this work. This work is supported by the NSFC of China (No. 11601522, 12201421); Natural Science Foundation of Shandong Province (ZR2021MA025, ZR2021MA028); the Hunan Province Natural Science Foundation of China (Grant Nos. 2022JJ30417, 2024JJ5288); the Provincial Natural Science Foundation of Hebei Grant No. A 2022203004; the Fund by Science Research Project of Hebei Education Department Grant No. QN2020203.
The authors declare that there are no conflicts of interest regarding the publication of this paper.
[1] |
G. S. Aragão, F. D. M. Bezerra, R. N. Figueroa-López, M. J. D. Nascimento, Continuity of pullback attractors for evolution processes associated with semilinear damped wave equations with time-dependent coefficients, J. Differ. Equ., 298 (2021), 30–67. https://doi.org/10.1016/j.jde.2021.06.036 doi: 10.1016/j.jde.2021.06.036
![]() |
[2] | A. Babin, M. I. Vishik, Attractors of evolutionary equations, Nauka, Moscow, 1989. |
[3] |
J. M. Ball, Global attractors for damped semilinear wave equations, Discr. Cont. Dyn. Syst., 10 (2004), 31–52. https://doi.org/10.3934/dcds.2004.10.31 doi: 10.3934/dcds.2004.10.31
![]() |
[4] |
F. D. M. Bezerra, R. N. Figueroa-López, M. J. D. Nascimento, Fractional oscillon equations: Continuity properties of attractors with respect to order of the equations, Nonlinearity, 36 (2023), 1218–1244. https://doi.org/10.1088/1361-6544/acad5c doi: 10.1088/1361-6544/acad5c
![]() |
[5] |
Q. Q. Chang, D. D. Li, C. Y. Sun, S. V. Zelik, Deterministic and random attractors for a wave equation with sign changing damping, (Russian) Izv. Ross. Akad. Nauk Ser. Mat., 87 (2023), 154–199. https://doi.org/10.4213/im9250e doi: 10.4213/im9250e
![]() |
[6] | V. V. Chepyzhov, M. I. Vishik, Attractors for equations of mathematical physics, Amer. Math. Soc. Colloq. Publ., Amer. Math. Soc., Providence, RI, 2002. |
[7] |
A. Cheskidov, Global attractors of evolutionary systems, J. Dyn. Differ. Equ., 21 (2009), 249–268. https://doi.org/10.1007/s10884-009-9133-x doi: 10.1007/s10884-009-9133-x
![]() |
[8] |
A. Cheskidov, L. Kavlie, Pullback attractors for generalized evolutionary systems, Discrete Cont. Dyn.-B, 20 (2015), 749–779. https://doi.org/10.3934/dcdsb.2015.20.749 doi: 10.3934/dcdsb.2015.20.749
![]() |
[9] |
A. Cheskidov, L. Kavlie, Degenerate pullback attractors for the 3D Navier-Stokes equations, J. Math. Fluid Mech., 17 (2015), 411–421. https://doi.org/10.1007/s40272-015-0146-0 doi: 10.1007/s40272-015-0146-0
![]() |
[10] |
A. Cheskidov, S. S. Lu, Uniform global attractors for the nonautonomous 3D Navier-Stokes equations, Adv. Math., 267 (2014), 277–306. https://doi.org/10.1016/j.aim.2014.09.005 doi: 10.1016/j.aim.2014.09.005
![]() |
[11] | I. Chueshov, Dynamics of quasi-stable dissipative dystems, Springer, New York, 2015. |
[12] |
M. Conti, V. Pata, R. Temam, Attractors for the processes on time-dependent spaces. Application to wave equations, J. Differ. Equations, 255 (2013), 1254–1277. https://doi.org/10.1016/j.jde.2013.05.013 doi: 10.1016/j.jde.2013.05.013
![]() |
[13] |
M. Conti, V. Pata, Asymptotic structure of the attractor for processes on time-dependent spaces, Nonlinear Anal.-Real, 19 (2014), 1–10. https://doi.org/10.1016/j.nonrwa.2014.02.002 doi: 10.1016/j.nonrwa.2014.02.002
![]() |
[14] |
V. Kalantarov, A. Savostianov, S. Zelik, Attractors for damped quintic wave equations in bounded domains, Ann. Henri Poincaré, 17 (2016), 2555–2584. https://doi.org/10.1007/s00023-016-0480-y doi: 10.1007/s00023-016-0480-y
![]() |
[15] |
X. Y. Mei, K. X. Zhu, Asymptotic behavior of solutions for hyperbolic equations with time-dependent memory kernels, Discrete Cont. Dyn.-B, 28 (2023), 1855–1885. https://doi.org/10.3934/dcdsb.2022150 doi: 10.3934/dcdsb.2022150
![]() |
[16] |
J. Metcalfe, D. Tataru, Global parametrices and dispersive estimates for variable coefficient wave equations, Math. Ann., 353 (2012), 1183–1237. https://doi.org/10.1007/s00208-011-0714-8 doi: 10.1007/s00208-011-0714-8
![]() |
[17] | A. Savostianov, Strichartz estimates and smooth attractors of dissipative hyperbolic equations, (Doctoral dissertation), University of Surrey, 2015. |
[18] |
C. Y. Sun, D. M. Cao, J. Q. Duan, Uniform attractors for non-autonomous wave equations with nonlinear damping, SIAM J. Appl. Dyn. Syst., 6 (2007), 293–318. https://doi.org/10.1126/science.318.5848.293 doi: 10.1126/science.318.5848.293
![]() |
[19] |
H. Uesaka, A pointwise oscillation property of semilinear wave equations with time-dependent coefficients Ⅱ, Nonlinear Anal., 47 (2001), 2563–2571. https://doi.org/10.1016/S0362-546X(01)00378-9 doi: 10.1016/S0362-546X(01)00378-9
![]() |
[20] |
S. Zelik, Asymptotic regularity of solutions of singularly perturbed damped wave equations with supercritical nonlinearities, Discrete Cont. Dyn.-A, 11 (2004), 351–392. https://doi.org/10.3934/dcds.2004.11.351 doi: 10.3934/dcds.2004.11.351
![]() |