This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
Introduction
Pullback attractors have become a suitable tool for studying non-autonomous dissipative dynamical systems generated by evolution equations arising in physical phenomena. Pullback global attractor is a family of compact subsets of the phase space which is strictly invariant under the evolution process and attracts all bounded subsets in a pullback sense. For pullback exponential attractor, we have a positively invariant family of compact subsets which have a uniformly bounded fractal dimension and pullback attract all bounded subsets at an exponential rate. We recall the definitions from [4]:
Definition 1
The two parameter family {U(t, s)|, t, s ∈ ℝ, t ≥ s} of continuous operators from the metric space X to itself is called an evolution process in X if it satisfies the properties
The family of non-empty subsets {𝒜(t)|, t ∈ ℝ} of X is called a pullback global attractor for the process {U(t, s)|, t ≥ s} if 𝒜(t) is compact for all t, the family {𝒜(t)|, t ∈ ℝ} is strictly invariant, that is
$$\begin{array}{}
\displaystyle
U(t,s)\mathscr{A}(s)=\mathscr{A}(t) \quad for\,\, all \,\,\,t\geq s,
\end{array}$$
it pullback attracts all bounded subsets of X, that is for every bounded D ⊂ X and t ∈ ℝ
and the family is minimal within the families of closed subsets that pullback attract all bounded subsets of X.
Definition 3
We call the family 𝓜 = {𝓜(t)|, t ∈ ℝ} a pullback exponential attractor for the process {U(t, s)|, t ≥ s} in X if
the subsets 𝓜(t) ⊂ X are non-empty and compact in X for all t ∈ X,
the family is positively semi-invariant, that is
$$\begin{array}{}
\displaystyle
U(t,s)\mathscr{M}(s)\subset \mathscr{M}(t)\quad for\,\,all\,\,t\geq s
\end{array}$$
the fractal dimension in X of the sections 𝓜(t), t ∈ ℝ, is uniformly bounded and
the family {𝓜(t)|, t ∈ ℝ} exponentially pullback attracts bounded subsets of X; that is, there exists a positive constant ω > 0 such that for every bounded subset D ⊂ X and t ∈ ℝ
For some constructions of these attractors refer for example to [1, 4, 6, 7, 8, 9]. Reaction-diffusion equation, in this regard, has been studied as an example of dissipative systems.
where the forcing term h(t) is allowed to have an exponential growth in time.
When the phase space is L2(Ω), in [1], it was obtained a uniform bound on the dimensions of sections of the pullback global attractor under a Lipschitz condition on the nonlinear term f(t, u).
In [7], under some requirements for the phase space and the process defined on it, it was constructed a uniform bound for the dimensions of sections of the pullback exponential attractor and the pullback global attractor contained in it and as an application, it was applied to the reaction-diffusion equation in the phase space
$\begin{array}{}
\displaystyle
H^1_0
\end{array}$
(Ω).
We use the bound obtained in these works for proving that the dimension of sections is zero providing a bound on the diameter of the domain Ω which brings about a unique global solution as the pullback global attractor.
We, moreover, release the Lipschitz condition on f and replace it with a condition on fu which affords us to deal with such problems with a polynomial nonlinearity as Chafee-Infante equation and obtain its trivial dynamics under a bounded perturbation.
Statement of the problem
Let us consider the initial boundary value problem for the non-autonomous reaction-diffusion equation as follows:
where f ∈ C1(ℝ2, ℝ), h(⋅) ∈
$\begin{array}{}
\displaystyle
L_{loc}^2
\end{array}$
(ℝ, L2(Ω)), Ω is a bounded open subset of ℝn with smooth boundary ∂ Ω, and there exists p ≥ 2, ci > 0, i = 1, …, 6, such that
and 0 ≤ α < λ1 where λ1 > 0 is the first eigenvalue of the operator A = −Δ where Δ is the Laplace operator in L2(Ω) with zero Dirichlet boundary condition [7].
When the phase space is L2(Ω), in [1], it was proved that the problem (1) has a pullback global attractor in L2(Ω) with f satisfying (2), (3), (4) and under the condition of a polynomial bound on the forcing term instead of (5):
and obtained a uniform bound on the fractal dimension of its sections under an additional assumption that there exists a positive and nondecreasing function ξ : ℝ → (0, ∞) such that:
Here, we briefly state the assumptions on the process to have a pullback exponential attractor [7]:
Let we have a Banach space V and U(t, s) an evolution process on it. We assume that
There is a positively invariant family of closed bounded subsets of V, B(t).
B(t) grows exponentially in the past, i.e. diam(B(t)) < M eγ0tt ≤ t0, γ0 ≥ 0.
The family pullback absorbs all bounded subsets of V.
Next, we assume that the process has the decomposition U(t, s) = C(t, s) + S(t, s), where, {C(t, s) : t0 ≥ t ≥ s} and {S(t, s) : t0 ≥ t ≥ s} are families of operators satisfying the following properties:
There exists t̃ > 0 such that C(t, t − t̃) are contractions within the absorbing sets, where the contraction constant λ is independent of time and
$\begin{array}{}
\displaystyle
0 \leq \lambda \leq \frac{1}{2}e^{-\gamma_0 \tilde{t}}
\end{array}$
with γ0 ≥ 0 from 𝒜2,
There exists an auxiliary normed space (W, ∥⋅∥W) such that V is compactly embedded into W and S(t − t̃) satisfies the smoothing property within the absorbing sets,
The process is Lipschitz continuous within the absorbing sets.
In [7], theorem 2.2, it was shown that under these assumptions, there exists a pullback exponential attractor for the process and it was applied to the above problem in
$\begin{array}{}
\displaystyle
H^1_0
\end{array}$
(Ω) and obtained a uniform bound on the dimension of sections of its pullback global attractor under the assumption (6). Here, we state a useful corollary of it, to which our results rely upon.
Corollary 1
[7] Assume that the process {U(t, s) : t ≥ s} on a Banach spaceVsatisfies (𝒜1)-(𝒜3), (𝓗3) and admits the above decomposition with (𝓗1) and letν ∈
$\begin{array}{}
\displaystyle
(0,\frac{1}{2}e^{-\gamma_0\tilde{t}}-\lambda)
\end{array}$
. Assume further that
there existsN = Nν ∈ ℕ such that for anyt ≤ t0, anyR > 0 and anyu ∈ B(t − t̃), there existsv1, …, vn ∈ Vsuch that
For the existence of global solutions u(t) for the problem on both L2(Ω) and
$\begin{array}{}
\displaystyle
H^1_0
\end{array}$
(Ω) refer to [7], Theorem 4.1.
Define U(t, s)us : = u(t), then we have the evolution process {U(t, s) : t ≥ s} in L2(Ω) and let us denote by ∥⋅∥ the L2(Ω) norm.
Main Results
A priori estimate method is widely used to capture the asymptotic behavior of PDE systems. In the following lemma, using some standard estimates and a new trick, we obtain a useful estimate of the solutions suitable for our purpose.
Lemma 2
Consider the initial boundary value problem(1)with the assumptions(2)-(5)and let {U(t, s) : t ≥ s} be the process of it inL2(Ω). letus, vs ∈ L2(Ω) are two initial values andu(t), v(t) are the solutions corresponding to them. we have:
Let Ω1 = {x ∈ Ω:u(t, x) ≠ v(t, x)}, since f ∈ C1 as a function of u, by Mean value theorem ∀x ∈ Ω1 ∃z :=
$\begin{array}{}
\displaystyle
z(t,x)\in \mathbb{R}: \frac{f(t,u)-f(t,v)}{u-v}=f_u(z,t)
\end{array}$
. Note that z is between u and v as values in ℝ and Ω1 is a Lebesgue measurable set. Note also that fu(t, z) as a function of x is not necessarily even measurable but fu(t, z)|ω|2 = (f(t, u) − f(t, v))ω is an integrable function on Ω1. Hence, we have
Thus, using the Gronwall lemma, the relation (9) follows.
Let Vn = span {e1, …, en} be the linear space spanned by the first n eigenfunctions of A = −Δ in L2(Ω) and let Pn : L2(Ω) → Vn denotes the orthogonal projection and Qn its complementary projection. For u ∈ L2(Ω) we write u = Pn(u) + Qn(u) = u1 + u2. We consider the difference ω = u − v of two solutions of (1) with us, vs ∈ L2(Ω).Taking the inner product in L2(Ω) with Qn(ω) = ω2, we obtain
As in [7] (section 4), we can obtain the absorbing sets B(t) for the problem so that diam(B(t)) ≤
$\begin{array}{}
\displaystyle
Le^{-\frac{\alpha}{2}t}
\end{array}$t ≤ t0 ≤ 0.
As refers to the uniform bound on the dimension of sections of pullback global attractor, in [2, 3], the Lipschitz condition on f is as a sufficient condition.
Here, we make our new assumption which replaces the Lipschitz condition on f with an assumption on fu.
As before in the Lemma 2, set Ω1 = {x ∈ Ω:u(t, x) ≠ v(t, x)} and z = z(t, x). we have
Let t̃ > 0. B(t − t̃) is absorbing, Hence, for the solutions starting at t − t̃ in B(t − t̃), according to [11] (Theorem 11.6 and lines after its proof), we can get t̃ so large that u(t), v(t) ∈ L∞(Ω) and thus |u(t)| ≤
$\begin{array}{}
\displaystyle
Le^{-\frac{\alpha}{2}t}
\end{array}$
and |v(t)| ≤
$\begin{array}{}
\displaystyle
Le^{-\frac{\alpha}{2}t}
\end{array}$
almost everywhere(t̃ needs to be further adjusted in the following lines). Since z is between u and v, then, |z(t)| ≤
$\begin{array}{}
\displaystyle
Le^{-\frac{\alpha}{2}t}
\end{array}$
a.e. Since fu is continuous w.r.t. u, so let
This is our new assumption which takes the place of the Lipschitz condition in dealing with (10). According to (12), there exists η0 ≥ 0 such that |fu(t, z)| ≤ η0 a.e. for t ≤ t0.
In the sequel, we try to separate the third term in (10), that is, (11) into positive and negative parts and drop the positive one to remain the expression negative.
We have
We have
$\begin{array}{}
\int_{\Omega_1^+}\omega_1\omega_2\, dx+\int_{\Omega_1^-}\omega_1\omega_2\, dx=\int_{\Omega_1}\omega_1\omega_2\, dx=-\int_{\Omega-\Omega_1}\omega_1^2\leq 0,
\end{array}$
Then (𝓗1) is satisfied with C(t, t − t̃) = QnU(t, t − t̃).
Hence, in fulfilling the requirements of Corollary 1 to obtain a pullback exponential attractor, for meeting (𝓗1), we can replace the assumption (6) with the assumption (12).
Now, we state the main result in L2(Ω):
Theorem 3
Let {U(t, s) : t ≥ s} be the process specified in the Lemma 2 and {𝒜(t) : t ∈ ℝ} inL2(Ω) be the pullback global attractor related to it with a uniform bound on the dimension of its sections. if
then, for anyt ∈ ℝ the section 𝒜(t) has a zero fractal dimension.
Proof
In [7], by checking the requirements of Corollary 1, it was established a pullback global attractor with a uniform bound on its sections. We again check (H2) using Lemma 2. we proceed as follows: From (9) we have:
If
$\begin{array}{}
\displaystyle
\lambda_1-c_5\gt \frac{\alpha}{2}
\end{array}$
, taking s so negative that, fixing t̃ = t − s, we have
$\begin{array}{}
\displaystyle
\nu=e^{-(\lambda_1-c_5)\tilde{t}}\lt \frac{1}{2}e^{-\frac{\alpha}{2}\tilde{t}}-\lambda
\end{array}$.
Now, for this ν, any R > 0 and u ∈ B(t − t̃)
Thus, (H2) is satisfied with Nν = 1. Moreover, by (8), the fractal dimension of its sections is 0. Hence, the pullback global attractor has zero dimensional sections.
We know that attractors are connected sets and zero dimensional connected sets are singletons. Then, we have the following corollary.
Corollary 4
The pullback global attractor under the assumption(13)consists only of one global solution.
The assumption (13), is interpreted as a restriction on the bounded domain Ω, since λ1 is only depended on the geometry of the bounded domain. We know from the classic work [10] that if Ω is a bounded, convex, Lipschitz domain with diameter d, then the Poincaré constant is at most
$\begin{array}{}
\frac{d}{\pi}~\text{
for}~~ p = 2 ~(\text{in}~~ W_0^{1,p}(\Omega))
\end{array}$
. It is also well-known that for a smooth, bounded domain Ω, since the Rayleigh quotient for the Laplace operator in
$\begin{array}{}
\displaystyle
H^1_0
\end{array}$
(Ω) is minimized by the eigenfunction corresponding to the first eigenvalue λ1 of −Δ, we have
$\begin{array}{}
\displaystyle
\|u\|\leq \lambda_1^{-1} \|\nabla u\|
\end{array}$.
Hence, for our case, If Ω is a bounded, smooth and convex domain, we have
$\begin{array}{}
\displaystyle
\lambda_1^{-1}\leq \frac{d}{\pi}~~ \text{or}~~ \lambda_1\geq \frac{\pi}{d}
\end{array}$.
Now, (13) says
$\begin{array}{}
\displaystyle
\lambda_1\gt c_5+\frac{\alpha}{2}
\end{array}$
, Hence if
$\begin{array}{}
\displaystyle
c_5+\frac{\alpha}{2}\lt \frac{\pi}{d}
\end{array}$
, the assumption (13) automatically holds and then the zero dimensionality of 𝓜(t) holds true for the convex domains with
$\begin{array}{}
\displaystyle
d\lt \pi(c_5+\frac{\alpha}{2})^{-1}
\end{array}$
and for a convex Ω with larger diameters, we must check (13) to be held for assuring zero dimensionality of sections of pullback global attractor of reaction-diffusion equation defined on it.
Corollary 5
Consider the problem(1)on a bounded, smooth and convex domain Ω with the assumptions(2)-(5)and(12)and let {𝒜(t) : t ∈ ℝ} be the pullback global attractor inL2(Ω) associated to it. If diam(Ω) <
$\begin{array}{}
\displaystyle
\pi(c_5+\frac{\alpha}{2})^{-1}
\end{array}$
, then 𝒜(t) = {u(t)} whereuis a global solution of the problem.
In what follows, As an application to Theorem 3, we examine the asymptotic behavior of a new version of non-autonomous Chafee-Infante equation respecting the new assumption (12). Some other versions of this equation in stronger spaces have arisen in [5], [3], [2] and [7]. In [3], the unperturbed version of the following equation has been studied and the trivial asymptotic dynamics has been shown when λ < 1. We show that under a bounded perturbation, this trivial dynamics remains valid.
Example 6
Consider the non-autonomous Chafee-Infante equation on the domain (0, π) as follows:
with the initial conditionu(s, x) = u0(x), x ∈ (0, π) and 0 < b0 ≤ β(t) ≤ B0isC1[3] and ∥h(t)∥ ≤ M. Obviously, this equation meets the conditions(2)-(5)withf(t, u) = β(t)u3 − λu, α = 0 andc5 = λ. For checking(12), We havefu(t, u) = 3β(t)u2 − λand so obviously, g(t) is bounded in the past. Hence, Corollary 5 shows that its pullback global attractor inL2(0, π) consists only of one global solution ifλ < 1. Note thatfdoes not meet the Lipschitz condition(6)and as a result, we could not prove the uniform bound on the dimensions of the sections of its pullback global attractor by the results in [1] or [7].
Also, [3] has obtained trivial attractor {0} for the unperturbed equation whenλ < 1, while here the perturbed version has a global solution {u(t)} as its pullback global attractor whenλ < 1 which generalizes that result.
Moreover, this result is consistent with the result about this problem in [12], Section 4, that is, ifh(t) = ϵh1(t)r(x) whereh1 : ℝ → ℝ is an almost periodic function andr(x) ∈ L2(0, π) thenλ ∈ (0, 1) impliesn = 0, and all solutions converge to its unique almost periodic solutionu(t), so we have the trivial pullback global attractor as {u(t)} which its sections are one-point sets [[12], Theorem 4.1]. In fact, our result shows that not only for almost periodic perturbations but also for all bounded perturbations, the pullback global attractor only contains a global solution which, in some sense, improves the result of [12] forλ ∈ (0, 1).
Remark 1
When the phase space is
$\begin{array}{}
\displaystyle
H^1_0
\end{array}$
(Ω), according to [7], Theorem 4.1, the assumption (6) needs for the existence of solutions and hence, for defining U(t, s) on
$\begin{array}{}
\displaystyle
H^1_0
\end{array}$
(Ω). Thus, here we could not replace it with the new assumption (12).
Remark 2
In the case of
$\begin{array}{}
\displaystyle
H^1_0
\end{array}$
(Ω), the results of theorem 3 and its corollaries are valid with some modifications for the expressions (9) and (13) and a new bound on the diameter of the domain.