Volume 2014, Issue 1 237027
Research Article
Open Access

Pullback Exponential Attractor for Second Order Nonautonomous Lattice System

Shengfan Zhou

Corresponding Author

Shengfan Zhou

Department of Mathematics, Zhejiang Normal University, Jinhua 321004, China zjnu.edu.cn

Search for more papers by this author
Hong Chen

Hong Chen

Department of Mathematics, Zhejiang Normal University, Jinhua 321004, China zjnu.edu.cn

Search for more papers by this author
Zhaojuan Wang

Zhaojuan Wang

School of Mathematical Science, Huaiyin Normal University, Huaiyin 223300, China

Search for more papers by this author
First published: 03 April 2014
Citations: 3
Academic Editor: Zengji Du

Abstract

We first present some sufficient conditions for the existence of a pullback exponential attractor for continuous process on the product space of the weighted spaces of infinite sequences. Then we prove the existence and continuity of a pullback exponential attractor for second order lattice system with time-dependent coupled coefficients in the weighted space of infinite sequences. Moreover, we obtain the upper bound of fractal dimension and attracting rate for the attractor.

1. Introduction

Lattice dynamical systems (LDSs), which include coupled systems of infinite ordinary differential equations and coupled map lattices, have drawn more and more attention because these systems appear in various fields [1, 2]. In recent years, global attractors, uniform attractors, pullback attractors (or kernel sections), and random attractor for autonomous, nonautonomous, and stochastic LDSs have been studied; see [312]. However, these attractors sometimes attract orbits at a relatively slow speed, so that it might take an unexpected long time to be reached. Besides, it is usually difficult to estimate the attracting rate in terms of physical parameters of the model. Appropriate alternatives are exponential attractors and pullback exponential attractors which contain the global attractors and pullback attractors and attract all bounded sets exponentially [1320]. For LDSs, [13, 16, 21] studied the existence of exponential attractors for first order, second order, and partly dissipative autonomous LDSs, respectively. Zhou and Han [20] presented some sufficient conditions for the existence of pullback exponential attractors for LDSs in and provided applications to first order and partly dissipative nonautonomous LDSs. As we know, there are no results on pullback exponential attractors for second order nonautonomous LDSs which have been studied extensively [47, 913, 2123]. In this paper, we study the following second order nonautonomous LDSs:
()
where for i  ∈ , ui, gi    C(, ); fi    C1( × , ); ηi,j(t),  j = −q, …, q  (q  ∈ ) are locally integrable in t, and λi,   α are positive constants. First we present some sufficient conditions for the existence of a pullback exponential attractor for a continuous process on the product space of infinite sequences, which are direct results of [20]. Then we prove the existence of a pullback exponential attractor for system (1) in weighted space ; moreover, we obtain the upper bound of fractal dimension and attracting rate for the attractor.

2. Preliminaries

In this section, we present some sufficient conditions for the existence of a pullback exponential attractor for a continuous process on the product space of m weighted space of infinite sequences.

Let ρj : + be positive-valued function such that 0 < ρj(i) = ρj,iMj0, where Mj0 are positive constants, j = 1, …, m, and let
()
which is Banach space with norm for u = (ui) i ∈, . Let be a product space of m spaces , j = 1, …, m. Write
()
Then is (2N + 1)-dimensional subspaces of . Let and define a bounded projection as follows: for , ,
()

Consider a two-parameter continuous process on E: W(t, τ) : EE, tτ   ∈ , satisfying the following: W(τ, τ) = I, for all τ; W(t, r)W(r, τ) = W(t, τ), − < τrt < ; (t, τ, φτ) → W(t, τ)φτ is continuous for − < τt < .

Definition 1. A family {(t)} t ∈ of subsets of is called a pullback exponential attractor for the continuous process {W(t, τ)} tτ, if

  • (i)

    each (t)  (t  ∈ ) is a compact set of and its fractal dimension is uniformly bounded in t; that is, sup⁡t ∈dim⁡f(t) < ;

  • (ii)

    it is positively invariant; that is, W(t, τ)(τ) ⊂ (t) for all − < τt < ;

  • (iii)

    there exist an exponent α > 0 and two positive-valued functions Q, 𝒯 : ++ such that for any bounded set

    ()

where “dh(·, ·)” is the Hausdorff semidistance between two subsets of .

As a direct consequence of Theorem  2 of [20], we have the following theorem.

Theorem 2. Let {W(t, τ)} tτ be a continuous process in . Assume that there exists a uniformly bounded absorbing set for {W(t, τ)} tτ: for any bounded set , there exists a constant TB ≥ 0 such that W(t, τ)BB0 for all tτ + TB, τ  ∈ . For any t  ∈ , set . If

  • (A1) there exist and such that, for every τ  ∈ , any t  ∈   [0, T*], and φτ, ψτ    X(τ),

    ()

  • (A2) there exist a positive constant γ  ∈   [0, 1/2) and a (2N + 1)m-dimensional bounded projection such that, for every τ  ∈ and φτ, ψτ    X(τ),

    ()

Then {W(t, τ)} tτ possesses a pullback exponential attractor {(t)} t ∈  R satisfying
()
and, for any bounded set ,
()
where Nθ is the minimal number of closed balls of E with radius θ covering the closed unit ball BN(0,1) of centered at 0, and with , α = −ln⁡θ/T*, and .

3. Pullback Exponential Attractor for Second Order Nonautonomous Lattice System

In this section, we study the existence of a pullback exponential attractor for the continuous processes associated with the second order nonautonomous lattice system (1). Note that system (1) with initial conditions can be written as a vector form
()
where u = (ui) i ∈, , λu = (λiui) i ∈, f(u, t) = (fi(ui, t)) i ∈, g(t) = (gi(t)) i ∈.
Throughout this section, let ρ be a positive weight function from to (0, M0) ⊂ + satisfying
  • (P0)

    0 < ρ(i) ≤ M0, ρi = ρ(i) ≤ c0ρ  (i ± 1), i  ∈ , for some positive constants M0 and c0.

Consider the weighted Hilbert space of infinite sequences
()
endowed with inner product (u, v) ρ = ∑i ∈  Zρiuivi and norm for u = (ui) i ∈  Z, . For any u, , define an inner product on by (u, v) λ,ρ = ∑i ∈  Zρiλiuivi, then the norm ∥·∥λ,ρ include by (·, ·) λ,ρ is equivalent to the norm ∥·∥ρ include by (·, ·) ρ. Let
()
We make the following assumptions on λi, ηi,j, fi, and gi, for i  ∈ , j = −q, …, q.
  • (H1)

    There exist two positive constants and such that

    ()

  • (H2)

    Let η(t) = {sup⁡|ηi,j(t)| : i  ∈ , j = −qq} <   (q  ∈ ) satisfy the following:

    • (H2a)

      in t;

    • (H2b)

      there exists a continuous positive valued function Q : ++ such that

      ()
      where is a positive constant;

    • (H2c)

      there exist I0 and η0 > 0 such that

      ()

  • (H3)

    For all i  ∈ , let fi satisfy the following:

    • (H3a)

      fi(x, t) is differentiable in x and continuous in t;

    • (H3b)

      for any i  ∈ , fi(0, t) = 0,   sup⁡x,t ∈ | fi(x, t)| ≤ βi(t);

    • (H3c)

      there exist functions Γ    C(+ × , +) and such that

      ()

  • (H4)

    g(t) = (gi(t)) i ∈  G, β(t) = (βi(t)) i ∈  G, where

    ()

and , , denotes the space of all continuous bounded functions from into .
Letting
()
then the system (10) is equivalent to the following evolution equation:
()
where
()

Definition 3. The function φ : [τ, τ + T) → E  (T > 0) is called a mild solution of the following lattice differential equations:

()
if φ  ∈   C([τ, τ + T), E) and
()

Theorem 4. Assume that (P0) and (H1)–(H4) hold. Then for any fixed τ  ∈ and any initial data φ(τ) = (uτ, vτ) ∈ E, problem (19) admits a unique mild solution φ(·, τ; φτ) ∈   C([τ, +), E) with φ(τ, τ; φτ) = φτ and φ(·, τ; φτ) being continuous in φτ    E, and the mapping

()
generates a continuous process {U(t, τ)} tτ on E.

Proof. Let

()
and then is continuous in u and locally integrable in t from into . For any bounded set BE with sup⁡φ ∈  Bφ∥ ≤ r, let
()
Then for φ = (u, v), φ(j) = (u(j), u(j))    B,  j = 1, 2, by (P0), (H1)–(H4), we have
()
By the approximating method [24], it follows that problem (19) possesses a unique local mild solution φ(·, τ; φτ)    C([τ, Tmax⁡), E). Similar to the proof of Theorem 5 below concerning the existence of a uniformly bounded absorbing set, we can prove that Tmax⁡ = +.

In the following part of this section, we assume that (P0) and (H1)–(H4) and
()
hold.

Theorem 5. Assume that (P0) and (H1)–(H4) hold. There exists a uniform bounded closed absorbing ball B0 = B0(0, r0) ⊂ E of {U(t, τ)} tτ centered at 0 with radius r0 (independent of τ, t) such that, for any bounded subset BE, there exists TB ≥ 0 yielding U(t + τ, τ)BB0 for all tTB.

Proof. Let be a mild solution of (19), tτ. Since the set of continuous functions is dense in , by (H2a), there exist sequences of continuous functions , m  ∈ , such that

()

Consider the following lattice differential equations:

()
where
()
Combining the continuity of Fm(φm, t) in t with the proof of Theorem 4, (36) has a unique strong solution φm(·, τ, φτ) = (um(·, τ, φτ), vm(·, τ, φτ))    C([τ, +), E)∩C1((τ, +), E) satisfying
()
Taking the inner product (·, ·) E of (29) with φm(t), we obtain
()
By some computation, we have that
()
()
Then we have that, for tτ,
()
Applying Gronwall’s inequality to (35) on [τ, t] (tτ), we obtain
()
It then follows that, for T > 0,
()
for some κ(T, τ) > 0 independent of m and i, and then, for any t, s  ∈   [τ, τ + T), m  ∈ ,
()
implies the equicontinuity of . By the Arzela-Ascoli Theorem, there exists a convergent subsequence of , such that
()
and is continuous in t  ∈   [τ, τ + T]; moreover, by (36), for t  ∈   [τ, τ + T]. By the Lebesgue Convergence Theorem, we have
()
Thus, by replacing m with mk in (31) and letting k → +, we have
()
By the uniqueness of the mild solutions of (19), we have for t  ∈   [τ, τ + T]. By replacing m with mk and letting k in (36), we have that for, tτ,
()
where
()
By (H2b), there exists T1 > 0 such that for, sT1 and t  ∈ ,
()
Let
()
and then
()
Thus, for any bounded subset B of E and φτ    B,
()
where
()
Again, by (H2b),
()
It then follows that
()
is a uniformly bounded closed absorbing set for {U(t, τ)} tτ.

Lemma 6. Assume that (P0) and (H1)–(H4) hold. For any ε > 0, there exist and I(ε, B0) ∈ such that, for tT(ε, B0), the mild solution U(τ + t, t)φτ = φ(τ + t, τ; φτ) = (ui(τ + t, τ; uτ), vi(τ + t, τ; vτ)) i ∈  E of system (19) with φτ    B0E satisfies

()

Proof. Let ξ  ∈   C1(R+, R) be a smooth increasing function which satisfies

()
Let be a solution of (29) with φm(τ) = φτ. Let MI0 + q (I0 is in (H2c)) be a suitable large integer, and set
()
Taking the inner product (·, ·) E of (29) with zm, we have
()
And we have the following estimates:
()
where
()
Therefore,
()
Applying Gronwall’s inequality to (57) on , we have that, for φτ    B0,
()
By (44)-(45), for , we have
()
as t → +. This means that, for any ε > 0, there exists such that when tT2(ε, B0),
()
By (t), β(t) ∈   G, and (H2b), there exists I1(ε) ∈ such that, for M > I1(ε),
()
From (47),
()
By (44), we have
()
And, for any ε > 0, there exists I2(ε, B0) > I0 + q such that, for M > I2(ε, B0),
()
Let T(ε, B0) = max⁡{T1, T2(ε, B0)} and I(ε, B0) = 2max⁡{I1(ε), I2(ε, B0). It follows immediately that
()

In the following, we prove the existence of a pullback exponential attractor for {U(t, τ)} tτ by applying Theorem 2. For any t  ∈ , set
()
where B0 is the uniform absorbing set defined in Theorem 5.

Theorem 7. Assume that (P0) and (H1)–(H4) hold.

  • (a)

    There exists a continuous positive value function Lt in t such that, for every τ  ∈ ,

    ()

  • (b)

    There exist positive constants T* > 0, γ  ∈   [0, 1/2), and a 2(2N + 1)-dimensional orthogonal projection PN : EEN  (N  ∈ ), such that, for every τ  ∈ and φτ, ψτ    Y(τ),

    ()

  • (c)

    For each fixed t  ∈ , .

Proof. For any τ  ∈ , initial data φτ = (uτ, vτ), ψτ = (xτ, yτ). Let φ(t) = U(t, τ)φτ, ψ(t) = U(t, τ)ψτ, and ϕ(t) = φ(t) − ψ(t) = (ϕi(t)) i ∈; then, φ(t), ψ(t), ϕ(t) ∈   C([τ, +), E).

(a) For any tτ, let φm(t), ψm(t) be two solutions of (29) with initial data φτ, ψτ and set ϕm(t) = φm(t) − ψm(t), and then ϕm(t) ∈   C([τ, +), E)∩C1((τ, +), E) satisfies

()

Taking the inner product (·, ·) E of (69) with ϕm, we have

()
Since φτ, ψτ    Y(τ), by (36) and (66), for tτ, φm(t) = (um(t), vm(t)), ψm(t) = (xm(t), ym(t)) ∈ Y(t)⊆B0, thus , , and for tτ. By (33), we have
()
and, by (H2)-(H3),
()
Thus,
()
Applying Gronwall’s inequality to (73) on [τ, τ + t], (t > 0), we have
()
that is,
()
where
()
By replacing m with mk in (75) and letting mk, we have
()

(b) For i  ∈ , let and , where ξ is as in (52) and M > I0 + q. Taking the inner product of (69) with ωm in E, we have that, for tτ,

()
For the terms in the right hand of (78), we have
()
where
()
By the continuity of Γ(u, t) and Γ(0, t) = 0 (see (H3c)), there exists such that
()
By Lemma 6 and for , there exist , such that, for ,
()
implying that, for ,
()
and thus
()
For and ,
()
Therefore, we obtain that, for all , ,
()
Applying Gronwall’s inequality to (86) from to τ + t  (tT2 + T1), we have
()
Thus, for tT2 + T1, we have
()
where
()
where
()
Thus, for , , we have
()
Taking
()
then
()
Thus, by (91), we have
()

(c) It follows from the following estimate:

()
and the continuity of Q(·), .

As a direct consequence of Theorem 2, Theorem 4, Theorem 5, Theorem 7, and Theorem  3.1 of [15], we have the following result.

Theorem 8. Assume that (P0) and (H1)–(H4) hold. The process {U(t, τ)} tτ associated with (19) possesses a pullback exponential attractor {𝒜(t)} t ∈ with the following properties:

  • (i)

    for any t  ∈ , 𝒜(t)⊆Y(t)⊆B0;

  • (ii)

    ;

  • (iii)

    and for any bounded set BE, dh(U(t, τ)Y(τ), , − < τ + TB < t ≤ +;

  • (iv)

    lim⁡tτdh(𝒜(t), 𝒜(τ)) = 0;

where
()
and is the minimal number of closed balls of E with radius θ covering the closed unit ball BN(0,1) of EN centered at 0.

Remark 9. It should be pointed out that, for some special cases, (H2) and (27) can be removed. For example, if the coupled linear operator A(t) ≡ A in (10) is a constant operator A satisfying −A = D*D = DD*, where , for all     (constant),        m0jm0, and D* is the adjoint of D. A simple example of such an operator A is (Au) i = ui−1 + ui+1 − 2ui and (Du) i = ui+1ui. Choose a positive weight function ρ : + satisfying the properties (P0) and |ρ(i ± 1) − ρ(i)| ≤ bρ(i), for all i  ∈ for some small b ≥ 0. Two examples of such weight functions are ρ(i) = 1/(1 + κ2i2) q, q > 1/2, and ρ(i) = eκ|i|, i  ∈ , with κ > 0 being a parameter. If the number b satisfies

()
where , then conditions (H2) and (27) can be removed.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

This work is supported by the National Natural Science Foundation of China under Grant no. 11071165, 11326114 and Zhejiang Normal University (ZC304011068).

      The full text of this article hosted at iucr.org is unavailable due to technical difficulties.