Almost Automorphic Mild Solutions to Neutral Parabolic Nonautonomous Evolution Equations with Nondense Domain
Abstract
Combining the exponential dichotomy of evolution family, composition theorems for almost automorphic functions with Banach fixed point theorem, we establish new existence and uniqueness theorems for almost automorphic mild solutions to neutral parabolic nonautonomous evolution equations with nondense domain. A unified framework is set up to investigate the existence and uniqueness of almost automorphic mild solutions to some classes of parabolic partial differential equations and neutral functional differential equations.
1. Introduction
Bochner has shown in the seminal work [1] that in certain situations it is possible to establish the almost periodicity of an object by first establishing its almost automorphy and then invoking auxiliary conditions which, when coupled with almost automorphy, give almost periodicity. From then on, automorphy has been widely investigated. Fundamental properties of almost automorphic functions on groups and abstract almost automorphic minimal flows were studied by Veech [2, 3] and others. Afterwards, Zaki [4] extended the notion of scalar-valued almost automorphy to the one of vector-valued almost automorphic functions, paving the road to many applications to differential equations and dynamical systems. Among other things, Shen and Yi [5] showed that almost automorphy is essential and fundamental in the qualitative study of almost periodic differential equations in the sense that almost automorphic solutions are the right class for almost periodic systems. We refer the readers to the monographs [6, 7] by N’Guérékata for more information on this topic.
A rich source of the literature exists on almost automorphic mild solutions to linear and semilinear evolution equations. However, to the best of our knowledge, there are few results available on the existence and uniqueness of almost automorphic mild solutions to neutral parabolic nonautonomous evolution equations (1) and (2), especially in the case of not necessarily dense domain and bounded perturbations. Nondensity occurs in many situations, from restrictions made on the space where the equation is considered or from boundary conditions. For example, the space C2[0, π] of twice continuously differential functions with null value on the boundary is nondense in C[0, π], the space of continuous functions. One can refer for this to [17–19] or Section 5 for more details. We further remark that our first main result (Theorem 18) recovers partly Theorem 2.2 in [15] and Theorem 3.2 in [16] in the parabolic case. Moreover, a unified framework is set up in the second main result (Theorem 21) to study the existence and uniqueness of almost automorphic mild solutions to some classes of parabolic partial differential equations and neutral functional differential equations. As one will see, the additional neutral term f(t, u(t)) greatly widens the applications of the main result since (2) is general enough to incorporate some classes of parabolic partial differential equations and neutral functional differential equations as special cases.
As a preparation, in Section 2 we fix our notation and collect some basic facts on evolution family and almost automorphy. Section 3 deals with the proof of the existence and uniqueness theorem of almost automorphic mild solutions to evolution equation (1). In Section 4, we study the existence and uniqueness of almost automorphic mild solutions to evolution equation (2) with bounded perturbations. Finally, the abstract results are applied to some classes of parabolic partial differential equations and neutral functional differential equations.
2. Preliminaries
Throughout this paper, ℕ, ℤ, ℝ, and ℂ stand for the sets of positive integer, integer, real, and complex numbers, and (𝕏, ∥·∥) stands for a Banach space. If (𝕐, ∥·∥𝕐) is another Banach space, the space B(𝕏, 𝕐) denotes the Banach space of all bounded linear operators from 𝕏 into 𝕐 equipped with the uniform operator topology. The resolvent operator R(λ, A) is defined by R(λ, A): = (λ − A) −1 for λ ∈ ρ(A), the resolvent set of a linear operator A.
2.1. Evolution Family and Exponential Dichotomy
Definition 1 (see [20], [21].)A family of bounded linear operators {U(t, s)} t≥s on a Banach space 𝕏 is called an evolution family if
- (1)
U(t, r)U(r, s) = U(t, s) and U(s, s) = I for all t ≥ r ≥ s and t, r, s ∈ ℝ;
- (2)
the map (t, s) ↦ U(t, s)x is continuous for all x ∈ 𝕏, t > s, and t, s ∈ ℝ.
Definition 2 (see [20], [21].)An evolution family {U(t, s)} t≥s on a Banach space 𝕏 has an exponential dichotomy (or is called hyperbolic) if there exist projections P(t), t ∈ ℝ, uniformly bounded and strongly continuous in t and constants M > 0, δ > 0 such that
- (1)
U(t, s)P(s) = P(t)U(t, s) for t ≥ s and t, s ∈ ℝ;
- (2)
the restriction UQ(t, s) : Q(s)𝕏 → Q(t)𝕏 of U(t, s) is invertible for t ≥ s (and we set UQ(s, t): = UQ(t, s) −1);
- (3)
for t ≥ s and t, s ∈ ℝ.
Remark 3. Exponential dichotomy is a classical concept in the study of the long-term behavior of evolution equations, combining forward exponential stability on some subspaces with backward exponential stability on their complements. Its importance relies in particular on the robustness; that is, exponential dichotomy persists under small linear or nonlinear perturbations (see, e.g., [20–24]).
2.2. Almost Automorphy and Bi-Almost Automorphy
Let C(ℝ, 𝕏) denote the collection of continuous functions from ℝ into 𝕏. Let BC(ℝ, 𝕏) denote the Banach space of all bounded continuous functions from ℝ into 𝕏 equipped with the sup norm ∥u∥∞ : = sup t∈ℝ∥u(t)∥. Similarly, C(ℝ × 𝕏, 𝕐) denotes the collection of all jointly continuous functions from ℝ × 𝕏 into 𝕐, and BC(ℝ × 𝕏, 𝕐) denotes the collection of all bounded and jointly continuous functions f : ℝ × 𝕏 → 𝕐.
Definition 5 (Bochner). A function f ∈ C(ℝ, 𝕏) is said to be almost automorphic if for any sequence of real numbers , there exists a subsequence {sn} n∈ℕ such that
Example 6 (Levitan). The function f(t) = sin(1/(2 + cos t + cos πt)), t ∈ ℝ, is almost automorphic but not almost periodic.
Remark 7. An almost automorphic function may not be uniformly continuous, while an almost periodic function must be uniformly continuous.
Lemma 8 (see [6], [7].)Assume that f, g : ℝ → 𝕏 are almost automorphic and λ is any scalar. Then the following holds true:
- (1)
f + g, λf are almost automorphic;
- (2)
the range Rf of f is precompact, so f is bounded;
- (3)
fτ defined by fτ(t) = f(t + τ), τ ∈ ℝ, is almost automorphic.
Lemma 9 (see [6], [7].)If {fn} is a sequence of almost automorphic functions and fn → f (n → ∞) uniformly on ℝ, then f is almost automorphic.
Lemma 10 (see [6].)The space AA(𝕏) equipped with sup norm ∥u∥∞ = sup t∈ℝ∥u(t)∥ is a Banach space.
Definition 11 (see [25].)A function f ∈ C(ℝ × 𝕏, 𝕏) is said to be almost automorphic if f is almost automorphic in t ∈ ℝ for each u ∈ 𝕏. That is to say, for every sequence of real numbers , there exists a subsequence {sn} n∈ℕ such that
Lemma 12 (see [6], Theorem 2.2.6.)Assume that f ∈ AA(ℝ × 𝕏, 𝕏) and there exists a constant Lf > 0 such that for all t ∈ ℝ and u, v ∈ 𝕏,
Corollary 13 (see [6], Corollary 2.1.6.)Assume that u ∈ AA(𝕏) and B ∈ B(𝕏). If for each t ∈ ℝ, v(t) = Bu(t), then v ∈ AA(𝕏).
Definition 14 (see [26].)A function f ∈ C(ℝ × ℝ, 𝕏) is called bi-almost automorphic if for every sequence of real numbers , one can extract a subsequence {sn} n∈ℕ such that
In other words, a function f ∈ C(ℝ × ℝ, 𝕏) is said to be bi-almost automorphic if for any sequence of real numbers , there exists a subsequence {sn} n∈ℕ such that
3. Neutral Parabolic Nonautonomous Evolution Equation
- (H1)
there exist constants λ0 ≥ 0, θ ∈ (π/2, π), L0, K0 ≥ 0, and α, β ∈ (0,1] with α + β > 1 such that
(14) -
for t, s ∈ ℝ, λ ∈ Σθ : = {λ ∈ ℂ∖{0} : |argλ| ≤ θ},
- (H2)
the evolution family {U(t, s)} t≥s generated by A(t) has an exponential dichotomy with dichotomy constants M > 0, δ > 0, dichotomy projections P(t), t ∈ ℝ, and Green’s function Γ(t, s),
- (H3)
Γ(t, s)x ∈ bAA(ℝ × ℝ, 𝕏) for each x ∈ 𝕏,
- (H4)
f ∈ AA(ℝ × 𝕏, 𝕏), and there exists a constant Lf > 0 such that for all t ∈ ℝ and u, v ∈ 𝕏,
(15) - (H5)
g ∈ AA(ℝ × 𝕏, 𝕏), and there exists a constant Lg > 0 such that for all t ∈ ℝ and u, v ∈ 𝕏,
(16)
Remark 15. Assumption (H1) is usually called “Acquistapace-Terreni” conditions, which was first introduced in [27] for λ0 = 0. If (H1) holds, then there exists a unique evolution family {U(t, s)} t≥s on 𝕏 such that (t, s) ↦ U(t, s) ∈ B(𝕏) is strongly continuous for t > s, U(·, s) ∈ C1((s, ∞), B(𝕏)), ∂tU(t, s) = A(t)U(t, s) for t > s. These assertions are established in Theorem 2.3 of [28]. See also [27, 29, 30].
Definition 16. A mild solution to (1) is a continuous function u : ℝ → 𝕏 satisfying integral equation
Lemma 17. Assume that (H1)–(H3) and (H5) hold. Define nonlinear operator Λ on AA(𝕏) by
Proof. Combining the ideas from Theorem 4.28 in [22], the technique of exponential dichotomy, composition theorem of almost automorphic functions, and the Lebesgue dominated convergence theorem, we strive for a more self-contained proof. Let u ∈ AA(𝕏). Then it follows from Lemma 12 [6, Theorem 2.2.6] that h : = g(·, u(·)) ∈ AA(𝕏), in view of (H5). Hence, (Λu)(t) can be rewritten as
To show Λu ∈ AA(𝕏), let us take a sequence of real numbers and show that there exists a subsequence {sn} n∈ℕ such that
Now we are in a position to state and prove the first main result of this paper.
Theorem 18. Suppose that (H1)–(H5) hold. If Θ = Lf + (2MLg/δ) < 1, then there exists a unique mild solution u ∈ AA(𝕏) to (1) such that
Proof. Firstly, define nonlinear operator Γ on AA(𝕏) by
Secondly, we will prove that Γ is a strict contraction on AA(𝕏). Let v, w ∈ AA(𝕏). By (H2), (H4), and (H5), we have
Finally, to prove that u satisfies (17) for all t ≥ s, all s ∈ ℝ. For this, we let
4. Bounded Perturbations
- (H6)
B, C ∈ B(𝕏) with max {∥B∥B(𝕏), ∥C∥B(𝕏)} = K.
Definition 19. A mild solution to (2) is a continuous function u : ℝ → 𝕏 satisfying integral equation
Lemma 20. Let assumptions (H1)–(H3), (H5), and (H6) hold. Define nonlinear operator Λ1 on AA(𝕏) by
Proof. Let u(·) ∈ AA(𝕏). By (H6) and Corollary 13, we obtain Cu(·) ∈ AA(𝕏). Then it follows from Lemma 12 [6, Theorem 2.2.6] that h1 : = g(·, Cu(·)) ∈ AA(𝕏), in view of (H5). The left is almost same as the proof of Lemma 17, remembering to replace Λ, h by Λ1, and h1, respectively. This ends the proof.
Now we are in a position to state and prove the second main result of this paper.
Theorem 21. Suppose that (H1)–(H6) hold. If Θ1 = K(Lf + (2MLg/δ)) < 1, then there exists a unique mild solution u ∈ AA(𝕏) to (2) such that
Proof. Firstly, define nonlinear operator Γ1 on AA(𝕏) by
Secondly, we will prove that Γ1 is a strict contraction on AA(𝕏) and apply Banach fixed point theorem. Let v, w ∈ AA(𝕏). Then it follows from (H2), (H4), (H5), and (H6) that
Finally, to prove that u satisfies (34) for all t ≥ s, all s ∈ ℝ. For this, we let
5. Applications to Parabolic Partial Differential Equations and Neutral Functional Differential Equations
In this section, two examples are given to illustrate the effectiveness and flexibility of Theorem 21. By a mild solution to a partial or neutral functional differential equation, we mean a mild solution to the corresponding evolution equation.
Example 22. Consider the following parabolic partial differential equation:
Let 𝕏 : = C[0, π] denote the space of continuous functions from [0, π] to ℝ equipped with the sup norm and define the operator A by
Define a family of linear operators A(t), t ∈ ℝ by
- (ST)
There exist constants θ ∈ (π/2, π), L0, K0 ≥ 0, and α ∈ (0,1] such that
(46) -
for t, s, r ∈ ℝ, λ ∈ Σθ : = {λ ∈ ℂ∖{0}:|argλ | ≤ θ}.
As for (H3), it is obvious that U(t, s)φ ∈ C(ℝ × ℝ, C[0, π]) for each φ ∈ C[0, π].
To show that U(t, s)φ ∈ bAA(ℝ × ℝ, C[0, π]) for each φ ∈ C[0, π], let us take a sequence of real numbers and show that there exists a subsequence {sn} n∈ℕ such that
By −3 + sinτ + sinπτ ∈ AA(ℝ), there exists a subsequence {sn} n∈ℕ such that pointwise for each τ ∈ ℝ,
Define the operators B, C by
In view of the above, (42) can be transformed into the abstract form (2), and assumptions (H1)–(H3) and (H6) are satisfied.
We add the following assumptions:
- (H4a)
f : ℝ × C[0, π] → C[0, π], (t, u) ↦ f(t, u) is almost automorphic, and there exists a constant Lf > 0 such that for all t ∈ ℝ, u(t, ·), v(t, ·) ∈ C[0, π],
(53) - (H5a)
g : ℝ × C[0, π] → C[0, π], (t, u) ↦ g(t, u) is almost automorphic, and there exists a constant Lg > 0 such that for all t ∈ ℝ, u(t, ·), v(t, ·) ∈ C[0, π],
(54)
Now, the following proposition is an immediate consequence of Theorem 21.
Proposition 23. Under assumptions (H4a) and (H5a), parabolic partial differential equation (42) admits a unique almost automorphic mild solution if
Example 24. Consider neutral functional differential equation
Take 𝕏, A, A(t), t ∈ ℝ, (H4a), and (H5a) as in Example 22. Define the operators B, C by
Now, (57) can be transformed into the abstract form (2) and assumptions (H1)–(H3) and (H6) are satisfied. Hence, Theorem 21 leads also to the following proposition.
Proposition 25. Under assumptions (H4a) and (H5a), neutral functional differential equation (57) admits a unique almost automorphic mild solution if
Acknowledgments
The authors would like to thank the referees for their careful reading of this paper. This work is supported by the National Science Foundation of China (11171314).