Hindawi Publishing Corporation Journal of Inequalities and Applications Volume 2007, Article ID 49125, 13 pages doi:10.1155/2007/49125 Research Article Semigroup Approach to Semilinear Partial Functional Differential Equations with Infinite Delay Hassane Bouzahir Received 6 November 2006; Revised 16 January 2007; Accepted 19 January 2007 Recommended by Simeon Reich We describe a semigroup of abstract semilinear functional differential equations with infinite delay by the use of the Crandall Liggett theorem. We suppose that the linear part is not necessarily densely defined but satisfies the resolvent estimates of the Hille-Yosida theorem. We clarify the properties of the phase space ensuring equivalence between the equation under investigation and the nonlinear semigroup. Copyright © 2007 Hassane Bouzahir. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 1. Introduction Most of the existing results about functional differential equations with finite delay have been recently under verification in the case of infinite delay. Our objective in this paper is to study the solution semigroup generated by the following partial functional differential equation with infinite delay: d x(t) = AT x(t) + F xt , dt x0 = φ ∈ Ꮾ, t ≥ 0, (1.1) where AT is a nondensely defined linear operator on a Banach space (E, | · |). The phase space Ꮾ can be the space Cγ , γ being a positive real constant, of all continuous functions φ : (−∞,0] → E such that limθ→−∞ eγθ φ(θ) exists in E, endowed with the norm φγ := supθ≤0 eγθ |φ(θ)|, φ ∈ Cγ . For every t ≥ 0, the function xt ∈ Ꮾ is defined by xt (θ) = x(t + θ), for θ ∈ (−∞,0]. (1.2) 2 Journal of Inequalities and Applications We assume the following. (H1) F : Ꮾ → E is globally Lipschitz continuous; that is, there exists a positive constant L such that |F(ψ1 ) − F(ψ2 )| ≤ Lψ1 − ψ2 Ꮾ for all ψ1 ,ψ2 ∈ Ꮾ. A typical example that can be transformed into (1.1) is the following: ∂2 ∂ w(t,ξ) = a 2 w(t,ξ) + bw(t,ξ) + c ∂t ∂ξ + f w(t − τ,ξ) , −∞ G(θ)w(t + θ,ξ)dθ t ≥ 0, 0 ≤ ξ ≤ π, w(t,0) = w(t,π) = 0, w(θ,ξ) = w0 (θ,ξ), 0 (1.3) t ≥ 0, −∞ < θ ≤ 0, 0 ≤ ξ ≤ π, where a, b, c, and τ are positive constants, f : R → R is a continuous function, G is a positive integrable function on (−∞,0], and w0 : (−∞,0] × [0,π] → R is an appropriate continuous function. Effectively, in [1], an abstract treatment of (1.3) as (1.1) leads to a characterization of exponential asymptotic stability near an equilibrium of (1.3) provided that the associated linearized semigroup is exponentially stable. For many quantitative studies of any problem of type (1.1) in a concrete space of functions mapping (−∞,0] into E, one should choose a space that verifies at least the fundamental axioms first introduced in [2]. That is, (Ꮾ, · Ꮾ ) is a (semi)normed abstract linear space of functions mapping (−∞,0] into E, which satisfies the following. (A) There is a positive constant H and functions K(·),M(·) : R+ → R+ , with K continuous and M locally bounded, such that for any σ ∈ R, a > 0, if x : (−∞,σ + a] → E, xσ ∈ Ꮾ, and x(·) is continuous on [σ,σ + a], then for every t in [σ,σ + a] the following conditions hold: (i) xt ∈ Ꮾ; (ii) |x(t)| ≤ H xt Ꮾ , which is equivalent to (ii) for each ϕ ∈ Ꮾ, |ϕ(0)| ≤ H ϕᏮ ; (iii) xt Ꮾ ≤ K(t − σ)supσ ≤s≤t |x(s)| + M(t − σ)xσ Ꮾ . (A1) For the function x(·) in (A), t → xt is a Ꮾ-valued continuous function for t in [σ,σ + a]. := Ꮾ/ · Ꮾ = {ϕ : ϕ ∈ Ꮾ} is (B) The space Ꮾ or the space of equivalence classes Ꮾ complete. However, to obtain interesting qualitative results, a concrete choice should be made on a space that verifies additional properties which are essential to investigate the equation. A class of employed spaces is called uniform fading memory spaces. They verify that the function K(·) is constant, limt→+∞ M(t) = 0, and the following extra property. (C) If {φn }n≥0 is a Cauchy sequence in Ꮾ with respect to the (semi)norm and if φn converges compactly to φ on (−∞,0], then φ is in Ꮾ and φn − φᏮ → 0 as n → +∞. There are many examples of concrete spaces that verify the above properties. In [3], it was proved, for instance, that if γ > 0, the above-defined space Cγ is a uniform fading Hassane Bouzahir 3 memory space. Another example is given by Cg0 := φ ∈ C (−∞,0];E : lim φ(θ) θ →−∞ g(θ) =0 , (1.4) equipped with the norm φ(θ) φg := sup −∞<θ ≤0 g(θ) , (1.5) where g : (−∞,0] → [1,+∞) is a continuous function such that (g1): g is nonincreasing and g(0) = 1, and (g2): the function G : [0,+∞) → [0,+∞) defined by G(t) := sup−∞<θ≤−t (g(t + θ)/g(θ)) tends to 0 as t tends to ∞. In general, set for any positive continuous function g on (−∞,0], φ(θ) Cg := φ ∈ C (−∞,0];E : LCg := φ ∈ Cg : lim θ →−∞ UCg := φ ∈ Cg : φ(θ) g(θ) g(θ) φ(θ) g(θ) is bounded , exists in E , (1.6) is uniformly continuous on (−∞,0] , such that (g3): G is locally bounded for t ≥ 0, (g4): g(θ) tends to ∞ as θ tends to −∞, and (g2) are satisfied. Then LCg is a uniform fading memory space; the additional condition (g5): logg(θ) is uniformly continuous on (−∞,0], ensuring that UCg is a uniform fading memory space. Precisely, K(t) = sup−t≤θ≤0 (1/g(θ)) and M(t) = G(t). Note that for the space Cγ , as defined above, g(θ) = e−γθ . On the contrary, despite its consideration in some recent separate publications concerned with abstract stability investigations, unfortunately, the space C0 := {φ ∈ C((−∞, 0];E) : limθ→−∞ φ(θ) = 0} is not a uniform fading memory space. In [4] for instance, some restrictive results about asymptotic behavior of solutions were obtained in the linear positive case on C0 . The followed method uses evolution semigroups, extrapolation spaces, and critical spectrum on Banach lattices spaces. For the basic discussion about the general phase space Ꮾ, we refer the reader to [3, especially Chapter 1] and [5, pages 401–406]. Although many authors have avoided repetitions by working on the abstract space Ꮾ, the delicacy of some investigations restricts their work on a class of concrete spaces that verify many properties such as Cγ with γ > 0. In [1, 6–8], we have considered (1.1) with AT being nondensely defined and satisfying the Hille-Yosida condition. Precisely, since D(AT ) is not densely defined, we have addressed the problems of existence, uniqueness, regularity, existence of global attractor, existence of periodic solutions, and local stability by means of the integrated semigroups theory. In this article, we use the Crandall Liggett approach. We show the relation between a nonlinear semigroup and integral solutions to (1.1). 4 Journal of Inequalities and Applications Notice that the most general results about functional differential equations with infinite delay are obtained notably in [9–15] and in [16] also in the situation where AT depends on t. Our results extend earlier ones which require the delay to be finite and AT to have a dense domain in E. In this paper, we proceed as follows. In Section 2, we recall some basic results on existence, uniqueness, and properties of integral solutions to (1.1). Then, in Section 3, we establish properties of the solution operator in nonlinear case. Next, we rely upon the well-known Crandall and Liggett theorem in order to compute the nonlinear solution semigroup by an exponential formula. Finally, we give the link between the semigroup given by the Crandall and Liggett theorem and the integral solution to (1.1). 2. Basic results Throughout, we assume that AT satisfies the Hille-Yosida condition: (H2) there exist two constants β ≥ 1 and ω0 ∈ R with (ω0 ,+∞) ⊂ ρ(AT ) and sup{(λ − ω0 )n R(λ,AT )n : λ > ω0 , n ∈ N} ≤ β, where ρ(AT ) is the resolvent set of AT and R(λ,AT ) = (λI − AT )−1 . Definition 2.1. A function x : (−∞,a] → E, a > 0, is an integral solution of (1.1) in (−∞,a] if the following conditions hold: (i) x is continuous on [0, a]; t (ii) 0 x(s)ds ∈ D(AT ), for t ∈ [0,a]; (iii) ⎧ t t ⎪ ⎨φ(0) + A x(s)ds + F s,xs ds, T x(t) = ⎪ 0 0 ⎩ φ(t), 0 ≤ t ≤ a, (2.1) −∞ < t ≤ 0. It follows from (ii) of the above definition that for an integral solution x, one has x(t) ∈ D(AT ) for all t ≥ 0. In particular, φ(0) ∈ D(AT ). Define the part A0 of AT in D(AT ) by D A0 = x ∈ D AT : AT x ∈ D AT , A0 x = AT x, for x ∈ D A0 . (2.2) Recall (cf. [17]) that A0 generates a C0 -semigroup (T0 (t))t≥0 on D(AT ). It is known (see [1, 6]) that under (H1) and (H2), for φ ∈ Ꮾ such that φ(0) ∈ D(AT ), (1.1) admits a unique integral solution x(·,φ) given by the following formula: ⎧ t ⎪ ⎨T (t)φ(0) + lim T0 (t − s)Bλ F xs (·,φ) ds, 0 λ→+∞ 0 x(t,φ) = ⎪ ⎩ φ(t), where Bλ = λR(λ,AT ). for t ≥ 0, for t ∈ (−∞,0], (2.3) Hassane Bouzahir 5 Set ᐄ := ϕ ∈ Ꮾ : ϕ(0) ∈ D AT . (2.4) Define ᐁ(t) on ᐄ for t ≥ 0 by ᐁ(t)ϕ = xt (·,ϕ), (2.5) where x(·,φ) is the integral solution of (1.1). The point of departure for our results is [1] where we have proved that (ᐁ(t))t≥0 is a strongly continuous semigroup satisfying the following properties: (i) (ᐁ(t))t≥0 satisfies, for t ≥ 0 and θ ∈ (−∞,0], the following translation property ⎧ ⎨ ᐁ(t + θ)ϕ (0), if t + θ ≥ 0, ᐁ(t)ϕ (θ) = ⎩ ϕ(t + θ), (2.6) if t + θ ≤ 0, (ii) there exist two positive locally bounded functions m(·),n(·) : R+ → R+ such that, for all ϕ1 ,ϕ2 ∈ X, and t ≥ 0, ᐁ(t)ϕ1 − ᐁ(t)ϕ2 ≤ m(t)en(t) ϕ1 − ϕ2 . Ꮾ Ꮾ (2.7) Moreover, if F is a bounded linear operator and Ꮾ is a subspace of C((−∞,0];E) satisfying axioms (A1), (A2), (B) and the following axiom which was introduced in [13]: (D) for a sequence (ϕn )n≥0 in Ꮾ, if ϕn Ꮾ → 0, then |ϕn (s)| → 0 for each s ∈ (−∞,0], then Aᐁ : D(Aᐁ ) ⊆ ᐄ → ᐄ such that Aᐁ ϕ = ϕ for any ϕ ∈ D(Aᐁ ), where D Aᐁ = ϕ ∈ ᐄ : ϕ is continuously differentiable, ϕ(0) ∈ D AT , ϕ ∈ ᐄ, ϕ (0) = AT ϕ(0) + F(ϕ) , (2.8) is the infinitesimal generator of (ᐁ(t))t≥0 . 3. Main results Our first main result can be considered as an extension of the above result to the case where F is nonlinear. The concrete choice of Ꮾ (LCg0 , LCg , or UCg ) seems to be best adapted to obtain our results. Here, we suppose sufficient conditions on g. The proof combines the ideas of [1, 18] or [19]. Proposition 3.1. Let Ꮾ = LCg0 , Ꮾ = LCg (with (g1)), or Ꮾ = UCg (with (g5)). Then Conditions (H1) and (H2) imply that Aᐁ is the infinitesimal generator of (ᐁ(t))t≥0 . 6 Journal of Inequalities and Applications Proof. Let ϕ ∈ ᐄ be continuously differentiable such that ϕ ∈ ᐄ, ϕ(0) ∈ D AT , ϕ (0) = AT ϕ(0) + F(ϕ). (3.1) Let x(·,ϕ) : (−∞,+∞) → E be the unique integral solution of (1.1). We have to show that limt→0+ (1/t)(ᐁ(t)ϕ − ϕ) exists in ᐄ and is equal to ϕ . By definition of xt (·,ϕ) and ᐁ, ⎧ ⎨ ᐁ(t)ϕ (0), x(t,ϕ) = ⎩ ϕ(t), if t ≥ 0, (3.2) if t ∈ (−∞,0], then ⎧ 1 ⎪ ⎪ ⎪ ⎨ x(t + θ,ϕ) − ϕ(θ) (0), 1 t ᐁ(t)ϕ − ϕ (θ) = ⎪ 1 t ⎪ ⎪ ϕ(t + θ) − ϕ(θ), ⎩ t t + θ > 0, (3.3) t + θ ∈ (−∞,0]. If t + θ ≤ 0, (1/t)(ᐁ(t)ϕ − ϕ)(θ) tends to D+ ϕ(θ) as t → 0+ , where D+ ϕ(θ) is the right derivative of ϕ in θ. If t + θ > 0, we have 1 ᐁ(t)ϕ − ϕ (θ) − ϕ (θ) t = 1 T0 (t + θ)ϕ(0) + lim t λ→∞ t+θ 0 (3.4) T0 (t + θ − s)Bλ F xs ds − ϕ(θ) − ϕ (θ). Let S(t), t ≥ 0, be the integrated semigroup associated with T0 (t), t ≥ 0. We obtain from the last equality 1 ᐁ(t)ϕ − ϕ (θ) − ϕ (θ) t = 1 ϕ(0) + S(t + θ)AT ϕ(0) + lim t λ→∞ t+θ + lim λ→∞ 0 t+θ 0 T0 (t + θ − s)Bλ F xs − F(ϕ) ds (3.5) T0 (t + θ − s)Bλ F(ϕ)ds − ϕ(θ) − ϕ (θ). Since T0 (t)ϕ(0) = ϕ(0) + AT S(t)ϕ(0) and t+θ lim λ→∞ 0 T0 (t + θ − s)Bλ F(ϕ)ds = lim S(t + θ)Bλ F(ϕ) λ→∞ = lim Bλ S(t + θ)F(ϕ) λ→∞ = S(t + θ)F(ϕ), (3.6) Hassane Bouzahir 7 we deduce that 1 ᐁ(t)ϕ − ϕ (θ) − ϕ (θ) t t+θ 1 = ϕ(0) + S(t + θ)ϕ (0) + lim T0 (t + θ − s)Bλ F xs − F(ϕ) ds − ϕ(θ) t λ→∞ 0 − ϕ (θ) t+θ t+θ 1 S(t + θ)ϕ (0) − ϕ (0)ds + lim T0 (t + θ − s)Bλ F xs − F(ϕ) ds t λ→∞ 0 0 1 1 t+θ + ϕ(0) + ϕ (0) − ϕ(θ) − ϕ (θ) t t t t+θ t+θ 1 = T0 (s)ϕ (0) − ϕ (0) ds + lim T0 (t + θ − s)Bλ F xs − F(ϕ) ds t 0 λ→∞ 0 0 1 0 1 + ϕ (s)ds − ϕ (θ) + ϕ (0) − ϕ (0)ds t θ t θ t+θ 1 t+θ = T0 (s)ϕ (0) − ϕ (0) ds + lim T0 (t + θ − s)Bλ F xs − F(ϕ) ds t 0 λ→∞ 0 1 0 + ϕ (s) − ϕ (0) ds + ϕ (0) − ϕ (θ). t θ (3.7) = Hence t+θ 1 T0 (s)ϕ (0) − ϕ (0)ds ᐁ(t)ϕ − ϕ (θ) − ϕ (θ) ≤ 1 t t 0 t+θ 1 ds T (t + θ − s)B F x F(ϕ) − + lim 0 λ s t λ→∞ 0 0 1 ϕ (s) − ϕ (0)ds + ϕ (0) − ϕ (θ). + t (3.8) θ Let ε > 0 and choose α > 0 small enough such that if 0 < t < α, −∞ < θ ≤ 0, and t + θ > 0, then 1 t t+θ 0 T0 (s)ϕ (0) − ϕ (0)ds < ε , 4 t+θ ε 1 < , T (t + θ − s)B F x F(ϕ) ds lim − 0 λ s 4 t λ→∞ 0 0 1 ϕ (s) − ϕ (0)ds + ϕ (0) − ϕ (θ) < ε , t 4 θ and if 0 < t < α, −∞ < θ ≤ 0, and t + θ ≤ 0, then (3.9) 1 1 ᐁ(t)ϕ − ϕ (θ) − ϕ (θ) ≤ 1 ᐁ(t)ϕ − ϕ (θ) − ϕ (θ) t g(θ) t 1 ε < . ϕ(t + θ) − ϕ(θ) ϕ (θ) − = t 4 (3.10) 8 Journal of Inequalities and Applications Consequently, if 0 < t < α, then 1 1 ᐁ(t)ϕ − ϕ (θ) − ϕ (θ) < 1 ᐁ(t)ϕ − ϕ (θ) − ϕ (θ) < ε. t g(θ) t Hence 1 ᐁ(t)ϕ − ϕ − ϕ < ε. t (3.11) (3.12) Ꮾ This proves that limt→0+ (1/t)(ᐁ(t)ϕ − ϕ) exists and is equal to ϕ . Then, ϕ ∈ D(Aᐁ ). Conversely, let ϕ ∈ ᐄ such that lim t →0+ 1 1 ᐁ(t)ϕ − ϕ = lim+ xt (·,ϕ) − ϕ = ψ = Aᐁ ϕ t →0 t t exists in ᐄ. (3.13) We can easily see that axiom (D) is verified by Cγ , LCg , and UCg which implies that lim+ t →0 1 xt (θ,ϕ) − ϕ(θ) t exists for all θ ≤ 0 and is equal to ψ(θ). (3.14) Then, for θ ∈ (−∞,0), we have ψ(θ) = lim+ t →0 1 ϕ(t + θ) − ϕ(θ) = D+ ϕ(θ); t (3.15) that is, D+ ϕ = ψ in (−∞,0). Since ψ is continuous, D+ ϕ is also continuous in (−∞,0). Let us recall the following result. Lemma 3.2 [20]. Let ϕ be continuous and differentiable on the right on [a,b). If D+ ϕ is continuous on [a,b), then ϕ is continuously differentiable on [a,b). From the above lemma, we deduce that the function ϕ is continuously differentiable in (−∞,0) and ϕ = ψ. On the other hand, for θ = 0, one has limθ→0− ϕ (θ) exists and equals ψ(0). From this we infer that the function ϕ is continuously differentiable in (−∞,0] and ϕ = ψ ∈ ᐄ. We also deduce that t → ᐁ(t)ϕ is continuously differentiable. On the other hand, we have x(t) = ϕ(0) + AT t 0 t x(s)ds + t 0 F xs ds. (3.16) t This implies that limt→0+ AT [((1/t) 0 x(s)ds) + (1/t) 0 F(xs )ds] exists and hence t t limt→0+ AT ((1/t) 0 x(s)ds) exists. From the closedness of AT and the fact that (1/t) 0 x(s)ds t ∈ D(AT ) for t > 0, we deduce that limt→0+ (1/t) 0 x(s)ds exists in D(AT ) and is equal to ϕ(0). Consequently, ϕ(0) ∈ D(AT ) and ϕ (0) = AT ϕ(0) + F(ϕ). This completes the proof of the proposition. The next result may be considered as an extension of a similar one in [19]. Our goal is to establish the Crandall and Liggett exponential formula lim n→+∞ t I − Aᐁ n −n ϕ = ᐁ(t)ϕ, ∀ϕ ∈ ᐄγ , t ≥ 0. (3.17) Hassane Bouzahir 9 We restrict our choice to ᐄγ := {φ ∈ Cγ : φ(0) ∈ D(AT )} with γ > 0. Recall that this specified space is a uniform fading memory one. Proposition 3.3. Let Ꮾ = Cγ with γ > 0. Suppose that (H1) and (H2) are satisfied. Then, the operator Aᐁ given by Proposition 3.1 satisfies the following Crandall Liggett conditions. (a) Im(I − λAᐁ ) = ᐄγ for all λ ∈ (0,1/(L + ω0 )). (b) For all ψ1 ,ψ2 ∈ ᐄγ and λ ∈ (0,1/(L + ω0 )), I − λAᐁ −1 ψ1 − I − λAᐁ −1 ψ2 ≤ γ 1 ψ1 − ψ 2 γ . 1 − λ L + ω0 (3.18) (c) D(Aᐁ ) is dense in ᐄγ . Proof. (a) It is well known that one can suppose without loss of generality that ω0 > −L and T0 (t) ≤ eω0 t . To prove (a), it is clear from the definition of Aᐁ that (I − λAᐁ )(D(Aᐁ )) ⊆ ᐄγ for λ > 0. On the other hand, for ψ ∈ ᐄγ and λ > 0, let us solve the following equation: I − λAᐁ ϕ = ψ, ϕ ∈ D Aᐁ . (3.19) Recall that with γ > 0 and λ > 0, the function W(1/λ)ψ(0) : θ → e(1/λ)θ ψ(0), θ ≤ 0, belongs to ᐄγ . Also, the fact that Ꮿγ with γ > 0 is a uniform fading memory space implies that the 0 function Mλ ψ : θ → (1/λ) θ e(1/λ)(θ−s) ψ(s)ds, θ ≤ 0, belongs to ᐄγ (see [21, 22]). Moreover, we can see that the solution of (3.19) is 1 1 ϕ(0) (θ) + Mλ ψ (θ) = e(1/λ)θ ϕ(0) + ϕ(θ) = W λ λ 0 θ e(1/λ)(θ−s) ψ(s)ds. (3.20) Next, we suppose that 0 < λω0 < 1. From (3.19) evaluated at 0 and the definition of Aᐁ we get ϕ(0) = I − λAT −1 ψ(0) + λF(ϕ) . (3.21) Introduce the following mapping Gλψ : E → E defined by Gλψ (x) = I − λAT −1 ψ(0) + λF W 1 x + Mλ ψ λ ∀x ∈ E. (3.22) For x, y ∈ E, we have λ G (x) − Gλ (y) ≤ R 1 ,AT F W 1 x + Mλ ψ − F W 1 y + Mλ ψ λ ψ ψ λ λ λL W 1 x − W 1 y ≤ 1 − λω0 λ λ γ (3.23) λL ≤ sup eγθ e(1/λ)θ (x − y) 1 − λω0 θ≤0 ≤ λL |x − y |. 1 − λω0 10 Journal of Inequalities and Applications Next, we suppose that λ ∈ (0,1/(L + ω0 )). Then, Gλψ is a strict contraction and it has a unique fixed point x in E. Knowing that (I − λAT )−1 (E) ⊆ D(AT ), we deduce that this fixed point belongs to D(AT ). Consequently, Im(I − λAᐁ ) = ᐄγ for all λ ∈ (0,1/(L + ω0 )). (b) Let λ ∈ (0,1/(L + ω0 )) be fixed. Set λ := (I − λAᐁ )−1 which is well defined from ᐄγ to D(Aᐁ ). We prove that λ is Lipschitz continuous with Lipschitz constant less than (1 − λω0 )/(1 − λ(L + ω0 )). In fact, let λ > 0 with λ ∈ (0,1/(L + ω0 )) and λ ψ1 := ϕ1 λ ψ2 := ϕ2 for ψ1 ,ψ2 ∈ ᐄγ . Given ε > 0, by definition, there exists θ ∈ (−∞,0] such that eγθ ϕ1 (θ) − ϕ2 (θ) > ϕ1 − ϕ2 γ − ε. (3.24) Using (3.21) and (H2), we get (1/λ)θ 1 1 R ,AT ψ1 (0) − ψ2 (0) + F ϕ1 − F ϕ2 eγθ ϕ1 (θ) − ϕ2 (θ) ≤ eγθ e λ λ 0 1 + e(1/λ)(θ−s) ψ1 (s) − ψ2 (s) ds λ θ 1 ψ1 − ψ2 + λLϕ1 − ϕ2 ≤ e(1/λ)θ γ γ 1 − λω0 0 1 (1/λ)θ + e(γ−1/λ)s ds λe ψ1 − ψ 2 γ θ (1/λ)θ λLe(1/λ)θ e ϕ1 − ϕ2 + 1 − e(1/λ)θ ψ1 − ψ2 γ + ≤ γ 1 − λω0 1 − λω0 (3.25) which implies that (1/λ)θ 1 − λω0 − λLe(1/λ)θ (1/λ)θ ϕ1 − ϕ2 ≤ ε + e ψ1 − ψ 2 , + 1 − e γ γ 1 − λω0 1 − λω0 (1/λ)θ 1 − λω0 e + 1 − λω0 − e(1/λ)θ + λω0 e(1/λ)θ ψ1 − ψ 2 γ 1 − λω0 − λL 1 − λω0 1 ψ1 − ψ 2 γ . ≤ 1 − λ ω0 + L ϕ1 − ϕ2 ≤ γ (3.26) To prove (c), using similar arguments as in [23, the proof of Proposition 3.5], we can verify that for all λ > 0 with λ ∈ (0,1/(L + ω0 )) and ψ ∈ ᐄγ , ψ − λ ψ ≤ γ λ λL F(0) ψ γ + 1 − λω 1 − λω −1 + ψ − ψ(0) − Mλ ψ − ψ(0) γ + I − λAT ψ(0) − ψ(0). (3.27) Since by its definition λ ψ belongs to D(Aᐁ ), assertion (c) follows from the fact that limλ→0+ |(I − λAT )−1 x − x| = 0 for all x ∈ D(AT ) (see [24]) and the following result. Lemma 3.4. Set ᐄγ0 := {φ ∈ ᐄγ such that φ(0) = 0}. Then for all λ > 0 and ψ ∈ ᐄγ0 the function Mλ ψ tends to ψ in ᐄγ0 as λ tends to zero. Hassane Bouzahir 11 Proof. Set ᐄγ0 := {φ ∈ ᐄγ such that φ(0) = 0}. We can see that the operator B0 : D(B0 ) ⊆ ᐄγ0 → ᐄγ0 such that B0 φ = φ for any φ ∈ D(B0 ) where D B0 = φ ∈ ᐄγ0 : φ is continuously differentiable and φ ∈ ᐄγ0 (3.28) is the infinitesimal generator of a C0 semigroup on ᐄγ0 . Moreover, Mλ ψ = I − λB0 −1 for any λ > 0, ψ ∈ ᐄγ0 . ψ (3.29) This implies that the assertion of the lemma is a direct consequence of a basic result that can be found, for instance, in [24, page 248]. Corollary 3.5. Let ᐄ = ᐄγ with γ > 0. Suppose that (H1) and (H2) are satisfied. Then, for all ϕ ∈ ᐄγ and t ≥ 0, lim n→+∞ t I − Aᐁ n −n ϕ = ᐁ(t)ϕ. (3.30) The proof of the above result is based on the following special case of the well-known Crandall and Liggett theorem (see [25]). Theorem 3.6 [25]. Let (Y , · ) be a Banach space and B a nonlinear operator with dense domain D(B) in Y . Suppose that there exists a positive real constant ω such that (a) Im(I − λB) = Y for all λ ∈ (0,1/ω), (b) for all y1 , y2 ∈ Y and λ ∈ (0,1/ω), (I − λB)−1 y1 − (I − λB)−1 y2 ≤ 1 y 1 − y 2 . 1 − λω (3.31) Then for all y ∈ Y , the limit W0 (t)y := lim n→+∞ t I− B n −n y (3.32) exists in Y . Moreover, the family of operators (W0 (t))t≥0 satisfies (i) W0 (0) = I, (ii) W0 (t1 + t2 ) = W0 (t1 )W0 (t2 ) for all t1 ,t2 ≥ 0, (iii) W0 (t)y1 − W0 (t)y2 ≤ eωt y1 − y2 for all y1 , y2 ∈ Y and t ≥ 0. Let us now give the link between the semigroup given by the Crandall and Liggett theorem and the integral solution to (1.1). The proof will be omitted because it is very similar to the case of finite delay (see [18] or [19]). Proposition 3.7. Let ᐄ = ᐄγ with γ > 0. Suppose that (H1) and (H2) are satisfied and the operator A : D(A) ⊆ ᐄγ → ᐄγ such that Aϕ = ϕ for any ϕ ∈ D(A), where D(A) = ϕ ∈ ᐄγ : ϕ is continuously differentiable, ϕ(0) ∈ D AT , ϕ ∈ ᐄγ , ϕ (0) = AT ϕ(0) + F(ϕ) , (3.33) 12 Journal of Inequalities and Applications satisfies the hypotheses of Crandall and Liggett theorem in ᐄγ . Let (U(t))t≥0 be the nonlinear semigroup given by U(t)ϕ = lim n→+∞ −n t I− A n ϕ ∀ϕ ∈ ᐄγ , t ≥ 0. (3.34) Then for all ϕ ∈ ᐄγ , the function y := y(·,ϕ) : (−∞,+∞) → R defined by ⎧ ⎨ U(t)ϕ (0), y(t,ϕ) = ⎩ ϕ(t), t ≥ 0, t ≤ 0, (3.35) is an integral solution of (1.1). Acknowledgments The author would like to thank Professors M. Adimy and K. Ezzinbi for helpful discussions. This research was supported by TWAS under contract no. 04-150 RG/MATHS/ AF/AC. References [1] M. Adimy, H. Bouzahir, and K. Ezzinbi, “Local existence and stability for some partial functional differential equations with infinite delay,” Nonlinear Analysis. Theory, Methods & Applications. Series A, vol. 48, no. 3, pp. 323–348, 2002. [2] J. K. Hale and J. Kato, “Phase space for retarded equations with infinite delay,” Funkcialaj Ekvacioj. Serio Internacia, vol. 21, no. 1, pp. 11–41, 1978. [3] Y. Hino, S. Murakami, and T. Naito, Functional-Differential Equations with Infinite Delay, vol. 1473 of Lecture Notes in Mathematics, Springer, Berlin, Germany, 1991. [4] S. Brendle and R. Nagel, “PFDE with nonautonomous past,” Discrete and Continuous Dynamical Systems. Series A, vol. 8, no. 4, pp. 953–966, 2002. [5] J. K. Hale and S. M. Verduyn Lunel, Introduction to Functional-Differential Equations, vol. 99 of Applied Mathematical Sciences, Springer, New York, NY, USA, 1993. [6] M. Adimy, H. Bouzahir, and K. Ezzinbi, “Existence for a class of partial functional differential equations with infinite delay,” Nonlinear Analysis. Theory, Methods & Applications. Series A, vol. 46, no. 1, pp. 91–112, 2001. [7] R. Benkhalti, H. Bouzahir, and K. Ezzinbi, “Existence of a periodic solution for some partial functional-differential equations with infinite delay,” Journal of Mathematical Analysis and Applications, vol. 256, no. 1, pp. 257–280, 2001. [8] H. Bouzahir and K. Ezzinbi, “Global attractor for a class of partial functional differential equations with infinite delay,” in Topics in Functional Differential and Difference Equations (Lisbon, 1999), T. Faria and P. Freitas, Eds., vol. 29 of Fields Inst. Commun., pp. 63–71, American Mathematical Society, Providence, RI, USA, 2001. [9] W. M. Ruess, “Linearized stability for nonlinear evolution equations,” Journal of Evolution Equations, vol. 3, no. 2, pp. 361–373, 2003. [10] W. M. Ruess, “Existence of solutions to partial functional evolution equations with delay,” in Functional Analysis (Trier, 1994), S. Dierolf, S. Dineen, and P. Domański, Eds., pp. 377–387, de Gruyter, Berlin, Germany, 1996. [11] W. M. Ruess, “Existence and stability of solutions to partial functional-differential equations with delay,” Advances in Differential Equations, vol. 4, no. 6, pp. 843–876, 1999. Hassane Bouzahir 13 [12] W. M. Ruess, “Existence of solutions to partial functional-differential equations with delay,” in Theory and Applications of Nonlinear Operators of Accretive and Monotone Type, A. G. Kartsatos, Ed., vol. 178 of Lecture Notes in Pure and Appl. Math., pp. 259–288, Dekker, New York, NY, USA, 1996. [13] W. M. Ruess and W. H. Summers, “Linearized stability for abstract differential equations with delay,” Journal of Mathematical Analysis and Applications, vol. 198, no. 2, pp. 310–336, 1996. [14] W. M. Ruess and W. H. Summers, “Operator semigroups for functional-differential equations with delay,” Transactions of the American Mathematical Society, vol. 341, no. 2, pp. 695–719, 1994. [15] W. M. Ruess and W. H. Summers, “Almost periodicity and stability for solutions to functionaldifferential equations with infinite delay,” Differential and Integral Equations, vol. 9, no. 6, pp. 1225–1252, 1996. [16] A. G. Kartsatos and M. E. Parrott, “The weak solution of a functional-differential equation in a general Banach space,” Journal of Differential Equations, vol. 75, no. 2, pp. 290–302, 1988. [17] H. R. Thieme, “Semiflows generated by Lipschitz perturbations of non-densely defined operators,” Differential and Integral Equations, vol. 3, no. 6, pp. 1035–1066, 1990. [18] M. Adimy, M. Laklach, and K. Ezzinbi, “Non-linear semigroup of a class of abstract semilinear functional differential equations with a non-dense domain,” Acta Mathematica Sinica (English Series), vol. 20, no. 5, pp. 933–942, 2004. [19] M. Laklach, “Contribution à l’étude des équations aux dérivées partielles à retard et de type neutre,” Thèse de doctorat, l’Université de Pau et des Pays de l’Adour, Pau, France, 2001. [20] A. Pazy, Semigroups of Linear Operators and Applications to Partial Differential Equations, vol. 44 of Applied Mathematical Sciences, Springer, New York, NY, USA, 1983. [21] T. Naito, N. V. Minh, and J. S. Shin, “Spectrum and (almost) periodic solutions of functional differential equations,” Vietnam Journal of Mathematics, vol. 30, supplement, pp. 577–589, 2002. [22] T. Naito, N. V. Minh, and J. S. Shin, “The spectrum of evolution equations with infinite delay,” unpublished. [23] C. C. Travis and G. F. Webb, “Existence and stability for partial functional differential equations,” Transactions of the American Mathematical Society, vol. 200, pp. 395–418, 1974. [24] K. Yosida, Functional Analysis, vol. 123 of Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], Springer, Berlin, Germany, 6th edition, 1980. [25] M. G. Crandall and T. M. Liggett, “Generation of semi-groups of nonlinear transformations on general Banach spaces,” American Journal of Mathematics, vol. 93, pp. 265–298, 1971. Hassane Bouzahir: National Engineering School of Applied Sciences (ENSA), University Ibn Zohr, P.O. Box 1136, Agadir 80000, Morocco Email address: bouzahir@ensa-agadir.ac.ma