Electronic Journal of Differential Equations, Vol. 2013 (2013), No. 08, pp. 1–10. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu EXISTENCE AND UNIQUENESS OF ANTI-PERIODIC SOLUTIONS FOR NONLINEAR THIRD-ORDER DIFFERENTIAL INCLUSIONS TOUMA HADDAD, TAHAR HADDAD Abstract. In this article, we study the existence of anti-periodic solutions for the third-order differential inclusion u000 (t) ∈ ∂ϕ(u0 (t)) + F (t, u(t)) u(0) = −u(T ), u0 (0) = −u0 (T ), a.e. on [0, T ] u00 (0) = −u00 (T ), where ϕ is a proper convex, lower semicontinuous and even function, and F is an upper semicontinuous convex compact set-valued mapping. Also uniqueness of anti-periodic solution is studied. 1. Introduction Existence and uniqueness of anti-periodic solutions for differential inclusions generated by the subdifferential of a convex lower semicontinuous even function appear in several articles; see [2, 3, 4, 5, 6, 11, 12]. Okochi [13] initiated the study of antiperiodic solutions of the differential inclusion f (t) ∈ u0 (t) + ∂ϕ(u(t)) a.e. t ∈ [0, T ] u(0) = −u(T ) (1.1) in Hilbert spaces, where ∂ϕ is the subdifferential of an even function ϕ on a real Hilbert space H and f ∈ L2 ([0, T ], H). It was shown in [14], by applying a fixed point theorem for nonexpansive mapping, that (1.1) has a unique solution. Later Aftabizadeh and al [1] studied the anti-periodic solution of third-order differential inclusion u000 (t) ∈ ∂ϕ(u0 (t)) + f (t) a.e. t ∈ [0, T ] (1.2) u(0) = −u(T ), u0 (0) = −u0 (T ), u00 (0) = −u00 (T ), by using maximal monotone operator theory. The aim of this article is to study the existence of anti-periodic solutions for the third-order differential inclusion u000 (t) ∈ ∂ϕ(u0 (t)) + F (t, u(t)) a.e. t ∈ [0, T ] (1.3) u(0) = −u(T ), u0 (0) = −u0 (T ), u00 (0) = −u00 (T ), 2000 Mathematics Subject Classification. 34C25, 34G20, 49J52. Key words and phrases. Anti-periodic solution; differential inclusions; subdifferential. c 2013 Texas State University - San Marcos. Submitted September 28, 2012. Published January 9, 2013. 1 2 T. HADDAD, T. HADDAD EJDE-2013/08 where ϕ : H →] − ∞, +∞] is a convex lower semicontinuous even function and F : [0, T ]×H → 2H is an upper semicontinuous convex compact set-valued mapping bounded above by L2 function. Furthermore, an existence and uniqueness result when F is single-valued is also studied. 2. Preliminaries Let H be a real Hilbert space with norm k · k and inner product h·, ·i. The open ball centered at x with radius r is defined by Br (x) = {y ∈ H : ky − xk < r}, where Br (x) denotes its closure. For a proper lower semicontinuous convex function ϕ : H →] − ∞, +∞], the set-valued mapping ∂ϕ : H → 2H defined by ∂ϕ(x) = {ξ ∈ H : ϕ(y) − ϕ(x) ≥ hξ, y − xi, ∀y ∈ H} which is the subdifferential of ϕ. Let us recall a classical closure type lemma from [7]. Lemma 2.1. Let H be a separable Hilbert space. Let ϕ be a convex lower semicontinuous function defined on H with values in ] − ∞, +∞]. Let (un )n∈N∪{∞} be a sequence of measurable mappings from [0, T ] into H such that un → u∞ pointwise with respect to the norm topology. Assume that (ξn )n∈N is a sequence in L2 ([0, T ], H) satisfying ξn (t) ∈ ∂ϕ(un (t)) a.e. t ∈ [0, T ] for each n ∈ N and converging weakly to ξ∞ ∈ L2 ([0, T ], H). Then we have ξ∞ (t) ∈ ∂ϕ(u∞ (t)) a.e. t ∈ [0, T ]. Let us recall a useful result. 1,2 Lemma 2.2 ([15]). Let H be a real Hilbert space. Let u ∈ Wloc (R, H) be 2T R 2T periodic and satisfying 0 u(t)dt = 0, then kukL2 ([0,2T ],H) ≤ T 0 ku kL2 ([0,2T ],H) . π 3. Main results We state and summarize some useful results for anti-periodic mappings that are crucial for our purpose. Proposition 3.1. Let H be a real Hilbert space. Let u ∈ W 3,2 ([0, T ], H) satisfying u(0) = −u(T ), u0 (0) = −u0 (T ), u00 (0) = −u00 (T ), then the following inequalities hold √ (A1) kukC([0,T ],H) ≤ 2T ku0 kL2 ([0,T ],H) ; (A2) kukL2 ([0,T ],H) ≤ Tπ ku0 kL2 ([0,T ],H) ; √ (B1) ku0 kC([0,T ],H) ≤ 2T ku00 kL2 ([0,T ],H) ; (B2) ku0 kL2 ([0,T ],H) ≤ Tπ ku00 kL2 ([0,T ],H) ; (C1) ku00 kC([0,T ],H) ≤ √ ku000 kL2 ([0,T ],H) . Rt RT Proof. (A1) Since u(t) = u(0) + 0 u0 (s)ds and u(t) = u(T ) − t u0 (s)ds, for all t ∈ [0, T ], by adding these equalities, by anti-periodicity, we obtain Z t Z T 0 2u(t) = u (s)ds − u0 (s)ds, ∀t ∈ [0, T ]. T 2 0 t EJDE-2013/08 EXISTENCE AND UNIQUENESS OF ANTI-PERIODIC SOLUTIONS 3 Hence we have t Z ku0 (s)kds + 2ku(t)k ≤ T Z 0 ku0 (s)kds = T Z t ku0 (s)kds, ∀t ∈ [0, T ], 0 and so, by Holder inequality √ T 0 ku kL2 ([0,T ],H) . 2 kukC([0,T ],H) = sup ku(t)k ≤ t∈[0,T ] (A2) For the sake of simplicity, we use the same notation u(t), t ∈ R, to denote the anti-periodic extension of u(t), t ∈ [0, T ], such that u(t + T ) = −u(t) and u0 (t + T ) = −u0 (t), for t ∈ R. Then u is 2T -periodic function since u(t + 2T ) = u(t + T + T ) = −u(t + T ) = u(t). Also, since Z 2T T Z u(t)dt = 0 Z 2T u(t)dt + Z 0 T T Z T u(t)dt − u(t)dt = 0 u(t)dt = 0, 0 so, by Lemma 2.2, Z T 2 2T 0 ku (t)k2 dt. π2 0 0 Let us observe that ku(t)k2 and ku0 (t)k2 are T-periodic because ku(t + T )k2 = k − u(t)k2 = ku(t)k2 and similarly ku0 (t + T )k2 = k − u0 (t)k2 = ku0 (t)k2 . Hence we deduce that Z 2T Z T Z 2T Z T 2 2 0 2 ku(t)k dt = 2 ku(t)k dt and ku (t)k dt = 2 ku0 (t)k2 dt. Z 0 2T ku(t)k2 dt ≤ 0 0 0 Finally we get the required inequality Z T Z T2 T 0 2 ku(t)k dt ≤ 2 ku (t)k2 dt. π 0 0 Similarly, we prove (B1), (B2) and (C1), using u0 (0) = −u0 (T ), and u00 (0) = −u00 (T ) respectively. The following result deal with convex compact valued perturbations of a thirdorder differential inclusion governed by subdifferential operators of convex lower semicontinuous functions with anti-periodic boundary conditions. First, to simplify we will assume that H = Rd . Theorem 3.2. Let H = Rd , ϕ : H →] − ∞, +∞] be a proper, convex, lower semicontinuous and even function. Let F : [0, T ] × H → 2H be a convex compact set-valued mapping, measurable on [0, T ] and upper semicontinuous on H satisfying: there is a function α(·) ∈ L2 ([0, T ], R+ ) such that F (t, x) ⊂ Γ(t) := Bα(t) (0) for all (t, x) ∈ [0, T ] × H. Then the problem u000 (t) ∈ ∂ϕ(u0 (t)) + F (t, u(t)) u(0) = −u(T ), 0 0 u (0) = −u (T ), a.e. t ∈ [0, T ], u00 (0) = −u00 (T ), has at least an anti-periodic W 3,2 ([0, T ], H) solution. 4 T. HADDAD, T. HADDAD EJDE-2013/08 Proof. Recall that a W 3,2 ([0, T ], H) function u : [0, T ] → H is a solution of the problem under consideration if there exists a function h ∈ L2 ([0, T ]; H) such that u000 (t) ∈ ∂ϕ(u0 (t)) + h(t) h(t) ∈ F (t, u(t)) u(0) = −u(T ), 0 a.e. t ∈ [0, T ], a.e. t ∈ [0, T ], 0 u00 (0) = −u00 (T ). u (0) = −u (T ), Let us denote by SΓ2 the set of all L2 ([0, T ]; H)-selection of Γ SΓ2 := {f ∈ L2 ([0, T ], H) : f (t) ∈ Γ(t) a.e. t ∈ [0, T ]}. By [2, Theorem 2.1], for all f ∈ SΓ2 , there is a unique W 3,2 ([0, T ], H) solution uf of 0 u000 f (t) ∈ ∂ϕ(uf (t)) + f (t) uf (0) = −uf (T ), u0f (0) = a.e. t ∈ [0, T ], −u0f (T ), u00f (0) = −u00f (T ), such that ku000 f kL2 ([0,T ],H) ≤ kf kL2 ([0,T ],H) . For each f ∈ SΓ2 , (3.1) let us define the set-valued mapping Ψ(f ) := {g ∈ L2 ([0, T ], H); g(t) ∈ F (t, uf (t)) a.e. t ∈ [0, T ]}. Then it is clear that Ψ(f ) is a nonempty convex weakly compact subset of SΓ2 , here the nonemptiness follows from [10, theorem VI.4]. From the above consideration, 2 we need to prove that the convex weakly compact set-valued mapping Ψ : SΓ2 → 2SΓ admits a fixed point. By Kakutani-Ky Fan fixed point theorem, it is sufficient to prove that Ψ is upper semicontinuous when SΓ2 is endowed with the weak topology of L2 ([0, T ], H). As L2 ([0, T ], H) is separable, SΓ2 is compact metrizable with respect to the weak topology of L2 ([0, T ], H). So it turns out to check that the graph gph(Ψ) is sequentially weakly closed in SΓ2 × SΓ2 . Let (fn , gn )n ∈ gph(Ψ) weakly converging to (f, g) ∈ SΓ2 × SΓ2 . From the definition of Ψ, that means ufn is the unique W 3,2 ([0, T ], H) solution of 0 u000 fn (t) ∈ ∂ϕ(ufn (t)) + fn (t) ufn (0) = −ufn (T ), u0fn (0) = a.e. t ∈ [0, T ], −u0fn (T ), u00fn (0) = −u00fn (T ), with fn ∈ SΓ2 and gn (t) ∈ F (t, ufn (t)) a.e. t ∈ [0, T ]. Taking into account the antiperiodicity of u00fn , u0fn and ufn , proposition 3.1 gives √ T 000 00 kufn kC([0,T ],H) ≤ kufn kL2 ([0,T ],H) , 2 √ T T 000 ku0fn kC([0,T ],H) ≤ kufn kL2 ([0,T ],H) , 2π √ T 2 T 000 kufn kL2 ([0,T ],H) . kufn kC([0,T ],H) ≤ 2π 2 for all n ≥ 1. Using the estimate (3.1), we have ku000 fn kL2 ([0,T ],H) ≤ kfn kL2 ([0,T ],H) ≤ kαkL2 ([0,T ],R) < +∞. We may conclude that sup ku00fn kC([0,T ],H) < +∞, n≥1 sup ku0fn kC([0,T ],H) < +∞, n≥1 EJDE-2013/08 EXISTENCE AND UNIQUENESS OF ANTI-PERIODIC SOLUTIONS 5 sup kufn kC([0,T ],H) < +∞. n≥1 By extracting suitable subsequences, we may assume that (u000 fn ) converges weakly 2 2 00 in L ([0, T ], H) to a function γ ∈ L ([0, T ], H) and (ufn ) converges pointwise to a function w, namely Z t 00 00 w(t) := lim ufn (t) = lim (ufn (0) + u000 fn (s)ds) n→+∞ n→+∞ 0 Z t = lim u00fn (0) + γ(s)ds, ∀t ∈ [0, T ]. n→+∞ 0 Then Z t v(t) := lim u0fn (t) = lim (u0fn (0) + u00fn (s)ds) n→+∞ n→+∞ 0 Z t = lim u0fn (0) + w(s)ds, ∀t ∈ [0, T ]. n→+∞ 0 So we have Z t u(t) := lim ufn (t) = lim (ufn (0) + u0fn (s)ds) n→+∞ n→+∞ 0 Z t = lim ufn (0) + v(s)ds, ∀t ∈ [0, T ]. n→+∞ 0 We conclude that u ∈ W 3,2 ([0, T ], H) with u0 = v, u00 = w and u000 = γ and satisfying the anti-periodic conditions u(0) = −u(T ), u0 (0) = −u0 (T ) and u00 (0) = −u00 (T ). Furthermore, we see that u0fn converges pointwise to u0 and u000 fn converges 000 2 to u with respect to the weak topology of L ([0, T ], H). Combining these facts and applying Lemma 2.1 to the inclusion 0 u000 fn (t) − fn (t) ∈ ∂ϕ(ufn (t)) a.e. t ∈ [0, T ] it yields u000 (t) − f (t) ∈ ∂ϕ(u0 (t)) a.e. t ∈ [0, T ]. By uniqueness [1, Theorem 2.1], we have u = uf . Further using the inclusion gn (t) ∈ F (t, ufn (t)) a.e. t ∈ [0, T ] and invoking the closure type Lemma in [10, theorem VI.4], we have g(t) ∈ F (t, uf (t)) a.e. t ∈ [0, T ]. We may then applying the Kakutani-Ky Fan fixed point theorem to the set-valued mapping Ψ to obtain some f ∈ SΓ2 such that f ∈ Ψ(f ) or f (t) ∈ F (t, u(t)) a.e. t ∈ [0, T ]. This means that u000 (t) ∈ ∂ϕ(u0 (t)) + f (t) f (t) ∈ F (t, u(t)) u(0) = −u(T ), The proof is complete. 0 a.e. t ∈ [0, T ], a.e. t ∈ [0, T ], 0 u (0) = −u (T ), u00 (0) = −u00 (T ). 6 T. HADDAD, T. HADDAD EJDE-2013/08 A more general version of the preceding result is available by introducing some inf-compactness assumption [2] on the function ϕ. Theorem 3.3. Let H be a separable Hilbert space, ϕ : H → [0, +∞] be a proper, convex, lower semicontinuous and even function satisfying: ϕ(0) = 0 and for each β1 , β2 > 0, the set {x ∈ D(ϕ) : kxk ≤ β1 , ϕ(x) ≤ β2 } is compact. Let F : [0, T ] × H → 2H be a convex compact set-valued mapping, measurable on [0, T ] and upper semicontinuous on H satisfying: there is α(·) ∈ L2 ([0, T ], R+ ) such that F (t, x) ⊂ Γ(t) := Bα(t) (0) ∀(t, x) ∈ [0, T ] × H Then the problem u000 (t) ∈ ∂ϕ(u0 (t)) + F (t, u(t)) 0 u(0) = −u(T ), 0 u (0) = −u (T ), a.e. t ∈ [0, T ], u00 (0) = −u00 (T ), has at least an anti-periodic W 3,2 ([0, T ], H) solution. Proof. Using the notation of the proof of Theorem 3.2, we have 0 u000 fn (t) − fn (t) ∈ ∂ϕ(ufn (t)) a.e. t ∈ [0, T ], for every fn ∈ SΓ2 . The absolute continuity of ϕ(u0fn (·)) and the chain rule theorem [9], yield d 00 00 hu000 ϕ(u0fn (t)), fn (t), ufn (t)i − hfn (t), ufn (t)i = dt for every fn ∈ SΓ2 , so that Z T Z T d 00 00 ϕ(u0f (t))dt. +∞ > sup |hu000 (t), u (t)i − hf (t), u (t)i|dt = sup n fn fn n n dt n≥1 0 n≥1 0 Further applying the classical definition of the subdifferential to convex function ϕ yields 0 = ϕ(0) ≥ ϕ(u0fn (t)) + h0 − u0fn (t), u000 fn (t) − fn (t)i or 0 ≤ ϕ(u0fn (t)) ≤ hu0fn (t), u000 fn (t) − fn (t)i. Hence sup |ϕ(u0fn )|L1R ([0,T ]) < +∞. n≥1 For all t ∈ [0, T ], we have ϕ(u0fn (t)) = ϕ(u0fn (0)) t Z + 0 d ϕ(u0fn (s))ds ≤ ϕ(u0fn (0)) + sup |ϕ(u0fn )|L1R ([0,T ]) . dt n≥1 ϕ(u0fn (t)) Now we assert that ≤ β2 for every t ∈ [0, T ], here β2 is a positive constant. Indeed for all t ∈ [0, T ], we have ϕ(u0fn (0)) ≤ |ϕ(u0fn (t)) − ϕ(u0fn (0))| + ϕ(u0fn (t)) Z T d ≤ | ϕ(u0fn (t))|dt + ϕ(u0fn (t)). dt 0 Hence ϕ(u0fn (0)) ≤ sup n≥1 Z T | 0 d 1 ϕ(u0fn (t))|dt + sup dt T n≥1 Z 0 T ϕ(u0fn (t))dt < +∞. EJDE-2013/08 EXISTENCE AND UNIQUENESS OF ANTI-PERIODIC SOLUTIONS 7 Whence we have β1 := sup sup ku0fn (t)k < +∞, n≥1 t∈[0,T ] β2 := sup sup ϕ(u0fn (t)) < +∞. n≥1 t∈[0,T ] (u0fn (t)) So that is relatively compact with respect to the norm topology of H using the inf-compactness assumption on ϕ. The proof can be therefore achieved as Theorem 3.2 by invoking Lemma2.1 and a closure type lemma in [10, Theorem VI-4]. Here is an existence and uniqueness result related to Theorem 3.3 when the perturbation is single-valued. Theorem 3.4. Let H be a separable Hilbert space, ϕ : H → [0, +∞] is a proper, convex, lower semicontinuous and even function satisfying: ϕ(0) = 0 and for each α, β > 0, the set {x ∈ D(ϕ) : kxk ≤ α, ϕ(x) ≤ β} is compact and f : [0, T ]×H → H is a Carathéodory mapping satisfying : (H1 ) kf (t, u) − f (t, v)k ≤ Lku − vk for all (t, u, v) ∈ [0, T ] × H × H, for some positive constant L > 0. (H2 ) There is a L2 ([0, T ]; R+ ) integrable function α : R → R+ such that kf (t, u)k ≤ π α(t) for all (t, u) ∈ [0, T ] × H. If 0 < T < √ , then the inclusion 3 L u000 (t) ∈ ∂ϕ(u0 (t)) + f (t, u(t)) u(0) = −u(T ), 0 0 u (0) = −u (T ), a.e. t ∈ [0, T ], u00 (0) = −u00 (T ), admits a unique W 3,2 ([0, T ]; H)-anti-periodic solution. Proof. Existence of at least one W 3,2 ([0, T ]; H)-anti-periodic solution is ensured by Theorem 3.3. Indeed, we put F (t, u) := {f (t, u)} for all (t, u) ∈ [0, T ] × H, As f is a Carathéodory function, then: u 7−→ f (t, u) is continuous, for almost all t ∈ [0, T ] and t 7−→ f (t, u) is Lebesgue measurable, for all u ∈ H. More, by assumption; there is a L2 ([0, T ]; R+ ) integrable function α(·) such that kf (t, u)k ≤ α(t) for all (t, u) ∈ [0, T ] × H. Therefore, F satisfies hypotheses of Theorem 3.3. To prove uniqueness, we assume that (u1 ) and (u2 ) are two solutions of the inclusion under consideration. 0 u000 1 (t) ∈ ∂ϕ(u1 (t)) + f (t, u1 (t)) u1 (0) = −u1 (T ), u01 (0) = −u01 (T ), a.e. t ∈ [0, T ], u001 (0) = −u001 (T ), and 0 u000 2 (t) ∈ ∂ϕ(u2 (t)) + f (t, u2 (t)) u2 (0) = −u2 (T ), u02 (0) = −u02 (T ), a.e. t ∈ [0, T ], u002 (0) = −u002 (T ). For simplicity, let us set v1 (t) = u000 1 (t) − f (t, u1 (t)), v2 (t) = u000 2 (t) − f (t, u2 (t)), ∀t ∈ [0, T ], ∀t ∈ [0, T ]. Then we have 000 v1 (t) − v2 (t) = u000 1 (t) − u2 (t) − f (t, u1 (t)) + f (t, u2 (t)), a.e. t ∈ [0, T ]. (3.2) 8 T. HADDAD, T. HADDAD EJDE-2013/08 Multiplying scalarly (3.2) by (u01 − u02 ) and integrating on [0, T ] yields T Z hv1 (t) − v2 (t), u01 (t) − u02 (t)idt 0 T Z 000 0 0 hu000 1 (t) − u2 (t), u1 (t) − u2 (t)idt = (3.3) 0 Z − T hf (t, u1 (t)) − f (t, u2 (t)), u01 (t) − u02 (t)idt 0 As v1 ∈ ∂ϕ(u01 ) and v2 ∈ ∂ϕ(u02 ) by monotonicity of (∂ϕ), (3.3) implies Z T 000 0 0 hu000 1 (t) − u2 (t), u1 (t) − u2 (t)idt 0 Z ≥ (3.4) T hf (t, u1 (t)) − f (t, u2 (t)), u01 (t) − u02 (t)idt. 0 By antiperiodicity, we have Z T 000 0 0 hu000 1 (t) − u2 (t), u1 (t) − u2 (t)idt 0 = hu001 (T ) − u002 (T ), u01 (T ) − u02 (T )i − hu001 (0) − u002 (0), u01 (0) − u02 (0)i Z T − hu001 (t) − u002 (t), u001 (t) − u002 (t)idt 0 Z =− T ku001 (t) − u002 (t)k2 dt. 0 The inequality (3.4) gives Z T Z ku001 (t) − u002 (t)k2 dt ≤ 0 T hf (t, u2 (t)) − f (t, u1 (t)), u01 (t) − u02 (t)idt 0 Z ≤L T ku1 (t) − u2 (t)k ku01 (t) − u02 (t)kdt. 0 By Holder’s inequality, we obtain ku001 − u002 k2L2 ([0,T ],H) ≤ Lku1 − u2 kL2 ([0,T ],H) ku01 − u02 kL2 ([0,T ],H) . Using the estimates (A1) and (A2) in proposition 3.1, we obtain T π2 0 ku1 − u02 k2L2 ([0,T ],H) ≤ L ku01 − u02 k2L2 ([0,T ],H) 2 T π or ku01 − u02 k2L2 ([0,T ],H) ≤ L T3 0 ku − u02 k2L2 ([0,T ],H) . π3 1 It follows from the choice of T that ku01 − u02 k2L2 ([0,T ],H) = 0. By inequality (A2) in lemma 2.2, we conclude that u1 (t) − u2 (t) = 0 for all t ∈ [0, T ]. This completes the proof. EJDE-2013/08 EXISTENCE AND UNIQUENESS OF ANTI-PERIODIC SOLUTIONS 9 4. Applications Let Ω be a bounded domain in Rn with a smooth boundary ∂Ω. Let γ be a maximal monotone operator on R such that γ = ∂j, where j : R → [0, +∞] is proper, convex, lower semicontinuous and even with j(0) = 0. We are concerned with the third-order boundary-value problem −uttt (t, x) − ∆x ut (t, x) + f (t, u(t, x)) = 0 in [0, T ] × Ω, ∂ut (4.1) (t, x) ∈ γ(ut (t, x)) on [0, T ] × ∂Ω, ∂ν u(0, x) = −u(T, x), ut (0, x) = −ut (T, x) utt (0, x) = −utt (T, x) in Ω, Pn ∂ 2 ∂ where ∂ν denotes outward normal derivative, ∆x = i=1 ∂x 2 , and f : [0, T ] × R → i R is a Carathéodory function satisfying: (i) |f (t, u) − f (t, v)| ≤ L|u − v| for all (t, u, v) ∈ [0, T ] × R2 , for some positive constant L > 0, (ii) There is an L2 ([0, T ]; R+ ) integrable function α : [0, T ] → R+ such that |f (t, u)| ≤ α(t) for all (t, u) ∈ [0, T ] × R. Let H = L2 (Ω), and define ϕ : H → [0, +∞] by ( R R 1 | grad u|2 dx + ∂Ω j(u)dσ, if u ∈ H 1 (Ω) and j(u) ∈ L1 (∂Ω), 2 Ω ϕ(u) = +∞, otherwise. According to Brézis [8, Theorem 12], ϕ is proper, convex and lower semicontinuous on H, with ∂ϕ(u) = −∆x u, and D(ϕ) = {u ∈ W 1,2 (Ω) : − ∂u ∂ν ∈ γ(u), a.e. on ∂Ω}. We consider u = u(t, x) = u(t)(x) and we rewriter the problem (4.1) in the abstract form −u000 (t) + ∂ϕ(u0 (t)) + f (t, u(t)) 3 0 u(0) = −u(T ), 0 0 u (0) = −u (T ), a.e. t ∈ [0, T ], 00 u (0) = −u00 (T ), or u000 (t) ∈ ∂ϕ(u0 (t)) + f (t, u(t)) u(0) = −u(T ), 0 0 u (0) = −u (T ), a.e. t ∈ [0, T ], u00 (0) = −u00 (T ). We remark that ϕ(0) = 0, ϕ is even and that the inf-compactness condition on ϕ holds because W 1,2 (Ω) is compactly imbedded in L2 (Ω). Then, we can applying π Theorem 3.4 to derive the existence of a solution to (4.1). If 0 < T < √ , then the 3 L solution is unique. Acknowledgments. The authors would like to thank the anonymous referees for their careful and thorough reading of the original manuscript. References [1] A. R. Aftabizadeh, Y. K. Huang, N.H. Pavel; Nonlinear Third-Order Differential Equations with Anti-Periodic Boundary Conditions and some Optimal Control Problems, J. Math. Anal. Appl. 192, (1995), 266-293. [2] S. Aizicovici, N. H. Pavel; Anti-Periodic Solutions to a class of Nonlinear Differential Equations in Hilbert Space, J. Fun. Anal. 99 (1991) 387-408. [3] S. Aizicovici, N. H. Pavel; Anti-Periodic boundary value problems for higher order differential equations in Hilbert spaces, Nonlinear Anal.18 (1992) 253-267. 10 T. HADDAD, T. HADDAD EJDE-2013/08 [4] S. Aizicovici, N. H. Pavel; On class of second order anti-periodic boundary value problem, J. Math. Appl. 171 (1992) 301-320. [5] S. Aizicovici, N. H. Pavel, I. I. Vrabie; Anti-Periodic Solutions to Strongly Nonlinear Evolutions Equations in Hilbbert Space, Analele Stiintifice Ale Universitatii ”A.I.Cuza” Iasi, Tomul XIV, s.I.a. Matematica, 1998. [6] C. Castaing, T. Haddad; Relaxation and Bolza Problem Involving a Second Order Evolution Inclusion, Yokohama Publishers, Volum 9, Number 2, 2008, 141-159. [7] C. Castaing, P. Raynaud de Fitte, A. Salvadori; Some variational convergence results with application to evolution inclusions, Adv. Math. Econ, 8 (2006), 33-73. [8] H. Brézis; Monotonicity methods in Hilbert space and some applications to nonlinear partial differential equations, Contribution to Nonlinear Functional Analysis, Acadimic press, New York, 1971, pp. 101-156. [9] H. Brézis; Opérateurs Maximaux Monotones et Semi-Groupes de Contractions dans les Espaces de Hilbert, North-Holland, Amsterdam, 1973. [10] C. Castaing, M. Valadier; Convex analysis and measurable multifunctions, Lectures Notes in Mathematics, Vol. 580, Springer-Verlag, Berlin (1977). [11] J. F Couchouron, R. Precup; Anti-periodic solution for second order differential inclusions, Electron. J. Differential Equations 2004 (2004) 1-17. [12] A. Haraux; Anti-periodic solutions of some nonlinear evolution equations, Manuscripta. Math. 63 (1989) 479-505. [13] H. Okochi; On the existence of periodic solution of some nonlinear abstract parabolic equations, J. Math. sOC. Japon 40 (1988) 541-553. [14] H. Okochi; On the existence of anti-periodic solutions to nonlinear evolutions equations associated with odd subdifferential operators, J. Funct. Aal. 90 (1990) 246-258. [15] P. Souplet; Optimal uniqueness condition for the anti-periodic solutions of some nonlinear parabolic equations, Nonlinear Analysis, Theory, Methods and Applications, Vol. 32, No. 2, pp. 279-286, 1998. Touma Haddad Département de Mathématiques, Faculté des Sciences, Université de Jijel, B.P. 98, Algérie E-mail address: touma.haddad@yahoo.com Tahar Haddad Laboratoire LMPA, Faculté des Sciences, Université de Jijel, B.P. 98, Algérie E-mail address: haddadtr2000@yahoo.fr