In this paper, we investigate the existence of solution for differential systems involving a ϕ−Laplacian operator which incorporates as a special case the well-known p−Laplacian operator. In this purpose, we use a variational method which relies on Szulkin’s critical point theory. We obtain the existence of solution when the corresponding Euler–Lagrange functional is coercive.
1. Introduction
This paper is devoted to the study of the following second-order differential systems:(1)−ϕu′t′+εϕut=∇Ft,ut a.e.on Ω=0,Tϕu′0,−ϕu′T∈∂ju0,uT,where ε≥0 is fixed, ϕ:ℝN⟶ℝN is a monotone homeomorphism, j:ℝN×ℝN⟶−∞,+∞ is a proper, convex, and lower semicontinuous function, F:Ω×ℝN⟶ℝN is a Caratheodory mapping, continuously differentiable with respect to the second variable and satisfies some usual growth conditions, ∇Ft,x is the gradient of Ft,. at x∈ℝN and ∂j denotes the subdifferential of j in the sense of convex analysis.
Second-order differential problems with multivalued boundary value conditions have been studied by many authors. In this direction, we can cite the works of Bader and Papageorgioua [1], Béhi et al. [2], Gasinski and Papageorgiou [4], Jebelean [5], Jebelean and Morosanu [6], and references therein. In [1, 4–6], the authors investigate differential systems driven by a homogeneous p−Laplacian operator while in problem (1), we deal with a nonhomogeneous ϕ−Laplacian operator which incorporates as a special case the p−Laplacian operator. As a consequence, problem (1) is a generalization of the following problem studied, in 2005, by Jebelean and Morosanu [6]:(2)−ϕpu′t′+εϕput=∇Ft,ut a.e. on Ω=0,Tϕpu′0,−ϕpu′T∈∂ju0,uT,where ϕp denotes the homogeneous operator p−Laplacian. Indeed, in our work, the ϕ−Laplacian operators are nonhomogeneous and are of the form ϕx=axϕpx with ϕp the p−Laplacian operator, p>1, and a:ℝN⟶ℝ+∗ a continuous map.
In order to obtain existence result for problem (1), we will use a variational method which relies on Szulkin’s critical point theory [7].
This paper is organized as follows: After introducing notations and preliminary results in Section 2, in Section 3 we give a variational approach to problem (1). In Section 4, using some results from Section 3, we prove our main result. In Section 4, we give an example to illustrate the applicability of our result. Finally, Section 5 is reserved for the conclusion.
2. Preliminaries
Let us recall some notions and results in the framework of Szulkin’s critical point theory [7] which are needed and also some notations which will be used in the sequel. At this end, W1,p0,T,ℝN is the Sobolev Banach space which will be endowed with the norm:(3)um=mupp+u′pp1/p,where m>0 and .p is the norm on Lp0,T,ℝN defined by(4)up=∫0Tupdt1/p.
We denote .∞, the norm on the set C0,T,ℝN:(5)u∞=maxut:t∈0,T.
Let X, be a real Banach space and B:X⟶−∞,+∞ be a functional of type(6)B=E+G,where E∈C1X,ℝ and G is proper, convex, and lower semicontinuous (lsc). A point u∈X is said to be a critical point of B if it satisfies the inequality(7)E′u,v−u+Gv−Gu≥0,∀v∈X.
A number c∈ℝ such that B−1c contains a critical point is called a critical value of B.
To establish existence results for (1), we will need the following proposition due to Szulkin (see [7] Proposition 1.1).
Proposition 1.
If B satisfies (7), each local minimum point of B is necessarily a critical point of B.
The functional B is said to satisfy the Palais-smale (in short, (PS)) condition if every sequence un⊂X for which Bun⟶c∈ℝ and(8)E′un,v−un+Gv−Gun≥εnv−un,∀v∈X,where εn⟶0 possesses a convergent subsequence.
3. A Variational Approach for Problem (1)
Before beginning the variational approach, we make the following hypotheses on the data of problem (1):
Ha: a:ℝN⟶ℝ+∗ is a continuous map such that there exists η,M>0 such that
(9)η≤ax≤M,∀x∈ℝN.
Hϕ: ϕ:ℝN⟶ℝN is a monotone homeomorphism such that
aϕx=axϕpx, ∀x∈ℝN, with ϕp the well-know p−Laplacian operator and ∇Φ=ϕ with Φ:ℝN⟶ℝ of class C1 on ℝN, strictly convex (i.e., Φ is a potential function corresponding to ϕ) and such that Φ0=0;
b there exist c1,c2,c3>0 such that
(10)c1xp≤Φx≤c2+c3xp,∀x∈ℝN,
where . denotes the Euclidean norm on ℝN.
Remark 1.
Suppose that ∀x∈ℝN, ax=1, we have ϕx=ϕpx=xp−2x,p>1. Then this function satisfies hypotheses Ha and Hϕ. Other cases that satisfy Hypotheses Ha and Hϕ are when η<ax<M with η,M>0 and the function ϕ:x⟼axxp−2x is a monotone homeomorphism on ℝN. This is the case, for example, when function a is equal to one of the following functions:(11)x⟼1+11+xp2or x⟼p+p−xex,∀x∈ℝN,with p>1. Indeed, for these examples, the potentials functions Φ corresponding to ϕ are respectively the functions x⟼1/p1+1/1+xpxp and x⟼1+1/exxp, ∀x∈ℝN.
HF: F:0,T×ℝN⟶ℝN is a Caratheodory mapping such that
HF1: F (t, .) is continuously differentiable, for a.e. t∈0,T
HF2: F.,0∈L10,T
HF3: for each R>0 there is some αR∈L10,T such that
(12)∇Ft,x≤αRt,a.e.t∈0,T,∀x∈ℝN with x≤R.
Our notion of solution of problem (1) is defined as follows:
Definition 1.
By a solution of the differential system (1), we will understand a function u:0,T⟶ℝN of class C1Ω,ℝN with ϕu′ absolutely continuous, which satisfies the equality in (1) a.e. on 0,T.
Now let us start the variational approach. In this purpose, for ε>0, let Kε:W1,p0,T,ℝN⟶ℝ be defined by(13)Kεu=∫0TΦu′s+εΦusds,∀u∈W1,p0,T,ℝN.
Lemma 1.
Kε is proper, convex, and lower semicontinuous.
Proof.
By hypothesis Hϕ, it follows(14)c1u′pp+εupp≤Kεu≤1+εc2+c3u′pp+εupp,∀u∈W1,p0,T,ℝN.
Whence Kε is proper. Also since Φ is convex, it follows that Kε is convex. Finally, because of the lower semicontinuity of the functional norm on Banach space, Kε is lower semicontinuous (in short, lsc).
Lemma 2.
Kε∈C1W1,p0,T,ℝN,ℝ and(15)K′εu,w=∫0Tϕu′t,w′tℝN+εϕut,wtℝNdt,∀u,w∈W1,p0,T,ℝN,where ℝN is the inner product in ℝN.
Proof.
Let us consider the product space H=Πi=1NLqΩ,ℝ, 1/p+1/q=1, equipped with the norm(16)xH=∑i=1Nxiqq1/q,∀x=x1,x2,…,xn−1,xn∈H.
We define gg1,g2,…,gn−1,gN:W1,p0,T,ℝN⟶H such that(17)gu=ϕu′+εϕu,∀u∈W1,p0,T,ℝN.
Let us show that g is bounded.
We have(18)giuqq=∫0Tϕiu′t+εϕiutqdt=∫0Tau′tu′tp−2ui′t+εaututp−2uitqdt≤Mq∫0Tu′tp−1q+εutp−1qdt=Mqu′pp+εupp=Mquεp≤Mqη1ump,for some η1>0.
It follows(19)guH≤MNη11/qump−1,∀u∈W1,pΩ.
So g is bounded.
Let us show that g is continuous.
Let unn≥1 be a sequence such that un⟶u in W1,pΩ. Then un⟶u in LpΩ and un′⇀u′ in LpΩ. Whence ϕiun′⟶ϕiu′ in LqΩ and ϕiun⟶ϕiu in LqΩ. We set(20)giun=ϕiun′+εϕiun.
By the previous arguments, the sequence giun⟶giu in LqΩ. We infer that the sequence gun⟶gu in H. So g is continuous.
We consider the functional:(21)Qv=∫0Tϕu′t,v′t+εϕut,vtdt,for all v∈W1,pΩ.
Let us show that the linear operator Q is continuous on W1,pΩ.
Using Hölder’s inequality, we obtain(22)Qv=∫0Tϕu′t,v′t+εϕut,vtdt≤∫0Tϕu′tqdt1/q∫0Tv′tpdt1/p+ε∫0Tϕutqdt1/q∫0Tvtpdt1/p≤M1+εump/qvm,∀u∈W1,pΩ.
So Q is continuous.
Let us show that Kε is Frechet differentiable in u∈W1,pΩ and Kε′u=ϕu′+εϕu in the sense of (15).
Using Fubini’s inequality, for arbitrary v∈W1,pΩ, we obtain(23)Kεu+v−Kεu−∫0Tϕu′t,v′t+εϕut,vtdt=∫0T∫01ddsΦu′t+sv′t+εddsΦut+svtds dt−∫0T∫01ϕu′t,v′t+εϕut,vtds dt=∫0T∫01ϕu′t+sv′t,v′t+εϕut+svt,vtds dt−∫0T∫01ϕu′t,v′t+εϕut,vtds dt=∫0T∫01ϕu′t+sv′t−ϕu′t,v′t+εϕut+svt−ϕut,vtds dt≤∫01∫0Tϕu′t+sv′t−ϕu′tq1/qdt∫0Tv′tp1/pdtds+ε∫01∫0Tϕut+svt−ϕutq1/qdt∫0Tvtp1/pdtds≤∫01ϕu′+sv′−ϕu′H+εϕu+sv−ϕuHdsvm≤∫01Lu+sv−LuH+εΛu+sv−ΛuHdsvm,with L:W1,pΩ⟶H and Lu=ϕu′; Λ:W1,pΩ⟶H and Λu=ϕu.
Arguing as in the proof of the continuity of g and the fact it is bounded on W1,pΩ, we show that L and Λ are continuous and bounded on W1,pΩ. Moreover, using Lebesgue’s dominated comvergence theorem, we have(24)Kεu+v−Kεu−∫0Tϕu′t,v′t+εϕut,vtdtvm≤∫01Lu+sv−LuH+εΛu+sv−ΛuHds⟶0,as vm⟶0.
Therefore, (15) is proved.
We know that Kε′:W1,pΩ⟶W1,pΩ∗ where W1,pΩ∗ denotes the dual of W1,pΩ.
Let us show that Kε′ is continuous.
We have(25)Kε′u−Kε′v,w=∫0Tϕu′t+εϕut−ϕv′t−εϕvt,wdt=∫0Tgut−gvt,wdt.
Using Hölder’s inequality, we have(26)Kε′u−Kε′v,w≤kgu−gvHwm,∀u,v,w∈W1,pΩ.
Whence(27)Kε′u−Kε′vW1,pΩ∗≤kgu−gvH.
Thus, since g is continuous, we obtain the continuity of Kε′.
We introduce also the functional: J:W1,p0,T,ℝN⟶−∞,+∞ defined by(28)Ju=ju0,uT,∀u∈W1,p0,T,ℝN.
Recall that j is proper, convex, and lsc. Then, J is also proper, convex, and lsc. Let us set(29)Δε=Kε+J.
Since Kε and J are proper, convex and lsc, it follows that Δε is proper, convex and lsc on W1,p0,T,ℝN. Assuming that hypotheses HF1,HF2 and HF3 on the Caratheodory function F hold, for each R>0, we obtain:(30)Ft,x≤RαRt+Ft,0,for a.e t∈0,T with x≤R,with αR∈L10,T. Equation (30) comes from inequality (12) and the estimation:(31)Ft,x=∫01ddsFt,sxds+Ft,0≤∫01∇Ft,sx,xℝNds+Ft,0≤x∫01∇Ft,sxds+Ft,0,equation (30) and the embedding W1,p0,T,ℝN⊂C0,T,ℝN allow us to introduce the functional: ΨF:W1,p0,T,ℝN⟶ℝ defined by(32)ΨFu=−∫0TFt,utdt+∫0TFt,0dt,∀u∈W1,p0,T,ℝN.
Lemma 3.
If the hypotheses HF1,HF2, and HF3 hold, then(33)ΨF∈C1W1,pΩ,ℝN,ℝ,(34)ΨF′u,v=−∫0T∇Ft,u,vdt,∀u,v∈W1,p0,T,ℝN.
Proof.
The proof is similar to the one of Theorem 2.6 of [2].
Considering the functional framework in Section 2, we set X=W1,p0,T,ℝN,E=ΨF in (32), G=Δε in (29) and BF,ε=B:(35)BF,ε=ΨF+Δε.
Proposition 2.
Let the Caratheodory function F:0,T×ℝN⟶ℝ satisfies HF1,HF2, and HF3 and let u∈W1,p0,T,ℝN. If u is a critical point of the functional BF,ε defined by (35), in the sense (7), i.e.,(36)ΨF′u,w−u+Δεw−Δεu≥0,∀w∈W1,p0,T,ℝN,then u is a solution of problem (1). The converse implication is also true.
Proof.
We suppose that u is a critical point of BF,ε. In the inequality (36), we take w=u+sv. Then dividing by s and letting s⟶0+, we obtain(37)ΨF′u,v+Kε′u,v+J′u;v≥0,∀v∈W1,p0,T,ℝN,where J′u;v is the directional derivative of the convex function J at u in the direction v. From (28) and (37), it follows(38)ΨF′u,v+Kε′u,v+j′u0,uT;v0,vT≥0,∀v∈W1,p0,T,ℝN.
Since C0∞0,T,ℝN⊂W1,p0,T,ℝN, using Hahn–Banach’s theorem (see Theorem I.1 of Brezis [3]), (38) implies(39)ΨF′u,v+Kε′u,v=0,∀v∈C0∞0,T,ℝN.
Using (15), (34), and (39), we obtain(40)∫0Tϕu′t,v′tℝNdt=∫0T−εϕut+∇Ft,ut,vtℝNdt,∀v∈C0∞0,T,ℝN.
From hypothesis Hϕb and u∈W1,p0,T,ℝN, it follows that(41)ϕu,ϕu′∈Lq0,T,ℝN,with 1p+1q=1.
Also, HF3 implies(42)∇F.,u∈L10,T,ℝN.
Equations (40) and (42) imply(43)ϕu′∈W1,10,T,ℝN,(44)−ϕu′t′=−εϕut+∇Ft,ut.
With ϕ being a homeomorphism, (43) ensures that u∈C10,T,ℝN. This together with (44) shows that u is the solution of the differential system (1). Furthermore, (38) and (44) yield(45)j′u0,uT;v0,vT≥ϕu′0,v0−ϕu′T,vT,∀v∈W1,p0,T,ℝN.
Thus,(46)j′u0,uT;x,y≥ϕu′0,x−ϕu′T,y,∀x,y∈ℝN,which, by a standard result from convex analysis (Theorem 23.2 of Rockafellar [8]), means that the boundary conditions in (1) are true.
Now let us show the converse implication.
By multiplying (1) with v∈W1,p0,T,ℝN and integrating by parts on 0,T, we obtain(47)∫0Tϕu′t,v′tℝN+εϕut,vtℝNdt+ϕu′0,v′0ℝN−ϕu′T,v′TℝN=∫0T∇Ft,ut,vtℝNdt.
Using inequalities (15) and (34) in (47), it follows(48)ΨF′u,v+Kε′u,v+ϕu′0,v′0ℝN−ϕu′T,v′TℝN=0.
By using Theorem 23.2 of Rockafellar [8], we obtain(49)ΨF′u,v+Kε′u,v+j′u0,uT;v0,vT≥0.
Using some previous arguments, we infer that(50)ΨF′u,w−u+Δεw−Δεu≥0,∀u∈W1,p0,T,ℝN.
4. Existence Results for Problem (1)
At first, let us introduce the constant(51)λ1=λ1p,j,ε=ε+infu′ppupp:u∈W1,p0,T,ℝN\0,u0,uT∈Dj,for ε≥0. If Dj=ℝN×ℝN, then λ1p,j,0=0 and if Dj=0;0, then λ1p,j,0>0. To obtain the existence result, we make the following hypothesis:
Hλ1λ1p,j,ε>0.
Proposition 3.
If Hλ1 is true, then(52)21/puλ1≤u′pp+εupp1/p≤uλ1,∀u∈DJ,where J is defined by (28).
Proof.
Since DJ=u∈W1,p0,T,ℝN:u0,uT∈Dj, by (51) we have(53)λ1=λ1p,j,ε=ε+infu′ppupp:u∈W1,p0,T,ℝN\0,u0,uT∈Dj.
Whence,(54)λ1≤ε+u′ppupp∀u∈W1,p0,T,ℝN\0,u0,uT∈Dj.
Then(55)λ1upp≤εupp+u′pp.
It follows(56)u′pp+λ1upp≤2u′pp+εupp.
Whence(57)2−1/Puλ1≤u′pp+εupp1/p.
Furthermore, we have(58)ε+u′ppupp≤λ1+u′ppupp.
Then(59)u′pp+εupp≤u′pp+λ1upp.
So(60)u′pp+εupp1/p≤uλ1.
Inequalities (57) and (60) yield the result.
If the nonlinearity F lies asymptotically on the left of λ1, then problem (1) is solvable.
Theorem 1.
Assume HF1,HF2, HF3, and Hλ1. If(61)limx⟶∞supFt,xc1xp<λ1,uniformly for a.e.t∈0,T,the problem (1) has at least a solution.
Proof.
Let us show that the functional BF,ε in (35) is sequentially weakly lsc and coercive on the space W1,p0,T,ℝN,λ1.
Let us show that ΨF in (32) is sequentially weakly continuous.
Let u,v∈W1,p0,T,ℝN be such that u∞,v∞≤M, with some M>0. By HF3, there is an α2M∈L10,T such that(62)∇Ft,x≤α2M,for a.e t∈0,T,∀x∈ℝN with α2M≤2M.
We estimate(63)ΨFu−ΨFv=∫0TFt,u−Ft,v=∫0T∫01ddsFt,u+sv−uds=∫0T∫01∇Ft,u+sv−u,v−uℝNds≤∫0T∫01∇Ft,u+sv−udsu−v∞,and by (62), it follows(64)ΨFu−ΨFv≤∫0Tα2Mtdtu−v∞.
By the compactness of the embedding W1,p0,T,ℝN⊂C0,T,ℝN and (64), it follows that ΨF is sequentially weakly continuous on W1,p0,T,ℝN. Then, by the weak lower semicontinuity of Δε in (29), BF,ε is sequentially weakly lower semicontinuous.
Furthermore, from (61), there are constant σ∈0,λ1 and ρ>0 such that(65)Ft,x≤λ1−σc1xp,for a.e t∈0,T,∀x∈ℝN with x>ρ.
Then, (30) and (65) yield(66)Ft,x≤ραpt+Ft,0+λ1−σc1xp,for a.e t∈0,T,∀x∈ℝN,which, by (32), give(67)ΨFu≥−γρ−λ1−σc1upp,∀u∈W1,p0,T,ℝN,with γρ=ρ∫0Tαptdt+2∫0TF0,tdt. Using (29), (67), and (53) and the Proposition 3, we estimate BF,ε on DJ as follows:(68)BF,εu=ΨFu+Δεu≥−γρ−λ1−σc1upp+c1u′pp+εupp+Ju≥−γρ−λ1−σc1u′pp+εuppλ1+c1u′pp+εupp+Ju=−γρ+σλ1c1u′pp+εupp+Ju≥−γρ+σ2λ1c1uλ1p+Ju,∀u∈DJ.
Since j is convex and lsc, it is bounded from below by an affine functional. Therefore, by (28) there are constants k1,k2,k3≥0 such that(69)BF,εu≥−γρ+σ2λ1c1uλ1p−k1u0−k2uT−k3,∀u∈DJ.
From (69) and the continuity embedding of W1,p0,T,ℝN⊂C0,T,ℝN, we obtain(70)BF,εu≥−γρ+σ2λ1c1uλ1p−k¯1uλ1−k¯2uλ1−k3,∀u∈DJ.with some k¯1,k¯2≥0. Consequently,(71)BF,εu⟶+∞,as uλ1⟶∞,meaning that BF,ε is coercive on W1,p,.λ1.
Since the functional BF,ε is sequentially weakly lsc and coercive, by a well-known result from calculus of variations, BF,ε is bounded from below and attains its infimum at some u∈W1,p0,T,ℝN. Then, by Proposition 1 and 3, u is a solution of problem (1).
Remarque 2. The problem (1) incorporates classical problems of Dirichlet, Neumann, Periodic, antiperiodic, and Sturm–Louiville (see Remark 3.9 of Jebelean [5]).
5. Example
We consider the following problem:(72)−1+11+u′tp2ϕpu′t′+ε1+11+u′tp−12ϕput=βtϕput,a.e on Ω=0,T1+11+u′0p2ϕpu′0,−1+11+u′Tp2ϕpuT∈∂ju0,uT,where ϕpx=xp−2x,∀x∈ℝN, is the well-known p−Laplacian operator. Let us set Φx=1/p1+1/1+xpxp. Whence ∇Φx=ϕx=1+1/1+xp2ϕpx. When, p>1, we see that ϕ is a monotone homeomorphism such that ϕ0=0. Moreover, Φ satisfies hypothesis Hϕb. Therefore, by Theorem 1, problem (72) has at least a solution.
6. Conclusion
In this article, using variational method which relies on Szulkin theory, we have established existence for second-order problems with multivalued boundary conditions. We give an example but more examples and applications can be given.
Data Availability
No data were used to support this study.
Conflicts of Interest
The authors declare no conflicts of interest regarding the publication of this paper.
BaderR.PapageorgiouaN. S.Quasilinear vector differential equations with maximal monotone terms and nonlinear boundary conditions2000731699210.4064/ap-73-1-69-92BéhiD. A.AdjéA.GoliK. C.Lower and upper solutions method for nonlinear second-order differential equations involving a ϕ-Laplacian operator20192212241BrezisH.1983Paris, FranceMassonGasinskiL.PapageorgiouN. S.Nonlinear second-order multivalued boundary value problems2003113329331910.1007/bf028296082-s2.0-0042763637JebeleanP.Variational methods for ordinary p-Laplacian systems with potential boundary conditions2008133-4273322JebeleanP.MorosanuG.Ordinary p-Laplacian systems with nonlinear boundary conditions2006313273875310.1016/j.jmaa.2005.04.0222-s2.0-28244484046SzulkinA.Minimax principles for lower semicontinuous functions and applications to nonlinear boundary value problems1986327710910.1016/s0294-1449(16)30389-42-s2.0-0002349602RockafellarR. T.1972Princeton, NJ, USAPrinceton University Press