By using the perturbation theories on sums of ranges of nonlinear accretive mappings of Calvert and Gupta (1978), the abstract result on the existence and uniqueness of the solution in Lp(Ω) of the generalized Capillarity equation with nonlinear Neumann boundary value conditions, where 2N/(N+1)<p<+∞ and
N≥1 denotes the dimension of RN, is studied. The equation discussed in this paper and the methods here are a continuation of and a complement to the previous corresponding results. To obtain the results, some new techniques are used in this paper.
1. Introduction and Preliminary
Since the p-Laplacian operator -Δp with p≠2 arises from a variety of physical phenomena, such as nonNewtonian fluids, reaction-diffusion problems, and petroleum extraction, it becomes a very popular topic in mathematical fields.
We began our study on this topic in 1995. We used a perturbation result of ranges for m-accretive mappings in Calvert and Gupta [1] to obtain a sufficient condition in Wei and He [2] so that the following zero boundary value problem, (1.1)-Δpu+g(x,u(x))=f(x),a.e.onΩ,-∂u∂n=0,a.e.onΓ,
has solutions in Lp(Ω), where 2≤p<+∞. Later on, a series work of ours has been done from different angles on this kind of equations, cf. [3–7], and so forth.
Especially, in 2008, as a summary of the work done in [2–6], we use some new techniques to work for the following problem with so-called generalized p-Laplacian operator:
(1.2)-div[(C(x)+|∇u|2)(p-2)/2∇u]+ε|u|q-2u+g(x,u(x))=f(x),a.e.inΩ,-〈ϑ,(C(x)+|∇u|2)(p-2)/2∇u〉∈βx(u(x)),a.e.onΓ,
where 0≤C(x)∈Lp(Ω), ε is a nonnegative constant and ϑ denotes the exterior normal derivative of Γ. We showed in Wei and Agarwal [7] that (1.2) has solutions in Ls(Ω) under some conditions, where 2N/(N+1)<p≤s<+∞, 1≤q<+∞ if p≥N, and 1≤q≤Np/(N-p) if p<N, for N≥1.
Capillarity equation is another important equation appeared in the capillarity phenomenon and we notice that in Chen and Luo [8], the authors studied the eigenvalue problem for the following generalized Capillarity equations:
(1.3)-div[(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u]=λ(|u|q-2u+|u|r-2u),inΩ,u=0,a.e.on∂Ω.
Their work inspired us and one idea came to our mind. Can we borrow the main ideas dealing with the nonlinear elliptic boundary value problems with the generalized p-Laplacian operator to study the nonlinear generalized Capillarity equation with Neumann boundary conditions?
We will answer the question in this paper. By using the perturbation results of ranges for m-accretive mappings in Calvert and Gupta [1] again, we will study the following one:
(1.4)-div[(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u]+λ(|u|q-2u+|u|r-2u)+g(x,u(x))=f(x),a.e.inΩ,-〈ϑ,(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u〉∈βx(u(x)),a.e.onΓ.
More details on (1.4) will be given in Section 2. Our methods and techniques are different from those in Chen and Luo [8].
Now, we list some basic knowledge we need in sequel.
Let X be a real Banach space with a strictly convex dual space X*. We use “→” and “w-lim” to denote strong and weak convergence, respectively. For any subset G of X, we denote by intG its interior and G¯ its closure, respectively. Let “X↪↪Y” denote that space X is embedded compactly in space Y and “X↪Y” denote that space X is embedded continuously in space Y. A mapping T:D(T)=X→X* is said to be hemicontinuous on X if w-limt→0T(x+ty)=Tx, for any x,y∈X. Let J denote the duality mapping from X into 2X* defined by
(1.5)J(x)={f∈X*:(x,f)=∥x∥·∥f∥,∥f∥=∥x∥},x∈X,
where (·,·) denotes the generalized duality pairing between X and X*. It is wellknown that J is a single-valued mapping since X* is strictly convex.
Let A:X→2X be a given multivalued mapping. We say that A is boundedly-inversely compact if for any pair of bounded subsets G and G′ of X, the subset G⋂A-1(G′) is relatively compact in X. The mapping A:X→2X is said to be accretive if (v1-v2,J(u1-u2))≥0, for any ui∈D(A) and vi∈Aui, i=1,2. The accretive mapping A is said to be m-accretive if R(I+μA)=X, for some μ>0.
Let B:X→2X* be a given multi-valued mapping. The graph of B, G(B), is defined by G(B)={[u,w]∣u∈D(B),w∈Bu}. Then B:X→2X* is said to be monotone if G(B) is a monotone subset of X×X* in the sense that
(1.6)(u1-u2,w1-w2)≥0,
for any [ui,wi]∈G(B), i=1,2. The monotone operator B is said to be maximal monotone if G(B) is maximal among all monotone subsets of X×X* in the sense of inclusion. The mapping B is said to be strictly monotone if the equality in (1.6) implies that u1=u2. The mapping B is said to be coercive if limn→+∞((xn,xn*)/∥xn∥)=∞ for all [xn,xn*]∈G(B) such that limn→+∞∥xn∥=+∞.
Definition 1.1.
The duality mapping J:X→2X* is said to be satisfying Condition (I) if there exists a function η:X→[0,+∞) such that
(I)∥Ju-Jv∥≤η(u-v),for∀u,v∈X.
Definition 1.2.
Let A:X→2X be an accretive mapping and J:X→X* be a duality mapping. We say that A satisfies Condition (*) if, for any f∈R(A) and a∈D(A), there exists a constant C(a,f) such that
(*)(v-f,J(u-a))≥C(a,f),foranyu∈D(A),v∈Au.
Lemma 1.3 (Li and Guo [9]).
Let Ω be a bounded conical domain in RN. Then we have the following results.
If mp>N, then Wm,p(Ω)↪CB(Ω); if mp<N and q=Np/(N-mp), then Wm,p(Ω)↪Lq(Ω); if mp=N and p>1, then for 1≤q<+∞,Wm,p(Ω)↪Lq(Ω).
If mp>N, then Wm,p(Ω)↪↪CB(Ω); if 0<mp≤N and q0=Np/(N-mp), then Wm,p(Ω)↪↪Lq(Ω), 1≤q<q0.
Lemma 1.4 (Pascali and Sburlan [10]).
If B:X→2X* is an everywhere defined, monotone and hemicontinuous operator, then B is maximal monotone. If B:X→2X* is maximal monotone and coercive, then R(B)=X*.
Lemma 1.5 (Pascali and Sburlan [10]).
If Φ:X→(-∞,+∞] is a proper convex and lower-semicontinuous function, then ∂Φ is maximal monotone from X to X*.
Lemma 1.6 (Pascali and Sburlan [10]).
If A and B are two maximal monotone operators in X such that (intD(A))⋂D(B)≠∅, then A+B is maximal monotone.
Proposition 1.7 (Calvert and Gupta [1]).
Let X=Lp(Ω) and Ω be a bounded domain in RN. For 2≤p<+∞, the duality mapping Jp:Lp(Ω)→Lp′(Ω) defined by Jpu=|u|p-1sgnu∥u∥p2-p, for u∈Lp(Ω), satisfies Condition (I); for 2N/(N+1)<p≤2 and N≥1, the duality mapping Jp:Lp(Ω)→Lp′(Ω) defined by Jpu=|u|p-1sgnu, for u∈Lp(Ω), satisfies Condition (I), where (1/p)+(1/p′)=1.
Lemma 1.8 (see Calvert and Gupta [1]).
Let Ω be a bounded domain in RN and g:Ω×R→R be a function satisfying Caratheodory's conditions such that
g(x,·) is monotonically increasing on R;
the mapping u∈Lp(Ω)→g(x,u(x))∈Lp(Ω) is well defined, where 2N/(N+1)<p<+∞ and N≥1.
Let Jp:Lp(Ω)→Lp'(Ω), (1/p)+(1/p′)=1 be the duality mapping defined by
(1.7)Jpu={|u|p-1sgnu,if2NN+1<p≤2,|u|p-1sgnu∥u∥p2-p,if2≤p<+∞,
for u∈Lp(Ω). Then the mapping B:Lp(Ω)→Lp(Ω) defined by (Bu)(x)=g(x,u(x)), for any x∈Ω satisfies Condition (*).
Theorem 1.9 (Calvert and Gupta [1]).
Let X be a real Banach space with a strictly convex dual X*. Let J:X→X* be a duality mapping on X satisfying Condition (I). Let A,B1:X→2X be accretive mappings such that
either both A,B1 satisfy Condition (*) or D(A)⊂D(B1) and B1 satisfies Condition (*),
A+B1 is m-accretive and boundedly inversely compact.
If B2:X→X be a bounded continuous mapping such that, for any y∈X, there is a constant C(y) satisfying (B2(u+y),Ju)≥-C(y) for any u∈X. Then.
[R(A)+R(B1)]¯⊂R(A+B1+B2)¯.
int[R(A)+R(B1)]⊂intR(A+B1+B2).
2. The Main Results 2.1. Notations and Assumptions of (1.4)
In the next of this paper, we assume 2N/(N+1)<p<+∞, 1≤q, r<+∞ if p≥N, and 1≤q, r≤Np/(N-p) if p<N, where N≥1. We use ∥·∥p, ∥·∥q, ∥·∥r, and ∥·∥1,p,Ω to denote the norms in Lp(Ω), Lq(Ω), Lr(Ω) and W1,p(Ω). Let (1/p)+(1/p')=1, (1/q)+(1/q')=1, and (1/r)+(1/r′)=1.
In (1.4), Ω is a bounded conical domain of a Euclidean space RN with its boundary Γ∈C1, (c.f. [4]). We suppose that the Green's Formula is available. Let |·| denote the Euclidean norm in RN, 〈·,·〉 the Euclidean inner-product and ϑ the exterior normal derivative of Γ. λ is a nonnegative constant.
Let φ:Γ×R→R be a given function such that, for each x∈Γ,
φx=φ(x,·):R→R is a proper, convex, lower-semicontinuous function with φx(0)=0.
βx=∂φx (: subdifferential of φx) is maximal monotone mapping on R with 0∈βx(0) and for each t∈R, the function x∈Γ→(I+μβx)-1(t)∈R is measurable for μ>0.
Let g:Ω×R→R be a given function satisfying Caratheodory's conditions such that for 2N/(N+1)<p<+∞ and N≥1, the mapping u∈Lp(Ω)→g(x,u(x))∈Lp(Ω) is defined. Further, suppose that there is a function T(x)∈Lp(Ω) such that g(x,t)t≥0, for |t|≥T(x), x∈Ω.
2.2. Existence and Uniqueness of the Solution of (1.4)Definition 2.1 (Calvert and Gupta [1]).
Define g+(x)=liminft→+∞g(x,t) and g-(x)=limsupt→-∞g(x,t).
Further, define a function g1:Ω×R→R by
(2.1)g1(x,t)={(infs≥tg(x,s))∧(t-T(x)),∀t≥T(x),0,∀t∈[-T(x),T(x)],(sups≤tg(x,s))∨(t+T(x)),∀t≤-T(x).
Then for all x∈Ω, g1(x,t) is increasing in t and limt→±∞g1(x,t)=g±(x). Moreover, g1:Ω×R→R satisfies Caratheodory's conditions and the functions g±(x) are measurable on Ω. And, if g2(x,t)=g(x,t)-g1(x,t) then g2(x,t)t≥0, for |t|≥T(x), x∈Ω.
Proposition 2.2 (see Calvert and Gupta [1]).
For 2N/(N+1)<p<+∞ and N≥1, define the mapping B1:Lp(Ω)→Lp(Ω) by (B1u)(x)=g1(x,u(x)), for all u∈Lp(Ω) and x∈Ω, then B1 is a bounded, continuous, and m-accretive mapping.
Moreover, Lemma 1.8 implies that B1 satisfies Condition (*).
Define B2:Lp(Ω)→Lp(Ω) by (B2u)(x)=g2(x,u(x)), where g2(x,t)=g(x,t)-g1(x,t), then B2 satisfies the inequality:
(2.2)(B2(u+y),Jpu)≥-C(y),
for any u,y∈Lp(Ω), where C(y) is a constant depending on y and Jp:Lp(Ω)→Lp'(Ω) denotes the duality mapping, where (1/p)+(1/p′)=1.
Lemma 2.3 (Wei and Agarwal [7]).
The mapping Φp:W1,p(Ω)→R defined by
(2.3)Φp(u)=∫Γφx(u|Γ(x))dΓ(x),
for any u∈W1,p(Ω), is a proper, convex, and lower-semicontinuous mapping on W1,p(Ω).
Moreover, Lemma 1.5 implies that ∂Φp, the subdifferential of Φp, is maximal monotone.
Lemma 2.4 (Wei and He [2]).
Let X0 denote the closed subspace of all constant functions in W1,p(Ω). Let X be the quotient space W1,p(Ω)/X0. For u∈W1,p(Ω), define the mapping P:W1,p(Ω)→X0 by Pu=(1/meas(Ω))∫Ωudx. Then, there is a constant C>0, such that for all u∈W1,p(Ω),
(2.4)∥u-Pu∥p≤C∥∇u∥(Lp(Ω))N.
Lemma 2.5.
Define the mapping Bp,q,r:W1,p(Ω)→(W1,p(Ω))* by
(2.5)(v,Bp,q,ru)=∫Ω〈(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u,∇v〉dx+λ∫Ω|u(x)|q-2u(x)v(x)dx+λ∫Ω|u(x)|r-2u(x)v(x)dx,
for any u,v∈W1,p(Ω). Then Bp,q,r is everywhere defined, strictly monotone, hemicontinuous, and coercive.
Proof.
Step 1. Bp,q,r is everywhere defined.
From Lemma 1.3, we know that W1,p(Ω)↪CB(Ω), when p>N. And, W1,p(Ω)↪Lq(Ω), W1,p(Ω)↪Lr(Ω), when p≤N. Thus, for all v∈W1,p(Ω), ∥v∥q≤k1∥v∥1,p,Ω, ∥v∥r≤k2∥v∥1,p,Ω, where k1,k2 are positive constants.
For u,v∈W1,p(Ω), we have
(2.6)|(v,Bp,q,ru)|≤2∫Ω|∇u|p-1|∇v|dx+λ∫Ω|u|q-1|v|dx+λ∫Ω|u|r-1|v|dx≤2∥|∇u|∥pp/p′∥|∇v|∥p+λ∥v∥q∥u∥qq/q′+λ∥v∥r∥u∥rr/r'≤2∥u∥1,p,Ωp/p′∥v∥1,p,Ω+k1′λ∥v∥1,p,Ω∥u∥1,p,Ωq/q′+k2′λ∥v∥1,p,Ω∥u∥1,p,Ωr/r',
where k1′ and k2′ are positive constants. Thus Bp,q,r is everywhere defined.
Step 2. Bp,q,r is strictly monotone.
For u,v∈W1,p(Ω), we have
(2.7)|(u-v,Bp,q,ru-Bp,q,rv)|=∫Ω〈(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u-(1+|∇v|p1+|∇v|2p)|∇v|p-2∇v,∇u-∇v〉dx+λ∫Ω(|u|q-2u-|v|q-2v)(u-v)dx+λ∫Ω(|u|r-2u-|v|r-2v)(u-v)dx=∫Ω{(1+|∇u|p1+|∇u|2p)|∇u|p-(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u∇v-(1+|∇v|p1+|∇v|2p)|∇v|p-2∇u∇v+(1+|∇v|p1+|∇v|2p)|∇v|p}dx+λ∫Ω(|u|q-2u-|v|q-2v)(u-v)dx+λ∫Ω(|u|r-2u-|v|r-2v)(u-v)dx≥∫Ω{(1+|∇u|p1+|∇u|2p)|∇u|p-1-(1+|∇v|p1+|∇v|2p)|∇v|p-1}(|∇u|-|∇v|)dx+λ∫Ω(|u|q-1-|v|q-1)(|u|-|v|)dx+λ∫Ω(|u|r-1-|v|r-1)(|u|-|v|)dx.
If, we let h(t)=(1+(t/1+t2))t(p-1)/p, for t≥0. Then we know that
(2.8)h′(t)=t(p-1)/p(1+t2)3/2+t-(1/p)(1+t1+t2)p-1p≥0,
since t≥0. And, h′(t)=0 if and only if t=0. Then h(t) is strictly monotone. Thus we can easily know that Bp,q,r is strictly monotone.
Step 3. Bp,q,r is hemicontinuous.
In fact, it suffices to show that, for any u,v,w∈W1,p(Ω) and t∈[0,1], (w,Bp,q,r(u+tv)-Bp,q,ru)→0, as t→0.
By Lebesque's dominated convergence theorem, it follows that
(2.9)0≤limt→0|(w,Bp,q,r(u+tv)-Bp,q,ru)|≤∫Ωlimt→0|(1+|∇u+t∇v|p1+|∇u+t∇v|2p)|∇u+t∇v|p-2(∇u+t∇v)-(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u||∇w|dx+λ∫Ωlimt→0||u+tv|q-2(u+tv)-|u|q-2u||w|dx+λ∫Ωlimt→0||u+tv|r-2(u+tv)-|u|r-2u||w|dx=0,
and hence Bp,q,r is hemicontinuous.
Step 4. Bp,q,r is coercive.
Now, for u∈W1,p(Ω), Lemma 2.4 implies that ∥u∥1,p,Ω→∞ is equivalent to ∥u-(1/meas(Ω))∫Ωudx∥1,p,Ω→∞ and hence we have the following result:
(2.10)(u,Bp,q,ru)∥u∥1,p,Ω=∫Ω(1+(|∇u|p/1+|∇u|2p))|∇u|pdx∥u∥1,p,Ω+λ∫Ω|u|qdx∥u∥1,p,Ω+λ∫Ω|u|rdx∥u∥1,p,Ω=∫Ω(|∇u|p+1+|∇u|2p)dx-∫Ω(1/1+|∇u|2p)dx∥u∥1,p,Ω+λ∫Ω|u|qdx∥u∥1,p,Ω+λ∫Ω|u|rdx∥u∥1,p,Ω≥2∫Ω|∇u|pdx-∫Ω(1/1+|∇u|2p)dx∥u∥1,p,Ω+λ∫Ω|u|qdx∥u∥1,p,Ω+λ∫Ω|u|rdx∥u∥1,p,Ω→+∞,
as ∥u∥1,p,Ω→+∞, which implies that Bp,q,r is coercive.
This completes the proof.
Remark 2.6.
Lemma 2.5 is a key result for later use.
Definition 2.7.
Define a mapping Ap:Lp(Ω)→2Lp(Ω) as follows:
(2.11)D(Ap)={u∈Lp(Ω)∣thereexistsanf∈Lp(Ω),suchthatf∈Bp,q,ru+∂Φp(u)}.
For u∈D(Ap), let Apu={f∈Lp(Ω)∣f∈Bp,q,ru+∂Φp(u)}.
Proposition 2.8.
The mapping Ap:Lp(Ω)→2Lp(Ω) is m-accretive.
Proof.
Step 1. Ap is accretive.
Case 1. If p≥2, the duality mapping Jp:Lp(Ω)→Lp′(Ω) is defined by Jpu=|u|p-1sgnu∥u∥p2-p for u∈Lp(Ω). It then suffices to prove that for any ui∈D(Ap) and vi∈Apui,i=1,2,
(2.12)(v1-v2,Jp(u1-u2))≥0.
To this, we are left to prove that both
(2.13)(|u1-u2|p-1sgn(u1-u2)∥u1-u2∥p2-p,Bp,q,ru1-Bp,q,ru2)≥0,(|u1-u2|p-1sgn(u1-u2)∥u1-u2∥p2-p,∂Φp(u1)-∂Φp(u2))≥0
are available.
Now take for a constant k>0, χk:R→R is defined by χk(t)=|(t∧k)∨(-k)|p-1sgnt. Then χk is monotone, Lipschitz with χk(0)=0 and χk′ is continuous except at finitely many points on R. This gives that
(2.14)(|u1-u2|p-1sgn(u1-u2)∥u1-u2∥p2-p,∂Φp(u1)-∂Φp(u2))=limk→+∞∥u1-u2∥p2-p(χk(u1-u2),∂Φp(u1)-∂Φp(u2))≥0.
Also,
(2.15)(|u1-u2|p-1sgn(u1-u2)∥u1-u2∥p2-p,Bp,q,ru1-Bp,q,ru2)=∥u1-u2∥p2-p×limk→+∞∫Ω〈(1+|∇u1|p1+|∇u1|2p)|∇u1|p-2∇u1-(1+|∇u2|p1+|∇u2|2p)|∇u2|p-2∇u2,∇u1-∇u2(1+|∇u1|p1+|∇u1|2p)〉×χk′(u1-u2)dx+λ∥u1-u2∥p2-p∫Ω(|u1|q-2u1-|u2|q-2u2)|u1-u2|p-1sgn(u1-u2)dx+λ∥u1-u2∥p2-p∫Ω(|u1|r-2u1-|u2|r-2u2)|u1-u2|p-1sgn(u1-u2)dx≥0,
the last inequality is available since χk is monotone and χk(0)=0.
Case 2. If 2N/(N+1)<p<2, the duality mapping Jp:Lp(Ω)→Lp′(Ω) is defined by Jpu=|u|p-1sgnu, for u∈Lp(Ω). It then suffices to prove that for any ui∈D(Ap) and vi∈Apui,i=1,2,
(2.16)(v1-v2,Jp(u1-u2))≥0.
To this, we define the function χn:R→R by
(2.17)χn(t)={|t|p-1sgnt,if|t|≥1n,(1n)p-2t,if|t|≤1n.
Then χn is monotone, Lipschitz with χn(0)=0 and χn′ is continuous except at finitely many points on R. So (χn(u1-u2),∂Φp(u1)-∂Φp(u2))≥0.
Then, for ui∈D(Ap), vi∈Apui,i=1,2, we have(2.18)(v1-v2,Jp(u1-u2))=(|u1-u2|p-1sgn(u1-u2),Bp,q,ru1-Bp,q,ru2)+(|u1-u2|p-1sgn(u1-u2),∂Φp(u1)-∂Φp(u2))=(|u1-u2|p-1sgn(u1-u2),Bp,q,ru1-Bp,q,ru2)+limn→∞(χn(u1-u2),∂Φp(u1)-∂Φp(u2))≥0.
Step 2. R(I+μAp)=Lp(Ω), for every μ>0.
First, define the mapping Ip:W1,p(Ω)→(W1,p(Ω))* by Ipu=u and (v,Ipu)(W1,p(Ω))*×W1,p(Ω)=(v,u)L2(Ω) for u,v∈W1,p(Ω), where (·,·)L2(Ω) denotes the inner product of L2(Ω). Then Ip is maximal monotone [7].
Secondly, for any μ>0, define the mapping Tμ:W1,p(Ω)→2(W1,p(Ω))* by Tμu=Ipu+μBp,q,ru+μ∂Φp(u), for u∈W1,p(Ω). Then similar to that in [7], by using Lemmas 1.4, 1.6, 2.3, and 2.5, we know that Tμ is maximal monotone and coercive, so that R(Tμ)=(W1,p(Ω))*, for any μ>0.
Therefore, for any f∈Lp(Ω), there exists u∈W1,p(Ω), such that
(2.19)f=Tμu=u+μBp,q,ru+μ∂Φp(u).
From the definition of Ap, it follows that R(I+μAp)=Lp(Ω), for all μ>0.
This completes the proof.
Lemma 2.9.
The mapping Ap:Lp(Ω)→2Lp(Ω) has a compact resolvent for 2N/(N+1)<p<2 and N≥1.
Proof.
Since Ap is m-accretive by Proposition 2.8, it suffices to prove that if u+μApu=f(μ>0) and {f} is bounded in Lp(Ω), then {u} is relatively compact in Lp(Ω). Now define functions χn,ξn:R→R by
(2.20)χn(t)={|t|p-1sgnt,if|t|≥1n(1n)p-2t,if|t|≤1n,ξn(t)={|t|2-(2/p)sgnt,if|t|≥1n,(1n)1-(2/p)t,if |t|≤1n.
Noticing that χn′(t)=(p-1)×(p′/2)p×(ξn′(t))p, for|t|≥1/n, where(1/p)+(1/p′)=1 and χn′(t)=(ξn′(t))p, for|t|≤1/n. We know that (χn(u),∂Φp(u))≥0 for u∈W1,p(Ω) since χn is monotone, Lipschitz with χn(0)=0, and χn′ is continuous except at finitely many points on R. Then
(2.21)(|u|p-1sgnu,Apu)=limn→∞(χn(u),Apu)≥limn→∞(χn(u),Bp,q,ru)=limn→∞∫Ω(1+|∇u|p1+|∇u|2p)|∇u|pχn′(u)dx+λlimn→∞∫Ω|u|q-2uχn(u)dx+λlimn→∞∫Ω|u|r-2uχn(u)dx≥limn→∞∫Ω|∇u|pχn′(u)dx≥const·limn→∞∫Ω|grad(ξn(u))|pdx≥const∫Ω|grad(|u|2-(2/p)sgnu)|pdx.
We now have from f=u+μApu that
(2.22)∥f∥p∥|u|2-(2/p)sgnu∥p2/2(p-1)p2/2(p-1)p′≥(|u|p-1sgnu,f)=(|u|p-1sgnu,u)+μ(|u|p-1sgnu,Apu)≥∥|u|2-(2/p)sgnu∥p2/2(p-1)p2/2(p-1)+μ·const·∥grad |u|2-(2/p)sgnu∥pp,
which gives that
(2.23)∥|u|2-(2/p)sgnu∥pp/2(p-1)≤∥|u|2-(2/p)sgnu∥p2/2(p-1)p/2(p-1)∥f∥p≤const,
in view of the fact that p<p2/2(p-1) when 2N/(N+1)<p<2 for N≥1. Again from (2.22), we have ∥grad(|u|2-(2/p)sgnu)∥p≤const. Hence {f} bounded in Lp(Ω) implies that {|u|2-(2/p)sgnu} is bounded in W1,p(Ω).
We notice that W1,p(Ω)↪↪Lp2/2(p-1)(Ω) when N≥2 and W1,p(Ω)↪↪CB(Ω) when N=1 by Lemma 1.3, hence{|u|2-(2/p)sgnu} is relatively compact in Lp2/2(p-1)(Ω). This gives that {u} is relatively compact in Lp(Ω) since the Nemytskii mapping u∈Lp2/2(p-1)(Ω)→|u|p/2(p-1)sgnu∈Lp(Ω) is continuous.
This completes the proof.
Remark 2.10.
Since Φp(u+α)=Φp(u), for any u∈W1,p(Ω) and α∈C0∞(Ω), we have f∈Apu implies that f=Bp,q,ru in the sense of distributions.
Proposition 2.11.
For f∈Lp(Ω), if there exists u∈Lp(Ω) such that f∈Apu, then u is the unique solution of (1.4).
Proof.
First, we show that
(2.24)-div[(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u]+λ|u|q-2u+λ|u|r-2u=f(x),a.e.x∈Ω
is available.
Now f∈Apu implies that f=Bp,q,ru+∂Φp(u). For all φ∈C0∞(Ω), by Remark 2.10, we have
(2.25)(φ,f)=(φ,Bp,q,ru+∂Φp(u))=(φ,Bp,q,ru)=∫Ω〈(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u,∇φ〉dx+λ∫Ω|u|q-2uφdx+λ∫Ω|u|r-2uφdx=∫Ω-div[(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u]φdx+λ∫Ω|u|q-2uφdx+λ∫Ω|u|r-2uφdx,
which implies that (2.24) is true.
Secondly, we show that
(2.26)-〈ϑ,(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u〉∈βx(u(x)),a.e.x∈Γ.
We will prove (2.26) under the additional condition |βx(u)|≤a|u|p/p′+b(x), where b(x)∈Lp′(Γ) and a∈R. Refer to the result of Brezis [11] for the general case.
Now, from (2.24), f∈Apu implies that f(x)=-div[(1+|∇u|p/1+|∇u|2p)|∇u|p-2∇u]+λ|u(x)|q-2u(x)+λ|u|r-2u∈Lp(Ω). By using Green's Formula, we have that for any v∈W1,p(Ω),
(2.27)∫Γ〈ϑ,(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u〉v|ΓdΓ(x)=∫Ωdiv[(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u]vdx+∫Ω〈(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u,∇v〉dx.
Then -〈ϑ,(1+|∇u|p/1+|∇u|2p)|∇u|p-2∇u〉∈W-(1/p),p′(Γ)=(W1/p,p(Γ))*, where W1/p,p(Γ) is the space of traces of W1,p(Ω).
Now let the mapping B:Lp(Γ)→Lp′(Γ) be defined by Bu=g(x), for any u∈Lp(Γ), where g(x)=βx(u(x)) a.e. on Γ. Clearly, B=∂Ψ where Ψ(u)=∫Γφx(u(x))dΓ(x) is a proper, convex, and lower-semicontinuous function on Lp(Γ). Now define the mapping K:W1,p(Ω)→Lp(Γ) by K(v)=v|Γ for any v∈W1,p(Ω). Then K*BK:W1,p(Ω)→(W1,p(Ω))* is maximal monotone since both K,B are continuous. Finally, for any u,v∈W1,p(Ω), we have
(2.28)Ψ(Kv)-Ψ(Ku)=∫Γ[φx(v|Γ(x))-φx(u|Γ(x))]dΓ(x)≥∫Γβx(u|Γ(x))(v|Γ(x)-u|Γ(x))dΓ(x)=(BKu,Kv-Ku)=(K*BKu,v-u).
Hence we get K*BK⊂∂Φp and so K*BK=∂Φp. Therefore, we have
(2.29)-〈ϑ,(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u〉∈βx(u(x)),a.e.onΓ.
Finally, we will show that u is unique.
If f∈Apu and f∈Apv, where u,v∈D(Ap). Then
(2.30)0≤(u-v,Bp,q,ru-Bp,q,rv)=(u-v,∂Φp(v)-∂Φp(u))≤0,
since Bp,q,r is strictly monotone and ∂Φp is maximal monotone, which implies that u(x)=v(x).
This completes the proof.
Remark 2.12.
If βx≡0 for any x∈Γ, then ∂Φp(u)≡0, for all u∈W1,p(Ω).
Proposition 2.13.
If βx≡0 for any x∈Γ, then {f∈Lp(Ω)∣∫Ωfdx=0}⊂R(Ap).
Proof.
We can easily know that R(Bp,q,r)=(W1,p(Ω))* in view of Lemmas 1.4 and 2.5. Note that for any f∈Lp(Ω) with ∫Ωfdx=0, the linear function u∈W1,p(Ω)→∫Ωfudx is an element of (W1,p(Ω))*. So there exists a u∈W1,p(Ω) such that
(2.31)∫Ωfvdx=∫Ω〈(1+|∇u|p1+|∇u|2p)|∇u|p-2∇u,∇v〉dx+λ∫Ω|u|q-2uvdx+λ∫Ω|u|r-2uvdx,
for any v∈W1,p(Ω). So f=Apu in view of Remark 2.12.
This completes the proof.
Definition 2.14 (see [1, 7]).
For t∈R, x∈Γ, let βx0(t)∈βx(t) be the element with least absolute value if βx(t)≠∅ and βx0(t)=±∞, where t>0 or <0, respectively, in case βx(t)=∅. Finally, let β±(x)=limt→±∞βx0(t) (in the extended sense) for x∈Γ. β±(x) define measurable functions on Γ, in view of our assumptions on βx.
Proposition 2.15.
Let f∈Lp(Ω) such that
(2.32)∫Γβ-(x)dΓ(x)<∫Ωfdx<∫Γβ+(x)dΓ(x).
Then f∈
Int
R(Ap).
Proof.
Let f∈Lp(Ω) and satisfy (2.32), by Proposition 2.8, there exists un∈Lp(Ω) such that, for each n≥1, f=(1/n)un+Apun. In the same reason as that in [1], we only need to prove that ∥un∥p≤const, for all n≥1.
Indeed, suppose to the contrary that 1≤∥un∥p→∞, as n→∞. Let vn=un/∥un∥p. Let ψ:R→R be defined by ψ(t)=|t|p, ∂ψ:R→R be its subdifferential and for μ>0, ∂ψμ:R→R denote the Yosida-approximation of ∂ψ. Let θμ:R→R denote the indefinite integral of [(∂ψμ)′]1/p with θμ(0)=0 so that (θμ′)p=(∂ψμ)′. In view of Calvert and Gupta [1], we have
(2.33)(∂ψμ(vn),∂Φp(un))≥∫Γβx((1+μ∂ψ)-1(un|Γ(x)))×∂ψμ(vn|Γ(x))dΓ(x)≥0.
Now multiplying the equation f=(1/n)un+Apun by ∂ψμ(vn), we get that
(2.34)(∂ψμ(vn),f)=(∂ψμ(vn),1nun)+(∂ψμ(vn),Bp,q,run)+(∂ψμ(vn),∂Φp(un)).
Since ∂ψμ(0)=0, it follows that (∂ψμ(vn),un)≥0. Also, we can know that
(2.35)(∂ψμ(vn),Bp,q,run)=∫Ω〈(1+|∇un|p1+|∇un|2p)|∇un|p-2∇un,∇vn〉(∂ψμ)′(vn)dx+λ∫Ω|un|q-2un∂ψμ(vn)dx+λ∫Ω|un|r-2un∂ψμ(vn)dx≥∫Ω|∇u|p∥un∥p(∂ψμ)′(vn)dx=∥un∥pp-1∫Ω|grad(θμ(vn))|pdx.
Then we get from (2.33) that
(2.36)∥un∥pp-1∫Ω|grad(θμ(vn))|pdx+∫Γβx((1+μ∂ψ)-1(un|Γ(x)))×∂ψμ(vn|Γ(x))dΓ(x)≤(∂ψμ(vn),f).
Since |∂ψμ(t)|≤|∂ψ(t)| for any t∈R and μ>0, we see from ∥vn∥p=1 for n≥1, that ∥∂ψμ(vn)∥p′≤C, for μ>0, where C is a constant which does not depend on n or μ and (1/p)+(1/p′)=1.
From (2.36), we have
(2.37)∫Ω|grad(θμ(vn))|pdx≤C∥un∥pp-1,forμ>0,n≥1.
Now we easily know that (θμ′)p=(∂ψμ)′→(∂ψ)′ as μ→0a.e. on R.
Letting μ→0, we see from Fatou's lemma and (2.37) that
(2.38)∫Ω|grad(|vn|2-(2/p)sgnvn)|pdx≤C∥un∥pp-1.
From (2.38), we know that |vn|2-(2/p)sgnvn→k (a constant) in Lp(Ω), as n→+∞.
Next we will show that k≠0 is in Lp(Ω) from two aspects.
If p≥2, since ∥|vn|2-(2/p)sgnvn∥p=∥vn∥2p-22-(2/p)≥∥vn∥p2-(2/p)=1, it follows that k≠0 in Lp(Ω),
if 2N/(N+1)<p<2, ∥|vn|2-(2/p)sgnvn∥p=∥vn∥2p-22-(2/p)≥∥vn∥p2-(2/p)=1, then {|vn|2-(2/p)sgnvn} is bounded in W1,p(Ω). By Lemma 1.3, W1,p(Ω)↪↪CB(Ω) when N=1 and W1,p(Ω)↪↪Lp2/2(p-1)(Ω), when N≥2. So {|vn|2-(2/p)sgnvn} is relatively compact in Lp2/2(p-1)(Ω). Then there exists a subsequence of {|vn|2-(2/p)sgnvn}, for simplicity, we denote it by {|vn|2-(2/p)sgnvn}, satisfying |vn|2-(2/p)sgnvn→g in Lp2/2(p-1)(Ω). Noticing that p≤p2/2(p-1) when 2N/(N+1)<p<2, it follows that k=ga.e. on Ω. Now,
(2.39)1=∥vn∥pp=∫Ω||vn|2-(2/p)sgnvn|p2/2(p-1)dx≤const∫Ω||vn|2-(2/p)sgnvn-g|p2/2(p-1)dx+const∥g∥p2/2(p-1)p2/2(p-1),
it follows that g≠0 in Lp(Ω) and then k≠0 in Lp(Ω). Assume, now, k>0, we see from (2.36) that
(2.40)∫Γβx((1+μ∂ψ)-1(un|Γ(x)))×∂ψμ(vn|Γ(x))dΓ(x)≤(∂ψμ(vn),f).
Choosing a subsequence so that un|Γ(x)→+∞ a.e. on Γ, we see letting n→+∞ that ∫Γβ+(x)dΓ(x)≤∫Ωf(x)dx, which is a contradiction with (2.32). Similarly, if k<0, it also leads to a contradiction. Thus f∈intR(Ap).
This completes the proof.
Proposition 2.16.
Ap+B1:Lp(Ω)→Lp(Ω) is m-accretive and has a compact resolvent.
Proof.
Using a theorem in Corduneanu [12], we know that Ap+B1:Lp(Ω)→Lp(Ω) is m-accretive.
To show that Ap+B1:Lp(Ω)→Lp(Ω) has a compact resolvent, we only need to prove that if w∈Apu+B1u with {w} and {u} being bounded in Lp(Ω), then {u} is relatively compact in Lp(Ω). Now we discuss it from two aspects.
If p≥2, since
(2.41)∫Ω|∇u|pdx≤(u,Bp,q,ru)=(u,Apu)-(u,∂Φp(u))≤(u,Apu)+(u,B1u)=(u,w)≤∥u∥p∥u∥p′≤const,
it follows that {u} is bounded in W1,p(Ω), where (1/p)+(1/p')=1. Then {u} is relatively compact in Lp(Ω) since W1,p(Ω)↪↪Lp(Ω);
if 2N/(N+1)<p<2, since w∈Apu+B1u with {w} and {u} being bounded in Lp(Ω), we have w-B1u∈Apu with {w-B1u} and {u} being bounded in Lp(Ω) which gives that {u} is relatively compact in Lp(Ω) since Ap is m-accretive by Proposition 2.8 and has a compact resolvent by Lemma 2.9.
This completes the proof.
Theorem 2.17.
Let f∈Lp(Ω) be such that
(2.42)∫Γβ-(x)dΓ(x)+∫Ωg-(x)dx<∫Ωf(x)dx<∫Γβ+(x)dΓ(x)+∫Ωg+(x)dx,
then (1.4) has a unique solution in Lp(Ω), where 2N/(N+1)<p<+∞ and N≥1.
Proof.
We want to use Theorem 1.9 to finish our proof. From Propositions 1.7, 2.2, 2.8, and 2.16, we can see that all of the conditions in Theorem 1.9 are satisfied. It then suffices to show that f∈int[R(Ap)+R(B1)] which ensures that f∈R(Ap+B1+B2). Thus Proposition 2.11 tells us (1.4) has a unique solution in Lp(Ω).
Using the similar methods as those in [2, 4, 7], by dividing it into two cases and using Propositions 2.13 and 2.15, respectively, we know that f∈int[R(Ap)+R(B1)].
This completes the proof.
Remark 2.18.
Compared to the work done in [1–7], not only the existence of the solution of (1.4) is obtained but also the uniqueness of the solution is obtained.
Acknowledgments
This work is supported by the National Natural Science Foundation of China (Grant no. 11071053), the Natural Science Foundation of Hebei Province (Grant no. A2010001482), the Key Project of Science and Research of Hebei Education Department (Grant no. ZH2012080), and the Youth Project of Science and Research of Hebei Education Department (Grant no. Q2012054 ).
CalvertB. D.GuptaC. P.Nonlinear elliptic boundary value problems in Lp-spaces and sums of ranges of accretive operators19782112610.1016/0362-546X(78)90038-X512651WeiL.HeZ.The applications of sums of ranges of accretive operators to nonlinear equations involving the p-Laplacian operator199524218519310.1016/0362-546X(94)E0051-H1312589WeiL.Existence of solutions of nonlinear elliptic boundary value problems2001, (Chinese)3133603641845781WeiL.HenZ.The applications of theories of accretive operators to nonlinear elliptic boundary value problems in Lp-spaces200146219921110.1016/S0362-546X(99)00457-51849790WeiL.ZhouH. Y.Existence of solutions of a family of nonlinear boundary value problems in L2-spaces200520217518210.1007/s11766-005-0050-42144744WeiL.ZhouH. Y.The existence of solutions of nonlinear boundary value problems involving the p-Laplacian operator in Ls-spaces20051845115212172435WeiL.AgarwalR. P.Existence of solutions to nonlinear Neumann boundary value problems with generalized p-Laplacian operator200856253054110.1016/j.camwa.2008.01.0132442671ChenZ. C.LuoT.The eigenvalue problem for p-Laplacian-like equations20034646316382010751LiL. K.GuoY. T.1981Shanghai, ChinaShanghai Science and Technology PressPascaliD.SburlanS.1978The Hague, The NetherlandsMartinus Nijhoff531036BrezisH.Integrales convexes dans les espaces de Sobolev1972131-29232-s2.0-5124918818610.1007/BF02760227CorduneanuA.Some remarks on the sum of two m-dissipative mappings19752044114140380523