INVOLVING DUALITY MAPPINGS VIA THE MOUNTAIN PASS THEOREM
JENIC ˘A CRˆINGANU
We derive existence results for operator equations having the form Jϕu=Nfu,
by using the Mountain Pass Theorem. HereJϕ is a duality mapping on a real reflexive and smooth Banach space, compactly imbedded in aLq-space, andNf
is the Nemytskii operator generated by a Carath´eodory functionfwhich satisfies some appropriate growth conditions.
AMS 2000 Subject Classification: 35J20, 35J60.
Key words: Duality mapping, Mountain Pass Theorem, Nemytskii operator, p- Laplacian.
1. INTRODUCTION
In [9] the existence of the weak solution in W01,p(Ω), 1 < p < +∞, for the Dirichlet problem
−∆pu=f(x, u) in Ω⊂RN, (1.1)
u= 0 on∂Ω (1.2)
was obtained using (among other methods) the Mountain Pass Theorem.
It is well known that the operator −∆p : W01,p(Ω) → W−1,p0(Ω) is in fact the duality mapping on W01,p(Ω) corresponding to the gauge function ϕ(t) = tp−1. In this paper we generalize the results from [9] by considering operator equations of the form
(1.3) Jϕu=Nfu,
where, instead of the special duality mapping (−∆p), we consider the case of an arbitrary duality mapping. More precisely, we consider equation (1.1) in the functional framework below.
MATH. REPORTS11(61),1 (2009), 11–20
(H1) Xis a real reflexive and smooth Banach space, compactly imbedded in Lq(Ω), where 1< q <+∞and Ω⊂RN,N ≥2, is a bounded domain with smooth boundary.
(H2) Jϕ :X→X∗ is the duality mapping corresponding to the gauge func- tion ϕ:R+→R+; we assume that Jϕ is continuous and satisfies the (S+) condition: ifxn* x(weakly) inXand lim sup
n→∞
hJϕxn, xn−xi ≤0, thenxn→x (strongly) inX.
(H3) Nf :Lq(Ω) → Lq0(Ω), 1q + q10 = 1, defined by (Nfu)(x) =f(x, u(x)), for u ∈ Lq(Ω), x ∈ Ω, is the Nemytskii operator generated by the Carath´eodory function f : Ω×R → R, which satisfies the growth condition
(1.4) |f(x, s)| ≤c|s|q−1+b(x), x∈Ω, s∈R, wherec >0 is constant and b∈Lq0(Ω).
By solution of equation (1.3) we mean an elementu∈X which satisfies
(1.5) Jϕu= (i0Nfi)u,
where i is the compact imbedding of X into Lq(Ω) and i0 : Lq0(Ω)→ X∗ is its adjoint.
Notice that by (1.4) the operator Nf is well-defined, continuous and bounded from Lq(Ω) into Lq0(Ω) (see, e.g., [8]), so that i0Nfi is also well- defined and compact from X intoX∗.
Moreover, the functionalψ:Lq(Ω)→R, defined byψ(u) =R
ΩF(x, u)dx, whereF(x, s) =Rs
0 f(x, t) dtisC1 onLq(Ω), hence onX and ψ0(u) =Nfu for all u∈X.
Notice also thatF is a Carath´eodory function and there exists a constant c1>0 and a function c∈Lq(Ω), c≥0, such that
(1.6) |F(x, s)| ≤c1|s|q+c(x), x∈Ω, s∈R.
Let us remark that (1.5) is equivalent to (1.7) hJϕu, vi=hNf(iu), iviLq(Ω),Lq0
(Ω)= Z
Ω
f(x, u)vdx, v∈X.
By Asplund’s theorem, Jϕu = ∂Φ(u) for each u ∈ X, where Φ(u) = Rkuk
0 ϕ(t)dt and ∂Φ : X → P(X∗) is the subdifferential of Φ in the sense of convex analysis, i.e.,
∂Φ(u) ={x∗∈X∗: Φ(v)−Φ(u)≥ hx∗, v−ui for allv∈X}
(see, e.g., [4], [5] or [12]). Since Φ is convex and X is smooth, Φ is Gˆateaux differentiable on X and ∂Φ(u) = {Φ0(u)} for all u ∈ X. So, Jϕu = Φ0(u)
for all u ∈ X and, by continuity of Jϕ (see (H2)), we have Φ∈ C1(X,R).
Consequently, the functional F :X →R defined by F(u) = Φ(u)−Ψ(u) =
Z kuk
0
ϕ(t)dt− Z
Ω
F(x, u)dx, u∈X,
is C1 on X and F0(u) = Φ0(u)−Ψ0(u) =Jϕu−Nfu for allu∈X. Therefore, u ∈X is a solution of equation (1.3) if and only if u is a critical point forF, i.e., F0(u) = 0.
To prove thatF has at least a critical point in X we use the Mountain Pass Theorem, which we recall below.
Theorem 1.1. Let X be a real Banach space and assumeI ∈ C1(X,R) satisfies the Palais-Smale (PS) condition. Suppose I(0) = 0 and that
(i) there are constantsρ, α >0 such that I|kuk=ρ≥α;
(ii)there is an element e∈X, kek> ρ such thatI(e)≤0.
Then I possesses a critical value c≥α. Moreover, c= inf
g∈Γ max
u∈g([0,1])I(u)
whereΓ ={g∈ C([0,1], X) :g(0) = 0, g(1) =e}. (It is obvious that each criti- cal point u at level c (I0(u) = 0, I(u) =c) is a nontrivial one.)
For theproof see, e.g., [2], [11] or [13].
Let us recall that the functional I ∈ C1(X,R) is said to satisfy the (PS) condition if any sequence (un) ⊂ X for which (I(un)) is bounded and I0(un)→0 as n→ ∞, possesses a convergent subsequence.
2. THE MAIN EXISTENCE RESULT
First, we establish some preliminary results.
Proposition 2.1. If (un) ⊂X is bounded and F0(un)→ 0 as n→ ∞, then (un) has a convergent subsequence.
Proof. This result is proved in [9], Lemma 2.
Theorem 2.1. Assume that there are θ >1 and s0>0 such that (2.1) θF(x, s)≤sf(x, s), x∈Ω, |s| ≥s0,
(2.2) lim inf
t→∞
Rt
0ϕ(s)ds tϕ(t) > 1
θ. Then F satisfies the (PS) condition.
Proof. According to Proposition 2.1 it is sufficient to show that any sequence (un)⊂Xfor which (F(un)) is bounded andF0(un)→0, is bounded.
As in the proof of Theorem 15 in [9], there exists a constant k > 0 such that
(2.3) Φ(un)−1
θ Z
Ω
f(x, un) undx≤k.
By F0(un) → 0, as in the proof of the same theorem, there exists n0 ∈ N such that
hJϕun, uni − Z
Ω
f(x, un)un
≤ kunk, (∀) n≥n0, which implies
(2.4) −1
θϕ(kunk)kunk+1 θ
Z
Ω
f(x, un)undx≤ 1
θkunk, (∀) n≥1.
From (2.3) and (2.4) we obtain Φ(un)− 1
θϕ(kunk)kunk −1
θkunk ≤k, (∀) n≥n0, so that
Z kunk
0
ϕ(s)ds−1
θϕ(kunk)kunk −1
θkunk ≤k, (∀) n≥n0
or, equivalently,
(2.5) ϕ(kunk)kunk
" Rkunk
0 ϕ(s)ds ϕ(kunk)kunk−1
θ − 1
θϕ(kunk)
#
≤k.
By condition (2.2) there are constants γ > 1θ and t0 >0 such that Rt
0ϕ(s)ds
tϕ(t) ≥γ, t≥t0.
If kunk → ∞there existsn1 ∈Nsuch that kunk ≥t0 forn≥n1 and then Rkunk
0 ϕ(s)ds
kunkϕ(kunk) ≥γ, n≥n1. From (2.5) we obtain
ϕ(kunk)kunk
γ−1
θ − 1
θϕ(kunk)
≤k, n≥n1.
a contradiction (because γ > 1θ and ϕ(kunk)→ ∞).
Consequently, (un) is bounded inX.
Theorem 2.2. Assume that there exists θ >1 such that
(2.6) lim
t→∞
Rt
0ϕ(s)ds tθ = 0 and either
(i) there existss1 >0 such that
(2.7) 0< θF(x, s)≤sf(x, s), x∈Ω, s≥s1, or
(ii)there exists s1<0 such that
(2.8) 0< θF(x, s)≤sf(x, s), x∈Ω, s≤s1. ThenF is unbounded from below.
Proof. We are going to prove the sufficiency of condition (i) (similar arguments may be used if (ii) holds).
Let u ∈ X, u > 0 (in fact i(u) > 0) be such that meas (M1(u)) > 0, where
M1(u) ={x∈Ω :u(x)≥s1} (in fact i(u)(x)≥s1).
We shall prove that F(λu)→ −∞ asλ→ ∞.
Forλ≥1, from the proof of Theorem 16 in [9] we have Z
Ω
F(x, λu)dx≥λθk1(u)−k2, where k1(u)>0, k2>0 are constants. Therefore,
F(λu) = Φ(λu)− Z
Ω
F(x, λu)dx≤Φ(λu)−λθk1(u) +k2
=λθ
" Rλkuk 0 ϕ(t)dt
λθ −k1(u)
# +k2.
By condition (2.6) there are constants α∈(0, k1(u)) andt0>0 such that Rt
0ϕ(s)ds
tθ kukθ≤α, t≥t0, and then
F(λu)≤λθ
" Rλkuk 0 ϕ(t)dt
(λkuk)θ kukθ−k1(u)
#
+k2≤λθ(α−k1(u)) +k2 for λkuk ≥t0. Sinceα−k1(u)<0, we haveF(λu)→ −∞asλ→ ∞.
Remark 2.1. By Theorem 2.2, sinceF is unbounded from below, for each ρ >0 there existse∈X with kek> ρsuch that F(e)≤0.
Now we are in a position to prove the main result of this section.
Theorem2.3. Assume that hypotheses(H1), (H2)and(H3)hold. More- over, assume that
(i) there areθ >1 and s0 >0 such that
(2.9) 0< θF(x, s)≤sf(x, s), x∈Ω,|s| ≥s0; (ii) lim inf
t→∞
Rt
0ϕ(s)ds tϕ(t) > 1
θ; (iii) lim
t→∞
Rt
0ϕ(s)ds tθ = 0;
(iv)there are r1, r2 ∈(1, ∞), r1 < r2, andc0 >0 such that F(u)≥c2kukrX1 −c3kukrX2 (∀) u∈X with kukX < c0, where c2, c3 >0 are constants.
Then equation (1.3)has at least a nontrivial solution in X.
Proof. Let us notice that condition (iv) is suggested by [10].
It is sufficient to show that F has at least a nontrivial critical point u∈X. To do it, we shall use Theorem 1.1. Clearly,F(0) = 0. By (i), (ii) and Theorem 2.1, F satisfies the (PS) condition. By (iv) we have
F(u)≥ kukrX1 c2−c3kukrX2−r1
for allu∈X,withkukX < c0,hence there are constantsρ, α >0,ρ < c0 such that F(u)≥α >0 provided thatkukX =ρ is small enough.
Finally, from (i), (iii) and Theorem 2.2 (see also Remark 2.1) there is an element e∈X,kek> ρ, such thatF(e)≤0, and the proof is complete.
Remark 2.2. A sufficient condition in order to have (iv) is the existence of real numbers r1, r2, 1< r1< r2,such that
(iv1) lim
s→0
Rs
0 ϕ(t)dt
|s|r1 >0, (iv2) lim
s→0
f(x, s)
|s|r2−1 <∞.
Indeed, by (iv1), Φ(u)≥c2kukr1 provided thatkuk>0 is small enough while, by (iv2),
Z
Ω
F(x, u)dx
≤c3kukr2,
provided that kuk>0 is small enough (here, c2, c3 >0 are constants). Con- sequently, F(u)≥c2kukr1 −c3kukr2 forkuk>0 sufficiently small.
3. EXAMPLES
Example 3.1 Consider the Dirichlet problem (1.1), (1.2), where 1< p <
∞, Ω ⊂ RN, N ≥ 2, is a bounded domain with smooth boundary, ∆pu = div(|∇u|p−2∇u) = PN
i=1 ∂
∂xi
|∇u|p−2∂x∂u
i
is the so-called p-Laplacian and f : Ω×R→Ris a Carath´eodory function which satisfies the growth condition (3.1) |f(x, s)| ≤c |s|q−1+ 1
, x∈Ω, s∈R, with c≥0 constant, 1< q < p∗=
( N p
N−p ifN > p, +∞ ifN ≤p.
Let us remark that thep-Laplacian operator−∆p :W01,p(Ω)→W−1,p0(Ω) defined by
−∆pu=−
N
X
i=1
∂
∂xi
|∇u|p−2 ∂u
∂xi
, u∈W01,p(Ω), or, equivalently,
h−∆pu, vi= Z
Ω
|∇u|p−2∇u ∇v, u, v∈W01,p(Ω), is the duality mapping
Jϕ :W01,p(Ω)→W−1,p0(Ω)
corresponding to the gauge function ϕ(t) =tp−1 (see, e.g., [9] or [12]).
On the other hand, the Nemytskii operatorNf is continuous and bounded from Lq(Ω) intoLq0(Ω).
By solution of the Dirichlet problem (1.1), (1.2) we mean an element u∈W01,p(Ω) which satisfies
(3.2) −∆pu= i0Nfi
u in W−1,p0(Ω), or, equivalently,
Z
Ω
|∇u|p−2∇u∇v= Z
Ω
f(x, u)v, v∈W01,p(Ω),
where i is the compact imbedding of W01,p(Ω) into Lq(Ω) and i0 : Lq0(Ω) → W−1,p0(Ω) is its adjoint. Consequently, (3.2) may be equivalently written as
(3.3) Jϕu=Nfu
(here, by Nf we mean i0Nfi).
We shall formulate sufficient conditions for equation (3.3) to admit a non- trivial solution, via Theorem 2.3. Take X=W01,p(Ω) with Ω⊂RN,N ≥2, a bounded domain with smooth boundary, 1< p <∞,q∈(1, p∗), ϕ(t) =tp−1,
t≥0,f : Ω×R→R a Carath´eodory function which satisfies the conditions below.
•The growth condition (3.1).
•There are θ > p ands0 >0 such that
(3.4) 0< θF(x, s)≤sf(x, s), x∈Ω, |s| ≥s0.
•
(3.5) lim sup
s→0
f(x, s)
|s|p−2s < λ1 uniformly inx∈Ω, where λ1 = inf
(kvkp1,p
kvkp0,p :v∈W01,p(Ω), v6= 0 )
is the first eigenvalue of −∆p in W01,p(Ω).
SinceW01,p(Ω) is a reflexive and smooth Banach space, compactly imbed- ded in Lq(Ω), hypothesis (H1) of Theorem 2.3 is satisfied. Since Jϕ =−∆p : W01,p(Ω)→W−1,p0(Ω) is continuous and satisfies condition (S+) (see, e.g., [9]) hypothesis (H2) of Theorem 2.3 is satisfied, too. By the growth condition (3.1), hypothesis (H3) of Theorem 2.3 is also satisfied. Since θ > p, conditions (ii) and (iii) are satisfied. Finally, condition (3.5) implies condition (iv) in Theorem 2.3 (see, e.g., [9]).
Consequently, Theorem 2.3 applies and gives the already known result (see, e.g., [9]) on the existence of a nontrivial solution for problem (1.1), (1.2).
Example 3.2. Consider the Neumann problem
−∆pu+|u|p−2u=f(x, u) in Ω, (3.6)
|∇u|p−2∂u
∂n = 0 on ∂Ω, (3.7)
where 1 < p < ∞, Ω ⊂ RN, N ≥ 2, is a bounded domain with smooth boundary and f : Ω×R→ R is a Carath´eodory function which satisfies the growth condition (3.1).
By solution of the Neumann problem (3.6), (3.7) we mean an element u∈W1,p(Ω) which satisfies
(3.8) Z
Ω
|∇u|p−2∇u∇v+ Z
Ω
|u|p−2uv = Z
Ω
f(x, u)v, v ∈W1,p(Ω).
AssumeX =W1,p(Ω) is endowed with the norm
|||u|||p1,p =kukp0,p+k |∇u| kp0,p, u∈W1,p(Ω),
which is equivalent to the standard norm on the space W1,p(Ω) (see [6]).
In this case, the duality mappingJϕ on W1,p(Ω),||| · |||1,p
corresponding to the gauge function ϕ(t) =tp−1 is defined (see [6]) by
Jϕ : W1,p(Ω),||| · |||1,p
→ W1,p(Ω),||| · |||1,p∗
, Jϕu=−∆pu+|u|p−2u, u∈W1,p(Ω).
It is easy to see that u ∈W1,p(Ω) is a solution of problem (3.6), (3.7) in the sense of (3.8) if and only if
(3.9) Jϕu=Nfu
(here, byNf we also meani0Nfi, whereiis the compact imbedding of (W1,p(Ω),
||| · |||1,p) intoLq(Ω) andi0 :Lq0(Ω)→ W1,p(Ω),||| · |||1,p∗
is its adjoint).
We shall formulate sufficient conditions for equation (3.9) to admit a nontrivial solution, via Theorem 2.3.
Take X = W1,p(Ω) with Ω ⊂ RN, N ≥ 2, a bounded domain with smooth boundary, 1< p <∞,q∈(1, p∗),ϕ(t) =tp−1,t≥0 andf : Ω×R→ R a Carath´eodory function which satisfies the conditions below.
•The growth condition (3.1).
•There are θ > p ands0 >0 such that
(3.10) 0< θF(x, s)≤sf(x, s), x∈Ω, |s| ≥s0.
•
(3.11) lim sup
s→0
f(x, s)
|s|p−2s < λ1 uniformly inx∈Ω, where λ1 = inf
(|||v|||p1,p
kvkp0,p :v∈W1,p(Ω), v6= 0 )
. Since W1,p(Ω),||| · |||1,p
is a reflexive and smooth Banach space, com- pactly imbedded inLq(Ω) (see [6]), hypothesis (H1) of Theorem 2.3 is satisfied.
Since Jϕ : W1,p(Ω),||| · |||1,p
→ W1,p(Ω),||| · |||1,p∗
is continuous and satisfies condition (S+) (see, e.g., [6]), hypothesis (H2) of Theorem 2.3 is satisfied, too.
By the growth condition (3.1), hypothesis (H3) of Theorem 2.3 is also satisfied.
Since θ > p, conditions (ii) and (iii) are satisfied. Finally, condition (3.11) im- plies condition (iv) in Theorem 2.3 (see, e.g., [6]). Consequently, Theorem 2.3 applies and gives the already known result (see, e.g., [6]) on the existence of a nontrivial solution for problem (3.6), (3.7).
REFERENCES
[1] R.A. Adams,Sobolev Spaces. Academic Press, New York–San Francisco–London, 1975.
[2] A. Ambrosetti and P.H. Rabinowitz,Dual variational methods in critical points theory and applications. J. Funct. Anal.14(1973), 349–381.
[3] H. Brezis,Analyse fonctionelle. Masson, Paris, 1983.
[4] F.E. Browder,Probl`emes non-lineaires. Les Presses de l’Universit´e de Montreal, 1964.
[5] Ioana Cioranescu,Duality Mapping in Nonlinear Functional Analysis. Publishing House of the Romanian Academy, Bucharest, 1974. (Romanian)
[6] J. Crˆınganu,Variational and topological methods for Neumann problems with p-Lapla- cian. Comm. Appl. Nonlinear Anal.11(2004), 1–38.
[7] J. Crˆınganu and G. Dinca,Multiple solutions for a class of nonlinear equations involving a duality mapping. Differential Integral Equations21(2008), 265–284.
[8] D.G. de Figueiredo, Lectures on Ekeland variational principle with applications and detours. Tata Institut of Fundamental Research, Springer-Verlag, 1989.
[9] G. Dinca, P. Jebelean and J. Mahwin,Variational and Topological methods for Dirichlet problems withp-Laplacian. Portugaliae Math.58(2001), 339–378.
[10] G. Dinca and P. Matei,Multiple solutions for operator equations involving duality map- pings on Orlicz-Sobolev spaces via the Mountain Pass Theorem. Rev. Roumaine Math.
Pures Appl.53(2008), 419–438.
[11] O. Kavian, Introduction a la th´eorie des points critiques. Springer Verlag, 1993.
[12] J.L. Lions, Quelques m´ethodes de r´esolution des probl`emes aux limites non-lineaires.
Dunod-Gauthier, Villars, Paris, 1969.
[13] J. Mawhin and M. Willem, Critical Point Theory and Hamiltonian Systems. Springer, 1989.
[14] P. H. Rabinowitz,Minimax Methods in Critical Point Theory with Applications to Dif- ferential Equations. CBMS Reg. Ser. Math. 65. Amer. Math. Soc., Providence, RI, 1986.
Received 14 November 2008 University of Galat¸i
Department of Mathematics 800008 Galat¸i, Romania
jcringanu@ugal.ro