Professional Documents
Culture Documents
variational inequalities
Raffaella Servadei, Enrico Valdinoci∗
Department of Mathematics
University of Rome ‘Tor Vergata’
via della Ricerca Scientifica, 00133 Rome, Italy
Abstract
We consider a class of variational inequalities and we give an existence
result of a nonnegative, not identically zero solution. Such result general-
izes the ones in [4] and [9], which were obtained by topological methods,
to nonlinear variational inequalities. We also obtain the existence of at
least two not identically zero solutions for a class of semilinear elliptic
variational inequalities which have been studied in [4] and [9]. Our proof
is based on the so called direct method, i.e., we introduce a suitable func-
tional and we prove that it has a minimum, which is a solution of the
variational inequality.
1 Introduction
Recently, some attention has been devoted to a class of semilinear, second order,
elliptic variational inequalities depending on a real parameter λ and related to
the obstacle problem. Existence results for such variational inequalities have
been obtained by making use of topological methods: namely, in [4] a mountain
pass argument has been applied to a penalized problem to prove the existence of
a nontrivial nonnegative solution for such variational inequalities, under suitable
hypotheses and for small enough values of λ. In [9], using a linking-type method,
a similar existence result has been proved to hold for all bigger values of λ. Many
existence, non-existence and multiplicity results have been obtained also in [5,
6, 8, 10, 11, 13, 14]. In [10, 11] a variational inequality is studied by the method
of supersolutions. In [14], the existence of multiple solutions for a variational
inequality involving a semilinear coercive operator is proved. Stability results for
semilinear elliptic variational inequalities have also been addressed in [2, 3, 7].
∗ e-mail:servadei@mat.uniroma2.it, valdinoci@mat.uniroma2.it
The authors were supported by MIUR, project ‘Variational Methods and Nonlinear Differen-
tial Equations’.
1
The purpose of this note is to extend the above mentioned results of [4] and
[9] to more general nonlinear variational inequalities (in particular we do not
need the inequality to be semilinear as it was assumed in the previous literature).
The class of operators that we deal with here includes also p̄-Laplacian-type
operators. Also, our result holds for any value of the real parameter λ.
Our proof is based on the so called direct method (i.e., on the choice of
minimizing sequences for a suitable functional).
Moreover, using the results of [4] and [9], we will obtain a multiplicity result
concerning the existence of at least two nontrivial solutions for the variational
inequalities dealt with in [4] and [9] by comparing suitable functional estimates.
This note is organized as follows. In §2, we recall the results of [4] and [9]. In
§3, we prove the existence of nonnegative nontrivial solution for a very general
class of variational inequalities. In §4, we compare our result with the ones in
[4, 9] and make some remarks on the assumptions considered both here and
in [4, 9]. The application to the semilinear problem of [4, 9] and the related
multiplicity result will be dealt with in §5.
N
X
c1 |ξ|2 ≤ aij (x)ξi ξj ≤ c2 |ξ|2 , ∀ ξ ∈ RN , x ∈ Ω.
i,j=1
In [4] and [9], the following class of semilinear elliptic variational inequalities
has been considered:
u ∈ H01 (Ω), u ≤ ψ in Ω
Z Z
aij (x)∂i u(x) ∂j (v − u)(x)dx − λ u(x)(v − u)(x)dx ≥
(P) ΩZ Ω
≥ p(x, u(x))(v − u)(x)dx
Ω
∀ v ∈ H01 (Ω), v ≤ ψ in Ω.
2
The summation over the indeces i, j is understood here above and in the sequel.
The ‘obstacle’ ψ belongs to H 1 (Ω) and satisfies ψ|∂Ω ≥ 0.
We define Z ξ
P (x, ξ) := p(x, t)dt, for x ∈ Ω, ξ ∈ R . (1)
0
In [4] and [9] the ‘forcing term’ p : Ω × R −→ R and the ‘forcing potential’
P : Ω × R −→ R are assumed to fulfill the following conditions:
(P 1) p is a Carathèodory function in Ω × R;
2N
with 1 < s, if N = 1, 2 and 1 < s < N −2 , if N ≥ 3;
Of course, in view of (P 3), the zero function is a trivial solution of problem (P).
Through this note we will be interested in nontrivial (i.e., not identically zero)
nonnegative solutions of variational inequalities.
In [4] and [9] the authors study the existence of a nontrivial solution ũ for
problem (P), respectively in case λ < λ1 and λ ≥ λ1 , where λ1 is the first eigen-
value of the uniformly elliptic operator in divergence form A(·) = −∂i (aij ∂j (·)),
with Dirichlet boundary conditions, by using a penalization method (see [1])
combined with some minimax theorems. More precisely, in [4] the following
theorem has been proved:
Theorem 2.1 Let λ < λ1 . Assume A satisfies (A1), (A2) and (A3) and p ver-
ifies (P 1), (P 2), (P 3) and (P 4). Moreover, let the following hypotheses hold:
(H2) ψ ≥ 0 in Ω;
3
Theorem 2.2 Let λ1 < λ2 ≤ λ3 ≤ . . . ≤ . . . λj ≤ . . . be the divergent sequence
of the eigenvalues of the operator A, where each λj has finite multiplicity coin-
ciding with the number of its different indexes and let e1 , . . . , ek be the eigen-
functions related to λ1 , . . . , λk such that {e1 , . . . , ek } is an L2 (Ω)- orthonormal
system of functions.
Let k ∈ N so that λk < λk+1 . Fix λ ∈ [λk , λk+1 ). Let x̄ and ck be suitable
positive constants (which may depend on k, a3 , s).
Assume A satisfies (A1), (A2) and (A3) and p verifies (P 1), (P 2), (P 3)
and (P 4). Moreover, let (H1), (H4) and the following hypothesis hold:
Remark 2.1 In [9] the authors need that the obstacle ψ is a subsolution of a
suitable Dirichlet problem (see hypothesis (H3) in Theorem 1 of [9]), but this
condition is not necessary (see Remark 17 of [12]).
The solution ũ given by Theorems 2.1 and 2.2 is the weak limit (up to subse-
quences) in H01 (Ω) of a sequence (u ) , where, for any > 0, u is a critical
point of the functional I : H01 (Ω) −→ R defined as follows
1 λ
Z Z
I (v) = aij (x) ∂i v(x) ∂j v(x)dx − v 2 (x)dx +
2 Ω 2 Ω
1
Z Z
+ 2
+ ((v − ψ) ) (x)dx − P (x, v(x))dx.
2 Ω Ω
4
(X1) a(x, 0, 0) = 0 and a(x, z, ζ) ≥ 0 , ∀(x, z, ζ) ∈ Ω × R × RN ;
We take p : Ω × R × RN −→ R satisfying
(Y 2) p(x, 0, 0) = 0
u ∈ W01,p̄ (Ω), u ≤ ψ in Ω
Z
∂z a(x, u(x), ∇u(x)) (v − u)(x) dx+
ΩZ
+ ∂ζi a(x, u(x), ∇u(x)) ∂i (v − u)(x)dx−
Ω
Z
(P ? ) −λ uq−1 (x)(v − u)(x)dx ≥
Z Ω
≥ p(x, u(x), ∇u(x))(v − u)(x)dx+
ZΩ
+ Pζi (x, u(x), ∇u(x))∂i (v − u)(x)dx
Ω
∀ v ∈ W01,p̄ (Ω), v ≤ ψ in Ω.
5
In order to deal with problem (P ? ), we consider the functional I : Kψ −→
R ∪ {+∞} defined as
Z h
λ i
I(v) = a(x, v(x), ∇v(x)) − (v + )q (x) − P (x, v + (x), ∇v + (x)) dx .
Ω q
Nonnegative minima of I are related to problem (P ? ) by the following observa-
tion:
Theorem 3.1 Let a satisfy hypotheses (X1), (X2) (X3) and (X4) and p :
Ω × R × RN −→ R verify conditions (Y 1) and (Y 2). Define P as in (5) and
assume that
6
as n goes to infinity.
By (6), we may assume that I(un ) ≤ I(0) = 0, for any n ∈ N and by Proposition
3.2 we may assume that un ≥ 0.
By
Z means of (X3) and recalling that 0 ≤ un ≤ ψ in Ω, we get that
sup |∇un (x)|p̄ dx < +∞, hence we may assume, possibly taking subsequences,
n∈N Ω
that un converges to u ≥ 0 weakly in W01,p̄ (Ω), strongly in Lq (Ω) and almost
everywhere in Ω. Hence, by (X4) and (X6), u is the desired minimum for I.
Notice that I(u) ≤ I(e) < 0 = I(0), thus u is not identically zero. 2
7
If u 6≡ 0, this gives λ1 ≤ λ + η + c(η)δ s−1 . Choosing now η > 0 small enough
so that λ + η < λ1 and then δ > 0 sufficiently small, a contradiction easily
follows. 2
In [9], the obstacle has been assumed suitably large by (H5). Theorem 3.1
shows that such assumption can be often weakened in the case λ > λ1 . We now
give a concrete example of such improvement: given any λ > λ1 , we fix > 0
so that λ > λ1 + . Take u∗ ∈ L∞ (Ω), u∗ ≥ 0 satisfying
Z Z
∗ 2
(λ1 + ) |u | (x)dx ≥ aij (x)∂i u∗ (x)∂j u∗ (x)dx .
Ω Ω
∗ ∗
Take the obstacle ψ = δu , for δ > 0. Consider the variational inequality
u ∈ H01 (Ω), u ≤ ψ ∗ in Ω
Z Z
aij (x)∂i u(x) ∂j (v − u)(x)dx − λ u(x)(v − u)(x)dx ≥
(P 0 ) ΩZ Ω
≥ u3 (x)(v − u)(x)dx
Ω
∀ v ∈ H01 (Ω), v ≤ ψ ∗ in Ω.
Then, problem (P 0 ) can not be attacked by the method in [9], since (H5) is not
fulfilled for small δ > 0. However, a direct application of Theorem 3.1 gives:
Theorem 3.1 also gives the existence of nontrivial solutions when condition
(P 4), which was needed in [4] and [9], is not fulfilled. For instance, the reader
may deduce from Theorem 3.1 the existence of a nonnegative nontrivial solution
for the problem
u ∈ H01 (Ω), u ≤ ψ in Ω
Z Z
aij (x)∂i u(x) ∂j (v − u)(x)dx − λ u(x)(v − u)(x)dx ≥
(P 00 ) ΩZ Ω
≥
u2 (x)(v − u)(x)dx
Ω
∀ v ∈ H01 (Ω), v ≤ ψ in Ω,
8
5 The multiplicity result for problem (P)
In this section we prove that problem (P) has at least two nontrivial solutions
ũ (obtained as in [4, 9]) and u given by Theorem 3.1. In order to get this result,
we compare the values I(ũ) and I(u).
Theorem 5.1 Let aij , i, j = 1, . . . , N , satisfy hypotheses (A1), (A2) and (A3)
and p verify condition (P 1), (P 2), (P 3) and (P 4). Let ψ ∈ L∞ (Ω). Moreover,
let (H1) and (H4) be verified. Assume that
either λ < λ1 and (H2) and (H3) hold;
or λ ≥ λ1 and (H5) hold.
Taking into account (3), (P 1) and (P 2) and by passing to the limit in (8) as
goes to zero, we have that
1
Z
lim (u − ψ)+ (x)u (x)dx =
→0 Ω
Z Z Z (9)
2
= p(x, ũ(x))ũ(x)dx − aij (x)∂i ũ(x) ∂j ũ(x)dx + λ ũ (x)dx.
Ω Ω Ω
Moreover, by using the same arguments and taking into account hypothesis
(H1), we get
1
Z
lim (u − ψ)+ (x)ψ(x)dx =
→0 Ω
Z Z Z (10)
= p(x, ũ(x))ψ(x)dx − aij (x)∂i uũ(x)∂j ψ(x)dx + λ ũ(x)ψ(x)dx
Ω Ω Ω
and
1
Z Z
lim (u − ψ)+ (x)(ψ − ũ)(x)dx = p(x, ũ(x))(ψ − ũ)(x)dx+
→0 Ω Ω
Z Z (11)
− aij (x)∂i ũ(x) ∂j (ψ − ũ)(x)dx + λ ũ(x)(ψ − ũ)(x)dx.
Ω Ω
9
By (9), (10) and (11) we easily deduce
1
Z
0 ≤ lim ((u − ψ)+ )2 (x)dx =
→0 Ω
1 1
Z Z
+
= lim (u − ψ) (x)u (x)dx − lim (u − ψ)+ (x)ψ(x)dx =
→0 Ω →0 Ω
Z Z
= aij (x)∂i ũ(x) ∂j (ψ − ũ)(x)dx − λ ũ(x)(ψ − ũ)(x)dx+
Ω Ω
1
Z Z
− p(x, ũ(x))(ψ − ũ)(x)dx = lim (u − ψ)+ (x)(ũ − ψ)(x)dx ≤ 0,
Ω →0 Ω
(12)
as ũ ≤ ψ in Ω.
So,
1
Z
lim ((u − ψ)+ )2 (x)dx = 0 (13)
→0 Ω
By using (3) and (13) and passing to the limit in (4) as goes to zero, we obtain
1
Z Z Z
λ
aij (x)∂i ũ(x) ∂j ũ(x)dx − ũ2 (x)dx − P (x, ũ(x))dx ≥ α > 0. (14)
2 Ω 2 Ω Ω
Taking into account the definition of positive part and (P 4), it is easy to check
that Z Z
+ 2
(ũ ) (x)dx ≤ ũ2 (x)dx (15)
Ω Ω
and Z Z
P (x, ũ+ (x))dx ≤ P (x, ũ(x))dx. (16)
Ω Ω
By (14), (15) and (16) we deduce that I(ũ) > 0. Then, we can conclude that
u 6= ũ, i.e. problem (P) has two nontrivial solutions.
2
References
[1] Bensoussan A.-Lions J.L.:‘Applications des inéquations variationelles en
contrôle stochastique’, Dunod, Paris (1978).
[2] Boccardo L.- Capuzzo Dolcetta I.:‘G-convergenza e problema di Dirichlet
unilaterale’, Boll. UMI (4), 12 (1975), pp.115-123.
[3] Boccardo L.- Capuzzo Dolcetta I.:‘Stabilitá delle soluzioni di disequazioni
variazionali ellittiche e paraboliche quasilineari’, Ann. Univ. Ferrara Sez.
7, Sc. Mat. 24 (1978), pp.99-111.
[4] Girardi M.-Mastroeni L.-Matzeu M.:‘Existence and regularity results for
nonnegative solutions of some semilinear elliptic variational inequalities
via Mountain Pass techniques’, ZAA, J. Anal. Appl. Vol.20, No.4 (2001),
845-857.
10
[5] Mancini G.-Musina R.:‘A free boundary problem involving limit Sobolev
exponents’, Manuscripta Math. 58 (1987), pp.77-93.
[6] Mancini G.-Musina R.:‘Holes and hobstacles’, Ann. Inst. Henri Poincaré,
Anal. Non Linéaire, vol.5, no.4 (1988), pp.323-345.
[7] Magrone P.-Servadei R.:‘A stability result for Mountain Pass type solu-
tions of semilinear elliptic variational inequalities, Nonlin. Studies Vol.9,
n.4 (2002), pp.387-405.
[8] Marino A.-Passaseo D.:‘A jumping behaviour induced by an obstacle’,
Progress in Variational Methods in Hamiltonian Systems and Elliptic Equa-
tions (M.Girardi-M.Matzeu-F.Pacella Eds.), Pitnam (1992), pp.127-143.
[9] Matzeu M.-Servadei R.:‘A Linking type method to solve a class of semilinear
elliptic variational inequalities’, Adv. Nonl. St. Vol.2, No.1 (2002), 1-17.
[10] Passaseo D.:‘Molteplicità di soluzioni per certe disequazioni variazionali di
tipo ellittico’, Boll. U.M.I. (7) 3 B (1989), pp.639-667.
[11] Passaseo D.:‘Molteplicità di soluzioni per disequazioni variazionali non
lineari di tipo ellittico’, Rend. Acc. Naz. Sc. detta dei XL, Memorie di
Mat., 109 vol.XV, fasc.2 (1991), pp.19-56.
[12] Servadei R.:‘Existence results for semilinear elliptic variational inequalities
with changing sign nonlinearities’, preprint, Roma 2003, submitted for the
publication.
[13] Szulkin A.:‘On the solvability of a class of semilinear variational inequali-
ties’, Rend. Mat. 4 (1984), pp.121-137.
[14] Szulkin A.:‘On a class of variational inequalities involving gradient opera-
tors’, J. Math. Anal. Appl. 100 (1984), pp.486-499.
11