Vous êtes sur la page 1sur 20

QUASILINEAR EQUATIONS ON RN : PERTURBATION

RESULTS

BENEDETTA PELLACCI1

Abstract. This paper deals with existence results for the quasilinear
(
elliptic problem
div((I + A(x, u))u) + u +H(x, u, u) = |u|p1 u, in RN ,
u H 1 (RN ) W 2,q (RN ), q > N
where 2 < p < (N + 2)/(N 2), N > 2 and the operator div((I +
A(x, u))u) +H(x, u, u) will be considered as a perturbation of the
Laplacian.
We use a perturbation method recently developed in [1], [2], [3] and we
get results both in the variational and in the nonvariational framework.

1. introduction
In this paper we study a class of nonlinear elliptic equations in RN of the
form
(
div((I + A(x, u))u) + u + H(x, u, u) = |u|p1 u, in RN ,
(P )
u(x) H 1 (RN ) W 2,q (RN ) q > N,
with 2 < p < (N + 2)/(N 2), N > 2 and is a small parameter so that
(P ) is a perturbation of the problem
(
u + u = |u|p1 u, in RN ,
(P0 )
u(x) H 1 (RN ) W 2,q (RN ) q > N.
A(x, s) = (aij (x, s)), aij (x, s) : RN R R is a matrix of class C 2 , with
respect of the variables (x, s). Moreover, we assume that for every compact
set C R there exists a positive constant = (C) such that for every s
in C, for almost every x in RN and for every i = 1, . . . , N
(A0 ) |A(x, s)| ,

(A1 ) |A0s (x, s)| , |i A(x, s)| ,

(A2 ) |A00s (x, s)| , |i A0s (x, s)| = |s i A(x, s)| .


where i A(x, s) is the matrix whose components are the partial derivative of
the matrix A(x, s) with respect to the variable xi . Regarding the function
1
Partially supported by Cofin. National Project Variational methods and nonlinear
differential equations.
1
2 BENEDETTA PELLACCI

H(x, s, ) : RN R RN R we assume that H is of class C 1 with respect


to the variables (s, ) for almost every x in RN . In addition, we suppose
that there exist a positive constant C0 and functions mi , di : RN R+ ,
i = 1, . . . , 3 such that
(
mi Lq (RN ) L2 (RN ), q > N,
(1.1)
di (x) L (RN ) + Lq (RN ) L2 (RN ), i = 1, 2, 3.
satisfying for every s R, for every RN and a.e. in RN
(H0 ) |H(x, s, )| C0 [|s|r + m1 (x) + d1 (x)||t ], r, t 1,

(H1 ) |Hs0 (x, s, )| C0 [|s|r + m2 (x) + d2 (x)||t ], r, t 1,

(H2 ) |H0 (x, s, )| C0 [|s|r + m3 (x) + d3 (x)||t ], r, t 1,


where H (x, s, ) is the vector in RN whose components are the partial
derivatives of H(x, s, ) with respect to i .
Notice that the growth conditions on the function H(x, s, ) with respect
to generate a lack of regularity in the problem. Indeed, consider, for
instance, the case in which H = ||2 , then H(x, u, u) L1 (RN ) even if
u H 1 (RN )C 1 (RN ). So that, the weak formulation of (P ) can be written
only for test functions v in H 1 (RN ) L (RN ). Moreover, the quadratic
dependence of H(x, s, ) is in some sense critical. Indeed, even if we consider
the simple example H(x, s, ) = s(1+||2 )1+ , with > 0 in [17] it is proved
that there can be no solution of problem (P ).
Neverthless, here we will prove existence results for problem (P ) for any
superlinear growth of H with respect to , moreover we remark that our
theorems apply also for the simple example H(x, s, ) = d(|x|)g(s)||t , where
d(|x|) is a monotonous C 1 function that belongs to L2 (RN ) Lq (RN ) +
L (RN ), g C 1 (R) and it does not change sign for every s R+ , t is any
number greater than one. We will set (P ) on H 1 (RN )W 2,q (RN ) in order to
reformulate it as a fixed point problem. We will apply a perturbation method
that has been recently developed in [1], [2], [3] and adapted to the Hilbertian
nonvariational framework in [14]. Notice that our problem is not set on an
Hilbert space so that in Section 2 we will first adapt the result of [14] in
order to deal with operators defined in Banach spaces (Theorem 2.1). By
following this approach we will apply the Implicit Function Theorem in order
to make a nonlinear finite dimensional reduction.
Denote with B(x, s) = I + A(x, s) and K(x, s, ) = H(x, s, ). Quasi
linear problem of this kind have been widely studied in bounded domains
in the nonperturbative setting (see [10], [4], [5], [12] and the references
therein).
In unbounded domains, Problem (P ) is studied, in the absence of the term
|u|p1 u, in [13] where the authors assume that B(x, s) is uniformly bounded
and elliptic, and the function K(x, s, ) has at most a quadratic growth with
respect to .
QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 3

The nonvariational case with the presence of the nonlinearity |u|p1 u has
been treated neither in bounded nor in unbounded domains. It is clear that
on unbounded domains Problem (P ) becomes more difficult because of the
lack of compactness. As a general feature of our approach, we remark that
no problem of compactness arises as we work on a finite dimensional mani-
fold near solutions of problem (P0 ). In the non-variational framework our
existence results are proved in Theorems 3.5, 3.7, 3.8, 3.9. The variational
case on RN is studied in [11] for uniformly bounded and elliptic matrices
B(x, s). Our existence results are proved in Theorems 4.1, 4.2, 4.4 (see
also Remark 4.3). In Section 5 we will consider perturbation with super
linear growth with respect to the gradient in the divergence operator (see
Theorems 5.4, 5.6, Remark 5.5 and Corollary 5.7).
Finally, we point out that we do not need to prove any regularity result for
the solutions of (P ) as we will work in H 1 (RN ) W 2,q (RN ), with q > N so
that any solution we find will be regular.
The paper is organized as follows.
In section 2 we will adapt the result of [14] in order to deal with operators
defined in Banach spaces and following the method developed in [1], [2] we
will apply the Implicit Function Theorem in order to reduce the problem
to a finite dimensional one. In section 3 we apply the results of section 2
in order to find existence results in the nonvariational case (see Theorems
3.5, 3.7, 3.8, 3.9). In section 4 we study the variational framework (see
Theorems 4.1, 4.4). Finally in section 5 we will consider perturbation with
superlinear growth on the gradient in the divergence operator.
Notation
In the sequel we will use the following notation:
Lp (RN ), 1 p + denotes the standard Lebesgue space over RN
and k kp its norm.
W m,r (RN ) is the usual Sobolev space endowed with the norm:

kukm,r = kukr + + kr ukr , m, r 1.

The norm in H 1 (RN ) = W 1,2 (RN ) will be denoted with kk1,2 = kk.
We will denote with (|) the inner product of H 1 (RN ).
H 1 (RN ) is the dual space of H 1 (RN ) and, for every f H 1 (RN ),
u H 1 (RN ), we denote with hf, ui the duality pairing between f
and u.
Br RN is the open ball centered in the origin of radius r.
L(E, F ) is the set of all the linear and continuous operator between
two Banach spaces E and F .
{ei } is a vector of the orthonormal basis of RN .
4 BENEDETTA PELLACCI

2. Abstract Setting
Let X be a Banach space and let E be an Hilbert space with X E. We
will deal with existence of solutions of the following functional equation
(2.1) Q (u) = S(u) + G(, u),
where S : X X satisfies the following hypotheses
(S0 ) S C 1 (X, X),
(S1 ) there exists : Rd X, continuous such that S(()) = 0, for every
in Rd ,
(S2 ) S 0 (()) is a compact perturbation of a linear homeomorphism and it
is such that dimKerS 0 (()) = d.
Note that as S 0 (()) is a compact perturbation of a linear homeomorphism
we have that imS 0 (()) is closed and the operator is Fredholm of index
0. From now on we will use the notation K = kerS 0 (()) and R =
imS 0 (()). As is continuous and S is C 1 , there exists orthonormal bases
(ei ())i=1,...,d and (ki ())i=1,...,d of R and K respectively, where R and
K are defined with respect of the inner scalar product of E. (ei ())i=1,...,d
and (ki ())i=1,...,d depend continuously on .
The operator G : R X X satisfies the following hypotheses
(G0 ) G(0, u) = 0 for all u X,
(G1 ) G is of class C 1 with respect to u in X,
(G2 ) G : R X X and G0 : R X L(X, X) are continuous.
Under the previous assumptions we can use the implicit function Theorem
in order to find a smooth map w = w(, ) K such that
Q (() + w) R .
More precisely we will prove the following theorem.
Theorem 2.1. Assume (S02 ) and (G02 ), then for all r > 0, there exists
r and a smooth map w : [r , r ] Br X such that w = w(, ) K
and
Q (() + w) R ,
Proof. Let H : R Rd X Rd X Rd be defined by
d
P
H1 (, , w, a) = S(() + w) + G(, () + w) ai ei (),
i=1
H2 ((, , w, a) = ((w|k1 ()), . . . , (w|kd ())).
We want to prove that near the point (0, , 0, 0) we can write w = w(, ),
and a = a(, ) such that H(, , w(, ), a(, )) = 0. Solving the equation
H1 = 0 implies that Q R , while solving the equation H2 = 0 means
that w K . By (S1 ) and (G0 ) we have that H(0, , 0, 0) = (0, 0), and
thanks to conditions (S0 ) and (G1 ) we get that H is of class C 1 . To apply
the implicit function Theorem we have to show that H(0,,0,0)
(w,a) is invertible.
QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 5

Note that
H1 (0,,0,0)
w [v] = S 0 (())v,
d
H1 (0,,0,0) P
a [b] = bi ei ().
i=1
H(0,,0,0)
Then (w,a) [v, b] = (w, c) iff
d
X
S 0 (())v bi ei () = w,
i=1
((v|k1 ()), . . . , (v|kd ())) = c.
Let w = w1 + w2 where w1 R and w2 R . Take b as the unique vector
such that bi = (w, ei ()) and consider v1 K such that S 0 (())v1 = w1 .
Finally take v2 as the unique vector such that (v2 , ki ()) = ci . This proves
that H(0,,0,0)
(w,a) is bijective.
By the implicit function Theorem there exists w = w(, ) such that the
equation H(, , w, a) = (0, 0) is satisfied in a neighbourhood of (0, 0 ) for
all 0 Rd . Since Br is compact it follows that there exists r > 0 and
w : [r , r ] Br X such that Q (() + w(, )) R for all (, )
[r , r ] Br .

t
u
Let [r , r ] and define
Zr = {() + w(, ) : Br }.
Remark 2.2. We want to point out that Theorem 2.1 holds even if there
is not an Hilbert space E with X E. Indeed as S 0 (()) is a Fredholm
operator of index zero, R has a topological supplement. So it is possible to
use projection operators instead of the scalar product of E. On the other
hand the presence of an Hilbert space will be fundamental in order to analize
Q on Z .
From now on it is possible to follow the same argument of [14] in order
to analize Q on Z .
We will need some informations on the dependence of G(, s) with respect
of , more precisely we will suppose that
(G3 ) There exists a positive number > 0 and a continuous function :
Rd Rd such that
(2.2) lim (G(, ())|ei ()) = (())i , uniformly in in Br .
0

Let : Rd Rd be defined by
(())i = (Q (() + w(, ))|ei ()) .
By the definition of w we get that the zeroes of in Br correspond to
the zeroes of Q in Zr . Moreover it is possible to prove that () can
6 BENEDETTA PELLACCI

be approximated by the map () for sufficiently small (see for more


details [14]). Therefore it holds the following existence result.
Theorem 2.3. suppose that S and G satisfy conditions (S02 ), (G03 ). If
there exists a bounded open set Rd such that deg(, ) 6= 0, then there
exists 1 such that for all , 0 < || < 1 , the equation (2.1) has at least a
solution in Z .
For the proof see [14].
Consider now the case in which the functional equation (2.1) has a vari-
ational structure, i.e. the operator Q is the derivative of a functional
f = f0 + g , where f00 = S and s g = G(, s). Then there exists also
a map () such that i () = i (). The counter part of Theorem 2.3 in
the variational framework is the following result proved in [1]
Theorem 2.4. Let (S02 ), (G03 ) hold and assume that has a proper
local minimum (or maximum) at some point 0 . Then there exists 1 such
that for all , 0 < || < 1 , the equation (2.1) has at least a solution which
corresponds to a critical point of the functional f .
For the proof see [1].

3. Existence Results: The nonvariational case


In this section we want to apply the results of the previous section to
problem (P ). Following the notation of the previous section we consider
E = H 1 (RN ) X = W 2,q (RN ) H 1 (RN ). Consider the operators S, G
defined by
S(u) = u K |u|p1 u

(3.1)

(3.2) G(, u) = K [div(A(x, u)u) + H(x, u, u)] ,


where K is defined by K := ( + 1)1 . Note that S : X X; in-
deed if u X, then, as 1 < p < (N + 2)/(N 2), |u|p belongs to
L1 (RN ) L (RN ), so that K |u|p1 u is in X. Moreover for every u X
we have that A(x, u)u belongs to W 1,q (RN ) L2 (RN ) for every q
(N, ), so that K (divA(x, u)u) belongs to X; finally, under assumption
(H0 ), we have that H(x, u, u) belongs to Lq (RN ) with q > N , which
implies K(H(x, u, u)) belongs to X. Note that S(u) = f00 (u) where the
functional f0 is defined by
1 1
f0 (u) = kuk2 kukp+1
p+1 ,
2 p+1
so that every u that is in KerS is a critical point of f0 . It is well known (see
[8], [15]) that Problem (P0 ) has a unique positive radial solution which we
call z0 (x). As (P0 ) is traslation invariant every z (x) = zeta(x + ) is also a
QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 7

solution. Moreover z0 (x) is strictly decreasing and decays exponentially to


zero as |x| +. Thus, f0 has a d-dimensional manifold of critical points
Z : = {z = z0 (x + ), Rd },
where z0 (x) is the positive radial solution (P0 ) So in this case : Rd
X is given by () = z . Since it is well known (see [18], [19]) that the
only solutions of the linearized equation of problem (P0 ) are the partial
derivatives of z , then R = Tz Z and ei () = i z . In order to apply
Theorem 2.1 we have to prove that G satisfies hypotheses (G02 ).
Lemma 3.1. Suppose that assumptions (A01 ), (H0 ) hold, then the operator
G defined in (3.2) is continuous between R X and X.
Proof. It is sufficient to show the continuity of G with respect of u. Take
un , u X such that un u in X; as K is an homeomorphism between
H 1 (RN ) and H 1 (RN ) and it is continuous between Lq (RN ) and W 2,q (RN ),
it is enough to prove that
A(x, un )un A(x, u)u, strongly in W 1,q (RN ) L2 (RN ),
H(x, un , un ) H(x, u, u), strongly in L2 (RN ) Lq (RN ).
From Sobolev imbedding Theorems we get that un u and un u
strongly in Ls (RN ) for every s [2, ]. Then A(x, un )un A(x, u)u,
almost everywhere in RN , moreover un and un live in compact sets and
we can use condition (A0 ) to deduce that
|A(x, un )un | |un |,
then by Lebesgue dominated convergence Theorem we have
(3.3) A(x, un )un A(x, u)u, strongly in Lr (RN ) with r [2, ).
Moreover condition (A1 ) implies that
|i (A(x, un )un ) | c0 (|un |(1 + kun k ) + |i un |) ,
again Lebesgue dominated convergence Theorem and (3.3) yield
A(x, un )un A(x, u)u, strongly in W 1,q (RN ) L2 (RN ).
Now let us deal with H(x, un , un ), as before we know that H(x, un , un )
converges to H(x, u, u) almost everywhere. From hypothesis (H0 ) we de-
rive
|H(x, un , un )| C0 (kun kr1 t1
|un | + m1 (x) + d1 (x)kun k |un |),
we conclude observing that the sequences on the right hand side strongly
converge in L2 (RN ) Lq (RN ).

t
u
Lemma 3.2. Suppose that assumptions (A02 ), (H02 ) hold, then G defined
in (3.2) is of class C 1 with respect to u in X, and G0 : R X L(X, X) is
continuous.
8 BENEDETTA PELLACCI

Proof. It is easy to see that it is possible to compute the Gateaux derivative


of G(, u) with respect to the variable u and it is given by
G
(3.4) (, u)[v] := K {div(A(x, u)v + A0s (x, u)uv)}
u n o
+K Hs0 (x, u, u)v + H0 (x, u, u)v .

In order to show that G/u is continuous in X, take a sequence vn which


strongly converges to v in X and note that it is sufficient to prove that for
every fixed u in X
(
A(x, u)vn A(x, u)v,
(a)
A0s (x, u)uvn A0s (x, u)uv,

strongly in W 1,q (RN ) L2 (RN ), and that


(
Hs0 (x, u, u)vn Hs0 (x, u, u)v,
(b)
H (x, u, u)vn , H0 (x, u, u)v,
0

strongly in L2 (RN ) Lq (RN ). By condition (H1 ) we have


|Hs0 (x, u, u)vn | C0 (kukr |vn | + m2 (x)kvn k )
(3.5)
+ C0 d2 (x)kukt |vn | ,


as the last term is strongly convergent in L2 (RN ) Lq (RN ) we get the first
assertion of (b). The term involving H0 (x, u, u)vn can be handled in the
same way thanks to condition (H2 ). Now let us prove (a). Note that (A01 )
imply
|A(x, u)vn + A0s (x, u)uvn | (|vn | + kuk |vn |) ,
and as the sequences on the right hand side strongly converge in L2 (RN )
L (RN ) we get the convergences in (a) in L2 (RN ) Lq (RN ). Moreover we
have
|i (A(x, u)vn ) | c0 (|vn | + kuk |vn | + |i vn |) ,
|i A0s (x, u)uvn | c0 kuk |vn | + kuk2 |vn |
 

+ c0 (kuk |vn | + kvn k |i u|) ,

so that applying Lebesgue dominated convergence Theorem we deduce (a).


In order to conclude it remains to show that the map G0 : R X L(X, X)
is continuous. Consider un u in X and take a sequence vn in X such that
kvn kX = 1; we have to prove that
(
(Hs0 (x, un , un ) Hs0 (x, u, u))vn 0,
(c) 0 0
(H (x, un , un ) H (x, u, u))vn 0,
QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 9

strongly in L2 (RN ) Lq (RN ). Regarding the terms involving A(x, s) and


A0s (x, s), we have to prove that
(
(A(x, un ) A(x, u))vn 0,
(d) 0 0
(As (x, un )un As (x, u)u)vn 0,

strongly in L2 (RN ) W 1,q (RN ). By condition (H2 ) we get

|H0 (x, un , un )vn | C0 kvn k (kun kr1



|un |
+ C0 d3 (x)kun kt1

|un | + m3 (x) ,

so that
|H0 (x, un , un )vn | C0 (kvn k (c1 |un | + c2 d3 (x)|un |) + m3 (x)) ,
and we can use Lebesgue dominated convergence Theorem to get the second
assertion of (c), while the first one can be obtained using (3.5). Now let us
deal with (d). Notice that the mean value theorem and condition (A01 )
imply
(3.6) kA(x, un ) A(x, u)k kun uk ,

(3.7) kA0s (x, un )un A0s (x, u)uk (k(un u)k )


+ (kuk kun uk ) .

Then (3.6) and (3.7) yield

(3.8) (A(x, un ) A(x, u))vn 0, in L2 (RN ) Lq (RN ),

(3.9) (A0s (x, un )un A0s (x, u)u)vn 0, in L2 (RN ) Lq (RN ).


Moreover from (A01 ) and (3.6), (3.7) we deduce
kx ((A(x, un ) A(x, u))vn )kq kvn kq kun uk (1 + kuk )
+(k(un u)k kvn kq )
+(kun uk k2 vn kq ),

proving, together with (3.8), the first assertion of (d). Computing the gra-
dient of (A0s (x, un )un A0s (x, u))u)vn gives
x ((A0s (x, un )un A0s (x, u)u)vn ) = (A0s (x, un )un A0s (x, u)u)vn
+ (A00s (x, un )un un A00s (x, u)u u)vn
+ (A0s (x, un )2 un A0s (x, u)2 u)vn
+ (x A0s (x, un )un x A0s (x, u)u)vn .
10 BENEDETTA PELLACCI

By applying hypotheses (A12 ) we obtain the following estimates on the


gradient of (A0s (x, un )un A0s (x, u))u)vn

kx ((A0s (x, un )un A0s (x, u)u)vn )kq c1 kvn k k(un u)kq
+c2 kvn k {kx (A0s (x, un ) A0s (x, u))ukq }
+c3 kvn k {kA00s (x, un )un A00s (x, u)ukq }
+c4 kvn k {kA0s (x, un )2 un A0s (x, u)2 ukq }
+c5 kvn k {k(A0s (x, un ) A0s (x, u))ukq },

and so we also get the second assertion in (d), proving the Lemma.

t
u

Lemma 3.1 and 3.2 allow us to apply Theorem 2.1. Moreover the operator
G satisfies condition (G3 ) with = 1, and by (2.2) we can evaluate the
components of the map (), obtaining
(3.10) Z
1
(())i = ( G(, z )|i z ) = A(x, z )z i z + H(x, z , z )i z dx .

RN

Moreover it is well known (see [8]) that z0 (x) is a radial function so that we
can define (r) = z0 (x) (r = |x|). Performing a simple change of variable
we have
Z
()i = A(x , z0 (x))z0 (x)i z0 (x) dx
RZN
(3.11)
+ H(x , z0 (x), z0 (x))i z0 (x) dx .
RN

From now on we will suppose that the function H(x, s, ) is of class C 1 with
respect to the variable x. Using (3.11) and computing the partial derivative
yield
Z
j ()i = xj A(x , z0 (x))z0 (x)i z0 (x) dx

ZR
N
(3.12)
xj H(x , z0 (x), z0 (x))i z0 (x) dx .
RN

Theorem 2.3 states that if we find a bounded open set Rd such that
deg(, ) 6= 0, then there exists 1 such that for all 0 < < 1 , the equation
QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 11

2.1 has at least a solution in Z . We define with 1 () and 2 () as follows


Z
1
(3.13) ()i = A(x , z0 (x))z0 (x)i z0 (x),
RZN
(3.14) 2 ()i = H(x , z0 (x), z0 (x))i z0 (x) dx .
RN
Applying the results of the previous section and by Lemma 3.1, 3.2 we get
the following existence result
Theorem 3.3. Assume conditions (A02 ), (H02 ) are satisfied. Suppose
that there exists a point in RN such that
i) ()i = 0, i = 1, . . . , N ,
ii) j ()i |= is invertible,
then there exists a positive 0 such that problem (P ) has a solution for every
with 0 < || < 0 .

Proof. By i) and ii) we get that there exists a bounded open set RN
such that deg(, ) is well defined and
deg(, ) 6= 0.
Then Theorem 2.3 yields the conclusion.

t
u

Now we will study some model examples in which it is possible to find a


point saisfying i) and ii) of Theorem 3.3. From now on we will suppose
that A(x, s) : RN R R is a scalar function, in addition H(x, s, ) and
A(x, s) satisfy the following conditions
(3.15) H(x, s, ) = H(|x|, s, ||) = H(r, s, ),

(3.16) A(x, s) = A(|x|, s) = A(r, s).


We will prove that 0 = 0 is an isolated zero of and deg((), BR0 , 0) 6= 0.

Remark 3.4. We remark that if either H(x, s, ) = 12 A0s (x, s)||2 or A(x, s)
0 and H does not depend on problem (P ) is variational and we will treat
this case in the following section.
Theorem 3.5. Suppose that conditions (H02 ), (3.15) hold and assume that
the map H(x, s, ) is derivable with respect to r and it satisfies the following
condition
(3.17) Hr0 (r, s, ||) > 0, or Hr0 (r, s, ||) < 0, s > 0.
12 BENEDETTA PELLACCI

then there exists a positive number 0 such that for every 0 < || < 0 there
is a solution of the problem
(
u + u + H(x, u, u) = |u|p1 u, in RN ,
(PH )
u H 1 (RN ) W 2,q (RN ).

Remark 3.6. Note that if H(x, s, ) = d(x)||t with d(x) > 0 and d(x) =
d(x) + d(x) where d L2 (RN ) Lq (RN ), d L (RN ), then condition
(3.17) implies that d 6 0.
Proof. First notice that in this case () 2 (). Conditions (H02 ) allow
us to apply Theorems 2.1, 2.3. It will be sufficient to prove that 0 = 0 is a
zero of ()i such that j ()i is invertible. From (3.15) it results
Z
(0)i = H(r, z0 (x), z0 (x))i z0 (x) dx
RZN
xi
= H(r, z0 (x), z0 (x)) 0 (r) dx = 0,
r
RN
because the previous integrals are odd in the variable xi . By (3.12) we have
Z
xi xj
j ((0)i ) = Hr0 (r, z0 (x), z0 (x)) 0 (r) 2 .
r
RN
6 j the last integral is zero since it is odd in some variable, while for
For i =
i = j we have
(xi )2
Z
i ((0)i ) = Hr0 (r, z0 (x), z0 (x)) 0 (r) 2
r
RN
(xi )2
Z
= Hr0 (r, z0 (x), z0 (x))| 0 (r)| 2 .
r
RN
Using condition (3.17) we get the conclusion applying Theorem 3.3.

t
u

We now consider problem (PA ).


Theorem 3.7. Suppose that the matrix A(x, s) satisfies (A02 ), (3.16).
Moreover assume that for every s in R+ and almost everywhere in RN one
of the following conditions is satisfied
(3.18) A0r (|x|, s) 0, s A0r (|x|, s) 0, A00r (|x|, s) 0,

(3.19) A0r (|x|, s) 0, s A0r (|x|, s) 0, A00r (|x|, s) 0.


QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 13

Then there exists a positive number 0 such that for every 0 < || < 0 there
is a solution of the problem
(
u + u div(A(x, u)u) = |u|p1 u, in RN ,
(PA )
u H 1 (RN ) W 2,q (RN ).

Proof. Notice that in this case ()i 1 ()i , moreover from (3.13) and
integrating by part we obtain
(k z0 (x))2
Z X
()i = i A(x , z0 (x)) dx.
2
RN k

Then 0 = 0 is a zero of ()i . Computing the partial derivatives of ()i


yields
( 0 (r))2 h 00
Z
xi xj xi xj i
j ((0)i ) = Ar (x, ) 2 + s A0r (x, ) 0 (r) dx
2 r r
R N

( 0 (r))2
Z   
ij xi xj
+ A0r (x, ) 3 dx.
2 r r
RN
Then we deduce that j ((0)i ) = 0 for every i 6= j, for i = j we have
( 0 (r))2 (xi )2 (xi )2
Z  
00 0 0
i ((0)i ) = Ar (r, ) + s Ar (r, ) (r) dx
2 r2 r
RN
( 0 (r))2 1 (xi )2
Z   
0
+ Ar (r, ) 3 dx.
2 r r
RN
Condition (3.18) or (3.19) and Theorem 3.3 yield the conclusion.

t
u

Finally let us consider the case in which both H(x, s, ) and A(x, s) are
present.
Theorem 3.8. Assume that conditions (A02 ), (H02 ), (3.15)(3.19) hold.
Then there exists a positive number 0 such that problem (P ) has at least a
solution for every with 0 < || < 0 .
Proof. The proof immediately follows from Theorems 3.5, 3.7.

t
u

Suppose now that H(x, s, ) = h(|x|, s)||2 . This case is particularly inter-
esting as there is an interaction between the functions H(x, s, ) and A(x, s)
and we can prove the following.
14 BENEDETTA PELLACCI

Theorem 3.9. Assume that conditions (A02 ), (H02 ), (3.15), (3.16) hold.
If A(x, s) and H(x, s, ) satisfy, for every s in R+ and almost everywhere
in RN , one of the following conditions
A00r (|x|, s) 0, A0r (|x|, s) 0, s A0r (|x|, s) h0r (|x|, s) 0,

(3.20)
or
A00r (|x|, s) 0, A0r (|x|, s) 0, s A0r (|x|, s) h0r (|x|, s) 0,

(3.21)
then there exists a positive number 0 such that problem (P ) has a solution
for every with 0 < || < 0 .
Proof. As before condition (3.16) and (3.15) imply that = 0 is a zero of
the function . Moreover we have
( 0 (r))2 00
Z
xi xj
j ((0)i ) = Ar (x, ) 2 dx
2 r
RN
( 0 (r))2
Z   
0 ij xi xj
+ Ar (x, ) 3 dx
2 r r
RN
( 0 (r))2 
Z
xi xj
s A0r (x, ) h0r (r, s) 0 (r)

+ dx.
2 r
R N

Condition (3.20) or (3.21) and Theorem 3.3 imply the conclusion.

t
u
Remark 3.10. In the non perturbative case problem (P ) has been studied
without the presence of the non-linearity |u|p1 u on unbounded domains in
[13] where the authors consider an uniformly bounded matrix A(x, s) and
the Hamiltonian satisfies the following growth condition
|H(x, s, )| %(x) + b(|s|)[k(x)|| + ||2 ]
where %(x) L2 (RN ) L (RN ) , k(x) Lp (RN ) L (RN ) and b : R+
R+ is an increasing function. We have showed that, in the perturbative
case and if H and A satisfy conditions (3.15)(3.19), we can admit any
superlinear growth on the function H and we can consider at the same time
bounded and unbounded maps A(x, s).

4. Existence Results : the variational case


Now let us consider the case in which
1
(4.1) H(x, s, ) = A0 (x, s)||2 .
2
Notice first that H satisfies condition (H02 ) with r = 0, mi = 0 di = 1 for
i = 1, . . . , 3 and t = 2. In this setting we can apply Theorem 2.4 in order to
QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 15

find a solution of (P ). Indeed it will be sufficient to find a local minimum


(or maximum) of the function
Z
1
(4.2) () = A(x, z (x))z (x)z (x).
2
RN
Theorem 4.1. Assume conditions (A02 ), (4.1). If there exists a point
such that
() 6= 0,
then there exists a positive number 0 such that problem (P ) has a solution
for every with 0 < || < 0 .
Proof. From the exponential decay of the function z0 (x) we derive
lim () = 0.
||

Then if there exists a such that () 6= 0 the map has a relative minimum
or maximum. t
u
The following existence result is a simple application of Theorem 4.1.
Theorem 4.2. Suppose that the matrix A(x, s) is definite for every s in R+
and for almost every x in RN . Then there exists 0 > 0 such that for every
, 0 < || < 0 there exists a solution of Problem (P ).
Remark 4.3. Problem (P ) has been studied in the nonperturbative case
and in the variational framework on buonded domains in [4], [5], [12]. While
on unbounded domains it has been studied in [11] where the authors consider
uniformly bounded matrices B(x, s) imposing that either B does not depend
on x or it is such that there exists (0, p 2) satisfying
(CG1 ) 0 sBs0 (x, s) B(x, s) , a.e. in RN , s R, RN ,

(CG2 ) B(x, s) ij , Bs0 (x, s)s 0, if|x| a.e. in RN , s R,

(CG3 ) B(x, s) ||2 .


We point out that condition (CG1 ) is assumed also in [4], [12] and it is fun-
damental in order to prove that any PalaisSmale sequence is bounded and
compact. Conditions (CG2 ), (CG3 ) are fundamental in order to recover the
compactness of the PalaisSmale sequences on unbounded domains.
We point out that here, thanks to the perturbative formulation of the prob-
lem, we do not require any of the assumptions (CG13 ), more precisely we
consider B(x, s) = I + A(x, s) and we do not impose any sign condition
on A as in (CG1 ), moreover we treat as well unbounded matrices which
obviously do not satisfy (CG23 ).
In the following result we deal with a model case in which it is possible
to obtain the existence of a solution without applying Theorem 4.1.
16 BENEDETTA PELLACCI

Theorem 4.4. Assume conditions (A02 ). Moreover we suppose that the


matrix A(x, s) = A(r, s), with r = |x| and that it satisfies one of the following
conditions
(
A0r (r, s) 0, s R+ r R+ ,
(4.3)
A00r (r, s) 0, s R+ r R+ ,

(
A0r (r, s) 0, s R+ r R+ ,
(4.4)
A00r (r, s) 0, s R+ r R+ ,
Then there exists a positive number 0 such that for every 0 < < 0 there
is a solution of the problem (P ).
Proof. The proof follows applying Theorem 3.9 in the particular case in
which h(r, s) = A0s (r, s). u
t
Remark 4.5. Consider a particular variational case i.e. A(x, s) = 0 and
H(x, s, ) = g(x, s), if we denote with G(x, t) the primitive with respect of
Rt
t of the function g(x, t), G(x, t) = 0 g(x, s)ds, the map will be defined as
follows Z
() = G(x, z (x)) dx.
RN
As before lim () = 0 so that we can apply Theorem 4.1, in order to find
||
a solution, for sufficiently small , of the problem
(
u + u = |u|p1 u g(x, u), in RN
(P )
u H 1 (RN ) W 2,q (RN ).

For example if g(x, s) has constant sign for almost every x in RN and for
every s in R+ , hypotheses of Theorem 4.1 are satisfied and we get a solution
for sufficiently small . We remark that Problem (P ) is studied in [9] as-
suming that G(x, s) C 2 (RN R, R), G(x, 0) = g(x, 0) = 0 and the second
derivative with respect to the variable s satisfies a natural growth condition.

5. SuperLinear Growth in the Divergence Operator


In this section we briefly consider the case of perturbation with general
divergence operator as div(a(x, u, u)); we obtain a model example if
a(x, s, ) = A(x, s)||m2 with m > 1. On the function a(x, s, ) : RN
R RN RN we assume that a is of class C 2 with respect of the variables
(x, s, ) and for every compact set C R we suppose that there exist =
(C), m 1 and 0 such that for almost every x in RN and for every
(s, ) in R RN the following conditions are satisfied
(a0 ) |a(x, s, )| ||m ,
QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 17

(
|xi a(x, s, )| ||m , |a0s (x, s, )| ||m ,
(a1 )
|i a(x, s, )| ||m ,
where ai (x, s, ) are the components of the map a(x, s, ) and we denote
with xi a(x, s, ) (i a(x, s, )) the vector whose components are the partial
derivatives of ai (x, s, ) with respect to xi (to i ), and with a0s (x, s, ) the
vector whose components are the partial derivatives of ai (x, s, ) with respect
to s.

0 m |xj j a(x, s, )| ||m ,
|xi as (x, s, )| || ,

(a2 ) 00 m
|as (x, s, ) || , |i a0s (x, s, ) || ,
|| .

|j i a(x, s, )|

We will be interested in the following problem


(
u + u + {div(a(x, u, u)) + H(x, u, u)} = |u|p1 u,
(P )
u H 1 (RN ) W 2,q (RN ), q > N,
where H(x, s, ) satisfies hypotheses (H02 ). Now the perturbation G(u) is
of the form G = G1 + G2 where Gi are defined by
G1 (u) = K(div(a(x, u, u))),
G2 (u) = K(H(x, u, u)).
By Lemma 3.1, 3.2 G2 is of class C 1 . Let us show that also G1 is of class
C 1.
Lemma 5.1. Under assumptions (a02 ) G1 is of class C 1 .
Proof. Let un u strongly in X. Then by (a0 ), up to a subsequence
a(x, un , un ) a(x, u, u), a.e. in RN ,
|a(x, un , un )| |un |m c0 h(x),
where h(x) belongs to L2 (RN )Lq (RN ) as un u strongly in L2 (RN )
Lq (RN ). From now on we denote with c possibly different positive constants.
(a1 ) yields
|xi [a(x, un , un )]| c(|un | + |xi un |).
Then a(x, un , un ) converges to a(x, u, u) in W 2,q (RN ) L2 (RN ) and this
implies that G1 is continuous. When we compute the derivative of G1 with
respect of u we get
G1
(u)[v] = K(div(a0s (x, u, u)v + a(x, u, u)v)),
u
where a(x, s, ) is the matrix whose components are the map i aj (x, s, ).
It is easy to show that Gu 1 (u)
(u)[v] is continuous with respect to v for every
fixed u in X. Consider now {un } and {vn } in X such that kvn kX = 1 and
un u strongly in X. We have by (a1 )
|a0s (x, un , un )vn + a(x, un , un )vn | ckvn kX |un |,
18 BENEDETTA PELLACCI

which implies that the sequences a0s (x, un , un ) and a(x, un , un ) stron-
gly converge in L2 (RN )Lq (RN ). Moreover by hypotheses (a12 ) we deduce
the following estimates
|xi [a0s (x, un , un )vn ]| kvn k [|un | (1 + |2 un |)]
+ 2kvn k kun km1
|un |,

|xi [ a(x, un , un )vn ]| kvn k c[|un | + |xi un |]


+ kxi vn kq | a(x, un , un )|.
Notice that a(x, un , un ) strongly converges in L (RN ), indeed by (a2 )
k a(x, un , un ) a(x, u, u)k kun k kun uk
+ kun k kun uk .
Therefore a(x, un , un )2 vn converges strongly in Lq (RN ), and this com-
pletes the proof.

t
u
Remark 5.2. Consider the particular interesting case in which a is of the
form a(x, s, ) = A(x, s)||m2 where A(x, s) satisfies hypotheses (A02 ).
Then if m 3 we obtain a perturbation of class C 1 .
As before we can apply Theorem 2.1. Moreover the operator G satisfies
condition (G3 ) with = 1, and by (2.2) we can evaluate the components of
the map () = 1 + 2 . 2 is given in (3.14), while 1 is defined by
Z
(5.1) (1 ())i = a(x, z (x), z (x))i z (x).
RN
We obtain the following existence result.
Theorem 5.3. Assume conditions (a02 ), (H02 ) are satisfied. Suppose
that there exists a point in RN such that
i) ()i = 0, i = 1, . . . , N ,
ii) j ()i |= is invertible,
then there exists a positive 0 such that problem (P ) has a solution for every
with 0 < < 0 .
From now on we will consider the following model case
(5.2) a(x, s, ) = A(r, s)||m2 ,
where A(r, s) is a map depending radially with respect of the variable x.
Theorem 5.4. Assume conditions (a02 ), (3.18), (3.19), (5.2). Then there
exists a positive 0 such that Problem (Pa )
(
u + u div(a(x, u, u)) = |u|p1 u, in RN ,
(Pa )
u H 1 (RN ) W 2,q (RN ) q > N,
QUASILINEAR EQUATIONS ON RN : PERTURBATION RESULTS 19

has a solution for every , 0 < || < 0 .


Proof. Integration by part gives
Z
1
1
()i = | 0 (r)|m i A(|x |, z0 (x)) dx.
m
RN
From now on it is possible to follow the argument used in the proof of
Theorem 3.7.

t
u
Remark 5.5. We remark that it is possible to obtain the same existence
results of Section 3 and 4 under hypotheses (3.15), (3.17). Clearly in this
case we will have interaction between H and the divergence operator if H
has a growth of order m, and it is possible to prove the same existence result
as Theorem 3.9 under conditions (3.20)(3.21).
Consider the case in which
(5.3) a(x, s, ) = h(x)||m2 .
Then Problem (Pa ) is variational and we can prove analogous versions of
Theorems 4.1, 4.2. More precisely, the following result holds.
Theorem 5.6. Assume conditions (a0 )-(a2 ) and (5.3). Define the map
Z
1
() = h(x )|z0 (x)|m dx.
m
RN
6 0. Then there exists a
Suppose that there exists a point such that () =
positive number 0 such that Problem (Pa ) has a solution for every with
0 < || < 0 .
Proof. As before, we have that () 0 when || . If there exists
a point such that () 6= 0, then the map has a relative minimum or
maximum. Theorem 2.4 yields the conclusion.

t
u
As a corollary we get the following existence result.
Corollary 5.7. Assume conditions (a0 )-(a2 ) and (5.3). Moreover, suppose
that for almost every x in RN the following condition holds
h(x) > 0, or h(x) < 0.
Then there exists 0 > 0 such that for every , 0 < || < 0 there exists a
solution of Problem (Pa ).
Proof. The conclusion follows applying Theorem 4.1.

t
u
20 BENEDETTA PELLACCI

References
[1] Ambrosetti, A., Badiale, M., Homoclinics : Poincare-Melnikov Type Results via a
Variational Approach. Ann. Ist. H.P. Analyse non Lineare 15 (2) (1998), 233-252.
[2] Ambrosetti, A., Badiale, M.,Variational Perturbative Methods and Bifurcations of
Bound State From The Essential Spectrum. Proc. Roy. Soc. Edin. Sect-A 128 (6)
(1998), 1131-1161.
[3] Ambrosetti, A., Badiale, M., Cingolani, S., Semiclassical States of Nonlinear
Schrodinger Equation. Arch. Rat. Mech. Anal. 140 (3) (1997), 285-300.
[4] Arcoya, D. Boccardo, L., Critical Points for Multiple Integral of the Calculus of
Variations. Archive Rat. Mech. Anal. 134 (3) (1996), 249-274.
[5] Arcoya, D., Boccardo, L., Some Remarks on Critical Point Theory for Nondifferen-
tiable Functionals. NoDEA 6 (1999), 79-100.
[6] Badiale, M., Duci, A., Concentrated Solutions for a non Variational Semilinear El-
liptic Equations. To appear on Houston Journ. of Math.
[7] Badiale, M. Garcia Azorero, J., Peral I., Perturbation Results for an Anisotropic
Schrodinger Equation via a Variational Method. NoDEA 7 (2000), 201-230.
[8] Beresticki, H., Lions, P.L., Nonlinear Scalar Fields Equation I. Existence of a Ground
State. Arch. Rat. Mech. Anal. 82 (1983), 313-346.
[9] Berti, M., Bolle, Homoclinics and Chaotic Behaviour for Perturbed Second Order
Systems. Annali di Mat. Pura e Appl. CLXXVI n.IV (1999), 323-378.
[10] Boccardo, L., Murat, F., Puel, J.P., L (RN ) Estimates for Some Nonlinear Partial
Differential Equations and an Application to an Existence Result. SIAM Journ. Math.
Anal. 23 (1992), 326-333.
[11] Conti, M. Gazzola, F., Positive Entire Solutions of Quasilinear Elliptic Problems via
Nonsmooth Critical Point Theory. Topol. Meth. in Nonlinear Anal. 8 (2) (1996),
275-294.
[12] Canino, A. Degiovanni, M., Non Smooth Critical Point Theory and Quasilinear Ellip-
tic Equations. Topological Methods in differential equations and Inclusions (Montreal,
1994), Nato, Asi Series Kluwer, Dordrecht.
[13] Donato, P., Giachetti, D., Quasilinear Elliptic Equations with Quadratic Growth in
Unbounded Domains. Nonlinear Anal. TMA 10 (8) (1986), 791-804.
[14] Henrard, M., Homoclinic and Multibump Solutions for Perturbed Second Order
Systems Using Topological Degree. Preprint n.66 December 1997 Institut de
Mathematique Pure et Appliquee, Univesite Catholique de Louvain.
[15] Kwong, M. K., Uniqueness of Positive Radial Solution of u u + up = 0 in RN .
Arch. Rat. Mech. Anal. 105 (1989), 495-505.
[16] Lions, P.L., Resolutions de Problemes Elliptiques Quasilineaires. Arch. Rat. Mech.
Anal. 74 (1980), 335-353.
[17] Nagumo, M., On Principally Linear Elliptic Equations of the Second Order. Osaka
Math. Journ. 6 (2) (1954), 207-229.
[18] Oh, Y.G., Existence of Semiclassical Bound State of Nonlinear Schrodinger equations.
Comm. Math. Phys. 209 (1993), 223-243.
[19] Weinstein, A. Modulation Stability of Ground State of Nonlinear Schrodinger Equa-
tions. SIAM Journ. Math. Anal. 16 (3) (1985), 472-491.

(Benedetta Pellacci) S.I.S.S.A. v. Beirut 2-4, 34013, Trieste


E-mail address, B. Pellacci: pellacci@mat.uniroma1.it

Vous aimerez peut-être aussi