# VARIATIONAL METHODS FOR NON-LOCAL OPERATORS OF ELLIPTIC .VARIATIONAL METHODS FOR NON-LOCAL OPERATORS

date post

07-Aug-2018Category

## Documents

view

214download

0

Embed Size (px)

### Transcript of VARIATIONAL METHODS FOR NON-LOCAL OPERATORS OF ELLIPTIC .VARIATIONAL METHODS FOR NON-LOCAL OPERATORS

VARIATIONAL METHODS

FOR NON-LOCAL OPERATORS OF ELLIPTIC TYPE

RAFFAELLA SERVADEI AND ENRICO VALDINOCI

Abstract. In this paper we study the existence of non-trivial solutions for equationsdriven by a non-local integrodifferential operator LK with homogeneous Dirichlet boundaryconditions. More precisely, we consider the problem

LKu + u + f(x, u) = 0 in u = 0 in Rn \ ,

where is a real parameter and the nonlinear term f satisfies superlinear and subcriticalgrowth conditions at zero and at infinity. This equation has a variational nature, and soits solutions can be found as critical points of the energy functional J associated to theproblem. Here we get such critical points using both the Mountain Pass Theorem andthe Linking Theorem, respectively when < 1 and > 1 , where 1 denotes the firsteigenvalue of the operator LK .

As a particular case, we derive an existence theorem for the following equation drivenby the fractional Laplacian

()su u = f(x, u) in u = 0 in Rn \ .

Thus, the results presented here may be seen as the extension of some classical nonlinearanalysis theorems to the case of fractional operators.

Contents

1. Introduction 22. Some preliminary results 52.1. Estimates on the nonlinearity and its primitive 52.2. The functional setting 63. An eigenvalue problem 74. Proofs of Theorems 1 and 2 94.1. The case < 1 : Mountain Pass type solutions for problem (1.9) 94.2. The case > 1 : Linking type solutions for problem (1.9) 144.3. Proof of Theorem 2 195. Some comments on the sign of the solutions of (1.9) 20Appendix A. Proof of Proposition 9 21References 29

Key words and phrases. Mountain Pass Theorem, Linking Theorem, variational techniques, integrodif-ferential operators, fractional Laplacian.

2010 AMS Subject Classification: Primary: 49J35, 35A15, 35S15; Secondary: 47G20, 45G05.The first author was supported by the MIUR National Research Project Variational and Topological

Methods in the Study of Nonlinear Phenomena, while the second one by the ERC grant (Elliptic Pdesand Symmetry of Interfaces and Layers for Odd Nonlinearities) and the FIRB project A&B (Analysis andBeyond).

1

2 R. SERVADEI AND E. VALDINOCI

1. Introduction

Recently, a great attention has been focused on the study of fractional and non-localoperators of elliptic type, both for the pure mathematical research and in view of con-crete applications, since these operators arise in a quite natural way in many different con-texts, such as, among the others, the thin obstacle problem, optimization, finance, phasetransitions, stratified materials, anomalous diffusion, crystal dislocation, soft thin films,semipermeable membranes, flame propagation, conservation laws, ultra-relativistic limits ofquantum mechanics, quasi-geostrophic flows, multiple scattering, minimal surfaces, materi-als science and water waves. For an elementary introduction to this topic and for a stillnot exhaustive list of related references see, e.g., [3].

In this work we consider the non-local counterpart of semilinear elliptic partial differentialequations of the type

(1.1)

{

u u = f(x, u) in u = 0 on ,

namely

(1.2)

{

()su u = f(x, u) in u = 0 in Rn \ .

Here, s (0, 1) is fixed and ()s is the fractional Laplace operator, which (up to normal-ization factors) may be defined as

(1.3) ()su(x) =1

2

Rn

u(x+ y) + u(x y) 2u(x)

|y|n+2sdy , x Rn

(see [3] and references therein for further details on the fractional Laplacian).Problem (1.1) has a variational nature and its solutions can be constructed as critical

points of the associated EulerLagrange functional. A natural question is whether or notthese topological and variational methods may be adapted to equation (1.2) and to itsgeneralization in order to extend the classical results known for (1.1) to a non-local setting.

To be precise, in the present paper we study the following equation

(1.4)

{

LKu+ u+ f(x, u) = 0 in u = 0 in Rn \ ,

where LK is the non-local operator defined as follows:

(1.5) LKu(x) =1

2

Rn

(

u(x+ y) + u(x y) 2u(x))

K(y) dy , x Rn .

Here K : Rn \ {0} (0,+) is a function such that

(1.6) mK L1(Rn), where m(x) = min{|x|2, 1} ;

(1.7) there exists > 0 and s (0, 1) such that K(x) > |x|(n+2s) for any x Rn \{0} ;

(1.8) K(x) = K(x) for any x Rn \ {0} .

A typical example for K is given by K(x) = |x|(n+2s). In this case LK is the fractionalLaplace operator ()s defined in (1.3).

In problem (1.4) the set Rn, n > 2s, is open, bounded and with Lipschitz boundary.The Dirichlet datum is given in Rn\ and not simply on , consistently with the non-localcharacter of the operator LK .

VARIATIONAL METHODS FOR NON-LOCAL OPERATORS 3

The weak formulation of (1.4) is given by the following problem (for this, it is convenientto assume (1.8))

(1.9)

R2n

(u(x) u(y))((x) (y))K(x y)dx dy

u(x)(x) dx

=

f(x, u(x))(x)dx X0

u X0 .

Here the functional space X denotes the linear space of Lebesgue measurable functionsfrom Rn to R such that the restriction to of any function g in X belongs to L2() and

the map (x, y) 7 (g(x) g(y))

K(x y) is in L2(

R2n \ (C C), dxdy

)

,

where C := Rn \ . Moreover,

X0 = {g X : g = 0 a.e. in Rn \ } .

We note that

(1.10) C20 () X0,

see, e.g., [6, Lemma 11] (for this we need condition (1.6)), and so X and X0 are non-empty.Finally, we suppose that the nonlinear term in equation (1.4) is a function f : R R

verifying the following conditions:

(1.11) f is continuous in R ;

(1.12)there exist a1, a2 > 0 and q (2, 2

), 2 = 2n/(n 2s) , such that

|f(x, t)| 6 a1 + a2|t|q1 for any x , t R ;

(1.13) limt0

f(x, t)

t= 0 uniformly in x ;

(1.14) tf(x, t) > 0 for any x , t R ;

(1.15)there exist > 2 and r > 0 such that for any x , t R, |t| > r

0 < F (x, t) 6 tf(x, t) ,

where the function F is the primitive of f with respect to its second variable, that is

(1.16) F (x, t) =

t

0f(x, )d .

As a model for f we can take the odd nonlinearity f(x, t) = a(x)|t|q2t, with a C(),a > 0 in , and q (2, 2) .

When dealing with partial differential equations driven by the Laplace operator (or, moregenerally, by uniformly elliptic operators) with homogeneous Dirichlet boundary conditions,the above assumptions are standard1 (see, for instance, [1, 5, 8]). In our framework, theexponent 2 plays the role of a fractional critical Sobolev exponent (see, e.g. [3, Theo-rem 6.5]).

We remark that f(x, 0) = 0, thanks to (1.13), therefore the function u 0 is a (trivial)solution of (1.4): our scope will be, then, to construct non-trivial solutions for (1.4). Forthis, we will exploit two different variational techniques: when < 1 (where, as usual,we denoted by 1 the first eigenvalue of LK , see Section 3), we construct a non-trivialsolution via the Mountain Pass Theorem; on the other hand, when > 1, we accomplish

1For the sake of completeness, we remark that condition (1.14) is not implied by the other ones. Indeed,we can consider a function C(R) such that |(t)| 6 1 for any t R, (t) = 1 if |t| > 1 and (t) = 1,when |t| 6 1/2. Taking f(t) = (t)|t|q2t with q (2, 2), it is easy to check that f satisfies (1.11)(1.13)and (1.15) (for instance with r = 1 and = q), but it does not verify condition (1.14) .

4 R. SERVADEI AND E. VALDINOCI

our purposes by using the Linking Theorem. These two different approaches are indeed thenon-local counterparts of the famous theory developed for the Laplace operator (see, e.g.,[1, 4, 5]).

The main result of the present paper is an existence theorem for equations driven bygeneral integrodifferential operators of non-local fractional type, as stated here below.

Theorem 1. Let s (0, 1), n > 2s and be an open bounded set of Rn with Lipschitzboundary. Let K : Rn \ {0} (0,+) be a function satisfying conditions (1.6)(1.8) andlet f : R R verify (1.11)(1.15) .

Then, for any R problem (1.9) admits a solution u X0 which is not identicallyzero.

In fact, if is small (i.e. < 1), we can find a non-negative (non-positive) solution ofproblem (1.9) (see Corollary 21).

When < 1, the thesis of Theorem 1 is still valid with weaker assumptions on f (see[7], where the case = 0 was considered).

In the non-local framework, the simplest example we can deal with is given by the frac-tional Laplacian ()s, according to the following result:

Theorem 2. Let s (0, 1), n > 2s and be an open bounded set of Rn with Lipschitzboundary. Consider the following equation

(1.17)

R2n

(u(x) u(y))((x) (y))

|x y|n+2sdx dy

u(x)(x) dx=

f(x, u(x))(x) dx

for any Hs(Rn) with = 0 a.e. in Rn \ .If f : R R is a function verifying (1.11)(1.15) , then, for any R problem (1.17)

admits a solution u Hs(Rn), which is not identically zero, and such that u = 0 a.e. inR

n \ .

We observe that (1.17) represents the weak formulation of the problem (1.2) . When s = 1,equation (1.2) reduces to the standard semilinear Laplace partial differential equation (1.1):in this sense Theorem 2 may be seen as the fractional version of the classical existence resultin [5, Theorem 5.16] (see also [1, 4, 8, 9]).

This classical result is an application of two critical points theorems (the Mountain PassTheorem and the Linking Theorem) to elliptic partial differential equations. In the presentpaper we prove that the geometry of these classical minimax theorems is respected by thenon-local framework: for this we develop

Recommended

*View more*