Home Mathematics Bounded Solutions for Nonlocal Boundary Value Problems on Lipschitz Manifolds with Boundary
Article Open Access

Bounded Solutions for Nonlocal Boundary Value Problems on Lipschitz Manifolds with Boundary

  • Ciprian G. Gal EMAIL logo and Mahamadi Warma
Published/Copyright: March 23, 2016

Abstract

We consider nonlinear nonlocal boundary value problems associated with fractional operators (including the fractional p-Laplace and the regional fractional p-Laplace operators) and subject to general (fractional-like) boundary conditions on bounded domains with Lipschitz boundary. Under suitable conditions on the nonlinearities of our system, we establish the existence of bounded solutions and provide explicit L-estimates of solutions which are optimal with respect to the inhomogeneous “sources” present in the system. As application, these results are shown to apply to a class of nonlinear nonlocal equations for the Dirichlet fractional p-Laplacian and regional fractional p-Laplace with a dissipative nonlinearity, and to a class of semilinear nonlocal boundary value problems with fractional Wentzell–Robin boundary conditions corresponding to the so-called fractional Wentzell Laplacian.

MSC 2010: 35R11; 35J92; 35A15

1 Introduction

In recent years, the mathematical community has developed a great appetite for fractional and nonlocal operators of elliptic type due to their interesting connections to concrete applications (see, e.g., [13, 28, 8, 17, 9, 4, 10] for an extensive but non-exhaustive list of references). In particular, the nonlocal system

(1.1) { ( - Δ ) p s u + f ( x , u ) = 0 in Ω , u = 0 on N Ω

is a subject of intensive mathematical research in the recent literature involving nonlocal integro-differential operators with a strongly singular (non-integrable) kernel. Here s(0,1) and p(1,) are given, ΩN is a bounded domain with Lipschitz continuous boundary Ω and (-Δ)ps is defined by the formula

(1.2) ( - Δ ) p s u ( x ) = lim ε 0 + { y N : | y - x | > ε } | u ( x ) - u ( y ) | p - 2 ( u ( x ) - u ( y ) ) | x - y | N + p s 𝑑 y , x N ,

for measurable functions u:N such that

N | u ( x ) | p - 1 ( 1 + | x | ) N + s p 𝑑 x < .

Recent important results for (1.1) range from existence of variational solutions for a subcritical/critical nonlinearity f of bad sign at infinity (i.e., f(,τ)cf|τ|q-2τ, as |τ| with cf<0 and some q>2) by employing various topological methods (see [6, 23, 28, 8, 32]), to Hölder global regularity results (see [26, 27] in the case f(x,τ)f(x) and p=2, or [5] when f0 with non-homogeneous Dirichlet boundary condition of the form u=g on NΩ, and [18] in the case when f(x,τ)f(x) and p2). The existence of weak solutions of (1.1) in the case where f(x,u)=α|u|qu+β|u|NpN-spu+h(x) (where α,β are constants, 0qNpN-sp-2, 1<p2) has been recently obtained in [32] by using some fixed point argument.

In this paper we wish to consider the nonlocal counterpart of quasilinear elliptic partial differential equations of the type

(1.3) - Δ p u + f ( x , u ) = F 1 ( x ) in Ω ,

subject to various nonlinear boundary conditions on Ω (such as Dirichlet, Neumann–Robin or Wentzell–Robin boundary conditions) for a given function F1. In (1.3), Δp is the classical p-Laplace operator and f=f(x,τ) is an unbounded “dissipative” nonlinearity which satisfies proper conditions at infinity (i.e., as |τ|). We emphasize that system (1.3) is in fact a simple prototype for a quasilinear elliptic system that was investigated thoroughly in [12, 29] (and the references therein), and where general existence and regularity results were provided. Thus, the nonlocal counterpart to (1.3) that we consider now reads

(1.4) A Ω , p , s u + f Ω ( x , u ) = F 1 ( x ) in Ω ,

subject to the boundary condition

(1.5) δ ( B Ω , q , l u + α | u | q - 2 u ) + D A , B u + β | u | p - 2 u + f Ω ( x , u ) = F 2 ( x ) on Ω ,

for some α>0, β>0, δ{0,1}, s,l(0,1) and p,q(1,). Here, the bulk AΩ,p,s and the boundary operators BΩ,q,l, DA,B correspond to certain classes of integro-differential operators on Ω and Ω, respectively, reflecting the nonlocal-character of problem (1.4)–(1.5). The spatially-dependent “sources” F1 and F2 are at least integrable over Ω and Ω, respectively, and the unbounded (in u) functions fΩ and fΩ satisfy some coercivity conditions (see Section 2). Our motivation for considering problem (1.4)–(1.5) is two-fold: first, as a property, the dissipative nature of fΩ, fΩ is frequently found in the application for the classical elliptic problem associated with the p-Laplace operator Δp (see [12] for further literature and applications), and secondly, it is also present for their parabolic counterparts (see [11, 13]) especially in the theory of infinite dimensional dynamical systems dealing with the long-term behavior of solutions (in terms of global attractors and convergence to equilibria) as time goes to infinity. Our main goal is then to derive a series of general existence and regularity results for problem (1.4)–(1.5) assuming that fΩ(,τ), fΩ(,τ) are properly behaved functions as |τ|. Namely, we show the existence of bounded solutions for (1.4)–(1.5) and provide explicit L-estimates which are also optimal with respect to the assumptions on the “sources” F1,F2. We emphasize that we do not place any growth restrictions (i.e., polynomial or otherwise) on the nonlinearities at infinity.

The structure of the paper is as follows. In Section 2 we give our assumptions and introduce the function spaces needed to investigate our problem. Section 3 contains the precise statement of our main result in Theorem 3.5. A complete and detailed proof of this theorem is contained in Section 5. Section 4 contains several important applications which illustrate the applicability of our main theorem. In particular, in Section 4.1 we apply our main result to a class of nonlocal elliptic equations involving the Dirichlet fractional p-Laplacian (1.2), to a class of boundary value problems for the regional fractional p-Laplacian with (fractional) Robin boundary conditions in Section 4.2, and finally in Section 4.3, to nonlocal problems involving the regional fractional Laplacian subject to nonlinear fractional Wentzell-type boundary conditions. The latter class of nonlocal operators was introduced recently by the authors in [14] to deal with parabolic and elliptic equations subject to fractional dynamic boundary conditions. In light of these applications, we think that our main theorem may be regarded as a natural extension of the classical existence theorems (see, for instance, [21]) for (1.3) to the nonlocal fractional framework.

2 Functional Framework and Assumptions

In this section we give our assumptions on the functions involved in our system and we introduce the Sobolev spaces needed throughout the paper.

2.1 General Assumptions

  1. Let 𝐚: be a measurable function satisfying the following conditions:

    1. 𝐚C() is non-decreasing and odd.

    2. There exist two constants c𝐚,C𝐚>0 such that, for all t,

      (2.1) 𝐚 ( t ) t c 𝐚 | t | p and | 𝐚 ( t ) | C 𝐚 | t | p - 1 ,

      for some p(1,).

Let ΩN be a bounded open set with a Lipschitz continuous boundary Ω and 0<s<1. We consider the operator AΩ,p,s formally given by

A Ω , p , s u ( x ) = 2 lim ε 0 { y Ω : | x - y | > ε } K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) 𝑑 y , x Ω ,

provided that the limit exists. We assume the following assumptions. The kernel KΩ:N×N is symmetric and there exist two constants 0<cΩCΩ depending only on Ω and KΩ such that

(2.2) c Ω K Ω ( x , y ) | x - y | N + p s C Ω for all x , y Ω , x y .

  1. Analogously, let 𝐛: be a measurable function satisfying the following conditions:

    1. 𝐛C() is non-decreasing and odd.

    2. There exist two constants c𝐛,C𝐛>0 such that, for all t,

      (2.3) 𝐛 ( t ) t c 𝐛 | t | q and | 𝐛 ( t ) | C 𝐛 | t | q - 1 ,

      for some q(1,).

For 0<l<1 and q(1,), we introduce the boundary operator BΩ,q,l formally given by

B Ω , q , l u ( x ) = 2 lim ε 0 { y Ω : | x - y | > ε } K Ω ( x , y ) 𝐛 ( u ( x ) - u ( y ) ) 𝑑 σ y , x Ω ,

where σ denotes the usual Lebesgue surface measure on Ω. The kernel KΩ:N-1×N-1 is symmetric and there exist two constants 0<cΩCΩ depending only on Ω and KΩ such that

(2.4) c Ω K Ω ( x , y ) | x - y | N - 1 + q l C Ω for all x , y Ω , x y .

We wish to investigate the existence of bounded weak solutions to the nonlocal problem

(2.5) A Ω , p , s u + f Ω ( x , u ) = F 1 in Ω ,

subject to the boundary condition

(2.6) δ ( B Ω , q , l u + α | u | u q - 2 ) + D A , B u + β | u | p - 2 u + f Ω ( x , u ) = F 2 on Ω ,

for some α,β>0, where DA,B is the corresponding coupling operator between (2.5) and (2.6), i.e.,

A Ω , p , s u , v Ω = Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x - D A , B u , v Ω ,

where ,S, S{Ω,Ω}, denotes the duality between an appropriate reflexive Banach space and its dual defined on S (see Section 3 below for more details). Here, δ{0,1} and we recall that s,l(0,1).

  1. We assume the following conditions on the nonlinearities fΩ and fΩ:

    1. fΩ:Ω× and fΩ:Ω× are measurable.

    2. fΩ(x,)C() for a.e. xΩ; fΩ(x,)C() for σ-a.e. xΩ.

    3. There exist some constants α0,β0,η0>0, α1,β10, and a positive function η:+ such that for a.e. xS, S{Ω,Ω}, and for all t with |t|t0, for some t0>0,

      (2.7) f Ω ( x , t ) t - α 0 η ( | t | ) - α 1 and f Ω ( x , t ) t - β 0 η ( | t | ) - β 1 ,

      as long as

      0 < η ( | t | ) | t | η 0 = η 0 ( t 0 ) for all | t | t 0 .

2.2 The Functional Setup

For a bounded open set ΩN with boundary Ω and for s(0,1), p[1,), we denote by

W s , p ( Ω ) := { u L p ( Ω ) : Ω Ω | u ( x ) - u ( y ) | p | x - y | N + s p 𝑑 x 𝑑 y < }

the fractional order Sobolev space endowed with the norm

u W s , p ( Ω ) := ( Ω | u | p 𝑑 x + Ω Ω | u ( x ) - u ( y ) | p | x - y | N + s p 𝑑 x 𝑑 y ) 1 p .

We let

W 0 s , p ( Ω ) = 𝒟 ( Ω ) ¯ W s , p ( Ω )

and

W 0 s , p ( Ω ¯ ) = { u W s , p ( N ) : u = 0 a.e. on N Ω } .

By definition, W0s,p(Ω¯) is always a subspace of Ws,p(Ω) but in general there is no obvious inclusion between W0s,p(Ω¯) and W0s,p(Ω). It has been shown in [15, Theorem 1.4.2.2] that if Ω is an open set with continuous boundary, then 𝒟(Ω) is dense in W0s,p(Ω¯). But if Ω is a bounded open set with Lipschitz continuous boundary, then when s1p we have W0s,p(Ω¯)=W0s,p(Ω). We refer to [15, Chapter 1] for further details on this topic.

Next, for every bounded open set ΩN, 0<s<1, 1<p< we have the following Sobolev embedding (see, e.g., [7]):

(2.8) { W 0 s , p ( Ω ) L p ( Ω ) , p := N p N - s p > p if N > s p , W 0 s , p ( Ω ) L q ( Ω ) for all q [ 1 , ) if N = s p , W 0 s , p ( Ω ) C 0 , s - N p ( Ω ¯ ) if N < s p .

All the embeddings in (2.8) also hold if W0s,p(Ω) is replaced by the space W0s,p(Ω¯).

The following result is contained in [2, Corollary 2.8 and Remark 2.3] for the case p=2 and in [15, 31] for general p(1,).

Theorem 2.1

Let ΩN be a bounded open set with a Lipschitz continuous boundary and p(1,). Then for every 0<s1p, the spaces Ws,p(Ω) and W0s,p(Ω) coincide with equivalent norms.

In view of Theorem 2.1, we have that if 0<s1p and Ω has a Lipschitz continuous boundary, then every uWs,p(Ω) is zero σ-a.e. on Ω, where we recall that σ denotes the usual Lebesgue surface measure on Ω. Therefore, to talk about traces of functions in Ws,p(Ω) that are not necessarily null on Ω, it is not a restriction to assume that 1p<s<1.

The following Sobolev embedding result can be found in [3, 7, 15].

Proposition 2.2

Let p(1,), 1p<s<1 and ΩN be a bounded open set with a Lipschitz continuous boundary Ω. Then the following assertions hold:

  1. If N>sp, then Ws,p(Ω)Lq(Ω) for every q[1,p] where p:=NpN-sp>p.

  2. If N=sp, then Ws,p(Ω)Lq(Ω) for every q[1,).

  3. If N<sp, then Ws,p(Ω)C0,s-Np(Ω¯).

  4. The continuous injection W s , p ( Ω ) L p ( Ω ) is also compact.

  5. There exists a linear continuous trace operator

    (2.9) Tr : W s , p ( Ω ) L q ( Ω ) ,

    such that Tr(u)=u on Ω for every uWs,p(Ω)C(Ω¯) with q=p=p(N-1)N-sp if N>sp, q is any number in [1,) if N=sp and q= if N<sp. Moreover, the continuous embedding Ws,p(Ω)Lp(Ω) is also compact.

Next, for l(0,1) and q[1,), we denote by

W l , q ( Ω ) := { u L q ( Ω ) : Ω Ω | u ( x ) - u ( y ) | q | x - y | N - 1 + q l 𝑑 σ x 𝑑 σ y < }

the fractional order Sobolev space endowed with the norm

u W l , q ( Ω ) := ( Ω | u | q 𝑑 σ + Ω Ω | u ( x ) - u ( y ) | q | x - y | N - 1 + l q 𝑑 σ x 𝑑 σ y ) 1 q .

Comparing with Proposition 2.2, we have the following embedding result.

Proposition 2.3

Let 0<l<1, q[1,) and ΩN be a bounded open set with a Lipschitz continuous boundary Ω. Then the following assertions hold:

  1. If N>1+lq, then Wl,q(Ω)Lr(Ω) for every r[1,q] with q:=q(N-1)N-1-ql>q.

  2. If N=1+lq, then Wl,q(Ω)Lr(Ω) for any r[1,).

  3. If N<1+lq, then Wl,q(Ω)C0,l-N-1q(Ω).

  4. The continuous injection W l , q ( Ω ) L q ( Ω ) is also compact.

For p,q(1,), 1p<s<1, 0<l<1 and δ{0,1}, we endow the Banach space

𝕎 δ l s , p , q ( Ω ¯ ) := { U = ( u , u | Ω ) : u W s , p ( Ω ) , δ u | Ω W l , q ( Ω ) }

with the norm

U 𝕎 l s , p , q ( Ω ¯ ) = u W s , p ( Ω ) + u W l , q ( Ω )

if δ=1, and

𝕎 0 s , p , q ( Ω ¯ ) = 𝕎 s , p ( Ω ¯ ) := { U = ( u , u | Ω ) : u W s , p ( Ω ) }

with the norm

U 𝕎 s , p ( Ω ¯ ) p = u W s , p ( Ω ) p + u W s - 1 p , p ( Ω ) p

if δ=0. Furthermore, it is well known that

U 𝕎 s , p ( Ω ¯ ) p = Ω Ω | u ( x ) - u ( y ) | p | x - y | N + s p 𝑑 x 𝑑 y + Ω | u | p 𝑑 σ

defines an equivalent norm on 𝕎s,p(Ω¯) (see, e.g., [30, Theorem 2.3]).

For r,q[1,] with 1r,q< or r=q= we endow the Banach space

𝕏 r , q ( Ω ¯ ) := L r ( Ω ) × L q ( Ω ) = { ( f , g ) , f L r ( Ω ) , g L q ( Ω ) }

with the norm

( f , g ) 𝕏 r , q ( Ω ¯ ) : = f L r ( Ω ) + g L q ( Ω ) , 1 r , q < , r q ,
( f , g ) 𝕏 r , r ( Ω ¯ ) : = ( f L r ( Ω ) r + g L r ( Ω ) r ) 1 r , 1 r < ,
( f , g ) 𝕏 , ( Ω ¯ ) : = max { f L ( Ω ) , g L ( Ω ) } .

We will simply write 𝕏r(Ω¯):=𝕏r,r(Ω¯). If Ω has a Lipschitz continuous boundary, by virtue of Proposition 2.2 and Proposition 2.3 we can summarize the corresponding continuous embeddings according to the following cases:

(2.10) { 𝕎 l s , p , q ( Ω ¯ ) 𝕏 r , t ( Ω ¯ ) , r [ 1 , p ] , t [ 1 , q ] if N > p s and N > l q + 1 , 𝕎 l s , p , q ( Ω ¯ ) 𝕏 r , t ( Ω ¯ ) , r [ 1 , p ] , t [ 1 , ) if N > p s and N = l q + 1 , 𝕎 l s , p , q ( Ω ¯ ) 𝕏 r , ( Ω ¯ ) , r [ 1 , p ] if N > p s and N < l q + 1 , 𝕎 l s , p , q ( Ω ¯ ) 𝕏 r , t ( Ω ¯ ) , r [ 1 , ) , t [ 1 , q ] if N = p s and N > l q + 1 , 𝕎 l s , p , q ( Ω ¯ ) 𝕏 r , t ( Ω ¯ ) , r [ 1 , ) , t [ 1 , ) if N = p s and N = l q + 1 , 𝕎 l s , p , q ( Ω ¯ ) 𝕏 r , ( Ω ¯ ) , r [ 1 , ) if N = p s and N < l q + 1 , 𝕎 l s , p , q ( Ω ¯ ) 𝕏 ( Ω ¯ ) if N < p s and l ( 0 , 1 )

for δ=1, and

(2.11) { 𝕎 s , p ( Ω ¯ ) 𝕏 r , t ( Ω ¯ ) , r [ 1 , p ] , t [ 1 , p ] if N > p s , 𝕎 s , p ( Ω ¯ ) 𝕏 r , t ( Ω ¯ ) , r [ 1 , ) , t [ 1 , ) if N = p s , 𝕎 s , p ( Ω ¯ ) 𝕏 ( Ω ¯ ) if N < p s

for δ=0.

Remark 2.4

It follows from (2.10) and (2.11) that the embeddings 𝕎ls,p,q(Ω¯)𝕏p,q(Ω¯) and 𝕎s,p(Ω¯)𝕏p(Ω¯) are compact.

For more information on fractional order Sobolev spaces we refer to [1, 7, 15, 19, 31] and their references.

3 Main Results

Recall that we are concerned with the existence of bounded weak solutions for the system

(3.1) { A Ω , p , s u + f Ω ( x , u ) = F 1 in Ω , δ ( B Ω , q , l u + α | u | q - 2 u ) + D A , B u + β | u | p - 2 u + f Ω ( x , u ) = F 2 on Ω .

The coupling operator DA,B is to be understood in the following sense. If uWs,p(Ω) and AΩ,p,su(Ws,p(Ω)), then DA,Bu(Ws-1p,p(Ω)) and one has the identity

(3.2) A Ω , p , s u , v Ω = Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x - D A , B u , v Ω

for every vWs,p(Ω), where ,Ω denotes the duality between (Ws,p(Ω)) and Ws,p(Ω), and ,Ω denotes the duality between (Ws-1p,p(Ω)) and Ws-1p,p(Ω).

Using (3.2), we first give a rigorous notion of a weak solution to system (3.1).

Definition 3.1

Let p,q(1,) and 0<l,s<1. A bounded weak solution to system (3.1) is a function u which satisfies

u W s , p ( Ω ) , δ u W l , q ( Ω ) , u L ( Ω ) , u | Ω L ( Ω )

and obeys the variational identity

Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x
+ δ Ω Ω K Ω ( x , y ) 𝐛 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 σ y 𝑑 σ x + Ω f Ω ( x , u ) v 𝑑 x
(3.3) + Ω ( f Ω ( x , u ) + α δ | u | q - 2 u + β | u | p - 2 u ) v 𝑑 σ = Ω F 1 v 𝑑 x + Ω F 2 v 𝑑 σ

for all vWs,p(Ω), δvWl,q(Ω).

If the solution is not necessarily bounded, then we shall just say that u is a weak solution and in that case, the test function v should in addition belong to L(Ω) and v|ΩL(Ω).

For p(1,) and 0<s<1, we also consider the following system with the homogeneous Dirichlet boundary condition:

(3.4) { A Ω , p , s u + f Ω ( x , u ) = F 1 in Ω , u = 0 on Ω .

In that case a weak solution is defined as follows.

Definition 3.2

A function u is said to be a bounded weak solution to (3.4) if

u W 0 s , p ( Ω ) L ( Ω )

and it satisfies the variational identity

Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x + Ω f Ω ( x , u ) v 𝑑 x = Ω F 1 v 𝑑 x

for all vW0s,p(Ω). If u is not necessarily bounded, then we shall say that u is a weak solution and in that case, v should in addition belong to L(Ω).

Remark 3.3

We notice that if 0<s1p, then it follows from Theorem 2.1 that (3.1) reduces to system (3.4). Therefore, throughout the paper, if 0<s1p, then the results stated for (3.1) are in fact for system (3.4).

Next, let us set rmin:=min{p,q}, rmax:=max{p,q}, and

(3.5) { r s = p = N p N - s p , r t = p = ( N - 1 ) p N - s p , if δ = 0 , r s = p = N p N - s p , r t = q = ( N - 1 ) q N - 1 - q l . if δ = 1 .

Remark 3.4

Note that from (3.5), it follows according to (2.10) and (2.11) that 𝕎δls,p,q(Ω¯)𝕏rs,rt(Ω¯) only when N>ps and N>lq+1. We view this as the critical case. Indeed, in the other remaining cases of (2.10) and (2.11), we can alternatively take either at least one of {r,t} to be arbitrarily large (but finite), or both equal to an arbitrarily large (but finite) number. In all these later cases, the conditions on s1,s2>1 of Theorem 3.5 can be effectively relaxed (cf. Remark 3.6 below).

The following theorem is the main result of this paper.

Theorem 3.5

Let δ{0,1} and suppose (H-a), (H-b) and (H-f) are satisfied. Let (s1,s2)(1,)2 be such that

(3.6) { s 1 > max { r s r s - r min , r s r s - r max } if δ = 1 , s 1 > r s r s - p = N s p if δ = 0 ,

and

(3.7) { s 2 > max { r t r t - r min , r t r t - r max } if δ = 1 , s 2 > r t r t - p = N - 1 s p - 1 if δ = 0 .

Then the following assertions hold:

  1. Let (F1,F2)𝕏s1,s2(Ω¯) and s(1/p,1), l(0,1). Then, there exists at least one bounded weak solution to system (3.1) in the sense of Definition 3.1. Moreover, there exist two constants C>0and ρ>1 (independent of U, F1,F2) such that

    (3.8) U 𝕏 ( Ω ¯ ) C ( ( F 1 , F 2 ) 𝕏 s 1 , s 2 ( Ω ¯ ) + 1 ) 1 ρ - 1 .

  2. Let F 1 L s 1 ( Ω ) with s 1 > N s p and s ( 0 , 1 ) . Then there exists at least one bounded weak solution to system ( 3.4 ) in the sense of Definition 3.2 . Moreover, there exists a constant C > 0 (independent of u and F 1 ) such that

    u L ( Ω ) C ( F 1 L s 1 ( Ω ) + 1 ) 1 p - 1 .

Remark 3.6

The conditions (3.6) and (3.7) are stated only in the critical case when N>ps and N>lq+1. Indeed, in order for the corresponding embedding of (2.10) and (2.11) to be satisfied it is only in this case that both {rs,rt} must be taken as in (3.5). For the other remaining cases in (2.10) and (2.11), at least one of {r=rs,t=rt} can now be chosen as an arbitrarily large but fixed number. The reader can easily check that the proof of Theorem 3.5 still works in these later cases; this observation also allows to relax the conditions for (s1,s2) of (3.6) and (3.7) such that the conclusion of Theorem 3.5 is still satisfied. For some (arbitrarily) small ε=ε(p,q)>0, these conditions read in fact as follows:

  1. If either δ=0 and N=ps, or δ=1, N=ps and Nlq+1, we can take s11+ε, s21+ε.

  2. If δ=1, N>ps and Nlq+1, we are allowed to take s21+ε with s1>1 satisfying (3.6).

  3. If δ=1, N=ps and N>lq+1, we are allowed to take s11+ε with s2>1 satisfying (3.7).

  4. Of course, whenever N<sp and δ{0,1}, every weak solution (if it exists) belongs to Ws,p(Ω) and so by Proposition 2.2 (iii), it is also globally Hölder continuous on Ω¯.

Remark 3.7

Assuming that 𝐚,𝐛C() are odd functions (see (H-a) and H-b) and KΩ, KΩ are symmetric is only required for the validity of the (weak) energy identity (3.2). If one merely begins with the weak formulation in Definition 3.1, then these conditions can be essentially dropped.

4 Applications

We give several applications of Theorem 3.5 when the operators involved are either the fractional p-Laplacian, or the regional fractional Laplacian with a fractional Wentzell boundary condition, which was recently introduced in [14], or the regional fractional p-Laplace operator which was also recently studied in [33]. In order to define these special operators, for 0<s<1 and p(1,) we set

s p - 1 ( X ) := { u : X measurable : X | u ( x ) | p - 1 ( 1 + | x | ) N + s p 𝑑 x < } .

where either X=Ω or X=N, and

s p - 1 ( Ω ) := { u : Ω measurable : Ω | u ( x ) | p - 1 ( 1 + | x | ) N - 1 + s p 𝑑 x < } .

We note for instance that sp-1(X) is nonempty since it contains L(X) and in particular smooth functions with compact support in X for X{N,Ω,Ω}.

4.1 The General Fractional p-Laplace Operator

Let p(1,), s(0,1), and let K:N×N be measurable, symmetric and satisfying

λ K ( x , y ) | x - y | N + s p Λ for all x , y N , x y ,

for some constants 0λΛ. For usp-1(N), xN and ε>0, we write

𝐚 , ε u ( x ) = 2 { y N : | y - x | > ε } K ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) 𝑑 y

such that 𝐚u is defined by the formula

𝐚 u ( x ) = lim ε 0 𝐚 , ε u ( x ) , x N ,

provided that the limit exists. If K(x,y)=|x-y|-N-sp and a(t)=|t|p-2t, then 𝐚=(-Δ)ps, that is, the fractional p-Laplace operator (see, e.g., [18, 26, 27, 28, 32]).

The corresponding boundary value problem for the operator 𝐚 with p(1,), s(0,1) and a dissipative source fΩ can be stated as follows:

(4.1) { 𝐚 u + f Ω ( x , u ) = F 1 in Ω , u = 0 on N Ω .

We note that dissipative sources fΩ like the ones we consider here are frequently found in the application for the classical quasilinear elliptic problem associated with the classical p-Laplacian Δp (see [12] for further literature and applications).

The analogue of Theorem 3.5 for Dirichlet boundary value problem (4.1) is the following result. Its proof is similar to that of Theorem 3.5 and follows in the same fashion with minor (inessential) modifications.

Theorem 4.1

Let ΩN be an arbitrary bounded open set. Suppose that F1Ls1(Ω) for s1>Nsp and that fΩ:Ω× is a measurable function such that fΩ(x,)C() for a.e. xΩ and

f Ω ( x , t ) t - α 0 η ( | t | ) - α 1 for all | t | t 0 , a.e. x Ω ,

for some α0,t0>0, α10. Here, η:+ satisfies 0<η(|t|)/|t|η0=η0(t0)< for all |t|t0. Then problem (4.1) possesses a weak solution uW0s,p(Ω¯)L(Ω), verifying

N N K ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x + Ω f Ω ( x , u ) v 𝑑 x = Ω F 1 v 𝑑 x

for all vW0s,p(Ω¯). Moreover, the estimate

u L ( Ω ) C ( 1 + F 1 L s 1 ( Ω ) ) 1 p - 1

holds for some C>0 independent of u and F1.

We mention that the general Dirichlet problem associated with the operator 𝐚, that is,

{ 𝐚 u = μ in Ω , u = g on N Ω ,

where μ is a measure, has been recently investigated in [5, 6, 20]. If 𝐚=(-Δ)ps and Ω has a Lipschitz continuous boundary, it turns out that every bounded weak solution to (4.1) is globally Hölder continuous in the following sense.

Theorem 4.2

Let all assumptions of Theorem 4.1 be satisfied and assume in addition that F1L(Ω) and Ω is of class C1,1. Then every bounded weak solution to (4.1) belongs to C0,γ(Ω¯) for some γ(0,s], and the estimate

u C 0 , γ ( Ω ¯ ) Q ( F 1 L ( Ω ) )

holds for some positive function Q independent of u and F1.

Proof.

The proof is a consequence of Theorem 4.1 and the application of [18, Theorem 1.1]. ∎

Remark 4.3

The condition s1>Nsp in the statement of Theorem 4.1 is for the case N>ps. In the case N=ps, it can be relaxed to s11+ε for some (arbitrarily) small ε>0 (cf. Remark 3.6). If N<sp, then there is nothing to prove since the Sobolev embedding theorem implies that every weak solution (if it exists) is in fact globally Hölder continuous on Ω¯.

4.2 The Regional Fractional p-Laplacian with Robin Boundary Conditions

Our second application of Theorem 3.5 is the fractional Robin type boundary value problem associated with the regional fractional p-Laplace operator. Let p(1,) and s(0,1). For usp-1(Ω), we define the regional fractional p-Laplacian (-Δ)Ω,psu by the formula

( - Δ ) Ω , p s u ( x ) = lim ε 0 C N , p , s { y Ω : | y - x | > ε } | u ( x ) - u ( y ) | p - 2 ( u ( x ) - u ( y ) ) | x - y | N + s p 𝑑 y , x Ω ,

provided that the limit exists, where the normalized constant CN,p,s is given by

C N , p , s = s 2 2 s Γ ( p s + p + N - 2 2 ) π N 2 Γ ( 1 - s ) ,

and has been introduced in [33, (1.2)] (cf. [16, 24] for p=2). We refer to [33, Section 4] for the class of functions for which the limit exists.

Next, let ΩN be a bounded domain of class C1,1 with boundary Ω. We define the following boundary operator.

Definition 4.4

For 0α<2, p(1,), uC1(Ω) and zΩ, we define the operator 𝒩pα on Ω by

𝒩 p α u ( z ) : = - lim t 0 | d u ( z + t n ( z ) ) d t | p - 2 d u ( z + t n ( z ) ) d t t α ( p - 1 )
(4.2) = - lim t 0 | d u ( z + t n ( z ) ) d t t α | p - 2 d u ( z + t n ( z ) ) d t t α ,

provided that the limit exists.

The operator 𝒩pαu is linear in u if and only if p=2 and 𝒩2α has been introduced in [16]. The extension of 𝒩2α to 𝒩pα for every p(1,) has been recently given in [33]. Let 0<s:=1-α2<1 so that α=2-2s. In [33], the function 𝒩pαu is called the (s,p)-normal derivative of u. Next for β>0 we define the space

C β 2 ( Ω ¯ ) := { u : u ( x ) = f ( x ) ρ ( x ) β - 1 + g ( x ) for all x Ω , for some f , g C 2 ( Ω ¯ ) } ,

where ρ(x)=dist(x,Ω), xΩ. When β(1,), we assume that Cβ2(Ω¯) is defined on Ω¯ by continuous extension.

The following characterization of 𝒩p2-β has been proved in [31] for p=2 and in [33, Lemma 3.5] for general p(1,).

Lemma 4.5

Let 1<β2, p(1,) and uCβ2(Ω¯). Then, for every zΩ,

𝒩 p 2 - β u ( z ) = - | β - 1 | p - 1 lim Ω x z ( | u ( x ) - u ( z ) ρ ( x ) β - 1 | p - 2 u ( x ) - u ( z ) ρ ( x ) β - 1 ) .

The following fractional Green formula for the operator (-Δ)Ω,ps is taken from [33, Theorem 3.9]. In the case p=2 it was given by [16], and further interior regularity results were proven in [24].

Theorem 4.6

Let p(1,), max{1p,p-1p}<s<1 and β:=ps-1p-1+1. Then, for every uCβ2(Ω¯) and vWs,p(Ω),

Ω v ( - Δ ) Ω , p s u 𝑑 x = C N , p , s 2 Ω Ω | u ( x ) - u ( y ) | p - 2 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) | x - y | N + p s 𝑑 x 𝑑 y - C p , s Ω v 𝒩 p 2 - β u 𝑑 σ ,

where 𝒩p2-β is the boundary operator defined in (4.2) and the constant Cp,s is given by

C p , s = ( p - 1 ) C 1 , p , s ( p s - ( p - 2 ) ) ( p s - ( p - 2 ) - 1 ) 0 ( 1 s ) p - p s - 1 - | 1 - s | ( p + 2 ) + 1 - p s s p - p s 𝑑 s .

Next, we introduce a weak formulation on non-smooth domains of an (s,p)-normal derivative.

Definition 4.7

Let p(1,), max{1p,p-1p}<s<1, β:=ps-1p-1+1 and ΩN be a bounded open set with Lipschitz continuous boundary Ω.

  1. Let uWs,p(Ω). We say that (-Δ)Ω,psuLpp-1(Ω) if there exists wLpp-1(Ω) such that

    C N , p , s 2 Ω Ω | u ( x ) - u ( y ) | p - 2 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) | x - y | N + p s 𝑑 x 𝑑 y = Ω w v 𝑑 x

    for all v𝒟(Ω), hence for all vW0s,p(Ω) by density. In that case we write (-Δ)Ω,psu=w.

  2. Let uWs,p(Ω) such that (-Δ)Ω,psuLpp-1(Ω). We say that u has a (weak) (s,p)-normal derivative in Lpp-1(Ω) if there exists gLpp-1(Ω) such that

    (4.3) Ω v ( - Δ ) Ω , p s u 𝑑 x = C N , p , s 2 Ω Ω | u ( x ) - u ( y ) | p - 2 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) | x - y | N + p s 𝑑 x 𝑑 y - Ω g v 𝑑 σ

    for all vWs,p(Ω)C(Ω¯), hence for all vWs,p(Ω) by density and by using (2.9). In that case, the function g is uniquely determined by (4.3), we write Cp,s𝒩p2-βu=g and call g the (weak) (s,p)-normal derivative of u.

Remark 4.8

It follows from Definition 4.7 that the Green’s type formula

(4.4) Ω v ( - Δ ) Ω , p s u 𝑑 x = C N , p , s 2 Ω Ω | u ( x ) - u ( y ) | p - 2 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) | x - y | N + p s 𝑑 x 𝑑 y - C p , s Ω v 𝒩 p 2 - β u 𝑑 σ

holds for all vWs,p(Ω) whenever uWs,p(Ω), (-Δ)Ω,psuLpp-1(Ω) and 𝒩p2-βu exists in Lpp-1(Ω).

If Ω is a bounded open set of class C1,1 and if uCβ2(Ω¯), then by [33, Lemma 3.3 and Theorem 4.1] 𝒩p2-βuLq(Ω) and (-Δ)Ω,psuLq(Ω) for every q[1,).

Next, for p(1,), max{1p,p-1p}<s<1 and β:=ps-1p-1+1, we consider the boundary value problem

(4.5) { ( - Δ ) Ω , p s u + f ( x , u ) = F 1 in Ω , C p , s 𝒩 p 2 - β u + β | u | p - 2 u + f Ω ( x , u ) = F 2 on Ω ,

where (F1,F2)𝕏s1,s2(Ω¯) is given. From the statement of Theorem 3.5, we can give the following solvability result for system (4.5), observing here that a(t)=|t|p-2t and δ=0.

Theorem 4.9

Assume that all conditions of (H-f) are satisfied and that ΩN is a bounded open set with Lipschitz continuous boundary Ω. Let (F1,F2)𝕏s1,s2(Ω¯) with

s 1 > N s p 𝑎𝑛𝑑 s 2 > N - 1 s p - 1 if N > p s

(when N=ps, one can take anys1,s2>1). Then there exists at least one bounded weak solution to system (4.5) in the sense that uWs,p(Ω)L(Ω), u|ΩL(Ω) and

C N , p , s 2 Ω Ω | u ( ( x ) - u ( y ) | p - 2 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) | x - y | N + s p d x d y
+ Ω f Ω ( x , u ) v 𝑑 x + Ω ( f Ω ( x , u ) + β | u | p - 2 u ) v 𝑑 σ = Ω F 1 v 𝑑 x + Ω F 2 v 𝑑 σ

for all vWs,p(Ω). Moreover, there exists a constant C>0 (independent of u, F1,F2) such that

u L ( Ω ) + u L ( Ω ) C ( ( F 1 , F 2 ) 𝕏 s 1 , s 2 ( Ω ¯ ) + 1 ) 1 p - 1 .

4.3 The Regional Fractional Laplacian with the Wentzell Type Boundary Conditions

The third application of Theorem 3.5 is for a nonlocal boundary value problem associated with a certain family of fractional operators that are a suitable self-adjoint realization in 𝕏2(Ω¯) of the regional fractional Laplacian (-Δ)Ωs=(-Δ)Ω,2s. The results given in Section 4.2 for p=2 are valid for the operator (-Δ)Ωs and 𝒩2α. In order to give the definition of the so-called fractional Wentzell operator(-Δ)W,δs on Ω¯, we need to recall one more definition. Let 0<l<1, ul1(Ω) and consider a fractional Laplace–Beltrami type operator formally given by

( - Δ ) Γ l u = lim ε 0 C N - 1 , l { y Ω : | x - y | > ε } u ( x ) - u ( y ) | x - y | N - 1 + 2 l 𝑑 σ y , x Ω .

For 12<s<1, 0<l<1, δ{0,1}, we define the bilinear symmetric form 𝐝Ωδ on the product space 𝕏2(Ω¯) with domain D(𝐝Ωδ)=𝕎δls,2,2(Ω¯), given for

U := ( u , u | Ω ) and Φ := ( φ , φ | Ω ) 𝕎 δ l s , 2 , 2 ( Ω ¯ )

by

𝐝 Ω δ ( U , Φ ) = C N , s 2 Ω Ω ( u ( x ) - u ( y ) ) ( φ ( x ) - φ ( y ) ) | x - y | N + 2 s 𝑑 x 𝑑 y + Ω β u φ 𝑑 σ
+ δ C N - 1 , l 2 Ω Ω ( u ( x ) - u ( y ) ) ( φ ( x ) - φ ( y ) ) | x - y | N - 1 + 2 l 𝑑 σ x 𝑑 σ y

for some β>0. Then one can associate with the closed form 𝐝Ωδ a suitable self-adjoint operator (-Δ)W,δs on 𝕏2(Ω¯) whose definition and characterization is taken from [14].

Definition 4.10

For U=(u,u|Ω)D((-Δ)W,δs), we have

( - Δ ) W , δ s U = ( ( - Δ ) Ω s 0 C s 𝒩 2 2 - 2 s δ ( - Δ ) Γ l + β ) ( u u | Ω )

where the domain D((-Δ)W,δs) consists of all functions U=(u,u|Ω)𝕎δls,2,2(Ω¯), such that (-Δ)ΩsuL2(Ω), δ(-Δ)Γl(u|Ω)L2(Ω), 𝒩22-2su exists in L2(Ω) and

C s 𝒩 2 2 - 2 s u + δ ( - Δ ) Γ l ( u | Ω ) + β u | Ω L 2 ( Ω ) .

We call (-Δ)W,δs the realization of the regional fractional Laplace operator (-Δ)Ωs with the fractional Wentzell–Robin type boundary conditions.

Next, for 12<s<1, 0<l<1, and δ{0,1}, we consider the nonlocal boundary value problem

(4.6) { ( - Δ ) Ω s u + f Ω ( x , u ) = F 1 in Ω , δ ( - Δ ) Γ l u + C s 𝒩 2 2 - 2 s u + β u + f Ω ( x , u ) = F 2 on Ω ,

where (F1,F2)𝕏s1,s2(Ω¯) is given. In light of the statement of Theorem 3.5, we can give the following solvability result for (4.6) owing to the fact that here a(t)b(t)t.

Theorem 4.11

Let δ{0,1} and assume that all conditions of (H-f) are satisfied and that ΩN is a bounded open set with Lipschitz continuous boundary Ω. Let (F1,F2)𝕏s1,s2(Ω¯) with

(4.7) s 1 > N 2 s 𝑎𝑛𝑑 s 2 > { N - 1 2 l if δ = 1 , N - 1 2 s - 1 if δ = 0 .

Then there exists at least one bounded solution to system (4.6) in the following sense:

U = ( u , u | Ω ) 𝕎 δ l s , 2 , 2 ( Ω ¯ ) 𝕏 ( Ω ¯ )

and

𝐝 Ω δ ( U , V ) + Ω f Ω ( x , u ) v 𝑑 x + Ω ( f Ω ( x , u ) + β u ) v 𝑑 σ = Ω F 1 v 𝑑 x + Ω F 2 v 𝑑 σ

for all vWs,2(Ω), δvWl,2(Ω). Moreover, there exists a positive function Q (independent of U, F1,F2) such that

u L ( Ω ) + u L ( Ω ) Q ( ( F 1 , F 2 ) 𝕏 s 1 , s 2 ( Ω ¯ ) + 1 ) .

Remark 4.12

(i) In addition to the assumptions of Theorem 4.11, let us assume further that F1L2(Ω), F2L2(Ω). Then owing to the statement of Theorem 4.11 and Green’s formula (4.4) with p=2, we also have (-Δ)ΩsuL2(Ω) and 𝒩2-2suL2(Ω). Therefore the estimate

(4.8) ( - Δ ) Ω s u L 2 ( Ω ) 2 + δ ( - Δ ) Γ l u L 2 ( Ω ) 2 + 𝒩 2 2 - 2 s u L 2 ( Ω ) 2 Q ( F 1 L s 1 ( Ω ) L 2 ( Ω ) + F 2 L s 2 ( Ω ) L 2 ( Ω ) + 1 )

holds for some positive function Q independent of U, F1 and F2. We refer to any bounded solution that satisfies (4.8), a strong solution.

(ii) Also, Theorem 4.11 assumes (4.7) only in the case when both N>2s and N>2l+1. In all the other remaining cases, it suffices to apply Remark 3.6 with p=q=2.

5 Proof of the Main Result

In what follows we shall only concentrate on proofs in the case when N>ps and N>lq+1; hence, the first case of (2.10) as well as of (2.11) hold. The other remaining cases in (2.10) and (2.11) follow similarly with some minor modifications since 𝕎δls,p,q(Ω¯)𝕏r,t(Ω¯) in those cases as well by choosing {r,t} in an appropriate way (see Remark 3.4). We first set for U=(u,u|Ω) and V=(v,v|Ω)𝕎δls,p,q(Ω¯),

δ ( U , V ) : = Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) d y d x
+ δ Ω Ω K Ω ( x , y ) 𝐛 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 σ y 𝑑 σ x
(5.1) + Ω ( δ α | u | q - 2 u + β | u | p - 2 u ) v 𝑑 σ .

Secondly, for u,vW0s,p(Ω) we set

(5.2) ( u , v ) := Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x .

Next, we give the definition of a weak solution to the system associated with the nonlinear form δ in (5.1) (one can do so analogously for in (5.2)).

Definition 5.1

Let p,q(1,) and 0<l,s<1. A weak solution to system (3.1) with fΩ0 and fΩ0 is a function u which satisfies

u W s , p ( Ω ) , δ u W l , q ( Ω )

and such that

Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x + δ Ω Ω K Ω ( x , y ) 𝐛 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 σ y 𝑑 σ x
+ Ω ( α δ | u | q - 2 u + β | u | p - 2 u ) v 𝑑 σ = Ω F 1 v 𝑑 x + Ω F 2 v 𝑑 σ

holds for all vWs,p(Ω), δvWl,q(Ω).

The following lemma establishes a useful comparison between various energy forms associated with a special class of test functions. It will become important in the proof of Theorem 3.5.

Lemma 5.2

Let δ{0,1} and let the assumptions (H-a) and (H-b) be satisfied. Then the following assertions hold:

  1. Let U = ( u , u | Ω ) 𝕎 δ l s , p , q ( Ω ¯ ) and k 0 a real number. Set u k := ( | u | - k ) + sgn ( u ) and U k := ( u k , u k | Ω ) . Then U k 𝕎 δ l s , p , q ( Ω ¯ ) and

    (5.3) δ ( U k , U k ) δ ( U , U k ) for all k 0 .

  2. Let u W 0 s , p ( Ω ) and k 0 a real number. Set u k := ( | u | - k ) + sgn ( u ) . Then u k W 0 s , p ( Ω ) and

    ( u k , u k ) ( u , u k ) for all k 0 .

Proof.

We give the proof of part (i) of the lemma, the proof of part (ii) follows in the same fashion. First, for U=(u,u|Ω)𝕎δls,p,q(Ω¯) and k0, it is easy to check that Uk=(uk,uk|Ω)𝕎δls,p,q(Ω¯), cf. [31, Lemma 2.6]. Next, let

A k := { x Ω ¯ : | u ( x ) | > k } and B k := Ω ¯ A k = { x Ω ¯ : | u ( x ) | k } .

Then, by definition, uk=(|u|-k)sgn(u) on Ak and uk=0 on Bk. We include the proof for the case δ=1 for the sake of completeness, the case δ=0 is similar.

First, we claim that

Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) 𝑑 y 𝑑 x
(5.4) Ω Ω K Ω ( x , y ) 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) 𝑑 y 𝑑 x for all k 0 .

Note that

Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) 𝑑 y 𝑑 x
= A k Ω A k Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) 𝑑 y 𝑑 x
    + A k Ω B k Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) 𝑑 y 𝑑 x
(5.5)     + B k Ω A k Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) 𝑑 y 𝑑 x .

Set now

A k + := { x Ω : u ( x ) > k } and A k - = { x Ω : u ( x ) < - k }

so that AkΩ=Ak+Ak-, and

B k + Ω := { x Ω : 0 u ( x ) k } and B k - := { x Ω : - k u ( x ) < 0 }

so that BkΩ=Bk+Bk-. On Ak+×Ak+={(x,y)Ω×Ω:u(x)>k,u(y)>k}, we have

𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) = 𝐚 ( u ( x ) - u ( y ) ) ( u ( x ) - k - u ( y ) + k )
= 𝐚 ( u ( x ) - k + k - u ( y ) ) ( u ( x ) - k - u ( y ) + k )
(5.6) = 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) .

On Ak-×Ak-={(x,y)Ω×Ω:u(x)<-k,u(y)<-k}, we have

𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) = 𝐚 ( u ( x ) - u ( y ) ) ( u ( x ) + k - u ( y ) - k )
= 𝐚 ( u ( x ) + k - k - u ( y ) ) ( u ( x ) + k - u ( y ) - k )
(5.7) = 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) .

In what follows, we frequently use the fact that function 𝐚() is non-decreasing.

On Ak+×Ak-={(x,y)Ω×Ω:u(x)>k,u(y)<-k}, using u(x)-u(y)-2k0, we have

𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) = 𝐚 ( u ( x ) - u ( y ) ) ( u ( x ) - k - u ( y ) - k )
𝐚 ( u ( x ) - k - u ( y ) - k ) ( u ( x ) - k - u ( y ) - k )
(5.8) = a ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) .

As in (5.8), on Ak-×Ak+={(x,y)Ω×Ω:u(x)<-k,u(y)>k}, using u(x)-u(y)+2k0, we have

(5.9) 𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) .

On Ak+×Bk+={(x,y)Ω×Ω:u(x)>k, 0u(y)k}, using u(x)-u(y)u(x)-k0 and uk(y)=0, we have

𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) = 𝐚 ( u ( x ) - u ( y ) ) ( u ( x ) - k )
𝐚 ( u ( x ) - k ) ( u ( x ) - k )
(5.10) = 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) .

On Ak-×Bk-={(x,y)Ω×Ω:u(x)<-k,-ku(y)<0}, using u(x)-u(y)u(x)+k0 and uk(y)=0, we have

𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) = 𝐚 ( u ( x ) - u ( y ) ) ( u ( x ) + k )
𝐚 ( u ( x ) + k ) ( u ( x ) + k )
(5.11) = 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) .

On Ak+×Bk-={(x,y)Ω×Ω:u(x)>k,-ku(y)<0}, using u(x)-u(y)u(x)-k0 and uk(y)=0, we have

𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) = 𝐚 ( u ( x ) - u ( y ) ) ( u ( x ) - k )
𝐚 ( u ( x ) - k ) ( u ( x ) - k )
(5.12) = 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) .

On Ak-×Bk+={(x,y)Ω×Ω:u(x)<-k, 0u(y)<k}, using u(x)-u(y)u(x)+k0 and uk(y)=0, we have

𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) = 𝐚 ( u ( x ) - u ( y ) ) ( u ( x ) + k )
𝐚 ( u ( x ) + k ) ( u ( x ) + k )
(5.13) = 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) .

Proceeding similarly to (5.10)–(5.13), we also deduce

(5.14) 𝐚 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) 𝐚 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) )

on Bk+×Ak+, Bk-×Ak-, Bk+×Ak- and Bk-×Ak+, respectively. Now combining (5.5)–(5.14), we easily obtain the claim (5.4). Finally,

Ω Ω K Ω ( x , y ) 𝐛 ( u ( x ) - u ( y ) ) ( u k ( x ) - u k ( y ) ) 𝑑 σ y 𝑑 σ x
(5.15) Ω Ω K Ω ( x , y ) 𝐛 ( u k ( x ) - u k ( y ) ) ( u k ( x ) - u k ( y ) ) 𝑑 σ y 𝑑 σ x for all k 0 .

The proof of estimate (5.15) follows verbatim the proof of (5.4). It is also easy to check that

(5.16) Ω A k | u ( x ) | l - 2 u ( x ) u k ( x ) 𝑑 x Ω A k | u k ( x ) | l 𝑑 x ,

where l{p,q}. Summarizing, estimate (5.3) follows from (5.4), (5.15) and (5.16). The proof of the lemma is finished. ∎

We recall the following lemma which can be found in, e.g., [25, Lemma 3.11].

Lemma 5.3

Let Ξ:[k0,) be a nonnegative, non-increasing function such that there are positive constants c, υ>0 and δ>1 such that

Ξ ( j ) c ( j - k ) - υ Ξ ( k ) δ for all j > k k 0 0 .

Then Ξ(k0+K)=0 with K=c1/υΞ(k0)(δ-1)/υ2δ(δ-1).

Our next result is stated for the nonlinear system associated with the nonlinear form δ in (5.1). It equally applies to in (5.2) as well, but we shall not repeat it here in order to avoid further redundancy.

Lemma 5.4

Let δ{0,1} and suppose (H-a) and (H-b) are satisfied. Let (s1,s2)(1,)2 be such that (3.6) and (3.7) are satisfied. Assume (F1,F2)𝕏s1,s2(Ω¯) and s(1/p,1), l(0,1). If u is a weak solution in the sense of Definition 5.1, which may be taken as a test function, then U𝕏(Ω¯)C, where C>0 only depends on s1,s2, on the norm of (F1,F2)𝕏s1,s2(Ω¯), and on the parameters in (3.6) and (3.7).

Now we are ready to give the proof of our main result.

Proof of Theorem 3.5.

We give the proof of part (i) of the theorem. The proof of part (ii) follows in a similar fashion. We proceed as follows:

  1. We approximate each nonlinear function fS, S{Ω,Ω}, by a globally bounded truncation fSε, ε>0, given for a.e. xS by

    f S ε ( x , t ) = { f S ( x , t ) for t [ - ε - 1 , ε - 1 ] , f S ( x , - ε - 1 ) for t < - ε - 1 f S ( x , ε - 1 ) for t > ε - 1 .

    It is clear that each fSε(x,)C() a.e. xS and |fSε(x,t)|Cε for all t and a.e. xS. Clearly, the constant Cε>0 generally explodes as ε0+. On the other hand, for sufficiently small ε1 the conditions (2.7) are also satisfied by fSε(,), S{Ω,Ω} with constants α0, β0, α1, β1 independent of ε. Indeed, these conditions yield, for |t|t0 and a.e. xS,

    (5.17) f Ω ε ( x , t ) sgn ( t ) - α 0 η ( | t | ) | t | - α 1 | t | - ( α 0 η 0 + α 1 t 0 ) = : - α ¯ 0

    as well as

    (5.18) f Ω ε ( x , t ) sgn ( t ) - β 0 η ( | t | ) | t | - β 1 | t | - ( β 0 η 0 + β 1 t 0 ) = : - β ¯ 0 .

  2. We replace each fS in (3.1) with fSε, S{Ω,Ω}, and consider the corresponding approximate nonlocal problem. In the first step we establish the existence of at least one weak solution Uε.

  3. In the second step, we establish bounds for the solution Uε that are uniform in ε>0.

  4. Finally, we let ε0+ and pass to the limit in the approximate nonlocal problem to deduce the existence of a bounded weak solution to the original problem (3.1).

For practical purposes, in this proof C>0 denotes a positive constant that is independent of ε, but which only depends on the other structural parameters. Such a constant may vary even from line to line. Further dependencies of this constant on other parameters will be pointed out as needed.

Step 1. We first show that the approximate nonlocal problem (3.1) with fS replaced by fSε, S{Ω,Ω}, has at least one weak solution. Here we will give a proof only in the case δ=1, the case δ=0 being similar. To this end, let 𝒜ε be the mapping defined for U,V𝕎δls,p,q(Ω¯) by

(5.19) 𝒜 ε ( U , V ) := δ ( U , V ) + Ω f Ω ε ( x , u ) v 𝑑 x + Ω f Ω ε ( x , u ) v 𝑑 σ .

Note that (5.19) is well defined since fSε is a bounded function on S for S{Ω,Ω}. It is clear that the mapping V𝒜ε(U,V) is linear on 𝕎ls,p,q(Ω¯). Using (H-a), (H-b), (2.2), (2.4), the classical Hölder inequality yields

| 𝒜 ε ( U , V ) | Ω Ω | u ( x ) - u ( y ) | p - 1 | v ( x ) - v ( y ) | | x - y | N + s p 𝑑 y 𝑑 x + Ω Ω | u ( x ) - u ( y ) | q - 1 | v ( x ) - v ( y ) | | x - y | N - 1 + l q 𝑑 σ y 𝑑 σ x
+ C ε Ω | v | 𝑑 x + C ε Ω | v | 𝑑 σ + ( δ α u L q ( Ω ) q - 1 v L q ( Ω ) + β u L p ( Ω ) p - 1 v L p ( Ω ) )
(5.20) C ε , Ω ( 1 + u W s , p ( Ω ) p - 1 + u W s , q ( Γ ) q - 1 ) V 𝕎 l s , p , q ( Ω ¯ ) .

Hence, for each ε>0 the mapping V𝒜ε(U,V) is also continuous on 𝕎ls,p,q(Ω¯). Therefore, there exists an operator

A ε : 𝕎 l s , p , q ( Ω ¯ ) ( 𝕎 l s , p , q ( Ω ¯ ) )

such that

A ε ( U ) , V = 𝒜 ε ( U , V ) for all V 𝕎 l s , p , q ( Ω ¯ ) ,

where , denotes the duality between 𝕎ls,p,q(Ω¯) and (𝕎ls,p,q(Ω¯)). It also follows from (5.20) that Aε maps bounded subsets of 𝕎ls,p,q(Ω¯) into bounded subsets of (𝕎ls,p,q(Ω¯)). Using (2.1), (2.3), (5.17) and (5.18), we can infer the existence of a constant C>0 (independent of ε>0) such that

𝒜 ε ( U , U ) C ( u W s , p ( Ω ) p + u W l , q ( Γ ) q ) - Ω α ¯ 0 | u | 𝑑 x - Ω β ¯ 0 | u | 𝑑 σ
(5.21) C ( u W s , p ( Ω ) p + u W l , q ( Γ ) q ) - C ( u W s , p ( Ω ) + u W l , q ( Γ ) ) .

Estimate (5.21) then yields

(5.22) lim U 𝕎 l s , p , q ( Ω ¯ ) 𝒜 ε ( U , U ) U 𝕎 l s , p , q ( Ω ¯ ) C lim u W s , p ( Ω ) + u W l , q ( Γ ) u W s , p ( Ω ) p + u W l , q ( Γ ) q u W s , p ( Ω ) + u W l , q ( Γ ) - C = + .

This means that the operator Aε is uniformly coercive with respect to ε>0. Moreover, we also observe that (2.1) and (2.3) give

(5.23) lim | s | 𝐚 ( s ) s | s | + | s | p - 1 = + and lim | t | 𝐛 ( t ) t | t | + | t | q - 1 = + .

By [22, Theorem 2] (cf. also [34, Theorem 27.A]) and exploiting (5.20), (5.22) and (5.23) together with the fact that the embedding 𝕎ls,p,q(Ω¯)𝕏p,q(Ω¯) is compact (by Remark 2.4), we deduce that the operator Aε is also surjective; in particular, for every F=(F1,F2)(𝕎ls,p,q(Ω¯)), there exists Uε𝕎ls,p,q(Ω¯) such that Aε(Uε)=F, or equivalently, 𝒜ε(Uε,V)=Aε(Uε),F=F,V for every V𝕎ls,p,q(Ω¯). The assumptions (3.5) on s1,s2 imply that (F1,F2)𝕏s1,s2(Ω¯)(𝕎ls,p,q(Ω¯)). Thus, we have shown that there exists a function Uε𝕎ls,p,q(Ω¯) satisfying the variational identity (3.3) with fS replaced by fSε, S{Ω,Ω}. In particular, we have shown that Uε obeys

Ω Ω K Ω ( x , y ) 𝐚 ( u ε ( x ) - u ε ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x
+ δ Ω Ω K Ω ( x , y ) 𝐛 ( u ε ( x ) - u ε ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 σ y 𝑑 σ x
+ Ω f Ω ε ( x , u ε ) v 𝑑 x + Ω f Ω ε ( x , u ε ) v 𝑑 σ x + Ω ( δ α | u ε | q - 2 u ε + β | u ε | p - 2 u ε ) v 𝑑 σ x
(5.24) = Ω F 1 v 𝑑 x + Ω F 2 v 𝑑 σ x

for all vWs,p(Ω), δvWl,q(Ω).

Step 2. Our goal is to provide a uniform bound for every weak solution Uε𝕎δls,p,q(Ω¯) in the space 𝕏(Ω¯). For the sake of simplicity of notation, we drop the ε-dependence of the solution Uε by simply writing U everywhere in this step.

Step 2.1 (δ=1). For U=(u,u|Ω)𝕎ls,p,q(Ω¯) and kk0:=t0, as before we define uk:=(|u|-k)+sgn(u) and Ak:={xΩ¯:|u(x)|>k}. It is clear that Uk=(uk,uk|Ω)𝕎ls,p,q(Ω¯) with uk=0 on Ω¯Ak (cf. Lemma 5.2). Taking v=uk as a test function in (5.24) and exploiting (5.3), we find

1 ( U k , U k ) - Ω A k α ¯ 0 | u k | 𝑑 x - Ω A k β ¯ 0 | u k | 𝑑 σ 1 ( U , U k ) + Ω A k f Ω ε ( x , u ) u k 𝑑 x + Ω A k f Ω ε ( x , u ) u k 𝑑 σ
(5.25) = Ω A k F 1 u k 𝑑 x + Ω A k F 2 u k 𝑑 σ ,

owing to the uniform bounds (5.17) and (5.18). Using (5.25) and the Hölder inequality, we get

1 ( U k , U k ) Ω A k ( F 1 u k + α ¯ 0 | u k | ) 𝑑 x + Ω A k ( F 2 u k + β ¯ 0 | u k | ) 𝑑 σ
Ω A k ( | F 1 | + α ¯ 0 ) | u k | 𝑑 x + Ω A k ( | F 2 | + β ¯ 0 ) | u k | 𝑑 σ
C u k L r s ( Ω ) | F 1 | + α ¯ 0 L s 1 ( Ω ) χ A k Ω L p 2 ( Ω ) + C u k L r t ( Ω ) | F 2 | + β ¯ 0 L s 2 ( Ω ) χ A k Ω L q 2 ( Ω )
(5.26) C U k 𝕎 l s , p , q ( Ω ¯ ) ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) ,

where C>0 depends only on Ω,Ω,N,p,q,s,l, where s1,s2,p2,q2 are such that

1 r s + 1 s 1 + 1 p 2 = 1 and 1 r t + 1 s 2 + 1 q 2 = 1 ,

and we recall that

r s = p = N p N - s p and r t = { p = p ( N - 1 ) N - s p if δ = 0 , t = q ( N - 1 ) N - 1 - q l if δ = 1 .

We notice that it follows from (H-a) and (H-b) that there exists a constant C>0 such that

(5.27) C ( u k W s , p ( Ω ) p + u k W l , q ( Ω ) q ) 1 ( U k , U k ) for all k 0 .

Case (i). Recall that rmin=min{p,q}>1. First, we assume that

(5.28) { u k W s , p ( Ω ) 1 for all k k 0 if r min = p , u k W l , q ( Ω ) 1 for all k k 0 if r min = q .

It follows from (5.27) and (5.28) that

(5.29) C ( u k W s , p ( Ω ) r min + u k W l , q ( Ω ) r min ) 1 ( U k , U k ) for all k 0 .

Since the mapping ξξrmin is convex, we have that

(5.30) 2 1 - r min ( ξ 1 + ξ 2 ) r min ( ξ 1 r min + ξ 2 r min ) for all ξ 1 , ξ 2 0 .

It follows from (5.29) and (5.30) that

(5.31) C ( u k W s , p ( Ω ) + u k W l , q ( Ω ) ) r min 1 ( U k , U k ) for all k 0 .

The estimates (5.26) and (5.31) imply that there is a constant C>0 such that

(5.32) U k 𝕎 l s , p , q ( Ω ¯ ) r min - 1 C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) for all k k 0 .

It follows from (2.10) (which states 𝕎ls,p,q(Ω¯)𝕏rs,rt(Ω¯)) and (5.32) that

(5.33) U k 𝕏 r s , r t ( Ω ¯ ) r min - 1 C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) for all k k 0 .

Let h>k. Then AhAk and on Ah we have |uk|h-k. From (5.33), we get

( h - k ) χ A h 𝕏 r s , r t ( Ω ¯ ) r min - 1 C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) .

This shows that

(5.34) χ A h 𝕏 r s , r t ( Ω ¯ ) r min - 1 C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) ( h - k ) - ( r min - 1 ) χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) .

Since

1 p 2 = 1 - 1 r s - 1 s 1 > r s r s - 1 r s - r s - r min r s = r min - 1 r s p 2 < r s r min - 1 ,

and, similarly, since q2<rtrmin-1 it follows that

ρ := min { r s p 2 , r t q 2 } > r min - 1

and thus

ρ 0 := ρ r min - 1 > 1 .

Let C:=1Ω¯Xrs,rt(Ω¯). Then, using the classical Hölder inequality, we get that

C - r s p 2 χ A k L p 2 ( Ω ) = C - 1 χ A k L r s ( Ω ) r s p 2 C - 1 χ A k L r s ( Ω ) ρ χ A k 𝕏 r s , r t ( Ω ¯ ) ρ C - ρ

and

C - r t q 2 χ A k L q 2 ( Ω ) = C - 1 χ A k L r t ( Ω ) r t q 2 C - 1 χ A k L r t ( Ω ) ρ χ A k 𝕏 r s , r t ( Ω ¯ ) ρ C - ρ .

This shows that for

C Ω := C r s p 2 - ρ + C r t q 2 - ρ

we have

χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) C Ω χ A k 𝕏 r s , r t ( Ω ¯ ) ρ .

Hence,

χ A h 𝕏 r s , r t ( Ω ¯ ) r min - 1 C ( | F 1 + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) ( h - k ) - ( r min - 1 ) χ A k 𝕏 r s , r t ( Ω ¯ ) ρ
= C ( | F 1 + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) ( h - k ) - ( r min - 1 ) [ χ A k 𝕏 r s , r t ( Ω ¯ ) r min - 1 ] ρ 0 .

The application of Lemma 5.3 with Ξ(h):=χAh𝕏rs,rt(Ω¯)rmin-1 yields the existence of a constant M>0 (independent of F1,F2 and ε>0) such that

χ A K 𝕏 r s , r t ( Ω ¯ ) r min - 1 = 0 with K := M [ C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) ] 1 r min - 1 .

This yields that |uε(x)|K for a.e xΩ¯ and proves our claim.

Case (ii). Next, assume that there is k10 such that

(5.35) { u k 1 W s , p ( Ω ) 1 if r min = p , u k 1 W l , q ( Ω ) 1 if r min = q .

Then, since AkAk1 and |uk||uk1| for kk1, it follows from (5.35) that

(5.36) { u k 1 W s , p ( Ω ) 1 for all k k 1 if r min = p , u k 1 W l , q ( Ω ) 1 for all k k 1 if r min = q .

In this step we also set k0:=max{t0,k1}>0 and recall that rmax=max{p,q}>1. In this case, in (5.26) we shall also take rs=p, rt=t. Then, the estimates (5.26), (5.27), (5.30) and (5.36) imply that there exists a constant C>0 such that

u k 𝕎 l s , p , q ( Ω ¯ ) r max - 1 ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) for all k k 0 .

Now, proceeding exactly in the same fashion as in (5.34), we deduce

(5.37) χ A h 𝕏 r s , r t ( Ω ) r max - 1 C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) ( h - k ) - ( r max - 1 ) χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) ,

with

1 p 2 = 1 - 1 r s - 1 s 1 > r s r s - 1 r s - r s - r max r s = r max - 1 r s p 2 < r s r max - 1

and

1 q 2 = 1 - 1 r t - 1 s 2 > r t r t - 1 r t - r t - r max r t = r max - 1 r t q 2 < r t r max - 1 .

Thus, arguing as in Case (i) to estimate the 𝕏p2,q2(Ω¯)-norm on the right-hand side of (5.37), we obtain

(5.38) χ A h 𝕏 r s , r t ( Ω ¯ ) r max - 1 C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) ( h - k ) - ( r max - 1 ) χ A k 𝕏 r s , r t ( Ω ¯ ) ρ ¯ ,

where ρ¯:=min{rsp2,rtq2}>rmax-1. Once again, if we set Ξ(h):=χAh𝕏rs,rt(Ω¯)rmax-1 in Lemma 5.3, we find that there exists a constant M>0 (independent of F1,F2 and ε>0) such that

χ A K + k 0 𝕏 r s , r t ( Ω ¯ ) r max - 1 = 0 with K := M [ C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) ] 1 r max - 1 .

Step 2.2 (δ=0). In this case, the corresponding estimate (5.26) becomes

0 ( U k , U k ) Ω A k ( | F 1 | + α ¯ 0 ) | u k | 𝑑 x + Ω A k ( | F 2 | + β ¯ 0 ) | u k | 𝑑 σ
C 1 U k 𝕎 s , p ( Ω ¯ ) ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) χ A k 𝕏 p 2 , q 2 ( Ω ¯ ) ,

where

1 r s + 1 s 1 + 1 p 2 = 1 and 1 r t + 1 s 2 + 1 q 2 = 1 .

Here we take rs=p and rt=p such that the analogue of (5.38) now reads

χ A h 𝕏 r s , r t ( Ω ¯ ) p - 1 C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) ( h - k ) - ( p - 1 ) χ A k 𝕏 r s , r t ( Ω ¯ ) ρ 1 ,

with ρ1:=min{rsp2,rtq2}>p-1. This is again enough to derive a uniform L-bound on the solution Uε.

Summing up, we conclude that Uε𝕏(Ω¯) uniformly with respect to ε>0 such that estimate (3.8) holds uniformly in ε>0. In particular, we obtain

(5.39) U ε 𝕏 ( Ω ¯ ) C ( | F 1 | + α ¯ 0 , | F 2 | + β ¯ 0 ) 𝕏 s 1 , s 2 ( Ω ¯ ) 1 ρ - 1

for some ρ>1 and some constant C>0, both independent of ε, F1,F2 and Uε. This completes the proof of Step 2.

Step 3. We are now ready to pass to the limit as ε0 in the energy identity (5.24). First, we observe that according to coercivity estimates (5.21) and (5.22), testing (5.24) again with v=uε, we also deduce

(5.40) U ε 𝕎 δ l s , p , q ( Ω ¯ ) C ,

owing to (5.39). By taking a subsequence if necessary, it follows from (5.39) and (5.40) that there exists a function U=(u,u|Ω)𝕎δls,p,q(Ω¯)𝕏(Ω¯) such that, as ε0+,

(5.41) U ε U weakly star in 𝕏 ( Ω ¯ ) ,
(5.42) U ε U weakly in 𝕎 δ l s , p , q ( Ω ¯ ) ,

and

(5.43) u ε u a.e. in Ω , u ε | Ω u | Ω a.e. on Ω ,

owing to the fact that 𝕎δls,p,q(Ω¯)𝕏p,q(Ω¯) is compact if δ=1, and 𝕎δls,p,q(Ω¯)𝕏p(Ω¯) is also compact when δ=0 (by Remark 2.4). The continuity of fSε(x,)C() for a.e. xS together with the pointwise convergence (5.43) (owing to the strong convergence UεU in 𝕏p,q(Ω¯) if δ=1, or in 𝕏p(Ω¯) if δ=0) yields, as ε0+,

f S ε ( x , u ε S ) f S ( x , u S ) a.e. in S { Ω , Ω } .

Moreover, fSε(x,uεS)fS(x,uS) weakly in 𝕏p,q(Ω¯) if δ=1 (and also converges weakly in 𝕏p(Ω¯), respectively if δ=0) by virtue of the (weak) Lebesgue dominated convergence theorem. Finally, in light of the growth conditions of assumptions (H-a) and (H-b) and the application of the Lebesgue dominated convergence theorem, we have, for every V𝕎ls,p,q(Ω¯),

lim ε 0 Ω Ω K Ω ( x , y ) 𝐚 ( u ε ( x ) - u ε ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y 𝑑 x = Ω Ω K Ω ( x , y ) 𝐚 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 y

and

lim ε 0 Ω Ω K Ω ( x , y ) 𝐛 ( u ε ( x ) - u ε ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 σ y 𝑑 σ x = Ω Ω K Ω ( x , y ) 𝐛 ( u ( x ) - u ( y ) ) ( v ( x ) - v ( y ) ) 𝑑 σ y 𝑑 σ x

owing to the convergence properties (5.41), (5.42) and (2.2), (2.4). Therefore, passing to the limit as ε0 in (5.24), we obtain that the limit solution U is a bounded weak solution in the sense of Definition 3.1. ∎

Funding statement: The research of the second author was partially supported by the AFOSR grant FA9550-15-1-0027.

References

[1] Adams D. R. and Hedberg L. I., Function Spaces and Potential Theory, Grundlehren Math. Wiss. 314, Springer, Berlin, 1996. 10.1007/978-3-662-03282-4Search in Google Scholar

[2] Bogdan K., Burdzy K. and Chen Z-Q., Censored stable processes, Probab. Theory Related Fields 127 (2003), no. 1, 89–152. 10.1007/s00440-003-0275-1Search in Google Scholar

[3] Danielli D., Garofalo N. and Nhieu D.-M., Non-doubling Ahlfors measures, perimeter measures, and the characterization of the trace spaces of Sobolev functions in Carnot–Carathéodory spaces, Mem. Amer. Math. Soc. 182 (2006). 10.1090/memo/0857Search in Google Scholar

[4] D’Elia M. and Gunzburger M., The fractional Laplacian operator on bounded domains as a special case of the nonlocal diffusion operator, Comput. Math. Appl. 66 (2013), 1245–1260. 10.1016/j.camwa.2013.07.022Search in Google Scholar

[5] Di Castro A., Kuusi T. and Palatucci G., Nonlocal Harnack inequalities, J. Funct. Anal. 267 (2014), 1807–1836. 10.1016/j.jfa.2014.05.023Search in Google Scholar

[6] Di Castro A., Kuusi T. and Palatucci G., Local behavior of fractional p-minimizers, Ann. Inst. H. Poincaré Anal. Non Linéaire (2015), 10.1016/j.anihpc.2015.04.003. 10.1016/j.anihpc.2015.04.003Search in Google Scholar

[7] Di Nezza E., Palatucci G. and Valdinoci E., Hitchhiker’s guide to the fractional Sobolev spaces, Bull. Sci. Math. 136 (2012), 521–573. 10.1016/j.bulsci.2011.12.004Search in Google Scholar

[8] Dipierro S., Medina M. and Valdinoci E., Fractional elliptic problems with critical growth in the whole of n, preprint 2015, http://arxiv.org/abs/1506.01748. Search in Google Scholar

[9] Du Q., Gunzburger M., Lehoucq R. B. and Zhou K., Analysis and approximation of nonlocal diffusion problems with volume constraints, SIAM Rev. 54 (2012), 667–696. 10.1137/110833294Search in Google Scholar

[10] Du Q., Gunzburger M., Lehoucq R. B. and Zhou K., A nonlocal vector calculus, nonlocal volume-constrained problems, and nonlocal balance laws, Math. Models Methods Appl. Sci. 23 (2013), 493–540. 10.1142/S0218202512500546Search in Google Scholar

[11] Gal C. G., On a class of degenerate parabolic equations with dynamic boundary conditions, J. Differential Equations 253 (2012), 126–166. 10.1016/j.jde.2012.02.010Search in Google Scholar

[12] Gal C. G. and Warma M., Existence of bounded solutions for quasilinear elliptic equations on manifolds with boundary, J. Differential Equations 255 (2013), 151–192. 10.1016/j.jde.2013.04.007Search in Google Scholar

[13] Gal C. G. and Warma M., Reaction-diffusion equations with fractional diffusion on non-smooth domains with various boundary conditions, Discrete Contin. Dyn. Syst. 36 (2016), 1279–1319. 10.3934/dcds.2016.36.1279Search in Google Scholar

[14] Gal C. G. and Warma M., Elliptic and parabolic equations with fractional diffusion and dynamic boundary conditions, Evol. Equ. Control Theory, to appear. Search in Google Scholar

[15] Grisvard P., Elliptic Problems in Nonsmooth Domains, Monogr. Stud. Math. 24, Pitman, Boston, 1985. Search in Google Scholar

[16] Guan Q. Y., Integration by parts formula for regional fractional Laplacian, Comm. Math. Phys. 266 (2006), 289–329. 10.1007/s00220-006-0054-9Search in Google Scholar

[17] Gunzburger M. and Lehoucq R. B., A nonlocal vector calculus with application to nonlocal boundary value problems, Multiscale Model. Simul. 8 (2010), 1581–1598. 10.1137/090766607Search in Google Scholar

[18] Iannizzotto A., Mosconi S. and Squassina M., Global Hölder regularity for the fractional p-Laplacian, Rev. Mat. Iberoam, to appear. 10.4171/RMI/921Search in Google Scholar

[19] Jonsson A. and Wallin H., Function spaces on subsets of N, Math. Rep. Ser. 2 (1984), no. 1. Search in Google Scholar

[20] Kuusi T., Mingione G. and Sire Y., Nonlocal equations with measure data, Comm. Math. Phys. 337 (2015), 1317–1368. 10.1007/s00220-015-2356-2Search in Google Scholar

[21] Ladyzenskaja O. A. and Uralceva N. N., Linear and Quasilinear Equations of Elliptic Type, Academic Press, New York, 1968. Search in Google Scholar

[22] Leray J. and Lions J. L., Quelques résultats de Visik sur les problèmes elliptiques nonlinéaires par les methodes de Minty–Browder, Bull. Soc. Math. France 93 (1965), 97–107. 10.24033/bsmf.1617Search in Google Scholar

[23] Mosconi S., Perera K., Squassina M. and Yang Y., The Brezis–Nirenberg problem for the fractional p-Laplacian, preprint 2015, http://arxiv.org/abs/1508.00700. 10.1007/s00526-016-1035-2Search in Google Scholar

[24] Mou C. and Yi Y., Interior regularity for regional fractional Laplacian, Comm. Math. Phys. 340 (2015), 233–251. 10.1007/s00220-015-2445-2Search in Google Scholar

[25] Murthy M. K. V. and Stampacchia G., Boundary value problems for some degenerate-elliptic operators, Ann. Mat. Pura Appl. (4) 80 (1968), 1–122. 10.1007/BF02413623Search in Google Scholar

[26] Ros-Oton X. and Serra J., The Dirichlet problem for the fractional Laplacian: Regularity up to the boundary, J. Math. Pures Appl. (9) 101 (2014), 275–302. 10.1016/j.matpur.2013.06.003Search in Google Scholar

[27] Ros-Oton X. and Serra J., The extremal solution for the fractional Laplacian, Calc. Var. Partial Differential Equations 50 (2014), 723–750. 10.1007/s00526-013-0653-1Search in Google Scholar

[28] Servadei R. and Valdinoci E., Variational methods for non-local operators of elliptic type, Discrete Contin. Dyn. Syst. 33 (2013), 2105–2137. 10.3934/dcds.2013.33.2105Search in Google Scholar

[29] Warma M., The p-Laplace operator with the nonlocal Robin boundary conditions on arbitrary open sets, Ann. Mat. Pura Appl. (4) 193 (2014), 203–235. 10.1007/s10231-012-0273-ySearch in Google Scholar

[30] Warma M., A fractional Dirichlet-to-Neumann operator on bounded Lipschitz domains, Commun. Pure Appl. Anal. 14 (2015), 2043–2067. 10.3934/cpaa.2015.14.2043Search in Google Scholar

[31] Warma M., The fractional relative capacity and the fractional Laplacian with Neumann and Robin boundary conditions on open sets, Potential Anal. 42 (2015), 499–547. 10.1007/s11118-014-9443-4Search in Google Scholar

[32] Warma M., Local Lipschitz continuity of the inverse of the fractional p-Laplacian, Hölder type continuity and continuous dependence of solutions to associated parabolic equations on bounded domains, Nonlinear Anal. 135 (2016), 129–157. 10.1016/j.na.2016.01.022Search in Google Scholar

[33] Warma M., The fractional Neumann and Robin boundary conditions for the regional fractional p-Laplacian, NoDEA Nonlinear Differential Equations Appl. (2016), 10.1007/s00030-016-0354-5. 10.1007/s00030-016-0354-5Search in Google Scholar

[34] Zeidler E., Nonlinear Functional Analysis and Applications. II/B: Nonlinear Monotone Operators, Springer, New York, 1990. 10.1007/978-1-4612-0981-2Search in Google Scholar

Received: 2015-09-17
Accepted: 2016-03-02
Published Online: 2016-03-23
Published in Print: 2016-08-01

© 2016 by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Downloaded on 1.3.2026 from https://www.degruyterbrill.com/document/doi/10.1515/ans-2015-5033/html
Scroll to top button