Local Elliptic Regularity for the Dirichlet Fractional Laplacian

Umberto Biccari; Mahamadi Warma; Enrique Zuazua

doi:10.1515/ans-2017-0014

Open Access Published by De Gruyter April 21, 2017

Local Elliptic Regularity for the Dirichlet Fractional Laplacian

Umberto Biccari , Mahamadi Warma and Enrique Zuazua

From the journal Advanced Nonlinear Studies

https://doi.org/10.1515/ans-2017-0014

Abstract

We prove the Wloc2⁢s,p local elliptic regularity of weak solutions to the Dirichlet problem associated with the fractional Laplacian on an arbitrary bounded open set of ℝN. The key tool consists in analyzing carefully the elliptic equation satisfied by the solution locally, after cut-off, to later employ sharp regularity results in the whole space. We do it by two different methods. First working directly in the variational formulation of the elliptic problem and then employing the heat kernel representation of solutions.

Keywords: Fractional Laplacian; Dirichlet Boundary Condition; Weak Solutions; Local Regularity

MSC 2010: 35B65; 35R11; 35S05

1 Introduction

The aim of the present paper is to study the local elliptic regularity of weak solutions to the following Dirichlet problem:

(1.1) { ( - Δ ) s ⁢ u = f in ⁢ Ω , u = 0 on ⁢ ℝ N ∖ Ω ,

where Ω⊂ℝN is an arbitrary bounded open set and s∈(0,1).

Here, f is a given distribution and (-Δ)s denotes the fractional Laplace operator, which is defined as the following singular integral:

(1.2) ( - Δ ) s ⁢ u ⁢ ( x ) := C N , s ⁢ P . V . ∫ ℝ N u ⁢ ( x ) - u ⁢ ( y ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y , x ∈ ℝ N .

In (1.2), CN,s is a normalization constant, given by

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

Γ being the usual Gamma function. Moreover, we have to mention that, for having a completely rigorous definition of the fractional Laplace operator, it is necessary to introduce also the class of functions u for which computing (-Δ)s⁢u makes sense. We postpone this discussion to the next section.

Models involving the fractional Laplacian or other types of non-local operators have been recently used in the description of several complex phenomena for which the classical local approach turns up to be inappropriate or limited. Among others, we mention applications in elasticity [8], turbulence [2], anomalous transport and diffusion [5, 21], porous media flow [32], image processing [14], wave propagation in heterogeneous high contrast media [33]. Also, it is well known that the fractional Laplacian is the generator of s-stable processes, and it is often used in stochastic models with applications, for instance, in mathematical finance [20, 23].

One of the main differences between these non-local models and classical partial differential equations is that the fulfilment of a non-local equation at a point involves the values of the function far away from that point.

Our concern in this article is the study of the local elliptic regularity for weak solutions of the Dirichlet problem (1.1). For this purpose, we firstly remind that, according to [19], we have the following definition of weak solutions.

Definition 1.1

Let f∈W-s,2⁢(Ω¯). A function u∈W0s,2⁢(Ω¯) is said to bea finite energy solution of the Dirichlet problem (1.1) if for every v∈W0s,2⁢(Ω¯), the equality

C N , s 2 ⁢ ∫ ℝ N ∫ ℝ N ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( v ⁢ ( x ) - v ⁢ ( y ) ) | x - y | N + 2 ⁢ s ⁢ 𝑑 x ⁢ 𝑑 y = 〈 f , v 〉 W - s , 2 ⁢ ( Ω ¯ ) , W 0 s , 2 ⁢ ( Ω ¯ )

holds.

We notice that, when 1<p<2, it is not natural to consider finite energy solutions for (1.1), and we shall rather introduce an alternative notion of solution. This will be given by duality with respect to the following class of test functions:

𝒯 ⁢ ( Ω ) = { ϕ : ( - Δ ) s ⁢ ϕ = ψ ⁢ in ⁢ Ω , ϕ = 0 ⁢ in ⁢ ℝ N ∖ Ω , ψ ∈ C 0 ∞ ⁢ ( Ω ) } .

Definition 1.2

Let 1<p<2. We say that u∈L1⁢(Ω) is a weak solution to (1.1) if, for f∈L1⁢(Ω) we have that

∫ Ω u ⁢ ψ ⁢ 𝑑 x = ∫ Ω f ⁢ ϕ ⁢ 𝑑 x

for any ϕ∈𝒯⁢(Ω) with ψ∈C0∞⁢(Ω).

The following Wloc2⁢s,2⁢(Ω)-regularity property is our first main result.

Theorem 1.3

Theorem 1.3 (L2-Local Regularity)

Let f∈W-s,2⁢(Ω¯) and let u∈W0s,2⁢(Ω¯) be the unique weak solution to the Dirichlet problem (1.1). If f∈L2⁢(Ω), then u∈Wloc2⁢s,2⁢(Ω).

This result can be extended to the Lp setting as follows.

Theorem 1.4

Theorem 1.4 (Lp-Local Regularity)

Let f∈W-s,2⁢(Ω¯) and let u∈W0s,2⁢(Ω¯) be the unique weak solution to the Dirichlet problem (1.1). If f∈Lp⁢(Ω) with 1<p<∞, then u∈Wloc2⁢s,p⁢(Ω).

We have to notice that these two theorems are already known when Ω is the whole space ℝN. In fact, they follow by combining several results on Fourier transform and singular integrals contained in [29, Chapter V]. This combination has been done in the reference [3]. On the other hand, when Ω is a bounded open set, to the best of our knowledge, such results are not yet available in the literature.

In Theorems 1.3 and 1.4, W0s,2⁢(Ω¯) denotes the fractional order Sobolev space which consists of all functions u∈Ws,2⁢(ℝN) which are zero on ℝN∖Ω, while W-s,2⁢(Ω¯) is its dual. We will give a more exhaustive description of these spaces in Appendix A at the end of this paper. Moreover, we comment that our results tell that, as it is for the classical Laplace operator (which corresponds to the case s=1), when the right-hand side is in Lp⁢(Ω) the corresponding solution of (1.1) gains locally the maximum possible regularity, that is, it gains locally up to 2⁢s derivatives in Lp⁢(Ω).

Our results complement some previous ones on local and global Sobolev regularity.

In [16, Section 7], Grubb proves that, under the restriction s>Np, the assumption f∈Wt,p⁢(Ω) for some t≥0 implies that the corresponding solution u of (1.1) belongs to Wloct+2⁢s,p⁢(Ω).
In [19, Theorem 17],Leonori, Peral, Primo and Soria show that, if f∈Lm⁢(Ω) for some m≥2⁢NN+2⁢s, then the weak solution u of (1.1) belongs to W0s⁢θ,p⁢(Ω) for some 0<θ<1 and p such that

1 p = 1 m + θ ⁢ ( 1 2 - 1 m ) - 2 ⁢ s ⁢ ( 1 - θ ) N .

This is proved by means of an interpolation argument between Ws,2⁢(Ω) and Lm⁢N/(N-2⁢m⁢s)⁢(Ω). Note however that this global regularity result does not achieve the maximal gain of regularity since 0<s⁢θ<s. On the other hand, a well-known example shows that the optimal global regularity fails (for more details, see [25, Remark 7.2]).
In [6], it is proved that if we take f∈L2⁢(Ω), then the corresponding weak solution of (1.1) satisfies u∈Wloc2⁢s-ε,2⁢(Ω) for all ε>0.

We conclude that our results complement those mentioned above, showing that optimal regularity holds locally away from the boundary, for all s∈(0,1).

Finally, we remind that, for the classical Laplace operator, maximum regularity holds globally provided that the open set is smooth enough. This result fails for the fractional Laplace operator. We refer to Section 5 for a full discussion on this topic and the possible remedies that should involve weighted estimates to take into account the boundary singularities.

The strategy that we will employ to prove our local regularity theorems does not involve interpolation techniques, as in the proof of [19, Theorem 17]. Instead, it will be based on a cut-off argument that will allow us to reduce the problem to the whole space case, for which, as we have mentioned above, the result is already known (see for example Theorem 2.7 below).

In order to develop this technique, the following proposition, which provides a formula for the fractional Laplacian of the product of two functions, will be fundamental (see, e.g., [24] and the references therein).

Proposition 1.5

Let u and v be such that (-Δ)s⁢u and (-Δ)s⁢v exist and

(1.3) ∫ ℝ N | ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( v ⁢ ( x ) - v ⁢ ( y ) ) | | x - y | N + 2 ⁢ s ⁢ 𝑑 y < ∞ .

Then (-Δ)s⁢(u⁢v) exists and is given by

(1.4) ( - Δ ) s ⁢ ( u ⁢ v ) = u ⁢ ( - Δ ) s ⁢ v + v ⁢ ( - Δ ) s ⁢ u - I s ⁢ ( u , v ) ,

where

(1.5) I s ⁢ ( u , v ) ⁢ ( x ) := C N , s ⁢ ∫ ℝ N ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( v ⁢ ( x ) - v ⁢ ( y ) ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y , x ∈ ℝ N .

Remark 1.6

We mention that for example if u,v∈Ws,2⁢(RN) with (-Δ)s⁢u,(-Δ)s⁢v∈L2⁢(RN), then one has (1.3) and thus formula (1.4) holds for such functions.

Formula (1.4), applied to the product of u with a cut-off function η, will be the principal tool for transforming our original problem (1.1) to one in the whole ℝN. Then, for the proof of our main results, we will need to carefully analyze the regularity of the remainder term Is. This analysis will be developed following two different approaches. In the first one, we will consider the fractional Laplacian (-Δ)s as defined in (1.2). In the second one, we will instead use the equivalent characterization of the fractional Laplace operator through the heat semigroup (et⁢Δ)t≥0, given by

(1.6) ( - Δ ) s ⁢ u := 1 Γ ⁢ ( - s ) ⁢ ∫ 0 + ∞ ( e t ⁢ Δ ⁢ u - u ) ⁢ d ⁢ t t 1 + s

(see for instance [30, Section 2.1] and the references therein).We recall that here, Γ⁢(1-s):=-s⁢Γ⁢(-s).

Finally, we mention that this careful analysis of the regularity of the remainder term had been partially developed already in [4, Lemma B1], as a technical tool for obtaining the results therein presented. This has been one of the main motivations that led to the development of the present work.

The paper is organized as follows. In Section 2, we present some preliminary tools that we shall use in the proof of our main results. In Section 3, we give the proof of Theorems 1.3 and 1.4 using the integral representation of the fractional Laplacian. In Section 4, we use the second approach which is based on the representation (1.6). Finally, in Section 5, we present some open problems and perspectives that are closely related to our work.

2 Preliminaries

In this section, we introduce some preliminary result that will be useful for the proof of our main Theorems 1.3 and 1.4.

We start by giving a more rigorous definition of the fractional Laplace operator, as we have anticipated in Section 1. Let

ℒ s 1 ⁢ ( ℝ N ) := { u : ℝ N → ℝ ⁢ measurable , ∫ ℝ N | u ⁢ ( x ) | ( 1 + | x | ) N + 2 ⁢ s ⁢ 𝑑 x < ∞ } .

For u∈ℒs1⁢(ℝN) and ε>0 we set

( - Δ ) ε s ⁢ u ⁢ ( x ) := C N , s ⁢ ∫ { y ∈ ℝ N : | x - y | > ε } u ⁢ ( x ) - u ⁢ ( y ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y , x ∈ ℝ N .

The fractional Laplace operator(-Δ)s is then defined by the following singular integral:

(2.1) ( - Δ ) s ⁢ u ⁢ ( x ) = C N , s ⁢ P . V . ∫ ℝ N u ⁢ ( x ) - u ⁢ ( y ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y = lim ε ↓ 0 ⁡ ( - Δ ) ε s ⁢ u ⁢ ( x ) , x ∈ ℝ N ,

provided that the limit exists.

We notice that if 0<s<12 and u is smooth, for example bounded and Lipschitz continuous on ℝN, then the integral in (2.1) is in fact not really singular near x (see, e.g., [7, Remark 3.1]). Moreover, ℒs1⁢(ℝN) is the right space for which v:=(-Δ)εs⁢u exists for every ε>0, v being also continuous at the continuity points of u.

The following result of existence and uniqueness of weak solutions to the Dirichlet problem (1.1) is by now well known (see, e.g., [19, Theorem 12]).

Proposition 2.1

Let Ω⊂RN be an arbitrary bounded open set and 0<s<1. Then for every f∈W-s,2⁢(Ω¯), the Dirichlet problem (1.1) has a unique finite energy solution u∈W0s,2⁢(Ω¯). In addition, there exists a constant C>0 such that

(2.2) ∥ u ∥ W 0 s , 2 ⁢ ( Ω ¯ ) ≤ C ⁢ ∥ f ∥ W - s , 2 ⁢ ( Ω ¯ ) .

Proof.

For the sake of completeness we include the proof. We recall that a complete description of the functional setting in which we are working is presented in Appendix A.Moreover, we recall that, according to Definition 1.1, a function u∈W0s,2⁢(Ω¯) is said to be a weak solution of the Dirichlet problem (1.1) if for every v∈W0s,2⁢(Ω¯), the equality

(2.3) C N , s 2 ⁢ ∫ ℝ N ∫ ℝ N ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( v ⁢ ( x ) - v ⁢ ( y ) ) | x - y | N + 2 ⁢ s ⁢ 𝑑 x ⁢ 𝑑 y = 〈 f , v 〉 W - s , 2 ⁢ ( Ω ¯ ) , W 0 s , 2 ⁢ ( Ω ¯ )

holds. Hence, given u,v∈W0s,2⁢(Ω¯) let us consider the bilinear form

(2.4) ℰ ⁢ ( u , v ) = C N , s 2 ⁢ ∫ ℝ N ∫ ℝ N ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( v ⁢ ( x ) - v ⁢ ( y ) ) | x - y | N + 2 ⁢ s ⁢ 𝑑 x ⁢ 𝑑 y ,

which is symmetric, continuous and coercive.

Thus by the classical Lax–Milgram theorem, for every

f ∈ ( W 0 s , 2 ( Ω ¯ ) ) ⋆ = : W - s , 2 ( Ω ¯ ) ,

there exists a unique u∈W0s,2⁢(Ω¯) such that the equality (2.3) holds for every v∈W0s,2⁢(Ω¯). We have shown that (1.1) has a unique weak solution u∈W0s,2⁢(Ω¯). Taking v=u as a test function in (2.3) and using (A.5), we get that

C ⁢ ∥ u ∥ W 0 s , 2 ⁢ ( Ω ¯ ) 2 = 〈 f , u 〉 W - s , 2 ⁢ ( Ω ¯ ) , W 0 s , 2 ⁢ ( Ω ¯ ) ≤ ∥ f ∥ W - s , 2 ⁢ ( Ω ¯ ) ⁢ ∥ u ∥ W 0 s , 2 ⁢ ( Ω ¯ ) .

We have shown (2.2) and the proof is finished.∎

Remark 2.2

Notice that also for 1<p<2 existence and uniqueness of a weak solution to problem (1.1)are guaranteed by [19, Theorem 28].

Remark 2.3

Notice that [19, Theorem 12] holds for a more general non-local operator where the kernel |x-y|-N-2⁢s is replaced by a general symmetric kernel K⁢(x,y) satisfying λ≤K⁢(x,y)⁢|x-y|N+2⁢s≤λ-1 for all (x,y)∈ℝN×ℝN, x≠y, and for some constant 0<λ≤1.

Remark 2.4

Let f∈W-s,2⁢(Ω¯) and let u∈W0s,2⁢(Ω¯) be the weak solution of the Dirichlet problem (1.1). We notice that it follows from the Sobolev embeddings (A.2) and (A.1) that if N<2⁢s, then u∈C0,s-N/2⁢(Ω¯) and if N=2⁢s, then u∈Lq⁢(Ω) for every 1≤q<∞.

The following lemma, giving a precise Lq-regularity of weak solutions and complementing the results in [19, Theorem 16] will be useful in the sequel.

Lemma 2.5

Assume that N>2⁢s and let f∈Lp⁢(Ω) for some p≥2⁢NN+2⁢s. Then (1.1) has a unique weak solution u. In addition the following assertions hold.

If p > N 2 ⁢ s , then u ∈ L ∞ ⁢ ( Ω ) and there exists a constant C > 0 such that

∥ u ∥ L ∞ ⁢ ( Ω ) ≤ C ⁢ ∥ f ∥ L p ⁢ ( Ω ) .
If 2 ⁢ N N + 2 ⁢ s ≤ p ≤ N 2 ⁢ s , then u ∈ L q ⁢ ( Ω ) for every q satisfying p ≤ q < N ⁢ p N - 2 ⁢ s ⁢ p and there exists a constant C > 0 such that

∥ u ∥ L q ⁢ ( Ω ) ≤ C ⁢ ∥ f ∥ L p ⁢ ( Ω ) .

Proof.

Let f∈Lp⁢(Ω) for some p≥2⁢NN+2⁢s. SinceW0s,2⁢(Ω¯)↪L2⁢N/(N-2⁢s)⁢(Ω) (see (A.1)), we have that

L p ⁢ ( Ω ) ↪ L 2 ⁢ N N + 2 ⁢ s ⁢ ( Ω ) ↪ W - s , 2 ⁢ ( Ω ¯ ) .

Thus (1.1) has a unique weak solution u.

Let ℰ with domain D⁢(ℰ)=W0s,2⁢(Ω¯) be the bilinear, symmetric, continuous and coercive form defined in (2.4). As we have shown in the proof of Proposition 2.1, for every f∈W-s,2⁢(Ω¯) there exists a unique u∈W0s,2⁢(Ω¯) such that

ℰ ⁢ ( u , v ) = 〈 f , v 〉 W - s , 2 ⁢ ( Ω ¯ ) , W 0 s , 2 ⁢ ( Ω ¯ ) for all ⁢ v ∈ W 0 s , 2 ⁢ ( Ω ¯ ) .

This defines an operator 𝒜:W0s,2⁢(Ω¯)→W-s,2⁢(Ω¯) which is continuous and coercive. Let AD be the part of 𝒜 in L2⁢(Ω), in the sense that

D ⁢ ( A D ) := { u ∈ W 0 s , 2 ⁢ ( Ω ¯ ) : 𝒜 ⁢ u ∈ L 2 ⁢ ( Ω ) } , A D ⁢ u = 𝒜 ⁢ u .

Using an integration by parts argument, one can show that AD is given precisely by

D ⁢ ( A D ) = { u ∈ W 0 s , 2 ⁢ ( Ω ¯ ) : ( - Δ ) s ⁢ u ∈ L 2 ⁢ ( Ω ) } , A D = ( - Δ ) s ⁢ u .

Then AD is the realization in L2⁢(Ω) of the operator (-Δ)s with the Dirichlet boundary condition u=0 on ℝN∖Ω.The operator AD has a compact resolvent (this follows from the compactness of the embedding from W0s,2⁢(Ω¯) into L2⁢(Ω), see (A.1)) and its first eigenvalue λ1>0.

In addition -AD generates a submarkovian strongly continuous semigroup (e-t⁢AD)t≥0 which is also ultracontractive in the sense that the semigroup maps Lr⁢(Ω) into Lm⁢(Ω) for every t>0 and 1≤r≤m≤∞. More precisely, following line by line the proof of [13, Theorem 2.16] by using the appropriate estimates, we get that, for every 1≤r≤m≤∞ there exists a constant C>0 such that for every f∈Lr⁢(Ω) and t>0,

(2.5) ∥ e - t ⁢ A D ⁢ f ∥ L m ⁢ ( Ω ) ≤ C ⁢ e - λ 1 ⁢ ( 1 r - 1 m ) ⁢ t - N 2 ⁢ s ⁢ ( 1 r - 1 m ) ⁢ ∥ f ∥ L r ⁢ ( Ω ) .

Since the operator AD is invertible, it follows from the abstract result in[10, p. 55, Theorem 1.10] that for every f∈L1⁢(Ω)∩W-s,2⁢(Ω¯), the unique solution u of the Dirichlet problem (1.1) is given by

u = A D - 1 ⁢ f = ∫ 0 ∞ e - t ⁢ A D ⁢ f ⁢ 𝑑 t .

(a) Assume that p>N2⁢s. Then applying (2.5) with r=p and m=∞, we get that

∥ u ∥ L ∞ Ω ) ≤ ∫ 0 ∞ ∥ e - t ⁢ A D ⁢ f ∥ L ∞ ⁢ ( Ω ) ⁢ 𝑑 t ≤ C ⁢ ∫ 0 ∞ e - λ 1 p ⁢ t - N 2 ⁢ s ⁢ p ⁢ ∥ f ∥ L p ⁢ ( Ω ) ⁢ 𝑑 t

= C ⁢ ( ∫ 1 ∞ e - λ 1 p ⁢ t - N 2 ⁢ s ⁢ p ⁢ 𝑑 t + ∫ 0 1 e - λ 1 p ⁢ t - N 2 ⁢ s ⁢ p ⁢ 𝑑 t ) ⁢ ∥ f ∥ L p ⁢ ( Ω ) .

The first integral in the right-hand side of the previous estimate is always finite. The second integral will be finite if 1-N2⁢s⁢p>0. This is equivalent to p>N2⁢s and we have shown part (a).

(b) Assume that 2⁢NN+2⁢s≤p≤N2⁢s and let p≤q<N⁢pN-2⁢s⁢p.Then applying (2.5) with r=p and m=q, we get that

∥ u ∥ L q ⁢ ( Ω ) ≤ ∫ 0 ∞ ∥ e - t ⁢ A D ⁢ f ∥ L q ⁢ ( Ω ) ⁢ 𝑑 t ≤ C ⁢ ∫ 0 ∞ e - λ 1 ⁢ ( 1 p - 1 q ) ⁢ t - N 2 ⁢ s ⁢ ( 1 p - 1 q ) ⁢ ∥ f ∥ L r ⁢ ( Ω ) ⁢ 𝑑 t

= C ⁢ ( ∫ 1 ∞ e - λ 1 ⁢ ( 1 p - 1 q ) ⁢ t - N 2 ⁢ s ⁢ ( 1 p - 1 q ) ⁢ 𝑑 t + ∫ 0 1 e - λ 1 ⁢ ( 1 p - 1 q ) ⁢ t - N 2 ⁢ s ⁢ ( 1 p - 1 q ) ⁢ 𝑑 t ) ⁢ ∥ f ∥ L p ⁢ ( Ω ) .

As above, the first integral is always finite and the second integral will be finite if 1-N2⁢s⁢(1p-1q)>0. This is equivalent to q<N⁢pN-2⁢s⁢p.We have shown part (b) and the proof is finished.∎

Remark 2.6

Assertion (b) has been also proved in [19, Theorem 16]. There,Leonori, Peral, Primo and Soria obtained the result adapting Moser’s method in [22], which allows to obtain the Lq⁢(Ω)-regularity of the solution u to (1.1) employing functions depending nonlinearly on the solution. To the best of our knowledge, our approach to the proof of Lemma 2.5 (b) is new.

In our discussion, the following result of regularity on the whole space ℝN will play an important role.

Theorem 2.7

Let F∈W-s,2⁢(RN):=(Ws,2⁢(RN))⋆ andlet u∈Ws,2⁢(RN) be the weak solution to the fractional Poisson type equation

(2.6) ( - Δ ) s ⁢ u = F in ⁢ ℝ N .

If F∈Lp⁢(RN) with 1<p<∞, then u∈W2⁢s,p⁢(RN).

Theorem 2.7 is a classical result whose proof can be done by combining several results on singular integrals and Fourier transform contained in [29, Chapter V, Section 3.3]; see, in particular, formulas (38), (40) and Theorem 3 therein. These mentioned results have been put together in the reference [3]. In particular, our Theorem 2.7 is obtained as a consequenceof [3, Corollary 6.7 and Lemma 6.9].

3 Proof of Theorems 1.3 and 1.4: First Approach

3.1 Proof of the L2-Local Regularity Theorem

Proof of Theorem 1.3.

As we have mentioned above, our strategy is based on a cut-off argument that will allow us to show that the solutions of the fractional Dirichlet problem in Ω, after cut-off, are solutions of the elliptic problem on the whole space ℝN, for which Theorem 2.7 holds. For this purpose, given two open subsets ω and ω~ of the domain Ω such that ω~⋐ω⋐Ω, we introduce a cut-off function η∈𝒟⁢(ω) such that

(3.1) { η ⁢ ( x ) ≡ 1 if ⁢ x ∈ ω ~ , 0 ≤ η ⁢ ( x ) ≤ 1 if ⁢ x ∈ ω ∖ ω ~ , η ⁢ ( x ) = 0 if ⁢ x ∈ ℝ N ∖ ω .

Let f∈W-s,2⁢(Ω¯) and let u∈W0s,2⁢(Ω¯) be the unique weak solution to the Dirichlet problem (1.1).

Let ω and η∈𝒟⁢(ω) be respectively the set and the cut-off function constructed in (3.1). We consider the function u⁢η. It is clear that u⁢η∈Ws,2⁢(ℝN). It follows from Proposition 1.5 and Remark 1.6 that

(3.2) ( - Δ ) s ⁢ ( u ⁢ η ) - η ⁢ ( - Δ ) s ⁢ u = u ⁢ ( - Δ ) s ⁢ η - I s ⁢ ( u , η ) .

Let

g := u ⁢ ( - Δ ) s ⁢ η - I s ⁢ ( u , η ) .

We claim that g∈L2⁢(ℝN). In fact, there exists a constant C>0, independent of u, such that

(3.3) ∥ g ∥ L 2 ⁢ ( ℝ N ) ≤ C ⁢ ∥ u ∥ W 0 s , 2 ⁢ ( Ω ¯ ) .

Since u=0 on ℝN∖Ω and (-Δ)s⁢η∈L∞⁢(ℝN), we have that

(3.4) ∥ u ⁢ ( - Δ ) s ⁢ η ∥ L 2 ⁢ ( ℝ N ) 2 = ∫ Ω | u ⁢ ( - Δ ) s ⁢ η | 2 ⁢ 𝑑 x ≤ ∥ ( - Δ ) s ⁢ η ∥ L ∞ ⁢ ( Ω ) 2 ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) 2 .

Now, recall from (1.5) that for a.e. x∈ℝN,

I s ⁢ ( u , η ) ⁢ ( x ) := C N , s ⁢ ∫ ℝ N ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( η ⁢ ( x ) - η ⁢ ( y ) ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y

= C N , s ⁢ ∫ Ω ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( η ⁢ ( x ) - η ⁢ ( y ) ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y ⏟ = ⁣ : 𝕀 1 ⁢ ( x ) + C N , s ⁢ η ⁢ ( x ) ⁢ ∫ ℝ N ∖ Ω u ⁢ ( x ) - u ⁢ ( y ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y ⏟ = ⁣ : 𝕀 2 ⁢ ( x ) , x ∈ ℝ N .

Let us start to estimate the term 𝕀1⁢(x). Using the Cauchy–Schwarz inequality, we get that

(3.5) | 𝕀 1 ⁢ ( x ) | ≤ C N , s ⁢ ( ∫ Ω | u ⁢ ( x ) - u ⁢ ( y ) | 2 | x - y | N + 2 ⁢ s ⁢ 𝑑 y ) 1 2 ⁢ ( ∫ Ω ( | η ( x ) - η ( y ) | 2 | x - y | N + 2 ⁢ s ⁢ 𝑑 y ) 1 2 .

Let x∈Ω be fixed and R>0 such that Ω⊂B⁢(x,R). Since η is a smooth function (in particular Lipschitz continuous on ℝN), we have that there exists a constant C>0 (depending on η) such that

∫ Ω | η ⁢ ( x ) - η ⁢ ( y ) | 2 | x - y | N + 2 ⁢ s ⁢ 𝑑 y ≤ C ⁢ ∫ Ω d ⁢ y | x - y | N + 2 ⁢ s - 2 ≤ C ⁢ ∫ B ⁢ ( x , R ) d ⁢ y | x - y | N + 2 ⁢ s - 2 ≤ C .

Using the preceding estimate and (3.5), we get that

(3.6) ∫ ℝ N | 𝕀 1 ⁢ ( x ) | 2 ⁢ 𝑑 x ≤ C ⁢ ∫ ℝ N ∫ Ω | u ⁢ ( x ) - u ⁢ ( y ) | 2 | x - y | N + 2 ⁢ s ⁢ 𝑑 y ⁢ 𝑑 x ≤ C ⁢ ∥ u ∥ W 0 s , 2 ⁢ ( Ω ¯ ) 2 .

Concerning the term 𝕀2, we notice that 𝕀2=0 on ℝN∖ω.In addition, using the Cauchy–Schwarz inequality, we get that

(3.7) | 𝕀 2 ⁢ ( x ) | 2 ≤ C N , s 2 ⁢ ∫ ℝ N ∖ Ω η 2 ⁢ ( x ) ⁢ d ⁢ y | x - y | N + 2 ⁢ s ⁢ ∫ ℝ N ∖ Ω | u ⁢ ( x ) - u ⁢ ( y ) | 2 | x - y | N + 2 ⁢ s ⁢ 𝑑 y .

For any y∈ℝN∖Ω, we have that

η 2 ⁢ ( x ) | x - y | N + 2 ⁢ s = χ ω ¯ ⁢ ( x ) ⁢ η 2 ⁢ ( x ) | x - y | N + 2 ⁢ s ≤ χ ω ¯ ⁢ ( x ) ⁢ η 2 ⁢ ( x ) ⁢ sup x ∈ ω ¯ ⁡ 1 | x - y | N + 2 ⁢ s .

Thus there exists a constant C>0 such that

(3.8) ∫ ℝ N ∖ Ω η 2 ⁢ ( x ) ⁢ d ⁢ y | x - y | N + 2 ⁢ s ≤ χ ω ¯ ⁢ ( x ) ⁢ η 2 ⁢ ( x ) ⁢ ∫ ℝ N ∖ Ω d ⁢ y dist ( y , ∂ ω ¯ ) N + 2 ⁢ s ≤ C ⁢ χ ω ¯ ⁢ ( x ) ⁢ η 2 ⁢ ( x ) ,

where we have used that the integral is finite which follows from the facts that dist⁡(∂⁡Ω,∂⁡ω¯)≥δ>0, that the distance function dist⁡(y,∂⁡ω¯) grows linearly as y tends to infinity and that N+2⁢s>N.

Since χω¯⁢η2∈L∞⁢(ω) and using (3.7) and (3.8), we get that there exists a constant C>0 such that

(3.9) ∫ ℝ N | 𝕀 2 ⁢ ( x ) | 2 ⁢ 𝑑 x = ∫ ω | 𝕀 2 ⁢ ( x ) | 2 ⁢ 𝑑 x ≤ C ⁢ ∫ ω ∫ ℝ N ∖ Ω | u ⁢ ( x ) - u ⁢ ( y ) | 2 | x - y | N + 2 ⁢ s ⁢ 𝑑 y ⁢ 𝑑 y ≤ C ⁢ ∥ u ∥ W 0 s , 2 ⁢ ( Ω ¯ ) 2 .

Now estimate (3.3) follows from (3.4), (3.6), (3.9), and the claim is proved. We have shown that η⁢u is a weak solution to the Poisson equation (2.6) with F given by F=η⁢(-Δ)s⁢u+g∈L2⁢(ℝN). It follows from Theorem 2.7 that (η⁢u)∈W2⁢s,2⁢(ℝN). Thus u∈Wloc2⁢s,2⁢(Ω) and the proof is complete.∎

3.2 Proof of the Lp-Local Regularity Theorem

We will now use Theorem 1.3 to prove our local regularity result in the general Lp setting.

Proof of Theorem 1.4.

We start by noticing that, assuming f∈Lp⁢(Ω)∩W-s,2⁢(Ω¯), we have that (1.1) has a unique weak solution u∈W0s,2⁢(Ω¯). We divide the proof into two steps.

Step 1: 1<p<2. If 1<p<2, then, according to Theorem A.2, u∈Ws,p⁢(Ω). In particular, u∈Lp⁢(Ω).

Let ω and η∈𝒟⁢(ω) be respectively the set and the cut-off function constructed in (3.1). We consider the function u⁢η∈Ws,p⁢(ℝN). As in the proof of Theorem 1.3, we have that (-Δ)s⁢(u⁢η) is given by

( - Δ ) s ⁢ u ⁢ η = η ⁢ f + u ⁢ ( - Δ ) s ⁢ η - I s ⁢ ( u , η ) ,

where the term Is⁢(u,η) has been introduced in (1.5). Let ω1,ω2 be open sets such that

ω ¯ ⊂ ω 1 ⊂ ω ¯ 1 ⊂ ω 2 ⊂ ω ¯ 2 ⊂ Ω .

Since the function η and the set ω in (3.1) are arbitrary, it follows that u∈Ws,p⁢(ω2). Thus we have u∈Ws,p⁢(ω2)∩Lp⁢(Ω). Let

g := u ⁢ ( - Δ ) s ⁢ η - I s ⁢ ( u , η ) .

We now claim that g∈Lp⁢(ℝN) and there exists a constant C>0 such that

(3.10) ∥ g ∥ L p ⁢ ( ℝ N ) ≤ C ⁢ ( ∥ u ∥ W s , p ⁢ ( ω 2 ) + ∥ u ∥ L p ⁢ ( Ω ) ) .

Indeed, it is clear that g is defined on all ℝN. Moreover,

(3.11) ∥ u ⁢ ( - Δ ) s ⁢ η ∥ L p ⁢ ( ℝ N ) p = ∫ Ω | u ⁢ ( - Δ ) s ⁢ η | p ⁢ 𝑑 x ≤ ∥ ( - Δ ) s ⁢ η ∥ L ∞ ⁢ ( Ω ) p ⁢ ∥ u ∥ L p ⁢ ( Ω ) p .

For estimating the term Is, we use the decomposition

I s ⁢ ( u , η ) ⁢ ( x ) := C N , s ⁢ ∫ ℝ N ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( η ⁢ ( x ) - η ⁢ ( y ) ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y

= C N , s ⁢ ∫ ω 1 ( u ⁢ ( x ) - u ⁢ ( y ) ) ⁢ ( η ⁢ ( x ) - η ⁢ ( y ) ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y ⏟ = ⁣ : 𝕀 1 ⁢ ( x ) + C N , s ⁢ η ⁢ ( x ) ⁢ ∫ ℝ N ∖ ω 1 u ⁢ ( x ) - u ⁢ ( y ) | x - y | N + 2 ⁢ s ⁢ 𝑑 y ⏟ = ⁣ : 𝕀 2 ⁢ ( x ) , x ∈ ℝ N .

Let p′:=pp-1. Using the Hölder inequality, we get that for a.e. x∈ℝN,

(3.12) | 𝕀 1 ⁢ ( x ) | ≤ C N , s ⁢ ( ∫ ω 1 | u ⁢ ( x ) - u ⁢ ( y ) | p | x - y | N + s ⁢ p ⁢ 𝑑 y ) 1 p ⁢ ( ∫ ω 1 | η ⁢ ( x ) - η ⁢ ( y ) | p ′ | x - y | N + s ⁢ p ′ ⁢ 𝑑 y ) 1 p ′ .

Let x∈ω1 be fixed and R>0 such that ω1⊂B⁢(x,R). Using the Lipschitz continuity of the function η, we obtain that there exists a constant C>0 such that

(3.13) ∫ ω 1 | η ⁢ ( x ) - η ⁢ ( y ) | p ′ | x - y | N + s ⁢ p ′ ⁢ 𝑑 y ≤ C ⁢ ∫ ω 1 d ⁢ y | x - y | N + s ⁢ p ′ - p ′ ≤ C ⁢ ∫ B ⁢ ( x , R ) d ⁢ y | x - y | N + s ⁢ p ′ - p ′ ≤ C .

Now, using (3.12), (3.13) and (A.7), we get that

∫ ℝ N | 𝕀 1 ⁢ ( x ) | p ⁢ 𝑑 x ≤ C ⁢ ( ∫ ω 2 ∫ ω 1 | u ⁢ ( x ) - u ⁢ ( y ) | p | x - y | N + s ⁢ p ⁢ 𝑑 y ⁢ 𝑑 x + ∫ ℝ N ∖ ω 2 ∫ ω 1 | u ⁢ ( x ) - u ⁢ ( y ) | p | x - y | N + s ⁢ p ⁢ 𝑑 y ⁢ 𝑑 x )

≤ C ⁢ ( ∥ u ∥ W s , p ⁢ ( ω 2 ) p + ∫ ℝ N ∖ ω 2 ∫ ω 1 | u ⁢ ( x ) | p + | u ⁢ ( y ) | p ( 1 + | x | ) N + s ⁢ p ⁢ 𝑑 y ⁢ 𝑑 x )

(3.14) ≤ C ⁢ ( ∥ u ∥ W s , p ⁢ ( ω 2 ) p + ∥ u ∥ L p ⁢ ( Ω ) p ) ,

where we have also used that u=0 on ℝN∖Ω.Recall that 𝕀2=0 on ℝN∖ω.Then using the Hölder inequality, we get that

(3.15) | 𝕀 2 ⁢ ( x ) | p ≤ C ⁢ ( ∫ ℝ N ∖ ω 1 η p ′ ⁢ ( x ) ⁢ d ⁢ y | x - y | N + s ⁢ p ′ ) p - 1 ⁢ ∫ ℝ N ∖ ω 1 | u ⁢ ( x ) - u ⁢ ( y ) | p | x - y | N + s ⁢ p ⁢ 𝑑 y .

For any y∈ℝN∖ω1, we have that

η p ′ ⁢ ( x ) | x - y | N + s ⁢ p ′ = χ ω ¯ ⁢ ( x ) ⁢ η p ′ ⁢ ( x ) | x - y | N + s ⁢ p ′ ≤ χ ω ¯ ⁢ ( x ) ⁢ η p ′ ⁢ ( x ) ⁢ sup x ∈ ω ¯ ⁡ 1 | x - y | N + s ⁢ p ′ .

So there exists a constant C>0 such that

(3.16) ∫ ℝ N ∖ ω 1 η p ′ ⁢ ( x ) ⁢ d ⁢ y | x - y | N + s ⁢ p ′ ≤ χ ω ¯ ⁢ ( x ) ⁢ η p ′ ⁢ ( x ) ⁢ ∫ ℝ N ∖ ω 1 d ⁢ y dist ( y , ∂ ω ¯ ) N + s ⁢ p ′ ≤ C ⁢ χ ω ¯ ⁢ ( x ) ⁢ η p ′ ⁢ ( x ) .

In (3.16) we have also used that the integral is finite which follows from the fact that dist⁡(∂⁡ω1,∂⁡ω¯)≥δ>0 together with the fact that dist⁡(y,∂⁡ω¯) grows linearly as y tends to infinity and N+s⁢p′>N.

Since χω¯⁢ηp′∈L∞⁢(ω) and using (3.15), (3.16) and (A.7), we also get that there exists a constant C>0 such that

∫ ℝ N | 𝕀 2 ⁢ ( x ) | p ⁢ 𝑑 x = ∫ ω | 𝕀 2 ⁢ ( x ) | p ⁢ 𝑑 x ≤ C ⁢ ∫ ω ∫ ℝ N ∖ ω 1 | u ⁢ ( x ) - u ⁢ ( y ) | p | x - y | N + s ⁢ p ⁢ 𝑑 y ⁢ 𝑑 x

(3.17) ≤ C ⁢ ∫ ω ∫ ℝ N ∖ ω 1 | u ⁢ ( x ) | p + | u ⁢ ( y ) | p ( 1 + | y | ) N + s ⁢ p ⁢ 𝑑 y ⁢ 𝑑 x ≤ C ⁢ ∥ u ∥ L p ⁢ ( Ω ) p ,

where we have used again that u=0 on ℝN∖Ω.Estimate (3.10) follows from (3.11), (3.14), (3.17) and we have shown the claim. We therefore proved that η⁢u is a weak solution to the Poisson equation (2.6) with F given by F=η⁢f+g. Since F∈Lp⁢(ℝN), it follows from Theorem 2.7 that η⁢u∈W2⁢s,p⁢(ℝN). Thus u∈Wloc2⁢s,p⁢(Ω) and the proof for 1<p<2 is concluded.

Step 2: p≥2.Let f∈W-s,2⁢(Ω¯) and let u∈W0s,2⁢(Ω¯) be the weak solution to the Dirichlet problem (1.1). Let ω and η∈𝒟⁢(ω) be respectively the set and the cut-off function constructed in (3.1). We consider the function u⁢η∈Ws,2⁢(ℝN). Assume that f∈Lp⁢(Ω) with p≥2. As in the proof of Theorem 1.3, we have that (-Δ)s⁢(u⁢η) is given by (3.2).Since by assumption f∈Lp⁢(Ω)↪L2⁢(Ω), it follows from Theorem 1.3 that u⁢η∈W2⁢s,2⁢(ℝN).

(a) Applying Theorem A.2 (a) with r=2⁢s and p=2, we get that W2⁢s,2⁢(ℝN)↪Ws,2⁢N/(N-2⁢s)⁢(ℝN). We have shown that u⁢η∈Ws,2⁢N/(N-2⁢s)⁢(ℝN). Let ω1,ω2 be open sets such that

ω ¯ ⊂ ω 1 ⊂ ω ¯ 1 ⊂ ω 2 ⊂ ω ¯ 2 ⊂ Ω .

Since the function η and the set ω in (3.1) are arbitrary, it follows from the observation u⁢η∈Ws,2⁢N/(N-2⁢s)⁢(ℝN) that u∈Ws,2⁢N/(N-2⁢s)⁢(ω2).Let q:=min⁡{p,2⁢NN-2⁢s}. We notice that q≥2. Applying again Theorem A.2 (a) with r=2⁢s and p=2 and the above q, we also get that W2⁢s,2⁢(ω2)↪Ws,q⁢(ω2). We have shown that u∈Ws,q⁢(ω2). Since by hypothesis, u∈W0s,2⁢(Ω¯), it follows from the Sobolev embedding (A.1) thatu∈L2⁢N/(N-2⁢s)⁢(Ω)↪Lq⁢(Ω). Thus u∈Ws,q⁢(ω2)∩Lq⁢(Ω).Let

g := u ⁢ ( - Δ ) s ⁢ η - I s ⁢ ( u , η ) .

Also in this case, it is possible to prove that g∈Lq⁢(ℝN) and there exists a constant C>0 such that

(3.18) ∥ g ∥ L q ⁢ ( ℝ N ) ≤ C ⁢ ( ∥ u ∥ W s , q ⁢ ( ω 2 ) + ∥ u ∥ L q ⁢ ( Ω ) ) .

We omit the proof of (3.18); it is totally analogous to the one we made in Step 1. As before, we have proved that η⁢u is a weak solution to the Poisson equation (2.6) with F given by F=η⁢(-Δ)s⁢u+g. Since F∈Lq⁢(ℝN), it follows from Theorem 2.7 that η⁢u∈W2⁢s,q⁢(ℝN). Thus u∈Wloc2⁢s,q⁢(Ω). If 2≤p≤2⁢NN-2⁢s, then the proof is finished.

(b) Assume that p>2⁢NN-2⁢s. Since u∈Wloc2⁢s,q⁢(Ω), we have that u∈W2⁢s,q⁢(ω2). This implies that u∈Ws,q1⁢(ω2) with q1:=min⁡{p,N⁢qN-s⁢q}=min⁡{p,2⁢NN-4⁢s}. It also follows from Lemma 2.5 that u∈Lq1⁢(Ω). We have shown that u∈Ws,q1⁢(ω2)∩Lq1⁢(Ω).Now proceeding as in part (a), we get that u∈Wloc2⁢s,q1⁢(Ω).Here also if 2≤p≤2⁢NN-4⁢s, then the proof is finished. Otherwise, iterating we will get that u∈Wloc2⁢s,qj⁢(Ω) with qj=min⁡{p,2⁢NN-s⁢j} for all j≥2. Hence, we can find j∈ℕ, j≥2, such that 2≤p≤2⁢NN-s⁢j. The proof of Theorem 1.4 is finished.∎

4 The Approach Using the Heat Semigroup Representation

One of the main passages in the proof of Theorems 1.3 and 1.4 has been to show that, after having applied the cut-off function η, the remainder g, that we obtain applying (1.4) to the product η⁢u, belongs to Lp⁢(ℝN) if f belongs to Lp⁢(Ω). In this section, we present an alternative proof of this fact, using the characterization of the fractional Laplacian through the heat semigroup introduced in (1.6).

The heat equation representation of the operator looks a priori local and this will allow us to give a very precise information on the commutator, in particular in terms of the order of regularity and the localization properties.

Before going further into our discussion, we first need to describe how the operator introduced in (1.6) behaves when it is applied to the function η⁢u. For simplicity of notation, let us define

ϱ ⁢ ( t ) := e t ⁢ Δ ⁢ ( η ⁢ u ) , t ≥ 0 .

Then, by definition, we have that ϱ satisfies the following heat equation on ℝN:

(4.1) ϱ t - Δ ⁢ ϱ = 0 , t > 0 , ϱ ⁢ ( 0 ) = η ⁢ u .

Furthermore, the solution of (4.1) can be written in the form ϱ=ϕ⁢η+z with

(4.2) ϕ t - Δ ⁢ ϕ = 0 , t > 0 , ϕ ⁢ ( 0 ) = u

and

(4.3) z t - Δ ⁢ z = 2 ⁢ div ⁡ ( ϕ ⁢ ∇ ⁡ η ) - ϕ ⁢ Δ ⁢ η , t > 0 , z ⁢ ( 0 ) = 0 .

Finally, we can trivially compute

( - Δ ) s ⁢ ( η ⁢ u ) = 1 Γ ⁢ ( - s ) ⁢ ∫ 0 + ∞ ( ϱ ⁢ ( t ) - ϱ ⁢ ( 0 ) ) ⁢ d ⁢ t t 1 + s

= 1 Γ ⁢ ( - s ) ⁢ ∫ 0 + ∞ ( η ⁢ ϕ ⁢ ( t ) + z ⁢ ( t ) - η ⁢ u ⁢ ( t ) ) ⁢ d ⁢ t t 1 + s

= η Γ ⁢ ( - s ) ⁢ ∫ 0 + ∞ ( e t ⁢ Δ ⁢ u - u ) ⁢ d ⁢ t t 1 + s + 1 Γ ⁢ ( - s ) ⁢ ∫ 0 + ∞ z ⁢ ( t ) t 1 + s ⁢ 𝑑 t .

Therefore we find an expression of the type

( - Δ ) s ⁢ ( η ⁢ u ) = η ⁢ ( - Δ ) s ⁢ u + g ,

where the remainder term g is given by

(4.4) g ⁢ ( x ) := 1 Γ ⁢ ( - s ) ⁢ ∫ 0 + ∞ z ⁢ ( x , t ) t 1 + s ⁢ 𝑑 t , x ∈ ℝ N .

4.1 Proof of the L2-Regularity of g

Keeping in mind the notations that we have just introduced, we can now prove the following result.

Lemma 4.1

Let u∈W0s,2⁢(Ω¯) and let η be the cut-off function introduced in (3.1). Moreover, let g be the remainder term in the expression

( - Δ ) s ⁢ ( η ⁢ u ) = η ⁢ ( - Δ ) s ⁢ u + g .

Then, there exists a constant C>0 (independent of u) such that

(4.5) ∥ g ∥ L 2 ⁢ ( ℝ N ) ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) .

Proof.

According to the expression (4.4), to estimate the L2-norm of g, it will be enough to obtain suitable bounds of the L2-norm of z. For this purpose, we notice that the solution of (4.3) can be computed explicitly as

(4.6) z ⁢ ( x , t ) = ∫ 0 t ∫ ℝ N G ⁢ ( x - y , t - τ ) ⁢ h ⁢ ( y , τ ) ⁢ 𝑑 y ⁢ 𝑑 τ = ∫ 0 t [ G ⁢ ( ⋅ , t - τ ) ∗ h ⁢ ( ⋅ , τ ) ] ⁢ ( x ) ⁢ 𝑑 τ , x ∈ ℝ N ,

where G is the Gaussian kernel

G ⁢ ( x , t ) := ( 4 ⁢ π ⁢ t ) - N 2 ⁢ exp ⁡ ( - | x | 2 4 ⁢ t ) , x ∈ ℝ N , t > 0 ,

and h is given by h:=2⁢div⁡(ϕ⁢∇⁡η)-ϕ⁢Δ⁢η. Hence, in particular, we have

(4.7) z ⁢ ( t ) = 2 ⁢ ∫ 0 t G ⁢ ( t - τ ) ∗ div ⁡ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ⁡ d ⁢ τ - ∫ 0 t G ⁢ ( t - τ ) ∗ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ⁢ 𝑑 τ := z 1 ⁢ ( t ) - z 2 ⁢ ( t ) .

In (4.7), since we are only interested in the behavior of z with respect to the variable t, and for keeping the notations lighter, we have omitted the dependence of z on the variable x. We will maintain this convention until the end of the proof. Finally, we have (recall that Γ⁢(1-s)=-s⁢Γ⁢(-s))

∥ g ∥ L 2 ⁢ ( ℝ N ) ≤ s Γ ⁢ ( 1 - s ) ⁢ ∫ 0 + ∞ ∥ z ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) t 1 + s ⁢ 𝑑 t

= s Γ ⁢ ( 1 - s ) ⁢ ∫ 0 1 ∥ z ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) t 1 + s ⁢ 𝑑 t + s Γ ⁢ ( 1 - s ) ⁢ ∫ 1 + ∞ ∥ z ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) t 1 + s ⁢ 𝑑 t

≤ s Γ ⁢ ( 1 - s ) ⁢ ∫ 0 1 ∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) t 1 + s ⁢ 𝑑 t ⏟ = ⁣ : A 1 1 + s Γ ⁢ ( 1 - s ) ⁢ ∫ 0 1 ∥ z 2 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) t 1 + s ⁢ 𝑑 t ⏟ = ⁣ : A 1 2

(4.8) + s Γ ⁢ ( 1 - s ) ⁢ ∫ 1 + ∞ ∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) t 1 + s ⁢ 𝑑 t ⏟ = ⁣ : A 2 1 + s Γ ⁢ ( 1 - s ) ⁢ ∫ 1 + ∞ ∥ z 2 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) t 1 + s ⁢ 𝑑 t ⏟ = ⁣ : A 2 2 .

We proceed now estimating the terms A11, A12, A21 and A22 separately.

Step 1: Preliminary Estimates.First of all, throughout the remainder of the proof, C will denote a generic positive constant depending only on Ω, η, s and N. This constant may change even from line to line.

Now, we observe that by using some classical energy estimates for solutions to the heat equation, we obtain that

(4.9) ∥ ϕ ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) ≤ ∥ u ∥ L 2 ⁢ ( Ω ) ,

(4.10) ∥ ϕ ⁢ ( t ) ∥ W s , 2 ⁢ ( ℝ N ) ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) for all ⁢ s ∈ ( 0 , 1 ) .

These inequalities can be easily proved by multiplying (4.2) by ϕ and (-Δ)s⁢ϕ, respectively, and integrating by parts. Moreover, to obtain (4.10) we also took into account that, according to [31, Lemma 16.3], we have

∥ ( - Δ ) s 2 ⁢ ϕ ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) = C ⁢ ∫ ℝ N ∫ ℝ N | ϕ ⁢ ( x , t ) - ϕ ⁢ ( y , t ) | 2 | x - y | N + 2 ⁢ s ⁢ 𝑑 x ⁢ 𝑑 y .

In our proof, we will also need the following classical property of convolution (see, e.g., [12, Proposition 8.9]). For all w1∈Lq1⁢(ℝN), w2∈Lq2⁢(ℝN) and for all q1,q2 and q3 satisfying

(4.11) 1 ≤ q 1 , q 2 , q 3 < + ∞ , 1 q 1 + 1 q 2 = 1 q 3 + 1 ,

we have that

(4.12) ∥ w 1 ∗ w 2 ∥ L q 3 ⁢ ( ℝ N ) ≤ ∥ w 1 ∥ L q 1 ⁢ ( ℝ N ) ⁢ ∥ w 2 ∥ L q 2 ⁢ ( ℝ N ) .

This is a straightforward consequence of the Young inequality. Finally, we recall that for all 1≤p<∞ and k≥0, the function G satisfies the following decay properties (see, e.g., [18]): there exists a constant C>0 such that

∥ D k ⁢ G ⁢ ( t ) ∥ L p ⁢ ( ℝ N ) ≤ C ⁢ t - N 2 ⁢ ( 1 - 1 p ) - k 2 .

Here, k=(k1,k2,…,kN) is a multi-index with modulus |k|=k1+k2+⋯+kNand we used the classical Schwartz notation

D k ⁢ ϕ ⁢ ( x ) = ∂ | k | ⁡ ϕ ⁢ ( x ) ∂ ⁡ x 1 k 1 ⁢ ∂ ⁡ x 2 k 2 ⁢ ⋯ ⁢ ∂ ⁡ x N k N .

In particular, we have that

∥ G ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) ≤ C ⁢ t - N 4 , ⁢ ∥ ∇ x ⁡ G ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) ≤ C ⁢ t - N 4 - 1 2 ,

∥ ( G ∗ h ) ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) ≤ C ⁢ ∥ h ∥ L 2 ⁢ ( Ω ) , ⁢ ∥ ( ∇ x ⁡ G ∗ h ) ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) ≤ C ⁢ t - 1 2 ⁢ ∥ h ∥ L 2 ⁢ ( Ω ) .

Step 2: Upper Bound of A1:=A11+A12.We start by estimating the contribution of z1. Using (4.12) with q1=1, q2=q3=2, and (4.10), we get that

∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) ≤ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ div ⁡ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∫ 0 t ∥ D 1 - s ⁢ G ⁢ ( t - τ ) ∗ D s ⁢ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∫ 0 t ( t - τ ) - 1 - s 2 ⁢ ∥ ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ∥ W s , 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ⁢ ∫ 0 t ( t - τ ) - 1 - s 2 ⁢ 𝑑 τ = C ⁢ t 1 + s 2 ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) .

In the previous computations, Ds denotes the differential operator with Fourier symbol |⋅|s, that is, Ds⁢ζ⁢(⋅)=ℱ-1⁢{|⋅|s⁢ℱ⁢ζ⁢(⋅)}for all functions ζ sufficiently smooth. Concerning the contribution of z2, instead, we have

∥ z 2 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) ≤ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ ≤ C ⁢ ∫ 0 t ∥ ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ ≤ C ⁢ t ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

Since 0<s<1, we have that

A 1 ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ⁢ ∫ 0 1 d ⁢ t t 1 + s 2 + C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 1 d ⁢ t t s ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) + C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) .

Step 3: Upper Bound of A21.We have to distinguish three cases: N=1, N=2 and N≥3.

Case 1: N=1.Since u∈L2⁢(Ω) and Ω is bounded, we also have u∈L1⁢(Ω). Hence, the quantity

m := ∫ ℝ u ⁢ 𝑑 x = ∫ Ω u ⁢ 𝑑 x

is well defined.

Let us now rewrite u=(u-m⁢δ0)+m⁢δ0, where δ0 is the Dirac delta at x=0. With this splitting in mind, we have that ϕ can be seen as the sum ϕ=ψ+m⁢G, with ψ solving

ψ t - ψ x ⁢ x = 0 , t > 0 , ψ ⁢ ( 0 ) = u - m ⁢ δ 0 .

Therefore, we obtain

z 1 ⁢ ( t ) = ∫ 0 t G ⁢ ( t - τ ) ∗ ( ψ ⁢ ( τ ) ⁢ η x ) x ⁢ 𝑑 τ + ∫ 0 t G ⁢ ( t - τ ) ∗ ( m ⁢ G ⁢ ( τ ) ⁢ η x ) x ⁢ 𝑑 τ := z 1 , ψ ⁢ ( t ) + z 1 , G ⁢ ( t ) .

Let us consider firstly the term z1,ψ. First of all, we notice that ψ=θx with θ solving

θ t - θ x ⁢ x = 0 , t > 0 , θ ⁢ ( 0 ) = ∫ - ∞ x ( u - m ⁢ δ 0 ) ⁢ 𝑑 ξ ,

and therefore,

z 1 , ψ ⁢ ( t ) = ∫ 0 t G ⁢ ( t - τ ) ∗ ( θ x ⁢ ( τ ) ⁢ η x ) x ⁢ 𝑑 τ .

Now

∥ z 1 , ψ ⁢ ( t ) ∥ L 2 ⁢ ( ℝ ) ≤ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ ( θ x ⁢ ( τ ) ⁢ η x ) x ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ

= ∫ 0 t ∥ G x ⁢ ( t - τ ) ∗ ( θ x ⁢ ( τ ) ⁢ η x ) ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ

≤ ∫ 0 t ( t - τ ) - 3 4 ⁢ ∥ θ x ⁢ ( τ ) ⁢ η x ∥ L 1 ⁢ ( ℝ ) ⁢ 𝑑 τ .

Moreover, we have

∥ θ x ⁢ ( τ ) ⁢ η x ∥ L 1 ⁢ ( ℝ ) ≤ C ⁢ ∥ θ x ⁢ ( τ ) ∥ L 1 ⁢ ( Ω ) ≤ C ⁢ τ - 1 2 ⁢ ∥ θ ⁢ ( 0 ) ∥ L 1 ⁢ ( Ω ) ≤ C ⁢ τ - 1 2 ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ,

where the last inequality is justified by the fact that the initial datum θ⁢(0) is well defined as an L1-function compactly supported in Ω, and there exists a constant C>0 such that

∥ θ ⁢ ( 0 ) ∥ L 1 ⁢ ( Ω ) ≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

See [9, Theorem 1] for more details. Hence,

∥ z 1 , ψ ⁢ ( t ) ∥ L 2 ⁢ ( ℝ ) ≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ( t - τ ) - 3 4 ⁢ τ - 1 2 ⁢ 𝑑 τ = C ⁢ t - 1 4 ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

Let us now analyze the term z1,G which, we remind, is defined as

z 1 , G ⁢ ( t ) = m ⁢ ∫ 0 t G ⁢ ( t - τ ) ∗ ( G ⁢ ( τ ) ⁢ η x ) x ⁢ 𝑑 τ .

We have

∥ z 1 , G ⁢ ( t ) ∥ L 2 ⁢ ( ℝ ) ≤ m ⁢ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ ( G ⁢ ( τ ) ⁢ η x ) x ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ = m ⁢ ∫ 0 t ∥ G x ⁢ ( t - τ ) ∗ ( G ⁢ ( τ ) ⁢ η x ) ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ .

Now, since u is compactly supported in Ω, the Cauchy–Schwarz inequality yields

m ≤ ∥ u ∥ L 1 ⁢ ( Ω ) ≤ | Ω | ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ,

where |Ω| is the Lebesgue measure of Ω; hence

∥ z 1 , G ⁢ ( t ) ∥ L 2 ⁢ ( ℝ ) ≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ∥ G x ⁢ ( t - τ ) ∗ ( G ⁢ ( τ ) ⁢ η x ) ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ .

Rewrite G⁢(τ)⁢ηx=(G⁢(τ)⁢η)x-Gx⁢(τ)⁢η. Then

∥ z 1 , G ⁢ ( t ) ∥ L 2 ⁢ ( ℝ ) ≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ∥ G x ⁢ ( t - τ ) ∗ ( G ⁢ ( τ ) ⁢ η ) x ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ ⏟ = ⁣ : J 1 + C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ∥ G x ⁢ ( t - τ ) ∗ ( G x ⁢ ( τ ) ⁢ η ) ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ ⏟ = ⁣ : J 2 .

Concerning J1, we have

J 1 ≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ∥ D 1 - s ⁢ G x ⁢ ( t - τ ) ∗ D s ⁢ ( G ⁢ ( τ ) ⁢ η ) ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ

≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ∥ D 1 - s ⁢ G x ⁢ ( t - τ ) ∥ L 1 ⁢ ( ℝ ) ⁢ ∥ D s ⁢ ( G ⁢ ( τ ) ⁢ η ) ∥ L 2 ⁢ ( ℝ ) ⁢ 𝑑 τ

≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ( t - τ ) - 2 - s 2 ⁢ τ - 1 4 - s 2 ⁢ 𝑑 τ = C ⁢ t - 1 4 ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

Finally, for J2 we have

J 2 ≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ∥ G x ⁢ ( t - τ ) ∥ L 2 ⁢ ( ℝ ) ⁢ ∥ G x ⁢ ( τ ) ⁢ η x ∥ L 1 ⁢ ( ℝ ) ⁢ 𝑑 τ

≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 0 t ( t - τ ) - 3 4 ⁢ τ - 1 2 ⁢ 𝑑 τ = C ⁢ t - 1 4 ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

Summarizing, we get that

∥ z 1 , G ⁢ ( t ) ∥ L 2 ⁢ ( ℝ ) ≤ C ⁢ t - 1 4 ⁢ ∥ u ∥ L 2 ⁢ ( Ω )

which, combined with the estimate that we have obtained before for z1,ψ, gives

∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ ) ≤ C ⁢ t - 1 4 ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

Therefore, since s>0, we finally get that

A 2 1 = s Γ ⁢ ( 1 - s ) ⁢ ∫ 1 + ∞ ∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ ) t 1 + s ⁢ 𝑑 t ≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 1 + ∞ d ⁢ t t s + 5 4 = C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

Case 2: N=2.Using again (4.10), (4.12) with q1=q3=2 and q2=1 and the fact that η has compact support, we get that

∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ 2 ) ≤ 2 ⁢ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ div ⁡ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L 2 ⁢ ( ℝ 2 ) ⁢ 𝑑 τ

≤ 2 ⁢ ∫ 0 t ∥ D 1 - s ⁢ G ⁢ ( t - τ ) ∗ D s ⁢ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L 2 ⁢ ( ℝ 2 ) ⁢ 𝑑 τ

≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ⁢ ∫ 0 t ( t - τ ) - 1 + s 2 ⁢ 𝑑 τ ≤ C ⁢ t s 2 ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) .

Since s>0, it follows that

A 2 1 = s Γ ⁢ ( 1 - s ) ⁢ ∫ 1 + ∞ ∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ 2 ) t 1 + s ⁢ 𝑑 t ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ⁢ ∫ 1 + ∞ d ⁢ t t 1 + s 2 = C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) .

Case 3: N≥3.This case is more delicate and we need to proceed in a slightly different way. For a given ε∈[0,1], we will apply again (4.12) but this time by choosing

(4.13) q 1 = 4 - 2 ⁢ ε 4 - 3 ⁢ ε , q 2 = 2 - ε , q 3 = 2 .

It is straightforward to check that q1, q2 and q3 given in (4.13) satisfycondition (4.11). In particular, we notice that q2∈[1,2]. With this particular choice of the parameters we have

∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ 2 ) ≤ 2 ⁢ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ div ⁡ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

= 2 ⁢ ∫ 0 t ∥ D 1 - s ⁢ G ⁢ ( t - τ ) ∗ D s ⁢ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∫ 0 t ( t - τ ) - N 2 ⁢ ε 4 - 2 ⁢ ε - 1 - s 2 ⁢ ∥ ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ∥ L 2 - ε ⁢ ( ℝ N ) ⁢ 𝑑 τ ≤ C ⁢ t 1 + s 2 - N 2 ⁢ ε 4 - 2 ⁢ ε ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ,

provided that

1 + s 2 - N 2 ⁢ ε 4 - 2 ⁢ ε > 0 ⟹ ε < 4 + 4 ⁢ s N + 2 + 2 ⁢ s .

Therefore,

A 2 1 = s Γ ⁢ ( 1 - s ) ⁢ ∫ 1 + ∞ ∥ z 1 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) t 1 + s ⁢ 𝑑 t ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ⁢ ∫ 1 + ∞ d ⁢ t t 1 + s 2 + N 2 ⁢ ε 4 - 2 ⁢ ε = C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ,

if we impose that

1 + s 2 + N 2 ⁢ ε 4 - 2 ⁢ ε > 1 ⟹ ε > 4 - 4 ⁢ s N + 2 - 2 ⁢ s .

Thus, we obtain a further condition on ε, namely

ε ∈ ( 4 - 4 ⁢ s N + 2 - 2 ⁢ s , 4 + 4 ⁢ s N + 2 + 2 ⁢ s ) .

Furthermore, we can easily check that, for all s∈(0,1) and N≥3 the set

[ 0 , 1 ] ∩ ( 4 - 4 ⁢ s N + 2 - 2 ⁢ s , 4 + 4 ⁢ s N + 2 + 2 ⁢ s ) ≠ ∅ .

Therefore, for any given s∈(0,1) and N≥3, we can always choose q1, q2 and q3 as in (4.13) such that

A 2 1 ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) .

Step 4: Upper Bound of A22.Using again (4.12), this time with q1=1, q2=q3=2 and the fact that η has compact support, for a given α∈(2-2⁢s,2) we can estimate

∥ z 2 ⁢ ( t ) ∥ L 2 ⁢ ( ℝ N ) ≤ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ ∫ 0 t ∥ D α ⁢ G ⁢ ( t - τ ) ∗ D - α ⁢ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ ∫ 0 t ∥ D α ⁢ G ⁢ ( t - τ ) ∥ L 1 ⁢ ( ℝ N ) ⁢ ∥ D - α ⁢ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∫ 0 t ( t - τ ) - α 2 ⁢ ∥ ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ∥ W - α , 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ ≤ C ⁢ t 1 - α 2 ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

Hence

A 2 2 ≤ C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) ⁢ ∫ 1 + ∞ d ⁢ t t s + α 2 = C ⁢ ∥ u ∥ L 2 ⁢ ( Ω ) .

Step 5: Conclusion.Collecting all the above estimates, we can finally conclude that there exists a constant C>0 such that (4.5) holds, and the proof of Lemma 4.1 is finished.∎

4.2 Proof of the Lp-Regularity of g

Lemma 4.1 provides an alternative proof of the L2⁢(ℝN)-regularity of the remainder term g which appears in the formula for the fractional Laplacian of the product η⁢u. Moreover, as we did before in Section 3, also this result can be generalized to the Lp setting. In particular, we can prove the following.

Lemma 4.2

Let u∈W0s,2⁢(Ω¯), p≥2, N≥2 and let η be the cut-off function introduced in (3.1). Moreover, let g be the remainder term in the expression

( - Δ ) s ⁢ ( η ⁢ u ) = η ⁢ ( - Δ ) s ⁢ u + g .

Then, there exists a constant C>0 (independent of u) such that

(4.14) ∥ g ∥ L p ⁢ ( ℝ N ) ≤ C ⁢ ( ∥ u ∥ L p ⁢ ( Ω ) + ∥ u ∥ W s , 2 ⁢ ( Ω ) ) .

Proof.

We recall that, according to (4.4), to estimate the Lp-norm of g we only need an appropriate bound for the Lp-norm of the function z introduced in (4.6). Moreover, also in this case we have

∥ g ∥ L p ⁢ ( ℝ N ) ≤ A 1 1 + A 1 2 + A 2 1 + A 2 2 ,

where, with some abuse of notations, the terms A11, A12, A21 and A22 are the same ones as in (4.8), after having replaced ∥z⁢(t)∥L2⁢(ℝN) with ∥z⁢(t)∥Lp⁢(ℝN).

Step 1: Preliminary Estimates.We recall that u∈W0s,2⁢(Ω¯) follows from Proposition 2.1,and u∈Lp⁢(Ω) follows from Lemma 2.5. Moreover, we observe that the classical energy decay estimates presented in (4.9) can be generalized to the Lp setting. In particular, we have

(4.15) ∥ ϕ ⁢ ( t ) ∥ L p ⁢ ( ℝ N ) ≤ ∥ u ∥ L p ⁢ ( Ω ) .

The proof of (4.15) is a straightforward application of (4.12), taking into account the fact that the solution of the heat equation (4.2) is given by the convolution ϕ⁢(t)=G⁢(t)∗u.

Step 2: Upper Bound of A11.First of all, throughout the remainder of the proof, C will denote a generic positive constant depending only on Ω, η, s, p and N. This constant may change even from line to line.

Now, using (4.12) with q1=2⁢p2+p, q2=2 and q3=p, we get that

∥ z 1 ⁢ ( t ) ∥ L p ⁢ ( ℝ N ) ≤ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ div ⁡ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L p ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∫ 0 t ∥ D 1 - s ⁢ G ⁢ ( t - τ ) ∗ D s ⁢ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L p ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∫ 0 t ∥ D 1 - s ⁢ G ⁢ ( t - τ ) ∥ L q 1 ⁢ ( ℝ N ) ⁢ ∥ D s ⁢ ( ϕ ⁢ ( τ ) ⁢ ∇ ⁡ η ) ∥ L 2 ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ⁢ ∫ 0 t ( t - τ ) - N 2 ⁢ ( 1 - 1 q 1 ) - 1 - s 2 ⁢ 𝑑 τ = C ⁢ t 1 + s 2 - N 2 ⁢ ( 1 - 1 q 1 ) ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ,

provided that

1 + s 2 - N 2 ⁢ ( 1 - 1 q 1 ) > 0 ⟹ q 1 < N N - 1 - s .

In view of the previous estimate, we have

A 1 1 ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ⁢ ∫ 0 1 d ⁢ t t 1 + s 2 + N 2 ⁢ ( 1 - 1 q 1 ) = C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ,

provided that

1 + s 2 + N 2 ⁢ ( 1 - 1 q 1 ) < 1 ⇒ q 1 < N N - 1 + s .

Finally, we notice that by hypothesis we have p≥2; this, according to the definition of q1 that we are considering, corresponds to the further condition 1≤q1<2. Hence, recollecting the conditions on q1 that we have encountered, we conclude that we have to impose

1 ≤ q 1 < min ⁡ { 2 , N N - 1 + s , N N - 1 - s } = N N - 1 + s = 1 + 1 - s N - 1 + s .

Summarizing, we have

A 1 1 ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ,

if in our computations we assume

1 ≤ q 1 < 1 + 1 - s N - 1 + s .

Step 3: Upper Bound of A12.We have

∥ z 2 ⁢ ( t ) ∥ L p ⁢ ( ℝ N ) ≤ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ∥ L p ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∫ 0 t ∥ ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ∥ L p ⁢ ( ℝ N ) ⁢ 𝑑 τ ≤ C ⁢ t ⁢ ∥ u ∥ L p ⁢ ( Ω ) .

Since 0<s<1, we have that

A 1 2 ≤ C ⁢ ∥ u ∥ L p ⁢ ( Ω ) ⁢ ∫ 0 1 d ⁢ t t s ≤ C ⁢ ∥ u ∥ L p ⁢ ( Ω ) .

Step 4: Upper Bound of A21.

Repeating the same computations that we did in Step 2, we get that

∥ z 1 ⁢ ( t ) ∥ L p ⁢ ( ℝ N ) ⁢ C ⁢ t 1 + s 2 - N 2 ⁢ ( 1 - 1 q 1 ) ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ,

provided that

1 + s 2 - N 2 ⁢ ( 1 - 1 q 1 ) > 0 ⟹ q 1 < N N - 1 - s .

Therefore

A 2 1 ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ⁢ ∫ 1 + ∞ d ⁢ t t 1 + s 2 + N 2 ⁢ ( 1 - 1 q 1 ) = C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ,

provided that

1 + s 2 + N 2 ⁢ ( 1 - 1 q 1 ) > 1 ⟹ q 1 > N N - 1 + s .

q 1 ∈ [ 1 , min ⁡ { 2 , N N - 1 - s } ) ∩ ( N N - 1 + s , + ∞ ) = ( N N - 1 + s , min ⁡ { 2 , N N - 1 - s } ) .

Summarizing, we have

A 1 2 ≤ C ⁢ ∥ u ∥ W s , 2 ⁢ ( Ω ) ,

if in our computations we assume

q 1 ∈ ( N N - 1 + s , min ⁡ { 2 , N N - 1 - s } ) .

Step 5: Upper Bound of A22.Using again (4.12), this time with q1=1, q2=q3=p and the fact that η has compact support, for a given α∈(2-2⁢s,2) we can estimate

∥ z 2 ⁢ ( t ) ∥ L p ⁢ ( ℝ N ) ≤ ∫ 0 t ∥ G ⁢ ( t - τ ) ∗ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ∥ L p ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ ∫ 0 t ∥ D α ⁢ G ⁢ ( t - τ ) ∗ D - α ⁢ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ∥ L p ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ ∫ 0 t ∥ D α ⁢ G ⁢ ( t - τ ) ∥ L 1 ⁢ ( ℝ N ) ⁢ ∥ D - α ⁢ ( ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ) ∥ L p ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ ∫ 0 t ( t - τ ) - α 2 ⁢ ∥ ϕ ⁢ ( τ ) ⁢ Δ ⁢ η ∥ W - α , p ⁢ ( ℝ N ) ⁢ 𝑑 τ

≤ C ⁢ t 1 - α 2 ⁢ ∥ u ∥ L p ⁢ ( Ω ) .

Hence

A 2 2 ≤ C ⁢ ∥ u ∥ L p ⁢ ( Ω ) ⁢ ∫ 1 + ∞ d ⁢ t t s + α 2 = C ⁢ ∥ u ∥ L p ⁢ ( Ω ) .

Step 6: Conclusion.Recollecting all the above estimates, we can finally conclude that there exists a constant C>0 such that (4.14) holds. The proof of Lemma 4.2 is finished.∎

Remark 4.3

Lemma 4.2 provides an alternative proof of the Lp⁢(ℝN)-regularity of the remainder term g which appears in the formula for the fractional Laplacian of the product η⁢u. However, in its proof, we are able to deal only with the case N≥2. When N=1, instead, we encounter some difficulties that, at the present stage, we are not able to overcome. We will present these difficulties with more details in Section 5, dedicated to open problems and perspectives. Nevertheless, we do not exclude that this regularity lemma could be extended also to the case of one-space dimension.

5 Open Problems and Perspectives

In the present paper we proved that weak solutions to the Dirichlet problem for the fractional Laplacian with a non-homogeneous right-hand side f∈Lp⁢(Ω) (1<p<∞)belong to Wloc2⁢s,p⁢(Ω).

The following comments are worth considering.

(a) In the proof of Lemma 4.2, which provides the Lp⁢(ℝN)-regularity of the remainder term g following the approach that employs the heat kernel characterization of the fractional Laplacian, we were not able to treat the cases 1<p<2 and N=1. In more detail, we cannot encounter appropriate bounds for the terms A11 and A12 (see (4.7) for more details on the notation). These difficulties are most likely related to the fact that, in this lower dimension case or for lower values of p, there is less diffusion and the decay rates that we shall employ are slower. On the other hand, we believe that there has to be a way to solve this problem.

(b) A natural interesting extension would be the analysis of the global elliptic regularity for weak solutions to (1.1). The problem is delicate however.

For the classical Dirichlet problem associated with the Laplace operator (the case s=1), it is well known that if Ω is smooth, say of class C2, then weak solutions to the associated problem belong to W2,p⁢(Ω).

But, unfortunately, at least for large p, this maximal global elliptic regularity is not true for the fractional Laplacian. That is, for problem (1.1), weak solutions do not necessarily belong to W2⁢s,p⁢(Ω). In fact, if this were the case, then for large p and 12<s<1, weak solutions would be at least β-Hölder continuous up to the boundary of Ω of order β>s. One can see that the latter property is not true by applying the Pohozaev identity obtained in [26] to the eigenfunctions of the Dirichlet fractional Laplacian. Indeed, let λk>0 be an eigenvalue of AD and uk the associated eigenfunction. Then, rewriting the identity in[26, Proposition 1.6] with uk by using the fact that AD⁢uk=λk⁢uk, we get

λ k ⁢ ∫ Ω u n ⁢ ( x ⋅ ∇ ⁡ u k ) ⁢ 𝑑 x = 2 ⁢ s - N 2 ⁢ λ k ⁢ ∫ Ω u k 2 ⁢ 𝑑 x - Γ ⁢ ( s + 1 ) 2 2 ⁢ ∫ ∂ ⁡ Ω ( u k ρ s ) 2 ⁢ ( x ⋅ ν ) ⁢ 𝑑 σ .

Integrating the term on the left-hand side by parts and using that uk=0 on ∂⁡Ω, we get that

(5.1) s ⁢ λ k ⁢ ∫ Ω u k 2 ⁢ 𝑑 x = Γ ⁢ ( s + 1 ) 2 2 ⁢ ∫ ∂ ⁡ Ω ( u k ρ s ) 2 ⁢ ( x ⋅ ν ) ⁢ 𝑑 σ .

Now if uk were β-Hölder continuous up to the boundary ∂⁡Ω of order β>s, then since 0<s<1 and s⁢λk>0, it would follow from (5.1) that ∫Ωuk2⁢𝑑x=0. Thus uk=0 on Ω, which contradicts the fact that uk is an eigenfunction. We have shown that uk cannot be β-Hölder continuous up to the boundary ∂⁡Ω of order β>s.A direct proof that uk cannot be Lipschitz continuous up to the boundary is also contained in [28] and the references therein, where it has been shown that the eigenfunctions are s-Hölder continuous up to the boundary and this regularity is optimal. Finally, a concrete example has been given in [25, Section 7].

(c) It has been shown in [24] that if f∈L∞⁢(Ω) with Ω of class C2 and u is a weak solution of (1.1), then u∈C0,s⁢(ℝN) and the function ρ-s⁢u, where ρ=dist⁡(x,∂⁡Ω) is the distance of a point x to the boundary of the domain Ω, belongs to C0,α⁢(Ω¯) for some 0<α<min⁡{s,1-s}. In addition one has the following precise regularity.

If Ω is of class C∞ and f∈C∞⁢(Ω¯), then ρ-s⁢u∈C∞⁢(Ω¯) (see, e.g., [24]).
If Ω is of class C2,β and f∈Cβ⁢(Ω¯), then ρ-s⁢u∈Cs+β⁢(Ω¯) (see, e.g., [27]).

Roughly speaking, these results just mentioned tell us that, if the domain Ω is regular enough, the solution u to (1.1) can be seen as u=ρs⁢v, where v is a function regular up to the boundary. By part (b), weak solutions are in general not in W2⁢s,p⁢(Ω). Nevertheless, compared with the above mentioned results, one could expect both ρ-s⁢u and ρ1-s⁢u to be smooth in the Lp⁢(Ω) context, i.e. to belong to W2⁢s,p⁢(Ω).In view of this, it would be natural to analyze whether this regularity property, which is not available in the literature, is actually true.Finally, more generally, it is also interesting to investigate for which β>0 we have ρβ⁢u∈W2⁢s,p⁢(Ω). Following our approach, we think that it is possible to show that ρβ⁢u∈W2⁢s,p⁢(Ω) for every β>s. However, the most interesting case is 0<β≤s. We mention that in this situation we have that ρβ⁢u is also a solution of a certain Dirichlet problem.

Dedicated to Ireneo Peral on the occasion of his 70th birthday: Gracias Ireneo por tantos años de amistad y ejemplo

Communicated by Antonio Ambrosetti and David Arcoya

Funding source: U.S. Air Force

Award Identifier / Grant number: FA9550-15-1-0027

Award Identifier / Grant number: FA9550-14-1-0214

Funding source: Ministerio de Economía y Competitividad

Award Identifier / Grant number: MTM2014-52347

Funding source: Agence Nationale de la Recherche

Award Identifier / Grant number: ICON

Funding statement: All authors were supported by the Air Force Office of Scientific Research through awardno. FA9550-15-1-0027.Umberto Biccari and Enrique Zuazua were supported by Ministerio de Economía y Competitividad(Spain) through grant MTM2014-52347 and by the European Research Council Executive Agencythrough Advanced Grant DYCON (Dynamic Control).Enrique Zuazua was supported by EOARD-AFOSR through award no. FA9550-14-1-0214 and byAgence Nationale de la Recherche (France) through ICON.

A Appendix

For the sake of completeness, we introduce some well-known facts about the fractional order Sobolev spaces, which are not so familiar as the classical integral order Sobolev spaces.

Let Ω⊂ℝN be an arbitrary open set. For p∈[1,∞) and s∈(0,1), we denote by

W s , p ⁢ ( Ω ) := { u ∈ L p ⁢ ( Ω ) : ∫ Ω ∫ Ω | u ⁢ ( x ) - u ⁢ ( y ) | p | x - y | N + p ⁢ s ⁢ 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 + p ⁢ s ⁢ 𝑑 x ⁢ 𝑑 y ) 1 p .

We set

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

where 𝒟⁢(Ω) is the space of all continuously infinitely differentiable functions with compact support in Ω.

The following result is taken from [15, p. 25, Theorem 1.4.2.4].

Theorem A.1

Let Ω⊂RN be a bounded open set with Lipschitz continuous boundary and 1<p<∞. Then for every 0<s≤1p, we have that Ws,p⁢(Ω)=W0s,p⁢(Ω) with equivalent norm.

It is well known (see, e.g., [7, 15]) that if Ω⊂ℝN is a bounded open set with a Lipschitz continuous boundary then

(A.1) W s , p ⁢ ( Ω ) ↪ L q ⁢ ( Ω ) with ⁢ { 1 ≤ q ≤ N ⁢ p N - s ⁢ p if ⁢ N > s ⁢ p , 1 ≤ q < ∞ if ⁢ N = s ⁢ p .

If N<s⁢p, then

(A.2) W s , p ⁢ ( Ω ) ↪ C 0 , s - N p ⁢ ( Ω ¯ ) .

Next, for 1<p<∞ and 0<s<1 we define

W 0 s , p ⁢ ( Ω ¯ ) := { u ∈ W s , p ⁢ ( ℝ N ) : u = 0 ⁢ on ⁢ ℝ N ∖ Ω } .

It has been shown in [7, Lemma 6.1] that for an arbitrary bounded open set Ω⊂ℝN, there exists a constant C>0 such that

(A.3) ∫ ℝ N ∖ Ω d ⁢ y | x - y | N + s ⁢ p ≥ C ⁢ | Ω | - s ⁢ p N .

Using (A.3), we get that there exists a constant C>0 such that for every u∈W0s,p⁢(Ω¯),

(A.4) ∫ ℝ N | u | p ⁢ 𝑑 x = ∫ Ω | u | p ⁢ 𝑑 x ≤ C ⁢ ∫ ℝ N | u ⁢ ( x ) | p ⁢ ∫ ℝ N ∖ Ω d ⁢ y | x - y | N + s ⁢ p ≤ C ⁢ ∫ ℝ N ∫ ℝ N | u ⁢ ( x ) - u ⁢ ( y ) | p | x - y | N + p ⁢ s ⁢ 𝑑 x ⁢ 𝑑 y .

It follows from (A.4) that for every 1<p<∞ and 0<s<1,

(A.5) ∥ u ∥ W 0 s , p ⁢ ( Ω ¯ ) = ( ∫ ℝ N ∫ ℝ N | u ⁢ ( x ) - u ⁢ ( y ) | p | x - y | N + p ⁢ s ⁢ 𝑑 x ⁢ 𝑑 y ) 1 p

defines an equivalent norm on W0s,p⁢(Ω¯). We shall denote by W-s,p′⁢(Ω¯) the dual of the reflexive Banach space W0s,p⁢(Ω¯), that is,

W - s , p ′ ⁢ ( Ω ¯ ) := ( W 0 s , p ⁢ ( Ω ¯ ) ) ⋆ where ⁢ p ′ := p p - 1 .

We remark that there is no obvious inclusion between W0s,p⁢(Ω) and W0s,p⁢(Ω¯). In fact, for an arbitrary bounded open set Ω⊂ℝN, the two spaces are different, since 𝒟⁢(Ω) is not always dense in W0s,p⁢(Ω¯) (see, e.g., [11]). But if Ω has a continuous boundary, then by [11, Theorem 6], 𝒟⁢(Ω) is dense in W0s,p⁢(Ω¯) and in addition we have that

(A.6) W 0 s , p ⁢ ( Ω ¯ ) = W 0 s , p ⁢ ( Ω ) for every ⁢ 1 p < s < 1 .

In fact, (A.6) follows by using the Hardy inequality for fractional order Sobolev spaces and the following estimate (see, e.g., [15, formula (1.3.2.12)]): there exist two constants 0<C1≤C2 such that

C 1 ( ρ ⁢ ( x ) ) p ⁢ s ≤ ∫ ℝ N ∖ Ω d ⁢ y | x - y | N + s ⁢ p ≤ C 2 ( ρ ⁢ ( x ) ) p ⁢ s , x ∈ Ω ,

where ρ⁢(x):=dist⁡(x,∂⁡Ω), x∈Ω.

We also notice that the continuous embeddings (A.1) and (A.2) hold with Ws,p⁢(Ω) replaced with W0s,p⁢(Ω) or W0s,p⁢(Ω¯) and this case without any regularity assumption on the open set Ω.

Next, if s>1 and is not an integer, then we write s=m+σ where m is an integer and 0<σ<1. In this case

W s , p ⁢ ( Ω ) := { u ∈ W m , p ⁢ ( Ω ) : D α ⁢ u ∈ W σ , p ⁢ ( Ω ) ⁢ for any ⁢ α ⁢ such that ⁢ | α | = m } .

Then Ws,p⁢(Ω) is a Banach space with respect to the norm

∥ u ∥ W s , p ⁢ ( Ω ) := ( ∥ u ∥ W m , p ⁢ ( Ω ) p + ∑ | α | = m ∥ D α ⁢ u ∥ W σ , p ⁢ ( Ω ) p ) 1 p .

If s=m is an integer, then Ws,p⁢(Ω) coincides with the Sobolev space Wm,p⁢(Ω). Comparing with (A.1), we have the following general embedding.

Theorem A.2

Let Ω⊂RN be a bounded open set with Lipschitz continuous boundary. Then the following assertions hold:

If 0 < s ≤ r and 1 < p ≤ q < ∞ are real numbers such that r - N p = s - N q ,then W r , p ⁢ ( ℝ N ) ↪ W s , q ⁢ ( ℝ N ) .
If 0 < s ≤ r and 1 < p ≤ q < ∞ are real numbers such that r - N p ≥ s - N q ,then W r , p ⁢ ( Ω ) ↪ W s , q ⁢ ( Ω ) .

For more information on fractional order Sobolev spaces, we refer to [1, 7, 15, 17]and references therein.

Finally, we mention the following estimate that has been used previously. Let A⊂ℝN be a bounded set and B⊂ℝN an arbitrary set. Then there exists a constant C>0 (depending on A and B) such that

(A.7) | x - y | ≥ C ⁢ ( 1 + | y | ) for all ⁢ x ∈ A , y ∈ ℝ N ∖ B , dist ⁡ ( A , ℝ N ∖ B ) = δ > 0 .

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] Bakunin O. G.,Turbulence and Diffusion: Scaling Versus Equations,Springer, Berlin, 2008.Search in Google Scholar

[3] Bernard C.,Regularity of solutions to the fractional Laplace equation,preprint 2014, http://math.uchicago.edu/~may/REU2014/REUPapers/Bernard.pdf.Search in Google Scholar

[4] Biccari U.,Internal control for non-local Schrödinger and wave equations involving the fractional Laplace operator,preprint 2017, https://arxiv.org/abs/1411.7800v2.Search in Google Scholar

[5] Bologna M., Tsallis C. and Grigolini P.,Anomalous diffusion associated with non-linear fractional derivative Fokker–Planck-like equation: Exact time-dependent solutions,Phys. Rev. E 62 (2000), 2213–2218.10.1103/PhysRevE.62.2213Search in Google Scholar

[6] Cozzi M.,Interior regularity of solutions of non-local equations in Sobolev and Nikol’skii spaces,Ann. Mat. Pura Appl. (2) 196 (2017), 555–578.10.1007/s10231-016-0586-3Search 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., Palatucci G. and Valdinoci E.,Dislocation dynamics in crystals: A macroscopic theory in a fractional Laplace setting,Comm. Math. Phys. 333 (2015), no. 2, 1061–1105.10.1007/s00220-014-2118-6Search in Google Scholar

[9] Duoandikoetxea J. and Zuazua E.,Moments, masses de Dirac et décomposition de fonctions,C. R. Acad. Sci. Paris Sér. 1 315 (1992), no. 6, 693–698.Search in Google Scholar

[10] Engel K.-J. and Nagel R.,One-Parameter Semigroups for Linear Evolution Equations,Grad. Texts in Math. 194,Springer, New York, 2000.Search in Google Scholar

[11] Fiscella A., Servadei R. and Valdinoci E.,Density properties for fractional Sobolev spaces,Ann. Acad. Sci. Fenn. Math. 40 (2015), 235–253.10.5186/aasfm.2015.4009Search in Google Scholar

[12] Folland G. B.,Real Analysis: Modern Techniques and Their Applications,John Wiley & Sons, New York, 2013.Search in Google Scholar

[13] Gal C. G. and Warma M.,Nonlocal transmission problems with fractional diffusion and boundary conditions on non-smooth interfaces,Comm. Partial Differential Equations (2017), 10.1080/03605302.2017.1295060.10.1080/03605302.2017.1295060Search in Google Scholar

[14] Gilboa G. and Osher S.,Nonlocal operators with applications to image processing,Multiscale Model. Simul. 7 (2008), no. 3, 1005–1028.10.1137/070698592Search in Google Scholar

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

[16] Grubb G.,Fractional Laplacians on domains, a development of Hörmander’s theory of μ-transmission pseudodifferential operators,Adv. Math. 268 (2015), 478–528.10.1016/j.aim.2014.09.018Search in Google Scholar

[17] Jonsson A. and Wallin H.,Function Spaces on Subsets of ℝN,Math. Rep. 2,Harwood Academic Publishers, Reading, 1984.Search in Google Scholar

[18] Kato T.,Strong Lp solutions of the Navier–Stokes equation in ℝm, with applications to weak solutions,Math. Z. 187 (1984), no. 4, 471–480.10.1007/BF01174182Search in Google Scholar

[19] Leonori T., Peral I., Primo A. and Soria F.,Basic estimates for solutions of a class of nonlocal elliptic and parabolic equations,Discrete Contin. Dyn. Syst. 35 (2015), 6031–6068.10.3934/dcds.2015.35.6031Search in Google Scholar

[20] Levendorski S.,Pricing of the American put under Lévy processes,Int. J. Theor. Appl. Finance 7 (2004), no. 3, 303–335.10.1142/S0219024904002463Search in Google Scholar

[21] Meerschaert M. M.,Fractional calculus, anomalous diffusion, and probability,Fractional Dynamics,World Scientific, Hackensack (2012), 265–284.10.1142/9789814340595_0011Search in Google Scholar

[22] Moser J.,A new proof of de Giorgi’s theorem concerning the regularity problem for elliptic differential equations,Comm. Pure Appl. Math. 13 (1960), 457–468.10.1002/cpa.3160130308Search in Google Scholar

[23] Pham H.,Optimal stopping, free boundary, and American option in a jump-diffusion model,Appl. Math. Optim. 35 (1997), no. 2, 145–164.10.1007/s002459900042Search in Google Scholar

[24] Ros-Oton X. and Serra J.,The Dirichlet problem for the fractional Laplacian: Regularity up to the boundary,J. Math. Pures Appl. 101 (2014), no. 9, 275–302.10.1016/j.matpur.2013.06.003Search in Google Scholar

[25] 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

[26] Ros-Oton X. and Serra J.,The Pohozaev identity for the fractional Laplacian,Arch. Ration. Mech. Anal. 213 (2014), 587–628.10.1007/s00205-014-0740-2Search in Google Scholar

[27] Ros-Oton X. and Serra J.,Boundary regularity for fully nonlinear integro-differential equations,Duke Math. J. 165 (2016), 2079–2154.10.1215/00127094-3476700Search in Google Scholar

[28] Servadei R. and Valdinoci E.,On the spectrum of two different fractional operators,Proc. Roy. Soc. Edinburgh Sect. A 144 (2014), 831–855.10.1017/S0308210512001783Search in Google Scholar

[29] Stein E.,Singular Integrals and Differentiability Properties of Functions,Princeton University Press, Princeton, 1970.10.1515/9781400883882Search in Google Scholar

[30] Stinga P. R.,Fractional powers of second order partial differential operators: Extension problem and regularity theory,Ph.D. Dissertation, Universidad Autónoma de Madrid, Spain, 2010.Search in Google Scholar

[31] Tartar L.,An Introduction to Sobolev Spaces and Interpolation Spaces,Springer, Berlin, 2007.Search in Google Scholar

[32] Vázquez J. L.,Nonlinear diffusion with fractional Laplacian operators,Nonlinear Partial Differential Equations (Oslo 2010),Abel Symp. 7,Springer, Heidelberg (2012), 271–298.10.1007/978-3-642-25361-4_15Search in Google Scholar

[33] Zhu T. and Harris J. M.,Modeling acoustic wave propagation in heterogeneous attenuating media using decoupled fractional Laplacians,Geophysics 79 (2014), no. 3, T105–T116.10.1190/geo2013-0245.1Search in Google Scholar

Received: 2017-02-06

Revised: 2017-04-13

Accepted: 2017-04-19

Published Online: 2017-04-21

Published in Print: 2017-05-01

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

Local Elliptic Regularity for the Dirichlet Fractional Laplacian

Abstract

1 Introduction

Theorem 1.3 (L2-Local Regularity)

Theorem 1.4 (Lp-Local Regularity)

2 Preliminaries

Proof.

Proof.

3 Proof of Theorems 1.3 and 1.4: First Approach

3.1 Proof of the L2-Local Regularity Theorem

Proof of Theorem 1.3.

3.2 Proof of the Lp-Local Regularity Theorem

Proof of Theorem 1.4.

4 The Approach Using the Heat Semigroup Representation

4.1 Proof of the L2-Regularity of g

Proof.

4.2 Proof of the Lp-Regularity of g

Proof.

5 Open Problems and Perspectives

A Appendix

References

Journal and Issue

Articles in the same Issue