Mediterr. J. Math.

(2023) 20:221
https://doi.org/10.1007/s00009-023-02419-w
1660-5446/23/040001-23
published online May 24, 2023
c The Author(s), under exclusive licence to Springer
Nature Switzerland AG 2023

On a Fractional p-Laplacian Equation

with Critical Fractional Sobolev Exponent
Ouarda Saifia and Jean Vélin

Abstract. In this article, we focus on the existence of solutions for a

nonlocal fractional p-Laplacian problem arising the critical fractional
Sobolev exponent. At least one solution is obtained as a local minimizer.

Mathematics Subject Classification. 35J50, 35J60, 35J61, 46E35, 47J05.

Keywords. Fractional p-Laplacian, critical fractional Sobolev exponent,
concentration-compactness, Dirichlet problem, Ekeland variational prin-
ciple.

1. Introduction
In this paper, we consider the following equation involving the fractional
p-Laplacian operator
 
(−Δ)sp u = |u|ps −2 u + f (x) in Ω
, (1)
u=0 on RN \Ω.

where (−Δ)sp is the fractional p-Laplacian operator s ∈ (0, 1), p > 1, N >
Np
sp > 1, ps = is the fractional critical Sobolev exponent and Ω is a
N − sp
bounded domain in R with a C 1,1 -boundary.
N

In the recent years, the study of a new nonlocal and nonlinear operator
namely the fractional p-Laplacian attracted a great attention (see for instance
[1–4,6,10,11,13,16,19,24]) and we refer [5,9,23] for more properties on the
fractional Sobolev space.
Among these studies, some of them deal with problems involving the
critical exponent Sobolev. Particularly, the problem (1) does not admit a
weak solution for any open set of RN . This fact has been discussed by Ros-
Oton and Serra in [25] for p = 2, f (x) ≡ 0 and a star-sharped domain Ω.
221 Page 2 of 23 O. Saifia and J. Vélin MJOM

Fiscella et al. (see [14]) overcame such difficulty by considering the non-
linear problem
 
(−Δ)s u = γ|u|2s −2 u + f (x, u) in Ω
u=0 on RN \Ω.

These authors deal with a problem driven by the fractional Laplace operator
(−Δ)s and involving a critical Sobolev exponent. They give some results on
the multiplicity of solutions under additional and suitable assumptions on
the function f.
However, in this study, the authors mention that contrary of the frac-
tional Laplacian operator, the extension to the nonlinear case p = 2 does not
remain valid because of the lack of knowledge about the spectral properties
for the fractional p-Laplacian operator. Having regard to this fact, unfortu-
nately the transfert of the technical decompositions performed for W0s,2 (Ω)
to the fractional p-Laplacian Sobolev space fails. Therefore, we mention a
result dealing with the decomposition due to Brasco et al. (see [7]).
Mosconi et al. [22] use the principle of concentration-compactness in
W0s,p (Ω) for proving that the following problem admits a weak solution

 
(−Δ)sp u = |u|ps −2−ε u in Ω
(2)
u=0 on RN \Ω.

Indeed, they consider an approximation nearly the critical Sobolev exponent.

Particularly, when Ω is an annual domain in RN , the authors prove that
there exists a continuum of positive solutions of (2). Therefore, they arrive
to conclude that the limit behavior of (2) admits only 0 as the unique weak
solution.
In the classical case (s = 1 and p = 2), the problem (1) has been treated
by Tarentello in [27] and it has been extended and simplified by Chabrowski
in [8].
Recently, Louis-Rose and Vélin [20] deal with the fractional p-Laplacian
equation

(−Δ)sp u = u|u|α−1 in Ω
u=0 on RN \Ω.

They established a nonexistence result for a star-shaped domain Ω, a critical

exponent (resp. supercritical) α+1 = ps (resp. > ps ) and a function f (x, u) =
f (x) ≡ 0 in Ω.
Motivated by the non-existence result obtained in [20], we place our-
selves in the critical situation α + 1 = ps and we propose to introduce a
suitable source term f depending only on the location x in Ω. We show that
the problem (1) admits at least one solution under some conditions on f.
MJOM On a Fractional p-Laplacian Equation with Page 3 of 23 221

2. Main Tools and Preliminaries

2.1. Definitions and Notations
For any q > 1, the classical Lebesgue’s space is
  
q q
L (Ω) = u : Ω −→ R measurable, |u| dx < ∞ .
Ω
This space is equipped with the norm
  q 1/q
 
uq = u(x) dx .
Ω

Given s ∈ (0, 1) and p ∈ [1, +∞[, we recall the usual definition of the frac-
tional Sobolev space
  
s,p
 N p
 N |u(x) − u(y)|p
W R = u∈L R ; N +ps
dxdy < +∞ .
RN ×RN |x − y|

Let Ω be a bounded domain in RN with a C 1,1 -boundary.


W0s,p (Ω) = {u ∈ W s,p RN ; u = 0 a.e. in RN \ Ω}
endowed with the norm
 1
|u(x) − u(y)|p p
uX0 = N +ps
dxdy .
RN ×RN |x − y|

The space W0s,p (Ω) is the closure of C0∞ (Ω) in W s,p RN .

2.2. Current Notations

p Np
For any p > 1, p = and ps = denote its associated conjugate
p−1 N − sp
exponent and the fractional Sobolev critical exponent, respectively.
The topological dual of W0s,p (Ω) is denoted by W −s,p (Ω). It is well


known that the immersion W0s,p (Ω) → Lq (Ω) is readability;

1. continuous for all q ∈ [1, ps ],
2. compact for all q ∈ [1, ps [.
Throughout the text and for an easy reading, we adopt the notation X0
(resp. X  ) instead of W0s,p (Ω) (resp. W −s,p (Ω)). The dual norm is denoted


by  −1 .
The notation , expresses the dual pairing between X0 and X  .
On X0 , we define:
• the functional

1 1 
I(u) = upX0 −  |u|ps dx − f, u (3)
p ps Ω

is currently named the energy functional for the problem (1). It is well
known (see [18]) that I is C 1 (X0 ; R). Its derivative denoted by I  is
defined as below

|u(x) − u(y)|p−2 (u(x) − u(y)
I  (u), ϕ = (ϕ(x) − ϕ(y))dxdy, ∀ϕ ∈ X0 ,
R2N |x − y|N +sp
221 Page 4 of 23 O. Saifia and J. Vélin MJOM

• the set
Λ = {u ∈ X0 \ {0}, I  (u), u = 0}.
• the functional
p
φ(u) = I  (u), u = upX0 − upss − f, u . (4)

Calculating the derivative of φ, we obtain

p
φ (u), u = (p − 1) upX0 − (ps − 1) upss .
Also, for the convenience, we split Λ into three parts
p
Λ>0 = u ∈ Λ, (p − 1) upX0 − (ps − 1) upss > 0 ,

p
Λ<0 = u ∈ Λ, (p − 1) upX0 − (ps − 1) upss < 0 ,

p
Λ0 = u ∈ Λ, (p − 1) upX0 − (ps − 1) upss = 0 .

2.3. Hypotheses
Along the text, we maintain the following hypotheses
1.
sp > 1. (5)
2. f satisfies the following condition
⎧
⎨ p − 1  S ps (p − 1)  pp−1
 −p
s
(H), 0 < f −1 < min ,
⎩ ps − p
 
ps − 1
⎫
⎛ ⎞ pps −p
−1
⎪
 ⎪
p − p ⎬
s

⎝ ps − 1 ⎠ s
p p−1
s p p−1 ⎪
⎪
Cp s −1
(p − 1) ⎭

upX0
the constant Cp will be fixed later and S = inf
X0 \{0}
  pp is
p
|u| dx
Ω
the best Sobolev constant (see [15,21]).
2.4. Preliminary Results
Theorem 1. [22] Let {un }n be a bounded sequence in X0 . Then, up to a sub-
sequence, there exists u ∈ X0 , two Borel regular measures μ and ν, J denu-
merable, xj ∈ Ω, νj , μj ≥ 0 with νj + μj > 0 j ∈ J , such that
un −→ u weakly in W0s,p (Ω) and strongly in Lp (Ω)
n→+∞
∗  ∗
|Ds un |p  dμ |un |ps  dν,

dμ ≥ |Ds u|p + μj δxj , μj = μ ({xj }) ,
j∈J
MJOM On a Fractional p-Laplacian Equation with Page 5 of 23 221

 
dν = |u|ps + νi δxj , νj = ν ({xj })
j∈J

and
p
p
μj ≥ Sνj s for all j ∈ J .

Lemma 1. [22] Let N > sp. For any u ∈ Lps (RN ), it holds

|u|p
lim δ N N +ps
dx = 0.
δ→0 BδC |x|

Lemma 2. [12, Theorem 13.44] Assume that Ω ⊂ RN is a bounded open.

Assume that

sup fn q ≤ K < +∞ and fn −→ f0 a.e in Ω.

n

If 1 < q < +∞, then fn  f0 in Lq (Ω).

We close this section by recalling the well-known algebraic inequality

Lemma 3. [17,26] Let ζ and ξ be in RM and p ∈ [1, +∞[. One has

⎧ p
 p−2 ⎨ C|ζ − ξ| if p ≥ 2
p−2
|ζ| ζ − |ξ| ξ (ζ − ξ) ≥ |ζ − ξ|2 (6)
⎩ 2−p if 1 < p < 2.
(|ζ| − |ξ|)

3. Main Result
Theorem 2. Under the assumptions (5) and (H), the problem (1) admits at
least one weak solution u0 . Moreover, u0 fulfills the following properties
1.

u0 ∈ Λ, (7)

2.

I(u0 ) = inf I(v) < 0, (8)

v∈Λ

3.
  1
(p − 1)(ps − p)tp0 p−1 1 p−1
f −1 ≤ u0 X0 ≤ p 1 −  f −1 (9)
2p(ps − 1) ps
t0 = t0 (uf ) is as below
  p1−p 
(p − 1) uf pX0 s
(−Δ)sp uf = f in Ω
0 < t0 < where
p
(ps − 1) uf pss uf = 0 on RN \Ω.
221 Page 6 of 23 O. Saifia and J. Vélin MJOM

4. Some Useful Lemmas

In order to prove our principal result, we need several techniques which are
ordered in the following way;
Lemma 4. The functional I is coercive and bounded below in Λ.
Proof. For every u ∈ Λ, one has

1 1 
I(u) = upX0 −  |u|p dx − f, u
p ps Ω
   
1 1 1
≥ −  upX0 − 1 −  f −1 uX0 .
p ps ps
Since p > 1, then I is coercive.
By Young inequality, we have
   
1 1 p 1
I(u) ≥ − uX0 − − 1 f −1 uX0
p ps ps
 
1 1 Θp
≥ −  upX0 − upX0
p ps p
 
1 p−1 p
− − 1 p f −1
p−1
for any Θ > 0.
ps pΘ p−1
1
p
p
By chosen Θ = 1 − , therefore we get
ps
 
1 p−1 p
I(u) ≥ − − 1 p f −1 .
p−1

ps pΘ p−1
Therefore,
 
1 p−1 p
c0 = inf I(u) ≥ − 
− 1 p f −1 .
p−1
(10)
Λ ps pΘ p−1
Consequently, I is bounded below on the set Λ. 
Now, for any u fixed in X0 , we define on ]0, +∞[, the function γ(u) as
follows
 
tp p tps 
γ(u)(t) = I(tu) = uX0 −  |u|ps dx − t f, u . (11)
p ps Ω
Obviously, it derives
1 
γ  (u)(t) = I  (tu), u = I (tu), tu .
t
So, it is immediate to notice that
t is critical point of γ(u) if and only if tu ∈ Λ.

Lemma 5. Under the hypothesis (H), the following assertions hold

1. the set Λ is nonempty.
MJOM On a Fractional p-Laplacian Equation with Page 7 of 23 221

2. If f, u > 0, then there exists a unique positive real t>0 (u) such that
t>0 (u)u ∈ Λ>0 .
Proof. 1. The definition (11) implies that
γ  (u)(t) = mu (t) − f, u
where for every u ∈ X0 , mu designates the auxiliary function
 p
mu (t) = tp−1 upX0 − tps −1 upss . (12)
Then, it is clear that the claim tu ∈ Λ is fulfilled if and only if the
identity γ  (u)(t) = 0 holds or again mu (t) = f, u . In this way, let us
observe that the following properties for the functions mu and γ(u)
(a) mu (0) = 0 and so γ  (u)(0) = − f, u ,
(b) lim γ  (u)(t) = lim mu (t) = −∞,
t→+∞ t→+∞
p
1)tp−2 upX0 − (ps − 1)tps −2 upss .

 
(c) γ (u)(t) = mu (t) = (p −
Some simple computations show that
⎛  1 ⎞ ⎛  1 ⎞
p p
(p − 1)u X0
ps −p
(p − 1)u X0
ps −p

γ  (u) ⎝ p 
⎠ = mu ⎝
p 
⎠ = 0. (13)
(ps − 1)upss (ps − 1)upss

Moreover, since p < ps , it gets

⎛  1 ⎞
(p − 1)upX0 ps −p
mu ⎝ p
⎠
(ps − 1)upss
 p
 pp−1
 −p p
s −1
p uX0
s
ps − 1 p
s −p ps − p
= uX0 p  > 0.
ups | p−1 p−1
s

On the other hand, it is obvious that

⎛  p1−p ⎞
p
(p − 1)u s

γ  (u) ⎝ X0
p
⎠
(ps − 1)upss
  pp−1 ps −1
upX0
 −p
p
s
ps − 1 ps −p ps − p
≥ uX0 p p − f −1 uX0 .
Cp s uXs p−1 p−1
0

Therefore, after using the theorem 6.5 in [9], more precisely, this last
estimate becomes
⎧
⎛  ⎞ ⎪⎡ ⎤ pps −p
−1
p 
1
⎪
⎨ s
(p − 1)uX0 ps −p 
ps − 1
γ  (u) ⎝ p
⎠ ≥ uX0 ⎣ ⎦ ×

(ps − 1)ups
s ⎪
⎪ p  p−1
s p
⎩ Cp s −1
(p − 1) (14)


p −p
× s − f −1 .
p−1
Thanks to the ⎛hypothesis (H), the right hand in (14) remains positive
 p
 p1−p ⎞
(p − 1)uX0 s

and so γ  (u) ⎝ p
⎠ > 0. Hence, the properties (b)
(ps − 1)upss
221 Page 8 of 23 O. Saifia and J. Vélin MJOM

and (c) and (14) induce successively that γ  (u) attains its maximum at
  1
(p − 1)upX0 ps −p
tmax = p and moreover, one has
(ps − 1)upss
 p
 pp−1
 −p p
s −1

 p u X0
s
p 
s − 1 p
s −p ps − p
γ (u)(tmax ) = uX0 p − f, u .
ups p−1 p−1
s

So, using the classical Intermediate Values Theorem, we conclude that

there exists tf satisfying tmax < tf < +∞ and γ  (u)(tf ) = 0. Thus, tf u
is in Λ.
Now, to end this first part of the proof, let us observe that γ  (u)
is decreasing in ]tmax , +∞[, so it derives that γ  (u)(t) = mu (t) < 0
for any t such that t > tmax and consequently taking t = tf , we get
1 
γ (u)(tf ) < 0. From definition of Λ<0 , it means that tf u ∈ Λ<0 .
t2f
2. Now, we deal with the second assertion in the announcement. To do it,
let us choose u ∈ X0 satisfying f, u > 0 (for instance u = uf , where
uf designates the unique solution of (−Δ)sp u = f in Ω and u = 0
in Ωc ). Clearly, from the property (a) and γ  (u)(tmax ) > 0, the use of
the Intermediate Value Theorem ensures that there is t0 ∈]0, tmax [ such
that γ  (u)(t0 ) = 0 which means that t0 u ∈ Λ. In addition, since γ  (u)
increases in ]0, tmax [, we deduce that γ  (u)(t0 ) > 0. Hence, according to
the definition of the set Λ>0 , we conclude that t0 u is in Λ>0 .

Lemma 6. Let f is as in the hypothesis (H). The set Λ0 is empty.
Proof. Argue by contradiction. We suppose that there exists an element u in
p
Λ such that φ (u), u = 0. Then, after substituting upss , it results that
0 ≤ (p − ps )upX0 + (ps − 1)f −1 uX0
thus,
1
ps − 1 p−1
uX0 ≤ f −1 . (15)
ps − p
On the other hand, since u ∈ Λ and φ (u), u = 0, we get


(p − 1)upX0 − (ps − 1) |u|ps dx = 0
Ω
1/p
and remember us that ups ≤ S uX0 , therefore, the estimate follows
1
p − 1 −ps p
s −p
S ≤ uX0 . (16)
ps − 1
From (15) and (16), we obtain
p−1

p − p (p − 1)S ps p
s −1
f −1 ≥ s
ps − 1 ps − 1
which contradicts the assumption (H). It results in the set Λ0 is empty. 
MJOM On a Fractional p-Laplacian Equation with Page 9 of 23 221

Lemma 7. Assume that f fulfills the condition (H), then for each u ∈ Λ,
there exist δ > 0 and a differentiable function ζ : B(0, δ) ⊂ X0 −→ R+ such
that
ζ(0) = 1, ζ(w)(u − w) ∈ Λ, for every w ∈ B(0, δ)
and

|u(x) − u(y)|p−2 (u(x) − u(y))
(w(x) − w(y)) dxdy
R2N |x − y|N +ps
ξ  (0), w = p p
 (p − 1)upX0 − (ps − 1)upss

(17)
ps |u|ps −2 uwdx + f, w
Ω
− p .
(p − 1)upX0 − (ps − 1)upss

Proof. For every u ∈ Λ, we define a function F (u) : R × X0 −→ R by

F (u)(ζ, w) = I  (ζ(u)(u − w)) , ζ(u)(u − w) .
Setting [u, w] = u − w, we have again

p |[u, w](x) − [u, w](y)|p
F (u)(ζ, w) = ζ dxdy
R2N |x − y|N +ps

 
−ζ ps |[u, w]|ps dx − ζf, [u, w] .
Ω
Consequently, the partial derivative of F (u) with respect to the second
variable is given by
F (u)(ζ, w + λv) − F (u)(ζ, w)
D2 F (u)(ζ, w), v = lim
λ→0 λ

|[u, w](x) − [u, w](y)|p−2 [[u, w](x) − [u, w](y)]
= pζ p (v(x) − v(y)) dxdy (18)
R2N |x − y|N +ps

 
−ps ζ ps |(u − w)(x)|ps −2 [u, w](x)vdx − ζ f, v , ∀v ∈ X0
Ω

and deriving the function F (u) with respect to the first variable, we obtain

dF |[u, w](x) − [u, w](y)|p
(ζ, w) = pζ p−1 dxdy
dζ R2N  |x − y|N +ps (19)
 
− ps ζ ps −1 |[u, w](x)|ps dx − f, [u, w] .
Ω
Therefore,
F (u)(1, 0) = I  (u), u = 0
and
 
dF (u) |u(x) − u(y)|p  
(1, 0) = p N +ps
dxdy − p s |u(x)|ps dx − f, u
dζ R2N |x − y| Ω

p 
= uX0 − |u(x)|ps dx − f, u + (p − 1)upX0 − (ps − 1)
Ω

p
|u(x)| s dx
Ω
221 Page 10 of 23 O. Saifia and J. Vélin MJOM

= I  (u), u + γ  (u)(1).
Let us notice that since u belongs to Λ, then the first term is zero. Moreover,
the second is φ (u), u and so from the lemma 6, it is non equal to zero.
dF (u)
Consequently, (1, 0) is non equal to zero and also to the definition
dζ
(11)
dF (u) p
(1, 0) = (p − 1)upX0 − (ps − 1)upss . (20)
dζ
Applying the Implicit function theorem, we deduce that there exists a func-
tion
ξ : B(0, δ) ⊂ X0 −→ R+
such that
1. ξ(0) = 1,
2. F (u)(ξ(w), w) = 0, ∀w ∈ B(0, ε),
D2 F (u)(ξ(w), w)
3. ξ  (w) = − , for each w ∈ B(0, δ).
D1 F (u)(ξ(w), w)
Then, we get
F (u)(ξ(w), w) = I  (ξ(w)(u − w)) , ξ(w)(u − w) = 0,
that is
ξ(w)(u − w) ∈ Λ, for every w ∈ B(0, δ).
Moreover,
D2 F (u)(1, 0), w
ξ  (0), w = − ,
D1 F (u)(1, 0)
or again

|u(x) − u(y)|p−2 (u(x) − u(y))
p (w(x) − w(y)) dxdy
R2N |x − y|N +ps
ξ  (0), w = p
 (p − 1)upX0 − (ps − 1)upss

ps |u|ps −2 uw dxdy + f, w
Ω
− p .
(p − 1)upX0 − (ps − 1)upss

Lemma 8. For each u ∈ Λ<0 , there exist δ > 0 and a differentiable function
ξ − : B(0, δ) ⊂ X0 −→ R+
such that ξ − (0) = 1 and ξ − (w)(u − w) ∈ Λ− .
Proof. According to the lemma 7, there exist δ > 0 and a differentiable
function ξ − : B(0, δ) ⊂ X0 −→ R+ such that ξ(0) = 1, ξ − (w)(u − w) ∈ Λ for
every w ∈ B(0, δ) and (17) holds.
Since,


γξ− (w)(u−w) (1) = (p − 1)ξ − (w)(u − w)pX0 − (ps − 1) |ξ − (w)(u − w)|ps dx,
Ω
MJOM On a Fractional p-Laplacian Equation with Page 11 of 23 221

by continuity of ξ − , we get
γξ− (w)(u−w) (1) < 0 for all w ∈ B(0, δ1 ), δ1 < δ is small enough.
That is ξ − (w)(u − w) ∈ Λ<0 . 

5. Proof of Theorem 2
Proof. The proof is divided into several steps.
Step 1 Existence of a minimizing solution for the problem (1).
From (10), the functional I is bounded below. So, it results that
 
1 p−1 p
c0 ≥ − 
− 1 p f −1 .
p−1

ps pΘ p−1
The Ekeland variational principle ensures there exists a sequence {un }n∈N in
Λ satisfying
(i)
1
I(un ) < c0 + , (21)
n
(ii)
1
I(w) ≥ I(un ) −un − w for all w ∈ Λ. (22)
n
Let uf be the unique solution of the Dirichlet-type condition problem

(−Δ)sp u = f in Ω
(23)
u = 0 on RN \Ω.
Therefore, after multiplying (23) by uf , it is easy to obtain
f, uf = uf pX0 > 0.
The lemma 5 implies that there exists t0 = t0 (uf ) such that t0 (uf )uf ∈ Λ>0
and consequently,

1 p p 1 ps 
I(t0 (uf )uf ) = t0 uf X0 −  t0 |uf |ps dx − t0 f, uf
p ps Ω
  
1 p p 1 p 
= − 1 t0 uf X0 − (  − 1)t0 s
|uf |ps dx
p ps
   Ω
1 p p p 1 p−1
≤ − 1 t0 uf X0 − t0 −1 uf pX0
p ps ps − 1
(p − 1)(p − ps ) p
≤ t0 uf pX0
pps
(p − 1)(p − ps ) p
= t0 f p−1 .
pps
We deduce that
(p − 1)(p − ps ) p
c0 ≤ t0 f p−1 < 0.
pps
221 Page 12 of 23 O. Saifia and J. Vélin MJOM

On the other hand, by taking n large enough and from (21), one has

1 p 1 
I(un ) = un X0 −  |un |ps dx − f, un (24)
p ps Ω
1 c0
≤ c0 + < < 0. (25)
n 2
Thus, taking into account that (un )n ∈ Λ, we rewrite I(un ) as follows
1 1 $ %
I(un ) = un pX0 −  un pX0 − f, un − f, un
p p
  s  
1 1 p 1 c0
= − un X0 − − 1 f, un < < 0.
p ps ps 2
Therefore, it is easy to deduce that
 
1 c0
− 1 −  f, un <
ps 2
which gives,
(p − 1)(p − ps )tp0 f p−1
f −1 un X0 ≥ f, un > − . (26)
2p(ps − 1)
Hence,
(p − 1)(ps − p)tp0
un X0 ≥ f p−1
−1 . (27)
2p(ps − 1)
Also, from (24), we obtain
 
1 p 1
un X0 − 1 −  f, un < 0.
p ps
Thus,
  1
p−1
1
un X0 ≤ p 1−  f −1 . (28)
ps
Consequently, from (27) and (28), we get
  1
(p − 1)(ps − p)tp0 p−1 1 p−1
f −1 ≤ u 
n X0 ≤ p 1 − f −1 . (29)
2p(ps − 1) ps
Step 2: Limiting behavior of the minimizing sequence of the problem (1).
Now, we show the claim
I  (un )−1 −→ 0. (30)
n→+∞

Indeed, by the lemma 7, for every n ∈ N , there exist δn > 0 and ξn such
that
ξn : B(0, δn ) ⊂ X0 −→ R+ ,
ξn (0) = 1, and ξ(w)(un − w) ∈ Λ, w ∈ B(0, δn ).
MJOM On a Fractional p-Laplacian Equation with Page 13 of 23 221

u
Let 0 < ρ < δn , u ∈ X0 \{0} and wρ = ρ .
uX0
Thus, wρ X0 = ρ, wρ ∈ B(0, δn ) and unρ = ξn (wρ )(un − wρ ) ∈ Λ.
From (22), we obtain
1
I(unρ ) − I(un ) ≥ − un − unρ X0 . (31)
n
Since wρ X0 = ρ, ξn is of class C(B(0, δn ); R+ ), ξn (0) = 1, then it results
that
unρ − un X0 =  (ξn (wρ ) − 1) un − ξn (wρ )wρ X0 ∼ ρ. (32)
0
Moreover, I is Fréchet differentiable, then
& '
I(unρ ) − I(un ) − I  (un ), unρ − un
lim = 0.
ρ→0 unρ − un X0
Therefore,
& '
I(unρ ) − I(un ) = I  (un ), unρ − un + Oρ (unρ − un X0 )
Oρ (unρ − un X0 )
with lim = 0. (33)
ρ→0 unρ − un X0
From (31) and (33), one has
1
I  (un ), ξn (wρ )(un − wρ ) − un + wρ − wρ ≥ − un − unρ X0
n
+ Oρ (unρ − un X0 ).
Therefore,
1
I  (un ), −wρ + (ξn (wρ ) − 1) I  (un ), un − wρ ≥ − un − unρ X0
n
+ Oρ (unρ − un X0 ).
&  '
Therefore, taking into account that unρ ∈ Λ i.e I (unρ ), unρ = 0,
& '
I  (un ), −wρ + (ξn (wρ ) − 1) I  (un ) − I  (unρ ), un − wρ
1
≥ − un − unρ X0 + Oρ (unρ − un X0 ).
n
u
We replace wρ by ρ in the left hand and we obtain
u
( )
 u 1 Oρ (unρ − un X0 )
I (un ), ≤ un − unρ X0 +
u nρ ρ
(ξn (wρ ) − 1) &  '
+ I (un ) − I  (unρ ), un − wρ
ρ
1 Oρ (unρ − un X0 )
≤ un − unρ X0 +
nρ ρ
ξn (wρ ) − 1 
+ [I (un ), un − I  (un ), wρ ]
ρ

ξn wρ ) − 1 1 &  '
− I (unρ ), unρ .
ρ ξn (wρ )
221 Page 14 of 23 O. Saifia and J. Vélin MJOM

Some calculations combined with (32) and (33), related to the sequence (unρ ),
bring the following estimate
 
I  (un ), u 1 Oρ (unρ − un X0 ) unρ − un X0
ξn (wρ ) ≤ + · .
u n unρ − un X0 ρ
Thus, since u is arbitrary in X0 \ {0}, thereby the claim (30) is attained.
That means,
I  (un )−1 −→ 0.
n→+∞

Step 3 : Convergence of the minimizing sequence to a solution of the

problem (1).
Before continuing, return to (29). This implies that there exists u0 in
X0 such that un weakly converges to u0 in X0 .
So, we claim that I  (un ) weakly converges to I  (u0 ) in X  .
Indeed, let ϕ be in X0 , we show that

|un (x) − un (y)|p−2 (un (x) − un (y))
lim (ϕ(x) − ϕ(y)) dxdy
n→+∞ R2N |x − y|N +ps

− |un |ps −2 un ϕ dxdy
 Ω
|u0 (x) − u0 (y)|p−2 (u0 (x) − u0 (y))
= (ϕ(x) − ϕ(y)) dxdy
R
2N |x − y|N +ps

− |u0 |ps −2 u0 ϕ dxdy.
Ω
More precisely, taking into account the Dirichlet-type condition on Ωc , the
integral over R2N is expressed as follows

|un (x) − un (y)|p−2 (un (x) − un (y))
(ϕ(x) − ϕ(y)) dxdy
R2N |x − y|N +ps
 
|un (x) − un (y)|p−2 (un (x) − un (y))
= (ϕ(x) − ϕ(y)) dxdy
RN \Ω RN \Ω |x − y|N +ps
 
|un (x) − un (y)|p−2 (un (x) − un (y))
+2 (ϕ(x) − ϕ(y)) dxdy
Ω RN \Ω |x − y|N +ps
 
|un (x) − un (y)|p−2 (un (x) − un (y))
+ (ϕ(x) − ϕ(y)) dxdy
Ω Ω |x − y|N +ps

=2 |un (x)|p−2 un (x)ϕ(x)VΩ (x)dx
Ω
 
|un (x) − un (y)|p−2 (un (x) − un (y))
+ (ϕ(x) − ϕ(y)) dxdy
Ω Ω |x − y|N +ps
where VΩ is currently named the potential and it is given by

dy
VΩ (x) = N +ps
. (34)
RN \Ω |x − y|
MJOM On a Fractional p-Laplacian Equation with Page 15 of 23 221

We need to point out that

 
|un (x) − un (y)|p−2 (un (x) − un (y))
lim (ϕ(x) − ϕ(y)) dxdy
n→+∞ Ω Ω |x − y|N +ps
 
|u0 (x) − u0 (y)|p−2 (u0 (x) − u0 (y))
= (ϕ(x) − ϕ(y)) dxdy.
Ω Ω |x − y|N +ps
Indeed, since
 p 1/p   1
|un (x) − un (y)| 1 p−1
dxdy = un X0 ≤ p 1−  f −1
R2N |x − y|N +ps ps
and using the decomposition, we can write
    
=2 +
R2N Ω RN \Ω Ω Ω

hence, the above formula becomes

   p−2
 p−1

p

|u
 n (x) − u n (y)| (u(x) − u(y)) 
  dxdy
Ω Ω 
p−1
|x − y|(N +ps) p (35)
 * + p   p
p−1
p−1 p−1 1
+2 |un (x)| VΩ (x)dx ≤ p 1 −  f −1 .
Ω ps
From above, we address some immediate consequences.
p−2
|un (x) − un (y)| (u(x) − u(y)) 
First, is bounded in Lp (Ω × Ω). In
(N +ps) p−1
|x − y| p

p−2
|un (x) − un (y)| (u(x) − u(y)) 
this way, weakly converges in Lp (Ω×Ω). A
(N +ps) p−1
|x − y| p

subsequence denoted again by un weak converges to u0 in W0s,p (Ω). Therefore,

recall that the immersion W0s,p (Ω) → Lr (Ω) for all r ∈ (1, ps [ is compact.
Then, the subsequence un strongly converges to u0 in Lr (Ω) for all r ∈
(1, ps [. Consequently, un converges to u0 almost everywhere in Ω. Consider
the sequence fn and f0 defined on Ω × Ω as
|un (x) − un (y)|p−2 (un (x) − un (y))
fn (x, y) = p−1
|x − y|(N +sp) p (36)
p−1
f0 (x, y) = |u0 (x) − u0 (y)|p−2 (u0 (x) − u0 (y))|x − y|(N +sp) p .
It results that
  p
p−1
1
sup fn  p−1 ≤ p 1 −  f −1
n p ps
and
fn (x, y) −→ f0 (x, y).
a.e in Ω×Ω

Thanks to a suitable adaptative high dimensional version of the lemma 2, we


can conclude that fn weakly converges to f0 in Lp (Ω). In other words, for
221 Page 16 of 23 O. Saifia and J. Vélin MJOM

ϕ(x) − ϕ(y)
any ϕ in X0 , since 1 ∈ Lp (Ω × Ω), then the limit below holds
|x − y|(N +sp) p
 
|un (x) − un (y)|p−2 (un (x) − un (y))
lim (ϕ(x) − ϕ(y)) dxdy
n→+∞ Ω Ω
  |x − y|N +sp (37)
|u0 (x) − u0 (y)|p−2 (u0 (x) − u0 (y))
= (ϕ(x) − ϕ(y)) dxdy.
Ω Ω |x − y|N +ps
p−1 p−1
Now, we are going to prove that |un | un weakly converges to |u0 | u0

in Lp (Ω). To do it, we should turn to the term VΩ expressed by (34). More
precisely, assume that sp > 1, because Ω is a bounded open subset of RN
with Lipschitz boundary, from the Hardy inequality, there exists a positive
constant
 c(Ω, s, p, N ) > 0 depending only on Ω, s, p and N such that the
term |un (x)|p dx is estimated as follows
Ω
   p
p sp |un (x) − un (y)|
|un | dx ≤ diam(Ω) c(Ω, s, p, N ) N +sp
dxdy. (38)
Ω Ω Ω |x − y|
Employing the spherical coordinates transformation, y = x−rΘ(θ1 , · · · , θN −1 )
where r ∈ ]ρ(x), +∞[ and Θ(θ1 , · · · , θN −1 ) ∈ RN −1 , θi ∈ [0, π[, for i =
1, · · · , N − 2, θN −1 ∈ [0, 2π[, we obtain
  ⎛N −2  ⎞
+∞ , π
VΩ (x) = r−(1+sp) ⎝ sinN −1−j (θj )dθj ⎠
ρ(x) j=1 0

where ρ(x) = dist (x, ∂Ω) , x ∈ Ω.

 0 such that VΩ (x) ≤
Then, it is clear that there exists a constant C >
−sp p
Cρ (x), for any x ∈ Ω. Consequently, the term |un (x)| VΩ (x)dx is
Ω
estimated as follows
    p
p |un |p |un (x) − un (y)|
|un (x)| VΩ (x)dx ≤ dx ≤ cΩ,s,p,N N +sp
dxdy.
Ω Ω ρ(x)sp Ω Ω |x − y|
p−1
To apply one more time the Lemma 2, we choose fn = un |un |p−2 VΩ p .
Hence, from this, the estimates (35) and (38) ensure that the sequence fn is in
  p
p−1
p 1
L (Ω) and moreover sup fn p ≤ p 1 −  f −1 and also fn → f0
n ps
almost everywhere in Ω.
On the other side, because ps > 1, un weakly converges to u0 in X0 ,
then passing to a subsequence again denoted by un , we can assume that un

a.e converges to u0 in Ω which implies that Un = un |un |ps −2 a.e converges to

U0 = u0 |u0 |ps −2 by continuity arguments. From the lemma 2, it results that
 p
Un weakly converges to U0 in Lqs (Ω) where qs =  s . Indeed, because
ps − 1
s,p p
the immersion W0 (Ω) → L (Ω) is continuous, there is a positive constant
s

C non-depending on n such that un ps ≤ Cun 1,s,p . Hence, from estimate
MJOM On a Fractional p-Laplacian Equation with Page 17 of 23 221

(29), it follows that

  1
p−1
1
un  ≤ C p 1 −  f −1
p
s
, ∀n ∈ N
ps
  1
p−1
1
which implies that Un qs ≤ C p 1 −  f −1 , ∀n ∈ N.
ps
Therefore, by applying Lemme 2, we deduce that the sequence Un

weakly converges to U0 in Lqs (Ω). In other words, we obtain particularly
that
 
p −2 
un (x)|un (x)| s wdx −→ u0 (x)|u0 (x)|ps −2 wdx, ∀w ∈ X0 . (39)
Ω n→+∞ Ω

The claim I (un ) weakly converges to I  (u0 ) in X  follows from (37) and (39).


Which implies that u0 obeys to the problem (1).

 
(−Δ)sp u0 = |u0 |ps −2 u0 + f (x) in Ω
u0 = 0 on RN \Ω.
and
c0 ≤ I(u0 ) ≤ lim inf I(un ) = c0 .
n→∞

Therefore, the assertion (8) in the statement of Theorem 2 is fulfilled.

In particular, u0 ∈ Λ. 

Lemma 9. The sequence un strongly converges to u0 in X0 .

Proof. We define the cut-off function

 
1
Φε (x) = Φ (x − xj )
ε
where Φ ∈ C0∞ (RN ) such that 0 ≤ Φ(x) ≤ 1, φ(x) = 1 in B1 (0), Φ(x) = 0 in
B2c (0) and |∇Φ(x)| ≤ 2 in RN .
In the theorem 1, we may assume that

un  u in X0 , |Ds un |p  dμ in measures and |un |ps  dν in measures(40)

|un (x) − un (y)|p
where |Ds un (x)|p = dy.
RN |x − y|N p+s
Now, consider the sequence {Φε un }n . The limit on I  (un ) given by (30)
induces also particularly that I  (un ), φε un tends to 0 when n goes to +∞,
and then, we obtain
 
Φε dμ − Φε dν − f, Φε u ≤ − lim In,ε (41)
Ω Ω n→+∞

where

|un (x) − un (y)|p−2 (un (x) − un (y))
In,ε = (Φε (x) − Φε (y)) un (y)dxdy.
R2N |x − y|N +ps
Thus, Hölder inequality and boundedness of {un }n imply
221 Page 18 of 23 O. Saifia and J. Vélin MJOM

 p−1
 p1
|un (x) − un (y)|p |Φε (x) − Φε (y)|p
p
|In,ε | ≤ dxdy |un (y)|p dxdy
R2N |x − y|N +ps R2N |x − y|N +ps
  p1
|Φε (x) − Φε (y)|p p
≤c |un (y)| dxdy .
R2N |x − y|N +ps
We will check that
  p1
|Φε (x) − Φε (y)|p
lim lim sup |un (y)|p dxdy = 0.
ε→0 n→∞ R2N |x − y|N +ps
Let ε > 0, Φε ∈ Cc∞ (B2ε (xj )) satisfies
C
Φε > 0, Φ|Bε = 1 |Φε | < 1 and |∇Φε | < .
ε
c c
Let A = (B2ε × B2ε ) ∪ (Bε × B2ε ) ∪ (B2ε × Bε ) .
2N
We notice that: on R \ A : |Φε (x) − Φε (y)| = 0.
Since |Φε (x) − Φε (y)| ≤ Cε−1 |x − y| on B2ε × B2ε , we have
  p
|Φε (x) − Φε (y)|p p |C p ε−p |z|
|u n (y)| dxdy ≤ |un (y)|p dydz
B2ε ×B2ε |x − y|N +ps B2ε ×B4ε |z|N +ps

  pp   pp
Cεp ε−p p s 1− pp p s
≤ N +ps |un | dx
s |B2ε | s ≤C |un | dx
s −→ 0.
|ε| B2ε B2ε ε→0

c
On B2ε × Bε , we get
 
|Φε (x) − Φε (y)|p p |un (y)|p
|u n (y)| dxdy ≤ C dxdy
B2ε ×B2ε |x − y|N +ps c ×B |x − xj |
B2ε ε
N +ps

  pp   pp
p s 1− pp 1 p s
≤C |un | dx
s |B2ε | s ≤C |un | dx
s .
B2ε εps B2ε
c
Finally, on Bε × B2ε

|y − yj |
|x − y| ≥
2
and thus
 
|Φε (x) − Φε (y)|p |un (y)|p
|un (y)|p dxdy ≤ εN dy
c
Bε ×B2ε |x − y|N +ps c
Bε ×B2ε |y − yj |N +ps


N |un (y)|p
≤ Cε dy −→ 0.
c
B2ε |y − yj |N +ps ε→0

Consequently, from (41), we obtain

 
Φε dμ − Φε dν ≤ lim f, Φε u .
Ω Ω ε→0
MJOM On a Fractional p-Laplacian Equation with Page 19 of 23 221

Which gives μj −νj ≤ 0 provided lim f, Φε u = 0. Indeed, the Riez represen-
ε→0

tation for Lebesgue’s integral
 ensures that there exists a unique g ∈ Lp (Ω)
such that f, Φε u = g Φε u dx. Furthermore, the description of Φε im-
 Ω 

plies that g Φε u dx = g Φε u dx. The right hand tends to 0 as ε.

Ω Bε (xj )
Therefore, the claim occurs from it.
Therefore, taking into account the theorem 1, we may compare the
masses νj and μj as follows
p

νj ≥ μj and μj ≥ Sνjp .
ps
Thus, arising from this, either on the other hand νj = 0 or νj ≥ S N .
ps
Assume by contradiction that νj ≥ S N with νj = 0 thus μj = 0. Else,
since it holds
  
dμ ≥ |Ds un |p + μj δxj dν = |u|ps + νj δxj μj ≥ νj S, (42)
j∈J j∈J
 
1 1
therefore, multiplying the first inequality in (42) by − , then we
p ps
obtain
     
1 
p
1 1 1 1 |u0 (x) − u0 (y)| 1
dμ −  ≥ −  dy + − μj δxj .
p ps p ps RN |x − y|N +sp p ps
j∈J

After integrating over RN , we get

     
1 1 p 1 1 p 1 1 
− lim un X0 ≥ − u0 X0 + − μ(xj ).
p ps n→+∞ p ps p ps
j∈J

Also, it yields
    
1 1 1
−  lim un pX0 − 
− 1 lim f un dx
p  ps n→+∞   ps  n→+∞  Ω  
1 1 1 1 1 
≥ −  u0 pX0 − − 1 lim f u n dx + − μ(xj ).
p ps ps n→+∞ Ω p ps
j∈J

In other words
     
1 1 1 1 1 
lim −  un pX0 − − 1 f u n dx ≥ I(u 0 )+ − μ(xj ).
n→+∞ p ps ps Ω p ps j∈J

Which means that

 
1 1
lim I(un ) ≥ I(u0 ) + − μ(xj ).
n→+∞ p ps
j∈J

Taking into account (21), we find

 
1 1
c0 ≥ lim I(un ) ≥ I(u0 ) + −  μ(xj )
n→+∞ p ps
j∈J
221 Page 20 of 23 O. Saifia and J. Vélin MJOM

since I(u0 ) = c0 , then

 
1 1
c0 + − μ(xj ) ≤ c0 .
p ps
j∈J

We deduce that μ(xj ) ≤ 0 and hence μ(xj ) = ν(xj ) = 0, ∀j ∈ J .
j∈J
p ps
Then, the assumption νj ≥ S 1− p = S N implies S = 0. This is impossible.

We conclude that νj = 0 ∀j ∈ J . Therefore, it holds lim |un |ps dx =
n→+∞ Ω


|u0 |ps dx.
Ω
By the well-known Brezis-Lieb’s lemma, we have


lim |un − u0 |ps dx = 0.
n→+∞ Ω
On the other hand, since
I  (un ) − I  (u0 ), un − u0 = 0 for every n in N,
that means

|un (x) − un (y)|p−2 (un (x) − un (y))
[(un − u0 )(x) − (un − u0 )(y)] dxdy
R2N |x − y|N +ps


− |un (x)|ps −2 un (x) · (un − u0 )(x) dx − f, un − u0 
Ω

|u0 (x) − u0 (y)|p−2 (u0 (x) − u0 (y))
− [(un − u0 )(x) − (un − u0 )(y)] dxdy
R2N |x − y|N +ps


+ |u0 (x)|ps −2 u0 (x) · (un − u0 )(x) dx + f, un − u0 
Ω

tends to 0.

Then, using the fact that un strongly converges to u0 in Lps (Ω), it
follows that

|un (x) − un (y)|p−2 (un (x) − un (y))
[(un − u0 )(x) − (un − u0 )(y)] dxdy
R2N |x − y|N +ps
tends to

|u0 (x) − u0 (y)|p−2 (u0 (x) − u0 (y))
[(un − u0 )(x) − (un − u0 )(y)] dxdy.
R2N |x − y|N +ps
or again, after some rearrangements, it holds

  p−2  p−2 
   
 un (x) − un (y)  un (x) − un (y)  u0 (x) − u0 (y)  u0 (x) − u0 (y)
  −  
R2N  |x − y| N +ps
p  N +ps
|x − y| p  |x − y| N +ps
p  N +ps
|x − y| p
 
un (x) − un (y) u0 (x) − u0 (y)
× N +ps − N +ps dxdy
|x − y| p |x − y| p
tends to 0.
MJOM On a Fractional p-Laplacian Equation with Page 21 of 23 221

Apply the algebraic inequality (6) of the lemma 3 by taking

un (x) − un (y) u0 (x) − u0 (y)
ζ= N +ps , ξ= N +ps and M = 1,
|x − y| p |x − y| p

we obtain
  
 u (x) − u (y) u (x) − u (y) p
 n n 0 0 
lim  −  dxdy = 0.
n→+∞ R2N  
N +ps N +ps
|x − y| p |x − y| p

Hence, we conclude that un strongly converges to u0 in X0 . The assertion

(9) in the statement of Theorem 2 involves by passing to the limit on n in
the estimate (29). The proof of the lemma 9 is achieved. 

Acknowledgements
The authors would like to thank Reviewer and Editor for their valuable com-
ments and suggestions which improved the quality of the paper.

Author contributions Not applicable.

Funding No funding.

Availability of Data and Materials Not applicable.

Declarations

Conflict of interest Not applicable.

Ethical Approval Not applicable.

Publisher’s Note Springer Nature remains neutral with regard to jurisdic-
tional claims in published maps and institutional affiliations.
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive
rights to this article under a publishing agreement with the author(s) or other
rightsholder(s); author self-archiving of the accepted manuscript version of
this article is solely governed by the terms of such publishing agreement and
applicable law.

References
[1] Ambrosio, V.: Multiple solutions for a fractional p-Laplacian equation with
sign-changing potential. Electron. J. Diff. Equ. 2016(151), 1–12 (2016)
[2] Ambrosio, V.: Nontrivial solutions for a fractional p-Laplacian problem via
Rabier Theorem. Complex Var. Elliptic Equ. 62(6), 838–847 (2017)
[3] Ambrosio, V.: A multiplicity result for a fractional p-Laplacian problem with-
out growth conditions. Riv. Math. Univ. Parma 9, 53–71 (2018)
[4] Ambrosio, V.: On the multiplicity and concentration of positive solutions for a
p-fractional Choquard equation in RN . Comput. Math. Appl. 78(8), 2593–2617
(2019)
221 Page 22 of 23 O. Saifia and J. Vélin MJOM

[5] Ambrosio, V.: Nonlinear fractional Schrödinger equations in RN , Frontiers in

Elliptic and Parabolic Problems. Birkhäuser/Springer (2021), xvii-662
[6] Ambrosio, V., Isernia, T.: Multiplicity and concentration results for some non-
linear Schrödinger equations with the fractional p-Laplacian, Discrete Contin.
Dyn. Syst. Vol. 38 (2018), No. 11, https://doi.org/10.3934/dcds.2018254.
[7] Brasco, L., Squassina, M., Yang, Y.: Global compactness results for nonlocal
problems, A.I.M.S 11 (2018), No. 3, 391–424
[8] Chabrowski, J.: On multiple solutions for the nonhomogeneous p-Laplacian
with a critical Sobolev exponent. Differ. Integr. Equ. 8(4), 705–716 (1995)
[9] Di Nezza, E., Palatucci, G., Valdinoci, E.: Hitchhiker’s guide to the fractional
Sobolev spaces. Bull. Sci. Math. 136, 521–573 (2012)
[10] Di Castro, A., Kuusi, T., Palatucci, G.: Nonlocal Harnack inequalities. J. Funct.
Anal. 267(6), 1807–1836 (2014)
[11] Di Castro, A., Kuusi, T., Palatucci, G.: Local behavior of fractional p-
minimizers. Ann. Inst. H. Poincaré Anal. Non Linéaire. 33(5), 1279–1299 (2016)
[12] Hewitt, E., Stromberg, K.: Real and Abstract Analysis. Springer Verlag, Berlin
(1975)
[13] Fiscella, A., Pucci, P.: Kirchhoff-Hardy fractional problems with lack of com-
pactness. Adv. Nonlinear Stud. 17(3), 429–456 (2017)
[14] Fiscella, A., Molica Bisci, G., Servadei, R.: Multiplicity results for fractional
Laplace problems with critical growth. Manuscr. Math. 155(3), 369–388 (2018)
[15] Fiscella, A., Pucci, P.: Kirchhoff-Hardy fractional problems with lack of com-
pactness. Adv. Nonlinear Stud. 17(3), 429–456 (2017)
[16] Franzina, G., Palatucci, G.: Fractional p−eigenvalues. Riv. Math. Univ. Parma
5(2), 373–386 (2014)
[17] Glowinski, R., Marroco, A.: Sur l’approximation, par éléments finis d’ordre
un, et la résolution, par pénalisation-dualité d’une classe de problèmes de
Dirichlet non linéaires Revue française d’ Automatique, Informatique et Rech.
Opérationnelle, 9, R2 (1975), 41–76
[18] Iannizzotto, A., Liu, S., Perera, K., Squassina, M.: Existence results for frac-
tional p-Laplacian problems via Morse theory. Adv. Calc. 9, 101–125 (2016)
[19] Lindgren, E., Lindqvist, P.: Fractional eigenvalues. Calc. Var. 49, 795–826
(2014)
[20] Louis-Rose, C., Vélin, J.: On a nonexistence result involving the fractional
p-Laplacian. Riv. Mat. Univ. Parma 6, 345–355 (2015)
[21] Mazýa, V., Shaposhnikova, T.: On the Bourgain, Brezis, and Mironescu the-
orem concerning limiting embeddings of fractional Sobolev spaces. J. Funct.
Anal. 195(2), 230–238 (2002)
[22] Mosconi, S., Squassina, M.: Nonlocal problems at nearly critical growth. Non-
linear Anal. 136, 84–101 (2016)
[23] Molica Bisci, G., Radulescu, V., Servadei, R.: Variational Methods for Nonlocal
Fractional Problems. Cambridge University Press, Cambridge (2016)
[24] Palatucci, G., Pisante, A.: Improved Sobolev embeddings, profile decomposi-
tion, and concentration-compactness for fractional Sobolev spaces. Calc. Var.
Partial. Differ. Equ. 50(3–4), 799–229 (2014)
MJOM On a Fractional p-Laplacian Equation with Page 23 of 23 221

[25] Ros-Oton, X., Serra, J.: Nonexistence results for nonlocal equations with crit-
ical and supercritical nonlinearities. Comm. Partial Differ. Equ. 40, 115–133
(2015)
[26] Simon, J.: Régularité de la solution d’un problème aux limites non-linéaires
dans RN . In Journées d’analyse non-linéaire. Proceedings, Besançon, France
(1977), Lectures notes in Mathematics no 665. Springer-Verlag
[27] Tarantello, G.: Nonhomogenous elliptic equations involving critical Sobolev
exponent. Ann. Inst. H. Poincare Anal. Non-linéaire. 9(3), 281–304 (1992)

Ouarda Saifia
Department of mathematics
Chadli Bendjedid University Eltarf, BP 73
Eltarf 36000
Algeria
e-mail: wsaifia@gmail.com

Jean Vélin
Department of Mathematics and Computer
Laboratory LAMIA, University of Antilles, Campus of Fouillole
Pointe-Pitre 97159
FWI
Guadeloupe
e-mail: jean.velin@univ-antilles.fr

Received: December 24, 2022.

Revised: April 6, 2023.
Accepted: May 6, 2023.