Int. J. Anal. Appl.

(2024), 22:107

A Fractional Elliptic System With Strongly Coupled Critical Terms and

Concave-Convex Nonlinearities

Rachid Echarghaoui, Abdelouhab Hatimi∗ , Mohamed Hatimi

Department of Mathematics, Faculty of Sciences, Ibn Tofail University, Kenitra B. P. 133, Morocco

∗ Corresponding author: [email protected]

Abstract. By the Nehari method and variational method, two positive solutions are obtained for a fractional elliptic
system with strongly coupled critical terms and concave-convex nonlinearities. Recent results from the literature are
extended to the fractional case.

1. Introduction

In this paper, we are concerned with the study of the existence of positive solutions for the
following elliptic system involving the fractional Laplacian for a given 0 < s < 1 :
η α η α
s u = 1 1 |u|α1 −2 |v|β1 u + 2 2 |u|α2 −2 |v|β2 u + λ |u|
q−2 u

( −∆ ) , x ∈ Ω,


∗ ∗ |x|γ

2s 2s



η β η β q−2 v

s v = 1 1 |u|α1 |v|β1 −2 v + 2 2 |u|α2 |v|β2 −2 v + µ |v|

(1.1)

 ( −∆ ) , x ∈ Ω,
2∗s 2∗s |x|γ





on ∂Ω,

u = v = 0,


where η1 , η2 , λ, µ, are positive, 2∗s := 2N

N−2s is the fractional Sobolev critical exponent, N > 2s,
α1 + β 1 = 2∗s , α2 + β2 = 2∗s , and (−∆)s is the spectral fractional Laplacian, in Ω, defined in section
2.
|u|αi −2 u|v|βi and |u|αi |v|βi −2 v, i = 1, 2 are called strongly coupled terms. When η1 = η2 = 1, α1 =
α2 = α, β1 = β2 = β and γ = 0, problem (1.1) becomes the following elliptic system:

(−∆)s u = α2α α−2 u|v|β + λ|u|q−2 u in Ω,
+β |u|





2β
(−∆)s v = α+β |u|α |v|β−2 v + µ|v|q−2 v in Ω,

(1.2)





on ∂Ω.

u = v = 0


Received: May 3, 2024.

2020 Mathematics Subject Classification. 35J60, 47J30.
Key words and phrases. Fractional Laplacian system; strongly coupled critical terms; Nehari manifold.

https://doi.org/10.28924/2291-8639-22-2024-107 © 2024 the author(s).

ISSN: 2291-8639
2 Int. J. Anal. Appl. (2024), 22:107

Systems similar to (1.2), but involving Laplacian or p-Laplacian operators, have been studied
extensively in recent years, for example, the authors in [13] proved that the system admits at least
two positive solutions when (λ, µ) belongs to certain subset of R2 . Later, Hsu [12] obtained the
same results for the p-Laplacian elliptic system. There are other multiplicity results for critical
elliptic equations involving concave–convex nonlinearities, see for example [2, 4].
Caffarelli and Silvestre [8] gave a new formulation of the fractional Laplacians through Dirichlet-
Neumann maps. This is commonly used in the recent literature since it allows us to write nonlocal
problems in a local way and this permits to us use the variational methods for those kinds. To our
best knowledge, there are just a few results in the literature on the fractional system (1.1) with both
concave-convex nonlinearities and the strongly-coupled critical terms. Motivated by the results
mentioned above, in this paper we extend the work [18], where the fibering and Nehari manifold
methods are applicable to obtain two positive solutions for
η1 α1 α −2 β η2 α2 α −2 β |u|q−2 u

2u + λ , x ∈ Ω,

−∆u = |u| 1 |v| 1u + |u| 2 |v|
|x|γ

2∗ 2∗



η1 β1 η2 β2 |v|q−2 v


−∆v = ∗ |u|α1 |v|β1 −2 v + ∗ |u|α2 |v|β2 −2 v + µ (1.3)

 , x ∈ Ω,
|x|γ




 2 2
x ∈ ∂Ω.

u = v = 0,


We point out that we adopt in the paper the spectral (or regional) definition of the fractional
laplacian in a bounded domain and not the integral definition. We shall refer the interested reader
to [14] for a careful comparison between these two different notions. To formulate the main result,
we consider
2 2
Cθ = {(λ, µ) ∈ R2+ : 0 < λ 2−q + µ 2−q < θ},

2 " # 2
 22∗s 2∗s −2
!− 2−q
2∗s − q 2−q 
Λ1 = ∗ Θ ks Sη,α, β (1.4)
2s − 2 (2∗s − q)
where ks is a normalization constant and Θ, Sη,α,β are constants that will be introduced later. We
assume that
qN
(F ) : 1 < q < 2 and 0 ≤ γ < N + sq − .
2
Our main results dealing with the problem (1.1) are the following.

Theorem 1.1. Assume that (F ) holds, Then

(i) system (1.1) has at least one positive solution for all (λ, µ) ∈ CΛ1 .
(ii) there is Λ2 < Λ1 such that (1.1) has at least two positive solutions for (λ, µ) ∈ CΛ2 .

The paper is organized as follows. In a preliminary Section 2 we describe the appropriate

functional setting for the study of problem (1.1), including the definition of an equivalent problem.
In Section 3 we show that the Palais-Smale condition holds for the energy functional associated
with (1.1) at energy levels in a suitable range related to the best Sobolev constants. In Section 4 we
investigate the existence of Palais-Smale sequences and we obtain solutions to some related local
minimization problems. Finally, the proof of Theorem 1.1 is given in Section 5.
Int. J. Anal. Appl. (2024), 22:107 3

2. Functional setting and definitions

In this section, we collect some preliminary facts in order to establish the functional setting. We
denote the upper half-space in RN+
+1
by
n o
RN+
+1
: = z = ( x, y ) = ( x 1 , · · · , x N , y ) ∈ R N +1
: y > 0 ,

the half cylinder standing on a bounded smooth domain Ω ⊂ RN by CΩ := Ω × (0, ∞) ⊂ RN +1

+ , the
cylinder with base Ω and its lateral boundary by ∂L CΩ := ∂Ω × (0, ∞). Let ek be an orthonnormal
basis of L2 (Ω) with kek kL2 (Ω) = 1 forming a spectral decomposition of −∆ in Ω with zero Dirichlet
boundary conditions and λk be the corresponding eigenvalues. Let
  1/2 

 X∞  X∞  

H0 ( Ω ) = 
s 2
Ω λ
2 s
<
   
u = a e ∈ L ( ) : kuk s (Ω) =  a ∞ .
 
j j H
 

 0  j j

 


 j=1 j=1 

For any u ∈ H0s (Ω), the spectral fractional Laplacian (−∆)s , is defined by
X
(−∆)s u = ai λsi ei .
i∈N

Note that kukHs (Ω) = (−∆)s/2 u L2 (Ω)

. The dual space H−s (Ω) is defined in the standard way, as
0
well as the inverse operator (−∆)−s .

Definition 2.1. We say that (u, v) ∈ H0s (Ω) × H0s (Ω) is a weak solution of (1.1) such that for all
(ϕ1 , ϕ2 ) ∈ H0s (Ω) × H0s (Ω), it holds
Z 
s s s s

(−∆) 2 u(−∆) 2 ϕ1 + (−∆) 2 v(−∆) 2 ϕ2 dx
Ω
η1 α1 α1 −2 β1 η2 α2 α2 −2 β2
Z !
= |u| |v| uϕ1 + ∗ |u| |v| uϕ1 dx
Ω 2∗s 2s
(2.1)
η1 β1 α1 β1 −2 η2 β2 α2 β2 −2
Z !
+ |u| |v| vϕ2 + ∗ |u| |v| vϕ2 dx
Ω 2∗s 2s
|u|q−2 u |v|q−2 v
Z !
+ λ ϕ1 + µ ϕ2 dx.
Ω |x|γ |x|γ
Associated with problem (1.1), we consider the energy functional
Z 
1 s s
 1 1
Jλ,µ (u, v) := |(−∆) 2 u|2 + |(−∆) 2 v|2 dx − ∗ Q(u, v) − K(u, v), (2.2)
2 Ω 2s q
where Z 
η1 |u|α1 |v|β1 + η2 |u|α2 |v|β2 dx

Q(u, v) :=
Ω
and
|u|q |v|q
Z  
K(u, v) := λ γ + µ γ dx.
Ω |x| |x|
The functional is well defined in H0s (Ω) × H0s (Ω), and moreover, the critical points of the functional
Jλ,µ correspond to weak solutions of (1.1).
4 Int. J. Anal. Appl. (2024), 22:107

We now conclude the main ingredients of a recently developed technique which can deal
with fractional power of the Laplacian. To treat the nonlocal problem (1.1), we will study a
corresponding extension problem, so that we can investigate problem (1.1) by studying a local
problem via classical variational methods.
We first define the extension operator and fractional Laplacian for functions in H0s (Ω). We refer
the reader to [3, 5, 9] and to the referencess therein.

Definition 2.2. For a function u ∈ H0s (Ω), we denote its s-harmonic extension w = Es (u) to the cylinder
CΩ as the solution of the problem
  

− div y1−2s ∇w = 0 in CΩ




on ∂L CΩ


w=0


on Ω × {0},

w = u.



and
∂w
(−∆)s u(x) = −ks lim+ y1−2s (x, y),
y→0 ∂y
s (C ) as the closure of C∞ (C ) under the norm
where ks is a normalization constant. Define H0,L Ω 0 Ω

Z !1/2
1−2s 2
kwkHs (CΩ ) := ks y |∇w| dxdy .
0,L
CΩ

We will use the following notations:

H := H0s (Ω) × H0s (Ω) , ˜ : = H s (CΩ ) × H s (CΩ ) ,

H 0,L 0,L

  ∂w ∂w
Ls w = − div y1−2s ∇w , := −ks lim y1−2s ,
∂ν s y→0 + ∂y

and k(w1 , w2 )k2˜ = kw1 k2Hs (CΩ )

+ kw2 k2Hs (CΩ )
.
H 0,L 0,L

By the arguments above, we can reformulate our problem (1.1) in terms of the extension problem
as follows

Ls w1 = 0, Ls w1 = 0 in CΩ





on ∂L CΩ




w1 = w2 = 0


on Ω × {0} (2.3)


w1 = u, w2 = v

∂w1 η1 α1 α1 −2 |w |β1 w + η2 α2 |w |α2 −2 |w |β2 w + λ |w1 | w1
q−2

on Ω × {0}

s = 2∗ |w1 |

∂ν 2 1 ∗ 1 2 1 γ



 s 2s |x|
η β η β
 ∂ws2 = 1 ∗ 1 |w1 |α1 |w2 |β1 −2 w2 + 2 ∗ 2 |w1 |α2 |w2 |β2 −2 w2 + µ |w2 | γ w2
 q−2
on Ω × {0},


∂ν 2s 2s |x|

where (w1 , w2 ) = (Es (u), Es (v)).

Int. J. Anal. Appl. (2024), 22:107 5

˜ such that
An energy solution to this problem is a function (w1 , w2 ) ∈ H
Z Z
1−2s
ks y ∇w1 · ∇ϕ1 dxdy + ks y1−2s ∇w2 · ∇ϕ2 dxdy
CΩ CΩ
η 1 α1 η 2 α2
Z !
α1 −2 β1 α2 −2 β2
= |w1 | |w2 | w1 ϕ1 + ∗ |w1 | |w2 | w1 ϕ1 dx
Ω×{0} 2∗s 2s
(2.4)
η1 β1 η2 β2
Z !
α1 β1 −2 α2 β2 −2
+ |w1 | |w2 | w2 ϕ2 + ∗ |w1 | |w2 | w2 ϕ2 dx
Ω×{0} 2∗s 2s
|w1 |q−2 w1 |w2 |q−2 w2
Z !
+ λ ϕ1 + µ ϕ2 dx,
Ω×{0} |x|γ |x|γ
˜
for all (ϕ1 , ϕ2 ) ∈ H.
If (w1 , w2 ) satisfies (2.3), then the trace (u, v) = (w1 (·, 0), w2 (·, 0)) ∈ H is an energy solution to
problem (1.1). The converse is also true. By the equivalence of these two formulations, we will use
both formulations in the sequel to their best advantage. Finally, the energy functional associated
with problem (2.3) is the following,

|w1 |q |w2 |q
Z
1 1
 
Iλ,µ (w) := Iλ,µ (w1 , w2 ) = k(w1 , w2 )k ˜ − 2
λ γ + µ γ dx
2 H q Ω×{0} |x| |x|
Z
1
η1 |w1 |α1 |w2 |β1 + η2 |w1 |α2 |w2 |β2 dx.
 
− ∗
2s Ω×{0}
˜ correspond to critical points of Jλ,µ : H → R.
Clearly, the critical points of Iλ,µ in H
The following lemma is due to [5], which reflect the relationship between the spaces H0s (Ω) and
s (C ) .
H0,L Ω

Lemma 2.1. i) kukHs (Ω) = Es (u) s (C )

H0,L
, for all u ∈ H0s (Ω).
0 Ω

ii)For any 1 ≤ r ≤ 2∗s and any w ∈ s

H0,L (CΩ ), it holds
Z ! 2r Z
r
|u(x)| dx ≤ C y1−2s |∇w(x, y)|2 dxdy, u := Tr(w), (2.5)
Ω CΩ

for some positive constant C = C(r, s, N, Ω). Furthermore, the space H0,L
s (C ) is compactly embedded
Ω
into Lr (Ω), for every r < 2∗s .

Remark. When r = 2∗s , the best constant in (2.5) is denoted by S(s, N ), that is
R
CΩ
y1−2s |∇w(x, y)|2 dxdy
S(s, N ) := inf
s (C )\{0}  2∗ . (2.6)
w∈H0,L
R
Ω 2∗ 2s
Ω
|w(x, 0)| s dx
 
It is not achieved in any bounded domain and, for all w ∈ Hs RN
+
+1
,
Z Z ! N−2s
N
2N
1−2s 2
y |∇w(x, y)| dxdy ≥ S(s, N ) |w(x, 0)| N−2s dx . (2.7)
RN
+
+1
RN
6 Int. J. Anal. Appl. (2024), 22:107

However, S(s, N ) is indeed achieved for the case Ω = RN when wε (x, y) = E(uε ), where uε takes
the form
ε(N−2s)/2
uε (x) := , ε > 0, x ∈ RN . (2.8)
(ε2 + |x|2 )(N−2s)/2
  2s−N
2
Let U (x) = 1 + |x|2 and let W be the extension of U [3, 13]. Then
Z
2s U (z)dz
W (x, y) = Es (U ) = CN,s y N +2s
,
RN (|x − z|2 + y2 ) 2

is the extreme function for the fractional Sobolev inequality (2.7). The constant S(s, N ) given in
(2.6) takes the exact value
     2sN
2πs Γ N +2s
Γ(1 − s) Γ N2
2
S(s, N ) =   .
Γ(s)Γ N−2s
2 (Γ(N ))s

Now, we consider the following minimization problem

R  
CΩ
y1−2s |∇w1 |2 + |∇w2 |2 dxdy
Sη,α,β := inf 2/2∗s . (2.9)
˜
R
(w1 ,w2 )∈H\{0} ( η |w | α1 |w |β1 + η |w |α2 |w |β2 ) dx
Ω×{0} 1 1 2 2 1 2

We define
1 + τ2
f (τ) : = 2
, τ > 0. (2.10)
( η1 τβ1 + η2 τβ2 ) 2s
∗

Since f is continuous on (0, ∞) such that lim f (τ) = lim f (τ) = +∞, then there exists τ0 > 0
τ→0+ τ→+∞
such that
f (τ0 ) := min f (τ) > 0. (2.11)
τ>0

Using ideas from [1], we establish a relationship between S(s, N ) and Sη,α,β .

Lemma 2.2. For the constants S(s, N ) and Sη,α,β introduced in (2.6) and (2.9), it holds

Sη,α,β = f (τ0 ) S(s, N ). (2.12)

In particular, the constant Sη,α,β is achieved for Ω = RN .

s (C ) be a minimization sequence for S(s, N ). Consider the sequences

Proof. Let {zn } ⊂ H0,L Ω
w1,n := zn and w2,n := τ0 zn in H0,L
s (C ). By (2.9), we have
Ω
R
1 + τ20 CΩ
y1−2s |∇zn (x, y)|2 dxdy
 β 2  2∗ ≥ Sη,α,β . (2.13)
β2 2∗s
R ∗
η1 τ0 + η2 τ0
1 2
|z (x, 0)| s dx
Ω n
2s

Letting n → +∞ yields
f (τ0 ) S(s, N ) ≥ Sη,α,β . (2.14)
Int. J. Anal. Appl. (2024), 22:107 7

On the other hand, let (w1,n , w2,n ) ⊂ H\{0} ˜ be a minimizing sequence for Sη,α,β . Set hn := sn w2,n
∗ ∗
R R
for sn > 0 with Ω×{0} |w1,n | s dx = Ω×{0} |hn | s dx. Then Young’s inequality yields
2 2

α β
Z Z Z
α β 2∗s ∗
|w1,n | |hn | dx ≤ ∗ |w1,n | dx + ∗ |hn |2s dx
Ω×{0} 2s Ω×{0} 2s Ω×{0}
Z Z
∗ ∗
= |hn |2s dx = |w1,n |2s dx.
Ω×{0} Ω×{0}

In turn, we can estimate

R  
y 1−2s |∇w (x, y)|2 + |∇w (x, y)|2 dxdy
CΩ 1,n 2,n
R 2/2∗s
( η |w | α1 |w |β1 + η |w |α2 |w |β2 ) dx
Ω×{0} 1 1,n 2,n 2 1,n 2,n
R
CΩ
y1−2s |∇w1,n (x, y)|2 dxdy
≥   2∗
−β −β
R
η1 sn 1 + η2 sn 2 Ω×{0} |w1,n |2s 2s
∗

R
s−2
n CΩ y
1−2s |∇h (x, y)|2 dxdy
n
+   2∗
−β −β
R
η1 sn 1 + η2 sn 2 Ω×{0} |hn |2s dx 2s
∗

 
≥ f s−1n S(s, N )

≥ f (τ0 ) S(s, N ).

Passing to the limit in the last inequality we obtain

f (τ0 ) S(s, N ) ≤ Sη,α,β .

Which together with (2.14) implies that

Sη,α,β = f (τ0 ) S(s, N ).


o
Let R0 > 0 be a constant such that Ω ⊂ B (0, R0 ), where B (0, R0 ) = {x ∈ RN : |x| < R0 . By
qN
Hölder’s inequality and (2.6), for all (w1 , w1 ) ∈ H̃, 1 < q < 2 and 0 ≤ γ < N + sq − 2 , we get

q ! q∗ ∗  2∗s2−q
∗
 2∗2−q
Z
s
Z
w1
Z 2∗ 2s

1 s  s
q· qs

dx ≤ |w1 | dx  dx

Ω×{0} |x|γ Ω×{0}

Ω |x|
γ 

∗  2∗s2−q
∗
 2∗2−q
Z
s
q  
1 s  s
−2 q
≤ (ks S(s, N )) kw1 kHs (C )  dx
 
γ
0,L Ω 
B(0,R0 ) |x|  (2.15)
  2∗s −q
2∗s
Z R0 N−1 
q
−2 q r
≤ (ks S(s, N )) kw1 kHs (C )  dr
 
Ω 2∗s γ
0,L  0 ∗ 
|r| 2s −q
q
= Θkw1 kHs (CΩ )
0,L
8 Int. J. Anal. Appl. (2024), 22:107

and
q
w2
Z
q
dx ≤ Θkw2 kHs , (2.16)
Ω×{0} |x|γ 0,L
(CΩ )

where
  2∗s −q
 2N − qN + 2sq  2∗s N−γ− qN +sq q
−2 (2.17)
Θ :=  R 2
( k S ( s, N )) .
 
0 s
 2N N − γ − qN + sq 
  
2

Now we are looking for the solutions of problem (1.1). Equivalently, we consider the solutions
˜ it is useful to consider the
of problem (2.3). Since the energy functional Iλ,µ is not bounded on H,
functional on the Nehari manifold
n D E o
˜
Nλ,µ := z ∈ H\{0} : I0λ,µ (z), z = 0 . (2.18)

Thus, z = (w1 , w2 ) ∈ Nλ,µ if and only if

D E
I0λ,µ (z), z = kzk2˜ − K(z) − Q(z) = 0, (2.19)
H

where
Z
η1 |w1 |α1 |w2 |β1 + η2 |w1 |α2 |w2 |β2 dx
 
Q(z) : =
Ω×{0}

and
|w1 |q |w2 |q
Z  
K (z) : = λ + µ dx.
Ω×{0} |x|γ |x|γ
 
Define Φ(z) = I0λ,µ (z), z , then for all z = (w1 , w2 ) ∈ Nλ,µ , we have

Φ0 (z), z = 2kzk2˜ − 2∗s Q(z) − qK(z)

H
= (2 − q)kzk2H˜ − (2∗s − q) Q(z) (2.20)
= (2 − 2∗s ) kzk2H˜ + (2∗s − q) K(z).

Thus, it is natural to split Nλ,µ into three parts corresponding to local minima, local maxima and
points of inflection, i.e.
n o
+
Nλ,µ = z ∈ Nλ,µ : Φ0 (z), z > 0 ,
n o
−
Nλ,µ = z ∈ Nλ,µ : Φ0 (z), z < 0 , (2.21)
n o
0
Nλ,µ = z ∈ Nλ,µ : Φ0 (z), z = 0 .

It is clear that all critical points of Iλ,µ must, lie on Nλ,µ and, as we will see below, local minimizers
on Nλ,µ are actually critical points of Iλ,µ . We have the following results.

Lemma 2.3. The energy functional Iλ,µ is bounded below and coercive on Nλ,µ .
Int. J. Anal. Appl. (2024), 22:107 9

Proof. Let z = (w1 , w2 ) ∈ Nλ,µ . From (2.15), (2.16) and (2.19) by the Hölder inequality, we get
! !
1 1 2 1 1
Iλ,µ (z) = − kzk ˜ − − K (z)
2 2∗s H q 2∗s
!
s 1 1

q q
≥ kzk ˜ −2
− ∗ λkw1 kHs (C ) + µkw2 kHs (C ) Θ (2.22)
N H q 2s 0,L Ω 0,L Ω

!  2−q
s 1 1 2 2 2 q
≥ kzk ˜ −2
− ∗ λ +µ2−q 2−q kzk ˜ Θ,
N H q 2s H

where Θ is given by (2.17). Since 1 < q < 2, the functional Iλ,µ is coercive and bounded below on
Nλ,µ . 

0
Lemma 2.4. (Natural Constraint). Suppose that z0 is a local minimizer of Iλ,µ on Nλ,µ and that z0 < Nλ,µ ,
˜ −1 .
then I0 (z0 ) = 0 in H
λ,µ

Proof. Suppose that z0 = (w0,1 , w0,2 ) is a local minimizer of Iλ,µ on Nλ,µ , then Iλ,µ (z0 ) =
min Iλ,µ (z) and (2.20) holds. Furthermore, by the theory of Lagrange multipliers, there exists
z∈Nλ,µ
θ ∈ R such that I0λ,µ (z0 ) = θΦ0 (z0 ). As z0 ∈ Nλ,µ , we get
D E
0 = I0λ,µ (z0 ) , z0 = θ Φ0 (z0 ) , z0 .

0
Since z0 < Nλ,µ ˜ −1 .
, Φ0 (z0 ) , z0 , 0. Consequently, θ = 0 and I0λ,µ (z0 ) = 0 in H 

Let Λ1 be the positive number defined in (1.4). Then we have the following result.

Lemma 2.5. Assume that (λ, µ) ∈ CΛ1 . Then Nλ,µ

0
= ∅.
2 2
Proof. Assume by contradiction that there exist λ > 0 and µ > 0 with 0 < λ 2−q + µ 2−q < Λ1 and
0 0
such that Nλ,µ , ∅. Let z ∈ Nλ,µ . Then, by virtue of (2.20), we get

2∗s − q 2∗s − q
kzk2˜ = Q(z), kzk2˜ = K (z).
H 2−q H 2∗s − 2
By Hölder inequality and the Sobolev embedding theorem, we have
# 1
 22∗s 2∗s −2
"
2−q 
kzkH˜ ≥ ks Sη,α,β ,
(2∗s − q)
! 1  12
2∗s − q 2−q  2−q 2 2
kzkH˜ ≤ ∗ Θ λ + µ 2−q ,
2s − 2
which leads to the inequality
2 "
!− 2−q ∗ # 2∗2−2
2 2 2∗s − q 2−q   22s s
λ 2−q +µ 2−q ≥ ∗ Θ k S
s η,α,β = Λ1 ,
2s − 2 (2∗s − q)
contradicting the assumption. 
10 Int. J. Anal. Appl. (2024), 22:107
2 2
From Lemma 2.5, if 0 < λ 2−q + µ 2−q < Λ1 , we can write Nλ,µ = Nλ,µ
+ − and define
∪ Nλ,µ

αλ,µ := inf Iλ,µ (z), αλ,µ

+
:= inf Iλ,µ (z), α−λ,µ := inf− Iλ,µ (z).
z∈Nλ,µ + z∈Nλ,µ
z∈Nλ,µ

Moreover, we have the following properties about the Nehari manifold Nλ,µ .

Theorem 2.1. The following facts holds

(i) If (λ, µ) ∈ CΛ1 , then we have αλ,µ ≤ αλ,µ
+
< 0;
(ii) If (λ, µ) ∈ C(q/2)2/(2−q) Λ1 , then we have α−λ,µ > c0 for some positive constant c0 depending on λ, µ, N, q, s
and Θ.

+
Proof. (i) Let z = (w1 , w2 ) ∈ Nλ,µ . By (2.18), (2.20) and (2.21), it follows that

2−q
kzk2 > Q(z). (2.23)
2∗s − q H˜
According to (2.18) and (2.23), we have that
! !
1 1 1 1
Iλ,µ (z) = − kzk2˜ + − Q(z)
2 q H q 2∗s
" ! ! #
1 1 1 1 2−q
< − + − kzk2˜
2 q q 2∗s 2∗s − q H

(2 − q)s 2
=− kzk ˜ < 0.
qN H

Therefore, by the definition of αλ,µ , αλ,µ+

, we can deduce that αλ,µ ≤ αλ,µ
+
< 0.
2
!
2 2  q  2−q
(ii) Suppose that λ 2−q + µ 2−q ∈ 0, 2 Λ1 and z = (w1 , w2 ) ∈ Nλ,µ
− . By (2.9), (2.20) and (2.21), one

has
∗
2−q  − 22s 2∗
∗ kzk ˜ < Q(z) ≤ ks Sη,α,β
2
kzk ˜s ,
2s − q H H

which implies that

! 1
2−q 2∗s −2   4sN
kzkH˜ > ∗ ks Sη,α,β . (2.24)
2s − q
From the last inequality we infer that
2∗s − 2 2 2∗s − q
Iλ,µ (z) = ∗ kzk ˜ − K (z)
22s H q2∗s

2∗s − q  2−q
!  2−q 
q   s 2−q 2 2 2
λ + µ 2−q Θ

≥ kzk ˜  kzk ˜ − ∗
H N H 2s q
! ∗q 
 2∗ − 2  ! 2−q
2−q 2s −2   qN s
 (2−q4s)N 2−q 2∗s −2
>
4s 
k S
s η,α, β

 22∗ k S
s η,α,β
2 (2∗s − q) 2 (2∗s − q)

s

2∗s − q  2−q 2 2
 2−q
2

λ + µ 2−q Θ .

− ∗
q2s
Int. J. Anal. Appl. (2024), 22:107 11
2 2 2
Thus, if λ 2−q + µ 2−q < (q/2) 2−q Λ1 , then

Iλ,µ (z) > c0 , −

for all z ∈ Nλ,µ ,

for some positive constant c0 = c0 (λ, µ, q, N, s, Θ). 

˜ with Q(z) > 0 then there exist

Theorem 2.2. Let (λ, µ) ∈ CΛ1 . Then, for every z = (w1 , w2 ) ∈ H
(unique) t− = t− (z) > 0 and t+ = t+ (z) > 0 such that
+
t+ z ∈ Nλ,µ , t− z ∈ Nλ,µ
−
.

In particular, we have
 ∗1
 (2 − q)kzk2H˜  2s −2

t+ < tmax < t− , tmax :=  ∗ 
(2 − q) Q(z) 

s

as well as
 
Iλ,µ t+ z = min Iλ,µ (tz), Iλ,µ (t− z) = max Iλ,µ (tz).
0<t<tmax t>0

˜ such that Q(z) > 0, and for all t ≥ 0, we have

Proof. For each z ∈ H
∗
D E
I0λ,µ (tz), tz = t2 kzk2˜ − t2s Q(z) − tq K(z).
H

We define g, h : R+ → R by
∗
g(t) := t2−q kzk2˜ − t2s −q Q(z) − K(z),
H
∗
h(t) : = 2−q 2
t kzk ˜ − t2s −q Q(z).
H

Clearly, we obtain h(0) = 0, and h(t) → −∞ as t → ∞. Because

∗
h i
h0 (t) = t1−q (2 − q)kzk2˜ − (2∗s − q) t2s −2 Q(z) , for all t > 0,
H

solving h0 (t) = 0, we obtain

 ∗1
 (2 − q)kzk2H˜  2s −2

t̄max =  ∗  > 0.
(2 − q) Q(z) 

s

Easy computations show that h0 (t) > 0 for all 0 < t < t̄max and h0 (t) < 0 for all t > t̄max . Thus h(t)
attains its maximum at t̄max , that is,
 2−q
 (2 − q)kzk2H˜  2s −2 2∗s − 2
 ∗

h (t̄max ) =  ∗
  kzk2 .
2∗s − q H˜

( 2s − q ) Q ( z ) 

Then from (2.17), (2.15) and (2.16), by the Holder inequality, one gets

g (t̄max ) = h (t̄max ) − K(w1 , w2 )

 2−q q (2.25)
 (2 − q)kzk2H˜  2s −2 2∗s − 2 q 
 ∗ 
 w1 w2 
Z
=  ∗
  2
kzk − λ γ + µ γ  dx

2∗s − q H˜

( 2s − q ) Q ( z )  Ω×{0} |x| |x|
12 Int. J. Anal. Appl. (2024), 22:107
  2−q
 2  2∗s −2
 ( 2 − q ) kzk ˜
 2∗s − 2 2
H
≥ 
 
kzk
 2∗s  2∗s − q H˜


 (2∗ − q) kzk2∗s ks Sη,α,β − 2 
 
s ˜
H
 
q q
− λkw1 kHs (C ) + µkw2 kHs (C ) Θ
0,L Ω 0,L Ω

2−q
 2∗s (2−q  2−q
)
2∗s − 2 q
!
2−q 2∗s −2
 2 2 2 q

≥ ∗ ks Sη,α,β 2(2∗s −2)
∗ kzk ˜ − λ + µ
2−q 2−q kzk ˜ Θ
2s − q 2s − q H H

> 0,
2 2
where Θ is as in (2.17) and the last inequality holds for every λ 2−q + µ 2−q ∈ (0, Λ1 ). It follows that
there exist t+ and t− such that
   
g t+ = g (t− ) and g0 t+ > 0 > g0 (t− ) ,

for 0 < t+ < t̄max < t− . We have t+ z ∈ Nλ,µ

+
, t− z ∈ Nλ,µ
− and

 
Iλ,µ (t− z) ≥ Iλ,µ (tz) ≥ Iλ,µ t+ z ,

for each t ∈ [t+ , t− ], and Iλ,µ (t+ z) ≤ Iλ,µ (tz) for each t ∈ [0, t+ ]. Thus
 
Iλ,µ t+ z = min Iλ,µ (tz), Iλ,µ (t− z) = max Iλ,µ (tz).
0≤t≤t̄max t≥t̄max

3. The Palais-Smale condition

In this section, we will find the range of c where the (PS)c condition holds for the functional
Iλ,µ .

Definition 3.1. Let c ∈ R and Iλ,µ ∈ C1 (H, ˜ R).

˜ for Iλ,µ if Iλ,µ (zn ) = c + o(1) and I0 (zn ) = o(1) strongly in H
(i) {zn } is a (PS)c -sequence in H ˜ −1 as
λ,µ
n → ∞.
(ii) We say that Iλ,µ satisfies the (PS)c condition if any (PS)c -sequence {zn } for Iλ,µ has a convergent
˜
subsequence in H.

We shall need the following preliminary result.

˜ is a (PS)c -sequence for Iλ,µ with zn * z in H

Lemma 3.1. (Uniform Lower Bound). Let {zn } ⊂ H ˜ and
I0λ,µ (z) = 0 and there exists a positive constant C0 such that
 2 2

Iλ,µ (z) ≥ −C0 λ + µ
2−q 2−q , (3.1)

where
" 2
! # 2−q
2−q 2N − qN + 2sq
C0 = Θ .
2 4s
Int. J. Anal. Appl. (2024), 22:107 13

˜ and z = (w1 , w2 ) ∈ H.
Proof. Consider zn = (w1,n , w2,n ) ⊂ H ˜ If {zn } is a (PS)c -sequence for Iλ,µ with
˜ then w1,n * w1 and w2,n * w2 in Hs (CΩ ), as n → ∞. Then, by virtue of Sobolev
zn * z in H, 0,L
embedding theorem (Lemma 2.1), we also have w1,n (·, 0) → w1 (·, 0) and w2,n (·, 0) → w2 (·, 0)
strongly in Lq (Ω), as n → ∞. Of course, up to a further subsequence, w1,n (·, 0) → w1 (·, 0) and
w2,n (·, 0) → w2 (·, 0) a.e. in Ω. It is standard to check that I0λ,µ (z) = 0 . This implies that
 
I0λ,µ (z), z = 0, namely
Z  
ks y1−2s |∇w1 |2 + |∇w2 |2 dxdy = K(w1 , w2 ) + Q(w1 , w2 ). (3.2)
CΩ

Consequently, we get
! Z
1 1  
Iλ,µ (z) = − ∗ ks y1−2s |∇w1 |2 + |∇w2 |2 dxdy
2 2s CΩ

|w1 |q |w2 |q
!Z
1 1
 
− − λ γ + µ γ dx.
q 2∗s Ω×{0} |x| |x|
Combining (2.15), (2.16) and the Young inequality, we have
 
q q
K(z) ≤ λ kw1 kHs (C ) + µ kw2 kHs (C ) Θ
0,L Ω 0,L Ω
q
!−1 − 2q
   
! −1 2

 2 s 1 1   2 s 1 1
   
q
λΘ
 
=  − ∗  kw1 kHs (C )   −
 
2∗s 

 q N q 2s Ω  q N q

0,L


!−1  2q !−1 − 2q
   
 2 s 1 1   2 s 1 1
   
q
− ∗  µΘ (3.3)
 
+  − ∗  kw2 kHs (C )  

 q N q 2s 0,L Ω  q N q 2s 
!−1 
s 1 1
  2 2

≤ − kw1 k2Hs 2
(CΩ ) + kw2 kHs
b λ 2−q + µ 2−q
+C
N q 2∗s 0,L 0,L
(CΩ )
!−1
s 1 1
 2 2

= − b λ 2−q + µ 2−q ,
kzk2˜ + C
N q 2∗s H

with
 2 2
!−1 − 2q  2−q
  ! 2q  2−q
2 − q  2 s 1 1
  2 − q  2N − qN + 2sq 
 Θ = Θ .

b=
C  −
   
2  q N q 2∗s 

2  4s



We obtain ! !
1 1 2 1 1
Iλ , µ ( z ) = − kzk ˜ − − K (z)
2 2∗s H q 2∗s
s s
 2 2

≥ kzk2˜ − kzk2˜ − C0 λ 2−q + µ 2−q
N H N H
 2 2

= −C0 λ 2−q + µ 2−q .
 
Then (3.1) follows from (3.3) with C0 = 1 − 1∗ C. b
q 2s 

.
14 Int. J. Anal. Appl. (2024), 22:107

˜ is a (PS)c -sequence for Iλ,µ , then {zn } is bounded in H.

Lemma 3.2. If {zn } ⊂ H ˜

˜ be a (PS)c -sequence for Iλ,µ and suppose, by contradiction, that

Proof. Let zn = (w1,n , w2,n ) ⊂ H
kzkH˜ → ∞, as n → ∞. Put
!
zn w1,n w2,n
zn = (w
e e2,n ) :=
e1,n , w = , .
kzkH˜ kzkH˜ kzkH˜

zn * e
We may assume thate z = (w
e1 , w ˜ This implies that w
e2 ) in H. e1,n (·, 0) → w
e1 (·, 0) and w
e2,n (·, 0) →
e2 (·, 0) strongly in Lr (Ω) for all 1 ≤ r < 2∗s and, thus,
w

w1,n |q µ|e
λ|e w2,n |q ) w1 |q µ|e
λ|e w2 |q )
Z Z
+ dx = + dx + on (1).
Ω×{0} |x|γ |x|γ Ω×{0} |x|
γ |x|γ

Since {zn } is a (PS)c sequence for Iλ,µ and kzkH˜ → ∞, there following hold:
q−2
ks
Z   kzk ˜ Z
w1,n |q µ|e
λ|e w2,n |q )
1−2s 2 2 H
y w1,n | + |∇e
|∇e w2,n | dxdy − + dx
2 CΩ q Ω×{0} |x|γ |x|γ
2∗ −2
(3.4)
kzk ˜s Z
w1,n |α1 |e
w2,n |β1 + η2 |e
w1,n |α2 |e
w2,n |β2 dx = on (1),
H
− η1 |e
2∗s Ω×{0}

and
w1,n |q µ|e
λ|e w2,n |q )
Z Z
q−2
 
1−2s 2 2
ks y w1,n | + |∇e
|∇e w2,n | dxdy − kzk ˜ + dx
CΩ H Ω×{0} |x|γ |x|γ
Z (3.5)
2∗s −2 α1 β1 α2 β2
 
− kzk ˜ η1 |e w2,n | + η2 |e
w1,n | |e w2,n |
w1,n | |e dx = on (1).
H Ω×{0}

Combining (3.4) and (3.5), as n → ∞, we obtain

Z  
ks y1−2s |∇e
w1,n |2 + |∇e
w2,n |2 dxdy
CΩ
(3.6)
2 (2∗s − q) q−2 w1,n |q µ|e
λ|e w2,n |q )
Z
= kzk + dx + on (1).
q (2∗s − 2) H˜ Ω×{0} |x|γ |x|γ

In view of 1 < q < 2 and kzkH˜ → ∞, (3.6) implies that

Z  
ks y1−2s |∇e w1,n |2 + |∇e
w2,n |2 dxdy → 0,
CΩ

zn = 1 for any n ≥ 1.
as n → ∞, which contradicts to the fact that e 

qN
  2sN
Lemma 3.3. Suppose that (F ) holds and 0 ≤ γ < N + sq − 2 , for all −∞ < c < c∞ = Ns Sη,α,β −
 2 2

˜ where C0 is given by Lemma 3.1.
C0 λ 2−q + µ 2−q , then Iλ,µ satisfies the (PS)c condition in H,

˜ be a (PS)c -sequence satisfying Iλ,µ (zn ) = c + o(1) and I0 (zn ) = o(1), where
Proof. Let {zn } ⊂ H λ,µ
( )
zn = w1,n , w2,n . By Lemma 3.2 , we see that {zn } is bounded in ˜
H. Passing to a subsequence (still
Int. J. Anal. Appl. (2024), 22:107 15

˜ such that
denoted by {zn } ), there exists z = (w1 , w2 ) ∈ H

w1,n * w1 , w2,n * w2 ,
s (C ) ,
weakly in H0,L Ω





w1,n (·, 0) → w1 (·, 0), w2,n (·, 0) → w2 (·, 0), strongly in Lr (Ω) (1 ≤ r < 2∗s ) , (3.7)




w1,n (·, 0) → w1 (·, 0), w2,n (·, 0) → w2 (·, 0), a.e. in Ω.



Hence, we have
λ|w1,n |q µ|w2,n |q ) λ|w1 |q µ|w2 |q )
Z Z
+ dx = + dx + on (1).
Ω×{0} |x|γ |x|γ Ω×{0} |x|γ |x|γ
b1,n := w1,n − w1 , w
Set w zn := (w
b2,n := w2,n − w2 and b b2,n ). From Brézis-Lieb’s lemma [6], it
b1,n , w
follows that
2
zn
b H˜ = kzk2H˜ − kzk2H˜ + on (1), (3.8)

and by Lemma (2.1) in [11] one has

Z Z Z
αi βi αi βi
w1,n | |b
|b w2,n | dx = |w1,n | |w2,n | dx − |w1 |αi |w2 |βi dx + o(1), i = 1, 2. (3.9)
Ω×{0} Ω×{0} Ω×{0}

Since Iλ,µ (zn ) = c + o(1) and I0λ,µ (zn ) = o(1), and by (3.7) to (3.9), we can deduce that
Z
1 2 1
w1,n |α1 |b
w2,n |β1 + η2 |b
w1,n |α2 |b
w2,n |β2 dx = c − Iλ,µ (z) + on (1),
 
zn H˜ − ∗
b η1 |b (3.10)
2 2s Ω×{0}
and Z
2
w1,n |α1 |b
w2,n |β1 + η2 |b
w1,n |α2 |b
w2,n |β2 dx
 
zn H˜
b − η1 |b
Ω×{0} (3.11)
D E D E
= I0λ, µ (zn ) , zn − I0λ,µ (z), z + on (1) = on (1).
Now, we can assume that
2
zn
lim b = lim Q (b
zn ) = l. (3.12)
n→∞ n→∞
If l = 0, the proof is complete. Assume l > 0, then it follows from (3.12) and the definition of Sη,α,β
that
2 2
∗
zn
b H˜ ≥ Sη,α,β Q 2s (b
zn ) ,
2 2 2
ks Sη,α,β ` 2s = ks Sη,α,β lim Q 2s (b = `,
∗ ∗
zn ) ≤ lim b
zn H˜
n→∞ n→∞
which implies that
  2sN
` ≥ ks Sη,α,β . (3.13)

In addition, from (3.12) to (3.13) and Lemma 3.1, we have

!
1 1 s   2sN  2 2

c= − ` + Iλ,µ (z) ≥ ks Sη,α,β − C0 λ + µ
2−q 2−q = c∞ ,
2 2∗s N
˜
which contradicts the definition of c. Therefore, l = 0 and (w1,n , w2,n ) → (w1 , w2 ) strongly in H.
The proof is complete. 
16 Int. J. Anal. Appl. (2024), 22:107

4. Existence of Palais-Smale sequences

Lemma 4.1. Let (λ, µ) ∈ CΛ1 . Then, for any z ∈ Nλ,µ , there exists r > 0 and a differentiable map
˜ → R+ such that ξ(0) = 1 and ξ(h)(z − h) ∈ Nλ,µ for every h ∈ B(0; r). Let us set
ξ : B(0; r) ⊂ H
Z
D1 := 2ks y1−2s (∇w1 · ∇h1 + ∇w2 · ∇h2 ) dxdy
Ω
λ|w1 |q−2 w1 h1 µ|w2 |q−2 w2 h2
Z !
D2 : = q + dx
Ω×{0} |x|γ |x|γ
Z
α1 η1 |w1 |α1 −2 w1 h1 |w2 |β1 + β1 η1 |w1 |α1 |w2 |β1 −2 w2 h2 dx
 
D3 :=
Ω×{0}
Z
α2 η2 |w1 |α2 −2 w1 h1 |w2 |β2 + β2 η2 |w1 |α2 |w2 |β2 −2 w2 h2 dx,
 
+
Ω×{0}

˜ and (w1 , w2 ) ∈ H.
for all (h1 , h2 ) ∈ H ˜ Then

D3 + D2 − D1
ξ0 (0), h = (4.1)
(2 − q)kzk2˜ − (2∗s − q) Q(w1 , w2 )
H

˜
for all (h1 , h2 ) ∈ H.

Proof. The proof is almost the same as in [17]. For z = (w1 , w2 ) ∈ Nλ,µ , define a function Fz :
R×H ˜ → R by
D E
Fz (ξ, w) = I0λ,µ (ξ(z − p)), ξ(z − p)
∗
=ξ2 kz − pk2H˜ − ξ2s Q(z − p) − ξq K(z − p).
 
0
Then Fz (1, 0) = Iλ,µ (z), z = 0 and, by Lemma 2.5 , we have

d
Fz (1, 0) = Φ0 (z), z
dξ
= (2 − 2∗s ) kzk2H˜ + (2∗s − q) Q(z) , 0.

According to the implicit function theorem, there exist η > 0 and a differentiable function ξ :
˜ → R such that ξ(0) = 1 and formula (4.1) holds, via direct computation. Moreover,
B(0; η) ⊂ H

Fz ( ξ ( h ) , h ) = 0 for all h ∈ B(0; η),

which is equivalent to
D E
I0λ,µ (ξ(h)(z − h)), ξ(h)(z − h) = 0 for all h ∈ B(0; η),

that is ξ(h)(z − h) ∈ Nλ,µ . 

Lemma 4.2. Let (λ, µ) ∈ CΛ1 . Then, for any z ∈ Nλ,µ − , there exists r > 0 and a differentiable map

˜ → R+ such that ξ− (0) = 1 and ξ− (h)(z − h) ∈ N − for every h ∈ B(0; r).

ξ− : B(0; r) ⊂ H λ,µ
Int. J. Anal. Appl. (2024), 22:107 17

Proof. Arguing as for the proof of Lemma 4.1, there exists r > 0 and a differentiable function
˜ → R+ such that ξ− (0) = 1, ξ− (h)(z − h) ∈ Nλ,µ for all h ∈ B(0; r) and formula (4.1)
ξ− : B(0; r) ⊂ H
holds. Since
Φ0 (z), z = (2 − q)kzk2˜ − (2∗s − q) Q(z) < 0,
H

by the continuity of the functions Φ0 and ξ− , up to reducing the size of r > 0, we get

Φ0 (ξ− (h)(z − h)) , ξ− (h)(z − h) < 0.

This implies that the functions ξ− (h)(z − h) belong to Nλ,µ

− . 

Proposition 4.1. The following facts hold.

(i) Let (λ, µ) ∈ CΛ1 . Then there is a (PS)αλ,µ -sequence {zn } ⊂ Nλ,µ for Iλ,µ .
(ii) Let (λ, µ) ∈ C(q/2)2/(2−q) Λ1 . Then there is a (PS)α− -sequence {zn } ⊂ Nλ,µ for Iλ,µ .
λ,µ

Proof. (i) By Lemma 2.3 and Ekeland Variational Principle [10], there exists a minimizing sequence
{zn } ⊂ Nλ,µ such that
1
Iλ,µ (zn ) < αλ,µ + ,
n

1
Iλ,µ (zn ) < Iλ,µ (w) + kw − zn kH˜ , for each w ∈ Nλ,µ . (4.2)
n
Taking n large and using αλ,µ < 0, we have

|w1,n |q |w2,n |q
! !Z !
1 1 1 1
Iλ,µ (zn ) = − 2
kzk ˜ − − λ +µ dx
2 2∗s H q 2∗s Ω×{0} |x|γ |x|γ
(4.3)
1 αλ,µ
< αλ,µ + < .
n 2
This yiclds that
q2∗s |w1,n |q |w2,n |q
Z !
− αλ,µ < λ +µ dx
2 (2∗s − q) Ω×{0} |x|γ |x|γ
(4.4)
 2 2
 2−q
2 q
≤ λ 2−q + µ 2−q kzk ˜ Θ.
H

Consequently, zn , 0 and combining with (4.3) and (4.4) and using Hölder inequality
  q−2  1q
q2∗s  2 2 2 
> − αλ,µ λ 2−q + µ 2−q

kzkH˜  ,

2 (2∗ − q) Θ
s

and
1

 2 (2∗s − q)  2 2
 2−q
2 
 2−q
kzkH˜ <  Θ λ 2−q + µ 2−q  . (4.5)

q (2∗s − 2) 

Now we prove that

I0λ,µ (zn ) ˜ −1
→ 0, as n → ∞.
H
18 Int. J. Anal. Appl. (2024), 22:107

Fix n ∈ N. By applying Lemma 4.2 to zn , we obtain the function ξn : B (0; rn ) → R+ for some rn > 0,
˜ with w , 0 and put h∗ = σw . We set
such that ξn (h) (zn − h) ∈ Nλ,µ . Take 0 < σ < rn . Let w ∈ H kwkH˜
hρ = ξn (h∗ ) (zn − h∗ ), then hσ ∈ Nλ,µ , and we have from (4.2) that
1
Iλ,µ (hσ ) − Iλ,µ (zn ) ≥ − khσ − zn kH˜ .
n
By the Mean Value Theorem, we get
D E   1
I0λ,µ (zn ) , hσ − zn + o khσ − zn kH˜ ≥ − khσ − zn kH˜ .
n
Thus, we have
D E D E 1  
I0λ,µ (zn ) , −h∗ + (ξn (h∗ ) − 1) I0λ,µ (zn ) , zn − h∗ ≥ − khσ − zn kH˜ + o khσ − zn kH˜ .
n
Whence, from ξn (h∗ ) (zn − h∗ ) ∈ Nλ,µ , it follows that
* +
w D E
0
−σ Iλ,µ (zn ) , + (ξn (h∗ ) − 1) I0λ,µ (zn ) − I0λ,µ (hσ ) , zn − h∗
kwkH˜
1  
≥ − khσ − zn kH˜ + o khσ − zn kH˜ .
n
So, we get
 
o khσ − zn kH˜
* +
w 1
I0λ,µ (zn ) , ≤ khσ − zn kH˜ +
kwkH˜ nσ σ (4.6)
(ξn (h ) − 1) D 0
∗ E
+ Iλ,µ (zn ) − I0λ,µ (hσ ) , zn − h∗ .
σ
Since khσ − zn kH˜ ≤ σ|ξn (h ) | + |ξn (h ) − 1|kzkH˜ and
∗ ∗

|ξn (h∗ ) − 1|
lim ≤ ξ0n (0) ˜ .
σ→0 σ H

For fixed n ∈ N, if we let σ → 0 in (4.6), then by virtue of (4.5) we can choose a constant C > 0
independent of σ such that
* +
w C 
I0λ,µ (zn ) , ≤ 1 + ξ0n (0) ˜
H
.
kwkH˜ n

Thus, we are done once we prove that ξ0n (0) ˜

H
remains uniformly bounded. By (4.1), (4.5) and
Hölder inequality, we have
C1 khkH˜
| ξ0n (0), h | ≤
(2 − q)kzk2˜ − Q(w1,n , w2,n )
H
for some C1 > 0. We only need to prove that

|(2 − q)kzk2˜ − (2∗s − q) − Q(w1,n , w2,n ) |≥ C2 ,

H

for some C2 > 0 and n large enough. We argue by contradiction. Suppose that there exists a
subsequence {zn } such that

(2 − q)kzk2H˜ − 2 (2∗s − q) Q(w1,n , w2,n ) = on (1). (4.7)

Int. J. Anal. Appl. (2024), 22:107 19

By virtue of (4.7) and the fact that zn ∈ Nλ,µ , we have

(2∗s − q) 2∗s − q
kzk2˜ = Q(w1,n , w2,n ) + on (1), kzk2˜ = K (zn ) + on (1).
H 2−q H 2∗s − 2

Taking into account that Iλ,µ (zn ) → αλ,µ < 0 as n → ∞, we have kzkH˜ 9 0 as n → ∞. Then,
arguing as in the proof of Lemma 2.5 yields (λ, µ) < CΛ1 , a contradiction. Then,
* +
0 w C
Iλ,µ (zn ) , ≤ .
kwkH˜ n

This proves (i). By Lemma 4.2, one can prove (ii), but we shall omit the details here. 

+
Now, we establish the existence of a local minimizer for Iλ,µ in Nλ,µ .

Proposition 4.2. Let (λ, µ) ∈ CΛ1 . Then Iλ,µ has a local minimizer z+ in Nλ,µ
+
satisfying the following
conditions:
(i) Iλ,µ (z+ ) = αλ,µ = αλ,µ
+
< 0;
(ii) z+ is a positive solution of (2.3).

Proof. By (i) of Proposition 4.1, there exists a minimizing sequence {zn } = (w1,n , w2,n ) for Iλ,µ in
Nλ,µ such that, as n → ∞,

˜ −1 .
Iλ,µ (zn ) = αλ,µ + on (1) and I0λ,µ (zn ) = on (1) in H (4.8)

˜ Then there exists a

By Lemma 2.3, we see that Iλ,µ is coercive on Nλ,µ , and {zn } is bounded in H.
 
subsequence, still denoted by {zn } and z+ = w+ , w+ ∈ H ˜ such that, as n → ∞,
1 2

w1,n * w1+ , w2,n * w2+ , weakly in H0,L

s (C ) ,



 Ω

w1,n → w1 , w2,n → w2 , strongy in L (Ω) for all 1 ≤ r < 2∗s ,
+ + r




 w → w+ , w → w+ , a.e. in Ω,

1,n 1 2,n 2

up to subsequences. This implies that, as n → ∞,

 
K (zn ) = K z+ + on (1). (4.9)

We claim that z+ is a nontrivial solution of (2.3). From (4.8) and (4.9), it is easy to verify that z+ is a
weak solution of (2.3). From zn ∈ Nλ,µ and (2.3) we deduce that

q (2∗s − 2) 2 q2∗s
K (zn ) = kzk − Iλ,µ (zn ) . (4.10)
2 (2∗s − q) H˜ 2∗s − q

Let n → ∞ in (4.10), by (4.8), (4.9) and αλ,µ < 0, we have

  q2∗s
K z+ ≥ − ∗ αλ,µ > 0.
2s − q
20 Int. J. Anal. Appl. (2024), 22:107

˜ and
Therefore, z+ ∈ Nλ,µ is a nontrivial solution of (2.3). Now we show that zn → z+ strongly in H
Iλ,µ (z+ ) = αλ,µ . Since z+ ∈ Nλ,µ , then by (4.10), we obtain
 
αλ,µ ≤ Iλ,µ z+
s 2∗ − q  
= kz+ k2H˜ − s ∗ K z+
N q2s
2∗s − q
!
s
≤ lim kzk2˜ − K (zn )
n→∞ N H q2∗s
= lim Iλ,µ (zn ) = αλ,µ .
n→∞

This implies that Iλ,µ (z+ ) = αλ,µ and limn→∞ kzk2˜ = kz+ k2˜ . Set b
zn = zn − z+ . Then, that
H H
2
zn H˜
b = kzk2H˜ − kz+ k2H˜ + on (1).
˜ We claim that z+ ∈ N + . Assume by contradiction that z+ ∈ N − . Then, by
Hence, zn → z+ in H. λ,µ λ,µ
Theorem 2.2 , there exist (unique) t1+ and t−1 with t1+ z+ ∈ Nλ,µ
+
and t−1 z+ ∈ Nλ,µ
− In particular, we have

t1+ < t−1 = 1. Since

d   d2  
Iλ,µ tz+ |t=t+ = 0, and 2 Iλ,µ tz+ |t=t+ > 0,
dt 1 dt 1
 
there exists t1 < t∗ ≤ t−1 such that Iλ,µ t1 z+ < Iλ,µ (t∗ z+ ). By Theorem 2.2, we have
+ +

       
Iλ,µ t1+ z+ < Iλ,µ t∗ z+ ≤ Iλ,µ t−1 z+ = Iλ,µ z+ ,
   
a contradiction. Since Iλ,µ (z+ ) = Iλ,µ |w1+ |, |w2+ | and |w1+ |, |w2+ | ∈ Nλ,µ , by Lemma 2.4 we may
assume that z+ is a nontrivial nonnegative solution of (2.3). Then by the Strong Maximum Principle
[ [9], Lemma 2.4], we have w1+ , w2+ > 0 in CΩ , hence, z+ is a positive solution for (2.3). 

Next we will use wε = Es (uε ), the family of minimizers for the trace inequality (2.7), where uε
is given in (2.8). Without loss of generality, we may assume that 0 ∈ Ω. We then define the cut-off
function φ ∈ C∞
0
(CΩ ) , 0 ≤ φ ≤ 1 and for small fixed ρ > 0,

1, (x, y) ∈ Bρ,


φ(x, y) = 


0, (x, y) < B ,

2ρ
n o
where Bρ = (x, y) : |x|2 + y2 < ρ2 , y > 0 . We take ρ so small that B2ρ ⊂ CΩ . Recall W is the
extension of U introduced in Section 2, we have (cf. [3]) |∇W (x, y)| ≤ Cy−1 W (x, y). Let
1
Uε (x) = N−2s
, ε > 0.
(ε2 + |x|2 ) 2

Then the extension of Uε (x) has the form

Z
Uε (z)dz x y
 
Wε (x, y) = cN,s y 2s
= ε2s−N W , .
RN (|x − z|2 + y2 )
N +2s
2 ε ε

Notice that φWε ∈ H0,L

s (C ), for ε > 0 small enough.
Ω
Int. J. Anal. Appl. (2024), 22:107 21

˜
Lemma 4.3. There is z ∈ H\{0} nonnegative and Λ∗ > 0 such that for (λ, µ) ∈ CΛ∗ .

sup Iλ,µ (tz) < c∞ ,

t≥0

where c∞ is given in Lemma 3.3. In particular, α−λ,µ < c∞ for all (λ, µ) ∈ CΛ∗ .

Proof. By an argument similar to that of the proof of [ [3], formula (3.26)], we get
Z
2
φWε Hs (C ) = ks y1−2s |∇Wε |2 dxdy + O(1)
Ω
0,L RN
+
+1
Z (4.11)
=ε 2s−N
ks y 1−2s
|∇W (x, y)| dxdy + O(1). 2
RN
+
+1

We notice that
∗
φ(x)2s
Z Z
2∗s 2∗s
φUε 2∗s
= |φUε | dx = dx,
Ω Ω (ε2 + |x|2 )N
Z
2∗ 1 2∗
kUε k2s∗ = dx = ε−N kUk2s∗
s
RN (ε2 + |x|2 )N s

Then, one has that

∗
φ2s (x) − 1
Z Z
2∗s 2∗ dx
φUε 2∗
− ε−N kUk2s∗ = N
dx − ,
s s
Ω (ε2 + |x|2 ) RN \Ω (ε2 + |x|2 )N
which yields
Z Z
2∗s 2∗ 1 dx
| φUε 2∗
− ε−N kUk2s∗ | ≤ N
dx +
s s
Ω\B(0;ρ) (ε2 + |x|2 ) RN \Ω (ε2 + |x|2 )N
Z Z
dx dx
= ≤ = C3 .
RN \B(0;ρ) (ε2 + |x|2 )N RN \B(0;ρ) |x|2N

This implies that

−2∗ 2∗s −2∗ −2∗
1 − C3 εN kUk2∗ s ≤ εN φUε 2∗s
kUk2∗ s ≤ 1 + C3 εN kUk2∗ s
s s s

−2∗s
Taking ε so small that C3 εN kUk 2∗s
< 1, since 2/2∗s = (N − 2s)/N < 1, we obtain

−2∗
  ∗
−2∗ 2/2s 2
1 − εN C3 kUk2∗ s ≤ 1 − εN C3 kUk2∗ s ≤ εN−2s φUε 2∗s
kUk−2
2∗
s s s
  ∗
−2∗ 2/2s −2∗
≤ 1 + εN C3 kUk2∗ s ≤ 1 + εN C3 kUk2∗ s .
s s

2  
Hence φUε 2∗s
= ε2s−N kUk22∗ + O ε2s . Since W = Es (U ) optimizes (2.7), by (4.11) we have
s
Z
φWε
2 ε2s−N k s y1−2s |∇W (x, y)|2 dxdy + O(1)
s (C )
H0,L Ω RN
+
+1
(4.12)
=
φUε
2 ε2s−N kUk22∗ + O (ε2s )
2∗s s
22 Int. J. Anal. Appl. (2024), 22:107
Z
ks y1−2s |∇W (x, y)|2 dxdy
RN +1   
1 + O εN−2s
+
=
kUk22∗
s
 
= ks S(s, N ) + O εN−2s
.
˜ → R defined by
Now we consider the function J : H
1 2 1
J (z) : = kzk ˜ − ∗ Q(z).
2 H 2s
˜ Notice that J (0) = 0, J (tz0 ) > 0 for
Set w0,1 := φWε , w0,2 := τ0 φWε and z0 := (w0,1 , w0,2 ) ∈ H.
t > 0 small and J (tz0 ) < 0 for t > 0 large. The map t 7→ J (tz0 ) maximizes at
 ∗1
 kz0 k2H˜  2s −2

t0 :=   . (4.13)
Q(z0 ) 


Then from (2.12), (4.12) and (4.13), we conclude that

22∗s
2∗s −2
1 1
! kz0 k ˜
H
sup J (tz0 ) = J (t0 z0 ) = −
2 2∗s 2
t≥0 (Q(z0 )) 2∗s −2
 Z  ∗2∗s
   2s −2
 1 + τ2 ks 1−2s
φW 2

0
y |∇ ( ε ) | dxdy 
s 
 
CΩ
= 

2

N   Z !
2∗s

β1 β2 ∗

 η1 τ0 + η2 τ0 2

|φUε | s dx
 
Ω
 Z  2sN (4.14)
 f (τ0 )ks y1−2s |∇ (φWε ) |2 
 
s 

CΩ 
=  ! 2s2

N  Z 
∗

|φUε |2s dx
 
 
Ω
s h   i 2sN
= f (τ0 ) ks S(s, N ) + O εN−2s
N
s   2sN  
= ks Sη,α,β + O εN−2s .
N
We now choose δ1 > 0 so small that, for all (λ, µ) ∈ Cδ1 , we get
s N 2 2
c∞ = (ks Sη,α,β ) 2s − C0 (λ 2−q + µ 2−q ) > 0.
N
By the definition of Iλ,µ and z0 , we have

t2
Iλ,µ (tz0 ) ≤ kz0 k2H˜ , for all t ≥ 0 and λ, µ > 0,
2
which implies that there exists t0 ∈ (0, 1) satisfying

sup Iλ,µ (tz0 ) < c∞ , for all (λ, µ) ∈ Cδ1 .

t∈[0,t0 ]
Int. J. Anal. Appl. (2024), 22:107 23

Hence, from (4.14) we see that

tq
!
sup Iλ,µ (tz0 ) = sup J (tz0 ) − K (tz0 )
t≥t0 t≥t0 q
(4.15)
s   2sN   tq0  q
Z |Uε |q
≤ ks Sη,α,β + O ε N−2s
− λ + µτ0 γ
dx.
N q B(0;ρ) |x|

Letting 0 < ε ≤ ρ, we have

|Uε |q
Z Z
1
dx = dx
|x|γ (N−2s)q
B(0;ρ) B(0;ρ) |x|γ (ε2 + |x|2 ) 2

ρ
rN−1
Z
= dr
0
  2  (N−2s
2
)q

rγ ε(N−2s)q 1 + εr
ρ
rN−1
Z
ε
=ε N−γ+(2s−N )q
(N−2s)q
dr (4.16)
0 rγ (1 + r2 ) 2

1
rN−1
Z
=ε N−γ+2sq−qN
(N−2s)q
dr
0 rγ ( 1 + r2 ) 2

ρ
rN−1
Z
ε
+ εN−γ+2sq−qN (N−2s)q
dr.
1 rγ (1 + r2 ) 2

From (4.16), we get

 qN

q


C3 εN−γ+sq− 2 , γ > N − (N − 2s)q,
t0 |Uε |q
Z 

 qN
dx ≥ C4 ε 2 −sq | ln ε|, γ = N − (N − 2s)q, (4.17)

q B(0;ρ) |x|γ 

 qN
C5 ε 2 −sq , γ < N − (N − 2s)q,



where Ci > 0(i = 3, 4, 5) are positive constants (Ci independent of ε).

The case of γ > N − (N − 2s)q, combining (4.14) with (4.17), one has
s   2sN    q
 qN
sup Iλ,µ (tz0 ) ≤ ks Sη,α,β + O εN−2s + C2 εN−2s − C3 λ + µτ0 εN−γ+sq− 2 .
t≥0 N
 2 1
 N−2s
2 2 2
Let λ 2−q + µ 2−q = εN−2s , that is, ε = λ 2−q + µ 2−q , then we can choose δ1 > 0 such that
   q
 qN
O εN−2s + C2 εN−2s − C3 λ + µτ0 εN−γ+sq− 2
 2 2
N−2s !  2 2
  2 2
 2N−2γ+2sq−qN
 q 2(N−2s)
=O λ + µ
2−q 2−q + C2 λ + µ
2−q 2−q − C3 λ + µτ0 λ + µ
2−q 2−q

 2 2

< − C0 λ 2−q + µ 2−q ,
2 2
for all λ 2−q + µ 2−q ∈ (0, δ1 ). Then, for (λ, µ) ∈ Cδ1 , one gets

sup Iλ,µ (tz0 ) < c∞ .

t≥0
24 Int. J. Anal. Appl. (2024), 22:107

The case of γ = N − (N − 2s)q, it follows from (4.14) and (4.17) that

s   2sN    q
 qN
sup Iλ,µ (tz0 ) ≤ ks Sη,α,β + O εN−2s + C2 εN−2s − C4 λ + µτ0 ε 2 −sq | ln ε|.
t≥0 N
 1
 N−2s
2 2 2 2 2 2
Let λ 2−q +µ 2−q = εN−2s , that is, ε = λ 2−q +µ 2−q , choosing δ2 > 0 such that for all λ 2−q + µ 2−q ∈
(0, δ2 ), then one has
   q
 qN
O εN−2s + C2 εN−2s − C4 λ + µτ0 ε 2 −sq | ln ε|
 2 2
  2 2
  2 2
 2q  2 2

q

=O λ + µ
2−q 2−q + C2 λ + µ
2−q 2−q − C4 λ + µτ0 λ + µ2−q 2−q | ln λ + µ
2−q 2−q |
 2 2

< − C0 λ 2−q + µ 2−q .

Consequently, for (λ, µ) ∈ Cδ2 , we obtain

sup Iλ,µ (tz0 ) < c∞ .

t≥0
n o
If we set Λ∗ = min δ1 , ρN−2s , δ2 > 0, then for (λ, µ) ∈ CΛ∗ ,

sup Iλ,µ (tz0 ) < c∞ . (4.18)

t≥0

Finally, we prove that α−λ,µ < c∞ for all (λ, µ) ∈ CΛ∗ . Recall that

z0 = (w0,1 , w0,2 ) = (φWε , τ0 φWε ) .

Since Q(z0 ) > 0, by Theorem 2.2 there exists t0 > 0 such that t0 z0 ∈ Nλ,µ
− . By the definition of

α−λ,µ and (4.18), we conclude that

α−λ,µ ≤ Iλ,µ (t0 z0 ) ≤ sup Iλ,µ (tz0 ) < c∞

t≥0

for all (λ, µ) ∈ CΛ∗ . 

Let Λ∗ be as in Lemma 4.3. We prove the existence a local minimizer for Iλ,µ on Nλ,µ
− .

Proposition 4.3. Let Λ∗ > 0 be as in Lemma 4.3 and set

 q 2 
Λ2 := min Λ∗ , ( ) 2−q Λ1 .
2
For (λ, µ) ∈ CΛ2 , Iλ,µ has a minimizer z− in Nλ,µ
− with I
λ,µ (z ) = αλ,µ . Furthermore, z is a positive
− − −

solution of (2.3).

Proof. By (ii) of Proposition 4.1, there is a (PS)α− sequence {zn } ⊂ Nλ− µ for Iλ,µ for all (λ, µ) ∈
λ,µ l
C q
2 . By Lemmas 3.3 and 4.3 and (ii) of Theorem 2.1, for all (λ, µ) ∈ CΛ∗ , Iλ,µ satisfies the
( 2 ) 2−q Λ1
(PS) condition and α−λ,µ > 0. Then, there exists a subsequence still denoted by {zn } and z− =
α−λ,µ
 
w−1 (x, y), w−2 (x, y) ∈ Nλ,µ ˜ and Iλ,µ (z− ) = α− > 0, for all
− such that z → z− strongly in H
n λ,µ
(λ, µ) ∈ CΛ2 . Arguing as in the proof of Proposition 4.2, for (λ, µ) ∈ CΛ2 , we obtain that z− is a
positive solution of (2.3). 
Int. J. Anal. Appl. (2024), 22:107 25

5. The proof of Theorem 1.1

By Proposition 4.2, for (λ, µ) ∈ CΛ1 , system (2.3) admits a positive solution z+ ∈ Nλ,µ
+
. By
− exists for all (λ, µ) ∈ C . Furthermore, since
Proposition 4.3, a positive solution z− ∈ Nλ,µ Λ2
N + ∩ Nλ,µ − = ∅, then z+ , z− are distinct positive solutions of system (2.3). In turn, (u± (x), v± (x)) =
 λ,µ 
w±1 (x, 0), w±2 (x, 0) are distinct positive solutions of (1.1).

Conflicts of Interest: The authors declare that there are no conflicts of interest regarding the
publication of this paper.

References
[1] C.O. Alves, D.C. de Morais Filho, M.A.S. Souto, On systems of elliptic equations involving subcritical or critical
Sobolev exponents, Nonlinear Anal.: Theory Meth. Appl. 42 (2000), 771–787. https://doi.org/10.1016/s0362-546x(99)
00121-2.
[2] C.O. Alves, Y.H. Ding, Multiplicity of positive solutions to a p-Laplacian equation involving critical nonlinearity,
J. Math. Anal. Appl. 279 (2003), 508–521. https://doi.org/10.1016/s0022-247x(03)00026-x.
[3] B. Barrios, E. Colorado, A. de Pablo, U. Sánchez, On some critical problems for the fractional Laplacian operator,
J. Diff. Equ. 252 (2012), 6133–6162. https://doi.org/10.1016/j.jde.2012.02.023.
[4] S. Benmouloud, R. Echarghaoui, Si.M. Sbaï, Multiplicity of positive solutions for a critical quasilinear elliptic
system with concave and convex nonlinearities, J. Math. Anal. Appl. 396 (2012), 375–385. https://doi.org/10.1016/j.
jmaa.2012.05.078.
[5] C. Brändle, E. Colorado, A. de Pablo, U. Sánchez, A concave–convex elliptic problem involving the fractional
Laplacian, Proc. R. Soc. Edinburgh: Sect. A Math. 143 (2013), 39–71. https://doi.org/10.1017/s0308210511000175.
[6] H. Brézis, E. Lieb, A relation between pointwise convergence of functions and convergence of functionals, Proc.
Amer. Math. Soc. 88 (1983), 486–490.
[7] H. Brézis, L. Nirenberg, Positive solutions of nonlinear elliptic equations involving critical Sobolev exponents,
Comm. Pure Appl. Math. 36 (1983), 437–477.
[8] L. Caffarelli, L. Silvestre, An extension problem related to the fractional Laplacian, Comm. Part. Diff. Equ. 32 (2007),
1245–1260. https://doi.org/10.1080/03605300600987306.
[9] A. Capella, J. Dávila, L. Dupaigne, Y. Sire, Regularity of radial extremal solutions for some non-local semilinear
equations, Comm. Part. Diff. Equ. 36 (2011), 1353–1384. https://doi.org/10.1080/03605302.2011.562954.
[10] I. Ekeland, On the variational principle, J. Math. Anal. Appl. 47 (1974), 324–353. https://doi.org/10.1016/
0022-247x(74)90025-0.
[11] P.G. Han, The effect of the domain topology on the number of positive solutions of elliptic systems involving
critical Sobolev exponents, Houston J. Math. 32 (2006), 1241–1257.
[12] T.S. Hsu, Multiple positive solutions for a critical quasilinear elliptic system with concave–convex nonlinearities,
Nonlinear Anal.: Theory Meth. Appl. 71 (2009), 2688–2698. https://doi.org/10.1016/j.na.2009.01.110.
[13] T.S. Hsu, H.L. Lin, Multiple positive solutions for a critical elliptic system with concave—convex nonlinearities,
Proc. R. Soc. Edinburgh: Sect. A Math. 139 (2009), 1163–1177. https://doi.org/10.1017/s0308210508000875.
[14] R. Servadei, E. Valdinoci, On the spectrum of two different fractional operators, Proc. R. Soc. Edinburgh: Sect. A
Math. 144 (2014), 831–855. https://doi.org/10.1017/s0308210512001783.
[15] J.L. Vázquez, Recent progress in the theory of nonlinear diffusion with fractional Laplacian operators, preprint,
(2014). http://arxiv.org/abs/1401.3640.
[16] Y. Wei, X. Su, Multiplicity of solutions for non-local elliptic equations driven by the fractional Laplacian, Calc. Var.
Part. Differ. Equ. 52 (2014), 95–124. https://doi.org/10.1007/s00526-013-0706-5.
26 Int. J. Anal. Appl. (2024), 22:107

[17] T.F. Wu, On semilinear elliptic equations involving concave-convex nonlinearities and sign-changing weight func-
tion, J. Math. Anal. Appl. 318 (2006), 253–270. https://doi.org/10.1016/j.jmaa.2005.05.057.
[18] X. Zhou, H.Y. Li, J.F. Liao, Multiplicity of positive solutions for a semilinear elliptic system with strongly cou-
pled critical terms and concave nonlinearities, Qual. Theory Dyn. Syst. 22 (2023), 126. https://doi.org/10.1007/
s12346-023-00825-9.