European Journal of Mathematics (2023) 9:54

https://doi.org/10.1007/s40879-023-00650-z

RESEARCH ARTICLE

A system with weights and with critical Sobolev exponent

Asma Benhamida1 · Rejeb Hadiji2,3

Received: 20 April 2022 / Revised: 11 January 2023 / Accepted: 4 March 2023

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2023

Abstract
We investigate the minimization problem
    
1 1
inf a(x)|∇u(x)|2 d x + b(x)|∇v(x)|2 d x − λ u(x)v(x) d x
u,v∈H01 () 2  2  
||u|| L q =1,v L q =1

where q = N2N −2 , N  4, a and b are two continuous positive weight functions.

We show the existence of solutions of the previous minimizing problem under some
conditions on a, b, the dimension of the space and the parameter λ.

Keywords Critical Sobolev exponent · Minimization problem · Non-linear effects ·

Variational problem

Mathematics Subject Classification 35A01 · 35A15 · 35J57 · 35J62

1 Introduction

Let  ⊂ R N be a given smooth bounded domain, with N  4. Throughout this paper,

we are concerned with the following nonlinear minimization problem:

Qλ = inf E λ (u, v), (1)

(u,v)∈(H01 ())2 \{(0,0)}

B Rejeb Hadiji
[email protected]
Asma Benhamida
[email protected]

1 Faculty of sciences of Tunis, Department of Mathematics, University of Tunis El Manar, Tunis,

Tunisia
2 Univ Paris Est Creteil, CNRS, LAMA, 94010 Creteil, France
3 Univ Gustave Eiffel, LAMA, 77447 Marne-la-Vallée, France

0123456789().: V,-vol 123

 54 Page 2 of 22 A. Benhamida, R. Hadiji

with
 
1 1
E λ (u, v) = a(x)|∇u(x)|2 d x + b(x)|∇v(x)|2 d x
2uq2  2v(x)q2 

λ
− u(x)v(x) d x,
uq vq 

where a and b are positive continuous functions on , λ is a real constant and q = 2N

N −2
is the critical exponent for the Sobolev embedding

H01 () → L q (). (2)

Note that positive minimizers u and v for (1) are non-trivial solutions of the Euler–
Lagrange equation associated to (1) namely:
⎧ ∗
⎪
⎪ − div(a(x)∇u) − λv = 1 u 2 −1 in ,
⎪
⎨ − div(b(x)∇v) − λu =  v 2∗ −1
2 in ,
⎪
⎪ u  0, v  0 in ,
⎪
⎩
u=v=0 on ∂,

where 1 , 2 ∈ R are the Lagrange multipliers associated to (1).

In the well-known article [5], Brezis and Nirenberg treated problem (3) in the
special case where the weights a and b are positive constant functions. They proved
that the problem has at least one positive solution for 0 < λ < λ1 when N  4 and
for λ∗ < λ < λ1 when N = 3, where λ1 denotes the first eigenvalue of − with
homogeneous Dirichlet conditions and λ∗ is a positive constant.
In the presence of a non-constant positive and bounded weight h, the scalar problem
was addressed in [16]. See also [17] where the authors considered the existence of
minimizers solutions for the equation

− div(h(x)∇u) = λu + |u|q−2 u in , u > 0 in , u = 0 on ∂. (3)

They proved that the existence of a solution depends on the first eigenvalue λ1h of
−div(h(x)∇ ·) in H01 (), on the behavior of the function h in the vicinity of its
minima, and on the geometry of the domain . In [12], Furtado and Souza generalized
problem (3), considering a non-homogeneous term |u|q−2 u. We refer also to [6, 14,
15] for more general weights which depend on u and x. There are other results in the
scalar case for this kind of problem, see the references [13–15].
In [1], the authors considered a critical and subcritical system with critical non-
linearity term u α−1 v β , where α > 1 and β > 1. The main difficulty faced when dealing
with this problem is the lack of compactness of the embedding (2). We overcame this
obstacle using the presence of linear perturbation terms λu and λv. In [7], the authors
considered an arbitrary system with l components, l  2, for the Yamabe equation on
a closed Riemannian manifold; see also [8].

123
A system with weights and with critical Sobolev exponent Page 3 of 22 54

Note that the shape of the domain can have a strong influence on the type of results
one can expect. See for example the seminal work of Coron [9].
Geometric and physical motivations, in particular in relation to the Yamabe problem,
can be found for example in [2, 11, 18–21].
In [21], the authors connect the Ginzburg–Landau functional used in the physics lit-
erature with the Yamabe functional known in mathematics. It is shown that, if properly
interpreted, both functionals upon minimization produce the same Ginzburg–Landau-
type equations used on critical phenomena.
In this paper, we consider the case where a and b are non-constant distinct weights.
We prove the existence of positive solutions (u, v) which depend, among others, on
the behavior of the weights a( · ) and b( · ) near their minima and the dimension of the
space.
In order to state the problem and to announce our main results, we give some
preliminaries.
Let us assume the existence of x0 in  such that, in a neighborhood of x0 , a and b
behave like

a(x) = a(x0 ) + Ak |x − x0 |k + |x − x0 |k θa (x), as x → x0 , (4)

b(x) = b(x0 ) + Bl |x − x0 | + |x − x0 | θb (x), as x → x0 ,
l l
(5)

where k, l > 0 and Ak , Bl are positive constants, θa (x) and θb (x) tend to 0 when x
tends to x0 .
The parameters k and l will play a critical role in the study of our problem. Indeed,
if N  4 the case (k, l) with k, l > 2 is treated through a classical procedure.
If 0 < k  2, 0 < l  2 the problem is more delicate, we have to assume that the
functions a and b satisfy the following additional conditions:

ã(x)
k Ak  for a.e. x ∈ , (6)
|x − x0 |k

and

b̃(x)
l Bl  for a.e. x ∈ , (7)
|x − x0 |l

where

ã(x) ..= ∇a(x) · (x − x0 ) and b̃(x) ..= ∇b(x) · (x − x0 ).

In order to highlight the difficulty, we consider the blow-up of u, v ∈ H01 () around
x = x0 , see [14, 15]. Depending on the parameters k, l and λ different situations occur
in the blow-up scale around the point where the weights reach their minimum. More
precisely, we consider the functions wε and z ε defined by

(N −2) x − x0 (N −2) x − x0
u(x) = ε− 2 wε , v(x) = ε− 2 z ε , ε > 0.
ε ε

123
 54 Page 4 of 22 A. Benhamida, R. Hadiji

One has wε , z ε ∈ H01 (ε ) with ε = {ε−1 y, y ∈ } and wε  L q (ε ) = z ε  L q (ε ) =
u L q () = v L q () , thus
 
1 1
E λ (u, v) = a(εy + x0 )|∇wε (y)| dy +
2
b(εy + x0 )|∇z ε (y)|2 dy
2 ε 2 ε

− λε 2
wε (y)z ε (y) dy.
ε

Consequently, the blow-up around x = x0 gives

 
1 1
E λ (u, v) ∼ a(x0 )|∇wε (y)|2 dy + b(x0 )|∇z ε (y)|2 dy
ε→0 2 ε 2 ε
 
Ak
− λε 2
wε (y)z ε (y) dy + |εy + x0 |k |∇wε (y)|2 dy
ε 2 ε
 
Bl
+ |εy + x0 | |∇z ε (y)| dy − λε
l 2 2
wε (y)z ε (y) dy.
2 ε ε

Then, as we can see the three last terms have different weights and the exponents
k = 2 and l = 2 are critical for our problem.
Let us finally point out that the lower dimension N = 3 and the exponents k < 2,
l < 2 could also be interesting for this problem, but it is not yet fully understood. For
dimension 3 in the scalar case we refer to [5, 10].
The rest of the paper is organized in the following way: in Sect. 2 we state our
main result. In Sect. 3 we give a sufficient condition for the existence of minimizers.
We give precise estimates of the energy in Sect. 4. In Sect. 5 we discuss the sign of
minimizers. Section 6 contains some non-existence results.

2 Main results

Let us suppose that there exists x0 ∈  such that

a(x0 ) = min{a(x), x ∈ }, b(x0 ) = min{b(x), x ∈ }.

In this section, we will prove the existence of a solution to problem (1). For this we
define

S = inf ∇u2L 2 ,
u∈H01 ()
u L q =1

which corresponds to the best constant for the Sobolev embedding H01 () into L q ().
We assume that a(x0 ) = b(x0 ), and we denote by γ0 this value.
We are now ready to state the main result of this paper.

123
A system with weights and with critical Sobolev exponent Page 5 of 22 54

Theorem 2.1 Assume that a, b ∈ H 1 () ∩ C() satisfy (4) and (5), respec-
tively. Let λa1 , respectively λb1 , be the first eigenvalue of −div(a(x)∇ ·), respectively
−div(b(x)∇ ·), on  with zero Dirichlet boundary condition. Let λ̃1 = min{λa1 , λb1 }.
We have:
(1) If N  4, k > 2 and l > 2, then Q λ is achieved for every λ ∈ (0, λa1 ) ∩ (0, λb1 ).
(2) If N  5, k = 2 and l = 2, then there exists a constant γ (N ) = (N −2)N (N +2)
8(N −1)
(A2 + B2 ) such that Q λ is achieved for every λ ∈ (γ (N ), λ̃1 ).
(3) If N  5, k = 2 and l > 2, respectively k > 2 and l = 2, then there exists a con-
stant γ̃ (N ) = N (N8(N
−2)(N +2)
−1) such that Q λ is achieved for every λ ∈ (γ̃ (N )A2 , λ̃1 ),
respectively λ ∈ (γ̃ (N )B2 , λ̃1 ).
(4) If N = 4, k = 2 and l > 2, respectively k > 2 and l = 2, then there exists
a minimizing solution of Q λ for every λ ∈ ( A2 , λ̃1 ), respectively for every λ ∈
( B2 , λ̃1 ), where A2 and B2 are some constants.

We now proceed with the proof of Theorem 2.1 as follows: First we show that
0  Q λ < γ0 S and then we prove that this implies that the infimum Q λ is achieved.

3 Sufficient conditions for the existence of minimizers

We first prove the existence of Q λ , which is guaranteed by the following result.

Proposition 3.1 Let ϕ1a , respectively ϕ1b , be the first eigenfunction of −div(a(x)∇ ·),
respectively −div(b(x)∇ ·), associated to the first eigenvalue λa1 , respectively λb1 . We
have
(i) Assume that 0 < λ < λ̃1 , then

Q λ  0.

ϕ1a  L q ϕ1b  L q 1− q2
(ii) For λ  || max(λa1 , λb1 ), one has
 ϕ1 ϕ1 d x
a b

Q λ  0.

Proof (i). Let 0 < λ < λ̃1 (), and let u and v be such that u L q = v L q = 1.
By the definitions of λa1 , λb1 and λ̃1 () one has


λ̃1 λ̃1
E λ (u, v)  u2L 2 + v2L 2 − λ uv d x.
2 2 

Applying the Cauchy–Schwarz inequality, we find

λ̃1 
E λ (u, v)  u2L 2 + v2L 2 − 2u L 2 v L 2 .
2

123
 54 Page 6 of 22 A. Benhamida, R. Hadiji

Thus

λ̃1  2
E λ (u, v)  u L 2 − v L 2  0.
2

Consequently, Q λ  0.
(ii). We have

ϕ1a ϕb
Q λ  Eλ , b1 ,
ϕ1  L q ϕ1  L q
a

which implies that

 
1 1
Qλ  a(x)|∇ϕ1a |2 d x + b(x)|∇ϕ1b |2 d x
2ϕ1a 2L q  2ϕ1b 2L q 

λ
− ϕ1a ϕ1b d x.
ϕ1a  L q ϕ1b  L q 

By the definitions of ϕ1a and ϕ1b , one has

  
1 1 λ
Qλ  λa |ϕ1a |2 d x + λb1 |ϕ1b |2 d x − ϕ1a ϕ1b d x.
2ϕ1a 2L q 1  2ϕ1b 2L q  ϕ1a  L q ϕ1b  L q 

1
− q1
Using the embedding of L q into L 2, there exists a positive constant C1 = || 2
such that

1 1 λ
Qλ  λa C12 ϕ1a 2L q + λb1 C12 ϕ1b 2L q − a ϕ1a ϕ1b d x.
2ϕ1a 2L q 1 2ϕ1b 2L q ϕ1  L q ϕ1b  L q 

Thus

1− q2 λ
Q λ  || max(λa1 , λb1 ) − ϕ1a ϕ1b d x.
ϕ1a  L q ϕ1b  L q 

Therefore, Q λ  0, completing the proof of the proposition.

Lemma 3.2 Let 0 < λ < λ̃1 . If Q λ < γ0 S, then the infimum in (1) is achieved.

Proof Let {Un } ⊂ (H01 ())2 be a minimizing sequence for (1) that is,

Un  L q = 1 (which means ||u n || L q = 1, ||vn || L q = 1), (8)

  
1 1
a(x)|∇u n |2 d x + b(x)|∇vn |2 d x − λ u n vn d x = Q λ + o(1) (9)
2  2  

123
A system with weights and with critical Sobolev exponent Page 7 of 22 54

as n → ∞. The sequence {Un } is bounded in (H01 ())2 . Indeed, from (9), we have
  
1 1
a(x)|∇u n | d x +
2
b(x)|∇vn | d x = λ
2
u n vn d x + Q λ + o(1).
2  2  

Using Hölder’s inequality we have


|u n ||vn | d x  u n  q vn  L q ,
 L q−1

then
  
1
a(x)|∇u n |2 d x + b(x)|∇vn |2 d x  λu n  q vn  L q + Q λ + o(1).
2   L q−1

Consequently, we have

(u n , vn )(H 1 ())2  C.

0

Then there is U = (u, v), up to a subsequence, still denoted by Un = (u n , vn ), such

that

(u n , vn )(u, v) weakly in (H01 ())2 ,

(u n , vn ) → (u, v) strongly in (L 2 ())2 ,
(u n , vn ) → (u, v) a.e. on ,

with u L q  1 and v L q  1.

Set wn = u n − u and z n = vn − v, so that

(wn , z n )(0, 0) weakly in (H01 ())2 ,

(wn , z n ) → (0, 0) strongly in (L 2 ())2 ,
(wn , z n ) → (0, 0) a.e. on .

By (8) and using the definition of S and the fact that min  a(x) = γ0 > 0 and
min  b(x) = γ0 > 0, we have
  
1 1
a(x)|∇u n | d x + 2
b(x)|∇vn | d x 2
2γ0 S.
2   2

By (9), we have
   
1
a(x)|∇u n |2 d x + b(x)|∇vn |2 d x − λ u n vn d x = Q λ + o(1),
2   

123
 54 Page 8 of 22 A. Benhamida, R. Hadiji

which implies that

   
1
a(x)|∇u n |2 d x + b(x)|∇vn |2 d x = Q λ + λ u n vn d x + o(1),
2   

which gives

λ uv d x  γ0 S − Q λ > 0


and so u ≡ 0 and v ≡ 0.
Again using (9) we obtain
    
1
a(x)|∇u| d x +
2
a(x)|∇wn | d x +2
b(x)|∇v| d x + b(x)|∇z n | d x
2 2
2    

− λ u n vn d x = Q λ + o(1). (10)


On the other hand, as {wn } and {z n } are bounded in L q () and wn → 0 for a.e. x
in , z n → 0 for a.e. x in , then we can use the Brezis–Lieb lemma, see [4],

q q q
u + wn  L q = u L q + wn  L q + o(1),
q q q
v + z n  L q = v L q + z n  L q + o(1).

Using (8) we have

q q q q
1 = u L q + wn  L q + o(1), 1 = v L q + z n  L q + o(1).

And so,

1  u2L q + wn 2L q + o(1), 1  v2L q + z n 2L q + o(1). (11)

We sum equations (11) and obtain

 
1
2  u2L q + v2L q + a(x)|∇wn |2 d x + b(x)|∇z n |2 d x .
γ0 S  

Since Q λ > 0, this implies

 
Qλ Qλ Qλ
Qλ  u2L q + v2L q + a(x)|∇wn |2 d x + b(x)|∇z n |2 d x . (12)
2 2 2γ0 S  

123
A system with weights and with critical Sobolev exponent Page 9 of 22 54

Adding (10) and (12), we obtain

 
1
a(x)|∇u| d x + a(x)|∇wn |2 d x
2
2  
   
+ b(x)|∇v| d x + b(x)|∇z n | d x − λ uv d x
2 2
  
 
Qλ Qλ Qλ
 u L q +
2
v L q +
2
a(x)|∇wn | d x + b(x)|∇z n |2 d x .
2
2 2 2γ0 S  

Thus

   
1
a(x)|∇u| d x +
2
b(x)|∇v| d x − λ
2
uv d x
2   
Qλ Qλ
 u2L q + v2L q
2  2   
Qλ 1
+ − a(x)|∇wn |2 d x + b(x)|∇z n |2 d x .
2γ0 S 2  

Hence

2E λ (u, v)  Q λ u2L q + Q λ v2L q

   
Qλ
+ −1 a(x)|∇wn |2 d x + b(x)|∇z n |2 d x + o(1),
γ0 S  

and since Q λ < γ0 S, we know that u = 0 and v = 0. We deduce

u v
2E λ ,
u L q v L q
 
1 1
 − 1 a(x)|∇u|2 d x + − 1 b(x)|∇v|2 d x
u2L q  v2L q 
 
uv
+ Q λ u2L q + Q λ u2L q + λ uv d x − λ d x.
  u L q v L q

123
 54 Page 10 of 22 A. Benhamida, R. Hadiji

Then
u v
2E λ ,
u L q v L q
   
  |∇u|2 uv
 u L q − 1 Q λ − a(x)
2
d x − 2λ dx
 u2L q  u L q v L q
   
  |∇v|2 uv
+ v L q − 1 Q λ − b(x)
2
d x − 2λ dx
 v2L q  u L q v L q
  
uv   uv
+ λ uv d x − λ d x + 2λ uq2 − 1 dx
 u v u L q v L q
 
L q L q

  uv
+ 2λ v2L q − 1 d x + 2Q λ .
 u L q v L q

On the one hand, we have u2L q − 1  0 and v2L q − 1  0, we obtain

u v
2E λ ,
u L q v L q
 
(u L q v L q − 1) + 2(u2L q − 1) + 2(v2L q − 1)
λ uv d x + 2Q λ .
u L q v L q 

On the other hand, we have u L q v L q − 1  0, then

u v
2E λ ,  2Q λ .
u L v L q
q

This means that (u, v) is a minimum of Q λ .

4 Precise estimates of the energy

Proposition 4.1 (a) For N  4, and k > 2, l > 2 we have

Q λ < γ0 S, for all λ > 0.

(b) For N = 4, and k = 2, l = 2 we have

Q λ < γ0 S, for all λ > A2 + B2 .

(c) For N  5, and k = 2, l = 2 we have

Q λ < γ0 S, for all λ > m N (A2 + B2 ).

(d) For N  5, and k = 2, l > 2 (respectively k > 2, l = 2) we have

Q λ < γ0 S, for all λ > m N A2 (respectively λ > m N B2 ).

123
A system with weights and with critical Sobolev exponent Page 11 of 22 54

(e) For N = 4, and k = 2, l > 2, respectively k > 2, l = 2 we have

Q λ < γ0 S, for all λ > A2 , respectively for all λ > B2 ,

where A2 , B2 are defined by (6) and (7), K 3 = 1

R N (1+|x|2 ) N −2 d x and m N =
N (N −2)(N +2)
8(N −1) .

Proof We shall estimate the ratio E λ (u, v) defined in (1), with u = u x0 ,ε

N −2
ε
= ζ Uε (x − x0 ), where for x ∈ R N, Uε (x) = 4
N −2 and ζ ∈ Cc∞ () with
(ε+|x|2 ) 2
ζ  0 and ϕ ≡ 1 on a neighborhood of x0 ; for more details see [5, 23].
We recall from [5] that


 N −2 
|∇u x0 ,ε (x)|2 d x = K 1 + O ε 2 ,

 2
q  N −2 
|u x0 ,ε (x)| d x
q
= K2 + O ε 2 , (13)

   N −2 
K3ε + O ε 2 , if N  5,
|u x0 ,ε (x)| d x = ω4
2
(14)
 2 ε|log ε| + o(ε| log ε|), if N = 4,

K1
where K 1 and K 2 are positive constants with K2 = S, w4 is a the area of S 3 and
K 3 = R N (1+|x|12 ) N −2 d x.
We know by [16], the following estimates:

N −2 
ε 2
a(x)|∇u x0 ,ε (x)|2 d x
2 
⎧  N −2 
⎪ a(x0 )K 1
⎪
⎪ +O ε 2 if N  4 and N − 2 < k,
⎨ a(x0 )K 1 + Ck ε k2 + oε k2 
⎪ 2
if N  4 and N − 2 > k,
 2 2

⎪ a(x0 )K 1 + (N −2)2 ω N (A N −2 +M) ε N 4−2 |log ε| + o ε N 2−2 |log ε|
 (15)
⎪
⎪ if N > 4 and k = N − 2,
⎪ 2
⎩ a(x0 )K 1
2
2 + A2 ω4 ε|log ε| + o(ε|log ε|) if N = 4 and k = 2

and

N −2
ε 2 b(x)|∇u x0 ,ε (x)|2 d x

⎧  N −2 
⎪ b(x0 )K 1
⎪
⎪ +O ε 2 if N  4 and N − 2 < l,
⎨ b(x0 )K 1 + Dl ε 2l + oε 2l 
⎪ 2
if N  4 and N − 2 > l,
 ⎪
2 2
N −2  N −2  (16)
+ (N −2) ω N4(B N −2 +M) ε 2 | log ε| + o ε 2 | log ε|
2
b(x0 )K 1
⎪
⎪ if N > 4 and l = N − 2,
⎪ 2
⎩ b(x0 )K 1
2 + B2 ω4 ε|log ε| + o(ε|log ε|) if N = 4 and l = 2,

123
 54 Page 12 of 22 A. Benhamida, R. Hadiji

where
  N −2
|y|2 dy N
K 1 = (N − 2) 2
dy, K 2 = ,
R N (1 + |y| ) R N (1 + |y| )
2 N 2 N
 
|y|k+2 |y|l+2
Ck = (N − 2)2 Ak dy, D k = (N − 2)2
B k dy,
R N (1 + |y| ) R N (1 + |y| )
2 N 2 N

M and M  are positive constants.

We claim that, as ε → 0, by adding (13), (14), (15) and (16), we obtain E(u x0 ,ε ) 
⎧ 
⎪
⎪ N  5,
⎪
⎪
a(x0 )K 1 +o(ε)
+ b(x0 )K 1 +o(ε)
−λ K K3
ε if
⎪
⎪ 2K 2 2K 2
k > 2, l > 2,
⎪
⎪
2

⎪
⎪
⎪
⎪ a(x0 )K 1 +C2 ε+o(ε) b(x0 )K 1 +D2 ε+o(ε) N  5,
⎪
⎪ + −λ K K3
ε if
⎪
⎪ 2K 2 2K 2
k = 2, l = 2,
⎪
⎪
2

⎪
⎪
⎪
⎪ N = 4,
⎪
⎪
⎪
a(x0 )K 1 +o(ε)
+ b(x0 )K 1 +o(ε)
− λε ω24 |log ε|
K 2 + o(ε|log ε|) if
⎪
⎪
2K 2 2K 2
k > 2, l > 2,
⎪
⎪ 
⎪
⎪ N = 4,
⎪ a(x0 )K 1 +2ω4 A2 ε|log ε|+o(|log ε|)
⎪
⎪
⎪ + b(x0 )K 1 +2ω4 B22K
ε|log ε|+o(ε|log ε|)
− λε ω4
|log ε| if
⎪
⎪
2K 2 2 2K 2
k = 2, l = 2,
⎪
⎪ 
⎪
⎪ N  5,
⎪
⎪ a(x0 )K 1 +C2 ε+o(ε)
⎪
⎨ + b(x0 )KK12+o(ε) − λ K K3
ε + O(1) if
2K 2 2 k = 2, l > 2,

⎪
⎪ N  5,
⎪
⎪
a(x0 )K 1 +o(ε)
+ b(x0 )K 12K
+D2 ε+o(ε)
− λK K3
ε + O(1) if
⎪
⎪ 2K 2
k > 2, l = 2,
⎪
⎪
2 2

⎪
⎪
⎪
⎪ N = 4,
⎪
⎪
⎪
a(x0 )K 1 +o(ε)
+ b(x0 )K 1 +2ω4 B22K
ε|log ε|+o(ε|log ε|)
− λε 2K ω4
|log ε| + o(ε|log ε|) if
⎪
⎪
2K 2 2 2 k > 2, l = 2,
⎪
⎪ 
⎪
⎪ N = 4,
⎪
⎪ a(x0 )K 1 +2ω4 A2 ε|log ε|+o(|log ε|)
⎪
⎪ + b(x0 )K 1 +o(ε)
− λε 2K ω4
|log ε| + o(ε|log ε|) if
⎪
⎪
2K 2 2K 2 2 k = 2, l > 2,
⎪
⎪ 
⎪
⎪ N = 4,
⎪
k k
⎪
⎪
⎪
a(x0 )K 1 +Ck ε 2 +o(ε 2 )
+ b(x0 )K 1 +2ω4 B22Kε|log ε|+o(ε|log ε|)
− λε 2K ω4
|log ε| + o(ε|log ε|) if
⎪
⎪
2K 2 2 2 k < 2, l = 2,
⎪
⎪ 
⎪
⎪ N = 4,
⎪
l l
⎪ a(x0 )K 1 +2ω4 A2 ε|log ε|+o(ε|log ε|)
⎪
⎩ + b(x0 )K 1 +D k ε 2 +o(ε 2 )
− λε ω24 |log ε| + o(ε|log ε|) if
2K 2 2K 2
k = 2, l < 2,

123
A system with weights and with critical Sobolev exponent Page 13 of 22 54

Thus Q λ 
⎧ 
⎪
⎪  a(x0 )+b(x0 )  N  5,
⎪
⎪ S−λ K K3
ε + o(ε) if
⎪
⎪ 2
k > 2, l > 2,
⎪
⎪
2

⎪
⎪
⎪
⎪  a(x0 )+b(x0 )   D2  K 3 N  5,
⎪
⎪ S − λ − 2KC2
− 2K ε + o(ε) if
⎪
⎪ 2 3 K2 k = 2, l = 2,
⎪
⎪
3

⎪
⎪
⎪
⎪ a(x0 )+b(x0 )  N = 4,
⎪
⎪ S−λ K ω4
ε|log ε| + o(ε|log ε|) if
⎪
⎪ 2
k > 2, l > 2,
⎪
⎪
2

⎪
⎪
⎪
⎪  a(x0 )+b(x0 )  N = 4,
⎪
⎪ S− Kω4
(λ − (A2 + B2 ))ε|log ε| + o(ε|log ε|) if
⎪
⎪ 2
k = 2, l = 2,
⎪
⎪
2

⎪
⎪
⎪
⎪  a(x0 )+b(x0 )   C2  K 3 N  5,
⎪
⎪ S − λ − 2K ε + o(ε) if
⎨ 2 3 K2 k = 2, l > 2,

⎪
⎪  a(x0 )+b(x0 )   D2  K 3 N  5,
⎪
⎪ S − λ − 2K ε + o(ε) if
⎪
⎪ 2 3 K2 k > 2, l = 2,
⎪
⎪ 
⎪
⎪
⎪ a(x )+b(x ) 
⎪ N = 4,
⎪
⎪ 0 0
S− Kω4
[λ − B2 ]ε|log ε| + o(ε|log ε|) if
⎪
⎪ 2
k > 2, l = 2,
⎪
⎪
2

⎪
⎪
⎪
⎪  a(x0 )+b(x0 )  N = 4,
⎪
⎪ S− Kω4
[λ − A2 ]ε|log ε| + o(ε|log ε|) if
⎪
⎪ 2
k = 2, l > 2,
⎪
⎪
2

⎪
⎪
⎪
⎪   N = 4,
⎪
⎪ a(x0 )+b(x 0)
S− Kω4 k k
[λ − B2 ]ε|log ε| + CKk2 ε 2 + o(ε|log ε|) + o(ε 2 ) if
⎪
⎪ 2
k < 2, l = 2,
⎪
⎪
2

⎪
⎪
⎪
⎪  a(x0 )+b(x0 )  N = 4,
⎪
⎪ S− Kω4
[λ − A2 ]ε|log ε| + KDk 2l l
ε + o(ε 2 ) + o(ε|log ε|) if
⎩ 2 2 2 k = 2, l < 2,

where CK23 = N (N4(N

−2)(N +2)
−1) A2 and D2
K3 = N (N −2)(N +2)
4(N −1) B2 . From these estimates we
get the desired result.

Combining Lemma 3.2 and Proposition 4.1 we conclude that the infimum in (1) is
achieved. This leads to the conclusion of Theorem 2.1.
a(x0 )+b(x0 )
Remark 4.2 We do not know if we can go strictly below 2 S in the case where
N  4, k < 2 and l < 2.

5 The sign of the minimizers

Now, we will discuss the positivity of the solutions.

Proposition 5.1 (i) If Q λ is achieved in (u, v) then uv  0.
(ii) There exist solutions u and v of the minimization problem (1) such that u  0 and
v  0.

Proof (i). Set

  
1 1
F(u, v) = a(x)|∇u|2 d x + b(x)|∇v|2 d x − λ u(x)v(x) d x.
2  2  

The inequality

F(u, v)  F(|u|, |v|)

123
 54 Page 14 of 22 A. Benhamida, R. Hadiji

gives that −λ  uv d x  −λ  |uv| d x, then


(u(x)v(x) − |u(x)v(x)|) d x  0. (17)


We have u(x)v(x) − |u(x)v(x)|  0 for all x ∈ . Hence


(u(x)v(x) − |u(x)v(x)|) d x  0. (18)


Combining (17) and (18) we obtain that uv − |uv| = 0 and hence |uv| = uv.
Finally for all x ∈ , u(x)v(x)  0.
(ii). From Lemma 3.2 we know there exists a minimum (u, v). By Proposition 5.1 (i)
we know that |uv| = uv.
On the other hand, we have
  
1 1
F(|u n |, |vn |) = a(x)|∇u n |2 d x + b(x)|∇vn |2 d x − λ |u n vn | d x = F(u n , vn ).
2  2  

Then if we have (u n , vn ) are solutions of (1) then (|u n |, |vn |) are also minimizing
solutions.
Thus, when dealing with (1), one can assume without loss of generality that u  0
and v  0.

Remark 5.2 Since u and v are minimizers for (1) we obtain Lagrange multipliers
1 , 2 ∈ R such that the Euler–Lagrange equation formula associated to (1) is
⎧
⎪ 2∗ −1
⎪−div(a(x)∇u) − λv = 1 u
⎪ in ,
⎪
⎪ ∗
⎪−div(b(x)∇v) − λu = 2 v 2 −1
⎨ in ,
u2∗ = u2∗ = 1, (19)
⎪
⎪
⎪
⎪ u  0, v  0 in ,
⎪
⎪
⎩u = v = 0 on ∂,

1 +1
such that 2 = Q λ  0 according to Proposition 3.1.

6 Non-existence results

In this section, we assume that a, b satisfy (4) and (5) with k > 2, l > 2 and (6), (7)
if k  2, l  2. we obtain in these cases a few non-existence results. We define

ω(a, b) ..= inf φa,b (u, v)

(u,v)∈(H01 ())2 \{0}

123
A system with weights and with critical Sobolev exponent Page 15 of 22 54

where
 
1  ã(x)|∇u|2 + b̃(x)|∇v|2 d x
φa,b (u, v) ..= .
4  u(x)v(x) d x

We see that ω(a, b) ∈ [−∞, +∞[.

The main goal of this section is the non-existence results.
Theorem 6.1 Assume that λ  ω(a, b) and  is a strictly star-shaped domain with
respect to x0 . Then (1) has no minimizers.
The proof of this result follows from the following Pohožaev identity; see [22].
Proposition 6.2 If (u, v) is a solution of (19) then (u, v) satisfies the following identity:
  
1 1
2λ u(x)v(x) d x − [∇b(x) · (x − x0 )]|∇v(x)|2 d x − [∇a(x) · (x − x0 )]|∇u(x)|2 d x
 2  2 
  2   2
1  ∂u  1  ∂v 
= a(x)[(x − x0 ) · n]   dσ (x) + b(x)[(x − x0 ) · n]   dσ (x),
2 ∂ ∂ν 2 ∂ ∂ν

where n denotes the outward normal to ∂.

Proof Suppose that (u, v) is a solution of problem (19). We multiply the first equation
in the system (19) by ∇u(x) · (x − x0 ) and we integrate by parts, leading to

N
1 |u(x)|q−2 u(x)∇u(x) · (x − x0 ) d x = − 1 , (20)
 q
  
λ v(x)∇u(x) · (x − x0 ) d x = − λ u(x)∇v(x) · (x − x0 ) d x − N λ v(x)u(x) d x (21)
  

and
 
N −2
−div(a(x)∇u(x))∇u(x) · (x − x0 ) d x = − a(x)|∇u(x)|2 d x
 2 

1
− ∇a(x) · (x − x0 )|∇u(x)|2 d x
2 
  2
1  ∂u 
− a(x)  ((x − x0 ) · n) dσ (x)
2 ∂ ∂ν
(22)

where n denotes the outward normal to ∂.

Combining (20), (21)and (22) we get
 
N −2 1
− a(x)|∇u(x)|2 d x − ∇a(x) · (x − x0 )|∇u(x)|2 d x
2  2 
  2
1  ∂u 
− a(x)  ((x − x0 ) · n) dσ (x) (23)
2 ∂ ∂ν

N
= λ v(x)∇u(x) · (x − x0 ) d x − 1 .
 q

123
 54 Page 16 of 22 A. Benhamida, R. Hadiji

Similarly, we multiply the second equation of (19) by ∇v(x) · (x − x0 ) and integrate

by parts, we get
 
N −2 1
− b(x)|∇v(x)| d x − 2
∇b(x) · (x − x0 )|∇v(x)|2 d x
2  2 
  2
1  ∂v 
− b(x)  ((x − x0 ) · n) dσ (x) (24)
2 ∂ ∂ν

N
= λ u(x)∇v(x) · (x − x0 ) d x − 2 .
 q

N −2 N −2
On the other hand, multiplying equations in (19) by 2 u and 2 v, respectively,
integrating and summing the obtained results, we get
 
N −2 N −2 N −2
a(x)|∇u(x)|2 d x = λ v(x)u(x) d x + 1 (25)
2  2  2

and
 
N −2 N −2 N −2
b(x)|∇v(x)|2 d x = λ u(x)v(x) d x + 2 . (26)
2  2  2

Combining (23) and (25), we obtain

 
N −2
− λ u(x)v(x) d x − λ v(x)[∇u(x) · (x − x0 )] d x
2  
   2 (27)
1 1  ∂u 
= |∇u| [∇a(x) · (x − x0 )] d x +
2  
a(x)  [(x − x0 ) · n] dσ (x).
2  2 ∂ ∂ν

On the other hand, combining (24) and (26), we get

 
N −2
− λ u(x)v(x) d x − λ u(x)[∇v(x) · (x − x0 )] d x
2  
   2 (28)
1 1  ∂v 
= |∇v(x)| [∇b(x) · (x − x0 )] d x +
2  
b(x)  [(x − x0 ) · n] dσ (x).
2  2 ∂ ∂ν

Adding (28) in (27) and using (21), we find

  
1 1
2λ u(x)v(x) d x − ∇b(x) · (x − x0 )|∇v(x)|2 d x − [∇a(x) · (x − x0 )]|∇u|2 d x
 2  2 
  2   2
1  ∂u  1  ∂v 
= a(x)  [(x − x0 ) · n] dσ (x) + b(x)  [(x − x0 ) · n] dσ (x).
2 ∂ ∂ν 2 ∂ ∂ν

Finally we get the proposition.

123
A system with weights and with critical Sobolev exponent Page 17 of 22 54

Now let us prove Theorem 6.1. Assume that λ  ω(a, b) and let  be strictly
starshaped domain with respect to x0 then

(x − x0 ) · n > 0, for all x ∈ ∂.

Suppose that (u, v) is a solution of (19). By Proposition 6.2 we get

 
1
2λ u(x)v(x) d x − ∇b(x) · (x − x0 )|∇v(x)|2 d x
 2 

1
− [∇a(x) · (x − x0 )]|∇u(x)|2 d x > 0.
2 

It follows that

1  ∇a(x) · (x − x0 )|∇u(x)|2 + ∇b(x) · (x − x0 )|∇v(x)|2 d x
λ> inf
4 (u,v)∈(H01 ())2 \{0}  u(x)v(x) d x

which is a contradiction.
Then problem (19) does not admit solutions. Consequently, we obtain the desired
result.

6.1 Estimates of !(a, b)

We end this section, using the techniques of [16] and give some estimates of ω(a, b).

Proposition 6.3 (1) Assume a, b ∈ C 1 () and there exists z 0 ∈  such that ã(z 0 ) +
b̃(z 0 ) < 0, then ω(a, b) = −∞.
(2) Assume a, respectively b ∈ H 1 () ∩ C(), satisfies (4), respectively (5).
(2.i) If k > 2, l > 2 and a, b ∈ C 1 (), then ω(a, b) = 0.
(2.ii) If k = 2, l > 2 or k > 2, l = 2, and a, b satisfy moreover (6) and (7) with
ã(x)  0, b̃(x)  0 for a.e. x ∈ , then

N2   A2
min A2 , l Bl (diam )l−2  ω(a, b)  λ1 (diam )2 (29)
16 2

and

N2   B2
min k Ak (diam )k−2, B2  ω(a, b)  λ1 (diam )2 . (30)
16 2

(2.iii) If 0 < k  2, 0 < l  2, a, b satisfy conditions (6) and (7), respectively, then

N2  
min k Ak (diam )k−2, l Bl (diam )l−2  ω(a, b).
16

123
 54 Page 18 of 22 A. Benhamida, R. Hadiji

Proof Let ϕ ∈ Cc (R N ) such that 0  ϕ  1 on R N, ϕ ≡ 1 on B(0, 2r ) and ϕ ≡ 0 on

R N \ B(0, 2r ), where 0 < r < 1. Set ϕ j (x) = ϕ( j(x − z 0 )) for j ∈ N∗, then we have

 (ã(x) + b̃(x))|∇ϕ j (x)|

2
1 dx
ω(a, b) 
 ϕ j (x) d x
4 2

B(b,2r / j) (ã(x) + b̃(x))|∇ϕ j (x)|

2
1 dx
 .
B(b,2r / j) ϕ j (x) d x
4 2

After the change of variable y = j(x − z 0 ), we get

 y   
j2 B(0,2r ) ã j + z 0 + b̃ yj + z 0 |∇ϕ(y)|2 dy
ω(a, b)  .
B(0,2r ) ϕ
4 2 (y) dy

Applying the Dominated Convergence Theorem, we obtain the desired result when j
goes to infinity.
Now we will prove (2.i). Using (4), (5) and since a, b ∈ C 1 () in a neighborhood
V of x1 , we can write

a(x) = a(x1 ) + Ak |x − x1 |k + θa (x), (31)

b(x) = b(x1 ) + Bl |x − x1 | + θb (x),
l
(32)

where θa (x) and θb (x) ∈ C 1 (V ) are such that

|θa (x)| |θb (x)|

lim = 0 and lim = 0. (33)
x→x1 |x − x1 |k x→x2 |x − x1 |l

From (33), we get the existence of r such that 0 < r < 1 and

|θa (x)|  |x − x0 |k and |θb (x)|  |x − x0 |l, for all x ∈ B(x0 , 2r ) ⊂ V . (34)

Let ϕ j (x) = ϕ( j(x − x0 )) be defined as in the proof of (1), we have

 (ã(x) + b̃(x))|∇ϕ j (x)|

2 dx
1
0  ω(a, b)  .
 ϕ j (x) d x
4 2

Using (31) and (32), we obtain

 
1 B(x0 ,2r / j) k Ak |x − x0 |k + l Bl |x − x0 |l |∇ϕ j (x)|2 d x
0  ω(a, b) 
B(x0 ,2r / j) ϕ j (x) d x
4 2
 
1 B(x0 ,2r / j) ∇θa (x) · (x − x0 ) + ∇θb (x) · (x − x0 ) |∇ϕ j (x)| d x
2
+ .
B(x0 ,2r / j) ϕ j (x) d x
4 2

123
A system with weights and with critical Sobolev exponent Page 19 of 22 54

By the change of variable y = j(x − x0 ) and integrating by parts, we obtain

y 
B(0,2r ) |y| B(0,2r ) θa j + x0 · div(y|∇ϕ(y)|2 ) dy
k |∇ϕ(y)|2 dy
k Ak j
0  ω(a, b) k−2 −
B(0,2r ) ϕ(y) B(0,2r ) ϕ(y) dy
4j 2 dy 4 2
y 
B(0,2r ) θb j + x 0 · div(y|∇ϕ(y)| ) dy
2
l Bl B(0,2r ) |y|l |∇ϕ(y)|2 dy j
+ − .
B(0,2r ) ϕ(y) dy B(0,2r ) ϕ(y) dy
4 j l−2 2 4 2

Using (34), we get

B(0,2r ) |y|
k ∇ϕ(y) dy
B(0,2r ) |y| dy
k
k Ak C
0  ω(a, b)  +
B(0,2r ) ϕ
2 (y) dy
4 j k−2 j k−1 B(0,2r ) ϕ (y) dy
2

B(0,2r ) |y| |∇ϕ(y)| dy B(0,2r ) |y| dy

l 2 l
l Bl C
+ l−2 + ,
B(0,2r ) ϕ (y) dy j l−1 B(0,2r ) ϕ (y) dy
4j 2 2

where C = max y∈B(0,2r ) |div(y|∇ϕ(y)|2 )|.

Therefore, for k > 2 and l > 2, when j tends to ∞ we obtain ω( p, q) = 0.
To prove (2.ii), we start with the case k = 2 and l > 2. We show the left-hand
of the inequality in (29). Since a and b satisfy (6) and (7), respectively, for all pairs
(u, v) ∈ E \ {0}, we have

1  |(x − x0 ) · ∇u(x)|2 d x l  |x − x0 |l−2 |(x − x0 ) · ∇v(x)|2 d x

φa,b (u, v) A2 + Bl
2  u(x)v(x) d x 4  u(x)v(x) d x
1  |(x − x0 ) · ∇u(x)|2 d x l  |(x − x0 ) · ∇v(x)|2 d x
 A2 + Bl (diam )l−2 .
2  u(x)v(x) d x 4  u(x)v(x) d x

Applying [16, Lemma 2.1] for t = 0, we find

 |u(x)|  |v(x)|
2 2d x 2 2 dx
1 N l N
φa,b (u, v)  A2 + Bl (diam )l−2 .
2 2  u(x)v(x) d x 4 2  u(x)v(x) d x

This implies that

N2  
ω(a, b)  min A2 , l Bl (diam )l−2 .
16

Similarly, we deduce in the case k > 2 and l = 2, that

N2  
ω(a, b)  min k Ak (diam )k−2, B2 .
16

Now we prove the right-hand of the inequality in (29) and (30). Let ψ j (x) =
ϕ1 ( j(x − x0 )) for j ∈ N large enough, where ϕ1 is a positive eigenfunction corre-
sponding to the first eigenvalue λ1 of the operator − in H01 ().

123
 54 Page 20 of 22 A. Benhamida, R. Hadiji

We have

 (ã(x) + b̃(x))|ψ j (x)|

2 dx
1
0  ω(a, b)  .
 ψ j (x) d x
4 2

Using (4) and (5), we obtain

 
x0 + 1j  2 A2 |x − x0 |2 + l Bl |x − x0 |l |∇ψ j (x)|2 d x
0  ω(a, b) 
x0 , 1j  ψ j (x) d x
4 2

 
x0 + 1j  ∇θa (x) · (x − x0 ) + ∇θb (x) · (x − x0 ) |∇ψ j (x)|2 d x
+ .
x0 + 1j  ψ j (x) d x
4 2

By the change of variable y = j(x − x0 ) and integrating by parts, we have by (33)

 |y|
2 |∇ϕ (y)|2 dy
 |y| dy
k
A2 1 C
0  ω(a, b)  +
 ϕ1 (y) dy j  ϕ1 (y)2 dy
2 2

l Bl  |y|l |∇ϕ1 (y)|2 dy C  |y| dy

l
+ + ,
 ϕ1 (y) dy
4 j l−2 2 j l−1  ϕ1 (y)2 dy

where C = max y∈ |div(y|∇ϕ1 (y)|2 )|. Letting j → ∞, we get

 |y|
2 |∇ϕ (y)|2 dy
A2 1
0  ω(a, b)  ,
2 ϕ
 1 (y) 2 dy

therefore

A2
0  ω(a, b)  λ1 (diam )2 .
2

Similarly, we deduce in the case k > 2 and l = 2 that

B2
0  ω(a, b)  λ1 (diam )2 .
2

Let us now prove (2.iii). Since a and b satisfy (6) and (7), respectively, for all
(u, v) ∈ E \ {0}, we have

k  |x − x0 |k−2 |(x − x0 ) · ∇u|2 d x l  |x − x0 |l−2 |(x − x0 ) · ∇v|2 d x

φa,b (u, v) Ak + Bl
4  uv d x 4  uv d x
k  |(x − x0 ) · ∇u|2 d x l  |(x − x0 ) · ∇v|2 d x
 Ak (diam )k−2 + Bl (diam )l−2 .
4  uv d x 4  uv d x

123
A system with weights and with critical Sobolev exponent Page 21 of 22 54

Applying [16, Lemma 2.1] for t = 0, we find

 |u|  |v|
2 2 dx 2 2 dx
k N l N
φa,b (u, v)  Ak (diam )k−2 + Bl (diam )l−2 .
4 2  uv d x 4 2  uv d x

This implies that

N2  
ω(a, b)  min k Ak (diam )k−2, l Bl (diam )l−2 .
16

References
1. Alves, C.O., de Morais Filho, D.C., Souto, M.A.S.: On systems of elliptic equations involving subcrit-
ical or critical Sobolev exponents. Nonlinear Anal. 42(5), 771–787 (2000)
2. Aubin, T.: Equations différentielles non linéaires et probleme de Yamabe concernant la courbure
scalaire. J. Math. Pures Appl. 55(3), 269–296 (1976)
3. Brezis, H.: Some variational problems with lack of compactness. In: Browder, F.E. (ed.) Nonlinear
Functional Analysis and its Applications, Part 1. Proceedings of Symposia in Pure Mathematics, vol.
45, pp. 165–201. American Mathematical Society, Providence (1986)
4. Brézis, H., Lieb, E.: A relation between pointwise convergence of functions and convergence of func-
tionals. Proc. Amer. Math. Soc. 88(3), 486–490 (1983)
5. Brézis, H., Nirenberg, L.: Positive solutions of nonlinear elliptic equations involving critical Sobolev
exponents. Commun. Pure Appl. Math. 36(4), 437–477 (1983)
6. Bae, S., Hadiji, R., Vigneron, F., Yazidi, H.: A non-linear problem involving a critical Sobolev exponent.
J. Math. Anal. Appl. 396(1), 98–107 (2012)
7. Clapp, M., Pistoia, A.: Yamabe systems and optimal partitions on manifolds with symmetries. Electron.
Res. Arch. 29(6), 4327–4338 (2021)
8. Conti, M., Terracini, S., Verzini, G.: Nehari’s problem and competing species systems. Ann. Inst. H.
Poincaré Anal. Non Linéaire 19(6), 871–888 (2002)
9. Coron, J.-M.: Topologie et cas limite des injections de Sobolev. C. R. Acad. Sci. Paris Sér. I Math.
299(7), 209–212 (1984)
10. Crouau, R., Hadiji, R., Lewandowski, R.: Critical Sobolev exponent and the dimension three. Houston
J. Math. 18(2), 189–204 (1992)
11. Cordero-Erausquin, D., Nazaret, B., Villani, C.: A mass-transportation approach to sharp Sobolev and
Gagliardo–Nirenberg inequalities. Adv. Math. 182(2), 307–332 (2004)
12. Furtado, M.F., Souza, B.N.: Positive and nodal solutions for an elliptic equation with critical growth.
Commun. Contemp. Math. 18(2), 1550021 (2016)
13. Hadiji, R.: Solutions positives de l’équation −Δu = u p + μu q dans un domaine à trou. Ann. Fac. Sci.
Toulouse Math. 11(3), 55–71 (1990)
14. Hadiji, R.: A nonlinear problem with a weight and a nonvanishing boundary datum. Pure Appl. Funct.
Anal. 5(4), 965–980 (2020)
15. Hadiji, R., Vigneron, F.: Existence of solutions of a non-linear eigenvalue problem with a variable
weight. J. Diff. Equ. 266(2–3), 1488–1513 (2019)
16. Hadiji, R., Yazidi, H.: Problem with critical Sobolev exponent and with weight. Chinese Ann. Math.
Ser. B 28(3), 327–352 (2007)
17. Hadiji, R., Molle, R., Passaseo, D., Yazidi, H.: Localization of solutions for nonlinear elliptic problems
with critical growth. C. R. Acad. Sci. Paris 334(11–12), 725–730 (2006)
18. Jost, J.: Riemannian Geometry and Geometric Analysis Universitext, 2nd edn. Springer, Berlin (1998)
19. Lions, P.-L.: The concentration compactness principle in the calculus of variations the limit case. I.
Rev. Mat. Iberoamericana 1(1), 145–201 (1985)
20. Lee, J.M., Parker, T.H.: The Yamabe problem. Bull. Amer. Math. Soc. 17(1), 37–91 (1987)
21. Kholodenko, A.L.: Towards physically motivated proofs of the Poincaré and geometrization conjec-
tures. J. Geom. Phys. 58(2), 259–290 (2008)
22. Pohožaev, S.I.: On the eigen functions of the equation Δu + λ f (u) = 0. Dokl. Akad. Nauk SSSR 165,
36–39 (1965) (in Russian)

123
 54 Page 22 of 22 A. Benhamida, R. Hadiji

23. Talenti, G.: Best constants in Sobolev inequality. Ann. Mat. Pura Appl. 110, 353–372 (1976)
24. Yazidi, H.: Private communication

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional 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.

123