Ground-state solutions of a two-component elliptic system in R
4with the Sobolev critical exponent
∗Yuanze Wu
a& Wenming Zou
baSchool of Mathematics, China University of Mining and Technology, Xuzhou 221116, P. R. China
bDepartment of Mathematical Sciences, Tsinghua University, Beijing 100084, P. R. China
Abstract:In this paper, we continue our study in [17] on the following elliptic system:
−ε2∆u1+λ1u1=α1up1−1+µ1u31+βu22u1 inΩ,
−ε2∆u2+λ2u2=α2up2−1+µ2u32+βu21u2 inΩ, u1, u2>0 inΩ, u1=u2= 0 on∂Ω,
whereΩ⊂R4is a bounded domain,µ1, µ2>0andβ̸= 0are constants,λ1, λ2, ε >0 are parameters and2 < p < 2∗ = 4. By further employing some canonical rescal- ings, we prove by variational methods that the above elliptic system has a ground-state solution forallβ ≥ −√µ1µ2 ifεandmin{λ1, λ2}both sufficiently small. We al- so observe that the ground-state solution willvanishwith the rate(max{λ1, λ2})p−21 if bothε → 0 andλ1, λ2 → 0. These results, together with that in [1, 17], suggest that the subcritical termsα1up1−1, α2up2−1 have strong effects on the structure of the ground-state solutions of the above system.
Keywords:Elliptic system; Ground-state solution; Variational method; Sobolev criti- cal exponent; Asymptotic property.
Mathematical Subject Classification 2010: 35B09; 35B33; 35B40; 35J50.
1 Introduction
In this paper, we continue our study in [17] on the following elliptic system:
−ε2∆u1+λ1u1=α1up1−1+µ1u31+βu22u1 inΩ,
−ε2∆u2+λ2u2=α2up2−1+µ2u32+βu21u2 inΩ, u1, u2>0 inΩ, u1=u2= 0 on∂Ω,
(1.1)
∗E-mails: [email protected] (Y. Wu); [email protected] (W. Zou)
whereΩ⊂R4is a bounded domain,µi, αi>0andβ̸= 0are constants,λi, ε >0are parameters and2< p <2∗= 4.
System (1.1) is critical since the cubic termsu31, u32and the coupling termsu22u1, u21u2
are all of critical growth in the sense of the Sobolev embedding. Such systems have been studied extensively in the literature, see, for example, [3–5, 11–17] and the refer- ences therein. By applying the Pohozaev identity, we can see that (1.1) hasnosolution whenΩis star-shaped forλ1, λ2 >0andα1=α2= 0. In the very recent work [17], by using variational methods, we proved the existence of one spiked solution to (1.1) forε >0sufficiently small, under one of the following three cases:
(1) β >0sufficiently large.
(2) −√µ1µ2< β <0orβ >0sufficiently small.
(3) β≤ −√µ1µ2andα1, α2>0sufficiently small.
Moreover, this solution is also a ground-state solution that has the same concentration behaviors of the ones in [9] asε→0. Our results suggest that the appearances of the subcritical termsα1up1−1, α2up2−1do affect the structure of the solutions of (1.1).
On the other hand, it is well-known that (1.1) withα1 =α2 = 0hasnosolutions in one of the following two cases:
(a) λ1≤λ2andµ2≤β ≤µ1with one strictly inequality holding.
(b) λ2≤λ1andµ1≤β ≤µ2with one strictly inequality holding.
Based on our very recent work [17], it is natural to ask thatcan the subcritical terms α1up1−1, α2up2−1affect the structure of the solutions of(1.1)strongly enough such that (1.1)withα1, α2>0has a solution in the above two cases(a)and(b)? Since to the best of our knowledge, this question has not been studied yet in the literature, the main purpose of this paper is to investigate this natural question.
Let us give some words about our strategies in studying the above question. We notice that in the recent work [10], the asymptotic properties of the unique ground- state solution of the following equation,
{ −∆u+εu=|u|p−2u− |u|q−2u inRN, u∈H1(RN)∩Lq(RN),
as ε → 0+, has been studied by variational arguments, where2 < p < q. It has been observed by employing some canonical rescalings, which are associated with the lowest order nonlinear term|u|p−2u, that the asymptotic properties depends strongly on the relation between the orderpand critical Sobolev exponent2∗. Now, motivated by [10], if we introduce the following canonical rescaling,
vi(x) =
( 1
min{λ1, λ2} )p−21
ui
(
x0+ ε
√min{λ1, λ2}x )
,
to (1.1), then we will have the following system:
−∆u1+ λ1
min{λ1, λ2}u1=α1up1−1+ (min{λ1, λ2})4−pp−2(µ1u31+βu22u1) inΩ˘ε,
−∆u2+ λ2
min{λ1, λ2}u2=α2up2−1+ (min{λ1, λ2})4−pp−2(µ2u32+βu21u2) inΩ˘ε, u1, u2>0 inΩ˘ε, u1=u2= 0 on∂Ω˘ε,
wherex0∈Ωand
Ω˘ε={y∈R4| ε
√min{λ1, λ2}y+x0∈Ω}.
As in [10], by lettingmin{λ1, λ2}sufficiently small,the coupled termsβu22u1, βu22u1
and the critical termsµ1u31, , µ2u32can be regard as small perturbations, which means that the lowest order termsα1up1−1, α2up2−1are the main terms in the above system.
Thus, we can find solutions in the cases(a)and(b)under some further assumptions.
If we go further in this direction by employing another canonical rescaling,
˜ vi(x) =
( 1
max{λ1, λ2} )p−21
ui
(
x0+ ε
√max{λ1, λ2}x )
,
to (1.1), then we will have the following system:
−∆u1+ λ1
max{λ1, λ2}u1=α1up1−1+ (max{λ1, λ2})4−pp−2(µ1u31+βu22u1) inΩ˜ε,
−∆u2+ λ2
max{λ1, λ2}u2=α2up2−1+ (max{λ1, λ2})4−pp−2(µ2u32+βu21u2) inΩ˜ε, u1, u2>0 inΩ˜ε, u1=u2= 0 on∂Ω˜ε,
wherex0∈Ωand
Ω˜ε={y∈R4| ε
√max{λ1, λ2}y+x0∈Ω}.
By letting ε → 0 first and max{λ1, λ2} → 0 next, we can obtain the scalar field equations
−∆u1+ ˜λ1u1=α1up1−1 inRN,
−∆u2+ ˜λ2u2=α2up2−1 inRN, u1, u2>0 inRN, u1, u2→0 as|x| →+∞,
whereλ˜i= limmax{λ1,λ2}→0max{λλi
1,λ2}. The above formal analysis yields that˜vi(x) will converge to the unique ground-state solution of the above scalar field equation and uiwill vanish in passing to the limit. These properties seems to be quite different from those ofε→0orλ1, λ2→0, only (cf. [4, 9, 13, 17]).
Let us give some necessary notations before we state our main results. LetHλi,ε,Ω
be the Hilbert space ofH01(Ω)equipped with the inner product
⟨u, v⟩λi,ε,Ω=
∫
Ω
ε2∇u∇v+λiuvdx.
Fori= 1,2, sinceλi >0andε >0,Hλi,ε,Ωis a Hilbert space and the corresponding norm is given by∥u∥λi,ε,Ω=⟨u, u⟩λ12i,ε,Ω. SetHε,Ω=Hλ1,ε,Ω× Hλ2,ε,Ω. ThenHε,Ω
is a Hilbert space with the inner product
⟨u,v⟩ε,Ω=
∑2 i=1
⟨ui, vi⟩λi,ε,Ω.
The corresponding norm is given by∥u∥ε,Ω = ⟨u,u⟩ε,Ω12 . Here, ui, vi are the ith component ofu,v, respectively. Define a functional inHε,Ωas follows:
Jε,Ω(u) =
∑2 i=1
(1
2∥ui∥2λi,ε,Ω−αi
p∥ui∥pΩ,p)−1 4
∑2 i=1
µi∥ui∥4Ω,4−β
2∥u21u22∥Ω,1. Here,∥ · ∥Ω,pis the standard norm inLp(Ω)for allp≥1which is given by∥u∥Ω,p= (∫
Ω|u|pdx)1p.
Definition 1.1. The vector(u1, u2) = u∈ Hε,Ωis called a nontrivial critical point ofJε,Ω(v)ifJε,Ω′ (u) = 0inH−ε,Ω1 withu1 ̸≡0andu2 ̸≡ 0. u ∈ Hε,Ωis called a semi-trivial critical point ofJε,Ω(v)ifJε,Ω′ (u) = 0inH−ε,Ω1 withu1̸≡0oru2̸≡0.
Here,Jε,Ω′ (u)is the Fr´echet derivative ofJε,Ω(u)andH−ε,Ω1 is the dual space ofHε,Ω. u∈ Hε,Ωis called a positive critical point ofJε,Ω(v)ifuis a nontrivial critical point andui>0for bothi= 1,2.
Clearly,Jε,Ω(u)is of classC2inHε,Ωand the positive critical points ofJε,Ω(u) are the solutions of (1.1).
Definition 1.2. uis called a ground-state solution of (1.1)ifuis a nontrivial solution of(1.1)andJε,Ω(u)≤ Jε,Ω(v)for any nontrivial solutionv.
Now, our main results can be stated as follows.
Theorem 1.1. Letβ ≥ −√µ1µ2 andλi, αi, µi > 0. LetD0 > 0 be an absolute constant. Then we have the following.
(1) There exists λ∗ > 0only dependent on D0,µ1, µ2 andβ such that(1.1)has a ground-state solution(u1,ε, u2,ε)for0 < min{λ1, λ2} < λ∗and 0 < ε <
D0√
min{λ1, λ2}.
(2) Letεn, λn1, λn2 →0+and(un1, un2)be the related solution obtained in(1). Sup- poseA= limn→∞√ εn
max{λn1,λn2} <+∞, then
uni(xni) =O((max{λn1, λn2})p−21 )
and
vni(x) =
( 1
max{λn1, λn2} )p−12
uni (
xni +√ εn
max{λn1, λn2}x )
converges to somevistrongly inH1(R4)asn→ ∞, whereviis the solution of the following equation
{ −∆u+λ∗iu=αi|u|p−2u inΩ∗i,
u∈H01(Ω∗i). (1.2)
Herexni is, respectively, the maximum point ofuni,Ω∗i is the limit of Ωi,n={y∈R4| εn
√max{λn1, λn2}y+xni ∈Ω}
asn→ ∞andλ∗i = limn→∞ λni
max{λn1,λn2}. Moreover, (i) IfA = limn→∞√ εn
max{λn1,λn2} > 0, thenΩ∗i are bounded and vi are a ground-state solution of (1.2), respectively.
(ii) IfA = limn→∞√ εn
max{λn1,λn2} = 0, thenΩ∗i = R4and vi is the unique radially ground-state solution of (1.2)forλ∗i >0whilevi≡0forλ∗i = 0.
Furthermore,
√max{λn1,λn2}|xn1−xn2|
εn →0asn→ ∞.
Remark 1.1. (a) By introducing the canonical re-scaling given by(2.2), we can re- scale system(1.1)into(2.3). In such system, the cubic nonlinearities and the coupled term can be treated as the perturbation terms forλ1small enough and ε ≪ √
λ1. Here, without loss of generality, we assumeλ1 = min{λ1, λ2}. Thus, under this idea, we can prove by the method of the Nehari manifold that system(1.1)has a ground-state solution for allβ ≥ −√µ1µ2under the condi- tions0< λ1< λ∗and0< ε < D0
√λ1, which is stated by(1)of Theorem 1.1.
(b) (2)of Theorem 1.1 yields that the scalar field equation(1.2)is the limit equation of system(1.1)if bothεn→0andmax{λn1, λn2} →0with
A= lim
n→∞
εn
√max{λn1, λn2} <+∞.
Moreover, the ground-state solutions will vanish at the rate of(max{λn1, λn2})p−21 . To the best of our knowledge, this is the first result about such asymptotic prop- erty of system(1.1).
This paper is organized as follows. In Section 2, we obtain an existence result of (1.1) by applying the method of the Nehari manifold. In Section 3, we study the concentration behavior of the ground state solution asε→0+andλ1, λ2→0+.
2 The existence results
2.1 Some preliminaries
In this section,we are interesting in finding a ground state solution of (1.1) for allβ ≥
−√µ1µ2. Without loss of generality, we assume that0∈Ω,λ1≤λ2andD0= 1. Let Ωε={x∈R4|εx∈Ω}.
Then it is easy to see thatΩε → R4 as ε → 0+. Moreover, it is easy to see that u= (u1, u2)is a solution of (1.1) if and only ifbu= (ub1,ub2)withbui(x) =ui(εx)is a solution of the following system
−∆bu1+λ1bu1=µ1bu31+α1ubp1−1+βbu22bu1 inΩε,
−∆bu2+λ2bu2=µ2bu32+α2ubp2−1+βbu21bu2 inΩε, b
u1,bu2>0 inΩ, ub1=ub2= 0 on∂Ωε,
(2.1)
We also define a canonical rescaling ofuband denote it byu. That is,e e
ui(x) = (1
λ1
)p−21 b ui
( x
√λ1
)
. (2.2)
Clearly, for everyui∈ Hλi,ε,Ω, we haveeui∈ Hλi
λ1,1,eΩε, where Ωeε={x∈R4| ε
√λ1x∈Ω}.
Moreover,ubis a solution of (2.1) if and only ifueis a solution of the following system
−∆eu1+ue1=α1eup1−1+ (λ1)4−pp−2(µ1ue31+βue22ue1) inΩeε,
−∆eu2+λ2 λ1
e
u2=α2eup2−1+ (λ1)4−pp−2(µ2ue32+βue21ue2) inΩeε, e
u1,ue2>0 inΩeε, eu1=eu2= 0 onΩeε,
(2.3)
By a direct calculation, we have
∫
Ωeε
|∇eui|2dx= (λ1)1−p−22
∫
Ωε
|∇bui|2dx= (λ1)1−p−22 ε−2
∫
Ω
|∇ui|2dx (2.4) and ∫
e Ωε
|eui|rdx= (λ1)2−p−2r
∫
Ωε
|bui|rdx= (λ1)2−p−2r ε−4
∫
Ω
|ui|rdx (2.5) for bothi= 1,2and all2≤r≤4. Let
EΩeε(eu) =
∑2 i=1
(1
2(∥∇eui∥2Ωe
ε,2+λi
λ1∥eui∥2Ωe
ε,2)−αi
p∥eui∥pΩe
ε,p))
−(λ1)4−pp−2(1 4
∑2 i=1
µi∥eui∥4Ωe
ε,4+β
2∥eu21eu22∥Ωeε,1).
Then it is easy to see thatEΩeε(eu)is of classC2 inH∗,eΩε = Hλ1
λ1,1,eΩε × Hλ2 λ1,1,eΩε
and positive critical points ofEeΩε(u)e are equivalent to solutions of (2.3). Moreover, by (2.4) and (2.5), we have
EeΩε(u) = (λe 1)1−p−22 J1,Ωε(bu) = (λ1)1−p−22 ε−4Jε,Ω(u). (2.6) Definition 2.1. ueis called a ground state solution of(2.3)ifueis a nontrivial solution of(2.3)andEeΩε(u)e ≤ EΩeε(ev)for any nontrivial solutionv.e
By (2.6) and Definition 1.2, ifwe is a ground state solution of (2.3), thenw is a ground state solution of (1.1), wherew= (w1, w2)withwi(x) =λ
1 p−2
1 wei(√ελ1x).
2.2 The Nehari manifold
Define the Nehari manifold ofEΩeε(u)e as follows:
NΩeε ={eu∈He∗,eΩε | EΩ′∗j,n(u)uk= 0, k= 1,2}, whereHe∗,eΩε = (Hλ1
λ1,1,eΩε\{0})×(Hλ2
λ1,1,eΩε\{0}),u1 = (u1,0)andu2 = (0, u2).
Clearly, all nontrivial solutions of (2.3) are contained inNeΩε. Thus, ifwe ∈ NΩeεattains the minimum ofEΩeε(u)e inNΩeεandwe is also a positive critical point ofEΩeε(u), thene
e
w is a ground state solution of (2.3). Here, we saywe is a positive critical point of EeΩε(u), ife we is a critical point ofEΩeε(u)e andwe = (we1,we2)withwei>0.
Lemma 2.1. Letβ≥ −√µ1µ2. ThenEΩeε(eu)≥0onNΩeε. Proof. Sinceβ ≥ −√µ1µ2, the functional
∑2 i=1
µi∥euni∥4Ωe
ε,4+ 2β∥(euni)2(euni)2∥Ωeε,1
is nonnegative definite inH∗,eΩεby the H¨older inequality. Thus, the conclusion follows immediately from2< p <4and a standard calculation.
By Lemma 2.1,
cε= inf
NΩeε
EΩeε(eu) (2.7)
is well defined.
Lemma 2.2. Let β ≥ −√µ1µ2. Then for 0 < ε < √
λ1, there exists C0 > 0 independent ofεandλ1such thatcε≤C0.
Proof. Since √ελ1 ≤1, we can see thatBC∗
2 (0)⊂Ωeε, whereC∗ =dist(0, ∂Ω). Let Ω1 andΩ2be two bounded domains withΩ1∩Ω2 = ∅andΩi ⊂ BC∗
2 (0)for both i= 1,2. Then it is well known that the following equation
{−∆u+λ2u=αi|u|p−2u inΩi,
u= 0 on∂Ωi
has a positive ground state solutionui. Denote the energy level ofuibyCi, thenCiis independent ofεandλ1. LetC0 =C1+C2, then byΩ1∩Ω2 =∅andλ1≤λ2, the conclusion immediately follows from (2.7).
By the Ekeland’s variational principle, there exists{eun} ⊂ NeΩεsuch that (1) EΩeε(eun) =cε+on(1),
(2) EΩeε(ev)≥ EΩeε(eun)−n1∥ev−uen∥ε,Ωfor allev∈ NΩeε.
Moreover, by settingueni ≡0outsideΩeε, we can regarduen= (eun1,uen2)∈D1,2(R4)× D1,2(R4).
Lemma 2.3. Letβ≥ −√µ1µ2. Then there existsλ∗>0such that
∥∇euni∥2R4,2+∥euni∥2R4,2≤C withnlarge enough for0< λ1 < λ∗and0 < ε <√
λ1, whereC >0is a constant independent ofεandλ1. Moreover,{eun} ⊂ NeΩεis also a(P S)sequence ofEΩeε(eun) at the energy levelcε.
Proof. By Lemma 2.2, we can see fromβ ≥ −√µ1µ2and2< p <4that C0+on(1) ≥ EΩeε(eun)−1
pEΩe′
ε(uen)eun
≥ p−2 2p
∑2 i=1
(∥∇euni∥2Ωe
ε,2+λi
λ1∥ueni∥2Ωe
ε,2).
Thus, we have{euni}is bounded inH1(R4)forn. Moreover, byλ1≤λ2, we also have
∥∇euni∥2R4,2+∥euni∥2R4,2≤2p(C0+ 1)
p−2 (2.8)
forn large enough, where 2p(Cp−02+1) > 0 is a constant independent ofε andλ1. It remains to show that{eun} ⊂ NΩeε is also a(P S)sequence ofEΩeε(eun)at the energy levelcε. Letwe ∈ H∗,eΩε. For everyn ∈ N, we consider the system Ψn(t, l) = 0, whereΨn(t, l) = (Ψn1(t, l),Ψn2(t, l))with
Ψni(t, l) = ∥∇(tieuni +lwei)∥2Ωe
ε,2+ λi
λ1∥tiueni +lwei∥2Ωe
ε,2−αi∥tiueni +lwei∥pΩe
ε,p
−λ
4−pp−2
1 (µi∥tieuni +lwei∥4eΩ
ε,4+β∥(t1uen1+lwe1)2(t2eun2 +lwe2)2∥Ωeε,1).
Clearly,Ψ(t, l)is ofC1. Moreover, since{eun} ⊂ NΩeε, we also have thatΨn(1,0) = 0. By a direct calculation, we have
∂Ψni(1,0)
∂ti
= 2(∥∇euni∥2Ωe
ε,2+λi
λ1∥euni∥2Ωe
ε,2)−pαi∥euni∥peΩ
ε,p
−λ
4−p p−2
1 (4µi∥euni∥4Ωe
ε,4+ 2β∥(eun1)2(eun2)2∥Ωeε,1)
= −(p−2)αi∥euni∥pΩe
ε,p−2µiλ
4−pp−2
1 ∥euni∥4Ωe
ε,4 (2.9)
respectively fori= 1,2and
∂Ψn2(1,0)
∂t1
= ∂Ψn1(1,0)
∂t2
= −2βλ
4−pp−2
1 ∥(ueni)2(ueni)2∥Ωeε,1 (2.10) Set
Θn = (θijn)i,j=1,2
withθnij =∂Ψni∂t(1,0)
j . By (2.8), we can see from the Sobolev inequality that
∥euni∥2R4,4≤2p(C0+ 1)
(p−2)S (2.11)
for nlarge enough, where S is best embedding constant from H1(R4) → L4(R4) defined by
S= inf{∥∇u∥2R4,2|u∈H1(R4),∥u∥2R4,4= 1}. (2.11) together with{un} ⊂ Nε,Ωand the H¨older inequality, implies
∥∇euni∥2Ωe
ε,2+ λi
λ1∥euni∥2Ωe
ε,2
= αi∥euni∥pΩe
ε,p+λ
4−pp−2
1 (µi∥euni∥4Ωe
ε,4+β∥(ueni)2(ueni)2∥Ωeε,1)
≤ αi∥euni∥pΩe
ε,p+λ
4−p p−2
1
2p(C0+ 1)
(p−2)S (µi+|β|)∥euni∥2eΩ
ε,4 (2.12)
fornlarge enough. Let λ∗=
( p−2
4p(C0+ 1)(µ1+µ2+|β|) )p−24−p
.
Then for0< λ1< λ∗, we have from (2.12) that
∥∇euni∥2Ωe
ε,2+ λi
λ1∥euni∥2Ωe
ε,2≤2αi∥euni∥pΩe
ε,p
fornlarge enough, which together withλ1≤λ2, implies αi∥euni∥pΩe
ε,p≥ (Sp
2 )p−2p
(2.13) forn large enough with0 < λ1 < λ∗. Here, Sp is best embedding constant from H1(R4)→Lp(R4)defined by
Sp= inf{∥∇u∥2R4,2+∥u∥2R4,2|u∈H1(R4),∥u∥pR4,p= 1}.
By2< p <4, we have from (2.9)–(2.10) that det(Θn)≥(p−2)2α1α2
(Sp
2 )p−22p
>0 (2.14)
fornlarge enough. Now, we can applying the implicit function theorem and the Tay- lor’s expansion in a standard way (cf. [2]) to show that{eun} ⊂ NeΩε is also a(P S) sequence ofEΩeε(uen)at the energy levelcεfor0< λ1< λ∗.
By Lemma 2.3,{euni}is bounded inH1(R4)forn. Without loss of generality, we assume thateuni ⇀ue∗i weakly inH1(R4)asn→ ∞.
Proposition 2.1. Letβ ≥ −√µ1µ2,0 < λ1 < λ∗and 0 < ε < √
λ1. Then there existsUe = (Ue1,Ue2)withUei>0inΩeεfor bothi= 1,2such thatUe is a critical point ofEΩeε(eu)at the energy levelcε.
Proof. Clearly, eu∗ = (ue∗1,ue∗2) is a critical point ofEΩeε(eu). Moreover, since Ωeε is bounded, we can see from the Sobolev embedding theorem and (2.13) thatueni → eu∗i strongly inLr(eΩε)for1≤r <4asn→ ∞andeu∗i ̸≡0for bothi= 1,2. Thus, we must haveue∗∈ NΩeε. It follows that
cε+on(1) = EΩeε(uen)−1 4EΩ′e
ε(eun)uen
= 1 4
∑2 i=1
(∥∇euni∥2Ωe
ε,2+ λi
λ1∥euni∥2Ωe
ε,2)−4−p 4p
∑2 i=1
αi∥euni∥pΩe
ε,p
≥ 1 4
∑2 i=1
∥∇eu∗i∥2Ωe
ε,2+λi
λ1∥eu∗i∥2Ωe
ε,2−4−p 4p
∑2 i=1
αi∥eu∗i∥pΩe
ε,p+on(1)
= EΩeε(ue∗)−1 4EeΩ′
ε(ue∗)eu∗+on(1)
≥ cε+on(1).
Hence,EΩeε(ue∗) = cε. That is,ue∗ is a minimum point ofEeΩε(u)e onNΩeε. Note that Ue = (Ue1,Ue2)withUei=|eu∗i|is also a minimum point ofEΩeε(u)e onNΩeε. Now, by a similar argument as used for (2.14), we can apply the method of Lagrange multipliers in a standard way (cf. [2]) to show thatNΩeεis a natural constraint inH∗,eΩε. Therefore, by the maximum principle,Ue is a critical point ofEΩeε(eu)at the energy levelcεwith Uei>0for bothi= 1,2.
3 Concentration behaviors
In this section, we are interesting in studying the concentration behavior of the ground state solution of (1.1) as ε → 0+ and λ = (λ1, λ2) → 0. For this purpose, we takeεn → 0+andλn = (λn1, λn2) → 0as n → ∞. Without loss of generality, we