construction of solutions to a nonlinear critical elliptic system via local pohozaev identities

Qidong Guo, Qingfang Wang^† and Wenju Wu School of Mathematics and Statistics, Central China Normal University, Wuhan, 430079, P. R. China [email protected] School of Mathematics and Computer Science, Wuhan Polytechnic University, Wuhan, 430079, P. R. China [email protected] School of Mathematics and Statistics, Central China Normal University, Wuhan, 430079, P. R. China [email protected]

Abstract.

In this paper, we investigate the following elliptic system with Sobolev critical growth

\begin{cases}-\Delta u+P(|y^{\prime}|,y^{\prime\prime})u=u^{2^{*}-1}+\frac{% \beta}{2}u^{\frac{2^{*}}{2}-1}v^{\frac{2^{*}}{2}},\ y\in\mathbb{R}^{N},\vspace% {0.12cm}\\ -\Delta v+Q(|y^{\prime}|,y^{\prime\prime})v=v^{2^{*}-1}+\frac{\beta}{2}v^{% \frac{2^{*}}{2}-1}u^{\frac{2^{*}}{2}},\ y\in\mathbb{R}^{N},\vspace{0.12cm}\\ u,\,v>0,\,u,\,\,v\in H^{1}(\mathbb{R}^{N}),\end{cases}

where $(y^{\prime},y^{\prime\prime})\in\mathbb{R}^{2}\times\mathbb{R}^{N-2}$ , $P(|y^{\prime}|,y^{\prime\prime}),Q(|y^{\prime}|,y^{\prime\prime})$ are bounded non-negative function in $\mathbb{R}^{+}\times\mathbb{R}^{N-2}$ , $2^{*}=\frac{2N}{N-2}$ . By combining a finite reduction argument and local Pohozaev type of identities, assuming that $N\geq 5$ and $r^{2}(P(r,y^{\prime\prime})+\kappa^{2}Q(r,y^{\prime\prime}))$ have a common topologically nontrivial critical point, we construct an unbounded sequence of non-radial positive vector solutions of synchronized type, whose energy can be made arbitrarily large. Our result extends the result of a single critical problem by [Peng, Wang and Yan,J. Funct. Anal. 274: 2606-2633, 2018]. The novelties mainly include the following two aspects. On one hand, when $N\geq 5$ , the coupling exponent $\frac{2}{N-2}<1$ , which creates a great trouble for us to apply the perturbation argument directly. This constitutes the main difficulty different between the coupling system and a single equation. On the other hand, the weaker symmetry conditions of $P(y)$ and $Q(y)$ make us not estimate directly the corresponding derivatives of the reduced functional in locating the concentration points of the solutions, we employ some local Pohozaev identities to locate them.

Key Words: critical elliptic systems; local Pohozeav identities; Lyapunov-Schmidt reduction; Synchronized vector solutions.

AMS Subject Classification: 35B33, 35J10, 35J60, 35B38.

^† Corresponding author: Qingfang Wang

1. Introduction and main results

In this paper, we are concerned with the multiplicity of solutions for the following nonlinear critical Schrödinger system

\begin{cases}-\Delta u+P(|y^{\prime}|,y^{\prime\prime})u=u^{2^{*}-1}+\frac{% \beta}{2}u^{\frac{2^{*}}{2}-1}v^{\frac{2^{*}}{2}},\ y\in\mathbb{R}^{N},\vspace% {0.12cm}\\ -\Delta v+Q(|y^{\prime}|,y^{\prime\prime})v=v^{2^{*}-1}+\frac{\beta}{2}v^{% \frac{2^{*}}{2}-1}u^{\frac{2^{*}}{2}},\ y\in\mathbb{R}^{N},\vspace{0.12cm}\\ u,\,v>0,\,u,\,\,v\in H^{1}(\mathbb{R}^{N}),\end{cases}

(1.1)

where $P(|y^{\prime}|,y^{\prime\prime}),Q(|y^{\prime}|,y^{\prime\prime})$ are bounded non-negative function in $\mathbb{R}^{+}\times\mathbb{R}^{N-2}$ , $2^{*}=\frac{2N}{N-2}$ , $\beta$ is a coupling constant. For the understanding of (1.1), let us begin with the single equation in $\mathbb{R}^{N}$ . It is related to the following well-known Brezis-Nirenberg problem in $S^{N}$

\displaystyle-\Delta_{S^{N}}u=u^{\frac{N+2}{N-2}}+\lambda u,u>0\,\,\,\,\hbox{% on}\,S^{N}.

(1.2)

Indeed, after using the stereographic projection, problem (1.2) can be reduced to

\displaystyle-\Delta u+V(y)u=u^{\frac{N+2}{N-2}},\,\,u>0,\,\,u\in H^{1}(% \mathbb{R}^{N})

with $V(y)=\frac{-4\lambda-N(N-2)}{(1+|y|^{2})^{2}}$ and $V(y)>0$ if $\lambda<-\frac{N(N-2)}{4}$ . In the pioneer work, Chen, Wei and Yan [6] applied the reduction argument to consider the Schrödinger equation with the critical exponent

\displaystyle-\Delta u+V(|x|)u=u^{\frac{N+2}{N-2}},\,\,\,\hbox{in}\,\,\,% \mathbb{R}^{N}.

If $V$ is radially symmetric, $r^{2}V(r)$ has a local maximum point or a local minimum point $r_{0}>0$ with $V(r_{0})>0$ , they proved the existence of infinitely many positive solutions. Do the same thing for the following equation

\displaystyle-\Delta u+V(|y^{\prime}|,y^{\prime\prime})u=u^{\frac{N+2}{N-2}},% \,\,\,\,\,\,u\in H^{1}(\mathbb{R}^{N}),

where $(y^{\prime},y^{\prime\prime})\in\mathbb{R}^{2}\times\mathbb{R}^{N-2}$ , $V(|y^{\prime}|,y^{\prime\prime})>0$ is bounded. By using the Pohozaev identities, Peng, Wang and Yan [23] overcome the difficulty appearing in using the standard reduction method to find algebraic equations which determine the location of bubbles. Moreover, He, Wang and Wang combining the reduction method and local Pohozaev identities proved non-degeneracy of multi-bubbling solutions of [14]. On the other hand, F. Du etc. [11] consider the multi-piece of bubble solutions for a nonlinear critical elliptic equation.

For the system, Peng and Wang discuss the following system in [25]

\begin{cases}-\Delta u+P(|x|)u=\mu u^{3}+\beta uv^{2},\,\,\,\hbox{in}\,\,\,% \mathbb{R}^{3},\vspace{0.12cm}\cr-\Delta v+Q(|x|)v=\nu v^{3}+\beta u^{2}v,\,\,% \,\hbox{in}\,\,\,\mathbb{R}^{3},\end{cases}

(1.3)

where $P(r),\,Q(r)$ are positive radial potential, $\mu>0,\,\nu>0$ . They examined the effect of nonlinear coupling on the solution structure. The authors use a finite dimensional reduction argument and found the location of bubbles by using a maximization procedure. They constructed an unbounded sequence of non-radial positive vector solutions of segregated type and an unbounded sequence of non-radial positive vector solution of synchronized type. The work of [18] gives solutions with one component peaking at the origin and the other having a finite number of peaks on a k-polygon. On the other hand, for the $\beta>0$ , $P=Q=1$ , it is known (for example [3, 2]) that there are special positive solutions with the two components being positive constant multiples of the unique positive solution of the scalar cubic Schrödinger equation $-\Delta w+w=w^{3},\,\,x\in\mathbb{R}^{3}$ . For the elliptic system of subcritical exponent, readers can refer to [15, 19, 26, 32, 33].

Systems of nonlinear Schrödinger equations with critical exponent have been the subject of extensive mathematical studies in recent years, for example, [1, 7, 17, 16, 20, 27, 28, 29, 36]. In [13], Guo, Li and Wei proved the existence of infinitely many positive non-radial solutions of the coupled elliptic system. Also, Chen, Medina and Pistoia [4] consider the segregated solutions for a critical elliptic system with a small interspecies repulsive force in $\mathbb{R}^{4}$ . More recently, Del Pino, Musso Pacard and Pistoia [9, 10] constructed sequences of sign-changing solutions with large energy and concentrating along some special sub-manifold of $\mathbb{R}^{N}$ . However, so far, the existing results of using the Lyapunov-Schmidt reduction argument to construct concentrated solutions of an elliptic system have been basically about low dimensional cases with $N<5$ , such as [5, 4, 13, 25, 24, 31], etc. Especially in [21], Peng, Peng and Wang consider the Dirichlet problem with Sobolev critical exponent

\begin{cases}-\Delta u=|u|^{2^{*}-2}u+\frac{\alpha}{2^{*}}|u|^{\alpha-2}u|v|^{% \beta},\,\,\,\,\hbox{in}\,\,\,\Omega,\vspace{0.12cm}\cr-\Delta v=|v|^{2^{*}-2}% v+\frac{\beta}{2^{*}}|u|^{\alpha}|v|^{\beta-2}v,\,\,\,\,\hbox{in}\,\,\,\Omega,% \end{cases}

(1.4)

where $\alpha+\beta=2^{*}=\frac{2N}{N-2}$ and $\Omega=\mathbb{R}^{N}$ or $\Omega$ is a smooth bounded domain in $\mathbb{R}^{N}$ . They obtain a uniqueness result on the least energy solutions and show that a manifold of the synchronized type of positive solutions is non-degenerate for the above system for some ranges of the parameters $\alpha,\beta,N$ .

Motivated by [12, 23], in this paper, we mainly want to study the existence of infinitely many solutions of system (1.1) with weaker symmetry of potentials. We would like to point out that very recently in [12] using the finite dimensional reduction method obtained infinitely many non-radial solutions for system (1.1) with radial potentials. Also, about the single critical Schrodinger equation $-\Delta u+V(y)u=u^{\frac{N+2}{N-2}},y\in\mathbb{R}^{N}$ involving with a potential linear term, there are some existence of solutions for it, one can refer [6, 11, 30], in particular, in [23] Peng, Wang and Yan first applied the finite dimensional reduction combining the local Pohozaev identities obtained infinitely many solutions of (1.1).

In this paper, we assume that $P(y),\,Q(y)$ satisfy the following assumptions:

$(i)$

The function $r^{2}(P(r,y^{\prime\prime})+\kappa^{2}Q(r,y^{\prime\prime}))$ have a common critical point $(r_{0},y_{0}^{\prime\prime})$ such that $r_{0}>0$ and $P(r_{0},y_{0}^{\prime\prime})>0$ and $Q(r_{0},y_{0}^{\prime\prime})>0$ ;
$(ii)$

$deg\big{(}\nabla(r^{2}P(r,y^{\prime\prime})+\kappa^{2}r^{2}Q(r,y^{\prime\prime% })),(r_{0},y_{0}^{\prime\prime})\big{)}\neq 0$ ;
$(iii)$

$\beta$ satisfies $1+\beta\kappa^{\frac{2^{*}}{2}}>0,$

where $\kappa$ satisfies

2+\kappa^{\frac{2^{*}}{2}}-\beta\kappa^{\frac{2^{*}}{2}-2}-2\kappa^{2^{*}-2}=0.

(1.5)

Our main result in this paper can be stated as follows.

Theorem 1.1.

Suppose $N\geq 5$ and $P\geq 0,\,Q\geq 0$ are bounded and belong to $C^{1}$ . If $P(r,y^{\prime\prime}),\,Q(r,y^{\prime\prime})$ satisfies $(i)$ and $(ii)$ , then problem (1.1) has infinitely many non-radial positive solutions whose energy can be made arbitrarily large.

Remark 1.2.

It is worth noting that in the higher dimensional space with $N\geq 5$ , the exponent $\frac{2}{N-2}<1$ from the worse couple terms will create a great trouble for us to apply the perturbation argument directly. To overcome this difficulty, we should ensure that the error term can be controlled by a small multiple of the approximate solution (see Lemma A.6).

Remark 1.3.

We would like to point out that the weaker symmetry conditions of $P(y)$ and $Q(y)$ make us not estimate directly the corresponding derivatives of the reduced functional in locating the concentration points of the solutions, we employ some local Pohozaev identities to locate them. Moreover, compared with [12], the system we consider here involving a linear term which causes the problem more difficult.

Remark 1.4.

The non-radial vector solutions of separated type for system (1.1) is also very important but more challenging and we will consider this topic in a future work.

It is well known that the only solutions of the problem

\displaystyle-\Delta w=w^{\frac{N+2}{N-2}},\,\,\,w>0\,\,\,\hbox{in}\,\,\,% \mathbb{R}^{N}

are of the form

\displaystyle w_{x,\lambda}(y):=\lambda^{\frac{N-2}{2}}w(\lambda(y-x))=(N(N-2)% )^{\frac{N-2}{4}}\Bigl{(}\frac{\lambda}{1+\lambda^{2}|y-x|^{2}}\Bigr{)}^{\frac% {N-2}{2}},\ \ \lambda>0,\ x\in\mathbb{R}^{N}.

Moreover, by [21], then $N+1$ dimensional manifold

\displaystyle\{(U_{x_{0},\lambda},V_{x_{0},\lambda})=(sw_{x_{0},\lambda},tw_{x% _{0},\lambda}),t=\kappa s,x_{0}\in\mathbb{R}^{N},\lambda>0\}

is a non-degenerate solution of the following equations

\begin{cases}-\Delta u=u^{2^{*}-1}+\frac{\beta}{2}u^{\frac{2^{*}}{2}-1}v^{% \frac{2^{*}}{2}},\ y\in\mathbb{R}^{N},\vspace{0.12cm}\\ -\Delta v=v^{2^{*}-1}+\frac{\beta}{2}v^{\frac{2^{*}}{2}-1}u^{\frac{2^{*}}{2}},% \ y\in\mathbb{R}^{N},\vspace{0.12cm}\\ u,\,v>0,\,u,\,\,v\in D^{1,2}(\mathbb{R}^{N}),\end{cases}

(1.6)

where $s$ satisfies that $s^{2^{*}-2}=\frac{2}{2+\beta\kappa^{\frac{2^{*}}{2}}}$ and $\kappa$ satisfies (1.5).

Let $y=(y^{\prime},y^{\prime\prime}),y^{\prime}\in\mathbb{R}^{2},y^{\prime\prime}% \in\mathbb{R}^{N-2}$ . We define

	$\displaystyle H_{s}=\Bigl{\{}u:u\in D^{1,2}(\mathbb{R}^{N}),$	$\displaystyle u\ \hbox{is\ even\ in}\ y_{h},h=2,\cdots,N,$
		$\displaystyle u(r\cos\theta,rsin\theta,y^{\prime\prime})=u(r\cos(\theta+\frac{% 2\pi j}{k}),r\sin(\theta+\frac{2\pi j}{k}),y^{\prime\prime})\Bigr{\}}.$

For any large integer $k>0$ , let

x_{j}=\Bigl{(}r\cos\frac{2(j-1)\pi}{k},r\sin\frac{2(j-1)\pi}{k},\mathbf{0}% \Bigr{)},\ \ j=1,\ldots,k,

where $\mathbf{0}$ is the zero vector in $\mathbb{R}^{N-2}$ .

Let $\delta>0$ is a small constant, such that

\displaystyle r^{2}P(r,y^{\prime\prime})>0,\,\,\,r^{2}Q(r,y^{\prime\prime})>0,% \,\,\,\hbox{if}\,\,\,|(r,y^{\prime\prime})-(r_{0},y_{0}^{\prime\prime})|\leq 1% 0\delta.

(1.7)

To deal with the slow decay of this function when $N$ is not big, we need to cut off this function. Let $\xi(y)=\xi(|y^{\prime}|,y^{\prime\prime})$ be a smooth function satisfying $\xi=1$ if $|(r,y^{\prime\prime})-(r_{0},y_{0}^{\prime\prime})|\leq\delta$ ; $\xi=0$ if $|(r,y^{\prime\prime})-(r_{0},y_{0}^{\prime\prime})|\geq 2\delta$ and $0\leq\xi\leq 1$ . Denote

\displaystyle W_{1,x_{j}}=\xi U_{x_{j},\lambda},\,\,\,W_{1,\lambda}^{*}=\sum% \limits_{j=1}^{k}U_{x_{j},\lambda},\,\,\,W_{1,\lambda}=\sum\limits_{j=1}^{k}W_% {1,x_{j}},

\displaystyle W_{2,x_{j}}=\xi V_{x_{j},\lambda},\,\,\,W_{2,\lambda}^{*}=\sum% \limits_{j=1}^{k}V_{x_{j},\lambda},\,\,\,W_{2,\lambda}=\sum\limits_{j=1}^{k}W_% {2,x_{j}}.

In this paper, we always assume that $k>0$ is a large integer, $\lambda\in\Bigl{[}L_{0}k^{\frac{N-2}{N-4}},L_{1}k^{\frac{N-2}{N-4}}\Bigr{]}$ for some constant $L_{1}>L_{0}>0$ . We will prove Theorem 1.1 by verifying the following result.

Theorem 1.5.

Under the assumptions of Theorem 1.1, there exists a positive integer $k_{0}>0$ such that for any integer $k\geq k_{0}$ , (1.1) has a solution $(u_{k},v_{k})$ of the form

\displaystyle u_{k}=W_{1,\lambda_{k}}+\varphi_{k}=\sum\limits_{j=1}^{k}\xi U_{% x_{j},\lambda_{k}}+\varphi_{k},\,\,\,v_{k}=W_{2,\lambda_{k}}+\psi_{k}=\sum% \limits_{j=1}^{k}\xi V_{x_{j},\lambda_{k}}+\psi_{k},

where $(\varphi_{k},\psi_{k})\in H_{s}\times H_{s}$ . Moreover, as $k\rightarrow+\infty$ , $\lambda_{k}\in\Bigl{[}L_{0}k^{\frac{N-2}{N-4}},L_{1}k^{\frac{N-2}{N-4}}\Bigr{]}$ , $(\bar{r}_{k},\bar{y}^{\prime\prime}_{k})\rightarrow(r_{0},y_{0}^{\prime\prime})$ and $\lambda_{k}^{-\frac{N-2}{2}}\|(\varphi_{k},\psi_{k})\|_{L^{\infty}}\to 0$ .

Before we close this section, let us explain the main ideas in the proof of Theorem 1.5. We will use a finite reduction argument in the proof. To deal with the large number of bubbles in the solution, similar as [34, 35], the reduction procedure is carried out in a weighted space instead of the standard Sobolev space. Since the critical points including the saddle points of the potentials, which makes us not estimate directly the corresponding derivatives of the reduced functional in locating the concentration points of the solutions, motivated by [23] we employ some local Pohozaev identities to locate them.

This paper is organized as follows. In section 2, we carry out the reduction procedure with weighted maximum norm to a finite dimensional setting. In section 3, we solve the corresponding finite dimensional probem and prove the main result Theorem 1.1. Some delicate estimates and tools are put in the appendix.

2. Finite-dimensional reduction

In this section, we perform a finite-dimensional reduction. For simplicity, we write $2^{*}=\frac{2N}{N-2}$ . Set

\displaystyle\|u\|_{*}=\sup_{y\in\mathbb{R}^{N}}\Big{(}\sum_{j=1}^{k}\frac{1}{% (1+\lambda|y-x_{j}|)^{\frac{N-2}{2}+\tau}}\Big{)}^{-1}\lambda^{-\frac{N-2}{2}}% |u(y)|

and

\displaystyle\|f\|_{**}=\sup_{y\in\mathbb{R}^{N}}\Big{(}\sum_{j=1}^{k}\frac{1}% {(1+\lambda|y-x_{j}|)^{\frac{N+2}{2}+\tau}}\Big{)}^{-1}\lambda^{-\frac{N+2}{2}% }|f(y)|,

where $\tau=\frac{N-4}{N-2}$ .

We also denote that $\|(u,v)\|_{*}=\|u\|_{*}+\|v\|_{*}$ and $\|(f,g)\|_{**}=\|f\|_{**}+\|g\|_{**}.$

Let

	$\displaystyle Y_{j,1}=\frac{\partial W_{1,x_{j}}}{\partial r},\,\,\,Y_{j,2}=% \frac{\partial W_{1,x_{j}}}{\partial\lambda},\,\,\,Y_{j,k}=\frac{\partial W_{1% ,x_{j}}}{\partial\bar{y}^{\prime\prime}_{k}},\,\,\,k=3,\cdots,N,$
	$\displaystyle Z_{j,1}=\frac{\partial W_{2,x_{j}}}{\partial r},\,\,\,Z_{j,2}=% \frac{\partial W_{2,x_{j}}}{\partial\lambda},\,\,\,Z_{j,k}=\frac{\partial W_{2% ,x_{j}}}{\partial\bar{y}^{\prime\prime}_{k}},\,\,\,k=3,\cdots,N.$

First we consider the linear problem

\displaystyle\begin{cases}L_{k}(\varphi,\psi)=(h,g)+\sum\limits_{i=1}^{N}c_{i}% \sum\limits_{j=1}^{k}(W_{1,x_{j}}^{2^{*}-2}Y_{j,i},W_{2,x_{j}}^{2^{*}-2}Z_{j,i% }),\\ (\varphi,\psi)\in H_{s}\times H_{s},\vspace{0.12cm}\\ \Bigl{\langle}(W_{1,x_{j}}^{2^{*}-2}Y_{j,l},W_{2,x_{j}}^{2^{*}-2}Z_{j,l}),(% \varphi,\psi)\Bigr{\rangle}=0,\ \ j=1,\cdots,k,\ l=1,2,\cdots,N\end{cases}

(2.1)

for some number $c_{i}$ , where $\langle(u_{1},u_{2}),(v_{1},v_{2})\rangle=\displaystyle\int_{\mathbb{R}^{N}}(u% _{1}v_{1}+u_{2}v_{2})$ ,

	$\displaystyle L_{k}(\varphi,\psi)=$	(2.2)
	$\displaystyle\left(\begin{matrix}-\Delta\varphi+P(r,y^{\prime\prime})\varphi-(% 2^{}-1)W_{1,\lambda}^{\frac{4}{N-2}}\varphi-\frac{\beta}{(N-2)}W_{1,\lambda}^% {\frac{4-N}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}\varphi-\frac{N\beta}{2(N-2)}W_{% 1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{2}{N-2}}\psi\\ -\Delta\psi+Q(r,y^{\prime\prime})\psi-(2^{}-1)W_{2,\lambda}^{\frac{4}{N-2}}% \psi-\frac{\beta}{(N-2)}W_{1,\lambda}^{\frac{4-N}{N-2}}W_{2,\lambda}^{\frac{N}% {N-2}}\psi-\frac{N\beta}{2(N-2)}W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{% \frac{2}{N-2}}\varphi\end{matrix}\right)^{\top}$
$\displaystyle=$	$\displaystyle(\varphi,\psi)\mathcal{L}_{k}^{\top},$	(2.3)

with

	$\displaystyle\mathcal{L}_{k}=$
	$\displaystyle\left(\begin{matrix}-\Delta+P(r,y^{\prime\prime})-(2^{}-1)W_{1,% \lambda}^{\frac{4}{N-2}}-\frac{\beta}{N-2}W_{1,\lambda}^{\frac{4-N}{N-2}}W_{2,% \lambda}^{\frac{N}{N-2}}&-\frac{N\beta}{2(N-2)}W_{1,\lambda}^{\frac{2}{N-2}}W_% {2,\lambda}^{\frac{2}{N-2}}\\ -\frac{N\beta}{2(N-2)}W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{2}{N-2% }}&-\Delta+Q(r,y^{\prime\prime})-(2^{}-1)W_{2,\lambda}^{\frac{4}{N-2}}-\frac{% \beta}{N-2}W_{2,\lambda}^{\frac{4-N}{N-2}}W_{1,\lambda}^{\frac{N}{N-2}}\end{% matrix}\right).$

Lemma 2.1.

Assume that $(\varphi,\psi)$ solves (2.1). If $\|(h_{1},h_{2})\|_{**}$ goes to zero as $k$ goes to infinity, so does $\|(\varphi,\psi)\|_{*}$ .

Proof.

By contradiction, we assume that there exist $k\rightarrow\infty$ , $(\varphi_{k},\psi_{k})$ , $\lambda_{k}\in\Bigl{[}L_{0}k^{\frac{N-2}{N-4}},L_{1}k^{\frac{N-2}{N-4}}\Bigr{]}$ , some $(\varphi_{k},\psi_{k})$ solving (2.1), with $\|(h_{1,k},h_{2,k})\|_{**}\rightarrow 0$ and $\|(\varphi_{k},\psi_{k})\|_{*}\geq c^{\prime}>0$ . We may assume that $\|(\varphi_{k},\psi_{k})\|_{*}=1$ . For simplicity, we drop the subscript $k$ and write (2.1) as

$\displaystyle\varphi(y)$	$\displaystyle=\int_{\mathbb{R}^{N}}\frac{(2^{*}-1)}{\|y-z\|^{N-2}}W_{1,\lambda}^% {\frac{4}{N-2}}\varphi(z)dz$
	$\displaystyle\quad+\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}\frac{1}{(% N-2)}W_{1,\lambda}^{\frac{4-N}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}\varphi(z)+% \frac{N}{2(N-2)}W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{2}{N-2}}\psi% (z)\Big{)}dz$
	$\displaystyle\quad+\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}h_{1}(z)+% \sum_{i=1}^{2}c_{i}\sum_{j=1}^{k}w_{x_{j},\lambda}^{2^{*}-2}Y_{j,i}\Big{)}dz$	(2.4)

and

$\displaystyle\psi(y)$	$\displaystyle=\int_{\mathbb{R}^{N}}\frac{(2^{*}-1)}{\|y-z\|^{N-2}}W_{2,\lambda}^% {\frac{4}{N-2}}\psi(z)dz$
	$\displaystyle\quad+\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}\frac{1}{(% N-2)}W_{2,\lambda}^{\frac{4-N}{N-2}}W_{1,\lambda}^{\frac{N}{N-2}}\psi(z)+\frac% {N}{2(N-2)}W_{2,\lambda}^{\frac{2}{N-2}}W_{1,\lambda}^{\frac{2}{N-2}}\varphi(z% )\Big{)}dz$
	$\displaystyle\quad+\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}h_{2}(z)+% \sum_{i=1}^{2}c_{i}\sum_{j=1}^{k}w_{x_{j},\lambda}^{2^{*}-2}Z_{j,i}\Big{)}dz.$	(2.5)

We are sufficed to deal with (2), since for (2), one can obtain similar estimates just in the same way. From Lemma A.5, we first get that

	$\displaystyle\Big{\|}\int_{\mathbb{R}^{N}}{\|y-z\|^{N-2}}W_{1,\lambda}^{\frac{4}{% N-2}}\varphi(z)dz\Big{\|}$
$\displaystyle\leq$	$\displaystyle C\\|\varphi\\|_{*}\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}W_{1,% \lambda}^{\frac{4}{N-2}}\sum_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+% \lambda\|z-x_{j}\|)^{\frac{N-2}{2}+\tau}}dz$	(2.6)
$\displaystyle\leq$	$\displaystyle C\\|\varphi\\|_{*}\lambda^{\frac{N-2}{2}}\sum_{j=1}^{k}\frac{1}{(1% +\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\theta}}.$	(2.7)

Recalling that $W_{2,\lambda}=\kappa W_{1,\lambda}$ , we obtain also

		$\displaystyle\Big{\|}\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}\frac{1}{% (N-2)}W_{1,\lambda}^{\frac{4-N}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}\varphi(z)+% \frac{N}{2(N-2)}W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{2}{N-2}}\psi% (z)\Big{)}dz\Big{\|}$
		$\displaystyle\leq C\lambda^{\frac{N-2}{2}}\\|(\varphi,\psi)\\|_{*}\sum_{j=1}^{k}% \frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\theta+\tau}}.$		(2.8)

Moreover, from Lemma A.4, there hold

\displaystyle\Big{|}\int_{\mathbb{R}^{N}}\frac{1}{|y-z|^{N-2}}h_{1}(z)dz\Big{|% }\leq C\lambda^{\frac{N-2}{2}}\|h_{1}\|_{**}\sum_{j=1}^{k}\frac{1}{(1+\lambda|% y-x_{j}|)^{\frac{N-2}{2}+\bar{\sigma}}}

(2.9)

and

\displaystyle\Big{|}\int_{\mathbb{R}^{N}}\frac{1}{|y-z|^{N-2}}\sum_{j=1}^{k}W_% {1,x_{j}}^{2^{*}-2}Y_{j,i}dz\Big{|}\leq C\lambda^{\frac{\lambda-2}{2}+n_{i}}% \sum_{j=1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N-2}{2}+\tau}},

(2.10)

where $n_{j}=1,j=2,\cdots,N$ , $n_{1}=-1$ .

Now, we estimate $c_{l},l=1,\cdots,N$ . Multiplying the two equations in (2.1) by $Y_{1,l}$ and $Z_{1,l}$ respectively and integrating, we find that

\displaystyle\sum_{i=1}^{N}\sum_{j=1}^{k}\left\langle(W_{1,x_{j}}^{2^{*}-2}Y_{% j,i},W_{2,x_{j}}^{2^{*}-2}Z_{j,i}),(Y_{1,l},Z_{1,l})\right\rangle c_{i}=\left% \langle L_{k}(\varphi_{1},\varphi_{2}),(Y_{1,l},Z_{1,l})\right\rangle-\left% \langle(h_{1},h_{2}),(Y_{1,l},Z_{1,l})\right\rangle.

(2.11)

From Lemma A.3 and since $\bar{\sigma}>1$ , we have

	$\displaystyle\left\|\left\langle(h_{1},h_{2}),(Y_{1,l},Z_{1,l})\right\rangle\right\|$	$\displaystyle\leq C\\|(h_{1},h_{2})\\|_{**}\int_{\mathbb{R}^{N}}\frac{\lambda^{% \frac{N-2}{2}+n_{l}}}{(1+\|z-x_{1}\|)^{N-2}}\sum_{j=1}^{k}\frac{\lambda^{\frac{N% +2}{2}}}{(1+\|z-x_{j}\|)^{\frac{N+2}{2}+\tau}}dz$
		$\displaystyle\leq C\lambda^{n_{l}}\\|(h_{1},h_{2})\\|_{**}.$

By direct computation, we have

\displaystyle|\langle P(r,y^{\prime\prime})\varphi,Y_{1,l}\rangle|\leq C\|% \varphi\|_{*}\int_{\mathbb{R}^{N}}\frac{\xi\lambda^{\frac{N-2}{2}+n_{i}}}{(1+% \lambda|z-x_{1}|)^{(N-2)}}\sum\limits_{j=1}^{k}\frac{\lambda^{\frac{N+2}{2}}}{% 1+\lambda|y-x_{j}|^{\frac{N+2}{2}+\tau}}=O\Bigl{(}\frac{\lambda^{n_{i}}\|% \varphi\|_{*}}{\lambda^{1+\epsilon}}\Bigr{)}.

(2.12)

Similarly, we have

\displaystyle\Bigl{|}\langle Q(r,y^{\prime\prime})\psi,Z_{1,l}\rangle\Bigr{|}% \leq O\Bigl{(}\frac{\lambda^{n_{l}}\|\psi\|_{*}}{\lambda^{1+\varepsilon}}\Bigr% {)}.

On the other hand, we also have

\displaystyle\left\langle L_{k}(\varphi,\psi),(Y_{1,l},Z_{1,l})\right\rangle=% \left\langle L_{k}(Y_{1,l},Z_{1,l}),(\varphi,\psi)\right\rangle.

(2.13)

Then

	$\displaystyle\eqref{eqs2.11}=$	$\displaystyle\frac{1}{(N-2)}\int_{\mathbb{R}^{N}}\Bigl{[}(N+2)\Bigl{(}W_{1,x_{% 1}}^{\frac{4}{N-2}}-W_{1,\lambda}^{\frac{4}{N-2}}\Bigr{)}+\beta\Bigl{(}W_{1,x_% {1}}^{\frac{4-N}{N-2}}W_{2,x_{1}}^{\frac{N}{N-2}}-W_{1,\lambda}^{\frac{4-N}{N-% 2}}W_{2,\lambda}^{\frac{N}{N-2}}\Bigr{)}$
		$\displaystyle+\frac{N\beta}{2}\Bigl{(}W_{1,x_{1}}^{\frac{2}{N-2}}W_{2,x_{1}}^{% \frac{2}{N-2}}-W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{2}{N-2}}\Bigr% {)}\Bigr{]}Y_{1,l}\varphi dy$
		$\displaystyle+\frac{1}{N-2}\int_{\mathbb{R}^{N}}\Bigl{[}(N+2)\Bigl{(}W_{1,x_{1% }}^{\frac{4}{N-2}}-W_{2,\lambda}^{\frac{4}{N-2}}\Bigr{)}+\Bigl{(}W_{2,x_{1}}^{% \frac{4-N}{N-2}}W_{1,x_{1}}^{\frac{N}{N-2}}-W_{2,\lambda}^{\frac{4-N}{N-2}}W_{% 1,\lambda}^{\frac{N}{N-2}}\Bigr{)}$
		$\displaystyle+\frac{N\beta}{2}\Bigl{(}W_{2,x_{1}}^{\frac{2}{N-2}}W_{1,x_{1}}^{% \frac{2}{N-2}}-W_{2,\lambda}^{\frac{2}{N-2}}W_{1,\lambda}^{\frac{2}{N-2}}\Bigr% {)}\Bigr{]}Z_{1,l}\psi dy.$

Then, we have

\displaystyle\left\langle L_{k}(\varphi,\psi),(Y_{1,l},Z_{1,l})\right\rangle=O% \Big{(}\frac{\lambda^{n_{l}}}{\lambda^{1+\varepsilon}}\Big{)}\|(\varphi,\psi)% \|_{*}+\|(h_{1},h_{2})\|_{**}.

(2.14)

But, observe that, there exists some $\bar{c}>0$ such that

\displaystyle\sum_{j=1}^{k}\big{\langle}(W_{1,x_{j}}^{2^{*}-2}Y_{j,i},W_{2,x_{% j}}^{2^{*}-2}Z_{j,i}),(Y_{1,l},Z_{1,l})\big{\rangle}=\lambda^{2n_{l}}(\bar{c}+% o(1))\delta_{il}.

Therefore, from (2.11), we get that

\displaystyle c_{l}=O\left(\frac{1}{\lambda^{n_{l}}}\|(\varphi,\psi)\|_{*}+\|(% h_{1},h_{2})\|_{**}\right).

(2.15)

Combining (2)-(2.15), we obtain that

\displaystyle\|(\varphi,\psi)\|_{*}\leq\Bigl{(}\|(h_{1},h_{2})\|_{**}+\frac{% \sum_{j=1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N-2}{2}+\tau+\theta}}}{\sum% _{j=1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N-2}{2}+\tau}}}\Bigr{)}.

(2.16)

Since $\|(\varphi,\psi)\|_{*}=1$ , we get from (2.16) that there exists some $R,a>0$ such that for some $j$

\displaystyle\|(\varphi,\psi)\|_{L^{\infty}(B_{R}(x_{j}))}\geq a>0.

(2.17)

However, transformation $(\bar{\varphi}(y),\bar{\psi}(y))=(\varphi(y-x_{j}),\psi(y-x_{j}))$ , notice that converges uniformly in any compact set to a solution $(u,v)$ which satisfies

\displaystyle\begin{cases}-\Delta u-(2^{*}-1)U_{0,\lambda}^{\frac{4}{N-2}}u-% \frac{1}{N-2}U_{0,\lambda}^{\frac{4}{N-2}}V_{0,\lambda}^{\frac{N}{N-2}}u-\frac% {N\beta}{2(N-2)}U_{0,\lambda}^{\frac{2}{N-2}}V_{0,\lambda}^{\frac{2}{N-2}}v=0,% \,\,\,y\in\mathbb{R}^{N},\\ -\Delta v-(2^{*}-1)V_{0,\lambda}^{\frac{4}{N-2}}v-\frac{1}{N-2}V_{0,\lambda}^{% \frac{4}{N-2}}U_{0,\lambda}^{\frac{N}{N-2}}v-\frac{N\beta}{2(N-2)}V_{0,\lambda% }^{\frac{2}{N-2}}U_{0,\lambda}^{\frac{2}{N-2}}u=0,\,\,\,y\in\mathbb{R}^{N}\end% {cases}

(2.18)

for some $\lambda\in[L_{1},L_{2}]$ . Moreover, $(u,v)$ is perpendicular to the kernel of (2.18). Hence, $(u,v)=(0,0)$ , which is a contradiction to (2.17). ∎

As a result of Lemma 2.1, applying the same argument as in the proof of proposition 4.1 in [8]. We can prove the following Proposition.

Proposition 2.2.

There exists $k_{0}>0$ and some constant $C>0$ , independent of $k$ , such that for all $k\geq k_{0}$ and all $(h_{1},h_{2})\in L^{\infty}(\mathbb{R}^{N})\times L^{\infty}(\mathbb{R}^{N})$ , the linear problem (2.1) has a unique solution $(\varphi,\psi)\equiv\mathbb{L}_{k}(h_{1},h_{2})$ . Moreover, there hold that

\displaystyle\|\mathbb{L}_{k}(h_{1},h_{2})\|_{*}\leq C\|(h_{1},h_{2})\|_{**},% \ \ \ \ |c_{l}|\leq\frac{1}{\lambda^{n_{l}}}\|(h_{1},h_{2})\|_{**}.

(2.19)

Now we consider the nonlinear problem

\displaystyle\begin{cases}-\Delta(W_{1,\lambda}+\varphi)+P(r,y^{\prime\prime})% (W_{1,\lambda}+\varphi)=(W_{1,\lambda}+\varphi)^{2^{*}-1}+\frac{\beta}{2}(W_{1% }+\varphi)^{\frac{2^{*}}{2}-1}(W_{2,\lambda}+\psi)^{\frac{2^{*}}{2}},\ \vspace% {0.12cm}\\ \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,% \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,+\sum\limits% _{i=1}^{2}c_{i}\sum\limits_{j=1}^{k}W_{1,x_{j}}^{2^{*}-2}Y_{j,i},\\ -\Delta(W_{2,\lambda}+\psi)+Q(r,y^{\prime\prime})(W_{2,\lambda}+\psi)=(W_{2,% \lambda}+\psi)^{2^{*}-1}+\frac{\beta}{2}(W_{2,\lambda}+\psi)^{\frac{2^{*}}{2}-% 1}(W_{1,\lambda}+\varphi)^{\frac{2^{*}}{2}},\ \vspace{0.12cm}\\ \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,% \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,% +\sum\limits_{i=1}^{2}c_{i}\sum\limits_{j=1}^{k}W_{2,x_{j}}^{2^{*}-2}Z_{j,i},% \\ (\varphi,\psi)\in H_{s}\times H_{s},\vspace{0.12cm}\\ \big{\langle}(W_{1,x_{j}}^{2^{*}-2}Y_{j,l},W_{2,x_{j}}^{2^{*}-2}Z_{j,l}),(% \varphi,\psi)\big{\rangle}=0,\ \ j=1,\ldots,k,\ \ l=1,2,\cdots,N.\end{cases}

(2.20)

In this section, we are aimed to prove that

Proposition 2.3.

There exists $k_{0}>0$ and some constant $C>0$ , independent of $k$ , such that for all $k\geq k_{0}$ , $\lambda\in[L_{0}k^{\frac{N-2}{N-4}},L_{1}k^{\frac{N-2}{N-4}}]$ , with $\bar{\theta}>0$ is a fixed small constant, problem (2.20) has a unique solution $(\varphi,\psi)=(\varphi(r,\lambda),\psi(r,\lambda))$ such that

\displaystyle\|(\varphi,\psi)\|_{*}\leq C\left(\frac{1}{\lambda}\right)^{1+% \epsilon},\,\,\,|c_{l}|\leq C\left(\frac{1}{\lambda}\right)^{1+\epsilon},\,\,% \,|\varphi|\leq\frac{1}{2}U,\,\,\,|\psi|\leq\frac{1}{2}V,

(2.21)

where $\epsilon>0$ small enough.

Rewrite the nonlinear problem (2.20) as

\displaystyle\begin{cases}L_{k}(\varphi,\psi)=R_{k}+N_{k}(\varphi,\psi),% \vspace{0.12cm}\\ (\varphi,\psi)\in H_{s}\times H_{s},\vspace{0.12cm}\\ \big{\langle}(W_{1,x_{j}}^{2^{*}-2}Y_{j,l},W_{2,x_{j}}^{2^{*}-2}Z_{j,l}),(% \varphi,\psi)\big{\rangle}=0,\ \ j=1,\ldots,k,\ \ l=1,\cdots,N,\end{cases}

(2.22)

where the operator $L_{k}$ is defined in (2.2), and

\displaystyle N_{k}(\varphi,\psi)=\left(N_{k}^{1,1}+2N_{k}^{1,2},N_{k}^{2,1}+2% N_{k}^{2,2}\right),\,\,\,R_{k}=(R_{1,k},R_{2,k}),

(2.23)

with

	$\displaystyle N_{k}^{1,1}=(W_{1,\lambda}+\varphi)^{2^{}-1}-W_{1,\lambda}^{2^{% }-1}-(2^{}-1)W_{1,\lambda}^{2^{}-2}\varphi,$
	$\displaystyle N_{K}^{1,2}=(W_{1,\lambda}+\varphi)^{\frac{2^{}}{2}-1}(W_{2,% \lambda}+\psi)^{\frac{2^{}}{2}}-W_{1,\lambda}^{\frac{2^{}}{2}-1}W_{2,\lambda% }^{\frac{2^{}}{2}}-\Bigl{(}\frac{2^{}}{2}-1\Bigr{)}W_{1,\lambda}^{\frac{2^{% }}{2}-2}W_{2,\lambda}^{\frac{2^{}}{2}}\varphi-\frac{2^{}}{2}W_{1,\lambda}^{% \frac{2^{}}{2}-1}W_{2,\lambda}^{\frac{2^{}}{2}-1}\psi,$
	$\displaystyle N_{K}^{2,1}=(W_{2,\lambda}+\psi)^{2^{}-1}-W_{2,\lambda}^{2^{}-% 1}-(2^{}-1)W_{2,\lambda}^{2^{}-2}\psi,$
	$\displaystyle N_{K}^{2,2}=(W_{2,\lambda}+\psi)^{\frac{2^{}}{2}-1}(W_{1,% \lambda}+\varphi)^{\frac{2^{}}{2}}-W_{2,\lambda}^{\frac{2^{}}{2}-1}W_{1,% \lambda}^{\frac{2^{}}{2}}-\Bigl{(}\frac{2^{}}{2}-1\Bigr{)}W_{2,\lambda}^{% \frac{2^{}}{2}-2}W_{1,\lambda}^{\frac{2^{}}{2}}\psi-\frac{2^{}}{2}W_{2,% \lambda}^{\frac{2^{}}{2}-1}W_{1,\lambda}^{\frac{2^{}}{2}-1}\varphi.$

Denote

	$\displaystyle R_{1,k}=$	$\displaystyle\Bigl{(}W_{1,\lambda}^{2^{}-1}-\sum\limits_{j=1}^{k}W_{1,x_{j}}^% {2^{}-1}\Bigr{)}-P(r,y^{\prime\prime})W_{1,\lambda}+W_{1,\lambda}^{}\Delta% \xi+2\nabla\xi\nabla W_{1,\lambda}^{}$
		$\displaystyle+\frac{\beta}{2}\Bigl{(}W_{2,\lambda}^{\frac{2^{}}{2}-1}W_{1,% \lambda}^{\frac{2^{}}{2}}-\sum\limits_{j=1}^{k}W_{2,x_{j}}^{\frac{2^{}}{2}}W% _{1,x_{j}}^{\frac{2^{}}{2}}\Bigr{)}$

and

	$\displaystyle R_{2,k}=$	$\displaystyle\Bigl{(}W_{2,\lambda}^{2^{}-1}-\sum\limits_{j=1}^{k}W_{2,x_{j}}^% {2^{}-1}\Bigr{)}-Q(r,y^{\prime\prime})W_{2,\lambda}+W_{2,\lambda}^{}\Delta% \xi+2\nabla\xi\nabla W_{2,\lambda}^{}$
		$\displaystyle+\frac{\beta}{2}\Bigl{(}W_{1,\lambda}^{\frac{2^{}}{2}-1}W_{2,% \lambda}^{\frac{2^{}}{2}}-\sum\limits_{j=1}^{k}W_{2,x_{j}}^{\frac{2^{}}{2}}W% _{1,x_{j}}^{\frac{2^{}}{2}}\Bigr{)}.$

In the following, we will use the contraction mapping theorem to show that there exists a unique solution to problem (2.22) in the set that $\|(\varphi,\psi)\|_{*}$ is small. In order to do this, we first estimate $N_{k}(\varphi,\psi)$ and $R_{k}$ , and, just as before, we may drop the subscript $k$ for convenience.

Lemma 2.4.

If $N\geq 5$ , then for any $(\varphi,\psi)$ satisfying $|\varphi|\leq\frac{1}{2}U_{1}(y),|\psi|\leq\frac{1}{2}V_{1}(y)$ there holds

\displaystyle\|N_{k}(\varphi,\psi)\|_{**}\leq C\|(\varphi,\psi)\|_{*}^{1+% \delta},

where $0<\delta<\frac{4}{N-2}$ .

Proof.

By definition of $N=N_{k}$ in (2.23), observing that for the terms $N^{1,1}$ and $N^{2,1}$ , we just follow the estimates as in Lemma 2.4 of [35] to get that, for $i=1,2$ ,

\displaystyle\|N^{i,1}\|_{**}\leq C\|(\varphi,\psi)\|_{*}^{2^{*}-1}.

Since for $p\geq\frac{N-2-m}{N-2}$ ,

\displaystyle\sum_{j=1}^{k}\frac{1}{(1+|y-x_{j}|)^{p}}\leq C+\sum_{j=2}^{k}% \frac{1}{|x_{j}-x_{1}|^{p}}\leq C,

combining Hölder inequalities, for $i=1,2$ , we obtain

\displaystyle\begin{split}|N^{i,1}|&\leq\|\varphi\|_{*}^{2^{*}-1}\Big{(}\sum_{% j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda|y-x_{j}|)^{\frac{N-2}{2}+% \tau}}\Big{)}^{2^{*}-1}\\ &\leq C\|\varphi\|_{*}^{2^{*}-1}\lambda^{\frac{N+2}{2}}\sum_{j=1}^{k}\frac{1}{% (1+\lambda|y-x_{j}|)^{\frac{N+2}{2}+\tau}}\Big{(}\sum_{j=1}^{k}\frac{1}{(1+% \lambda|y-x_{j}|)^{\tau}}\Big{)}^{\frac{4}{N-2}}\\ &\leq C\|(\varphi,\psi)\|_{*}^{2^{*}-1}\lambda^{\frac{N+2}{2}}\sum_{j=1}^{k}% \frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N+2}{2}+\tau}}.\end{split}

(2.24)

Next for the terms $N^{1,2}$ and $N^{2,2}$ in (2.23), it suffices to deal with $N^{1,2}$ , since the other one can be estimated in the same way.

Rewrite

	$\displaystyle N^{1,2}=$	$\displaystyle(W_{1,\lambda}+\varphi)^{\frac{2}{N-2}}(W_{2,\lambda}+\psi)^{% \frac{N}{N-2}}-W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}-% \frac{2}{N-2}W_{1,\lambda}^{\frac{4-N}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}\varphi$		(2.25)
		$\displaystyle-\frac{N}{N-2}W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{2% }{N-2}}\psi=I+II+III,$		(2.26)

where

	$\displaystyle I=(W_{1,\lambda}+\varphi)^{\frac{2}{N-2}}\Big{(}(W_{2,\lambda}+% \psi)^{\frac{N}{N-2}}-W_{2,\lambda}^{\frac{N}{N-2}}-\frac{N}{N-2}W_{2,\lambda}% ^{\frac{2}{N-2}}\psi\Big{)},$
	$\displaystyle II=\Big{(}(W_{1,\lambda}+\varphi)^{\frac{2}{N-2}}-W_{1,\lambda}^% {\frac{2}{N-2}}-\frac{2}{N-2}W_{1,\lambda}^{\frac{4-N}{N-2}}\varphi\Big{)}W_{2% ,\lambda}^{\frac{N}{N-2}},$
	$\displaystyle III=\frac{N}{N-2}\Big{(}(W_{1,\lambda}+\varphi)^{\frac{2}{N-2}}-% W_{1,\lambda}^{\frac{2}{N-2}}\Big{)}W_{2,\lambda}^{\frac{2}{N-2}}\psi.$

First for $I$ , we have

\displaystyle|I|\leq C\Bigl{(}|\psi|^{\frac{N}{N-2}}W_{1,\lambda}^{\frac{2}{N-% 2}}+|\psi|^{\frac{N}{N-2}}|\varphi|^{\frac{2}{N-2}}\Bigr{)}:=I_{1}+I_{2}.

We estimate by Hölder inequalities,

	$\displaystyle\|I_{1}\|\leq$	$\displaystyle C\\|\psi\\|_{*}^{\frac{N}{N-2}}\Big{(}\sum_{j=1}^{k}\frac{\lambda^% {\frac{N-2}{2}}}{(1+\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\tau}}\Big{)}^{\frac{N}{N% -2}}\Big{(}\sum_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda\|y-x_{j}\|)^{% N-2}}\Big{)}^{\frac{2}{N-2}}$
	$\displaystyle\leq$	$\displaystyle C\lambda^{\frac{N+2}{2}}\\|\psi\\|_{}^{\frac{N}{N-2}}\Big{(}\sum_% {j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\tau}}\Big{)}^{2^{}-1}$
	$\displaystyle\leq$	$\displaystyle C\lambda^{\frac{N+2}{2}}\\|\psi\\|_{*}^{\frac{N}{N-2}}\sum_{j=1}^{% k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}}\Big{(}\sum_{j=1}^{K}% \frac{1}{(1+\lambda\|y-x_{j}\|)^{\tau}}\Big{)}^{\frac{4}{N-2}}$
	$\displaystyle\leq$	$\displaystyle C\lambda^{\frac{N+2}{2}}\\|(\varphi,\psi)\\|_{*}^{\frac{N}{N-2}}% \sum_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}}.$

Moreover, we have

	$\displaystyle\|I_{2}\|$	$\displaystyle\leq C\\|\psi\\|_{}^{\frac{N}{N-2}}\\|\varphi\\|_{}^{\frac{2}{N-2}}% \Big{(}\sum_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda\|y-x_{j}\|)^{% \frac{N-2}{2}+\tau}}\Big{)}^{\frac{N+2}{N-2}}$
		$\displaystyle\leq C\lambda^{\frac{N+2}{2}}\\|\psi\\|_{}^{\frac{N}{N-2}}\\|% \varphi\\|_{}^{\frac{2}{N-2}}\sum_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{% \frac{N+2}{2}+\tau}}\Big{(}\sum_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\tau}}% \Big{)}^{\frac{4}{N-2}}$
		$\displaystyle\leq C\lambda^{\frac{N+2}{2}}\\|\psi\\|_{}^{\frac{N}{N-2}}\\|% \varphi\\|_{}^{\frac{2}{N-2}}\sum_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{% \frac{N+2}{2}+\tau}}.$

Therefore, we get

\displaystyle|I|\leq C\lambda^{\frac{N+2}{2}}\Bigl{(}\|\psi\|_{*}^{\frac{N}{N-% 2}}+\|\psi\|_{*}^{\frac{N}{N-2}}\|\varphi\|_{*}^{\frac{2}{N-2}}\Bigr{)}\sum_{j% =1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N+2}{2}+\tau}}.

For $II$ , we will deal with it by dividing into the following two cases. It follows from Lemma A.6 that $|\frac{\varphi_{i}}{W_{i}}|<\frac{1}{2}.$ Using the basic inequality

\displaystyle(1+t)^{\frac{2}{N-2}}-1-\frac{2}{N-2}t=O(t^{2}),\,\,\,\text{for}% \,\,\,|t|<\frac{1}{2},

(2.27)

we obtain if $N=5$ by using Hölder again,

$\displaystyle\|II\|=$	$\displaystyle\Big{\|}\Big{(}(W_{1,\lambda}+\varphi)^{\frac{2}{N-2}}-W_{1,% \lambda}^{\frac{2}{N-2}}-\frac{2}{N-2}W_{1,\lambda}^{\frac{4-N}{N-2}}\varphi% \Big{)}W_{2,\lambda}^{\frac{N}{N-2}}\Big{\|}$	(2.28)
$\displaystyle=$	$\displaystyle W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}\Big{\|% }\Bigl{(}1+\frac{\varphi}{W_{1,\lambda}}\Bigr{)}^{\frac{2}{N-2}}-1-\frac{2}{N-% 2}\frac{\varphi}{W_{1,\lambda}}\Big{\|}\leq CW_{1,\lambda}^{\frac{2}{N-2}-2}W_{% 2,\lambda}^{\frac{N}{N-2}}\|\varphi\|^{2}$	(2.29)
$\displaystyle\leq$	$\displaystyle C\\|\varphi\\|_{*}^{2}\lambda^{\frac{N+2}{2}}\Big{(}\sum_{j=1}^{k}% \frac{1}{(1+\lambda\|y-x_{j}\|)^{N-2}}\Big{)}^{\frac{6-N}{N-2}}\Big{(}\sum_{j=1}% ^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\tau}}\Big{)}^{2}$	(2.30)
$\displaystyle\leq$	$\displaystyle C\lambda^{\frac{N+2}{2}}\\|\varphi\\|_{*}^{2}\Big{(}\sum_{j=1}^{k}% \frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\tau}}\Big{)}^{\frac{N+2}{N-2}}$	(2.31)
$\displaystyle\leq$	$\displaystyle C\lambda^{\frac{N+2}{2}}\\|\varphi\\|_{*}^{2}\sum_{j=1}^{k}\frac{1% }{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}}.$	(2.32)

Also for $|t|<1$ and $N\geq 6,$ there holds $(1+t)^{\frac{2}{N-2}}-1-\frac{2}{N-2}t=O(t^{1+\delta}),$ where $\frac{4}{N-2}-\delta>0$ . Hence when $N\geq 6,$ we have

\displaystyle\begin{split}|II|&=\Big{|}\Big{(}(W_{1,\lambda}+\varphi)^{\frac{2% }{N-2}}-W_{1,\lambda}^{\frac{2}{N-2}}-\frac{2}{N-2}W_{1,\lambda}^{\frac{4-N}{N% -2}}\varphi\Big{)}W_{2,\lambda}^{\frac{N}{N-2}}\Big{|}\cr&=W_{1,\lambda}^{% \frac{2}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}\Big{|}\Bigl{(}1+\frac{\varphi}{W_{% 1,\lambda}}\Bigr{)}^{\frac{2}{N-2}}-1-\frac{2}{N-2}\frac{\varphi}{W_{1,\lambda% }}\Big{|}\leq CW_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}\Big{% |}\frac{\varphi}{W_{1,\lambda}}\Big{|}^{1+\delta}\cr&\leq C\|\varphi\|_{*}^{1+% \delta}\Big{(}\sum\limits_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda|y% -x_{j}|)^{N-2}}\Big{)}^{\frac{4}{N-2}-\delta}\Big{(}\sum\limits_{j=1}^{k}\frac% {\lambda^{\frac{N-2}{2}}}{(1+\lambda|y-x_{j}|)^{\frac{N-2}{2}+\tau}}\Big{)}^{1% +\delta}\cr&\leq C\lambda^{\frac{N+2}{2}}\|\varphi\|_{*}^{1+\delta}\Big{(}\sum% \limits_{j=1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N-2}{2}+\tau}}\Big{)}^{% \frac{N+2}{N-2}}\leq C\lambda^{\frac{N+2}{2}}\|\varphi\|_{*}^{1+\delta}\sum% \limits_{j=1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N+2}{2}+\tau}},\end{split}

since

	$\displaystyle\Big{(}\sum\limits_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\bar{% \sigma}}}\Big{)}^{\frac{N+2}{N-2}}$	$\displaystyle\leq\Big{(}\sum\limits_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{% \frac{N+2}{2}+\tau}}\Big{)}\Big{(}\sum\limits_{j=1}^{k}\frac{1}{(1+\lambda\|y-x% _{j}\|)^{\tau}}\Big{)}^{\frac{4}{N-2}}$
		$\displaystyle\leq C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}\sum\limits_{% j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}},$

which means that

\displaystyle\begin{split}|II|\leq C\lambda^{\frac{N+2}{2}}\|\varphi\|_{*}^{1+% \delta}\sum_{j=1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N+2}{2}+\tau}}.\end{split}

(2.33)

Finally, for $III$ , we have

\displaystyle\begin{split}|III|&\leq C|\varphi|^{\frac{2}{N-2}}|\psi|W_{2,% \lambda}^{\frac{2}{N-2}}\cr&\leq C\|\varphi\|_{*}^{\frac{2}{N-2}}\|\psi\|_{*}% \Big{(}\sum_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda|y-x_{j}|)^{% \frac{N-2}{2}+\tau}}\Big{)}^{\frac{N}{N-2}}\Big{(}\sum_{j=1}^{k}\frac{\lambda^% {\frac{N-2}{2}}}{(1+\lambda|y-x_{j}|)^{N-2}}\Big{)}^{\frac{2}{N-2}}\cr&\leq C% \lambda^{\frac{N+2}{2}}\|\varphi\|_{*}^{\frac{2}{N-2}}\|\psi\|_{*}\Big{(}\sum_% {j=1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N-2}{2}+\tau}}\Big{)}^{\frac{N+2% }{N-2}}\cr&\leq C\lambda^{\frac{N+2}{2}}\|\varphi\|_{*}^{\frac{2}{N-2}}\|\psi% \|_{*}\sum_{j=1}^{k}\frac{1}{(1+\lambda|y-x_{j}|)^{\frac{N+2}{2}+\tau}}.\end{split}

(2.34)

From (2.24) to (2.34), we obtain that

\displaystyle\|N(\varphi,\psi)\|_{**}\leq C\|(\varphi,\psi)\|_{*}^{1+\delta}.

∎

Next, we estimate $R_{k}$ .

Lemma 2.5.

Then, there exists some small $\epsilon>0$ such that

\displaystyle\|R_{k}\|_{**}\leq C\left(\frac{1}{\lambda}\right)^{1+\epsilon}.

(2.35)

Proof.

We define

\displaystyle\Omega_{j}=\left\{y:y=(y^{\prime},y^{\prime\prime})\in\mathbb{R}^% {2}\times\mathbb{R}^{N-2},\langle\frac{y^{\prime}}{|y^{\prime}|},\frac{x_{j}}{% |x_{j}|}\rangle\geq\cos\frac{\pi}{k}\right\}.

Recall by (2.23) that

\displaystyle\begin{split}R_{1,K}&=\Bigl{(}W_{1,\lambda}^{2^{*}-1}-\sum_{j=1}^% {k}W_{1,x_{j}}^{2^{*}-1}\Bigr{)}-P(r,y^{\prime\prime})W_{1,\lambda}+W^{*}_{1,% \lambda}\Delta\xi+2\nabla\xi\nabla W_{1,\lambda}^{*}+\frac{\beta}{2}\Big{(}W_{% 1,\lambda}^{\frac{2^{*}}{2}-1}W_{2,\lambda}^{\frac{2^{*}}{2}}-\sum_{j=1}^{k}W_% {1,x_{j}}^{\frac{2^{*}}{2}-1}W_{2,x_{j}}^{\frac{2^{*}}{2}}\Big{)}\\ &:=J_{1}+J_{2}+J_{3}+J_{4}+J_{5}.\end{split}

(2.36)

By symmetry, we might as well assume that $y\in\Omega_{1}$ , then $|y-x_{j}|\geq|y-x_{1}|$ . Therefore,

\displaystyle\begin{split}|J_{1}|\leq&CW_{1,x_{1}}^{\frac{4}{N-2}}\sum\limits_% {j=2}^{k}W_{1,x_{j}}+C\Bigl{(}\sum\limits_{j=2}^{k}W_{1,x_{j}}\Bigr{)}^{2^{*}-% 1}\cr\leq&C\frac{\lambda^{2}}{(1+\lambda|y-x_{1}|)^{4}}\sum_{j=2}^{k}\frac{% \lambda^{\frac{N-2}{2}}}{(1+\lambda|y-x_{j}|)^{N-2}}+C\Big{(}\sum_{j=2}^{k}% \frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda|y-x_{j}|)^{N-2}}\Big{)}^{2^{*}-1}.% \end{split}

(2.37)

By Lemma A.3, taking $0<\alpha\leq\min\{4,N-2\}$ , we obtain that for any $y\in\Omega_{1}$ and $j>1$ , then

\displaystyle\begin{split}\frac{1}{(1+\lambda|y-x_{1}|)^{4}}\sum_{j=2}^{k}% \frac{1}{(1+\lambda|y-x_{j}|)^{N-2}}\leq C\frac{1}{(1+\lambda|y-x_{1}|)^{N+2-% \alpha}}\sum_{j=2}^{k}\frac{1}{|\lambda(x_{j}-x_{1})|^{\alpha}}.\end{split}

(2.38)

We choose $\alpha>\frac{N}{2}$ satisfying $\frac{N-2}{2}\leq\alpha\leq 4-\tau$ to obtain that

\displaystyle\frac{1}{(1+\lambda|y-x_{1}|)^{4}}\sum_{j=2}^{k}\frac{1}{(1+% \lambda|y-x_{j}|)^{N-2}}\leq\frac{1}{(1+\lambda|y-x_{1}|)^{N-2+\tau}}\Bigl{(}% \frac{k}{\lambda}\Bigr{)}^{\alpha}\leq\frac{1}{C}\frac{1}{(1+\lambda|y-x_{1}|)% ^{\frac{N+2}{2}+\tau}}\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}.

Next, as in [23], for $y\in\Omega_{1}$ , we can derive

$\displaystyle\Bigl{(}\sum\limits_{j=2}^{k}\frac{1}{1+\lambda\|y-x_{j}\|^{N-2}}% \Bigr{)}^{2^{*}-1}\leq$	$\displaystyle\sum\limits_{j=2}^{k}\frac{1}{1+\lambda\|y-x_{j}\|^{\frac{N+2}{2}+% \tau}}\Bigl{(}\sum\limits_{j=2}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{4% }(\frac{N-2}{2}-\tau(\frac{N-2}{N+2}))}}\Bigr{)}^{\frac{4}{N-2}}$	(2.39)
$\displaystyle\leq$	$\displaystyle\Bigl{(}\frac{k}{\lambda}\Bigr{)}^{\frac{N+2}{4}\Bigl{(}\frac{N-2% }{2}-\tau(\frac{N-2}{N+2})\Bigr{)}}\sum\limits_{j=2}^{k}\frac{1}{(1+\lambda\|y-% x_{j}\|)^{\frac{N+2}{2}+\tau}}$	(2.40)
$\displaystyle\leq$	$\displaystyle\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}\sum\limits_{j=2}^{% k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}}.$	(2.41)

Hence (2.37)-(2.39) imply that

\displaystyle\|J_{1}\|_{**}\leq\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}.

On the other hand, we obtain

	$\displaystyle\|J_{2}\|\leq$	$\displaystyle C\sum\limits_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}\xi}{(1+% \lambda\|y-x_{j}\|)^{N-2}}\leq C\lambda^{\frac{N+2}{2}}\sum\limits_{j=1}^{k}% \frac{\xi}{\lambda^{2}(1+\lambda\|y-x_{j}\|)^{N-2}}$		(2.42)
	$\displaystyle\leq$	$\displaystyle C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}\sum\limits_{j=1}% ^{k}\frac{\lambda^{\frac{N+2}{2}}\xi}{\lambda^{1-\epsilon}(1+\lambda\|y-x_{j}\|)% ^{N-2}}\leq C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}\sum\limits_{j=1}^{% k}\frac{\lambda^{\frac{N+2}{2}}}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}},$		(2.43)

since $N-1-\epsilon>\frac{N+2}{2}+\tau$ if $\epsilon>0$ is small. Hence $\|J_{2}\|_{**}\leq C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}$ .

Similar to (2.42), we have

	$\displaystyle\|J_{3}\|\leq$	$\displaystyle C\sum\limits_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}\|\Delta\xi\|}{% (1+\lambda\|y-x_{j}\|)^{N-2}}\leq C\lambda^{\frac{N+2}{2}}\sum\limits_{j=1}^{k}% \frac{\|\Delta\xi\|}{\lambda^{2}(1+\lambda\|y-x_{j}\|)^{N-2}}$		(2.44)
	$\displaystyle\leq$	$\displaystyle C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}\sum\limits_{j=1}% ^{k}\frac{\|\Delta\xi\|\lambda^{\frac{N+2}{2}}}{\lambda^{1-\epsilon}(1+\lambda\|y% -x_{j}\|)^{N-2}}\leq C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}\sum\limits% _{j=1}^{k}\frac{\lambda^{\frac{N+2}{2}}}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+% \tau}},$		(2.45)

which yields that $\|J_{3}\|_{**}\leq C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}$ .

Moreover, we can check that

$\displaystyle\|J_{4}\|\leq$	$\displaystyle C\sum\limits_{j=1}^{k}\frac{\lambda^{\frac{N}{2}}\|\nabla\xi\|}{(1% +\lambda\|y-x_{j}\|)^{N-1}}\leq C\lambda^{\frac{N+2}{2}}\sum\limits_{j=1}^{k}% \frac{\|\nabla\xi\|}{\lambda(1+\lambda\|y-x_{j}\|)^{N-1}}$	(2.46)
$\displaystyle\leq$	$\displaystyle\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}\sum\limits_{j=1}^{% k}\frac{\|\Delta\xi\|\lambda^{\frac{N+2}{2}}}{\lambda^{-\epsilon}(1+\lambda\|y-x_% {j}\|)^{N-2}}$	(2.47)
$\displaystyle\leq$	$\displaystyle C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}\sum\limits_{j=1}% ^{k}\frac{\lambda^{\frac{N+2}{2}}}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}},$	(2.48)

since $\delta\leq|(r,y^{\prime\prime})-(r_{0},y_{0}^{\prime\prime})|\leq 2\delta$ and (1.7) implies that

\displaystyle\frac{1}{1+\lambda|y-x_{j}|}\leq\frac{1}{\lambda}.

Hence, we obtain $\|J_{4}\|_{**}\leq C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}$ .

As a result, from (2.37) to (2.46), we have

\displaystyle\|l_{k}\|_{**}\leq C\Bigl{(}\frac{1}{\lambda}\Bigr{)}^{1+\epsilon}

∎

Now, we are ready to prove Proposition 2.3.

Proof of Proposition 2.3:.

Recall that $\lambda=k^{\frac{N-2}{N-4}}$ . Let

	$\displaystyle E=\Big{\{}(\varphi_{1},\varphi_{2})\in$	$\displaystyle(C(\mathbb{R}^{N})\cap H_{s})^{2},\ \\|(\varphi,\psi)\\|_{*}\leq% \frac{1}{\lambda},\ \|\varphi_{i}\|\leq\frac{1}{2}W_{i},\ i=1,2,$
		$\displaystyle\big{\langle}(w_{1,x_{j}}^{2^{}-2}Y_{j,l},w_{2,x_{j}}^{2^{}-2}Z% _{j,l}),(\varphi,\psi)\big{\rangle}=0,\ j=1,\ldots,k,\ l=1,2\Big{\}}.$

Then we are sufficed to solve

\displaystyle(\varphi,\psi)=A(\varphi,\psi):=\mathbb{L}_{k}(N(\varphi,\psi))+% \mathbb{L}_{k}(R_{k})

with $\mathbb{L}_{k}$ defined as in Proposition 2.2. We will prove that $A$ is a contraction map from $E$ to $E$ .

First, by Proposition 2.2, Lemma 2.4, Lemma 2.5, $A$ maps $E$ to $E$ and

	$\displaystyle\\|A(\varphi,\psi)\\|_{*}$	$\displaystyle\leq C\\|N(\varphi,\psi)\\|_{}+C\\|R_{k}\\|_{}$
		$\displaystyle\leq C\\|(\varphi,\psi)\\|_{*}^{1+\delta}+\Bigl{(}\frac{1}{\lambda}% \Bigr{)}^{1+\epsilon}\leq\frac{1}{\lambda}.$

Next it is obvious

\displaystyle\|A(\varphi,\psi)-A(\tilde{\varphi},\tilde{\psi})\|_{*}

\displaystyle\leq C\|N(\varphi,\psi)-N(\tilde{\varphi},\tilde{\psi})\|_{**}.

Recall the definition of (2.23). We first deal with $N_{K}^{i,2}\ (i=1,2)$ by considering

f(t,s):=(1+t)^{\frac{2}{N-2}}(1+s)^{\frac{N}{N-2}}-1-\frac{2}{N-2}t-\frac{N}{N% -2}s.

It is easy to have

	$\displaystyle f^{\prime}_{t}(t,s)=\frac{2}{N-2}(1+t)^{\frac{4-N}{N-2}}(1+s)^{% \frac{N}{N-2}}-\frac{2}{N-2},$
	$\displaystyle f^{\prime}_{s}(t,s)=\frac{N}{N-2}(1+t)^{\frac{2}{N-2}}(1+s)^{% \frac{2}{N-2}}-\frac{N}{N-2}.$

For the case $t+1,s+1>\frac{1}{2}$ , $|f^{\prime}_{t}(t,s)|$ and $|f^{\prime}_{s}(t,s)|=O(|s|+|t|)$ , so we have

	$\displaystyle\|N^{i,2}(\varphi,\psi)-N^{i,2}(\tilde{\varphi},\tilde{\psi})\|$
	$\displaystyle\leq CW_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}% \Bigl{(}\Bigl{\|}\frac{\varphi}{W_{1,\lambda}}\Bigr{\|}+\Bigl{\|}\frac{\psi}{W_{2% ,\lambda}}\Bigr{\|}\Bigr{)}\Bigl{(}\frac{\|\varphi-\tilde{\varphi}\|}{W_{1,% \lambda}}+\frac{\|\psi-\tilde{\psi}\|}{W_{2,\lambda}}\Bigr{)}$
	$\displaystyle\leq C\Big{(}\sum_{j=1}^{k}\frac{1}{(1+\|y-x_{j}\|)^{N-2}}\Big{)}^{% \frac{4}{N-2}}\Big{(}\sum_{j=1}^{k}\frac{1}{(1+\|y-x_{j}\|)^{\bar{\sigma}}}\Big{% )}^{2}$
	$\displaystyle\quad\times(\\|\varphi\\|_{}+\\|\psi\\|_{})(\\|\varphi-\tilde{% \varphi}\\|_{}+\\|\psi-\tilde{\psi}\\|_{})$
	$\displaystyle\leq(\\|\varphi\\|_{}+\\|\psi\\|_{})(\\|\varphi-\tilde{\varphi}\\|_{% }+\\|\psi-\tilde{\psi}\\|_{})\Big{(}\sum_{j=1}^{k}\frac{1}{(1+\|y-x_{j}\|)^{\bar{% \sigma}}}\Big{)}^{\frac{N+2}{N-2}}$
	$\displaystyle\leq\frac{1}{2}(\\|\varphi-\tilde{\varphi}\\|_{}+\\|\psi-\tilde{% \psi}\\|_{}).$

We can deal with $N^{1,1}$ and $N^{2,1}$ similarly.

Therefore, $A$ is a contraction map and it follows from the contraction mapping theorem that there exists a unique $(\varphi,\psi)\in E,$ such that

(\varphi,\psi)=A(\varphi,\psi).

Moreover, from Proposition 2.2 we get

\|(\varphi,\psi)\|_{*}\leq C\Big{(}\frac{1}{\lambda}\Big{)}^{1+\epsilon}.

∎

3. Proof of the main result

Let

\displaystyle F(r,\lambda)=I(W_{1,\lambda}+\varphi,W_{2,\lambda}+\psi),

where $r=|x_{1}|$ , $(\varphi_{1},\varphi_{2})$ is obtained in Proposition 2.3, and

\displaystyle I(u,v):=\frac{1}{2}\int_{\mathbb{R}^{N}}(|\nabla u|^{2}+P(r,y^{% \prime\prime})u^{2}+|\nabla v|^{2}+Q(r,y^{\prime\prime})v^{2})-\frac{1}{2^{*}}% \int_{\mathbb{R}^{N}}\left(|u|^{2^{*}}+|v|^{2^{*}}+\beta|u|^{\frac{2^{*}}{2}}|% v|^{\frac{2^{*}}{2}}\right).

In this section, we will choose suitable $(\bar{r},\bar{y}^{\prime\prime},\lambda)$ so that $(W_{1,\lambda}+\varphi,W_{2,\lambda}+\psi)$ is a solution of (1.1). For this purpose, we need the following result.

Lemma 3.1.

Suppose that $(r,y^{\prime\prime},\lambda)$ satisfies

		$\displaystyle\int_{D_{\rho}}\Bigl{(}-\Delta u_{k}+P(r,y^{\prime\prime})u_{k}-u% _{k}^{2^{}}-\frac{\beta}{2}u_{k}^{\frac{2^{}}{2}-1}v_{k}^{\frac{2^{*}}{2}}% \Bigr{)}(y,\nabla u_{k})=0,$		(3.1)
		$\displaystyle\int_{D_{\rho}}\Bigl{(}-\Delta v_{k}+Q(r,y^{\prime\prime})v_{k}-v% _{k}^{2^{}}-\frac{\beta}{2}v_{k}^{\frac{2^{}}{2}-1}u_{k}^{\frac{2^{*}}{2}}% \Bigr{)}(y,\nabla v_{k})=0,$		(3.2)

		$\displaystyle\int_{D_{\rho}}\Bigl{(}-\Delta u_{k}+P(r,y^{\prime\prime})u_{k}-u% _{k}^{2^{}}-\frac{\beta}{2}u_{k}^{\frac{2^{}}{2}-1}v_{k}^{\frac{2^{*}}{2}}% \Bigr{)}\frac{\partial u_{k}}{\partial y_{i}}=0,$		(3.3)
		$\displaystyle\int_{D_{\rho}}\Bigl{(}-\Delta v_{k}+Q(r,y^{\prime\prime})v_{k}-v% _{k}^{2^{}}-\frac{\beta}{2}v_{k}^{\frac{2^{}}{2}-1}u_{k}^{\frac{2^{*}}{2}}% \Bigr{)}\frac{\partial v_{k}}{\partial y_{i}}=0,$		(3.4)

and

		$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}-\Delta u_{k}+P(r,y^{\prime\prime})u% _{k}-u_{k}^{2^{}}-\frac{\beta}{2}u_{k}^{\frac{2^{}}{2}-1}v_{k}^{\frac{2^{*}}% {2}}\Bigr{)}\frac{\partial u_{k}}{\partial\lambda}=0,$		(3.5)
		$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}-\Delta v_{k}+Q(r,y^{\prime\prime})v% _{k}-v_{k}^{2^{}}-\frac{\beta}{2}v_{k}^{\frac{2^{}}{2}-1}u_{k}^{\frac{2^{*}}% {2}}\Bigr{)}\frac{\partial v_{k}}{\partial\lambda}=0,$		(3.6)

where $u_{k}=W_{1,\lambda}+\varphi$ , $v_{k}=W_{2,\lambda}+\psi$ and $D_{\rho}=\displaystyle\{(r,y^{\prime\prime}):|(r,y^{\prime\prime})-(r_{0},y_{0% }^{\prime\prime})|\leq\rho\}$ with $\rho\in(2\delta,5\delta)$ . Then $c_{l}=0,l=1,2,\cdots,N$ .

Proof.

Since $(W_{1,\lambda},W_{2,\lambda})=(0,0)$ in $\mathbb{R}^{N}\backslash D_{\rho}$ , we see that if (3.1)-(3.5) hold, then

\displaystyle\sum\limits_{i=1}^{N}c_{i}\sum\limits_{j=1}^{k}(W_{1,x_{j}}^{2^{*% }-2}Y_{j,i}u,W_{2,x_{j}}^{2^{*}-2}Z_{j,i}v)=(0,0),

(3.7)

for $(u,v)=\Bigl{(}\langle y,\nabla u_{k}\rangle,\langle y,\nabla v_{k}\rangle\Bigr% {)}$ , $(u,v)=\Bigl{(}\frac{\partial u_{k}}{\partial y_{i}},\frac{\partial v_{k}}{% \partial y_{i}}\Bigr{)}$ and $(u,v)=\Bigl{(}\frac{\partial W_{1,\lambda}}{\partial\lambda},\frac{\partial W_% {2,\lambda}}{\partial\lambda}\Bigr{)}.$

For any function $f,\varphi\in H^{1}(\mathbb{R}^{N})$ , we have

\displaystyle\int_{\mathbb{R}^{N}}f\frac{\partial\varphi}{\partial y_{i}}=-% \int_{\mathbb{R}^{N}}\varphi\frac{\partial f}{\partial y_{i}}.

(3.8)

Using (3.8), we can prove that from (3.7) that

\displaystyle\sum\limits_{i=1}^{N}c_{i}\sum\limits_{j=1}^{k}(W_{1,x_{j}}^{2^{*% }-2}Y_{j,i}u,W_{2,x_{j}}^{2^{*}-2}Z_{j,i}v)(1+o(1))=(0,0)

(3.9)

holds for $(u,v)=\Bigl{(}\langle y,\nabla u_{k}\rangle,\langle y,\nabla v_{k}\rangle\Bigr% {)}$ , $(u,v)=\Bigl{(}\frac{\partial u_{k}}{\partial y_{i}},\frac{\partial v_{k}}{% \partial y_{i}}\Bigr{)}$ and $(u,v)=\Bigl{(}\frac{\partial W_{1,\lambda}}{\partial\lambda},\frac{\partial W_% {2,\lambda}}{\partial\lambda}\Bigr{)}$ .

We have

\displaystyle\langle y,\nabla W_{i,\lambda}\rangle=\langle y^{\prime},\nabla_{% y^{\prime}}W_{i,\lambda}\rangle+\langle y^{\prime\prime},\nabla_{y^{\prime% \prime}}W_{i,\lambda}\rangle,\,\,\,i=1,2.

So we have

		$\displaystyle\sum\limits_{l=1}^{N}c_{l}\sum\limits_{j=1}^{k}\int_{\mathbb{R}^{% N}}(W_{1,x_{j}}^{2^{}-2}Y_{j,l}\langle y,\nabla W_{1,\lambda}\rangle,W_{2,x_{% j}}^{2^{}-2}Z_{j,l}\langle y,\nabla W_{2,\lambda}\rangle)$		(3.10)
	$\displaystyle=$	$\displaystyle\sum\limits_{j=1}^{k}\int_{\mathbb{R}^{N}}(W_{1,x_{j}}^{2^{}-2}Y% _{j,2}\langle{y^{\prime}},\nabla_{y^{\prime}}W_{1,\lambda}\rangle,W_{2,x_{j}}^% {2^{}-2}Z_{j,2}\langle y,\nabla W_{2,\lambda}\rangle)c_{2}+o(1)\sum\limits_{t% \neq 2}\|c_{t}\|$		(3.11)

and

		$\displaystyle\sum\limits_{l=1}^{N}\sum\limits_{j=1}^{k}\Bigl{(}W_{1,x_{j}}^{2^% {}-2}Y_{j,l}\frac{\partial W_{1,\lambda}}{\partial y_{i}},W_{2,x_{j}}^{2^{}-% 2}Z_{j,l}\frac{\partial W_{2,\lambda}}{\partial y_{i}}\Bigr{)}$		(3.12)
	$\displaystyle=$	$\displaystyle\sum\limits_{j=1}^{k}\int_{\mathbb{R}^{N}}\Bigl{(}W_{1,x_{j}}^{2^% {}-2}Y_{j,l}\frac{\partial W_{1,\lambda}}{\partial y_{i}},W_{2,x_{j}}^{2^{}-% 2}Z_{j,2}\Bigr{)}c_{i}+o(1)\sum\limits_{t\neq i}^{i=3,\cdots,N}\|c_{t}\|.$		(3.13)

Since

\displaystyle\sum\limits_{j=1}^{k}\int_{\mathbb{R}^{N}}\Bigl{(}W_{1,x_{j}}Y_{j% ,2}\langle y^{\prime},\nabla_{y^{\prime}}W_{1,\lambda}\rangle,W_{2,x_{j}}Z_{j,% 2}\langle y^{\prime},\nabla_{y^{\prime}}W_{2,\lambda}\rangle\Bigr{)}=k(a_{1}+o% (1))\lambda^{2}

and

\displaystyle\sum\limits_{j=1}^{k}\int_{\mathbb{R}^{N}}\Bigl{(}W_{1,x_{j}}^{2^% {*}-2}Y_{j,i}\frac{\partial W_{1,\lambda}}{\partial y_{i}},W_{2,x_{j}}^{2^{*}-% 2}Z_{j,i}\frac{\partial W_{2,\lambda}}{\partial y_{i}}\Bigr{)}=k(a_{2}+o(1))% \lambda^{2}

for some constants $a_{1}\neq 0$ and $a_{2}\neq 0$ , we find from (3.10)-(3.12)

\displaystyle c_{i}=o\Bigl{(}\frac{1}{\lambda^{2}}\Bigr{)}c_{1},\,\,\,i=1,2,% \cdots,N.

Now, we have

	$\displaystyle 0=\sum\limits_{l=1}^{N}c_{l}\sum\limits_{j=1}^{k}\int_{\mathbb{R% }^{N}}W_{1,x_{j}}^{2^{*}-2}Y_{j,l}\frac{\partial W_{1,\lambda}}{\partial% \lambda}=$	$\displaystyle\sum\limits_{j=1}^{k}\int_{\mathbb{R}^{N}}W_{1,x_{j}}^{2^{*}-2}Y_% {j,1}\frac{\partial W_{1,\lambda}}{\partial\lambda}c_{1}+O\Bigl{(}\frac{k}{% \lambda^{2}}c_{1}\Bigr{)}$
	$\displaystyle=$	$\displaystyle k(a_{3}+o(1))c_{1}+O\Bigl{(}\frac{k}{\lambda^{2}}\Bigr{)}c_{1}$

for some constant $a_{3}\neq 0$ . So $c_{1}=0$ . ∎

Lemma 3.2.

We have

	$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}-\Delta u_{k}+P(r,y^{\prime\prime})u% _{k}-u_{k}^{2^{}-1}-\frac{\beta}{2}u_{k}^{\frac{2^{}}{2}-1}v_{k}^{\frac{2^{*% }}{2}}\Bigr{)}\frac{\partial W_{1,\lambda}}{\partial\lambda}$	(3.14)
	$\displaystyle+\int_{\mathbb{R}^{N}}\Bigl{(}-\Delta v_{k}+Q(r,y^{\prime\prime})% v_{k}-v_{k}^{2^{}-1}-\frac{\beta}{2}v_{k}^{\frac{2^{}}{2}-1}u_{k}^{\frac{2^{% *}}{2}}\Bigr{)}\frac{\partial W_{2,\lambda}}{\partial\lambda}$	(3.15)
$\displaystyle=$	$\displaystyle k\Bigl{[}-\frac{B_{1}}{\lambda^{3}}P(r,y^{\prime\prime})-\frac{B% _{2}}{\lambda^{3}}Q(r,y^{\prime\prime})+\sum\limits_{j=1}^{k}\frac{(C_{1}+% \beta C_{2})(N-2)}{\lambda^{N-1}\|x_{1}-x_{j}\|^{N-2}}+O\Bigl{(}\frac{1}{\lambda% ^{3+\epsilon}}\Bigr{)}\Bigr{]},$	(3.16)

where $B_{1}=\displaystyle\int_{\mathbb{R}^{N}}U_{0,1}^{2},\,B_{2}=\int_{\mathbb{R}^{% N}}V_{0,1}^{2}$ and $C_{1}+\beta C_{2}>0$ are integer.

Proof.

We have

	$\displaystyle\hbox{left the hand side of}\,\eqref{wqf3.1}=$	$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}-\Delta W_{1,\lambda}+P(r,y^{\prime% \prime})W_{1,\lambda}-W_{1,\lambda}^{2^{}-1}-\frac{\beta}{2}W_{1,\lambda}^{% \frac{2^{}}{2}-1}W_{2,\lambda}^{\frac{2^{*}}{2}}\Bigr{)}\frac{\partial W_{1,% \lambda}}{\partial\lambda}\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,% \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$
		$\displaystyle+\int_{\mathbb{R}^{N}}\Bigl{(}-\Delta W_{2,\lambda}+Q(r,y^{\prime% \prime})W_{2,\lambda}-W_{2,\lambda}^{2^{}-1}-\frac{\beta}{2}W_{1,\lambda}^{% \frac{2^{}}{2}}W_{2,\lambda}^{\frac{2^{*}}{2}-1}\Bigr{)}\frac{\partial W_{2,% \lambda}}{\partial\lambda}$
		$\displaystyle+\int_{\mathbb{R}^{N}}\Bigl{[}\Bigl{(}-\Delta\varphi+P(r,y^{% \prime\prime})\varphi-(2^{}-1)W_{1,\lambda}^{2^{}-2}\varphi\Bigr{)}\frac{% \partial W_{1,\lambda}}{\partial\lambda}$
		$\displaystyle+\Bigl{(}-\Delta\psi+Q(r,y^{\prime\prime})\psi-(2^{}-1)W_{2,% \lambda}^{2^{}-2}\psi\Bigr{)}\frac{\partial W_{2,\lambda}}{\partial\lambda}% \Bigr{]}$
		$\displaystyle-\int_{\mathbb{R}^{N}}\Bigl{[}\Bigl{(}(W_{1,\lambda}+\varphi)^{2^% {}-1}-W_{1,\lambda}^{2^{}-1}-(2^{}-1)W_{1,\lambda}^{2^{}-2}\varphi\Bigr{)}% \frac{\partial W_{1,\lambda}}{\partial\lambda}$
		$\displaystyle-\Bigl{(}(W_{2,\lambda}+\psi)^{2^{}-1}-W_{2,\lambda}^{2^{}-1}-(% 2^{}-1)W_{2,\lambda}^{2^{}-2}\psi\Bigr{)}\frac{\partial W_{2,\lambda}}{% \partial\lambda}\Bigr{]}$
		$\displaystyle+\frac{\beta}{2}\int_{\mathbb{R}^{N}}\Bigl{[}\Bigl{(}(W_{1,% \lambda}+\varphi)^{\frac{2^{}}{2}-1}(W_{2,\lambda}+\psi)^{\frac{2^{}}{2}}-W_% {1,\lambda}^{\frac{2^{}}{2}-1}W_{2,\lambda}^{\frac{2^{}}{2}}\Bigr{)}\frac{% \partial W_{1,\lambda}}{\partial\lambda}$
		$\displaystyle+\Bigl{(}(W_{1,\lambda}+\varphi)^{\frac{2^{}}{2}}(W_{2,\lambda}+% \psi)^{\frac{2^{}}{2}-1}-W_{1,\lambda}^{\frac{2^{}}{2}}W_{2,\lambda}^{\frac{% 2^{}}{2}-1}\Bigr{)}\frac{\partial W_{2,\lambda}}{\partial\lambda}\Bigr{]}$
	$\displaystyle=$	$\displaystyle:I_{1}+I_{2}+I_{3}+I_{4}.$

By Lemma A.1, we obtain

\displaystyle I_{1}=k\Bigl{(}-\frac{B_{1}}{\lambda^{3}}P(r,y^{\prime\prime})-% \frac{B_{2}}{\lambda^{3}}Q(r,y^{\prime\prime})+\sum\limits_{j=2}^{k}\frac{B_{2% }}{\lambda^{N-1}|x_{1}-x_{j}|^{N-2}}\Bigr{)}+O\Bigl{(}\frac{1}{\lambda^{3+% \lambda}}\Bigr{)}.

(3.17)

Using (2.12) and (2.13), we have

\displaystyle\begin{split}I_{2}&=\int_{\mathbb{R}^{N}}\Bigl{(}-\Delta\varphi+P% (r,y^{\prime\prime})\varphi-(2^{*}-1)W_{1,\lambda}^{2^{*}-2}\varphi\Bigr{)}% \frac{\partial W_{1,\lambda}}{\partial\lambda}\\ &\quad+\Bigl{(}-\Delta\psi+Q(r,y^{\prime\prime})\psi-(2^{*}-1)W_{2,\lambda}^{2% ^{*}-2}\psi\Bigr{)}\frac{\partial W_{2,\lambda}}{\partial\lambda}\\ &=O\Bigl{(}\frac{\|(\varphi,\psi)\|_{*}}{\lambda^{2+\epsilon}}\Bigr{)}=O\Bigl{% (}\frac{1}{\lambda^{3+\epsilon}}\Bigr{)}.\end{split}

(3.18)

Supposing that $N\geq 6$ and denoting $I_{3}=I_{31}+I_{32}$ , then we have

$\displaystyle\|I_{31}\|=$	$\displaystyle\Bigl{\|}\int_{\mathbb{R}^{N}}\Bigl{(}(W_{1,\lambda}+\varphi)^{2^{% }-1}-W_{1,\lambda}^{2^{}-1}-(2^{}-1)W_{1,\lambda}^{2^{}-2}\varphi\Bigr{)}% \frac{\partial W_{1,\lambda}}{\partial\lambda}\Bigr{\|}$	(3.19)
$\displaystyle\leq$	$\displaystyle C\int_{\mathbb{R}^{N}}W_{1,\lambda}^{2^{}-3}\varphi^{2}\Bigl{\|}% \frac{\partial W_{1,\lambda}}{\partial\lambda}\Bigr{\|}\leq\frac{C}{\lambda}% \int_{\mathbb{R}^{N}}\Bigl{(}\sum\limits_{j=1}^{k}W_{1,x_{j}}\Bigr{)}^{2^{}-2% }\varphi^{2}$	(3.20)
$\displaystyle\leq$	$\displaystyle\frac{\\|\varphi\\|_{}^{2}}{\lambda}\int_{\mathbb{R}^{N}}\Bigl{(}% \sum\limits_{j=1}^{k}W_{1,x_{j}}\Bigr{)}^{2^{}-2}\Bigl{(}\sum\limits_{i=1}^{k% }\frac{\lambda^{\frac{N-2}{2}}}{1+\lambda\|y-x_{i}\|^{\frac{N-2}{2}+\tau}}\Bigr{% )}^{2}$	(3.21)
$\displaystyle=$	$\displaystyle O\Bigl{(}\frac{k}{\lambda^{3+\epsilon}}\Bigr{)}.$	(3.22)

Similarly, for $N=5$ , we have

\displaystyle|I_{31}|\leq\int_{\mathbb{R}^{N}}\Bigl{(}W_{1,\lambda}^{2^{*}-3}% \varphi^{2}\Bigl{|}\frac{\partial W_{1,\lambda}}{\partial\lambda}\Bigr{|}+|% \varphi|^{2^{*}}\frac{\partial W_{1,\lambda}}{\partial\lambda}\Bigr{)}=O\Bigl{% (}\frac{k}{\lambda^{3+\epsilon}}\Bigr{)}.

(3.23)

Similarly, we have

\displaystyle|I_{32}|=\int_{\mathbb{R}^{N}}\Bigl{(}(W_{2,\lambda}+\psi)^{2^{*}% -1}-W_{2,\lambda}^{2^{*}-1}-(2^{*}-1)W_{2,\lambda}^{2^{*}-2}\psi\Bigr{)}\frac{% \partial W_{2,\lambda}}{\partial\lambda}=O\Bigl{(}\frac{k}{\lambda^{3+\epsilon% }}\Bigr{)},

(3.24)

then $|I_{3}|=|I_{31}+I_{32}|=O\Bigl{(}\frac{k}{\lambda^{3+\epsilon}}\Bigr{)}$ .

On the other hand, we have

$\displaystyle I_{41}=$	$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}(W_{1,\lambda}+\varphi)^{\frac{2^{}% }{2}-1}(W_{2,\lambda}+\psi)^{\frac{2^{}}{2}}-W_{1,\lambda}^{\frac{2^{}}{2}}W% _{2,\lambda}^{\frac{2^{}}{2}}\Bigr{)}\frac{\partial W_{1,\lambda}}{\partial\lambda}$	(3.25)
$\displaystyle\leq$	$\displaystyle C\int_{\mathbb{R}^{N}}\Bigl{(}W_{1,\lambda}^{2^{}-2}\varphi+W_{% 2,\lambda}^{2^{}-3}\varphi\psi+W_{1,\lambda}^{2^{*}-2}\psi\Bigr{)}\frac{% \partial W_{1,\lambda}}{\partial\lambda}$	(3.26)
$\displaystyle\leq$	$\displaystyle\frac{1}{\lambda}\int_{\mathbb{R}^{N}}\Bigl{(}\sum\limits_{j=1}^{% k}W_{1,x_{j}}\Bigr{)}^{2^{*}-1}\varphi=O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}% \Bigr{)}.$	(3.27)

Similarly, we can also estimate

\displaystyle I_{42}=\int_{\mathbb{R}^{N}}\Bigl{(}(W_{1,\lambda}+\varphi)^{% \frac{2^{*}}{2}}(W_{2,\lambda}+\psi)^{\frac{2^{*}}{2}-1}-W_{1,\lambda}^{\frac{% 2^{*}}{2}}W_{2,\lambda}^{\frac{2^{*}}{2}-1}\Bigr{)}\frac{\partial W_{2,\lambda% }}{\partial\lambda}=O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}\Bigr{)}.

(3.28)

Combining (3.17)-(3.28), we complete the proof. ∎

Next, we estimate (3.1) and (3.3). Let us point out that the left hand side of (3.3) is the local Pohozaev identities generating from translation. We integrate by parts to find that (3.1) is equivalent to

\displaystyle\int_{D_{\rho}}\frac{\partial P(r,y^{\prime\prime})}{\partial y_{% i}}u_{k}^{2}=O\Bigl{(}\int_{\partial D_{\rho}}|\nabla\varphi|^{2}+\varphi^{2}+% |\varphi|^{2^{*}}\Bigr{)},\,\,\,\int_{D_{\rho}}\frac{\partial Q(r,y^{\prime% \prime})}{\partial y_{i}}v_{k}^{2}=O\Bigl{(}\int_{\partial D_{\rho}}|\nabla% \psi|^{2}+\psi^{2}+|\psi|^{2^{*}}\Bigr{)},

(3.29)

since $(u_{k},v_{k})=(\varphi,\psi)$ on $\partial D_{\rho}$ .

On the other hand, we know that (3.1) is equivalent to

	$\displaystyle\frac{N-2}{2}\int_{\Omega}\Bigl{(}\|\nabla u\|^{2}+\|\nabla v\|^{2}% \Bigr{)}-\frac{1}{2}\int_{\Omega}(NP(y)+\langle y,\nabla P\rangle)u^{2}+\frac{% N}{2^{}}\int_{\Omega}\Bigl{(}u^{2^{}}+v^{2^{}}+\beta u^{\frac{2^{}}{2}}v^{% \frac{2^{*}}{2}}\Bigr{)}$	(3.30)
	$\displaystyle-\frac{1}{2}\int_{\Omega}(NQ(y)+\langle y,\nabla Q\rangle)v^{2}$	(3.31)
$\displaystyle=$	$\displaystyle O\Bigl{(}\int_{\partial\Omega}\|\nabla\varphi\|^{2}+\varphi^{2}+\|% \varphi\|^{2^{}}+\|\nabla\psi\|^{2}+\psi^{2}+\|\psi\|^{2^{}}\Bigr{)}.$	(3.32)

From (2.20), we obtain

	$\displaystyle\int_{D_{\rho}}\|\nabla u_{k}\|^{2}+P(y)u_{k}^{2}=$	$\displaystyle\int_{D_{\rho}}u_{k}^{2^{}}+\frac{\beta}{2}\int_{D_{\rho}}u_{k}^% {\frac{2^{}}{2}}v_{k}^{\frac{2^{}}{2}}+\sum\limits_{l=1}^{N}c_{l}\sum\limits% _{j=1}^{k}\int_{\mathbb{R}^{N}}W_{1,x_{j}}^{2^{}-2}Y_{j,l}u_{k}$		(3.33)
		$\displaystyle+O\Bigl{(}\int_{\partial D_{\rho}}\|\nabla\varphi\|^{2}+\varphi^{2}% \Bigr{)}$		(3.34)

and

	$\displaystyle\int_{D_{\rho}}\|\nabla v_{k}\|^{2}+Q(y)v_{k}^{2}=$	$\displaystyle\int_{D_{\rho}}v_{k}^{2^{}}+\frac{\beta}{2}\int_{D_{\rho}}u_{k}^% {\frac{2^{}}{2}}v_{k}^{\frac{2^{}}{2}}+\sum\limits_{l=1}^{N}c_{l}\sum\limits% _{j=1}^{k}\int_{\mathbb{R}^{N}}W_{2,x_{j}}^{2^{}-2}Z_{j,l}u_{k}$		(3.35)
		$\displaystyle+O\Bigl{(}\int_{\partial D_{\rho}}\|\nabla\psi\|^{2}+\psi^{2}\Bigr{% )}.$		(3.36)

Using (3.30)-(3.35), we can get the following estimate for $c_{i}$ ,

\displaystyle c_{i}=O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}\Bigr{)},\,\,\,i=2,% \cdots,N,\,\,\,c_{1}=o\Bigl{(}\frac{1}{\lambda}\Bigr{)}.

(3.37)

Noting

\displaystyle\int_{\mathbb{R}^{N}}W_{1,x_{j}}^{2^{*}-1}Y_{j,l}=O\Bigl{(}\frac{% 1}{\lambda^{N}}\Bigr{)},\,\,\,\int_{\mathbb{R}^{N}}W_{2,x_{j}}Z_{j,l}=O\Bigl{(% }\frac{1}{\lambda^{N}}\Bigr{)},

(3.38)

we find from (3.37) and (3.38) that

		$\displaystyle\int_{D_{\rho}}P(y)+\frac{1}{2}\langle y,\nabla P(y)\rangle u_{k}% ^{2}+Q(y)+\frac{1}{2}\langle y,\nabla Q(y)\rangle v_{k}^{2}$		(3.39)
	$\displaystyle=$	$\displaystyle O\Bigl{(}\frac{k}{\lambda^{2+\epsilon}}\Bigr{)}+O\Bigl{(}\int_{% \partial D_{\rho}}\|\nabla\varphi\|^{2}+\varphi^{2}+\|\varphi\|^{2^{}}+\|\nabla% \psi\|^{2}+\psi^{2}+\|\psi\|^{2^{}}\Bigr{)}$		(3.40)

for some small $\epsilon>0$ .

From (3.29), we can rewrite (3.39),

		$\displaystyle\int_{D_{\rho}}\Bigl{(}P(y)+\frac{1}{2}r\frac{\partial P(r,y^{% \prime\prime})}{\partial r}\Bigr{)}u_{k}^{2}+\Bigl{(}Q(y)+\frac{1}{2}r\frac{% \partial Q(r,y^{\prime\prime})}{\partial r}\Bigr{)}v_{k}^{2}$
	$\displaystyle=$	$\displaystyle O\Bigl{(}\frac{k}{\lambda^{2+\epsilon}}\Bigr{)}+O\Bigl{(}\int_{% \partial D_{\rho}}\|\nabla\varphi\|^{2}+\varphi^{2}+\|\varphi\|^{2^{}}+\|\nabla% \psi\|^{2}+\psi^{2}+\|\psi\|^{2^{}}\Bigr{)}.$

That is

\displaystyle\int_{D_{\rho}}\frac{1}{2r}\Bigl{(}\frac{\partial r^{2}P(r,y^{% \prime\prime})}{\partial r}u_{k}^{2}+\frac{\partial r^{2}Q(r,y^{\prime\prime})% }{\partial r}v_{k}^{2}\Bigr{)}=O\Bigl{(}\frac{k}{\lambda^{2}}+\int_{\partial D% _{\rho}}|\nabla\varphi|^{2}+\varphi^{2}+|\varphi|^{2^{*}}+|\nabla\psi|^{2}+% \psi^{2}+|\psi|^{2^{*}}\Bigr{)}.

Lemma 3.3.

It holds

\displaystyle\int_{\mathbb{R}^{N}}|\nabla\varphi|^{2}+P(r,y^{\prime\prime})|% \varphi|^{2}+|\nabla\psi|^{2}+Q(r,y^{\prime\prime})\psi^{2}=O\Bigl{(}\frac{k}{% \lambda^{2+\epsilon}}\Bigr{)}.

Proof.

By the system, we prove

		$\displaystyle\int_{\mathbb{R}^{N}}\|\nabla\varphi\|^{2}+P(r,y^{\prime\prime})% \varphi^{2}+\|\nabla\psi\|^{2}+Q(r,y^{\prime\prime})\psi^{2}$
	$\displaystyle=$	$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{[}(W_{1,\lambda}+\varphi)^{2^{}-1}+(W% _{2,\lambda}+\psi)^{2^{}-1}+\frac{\beta}{2}\Bigl{(}(W_{1,\lambda}+\varphi)^{% \frac{2^{}}{2}-1}(W_{2,\lambda}+\psi)^{\frac{2^{}}{2}}$
		$\displaystyle+(W_{1,\lambda}+\varphi)^{\frac{2^{}}{2}}(W_{2,\lambda}+\psi)^{% \frac{2^{}}{2}-1}\Bigr{)}-P(r,y^{\prime\prime})W_{1,\lambda}-Q(r,y^{\prime% \prime})W_{2,\lambda}+\Delta W_{1,\lambda}\varphi+\Delta W_{2,\lambda}\psi% \Bigr{]}$
	$\displaystyle=$	$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}(W_{1,\lambda}+\varphi)^{2^{}-1}-W_% {1,\lambda}^{2^{}-1}\Bigr{)}\varphi+\Bigl{(}(W_{2,\lambda}+\psi)^{2^{}-1}-W_% {2,\lambda}^{2^{}-1}\Bigr{)}\psi$
		$\displaystyle+\frac{\beta}{2}\int_{\mathbb{R}^{N}}\Bigl{(}(W_{1,\lambda}+% \varphi)^{\frac{2^{}}{2}-1}(W_{2,\lambda}+\psi)^{\frac{2^{}}{2}}\varphi-W_{1% ,\lambda}^{\frac{2^{}}{2}-1}W_{2,\lambda}^{\frac{2^{}}{2}}\varphi$
		$\displaystyle+(W_{1,\lambda}+\varphi)^{\frac{2^{}}{2}}(W_{2,\lambda}+\psi)^{% \frac{2^{}}{2}-1}\psi-W_{1,\lambda}^{\frac{2^{}}{2}}W_{2,\lambda}^{\frac{2^{% }}{2}-1}\psi\Bigr{)}$
		$\displaystyle-\int_{\mathbb{R}^{N}}P(r,y^{\prime\prime})W_{1,\lambda}\varphi+Q% (r,y^{\prime\prime})W_{2,\lambda}\psi$
		$\displaystyle+\int_{\mathbb{R}^{N}}\Bigl{(}\Delta W_{1,\lambda}-W_{1,\lambda}^% {2^{}-1}-\frac{\beta}{2}W_{1,\lambda}^{\frac{2^{}}{2}-1}W_{2,\lambda}^{\frac% {2^{}}{2}}\Bigr{)}\varphi+\Bigl{(}\Delta W_{2,\lambda}-W_{2,\lambda}^{2^{}-1% }-\frac{\beta}{2}W_{1,\lambda}^{\frac{2^{}}{2}}W_{2,\lambda}^{\frac{2^{}}{2}% -1}\Bigr{)}\psi$
	$\displaystyle=$	$\displaystyle:M_{1}+M_{2}+M_{3}+M_{4}.$

We have

$\displaystyle\|M_{1}\|\leq$	$\displaystyle C\int_{\mathbb{R}^{N}}W_{1,\lambda}^{2^{}-2}\varphi^{2}+\|% \varphi\|^{2^{}}+W_{2,\lambda}^{2^{}-2}\psi^{2}+\|\psi\|^{2^{}}$	(3.41)
$\displaystyle\leq$	$\displaystyle C\\|(\varphi,\psi)\\|_{}^{2}\int_{\mathbb{R}^{N}}W_{1,\lambda}^{2% ^{}-2}\Bigl{(}\sum\limits_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda\|% y-x_{j}\|)^{\frac{N-2}{2}+\tau}}\Bigr{)}^{2}$	(3.42)
	$\displaystyle+\\|(\varphi,\psi)\\|_{}^{2^{}}\int_{\mathbb{R}^{N}}\Bigl{(}\sum% \limits_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda\|y-x_{j}\|)^{\frac{N-% 2}{2}+\tau}}\Bigr{)}^{2^{*}}$	(3.43)
$\displaystyle\leq$	$\displaystyle C\Bigl{(}\\|(\varphi,\psi)\\|_{}^{2}+\\|(\varphi,\psi)\\|_{}^{2^{% }}\Bigr{)}\int_{\mathbb{R}^{N}}\Bigl{(}\sum\limits_{j=1}^{k}\frac{\lambda^{% \frac{N-2}{2}}}{(1+\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\tau}}\Bigr{)}^{2^{}}$	(3.44)
$\displaystyle\leq$	$\displaystyle Ck\\|(\varphi,\psi)\\|_{}^{2}\int_{\mathbb{R}^{N}}\Bigl{(}\frac{% \lambda^{\frac{N-2}{2}}}{(1+\lambda\|y-x_{1}\|)^{\frac{N-2}{2}+\theta}}\Bigr{)}^% {2^{}}\Bigl{(}1+\sum\limits_{j=2}^{k}\frac{1}{(\lambda\|x_{j}-x_{1}\|)^{\tau-% \theta}}\Bigr{)}^{2^{*}}$	(3.45)
$\displaystyle\leq$	$\displaystyle Ck\\|(\varphi,\psi)\\|_{}^{2}\lambda^{2^{}\theta}\leq\frac{Ck}{% \lambda^{2+2\epsilon-2^{*}\theta}},$	(3.46)

where $\theta>0$ is any fixed small constant. Similar as (3.41), for $W_{2,\lambda}=\kappa W_{1,\lambda}$ , we obtain

\displaystyle|M_{2}|\leq Ck\|(\varphi,\psi)\|_{*}^{2}\lambda^{2^{*}\theta}\leq% \frac{Ck}{\lambda^{2+2\epsilon-2^{*}\theta}}.

(3.47)

Similar to the proof of (2.37), we can deduce

$\displaystyle\|M_{3}\|\leq$	$\displaystyle C\\|(\varphi,\psi)\\|_{*}\int_{\mathbb{R}^{N}}\sum\limits_{j=1}^{m% }\frac{\lambda^{\frac{N-2}{2}}\xi}{(1+\lambda\|y-x_{j}\|)^{N-2}}\sum\limits_{i=1% }^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda\|y-x_{i}\|)^{\frac{N-2}{2}+\tau}}$	(3.48)
$\displaystyle\leq$	$\displaystyle C\frac{\\|(\varphi,\psi)\\|_{*}}{\lambda^{1+\epsilon}}\int_{% \mathbb{R}^{N}}\sum\limits_{j=1}^{k}\frac{\lambda^{\frac{N+2}{2}}}{(1+\lambda\|% y-x_{j}\|)^{\frac{N+2}{2}+\tau}}\sum\limits_{i=1}^{k}\frac{\lambda^{\frac{N-2}{% 2}}}{(1+\lambda\|y-x_{i}\|)^{\frac{N-2}{2}+\tau}}$	(3.49)
$\displaystyle\leq$	$\displaystyle\frac{\\|(\varphi,\psi)\\|_{*}}{\lambda^{1+\epsilon}}\leq\frac{Ck}{% \lambda^{2+2\epsilon}}.$	(3.50)

Moreover, from the estimates of $J_{1},\,J_{3},\,J_{4}$ in Lemma A.5, we obtain

	$\displaystyle\|M_{4}\|\leq$	$\displaystyle C\frac{\\|(\varphi,\psi)\\|_{*}}{\lambda^{1+\epsilon}}\int_{% \mathbb{R}^{N}}\sum\limits_{j=1}^{k}\frac{\lambda^{\frac{N+2}{2}}}{(1+\lambda\|% y-x_{j}\|)^{\frac{N+2}{2}+\tau}}\sum\limits_{i=1}^{k}\frac{\lambda^{\frac{N-2}{% 2}}}{(1+\lambda\|y-x_{i}\|)^{\frac{N-2}{2}+\tau}}$		(3.51)
	$\displaystyle\leq$	$\displaystyle C\frac{k\\|(\varphi,\psi)\\|_{*}}{\lambda^{1+\epsilon}}\leq C\frac% {k}{\lambda^{2+2\epsilon}}.$		(3.52)

Combining (3.41)-(3.51), we deduce

\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}|\nabla\varphi|^{2}+P(r,y^{\prime% \prime})\varphi^{2}+|\nabla\psi|^{2}+Q(r,y^{\prime\prime})\psi^{2}\Bigr{)}=O% \Bigl{(}\frac{k}{\lambda^{2+\epsilon}}\Bigr{)}.

∎

Lemma 3.4.

For any $C^{1}$ function $g(r,y^{\prime\prime})$ , it holds

\displaystyle\int_{D_{\rho}}g(r,y^{\prime\prime})u_{k}^{2}=k\Bigl{(}\frac{1}{% \lambda^{2}}g(r,y^{\prime\prime})\int_{\mathbb{R}^{N}}U_{0,1}^{2}+o\Bigl{(}% \frac{1}{\lambda^{2}}\Bigr{)}\Bigr{)}.

Proof.

The proof is similar to Lemma 3.3 of [23], so we omit it. ∎

Lemma 3.5.

It holds

\displaystyle\int_{D_{4\delta}\setminus D_{3\delta}}(|\nabla\varphi|^{2}+|% \nabla\psi|^{2})=O\Bigl{(}\frac{k}{\lambda^{2+\epsilon}}\Bigr{)}.

Proof.

Noting that $W_{1,\lambda}=0$ in $D_{5\delta}\setminus D_{2\delta}$ , we find from (2.37) that

	$\displaystyle\int_{D_{4\delta}\setminus D_{3\delta}}(\|\nabla\varphi\|^{2}+\|% \nabla\psi\|^{2})\leq$	$\displaystyle C\int_{D_{5\delta}\setminus D_{2\delta}}(\|\varphi\|^{2}+\|\psi\|^{2% }+\|\varphi\|^{2^{}}+\|\psi\|^{2^{}})$
	$\displaystyle\leq$	$\displaystyle C\int_{D_{5\delta}\setminus D_{3\delta}}(\varphi^{2}+\psi^{2})=O% \Bigl{(}\frac{k}{\lambda^{2+\epsilon}}\Bigr{)}.$

∎

From Lemma 3.3 and Lemma 3.5, we know

\displaystyle\int_{D_{4\delta}\setminus D_{3\delta}}\Bigl{(}|\nabla\varphi|^{2% }+|\varphi|^{2}+|\varphi|^{2^{*}}+|\nabla\psi|^{2}+|\psi|^{2}+|\psi|^{2^{*}}% \Bigr{)}=O\Bigl{(}\frac{k}{\lambda^{2+\epsilon}}\Bigr{)}.

As a result, we can find a $\rho\in(3\delta,4\delta)$ , such that

\displaystyle\int_{\partial D_{\rho}}\Bigl{(}|\nabla\varphi|^{2}+|\varphi|^{2}% +|\varphi|^{2^{*}}+|\nabla\psi|^{2}+|\psi|^{2}+|\psi|^{2^{*}}\Bigr{)}=O\Bigl{(% }\frac{k}{\lambda^{2+\epsilon}}\Bigr{)}.

(3.53)

Now we apply Lemma 3.3 and (3.53) to obtain

\displaystyle k\Bigl{[}\frac{s^{2}}{\lambda^{2}}\Bigl{(}\frac{\partial P(r,y^{% \prime\prime})}{\partial y_{i}}+\kappa^{2}\frac{\partial Q(r,y^{\prime\prime})% }{\partial y_{i}}\Bigr{)}\int_{\mathbb{R}^{N}}w_{0,1}^{2}+o\Bigl{(}\frac{1}{% \lambda^{2}}\Bigr{)}\Bigr{]}=o\Bigl{(}\frac{k}{\lambda^{2}}\Bigr{)},

and

\displaystyle k\Bigl{[}\frac{s^{2}}{\lambda^{2}}\frac{1}{2\bar{r}}\Bigl{(}% \frac{\partial\bar{r}^{2}P(r,y^{\prime\prime})}{\partial\bar{r}}+\kappa^{2}% \frac{\partial\bar{r}^{2}Q(r,y^{\prime\prime})}{\partial\bar{r}}\Bigr{)}\int_{% \mathbb{R}^{N}}w_{0,1}^{2}+o\Bigl{(}\frac{1}{\lambda^{2}}\Bigr{)}\Bigr{]}=o% \Bigl{(}\frac{k}{\lambda^{2}}\Bigr{)}.

Therefore, the equations to determine $(\bar{r},y^{\prime\prime})$ are

\displaystyle\frac{\partial(P(\bar{r},\bar{y}^{\prime\prime})+\kappa^{2}Q(\bar% {r},\bar{y}^{\prime\prime}))}{\partial\bar{y}_{i}}=o(1),\,\,\,\frac{\partial r% ^{2}P(\bar{r},\bar{y}^{\prime\prime})+\kappa^{2}\partial r^{2}Q(\bar{r},\bar{y% }^{\prime\prime})}{\partial\bar{r}}=o(1),

and

\displaystyle-\frac{B_{1}(P(\bar{r},\bar{y}^{\prime\prime})+\kappa^{2}Q(\bar{r% },\bar{y}^{\prime\prime}))}{\lambda^{3}}+\frac{(C_{1}+\beta\kappa^{\frac{2^{*}% }{2}}C_{1})k^{N-2}}{\lambda^{N-1}}=O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}% \Bigr{)}.

(3.54)

Letting $\lambda=tk^{\frac{N-2}{N-4}}$ , then $t\in[L_{0},L_{1}]$ since $\lambda\in[L_{0}k^{\frac{N-2}{N-4}},L_{1}k^{\frac{N-2}{N-4}}]$ . Then from (3.54), we get

\displaystyle-\frac{B_{1}(P(\bar{r},\bar{y}^{\prime\prime})+\kappa^{2}Q(\bar{r% },\bar{y}^{\prime\prime}))}{t^{3}}+\frac{(C_{1}+\beta\kappa^{\frac{2^{*}}{2}}C% _{1})k^{N-2}}{t^{N-1}}=o(1),\,\,\,t\in[L_{0},L_{1}].

Let

\displaystyle F(t,\bar{r},\bar{y}^{\prime\prime})=\Bigl{(}\nabla_{\bar{r},\bar% {y}^{\prime\prime}}(P+\kappa^{2}Q)(\bar{r},\bar{y}^{\prime\prime}),-\frac{B_{1% }}{t^{3}}+\frac{C_{1}+\beta\kappa^{\frac{2^{*}}{2}}C_{1}}{t^{N-1}}\Bigr{)},

(3.55)

then

\displaystyle deg\Bigl{(}F(t,\bar{r},\bar{y}^{\prime\prime}),[D_{1},D_{2}]% \times B_{\frac{1}{\mu^{1-\theta}}}(r_{0},y_{0}^{\prime\prime})\Bigr{)}=deg% \Bigl{(}\nabla_{\bar{r},\bar{y}^{\prime\prime}}(P+\kappa^{2}Q)(\bar{r},\bar{y}% ^{\prime\prime}),B_{\frac{1}{\mu^{1-\theta}}}(r_{0},y_{0}^{\prime\prime})\Bigr% {)}\neq 0.

So, (3.55) have a solution $t_{k}\in[L_{0},L_{1}]$ , $(\bar{r}_{k},\bar{y}^{\prime\prime})\in B_{\theta}(r_{0},y_{0}^{\prime\prime})$ .

Appendix A energy estimates and some known results

In this section, we mainly do some energy expansions and give some known results which are used before.

Lemma A.1.

If $N\geq 5$ , then

\displaystyle I_{1}=k\Bigl{(}-\frac{B_{1}}{\lambda^{3}}(P(r,y^{\prime\prime})+% \kappa^{2}Q(r,y^{\prime\prime}))+\sum\limits_{j=2}^{k}\frac{C_{1}(1+\beta% \kappa^{\frac{2^{*}}{2}})}{\lambda^{N-1}|x_{1}-x_{j}|^{N-2}}\Bigr{)}+O\Bigl{(}% \frac{1}{\lambda^{3+\epsilon}}\Bigr{)},

where $B_{1}=\displaystyle\int_{\mathbb{R}^{N}}w_{0,1}^{2}$ , $C_{1}=\displaystyle\int_{\mathbb{R}^{N}}w_{0,1}^{2^{*}}$ and $1+\beta\kappa^{\frac{2^{*}}{2}}>0$ .

Proof.

We have

\frac{\partial I(W_{1,x_{j}})}{\partial\lambda}=\frac{\partial W^{*}_{1,x_{j}}% }{\partial\lambda}+O\Bigl{(}\frac{k}{\lambda^{N-2}}\Bigr{)}.

Direct calculations show

\displaystyle-\Delta\sum\limits_{j=1}^{k}W^{*}_{1,x_{j}}=\sum\limits_{j=1}^{k}% {W^{*}}_{1,x_{j}}^{2^{*}-1}+\frac{\beta}{2}\sum\limits_{j=1}^{k}{W^{*}}_{1,x_{% j}}^{\frac{2^{*}}{2}-1}{W^{*}}_{2,x_{j}}^{\frac{2^{*}}{2}}

and

\displaystyle-\Delta\sum\limits_{j=1}^{k}W^{*}_{2,x_{j}}=\sum\limits_{j=1}^{k}% {W^{*}}_{2,x_{j}}^{2^{*}-1}+\frac{\beta}{2}\sum\limits_{j=1}^{k}{W^{*}}_{1,x_{% j}}^{\frac{2^{*}}{2}}{W^{*}}_{2,x_{j}}^{\frac{2^{*}}{2}-1}.

Then, we have

	$\displaystyle I_{1}=$	$\displaystyle\int_{\mathbb{R}^{N}}P(y)W^{}_{1,\lambda}\frac{\partial W^{}_{1% ,\lambda}}{\partial\lambda}-\Bigl{(}{W^{}}_{1,\lambda}^{2^{}-1}-\sum\limits_% {j=1}^{k}{W^{}}_{1,x_{j}}^{2^{}-1}\Bigr{)}\frac{\partial{W^{}}_{1,\lambda}}% {\partial\lambda}+Q(y)W^{}_{2,\lambda}\frac{\partial W^{*}_{2,\lambda}}{% \partial\lambda}$
		$\displaystyle-\Bigl{(}{W^{}}_{2,\lambda}^{2^{}-1}-\sum\limits_{j=1}^{k}{W^{% }}_{2,x_{j}}^{2^{}-1}\Bigr{)}\frac{\partial{W^{}}_{2,\lambda}}{\partial% \lambda}+\frac{\beta}{2}\int_{\mathbb{R}^{N}}\Bigl{(}{W^{}}_{1,\lambda}^{% \frac{2^{}}{2}-1}{W^{}}_{2,\lambda}^{\frac{2^{}}{2}}-\sum\limits_{j=1}^{k}{% W^{}}_{1,x_{j}}^{\frac{2^{}}{2}-1}{W^{}}_{2,x_{j}}^{\frac{2^{}}{2}}\Bigr{)% }\frac{\partial{W^{}}_{1,\lambda}}{\partial\lambda}$
		$\displaystyle+\Bigl{(}{W^{}}_{1,\lambda}^{\frac{2^{}}{2}}{W^{}}_{2,\lambda}% ^{\frac{2^{}}{2}-1}-\sum\limits_{j=1}^{k}{W^{}}_{1,x_{j}}^{\frac{2^{}}{2}}{% W^{}}_{2,x_{j}}^{\frac{2^{}}{2}-1}\Bigr{)}\frac{\partial{W^{*}}_{2,\lambda}}% {\partial\lambda}.$

It is easy to check that

		$\displaystyle\int_{\mathbb{R}^{N}}P(y)W^{}_{1,\lambda}\frac{\partial W^{}_{1% ,\lambda}}{\partial\lambda}=k\Bigl{[}\int_{\mathbb{R}^{N}}P(y)W^{}_{1,x_{1}}% \frac{\partial W^{}_{1,x_{1}}}{\partial\lambda}+O\Bigl{(}\frac{1}{\lambda}% \int_{\mathbb{R}^{N}}W^{}_{1,x_{1}}\sum\limits_{j=2}^{k}mW^{}_{1,x_{j}}\Bigr% {)}\Bigr{]}$
	$\displaystyle=$	$\displaystyle k\Bigl{[}-\frac{P(\bar{r},\bar{y}^{\prime\prime})}{\lambda^{3}}% \int_{\mathbb{R}^{N}}U_{0,1}^{2}+O\Bigl{(}\frac{1}{\lambda^{3}}\sum\limits_{j=% 1}^{k}\frac{1}{(\lambda\|x_{1}-x_{j}\|)^{N-4}}\Bigr{)}+\frac{1}{\lambda^{2+% \epsilon}}\Bigr{]}$
	$\displaystyle=$	$\displaystyle k\Bigl{[}-\frac{P(\bar{r},\bar{y}^{\prime\prime})}{\lambda^{3}}% \int_{\mathbb{R}^{N}}U_{0,1}^{2}+O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}\Bigr{% )}\Bigr{]}.$

Similarly, we have

\displaystyle\int_{\mathbb{R}^{N}}Q(y)W^{*}_{2,\lambda}\frac{\partial W^{*}_{2% ,\lambda}}{\partial\lambda}=k\Bigl{[}-\frac{Q(\bar{r},\bar{y}^{\prime\prime})}% {\lambda^{3}}\int_{\mathbb{R}^{N}}V_{0,1}^{2}+O\Bigl{(}\frac{1}{\lambda^{3+% \epsilon}}\Bigr{)}\Bigr{]}.

On the other hand, we have

		$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}{W^{}}_{1,\lambda}^{2^{}-1}-\sum% \limits_{j=1}^{k}{W^{}}_{1,x_{j}}^{2^{}-1}\Bigr{)}\frac{\partial W^{}_{1,% \lambda}}{\partial\lambda}=k\int_{\Omega_{1}}\Bigl{(}{W^{}}_{1,\lambda}^{2^{% }-1}-\sum\limits_{j=1}^{k}{W^{}}_{1,x_{j}}\Bigr{)}\frac{\partial{W^{*}}_{1,% \lambda}}{\partial\lambda}$
	$\displaystyle=$	$\displaystyle k\int_{\Omega_{1}}(2^{}-1){W^{}}_{1,x_{1}}^{2^{}-2}\sum% \limits_{j=2}^{k}{W^{}}_{1,x_{j}}\frac{\partial{W^{*}}_{1,x_{1}}}{\partial% \lambda}+O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}\Bigr{)}$
	$\displaystyle=$	$\displaystyle k\Bigl{(}-\sum\limits_{j=2}^{k}\frac{C_{1}}{\lambda^{(N-1)}\|x_{1% }-x_{j}\|^{N-2}}\Bigr{)}+O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}\Bigr{)}$

for some constant $C_{1}=\displaystyle\int_{\mathbb{R}^{N}}W_{0,1}^{2^{*}}>0$ .

In view of $W^{*}_{2,\lambda}=kW^{*}_{1,\lambda}$ , we estimate

	$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}{W^{}}_{1,\lambda}^{\frac{2^{}}{2}% -1}{W^{}}_{2,\lambda}^{\frac{2^{}}{2}}$	$\displaystyle-\sum\limits_{j=1}^{k}{W^{}}_{1,x_{j}}^{\frac{2^{}}{2}-1}{W^{}% }_{2,x_{j}}^{\frac{2^{}}{2}}\Bigr{)}\frac{\partial{W^{}}_{1,\lambda}}{% \partial\lambda}=O\int_{\mathbb{R}^{N}}\Bigl{(}{W^{}}_{1,\lambda}^{2^{}-1}-% \sum\limits_{j=1}^{k}{W^{}}_{1,x_{j}}^{2^{}-1}\Bigr{)}\frac{\partial{W^{}}_% {1,\lambda}}{\partial\lambda}$
	$\displaystyle=$	$\displaystyle kO\Bigl{(}-\sum\limits_{j=2}^{k}\frac{\kappa^{\frac{2^{}}{2}}% \displaystyle\int_{\mathbb{R}^{N}}W_{0,1}^{2^{}}}{\lambda^{N-1}\|x_{1}-x_{j}\|^% {N-2}}\Bigr{)}+O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}\Bigr{)}$

and

	$\displaystyle\int_{\mathbb{R}^{N}}\Bigl{(}W_{1,\lambda}^{\frac{2^{}}{2}}W_{2,% \lambda}^{\frac{2^{}}{2}-1}-$	$\displaystyle\sum\limits_{j=1}^{k}W_{1,x_{j}}^{\frac{2^{}}{2}}W_{2,x_{j}}^{% \frac{2^{}}{2}-1}\Bigr{)}\frac{\partial W_{2,\lambda}}{\partial\lambda}=O\int% _{\mathbb{R}^{N}}\Bigl{(}W_{1,\lambda}^{2^{}-1}-\sum\limits_{j=1}^{k}W_{1,x_{% j}}^{2^{}-1}\Bigr{)}\frac{\partial W_{2,\lambda}}{\partial\lambda}$
	$\displaystyle=$	$\displaystyle kO\Bigl{(}-\sum\limits_{j=2}^{k}\frac{\kappa^{\frac{2^{}}{2}% \displaystyle\int_{\mathbb{R}^{N}}W_{0,1}^{2^{}}}}{\lambda^{N-1}\|x_{1}-x_{j}\|% ^{N-2}}\Bigr{)}+O\Bigl{(}\frac{1}{\lambda^{3+\epsilon}}\Bigr{)}.$

So, we obtain

\displaystyle I_{1}=k\Bigl{(}-\frac{B_{1}}{\lambda^{3}}\bigl{(}P(\bar{r},\bar{% y}^{\prime\prime})+\kappa^{2}Q(\bar{r},\bar{y}^{\prime\prime})\bigr{)}+\sum% \limits_{j=2}^{k}\frac{C_{1}(1+\beta\kappa^{\frac{2^{*}}{2}})}{\lambda^{N-1}|x% _{1}-x_{j}|^{N-2}}\Bigr{)}+O\Bigl{(}\frac{1}{\lambda^{2+\epsilon}}\Bigr{)},

where $B_{1}=\displaystyle\int_{\mathbb{R}^{N}}W_{0,1}^{2}$ , $C_{1}=\displaystyle\int_{\mathbb{R}^{N}}W_{0,1}^{2^{*}}$ and $1+\beta\kappa^{\frac{2^{*}}{2}}>0$ . ∎

Lemma A.2.

If $N\geq 5$ , then

\displaystyle\frac{\partial I(W_{1,\lambda},W_{2,\lambda})}{\partial\bar{r}}=k% \Bigl{(}\frac{B_{1}}{\lambda^{2}}\frac{\partial(P(\bar{r},\bar{y}^{\prime% \prime})+\kappa^{2}Q(\bar{r},\bar{y}^{\prime\prime}))}{\partial\bar{r}}+\sum% \limits_{j=2}^{k}\frac{C_{1}(1+\beta\kappa^{\frac{2^{*}}{2}})}{\lambda^{N-2}|x% _{1}-x_{j}|^{N-2}}\Bigr{)}

and

\displaystyle\frac{\partial I(W_{1,\lambda},W_{2,\lambda})}{\partial\bar{y}^{% \prime\prime}}=k\Bigl{(}\frac{B_{1}}{\lambda^{2}}\frac{\partial(P(\bar{r},\bar% {y}^{\prime\prime})+\kappa^{2}Q(\bar{r},\bar{y}^{\prime\prime}))}{\partial\bar% {y}^{\prime\prime}_{k}}\Bigr{)}+O\Bigl{(}\frac{1}{\lambda^{1+\epsilon}}\Bigr{)},

where $B_{1}$ and $C_{1}$ are the same positive constants in Lemma A.1.

For each fixed $k$ and $j$ , we consider the following function

\displaystyle g_{k,j}=\frac{1}{(1+|y-x_{j}|)^{\alpha}}\frac{1}{(1+|y-x_{k}|)^{% \beta}},

where $\alpha\geq 1$ and $\beta\geq 1$ are two constants.

Lemma A.3.

([35], Lemma B.1) For any constants $0<\delta\leq\min\{\alpha,\beta\}$ , there is a constant $C>0$ , such that

\displaystyle g_{k,j}(y)\leq\frac{C}{|x_{k}-x_{j}|^{\delta}}\Bigl{(}\frac{1}{(% 1+|y-x_{k}|)^{\alpha+\beta-\delta}}+\frac{1}{(1+|y-x_{j}|)^{\alpha+\beta-% \delta}}\Bigr{)}.

Lemma A.4.

([35], Lemma B.2) For any constant $0<\delta<N-2$ , there is a constant $C>0$ , such that

\displaystyle\int_{\mathbb{R}^{N}}\frac{1}{|y-z|^{N-2}}\frac{1}{(1+|z|)^{2+% \delta}}\leq\frac{C}{(1+|y|)^{\delta}}.

Lemma A.5.

([35], Lemma B.3) Suppose that $N\geq 5$ , then there is a small constant $\theta>0$ , such that

\displaystyle\int_{\mathbb{R}^{N}}\frac{1}{|y-z|^{N-2}}W_{1,\lambda}^{\frac{4}% {N-2}}\sum\limits_{j=1}^{k}\frac{1}{(1+\lambda|z-x_{j}|)^{\frac{N-2}{2}+\tau}}% \leq\sum\limits_{j=1}^{k}\frac{C}{1+\lambda|y-x_{j}|^{\frac{N-2}{2}+\tau+% \theta}}.

Lemma A.6.

For any solution $(\varphi,\psi)$ to problem (2.20) with $\|(\varphi,\psi)\|_{*}<+\infty$ , there must hold further that

\displaystyle|\varphi|\leq\frac{1}{2}U,\,\,\,\,|\psi|\leq\frac{1}{2}V.

Proof.

Since the proof is similar as Lemma of [12], we omit it. ∎

Acknowledgement: The authors would like to thank Professor Chunhua Wang for the helpful discussion with her. This paper is supported by NSFC(No. 12071169 and No.12471106).

References

[1] A. Ambrosetti, E. Colorado, Bound and ground states of coupled nonlinear Schrödinger equations. C. R. Math. Acad. Sci. Paris, 342 (2006), 453-458.
[2] T. Bartsch, N. Dancer, Z. Wang, A Liouville theorem, a-priori bounds, and bifurcating branches of positive solutions for a nonlinear elliptic system. Calc. Var. Partial Diff. Equ., 37 (2010), 345-361.
[3] T. Bartsch, Z. Wang, Note on ground states of nonlinear Schrödinger systems. J. Part. Diff. Equ., 19 (2006), 200-207.
[4] H. Chen, M. Medina, A. Pistoia, Segregated solutions for a critical elliptic system with a small interspecies repulsive force. arXiv:2203.10990.
[5] H. Chen, A. Pistoia, G. Vaira, Segregated solutions for some non-linear Schrödinger systems with critical growth. arXiv:2206.09637.
[6] W. Chen, J. Wei, S. Yan, Infinitely many positive solutions for the Schrödinger equations in $\mathbb{R}^{N}$ with critical growth. J. Differ. Equ., 252 (2012), 2425-2447.
[7] E. N. Dancer, J. Wei, T. Weth, A priori bounds versus multiple existence of positive solutions for a nonlinear Schrödiner system. Ann. Inst. H. Poincaré Anal. Non Linéaire, 27 (2010), 953-969.
[8] M. delpino, P. Felmer, M. Musso, Two-bubble solutions in the supercritical Bahri-Coron’s problem. Calc. Var. Partial Diff. Equ., 16 (2003), 113-145.
[9] M. delpino, M. Musso, F. Pacard, A.Pistoia, Large energy entire solutions for the Yamabe equation. J. Diff. Equ., 251 (2011), 2568-2597.
[10] M. delpino, M. Musso, F. Pacard, A.Pistoia, Torus action on $S^{N}$ and sign changing solutions for conformally invariant equations. Ann.Super. Pisa Cl. Sci., 12 (2013), 209-237.
[11] F. Du, Q. Hua, C. Wang, Q. Wang, Multi-piece of bubble solutions for a nonlinear critical elliptic equation. J. Diff. Equ., 393 (2024), 102-138.
[12] Q. Guo, C. Wang, Q. Wang, Synchronized vector solutions for critical nonlinear elliptic systems in higher dimensions. Preprint.
[13] Y. Guo, B. Li, J. Wei, Entire nonradial solutions for non-cooperative coupled elliptic system with critical exponents in $\mathbb{R}^{3}$ . J. Diff. Equ., 256 (2014), 3463-3495.
[14] Q. He, C. Wang, Q. Wang, New type of positive bubble solutions for a critical Schrödinger equation. The Journal of Geometric Analysis, 32 (2022), 278.
[15] T. Li, J. Wei, Y. Wu, Infinitely many nonradail positive solutions for multi-species nonlinear Schrödinger system in $\mathbb{R}^{N}$ . J. Diff. Equ., 381(2024), 340-396.
[16] T. Lin, J. Wei, Spikes in two coupled nonlinear Schrödinger equations. Ann. Inst. H. Poincaré Anal. Non Linéaire, 22 (2005), 403-439.
[17] T. Lin, J. Wei, Ground state of $N$ coupled nonlinear Schrödinger equations in $\mathbb{R}^{n}$ , $n\leq 3$ . Comm. Math. Phys., 255 (2005), 629-653.
[18] T. C. Lin, J. Wei, Solitary and self-similar solutions of two-component systems of nonlinear Schrödinger equations. Physica D: Nonlinear Phenomena, 220 (2006), 99-115.
[19] S. Liu, C. Wang, Q. Wang, On vector solutions of nonlinear Schrödinger systems with mixed potentials. J. Diff. Equ., 411 (2024), 506-530.
[20] Z. Liu, Z. Wang, Multiple bound states and bound states of nonlinear Schrödinger systems. Comm. Math. Phys., 282 (2008), 721-731.
[21] S. Peng, Y. Peng, Z. Wang, On elliptic systems with Sobolev critical growth. Calc. Var. Partial Differential Equations, 55 (2016), 142.
[22] S. Peng, C. Wang, S. Wei, Constructing solutions for the prescribed scalar curvature problem via local Pohozaev identities. J. Differential Equations, 274 (2019), 2503-2530.
[23] S. Peng, C. Wang, S. Yan, Construction of solutions via local Pohozaev identities. J. Funct. Anal., 274 (2018), 2606-2633.
[24] S. Peng, Q. Wang, Z. Wang, On coupled nonlinear Schrödinger systems with mixed couplings. Transactions of the American Mathematical Society, (2019), 7559-7583.
[25] S. Peng, Z. Wang, Segregated and synchronized vector solutions for nonlinear Schrödinger systems. Arch. Rational Mech. Anal., 208 (2013), 305-339.
[26] A. Pistoia, G. Vaira, Segregated solutions for nonlinear Schrödinger systems with weak interspecies force. Comm. Partial Diff. Equ., 47(2022), 2146-2179.
[27] B. Sirakov, Least energy solitary waves for a system of nonlinear Schrödinger equations in $\mathbb{R}^{n}$ . Comm. Math. Phys., 271 (2007), 199-221.
[28] S. Terracini, G. Verzini, Multipulse phase in k-mixtures of Bose-Einstein condenstates. Arch. Rational Mech. Anal., 194 (2009), 717-741.
[29] R. Tian, Z. Wang, Multiple solitary wave solutions of nonlinear Schrödinger systems. Top Methods Nonlinear Anal., 37 (2011), 203-223.
[30] C, Wang, Q. Wang, J. Yang, Existence and properties of bubbling solutions for a critical nonlinear elliptic equation. J. Fixed Point Theory Apply, 25 (2023), 56.
[31] C. Wang, D. Xie, L. Zhan, L. Zhang, L. Zhao, Segregated vector solutions for nonlinear Schrödinger systems in $\mathbb{R}^{2}$ . (English summary) Acta Math. Sci. Ser. B (Engl. Ed.), 35 (2015), 383-398.
[32] Q. Wang, D. Ye, Infinitely many solutions with simulaeous synchronized and segregated components for nonlinear Schrödinger system. arXiv:2303.10324.
[33] J. Wei, Y. Wu, Groud states of nonlinear Schrödinger systems with mixed couplings. Journal de Math $\acute{e}$ matique Pures at Appliqu $\acute{e}$ s, 141 (2020), 50-88.
[34] J. Wei, S. Yan, Infinitely many positive solutions for the nonlinear Schrödinger equations in $\mathbb{R}^{N}$ . Calc. Var. Partial Differential Equations, 258 (2010), 423-439.
[35] J. Wei, S. Yan, Infinitely many solutions for the prescribed scalar curvature problem on $S^{N}$ . J. Funct. Anal., 258 (2010), 3048-3081.
[36] J. Wei, W. Yao, Uniqueness of positive solutions to some coupled nonlinear Schrödinger equations. Commun. Pure Appl. Anal., 11 (2012), 1003-1011.

$\displaystyle\varphi(y)$	$\displaystyle=\int_{\mathbb{R}^{N}}\frac{(2^{*}-1)}{\|y-z\|^{N-2}}W_{1,\lambda}^% {\frac{4}{N-2}}\varphi(z)dz$
	$\displaystyle\quad+\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}\frac{1}{(% N-2)}W_{1,\lambda}^{\frac{4-N}{N-2}}W_{2,\lambda}^{\frac{N}{N-2}}\varphi(z)+% \frac{N}{2(N-2)}W_{1,\lambda}^{\frac{2}{N-2}}W_{2,\lambda}^{\frac{2}{N-2}}\psi% (z)\Big{)}dz$
	$\displaystyle\quad+\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}h_{1}(z)+% \sum_{i=1}^{2}c_{i}\sum_{j=1}^{k}w_{x_{j},\lambda}^{2^{*}-2}Y_{j,i}\Big{)}dz$	(2.4)

$\displaystyle\psi(y)$	$\displaystyle=\int_{\mathbb{R}^{N}}\frac{(2^{*}-1)}{\|y-z\|^{N-2}}W_{2,\lambda}^% {\frac{4}{N-2}}\psi(z)dz$
	$\displaystyle\quad+\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}\frac{1}{(% N-2)}W_{2,\lambda}^{\frac{4-N}{N-2}}W_{1,\lambda}^{\frac{N}{N-2}}\psi(z)+\frac% {N}{2(N-2)}W_{2,\lambda}^{\frac{2}{N-2}}W_{1,\lambda}^{\frac{2}{N-2}}\varphi(z% )\Big{)}dz$
	$\displaystyle\quad+\int_{\mathbb{R}^{N}}\frac{1}{\|y-z\|^{N-2}}\Big{(}h_{2}(z)+% \sum_{i=1}^{2}c_{i}\sum_{j=1}^{k}w_{x_{j},\lambda}^{2^{*}-2}Z_{j,i}\Big{)}dz.$	(2.5)

	$\displaystyle N_{k}^{1,1}=(W_{1,\lambda}+\varphi)^{2^{}-1}-W_{1,\lambda}^{2^{% }-1}-(2^{}-1)W_{1,\lambda}^{2^{}-2}\varphi,$
	$\displaystyle N_{K}^{1,2}=(W_{1,\lambda}+\varphi)^{\frac{2^{}}{2}-1}(W_{2,% \lambda}+\psi)^{\frac{2^{}}{2}}-W_{1,\lambda}^{\frac{2^{}}{2}-1}W_{2,\lambda% }^{\frac{2^{}}{2}}-\Bigl{(}\frac{2^{}}{2}-1\Bigr{)}W_{1,\lambda}^{\frac{2^{% }}{2}-2}W_{2,\lambda}^{\frac{2^{}}{2}}\varphi-\frac{2^{}}{2}W_{1,\lambda}^{% \frac{2^{}}{2}-1}W_{2,\lambda}^{\frac{2^{}}{2}-1}\psi,$
	$\displaystyle N_{K}^{2,1}=(W_{2,\lambda}+\psi)^{2^{}-1}-W_{2,\lambda}^{2^{}-% 1}-(2^{}-1)W_{2,\lambda}^{2^{}-2}\psi,$
	$\displaystyle N_{K}^{2,2}=(W_{2,\lambda}+\psi)^{\frac{2^{}}{2}-1}(W_{1,% \lambda}+\varphi)^{\frac{2^{}}{2}}-W_{2,\lambda}^{\frac{2^{}}{2}-1}W_{1,% \lambda}^{\frac{2^{}}{2}}-\Bigl{(}\frac{2^{}}{2}-1\Bigr{)}W_{2,\lambda}^{% \frac{2^{}}{2}-2}W_{1,\lambda}^{\frac{2^{}}{2}}\psi-\frac{2^{}}{2}W_{2,% \lambda}^{\frac{2^{}}{2}-1}W_{1,\lambda}^{\frac{2^{}}{2}-1}\varphi.$

	$\displaystyle\|I_{1}\|\leq$	$\displaystyle C\\|\psi\\|_{*}^{\frac{N}{N-2}}\Big{(}\sum_{j=1}^{k}\frac{\lambda^% {\frac{N-2}{2}}}{(1+\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\tau}}\Big{)}^{\frac{N}{N% -2}}\Big{(}\sum_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda\|y-x_{j}\|)^{% N-2}}\Big{)}^{\frac{2}{N-2}}$
	$\displaystyle\leq$	$\displaystyle C\lambda^{\frac{N+2}{2}}\\|\psi\\|_{}^{\frac{N}{N-2}}\Big{(}\sum_% {j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N-2}{2}+\tau}}\Big{)}^{2^{}-1}$
	$\displaystyle\leq$	$\displaystyle C\lambda^{\frac{N+2}{2}}\\|\psi\\|_{*}^{\frac{N}{N-2}}\sum_{j=1}^{% k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}}\Big{(}\sum_{j=1}^{K}% \frac{1}{(1+\lambda\|y-x_{j}\|)^{\tau}}\Big{)}^{\frac{4}{N-2}}$
	$\displaystyle\leq$	$\displaystyle C\lambda^{\frac{N+2}{2}}\\|(\varphi,\psi)\\|_{*}^{\frac{N}{N-2}}% \sum_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\frac{N+2}{2}+\tau}}.$

	$\displaystyle\|I_{2}\|$	$\displaystyle\leq C\\|\psi\\|_{}^{\frac{N}{N-2}}\\|\varphi\\|_{}^{\frac{2}{N-2}}% \Big{(}\sum_{j=1}^{k}\frac{\lambda^{\frac{N-2}{2}}}{(1+\lambda\|y-x_{j}\|)^{% \frac{N-2}{2}+\tau}}\Big{)}^{\frac{N+2}{N-2}}$
		$\displaystyle\leq C\lambda^{\frac{N+2}{2}}\\|\psi\\|_{}^{\frac{N}{N-2}}\\|% \varphi\\|_{}^{\frac{2}{N-2}}\sum_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{% \frac{N+2}{2}+\tau}}\Big{(}\sum_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{\tau}}% \Big{)}^{\frac{4}{N-2}}$
		$\displaystyle\leq C\lambda^{\frac{N+2}{2}}\\|\psi\\|_{}^{\frac{N}{N-2}}\\|% \varphi\\|_{}^{\frac{2}{N-2}}\sum_{j=1}^{k}\frac{1}{(1+\lambda\|y-x_{j}\|)^{% \frac{N+2}{2}+\tau}}.$