数学物理学报  2017, Vol. 37 Issue (6): 1105-1118   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
康东升
熊萍
一类带有不同Rellich项的临界双调和方程组的非平凡解
康东升, 熊萍     
中南民族大学数学与统计学学院 武汉 430074
摘要:该文研究一类带有多重临界Sobolev指数和不同Rellich位势项的双调和方程组.利用变分法得到在一定条件下相关最佳常数的达到函数对的存在形式和基本性质,并证明双调和方程组非平凡解的存在性.
关键词双调和方程组        临界Sobolev指数    Rellich位势项    变分法    
Biharmonic Systems Involving Critical Nonlinearities and Different Rellich-Type Potentials
Kang Dongsheng, Xiong Ping     
School of Mathematics and Statistics, South-Central University for Nationalities, Wuhan 430074
Abstract: In this paper, a system of biharmonic equations is investigated, which involves multiple critical Sobolev nonlinearities and different Rellich-type terms. The minimizers of the related best Soblev constant are found under certain assumptions and the existence of solutions to the system is established by variational arguments.
Key words: Biharmonic system     Solution     Critical exponent     Rellich potential     Variational method    
1 引言

本文研究如下双调和方程组

$\begin{eqnarray}\label{q1} \hspace {-1cm}\left\{ \begin{array}{ll} \Delta^2 u -\displaystyle \mu_1\frac{ u}{|x|^4} = |u|^{2^*-2}u + \frac{\eta \alpha }{2^*} |u|^{\alpha -2} |v|^\beta u + a_1 u +a_2v ,\ \ &x\in\Omega ,\\ \Delta^2 v -\displaystyle \mu_2\frac{ v}{|x|^4} = |v|^{2^*-2}v + \frac{\eta \beta}{2^* } |u|^{\alpha }|v|^{\beta -2}v + a_2 u +a_3v,\ \ &x\in\Omega ,\\ \displaystyle u =v =\frac{\partial u}{\partial n} =\frac{\partial v}{\partial n} =0,\ \ &x\in \partial \Omega , \end{array} \right. \end{eqnarray}$ (1.1)

其中 $\Omega\subset {\mathbb R}^N (N\geq 5)$ 是具有光滑边界 $\partial \Omega$ 的有界域, $\frac{\partial }{\partial n}$ 是外法向导数, $ 0\in \Omega ,$ $ a_1,a_2,a_3 \in {\mathbb R} ,$ $\eta >0,\ 0\le \mu_1,\mu_2<\bar{\mu},\ \alpha ,\beta >1,\ \alpha +\beta =2^*,$ $ \ 2^* := \frac{2N}{N-4} $ 是临界Sobolev指数, $\bar{\mu }:=\bigr(\frac{N(N-4)}{4}\bigr)^2$ 是最佳Rellich常数.

我们用 $H:= H^2_0(\Omega ) $ 表示 $C_0^\infty(\Omega )$ 关于范数 $(\int_{\Omega } |\Delta \cdot |^2 \textrm{d}x )^{1/2}$ 的完备化空间.在空间 $H\times H$ 上, 定义双调和方程组(1.1)的能量泛函

$\begin{eqnarray}\label{1.5} \displaystyle J(u,v)& :=& \frac{1}{2}\int_\Omega \Bigr ( |\Delta u |^2 + |\Delta v |^2 - \frac{ \mu_1 u^2+ \mu_2 v^2 }{|x|^4} \Bigr )\textrm{d}x - \frac{ \eta }{2^*}\int_\Omega |u|^\alpha |v|^\beta \textrm{d}x \\ \displaystyle &&- \frac{1}{2^*}\int_\Omega \bigr (|u|^{2^*} + |v|^{2^*} \bigr ) \textrm{d}x - \frac{ 1}{2}\int_\Omega \bigr ( a_1u^2+2a_2uv +a_3v^2\bigr ) \textrm{d}x , \end{eqnarray}$ (1.2)

$J(u,v)\in C^1 (H\times H,{\mathbb R} ) $ .在空间 $H\times H$ 和它的对偶空间 $(H\times H)^{-1}$ 之间定义对偶积

$\begin{eqnarray*} \langle J'(u,v),(\varphi,\phi ) \rangle &:=&\int_\Omega \Bigr ( \Delta u \Delta \varphi + \Delta v \Delta \phi - \mu_1\frac{ u \varphi }{|x|^4} - \mu_2\frac{ v \phi }{|x|^4} \Bigr )\textrm{d}x \\ &&- \int_\Omega \bigr ( a_1 u \varphi + a_2 v \varphi + a_2 u \phi + a_3 v \phi \bigr )\textrm{d}x \\ &&- \int_\Omega \Bigr ( \frac{\eta \alpha }{2^*} |u|^{\alpha -2} |v|^\beta u \varphi + \frac{\eta \beta}{2^*} |u|^{\alpha } |v|^{\beta -2}v \phi \Bigr )\textrm{d}x \\ &&- \int_\Omega \Bigr (|u|^{2^*-2}u \varphi + |v|^{2^*-2}v \phi \Bigr )\textrm{d}x, \end{eqnarray*}$

其中 $J^{\prime}(u,v)$ 是能量泛函 $J$ $(u,v)$ 处的Fréchet导数, $u,v,\varphi,\phi \in H$ .如果函数组 $(u,v)\in H\times H$ 且满足

$\begin{eqnarray} (u,v)\neq (0,0),\ \ \ \ \ \langle J'(u,v),(\varphi,\phi ) \rangle =0 ,\ \ \forall (\varphi,\phi) \in H\times H, \end{eqnarray}$

则称 $(u,v)$ 是双调和方程组(1.1)的一组解.即方程组(1.1)的解等价于能量泛函 $J(u,v)$ 的非零临界点.

本文所研究的双调和方程组(1.1)与下列著名的Rellich不等式相关[1]

$\begin{eqnarray} \int_{{\mathbb R}^N}\frac{u^2}{|x |^{4}} \textrm{d}x \le \frac{1}{\bar{\mu }}\int_{{\mathbb R}^N}|\Delta u|^2 \textrm{d}x ,\hspace {.3cm}\forall u\in D^{2,2}({\mathbb R}^N) , \end{eqnarray}$ (1.3)

其中 $D:= D^{2,2}({\mathbb R}^N) $ 是空间 $C_0^\infty({\mathbb R}^N)$ 关于 $ (\int_{{\mathbb R}^N} |\Delta u|^2 \textrm{d}x )^{1/2}$ 的完备化空间.由Rellich不等式可知, 当 $\mu <\bar{\mu } $ 时算子 $L:=(-\Delta^2 \cdot -\mu \frac{ \cdot }{|x|^4})$ 是正算子.于是如下方程的第一特征值 $\Lambda _1(\mu)$ 可以被定义

$\begin{eqnarray*} \hspace {-1cm}\left\{ \begin{array}{ll} \Delta^2 u -\displaystyle \mu \frac{ u}{|x|^4} = \lambda u,\ \ & x\in\Omega ,\\ \displaystyle u = \frac{\partial u}{\partial n}=0,\ \ &x\in \partial \Omega . \end{array} \right. \end{eqnarray*}$

并且我们还可以定义下列最佳常数

$\begin{eqnarray} S(\mu ) := \inf\limits_{u\in D \setminus \{0 \}}\frac{ \int_{{\mathbb R}^N}\bigr ( |\Delta u|^2- \mu \frac{ u^2}{|x|^{4 }} \bigr)\textrm{d}x } {\bigr ( \int_{{\mathbb R}^N} {|u|^{2^{*} }} \textrm{d}x \bigr )^{\frac{2}{2^* }}} ,\ \ \mu \in (-\infty,\bar{\mu }) . \end{eqnarray}$

$S( 0) $ 是著名的最佳Sobolev常数[2].当 $0\le \mu <\bar{\mu } $ 时, 由文献[3]和[4]可知 $S(\mu ) $ 有达到函数

$\begin{eqnarray} \Bigr \{ V_{\mu}^\varepsilon (x) := \varepsilon ^\frac{4-N}{2}U_\mu ( {\varepsilon }^{-1}x ) ,\ \varepsilon >0\Bigr \}, \end{eqnarray}$ (1.4)

其中 $U_\mu(x) =U_\mu(|x|)$ 是径向对称函数, 关于 $S(\mu ) $ 的达到函数的详细性质将在本文的引理2.2中给出.当 $ \mu_1,\mu_2 <\bar{\mu },\ \alpha ,\beta >1$ , $ \alpha +\beta = 2^*,$ 时, 由Rellich不等式、Young不等式和Sobolev不等式, 定义空间 ${\cal D}:= D ^2\setminus \{(0,0)\}$ 上的最佳常数

$\begin{eqnarray} \displaystyle S (\mu_1,\mu_2)\label{s12} := \inf\limits_{(u,v)\in {\cal D} }\frac{ \int_{{\mathbb R}^N}\bigr ( |\Delta u|^2 + |\Delta v|^2 - \frac{\mu_1 u^2+\mu_2v^2}{|x|^{4 }} \bigr )\textrm{d}x } {\bigr ( \int_{{\mathbb R}^N} \bigr ( |u|^{2^*} + |v|^{2^*}+ \eta |u|^{\alpha } |v|^\beta \bigr ) \textrm{d}x \bigr )^{\frac{2}{2^*}}} . \end{eqnarray}$ (1.5)

考虑下列双调和方程

$\begin{eqnarray} \hspace {-1cm}\left\{ \begin{array}{ll} \Delta^2 u -\displaystyle \mu \frac{ u}{|x|^4} = |u|^{2^*-2}u +\lambda u ,\ \ &x\in\Omega ,\\ \displaystyle u =\frac{\partial u}{\partial n} =0,\ \ &x\in \partial \Omega , \end{array} \right. \end{eqnarray}$ (1.6)

其中 $\mu \in (0,\bar \mu),\ \lambda \in (0,\Lambda _1(\mu) ) $ , $\Omega\subset {\mathbb R}^N (N\geq 5)$ 是具有光滑边界 $\partial \Omega $ 的有界域, 并且 $ 0\in \Omega $ .在文献[4]中, 作者研究了最佳常数 $S(\mu ) $ 达到函数的渐近性, 并且证明了方程(1.6)解的存在性.文献[3]和[5]也分别研究了带有Rellich项的双调方程.

下面我们给出Hardy不等式[6]

$\begin{eqnarray*} \int_{{\mathbb R}^N}\frac{|u|^2}{|x |^{2}} \textrm{d}x \le \Bigr (\frac{2}{N-2}\Bigr )^2 \int_{{\mathbb R}^N}| \nabla u|^2 \textrm{d}x ,\hspace {.3cm}\forall u\in D^{1,2}({\mathbb R}^N) , \end{eqnarray*}$

其中 $N\ge 3$ , $ D^{1,2}({\mathbb R}^N) $ $C_0^\infty({\mathbb R}^N)$ 关于范数 $ (\int_{{\mathbb R}^N} | \nabla \cdot |^2 \textrm{d}x )^{1/2}$ 的完备化空间.近年来, 带有Hardy项的二阶椭圆方程已经被大量研究, 并得出许多结果.例如文献[7-14]以及其参考文献.双调方程(1.1)的结构是受到文献[7, 9]和[11]的启发.到目前为止, 带有Rellich项的临界双调和方程及方程组的研究相对较少.因此对于方程组(1.1)的深入研究具有一定理论及实际意义.

本文将需要下列假设条件.

$ ({\cal H}_1) \ \ N\ge 5,\ \eta >0,\ \ 0\le \mu_2\le \mu_1<\bar\mu ,\ (\mu_1,\mu_2)\not =(0,0), \ \alpha ,\beta>1,\ \alpha +\beta =2^*;$

$ ({\cal H}_2) \ \ a_1> 0,\ a_2\not =0, \ a_1a_3- a_2^2>0,\ \ 0< \lambda _1 \le \lambda _2<\Lambda _1(\mu_1),$ 其中 $\lambda _1$ , $\lambda _2$ 是矩阵 $ A:= \bigg( \begin{array}{ll} a_1 \ a_2 \\ a_2 \ a_3 \end{array} \bigg) $ 的特征值.

在假设条件 $({\cal H}_1)$ 下我们定义

$\begin{eqnarray} \mu^*:= \bar\mu- 8 \sqrt{\bar\mu}, \end{eqnarray}$
$\begin{eqnarray} F(s,t):= |s|^{2^*}+|t|^{2^*}+\eta |s|^\alpha |t|^\beta ,\ \ s,t \in {\mathbb R} , \end{eqnarray}$
$\begin{eqnarray} f(\tau ):= \frac{1+\tau^2}{F(1,\tau )^\frac{2}{2^*}},\ \ \tau\ge 0, \end{eqnarray}$
$\begin{eqnarray} \bar{g}(\mu_1,\mu_2):= \min\limits_{\tau \ge 0 } g(\tau),\ \ \ g (\tau ) := \frac{S(\mu_1)+S(\mu_2) \tau^2}{F( 1,\tau )^\frac{2}{2^*} },\ \tau\ge 0, \end{eqnarray}$ (1.7)
$\begin{eqnarray} \bar{h}(\mu_1,\mu_2):= \min\limits_{\tau \ge 0 } h(\tau),\ \ \ h(\tau ) := \frac{ S(\mu_1)^\frac{N}{4}+ S(\mu_2)^\frac{N}{4} \tau^2 }{ \bigr ( \int_{{\mathbb R}^N } F(V_{\mu_1}^\varepsilon ,\tau V_{\mu_2}^\varepsilon ) \bigr ) ^\frac{2}{2^*} } ,\ \tau\ge 0, \end{eqnarray}$ (1.8)
$\begin{eqnarray} f(\tau_{\min}) := \min\limits_{\tau \ge 0 } f(\tau )>0,\ \ \ h(\hat{\tau}_{\min}) := \min\limits_{\tau \ge 0 } h(\tau )>0, \end{eqnarray}$ (1.9)
$\begin{eqnarray} \bar{\eta} := \inf\limits_{\tau> 0}\Bigg(\frac{ S(\mu_1)^{-\frac{2^*}{2}} \bigr (S(\mu_1)^\frac{N}{4} + S(\mu_2)^\frac{N}{4} \tau^{2 } \bigr ) ^{ \frac{2^*}{2}} - \bigr ( S(\mu_1)^\frac{N}{4} + S(\mu_2)^\frac{N}{4} \tau^{2^*} \bigr ) }{\tau^{ \beta } \int_{{\mathbb R}^N} |V_{\mu_1} ^\varepsilon |^{\alpha } |V_{\mu_2} ^\varepsilon |^{\beta } }\Bigg), \end{eqnarray}$ (1.10)

其中 $\tau_{\min},\hat{\tau}_{\min} \ge 0$ 分别是 $f(\tau )$ $h(\tau)$ 的极小值点.直接通过计算可知 $\bar\eta>2 $ , 并且下确界 $\bar\eta$ 能在某有穷点 $\tau ^*>0$ 处达到.特别地当 $\mu_1=\mu_2 $ 时, 我们可以把 $\bar{\eta}$ 定义为下列更简单的常数

$\begin{eqnarray} \hat{ {\eta}} := \inf\limits_{\tau> 0} \Bigr ( \tau^ {- \beta } \bigr( { (1 + \tau^{2 } ) ^{ \frac{2^*}{2}} - ( 1+ \tau^{2^*} ) } \bigr)\Bigr ) . \end{eqnarray}$ (1.11)

下面我们给出本文需要的第三个假设条件.

$ ({\cal H}_3) $ $ \beta <2 $ 或者 $ \eta >\bar \eta.$

最后我们把本文的主要结论归纳为以下定理.

定理1.1 假设 $({\cal H}_1)$ $({\cal H}_3)$ 成立.则 $S (\mu_1 ,\mu_2 ) <S(\mu_1) $ 并且 $S (\mu_1,\mu_2) $ 有正的、单调递减的、径向对称的达到函数对.

定理1.2 假设 $({\cal H}_1)$ 成立, $N\le 8,\ \eta \le 2,\ \min\{\alpha ,\beta \} \ge 2 . $ $ S(\mu_1 ,\mu_2 ) = S(\mu_1)$ 并且 $S(\mu_1 ,\mu_2 )$ 有半平凡达到函数对 $ \bigr \{( V_{\mu_1} ^\varepsilon (x),0), \varepsilon >0 \bigr \} . $

定理1.3 假设 $({\cal H}_1)$ $({\cal H}_3)$ 成立, 且 $\bar{g}(\mu_1,\mu_2)=\bar{h}(\mu_1,\mu_2). $ 则当 $\hat{\tau}_{\min}>0$ 时, $S(\mu_1 ,\mu_2 ) = h(\hat{\tau}_{\min})$ , 并且 $S(\mu_1 ,\mu_2 )$ 有达到函数对 $\bigr \{( V_{\mu_1} ^\varepsilon (x),\ \hat{\tau}_{ \min} V_{\mu_2} ^\varepsilon (x) ), \varepsilon >0 \bigr \} . $

推论1.4 假设 $({\cal H}_1)$ 成立, $\mu\in [0,\bar \mu).$ 则当 $\eta > \hat{ \eta}$ 或者 $\min\{\alpha ,\beta \} <2$ 时, 我们有 $ S(\mu ,\mu )$ $ = f(\tau_{\min}) S(\mu)$ , 并且 $ S(\mu ,\mu ) $ 有达到函数对 $ \{( V_\mu ^\varepsilon (x), \tau_{ \min} V_\mu ^\varepsilon (x) ),\varepsilon >0 \}. $

定理1.5 假设 $({\cal H}_1) $ - $({\cal H}_3)$ 成立, 并且 $ N\ge 9,\ \mu_1 \le \mu^* ,$ $\bar{g}(\mu_1,\mu_2)=\bar{h}(\mu_1,\mu_2) $ .则方程组(1.1)存在一个解 $(u_0,v_0) \in (H\setminus \{0\})^2$ .

注1.6 需要特别说明的是定理1.2和定理1.5中条件 $\bar{g}(\mu_1,\mu_2)= \bar{h}(\mu_1,\mu_2)$ 是可以成立的.例如, 当 $\bar{g}(\mu_1, \mu_1)= \bar{h}(\mu_1,\mu_1)$ 时, 由隐函数理论可知, 在一定条件下等式 $\bar{g}(\mu_1,\mu_2)=\bar{h} (\mu_1,\mu_2) $ 成立.

本篇文章共有四个部分.在文章第二部分中, 我们建立一些预备引理, 为后面的研究作准备.在第三部分中, 完成定理1.1和定理1.3的证明.最后我们将在文章的第四部分给出定理1.5的证明.在下文中, $\|u\|=(\int_{\Omega } |\Delta u|^2 \textrm{d}x )^{1/2}$ 表示空间 $H$ 上的范数, $\|(u,v)\|_{H\times H}=(\|u\|^2+\|v\|^2)^{1/2}$ 表示空间 $H \times H$ 上的范数.对于任意的 $t>0$ , 当 $\varepsilon \to 0 $ 时, 我们用 $O(\varepsilon ^t)$ 表示 $|O(\varepsilon ^t)|/\varepsilon ^t \le C$ , $o(\varepsilon ^t)$ 表示 $|o(\varepsilon ^t)|/\varepsilon ^t \to 0 $ , $o(1)$ 表示无穷小量. $O_1(\varepsilon ^t)$ 表示当 $\varepsilon$ 足够小时, 存在常数 $C_1,C_2>0$ , 使得 $C_1\varepsilon ^t \le O_1(\varepsilon ^t) \le C_2\varepsilon ^t\ $ .我们通常用 $ C $ 表示正常数.为了方便起见, 我们有时省略积分式中的 $ \textrm{d}x $ .我们用 $\rightharpoonup$ 表示在对应空间上的弱收敛, 用 $\rightarrow$ 表示在对应空间上的强收敛.

2 预备引理

为了方便起见, 我们设

$\begin{eqnarray} E(u,v):= |\Delta u|^2+ |\Delta v|^2 -\frac{\mu_1u^2+\mu_2v^2}{|x|^4},\ \ \mu_1,\mu_2<\bar \mu. \end{eqnarray}$

引理2.1 假设条件 $({\cal H}_1)$ $({\cal H}_2)$ 成立.则对于任意 $c<c^* := \frac{2}{N} S (\mu_1,\mu_2) ^ {N/ 4} $ , 泛函 $J$ 在积空间 $H\times H$ 上满足 $(PS)_c$ 条件.

 假设序列 $\{(u_n,v_n)\}\subset H\times H$ 满足

$\begin{eqnarray} J(u_n,v_n)\rightarrow c<c^* ,\ \ \ \ J '(u_n,v_n) \rightarrow 0 \ \ \mbox{在对偶空间上$ (H\times H)^{-1} .$} \end{eqnarray}$

则易证 $\{(u_n,v_n)\}$ 是空间 $H\times H$ 上的有界序列.于是存在某个子列(仍记为 $\lbrace(u_{n},v_{n})\rbrace$ )和 $(u,v) \in H\times H $ 使得

$\begin{eqnarray} (u_n,v_n) \rightharpoonup (u,v) \mbox{在$ H\times H$ 中,} \end{eqnarray}$
$\begin{eqnarray} (u_n,v_n) \to (u,v)\ \mbox{ a.e. 在 $\Omega $ 中,} \end{eqnarray}$
$\begin{eqnarray} (u_n,v_n) \rightharpoonup (u,v) \mbox{在$ L^{2^*}(\Omega )\times L^{2^*}(\Omega )$中,} \end{eqnarray}$
$\begin{eqnarray}(u_n,v_n) \to (u,v) \mbox{在$ L^{2}(\Omega ,|x|^{-4} ) \times L^{ 2}(\Omega ,|x|^{-4} ) $中,} \end{eqnarray}$
$\begin{eqnarray}(u_n,v_n) \to (u,v) \ \ \mbox{在$ L^{q_1}(\Omega ) \times L^{q_2}(\Omega )$中,} \ \forall q_1,q_2 \in [1, 2^*). \end{eqnarray}$

由集中紧性原理[15-17], 存在至多可数的集合 ${\cal J}$ , 点 $x_j\in \Omega\backslash \{0\}$ , 实数 $ \sigma_0,\nu _0, \rho_0,\sigma_{x_j},\rho_{x_j},j\in {\cal J},$ 使得下列收敛性在测度意义下成立

$\begin{eqnarray}\label{2.3} |\Delta u_n |^2 + |\Delta v_n |^2 \rightharpoonup \textrm{d}\sigma \ge |\Delta {u} |^2 + |\Delta {v} |^2 + \sigma_0 \delta_{0} +\sum\limits_{j\in {\cal J}}\sigma_{x_j}\delta_{x_j}, \end{eqnarray}$ (2.1)
$\begin{eqnarray}\label{2.7} \frac{\mu_1|u_n|^2+\mu_2 |v_n|^2 }{|x|^4} \rightharpoonup \textrm{d}\nu = \frac{\mu_1u^2+\mu_2 v^2}{|x|^4} + \nu_0 \delta_{0}, \end{eqnarray}$ (2.2)
$\begin{eqnarray}\label{2.4} F( u_n,v_n )\rightharpoonup \textrm{d} \rho = F(u,v)+ \rho_0\delta_0 +\sum\limits_{j\in {\cal J}}\rho_{x_j}\delta_{x_j} , \end{eqnarray}$ (2.3)

其中 $\delta _x$ 表示在点 $x$ 处的Dirac质量, 由(1.5)式和(2.1)-(2.3)式可得

$\begin{eqnarray}\label{2.1} \sigma_0 - \nu_0 \ge S(\mu_1,\mu_2) (\rho_0) ^\frac{2}{2^* } , \end{eqnarray}$ (2.4)
$\begin{eqnarray} \sigma_{x_j} \ge S (0,0 ) (\rho_{x_j}) ^\frac{2}{2^* } ,\ \ j \in {\cal J} . \end{eqnarray}$ (2.5)

我们先考虑在原点的情况.取 $\varepsilon >0$ 足够小, $\varphi_\varepsilon \in C_0^\infty (B_\varepsilon (0))$ , 使得在球 $B_{\varepsilon /2}(0)$ $ \varphi_\varepsilon =1$ , 并且 $ 0\le \varphi _\varepsilon \le 1,|\Delta \varphi_\varepsilon |\le \frac{2}{\varepsilon } .$ 则有

$\begin{eqnarray} \lim\limits_{\varepsilon \to 0}\lim\limits_{n\to \infty } \int_\Omega \Bigr ( {|\Delta {u}_n |^2} + {|\Delta {v}_n |^2} \Bigr ) \varphi_\varepsilon =\lim\limits_{\varepsilon \to 0}\int_\Omega \varphi_\varepsilon \textrm{d}\sigma \ge \sigma_0 , \end{eqnarray}$
$\begin{eqnarray} \lim\limits_{\varepsilon \to 0}\lim\limits_{n\to \infty } \int_\Omega \frac{\mu_1|u_n|^2+\mu_2 |v_n|^2 }{|x |^{4 }}\varphi_\varepsilon =\lim\limits_{\varepsilon \to 0}\int_\Omega \varphi_\varepsilon \textrm{d}\nu = \nu_0 , \end{eqnarray}$
$\begin{eqnarray} \lim\limits_{\varepsilon \to 0}\lim\limits_{n\to \infty } \int_\Omega F( u_n,v_n )\varphi_\varepsilon =\lim\limits_{\varepsilon \to 0}\int_\Omega \varphi_\varepsilon \textrm{d} \rho = \rho_0 , \end{eqnarray}$
$\begin{eqnarray} \lim\limits_{\varepsilon \to 0}\lim\limits_{n\to \infty } \int_\Omega {u_n \Delta u_n\Delta \varphi_\varepsilon } = \lim\limits_{\varepsilon \to 0}\lim\limits_{n\to \infty } \int_\Omega {v_n \Delta v_n\Delta \varphi_\varepsilon } = 0 , \end{eqnarray}$
$\begin{eqnarray} \lim\limits_{\varepsilon \to 0}\lim\limits_{n\to \infty } \int_\Omega \bigr (|u_n|^2+|v_n|^2+ |u_nv_n| \bigr )\varphi_\varepsilon = 0 . \end{eqnarray}$

因此

$\begin{eqnarray}\label{2.2} 0=\lim\limits_{\varepsilon \to 0} \lim\limits_{n\to \infty } \langle J'(u_n,v_n) , (u_n\varphi_\varepsilon ,v_n\varphi_\varepsilon )\rangle \ge \sigma_0- \nu_0 - \rho _0. \ \ \ \end{eqnarray}$ (2.6)

根据(2.4)和(2.6)式可知 $S(\mu_1,\mu_2) \rho_0^\frac{2}{2^* }\le \rho_0 $ , 由此可以推出

$\begin{eqnarray} \rho_0=0 \ \mbox{或者} \ \rho_0\ge S(\mu_1,\mu_2)^\frac{N}{4} . \end{eqnarray}$ (2.7)

接下来我们考虑在点 $x_j$ ( $j\in {\cal J}$ )处的情况.取 $\varepsilon >0$ 足够小, $\phi_\varepsilon \in C_0^\infty (B_\varepsilon (x_j))$ , 使得在 $B_{\varepsilon /2}(x_j)$ $ \phi_\varepsilon =1$ , 并且 $ 0\le \phi _\varepsilon \le 1 $ , $ |\Delta \phi_\varepsilon |\le \frac{2}{\varepsilon } .$ 则有

$\begin{eqnarray} \lim\limits_{\varepsilon \to 0}\lim\limits_{n\to \infty } \int_\Omega \frac{\mu_1|u_n|^2+\mu_2| v_n|^2 }{|x |^{4 }}\phi_\varepsilon =\lim\limits_{\varepsilon \to 0}\int_\Omega \phi_\varepsilon \textrm{d}\nu = 0 . \end{eqnarray}$

类似于(2.6)式, 我们可以得到

$\begin{eqnarray} 0=\lim\limits_{\varepsilon \to 0} \lim\limits_{n\to \infty } \langle J'(u_n,v_n) , (u_n\phi_\varepsilon ,v_n\phi_\varepsilon )\rangle \ge \sigma_{x_j} - \rho _{x_j}. \ \ \ \end{eqnarray}$ (2.8)

由(2.5)和(2.8)式可得 $ S (0,0) (\rho_{x_j})^\frac{2}{2^* }\le \rho_{x_j} $ , 由此可以推出

$\begin{eqnarray} \rho_{x_j}=0 \ \mbox{或者}\ \rho_{x_j}\ge S (0,0)^\frac{N }{4 }> S(\mu_1,\mu_2)^\frac{N }{4} . \end{eqnarray}$ (2.9)

于是集合 $ {\cal J}$ 必为有限集.

另一方面, 我们有

$\begin{eqnarray} c& =& J(u_n,v_n)-\frac{1}{2} \langle J'(u_n,v_n) ,(u_n,v_n) \rangle+o(1) \\ &=& \frac{2}{N} \int_\Omega F( u_n,v_n ) +o(1) \\ & =& \frac{2}{N}\bigg(\int_\Omega F(u,v)+\rho_0 + \sum\limits_{j\in {\cal J} }\rho_{x_j} \bigg) . \end{eqnarray}$ (2.10)

由(2.7)-(2.10)式和假设条件 $c<c ^* $ 可知 $\rho_0=\rho_{x_j} = 0 ,\ \forall j\in {\cal J}.$ 则在空间 $H\times H$ 上子列 $(u_n,v_n)$ 强收敛到 $(u,v)$ .引理证毕.

下面我们定义

$\begin{eqnarray} \lambda (\mu ) := \psi(\mu /\bar\mu),\ \ \nu (\mu ):=\frac{ 1 }{2} (N-4)\lambda(\mu ) ,\ \ \ \gamma (\mu):=\frac{ 1 }{2} (N-4) \bigr (2- \lambda(\mu ) \bigr ), \end{eqnarray}$

其中 $\psi: [0, 1] \mapsto [0, 1]$ 定义如下

$\begin{eqnarray} \psi (t) := 1- \frac{\sqrt{(N-2)^2+4- \sqrt{16(N-2)^2 +t(N-4)^2N^2}}}{N-4} . \end{eqnarray}$

引理2.2[4] 对于所有的 $0< \mu <\bar{\mu } ,$ $S(\mu ) $ 有达到函数 $ \bigr \{ V_{\mu}^\varepsilon (x) := \varepsilon ^\frac{4-N}{2}U_\mu ( x/\varepsilon ) ,\varepsilon >0 \bigr \},$ 其中 $U_\mu(x) =U_\mu(|x|)$ 是正的、单调递减的和径向对称的, 并且满足

$\begin{eqnarray} U_\mu (\rho) = O_1 (\rho^{-\nu (\mu )}),\ \ \ \rho\to 0^+; \end{eqnarray}$
$\begin{eqnarray} U_\mu (\rho) = O_1 (\rho^{-\gamma (\mu)}),\ \ \ U'_\mu (\rho) = O_1 (\rho^{-\gamma (\mu)-1}),\ \ \ \rho\to +\infty. \end{eqnarray}$

函数 $V_{\mu}^\varepsilon (x)$ 是下列方程的解

$\begin{eqnarray} \Delta^2 u- \mu \frac{ u}{|x|^4} = { u ^ {2^* -1}} ,\ \ u>0,\ \ x\in {\mathbb R} ^ N \setminus \{0\}, \end{eqnarray}$

且满足下列等式

$\begin{eqnarray} \int_{{\mathbb R}^N}\bigg( |\Delta V_{\mu}^\varepsilon (x) |^ 2-\mu \frac{|V_{ \mu}^\varepsilon (x)|^ 2}{|x|^ 4} \bigg) = \int_{{\mathbb R}^N} |V_{ \mu}^\varepsilon (x)|^{2^* }= S(\mu)^\frac{N }{4 }. \end{eqnarray}$

进而有

$\begin{eqnarray} \Bigr ( 1- \frac{\mu}{\bar{\mu}} \Bigr ) S(0) < S(\mu)<S(0). \end{eqnarray}$

$\rho>0$ 足够小使得 $B_\rho (0) \in \Omega $ .选取截断函数 $\varphi(x)= \varphi(|x|) \in C_0^\infty (B_\rho ( 0))$ , 使得在球 $B_{\rho} (0)$ $\varphi (x)$ 满足 $ 0 \le \varphi (x) \le 1 $ , 在球 $B_{\rho/2} (0)$ 中满足 $ \varphi(x) = 1 $ .设 $u_{\mu }^\varepsilon (x)=\varphi (x)V_{\mu}^\varepsilon (x)$ .我们有下面的渐近估计结果.

引理2.3[4] 假设 $N\ge 5,\ \mu\in [0,\bar\mu).$ $\varepsilon \rightarrow 0$ 时, 下列估计成立

$\begin{eqnarray} \int_{\Omega }\Bigr (|\Delta u_{\mu }^\varepsilon |^2 -\mu\frac{|u_{\mu }^\varepsilon |^2}{|x|^4} \Bigr )= S(\mu) ^{\frac{N}{4}}+O\bigr (\varepsilon ^{2(\gamma (\mu)- \frac{N-4}{2} )}\bigr ); \end{eqnarray}$
$\begin{eqnarray} \int_{\Omega } |u_{\mu }^\varepsilon |^{2^*} = S(\mu) ^{\frac{N}{4}}+ O\bigr (\varepsilon ^{2^*(\gamma (\mu)- \frac{N-4}{2} )}\bigr ). \end{eqnarray}$

进一步地, 当 $N\ge 9$ 时有

$\begin{eqnarray} \int_{{\Omega }} {|u_{\mu }^\varepsilon |^{2 }} = \left\{ \begin{array}{ll} O_1( \varepsilon ^{4}) , &0\le \mu <\mu^*,\\ O_1\bigr ( \varepsilon ^{4}|\ln\varepsilon |\bigr ) , &\mu =\mu^*. \end{array} \right.\end{eqnarray}$
3 最佳常数 $S (\mu_1,\mu_2)$ 的达到函数对

引理3.1 假设 $({\cal H}_1)$ 成立, 并且 $S (\mu_1,\mu_2) <S(\mu_1)$ .则 $S (\mu_1,\mu_2) $ 有径向对称且单调递减的达到函数对.

 由Ekeland变分原理, 我们可以选取 $S (\mu_1,\mu_2)$ 的极小序列 $\{ (\tilde{u}_n,\tilde{v}_n) \} \subset {\cal D}$ , 满足

$\begin{eqnarray} \left. \begin{array}{l} \displaystyle \int_{{\mathbb R}^N} E( \tilde{u}_n,\tilde{v}_n) = \int_{{\mathbb R}^N} F( \tilde{u}_n,\tilde{v}_n)+o(1) = S (\mu_1, \mu_2)^\frac{N}{4}+o(1), \\ \Delta^2 \tilde{u}_n -\displaystyle \mu_1\frac{ \tilde{u}_n }{|x|^4} = |\tilde{u}_n|^{2^*-2}\tilde{u}_n + \frac{\eta \alpha }{2^*} |\tilde{u}_n|^{\alpha -2} |\tilde{v}_n|^\beta \tilde{u}_n +f_n ,\\ \Delta^2 \tilde{v} _n-\displaystyle \mu_2\frac{ \tilde{v}_n}{|x|^4} = |\tilde{v}_n|^{2^*-2}\tilde{v}_n + \frac{\eta \beta}{2^* } |\tilde{u}_n|^{\alpha }|\tilde{v}_n|^{\beta -2}\tilde{v}_n+g_n, \end{array} \right. \end{eqnarray}$ (3.1)

其中 $f_n \to 0$ , $g_n \to 0$ ( $n\to \infty$ )在空间 $D $ 的对偶空间 $D^{-1}$ 上.

定义

$\begin{eqnarray} \bigr ( {u}_n(x),{v}_n (x)\bigr ):= R_n^\frac{N-4}{2}\bigr (\tilde{u}_n(R_nx),\tilde{v}_n (R_nx)\bigr )\ge 0, \ \ \ R_n>0. \end{eqnarray}$

$\int_{{\mathbb R}^N} E( \cdot,\cdot) $ $\int_{{\mathbb R}^N} F( \cdot,\cdot)$ 的伸缩不变性及(3.1)式可得

$\begin{eqnarray} \int_{{\mathbb R}^N} E( {u}_n,{v}_n) = \int_{{\mathbb R}^N} F( {u}_n,{v}_n)+o(1) = S (\mu_1, \mu_2)^\frac{N}{4}+o(1) . \end{eqnarray}$ (3.2)

通过选取合适的 $R_n$ , 根据(3.1)和(3.2)式, 我们可以假设

$\begin{eqnarray} \int_{B_1(0 )}\hspace {-.1cm} F( {u}_n,{v}_n)= \int_{ B_{R_n}(0 )} \hspace {-.1cm} F( \tilde{u}_n(x),\tilde{v}_n(x)) = \frac{1}{2} S (\mu_1, \mu_2)^\frac{N}{4} +o(1) . \end{eqnarray}$ (3.3)

再由(1.3)和(3.2)式可知 $\{ (u_n,v_n) \} $ 在空间 $ D\times D $ 上有界.类似于引理2.1, 对于某个 $(u ,v)\in D\times D $ , 存在至多可数集 $\tilde{{\cal J}},x_j\in {\mathbb R}^N\backslash \{0\}$ , 实数 $ {\tilde\rho}_{x_j},{\tilde\nu}_{x_j},$ $j\in \tilde{{\cal J}}$ ${\tilde\rho}_{0},{\tilde\nu} _{0}, {\tilde\gamma}_{0} $ , 使得下列收敛性在测度意义下成立

$\begin{eqnarray}\label{Equ11} \left. \begin{array}{l} \displaystyle | \Delta u_n|^2+|\Delta v_n|^2 \rightharpoonup d{\tilde\rho} \ge |\Delta u |^2 +|\Delta v|^2 + {\tilde\rho}_{0}\delta_{0} +\sum\limits_{j\in \tilde{{\cal J}}}{\tilde\rho}_{x_j}\delta_{x_j}, \\ \displaystyle F(u_n,v_n) \rightharpoonup d{\tilde\nu} = F(u,v) + {\tilde\nu} _0\delta_0 +\sum\limits_{j\in \tilde{{\cal J}}}{\tilde\nu} _{x_j}\delta_{x_j}, \\ \displaystyle \frac{ \mu_1 u_n^2+\mu_2v_n^2 }{|x |^4} \rightharpoonup d{\tilde\gamma} = \frac{\mu_1 u ^2+\mu_2v ^2 }{|x |^4} + {\tilde\gamma}_{0}\delta_{0}, \end{array} \right. \end{eqnarray}$ (3.4)

其中 $\delta _x$ 表示在点 $x$ 处的Dirac质量.为了讨论无穷点处的情况, 我们定义

$\begin{eqnarray} \left. \begin{array}{ll} & \displaystyle {\tilde\rho}_\infty =\lim\limits_{R\to\infty}\limsup\limits_{n\to \infty }\int_{|x|>R} (| \Delta u_n|^2+|\Delta v_n|^2) , \\ & \displaystyle {\tilde\nu}_\infty =\lim\limits_{R\to\infty}\limsup\limits_{n\to \infty }\int_{|x|>R} F(u_n,v_n), \\ & \displaystyle {\tilde\gamma }_\infty =\lim\limits_{R\to\infty}\limsup\limits_{n\to \infty }\int_{|x|>R} \frac{\mu_1 u_n^2+\mu_2v_n^2 }{|x |^4}. \end{array} \right. \end{eqnarray}$ (3.5)

则有

$\begin{eqnarray} S (\mu_1,\mu_2)^\frac{N}{4} = \limsup\limits_{n\to \infty} \int_{{\mathbb R}^N}\hspace {-.1cm} F( {u}_n,{v}_n) = \int_{{\mathbb R}^N} \hspace {-.1cm}F(u,v) + {\tilde\nu} _0 + {\tilde\nu} _\infty +\sum\limits_{j\in \tilde{{\cal J}}}{\tilde\nu} _{x_j} . \end{eqnarray}$ (3.6)

类似于引理2.1, 由(3.1)式可得

$\begin{eqnarray} \left. \begin{array}{ll} & S(0,0 ) ({\tilde\nu} _{x_j} )^\frac{2}{2^*}\le {\tilde\rho}_{x_j} \le {\tilde\nu}_{x_j} ,\ j\in \tilde{{\cal J}},\\ & \displaystyle S(\mu_1,\mu_2) ({\tilde\nu} _0 )^\frac{2}{2^*}\le {\tilde\rho}_0 - {\tilde\gamma } _{0} \le {\tilde\nu} _0 , \\ & \displaystyle S(\mu_1,\mu_2) ({\tilde\nu} _\infty )^\frac{2}{2^*}\le {\tilde\rho}_\infty - {\tilde\gamma } _\infty \le {\tilde\nu} _\infty , \end{array} \right. \end{eqnarray}$ (3.7)

$ {\tilde\nu} _{x_j},j\in \tilde{{\cal J}},\ {\tilde\nu} _{0} ,\ {\tilde\nu} _\infty,$ 等于 $0$ 或者不小于 $ S(\mu_1,\mu_2) ^{N/4} .$ $ \tilde{{\cal J}}$ 是有限集.

由(3.1)-(3.7)式可以得出

$\begin{eqnarray} \displaystyle S(\mu_1,\mu_2) ^\frac{N}{4} & =& \limsup\limits_{n\to \infty} \int_{{\mathbb R}^N} E( {u}_n,{v}_n) \\ & \displaystyle \ge& S(\mu_1,\mu_2) \limsup\limits_{n\to \infty} \Bigr ( \int_{{\mathbb R}^N}\hspace {-.1cm} F( {u}_n,{v}_n) \Bigr )^\frac{2}{2^*} \\ & \displaystyle =& S(\mu_1,\mu_2) \bigg( \int_{{\mathbb R}^N} F( {u} ,{v} ) + {\tilde\nu} _0 + {\tilde\nu} _\infty + \sum\limits_{j\in \tilde{{\cal J}}} {\tilde\nu} _{x_j} \bigg)^\frac{2}{2^*} \\ & \displaystyle =& S(\mu_1,\mu_2) ^\frac{N}{4}. \end{eqnarray}$ (3.8)

由(3.6)-(3.8)式可得 $\int_{{\mathbb R}^N}\hspace {-.1cm}F(u ,v ),\ {\tilde\nu} _{0} ,\ {\tilde\nu} _\infty,\ {\tilde\nu} _{x_j},j\in \tilde{{\cal J}},$ 等于 $0$ 或者等于 $ S(\mu_1,\mu_2) ^{N/4} .$ 再由(3.7)式可以得出 ${\tilde\nu}_{x_j}=0 $ 或者 $ {\tilde\nu}_{x_j} \ge S(0,0) ^{N/4} > S(\mu_1,\mu_2) ^{N/4},$ 即我们可以推出 ${\tilde\nu}_{x_j}=0 ,\ j\in \tilde{{\cal J}}$ .此外, 由(3.2)和(3.3)式可知 $ \min \{{\tilde\nu}_0,{\tilde\nu} _\infty \}\le \frac{1}{2} S(\mu_1,\mu_2) ^{N/4} ,$ ${\tilde\nu}_0= {\tilde\nu} _\infty=0.$ 因此

$\begin{eqnarray} \int_{{\mathbb R}^N}F(u ,v ) = S(\mu_1,\mu_2) ^\frac{N}{4}<S(\mu_1)\le S(\mu_2). \end{eqnarray}$

由(1.5), (3.2)和(3.4)式可得

$\begin{eqnarray} S(\mu_1,\mu_2) ^\frac{N}{4}\ge \int_{{\mathbb R}^N}E(u ,v ) \ge S(\mu_1,\mu_2)\Bigr (\int_{{\mathbb R}^N}F(u ,v )\Bigr )^\frac{2}{2^*}=S(\mu_1,\mu_2) ^\frac{N}{4}, \end{eqnarray}$

则有

$\begin{eqnarray} \int_{{\mathbb R}^N}E(u ,v ) = \int_{{\mathbb R}^N}F(u ,v ) = S(\mu_1,\mu_2) ^\frac{N}{4}<S(\mu_1)\le S(\mu_2). \end{eqnarray}$

所以 $(u_n,v_n)$ 强收敛于 $(u,v)$ , 并且 $(u,v)$ $S(\mu_1,\mu_2)$ 的达到函数对, 其中 $ u\not =0,v \not =0.$ 同时 $(u,v)$ 也是下列方程组的解

$\begin{eqnarray}\left\{ \begin{array}{ll} \Delta^2 u -\displaystyle \mu_1\frac{ u}{|x|^4} = |u|^{2^*-2}u + \frac{\eta \alpha }{2^*} |u|^{\alpha -2} |v|^\beta u ,\ \ & x\in{\mathbb R}^N ,\\ \Delta^2 v -\displaystyle \mu_2\frac{ v}{|x|^4} = |v|^{2^*-2}v + \frac{\eta \beta}{2^* } |u|^{\alpha }|v|^{\beta -2}v,\ \ \ & x\in{\mathbb R}^N . \end{array} \right. \end{eqnarray}$

最后我们来证明 $(|u|,|v|)=(u^*,v^*)$ , 其中 $(u^*,v^*)$ $(u,v)$ 的Schwartz对称化.

我们用反证法.假设 $(|u|,|v|)\not =(u^*, v^*) $ , 则有以下三种情况: (ⅰ) $|u| \not = u^*, \ |v| = v^*$ , (ⅱ) $|u| = u^*, \ |v| \not = v^*,$ (ⅲ) $|u| \not = u^*, \ |v| \not = v^*.$ 接下来将只对(ⅰ)进行证明. (ⅱ), (ⅲ)的证明类似, 我们省略掉.

因为 $ (u^2)^* =(u^*)^2 ,$ $(v^2)^* = (v^*)^2 ,$ 则有

$\begin{eqnarray} (u^2,v^2) \not =((u^*)^2,(v^*)^2)= ((u^2)^*,(v^2)^*). \end{eqnarray}$

$\tilde{u},\tilde{v} \in L^{2^*} ({\mathbb R}^N)$ 分别是下列方程的解

$\begin{eqnarray} - \Delta \tilde{u} = (- \Delta u)^*,\ \ \ - \Delta \tilde{v} = (- \Delta v)^*. \end{eqnarray}$ (3.9)

由于 $\Delta u,\Delta v\in L^2({\mathbb R}^N),$ 根据文献[4]可知 $ \tilde{u},\tilde{v} \in D^{2,2}({\mathbb R}^N) $ 并且 $\tilde{u}\ge u^*, \tilde{v}\ge v^*$ .再由(3.9)式可得(参见文献[18]和[19])

$\begin{eqnarray} &&\displaystyle \int_{{\mathbb R}^N} |\Delta \tilde{u}|^2 = \int_{{\mathbb R}^N} |(-\Delta u)^*|^2= \int_{{\mathbb R}^N} |\Delta u|^2 , \\ &&\displaystyle\int_{{\mathbb R}^N} |\Delta \tilde{v}|^2 = \int_{{\mathbb R}^N} |(-\Delta v)^*|^2= \int_{{\mathbb R}^N} |\Delta v|^2 , \\ &&\displaystyle \int_{{\mathbb R}^N} \frac{\tilde{u} ^2}{|x|^4} \ge \int_{{\mathbb R}^N} \frac{(u^*)^{2 }}{|x|^4} \ge \int_{{\mathbb R}^N} \frac{u^{2 }}{|x|^4},\\ &&\displaystyle \int_{{\mathbb R}^N} \frac{\tilde{v} ^2}{|x|^4} \ge \int_{{\mathbb R}^N} \frac{(v^*)^{2 }}{|x|^4} \ge \int_{{\mathbb R}^N} \frac{v^{2 }}{|x|^4}, \\ \displaystyle &&\int_{{\mathbb R}^N} \tilde{ u} ^\alpha \tilde{v}^\beta \ge \int_{{\mathbb R}^N} (u^*)^{\alpha }(v^*)^{\beta } \ge \int_{{\mathbb R}^N} |u|^\alpha |v|^{\beta } , \\ \displaystyle &&\int_{{\mathbb R}^N} \tilde{ u} ^{2^*}\ge \int_{{\mathbb R}^N} (u^*)^{2^*}= \int_{{\mathbb R}^N} |u |^{2^*}, \\ && \displaystyle \int_{{\mathbb R}^N} \tilde{ v} ^{2^*}\ge \int_{{\mathbb R}^N} (v^*)^{2^*}= \int_{{\mathbb R}^N} |v |^{2^*}. \end{eqnarray}$ (3.10)

另外, 如果 $f,g$ 是可测函数, 并且 $f>0$ 是径向递减, $g\ge 0$ 满足 $g\not =g^*$ , 由文献[18]可得

$\begin{eqnarray} \int_{{\mathbb R}^N} f g^* > \int_{{\mathbb R}^N} f g . \end{eqnarray}$

$f= |x|^{-4} $ , $g=u^2 $ , 则有

$\begin{eqnarray} \int_{{\mathbb R}^N} \frac{(u^*) ^2}{|x|^4} = \int_{{\mathbb R}^N} \frac{(u^2)^{* }}{|x|^4} > \int_{{\mathbb R}^N} \frac{u ^2}{|x|^4}. \end{eqnarray}$ (3.11)

再由(3.10)和(3.11)式可得

$\begin{eqnarray} S(\mu_1,\mu_2) = \frac{\int_{{\mathbb R}^N} E(u,v)}{ (\int_{{\mathbb R}^N} F(u,v))^\frac{2}{2^*} } >\frac{\int_{{\mathbb R}^N} E(\tilde{u},\tilde{v})}{ (\int_{{\mathbb R}^N} F(\tilde{u},\tilde{v}))^\frac{2}{2^*} } \ge S(\mu_1, \mu_2) , \end{eqnarray}$

由此得出矛盾.所以就有 $(|u|,|v|)=(u^*,v^*)$ .

因为 $(|u|,|v|)=(u^*,v^*)$ 是径向对称且不增的, 所以有

$\begin{eqnarray}(u,v)=(u^*,v^*)\ \mbox{或者}\ (u,v)=-(u^*,v^*).\end{eqnarray}$

我们选取 $(u,v)=(u^*,v^*)$ 作为 $S(\mu_1,\mu_2)$ 的非负达到函数对.注意到 $\mu_1,\mu_2, \eta $ 均为正的, 则 $(u,v)=(u^*,v^*)$ 且满足下列不等式

$\begin{eqnarray*} \hspace {-1cm}\left\{ \begin{array}{ll} \Delta^2 u \ge u^{2^*-1} ,\ \ & u\ge 0 ,\ x\in {\mathbb R}^N ,\\ \Delta^2 v \ge v^{2^*-1} ,\ \ &v\ge 0 ,\ x\in {\mathbb R}^N . \end{array} \right. \end{eqnarray*}$

由文献[4, 定理2]可知 $(u,v)=(u^*,v^*)$ 是严格正的且单调递减的.引理证毕.

定理1.1的证明 考虑定义在 $[0,\infty)$ 上的下列函数

$\begin{eqnarray*} {k} (\tau)&:=& 2^* S(\mu_1)^\frac{N}{4}S(\mu_2)^\frac{N}{4} \tau^{2-\beta } + \eta \alpha S(\mu_2)^\frac{N}{4}\int_{{\mathbb R}^N} |V_{\mu_1} ^\varepsilon |^{\alpha } |V_{\mu_2} ^\varepsilon |^{\beta } \tau^2 \\ && - 2^* S(\mu_1)^\frac{N}{4}S(\mu_2)^\frac{N}{4} \tau^\alpha -\eta \beta S(\mu_1)^\frac{N}{4}\int_{{\mathbb R}^N} |V_{\mu_1} ^\varepsilon |^{\alpha } |V_{\mu_2} ^\varepsilon |^{\beta } . \end{eqnarray*}$

由(1.8)式可得

$\begin{eqnarray} h '(\tau)= \frac{2 \tau^{\beta -1} k (\tau ) }{ 2^*\bigr ( \int_{{\mathbb R}^N } F(V_{\mu_1}^\varepsilon ,\tau V_{\mu_2}^\varepsilon ) \bigr ) ^{\frac{ 2}{2^*}+1}} , \ \ \ \tau\ge 0. \end{eqnarray}$

对于任意的 $\beta <2$ , 当 $\tau\to 0^+$ 时, 则有 $h'(\tau)<0$ .并由假设条件 $ ({\cal H}_1) $ , 我们可以知道

$\begin{eqnarray} \bar{h}(\mu_1,\mu_2) = \min\limits_{\tau \ge 0 } h(\tau)<h(0)=S(\mu_1)\le S(\mu_2)=h(\infty). \end{eqnarray}$

则可以推出 $\bar{h}(\mu_1,\mu_2)$ 一定能在某有穷点 $\hat{\tau}_{\min} >0 $ 处达到.将 $(V_{\mu_1}^\varepsilon ,\hat{\tau}_{\min} V_{\mu_2}^\varepsilon ) $ 代入(1.5)式可得

$\begin{eqnarray} S (\mu_1,\mu_2) \le h( \hat{\tau}_{\min} )= \bar{h}(\mu_1,\mu_2) < \min \{ S(\mu_1) ,S(\mu_2)\}. \end{eqnarray}$ (3.12)

如果 $\eta> \bar{\eta} $ , 由(1.10)式可知存在 $\tau_1\in (0,\infty)$ 使得

$\begin{eqnarray} \min\limits_{\tau\ge 0} h(\tau) \le h(\tau_1) <S(\mu_1) = \min\bigr \{ S(\mu_1), S(\mu_2)\bigr \} =\min \bigr \{ h(0),h(\infty) \bigr \} , \end{eqnarray}$

即(3.12)式仍然成立.

最后, 由(3.12)式和引理3.1可知定理1.1成立.

定理1.2的证明 证明与文献[20, 定理1.4]证明类似, 在此省略.

定理1.3的证明 因为 $S(\mu_1,\mu_2) <S(\mu_1)\le S(\mu_2)$ , 所以 $S(\mu_1,\mu_2)$ 没有形如 $(u,0),(0,v)$ 的半平凡达到函数对, 其中 $u,v\in D\setminus \{0\}$ .

对于任意的 $ (u ,v)\in (D\setminus \{0\})^2 $ , 定义 $z=sv$ , 其中

$\begin{eqnarray} s= \bigg(\Bigr ( \int_{{\mathbb R}^N} |v|^{2^*}\Bigr )^{-1} \int_{{\mathbb R}^N} |u|^{2^* } \bigg)^\frac{1}{2^*}, \end{eqnarray}$

则有

$\begin{eqnarray} \int_{{\mathbb R}^N} |z|^{2^* } = \int_{{\mathbb R}^N} |u|^{2^* }. \end{eqnarray}$ (3.13)

由Young不等式可得

$\begin{eqnarray} \int_{{\mathbb R}^N} |u|^\alpha |z|^\beta \le \frac{\alpha }{2^*} \int_{{\mathbb R}^N} |u|^{2^* } + \frac{\beta }{2^*} \int_{{\mathbb R}^N} |z|^{2^* } , \end{eqnarray}$

因此

$\begin{eqnarray} \int_{{\mathbb R}^N} |u|^\alpha |z|^\beta \le \int_{{\mathbb R}^N} |u|^{2^*} = \int_{{\mathbb R}^N} |z|^{2^*} . \end{eqnarray}$ (3.14)

对于任意的 $ (u ,v)\in (D\setminus \{0\})^2 $ , 由(1.7), (3.13)和(3.14)式可得

$\begin{eqnarray*} && \frac{ \displaystyle\int_{{\mathbb R}^N}\Bigr ( |\Delta u|^2 + |\Delta v|^2 - \mu_1\frac{ u^2}{|x|^{4}}- \mu_2\frac{ v^2}{|x |^{4 }} \Bigr) } {\Bigr( \displaystyle \int_{{\mathbb R}^N} |u|^{2^* } + \eta |u|^{\alpha } |v|^\beta + |v|^{2^* } \Bigr)^{\frac{2}{2^* }}} \\ & \ge& \frac{ \displaystyle\int_{{\mathbb R}^N}\Bigr ( |\Delta u|^2 + |\Delta v|^2 - \mu_1\frac{u^2}{|x |^{4 }}- \mu_2\frac{ v^2}{|x|^{4 }} \Bigr) } { \Bigr( (1+\eta s^{-\beta }+s^{- 2^* })\displaystyle\int_{{\mathbb R}^N} |u|^{2^*} \Bigr)^{\frac{2}{2^* }}} \\ &\ge& \frac{ \displaystyle\int_{{\mathbb R}^N}\Bigr ( |\Delta u|^2 - \mu_1\frac{ u^2}{|x|^{4}} \Bigr) } {\Bigr((1+\eta s^{-\beta }+s^{- 2^* }) \displaystyle\int_{{\mathbb R}^N} |u|^{2^* } \Bigr)^{\frac{2}{2^* }}} + \frac{ s^{ -2}\displaystyle\int_{{\mathbb R}^N}\Bigr ( |\Delta z|^2 - \mu_2\frac{ z^2}{|x|^{2 }} \Bigr) } {\Bigr( (1+\eta s^{-\beta }+s^{- 2^* })\displaystyle\int_{{\mathbb R}^N} |z|^{2^*} \Bigr)^{\frac{2}{2^* }}} \\ & \ge& g(s^{-1}) \ge \bar{g}(\mu_1,\mu_2) . \end{eqnarray*}$

由(1.5)式可知

$\begin{eqnarray} \ S(\mu_1,\mu_2)\ge \bar g(\mu_1,\mu_2). \end{eqnarray}$ (3.15)

最后, 根据(3.12)和(3.15)式我们可以得出

$\begin{eqnarray} \bar g(\mu_1,\mu_2) \le S(\mu_1,\mu_2) \le \bar h(\mu_1,\mu_2). \end{eqnarray}$ (3.16)

再由(3.16)式和假设条件 $\bar g(\mu_1,\mu_2) = \bar h(\mu_1,\mu_2)$ 可知定理1.3成立.

推论1.4的证明 因为 $\mu_1=\mu_2=\mu\in [0,\bar\mu)$ , 则有

$\begin{eqnarray} g(\tau)= h(\tau )= f(\tau)S(\mu ) ,\ \tau\ge 0. \end{eqnarray}$

即由定理1.3可知推论1.4成立.

4 双调和方程组(1.1)非平凡解的存在性

引理4.1 在定理1.5的条件下, 当 $ \varepsilon\rightarrow0^+$ 时, 下列估计成立

$ {\rm (ⅰ)} \ \int _{\Omega }E( u_{\mu_1 }^{\varepsilon },\hat{t}_{\min} u_{\mu_2 }^{\varepsilon } ) = S(\mu_1,\mu_2)^{\frac{N }{4 }} +O\bigr (\varepsilon ^{2(\gamma (\mu_1)- \frac{N-4}{2} )}\bigr ) ;$

$ {\rm (ⅱ)} \ \int _{\Omega }F( u_{\mu_1 }^{\varepsilon },\hat{t}_{\min} u_{\mu_2 }^{\varepsilon } ) = S(\mu_1,\mu_2)^{\frac{N }{4 }} +O\bigr (\varepsilon ^{2^*(\gamma (\mu_1)- \frac{N-4}{2} )}\bigr ) . $

 根据定理1.2可知 $S(\mu_1,\mu_2)$ 有达到函数对 $(V_{\mu_1 }^{\varepsilon },\hat{t}_{\min} V_{\mu_2 }^{\varepsilon } )$ .所以有

$\begin{eqnarray} \int_{{\mathbb R}^N}E( V_{\mu_1 }^{\varepsilon },\hat{t}_{\min} V_{\mu_2 }^{\varepsilon } ) =\int _{{\mathbb R}^N}F( V_{\mu_1 }^{\varepsilon },\hat{t}_{\min} V_{\mu_2 }^{\varepsilon } ) = S(\mu_1,\mu_2)^{\frac{N }{4 }}. \end{eqnarray}$ (4.1)

再由引理2.2和(4.1)式可得

$\begin{eqnarray} S(\mu_1)^\frac{N}{4} + (\hat{t}_{\min})^2 S(\mu_2)^\frac{N}{4}= S(\mu_1,\mu_2)^{\frac{N }{4 }}, \end{eqnarray}$ (4.2)
$\begin{eqnarray} S(\mu_1)^\frac{N}{4} + (\hat{t}_{\min})^{2^*} S(\mu_2)^\frac{N}{4}+ (\hat{t}_{\min})^\beta \int _{{\mathbb R}^N}|V_{\mu_1}^\varepsilon | ^\alpha |V_{\mu_2}^\varepsilon |^\beta = S(\mu_1,\mu_2)^{\frac{N }{4 }}. \end{eqnarray}$ (4.3)

(ⅰ) 因为 $\gamma (\mu)$ $[0,\bar\mu) $ 上单调递减, 我们根据(4.1)式, (4.2)式及引理2.3直接可知结论成立.

(ⅱ)由(4.1)式, (4.3)式和引理2.3可得

$\begin{eqnarray*} \int _{\Omega }F(u_{\mu_1 }^{\varepsilon },\hat{t}_{\min} u_{\mu_2 }^{\varepsilon } ) &=& \int _{{\mathbb R}^N}F( V_{\mu_1 }^{\varepsilon },\hat{t}_{\min} V_{\mu_2 }^{\varepsilon } )+ \ O(\varepsilon ^{2^{*} (\gamma (\mu_1)-\frac{N-4}{2})}) \\ &&+ \eta \hat{t}_{\min}^\beta \int_{{\mathbb R}^N}\Bigr ( |u_{\mu_1}^\varepsilon |^\alpha |u_{\mu_2}^\varepsilon |^\beta - |V_{\mu_1}^\varepsilon | ^\alpha |V_{\mu_2}^\varepsilon |^\beta \Bigr ) \\ & =&S(\mu_1,\mu_2)^\frac{N}{4} + O(\varepsilon ^{2^{*} (\gamma (\mu_1)-\frac{N-4}{2})}) \\ && - C \int _{ \frac{\rho}{2}<|x|<\rho }\bigr ( 1- \varphi ^{p^{*} } \bigr ) |V_{\mu_1}^\varepsilon | ^\alpha |V_{\mu_2}^\varepsilon |^\beta -C\int _{{\mathbb R}^{N}\setminus{B_{\rho}(0)}}|V_{\mu_1}^\varepsilon | ^\alpha |V_{\mu_2}^\varepsilon |^\beta \\ & =&S(\mu_1,\mu_2)^\frac{N}{4} + O(\varepsilon ^{2^{*} (\gamma (\mu_1)-\frac{N-4}{2})}) \\ && + O(\varepsilon ^{ \alpha \gamma (\mu_1)+ \beta \gamma (\mu_2) -N}) \int _{ \frac{\rho}{2}<|x|<\rho }|x|^{- (\alpha \gamma (\mu_1)+ \beta \gamma (\mu_2))} \\ && + O(\varepsilon ^{ \alpha \gamma (\mu_1)+ \beta \gamma (\mu_2) - N )}) \int _{{\mathbb R}^{N}\setminus {B_{\rho}(0)}}|x|^{- (\alpha \gamma (\mu_1)+ \beta \gamma (\mu_2))} \\ & =&S(\mu_1,\mu_2)^\frac{N}{4} + O(\varepsilon ^{2^{*} (\gamma (\mu_1)- \frac{N-4}{2})}) \\ && + O(\varepsilon ^{ \alpha \gamma (\mu_1)+ \beta \gamma (\mu_2) - N )}) \int _{ \frac{\rho}{2} }^\rho r^{N-1 - (\alpha \gamma (\mu_1)+ \beta \gamma (\mu_2))} {\rm d}r\\ && + O(\varepsilon ^{ \alpha \gamma (\mu_1)+ \beta \gamma (\mu_2) - N )}) \int _{\rho }^{+\infty } r ^{N-1 - (\alpha \gamma (\mu_1)+ \beta \gamma (\mu_2))} {\rm d}r \\ & =&S(\mu_1,\mu_2)^\frac{N}{4} + O(\varepsilon ^{2^{*} (\gamma (\mu_1)- \frac{N-4}{2})}). \end{eqnarray*}$

引理4.1证毕.

引理4.2 在定理1.5的条件下, 可以得出

$\begin{eqnarray} \sup\limits_{t\geq0} J(tu_{\mu_1 }^\varepsilon ,t\hat{\tau}_{\min}u_{\mu_2 }^\varepsilon )< c^* = \frac{ 2}{N} S (\mu_1,\mu_2)^{ {N}/{4}} . \end{eqnarray}$

 假设条件 $({\cal H}_2)$ 成立, 则下列二次型

$\begin{eqnarray} Q(u,v):= (u,v) A \bigg( \begin{array}{ll} u \\ v \end{array} \bigg) = a_1u^2+2a_2 uv+a_3 v^2 \end{eqnarray}$

是正定的并且满足

$\begin{eqnarray} \lambda _1 (u^2+v^2)\le Q(u,v)\le \lambda _2 (u^2+v^2),\ \ \ \forall (u, v) \in H\times H, \end{eqnarray}$

由此可以推出

$\begin{eqnarray*} g_1(t)&:=&J(tu_{\mu_1 }^\varepsilon ,t\hat{\tau}_{\min}u_{\mu_2 }^\varepsilon )\\ &\le& \frac{t^2}{2} \int_{\Omega }\Bigr ( E( u_{\mu_1 }^{\varepsilon }, \hat{t}_{\min} u_{\mu_2 }^{\varepsilon } ) - \lambda _1 ( |u_{\mu_1 }^\varepsilon |^2 +\hat{t}_{\min}^2 |u_{\mu_2 }^\varepsilon |^2 )\Bigr ) - \frac{ t^{2^*}}{2^*}\int_{\Omega } F( u_{\mu_1 }^{\varepsilon },\hat{t}_{\min} u_{\mu_2 }^{\varepsilon } ) . \end{eqnarray*}$

通过计算我们可以得出

$\begin{eqnarray} 0\le \mu_1 \le \mu^* \ \Leftarrow \hspace {-.2cm}=\hspace {-.2cm}\Rightarrow \ \gamma (\mu_1) - \frac{N-4}{2} \ge 2 , \end{eqnarray}$ (4.4)
$\begin{eqnarray} \max\limits_{t\geq 0}\Bigr (\frac{t^2}{2}C_1-\frac{t^{2^*}}{2^*}C_2\Bigr )=\frac{2}{N} \Bigr (C_1C_2^{-\frac{2 }{2^*}}\Bigr ) ^{\frac{N}{4}} ,\ \ C_1>0,\ C_2>0. \end{eqnarray}$ (4.5)

由(4.4)式, (4.5)式, 引理2.3和引理4.1可得

$\begin{eqnarray*} \sup\limits_{t\geq0} g_1(t ) &\leq& \frac{2}{N} \Bigg( \frac{ \int_{\Omega }\bigr ( E( u_{\mu_1 }^{\varepsilon }, \hat{t}_{\min} u_{\mu_2 }^{\varepsilon } ) - \lambda _1 ( |u_{\mu_1 }^\varepsilon |^2 +\hat{t}_{\min}^2 |u_{\mu_2 }^\varepsilon |^2 )\bigr ) }{ \bigr ( \int_{\Omega }F( u_{\mu_1 }^{\varepsilon }, \hat{t}_{\min} u_{\mu_2 }^{\varepsilon } ) \bigr )^{\frac{2}{2^*}} } \Bigg)^{\frac{N}{4}} \\ &\leq& \frac{2}{N} \Bigg( \frac{S(\mu_1,\mu_2)^{\frac{N }{4 }} +O (\varepsilon ^{2(\gamma (\mu_1)- \frac{N-4}{2} )} )-O_1( \int_{\Omega }|u_{\mu_1 }^\varepsilon |^2 )}{ (S(\mu_1,\mu_2)^{\frac{N }{4 }} +O (\varepsilon ^{2^*(\gamma (\mu_1)- \frac{N-4}{2}) }) )^\frac{2}{2^*} } \Bigg)^{\frac{N}{4}} \\ & \leq &\frac{ 2}{N} S(\mu_1,\mu_2)^{\frac{N}{4}}+O(\varepsilon ^{2(\gamma (\mu_1)- \frac{N-4}{2}) }) -O_1\Bigr ( \int_{\Omega }|u_{\mu_1 }^\varepsilon |^2\Bigr ) \\ & <& \frac{ 2}{N} S(\mu_1,\mu_2) ^{\frac{N}{4}}. \end{eqnarray*}$

引理证毕.

定理1.5的证明 由山路引理[21-22], 引理2.1和引理4.2, 可得能量泛函 $J $ 的临界点 $(u_0,v_0)\in H\times H$ 满足双调和方程组(1.1).另外, 由假设条件 $({\cal H}_2)$ $a_i\not =0 $ 可知 $u_0\not \equiv0,$ $v_0\not \equiv0 $ .定理证毕.

参考文献
[1] Davies E, Hinz A. Explicit constants for Rellich ineqaulities in Lp(Ω). Math Z, 1998, 227(2): 511–523.
[2] Edmunds D, Fortunato D, Jannelli E. Critical exponents, critical dimensions and the biharmonic operator. Arch Ration Mech Anal, 1990, 112(1): 269–289.
[3] Bhakta M, Musina R. Entire solutions for a class of variational problems involving the biharmonic operator and Rellich potentials. Nonlinear Anal, 2012, 75(12): 3836–3848.
[4] D'Ambrosio L, Jennelli E. Nonlinear critical problems for the biharmonic operator with Hardy potential. Calc Var Partial Differential Equations, 2015, 54(1): 365–396. DOI:10.1007/s00526-014-0789-7
[5] Bhakta M. Caffarelli-Kohn-Nirenberg type equations of fourth order with the critical exponent and Rellich potential. J Math Anal Appl, 2016, 433(1): 681–700. DOI:10.1016/j.jmaa.2015.07.042
[6] Hardy G, Littlewood J, Polya G. Inequalities. Cambridge: Cambridge University Press, 1952.
[7] Chen Z, Zou W. Existence and symmetry of positive ground states for a doubly critical Schrödinger system. Trans Amer Math Soc, 2015, 367(11): 3599–3646.
[8] Felli V, Terracini S. Elliptic equations with multi-singular inverse-square potentials and critical nonlinearity. Comm Partial Differential Equations, 2006, 31(2): 469–495.
[9] Abdellaoui B, Felli V, Peral I. Some remarks on systems of elliptic equations doubly critical in the whole $\mathbb{R}^{N}$. Calc Var Partial Differential Equations, 2009, 34(1): 97–137. DOI:10.1007/s00526-008-0177-2
[10] Jannelli E. The role played by space dimension in elliptic critcal problems. J Differential Equations, 1999, 156(2): 407–426. DOI:10.1006/jdeq.1998.3589
[11] Kang D. Positive minimizers of the best constants and solutions to coupled critical quasilinear systems. J Differential Equations, 2016, 260(1): 133–148. DOI:10.1016/j.jde.2015.08.042
[12] Kang D, Luo J, Shi X L. Solutions to elliptic systems involving doubly critical nonlinearities and Hardy-type potentials. Acta Mathematica Scientia, 2015, 35B(2): 423–438.
[13] Zhang J, Ma S W. Infinitely many sign-changing solutions for the Brezis-Nirenberg problem involving Hardy-potential. Acta Mathematica Scientia, 2016, 36B(2): 527–536.
[14] Terracini S. On positive entire solutions to a class of equations with a singular coefficient and critical exponent. Adv Differential Equations, 1996, 1(2): 241–264.
[15] Kang D. Concentration compactness principles for the systems of critical elliptic equations. Differ Equ Appl, 2012, 4(2): 435–444.
[16] Lions P L. The concentration compactness principle in the calculus of variations, the limit case (I). Rev Mat Iberoamericana, 1985, 1(1): 145–201.
[17] Lions P L. The concentration compactness principle in the calculus of variations, the limit case (Ⅱ). Rev Mat Iberoamericana, 1985, 1(2): 45–121.
[18] Lieb E, Loss M. Analysis. Providence, RI: American Mathematical Society, 2001.
[19] Willem M. Analyse Fonctionnelle Élémentaire. Paris: Cassini Éditeurs, 2003.
[20] Kang D. Systems of quasilinear elliptic equations involving multiple homogeneous nonlinearities. Appl Math Lett, 2014, 37(1): 1–6.
[21] Ambrosetti A, Rabinowitz H. Dual variational methods in critical point theory and applications. J Funct Anal, 1973, 14(2): 349–381.
[22] Brezis H, Nirenberg L. Positive solutions of nonlinear elliptic equations involving critical Sobolev exponents. Comm Pure Appl Math, 1983, 36(2): 437–477.