数学物理学报, 2024, 44(1): 60-79

次临界 Choquard 方程的多解

温瑞江,1, 刘范琴,2, 徐子怡,3,*

1.江西师范大学数学与统计学院 南昌 330022

2.首都师范大学数学科学学院 北京 100048

3.兰州大学数学与统计学院 兰州 730000

Multiplicity of Positive Solutions to Subcritical Choquard Equation

Wen Ruijiang,1, Liu Fanqin,2, Xu Ziyi,3,*

1. School of Mathematics and Statistics, Jiangxi Normal University, Nanchang 330022

2. School of Mathematical Sciences, Capital Normal University, Beijing 100048

3. School of Mathematics and Statistics, Lanzhou University, Lanzhou 730000

通讯作者: 徐子怡, E-mail:XuZiyi0822@outlook.com

收稿日期: 2022-09-20   修回日期: 2023-10-16  

Received: 2022-09-20   Revised: 2023-10-16  

作者简介 About authors

温瑞江,E-mail:ruijiangwen@126.com

刘范琴,E-mail:fanqliu@163.com

摘要

该文考虑次临界 Choquard 方程$\begin{cases} -{\Delta}{u}+(\lambda V(x)+1)u=\Big(\int_{\mathbb{R}^N} \frac{|u(y)|^{p_{\varepsilon}}}{|x-y|^\mu} {\rm d}y\Big)|u|^{p_{\varepsilon}-2}u,\quad x\in\mathbb{R}^N,\\{u\in{{H}}^{1}{(\mathbb{R}^N)}} \end{cases}$

多解的存在性, 其中$N>3$,$\lambda$是正实参数,$p_{\varepsilon}=2^\ast_{\mu}-\varepsilon$,$\varepsilon>0$,$0<\mu<N$,$2^\ast_{\mu}=\frac{2N-\mu}{N-2}$是 Hardy-Littlewood-Sobolev 不等式意义下的临界指数. 假定$\Omega:={\rm int}\,V^{-1}(0)$是$\mathbb{R}^N$中非空带光滑边界的有界区域, 利用 Lusternik-Schnirelman 定理,该文证明了当$\lambda$足够大及$\varepsilon$充分小时, 方程(0.1)至少有$cat_\Omega(\Omega)$个正解.

关键词: 次临界 Choquard 方程; Lusternik-Schnirelman 定理; 解的多重性

Abstract

In this paper, we are concerned with the multiplicity of solutions for the following subcritical Choquard equation$\begin{equation*} \begin{cases} -{\Delta}{u}+(\lambda V(x)+1)u=\Big(\int_{\mathbb{R}^N}\frac{|u(y)|^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}y\Big)|u|^{p_{\varepsilon}-2}u,\quad x\in\mathbb{R}^N,\\{u\in{{H}}^{1}{(\mathbb{R}^N)}}, \end{cases}\end{equation*}$

where$N>3$,$\lambda$is a real parameter,$p_{\varepsilon}=2^\ast_{\mu}-\varepsilon$,$\varepsilon>0$,$\mu\in(0,N)$and$2^\ast_{\mu}=\frac{2N-\mu}{N-2}$is the critical Hardy-Littlewood-Sobolev exponent. Suppose that$\Omega:={\rm int}\,V^{-1}(0)$is a nonempty bounded domain in$\mathbb{R}^N$with smooth boundary, using Lusternik-Schnirelman theory, we prove the problem (0.1) has at least$cat_\Omega(\Omega)$positive solutions for$\lambda$large and$\varepsilon$small enough.

Keywords: Subcritical Choquard equation; Lusternik-Schnirelman theory; Multiplicity of solutions

PDF (666KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

温瑞江, 刘范琴, 徐子怡. 次临界 Choquard 方程的多解[J]. 数学物理学报, 2024, 44(1): 60-79

Wen Ruijiang, Liu Fanqin, Xu Ziyi. Multiplicity of Positive Solutions to Subcritical Choquard Equation[J]. Acta Mathematica Scientia, 2024, 44(1): 60-79

1 引言

本文研究次临界 Choquard 方程

$\begin{cases} -{\Delta}{u}+(\lambda V(x)+1)u=\Big(\int_{\mathbb{R}^N}\frac{|u(y)|^{p_{\varepsilon}}}{|x-y|^\mu} {\rm d}y\Big)|u|^{p_{\varepsilon}-2}u,\quad x\in\mathbb{R}^N,\\{u\in{{H}}^{1}{(\mathbb{R}^N)}}\end{cases}$

多解的存在性, 其中$N>3$,$\lambda$是正实参数,$p_{\varepsilon}=2^\ast_{\mu}-\varepsilon$,$\varepsilon>0$,$0<\mu<N$,$2^\ast_{\mu}=\frac{2N-\mu}{N-2}$是 Hardy-Littlewood-Sobolev 不等式意义下的临界指数.

1954 年, 物理学家 Pekar[27] 研究了极化子处于静止时的理论, 并首次提出了 Choquard 方程, 随后 Choquard 在 1976 年将它引入到等离子体的 Hartree-Fock 方程.

1977 年, Lieb[18] 证明了次临界 Choquard 方程

$-{\Delta}{u}+u=(|x|^{-1}\ast|u|^{2})u,\quad\quad u>0,\quad x\in\mathbb{R}^{3}$

有唯一解. Choquard 方程的理论发展至今, 研究成果非常丰富[10,13-17,21-25,29].

2000 年, Bartsch 和 Wang[6]研究了次临界椭圆方程

$\begin{aligned} -{\Delta}{u}+(\lambda V(x)+1)u&=u^{p-1},\quad &x\in\mathbb{R}^N,\\ u&>0,\quad\quad &x\in\mathbb{R}^N,\\ \end{aligned}$

其中$V(x)$满足下列条件

(A$_{1})$$V(x)\in \mathcal{C}(\mathbb{R}^N, \mathbb{R})$,$V(x)\geq0$,$\Omega:={\rm int}\, V^{-1}(0)$是$\mathbb{R}^N$中非空带光滑边界的有界集,$\bar{\Omega}=V^{-1}(0)$.

(A$_{2})$存在常数$M_{0}>0$, 有

$\begin{equation*}\mu(\{x\in\mathbb{R}^N:V(x)\leq M_{0}\})<\infty,\end{equation*}$

其中$\mu$是$\mathbb{R}^N$上的 Lebesgue 测度. Bartsch 等利用极限方程

$\begin{equation}-{\Delta}{u}+u=u^{p-1},\quad\quad u\in{{H}}^{1}_{0}{(\Omega)}\end{equation}$

在指数$p$靠近临界指数$2^*:=\frac{2N}{N-2}$时存在$cat_{\Omega}(\Omega)$个正解, 来证明当方程(1.3)中$\lambda$较大同时指数$p$靠近$2^*$时, 方程(1.3)有$cat_{\Omega}(\Omega)$个正解, 这里$cat_X(Y)$指$Y$在$X$中的 Lusternik-Schnirelman 畴数, 即在拓扑空间$X$中能覆盖闭子集$Y\subset X$的最少的子集个数, 这些用来覆盖的子集要求是闭集并且可收缩.

而在文献[11]中, Clapp 和丁彦恒沿用 (A$_{1})$, (A$_{2})$对位势的假设考虑了$p=2^*$的情形, 并利用极限方程的多解性以及 Lusternik-Schnirelman 定理, 证明了当$\lambda$足够大时, 问题

$\begin{equation}-{\Delta}{u}+\lambda V(x)u=\mu u+u^{2^{\ast}-1},\quad\quad x\in\mathbb{R}^N\end{equation}$

至少有$cat_{\Omega}(\Omega)$个正解. Alves 和 Barro[2]进一步证明了临界薛定谔方程

$\begin{equation}-{\Delta}{u}+\lambda V(x)u=\mu u^{q-1}+u^{2^{\ast}-1},\quad\quad x\in\mathbb{R}^N\end{equation}$

在$q\in(2,2^{\ast})$时, 至少有$cat_{\Omega}(\Omega)$个正解. Ma 等[20]用类似的方法考虑了分数阶拉普拉斯方程

$\begin{equation}(-{\Delta})^{s}{u}+\lambda V(x)u=\mu u^{q-1}+u^{2^{\ast}-1},\quad\quad x\in\mathbb{R}^N\end{equation}$

并证明方程(1.7)有$cat_{\Omega}(\Omega)$个正解. 最近, 在文献[28]中已经考虑了临界 Choquard 问题,

$\begin{equation}-{\Delta}{u}+(\lambda V(x)-\beta)u=(|x|^{-\mu}\ast|u|^{2^*_\mu})|u|^{2^*_\mu-2}u,\quad\quad x\in\mathbb{R}^N\end{equation}$

并得到多解的结果, 进一步在文献[30]中考虑了多临界的情形. 更多关于多解问题的研究, 见文献[1,3-5,7-9,31]等.

受到以上文章的启发, 本文将证明问题(1.1)在(A$_{1})$和(A$_{2})$假设下的多解性结果. 在空间$E_{\lambda}=(E,\|\cdot\|_{\lambda})$中考虑解的存在性, 其中

$\begin{equation*}E=\bigg\{u\in{{H}}^{1}{(\mathbb{R}^N)}:\int_{\mathbb{R}^N}V(x)u^{2}\,{\rm d}x<\infty\bigg\},\end{equation*}$

其上内积定义为

$\begin{equation*}\langle u,v\rangle_{\lambda}=\int_{\mathbb{R}^N}(\nabla u\nabla v+(\lambda V(x)+1)uv)\,{\rm d}x.\end{equation*}$

由内积诱导的范数为$\|\cdot\|_{\lambda}$,

$\begin{equation*}\|u\|_{\lambda}=\bigg(\int_{\mathbb{R}^N}(|\nabla u|^{2}+(\lambda V(x)+1)u^{2}\bigg)\,{\rm d}x)^{\frac{1}{2}}.\end{equation*}$

为了寻找方程(1.1)的正解, 类似文献[14,命题 3.1]中, 考虑方程(1.1)对应的能量泛函为

$\begin{equation*}I_{\lambda,p_{\varepsilon}}(u):=\frac{1}{2}\int_{\mathbb{R}^N}(|\nabla u|^{2}+(\lambda V(x)+1)u^{2}){\rm d}x-\frac{1}{2p_{\varepsilon}}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N} \frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^{\mu}}{{\rm d}x{\rm d}y},\end{equation*}$

其中$u^+(x):=\max\{u(x),0\}$是函数$u$的正部. 由命题 2.1 (见第 2 节) 知, 泛函$I_{\lambda,p_{\varepsilon}}$在空间$E_\lambda$中取值有意义并且可证$I_{\lambda,p_{\varepsilon}}\in \mathcal{C}^{1}(E_{\lambda},\mathbb{R})$. 设$\mathcal{M}_{\lambda,p_{\varepsilon}}$是与$I_{\lambda,p_{\varepsilon}}$有关的 Nehari 流形, 即

$\begin{equation*}\mathcal{M}_{\lambda,p_{\varepsilon}}:=\bigg\{u\in E_{\lambda}\setminus\{0\}:\int_{\mathbb{R}^N}(|\nabla u|^{2}\!+\!(\lambda V(x)\!+\!1)u^{2}){\rm d}x=\!\int_{\mathbb{R}^N}\int_{\mathbb{R}^N} \frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^{\mu}}{{\rm d}x{\rm d}y}\bigg\}.\end{equation*}$

$\begin{equation*}c_{\lambda,p_{\varepsilon}}:=\inf\{I_{\lambda,p_{\varepsilon}}(u):u\in\mathcal{M}_{\lambda,p_{\varepsilon}}\}.\end{equation*}$

当$\lambda$足够大时, 问题(1.1)的极限方程为

$\begin{cases} -{\Delta}{u}+u=\Big(\int_{\Omega}\frac{|u(y)|^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}y\Big) |u|^{p_{\varepsilon}-2}u,\quad x\in\Omega,\\ u\in H^1_0(\Omega). \end{cases}$

对于方程(1.1)的解序列$u_{\lambda_n}$, 当$\lambda_n\rightarrow\infty$时, 若$u_{\lambda_n}$在$H^1(\mathbb{R}^N)$中强收敛到方程(1.9)的解$u$, 则称方程(1.1)的解序列$u_{\lambda_n}$集中到方程(1.9)的解$u$.

本文首先得到方程(1.1)基态解的存在性, 并给出了当$\lambda_n\rightarrow\infty$时, 基态解$u_{\lambda_n}$的集中性.

定理 1.1 假设 (A$_{1})$和 (A$_{2})$成立, 则当$\lambda$足够大时, 方程(1.1)存在对应于$c_{\lambda,p_{\varepsilon}}$的非负基态解$u_\lambda$, 并且对任意序列$\lambda_{n}\rightarrow\infty$, 解序列$u_{{\lambda_n}}$有子列在${{H}}^{1}{(\mathbb{R}^N)}$中收敛到极限方程(1.9)的基态解.

利用变分法来证明定理 1.1, 其中主要的困难是克服失紧的问题, 而关于$V(x)$的条件 (A$_2)$对紧性的证明起了关键作用.

对于更一般情形下方程(1.1)解的极限行为, 有如下结果.

定理 1.2 假设 (A$_{1})$和 (A$_{2})$成立, 当$\lambda_{n}\rightarrow\infty$且$\underset{n\rightarrow\infty}{\lim\sup}\,I_{\lambda_{n},p_{\varepsilon}}(u_n)<\infty$, 其中$\{u_{n}\}_{n\in \mathbb{N}}$是方程(1.1)的解序列, 则$u_{n}$在${{H}}^{1}{(\mathbb{R}^N)}$中有子列收敛到方程(1.9)在${{H}}^{1}_{0}{(\Omega)}$中的一个解.

由定理 1.1 知, 当$\lambda_n$较大时, 方程(1.1)的解会集中到方程(1.9)的解, 而在文献[14]中已经证明了方程(1.9)有$cat_\Omega(\Omega)$个解, 那么考虑问题(1.1) 是否也会有类似的结果, 接下来给出定理 1.3.

定理 1.3 假设 (A$_{1})$和 (A$_{2})$成立, 则存在$\varepsilon_{0}>0$, 使得对任意$\varepsilon\in(0,\varepsilon_{0}]$, 当$\lambda\geq\Lambda(\varepsilon)$时, 方程(1.1)至少有$cat_{\Omega}(\Omega)$个正解.

多解的结果是利用 Lusternik-Schnirelman 定理来证明的, 关键是建立泛函在 Nehari 流形上的水平集与$\Omega$之间的联系, 最后得到泛函有$cat_{\Omega}(\Omega)$个临界点. 遇到的困难主要有以下几点: 临界情形下的紧性缺失问题; 对能量泛函的估计; 建立区域拓扑与解个数的关系.

本文结构如下: 在第 2 节给出预备知识; 第 3 节证明$(PS)_{c}$条件并证明定理 1.1; 在第 4 节先证明解的集中性而后给出多解的结果.

2 预备知识

为了用变分法研究问题(1.1), 需要以下 Hardy-Littlewood-Sobolev 不等式.

命题 2.1[19] 设$t, r>1$且$0<\mu<N$,$\frac{1}{t}+\frac{\mu}{N}+\frac{1}{r}=2$,$f\in L^{t}({\mathbb{R}}^N)$,$h\in L^{r}({\mathbb{R}}^N)$, 则存在与$f, h$无关的常数$C(t,r,\mu,N)$, 使得

$\begin{equation}\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}\frac{f(x)h(y)}{|x-y|^\mu}{\rm d}x{\rm d}y\leq C(t,r,\mu,N)|f|_{L^t} |h|_{L^r}.\end{equation}$

若$t=r=\frac{2N}{2N-\mu}$, 则

$\begin{equation*}C(t,r,\mu,N)=C(N,\mu)=\pi^{\frac{\mu}{2}}\frac{\Gamma(\frac{N}{2}-\frac{\mu}{2})}{\Gamma(N-\frac{\mu}{2})}\bigg\{\frac{\Gamma(\frac{N}{2})}{\Gamma(N)}\bigg\}^{-1+\frac{\mu}{N}}.\end{equation*}$

(2.1)式不等式成立当且仅当$f\equiv Ch$, 且达到函数为

$\begin{equation*}h(x)=A(\gamma^2 +|x-a|^{2})^{\frac{-(2N-\mu)}{2}},\end{equation*}$

其中$A\in \mathbb{C}$,$0\neq \gamma \in \mathbb{R}$,$a\in \mathbb{R}^N$.

根据 Hardy-Littlewood-Sobolev 不等式(2.1), 可以得到

$\begin{equation*}\int_{\Omega}\int_{\Omega} \frac{|u(x)|^{p_{\varepsilon}}|u(y)|^{p_{\varepsilon}}}{|x-y|^{\mu}}{{\rm d}x{\rm d}y}\leq C(N,\mu)|u|^{2p_{\varepsilon}}_{\frac{2Np_{\varepsilon}}{2N-\mu}}.\end{equation*}$

可以验证, 对于$v\in E_\lambda$, 有

$\langle I'_{\lambda,p_\varepsilon}(u),v\rangle=\int_{\mathbb{R}^N}\nabla u\nabla v+(\lambda V(x)+1)uv\,{\rm d}x-\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}\frac{(u^+(x))^{p_\varepsilon}(u^+(y))^{p_\varepsilon-1}v(y)}{|x-y|^\mu}{\rm d}x{\rm d}y.$

称$u\in E_\lambda$是方程(1.1)的弱解, 如果对于任意$v\in E_\lambda$, 有

$\int_{\mathbb{R}^N}\nabla u\nabla v+(\lambda V(x)+1)uv\,{\rm d}x-\int_{\mathbb{R}^{N}}\int_{\mathbb{R}^{N}}\frac{(u^+(x))^{p_\varepsilon}(u^+(y))^{p_\varepsilon-1}v(y)}{|x-y|^\mu}{\rm d}x{\rm d}y=0,$

即$\langle I'_{\lambda,p_\varepsilon}(u),v\rangle=0$.

3$(PS)_{c}$条件

本节将给出$(PS)_{c}$条件的证明.

引理 3.1 令$\{u_{n}\}\subset E_{\lambda}$是泛函$I_{\lambda,p_{\varepsilon}}$的$(PS)_{c}$序列, 则当$n\rightarrow\infty$时,$\{u_{n}\}$在$E_\lambda$中有界, 并且$c\geq0$.

因为$\{u_{n}\}$是$(PS)_{c}$列, 即满足

$\begin{equation*}I_{\lambda,p_{\varepsilon}}(u_{n})\rightarrow c,\ \ I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n})\rightarrow 0,\end{equation*}$

则当$n$很大时,

$\begin{equation*}c+1+o(1){\rVert u_{n} \rVert} _{\lambda}\geq I_{\lambda,p_{\varepsilon}}(u_{n})-\frac{1}{2p_{\varepsilon}}\langle{I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle=(\frac{1}{2}-\frac{1}{2p_{\varepsilon}}){\rVert u_{n} \rVert}^{2} _{\lambda},\end{equation*}$

因此${\rVert u_{n} \rVert}_{\lambda}$有界. 另一方面,

$\begin{equation*}0\leq(\frac{1}{2}-\frac{1}{2p_{\varepsilon}}){\rVert u_{n} \rVert}^{2} _{\lambda}=I_{\lambda,p_{\varepsilon}}(u_{n})-\frac{1}{2p_{\varepsilon}}\langle{I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle\leq c+o_{n}(1),\end{equation*}$

所以可得$c\geq0$. 证毕.

引理 3.2 令$c>0$,$\{u_{n}\}$是$(PS)_{c}$序列. 若在$E_{\lambda}$中$u_{n}\rightharpoonup u$且$u\not\equiv0$, 则

$I_{\lambda,p_{\varepsilon}}(v_{n})-I_{\lambda,p_{\varepsilon}}(u_{n})+I_{\lambda,p_{\varepsilon}}(u)=o_{n}(1),$
$\|I_{\lambda,p_{\varepsilon}}^{\prime}(v_{n})-I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n})+I_{\lambda,p_{\varepsilon}}^{\prime}(u)\|_{E_{\lambda}^{\prime}}=o_{n}(1),$

其中$E_{\lambda}^{\prime}$是$E_{\lambda}$的对偶空间,$v_{n}=u_{n}-u$, 并且$\{v_{n}\}$是一个$(PS)_{c-I_{\lambda,p_{\varepsilon}}(u)}$序列.

首先证明

$\begin{equation*}I_{\lambda,p_{\varepsilon}}(v_{n})-I_{\lambda,p_{\varepsilon}}(u_{n})+I_{\lambda,p_{\varepsilon}}(u)=o_{n}(1).\end{equation*}$

$\begin{align*}&I_{\lambda,p_{\varepsilon}}(v_{n})-I_{\lambda,p_{\varepsilon}}(u_{n})+I_{\lambda,p_{\varepsilon}}(u)\\=\ &\frac{1}{2}({\rVert v_{n} \rVert}^{2}_{\lambda}-{\rVert u_{n} \rVert}^{2}_{\lambda}+{\rVert u \rVert}^{2}_{\lambda})\\&-\frac{1}{2p_{\varepsilon}}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(x))^{p_{\varepsilon}}(v_{n}^+(y))^{p_{\varepsilon}}-(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}}+(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y,\end{align*}$

又因为在$E_{\lambda}$中,$u_{n}\rightharpoonup u$, 所以有

$\begin{align*}&I_{\lambda,p_{\varepsilon}}(v_{n})-I_{\lambda,p_{\varepsilon}}(u_{n})+I_{\lambda,p_{\varepsilon}}(u)\\=\ &o_{n}(1)+\frac{1}{2p_{\varepsilon}}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(-(v_{n}^+(x))^{p_{\varepsilon}}+(u_{n}^+(x))^{p_{\varepsilon}}-(u^+(x))^{p_{\varepsilon}})}{|x-y|^\mu}{\rm d}x{\rm d}y\\&+\frac{1}{2p_{\varepsilon}}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u_{n}^+(y))^{p_{\varepsilon}}(-(v_{n}^+(x))^{p_{\varepsilon}}+(u_{n}^+(x))^{p_{\varepsilon}}-(u^+(x))^{p_{\varepsilon}})}{|x-y|^\mu}{\rm d}x{\rm d}y\\&+\frac{1}{2p_{\varepsilon}}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u^+(y))^{p_{\varepsilon}}(-(v_{n}^+(x))^{p_{\varepsilon}}+(u_{n}^+(x))^{p_{\varepsilon}}-(u^+(x))^{p_{\varepsilon}})}{|x-y|^\mu}{\rm d}x{\rm d}y\\&+\frac{1}{p_{\varepsilon}}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(u^+(x))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y.\end{align*}$

利用 Hardy-Littlewood-Sobolev 不等式,

$\begin{align*}&\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}((v_{n}^+(x))^{p_{\varepsilon}}-(u_{n}^+(x))^{p_{\varepsilon}}+(u^+(x))^{p_{\varepsilon}})}{|x-y|^\mu}{\rm d}x{\rm d}y\\\leq\ & C\bigg(\int_{\mathbb{R}^N}(v_{n}^+)^{\frac{p_{\varepsilon}2N}{2N-\mu}}{\rm d}x\bigg)^{\frac{2N-\mu}{2N}}\bigg(\int_{\mathbb{R}^N}|(v_{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}|^{\frac{2N}{2N-\mu}}{\rm d}x\bigg)^{\frac{2N-\mu}{2N}}.\end{align*}$

注意到

$\begin{align*}&\int_{\mathbb{R}^N}|(v_{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}|^{\frac{2N}{2N-\mu}}{\rm d}x\\=&\int_{B_{R}(0)}|(v_{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}|^{\frac{2N}{2N-\mu}}{\rm d}x+\int_{\mathbb{R}^N\backslash B_{R}(0)}|(v_{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}|^{\frac{2N}{2N-\mu}}{\rm d}x,\end{align*}$

其中$R>0$. 因为在$E_{\lambda}$中,$u_{n}\rightharpoonup u$, 可得

$\begin{align*}&u_{n}\rightarrow u\quad L^{\frac{p_{\varepsilon}2N}{2N-\mu}}(B_{R}(0)),\\&u_{n}\rightarrow u\quad a.e.\ \mathbb{R}^N,\end{align*}$

不妨假设$u^+\not\equiv0$, 所以

$\begin{align*}&u_{n}^+\rightarrow u^+\quad L^{\frac{p_{\varepsilon}2N}{2N-\mu}}(B_{R}(0)),\\&u_{n}^+\rightarrow u^+\quad a.e.\ \mathbb{R}^N,\end{align*}$

那么存在$h_{1}\in L^{\frac{p_{\varepsilon}2N}{2N-\mu}}(B_{R}(0))$使得

$\begin{equation*}u_{n}^+(x)\leq h_{1}(x)\quad a.e.\ B_{R}(0).\end{equation*}$

因此

$\begin{equation*}(v_{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}\rightarrow0\ \quad a.e.\ B_{R}(0),\end{equation*}$

并且

$\begin{equation*}|(v_{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}|^{\frac{2N}{2N-\mu}}\leq(2^{p_{\varepsilon}}+1)^{\frac{2N}{2N-\mu}}(h_{1}+u^+)^{\frac{2Np_{\varepsilon}}{2N-\mu}}.\end{equation*}$

再利用勒贝格控制收敛定理, 则

$\begin{equation} \int_{B_{R}(0)}|(v_{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}|^{\frac{2N}{2N-\mu}}{\rm d}x\rightarrow0.\end{equation}$

同样可得

$\begin{equation*}|((u_{n}-u)^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}|\leq p_{\varepsilon}2^{p_{\varepsilon}-1}(|u_{n}|^{p_{\varepsilon}-1}|u|+|u|^{p_{\varepsilon}}).\end{equation*}$

因此有

$\begin{align*}&\int_{\mathbb{R}^N\backslash B_{R}(0)}|(v_{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}|^{\frac{2N}{2N-\mu}}{\rm d}x\\\leq\ & C\int_{\mathbb{R}^N\backslash B_{R}(0)}|u_{n}|^{\frac{2N(p_{\varepsilon}-1)}{2N-\mu}}|u|^{\frac{2N}{2N-\mu}}{\rm d}x+C\int_{\mathbb{R}^N\backslash B_{R}(0)}|u|^{\frac{2Np_{\varepsilon}}{2N-\mu}}{\rm d}x.\end{align*}$

任意给定$\delta>0$, 选择$R>0$使得

$\begin{equation*}\int_{\mathbb{R}^N\backslash B_{R}(0)}|u|^{\frac{2Np_{\varepsilon}}{2N-\mu}}{\rm d}x\leq\delta.\end{equation*}$

再根据 Hölder 不等式以及$\{u_{n}\}$有界可知

$\begin{equation}\int_{\mathbb{R}^N\backslash B_{R}(0)}|(v _{n}^+)^{p_{\varepsilon}}-(u_{n}^+)^{p_{\varepsilon}}+(u^+)^{p_{\varepsilon}}|^{\frac{2N}{2N-\mu}}{\rm d}x\leq\delta.\end{equation}$

结合 (3.1), (3.2) 式以及$\{v_{n}\}$有界, 当$n\rightarrow\infty$时, 有

$\begin{equation*}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(-(v_{n}^+(x))^{p_{\varepsilon}}+(u_{n}^+(x))^{p_{\varepsilon}}-(u^+(x))^{p_{\varepsilon}})}{|x-y|^\mu}{\rm d}x{\rm d}y\rightarrow0.\end{equation*}$

类似可证

$\begin{equation*}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u_{n}^+(y))^{p_{\varepsilon}}(-(v_{n}^+(x))^{p_{\varepsilon}}+(u_{n}^+(x))^{p_{\varepsilon}}-(u^+(x))^{p_{\varepsilon}})}{|x-y|^\mu}{\rm d}x{\rm d}y\rightarrow0\end{equation*}$

$\begin{equation*}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u^+(y))^{p_{\varepsilon}}(-(v_{n}^+(x))^{p_{\varepsilon}}+(u_{n}^+(x))^{p_{\varepsilon}}-(u^+(x))^{p_{\varepsilon}})}{|x-y|^\mu}{\rm d}x{\rm d}y\rightarrow0.\end{equation*}$

最后,只需证明

$\begin{equation*}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(u^+(x))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y\rightarrow0.\end{equation*}$

因为在$E_{\lambda}$中,$v_{n}\rightharpoonup 0$,$p_{\varepsilon}=2^\ast_{\mu}-\varepsilon$,$\{|v_{n}|^{p_{\varepsilon}}\}$在$L^{\frac{2N}{2N-\mu}}(\mathbb{R}^N)$中有界; 再根据$v_{n}\rightarrow 0$a.e.$\mathbb{R}^N$, 在$L^{\frac{2N}{2N-\mu}}(\mathbb{R}^N)$中$|v_{n}|^{p_{\varepsilon}}\rightharpoonup0$, 那么有$(v_{n}^+)^{p_{\varepsilon}}\rightharpoonup0$. 由 Hardy-Littlewood-Sobolev 不等式, 可知$F : L^{\frac{2N}{2N-\mu}}(\mathbb{R}^N)\rightarrow \mathbb{R}$,

$\begin{equation*}F(w)=\int_{\mathbb{R}^N}(\frac{1}{|x|^{\mu}}\ast w)|u|^{p_{\varepsilon}}{\rm d}x\end{equation*}$

是连续的. 因此$F((v_{n}^+)^{p_{\varepsilon}})\rightarrow0$. 于是

$\begin{equation*}I_{\lambda,p_{\varepsilon}}(v_{n})-I_{\lambda,p_{\varepsilon}}(u_{n})+I_{\lambda,p_{\varepsilon}}(u)=o_{n}(1),\end{equation*}$

从而有

$\begin{equation*}I_{\lambda,p_{\varepsilon}}(v_{n})\rightarrow c-I_{\lambda,p_{\varepsilon}}(u).\end{equation*}$

其次证明

$\begin{equation*}\|I_{\lambda,p_{\varepsilon}}^{\prime}(v_{n})-I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n})+I_{\lambda,p_{\varepsilon}}^{\prime}(u)\|_{E_{\lambda}^{\prime}}=o_{n}(1).\end{equation*}$

对任意$w\in E_{\lambda}$,

$\begin{align*}&\langle I_{\lambda,p_{\varepsilon}}^{\prime}(v_{n})-I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n})+I_{\lambda,p_{\varepsilon}}^{\prime}(u),w\rangle\\=\ &\langle v_{n}-u_{n}+u,w\rangle_{\lambda}-\int_{\mathbb{R}^N}(\frac{1}{|x|^{\mu}}\ast (v_{n}^+)^{p_{\varepsilon}})(v_{n}^+(x))^{p_{\varepsilon}-1}w{\rm d}x\\&+\int_{\mathbb{R}^N}(\frac{1}{|x|^{\mu}}\ast (u_{n}^+)^{p_{\varepsilon}})(u_{n}^+(x))^{p_{\varepsilon}-1}w{\rm d}x-\int_{\mathbb{R}^N}(\frac{1}{|x|^{\mu}}\ast (u^+)^{p_{\varepsilon}})(u^+(x))^{p_{\varepsilon}-1}w{\rm d}x,\end{align*}$

因此需要证明当$n\rightarrow\infty$时,

$\begin{align*}&\underset{\|w\|_{\lambda}\leq1}{\sup}\bigg|\int_{\mathbb{R}^N}[-(\frac{1}{|x|^{\mu}}\ast (v_{n}^+)^{p_{\varepsilon}})(v_{n}^+(x))^{p_{\varepsilon}-1}\\&+(\frac{1}{|x|^{\mu}}\ast (u_{n}^+)^{p_{\varepsilon}})(u_{n}^+(x))^{p_{\varepsilon}-1}-(\frac{1}{|x|^{\mu}}\ast (u^+)^{p_{\varepsilon}})(u^+(x))^{p_{\varepsilon}-1}]w\,{\rm d}x\bigg|\rightarrow0.\end{align*}$

因为

$\begin{align*}\langle &I_{\lambda,p_{\varepsilon}}^{\prime}(v_{n})-I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n})+I_{\lambda,p_{\varepsilon}}^{\prime}(u),w\rangle\\=\ &\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(-(v_{n}^+(x))^{p_{\varepsilon}-1}+(u_{n}^+(x))^{p_{\varepsilon}-1}-(u^+(x))^{p_{\varepsilon}-1})w}{|x-y|^\mu}{\rm d}x{\rm d}y\\&+\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(-(v_{n}^+(y))^{p_{\varepsilon}}+(u_{n}^+(y))^{p_{\varepsilon}}-(u^+(y))^{p_{\varepsilon}})(u_{n}^+(x))^{p_{\varepsilon}-1}w}{|x-y|^\mu}{\rm d}x{\rm d}y\\&+\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u^+(y))^{p_{\varepsilon}}(-(v_{n}^+(x))^{p_{\varepsilon}-1}+(u_{n}^+(x))^{p_{\varepsilon}-1}-(u^+(x))^{p_{\varepsilon}-1})w}{|x-y|^\mu}{\rm d}x{\rm d}y\\&+\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(u^+(x))^{p_{\varepsilon}-1}w}{|x-y|^\mu}{\rm d}x{\rm d}y+\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u^+(y))^{p_{\varepsilon}}(v_n^+(x))^{p_{\varepsilon}-1}w}{|x-y|^\mu}{\rm d}x{\rm d}y\\=&:K_{1}+K_{2}+K_{3}+K_{4}+K_{5}.\end{align*}$

由 Hardy-Littlewood-Sobolev 不等式和 Hölder 不等式,

$\begin{align*}K_{1}\leq\ & C\bigg(\int_{\mathbb{R}^N}|v_{n}|^{\frac{p_{\varepsilon}2N}{2N-\mu}}{\rm d}x\bigg)^{\frac{2N-\mu}{2N}}\\&\times \bigg(\int_{\mathbb{R}^N}(|(v_{n}^+(x)^{p_{\varepsilon}-1} -(u_{n}^+(x))^{p_{\varepsilon}-1}+(u^+(x))^{p_{\varepsilon}-1})w|^{\frac{2N}{2N-\mu}}{\rm d}x\bigg)^{\frac{2N-\mu}{2N}}\\ \leq\ & C\bigg(\int_{\mathbb{R}^N}|v_{n}|^{\frac{p_{\varepsilon}2N}{2N-\mu}}{\rm d}x\bigg)^{\frac{2N-\mu}{2N}}\\&\times\bigg(\int_{\mathbb{R}^N}|(v_{n}^+(x))^{p_{\varepsilon}-1} -(u_{n}^+(x))^{p_{\varepsilon}-1}+(u^+(x))^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}{\rm d}x\bigg)^{\frac{N-\mu+2}{2N}}\|w\|_{\lambda}.\end{align*}$

注意到

$\begin{align*}&\int_{\mathbb{R}^N}|(v_{n}^+(x))^{p_{\varepsilon}-1}-(u_{n}^+(x))^{p_{\varepsilon}-1}+(u^+(x))^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}{\rm d}x\\=\ &\int_{ B_{R}(0)}|(v_{n}^+(x))^{p_{\varepsilon}-1}-(u_{n}^+(x))^{p_{\varepsilon}-1}+(u^+(x))^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}{\rm d}x\\&+\int_{\mathbb{R}^N\backslash B_{R}(0)}|(v_{n}^+(x))^{p_{\varepsilon}-1}-(u_{n}^+(x))^{p_{\varepsilon}-1}+(u^+(x))^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}{\rm d}x,\end{align*}$

其中$R>0$. 因为在$E_{\lambda}$中,$u_{n}\rightharpoonup u$, 可得

$\begin{align*}&u_{n}\rightarrow u\quad L^{\frac{2N(p_{\varepsilon}-1)}{N-\mu+2}}(B_{R}(0)),\\&u_{n}\rightarrow u\quad a.e.\ \mathbb{R}^N.\end{align*}$

所以

$\begin{align*}&u_{n}^+\rightarrow u^+\quad L^{\frac{2N(p_{\varepsilon}-1)}{N-\mu+2}}(B_{R}(0)),\\&u_{n}^+\rightarrow u^+\quad a.e.\ \mathbb{R}^N,\end{align*}$

那么存在$h_{2}\in L^{\frac{2N(p_{\varepsilon}-1)}{N-\mu+2}}(B_{R}(0))$使得

$\begin{equation*}|u_{n}^+(x)|\leq h_{2}(x)\quad a.e.\ B_{R}(0).\end{equation*}$

由此可知

$\begin{equation*}(v_{n}^+)^{p_{\varepsilon}-1}-(u_{n}^+)^{p_{\varepsilon}-1}+(u^+)^{p_{\varepsilon}-1}\rightarrow0\ \quad a.e.\ B_{R}(0),\end{equation*}$

并且

$\begin{align*}& |(v_{n}^+(x))^{p_{\varepsilon}-1}-(u_{n}^+(x))^{p_{\varepsilon}-1}+(u^+(x))^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}\\&\leq C|(|u_{n}|^{p_{\varepsilon}-2}+|u|^{p_{\varepsilon}-2})|u|+|u_{n}|^{p_{\varepsilon}-1}+|u|^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}\\&\leq C(h_{2}(x)+|u(x)|)^{\frac{2N(p_{\varepsilon}-1)}{N-\mu+2}}.\end{align*}$

再利用勒贝格控制收敛定理

$\begin{equation}\int_{ B_{R}(0)}|(v_{n}^+(x))^{p_{\varepsilon}-1}-(u_{n}^+(x))^{p_{\varepsilon}-1}+(u^+(x))^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}{\rm d}x\rightarrow0.\end{equation}$

又因为

$\begin{equation*}|((u_{n}(x)-u(x))^+)^{p_{\varepsilon}-1}-(u_{n}^+(x))^{p_{\varepsilon}-1}|\leq C(|u(x)|^{p_{\varepsilon}-1}+|u(x)||u_{n}(x)|^{p_{\varepsilon}-2}),\end{equation*}$

因此

$\begin{align*}&\int_{\mathbb{R}^N\backslash B_{R}(0)}|((u_{n}(x)-u(x))^+)^{p_{\varepsilon}-1}-(u_{n}^+(x))^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}{\rm d}x\\\leq\ & C\bigg(\int_{\mathbb{R}^N\backslash B_{R}(0)}|u_{n}(x)|^{\frac{2N(p_{\varepsilon}-2)}{N-\mu+2}}|u(x)|^{\frac{2N}{N-\mu+2}}+|u(x)|^{\frac{2N(p_{\varepsilon}-1)}{N-\mu+2}}{\rm d}x\bigg).\end{align*}$

任意给定$\delta_{1}>0$, 选择$R>0$使得

$\begin{equation*}\int_{\mathbb{R}^N\backslash B_{R}(0)}|u|^{\frac{2N(p_{\varepsilon}-1)}{N-\mu+2}}{\rm d}x\leq\delta_{1},\end{equation*}$

再根据 Hölder 不等式以及$\{u_{n}\}$有界可知

$\begin{equation}\int_{\mathbb{R}^N\backslash B_{R}(0)}|(v_{n}^+(x))^{p_{\varepsilon}-1}-(u_{n}^+(x))^{p_{\varepsilon}-1}+(u^+(x))^{p_{\varepsilon}-1}|^{\frac{2N}{N-\mu+2}}{\rm d}x\leq\delta_{1}.\end{equation}$

综合 (3.3), (3.4) 式以及$\{v_{n}\}$有界, 当$n\rightarrow\infty$时, 有

$\begin{equation*}\underset{\|w\|_{\lambda}\leq1}{\sup}|K_{1}|\rightarrow 0,\end{equation*}$

类似可得

$\begin{equation*}\underset{\|w\|_{\lambda}\leq1}{\sup}|K_{2}|\rightarrow 0 \mbox{和} \underset{\|w\|_{\lambda}\leq1}{\sup}|K_{3}|\rightarrow 0.\end{equation*}$

然后证明

$\begin{equation*}\underset{\|w\|_{\lambda}\leq1}{\sup}\bigg|\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(u^+(x))^{p_{\varepsilon}-1}w}{|x-y|^\mu}{\rm d}x{\rm d}y\bigg|\rightarrow 0.\end{equation*}$

$\begin{align*}&\bigg|\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(u^+(x))^{p_{\varepsilon}-1}w}{|x-y|^\mu}{\rm d}x{\rm d}y\bigg|\\\leq\ &\int_{B_{R}(0)}F_{n}(x)(u^+(x))^{p_{\varepsilon}-1}|w|{\rm d}x+\int_{B_{R}^{c}(0)}F_{n}(x)(u^+(x))^{p_{\varepsilon}-1}|w|{\rm d}x,\end{align*}$

其中

$\begin{equation*}F_{n}(x):=\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}y.\end{equation*}$

当$R\rightarrow\infty$以及$n\rightarrow\infty$时,

$\begin{align*}F_{n}(x)&=\int_{B_{R}(x)}\frac{(v_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}y+\int_{B_{R}^{c}(x)}\frac{(v_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}y\\&\leq\bigg(\int_{B_{R}(x)}|v_{n}(y)|^{\frac{p_{\varepsilon}2N}{2N-\mu}}{\rm d}y\bigg)^{\frac{2N-\mu}{2N}}\bigg(\int_{B_{R}(x)}\frac{1}{|x-y|^{2N}}{\rm d}y\bigg)^{\frac{\mu}{2N}}\\& \ +\bigg(\int_{B_{R}^{c}(x)}|v_{n}(y)|^{\frac{p_{\varepsilon}2N}{2N-\mu}}{\rm d}y\bigg)^{\frac{2N-\mu}{2N}}\bigg(\int_{B_{R}^{c}(x)}\frac{1}{|x-y|^{2N}}{\rm d}y\bigg)^{\frac{\mu}{2N}}\rightarrow0,\end{align*}$

其中$x\in B_{R}(0)$,因此

$\begin{equation*}\bigg|\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(y))^{p_{\varepsilon}}(u^+(x))^{p_{\varepsilon}-1}w}{|x-y|^\mu}{\rm d}x{\rm d}y\bigg|\leq o(1)\|w\|_{\lambda}.\end{equation*}$

类似的对于$K_5$. 因此结合$K_1,K_2,K_3,K_4,K_5$的估计, 有

$\begin{equation*}\begin{aligned}\|I_{\lambda,p_{\varepsilon}}^{\prime}(v_{n})-I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n})+I_{\lambda,p_{\varepsilon}}^{\prime}(u)\|_{E_{\lambda}^{\prime}}=\underset{\|w\|_{\lambda}\leq1}{\sup}|\langle I_{\lambda,p_{\varepsilon}}^{\prime}(v_{n})-I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n})+I_{\lambda,p_{\varepsilon}}^{\prime}(u),w\rangle|=o_{n}(1),\end{aligned}\end{equation*}$

所以

$\begin{equation*}\|I_{\lambda,p_{\varepsilon}}^{\prime}(v_{n})-I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n})+I_{\lambda,p_{\varepsilon}}^{\prime}(u)\|_{E_{\lambda}^{\prime}}=o_{n}(1).\end{equation*}$

证毕.

引理 3.3 假设$\{u_{n}\}$是$(PS)_{c}$序列, 则$c=0$或存在$c_{0}>0$, 对任意的$\lambda>0$, 有$c\geq c_{0}$.

由引理 3.1 可知$c\geq0$. 现假设$c>0$, 易知

$\begin{equation*}c+o_{n}(1)\|u_{n}\|_{\lambda}\geq I_{\lambda,p_{\varepsilon}}(u_{n})-\frac{1}{2p_{\varepsilon}}\langle{I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle=(\frac{1}{2}-\frac{1}{2p_{\varepsilon}}){\rVert u_{n} \rVert}^{2} _{\lambda},\end{equation*}$

从而

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\sup}{\rVert u_{n} \rVert}^{2} _{\lambda}\leq\frac{2p_{\varepsilon}c}{p_{\varepsilon}-1}.\end{equation*}$

根据 Hardy-Littlewood-Sobolev 不等式以及 Sobolev 嵌入定理,

$\begin{equation*}\langle{I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle\geq\frac{1}{2}\|u_{n}\|_{\lambda}^{2}-C\|u_{n}\|_{\lambda}^{2p_{\varepsilon}},\end{equation*}$

其中$C$是正常数. 那么存在$\sigma>0$, 当$\|u_{n}\|_{\lambda}^{2}<\sigma$时, 有

$\begin{equation*}\langle{I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle\geq\frac{1}{4}\|u_{n}\|_{\lambda}^{2}.\end{equation*}$

若$c<c_{0}:=\sigma^{2}\frac{p_{\varepsilon}-1}{2p_{\varepsilon}}$, 则当$n$足够大时, 有

$\begin{equation*}\|u_{n}\|_{\lambda}^{2}\leq\sigma.\end{equation*}$

所以

$\begin{equation*}o(1)\|u_{n}\|_{\lambda}=\langle{I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle\geq\frac{1}{4}\|u_{n}\|_{\lambda}^{2},\end{equation*}$

由此可得, 当$n\rightarrow\infty$时,$\|u_{n}\|_{\lambda}\rightarrow0$; 进一步得$I_{\lambda,p_{\varepsilon}}(u_{n})\rightarrow0$, 这与$c>0$矛盾. 故$c\geq c_{0}$或$c=0$. 证毕.

引理 3.4 存在与$\lambda$无关的$\delta_{0}>0$, 当$\lambda\geq0$,$c>0$时,$(PS)_{c}$序列$\{u_{n}\}$满足

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\inf}{|u_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu}}\geq\delta_{0}c.\end{equation*}$

易知

$\begin{aligned}c=\underset{n\rightarrow\infty}{\lim}(I_{\lambda,p_{\varepsilon}}(u_{n})-\frac{1}{2}\langle{I_{\lambda,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle)=(\frac{1}{2}-\frac{1}{2p_{\varepsilon}})\underset{n\rightarrow\infty}{\lim}\int_{\mathbb{R}^N}(\frac{1}{|x|^{\mu}}\ast(u_{n}^+)^{p_{\varepsilon}})(u_{n}^+)^{p_{\varepsilon}}{\rm d}x.\end{aligned}$

由 Hardy-Littlewood-Sobolev 不等式可得

$\begin{equation*}c\leq(\frac{1}{2}-\frac{1}{2p_{\varepsilon}})C\underset{n\rightarrow\infty}{\lim\inf}{|u_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu}},\end{equation*}$

从而, 选择$\delta_{0}=\frac{2p_{\varepsilon}}{C(p_{\varepsilon}-1)}>0$, 有$\underset{n\rightarrow\infty}{\lim\inf}{|u_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu}}\geq\delta_{0}c.$证毕.

引理 3.5 对任意$C_{1}>0$,$\sigma_{1}>0$, 存在$\Lambda=\Lambda(\sigma_{1})$和$R=R(\sigma_{1},C_{1})$, 当$c\in[C_{1}]$,$\lambda\geq\Lambda$时,$(PS)_{c}$序列$\{u_{n}\}$满足

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\sup}{|u_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu},B_{R}^{c}(0)}\leq\sigma_{1}.\end{equation*}$

利用文献[6,引理 2.5]中的想法. 令$R>0$, 定义

$A(R):=\{x\in\mathbb{R}^N:\ |x|>R,\ V(x)\geq M_{0}\},$
$B(R):=\{x\in\mathbb{R}^N:\ |x|>R,\ V(x)< M_{0}\}.$

由引理 3.3 知, 当$n\rightarrow\infty$时,

$\begin{align*}\int_{A(R)}u_{n}^{2}{\rm d}x&\leq\frac{1}{\lambda M_{0}+1}\int_{\mathbb{R}^N}(\lambda V(x)+1)u_{n}^{2}{\rm d}x\leq\frac{1}{\lambda M_{0}+1}\|u_{n}\|_{\lambda}^{2}\\&\leq\frac{1}{\lambda M_{0}+1}\bigg[\frac{2p_{\varepsilon}}{p_{\varepsilon}-1}C_{1}+o_{n}(1)\bigg].\end{align*}$

利用 Hölder 不等式以及$E_{\lambda}\hookrightarrow L^{2q}(\mathbb{R}^N)$, 其中$q\in[\frac{N}{N-2}]$, 可得

$\begin{align*}\int_{B(R)}u_{n}^{2}{\rm d}x&\leq\bigg(\int_{\mathbb{R}^N}|u_{n}|^{2q}{\rm d}x\bigg)^{\frac{1}{q}}\bigg(\int_{B(R)}{\rm d}x\bigg)^{\frac{1}{q^{\prime}}}\leq C\|u_{n}\|_{\lambda}^{2}\mu(B(R))^{\frac{1}{q^{\prime}}}\\&\leq C\frac{2p_{\varepsilon}}{p_{\varepsilon}-1}C_{1}\mu(B(R))^{\frac{1}{q^{\prime}}},\end{align*}$

其中$\frac{1}{q}+\frac{1}{q^{\prime}}=1$. 令$\theta=\frac{Np_{\varepsilon}-2N+\mu}{2p_{\varepsilon}}$, 再根据 Gagliardo-Nirenberg 不等式有

$\begin{align*}\bigg(\int_{B_{R}^{c}(0)}|u_{n}|^{\frac{2Np_{\varepsilon}}{2N-\mu}}\bigg)^{\frac{2N-\mu}{2Np_{\varepsilon}}}&\leq C|\nabla u_{n}|^{\theta}_{2,B_{R}^{c}}|u_{n}|^{1-\theta}_{2,B_{R}^{c}}\leq C\|u_{n}\|_{\lambda}^{\theta}\bigg[\int_{A(R)}u_{n}^{2}{\rm d}x+\int_{B(R)}u_{n}^{2}{\rm d}x\bigg]^{\frac{1-\theta}{2}}\\&\leq C(\frac{2C_{1}p_{\varepsilon}}{p_{\varepsilon}-1})^{\theta}\bigg[\frac{1}{\lambda M_{0}+1}(\frac{2C_{1}p_{\varepsilon}}{p_{\varepsilon}-1}+o_{n}(1))+C\frac{2C_{1}p_{\varepsilon}}{p_{\varepsilon}-1}\mu(B(R))^{\frac{1}{q^{\prime}}}\bigg]^{\frac{1-\theta}{2}}.\end{align*}$

当$R$与$\lambda$很大时, 注意到上式中$\mu(B(R))$较小, 因此有$\underset{n\rightarrow\infty}{\lim\sup}{|u_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu},B_{R}^{c}(0)}\leq\sigma_{1}.$证毕.

下面证明泛函$I_{\lambda,p_\varepsilon}$满足$(PS)_{c}$条件.

引理 3.6 令$C_{1}>0$, 存在$\Lambda_{1}=\Lambda(C_{1})>0$, 当$\lambda\geq\Lambda_{1}$时, 对所有的$c\in[C_{1}]$,$I_{\lambda,p_{\varepsilon}}$满足$(PS)_{c}$条件.

令$\{u_{n}\}$是$(PS)_{c}$序列, 由引理3.2可得$\{u_{n}\}$有界, 则存在子列仍定义为$\{u_{n}\}$有

$\begin{align*}&u_{n}\rightharpoonup u\quad\quad E_{\lambda},\\&u_{n}\rightarrow u\quad L_{loc}^{q}(\mathbb{R}^N),\quad 1\leq q<2^{\ast},\\&u_{n}\rightarrow u\quad\quad a.e. \quad\mathbb{R}^N,\end{align*}$

从而$I_{\lambda,p_{\varepsilon}}^{\prime}(u)=0$和$I_{\lambda,p_{\varepsilon}}(u)\geq0$. 令$v_{n}=u_{n}-u$, 由引理 3.2 可知$\{v_{n}\}$是$(PS)_{d}$序列, 其中$d=c-I_{\lambda,p_{\varepsilon}}(u)$, 故

$\begin{equation*}0\leq d=c-I_{\lambda,p_{\varepsilon}}(u)\leq c\leq C_{1}.\end{equation*}$

断言$d=0$. 否则, 若$d>0$, 通过引理 3.3 有$d\geq c_{0}$, 再由引理3.4, 可得

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\inf}{|v_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu}}\geq\delta_{0}c_{0}>0.\end{equation*}$

结合引理 3.5,$\sigma_{1}=\frac{\delta_{0}c_{0}}{2}>0$, 因此, 存在$\Lambda_{1}$,$R>0$, 当$\lambda\geq\Lambda_{1}$,有

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\sup}{|v_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu},B_{R}^{c}(0)}\leq\frac{\delta_{0}c_{0}}{2},\end{equation*}$

所以与

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\inf}{|v_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu},B_{R}(0)}\geq\frac{\delta_{0}c_{0}}{2}>0\end{equation*}$

矛盾, 因为在$E_{\lambda}$中$v_{n}\rightharpoonup0$, 嵌入$E_{\lambda}\hookrightarrow L^{\frac{2Np_{\varepsilon}}{2N-\mu}}(B_{R}(0))$是紧的, 从而

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\inf}{|v_{n}|}^{2p_{\varepsilon}} _{\frac{2Np_{\varepsilon}}{2N-\mu},B_{R}(0)}=0.\end{equation*}$

由此可得$d=0$,$\{v_{n}\}$是$(PS)_{0}$序列, 即在$E_{\lambda}$中$v_{n}\rightarrow0$. 当$\lambda$足够大时, 对所有的$c\in[C_{1}]$,$I_{\lambda,p_{\varepsilon}}$满足$(PS)_{c}$条件. 证毕.

接下来证明方程(1.1)存在非平凡解关于山路值

$ m_{\lambda,p_\varepsilon}=\inf\limits_{\gamma\in\Gamma} \max\limits_{t\in[0,1]}I_{\lambda,p_\varepsilon}(\gamma(t)),$

其中

$\Gamma:=\{\gamma\in C([0,1], E_\lambda): \gamma(0) =0, I_{\lambda,p_\varepsilon}(\gamma(1)) \leq 0, \gamma(1)\not = 0\}.$

并定义

$ m^s_{\lambda,p_\varepsilon}=\inf_{u\in E_\lambda\setminus \{0\}}\sup_{t>0}I_{\lambda,p_\varepsilon}(tu).$

类似在文献[29,定理 4.2] 证明, 可以证明

$\begin{equation}c_{\lambda,p_\varepsilon} = m_{\lambda,p_\varepsilon}=m^s_{\lambda,p_\varepsilon}.\end{equation}$

接下来给出定理 1.1 前半部分的证明, 后半部分将在定理 1.2 的证明中给出.

定理 1.1 的证明 易证$I_{\lambda, p_{\varepsilon}}$满足山路定理条件, 因此存在一串$(PS)_{m_{\lambda,p_\epsilon}}$列, 记为$\{u_{n}\}$; 由引理3.6可知,$\{u_{n}\}$的子列存在弱收敛极限, 记为$u_{\lambda,p_\varepsilon}$; 由(3.5)式, 那么$u_{\lambda,p_{\varepsilon}}$是问题(1.1)的基态解. 又因为$I_{\lambda,p_{\varepsilon}}(u_{\lambda,p_{\varepsilon}})=I_{\lambda,p_{\varepsilon}}(|u_{\lambda,p_{\varepsilon}}|)$, 可以假设$u_{\lambda,p_{\varepsilon}}\geq0$. 再利用强极值原理, 则$u_{\lambda,p_{\varepsilon}}$是正解.

4 主要结果的证明

方程(1.1)的极限方程(1.9)的 Nehari 流形定义为

$\begin{equation*}\mathcal{M}(p_{\varepsilon},\Omega)=\bigg\{u\in{{H}}^{1}_{0}{(\Omega)}\backslash\{0\}:\int_{\Omega}(|\nabla u|^{2}+u^{2}){\rm d}x=\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y\bigg\}.\end{equation*}$

$\begin{equation*}c(p_{\varepsilon},\Omega)=\underset{u\in\mathcal{M}(p_{\varepsilon},\Omega)}{\inf}I_{p_{\varepsilon},\Omega}(u),\end{equation*}$

其中

$\begin{equation*}I_{p_{\varepsilon},\Omega}(u)=\frac{1}{2}\int_{\Omega}(|\nabla u|^{2}+u^{2}){\rm d}x-\frac{1}{2p_{\varepsilon}}\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y.\end{equation*}$

再令

$\begin{equation*}c(p_{\varepsilon},r):=c(p_{\varepsilon},B_{r}),\end{equation*}$

其中$B_{r}=\{x\in\mathbb{R}^N:|x|<r\}$.

引理 4.1[14] 当$\varepsilon\rightarrow0$时,$c(p_{\varepsilon},\Omega)\rightarrow(\frac{1}{2}-\frac{1}{2\cdot2^{\ast}_{\mu}})S_{H,L}^{\frac{2^{\ast}_{\mu}}{2^{\ast}_{\mu}-1}}$, 其中

$\begin{equation*}S_{H,L}:=\underset{u\in D^{1,2}(\mathbb{R}^N)\backslash\{0\}}{\inf}\frac{\int_{\mathbb{R}^N}|\nabla u|^{2}{\rm d}x}{(\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{|u(x)|^{2^{\ast}_{\mu}}|u(y)|^{2^{\ast}_{\mu}}}{|x-y|^{\mu}}{\rm d}x{\rm d}y)^{\frac{N-2}{2N-\mu}}}\end{equation*}$

是最佳常数.

若序列$\{u_{n}\}\subset{H}^{1}{(\mathbb{R}^N)}$满足

$u_{n}\in E_{\lambda_{n}},\quad\quad\lambda_{n}\rightarrow\infty,$
$I_{\lambda_{n},p_{\varepsilon}}(u_{n})\rightarrow c,\quad\quad c\in \mathbb{R},$
$\|I_{\lambda_{n},p_{\varepsilon}}^{\prime}(u_{n})\|_{(E_{\lambda_{n}})^{\prime}}\rightarrow 0,$

则称$\{u_{n}\}$是$(PS)_{c,\infty}$序列.

命题 4.1 假设$\{u_{n}\}$是$(PS)_{c,\infty}$序列,$c\in(0,(\frac{1}{2}-\frac{1}{2\cdot2^{\ast}_{\mu}})S_{H,L}^{\frac{2^{\ast}_{\mu}}{2^{\ast}_{\mu}-1}})$, 那么存在子列仍记为$\{u_{n}\}$以及$u$, 满足

$\begin{equation*}u_{n}\rightharpoonup u\quad\quad {H}^{1}{(\mathbb{R}^N)},\end{equation*}$

并且

(i) 在$\mathbb{R}^N\backslash\Omega$中$u\equiv0$,$u$是方程

$\begin{equation*} \begin{cases} -{\Delta}{u}+u=\Big(\int_{\Omega}\frac{|u(y)|^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}y\Big)|u|^{p_{\varepsilon}-2}u, \quad & x\in\Omega,\\u=0,\quad \quad &x\in\partial\Omega \end{cases}\end{equation*}$

的解;

(ii) 在${H}^{1}{(\mathbb{R}^N)}$中$u_{n}\rightarrow u$;

(iii) 当$n\rightarrow\infty$时,$\lambda_{n}\int_{\mathbb{R}^N}V(x)|u_{n}|^{2}{\rm d}x\rightarrow0$;

(iv) 当$n\rightarrow\infty$时,

$\begin{equation*}I_{\lambda_{n},p_{\varepsilon}}(u_{n})\rightarrow \frac{1}{2}\int_{\Omega}(|\nabla u|^{2}+|u|^{2}){\rm d}x-\frac{1}{2p_{\varepsilon}}\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y.\end{equation*}$

先证(i), 由引理 3.3,

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\sup}{\rVert u_{n} \rVert}^{2} _{\lambda_n}\leq\frac{2p_{\varepsilon}c}{p_{\varepsilon}-1},\end{equation*}$

可知$\{\|u_{n}\|_{\lambda_{n}}\}$在$\mathbb{R}$上有界. 对任意的$n\in\mathbb{N}$, 有$\|u_{n}\|_{\lambda_{n}}\geq\|u_{n}\|$, 故$\{u_{n}\}$在${H}^{1}{(\mathbb{R}^N)}$中有界, 那么存在子列仍记为$\{u_{n}\}$, 有$u\in {H}^{1}{(\mathbb{R}^N)}$满足

$\begin{align*}&u_{n}\rightharpoonup u\quad {H}^{1}{(\mathbb{R}^N)},\\&u_{n}\rightarrow u\quad\quad a.e. \quad\mathbb{R}^N\end{align*}$

以及

$\begin{align*}&u_{n}^+\rightharpoonup u^+\quad {H}^{1}{(\mathbb{R}^N)},\\&u_{n}^+\rightarrow u^+\quad\quad a.e. \quad\mathbb{R}^N.\end{align*}$

对每个$m\in\mathbb{N}$, 定义$C_{m}=\{x\in\mathbb{R}^N:V(x)\geq\frac{1}{m}\}$, 易知

$\begin{equation*}\mathbb{R}^N\backslash\bar{\Omega}=\bigcup_{m=1}^{+\infty}{C_{m}}.\end{equation*}$

注意到

$\begin{equation*}\int_{C_{m}}|u_{n}|^{2}{\rm d}x=\int_{C_{m}}\frac{\lambda_{n}V(x)}{\lambda_{n}V(x)}|u_{n}|^{2}{\rm d}x\leq\frac{m}{\lambda_{n}}\|u_{n}\|^{2}_{\lambda_{n}}\leq\frac{c}{\lambda_{n}}.\end{equation*}$

根据 Fatou 引理

$\begin{equation*}\int_{C_{m}}|u|^{2}{\rm d}x\leq\underset{n\rightarrow\infty}{\lim\inf}\int_{C_{m}}|u_{n}|^{2}{\rm d}x\leq\underset{n\rightarrow\infty}{\lim\inf}\frac{c}{\lambda_{n}}=0,\end{equation*}$

从而$u$在$C_{m}$中几乎处处为$0$, 即$u$在$\mathbb{R}^N\backslash\bar{\Omega}$中几乎处处为$0$. 令$\varphi\in C_{0}^{\infty}(\Omega)$, 有

$\begin{equation*}\langle{I_{\lambda_n,p_{\varepsilon}}^{\prime}(u_{n}),\varphi}\rangle=\int_{\Omega}(\nabla u_{n}\nabla\varphi+u_{n}\varphi) {\rm d}x-\int_{\Omega}\int_{\Omega}\frac{(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}-1}\varphi}{|x-y|^\mu}{\rm d}x{\rm d}y,\end{equation*}$

由于$\{u_{n}\}$是$(PS)_{c,\infty}$序列, 所以

$\begin{equation*}\langle{I_{\lambda_n,p_{\varepsilon}}^{\prime}(u_{n}),\varphi}\rangle\rightarrow0.\end{equation*}$

因为在${H}^{1}{(\mathbb{R}^N)}$中$u_{n}\rightharpoonup u$, 故

$\begin{equation*}\int_{\mathbb{R}^N}(\nabla u_{n}\nabla\varphi+u_{n}\varphi) {\rm d}x\rightarrow\int_{\Omega}(\nabla u\nabla\varphi+u\varphi) {\rm d}x,\end{equation*}$

以及

$\begin{equation*}\int_{\Omega}\int_{\Omega}\frac{(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}-1}\varphi}{|x-y|^\mu}{\rm d}x{\rm d}y\rightarrow\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}-1}\varphi}{|x-y|^\mu}{\rm d}x{\rm d}y.\end{equation*}$

因此对$\forall\varphi\in C_{0}^{\infty}(\Omega)$,

$\begin{equation*}\int_{\Omega}(\nabla u\nabla\varphi+u\varphi) {\rm d}x=\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}-1}\varphi}{|x-y|^\mu}{\rm d}x{\rm d}y,\end{equation*}$

那么$u$是问题

$\begin{equation*} \begin{cases} -{\Delta}{u}+u=\Big(\int_{\Omega}\frac{|u(y)|^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}y\Big)|u|^{p_{\varepsilon}-2}u,\quad & x\in\Omega,\\u=0,\quad \quad \ \ \,&x\in\partial\Omega \end{cases}\end{equation*}$

的弱解.

其次证 (ii), 由 (i) 可得, 在$\mathbb{R}^N\backslash\Omega$中$u\equiv0$, 并且$V(x)$在$\Omega$中等于$0$, 所以

$\begin{equation*}\|u\|_{\lambda_{n}}^{2}=\int_{\Omega}(|\nabla u|^{2}+u^{2}){\rm d}x.\end{equation*}$

由于$u$的支集在$\Omega$中,$V(x)$在$\Omega$中等于$0$, 所以即使$\lambda_{n}$很大,$\int_{\mathbb{R}^N}\lambda_{n}V(x)u_{n}u{\rm d}x=0$, 那么

$\begin{equation*}\int_{\mathbb{R}^N}(\nabla u_{n}\nabla u+(\lambda_{n}V(x)+1)u_{n}u){\rm d}x=\int_{\Omega}(|\nabla u|^{2}+u^{2}){\rm d}x+o_{n}(1),\end{equation*}$

因此

$\begin{align*}\|u_{n}-u\|_{\lambda_{n}}^{2}&=\|u_{n}\|_{\lambda_{n}}^{2}+\|u\|_{\lambda_{n}}^{2}-2\int_{\mathbb{R}^N}(\nabla u_{n}\nabla u+(\lambda_{n}V(x)+1)u_{n}u){\rm d}x\\&=\|u_{n}\|_{\lambda_{n}}^{2}-\int_{\Omega}(|\nabla u|^{2}+u^{2}){\rm d}x+o_{n}(1).\end{align*}$

结合$\{|u_{n}|_{\lambda_{n}}\}$有界与$\|I_{\lambda_{n},p_{\varepsilon}}^{\prime}(u_{n})\|_{(E_{\lambda_{n}})^{\prime}}\rightarrow 0$可知$\langle{I_{\lambda_n,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle\rightarrow0,$从而

$\begin{align*}\|u_{n}\|_{\lambda_{n}}^{2}&=\langle{I_{\lambda_n,p_{\varepsilon}}^{\prime}(u_{n}),u_{n}}\rangle+\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y\\&=o_{n}(1)+\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y.\end{align*}$

再由$\langle{I_{\lambda_n,p_{\varepsilon}}^{\prime}(u_{n}),u}\rangle\rightarrow0$可得

$\begin{equation*}\int_{\Omega}(|\nabla u|^{2}+u^{2}){\rm d}x=\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y,\end{equation*}$

故有

$\begin{align*}\|u_{n}-u\|_{\lambda_{n}}^{2}&=\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y-\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y+o_{n}(1)\\&=\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{((u_{n}(x)-u(x))^+)^{p_{\varepsilon}}((u_{n}(y)-u(y))^+)^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y+o_{n}(1),\end{align*}$

因此

$\begin{equation*}\|v_{n}\|_{\lambda_{n}}^{2}=\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(v_{n}^+(x))^{p_{\varepsilon}}(v_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y+o_{n}(1).\end{equation*}$

下证:$u_n\rightarrow u$在$L^p(\mathbb{R}^N)$对于$2<p<2^*$. 否则由文献[29,引理 1.21], 存在$\delta>0,$$\rho>0$,$x_n\in\mathbb{R}^N$且$|x_n|\rightarrow\infty$, 有

$\liminf\limits_{n\rightarrow\infty}\int_{B_\rho(x_n)}|u_n-u|^2\,{\rm d}x\geq\delta>0,$

那么

$\begin{align*}& I_{\lambda_n,p_\varepsilon}(u_n)-\frac{1}{2p_\varepsilon}\langle I'_{\lambda_n,p_\varepsilon}(u_n),u_n\rangle\\&\geq C\lambda_n\int_{B_\rho(x_n)\cap\{x:V(x)\geq M_0\}}V(x)u_n^2\,{\rm d}x\\&=C\lambda_n\int_{B_\rho(x_n)\cap\{x:V(x)\geq M_0\}}V(x)|u_n-u|^2\,{\rm d}x\\&\geq C\lambda_n\bigg(\int_{B_\rho(x_n)}V(x)|u_n-u|^2\,{\rm d}x-M_0\int_{B_\rho(x_n)\cap\{x:V(x)\leq M_0\}}|u_n-u|^2\,{\rm d}x\bigg)\\&\geq C\lambda_n\bigg(M_0\int_{B_\rho(x_n)}|u_n-u|^2\,{\rm d}x-o(1)\bigg),\end{align*}$

最后一个不等式利用了 Hölder 不等式及当$n\rightarrow\infty$时,$\mu(B_\rho(x_n)\cap\{x:V(x)\leq M_0\})\rightarrow0,$因此有$u_n\rightarrow u$在$L^p(\mathbb{R}^N)$对于$2<p<2^*$. 由 Hardy-Littlewood-Sobolev 不等式, 有$\|v_{n}\|_{\lambda_{n}}^{2}\rightarrow0$, 即在${H}^{1}{(\mathbb{R}^N)}$中,$u_{n}\rightarrow u$.

再证 (iii), 由

$\begin{equation*}\lambda_{n}\int_{\mathbb{R}^N}V(x)|u_{n}|^{2}{\rm d}x=\int_{\mathbb{R}^N}\lambda_{n}V(x)|u_{n}-u|^{2}{\rm d}x\leq\|u_{n}-u\|_{\lambda_{n}}^{2},\end{equation*}$

所以$\lambda_{n}\int_{\mathbb{R}^N}V(x)|u_{n}|^{2}{\rm d}x\rightarrow0$.

最后证 (iv): 下面改写泛函$I_{\lambda_{n},p_{\varepsilon}}$,

$\begin{align*}I_{\lambda_{n},p_{\varepsilon}}(u_{n})&=\frac{1}{2}\int_{\Omega}(|\nabla u_{n}|^{2}+(\lambda_{n}V(x)+1)|u_{n}|^{2}){\rm d}x\!+\!\frac{1}{2}\int_{\mathbb{R}^N\backslash\Omega}(|\nabla u_{n}|^{2}\!+\!(\lambda_{n}V(x)+1)|u_{n}|^{2}){\rm d}x\\& \ -\frac{1}{2p_{\varepsilon}}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y.\end{align*}$

由结论 (ii) 在${H}^{1}{(\mathbb{R}^N)}$中,$u_{n}\rightarrow u$, 以及结论 (iii) 可得

$\begin{equation*}\frac{1}{2}\int_{\Omega}(|\nabla u_{n}|^{2}+(\lambda_{n}V(x)+1)|u_{n}|^{2}){\rm d}x\rightarrow\frac{1}{2}\int_{\Omega}(|\nabla u|^{2}+|u|^{2}){\rm d}x,\end{equation*}$
$\begin{equation*}\frac{1}{2}\int_{\mathbb{R}^N\backslash\Omega}(|\nabla u_{n}|^{2}+(\lambda_{n}V(x)+1)|u_{n}|^{2}){\rm d}x\rightarrow0,\end{equation*}$
$\begin{equation*}\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y\rightarrow\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y,\end{equation*}$

因此

$\begin{equation*}I_{\lambda_{n},p_{\varepsilon}}(u_{n})\rightarrow \frac{1}{2}\int_{\Omega}(|\nabla u|^{2}+|u|^{2}){\rm d}x-\frac{1}{2p_{\varepsilon}}\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y.\end{equation*}$

证毕.

接下来证明关于$c(p_{\varepsilon},\Omega)$与$c_{\lambda,p_{\varepsilon}}$之间的估计.

引理 4.2 若$\lambda_{n}\rightarrow+\infty$, 则$c_{\lambda_{n},p_{\varepsilon}}\rightarrow c(p_{\varepsilon},\Omega)$.

不等式$c_{\lambda_{n},p_{\varepsilon}}<c(p_{\varepsilon},\Omega)$是严格的. 如果$c_{\lambda_{n},p_{\varepsilon}}=c(p_{\varepsilon},\Omega)$, 可找到问题(1.1)的非负解, 并且解在$\Omega$外部为$0$, 由极大值原理可知这是不可能的. 假设当$\lambda_{n}\rightarrow\infty$时, 存在$K$, 使得

$\begin{equation*}\underset{n\rightarrow\infty}{\lim}c_{\lambda_{n},p_{\varepsilon}}=K< c(p_{\varepsilon},\Omega),\end{equation*}$

根据引理 3.3,$K>0$. 由引理 3.6 可知, 存在$\Lambda_{1}(p_{\varepsilon})>0$, 使得对所有的$\lambda_{n}\geq\Lambda_{1}(p_{\varepsilon})$,$I_{\lambda_{n},p_{\varepsilon}}$满足山路定理条件, 因此$c_{\lambda_{n},p_{\varepsilon}}$可达, 记达到函数为$u_{n}$. 由$\{u_{n}\}\subset {H}^{1}{(\mathbb{R}^N)}$有界以及命题 4.1, 存在$\{u_{n}\}$的子列仍记为$\{u_{n}\}$和$u\in {H}^{1}_{0}{(\Omega)}\backslash\{0\}$使得

$\begin{align*}&\|u_{n}-u\|_{\lambda_{n}}^{2}\rightarrow0,\\&u_{n}\rightarrow u\quad {H}^{1}{(\mathbb{R}^N)},\\&u_{n}\rightarrow u\quad L^{q}{(\mathbb{R}^N)},\ \ q\in[2^{\ast}],\\&I_{p_{\varepsilon},\Omega}^{\prime}(u)=0.\end{align*}$

由于$u\neq0$,$\langle{I_{p_{\varepsilon},\Omega}^{\prime}(u),u}\rangle=0$, 可得$u\in\mathcal{M}(p_{\varepsilon},\Omega)$, 因此$I_{p_{\varepsilon},\Omega}(u)\geq c(p_{\varepsilon},\Omega)$, 所以$I_{\lambda_{n},p_{\varepsilon}}(u_{n})+o_{n}(1)\geq c(p_{\varepsilon},\Omega)$; 进一步可得当$\lambda_{n}\rightarrow+\infty$时,$c_{\lambda_{n},p_{\varepsilon}}\rightarrow c(p_{\varepsilon},\Omega)$. 证毕.

定理 1.2 的证明 假设$\{u_{n}\}_{n\in\mathbb{N}}$是问题(1.1)在$\lambda_{n}\rightarrow+\infty$时的解序列, 使得

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\sup}\,I_{\lambda_{n},p_{\varepsilon}}(u_{n})<\infty,\end{equation*}$

通过命题 4.1, 可知在${H}^{1}{(\mathbb{R}^N)}$中,$u_{n}\rightarrow u$,$u$是问题(1.9)在${H}^{1}_{0}{(\Omega)}$中的解. 类似的, 当序列$u_{\lambda_n}$为关于$c_{\lambda_n,p_\varepsilon}$的基态解序列时, 由命题 4.1 和引理 4.2, 那么$u_{\lambda_n}$在$H^1(\mathbb{R}^N)$中强收敛到$u$,并且$u$是方程(1.9)的基态解. 证毕.

现在假设$\Omega$是$V^{-1}(0)$的内部. 由 (A$_{1})$可知存在$r>0$, 使得$\Omega^{-}_{r}\hookrightarrow\Omega\hookrightarrow\Omega^{+}_{r}$是同伦等价的, 这里

$\begin{equation*}\Omega^{+}_{r}:=\{x\in\mathbb{R}^N:\text{dist}(x,\Omega)\leq r\},\end{equation*}$
$\begin{equation*}\Omega^{-}_{r}:=\{x\in\Omega:\text{dist}(x,\partial\Omega)\geq r\},\end{equation*}$

其中$r$固定.

设$\eta\in C_{c}^{\infty}(\mathbb{R}^N)$, 对于$x\in\Omega$满足$\eta(x)=x$. 定义非零函数$u\in {H}_{0}^{1}(\Omega)$的重心为

$\begin{equation*}\beta(u)=\frac{\int_{\mathbb{R}^N}\eta(x)|\nabla u|^{2}{\rm d}x}{\int_{\mathbb{R}^N}|\nabla u|^{2}{\rm d}x}.\end{equation*}$

那么, 在文献[14]中已经得到下列结果.

引理4.3[14] 对任意的$r^{\prime}<r$, 存在$\varepsilon(r^{\prime})$使得对$\forall\varepsilon\in(0,\varepsilon(r^{\prime})]$, 当$u\in\mathcal{M}(p_{\varepsilon},\Omega)$并且$I_{p_{\varepsilon},\Omega}(u)\leq c(p_{\varepsilon},r^{\prime})$时,$\beta(u)\in\Omega^{+}_{r^{\prime}}$.

现在, 类似文献[6]中重心函数的定义, 选取$R_0>0$满足$\bar{\Omega}\subset B_{R_0}$并且令

$\begin{equation*}\xi(t)=\begin{cases}1,\quad & 0\leq t\leq R_0;\\R_0/t,\quad &R_0\leq t.\end{cases}\end{equation*}$

然后定义$u\in\mathcal{M}_{\lambda,p_{\varepsilon}}$的重心函数为

$\begin{equation*}\beta_0(u)=\frac{\int_{\mathbb{R}^N}\xi(|x|)x|\nabla u|^{2}{\rm d}x}{\int_{\mathbb{R}^N}|\nabla u|^{2}{\rm d}x}.\end{equation*}$

那么有下列结果.

引理 4.4 存在$\varepsilon_{2}>0$, 则对任意的$\varepsilon\in(0,\varepsilon_{2}]$, 存在$\Lambda_{2}(\varepsilon)>0$, 对所有的$\lambda\geq\Lambda_{2}(\varepsilon)$以及任意的$u\in\mathcal{M}_{\lambda,p_{\varepsilon}}$, 满足$I_{\lambda,p_{\varepsilon}}(u)\leq c(p_{\varepsilon},r)$, 使得$\beta_0(u)\in\Omega^{+}_{r}$.

利用反证法. 令引理4.3中的$\varepsilon(\frac{r}{2})=\varepsilon_{2}$, 存在序列$\lambda_{n}\rightarrow+\infty$和$u_{n}\in\mathcal{M}_{\lambda_{n},p_{\varepsilon}}$, 使得$I_{\lambda_{n},p_{\varepsilon}}(u_{n})\leq c(p_{\varepsilon},r)$并且$\beta_0(u_{n})\notin\Omega^{+}_{r}$. 已知

$\begin{equation*}\underset{n\rightarrow\infty}{\lim\inf}\int_{\mathbb{R}^N}|\nabla u_{n}|^{2}{\rm d}x\geq\int_{\mathbb{R}^N}|\nabla u|^{2}{\rm d}x,\end{equation*}$

并且, 当$\lambda_{n}\rightarrow+\infty$时, 由命题 4.1 的结论, 在$\mathbb{R}^N\backslash\Omega$中$u\equiv0$, 因此

$\begin{align*}\int_{\Omega}(|\nabla u|^{2}+u^{2}){\rm d}x&=\int_{\mathbb{R}^N}(|\nabla u|^{2}+u^{2}){\rm d}x\\&\leq\underset{n\rightarrow\infty}{\lim\inf}\int_{\mathbb{R}^N}(|\nabla u_{n}|^{2}+u_{n}^{2}){\rm d}x\\&=\underset{n\rightarrow\infty}{\lim\inf}\bigg(\int_{\mathbb{R}^N}-\lambda_{n}V(x)u_{n}^{2}{\rm d}x+\int_{\mathbb{R}^N}\int_{\mathbb{R}^N}\frac{(u_{n}^+(x))^{p_{\varepsilon}}(u_{n}^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y\bigg)\\&=\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y.\end{align*}$

因此, 可得

$\begin{equation*}\alpha=\bigg(\frac{\int_{\Omega}(|\nabla u|^{2}+u^{2}){\rm d}x}{\int_{\Omega}\int_{\Omega}\frac{(u^+(x))^{p_{\varepsilon}}(u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y}\bigg)^{\frac{1}{2p_{\varepsilon}-2}}\leq1,\end{equation*}$

所以存在$\alpha\in(0,1]$使得

$\begin{equation*}\int_{\Omega}(|\nabla (\alpha u)|^{2}+(\alpha u)^{2}){\rm d}x=\int_{\Omega}\int_{\Omega}\frac{(\alpha u^+(x))^{p_{\varepsilon}}(\alpha u^+(y))^{p_{\varepsilon}}}{|x-y|^\mu}{\rm d}x{\rm d}y,\end{equation*}$

即$\alpha u\in\mathcal{M}(p_{\varepsilon},\Omega)$, 那么有

$\begin{align*}I_{p_{\varepsilon},\Omega}(\alpha u)&=(\frac{1}{2}-\frac{1}{2p_{\varepsilon}})\int_{\Omega}(|\nabla (\alpha u)|^{2}+(\alpha u)^{2}){\rm d}x\\& \leq(\frac{1}{2}-\frac{1}{2p_{\varepsilon}})\int_{\mathbb{R}^N}(|\nabla u|^{2}+u^{2}){\rm d}x\\& \leq(\frac{1}{2}-\frac{1}{2p_{\varepsilon}})\underset{n\rightarrow\infty}{\lim\inf}\int_{\mathbb{R}^N}(|\nabla u_{n}|^{2}+(\lambda_{n}V(x)+1)u_{n}^{2}){\rm d}x\\& =\underset{n\rightarrow\infty}{\lim\inf}\,I_{\lambda_{n},p_{\varepsilon}}(u_{n})\\&\leq c(p_{\varepsilon},r).\end{align*}$

根据引理 4.3,$\beta(\alpha u)\in\Omega^{+}_{\frac{r}{2}}$, 并且因为在$H^1(\mathbb{R}^N)$中$u_n\rightarrow u$, 所以

$\begin{equation*}\lim\limits_{n\rightarrow\infty}\beta_0(u_n)=\beta(u)=\beta(\alpha u)\in\Omega^{+}_{\frac{r}{2}},\end{equation*}$

这与$\beta_0(u_{n})\notin\Omega^{+}_{r}$矛盾. 证毕.

由定理 1.1 和极值原理可知存在问题(1.9)的基态解$u_{r}>0$, 再根据文献[23]知$u_{r}$是球对称的, 即

$\begin{equation*}I_{p_{\varepsilon},\Omega}(u_{r})=c(p_{\varepsilon},r),\quad I_{p_{\varepsilon},\Omega}^{\prime}(u_{r})=0.\end{equation*}$

对给定的$u_{r}$, 定义$\psi_{r}:\Omega^{-}_{r}\rightarrow {{H}}^{1}_{0}{(\Omega)}$,

$\begin{equation}\psi_{r}(y)(x)=\begin{cases}u_{r}(x-y), \quad&\text{若 }x\in B_{r}(y),\\0, \quad&\text{若 }x\notin B_{r}(y).\end{cases}\end{equation}$

因为$u_{r}$是球对称的, 对任意$y\in\Omega^{-}_{r}$, 可以验证$\beta_0(\psi_{r}(y))=y$.

下面说明$\Omega$与水平集的畴数之间的关系.

引理 4.5 存在$\bar{\varepsilon}>0$, 对任意的$\varepsilon\in(0,\bar{\varepsilon})$, 相应的存在$\Lambda_{2}(\varepsilon)>0$, 对所有的$\lambda\geq\Lambda_{2}(\varepsilon)$, 有

$\begin{equation*}cat_{I_{\lambda,p_{\varepsilon}}^{c(p_{\varepsilon},r)}}(I_{\lambda,p_{\varepsilon}}^{c(p_{\varepsilon},r)})\geq cat_{\Omega}(\Omega),\end{equation*}$

其中

$\begin{equation*}I_{\lambda,p_{\varepsilon}}^{c(p_{\varepsilon},r)}:=\{u\in\mathcal{M}_{\lambda,p_{\varepsilon}}:I_{\lambda,p_{\varepsilon}}(u)\leq c(p_{\varepsilon},r)\}.\end{equation*}$

取$\bar{\varepsilon}=\varepsilon(r^{\prime})$, 设$I_{\lambda,p_{\varepsilon}}^{c(p_{\varepsilon},r)}=P_{1}\cup P_{2}\cdots\cup P_{n}$, 其中$P_{j}$,$j=1,2,\cdots,n$, 是$I_{\lambda,p_{\varepsilon}}^{c(p_{\varepsilon},r)}$中的可压缩集, 即存在$h_{j}\in \mathcal{C}([0,1]\times P_{j},I_{\lambda,p_{\varepsilon}}^{c(p_{\varepsilon},r)}) $, 使得对$\forall u,v\in P_{j}$, 有

$\begin{equation*}h_{j}(0,u)=u,\quad\quad\quad h_{j}(1,u)=h_{j}(1,v).\end{equation*}$

由 (4.1) 式定义的$\psi_{r}$, 注意到$\psi_{r}(y)(x)\in\mathcal{M}(p_\varepsilon,B_r(0))$, 进一步将其延拓至$\mathbb{R}^N$, 那么有$\psi_{r}(y)(x)\in\mathcal{M}_{\lambda,p_\varepsilon}$. 接下来令$B_{j}=\psi_{r}^{-1}(P_{j})$,$j=1,2,\cdots,n$, 则$B_{j}$是闭集且$B_{j}\subset\Omega_{r}^{-}$. 对$\forall x\in\Omega_{r}^{-}$, 有$\psi_{r}(x)\in I_{p_{\varepsilon},\Omega}^{c(p_{\varepsilon},r)}\subset\bigcup\limits_{j=1}^n{P_{j}}$, 所以存在$j_{0}$, 使得$\psi_{r}(x)\in P_{j_{0}}$, 即$x\in\psi_{r}^{-1}(P_{j_{0}})=B_{j_{0}}$. 因此

$\begin{equation*}\Omega_{r}^{-}=\bigcup\limits_{j=1}^n{B_{j}}.\end{equation*}$

再证$B_{j}$在$\Omega_{r}^{+}$上可收缩. 令$g_{j}(t,x)\in \mathcal{C}([0,1] \times B_{j},\Omega_{r}^{+}) $,

$\begin{equation*}g_{j}(t,x)=\beta_0(h_{j}(t,\psi_{r}(x))),\quad 1\leq j\leq n,\end{equation*}$

有$g_{j}(0,x)=\beta_0(h_{j}(0,\psi_{r}(x)))=\beta_0(\psi_{r}(x))$,$\forall x,y\in B_j$,

$\begin{equation*}g_{j}(1,x)=\beta_0(h_{j}(1,\psi_{r}(x)))=\beta_0(h_{j}(1,\psi_{r}(y)))=g_{j}(1,y). 且\end{equation*}$

所以$B_{j}$在$\Omega_{r}^{+}$上可收缩, 即可得$cat_{\Omega}(\Omega)=cat_{\Omega_{r}^{+}}(\Omega_{r}^{-})\leq cat_{I_{\lambda,p_{\varepsilon}}^{c(p_{\varepsilon},r)}}(I_{\lambda,p_{\varepsilon}}^{c(p_{\varepsilon},r)})$. 证毕.

最后有如下关于方程(1.1)的多解结果.

定理 1.3 的证明 令$\varepsilon_{0}=\min\{\varepsilon_{1},\varepsilon_{2}\}$, 对任意的$\varepsilon\in(0,\varepsilon_{0}]$,$\Lambda(\varepsilon)=\max\{\Lambda_{1}{(\varepsilon)},\Lambda_{2}(\varepsilon)\}$, 因为$c(p_{\varepsilon},r)<(\frac{1}{2}-\frac{1}{22^{\ast}_{\mu}})S_{H,L}^{\frac{2^{\ast}_{\mu}}{2^{\ast}_{\mu}-1}}$,由引理 3.6, 对所有的$c\leq c(p_{\varepsilon},r)$,$I_{\lambda,p_{\varepsilon}}(u)$满足$(PS)_{c}$条件, 再根据文献[29,定理 5.20] 可知问题(1.1)至少有$cat_\Omega(\Omega)$个非平凡弱解, 结合$I_{\lambda,p_\varepsilon}$定义及强极值原理和标准的椭圆理论, 那么有$cat_\Omega(\Omega)$个正解. 证毕.

参考文献

Alves C O, Nóbrega A B, Yang M B.

Multi-bump solutions for Choquard equation with deepening potential well

Calc Var Partial Differ Equ, 2016, 55(3): 1-28

DOI:10.1007/s00526-015-0942-y      URL     [本文引用: 1]

Alves C O, Barros L M.

Existence and multiplicity of solutions for a class of elliptic problem with critical growth

Monatsh Math, 2018, 187(2): 195-215

DOI:10.1007/s00605-017-1117-z      [本文引用: 1]

Ambrosetti A, Brezis H, Cerami G.

Combined effects of concave and convex nonlinearities in some elliptic problems

J Funct Anal, 1994, 122(2): 519-543

DOI:10.1006/jfan.1994.1078      URL     [本文引用: 1]

Alves C O, Ding Y H.

Multiplicity of positive solutions to a$p$-Laplacian equation involving critical nonlinearity

J Math Anal Appl, 2003, 279(2): 508-521

DOI:10.1016/S0022-247X(03)00026-X      URL     [本文引用: 1]

Bahri A, Coron J M.

On a nonlinear elliptic equation involving the Sobolev exponent: The effect of the topology of the domain

Comm Pure Appl Math, 1988, 41(3): 253-294

DOI:10.1002/cpa.v41:3      URL     [本文引用: 1]

Bartsch T, Wang Z Q.

Multiple positive solutions for a nonlinear Schrödinger equation

Z Angew Math Phys, 2000, 51(3): 366-384

DOI:10.1007/PL00001511      URL     [本文引用: 3]

Benci V, Cerami G.

Positive solutions of some nonlinear elliptic problems in exterior domains

Arch Rational Mech Anal, 1987, 99(4): 283-300

DOI:10.1007/BF00282048      URL     [本文引用: 1]

Benci V, Cerami G.

The effect of the domain topology on the number of positive solutions of nonlinear elliptic problems

Arch Rational Mech Anal, 1991, 114(1): 79-93

DOI:10.1007/BF00375686      URL     [本文引用: 1]

Benci V, Cerami G.

Multiple positive solutions of some elliptic problems via the Morse theorem and the domain topology

Calc Var Partial Differ Equ, 1994, 2: 29-48

DOI:10.1007/BF01234314      URL     [本文引用: 1]

Cingolani S, Clapp M, Secchi S.

Multiple solutions to a magnetic nonlinear Choquard equation

Z Angew Math Phys, 2012, 63: 233-248

DOI:10.1007/s00033-011-0166-8      URL     [本文引用: 1]

Clapp M, Ding Y H.

Positive solutions for a Schrödinger equation with critical nonlinearity

Z Angew Math Phys, 2004, 55(4): 592-605

DOI:10.1007/s00033-004-1084-9      URL     [本文引用: 1]

Gao F S, Da Silva E D, Yang M B, Zhou J Z.

Existence of solutions for critical Choquard equations via the concentration compactness method

Proc R Soc Edinb Sect A Math, 2020, 150(2): 921-954

DOI:10.1017/prm.2018.131      URL    

In this paper, we consider the nonlinear Choquard equation$$-\\Delta u + V(x)u = \\left( {\\int_{{\\open R}^N} {\\displaystyle{{G(u)} \\over { \\vert x-y \\vert ^\\mu }}} \\,{\\rm d}y} \\right)g(u)\\quad {\\rm in}\\;{\\open R}^N, $$where 0 &lt; μ &lt;N,N⩾ 3,g(u) is of critical growth due to the Hardy–Littlewood–Sobolev inequality and$G(u)=\\int ^u_0g(s)\\,{\\rm d}s$. Firstly, by assuming that the potentialV(x) might be sign-changing, we study the existence of Mountain-Pass solution via a nonlocal version of the second concentration- compactness principle. Secondly, under the conditions introduced by Benci and Cerami, we also study the existence of high energy solution by using a nonlocal version of global compactness lemma.

Gao F S, Yang M B.

On the Brezis-Nirenberg type critical problem for nonlinear Choquard equation

Sci China Math, 2018, 61(7): 1219-1242

DOI:10.1007/s11425-016-9067-5      [本文引用: 1]

Ghimenti M, Pagliardini D.

Multiple positive solutions for a slightly subcritical Choquard problem on bounded domains

Calc Var Partial Differ Equ, 2019, 58(5): 1-21

DOI:10.1007/s00526-018-1462-3      [本文引用: 6]

Goel D.

The effect of topology on the number of positive solutions of elliptic equation involving Hardy-Littlewood-Sobolev critical exponent

Topol Methods Nonlinear Anal, 2019, 54(2): 751-771

[本文引用: 1]

Goel D, Rădulescu V D, Sreenadh K.

Coron problem for nonlocal equations involving Choquard Nonlinearity

Adv Nonlinear Stud, 2020, 20(1): 141-161

DOI:10.1515/ans-2019-2064      URL     [本文引用: 1]

We consider the following Choquard equation:

Goel D, Sreenadh K.

Critical growth elliptic problems involving Hardy-Littlewood-Sobolev critical exponent in non-contractible domains

Adv Nonlinear Anal, 2020, 9(1): 803-835

[本文引用: 1]

Lieb E H.

Existence and uniqueness of the minimizing solution Choquard's nonlinear equations

Stud Appl Math, 1977, 57: 93-105

DOI:10.1002/sapm.v57.2      URL     [本文引用: 1]

Lieb E H, Loss M.

Analysis

Rhode Island: AMS, 2001

[本文引用: 1]

Ma P, Zhang J H.

Existence and multiplicity of solutions for fractional Choquard

Nonlinear Anal, 2017, 164: 100-117

DOI:10.1016/j.na.2017.07.011      URL     [本文引用: 1]

Lions P L.

The Choquard equation and related questions

Nonlinear Anal, 1980, 4(6): 1063-1072

DOI:10.1016/0362-546X(80)90016-4      URL     [本文引用: 1]

Moroz V, Van Schaftingen J.

A guide to the Choquard equation

J Fixed Point Theory Appl, 2017, 19(1): 773-813

DOI:10.1007/s11784-016-0373-1      URL     [本文引用: 1]

Moroz V, Van Schaftingen J.

Ground states of nonlinear Choquard equations: Existence, qualitative properties and decay asymptotics

J Funct Anal, 2013, 265: 153-184

DOI:10.1016/j.jfa.2013.04.007      URL     [本文引用: 2]

Moroz V, Van Schaftingen J.

Existence of groundstates for a class of nonlinear Choquard equations

Trans Am Math Soc, 2015, 367(9): 6557-6579

DOI:10.1090/tran/2015-367-09      URL     [本文引用: 1]

Moroz V, Van Schaftingen J.

Semi-classical states for the Choquard equation

Calc Var Partial Differ Equ, 2015, 52(1/2): 199-235

DOI:10.1007/s00526-014-0709-x      URL     [本文引用: 1]

Palatucci G, Pisante A.

Improved Sobolev embeddings, profile decomposition and concentration compactness fractional Sobolev space

Calc Var Partial Differ Equ, 2014, 50(3/4): 799-829

DOI:10.1007/s00526-013-0656-y      URL    

Pekar S.

Untersuchung über die Elektronentheorie der Kristalle

Berlin: Akademie Berlag, 1954

[本文引用: 1]

Shen Z F, Gao F S, Yang M B.

On critical Choquard equation with potential well

Discrete Contin Dyn Syst, 2018, 38(7): 3567-3593

DOI:10.3934/dcds.2018151      URL     [本文引用: 1]

Willem M. Minimax Theorems, Progress in Nonlinear Differential Equations and Their Applications. Boston: Birkhäuser, 1996, 24

[本文引用: 4]

Xu Z Y, Yang J F.

Multiple solutions to multi-critical Schrödinger equations

Adv Nonlinear Stud, 2022, 22(1): 273-288

DOI:10.1515/ans-2022-0014      URL     [本文引用: 1]

In this article, we investigate the existence of multiple positive solutions to the following multi-critical Schrödinger equation: \n (0.1)\n \n \n \n \n \n \n \n \n −\n Δ\n u\n +\n λ\n V\n \n (\n \n x\n \n )\n \n u\n =\n μ\n ∣\n u\n \n \n \n ∣\n \n \n p\n −\n 2\n \n \n u\n +\n \n \n \n \n ∑\n \n \n \n \n i\n =\n 1\n \n \n k\n \n \n \n (\n \n ∣\n x\n \n \n \n ∣\n \n \n −\n \n (\n \n N\n −\n \n \n α\n \n \n i\n \n \n \n )\n \n \n \n ∗\n ∣\n u\n \n \n \n ∣\n \n \n \n \n 2\n \n \n i\n \n \n ∗\n \n \n \n \n \n )\n \n ∣\n u\n \n \n \n ∣\n \n \n \n \n 2\n \n \n i\n \n \n ∗\n \n \n −\n 2\n \n \n u\n \n \n \n in\n \n \n \n \n \n R\n \n \n N\n \n \n,\n \n \n \n \n \n u\n \n ∈\n \n \n H\n \n \n 1\n \n \n \n (\n \n \n \n R\n \n \n N\n \n \n \n )\n \n,\n \n \n \n \n \n \n \n \\left\\{\\begin{array}{l}-\\Delta u+\\lambda V\\left(x)u=\\mu | u{| }^{p-2}u+\\mathop{\\displaystyle \\sum }\\limits_{i=1}^{k}\\left(| x{| }^{-\\left(N-{\\alpha }_{i})}\\ast | u{| }^{{2}_{i}^{\\ast }})| u{| }^{{2}_{i}^{\\ast }-2}u\\hspace{1.0em}\\hspace{0.1em}\\text{in}\\hspace{0.1em}\\hspace{0.33em}{{\\mathbb{R}}}^{N},\\hspace{1.0em}\\\\ u\\hspace{0.33em}\\in {H}^{1}\\left({{\\mathbb{R}}}^{N}),\\hspace{1.0em}\\end{array}\\right.\n \n where \n \n \n \n λ\n,\n μ\n ∈\n \n \n R\n \n \n +\n \n \n,\n N\n ≥\n 4\n \n \\lambda,\\mu \\in {{\\mathbb{R}}}^{+},N\\ge 4\n \n, and \n \n \n \n \n \n 2\n \n \n i\n \n \n ∗\n \n \n =\n \n \n N\n +\n \n \n α\n \n \n i\n \n \n \n \n N\n −\n 2\n \n \n \n {2}_{i}^{\\ast }=\\frac{N+{\\alpha }_{i}}{N-2}\n \n with \n \n \n \n N\n −\n 4\n &lt;\n \n \n α\n \n \n i\n \n \n &lt;\n N\n \n N-4\\lt {\\alpha }_{i}\\lt N\n \n, \n \n \n \n i\n =\n 1\n,\n 2\n,\n \n …\n \n,\n k\n \n i=1,2,\\ldots,k\n \n are critical exponents and \n \n \n \n 2\n &lt;\n p\n &lt;\n \n \n 2\n \n \n min\n \n \n ∗\n \n \n =\n min\n \n {\n \n \n \n 2\n \n \n i\n \n \n ∗\n \n \n :\n i\n =\n 1\n,\n 2\n,\n \n …\n \n,\n k\n \n }\n \n \n 2\\lt p\\lt {2}_{\\min }^{\\ast }={\\rm{\\min }}\\left\\{{2}_{i}^{\\ast }:i=1,2,\\ldots,k\\right\\}\n \n. Suppose that \n \n \n \n Ω\n =\n int\n \n \n \n V\n \n \n −\n 1\n \n \n \n (\n \n 0\n \n )\n \n ⊂\n \n \n R\n \n \n N\n \n \n \n \\Omega ={\\rm{int}}\\hspace{0.33em}{V}^{-1}\\left(0)\\subset {{\\mathbb{R}}}^{N}\n \n is a bounded domain, we show that for \n \n \n \n λ\n \n \\lambda \n \n large, problem (0.1) possesses at least \n \n \n \n \n \n cat\n \n \n Ω\n \n \n \n (\n \n Ω\n \n )\n \n \n {{\\rm{cat}}}_{\\Omega }\\left(\\Omega )\n \n positive solutions.

Yang J F, Zhu L P.

Multiple solutions to Choquard equation in exterior domain

J Math Anal Appl, 2022, 507(1): 125726

DOI:10.1016/j.jmaa.2021.125726      URL     [本文引用: 1]

/