数学物理学报  2016, Vol. 36 Issue (4): 639-648   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
胡良根
非线性Hénon方程解的Liouville定理
胡良根     
宁波大学数学系 浙江宁波 315211
摘要:该文研究了二阶和四阶非线性Hénon-Lane-Emden方程有限Morse指标解的Liouville定理.利用一种新方法,即使用单调公式、Pohozaev恒等式和doubling引理等相结合证明了其结果.
关键词有限Morse指标解     单调公式     Liouville定理     Doubling引理     Pohozaev恒等式    
Liouville Type Theorems of Solutions for the Nonlinear Hénon Equations
Hu Lianggen     
Department of Mathematics, Ningbo University, Zhejiang Ningbo 315211
Abstract: In this paper, Liouville theorem of finite Morse index solutions for the second order and fourth order nonlinear Hénon-Lane-Emden equations is considered. Adopting a new method, i.e., the monotonicity formula, Pohozaev identity combining with doubling lemma, the main results is proved.
Key words: Finite Morse index solution     Monotonicity formula     Liouville theorem     Doubling lemma     Pohozaev identity    
1 引言

本文考虑二阶非线性Hénon-Lane-Emden方程

$ - \Delta u=|x|^{\alpha} |u|^{p-1}u, \; x \in \Omega $ (1.1)

和四阶的

$ \Delta^2 u=|x|^{\alpha} |u|^{p-1}u, \; x \in \Omega $ (1.2)

有限Morse指标解的Liouville定理, 其中$\alpha \ge 0$, $p>1$$\Omega \subset {\Bbb R}^n$是一个无界定义域.

定义 1.1 方程(1.1)和方程(1.2)的能量函数分别为

$ {\mathfrak E}(u)= \int_{\Omega} \left [\frac{1}{2} |\nabla u|^2-\frac{1}{p+1} |x|^{\alpha} |u|^{p+1} \right] {\rm d}x $

$ {\cal E}(u)= \int_{\Omega} \left [\frac{1}{2} |\Delta u|^2-\frac{1}{p+1} |x|^{\alpha} |u|^{p+1} \right] {\rm d}x, $

其临界点$u \in C^2 (\Omega)$ ($u \in C^4 (\Omega))$定义为

(ⅰ) 方程$(1.1)$ (方程(1.2))的稳定解, 如果对任意的$\phi \in C_0^1( \Omega )$ ($\psi \in C_0^2( \Omega ))$, 有${\mathfrak E}_{u}(\phi)= \int_{\Omega} \left ( |\nabla \phi|^2- p|x|^{\alpha} |u|^{p-1} \phi^2 \right ) {\rm d}x\ge 0({\cal E}_{uu}(\psi)= \int_{\Omega} \left( |\Delta \psi|^2 - p |x|^{\alpha} |u|^{p-1} \psi^2 \right ) {\rm d}x \ge 0).$

(ⅱ) 方程$(1.1)$ (方程(1.2))具有Morse指标$k \ge 0$解, 如果$k$是子空间$X_k \subset C_0^1(\Omega)$ ($X_k \subset C_0^2(\Omega))$的最大维数, 使得对所有的$\phi \in X_k \backslash \{0\} (\psi \in X_k \backslash \{0\})$, 有${\mathfrak E}_{u}(\phi) < 0 ({\cal E}_{uu}(\psi) < 0)$.因此, $u$是稳定的, 当且仅当它的Morse指标等于$0$.

(ⅲ) 方程$(1.1)$ (方程(1.2))在紧集$\digamma \subset \Omega$外的稳定解, 如果对任意的$\phi \in C_0^1(\Omega \backslash \digamma )(\psi \in C_0^2 (\Omega \backslash \digamma))$, 有${\mathfrak E}_{u} (\phi) \ge 0 ({\cal E}_{uu} (\psi) \ge 0)$.这意味着任何有限Morse指标解在某个紧集$\digamma \subset \Omega$外是稳定的.

1992年, Bahri和Lions[1]首次阐述了半线性椭圆方程解的Morse指标, 获得了许多有关解和有限Morse指标的性质. 2004年, Dancer[2]${\Bbb R}^n$(其中$n=2$$n=3$)中研究了非线性方程$-\Delta u=f(u)$的有界解, 证明了对于许多$f$, 其弱稳定解和有限Morse指标解是单一的.再使用其结果研究了具有Dirichlet或Neumann边界条件的$-\varepsilon^2 \Delta u=f(u)$稳定解和有限Morse指标解的性质.其后非线性椭圆方程有限Morse指标解引理广泛关注. 2007年, Farina[3]对二阶椭圆方程解的Liouville定获得了重要成果, 即在有界和无界定义域$\Omega \subset {\Bbb R}^n$, 且$n \ge 2$$p < 1$, 他给出了考虑了Lane-Emden方程

$ - \Delta u =|u|^{p-1}u, \;\ x \in \Omega $

有限Morse指标(正或符号改变)解的完全分类.其主要的证明方法是使用Moser迭代方法.应用这种广义的Moser迭代方法, 许多专家讨论了调和和双调和椭圆方程, 如文献[4-6]等.另一方面, 研究解的Liouvile定理的临界指数有很多应用, 如文献[7]等.

但这种经典的Moser迭代技术对双调和方程

$ \Delta ^2 u=|u|^{p-1}u, \;\ \Omega \subset {\Bbb R}^n $ (1.3)

有限Morse指标解并不能给出完全分类.为解决这个问题, 最近, Dávila, Dupaigne, Wang和Wei[8]推导出方程(1.3)解的一个单调公式, 再结合爆缩方法, 正则性等理论给出了方程(1.3)稳定解、有限Morse指标解的完全分类.

另一方面, Falzy和Ghoussoub[9]利用Souplet的方法[10-11]研究了方程(1.1)和方程(1.2)有限Morse指标解的Liouville定理.在文献[9]中定理2的证明Step 3中, 可以发现不等式

$ \left |\int_{B_{2R}} \nabla \zeta_R \cdot \nabla uu \right | \le R^{-1}\int_{B_{2R}}|\nabla u| \left (|x|^{\frac{\alpha}{p+1}}u \right )|x|^{-\frac{\alpha}{p+1}}\\ \le R^{-1}\left (\int_{B_{2R}} |\nabla u|^2 \right )^{\frac{1}{2}} \left (\int_{B_{2R}}|x|^{\alpha} u^{p+1}\right )^{\frac{1}{p+1}} \left (\int_{B_{2R}} |x|^{-\frac{2\alpha}{p-1}}\right )^{\frac{p-1}{2(p+1)}} $

和定理3的证明Step 3中, 有不等式

$ I_1(R) \le R^{-2}\int_{B_{2R}} |\Delta u|\left (|x|^{\frac{\alpha}{p+1}}u \right )|x|^{-\frac{\alpha}{p+1}}\\ \le R^{-2} \left (\int_{B_{2R}} |\Delta u|^2 \right )^{\frac{1}{2}} \left (\int_{B_{2R}} |x|^{\alpha} u^{p+1} \right )^{\frac{1}{p+1}} \left (\int_{B_{2R}} |x|^{-\frac{2\alpha}{p-1}} \right )^{\frac{p-1}{2(p+1)}} $

$ I_4(R) \le R^{-4} \int_{B_{2R}} |x|^{\frac{\alpha}{p+1}}u^2 |x|^{-\frac{\alpha}{p+1}} \\ \ R^{-4}\left (\int_{B_{2R}} |x|^{\alpha}u^{p+1} \right )^{\frac{2}{p+1}}\left ( \int_{B_{2R}} |x|^{-\frac{2\alpha}{p-1}}\right )^{\frac{p-1}{p+1}}. $

仔细检查发现当$n-\frac{2\alpha}{p-1} \le 0$时, 积分$ \int_{B_{2R}} |x|^{-\frac{2\alpha}{p-1}}{\rm d}x $中有奇性.

本文研究了当$p$属于次临界情形、临界情形时非线性调和和双调和Hénon-Lane-Emden方程有限Morse指标解的Liouville定理.文中使用一种新的证明方法, 即利用单调公式、Pohozaev恒等式和doubling引理等相结合证明了其结果, 从而避免讨论文献[9]中出现的问题.

2 预备知识和引理

采用文献[8, 定理1.1]和文献[12, 定理2.1]的方法, 可以建立方程(1.1)的单调公式, 即对任给定的$x \in \Omega$, 设$0 < r < R$$B_r(x) \subset B_R(x) \subset \Omega$.取$u \in W^{2, 2}(\Omega)$$|x|^{\alpha} |u|^{p+1} \in L^1 (\Omega)$, 定义

$ {\mathfrak M}(r;x, u) : = r^{\frac{2(p+1)+2\alpha}{p-1}-n} \int_{B_r(x)} \left [\frac{1}{2} |\nabla u|^2-\frac{1}{p+1}|x|^{\alpha}|u(x)|^{p+1} \right] {\rm d}x \\ +\frac{2+\alpha}{2(p-1)} \left (r^{\frac{4+2\alpha}{p-1}+1-n} \int_{\partial B_r(x)} u^2{\rm d}S \right ). $

定理 2.1 假定$n\ge 3$, $\alpha \ge 0$, $p>1$, $u \in W_{loc}^{2, 2}(\Omega)$是方程$(1.1)$的一个弱解, 且$|x|^{\alpha}|u|^{p+1} \in L^1(\Omega)$, 则对任意的$r \in (0, R)$, ${\mathfrak M}(r;x, u)$是非减的且有

$ \frac{{\rm d}}{{\rm d} r}{\mathfrak M}(r;0, u)=r^{\frac{4+2\alpha}{p-1}+2-n} \int_{\partial B_r} \left (\frac{2+\alpha}{p-1} r^{-1}u+\frac{\partial u}{\partial r} \right )^2 {\rm d}S \ge 0. $ (2.1)

 由于对每个$B_r(x) \subset B_R(0)$, $u\in W^{2, 2}(B_R(0))$$|x|^{\alpha}|u|^{p+1} \in L^1(B_R(0))$, 从而有$u \in W^{1, 2}(\partial B_r(x))$.因此, 我们可以假定$x=0$.又方程(1.1)是变分的, 其能量函数定义为

$ \widetilde{{\mathfrak M}}(\lambda):=\lambda^{\frac{2(p+1)+2\alpha}{p-1}-n}\int_{B_{\lambda}}\left [\frac{1}{2}(\nabla u)^2-\frac{1}{p+1} |x|^{\alpha} |u|^{p+1}\right]{\rm d}x. $

在标度变换$u^{\lambda}(x):=\lambda^{\frac{2+\alpha}{p-1}}u(\lambda x)$, $\forall \lambda >0$下, 方程(1.1)的解是不变的.事实上

$ -\Delta u^{\lambda}(x)=-\lambda^{\frac{2+\alpha}{p-1}+2} \Delta u(\lambda x)= |x|^{\alpha} \lambda^{p\frac{2+\alpha}{p-1}}|u(\lambda x)|^{p-1}u(\lambda x)=|x|^{\alpha} |u^{\lambda}(x)|^{p-1}u^{\lambda}(x). $

特别, 对能量函数进行重标度变换可得

$ \widetilde{{\mathfrak M}}(\lambda) = \lambda^{\frac{2(p+1)+2\alpha}{p-1}-n}\int_{B_{\lambda}} \left [\frac{1}{2}(\nabla u)^2-\frac{1}{p+1} |x|^{\alpha} |u|^{p+1}\right] {\rm d}x \\ = \int_{B_1} \left ( \frac{1}{2}|\nabla u^{\lambda}|^2-\frac{1}{p+1} |x|^{\alpha}|u^{\lambda}|^{p+1} \right ) {\rm d}x. $

对上式关于λ求导可得

$ \frac{{\rm d} \widetilde{{\mathfrak M}}(\lambda)}{{\rm d} \lambda} \int_{B_1} \left [\nabla u^{\lambda}\cdot \nabla \frac{{\rm d}u^{\lambda}}{{\rm d} \lambda} -|x|^{\alpha} |u^{\lambda}|^{p-1}u^{\lambda}\frac{{\rm d} u^{\lambda}}{{\rm d} \lambda}\right] {\rm d}x \\ = -\int_{B_1} \Delta u^{\lambda} \frac{{\rm d} u^{\lambda}}{{\rm d} \lambda} {\rm d}x+\int_{\partial B_1} \frac{\partial u^{\lambda}}{\partial r} \frac{{\rm d} u^{\lambda}}{{\rm d} \lambda} {\rm d}S +\int_{B_1} \Delta u^{\lambda}\frac{{\rm d} u^{\lambda}}{{\rm d} \lambda} {\rm d}x \\ = \int_{\partial B_1}\frac{\partial u^{\lambda}}{\partial r}\frac{{\rm d} u^{\lambda}}{{\rm d}\lambda} {\rm d}S. $ (2.2)

下面就$r=|x|$关于$u^{\lambda}$的所有导数用$\lambda$关于$λ$变量的导数来表示.由方程(1.1)可得

$ \frac{{\rm d}u^{\lambda}}{{\rm d} \lambda} = \frac{2+\alpha}{p-1} \lambda^{\frac{2+\alpha}{p-1}-1}u(\lambda x) +\lambda^{\frac{2+\alpha}{p-1}-1} r \frac{\partial u}{\partial r}(\lambda x) \\ = \frac{1}{\lambda} \left (\frac{2+\alpha}{p-1} u^{\lambda}(x)+r\frac{\partial u^{\lambda}}{\partial r} (x) \right ). $

$ r \frac{\partial u^{\lambda}}{\partial r}=\lambda \frac{{\rm d} u^{\lambda}}{{\rm d} \lambda}-\frac{2+\alpha}{p-1}u^{\lambda}. $

因此有

$ \frac{{\rm d} \widetilde{{\mathfrak M}}(\lambda)}{{\rm d} \lambda} = \int_{\partial B_1} \left ( \lambda \frac{{\rm d} u^{\lambda}}{{\rm d} \lambda}-\frac{2+\alpha}{p-1} u^{\lambda} \right ) \frac{{\rm d} u^{\lambda}}{{\rm d} \lambda} {\rm d}S \\ = \int_{\partial B_1} \lambda \left (\frac{{\rm d} u^{\lambda}}{{\rm d} \lambda} \right )^2 {\rm d}S-\frac{2+\alpha}{2(p-1)} \int_{\partial B_1} \frac{{\rm d} (u^{\lambda})^2}{{\rm d} \lambda} {\rm d}S \\ \ge -\frac{2+\alpha}{2(p-1)} \int_{\partial B_1} \frac{{\rm d} (u^{\lambda})^2}{{\rm d} \lambda} {\rm d}S. $

由重标度变换可得

$ \int_{\partial B_1} \frac{{\rm d} (u^{\lambda})^2}{{\rm d} \lambda} {\rm d}S=\frac{{\rm d}}{{\rm d}\lambda} \left ( \int_{\partial B_1} (u^{\lambda})^2 {\rm d}S \right ) =\frac{{\rm d} }{{\rm d} \lambda} \left (\lambda^{\frac{4+2\alpha}{p-1}+1-n} \int_{\partial B_{\lambda}} u^2 {\rm d}S \right ) $

$ \int_{\partial B_1} \lambda \left (\frac{{\rm d} u^{\lambda}}{{\rm d} \lambda} \right )^2 {\rm d}S = \int_{\partial B_1} \lambda \left (\frac{2+\alpha}{p-1} \lambda^{\frac{2+\alpha}{p-1}} u(\lambda x)+\lambda^{\frac{2+\alpha}{p-1}} \lambda r\frac{\partial u(\lambda x)}{\partial (\lambda r)} \right )^2 {\rm d}S \\ = \lambda^{\frac{4+2\alpha}{p-1}+2-n}\int_{\partial B_{\lambda}} \left (\frac{2+\alpha}{p-1}r^{-1}u+\frac{\partial u}{\partial r} \right )^2 {\rm d}S. $

联合(2.2)式可得

$ \frac{{\rm d} {\mathfrak M}}{{\rm d} r}(r;0, u)=r^{\frac{4+2\alpha}{p-1}+2-n}\int_{\partial B_r} \left (\frac{2+\alpha}{p-1} r^{-1}u + \frac{\partial u}{\partial r} \right )^2 {\rm d}S \ge 0. $

证毕.

由积分的性质可以得到定理的结论.

注 2.1 在定理2.1的条件下, 设对所有的$\delta \in (r, R)$, 有${\mathfrak M}(\delta;0, u) \equiv \mbox{常数, }$$u$$B_R \backslash B_r$中是齐次的, 即对$x \in B_R \backslash B_r$, 有

$ u(x)=|x|^{-\frac{2+\alpha}{p-1}}u\left (\frac{x}{|x|} \right ). $

 任取$r_1, r_2 \in (r, R)$$r_1 < r_2$, 从(2.1)式可知

$ 0 = {\mathfrak M}(r_2;0, u)-{\mathfrak M}(r_1;0, u)=\int_{r_1}^{r_2} \frac{{\rm d}}{{\rm d}\delta} {\mathfrak M}(\delta;0, u) {\rm d}\delta \\ = \int_{B_{r_2} \backslash B_{r_1}}\frac{\left (\frac{2+\alpha}{p-1} \delta^{-1} u+\frac{\partial u}{\partial \delta} \right )^2}{|x|^{n-2-\frac{4+2\alpha}{p-1}}} {\rm d}x. $

$ \frac{2+\alpha}{p-1}\delta^{-1}u+\frac{\partial u}{\partial \delta}=0\;\Longrightarrow\; u(x)=|x|^{-\frac{2+\alpha}{p-1}}u \left (\frac{x}{|x|} \right ), \;\ x \in B_R \backslash B_r. $

证毕

采用文献[12, 定理1.1]相同的证明方法, 利用定理2.1(单调公式)、注2.1和Laplace-Beltrami算子可以得到:

定理 2.2 对任意的$1 < p \le P_+(n, \alpha)$, 假定$u\in C^2({\Bbb R}^n)$是方程$(1.1)$${\Bbb R}^n$中的稳定解, 则$u \equiv 0$. (其中$P_+(n, \alpha)$在文献[4, p3286]中给出.)

由文献[9, 引理8]可知:

引理 2.1 设$\Omega \subset {\Bbb R}^n$, $u \in C^2(\Omega)$是方程$(1.1)$的稳定解, 则对任意的$t \in [1, 2p+2\sqrt{p(p-1)}-1)$, $\xi \in C_c^1(\Omega)$$0\le \xi \le 1$, 当$m$足够大时, 有

$ \int_{\Omega} \left (|\nabla u|^2 u^{t-1}+|x|^{\alpha} |u|^{t+p} \right )\xi^{2m} {\rm d}x \le C \int_{\Omega}|x|^{-\frac{t+1}{p-1}\alpha}|\nabla \xi|^{2\frac{t+p}{p-1}} {\rm d}x, $

其中常数$C$$u$和Ω无关.

注 2.2 如果在引理2.1中令$t=1$, 选取$\xi \in C_c^1(B_{2R}(0))$, $0 \le \xi \le 1$$|\nabla \xi | \le \frac{C}{R}$, 且在$B_R(0)$中有$B_R(0)$$\xi \equiv 1$, 则

$ \int_{B_R(x)}\left (|\nabla u|^2+|x|^{\alpha} |u|^{p+1} \right ) {\rm d}x \le CR^{n-\frac{2(p+1)+2\alpha}{p-1}}. $ (2.3)

由文献[12, 引理2.2]可知:

引理 2.2 设$u\in C^4({\Bbb R}^n)$是方程(1.2)的稳定解.则对充分大的$m$和对所有的$\psi \in C^4_0({\Bbb R}^n)$$0 \le \psi \le 1$, 有

$ \int_{{\Bbb R}^n} \left ( |\Delta u|^2+|x|^{\alpha} |u|^{p+1} \right )\psi^{2m}{\rm d}x \le C \int_{{\Bbb R}^n} |x|^{-\frac{2\alpha}{p-1}}|F(\psi^m)|^{\frac{p+1}{p-1}} {\rm d}x, $

其中$F(\psi^m)=|\nabla \psi|^4+\psi^{2(2-m)}\Big [|\nabla (\Delta \psi^m)\cdot \nabla \psi^m|+|\Delta \psi^m|^2+|\Delta |\nabla \psi^m|^2| \Big]$, 常数$C$$u$无关.

3 主要结果

引理 3.1 (Pohozaev型恒等式)方程$(1.1)$和方程(1.2)的Pohozaev型恒等式分别为

$ \frac{n+\alpha}{p+1} \int_{\Omega} |x|^{\alpha} |u|^{p+1} {\rm d}x -\frac{n-2}{2} \int_{\Omega} |\nabla u|^2 {\rm d}x \nonumber \\ = \frac{1}{p+1}\int_{\partial \Omega} |x|^{\alpha} |u|^{p+1} x \cdot \nu {\rm d}S+\int_{\partial \Omega} x\cdot \nabla u \nu \cdot \nabla u {\rm d}S-\frac{1}{2}\int_{\partial \Omega}|\nabla u|^2x \cdot \nu {\rm d}S $ (3.1)

$ \int_{\Omega} \left (\frac{n+\alpha}{p+1} |x|^{\alpha} |u|^{p+1}-\frac{n-4}{2} |\Delta u|^2 \right ) {\rm d}x \nonumber \\ = \frac{1}{p+1}\int_{\partial \Omega} |x|^{\alpha}|u|^{p+1}x \cdot \nu {\rm d}S -\int_{\partial \Omega} \nabla u \cdot x \nabla (\Delta u)\cdot \nu {\rm d}S \nonumber \\ -\frac{1}{2}\int_{\partial \Omega} |\Delta u|^2 x \cdot \nu {\rm d}S +\int_{\partial \Omega} \Delta u \nabla(x \cdot \nabla u)\cdot \nu {\rm d}S. $ (3.2)

利用文献[13, 引理5.1]和定理2.2, 可以得到下面点点估计.

引理 3.2 设$u \in C^2({\Bbb R}^n)$是方程$(1.1)$的有限Morse指标解.则存在常数$C$$R_*$, 使得对所有的$x \in B_{R_*}^c$, 有

$ |u(x)| \le C |x|^{-\frac{2+\alpha}{p-1}}, $ (3.3)

$\forall x \in B_{3R_*}^c$, 有

$ |\nabla u(x)| \le C|x|^{-\frac{2+\alpha}{p-1}-1}. $ (3.4)

 由于$u$是方程(1.1)的有限Morse指标解, 从而可以假定$u$在球$B_{R_*}$外是稳定的.

假设结论不成立, 则对所有的$x \in B_{R_*}^c$, 设$\omega(x)=|u(x)|^{\frac{p-1}{2+\alpha}}$$d(x)=|x|-R_*$, 存在一个序列$\{x_m\} \subseteq B_{R_*}^c$, 使得

$ \omega (x_m)d(x_m) \ge 2m. $

利用文献[13, 引理5.1]的doubling引理可以找到另一个序列$\{y_m\}\ \subseteq B_{R_*}^c$, 使得

(a) $\omega(y_m)d(y_m) \ge 2m$;

(b) $\omega(y_m) \ge \omega(x_m)$;

(c) 对任意的$z\in B_{R_*}^c$, 使得$|z-y_m|\le \frac{m}{\omega(y_m)}$, 有$\omega (z) \le 2 \omega(y_m)$.

由于在${\Bbb R}^n$的任何紧集中$u$是有界的, 从而可推出$d(x_m) \to \infty$.

对任意的$x\in B_m(0)$, 设$ u_m(x):=\omega(y_m)^{-\frac{2+\alpha}{p-1}}u \left (y_m+\omega(y_m)^{-1}x \right )$, 则$|u_m(0)|=1$, 且从(c)可推出在$B_m(0)$$|u_m|\le 2^{\frac{2+\alpha}{p-1}}$.

另一方面, 应用(a)可知$B_{\frac{m}{\omega}(y_m)}(y_m) \cap B_{R_*} = \emptyset$, 从而解$u$$B_{\frac{m}{\omega}(y_m)}(y_m)$中是稳定的, 因此$u_m$$B_m(0)$中是稳定的.由椭圆正则性理论可知$u_m$$W^{2, p+1}(B_m(0))$中是一致有界的, 因此由嵌入定理和椭圆Schauder估计(选取子序列), 可以假定$u_m$$C_{loc}^2({\Bbb R}^n)$中收敛于$u_{\infty}$, 且$u_{\infty}$满足

(ⅰ) $|u_{\infty}(0)|=1$;

(ⅱ) 在${\Bbb R}^n$中, $u_{\infty}$是方程(1.1)的稳定解且有$|u_{\infty}|\le 2^{\frac{2+\alpha}{p-1}}$.

定理2.2意味着矛盾$u_{\infty} \equiv 0$.则(3.3)式成立.

下面证明(3.4)式:任选取$\overline{x}$, 使得$|\overline{x}|>3R_*$$\lambda =\frac{|\overline{x}|}{2}$, 且记

$ \varpi(x):=\lambda^{\frac{2+\alpha}{p-1}}u(\overline{x}+\lambda x). $

计算可得

$ |\varpi(x)|\le C\lambda^{\frac{2+\alpha}{p-1}}|\overline{x}+\lambda x|^{-\frac{2+\alpha}{p-1}} \le C_1, \;\ \forall x\in B_1(0). $

对任意的$x \in B_1(0)$, 由椭圆估计可得$|\nabla \varpi(x)| \le C_2$.则从$\nabla \varpi(x)=\lambda^{\frac{2+\alpha}{p-1}+1} \nabla u(\overline{x}+\lambda x)$可推出

$ |\nabla u(x)| \le C|x|^{-\frac{2+\alpha}{p-1}-1}, \;\ \forall x \in B_{3R_*}^c. $

证毕

定理 3.1 假设$u \in C^2({\Bbb R}^n)$是方程$(1.1)$的有限Morse指标解.

$\bullet$ 如果$p \in \big (1, \frac{n+2+2\alpha}{n-2} \big )$, 则$u \equiv 0$.

$\bullet$ 如果$p=\frac{n+2+2\alpha}{n-2}$, 则$u$有有限能量, 即

$ \int_{{\Bbb R}^n} \left |\nabla u \right |^2 {\rm d}x=\int_{{\Bbb R}^n} |x|^{\alpha} |u|^{p+1} {\rm d}x < +\infty. $

 Ⅰ.次临界情形:$1 < p < \frac{n+2+2\alpha}{n-2}$.

首先应用(3.3)式和(3.4)式估计(3.1)式右边可得

$ \frac{1}{p+1}\int_{\partial B_R} R^{1+\alpha}|u|^{p+1} +\frac{3}{2}\int_{\partial B_R} R |\nabla u|^2 \to 0, \;\; R \to +\infty. $ (3.5)

因为$u$$\Omega \subset {\Bbb R}^n$紧集外的稳定解, 则对$R>\max \{R_*+5, 2R_*\}$$\Omega \subset B_{R_*}$, 选择测试函数$\xi_R \in C^2_c ({\Bbb R}^n \backslash \Omega)$满足

$ \xi_R(x)= \left\{\begin{array}{ll} 0, \mbox{如果}\; |x| < R_*+1\; \mbox{或}\; |x|>2R, \\ 1, \mbox{如果}\; R_*+2 < |x| < R, \end{array}\right. $

$0 \le \xi_R \le 1$, $\|\nabla \xi_R\|_{L^{\infty}(B_{2R}\backslash B_R)} \le \frac{C}{R}$$\|\nabla \xi_R\|_{L^{\infty}(B_{R_*+2} \backslash B_{R_*+1}) } \le C_{R_*}$.从(2.3)式可得

$ \int_{R_*+2 < |x| < R} \left (|\nabla u|^2+|x|^{\alpha} |u|^{p+1} \right ) {\rm d}x \le C_{R_*}+C R^{n-\frac{2(p+1)+2\alpha}{p-1}}. $

因此条件$p < \frac{n+2+2\alpha}{n-2}$意味着

$ \int_{{\Bbb R}^n} \left ( (\Delta u)^2 +|x|^{\alpha} |u|^{p+1} \right ) {\rm d}x < +\infty. $

联合(3.1)式和(3.5)式可得

$ \int_{{\Bbb R}^n} \left [\frac{n-2}{2} |\nabla u|^2-\frac{n+\alpha}{p+1} |x|^{\alpha} |u|^{p+1}\right] {\rm d}x=0. $ (3.6)

下面将证明

$ \int_{{\Bbb R}^n} |\nabla u|^2 {\rm d}x = \int_{{\Bbb R}^n} |x|^{\alpha} |u|^{p+1} {\rm d}x. $ (3.7)

事实上, 选取$\zeta_R \in C^2_0(B_{2R})$, 使得对$|x| < R$, 有$\zeta_R(x)=1$, 对$|x|>2R$, 有$\zeta_R(x)=0$, 以及$0 \le \zeta_R \le 1$且满足$\|\nabla \zeta_R\|_{L^{\infty}} \le \frac{C}{R}$.用函数$u \zeta_R$乘方程(1.1)并计算可得

$ \int_{{\Bbb R}^n} \left ( |x|^{\alpha} |u|^{p+1}-|\nabla u|^2 \right )\zeta_R {\rm d}x=\int_{{\Bbb R}^n} \nabla \zeta_R \cdot \nabla uu {\rm d}x. $

选取$R>3R_*$, 由(3.3)式和(3.4)式可得

$ \left |\int_{{\Bbb R}^n} \nabla \zeta_R \cdot \nabla uu {\rm d}x\right | \le C R^{-1} \int_{B_{2R}\backslash B_{3R_*}} |x|^{-2\frac{2+\alpha}{p-1}-1} {\rm d}x \le C R^{n-\frac{2(p+1+\alpha)}{(p+1)}}. $

因为$n < \frac{2(p+1)+2\alpha}{p-1}$, 则(3.7)式成立.

联合(3.6)式和(3.7)式可得

$ \left (\frac{n-2}{2}-\frac{n+\alpha}{p+1} \right ) \int_{{\Bbb R}^n} |u|^{p+1} {\rm d}x =0, $

从而推出$u \equiv 0$.

Ⅱ.临界情形:$n=\frac{2(p+1)+2\alpha}{p-1}$.

由于$u$$B_{R_*}$外的稳定解, 则对$\forall R>3R_*$, 采用同次临界情形一样的处理方法可得

$ \int_{B_R \backslash B_{3R_*}}\left [(\nabla u)^2 +|x|^{\alpha}|u|^{p+1} \right] {\rm d}x\le C. $

$ \int_{{\Bbb R}^n} \left [(\nabla u)^2 +|x|^{\alpha} |u|^{p+1} \right] {\rm d}x < +\infty $

$ \lim\limits_{R \to \infty} \int_{B_{2R}(x)\backslash B_R(x)} \left [(\nabla u)^2 +|x|^{\alpha} |u|^{p+1} \right] {\rm d}x =0. $

因此容易推出

$ \int_{R^n} \left [(\nabla u)^2-|x|^{\alpha} |u|^{p+1} \right] {\rm d}x =0. $

证毕

利用文献[13, 引理5.1]和文献[12, 定理1.1], 可以得到下面估计(证明过程与引理3.2完全类似):

引理 3.3 设$u$是方程(1.2)的一个有限Morse指标解.则存在常数$C$$\widetilde{R}$, 使得

$ |u(x)|\le C|x|^{-\frac{4+\alpha}{p-1}}, \;\ \mbox{ 对所有的 }\ x \in B_{\widetilde{R}}^c, $ (3.8)

$ |\nabla^ku(x)|\le C|x|^{-\frac{4+\alpha}{p-1}-k}, \;\ k=1, 2, 3 \mbox{ 和对所有的 }\ x \in B_{3\widetilde{R}}^c. $ (3.9)

定理 3.2 设$u \in C^4 ({\Bbb R}^n)$是方程(1.2)的有限Morse指标解.

$\bullet$ 如果$1 < p < \frac{n+4+2\alpha}{n-4}$, 则$u \equiv 0$.

$\bullet$ 如果$p=\frac{n+4+2\alpha}{n-4}$, 则$u$具有有限能量, 即

$ \int_{{\Bbb R}^n} \left (\Delta u \right )^2 {\rm d}x =\int_{{\Bbb R}^n} |x|^{\alpha} |u|^{p+1} {\rm d}x < +\infty. $

这个定理的证明方法类似于文献[12, 定理1.2].为了完整性, 我们重新给出证明.

 Ⅰ.次临界情形:$1 < p < \frac{n+4+2\alpha}{n-4}$.

利用(3.8)式和(3.9)式可得(3.2)式右边的估计为

$ \int_{\partial B_R} \frac{R^{1+\alpha}}{p+1} |u|^{p+1}+R \frac{\partial u}{\partial r} \frac{\partial \Delta u}{\partial r}+\frac{R}{2} (\Delta u)^2+\left |\Delta u \frac{\partial (x \cdot \nabla u) }{\partial r} \right |\to 0, \;\; R \to +\infty. $

由于$u$$\Omega \subset {\Bbb R}^n$紧集外是稳定的, 则对$R>\frac{5}{3}\widetilde{R}$$\Omega \subset B_{\widetilde{R}}$, 使得

$ \psi_R(x)= \left\{\begin{array}{ll} 0, \mbox{如果}\ |x| < \frac{10}{9}\widetilde{R}\; \mbox{或}\; |x|>2R, \\ 1, \mbox{如果}\frac{5}{4}\widetilde{R} < |x| < R, \end{array}\right. $

且满足$0 \le \psi_R \le 1$, $\|\nabla^i \psi_R\|_{L^{\infty}(B_{2R}\backslash B_R)} \le \frac{C}{R^i}$$\|\nabla^i \psi_R\|_{L^{\infty}(B_{\frac{5}{4}\widetilde{R}} \backslash B_{\frac{10}{9}\widetilde{R}}) } \le C_{\widetilde{R}}$, $i$=1, 2, 3, 4.再应用引理2.2和$n < \frac{4(p+1)+2\alpha}{p-1}$, 可得

$ \int_{\frac{5}{4}\widetilde{R} < |x| < R} \left (|\Delta u|^2+|x|^{\alpha} |u|^{p+1} \right ) {\rm d}x \le C_{\widetilde{R}}+C R^{n-\frac{4(p+1)+2\alpha}{p-1}} < +\infty, \;\ R \to \infty, $

从而有

$ \int_{{\Bbb R}^n} \left [(\Delta u)^2 +|x|^{\alpha} |u|^{p+1} \right] {\rm d}x < +\infty. $

因此, 在(3.2)式中取极限可得

$ \int_{{\Bbb R}^n} \left [\frac{n-4}{2} |\Delta u|^2-\frac{n+\alpha}{p+1} |x|^{\alpha} |u|^{p+1}\right] {\rm d}x =0. $ (3.10)

再选取$\eta_R \in C^2_0(B_{2R})$, 使得对$|x| < R$, 有$\eta_R(x)=1$, 对$|x|>2R$, 有$\eta_R(x)=0$, 以及$0 \le \eta_R \le 1$且满足

$ \|\nabla^i \eta_R\|_{L^{\infty}} \le \frac{C}{R^i}, ~~i=1, 2. $

$u\eta_R$乘方程1.2且对任意的$R>3\widetilde{R}$, 由引理3.3计算可得

$ \left | \int_{{\Bbb R}^n} \left ( |x|^{\alpha} |u|^{p+1} -(\Delta u)^2\right ) \eta_R {\rm d}x \right |=\left |\int_{{\Bbb R}^n} \left ( u \Delta u \Delta \eta_R+ 2\Delta u\nabla u \cdot \nabla \eta_R \right ) {\rm d}x\right | \\ \le C R^{-2} \int_{B_{2R}\backslash B_{3\widetilde{R}}} |\Delta u||u| {\rm d}x +CR^{-1}\int_{{\Bbb R}^n}|\Delta u|\cdot |\nabla u| {\rm d}x \\ \le CR^{-2} \int_{B_{2R}\backslash B_{3\widetilde{R}}} |x|^{-\frac{4+\alpha}{p-1}-2}|x|^{-\frac{4+\alpha}{p-1}} {\rm d}x + CR^{-1} \int_{{B_{2R}\backslash B_{3\widetilde{R}}}} |x|^{-\frac{4+\alpha}{p-1}-2}|x|^{-\frac{4+\alpha}{p-1}-1} {\rm d}x \\ \le CR^{n-\frac{4(p+1)+2\alpha}{p-1}}. $

由于$n < \frac{4(p+1)+2\alpha}{p-1}$, 则有

$ \int_{{\Bbb R}^n} |\Delta u|^2 {\rm d}x= \int_{{\Bbb R}^n} |x|^{\alpha} |u|^{p+1} {\rm d}x. $ (3.11)

联合(3.10)式和(3.11)式可得

$ \left (\frac{n-4}{2}-\frac{n+\alpha}{p+1} \right )\int_{{\Bbb R}^n} |u|^{p+1} {\rm d}x =0\Longrightarrow u\equiv 0. $

Ⅱ.临界情形: $n=\frac{4(p+1)+2\alpha}{p-1}$.

因为$u$$B_{\widetilde{R}}$外是稳定的, 则$\forall R>3\widetilde{R}$, 类似于次临界情形一样处理可得

$ \int_{B_R \backslash B_{3\widetilde{R}}}\left [(\Delta u)^2 +|x|^{\alpha}|u|^{p+1} \right] {\rm d}x \le C $

$ \int_{{\Bbb R}^n} \left [(\Delta u)^2 +|x|^{\alpha} |u|^{p+1} \right] {\rm d}x < +\infty. $

由椭圆正则性理论可得

$ \lim\limits_{R \to \infty} \int_{B_{2R}\backslash B_R} \left [R^{-1} |\nabla u| +R^{-2} |u|\right] {\rm d}x=0. $

则容易得到

$ \int_{R^n} \left [(\Delta u)^2-|x|^{\alpha} |u|^{p+1}\right] {\rm d}x =0. $

证毕

注 3.1 相比于文献[9, 定理2, 定理3], 我们采用一种新方法证明了相同的结果, 但避免去讨论积分$ \int_{B_R} |x|^{-\frac{2\alpha}{p-1}} {\rm d}x $的可积性.

参考文献
[1] Bahri A, Lions P L. Solutions of superlinear elliptic equations and their Morse indicies. Comm Pure Appl Math , 1992, 45 : 1205–1215. DOI:10.1002/(ISSN)1097-0312
[2] Dancer E N. Stable and finite Morse index solutions on or on bounded domains with small diffusion. Trans Amer Math Soc , 2004, 357 : 1225–1243.
[3] Farina A. On the classification of solutions of Lane-Emden equation on unbounded domains of Rn. J Math Pures Appl , 2007, 87 : 537–561. DOI:10.1016/j.matpur.2007.03.001
[4] Dancer E N, Du Y H, Guo Z M. Finite Morse index solutions of an elliptic equation with supercritical exponent. J Differential Equations , 2011, 250 : 3281–3310. DOI:10.1016/j.jde.2011.02.005
[5] Wang C, Ye D. Some Liouville theorems for Hénon type elliptic equations. J Funct Anal , 2012, 262 : 1750–1727.
[6] Wei J C, Ye D. Liouville Theorems for finite Morse index solutions of Biharmonic problem. Math Ann , 2013, 356 : 1599–1612. DOI:10.1007/s00208-012-0894-x
[7] LiangZ P, SuJ B. Solutions to inhomogeneous quasilinear elliptic problems with concave-convex type nonlinearities. Acta Math Sci , 2014, 34A (2) : 217–226.
[8] Dávila J, Dupaigne L, Wang K L, Wei J C. A monotonicity formula and a Liouville-type theorem for a forth order supercritical problem. Adv Math , 2014, 258 : 240–285. DOI:10.1016/j.aim.2014.02.034
[9] Falzy M, Ghoussoub N. On the Hénon-Lane-Emden conjecture. Disc Cont Dyn Syst , 2014, 34 : 2513–2533.
[10] Souplet Ph. The proof of the Lane-Emden conjecture in four space dimensions. Adv Math , 2009, 221 : 1409–1427. DOI:10.1016/j.aim.2009.02.014
[11] Phan Q H, Souplet Ph. Liouville-type theorems and bounds of solutions of Hardy-Hénon equations. J Differential Equations , 2012, 252 : 2544–2562. DOI:10.1016/j.jde.2011.09.022
[12] Hu L G. Liouville-type theorems for the fourth order nonlinear elliptic equation. J Differential Equation , 2014, 256 : 1817–1746. DOI:10.1016/j.jde.2013.12.001
[13] Polácik P, Quittner P, Souplet Ph. Singularity and decay estimates in superliner problems via Liouville-type theorems I. elliptic equations and systems. Duke Math J , 2007, 139 : 555–579. DOI:10.1215/S0012-7094-07-13935-8