调和映射的Pohozaev恒等式

$\newcommand{\div}{\mathrm{div}\,}$
我们知道调和映射的方程为
$$
\Delta u+A(u)(\nabla u,\nabla u)=0,
$$
其中$u:M^2\to N$是黎曼流形间的映射而$A(u)$是$N\hookrightarrow \mathbb{R}^n$在$u$处的第二基本形式。

我们将用两种办法来证明如下的Pohozaev恒等式。

Theorem 1 (Pohozaev恒等式). 假设$u:M^2\to N$是光滑调和映射,则有
\[
\int_{\partial B_\rho} \lvert u_r \rvert^2=\int_{B_\rho}r^{-2} \lvert u_\theta \rvert^2.
\]


1. 极坐标系下的Pohozaev恒等式 局部地,若有坐标$(x,y)$, 则可以考察极坐标
$$
\begin{cases}
x=r\cos\theta,\\
y=r\sin\theta.
\end{cases}
$$
容易知道,
$$
ds^2=dx^2+dy^2=dr^2+r^2d\theta^2.
$$
单位外法向量为$\nu=\partial_r$. 梯度为
\[
\nabla u=\partial_ru\partial_r+r^{-2}\partial_\theta u\partial_\theta=u_r\partial_r+r^{-1}u_\theta\partial_\theta,
\]
从而
\[
\lvert \nabla u \rvert^2= \lvert \partial_r u \rvert^2+r^{-2} \lvert \partial_\theta u \rvert^2:= \lvert u_r \rvert^2+r^{-2} \lvert u_\theta \rvert^2.
\]

现在, 注意到$ru_r\in \Gamma(u^*TN)$, 因此我们得到
\begin{align*}
0&=\int_{B_\rho}ru_r\cdot \Delta u
=\int_{B_\rho}\div(ru_r\cdot \nabla u)-\int_{B_\rho}\nabla(ru_r)\cdot \nabla u\\
&=\int_{\partial B_\rho}ru_r\cdot \frac{\partial u}{\partial \nu}-\int_{B_\rho}\nabla(ru_r)\cdot \nabla u
=\int_{\partial B_\rho}r \lvert u_r \rvert^2-\int_{B_\rho}\nabla(ru_r)\cdot \nabla u.
\end{align*}
注意到
\begin{align*}
\nabla(r u_r)&=\partial_r(ru_r)\partial_r+r^{-2}\partial_\theta(ru_r)
=(u_r+r\partial_r(u_r))\partial_r+r\cdot r^{-2}\partial_\theta (u_r)\\
&=u_r\partial_r+r\nabla u_r,\\
r \left\langle \nabla u_r,\nabla u \right\rangle
&=r\left( \left\langle u_r, u_{rr} \right\rangle+r^{-2} \left\langle u_\theta, u_{r\theta} \right\rangle \right)\\
\frac{1}{2}r\partial_r \lvert \nabla u \rvert^2
&= \frac{1}{2}r\partial_r\left( \lvert u_r \rvert^2+r^{-2} \lvert u_\theta \rvert^2 \right)\\
&=r\left( \left\langle u_r, u_{rr} \right\rangle+r^{-2} \left\langle u_\theta, u_{r\theta} \right\rangle \right)-r^{-2} \lvert u_\theta \rvert^2\\
&=r \left\langle \nabla u_r,\nabla u \right\rangle-r^{-2} \lvert u_\theta \rvert^2.
\end{align*}

\begin{align*}
\nabla(ru_r)\cdot \nabla u&= \left\langle u_r\partial_r +r\nabla u_r,\nabla u \right\rangle
= \lvert u_r \rvert^2+r \left\langle \nabla u_r,\nabla u \right\rangle\\
&= \lvert u_r \rvert^2+r^{-2} \lvert u_\theta \rvert^2+ \frac{1}{2}r\partial_r \lvert \nabla u \rvert^2\\
&= \lvert \nabla u \rvert^2+ \frac{1}{2}r\partial_r \lvert \nabla u \rvert^2,
\end{align*}
从而,我们得到
\[
\int_{B_\rho}\nabla(ru_r)\cdot \nabla u
=\int_{B_\rho} \lvert \nabla u \rvert^2+ \frac{1}{2}r\partial_r \lvert \nabla u \rvert^2.
\]
注意到
\begin{align*}
\int_{B_\rho}r\partial_r \lvert \nabla u \rvert^2
&=\int_0^\rho\int_{S^1} r\partial_r\lvert \nabla u \rvert^2 rd\theta dr
=\int_0^\rho r^2\partial_r\int_{S^1} \lvert \nabla u \rvert^2 d\theta dr\\
&=\left. \left[ r^2\int_{S^1} \lvert \nabla u \rvert^2d\theta \right] \right|_{0}^\rho-2\int_0^\rho r\int_{S^1} \lvert \nabla u \rvert^2 d\theta dr\\
&=\int_{\partial B_\rho}r \lvert \nabla u \rvert^2 d\sigma-2\int_{B_\rho} \lvert \nabla u \rvert^2.
\end{align*}
因此
\begin{align*}
\int_{B_\rho}\nabla(r u_r)\cdot \nabla u
&=\int_{B_\rho} \lvert \nabla u \rvert^2+ \frac{1}{2}\int_{\partial B_\rho}r \lvert \nabla u \rvert^2 d\sigma-\int_{B_\rho} \lvert \nabla u \rvert^2\\
&= \frac{1}{2}\int_{\partial B_\rho}r \lvert \nabla u \rvert^2d\sigma.
\end{align*}
进而
\begin{align*}
0&=\int_{\partial B_\rho}r \lvert u_r \rvert^2-\int_{B_\rho}\nabla (r u_r)\cdot \nabla u\\
&=\frac{1}{2}\int_{\partial B_\rho}r\left( \lvert u_r \rvert^2- r^{-2} \lvert u_\theta \rvert^2 \right),
\end{align*}

\[
\int_{\partial B_\rho} \lvert u_r \rvert^2=\int_{\partial B_\rho} r^{-2} \lvert u_\theta \rvert^2,
\]
这就是调和映射的Pohozaev恒等式。
2. 柱坐标系下的Pohozaev恒等式 注意,若令$r=e^{-t}$, 则$B_\rho\to (-\infty,-\ln\rho]\times\theta$, $(r,\theta)\mapsto (t,\theta)$. 容易知道$dr=-rdt$, 从而
\[
ds^2=dx^2+dy^2=dr^2+r^2d\theta^2=e^{-2t}\left( dt^2+d\theta^2 \right),
\]
以及
\begin{align*}
ru_r&=r \frac{\partial t}{\partial r}u_t=-u_t,\\
rdr\wedge d\theta&=-r^2 dt\wedge d\theta=-e^{-2t}dt\wedge d\theta,\\
\Delta u&=u_{xx}+u_{yy}= \frac{1}{\sqrt{g}}\partial_\alpha\left( g^{\alpha\beta}\sqrt{g}\partial_\beta u \right)
=e^{2t}\left( u_{tt}+u_{\theta\theta} \right).
\end{align*}
因此
\begin{align*}
0&=\int_{B_\rho}ru_r\cdot \Delta u
=\int_{-\infty}^{-\ln\rho}\int_{S^1} u_t\cdot \left( u_{tt}+u_{\theta\theta} \right) d\theta dt\\
&=\int_{\ln\rho}^\infty\int_{S^1} u_t\cdot \left( u_{tt}+u_{\theta\theta} \right) d\theta dt\\
&=\int_{\ln\rho}^\infty\int_{S^1} \frac{1}{2}\partial_t \lvert u_t \rvert^2-u_\theta\cdot u_{t\theta}+ \int_{\ln\rho}^\infty\left. \left[ u_t\cdot u_\theta \right] \right|_{0}^{2\pi}dt\\
&= \frac{1}{2}\int_{\ln\rho}^\infty\int_0^{2\pi} \partial_t\left( \lvert u_t \rvert^2- \lvert u_\theta \rvert^2 \right).
\end{align*}
由$\rho$的任意性,我们得到
\[
\partial_t\int_{S^1}\left( \lvert u_t \rvert^2- \lvert u_\theta \rvert^2 \right) d\theta=0.
\]
在$[\ln\rho,+\infty)$积分得到
\[
\int_{\partial B_{\rho}}\left( \lvert u_t \rvert^2- \lvert u_\theta \rvert^2 \right) d\sigma=
\int_{S^1}\left( \lvert u_t(\rho,\cdot) \rvert^2- \lvert u_\theta(\rho,\cdot) \rvert^2 \right) \rho d\theta=0,
\]
这里我们使用了$u$光滑,故
\[
\lvert \nabla u \rvert^2\leq C\implies \lvert u_t \rvert^2+ \lvert u_{\theta} \rvert^2\leq C e^{-2t}=C \rho^2,
\]
则表明
\[
\lim_{\rho\to 0}\int_{S^1}\left( \lvert u_t(\rho,\cdot) \rvert^2- \lvert u_\theta(\rho,\cdot) \rvert^2 \right)=0.
\]

Leave a Reply

Your email address will not be published. Required fields are marked *