数学物理学报, 2023, 43(4): 1037-1061

广义地表准地转方程的非对称行波涡对解

吴文菊,1, 范伯全,2,*

1广州大学数学与信息科学学院 广州 510006

2中国科学院数学与系统科学研究院 北京 100190

Traveling Asymmetric Vortex Pairs of the Generalized Surface Quasi-Geostrophic Equation

Wu Wenju,1, Fan Boquan,2,*

1School of Mathematics and Information Science, Guangzhou University, Guangzhou 510006

2Institute of Applied Mathematics, Chinese Academy of Sciences, Beijing 100190

通讯作者: *范伯全, E-mail: 2225409634@qq.com

收稿日期: 2022-09-20   修回日期: 2023-02-14  

Received: 2022-09-20   Revised: 2023-02-14  

作者简介 About authors

吴文菊,E-mail:wuchrysanthemum@163.com

摘要

该文利用涡方法研究广义地表准地转(gSQG)方程具有不同尺度、不同分布的反向旋转涡对的存在性. 利用变分方法构造一族非对称的行波涡对解, 并对该族涡对解的渐近行为进行了分析.

关键词: gSQG方程; 行波解; 涡方法

Abstract

In this paper, we study the existence of counter-rotating vortex pairs with different scales and different distributions for the generalized surface quasi-geostrophic (gSQG) equation by using vorticity method. We construct a family of traveling asymmetric vortex pairs by using the variational method, and analyze the asymptotic behaviors of the family of vortex pairs.

Keywords: The generalized surface quasi-geostrophic equation; Traveling-wave solutions; Vortex method

PDF (492KB) 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

吴文菊, 范伯全. 广义地表准地转方程的非对称行波涡对解[J]. 数学物理学报, 2023, 43(4): 1037-1061

Wu Wenju, Fan Boquan. Traveling Asymmetric Vortex Pairs of the Generalized Surface Quasi-Geostrophic Equation[J]. Acta Mathematica Scientia, 2023, 43(4): 1037-1061

1 简介与主要结果

本文考虑近年来受到广泛关注的广义地表准地转(gSQG)方程

$\begin{matrix} \left\{ \begin{array}{lll} \partial_{t}\theta+{v}\cdot\nabla\theta=0,&(x,t)\in {\mathbb{R} }^{2} \times \mathbb{R} _{+},\\ {v}=\nabla^{\perp}(-\Delta)^{-s}\theta,\\ \end{array} \right.\end{matrix}$

其中$0<s<1$, $v(x,t):\mathbb{R} ^2\times \mathbb{R} _{+}\to \mathbb{R} ^2$ 是速度场, $\theta(x,t):\mathbb{R} ^2\times \mathbb{R} _{+}\to \mathbb{R} $是在自身产生的速度场$v$ 中输运的活动标量, $\nabla^{\perp}=\left(\partial_{2},-\partial_{1}\right)$, $(-\Delta)^{-s}$是分数阶拉普拉斯算子的逆算子, 定义为

$\begin{equation} (-\Delta)^{-s}\theta(x)={\cal G}_s\theta(x)=\int_{\mathbb{R} ^{2}} G_{s}(x-y) \theta(y){\rm d}y,\end{equation}$

其中$G_{s}$是Riesz位势, 由下式给出

$\begin{equation}G_{s}(z)=\frac{c_{s}}{|z|^{2-2s}}, uad c_{s}=\frac{\Gamma(1-s)}{2^{2s} \pi \Gamma(s)}.\end{equation}$

这里的$\Gamma$是伽马函数.

对于$s=\frac{1}{2}$的情形, 方程(1.1)为无粘面准地转(SQG)方程, 用于模拟一般准地转系统中的大气和大气流的温度演变[1-3].对于$s\rightarrow 1$的特殊情形, 可以得到二维不可压缩欧拉方程的涡方程[4].对于$s\rightarrow 0$的极限情形, 得到稳态方程$\partial_{t}\theta=0$, 从而产生稳态解.方程(1.1)与三维不可压缩欧拉方程具有一些相似的数学机制[1].因此, 近年来人们对gSQG方程产生了浓厚的兴趣.

众所周知, 所有径向对称函数$\theta$都是gSQG方程的稳态解.我们在本文中将考虑该方程是否存在其他的全局经典解.据我们所知, 建立gSQG方程经典解的全局适定性, 或者解是否会在有限时间爆破, 仍然是一个公开问题[5-6].

在文献[5]中, Castro等从一个特殊的径向对称函数出发, 利用分歧理论首次构造出了SQG 方程的非平凡全局光滑解, 该光滑解是通过围绕质心以恒定的角速度旋转而演化的匀速旋转解. Gravejat和Smets[7]提出了另一种构造特殊的光滑全局解族的方法.他们利用变分法构造了SQG方程的光滑行波解族, 该结果由Godard-Cadillac在文献[8]中推广到了gSQG 方程. Ao等[9]成功地应用Lyapunov-Schmidt约化方法构造了gSQG方程的旋转经典解和行波经典解.除了光滑的全局解, 人们还对匀速旋转的涡斑进行了广泛的研究, 参见文献[10-12]及其参考文献. 关于gSQG方程的弱解, 有关结果请参阅文献[13-15].最近, Cao 等[17]研究了对称的行波涡对解的存在性, 得到了方程(1.1)的一些新的行波解族.

本文的主要目的是研究具有不同尺度、不同分布的非对称行波涡对的存在性.在说明我们的主要结果之前, 我们首先推导行波解所满足的方程.通过旋转, 可假设这些波在竖直方向上的速度是$-W$, 那么解具有如下的形式

$\begin{equation} \begin{array}{lll} \theta(x, t)=\Theta\left(x_{1}, x_{2}+W t\right). \end{array}\end{equation}$

将(1.4)式代入(1.1)式中的第一个方程, 则该方程被简化为如下稳态方程

$\begin{equation} \begin{array}{lll} \nabla^{\perp}\left({\cal G}_{s} \Theta-W x_{1}\right) \cdot \nabla \Theta=0. \end{array}\end{equation} $

在(1.5)式两边乘上检验函数$\varphi \in C_{0}^{\infty}\left(\mathbb{R} ^{2}\right)$, 形式上的分部积分就可以得到如下弱解的定义.

定义 1.1$s \in[\frac{1}{2},1)$, $\Theta\in{ L_c^\infty}(\mathbb{R} ^2)$, 如果对任意 $\varphi\in C_0^{\infty}(\mathbb{R} ^2)$ 都成立

$\begin{equation} \int_{\mathbb{R} ^2}\Theta\nabla^\perp({\cal G}_s\Theta-Wx_{1})\cdot\nabla\varphi {\rm d}x=0. \end{equation}$

我们就称$\Theta$ 是方程(1.5)的弱解.

根据Riesz位势的正则性理论, 对所有的 $1<p<\infty$, ${\cal G}_{s} \Theta \in \dot{W}^{2s, p}\left(\mathbb{R} ^{2}\right)$, 所以(1.6)式中的积分只有在$s \in[\frac{1}{2},1)$时是有意义的. 如果$0<s <\frac{1}{2}$, ${\cal G}_{s} \Theta$ 的正则性不足以为(1.6)式提供严格意义. 因此, 我们将只考虑方程(1.5)在$s \in[\frac{1}{2},1)$上的弱解.

正如Arnol'd在文献[18]中所提到的, 获得稳态问题(1.5)的解的一种自然的方法, 是让$\Theta$${\cal G}_s\Theta-Wx_{1}$局部函数相关, 即假设存在某个Borel可测函数$f:\mathbb{R} →\mathbb{R} $,使得

$\begin{equation} \Theta=f\left({\cal G}_{s} \Theta-W x_{1}\right).\end{equation}$

本文将采用Arnol'd的思想构造非对称涡对.我们假设

(A) $ f(\tau)=0, \forall\tau \leq 0 ; uad f(\tau)>0, \forall\tau>0.$

$\Pi_{r}=\left\{x \in \mathbb{R} ^{2} \mid x_{1}>0\right\}, \Pi_{\ell}=\left\{x \in \mathbb{R} ^{2} \mid x_{1}<0\right\}. $

本文的主要结果如下.

定理 1.1$\frac{1}{2}\leq s<1$, $\kappa>0$$W>0$, 假设 $f_{r}$$f_{\ell}$ 是两个单调递增的有界连续函数, 且均满足假设 (A). 则存在$\varepsilon_{0}>0$, 使得对所有的 $\varepsilon, \nu \in\left(0, \varepsilon_{0}\right]$, 方程(1.1)有以下形式的行波解

$\theta_{\varepsilon, \nu}(x, t)=\Theta_{\varepsilon, \nu}\left(x_{1}, x_{2}+Wt\right).$

此外, 如下结论成立

(i) $\Theta_{\varepsilon, \nu}$ 具有以下形式$\Theta_{\varepsilon, \nu}=\Theta_{r, \varepsilon}-\Theta_{\ell, \nu}, $这里

$\Theta_{r, \varepsilon}(x)=\frac{1}{\varepsilon^{2}} f_{r}\left({\cal G}_{s} \Theta_{\varepsilon, \nu}-W x_{1}-\mu_{r, \varepsilon}\right), \forall x \in \Pi_{r}, $
$\Theta_{\ell, \nu}(x)=\frac{1}{\nu^{2}} f_{\ell}\left(-{\cal G}_{s} \Theta_{\varepsilon, \nu}+W x_{1}-\mu_{\ell, \nu}\right), \forall x \in \Pi_{\ell},$

其中 $\mu_{r, \varepsilon}$$\mu_{\ell, \nu}$ 均是正常数.

(ii) 存在不依赖于 $\varepsilon, \nu$ 的正常数 $C$, 使得

${\rm diam}\left({\rm supp} \left(\Theta_{r, \varepsilon}\right)\right) \leq C \varepsilon, uad {\rm diam}\left({\rm supp}\left(\Theta_{\ell, \nu}\right)\right) \leq C \nu.$

(iii) 当 $\max \{\varepsilon, \nu\} \rightarrow 0$ 时, 在测度的意义下成立

$\Theta_{r, \varepsilon}(x) \rightharpoonup \kappa {\delta}\left(x-b_{1}\right), uad \Theta_{\ell, \nu}(x) \rightharpoonup \kappa {\delta}\left(x-b_{2}\right),$

其中

$L_{*}=\left(\frac{\kappa}{4 \pi W} \frac{\Gamma(2-s)}{\Gamma(s)}\right)^{\frac{1}{3-2 s}},\ b_{1}=L_{*} {\bf e}_{1},uad b_{2}=-L_{*} {\bf e}_{1},\ {\bf e}_{1}=(1,0).$

(iv) 对于 $\mu_{r, \varepsilon}$$\mu_{\ell, \nu}$

$0<\liminf _{\max \{\varepsilon, \nu\} \rightarrow 0^{+}} \varepsilon^{2-2 s} \mu_{r, \varepsilon} \leq \limsup _{\max \{\varepsilon, \nu\} \rightarrow 0^{+}} \varepsilon^{2-2 s} \mu_{r, \varepsilon}<+\infty, $

$0<\liminf _{\max \{\varepsilon, \nu\} \rightarrow 0^{+}} \nu^{2-2 s} \mu_{\ell, \nu} \leq \limsup _{\max \{\varepsilon, \nu\} \rightarrow 0^{+}} \nu^{2-2 s} \mu_{\ell, \nu}<+\infty.$

注 1.1 如果 $f_{r}$$f_{\ell}$是光滑的, 那么根据标准的椭圆方程正则性理论可知上述行波解也是光滑的. 这里的 $f_{r}$$f_{\ell}$ 可以是不连续的.如果 $f_{r}$$f_{\ell}$ 都是Heaviside函数, 那么上述行波解为非对称涡对形式的涡斑.

注 1.2 如定理1.1所述, 当 $\max \{\varepsilon, \nu\}$ 趋于零时, 上述行波解在测度的意义下趋向于非对称平移点涡对.因此, 上述行波解族构成了一对强度相等、符号相反的点涡对的去奇异化.

2 定理1.1的证明

在本节中, 我们将给出gSQG方程的非对称涡对解的变分构造, 并给出定理1.1的证明.对于行波解的存在性的证明, 我们参考了Cao-Lai-Zhan[19]的对偶变分原理.在文献[19]中对于对称涡对, 可以将问题简化为半平面上的半线性椭圆问题.而本文研究非对称涡对的存在性, 需要在整个平面上考虑, 所以我们需要发展一些技术来摆脱对对称性的依赖, 该想法来源于Cao-Qin-Zhan-Zou[16]关于欧拉方程的非对称涡对解的存在性的研究.对于渐近性的证明部分, 我们参考了Cao-Qin-Zhan-Zou[17]证明gSQG方程的旋转解的渐近性的方法.

2.1 变分问题

首先引入一些符号.用 $\bar{x}=\left(x_{1},-x_{2}\right)$ 表示 $x=\left(x_{1},x_{2}\right)$ 关于 $x_{1}$ 轴对称的点.设 $\kappa$$W$ 是两个固定的正数, 且 $L_{*}=\left(\frac{\kappa}{4 \pi W} \frac{\Gamma(2-s)}{\Gamma(s)}\right)^{\frac{1}{3-2 s}}$, 记$D_{r}=\overline{B_{\frac{L_{*}}{2}}\left(L_{*} {\bf e}_{1}\right)}, uad D_{\ell}=\overline{B_{\frac{L_{*}}{2}}\left(-L_{*} {\bf e}_{1}\right)}.$

我们考虑以下容许类

${\cal A}_{r}=\left\{\Theta \in L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right) \mid \Theta \geq 0, \mbox{ a.e., } {\rm supp}(\Theta) \subseteq D_{r}, \int_{\mathbb{R} ^{2}} \Theta {\rm d} x=\kappa, \Theta(\bar{x})=\Theta(x), \mbox{ a.e. }\right\},$
${\cal A}_{\ell}=\left\{\Theta \in L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right) \mid \Theta \geq 0, \mbox{ a.e., } {\rm supp}(\Theta) \subseteq D_{\ell}, \int_{\mathbb{R} ^{2}} \Theta {\rm d} x=\kappa, \Theta(\bar{x})=\Theta(x), \mbox{ a.e. }\right\}.$

定义流体的动能

${\rm E}_{s}(\Theta):=\frac{1}{2} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} G_{s}\left(x-x^{\prime}\right) \Theta(x) \Theta\left(x^{\prime}\right){\rm d} x {\rm d} x^{\prime},$

及其冲量

${\cal P}(\Theta)=\int_{\mathbb{R} ^{2}} x_{1} \Theta(x) {\rm d}x.$

$f_{r}$$f_{\ell}$ 是两个单调递增的有界连续函数, 且均满足假设(A).定义

$F_{r}(\tau)=\int_{0}^{\tau} f_{r}\left(\tau^{\prime}\right) {\rm d}\tau^{\prime},uad F_{\ell}(\tau)=\int_{0}^{\tau} f_{\ell}\left(\tau^{\prime}\right) {\rm d}\tau^{\prime}.$

容易验证 $F_{r}$$F_{\ell}$ 均是凸函数.从而可以定义 $F_{r}$$F_{\ell}$ 的共轭凸函数(参见文献[20])

$J_{r}(\tau)=\sup\limits_{\tau^{\prime} \in \mathbb{R} }\left[\tau \tau^{\prime}-F_{r}\left(\tau^{\prime}\right)\right],uad J_{\ell}(\tau)=\sup\limits_{\tau^{\prime} \in \mathbb{R} }\left[\tau \tau^{\prime}-F_{\ell}\left(\tau^{\prime}\right)\right].$

注意到 $J_{r}$$J_{\ell}$ 均是 $\mathbb{R} $ 上的下半连续的凸函数. 设 $\varepsilon$$\nu$ 是两个正数, 我们定义

${\cal J}_{r, \varepsilon}\left(\Theta_{r}\right):=\frac{1}{\varepsilon^{2}} \int_{\mathbb{R} ^{2}} J_{r}\left(\varepsilon^{2} \Theta_{r}\right), uad \Theta_{r} \in {\cal A}_{r},$
${\cal J}_{\ell, \nu}\left(\Theta_{\ell}\right):=\frac{1}{\nu^{2}} \int_{\mathbb{R} ^{2}} J_{\ell}\left(\nu^{2} \Theta_{\ell}\right), uad \Theta_{\ell} \in {\cal A}_{\ell}.$

为了保证 ${\cal J}_{r, \varepsilon}$${\cal J}_{\ell, \nu}$ 不恒等于 $+\infty $, 我们假设

$\begin{equation} \frac{f_{r}\left(1\right)}{2 \varepsilon^{2}} {\rm meas}\left(D_{r}\right)+\frac{f_{\ell}\left(1\right)}{2 \nu^{2}} {\rm meas}\left(D_{\ell}\right)>\kappa,\end{equation}$

其中 ${\rm meas}\left(K\right)$ 表示 $K$ 的二维勒贝格测度.

定义能量泛函

${\cal E}_{\varepsilon, \nu}\left(\Theta _{r}, \Theta _{\ell}\right)=E_{s}\left(\Theta _{r}-\Theta _{\ell}\right)-W {\cal P}\left(\Theta _{r}-\Theta _{\ell}\right)-{\cal J}_{r, \varepsilon}\left(\Theta _{r}\right)-{\cal J}_{\ell, \nu}\left(\Theta _{\ell}\right), \ \left(\Theta _{r}, \Theta _{\ell}\right) \in {\cal A}_{r} \times {\cal A}_{\ell}.$

下面我们研究以下极大化问题

$\begin{equation} {\cal C}_{\varepsilon, \nu}:=\sup \left\{{\cal E}_{\varepsilon, \nu}\left(\Theta _{r}, \Theta _{\ell}\right):\left(\Theta _{r}, \Theta _{\ell}\right) \in {\cal A}_{r} \times {\cal A}_{\ell}\right\}.\end{equation}$

注意到能量泛函 ${\cal E}_{\varepsilon, \nu}$${\cal A}_{r} \times {\cal A}_{\ell}$ 上可以取到 $-\infty $, 并且它不是 $C^1$ 光滑的.这导致了一些技术性的困难, 下面我们将发展一种摄动的方法来克服这些困难.

2.2 摄动问题

本节我们将对极大化问题(2.2)进行微小扰动, 并研究由此产生的摄动问题.

$\rho$ 是一个充分小的正数, 定义

$f_{r, \rho}(\tau)=f_{r}(\tau)+\rho\tau_{+}^{s}, uad f_{\ell, \rho}(\tau)=f_{\ell}(\tau)+\rho \tau_{+}^{s}, $
$F_{r, \rho}(\tau)=\int_{0}^{\tau} f_{r, \rho}\left(\tau^{\prime}\right) {\rm d}\tau^{\prime}, uad F_{\ell, \rho}(\tau)=\int_{0}^{\tau} f_{\ell, \rho}\left(\tau^{\prime}\right) {\rm d}\tau^{\prime}.$

显然 $f_{r, \rho}$$f_{\ell, \rho}$ 均是非负的, 单增的, 无界的连续函数.

同样地, 我们可以用 $J_{r, \rho}$$J_{\ell, \rho}$ 分别表示 $F_{r, \rho}$$F_{\ell, \rho}$ 的共轭凸函数, 从而$J_{r, \rho}$$J_{\ell, \rho}$ 均是 $\mathbb{R} $ 上的下半连续的凸函数. 通过简单的计算知, 当 $\tau<0$ 时, $J_{r, \rho}$, $J_{\ell, \rho}(\tau)=+\infty$. 为了方便, 我们重新定义 $J_{r, \rho}$$J_{\ell, \rho}$ 在负半轴上的值恒为0.根据 $J_{r, \rho}$$J_{\ell, \rho}$ 的定义, 容易验证它们均是连续可微的.令

${\cal J}_{r, \varepsilon, \rho}\left(\Theta _{r}\right)=\frac{1}{\varepsilon^{2}} \int_{\mathbb{R} ^{2}} J_{r, \rho}\left(\varepsilon^{2} \Theta _{r}\right) {\rm d} x, uad \Theta _{r} \in {\cal A}_{r}, $
${\cal J}_{\ell, \nu, \rho}\left(\Theta _{\ell}\right)=\frac{1}{\nu^{2}} \int_{\mathbb{R} ^{2}} J_{\ell, \rho}\left(\nu^{2} \Theta _{\ell}\right) {\rm d} x, uad \Theta _{\ell} \in {\cal A}_{\ell}.$

类似地, 定义扰动后的能量泛函为

${\cal E}_{\varepsilon, \nu, \rho}\left(\Theta _{r}, \Theta _{\ell}\right)=E_{s}\left(\Theta _{r}-\Theta _{\ell}\right)-W {\cal P}\left(\Theta _{r}-\Theta _{\ell}\right)-{\cal J}_{r, \varepsilon, \rho}\left(\Theta _{r}\right)-{\cal J}_{\ell, \nu, \rho}\left(\Theta _{\ell}\right), \ \left(\Theta _{r}, \Theta _{\ell}\right) \in {\cal A}_{r} \times {\cal A}_{\ell}.$

相应的极大值问题为

$\begin{equation} {\cal C}_{\varepsilon, \nu,\rho}:=\sup \left\{{\cal E}_{\varepsilon, \nu,\rho}\left(\Theta _{r}, \Theta _{\ell}\right):\left(\Theta _{r}, \Theta _{\ell}\right) \in {\cal A}_{r} \times {\cal A}_{\ell}\right\}.\end{equation}$

我们将把极大值问题(2.2)作为摄动问题(2.3)在 $\rho\rightarrow 0$ 的极限来处理.

通过直接计算, 可得如下引理.

引理 2.1 存在常数 $C_{s}>0$, 对任意的 $\Theta \in L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)$, 且满足${\rm supp}(\Theta) \subseteq B_{4L_{*}}(0)$, 成立

$\|{\cal G}_{s} \Theta \|_{L^{\infty}\left(B_{4 L_{*}}(0)\right)} \leq C_{s}\|\Theta \|_{L^{1+\frac{1}{s} }\left(B_{4 L_{*}}(0)\right)}.$

对任意的 $\Theta \in L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)$ 以及任意的 $x \in B_{4L_{*}}(0)$,利用 Hölder 不等式可得

$\begin{matrix}\left|{\cal G}_{s} \Theta \right|&=&\bigg|\int_{B_{4L_{*}}(0)} \frac{c_{s}}{\left|x-x^{\prime}\right|^{2-2s}}\Theta \left(x^{\prime}\right){\rm d}x^{\prime}\bigg|\\&\le&C_{s}\bigg[\int_{B_{4L_{*}}(0)}\bigg(\frac{1}{\left|x-x^{\prime}\right|^{2-2s}}\bigg)^{s+1}{\rm d}x^{\prime}\bigg]^{\frac{1}{s+1}} \|\Theta \|_{L^{1+\frac{1}{s} }\left(B_{4 L_{*}}(0)\right)}\\&\le&C_{s}\bigg[\int_{B_{8L_{*}}(x)}\frac{1}{\left|x-x^{\prime}\right|^{2-2s^{2}}}{\rm d}x^{\prime}\bigg]^{\frac{1}{s+1}} \|\Theta \|_{L^{1+\frac{1}{s} }\left(B_{4 L_{*}}(0)\right)}\\&=&C_{s}\bigg(\int_{B_{8L_{*}}(0)} \frac{1}{\left|y\right|^{2-2s^2}}{\rm d}y\bigg)^{\frac{1}{s+1}} \|\Theta \|_{L^{1+\frac{1}{s} }\left(B_{4 L_{*}}(0)\right)}\\&\le&C_{s}\|\Theta \|_{L^{1+\frac{1}{s} }\left(B_{4 L_{*}}(0)\right)}.\end{matrix}$

因此, 有$\|{\cal G}_{s} \Theta \|_{L^{\infty}\left(B_{4 L_{*}}(0)\right)} \leq C_{s}\|\Theta \|_{L^{1+\frac{1}{s} }\left(B_{4L_{*}}(0)\right)}.$ 证毕.

我们可以用直接法证明摄动问题(2.3)的极大元的存在性.

引理 2.2 存在 $\left(\Theta _{r, \varepsilon, \rho}, \Theta _{\ell, \nu, \rho}\right) \in {\cal A}_{r} \times {\cal A}_{\ell}$ 使得

${\cal E}_{\varepsilon, \nu, \rho}\left(\Theta _{r, \varepsilon, \rho}, \Theta _{\ell, \nu, \rho}\right)=\max _{{\cal A}_{r} \times {\cal A}_{\ell}} {\cal E}_{\varepsilon, \nu, \rho} \in \mathbb{R}.$

首先说明 ${\cal E}_{\varepsilon, \nu, \rho}$${\cal A}_{r} \times {\cal A}_{\ell}$ 上有上界. 对任意的 $\left(\Theta_{r}, \Theta_{\ell}\right) \in {\cal A}_{r} \times {\cal A}_{\ell}$,由引理2.1和 Young 不等式可得

$\begin{matrix} E_{s}\left(\Theta _{r}-\Theta _{\ell}\right) &=&\frac{1}{2} \int_{\mathbb{R} ^{2}}\left[\Theta _{r}(x)-\Theta _{\ell}(x)\right] {\cal G}_{s}\left(\Theta _{r}-\Theta _{\ell}\right)(x) {\rm d}x \\&\le&\frac{C_{s}}{2}\left\|\Theta _{r}-\Theta _{\ell}\right\|_{L^{1}\left(B_{4L_{*}}(0)\right)} \left\|\Theta _{r}-\Theta _{\ell}\right\|_{L^{1+\frac{1}{s}}\left(B_{4L_{*}}(0)\right)}\\&\le&{C_{s}}{\kappa }(\left\|\Theta _{r}\right\|_{L^{1+\frac{1}{s}}\left(B_{4L_{*}}(0)\right)}+\left\|\Theta _{\ell}\right\|_{L^{1+\frac{1}{s}}\left(B_{4L_{*}}(0)\right)} )\\&\le&\frac{4^{s}}{s+1} C_{s}^{1+s} \kappa^{1+s}\left(\varepsilon^{-2}+\nu^{-2}\right) \rho+\frac{\varepsilon^{\frac{2}{s} }}{4(1+\frac{1}{s}) \rho^{\frac{1}{s}}}\left\|\Theta _{r}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}\\&& +\frac{\nu^{\frac{2}{s} }}{4(1+\frac{1}{s}) \rho^{\frac{1}{s}}}\left\|\Theta _{\ell}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}.\end{matrix}$

又因为 ${\rm supp}(\Theta_{r}) \subseteq D_{r}$, ${\rm supp}(\Theta_{\ell}) \subseteq D_{\ell}$, 所以

${\cal P}\left(\Theta _{r}-\Theta _{\ell}\right)=\int_{\mathbb{R} ^{2}}x_{1}\left(\Theta _{r}-\Theta _{\ell}\right)(x){\rm d}x=\int_{D_{r}}x_{1}\Theta _{r}(x){\rm d}x-\int_{D_{\ell}}x_{1}\Theta _{\ell}(x){\rm d}x,$

从而有

$\begin{equation}L_{*}\kappa \leq {\cal P}\left(\Theta _{r}-\Theta _{\ell}\right) \leq 3L_{*}\kappa.\end{equation}$

此外, 不妨设 $ \Lambda_{r}:=\sup\limits_{\mathbb{R} } f_{r} $, $ \Lambda_{\ell}:=\sup\limits_{\mathbb{R} } f_{\ell} $, 那么

$\begin{equation} \rho {\tau_{+}^{s}} \leq f_{r, \rho}(\tau) \leq \Lambda_{r}+\rho {\tau_{+}^{s}}, uad \rho {\tau_{+}^{s}} \leq f_{\ell, \rho}(\tau) \leq \Lambda_{\ell}+\rho {\tau_{+}^{s}}, uad \forall \tau \geq 0.\end{equation}$

从而有

$\begin{equation} \rho {\frac{\tau_{+}^{s+1}}{s+1}} \leq F_{r, \rho}(\tau) \leq \Lambda_{r}{\tau_{+}}+\rho {\frac{\tau_{+}^{s+1}}{s+1}}, uad \rho {\frac{\tau_{+}^{s+1}}{s+1}} \leq F_{\ell, \rho}(\tau) \leq \Lambda_{\ell}{\tau_{+}}+\rho {\frac{\tau_{+}^{s+1}}{s+1}}, uad \forall \tau \geq 0.\end{equation}$

进而对 $\forall \tau \geq 0$

$\begin{equation}\frac{\left(\tau-\Lambda_{r}\right)_{+}^{1+\frac{1}{s}}}{\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}} \leq J_{r, \rho}(\tau) \leq \frac{\tau_{+}^{1+\frac{1}{s}}}{\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}}, \\uad \frac{\left(\tau-\Lambda_{\ell}\right)_{+}^{1+\frac{1}{s}}}{\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}} \leq J_{\ell, \rho}(\tau) \leq \frac{\tau_{+}^{1+\frac{1}{s}}}{\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}}.\end{equation}$

因此, 存在常数 $c_{0}>0$, 使得

$\begin{equation} J_{r, \rho}(\tau), uad J_{\ell, \rho}(\tau) \geq \frac{\tau_{+}^{1+\frac{1}{s}}}{2\left(1+\frac{1}{s}\right)\rho^{\frac{1}{s}}}-\frac{c_{0}}{\rho^{\frac{1}{s}}}, uad \forall \tau \geq 0.\end{equation}$

根据(2.9), (2.10)式得

$\begin{equation} \frac{\varepsilon^{\frac{2}{s}}}{2\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}}\left\|\Theta_{r}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}-\frac{c_{0} {\rm meas}\left(D_{r}\right)}{\varepsilon^{2} \rho^{\frac{1}{s}}} \leq {\cal J}_{r, \varepsilon, \rho}\left(\Theta_{r}\right) \leq \frac{\varepsilon^{\frac{2}{s}}}{\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}}\left\|\Theta_{r}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}\end{equation}$

$ \begin{equation} \frac{\nu^{\frac{2}{s}}}{2\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}}\left\|\Theta_{\ell}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}-\frac{c_{0} {\rm meas}\left(D_{\ell}\right)}{\nu^{2} \rho^{\frac{1}{s}}} \leq {\cal J}_{\ell, \nu, \rho}\left(\Theta_{\ell}\right) \leq \frac{\nu^{\frac{2}{s}}}{\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}}\left\|\Theta_{\ell}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}, \end{equation}$

结合(2.5), (2.6), (2.11)和(2.12)式可得

$-\infty<{\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}, \Theta_{\ell}\right) \leq C_{s,\varepsilon, \nu, \rho}, uad \forall\left(\Theta_{r}, \Theta_{\ell}\right) \in {\cal A}_{r} \times {\cal A}_{\ell},$

这里的 $C_{s,\varepsilon, \nu, \rho}$ 表示依赖于 $s,\varepsilon, \nu, \rho$ 的正常数.这说明了 $\sup\limits_{{\cal A}_{r} \times {\cal A}_{\ell}} {\cal E}_{\varepsilon, \nu, \rho} \in \mathbb{R} $.从而我们可以取极大化序列 $\left\{\left(\Theta_{r}^j,\Theta_{\ell}^j\right)\right\}_{j\in{{\Bbb N}_{+}}} \subseteq {\cal A}_{r} \times {\cal A}_{\ell} $ 使得

$\lim_{j \to \infty}{\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^j,\Theta_{\ell}^j\right)=\sup _{{\cal A}_{r} \times {\cal A}_{\ell}} {\cal E}_{\varepsilon, \nu, \rho}.$

我们断言 $\left\{\left(\Theta_{r}^j,\Theta_{\ell}^j\right)\right\}_{j\in{{\Bbb N}_{+}}}$${\cal A}_{r} \times {\cal A}_{\ell} $ 中的有界序列.事实上, 由 ${\cal E}_{\varepsilon, \nu, \rho}$ 的定义以及(2.6)式, 有

$\begin{equation} {\cal J}_{r, \varepsilon, \rho}\left(\Theta_{r}^j\right)+{\cal J}_{\ell, \nu, \rho}\left(\Theta_{\ell}^j\right) \leq E_{s}\left(\Theta _{r}^j-\Theta _{\ell}^j\right)-{\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^j,\Theta_{\ell}^j\right).\end{equation}$

结合(2.5), (2.11), (2.12)和(2.13)式, 有

$\begin{eqnarray*} &&\frac{\varepsilon^{\frac{2}{s}}}{2\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}}\left\|\Theta_{r}^{j}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}+\frac{\nu^{\frac{2}{s}}}{2\left({1+\frac{1}{s}}\right) \rho^{\frac{1}{s}}}\left\|\Theta_{\ell}^{j}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}-\frac{c_{0} {\rm meas}\left(D_{r}\right)}{\varepsilon^{2} \rho^{\frac{1}{s}}}-\frac{c_{0} {\rm meas}\left(D_{\ell}\right)}{\nu^{2} \rho^{\frac{1}{s}}}\\&\leq& \frac{4^{s}}{s+1} C_{s}^{1+s} \kappa^{1+s}\left(\varepsilon^{-2}+\nu^{-2}\right) \rho+\frac{\varepsilon^{\frac{2}{s} }}{4(1+\frac{1}{s}) \rho^{\frac{1}{s}}}\left\|\Theta_{r}^j\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}\\&&+\frac{\nu^{\frac{2}{s} }}{4(1+\frac{1}{s}) \rho^{\frac{1}{s}}}\left\|\Theta _{\ell}^j\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}-{\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^j,\Theta_{\ell}^j\right).\end{eqnarray*}$

$\left\|\Theta_{r}^{j}\right\|_{L^{1+\frac{1}{s}\left(\mathbb{R}^{2}\right)}}^{1+\frac{1}{s}}+\left\|\Theta_{\ell}^{j}\right\|_{L^{1+\frac{1}{s}\left(\mathbb{R}^{2}\right)}}^{1+\frac{1}{s}} \leq C_{s, \varepsilon, \nu}\left[1+\rho^{1+\frac{1}{s}}-\rho^{\frac{1}{s}} \mathcal{E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^{j}, \Theta_{\ell}^{j}\right)\right]$

因此, 选取 $\left(\Theta_{r}^0,\Theta_{\ell}^0\right) \in {\cal A}_{r} \times {\cal A}_{\ell} $ 使得 ${\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^0,\Theta_{\ell}^0\right) \leq {\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^j,\Theta_{\ell}^j\right)+1$, 那么有

$\left\|\Theta_{r}^{j}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R}^{2}\right)}^{1+\frac{1}{s}}+\left\|\Theta_{\ell}^{j}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R}^{2}\right)}^{1+\frac{1}{s}} \leq C_{s, \varepsilon, \nu}\left[1+\rho^{1+\frac{1}{s}}-\rho^{\frac{1}{s}} \mathcal{E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^{0}, \Theta_{\ell}^{0}\right)\right]$

其中 $C_{s,\varepsilon,\nu}$ 是依赖于 $s$, $\varepsilon$, $\nu$, 但不依赖于 $\rho$ 的正常数. 这说明了 $\left\{\left(\Theta_{r}^j,\Theta_{\ell}^j\right)\right\}_{j\in{{\Bbb N}_{+}}}$$L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right) \times L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)$ 中有界, 所以存在子序列 $\left\{\left(\Theta_{r}^{j_k},\Theta_{\ell}^{j_k}\right)\right\}_{k\in{{\Bbb N}_{+}}}$$\left(\Theta_{r,\varepsilon,\rho},\Theta_{\ell,\nu,\rho}\right)\in L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right) \times L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)$, 使得当 $k\rightarrow +\infty $ 时, 在 $L^{1+\frac{1}{s}}$ 的弱拓扑下有

$\Theta_{r}^{j_k}\rightharpoonup\Theta_{r,\varepsilon,\rho}, uad \Theta_{\ell}^{j_k}\rightharpoonup\Theta_{\ell,\nu,\rho}.$

接下来证明 $\left(\Theta_{r,\varepsilon,\rho},\Theta_{\ell,\nu,\rho}\right)\in{\cal A}_{r} \times {\cal A}_{\ell}$, 这里我们只证明 $\Theta_{r,\varepsilon,\rho}\in{\cal A}_{r}$,$\Theta_{\ell,\nu,\rho}\in{\cal A}_{\ell}$ 类似可证. 即要证$\Theta_{r,\varepsilon,\rho}$ 满足以下四点.

(1) $\Theta_{r,\varepsilon,\rho} \geq 0$, a.e..

我们采用反证法. 假设 ${\rm meas}\left(\{\Theta_{r,\varepsilon,\rho}<0\}\right)>0$, 则存在一个 $n_{0}\in {\Bbb N}_{+}$, 使得

${\rm meas}\left(\{\Theta_{r,\varepsilon,\rho}<-\frac{1}{n_{0}}\}\right)>0, $

取集合 $\{\Theta_{r,\varepsilon,\rho}<-\frac{1}{n_{0}}\}$ 的特征函数 $\phi=I_{\{\Theta_{r,\varepsilon,\rho}<-\frac{1}{n_{0}}\}}$, 则

$\begin{equation} \int_{\mathbb{R} ^{2}}\Theta_{r}^{j_k}\phi {\rm d}x\ge 0.\end{equation}$

$k\to\infty$, 由 $\Theta_{r}^{j_k}$ 的弱收敛性得

$\begin{equation}\int_{\mathbb{R} ^{2}}\Theta_{r,\varepsilon,\rho}\phi {\rm d}x \ge 0.\end{equation}$

另一方面

$\begin{equation}\int_{\mathbb{R} ^{2}}\Theta_{r,\varepsilon,\rho}\phi {\rm d}x=\int_{{\{\Theta_{r,\varepsilon,\rho}<-\frac{1}{n_{0}}\}}}\Theta_{r,\varepsilon,\rho} {\rm d}x<(-\frac{1}{n_{0}}){\rm meas}\left(\{\Theta_{r,\varepsilon,\rho}<-\frac{1}{n_{0}}\}\right)<0,\end{equation}$

然而这导致了矛盾.因此我们得到了 $\Theta_{r,\varepsilon,\rho} \geq 0$, a.e..

(2) ${\rm supp}(\Theta_{r,\varepsilon,\rho}) \subseteq D_{r}$.

${\rm supp}(\Theta_{r}^{j_k}) \subseteq D_{r}$ 以及 $\Theta_{r}^{j_k}$ 的弱收敛性, 易得 ${\rm supp}(\Theta_{r,\varepsilon,\rho}) \subseteq D_{r}$.

(3) $\int_{\mathbb{R} ^{2}} \Theta_{r,\varepsilon,\rho} {\rm d}x=\kappa$.

这个结论可由 $\int_{\mathbb{R} ^{2}} \Theta_{r}^{j_k} {\rm d}x=\kappa$ 以及 $\Theta_{r}^{j_k}$ 的弱收敛性立即得到.

(4) $\Theta_{r,\varepsilon,\rho}(\bar{x})=\Theta_{r,\varepsilon,\rho}(x)$, a.e..

对任意的 $\phi\in C_{0}^\infty(\mathbb{R} ^{2})$, 我们有

$\begin{matrix}\int_{\mathbb{R} ^{2}}\Theta_{r,\varepsilon,\rho}(x_{1},-x_{2})\phi(x_{1},x_{2}){\rm d}x&=&\int_{\mathbb{R} ^{2}}\Theta_{r,\varepsilon,\rho}(x_{1},x_{2})\phi(x_{1},-x_{2}){\rm d}x\\&=&\lim_{k\to\infty}\int_{\mathbb{R} ^{2}}\Theta_{r}^{j_k}(x_{1},x_{2})\phi(x_{1},-x_{2}){\rm d}x\\&=&\lim_{k\to\infty}\int_{\mathbb{R} ^{2}}\Theta_{r}^{j_k}(x_{1},-x_{2})\phi(x_{1},x_{2}){\rm d}x\\&=&\lim_{k\to\infty}\int_{\mathbb{R} ^{2}}\Theta_{r}^{j_k}(x_{1},x_{2})\phi(x_{1},x_{2}){\rm d}x\\&=&\int_{\mathbb{R} ^{2}}\Theta_{r,\varepsilon,\rho}(x_{1},x_{2})\phi(x_{1},x_{2}){\rm d}x,\end{matrix}$

因此 $\Theta_{r,\varepsilon,\rho}(x_{1},-x_{2})= \Theta_{r,\varepsilon,\rho}(x_{1},x_{2})$, a.e..

最后, 只需说明 ${\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r,\varepsilon,\rho},\Theta_{\ell,\nu,\rho}\right)=\sup\limits_{{\cal A}_{r} \times {\cal A}_{\ell}} {\cal E}_{\varepsilon, \nu, \rho}$. 因为$G_{s}(\cdot,\cdot)\in L^{1+s}(\mathbb{R} ^{2}\times \mathbb{R} ^{2})$ 以及 $x_{1}\in L^{1+s}(\mathbb{R} ^{2})$, 则有

$\begin{equation} \lim_{k \rightarrow+\infty}{\rm E}_{s}\left(\Theta_{r}^{j_k}-\Theta_{\ell}^{j_k}\right)=\lim _{k \rightarrow+\infty}{\rm E}_{s}\left(\Theta_{r,\varepsilon,\rho}-\Theta_{\ell,\nu,\rho}\right).\end{equation}$
$\begin{equation} \lim_{k \rightarrow+\infty} {\cal P}\left(\Theta_{r}^{j_k}-\Theta_{\ell}^{j_k}\right)=\lim _{k \rightarrow+\infty} {\cal P}\left(\Theta_{r,\varepsilon,\rho}-\Theta_{\ell,\nu,\rho}\right).\end{equation}$

另一方面, 根据 $J_{r, \rho}$$J_{\ell, \rho}$ 的下半连续性, 有

$\begin{equation} \liminf _{k \rightarrow+\infty} {\cal J}_{r, \varepsilon, \rho}\left(\Theta _{r}^{j_k}\right) \geq {\cal J}_{r, \varepsilon, \rho}\left(\Theta _{r, \varepsilon, \rho}\right), uad \liminf _{k \rightarrow+\infty} {\cal J}_{\ell, \nu, \rho}\left(\Theta _{\ell}^{j_k}\right) \geq {\cal J}_{\ell, \nu, \rho}\left(\Theta _{\ell, \nu, \rho}\right).\end{equation} $

因此, 结合(2.19), (2.20), (2.21)式得

${\cal E}_{\varepsilon, \nu, \rho}\left(\Theta _{r, \varepsilon, \rho},\Theta _{\ell, \nu, \rho}\right)\geq\sup _{{\cal A}_{r} \times {\cal A}_{\ell}} {\cal E}_{\varepsilon, \nu, \rho}.$

从而容易得到

${\cal E}_{\varepsilon, \nu, \rho}\left(\Theta _{r, \varepsilon, \rho},\Theta _{\ell, \nu, \rho}\right)=\sup\limits_{{\cal A}_{r} \times {\cal A}_{\ell}} {\cal E}_{\varepsilon, \nu, \rho}\in \mathbb{R} $. 证毕.

现在假设 $\left(\Theta _{r, \varepsilon, \rho},\Theta _{\ell, \nu, \rho}\right)$ 为变分问题(2.3)的解. 为了估计一些与 $\left(\Theta _{r, \varepsilon, \rho},\Theta _{\ell, \nu, \rho}\right)$ 相关的量,我们通过选取适当的检验函数来得到变分问题(2.3)对应的欧拉—拉格朗日方程.

引理 2.3 存在两个拉格朗日乘子$\mu_{r, \varepsilon, \rho}$, $\mu _{\ell, \nu, \rho}\in \mathbb{R} $, 使得

$\begin{equation} \Theta _{r, \varepsilon, \rho}(x)=\frac{1}{\varepsilon^{2}} f_{r, \rho}\left({\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)(x)-W x_{1}-\mu_{r, \varepsilon, \rho}\right), uad \mbox{ a.e. } x \in D_{r},\end{equation} $
$\begin{equation} \Theta _{\ell, \nu, \rho}(x)=\frac{1}{\nu^{2}} f_{\ell, \rho}\left(-{\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)(x)+W x_{1}-\mu_{\ell, \nu, \rho}\right), uad \mbox{ a.e. } x \in D_{\ell}. \end{equation}$

我们只证明(2.22)式, 对于(2.23)式类似可证.取 $\delta_0>0$ 使得

$S_{\delta_{0}}=\left\{x \in D_{r} \mid \Theta _{r, \varepsilon, \rho}(x) \geq \delta_{0}\right\}$

具有正测度, 则存在函数 $\phi \in L^{\infty}\left ( \mathbb{R} ^{2} \right )$, 满足

$ \begin{matrix} \left\{ \begin{array}{lll}\int_{\mathbb{R} ^{2}}\phi(x){\rm d}x=1,\\[2mm] \phi(\bar{x})=\phi(x),\\ {\rm supp}(\phi) \subseteq S_{\delta_0}. \end{array} \right. \end{matrix} $

$0<\delta<\delta_0$, 记 $S_{\delta}=\left\{x \in D_{r} \mid \Theta _{r, \varepsilon, \rho}(x) \geq \delta\right\}$.对任意给定的函数 $\Theta \in L^{\infty }\left ( \mathbb{R} ^{2} \right )$, 满足

$\begin{matrix} \left\{ \begin{array}{lll} \Theta \geq 0 uad \forall x\in D_{r} \backslash S_{\delta},\\ \Theta(\bar{x})=\Theta(x),\\ {\rm supp}(\Theta) \subseteq D_{r}. \end{array} \right.\end{matrix}$

定义 $\Theta _{r, \varepsilon, \rho}$ 的一个扰动

$\Theta _{\tau}=\Theta _{r,\varepsilon, \rho }+\tau\left ( \Theta -\phi\int_{\mathbb{R} ^{2}}\Theta (x){\rm d}x\right ), \tau>0.$

则可以验证当 $\tau$ 充分小时(比如 $0<\tau\leq \frac{4\delta}{4{\Vert \Theta\Vert_{\infty}}+\pi L_{*}^{2}{\Vert \phi\Vert_{\infty}}{\Vert \Theta\Vert_{\infty}}}$), 有 $\Theta _{\tau}\in {\cal A}_{r}$. 因此可得

$\begin{matrix}0 & \geq&\left.\frac{\rm d}{{\rm d}\tau}\right|_{\tau=0^{+}} {\cal E}_{\varepsilon, \nu, \rho}\left(\Theta _{\tau}, \Theta _{\ell, \nu, \rho}\right) \\&=&\int_{\mathbb{R} ^{2}}\left[{\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)-W x_{1}-\mu_{r, \varepsilon, \rho}-J_{r, \rho}^{\prime}\left(\varepsilon^{2} \Theta _{r, \varepsilon, \rho}\right)\right] \Theta (x) {\rm d}x.\end{matrix}$

其中

$\mu_{r, \varepsilon, \rho}=\int_{\mathbb{R} ^{2}}\left[{\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)-W x_{1}-J_{r, \rho}^{\prime}\left(\varepsilon^{2} \Theta _{r, \varepsilon, \rho}\right)\right] \phi(x) {\rm d}x.$

定义

$\begin{equation}\psi_{r, \varepsilon, \rho}:={\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)-W x_{1}-\mu_{r, \varepsilon, \rho}.\end{equation}$

从而有

$\int_{\mathbb{R} ^{2}}\left[\psi_{r, \varepsilon, \rho}-J_{r, \rho}^{\prime}\left(\varepsilon^{2} \Theta _{r, \varepsilon, \rho}\right)\right] \Theta (x) {\rm d}x \le 0.$

根据 $\Theta$ 的任意性可得

$\psi_{r, \varepsilon, \rho}=J_{r, \rho}^{\prime}\left(\varepsilon^{2} \Theta _{r, \varepsilon, \rho}\right), uad x\in S_{\delta},uad \mbox{且} uad\psi_{r, \varepsilon, \rho} \leq J_{r, \rho}^{\prime}\left(\varepsilon^{2} \Theta _{r, \varepsilon, \rho}\right), uad x\in D_{r} \backslash S_{\delta}.$

$\delta \rightarrow 0^{+}$, 则

$\psi_{r, \varepsilon, \rho}=J_{r, \rho}^{\prime}\left(\varepsilon^{2} \Theta _{r, \varepsilon, \rho}\right), uad x\in\left\{x \in D_{r} \mid \Theta _{r, \varepsilon, \rho}(x)>0\right\}.$

$\psi_{r, \varepsilon, \rho} \leq 0, uad x\in\left\{x \in D_{r} \mid \Theta_{r, \varepsilon, \rho}(x)=0\right\}.$ 因为 $J_{r, \rho}^{\prime}$$f_{r, \rho}$ 互为逆映射, 所以有 $\Theta _{r, \varepsilon, \rho}(x)=\frac{1}{\varepsilon^{2}} f_{r, \rho}\left(\psi_{r, \varepsilon, \rho} \right), uad \mbox{ a.e. } x \in D_{r}.$

故(2.22)式成立. 证毕.

2.3 一致有界性

我们希望通过计算变分问题(2.3)的解在$\rho \rightarrow 0$ 的极限值, 来得到变分问题(2.2)的解.为此, 需要先建立一些关于 $\rho \rightarrow 0$ 时的一致有界性.为了方便, 记

$\psi_{r, \varepsilon, \rho}(x)={\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)(x)-W x_{1}-\mu_{r, \varepsilon, \rho},$$\psi_{\ell, \nu, \rho}(x)=-{\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)(x)+W x_{1}-\mu_{\ell, \nu, \rho}.$

引理 2.4 存在不依赖于 $\rho$ 的正常数 $C$, 使得对所有的 $\rho \in\left ( 0,1 \right )$

$\begin{eqnarray*}&&\left|\mu_{r, \varepsilon, \rho}\right|+\left|\mu_{\ell, \nu, \rho}\right|+\left\|\psi_{r, \varepsilon, \rho}\right\|_{L^{\infty }\left(B_{4 L_{*}}(0)\right)}+\left\|\psi_{\ell, \nu, \rho}\right\|_{L^{\infty }\left(B_{4 L_{*}}(0)\right)}\\&&+\left\|\Theta _{r, \varepsilon, \rho}\right\|_{L^{\infty}\left(B_{4 L_{*}}(0)\right)}+\left\|\Theta _{\ell, \nu, \rho}\right\|_{L^{\infty}\left(B_{4 L_{*}}(0)\right)} \leq C.\end{eqnarray*}$

根据(2.14)式有

$\left\|\Theta_{r, \varepsilon, \rho}\right\|_{L^{1+\frac{1}{s}\left(\mathbb{R}^{2}\right)}}^{1+\frac{1}{s}}+\left\|\Theta_{\ell, \nu, \rho}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R}^{2}\right)}^{1+\frac{1}{s}} \leq C_{s, \varepsilon, \nu}\left[1+\rho^{1+\frac{1}{s}}-\rho^{\frac{1}{s}} \mathcal{E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^{0}, \Theta_{\ell}^{0}\right)\right]$.

又由(2.5), (2.6), (2.11)和(2.12)式可得 $\left|{\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}^0,\Theta_{\ell}^0\right)\right| \leq \frac{C_{s,\varepsilon, \nu}}{\rho^{\frac{1}{s}}}.$ 所以对所有的 $\rho \in\left ( 0,1 \right )$$\left\|\Theta_{r, \varepsilon, \rho}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}}+\left\|\Theta _{\ell, \nu, \rho}\right\|_{L^{1+\frac{1}{s}}\left(\mathbb{R} ^{2}\right)}^{1+\frac{1}{s}} \leq C_{s,\varepsilon, \nu}.$

再由引理2.1, 对所有的 $\rho \in\left ( 0,1 \right )$

$\begin{equation} \|{\cal G}_{s} \left(\Theta_{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)\|_{L^{\infty}\left(B_{4 L_{*}}(0)\right)} \leq C_{s}\|\Theta_{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho} \|_{L^{1+\frac{1}{s} }\left(B_{4 L_{*}}(0)\right)}\leq C_{s,\varepsilon, \nu}.\end{equation}$

我们现在来证明当 $\rho$ 充分小时, $\mu_{r, \varepsilon, \rho}$$\mu_{\ell, \nu, \rho}$ 是一致有界的.

$\psi_{r, \varepsilon, \rho}(x)={\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)(x)-W x_{1}-\mu_{r, \varepsilon, \rho}$

$\begin{equation} \mu_{r, \varepsilon, \rho}={\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)(x)-W x_{1}-\psi_{r, \varepsilon, \rho}(x),uad x\in D_{r}.\end{equation}$

因为 ${\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)$$W x_{1}$$D_{r}$ 上一致有界, 所以只需证明

$\inf\limits_{D_{r}} \psi_{r, \varepsilon, \rho} \leq 1 uad \mbox{ 和 } uad \sup\limits_{D_{r}} \psi_{r, \varepsilon, \rho} \geq 0.$

事实上, 一方面, 根据(2.22)式, 一定有 $\inf\limits_{D_{r}} \psi_{r, \varepsilon, \rho} \leq 1$.否则, 若假设$\inf\limits_{D_{r}} \psi_{r, \varepsilon, \rho} > 1$, 那么有

$\Theta _{r, \varepsilon, \rho}(x)=\frac{1}{\varepsilon^{2}} f_{r, \rho}\left(\psi_{r, \varepsilon, \rho} \right)>\frac{f_{r}\left(1 \right)}{\varepsilon^{2}},$

从而有

$\begin{equation}\frac{1}{2}\int_{D_{r}} \Theta _{r, \varepsilon, \rho} {\rm d}x>\frac{f_{r}\left(1 \right)}{2\varepsilon^{2}}{\rm meas}\left(D_{r}\right). \end{equation} $

同理可得

$\begin{equation}\frac{1}{2}\int_{D_{\ell}} \Theta _{\ell, \nu, \rho} {\rm d}x>\frac{f_{\ell}\left(1 \right)}{2\nu^{2}}{\rm meas}\left(D_{\ell}\right). \end{equation}$

因此由(2.30), (2.31)式以及 $\left(\Theta _{r, \varepsilon, \rho},\Theta _{\ell, \nu, \rho}\right)\in{\cal A}_{r} \times {\cal A}_{\ell}$ 可得

$\frac{f_{r}\left(1\right)}{2 \varepsilon^{2}} {\rm meas}\left(D_{r}\right)+\frac{f_{\ell}\left(1\right)}{2 \nu^{2}} {\rm meas}\left(D_{\ell}\right)<\kappa,$

这与(2.1)式矛盾.

另一方面, 根据 $\int_{D_{r}} \Theta _{r, \varepsilon, \rho} {\rm d}x=\kappa$, 一定有 $\sup\limits_{D_{r}} \psi_{r, \varepsilon, \rho} \geq 0$.否则, 若假设 $\sup\limits_{D_{r}} \psi_{r, \varepsilon, \rho}<0$, 那么有 $\int_{D_{r}} \Theta _{r, \varepsilon, \rho} {\rm d}x=0$, 矛盾.因此当 $\rho$ 充分小时, $\mu_{r, \varepsilon, \rho}$ 是一致有界的, 类似地可得 $\mu_{\ell, \nu, \rho}$ 也是一致有界的.

此外, 通过(2.28), (2.29)式可得

$\left\|\psi_{r, \varepsilon, \rho}\right\|_{L^{\infty }\left(B_{4 L_{*}}(0)\right)}\leq C_{s,\varepsilon, \nu},uad \rho \in\left (0,1 \right ).$

同理可得

$\left\|\psi_{\ell, \nu, \rho}\right\|_{L^{\infty }\left(B_{4 L_{*}}(0)\right)}\leq C_{s,\varepsilon, \nu},uad \rho \in\left ( 0,1 \right ).$

再结合引理易得

$\left\|\Theta _{r, \varepsilon, \rho}\right\|_{L^{\infty}\left(B_{4 L_{*}}(0)\right)}+\left\|\Theta _{\ell, \nu, \rho}\right\|_{L^{\infty}\left(B_{4 L_{*}}(0)\right)} \leq C_{s,\varepsilon, \nu},uad \rho \in\left ( 0,1 \right ).$

引理得证.

2.4 $\left(\Theta _{r, \varepsilon, \rho},\Theta _{\ell, \nu, \rho}\right)$${\cal E}_{\varepsilon, \nu, \rho}\left(\Theta _{r, \varepsilon, \rho},\Theta _{\ell, \nu, \rho}\right)$ 的收敛性

有了引理2.4, 我们现在可以利用紧性论证获得 $\rho \rightarrow 0$$\left(\Theta _{r, \varepsilon, \rho},\Theta _{\ell, \nu, \rho}\right)$ 以及 ${\cal E}_{\varepsilon, \nu, \rho}\big(\Theta _{r, \varepsilon, \rho},$$\Theta _{\ell, \nu, \rho}\big)$ 的极限.

引理 2.5 存在 $\left(\Theta _{r, \varepsilon}, \Theta _{\ell, \nu}\right) \in {\cal A}_{r} \times {\cal A}_{\ell}$ 使得

${\cal E}_{\varepsilon, \nu}\left(\Theta _{r, \varepsilon}, \Theta _{\ell, \nu}\right)=\max _{{\cal A}_{r} \times {\cal A}_{\ell}} {\cal E}_{\varepsilon, \nu}.$

此外, 存在两个拉格朗日乘子$\mu_{r, \varepsilon}$, $\mu _{\ell, \nu}\in \mathbb{R} $, 使得

$ \begin{equation} \Theta _{r, \varepsilon}(x)=\frac{1}{\varepsilon^{2}} f_{r}\left({\cal G}_s\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right)(x)-W x_{1}-\mu_{r, \varepsilon}\right), uad \mbox{ a.e. } x \in D_{r}, \end{equation} $
$\begin{equation} \Theta _{\ell, \nu}(x)=\frac{1}{\nu^{2}} f_{\ell}\left(-{\cal G}_s\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right)(x)+W x_{1}-\mu_{\ell, \nu}\right), uad \mbox{ a.e. } x \in D_{\ell}. \end{equation}$

因此

$\left\|\Theta _{r, \varepsilon}\right\|_{L^{\infty}\left(\mathbb{R} ^{2}\right)} \leq \frac{\sup_\mathbb{R} f_{r}}{\varepsilon^{2}},uad \left\|\Theta _{\ell, \nu}\right\|_{L^{\infty}\left(\mathbb{R} ^{2}\right)} \leq \frac{\sup_\mathbb{R} f_{\ell}}{\nu^{2}}.$

根据引理2.4, 知 $\mu_{r, \varepsilon, \rho}$$\mu_{\ell, \nu, \rho}$ 有收敛的子列(仍记为$\mu_{r, \varepsilon, \rho}$$\mu_{\ell, \nu, \rho}$), 使得当 $\rho \rightarrow 0$

$\begin{equation}\mu_{r, \varepsilon, \rho} \rightarrow \mu_{r, \varepsilon},uad \mu_{\ell, \nu, \rho} \rightarrow \mu_{\ell, \nu}.\end{equation}$

$\Theta _{r, \varepsilon, \rho}$$\Theta _{\ell, \nu, \rho}$ 有弱 $*$ 收敛的子列(仍记为$\Theta _{r, \varepsilon, \rho}$$\Theta _{\ell, \nu, \rho}$), 在 $L^{\infty}$ 的弱 $*$ 拓扑下有

$\begin{equation} \Theta _{r, \varepsilon, \rho} \rightharpoonup \Theta _{r, \varepsilon},uad \Theta _{\ell, \nu, \rho} \rightharpoonup \Theta _{\ell, \nu}.\end{equation} $

根据正则性理论, 对任意的 $1<p<\infty$, 存在一个依赖于$s$$p$ 的正常数 $C_{s,p}$, 使得

$\left\|{\cal G}_s\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)\right\|_{\dot{W}^{2s,p}\left(\mathbb{R} ^{2}\right)}\leq C_{s,p}\left\|\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right\|_{L^{p}\left(B_{4 L_{*}}(0)\right)}.$

结合$\psi_{r, \varepsilon, \rho}(x)$的表达式知, 存在不依赖于 $\rho$ 的常数 $C\left(R\right)$, 使得对任意的 $1<p<\infty$ 以及任意的正数 $R$

$\left\|\psi_{r, \varepsilon, \rho}\right\|_{W^{2s,p}\left(B_{R}(0)\right)}\leq C\left(R\right).$

因此, 存在子序列(仍然记为 $\psi_{r, \varepsilon, \rho}$)和连续函数$\psi_{r, \varepsilon}: \mathbb{R} ^{2} \rightarrow \mathbb{R} $,使得当 $\rho \rightarrow 0$ 时, 在 $L^{\infty}\left(B_{R}(0)\right)$的强拓扑下有

$\begin{equation}\label{eq:e5} \psi_{r, \varepsilon, \rho} \rightarrow \psi_{r, \varepsilon}. \end{equation} $

由(2.34), (2.35)及(2.36)式, 显然有

$\psi_{r, \varepsilon}(x)={\cal G}_s\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right)(x)-W x_{1}-\mu_{r, \varepsilon}.$

根据(2.22)式可得

$\begin{equation} \Theta _{r, \varepsilon, \rho}(x)=\frac{1}{\varepsilon^{2}} f_{r, \rho}\left(\psi_{r, \varepsilon, \rho} \right)= \frac{1}{\varepsilon^{2}} f_{r}\left(\psi_{r, \varepsilon, \rho}\right)+\frac{\rho}{\varepsilon^{2}}\left(\psi_{r, \varepsilon, \rho}\right)_{+}^{\frac{1}{s}}, uad \mbox{ a.e. } x \in D_{r}, \end{equation} $

$\rho \rightarrow 0$

$\Theta _{r, \varepsilon}(x)=\frac{1}{\varepsilon^{2}} f_{r}\left(\psi_{r, \varepsilon}\right), uad \mbox{ a.e. } x \in D_{r}.$

因此(2.32)式得证, (2.33)式可以用类似的方法得到.

接下来, 要证 $\left(\Theta _{r, \varepsilon}, \Theta _{\ell, \nu}\right)$${\cal E}_{\varepsilon, \nu}$${\cal A}_{r} \times {\cal A}_{\ell}$ 上的极大元. 对任意的 $\left(\Theta _{r}, \Theta _{\ell}\right)\in {\cal A}_{r} \times {\cal A}_{\ell}$, 有

$ \begin{matrix} {\cal E}_{\varepsilon, \nu}\left(\Theta _{r}, \Theta _{\ell}\right) &=&E_{s}\left(\Theta _{r}-\Theta _{\ell}\right)-W {\cal P}\left(\Theta _{r}-\Theta _{\ell}\right)-{\cal J}_{r, \varepsilon}\left(\Theta _{r}\right)-{\cal J}_{\ell, \nu}\left(\Theta _{\ell}\right) \\ & \leq& E_{s}\left(\Theta _{r}-\Theta _{\ell}\right)-W {\cal P}\left(\Theta _{r}-\Theta _{\ell}\right)-{\cal J}_{r, \varepsilon,\rho }\left(\Theta _{r}\right)-{\cal J}_{\ell, \nu,\rho}\left(\Theta _{\ell}\right)\\ & =& {\cal E}_{\varepsilon, \nu, \rho}\left(\Theta_{r}, \Theta_{\ell}\right) \leq {\cal E}_{\varepsilon, \nu, \rho}\left(\Theta _{r, \varepsilon, \rho}, \Theta _{\ell, \nu, \rho}\right). \end{matrix}$

因此, 只需证明

$\begin{equation}\label{eq:e6} \lim _{\rho \rightarrow 0} {\cal E}_{\varepsilon, \nu, \rho}\left(\Theta _{r, \varepsilon, \rho}, \Theta _{\ell, \nu, \rho}\right)={\cal E}_{\varepsilon, \nu}\left(\Theta _{r, \varepsilon}, \Theta _{\ell, \nu}\right). \end{equation}$

一方面, 根据(2.35)式可得

$\begin{matrix} & &\lim_{\rho \rightarrow 0} E_{s}\left(\Theta_{r, \varepsilon, \rho}-\Theta_{\ell, \nu, \rho}\right)=E_{s}\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right),\\ &&\lim _{\rho \rightarrow 0} {\cal P}\left(\Theta _{r, \varepsilon, \rho}-\Theta _{\ell, \nu, \rho}\right)={\cal P}\left(\Theta_{r, \varepsilon}-\Theta _{\ell, \nu}\right). \end{matrix}$

另一方面, 根据(2.22), (2.23)式以及共轭函数的性质得

$J_{r, \rho}\left(\varepsilon^{2} \Theta_{r, \varepsilon, \rho}(x)\right)=\varepsilon^{2} \Theta_{r, \varepsilon, \rho}(x)\psi_{r, \varepsilon, \rho}(x)-F_{r, \rho}\left(\psi_{r, \varepsilon, \rho}(x)\right), \mbox{ a.e. } x \in D_{r},$
$J_{\ell, \rho}\left(\nu^{2} \Theta_{\ell, \nu, \rho}(x)\right)=\nu^{2} \Theta_{\ell, \nu, \rho}(x)\psi_{\ell, \nu, \rho}(x)-F_{\ell, \rho}\left(\psi_{\ell, \nu, \rho}(x)\right), {\rm a.e. }\ x \in D_{\ell}.$

结合(2.32), (2.33), (2.35)和(2.36)式, 可得

$\begin{equation} \lim _{\rho \rightarrow 0} {\cal J}_{r, \varepsilon, \rho}\left(\Theta_{r, \varepsilon, \rho}\right)={\cal J}_{r, \varepsilon}\left(\Theta_{r, \varepsilon}\right), uad \lim _{\rho \rightarrow 0} {\cal J}_{\ell, \nu, \rho}\left(\Theta_{\ell, \nu, \rho} \right)={\cal J}_{\ell, \nu}\left(\Theta_{\ell, \nu}\right).\end{equation}$

由(2.40)和(2.41)式得(2.39)式成立.

2.5 涡的能量下界估计及其相关估计

为了得到拉格朗日乘子和$(\Theta_{r, \varepsilon},\Theta_{\ell, \nu})$ 的支集的估计, 我们需要对它们的能量进行下界估计, 并且推出一些相关的估计.由于 $\left(\Theta_{r, \varepsilon},\Theta_{\ell, \nu}\right)$ 是能量极大元, 因此可以通过选取适当的检验函数得到能量的下界估计.根据能量的定义

${\cal E}_{\varepsilon, \nu}\left(\Theta _{r}, \Theta _{\ell}\right)=E_{s}\left(\Theta _{r}-\Theta _{\ell}\right)-W {\cal P}\left(\Theta _{r}-\Theta _{\ell}\right)-{\cal J}_{r, \varepsilon}\left(\Theta _{r}\right)-{\cal J}_{\ell, \nu}\left(\Theta _{\ell}\right), \ \left(\Theta _{r}, \Theta _{\ell}\right) \in {\cal A}_{r} \times {\cal A}_{\ell}.$

等价地, 有

${\cal E}_{\varepsilon, \nu}\left(\Theta_{r}, \Theta_{\ell}\right)={\cal E}_{r, \varepsilon}\left(\Theta_{r}\right)+{\cal E}_{\ell, \nu}\left(\Theta_{\ell}\right)-{\cal R}\left(\Theta_{r}, \Theta_{\ell}\right)={\cal E}_{r, \varepsilon}\left(\Theta_{r}\right)+{\cal E}_{\ell, \nu}\left(\Theta_{\ell}\right)+O(1).$

其中

${\cal E}_{r, \varepsilon}\left(\Theta_{r}\right):=E_{s}\left(\Theta_{r}\right)-W {\cal P}\left(\Theta _{r}\right)-{\cal J}_{r, \varepsilon}\left(\Theta_{r}\right), uad {\cal E}_{\ell, \nu}\left(\Theta_{\ell}\right):=E_{s}\left(\Theta_{\ell}\right)+W {\cal P}\left(\Theta _{\ell}\right)-{\cal J}_{\ell, \nu}\left(\Theta_{\ell}\right), $
${\cal R}\left(\Theta_{r}, \Theta_{\ell}\right)=c_{s} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{1}{{\left|x-x^{\prime}\right|}^{2-2s}} \Theta_{r}(x) \Theta_{\ell}\left(x^{\prime}\right) {\rm d} x {\rm d} x^{\prime}.$

显然可见

${\cal E}_{r, \varepsilon}\left(\Theta_{r, \varepsilon}\right)=\sup\limits_{{\cal A}_{r}} {\cal E}_{r, \varepsilon}+O(1), uad {\cal E}_{\ell, \nu}\left(\Theta_{\ell, \nu}\right)=\sup\limits_{{\cal A}_{\ell}} {\cal E}_{\ell, \nu}+O(1).$

$A_{s}=c_{s} \pi^{-1-s} \int_{B_{1}(0)} \int_{B_{1}(0)} \frac{1}{{\left|x-x^{\prime}\right|^{2-2s}}} {\rm d}x {\rm d}x^{\prime}.$

引理 2.6 对任意的 $\delta_{r} \in \left(0,\sup_{R}f_{r}\right)$, $\delta_{\ell} \in \left(0,\sup_{R}f_{\ell}\right)$, 存在只依赖于 $\delta_{r}$$\delta_{\ell}$ 的常数 $C_{\delta_{r}}$, $C_{\delta_{\ell}}>0$, 使得

$\begin{equation}{\cal E}_{r,\varepsilon}\left(\Theta_{r,\varepsilon}\right) \geq \frac{\delta_{r}^{1-s}\kappa^{1+s} A_{s}}{2 \varepsilon^{2-2s}}-C_{\delta_{r}}, \end{equation} $
$\begin{equation} {\cal E}_{\ell, \nu}\left(\Theta_{\ell, \nu}\right) \geq \frac{\delta_{\ell}^{1-s}\kappa^{1+s} A_{s}}{2 \nu^{2-2s}}-C_{\delta_{\ell}}. \end{equation}$

我们首先证明(2.42)式, 其关键在于选取合适的检验函数.取 $x_{0}=\left(a,0\right)\in D_{r}$, 记

$\Theta_{r,\varepsilon}^{\delta_{r}}=\frac{\delta_{r}}{\varepsilon^{2}} I_{B_{\varepsilon \sqrt\frac{\kappa }{{\pi \delta_{r}}}} \left(x_{0}\right)},$

$\varepsilon$ 充分小时 $\Theta_{r,\varepsilon}^{\delta_{r}}\in {\cal A}_{r}$.

容易验证当$\tau \notin\left[0, \sup _{\mathbb{R}} f_{r}\right]$时, $J_{r}(\tau)$ 等于正无穷, 因此 $J_{r}$ 的有效域包含在$\left[sup_{\mathbb{R} }f_{r}\right]$中.此外, 不难发现 $J_{r}$$\left[\delta_{r}\right]$ 上是有界的. 通过直接计算得

${\cal E}_{r,\varepsilon}\left(\Theta_{r,\varepsilon}^{\delta_{r}}\right)\geq \frac{\delta_{r}^{1-s}\kappa^{1+s} A_{s}}{2\varepsilon^{2-2s}}-C_{\delta_{r}}.$

因为 $\Theta_{r,\varepsilon}$ 是极大元, 所以 ${\cal E}_{r,\varepsilon}\left(\Theta_{r,\varepsilon}\right) \geq {\cal E}_{r,\varepsilon}\left(\Theta_{r,\varepsilon}^{\delta_{r}}\right).$ 故(2.42)式得证, 同理可得(2.43)式.

$\begin{equation} \widetilde{\Theta}_{r,\varepsilon}\left(x\right) ={\varepsilon}^{2}\Theta_{r,\varepsilon}\left(\varepsilon x\right),uad \widetilde{\Theta}_{\ell, \nu}\left(x\right) =\nu^{2}\Theta_{\ell, \nu}\left(\nu x\right),uad \forall x \in \mathbb{R} ^{2}.\end{equation}$

那么

$\begin{equation}0 \leq \widetilde{\Theta}_{r,\varepsilon} \leq \sup_{\mathbb{R} }f_{r},uad {\rm supp}(\widetilde{\Theta}_{r,\varepsilon}) \subseteq B_{\frac{L_{*}}{2\varepsilon}}\left(L_{*}e_{1}\right),uad \int_{\mathbb{R} ^{2}} \widetilde{\Theta}_{r,\varepsilon}\left(x\right) {\rm d}x=\kappa,\end{equation}$
$\begin{equation}0 \leq \widetilde{\Theta}_{\ell, \nu} \leq \sup_{\mathbb{R} }f_{\ell},uad {\rm supp}(\widetilde{\Theta}_{\ell, \nu}) \subseteq B_{\frac{L_{*}}{2\nu}}\left(-L_{*}e_{1}\right),uad \int_{\mathbb{R} ^{2}} \widetilde{\Theta}_{\ell, \nu}\left(x\right) {\rm d}x=\kappa.\end{equation}$

通过简单的计算有

$E_{s}\left(\Theta _{r,\varepsilon}\right)=\frac{c_{s}}{2 \varepsilon^{2-2s}} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{ \widetilde{\Theta}_{r,\varepsilon}(x) \widetilde{\Theta}_{r,\varepsilon}\left(x^{\prime}\right)}{|x-x^{\prime}|^{2-2s}} {\rm d}x {\rm d}x^{\prime},$
$E_{s}\left(\Theta _{\ell, \nu}\right)=\frac{c_{s}}{2 \nu^{2-2s}} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{\widetilde{\Theta}_{\ell, \nu}(x) \widetilde{\Theta}_{\ell, \nu}\left(x^{\prime}\right)}{|x-x^{\prime}|^{2-2s}} {\rm d}x {\rm d}x^{\prime}.$

$\begin{equation}{\cal I}_{r,s}(\widetilde{\Theta}_{r})=c_{s} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{\widetilde{\Theta}_{r}(x) \widetilde{\Theta}_{r}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2 s}} {\rm d}x {\rm d}x^{\prime},\end{equation}$
$\begin{equation} {\cal I}_{\ell,s}(\widetilde{\Theta}_{\ell})=c_{s} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{\widetilde{\Theta}_{\ell}(x) \widetilde{\Theta}_{\ell}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2 s}} {\rm d}x {\rm d}x^{\prime}.\end{equation}$

由引理2.6, 我们可以得到以下结果.

引理 2.7 对任意的 $\delta_{r} \in \left(0,\sup_{\mathbb{R} }f_{r}\right)$, $\delta_{\ell} \in \left(0,\sup_{\mathbb{R} }f_{\ell}\right)$, 存在只依赖于 $\delta_{r}$$\delta_{\ell}$ 的常数 $C_{\delta_{r}}$, $C_{\delta_{\ell}}>0$, 使得

$\begin{equation}{\cal I}_{r,s}(\widetilde{\Theta}_{r,\varepsilon})\geq \delta_{r}^{1-s}\kappa^{1+s}A_{s}-C_{\delta_{r}}\varepsilon^{2-2s},\end{equation}$
$\begin{equation}{\cal I}_{\ell,s}(\widetilde{\Theta}_{\ell, \nu})\geq \delta_{\ell}^{1-s}\kappa^{1+s}A_{s}-C_{\delta_{\ell}}\nu^{2-2s}.\end{equation}$

引理 2.8 对任意的 $\eta_{r}$, $\eta_{\ell} \in \left ( 0,\kappa \right )$, 存在正数 $R_{r}$, $R_{\ell}$ 使得

$\begin{equation} \sup\limits_{y \in \mathbb{R} ^{2}} \int_{B_{R_{r}}(y)} \widetilde{\Theta}_{r,\varepsilon} {\rm d}x>\kappa -\eta_{r}, uad \forall \varepsilon>0,\end{equation}$
$\begin{equation}\sup\limits_{y \in \mathbb{R} ^{2}} \int_{B_{R_{\ell}}(y)} \widetilde{\Theta}_{\ell,\nu} {\rm d}x>\kappa -\eta_{\ell}, uad \forall \nu>0.\end{equation}$

为了证明引理2.8, 我们需要介绍两个辅助引理.第一个是Lions[21]中的集中紧性引理.

引理 2.9$\left\{\xi_{n} \right\}_{n=1}^{\infty}$ 是定义在 $\mathbb{R} ^{2}$ 上的非负可积的函数序列, 对某个 $0<\beta<\infty$

$\limsup_{n\rightarrow \infty} \int_{\mathbb{R} ^{2}}\xi _{n}{\rm d}x\rightarrow \beta.$

那么, 存在一个子序列(仍然记为 $\left\{\xi_{n} \right\}_{n=1}^{\infty}$), 使得下述情形之一成立

(i) (紧性) 存在 $\mathbb{R} ^{2}$ 中的序列 $\left\{y_{n} \right\}_{n=1}^{\infty}$, 使得对任意的 $\varepsilon>0$, 存在 $R>0$ 满足

$\int_{B_{R}(y_{n})}\xi _{n}{\rm d}x\geq \beta-\varepsilon, uad \forall n \geq 1.$

(ii) (消失) 对每个 $R>0$, 有

$\lim_{n \to \infty} \sup_{y\in \mathbb{R} ^{2}} \int_{B_{R}(y)}\xi _{n}{\rm d}x=0.$

(iii) (分裂) 存在一个数 $\beta_{1}\in (0,\beta)$, 使得对任意的 $\varepsilon>0$, 存在 $N=N(\varepsilon) \geq 1$ 以及 $0 \leq \xi _{i,n} \leq \xi _{n}$, $i=1,2$, 满足

$\begin{matrix}\ \left\{ \begin{array}{lll} \left\|\xi_{n}-\xi_{1, n}-\xi_{2, n}\right\|_{1}+\left|\beta_{1}-\int_{\mathbb{R} ^{2}} \xi_{1, n}{\rm d} x\right|+\left|\beta-\beta_{1}-\int_{\mathbb{R} ^{2}} \xi_{2, n} {\rm d} x\right|<\epsilon, uad n \geq N, \\ d_{n}:={\rm dist}\left({\rm supp}\left(\xi_{1, n}\right), {\rm supp}\left(\xi_{2, n}\right)\right) \rightarrow \infty, uad n \rightarrow \infty.\end{array}\right.\end{matrix}$

另外, 还需要以下结果.

引理 2.10 (Riesz 积分的浴缸原理) 设 $s \in \left(0,1\right)$, $\beta_{1}$, $\beta_{2}$ 是两个正数, 给定一个正数 $\eta$, 记

$\mathcal{B}_{\eta}:=\left\{\xi \in L^{\infty}\left(\mathbb{R}^{2},[0,1]\right): \int_{\mathbb{R}^{2}} \xi \mathrm{d} x \leq \eta\right\}$

那么极大值问题

${\cal M}_{\beta_{1}, \beta_{2}}:=\sup\limits_{\left(\xi_{1}, \xi_{2}\right) \in {\cal B}_{\beta_{1}} \times {\cal B}_{\beta_{2}}} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{1}{|x-y|^{2-2 s}} \xi_{1}(x) \xi_{2}(y) {\rm d} x {\rm d}y$

的解为

$\xi_{i}(x)=I_{B_{R_{i}}(0)}, uad R_{i}=\sqrt{\frac{\beta_{i}}{\pi}}, uad i=1,2.$

注意到引理2.10是浴缸原理和 $Riesz$ 重排不等式的一个简单应用.此处略去证明, 请读者参考文献[22,定理1.14和定理3.9].

下面, 我们给出引理2.8的证明.

我们只证明(2.51)式, 同理可得(2.52)式.我们采用反证法, 假设存在 $\eta_{0}\in \left(0,\kappa \right)$, 使得对于任何整数 $n\geq1$, 存在 $\varepsilon_{n}>0$, 满足当 $n \rightarrow \infty$ 时, $\varepsilon_{n}\rightarrow 0^{+}$, 使得

$\sup\limits_{y \in \mathbb{R} ^{2}} \int_{B_{n}(y)} \widetilde{\Theta}_{r,\varepsilon_{n}} {\rm d}x \leq \kappa -\eta_{0}, uad \forall n \geq 1.$

$\xi_{n}=\frac{\widetilde{\Theta}_{r,\varepsilon_{n}}}{\sup_{R} f_{r}}$, 则根据引理2.9, 存在子序列(仍记为$\xi_{n}$)满足引理2.9中的三种情况之一.

$\left\{\xi_{n} \right\}_{n=1}^{\infty}$ 满足引理2.9中的紧性情形, 那么这与假设相矛盾, 从而完成证明.因此, 现只需排除消失情形和分裂情形.

首先排除消失情形.假设对每个固定的 $R>0$, 有

\begin{matrix} \lim_{n \to \infty} \sup_{y\in \mathbb{R} ^{2}} \int_{B_{R}(y)}\xi_{n}{\rm d}x=0. \end{matrix}

则有 $\lim_{n \rightarrow \infty}{\cal I}_{r,s}(\xi_{n})=0$, 这与(2.49)式矛盾.

事实上, 对任意的 $x\in \mathbb{R} ^{2}$$R>0$, 由Hölder 不等式可得

$\begin{matrix}\int_{\mathbb{R} ^{2}} \frac{\xi_{n}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2 s}} {\rm d} x^{\prime}&=&\int_{\left|x-x^{\prime}\right|<R} \frac{\xi_{n}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2 s}} {\rm d} x^{\prime}+\int_{\left|x-x^{\prime}\right| \geq R} \frac{\xi_{n}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2 s}} {\rm d} x^{\prime} \\&\leq& \bigg(\int_{\left|x-x^{\prime}\right|<R} \frac{1}{\left|x-x^{\prime}\right|^{2-2s^{2}}} {\rm d} x^{\prime}\bigg)^{\frac{1}{1+s}}\left\|\xi_{n}\right\|_{L^{1+\frac{1}{s}}\left(B_{R}(x)\right)}+\frac{\kappa}{R^{2-2s}\sup_{\mathbb{R} }f_{r}} \\& \leq& C_{s} R^{2 s^{2}}\bigg(\sup\limits_{y \in \mathbb{R} ^{2}} \int_{B_{R}(y)} \xi_{n} {\rm d} x\bigg)^{\frac{s}{1+s}}+\frac{\kappa}{R^{2-2s}\sup_{\mathbb{R} } f_{r}},\end{matrix}$

其中 $C_{s}$ 是只依赖于 $s$ 的正常数.因此有

$\begin{matrix} {\cal I}_{r,s}\left(\xi_{n}\right) &=&c_{s} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{\xi_{n}(x){\xi_{n}} \left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2s}} {\rm d}x {\rm d} x^{\prime} \\ & \leq &C_{s} R^{2 s^{2}}\bigg(\sup\limits_{y \in \mathbb{R} ^{2}} \int_{B_{R}(y)} \xi_{n} {\rm d} x\bigg)^{\frac{s}{1+s}}+\frac{C_{s}}{R^{2-2 s}}.\end{matrix}$

先令 $n \rightarrow \infty$, 再令 $R \rightarrow \infty$, 由(2.54)式可得

$\lim_{n \rightarrow \infty}{\cal I}_{r,s}(\xi_{n})=0.$

然后, 我们来排除分裂情形.假设存在 $\beta\in (0,\frac{\kappa}{\sup_{\mathbb{R} } f_{r}})$, 使得对任意的 $\varepsilon>0$, 存在 $N=N(\varepsilon) \geq 1$ 以及 $0 \leq {\xi}_{r,\varepsilon_{n}}^{i} \leq \xi_{n}$, $i=1,2,3$, 满足

$\begin{matrix} \left\{ \begin{array}{lll} \xi_{n}={\xi}_{r,\varepsilon_{n}}^{1}+{\xi}_{r,\varepsilon_{n}}^{2}+{\xi}_{r,\varepsilon_{n}}^{3},\\[2mm] \left|\beta-\beta_{1, n}\right|+\left|\frac{\kappa}{\sup_{\mathbb{R} } f_{r}}-\beta-\beta_{2, n}\right|+\left|\beta_{3, n}\right|<\epsilon, uad n \geq N(\epsilon),\\[2mm] d_{n}:={\rm dist}\left({\rm supp}{\xi}_{r,\varepsilon_{n}}^{1}, {\rm supp}{\xi}_{r,\varepsilon_{n}}^{2}\right) \rightarrow \infty, uad n \rightarrow \infty. \end{array} \right.\end{matrix}$

其中 $\beta_{i, n}=\left\|{\xi}_{r,\varepsilon_{n}}^{i}\right\|_{L^{1}\left(\mathbb{R} ^{2}\right)}$, $i=1,2,3$. 使用对角线方法, 可得一个子序列, 仍然记为 $\left\{\xi_{n} \right\}_{n=1}^{\infty}$, 使得

$\begin{matrix} \left\{ \begin{array}{lll} \xi_{n}={\xi}_{r,\varepsilon_{n}}^{1}+{\xi}_{r,\varepsilon_{n}}^{2}+{\xi}_{r,\varepsilon_{n}}^{3},\\[2mm] \left|\beta-\beta_{1, n}\right|+\left|\frac{\kappa}{\sup_{\mathbb{R} } f_{r}}-\beta-\beta_{2, n}\right|+\left|\beta_{3, n}\right|\rightarrow 0, uad n \rightarrow \infty,\\[2mm] d_{n}:={\rm dist}\left({\rm supp}{\xi}_{r,\varepsilon_{n}}^{1}, {\rm supp}{\xi}_{r,\varepsilon_{n}}^{2}\right) \rightarrow \infty, uad n \rightarrow \infty. \end{array} \right. \end{matrix}$

根据 ${\cal I}_{r,s}$ 的对称性可得

$\begin{matrix} {\cal I}_{r,s}\left(\xi_{n}\right)&=& {\cal I}_{r,s}\left({\xi}_{r,\varepsilon_{n}}^{1}+{\xi}_{r,\varepsilon_{n}}^{2}+{\xi}_{r,\varepsilon_{n}}^{3}\right) \\ &=& c_{s} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{{\xi}_{r,\varepsilon_{n}}^{1}(x) {\xi}_{r,\varepsilon_{n}}^{1}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2 s}} {\rm d} x {\rm d} x^{\prime}+c_{s} \int_{\mathbb{R} ^{2}} \frac{{\xi}_{r,\varepsilon_{n}}^{2}(x) {\xi}_{r,\varepsilon_{n}}^{2}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2 s}} {\rm d} x {\rm d} x^{\prime} \\ &&+2 c_{s} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{{\xi}_{r,\varepsilon_{n}}^{1}(x) {\xi}_{r,\varepsilon_{n}}^{2}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2s}} {\rm d}x {\rm d}x^{\prime} \\ &&+c_{s} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{\left(2 \xi_{n}(x)-{\xi}_{r,\varepsilon_{n}}^{3}(x)\right) {\xi}_{r,\varepsilon_{n}}^{3}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2s}} {\rm d}x {\rm d}x^{\prime} \\& =&:I_{1}+I_{2}+I_{3}+I_{4}. \end{matrix}$

$r^{\prime}=\sqrt{\frac{\left \| {\xi}_{r,\varepsilon_{n}}^{1} \right \|_{L^{1}(\mathbb{R} ^{2})}}{\pi }}$, 利用引理2.10, 有

$I_{1}\leq c_{s}\int_{B_{r^{\prime}}(0)} \int_{B_{r^{\prime}}(0)} \frac{1}{\left|x-x^{\prime}\right|^{2-2 s}} {\rm d} x {\rm d}x^{\prime}=\beta_{1, n}^{1+s}A_{s}.$

同样地, 可得

$I_{2}\leq \beta_{2, n}^{1+s}A_{s}.$

对于 $I_{3}$, 当 $n$ 充分大时有

$I_{3}\leq \frac{2c_{s}\kappa^2}{(\sup_{\mathbb{R} }f_r)^2} d_{n}^{2s-2}=o_{n}\left ( 1 \right ).$

最后对于 $I_{4}$, 当 $n$ 充分大时有

$\begin{matrix} I_{4}&\leq& 2c_{s} \int_{\mathbb{R} ^{2}} \int_{\mathbb{R} ^{2}} \frac{ \xi_{n}(x) {\xi}_{r,\varepsilon_{n}}^{3}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2s}} {\rm d}x {\rm d}x^{\prime}\\ &=&2c_{s}\int_{\mathbb{R} ^{2}}\xi_{n}(x)\bigg(\int_{\left|x-x^{\prime}\right|\geq 1}+\int_{\left|x-x^{\prime}\right|< 1}\frac{ {\xi}_{r,\varepsilon_{n}}^{3}\left(x^{\prime}\right)}{\left|x-x^{\prime}\right|^{2-2s}} {\rm d}x^{\prime}\bigg){\rm d}x\\ &\leq &2c_{s}\int_{\mathbb{R} ^{2}}\xi_{n}(x)\left(\left \| {\xi}_{r,\varepsilon_{n}}^{3} \right \|_{L^{1}(\mathbb{R} ^{2})}+C\left \| {\xi}_{r,\varepsilon_{n}}^{3} \right \|_{L^{1+\frac{1}{s}}(\mathbb{R} ^{2})}\right){\rm d}x\\ &\leq &C_{s}\left(\left \| {\xi}_{r,\varepsilon_{n}}^{3} \right \|_{L^{1}(\mathbb{R} ^{2})}+\left \| {\xi}_{r,\varepsilon_{n}}^{3} \right \|_{L^{1}(\mathbb{R} ^{2})}^{\frac{s}{1+s}}\right)\\ &=&o_{n}\left ( 1 \right ). \end{matrix}$

通过以上的计算可得

${\cal I}_{r,s}\left(\xi_{n}\right)\leq \left(\beta_{1, n}^{1+s}+\beta_{2, n}^{1+s}\right)A_{s}+o_{n}\left (1 \right ).$

$n$ 充分大时, 由(2.58)式可得

${\cal I}_{r,s}\left(\xi_{n}\right)\leq \left[\beta^{1+s}+\left(\frac{\kappa}{\sup_{\mathbb{R} } f_{r}}-\beta\right)^{1+s}\right]A_{s}+o_{n}\left ( 1 \right ).$

另一方面, 根据引理2.7, 对任意的 $\delta_{r} \in \left(0,\sup_{\mathbb{R} }f_{r}\right)$

$\begin{equation} {\cal I}_{r,s}\left(\xi_{n}\right)\geq \frac{\delta _{r}^{1-s}\kappa^{1+s}}{\left(\sup_{\mathbb{R} }f_{r}\right)^2}A_{s}+o_{n}\left ( 1 \right ). \end{equation}$

则对任意的 $\delta_{r} \in \left(0,\sup_{\mathbb{R} }f_{r}\right)$, 存在 $\beta\in (0,\frac{\kappa}{\sup_{\mathbb{R} } f_{r}})$, 使得

$\frac{\delta _{r}^{1-s}\kappa^{1+s}}{\left(\sup_{\mathbb{R} }f_{r}\right)^2}\leq \beta^{1+s}+\left(\frac{\kappa}{\sup_{\mathbb{R} } f_{r}}-\beta\right)^{1+s},$

由此容易推出 $s\leq 0$, 这与 $0<s<1$ 矛盾.故假设不成立. 引理得证.

引理 2.11 存在 $x_{1}$ 轴上的点 $\left\{y_{r,\varepsilon}\right\}_{\varepsilon>0}$$\left\{y_{\ell,\nu}\right\}_{\nu>0}$, 使得对任意的 $\eta_{r}$, $\eta_{\ell} \in \left ( 0,\kappa \right )$, 存在正数 $R_{\eta_r}$, $R_{\eta_\ell}$ 满足

$ \begin{equation} \int_{B_{R_{\eta_r}}(y_{r,\varepsilon})} \widetilde{\Theta}_{r,\varepsilon} {\rm d}x>\kappa -\eta_{r}, uad \forall \varepsilon>0, \end{equation}$
$\begin{equation} \int_{B_{R_{\eta_{\ell}}}(y_{\ell,\nu})} \widetilde{\Theta}_{\ell,\nu} {\rm d}x>\kappa -\eta_{\ell}, uad \forall \nu>0. \end{equation}$

根据引理2.8以及 $\widetilde{\Theta}_{r,\varepsilon}$ 关于 $x_{1}$ 轴对称性, 可以取 $\left\{y_{\varepsilon}^{0}\right\}_{\varepsilon>0}\in \mathbb{R} \times \left \{ 0 \right \}$$R_{0}>0$, 使得

$\int_{B_{R_{0}}(y_{\varepsilon}^{0})}\widetilde{\Theta}_{r,\varepsilon}{\rm d}x>\frac{5\kappa}{6}, uad \forall \varepsilon>0.$

对任意的 $\eta_r \in \left(0,\frac{\kappa}{2}\right)$, 我们将证明存在 $R_{\eta_r}>R_{0}$, 使得

$\int_{B_{R_{\eta_r}}(y_{\varepsilon}^{0})}\widetilde{\Theta}_{r,\varepsilon}{\rm d}x>\kappa -\eta_{r}, uad \forall \varepsilon>0.$

事实上, 由引理2.8知, 存在 $\widetilde{R}_{\eta_r}>R_{0}$$\left\{y_{\varepsilon}^{\eta_r}\right\}_{\varepsilon>0}\subseteq \mathbb{R} ^{2}$, 使得

$\int_{B_{\widetilde{R}_{\eta_r}}(y_{\varepsilon}^{\eta_r})}\widetilde{\Theta}_{r,\varepsilon}{\rm d}x>\kappa -\eta_{r}, uad \forall \varepsilon>0.$

我们断言 $\left|y_{\varepsilon}^{\eta_r} - y_{\varepsilon}^{0}\right| \leq 2 \widetilde{R}_{\eta_r}, \ \forall \varepsilon>0.$ 下面采用反证法, 若 $\left|y_{\varepsilon}^{\eta_r} - y_{\varepsilon}^{0}\right| > 2 \widetilde{R}_{\eta_r}, \ \forall \varepsilon>0$, 则有

$B_{\widetilde{R}_{\eta_r}}\left(y_{\varepsilon}^{0}\right) \cap B_{\widetilde{R}_{\eta_r}}\left(y_{\varepsilon}^{\eta_r}\right)=\varnothing.$

从而

$\kappa =\int_{\mathbb{R} ^{2}} \widetilde{\Theta}_{r,\varepsilon} {\rm d} x \geq \int_{B_{\widetilde{R}_{\eta_r}}\left(y_{\varepsilon}^{0}\right)} \widetilde{\Theta}_{r,\varepsilon}{\rm d}x+\int_{B_{\widetilde{R}_{\eta_r}}\left(y_{\varepsilon}^{\eta_r}\right)} \widetilde{\Theta}_{r,\varepsilon} {\rm d} x \geq \frac{5\kappa}{6}+\kappa-\eta_{r}>\kappa.$

这导致了矛盾.

$R_{\eta_r}=3\widetilde{R}_{\eta_r}$, 那么有

$\int_{B_{R_{\eta_r}}(y_{\varepsilon}^{0})}\widetilde{\Theta}_{r,\varepsilon}{\rm d}x\geq \int_{B_{\widetilde{R}_{\eta_r}}(y_{\varepsilon}^{\eta_r})}\widetilde{\Theta}_{r,\varepsilon}{\rm d}x >\kappa -\eta_{r}, uad \forall \varepsilon>0.$

因此(2.62)式得证, 同理可得(2.63)式. 引理得证.

接下来, 我们考虑函数 $\Theta_{r,\varepsilon}$$\Theta_{\ell,\nu}$, 可立即得到以下引理.

引理 2.12 存在 $x_{1}$ 轴上的点 $\left\{z_{r,\varepsilon}\right\}_{\varepsilon>0}$$\left\{z_{\ell,\nu}\right\}_{\nu>0}$, 使得对任意的 $\eta_{r}$, $\eta_{\ell} \in \left ( 0,\kappa \right )$, 存在正数 $R_{\eta_r}$, $R_{\eta_\ell}$ 满足

$\begin{equation} \int_{B_{R_{\eta_r}\varepsilon}(z_{r,\varepsilon})} \Theta_{r,\varepsilon} {\rm d}x>\kappa -\eta_{r}, uad \forall \varepsilon>0, \end{equation}$
$\begin{equation} \int_{B_{R_{\eta_{\ell}}\nu}(z_{\ell,\nu})} \Theta_{\ell,\nu} {\rm d}x>\kappa -\eta_{\ell}, uad \forall \nu>0. \end{equation}$

根据(2.44)式以及引理2.11得

$\begin{matrix}\int_{B_{R_{\eta_r}}(y_{r,\varepsilon})}\widetilde{\Theta}_{r,\varepsilon} {\rm d}x&=&\int_{B_{R_{\eta_r}}(y_{r,\varepsilon})}{\varepsilon}^{2}\Theta_{r,\varepsilon}\left(\varepsilon x\right) {\rm d}x\\ &=&\int_{B_{R_{\eta_r}\varepsilon}(y_{r,\varepsilon})}\Theta_{r,\varepsilon}\left(x\right) {\rm d}x >\kappa -\eta_{r}. \end{matrix}$

因此存在 $z_{r,\varepsilon}=y_{r,\varepsilon}\in \mathbb{R} \times \left\{0\right\}$, 使得(2.64)式成立, 同理可得(2.65)式. 证毕.

2.6 拉格朗日乘子 $\mu _{r,\varepsilon}$$\mu _{\ell,\nu}$ 的估计

有了引理2.12, 我们就可以给出拉格朗日乘子 $\mu _{r,\varepsilon}$$\mu _{\ell,\nu}$ 的估计.

引理 2.13$\frac{1}{2} \leq s<1$, 则

$\begin{equation} 0<\liminf _{\varepsilon, \nu \rightarrow 0^{+}} \varepsilon^{2-2 s} \mu_{r, \varepsilon} \leq \limsup _{\varepsilon, \nu \rightarrow 0^{+}} \varepsilon^{2-2 s} \mu_{r, \varepsilon}<+\infty,\end{equation}$
$\begin{equation} 0<\liminf _{\varepsilon, \nu \rightarrow 0^{+}} \nu^{2-2 s} \mu_{\ell, \nu} \leq \limsup _{\varepsilon, \nu \rightarrow 0^{+}} \nu^{2-2 s} \mu_{\ell, \nu}<+\infty. \end{equation}$

对任意的 $x\in D_{r}$, 利用重排不等式得

${\cal G}_s\Theta_{r,\varepsilon}(x)=\int_{\mathbb{R} ^{2}} G_{s}(x-x^{\prime}) \Theta_{r,\varepsilon}(x^{\prime}){\rm d}x^{\prime}\leq \frac{c_{s}\pi^{1-s}\kappa^{s}(\sup_{\mathbb{R} } f_{r})^{1-s}}{s\varepsilon^{2-2s}}.$

从而

$\psi_{r, \varepsilon}(x)={\cal G}_s\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right)(x)-W x_{1}-\mu_{r, \varepsilon}\leq \frac{c_{s}\pi^{1-s}\kappa^{s}(\sup_{\mathbb{R} }f_{r})^{1-s}}{s\varepsilon^{2-2s}}-\mu_{r, \varepsilon}+C.$

一方面, 我们断言对任意的 $x\in D_{r}$, 有$\psi_{r, \varepsilon}(x)\geq 0$.否则, 假设在$D_{r}$上有 $\psi_{r, \varepsilon}(x)< 0$, 那么 $\Theta_{r,\varepsilon}\equiv 0$, 这与 $\int_{D_{r}} \Theta_{r,\varepsilon} {\rm d}x=\kappa$ 矛盾.因此

$\limsup _{\varepsilon, \nu \rightarrow 0^{+}} \varepsilon^{2-2 s} \mu_{r, \varepsilon}\leq \frac{c_{s}\pi^{1-s}\kappa^{s}(\sup_{\mathbb{R} }f_{r})^{1-s}}{s}<+\infty.$

另一方面, 由引理2.12知, 存在不依赖于 $\varepsilon$$R_{0}>\left[f_{r}(1)\pi\right]^{-\frac{1}{2}}$, 使得

$\int_{B_{R_{0}\varepsilon}(z_{r,\varepsilon})} \Theta_{r,\varepsilon} {\rm d}x\geq \frac{\kappa}{2}.$

则对任意的 $x\in B_{R_{0}\varepsilon}(z_{r,\varepsilon})$, 有

${\cal G}_s\Theta_{r,\varepsilon}(x)\geq \int_{B_{R_{0}\varepsilon}(z_{\varepsilon})} \frac{c_{s}}{\left|x-x^{\prime}\right|^{2-2s}}\Theta_{r,\varepsilon}(x^{\prime}) {\rm d} x^{\prime}\geq \frac{C_{s}}{\varepsilon^{2-2s}},$

其中 $C_{s}$ 是只与 $s$ 有关的常数.从而

$\psi_{r, \varepsilon}(x)\geq \frac{C_{s}}{\varepsilon^{2-2s}}-\mu_{r, \varepsilon}-C.$

由引理2.4的证明过程知 $\inf\limits_{D_{r}}\psi_{r, \varepsilon}\leq 1$ 因此结合 $R_{0}$ 的选取, 存在 $x_{r,\varepsilon}\in B_{R_{0}\varepsilon}(z_{r,\varepsilon})$, 使得 $\psi_{r, \varepsilon}(x_{r,\varepsilon})\leq 1$, 从而有

$1\geq \psi_{r, \varepsilon}(x_{r,\varepsilon})\geq \frac{C_{s}}{\varepsilon^{2-2s}}-\mu_{r, \varepsilon}-C.$

进而有

$\liminf _{\varepsilon, \nu \rightarrow 0^{+}} \varepsilon^{2-2 s} \mu_{r, \varepsilon}\geq C_{s}>0.$

故(2.67)式得证,同理可得(2.68)式.

2.7 ${\rm supp}(\Theta_{r,\varepsilon})$${\rm supp}(\Theta_{\ell,\nu})$的直径估计

本小节结合引理2.12和拉格朗日乘子的估计, 将得到涡对的支集估计.

引理 2.14$\frac{1}{2} \leq s<1$, 存在不依赖于 $\varepsilon$, $\nu$的正常数 $C$, 使得

$\operatorname{diam}\left(\operatorname{supp}\left(\Theta_{r, \varepsilon}\right)\right) \leq C \varepsilon$
$\operatorname{diam}\left(\operatorname{supp}\left(\Theta_{\ell, \nu}\right)\right) \leq C \nu$

$\Lambda >1$ 是一个待定的数.根据引理2.12可知, 对于每个 $\sigma _{r} \in \left ( 0,\kappa \right )$, 存在只依赖于 $\sigma _{r}$ 的正数 $R_{\sigma _r}$, 使得

$\int_{B_{R_{\sigma_r}\varepsilon}(z_{r,\varepsilon})} \Theta_{r,\varepsilon} {\rm d}x>\kappa -\sigma_{r}, uad \forall \varepsilon>0.$

$x\in D_{r}\backslash B_{2 \Lambda R_{\sigma_r} \varepsilon}\left(z_{r,\varepsilon}\right)$, 那么由重排不等式可得

$\begin{matrix}{\cal G}_s\Theta_{r,\varepsilon}(x)&=&c_{s}\int_{B_{R_{\sigma_r}\varepsilon}(z_{r,\varepsilon})} \frac{\Theta_{r,\varepsilon}(x^{\prime})}{\left|x-x^{\prime}\right|^{2-2s}} {\rm d} x^{\prime}+c_{s}\int_{D_{r}\backslash {B_{R_{\sigma_r}\varepsilon}(z_{r,\varepsilon})}} \frac{\Theta_{r,\varepsilon}(x^{\prime})}{\left|x-x^{\prime}\right|^{2-2s}} {\rm d} x^{\prime}\\&\leq&\bigg(\frac{c_{s}\kappa}{\left(\Lambda R_{\sigma_r}\right)^{2-2s}}+\frac{c_{s}\pi^{1-s}\sigma_r^{1-s}\kappa^s}{s}\bigg)\frac{1}{\varepsilon^{2-2s}}. \end{matrix}$

另一方面, 由引理2.13知, 存在不依赖于 $\varepsilon$ 的正数 $\eta_{0}$, 使得当 $\varepsilon$ 充分小时, 有

$\begin{equation} \mu_{\varepsilon}\geq \frac{\eta_{0}}{\varepsilon^{2-2s}}.\end{equation} $

$\sigma_r=\sigma_0$ 满足

$\frac{c_{s}\pi^{1-s}\sigma_r^{1-s}\kappa^s}{s}\leq \frac{\eta_{0}}{3}.$

$\Lambda=\frac{1}{R_{\sigma_0}}\left(\frac{3c_{s}\kappa}{\eta_{0}}\right)^{\frac{1}{2-2s}}.$

结合(2.71)和(2.72)式, 我们可得当 $\varepsilon$ 充分小时, 对任意的 $x\in D_{r}\backslash B_{2 \Lambda R_{\sigma_r} \varepsilon}\left(z_{\varepsilon}\right)$

$\psi_{r, \varepsilon}(x)\leq {\cal G}_s\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right)(x)-\mu_{r, \varepsilon}+C\leq -\frac{\eta_{0}}{3}\frac{1}{\varepsilon^{2-2s}}+C<0.$

从而对任意的$x\in D_{r}\backslash B_{2 \Lambda R_{\sigma_r} \varepsilon}\left(z_{r,\varepsilon}\right)$$\Theta_{r,\varepsilon}(x)=0$, 因此${\rm supp}\left(\Theta_{r, \varepsilon}\right)\subseteq B_{2 \Lambda R_{\sigma_r} \varepsilon}(z_{r,\varepsilon}).$$C=4\Lambda R_{\sigma_r}$, 则(2.69)式成立, 同理可得(2.70)式.

2.8 $(\Theta_{r,\varepsilon},\Theta_{\ell,\nu})$的渐近行为

现在我们寻找涡对 $\left(\Theta_{r, \varepsilon},\Theta_{\ell, \nu}\right)$ 的极限位置.根据引理2.14可知涡对支集的直径会随着 $\max \{\varepsilon, \nu\} $ 趋于零而趋于零, 这说明其支集可能会收缩到一个点.为了研究这种点所在的位置, 定义涡的中心

$x^{r, \varepsilon}:=\frac{1}{\kappa } \int_{\mathbb{R} ^{2}} x \Theta _{r, \varepsilon}(x) {\rm d} x, uad x^{\ell, \nu}:=\frac{1}{\kappa }\int_{\mathbb{R} ^{2}} x \Theta _{\ell, \nu}(x) {\rm d} x.$

根据 $\Theta_{r, \varepsilon}$$\Theta_{\ell, \nu} $ 关于 $x_{2}$ 轴对称的事实, 可知 $x^{r, \varepsilon}$$x^{\ell, \nu}$ 均在 $x_{1}$ 轴上.下面的引理表明正涡与负涡将会互相平衡, 并将趋近于两个点涡之间的平衡状态.

引理 2.15$\max\left\{\varepsilon,\nu\right\}\rightarrow 0$ 时成立

${\rm dist}\left({\rm supp}\left(\Theta _{r, \varepsilon}\right), {\rm supp}\left(\Theta _{\ell, \nu}\right)\right) \rightarrow 2 L_{*}.$

首先说明 $x^{r, \varepsilon}\in {\rm supp}\left(\Theta _{r, \varepsilon}\right)$$x^{\ell, \nu}\in {\rm supp}\left(\Theta _{\ell, \nu}\right)$.事实上, 对任意的 $x\in {\rm supp}\left(\Theta _{r, \varepsilon}\right)$

$\begin{matrix} \left|x-x^{r, \varepsilon}\right|&=&\left|\frac{1}{\kappa} \int_{\mathbb{R} ^{2}} x \Theta _{r, \varepsilon}(y){\rm d}y-\frac{1}{\kappa} \int_{\mathbb{R} ^{2}} y \Theta _{r, \varepsilon}(y){\rm d}y\right|\\ &\leq& \max_{x,y \in {\rm supp}\left(\Theta_{r, \varepsilon}\right)} \left|x-y\right| \leq C\varepsilon. \end{matrix}$

因此$x^{r, \varepsilon}\in {\rm supp}\left(\Theta _{r, \varepsilon}\right)$, 同理可得 $x^{\ell, \nu}\in {\rm supp}\left(\Theta _{\ell, \nu}\right)$.

$d_{\varepsilon,\nu}=\big|x_{1}^{r, \varepsilon}-x_{1}^{\ell, \nu}\big|$. 结合引理2.14, 只需证明当$\max\left\{\varepsilon,\nu\right\}\rightarrow 0$$d_{\varepsilon,\nu}\rightarrow 2 L_{*}$.利用反证法, 假设当$\max\left\{\varepsilon,\nu\right\}\rightarrow 0$$d_{\varepsilon,\nu}\rightarrow d_{*}\ne 2 L_{*}$.

$\overline{\Theta } _{r,\varepsilon }=\Theta _{r, \varepsilon}\left(x^{r, \varepsilon}-b_{1}+\cdot \right)\in {\cal A}_{r},uad \overline{\Theta } _{\ell,\nu}=\Theta _{r, \varepsilon}\left(x^{\ell,\nu}-b_{2}+\cdot \right)\in {\cal A}_{\ell},$

那么有 ${\cal E}_{\varepsilon, \nu}\left(\Theta _{r, \varepsilon}, \Theta _{\ell, \nu}\right)\geq {\cal E}_{\varepsilon, \nu}\left(\overline{\Theta } _{r,\varepsilon }, \overline{\Theta } _{\ell,\nu}\right)$.通过简单的计算可得

$E_{s}\left(\overline{\Theta } _{r,\varepsilon }\right)-{\cal J}_{r, \varepsilon}\left(\overline{\Theta } _{r,\varepsilon }\right)=E_{s}\left(\Theta_{r,\varepsilon }\right)-{\cal J}_{r, \varepsilon}\left(\Theta_{r,\varepsilon }\right),$
$E_{s}\left(\overline{\Theta } _{\ell,\nu}\right)-{\cal J}_{\ell,\nu}\left(\overline{\Theta } _{\ell,\nu}\right)=E_{s}\left(\Theta_{\ell,\nu}\right)-{\cal J}_{\ell,\nu}\left(\Theta_{\ell,\nu}\right).$

从而有

$\begin{matrix}&&\int_{\mathbb{R} ^{2}}\int_{\mathbb{R} ^{2}}\frac{c_{s}}{\left|x-x^{\prime}\right|^{2-2s}}\Theta_{r,\varepsilon }(x)\Theta_{\ell,\nu}(x^{\prime}){\rm d}x{\rm d}x^{\prime}+W {\cal P}\left(\Theta_{r,\varepsilon }-\Theta_{\ell,\nu}\right)\\&\leq& \int_{\mathbb{R} ^{2}}\int_{\mathbb{R} ^{2}}\frac{c_{s}}{\left|x-x^{\prime}\right|^{2-2s}}\overline{\Theta } _{r,\varepsilon }(x)\overline{\Theta } _{\ell,\nu}(x^{\prime}){\rm d}x{\rm d}x^{\prime}+W {\cal P}\left(\overline{\Theta } _{r,\varepsilon }-\overline{\Theta } _{\ell,\nu}\right).\end{matrix}$

$\max\left\{\varepsilon,\nu\right\}\rightarrow 0$, 得

$\frac{\kappa^{2}c_{s}}{d_{*}^{2-2s}}+\kappa Wd_{*}\leq \frac{\kappa^{2}c_{s}}{\left(2L_{*}\right)^{2-2s}}+2\kappa WL_{*}.$

$h_{s}(\tau)=\frac{\kappa^{2} c_{s}}{\tau^{2-2 s}}+\kappa W \tau, \tau \in \mathbb{R} _{+}.$

通过计算知, 函数 $h_{s}(\tau)$ 有唯一的极小值点

$\tau=2\left(\frac{\kappa}{4 \pi W} \frac{\Gamma(2-s)}{\Gamma(s)}\right)^{\frac{1}{3-2 s}}=2L_{*},$

从而$d_{*}=2 L_{*}$, 这与假设矛盾.

注意到能量泛函 ${\cal E}_{\varepsilon, \nu}$ 关于 $x_1$ 方向具有平移不变性, 即对于任意的数 $\tau$, 设 $\left(\Theta _{r}, \Theta _{\ell}\right)$, $\left(\Theta _{r}\left(\cdot -\tau e_{1}\right), \Theta _{\ell}\left(\cdot -\tau e_{1}\right)\right)\in {\cal A}_{r} \times {\cal A}_{\ell}$, 那么

${\cal E}_{\varepsilon, \nu}\left(\Theta _{r}, \Theta _{\ell}\right)={\cal E}_{\varepsilon, \nu}\left(\Theta _{r}\left(\cdot -\tau e_{1}\right), \Theta _{\ell}\left(\cdot -\tau e_{1}\right)\right).$

结合引理2.14和引理, 可见在 $x_{1}$ 方向上进行适当的平移之后(如果需要的话), 在这里及下文中, 我们可以假设对所有充分小的 $\varepsilon$$\nu$, 使得

$x^{r, \varepsilon}=-x^{\ell, \nu}.$

因此当 $\max\left\{\varepsilon,\nu\right\}\rightarrow 0$$x^{r, \varepsilon}\rightarrow b_{1}$, $x^{\ell, \nu}\rightarrow b_{2}$, 从而我们有以下引理.

引理 2.16$\max\left\{\varepsilon,\nu\right\}\rightarrow 0$ 时, 在测度的意义下成立

$\Theta_{r, \varepsilon}(x) \rightharpoonup \kappa {\delta}\left(x-b_{1}\right), uad \Theta_{\ell, \nu}(x) \rightharpoonup \kappa {\delta}\left(x-b_{2}\right).$

对任意的 $\phi \in C_{0}^{\infty}\left(\mathbb{R} ^{2}\right)$

$\begin{matrix} I&=&\left|\int_{\mathbb{R} ^{2}}\Theta_{r, \varepsilon}(x)\phi(x){\rm d}x-\kappa \phi(b_{1})\right|\\ &=&\left|\int_{\mathbb{R} ^{2}}\Theta_{r, \varepsilon}(x)\phi(x){\rm d}x-\int_{\mathbb{R} ^{2}}\Theta_{r, \varepsilon}(x) \phi(b_{1}){\rm d}x\right|\\ &\leq &\int_{{\rm supp}\left(\Theta _{r, \varepsilon}\right)}\Theta_{r, \varepsilon}(x)\left|\phi(x)-\phi(b_{1})\right|{\rm d}x\\ &\leq& \kappa \max_{x \in {\rm supp}\left(\Theta _{r, \varepsilon}\right)} \left|\phi(x)-\phi(b_{1})\right|. \end{matrix}$

注意到对任意的 $x \in {\rm supp}\left(\Theta _{r, \varepsilon}\right)$, 由(2.73)式得

$\begin{matrix} \left|x-b_{1}\right| \leq \left|x-x^{r, \varepsilon}\right|+\left|x^{r, \varepsilon}-b_{1}\right| \leq C\varepsilon+\left|x^{r, \varepsilon}-b_{1}\right| \rightarrow 0 uad \left(\varepsilon \rightarrow 0\right). \end{matrix}$

根据 $\phi$ 的连续性, 容易验证当 $\varepsilon \rightarrow 0$ 时, 有 $I \rightarrow 0$. 因此按测度的意义有

$\Theta_{r,\varepsilon}(x)\rightharpoonup\kappa{\delta}\left(x-b_{1}\right).$

同理可得当 $\max\left\{\varepsilon,\nu\right\}\rightarrow 0$ 时, 有$\Theta_{\ell,\nu}(x)\rightharpoonup\kappa{\delta}\left(x-b_{2}\right).$

2.9 弱解

根据我们已经得到的渐近估计, 我们将(2.32)和(2.33)式进行如下的改进.

引理 2.17 对于所有充分小的 $\varepsilon$$\nu$, 成立

$ \begin{equation} \Theta _{r, \varepsilon}(x)=\frac{1}{\varepsilon^{2}} f_{r}\left({\cal G}_s\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right)(x)-W x_{1}-\mu_{r, \varepsilon}\right), uad \mbox{ a.e. } x \in \Pi_{r}, \end{equation} $
$\begin{equation} \Theta _{\ell, \nu}(x)=\frac{1}{\nu^{2}} f_{\ell}\left(-{\cal G}_s\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right)(x)+W x_{1}-\mu_{\ell, \nu}\right), uad \mbox{ a.e. } x \in \Pi_{\ell}. \end{equation}$

根据引理2.14和引理2.16可得, 对于充分小的 $\varepsilon$$\nu$

${\rm supp}\left(\Theta_{r, \varepsilon}\right)\subseteq B_{\frac{L_{*}}{4}}(b_{1}),uad {\rm supp}\left(\Theta_{\ell, \nu}\right)\subseteq B_{\frac{L_{*}}{4}}(b_{2}).$

$\varepsilon$$\nu$ 充分小时, 对任意的 $x \in \Pi_{r} \backslash D_{r}$, 一方面有

$\left|{\cal G}_{s}\left(\Theta _{r, \varepsilon}-\Theta _{\ell, \nu}\right)(x)\right|\leq C.$

另一方面, 由引理2.13知$\mu_{r, \varepsilon} \rightarrow +\infty$. 从而易得对任意的 $x \in \Pi_{r} \backslash D_{r}$$\psi_{r, \varepsilon}\leq 0$. 因此结合(2.32)式可知(2.77)式成立, 同理可得(2.78)式成立. 证毕.

$\Theta_{\varepsilon, \nu}=\Theta_{r, \varepsilon}-\Theta_{\ell, \nu}$.我们现在可以说明 $\Theta_{\varepsilon, \nu}$ 是(1.5)在(1.6)意义下的解.

引理 2.18$\frac{1}{2}\leq s<1$, 当 $\varepsilon$$\nu$ 充分小时, 成立

$\begin{equation} \int_{\mathbb{R} ^2}\Theta_{\varepsilon, \nu}\nabla^\perp({\cal G}_s\Theta_{\varepsilon, \nu}-Wx_{1})\cdot\nabla\varphi {\rm d}x=0,uad \forall \varphi\in C_0^{\infty}(\mathbb{R} ^2). \end{equation}$

根据引理2.17的证明过程容易得到当 $\varepsilon$$\nu$ 充分小时

$F_{r}\left(\psi_{r, \varepsilon}\right)=0,uad \forall x\in \partial D_{r} ; F_{\ell}\left(\psi_{\ell, \nu}\right)=0,uad \forall x\in \partial D_{\ell}.$

则对任意的 $\varphi\in C_0^{\infty}(\mathbb{R} ^2)$, 利用分部积分可得

$\begin{matrix}&&\int_{\mathbb{R} ^2}\Theta_{\varepsilon, \nu}\nabla^\perp({\cal G}_s\Theta_{\varepsilon, \nu}-Wx_{1})\cdot\nabla\varphi {\rm d}x\\ &=&\int_{D_{r}}\Theta_{r, \varepsilon}\nabla^\perp\psi_{r, \varepsilon}\cdot\nabla\varphi {\rm d}x + \int_{D_{\ell}}\Theta_{\ell, \nu}\nabla^\perp\psi_{\ell, \nu}\cdot\nabla\varphi {\rm d}x\\ &=&\int_{D_{r}}\left[-\frac{1}{\varepsilon^2}F_{r}(\psi_{r, \varepsilon})\partial_{x_{2}} \partial_{x_{1}} \varphi+\frac{1}{\varepsilon^2}F_{r}(\psi_{r, \varepsilon})\partial_{x_{1}} \partial_{x_{2}} \varphi\right]{\rm d}x\\&& +\int_{D_{\ell}}\left[-\frac{1}{\nu^2}F_{\ell}(\psi_{\ell, \nu})\partial_{x_{2}} \partial_{x_{1}} \varphi+\frac{1}{\nu^2}F_{\ell}(\psi_{\ell, \nu})\partial_{x_{1}} \partial_{x_{2}} \varphi\right]{\rm d}x\\ &=&0. \end{matrix}$

证毕.

定理1.1的证明$\theta_{\varepsilon, \nu}(x, t)=\Theta_{\varepsilon, \nu}\left(x_{1}, x_{2}+Wt\right)$, 则由以上引理即可得到定理1.1的证明.

3 涡度在伸缩变化下的渐近行为

在上一节中证明了 $\Theta_{r,\varepsilon}$$\Theta_{\ell,\nu}$ 在测度的意义下收敛到一个Dirac 测度, 但这个结果是比较粗糙的. 本节将给出更精确的渐近行为, 为此定义

$\vartheta_{r,\varepsilon}(x)=\varepsilon^2\Theta_{r,\varepsilon}(x^{r,\varepsilon}+\varepsilon x),uad \vartheta_{\ell,\nu}(x)=\nu^2\Theta_{\ell,\nu}(x^{\ell,\nu}+\nu x),uad \forall x \in \mathbb{R} ^{2}.$

那么

$0\leq \vartheta_{r,\varepsilon}\leq \sup_{\mathbb{R} }f_{r},uad \int_{\mathbb{R} ^{2}}\vartheta_{r,\varepsilon}{\rm d}x=\kappa,uad \int_{\mathbb{R} ^{2}}x \vartheta_{r,\varepsilon}{\rm d}x=0,uad {\rm supp}(\vartheta_{r,\varepsilon}) \subseteq B_{C}(0),$
$0\leq \vartheta_{\ell,\nu}\leq \sup_{\mathbb{R} }f_{\ell},uad \int_{\mathbb{R} ^{2}}\vartheta_{\ell,\nu}{\rm d}x=\kappa,uad \int_{\mathbb{R} ^{2}}x \vartheta_{\ell,\nu}{\rm d}x=0,uad {\rm supp}(\vartheta_{\ell,\nu}) \subseteq B_{C}(0).$

我们用 $\vartheta_{r,\varepsilon}^{*}$$\vartheta_{\ell,\nu}^{*}$ 分别表示$\vartheta_{r,\varepsilon}$$\vartheta_{\ell,\nu}$的勒贝格重排函数, 且它们均是以原点为心的径向对称的递减函数.下面将考虑 $\vartheta_{r,\varepsilon}$$\vartheta_{\ell,\nu}$ 的渐近行为.

引理 3.1$\frac{1}{2}\leq s<1$, 当 $\max\left\{\varepsilon,\nu\right\}\rightarrow0$ 时, 在 $L^{\infty}(\mathbb{R} ^{2})$ 的弱 $*$ 拓扑下, $\left\{\vartheta_{r,\varepsilon}\right\}_{\varepsilon>0}$$\left\{\vartheta_{\ell,\nu}\right\}_{\nu>0}$ 的极限点必是一个径向递减的函数.

因为 $\left\{\vartheta_{r,\varepsilon}\right\}_{\varepsilon>0}$, $\left\{\vartheta_{\ell,\nu}\right\}_{\nu>0}$, $\left\{\vartheta_{r,\varepsilon}^{*}\right\}_{\varepsilon>0}$$\big\{\vartheta_{\ell,\nu}^{*}\big\}_{\nu>0}$ 均是$L^{\infty}(\mathbb{R} ^{2})$ 中的有界序列, 相应的存在子序列(仍然记为$\left\{\vartheta_{r,\varepsilon}\right\}_{\varepsilon>0}$,$\left\{\vartheta_{\ell,\nu}\right\}_{\nu>0}$, $\left\{\vartheta_{r,\varepsilon}^{*}\right\}_{\varepsilon>0}$$\big\{\vartheta_{\ell,\nu}^{*}\big\}_{\nu>0}$), 以及 $\bar{\vartheta}_{r}$, $\bar{\vartheta}_{\ell}$, $\bar{\vartheta}_{r}^{*}$, $\bar{\vartheta}_{\ell}^{*}\in L^{\infty}(\mathbb{R} ^{2})$, 使得当 $\max\left\{\varepsilon,\nu\right\}\rightarrow0$ 时, 在$L^{\infty}(\mathbb{R} ^{2})$ 的弱 $*$ 拓扑下有

$\vartheta_{r,\varepsilon}\rightharpoonup \bar{\vartheta}_{r},uad \vartheta_{r,\varepsilon}^{*}\rightharpoonup \bar{\vartheta}_{r}^{*},uad \vartheta_{\ell,\nu}\rightharpoonup \bar{\vartheta}_{\ell},uad \vartheta_{\ell,\nu}^{*}\rightharpoonup \bar{\vartheta}_{\ell}^{*}.$

由 Riesz 重排不等式得

$c_{s}\int_{\mathbb{R} ^{2}}\int_{\mathbb{R} ^{2}}\frac{\vartheta_{r,\varepsilon}(x)\vartheta_{r,\varepsilon}(x^\prime)}{\left|x-x^\prime\right|^{2-2s}}{\rm d}x{\rm d}x^\prime\leq c_{s}\int_{\mathbb{R} ^{2}}\int_{\mathbb{R} ^{2}}\frac{\vartheta_{r,\varepsilon}^{*}(x)\vartheta_{r,\varepsilon}^{*}(x^\prime)}{\left|x-x^\prime\right|^{2-2s}}{\rm d}x{\rm d}x^\prime,$

${\cal I}_{r,s}(\vartheta_{r,\varepsilon})\leq {\cal I}_{r,s}(\vartheta_{r,\varepsilon}^{*})$, 同理有 ${\cal I}_{\ell,s}(\vartheta_{\ell,\nu})\leq {\cal I}_{\ell,s}(\vartheta_{\ell,\nu}^{*})$.则当 $\max\left\{\varepsilon,\nu\right\}\rightarrow0$

$\begin{equation} {\cal I}_{r,s}(\bar{\vartheta}_{r})\leq {\cal I}_{r,s}(\bar{\vartheta}_{r}^{*}),uad {\cal I}_{\ell,s}(\bar{\vartheta}_{\ell})\leq {\cal I}_{\ell,s}(\bar{\vartheta}_{\ell}^{*}). \end{equation}$

定义

$\Theta_{r,\varepsilon}^{*}(x):=\varepsilon^{-2}\vartheta_{r,\varepsilon}^{*}\left(\varepsilon^{-1}\left(x-x^{r,\varepsilon}\right)\right),$
$\Theta_{\ell,\nu}^{*}(x):=\nu^{-2}\vartheta_{\ell,\nu}^{*}\left(\nu^{-1}\left(x-x^{\ell,\nu}\right)\right).$

$\max\left\{\varepsilon,\nu\right\}\rightarrow0$ 时, 通过直接计算可得

${\cal E}_{\varepsilon, \nu}\left(\Theta_{r,\varepsilon}, \Theta_{\ell,\nu}\right)=\frac{{\cal I}_{r,s}(\vartheta_{r,\varepsilon})}{2\varepsilon^{2-2s}}+\frac{{\cal I}_{\ell,s}(\vartheta_{\ell,\nu})}{2\nu^{2-2s}}-{\cal J}_{r,\varepsilon}(\vartheta_{r,\varepsilon})-{\cal J}_{\ell,\nu}(\vartheta_{\ell,\nu})-\bigg(\frac{c_{s}\kappa^2}{\left(2L^{*}\right)^{2-2s}}+2\kappa WL^{*}\bigg)+o(1),$
${\cal E}_{\varepsilon, \nu}\left(\Theta_{r,\varepsilon}^{*}, \Theta_{\ell,\nu}^{*}\right)=\frac{{\cal I}_{r,s}(\vartheta_{r,\varepsilon}^{*})}{2\varepsilon^{2-2s}}+\frac{{\cal I}_{\ell,s}(\vartheta_{\ell,\nu})^{*}}{2\nu^{2-2s}}-{\cal J}_{r,\varepsilon}(\vartheta_{r,\varepsilon}^{*})-{\cal J}_{\ell,\nu}(\vartheta_{\ell,\nu}^{*})-\bigg(\frac{c_{s}\kappa^2}{\left(2L^{*}\right)^{2-2s}}+2\kappa WL^{*}\bigg)+o(1).$

因为${\cal J}_{r,\varepsilon}(\vartheta_{r,\varepsilon})={\cal J}_{r,\varepsilon}(\vartheta_{r,\varepsilon}^{*})$, ${\cal J}_{\ell,\nu}(\vartheta_{\ell,\nu})={\cal J}_{\ell,\nu}(\vartheta_{\ell,\nu}^{*})$, 所以我们可以由${\cal E}_{\varepsilon, \nu}\left(\Theta_{r,\varepsilon}^{*}, \Theta_{\ell,\nu}^{*}\right)\leq {\cal E}_{\varepsilon, \nu}\left(\Theta_{r,\varepsilon}, \Theta_{\ell,\nu}\right)$ 推出

$\frac{{\cal I}_{r,s}(\vartheta_{r,\varepsilon}^{*})}{2\varepsilon^{2-2s}}+\frac{{\cal I}_{\ell,s}(\vartheta_{\ell,\nu}^{*})}{2\nu^{2-2s}}\leq \frac{{\cal I}_{r,s}(\vartheta_{r,\varepsilon})}{2\varepsilon^{2-2s}}+\frac{{\cal I}_{\ell,s}(\vartheta_{\ell,\nu})}{2\nu^{2-2s}}+o(1),$

从而当$\max\left\{\varepsilon,\nu\right\}\rightarrow0$ 时, 有

$\frac{{\cal I}_{r,s}(\bar{\vartheta}_{r}^{*})}{2\varepsilon^{2-2s}}+\frac{{\cal I}_{\ell,s}(\bar{\vartheta}_{\ell}^{*})}{2\nu^{2-2s}}\leq \frac{{\cal I}_{r,s}(\bar{\vartheta}_{r})}{2\varepsilon^{2-2s}}+\frac{{\cal I}_{\ell,s}(\bar{\vartheta}_{\ell})}{2\nu^{2-2s}}.$

结合(3.1)式可得

$ {\cal I}_{r,s}(\bar{\vartheta}_{r})={\cal I}_{r,s}(\bar{\vartheta}_{r}^{*}),uad {\cal I}_{\ell,s}(\bar{\vartheta}_{\ell})={\cal I}_{\ell,s}(\bar{\vartheta}_{\ell}^{*}).$

根据文献[23,引理3.2]可知, 存在 $\mathbb{R} ^{2}$ 上的平移 ${\cal T}_1$${\cal T}_2$, 使得 ${\cal T}_1 \circ \bar{\vartheta}_{r}=\bar{\vartheta}_{r}^{*},$$ {\cal T}_2 \circ \bar{\vartheta}_{\ell}=\bar{\vartheta}_{\ell}^{*}$.注意到

$\int_{R^2}x \bar{\vartheta}_{r}{\rm d}x=\int_{R^2}x \bar{\vartheta}_{r}^{*}{\rm d}x=0,$
$\int_{R^2}x \bar{\vartheta}_{\ell}{\rm d}x=\int_{R^2}x \bar{\vartheta}_{\ell}^{*}{\rm d}x=0.$

$\bar{\vartheta}_{r}=\bar{\vartheta}_{r}^{*}$, $\bar{\vartheta}_{\ell}=\bar{\vartheta}_{\ell}^{*}$.

参考文献

Constantin P, Majda A J, Tabak E.

Formation of strong fronts in the 2-D quasigeostrophic thermal active scalar

Nonlinearity, 1994, 7(6): 1495-1533

DOI:10.1088/0951-7715/7/6/001      URL     [本文引用: 2]

Held I M, Pierrehumbert R T, Garner S T, Swanson K L.

Surface quasi-geostrophic dynamics

J Fluid Mech, 1995, 282: 1-20

DOI:10.1017/S0022112095000012      URL     [本文引用: 1]

The dynamics of quasi-geostrophic flow with uniform potential vorticity reduces to the evolution of buoyancy, or potential temperature, on horizontal boundaries. There is a formal resemblance to two-dimensional flow, with surface temperature playing the role of vorticity, but a different relationship between the flow and the advected scalar creates several distinctive features. A series of examples are described which highlight some of these features: the evolution of an elliptical vortex; the start-up vortex shed by flow over a mountain; the instability of temperature filaments; the ‘edge wave’ critical layer; and mixing in an overturning edge wave. Characteristics of the direct cascade of the tracer variance to small scales in homogeneous turbulence, as well as the inverse energy cascade, are also described. In addition to its geophysical relevance, the ubiquitous generation of secondary instabilities and the possibility of finite-time collapse make this system a potentially important, numerically tractable, testbed for turbulence theories.

Lapeyre G.

Surface quasi-geostrophy

Fluids, 2017, 2(1): 7

DOI:10.3390/fluids2010007      URL     [本文引用: 1]

Majda A J, Bertozzi A L.

Vorticity and Incompressible Flow

Cambridge: Cambridge UP, 2002

[本文引用: 1]

Castro A, Córdoba D, Gómez-Serrano J.

Global Smooth Solutions for the Inviscid SQG Equation

Providence, RI: Amer Math Soc, 2020

[本文引用: 2]

Kiselev A, Ryzhik L, Yao Y, Zlato A.

Finite time singularity for the modified SQG patch equation

Ann Math, 2016, 184: 909-948

DOI:10.4007/annals.2016.184-3      URL     [本文引用: 1]

Gravejat P, Smets D.

Smooth travelling-wave solutions to the inviscid surface quasigeostrophic equation

Int Math Res Not, 2019, 6: 1744-1757

[本文引用: 1]

Godard-Cadillac L.

Smooth traveling-wave solutions to the inviscid surface quasi-geostrophic equations

C R Math Acad Sci Paris, 2021, 359: 85-98

[本文引用: 1]

Ao W, Dávila J, Pino L D, et al.

Travelling and rotating solutions to the generalized inviscid surface quasi-geostrophic equation

Trans Amer Math Soc, 2021, 374(9): 6665-6689

DOI:10.1090/tran/2021-374-09      [本文引用: 1]

Castro A, Córdoba D, Gómez-Serraon J.

Existence and regularity of rotating global solutions for the generalized surface quasi-geostrophic equations

Duke Math J, 2016, 165(5): 935-984

[本文引用: 1]

Hassainia Z, Hmidi T.

On the V-states for the generalized quasi-geostrophic equations

Commun Math Phys, 2015, 337(1): 321-377

DOI:10.1007/s00220-015-2300-5      URL     [本文引用: 1]

Hassainia Z, Masmoudi W N, Miles H.

Global bifurcation of rotating vortex patches

Commun Pure Appl Math, 2020, 73(9): 1933-1980

DOI:10.1002/cpa.v73.9      URL     [本文引用: 1]

Buckmaster T, Shkoller S, Vicol V.

Nonuniqueness of weak solutions to the SQG equation

Commun Pure Appl Math, 2019, 72(9): 1809-1874

DOI:10.1002/cpa.v72.9      URL     [本文引用: 1]

Marchand F.

Existence and regularity of weak solutions to the quasi-geostrophic equations in the spaces $L^p$ or $H^{-\frac{1}{2}}$

Commun Math Phys, 2008, 277(1): 45-67

DOI:10.1007/s00220-007-0356-6      URL     [本文引用: 1]

Resnick S.

Dynamical Problems in Non-linear Advective Partial Differential Equations

Chicago: University of Chicago, 1995

[本文引用: 1]

Cao D, Qin G, Zhan W, Zou C.

Existence of traveling asymmetric vortex pairs in an ideal fluid

J Differ Equations, 2023, 351(5): 131-155

DOI:10.1016/j.jde.2022.12.024      URL     [本文引用: 1]

Cao D, Qin G, Zhan W, Zou C.

On the global classical solutions for the generalized SQG equation

J Funct Anal, 2022, 283(2): 109503

DOI:10.1016/j.jfa.2022.109503      URL     [本文引用: 2]

Arnol'd V I. Conditions for Nonlinear Stability of Stationary Plane Curvilinear Flows of an Ideal Fluid// Givental A, Khesin B, Varchenko A, et al. Vladimir I. Arnold - Collected Works. Hedelberg: Springer, 1965: 19-23

[本文引用: 1]

Cao D, Lai S, Zhan W.

Traveling vortex pairs for 2D incompressible Euler equations

Calc Var Partial Differential Equations, 2021, 60(5): 16

DOI:10.1007/s00526-020-01890-7      [本文引用: 2]

Rockafellar T. Convex Analysis. Princeton, NJ: Princeton Univ Princeton Press, 1970

[本文引用: 1]

Lions P L.

The concentration-compactness principle in the calculus of variations

The locally compact case I. Ann Inst He Poincaré, 1984, 1(2): 109-145

[本文引用: 1]

Lieb E H, Loss M.

Analysis. Providence

RI: American Mathematical Society, 2001

[本文引用: 1]

Burchard A, Guo Y.

Compactness via symmetrization

J Funct Anal, 2004, 214: 40-73

DOI:10.1016/j.jfa.2004.04.005      URL     [本文引用: 1]

/