数学杂志  2021, Vol. 41 Issue (3): 219-226   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
朱茂春
刘杰
一维直线上的奇异型Trudinger-Moser不等式
朱茂春, 刘杰    
江苏大学数学科学学院, 江苏 镇江 212013
摘要:本文研究了一维直线上的奇异型Trudinger-Moser不等式.利用分数次Sobolev空间上函数的Green表示公式,得到了一类奇异型Trudinger-Moser不等式.进一步利用合适的测试函数序列验证了不等式中常数的最佳性.这一结果将高维空间上的奇异型Trudinger-Moser不等式推广到了一维情形.
关键词Trudinger-Moser不等式    分数次Sobolev空间    重排    最佳常数    
A FRACTIONAL SINGULAR TRUDINGER-MOSER INEQUALITY IN DIMENSION ONE
ZHU Mao-chun, LIU Jie    
School of Mathematical Sciences, Jiangsu University, Zhenjiang 212013, China
Abstract: This paper is devoted to studying a singular fractional Trudinger-Moser inequality in dimension one. By the Green representation formula for functions in the fractional Sobolev spaces, we get a singular fractional Trudinger-Moser type inequality. Furthermore, by using a suitable test functions sequence, we can show that the constant in the inequality is optimal. This result extends the singular Trudinger-Moser inequality in the high-dimensional spaces to the space of dimension one.
Keywords: Trudinger-Moser inequality     fractional Sobolev space     rearrangement     optimal constant    
1 引言

$ \Omega \subset R^{n} $是具有$ C^{\infty} $的有界区域, $ W_{0}^{1, p}(\Omega) $是定义在$ \Omega $上的Sobolev空间, 其范数为$ \|u\|_{W^{1, p}} = \left(\int_{\Omega}\left(|\nabla u|^{p}+|u|^{p}\right) d x\right)^{\frac{1}{p}} $. 由经典的Sobolev嵌入定理可知($ n \geq 2 $):

(1) 当$ p<n $时: $ W_{0}^{1, n}(\Omega) $嵌入到$ L^{\frac{n p}{n-p}}(\Omega) $;

(2) 当$ p>n $时: $ W_{0}^{1, n}(\Omega) $嵌入到$ C^{\alpha}(\Omega) $, 其中$ \alpha = 1-\frac{n}{p} $;

(3) 当$ p = n $时: $ W_{0}^{1, n}(\Omega) $嵌入到任意的$ L^{q}(\Omega) $, 其中$ q \geq 1 $,

但是, 在$ p = n $时, 并不能得到$ W_{0}^{1, n}(\Omega) \subset L^{\infty}(\Omega) $. 1967年Trudinger通过幂级数展开法建立了一个最佳嵌入不等式, 即存在$ \alpha_{0}>0 $, 只要$ \alpha<\alpha_{0} $, 就有

$ \sup\limits _{\|u\|_{W_{0}^{1, n}(\Omega)} \leq 1} \int_{\Omega} \exp \left(\alpha|u|^{\frac{n}{n-1}}\right) d x<+\infty. $

1971年, Moser[1]通过对称重排方法确定了使得上式成立的最大常数$ \alpha_{n} $, $ \alpha_{n} = n \omega_{n-1}^{\frac{1}{n-1}} $, 其中$ \omega_{n-1} $$ \mathbb{R}^{n} $上单位球面的测度. 即当$ 0 \leq \alpha \leq \alpha_{n} $时, 上式成立; 但$ \alpha>\alpha_{n} $时, 上式中的上确界可以任意大. 现在称Moser得到的结果为Trudinger-Moser不等式, 即

$ \sup\limits _{\|\nabla u\|_{n} \leq 1} \int_{\Omega}\left\{\exp \left(\alpha_{\mathrm{n}}|u|^{\frac{n}{n-1}}\right)-1\right\} d x<C(\Omega, \alpha). $

后来Adimurthi和K.Sandeep[2]建立了有界区域上Dirichlet范数约束下的奇异型Trudinger-Moser不等式: 即当且仅当$ \alpha $$ \beta $满足$ \frac{\alpha}{\alpha_{n}}+\frac{\beta}{n} \leq 1 $, $ \alpha>0 $, $ \beta \in[0, n) $时, 有

$ \sup\limits _{\|\nabla u\|_{n} \leq 1} \int_{\Omega} \frac{\exp \left(\alpha|u|^{\frac{n}{n-1}}\right)}{|x|^{\beta}} d x<\infty. $

关于Trudinger-Moser不等式的其他重要研究可以参见文献[3-8].

一个自然的问题是当$ n = 1 $时, 有没有类似的Trudinger-Moser不等式呢? 2015年, Martinazzi, Iula以及Maalaoui在文献[9]中给出了一个肯定的回答, 他们得到了定义在直线上的分数次Sobolev空间上的Trudinger-Moser不等式, 即若$ p \in(1, \infty) $, 则对于任何满足$ |I|<\infty $的开区间$ I \subset \mathbb{R} $, 当$ \alpha \leq \alpha_{p} $时, 有

$ \sup\limits _{u \in \tilde{H}^{\frac{1}{p}, p}(I) , \left\|(-\Delta)^{\frac{1}{2p} } u\right\|_{L^{p}(I)} \leq 1} \int_{\mathbb{R}} \exp \left(\alpha|u|^{\frac{p}{p-1}}\right) d x = C_{p}|I| $

成立, 其中$ \alpha_{p} = \frac{1}{2}\left(2 \Gamma\left(\frac{1}{p}\right) \cos \left(\frac{\pi}{2 p}\right)\right)^{\frac{p}{p-1}} $, $ \Gamma(z) = \int_{0}^{+\infty} t^{z-1} e^{-t} d t $.

进一步, 他们证明了常数$ \alpha_{p} $是最佳的, 即当$ \alpha>\alpha_{p} $时, 上式中的上确界可以无穷大.

但一维情形下分数次Sobolev空间上奇异型Trudinger-Moser不等式的研究仍是空白. 本文将着力关注、探讨并尝试建立一维情形下有界区间上的奇异型Trudinger-Moser不等式并讨论常数的最佳性.

为了叙述主要结果, 首先介绍分数次Sobolev空间$ H^{\mathrm{s}, p} $. 令$ s \in(0, 1) $, 考虑函数空间$ L_{s}(\mathbb{R}) $:

$ L_{s}(\mathbb{R}) = \left\{u \in L_{l o c}^{1}(\mathbb{R}): \int_{\mathbb{R}} \frac{|u(x)|}{1+|x|^{1+2 s}} d x<\infty\right\}. $

对于$ u \in L_{s}(\mathbb{R}) $, 定义$ (-\Delta)^{s} u $如下:

$ \langle(-\Delta)^{s} u, \varphi\rangle: = \int_{\mathbb{R}} u(-\Delta)^{s} \varphi d x, \varphi \in \mathcal{S}, $

其中$ \mathcal{S} $为光滑的速降函数空间, 并且对于$ \varphi \in \mathcal{S} $, 定义

$ (-\Delta)^{s} \varphi = \mathcal{F}^{-1}\left(|\cdot|^{2 s} \hat{\varphi}\right), $

其中$ \mathcal{F}^{-1} $为Fourier逆变换, $ \hat{\varphi} $$ \varphi $的Fourier变换. 对于$ s \in(0, 1) $$ p \in[1, \infty) $, $ s = \frac{1}{p} $, 定义Bessel位势空间

$ H^{s, p}(\mathbb{R}) = \left\{u \in L^{p}(\mathbb{R}):(-\Delta)^{\frac{s}{2}} u \in L^{p}(\mathbb{R})\right\} $

以及它的子空间

$ \widetilde{H}^{s, p}(I) = \left\{u \in L^{p}(\mathbb{R}): \text{ 在 } \mathbb{R} \backslash I \text{ 上, } u \equiv 0, (-\Delta)^{\frac{s}{2}} u \in L^{p}(\mathbb{R})\right\}, $

其中$ I \subset \mathbb{R} $是一个有界区间. 上述两个空间的范数是:

$ \|u\|_{H^{s, p}(\mathbb{R})}^{p}: = \left\|(-\Delta)^{\frac{s}{2}} u\right\|_{L^{p}(\mathbb{R})}^{p}+\|u\|_{L^{p}(\mathbb{R})}^{p}. $

$ \|u\|^{*}: = \left\|(-\Delta)^{\frac{s}{2}} u\right\|_{L^{p}(I)} $, 当$ u \in \widetilde{H}^{s, p}(I) $时, 由[1]可知存在$ C $使得

$ \|u\|_{H^{s, p}(\mathbb{R})} \leq C\|u\|^{*}, $

也即在$ \widetilde{H}^{s, p}(I) $$ \|u\|_{H^{s, p}(\mathbb{R})} $$ \|u\|^{*} = \left\|(-\Delta)^{\frac{s}{2}} u\right\|_{L^{p}(I)} $是等价的. 所以上述$ \widetilde{H}^{s, p}(I) $空间的范数可以由$ \|u\|^{*} = \left\|(-\Delta)^{\frac{s}{2}} u\right\|_{L^{p}(I)} $替代.

本文的主要结果如下.

定理1.1   对于任意有限区间$ I \subset \mathbb{R} $, $ p \in[1, \infty) $$ \beta \in[0, 1) $, 存在常数$ C(p, \beta) $, 满足对任意的$ 0 \leq \alpha \leq \alpha_{p} $, 有

$ \begin{equation} \sup\limits _{u \in \tilde{H}^{\frac{1}{p}, p} (I), \left\|(-\Delta)^{\frac{1}{2 p}} u\right\|_{L^{p}(I)}{\leq 1} }\frac{1}{|I|^{1-\beta}} \int_{\mathbb{R}} \frac{\exp \left[(1-\beta) \alpha_{p}|u|^{\frac{p}{p-1}}\right]}{|x|^{\beta}} d x \leq C(p, \beta) \end{equation} $ (1.1)

成立, 其中

$ \alpha_{p}: = \frac{1}{2}\left[2 \cos \left(\frac{\pi}{2 p}\right) \Gamma\left(\frac{1}{p}\right)\right]^{p^{\prime}}, p^{\prime} = \frac{p}{p-1}, \Gamma(z): = \int_{0}^{+\infty} t^{z-1} e^{-t} d t. $

进一步, (1)式中的常数$ \alpha_{p} $是最佳的, 即如果$ \alpha>\alpha_{p} $, (1) 式中的上确界可以任意大.

2 定理1.1的证明

为了证明定理1.1, 首先给出下面的重排定义及处理卷积重排问题时常用的O'Neil引理.

$ \lambda(s) = |\{x \in \operatorname{R}: f(x)>s\}| $, 则$ f^{*}(t) = \inf \{s>0: \lambda(s) \leq t\} $$ f $的径向递减重排, $ f^{* *}(t) = t^{-1} \int_{0}^{t} f^{*}(s) d s $为极大值函数. 下面给出著名的O'Neil引理(见[10]), 即若$ \varphi = f * g $, 则对于任意$ t>0 $均有

$ \varphi^{*}(t) \leq \varphi^{* *}(t) \leq t f^{* *}(t) g^{* *}(t)+\int_{t}^{\infty} f^{*}(s) g^{*}(s) d s. $

引理2.1  设$ I $$ \mathbb{R} $的有界子集, $ u \in \tilde{H}^{\frac{1}{p}, p}(I) $. 则

$ \begin{equation} u^{*}(t) \leq\left(\alpha_{p}\right)^{\frac{1-p}{p}}\left[p(t)^{-\frac{1}{p^{\prime}}} \int_{0}^{t}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(r) d r+\int_{t}^{|I|}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(r) \cdot(r)^{-\frac{1}{p^{\prime}}}\right] \end{equation} $ (2.1)

  当$ u \in \tilde{H}^{\frac{1}{p}, p}(I) $, 可知

$ u(x) = \int_{I} G_{\frac{1}{2 p}}(x, y)(-\Delta)^{\frac{1}{2 p}} u(y) d y, $

其中$ G_{\frac{1}{2 p}}(x, y) $$ (-\Delta)^{\frac{1}{2 p}} $$ I $上的Green函数, 即对于$ x \in I $, 满足

$ \left\{\begin{array}{ll} (-\Delta)^{\frac{1}{2 p}} G_{\frac{1}{2 p}}(x, .) = \delta_{x}, & y \in I \\ G_{\frac{1}{2 p}}(x, y) = 0, & y \in R \backslash I \end{array}\right. $

进一步, 不难验证$ G_{\frac{1}{2 p}}(x, y) $满足

$ 0 \leq\left(2 \alpha_{p}\right)^{\frac{p-1}{p}} G_{\frac{1}{2 p}}(x, y) \leq {I_\frac{1}{p}}(x-y) = |x-y|^{\frac{1}{p}-1}, $

其中$ I_{\frac{1}{p}}(x) = |x|^{\frac{1}{p}-1} $. 上述两式证明参照文献[1, Lemma 2.2] 以及文献[1, Theorem 1.1]的证明, 此处从略.

$ x \in I $时, 记$ f = \left|(-\Delta)^{\frac{1}{2 p}} u\right| $, $ g(x) = |x|^{\frac{1}{p}-1} $, 则由简单计算可得

$ g^{*}(t) = \left(\frac{t}{2}\right)^{-\frac{1}{p^{\prime}}}, g^{* *}(t) = p\left(\frac{t}{2}\right)^{-\frac{1}{p^{\prime}}}, $

下面估计$ u(x) $的重排函数. 由于$ f \geq 0 $利用O'Neil引理, 可以得到

$ \begin{align*} u^{*}(t)& \leq u^{* *}(t) \leq\left(2 \alpha_{p}\right)^{\frac{1-p}{p}}\left[t f^{* *}(t) g^{* *}(t)+\int_{t}^{\infty} f^{*}(s) g^{*}(s) d s\right] \\& = \left(2 \alpha_{p}\right)^{\frac{1-p}{p}}\left[t t^{-1} p\left(\frac{t}{2}\right)^{-\frac{1}{p^{\prime}}} \int_{0}^{t}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(r) d r+\int_{t}^{|I|}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(r) \cdot\left(\frac{r}{2}\right)^{-\frac{1}{p^{\prime}}} d r \mid\right. \\& = \left(2 \alpha_{p}\right)^{\frac{1-p}{p}}\left[p\left(\frac{t}{2}\right)^{-\frac{1}{p^{\prime}}} \int_{0}^{t}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(r) d r+\int_{t}^{|I|}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(r) \cdot\left(\frac{r}{2}\right)^{-\frac{1}{p^{\prime}}}\right] \\& = \left(\alpha_{p}\right)^{\frac{1-p}{p}}\left[p(t)^{-\frac{1}{p^{\prime}}} \int_{0}^{t}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(r) d r+\int_{t}^{|I|}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(r) \cdot(r)^{-\frac{1}{p^{\prime}}}\right] \end{align*} $

证毕.

引理2.2[11]  令$ 0<\alpha \leq 1 $, $ p \in (1, \infty) $, $ a(s, t) $是定义在$ (-\infty, \infty) \times[0, \infty) $的非负可测函数且满足

$ \left\{\begin{array}{ll} a(s, t) \leq 1, \text{ 当 } 0<s<t \text{ 时, } \\ \sup _{t>0}\left(\int_{-\infty}^{0} a(s, t)^{p^{\prime}} d s+\int_{t}^{\infty} a(s, t)^{p^{\prime}} d s\right)^{\frac{1}{p^{\prime}}} = b<\infty \end{array}\right. $

$ \Phi \geq 0 $且满足$ \int_{-\infty}^{\infty} \Phi(s)^{p} d s \leq 1, $那么一定存在某个常数$ C_{0} $满足

$ \int_{0}^{+\infty} \exp \left(-F_{\alpha}(t)\right) d t, $

其中

$ F_{\alpha}(t) = \alpha t-\alpha\left(\int_{-\infty}^{+\infty} a(s, t) \Phi(s) d s\right)^{p^{\prime}}. $

定理1.1的证明   对(2.1)式做变量替换$ t = e^{-s}|I| $, 可以很容易得到

$ \begin{equation} \begin{aligned} u^{*}\left(e^{-s}|I|\right) \leq & \left(\alpha_{p}\right)^{\frac{1-p}{p}}|I|^{\frac{1}{p}}\left[p\left(e^{-s}\right)^{\frac{1-p}{p}} \int_{s}^{\infty}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}\left(e^{-r}|I|\right) \cdot e^{-r} d r+\right.\\ &\left.\int_{0}^{s}\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}\left(e^{-r}|I|\right) \cdot\left(e^{-r}\right)^{\frac{1}{p}} d r\right] \end{aligned} \end{equation} $ (2.2)

$ \begin{equation} \Phi(r) = \left\{\begin{array}{ll} \left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}\left(e^{-r}|I|\right) \cdot\left(e^{-r}|I|\right)^{\frac{1}{p}}, & \text{ 若 } r \geq 0 \\ 0, & \text{ 若 } r<0 \end{array}\right. \end{equation} $ (2.3)

以及

$ \begin{equation} a(r, s) = \left\{\begin{array}{ll} 0, & \text{ 若 } r \leq 0 \\ 1, & \text{ 若 } 0<r \leq s \\ p e^{\frac{s-r}{p^{\prime}}}, & \text{ 若 } s<r<\infty \end{array}\right. \end{equation} $ (2.4)

结合(2.2)–(2.4)式, 有

$ u^{*}\left(e^{-s}|I|\right) \leq\left(\alpha_{p}\right)^{\frac{1-p}{p}} \int_{-\infty}^{+\infty} \Phi(r) \cdot a(r, s) d r. $

通过简单的计算, 可知

$ \left(|x|^{-\beta}\right)^{*}(t) = \left(\frac{t}{2}\right)^{-\beta}, $

并且由重排的基本性质, 可知

$ \left(\exp \left[(1-\beta) \alpha|u|^{\frac{p}{p-1}}\right]\right)^{*}(t) = \exp \left[(1-\beta) \alpha\left|u^{*}\right|^{\frac{p}{p-1}}\right]. $

对于任意的$ 0 \leq \alpha \leq \alpha_{p} $, 利用Hardy-Littlewood不等式[12], 有

$ \begin{aligned} & \frac{1}{|I|^{1-\beta}} \int_{I} \frac{\exp \left[(1-\beta) \alpha|u|^{\frac{p}{p-1}}\right]}{|x|^{\beta}} d x \\ \leq &\frac{1}{|I|^{1-\beta}} \int_{0}^{|I|} \exp \left[(1-\beta) \alpha\left|u^{*}\right|^{\frac{p}{p-1}}\right]\left(\frac{t}{2}\right)^{-\beta} d t \\ = &\frac{1}{|I|^{1-\beta}} \int_{0}^{\infty} \exp \left[(1-\beta) \alpha\left|u^{*}\left(e^{-s}|I|\right)\right|^{\frac{p}{p-1}}\right] \cdot\left(\frac{e^{-s}|I|}{2}\right)^{-\beta} \cdot e^{-s}|I| d s \\ = &2^{\beta} \int_{0}^{\infty} \exp \left[(1-\beta) \alpha\left|u^{*}\left(e^{-s}|I|\right)\right|^{\frac{p}{p-1}}\right] \cdot e^{-s(1-\beta)} d s \\ = &2^{\beta} \int_{0}^{\infty} \exp \left[(1-\beta) \alpha\left[\left(\alpha_{p}\right)^{\frac{1-p}{p}} \int_{-\infty}^{+\infty} \Phi(r) \cdot a(r, s) d r\right]^{\frac{p}{p-1}}\right] \cdot e^{-s(1-\beta)} d s\\ \leq& 2^{\beta} \int_{0}^{\infty} \exp \left[(1-\beta)\left[\int_{-\infty}^{+\infty} \Phi(r) \cdot a(r, s) d r\right]^{\frac{p}{p-1}}-s(1-\beta)\right] d s \\ \leq &2^{\beta} \int_{0}^{\infty} \exp \left[-F_{1-\beta}(s)\right] d s, \end{aligned} $

其中

$ F_{1-\beta}(s) = s(1-\beta)-(1-\beta)\left[\int_{-\infty}^{+\infty} \Phi(r) \cdot a(r, s) d r\right]^{\frac{p}{p-1}}. $

因为(2.4)中的$ a(s, r) $满足当$ 0<r<s $, $ a(r, s)\leq 1 $, 以及

$ \int_{-\infty}^{0} a(r, s)^{\frac{p}{p-1}} d r+\int_{s}^{\infty} a(r, s)^{\frac{p}{p-1}} d r = p^{p^{\prime}} \int_{s}^{\infty} e^{s-r} d r = p^{p^{\prime}}<\infty, $

且有

$ \begin{aligned} \int_{-\infty}^{+\infty} \Phi(s)^{p} d s & = \int_{0}^{+\infty}|I|\left[\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}\left(e^{-r}|I|\right)\right]^{p} \cdot e^{-r} d r \\ & = \int_{0}^{|I|}\left[\left|(-\Delta)^{\frac{1}{2 p}} u\right|^{*}(t)\right]^{p} \cdot t \cdot \frac{1}{t} d t \\ & = \left\|(-\Delta)^{\frac{1}{2 p}} u\right\|_{L^{p}(I)}^{p} \leq 1. \end{aligned} $

于是利用引理2.2, 可以得到

$ \int_{0}^{\infty} \exp \left[-F_{1-\beta}(s)\right] d s \leq C_{0}. $

因此, 当$ 0 \leq \alpha \leq \alpha_{p} $, 可以得到

$ \frac{1}{|I|^{1-\beta}} \int_{I} \frac{\exp \left[(1-\beta) \alpha|u|^{\frac{p}{p-1}}\right]}{|x|^{\beta}} \leq C(p, \beta). $

下面证明常数$ \alpha_{p} $是最佳的. 通过简单的伸缩变换, 不难发现只需证明该结论在特定区间$ I = (-1, 1) $上成立即可. 证明过程分为三步.

第一步  测试函数的构造. 固定$ \tau \geq 1 $, 令$ f(y) = f_{\tau}(y): = \frac{1}{2 \tau}|y|^{-\frac{1}{p}} \chi_{\left[-\frac{1}{2}, -r\right] \cup\left[r, \frac{1}{2}\right]}, r: = \frac{e^{-\tau}}{2} $直接计算可得

$ \|f\|_{L_{p}}^{p} = \frac{2}{(2 \tau)^{p}} \int_{r}^{\frac{1}{2}} \frac{d y}{y} = \frac{1}{(2 \tau)^{p-1}}. $

$ u = u_{\tau} \in \widetilde{H}^{\frac{1}{2p}, 2}(I) $为下列方程的解

$ \left\{\begin{array}{ll} (-\Delta)^{\frac{1}{2 p}} u = f, & \text{ 当 } x \in I \\ u \equiv 0, & \text{ 当 } x \in I^{c} \end{array}\right. $

第二步  证明$ u \in \widetilde{H}^{\frac{1}{p}, p}(I) $. 这里的证明与[9]中类似, 这里省略.

第三步  当$ x \in I $时,

$ \begin{aligned} u(x) & = \int_{I} G_{\frac{1}{2 p}}(x, y) f(y) d y \\ & = \frac{1}{2 \tau\left(2 \alpha_{p}\right)^{\frac{p-1}{p}}}\int_{r<|y|<\frac{1}{2}}\frac{d y}{|x-y|^{1-\frac{1}{p}}|y|^{\frac{1}{p}}}-\int_{r<|y|<\frac{1}{2}} H_\frac{{1}}{2 p}(x, y) f(y) d y\\ & = :u_{1}(x)+u_{2}(x) \end{aligned} $

其中$ H_{\frac{1}{2 p}}(x, \cdot) \in \tilde{H}^{\frac{1}{2 p}, {2}}(I)+g_{x} $是下述方程的解

$ \left\{\begin{array}{ll} (-\Delta)^{s} H_{s}(x, \cdot) = 0, & \text{ 当 } x \in I \\ H_{s}(x, \cdot) = g_{x}, & \text{ 当 } x \in I^{c} \end{array}\right. $

其中$ g_{x} \in C^{\infty}(\mathbb{R}) $且当$ y \in I^{c} $时,

$ g_{x}(y) = F_{\frac{1}{2 p}}(x-y) = \frac{1}{2 \cos \left(\frac{1}{2 p} \pi\right) \Gamma\left(\frac{1}{p}\right)|x-y|^{1-\frac{1}{p}}}. $

类似于[9], 可以得到$ u $$ [-r, r] $上的一个下界估计.

$ \begin{aligned} u_{1}(x) = &\frac{1}{2 \tau\left(2 \alpha_{p}\right)^{\frac{p-1}{p}}}\left(\int_{r}^{1 / 2} \frac{d y}{|x-y|^{1-\frac{1}{p}}|y|^{\frac{1}{p}}}+\int_{-1 / 2}^{-r} \frac{d y}{|y-x|^{1-\frac{1}{p}}|y|^{\frac{1}{p}}}\right) \\ \geq &\frac{1}{2 \tau\left(2 \alpha_{p}\right)^{\frac{p-1}{p}}}\left(\int_{r}^{1 / 2} \frac{d y}{y}+\int_{r}^{1 / 2} \frac{d y}{y+x}\right) \\ = &\frac{1}{2 \tau\left(2 \alpha_{p}\right)^{\frac{p-1}{p}}}\left(2 \tau+\log \left(\frac{1+2 x}{1+\frac{x}{r}}\right)\right) \\ = &\frac{1}{\left(2 \alpha_{p}\right)^{\frac{p-1}{p}}}+O\left(\tau^{-1}\right). \end{aligned} $

因为$ H_{\frac{1}{2 p}} $$ [-r, r] \times\left[-\frac{1}{2}, \frac{1}{2}\right] $上有界, 可知

$ \left|u_{2}(x)\right| \leq C \int_{r}^{\frac{1}{2}} f(y) d y \leq C \tau^{-1} \int_{0}^{\frac{1}{2}}|y|^{-\frac{1}{p}} d y = O\left(\tau^{-1}\right), \quad x \in[-r, r]. $

于是, 对于任意的$ x \in[-r, r] $, 当$ \tau \rightarrow \infty $时,

$ u = u_{\tau} \geq \frac{1}{\left(2 \alpha_{p}\right)^{\frac{p-1}{p}}}+O\left(\tau^{-1}\right). $

$ w_{\tau}: = (2 \tau)^{\frac{p-1}{p}} u_{\tau} \in \widetilde{H}^{\frac{1}{p}, p}(I), $不难验证$ \|(-\Delta)^{\frac{1}{2 p}} w_{\tau} \|_{L_{p}(I)} = 1 $, 于是

$ \begin{align*} \int_{I} \frac{\exp \left[(1-\beta) \alpha\left|w_{\tau}\right|^{p^{\prime}}\right]}{|x|^{\beta}} d x \geq& \int_{-r}^{r} \frac{\exp \left[(1-\beta) \frac{\alpha}{\alpha_{p}}(\tau+O(1))\right]}{|x|^{\beta}} dx \\ = & 2 \int_{0}^{r} \frac{\exp \left[(1-\beta) \frac{\alpha}{\alpha_{p}}(\tau+O(1))\right]}{x^{\beta}} d x\\ = &\frac{2}{1-\beta} r^{1-\beta} \exp \left[(1-\beta) \frac{\alpha}{\alpha_{p}}(\tau+O(1))\right] \\ = &\frac{2^{\beta}}{1-\beta} \exp \left[(1-\beta) \tau\left(\frac{\alpha}{\alpha_{p}}-1\right)+(1-\beta) \frac{\alpha}{\alpha_{p}} O(1)\right] \\ \rightarrow & \infty \quad(\tau \rightarrow \infty) \end{align*} $
参考文献
[1] Moser J. A sharp form of an inequality by N. Trudinger[J]. Indiana Univ. math. j., 1970-1971, 20: 1077–1092.
[2] Adimurthi, Sandeep K. A singular Moser-Trudinger embedding and its applications[J]. Nodea Nonlinear Differential Equations and Applications, 2007, 13(5): 585–603. DOI:10.1007/s00030-006-4025-9
[3] Li Yuxiang, Ruf B. A sharp Trudinger-Moser type inequality for unbounded domains in RN[J]. Indiana University Mathematics Journal, 2008, 57(1): 451–480. DOI:10.1512/iumj.2008.57.3137
[4] Lam N, Lu Guozhen. A new approach to sharp Moser-Trudinger and Adams type inequalities: A rearrangement-free argument[J]. Journal of Differential Equations, 2013, 255(3): 298–325. DOI:10.1016/j.jde.2013.04.005
[5] Adachi S, Tanaka K. Trudinger type inequalities in R-N and their best exponent[J]. Proceedings of the American Mathematical Society, 2000, 128(7): 2051–2058.
[6] Calanchi M, Ruf B. Trudinger-Moser type inequalities with logarithmic weights in dimension[J]. Nonlinear Analysis Theory Methods and Applications, 2015.
[7] Cassani D, Tarsi C. A Moser-type inequality in Lorentz-Sobolev spaces for unbounded domains in RN[J]. Asymptotic Analysis, 2009, 64(1-2): 29–51. DOI:10.3233/ASY-2009-0934
[8] Alberico A. Moser type inequalities for higher-order derivatives in lorentz spaces[J]. Potential Analysis, 2008, 28(4): 389–400. DOI:10.1007/s11118-008-9085-5
[9] Feller W. An introduction to probability theory and its applications (3rd ed.) [M]. New York: Wiley, 1969.
[10] O'Neil R. Convolution operators and L(p, q) spaces[J]. Duke Mathematical Journal, 1963, 30(1): 129–142.
[11] Lu Guozhen, Tang Hanli. Sharp singular trudinger-moser inequalities in Lorentz-Sobolev spaces[J]. Advanced Nonlinear Studies, 2016, 16(3): 581–601.
[12] Iwaniec, Nolder C A. Hardy-Littlewood inequality for quasi-regular mappings in certain domains in R[J]. R. N., 1985, 10: 267–282.