\documentclass[reqno]{amsart} \usepackage{hyperref} \usepackage{cite} \AtBeginDocument{{\noindent\small \emph{Electronic Journal of Differential Equations}, Vol. 2015 (2015), No. 83, pp. 1--9.\newline ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu \newline ftp ejde.math.txstate.edu} \thanks{\copyright 2015 Texas State University - San Marcos.} \vspace{9mm}} \begin{document} \title[\hfilneg EJDE-2015/83\hfil Uniqueness of self-similar very singular solution] {Uniqueness of self-similar very singular solution for non-Newtonian polytropic filtration equations with gradient absorption} \author[H. Ye, J. Yin \hfil EJDE-2015/83\hfilneg] {Hailong Ye, Jingxue Yin} \address{Hailong Ye \newline School of Mathematical Sciences, South China Normal University, Guangzhou, Guangdong 510631, China} \email{ye2006hailong@yeah.net} \address{Jingxue Yin \newline School of Mathematical Sciences, South China Normal University, Guangzhou, Guangdong 510631, China} \email{yjx@scnu.edu.cn} \thanks{Submitted November 20, 2014. Published April 7, 2015.} \subjclass[2000]{35K65, 35K92, 35K15} \keywords{Polytropic filtration; gradient absorption; uniqueness; self-similar; \hfill\break\indent very singular} \begin{abstract} Uniqueness of self-similar very singular solutions with compact support are proved for the non-Newtonian polytropic filtration equation with gradient absorption $$ \frac{\partial u}{\partial t} =\operatorname{div}(|\nabla u^m|^{p-2}\nabla u^m) -|\nabla u|^{q},\quad x\in \mathbb{R}^N,\quad t>0, $$ where $m>0$, $p>1$, $m(p-1)>1$ and $q>1$. \end{abstract} \maketitle \numberwithin{equation}{section} \newtheorem{theorem}{Theorem}[section] \newtheorem{lemma}[theorem]{Lemma} \newtheorem{remark}[theorem]{Remark} \allowdisplaybreaks \section{Introduction} This article concerns the non-Newtonian polytropic filtration equation with gradient absorption \begin{equation}\label{problem} \frac{\partial u}{\partial t} =\operatorname{div}(|\nabla u^m|^{p-2}\nabla u^m) -|\nabla u|^{q},\qquad x\in \mathbb{R}^N,\quad t>0, \end{equation} where $m>0, p>1$, $m(p-1)>1$ and $q>1$. Such an equation, especially the case $m=1$ and $p=2$, appears as the viscosity approximation to the well-known Hamilton-Jacobi equation, in the stochastic control theory, as well as in a number of interesting and different physical considerations. For more details, see \cite{PBM,MKG,JKH} and the references therein. In this article, we pay attention to self-similar very singular solutions of \eqref{problem}. Due to the possible degeneracy and singularity, it is necessary to clarify the concept of weak solutions of \eqref{problem}. A non-negative function $u$ is said to be a weak solution of \eqref{problem}, if $u\in C_{\mathrm{loc}}(0,\infty; L^2(\mathbb{R}^N))$, $u^m\in L_{\mathrm{loc}}^p(0,\infty; W_{loc}^{1,p}(\mathbb{R}^N))$, $|\nabla u|\in L_{\mathrm{loc}}^q(\mathbb{R}^N\times(0,\infty))$ and $u$ satisfies \eqref{problem} in the sense of distributions in $\mathbb{R}^N\times(0,\infty)$. Further, by a very singular solution $u$, we mean a weak solution with $u\in C(\mathbb{R}^N\times[0,\infty)\backslash\{(0,0)\})$ satisfying \begin{equation}\label{singular} \lim_{t\to0}\sup_{|x|>\varepsilon}u(x,t)=0 \end{equation} and \begin{equation}\label{verysingular} \lim_{t\to0}\int_{|x|<\varepsilon}u(x,t)\,\mathrm{d}x=\infty \end{equation} for any $\varepsilon>0$. In 1986, Brezis et al \cite{HBL} investigated the semilinear heat equation with concentration absorption \begin{equation*} \frac{\partial u}{\partial t}=\Delta u-u^q; \end{equation*} they proved the existence and uniqueness of self-similar very singular solutions when $11$. More precisely, they proved that there exists a unique self-similar very singular solution of \eqref{p-Laplace} when $\frac{2N}{N+1}2$, Shi \cite{PS0} obtained existence of self-similar very singular solutions with compact support of \eqref{p-Laplace} when $p-10, \quad \beta=\frac{q-m(p-1)}{p-q}>0, $$ and implies that the profile $f$ is a solution of the ordinary differential equation \begin{equation}\label{ODE} \big(|(f^m)'|^{p-2}(f^m)'\big)'+\frac{n-1}{r}|(f^m)'|^{p-2}(f^m)'+\beta rf'+f-|f'|^q=0, \ r>0 \end{equation} with \begin{equation}\label{zeroderivative} (f^m)'(0)=0, \quad f(0)=a, \end{equation} where $a$ is a positive constant to be determined. Note that condition \eqref{singular} is equivalent to, if $u$ is given by \eqref{self-similarsolution}, \begin{equation}\label{condition} \lim_{r\to\infty}r^{1/{\beta}}f(r)=0. \end{equation} In addition, it is easy to see that if $N\beta<1$ (i.e. $q<\frac{p+Nm(p-1)}{N+1}$) and the solution $f$ of \eqref{ODE} satisfies \eqref{zeroderivative}--\eqref{condition}, then $u$ given explicitly by \eqref{self-similarsolution} satisfies \eqref{verysingular} automatically. According to \cite[Lemma 3.1]{PSM}, however, the condition \eqref{condition} does not hold if $N\beta\ge1$, that is, there is no self-similar singular solution. Let $z=f^m, a^m=b$, then the problem \eqref{ODE}--\eqref{zeroderivative} is replaced by the following problem with respect to $z$, \begin{equation}\label{problem3} \begin{gathered} (|z'|^{p-2}z')'+\frac{n-1}{r}|z'|^{p-2}z'+\beta r(z^{1/m})'+z^{1/m}-|(z^{1/m})'|^q=0, \quad r>0,\\ z(0)=b>0,\quad z'(0)=0 \end{gathered} \end{equation} and the condition \eqref{condition} is replaced by \begin{equation}\label{conditionz} \lim_{r\to\infty}r^{1/{\beta}}z^{1/m}(r)=0. \end{equation} By the standard theory of ordinary differential equations, the local existence and uniqueness of solution for \eqref{problem3} is easy to be obtained. Let $z(\cdot;b)$ be the solution of \eqref{problem3} and define $$ R(b):=\sup\{r_0>0:z(r;b)>0,\quad r\in[0,r_0)\}. $$ In the sequel, where there is no confusion, we will omit $b$ and let $z=z(\cdot;b)$. Before going further, we present some basic properties of $z$ which have already been proved in \cite{PSM}. \begin{lemma}\label{lemma} Assume that $\alpha>0$, $\beta>0$ and $b>0$. Let $z$ be a solution to \eqref{problem3} with support $[0,R(b))$. Then \begin{itemize} \item[(i)] $z'(r)<0$ in $(0,R(b))$; \item[(ii)] $\lim_{r\to R(b)^-}z(r)=0$; \item[(iii)] $\lim_{r\to R(b)^-}z'(r)=0$ when $R(b)=\infty$. \end{itemize} \end{lemma} Next, we prove the monotonicity of solutions of \eqref{problem3} with respect to $b$ in the sense that two positive orbits do not intersect each other. \begin{lemma}\label{monotonicity} Assume that $\alpha, \beta>0$, $z_i$ are solutions of \eqref{problem3} on $[0,R_i)$ with initial data $z_i(0)=b_i, i=1,2$ and $\min\{R_1, R_2\}<\infty$, where $[0,R_i)$ denotes the maximal existence interval of $z_i$ and the $R_i>0$ are possibly infinity. If $b_10$ and then $g_k(r)$ solves \begin{equation}\label{1} \begin{aligned} &(|g'_k|^{p-2}g'_k)'+\frac{N-1}{r}|g'_k|^{p-2}g'_k+\beta r(g_k^{1/m})'\\ &+g_k^{1/m}-k^{\frac{q(p+1-m(p-1))-p}{m(p-1)-1}}|(g_k^{1/m})'|^q=0 \end{aligned} \end{equation} Note that $g_k$ is strictly decreasing with respect to $k$, and $\lim_{k\to0}g_k(r)=+\infty$ for any $r\in[0,R]$, then there exists a small $k_0>0$ such that $$ z_2(r)0;z_2(r)0$, that is, $p>q>\frac{p}{p+1-m(p-1)}$, we deduce from \eqref{problem3} that \begin{align*} \lefteqn{(|g'_\tau|^{p-2}g'_\tau)'(r_0)-(|z_2'|^{p-2}z_2')'(r_0)}\\ & = (\tau^{\frac{q(p+1-m(p-1))-p}{m(p-1)-1}}-1)|(z_2^{1/m})'|^q\\ < 0 \end{align*} which contradicts $g''_\tau(r_0)\ge z''_2(r_0)$. Thus, $r_0=0$ and $g_\tau(r)>z_2(r)$ for $r\in(0,R_0]$. Then we have $$ g_\tau(0)=z_2(0), $$ $$ \lim_{r\to0^+}g_\tau'(r)=\lim_{r\to0^+}z_2'(r)=0, $$ and $$ \lim_{r\to0^+}(|g'_\tau|^{p-2}g'_\tau)'(r)=\lim_{r\to0^+}(|z_2'|^{p-2}z_2')'(r)=-\frac{b_2^{1/m}}{N}<0. $$ By continuity there exists $\varepsilon>0$ such that $$ g_\tau(r)>z_2(r)>0\quad\text{and}\quad 0>g'_\tau(r)>z'_2(r) $$ for $r\in(0,\varepsilon)$. Further, we can choose $\varepsilon>0$ small enough such that the following inequalities hold for $r\in(0,\varepsilon)$, \begin{gather*} (|g'_\tau|^{p-2}g'_\tau)'(r)-(|z_2'|^{p-2}z_2')'(r) >0,\\ (g^{1/m}_\tau)'(r)-(z^{1/m}_2)'(r) >0,\\ |(z_2^{1/m})'|^q-\tau^{\frac{q(p+1-m(p-1))-p}{m(p-1)-1}}|(g_{\tau}^{1/m})'|^q >0. \end{gather*} Thus, we obtain that \begin{align*} 0 &=\Big((|g'_\tau|^{p-2}g'_\tau)'-(|z_2'|^{p-2}z_2')'\Big)(r)+\frac{N-1}{r} \Big(|g'_\tau|^{p-2}g'_\tau-|z_2'|^{p-2}z_2'\Big)(r) \\ &\quad +\beta r\Big((g^{1/m}_\tau)'-(z^{1/m}_2)'\Big)(r)+\Big((g^{1/m}_\tau)-(z^{1/m}_2)\Big)(r)& \\ &\quad +\Big(|(z_2^{1/m})'|^q-\tau^{\frac{q(p+1-m(p-1))-p}{m(p-1)-1}}|(g_{\tau}^{1/m})'|^q\Big)(r)>0& \\ \end{align*} for $r\in(0,\varepsilon)$, which is impossible. Summing up, we completed the proof of Lemma \ref{monotonicity}. \end{proof} According to Lemma \ref{monotonicity}, we can define three sets for every $b>0$, \begin{gather*} \mathcal{A}=\{b>0;R(b)<\infty\ {\rm and}\ z'(R(b))<0\},\\ \mathcal{B}=\{b>0;R(b)<\infty\ {\rm and}\ z'(R(b))=0\},\\ \mathcal{C}=\{b>0;R(b)=\infty\ {\rm and}\ z(r)>0, r\ge0\}. \end{gather*} Obviously, these sets are disjoint and $\mathcal{A}\cup\mathcal{B}\cup\mathcal{C}=(0,\infty)$. From \cite[Theorem 1.1]{PSM}, we have the following lemma. \begin{lemma}\label{threesets} Assume that $N\beta<1$, then \begin{itemize} \item[(i)] set $\mathcal{A}$ is nonempty and open; \item[(ii)] set $\mathcal{B}$ is nonempty and closed, and the interface relation \begin{equation}\label{interface} \lim_{r\to R(b)^-}\Big(z^{\frac{m(p-1)-1}{m(p-1)}}\Big)'(r;b) =-\frac{m(p-1)-1}{m(p-1)}(\beta R(b))^{1/(p-1)} \end{equation} holds if $b\in\mathcal{B}$; \item[(iii)] set $\mathcal{C}$ is nonempty and open, and $\lim_{r\to\infty}r^{1/{\beta}}z^{1/m}(r;b)>0$ if $b\in\mathcal{C}$. \end{itemize} \end{lemma} \begin{remark}\label{remark} \rm By Lemma \ref{threesets}, it is easy to see that the solution $z(\cdot;b)$ of the problem \eqref{problem3} satisfies \eqref{conditionz} if and only if $b\in\mathcal{B}$. That is to say, to obtain the uniqueness of self-similar very singular solution of \eqref{problem}, it is suffice to show that the set $\mathcal{B}$ consists only one element. \end{remark} \section{Proof of the Theorem \ref{uniqueness}} We need an auxiliary lemma. Let $z(\cdot;b)$ be a solution of \eqref{problem3} satisfying $b\in\mathcal{B}$, then $R(b)<\infty$ and \eqref{conditionz} holds. Denote $\xi_0=R(b)$ and define $$ U(x,t)=k^{1/m}(\frac{\alpha}{t})^{\alpha}z^{1/m}(\xi), $$ where $\xi=k^{-\gamma}|x|(\frac{\alpha}{t})^{\alpha\beta}$ and $\gamma=\frac{m(p-1)-1}{mp}$, then $$ \operatorname{supp}U=\big\{(x,t)\in\mathbb{R}^N\times(0,\infty); |x|\le\xi_0k^{\gamma}(\frac{\alpha}{t})^{-\alpha\beta}\big\}. $$ \begin{lemma}\label{U} For $t>0$ fixed and $\delta>0$ small enough there exists $\theta=\theta(\delta)\in(0,1)$ such that $U(x,t)0$ small enough we can deduce the existence of $\xi_1\in(\tilde{\xi}, \xi_0)$ such that \eqref{4} hold on $[\xi_1, \xi_0]$. Denote $\theta=\theta(\delta)=\xi_1/\xi_0$ and $\theta_0=\tilde{\xi}/\xi_0$, it is obvious that $$ \lim_{\delta\to0}\xi_1=\tilde{\xi} \quad\text{and}\quad \lim_{\delta\to0}\theta(\delta)=\theta_0\in(0,1). $$ The proof is complete. \end{proof} Now we give the proof of the main result. \begin{proof}[Proof of Theorem \ref{uniqueness}] By Remark \ref{remark}, it is suffice to show that the set $\mathcal{B}$ consists only one element. We give the proof by contradiction. Without loss of generality, assume that $z$ and $Z$ are two solutions of \eqref{problem3} satisfying $z(0),Z(0)\in\mathcal{B}$ and $z(0)1$ in the following proof. By the definition of $\tau$, $z_{\tau}(r)$ must touch $Z(r)$ at $r_0\in[0,R_2]$ from the above, so we divide the next proof into two cases: $r_0\in[0,R_2)$ and $r_0=R_2$. \smallskip \noindent\textbf{Case (i).} If $z_{\tau}(r)$ touch $Z(r)$ at $r_0\in[0,R_2)$, by the similar proof to that of Proposition \ref{monotonicity}, we will derive a contradiction, so $z_{\tau}(r)$ can not touch $Z(r)$ at $r_0\in[0,R_2)$. \smallskip \noindent\textbf{Case (ii).} We firstly define the functions $u, U_{\tau}$ corresponding to $Z$ and $z_\tau$ by \begin{gather*} u(x,t):=(\frac{\alpha}{t})^{\alpha}Z^{1/m}(r), \\ U_\tau(x,t):=(\frac{\alpha}{t})^{\alpha}z_\tau^{1/m}(r)=\tau^{1/m} (\frac{\alpha}{t})^{\alpha}z^{1/m}(\tau^{-\gamma}r). \end{gather*} Then $u$ is a solution of \eqref{problem} and $U_\tau$ is a supersolution. Indeed, a straightforward computation shows that $$ \frac{\partial U_\tau}{\partial t} -\operatorname{div}(|\nabla U_\tau^m|^{p-2}\nabla U_\tau^m) +|\nabla U_\tau|^{q}=(1-\tau^{\frac{q(m(p-1)-p-1)+p}{mp}}|\nabla U_\tau|^{q})\ge0. $$ By Lemma \ref{U}, for sufficiently small $\delta>0$, there exist $\theta_0, \theta(\delta)\in(0,1)$ such that $$ U_\tau(x,1)0$, there exists $\kappa\in(0,1)$ such that $$ Z(|x|)<\kappa z_\tau(|x|),\quad |x|<(1-\varepsilon_1)R_2\tau^{\gamma}; $$ that is, \begin{equation}\label{onlyone2} u(x,1)<\kappa U_\tau(x,1),\quad |x|<(1-\varepsilon_1)R_2\tau^{\gamma}. \end{equation} Now we choose sufficiently small $\varepsilon_1>0$ and $\delta_0>0$ such that $$ \theta(\delta)<1-\varepsilon_1 $$ for $\delta\in(0,\delta_0)$ and $$ \theta_0<1-\varepsilon_1. $$ So we obtain that \begin{equation}\label{onlyone5} \theta(\delta)R_2\tau^{\gamma}<(1-\varepsilon_1)R_2\tau^{\gamma}, \quad \delta\in[0,\delta_0). \end{equation} By continuity of $U_\tau$, there exists $\delta_1\in(0,\delta_0)$ such that $$ \kappa U_\tau(x,1)\le U_\tau(x,1+\delta) $$ for $\delta\in(0,\delta_1)$ and $|x|<(1-\varepsilon_1)R_2\tau^{\gamma}$. Combining with \eqref{onlyone2}, we have \begin{equation}\label{onlyone3} u(x,1)