Critical conditions of nonlinearities in the John problem
Extracted source text · arXiv 2405.12761
%\documentclass[12pt,a4paper,reqno]{amsart}
\documentclass{amsart}
\usepackage{amssymb,bbm,color,xcolor}
\usepackage{amscd}
\usepackage{enumerate}
%\usepackage{showkeys}
\usepackage{siunitx}
\usepackage{bm}
\usepackage{ulem}
\usepackage{float, graphicx}%图片浮动，插图包
\usepackage{subfigure}
\numberwithin{equation}{section}
%\usepackage{mathabx}
\usepackage[pagebackref]{hyperref}
\usepackage{mathtools}% http://www.ctan.org/pkg/mathtools
\usepackage[tableposition=top]{caption}% http://www.ctan.org/pkg/caption
\usepackage{booktabs,dcolumn}% http://www.ctan.org/pkg/dcolumn + http://www.ctan.org/pkg/booktabs
\DeclareFontFamily{OT1}{rsfs}{}
\DeclareFontShape{OT1}{rsfs}{n}{it}{<-> rsfs10}{}
\DeclareMathAlphabet{\mathscr}{OT1}{rsfs}{n}{it}
\theoremstyle{plain}
\newtheorem{thm}{Theorem}[section]
%\newtheorem{theorem}[theorem]{Theorem}
\newtheorem{prop}[thm]{Proposition}
\newtheorem{lem}[thm]{Lemma}
\newtheorem{cor}[thm]{Corollary}
\newtheorem{conjecture}[thm]{Conjecture}
\newtheorem{heuristic}[thm]{Heuristic}
\newtheorem{principle}[thm]{Principle}
\newtheorem{claim}[thm]{Claim}
\theoremstyle{definition}
%\newtheorem{roughdef}[subsection]{Rough Definition}
\newtheorem{definition}[thm]{Definition}
\newtheorem{remark}[thm]{Remark}
\newtheorem{remarks}[thm]{Remarks}
\newtheorem{example}[thm]{Example}
\newtheorem{examples}[thm]{Examples}
%\newtheorem{problem}[subsection]{Problem}
%\newtheorem{question}[subsection]{Question}
%% aligned display, no label
\newcommand{\bal}{\[\begin{aligned}}
\newcommand{\eal}{\end{aligned}\]}
\newcommand{\beeq}{\begin{equation}}\newcommand{\eneq}{\end{equation}}
\newcommand{\beq}{\begin{equation*}}\newcommand{\eeq}{\end{equation*}}
\newcommand{\beqa}{\begin{eqnarray*}}\newcommand{\eeqa}{\end{eqnarray*}}
\newcommand{\beeqa}{\begin{eqnarray}}\newcommand{\eneqa}{\end{eqnarray}}
%% single line, with or without a label
%\newcommand{\beq}{\begin{equation}}
%\newcommand{\eeq}{\end{equation}}
%变体
\newcommand{\SN}{{\mathbb{N}}}
%\newcommand{\SZ}{{\mathbb{Z}}}
\def\<{\langle} \def\>{\rangle}
\newcommand{\al}{\alpha} \newcommand{\be}{\beta}
%\newcommand{\de}{\delta} %\newcommand{\De}{\Delta}
\newcommand{\ep}{\epsilon} \newcommand{\vep}{\varepsilon}
\newcommand{\La}{\Lambda}
\newcommand{\la}{\lambda}
\newcommand{\sig}{\sigma}
\newcommand{\Si}{\Sigma}
\newcommand{\om}{\omega} \newcommand{\Om}{\Omega}
\newcommand{\ga}{\Gamma} % \newcommand{\Ga}{\Gamma}
\newcommand{\gam}{\gamma}
\newcommand{\R}{\mathbb{R}}\newcommand{\Z}{\mathbb{Z}}
\newcommand{\N}{\mathbb{N}}
\newcommand{\CC}{\mathbb{C}}
%\newcommand{\Sp}{\mathbb{S}}
\newcommand{\ms}{\mathbb{S}}\newcommand{\ti}{\tilde }
\newcommand{\pt}{\partial_t}\newcommand{\pa}{\partial}
\newcommand{\D}{{\mathrm D}}\newcommand{\les}{{\lesssim}}
\newcommand{\ges}{{\gtrsim}}
%新定义符号
\newcommand{\hn}{\mathbb{H}^n}
\newcommand{\hth}{\mathbb{H}^3}
\newcommand{\e}{\eta}
\newcommand{\De}{\Delta}
\newcommand{\de}{\delta}
\newcommand{\Th}{\Theta}
\newcommand{\ra}{\Longrightarrow}
\newcommand{\sht}{\frac{1}{\sinh{t}}}
\newcommand{\shr}{\frac{1}{\sinh{r}}}
\newcommand{\shs}{\frac{1}{\sinh{s}}}
\newcommand{\shu}{\frac{1}{\sinh{u}}}
\newcommand{\hf}{\frac{1}{2}}
\newcommand{\f}{\mathscr{F}}
\newcommand{\msu}{\mathscr{U}}
\newcommand{\gm}{\Gamma^{-}(O,t_0)}
\newcommand{\gp}{\Gamma^{+}(O,t_0)}
\newcommand{\one}{\uppercase\expandafter{\romannumeral1}}
\newcommand{\two}{\uppercase\expandafter{\romannumeral2}}
\newcommand{\three}{\uppercase\expandafter{\romannumeral3}}
\newcommand{\four}{\uppercase\expandafter{\romannumeral4}}
\newcommand{\five}{\uppercase\expandafter{\romannumeral5}}
\newcommand{\six}{\uppercase\expandafter{\romannumeral6}}
\newcommand{\sev}{\uppercase\expandafter{\romannumeral7}}
\newcommand{\eig}{\uppercase\expandafter{\romannumeral8}}
\newcommand{\wt}{\widetilde}
\newcommand{\ol}{\overline}
\newcommand{\pri}{\prime}
\newcommand{\nb}{\nabla}
\newcommand{\ps}{\frac1{p_S}}
\newcommand{\pss}{\frac1{p_S^{\pri}}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\title
[Generalized Strauss conjecture for semilinear wave equations on $\R^3$]{
%Concerning the
Generalized Strauss conjecture for semilinear wave equations on $\R^3$
}
%{Nonlinear wave equations with logarithmic nonlinearity on hyperbolic spaces}
%{Blow-up of solutions of nonlinear wave equations on hyperbolic spaces}
\author{Chengbo Wang}
\address{School of Mathematical Sciences\\ Zhejiang University\\Hangzhou 310058, P. R. China}\email{wangcbo@zju.edu.cn }
\author{Xiaoran Zhang$^{*}$}\thanks{* Corresponding author}
\address{School of Mathematical Sciences\\ Zhejiang University\\ Hangzhou 310058,P.R.China}
\email{1025391337@qq.com}
\thanks{The authors were supported by
NSFC 12141102. }
\date{\today}
%\dedicatory{} \commby{}
\begin{document}
%\newcommand{\part}{\value{parta}}
%\bibliographystyle{elsarticle-num-names}%plain}
\bibliographystyle{plain}
\begin{abstract}
In this manuscript, we focus on the more delicate nonlinearity of the semilinear wave equation
$$\partial_{t}^2 u-\De_{\R^3}u=|u|^{p_S}\mu(|u|)\ ,u(0,x)=\vep u_0,\ u_t(0,x)=\vep u_1\ ,$$
where $p_S=1+\sqrt{2}$ is the Strauss critical index in $n=3$, and $\mu$ is a modulus of continuity. Inspired by Chen, Reissig\cite{Chen_2024} and Ebert, Girardi, Reissig\cite{MR4163528}, we investigate the sharp condition of $\mu$ as the threshold between the global existence and blow up with small data. We obtain the almost sharp results in this paper,
which in particular disproves the conjecture in \cite{Chen_2024}. %This is joint work with Professor Chengbo Wang.
%give a counterexample
\end{abstract}
\keywords{Strauss conjecture; sharp critical condition; global existence; blow up.}
%%%%%%%%%%
\subjclass[2010]{35L71, 35L05, 35B33, 35A0, 35B44 }
\maketitle
%\tableofcontents
\section{Introduction}
We consider the semilinear wave equation in Euclidean spaces
\beeq\label{nlw}
\begin{cases}
\partial_{t}^2 u-\De_{\R^n}u=F(u)\ ,\\
u(0,x)=\vep u_0,\ u_t(0,x)=\vep u_1.
\end{cases}
\eneq
If we consider the power type nonlinearity $F(u)=|u|^p$,
Strauss conjecture tells us the critical index is $p=p_S(n)$, which is the positive root of
$$(n-1)p^2-(n+1)p-2=0.$$
It has been studied for a long time and verified by proving small data global existence for $p>p_S(n)$ and blow up with some initial data for $1<p\le p_S(n)$. We refer interested readers to the classical papers, \cite{MR1481816}, \cite{MR1804518} for $p>p_S(n)$, \cite{MR744303} for $1<p<p_S(n)$, \cite{MR824205}, \cite{MR2195336}, \cite{MR2316656} for $p=p_S(n)$, and also \cite{Chen_2024}, \cite{wang2023wave} and references therein.
The history shows that if $F(u)=|u|^{p_S}$, there exists some initial data $(u_0,u_1)$ such that the solution blows up at finite time, and if it is changed a little to $F(u)=|u|^{p_S+\ep}$ in the scale of power function for any $\ep>0$, the solution turns to global existence. This implies Strauss conjecture is sharp in the scale of power type nonlinearity.
Actually, we want to focus on the more delicate nonlinearity, which behaves like
\beeq\label{nonlin}
F(u)=|u|^{p_S}\mu(|u|),
\eneq
where $\mu: [0,\infty)\to [0,\infty)$ is a continuous and increasing function satisfying $\mu(0)=0$, which we call $\mu$ modulus of continuity. Our goal is to investigate the real critical condition for $F(u)$.
\begin{remark}
In \cite{wang2023wave} and \cite{zhang2023critical}, we consider semilinear wave equation on hyperbolic sapce
\beeq\label{nlw-hy}
%\tag{NLW}
\begin{cases}
\partial_{t}^2 u-(\De_{\hn}+\rho^2)u=F(u)\ ,\\
u(0,x)=\vep u_0, u_t(0,x)=\vep u_1\ ,
\end{cases}
\eneq
where $\rho=\frac{n-1}{2}$. It is recalled that if $F(u)=|u|^p$ for $p>1$, small data global existence always holds, due to the fact that the linear solution has exponentially decay estimates. In other words, the critical index is $p=1$ in the scale of polynomial nonlinearity. Then we explore the critical phenomenon for the new nonlinearity
$$F(u)=|u|\mu(|u|)$$
with $\mu(\la)$ behaving like $(\log\frac1\la)^{-\gam}$ $(\gam>0)$ when $\la\ll1$. In the end, we find the sharp critical index $\gam=2$ in the scale of logarithm in $n=2,3$. Furthermore, it is also a rough sense and the real critical condition for $\mu$ is still open.
\end{remark}
\subsection{Chen-Reissig’s Conjecture}
Chen and Reissig in \cite{Chen_2024} first consider the Cauchy problem \eqref{nlw} with nonlinearity \eqref{nonlin}. They have proved if
$$\mu(\la)=c_l(\log\frac1\la)^{-\gam}$$
with $\gam>0$, then the critical exponent is $\gam=\ps$, with the global result for $\gam>\ps$ in $\R^3$ and the blowup result for $\gam\le\ps$ and $c_l\gg1$ in $\R^n$. It implies the almost sharp condition of nonlinearity in the scale of logarithm in $\R^3$.
Based on these results, the authors in \cite{Chen_2024} propose their conjecture for the critical phenomenon:
\begin{conjecture}[Chen-Reissig’s Conjecture]\label{Chen-Reissig-Conjecture}
The general threshold is described by the quantity:
\beeq\label{chen-conj-index}
C_{Str}=\lim_{\la\to0^+}\mu(\la)(\log\frac1\la)^{\ps}.
\eneq
If $C_{Str}>0$, the solution of the Cauchy problem \eqref{nlw} with nonlinearity \eqref{nonlin} will blow up at finite time for some initial data. And on the other hand if $C_{Str}=0$, the small data global existence hold.
\end{conjecture}
%Now we list some typical examples which satisfy Chen-Reissig’s conjecture:
%\begin{example}\label{chen-conj-eg}
%Let $\de_0>0$.
%\begin{itemize}
%\item[(1)] Global part holds when
%$$\mu(\la)=|\la|^{\de_0}\ ,\ \mu(\la)=(\log\frac1\la)^{-\ps+\de_0}\ ,$$
%$$\mu(\la)=(\log\frac1\la)^{-\ps}(\log\log\frac1\la)^{-k}\ with\ k\ge1\ .$$
%Blowup part holds when
%$$\mu(\la)=(\log\frac1\la)^{-\ps}(\log\log\frac1\la)^{\de_0}\ .$$
%\item[(2)] Blowup part holds when
%$$\mu(\la)=(\log\frac1\la)^{-\ps}\ .$$
%\item[(3)] Global part holds when
%$$\mu(\la)=(\log\frac1\la)^{-\ps}(\log\log\frac1\la)^{-\ps}\ .$$
%\end{itemize}
%\end{example}
In fact, they have verified their conjecture partially by proving blowup part when $C_{Str}\gg1$ and global existence part when $C_{Str}=0$ with additional “technical” decay assumption
\beeq\label{chen-add-assume-global}
\mu(\la)(\log\frac1\la)^{\ps}\les(\log\log\frac1\la)^{-1}\ .
\eneq
From the additional decay assumption \eqref{chen-add-assume-global}, we can see the gap between their result and conjecture in small data global existence.
\subsection{Investigate the sharp condition}
To investigate the sharp condition for $F(u)$, we start from the results in \cite{Chen_2024} in $\R^3$. Actually,
it is unnatural to introduce the additional decay assumption \eqref{chen-add-assume-global}, which the authors in \cite{Chen_2024} conjecture is a technical restriction. In fact, it is easy to translate \eqref{chen-add-assume-global} to a weaker condition
\beeq\label{new-add-assume-global}
\mu(\la)(\log\frac1\la)^{\ps}(\log\log\frac1\la)^{\ps}\les(\log\log\log\frac1\la)^{-1}\ ,\eneq
in their proof of global part.
Here we can give a sketch from perspective of iteration. It is known that
%the linear solution $u^0$ behaves
\beq
\begin{split}
u(t,r)&=\vep u^0(t,r)+\frac1{2r}\int_0^t\int_{|t-s-r|}^{t-s+r}(|u|^{p_S}\mu(|u|))(s,\la)\la d\la ds\\
&=\vep u^0(t,r)+\frac1{4r}\int_{|t-r|}^{t+r}\int_{-\be}^{t-r}(|u|^{p_S}\mu(|u|))(\al,\be)\frac{\be-\al}2d\al d\be\ ,
\end{split}
\eeq
where $u^0$ is the homogeneous solution, $\al=s-\la$, $\be=s+\la$ . Let initial data be radial and supported in $B_R(O)$, we recall the classical estimate
$$|u^0(t,r)|\les\<t+r\>^{-1}\chi_{|t-r|\le R}\ ,$$
where $\chi$ is the characteristic funtion. Then by $u^n=\vep u^0+LF(u^{n-1})$, we have
\beq
\begin{split}
|LF(u^0)(t,r)|&\les \frac{\vep^{p_S}}{2r}\int_{|t-r|}^{t+r}\<\be\>^{1-p_S}\mu(\<\be\>^{-1})d\be\\
&\les\frac{\vep^{p_S}}{2r}\int_{|t-r|}^{t+r}\<\be\>^{1-p_S}d\be\mu(\<t-r\>^{-1})\\
&\les\vep^{p_S}\<t+r\>^{-1}\<t-r\>^{-\ps}\mu(\<t-r\>^{-1})\ ,
\end{split}
\eeq
through a simple discussion of the relation of $r,t$. Then it leads to
\beq
|LF(u^1)(t,r)|\les \frac{\vep^{p_S}}{2r}\int_{|t-r|}^{t+r}\<\be\>^{1-p_S}\mu(\<\be\>^{-1})\int_{-\be}^{t-r}\<\al\>^{-1}\mu^{p_S}(\<\al\>^{-1})d\al d\be\ .
%&\les\vep^{p_S}\<t+r\>^{-1}\<t-r\>^{-\ps}\mu(\<t-r\>^{-1})\ ,
\eeq
It suggests that global existence might have a strong connection with the following condition
$$\int_0^{\la_0}\frac{\mu^{p_S}(\la)}{\la}d\la<\infty\ ,$$
which generalizes the additional decay assumption \eqref{chen-add-assume-global} and \eqref{new-add-assume-global}.
On the other hand, for blowup part, by Duhamel’s principle, we recall the expression of solution
$$u(t,r)=\frac{1}{2r}\int_{t-r}^{t+r}u_1(\la)\la d\la+\frac1{4r}\int_{|t-r|}^{t+r}\int_{-\be}^{t-r}(|u|^{p_S}\mu(|u|))(\al,\be)\frac{\be-\al}2d\al d\be\ ,$$
for $u_0=0$ and radial, non-negative, smooth $u_1$ with compact support. From the perspective of iteration, and the initial lower bound
$$u(t,r)\ges \<t+r\>^{-1}$$
near the light cone $0\le t-r\le r_0$, we have the first lower bound estimate
$$u(t,r)\ges \frac1{2r}\int_{t-r}^{t+r}\be^{1-p_S}\mu(\be^{-1})d\be\ges(t+r)^{-1}(t-r)^{-\ps}\mu((t-r)^{-1})\ ,$$
for $t-r\ge R_0\gg1$. Then the second iteration arrives at
$$u(t,r)\ges \frac1{2r}\int_{t-r}^{t+r}\be^{-p_S}(\be-(t-r))\mu(\be^{-2})d\be\int_{R_0}^{t-r}\al^{-1}\mu^{p_S}(\al^{-1})d\al\ .$$
It suggests that blow up also has a strong relation with the critical integral
$$\int_0^{\la_0}\frac{\mu^{p_S}(\la)}{\la}d\la\ .$$
To sum up, it is natural for us to propose the following conjecture:
\begin{conjecture}\label{our-conj}
We consider the Cauchy problem \eqref{nlw} with nonlinearity \eqref{nonlin}. Then we have small data global existence if
$$\int_0^{\la_0}\frac{\mu^{p_S}(\la)}{\la}d\la<\infty\ ,$$
for some $\la_0>0$. And on the other hand,
if
$$\int_0^{\la_0}\frac{\mu^{p_S}(\la)}{\la}d\la=\infty\ ,$$
the solution blows up at finite time in general.
\end{conjecture}
\subsection{Our results}
In this manuscript, we consider the simplest case $n=3$.
Then we have our main results:
%Then we have our conjecture:
%\begin{conjecture}
%If we consider Cauchy problem \eqref{nlw} with \eqref{nonlin}, we
%\end{conjecture}
\begin{thm}\label{global-thm}
Let us consider the Cauchy problem \eqref{nlw} for $n=3$ with nonlinearity \eqref{nonlin}, and assume $F\in C^2$ and
the modulus of continuity $\mu=\mu(\la)$ satisfies
\beeq\label{global-mu}
\int_0^{\la_0}\frac{\mu^{p_S}(\la)}{\la}d\la<\infty\ ,
\eneq
for some constant $\la_0>0$. Then for any $(u_0,u_1)\in C^3_c\times C^2_c$, there exists $\vep_0>0$ so that the problem admits a global classical solution $u\in C^2(\R\times\R^3)$ for any $|\vep|<\vep_0$.
\end{thm}
%%这里直接得C^1的结论好了
Conversely, we consider the classical solution to the Cauchy problem
%\begin{definition}\label{def-energysolu}
%Let $(u_0,u_1)\in H^1\times L^2$, we define that $u=u(t,x)$ is an energy solution to the semilinear Cauchy problem
%\beeq\label{blow-eq}
%\partial_{t}^2 u-\De_{\R^3}u=F(u)\ ,\ u(0,x)= u_0,\ u_t(0,x)= u_1\ .
%\eneq
%on $[0,T)$ if
%$$u\in C([0,T),H^1)\cap C^1([0,T),L^2)\ ,\quad F(u)\in L^1_{loc}([0,T)\times\R^3)\ ,$$
%fulfills the next relation:
%\beeq\label{energy-eq}
%\begin{split}
%&\int_{\R^3}u_t(t,x)\phi(t,x)dx+\int_0^t\int_{\R^3}(\nb u(s,x)\cdot \nb\phi(s,x)-u_s(s,x)\phi_s(s,x))dxds\\
%&=\int_{\R^3}u_1\phi(0.x)dx+\int_0^t\int_{\R^3}F(u)(s,x)\phi(s,x)dxds
%\end{split}
%\eneq
%for any $\phi\in C_0^{\infty}([0,T)\times\R^3)$ and any $t\in[0,T)$.
%\end{definition}
\beeq\label{blow-eq}
\partial_{t}^2 u-\De_{\R^3}u=|u|^{p_S}\mu(|u|)\ ,\ u(0,x)= 0,\ u_t(0,x)= u_1\ ,
\eneq
and obatin blow up part for some initial data.
\begin{thm}\label{blowup-thm}
Let us consider the Cauchy problem \eqref{blow-eq}. Let $k\in\N_+$, $\log^k$ denote the k times iterated logarithm and
the modulus of continuity $\mu=\mu(\la)$ be
%\beeq\label{mu-deri}
%s^l|\mu^{(l)}(s)|\le c_*\mu(s)\quad s\in(0,s_0]\ ,\ l=1,2\ ,
%\eneq
\beeq\label{blowup-mu}
\mu(\la)=\mu_k(\la):= (\log\frac1\la)^{-\frac1{p_S}}(\log\log\frac1\la)^{-\frac1{p_S}}\cdots(\log^k\frac1\la)^{-\frac1{p_S}}\ ,\ \la\in(0,\la_k)\ ,
\eneq
for some $\la_k>0$. Then the classical solution $u$ blows up at finite time for any radial, non-negative, non-trivial data $u_1\in C_c^2$.
% in the form of 对于固定\mu而言，已经计算有$g:\tau\in\R\to g(\tau):=\tau\mu^{\frac1{p_S}}(|\tau|)$ is convex and
% energy solution defined in Definition \ref{def-energysolu} .
\end{thm}
\begin{remark}
For blowup case, we verify our conjecture from the perspective of some typical examples.
We introduce the pointwise lower bound estimate of classical solution and
%give our method to with the help of the pointwise lower bound estimate of classical solution in physical case $n=3$.
finally, we find a counterexample to Chen-Reissig’s Conjecture (Conjecture \ref{Chen-Reissig-Conjecture}), such as
$$\mu(\la)=(\log\frac1\la)^{-\ps}(\log\log\frac1\la)^{-\ps}\ ,$$
which will lead to blowup of the solution at finite time.
\end{remark}
\subsection*{Notation}
\begin{itemize}
\item We still use $\<x\>$ to denote $(R_k^2+|x|^2)^{\hf}$ for fixed $k\in\N_+$ and $\log^{k+1}R_k=3$ in the proof of Theorem \ref{blowup-thm}.
\item We use $A\lesssim B$ to denote
$A\leq CB$ for some fixed large constant C which may vary from line to
line, and similarly we use $A\ll B$
to denote $A\leq C^{-1} B$. We employ $A\sim B$ when $A\lesssim
B\lesssim A$. Particularly, when $C$ depends on parameter $k$, we use $A\les_k B$ and $A\sim_k B$.
\end{itemize}
\section{Preliminary}\label{pre}
Considering the linear wave equations \beeq\label{wave}
%\tag{NLW}
\begin{cases}
\partial_{t}^2 u-\De u=F\ , x\in\R^n\ ,\\
u(0,x)= u_0, u_t(0,x)= u_1\ ,
\end{cases}
\eneq
%it is well known that
Duhamel's principle tells us that
\eqref{wave} is equivalent to the integral equation
\beeq\label{ie}
%\tag{IE}
u(t,x)= u^0(t,x)+(LF)(t,x)= u^0(t,x)+\int_0^t S(\tau)F(t-\tau)(x)d\tau,
\eneq
where $u^0= \pt S(t)u_0(x)+ S(t)u_1(x)$, and $S(t)u_1(x)$ is the solution for the linear homogeneous equation with data $u_0=0$.
Under the assumption of the initial data, we can easily obtain the following classical a priori dispersive estimate for the linear solution $u^0$ in $n=3$ by fundamental solution.% which is available in Chen, Reissig \cite[Proposition 4.1]{Chen_2024}.
%For completeness, we present our proof for $n=3$.
%\begin{lem}[Linear estimates]\label{linear-esti}
%Let $n=3$,
%there exists $N>0$ so that
%we have the estimate $(|x|=r)$
%\beeq\label{eq-linear}
%|\pa^{\gam}u^0(t,x)|\le N\<t+r\>^{-1}\chi_{|t-r|\le R}\ ,\ \gam=\{0,1\}\ ,\ \pa=(\pa_t,\bm{\pa_x})\ ,
%\eneq
%for any solutions to \eqref{wave} with $F=0$, whenever the initial data $(u_0, u_1)\in C^2_c\times C^1_c$ with compact support $B_R(O)$.
%\end{lem}
\begin{lem}[Linear estimates]\label{linear-esti}
Let $n=3$, and the initial data $(u_0, u_1)\in C^3_c\times C^2_c$ is supported in $B_R(O)$. Then
there exists $N>0$ so that $(|x|=r)$
%we have the estimate
\beeq\label{eq-linear}
|u^0(t,x)|\le N\<t+r\>^{-1}\chi_{|t-r|\le R}\ ,
\eneq
for the solution to \eqref{wave} with $F=0$.
\end{lem}
\begin{remark}
One can extend the result to noncompact supported initial data, see Chen, Reissig \cite[Proposition 4.1]{Chen_2024} and Wang, Zhang\cite[Lemma 2.1]{wang2023wave}.
\end{remark}
%To give the proof of Theorem \ref{blowup-thm}, we introduce the following
Next we establish the preliminary of the proof of Theorem \ref{blowup-thm}. Let us set a critical parameter
$$q=1-\frac1{p_S}\ .$$
Under the assumption of Theorem \ref{blowup-thm}, without loss of generality, we set $u_1$ is supported in $B_1(O)$.
%And
%for the fixed constants $c_0>0$, $k\in\N_+$, we have
%$$\mu(\la)\ge c_0(\log\frac1\la)^{-\frac1{p_S}}(\log\log\frac1\la)^{-\frac1{p_S}}\cdots(\log^{k}\frac1\la)^{-\frac1{p_S}}:=\mu_k(\la)\ ,$$
We recall the well-known test function $\Phi=\Phi(x)$ in \cite{MR2195336} defined by
$$\Phi:x\in\R^3\to\Phi(x):=\int_{\ms^2}e^{x\cdot\om}d\sig_{\om}\ .$$
It is simple to obtain that $\Phi$ is
positive, smooth and satisfies $\De\Phi=\Phi$ with
\beeq\label{phi-asy}
\Phi(x)\sim\<x\>^{-1}e^{|x|}\ .
\eneq
%引用文献
We set a modified function
$$M_k(\la):=(\log\frac1\la)(\log\log\frac1\la)\cdots(\log^{k-1}\frac1\la)\ , k\ge2\ ,\ M_1=1\ .$$
%%%突然发现u_0没有了不需要\xi_q部分了，之后看看再改吧
Then we introduce a family of modified auxiliary functions from \cite{MR3912752}:
%\beeq\label{xi-q}
%\xi_q^k(t,x):=\int_0^{\la_1}e^{-\la(R+t)}\cosh(\la t)\Phi(\la x)\la^qM_k^{-q}(\la)d\la\ ,
%\eneq
\beeq\label{eta-q}
\e_q^k(t,s,x):=\int_0^{\la_k}e^{-\la(1+t)}\frac{\sinh(\la(t-s))}{\la(t-s)}\Phi(\la x)\la^qM_k^{-q}(\la)d\la\ ,
\eneq
where %$M_k(\la)$ is the modified function depending on $\mu_k(\la)$,
$\la_k$ is a fixed positive parameter to be determined.
% and the test function
%Then by separation of variables,
%\beq
%\Psi:(t,x)\in[0,+\infty)\times\R^3\to\Psi(t,x):=e^{-t}\Phi(x)
%\eeq
%is a solution to the free wave equation $\pt^2\Psi-\De\Psi=0$.
\begin{lem}\label{pre-e}
Let $\log^{k-1}\frac1{\la_k}=2$, then the following properties hold:
\begin{itemize}
%\item if $|x|\le R$ and $t\ge 0$, then
%$$\xi_q^k(t,x)\ge A_k\ ,$$
%$$\e_q^k(t,0,x)\ges_k\<t\>^{-1}\ .$$
\item if $|x|\le 1+s$ and $t>s\ge0$, then
\beeq\label{lower-e-ts}
\e_q^k(t,s,x)\ges_k\<t\>^{-1}\<s\>^{-q}M_k^{-q}(\<s\>^{-1})\ .\eneq
\item if $|x|\le 1+t$ and $t>0$, then
\beeq\label{lower-e-tt}
\e_q^k(t,t,x)\sim_k\<t\>^{-1}\<t-|x|\>^{-q}M_k^{-q}(\<t-|x|\>^{-1})\ .\eneq
\end{itemize}
%Here $A_k$ are fixed constant depending only on $k$.
\end{lem}
\begin{proof}
When $k=1$, $M_1=1$, the above results can be found in \cite[Lemma 3.1]{MR3912752} except
$$\e_q^1(t,t,x)\ges\<t\>^{-1}\<t-|x|\>^{-q}\ .$$
In this case, if $|x|\le t/2$, it follows from \eqref{lower-e-ts}, we focus on $t/2<|x|\le t+1$.
Without loss of generality, let $t,t-|x|\gg1$, then we have
\beeq\label{k=1-lower}
\begin{split}
\e_q^1(t,t,x)&\ges \int_0^{\la_1}e^{-\la(1+t-|x|)}\<\la|x|\>^{-1}\la^qd\la\\
&\ges t^{-1}\int_{t^{-1}}^{(1+t-|x|)^{-1}}\la^{q-1}d\la\\
&\ges t^{-1}((1+t-|x|)^{-q}-t^{-q})\ges t^{-1}(t-|x|)^{-q}\ .
\end{split}
\eneq
For fixed $k\ge2$, let $\log^{k-1}\frac1{\la_k}=2$, then $M_k^{-q}(\la)$ is positive and
\beq
\begin{split}
\e_q^k(t,s,x)\ges_k\int_0^{\la_k}e^{-\la(1+s-|x|)}\<\la|x|\>^{-1}\frac{1-e^{-2\la(t-s)}}{2(t-s)}\la^{q-1}M_k^{-q}(\la)d\la\ .
%&\ges \int_0^{\la_1}\la^qM_k^{-q}(\la)d\la\ .
\end{split}
\eeq
WLOG, let $t\gg R_k$, then for $|x|\le s+1$, if $s\le R_k$, we have
$$\e_q^k(t,s,x)\ges_k t^{-1}\int_{R_k^{-1}}^{\la_k}\la^{q-1}M_k^{-q}(\la)d\la\ges_k t^{-1}\ ,$$
with $e^{-\la(1+s-|x|)}\<\la|x|\>^{-1}\sim_k1$.
And if $s>R_k$, %with reference to \cite{MR3912752},
we have
\beq
\begin{split}
\e_q^k(t,s,x)&\ges_k\int_{\hf \<s\>^{-1}}^{\<s\>^{-1}}\frac{1-e^{-2\la(t-s)}}{2(t-s)}\la^{q-1}M_k^{-q}(\la)d\la\\
&\ges_k\frac{1-e^{-(t-s)/\<s\>}}{2(t-s)}\int_{\hf \<s\>^{-1}}^{\<s\>^{-1}}\la^{q-1}M_k^{-q}(\la)d\la\ ,
\end{split}
\eeq
with $e^{-(1+s-|x|)/\<s\>}\<|x|/\<s\>\>^{-1}\sim_k1$.
Since $\la^{q-1}M_k^{-q}(\la)$ is decreasing function, we arrive at
$$\e_q^k(t,s,x)\ges_k\frac{1-e^{-(t-s)/\<s\>}}{2(t-s)}\<s\>^{-q}M_k^{-q}(\<s\>^{-1})\ges_k\<t\>^{-1}\<s\>^{-q}M_k^{-q}(\<s\>^{-1})\ .$$
Actually, more precisely, we have
\beeq\label{deri-q-1}
(\la^{q-1}M_k^{-q}(\la))^{\pri}=(q-1)\la^{q-2}M_k^{-q}+q\la^{q-2}M_k^{-q}(\la)(M_2^{-1}+\cdots+M_{k}^{-1})(\la)<0\ ,\eneq
due to $(M_2^{-1}+\cdots+M_{k}^{-1})(\la)\le(k-1)M_2^{-1}(\la)\le\frac{k-1}{2e^{k-2}}\le\hf$ for $\la\le \la_k$.
Finally, we prove \eqref{lower-e-tt}. WLOG, let $t,t-|x|>R_k$. If $|x|\le t/2$, we have the lower bound estimate by \eqref{lower-e-ts}, then we consider $t/2<|x|\le t+1$ like the proof for $k=1$. By \eqref{k=1-lower}, it is clear that
\beq
\begin{split}
\e_q^k(t,t,x)&\ges_k t^{-1}\int_{\<t\>^{-1}}^{\<t-|x|\>^{-1}}\la^{q-1}M_k^{-q}(\la)d\la\\
&\ges_k t^{-1}\<t-|x|\>^{-q}M_k^{-q}(\<t-|x|\>^{-1})\ .
\end{split}
\eeq
Now we pay attention to the upper bound. %WLOG, let $\<t\>,\<t-|x|\>>\frac1{\la_1}$.
For $|x|\le t/2$, we obtain
\beq
\begin{split}
\e_q^k(t,t,x)\les_k&\int_0^{\<t-|x|\>^{-1}}\la^qM_k^{-q}(\la)d\la+\int_{\<t-|x|\>^{-1}}^{\la_k}e^{-\la\<t-|x|\>}\la^qM_k^{-q}(\la)d\la\\
\les_k&\<t-|x|\>^{-1-q}M_k^{-q}(\<t-|x|\>^{-1})\\
&+\<t-|x|\>^{-q}M_k^{-q}(\<t-|x|\>^{-1})\int_{\<t-|x|\>^{-1}}^{\la_k}e^{-\la\<t-|x|\>}\la\<t-|x|\>d\la\\
\les_k&\<t\>^{-1}\<t-|x|\>^{-q}M_k^{-q}(\<t-|x|\>^{-1})\ ,
\end{split}
\eeq
with the fact that $\la^qM_k^{-q}(\la)$ is a increasing function and $\la^{q-1}M_k^{-q}(\la)$ is a decreasing function by \eqref{deri-q-1}.
For $|x|\ge t/2$, we finish the proof with
\beq
\begin{split}
\e_q^k(t,t,x)\les_k&t^{-1}(\int_0^{\<t-|x|\>^{-1}}\la^{q-1}M_k^{-q}(\la)d\la+\int_{\<t-|x|\>^{-1}}^{\la_k}e^{-\la\<t-|x|\>}\la^{q-1}M_k^{-q}(\la)d\la)\\
\les_k&t^{-1}\int_0^{\<t-|x|\>^{-1}}(\la^{q}M_k^{-q}(\la))^{\pri}d\la\\
&+t^{-1}\<t-|x|\>^{-q+1}M_k^{-q}(\<t-|x|\>^{-1})\int_{\<t-|x|\>^{-1}}^{\la_k}e^{-\la\<t-|x|\>}d\la\\
\les_k&\<t\>^{-1}\<t-|x|\>^{-q}M_k^{-q}(\<t-|x|\>^{-1})\ ,
\end{split}
\eeq
where
$$\left(\frac{\la^qM_k^{-q}(\la)}q\right)^{\pri}=\la^{q-1}M_k^{-q}(\la)(1+M_2^{-1}+\cdots+M_{k}^{-1})\ge\la^{q-1}M_k^{-q}\ .$$
\end{proof}
\section{Proof of Theorem \ref{global-thm}}
In this section, we give the proof of Theorem \ref{global-thm}. By \cite[Chapter \one, Theorem 5.1]{MR1715192}, we have the classical local solution $u\in C^2([0,T]\times\R^3)$ due to $F\in C^2$ and $F(0)=0$.
%In fact, with the fact of \eqref{mu-deri}, we obtain
%\beeq\label{F-2}
%|F^{(l)}(u)|\les |u|^{p_S-l}\mu(|u|)\ ,\ l=1,2\ ,\ \forall 0<|u|\le s_0\ .
%\eneq
To obtain the global existence, we only need to prove there exists a constant $C$ so that
$$\sup_{x}|u(t,x)|\le C\ ,$$
for all $t>0$.
With the help of the monotonicity, continuous of $\mu$ and $\mu(0)=0$, there exists $0<s_1\le1$ so that $\mu(s)\le1$ for $0<s\le s_1$. Let $\<y\>:=(s_1^{-2}+|y|^2)^{\hf}$, then $\mu(\<y\>^{-1})\le1$. WLOG, let $s_1=1$.
By Lemma \ref{linear-esti}, there exists a constant $C_0>0$ such that
$$|u^0(t,x)|\le \vep C_0\<t+r\>^{-1}\<t-r\>^{-\ps}\mu(\<t-r\>^{-1})=: \vep C_0\Phi_0(t,r)\ .$$
%直接从迭代的角度得到有界
%In light of Banach’s contraction principle, we define the complete solution space and the solution map $Tu$ as follows
%$$X_{\vep}=\{u\in C([0,\infty)\times\R^3):\|u\|\le 2\vep C_0\}\ ,$$
%$$\|u\|:=\sup_{t,x}\frac{\<t+|x|\>\<t-|x|\>^{\ps}}{\mu(\<t-|x|\>^{-1})}|\pa^{\gam} u(t,x)|\ , \ \gam=\{0,1\}\ ,\ \pa=(\pa_t,\bm{\pa_x})\ ,$$
%From the perspective of iteration by \eqref{ie}, we have
%$$u^n(t,x)=u^0(t,x)+LF(u^{n-1})(t,x)\ , n\ge1\ .$$
%Then our aim is to
%$$\|u\|:=\sup_{t,x}\frac{\<t+|x|\>\<t-|x|\>^{\ps}}{\mu(\<t-|x|\>^{-1})}| u(t,x)|\ ,$$
%$$(Tu)(t,x)=\vep u^0+LF(u)\ .$$
%which means the estimate of contraction part is the same as
% the boundness of $u$.
Then bootstrap argument tells us that the proof is reduced to the following boundness estimate.
\begin{lem}\label{bound-lem}
Under the assumption of Theorem \ref{global-thm}, there exists $\vep_0>0$ so that for any $\vep\in(0,\vep_0]$, we have
\beeq\label{bound}
|LF(u)(t,x)|\le \vep C_0\Phi_0(t,r)\ ,
\eneq
whenever $|u(t,x)|\le 2\vep C_0\Phi_0(t,r)$.
\end{lem}
\begin{proof}
%Since $u\in X_{\vep}$, we have
%%$$|\pa^{\gam}u(t,x)|\le 2\vep C_0\<t+r\>^{-1}\<t-r\>^{-\ps}\mu(\<t-r\>^{-1})=:\Phi_0(t,r)\ .$$
%$$|u(t,x)|\le 2\vep C_0\<t+r\>^{-1}\<t-r\>^{-\ps}\mu(\<t-r\>^{-1})=:\Phi_0(t,r)\ .$$
By Duhamel’s principle and the well-known solution formula of linear wave equation for radial data
$$u^0(t,x)=S(t)u_1+\pt S(t)u_0\ ,$$
\beeq\label{radial-fom}
S(t)u_1(x)=tM^tu_1(x)=\frac{t}{|\ms^2|}\int_{\ms^2}u_1(x+ty)d\sig_y=\frac1{2r}\int_{|t-r|}^{t+r}u_1(\la) \la d\la\ ,
\eneq
we obtain for $2\vep C_0\le1$
\beq
\begin{split}
|LF(u)(t,x)|&\le\frac1{2r}\int_0^t\int_{|t-s-r|}^{t-s+r}F(2\vep C_0\Phi_0)(s,\la)\la d\la ds\\
&\le\frac{(2\vep C_0)^{p_S}}{2r}\int_0^t\int_{|t-s-r|}^{t-s+r}\<s+\la\>^{-p_S}\<s-\la\>^{-1}\mu^{p_S}(\<s-\la\>^{-1})\mu(\Phi_0)\la d\la ds.
\end{split}
\eeq
Introducing new variables of integration
\beeq\label{new-vari}
\al=s-\la, \be=s+\la,
\eneq
we have
\beq
%\begin{split}
|LF(u)(t,x)|\le\frac{(2\vep C_0)^{p_S}}{2r}\int_{|t-r|}^{t+r}\int_{-\be}^{t-r}\<\be\>^{-p_S}\<\al\>^{-1}\mu^{p_S}(\<\al\>^{-1})\mu(\Phi_0(\al,\be))(\be-\al)d\al d\be.
%\end{split}
\eeq
Since $\Phi_0(\al,\be)\le \<t-r\>^{-1}$,
there exists a constant $C_1>0$ such that
\beq
|LF(u)(t,x)|\le\frac{C_1\vep^{p_S}}{r}\mu(\<t-r\>^{-1})\int_{|t-r|}^{t+r}\<\be\>^{1-p_S}\int_{-\be}^{t-r}\<\al\>^{-1}\mu^{p_S}(\<\al\>^{-1})d\al d\be\ ,
\eeq
where $C_1$ may change line by line in the following computation.
Without loss of generality, we consider $t+r\gg1$.
Under the assumption \eqref{global-mu} of $\mu$, it is clear that
\beq
\begin{split}
&\int_{-\be}^{t-r}\<\al\>^{-1}\mu^{p_S}(\<\al\>^{-1})d\al\\
\le&\int_{|\al|\le1}+\int_{[-\be,t-r]\cap[-1,1]^c}\<\al\>^{-1}\mu^{p_S}(\<\al\>^{-1})d\al\\
%\les&1+\int_{[-\be,t-r]\cap[-1,1]^c}\al^{-1}\mu^{p_S}(\al^{-1})d\al\\
\les&1+\int_0^{\la_0}\frac{\mu^{p_S}(\la)}{\la}d\la<\infty\ .
\end{split}
\eeq
Then we deduce that
\beq
|LF(u)(t,x)|\le\frac{C_1\vep^{p_S}}{r}\mu(\<t-r\>^{-1})\int_{|t-r|}^{t+r}\<\be\>^{1-p_S}d\be\ .
\eeq
When $r\le t/2$, we have
$$\frac1r\int_{t-r}^{t+r}\<\be\>^{1-p_S}d\be\les\<t-r\>^{1-p_S}\les\<t+r\>^{-1}\<t-r\>^{2-p_S}\ ,$$
for $r\ge t/2$, we have
$$\frac1r\int_{|t-r|}^{t+r}\<\be\>^{1-p_S}d\be\les \<t+r\>^{-1}\<t-r\>^{2-p_S}\ ,$$
%Since
%\beq
%\begin{split}
%|\pa LF(u)(t,x)|&=|\int_0^t s\pa M^sF(u(t-s))(x)ds|\\
%&\le\frac1{|\ms^2|}\int_0^ts\int_{\ms^2}|F^{\pri}(u)||\pa u|(t-s,x+sy)d\sig_y\\
%&\les\int_0^tsM^sF(\Phi_0(t-s))(x)ds=LF(\Phi_0)(t,x)\ ,
%\end{split}
%\eeq
%we can do the same estimate to the derivative term,
then
with the fact of $2-p_S=-\ps$, we finish the proof for sufficiently small $\vep_0$.
\end{proof}
\section{Proof of Theorem \ref{blowup-thm}}
Under the assumption of initial data and $F\in C^2$, we have the unique classical local solution $u\in C^2([0,T_0]\times\R^3)$. %which is also an energy solution defined by Definition \ref{def-energysolu} by finite propagation speed.
%Then by contradiction, we suppose that the solution is global.
Then we will follow the method in \cite{MR3912752} by
using an iteration argument to derive lower bound estimates for an appropriate functional.
% with the
%following iteration frame, which is modified from the one introduced in \cite{Chen_2024}, with a better initial lower bound.% in Proposition \ref{pointwise-low}.
\subsection{Iteration frame}
From finite propagation speed, %and Definition \ref{def-energysolu}
we have the following integral identity for the classical solution $u\in C^2([0,T)\times\R^3)$ in Theorem \ref{blowup-thm}
\beeq\label{def-u}
\begin{split}
\int_{\R^3}u(t,x)\e_q^k(t,t,x)dx=&t\int_{\R^3}u_1(x)\e_q^k(t,0,x)dx\\
&+\int_0^t(t-s)\int_{\R^3}|u(s,x)|^{p_S}\mu(|u(s,x)|)\e_q^k(t,s,x)dxds\ ,
\end{split}
\eneq
for any $t\in(0,T)$, where %$\xi_q^k$ and
$\e_q^k$ is defined in \eqref{eta-q}. One can see \cite[Proposition 4.1]{MR3912752} for detail.
Let us introduce a non-negative time-dependent functional
\beeq\label{functional}
H:t\in[0,T)\to H(t):=\int_{\R^3}u(t,x)\e_q^k(t,t,x)dx\ ,
\eneq
where $u$ is non-negative obviously by fundamental solution and Duhamel’s principle.
Using H${\rm \ddot{o}}$lder’s inequality, we obtain
\beq
\begin{split}
\msu(s):&=\int_{\R^3}u(s,x)\mu^{\ps}(|u(s,x)|)\e_q^k(s,s,x)dx\\
&\le(\int_{\R^3}|u|^{p_S}\mu(|u|)\e_q^k(t,s,x)dx)^{\ps}
\cdot \left(\int_{B_{1+s}}\left(\frac{\e_q^k(s,s,x)}{(\e_q^k(t,s,x))^{\ps}}\right)^{p_S^{\pri}}dx\right)^{\pss}\ .
\end{split}
\eeq
According to \eqref{def-u}, we have
\beeq\label{H-compute}
H(t)\ge\int_0^t(t-s)\msu^{p_S}(s)\left(\int_{B_{1+s}}\left(\frac{\e_q^k(s,s,x)}{(\e_q^k(t,s,x))^{\ps}}\right)^{p_S^{\pri}}dx\right)^{-\frac{p_S}{p_S^{\pri}}}ds\ .\eneq
On the other hand, the convex function $g$ gives the following relation between $H$ and $\msu$ by Lemma \ref{jensen-ineq},
$$g\left(\frac{\int u(s,x)\e_q^k(s,s,x)dx}{\int_{B_{1+s}}\e_q^k(s,s,x)dx}\right)\le\frac{\int u(s,x)\mu^{\ps}(|u(s,x)|)\e_q^k(s,s,x)dx}{\int_{B_{1+s}}\e_q^k(s,s,x)dx}\ ,$$
that is
\beeq\label{msu-compute}
\msu(s)\ge H(s)\mu^{\ps}\left(\left|\frac{H(s)}{\int_{B_{1+s}}\e_q^k(s,s,x)dx}\right|\right)\ .\eneq
%$g(\tau)=\tau\mu^{\ps}(|\tau|)$
%To sum up, we obtain
%\beq
%\begin{split}
%&H(t)\\
%\ge&\int_0^t(t-s)H^{p_S}(s)\mu\left(\left|\frac{H(s)}{\int_{B_{R+s}}\e_q^k(s,s,x)dx}\right|\right)\left(\int_{B_{R+s}}\left(\frac{\e_q^k(s,s,x)}{(\e_q^k(t,s,x))^{\ps}}\right)^{p_S^{\pri}}dx\right)^{-\frac{p_S}{p_S^{\pri}}}ds\ .
%\end{split}
%\eeq
WLOG, let $s>2$, then from Lemma \ref{pre-e}, it is clear that
\beq
\begin{split}
&\int_{B_{1+s}}\e_q^k(s,s,x)dx\\
\les_k&\<s\>^{-1}(\int_0^{s/2}+\int_{s/2}^{s-1}+\int_{s-1}^{s+1})r^2\<s-r\>^{-q}M_k^{-q}(\<s-r\>^{-1})dr\\
\les_k&\<s\>^{1+\ps}M_k^{-q}(\<s/2\>^{-1})+\<s\>\int_{s/2}^{s-1}\<s-r\>^{-q}M_k^{-q}(\<s-r\>^{-1})dr+\<s\>\ .
%\les_k&\<s\>^{1+\ps}M_k^{-q}(\<s\>^{-1})\les_k \<s\>^2\ ,
\end{split}
\eeq
By \eqref{deri-q-1}, we obtain
\beq
\begin{split}
\int_{s/2}^{s-1}\<s-r\>^{-q}M_k^{-q}(\<s-r\>^{-1})dr&\sim_k\int_{\<s/2\>^{-1}}^{\<1\>^{-1}}\la^{q-2}M_k^{-q}(\la)d\la\\
&\sim_k\int_{\<s/2\>^{-1}}^{\<1\>^{-1}}\left(\frac{\la^{q-1}M_k^{-q}(\la)}{q-1}\right)^{\pri}d\la\\
&\les_k\<s\>^{1-q}M_k^{-q}(\<s/2\>^{-1})\ .
\end{split}
\eeq
Then it leads to
$$\int_{B_{1+s}}\e_q^k(s,s,x)dx\les_k\<s\>^{1+\ps}M_k^{-q}(\<s\>^{-1})\les_k \<s\>^2\ ,$$
with the fact that $\<s/2\>^{-1}\le2\<s\>^{-1}$, $\<s\>\ge R_k\gg1$, which means $M_k^{-q}(\<s/2\>^{-1})\les_kM_k^{-q}(\<s\>^{-1})$.
To reduce \eqref{H-compute}, we calculus the following integral with the help of $q=1-\ps=\frac1{p_S^{\pri}}$, $-p_S^{\pri}+q\frac{p_S^{\pri}}{p_S}+2=\frac{p_S^{\pri}}{p_S}$
\beq
\begin{split}
&\int_{B_{1+s}}\left(\frac{\e_q^k(s,s,x)}{(\e_q^k(t,s,x))^{\ps}}\right)^{p_S^{\pri}}dx\les_k\int_0^{1+s}r^2\left(\frac{\<s\>^{-1}\<s-r\>^{-q}M_k^{-q}(\<s-r\>^{-1})}{[\<t\>^{-1}\<s\>^{-q}M_k^{-q}(\<s\>^{-1})]^{\frac1{p_S}}}\right)^{p_S^{\pri}}dr\\
\les_k&\<t\>^{\frac{p_S^{\pri}}{p_S}}\<s\>^{-p_S^{\pri}+q\frac{p_S^{\pri}}{p_S}}(M_k(\<s\>^{-1}))^{q\frac{p_S^{\pri}}{p_S}}\times\\
&\left(\<s\>^{2}M_k^{-1}(\<s/2\>^{-1})+\<s\>^{2}\int_{s/2}^{s-1}\<s-r\>^{-1}M_k^{-1}(\<s-r\>^{-1})dr\right)\\
\les_k&\<t\>^{\frac{p_S^{\pri}}{p_S}}\<s\>^{\frac{p_S^{\pri}}{p_S}}M_k^{\ps}(\<s\>^{-1})\log^k\<s\>\ ,
\end{split}
\eeq
where
\beq
\begin{split}
\int_{s/2}^{s-1}\<s-r\>^{-1}M_k^{-1}(\<s-r\>^{-1})dr\sim_k\int_{\<s/2\>^{-1}}^{\<1\>^{-1}}\la^{-1}M_k^{-1}(\la)d\la\les_k\log^k\<s\>\ .
\end{split}
\eeq
In conclusion, by \eqref{H-compute} and \eqref{msu-compute}, we deduce our iteration frame
\beeq\label{iter-frame}
H(t)\ge C_k\<t\>^{-1}\int_0^t(t-s)\<s\>^{-1}M_k^{-q}(\<s\>^{-1})(\log^k\<s\>)^{-\frac{p_S}{p_S^{\pri}}}H^{p_S}(s)\mu(H(s)\<s\>^{-2})ds\ ,
\eneq
for any $t\in[0,T)$ with positive constant $C_k$ depending only on $k$.
\subsection{Initial lower bound}
%In \cite{Chen_2024}, they
%%对比陈的初始下界结论以及过程
From the perspective of iteration, we deduce the following pointwise lower bound estimate.
\begin{prop}\label{pointwise-low}
Under the assumption of Theorem \ref{blowup-thm}, the classical solution of Cauchy problem \eqref{blow-eq} has the pointwise lower bound
$$u(t,r)\ge C(t+r)^{-1}(t-r)^{-\ps}\mu((t-r)^{-2})\int_{C(t-r)^{-1}}^{\frac{C}2}\frac{\mu^{p_S}(\la)}{\la}d\la$$
for $t-r>R_0^k$ and $r\ge \frac{2^6-1}{2^6+1}t=:k_0t$, $R_0^k\gg1$ and $0<C<1$.
\end{prop}
We leave the proof to the end of this section. And now we want to
give a better first lower bound estimate of the functional $H(t)$ due to the pointwise estimate. There exists $t_0\gg \max\{R_k,R_0^k\}$, which we still denote by $R_k$,
such that for $t>t_0$, we have
\beq
\begin{split}
H(t)&=\int_{B_{1+t}}u(t,x)\e_q^k(t,t,,x)dx\\
&\ges_k\int_{k_0t}^{t-R_k} (t-r)^{-1}\mu((t-r)^{-2})M_k^{-q}((t-r)^{-1})\int_{C(t-r)^{-1}}^{\frac{C}2}\frac{\mu^{p_S}(\la)}{\la}d\la dr\ ,
\end{split}
\eeq
WLOG, we set $\log^{k+1}R_k>2$, and $\log^{k+1}((1-k_0)t_0)\ge 2\log^{k+1}R_k$.
Since $\mu=\mu_k$, let
\beeq\label{ti-def}
\ti{\mu}(\la)=(\log\frac1\la)^{\ps}\cdots(\log^{k-1}\frac1\la)^{\ps}\mu(\la)\ ,\eneq
%=(\log^k\frac1\la)^{-\ps}
then we arrive at
\beeq\label{muM_k}
\mu((t-r)^{-2})M_k^{-q}((t-r)^{-1})\ges_k(\log(t-r))^{-1}\cdots(\log^{k-1}(t-r))^{-1}\ti{\mu}((t-r)^{-2})\eneq
with the fact that
$$\left(\frac{\log(t-r)}{\log((t-r)^2)}\right)^{\ps}\cdots\left(\frac{\log^{k-1}(t-r)}{\log^{k-1}((t-r)^2)}\right)^{\ps}\ges_k1\ ,\ for\ t-r\ge R_k\ .$$
Due to $\mu=\mu_k$, we have
$$\int_{C(t-r)^{-1}}^{\frac{C}2}\frac{\mu^{p_S}(\la)}{\la}d\la\ges_k\log^{k+1}(t-r)\ .$$
Finally, we obtain our lower bound estimate
\beq
\begin{split}
H(t)&\ges_k\int_{k_0t}^{t-R_k}(t-r)^{-1}(\log(t-r))^{-1}\cdots(\log^{k-1}(t-r))^{-1}\log^{k+1}(t-r)dr\ti{\mu}(t^{-2})\\
&\ges_k\int_{R_k}^{(1-k_0)t}\left((\log^k\la)(\log^{k+1}\la)\right)^{\pri}d\la\ti{\mu}(t^{-2})\\
&\ges_k(\log^kt)(\log^{k+1}t)\ti{\mu}(t^{-2})\ ,
\end{split}
\eeq
where
\beq\begin{split}\left((\log^k\la)(\log^{k+1}\la)\right)^{\pri}&=\la^{-1}(\log\la)^{-1}\cdots(\log^{k-1}\la)^{-1}\left(\log^{k+1}\la+(\log^{k}\la)^{-1}\right)\\
&\le 2\la^{-1}(\log\la)^{-1}\cdots(\log^{k-1}\la)^{-1}\log^{k+1}\la\ ,\ for\ \la\ge R_k\ .
\end{split}
\eeq
That is
\beeq\label{first-lower}
H(t)\ge B_k(\log^kt)(\log^{k+1}t)\ti{\mu}(t^{-2})\ ,
\eneq
for $t>t_0$ and some constant $B_k>0$ depending only on $k$.
\begin{remark}
Compared with the first lower bound in \cite{Chen_2024},
for $k=1$, our new bound have more increase with “$\log\log t$”, which remove the additional assumption
$$(\log\frac1\la)^{\ps}\mu(\la)\ge C_{Str}\gg1$$
in \cite{Chen_2024}.
\end{remark}
To end this section, we give our proof of Proposition \ref{pointwise-low}.\\
\textbf{Proof of Proposition \ref{pointwise-low}:}
Suppose the initial data has compact support in $B_1(O)$.
By \eqref{ie} and \eqref{radial-fom},
we first have the lower bound for $t>r$
$$u(t,r)\ge u^0(t,r)= \frac{1}{2r}\int_{t-r}^{t+r}u_1(\la)\la d\la\ .$$
Since $u_1$ is non-negative, there exists $r_0\in(0,1)$ and $0<C<1$ ($C$ may change line by line) such that
for $t+r>1$ we have
$$\int_{r_0}^{t+r}u_1(\la)\la d\la \ge C\ .$$
Then for $0<t-r<r_0$ and $t+r>1$, we obtain
\beeq\label{initial-bd}
u(t,r)\ge C(t+r)^{-1}\ .
\eneq
We next iterate by the following form
\beeq\label{iter-lb}
u(t,r)\ge\int_0^tS(t-s)F(u)(s)ds=\frac1{2r}\int_0^t\int_{|t-s-r|}^{t-s+r}(|u|^{p_S}\mu(|u|))(s,\la)\la d\la ds\ .
\eneq
With the initial lower bound \eqref{initial-bd}, the monotonicity of $F(u)$ and the new coordinate $(\al,\be)$ in \eqref{new-vari}, we have
\beq
\begin{split}
u(t,r)&\ge \frac{C}{2r}\int_{t-r}^{t+r}\int_0^{r_0} \be^{-p_S}\mu(C\be^{-1})(\be-\al)d\al d\be\\
&\ge \frac{C}{2r}\int_{t-r}^{t+r}\be^{1-p_S}\mu(C\be^{-1})d\be\ ,
\end{split}
\eeq
for $t-r>2>(1\vee 2r_0)$. Now we deal with the last integral.
Since $\mu=\mu_k$ in \eqref{blowup-mu}, we have
\beeq\label{mu-deri}
\la|\mu^{\pri}(\la)|\le c_*\mu(\la)\ ,\quad \la\in(0,\la_k]\ ,\ c_*\ll1\ ,
\eneq
for some $\la_k$, which means
\beeq\label{deri-main}
\begin{split}
\left(\frac{\be^{2-p_S}\mu(C\be^{-1})}{2-p_S}\right)^{\pri}&=\be^{1-p_S}\mu(C\be^{-1})+\frac{1}{p_S-2}\be^{1-p_S}((C\be^{-1})\mu^{\pri}(C\be^{-1}))\\
&\le(1+\frac{c_*}{p_S-2})\be^{1-p_S}\mu(C\be^{-1})\ .
\end{split}
\eneq
Then for $t/2\le r<t$, the last integral implies
\beq
\begin{split}
u(t,r)&\ge \frac{C}{2r}\int_{t-r}^{t+r}\left(\frac{\be^{2-p_S}\mu(C\be^{-1})}{2-p_S}\right)^{\pri}d\be\\
&\ge Ct^{-1}\left[(t-r)^{2-p_S}\mu(C(t-r)^{-1})-(t+r)^{2-p_S}\mu(C(t+r)^{-1})\right]\\
&\ge C(t+r)^{-1}(t-r)^{-\ps}\mu(C(t-r)^{-1})\
\end{split}
\eeq
On the other hand, for $r<t/2$, it is direct that
$$u(t,r)\ge C(t+r)^{1-p_S}\mu(C(t+r)^{-1})\ge C(t+r)^{-1}(t-r)^{-\ps}\mu(C(t-r)^{-1})\ .$$
%After one iteration by \eqref{iter-lb}, we get
%$$u(t,r)\ge C(t+r)^{-1}(t-r)^{-\ps}\mu(C(t-r)^{-1})\ ,$$
%for $t-r>(R\vee 2r_0)$,
Then we make another iteration and get
\beq
\begin{split}
u(t,r)&\ge\frac{C}{2r}\int_{t-r}^{t+r}\int_{2}^{t-r}\be^{-p_S}\al^{-1}\mu^{p_S}(C\al^{-1})\mu(C\be^{-2})(\be-\al)d\al d\be\\
&\ge\frac{C}{2r}\int_{t-r}^{t+r}\be^{-p_S}\mu(C\be^{-2})(\be-(t-r))d\be\int_{2}^{t-r}\al^{-1}\mu^{p_S}(C\al^{-1})d\al\\
&\ge\frac{C}{2r}\int_{t-r}^{t+r}(\be^{1-p_S}-(t-r)\be^{-p_S})\mu(C\be^{-2})d\be\int_{C(t-r)^{-1}}^{\frac{C}2}\frac{\mu^{p_S}(\la)}{\la}d\la\ .
\end{split}
\eeq
Similar to \eqref{deri-main}, we have
\beq
(\be^{1-p_S}-(t-r)\be^{-p_S})\mu(C\be^{-2})=\left(\frac{\be^{2-p_S}\mu(C\be^{-2})}{2-p_S}-\frac{(t-r)\be^{1-p_S}\mu(C\be^{-2})}{1-p_S}\right)^{\pri}+e(\be)
\eeq
where the error term $e(\be)$ is
\beq
\begin{split}
|e(\be)|&=\left|\frac{2(t-r)\be^{-p_S}(C\be^{-2}\mu^{\pri}(C\be^{-2}))}{p_S-1}-\frac{2\be^{1-p_S}(C\be^{-2}\mu^{\pri}(C\be^{-2}))}{p_S-2}\right|\\
&\le \ti{c_*}\be^{1-p_S}\mu(C\be^{-2})
\end{split}
\eeq
with $\ti{c_*}\ll1$ (at least for $t-r>R_0^k\gg1$). Then the last integral is reduced to
\beq
\begin{split}
&\int_{t-r}^{t+r}(\be^{1-p_S}-(t-r)\be^{-p_S})\mu(C\be^{-2})d\be\\
\ge&\int_{t-r}^{t+r}\left(\frac{\be^{2-p_S}\mu(C\be^{-2})}{2-p_S}-\frac{(t-r)\be^{1-p_S}\mu(C\be^{-2})}{1-p_S}\right)^{\pri}-\ti{c_*}\be^{1-p_S}\mu(C\be^{-2})d\be\\
\ge&\frac{(t-r)^{2-p_S}\mu(C(t-r)^{-2})}{(p_S-2)(p_S-1)}+\left[\frac{1}{p_S-1}(\frac{t-r}{t+r})-\frac{1}{p_S-2}\right](t+r)^{2-p_S}\mu(C(t+r)^{-2})\\
&-\frac{\ti{c_*}}{p_S-2-2c_*}\left[(t-r)^{2-p_S}\mu(C(t-r)^{-2})-(t+r)^{2-p_S}\mu(C(t+r)^{-2})\right]\ ,
\end{split}
\eeq
where
$$\be^{1-p_S}\mu(C\be^{-2})\le\left(\frac{-\be^{2-p_S}\mu(C\be^{-2})}{p_S-2-2c_*}\right)^{\pri}$$
by \eqref{deri-main}.
It means for $t-r>R_0^k\gg1$ and $\frac{2^6-1}{2^6+1}t\le r<t$, we have
$$u(t,r)\ge C(t+r)^{-1}(t-r)^{-\ps}\mu(C(t-r)^{-2})\int_{C(t-r)^{-1}}^{\frac{C}2}\frac{\mu^{p_S}(\la)}{\la}d\la\ .$$
%On the other hand, for $r<k_0t$,
\subsection{Iteration procedure}
Now we return to the proof of Theorem \ref{blowup-thm} with iteration frame \eqref{iter-frame} and initial lower bound \eqref{first-lower}.
Since we have the initial lower bound \eqref{first-lower}, which means for $t\ge t_0\gg1$, we have
$$H(t)\ge B_k(\log^kt)^{1-\ps}\log^{k+1} t\ge 1\ .$$
Thus by \eqref{ti-def} and \eqref{muM_k}, \eqref{iter-frame} yields
\beeq\label{iter-newform}
H(t)\ge C_kt^{-1}\int_{t_0}^t(t-s)s^{-1}(\log s)^{-1}\cdots(\log^{k-1}s)^{-1}H^{p_S}(s)ds\ti{\mu}(t^{-2})(\log^kt)^{-\frac{p_S}{p_S^{\pri}}}\ .
\eneq
Let us choose the sequence $\{m_j\}_{j\in\N}$ with $m_j:=2-2^{-(j+1)}$. And our goal is to verify that the functional $H(t)$ has the following lower bound estimate:
\beeq\label{iter-ineq}
H(t)\ge N_j(\log^k \frac{r_kt}{m_{2j}t_0})^{a_j}(\log^kt)^{-l_j}(\log^{k+1}\frac{m_0}{m_{2j}}t)^{b_j}\ti{\mu}^{\sig_j}(L_0t^{-2})
\eneq
for $t\ge m_{2j}t_0$, where $\log^kr_k=0$,
$N_j,a_j,l_j,b_j,\sig_j\ge0$, and
\beeq\label{initial-j}
a_0=1\ ,\ l_0=0\ ,\ b_0=1\ ,\ \sig_0=1\ ,\ N_0=B_k\ .
\eneq
\begin{prop}\label{H-iter-j+1}
Under the assumption of Theorem \ref{blowup-thm}, and
suppose that \eqref{iter-ineq} holds for some $j\in\N$, with $b_j\le a_j$, then we obtain \eqref{iter-ineq} holds for $j+1$, with
\beq
\begin{split}
a_{j+1}&=a_jp_S+1\ ,\\
l_{j+1}&=l_jp_S+\frac{p_S}{p_S^{\pri}}\ ,\\
b_{j+1}&=b_jp_S\ ,\\
\sig_{j+1}&=\sig_jp_S+1\ ,\\
N_{j+1}&=\frac{C_k(m_{2j+2}-m_{2j})}{L_k(a_jp_S+1)m_{2j+2}}N_j^{p_S}\ ,
\end{split}
\eeq
where $L_k$ is a constant depending only on $k$.
\end{prop}
\begin{proof}
By \eqref{iter-newform} and \eqref{iter-ineq}, we have
\beq
\begin{split}
H(t)\ge&C_kN_j^{p_S}t^{-1}\ti{\mu}^{\sig_jp_S+1}(t^{-2})(\log^kt)^{-\frac{p_S}{p_S^{\pri}}-l_jp_S}\times\\
&\int_{m_{2j}t_0}^t(t-s)s^{-1}(\log s)^{-1}\cdots(\log^{k-1}s)^{-1}(\log^k \frac{r_ks}{m_{2j}t_0})^{a_jp_S}(\log^{k+1}\frac{m_0}{m_{2j}}s)^{b_jp_S}ds\ .
\end{split}
\eeq
Observe that
\beq
\begin{split}
&\left(\frac{(\log^k \frac{r_ks}{m_{2j}t_0})^{a_jp_S+1}(\log^{k+1}\frac{m_0}{m_{2j}}s)^{b_jp_S}}{a_jp_S+1}\right)^{\pri}\\
&=s^{-1}(\log s)^{-1}\cdots(\log^{k-1}s)^{-1}(\log^k \frac{r_ks}{m_{2j}t_0})^{a_jp_S}(\log^{k+1}\frac{m_0}{m_{2j}}s)^{b_jp_S}\times\\
&\left(\frac{\log s}{\log\frac{r_ks}{m_{2j}t_0}}\cdots\frac{\log^{k-1} s}{\log^{k-1}\frac{r_ks}{m_{2j}t_0}}+\frac{b_jp_S}{a_jp_S+1}\frac{\log^k\frac{r_ks}{m_{2j}t_0}}{\log^k \frac{m_0}{m_{2j}}s}\frac{\log s}{\log\frac{m_0}{m_{2j}}s}\cdots\frac{\log^{k-1} s}{\log^{k-1}\frac{m_0}{m_{2j}}s}(\log^{k+1}\frac{m_0}{m_{2j}}s)^{-1}\right)\ ,
\end{split}
\eeq
it is easy to obtain that there exists a constant $L_{k,j}>0$ so that
\beeq\label{L-kj}
\frac{\log s}{\log\frac{r_ks}{m_{2j}t_0}}\cdots\frac{\log^{k-1} s}{\log^{k-1}\frac{r_ks}{m_{2j}t_0}}+\frac{b_jp_S}{a_jp_S+1}\frac{\log^k\frac{r_ks}{m_{2j}t_0}}{\log^k \frac{m_0}{m_{2j}}s}\frac{\log s}{\log\frac{m_0}{m_{2j}}s}\cdots\frac{\log^{k-1} s}{\log^{k-1}\frac{m_0}{m_{2j}}s}(\log^{k+1}m_0t_0)^{-1}\le L_{k,j}\eneq
for $s\ge m_{2j}t_0$.
Thus we arrive at
\beq
\begin{split}
H(t)\ge&\frac{C_k}{L_{k,j}}N_j^{p_S}t^{-1}\ti{\mu}^{\sig_jp_S+1}(t^{-2})(\log^kt)^{-\frac{p_S}{p_S^{\pri}}-l_jp_S}\times\\
&\int_{m_{2j}t_0}^t(t-s)\left(\frac{(\log^k \frac{r_ks}{m_{2j}t_0})^{a_jp_S+1}(\log^{k+1}\frac{m_0}{m_{2j}}s)^{b_jp_S}}{a_jp_S+1}\right)^{\pri}ds\\
\ge&\frac{C_k}{L_{k,j}}N_j^{p_S}t^{-1}\ti{\mu}^{\sig_jp_S+1}(t^{-2})(\log^kt)^{-\frac{p_S}{p_S^{\pri}}-l_jp_S}\times\\
&\int_{\frac{m_{2j}t}{m_{2j+2}}}^t\frac{(\log^k \frac{r_ks}{m_{2j}t_0})^{a_jp_S+1}(\log^{k+1}\frac{m_0}{m_{2j}}s)^{b_jp_S}}{a_jp_S+1}ds\\
\ge&\frac{C_kN_j^{p_S}(m_{2j+2}-m_{2j})}{L_{k,j}(a_jp_S+1)m_{2j+2}}\times\\
&(\log^k \frac{r_kt}{m_{2j+2}t_0})^{a_jp_S+1}(\log^kt)^{-\frac{p_S}{p_S^{\pri}}-l_jp_S}(\log^{k+1}\frac{m_0}{m_{2j+2}}t)^{b_jp_S}\ti{\mu}^{\sig_jp_S+1}(t^{-2})
\end{split}
\eeq
for $t\ge m_{2j+2}t_0$, that is $\frac{m_{2j}t}{m_{2j+2}}\ge m_{2j}t_0$.
Finally, we need to show that actually there exists a uniform upper bound $L_k>0$ in \eqref{L-kj} for all $j\in\N$.
%there exists $L_k>0$ such that $L_{k,j}\le L_k$ for all $j\in\N$.
In fact, due to
$$b_0=a_0\ ,\ b_{j+1}=b_jp_S\ ,\ a_{j+1}=a_jp_S+1\ \ra\ \frac{b_jp_S}{a_jp_S+1}\le1\ ,$$
we finsh the proof
with the fact that $m_{2j}\in[\frac32,2]$ and $\frac{\log^l s}{\log^l\frac{s}{a}}$ is a continuous function of $(s,a)$ in the domain.
\end{proof}
By Proposition \ref{H-iter-j+1}, we deduce that
\beq
\begin{split}
a_{j+1}&=a_jp_S+1=p^{j+1}+p^{j}+\cdots+1=\frac{p_S^{j+2}-1}{p_S-1}\ ,\\
l_{j+1}&=l_jp_S+\frac{p_S}{p_S^{\pri}}=p_S^{j+1}-1\ ,\\
b_{j+1}&=b_jp_S=p_S^{j+1}\ ,\\
\sig_{j+1}&=\sig_jp_S+1=\frac{p_S^{j+2}-1}{p_S-1}\ ,\\
N_{j+1}&=\frac{C_k(m_{2j+2}-m_{2j})}{L_k(a_jp_S+1)m_{2j+2}}N_j^{p_S}\ge\frac{3C_k2^{-(2j+3)}}{4L_kp^{j+1}}N_j^{p_S}\\
&\ge \frac{3C_k}{8L_k(4p_S)^{(j+1)}}N_j^{p_S}\ ,
\end{split}
\eeq
with $\frac{p_S}{p_S^{\pri}}=p_S-1$. Thus, the estimate of $N_j$ yields
\beq
\begin{split}
\log N_j&\ge \log\frac{3C_k}{8L_k}-j\log 4p_S+p_S\log N_{j-1}\\
&\ge p_S^j\left(\log N_0-\log 4p_S\sum_{l=1}^j\frac{l}{p_S^l}+\log\frac{3C_k}{8L_k}\sum_{l=1}^j\frac{1}{p_S^l}\right)=:\ti{C}_jp_S^j\ .
\end{split}
\eeq
Finally, we have
$$H(t)\ge \exp\left[{p_S^j}\left(\ti{C}_j+\log\left(A_k(\log^kt)^{\frac{1}{p_S-1}}(\log^{k+1}t)\ti{\mu}^{\frac{p_S}{p_S-1}}(t^{-2})\right)\right)\right](\log^k t)^{\frac{p_S-2}{p_S-1}}\ti{\mu}^{-\frac1{p_S-1}}(t^{-2})
$$
for sufficiently large $t>t_0$, and a constant $A_k>0$.
%where
%$$\ti{C_0}=\log M_0-\log4p_S\sum_{l=1}^{\infty}\frac{l}{p_S^l}+\log C_0\sum_{l=0}^{\infty}\frac1{p_S^l}\ .$$
Since $\ti{\mu}(\la)=(\log^k\frac1{\la})^{-\ps}$ and $\ti{C}_j$ will converge to a constant, we obtain $H(t)$ blows up at some sufficiently large time $t>t_0$ as $j\to\infty$.
%\begin{remark}
%From Proposition \ref{pointwise-low}, we observe that the critical term
%$$\int_0^{\la_0}\frac{\mu^{p_S}(\la)}{\la}d\la$$
%will appear inherently. With the analysis of global existence, we conclude our conjecture of the critical condition.
%\end{remark}
\section{Appendix}
In this section we include some useful results. And the first is the following generalized version of Jensen’s inequality \cite{MR3024912}.%还没翻过这本书，去找一下
\begin{lem}\label{jensen-ineq}
Let $g=g(\tau)$ be a convex function on $\R$ and $\al=\al(x)$ be defined non-negative almost everywhere on $\Om$, such that $\al$ is positive in a set of positive measure. Then it holds
$$g\left(\frac{\int_{\Om}v(x)\al(x)dx}{\int_{\Om}\al(x)dx}\right)\le\frac{\int_{\Om}g(v(x))\al(x)dx}{\int_{\Om}\al(x)dx}$$
for all non-negative functions $v=v(x)$ provided that all the integral terms are meaningful.
\end{lem}
\bibliography{bib.bib}
\end{document}
\begin{thebibliography}{10}
\bibitem{Chen_2024}
Wenhui Chen and Michael Reissig.
\newblock On the critical regularity of nonlinearities for semilinear classical
wave equations.
\newblock {\em Mathematische Annalen}, April 2024.
\bibitem{MR4163528}
M.~R. Ebert, G.~Girardi, and M.~Reissig.
\newblock Critical regularity of nonlinearities in semilinear classical damped
wave equations.
\newblock {\em Math. Ann.}, 378(3-4):1311--1326, 2020.
\bibitem{MR1481816}
Vladimir Georgiev, Hans Lindblad, and Christopher~D. Sogge.
\newblock Weighted {S}trichartz estimates and global existence for semilinear
wave equations.
\newblock {\em Amer. J. Math.}, 119(6):1291--1319, 1997.
\bibitem{MR3024912}
Lubo\v{s} Pick, Alois Kufner, Old\v{r}ich John, and Svatopluk Fu\v{c}\'{\i}k.
\newblock {\em Function spaces. {V}ol. 1}, volume~14 of {\em De Gruyter Series
in Nonlinear Analysis and Applications}.
\newblock Walter de Gruyter \& Co., Berlin, extended edition, 2013.
\bibitem{MR824205}
Jack Schaeffer.
\newblock The equation {$u_{tt}-\Delta u=|u|^p$} for the critical value of
{$p$}.
\newblock {\em Proc. Roy. Soc. Edinburgh Sect. A}, 101(1-2):31--44, 1985.
\bibitem{MR744303}
Thomas~C. Sideris.
\newblock Nonexistence of global solutions to semilinear wave equations in high
dimensions.
\newblock {\em J. Differential Equations}, 52(3):378--406, 1984.
\bibitem{MR1715192}
Christopher~D. Sogge.
\newblock {\em Lectures on nonlinear wave equations}.
\newblock Monographs in Analysis, II. International Press, Boston, MA, 1995.
\bibitem{MR1804518}
Daniel Tataru.
\newblock Strichartz estimates in the hyperbolic space and global existence for
the semilinear wave equation.
\newblock {\em Trans. Amer. Math. Soc.}, 353(2):795--807, 2001.
\bibitem{MR3912752}
Kyouhei Wakasa and Borislav Yordanov.
\newblock Blow-up of solutions to critical semilinear wave equations with
variable coefficients.
\newblock {\em J. Differential Equations}, 266(9):5360--5376, 2019.
\bibitem{wang2023wave}
Chengbo Wang and Xiaoran Zhang.
\newblock Wave equations with logarithmic nonlinearity on hyperbolic spaces,
arXiv:2304.01595, 2023.
\bibitem{MR2195336}
Borislav~T. Yordanov and Qi~S. Zhang.
\newblock Finite time blow up for critical wave equations in high dimensions.
\newblock {\em J. Funct. Anal.}, 231(2):361--374, 2006.
\bibitem{zhang2023critical}
Xiaoran Zhang.
\newblock The critical conjecture of wave equations with logarithmic
nonlinearity on $\mathbb{H}^2$, arXiv:2309.10384, 2023.
\bibitem{MR2316656}
Yi~Zhou.
\newblock Blow up of solutions to semilinear wave equations with critical
exponent in high dimensions.
\newblock {\em Chinese Ann. Math. Ser. B}, 28(2):205--212, 2007.
\end{thebibliography}