The blow-up rate for a loglog non-scaling invariant semilinear wave equation

Extracted source text · arXiv 2308.12220

%\documentclass[a4paper,11pt]{article}
\documentclass{amsart}
\usepackage{amsthm}
\usepackage{amssymb, latexsym}
%\usepackage[notcite,notref]{showkeys}
\usepackage{enumerate}
\usepackage{mathtools,color}
\usepackage{dsfont}
\usepackage{bbm}
\usepackage{ulem}

\usepackage{hyperref}

\usepackage{bbold}

%\bibliographystyle{amsplain}

\newcommand\dive{\operatorname{div}}
\newcommand\supp{\operatorname{supp}}
\newcommand\card{\operatorname{card}}
\newcommand\gu{\operatorname{g(u)}}
\newcommand\gM{\operatorname{g (M)}}
\newcommand\gfivesixM{\operatorname{g^{\frac{5}{6}}(M) }}
\newcommand\gfivesixu{\operatorname{g^{\frac{5}{6}} (u)}}
\newcommand\gonesixM{\operatorname{g^{\frac{1}{6}} (M) }}
\newcommand\gonesixu{\operatorname{g^{\frac{1}{6}} (u) }}
\newcommand\gfivetwfourM{\operatorname{g^{\frac{5}{24}}(M) }}
\newcommand\loglog{\operatorname{loglog}}

\newcommand {\bc}{\color{black}}
\newcommand {\rc}{\color{red}}
\newcommand {\blc}{\color{blue}}

\newcommand \D{< \hspace{-1mm} D \hspace{-1mm} >}
\newcommand \Dfrac{< \hspace{-0.1mm} D \hspace{-0.1mm} >}

\newcommand \lesssimo{\lesssim_{ _{ {\underset { \| u_{0} \|_{H^{2}} } {\| u_{1} \|_{H^{1}}}}} }}
\newcommand \veryless{<<_{ _{ {\underset { \| u_{0} \|_{H^{2}} } {\| u_{1} \|_{H^{1}}}}} }}


\newcommand \tb {\tau}
\newcommand \Ivl {\left( [0, \, \tau] \right) \, }
\newcommand \Tb {\overline{T}}

\newcommand{\m}[1]{\mathbb{#1}}

\newcommand \maxJ {\max{(1,|J|)}}
\newcommand \maxldt {\max{\left(1,\lambda T \right)}}
\newcommand \Nlow{\left( N_{4} \right)^{*}}
\newcommand \Nhigh{\left( N_{1} \right)^{*}}


\newcommand{\q}[1]{\mathcal{#1}}

\newcommand \ulJ{u^{l,J}}
\newcommand \unlJ{u^{nl,J}}


%\theoremstyle{plain}
%\newtheorem{name}{Printed output}
\newtheorem{prop}{Proposition}
\newtheorem{cor}{Corollary}
\newtheorem{thm}[prop]{Theorem}
\newtheorem{rem}{Remark}
%\newtheorem{definition}[thm]{Definition}
\newtheorem{lem}[prop]{Lemma}
\newtheorem{res}{Result}
%\newtheorem{corollary}[thm]{Corollary}
%\newtheorem{assumption}[thm]{Assumption}
\newtheorem{ack}{Acknowledgments}
\newtheorem*{nb}{Remark}

\numberwithin{equation}{section}
%\numberwithin{thm}{section}

\begin{document}

\title[Non-scaling invariant semilinear wave equations] {The blow-up rate for a loglog non-scaling invariant semilinear wave equation}

\author[Tristan Roy and Hatem Zaag]{{Tristan Roy}\\
 \textit{\tiny American University of Beirut, Department of Mathematics}\\
{Hatem Zaag}\\
{\it \tiny
  Universit\'e Sorbonne Paris Nord,  LAGA, CNRS (UMR 7539)},\\
 {\it \tiny F-93430, Villetaneuse, France}
}

\address{TR: American University of Beirut, Department of Mathematics.
HZ: Universit\'e Sorbonne Paris Nord, Institut Galil\'ee,
Laboratoire Analyse, G\'eom\'etrie et Applications, CNRS UMR 7539,
99 avenue J.B. Cl\'ement, 93430 Villetaneuse, France.
}
\email{tr14@aub.edu.lb, hatem.zaag@univ-paris13.fr}

\begin{abstract}

  We consider blow-up solutions of a semilinear wave equation with a $\log\log$ perturbation of the power
  nonlinearity in the subconformal case, and show that the blow-up rate is given by
  the solution of the associated ODE which has the same blow-up time. In fact, our result shows an
  upper bound and a lower bound of the blow-up rate, both proportional to the blow-up solution of the
  associated ODE. As for the upper bound, the main difficulty comes from the fact that the PDE is not
  scaling invariant. Our argument is a delicate adaptation of the
  original approach introduced by Hamza and Zaag for the $\log$
  perturbation of the pure power nonlinearity (see
  \cite{HZjmaa20}). It mainly relies on the design of a Lyapunov
  functional and the use of a  blow-up criterion in the similarity
  variables' setting, together with some energy and interpolation
  estimates. As for the proof of the lower bound, we consider
  our argument as a major contribution, since the argument for the
  pure power case cannot be adapted when the scale invariance breaks
  down.
\end{abstract}

\maketitle

% \textbf{Id\'ees:\\
%   - Citer d'autres travaux sur log log pour d'autres \'equations, en
%   particulier le travail r\'ecent de Souplet..}. Par exemple:\\
% T. Roy. Global existence of smooth solutions of a 3D log-log energy-supercritical
% wave equation, Anal. PDE, 2(3), 261–280, 2009.

%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%
%\subsection{Earlier results}
%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%

Many papers are devoted to the study of finite time blow-up for
semilinear wave equations of the type
\begin{equation}\label{equ}
\partial_{tt} u -  \Delta u = f(u)
\end{equation}
considered in the whole space $\m R^N$,
where $f:\m R \longrightarrow \m R$ is regular and chosen so that the
associated ODE $u''=f(u)$ exhibits blow-up. The simplest example is
certainly the case where
\begin{equation}\label{fup}
  f(u)=|u|^{p-1}u \mbox{ with }p>1,
\end{equation}
  considered by
many authors (see John \cite{Jmm79},...).

\medskip

The Cauchy problem for \eqref{equ}-\eqref{fup} can be solved in
$H^1\times L^2(\m R^N)$, locally in time, at least for Sobolev subcritical $p$ (see for
example Shatah and Struwe \cite{SSnyu98}). If $u$ is a maximal
solution whose existence time interval is finite, following Alinhac
\cite{Apndeta95} (see also Appendix \ref{Sec:InflDomain}),  we may use the finite speed of propagation to
extend the domain of definition of $u$ to the set
\begin{equation}\label{defD}
\q D= \{(x,t)\;|\;0\le t<T(x)\},
\end{equation}
for some $1$-Lipschitz function $x\longmapsto T(x)$ with the following
property: the $H^1\times L^2$ average of the solution on sections of
any light cone with vertex $(x,T(x))$ blows up as the section
approaches the vertex. The boundary $\{t=T(x)\}$ of $\q D$ is called
the ``blow-up surface''.
% Locally in backward
% light cones with vertex $(x,T(x))$ for any $x\in \m R^N$, we may
% derive from the wellposedness in $H^1\times L^2(\m R^N)$ togethe
In other words, the solution $u$ enjoys a ``local'' blow-up time $T(x)$, near
each $x\in \m R^N$. Note that the $1$-Lipschitz property of $T$
implies that $u$ is defined
on the backward light cone $\q C_{x,1}$ for any $x\in \m R^N$, where for any $\delta\ge
0$, $\q C_{x,\delta}$ is the backward cone with vertex $(x, T(x))$ and
slope $\delta$, formally defined by
\[
\q C_{x,\delta}=\{(\xi,\tau)\;|\; 0\le \tau < T(x) -
\delta |\xi-x|\}.
\]
If ever $x\in \m R^N$ is such that $u$ is defined on $\q C_{x,\delta}$
for some $\delta<1$, then $x$ and  $\q C_{x,\delta}$ are called
``non-characteristic''. Otherwise, we call them ``characteristic''. The set of all
non-characteristic points is denoted by $\q R$ and its complementary
by $\q S$.

\medskip

% COMMENTS
%textbf{Remarque pour Tristan : Dans ta version dat\'ee du 15
 % d\'ecembre 2022, tu d\'efinis un point caract\'eristique ainsi :\\
%We recall that a point $ x_{0}$ is a non-characteristic point if}
%\begin{equation*}
%\begin{array}{l}
%\exists \; \left( \delta_{0}(x_{0}),\bar{t} \right) \in (0,1) \times %[0,\infty) \;  \text{such that} \; u \; \text{is defined on} \;
%C_{x_{0}} :=  \left\{ (t,x): \;  \bar{t} \leq t < T(x_{0}) - %\delta_{0}(x_{0}) |x- x_{0}| \right\} \cdot
%\end{array}
%\label{Eqn:DefNonCharac}
%\end{equation*}
%\textbf{En fait, on peut prendre comme d\'efinition
%\'equivalente la m\^eme chose avec $\bar t=0$, quitte \`a changer le
%$\delta_0(x_0)$. Du coup, c'est ce que je propose dans cette version.}

\medskip

Following these definitions, it is natural to ask whether one can derive
the ``local blow-up rate''. Such a question was solved by Merle and
Zaag in \cite{MZajm03} and \cite{MZimrn05}, in the subconformal case, where
\begin{equation}\label{condp}
p<\frac{N+3}{N-1}\mbox{ if } N\ge 2.
\end{equation}
Indeed, in \cite{MZimrn05}, given $x_0\in \q R$, the
authors proved that for any $t\in [0,T(x_0))$,
% \begin{equation}
%   \begin{array}{rl}
    \begin{align}
  k (T(x_0)-t)^{-\frac 2{p-1}}
  \leq& \frac{\| u(t) \|_{L^{2} \left( B \left( x_{0}, T(x_{0}) - t \right) \right) } }
{ \left(  T(x_{0}) - t \right)^{\frac{N}{2}}}\nonumber\\
+&(T(x_0)-t)
\frac{ \| \nabla u(t) \|_{L^{2} \left( B(x_{0},T(x_{0})-t) \right)} +  \| \partial_{t} u(t) \|_{L^{2} \left( B \left( x_{0}, T(x_{0}) - t \right) \right)}} {\left(  T(x_{0}) - t \right)^{\frac{N}{2}}}\nonumber\\
      \leq \; & K (T(x_0)-t)^{-\frac 2{p-1}} ,\label{2bounds}
    \end{align}
% \end{array}
% \label{Eqn:BlowUpRate}
% \end{equation}
for some $0<k(N,p)\le K$, where $K$ depends only on an upper bound on
$T(x_0)$, $1/T(x_0)$, the
local initial norm $\| \left( u
(0), \partial_{t} u (0) \right) \|_{H^{1} \times L^{2}  \left( B \left(
      x_{0},  \frac{T(x_{0})}{\delta_{0}(x_{0})} \right)\right)}$
% $K=K  \left( \delta_{0}(x_{0}), T(x_0),
% \| u (0), \partial_{t} u (0) \|_{H^{1} \times L^{2}  \left( B \left(
%       x_{0},  \frac{T(x_{0})}{\delta_{0}(x_{0})} \right) \right) }
% \right) $
and $\delta_0(x_0)<1$, the slope of some non-characteristic
cone $\q C_{x_0,\delta_0(x_0)}\subset \q D$. Note that up to some factor, both the lower
and the upper bounds show the solution of the associated ODE
$u''=|u|^{p-1}u$ which blows up at time $T(x_0)$. The proof is given
in the framework of similarity variables first introduced in Antonini
and Merle \cite{AMimrn01} and defined for any $x_0\in \m
R^N$ and $T_0\in[0,T(x_0))$ by
\begin{equation}\label{defw}
\begin{array}{l}
y = \frac{x-x_{0}}{T_{0} - t}, \; s= - \log(T_{0} - t), \; \text{and}
  \;
  u(x,t) = (T_{0}-t)^{\frac 2{p-1}} w_{x_{0},T_{0}} (y,s) \cdot
\end{array}
\end{equation}
Given that $u$ is defined (at least) in the backward light-cone $\q
C_{x_0,T_0}$, as we have just mentioned, it follows that $w_{x_0,T_0}$
(or $w$ for simplicity) satisfies the following equation for all $s\ge
-\log T_0$ and $y\in B(0,1)$:
\begin{equation}\label{eqw}
\begin{array}{ll}
\partial_{s}^{2} w  & =  \frac{1}{\rho} \dive \left( \rho \nabla w - \rho (y \cdot \nabla w) y \right)
 - \frac{2(p+1)}{(p-1)^{2}} w +|w|^{p-1}w- \frac{p+3}{p-1} \partial_{s} w
 - 2 y \cdot \nabla \partial_{s} w
\end{array}
\end{equation}
where
\begin{equation}\label{Eqn:ValAlpha}
\rho(y):= (1 - |y|^{2})^{\alpha}\mbox{ with }\alpha := \frac{2}{p-1} - \frac{N-1}{2} > 0.
\end{equation}
Again from \cite{AMimrn01}, we know that equation \eqref{eqw} has the
following Lyapunov functional
\begin{align}
E(w(s)) :=
\int_{B(0,1)} &\left( \frac{1}{2} (\partial_{s} w(s) )^{2} + \frac{1}{2} \left( |\nabla w(s)|^{2} - |y \cdot \nabla w(s)|^{2} \right)
  +  \frac{p+1}{(p-1)^{2}} w^{2}\right. \label{defE}\\
              &\left. - |w|^{p+1}\right) \rho \; dy.\nonumber
\end{align}
This functional, combined to a blow-up criterion in the similarity
variables' setting, makes the core of the argument of \cite{MZajm03}
and \cite{MZimrn05}, together with some involved energy and interpolation estimates.

\medskip

Further refinements (both in the non-characteristic and
the the characteristic cases) or generalizations (in particular to the conformal and
superconformal cases) are available in
\cite{MZma05}, \cite{MZjfa07}, \cite{MZcmp08}, \cite{MZajm12},
\cite{MZdmj12}, \cite{HZdcds13}, \cite{CZcpam13}, \cite{MZtams16}, \cite{MZcmp15} and
\cite{MZcpam18} (see also the notes \cite{MZsnp09},  \cite{MZxedp10}, \cite{MZsls17}).
Since our focus in this paper is the blow-up rate in the
non-characteristic case for general nonlinearities in \eqref{equ}, we
won't mention those developments.

% Since our focus in this paper is the blow-up rate in the
% non-characteristic case for general nonlinearities in \eqref{equ}, we
% won't mention further refinements (both in the non-characteristic and
% the the characteristic cases) or generalizations (in particular to the conformal and
% superconformal cases). We will instead refer the
% interested reader to the following papers
% \cite{MZma05}, \cite{MZjfa07}, \cite{MZcmp08}, \cite{MZajm12},
% \cite{MZdmj12}, \cite{HZdcds13}, \cite{CZcpam13}, \cite{MZtams16},
% \cite{MZcmp15} and \cite{MZcpam18} (see also the notes
% \cite{MZsnp09},  \cite{MZxedp10}, \cite{MZsls17}).

\medskip

Following our result in the case \eqref{fup}, we wonder whether the
same holds for other nonlinearities, namely that the blow-up rate near
non-characteristic points is given by the solution of the associated
ODE
\begin{equation}\label{ode}
u"=f(u).
\end{equation}
Given that the Lypunov functional $E(w)$ \eqref{defE} is central in
the argument, and that its existence seems natural, from the fact that
equation \eqref{eqw} is autonomous, we thought for a while that
extending the result to other nonlinearities would be impossible,
unless the similarity variables' equation is autonomous.

\medskip

That belief was proven to be wrong thanks to the work of Hamza and
Zaag in \cite{HZnonl12} together with Hamza and Saidi \cite{HSjdde14},
devoted to the case where
% The first answer outside the case \eqref{fup} was given by Hamza and
% Zaag in \cite{HZnonl12} and then extended by Hamza and Saidi \cite{HSjdde14}, for the case
\begin{equation}\label{pert1}
  f(u) = |u|^{p-1}u+g(u)
\end{equation}
with $p$ subconformal in the sense \eqref{condp} and $g$ $C^1$ satisfying
\begin{equation}\label{condg}
  |g(u)|\le M\left(1+\frac{|u|^p}{[\log(2+|u|^2)]^a}\right)\mbox{ for some }a\ge 1.
%  0\le q<p.
\end{equation}
Indeed, the statement \eqref{2bounds} holds in this case, and even for more
general nonlinearities involving $x$, $t$ and also $\nabla u$ or $\partial_t u$, with a
sublinear growth, despite the fact that the similarity variables'
equation, which looks as a perturbation of \eqref{eqw}, is not
autonomous, as one may easily check. In fact, the proof in the
case \eqref{pert1} follows from a delicate and involved
adaptation of the argument of \cite{MZajm03} given in the pure power case
\eqref{fup}. It relies in particular on the design of a new Lyapunov
functional, which appears to be a clever perturbation of \eqref{defE},
in order to take into account the time dependent terms one gets in
equation \eqref{eqw}, due to the additional term in \eqref{pert1}.

\medskip

Following \cite{HZnonl12} and \cite{HSjdde14} devoted to the case
\eqref{pert1}, we still had the impression that we were handling a case
where the similarity variables' equation is asymptotically autonomous,
as one may easily check, since the main term of the nonlinearity at
infinity is a pure power. In consequence, the next challenge was to
handle a nonlinear term where the main term at infinity is not a pure power.


% ........


% The proof in the
% case \eqref{pert1} follows from a delicate and involved
% adaptation of the argument of \cite{MZajm03} in the pure power case
% \cite{fup}. It relies in particular on the design of a new Lyapunov
% functional, which appears to be a clever perturbation of the
% functional used in \cite{MZajm03}. However, one may note that the
% blow-up rate of the ODE \eqref{ode}-\eqref{pert1} is the same as in the pure power
% case, because the additional term $g(u)$ in \eqref{pert1} appears to be
% of lower order at infinity, thanks to \eqref{condg}.
% % hence doesn't change affect the main term of the solution at blow-up.
% % In other words, the blow-up rate is still given by the main nonlinear term,
% % namely $|u|^{p-1}u$, and we don't exhibit a new blow-up rate with this
% % example.

% \medskip

% In the fact that the main term is a pure power necessary (that gives
% an autonomous PDE is similarity variables, though with some
% pertrubations....

\medskip

There comes the recent contribution of Hamza and Zaag in
\cite{HZjmaa20} and \cite{HZna21}, where the authors consider the case
\[
f(u)=|u|^{p-1}u[\log(2+u^2)]^a\mbox{ with }p\mbox{ satisfying
\eqref{condp} and }a\in \m R.
\]
In that case, the authors prove an analogous statement to
\eqref{2bounds}, where the factor $(T(x_0)-t)^{-\frac 2{p-1}}$ is
replaced by $\psi(T(x_0)-t)$, and $\psi$ is a solution of the
associated ODE \eqref{ode} blowing up at time $t=0$. Although the
proof follows the strategy of the pure power case \eqref{fup}, the
adaptation is not easy to guess, in the sense that the similarity
variables' definition has to be adapted, and that the new Lyapunov
functional is not a perturbation of \eqref{defE}.

%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%
%\subsection{Non scaling invariant nonlinearities}
% \subsection{Slowly varying nonlinearities}
%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%

\bigskip

In this paper, we further investigate the case where
the nonlinearity is not a pure power, and focus on the following
``$\log \log$'' perturbation of \eqref{fup}, where
\begin{equation}\label{pert2}
f(u) =  |u|^{p-1} u g(u) \mbox{ with } g(u) := \log^{a} \left( \log
  \left( 10 + u^{2} \right) \right)\mbox{ and } a \in \mathbb{R}.
\end{equation}
Note that $\log \log$ nonlinearities of this style have already been
considered in the literature, in particular by Roy in \cite{Rapde09}.
% Let us note that with both examples \eqref{pert1} and \eqref{pert2},
%  equation \eqref{equ} is not scaling inviariant.
%
% have
% \textit{``regular variation at infinity with index $p$''}, according to the terminology of
% Karamata \cite{Kmath30}.
%

\medskip

The local-in-time Cauchy problem for equation
\eqref{equ}-\eqref{pert2} %in $H^1\times L^2(\m R^N)$
is essentially well-known. %see e.g \cite{Tna00}.
For the reader's convenience
we write a proof in Appendix \ref{Sec:LwpH1L2}. The existence of blow-up
solutions follows from some cut-off argument applied to the solution
of the ODE \eqref{ode}. One may also use the concavity argument of
Levine \cite{Ltams74} to derive such solutions (see also Levine and
Todorova \cite{LTpams01}). Given a
blow-up solution, one may extend its domain of definition to a set $\q D$
\eqref{defD},  for some $1$-Lipschitz function $x\longmapsto T(x)$, as
for the pure-power case \eqref{fup}. The notion of characteristic and
non-characteristic points are introduced similarly.

\medskip

On the contrary, we take a different definition of similarity
variables, for any $x_{0} \in \mathbb{R}^N$ and $ T_0\in (0,T(x_{0})] $:
%
% We consider the following equation
%
% \begin{equation}
% \begin{array}{l}
% \partial_{tt} u -  \triangle u = f(u)
% \end{array}
% \label{Eqn:WaveFoc}
% \end{equation}
% with $ f(u) := |u|^{p-1} u g(u) $. Here $ g(x) := \log^{a} \left( \log \left( 10 + x^{2} \right) \right)$ with $ a \in \mathbb{R} $. Let $F$ denote the following function
%
% \begin{equation}
% \begin{array}{l}
% F(x) := \int_{0}^{x} f(x') \; dx' \cdot
% \end{array}
% \nonumber
% \end{equation}
%Let $x_{0} \in \mathbb{R}$. Let $ 0 < T_{0} < T(x_{0}) $. We define the similarity variables
%
\begin{equation}
\begin{array}{l}
y = \frac{x-x_{0}}{T_{0} - t}, \; s= - \log(T_{0} - t), \; \text{and} \; u(x,t) = \psi_{T_{0}}(t) w_{x_{0},T_{0}} (y,s) \cdot
\end{array}
\label{Eqn:SimilVar}
\end{equation}
Here $ \psi_{T_{0}}(t) := (T_{0} - t)^{-\frac{2}{p-1}}
\log^{-\frac{a}{p-1}} \left( - \log(T_{0}-t) \right) $, is the main
term of the solution of the ODE \eqref{ode} which blows up at time
$T_0$ (see Lemma \ref{Lem:OdeAsymp}). Using equation
\eqref{equ}-\eqref{pert2}, we see
%
% which is the asymptotic behavior of solutions
% of (\ref{Eqn:ProblemOdeV}) that contains the ODE associated to (\ref{Eqn:WaveFoc}): see Lemma \ref{Lem:OdeAsymp}. \\
%We see from (\ref{Eqn:WaveFoc})
that $w_{x_{0},T_{0}}$ satisfies the following equation:\footnote{We write $w$ for $w_{x_{0},T_{0}}$ for sake of simplicity.}

\begin{equation}
\begin{array}{ll}
\partial_{s}^{2} w  & =  \frac{1}{\rho} \dive \left( \rho \nabla w - \rho (y \cdot \nabla w) y \right) + \frac{2a}{(p-1) s \log(s)} y \cdot \nabla w  \\
& - \frac{2(p+1)}{(p-1)^{2}} w + \gamma(s) w - \left( \frac{p+3}{p-1} - \frac{2a}{(p-1)s \log(s)} \right) \partial_{s} w
 - 2 y \cdot \nabla \partial_{s} w \\
& + e^{-\frac{2sp}{p-1}} \log^{\frac{a}{p-1}}(s) f(\phi w),
\end{array}
\label{Eqn:Wavew}
\end{equation}
where $\rho(y)$ is defined in \eqref{Eqn:ValAlpha},
%
% Here $\rho(y):= (1 - |y|^{2})^{\alpha}$ with
%
\begin{align}
%\alpha := \frac{2}{p-1} - \frac{N-1}{2} > 0, \nonumber\\%\label{Eqn:ValAlpha} \\
%\nonumber \\
\gamma(s) &:= \frac{a(p+3)}{(p-1)^{2}s \log(s)} - \frac{a(a+p-1)}{(p-1)^{2} s^{2} \log^{2}(s)} - \frac{a}{(p-1) \log(s) s^{2}}   , \; \text{and} \label{Eqn:ValGamma}    \\
\nonumber \\
\phi(s) &:= e^{\frac{2s}{p-1}} (\log(s))^{-\frac{a}{p-1}} \cdot \label{Eqn:Defphi}
\end{align}
We also introduce the functional
% We define
$\mathcal{E}(w) : s \rightarrow \mathcal{E} (w(s)) $, which will serve
as a basis to derive a Lyapunov functional for equation
\eqref{Eqn:Wavew} (see below \eqref{Eqn:DefHN} together with Lemma
\ref{lemlyap}):
\begin{align}
\mathcal{E}(w(s)):=
\int_{B(O,1)} &\left( \frac{1}{2} (\partial_{s} w(s) )^{2} + \frac{1}{2} \left( |\nabla w(s)|^{2} - |y \cdot \nabla w(s)|^{2} \right)
  +  \frac{p+1}{(p-1)^{2}} w^{2}\right. \label{Eqn:Energyw}\\
 &\left. - e^{-\frac{2(p+1)s}{p-1}} (\log(s))^{\frac{2a}{p-1}} F(\phi w(s))  \right) \rho \; dy,\nonumber
\end{align}
where
\begin{equation}
\begin{array}{l}
F(x) := \int_{0}^{x} f(x') \; dx' \cdot
\end{array}
\nonumber
\end{equation}
%
% \begin{equation}
% \begin{array}{l}
% \mathcal{E}(w(s)) := \\
% \int_{B(O,1)} \left( \frac{1}{2} (\partial_{s} w(s) )^{2} + \frac{1}{2} \left( |\nabla w(s)|^{2} - |y \cdot \nabla w(s)|^{2} \right)
% +  \frac{p+1}{(p-1)^{2}} w^{2} - e^{-\frac{2(p+1)s}{p-1}} (\log(s))^{\frac{2a}{p-1}} F(\phi w(s))  \right) \rho \; dy
% \end{array}
% \label{Eqn:Energyw}
% \end{equation}
%
% \textbf{TO BE COMPLETED: BEGINNING}
%
% Define a blow-up graph;
%
% \textbf{TO BE COMPLETED: END}
%
% We recall that a point $ x_{0}$ is a non-characteristic point if
% there exists $\delta_{0}(x_{0}) \in (0,1)$ such that $u$ is defined on
% \begin{equation}\label{Eqn:DefNonCharac}
% C_{x_{0}} :=  \left\{ (t,x): \;  0 \leq t < T(x_{0}) - \delta_{0}(x_{0}) |x- x_{0}| \right\}.
% \end{equation}
% \begin{equation}
% \begin{array}{l}
% \exists \; \delta_{0}(x_{0}) \in (0,1) \;  \text{such that} \; u \; \text{is defined on} \;
% C_{x_{0}} :=  \left\{ (t,x): \;  0 \leq t < T(x_{0}) - \delta_{0}(x_{0}) |x- x_{0}| \right\} \cdot
% \end{array}
% \label{Eqn:DefNonCharac}
% \end{equation}
%
% \textbf{TO BE COMPLETED: BEGINNING} \\
% \\
% Literature; \\
% \\
% \textbf{TO BE COMPLETED: END} \\
% \\
Thanks to a delicate adaptation of the previous literature, we are
able to confirm that the blow-up rate in the non-characteristic case
is given by the solution of the associated ODE \eqref{ode}. More
precisely, this is our main statement:
%
%We now state the main theorem of this paper:
%
%\begin{equation}
%\begin{array}{l}
%L(w(s)) := E(w(s)) - \frac{1}{s \sqrt{s}} \int_{-1}^{1} \partial_{s} w w \rho(y) \; dy
%\end{array}
%\label{Eqn:FuncL}
% \end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{thm}\label{Thm:BlowUpRate}{\rm (The blow-up rate for a $\log\log$ perturbation of the
  power nonlineairty).}
  Let $u$ be a solution of equation \eqref{equ}-\eqref{pert2} defined
  on some set $\q D$ \eqref{defD} for some $1$-Lipschitz function
  $x\longmapsto T(x)$.
  %(\ref{Eqn:WaveFoc})  with blow-up graph $\Gamma: \left\{ x
  %  \rightarrow T(x) \right\}$.
  Let $x_{0}$ be a non-characteristic
point. Then there exists  $t_{0}(x_{0}) \in [ 0, T(x_{0}) ) $ such that
for all $ t \in \left[ t_0(x_{0}), T(x_{0}) \right)$,
\begin{equation}
\begin{array}{rl}
k  \leq &\frac{1}{ \psi_{T(x_{0})}(t)}  \frac{\| u(t) \|_{L^{2} \left( B \left( x_{0}, T(x_{0}) - t \right) \right) } }
{ \left(  T(x_{0}) - t \right)^{\frac{N}{2}}}\\
&+ \frac{1} {\psi_{T(x_{0})}(t)}
\frac{ \| \nabla u(t) \|_{L^{2} \left( B(x_{0},T(x_{0})-t) \right)} +  \| \partial_{t} u(t) \|_{L^{2} \left( B \left( x_{0}, T(x_{0}) - t \right) \right)}} {\left(  T(x_{0}) - t \right)^{\frac{N}{2} -1}}
\leq  K,
\end{array}
\label{Eqn:BlowUpRate}
\end{equation}
for some $0< k(N,p,a)\le K$, where $K$ depends only on an upper bound
on $T(x_0)$, $1/T(x_0)$, the norm $\| \left( u
(t_0(x_0)), \partial_{t} u (t_0(x_0)) \right) \|_{H^{1} \times L^{2}  \left( B \left(
      x_{0},  \frac{T(x_{0})-t_0(x_0))}{\delta_{0}(x_{0})} \right)\right)}$
% $K=K  \left( \delta_{0}(x_{0}), T(x_0),
% \| u (0), \partial_{t} u (0) \|_{H^{1} \times L^{2}  \left( B \left(
%       x_{0},  \frac{T(x_{0})}{\delta_{0}(x_{0})} \right) \right) }
% \right) $
and $\delta_0(x_0)<1$, the slope of some non-characteristic
cone $\q C_{x_0,\delta_0(x_0)}\subset \q D$.

% $ K := K  \left( \delta_{0}(x_{0}),
% \| u (t(x_{0})), \partial_{t} u (t(x_{0})) \|_{H^{1} \times L^{2}  \left( B \left(  x_{0},  \frac{T(x_{0}) - t_{0}(x_{0})}{\delta_{0}(x_{0})} \right) \right) }
% \right) $ such that $T(x_{0}) - t_{0}(x_{0}) \ll 1 $; and for all $ t \in \left[ t(x_{0}), T(x_{0}) \right)$

% \begin{equation}
% \begin{array}{rl}
% k  \leq &\frac{1}{ \psi_{T(x_{0})}(t)}  \frac{\| u(t) \|_{L^{2} \left( B \left( x_{0}, T(x_{0}) - t \right) \right) } }
% { \left(  T(x_{0}) - t \right)^{\frac{N}{2}}}\\
% &+ \frac{1} {\psi_{T(x_{0})}(t)}
% \frac{ \| \nabla u(t) \|_{L^{2} \left( B(x_{0},T(x_{0})-t) \right)} +  \| \partial_{t} u(t) \|_{L^{2} \left( B \left( x_{0}, T(x_{0}) - t \right) \right)}} {\left(  T(x_{0}) - t \right)^{\frac{N}{2} -1}}
% \leq  K \cdot
% \end{array}
% \label{Eqn:BlowUpRate}
% \end{equation}
% Here $ \delta_{0}(x_{0}) $ is the number defined in (\ref{Eqn:DefNonCharac}).
\end{thm}
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{nb}
Unlike in the pure power case treated by Merle and Zaag in
\cite{MZajm03}, the lower bound is not a direct consequence of the
local in time wellposedness in the energy space combined with the
 invariance by scaling %(see page 1149 in that paper).
 In fact, due to the lack of that invariance, we had to introduce a new argument
 below in Section \ref{Sec:LowerBoundProof}, which makes a major novelty of
our paper. Our argument relies on a delicate
application of the local well-posedness theory for equation
\eqref{equ}-\eqref{pert2} together with a modification of the scaling argument.
We would like to mention that our method is valid not
only for the $\log \log$ nonlinearity. It extends in fact to other
types of nonlinearities, including the $\log$ nonlinearity
considered by Hamza and Zaag in \cite{HZna21}, where the lower bound
was given with no details of the proof.
\end{nb}

%\begin{nb}
% \textbf{COMMENTS: BEGINNING} \\
% \\
%We only prove the upper bound in (\ref{Eqn:BlowUpRate})
% \\
% Do $k$, $K$ depend also on $t_{0}(x_{0})$ and $T(x_{0})$? I think that the dependance on $N$ and $p$ is not necessary to be mentioned. \\
% \\
% \textbf{COMMENTS:END}\\
%\end{nb}
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
In order to find the upper bound of (\ref{Eqn:BlowUpRate}), we first find a bound on average on $s$-unit intervals $[s,s+1]$ of
$ \left\|  \left( w(\tau), \partial_{s} w(\tau \right)) \right\|^{2}_{H^{1}(B(O,1)) \times L^{2} (B(O,1))}$. More precisely, we prove the proposition below:
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{prop}[Bounds on time averages of $(w,\partial_sw)$ in
  $H^1\times L^2(B(O,1))$]\label{Prop:BoundPolynws}
  Let $u$ be a solution of \eqref{equ}-\eqref{pert2}  defined
  on some set $\q D$ \eqref{defD} for some $1$-Lipschitz function
  $x\longmapsto T(x)$.
%(\ref{Eqn:WaveFoc}) with blow-up graph $\Gamma: \left\{ x \rightarrow
%  T(x) \right\}$.
Then there  exists $b> 0$ such that for any $x_{0}$ non-characteristic point there exists $t_{0}(x_{0}) \in [ 0, T(x_{0}) ) $ such that
  % $K := K \left( \delta_{0}(x_{0}),
% \left\| \left( u(t_{0}(x_{0})), \partial_{t} u(t_{0}(x_{0})) \right) \right\|_{H^{1} \times L^{2} \left( B \left(  x_{0},  \frac{T(x_{0}) - t_{0}(x_{0})}{\delta_{0}(x_{0})} \right) \right)}
% \right) $, and   such that $T(x_{0}) - t_{0}(x_{0}) \ll 1 $; and
for all $ T_{0} \in \left[  \frac{ t_{0}(x_{0}) + T(x_{0}) }{2}, T(x_{0}) \right) $, $ x \in \mathbb{R}^N: | x - x_{0} | \leq
\frac{T_{0} - \frac{ t_{0}(x_{0}) + T(x_{0}) }{2}}{\delta(x_{0})} $, and  $ s \geq - \log \left( T^{*}(x) - t_{0}(x_{0}) \right) $  we have

\begin{equation}
\begin{array}{l}
\int_{s}^{s+1} \int_{B(O,1)} \left( ( \partial_{\tau} w )^{2} + \left| \nabla w \right|^{2} + w^{2} \right) \; dy d \tau
\leq K s \log^{1+b}(s) \;
\end{array}
\label{Eqn:EstKinetw}
\end{equation}
where $ w :=  w_{x,T^{*}(x)}$ is the function defined in
(\ref{Eqn:SimilVar}) with $ T^{*}(x) := T_{0} - \delta_{0}(x_{0})
|x-x_{0}| $, $K$ depends only on an upper bound
on $T(x_0)$, $1/T(x_0)$, the local norm \\
$\| u (t_0(x_0)), \partial_{t} u (t_0(x_0)) \|_{H^{1} \times L^{2}  \left( B \left(
      x_{0},  \frac{T(x_{0})-t_0(x_0)}{\delta_{0}(x_{0})} \right)\right)}$,
and $\delta_0(x_0)<1$, the slope of some non-characteristic
cone $\q C_{x_0,\delta_0(x_0)}\subset \q D$.
% and $ \delta_{0}(x_{0}) $  is the number defined in (\ref{Eqn:DefNonCharac}).
\end{prop}
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%

% COMMENTS
%\textbf{Remarque pour Tristan : Dans ta version du 15 d\'ecembre, tu %as
%  restreint $T_0$ \`a l'intervalle $\left[\frac{T_0+  t_{0}(x_{0})}2,
%    T(x_{0}) \right] $, et du coup tu as restreint $|x-x_0|$ \`a
%  \^etre inf\'erieur \`a $\frac{T_0-
%    \frac{T_0+t_0(x_0)}2}{\delta_0(x_0)}$. Je n'ai pas encore
%  v\'erifi\'e, mais j'ai l'impression qu'il y a moyen d'adopter la
%  formulation que j'ai prise dans cette version, \`a l'instar de ce
%  que j'ai fait avec Hamza. Je ne pense pas que le $\log \log$ change
%  quelque chose ici.....}

\medskip


We then prove pointwise bounds of $ \left\|  \left( w(s), \partial_{s} w(s) \right) \right\|^{2}_{H^{1}(B(O,1)) \times L^{2} (B(O,1))}$. More precisely, we prove
the following proposition:

%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{prop}[Bounds on $(w,\partial_sw)$ in $H^1\times L^2(B(O,1))$] \label{Prop:BoundFinalWs}
  Let $u$ be a solution of \eqref{equ}-\eqref{pert2}  defined
  on some set $\q D$ \eqref{defD} for some $1$-Lipschitz function
  $x\longmapsto T(x)$.  Let $x_{0}$ be a non-characteristic
  point. Then, there exists $t_{0}(x_{0}) \in [ 0, T(x_{0}) ) $ such
  that for all $T_{0} \in \left[ \frac{t_{0}(x_{0}) + T(x_{0})}{2},
  T(x_{0}) \right) $,  $ x \in \mathbb{R}^N: | x - x_{0} | \leq
\frac{T_{0} - \frac{ t_{0}(x_{0}) + T(x_{0}) }{2}}{\delta(x_{0})} $,
  and $  s \geq - \log
  \left( T^{*}(x) - t_{0}(x_{0})
  \right) $, we have

\begin{equation}
\begin{array}{l}
\| w(s) \|_{H^{1} \left( B(O,1) \right) }^{2} + \left\| \partial_{s} w
  (s) \right\|_{L^{2} \left( B(O,1) \right)}^{2} \le K,
\end{array}
\label{Eqn:BoundFinalWs}
\end{equation}
where $ w :=  w_{x, T^{*}(x)}$ is the function defined in
(\ref{Eqn:SimilVar}), $K$ depends only on an upper bound
on $T(x_0)$, $1/T(x_0)$, the local norm
$\| u(t_{0}), \partial_{t} u (t_{0}) \|_{H^{1} \times L^{2}  \left( B \left( x_{0},  \frac{T_{0} - t_{0}(x_{0})}{\delta_{0}(x_{0})} \right)\right)}$,
and $\delta_0(x_0)<1$, the slope of some non-characteristic
cone $\q C_{x_0,\delta_0(x_0)}\subset \q D$.


\medskip

% COMMENTS
% textbf{Remarque pour Tristan : Dans ta version du 15 d\'ecembre 2022,
% tu as \'enonc\'e ce r\'esultat pour $w_{x,T^*(x)}$ o\`u $x$ est
% proche de $x_0$, et pas juste pour $w_{x_0,T(x_0)}$. Est-ce %n\'ecessaire ?}

\medskip

%
% Let $u$ be a solution of (\ref{Eqn:WaveFoc}) with blow-up graph $\Gamma: \left\{ x \rightarrow T(x) \right\}$. Let $x_{0}$ be a non-characteristic
% point. Then there exist $t_{0}(x_{0}) \in [ 0, T(x_{0}) ) $ and $K := K \left( \delta_{0}(x_{0}),
% \left\| \left( u(t_{0}(x_{0})), \partial_{t} u(t_{0}(x_{0})) \right) \right\|_{H^{1} \times L^{2} \left( B \left(  x_{0}, \frac{T(x_{0}) -t_{0}(x_{0})}{\delta_{0}(x_{0})} \right) \right)}
% \right) $ such that $T(x_{0}) - t_{0}(x_{0}) \ll 1 $ and for all $  s \geq - \log \left( T(x_{0}) - t_{0}(x_{0}) \right) $ we have

% \begin{equation}
% \begin{array}{l}
% \| w(s) \|_{H^{1} \left( B(O,1) \right) }^{2} + \left\| \partial_{s} w (s) \right\|_{L^{2} \left( B(O,1) \right)}^{2} \lesssim 1 \cdot
% \end{array}
% \label{Eqn:BoundFinalWs}
% \end{equation}
% Here $ w :=  w_{x_{0}, T(x_{0})}$ is the function defined in (\ref{Eqn:SimilVar}).
\end{prop}
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
Proposition \ref{Prop:BoundFinalWs} yields the upper bound in Theorem
\ref{Thm:BlowUpRate} by getting back to the $x$ and $t$ variables from
the $y$ and $s$ variables thanks to the similarity variables'
definition \eqref{defw} and using the monotone convergence theorem.
%Since the lower bound in that theorem follows
%from the local well-posedness in $H^1\times L^2$, as we explained in
%the remark following the statement of the theorem, we won't prove %that statement, and focus only on the proofs of Propositions
%\ref{Prop:BoundPolynws} and \ref{Prop:BoundFinalWs} in our paper.

%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Notation}
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
We recall some notation, some of which was already used in the
introduction. If $X\in \m R$,  then $\langle X \rangle =
\left(1+X^2\right)^{\frac 12}$.\\
We write $X\ll Y$ if the value of $X$ is much smaller that that of $Y$, $X \gg Y$ if the value of $X$ is much larger than that of $Y$, and $X \approx Y $ if  $X \ll Y$ and $Y \ll X$ are not true. We write
$X = o(Y)$ if there exists a constant $0 < c \ll 1$ such that  $|X| \leq c |Y|$. We define
$X+ = X + \epsilon$ for $ 0 < \epsilon \ll 1$.\\
We also write $X\lesssim Y$ if $X\le CY$ for some $C>0$ that does not depend on
$X$ and $Y$. %depending only %on $N$, $p$ and $a$, the constants of the problem appearing in \eqref{equ} and \eqref{pert2}.
If $Y+$ appears in the mathematical expression
$X \leq C Y+$ then $ C $ may depend on $ \epsilon $ but not on $X$ and $Y$.
%we ignore the dependence of $C$ on $\epsilon$ in order to make our presentation simple.
The origin of $\m R^N$ is denoted by
$O$ (this is mostly used to
denote by $B(O,r)$, the ball centered at the origin with radius $r>0$).
Unless otherwise specified, we let in the sequel $f$ (resp. $u$) be a function depending on the space variable (resp.
the space variable and the time variable). Unless otherwise specified, for sake of simplicity, we do not mention the spaces to which $f$ and $u$
belong in the estimates: this exercise is left to the reader.
% \\
% \\
% \textbf{TO BE COMPLETED: BEGINNING} \\
% \\
% Other notation \\
% \\
% \textbf{TO BE COMPLETED: END} \\'


\section{Proof of the lower bound in (\ref{Eqn:BlowUpRate})}
\label{Sec:LowerBoundProof}

Let $u$ be a solution of \eqref{equ}-\eqref{pert2}
%(\ref{Eqn:WaveFoc})
with data $ \left( u(0),\partial_{t}u(0) \right) := (u_{0},u_{1}) \in H^{1}_{loc} \times L^{2}_{loc}(\mathbb{R}^{N})$ and with a $1-$ Lipschitz  graph. \\
\\
We claim that the ``$\leq $'' inequality of (\ref{Eqn:BlowUpRate})
holds. We may assume
without loss of generality
%WLOG
that $k$ is small enough so that all the statements below are true.
Let $ A (x):= \log^{- \frac{a}{p-1}} (- \log(x))$ if $x \ll  1 $ and  $A(x) := 1 $ if $x \gtrsim 1 $. \\
Assume that this is not the case. This means that there exists  $t_{1} \in [0, T(x_{0}))$ such that

\begin{equation}
\begin{array}{l}
\frac{\| u(t_{1}) \|_{L^{2} \left( B(x_{0}, T(x_{0}) - t_{1}) \right) } }{ \left(  T(x_{0}) - t_{1} \right)^{\frac{N}{2} - \frac{2}{p-1}}} +
\frac{ \| \nabla u(t_{1}) \|_{L^{2} \left( B(x_{0}, T(x_{0}) - t_{1}) \right)} +  \| \partial_{t} u(t_{1}) \|_{L^{2} \left( B(x_{0}, T(x_{0}) - t_{1}) \right)} } {\left(  T(x_{0}) - t_{1} \right)^{\frac{N}{2}- \frac{p+3}{p-1}}} \leq k A \left( T(x_{0})- t_{1} \right) \cdot
\end{array}
\nonumber
\end{equation}
Since $x_{0}$ is a non-characteristic point, we can use the dominated convergence theorem to infer that there exists $ 1 \gg \epsilon > 0$  such that

\begin{equation}
\begin{array}{l}
\frac{\| u(t_{1}) \|_{L^{2} \left( B \left( x_{0}, (T(x_{0}) - t_{1}) (1 + \epsilon) \right) \right) } }{ \left(  T(x_{0}) - t_{1} \right)^{\frac{N}{2} - \frac{2}{p-1}}} +
\frac{ \| \nabla u(t_{1}) \|_{L^{2} \left( B(x_{0}, T(x_{0}) - t_{1}) (1 + \epsilon) \right)} +
\| \partial_{t} u(t_{1}) \|_{L^{2} \left( B(x_{0}, T(x_{0}) - t_{1}) (1+ \epsilon) \right)} } {\left(  T(x_{0}) - t_{1} \right)^{\frac{N}{2}- \frac{p+3}{p-1}}} \leq 2 k A \left( T(x_{0})- t_{1} \right) \cdot
\end{array}
\nonumber
\end{equation}
%$u_{\lambda} (t,x) := \lambda^{\frac{2}{p-1}} u \left( \lambda t, \lambda \left( x + \frac{x_{0}}{\lambda}  \right)  \right)$.
A computation (using an appropriate change of variable) shows that for $\lambda = T(x_{0}) - t_{1}$, we get
\begin{equation}
\begin{array}{l}
\left\| \left(  f_{\lambda} ,  g_{\lambda} \right) \right\|_{H^{1}  \left( B(O,1 + \epsilon) \right) \times L^{2}  \left( B(O,1 + \epsilon) \right)} \lesssim k A (\lambda)
\end{array}
\label{Eqn:Estulambdalambdat1}
\end{equation}
with $ \left( f_{\lambda}(x), g_{\lambda}(x) \right) := \lambda^{\frac{2}{p-1}} \left( u \left(  t_{1}, \lambda x + x_{0}  \right),
 \lambda \partial_{t} u  \left( t_{1} , \lambda x + x_{0} \right) \right) $.
%Observe that $ v(t,x) := u_{\lambda} \left( \frac{t_{1}}{\lambda} +  t ,  x  \right)$
Let  $C^{'} \gg 1$ be a constant large enough such that all the statements below are true. Let $K:= \left\{ (t,x): t \in [0,1 + \epsilon), \, \text{and} \, t < 1 + \epsilon - |x| \right\}$.  We claim that there exists a unique solution \\
 $ v \in \mathcal{X} := \left\{ w: \; w \in \mathcal{C}_{t} H^{1} (K) \cap \mathcal{C}^{1}_{t} L_{x}^{2} (K), \; \text{and} \;
\| v \|_{\mathcal{X}} \leq C^{'} k A(\lambda)  \right\} $  such that  $ v = \tilde{\Psi}(v) $  with
$ \| v \|_{\mathcal{X}} := \sup \limits_{t \in [ 0,1 + \epsilon) } \sup \limits_{ 1 + \epsilon - t > R > 0}
\left( \| v(t) \|_{H^{1}(B(O,R))},  \| \partial_{t} v(t) \|_{L^{2}(B(O,R))} \right)  $ and

\begin{equation}
\begin{array}{l}
\tilde{\Psi}(w)(t) :=  \partial_{t} \mathcal{R}(t)* f_{\lambda} + \mathcal{R}(t) * g_{\lambda} + \int_{0}^{t} \mathcal{R}(t-s)*
h_{\lambda} \left( w(s) \right) \; ds \cdot
\end{array}
\nonumber
\end{equation}
Here $ h_{\lambda}(f) :=  |f|^{p-1} f \log^{a} \left( \log \left( 10 + \lambda^{-\frac{4}{p-1}} f^{2} \right) \right)  $.
%In the equal

%\begin{equation}
%\left\{
%\begin{array}{l}
%\partial_{tt} w - \triangle w = |w|^{p-1} w \log^{a} \left(  \log \left( 10 +  \lambda^{-\frac{4}{p-1}} w^{2} \right) \right) \\
%\left( w(0),\partial_{t} w(0) \right) := \lambda^{\frac{2}{p-1}} \left(  u \left(  t_{1}, \lambda x + x_{0}  \right), \lambda
%\partial_{t} u  \left( t_{1} , \lambda x + x_{0} \right) \right) \cdot
%\end{array}
%\right.
%\nonumber
%\end{equation}
Let $\left( w, w_{1}, w_{2} \right) \in \mathcal{X}^{3}$. In the sequel, we implicitly use elementary estimates involving the logarithm \footnote{such as $\log(xy) = \log(x) + \log(y)$;
$\log(x^{-1})= - \log(x) $;   $\log(x) \lesssim x^{0+}$, $\log(x) \lesssim x^{0-}$  for $x \gtrsim 1$, $x \lesssim 1$ respectively .} and the Sobolev embedding
$ H^{1} \left( B (O,R) \right)  \hookrightarrow L^{s} \left( B (O,R) \right)$ for $ 2 \leq s \leq (2p)+$. \\
We first estimate $ \| \tilde{\Psi}(w_{2}) - \tilde{\Psi}(w_{1}) \|_{\mathcal{X}} $. Assume that $\lambda \gtrsim 1$. We have
$ \| \tilde{\Psi}(w_{2}) - \tilde{\Psi}(w_{1}) \|_{\mathcal{X}} \lesssim   \left\| h_{\lambda}(w_{1}) - h_{\lambda}(w_{2}) \right\|_{L_{t}^{1} L_{x}^{2}(K)} $. Hence, the fundamental theorem of calculus and the Holder inequality show that
\begin{equation}
\begin{array}{ll}
\| \tilde{\Psi}(w_{2}) - \tilde{\Psi}(w_{1}) \|_{X} & \lesssim  (1 + \epsilon) \sum \limits_{\substack{\bar{w} \in \{ w_{1}, w_{2} \} \\ r \in \{  p-1, (p-1)+ \} }}
\| \bar{w} \|^{r}_{L_{t}^{\infty} L_{x}^{2r} (K)} \| w_{2} - w_{1} \|_{\mathcal{X}} \cdot
\end{array}
\nonumber
\end{equation}
Assume that $\lambda \ll 1$. Let $a > 0$. Then

\begin{equation}
\begin{array}{l}
\| \tilde{\Psi}(w_{2}) - \tilde{\Psi}(w_{1}) \|_{\mathcal{X}} \\
\lesssim (1 + \epsilon)
\left(
\sum \limits_{\substack{\bar{w} \in \{ w_{1}, w_{2} \} \\ r \in \{ p \pm , (p-1)\pm \}}}
\| \bar{w} \|^{r}_{L_{t}^{\infty} L_{x}^{2r}(K)} + \log^{a} \left(- \log ( \lambda) \right)
\sum \limits_{\bar{w} \in \{ w_{1},w_{2} \}}  \| \bar{w} \|^{p-1}_{L_{t}^{\infty} L_{x}^{2(p-1)}(K)}
\right) \| w_{2} - w_{1} \|_{\mathcal{X}} \cdot
\end{array}
\nonumber
\end{equation}
Let $a < 0 $. Let $\bar{w} \in \{ w_{1}, w_{2} \}$. Then $ \left\| h_{\lambda}(w_{1}) - h_{\lambda}(w_{2}) \right\|_{L_{t}^{1} L_{x}^{2}(K)} $ is bounded by terms of a finite sum of terms of the form $ X_{1}(\lambda) $ and $ X_{2}(\lambda) $ with

$ X_{1}(\lambda) := \left\| \log^{a} \left( \log \left( 10 + \frac{\bar{w}^{2}}{\lambda^{\frac{4}{p-1}}} \right) \right) |\bar{w}|^{p-1} \left( w_{2} - w_{1} \right)
\right\|_{L_{t}^{1} L_{x}^{2} ( K \cap \{ \bar{w}^{2} \geq  \lambda^{\frac{4}{p-1}-} \} )} $, and \\
$ X_{2}(\lambda) := \left\|  \log^{a} \left( \log \left( 10 + \frac{\bar{w}^{2}}{\lambda^{\frac{4}{p-1}}} \right) \right) |\bar{w}|^{p-1} \left( w_{2} - w_{1} \right) \right\|_{L_{t}^{1} L_{x}^{2} ( K  \cap \{ \bar{w}^{2} \leq \lambda^{\frac{4}{p-1}-}\}
)}$. We have

\begin{equation}
\begin{array}{ll}
X_{1}(\lambda)  &  \lesssim (1 + \epsilon) \log^{a} \left( -\log(\lambda) \right)
\| \bar{w} \|^{p-1}_{L_{t}^{\infty} L_{x}^{2(p-1)} (K) }
\| w_{2} - w_{1} \|_{\mathcal{X}} \cdot
\end{array}
\nonumber
\end{equation}
We have

\begin{equation}
\begin{array}{ll}
X_{2} (\lambda) & \lesssim (1 + \epsilon) \| \bar{w} \|^{p-1}_{L_{t}^{\infty} L_{x}^{\infty} ( K  \cap \{ \bar{w}^{2} \leq \lambda^{\frac{4}{p-1}-}  \} )}
\| w_{2} - w_{1} \|_{L_{t}^{\infty} L_{x}^{2}(K)} \\
& \lesssim (1 + \epsilon) \lambda^{0+} \| w_{2} - w_{1} \|_{\mathcal{X}}.
\end{array}
\nonumber
\end{equation}
Hence, the estimates above show that  $\tilde{\Psi}$ is a contraction. \\
\\
We then estimate $\left\| \tilde{\Psi}(w)  \right\|_{\mathcal{X}}$.  Assume that $\lambda \gtrsim 1$. Then  (\ref{Eqn:LocNrjEst}), (\ref{Eqn:Estulambdalambdat1}), and the H\"older inequality show that

\begin{equation}
\begin{array}{ll}
\left\| \tilde{\Psi}(w)  \right\|_{\mathcal{X}}
& \lesssim  k A(\lambda) + (1 + \epsilon) A(\lambda) \sum \limits_{ r \in \{ p ,p+ \} }
\| w \|^{r}_{L_{t}^{\infty} H^{1}(K)}.
\end{array}
\label{Eqn:EstTildePsi}
\end{equation}
Assume that $\lambda \ll 1$. Let $a > 0$. Then (\ref{Eqn:EstTildePsi}) holds. Let $a < 0$. Let  $X(w) :=\left\| |v|^{p-1} v \log^{a} \left(  \log \left( 10 +  \lambda^{-\frac{4}{p-1}} w^{2} \right) \right) \right\|_{L_{t}^{1} L_{x}^{2}(K)}$. Then $ X(w) \lesssim  X_{1}(w) + X_{2}(w) $  with

\begin{equation}
\begin{array}{l}
X_{1}(w) := \left\| |w|^{p-1} w \log^{a} \left(  \log \left( 10 +  \lambda^{-\frac{4}{p-1}} w^{2} \right) \right) \right\|_{L_{t}^{1} L_{x}^{2} (K \cap
\{ w^{2} \geq \lambda^{\frac{4}{p-1}-}  \} )}, \; \text{and} \\
X_{2}(w) := \left\| |w|^{p-1} w \log^{a}
\left(  \log \left( 10 +  \lambda^{- \frac{4}{p-1}} w^{2} \right) \right) \right\|_{L_{t}^{1} L_{x}^{2}(K
\cap \{ w^{2} \leq \lambda^{\frac{4}{p-1}-} \} )} \cdot
\end{array}
\nonumber
\end{equation}
We have $ X_{1}(w) \lesssim (1+ \epsilon) \log^{a} \left(  - \log (\lambda) \right)  \| w \|^{p}_{L_{t}^{\infty} H^{1}(K)} $. By interpolation there exists $\theta:= \theta(p) > 0 $ such that

\begin{equation}
\begin{array}{ll}
X_{2}(w) & \lesssim (1+ \epsilon) \| w \|^{p}_{L_{t}^{\infty} L_{x}^{2p} (K  \cap \{
w^{2}  \leq \lambda^{\frac{4}{p-1}-}  \}) } \\
& \lesssim (1+ \epsilon)
\| w \|^{p \theta}_{L_{t}^{\infty} L_{x}^{\infty}( K \cap \{ w^{2} \leq \lambda^{\frac{4}{p-1}-}  \} )}
\| w \|^{p (1- \theta)}_{L_{t}^{\infty} L_{x}^{2} (K) } \\
& \lesssim (1+ \epsilon) \lambda^{0+}  \| w \|^{p(1- \theta)}_{L_{t}^{\infty} H^{1} (K)}
\end{array}
\nonumber
\end{equation}
Hence from the estimates above  we see that $\tilde{\Psi}(\mathcal{X}) \subset \mathcal{X} $. \\
\\
Hence by the fixed point theorem we see that there exists $v \in \mathcal{X}$ such that $ v = \tilde{\Psi}(v) $.\\
\\
Let \\
$\Omega := \left\{ (t,x): \, 0 \leq  t  <
T(x_{0}) + \epsilon \left( T(x_{0}) - t_{1} \right),  \;
|x - x_{0}| < T(x_{0}) + \epsilon \left( T(x_{0}) -t_{1} \right) - t \right\}$. Let $w$ be defined by
$ w(t) := u(t)$ if $ t \in [0, t_{1}]$ and $ w (t) := \frac{1}{\lambda^{\frac{2}{p-1}}} v \left( \frac{t- t_{1}}{\lambda}, \frac{x-x_{0}}{\lambda} \right) $ if $t \in [ t_{1}, T(x_{0})  + \epsilon (T(x_{0}) - t_{1})  ) $. By the
finite speed of propagation we infer that $ w(t) = u(t)$ on $ \Omega \cap ( t_{1} - \alpha, t_{1} + \alpha) $ for $\alpha > 0 $ small.  Hence $ w \in \mathcal{C}_{t} H^{1} (\Omega) \cap \mathcal{C}_{t}^{1} L_{x}^{2} (\Omega)$ and it satisfies $\Psi(w)=w$, with
$\Psi$ defined in (\ref{Eqn:DefPsi}) . This contradicts the definition of $T(x_{0})$.


%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Design of a Lyapunov functional in similarity variables}
%\section{Proof of Proposition \ref{Prop:BoundPolynws} }
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
This section is at the heart of our argument. We design here a
Lyapunov functional for the similarity variables' PDE
\eqref{Eqn:Wavew}, based on the functional $\q E(w(s))$ \eqref{Eqn:Energyw}, and prove its nonnegativity, thanks to a blow-up
criterion. This way, the Lyapunov functional will turn to be bounded
from above and below. Using some intricate interpolation arguments, we
prove the boundedness of the solution in $H^1\times L^2(B(O,1)$ in
time averages, which yields Proposition
\ref{Prop:BoundPolynws}. If we follow the spirit of earlier papers on
the subject, the $\log \log$ factor in the nonlinearity \eqref{pert2}
makes our argument very delicate.

\medskip

We proceed in 4 subsections, first bounding the derivative of $\q
E(w(s))$ \eqref{Eqn:Energyw}, then introducing a corrective
term. Combining it with $\q E(w(s))$, we define a Lyapunov functional
in the third subsection and prove its nonnegativity. Finally, the last subsection is devoted to
the proof of  Proposition \ref{Prop:BoundPolynws}. \\
%In this section we prove Proposition \ref{Prop:BoundPolynws}.

%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Estimate of derivative of the energy}
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%

The proposition below yields an estimate of the derivative of $\mathcal{E}(w)$:
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{lem}[Upper bound on the derivative of  $\q E(w(s))$
  \eqref{Eqn:Energyw}]
  Consider $w$ a solution of equation \eqref{Eqn:Wavew}.
Let $s \gg 1$. Then there exists a large positive constant $C$ such that

\begin{equation}
\begin{array}{l}
\frac{d \mathcal{E}(w)}{ds} \leq - \frac{3 \alpha}{2} \int_{B(O,1)} (\partial_{s} w)^{2} \frac{1}{1 - |y|^{2}} \; \rho dy +
\frac{C}{s \log^{a+1}(s)} \int_{B(O,1)} |w|^{p+1} g(\phi w) \; \rho
  dy + \Sigma,
  % \text{with}
\end{array}
\label{Eqn:EstDerivE}
\end{equation}
where $\phi$ is defined in \eqref{Eqn:Defphi} and
%with
\begin{equation}
\begin{array}{l}
\Sigma \leq \frac{C}{s^{2}}  \int_{B(O,1)} |\nabla w|^{2} (1 - |y|^{2}) \; \rho  dy
+ \frac{C}{s^{2}} \int_{B(O,1)} w^{2} \; \rho  dy + C e^{-s} \cdot
\end{array}
\nonumber
\end{equation}

\label{Lem:EstDerivE}
\end{lem}
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{proof}

We multiply (\ref{Eqn:Wavew})  by  $ \partial_{s} w  \rho $ and we integrate over $B(O,1)$. We get

\begin{equation}
\begin{array}{l}
\Sigma_{l} = \Sigma_{r,1} + ... + \Sigma_{r,8} \nonumber
\end{array}
\nonumber
\end{equation}
Here  $ \Sigma_{l} :=  \int_{B(O,1)} \partial_{s}^{2} w \rho \partial_{s} w \; dy $, $ \Sigma_{r,2} := \int_{B(O,1)}
\frac{1}{\rho} \dive \left( \rho \nabla w - \rho (y \cdot \nabla w) y \right) \rho \partial_{s} w \; dy $, \\
$ \Sigma_{r,3}  := \frac{2a}{(p-1)s \log(s)} \int_{B(O,1)}  (y \cdot \nabla w) \rho \partial_{s} w \; dy $,
$ \Sigma_{r,4} := - \frac{2(p+1)}{(p-1)^{2}} \int_{B(O,1)} w \partial_{s} w  \rho \; dy $, \\
$ \Sigma_{r,5} := \gamma(s) \int_{B(O,1)} w \partial_{s} w \; \rho \; dy $,  $\Sigma_{r,6} := - \left( \frac{p+3}{p-1} - \frac{2a}{(p-1)s \log(s)} \right) \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy $, \\
$\Sigma_{r,7} : = - 2 \int_{B(O,1)} y \cdot \nabla \partial_{s} w \rho \; dy $ and\\  $\Sigma_{r,8} := - e^{-\frac{2sp}{p-1}} \log^{\frac{a}{p-1}}(s) \int_{B(O,1)} f(\phi w)  \partial_{s} w  \rho \; dy  $. \\
Integrating by parts whenever it is necessary, and using the inequality $ab \leq \frac{a^{2}}{2} + \frac{b^{2}}{2}$, we get
\begin{equation}
\begin{array}{l}
\Sigma_{r,1} := \frac{d}{ds} \left(  \frac{1}{2} \int_{B(O,1)}
  (\partial_{s} w)^{2} \rho \; dy \right), \\
  \Sigma_{r,2} = - \frac{d}{ds} \left( \frac{1}{2} \int_{B(O,1)} \left( |\nabla w|^{2}  - |\nabla w \cdot y|^{2} \right) \rho \; dy \right), \\
\\
\left| \Sigma_{r,3} \right| \lesssim  \frac{1}{s^{2}} \int_{B(O,1)} |\nabla w|^{2} (1 - |y|^{2}) \rho \; dy + \frac{1}{\log^{2}(s)} \int_{B(O,1)} (\partial_{s} w )^{2} \; \frac{\rho}{1-|y|^{2}} \; dy, \\
\\
\Sigma_{r,4}  = - \frac{d}{ds} \left(  \int_{B(O,1)} \frac{p+1}{(p-1)^{2}} w^{2} \rho \; dy \right), \\
\\
\left| \Sigma_{r,5}(s) \right| \lesssim \frac{1}{s^{2}} \int_{B(O,1)} w^{2} \rho \; dy
+ \frac{1}{\log^{2}(s)} \int_{B(O,1)} (\partial_{s} w )^{2} \; \frac{\rho}{1-|y|^{2}} \; dy, \; \text{and} \\
\\
\Sigma_{r,6}(s) = - \frac{p+3}{p-1} \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy +
O \left( \frac{1}{s \log(s)} \right) \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy  \cdot
\end{array}
\nonumber
\end{equation}
%\textbf{Question pour Tristan : pour la ligne avec $\Sigma_{r,5}(s)$,
%  est-ce un $\frac 1{s^2}$ ou bien $\frac 1{s\log s}$ ? Je ne sais
%  dans quelle mesure ce taux est important ou pas (je pense \`a l'integrabilit\'e).} \\
We then estimate $ \Sigma_{r,7}(s) $. Observe from (\ref{Eqn:ValAlpha}) that  $ 2 \alpha + N = \frac{p+3}{p-1} $. Hence

\begin{equation}
\begin{array}{ll}
\Sigma_{r,7} & = - \int_{B(O,1)} y \cdot \nabla \left( (\partial_{s} w)^{2} \right) \rho \; dy \\
& = \frac{p+3}{p-1} \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy - 2 \alpha  \int_{B(O,1)} \frac{(\partial_{s} w)^{2}}{1- |y|^{2}} \rho \; dy
\end{array}
\nonumber
\end{equation}
We then estimate $\Sigma_{r,8} $. We have $\Sigma_{r,8} = I + J $ with

\begin{equation}
\begin{array}{l}
I := - \int_{B(O,1)} e^{- \frac{2sp}{p-1}} \log^{\frac{a}{p-1}}(s) \phi^{-1}  \frac{d}{ds} \left( F(\phi w) \right) \rho \; dy, \; \text{and} \\
J :=  \int_{B(O,1)} e^{- \frac{2sp}{p-1}} \log^{\frac{a}{p-1}}(s)  F^{'}(\phi w) \phi^{-1} \phi^{'} w  \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
By integrating by parts, we see that
\begin{equation}
\begin{array}{ll}
I & =  - \frac{d}{ds} \left( \int_{B(O,1)} e^{-\frac{2(p+1)s}{p-1}} \log^{\frac{2a}{p-1}}(s) F(\phi w) \rho \; dy  \right)\\
&- \frac{2(p+1)}{p-1} e^{- \frac{2(p+1)s}{p-1}}  \log^{\frac{2a}{p-1}}(s) \int_{B(O,1)} F(\phi w)  \rho) \; dy \\
& + \frac{2a}{p-1}  e^{- \frac{2(p+1)s}{p-1}} \frac{ \log^{\frac{2a}{p-1} - 1}(s)}{s}  \int_{B(O,1)} F(\phi w) \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
We also see from (\ref{Eqn:Defphi}) that

\begin{equation}
\begin{array}{rl}
J =& \frac{2(p+1)}{p-1} e^{- \frac{2(p+1)s}{p-1}} \log^{\frac{2a}{p-1}}(s) \int_{B(O,1)} \frac{\phi w f(\phi w)}{p+1} \rho \; dy\\
&- \frac{2a}{(p-1)s} e^{- \frac{2(p+1)s}{p-1}} \log^{\frac{2a}{p-1}}(s)  \int_{B(O,1)} \frac{\phi w f(\phi w)}{2} \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
Hence,
\begin{equation}
\begin{array}{l}
\Sigma_{r,8} = \frac{d}{ds} \left( \int_{B(O,1)} e^{-\frac{2(p+1)s}{p-1}} \log^{\frac{2a}{p-1}}(s) F(\phi w) \rho \; dy  \right) + X + Y, \; \text{with}
\end{array}
\nonumber
\end{equation}

\begin{equation}
\begin{array}{l}
X :=  \frac{2(p+1)}{p-1} e^{- \frac{2(p+1) s}{p-1}} \log^{\frac{2a}{p-1}}(s) \int_{B(O,1)} \left( F(\phi w) - \frac{\phi w f(\phi w) }{p+1} \right) \rho \; dy \; \text{and} \\
Y := - \frac{2a}{p-1} e^{- \frac{2(p+1)s}{p-1}} \frac{\log^{\frac{2a}{p-1} - 1}(s)}{s} \int_{B(O,1)}  \left( F(\phi w) - \frac{\phi w f(\phi w) }{2} \right) \rho \; dy \cdot
\end{array}
\label{Eqn:DefXY}
\end{equation}
It remains to estimate $X$ and $Y$.  To this end we estimate $ F(\phi w) - \frac{\phi w  f(\phi w)}{p+1} $  by using
(\ref{Eqn:Defphi}), (\ref{Eqn:DefF1}), (\ref{Eqn:DecompF}), (\ref{Eqn:ApproxPot}), $|F(x)| \lesssim |x|^{p+1} \left( 1 + g(x) \right)$, and elementary estimates. Since
$ \{ \phi w^{2} \gg 1 \}  \subset \{ \phi^{2} w^{2} \gtrsim \phi \} $ and  $ \{ \phi w^{2}  \lesssim 1 \}  \subset \{ \phi w  \lesssim \phi^{\frac{1}{2}} \} $ we get

\begin{equation}
\begin{array}{l}
\phi w^{2} \gg 1: \;  \left| F(\phi w) - \frac{\phi w f(\phi w)}{p+1} \right| \lesssim \frac{  \left| \phi w  f(\phi w ) \right|}{ s \log(s) }  \\
\phi w^{2} \lesssim 1: \; \left| F(\phi w) - \frac{\phi w f(\phi w)}{p+1} \right| \lesssim \phi^{\frac{p+1}{2}} \left(  1 + g(x) \right) \cdot
\end{array}
\label{Eqn:ErrorFPhi} % Eq 1
\end{equation}
Hence

\begin{equation}
\begin{array}{l}
X \lesssim \frac{1}{s \log^{a+1}(s)}  \int_{B(O,1)} |w|^{p+1} g(\phi w) \; dy + e^{-s}  \cdot
\end{array}
\nonumber
\end{equation}
We also see from (\ref{Eqn:ApproxPot}) and the triangle inequality that

\begin{equation}
\begin{array}{l}
Y \lesssim \frac{1}{s \log^{a+1}(s)}  \int_{B(O,1)} |w|^{p+1} g(\phi w) \; dy  \cdot
\end{array}
\nonumber
\end{equation}
Hence, collecting all the estimates above we see that
(\ref{Eqn:EstDerivE}) holds. This concludes the proof of Lemma \ref{Lem:EstDerivE}.

% COMMENTS Eq 1
% Observe that $\frac{p+1}{p-1} \geq 2 $ if $ p < 3 $

% COMMENTS: PREV VERS Eq 1 (WRONG)

%\begin{equation}
%\begin{array}{l}
%F(\phi w) - \frac{\phi w  f(\phi w)}{p+1} \lesssim 1 +  \frac{\phi w f(\phi w)}{s \log(s)} :
%\end{array}
%\label{Eqn:DiffFComp}
%\end{equation}
%indeed, if $ \phi(s) w(y,s) \ll 1 $ then this is clearly bounded by a constant from the continuity of $F_{2}$;
%if $ \phi(s) w(y,s) \gtrsim 1 $ and $ \phi(s) w^{2}(y,s) \lesssim 1 $ then this is clear; and if $ \phi(s) w(y,s) \gtrsim 1 $  and $ \phi(s) w^{2}(y,s) \gg 1 $ then this follows %from $\phi^{2}(s) w^{2}(y,s) \gtrsim \phi(s)$.


\end{proof}

%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%
\subsection{Corrective term}
%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%
In this subsection, we introduce a corrective term which will be combined to the
functional  $\q E(w(s))$ \eqref{Eqn:Energyw} in the next subsection to
provide us with a Lyapunov functional for equation
\eqref{Eqn:Wavew}. More precisely, introducing
% Let $u$ be a  solution of  with blow-up graph $\Gamma : \{ x \rightarrow T(x) \}$. Let $\mathcal{J}(w)$ denote the map below
\begin{equation}
\begin{array}{l}
\mathcal{J}(w) : s \rightarrow \mathcal{J}(w(s)) := - \frac{1}{s \log(s)} \int_{B(O,1)} w(s) \partial_{s} w(s)  \rho \; dy,
\end{array}
\label{Eqn:DefJ}
\end{equation}
we have the following statement:
%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%
\begin{lem}[Upper bound on the derivative  of the corrective term]
Let $s \gg 1$. There exists a positive constant $C > 0$ such that

\begin{equation}
\begin{array}{ll}
\frac{d \mathcal{J}(w)}{ds} & \leq \frac{p+3}{2 s \log(s)} \mathcal{E}(w)  - \frac{p+7}{4 s \log(s)} \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy \\
& - \frac{p+1}{2(p-1) s \log(s)} \int_{B(O,1)} w^{2} \rho \; dy  - \frac{p-1}{4s \log(s)} \int_{B(O,1)}  \left( |\nabla w|^{2} - (y \cdot \nabla w)^{2} \right) \rho \; dy \\
& - \frac{p-1} {2(p+1) s \log^{a+1}(s)} \int_{B(O,1)} |w|^{p+1} g(\phi w) \rho \; dy  + \Sigma, \; \text{with}
\end{array}
\label{Eqn:DerivJ}
\end{equation}

\begin{equation}
\begin{array}{l}
\Sigma \leq \frac{C}{s^{\frac{1}{2}}} \int_{B(O,1)} \frac{(\partial_{s} w)^{2}}{1-|y|^{2}} \rho \; dy
+ \frac{C}{s^{\frac{3}{2}} \log^{2}(s)} \int_{B(O,1)} w^{2} \rho \; dy \\
+ C \frac{1}{s^{2} (\log(s))^{a+2}} \int_{B(O,1)} |w|^{p+1} g(\phi w) \rho \; dy + C  e^{-s},
\end{array}
\nonumber
\end{equation}
where $\phi$ is introduced in \eqref{Eqn:Defphi}.
\label{Lem:EstDerivJ}
\end{lem}
%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%

\begin{proof}

We have  $ \frac{d \mathcal{J}(w)}{ds}=  - \frac{1}{s \log(s)} \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy
+ \Sigma_{1} + \Sigma_{2} $ with

\begin{equation}
\begin{array}{l}
\Sigma_{1} := - \frac{1}{s \log(s) }  \int_{B(O,1)} w \partial_{s}^{2} w \rho \; dy, \; \text{and} \\
\Sigma_{2} := \frac{1}{s^{2} \log(s)} \left( 1 + \frac{1}{\log(s)} \right) \int_{B(O,1)} w \partial_{s} w \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
We estimate $\Sigma_{2}$. The inequality $ab \leq \frac{a^{2}}{2} + \frac{b^{2}}{2}$ shows that

\begin{equation}
\begin{array}{l}
\Sigma_{2} \leq \frac{1}{2s^{2} \log(s)}  \int_{B(O,1)} w^{2} \rho \; dy + \frac{1}{2s^{2} \log(s)} \int_{B(0,1)} (\partial_{s} w)^{2} \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
We now estimate $\Sigma_{1}$. We see from (\ref{Eqn:Wavew}) that

\begin{equation}
\Sigma_{1} = I_{1}+...+ I_{6} ,
\end{equation}
with $I_{1} := -\frac{1}{s \log(s)} \int_{B(O,1)} \frac{1}{\rho} \dive \left( \rho \nabla w  - \rho (y \cdot \nabla w ) y \right) w \rho \; dy  $, \\
$ I_{2} := - \frac{2a}{(p-1)s^{2} \log^{2}(s)} \int_{B(O,1)}  y \cdot \nabla w w  \rho \; dy $,
$ I_{3} := \frac{1}{s \log(s)} \left( \frac{2(p+1)}{(p-1)^{2}}  - \gamma \right) \int_{B(O,1)}  w^{2}   \rho \; dy  $, \\
$ I_{4} :=  \frac{1}{s \log(s)} \left( \frac{p+3}{p-1} - \frac{2a}{(p-1)s \log(s)} \right) \int_{B(O,1)}  \partial_{s} w  w \rho \; dy $, \\
$ I_{5}  := \frac{2}{s \log(s)} \int_{B(O,1)} y \cdot \nabla \partial_{s} w  w \rho \; dy  $,  and\\
$ I_{6} := - \frac{1}{s \log(s)} e^{-\frac{2ps}{p-1}} \log^{\frac{a}{p-1}}(s) \int_{B(O,1)} f(\phi w) w \rho \; dy $. \\
An integration by parts shows that

\begin{equation}
\begin{array}{l}
I_{1} = \frac{1}{s \log(s)} \int_{B(O,1)} \left( |\nabla w|^{2} - |y \cdot \nabla w|^{2} \right) \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
Writing  $w\nabla w = \nabla \left( \frac{w^{2}}{2}\right) $ we have

\begin{equation}
\begin{array}{l}
I_{2} = \frac{a}{(p-1)s^{2} \log^{2}(s)} \left(  N \int_{B(O,1)} w^{2} \rho \; dy - \int_{B(O,1)} \frac{2 |y|^{2} \alpha}{1- |y|^{2}} w^{2} \rho \; dy \right)
\end{array}
\nonumber
\end{equation}
Hence $  | I_{2} | \lesssim  \frac{1}{s^{2}\log^{2}(s)} \int_{B(O,1)}
w^{2} \rho \; dy $. \\
Clearly, $|I_{3}| \leq \left( \frac{2(p+1)}{(p-1)^{2} s\log(s)} + O \left( \frac{1}{s^{2} \log^{2}(s)} \right) \right) \int_{B(O,1)} w^{2} \rho \; dy $. The inequality
$AB \leq \frac{A^{2}}{2} + \frac{B^{2}}{2} $
%$ab \leq \frac{a^{2}}{2} + \frac{b^{2}}{2} $
shows that

\begin{equation}
\begin{array}{l}
I_{4} \lesssim \frac{1}{s^{\frac{1}{2}}} \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy
+ \frac{1}{s^{\frac{3}{2}} \log^{2}(s)} \int_{B(O,1)} w^{2} \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
From $\nabla \partial_{s} w  w = \frac{1}{2} \nabla \left( \partial_{s} (w^{2}) \right) $  we have

\begin{equation}
\begin{array}{ll}
I_{5} & = - \frac{2N}{s \log(s)} \int_{B(O,1)} w \partial_{s} w \rho \; dy  + \frac{2 \alpha}{s \log(s)} \int_{B(O,1)} w \partial_{s} w \frac{|y|^{2}}{1- |y|^{2}}
\rho \; dy \\
& = X + Y
\end{array}
\nonumber
\end{equation}
We have

\begin{equation}
\begin{array}{l}
X \lesssim \frac{1}{s^{\frac{3}{2}} \log^{2}(s)} \int_{B(O,1)} w^{2} \rho \; dy + \frac{1}{s^{\frac{1}{2}}} \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
In order to estimate $Y$ we use the Hardy-type inequality  (see e.g \cite{MZajm03})
\begin{equation}
\begin{array}{l}
\int_{B(O,1)} w^{2} \frac{|y|^{2}}{1- |y|^{2}} \rho \; dy \lesssim \int_{B(O,1)} |\nabla w|^{2} (1-|y|^{2}) \rho \; dy
+ \int_{B(O,1)} w^{2} \rho \; dy
\end{array}
\label{Eqn:HardyType}
\end{equation}
to get

\begin{equation}
\begin{array}{l}
Y \lesssim \frac{1}{s^{\frac{3}{2}} \log^{2}(s)} \int_{B(O,1)} w^{2} \rho \; dy + \frac{1}{s^{\frac{1}{2}}} \int_{B(O,1)}
\frac{(\partial_{s} w)^{2}}{1- |y|^{2}} \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
We turn now our attention to $I_{6}$. We get from (\ref{Eqn:Defphi})

\begin{equation}
\begin{array}{ll}
I_{6} & =  - \frac{1}{s \log^{a+1}(s)} \int_{B(O,1)} |w|^{p+1} g(\phi w) \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}
Hence we see from the estimates above and from (\ref{Eqn:ErrorFPhi})
that (\ref{Eqn:DerivJ}) holds. This concludes the proof of Lemma \ref{Lem:EstDerivJ}.

\end{proof}

%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%
\subsection{A new functional: decreasing and nonnegativity properties}
%\subsection{Functional $\mathcal{N}$: decreasing property and non-negativity property }
%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%
Using the previous subsections,
% Let $u$ be a  solution of equation \eqref{equ}-\eqref{pert2} defined
% on some domain $\q D$ for some $1$-Lipschitz function $x\mapsto T(x)$.
% (\ref{Eqn:WaveFoc}) with blow-up graph $\Gamma : \{ x \rightarrow
% T(x) \}$.
we introduce the following maps $\mathcal{N}_{m}(w)$
and $\mathcal{H}_{m}(w)$ (with $m \in \mathbb{R}^{+}$):
\begin{equation}
\begin{array}{ll}
\mathcal{N}_{m}(w): s \rightarrow \mathcal{N}_{m}(w(s)):=
  (\log(s))^{-\frac{m(p+3)}{2}} \mathcal{H}_{m}(w(s)) + m^{2} e^{-s},\\
%  \; \text{and} \\
\mathcal{H}_{m}(w) : s \rightarrow \mathcal{E}(w(s)) + m \mathcal{J}(w(s)) \cdot
\end{array}
\label{Eqn:DefHN}
\end{equation}
The lemma below shows that $ s \rightarrow \mathcal{N}_{m} (w(s))$ has a decreasing property (for a large $m$):
%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%
\begin{lem}[A nonnegative Lyapunov functional for equation
  \eqref{Eqn:Wavew}]\label{lemlyap}

There exist $m \gg 1$, $s_{0} \gg 1$, and $C > 0$ such that for all $s \geq s_{0} $ we have\\
\\
$(i)$:

\begin{equation}
\begin{array}{ll}
\mathcal{N}_{m}(w(s+1)) &- \mathcal{N}_{m}(w(s)) \leq - \frac{C}{s \log^{1+b}(s) } \int_{s}^{s+1} \int_{B(O,1)} \frac{(\partial_{\tau} w)^{2}}{1-|y|^{2}} \rho dy d \tau \\
& - \frac{C}{s \log^{1+b}(s)} \int_{s}^{s+1} \int_{B(O,1)} \left( |\nabla w|^{2} - |\nabla w \cdot y|^{2} \right) \rho  dy d \tau \\
& - \frac{C}{s \log^{a + 1 + b }(s)} \int_{s}^{s+1} \int_{B(O,1)} |w|^{p+1} \log^{a} \left( \log(10 + \phi^{2} w^{2}) \right) \rho  dy d \tau   \\
& - \frac{C}{\log^{1+b}(s)} \int_{s}^{s+1} \int_{B(O,1)} w^{2} \rho  dy, \; \text{where}
\end{array}
\label{Eqn:EstDecreaseAvN}
\end{equation}
$b := \frac{m(p+3)}{2}$. \\

$(ii)$:

\begin{equation}
\begin{array}{l}
\mathcal{N}_{m}(w(s)) \geq 0
\end{array}
\label{Eqn:EstPosN}
\end{equation}

\label{Lem:DecreaseN}
\end{lem}
%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%

\begin{proof}

(i) Let $s_{0} \gg 1$ be constant large enough such that all the
statements below are true. A computation shows that
\[
   \frac{d \mathcal{N}_{m}(w)}{ds} = (\log(s))^{-\frac{m(p+3)}{2}}
\left( \frac{d \mathcal{H}_{m}(w)}{ds} - \frac{m(p+3)}{2 s \log(s)}
  \mathcal{H}_{m}(w) \right) - m^{2} e^{-s}.
\]
Hence, by choosing $ m \gg 1$ large enough we
see from (\ref{Eqn:EstDerivE}), (\ref{Eqn:DerivJ}), and the estimate

\begin{equation}
\begin{array}{l}
\left| \mathcal{J}(w(s)) \right| \leq \frac{1}{2s \log(s)}
\left( \int_{B(O,1)} w^{2} \rho \; dy + \int_{B(O,1)} (\partial_{s} w)^{2} \rho \; dy \right)
\end{array}
\label{Eqn:EstJw}
\end{equation}
derived from $a b \leq \frac{a^{2}}{2} + \frac{b^{2}}{2}$ that there exists $ C > 0 $ such that

\begin{equation}
\begin{array}{ll}
\frac{d \mathcal{N}_{m}(w)}{ds} & \leq  - \frac{C}{\log^{b}(s)} \int_{B(0,1)} (\partial_{s} w)^{2} \; \rho  dy\\
&- \frac{C}{s \log^{1 + b }(s)} \int_{B(O,1)} \left( |\nabla w|^{2} - |\nabla w \cdot y|^{2} \right) \; \rho  dy \\
& - \frac{C}{s \log^{1+b}(s)} \int_{B(O,1)} w^{2} \rho \; dy - \frac{C}{s \log^{a+1 + b}(s)} \int_{B(O,1)} |w|^{p+1} g(\phi w) \; \rho  dy \cdot
\end{array}
\label{Eqn:EstDerivN}
\end{equation}
Hence integrating (\ref{Eqn:EstDerivN}) from $s$ to $s+1$ we get (\ref{Eqn:EstDecreaseAvN}). \\
(ii) In order to prove the second statement we argue by contradiction. Assume that there exists $S_{1} \geq s_{0} $ such that
$\mathcal{N}_{m} (w(S_{1})) < 0$. Let $ \delta > 0 $  be a constant so small that all the estimates below are true.
Let  $\tilde{w}^{\delta}(y,s) := w_{x, T^{*}(x) - \delta}(y,s)$. By (\ref{Eqn:SimilVar}) we have

\begin{equation}
\begin{array}{ll}
\tilde{w}^{\delta}(y,s) & := \frac{\phi \left(-\log(\delta + e^{-s}) \right) }{\phi(s)} w \left( \frac{y}{1 + \delta e^{s}}, -\log \left( \delta + e^{-s} \right) \right) \cdot
\end{array}
\label{Eqn:Reltildewandw}
\end{equation}
Here $ \phi(t) := \phi_{T^{*}(x) - \delta}(t)$.\\
Observe that $\tilde{w}^{\delta} $ is defined on $[S_{1} + 1, \infty)$, since
$- \log \left( \delta + e^{-( 1+S_{1} )} \right) \geq S_{1}$. Observe that

\begin{equation}
\begin{array}{l}
\mathcal{N}_{m} \left( \tilde{w}^{\delta}(1+ S_{1}) \right) < 0:
\end{array}
\label{Eqn:AssNNeg}
\end{equation}
this follows from the continuity of the map $ \delta^{'} \rightarrow \mathcal{N}_{m} \left( \tilde{w}^{\delta^{'}}(1+ S_{1}) \right)$ for $\delta^{'}$ positive and small enough. Since $ s \rightarrow N_{m} \left( \tilde{w}^{\delta}(s) \right) $ is decreasing for $s \geq S_{1} + 1$ we get

\begin{equation}
\begin{array}{l}
\lim \limits_{s \rightarrow \infty} \mathcal{N}_{m} \left( \tilde{w}^{\delta}(s) \right) \leq \mathcal{N}_{m} \left( \tilde{w}^{\delta}(1 + S_{1}) \right) \cdot
\end{array}
\nonumber
\end{equation}
We also see from the estimate $ab \leq \frac{a^{2}}{2} + \frac{b^{2}}{2}$ that

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} \tilde{w}^{\delta} \partial_{s} \tilde{w}^{\delta} \; \rho  dy \lesssim  \int_{B(O,1)} ( \tilde{w}^{\delta})^{2} \; \rho  dy
+ \int_{B(O,1)} \left( \partial_{s} \tilde{w}^{\delta} \right)^{2} \; \rho dy \cdot
\end{array}
\label{Eqn:ContPerturbTildew}
\end{equation}
Hence

\begin{equation}
\begin{array}{ll}
\mathcal{N}_{m}  \left( \tilde{w}^{\delta}(s) \right) & \geq
- e^{-\frac{2(p+1)}{p-1} s} (\log(s))^{\frac{2a}{p-1}- b} \int_{B(O,1)} F(\phi \tilde{w}^{\delta}) \; \rho  dy
\end{array}
\nonumber
\end{equation}
We see from the boundedness of $F(\phi \tilde{w}^{\delta})$ on $ \{ \phi \tilde{w}^{\delta} \lesssim  1 \}$  and
$F(\phi \tilde{w}^{\delta}) \lesssim  | \phi \tilde{w}^{\delta}|^{(p+1)+} $ derived from (\ref{Eqn:ApproxPot}) on
$ \{ \phi \tilde{w}^{\delta} \gg 1 \} $, and (\ref{Eqn:Reltildewandw}), that there exists a large constant $C > 0$ such that

\begin{equation}
\begin{array}{ll}
\mathcal{N}_{m} \left( \tilde{w}^{\delta}(s) \right) & \geq - C \left(  1 + \delta e^{s} \right)^{N} e^{-\frac{2(p+1)s}{p-1}} (\log(s))^{\frac{2a}{p-1} - b}
\phi^{(p+1)+} \left( \log( \delta + e^{-s} ) \right) \times  \\
& \int_{B(O,1)} \left| w \left( z, -\log(\delta + e^{-s}) \right) \right|^{(p+1)+} \;  dy
 - C e^{-2s} \cdot
 \end{array}
\nonumber
\end{equation}
The estimate above and the embedding $ H^{1} \left( B(O,1) \right) \hookrightarrow  L^{(p+1)+}(B(O,1))  $ show that

\begin{equation}
\begin{array}{ll}
\lim \limits_{s \rightarrow \infty} \mathcal{N}_{m} \left( \tilde{w}^{\delta}(s) \right)  & = 0
\end{array}
\nonumber
\end{equation}
Given that the functional $s\mapsto \q N_m(\tilde w^\delta(s))$ is
decreasing by item (i), this impossible, because of
(\ref{Eqn:AssNNeg}). Thus, a contradiction follows and item (ii)
follows. This concludes the proof of Lemma \ref{lemlyap}.
\end{proof}


\subsection{ Proposition \ref{Prop:BoundPolynws}: the proof}

In this subsection we prove Proposition \ref{Prop:BoundPolynws}. First we deduce from Lemma \ref{Lem:DecreaseN} a polynomial bound of averages (in space and in time) of $w$.

\begin{lem}
If  $s \geq s_{0}$  then the following holds:

\begin{equation}
\begin{array}{l}
0 \leq \mathcal{N}_{m} \left( w(s) \right) \leq \mathcal{N}_{m} \left( w(s_{0}) \right), \; \text{and}
\end{array}
\label{Eqn:EstDecN}
\end{equation}

\begin{equation}
\begin{array}{rl}
&\int_{s}^{s+1} \int_{B(O,1)} \left( |\nabla w|^{2} (1 - |y|^{2}) +  \frac{ \left( \partial_{\tau} w \right)^{2} }{1- |y|^{2}}
+ w^{2} \right)
  \rho \; dy \; d \tau\\
  \lesssim & \mathcal{N}_{m} \left( w(s_{0}) \right) \left( s \log^{1+b}(s) \right) \cdot
\end{array}
\label{Eqn:EstAverNablaW}
\end{equation}
Here $s_{0}$ is the number defined in Lemma \ref{Lem:DecreaseN}.

\label{Lem:BoundPoly}
\end{lem}

\begin{proof}
(\ref{Eqn:EstDecN}) follows from (\ref{Eqn:EstDerivN});
(\ref{Eqn:EstDecreaseAvN}) and (\ref{Eqn:EstDecN}) yield
(\ref{Eqn:EstAverNablaW}).

\end{proof}

We now prove Proposition \ref{Prop:BoundPolynws}. Recall from the
covering lemma (see e.g Proposition 3.4 page 1138 in \cite{MZimrn05}) that there exists
$C:= C(\delta_{0}) > 0 $ such that

\begin{equation}
\begin{array}{l}
\sup \limits_{ \left\{ x: \, |x -x_{0}| \leq
\frac{T_{0} - \frac{t_{0}(x_{0}) + T(x_{0})}{2}}{\delta_{0}} \right\} } \int_{B(O,1)} (\partial_{s} w(s))^{2} +
|\nabla w(s)|^{2} + w^{2}(s)  \; dy  \\
\leq C \sup \limits_{
\left\{ x: \, |x -x_{0}| \leq \frac{T_{0} - \frac{t_{0}(x_{0}) + T(x_{0})}{2}}{\delta_{0}}  \right\} }
\int_{B \left( 0, \frac{1}{2} \right)}  (\partial_{s} w(s))^{2} + |\nabla w(s)|^{2}  + w^{2}(s)  \; dy \cdot
\end{array}
\label{Eqn:CoverConseqGradw}
\end{equation}
Hence combining this estimate with Lemma \ref{Lem:BoundPoly} we see that (\ref{Eqn:EstKinetw}) holds. This proves Proposition \ref{Prop:BoundPolynws}.


\section{Proof of Proposition \ref{Prop:BoundFinalWs}}

In this section we prove Proposition \ref{Prop:BoundFinalWs}.

\subsection{Polynomial pointwise bound of $w(s)$}

First we prove a polynomial pointwise bound of $ \left\| ( w(s),\partial_{s} w(s) ) \right\|^{2}_{H^{1} (B(O,1)) \times L^{2} (B(O,1))}$ :

\begin{lem}

%  Let $u$ be a solution of
%\eqref{equ}-\eqref{pert2} defined on a domain $\q D$ for some
%$1$-Lipschitz nonnegative function $x\mapsto T(x)$.
%(\ref{Eqn:WaveFoc})  with blow-up graph $\Gamma : x \rightarrow
 %\left\{ x \rightarrow T(x) \right\}$.
%Let $x_{0}$ be a non-characteristic point. Then there exist $C > %0$, $t_{0}(x_{0}) \in \left[ 0, T(x_{0}) \right)$ and \\
%$K := K \left( \delta_{0}(x_{0}), \left\| \left( u(t_{0}(x_{0})), %\partial_{t} u(t_{0}(x_{0})) \right) \right\|_{H^{1} \times L^{2} %\left( B \left(  x_{0},  \frac{T(x_{0}) -t_{0}(x_{0})}{\delta_{0}%(x_{0})} \right) \right)}
%\right)$ such that\\
%$T(x_{0}) - t_{0}(x_{0}) \ll 1$; and for all
%$T_{0} \in \left( t_{0}(x_{0}), T(x_{0}) \right) $, for all $x \in %\mathbb{R}^{N}$ such that
%$|x - x_{0}| \leq \frac{T_{0} -t_{0}(x_{0})}{\delta(x_{0})}$, and %for all $ s \geq - \log \left( T^{*}(x) - t_{0}(x_{0}) \right) $ we %have

There exists $C > 0$  such that for all $s \gg 1$

\begin{equation}
\begin{array}{ll}
\| w(s) \|_{H^{1} \left( B(O,1) \right)} + \| \partial_{s} w(s) \|_{L^{2} \left( B(O,1) \right)} \lesssim  s^{C} \cdot
\end{array}
\label{Eqn:Bound1Ws}
\end{equation}
%Here $w := w_{x, T^{*}(x)}$ is the function defined in
%(\ref{Eqn:SimilVar}) with $T^{*}(x) := T_{0} - \delta_{0}(x_{0}) %|x- x_{0}| $ and
%$\delta_{0}(x_{0})$ is such that the cone $\q C_{x_0, %\delta_0}\subset
%\q D$.
% $\delta_{0}(x_{0})$ is the number defined in
%(\ref{Eqn:DefNonCharac}).

\label{Lem:PointwiseEstw}
\end{lem}

\begin{proof}

We see from (\ref{Eqn:CoverConseqGradw}) and Proposition \ref{Prop:BoundPolynws} that

\begin{equation}
\begin{array}{l}
\int_{s}^{s+1} \int_{B(O,1)} \left( ( \partial_{\tau} w )^{2} + |\nabla w |^{2} + w^{2} \right) \; \rho dy d \tau  \lesssim  s \log^{1+b}(s) \cdot
\end{array}
\label{Eqn:EstAverw}
\end{equation}
Let $ q  > 1 $ to be chosen later. The mean value theorem shows that there exists $ \sigma(s) \in [s,s+1] $ such that

\begin{equation}
\begin{array}{l}
\int_{B(O,1)}  \left| w (\sigma(s)) \right|^{q} \; \rho dy  =  \int_{s}^{s+1} \int_{B(O,1)}  | w |^{q} \;  dy d \tau
\end{array}
\nonumber
\end{equation}
Hence the combination of the equality above with the fundamental theorem of calculus shows that

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} |w(s)|^{q} \; \rho dy =  \int_{s}^{s+1} \int_{B(O,1)}  | w |^{q} \;  dy d \tau +
\int_{\sigma(s)}^{s}  \frac{d}{d \tau} \int_{B(O,1)} |w(\tau)|^{q} \; \rho dy d \tau \cdot
\end{array}
\label{Eqn:EstLpw}
\end{equation}
We then claim that the following holds: \\
\\
\underline{Claim}: there exists $ C > 0 $ such that the following holds:

\begin{equation}
\begin{array}{l}
N =1, \infty > r \geq 2: \; \int_{B(O,1)} |w(s)|^{r} \; dy  \lesssim s^{C} \\
N > 1: \; \int_{B(O,1)} |w(s)|^{\frac{2N}{N-1}} \; dy  \lesssim s^{C} \cdot
\end{array}
\label{Eqn:LogControlBoundLpw}
\end{equation}

\begin{proof}

Assume that $N=1$ (resp. $N > 1$). Apply (\ref{Eqn:EstLpw}) with $ q = r $ (resp.  $q = \frac{2N}{N-1}$). In order to estimate the first term of the RHS of (\ref{Eqn:EstLpw}), we use the embedding $H^{1} \left( (s,s+1) \times B(O,1) \right) \hookrightarrow L^{r} $  (resp. $x^{q} \lesssim 1 + x^{2^{*}} $ for $x \geq 0$ ). In order to estimate the second term of the RHS of (\ref{Eqn:EstLpw}) we proceed as follows: we first use the composition rule for the derivative; then we apply the estimate $ a b  \leq \frac{a^{2}}{2} + \frac{b^{2}}{2}$ with $ (a,b):= \left( |w(y,\tau)|^{q-1}, |\partial_{\tau} w(y,\tau)| \right) $,  then we use the embedding
$ H^{1} \left( (s,s+1) \times B(O,1) \right) \hookrightarrow L^{2(q-1)} $. Hence using also (\ref{Eqn:EstAverw}) we see that (\ref{Eqn:Bound1Ws}) holds.

\end{proof}

We then claim the following: \\
\\
\underline{Claim}: Let $N > 1$. Then there exists  $\bar{\epsilon} := \bar{\epsilon} (p) > 0 $ for which the following holds: there exist $ \left( \beta, \gamma, C \right) \in \mathbb{R}^{+} \times (0,1) \times \mathbb{R}^{+} $ such that

\begin{equation}
\begin{array}{ll}
p + 1 + \bar{\epsilon} \leq \frac{2N}{N-1}: & \int_{B(0,1)} |w(s)|^{p + 1 + \bar{\epsilon}} \; dy \lesssim s^{C}    \\
p + 1 + \bar{\epsilon} > \frac{2N}{N-1}: &  \int_{B(0,1)} |w(s)|^{p + 1 + \bar{\epsilon}} \; dy  \lesssim \\
&   \left( \int_{B(O,1)} |w(s)|^{\frac{2N}{N-1}} \; dy \right)^{\beta} \left( \int_{B(O,1)} |w(s)|^{2} + |\nabla w(s)|^{2} \; dy \right)^{\gamma}  \cdot
\end{array}
\label{Eqn:BoundLpEpsw}
\end{equation}

\begin{proof}

Indeed let $\bar{\epsilon}> 0$ be such that $ \frac{p + 1 + \bar{\epsilon}}{2} \theta < 1 $ with $\theta$ such that
$ \frac{1}{p + 1 + \bar{\epsilon}} = \frac{1- \theta}{\frac{2N}{N-1}} + \frac{\theta}{\frac{2N}{N-2}} $. Such a choice of $\bar{\epsilon}$
is always possible since $ p < 1 + \frac{4}{N-1} $. If  $ p + 1 + \bar{\epsilon} > \frac{2N}{N-1} $ then the Gagliardo-Nirenberg inequality yields

\begin{equation}
\begin{array}{l}
\| w(s) \|_{L^{p+1 + \bar{\epsilon}}(B(O,1))} \lesssim \| w(s) \|^{1- \theta}_{L^{\frac{2N}{N-1}} (B(O,1))}   \| w (s) \|^{\theta}_{H^{1} (B(O,1))}
\end{array}
\end{equation}
If $ p + 1 + \bar{\epsilon} \leq \frac{2N}{N-1} $ then the H\"older inequality applied to (\ref{Eqn:LogControlBoundLpw}) shows that
(\ref{Eqn:BoundLpEpsw}) holds.

\end{proof}

Elementary estimates show that

\begin{equation}
\begin{array}{l}
a > 0: \\
\begin{array}{ll}
g(\phi w(s)) & \lesssim \log^{a} \left( \log(\phi(s)) \right) \mathbbm{1}_{\{ \phi(s) \geq |w(s)| \}} +
\log^{a} \left( \log(|w(s)|) \right)  \mathbbm{1}_{ \{ |w(s)| \geq \phi(s) \} } \\
& \lesssim  \log^{a}(s) + |w(s)|^{\bar{\epsilon}}
\end{array}
\\
a < 0, \;  \phi w^{2}(s) \gg 1: \, g(\phi w(s)) \lesssim \log^{a} \left( \log(\phi(s)) \right) \\
\end{array}
\label{Eqn:ElemEstgw}
\end{equation}
Moreover we see from (\ref{Eqn:EstDecN})  and (\ref{Eqn:EstJw}) that there exists $C > 0$ such that

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} \left( (\partial_{s} w(s))^{2} +  |\nabla w(s)|^{2} (1 -|y|^{2}) + w^{2}(s) \right) \; \rho  dy \\
\lesssim  e^{-\frac{2(p+1)s}{p-1}} (\log(s))^{\frac{2a}{p-1}} \int_{B(O,1)} F(\phi w(s)) \; \rho  dy  + O \left( (\log(s))^{C} \right) \cdot
\end{array}
\label{Eqn:EstKinetPrelim}
\end{equation}
We then estimate $ \int_{B(O,1)} \left( (\partial_{s} w(s))^{2} +  |\nabla w|^{2} (1 -|y|^{2}) + w^{2}(s) \right) \; \rho  dy $ . Using
(\ref{Eqn:Defphi}) and (\ref{Eqn:ApproxPot}) on $ \{ \phi w^{2}(s) \gg
1 \} $, then %and
$ F(\phi w(s)) \lesssim  (\phi w(s))^{p+1} \left( g(\phi w(s)) + 1
\right)  $ on\\
$ \{ \phi w^{2}(s) \lesssim 1 \} \subset \{ w(s) \lesssim \phi^{-\frac{1}{2}}(s) \} $,    we get

\begin{equation}
\begin{array}{rl}
&e^{-\frac{2(p+1)s}{p-1}} (\log(s))^{\frac{2a}{p-1}} \int_{B(O,1)}
  F(\phi w(s)) \rho \; dy\\
  \lesssim
&\frac{1}{(\log(s))^{a}} \int_{ B(O,1) \cap \{ \phi w^{2}(s) \gg 1 \} } |w(y,s)|^{p+1} g (\phi w(s)) \; \rho dy + e^{-s}.
\end{array}
\label{Eqn:PrelimBoundPot}
\end{equation}
We see from (\ref{Eqn:LogControlBoundLpw}), (\ref{Eqn:BoundLpEpsw}), (\ref{Eqn:ElemEstgw}), and (\ref{Eqn:PrelimBoundPot}) that there exists $C > 0$ such that

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} \left( (\partial_{s} w)^{2}(s) +  |\nabla w(s)|^{2} (1 - |y|^{2}) + w^{2}(s) \right) \; \rho  dy \\
\leq   s^{C} \left( 1 +  \left( \int_{B(O,1)}  (\partial_{s} w)^{2}(s) + |\nabla w|^{2}(s) + w^{2}(s) \; dy \right)^{\gamma} \right) \cdot
\end{array}
\label{Eqn:BoundGradw}
\end{equation}
Hence, combining (\ref{Eqn:CoverConseqGradw}) with (\ref{Eqn:BoundGradw}) we see that there exists $ C := C(\delta_{0}) > 0 $ such that

\begin{equation}
\begin{array}{l}
\sup \limits_{ \left\{ x: \, |x-x_{0}| \leq \frac{T_{0} -
\frac{T(x_{0})+t_{0}(x_{0})}{2}} {\delta_{0}} \right\}  } \int_{B(O,1)}  (\partial_{s} w(s))^{2} +
|\nabla w(s)|^{2} + w^{2}(s)  \; dy  \leq  s^{C} \cdot
\end{array}
\label{Eqn:LastStepBoundH1Normw}
\end{equation}
Hence there exists $ C > 0 $ such that (\ref{Eqn:Bound1Ws}) holds. \\


\end{proof}


\subsection{Improved estimate of derivative of $\mathcal{E}(w)$}

The lemma below yields a better estimate of the derivative of $\mathcal{E}(w)$  than (\ref{Eqn:EstDerivE}):

\begin{lem}

  There exists a constant $C > 0$ such that for all $ s \gg 1  $  we have:

\begin{equation}
\begin{array}{ll}
\frac{d \mathcal{E}(w)}{ds} & \leq  - \frac{3 \alpha}{2} \int_{B(O,1)} \frac{(\partial_{s} w)^{2}}{1- |y|^{2}} \rho \; dy \\
+&  \frac{C \log \left( \log( \log(s) ) \right)}{s \log^{a+2}(s)} \int_{B(O,1)} |w|^{p+1} g(\phi w) \rho \; dy + \frac{C}{s^{2}} \int_{B(O,1)} |\nabla w|^{2}(1- |y|^{2}) \rho \; dy \\
+& \frac{C}{s^{2}} \int_{B(O,1)} w^{2} \rho \; dy + \frac{C}{s^{\frac{7}{4}}} \cdot
\end{array}
\label{Eqn:SndDerivE}
\end{equation}

\label{Lem:SndDerivE}
\end{lem}

\begin{proof}

We already know from the proof of Lemma \ref{Lem:EstDerivE} that there exists a constant $C > 0$ such that

\begin{equation}
\begin{array}{ll}
\frac{d \mathcal{E}(w)}{ds} & \leq  - \frac{3 \alpha}{2} \int_{B(O,1)} \frac{(\partial_{s} w)^{2}}{1- |y|^{2}} \rho \; dy
+ \frac{C}{s^{2}} \int_{B(O,1)} |\nabla w|^{2}(1- |y|^{2}) \rho \; dy \\
& + \frac{C}{s^{2}} \int_{B(O,1)} w^{2} \rho \; dy + C e^{-s} + X + Y,
\end{array}
\nonumber
\end{equation}
with $X$ and $Y$ defined in (\ref{Eqn:DefXY}). We have $ X + Y  = \chi_{1}(s) + \chi_{2}(s) $ with

\[
  \begin{array}{rl}
    \chi_{1}(s) := \frac{a}{(p+1) s \log^{a+1}(s) }  \int_{B(O,1)}
  &|w|^{p+1} \log^{a} \left( \log ( 10 + \phi^{2} w^{2} ) \right)\\
&\times    \left( 1 -  \frac{4 s \log(s)}{ (p-1) \log \left( \log ( 10 + \phi^{2} w^{2} ) \right) \log ( 10+ \phi^{2} w^{2}) }   \right) \rho \; dy,
% \chi_{1}(s) := \frac{a}{(p+1) s \log^{a+1}(s) }  \int_{B(O,1)} |w|^{p+1} \log^{a} \left( \log ( 10 + \phi^{2} w^{2} ) \right)
% \left( 1 -  \frac{4 s \log(s)}{ (p-1) \log \left( \log ( 10 +
% \phi^{2} w^{2} ) \right) \log ( 10+ \phi^{2} w^{2}) }   \right) \rho
% \; dy,    \\
  \end{array}
  \]
  and
  \[
\begin{array}{rl}
    \chi_{2}(s)   := &
  \frac{2 e^{-\frac{2(p+1)s}{p-1}}}{p-1} \log^{\frac{2a}{p-1}}(s)\\
  &\times\int_{B(O,1)}
\left( (p+1) F_{2}(\phi w) - \frac{a}{s \log(s)} F_{1}(\phi w) - \frac{a}{s \log(s)} F_{2}(\phi w) \right) \rho \; dy \cdot
\end{array}
\]
% \nonumber
% \end{equation}
% where $\alpha =  1 -  \frac{4 s \log(s)}{ (p-1) \log \left( \log ( 10 + \phi^{2} w^{2} ) \right) \log ( 10+ \phi^{2} w^{2}) }$.\\
We divide the region of integration into two regions:\\
$ X_{<} := \left\{ y \in B(O,1): \; \phi(s) w^{2}(y,s) \lesssim 1  \right\} $, and\\
$ X_{>} := \left\{ y \in B(O,1): \; \phi(s) w^{2}(y,s) \gg 1  \right\} $. \\
\\
We first estimate $\chi_{1}(s)$. \\
\\
Let

\begin{equation}
\begin{array}{ll}
\chi_{1,<}(s) & := \frac{a}{(p+1) s \log^{a+1}(s)} \times \\
& \int_{X_{<}}  |w|^{p+1} \log^{a} \left( \log ( 10 + \phi^{2} w^{2} ) \right)
\left( 1 -  \frac{4 s \log(s)}{ (p-1) \log \left( \log ( 10 + \phi^{2} w^{2} ) \right) \log ( 10+ \phi^{2} w^{2}) }   \right) \rho \; dy, \; \text{and} \\
& \\
\chi_{1,>}(s) & :=  \frac{a}{(p+1) s \log^{a+1}(s)}  \times \\
& \int_{X_{>}} |w|^{p+1} \log^{a} \left( \log ( 10 + \phi^{2} w^{2} ) \right) \left( 1 -  \frac{4 s \log(s)}{ (p-1) \log \left( \log ( 10 + \phi^{2} w^{2} ) \right) \log ( 10+ \phi^{2} w^{2}) }   \right) \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}

We first control $\chi_{1,<}(s)$. Clearly

\begin{equation}
\begin{array}{l}
|w|^{p+1} \log^{a} \left( \log (  10 +  \phi^{2} w^{2}) \right) \lesssim \phi^{-\frac{p+1}{2}} \log^{a} \left( \log (  10 +  \phi) \right)
\lesssim e^{-s} \cdot
\end{array}
\label{Eqn:Pointwisew}
\end{equation}
Hence taking into account that $  1 -  \frac{4 s \log(s)}{ (p-1) \log \left( \log ( 10 + \phi^{2} w^{2} ) \right) \log ( 10+ \phi^{2} w^{2}) } \lesssim 1 $ we get

\begin{equation}
\begin{array}{l}
\chi_{1,<}(s) \lesssim e^{-s} \cdot
\end{array}
\nonumber
\end{equation}

We then control $\chi_{1,>}(s)$. We first observe that

\begin{equation}
\begin{array}{l}
\log (10 + \phi^{2} w^{2}) \gtrsim \log(\phi) \gtrsim s, \; \text{and} \\
\log \left( \log (10 + \phi^{2} w^{2}) \right) \gtrsim  \log \left( \log(\phi) \right) \gtrsim \log(s) \cdot
\end{array}
\label{Eqn:LowerBoundDen}
\end{equation}
We have

\begin{equation}
\begin{array}{l}
\log(10 + \phi^{2} w^{2}) = \log(\phi^{2})  +  \log (10 \phi^{-1} + w^{2} ), \; \text{and} \\
\log  \left( \log(10 + \phi^{2} w^{2} ) \right) =  \log \left( \log( \phi^{2} ) \right) + \log \left( 1 + \frac{\log \left(w^{2} + 10 \phi^{-2} \right)}{\log(\phi^{2})} \right) \cdot
\end{array}
\nonumber
\end{equation}
Also $ \log(\phi^{2}) \log \left( \log( \phi^{2} ) \right) = \frac{4s}{p-1} \log(s) + O \left( s \log ( \log (\log(s)) ) \right) $. Hence
$(p-1) \log \left(  \log( 10 + \phi^{2} w^{2} ) \right) \log (10 + \phi^{2} w^{2})  - 4s \log(s) =  X  + Y  + Z + O \left( s \log \left( \log \left( \log(s) \right) \right) \right) $
with

\begin{equation}
\begin{array}{l}
X := (p-1) \log ( w^{2} + 10 \phi^{-2}) \log \left( \log(\phi^{2}) \right), \\
Y  := (p-1) \log(\phi^{2}) \log \left( 1 + \frac{\log (10 \phi^{-2} +  w^{2})}{\log(\phi^{2})} \right), \; \text{and} \\
Z:= (p-1) \log ( w^{2} + 10 \phi^{-2} ) \log \left( 1 + \frac{ \log(w^{2} + 10 \phi^{-2})}{\log(\phi^{2})} \right) \cdot
\end{array}
\nonumber
\end{equation}
Hence using also the elementary inequality $ \log(1 + x) \lesssim  x $ for $x \geq 0$ we get

\begin{equation}
\begin{array}{l}
(p-1) \log \left(  \log( 10 + \phi^{2} w^{2} ) \right) \log (10 + \phi^{2}  w^{2})  - 4s \log(s)  \\
= O \left( s \log ( \log (\log(s)) ) \right) + O \left( \log(s) \log(w^{2} +  10 \phi^{-2} ) \right) \\
+ O \left( \frac{\log^{2}(w^{2} + 10 \phi^{-2})}{s} \right)  \cdot
\end{array}
\nonumber
\end{equation}
Next we claim that the following holds: \\
\\
\underline{Claim}: Let $q \in \{ 1,2 \}$. Then there exists $C > 0 $ such that

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} |w(s)|^{p+1} g(\phi w(s)) \log^{q} \left( 10 + w^{2}(s) \right) \; \rho dy \lesssim  s^{\frac{1}{8}}
+ \int_{B(O,1)} |w(s)|^{p+1} g(\phi w(s)) \; \rho dy
\end{array}
\label{Eqn:BoundPotwLogCorr}
\end{equation}

\begin{proof}

Let $ 0 < \epsilon \ll 1$ and $ 0 < \epsilon^{'} \ll \epsilon $ such that all the statements in the proof are correct. From
$ \log (10 + x^{2}) \lesssim 1 + |x|^{\epsilon^{'}}$ we get

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} |w(s)|^{p+1} g( \phi w(s) ) \log^{q} \left( 10 + w^{2}(s) \right)  \; \rho dy \\
\lesssim \int_{B(O,1)} |w(s)|^{p+1} g( \phi w(s) ) \; \rho dy + \int_{B(O,1)} |w(s)|^{p + 1 + \epsilon'} g(\phi w(s)) \; \rho dy
\end{array}
\nonumber
\end{equation}
Moreover the interpolation inequality w.r.t the measure $ d \mu  := g (\phi w(s)) \rho dy $  shows that there exists $ 0 <  \theta \ll 1 $ such that

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} |w(s)|^{p + 1 + \epsilon'} g(\phi w(s)) \; \rho dy \\
\lesssim   \left( \int_{B(O,1)} |w(s)|^{p+1} g(\phi w(s)) \; \rho dy \right)^{1- \theta}  \left( \int_{B(O,1)} |w(s)|^{p+1 + \epsilon} g(\phi w(s)) \; \rho dy \right)^{\theta}
\end{array}
\nonumber
\end{equation}
We then estimate $g(\phi w)$. If $ a > 0 $ then $ g( \phi w (s) ) \lesssim  \log^{a}(s) + |w(s)|^{\epsilon} $  by proceeding similarly as in (\ref{Eqn:ElemEstgw}).
If $a < 0$ then we clearly have $g(\phi w(s)) \lesssim 1 $. Hence using also the embeddings $ H^{1} ( B(O,1) ) \hookrightarrow L^{p + 1 + \epsilon} (B(O,1)) $
and $ H^{1} ( B(O,1) ) \hookrightarrow L^{p + 1 + 2 \epsilon} (B(O,1)) $, and (\ref{Eqn:Bound1Ws}) we see that there exists $C > 0$ such that

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} |w(s)|^{p+1 + \epsilon} g(\phi w(s)) \; \rho dy \lesssim s^{C} \cdot
\end{array}
\label{Eqn:Boundpowerwplusg}
\end{equation}
Hence we get (\ref{Eqn:BoundPotwLogCorr})  from $s^{C} = s^{C- \frac{1}{8}} s^{\frac{1}{8}} $  (with $C$ defined in (\ref{Eqn:Boundpowerwplusg})) and $ u^{\theta} v^{1 - \theta} \leq \theta u + (1 - \theta) v $.

\end{proof}

We then estimate $\chi_{2}(s)$.  To this end we write $\chi_{2}(s) = \chi_{2, <}(s) + \chi_{2, >}(s) $ with

\begin{equation}
\begin{array}{l}
\chi_{2, <}(s) :=  \frac{2 e^{-\frac{2(p+1)s}{p-1}}}{p-1} \log^{\frac{2a}{p-1}}(s) \int_{X_{<}}
\left( (p+1) F_{2}(\phi w) - \frac{a}{s \log(s)} F_{1}(\phi w) - \frac{a}{s \log(s)} F_{2}(\phi w) \right) \rho \; dy, \; \text{and} \\
\chi_{2, >}(s) := \frac{2 e^{-\frac{2(p+1)s}{p-1}}}{p-1} \log^{\frac{2a}{p-1}}(s) \int_{X_{>}}
\left( (p+1) F_{2}(\phi w) - \frac{a}{s \log(s)} F_{1}(\phi w) - \frac{a}{s \log(s)} F_{2}(\phi w) \right) \rho \; dy \cdot
\end{array}
\nonumber
\end{equation}

We first control $\chi_{2,<}(s)$. By proceeding similarly as in (\ref{Eqn:ErrorFPhi}) we get

\begin{equation}
\begin{array}{ll}
\chi_{2,<}(s) \lesssim e^{-s} \cdot
\end{array}
\nonumber
\end{equation}

We then control $\chi_{2, >}(s)$. We see from  (\ref{Eqn:DefF1}), (\ref{Eqn:ApproxPot}), and (\ref{Eqn:LowerBoundDen}) that

\begin{equation}
\begin{array}{l}
\chi_{2,>}(s) \lesssim \frac{1}{\log^{a+1}(s) s^{2}} \int_{B(O,1)} |w|^{p+1} g(\phi w) \; \rho  dy \cdot
\end{array}
\nonumber
\end{equation}
\vphantom \\
\\
Hence by combining all the estimates above we get (\ref{Eqn:SndDerivE}).




% \color{red} COMMENTS; PROOF USING EMBEDDING IN L INFTY IN N=1


%Assume now that $N=1$. Then from (\ref{Eqn:Bound1Ws}) and the embedding $H^{1}(-1, 1) \hookrightarrow L^{\infty}(-1,1)$ we see that exists $ C > 0 $
%such that

%\begin{equation}
%\begin{array}{l}
%\| w(s) \|_{L^{\infty}(-1,1)} \lesssim (\log(s))^{C} \cdot
%\end{array}
%\label{Eqn:PolyLinf}
%\end{equation}
%By (\ref{Eqn:PolyLinf}) and the product rule of the $log$ we see that there exists $C > 0$ such that

%\begin{equation}
%\begin{array}{ll}
%\log \left( \log ( 10 + \phi^{2} w^{2} ) \right) \log \left( 10+ \phi^{2} w^{2} \right) & \lesssim \log( C \phi^{2} w^{2} )
%\log \left( \log ( C \phi^{2} w^{2}) \right)  \\
%& \leq \frac{4s \log(s)}{p-1} + O \left( s \log ( \log(\log(s)) ) \right)
%\end{array}
%\nonumber
%\end{equation}

%Hence

%\begin{equation}
%\begin{array}{l}
%\chi_{1,>} (s) \lesssim \frac{\log \left( \log(s) \right)}{\log^{a+2}(s) s} \int_{B(O,1)} |w|^{p+1}  g(\phi w) \rho(y) \; dy  \cdot
%\end{array}
%\end{equation}
%\vphantom \\
%\\
%We then estimate $\chi_{2}(s)$.  To this end we write $\chi_{2}(s) = \chi_{2, <}(s) + \chi_{2, >}(s) $ with

%\begin{equation}
%\begin{array}{l}
%\chi_{2, <}(s) :=  \frac{2 e^{-\frac{2(p+1)s}{p-1}}}{p-1} \log^{\frac{2a}{p-1}}(s) \int_{X_{<}}
%\left( (p+1) F_{2}(\phi w) - \frac{a}{s \log(s)} F_{1}(\phi w) - \frac{a}{s \log(s)} F_{2}(\phi w) \right) \rho \; dy, \; \text{and} \\
%\chi_{2, >}(s) := \frac{2 e^{-\frac{2(p+1)s}{p-1}}}{p-1} \log^{\frac{2a}{p-1}}(s) \int_{X_{>}}
%\left( (p+1) F_{2}(\phi w) - \frac{a}{s \log(s)} F_{1}(\phi w) - \frac{a}{s \log(s)} F_{2}(\phi w) \right) \rho \; dy \cdot
%\end{array}
%\nonumber
%\end{equation}

%We first control $\chi_{2,<}(s)$. Clearly in view of (\ref{Eqn:DecompF}), (\ref{Eqn:DefF1}), and $|F(x)| \lesssim |x|^{p+1} (1+ g(x)) $ we get

%\begin{equation}
%\begin{array}{ll}
%\chi_{2,<}(s) \lesssim  \frac{1}{\log^{a}(s)} \int_{X_{<}} |w|^{p+1} g(\phi w) \; \rho dy  \lesssim e^{-s} \cdot
%\end{array}
%\nonumber
%\end{equation}

%We then control $\chi_{2, >}(s)$. We see from  (\ref{Eqn:DefF1})  and (\ref{Eqn:ApproxPot}) that

%\begin{equation}
%\begin{array}{l}
%\chi_{2,>}(s) \lesssim \frac{1}{\log^{a+1}(s) s^{2}} \int_{B(O,1)} |w|^{p+1} g(\phi w) \; \rho  dy \cdot
%\end{array}
%\nonumber
%\end{equation}
%\vphantom \\
%\\
%Hence by combining all the estimates above we get (\ref{Eqn:SndDerivE}). \color{black}

\end{proof}

\subsection{Decreasing property of a Lyapunov functional}

Let $u $ be a solution of \eqref{equ}-\eqref{pert2}.
%(\ref{Eqn:WaveFoc}).
We define $\mathcal{L}_{0}(w)$ in the following fashion:

\begin{equation}
\mathcal{L}_{0}(w): s \rightarrow \mathcal{L}_{0}(w(s)) := \mathcal{E}(w(s))  - \frac{1}{s \log^{\frac{3}{2}}(s)}
\int_{B(O,1)} w(s) \partial_{s} w(s) \rho \; dy \cdot
\label{Eqn:DefL0}
\end{equation}
We prove the following lemma:

\begin{lem}

There exists a constant $ C > 0 $ such that for all $ s \gg 1 $ we have

\begin{equation}
\begin{array}{ll}
\frac{d \mathcal{L}_{0}(w)}{ds} & \leq - \alpha  \int_{B(O,1)}  \frac{(\partial_{s} w)^{2}}{1-|y|^{2}} \rho \; dy
+ \frac{C}{s \log^{\frac{3}{2}}(s)} \mathcal{L}_{0} (w) +  \frac {C}{s^{\frac{7}{4}}} \cdot
\end{array}
\label{Eqn:L0Bound}
\end{equation}

\label{Lem:DerivL0}
\end{lem}

\begin{proof}

Observe that $ \mathcal{L}_{0} (w) = \mathcal{E}(w) + \frac{1}{\log^{\frac{1}{2}}(s)} \mathcal{J} (w) $. Hence

\begin{equation}
\begin{array}{l}
\frac{d \mathcal{L}_{0}(w)}{ds} = \frac{d \mathcal{E}(w)}{ds} - \frac{1}{2 s \log^{\frac{3}{2}}(s)} \mathcal{J}(w)
+ \frac{1}{\log^{\frac{1}{2}}(s)} \frac{d \mathcal{J}(w)}{ds}
\end{array}
\label{Eqn:DerivL0}
\end{equation}
We see from (\ref{Eqn:EstJw}) that $ \mathcal{E}(w) \approx \mathcal{L}_{0}(w) $. This estimate combined with (\ref{Eqn:DerivL0}), (\ref{Eqn:SndDerivE}), (\ref{Eqn:DerivJ}),
and (\ref{Eqn:EstJw}) yields (\ref{Eqn:L0Bound}).

\end{proof}

Let $C$ be the constant defined in Lemma \ref{Lem:DerivL0}. Let $m \in \mathbb{R}^{+}$. We define  the following map:

\begin{equation}
\tilde{\mathcal{L}}_{m}(w):
s \rightarrow \tilde{\mathcal{L}}_{m}(w(s)) :=  e^{\frac{2 C}{\log^{\frac{1}{2}}(s)}} \mathcal{L}_{0} (w(s))  + \frac{m}{s^{\frac{1}{2}}}  \cdot
\label{Eqn:DefL}
\end{equation}
This map has a decreasing property for $m \gg 1$

\begin{lem}

There exist $m \gg 1$ and a constant $ \bar{C} > 0 $ such that for all $s_{2} \geq s_{1} \gg 1 $

\begin{equation}
\begin{array}{l}
\tilde{\mathcal{L}}_{m}(w(s_{2})) - \tilde{\mathcal{L}}_{m} (w(s_{1})) \leq
- \bar{C} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} \frac{(\partial_{\tau} w)^{2}}{1 - |y|^{2}} \rho \; dy \; ds
\end{array}
\label{Eqn:DecL}
\end{equation}

\label{Lem:DecL}
\end{lem}

\begin{proof}

Observe that $f(s) := e^{\frac{2 C}{ \log^{\frac{1}{2}}(s)}}$ satisfies $f^{'}(s) = -\frac{C}{s \log^{\frac{3}{2}}(s)} f(s) $. Hence
by using the operation rules of the derivative we get

\begin{equation}
\begin{array}{l}
\frac{d \tilde{\mathcal{L}}_{m}(w)}{d s} \leq - \alpha e^{\frac{2 C}{ \log^{\frac{1}{2}}(s)}}
 \int_{B(O,1)} \frac{(\partial_{s} w)^{2}}{1 - |y|^{2}} \rho \; dy  + C  e^{\frac{2 C}{ \log^{\frac{1}{2}}(s)}}
\frac{1}{s^{\frac{7}{4}}} - \frac{m}{2} \frac{1}{s^{\frac{3}{2}}}
\end{array}
\nonumber
\end{equation}
Hence we see that (\ref{Eqn:DecL}) holds.

\end{proof}

\subsection{Bounds}

In the corollary below we derive some bounds from Lemma \ref{Lem:DecL}. We have

\begin{cor}

There exists a constant $ C > 0 $ and $ s_{0} \gg 1$ such that
for all $s \geq s_{0} $ we have

\begin{equation}
\begin{array}{l}
-C \leq \mathcal{L}_{0}(w(s)) \leq C \left( \mathcal{L}_{0}(w(s_{0})) + 1 \right) \cdot
\end{array}
\label{Eqn:BoundL0}
\end{equation}
Moreover for all $s_{2} \geq s_{1} \geq s_{0}$ we have

\begin{equation}
\begin{array}{l}
\int_{s_{1}}^{s_{2}} \int_{B(O,1)} (\partial_{\tau} w)^{2} \frac{1}{1- |y|^{2}} \rho dy ds \lesssim  \mathcal{L}_{0}(w(s_{0})) + 1
\end{array}
\label{Eqn:BoundDerivw}
\end{equation}

\label{Cor:BoundL0}
\end{cor}

\begin{proof}

The upper bound in (\ref{Eqn:BoundL0}) is a direct consequence of the decreasing property of $s \rightarrow \tilde{\mathcal{L}}_{m} (w(s)) $ mentioned in  Lemma \ref{Lem:DecL}. It remains to prove the lower bound. By using the same argument to prove that
$\mathcal{N}_{m}(w(s)) \geq 0$ ( see proof of Lemma \ref{Lem:DecreaseN} ) we see that $ \tilde{\mathcal{L}}_{m} (w(s)) \geq 0 $ for all $ s \geq s_{0} $. But
then we see from (\ref{Eqn:DefL}) that the lower bound in (\ref{Eqn:BoundL0}) holds. But then combining (\ref{Eqn:BoundL0}) with (\ref{Eqn:DecL}) we get
(\ref{Eqn:BoundDerivw}).

\end{proof}


\section{Proof of Proposition \ref{Prop:BoundFinalWs} }

We then prove Proposition \ref{Prop:BoundFinalWs}. To this end we first bound the $s-$ average of the $L^{p+1}\left( B(O,1) \right))$  norm of $w$ on
intervals of size roughly equal to one. More precisely the following lemma holds:

\begin{lem}
Let $s \gg 1$. There exist $C > 0,$  $s_{1} \in [s-1,s] $, and $s_{2} \in [s, s+1]$ such that

\begin{equation}
\begin{array}{l}
\int_{s_{1}}^{s_{2}} \int_{B(O,1)} e^{-\frac{2(p+1)\tau}{p-1}} (\log(\tau))^{\frac{2a}{p-1}} F(\phi w) \; \rho dy d \tau \leq C \cdot
\end{array}
\label{Eqn:BoundAverPotNrj}
\end{equation}

\label{Lem:BoundAvPot}
\end{lem}

The proof is given in Subsection \ref{Subsec:ProofLemBoundAvPot}. \\
With this lemma in mind, we can now prove Proposition \ref{Prop:BoundFinalWs}. \\
\vphantom \\
We see from (\ref{Eqn:Ident0}), (\ref{Eqn:BoundAverPotNrj}), (\ref{Eqn:DecL}), (\ref{Eqn:EstJw}), and
$ |\nabla w(s,y)|^{2} (1 - |y|^{2}) \leq |\nabla w(s,y)|^{2} - |\nabla w(s,y) \cdot y|^{2} $ that

\begin{equation}
\begin{array}{l}
\int_{s}^{s+1} \int_{B(O,1)} \left( (\partial_{s} w )^{2} + |\nabla w|^{2} ( 1 - |y|^{2} ) + w^{2} \right) \; \rho dy d \tau \lesssim 1 \cdot
\end{array}
\label{Eqn:BoundKinetwAver}
\end{equation}
Next we follow again the steps from (\ref{Eqn:EstAverw}) to (\ref{Eqn:LastStepBoundH1Normw}),  except that we use (\ref{Eqn:BoundL0}) instead
of (\ref{Eqn:EstDecN}) to get

\begin{equation}
\begin{array}{l}
\int_{B(O,1)} \left( (\partial_{s} w(s))^{2} +  |\nabla w(s)|^{2} (1 -|y|^{2}) + w^{2}(s) \right) \; \rho  dy \\
\lesssim  e^{-\frac{2(p+1)s}{p-1}} (\log(s))^{\frac{2a}{p-1}} \int_{B(O,1)} F(\phi w(s)) \; \rho  dy  + O(1)
\end{array}
\nonumber
\end{equation}
instead of (\ref{Eqn:EstKinetPrelim}). Observe that the bounds in the process do no longer depend on $s$ since the bound in
(\ref{Eqn:BoundKinetwAver}) does no longer depend on $s$. \\
\\
Hence (\ref{Eqn:BoundFinalWs}) holds.



\subsection{Proof of Lemma \ref{Lem:BoundAvPot}}
\label{Subsec:ProofLemBoundAvPot}

Let $s_{1} \in [s-1,s]$ and $s_{2} \in [s+1,s+2]$ to be chosen. We first state some identities. Integrating $\mathcal{L}_{0}(w(\tau))$ from $s_{1}$ to $s_{2}$ we get from (\ref{Eqn:DefL0}) and (\ref{Eqn:Energyw})

\begin{equation}
\begin{array}{ll}
\int_{s_{1}}^{s_{2}} \mathcal{L}_{0} (w) \; d \tau & =
\int_{s_{1}}^{s_{2}} \int_{B(O,1)}
\left( \frac{1}{2} (\partial_{\tau} w)^{2} + \frac{p+1}{(p-1)^{2}} w^{2} - e^{-\frac{2(p+1) \tau}{p-1}}
(\log(\tau))^{\frac{2a}{p-1}} F(\phi w) \right) \; \rho  dy d \tau \\
& + \frac{1}{2}  \int_{s_{1}}^{s_{2}} \int_{B(O,1)} |\nabla w|^{2} ( 1 - |y|^{2})  \;  \rho  dy \ d \tau
- \int_{s_{1}}^{s_{2}} \frac{1}{\tau \log^{\frac{3}{2}}(\tau)} \int_{B(0,1)} w \partial_{\tau} w \;  \rho dy  d \tau  \cdot
\end{array}
\label{Eqn:Ident0}
\end{equation}
We then multiply (\ref{Eqn:Wavew}) by $w \rho$ and we integrate on $B(O,1) \times [s_{1},s_{2}]$. We get from (\ref{Eqn:ValAlpha}) and some integrations
by parts

\begin{equation}
\begin{array}{l}
\left[ \int_{B(O,1)} \left( w \partial_{\tau} w +  \left( \frac{p+3}{2(p-1)} - N \right) w^{2} \right) \; \rho dy  \right]_{s_{1}}^{s_{2}}  = \\
\\
\int_{s_{1}}^{s_{2}} \int_{B(O,1)} (\partial_{\tau} w)^{2} \; \rho  dy  d \tau
- \int_{s_{1}}^{s_{2}} \int_{B(O,1)}  \left( |\nabla w|^{2}  - |y \cdot \nabla w|^{2} \right) \; \rho  dy  d \tau \\
\\
- \frac{2(p+1)}{(p-1)^{2}} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} w^{2} \; \rho dy  d \tau
+ \int_{s_{1}}^{s_{2}} \int_{B(O,1)} e^{-\frac{2p \tau}{p-1}} (\log(\tau))^{\frac{a}{p-1}} w f(\phi w) \; \rho d y d \tau \\
\\
- 4 \alpha \int_{s_{1}}^{s_{2}} \int_{B(O,1)} w \partial_{\tau} w \frac{|y|^{2}}{1 - |y|^{2}} \; \rho dy d \tau
+ 2 \int_{s_{1}}^{s_{2}} \int_{B(O,1)} (\nabla w \cdot y) \partial_{\tau} w \; \rho  dy  d \tau \\
\\
+ \frac{2a}{p-1} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} \frac{1}{\tau \log(\tau)} (y \cdot \nabla w)  w \; \rho dy d  \tau
+ \int_{s_{1}}^{s_{2}} \int_{B(O,1)} \gamma w^{2} \; \rho dy d \tau \\
\\
+ \frac{2a}{p-1} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} \frac{1}{\tau \log(\tau)} \partial_{\tau} w  w \; \rho  dy d \tau \cdot
\end{array}
\label{Eqn:Ident1}
\end{equation}
Adding (\ref{Eqn:Ident0}) to $ \frac{1}{2} \times \text{(\ref{Eqn:Ident1})}$ we get, by (\ref{Eqn:DecompF}) and (\ref{Eqn:Defphi})

\begin{equation}
\begin{array}{l}
\frac{p-1}{2} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} e^{-\frac{2(p+1) \tau}{p-1}} (\log(\tau))^{\frac{2a}{p-1}} F(\phi w) \; \rho  dy d \tau = \\
\\
\frac{1}{2} \left[  \int_{B(O,1)} \left( w \partial_{\tau} w + \frac{5-p}{2(p-1)} w^{2} \right) \; \rho  dy  \right]_{s_{1}}^{s_{2}}
- \int_{s_{1}}^{s_{2}} \int_{B(O,1)} (\partial_{\tau} w)^{2} \; \rho dy d \tau  \\
\\
+ \int_{s_{1}}^{s_{2}} \mathcal{L}_{0} (w(\tau)) \; d \tau + \frac{4}{p-1} \int_{s_{1}}^{s_{2}} \int_{B(O,1)}
w \partial_{\tau} w \frac{|y|^{2} \rho}{1-|y|^{2}} \; \rho dy d \tau \\
\\
- \int_{s_{1}}^{s_{2}} \int_{B(O,1)} (\nabla w \cdot y) \partial_{\tau} w \; \rho dy  d \tau  + X_{1}+ ... + X_{6}  \cdot
\end{array}
\label{Eqn:CombinEq}
\end{equation}
Here

\begin{equation}
\begin{array}{l}
X_{1} := - \frac{a}{p-1} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} \frac{1}{\tau \log(\tau)} y \cdot \nabla w w \; \rho  d y d \tau, \; X_{2} := - \frac{1}{2} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} \gamma(s) w^{2} \; \rho dy d \tau, \\
\\
X_{3} := -\frac{a}{p-1} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} \frac{1}{\tau \log(\tau)} \partial_{\tau} w  w \; \rho dy d \tau, \;
X_{4} := \int_{s_{1}}^{s_{2}} \frac{1}{\tau \log^{\frac{3}{2}}(\tau)} \int_{B(O,1)} w \partial_{\tau} w \; \rho dy d \tau , \\
\\
X_{5} := \frac{p+1}{2} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} e^{-\frac{2(p+1)}{p-1} \tau} (\log(\tau))^{\frac{2a}{p-1}} F_{1}(\phi w) \; \rho d y d \tau, \; \text{and} \\
\\
X_{6} := \frac{p+1}{2} \int_{s_{1}}^{s_{2}} \int_{B(O,1)} e^{-\frac{2(p+1)}{p-1} \tau} (\log(\tau))^{\frac{2a}{p-1}} F_{2}(\phi w) \; \rho d y d \tau \cdot
\end{array}
\nonumber
\end{equation}
Let $ 1 \gg \epsilon > 0 $ (resp.  $C > 0 $) be a small (resp. large) constant such that all the statements below are true.
Let $ A :=   \epsilon^{-1}  +  \epsilon \int_{s_{1}}^{s_{2}} \int_{B(O,1)} |w|^{p+1}  \; \rho dy d \tau $ and
$ B :=  \epsilon^{-1}  + \epsilon \int_{s_{1}}^{s_{2}} \int_{B(O,1)} e^{- \frac{2(p+1)\tau}{p-1}} (\log(\tau))^{\frac{2a}{p-1}} F( \phi w) \; \rho dy d \tau $. \\
\\
We see from (\ref{Eqn:BoundDerivw}) and similar arguments as those in \cite{MZajm03} (see Section $2$) that

\begin{equation}
\begin{array}{l}
\sup \limits_{s_{1} \leq s \leq s_{2}} \int_{B(O,1)} w^{2}(s) \; \rho dy \lesssim A \cdot
\end{array}
\label{Eqn:BoundMassw}
\end{equation}
Hence using also (\ref{Eqn:Ident0}), (\ref{Eqn:BoundL0}), the inequality $ ab \leq \frac{a^{2}}{2} + \frac{b^{2}}{2}$,
and (\ref{Eqn:BoundDerivw}) we see that the following holds

\begin{equation}
\begin{array}{l}
\int_{s_{1}}^{s_{2}} \int_{B(O,1)} |\nabla w|^{2} (1 - |y|^{2}) \rho \; dy d \tau  \\
\lesssim 1 +  \int_{s_{1}}^{s_{2}} e^{-\frac{2(p+1) \tau}{p-1}} (\log(\tau))^{\frac{2a}{p-1}} \int_{B(O,1)} F (\phi w)  \; \rho dy d \tau \\
+ \frac{1}{s \log^{\frac{3}{2}}(s)} \left(  \int_{s_{1}}^{s_{2}} \int_{B(O,1)} w^{2} \rho \; dy  d \tau
+ \int_{s_{1}}^{s_{2}} \int_{B(O,1)} (\partial_{\tau} w)^{2} \rho \; dy d \tau  \right) \\
\leq C  + C \int_{s_{1}}^{s_{2}} e^{-\frac{2(p+1) \tau}{p-1}} (\log(\tau))^{\frac{2a}{p-1}} \int_{B(O,1)} F(\phi w)  \; \rho dy d \tau + A  \cdot
\end{array}
\label{Eqn:BoundMassGradientw}
\end{equation}
Hence, using also (\ref{Eqn:HardyType})

\begin{equation}
\begin{array}{ll}
|X_{1}| & \lesssim \frac{1}{s \log(s)}
\left(
\int_{s_{1}}^{s_{2}} \int_{B(O,1)}  |\nabla w|^{2} ( 1- |y|^{2}) \rho \; dy d \tau
+ \int_{s_{1}}^{s_{2}} \int_{B(O,1)}  w^{2} \frac{|y|^{2}}{1 - |y|^{2}} \; \rho dy d \tau \right) \\
& \lesssim A  \cdot
\end{array}
\nonumber
\end{equation}
By using (\ref{Eqn:BoundDerivw}), (\ref{Eqn:BoundMassGradientw}), and similar arguments as those in \cite{MZajm03} (see Section $2$) we get

\begin{equation}
\begin{array}{l}
\int_{s_{1}}^{s_{2}} \int_{B(O,1)} |\partial_{\tau} w|  |\nabla w \cdot y| \; \rho dy ds  +
\int_{s_{1}}^{s_{2}} \int_{B(O,1)} |\partial_{\tau} w| |w| \frac{|y|^{2}}{1- |y|^{2}} \; \rho dy ds \lesssim A + B \cdot
\end{array}
\nonumber
\end{equation}
We get from (\ref{Eqn:BoundMassw}) and (\ref{Eqn:BoundDerivw})

\begin{equation}
\begin{array}{l}
| X_{2} | + | X_{3}| + | X_{4} | \lesssim A  \cdot
\end{array}
\nonumber
\end{equation}
We then estimate $|X_{5}|$ and $|X_{6}|$. To this end we proceed similarly as in (\ref{Eqn:ErrorFPhi}) by dividing into the
cases $ \{ \phi w^{2} \gg 1 \} $ and $ \{ \phi w^{2} \lesssim  1 \} $ for all $ \tau \in [s_{1},s_{2}] $ to get

\begin{equation}
\begin{array}{l}
|X_{5}| + |X_{6}| \lesssim e^{-s} + \frac{1}{s \log(s)} \int_{s_{1}}^{s_{2}} \int_{B(O,1)}  e^{-\frac{2(p+1) \tau}{p-1}}
(log(\tau))^{\frac{2a}{p-1}} F(\phi w ) \; \rho  dy  d \tau \lesssim B  \cdot
\end{array}
\nonumber
\end{equation}
Now we prove the following claim: \\
\\
\underline{Claim}: $ A \lesssim  B + 1 $. \\
\\
Indeed by dividing again into the cases $ \{ \phi w^{2} \gg 1 \} $ and $ \{ \phi w^{2} \lesssim  1 \} $
for all $ s \in [s_{1},s_{2}] $ and by using the estimate
$ g (\phi w) \gtrsim \log^{a}(\tau) $ in the region $ \{ \phi w^{2} \gg 1 \}$, we see that

\begin{equation}
\begin{array}{l}
\int_{s_{1}}^{s_{2}} \int_{B(O,1)} |w|^{p+1} \; \rho dy d \tau \lesssim 1 + \int_{s_{1}}^{s_{2}}
\int_{B(O,1) \cap \{ \phi w^{2} \gg 1 \} } \frac{1}{\log^{a}(\tau)} |w|^{p+1} g(\phi w) \; \rho dy d \tau
\end{array}
\nonumber
\end{equation}
From the estimate above  and (\ref{Eqn:ApproxPot}) we see that the claim holds.  \\
\\
Hence we can replace $A$ with $B + 1$ in `` $\lesssim A $ ''  of  all the inequalities above where `` $\lesssim A $ '' appears. \\
\\
Finally we estimate the boundary terms of (\ref{Eqn:CombinEq}) for some particular values of $s_{1}$ and $s_{2}$.
The mean value shows that there exist $s_{1} \in [s-1,s]$  and  $s_{2} \in [s,s+1]$ such that

\begin{equation}
\begin{array}{l}
\int_{s-1}^{s} \int_{B(O,1)} \frac{(\partial_{\tau} w)^{2}}{1- |y|^{2}} \rho \; dy d \tau = \int_{B(O,1)}
\frac{( \partial_{\tau} w)^{2}(s_{1})}{1- |y|^{2}} \; \rho dy \; \text{and} \\
\int_{s}^{s+1} \int_{B(O,1)} \frac{( \partial_{\tau} w)^{2}}{1- |y|^{2}} \rho \; dy d \tau = \int_{B(O,1)} \frac{( \partial_{\tau} w)^{2}(s_{1})}{1- |y|^{2}} \; \rho  dy  \cdot
\end{array}
\nonumber
\end{equation}
Hence using also (\ref{Eqn:BoundDerivw}), the inequality $ab \leq \frac{a^{2}}{2} + \frac{b^{2}}{2}$, and (\ref{Eqn:BoundMassw}), we see that

\begin{equation}
\begin{array}{l}
\left| \left[  \int_{B(O,1)} \left( w \partial_{\tau} w + \frac{5-p}{2(p-1)} w^{2} \right) \rho \; dy  \right]_{s_{1}}^{s_{2}} \right| \lesssim B \cdot
\end{array}
\nonumber
\end{equation}

We can now conclude from all the estimates above that (\ref{Eqn:BoundAverPotNrj}) holds.


% COMMENTS

\appendix
\section{Estimates of functions}

In this section we prove some pointwise estimates for $F$ and other functions. \\
\\
Let $F_{1}$ be the function defined below

\begin{equation}
\begin{array}{ll}
F_{1}(x) := - \frac{2a}{(p+1)^{2}} |x|^{p+1} \frac{\log^{a-1} \left( \log(10 + x^{2}) \right)}{\log(10+x^{2})} \cdot
\end{array}
\label{Eqn:DefF1}
\end{equation}
Let $F_{2}$ be the function such that

\begin{equation}
\begin{array}{l}
F(x) = \frac{x f(x)}{p+1} + F_{1}(x) + F_{2}(x) \cdot
\end{array}
\label{Eqn:DecompF}
\end{equation}

We prove the following lemma:

\begin{lem}
There exists a large constant $A > 0$ such that if $ |u| \geq A $ then the estimates below hold:

\begin{equation}
\begin{array}{l}
F(u) \approx |u|^{p+1} \log^{a} \left( \log (10 + u^{2}) \right), \; \text{and} \\
\\
% \approx \frac{|u|^{p+1} } \log^{a} \left( \log (10+ u^{2}) \right), \; \text{and} \\
F_{2}(u) \lesssim  |u|^{p+1} \frac{ \log^{a-1} \left( \log(10 + u^{2}) \right)}{\log^{2}(10+u^{2})} %\frac{u f(u)}{\log^{a} \left( \log (10 + u^{2}) \right)} \cdot
\end{array}
\label{Eqn:ApproxPot}
\end{equation}

\end{lem}


\begin{proof}

Let $u \in \mathbb{R}$. Let $ I :=  \frac{2a}{p+1} \int_{0}^{u} \frac{|x|^{p+1} x}{ (10 + x^{2}) \log(10 + x^{2})}  \log^{a-1} \left( \log (10 + x^{2} ) \right) \; dx $.
An integration by parts shows that

\begin{equation}
\begin{array}{l}
\int_{0}^{u} |x|^{p-1} x \log^{a} \left( \log( 10+ x^{2}) \right) \; dx = \frac{|u|^{p+1}}{p+1} \log^{a} \left( \log(10+ u^{2}) \right) - I \cdot
\end{array}
\nonumber
\end{equation}
Observe that for $ |x| \leq |u|$ we have  $ \left|  \frac{|x|^{p+1} x}{ (10 + x^{2}) \log(10 + x^{2})} \log^{a-1}  \left( \log (10 + x^{2} ) \right) \right|
\lesssim 1 + |x|^{p} \frac{ \log^{a-1}  \left( \log (10 + x^{2} ) \right)}{\log(10 + x^{2})}  \lesssim
1 + |u|^{p} \frac{ \log^{a-1}  \left( \log (10 + u^{2} ) \right)}{\log(10 + u^{2})} $: this follows from the increasing property of
$r \rightarrow r^{p} \frac{ \log^{a-1}  \left( \log (10 + r^{2} ) \right)}{\log(10 + r^{2})} $  for $r \gg 1$. Hence by
integration we see that if $|u| \geq A $ then  $ I \lesssim 1 + |u|^{p+1} \frac{ \log^{a-1}  \left( \log (10 + u^{2} ) \right)}{\log(10 + u^{2})} $ and the first estimate of (\ref{Eqn:ApproxPot}) holds. \\
\\
We turn now our attention to the second estimate of (\ref{Eqn:ApproxPot}). We write $ I = J + K $ with

\begin{equation}
\begin{array}{l}
J := \frac{2a}{p+1} \int_{0}^{u} \frac{|x|^{p-1} x} {\log(10+x^{2})}  \log^{a-1} \left( \log (10 + x^{2} ) \right) \; dx
\end{array}
\nonumber
\end{equation}
We first estimate $J$. An integration by parts shows that there exists a
$\mathcal{C}^{0}([0,u])-$ function $h$ such that $|h(x)| \lesssim \frac{\log^{a-1} \left( \log (10 + x^{2}) \right)}{\log^{2}(10+x^{2})}$ for $|x| \geq 1$
and such that

\begin{equation}
\begin{array}{l}
J = \frac{2a}{(p+1)^{2}} \frac{|x|^{p+1} \log^{a-1} \left( \log(10+x^{2}) \right)}{\log(10 + x^{2})} - \int_{0}^{u} |x|^{p-1} x  h(x) \; dx \\
\end{array}
\nonumber
\end{equation}
Hence if $|u| \geq A$ then

\begin{equation}
\begin{array}{l}
\left| J - \frac{2a}{(p+1)^{2}} \frac{|u|^{p+1} \log^{a-1} \left( \log(10+ u^{2}) \right)}{\log(10 + x^{2})} \right| \lesssim
|u|^{p+1} \frac{ \log^{a-1} \left( \log(10 + u^{2}) \right)}{\log^{2}(10+u^{2})}
\end{array}
\nonumber
\end{equation}
We then write a (rough) estimate of $K$. We have $ K = - \frac{20a}{p+1} \int_{0}^{u} \frac{|x|^{p-1} x}{(10+ x^{2}) \log{(10 +x^{2})}} \log^{a-1} \left( \log (10 + x^{2} ) \right) \; dx $. If $ |x| \leq |u|$ then elementary considerations show that $ \left| \frac{ |x|^{p-1} x }{(10+ x^{2}) \log{(10 +x^{2})}} \log^{a-1} \left( \log (10 + x^{2} ) \right) \right| \lesssim 1 + |u|^{p} \frac{ \log^{a-1}  \left( \log (10 + u^{2} ) \right)}{\log^{2}(10 + u^{2})} $ . Hence by integration if $ |u| \geq A $ then

\begin{equation}
\begin{array}{l}
|K| \lesssim |u|^{p+1} \frac{ \log^{a-1} \left( \log(10 + u^{2}) \right)}{\log^{2}(10+u^{2})}
\end{array}
\nonumber
\end{equation}


\end{proof}


\section{Study of an ODE}

In this section we study the ODE associated to the equation
(\ref{equ}). \\
\\
We consider the following problem

\begin{equation}
\left\{
\begin{array}{l}
v^{''}(t) := |v(t)|^{p-1} v(t) g(t)  \\
v(0) := A > 0 \; \text{and} \; v^{'}(0) := B > 0
\end{array}
\right.
\label{Eqn:ProblemOdeV}
\end{equation}

\begin{lem}
There exists one positive solution to (\ref{Eqn:ProblemOdeV}). Moreover there exists $T < \infty$ such that this solution blows up in finite time $T$ and, moreover, it has the following asymptotic

\begin{equation}
\begin{array}{l}
v(t) \approx (T-t)^{-\frac{2}{p-1}} \log^{-\frac{a}{p-1}}(-\log(T-t)), \; \text{as} \; t \rightarrow T^- \cdot
\end{array}
\label{Eqn:Asympv}
\end{equation}

\label{Lem:OdeAsymp}
\end{lem}

\begin{proof}
The local existence and uniqueness follows from the Picard-Lindel\"of theorem (observe that $y \rightarrow |y|^{p-1} y
\log^{a} \left( \log (10 + y^{2}) \right) $ is locally Lipschitz).  Let $T$ be the maximal positive time of existence of the solution.\\
\\
We claim that $ T < \infty $. Indeed multiplying (\ref{Eqn:ProblemOdeV}) by $v^{'}(t)$ and integrating we see that there exists
a constant $C \in \mathbb{R}$ such that

\begin{equation}
\begin{array}{l}
(v^{'}(t))^{2} = 2F(v(t)) + C
\end{array}
\nonumber
\end{equation}
We claim that $v$ is positive on $[0, \infty)$. If not introducing $ t_{0} := \min \left\{ t \in [0, \infty) : v(t) \leq 0 \right\} $ we see that
for $t < t_{0}$, $ v(t) > 0 $; hence, using also (\ref{Eqn:ProblemOdeV})),  $v^{'}(t) > 0$, which contradicts $v(t_{0}) =0$. Hence
using again  (\ref{Eqn:ProblemOdeV})), we see that $ v^{'}(t) > 0 $ for $t \in [0, \infty)$. Hence

\begin{equation}
\begin{array}{l}
v^{'}(t)= \sqrt{2 F(v(t)) + C} \cdot
\end{array}
\label{Eqn:DerivPrime}
\end{equation}
Hence integrating (\ref{Eqn:DerivPrime}) from $0$ to $t$ we get

\begin{equation}
\int_{0}^{t} \frac{v^{'}(t)}{\sqrt{2 F(v(t)) + C}} \; ds = t \cdot
\label{Eqn:EqualContr}
\end{equation}
We see from (\ref{Eqn:ApproxPot}) that $ \int_{0}^{t} \frac{v^{'}(t)}{\sqrt{2 F(v(t)) + C}} \; ds $ is bounded as $t \rightarrow \infty$. This
contradicts (\ref{Eqn:EqualContr}). Hence $T < \infty$. \\
\\
The blow-up criterion (i.e $ |v(t)| + |v^{'}(t)| \rightarrow \infty $ as $ t \rightarrow T^{-}$), the increasing and positive properties of $v$ and $v^{'}$,
and (\ref{Eqn:DerivPrime}) combined with (\ref{Eqn:ApproxPot}) show that $ v(t) \rightarrow \infty $ as $ t \rightarrow T^{-}$. Integrating (\ref{Eqn:DerivPrime}) from $t$ to $T$ we see that

\begin{equation}
\begin{array}{l}
T - t = \int_{t}^{\infty} \frac{v^{'}(s)}{\sqrt{F(v(s)) +C}} \; ds =  \int_{v(t)}^{\infty} \frac{dy}{\sqrt{F(y) + C}} \cdot
\end{array}
\label{Eqn:EstTmint}
\end{equation}
Hence for $t < T $ close enough to $T$ we get

\begin{equation}
\begin{array}{l}
\int_{v(t)}^{\infty} \frac{dy}{y^{\frac{p+1}{2}+}} \; dy  \leq T - t \leq \int_{v(t)}^{\infty} \frac{dy}{y^{\frac{p+1}{2}-}} \; dy \cdot
\end{array}
\nonumber
\end{equation}
Hence

\begin{equation}
\begin{array}{l}
\frac{1}{(T-t)^{\frac{p-1}{2}+}} \lesssim  v(t) \lesssim \frac{1}{(T - t)^{\frac{p-1}{2}-}} \cdot
\end{array}
\nonumber
\end{equation}
Hence

\begin{equation}
\begin{array}{l}
\log \left( \log \left( 10 + v^{2}(t) \right) \right) \approx \log \left( - \log (T-t) \right) \cdot
\end{array}
\nonumber
\end{equation}
Hence, plugging the estimate above in (\ref{Eqn:EstTmint}) we see that (\ref{Eqn:Asympv}) holds.

\end{proof}

\section{Local well-posedness theory for
\eqref{equ}-\eqref{pert2}}

\label{Sec:LwpH1L2}

We recall the local well-posedness theory of solutions of
\eqref{equ}-\eqref{pert2}
%(\ref{Eqn:WaveFoc})
with data  $ \left( u(0),\partial_{t} u(0) \right) := (u_{0},u_{1}) \in H^{1}_{loc} \times L^{2}_{loc} (\mathbb{R}^{N})$.  \\

First we recall the localized energy estimates (\ref{Eqn:LocNrjEst})
(see e.g. Shatah and Struwe \cite{SSnyu98}).
%(see e.g \cite{shatahstruwe}).
Let $u$ be a solution of $\partial_{tt}
u - \triangle u = F $ with data $\left( u(0),\partial_{t} u(0)
\right):= (u_{0},u_{1})$. Let $t > 0$. Then we get (formally) the
localized energy estimates:
% (see e.g. Shatah and Struwe \cite{SSnyu98}), i.e.
% %(see e.g \cite{shatahstruwe}), i.e

\begin{equation}
\begin{array}{l}
\left\| \left( u(t), \partial_{t} u(t) \right) \right\|_{H^{1} \times L^{2} (|x| < R)} \lesssim  \| (u_{0},u_{1}) \|_{H^{1} (|x| < R + t) \times L^{2} (|x| < R + t|)} +
\| F \|_{L_{t}^{1} L_{x}^{2} \left( (t ^{'}, x): \, t^{'} \in [0,t], |x| < R + t^{'} \right)}
\end{array}
\label{Eqn:LocNrjEst}
\end{equation}
Let $t_{0} > 0$ (resp. $C^{'} > 0 $) be a constant small enough (resp. large enough) such that all the statements below are true.
Let $x_{0} \in \mathbb{R}^{N}$ and $B(x_{0},R)$ be the closed ball with center $x_{0}$ and radius $R$. Let $ K := \left\{ (t,x):  0  \leq  t  <  t_{0}, \, |x - x_{0}| <  t_{0} - t  \right\}$ and
$ \mathcal{X} := \left\{ v: \, v \in \mathcal{C}_{t} H^{1} (K) \cap \mathcal{C}^{1}_{t} L_{x}^{2} (K) \; \text{and} \; \| v \|_{X} \leq  C^{'} \| (u_{0},u_{1}) \|_{H^{1} ( B(x_{0},t_{0}) )  \times L^{2} ( B(x_{0},t_{0})) }   \right) $ be the Banach space endowed with the norm \\
$ \| v \|_{\mathcal{X}} := \sup \limits_{t \in [0,t_{0})} \sup \limits_{ t_{0} - t > R > 0} \max \left( \| v(t) \|_{H^{1}(B(x_{0},R))}, \| \partial_{t} v (t)\|_{L^{2}(B(x_{0},R))} \right) $.  \\

\begin{rem}
We say that $ v \in \mathcal{C}_{t} L_{x}^{2} (K) $ if and only if for all $ \bar{t} \in [ 0, t_{0} ) $ and for all open ball $ B(x_{0},R) \subsetneq
B \left( x_{0}, t_{0} - \bar{t} \right)$  we have $ \lim \limits_{t \rightarrow \bar{t}} \left\| v(t) -  v(\bar{t}) \right\|_{L^{2}( B(x_{0},R))} = 0 $. A similar definition holds for $ u \in \mathcal{C}_{t} H^{1}(K) $.
\end{rem}
Let $\Psi$ be defined by the formula below:
\begin{equation}
\begin{array}{l}
u \in \mathcal{X}   \rightarrow \Psi(u), \; \text{with} \; \Psi(u)(t) :=  \partial_{t} \mathcal{R}(t) * u_{0} + \mathcal{R}(t) * u_{1} + \int_{0}^{t} \mathcal{R}(t-s) * f(u(s)) \; ds
\end{array}
\label{Eqn:DefPsi}
\end{equation}
Here $\mathcal{R}$ denotes the fundamental solution, that is
$\mathcal{R}$  is the solution of
\eqref{equ} with zero right-hand side and
%(\ref{Eqn:WaveFoc})
initial data
$ \left( \mathcal{R}(0),\partial_{t} \mathcal{R}(0) \right) := (0,\delta)$ ( Here $ \delta $ is the standard delta function, i.e $\delta(\phi) := \phi(0)$ for every test function
$\phi$ ).  It is well-known (see e.g \cite{FollPDEsBook}) that if $N=1$ then  $ \langle \mathcal{R}(t), \phi \rangle := \frac{1}{2} \int_{-t}^{t} \phi \; dx $ and for $N > 1$

\begin{equation}
\begin{array}{l}
N = 2k, \; k \in \{ 1,.2,... \}: \;  \mathcal{R}(t)  := \frac{1} {1 \times 3 .... \times (N-1)}  \left[ t^{-1} \partial_{t} \right]^{\frac{N-2}{2}} t^{N-2} T_{t}, \; \text{and} \\
N = 2 k + 1, \;  k \in \{ 1, 2,...\}: \;  \mathcal{R}(t) := \frac{1}{1 \times 3... \times (N-2)} \left[ t^{-1} \partial_{t} \right]^{\frac{N-3}{2}} \left[ t^{N-2} \Sigma_{t} \right]
\end{array}
\label{Eqn:FormulaR}
\end{equation}
Here  $ \langle T_{t}, \phi \rangle :=  \frac{1}{\omega_{N+1}} \int_{|y| < 1} \frac{\phi(ty)}{\sqrt{1 - |y|^{2}}} \; dy $ and
$ \langle \Sigma_{t}, \phi \rangle :=  \frac{1}{\omega_{N}} \int_{|y|=1} \phi(ty)  d \sigma(y) $.   In particular  if $N=2$ then
$ \langle R(t), \phi \rangle := \frac{t}{2 \pi} \int_{|y| < 1} \frac{\phi(ty)}{\sqrt{1- |y|^{2}}} \; dy $ and if $N=3$
then  $ \langle R(t), \phi \rangle := \frac{1}{4 \pi}  \int_{S^{2}}  \phi(ty) \; d \sigma(y) $.  \\
Let $(w,w_{1},w_{2}) \in \mathcal{X}^{3}$. \\
From (\ref{Eqn:LocNrjEst}), the embedding $L^{2p} ( B(0,t_{0})) \hookrightarrow H^{1} ( B(0,t_{0})) $, and the estimate $ g(x) \lesssim 1 + |x|^{0+} $ show that

\begin{equation}
\begin{array}{ll}
\| \Psi(w) \|_{L_{t}^{\infty} H^{1}(K) \times L_{t}^{\infty} L_{x}^{2}(K)} & \lesssim \left\| (u_{0},u_{1}) \right\|_{H^{1} ( B(0,t_{0}) ) \times L^{2}  ( B(0,t_{0}))} + \| f(w) \|_{L_{t}^{1} L_{x}^{2}(K)} \\
& \lesssim \left\| (u_{0},u_{1}) \right\|_{H^{1} ( B(0,t_{0}) ) \times L^{2}  ( B(0,t_{0}))} +
t_{0} \left( \| w \|^{p}_{L_{t}^{\infty} L_{x}^{2p} (K)} +
\| w \|^{p+}_{L_{t}^{\infty} L_{x}^{2p+} (K)} \right)  \\
& \leq C^{'} \left\| ( u_{0},u_{1} ) \right\|_{H^{1} (B(0,t_{0})) \times L^{2} (B(0,t_{0}))} \cdot
\end{array}
\label{Eqn:ExpNonlin}
\end{equation}
We then claim that $ \Psi(w) \in  \mathcal{C}_{t} H^{1} (K) \cap
\mathcal{C}^{1}_{t} L_{x}^{2} (K) $. Translating in space and in time
if necessary we may assume
without loss of generality
%WLOG
that $x_{0}=0$ and $\bar{t}=0 $. \\
We first prove that $w_{l}$, solution of the linear wave equation with
data $ \left( u_{l}(0),\partial_{t} u_{l}(0) \right) :=
(u_{0},u_{1})$, lies in $ \mathcal{C}_{t} H^{1} (K) \cap
\mathcal{C}^{1}_{t} L_{x}^{2} (K)$. If $u_{0}$ and $u_{1}$ are smooth
functions (say, in $\mathcal{C}^{\infty}(B(O,R))$), then it is clear from (\ref{Eqn:FormulaR}) and elementary considerations (such as the Minkowski inequality for integrals and the dominated convergence theorem) that $\mathcal{P}(w_{l})$ holds with $ \mathcal{P}(v): \; \lim \limits_{t \rightarrow \bar{t}} \left\| v(t) -  v(\bar{t}) \right\|_{L^{2}( B(O,R))} = 0 $. In the general case $(u_{0},u_{1}) \in H^{1}_{loc} \times L^{2}_{loc}(\mathbb{R}^{N})$,  $\mathcal{P}(u_{l})$ also holds since we can find smooth functions $\left( u_{0,\epsilon}, u_{1,\epsilon} \right)$
within a $ H^{1} (B(O, R+ t_{0})) \times  L^{2}( B(O, R +t_{0}) ) - $ distance of $\epsilon > 0 $ from $(u_{0},u_{1})$
(with $\epsilon$ that can be chosen arbitrary small), then use (\ref{Eqn:LocNrjEst}) to the solution of the linear wave equation with data
$\left( u_{0}- u_{0,\epsilon},u_{1} - u_{1,\epsilon} \right)$, and then use $\mathcal{P} \left( u_{l,\epsilon} \right)$ with $u_{l,\epsilon}$ solution of the linear wave equation with data $ \left( u_{0,\epsilon}, u_{\epsilon} \right)$. \\
Next we consider  $u_{nl}$, solution of  $ \partial_{tt} w - \triangle w  = f(u) $ with data $ \left( u_{nl}(0),\partial_{t} u_{nl}(0) \right) := (0,0)$. From (\ref{Eqn:LocNrjEst}), similar steps as (\ref{Eqn:ExpNonlin}), and the dominated convergence theorem,  we see that $ \mathcal{P} \left( u_{nl} \right)$ holds. \\
The fundamental theorem of calculus and the Holder inequality  show that

\begin{equation}
\begin{array}{ll}
\| \Psi(w_{2}) - \Psi(w_{1}) \|_{X} & \lesssim   \left\| f(w_{2}) - f(w_{1}) \right\|_{L_{t}^{1} L_{x}^{2}(K)} \\
& \lesssim t_{0} \| w_{2} - w_{2} \|_{L_{t}^{\infty} L_{x}^{2p}(K)} \sum \limits_{ \bar{w} \in \{ w_{1}, w_{2} \} }
\left( \| \bar{w} \|^{p-1}_{L_{t}^{\infty} L_{x}^{2p}(K)} + \| \bar{w} \|^{(p-1)+}_{L_{t}^{\infty} L_{x}^{2p+} (K)}  \right) \cdot
\end{array}
\nonumber
\end{equation}
Hence $\Psi$ is contraction and we can apply the fixed point argument to conclude that there exist a unique $u \in \mathcal{X}$ such that $u = \Psi(u)$.

\section{Influence domain, maximal influence domain}
\label{Sec:InflDomain}

In this section we recall the notion of influence domain and maximal influence domain. \\
Recall (see \cite{Apndeta95})
%(see \cite{alinhac})
\footnote{There are slight differences between the definition we give
  and the one in
\cite{Apndeta95}.
  %\cite{alinhac}.
  We refer to Section $2$ in Chapter $5$ of this book.}
  %\cite{alinhac}.}
that a set $\Omega$ is an influence domain if it is an open set of $\{ t \geq 0 \}$ and if $ (\bar{x},\bar{t}) \in \Omega $ implies that
$K(\bar{x},\bar{t}) := \left\{ (t,x) :  0 \leq  t <  \bar{t} - |x- \bar{x}| \right\} \subset \Omega $. We define
$ \q D $, the maximal influence domain, to be the union of the influence domains $ \Omega $ such that there exists a solution $ u \in  \mathcal{C}_{t} H^{1} (\Omega) \cap \mathcal{C}^{1}_{t} L_{x}^{2} (\Omega) $ with data $ \left( u(0), \partial_{t} u(0) \right) := (u_{0},u_{1}) \in H^{1}_{loc} \times L^{2}_{loc} (\mathbb{R}^{N}) $ that satisfies $ u = \Psi(u) $.

\begin{rem}
We say that $ v \in \mathcal{C}_{t} L_{x}^{2} (\Omega) $ if and only for all $\bar{t}$ such that $ \{ t = \bar{t} \}  \cap  \Omega  \neq \emptyset $ and for all compact set
$ \mathcal{K} $ such that $ \bar{t} \in \mathcal{K} \subset \{ t = \bar{t} \}  \cap  \Omega $ we have $ \lim \limits_{t \rightarrow \bar{t}} \left\| v(t) - v(\bar{t}) \right\|_{L^{2}(\mathcal{K})} = 0 $. A similar definition holds for $ v \in \mathcal{C}_{t} H^{1}(\mathcal{K}) $.
\end{rem}
We make the elementary observation that if a cone is an influence domain, then a subcone of this cone is also an
influence domain. Hence, using also Appendix \ref{Sec:LwpH1L2} and the definition of
 $ \q D $, we see that there exists $T: = T(x) > 0 $  such that we can write
 $ \q D =  := \left\{ (t,x):  0 \leq t < T(x) \right\}$. Moreover there are two options. Either there exists $x_{0} \in \mathbb{R}^{N}$ such that $T(x_{0}) =  \infty$  and, in this case, $\q D = \left\{ t \geq 0 \right\}$; or the graph $ x \rightarrow T(x)$ is $1-$Lipschitz.





%\begin{rem}
%The proof of Theorem \ref{thm:main} also shows that
%$ \| u \|_{L_{t}^{\frac{2(n+2)}{n-2}} L_{x}^{\frac{2(n+2)}{n-2}}(\mathbb{R}^{+})} \leq \bar{C}^{\bar{C}^{\delta^{-\frac{1}{2}}}} $.
%\end{rem}


%Now we explain the main ideas of this paper and how it is organized. In Section \ref{Sec:Thmmain} we prove the main result of this paper, i.e Theorem \ref{thm:main}. The
%proof relies upon the following estimate of $\| u \|_{L_{t}^{\frac{2(n+2)}{n-2}} L_{x}^{\frac{2(n+2)}{n-2}} }$ on an arbitrarily long-time interval $J$, assuming that
%an \textit{a priori} bound of some norms at $\tilde{H}^{k}-$ regularity and a smallness condition hold
%\footnote{For a definition of $Q(J,u)$ and $\tilde{h}$ we refer to Section \ref{Sec:Prelim} and to Section \ref{Sec:ProofProp}.}:


%\section{Appendixes}



%\section{Appendix $B$}

%In this appendix we prove Proposition \ref{Prop:NonlinFracSmooth1}. We also point out an issue in \cite{triroyrad} and we explain
%how Proposition \ref{Prop:NonlinFracSmooth2} can resolve this issue.

%\subsection{Proof of Proposition \ref{Prop:NonlinFracSmooth1}}

%We recall the usual product rule for fractional derivatives

%\begin{equation}
%\begin{array}{ll}
%\| D^{\alpha_{1}} (fg) \|_{L^{q}} & \lesssim \| D^{\alpha_{1}} f \|_{L^{q_{1}}} \| g \|_{L^{q_{2}}} + \| f \|_{L^{q_{3}}} \| D^{\alpha_{1}} g
%\|_{L^{q_{4}}}
%\end{array}
%\label{Eqn:FracProd}
%\end{equation}
%and the usual Leibnitz rule for fractional derivatives :

%\begin{equation}
%\begin{array}{ll}
%\| D^{\alpha_{2}} H(f) \|_{L^{q}} & \lesssim \| \tilde{H}(f) \|_{L^{q_{1}}}  \| D^{\alpha_{2}} f \|_{L^{q_{2}}}
%\end{array}
%\label{Eqn:DerivComp}
%\end{equation}
%if $H$ is $C^{1}$ and it satisfies $ \tau \in [0,1]: \, \left| H^{'}\left( \tau x + (1- \tau) y \right) \right| \lesssim
%\tilde{H}(x) + \tilde{H}(y)$, $ 0 \leq \alpha_{1} < \infty$, $ 0 < \alpha_{2}  \leq 1 $, $ (q,q_4) \in (1, \infty)^{2}$, $q_{3} \in (1,\infty]$,
%$(q_1,q_2) \in (1,\infty) \times (1,\infty] $ in (\ref{Eqn:FracProd}), $(q_1,q_2) \in (1,\infty] \times (1,\infty) $ in (\ref{Eqn:DerivComp}),
%$\frac{1}{q}= \frac{1}{q_{1}} + \frac{1}{q_{2}}$, and
%$\frac{1}{q}= \frac{1}{q_3} + \frac{1}{q_4}$ (see e.g. Christ-Weinstein \cite{christwein}, Taylor
%\cite{taylor} and references in \cite{taylor}). \\
%\\
%We prove Proposition \ref{Prop:NonlinFracSmooth1} by induction on $k'$, in a similar fashion as the proof of Proposition \ref{Prop:NonlinFracSmooth1}
%in \cite{triroyrad}. \\
%Let $k'=2$. Then

%\begin{equation}
%\begin{array}{ll}
%\left\| D^{2- 1 + \alpha} \left( G(f,\bar{f}) F(|f|^{2}) \right)  \right\|_{L^{r}}  & \approx
%\left\| D^{\alpha}  \nabla \left( G(f,\bar{f}) F(|f|^{2}) \right)   \right\|_{L^{r}} \\
%& \lesssim \left\| D^{\alpha} \left( \partial_{z} G(f,\bar{f}) \nabla f  F( |f|^{2})  \right) \right\|_{L^{r}}
%+ \left\| D^{\alpha} \left( \partial_{\bar{z}} G(f,\bar{f}) \overline{\nabla f} \nabla  F( |f|^{2})  \right) \right\|_{L^{r}} \\
%& + \left\| D^{\alpha} \left( F^{'}(|f|^{2}) 2 \Re( \bar{f} \nabla f) G(f,\bar{f}) \right)  \right\|_{L^{r}} \\
%& \lesssim A_{1} + A_{2} + A_{3}
%\end{array}
%\end{equation}
%We estimate $A_{1}$. We have

%\begin{equation}
%\begin{array}{ll}
%A_{1} & \lesssim  \left\| D^{\alpha} \left( \partial_{z} G(f,\bar{f}) F(|f|^{2}) \right) \right\|_{L^{r_{4}}}
%\| D f  \|_{L^{r_{5}}} +  \left\| \partial_{z} G( f,\bar{f} ) F (|f|^{2}) \right\|_{L^{?}}  \| D^{2-1 + \alpha} \|_{L^{?}} \\
%& \lesssim \| F(|f|^{2}) \|_{L^{\infty}} \| D^{\alpha} f \|_{L^{r_{2}}}
%\end{array}
%\end{equation}
%with $\frac{1}{r} =  \frac{1}{r_{4}} + \frac{1}{r_{5}}$,

%By complex interpolation we have

%\begin{equation}
%\begin{array}{ll}
%\| D f \|_{L^{?}} & \lesssim \| f \|^{\theta_{1}}_{L^{?}} \| D^{2-1 + \alpha} f \|^{1 - \theta_{1}}_{L^{?}} \\
%\| D^{\alpha} f \|_{L^{?}} & \lesssim \| f \|^{\theta_{2}}_{L^{?}}  \| D^{2-1 + \alpha} f \|^{1 - \theta_{2}}_{L^{?}}
%\end{array}
%\nonumber
%\end{equation}
%$A_{2}$ is estimated similarly: therefore we omit the proof. We estimate $A_{3}$.

%\begin{equation}
%\begin{array}{ll}
%? & ?
%\end{array}
%\nonumber
%\end{equation}


%\section{proof of Theorem \ref{thm:main}, Case $2$}

%It is sufficient to prove that for $\gamma$ satisfying (\ref{Eqn:Ass4}), we cannot have $T_{+} = \infty $ and

%\begin{equation}
%\begin{array}{l}
%\| u(t) \|_{L^{2^{*}}} \leq A, \; t \in [0,\infty). \\
%\end{array}
%\label{Eqn:AssBound}
%\end{equation}

%\begin{prop}{\textbf{`` Bound of $L_{t}^{\frac{2(n+2)}{2(n-2)}} L_{x}^{\frac{2(n+2)}{2(n-2)}}$ norm, $\| u_0 \|_{L^{2^{*}}} >
%\| W \|_{L^{2^{*}}}$. ''}}
%Let $u$ be an $\tilde{H}^{k}$ solution of (\ref{Eqn:BarelySchrod}) such that (\ref{Eqn:Ass0}), Case $1$, and (\ref{Eqn:AssBound}) hold.
%Assume that for all $t \in [0,\infty)$

%\begin{equation}
%\begin{array}{l}
%g(u(t)) - 1   \ll \delta  \cdot
%\end{array}
%\label{Eqn:Constg2}
%\end{equation}
%Then there exists a constant $C_{0} \gg 1$  such that if $\| u \|_{L_{t}^{\infty} \tilde{H}^{k}([0,\infty))} \leq M$ for some $M \gg 1$, then

%\begin{equation}
%\begin{array}{ll}
%\| u \|^{\frac{2(n+2)}{n-2}}_{ L_{t}^{\frac{2(n+2)}{n-2}} L_{x}^{\frac{2(n+2)}{n-2}} (J)} & \leq
%C_{0}^{C_{0}^{ \delta^{-\frac{1}{2}}g^{a_{n}+}(M)}}
%\end{array}
%\label{Eqn:BoundLong2}
%\end{equation}
%\label{prop:BoundLong2}
%\end{prop}

%\begin{lem}
%We have for all time $t \in J$

%\begin{equation}
%\begin{array}{ll}
%\tilde{K}(u(t)) & \leq  - \delta
%\end{array}
%\nonumber
%\end{equation}

%\end{lem}

%\begin{proof}
%Following the proof of Lemma \ref{lem:Mounoton} we get $\tilde{E}(u(t)) < (1- \delta) \tilde{E}(W)$. Now we see from the sharp Sobolev inequality that
%if $ \| u(t) \|_{L^{2^{*}}} > \| W \|_{L^{2^{*}}}$ then $\| \nabla u(t) \|_{L^{2}} > \| \nabla W \|_{L^{2}}$.



%\end{proof}



%\begin{equation}
%\begin{array}{ll}
%2m \; \partial_{t} \langle \partial_{x_j} \phi \left( \frac{|x|}{m} \right) u, -i \partial_{x_j} u  \rangle & =
%8 \int_{|x| \leq m} |\nabla u|^{2} - |u|^{2^{*}} g(|u(t)|) \; dx \\
%& - 4(n-2) \int_{|x| \leq m} \int_{0}^{|u|^{2}(t)} s^{\frac{n}{n-2}} \tilde{g}^{'}(s) \; ds \; dx + O(Y_m)
%\end{array}
%\nonumber
%\end{equation}
%In other words

%\begin{equation}
%\begin{array}{ll}
%2m \; \partial_{t} \langle \partial_{x_j} \phi \left( \frac{|x|}{m} \right) u, -i \partial_{x_j} u  \rangle & =
%8 K (\chi_{m} u)  - 4(n-2) \int_{|x| \leq m} \int_{0}^{|u|^{2}(t,x)} s^{\frac{n}{n-2}} \tilde{g}^{'}(s) \; ds \; dx + O(Y_m)
%\end{array}
%\nonumber
%\end{equation}
%Integrating by parts $F(u,\bar{u})$ once we see that

%\begin{equation}
%\begin{array}{ll}
%E(f)  \geq \frac{1}{2} \int_{\mathbb{R}^{n}} | \nabla f |^{2} \; dx  - \frac{1}{2^{*}} \int_{\mathbb{R}^{n}} |f|^{2^{*}}
%g(|f|) \; dx
%\end{array}
%\end{equation}
%Hence

%\begin{equation}
%\begin{array}{ll}
%K(\chi_{m} u) & \leq  2 E(\chi_{m} u)
%- 2 \left( \frac{1}{2} - \frac{1}{2^{*}} \right) \int_{\mathbb{R}^{n}} |\chi_{m}|^{2^{*}} g(|\chi_{m} u|) \; dx
%\end{array}
%\nonumber
%\end{equation}
%Hence we get  \color{red} ??: define $c$ \color{black}

%\begin{equation}
%\begin{array}{ll}
%K(\chi_{m} u) & \leq - c \left( 1 + \int_{\mathbb{R}^{n}} |\chi_{m} u|^{2^{*}} g(|\chi_{m} u|) \; dx \right)
%\end{array}
%\nonumber
%\end{equation}
%provided that we can prove that

%\begin{equation}
%\begin{array}{ll}
%K(\chi_{m} u) & \leq - c
%\end{array}
%\nonumber
%\end{equation}
%Let $I \subset [0,\infty)$ and $t \in I$. Clearly $ \| \nabla (\chi_{m} u(t)) \|_{L^{2}}^{2}
%\lesssim \| \nabla u(t) \|^{2}_{L^{2}} + \int_{|x| \sim m } \frac{|u(t)|^{2}}{|x|^{2}} \; dx \cdot $. Integrating the second term of the previous estimate on %$I$, summing  over $m$, we see from the Hardy inequality and the pigeonhole principle that there exists $C \gg 1$ such that
%$m \leq C \delta^{-1} $ and

%\begin{equation}
%\begin{array}{ll}
%\| \nabla (\chi_m u(t)) \|_{L^{2}}^{2} \; dt \leq (1 + o(\delta) )  \| \nabla u(t) \|^{2}_{L^{2}} \cdot
%\end{array}
%\nonumber
%\end{equation}
%Let $X := \int_{\mathbb{R}^{n}} F(u,\bar{u}) - F(\chi_{m} u , \chi_{m} \bar{u}) \; dx $. But

%\begin{equation}
%\begin{array}{ll}
%|X| & \int_{|x| \gtrsim m} \int_{0}^{|u(t)|} s^{\frac{n+2}{n-2}} g(s) \; dx \\
%& \lesssim g(M) \int_{|x| \gtrsim m} |u(t)|^{2^{*}} \; dx
%\end{array}
%\nonumber
%\end{equation}
%Now

%\begin{equation}
%\begin{array}{ll}
%\int_{|x| \gtrsim m} |u(t)|^{2^{*}} \; dx &  \lesssim  \| r^{\frac{n}{2} -1} u(t) \|^{2^{*} -2}_{L^{\infty}(|x| \gtrsim m)}
%\left\| \frac{u(t)} {|x|} \right\|^{2}_{L^{2}(|x| \gtrsim m)} \\
%& \lesssim \| \nabla u(t) \|^{\frac{2^{*}}{2}-1}_{L^{2}(|x| \gtrsim m)} \left\| \frac{u}{r} \right\|^{\frac{2^{*}}{2} + 1}_{L^{2}(|x| \gtrsim m)} \\
%& \ll \| \nabla u(t) \|^{2}_{L^{2}} \ll \| u(t) \|^{2^{*}}_{L^{2^{*}}}
%\end{array}
%\nonumber
%\end{equation}
%Let $\tilde{J}:= J_{i_0} \cup... \cup J_{i_1}$ be a sequence of consecutive consecutive unexceptional intervals.
%Observe that from Lemma \ref{Lem:}

%\begin{equation}
%\begin{array}{ll}
%Number \; of \; \tilde{J}s & \lesssim \eta_{2}^{-1} \sup_{t \in [0,\infty)} \| \nabla u(t) \|^{\frac{2(n+2)}{n-2}}_{\dot{H}^{1}} \\
%& \lesssim \eta_{2}^{-1} A^{\frac{2n(n+2)}{(n-2)^{2}}}
%\end{array}
%\nonumber
%\end{equation}
%We claim that one one of these intervals $J_{l} \in \tilde{J}$ is large. More precisely

%From we see that there exists $m := |I|^{\frac{1}{2}}$ such that

%\begin{equation}
%\begin{array}{ll}
%1 & 1
%\end{array}
%\nonumber
%\end{equation}

%\begin{equation}
%\begin{array}{ll}
%\left| \int_{|x| \geq m} \frac{|u(t)|^{2}}{|x|^{2}} \; dx - \int_{|x| \geq m} \frac{|u_0|^{2}}{|x|^{2}} \; dx  \right|
%& \lesssim \int_{I} \int_{|x| \geq m} \frac{m^{2}}{|x|^{2}} \left( |\nabla u(t)|^{2} + \frac{|u(t)|^{2}}{|x|^{2}} \right) \; dx \; dt
%\end{array}
%\nonumber
%\end{equation}

%\begin{equation}
%\begin{array}{ll}
%E(\chi_{m} u(t) ) & \leq E(u) + O (Y_{m}) + o \left( g(M) \| u(t) \|^{2^{*}}_{L^{2^{*}}} \right) \\
%\end{array}
%\end{equation}


%\begin{equation}
%\begin{array}{ll}
%K(\chi_{m} u(t)) & \leq 2(1- 2 \delta) \tilde{E}(W) + O (Y_m) + o \left( g(M) \| u(t) \|^{2^{*}}_{L^{2^{*}}} \right) \\
%& - 2 \left( \frac{1}{2} - \frac{1}{2^{*}} \right) \int_{\mathbb{R}^{n}} |\chi_{m} u(t)|^{2^{*}} g(|\chi_{m} u(t)|^{2^{*}}) \; dx
%\end{array}
%\nonumber
%\end{equation}


% \bibliography{../ref}
% \end{document}

\def\cprime{$'$} \def\cprime{$'$}
\providecommand{\bysame}{\leavevmode\hbox to3em{\hrulefill}\thinspace}
\providecommand{\MR}{\relax\ifhmode\unskip\space\fi MR }
% \MRhref is called by the amsart/book/proc definition of \MR.
\providecommand{\MRhref}[2]{%
  \href{http://www.ams.org/mathscinet-getitem?mr=#1}{#2}
}
\providecommand{\href}[2]{#2}
\begin{thebibliography}{10}

\bibitem{Apndeta95}
S.~Alinhac, \emph{Blowup for nonlinear hyperbolic equations}, Progress in
  Nonlinear Differential Equations and their Applications, vol.~17,
  Birkh\"auser Boston Inc., Boston, MA, 1995. \MR{96h:35109}

\bibitem{AMimrn01}
C.~Antonini and F.~Merle, \emph{Optimal bounds on positive blow-up solutions
  for a semilinear wave equation}, Internat. Math. Res. Notices (2001), no.~21,
  1141--1167. \MR{2002h:35027}

\bibitem{CZcpam13}
R.~C\^ote and H.~Zaag, \emph{Construction of a multisoliton blowup solution to
  the semilinear wave equation in one space dimension}, Comm. Pure Appl. Math.
  \textbf{66} (2013), no.~10, 1541--1581. \MR{3084698}

  \bibitem{FollPDEsBook}  G. Folland, \emph{Introduction to partial differential equations}, Princeton University Press, second edition, 1995.

\bibitem{HSjdde14}
M.~A. Hamza and O.~Saidi, \emph{The blow-up rate for strongly perturbed
  semilinear wave equations}, J. Dynam. Differential Equations \textbf{26}
  (2014), no.~4, 1115--1131. \MR{3285649}

\bibitem{HZjmaa20}
M.~A. Hamza and H.~Zaag, \emph{The blow-up rate for a non-scaling invariant
  semilinear wave equations}, J. Math. Anal. Appl. \textbf{483} (2020), no.~2,
  123652, 34. \MR{4037583}

\bibitem{HZna21}
\bysame, \emph{The blow-up rate for a non-scaling invariant semilinear wave
  equations in higher dimensions}, Nonlinear Analysis \textbf{212} (2021),
  112445.

\bibitem{HZnonl12}
M.A. Hamza and H.~Zaag, \emph{A {L}yapunov functional and blow-up results for a
  class of perturbed semilinear wave equations}, Nonlinearity \textbf{25}
  (2012), no.~9, 2759--2773.

\bibitem{HZdcds13}
\bysame, \emph{Blow-up results for semilinear wave equations in the
  super-conformal case}, Discrete Contin. Dyn. Syst. Ser. B \textbf{18} (2013),
  no.~9, 2315--2329.

\bibitem{Jmm79}
F.~John, \emph{Blow-up of solutions of nonlinear wave equations in three space
  dimensions}, Manuscripta Math. \textbf{28} (1979), no.~1-3, 235--268.
  \MR{80i:35114}

\bibitem{Ltams74}
H.~A. Levine, \emph{Instability and nonexistence of global solutions to
  nonlinear wave equations of the form {$Pu_{tt}=-Au+{\mathcal F}(u)$}}, Trans.
  Amer. Math. Soc. \textbf{192} (1974), 1--21. \MR{0344697}

\bibitem{LTpams01}
H.~A. Levine and G.~Todorova, \emph{Blow up of solutions of the {C}auchy
  problem for a wave equation with nonlinear damping and source terms and
  positive initial energy}, Proc. Amer. Math. Soc. \textbf{129} (2001), no.~3,
  793--805. \MR{1792187 (2001k:35212)}

\bibitem{MZajm03}
F.~Merle and H.~Zaag, \emph{Determination of the blow-up rate for the
  semilinear wave equation}, Amer. J. Math. \textbf{125} (2003), 1147--1164.

\bibitem{MZimrn05}
\bysame, \emph{Blow-up rate near the blow-up surface for semilinear wave
  equations}, Internat. Math. Res. Notices (2005), no.~19, 1127--1156.

\bibitem{MZma05}
\bysame, \emph{Determination of the blow-up rate for a critical semilinear wave
  equation}, Math. Annalen \textbf{331} (2005), no.~2, 395--416. \MR{MR2115461}

\bibitem{MZjfa07}
\bysame, \emph{Existence and universality of the blow-up profile for the
  semilinear wave equation in one space dimension}, J. Funct. Anal.
  \textbf{253} (2007), no.~1, 43--121. \MR{MR2362418}

\bibitem{MZcmp08}
\bysame, \emph{Openness of the set of non characteristic points and regularity
  of the blow-up curve for the $1$ d semilinear wave equation}, Comm. Math.
  Phys. \textbf{282} (2008), 55--86.

\bibitem{MZsnp09}
\bysame, \emph{On characteristic points at blow-up for a semilinear wave
  equation in one space dimension}, Singularities in Nonlinear Problems, Kyoto,
  2009.

\bibitem{MZxedp10}
\bysame, \emph{Isolatedness of characteristic points for a semilinear wave
  equation in one space dimension}, S{\'e}minaire sur les {\'E}quations aux
  D{\'e}riv{\'e}es Partielles, 2009--2010, {\'E}cole Polytech., Palaiseau,
  2010, pp.~Exp.\ No.\ 11, 10p.

\bibitem{MZajm12}
\bysame, \emph{Existence and classification of characteristic points at blow-up
  for a semilinear wave equation in one space dimension}, Amer. J. Math.
  \textbf{134} (2012), no.~3, 581--648.

\bibitem{MZdmj12}
\bysame, \emph{Isolatedness of characteristic points for a semilinear wave
  equation in one space dimension}, Duke Math. J. \textbf{161} (2012), no.~15,
  2837--2908.

\bibitem{MZcmp15}
\bysame, \emph{On the stability of the notion of non-characteristic point and
  blow-up profile for semilinear wave equations}, Comm. Math. Phys. (2015),
  1--34.

\bibitem{MZtams16}
\bysame, \emph{Dynamics near explicit stationary solutions in similarity
  variables for solutions of a semilinear wave equation in higher dimensions},
  Trans. Amer. Math. Soc. \textbf{368} (2016), no.~1, 27--87. \MR{3413856}

\bibitem{MZsls17}
\bysame, \emph{Solution to the semilinear wave equation with a pyramid-shaped
  blow-up surface}, S\'{e}minaire {L}aurent {S}chwartz---\'{E}quations aux
  d\'{e}riv\'{e}es partielles et applications. {A}nn\'{e}e 2016--2017, Ed.
  \'{E}c. Polytech., Palaiseau, 2017, pp.~Exp. No. VI, 13. \MR{3790935}

\bibitem{MZcpam18}
\bysame, \emph{Blowup solutions to the semilinear wave equation with a stylized
  pyramid as a blowup surface}, Comm. Pure Appl. Math. \textbf{71} (2018),
  no.~9, 1850--1937. \MR{3847768}


\bibitem{Rapde09}
T.~Roy, \emph{Global existence of smooth solutions of a 3d log-log
  energy-supercritical wave equation}, Anal. PDE \textbf{2} (2009), no.~3,
  261--280 (English).

\bibitem{SSnyu98}
J.~Shatah and M.~Struwe, \emph{Geometric wave equations}, New York University
  Courant Institute of Mathematical Sciences, New York, 1998. \MR{2000i:35135}

\bibitem{Tna00}
G.~Todorova, \emph{Cauchy problem for a nonlinear wave equation with nonlinear
  damping and source terms}, Nonlinear Anal. \textbf{41} (2000), no.~7-8, Ser.
  A: Theory Methods, 891--905. \MR{1764026 (2001c:35160)}

\end{thebibliography}


%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%
\end{document}
%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{thebibliography}{1}

\bibitem{HZjhde12} M. A. Hamza and H. Zaag, \emph{A Lyapunov functional and blow-up results for a class of perturbation for semilinear wave equations}, J. Hyperbolic
Differ. Equ., 9m 195-211, 2012.

\bibitem{HZjmaa20} M. A. Hamza and H. Zaag, \emph{The blow-up rate for a non-scaling invariant semilinear wave equation}, J. Math. Anal. Appl. 483, no. 2, 34p. (2020).

\bibitem{HZna21} M.A. Hamza and H. Zaag. The blow-up rate for a non-scaling invariant semilinear wave equations in higher dimensions. Nonlinear Analysis 212 (2021), 112445.

\bibitem{MZajm03}  F. Merle and H. Zaag, \emph{Determination of the blow-up rate for the semilinear wave equation}, Amer. J. Math.,
125, 1147-1164, 2003.

\bibitem{MZimrn05} F. Merle and H. Zaag, \emph{Blow-up rate near the blow-up surface for semilinear wave equations}, Internat. Math.
Res. Notices., 19, 1127-1156, 2005.

%\bibitem{aubin} Aubin, \emph{Equations differentielles nonlineaires et probleme de Yamabe concernant la courbure scalaire}, J. Math.
%Pures. Appl. (9), 55, 1976, 3, 269-296.

%\bibitem{bourg} J. Bourgain, \emph{Global well-posedness of defocusing  critical NLS in the radial case}, JAMS 12 (1999), no. 1, 145-171.

%\bibitem{cazweiss} T. Cazenave, F.B. Weissler, \emph{The Cauchy problem for the critical nonlinear  Schr\"odinger equation in $H^{s}$}, Nonlinear Anal. 14 (1990), no. 10,
%    807-836.

%\bibitem{collkeelstafftaktao}  J. Colliander, M. Keel, G. Staffilani, H. Takaoka, T. Tao, \emph{Global well-posedness and scattering in the energy
%space for the critical nonlinear Schr\"odinger equation in $\mathbb{R}^{3}$}, Annals of Math. 167 (2008), 767-865.

%\bibitem{duymerle} T.~Duyckaerts and F.~Merle, \emph{Dynamic of threshold solutions for energy-critical NLS}, Geom. Funct. Anal. 18 (2009), no.~6,
%1787-1840.

%\bibitem{shihhsiwei} S. Hsi-Wei, \textit{Some results on scattering for log-subcritical and log-supercritical nonlinear wave equations}, Anal. PDE 6 (2013), no. 1, 1–24.

%\bibitem{foll} G. Folland, \emph{Real Analysis, Modern Techniques and Their Applications}
%\bibitem{foshi} D. Foschi, \emph{Inhomogeneous Strichartz estimates}, J. Hyperbolic. Differ. Equ. 2 (2005), 1-24.

%\bibitem{grill} M. Grillakis, \emph{On nonlinear Schr\"odinger equations}, Comm. Partial Differential Equations 25 (2000), no 9-10, 1827-1844

%\bibitem{keeltao} M. Keel and T. Tao, \emph{Endpoint Strichartz estimates}, Amer. Math. J. 120 (1998), 955-980.
%\bibitem{kenigmerle} C.Kenig and F.Merle, \emph{Nondispersive Radial Solutions To Energy Supercritical Non-Linear Wave Equations, With
%Applications}, preprint, 2008

%\bibitem{kenmer} C.~E.~Kenig and F.~Merle, \emph{Global well-posedness, scattering, and blow-up for the energy-critical, focusing, non-linear Schr\"odinger
%    equation in the radial case}, Invent. Math., 166 (2006), no. 3, 645-675.

%\bibitem{nakschlagkg} K. Nakanishi and W. Schlag, \emph{Global dynamics above the ground state energy for the focusing nonlinear Klein-Gordon equation}, J. Diff. Eq. 250 (2011), 2299-2333.

%\bibitem{roynak} K. Nakanishi and T. Roy, \textit{Global dynamics above the ground state for the energy-critical Schr\"odinger
%equation with radial data}, Communications on Pure and Applied Analysis, Volume 15, Issue 6, November 2016, 2023-2058.

%\bibitem{taylor} M. Taylor, \textit{Tools for PDE. Pseudodifferential Operators, Paradifferential Operators, and Layer Potentials},
%Mathematical Surveys and Monographs 81, AMS, Providence, R.I. 2000.

%\bibitem{yeh} J. Yeh, \textit{Real analysis, Theory of measure and integration}, World Scientific, 2014 ($3^{rd}$ edition).

%\bibitem{talenti} Talenti, Giorgio, \textit{Best constant in Sobolev inequality}, Ann. Mat. Pura Appl.,(4)  1976, 353-372.

%\bibitem{triroyrad} T. Roy, \textit{Scattering above energy norm of solutions of a loglog energy-supercritical Schr\"odinger
%equation with radial data}, Journal of Differential Equations, 250 (2011), no. 1, 292-319. Corrigendum, 264 (2018), no. 9, 6013-6024.

%\bibitem{triroyjensen} T. Roy, \textit{On Jensen-type inequalities for nonsmooth radial scattering solutions of a loglog
%energy-supercritical Schr\"odinger equation}, International Mathematics Research Notices, Volume 2020, Issue 8, April 2020, 2501–2541.

%\bibitem{triroysmooth} T. Roy, \textit{Global existence of smooth solutions to a $3D$
%loglog energy-supercritical wave equation}, Analysis and PDE, Vol 2, Num 3,
%261-280, 2009.

%\bibitem{taorad} T. Tao, \textit{Global well-posedness and scattering for the higher-dimensional energy-critical non-linear Schr\"odinger
%equation for radial data}, New York J. Math., 11, 2005, 57-80.

%\bibitem{taofoc} T. Tao, \emph{The Kenig-Merle scattering result for the energy-critical focusing NLS }, unpublished note.

%\bibitem{taolog} T. Tao, \textit{Global regularity for a logarithmically supercritical defocusing nonlinear wave equation
%for spherically symmetric data}, J. Hyperbolic Differ. Equ. 4 (2007), no. 2, 259-265.

\end{thebibliography}

\bibliography{../ref}

\end{document}



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%We have
%\begin{equation}
%\begin{array}{ll}
%\| v_{1,h} - v_{1} \|_{L_{t}^{\infty} L_{x}^{\frac{8n(n+2)}{(n-2)(8 + 6n - n^{2})}} ([t_{*} |I_{j}|, |I_{j}|  ]) } & \lesssim \| D v_{1}
%\|_{L_{t}^{\infty} L_{x}^{\frac{8n(n+2)}{(n-2)(8 + 6n - n^{2})}} ([t_{*} |I_{j}|, |I_{j}|  ]) } |h|
%\end{array}
%\end{equation}
%and
%\begin{equation}
%\begin{array}{ll}
%\| D v_{1} \|_{ L_{t}^{\infty} L_{x}^{\frac{8n(n+2)}{(n-2)(8 + 6n - n^{2})}} ([t_{*} |I_{j}|, |I_{j}|  ]) } & \lesssim \| D v_{1}
%\|^{}_{L_{t}^{\infty} L_{x}^{2} ([t_{*} |I_{j}|, |I_{j}|  ]) }   \| D v_{1} \|^{}_{L_{t}^{\infty} L_{x}^{\infty}}
%\end{array}
%\end{equation}
%By  Result \ref{res:res1}

%\begin{equation}
%\begin{array}{ll}
%\| D (|u|^{\frac{4}{n-2}} u g(|u|)) \|_{L_{s}^{\infty} L_{x}^{1} ([t_{1}, (t_{*} - \eta_{3})|I_{j}|])} & \lesssim \| D u \|_{L_{s}^{\infty}
%L_{x}^{2} ([t_{1}, (t_{*} - \eta_{3}) |I_{j}|])} \| u g^{\frac{n-2}{2n}} (|u|) \|^{\frac{4}{n-2}}_{L_{s}^{\infty} L_{x}^{\frac{2n}{n-2}}
%([t_{1}, (t_{*} - \eta_{3}) |I_{j}|]) }  \\
%& \left( E^{\frac{1}{2}} + g^{\frac{n-2}{n}} (
%\| u \|_{L_{t}^{\infty} L_{x}^{\frac{2n}{n-4}} ([t_{1}, (t_{*} - \eta_{3}) |I_{j}|])} ) \right) \\
%& \lesssim_{E} g^{\frac{n-2}{n}} (M)
%\end{array}
%\end{equation}
%and by the dispersive inequality (\ref{Eqn:DispIneq})

%\begin{equation}
%\begin{array}{ll}
%\| D v_{1} \|_{L_{t}^{\infty} L_{x}^{1} ([t_{*}|I_{j}|, |I_{j}|  ])}  & \lesssim \left\| \int_{t_{1}}^{(t_{*} - \eta_{3})|I_{j}|} \frac{1}{|t -
%s|^{\frac{n}{2}}} \| D ( |u(s)|^{\frac{4}{n-2}} u(s)  g(|u(s)|)) \|_{
%L_{x}^{1}}  \, ds \right\|_{L_{t}^{\infty} ( [t_{*}|I_{j}|, |I_{j}|]  ) } \\
%& \lesssim g^{\frac{n-2}{2}} (M) \eta_{3}^{\frac{2-n}{2}} |I_{j}|^{\frac{2 -n}{2}}
%\end{array}
%\end{equation}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%\begin{prop}{\textbf{``Estimates for the norm $\| D^{\theta} u \|_{L_{t}^{q} L_{x}^{q}}$'' }}
%Let $(\theta,q,\tilde{q})$ be such that $\frac{n-2}{2(n+2)} \leq \theta \leq 1$, $\frac{1}{q}:= \frac{\theta}{\frac{2(n+2)}{n}} + \frac{1
%-\theta}{\frac{2(n+2)}{n-2}} $ and $ \frac{1}{q} + \frac{1}{\tilde{q}} =\frac{n}{n+2} $. if $u$ is a solution of (\ref{Eqn:BarelySchrod}) then
%we have

%\begin{itemize}

%\item [$\bullet$] ( Inhomogeneous Strichartz estimates)

%\begin{equation}
%\begin{array}{ll}
%\| D^{\theta} u_{nl} \|_{L_{t}^{q} L_{x}^{q} (J)} & \lesssim \| D^{\theta} (|u|^{\frac{4}{n-2}} u g(|u|) ) \|_{L_{t}^{\tilde{q}^{'}}
%L_{x}^{\tilde{q}^{'}} (J)}
%\end{array}
%\end{equation}

%\item [$\bullet$] ( control of $\| \nabla u \|_{ L_{t}^{\frac{2(n+2)}{n}} L_{x}^{\frac{2(n+2)}{n}} (J)  }$ and
%lower bound of $\| D^{ \theta} u \|_{L_{t}^{q} L_{x}^{q} (J)} $ $ \Rightarrow $ lower bound of $ \| u  \|_{L_{t}^{} L_{x}^{} }  $   )

%\begin{equation}
%\begin{array}{ll}
%\| D^{\theta} u \|_{L_{t}^{q} L_{x}^{q} (J) } & \lesssim \| u \|^{1- \theta}_{L_{t}^{\frac{2(n+2)}{n-2}} L_{x}^{\frac{2(n+2)}{n-2}} (J)  } \|
%\nabla u \|^{\theta}_{ L_{t}^{\frac{2(n+2)}{n}} L_{x}^{\frac{2(n+2)}{n}} (J)  }
%\end{array}
%\label{Eqn:Dtheta1}
%\end{equation}

%\item [$\bullet$]

%\begin{equation}
%\begin{array}{ll}
%\| u \|_{L_{t}^{\frac{2(n+2)}{n-2}} L_{x}^{\frac{2(n+2)}{n-2}} (J) } & \leq  E^{\frac{1-\tilde{\theta}}{2}} \| D^{\theta} u
%\|^{\tilde{\theta}}_{L_{t}^{q} L_{x}^{q} (J)}
%\end{array}
%\label{Eqn:Dtheta2}
%\end{equation}
%with $\tilde{\theta}$ such that $\frac{ \tilde{\theta} }{q} := \frac{n-2}{2(n+2)}$ \label{lem:EstLtqLxq}
%\end{itemize}
%\label{prop:EstNorm}
%\end{prop}

%and if $n \in \{ 5,6 \}$ then
%\begin{equation}
%\begin{array}{ll}
%\| u \|_{L_{t}^{\infty} L_{x}^{\frac{2n}{n-4}} (J) } & \lesssim  \| u \|_{L_{t}^{\infty} \tilde{H}^{k} (J) }
%\end{array}
%\label{Eqn:SobolevIneq3}
%\end{equation}
%and
%\begin{equation}
%\begin{array}{ll}
%\| u \|_{L_{t}^{\frac{2(n+2)}{n-2}} L_{x}^{\frac{2n(n+2)}{n^{2} - 4n -4}} (J) } & \lesssim \| D u \|_{L_{t}^{\frac{2(n+2)}{n-2}}
%L_{x}^{\frac{2n(n+2)}{n^{2} + 4}} (J) }  + \| D^{k} u \|_{L_{t}^{\frac{2(n+2)}{n-2}} L_{x}^{\frac{2n(n+2)}{n^{2} + 4}}(J)}
%\end{array}
%\label{Eqn:SobolevIneq4}
%\end{equation}


%In Appendix $C$, we prove the proposition below (which may be seen as an extension of Proposition \ref{Prop:NonlinFracSmooth} to the case $\beta = k'-1$):

%\begin{prop}
%Let $ 0  \leq \alpha < 1$, $k'$  be an integer such that $k' \geq 2$, $(r ,r_{2}) \in (1,\infty)^{2}$,
%$r_{1} \in (1, \infty]$ be such that $\frac{1}{r} = \frac{k'-1}{r_{1}} + \frac{1}{r_{2}}  $. Let $F: \mathbb{R}^{+} \rightarrow \mathbb{R}$ be a $C^{k'}$- function and let %$G:=\mathbb{R}^{2} \rightarrow \mathbb{R}^{2}$ be a $C^{k'}$- function such that

%\begin{equation}
%\begin{array}{l}
%F^{[i]}(x) =O \left( \frac{F(x)}{x^{i}} \right), \; \tau \in [0,1]: \;
%\left| F \left( |\tau x + (1-\tau)y|^{2} \right) \right|
%\lesssim  \left| F(|x|^{2}) \right| + \left| F(|y|^{2}) \right|,
%\end{array}
%\label{Eqn:Cdtionf}
%\end{equation}
%and

%\begin{equation}
%G^{[i]}(x,\bar{x})  = O (|x|^{k' -i})
%\label{Eqn:CdtionG}
%\end{equation}
%for $ 0 \leq i \leq k'$. Then

%\begin{equation}
%\begin{array}{ll}
%\left\| D^{ k' -1 + \alpha} ( G(f,\bar{f}) F(|f|^{2}) \right\|_{L^{r}} & \lesssim \| f \|^{k'-1}_{L^{r_{1}}} \| D^{k' -1  + \alpha} f \|_{L^{r_{2}}}
%\| F(|f|^{2}) \|_{L^{\infty}}
%\end{array}
%\label{Eqn:EstToProveFrac}
%\end{equation}

%\label{Prop:NonlinFracSmooth2}
%\end{prop}


%\begin{prop}
%Let $ 0  \leq \alpha < 1$, $k'$ and $\beta$ be integers such that $k' \geq 2$ and $\beta \geq k'-1 $, $(r ,r_{2}) \in (1,\infty)^{2}$,
%$(r_{1},r_{3}) \in (1, \infty]^{2}$ be such that $\frac{1}{r}= \frac{\beta}{r_{1}} + \frac{1}{r_{2}} +\frac{1}{r_{3}}$. Let $F: \mathbb{R}^{+} \rightarrow \mathbb{R}$ be a %$\mathcal{C}^{k'}$- function and let $ G := \mathbb{R}^{2} \rightarrow \mathbb{R}^{2} $ be a $\mathcal{C}^{k'}-$ function that satisfy the following assumptions:

%\begin{equation}
%\begin{array}{l}
%(a): \; \tau \in [0,1]: \; \left| F \left( |\tau x + (1-\tau)y|^{2} \right) \right|
%\lesssim  \left| F(|x|^{2}) \right| + \left| F(|y|^{2}) \right| \\
%(b): i \in \{0,...,k^{'} \}:  \; F^{[i]}(x) = O \left( \frac{F(x)}{x^{i}} \right) \; \text{and} \\
%(c) : \; i \in \{ 0, ..., k^{'} \}: \; G^{[i]}(x,\bar{x})   = O (|x|^{\beta + 1 -i})
%\end{array}
%\nonumber
%\end{equation}
%Then

%\begin{equation}
%\begin{array}{ll}
%\left\| D^{ k' -1 + \alpha} ( G(f,\bar{f}) F(|f|^{2}) \right\|_{L^{r}} & \lesssim \| f \|^{\beta}_{L^{r_{1}}} \| D^{k' -1  + \alpha} f \|_{L^{r_{2}}}
%\| F(|f|^{2}) \|_{L^{r_{3}}}
%\end{array}
%\label{Eqn:EstToProveFrac}
%\end{equation}
%Here $F^{[i]}$ and $G^{[i]}$ denote the $i^{th}-$ derivatives of $F$ and $G$ respectively.

%\label{Prop:NonlinFracSmooth}
%\end{prop}

%\begin{rem}

%In \cite{triroyrad}, Proposition \ref{Prop:NonlinFracSmooth} was stated and proved with the assumptions
%``$ \beta > k' - 1 $'' instead of ``$ \beta \geq k' - 1 $''. In Appendix $C$, we explain that
%(\ref{Eqn:EstToProveFrac}) also holds in the limit case ``$ \beta = k' - 1 $'' by modifying slightly arguments
%in \cite{triroyrad}.

% We also point out a technical issue that appears in \cite{triroyrad} and that is related to this observation,
%and we explain how to resolve this issue.

%\label{Rem:NonlinFracSmooth}
%\end{rem}


%
%In \cite{triroyrad}, Proposition \ref{Prop:NonlinFracSmooth} was stated with the assumption
%$(a)$, $(b)$, and ``$ \text{let} \; \tilde{F} : \mathbb{R}^{+} \rightarrow \mathbb{R} ...  \text{``F  is not a nondecreasing function''. }$'',
%replaced with ``$ \tau \in [0,1]: \; \left| F \left( |\tau x + (1-\tau)y|^{2} \right) \right| \lesssim  \left| F(|x|^{2}) \right| + \left| F( |y|^{2}) \right|, \; \text{and} \; %i \in \{0,...,k^{'} \}:  \;  F^{[i]}(x) = O \left( \frac{F(x)}{x^{i}} \right) $ '', and
%`` $ G \in \mathcal{C}^{k^{'}} \left( \mathbb{R}^{2} \right) \, \text{and} \, i \in \{0,...,k^{'} \}: \;  G^{[i]}(x,\bar{x}) = O \left( |x|^{\beta + 1  - i} \right) $ '',
%`` $ \text{Let} \; \tilde{F} : \mathbb{R}^{+} \rightarrow \mathbb{R} \; \text{be a} \; \mathcal{C}^{k}- \, \text{function}$ '' respectively.

%%%%%%%%%%%%%%%%%%%%%

%\begin{prop}
%Let $ 0 \leq \alpha < 1$, $k'$ and $\beta$ be integers such that $k' \geq 2$ and $\beta > k'-1 $,  $(r , r_{2}) \in (1,\infty)^{2}$,
%$(r_{1},r_{3}) \in (1, \infty]^{2}$ be such that $\frac{1}{r}=
%\frac{\beta}{r_{1}} + \frac{1}{r_{2}} +\frac{1}{r_{3}}$. Let $F: \mathbb{R}^{+} \rightarrow \mathbb{R}$ be a $C^{k}$- function and let $G:=\mathbb{R}^{2} \rightarrow %\mathbb{R}^{2}$ be a $C^{k'}$- function such that

%\begin{equation}
%\begin{array}{l}
%F^{[i]}(x) =O \left( \frac{F(x)}{x^{i}} \right), \; \tau \in [0,1]: \;
%\left| F \left( |\tau x + (1-\tau)y|^{2} \right) \right|
%\lesssim \left| F(|x|^{2}) \right| + \left| F(|y|^{2}) \right|,
%\end{array}
%\label{Eqn:Cdtionf}
%\end{equation}
%and

%\begin{equation}
%G^{[i]}(x,\bar{x})  = O (|x|^{\beta + 1 -i})
%\label{Eqn:CdtionG}
%\end{equation}
%for $ 0 \leq i \leq k' $. Then

%\begin{equation}
%\begin{array}{ll}
%\left\| D^{ k -1 + \alpha} ( G(f,\bar{f}) F(|f|^{2}) \right\|_{L^{r}} & \lesssim \| f \|^{\beta}_{L^{r_{1}}} \| D^{k -1  + \alpha} f \|_{L^{r_{2}}}
%\| F(|f|^{2}) \|_{L^{r_{3}}}
%\end{array}
%\label{Eqn:EstToProveFrac}
%\end{equation}
%Here $F^{[i]}$ and $G^{[i]}$ denote the $i^{th}$- derivatives of $F$ and $G$ respectively. \\
%More generally, let $\tilde{F}: \mathbb{R}^{+} \rightarrow \mathbb{R}$ be a $C^{k}$ function. Substitute $F$ with
%$\tilde{F}$ on the right-hand side of the equality of (\ref{Eqn:Cdtionf}), in the inequality of (\ref{Eqn:Cdtionf}), and on the right-hand side of
%(\ref{Eqn:EstToProveFrac}). With these substitutions made, if
%$F$, $\tilde{F}$, and $G$ satisfy (\ref{Eqn:Cdtionf}) and (\ref{Eqn:CdtionG}), then $F$ and $G$ satisfy (\ref{Eqn:EstToProveFrac}).
% \label{Prop:FracLeibn}

%\end{prop}