Blow-up for the 3D intercritical inhomogeneous NLS with inverse-square potential
Extracted source text · arXiv 2305.06971
\documentclass[11pt, reqno]{amsart}
\usepackage[english]{babel}
\usepackage[T1]{fontenc}
\usepackage{amsmath}
\usepackage{amsthm}
\usepackage{amsfonts}
\usepackage{amssymb}
%\usepackage{arial}
%\usepackage{euller}
%\usepackage{enumerate}
\usepackage{lmodern} %Type1-font for non-english texts and characters
\usepackage{indentfirst} % Indenta o primeiro parágrafo de cada seção.
\usepackage{microtype} % para melhorias de justificação
\usepackage[numbers]{natbib}
\usepackage{hyphenat}
\usepackage{pgfplots}
\usepackage[utf8]{inputenc}
\usepackage{indentfirst}
%\usepackage[notref,notcite]{showkeys}
\definecolor{blackgreen}{RGB}{0,80,0}
\usepackage[plainpages=false,pdfpagelabels,pdfusetitle,pdfdisplaydoctitle, pdfstartpage=1, pdfstartview={{XYZ null null 1.00}}]{hyperref}
\hypersetup{
pdftitle={INLSa},
pdfauthor={Luccas Campos, Mykael Cardoso and Luiz Gustavo Farah},
pdfnewwindow=true,
colorlinks=true,
linkcolor=blackgreen,
citecolor=red}
\usepackage[top=2cm, bottom=2cm, left=2cm, right=2cm]{geometry}
\usepackage{mathtools}
%\usepackage{mathabx}
\mathtoolsset{showonlyrefs}
% O tamanho do parágrafo é dado por:
\setlength{\parindent}{1cm}
% Controle do espaçamento entre um parágrafo e outro:
%\setlength{\parskip}{\baselineskip} % tente também \onelineskip
% ---
%\renewcommand{\baselinestretch}{1.5}
%\geometry{verbose,a4paper,tmargin=3.0cm,bmargin=2.0cm,%%%% Normas ABNT
%lmargin=3.0cm,rmargin=2.0cm,headsep=5mm,footskip=1cm}
%%%%Formatação dos rodapés
%\usepackage{fancyhdr}
%\pagestyle{fancy}
%\fancyhead[LO,LE]{{\bf \nouppercase \leftmark}}
%\fancyhead[RO,RE]{{\sc \rightmark \,\,\,\thepage
%}}
%\fancyfoot[LO,LE]{}
%%%%%%%%
%\renewcommand{\footrulewidth}{0.4pt}
%\renewcommand{\headrulewidth}{0.4pt}
\newcommand{\Real}{\mathbb R}
\newtheorem{thm}{Theorem}[section]
\newtheorem{prop}[thm]{Proposition}
\newtheorem{lemma}[thm]{Lemma}
\newtheorem{claim}[thm]{Claim}
\newtheorem{rem}[thm]{Remark}
\newtheorem{definition}[thm]{Definition}
\newtheorem{coro}[thm]{Corollary}
\numberwithin{equation}{section}
\title{Blow-up for the 3D intercritical inhomogeneous NLS with inverse-square potential}
\author{Luccas Campos, Mykael Cardoso and Luiz Gustavo Farah} % Autor
\date{} % Data
\begin{document}
\maketitle
\begin{abstract}\noindent
In this paper we study the focusing inhomogeneous 3D nonlinear Schr\"odinger equation with inverse-square potential in the mass-supercritical and energy-subcritical regime. We first establish local well-posedness in $\dot{H}_a^{s_c}\cap \dot{H}_a^1$, with $s_c=3/2-(2-b)/2\sigma$. Next, we prove the blow-up of the scaling invariant Lebesgue norm for radial solutions and also, with an additional restriction, in the non-radial case.
\end{abstract}
\section{Introduction}
We consider the initial value problem (IVP) for the inhomogeneous nonlinear Schr\"odinger equation with inverse-square potential $(\mbox{INLS})_a$
\begin{equation}
\begin{cases}
i \partial_tu +\Delta u -\frac{a}{|x|^2}u + |x|^{-b} |u|^{2 \sigma}u = 0, \,\,\, x \in \mathbb{R}^3, \,t>0,\\
u(0) = u_0.
\end{cases}
\label{PVI}
\end{equation}
%The case $a=0$ has been object of intense study by many authors in last few years and is referred as the INLS equation
The (INLS)$_a$ model can be interpreted as an extension to the nonlinear Schr\"odinger equation. For instance, when $a=b=0$ we have the classical nonlinear Schr\"odinger equation, extensively studied over the last decades (see Sulem and Sulem \citep{Sulem}, Bourgain \citep{Bo99}, Cazenave \citep{cazenave}, Linares and Ponce \citep{LiPo15}, Fibich \citep{Fi15}, Tao \cite{TaoBook} and the references therein).
%(see \cite{FIBI1}, \cite{Bourgain}, \cite{CAZENAVEBOOK}, \cite{FELGUS}, \cite{Tao} and the references therein).
The case $b=0$ and $a\neq 0$ is known as the NLS equation with inverse square potential, denoted by (NLS)$_{a}$,
\begin{equation}\label{NLSa}
i\partial_tu +\Delta u -\frac{a}{|x|^2}u + |u|^{2\sigma} u =0.
\end{equation}
This equation has also been attracting attention in recent years (see for instance, \cite{Burq}, \cite{Murphy}, \cite{Murphy1}, \cite{Okazawa2012}, \cite{Zhang}). The equation \eqref{NLSa} appears in several physical settings, such as in quantum field equations or black hole solutions of the Einstein's equations (see e.g. \cite{MR0397192}).
Alternatively, if $a=0$ and $b\neq 0$ we have the inhomogeneous NLS equation, denoted by INLS, i.e.,
\begin{equation}\label{INLS}
i\partial_tu+\Delta u +|x|^{-b} |u|^{2\sigma} u =0,
\end{equation}
which also has received substantial attention recently (see e.g. \cite{g_8}, \cite{CARLOS}, \cite{paper2}, \cite{CHL20},\cite{Campos}, \cite{DK21}, \cite{MMZ21}, \cite{M_Proc22}, \cite{AT21}, \cite{AK21}, \cite{KLS21}, \cite{LS21}. It can model, for example, nonlinear optical systems with spatially dependent interactions (see e.g. \cite{belmonte2007lie} and the references therein). In particular, it can be thought of as modeling
inhomogeneities in the medium in which the wave propagates (see for instance \cite{MALOMED1}). The (INLS)$_a$ can thus be seen as a general model for various physical contexts, and it has been already studied in \cite{CG_Zamp_21} and \cite{AJK_DCDS_22}.
Here, we are interested in the following range of the parameters: %$N\geq 3$ and
\begin{equation}\label{restrictions}
a > 0\, , \quad 0<b<\frac32 \quad \text{and} \quad \dfrac{2-b}{3} < \sigma <2-b.
\end{equation}
%\begin{itemize}
%\item[($I$)]\label{itm:I} $a>0,$
%\vspace*{.4cm}
%\item[($II$)] $\dfrac{2-b}{3} < \sigma <2-b.$
%\end{itemize}
The first restriction in \eqref{restrictions} is related to the dispersive estimate \eqref{dispersive_decay}. It is necessary to deduce Lemma \ref{Lemma-Str2}, which is used, in turn, to prove our local well-posedness result (Theorem \ref{LWP}). The restriction $a>0$ also ensures, via the sharp Hardy inequality
\begin{align}\label{SHI}
\||x|^{-1}u\|_{L^2}\leq 2\|\nabla u\|_{L^2},\,\,\,\mbox{for all}\,\,\, u \in H^1(\Real^3). \,\,\, %\mbox{with} \,\,\,N\geq 3,
\end{align}
the coercivity and boundedness of the linear operator in \eqref{PVI}.
Indeed, consider the quadratic form
$$
Q_a(u)=\int |\nabla u(x)|^2+\frac{a}{|x|^2}|u(x)|^2\,dx
$$
defined in $C^{\infty}_c(\mathbb{R}^3\backslash\{0\})$ and its Friedrichs extension denoted by
$$
\mathcal L_a=-\Delta+\frac{a}{|x|^2}
$$
(see Killip, Miao, Visan, Zhang and Zheng \cite{KMVZZ18} and the references therein for more details).
The restriction $a>0$ (or, more broadly, $a>-{1}/{4}$) guarantees that $\mathcal L_a > 0$ and, defining $\|f\|^2_{\dot{H}^1_a}=\|\mathcal L_a^{\frac12}
f\|^2_{L^2}=(f,\mathcal L_a f)_{L^2}=Q_a(f)$, we also have
\begin{align}\label{equiv}
\|f\|_{\dot{H}^1_a}\sim \|\nabla f\|_{L^2}, \,\,\,\mbox{for all} \,\,\,f\in \dot H^1(\Real^3).
\end{align}
This is a particular case of the $L^p$- based Sobolev spaces adapted to the Schr\"odinger operator with inverse-square potential $\mathcal L_a$ equipped with the norms
$$
\|f\|_{H_a^{s,p}}=\|(1+\mathcal L_a)^{\frac s2}f\|_{L^p}\,\,\, \mbox{and}\,\,\, \|f\|_{\dot{H}_a^{s,p}}=\|\mathcal L_a^{\frac s2}f\|_{L^p}.
$$
We use the usual convention for $p=2$, i.e. $\|f\|_{H_a^{s}}=\|f\|_{H_a^{s,2}}$ and $\|f\|_{\dot{H}_a^{s,2}}=\|f\|_{\dot{H}_a^{s}}$. In \cite{KMVZZ18} the authors studied the relations between these spaces and those defined via $(\Delta)^{s/2}$ and we recall their results in Lemma \ref{EquivHs} below. For instance, when $s>0$ and we have
\begin{align}\label{equiv0}
\|\mathcal L_a^\frac s2 f\|_{L^r}\sim \|D^s f\|_{L^r}, %\lesssim \|\mathcal L_a^\frac s2 f\|_{L^r},
\end{align}
for $0<s<2$ and $1< r<3/s$. Note that the case $s=1$ is a direct consequence of the sharp Hardy inequality as explained above. The relation \eqref{equiv0} will be one of the key ingredients to deduce the local well-posedness result in Theorem \ref{LWP}.
On the other hand, the second and third conditions in \eqref{restrictions} are related to the local well-posedness of \eqref{PVI}. Indeed, the scale-invariant Sobolev space for the initial data in view of the scale symmetry
$$
\lambda\mapsto u_{\lambda}(x,t):=\lambda^{\frac{2-b}{2\sigma}} u(\lambda x,\lambda^2t),
$$
corresponds to $\dot{H}^{s_c}(\Real^3)$, where $s_c=\frac{3}{2}-\frac{2-b}{2\sigma}$. The equation is called intercritical if $0<s_c<1$, which is equivalent to the third condition in \eqref{restrictions}. The condition $0<b<3/2$ stems from the more general restriction $0<b<\min\{N/2,2\}$, for $u_0 \in \dot H^1(\mathbb R^N)$, and it is exploited in detail by Correia and the first and third authors \cite{CCF_Arxiv_22}.
Next, we recall mass and energy conservation along the flow of the $(\mbox{INLS})_a$ equation. If $u(t)$ is a solution to \eqref{PVI} on some time interval $I\subset \mathbb{R}$, $0\in I$, then for any $t\in I$
\begin{equation}\label{Mass}
M[u(t)] =\int |u(x,t)|^2\, dx =M[u_0]
\end{equation}
and
\begin{equation}\label{Energy}
E_a[u(t)] =\frac12 \int |\nabla u(x,t)|^2\, dx +\frac a2\int |x|^{-2}|u(x,t)|^2\,dx - \frac{1}{2\sigma+2}\int |x|^{-b}|u(x,t)|^{2\sigma+2}\,dx=E_a[u_0].
\end{equation}
Recently, Guzmán and the first author \cite{LUCCASCARLOS} obtained the local well-posedness for the $(\mbox{INLS})_a$ equation in $H^1(\Real^N)$ via Kato's method, which is based on the contraction mapping principle and the Strichartz
estimates. More precisely, they considered $N\geq 3$ and showed that for initial data $u_0\in H^1_a(\Real^N)$ and
$$
\left\{
\begin{array}{ll}
a>-\frac{(N-2)^2}{4},& \mbox{ if }\frac{1-b}{N}<\sigma\leq \frac{1-b}{N-2}\mbox{ and }0<b<1\\
a>-\frac{(N-2)^2}{4}+\left(\frac{\sigma(N-2)-(1-b)}{2\sigma+1}\right)^2,& \mbox{if} \max\{0,\frac{1-b}{N}\}<\sigma<\frac{2-b}{N-2}
\end{array}
\right.
$$
there exists a $T=T(\|u_0\|_{\dot H^1_a}, N, \sigma, b)$ and a unique solution of \eqref{PVI} satisfying
\begin{align}
u\in C\left([0,T];H^1_a(\Real^N)\right)\cap L^q\left([0,T]; H^{1,r}_{a}(\Real^N)\right)
\end{align}
where $(q,r)$ is a $L^2$-admissible pair. \\
They also proved the existence of a ground state and a dichotomy between global existence and blow up below the ground state threshold. Furthermore, they proved small data scattering, constructed wave operators in $H^1_a$ and proved that suitable space-time bounds imply scattering in the intercritical case. Later on, An, Jan and Kim
\cite{AJK_DCDS_22} extended the results of Duyckaerts and Roudenko \cite{DR_Going} for the NLS and of the first and second authors \cite{CC_Going} for the INLS to include a description of the global behavior of solutions to (INLS)$_a$ with mass-energy possibly greater than or equal to the ground state threshold.
%Let $T^*>0$ be the maximal time of existence for a solution $u_0\in C([0,T)]; H^1(\Real^N))$ to \eqref{PVI}. If $T^*=\infty$ we say that corresponding solution $u$ is global. Case $T^*<\infty$, then we say the solution $u$ blows up in finite time.
%Using the ideas of \citet{Cbook} (see also \cite{CFG20}) and the proof of the well-posedness of \cite{LUCCASCARLOS}, for $(\mbox{INLS})_a$ in $H^1_a(\Real^N)$ is expected to obtain the following blowup alternative for blow-up solutions in finite time
%\begin{align}
% \left\|\mathcal L_a^{\frac12}
% u(t)\right\|_{L^2}\geq \frac{c}{(T^{\ast}-t)^{\frac{1-s_c}{2}}},
%\end{align}
We are interested here in studying blow-up rates for (INLS)$_a$. To that end, we first obtain the local theory in $\dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)$. We exclude the $L^2$ norm from the well-posedness theorem since it is subcritical with respect to the scaling of the equation. Therefore, the procedure adopted in the proof of Theorem \ref{scteo2} produces a sequence of functions whose mass is unbounded. In order to be able to obtain uniform bounds, we work with a Morrey-Campanato type semi-norm $\rho$, defined by \eqref{defrho}, which somehow mimics the mass and is defined for functions merely in $\dot{H}^{s_c}$. Moreover, the control of this quantity allows us to show that global solutions must have positive energy in the intercritical case (see Theorem \ref{scteo1} below). Next, we state our local theory result.
\begin{thm}\label{LWP} Let $a>0$, $0<b<\frac{3}{2}$ and $\frac{2-b}{3}<\sigma<2-b$. If $u_0 \in \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3) $, then there exist $T=T(\|u_0\|_{\dot{H}^{s_c}_a\cap \dot{H}^1_a }) > 0$ and a unique solution $u$ to the IVP \eqref{PVI} satisfying
$$
u\in C\left([-T,T];\dot{H}^{s_c}_a\cap \dot{H}^1_a \right)\bigcap L^{q}\left([-T,T];\dot{H}_a^{s_c,p}\cap \dot{H}_a^{1,p} \right)\bigcap L^{\overline{q}}\left([-T,T]; L^{\overline{p}} \right),
$$
for any ($q,p$) $L^2$-admissible and $(\overline{q},\overline{p})$ $\dot H^{s_c}$-admissible.
\end{thm}
We note that the nonlinear estimates used in the proof of the previous theorem are more refined then the previous ones that appeared in the literature (see, for instance, \cite[Lemmas 3.2-3.4]{CFG20}). Indeed, in Lemmas \ref{est0}, \ref{est2}, \ref{est1} we are able to achieve the power of $T$ predicted by the scaling of the equation. Such precise estimate implies the desired blow-up rate \eqref{BUA2} used as an additional assumption in our previous works \cite[Theorem 1.2.]{CF20}, \cite[Theorem 1.4]{CF23}. The strategy employed here is inspired by the new ideas introduced in \cite{CCF_Arxiv_22} and we split the spacial domain in a ball of radius $R$ and its complement and then optimize the choice of $R$.
%Tivemos mais cuidado na escolha dos pares nesse trabalho para determinar a potencia correta do T nas estimativas não lineares o que implica a taxa inferior para o blow-up. Para obter a potencia esperada dividimos o dominio espacial em bolas de tamanho R e seu complementar para depois otimizar essa R obtendo a taxa desejada.
%
%Nos trabalhos anteriores, a escolha da potencia no T não é feita com o cuidado, uma vez que não interessa o taxa inferior de blow-up
Next, we study the existence of blow-up solutions and the lower bound for the blow-up rate.
\begin{thm}\label{scteo1} Let $a>0$, $0<b<\frac{3}{2}$ and $\frac{2-b}{3}<\sigma<2-b$. Consider $u_0\in \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3) $ such that $E(u_0)\leq 0$.Then the maximal time of existence $T^{\ast}>0$ of the corresponding solution $u(t)$ to \eqref{PVI} is finite in each one of the following cases
\begin{itemize}
\item[$(a)$] $u_0$ is radial;
\item[$(b)$] $\sigma<\min\left\{2-b,\frac{2}{3}\right\}$.
\end{itemize}
\end{thm}
It should be emphasized that in case $(b)$ the radial assumption is not necessary and instead we need an extra restriction for the parameter $\sigma$. However, if $b>4/3$, then ${2-b}<\frac{2}{3}$ and the result covers all the intercritical range for $\sigma$ (see \eqref{restrictions}).
Let $\sigma_c=\frac{6\sigma}{2-b}$. In view of the Sobolev embedding $\dot{H}^{s_c} \subset L^{\sigma_c}$ and \eqref{equiv0}, we have
\begin{equation}\label{equiv_sigma_c}
\|u(t)\|_{L^{\sigma_c}}\lesssim \|u(t)\|_{\dot{H}^{s_c}}\sim \|u(t)\|_{\dot{H}_a^{s_c}}
\end{equation}
and then $\dot{H}_a^{s_c} \subset L^{\sigma_c}$. The next result states that this scaling invariant Lebesgue norm blows up with a certain lower bound.
\begin{thm}\label{scteo2}
%Let $\sigma_c=\frac{6\sigma}{2-b}$ such that $\dot{H}^{s_c} \subset L^{\sigma_c}$.
Assume $a>0$, $0<b<\frac{3}{2}$ and $\frac{2-b}{3}<\sigma<2-b$. Let $u_0\in \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)$ be such that the maximal time of existence $T^{\ast}>0$ of the corresponding solution $u$ to \eqref{PVI} is finite.
% and satisfies %either
%\begin{align}\label{finitegrad}
%\sup_{t\in [0,T^{\ast})}\left\|\nabla u(t)\right\|_{L^2} \,\,\, \textit{is finite}
%\end{align}
%or
%\begin{align}\label{esttl}
%\left\|\mathcal L_a^{\frac12}
% u(t)\right\|_{L^2}\geq \frac{c}{(T^{\ast}-t)^{\frac{1-s_c}{2}}},
%\end{align}
%for some constant $c=c(N,\sigma)$ and $t$ close enough to $T^{\ast}$.
Suppose that either $u_0$ is radial or $\sigma<\min\left\{2-b,\frac{2}{3}\right\}$. Then there exists $\alpha=\alpha(N,\sigma,b)>0$ such that
\begin{align}\label{rate}
\|u(t)\|_{L^{\sigma_c}}\geq |\log (T^{\ast}-t)|^{\alpha},\,\,\,as\,\,\,t\to T^{\ast}.
\end{align}
\end{thm}
%Let $\sigma_c=\frac{6\sigma}{2-b}$ such that $\dot{H}^{s_c} \subset L^{\sigma_c}$.
%In view of the Sobolev embedding $\dot{H}^{s_c} \subset L^{\sigma_c}$ and \eqref{equiv0}, we have
%\begin{equation}\label{equiv_sigma_c}
%\|u(t)\|_{L^{\sigma_c}}\lesssim \|u(t)\|_{\dot{H}^{s_c}}\sim \|u(t)\|_{\dot{H}_a^{s_c}}.
%\end{equation}
%So that $\dot{H}_a^{s_c} \subset L^{\sigma_c}$ and Theorem \ref{scteo2} also implies the blow-up of the critical Sobolev norm $\dot{H}_a^{s_c}$.
Note that this phenomenon does not occur in the mass-critical case ($s_c = 0$), since the scaling invariant $L^2$-norm is preserved along the flow by the mass conservation \eqref{Mass}.
We emphasize that our previous studies about the INLS model in \cite{CF20, CF23} prove the lower bound \eqref{rate} upon imposing an additional assumption and only implies $\limsup_{t\to T^*} \|u(t)\|_{\dot{H}^{s_c}}=\infty$. The novelty here is that we prove that this assumption always holds in the intercritical case, see Proposition \ref{Blowalt} below. Using similar ideas, the same statement can also be shown to hold for the INLS setting, improving our previous results in this case.
We expect that our blow-up results in the last two theorems extend to higher dimensions, if one can generalize the local theory in Theorem \ref{LWP} to dimension $N\geq 3$. The main obstacle is the need of Strichartz estimates for exotic pairs, especially the non $L^2$-admissible ones. This can be achieved, for example, by proving the dispersive $L^1$-$L^\infty$ decay for the linear evolution, which can then be combined with the theory developed by Foschi \cite{Foschi05} to prove the desired estimates.
This paper is organized as follows. In Section \ref{Prel}, we introduce some notation and preliminary estimates. The well-posedness theory in $\dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)$ is presented in Section \ref{WPT}. The last section is devoted to prove the existence of blow-up solutions and also a lower bound for the blow-up rate.
\section{Preliminaries}\label{Prel}
We start this section by introducing the notation used throughout the paper. We use $c$ to denote various constants that may vary line by line. Given any positive quantities $a$ and $b$, the notation $a\lesssim b$ means that $a \leq cb$, with $c$ uniform with respect to the set where $a$ and $b$ vary. We denote by $p'$ the H\"older conjugate of $1 \leq p \leq \infty$ and we
use $a^+$ and $a^-$ to denote $a +\varepsilon$ and $a -\varepsilon$, respectively, for a sufficiently small $\varepsilon > 0$.
The following fractional Leibniz and Chain rules obtained will be used in the sequel (see also the generalizations suitable for singular weights proved in \cite{CCF_Arxiv_22}).
\begin{lemma}[Fractional calculus \cite{CW91, KPV93}]\label{leibniz}
Let $N\geq 1$, $s \in (0,1]$, $p, q_1,q_2, r_1, r_2\in (1,\infty)$ with $\frac{1}{p} = \frac{1}{q_i} + \frac{1}{r_i}$, $i = 1, 2$ and $G\in C^1(\mathbb{C})$. Then
\begin{itemize}
\item[(i)] $\displaystyle \|D^s(fg)\|_{L^p(\mathbb R^N)} \lesssim \|D^sf\|_{L^{q_1}(\mathbb R^N)}\|g\|_{L^{r_1}(\mathbb R^N)}+\|f\|_{L^{p_2}(\mathbb R^N)}\|D^s g\|_{L^{q_2}(\mathbb R^N)}.$\\
\item[(ii)] $\displaystyle \|D^sG(f)\|_{L^p(\mathbb R^N)} \lesssim \|G'(f)\|_{L^{q_1}(\mathbb R^N)}\|D^sf\|_{L^{r_1}(\mathbb R^N)}.$
\end{itemize}
%Let $N\geq 1$, $0<s<1$ and $\frac{1}{p} = \frac{1}{q_i} + \frac{1}{r_i}$, with $p \in (1,\infty)$, $q_i,r_i\in (1,\infty]$, $i = 1, 2$. If $f, g \in \mathcal{S}(\mathbb{R}^N)$, then
%\begin{align}
% \|D^s(fg)\|_{L^p} &\lesssim
% \|D^sf\|_{L^{q_1}}\|g\|_{L^{r_1}}
% +\|f\|_{L^{p_2}}\|D^s g\|_{L^{q_2}}.
%\end{align}
\end{lemma}
Next, we recall an important tool proved by Killip, Miao, Visan, Zhang and Zheng. It was proved for a large range of parameters, and we only state the case that interests us here.
\begin{lemma}[Equivalence of Sobolev norms \cite{KMVZZ18}]\label{EquivHs} If $a >0$, $0<s<1$, and $1<p< 3/s$ then
\begin{equation}
\|D^s f\|_{L^p}\lesssim \| \mathcal{L}_a^{\frac{s}{2}} f\|_{L^p} \lesssim \|D^s f\|_{L^p}\;\textnormal{for all}\;f\in \mathbb{C}^\infty_c(\mathbb{R}^3 ).
\end{equation}
\end{lemma}
Now, we define the notion of admissible pairs.
\begin{definition}\label{Hs_adm} If $s \in (-1,1)$, the pair $(q,r)$ is called $\dot{H}^s$\textit{-admissible} if it satisfies the condition
\begin{equation}\label{hs_adm_eq}
\frac{2}{q} = \frac{3}{2}-\frac{3}{r}-s,
\end{equation}
where
$$
2 \leq q,r \leq \infty.
$$
In particular, if $s=0$, we say that the pair is $L^2$-admissible.
\end{definition}
\begin{definition}\label{As}
Consider the set
\begin{equation}
\mathcal{A}_0 = \left\{(q,r)\text{ is } L^2\text{-admissible} \left|
%\begin{cases}
\, 2 \leq r \leq 6
%, & N \geq 3\\
%\frac{2}{1-s} \leq q \leq \left(\left(\frac{2}{1-s}\right)^+\right)', & N = 2\\
%\frac{2}{1-2s} \leq q \leq \infty, &N = 1
%\end{cases}
\right.
\right\}.
\end{equation}
For $s \in (-1,1)$, consider also
\begin{equation}
\mathcal{A}_s = \left\{(q,r)\text{ is } \dot{H}^{s}\text{-admissible} \left|\,
%\begin{cases}
\left(\frac{6}{3-2|s|}\right)^+ \leq r \leq 6^-
%, & N \geq 3\\
%\frac{2}{1-s} \leq q \leq \left(\left(\frac{2}{1-s}\right)^+\right)', & N = 2\\
%\frac{2}{1-2s} \leq q \leq \infty, &N = 1
%\end{cases}
\right.
\right\}.
\end{equation}
%and
%\begin{equation}
% \mathcal{A}_{-s} = \left\{(q,r) \text{ is } \dot{H}^{-s}\text{-admissible} \left|
% %\begin{cases}
% \left(\frac{2N}{N-2s}\right)^+ \leq r \leq \left(\frac{2N}{N-2}\right)^-
% %, & N \geq 3\\
% %\left(\frac{2}{1-s}\right)^+ \leq q \leq \left(\left(\frac{2}{1+s}\right)^+\right)', & N = 2\\
% %\left(\frac{2}{1-2s}\right)^+ \leq q \leq \infty, &N = 1
% %\end{cases}
% \right.\right\}.
%\end{equation}
If $s \in [0,1)$, we define the following Strichartz norm
\begin{equation}
\|u\|_{S(\dot{H}^s,I)} = \sup_{(q,r)\in \mathcal{A}_s}\|u\|_{L_I^qL_x^r},
\end{equation}
and the dual Strichartz norm
\begin{equation}
\|u\|_{S'(\dot{H}^{-s},I)} = \inf_{(q,r)\in \mathcal{A}_{-s}}\|u\|_{L_I^{q'}L_x^{r'}}.
\end{equation}
For $s=0$, we shall write $S(\dot{H}^0,I) = S(L^2,I)$ and $S'(\dot{H}^0,I) = S'(L^2,I)$. If $I=\mathbb{R}$, we will omit $I$.
\end{definition}
Here, we are mainly interested in the case $a>0$ and $x\in \mathbb{R}^3$. In particular, the relation \eqref{equiv0} holds for $a>0$, $0<s<2$ and $r<3/s$, which implies
\begin{align}\label{equivsc1}
\|f\|_{\dot{H}^{s_c}_a}\sim \|f\|_{\dot{H}^{s_c}} \,\,\, \mbox{and} \,\,\, \|f\|_{\dot{H}^{1}_a}\sim \|f\|_{\dot{H}^{1}}.
\end{align}
%\begin{rem}\label{equivalence}
% Let $0 <s< 2$. It is easy to see that, if $a>0$ then $\|f\|_{H_a^{s,r}} \sim \|f\|_{H^{s,r}}$, provided that $1<r<\frac{N}{s}$. When $-\frac{(N-2)^2}{2}\leq a<0$ we have $0<\rho<\frac{(N-2)}{2}$ and so $\|f\|_{H_a^{s,r}} \sim \|f\|_{H^{s,r}}$ if $\frac{N}{N-\rho}<r<\frac{N}{s+\rho}$.
%\end{rem}
%
%\begin{rem}
%In most of the works related to the NLS equation, the set $\mathcal{A}_0$ is usually defined for $2\leq r\leq 6$. Here we need more restriction since we will make use of the following norm equivalence property on Sobolev spaces: For $a>0$, $0<s<2$ and $r<3/s$
%\begin{align}\label{equiv0}
%\|\mathcal L_a^\frac s2 f\|_{L^r}\lesssim \|D^s f\|_{L^r}\lesssim \|\mathcal L_a^\frac s2 f\|_{L^r}.
%\end{align}
%See Killip, Miao, Visan, Zhang and Zheng \cite{KMVZZ}. Note that the case $s=1$ is a direct consequence of the sharp Hardy inequality.\\
%\end{rem}
In this work, we consider the following Strichartz estimates for $s \in [0,1)$ (see Kato \cite{Kato94}, Cazenave \cite{cazenave}, Keel and Tao \cite{KT98}, Foschi \cite{Foschi05})
\begin{equation}\label{S2}
\|e^{it\Delta}f\|_{S(\dot{H}^s)} \lesssim\|f\|_{\dot{H}^s},
\end{equation}
and
\begin{equation}\label{KS2}
\left\|\int_\mathbb{R}e^{i(t-\tau)\Delta}g(\cdot,\tau) \, d\tau\right\|_{S(\dot{H}^s,I)} + \left\|\int_0^te^{i(t-\tau)\Delta}g(\cdot,\tau) \, d\tau\right\|_{S(\dot{H}^s,I)}\lesssim\|g\|_{S'(\dot{H}^{-s},I)}.
\end{equation}
\begin{lemma}\label{Lemma-Str1}
Let $s \in [0,1)$ and $a>-\frac{(N-2)^2}{4}$. Then,
\begin{eqnarray}
%\| e^{-it \mathcal{L}_a}f \|_{S(L^2;I)} &\lesssim & \|f\|_{L^2},\label{SE1} \\
\| e^{-it \mathcal{L}_a}f \|_{S(\dot{H}^s;I)} &\lesssim & \|f\|_{\dot{H}^s_a} \label{SE2}\\
\left \| \int_{t_0}^t e^{-i(t-t')\mathcal{L}_a}g(\cdot,t') dt' \right \|_{S(L^2;I) } &\lesssim& \|g\|_{S'(L^2;I)} \label{SE3}%\\
%\left \| \int_{t_0}^t e^{-i(t-t')\mathcal{L}_a}g(\cdot,t') dt' \right \|_{S(\dot{H}^{s};I)} &\lesssim &\|D^s g\|_{S'(L^2;I)}, \label{SE4}
\end{eqnarray}
%\begin{equation}\label{SE1}
%\| e^{-it \mathcal{L}_a}f \|_{S(L^2;I)} \lesssim \|f\|_{L^2},
%\end{equation}
%\begin{equation}\label{SE2}
%\| e^{-it \mathcal{L}_a}f \|_{S(\dot{H}^s;I)} \lesssim \|f\|_{\dot{H}^s_a}.
%\end{equation}
%\begin{equation}\label{SE3}
%\left \| \int_{t_0}^t e^{-i(t-t')\mathcal{L}_a}g(\cdot,t') dt' \right \|_{S(L^2;I) } \lesssim \|g\|_{S'(L^2;I)},
%\end{equation}
%\begin{equation}\label{SE4}
%\left \| \int_{t_0}^t e^{-i(t-t')\mathcal{L}_a}g(\cdot,t') dt' \right \|_{S(\dot{H}^{-s};I)} \lesssim \|D^s g\|_{S'(L^2;I)},
%\end{equation}
%where $I\subset\mathbb{R}$ be an interval and $t_0\in I$.
%\end{itemize}
\begin{proof}
See Burq, Planchon, Stalker and Tahvildar-Zadeh \cite{BPST03} and Zhang and Zheng \cite{ZZ14}.
\end{proof}
\end{lemma}
\begin{lemma}\label{Lemma-Str2}
Let $s \in [0,1)$ and $a> 0$. Then,
\begin{equation}\label{SE5}
\left \| \int_{t_0}^t e^{-i(t-t')\mathcal{L}_a}g(\cdot,t') dt' \right \|_{S(\dot{H}^s;I) } \lesssim \|g\|_{S'(\dot{H}^{-s};I)},
\end{equation}
where $I\subset\mathbb{R}$ be an interval and $t_0\in I$.
\begin{proof}
The Application 2 in Fanelli, Felli, Fontelos and Primo \cite{FFFP13} proves
\begin{equation}\label{dispersive_decay}
\|e^{-it \mathcal{L}_a} f\|_{L^\infty_x} \lesssim \frac{1}{|t|^{\frac{3}{2}}}\|f\|_{L^1_x},
\end{equation}
which corresponds to the same decay as for the free ($a=0$) Schr\"odinger evolution. The estimate \eqref{SE5} is then a corollary of the main result in Foschi \cite{Foschi05}.
\end{proof}
\end{lemma}
\begin{rem}
The result in Lemma \ref{Lemma-Str1} also holds in general space dimension $N>3$. On the other hand, Lemma \eqref{Lemma-Str2} is still an open problem in this general setting.
\end{rem}
%PORQUE OS RESULTADOS DE [1] E [5] NAO IMPLICAM A DESIGUALDADE DO LEMMA 2.5???
\section{Well-posedness in {$\dot{H_a^{s_c}}\cap \dot H_a^1$}}\label{WPT}
In this section we study the well-posedness to $(\mbox{INLS})_a$ in homogeneous Sobolev spaces $\dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)$. The main ingredients are the equivalence of Sobolev norms (Lemma \ref{EquivHs}) and Strichartz estimates (Lemmas \ref{Lemma-Str1}-\ref{Lemma-Str2}).
\subsection{Nonlinear estimates}
We start by proving some preliminary estimates. The novelty here is that we can optimize the estimates around and outside the origin in order to establish the exact power of $T$ in the right hand side of these inequalities. This will be important to deduce a lower bound for the blow-up rate for finite time solutions (see Proposition \ref{Blowalt}).
\begin{lemma}\label{est0} Let $a>0$, $0<b<\frac{3}{2}$ and $\frac{2-b}{3}<\sigma<2-b$. Then, for $I=[-T,T]$ we have
\begin{equation}
\left\|\mathcal L_a^{\frac12}\left(|x|^{-b}|u|^{2\sigma}u\right) \right\|_{S'(L^2;I)} \lesssim T^{\sigma(1-s_c)}\|\mathcal L_a^\frac12 u\|^{2\sigma+1}_{S(L^2;I)}.
\end{equation}
\end{lemma}
\begin{proof}
Let $R>0$ to be chosen later and denote by $B_R$ the closed ball of radius $R$ in $\mathbb{R}^3$. We consider the sets $B^R_+=B_R$ and $B^R_-=\Real^3\backslash B_R$. Let $\eta\in(0,1)$ be such that $(1-2b)/3 < \eta < (3-2b)/2$ (which is possible since $0 < b < 3/2$). Let also $0<\varepsilon\ll\min\{(3-2b)/2-\eta, \eta - (1-2b)/3\}$. Define
\begin{equation}\label{palpha}
p = 6/(1+2 \eta) \quad \mbox{and} \quad \alpha_{\pm} = \frac{6(2\sigma+1)}{5-2b+4\sigma - 2\eta \mp 2\varepsilon}.
\end{equation}
From the choices of $\eta$ and $\varepsilon$, together with $(2-b)/3<\sigma<2-b$ and $0 < b < 3/2$, one gets that $2< p <6 $ and $2<\alpha_{\pm} <3$. Moreover, the following relation holds
\begin{eqnarray}\label{rel1}
\frac{1}{p'}&=&\frac{b+1\pm\varepsilon}{3}+(2\sigma+1)\left(\frac{1}{\alpha_{\pm}}-\frac{1}{3}\right)\\
&=&\frac{b\pm\varepsilon}{3}+2\sigma\left(\frac{1}{\alpha_{\pm}}-\frac{1}{3}\right)+\frac{1}{\alpha_{\pm}}.
\end{eqnarray}
Therefore, from the H\"older inequality and Sobolev embedding we obtain
\begin{align}
\left\|\nabla\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{p'}_x(B^R_{\pm})}
&\lesssim \left(\left\||x|^{-b}\right\|_{L^{\frac{3}{b\pm \varepsilon}}_x(B^R_{\pm})}+\left\||x|^{-b-1}\right\|_{L^{\frac{3}{b+1\pm \varepsilon}}_x(B^R_{\pm})}\right)\|\nabla u\|_{L^{\alpha_{\pm}}_x}^{2\sigma+1}\\
&\lesssim R^{\pm \varepsilon}\|\nabla u\|_{L^{\alpha_{\pm}}_x}^{2\sigma+1}.\label{estle1ger}
\end{align}
Now, let $q$, $\beta_{\pm}$ be such that $(q,p)$, $(\beta_+, \alpha_+)$ and $(\beta_-, \alpha_-)$ are $L^2$-admissible pairs. One can easily check that, for $\theta_\pm=\sigma(1-s_c)\mp \varepsilon/2$, the following condition is satisfied
\begin{eqnarray}\label{rel2}
\frac{1}{q'}=\theta_\pm+\frac{2\sigma+1}{\alpha_\pm},
\end{eqnarray}
then, by the H\"older inequality in the time variable and the equivalence of Sobolev norms, we have
\begin{align}
\left\|\mathcal L_a^{\frac12}\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{p'}_x}
&\lesssim \left(T^{\theta_{+}}R^{\varepsilon}+T^{\theta_{-}}R^{-\varepsilon}\right)\|\mathcal L^\frac12_a u\|_{S(L^2;I)}^{2\sigma+1}.
\end{align}
Finally, taking $R=T^{\frac12}$, we have the desired estimate.
%Given $\frac{2-b}{3}<\sigma<2-b$ and $0<b<\frac{3}{2}$, for $0<\varepsilon\ll \min\{\frac{1+2\sigma}{4},b, \frac 32-b\}$, let $(q,p)$ $L^2$-admissible such that
%\begin{itemize}
% \item $\max\{2,\frac{3}{2-b-\varepsilon}\}< p<6,$ if $3-2\sigma-2b+2\varepsilon\leq 0;.$
% \item $\max\{2,\frac{3}{2-b-\varepsilon}\}< p<\min\{6,\frac{6}{3-2\sigma-2b+2\varepsilon}\},$ if $3-2\sigma-2b+2\varepsilon> 0.$
%\end{itemize}
%
%Since $p'<2$, we use the equivalence of Sobolev norms (Lemma \ref{EquivHs}) to first deduce
%\begin{equation}\label{ESS}
%\left\|\mathcal L_a^{\frac12}\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{p'}_x}\lesssim \left\|\nabla\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{p'}_x}.
%\end{equation}
%
%Now, considering $(\beta_+,\alpha_+)$ and $(\beta_-,\alpha_-)$ $L^2$-admissible pairs where
%\begin{align}\label{alpha+-}
% \alpha_{\pm}=\frac{3p(2\sigma+1)}{p(3+2\sigma-b\mp\varepsilon)-3},
%\end{align}
%we have $2< \alpha_{\pm}<3$. Then Sobolev embedding and H\"older's inequality yield
%\begin{align}
%\left\|\nabla\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{p'}_x(A^R_{\pm})}
% &\lesssim \left\|\left(\left\||x|^{-b}\right\|_{L^{\gamma_{\pm}}_x(A^R_{\pm})}+\left\||x|^{-b-1}\right\|_{L^{d_{\pm}}_x(A^R_{\pm})}\right)\|\nabla u\|_{L^{\alpha_{\pm}}_x}^{2\sigma+1}\right\|_{L^{q'}_T}\\
% &\lesssim T^{\theta_{\pm}} \left(\left\||x|^{-b}\right\|_{L^{\gamma_\pm}_x(A^R_{\pm})}+\left\||x|^{-b-1}\right\|_{L^{d_{\pm}}_x(A^R_{\pm})}\right)\|\nabla u\|_{L^{\beta_{\pm}}_TL^{\alpha_{\pm}}_x}^{2\sigma+1},\label{estle1ger}
%\end{align}
%if the following conditions are satisfied
%\begin{align}
% \frac{3}{d_\pm}-b-1=\frac{3}{\gamma_\pm}-b=3+2\sigma-b-3\left(\frac{1}{p}+\frac{2\sigma+1}{\alpha_\pm}\right)
%\end{align}
%and
%\begin{align}
% 2\theta_\pm=-1-3\sigma+3\left(\frac1p+\frac{2\sigma+1}{\alpha_\pm}\right).
%\end{align}
%More precisely, due to the choice of $\varepsilon$ and $\alpha_\pm$, we obtain
%\begin{align}
%\frac{3}{d_\pm}-b-1=\frac{3}{\gamma_\pm}-b=\pm\varepsilon\,\,\,\,\mbox{ and }\,\,\,\,2\theta_\pm=2\sigma(1-s_c)\mp \varepsilon.
%\end{align}
%Thus, in view of \eqref{ESS}, we can rewrite the inequality \eqref{estle1ger} as
%\begin{align}
% \left\|\mathcal L_a^{\frac12}\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{p'}_x}
% &\lesssim \left(T^{\theta_{+}}R^{\frac{3}{\gamma_+}-b}+T^{\theta_{-}}R^{\frac{3}{\gamma_-}-b}\right)\|\mathcal L^\frac12_a u\|_{S(L^2;I)}^{2\sigma+1}.
%\end{align}
%Finally, taking $R=T^{\frac12}$, we have the desired estimate.
\end{proof}
\begin{lemma}\label{est2} Let $a>0$, $0<b<\frac32$ and $\frac{2-b}{3}<\sigma<2-b$. Then, for $I=[-T,T]$ we have
\begin{equation}
\left\|\mathcal L_a^{\frac{s_c}{2}}\left(|x|^{-b}|u|^{2\sigma}u\right) \right\|_{S'(L^2;I)} \lesssim T^{\sigma(1-s_c)}\|\mathcal L_a^\frac12 u\|_{S(L^2;I)}^{2\sigma}\|\mathcal L_a^\frac{s_c}{2} u\|_{S(L^2;I)}.
\end{equation}
\end{lemma}
\begin{rem}
Here, due to the product rule, we are required to estimate quantities comparable to $(D^{s_c}|x|^{-b}) |u|^{2\sigma}u$ and to $|x|^{-b} |u|^{2\sigma}D^{s_c}u$. However, unlike the previous Lemma, since fractional derivatives are not localized in space, one cannot simply split the estimate on a ball and its exterior. Indeed, it is \textbf{not} expected that
\begin{equation}
\|D^{s_c}(|x|^{-b}f)\|_{L_x^p(B_\pm^R)} \lesssim \|D^{s_c}|x|^{-b}\|_{L_x^{p_1}(B_\pm^R)} \|f\|_{L^{q_1}} + \||x|^{-b}\|_{L_x^{p_2}(B_\pm^R)} \|D^s f\|_{L^{q_2}}
\end{equation}
for general $f$, even for suitable $p,p_j,q_j$, due to the non-local character of the fractional derivatives. A more appropriate version of the Leibniz rule, adapted to deal with this case, can be found in \cite{CCF_Arxiv_22}, and the nonlinear estimates there can be readily adapted to the proof of this Lemma. Nevertheless, we present another approach here, involving a well-applied localization, and a Gagliardo-Nirenberg inequality. Indeed, we let $\psi$ be a smooth, nonincreasing, radially symmetric function such that
\begin{equation}
\psi(x) = \begin{cases}
1, & |x| \leq 1\\
0, & |x| \geq 2,
\end{cases}
\end{equation}
and let $\psi_R (x) = \psi (x/R)$ for all $x$ and $R>0$. We then claim
\end{rem}
\begin{claim}\label{claim_x_b}
We have, for all $0\leq s \leq 1$,
\begin{equation}
\|D^{s}(|x|^{-b}\psi_R)\|_{L^{\frac{3}{b+s+\varepsilon}}} \lesssim R^{\varepsilon},
\end{equation}
and
\begin{equation}
\|D^{s}(|x|^{-b}(1-\psi_R))\|_{L^{\frac{3}{b+s-\varepsilon}}} \lesssim R^{-\varepsilon}.
\end{equation}
\end{claim}
\begin{proof}[Proof of the Claim]
%The work of Brezis and Mironescu \cite[Theorem 1]{BM_18} immediately imply that, if $p, p_1, p_2 > 1$ and $f \in C^\infty_0(\mathbb{R}^3)$ are such that
%\begin{equation}
% \frac{1}{p} = \frac{1-s}{p_1} + \frac{s}{p_2},
%\end{equation}
%then
%\begin{equation}
% \|D^{s} f\|_{L^p} \lesssim \|f\|_{L^{p_1}}^{1-s}\left(\|\nabla f\|_{L^{p_2}}+ \|f\|_{L^{p_2}}\right)^{s}.
%\end{equation}
%A standard scaling argument lets us drop the last term inside the parentheses, leading to
% \begin{equation}\label{GN_Brezis}
% \|D^{s} f\|_{L^p} \lesssim \|f\|_{L^{p_1}}^{1-s}\|\nabla f\|_{L^{p_2}}^{s}.
% \end{equation}
We use Sobolev embedding to deduce
\begin{equation}
\|D^{s}(|x|^{-b}\psi_R)\|_{L^{\frac{3}{b+s+\varepsilon}}}\lesssim \|\nabla(|x|^{-b}\psi_R)\|_{L^{\frac{3}{b+1+\varepsilon}}} \lesssim R^{\varepsilon},
\end{equation}
since $|\nabla |x|^{-b}||\psi_R| = |x|^{-b-1}|\psi_R|$ and since $|x|^{-b}|\nabla \psi_R| \lesssim R^{-b-1}$ and $\nabla \psi_R$ is supported on $R \leq |x| \leq 2R$. Similar calculations give the estimate for the term with $(1-\psi_R)$.
\end{proof}
\begin{proof}[Proof of Lemma \ref{est2}]
%Let $\eta\in(0,1)$ be such that $(1-2b)/3 < \eta < (3-2b)/2$ (which is possible since $0 < b < 3/2$). Let also $0<\varepsilon\ll\min\{(3-2b)/2-\eta, \eta - (1-2b)/3\}$. Taking $p = 6/(1+2 \eta)$ we use definition \eqref{alpha+-} to get
%\begin{equation}
%\alpha_{\pm} = \frac{6(2\sigma+1)}{5-2b+4\sigma - 2\eta \mp 2\varepsilon}.
%\end{equation}
%
%From the choices of $\eta$ and $\varepsilon$, together with $(2-b)/3<\sigma<2-b$ and $0 < b < 3/2$, one gets that $2< p <6 $ and $2<\alpha_{\pm} <3$.
As in the proof of Lemma \ref{est0}, consider the $L^2$-admissible pairs $(q,p)$, $(\beta_+, \alpha_+)$ and $(\beta_-, \alpha_-)$ given by \eqref{palpha}. From relation \eqref{rel1}, we also have
\begin{eqnarray}
\frac{1}{p'}&=&\frac{b+s_c\pm\varepsilon}{3}+2\sigma\left(\frac{1}{\alpha_{\pm}}-\frac{1}{3}\right)+\left(\frac{1}{\alpha_{\pm}}-\frac{s_c}{3}\right).%\\
%&=&\frac{b\pm\varepsilon}{3}+2\sigma\left(\frac{1}{\alpha_{\pm}}-\frac{1}{3}\right)+\frac{1}{\alpha_{\pm}},
\end{eqnarray}
The Lemma \ref{leibniz} (Leibniz and Chain rules), H\"older and Sobolev inequalities yield
\begin{align}
\|D^{s_c}(|x|^{-b}|u|^{2\sigma}u)\|_{L_x^{p'}}
&\lesssim \|D^{s_c}(|x|^{-b}\psi_R)\|_{L_x^{\frac{3}{b+s_c+\varepsilon}}} \|\nabla u\|^{2\sigma}_{L_x^{\alpha_+}}\|D^{s_c}u\|_{L_x^{\alpha_+}} \\
&\quad+ \||x|^{-b}\psi_R\|_{L_x^{\frac{3}{b+\varepsilon}}} \|\nabla u\|^{2\sigma}_{L_x^{\alpha_+}}\|D^{s_c}u\|_{L_x^{\alpha_+}}\\
&\quad +\|D^{s_c}(|x|^{-b}(1-\psi_R))\|_{L_x^{\frac{3}{b+s_c-\varepsilon}}} \|\nabla u\|^{2\sigma}_{L_x^{\alpha_-}}\|D^{s_c}u\|_{L_x^{\alpha_-}}\\
&\quad+ \||x|^{-b}(1-\psi_R)\|_{L_x^{\frac{3}{b-\varepsilon}}} \|\nabla u\|^{2\sigma}_{L_x^{\alpha_-}}\|D^{s_c}u\|_{L_x^{\alpha_-}}.
%\\
% &\lesssim R^{\varepsilon}\|\nabla u\|^{2\sigma}_{L_x^{\alpha_+}}\|D^{s_c}u\|_{L_x^{\alpha_+}} +
% R^{-\varepsilon}\|\nabla u\|^{2\sigma}_{L_x^{\alpha_-}}\|D^{s_c}u\|_{L_x^{\alpha_-}}.
\end{align}
Next, from Claim \ref{claim_x_b}, we get
$$
\|D^{s_c}(|x|^{-b}|u|^{2\sigma}u)\|_{L_x^{p'}} \lesssim R^{\varepsilon}\|\nabla u\|^{2\sigma}_{L_x^{\alpha_+}}\|D^{s_c}u\|_{L_x^{\alpha_+}} +
R^{-\varepsilon}\|\nabla u\|^{2\sigma}_{L_x^{\alpha_-}}\|D^{s_c}u\|_{L_x^{\alpha_-}}.
$$
Finally, using the relation \eqref{rel2}, the H\"older inequality in the time variable and the equivalence of Sobolev norms imply
\begin{equation}
\|\mathcal L^\frac{s_c}{2}_a(|x|^{-b}|u|^{2\sigma}u)\|_{L^{q'}_TL^{p'}_x} \lesssim \left(T^{\theta_{+}}R^{\varepsilon}+T^{\theta_{-}}R^{-\varepsilon}\right)\|\mathcal L^\frac12_a u\|_{S(L^2;I)}^{2\sigma}\|\mathcal L^\frac{s_c}{2}_a u\|_{S(L^2;I)}
\end{equation}
and the choice $R = T^{\frac12}$ finishes the proof.
\end{proof}
%%%%%%% Lema 3.2 anterior
% \begin{lemma}\label{est2} Let $N=3$, $a>0$, $0<b<\frac32$ and $\frac{2-b}{3}<\sigma<2-b$. Then, for $I=[-T,T]$ we have
% \begin{equation}
% \left\|\mathcal L_a^{\frac{s_c}{2}}\left(|x|^{-b}|u|^{2\sigma}u\right) \right\|_{S'(L^2;I)} \lesssim T^{\frac{1-s_c}{2}}\|\mathcal L_a^\frac12 u\|_{S(L^2;I)}\|\mathcal L_a^\frac{s_c}{2} u\|^{2\sigma}_{S(L^2;I)}.
% \end{equation}
% \end{lemma}
% \begin{proof}
% Given $\sigma>0$ and $0<b<\frac{3}{2}$, for $\varepsilon>0$ small enough, consider $r>0$ such that\footnote{Note that this choose is possible since $\sigma+\frac12-s_c>0$, in view of the assumption $0<b<\frac{3}{2}$.}
% \begin{align}\label{condr}
% r>\max\left\{\frac{3}{\frac{1}{2}+s_c},\frac{3}{\sigma-\varepsilon}\right\}\,\,\,\,\mbox{ and }\,\,\,\,r\left(s_c-\frac{1}{2}\right)<3.
% \end{align}
% Let also $q>2$ such that $\frac3r=\frac12-\frac{2}{q}+s_c$. We start showing
% the following inequality
% \begin{align}
% \left\|\nabla\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{r'}_x}\lesssim T^{\frac{1-s_c}{2}}\|\mathcal L_a^\frac12 u\|_{S(L^2;I)}\|\mathcal L_a^\frac{s_c}{2} u\|^{2\sigma}_{S(L^2;I)}.\label{est21}
% \end{align}
% Consider the $L^2$-admissible pairs $(\beta_-,\alpha_-)$ and $(\beta_+,\alpha_+)$ such that
% \begin{align}
% \alpha_{\pm}=\frac{3r(2\sigma+1)}{r(1+3\sigma\mp\varepsilon)-3}.
% \end{align}
% Note that, for $\varepsilon>0$ small enough, if necessary, we have $2<\alpha_{\pm}<3$. For $R>0$, using the Hlder's inequality, embedding Sobolev and equivalence of Sobolev norm we get
% %
% %
% \begin{align} \left\|\nabla\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{r'}_x(A^R_{\pm})}
% &\lesssim \left\|\left(\left\||x|^{-b}\right\|_{L^{\gamma_{\pm}}_x(A^R_{\pm})}+\left\||x|^{-b-1}\right\|_{L^{d_{\pm}}_x(A^R_{\pm})}\right)\|\nabla u\|_{L^{\alpha_{\pm}}_x}\|D^{s_c} u\|_{L^{\alpha_{\pm}}_x}^{2\sigma}\right\|_{L^{q'}_T}\\
% &\lesssim T^{\theta_{\pm}} \left(\left\||x|^{-b}\right\|_{L^{\gamma_\pm}_x(A^R_{\pm})}+\left\||x|^{-b-1}\right\|_{L^{d_{\pm}}_x(A^R_{\pm})}\right)\|\mathcal L_a^{\frac{1}{2}} u\|_{L^{\beta_{\pm}}_TL^{\alpha_{\pm}}_x}\|\mathcal L_a^{\frac{s_c}{2}} u\|_{L^{\beta_{\pm}}_TL^{\alpha_{\pm}}_x}^{2\sigma},
% \label{estbef1}
% \end{align}
% where
% \begin{align}
% \frac{3}{d_\pm}-1=\frac{3}{\gamma_{\pm}}=3+2\sigma s_c-3\left(\frac{1}{r}+\frac{2\sigma+1}{\alpha_\pm}\right)=b\pm \varepsilon
% \end{align}
% and
% \begin{align}
% 2\theta=-3\sigma- s_c+3\left(\frac{1}{r}+\frac{2\sigma+1}{\alpha_\pm}\right)=1-s_c\mp\varepsilon.
% \end{align}
% Now, since
% \begin{align}
% \left\||x|^{-b}\right\|_{L^{\gamma_{\pm}}_x(A^R_\pm)}=\left\||x|^{-b-1}\right\|_{L^{d_\pm}_x(A^R_\pm)}=R^{\pm\varepsilon}
% \end{align}
% so setting $R=T^\frac12$ we obtain the estimate \eqref{est21} from \eqref{estbef1}. To conclude, let $p=\frac{3r}{3+r(1-s_c)}$. From \eqref{condr}, we have that $2<p<6$. Thus, $(q,p)$ is $L^2$-admissible and $\dot H^{1-s_c,r'}\hookrightarrow L^{p'}$. Consequently, the equivalence of Sobolev norm \eqref{equiv0} and Sobolev embedding yield
% \begin{align}
% \left\|\mathcal L_a^{\frac{s_c}{2}}\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{p'}_x}&\lesssim \left\|\nabla\left(|x|^{-b}|u|^{2\sigma}u\right)\right\|_{L^{q'}_TL^{r'}_x},
% \end{align}
% which concludes the proof.
% \end{proof}
\begin{lemma}\label{est1} Let $a>0$, $0<b<2$, $\frac{2-b}{3}<\sigma<2-b$. Then, there exists $\delta \in (0,1)$ such that, for $I=[-T,T]$ we have
\begin{equation}
\left\|\mathcal |x|^{-b}|u|^{2\sigma}v \right\|_{S'(\dot H^{-s_c};I)} \lesssim \left[\left(T^{\frac{1-s_c}{2}}\|\mathcal L_a^\frac{1}{2} u\|_{S(L^2;I)}\right)^{1-\delta}\left(\|\mathcal L_a^\frac{s_c}{2} u\|_{S(L^2;I)}\right)^\delta\right]^{2\sigma}\|v\|_{S(\dot H^{s_c}; I)}.
\end{equation}
\end{lemma}
\begin{proof}
For a fixed $\sigma>0$, take $s\in (s_c,1)$ such that
\begin{equation}\label{sigma_cond}
\sigma<\frac{1-s_c}{s-s_c}.
\end{equation}
So, since $2\sigma(s-s_c)+1<3-2s_c$, we can find $\ell$ satisfying
$$\max\left\{\frac{6}{3-2s_c},\frac{6}{2\sigma(s-s_c)+1+\sigma-\varepsilon}\right\}< \ell<\frac{6}{2\sigma(s-s_c)+1+\varepsilon},
$$
for some $0<\varepsilon\ll 1$, such that $(m,\ell)$ is $\dot H^{-s_c}$-admissible and $(\tau,\ell)$ is $\dot H^{s_c}$-admissible. Also consider the pairs $(\nu_\pm,\mu_\pm)$ such that
\begin{align}
\mu_{\pm}=\frac{6\sigma \ell}{\ell(3+2\sigma s-b\mp\varepsilon)-6}.
\end{align}
It is straightforward to check that $2<\mu_\pm<3$, so we can define the $L^2$-admissible pairs $(\nu_-,\mu_-)$ and $(\nu_+,\mu_+)$. Now, setting
$$
\zeta_\pm=\sigma(s-s_c)\mp\frac{\varepsilon}{2},
$$
we get
\begin{align}
\frac{1}{\ell'}=\frac{b\pm\varepsilon}{3}+2\sigma\left(\frac{1}{\mu_\pm}-\frac{s}{3}\right)+\frac{1}{\ell}
\,\,\,\,
\mbox{and}
\,\,\,\,
\frac{1}{m'}=\zeta_\pm+\frac{2\sigma}{\nu_\pm}+\frac1\tau.
\end{align}
Then, applying H\"older's inequality and Sobolev embedding we have, using the same notation of the previous lemmas, that
\begin{align}
\left\||x|^{-b}|u|^{2\sigma}v\right\|_{L^{m'}_TL^{\ell'}_x(B^R_{\pm})}
&\lesssim T^{\zeta_{\pm}} R^{\pm \varepsilon}\|D^s u\|_{L^{\nu_{\pm}}_TL^{\mu_{\pm}}_x}^{2\sigma}\|v\|_{L^{\tau}_TL^{\ell}_x}.
\end{align}
Moreover, by interpolation and the equivalence of Sobolev norms \eqref{equiv0} we obtain
\begin{align}
\|D^s u\|_{L^{\nu_{\pm}}_TL^{\mu_{\pm}}_x}&\lesssim \|D u\|^{1-\delta}_{L^{\nu_{\pm}}_TL^{\mu_{\pm}}_x}\|D^{s_c} u\|^{\delta}_{L^{\nu_{\pm}}_TL^{\mu_{\pm}}_x}\\
&\lesssim \|\mathcal L^{\frac{1}{2}}_a u\|_{L^{\nu_{\pm}}_TL^{\mu_{\pm}}_x}^{1-
\delta}\|\mathcal L^{\frac{s_c}{2}}_a u\|_{L^{\nu_{\pm}}_TL^{\mu_{\pm}}_x}^{\delta},
\end{align}
for $\delta = (1-s)/(1-s_c)\in (0,1)$. Collecting the previous two estimates, we obtain
\begin{align}
\left\||x|^{-b}|u|^{2\sigma}v\right\|_{L^{m'}_TL^{\ell'}_x}
&\lesssim \left(T^{\zeta_{+}}R^{\varepsilon}+T^{\zeta_{-}}R^{-\varepsilon}\right)\left(\|\mathcal L_a^\frac{1}{2} u\|_{S(L^2;I)}^{1-\delta}\|\mathcal L_a^\frac{s_c}{2} u\|_{S(L^2;I)}^{\delta}\right)^{2\sigma}\|v\|_{S(\dot H^{s_c}; I)}.
\end{align}
Again setting $R=T^{\frac12}$, we find
\begin{equation}\label{T_cond}
T^{\zeta_{+}}R^{\varepsilon}+T^{\zeta_{-}}R^{-\varepsilon}=T^{\sigma(s-s_c)} = T^{\sigma(1-\delta)(1-s_c)},
\end{equation}
completing the proof.
%Now, if $\sigma\leq 1/2$, then the condition \eqref{sigma_cond} is satisfied for all $s\in (s_c,1)$ and taking $s\to 1$ we find $T^{\sigma(s-s_c)} \to T^{\sigma(1-s_c)}$. On the other hand, if $\sigma> 1/2$, then we can find $s_0\in (s_c,1)$ such that $\sigma(s_0-s_c)=(1-s_c)/2$. So, in this case, \eqref{sigma_cond} holds and the quantity in \eqref{T_cond} is given by $T^{\frac{1-s_c}{2}}$. This concludes the proof of Lemma \ref{est1}.
\end{proof}
%\begin{rem}
%Note that the assumption $s\in (s_c,1)$ is crucial in the proof of Lemma \ref{est1}. Indeed, if $s=s_c$, then $\theta_+=-\frac{\varepsilon}{2}<0$. On the other hand, if $s=1$, then the restriction on $p$ implies $\sigma<1$ and we can not cover all the intercritical range $\frac{2-b}{3}<\sigma<2-b$.
%\end{rem}
\subsection{Local theory}
Next, we use the above nonlinear estimates to obtain the local well-posedness in $ \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3) $.
\begin{proof}[Proof of Theorem \ref{LWP}]
Let
$$
X= \left( \bigcap_{(q,p)\in \mathcal{A}_0}L^q\left([-T,T];\dot{H}^{s_c,p}_a\cap \dot{H}^{1,p}_a\right)\right)\bigcap \left( \bigcap_{(\overline{q},\overline{p})\in \mathcal{A}_{s_c}}L^{\overline{q}}\left([-T,T]; L^{\overline p} \right)\right)$$ and
\begin{equation*}\label{NHs} %NHS = norma Hs
\|u\|_T=\|\mathcal L_a^{\frac12} u\|_{S\left(L^2;I\right)}+\|\mathcal L_a^{\frac{s_c}{2}} u\|_{S\left(L^2;I\right)}+\|u\|_{S\left(\dot H^{s_c};I\right)},
\end{equation*}
where $I=[-T,T]$.
For {$m>1$} and $T>0$ define the set
\begin{equation*}
S(m,T)=\{u \in X : \|u\|_T\leq m \}
\end{equation*}
with the metric
$$
d_T(u,v)=\|u-v\|_{S\left(\dot H^{s_c};I\right)}.
$$
By standards arguments $(S(m,T), d_T)$ is a complete metric space (see Cazenave \cite[Theorem 4.4.1]{cazenave} and also \cite[Appendix A]{CFG20}). We shall show that $G=G_{u_0}$ defined by
\begin{equation}\label{OPERATOR} % IE= integral equation
G(u)(t)=e^{-it\mathcal L_a}u_0+i\lambda \int_0^t e^{-i(t-s)\mathcal L_a}\left(|x|^{-b}|u(s)|^{2\sigma} u(s)\right)ds,
\end{equation}
has a fixed point in $S(m,T)$ for a suitable choice of $m$ and $T$. Indeed, it follows from the Strichartz inequalities in Lemma \ref{Lemma-Str1}, and also, from Lemmas \ref{est0} and \ref{est2}, that, for some $c>1$,
\begin{align}\label{esa}
\|\mathcal L_a^{\frac12}G(u)\|_{S(L^2;I)}
&\leq c \|\mathcal L_a^\frac12 u_0\|_{L^2}+cT^{\sigma(1-s_c)} \|\mathcal L_a^\frac12 u\|^{2\sigma+1}_{S(L^2;I)},
\end{align}
and
\begin{align}\label{esa0}
\|\mathcal L_a^{\frac{s_c}{2}}G(u)\|_{S(L^2;I)}
&\leq c \|\mathcal L_a^\frac{s_c}{2} u_0\|_{L^2}+cT^{\sigma(1-s_c)}\|\mathcal L_a^\frac12 u\|_{S(L^2;I)}^{2\sigma}\|\mathcal L_a^\frac{s_c}{2} u\|_{S(L^2;I)}.
\end{align}
Moreover, from Lemma \ref{Lemma-Str2} and Lemma \ref{est1}, we also have
$$
\|G(u)\|_{S\left(\dot H^{s_c};I\right)}\leq c \|\mathcal L_a^\frac{s_c}{2} u_0\|_{L^2}+c\left[\left(T^{\frac{1-s_c}{2}}\|\mathcal L_a^\frac{1}{2} u\|_{S(L^2;I)}\right)^{1-\delta}\left(\|\mathcal L_a^\frac{s_c}{2} u\|_{S(L^2;I)}\right)^\delta\right]^{2\sigma}\|u\|_{S(\dot H^{s_c}; I)}.
$$
%\footnote{Note that it is enough to choose $m$ such that $m\geq c\|u_0\|_{\dot{H}^{s_c}\cap \dot{H}^1 }$. Thus, we obtain the same existence time in \eqref{CTHs} for all initial data inside a ball.}
%By enlarging $c$ (depending on $\delta$ and $\sigma$), if necessary, we can use Young's inequality to write \eqref{esa0} as
%\begin{align}\label{esa0new}
% \|\mathcal L_a^{\frac{s_c}{2}}G(u)\|_{S(L^2;I)}
% &\leq c \|\mathcal L_a^\frac{s_c}{2} u_0\|_{L^2}+cT^{\sigma(1-s_c)}\|\mathcal L_a^\frac12 u\|_{S(L^2;I)}^{2\sigma}\|\mathcal L_a^\frac{s_c}{2} u\|_{S(L^2;I)}.
% \end{align}
Now, by fixing $m\geq \max\{1, 2c\|u_0\|_{\dot{H}^{s_c}_a\cap \dot{H}^1_a }\}$ and choosing $T\in (0,1)$ such that
\begin{equation}\label{CTHs} %CTHs=condi do T Hs
T^{\sigma(1-s_c)} < \left(\frac{1}{4c m^{2\sigma}}\right)^{\frac{1}{1-\delta}},
\end{equation}
we obtain that $G(u)\in S(m,T)$, if $u \in S(m,T)$ and therefore $G$ is well defined.
To prove that $G$ is a contraction we first recall the elementary inequality
\begin{equation}\label{nonlinearity}% first elementary inequality
\left||x|^{-b}|u|^{2\sigma} u-|x|^{-b}|v|^{2\sigma} v\right|\lesssim |x|^{-b}\left( |u|^{2\sigma}+ |v|^{2\sigma} \right)|u-v|.
\end{equation}
Then, again from Lemma \ref{Lemma-Str2} and Lemma \ref{est1}, we get for $u,v\in S(m,T)$ that
\begin{align*}
d_T(G(u),G(v)) \leq 2c T^{\sigma(1-s_c)(1-\delta)}m^{2\sigma} d_T(u,v).
\end{align*}
Therefore, for $T\in (0,1)$ satisfying \eqref{CTHs}, $G$ is a contraction on $S(m,T)$. Finally, by the contraction mapping principle we have a unique $u \in S(m,T)$ such that $G(u)=u$ and the proof is completed.
\end{proof}
\subsection{A Gagliardo-Nirenberg type inequality and Blow-up alternative}
We start recalling the celebrated Caffareli-Khon-Nirenberg inequality in $\mathbb{R}^3$.
\begin{prop}[Caffareli-Khon-Nirenberg inequality]\label{GNcrit2} There exists $C>0$ such that the following inequality holds for all $f \in \dot H^1(\mathbb{R}^3)\cap L^p(\Real^3)$
\begin{equation}\label{GNp}
\int |x|^{-b}|f|^{2\sigma+2} \, dx\leq C \|\nabla f\|^\theta_{L^2}\|f\|^{2\sigma+2-\theta}_{L^{p}}.
\end{equation}
if and only, if the following relations hold
\begin{align}
\frac{3-b}{3}=\frac{\theta}{6}+\frac{2\sigma+2-\theta}{p}
\end{align}
where $b<3$, $p\geq 1$ and $0<\theta<2\sigma+2$.
\end{prop}
\begin{proof}
This is the 3D version of the main result in Caffareli, Khon and Nirenberg \cite{CKN84}.
\end{proof}
\begin{rem}We note that:
\begin{itemize}
\item[1.] Taking $p=2$ in the inequality \eqref{GNp}, we have the well-known Gagliardo-Nirenberg inequality
\begin{align}\label{GNF}
\int|x|^{-b}|f|^{2\sigma+2}\,dx\leq C\|\nabla f\|_{L^2}^{2\sigma s_c+2}\|f\|_{L^2}^{2\sigma(1-s_c)},
\end{align}
where the best constant was established by the third author in \cite{Farah}.
\item[2.] For $p=\sigma_c=\frac{2N\sigma}{2-b}$, the inequality \eqref{GNp} is given by \begin{equation}\label{GNcrit}
\int |x|^{-b}|f|^{2\sigma+2} \, dx\leq C \|\nabla f\|^2_{L^2}\|f\|^{2\sigma}_{L^{\sigma_c}}.
\end{equation}
In \cite{CFG20}, joint with Guzm\'an, the sharp constant for above inequality was obtained.
\item[3.] Since $a>0$, we have that (see relation \eqref{equiv})
$$\|\mathcal L_a^{\frac12} u\|_{L^2}\sim \|\nabla u\|_{L^2}.$$
Thus, it follows from inequality \eqref{GNp} that there exists an optimal constant $K_{a}>0$ such that
\begin{equation}\label{GNacrit2}
\int |x|^{-b}|f|^{2\sigma+2} \, dx\leq \frac{\sigma+1}{K_a^{2\sigma}} \|\mathcal L_a^{\frac12}
f\|^2_{L^2}\|f\|^{2\sigma}_{L^{\sigma_c}}.
\end{equation}
\end{itemize}
We will refer to the constant $K_{a}$ later in this work (see Proposition \ref{prop1}).
\end{rem}
Next, we prove a blow-up alternative which follows as a consequence of the local theory obtained in Theorem \ref{LWP}.
\begin{prop}[Blow-up alternative]\label{Blowalt}
Let $a>0$, $0<b<\frac32$ and $\frac{2-b}{3}<\sigma<2-b$. If $u$ is a solution to the IVP \eqref{PVI} in $\dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)$ with finite maximal positive time of existence $0<T^*<+\infty$, then there exists a constant $c>0$ such that
%\begin{align}\label{BUA1}
%\|u(t)\|_{\dot H_a^{s_c}\cap \dot H_a^1}=\|u(t)\|_{\dot H_a^{s_c}}+
%{\color{blue}\|u(t)\|_{\dot H_a^1}\geq \frac{c}{(T^*-t)^{\frac{1-s_c}{2}}}, \,\,\, \textit{for all} \,\,\, t\in [0,T^*).}
%\end{align}
%Moreover, if $\displaystyle\limsup_{t\to T^*}\|u(t)\|_{\dot H_a^1}=\infty$
% supposing $\displaystyle\lim_{t\to T^*}\|\mathcal L_a^\frac12 u\|_{L^2}=\infty$,
%then there exist a constant $c>0$ such that the following lower bound is %satisfied
\begin{align}\label{BUA2}
\|\mathcal L_a^\frac12 u(t)\|_{L^2}\geq \frac{c}{(T^*-t)^{\frac{1-s_c}{2}}}, \,\,\, \textit{for all} \,\,\, t\in [0,T^*).
%\,\,\,\mbox{ or }\,\,\,\,\|\mathcal L_a^\frac{s_c}{2} u(t)\|_{L^2}> \frac{c}{(T^*-t)^{\frac{1-s_c}{4\sigma}}} ,
\end{align}
\end{prop}
\begin{proof}
%First note that the time of existence given by Theorem \ref{LWP} depends on the $\dot{H}^{s_c}_a\cap \dot{H}^1_a$-norm of the initial data. Now, take $\{t_k\}_{k\in\mathbb N}$ an arbitrary sequence of times such that $t_k\nearrow T^*$ and assume that there exists $M>0$ with $\|u(t_k)\|_{\dot H_a^{s_c}\cap \dot H_a^1}<M$ for all $k\in \mathbb N.$ Therefore, there exists a time $T(M)>0$ (independent of $k$) such that the solution $u\in C([t_k,t_k+T(M)],\dot H_a^{s_c}\cap \dot H_a^1)$ exists. This contradicts the maximality of $T^*$, since
% $$T(M)+t_k>T^*,$$
%for $k$ enough large. Thus, since $\{t_k\}_{k\in\mathbb N}$ is an arbitrary sequence, we have
%\begin{equation}\label{Blowalt2}
%\|u(t)\|_{\dot H_a^{s_c}\cap \dot H_a^1}\to \infty, \quad \mbox{as} \quad t\nearrow T^*.
%\end{equation}
%The proof of \eqref{BUA1} is a direct consequence of the well-posedness theory obtained in Theorem \ref{LWP} and was first introduced by Cazenave and Weissler \cite{CW90}. We refer the reader to \cite[Corollary 3.5]{CFG20} for a detailed proof in the {INLS} setting.
Assume,
%$\displaystyle\limsup_{t\to T^*}\|u(t)\|_{\dot H_a^1}=\infty$ and
by contradiction, that there exists a sequence $t_n\nearrow T^{\ast}$ such that, for all $n$,
\begin{equation}\label{bound_0}
(T^*-t_n)^{\frac{1-s_c}{2}}\|\mathcal L_a^\frac12 u(t_n)\|_{L^2} < \frac{1}{n}.
\end{equation}
Using the same notation as in the proof of Lemma \ref{est0}, we define
\begin{equation}
X_n(t) =\|\mathcal L_a^{\frac12}u\|_{L^{\beta_+}_{[t_n,t]}L^{\alpha_+}_x} + \|\mathcal L_a^{\frac12}u\|_{L^{\beta_-}_{[t_n,t]}L^{\alpha_-}_x}.
\end{equation}
From the Duhamel formula and estimate \eqref{bound_0}, for $t \in [t_n, T^*)$, there exists $C_0>0$ such that
\begin{equation}\label{ContArg}
%\|\mathcal L_a^{\frac12}u\|_{S(L^2;[t_n,t])}
\|\mathcal L_a^{\frac12}u\|_{L^{\infty}_{[t_n,t]}L^{2}_x} +
X_n(t)\leq
C_0\|\mathcal L_a^\frac12 u(t_n)\|_{L^2} + \frac{C_0}{(n \|\mathcal L_a^\frac12 u(t_n)\|_{L^2})^{2\sigma}} [X_n(t)]^{2\sigma+1}.
\end{equation}
Consider the function $f(x)=x-C_0\|\mathcal L_a^\frac12 u(t_n)\|_{L^2} - \frac{C_0}{(n \|\mathcal L_a^\frac12 u(t_n)\|_{L^2})^{2\sigma}} x^{2\sigma+1}$. A simple computation revels that it has a global maximum at $x_n^{\ast}=\frac{n\|\mathcal L_a^\frac12 u(t_n)\|_{L^2}}{[C_0(2\sigma+1)]^{1/2\sigma}}$, $f(x_n^{\ast})=\left(\frac{2\sigma n}{(2\sigma+1)[C_0(2\sigma+1)]^{1/2\sigma}}-C_0\right)\|\mathcal L_a^\frac12 u(t_n)\|_{L^2}$ and $f(0)<0$. Thus, for $n_0> \frac{\left[C_0(2\sigma+1)\right]^{\frac{2\sigma+1}{2\sigma}}}{2\sigma}$ we have that $f(x_{n_0}^{\ast})>0$.
Now, from \eqref{ContArg} we have that $f(X_{n_0}(t))\leq 0$, for all $t \in [t_{n_0}, T^*)$. A continuity argument then implies that $X_{n_0}(t)\leq x_{n_0}^{\ast}$ and therefore, from inequality \eqref{ContArg}, we deduce
\begin{eqnarray}\label{bound_1}
\|\mathcal L_a^{\frac12}u\|_{L^{\infty}_{[t_{n_0},T^*)}L^2_x}
&\leq& C_0\|\mathcal L_a^\frac12 u(t_{n_0})\|_{L^2}+ \frac{C_0}{(n_0 \|\mathcal L_a^\frac12 u(t_{n_0})\|_{L^2})^{2\sigma}} (x_{n_0}^*)^{2\sigma+1}\\
& =& \left(1+\frac{n_0}{[C_0(2\sigma+1)]^{(2\sigma+1)/2\sigma}} \right)C_0\|\mathcal L_a^\frac12 u(t_{n_0})\|_{L^2},
%\leq\frac{n_0\|\mathcal L_a^\frac12 u(t_{n_0})\|_{L^2}}{[C_0(2\sigma+1)]^{1/2\sigma}},
\end{eqnarray}
which implies
%$\|u\|_{L^{\infty}_{[0,T^*)}\dot H_a^1}
$\|\mathcal{L}_a^{\frac{1}{2}}u\|_{L^{\infty}_{[0,T^*)}L^2_x} <\infty$, by continuity. % and the definition of $S(L^2;[t_{n_0},T^*))$.
On the other hand, the bound $X_{n_0}(t)\leq x_{n_0}^{\ast}$ and assumption \eqref{bound_0} also implies
\begin{equation}
(t - t_{n_0})^{\frac{1-s_c}{2}}X_{n_0}(t)< \frac{1}{[C_0(2\sigma+1)]^{1/2\sigma}}, \,\,\mbox{for all}\,\, t\in [0,T^*).
\end{equation}
Moreover, again from Duhamel and the proof of estimate \eqref{esa0}, we can write, for $t_{n_0}\leq t < T^*$,
\begin{align}
\|\mathcal L_a^{\frac{s_c}2}u\|_{S(L^2;[t_{n_0},t])} &\leq
C_0\|\mathcal L_a^\frac{s_c}2 u(t_{n_0})\|_{L^2} + C_0 \left[(t-t_{n_0})^{\frac{1-s_c}{2}}X_{n_0}(t)\right]^{2\sigma}\|\mathcal L_a^{\frac{s_c}2}u\|_{S(L^2;[t_{n_0},t])}\\
&\leq C_0\|\mathcal L_a^\frac{s_c}2 u(t_{n_0})\|_{L^2} + \frac{1}{2\sigma+1}\|\mathcal L_a^{\frac{s_c}2}u\|_{S(L^2;[t_{n_0},t])}
.
\end{align}
We can then absorb the last term of the right-hand side in the left-hand side and let $t \nearrow T^*$ to obtain
\begin{equation}
\|\mathcal L_a^{\frac{s_c}2}u\|_{S(L^2;[t_{n_0},T^{*}))} \leq
\frac{(2\sigma+1) C_0}{2\sigma}\|\mathcal L_a^\frac{s_c}2 u(t_{n_0})\|_{L^2},
\end{equation}
which is a contradiction, since $\|\mathcal{L}_a^{\frac{1}{2}}u\|_{L^{\infty}_{[0,T^*)}L^2_x} + \|\mathcal{L}_a^{\frac{s_c}{2}}u\|_{L^{\infty}_{[0,T^*)}L^2_x}$ cannot be finite, otherwise $T^{*} = +\infty$.
\begin{rem}
The constant $c>0$ in the previous proposition can be chosen independently of $u$. Indeed, from its proof,one can choose
\begin{equation}
c = \frac{2\sigma}{\left[C_0(2\sigma+1)\right]^{\frac{2\sigma+1}{2\sigma}}}.
\end{equation}
\end{rem}
\begin{rem}
The previous proposition can be adapted to the INLS equation \eqref{INLS}, showing that the lower bound
$$
\|\mathcal \nabla u(t)\|_{L^2}\geq \frac{c}{(T^*-t)^{\frac{1-s_c}{2}}}, \,\,\, \textit{for all} \,\,\, t\in [0,T^*),
$$
always hold for $\dot{H}^{s_c}\cap \dot{H}^1$ blow-up solution in the INLS setting, improving our previous results \cite[Theorem 1.2 and Corollary 1.3]{CF20} and \cite[Theorem 1.4 and Corollary 1.5]{CF23}.
\end{rem}
%Next, we estimate the norm $\|\mathcal L_a^{\frac{s_c}{2}}u\|_{S(L^2;[t_{n_0},t])}$. Indeed, by similar computations as the ones used to obtain \eqref{esa0} and inserting the bounds \eqref{bound_0} and \eqref{bound_1}, we have for $n_0$ large enough and $t \in [t_{n_0}, T)$
%\begin{equation}\label{ContArg2}
% \|\mathcal L_a^{\frac{s_c}{2}}u\|_{S(L^2;[t_{n_0},t])}\leq
% C_0\|\mathcal L_a^{\frac{s_c}{2}} u(t_{n_0})\|_{L^2} + \frac{C_0}{[C_0(2\sigma+1)]^{1/2\sigma}}\|\mathcal L_a^{\frac{s_c}{2}}u\|_{S(L^2;[t_{n_0},t])}^{2\sigma}.
%\end{equation}
%
%We need to study all the possible values for $2\sigma$. Indeed, if $2\sigma=1$, we can absorb the last term in the RHS in the LRS to deduce
%$$
%\|\mathcal L_a^{\frac{s_c}{2}}u\|_{S(L^2;[t_{n_0},T))}\leq \frac{C_0(2\sigma+1)}{2\sigma}
%\|\mathcal L_a^{\frac{s_c}{2}} u(t_{n_0})\|_{L^2}.
%$$
%Now, if $2\sigma<1$, then, since the RHS of the inequality \eqref{ContArg2} is sublinear, $\|\mathcal L_a^{\frac{s_c}{2}}u\|_{S(L^2;[t_{n_0},T))}$ is finite. Finally, if $2\sigma>1$, ???
%
%we consider the function $g(y)=y-C_0\|\mathcal L_a^{\frac{s_c}{2}} u(t_{n_0})\|_{L^2} - \frac{C_0}{[C_0(2\sigma+1)]^{1/2\sigma}}y^{2\sigma}$. Setting $y(t)=\|\mathcal L_a^{\frac{s_c}{2}}u\|_{S(L^2;[t_{n_0},t])}$, from \eqref{ContArg2} we have that $g(y(t))\leq 0$, for all $t \in [t_{n_0}, T)$ and since $g(y(t_n))=g(0)< 0$ a continuity argument also implies that $\|\mathcal L_a^{\frac{s_c}{2}}u\|_{S(L^2;[t_{n_0},T))}$ is finite. In all the three possibilities we reach a contradiction with \eqref{Blowalt2} concluding the proof of Proposition \ref{Blowalt}.
\end{proof}
\section{Blow-up solutions in $\dot H_a^{s_c}\cap \dot H_a^1$}\label{sec4}
\subsection{A virial-type estimate}
Consider $u\in C([0,\tau_*]: \dot H_a^{s_c}\cap \dot H_a^1)$ a solution to \eqref{PVI} and define the function
\begin{align}\label{virial}
z(t)=\displaystyle\int\phi|u(t)|^2\,dx,
\end{align}
where $\phi$ is cut-off function in $\Real^3$.
In \cite{LUCCASCARLOS}, the second author and Guzm\'an obtained
\begin{equation}\label{zR'2}
z'(t)=2\,\mbox{Im}\int \nabla\phi\cdot\nabla u(t)\overline{u}(t)\,dx
\end{equation}
and
\begin{align}\label{zR''}
z''(t)=&4\,\mbox{Re} \sum_{j,k=1}^{3}\int \partial_ju(t)\,\partial_k\overline u(t)\,\partial^2_{jk}\phi\,dx-\int |u(t)|^2 \Delta^2\phi\,dx-2a\int \nabla (|x|^{-2})\cdot \nabla \phi|u(t)|^{2}\,dx\nonumber\\
&-\frac{2\sigma}{\sigma+1}\int|x|^{-b}|u(t)|^{2\sigma+2}\Delta\phi\,dx+\frac{2}{\sigma+1}\int\nabla\left(|x|^{-b}\right)\cdot \nabla\phi|u(t)|^{2\sigma+2}\,dx.
\end{align}
Set $\phi$ to be a non-negative radial function $\phi\in C^{\infty}_0(\Real^3)$, such that
\begin{align}\label{phi}
\phi(x)=
\left\{
\begin{array}{ll}
|x|^2,&\mbox{ se }|x|\leq 2\\
0,&\mbox{ se }|x|\geq 4
\end{array}
\right.
\end{align}
satisfying
\begin{align}\label{nablaphi}
\phi(x)\leq c|x|^2, \quad |\nabla \phi (x)|^2\leq c\phi(x) \quad \mbox{and} \quad \partial_r^2\phi(x)\leq 2, \quad \mbox{for all } |x|>2,
\end{align}
with $r=|x|$. Then, define $\phi_R(x)=R^2\phi\left(\frac{x}{R}\right)$. %(see Appendix \ref{appB} for the construction of such function $\phi$).
Since $\phi$ is radial, we have
\begin{align}\label{virial1}
z'(t)=2\,\mbox{Im}\int \partial_r\phi_R\frac{x\cdot \nabla u(t)}{r}\overline{u}(t)\,dx
\end{align}
and
\begin{align}\label{virial2}
z_R ''(t)=&4\int \frac{\partial_r\phi_R}{r}|\nabla u(t)|^2\,dx+4\int \left(\frac{\partial_r^2\phi_R}{r^{2}}-\frac{\partial_r \phi_R}{r^3}\right)|x\cdot \nabla u(t)|^2\,dx+4a\int \frac{\partial_r \phi_R}{r}|x|^{-2}|u(t)|^2\,dx \nonumber\\
&-\int|u(t)|^2 \Delta^2\phi_R
\,dx-\frac{2\sigma}{\sigma+1}\int \left[\partial^2_r\phi_R +\left(N-1+\frac{b}{\sigma}\right)\frac{\partial_r \phi_R}{r}\right]|x|^{-b}|u(t)|^{2\sigma+2}\,dx,
\end{align}
where $\partial_r$ denotes the derivative with respect to $r=|x|$.
In the following lemma we prove a virial type estimate adapted to the $(\mbox{INLS})_a$ setting.
\begin{lemma}\label{lemintradial}
%Let $N\geq3$, $0<b<\min\{\frac{N}{2},2\}$ and $\frac{2-b}{N}\leq\sigma<\min\left\{\frac{2-b}{N-2},\frac{2}{N}\right\}$.
If $u\in C([0,\tau_*]: \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3) )$ is a solution to \eqref{PVI} with initial data $u(0)=u_0$, then there exists $c>0$ depending only on $N, \sigma, b$ such that for all $R>0$ and $t\in[0, \tau_\ast]$ we have
\begin{align}
8\sigma s_c\|\mathcal L_a^{\frac12}
u(t)\|^2_{L^2}&+z''_R(t)-16(\sigma s_c+1)E[u_0]\\
\lesssim &\left(\frac{1}{R^2}\int_{2R\leq |x|\leq 4R}|u(t)|^2\,dx+a\int_{|x|\geq R} |x|^{-2}|u(t)|^2\,dx+\int_{|x|\geq R}|x|^{-b}|u(t)|^{2\sigma+2}\,dx\right).\label{estintradial}
\end{align}
\end{lemma}
\begin{proof} Define
\begin{align}
P[u]&=\int |\nabla u|^{2}\,dx+a\int |x|^{-2}|u|^2\,dx-\frac{3\sigma+b}{2\sigma+2}\int |x|^{-b}|u|^{2\sigma+2}\,dx,\label{P}
\end{align}
then can rewrite $z_R''(t)$ in \eqref{virial2} as
\begin{align}\label{z''K}
z_R''(t)=8P[u(t)]+K_1+K_2+K_3+K_4,
\end{align}
where
\begin{align}
%P(u)=&\|\nabla u\|_{L^2}^2-\frac{N(2\sigma)+2b}{2(2\sigma+2)}\int |x|^{-b}|u|^{2\sigma+2}\,dx,\\
K_1=&4\int \left(\frac{\partial_r\phi_R}{r}-2\right)|\nabla u(t)|^2\,dx+4\int \left(\frac{\partial^2_r\phi_R}{r^2}-\frac{\partial_r \phi_R}{r^3}\right)|x\cdot \nabla u(t)|^2\,dx,\label{K1}\\
K_2=&4a\int \left(\frac{\partial_r \phi_R}{r}-2\right)|x|^{-2}|u(t)|^{2}\,dx\label{K2}\\
K_3=&-\frac{2\sigma}{\sigma+1}\int\left[\partial^2_r\phi_R+\left(2+\frac{b}{\sigma}\right)\frac{\partial_r \phi_R}{r}-6-\frac{2b}{\sigma}\right]|x|^{-b}|u(t)|^{2\sigma+2}\,dx,\label{K3}\\
%&+\frac{4\left(N\sigma+b\right)}{\sigma+1}\int|x|^{-b}|u|^{2\sigma+2}\,dx,\\
K_4=&-\int|u(t)|^2\Delta^2\phi_R\,dx\label{K4}.
\end{align}
First note, by definition of $\phi_R$, that
\begin{align}
K_4\lesssim \frac{1}{R^{2}}\int_{2R\leq |x|\leq 4R} |u(t)|^{2}\,dx\label{K_4}.
\end{align}
Moreover, if $0\leq |x|\leq 2R$, then
\begin{align}
\partial_r\phi_R(x)=2|x|=2r,\,\,\,\,\,\,\partial^2_r\phi_R(x)=2,
\end{align}
which implies
\begin{align}
\mbox{supp} \left[\partial^2_r\phi_R+\left(2+\frac{b}{\sigma}\right)\frac{\partial_r\phi_R}{r}-6-\frac{2b}{\sigma}\right]\,\cup\,\, \mbox{supp} \left[\frac{\partial_r\phi_R}{r}-2\right]\subset (2R,\infty).
\end{align}
Therefore, there exists $c>0$ such that
\begin{align}
K_3\lesssim& \int_{|x|\geq R}|x|^{-b}|u(t)|^{2\sigma+2}\,dx\label{K_3}
\end{align}
and
\begin{align}
K_2\lesssim a\int_{|x|\geq R} |x|^{-2}|u(t)|^2\,dx.\label{K_2}
\end{align}
Now, we set
\begin{align}
\Omega=\left\{x\in \mathbb R^3;\,\,\frac{\partial^2_r\phi_R(x)}{r^2}-\frac{\partial_r\phi_R(x)}{r^3}\leq 0\right\}.
\end{align}
Since $\partial^2_r\phi_R\leq 2$ it follows that $\partial_r\phi_R(|x|)\leq 2|x|$, for all $x\in \mathbb R^3$. Then, using the Cauchy-Schwartz inequality, we get
\begin{align}
K_1=&\,\,4\int\left(\frac{\partial_r\phi_R}{r}-2\right)|\nabla u(t)|^2\,dx+4\int\left(\frac{\partial_r^2\phi_R}{r^2}-\frac{\partial_r\phi_R}{r^3}\right) |x\cdot \nabla u(t)|^2\,dx\nonumber\\
=&\,\,4\int_{\Omega} \left(\frac{\partial_r\phi_R}{r}-2\right)|\nabla u(t)|^2\,dx+4\int_{\Omega} \left(\frac{\partial_r^2\phi_R}{r^2}-\frac{\partial_r\phi_R}{r^3}\right) |x\cdot \nabla u(t)|^2\,dx\nonumber\\
&+4\int_{\mathbb R^3\backslash\Omega} \left(\frac{\partial_r\phi_R}{r}-2\right)|\nabla u(t)|^2\,dx+4\int_{\mathbb R^3\backslash\Omega} \left(\frac{\partial_r^2\phi_R}{r^2}-\frac{\partial_r\phi_R}{r^3}\right) |x\cdot \nabla u(t)|^2\,dx\nonumber\\
\leq&\,\, 4\int_{\mathbb R^3\backslash\Omega} \left(\frac{\partial_r\phi_R}{r}-2\right)|\nabla u(t)|^2\,dx+4\int_{\mathbb R^3\backslash\Omega} \left(\partial_r^2\phi_R-\frac{\partial_r\phi_R}{r}\right)\frac{|x|^2}{r^2}|\nabla u(t)|^2\,dx\nonumber\\
=&\,\, 4\int_{\Real^3\backslash \Omega} \left(\partial_r^2\phi_R-2\right)|\nabla u(t)|^2\,dx\leq 0,\label{K_1}
\end{align}
where in the last inequality we have used the assumption \eqref{nablaphi}.
Collecting \eqref{K_4}, \eqref{K_3}, \eqref{K_2} and \eqref{K_4}, there exists $c>0$ such that
\begin{align}
z_R''(t)-8P[u(t)]\lesssim \frac{1}{R^2}\int_{2R\leq |x|\leq 4R}|u(t)|^2\,dx+ a\int_{|x|\geq R} |x|^{-2}|u(t)|^2\,dx+\int_{|x|\geq R}|x|^{-b}|u(t)|^{2\sigma+2}\,dx\label{zR8Q}.
\end{align}
Finally, from the energy conservation \eqref{Energy}, we can write
\begin{align}
P[u(t)]= -\sigma s_c\|\mathcal L_a^{\frac12}
u(t)\|_{L^2}^2+2(\sigma s_c+1)E[u_0]
\end{align}
and combining the last two relations, we deduce the estimate \eqref{estintradial}.
\end{proof}
%\begin{align}
% 8\sigma s_c\|\nabla u(t)\|_{L^2}^2+z_R''(t)-16(\sigma s_c+1)E[u_0]\leq c\left(\frac{1}{R^2}\int_{2R\leq |x|\leq 4R}|u(t)|^2\,dx+\int_{|x|\geq R}|x|^{-b}|u(t)|^{2\sigma+2}\,dx\right).
%\end{align}
%For future reference we state this result in the next lemma.
%\subsection{Existence of non-radial blow-up solutions in $H^1$}
%In this section, we extend the result of \citet{LUCCASCARLOS} for non-radial case. The main tool is the inequality in Lemma \ref{lemintradial}. On the condictions \eqref{ME} and \eqref{MC}, \citet{LUCCASCARLOS} showed the followings lemmas.
%\begin{lemma}[\cite{LUCCASCARLOS}]
%Let $N \geq 3$, $0<b<2$, $\frac{2-b}{N}<\sigma<\frac{2-b}{N-2}$ and $f \in H_a^1(\Real^N).$ Assume that, for some $\delta_0 > 0$,
%\begin{align}
% M[f]^\frac{1-s_c}{s_c}E_a[f] \leq (1 - \delta_0)M[Q]^\frac{1-s_c}{s_c}E_a[Q],
%\end{align}
%and
%\begin{align}
% \|f\|^{\frac{1-s_c}{s_c}}_{L^2}\|\mathcal L_a^{\frac12}
%f\|_{L^2}>\|Q\|_{L^2}^{\frac{1-s_c}{s_c}}\|\mathcal L_a^{\frac12}
% Q\|_{L^2}
%\end{align}
%then there exists $\delta = \delta(\delta_0, N, \sigma, Q, a, b)>0$ such that
%\begin{align}
%\|f\|^{\frac{1-s_c}{s_c}}_{L^2}\|\mathcal L_a^{\frac12}
%f\|_{L^2}\geq (1+\delta) \|Q\|_{L^2}^{\frac{1-s_c}{s_c}}\|\mathcal L_a^{\frac12}
%Q\|_{L^2}.
%\end{align}
%\end{lemma}
%\begin{lemma}[\cite{LUCCASCARLOS}]\label{LC2} Under the conditions of the previous lemma, for some $\eta=\eta(\delta_0, Q, M(f))>0$ and $\varepsilon=\varepsilon(\delta_0, Q)>0$,
%\begin{align}
%(1+\varepsilon)\int |\mathcal L_a^{\frac12}
% u(t)|^2\,dx-\frac{N\sigma+b}{2\sigma+2}\int |x|^{-b}|u|^{2\sigma+2}\,dx\leq -\eta<0.
%\end{align}
%\end{lemma}
%\begin{proof}[Proof of Theorem \ref{crith1}] Let $R,\varepsilon>0$ real numbers to be chosen later. Using the Gagliardo-Nirenberg inequality \eqref{GNF} with $b=0$,% \begin{align}\label{RegGN}
%% \int_{|x|\geq R}|x|^{-b}|u(t)|^{2\sigma+2}\,dx&\leq c\frac{1}{R^b}\|u(t)\|_{L^{\frac{2N}{N-2}}}^{\sigma N}\|u(t)\|_{L^2}^{2\sigma+2-\sigma N}\\
%% &\leq c\frac{1}{R^b}\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^{\sigma N}\|u(t)\|_{L^2}^{2\sigma+2-\sigma N}.
%% \end{align}
%% Then,
%Young inequality and mass conservation \eqref{Mass} we conclude, for some constant $C_{\varepsilon}>0$, that
% \begin{align}
% \int_{|x|\geq R}|x|^{-b}|u(t)|^{2\sigma+2}\,dx&\leq \frac{c}{R^{b}}\|\mathcal L_a^{\frac12}
% u\|_{L^{2}}^{\sigma N}\|u_0\|^{2\sigma+2-\sigma N}_{L^2} \leq 8\varepsilon \|\mathcal L_a^{\frac12}
% u\|_{L^2}^2+C_{\varepsilon}\frac{\|u_0\|_{L^2}^{\frac{2(2\sigma+2-\sigma N)}{2-\sigma N}}}{R^{\frac{2b}{2-\sigma N}}},\label{epsgradR}
% \end{align}
% where we have used that $\sigma <2/N$.
%Collecting the inequalities \eqref{estintradial} and \eqref{epsgradR}, for all $\varepsilon>0$, we obtain
%\begin{align}
% z''_R(t)\leq 8\left((1+\varepsilon)\int |\mathcal L_a^{\frac12}
% u(t)|^2\,dx-\frac{N\sigma+b}{2\sigma+2}\int |x|^{-b}|u|^{2\sigma+2}\,dx\right)+C_{\varepsilon}\frac{\|u_0\|_{L^2}^{\frac{2(2\sigma+2-\sigma N)}{2-\sigma N}}}{R^{\frac{2b}{2-\sigma N}}}+c\frac{\|u_0\|_{L^2}^2}{R^2}.
%\end{align}
%%On the other hand, if $u_0$ satisfies the condictions \eqref{ME} and \eqref{MC}, then by Lemma \ref{LC2} for some $\eta>0$ and $\varepsilon>0$
%%\begin{align}
%%(1+\varepsilon)\int |\mathcal L_a^{\frac12}
% u(t)|^2\,dx-\frac{N\sigma+b}{2\sigma+2}\int |x|^{-b}|u|^{2\sigma+2}\,dx\leq -\eta<0.
%%\end{align}
%Therefore, for $\varepsilon$ small enough and $R$ suitable, from Lemma \ref{LC2} we have
%\begin{align}
% z''_R(t)\leq -\eta
%\end{align}
%which implies blowup in finite positive and negative times.
%\end{proof}
%\subsection{Upper bound for the blow-up rate}
%%Now, we are able to prove Theorem \ref{thmblowdisper}. This is an extension of the result obtained by Merle, Rapha\"{e}l and Szeftel \cite[Theorem 1.1]{MRS2014} for the classical NLS equation to the $(\mbox{INLS})_a$ model and is a direct consequence of Lemma \ref{lemintradial}.
%
%As an application of Lemma \ref{lemintradial} we obtain our first main result.
%
%\begin{proof}[Proof of Theorem \ref{thmblowdisper}] Let $R,\varepsilon>0$ real numbers to be chosen later.
%First, we consider $\sigma<\min\left\{\frac{2-b}{N-2},\frac{2}{N}\right\}$. Combining the inequality \eqref{estintradial}, energy and mass conservation \eqref{Energy}-\eqref{Mass} and the inequalities \eqref{R2M}-\eqref{epsgradR} we deduce that
% \begin{align}
% 8\sigma s_c\int |\mathcal L_a^{\frac12}
% u(t)|^2
% \,dx+z''_R(t)\leq C_{u_0}\left(1+\frac{1}{R^2}+8\varepsilon\int|\mathcal L_a^{\frac12}
% u(t)|^2\,dx+\frac{C_{\varepsilon}}{R^{\frac{2b}{2-\sigma N}}}\right).
% \end{align}
%Fix $\varepsilon>0$ small enough. Therefore, there exists a universal constant $C=C(u_0, \sigma, b, N)>0$ such that
%\begin{align}
% \sigma s_c\int|\mathcal L_a^{\frac12}
% u(t)|^2\,dx+z''_R(t)\leq C_{u_0}\left(1+\frac{1}{R^2}+\frac{1}{R^{\frac{2b}{2-\sigma N}}}\right).
% \end{align}
%Now, since $\frac{2b}{2-\sigma N}>2$ (or $\sigma>\frac{2-b}{N}$) , if $R\ll1$, then
% \begin{align}\label{ctR1}
% \sigma s_c\int|\mathcal L_a^{\frac12}
% u(t)|^2\,dx+z''_R(t)\leq \frac{C_{u_0}}{R^{\frac{2b}{2-\sigma N}}}.
% \end{align}
%Define $\beta=\frac{2-\sigma N}{b}$. Integrating \eqref{ctR1} from $t$ to $\tau $ and then from $t$ to $t_2$, in time, we get
%\begin{align}
% \int_{t}^{t_2}(t_{2}-\tau)\|\mathcal L_a^{\frac12}
% u(\tau&)\|_{L^2}^2\,d\tau +\frac{1}{2}\int\phi_R|u(t_2)|^2\,dx\\
% &\leq C_{u_0} \frac{(t_2-t)^2}{2R^{\frac{2}{\beta}}}+\frac{1}{2}\int\phi_R|u(t)|^2\,dx+(t_{2}-t)\textit{Im}\int \nabla \phi_R\nabla u \overline{u}(t)\,dx\nonumber\\
% &\leq C_{u_0}\left(\frac{(t_2-t)^2}{R^{\frac{2}{\beta}}}+R^2+R(t_2-t)\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2} \right),
%\end{align}
%where we have used Cauchy-Schwarz inequality, the mass conservation \eqref{Mass} and the fact that $a>-\frac{(N-2)^2}{4}$ in the last inequality. Taking $t_2\to T^{\ast}$ in the above inequality, we conclude that the first integral on the left hand side converges and
%\begin{align}\label{cotaint}
% \int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\mathcal L_a^{\frac12}
% u(\tau&)\|_{L^2}^2\,d\tau
% \leq C_{u_0}\left(\frac{(T^{\ast}-t)^2}{R^{\frac{2}{\beta}}}+R^2+R(T^{\ast}-t)\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}\right).
%\end{align}
%Thus, choosing $R\ll1$ so that the first two terms on the right hand side are equals, more precisely
%\begin{align}
% R(t)=(T^{\ast}-t)^{\frac{\beta}{1+\beta}},
%\end{align}
%the inequality \eqref{cotaint} can be rewritten as
%\begin{align}
% \int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\mathcal L_a^{\frac12}
% u(\tau)\|_{L^2}^2\,d\tau
% &\leq C_{u_0}\left((T^{\ast}-t)^{\frac{2\beta}{1+\beta}}+(T^{\ast}-t)^{\frac{\beta}{1+\beta}}(T^{\ast}-t)\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}\right)\\
% &\leq C_{u_0} (T^{\ast}-t)^{\frac{2\beta}{1+\beta}}+(T^{\ast}-t)^2\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^2.\label{cotaint2}
%\end{align}
%where we have used Young's inequality in the last step.
%
%Define
%\begin{align}
% h(t)=\int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\mathcal L_a^{\frac12}
% u(\tau)\|_{L^2}^{2}\,d\tau.
%\end{align}
%Note that inequality \eqref{cotaint2} implies that $h(0)\leq C_{u_0}$ and moreover
%\begin{align}
% h(t)\leq C_{u_0}(T^{\ast}-t)^{\frac{2\beta}{1+\beta}}-(T^{\ast}-t)h'(t),
%\end{align}
%that is,
%\begin{align}
% \left(\frac{h(t)}{T^{\ast}-t}\right)'=\frac{1}{(T^{\ast}-t)^2}((T^{\ast}-t)h'(t)+h(t))\leq C_{u_0} \frac{1}{(T^{\ast}-t)^{\frac{2}{1+\beta}}}.
%\end{align}
%Integrating the above inequality from $0$ to $t$,
%\begin{align}
% \frac{h(t)}{T^{\ast}-t}\leq \frac{h(0)}{T^{\ast}}+ \frac{C_{u_0}}{(T^{\ast}-t)^{\frac{1-\beta}{1+\beta}}}-\frac{C_{u_0}}{(T^{\ast})^{\frac{1-\beta}{1+\beta}}}\leq \frac{C_{u_0}}{(T^{\ast}-t)^{\frac{1-\beta}{1+\beta}}},
% %\leq \frac{C_{u_0}}{T^{\ast}-t}\left((T^{\ast}-t)+(T^{\ast}-t)^{\frac{2\beta}{1+\beta}}\right)
%\end{align}
%for $t$ close enough to $T^{\ast}$, since $0<\beta<1 $. Therefore
%\begin{align}
% h(t)\leq C_{u_0}(T^{\ast}-t)^{\frac{2\beta}{1+\beta}},
%\end{align}
%which implies \eqref{blowupestdisp}.
%
%Now, if the initial data $u_0\in H^1(\Real^N)$ is radial, since $a>-\frac{(N-2)^2}{4}$, we have by the Strauss inequality for radial functions \cite{Strauss} that (see, for instance, \cite[Lemma 4.1]{FG20} for a proof)
%\begin{align}
% \|u\|_{L^{\infty}(|x|\geq R)}\leq c\frac{\|\mathcal L_a^{\frac12}
% u\|_{L^2}^{\frac{1}{2}}\|u\|_{L^2}^{\frac{1}{2}}}{R^{\frac{N-1}{2}}}.
%\end{align}
%Then, by the mass conservation \eqref{Mass} and Young's inequality
%\begin{align}
% \int_{|x|\geq R}|x|^{-b}|u|^{2\sigma+2}\,dx&\leq \frac{1}{R^{b}}\|u\|_{L^{\infty}(|x|\geq R)}^{2\sigma}\|u\|^{2}_{L^2(|x|\geq R)}\leq \frac{C(u_0)}{R^{\sigma(N-1)+b}}\|\mathcal L_a^{\frac12}
% u\|_{L^2}^{\sigma}\|u_0\|_{L^2}^{\sigma+2}\\& \leq \varepsilon \|\mathcal L_a^{\frac12}
% u\|_{L^2}^2+\frac{C(u_0,\sigma,\varepsilon)}{R^{\frac{2\sigma(N-1)+2b}{2-\sigma}}},\label{case2}
%\end{align}
%where we have used that $\sigma<2$. Hence, fixing $\varepsilon>0$ small enough, by laws conservation \eqref{Mass}, \eqref{Energy} and injecting the inequality above in \eqref{estintradial}, we get
% \begin{align}\label{ctR2}
% \sigma s_c\int|\mathcal L_a^{\frac12}
% u(t)|^2\,dx+z''_R(t)\leq \frac{C}{R^{\frac{2\sigma(N-1)+2b}{2-\sigma}}},
% \end{align}
%if $R\ll 1$ (see that $\sigma(N-1)+b>2-\sigma$). Following the same steps after equation \eqref{ctR1}, we obtain \eqref{blowupestdisp}, and thus, we conclude the proof.
%
%%To simplify the notation let $\beta=\frac{2-\sigma N}{b}$ (in this case $0<\beta<1$). Integrating \eqref{ctR} from $t$ to $s$ in a variable $\tau$ we get
%%\begin{align}
%% \sigma s_c\int_{t}^{s}\|\nabla u(\tau)\|_{L^2}^2\,d\tau +z'_R(s) \leq C \frac{(s-t)}{R^{\frac{2}{\beta}}}+z'_R(t).
%%\end{align}
%%Next, integrating from $t$ to $t_2$ in the variable $s$ and using Fubini's theorem we obtain
%% \begin{align}
%% \sigma s_c\int_{t}^{t_2}(t_{2}-\tau)\|\nabla u(\tau&)\|_{L^2}^2\,d\tau +z_R(t_2)\\
%% &\leq C \frac{(t_2-t)^2}{2R^{\frac{2}{\beta}}}+z_R(t)+{(t_{2}-t)}z'_R(t)\\
%% &\leq C\left(\frac{(t_2-t)^2}{R^{\frac{2}{\beta }}}+R^2+R(t_2-t)\|\nabla u(t)\|_{L^2} \right),
%% \end{align}
%%where we have used the estimates \eqref{zR-bound} and \eqref{zR'-bound} in the last inequality. So taking $t_2\to T^{\ast}$, we deduce
%% \begin{align}\label{cotaint}
%% \int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\nabla u(\tau&)\|_{L^2}^2\,d\tau
%% \leq C\left(\frac{(T^{\ast}-t)^2}{R^{\frac{2}{\beta}}}+R^2+R(T^{\ast}-t)\|\nabla u(t)\|_{L^2}\right).
%% \end{align}
%% Thus, choosing $R\ll1$ so that the first two terms on the right hand side are equal, more precisely
%% \begin{align}
%% R(t)=(T^{\ast}-t)^{\frac{\beta}{1+\beta}},
%% \end{align}
%%the inequality \eqref{cotaint} can be rewritten as
%% \begin{align}
%% \int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\nabla u(\tau)\|_{L^2}^2\,d\tau
%% &\leq C\left((T^{\ast}-t)^{\frac{2\beta}{1+\beta}}+(T^{\ast}-t)^{\frac{\beta}{1+\beta}}(T^{\ast}-t)\|\nabla u(t)\|_{L^2}\right)\\
%% &\leq C (T^{\ast}-t)^{\frac{2\beta}{1+\beta}}+(T^{\ast}-t)^2\|\nabla u(t)\|_{L^2}^2.\label{cotaint2}
%% \end{align}
%%where we have used Young's inequality in the last step.
%%
%% Define
%% \begin{align}
%% h(t)=\int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\nabla u(\tau)\|_{L^2}^{2}\,d\tau.
%% \end{align}
%% Note that inequality \eqref{cotaint2} implies that $h(0)\leq C$ and moreover
%% \begin{align}
%% h(t)\leq C(T^{\ast}-t)^{\frac{2\beta}{1+\beta}}-(T^{\ast}-t)h'(t),
%% \end{align}
%% that is,
%% \begin{align}
%% \left(\frac{h(t)}{T^{\ast}-t}\right)'=\frac{1}{(T^{\ast}-t)^2}((T^{\ast}-t)h'(t)+h(t))\leq \frac{1}{(T^{\ast}-t)^{\frac{2}{1+\beta}}}.
%% \end{align}
%% Integrating the above inequality from $0$ to $t$,
%% \begin{align}
%% \frac{h(t)}{T^{\ast}-t}\leq \frac{h(0)}{T^{\ast}}+ \frac{C}{(T^{\ast}-t)^{\frac{1-\beta}{1+\beta}}}-\frac{C}{(T^{\ast})^{\frac{1-\beta}{1+\beta}}}\leq \frac{C}{(T^{\ast}-t)^{\frac{1-\beta}{1+\beta}}},
%%%\leq \frac{C}{T^{\ast}-t}\left((T^{\ast}-t)+(T^{\ast}-t)^{\frac{2\beta}{1+\beta}}\right)
%% \end{align}
%%for $t$ close enough to $T^{\ast}$, since $0<\beta<1 $. Therefore
%% \begin{align}
%% h(t)\leq C(T^{\ast}-t)^{\frac{2\beta}{1+\beta}},
%% \end{align}
%%which implies \eqref{blowupestdisp} and completes the proof.
%\end{proof}
%\subsection{$L^{\sigma_c}$-norm concentration}
%
%Let $\phi \in C^{\infty}_0(\Real^N)$ be a positive radial cut-off solution such that
%$$
%\phi(x)=
%\left\{
%\begin{array}{ll}
% 1&,\,\mbox{ if } |x|\leq 1,\\
% 0&,\,\mbox{ if } |x|\geq 2.
%\end{array}
%\right.
%$$
%For $R(t)>0$, define the inner and outer spatial localizations of $u(x,t)$ at radius $R(t)$ as
%$$u_1(x,t)=\phi\left(\frac{x}{R(t)}\right)u(x,t)\,\,\,\,\mbox{ and }\,\,\,\,u_2(x,t)=\left(1-\phi\left(\frac{x}{R(t)}\right)\right)u(x,t
%).$$
%Moreover, let $\chi \in C^{\infty}_0(\Real^N)$ be a radial function such that $\chi(x)=0$ for $|x|\geq 1$ and $\widehat{\chi}(0)=1$. For $\rho(t)$, define the inner and outer frequency localizations at radius $\rho(t)$ of $u_1(x,t)$ as $$\widehat{u}_{1L}(\xi,t)=\widehat{\chi}\left(\frac{\xi}{\rho(t)}\right)\widehat{u}_1(\xi,t)\,\,\,\,\mbox{ and }\,\,\,\,\widehat{u}_{1H}(\xi,t)=\left(1-\widehat{\chi}\left(\frac{\xi}{\rho(t)}\right)\right)\widehat{u}_1(\xi,t).$$
%By direct computation, it is clear that if $a>-\frac{(N-2)^2}{4}$, then
%\begin{equation}\label{u1H4}
%\|\mathcal L_a^{\frac12}
% {u}_{1L}(t)\|_{L^2}\leq c\|\mathcal L_a^{\frac12}
% {u}_{1}(t)\|_{L^2} \quad \mbox{ and }\quad \|\mathcal L_a^{\frac12}
% {u}_{1H}(t)\|_{L^2}\leq c\|\mathcal L_a^{\frac12}
% {u}_{1}(t)\|_{L^2}.
%\end{equation}
%\begin{proof}[Proof of Theorem \ref{Tlpc}]
%From the blow-wp alternative \eqref{GradBU1} and the energy conservation \eqref{Energy} we deduce
%$$
%\lim_{t\to T^{\ast}}\frac{\displaystyle\int |x|^{-b}|u(t)|^{2\sigma+2}\,dx}{\|\mathcal L_a^{\frac12}
% u(t)\|^2_{L^2}}=\sigma + 1.
%$$
%Hence, for $t$ close to $T^{\ast}$, we have
% \begin{align}\label{equ}
% \|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^{2}&\leq \int |x|^{-b}|u(t)|^{2\sigma+2}\,dx\nonumber\\
% &\leq \int |x|^{-b}|u_{1L}(t)|^{2\sigma+2}\,dx+\int |x|^{-b}|u_{1H}(t)|^{2\sigma+2}\,dx+\int |x|^{-b}|u_{2}(t)|^{2\sigma+2}\,dx,
% \end{align}
%where in the last line we have used that $u=u_{1L}+u_{1H}+u_2$ with pairwise disjoint supports.
%
%From the Gagliardo-Nirenberg inequality \eqref{GNcrit2} and the Sobolev embedding $\dot{H}^{s_c} \subset L^{\sigma_c}$, we get
% \begin{align}\label{u1H1}
% \int |x|^{-b}|u_{1H}(t)|^{2\sigma+2}\,dx&\leq c \|\mathcal L_a^{\frac12}
% u_{1H}(t)\|_{L^2}^2\|u_{1H}(t)\|_{L^{\sigma_c}}^{2\sigma}\leq c\|\mathcal L_a^{\frac12}
% u_{1H}(t)\|_{L^2}^2\|u_{1H}(t)\|_{\dot{H}^{s_c}}^{2\sigma}\nonumber\\
% &=c\|\mathcal L_a^{\frac12}
% u_{1H}(t)\|_{L^2}^2\left\||\xi|^{s_c}\left(1-\widehat{\chi}\left(\frac{\xi}{\rho(t)}\right)\right)\widehat{u}_1(t)\right\|_{L^2}^{2\sigma}
% \end{align}
%
%By the mean value theorem
% $$|1-\widehat{\chi}(\xi)|=|\widehat{\chi}(0)-\widehat{\chi}(\xi)|\leq c\min\{1,|\xi|\}.$$
%Therefore, for $|\xi|\leq \rho(t)$, we have
% $$|\xi|^{s_c}\left|1-\widehat{\chi}\left(\frac{\xi}{\rho(t)}\right)\right|\leq c|\xi|^{s_c}\frac{|\xi|}{\rho(t)}\leq c\frac{|\xi|}{\rho(t)^{1-s_c}}$$
% and if $|\xi|\geq \rho(t)$, then
% $$|\xi|^{s_c}\left|1-\widehat{\chi}\left(\frac{\xi}{\rho(t)}\right)\right|\leq c|\xi|^{s_c}= |\xi|^{s_c}\frac{|\xi|}{|\xi|}\leq \frac{|\xi|}{\rho(t)^{1-s_c}}.$$
%This implies that
%\begin{equation}\label{u1H2}
% |\xi|^{s_c}\left(1-\widehat{\chi}\left(\frac{\xi}{\rho(t)}\right)\right)\leq c \frac{|\xi|}{\rho(t)^{1-s_c}}.
%\end{equation}
%For a constant $a_2>0$, let
%\begin{equation}\label{rho(t)}
% \rho(t)=a_2\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^{\frac{1}{1-s_c}}.
%\end{equation}
%Collecting the estimates \eqref{u1H4}, \eqref{u1H1}, \eqref{u1H2}, \eqref{u1H3} and choosing $a_2>0$ sufficiently large we have
%\begin{align}
% \int |x|^{-b}|u_{1H}(t)|^{2\sigma+2}\,dx&\leq c\|\mathcal L_a^{\frac12}
% u_1(t)\|_{L^2}^2\left\|\frac{|\xi|}{\rho(t)^{1-s_c}}\widehat{u}_{1}(t)\right\|_{L^2}^{2\sigma}\leq c\frac{\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^{2\sigma+2}}{\rho(t)^{2\sigma(1-s_c)}}\\
% &\leq \frac{1}{4}\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^2.\label{eq_410}
%\end{align}
%Now, we will go to divide the proof in two cases.
%\begin{itemize}
% \item[\textbf{Case 1:}] \textbf{$\frac{2-b}{N}<\sigma<\min\{\frac{2-b}{N-2},\frac{2}{N}\}$ and $u_0$ is a function in $H^1(\Real^N)$}
%%\item[\textbf{Case 1:}] \textbf{$u_0$ is a radial initial data in $H^1$}
%\end{itemize}
%For a constant $a_1>0$, define
%\begin{equation}\label{R(t)}
% R(t)=a_1\frac{\max\left\{\|u_0\|_{L^2}, \|u_0\|_{L^2}^{\frac{2\sigma+2-\sigma N}{b}}\right\}}{\|\mathcal L_a^{\frac12}
% u(t)\|^{\frac{2-\sigma N}{b}}_{L^2}}.
%\end{equation}
%Moreover, since $a>-\frac{(N-2)^2}{4}$, $\sigma>\frac{2-b}{N}$ and $\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}>1$ for $t$ close enough to $T^{\ast}$, from the definition of $R(t)$ \eqref{R(t)} we get
% \begin{align}
% \left\|\nabla u_1(t)\right\|_{L^2} &\leq \frac{c}{R(t)}\|u_0\|_{L^2}+c\|\nabla u(t)\|_{L^2}\\
% &\leq c\,\left(\|\mathcal L_a^{\frac12}
% u(t)\|^{\frac{2-\sigma N}{b}}_{L^2}+\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}\right)\\
% &\leq c\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}.
% \end{align}
%Again, since $a>-\frac{(N-2)^2}{4}$, from inequality above, we obtain
%\begin{align}
% \left\|\mathcal L_a^{\frac12}
% u_1(t)\right\|_{L^2}
% &\leq c\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}. \label{u1H3}
%\end{align}
%Since $\sigma<\min\left\{\frac{2-b}{N-2},\frac{2}{N}\right\}$, applying the first inequality in \eqref{epsgradR}, we obtain
%\begin{align}\label{eq_482}
% \int |x|^{-b}|u_{2}(t)|^{2\sigma+2}\,dx&=\int |x|^{-b}\left|\left(1-\phi\left(\frac{x}{R(t)}\right)\right)u(t
% )\right|^{2\sigma+2}\,dx\nonumber\\&\leq c \int_{|x|\geq R(t)}|x|^{-b}|u(t)|^{2\sigma+2}\,dx \leq \frac{c}{R(t)^b}\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^{\sigma N}\|u_0\|_{L^2}^{2\sigma +2-\sigma N}\\
% &\leq \frac{c}{a^b_{1}} \|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^2
% \leq \frac{1}{4}\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^2,
%\end{align}
%where we have chosen $a_1>0$ sufficiently large such that the last inequality holds.
%\begin{itemize}
% \item[\textbf{Case 2:}] $\frac{2-b}{N}<\sigma<\frac{2-b}{N-2}$ and \textbf{$u_0$ is a radial function in $H^1(\Real^N)$}
%\end{itemize}
%Again, since $a>-\frac{(N-2)^2}{4}$, $\sigma>\frac{2-b}{N}$ and $\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}>1$ for $t$ close enough to $T^{\ast}$, if we define $R(t)$ as
% $$R(t)=a_{1}\frac{\max\left\{\|u_0\|_{L^2},\|u_0\|_{L^2}^{\frac{\sigma+2}{\sigma(N-1)+b}}\right\}}{\|\mathcal L_a^{\frac12}
%\|_{L^2}^\frac{\sigma}{\sigma(N-1)+b}}$$ $a_{1}$ is chosen small enough, then
%\begin{align}
% \left\|\nabla u_1(t)\right\|_{L^2} &\leq \frac{c}{R(t)}\|u_0\|_{L^2}+c\|\nabla u(t)\|_{L^2}\\
% &\leq c\,\left(\|\mathcal L_a^{\frac12}
% u(t)\|^{\frac{\sigma}{\sigma(N-1)+b}}_{L^2}+\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}\right)\\
% &\leq c\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}.
%\end{align}
%Thus, since $a>-\frac{(N-2)^2}{4}$, we get
%\begin{align}
% \left\|\mathcal L_a^{\frac12}
% u_1(t)\right\|_{L^2}
% &\leq c\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}. \label{u1H}
%\end{align}
%Using the second inequality in \eqref{case2} we have
%\begin{align}\label{eq_48}
% \int |x|^{-b}|u_{2}(t)|^{2\sigma+2}\,dx&\leq c \int_{|x|\geq R(t)}|x|^{-b}|u(t)|^{2\sigma+2}\,dx \leq \frac{c}{R(t)^{\sigma(N-1)+b}}\|\mathcal L_a^{\frac12}
% u\|_{L^2}^{\sigma}\|u_0\|_{L^2}^{\sigma+2}\\
% &\leq \frac{c}{a^{\sigma(N-1)+b}_{1}} \|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^2
% \leq \frac{1}{4}\|\mathcal L_a^{\frac12}
% u(t)\|_{L^2}^2.
%\end{align}
%
%
% Therefore, in both cases, in view of the inequality \eqref{equ}, \eqref{eq_410}, \eqref{eq_482} and \eqref{eq_48}, we deduce that
% \begin{equation}\label{inq}
% \left\|\mathcal L_a^{\frac12}
% u(t)\right\|_{L^{2}}^{2}\leq c\int |x|^{-b}|u_{1L}(t)|^{2\sigma+2}\,dx.
% \end{equation}
%Next, using again the Gagliardo-Nirenberg inequality \eqref{GNcrit} and the estimates \eqref{u1H4}, \eqref{u1H3} and \eqref{inq}, we obtain
% \begin{align*}
% \int |x|^{-b}|u_{1L}(t)|^{2\sigma+2}\,dx&\leq c\|\nabla u_{1L}(t)\|_{L^2}^{2}\|u_{1L}(t)\|_{L^{\sigma_c}}^{2\sigma}\\
% &\leq c\|u_{1L}(t)\|_{L^{\sigma_c}}^{2\sigma}\int |x|^{-b}|u_{1L}(t)|^{2\sigma+2}\,dx.
% \end{align*}
%Moreover, Young's convolution inequality yields
% $$\|u_{1L}(t)\|_{L^{\sigma_c}}=\|\rho^N\chi(\rho\,\cdot)\ast u_1(t)\|_{L^{\sigma_c}}\leq c\|u_1(t)\|_{L^{\sigma_c}}.$$
%From the last two inequalities it follows that $\|u_1(t)\|_{L^{\sigma_c}}$ is bounded from below by a universal constant $c>0$ independent of the initial data $u_0$ and time $t$. Finally, by definition of $u_1(t)$, we get
%$$
%c^{\sigma_c}\leq \|u_1(t)\|_{L^{\sigma_c}}^{\sigma_c}=\left\|\phi\left(\frac{x}{R(t)}\right)u(t)\right\|_{L^{\sigma_c}}^{\sigma_c}\leq \int_{|x|\leq2R(t)}|u(t)|^{\sigma_c}\,dx=\|u(t)\|^{\sigma_c}_{L^{\sigma_c}_{\{|x|\leq 2R(t)\}} },
%$$
%which, by the definition of $R(t)$ \eqref{R(t)}, implies \eqref{fint} and complete the proof of Theorem \ref{Tlpc}.
%
%
%%\textbf{Case 1.} If $\|u_1(t)\|_{L^{\sigma_c}}$ is not bounded (so $\| u_1(t)\|_{L^{\gamma}} $ is also unbounded for any $\gamma > \sigma_c$), then there exist a sequence $t_n\to T^{\ast}$ such that $\|u_1(t_n)\|_{L^{\sigma_c}}\to +\infty.$
%%
%% Since
%% $$\|u_1(t)\|_{L^{\sigma_c}}^{\sigma_c}=\left\|\phi\left(\frac{x}{R(t)}\right)u(t)\right\|_{L^{\sigma_c}}^{\sigma_c}\leq \int_{|x|\leq2R(t)}|u(t)|^{\sigma_c}\,dx=\|u(t)\|^{\sigma_c}_{L^{\sigma_c}_{\{|x|\leq 2R(t)\}} },$$
%% we have \eqref{inft}.\\
%%
%%\textbf{Case 2.} If, on the other hand, $\| u_1(t)\|_{L^{\gamma}} \leq c^*$, for some $\gamma > \sigma_c$, then we have \eqref{itemii}. Since $|x_0(t)|\,\rho(t)\leq c(c^*)^{\frac{\gamma}{N-1}+\frac{\sigma_c\gamma(\gamma-2)}{(N-1)(N(2\sigma+2)-\sigma_c(N-b))}}$, then$$
%% \frac{c}{(c^*)^{\frac{\gamma N}{N(2\sigma+2)-\sigma_c(N-b)}}}
%% \leq \left\|u_{1}(t)\right\|_{L^{\sigma_c}\left(|x-x_0(t)|\leq \rho(t)^{-1}\right)}
%% \leq \left\|u_{1}(t)\right\|_{L^{\sigma_c}\left(|x|\leq c(c^*)^{\frac{\gamma}{N-1}+\frac{\sigma_c\gamma(\gamma-2)}{(N-1)(N(2\sigma+2)-\sigma_c(N-b))}} \rho(t)^{-1}\right)},
%% $$
%% which gives \eqref{fint}.
%\end{proof}
%
%As a consequence of the above argument, if we further assume that $\|u_1(t)\|_{L^{\sigma_c}}$ is not bounded above, then there exist a sequence $t_n\to T^{\ast}$ such that
%\begin{equation}\label{inft}
%\int_{|x|\leq c_{u_0}\|\nabla u(t_n)\|^{-\frac{2-\sigma N}{b}}_{L^2}}|u(x,t_n)|^{\sigma_c}\,dx\to +\infty.
%\end{equation}
%On the other hand, if $\|u_1(t)\|_{L^{\sigma_c}}$ is bounded, then it may be possible to prove that the concentration window shrinks at a different rate following the strategy of \citet[Theorem 1.2]{HR07}. We have decided not to explore this for the moment, since our main purpose here is to show how to remove the radial assumption and still obtain a concentration result around the origin for the $(\mbox{INLS})_a$ equation.
\subsection{Gagliardo-Nirenberg inequality}
We first recall the following scaling invariant Morrey-Campanato type semi-norm, used in \citet{MR_Bsc}
\begin{equation}\label{defrho}
\rho(u,R)=\sup_{R'\geq R}\frac{1}{(R')^{2s_c}}\int_{R'\leq |x|\leq 2R'}|u|^2\,dx.
\end{equation}
It is easy to see that $\rho(u,R)$ is non-increasing in $R>0$. Moreover, by H\"older's inequality, there exists a universal constant $c>0$ such that for all $u\in L^{\sigma_c}$ and $R>0$
\begin{equation}\label{radial1}
\frac{1}{R^{2s_c}}\int_{|x|\leq R}|u|^2\,dx\leq c\|u\|_{L^{\sigma_c}}^2
\end{equation}
and
\begin{equation}\label{radial2}
\lim_{R\to+\infty}\frac{1}{R^{2s_c}}\int_{|x|\leq R}|u|^2\,dx=0
\end{equation}
(c.f. Merle and Raph\"ael in \citep[Lemma 1]{MR_Bsc} and also \cite[Lemma 2.1]{CF20}). In \cite{CF20} and \cite{CF23}, the last two authors proved some interpolation inequalities adapted to the inhomogeneous case in $\dot{H}^{s_c}\cap\dot{H}^{1}$. The relations \eqref{equivsc1} imply that these results also hold in the space $\dot{H}_a^{s_c}\cap\dot{H}_a^{1}$, for $a>0$ and $x\in \mathbb{R}^3$. Below we recall such estimates.
\begin{lemma}[Gagliardo-Nirenberg type inequality \cite{CF20}, \cite{CF23}]\label{lemaradialGN1} Assume $0<b<3/2$. Then, for all $\eta>0$, there exists a constant $C_\eta>0$ such that
\begin{itemize}
\item[(i)][Radial] for all $u\in \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)$ with radial symmetry the following inequality holds
\begin{equation}\label{GNradial}
\int_{|x|\geq R}|x|^{-b}|u|^{2\sigma+2}\,dx\leq \eta\|\nabla u\|_{L^{2}}^{2}+\frac{C_{\eta}}{R^{2(1-s_ c)}}\left\{[\rho(u,R)]^{\frac{2+\sigma}{2-\sigma}}+[\rho(u,R)]^{\sigma+1}\right\},
\end{equation}
if $\frac{2-b}{3}<\sigma<2-b$.
\item[(ii)][Non-radial] for all $u\in \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)$ the following inequality holds
\begin{equation}\label{GNradial2}
\int_{|x|\geq R}|x|^{-b}|u|^{2\sigma+2}\,dx\leq \eta\|\nabla u\|_{L^{2}}^{2}+\frac{C_{\eta}}{R^{2(1-s_ c)}}[\rho(u,R)]^{\frac{2-\sigma}{2-3\sigma}},
\end{equation}
if $\frac{2-b}{3}<\sigma<\min\{2-b, \frac{2}{3}\}$.
\end{itemize}
\end{lemma}
%\begin{lemma}[\cite{CF20} Radial Gagliardo-Nirenberg inequality]\label{lemaradialGN1} Suppose that $N\geq 3$, $0<b<\min\left\{\frac{N}{2},2\right\}$, $\frac{2-b}{N}<\sigma<\frac{2-b}{N-2}$, $s_c=\frac{N}{2}-\frac{2-b}{2\sigma}$ and $\sigma_c=\frac{2N\sigma}{2-b}$. The following statements hold.
% Then, for all $\eta>0$, there exists a constant $C_\eta>0$ such that for all $u\in \dot{H}_a^{s_c}\cap\dot{H}_a^{1}$ the following inequality holds
% \begin{equation}\label{GNradial}
% \int_{|x|\geq R}|x|^{-b}|u|^{2\sigma+2}\,dx\leq \eta\|\nabla u\|_{L^{2}}^{2}+\frac{C_{\eta}}{R^{2(1-s_ c)}}\left\{[\rho(u,R)]^{\frac{2+\sigma}{2-\sigma}}+[\rho(u,R)]^{\sigma+1}\right\}.
% \end{equation}
%\end{lemma}
%\begin{lemma}[\cite{CF23} Non-radial Gagliardo-Nirenberg inequality] \label{lemaradialGN2} Suppose that $N\geq 3$, $0<b<2$, $\frac{2-b}{N}<\sigma<\min\{\frac{2}{N},\frac{2-b}{N-2}\}$.
% Then, for all $\eta>0$, there exists a constant $C_\eta>0$ such that for all $u\in \dot{H}_a^{s_c}\cap\dot{H}_a^{1}$ the following inequality holds
% \begin{equation}\label{GNradial2}
% \int_{|x|\geq R}|x|^{-b}|u|^{2\sigma+2}\,dx\leq \eta\|\nabla u\|_{L^{2}}^{2}+\frac{C_{\eta}}{R^{2(1-s_ c)}}[\rho(u,R)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}.
% \end{equation}
%\end{lemma}
Next, we use the previous lemma to refine the estimate \eqref{estintradial}. Indeed, since $\rho$ non-increasing, for all $u\in \dot{H}_a^{s_c} \subset L^{\sigma_c}$ we have
\begin{align}
\int_{|x|\geq R} |x|^{-2}|u|^2\,dx&=\sum_{j=0}^\infty \int _{2^jR\leq |x|\leq 2^{j+1}R}|x|^{-2}|u|^2\,dx\leq \sum_{j=0}^\infty\dfrac{1}{(2^jR)^2}\int_{2^jR\leq |x|\leq 2^{j+1}R}|u|^{2}\,dx\\
&\leq \frac{1}{R^{2(1-s_c)}}\rho(u,R)\sum_{j=0}^\infty \frac{1}{(2^{2(1-s_c)})^j}\leq \frac{C}{R^{2(1-s_c)}}\rho(u,R). \label{GNradial3}
\end{align}
Define the number $\gamma=\gamma(\sigma, N)$ by
\begin{align}
\displaystyle\gamma=\left\{
\begin{array}{cc}
\dfrac{2+\sigma}{2-\sigma},& \mbox{ if $\frac{2-b}{3}<\sigma<2-b$ and $u\in \dot{H}_a^{s_c}\cap\dot{H}_a^{1}$ is radial},\\
&\\
\dfrac{2-\sigma}{2-3\sigma },& \mbox{ if $\frac{2-b}{3}<\sigma<\min\{2-b, \frac{2}{3}\}$ and $u\in \dot{H}_a^{s_c}\cap\dot{H}_a^{1}$}.
\end{array}
\right.
\end{align}
Note that in both cases, we have
\begin{align}
\gamma>\sigma+1>1.
\end{align}
Thus, inserting \eqref{GNradial}, \eqref{GNradial2}, \eqref{GNradial3} in \eqref{estintradial} we deduce
\begin{align}
8\sigma s_c\|\mathcal L_a^{\frac12}
u(t)\|^2_{L^2}+z''_R(t)&-16(\sigma s_c+1)E[u_0]\\\leq &\eta\|\mathcal L_a^{\frac12}
u(t)\|_{L^2}^2+\frac{C_{\eta}}{R^{2(1-s_c)}}\left\{[\rho(u(t), R)]^{\gamma}+\rho(u(t),R)\right\}.\label{estint}
\end{align}
Now, assume that $v\in C([0,\tau_*]: \dot H_a^{s_c}\cap \dot H_a^1)$ and fix $\tau_0\in [0,\tau_*]$. Choosing $\eta>0$ we can absorb the term $\eta\|\mathcal L_a^{\frac12}
v(t)\|_{L^2}^2$ in the left hand side of the above inequality, moreover integrating in $t$ from $0$ to $\tau$ and then in $\tau$ from $0$ to $\tau_0$, we obtain\footnote{See the proof of estimate (4.15) in \cite{CF20} for more details.}
\begin{align}
4\sigma s_c\int_{0}^{\tau_0}(\tau_0-\tau)&\|\mathcal L_a^{\frac12}
v(\tau)\|_{L^2}^2\,d\tau+z_R(\tau_0)\\
\leq& z_R(0)+\tau_0\left(z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right)\\
\quad &+\frac{1}{R^{2(1-s_c)}}\int_{0}^{\tau_0}\int_{0}^{\tau}\left\{[\rho(v(t), R)]^{\gamma}+\rho(v(t),R)\right\}\,dtd\tau\\\leq& cR^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2+\tau_0\left(z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right)\\
\quad &+\frac{1}{R^{2(1-s_c)}}\int_{0}^{\tau_0}\int_{0}^{\tau}\left\{[\rho(v(t), R)]^{\gamma}+\rho(v(t),R)\right\}\,dtd\tau.\label{ineq1}
\end{align}
where we have used, from inequality \eqref{radial1}, that
\begin{align}\label{phiR1}
z_R(0)\leq cR^{2}\int_{|x|\leq 4R}|v_0|^2\,dx\leq cR^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2.
\end{align}
Next, observe that for all $R>0$ the definition of $\phi$ (see \eqref{phi}) yields
\begin{align}\label{zRt0}
z_R(\tau_0)%&=\frac{1}{R^{2}}\int_{|x|\leq 4R} \phi\left(\frac{x}{R}\right)|v(\tau_0)|^2\,dx\\
&\geq \int_{R\leq |x|\leq 2R}|x|^2|v(\tau_0)|^{2}\,dx\geq {R^2}\int_{R\leq|x|\leq 2R}|v(\tau_0)|^2\,dx.
\end{align}
Thereby, injecting \eqref{zRt0} into \eqref{ineq1} we have
\begin{align}
4\sigma s_c\int_{0}^{\tau_0}(\tau_0-\tau)&\|\mathcal L_a^{\frac12}
v(\tau)\|_{L^2}^2\,d\tau+{R^2}\int_{R\leq|x|\leq 2R}|v(\tau_0)|^2\,dx
\\\leq& cR^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2+\tau_0\left(z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right)\\
\quad &+\frac{1}{R^{2(1-s_c)}}\int_{0}^{\tau_0}\int_{0}^{\tau}\left\{[\rho(v(t),R)]^{\gamma}+\rho(v(t),R)\right\}\,dtd\tau.\label{ineq2}
\end{align}
\subsection{Preliminary lemmas}
We now recall two results that we will be used in the proof of our main theorems. The first one provides a uniform control of the $\rho$ norm and a global dispersive estimate.
\begin{prop}\label{prop1}
Let $0<b<\frac{3}{2}$. Assume either $\frac{2-b}{3}<\sigma<2-b$ and $v_0\in \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)$ is radial or $\frac{2-b}{3}<\sigma<\min\{2-b, \frac{2}{3}\}$ and $v_0\in \dot{H}_a^{s_c}\cap\dot{H}_a^{1}$. Let $v\in C\left([0,\tau_\ast]: \dot{H_a^{s_c}}(\Real^3)\cap \dot H_a^1(\Real^3)\right)$ be a solution to \eqref{PVI} such that the initial data $v(0)=v_0$ satisfies
\begin{equation}\label{hpprop1i}
\tau_\ast^{1-s_c}\max\{E[v_0],0\}<1
\end{equation}
and
\begin{align}\label{hpprop1ii}
M_0:=\frac{4\|v_0\|_{L^{\sigma_c}}}{K_a}\geq 2,
\end{align}
where $K_a$ is a optimal constant in \eqref{GNacrit2}. Then, there exist universal constants $C_1,\alpha_1,\alpha_2>0$ depending only on $N,\sigma$ and $b$ such that, for all $\tau_0\in [0,\tau_\ast]$, the following uniform estimates hold
\begin{align}\label{prop1ii}
\rho(v(\tau_0),M_0^{\alpha_1}\sqrt{\tau_0})\leq C_1M_0^2
\end{align}
and
\begin{align}\label{prop1i}
\int_0^{\tau_0}(\tau_{0}-\tau)\|\mathcal L_a^{\frac12}
v(\tau)\|_{L^2}^2\,d\tau \leq M_0^{\alpha_2}\tau_{0}^{1+s_c}.
\end{align}
\end{prop}
%\begin{proof} For all $A>0$ and $\tau_0\in [0,\tau_{\ast}]$, let $R=A\sqrt{\tau_0}$ and $M_{\infty}$ defined by
% \begin{align}\label{Minfty}
% M^2_{\infty}(A,\tau_0)=\max_{\tau\in[0,\tau_0]}\rho(v(\tau),A\sqrt{\tau}).
% \end{align}
%Then using the inequality \eqref{ineq2}, we obtain the existence of a universal constant $c>0$ such that
% \begin{align}
% 8\sigma s_c\int_{0}^{\tau_0}(\tau_0-\tau)\|\mathcal L_a^{\frac12}
% v(\tau)\|_{L^2}^2\,d\tau&\leq c\tau_0^{1+s_c}A^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2
% +2\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]\\ &\quad\quad+c\frac{\tau_0^{1+s_c}}{A^{2(1-s_c)}}\left[[M^2_{\infty}(A,\tau_0)]^{\gamma}+M^2_{\infty}(A,\tau_0)\right]\label{lemi}
% \end{align}
% and
% \begin{align}
% \frac{1}{R^{2s_c}}\int_{R\leq |x|\leq 2R}|v(\tau_0)|^2\,dx&\leq c\|v_0\|_{L^{\sigma_c}}^2+\frac{c}{A^4}\left[M^2_{\infty}(A,\tau_0)]^{\gamma}+M^2_{\infty}(A,\tau_0)\right]\\
% &\quad\quad+\frac{2}{\tau_0^{s_c}A^{2(1+s_c)}}\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]. \,\,\,\,\,\, \label{lemii}
% \end{align}
%
%
%Now, let $\varepsilon>0$ a fixed small enough real number to be chosen later and define%depending only of $\sigma_c$ and $v_0$
%\begin{align}\label{Gep}
%G_{\varepsilon}=M_0^{\frac{1}{\varepsilon}}\,\,\, \mbox{and}\,\,\,A_{\varepsilon}=\left(\frac{\varepsilon G_{\varepsilon}}{M_0^2}\right)^{\frac{1}{2(1+s_c)}},
%\end{align}
%where $M_0$ is given in \eqref{hpprop1ii}. Consider the following estimates
%\begin{align}\label{dispersioneps}
%\int_{0}^{\tau_0}(\tau_0-\tau)\|\mathcal L_a^{\frac12}
% v(\tau)\|_{L^2}^{2}\,d\tau\leq G_{\varepsilon}\tau_0^{1+s_c}
%\end{align}
%and
%\begin{align}\label{Minftyep}
%M_\infty^2(A_\varepsilon,\tau_0)\leq \frac{2M_0^2}{\varepsilon}.
%\end{align}
%We define
%\begin{align}\label{tau11}
%S_{\varepsilon}= \left\{\tau\in [0,\tau_*];\,\,\eqref{dispersioneps}\mbox{ and }\eqref{Minftyep} \mbox{ hold for all }\tau_0\in [0,\tau]\right\}
%\end{align}
%and
%\begin{align}\label{tau1}
%\tau_1=\max_{\tau\in [0,\tau_*]}S_{\varepsilon}.
%\end{align}
%Note that $S_{\varepsilon}\neq \emptyset$, for $\varepsilon>0$ sufficiently small, since $v\in C\left([0,\tau_\ast]: \dot{H}_a^{s_c}\cap\dot{H}^1_a\right)$ and also by the definition of $M_0$ in \eqref{hpprop1ii} and the Sobolev embedding $\dot{H}_a^{s_c} \subset L^{\sigma_c}$.
%%In fact, by the regularity of $v(\tau)$, there exists $0<\delta'<\min\{1,\tau_*\}$ such that
%%\begin{align}
%%\|\nabla v(\tau)\|_{L^2}^2<1+\|\nabla v_0\|_{L^2}^2,\,\,\,\mbox{for all}\,\,\, \tau\in [0,\delta'].
%%\end{align}
%%%definition of energy, the Gagliardo-Nirenberg inequality, the Young inequality and the definition \eqref{hpprop1ii} of $M_0$, we have
%%%\begin{align}
%%%E[v_0]\geq \frac{1}{2}\|\nabla v_0\|_{L^2}^2-\frac{1}{2}\|\nabla v_0\|_{L^2}^2\left(\frac{\|v_0\|_{L^{\sigma_c}}}{\|V\|_{L^{\sigma_c}}}\right)^{2\sigma}\geq \frac{1}{2}\|\nabla v_0\|_{L^2}^2-\frac{1}{2}\|\nabla v_0\|_{L^2}^4-\frac{1}{16}\left(\frac{M_0}{4}\right)^{4\sigma},
%%%\end{align}
%%%more precisely,
%%%\begin{align}
%%%-\|\nabla v_0\|_{L^2}^4+\|\nabla v_0\|_{L^2}^2\leq 2E[v_0]+\frac{1}{8}\left(\frac{M_0}{4}\right)^{4\sigma}.
%%%\end{align}
%%%Thus, there exists $C(v_0)>0$ such that $\|\nabla v_0\|_{L^2}<C(v_0)$.
%%Next, multiplying the last inequality by $(\tau_0-\tau)$ and integrating from $0$ to $\tau_0$ with $\tau_0\in [0,\delta']$, we have
%%\begin{align}
%%\int_{0}^{\tau_0}(\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau\leq \left(\frac{1+\|\nabla v_0\|_{L^2}^2}{2}\right)\tau_0^{1+s_c}\leq M_0^{\frac{1}{\varepsilon}}\tau_0^{1+s_c},
%%\end{align}
%%for sufficiently small $\varepsilon>0$, since $M_0\geq 2$. Furthermore, we recall from \eqref{radial1} and \eqref{hpprop1ii} that there exist universal constants $c,C>0$ such that for every $R>0$
%%%\begin{align}\label{rhoM_0}
%%%\frac{1}{R^{2s_c}}\int_{|x|\leq 4R}|v_0|^{2}\,dx=\frac{4^{2s_c}}{(4R)^{2s_c}}\int_{|x|\leq4R}|v_0|^2\,dx\leq c \|v_0\|_{L^{\sigma_c}}^2\leq c M_0^2.
%%%\end{align}
%%\begin{align}
%%\rho(v_0,R)\leq c\|v_0\|_{L^{\sigma_c}}^2\leq C M_0^2.
%%\end{align}
%%Again using the regularity of $v\in C([0,\tau_*]:\dot H^{s_c}\cap\dot H^1 )$ and definition \eqref{Minfty}, there exists $0<\delta''\leq\tau_*$ such that for all $\tau\in [0,\delta'']$
%%\begin{align}
%%%\|\nabla v(\tau)\|_{L^2}<C(v_0)\mbox{ and }
%%M_{\infty}^2(A_{\varepsilon},\tau)<2CM_0^2\leq \frac{2M_0^2}{\varepsilon},
%%\end{align}
%%for sufficiently small $\varepsilon>0$.
%%Hence, there exists $\varepsilon_0>0 $, depending only on $v_0$, such that $\min\{\delta',\delta''\}\in S_{\varepsilon}$, for all $\varepsilon\leq \varepsilon_0$.
%
%The goal is to show that $\tau_1=\tau_\ast$ and therefore inequalities \eqref{dispersioneps} and \eqref{Minftyep} hold at the maximal time $\tau_{\ast}$, which clearly imply \eqref{prop1ii} and \eqref{prop1i}, in view of definition \eqref{Gep}. Indeed, from \eqref{Gep} and \eqref{Minftyep} it is easy to see that
% \begin{align}\label{110}
% \frac{[M^2_{\infty}(A_{\varepsilon},\tau_0)]^{\gamma}+M^2_{\infty}(A_{\varepsilon},\tau_0)}{A_{\varepsilon}^{2(1-s_c)}}
%\leq \frac{\left(\frac{2M_0^2}{\varepsilon}\right)^{\gamma}+\frac{2M_0^2}{\varepsilon}}{\left(\frac{\varepsilon G_{\varepsilon}}{M_0^2}\right)^{\frac{1-s_c}{1+s_c}}}\leq c\frac{1}{M_0^{\frac{1}{\varepsilon}\cdot \frac{1-s_c}{1+s_c}}} \left(\frac{M_0^2}{\varepsilon}\right)^{\gamma+\frac{1-s_c}{1+s_c}}\leq \frac{1}{10}, %\\
%% &\leq \frac{\left(\frac{M_0^2}{\varepsilon}\right)^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+\frac{M_0^2}{\varepsilon}}{\left(\frac{\varepsilon}{M_0^2}\right)^{\frac{1-s_c}{1+s_c}}G_{\varepsilon}^{\frac{1-s_c}{1+s_c}}}
%% \leq \left(\frac{M_0}{\varepsilon}\right)^C\frac{1}{G_{\varepsilon}^{\frac{1-s_c}{1+s_c}}}\leq \frac{1}{2^{\frac{1}{C_{\varepsilon}}}\varepsilon^C}\leq \frac{1}{10},
% \end{align}
%for $\varepsilon>0$ small enough, where we have used the assumption \eqref{hpprop1ii}. Combining the last inequality with \eqref{lemi} and using again definition \eqref{Gep} and assumption \eqref{hpprop1ii}, we have, for $R=A_{\varepsilon}\sqrt{\tau_0}$, that
% \begin{align}
% \int_{0}^{\tau_0}(\tau_0-\tau)\|\mathcal L_a^{\frac12}
% v(\tau)\|_{L^2}^2\,d\tau&\leq c\tau_0^{1+s_c}\left[M_0^2A_{\varepsilon}^{2(1+s_c)}+\frac{[M^2_{\infty}(A_{\varepsilon},\tau_0)]^{\gamma}+M^2_{\infty}(A_{\varepsilon},\tau_0)}{A_{\varepsilon}^{2(1-s_c)}}\right]\nonumber\\
% &\quad+2c\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]\\
% & \leq c\tau_0^{1+s_c}\left[\varepsilon G_{\varepsilon}+\frac{1}{10}\right]+2c\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]\nonumber\\
% & \leq G_{\varepsilon}\tau_0^{1+s_c}\left\{c\varepsilon+\frac{c}{10G_{\varepsilon}}+\frac{2c}{G_{\varepsilon}\tau_0^{s_c}}\left[z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right]\right\}\nonumber\\
% &\leq G_{\varepsilon}\tau_0^{1+s_c}\left\{\frac{1}{10}+\frac{2c}{G_{\varepsilon}\tau_0^{s_c}}\left[z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right]\right\},\nonumber
% \end{align}
%for $\varepsilon>0$ small enough.
%
%Next, from Lemma 4.4 in \cite{CF20}\footnote{In the proof of \cite[Lemma 4.4]{CF20} does not require a radial assumption.}, there exists $\varepsilon_0>0 $ small enough and $c>0$ a universal constant such that, for all $\tau_0\in[0,\tau_1]$, $0<\varepsilon\leq \varepsilon_0$, $A\geq A_\varepsilon$ and $R=A\sqrt{\tau_0}$, the following inequality holds
% \begin{align}\label{ImE}
% z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\leq c \frac{M_0^2A^{2(1+s_c)}}{\varepsilon^{\frac{1}{1+s_c}}}\tau_0^{s_c}.
% \end{align}
%Therefore, from estimate \eqref{ImE} with $A=A_\varepsilon$, for all $\tau_0\in [0,\tau_1]$ and since $R=A_{\varepsilon}\sqrt{\tau_0}$, we get
% \begin{align}\label{dispersioneps2}
% \int_{0}^{\tau_0}(\tau_0-\tau)\|\mathcal L_a^{\frac12}
% v(\tau)\|^2_{L^2}\,d\tau&\leq G_{\varepsilon}\tau_0^{1+s_c}\left[\frac{1}{10}+c\frac{M_0^2A_{\varepsilon}^{2(1+s_c)}}{G_{\varepsilon}\varepsilon^{\frac{1}{1+s_c}}}\right]\\
% &=G_{\varepsilon}\tau_0^{1+s_c}\left[\frac{1}{10}+c\varepsilon^{\frac{s_c}{1+s_c}}\right]\leq \frac{G_{\varepsilon}}{2}\tau_0^{1+s_c}.
% \end{align}
%Moreover, let $A\geq A_{\varepsilon}$ and $R=A\sqrt{\tau_0}$. First, since $\rho$ is non-increasing in $R$, then for all $\tau_0\in [0,\tau_1]$ we obtain
% \begin{align}
% M^2_{\infty}(A,\tau_0)\leq M^2_{\infty}(A_{\varepsilon},\tau_0)\leq M^2_{\infty}(A_{\varepsilon},\tau_1)\leq \frac{2M_0^2}{\varepsilon},
% \end{align}
%where in the last inequality we have used \eqref{Minftyep}.\\
%Combining \eqref{lemii} with \eqref{hpprop1ii}, the last inequality and \eqref{ImE} we deduce
% \begin{align}\label{Minftyep2}
% \frac{1}{R^{2s_c}}\int_{R\leq |x|\leq 2R}|v(\tau_0)|^2\,dx&\leq cM^2_0+\frac{c}{A^4}\left[[M^2_{\infty}(A,\tau_0)]^{\gamma}+M^2_{\infty}(A,\tau_0)\right]\\
% &\quad +\frac{2}{\tau_0^{s_c}A^{2(1+s_c)}}\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]\nonumber\\
% &\leq cM_0^2+c\frac{\left(\frac{2M_0^2}{\varepsilon}\right)^{\gamma}+\frac{2M_0^2}{\varepsilon}}{\left(\frac{\varepsilon G_{\varepsilon}}{M_0^2}\right)^{\frac{2}{1+s_c}}}+\frac{cM_0^2}{\varepsilon^{\frac{1}{1+s_c}}}\\
% &=\frac{M_0^2}{\varepsilon}\left[c\varepsilon+\frac{c\varepsilon}{M_0^{2+\frac{1}{\varepsilon}\cdot \frac{2}{1+s_c}}} \left(\frac{M_0^2}{\varepsilon}\right)^{\gamma+\frac{2}{1+s_c}}+c\varepsilon^{\frac{s_c}{1+s_c}}\right]< \frac{M_0^2}{\varepsilon},
% \end{align}
%for $\varepsilon>0$ small enough.
%
%Finally, in view of the regularity of $v$ and the definition of $S_{\varepsilon}$ in \eqref{tau11}, the inequalities \eqref{dispersioneps2} and \eqref{Minftyep2} imply that $\tau_1=\tau_\ast$ and the proof of Proposition \ref{prop1} is completed.
%\end{proof}
The proof of Proposition \ref{prop1} follows exactly the same outline as the proof of Proposition 3.2 in \cite{CF23}, just using the new estimate \eqref{ineq2} and the equivalence \eqref{equiv}. The next result shows that restricting the initial data in a neighborhood of the origin we can deduce a lower bound for its $L^2$ norm.
\begin{prop}\label{prop2} Under the same assumptions of Proposition \ref{prop1}, let
\begin{align}\label{hpprop2i}
\tau_0\in\left[0,\frac{\tau_*}{2}\right].
\end{align}
Define $\lambda_v(\tau)=\|\mathcal L_a^{\frac12}
v(\tau)\|^{-\frac{1}{1-s_c}}_{L^2}$ and assume that
\begin{align}\label{Ev}
E[v_0]\leq \frac{\|\mathcal L_a^{\frac12}
v(\tau_0)\|^2_{L^2}}{4}=\frac{1}{4\lambda_v^{2(1-s_c)}(\tau_0)}.
\end{align}
Then, there exist universal constants $C_2, \alpha_3>0$ depending only on $N,\sigma$ and $b$ such that if
\begin{align}\label{F}
F_*=\frac{\sqrt{\tau_0}}{\lambda_v(\tau_0)}\,\,\,\mbox{ and }\,\,\,D_*=M_0^{\alpha_3}\max[1,F_*^{\frac{1+s_c}{1-s_c}}],
\end{align}
then
\begin{align}\label{tprop2}
\frac{1}{\lambda_v^{2s_c}(\tau_0)}\int_{|x|\leq D_*\lambda_v(\tau_0)}|v_0|^2\,dx\geq C_2.
\end{align}
\end{prop}
%Let $T>0$ and $\phi$ be a real-function such that $\phi\geq 0$, $\phi\in C^{\infty}_0 $. For $R>0$ and $t\in [0,T]$, we define the function
%We postpone the proofs of the above propositions to the subsection \ref{5.3.3}. Next, we state and prove some preliminary results.
%\subsection{Auxiliary lemmas}
%In the next result we use Lemma \ref{lemaradialGN} and Lemma \ref{lemintradial} to deduce a first control in the quantities that appears in the left hand side of \eqref{prop1i} and in the definition of the semi norm $\rho$ \eqref{defrho} for a suitable choice of $R>0$.
%Verificar quem é M_0 do próximo Lemma!
%\begin{lemma}\label{lemma34} Let $v\in C([0,\tau_*]: \dot H_a^{s_c}\cap \dot H_a^1 )$ be a solution to \eqref{PVI} with initial data $v_0\in \dot H_a^{s_c}\cap \dot H_a^1$. For all $A>0$ and $\tau_0\in [0,\tau_{\ast}]$, let $R=A\sqrt{\tau_0}$ and $M_{\infty}$ defined by
% \begin{align}\label{Minfty}
% M^2_{\infty}(A,\tau_0)=\max_{\tau\in[0,\tau_0]}\rho(v(\tau),A\sqrt{\tau}).
% \end{align}
%Then, there exists a universal constant $c>0$ such that
% \begin{align}
% 8\sigma s_c\int_{0}^{\tau_0}(\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau&\leq c\tau_0^{1+s_c}\left[A^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2+\frac{[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A,\tau_0)}{A^{2(1-s_c)}}\right]\nonumber\\
% &\quad\quad+2\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right] \label{lemi}
% \end{align}
% and
% \begin{align}
% \frac{1}{R^{2s_c}}\int_{R\leq |x|\leq 2R}|v(\tau_0)|^2\,dx&\leq c\|v_0\|_{L^{\sigma_c}}^2+\frac{c}{A^4}\left[[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A,\tau_0)\right]\\
% &\quad\quad+\frac{2}{\tau_0^{s_c}A^{2(1+s_c)}}\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]. \,\,\,\,\,\, \label{lemii}
% \end{align}
%\end{lemma}
%\begin{proof}
% By the definition of the semi-norm $\rho$ and $M_{\infty}$ with $R=A\sqrt{\tau_0}$, see \eqref{defrho} and \eqref{Minfty} respectively, we have
% \begin{align}
% \rho(v(\tau),R)=\rho(v(\tau),A\sqrt{\tau_0})\leq \rho(v(\tau),A\sqrt{\tau})\leq M_{\infty}^2(A,\tau_0), \,\,\mbox{for all}\,\,\tau \in[0,\tau_0].
% \end{align}
% From the estimate \eqref{GNradial} with $R=A\sqrt{\tau_0}$, for all $\tau\in [0,\tau_0]$ and $\eta>0$ we obtain
% \begin{align}
% \int_{|x|\geq R}|x|^{-b}|v(\tau)|^{2\sigma+2}\,dx\leq \eta\|\nabla v(\tau)\|_{L^2}^2+\frac{C_\eta}{R^{2(1-s_c)}}[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}.\label{GN1}
% \end{align}
% Moreover, we deduce the following estimate
% \begin{align}
% \frac{1}{R^{2}}\int_{2R\leq|x|\leq 4R} |v(\tau)|^2\,dx&\leq \frac{(2R)^{2s_c}}{R^2}\frac{1}{(2R)^{2s_c}}\int_{2R\leq|x|\leq 4R} |v(\tau)|^2\,dx\leq \frac{c}{R^{2(1-s_c)}}\rho(v(\tau),2R)\\&\leq \frac{c}{R^{2(1-s_c)}}\rho(v(\tau),R)\leq \frac{c}{R^{2(1-s_c)}}M_{\infty}^2(A,\tau_0).\label{Minfty2}
% \end{align}
%Now, choosing $\eta>0$ small enough, we inject \eqref{GN1} and \eqref{Minfty2} into \eqref{estintradial} to obtain
% \begin{align}
% 4\sigma s_c\|\nabla v(\tau) \|^2_{L^2}+z''_R(\tau)
% &-16(\sigma s_c+1)E[v_0]\leq c \frac{[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A,\tau_0)}{R^{2(1-s_c)}}.\nonumber
% \end{align}
% Given $\tau\in [0,\tau_0]$ and integrating the above inequality from $0$ to $\tau$, we get
% \begin{align}
% 4\sigma s_c\int_0^\tau\|\nabla v(s)\|^2_{L^2}\,ds&+z'_R(\tau)\leq z'_R(0)+16\tau(\sigma s_c+1)E[v_0] +c\tau\frac{[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A,\tau_0)}{R^{2(1-s_c)}}.
% \end{align}
% We integrate one more time from $0$ to $\tau_0$ to obtain
% \begin{align}
% 4\sigma s_c\int_{0}^{\tau_0}\int_0^\tau\|\nabla v(s)\|^2_{L^2}\,ds d\tau+z_R(\tau_0)-z_R(0)\leq &\tau_0z'_R(0)+8\tau_0^2(\sigma s_c+1)E[v_0]\\
% & +c\frac{\tau_0^2}{2}\frac{[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A,\tau_0)}{R^{2(1-s_c)}}.\label{dblinte}
% \end{align}
% Now, note that integration by parts yields
% \begin{align}
% \int_{0}^{\tau_0}\int_{0}^{\tau}\|\nabla v(s)\|_{L^2}^2\,ds d\tau&=\int_{0}^{\tau_0}1\int_{0}^{\tau} \|\nabla v(s)\|_{L^2}^2\,ds d\tau\\
% &=\tau_0\int_{0}^{\tau_0}\|\nabla v(\tau)\|_{L^2}^{2}\,d\tau-\int_{0}^{\tau_0}\tau\|\nabla v(\tau)\|^2_{L^2}\,d\tau\\
% &=\int_{0}^{\tau_0}(\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^{2}\,d\tau.
% \end{align}
% Then, we can rewrite \eqref{dblinte} as
% \begin{align}
% 8\sigma s_c\int_{0}^{\tau_0}(\tau_0-\tau)&\|\nabla v(\tau)\|_{L^2}^{2}\,d\tau+2z_R(\tau_0)\\
% &\leq 2z_R(0)+2\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right]\nonumber\\
% &\quad+c\tau_0^2\frac{[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M_{\infty}^2(A,\tau_0)}{R^{2(1-s_c)}}\\
% & \leq cR^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2+2\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right]\nonumber\\
% &\quad+c\tau_0^2\frac{[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M_{\infty}^2(A,\tau_0)}{R^{2(1-s_c)}}\label{prinlem01},
% \end{align}
% where we have used, from inequality \eqref{radial1}, that
% \begin{align}\label{phiR1}
% z_R(0)\leq cR^{2}\int_{|x|\leq 4R}|v_0|^2\,dx\leq cR^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2.
% \end{align}
% Now, since $R=A\sqrt{\tau_0}$, we deduce from \eqref{prinlem01} that
% \begin{align}
% 8\sigma s_c\int_{0}^{\tau_0}(\tau_0-\tau)&\|\nabla v(\tau)\|_{L^2}^{2}\,d\tau+2z_R(\tau_0)\\
% & \leq c\tau_0^{1+s_c}A^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2+2\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right]\nonumber\\&\quad+c\tau_0^{1+s_c}\frac{[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M_{\infty}^2(A,\tau_0)}{A^{2(1-s_c)}}\label{prinlem},
% \end{align}
% which yields \eqref{lemi}.
% Next, observe that for all $R>0$ the definition of $\phi$ (see \eqref{phi}) yields
% \begin{align}
% \frac{z_R(\tau_0)}{R^{2(1+s_c)}}&=\frac{1}{R^{2s_c}}\int_{|x|\leq 4R} \phi\left(\frac{x}{R}\right)|v(\tau_0)|^2\,dx\\
% &\geq \frac{1}{R^{2s_c}}\int_{R\leq |x|\leq 2R}\left|\frac{x}{R}\right|^2|v(\tau_0)|^{2}\,dx\geq \frac{1}{R^{2s_c}}\int_{R\leq|x|\leq 2R}|v(\tau_0)|^2\,dx.
% \end{align}
% Dividing inequality \eqref{prinlem} by $R^{2(1+s_c)}$ where $R=A\sqrt{\tau_0}$, we have
% \begin{align}
% \frac{1}{R^{2s_c}}\int_{R\leq|x|\leq 2R}|v(\tau_0)|^2\,dx
% &\leq c\|v_0\|_{L^{\sigma_c}}^2+\frac{c}{A^4}\left[[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A,\tau_0)\right]\nonumber\\
% &\quad+\frac{2}{\tau_0^{s_c}A^{2(1+s_c)}}\left[z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right],
% \end{align}
%which implies \eqref{lemii} and concludes the proof.
%\end{proof}
% \begin{align}\label{phiR1}
% z_R(0)\leq cR^{2}\int_{|x|\leq 4R}|v_0|^2\,dx\leq cR^{2(1+s_c)}\|v_0\|_{L^{\sigma_c}}^2.
% \end{align}
%
%
% \begin{align}
% \frac{1}{R^{2}}\int_{2R\leq|x|\leq 4R} |v(\tau)|^2\,dx&\leq \frac{(2R)^{2s_c}}{R^2}\frac{1}{(2R)^{2s_c}}\int_{2R\leq|x|\leq 4R} |v(\tau)|^2\,dx\leq \frac{c}{R^{2(1-s_c)}}\rho(v(\tau),2R)\\&\leq \frac{c}{R^{2(1-s_c)}}\rho(v(\tau),R)\leq \frac{c}{R^{2(1-s_c)}}M_{\infty}^2(A,\tau_0).\label{Minfty2}
% \end{align}
%
%
%
%In what follows, let $v\in C\left([0,\tau_*]:\dot H_a^{s_c}\cap \dot H_a^1 \right)$ be a solution to \eqref{PVI} with initial data $v_0\in \dot H_a^{s_c}\cap \dot H_a^1$ and $\varepsilon>0$ a fixed small enough real number to be chosen later. For $M_0$ as in \eqref{hpprop1ii}%depending only of $\sigma_c$ and $v_0$
%, we define
%\begin{align}\label{Gep}
%G_{\varepsilon}=M_0^{\frac{1}{\varepsilon}}\,\,\, \mbox{and}\,\,\,A_{\varepsilon}=\left(\frac{\varepsilon G_{\varepsilon}}{M_0^2}\right)^{\frac{1}{2(1+s_c)}}.
%\end{align}
%
%Consider the following estimates
%\begin{align}\label{dispersioneps}
%\int_{0}^{\tau_0}(\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^{2}\,d\tau\leq G_{\varepsilon}\tau_0^{1+s_c}
%\end{align}
%and
%\begin{align}\label{Minftyep}
%M_\infty^2(A_\varepsilon,\tau_0)\leq \frac{2M_0^2}{\varepsilon}.
%\end{align}
%We define
%\begin{align}\label{tau11}
%S_{\varepsilon}= \left\{\tau\in [0,\tau_*];\,\,\eqref{dispersioneps}\mbox{ and }\eqref{Minftyep} \mbox{ hold for all }\tau_0\in [0,\tau]\right\}
%\end{align}
%and
%\begin{align}\label{tau1}
%\tau_1=\max_{\tau\in [0,\tau_*]}S_{\varepsilon}.
%\end{align}
%%Note that $S_{\varepsilon}\neq \emptyset$ for $\varepsilon>0$ sufficiently small. In fact, by the regularity of $v(\tau)$, there exists $0<\delta'<\min\{1,\tau_*\}$ such that
%%\begin{align}
%%\|\nabla v(\tau)\|_{L^2}^2<1+\|\nabla v_0\|_{L^2}^2,\,\,\,\mbox{for all}\,\,\, \tau\in [0,\delta'].
%%\end{align}
%%%definition of energy, the Gagliardo-Nirenberg inequality, the Young inequality and the definition \eqref{hpprop1ii} of $M_0$, we have
%%%\begin{align}
%%%E[v_0]\geq \frac{1}{2}\|\nabla v_0\|_{L^2}^2-\frac{1}{2}\|\nabla v_0\|_{L^2}^2\left(\frac{\|v_0\|_{L^{\sigma_c}}}{\|V\|_{L^{\sigma_c}}}\right)^{2\sigma}\geq \frac{1}{2}\|\nabla v_0\|_{L^2}^2-\frac{1}{2}\|\nabla v_0\|_{L^2}^4-\frac{1}{16}\left(\frac{M_0}{4}\right)^{4\sigma},
%%%\end{align}
%%%more precisely,
%%%\begin{align}
%%%-\|\nabla v_0\|_{L^2}^4+\|\nabla v_0\|_{L^2}^2\leq 2E[v_0]+\frac{1}{8}\left(\frac{M_0}{4}\right)^{4\sigma}.
%%%\end{align}
%%%Thus, there exists $C(v_0)>0$ such that $\|\nabla v_0\|_{L^2}<C(v_0)$.
%%Next, multiplying the last inequality by $(\tau_0-\tau)$ and integrating from $0$ to $\tau_0$ with $\tau_0\in [0,\delta']$, we have
%%\begin{align}
%%\int_{0}^{\tau_0}(\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau\leq \left(\frac{1+\|\nabla v_0\|_{L^2}^2}{2}\right)\tau_0^{1+s_c}\leq M_0^{\frac{1}{\varepsilon}}\tau_0^{1+s_c},
%%\end{align}
%%for sufficiently small $\varepsilon>0$, since $M_0\geq 2$. Furthermore, we recall from \eqref{radial1} and \eqref{hpprop1ii} that there exist universal constants $c,C>0$ such that for every $R>0$
%%%\begin{align}\label{rhoM_0}
%%%\frac{1}{R^{2s_c}}\int_{|x|\leq 4R}|v_0|^{2}\,dx=\frac{4^{2s_c}}{(4R)^{2s_c}}\int_{|x|\leq4R}|v_0|^2\,dx\leq c \|v_0\|_{L^{\sigma_c}}^2\leq c M_0^2.
%%%\end{align}
%%\begin{align}
%%\rho(v_0,R)\leq c\|v_0\|_{L^{\sigma_c}}^2\leq C M_0^2.
%%\end{align}
%%Again using the regularity of $v\in C([0,\tau_*]:\dot H^{s_c}\cap\dot H^1 )$ and definition \eqref{Minfty}, there exists $0<\delta''\leq\tau_*$ such that for all $\tau\in [0,\delta'']$
%%\begin{align}
%%%\|\nabla v(\tau)\|_{L^2}<C(v_0)\mbox{ and }
%%M_{\infty}^2(A_{\varepsilon},\tau)<2CM_0^2\leq \frac{2M_0^2}{\varepsilon},
%%\end{align}
%%for sufficiently small $\varepsilon>0$.
%%Hence, there exists $\varepsilon_0>0 $, depending only on $v_0$, such that $\min\{\delta',\delta''\}\in S_{\varepsilon}$, for all $\varepsilon\leq \varepsilon_0$.
%In the next lemma, we obtain an initial control on the virial quantity \eqref{zR'2}, assuming an a priori bound on the energy.
%%reducing this $\varepsilon_0>0$ if necessary
%
%%Also note that, from \eqref{radial1} and the definition \eqref{hpprop1ii} of $M_0$ , we have
%%\begin{align}\label{rhoM_0}
%%\forall R>0, \,\,\,\,\frac{1}{R^{2s_c}}\int_{|x|\leq 4R}|v_0|^{2}\,dx=\frac{4^{2s_c}}{(4R)^{2s_c}}\int_{|x|\leq4R}|v_0|^2\,dx\leq c \|v_0\|_{L^{\sigma_c}}^2\leq c M_0^2.
%%\end{align}
%
%\begin{lemma}\label{lemimint}
% Consider $0<b<2$ and $\frac{2-b}{N}<\sigma<\min\left\{\frac{2-b}{N-2},\frac{2}{N}\right\}$. Let $v_0\in \dot H_a^{s_c}\cap \dot H_a^1 $ and $v\in C([0,\tau_*]: \dot H_a^{s_c}\cap \dot H_a^1 )$ the corresponding solution to \eqref{PVI}. In addition, assume the energy control \eqref{hpprop1i} and let $\tau_1>0$ and $A_\varepsilon>0$ given by \eqref{tau1} and \eqref{Gep}, respectively.
%% \begin{align}\label{tau*}
%% \tau_*^{1-s_c}\max\{E[v_0],0\}<1.
%% \end{align}
%Then, there exists $\varepsilon_0>0 $ small enough and $c>0$ a universal constant such that, for all $\tau_0\in[0,\tau_1]$, $0<\varepsilon\leq \varepsilon_0$, $A\geq A_\varepsilon$ and $R=A\sqrt{\tau_0}$, the following inequality holds
% \begin{align}\label{ImE}
% z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\leq c \frac{M_0^2A^{2(1+s_c)}}{\varepsilon^{\frac{1}{1+s_c}}}\tau_0^{s_c}.
% \end{align}
%\end{lemma}
%\begin{proof}
%We split the proof in several steps.\\
% \textbf{Step 1.}
%Let $\tau_0\in[0,\tau_1]$ and $\varepsilon>0$ small enough. We first claim that there exists a universal constant $K(N,\sigma,b)>0$ and a time $t_0$ such that
% \begin{align}\label{thetaK}
% t_0 \in \left[\frac{\varepsilon^{\frac{1}{1+s_c}}}{4}\tau_0,\frac{\varepsilon^{\frac{1}{1+s_c}}}{2}\tau_0\right]\quad\mbox{ and }\quad \|\nabla v(t_0)\|_{L^2}^2\leq\frac{KG_\varepsilon}{t_0^{1-s_c}}.
% \end{align}
% In fact, let $t=\varepsilon^{\frac{1}{1+s_c}}\tau_0\leq\tau_0\leq\tau_1$. If \eqref{thetaK} does not occur, then from \eqref{dispersioneps} we have
% \begin{align}
% G_{\varepsilon}t^{1+s_c}\geq \int^{t}_{0}(t-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau\geq\frac{t}{2}\int_{\frac{t}{4}}^{\frac{t}{2}}\|\nabla v(\tau)\|_{L^2}^2\,d\tau> K G_{\varepsilon}\frac{t}{2}\int_{\frac{t}{4}}^{\frac{t}{2}}\frac{d\tau}{\tau^{1-s_c}}\geq cKG_{\varepsilon}t^{1+s_c}
% \end{align}
% which yields a contradiction for $K>0$ large enough and thus \eqref{thetaK} is valid.\\
% \textbf{Step 2.} For $A\geq A_{\varepsilon}$, let $A_1>0$ so that $R=A\sqrt{\tau_0}=A_1\sqrt{t_0}$, and thus,
% \begin{align}\label{epsA1}
% \frac{\varepsilon^{\frac{1}{2(1+s_c)}}}{2}\leq \frac{A}{A_1}\leq \varepsilon^{\frac{1}{2(1+s_c)}}.
% \end{align}
% Next, we show the following inequality
% \begin{align}\label{intphi1}
% \frac{z_R(t_0)}{R^{2(1+s_c)}}\leq c\left[M_0^2+\frac{G_\varepsilon}{A_1^{2(1+s_c)}}\right].
% \end{align}
% Indeed, by the virial identity \eqref{zR'2}, Cauchy-Schwarz inequality and the properties of $\phi$ \eqref{nablaphi} we obtain
% \begin{align}\label{phiRL2}
% z'_R(\tau)\leq 2\left|\textit{Im}\left(\int\nabla\phi_R\cdot\nabla v(\tau)\overline v(\tau)\,dx\right)\right|\leq c\|\nabla v(\tau)\|_{L^2}\left(z_R(\tau)\right)^{\frac{1}{2}}
% \end{align}
%and therefore
% \begin{align}
% \frac{d}{d\tau}\left[\left(z_R(\tau)\right)^{\frac{1}{2}}\right]\leq c\|\nabla v(\tau)\|_{L^2}.
% \end{align}
% Integrating the above inequality from $0$ to $t_0$, square both sides, using inequality \eqref{phiR1}, the definition of $M_0$ \eqref{hpprop1ii} and Cauchy-Schwarz inequality
% \begin{align}
% z_R(t_0)&\leq c \left[z_R(0)+\left(\int_0^{t_0}\|\nabla v(\tau)\|_{L^2}\,d\tau\right)^2\right]\\
% &
% \leq c\left[R^{2(1+s_c)}M_0^2+t_0\int_{0}^{t_0}\|\nabla v(\tau)\|_{L^2}^2\,d\tau\right].\label{intphi}
% \end{align}
% Since $2t_0\leq \varepsilon^{\frac{1}{1+s_c}}\tau_0\leq\tau_0\leq\tau_1$ and using estimate \eqref{dispersioneps}, we get
% \begin{align}\label{intzeta0}
% \int_0^{t_0}\|\nabla v(\tau)\|_{L^2}^2\,d\tau\leq \frac{1}{t_0}\int_{0}^{2t_0}(2t_0-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau\leq cG_\varepsilon t_0^{s_c}.
% \end{align}
% Thus, recalling that $R=A_1\sqrt{t_0}$, we combine \eqref{intzeta0} and \eqref{intphi} to obtain
% \begin{align}
% z_R(t_0)\leq c[R^{2(1+s_c)}M_0^2+G_{\varepsilon}t_0^{1+s_c}]\leq cR^{2(1+s_c)}\left[M_0^2+\frac{G_\varepsilon}{A_1^{2(1+s_c)}}\right]
% \end{align}
%completing the proof of \eqref{intphi1}.\\
% \textbf{Step 3.} Now, we control the virial quantity \eqref{zR'2} at time $t_0$. From the inequality in \eqref{phiRL2}, we apply \eqref{thetaK} and \eqref{intphi1} to deduce for $R=A\sqrt{\tau_0}=A_1\sqrt{t_0}$ that
% \begin{align}
% \left|z'_R(t_0)\right|&\leq c\left(\frac{KG_\varepsilon}{t_0^{1-s_c}}\right)^{\frac{1}{2}}R^{1+s_c }\left[M_0^2+\frac{G_\varepsilon}{A_1^{2(1+s_c)}}\right]^{\frac{1}{2}}\\
% &\leq c\frac{G_\varepsilon^{\frac{1}{2}}}{\left(\tau_0\frac{A^2}{A_1^2}\right)^{\frac{1-s_c}{2}}}A^{1+s_c}\tau_0^{\frac{1+s_c}{2}}\left[M_0+\frac{G_\varepsilon^{\frac{1}{2}}}{A_1^{1+s_c}}\right]\\\nonumber
% &= cM_0^2A^{2(1+s_c)}\tau_0^{s_c}\left[\left(\frac{G_\varepsilon}{A^{2(1+s_c)}M_0^2}\right)^{\frac{1}{2}}\left(\frac{A_1}{A}\right)^{1-s_c}\!\!\!+\frac{G_\varepsilon}{A^{2(1+s_c)}M_0^2}\left(\frac{A}{A_1}\right)^{2s_c}\right].\\
%\end{align}
%So, for all $A\geq A_\varepsilon$, we deduce
%\begin{align}
% \left|z'_R(t_0)\right|
%&\leq cM_0^2A^{2(1+s_c)}\tau_0^{s_c}\left[\left(\frac{G_\varepsilon}{A_\varepsilon^{2(1+s_c)}M_0^2}\right)^{\frac{1}{2}}\left(\frac{A_1}{A}\right)^{1-s_c}\!\!\!+\frac{G_\varepsilon}{A_\varepsilon^{2(1+s_c)}M_0^2}\left(\frac{A}{A_1}\right)^{2s_c}\right]\\&
% \leq c\frac{M_0^2A^{2(1+s_c)}}{\varepsilon^{\frac{1}{1+s_c}}}\tau_0^{s_c},\label{intImzeta}
% \end{align}
%where in the last inequality we have used definition \eqref{Gep} and relation \eqref{epsA1}.\\
% \textbf{Step 4.} Finally, we prove an initial control on the virial quantity \eqref{zR'2}. Recall that $A\geq A_{\varepsilon}$, $R=A\sqrt{\tau_0}$ and $2t_0\leq \varepsilon^{\frac{1}{1+s_c}}\tau_0$. From \eqref{Minfty2}, \eqref{Gep} and \eqref{Minftyep}, for all $\tau_0\in [0,\tau_1]$
% \begin{align}
% \left|\int\Delta^2\phi_R|v(\tau_0)|^2\,dx\right|&\leq \frac{c}{R^2}\int_{2R\leq|x|\leq 4R}|v(\tau_0)|^2\,dx\leq c\frac{M_\infty^2(A_\varepsilon,\tau_0)}{R^{2(1-s_c)}}\\&\leq c\frac{M_0^2}{\varepsilon}\frac{\varepsilon^{\frac{1-s_c}{1+s_c}}}{A_\varepsilon^{2(1-s_c)}t_0^{1-s_c}}
% \leq c\frac{M_0^2}{\varepsilon}\frac{M_0^{\frac{2(1-s_c)}{1+s_c}}}{G_\varepsilon^{\frac{1-s_c}{1+s_c}}}\frac{1}{t_0^{1-s_c}}\leq \frac{1}{t_0^{1-s_c}},\label{lapacian2}
% \end{align}
% for all $\varepsilon\leq \varepsilon_0$ with $\varepsilon_0>0$ small enough. Moreover, by the properties of $\phi$ \eqref{nablaphi} and the energy conservation \eqref{Energy} we have the crude estimates
% \begin{align}
% \left| \int\nabla\left(|x|^{-b}\right)\cdot \nabla\phi_R|v(\tau_0)|^{2\sigma+2}\,dx \right|&\leq c\int |x|^{-b} |v(\tau_0)|^{2\sigma+2}\,dx\label{lapacian3}
% \end{align}
%and
%\begin{equation}
%\int |x|^{-b} |v(\tau_0)|^{2\sigma+2}\,dx\leq c\left( \|\nabla v(\tau_0)\|^2_{L^2} +|E[v_0]|\right)
%\end{equation}
%Inserting the previous three inequalities in the virial quantity \eqref{zR''}, we deduce%\footnote{The proof of this estimate is very close to that of Lemma \ref{lemintradial}}
% \begin{align}
% \left|z''_R(\tau_0)\right|\leq c\left[\|\nabla v(\tau_0)\|^2_{L^2}+|E[v_0]|+\frac{1}{t_0^{1-s_c}}\right], \,\,\, \mbox{for all} \,\,\, \tau_0\in [0,\tau_1].
% \end{align}
% Now, integrating from $0$ to $t_0$, we obtain
% \begin{align}
% \left|z'_R(0)\right|&\leq \left| z'_R(t_0) \right|+c\left[\int_0^{t_0}\|\nabla v(\tau)\|_{L^2}^2\,d\tau+t_0|E[v_0]|+t_0^{s_c}\right].
% \end{align}
%In view of inequalities \eqref{intzeta0} and \eqref{intImzeta} we deduce
% \begin{align}\label{Imzeta}
% z'_R(0) \leq& c\frac{M_0^2A^{2(1+s_c)}}{\varepsilon^{\frac{1}{1+s_c}}}\tau_0^{s_c}+ cG_\varepsilon t_0^{s_c}+ct_0|E[v_0]|.
% \end{align}
%To estimate the second term in the right hand side of \eqref{Imzeta} we use \eqref{Gep} and \eqref{thetaK} to obtain
% \begin{align}\label{Gepzeta}
% G_\varepsilon t_0^{s_c}\leq \frac{A_\varepsilon^{2(1+s_c)} M_0^2}{\varepsilon}\frac{\varepsilon^{\frac{s_c}{1+s_c}}}{2^{s_c}}\tau_0^{s_c}\leq c\frac{M_0^2A^{2(1+s_c)}}{\varepsilon^{\frac{1}{1+s_c}}}\tau_0^{s_c}.
% \end{align}
%On the other hand, to control the last term in the right hand side of \eqref{Imzeta}, we observe that
%$$
%8(\sigma s_c+1)E[v_0]+c\varepsilon^{\frac{1}{1+s_c}}|E[v_0]|\leq c \max\{E[v_0],0\},
%$$
%for all $\varepsilon>0$ small enough. Therefore, using again \eqref{thetaK} and the assumption \eqref{hpprop1i} we have
%\begin{align}\label{maxE}
% 8\tau_0(\sigma s_c+1)E[v_0]+ct_0|E[v_0]|&\leq \left[8(\sigma s_c+1)E[v_0]+c\varepsilon^{\frac{1}{1+s_c}}|E[v_0]|\right]\tau_0\\\nonumber
% &\leq c\max\{E[v_0],0\}\tau_0^{1-s_c}\tau_0^{s_c}<c\tau_0^{s_c}.
% \end{align}
%Finally, collecting \eqref{Imzeta}, \eqref{Gepzeta} and \eqref{maxE} we complete the proof of estimate \eqref{ImE}.
%\end{proof}
%\subsection{Upper bound on the blow-up rate for radial data in $H^1$}
%Before we proceed to the proof of Propositions \ref{prop1} and \ref{prop2}, if we assume in addition that $u_0\in H^1 $, we obtain the following upper bound on blow-up rate for radially symmetric initial data, which is a consequence of the localized virial identity \eqref{virial}. The proof of the theorem below is based on the ideas of Merle, Rapha\"{e}l, and Szeftel \cite{MRS2014}.
%\begin{thm}\label{thmblowdisper}
% Let $N\geq 3$, $0<b<\min\left\{\frac{N}{2},2\right\}$ and $\frac{2-b}{N}<\sigma<\frac{2-b}{N-2}$. Let $u_0\in H^1 $ with radial symmetry and assume that the maximal time of existence $T^{\ast}>0$ for the corresponding solution $u\in C([0,T^{\ast}):H^1 )$ of \eqref{PVI} is finite. Define $\beta=\frac{2-\sigma}{\sigma(N-1)+b}$, then the following space-time upper bound holds
% \begin{align}\label{blowupestdisp}
% \int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\nabla u(\tau)\|_{L^2}^2\,d\tau\leq C_{u_0}(T^{\ast}-t)^{\frac{2\beta}{1+\beta}},
% \end{align}
%for $t$ close enough to $T^{\ast}$.
%\end{thm}
%\begin{rem}
% As a consequence of \eqref{blowupestdisp} we have the following upper bound
% \begin{align}
% \liminf_{t\uparrow T^{\ast}}(T^{\ast}-t)^{\frac{1}{1+\beta}}\|\nabla u(t)\|_{L^2}<+\infty.
% \end{align}
% In particular, there exists a sequence $\{t_n\}_{n=1}^{+\infty}\subset [0,T^{\ast})$ with $t_n\to T^{\ast}$ such that
% \begin{align}
% \|\nabla u(t_n)\|_{L^2}\leq \frac{C}{(T^{\ast}-t_n)^{\frac{1}{1+\beta}}},\,\,\,\, \mbox{ as }\,\,\,\, n\to+\infty.
% \end{align}
% It should be emphasized that it is an open problem to show the previous upper bound for every sequence $\{t_n\}_{n=1}^{+\infty}$ with $t_n\to T^{\ast}$.
%\end{rem}
%
%
%
%
%\begin{proof}[Proof of Theorem \ref{thmblowdisper}] Let $R,\varepsilon>0$ real number to be chosen later. By the Strauss inequality for radial functions \cite{Strauss} (see, for instance, \cite[Lemma 4.1]{FG20} for a complete proof) we first have
% \begin{align}
% \|u\|_{L^{\infty}(|x|\geq R)}\leq c\frac{\|\nabla u\|_{L^2}^{\frac{1}{2}}\|u\|_{L^2}^{\frac{1}{2}}}{R^{\frac{N-1}{2}}}.
% \end{align}
% Then, by the mass conservation \eqref{Mass} and Young's inequality
% \begin{align}
% \int_{|x|\geq R}|x|^{-b}|u|^{2\sigma+2}\,dx&\leq \frac{1}{R^{b}}\|u\|_{L^{\infty}(|x|\geq R)}^{2\sigma}\|u\|^{2}_{L^2(|x|\geq R)}\leq \frac{C(u_0)}{R^{\sigma(N-1)+b}}\|\nabla u\|_{L^2}^{\sigma}\|u_0\|_{L^2}^{\sigma+2}\\& \leq \varepsilon \|\nabla u\|_{L^2}^2+\frac{C(u_0,\sigma,\varepsilon)}{R^{\frac{2\sigma(N-1)+2b}{2-\sigma}}},\label{case2}
% \end{align}
%where we have used that $\sigma<2$.
% Let $\beta=\frac{2-\sigma}{\sigma(N-1)+b}$. Combining inequality \eqref{estintradial}, energy and mass conservation \eqref{Energy}-\eqref{Mass} and the previous inequality we deduce
% \begin{align}
% 2\sigma s_c\int |\nabla u|^2
% \,dx+\frac{d}{d\tau}\textit{Im}&\int \nabla \phi_R\nabla u\overline{u}\,dx\\&\leq C_{u_0}\left(
% 1+\frac{1}{R^2}+\int_{|x|\geq R}|x|^{-b}|u|^{2\sigma+2}\,dx\right)\nonumber
% \\
% &\leq C_{u_0}\left(1+\frac{1}{R^2}+\varepsilon\int|\nabla u|^2\,dx+\frac{1}{R^{\frac{2}{\beta}}}\right).
% \end{align}
% Fix $\varepsilon>0$ small enough. If $R\ll1$, since $0<\beta<1$, then
% \begin{align}\label{ctR}
% \sigma s_c\int|\nabla u|^2\,dx+\frac{d}{d\tau}\textit{Im}\int\nabla \phi_R\nabla u\overline{u}\,dx\leq \frac{C_{u_0}}{R^{\frac{2}{\beta}}}.
% \end{align}
% Integrating \eqref{ctR} from $t$ to $\tau $ and then from $t$ to $t_2$, in time, we get
% \begin{align}
% \int_{t}^{t_2}(t_{2}-\tau)\|\nabla u(\tau&)\|_{L^2}^2\,d\tau +\frac{1}{2}\int\phi_R|u(t_2)|^2\,dx\\
% &\leq C_{u_0} \frac{(t_2-t)^2}{2R^{\frac{2}{\beta}}}+\frac{1}{2}\int\phi_R|u(t)|^2\,dx+(t_{2}-t)\textit{Im}\int \nabla \phi_R\nabla u \overline{u}(t)\,dx\nonumber\\
% &\leq C_{u_0}\left(\frac{(t_2-t)^2}{R^{\frac{2}{\beta }}}+R^2+R(t_2-t)\|\nabla u(t)\|_{L^2} \right),
% \end{align}
%where we have used Cauchy-Schwarz inequality and the mass conservation \eqref{Mass} in the last inequality. Taking $t_2\to T^{\ast}$ in the above inequality, we conclude that the first integral on the left hand side converges and
% \begin{align}\label{cotaint}
% \int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\nabla u(\tau&)\|_{L^2}^2\,d\tau
% \leq C_{u_0}\left(\frac{(T^{\ast}-t)^2}{R^{\frac{2}{\beta}}}+R^2+R(T^{\ast}-t)\|\nabla u(t)\|_{L^2}\right).
% \end{align}
% Thus, choosing $R\ll1$ so that the first two terms on the right hand side are equals, more precisely
% \begin{align}
% R(t)=(T^{\ast}-t)^{\frac{\beta}{1+\beta}},
% \end{align}
%the inequality \eqref{cotaint} can be rewritten as
% \begin{align}
% \int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\nabla u(\tau)\|_{L^2}^2\,d\tau
% &\leq C_{u_0}\left((T^{\ast}-t)^{\frac{2\beta}{1+\beta}}+(T^{\ast}-t)^{\frac{\beta}{1+\beta}}(T^{\ast}-t)\|\nabla u(t)\|_{L^2}\right)\\
% &\leq C_{u_0} (T^{\ast}-t)^{\frac{2\beta}{1+\beta}}+(T^{\ast}-t)^2\|\nabla u(t)\|_{L^2}^2.\label{cotaint2}
% \end{align}
%where we have used Young's inequality in the last step.
%
% Define
% \begin{align}
% h(t)=\int_{t}^{T^{\ast}}(T^{\ast}-\tau)\|\nabla u(\tau)\|_{L^2}^{2}\,d\tau.
% \end{align}
% Note that inequality \eqref{cotaint2} implies that $h(0)\leq C_{u_0}$ and moreover
% \begin{align}
% h(t)\leq C_{u_0}(T^{\ast}-t)^{\frac{2\beta}{1+\beta}}-(T^{\ast}-t)h'(t),
% \end{align}
% that is,
% \begin{align}
% \left(\frac{h(t)}{T^{\ast}-t}\right)'=\frac{1}{(T^{\ast}-t)^2}((T^{\ast}-t)h'(t)+h(t))\leq C_{u_0} \frac{1}{(T^{\ast}-t)^{\frac{2}{1+\beta}}}.
% \end{align}
% Integrating the above inequality from $0$ to $t$,
% \begin{align}
% \frac{h(t)}{T^{\ast}-t}\leq \frac{h(0)}{T^{\ast}}+ \frac{C_{u_0}}{(T^{\ast}-t)^{\frac{1-\beta}{1+\beta}}}-\frac{C_{u_0}}{(T^{\ast})^{\frac{1-\beta}{1+\beta}}}\leq \frac{C_{u_0}}{(T^{\ast}-t)^{\frac{1-\beta}{1+\beta}}},
%%\leq \frac{C_{u_0}}{T^{\ast}-t}\left((T^{\ast}-t)+(T^{\ast}-t)^{\frac{2\beta}{1+\beta}}\right)
% \end{align}
%for $t$ close enough to $T^{\ast}$, since $0<\beta<1 $. Therefore
% \begin{align}
% h(t)\leq C_{u_0}(T^{\ast}-t)^{\frac{2\beta}{1+\beta}},
% \end{align}
%which implies \eqref{blowupestdisp} and completes the proof.
%\end{proof}
%\subsection{The proofs of Propositions \ref{prop1} and \ref{prop2}}\label{5.3.3}
%\begin{proof}[Proof of Proposition \ref{prop1}.] We will show that the estimates \eqref{prop1ii} and \eqref{prop1i} are consequence of \eqref{lemi} and \eqref{lemii} obtained in Lemma \ref{lemma34}. Recall inequalities \eqref{dispersioneps} and \eqref{Minftyep} used in the definition of $\tau_1$ in \eqref{tau1}. We first claim that for all $\tau_0\in [0,\tau_1]$, we have
% \begin{align}\label{intep1}
% \int_{0}^{\tau_0}(\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^{2}\,d\tau\leq \frac{G_{\varepsilon}}{2}\tau_0^{1+s_c}
% \end{align}
% and
% \begin{align}\label{intep2}
% M_{\infty}^2(A_{\varepsilon},\tau_0)=\max_{\tau\in [0,\tau_0]}\rho(v(\tau),A_{\varepsilon}\sqrt{\tau})\leq \frac{M^2_0}{\varepsilon}.
% \end{align}
% for $\varepsilon>0$ small enough. Assuming that \eqref{intep1} and \eqref{intep2} hold, by the regularity of $v$, we obtain a contradiction with the definition of $\tau_1$. Therefore inequalities \eqref{dispersioneps} and \eqref{Minftyep} hold at the maximal time $\tau_{\ast}$, which clearly imply \eqref{prop1ii} and \eqref{prop1i}, in view of definitions \eqref{Gep}.
%% for $0<\tau_1<\tau_2<\tau_*$, by the regularity of $v$, we obtain
%% \begin{align}
%% \int^{\tau_2}_0(\tau_2-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau&=\int^{\tau_1}_0(\tau_1-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau+\int_{\tau_1}^{\tau_2}(\tau_2-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau\nonumber\\&\quad+(\tau_2-\tau_1)\int_{0}^{\tau_1}\|\nabla v(\tau)\|_{L^2}^2\,d\tau\\
%% & \leq \frac{G_{\varepsilon}}{2}\tau_1^{1+s_c}+C\left(\tau_2-\tau_1\right)\leq G_{\varepsilon}\tau_2^{1+s_c},\label{inttau2}
%% \end{align}
%% for $0<\tau_2-\tau_1$ small enough. Moreover, from \eqref{intep2} we have that
%% $$\rho(v(\tau_1),A_{\varepsilon}\sqrt{\tau_1})<\frac{M_0^2}{\varepsilon}.$$
%%
%% Thus, again by the regularity of $v$ and taking $0<\tau_2-\tau_1$ small enough
%% \begin{align}\label{Mtau2}
%% M_{\infty}^2(A_{\varepsilon},\tau_2)\leq \max_{\tau\in[0,\tau_1]}\rho(v(\tau),A_{\varepsilon}\sqrt{\tau})+\max_{\tau\in[\tau_1,\tau_2]}\rho(v(\tau),A_{\varepsilon}\sqrt{\tau})\leq 2\frac{M_0^2}{\varepsilon}.
%% \end{align}
%%However \eqref{inttau2} and \eqref{Mtau2} contradicts the definition of $\tau_1$.
%
%Thus, it remains to show \eqref{intep1} and \eqref{intep2}. First observe that from \eqref{Gep} and \eqref{Minftyep} we deduce
% \begin{align}\label{110}
% \frac{[M^2_{\infty}(A_{\varepsilon},\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A_{\varepsilon},\tau_0)}{A_{\varepsilon}^{2(1-s_c)}}
%&\leq \frac{\left(\frac{2M_0^2}{\varepsilon}\right)^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+\frac{2M_0^2}{\varepsilon}}{\left(\frac{\varepsilon G_{\varepsilon}}{M_0^2}\right)^{\frac{1-s_c}{1+s_c}}}\leq c\frac{1}{M_0^{\frac{1}{\varepsilon}\cdot \frac{1-s_c}{1+s_c}}} \left(\frac{M_0^2}{\varepsilon}\right)^{\frac{2\sigma+2-\sigma N}{2-\sigma N}+\frac{1-s_c}{1+s_c}}\leq \frac{1}{10}, %\\
%% &\leq \frac{\left(\frac{M_0^2}{\varepsilon}\right)^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+\frac{M_0^2}{\varepsilon}}{\left(\frac{\varepsilon}{M_0^2}\right)^{\frac{1-s_c}{1+s_c}}G_{\varepsilon}^{\frac{1-s_c}{1+s_c}}}
%% \leq \left(\frac{M_0}{\varepsilon}\right)^C\frac{1}{G_{\varepsilon}^{\frac{1-s_c}{1+s_c}}}\leq \frac{1}{2^{\frac{1}{C_{\varepsilon}}}\varepsilon^C}\leq \frac{1}{10},
% \end{align}
%for $\varepsilon>0$ small enough, where we have used the assumption \eqref{hpprop1ii}. Combining the last inequality with \eqref{lemi} and using again definitions \eqref{Gep} and assumption \eqref{hpprop1ii}, we have, for $R=A_{\varepsilon}\sqrt{\tau_0}$, that
% \begin{align}
% \int_{0}^{\tau_0}(\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau&\leq c\tau_0^{1+s_c}\left[M_0^2A_{\varepsilon}^{2(1+s_c)}+\frac{[M^2_{\infty}(A_{\varepsilon},\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A_{\varepsilon},\tau_0)}{A_{\varepsilon}^{2(1-s_c)}}\right]\nonumber\\
% &\quad+2c\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]\\
% & \leq c\tau_0^{1+s_c}\left[\varepsilon G_{\varepsilon}+\frac{1}{10}\right]+2c\tau_0\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]\nonumber\\
% & \leq G_{\varepsilon}\tau_0^{1+s_c}\left\{c\varepsilon+\frac{c}{10G_{\varepsilon}}+\frac{2c}{G_{\varepsilon}\tau_0^{s_c}}\left[z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right]\right\}\nonumber\\
% &\leq G_{\varepsilon}\tau_0^{1+s_c}\left\{\frac{1}{10}+\frac{2c}{G_{\varepsilon}\tau_0^{s_c}}\left[z'_R(0)+8\tau_0(\sigma s_c+1)E[v_0]\right]\right\},\nonumber
% \end{align}
%for $\varepsilon>0$ small enough. Now, from Lemma \ref{lemimint} with $A=A_\varepsilon$, for all $\tau_0\in [0,\tau_1]$ and since $R=A_{\varepsilon}\sqrt{\tau_0}$, we get
% \begin{align}
% \int_{0}^{\tau_0}(\tau_0-\tau)\|\nabla v(\tau)\|^2_{L^2}\,d\tau&\leq G_{\varepsilon}\tau_0^{1+s_c}\left[\frac{1}{10}+c\frac{M_0^2A_{\varepsilon}^{2(1+s_c)}}{G_{\varepsilon}\varepsilon^{\frac{1}{1+s_c}}}\right]\\
% &=G_{\varepsilon}\tau_0^{1+s_c}\left[\frac{1}{10}+c\varepsilon^{\frac{s_c}{1+s_c}}\right]\leq \frac{G_{\varepsilon}}{2}\tau_0^{1+s_c},
% \end{align}
%and thus \eqref{intep1} is proved.
%
%To show \eqref{intep2}, let $A\geq A_{\varepsilon}$ and $R=A\sqrt{\tau_0}$. First, since $\rho$ is non-increasing in $R$, then for all $\tau_0\in [0,\tau_1]$ we obtain
% \begin{align}
% M^2_{\infty}(A,\tau_0)\leq M^2_{\infty}(A_{\varepsilon},\tau_0)\leq M^2_{\infty}(A_{\varepsilon},\tau_1)\leq \frac{2M_0^2}{\varepsilon},
% \end{align}
%where in the last inequality we have used \eqref{Minftyep}.\\
%Combining \eqref{lemii} with \eqref{hpprop1ii}, the last inequality and \eqref{ImE} we deduce
% \begin{align}
% \frac{1}{R^{2s_c}}\int_{R\leq |x|\leq 2R}|v(\tau_0)|^2\,dx&\leq cM^2_0+\frac{c}{A^4}\left[[M^2_{\infty}(A,\tau_0)]^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+M^2_{\infty}(A,\tau_0)\right]\\
% &\quad +\frac{2}{\tau_0^{s_c}A^{2(1+s_c)}}\left[z'_R(0)+8\tau_0(\sigma s_c+1) E[v_0]\right]\nonumber\\
% &\leq cM_0^2+c\frac{\left(\frac{2M_0^2}{\varepsilon}\right)^{\frac{2\sigma+2-\sigma N}{2-\sigma N}}+\frac{2M_0^2}{\varepsilon}}{\left(\frac{\varepsilon G_{\varepsilon}}{M_0^2}\right)^{\frac{2}{1+s_c}}}+\frac{cM_0^2}{\varepsilon^{\frac{1}{1+s_c}}}\\
% &=\frac{M_0^2}{\varepsilon}\left[c\varepsilon+\frac{c\varepsilon}{M_0^{2+\frac{1}{\varepsilon}\cdot \frac{2}{1+s_c}}} \left(\frac{M_0^2}{\varepsilon}\right)^{\frac{2\sigma+2-\sigma N}{2-\sigma N}+\frac{2}{1+s_c}}+c\varepsilon^{\frac{s_c}{1+s_c}}\right]< \frac{M_0^2}{\varepsilon},
% \end{align}
% for $\varepsilon>0$ small enough. The last inequality implies \eqref{intep2} and completes the proof of Proposition \ref{prop1}.
%\end{proof}
%\begin{proof}[Proof of Proposition \ref{prop2}]
\begin{proof}
% The argument is quite similar to the one in \cite[Proposition 4.2]{CF20} and we only sketch the details.
Given $v(\tau_0)$ we define
\begin{align}
w(x)=\lambda_v^{\frac{2-b}{2\sigma}}(\tau_0)v(\lambda_v(\tau_0)x,\tau_0),
\end{align}
where $\lambda_v(\tau_0)=\|\mathcal L_a^{\frac12}
v(\tau_0)\|^{-\frac{1}{1-s_c}}_{L^2}$,
and
\begin{align}\label{Aep}
J_{*}=C_{*}\max[M_0^{\alpha_1}F_*,M_0^{\frac{\gamma}{1-s_c}}],
\end{align}
for $C_*>1$ large enough.
%By a straightforward calculation and assumption \eqref{Ev}, we have
%\begin{equation}\label{wgrad}
%\|\mathcal L_a^{\frac12}
% w\|_{L^2}^2=1\,\,\, \mbox{and} \,\,\, E[w]\leq \frac{1}{4}.
%\end{equation}
%Thus, by the definition of the energy \eqref{Energy}, we get
%\begin{equation}\label{xbw}
%\int |x|^{-b}|w|^{2\sigma+2}\,dx\geq \frac{\sigma+1}{2}.
%\end{equation}
%Recalling the definition of $F_{\ast}$ in \eqref{F}, for $J\geq M_0^{\alpha_1}F_*$, the estimate \eqref{prop1ii} obtained in Proposition \ref{prop1} and the definition of the semi-norm $\rho$ \eqref{defrho} imply
%\begin{equation}\label{wJast}
%\rho(w,J)\leq \rho(v(\tau_0),M_0^{\alpha_1}\sqrt{\tau_0})\leq c M_0^2.
%\end{equation}
%On the other hand, for $J\geq CM_0^{\frac{\gamma}{1-s_c}}$ and all $\eta>0$, since $a>-\frac{(N-2)^2}{4}$, from estimates \eqref{GNradial} and \eqref{GNradial2} we get
% \begin{align}
% \int_{|x|\geq J}|x|^{-b}|w|^{2\sigma+2}\,dx&\leq \eta \|\mathcal L_a^{\frac12}
% w\|_{L^2}^2+\frac{C_\eta}{J^{2(1-s_c)}}\left\{[\rho(w,J)]^{\gamma}+[\rho(w,J)]^{\sigma+1}\right\}\\
% &\leq \eta+\frac{C_\eta}{C^{2(1-s_c)}M_0^{2\gamma}}\left\{[\rho(w,J)]^{\gamma}+[\rho(w,J)]^{\sigma+1}\right\}.
% \end{align}
%Now, taking
%\begin{align}\label{Aep}
% J_{*}=C_{*}\max[M_0^{\alpha_1}F_*,M_0^{\frac{\gamma}{1-s_c}}],
% \end{align}
%for $C_*>1$ large enough, the previous two inequalities (with $\eta>0$ small enough) and \eqref{xbw} yield
% \begin{align}\label{sigma+14}
% \int_{|x|\leq J_*}|x|^{-b}|w|^{2\sigma+2}\,dx\geq\frac{\sigma+1}{4}.
% \end{align}
%
% Let $\varphi\in C^{\infty}_0(\Real^N)$ be a cut-off function such that
% \begin{align}\label{fi}
% \varphi(x)=\left\{
% \begin{array}{ll}
% 1,&|x|\leq 1\\
% 0,&|x|\geq 2
% \end{array}
% \right.
% \,\,\,\mbox{ and }\,\,\, \varphi_{A}(x)=\varphi\left(\frac{x}{{A}}\right).
% \end{align}
%Using \eqref{wgrad} we get
% \begin{align}\label{prodcorte}
% \|\nabla (\varphi_{J_*} w)\|_{L^2}\leq \|\nabla \varphi_{J_*} w\|_{L^2}+\|\varphi_{J_*}\nabla w\|_{L^2}\leq c(\|w\|_{L^2(|x|\leq 2J_*)}+1).
% \end{align}
%Combining \eqref{GNF}, \eqref{sigma+14}
%and \eqref{prodcorte}, we deduce
% \begin{align}
% \frac{\sigma+1}{4}&\leq \int_{|x|\leq J_*} |x|^{-b}|w|^{2\sigma+2}\,dx\leq \int|x|^{-b}|\varphi_{J_*} w|^{2\sigma+2}\,dx\nonumber\\&\leq c\|\nabla (\varphi_{J_*} w)\|_{L^2}^{2\sigma s_c+2}\|\varphi_{J_*} w\|_{L^2}^{2\sigma(1-s_c)}\leq c\left(\|w\|_{L^2(|x|\leq 2J_*)}^{2\sigma+2}+\|w\|_{L^2(|x|\leq 2J_*)}^{2\sigma(1-s_c)}\right).
% \end{align}
Following the same arguments as in the proof of Proposition 3.3 in \cite{CF23}, there exists a constant $c>0$ such that
\begin{align}\label{364}
c\leq \int_{|y|\leq 2 J_*}|w|^2\,dx=\frac{1}{\lambda_v^{2s_c}(\tau_0)}\int_{|x|\leq 2J_*\lambda_v(\tau_0)}|v(\tau_0)|^2\,dx.
\end{align}
Now, we want to verify that the above estimate is also verified for the initial data $v_0$. Indeed, for $R>0$, applying H\"older's inequality and equivalence \eqref{equiv} we have
\begin{align}
\left|\frac{d}{d\tau}\int \varphi_{R}|v|^2\,dx\right|=2\left|\textit{Im}\left(\int \nabla \varphi_{{R}}\cdot \nabla v\overline{v}\,dx\right)\right|&\leq \frac{c}{{R}}\|\mathcal L_a^{\frac12}
v(\tau)\|_{L^2}\left(\int_{{R}\leq |x|\leq 2{R}}|v(\tau)|^2\,dx\right)^{\frac{1}{2}}
\\&\leq c\frac{\rho(v(\tau),{R})^{\frac{1}{2}}}{{{R}}^{1-s_c}}\|\mathcal L_a^{\frac12}
v(\tau)\|_{L^2}.
\end{align}
Next, integrating the above inequality from $0$ to $\tau_0\in[0,\frac{\tau_*}{2}]$ and using Cauchy-Schwarz inequality and \eqref{prop1i} we get
\begin{align}
\left|\int \varphi_{{R}}|v(\tau_0)|^2\,dx\right.&\left.-\int \varphi_{{R}}|v_0|^2\,dx\right|\\
&\leq c\frac{\rho(v(\tau),{R})^{\frac{1}{2}}}{{R}^{1-s_c}}\int_0^{\tau_0}\|\mathcal L_a^{\frac12}v(\tau)\|_{L^2}\,d\tau \leq c\frac{\rho(v(\tau),{R})^{\frac{1}{2}}}{{R}^{1-s_c}}\left(\tau_0\int_0^{\tau_0}\|\mathcal L_a^{\frac12}
v(\tau)\|_{L^2}^2\,d\tau\right)^{\frac12}\\
&\leq c\frac{\rho(v(\tau),{R})^{\frac{1}{2}}}{{R}^{1-s_c}}\left(\int_0^{2\tau_0}(2\tau_0-\tau)\|\mathcal L_a^{\frac12}
v(\tau)\|_{L^2}^2\,d\tau\right)^{\frac{1}{2}} \leq c\rho(v(\tau),{R})^{\frac{1}{2}} M_0^{\frac{\alpha_2}{2}}{R}^{2s_c}\left(\frac{\sqrt{\tau_0}}{R}\right)^{1+s_c}.\label{prop21}
\end{align}
Given $0<\varepsilon\ll 1$, define
\begin{align}\label{Dep}
D_\varepsilon=\frac{1}{\varepsilon} \max\left\{F_*,F_*^{\frac{1+s_c}{1-s_c}}\right\}\max\left\{M_0^{\alpha_1},M_0^{\frac{2+\alpha_2}{2(1-s_c)}}\right\},
\end{align}
and for $D\geq D_\varepsilon$ consider
\begin{align}\label{Rtil}
\widetilde{R}=\widetilde{R}(D,\tau_0)=D\lambda_v(\tau_0).
\end{align}
From \eqref{F} and \eqref{Rtil}, we get
\begin{align}
\tilde{R}=D\lambda_v(\tau_0)\geq D\frac{\lambda_v(\tau_0)}{\sqrt{\tau_0}}\sqrt{\tau_0}\geq \frac{D}{F_*}\sqrt{\tau}%\geq M_0^{\alpha_1}\sqrt{\tau}
, \,\,\,\mbox{for all}\,\,\, \tau\in [0,\tau_0],
\end{align}
and thus, using \eqref{Dep}, the monotonicity of $\rho$ and \eqref{prop1ii}, we have
\begin{align}
\rho(v(\tau),\tilde{R})\leq \rho(v(\tau),M_0^{\alpha_1}\sqrt{\tau})\leq cM_0^2, \,\,\,\mbox{for all}\,\,\, \tau\in [0,\tau_0].
\end{align}
Thereby, from \eqref{prop21}
\begin{align}
\frac{1}{\lambda^{2s_c}_v(\tau_0)}\left|\int \varphi_{\tilde{R}}|v(\tau_0)|^2\,dx-\int \varphi_{\tilde{R}}|v_0|^2\,dx
\right|
\leq cM_0^{1+\frac{\alpha_2}{2}}D^{2s_c}\left(\frac{F_*}{D}\right)^{1+s_c}=cM_0^{\frac{2+\alpha}{2}}\frac{F_*^{1+s_c}}{D^{1-s_c}}.
\end{align}
Finally, since $D\geq D_{\varepsilon}\geq \frac{1}{\varepsilon} F_*^{\frac{1+s_c}{1-s_c}}M_0^{\frac{2+\alpha_2}{2(1-s_c)}}$ by \eqref{Dep}, we obtain \begin{align}
\frac{1}{\lambda^{2s_c}_v(\tau_0)}\left|\int \varphi_{\tilde{R}}|v(\tau_0)|^2\,dx-\int \varphi_{\tilde{R}}|v_0|^2\,dx
\right|
< \varepsilon
\end{align}
which concludes the proof of Proposition \ref{prop2}.
% To complete the proof of Proposition \ref{prop2}, we need to show that the above estimate is also verified at $0$. For this purpose, we show that for all $\varepsilon>0$ there exists $\tilde{R}>0$ such that the following inequality holds
% \begin{align}\label{374}
% \frac{1}{\lambda_v^{2s_c}(\tau_0)}\left|\int\varphi_{\tilde{R}}|v(\tau_0)|^2\,dx-\int\varphi_{\tilde{R}}|v_0|^2\,dx\right|<\varepsilon,
% \end{align}
%with $\varphi_{\tilde{R}}$ as in \eqref{fi}. More precisely, we show that for all $\varepsilon>0$ there exists $C_{\varepsilon}>0$ such that if
% \begin{align}\label{Dep}
% D_\varepsilon=C_\varepsilon \max\left[F_*,F_*^{\frac{1+s_c}{1-s_c}}\right]\max\left[M_0^{\alpha_1},M_0^{\frac{2+\alpha_2}{2(1-s_c)}}\right],
% \end{align}
%and $D\geq D_\varepsilon$, then the inequality \eqref{374} holds for
% \begin{align}\label{Rtil}
% \widetilde{R}=\widetilde{R}(D,\tau_0)=D\lambda_v(\tau_0).
% \end{align}
%
%
% Assuming that \eqref{374} is true, from \eqref{364} and since $\|\varphi_R\|_{L^{\infty}}=1$, we have
% \begin{align}
% \frac{1}{\lambda_v^{2s_c}(\tau_0)}\int_{|x|\leq 2\tilde R} |v_0|^2\,dx> \frac{1}{\lambda_v^{2s_c}(\tau_0)}\int\varphi_{\tilde{R}}|v(\tau_0)|^2\,dx-\varepsilon\geq c_3>0,
% \end{align}
%for $\varepsilon>0$ small enough and $\widetilde{R}=D\lambda_v(\tau_0)\geq 2J_*\lambda_v(\tau_0)$. So, choosing $\alpha_3>0$ such that
%$$
%\frac{M_0^{\alpha_3}}{2}\max[1,F_*^{\frac{1+s_c}{1-s_c}}]\geq \max\{D_{\varepsilon},2J_*\},
%$$
%we deduce the desired inequality (\ref{tprop2}). Thus, it only remains to show \eqref{374}. Indeed, given $\varepsilon>0$, from \eqref{F}, \eqref{Dep} and \eqref{Rtil}, we get for $C_{\varepsilon}>1$
% \begin{align}
%\tilde{R}=D\lambda_v(\tau_0)\geq D_\varepsilon\frac{\lambda_v(\tau_0)}{\sqrt{\tau_0}}\sqrt{\tau_0}\geq \frac{D_\varepsilon}{F_*}\sqrt{\tau}\geq M_0^{\alpha_1}\sqrt{\tau}, \,\,\,\mbox{for all}\,\,\, \tau\in [0,\tau_0],
% \end{align}
% and thus, using \eqref{prop1ii} and the monotonicity of $\rho$ we have
% \begin{align}
%\rho(v(\tau),\tilde{R})\leq \rho(v(\tau),M_0^{\alpha_1}\sqrt{\tau})\leq cM_0^2, \,\,\,\mbox{for all}\,\,\, \tau\in [0,\tau_0].
% \end{align}
%Now, using the Cauchy-Schwarz inequality
% \begin{align}
% \left|\frac{d}{d\tau}\int \varphi_{\tilde{R}}|v(\tau)|^2\,dx\right|=&2\left|\textit{Im}\left(\int \nabla \varphi_{\tilde{R}}\cdot \nabla v(\tau)\overline{v}(\tau)\,dx\right)\right|\\
% &\leq \frac{c}{\tilde{R}}\|\nabla v(\tau)\|_{L^2}\left(\int_{\tilde{R}\leq |x|\leq 2\tilde{R}}|v(\tau)|^2\,dx\right)^{\frac{1}{2}}\\
% &\leq c\frac{(\rho(v(\tau),\tilde{R}))^{1/2}}{{\tilde{R}}^{1-s_c}}\|\nabla v(\tau)\|_{L^2}\\
% &\leq c\frac{M_0}{{\tilde{R}}^{1-s_c}}\|\nabla v(\tau)\|_{L^2}.
% \end{align}
% Next, we integrate from $0$ to $\tau_0\in[0,\frac{\tau_*}{2}]$ and use \eqref{prop1i} to obtain
% \begin{align}
% \frac{1}{\lambda_v^{2s_c}(\tau_0)}\left|\int \varphi_{\tilde{R}}|v(\tau_0)|^2\,dx-\int \varphi_{\tilde{R}}|v_0|^2\,dx
% \right|
% &\leq c\frac{D^{2s_c}}{\tilde{R}^{2s_c}}\frac{M_0}{\tilde{R}^{1-s_c}}\int_0^{\tau_0}\|\nabla v(\tau)\|_{L^2}\,d\tau\\
% &\leq c\frac{D^{2s_c}M_0}{\tilde{R}^{1+s_c}}\left(\int_0^{\tau_0}\tau_0\|\nabla v(\tau)\|_{L^2}^2\,d\tau\right)^{\frac{1}{2}}\\
% &\leq c\frac{D^{2s_c}M_0}{\tilde{R}^{1+s_c}}\left(\int_0^{\tau_0}(2\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau\right)^{\frac{1}{2}}\\
% &\leq c\frac{D^{2s_c}M_0}{\tilde{R}^{1+s_c}}\left(\int_0^{2\tau_0}(2\tau_0-\tau)\|\nabla v(\tau)\|_{L^2}^2\,d\tau\right)^{\frac{1}{2}}\\
% &\leq cM_0^{\frac{2+\alpha_2}{2}}D^{2s_c}\left(\frac{\tau_0}{\tilde{R}^2}\right)^{\frac{1+s_c}{2}}\\
% & =cM_0^{\frac{2+\alpha_2}{2}}\frac{F_*^{1+s_c}}{D^{1-s_c}},
% \end{align}
%where we have used definitions \eqref{F} and \eqref{Rtil} in the last step.
%
%Finally, since $D\geq D_{\varepsilon}\geq C_\varepsilon F_*^{\frac{1+s_c}{1-s_c}}M_0^{\frac{2+\alpha_2}{2(1-s_c)}}$ by \eqref{Dep}, taking $C_\varepsilon>1$ large enough we conclude \eqref{374} and complete the proof of Proposition \ref{prop2}.
\end{proof}
\subsection{Existence of blow-up solutions and lower bound for the blow-up rate}
We are now in position to prove our main results. We start with the existence of blow-up solutions.
\begin{proof}[Proof of Theorem \ref{scteo1}] Let $u\in C([0,+\infty):\dot H_a^{s_c}\cap \dot H_a^1 )$ be a global solution with $E[u_0]\leq 0$. In view of the Gagliardo-Nirenberg type inequality \eqref{GNacrit2} and $a>0$, we deduce that
\begin{align}
0\geq E[u_0]\geq \frac{1}{2}\|\mathcal L_a^{\frac12} u_0\|_{L^2}^2\left[1-\left(\frac{\|u_0\|_{L^{\sigma_c}}}{K_a}\right)^{2\sigma}\right].
\end{align}
and therefore the assumptions of Proposition \ref{prop1} are satisfied which implies
\begin{align}
\int_{0}^{\tau_*}(\tau_*-\tau)\|\mathcal L_a^{\frac12} u(\tau)\|_{L^2}^{2}\,d\tau\leq C_{u_0}\tau_*^{1+s_c},\,\, \mbox{for all} \,\, \tau_*>0.
\end{align}
Define $\lambda_u(\tau)=\|\mathcal L_a^{\frac12} u(\tau)\|^{-\frac{1}{1-s_c}}_{L^2}$. By previous inequality there exists a sequence $\tau_n\nearrow +\infty$ such that
\begin{align}\label{lambdatau}
\lambda_u(\tau_n)\geq C_{u_0}\sqrt{\tau_n},\,\,\,\mbox{for all}\,\,\, n\in \mathbb{N},
\end{align}
and in particular $\lambda_u(\tau_n)\nearrow +\infty$. Therefore, since $u_0\in \dot H_a^{s_c} \subset L^{\sigma_c}$, the relation \eqref{radial2} implies
\begin{align}\label{contrteo11}
\lim_{n\to +\infty}\frac{1}{(D\lambda_u(\tau_n))^{2s_c}}\int_{|x|\leq D\lambda_u(\tau_n)}|u_0|^2\,dx=0, \,\,\,\mbox{for all}\,\,\, D\geq 0.
\end{align}
On the other hand, we can apply Proposition \ref{prop2} at time $\tau_n$, for all $n\in \mathbb{N}$. Indeed, since $F_n = \frac{\sqrt{\tau_n}}{\lambda_u(\tau_n)} \leq C_{u_0}$, there exists $D_*$, independent of $n$, such that
\begin{align}\label{contrteo1}
\frac{1}{\lambda_u^{2s_c}(\tau_n)}\int_{|x|\leq D_*\lambda_u(\tau_n)}|u_0|^2\,dx\geq C_2>0,
\end{align}
reaching a contradiction with \eqref{contrteo11}. Therefore the solution cannot exist for all positive time completing the proof.
\end{proof}
%\subsection{Blow-up of solutions with non-negative energy}\label{sec5}
%In this section, we prove Theorem \ref{scteo1}, which follows as a direct consequence of Propositions \ref{prop1} and \ref{prop2}.%and Corollary \ref{corosc}.
%\begin{proof}[Proof of Theorem \ref{scteo1}] Assume by contradiction that there exists $u\in C([0,+\infty):\dot H_a^{s_c}\cap \dot H_a^1 )$ global solution to \eqref{PVI} with non-positive energy. From the Gagliardo-Nirenberg type inequality \eqref{GNine} we deduce that
% \begin{align}
% E[u_0]\geq \frac{1}{2}\|\nabla u_0\|_{L^2}^2\left[1-\left(\frac{\|u_0\|_{L^{\sigma_c}}}{\|V\|_{L^{\sigma_c}}}\right)^{2\sigma}\right].
% \end{align}
% Since $E[u_0]\leq 0$, the conditions \eqref{hpprop1i} and \eqref{hpprop1ii} are satisfied. Thus, for all $\tau_*>0$, we apply Proposition \ref{prop1} and inequality \eqref{prop1i} to deduce
% \begin{align}
% \int_{0}^{\tau_*}(\tau_*-\tau)\|\nabla u(\tau)\|_{L^2}^{2}\,d\tau\leq C_{u_0}\tau_*^{1+s_c}.
% \end{align}
% In particular, we obtain the existence of a sequence $\{\tau_n\}_{n=1}^{+\infty}$ such that $\tau_n\to +\infty$, as $n\to+\infty$, and
% \begin{align}
% \|\nabla u(\tau_n)\|_{L^2}\leq \frac{C_{u_0}}{\tau_n^{\frac{1-s_c}{2}}},\,\,\,\mbox{for all}\,\,\, n\in \mathbb{N}.
% \end{align}
%Therefore for $\lambda_u(\tau)=\|\nabla u(\tau)\|^{-\frac{1}{1-s_c}}_{L^2}$ we have
% \begin{align}\label{lambdatau}
% \lambda_u(\tau_n)=\left(\frac{1}{\|\nabla u(\tau_n)\|_{L^2}}\right)^{\frac{1}{1-s_c}}\geq C_{u_0}\sqrt{\tau_n}.
% \end{align}
% Again from the conservation of the energy, we have that the solution $u(\tau)\in \dot H_a^{s_c}\cap \dot H_a^1 $ satisfies conditions \eqref{hpprop1i}, \eqref{hpprop1ii} and \eqref{Ev} with $\tau_0=\tau_n$. Now, we apply Proposition \ref{prop2} at time $\tau_n$. Indeed, denoting $F_*=F_n = \frac{\sqrt{\tau_n}}{\lambda_u(\tau_n)}$, from \eqref{lambdatau}, $F_n \leq C_{u_0}$ for all $n\in \mathbb{N}$ and hence there exists $D_*$, independent of $n$, such that
% \begin{align}\label{contrteo1}
% \frac{1}{\lambda_u^{2s_c}(\tau_n)}\int_{|x|\leq D_*\lambda_u(\tau_n)}|u_0|^2\,dx\geq C_2>0,
% \end{align}
%in view of inequality \eqref{tprop2}.
%
% On the other hand, in view of \eqref{lambdatau}, $\lambda_u(\tau_n)\to+\infty$, as $n\to+\infty$. Moreover, since $u_0\in \dot H^{s_c} \subset L^{\sigma_c}$, we apply the relation \eqref{radial2} with $R=D_*\lambda_u(\tau_n)$, to obtain
% \begin{align}
% \lim_{n\to +\infty}\frac{1}{(D_*\lambda_u(\tau_n))^{2s_c}}\int_{|x|\leq D_*\lambda_u(\tau_n)}|u_0|^2\,dx=0,
% \end{align}
%which is a contradiction with \eqref{contrteo1}. Therefore, the maximal time of existence $T^{\ast}>0$ of the solution $u(t)$ is finite and the proof of Theorem \ref{scteo1} is complete.
%\end{proof}
%\subsection{A lower bound for the blow-up rate}\label{sec6}
%Here, we show that Theorem \ref{scteo2} is also a consequence of Propositions \ref{prop1} and \ref{prop2}. We start with a preliminary and simple result.
%\begin{lemma}\label{vteo2}
% Let $0<b<2$ and $\frac{2-b}{N}<\sigma<\min\left\{\frac{2-b}{N-2},\frac{2}{N}\right\}$.
% Consider $u_0\in\dot H^{s_c}\cap\dot H^{1} $ and assume that the maximal time of existence $T^{\ast}>0$ of the corresponding solution $u(t)$ to \eqref{PVI} is finite. Also assume that
%\begin{align}\label{esttl10}
%\lim_{t\to T^{\ast}}\left\|\nabla u(t)\right\|_{L^2}=\infty.
%\end{align}
%For $t$ close enough to $T^{\ast}$, consider the following renormalization of $u$
% \begin{align}\label{v}
% v^{(t)}(x,\tau)=\lambda_u^{\frac{2-b}{2\sigma}}(t)\bar{u}(\lambda_u(t)x,t-\lambda^2_u(t)\tau),
% \end{align}
% where $\lambda_u$ is given by
% \begin{align}\label{lambu}
% \lambda_u(t)=\left(\frac{1}{\|\nabla u(t)\|_{L^2}}\right)^{\frac{1}{1-s_c}}.
% \end{align}
% %By local well-posedness theory, we have
% %\begin{align}\label{esttl}
% %\lambda_u(t)\leq C(N,\sigma)\sqrt{T-t},
% %\end{align}
% %that given us $\lambda_u(t)\to 0$ as $t\to T$.
% Then, $v^{(t)}(\tau)\in C([0,\tau^{(t)}]:\dot H_a^{s_c}\cap \dot H_a^1)$ is a solution to \eqref{PVI} with $\tau^{(t)}=\frac{t}{\lambda_u^2(t)}$ and $v^{(t)}(\tau_0)$ satisfies the conditions of Propositions \ref{prop1} and \ref{prop2} for any $\tau_0$ such that
% \begin{align}
% \tau_0\in\left[0,\frac{1}{\lambda_u(t)}\right].
% \end{align}
%\end{lemma}
%\begin{proof}
% For $t>0$ fixed, to simplify the notation we write $v=v^{(t)}$. From scaling symmetry $u(x,t)\mapsto \lambda^{\frac{2-b}{2\sigma}}u(\lambda x, \lambda^2 t)$ it is easy to see that $v(\tau)$ is a solution to \eqref{PVI} on the time interval $[0, \frac{t}{\lambda_u^2(t)}]$. Moreover, from the assumption \eqref{esttl10} we get
% $$
% \frac{t}{\lambda_u^2(t)}\geq \frac{2}{\lambda_u(t)},
% $$
%for $t$ close enough to $T^{\ast}$. Set $\tau_\ast=\frac{2}{\lambda_u(t)}$, then $v(\tau)$ is a solution to \eqref{PVI} on the time interval $[0,\tau_\ast]$. Now, we need to check that \eqref{hpprop1i}, \eqref{hpprop1ii}, \eqref{hpprop2i} and \eqref{Ev} are satisfied. First note that if $\tau_0\in \left[0,\frac{1}{\lambda_u(t)}\right]=\left[0,\frac{\tau_*}{2}\right]$, then \eqref{hpprop2i} holds. Also, it is clear that
% \begin{align}\label{vt}
% \|\nabla v(0)\|_{L^2}=1\,\,\,\,\mbox{ and }\,\,\,\,\,E[v(0)]=\lambda_u(t)^{2(1-s_c)}E[u_0].
% \end{align}
% In particular,
% \begin{align}
% \tau_*^{1-s_c}|E[v(0)]|=\left(\tau_*\lambda_u^2(t)\right)^{1-s_c}|E[u_0]|=(2\lambda_u(t))^{1-s_c}|E[u_0]|\to 0,\,\,\,\mbox{ as }\,\,\,t\to T^{\ast}
% \end{align}
% and \eqref{hpprop1i} follows for $t$ close enough to $T^{\ast}$. In addition, from the Gagliardo-Nirenberg type inequality \eqref{GNine}, we have
% \begin{align}
% E[v(0)]\geq\frac{1}{2}\left(1-\left(\frac{\|v(0)\|_{L^{\sigma_c}}}{\|V\|_{L^{\sigma_c}}}\right)^{2\sigma}\right),
% \end{align}
%which implies \eqref{hpprop1ii}, for $t$ close enough to $T^{\ast}$, since $E[v(0)]\to 0$, as $t\to T^{\ast}$ by \eqref{vt}. Next, for all $\tau_0\in \left[0,\frac{1}{\lambda_u(t)}\right]$, we have
% \begin{align}
% \lambda_v(\tau_0)=\left(\frac{1}{\|\nabla v(\tau_0)\|_{L^2}}\right)^{\frac{1}{1-s_c}}=\left(\frac{1}{\lambda_u(t)^{1-s_c}\|\nabla u(t-\lambda_u^2(t)\tau_0)\|_{L^2}}\right)^{\frac{1}{1-s_c}}=\frac{\lambda_u(t-\lambda_u^2(t)\tau_0)}{\lambda_u(t)}.
% \end{align}
%Moreover
%$$
%t-\lambda^2_u(t)\tau_0\geq t-\lambda_u(t)\,\,\, \mbox{and} \,\,\,
%t-\lambda_u(t) \to T^{\ast}, \,\,\,\mbox{ as }\,\,\,t\to T^{\ast}.
%$$
%Combining these last two relation with \eqref{vt}, we deduce
% \begin{align}
% \lambda_v^{2(1-s_c)}(\tau_0)|E[v(0)]|=&(\lambda_v(\tau_0)\lambda_u(t))^{2(1-s_c)}|E[u_0]|\\
% =&(\lambda_u(t-\lambda_u^2(t)\tau_0))^{2(1-s_c)}|E[u_0]|\to 0, \,\,\,\mbox{ as }\,\,\,t\to T^{\ast},
% \end{align}
%and then \eqref{Ev} holds for $t$ close enough to $T^{\ast}$.
%\end{proof}
%%\begin{rem}
%%Note that the universal constants $C_1, \alpha_1$ and $\alpha_2$ obtained in Proposition \ref{prop1} applied to the rescaled solution $v^{(t)}$ given by \eqref{v} are independent of $t$, since $\|\nabla v^{(t)}(0)\|_{L^2}=1$, for all $t\in [0,T^{\ast})$.
%%\end{rem}
%
Next, we refine the analysis to show a lower bound for the blow-up rate of the critical Sobolev norm.
%
\begin{proof}[Proof of Theorem \ref{scteo2}]
Let $u_0\in \dot{H_a^{s_c}}\cap \dot H_a^1$ so that the maximal time of existence $T^{\ast}>0$ of the corresponding solution $u$ to \eqref{PVI} is finite. The inequality \eqref{BUA2} yields
\begin{align}\label{HYPLB}
\|\mathcal L_a^\frac12 u(t)\|_{L^2}\geq \frac{c}{(T^*-t)^{\frac{1-s_c}{2}}}, \,\,\, \textit{for all} \,\,\, t\in [0,T^*).
\end{align}
For $\sigma_c=\frac{6\sigma}{2-b}$, we shall prove that
\begin{align}\label{rate2}
\|u(t)\|_{L^{\sigma_c}}\geq |\log (T^{\ast}-t)|^{\alpha},\,\,\,as\,\,\,t\to T^{\ast},
\end{align}
for a universal constant $\alpha=\alpha(N,\sigma,b)>0$, which implies the same lower bound for the critical Sobolev norm $\dot{H}_a^{s_c}$. The proof follows the ideas introduced by Merle and Raph\"ael \cite{MR_Bsc} and is very similar to the proof of Theorem 1.2 in \cite{CF20}. Below we sketch the main steps. First, for $t$ close enough to $T^*$, define the renormalization of $u$ given by
\begin{align}\label{v}
v^{(t)}(x,\tau)=\lambda_u^{\frac{2-b}{2\sigma}}(t)\bar{u}(\lambda_u(t)x,t-\lambda^2_u(t)\tau),
\end{align}
where $\lambda_u(t)=\|\mathcal L_a^{\frac12} u(t)\|^{-\frac{1}{1-s_c}}_{L^2}$.
One can check that $v^{(t)}(\tau)\in C([0,\tau^{(t)}]:\dot H^{s_c}\cap \dot H^1)$ is a solution to \eqref{PVI} with $\tau^{(t)}={t}/{\lambda_u^2(t)}$ and also Propositions \ref{prop1} and \ref{prop2} can be applied to $v^{(t)}(\tau_0)$ for any $\tau_0\in\left[0,{1}/{\lambda_u(t)}\right]$ (see \cite[Lemma 6.1]{CF20}). From now on, to simplify the notation, we write $v=v^{(t)}$.
Define $N(t)=-\log \lambda_u(t)$. Since $\|u(t)\|_{L^{\sigma_c}}=\|v(0)\|_{L^{\sigma_c}}$, in view of inequality \eqref{HYPLB}, the problem is reduced to find $\alpha=\alpha(N,\sigma, b)>0$ such that, for $t$ close enough to $T^{\ast}$, we have
\begin{align}\label{N1}
\|v(0)\|_{L^{\sigma_c}}\geq [N(t)]^{\alpha}.
\end{align}
Recalling the constants $K_a$ and $\alpha_2$ given in \eqref{GNacrit2} and \eqref{prop1i}, respectively, we define
\begin{align}\label{Mt1}
L(t)=[100[M(t)]^{\alpha_2}]^{\frac{1}{2(1-s_c)}}, \,\,\,\mbox{where}\,\,\, M(t)=\frac{4\|v(0)\|_{L^{\sigma_c}}}{K_a}\geq 2.
\end{align}
If $L(t)\geq e^{\frac{\sqrt{N(t)}}{2}}$, then it is easy to see that \eqref{N1} holds with $\alpha=1$. On the other hand, if $L(t)<e^{\frac{\sqrt{N(t)}}{2}}$, we pick times $\tau_i\in[0,e^i]$, for all $i\in [\sqrt{N(t)},N(t)]$, such that
\begin{align}\label{Ftaui}
F(\tau_i)\leq L(t)\,\,\,\,\mbox{ and }\,\,\,\,\,\frac{1}{10L(t)}e^{\frac{i-1}{2}}\leq \lambda_v(\tau_i)\leq \frac{10}{L(t)}e^{\frac{i}{2}},
\end{align}
where $F(\tau)=\sqrt{\tau}\left/\right.\lambda_v(\tau)$ (see the proof of inequality (6.10) in \cite{CF20}). Moreover, for this choice of $\tau_i$, there exist universal positive constants $\alpha_4(N,\sigma, b)$ and $c_4(N,\sigma,b)$ such that
\begin{align}\label{teste}
\int_{\mathcal{C}_i}|v(0)|^{\sigma_c}\,dx\geq \frac{c_4}{[M(t)]^{\alpha_4s_c\sigma_c}}
\end{align}
where
\begin{align}\label{Rti1}
\mathcal{C}_i=\left\{x\in \Real^N;\, \frac{\lambda_v(\tau_i)}{[M(t)]^{\alpha_4}}\leq |x|\leq [M(t)]^{\alpha_4}\lambda_v(\tau_i)\right\},
\end{align}
(see the proof of inequality (6.12) in \cite{CF20}).
To finish the proof, let $p(t)>1$ an integer such that
\begin{align}\label{cotapt}
10^3[M(t)]^{2\alpha_4}\leq e^{\frac{p(t)}{2}}\leq 10^7[M(t)]^{2\alpha_4}.
\end{align}
If $p(t)\geq \sqrt{N(t)}$, then simple computations imply \eqref{N1} with $\alpha=1/4\alpha_4$. In the remaining case $p(t)<\sqrt{N(t)}$ consider $(i,i+p(t))\in [\sqrt{N(t)},N(t)]\times [\sqrt{N(t)},N(t)]$. From\eqref{Ftaui} and \eqref{cotapt} we get
\begin{align}
\lambda_v(\tau_{i+p(t)})&\geq \frac{e^{\frac{i+p(t)-1}{2}}}{10L(t)}
\geq\frac{10^3[M(t)]^{2\alpha_4}e^{\frac{i-1}{2}}}{10L(t)}\geq [M(t)]^{2\alpha_4}\frac{10e^{\frac{i}{2}}}{L(t)}\geq [M(t)]^{2\alpha_4}\lambda_v(\tau_i),
\end{align}
% \begin{align}
% \frac{\lambda_v(\tau_{i+p(t)})}{[M(t)]^{\alpha_4}}&\geq \frac{1}{[M(t)]^{\alpha_4}}\frac{e^{\frac{i+p(t)-1}{2}}}{10L(t)}
% \geq\frac{1000[M(t)]^{\alpha_4}e^{\frac{i-1}{2}}}{10L(t)}\geq [M(t)]^{\alpha_4}\frac{10e^{\frac{i}{2}}}{L(t)}\geq [M(t)]^{\alpha_4}\lambda_v(\tau_i),
% \end{align}
which implies that the annuli $\mathcal{C}_i$ and $\mathcal{C}_{i+p(t)}$ given by \eqref{Rti1} are disjoints. In this case, for $t$ close enough to $T^{\ast}$, there are at least $\frac{N(t)}{10p(t)}$ disjoint annuli, thus \eqref{teste} yields
\begin{align}
\|v(0)\|_{L^{\sigma_c}}^{\sigma_c}\geq \sum_{k=0}^{\frac{N(t)}{10p(t)}}\int_{\mathcal{C}_{kp(t)+\sqrt{N(t)}}}|v(0)|^{\sigma_c}\,dx\geq \frac{c_4}{[M(t)]^{\alpha_4\sigma_cs_c}}\frac{\sqrt{N(t)}}{10}=C\frac{\sqrt{N(t)}}{\|v(0)\|_{L^{\sigma_c}}^{\alpha_4\sigma_cs_c}},
\end{align}
which implies \eqref{N1}, finishing the proof of Theorem \ref{scteo2}.
\end{proof}
%
%Finally, we prove our last result, which asserts that even without the assumption \eqref{esttl} the $\dot H^{s_c}$ norm of finite time solutions in $\dot H_a^{s_c}\cap \dot H_a^1$ is never uniformly bounded in time.
%
%\begin{proof}[Proof of Corollary \ref{cor13}]
%Let $u\in C([0,T^{\ast}): \dot H_a^{s_c}\cap \dot H_a^1 )$ be a solution to \eqref{PVI} with finite maximal time existence $T^{\ast}>0$ such that
% \begin{align}\label{Msup}
% \sup_{t\in [0,T^{\ast})}\|u(t)\|_{\dot H^{s_c}}=M<+\infty,
% \end{align}
%Then for any $t\in [0,T^{\ast})$, we consider the following scaling of $u$
% \begin{align}\label{scaling}
% v^{(t)}(x,\tau)=\rho^{\frac{2-b}{2\sigma}}(t)u(\rho(t)x,t+\rho^2(t)\tau),
% \end{align}
% where $\rho(t)^{1-s_c}\|\nabla u(t)\|_{L^2}=1$. Hence,
%$$
%\|v^{(t)}(0)\|_{\dot H^{s_c}}=\|u(t)\|_{\dot H^{s_c}}\leq M\,\,\,\mbox{ and }\,\,\,\,\|\nabla v^{(t)}(0)\|_{L^2}=\rho(t)^{1-s_c}\|\nabla u(t)\|_{L^2}=1
%$$
%and thus, from the local Cauchy theory in $\dot H^{s_c}\cap\dot H^1 $ (see \cite[Theorem 1.2.]{CFG20}), there exists $\tau_0$, independent of $t,$ such that $v^{(t)}$ is defined on $[0,\tau_0]$. Therefore, $t+\rho^2(t)\tau_0<T^{\ast}$ and this is \eqref{esttl}. From Theorem \ref{scteo2}, the lower bound \eqref{rate} holds and, in particular, we have that
%$$
%\lim_{t\uparrow T^{\ast}}\|u(t)\|_{\dot H^{s_c}}=+\infty,
%$$
%which is a contradiction with \eqref{Msup}.
%\end{proof}
%\section{Critical norm concentrion}
\vspace{0.5cm}
\noindent
\textbf{Acknowledgments.} L.C. was partially supported by Coordena\c{c}\~ao de Aperfei\c{c}oamento de Pessoal de N\'ivel Superior - CAPES. M.C. was partially supported by Coordena\c{c}\~ao de Aperfei\c{c}oamento de Pessoal de N\'ivel Superior - CAPES and Funda\c{c}\~ao de Amparo a Pesquisa do Estado do Piau\'i - FAPEPI. L.G.F. was partially supported by Coordena\c{c}\~ao de Aperfei\c{c}oamento de Pessoal de N\'ivel Superior - CAPES, Conselho Nacional de Desenvolvimento Cient\'ifico e Tecnol\'ogico - CNPq and Funda\c{c}\~ao de Amparo a Pesquisa do Estado de Minas Gerais - FAPEMIG. %C.G. was partially supported by ???.
%\appendix
%\section{Cut-off function}\label{appB}
%In this section, we construct the non-negative radially symmetric function $\phi\in C^{\infty}_0(\Real^N)$ such that
%\begin{align}\label{app1}
%\phi(x)=
%\left\{
%\begin{array}{ll}
%\frac{|x|^2}{2},&\mbox{ se }|x|\leq 2\\
%0,&\mbox{ se }|x|\geq 4
%\end{array}
%\right.,
%\quad\quad\partial^2_r\phi(x)\leq 1
%\end{align}
%and there exists $c>0$ satisfying
%\begin{align}\label{app2}
%&\phi(x)\leq c|x|^2\quad\quad\mbox{ and }\quad\quad|\nabla \phi (x)|^2\leq c\phi(x).
%\end{align}
%Choose $\eta\in C_0^{\infty}(\Real^N)$ be a function radially symmetric so that $0\leq\eta\leq 1,$ $\eta'\leq 0$ and
%\begin{align}
%\eta(x)=
%\left\{
%\begin{array}{ll}
%1,&\mbox{ if }|x|\leq 2\\
%0,&\mbox{ if }|x|\geq 4
%\end{array}
%\right..
%\end{align}
%Making $\partial_r^2\phi(x):=\eta(x)$ with $\phi(0)=0,$ $\phi(2)=2$ and $\phi(x)=0$ if $|x|\geq4$, we get that $\phi$ satisfies the conditions in \eqref{app1}. Clearly $\phi$ is a bounded function, and thus, there exists $c>0$ such that $\phi(x)\leq c|x|^2$. Therefore, it is enough to show that the second condition in \eqref{app2} holds. Indeed, if $|x|< 2$ or $|x|> 4$ it is immediate. Suppose that for every $n\in \mathbb N$ there exists
%$$x_n\in D:=\{x\in \Real^N;\,\,2\leq|x|\leq 4\}, \quad |x_n|=r_n$$ such that
%$$|\phi'(r_n)|=|\nabla \phi(x_n)|>n\phi^{\frac{1}{2}}(x_n)=n\phi^{\frac{1}{2}}(r_n),$$
%that is,
%$$\left|\partial_r\phi^{\frac{1}{2}}(r_n)\right|> n.$$
%Since $\partial_r\phi$ is uniformly continuous in the compact $\overline{D}$, then there exists $\delta>0$ such that
%\begin{align}
%\left|\partial_r\phi^{\frac{1}{2}}(y)\right|>\left|\partial_r\phi^{\frac{1}{2}}(x)\right|-\frac{n}{2}\quad\mbox{ wherenever }\quad|x-y|<\delta,
%\end{align}
%and thus,
%\begin{align}
%\left|\partial_r\phi^{\frac{1}{2}}(x)\right|\geq \frac{n}{2},\quad\quad\forall x\in (r_n-\delta,r_n+\delta).
%\end{align}
%Let $a_n<b_n$ such that $a_n,b_n\in(r_n-\delta,r_n+\delta)$, then integrating from $a_n$ to $b_n$
%\begin{align}
%\phi^{\frac{1}{2}}(b_n)\leq -\frac{n}{2}(b_n-a_n)&+\phi^{\frac{1}{2}}(a_n)<-n\delta+\phi^{\frac{1}{2}}(a_n)\\
%&\mbox{or}\\
%\phi^{\frac{1}{2}}(b_n)\geq \frac{n}{2}(b_n&-a_n)+\phi^{\frac{1}{2}}(a_n)> n\delta,
%\end{align}
%which is a contradiction, since $\phi$ is a bounded function.
%\bibliography{biblio2}
%\bibliographystyle{abbrvnat}
%%\usepackage[abbrev]{amsrefs}
\begin{thebibliography}{47}
\providecommand{\natexlab}[1]{#1}
\providecommand{\url}[1]{\texttt{#1}}
\expandafter\ifx\csname urlstyle\endcsname\relax
\providecommand{\doi}[1]{doi: #1}\else
\providecommand{\doi}{doi: \begingroup \urlstyle{rm}\Url}\fi
\bibitem[Aloui and Tayachi(2021)]{AT21}
L.~Aloui and S.~Tayachi.
\newblock Local well-posedness for the inhomogeneous nonlinear
{S}chr\"{o}dinger equation.
\newblock \emph{Discrete Contin. Dyn. Syst.}, 41\penalty0 (11):\penalty0
5409--5437, 2021.
\newblock ISSN 1078-0947.
\newblock \doi{10.3934/dcds.2021082}.
\newblock URL \url{https://doi.org/10.3934/dcds.2021082}.
\bibitem[An and Kim(2021)]{AK21}
J.~An and J.~Kim.
\newblock Local well-posedness for the inhomogeneous nonlinear
{S}chr\"{o}dinger equation in {$H^s(\mathbb R^n)$}.
\newblock \emph{Nonlinear Anal. Real World Appl.}, 59:\penalty0 Paper No.
103268, 21, 2021.
\newblock ISSN 1468-1218.
\newblock \doi{10.1016/j.nonrwa.2020.103268}.
\newblock URL \url{https://doi.org/10.1016/j.nonrwa.2020.103268}.
\bibitem[An et~al.(2023)An, Jang, and Kim]{AJK_DCDS_22}
J.~An, R.~Jang, and J.~Kim.
\newblock Global existence and blow-up for the focusing inhomogeneous nonlinear
{S}chr\"{o}dinger equation with inverse-square potential.
\newblock \emph{Discrete Contin. Dyn. Syst. Ser. B}, 28\penalty0 (2):\penalty0
1046--1067, 2023.
\newblock ISSN 1531-3492.
\newblock \doi{10.3934/dcdsb.2022111}.
\newblock URL \url{https://doi.org/10.3934/dcdsb.2022111}.
\bibitem[Belmonte-Beitia et~al.(2007)Belmonte-Beitia, P{\'e}rez-Garc{\'\i}a,
Vekslerchik, and Torres]{belmonte2007lie}
J.~Belmonte-Beitia, V.~M. P{\'e}rez-Garc{\'\i}a, V.~Vekslerchik, and P.~J.
Torres.
\newblock Lie symmetries and solitons in nonlinear systems with spatially
inhomogeneous nonlinearities.
\newblock \emph{Physical review letters}, 98\penalty0 (6):\penalty0 064102,
2007.
\bibitem[Bourgain(1999)]{Bo99}
J.~Bourgain.
\newblock \emph{Global solutions of nonlinear {S}chr\"{o}dinger equations},
volume~46 of \emph{American Mathematical Society Colloquium Publications}.
\newblock American Mathematical Society, Providence, RI, 1999.
\newblock ISBN 0-8218-1919-4.
\newblock \doi{10.1090/coll/046}.
\newblock URL \url{https://doi.org/10.1090/coll/046}.
\bibitem[Burq et~al.(2003{\natexlab{a}})Burq, Planchon, Stalker, and
Tahvildar-Zadeh]{BPST03}
N.~Burq, F.~Planchon, J.~G. Stalker, and A.~S. Tahvildar-Zadeh.
\newblock Strichartz estimates for the wave and {S}chr\"{o}dinger equations
with the inverse-square potential.
\newblock \emph{J. Funct. Anal.}, 203\penalty0 (2):\penalty0 519--549,
2003{\natexlab{a}}.
\newblock ISSN 0022-1236.
\newblock \doi{10.1016/S0022-1236(03)00238-6}.
\newblock URL \url{https://doi.org/10.1016/S0022-1236(03)00238-6}.
\bibitem[Burq et~al.(2003{\natexlab{b}})Burq, Planchon, Stalker, and
Tahvildar-Zadeh]{Burq}
N.~Burq, F.~Planchon, J.~G. Stalker, and A.~S. Tahvildar-Zadeh.
\newblock Strichartz estimates for the wave and {S}chr\"{o}dinger equations
with the inverse-square potential.
\newblock \emph{J. Funct. Anal.}, 203\penalty0 (2):\penalty0 519--549,
2003{\natexlab{b}}.
\newblock ISSN 0022-1236.
\newblock \doi{10.1016/S0022-1236(03)00238-6}.
\newblock URL \url{https://doi.org/10.1016/S0022-1236(03)00238-6}.
\bibitem[Caffarelli et~al.(1984)Caffarelli, Kohn, and Nirenberg]{CKN84}
L.~Caffarelli, R.~Kohn, and L.~Nirenberg.
\newblock First order interpolation inequalities with weights.
\newblock \emph{Compositio Math.}, 53\penalty0 (3):\penalty0 259--275, 1984.
\newblock ISSN 0010-437X.
\newblock URL \url{http://www.numdam.org/item?id=CM_1984__53_3_259_0}.
\bibitem[Campos(2021)]{Campos}
L.~Campos.
\newblock Scattering of radial solutions to the inhomogeneous nonlinear
{S}chr\"{o}dinger equation.
\newblock \emph{Nonlinear Anal.}, 202:\penalty0 1--17, 2021.
\newblock ISSN 0362-546X.
\newblock \doi{10.1016/j.na.2020.112118}.
\newblock URL \url{https://doi.org/10.1016/j.na.2020.112118}.
\bibitem[Campos and Cardoso(2021)]{CC_Going}
L.~Campos and M.~Cardoso.
\newblock Blowup and scattering criteria above the threshold for the focusing
inhomogeneous nonlinear {S}chr\"{o}dinger equation.
\newblock \emph{NoDEA Nonlinear Differential Equations Appl.}, 28\penalty0
(6):\penalty0 Paper No. 69, 33, 2021.
\newblock ISSN 1021-9722.
\newblock \doi{10.1007/s00030-021-00725-4}.
\newblock URL \url{https://doi.org/10.1007/s00030-021-00725-4}.
\bibitem[Campos and Guzm\'{a}n(2021{\natexlab{a}})]{CG_Zamp_21}
L.~Campos and C.~M. Guzm\'{a}n.
\newblock On the inhomogeneous {NLS} with inverse-square potential.
\newblock \emph{Z. Angew. Math. Phys.}, 72\penalty0 (4):\penalty0 Paper No.
143, 29, 2021{\natexlab{a}}.
\newblock ISSN 0044-2275.
\newblock \doi{10.1007/s00033-021-01560-4}.
\newblock URL \url{https://doi.org/10.1007/s00033-021-01560-4}.
\bibitem[Campos and Guzm\'{a}n(2021{\natexlab{b}})]{LUCCASCARLOS}
L.~Campos and C.~M. Guzm\'{a}n.
\newblock On the inhomogeneous {NLS} with inverse-square potential.
\newblock \emph{Z. Angew. Math. Phys.}, 72\penalty0 (4):\penalty0 Paper No.
143, 2021{\natexlab{b}}.
\newblock ISSN 0044-2275.
\newblock \doi{10.1007/s00033-021-01560-4}.
\newblock URL \url{https://doi.org/10.1007/s00033-021-01560-4}.
\bibitem[Campos et~al.(2022)Campos, Correia, and Farah]{CCF_Arxiv_22}
L.~Campos, S.~Correia, and L.~G. Farah.
\newblock Some well-posedness and ill-posedness results for the inls equation.
\newblock \emph{arXiv preprint arXiv:2210.07060}, 2022.
\bibitem[Cardoso and Farah(2021)]{CF20}
M.~Cardoso and L.~G. Farah.
\newblock Blow-up of radial solutions for the intercritical inhomogeneous {NLS}
equation.
\newblock \emph{J. Funct. Anal.}, 281\penalty0 (8):\penalty0 Paper No. 109134,
38, 2021.
\newblock ISSN 0022-1236.
\newblock \doi{10.1016/j.jfa.2021.109134}.
\newblock URL \url{https://doi.org/10.1016/j.jfa.2021.109134}.
\bibitem[Cardoso and Farah(2023)]{CF23}
M.~Cardoso and L.~G. Farah.
\newblock Blow-up solutions of the intercritical inhomogeneous {NLS} equation:
the non-radial case.
\newblock \emph{Math. Z.}, 303\penalty0 (3):\penalty0 Paper No. 63, 18, 2023.
\newblock ISSN 0025-5874.
\newblock \doi{10.1007/s00209-023-03212-x}.
\newblock URL \url{https://doi.org/10.1007/s00209-023-03212-x}.
\bibitem[Cardoso et~al.(2021)Cardoso, Farah, and Guzm{\'a}n]{CFG20}
M.~Cardoso, L.~G. Farah, and C.~M. Guzm{\'a}n.
\newblock On well-posedness and concentration of blow-up solutions for the
intercritical inhomogeneous {NLS} equation.
\newblock \emph{Journal of Dynamics and Differential Equations}, 2021.
\newblock \doi{10.1007/s10884-021-10045-x}.
\newblock URL \url{https://doi.org/10.1007/s10884-021-10045-x4}.
\bibitem[Cazenave(2003)]{cazenave}
T.~Cazenave.
\newblock \emph{Semilinear {S}chr\"{o}dinger equations}, volume~10 of
\emph{Courant Lecture Notes in Mathematics}.
\newblock New York University, Courant Institute of Mathematical Sciences, New
York; American Mathematical Society, Providence, RI, 2003.
\newblock ISBN 0-8218-3399-5.
\newblock \doi{10.1090/cln/010}.
\newblock URL \url{https://doi.org/10.1090/cln/010}.
\bibitem[Cho et~al.(2020)Cho, Hong, and Lee]{CHL20}
Y.~Cho, S.~Hong, and K.~Lee.
\newblock On the global well-posedness of focusing energy-critical
inhomogeneous {NLS}.
\newblock \emph{J. Evol. Equ.}, 20\penalty0 (4):\penalty0 1349--1380, 2020.
\newblock ISSN 1424-3199.
\newblock \doi{10.1007/s00028-020-00558-1}.
\newblock URL \url{https://doi.org/10.1007/s00028-020-00558-1}.
\bibitem[Christ and Weinstein(1991)]{CW91}
F.~M. Christ and M.~I. Weinstein.
\newblock Dispersion of small amplitude solutions of the generalized
{K}orteweg-de {V}ries equation.
\newblock \emph{J. Funct. Anal.}, 100\penalty0 (1):\penalty0 87--109, 1991.
\newblock ISSN 0022-1236.
\newblock \doi{10.1016/0022-1236(91)90103-C}.
\newblock URL \url{https://doi.org/10.1016/0022-1236(91)90103-C}.
\bibitem[Dinh and Keraani(2021)]{DK21}
V.~D. Dinh and S.~Keraani.
\newblock Long time dynamics of nonradial solutions to inhomogeneous nonlinear
{S}chr\"{o}dinger equations.
\newblock \emph{SIAM J. Math. Anal.}, 53\penalty0 (4):\penalty0 4765--4811,
2021.
\newblock ISSN 0036-1410.
\newblock \doi{10.1137/20M1383434}.
\newblock URL \url{https://doi.org/10.1137/20M1383434}.
\bibitem[Duyckaerts and Roudenko(2015)]{DR_Going}
T.~Duyckaerts and S.~Roudenko.
\newblock Going beyond the threshold: scattering and blow-up in the focusing
{NLS} equation.
\newblock \emph{Comm. Math. Phys.}, 334\penalty0 (3):\penalty0 1573--1615,
2015.
\newblock ISSN 0010-3616.
\newblock \doi{10.1007/s00220-014-2202-y}.
\newblock URL \url{https://doi.org/10.1007/s00220-014-2202-y}.
\bibitem[Fanelli et~al.(2013)Fanelli, Felli, Fontelos, and Primo]{FFFP13}
L.~Fanelli, V.~Felli, M.~A. Fontelos, and A.~Primo.
\newblock Time decay of scaling critical electromagnetic {S}chr\"{o}dinger
flows.
\newblock \emph{Comm. Math. Phys.}, 324\penalty0 (3):\penalty0 1033--1067,
2013.
\newblock ISSN 0010-3616.
\newblock \doi{10.1007/s00220-013-1830-y}.
\newblock URL \url{https://doi.org/10.1007/s00220-013-1830-y}.
\bibitem[Farah(2016)]{Farah}
L.~G. Farah.
\newblock Global well-posedness and blow-up on the energy space for the
inhomogeneous nonlinear {S}chr\"{o}dinger equation.
\newblock \emph{J. Evol. Equ.}, 16\penalty0 (1):\penalty0 193--208, 2016.
\newblock ISSN 1424-3199.
\newblock \doi{10.1007/s00028-015-0298-y}.
\newblock URL \url{https://doi.org/10.1007/s00028-015-0298-y}.
\bibitem[Farah and Guzm\'an(2017)]{paper2}
L.~G. Farah and C.~M. Guzm\'an.
\newblock Scattering for the radial 3{D} cubic focusing inhomogeneous nonlinear
{S}chr\"odinger equation.
\newblock \emph{J. Differential Equations}, 262\penalty0 (8):\penalty0
4175--4231, 2017.
\newblock ISSN 0022-0396.
\newblock \doi{10.1016/j.jde.2017.01.013}.
\newblock URL \url{http://dx.doi.org/10.1016/j.jde.2017.01.013}.
\bibitem[Fibich(2015)]{Fi15}
G.~Fibich.
\newblock \emph{The nonlinear {S}chr\"{o}dinger equation}, volume 192 of
\emph{Applied Mathematical Sciences}.
\newblock Springer, Cham, 2015.
\newblock ISBN 978-3-319-12747-7; 978-3-319-12748-4.
\newblock \doi{10.1007/978-3-319-12748-4}.
\newblock URL \url{https://doi.org/10.1007/978-3-319-12748-4}.
\newblock Singular solutions and optical collapse.
\bibitem[Foschi(2005)]{Foschi05}
D.~Foschi.
\newblock Inhomogeneous {S}trichartz estimates.
\newblock \emph{J. Hyperbolic Differ. Equ.}, 2\penalty0 (1):\penalty0 1--24,
2005.
\newblock ISSN 0219-8916.
\newblock \doi{10.1142/S0219891605000361}.
\newblock URL \url{https://doi.org/10.1142/S0219891605000361}.
\bibitem[Genoud and Stuart(2008)]{g_8}
F.~Genoud and C.~A. Stuart.
\newblock Schr\"{o}dinger equations with a spatially decaying nonlinearity:
existence and stability of standing waves.
\newblock \emph{Discrete Contin. Dyn. Syst.}, 21\penalty0 (1):\penalty0
137--186, 2008.
\newblock ISSN 1078-0947.
\newblock \doi{10.3934/dcds.2008.21.137}.
\newblock URL \url{https://doi.org/10.3934/dcds.2008.21.137}.
\bibitem[Guzm\'{a}n(2017)]{CARLOS}
C.~M. Guzm\'{a}n.
\newblock On well posedness for the inhomogeneous nonlinear {S}chr\"{o}dinger
equation.
\newblock \emph{Nonlinear Anal. Real World Appl.}, 37:\penalty0 249--286, 2017.
\newblock ISSN 1468-1218.
\newblock \doi{10.1016/j.nonrwa.2017.02.018}.
\newblock URL \url{https://doi.org/10.1016/j.nonrwa.2017.02.018}.
\bibitem[Kalf et~al.(1975)Kalf, Schmincke, Walter, and W\"{u}st]{MR0397192}
H.~Kalf, U.-W. Schmincke, J.~Walter, and R.~W\"{u}st.
\newblock On the spectral theory of {S}chr\"{o}dinger and {D}irac operators
with strongly singular potentials.
\newblock In \emph{Spectral theory and differential equations}, pages 182--226.
Lecture Notes in Math., Vol. 448, 1975.
\bibitem[Kartashov et~al.(2017)Kartashov, Malomed, Vysloukh, Belic, and
Torner]{MALOMED1}
Y.~V. Kartashov, B.~A. Malomed, V.~A. Vysloukh, M.~R. Belic, and L.~Torner.
\newblock Rotating vortex clusters in media with inhomogeneous defocusing
nonlinearity.
\newblock \emph{Optic Letters.}, 42\penalty0 (3):\penalty0 446--449, 2017.
\bibitem[Kato(1994)]{Kato94}
T.~Kato.
\newblock An {$L^{q,r}$}-theory for nonlinear {S}chr\"{o}dinger equations.
\newblock In \emph{Spectral and scattering theory and applications}, volume~23
of \emph{Adv. Stud. Pure Math.}, pages 223--238. Math. Soc. Japan, Tokyo,
1994.
\bibitem[Keel and Tao(1998)]{KT98}
M.~Keel and T.~Tao.
\newblock Endpoint {S}trichartz estimates.
\newblock \emph{Amer. J. Math.}, 120\penalty0 (5):\penalty0 955--980, 1998.
\newblock ISSN 0002-9327.
\newblock URL
\url{http://muse.jhu.edu/journals/american_journal_of_mathematics/v120/120.5keel.pdf}.
\bibitem[Kenig et~al.(1993)Kenig, Ponce, and Vega]{KPV93}
C.~E. Kenig, G.~Ponce, and L.~Vega.
\newblock Well-posedness and scattering results for the generalized
{K}orteweg-de {V}ries equation via the contraction principle.
\newblock \emph{Comm. Pure Appl. Math.}, 46\penalty0 (4):\penalty0 527--620,
1993.
\newblock ISSN 0010-3640.
\newblock \doi{10.1002/cpa.3160460405}.
\newblock URL \url{https://doi.org/10.1002/cpa.3160460405}.
\bibitem[Killip et~al.(2017)Killip, Murphy, Visan, and Zheng]{Murphy}
R.~Killip, J.~Murphy, M.~Visan, and J.~Zheng.
\newblock The focusing cubic {NLS} with inverse-square potential in three space
dimensions.
\newblock \emph{Differential Integral Equations}, 30\penalty0 (3-4):\penalty0
161--206, 2017.
\newblock ISSN 0893-4983.
\newblock URL \url{http://projecteuclid.org/euclid.die/1487386822}.
\bibitem[Killip et~al.(2018)Killip, Miao, Visan, Zhang, and Zheng]{KMVZZ18}
R.~Killip, C.~Miao, M.~Visan, J.~Zhang, and J.~Zheng.
\newblock Sobolev spaces adapted to the {S}chr\"{o}dinger operator with
inverse-square potential.
\newblock \emph{Math. Z.}, 288\penalty0 (3-4):\penalty0 1273--1298, 2018.
\newblock ISSN 0025-5874.
\newblock \doi{10.1007/s00209-017-1934-8}.
\newblock URL \url{https://doi.org/10.1007/s00209-017-1934-8}.
\bibitem[Kim et~al.(2021)Kim, Lee, and Seo]{KLS21}
J.~Kim, Y.~Lee, and I.~Seo.
\newblock On well-posedness for the inhomogeneous nonlinear {S}chr\"{o}dinger
equation in the critical case.
\newblock \emph{J. Differential Equations}, 280:\penalty0 179--202, 2021.
\newblock ISSN 0022-0396.
\newblock \doi{10.1016/j.jde.2021.01.023}.
\newblock URL \url{https://doi.org/10.1016/j.jde.2021.01.023}.
\bibitem[Lee and Seo(2021)]{LS21}
Y.~Lee and I.~Seo.
\newblock The {C}auchy problem for the energy-critical inhomogeneous nonlinear
{S}chr\"{o}dinger equation.
\newblock \emph{Arch. Math. (Basel)}, 117\penalty0 (4):\penalty0 441--453,
2021.
\newblock ISSN 0003-889X.
\newblock \doi{10.1007/s00013-021-01632-x}.
\newblock URL \url{https://doi.org/10.1007/s00013-021-01632-x}.
\bibitem[Linares and Ponce(2015)]{LiPo15}
F.~Linares and G.~Ponce.
\newblock \emph{Introduction to nonlinear dispersive equations}.
\newblock Universitext. Springer, New York, second edition, 2015.
\newblock ISBN 978-1-4939-2180-5; 978-1-4939-2181-2.
\newblock \doi{10.1007/978-1-4939-2181-2}.
\newblock URL \url{https://doi.org/10.1007/978-1-4939-2181-2}.
\bibitem[Lu et~al.(2018)Lu, Miao, and Murphy]{Murphy1}
J.~Lu, C.~Miao, and J.~Murphy.
\newblock Scattering in {$H^1$} for the intercritical {NLS} with an
inverse-square potential.
\newblock \emph{J. Differential Equations}, 264\penalty0 (5):\penalty0
3174--3211, 2018.
\newblock ISSN 0022-0396.
\newblock \doi{10.1016/j.jde.2017.11.015}.
\newblock URL \url{https://doi.org/10.1016/j.jde.2017.11.015}.
\bibitem[Merle and Rapha\"{e}l(2008)]{MR_Bsc}
F.~Merle and P.~Rapha\"{e}l.
\newblock Blow up of the critical norm for some radial {$L^2$} super critical
nonlinear {S}chr\"{o}dinger equations.
\newblock \emph{Amer. J. Math.}, 130\penalty0 (4):\penalty0 945--978, 2008.
\newblock ISSN 0002-9327.
\newblock \doi{10.1353/ajm.0.0012}.
\newblock URL \url{https://doi.org/10.1353/ajm.0.0012}.
\bibitem[Miao et~al.(2021)Miao, Murphy, and Zheng]{MMZ21}
C.~Miao, J.~Murphy, and J.~Zheng.
\newblock Scattering for the non-radial inhomogeneous {NLS}.
\newblock \emph{Math. Res. Lett.}, 28\penalty0 (5):\penalty0 1481--1504, 2021.
\newblock ISSN 1073-2780.
\newblock \doi{10.4310/mrl.2021.v28.n5.a9}.
\newblock URL \url{https://doi.org/10.4310/mrl.2021.v28.n5.a9}.
\bibitem[Murphy(2022)]{M_Proc22}
J.~Murphy.
\newblock A simple proof of scattering for the intercritical inhomogeneous
{NLS}.
\newblock \emph{Proc. Amer. Math. Soc.}, 150\penalty0 (3):\penalty0 1177--1186,
2022.
\newblock ISSN 0002-9939.
\newblock \doi{10.1090/proc/15717}.
\newblock URL \url{https://doi.org/10.1090/proc/15717}.
\bibitem[Okazawa et~al.(2012)Okazawa, Suzuki, and Yokota]{Okazawa2012}
N.~Okazawa, T.~Suzuki, and T.~Yokota.
\newblock Energy methods for abstract nonlinear {S}chr\"{o}dinger equations.
\newblock \emph{Evol. Equ. Control Theory}, 1\penalty0 (2):\penalty0 337--354,
2012.
\newblock ISSN 2163-2472.
\newblock \doi{10.3934/eect.2012.1.337}.
\newblock URL \url{https://doi.org/10.3934/eect.2012.1.337}.
\bibitem[Sulem and Sulem(1999)]{Sulem}
C.~Sulem and P.-L. Sulem.
\newblock \emph{The nonlinear {S}chr\"{o}dinger equation}, volume 139 of
\emph{Applied Mathematical Sciences}.
\newblock Springer-Verlag, New York, 1999.
\newblock ISBN 0-387-98611-1.
\newblock Self-focusing and wave collapse.
\bibitem[Tao(2006)]{TaoBook}
T.~Tao.
\newblock \emph{Nonlinear dispersive equations: local and global analysis}.
\newblock Number 106. American Mathematical Soc., 2006.
\bibitem[Zhang and Zheng(2014{\natexlab{a}})]{ZZ14}
J.~Zhang and J.~Zheng.
\newblock Scattering theory for nonlinear {S}chr\"{o}dinger equations with
inverse-square potential.
\newblock \emph{J. Funct. Anal.}, 267\penalty0 (8):\penalty0 2907--2932,
2014{\natexlab{a}}.
\newblock ISSN 0022-1236.
\newblock \doi{10.1016/j.jfa.2014.08.012}.
\newblock URL \url{https://doi.org/10.1016/j.jfa.2014.08.012}.
\bibitem[Zhang and Zheng(2014{\natexlab{b}})]{Zhang}
J.~Zhang and J.~Zheng.
\newblock Scattering theory for nonlinear {S}chr\"{o}dinger equations with
inverse-square potential.
\newblock \emph{J. Funct. Anal.}, 267\penalty0 (8):\penalty0 2907--2932,
2014{\natexlab{b}}.
\newblock ISSN 0022-1236.
\newblock \doi{10.1016/j.jfa.2014.08.012}.
\newblock URL \url{https://doi.org/10.1016/j.jfa.2014.08.012}.
\end{thebibliography}
\newcommand{\Addresses}{{% additional braces for segregating \footnotesize
\bigskip
\footnotesize
LUCCAS CAMPOS, \textsc{Department of Mathematics, UFMG, Brazil}\par\nopagebreak
\textit{E-mail address:} \texttt{luccas@ufmg.br}
\medskip
MYKAEL A. CARDOSO, \textsc{Department of Mathematics, UFPI, Brazil}\par\nopagebreak
\textit{E-mail address:} \texttt{mykael@ufpi.edu.br}
\medskip
LUIZ G. FARAH, \textsc{Department of Mathematics, UFMG, Brazil}\par\nopagebreak
\textit{E-mail address:} \texttt{farah@mat.ufmg.br}
}}
\setlength{\parskip}{0pt}
\Addresses
\end{document}