Tail probability of maximal displacement in critical and subcritical branching stable processes
Extracted source text ยท arXiv 2504.05540
\documentclass[11pt]{article}
\usepackage{tikz}
\usepackage{tikz-qtree}
\usepackage[page, header]{appendix}
\usepackage{titletoc}
\usepackage{amsmath, amsfonts, amssymb, amsthm, amsbsy, amscd, bm, bbm,mathrsfs}
\usepackage{color}
\usepackage[colorlinks,linkcolor=blue,citecolor=blue]{hyperref}
\usepackage[top=1.5cm, bottom=1.5cm, outer=3cm, inner=3cm, heightrounded, marginparwidth=3cm, marginparsep=3cm]{geometry}
\newcommand\myshade{70}
\usepackage{comment}
\usepackage{mathtools}
\mathtoolsset{showonlyrefs}
\newtheorem{assumption}{Assumption}
%% Operators
\def\iid{\textrm{i.i.d.~}}
\def\wp{\textit{w.p.~}}
\def\st{\mathop{\hbox{subject to}}}
%% Graphs
\newcommand{\Var}{\operatorname{\textrm{Var}}}
\newcommand{\goto}{\rightarrow}
\newcommand{\sgn}{\mathop\mathrm{sign}}
\newcommand{\KL}{\operatorname{KL}}
\newcommand{\law}{\operatorname{Law}}
\newcommand{\loc}{{\operatorname{loc}}}
\newcommand{\semiloc}{{\operatorname{semiloc}}}
\newcommand{\orth}{\operatorname{orth}}
\newcommand{\pa}{\operatorname{pa}}
\newcommand{\lip}{\operatorname{lip}}
\newcommand{\op}{\operatorname{op}}
\newcommand{\e}{\mathrm{e}}
\renewcommand\d{\operatorname{\mathrm{d}}}
\newcommand{\deq}{\overset{\mathrm{d}}{=}}
\newcommand{\rad}{\operatorname{Rad}_n}
\newcommand{\erad}{\widehat{\operatorname{Rad}}}
% \newcommand{\deq}
% Specific commands
\renewcommand{\AA}{\mathbb{A}}
\newcommand{\BB}{\mathbb{B}}
\newcommand{\CC}{\mathbb{C}}
\newcommand{\DD}{\mathbb{D}}
\newcommand{\EE}{\operatorname{\mathbb{E}}}
\newcommand{\FF}{\mathbb{F}}
\newcommand{\GG}{\mathbb{G}}
\newcommand{\HH}{\mathbb{H}}
\newcommand{\II}{\mathbb{I}}
\newcommand{\JJ}{\mathbb{J}}
\newcommand{\KK}{\mathbb{K}}
\newcommand{\LL}{\mathbb{L}}
\newcommand{\MM}{\mathbb{M}}
\newcommand{\NN}{\mathbb{N}}
\newcommand{\OO}{\mathbb{O}}
\newcommand{\PP}{\mathbb{P}}
\newcommand{\QQ}{\mathbb{Q}}
\newcommand{\RR}{\mathbb{R}}
\renewcommand{\SS}{\mathbb{S}}
\newcommand{\TT}{\mathbb{T}}
\newcommand{\UU}{\mathbb{U}}
\newcommand{\VV}{\mathbb{V}}
\newcommand{\WW}{\mathbb{W}}
\newcommand{\XX}{\mathbb{X}}
\newcommand{\YY}{\mathbb{Y}}
\newcommand{\ZZ}{\mathbb{Z}}
\newcommand{\fbf}[1]{\mathbf{#1}}
\newcommand{\bA}{\fbf{A}}
\newcommand{\bB}{\fbf{B}}
\newcommand{\bC}{\fbf{C}}
\newcommand{\bD}{\fbf{D}}
\newcommand{\bE}{\fbf{E}}
\newcommand{\bF}{\fbf{F}}
\newcommand{\bG}{\fbf{G}}
\newcommand{\bH}{\fbf{H}}
\newcommand{\bI}{\fbf{I}}
\newcommand{\bJ}{\fbf{J}}
\newcommand{\bK}{\fbf{K}}
\newcommand{\bL}{\fbf{L}}
\newcommand{\bM}{\fbf{M}}
\newcommand{\bN}{\fbf{N}}
\newcommand{\bO}{\fbf{Q}}
\newcommand{\bP}{\fbf{P}}
\newcommand{\bQ}{\fbf{Q}}
\newcommand{\bR}{\fbf{R}}
\newcommand{\bS}{\fbf{S}}
\newcommand{\bT}{\fbf{T}}
\newcommand{\bU}{\fbf{U}}
\newcommand{\bV}{\fbf{V}}
\newcommand{\bW}{\fbf{W}}
\newcommand{\bX}{\fbf{X}}
\newcommand{\bY}{\fbf{Y}}
\newcommand{\bZ}{\fbf{Z}}
\newcommand{\ba}{\fbf{a}}
\newcommand{\bb}{\fbf{b}}
\newcommand{\bc}{\fbf{c}}
\newcommand{\bd}{\fbf{d}}
\newcommand{\be}{\mathrm{\fbf{e}}}
\newcommand{\bmf}{\fbf{f}}
\newcommand{\bg}{\fbf{g}}
\newcommand{\bh}{\bar{h}}
\newcommand{\bi}{\fbf{i}}
\newcommand{\bj}{\fbf{j}}
\newcommand{\bl}{\fbf{l}}
\newcommand{\bmm}{\fbf{m}}
\newcommand{\bk}{\fbf{k}}
\newcommand{\bo}{\fbf{o}}
\newcommand{\bp}{\fbf{p}}
\newcommand{\bq}{\fbf{q}}
\newcommand{\br}{\fbf{r}}
\newcommand{\bs}{\fbf{s}}
\newcommand{\bt}{\fbf{t}}
\newcommand{\bu}{\fbf{u}}
\newcommand{\bv}{\fbf{v}}
\newcommand{\bw}{\fbf{w}}
\newcommand{\bx}{\fbf{x}}
\newcommand{\by}{\fbf{y}}
\newcommand{\bz}{\fbf{z}}
\renewcommand{\leq}{\leqslant}
\renewcommand{\geq}{\geqslant}
\newcommand{\balpha}{\bm{\alpha}}
\newcommand{\bbeta}{\bm{\beta}}
\newcommand{\bgamma}{\bm{\gamma}}
\newcommand{\bomega}{\bm{\omega}}
\newcommand{\bpsi}{\bm{\psi}}
\newcommand{\bxi}{\bm{\xi}}
\newcommand{\bpi}{\bm{\pi}}
\newcommand{\bepsilon}{\bm{\epsilon}}
\newcommand{\bvarepsilon}{\bm{\varepsilon}}
\newcommand{\cA}{\mathcal{A}}
\newcommand{\cB}{\mathcal{B}}
\newcommand{\cC}{\mathcal{C}}
\newcommand{\cD}{\mathcal{D}}
\newcommand{\cE}{\mathcal{E}}
\newcommand{\cF}{\mathcal{F}}
\newcommand{\cG}{\mathcal{G}}
\newcommand{\cH}{\mathcal{H}}
\newcommand{\cI}{\mathcal{I}}
\newcommand{\cJ}{\mathcal{J}}
\newcommand{\cK}{\mathcal{K}}
\newcommand{\cL}{\mathcal{L}}
\newcommand{\cM}{\mathcal{M}}
\newcommand{\cN}{\mathcal{N}}
\newcommand{\cO}{\mathcal{O}}
\newcommand{\cP}{\mathcal{P}}
\newcommand{\cR}{\mathcal{R}}
\newcommand{\cS}{\mathcal{S}}
\newcommand{\cT}{\mathcal{T}}
\newcommand{\cW}{\mathcal{W}}
\newcommand{\cU}{\mathcal{U}}
\newcommand{\cV}{\mathcal{V}}
\newcommand{\cX}{\mathcal{X}}
\newcommand{\cY}{\mathcal{Y}}
\newcommand{\cOmega}{\mathcal{\Omega}}
\newcommand{\ha}{\hat{a}}
\newcommand{\hb}{\hat{b}}
\newcommand{\hc}{\hat{c}}
\newcommand{\hd}{\hat{d}}
\newcommand{\he}{\hat{e}}
\newcommand{\hf}{\hat{f}}
\newcommand{\hg}{\hat{g}}
\newcommand{\hh}{\hat{h}}
\newcommand{\hi}{\hat{i}}
\newcommand{\hj}{\hat{j}}
\newcommand{\hk}{\hat{k}}
\newcommand{\hhl}{\hat{l}}
\newcommand{\hhm}{\hat{m}}
\newcommand{\hn}{\hat{n}}
\newcommand{\ho}{\hat{o}}
\newcommand{\hp}{\hat{p}}
\newcommand{\hq}{\hat{q}}
\newcommand{\hr}{\hat{r}}
\newcommand{\hs}{\hat{s}}
\newcommand{\hht}{\hat{t}}
\newcommand{\hu}{\hat{u}}
\newcommand{\hv}{\hat{v}}
\newcommand{\hw}{\hat{w}}
\newcommand{\hx}{\hat{x}}
\newcommand{\hy}{\hat{y}}
\newcommand{\hz}{\hat{z}}
\newcommand{\hA}{\hat{A}}
\newcommand{\hB}{\hat{B}}
\newcommand{\hC}{\hat{C}}
\newcommand{\hD}{\hat{D}}
\newcommand{\hE}{\hat{E}}
\newcommand{\hF}{\hat{F}}
\newcommand{\hG}{\hat{G}}
\newcommand{\hH}{\hat{H}}
\newcommand{\hI}{\hat{I}}
\newcommand{\hJ}{\hat{J}}
\newcommand{\hK}{\hat{K}}
\newcommand{\hL}{\hat{L}}
\newcommand{\hM}{\hat{M}}
\newcommand{\hN}{\hat{N}}
\newcommand{\hO}{\hat{O}}
\newcommand{\hP}{\hat{P}}
\newcommand{\hQ}{\hat{Q}}
\newcommand{\hR}{\hat{R}}
\newcommand{\hS}{\hat{S}}
\newcommand{\hT}{\hat{T}}
\newcommand{\hU}{\hat{U}}
\newcommand{\hV}{\hat{V}}
\newcommand{\hW}{\hat{W}}
\newcommand{\hX}{\hat{X}}
\newcommand{\hY}{\hat{Y}}
\newcommand{\hZ}{\hat{Z}}
\newcommand{\htheta}{\hat{\theta}}
\newcommand{\hrho}{\hat{\rho}}
\newcommand{\transpose}{^{\operatorname{T}}}
\newcommand{\step}{\texttt{step}}
\newcommand{\relu}{\texttt{ReLU}}
\newcommand{\<}{\left\langle}
\renewcommand{\>}{\right\rangle}
\newcommand{\tri}{\mathrm{tri}}
\newcommand{\dd}{\mathop{}\!\mathrm{d}}
\newcommand{\risk}{\mathcal{R}}
\renewcommand{\ge}{\mathcal{E}}
\newcommand{\erisk}{\hat{\mathcal{R}}}
\newcommand{\wrt}{with respect to }
\newcommand{\fn}{\frac{1}{n}}
\newcommand{\fm}{\frac{1}{m}}
\newcommand{\fM}{\frac{1}{M}}
\newcommand{\sumin}{\sum_{i=1}^n}
\newcommand{\sumim}{\sum_{i=1}^m}
\newcommand{\sumjm}{\sum_{j=1}^m}
\newcommand{\sumiM}{\sum_{i=1}^M}
\newcommand{\sumjM}{\sum_{j=1}^M}
\newcommand{\sumtm}{\sum_{t=1}^m}
\newcommand{\barj}{\bar{j}}
\newcommand{\hl}{\cH_{\ell}}
\newcommand{\ttheta}{\tilde{\theta}}
\newcommand{\half}{\frac{1}{2}}
\newcommand{\de}[2]{\frac{{\rm d}#1}{{\rm d}#2}}
\newcommand{\dde}[2]{\frac{{\rm d}^2#1}{{\rm d}#2^2}}
\newcommand{\ddde}[3]{\frac{{\rm d}^2#1}{{\rm d}#2{\rm d}#3}}
\newcommand{\pd}[2]{\frac{\partial#1}{\partial#2}}
\newcommand{\ppd}[2]{\frac{\partial^2#1}{\partial#2^2}}
\newcommand{\pppd}[3]{\frac{\partial^2#1}{\partial#2\partial#3}}
% trace, rank, diag, domain
\newcommand{\tr}{{\rm tr}}
\newcommand{\rank}{{\rm rank}}
\newcommand{\diag}[1]{{\rm diag}\left(#1\right)}
\newcommand{\dom}{{\rm dom}}
% abs, norm, inner product
\newcommand{\abs}[1]{\left\vert#1\right\vert}
\newcommand{\norm}[2]{\left\Vert#1\right\Vert_{#2}}
\newcommand{\inn}[1]{\left\langle#1\right\rangle}
\newcommand{\cnn}{\mathrm{cnn}}
\newcommand{\lcn}{\mathrm{lcn}}
\newcommand{\fcn}{\mathrm{fcn}}
% +-infinity
\newcommand{\pif}{{+\infty}}
\newcommand{\nif}{{-\infty}}
% column vector 1
\newcommand{\one}{{\bf 1}}
% left/right hand side
\newcommand{\lhs}{\textsc{l.h.s.}}
\newcommand{\rhs}{\textsc{r.h.s.}}
% transpose
\newcommand{\tp}{^{\mathrm{T}}}
\newcommand{\ntp}{^{-\mathrm{T}}}
% projection
\newcommand{\proj}[2]{\mathrm{proj}_{#1}\left(#2\right)}
% X, x, Y, y
\newcommand{\Xbf}{\mathbf{X}}
\newcommand{\xbf}{\mathbf{x}}
\newcommand{\ybf}{\mathbf{y}}
% epsilon
\newcommand{\ep}{{\epsilon}}
\newtheorem{theorem}{Theorem}
\newtheorem{corollary}{Corollary}
\newtheorem{case}{Case}
\newtheorem{lemma}{Lemma}
\newtheorem{proposition}{Proposition}
\theoremstyle{definition}
\newtheorem{definition}{Definition}
\newtheorem{example}{Example}
\newtheorem{remark}{Remark}
\providecommand{\keywords}[1]
{
\small
\textbf{\textit{Keywords---}} #1
}
\providecommand{\MSC}[2]
{
\small
\textbf{\textit{2020 MSC---}} #1
}
\title{Tail probability of maximal displacement in critical and subcritical branching stable processes}
\author{ \bf Haojie Hou\footnote{The research of this author is supported by the China Postdoctoral Science Foundation (No. 2024M764112).}
\hspace{1mm}\hspace{1mm}
Yiyang Jiang \hspace{1mm}\hspace{1mm}
Yan-Xia Ren\footnote{The research of this author is supported by NSFC (Grant No 12231002) and The Fundamental
Research Funds for the Central Universities, Peking University LMEQF.
} \hspace{1mm}\hspace{1mm} and \hspace{1mm}\hspace{1mm}
Renming Song\thanks{Research supported in part by a grant from the Simons
Foundation
(\#960480, Renming Song).}
\hspace{1mm} }
\begin{document}
\maketitle
\begin{abstract}
In this paper, we study critical and subcritical branching $\alpha$-stable processes, $\alpha \in (0, 2)$. We obtain the exact asymptotic behaviors of the tails of the maximal positions of
all subcritical branching $\alpha$-stable processes with positive jumps.
In the case of subcritical branching spectrally negative $\alpha$-stable processes, we obtain the exact asymptotic behaviors of the tails of the maximal positions under the assumption that the offspring distributions satisfy the $L\log L$ condition.
For critical branching $\alpha$-stable processes, we obtain the exact asymptotic behaviors of the tails under the assumption that the offspring distributions belong to the domain of attraction of a $\gamma$-distribution, $\gamma\in (1, 2]$.
\end{abstract}\hspace{10pt}
\keywords{Branching stable process; rightmost position; integral equation}
\MSC{60J80; 60G52; 60G51; 60G70.}
\section{Introduction and main results}
\subsection{Introduction}
A branching L\'evy process is a continuous-time Markov process defined as follows. At time 0, there is a particle at $x\in \mathbb{R}$ and it moves according to a L\'evy process $(\xi_t, \mathbf{P}_x)$. After an exponential time with parameter $1$, independent of the motion, it dies and produces $k$ offspring with probability $p_k$, $k\geq 0$. The offspring move independently according to $\xi$ from the place where they are born and obey the same branching mechanism as their parent. This procedure goes on. For $t\geq 0$, let $N_t$ be the collection of particles alive at time $t$. For $u\in N_t$, we use $X_u(t)$ to denote the position of the particle $u$ at time $t$. The point process $(Z_t)_{t\geq 0}$ defined by
$$
Z_t:=\sum_{u\in N_t}\delta_{X_u(t)}, \qquad t \geq 0,
$$
is called a branching L\'evy process. We will denote the law of $(Z_t)_{t\geq 0}$ by $\mathbb{P}_x$ and write $\mathbb{P}$ for $\mathbb{P}_0$ for simplicity.
When $\xi$ is a Brownian motion, $(Z_t)_{t\geq 0}$ is called a branching Brownian motion. When $\xi$ is an $\alpha$-stable process, $\alpha\in (0, 2)$, $(Z_t)_{t\geq 0}$ is called a branching $\alpha$-stable process.
Let $m:=\sum^\infty_{k=0}kp_k$ be the mean of the offspring distribution. When $m>1(=1, <1)$, we say that
the branching L\'evy process $(Z_t)_{t\geq 0}$ is supercritical (critical, subcritical). It is well known that in the critical and subcritical cases, $(Z_t)_{t\geq 0}$ will die out in finite time with probability 1. Thus in this case we can define the maximal position of $(Z_t)_{t\geq 0}$ by
$$
M:=\sup_{t>0}\sup_{u\in N_t} X_u(t).
$$
When $m=1$ and $\xi$ is a standard Brownian motion, Sawyer and Fleischman \cite{Sawyer} proved that, under the assumption $\sum_{k=0}^\infty k^3 p_k<\infty$,
\begin{align}\label{Tail-probability1}
\lim_{x\to +\infty} x^2 \mathbb{P}\left(M\geq x\right) =\frac{6}{\sigma^2 },
\end{align}
where $\sigma^2$ is the variance of the offspring distribution. Profeta \cite{profeta:hal-03736526} extended the result above to some
critical branching spectrally negative L\'evy processes under the same third moment condition on the offspring distribution. When $m=1$, $\xi$ is an $\alpha$-stable process with positive jumps and $\sum_{k=0}^\infty k^3 p_k<\infty$, Lalley and Shao \cite{lalley}, and Profeta \cite{profeta}
proved that
\begin{align} \label{result1}
\lim_{x\to +\infty} x^{\alpha/2}\mathbb{P}\left(M\geq x\right) =c_1(\alpha),
\end{align}
where $c_1(\alpha)$ is an explicit positive constant. For critical branching spectrally negative $\alpha$-stable processes
with
$\alpha\in (1, 2)$, Profeta \cite[Corollary 1.3]{profeta:hal-03736526} proved that,
when $\sum_{k=0}^\infty k^3 p_k<\infty$,
\begin{align}\label{Profeta-bound}
0< \liminf_{x\to\infty} x^{\alpha} \mathbb{P}\left(M\geq x\right)\leq \limsup_{x\to\infty} x^{\alpha} \mathbb{P}\left(M \geq x\right)<\infty.
\end{align}
When $m<1$ and $\xi$ is a standard Brownian motion, Sawyer and Fleischman \cite{Sawyer} proved that, under the assumption $\sum_{k=0}^\infty k^3 p_k<\infty$, there exists a function $c(\cdot)$ bounded between two positive constants such that
\begin{align}\label{Tail-probability2}
\lim_{x\to +\infty} \frac{\mathbb{P}\left(M\geq x\right)}{(1-m)c(x)e^{-\sqrt{2(1-m)}x}}=1.
\end{align}
For subcritical branching $\alpha$-stable processes with positive jumps,
Profeta \cite{profeta} proved that,
under the assumption $\sum_{k=0}^\infty k^3 p_k<\infty$,
\begin{align}\label{result2}
\lim_{x\to +\infty} x^{\alpha}\mathbb{P}\left(M\geq x\right)=
\frac{c_2(\alpha)}{1-m}
\end{align}
for some explicit constant $c_2(\alpha)$.
For subcritical branching spectrally negative processes,
Profeta \cite[Theorem 1]{profeta:hal-03736526}
proved that, under the assumption $\sum_{k=0}^\infty k^3 p_k<\infty$,
\begin{align}\label{Profeta-limit}
\lim_{x\to\infty} e^{c_4x} \mathbb{P}\left(M\geq x\right) = c_3
\end{align}
for some explicit positive constants $c_3, c_4$.
For results on the asymptotic behaviors of the tails of maximal positions of (sub)critical branching random walks, see \cite{lalleywalk, neuman2017maximal}.
The purpose of this paper is to
establish the exact asymptotic behaviors of the tails of the maximal positions of (sub)critical branching $\alpha$-stable processes under minimal conditions on the offspring distributions. More precisely, we will obtain (i) the exact asymptotic behaviors of the tails of the maximal positions of subcritical branching $\alpha$-stable processes with positive jumps without any extra assumption on the offspring distributions; (ii) the exact asymptotic behaviors of the tail of the maximal positions of subcritical branching spectrally negative $\alpha$-stable processes under the assumption that the offspring distributions satisfy the $L\log L$ condition; (iii) the exact asymptotic behaviors of the tails of critical branching $\alpha$-stable processes under the assumption that the offspring distributions belong to the domain of attraction of a $\gamma$-distribution, $\gamma\in (1, 2]$.
\subsection{Main results}
Before we state our main results, we recall some useful facts about stable processes.
In this paper, we always assume that the spatial motion $\xi$ is a (strictly) $\alpha$-stable process, $\alpha\in (0, 2)$. For basic information on $\alpha$-stable process, see, for instance, \cite[Chap. VIII]{bertoin}. We assume that the L\'{e}vy measure of $\xi$ is given by
\begin{equation}\label{def-v}
v_\alpha(\mathrm{d}x):=c_+x^{-(1+\alpha)}{\bf 1}_{(0,\infty)}(x){\rm d} x+c_-|x|^{-(1+\alpha)}{\bf 1}_{(-\infty,0)}(x){\rm d} x,
\end{equation}
where $c_+$ and $c_-$ are non-negative numbers with at least one of them being positive.
Let
$\Psi(\theta):=-\ln \mathbf{E}(e^{\mathrm{i}\theta \xi_{1} })$
be the characteristic exponent of $\xi$. It is known (see, for instance \cite[Chapter 1.2.6]{kyprianou2014fluctuations}) that,
$$
\Psi(\theta)=
\begin{cases}
c_*|\theta|^\alpha\left(1-\mathrm{i} \beta \tan \frac{\pi \alpha}{2} \operatorname{sgn} \theta\right),
& \text { for } \alpha \in(0,1) \cup(1,2) \text { and } \beta \in[-1,1],\\
c_*|\theta|+\mathrm{i} \theta \eta, & \text { for } \alpha=1
\end{cases}
$$
where $ \eta \in \mathbb{R}$, $c_*:=-\left(c_++c_-\right) \Gamma(-\alpha) \cos (\pi \alpha / 2)$, and $\beta=\left(c_+-c_-\right) /\left(c_++c_-\right)$ if $\alpha \in(0,1) \cup(1,2)$. Here $\operatorname{sgn} \theta=1_{(\theta>0)}-1_{(\theta<0)}$.
Note that, for (strictly) 1-stable process, $c_+=c_-$
(see, for example, \cite[Theorem 14.7 (v)]{Sato}).
It follows from \cite[Example 1.1, Lemma 2.1 and the last sentence in Subsection 1.1]{Ren_Song_Zhang_2023} (see also \cite[Proposition VIII.4]{bertoin}) that, when $c_+>0$,
\begin{align}\label{Tail-of-xi}
\lim_{x\to\infty} x^{\alpha} \mathbf{P}_0(\xi_1\geq x) = \nu_\alpha ((1,\infty))= \frac{c_+}{\alpha}.
\end{align}
When $c_+=0$ and $\alpha\in (0,1)$, $-\xi$ is a subordinator and $M=0$ almost surely.
A spectrally negative 1-stable process reduces to the drift $-\eta t$. When $\eta\geq 0$, $M=0$.
Thus in the case of branching spectrally negative $\alpha$-stable processes,
exclude the above cases and assume
either $\alpha\in (1,2)$, or $\alpha=1$ and $\eta<0$.
In this case, it is easy to see that
for any $\lambda>0$,
\begin{align}\label{Def-of-Upsilon}
\mathbf{E}_0(e^{\lambda \xi_1})=
\exp\left\{C_1(\alpha) \lambda^\alpha \right\},
\quad \mbox{where}\quad C_1(\alpha)=
-\Psi(-\mathrm{i}).
\end{align}
Define $\tau_{y} = \inf\{t>0, \xi_t\geq y\}$. For any $x<y$, combining \eqref{Def-of-Upsilon} with the fact that
$\mathbf{P}_x\left(\xi_{\tau_y}=y\right)=1$,
we have that for any $\lambda>0$,
\begin{align}\label{Laplace-of-stopping-time}
\mathbf{E}_x\left(e^{-\lambda \tau_y} \right) = \exp\left\{-
C_1(\alpha)^{-1/\alpha}\lambda^{1/\alpha}
(y-x)\right\}.
\end{align}
In the critical case, we will need the following assumption on the offspring distribution:
\begin{itemize}
\item [{\bf(H)}]
The offspring distribution $\{p_k: k\geq 0\}$ belongs to the domain of attraction of a $\gamma$-stable, $\gamma\in (1,2]$, distribution.
More precisely, either there exist $\gamma\in (1,2)$ and $\kappa_{\gamma}\in(0,\infty)$ such that
\[
\lim_{n\to\infty} n^\gamma \sum_{k=n}^\infty p_k = \kappa_{\gamma},
\]
or that (corresponding $\gamma=2$)
\[
\sum_{k=0}^\infty k^2 p_k<\infty.
\]
\end{itemize}
Recall that $\sigma^2$ is the variance of the offspring distribution. We define
\begin{align}\label{Def-of-Lambda-gamma}
C_2(\gamma): =\begin{cases}
\frac{\Gamma(2-\gamma)}{\gamma-1}\kappa_{\gamma} & \text { for } \gamma \in(1,2);\\
\frac{1}{2}\sigma^2& \text { for } \gamma=2.
\end{cases}
\end{align}
Our first main result is on (sub)critical branching stable processes with positive jumps.
\begin{theorem}\label{t1}
Suppose that $c_+>0$.
i) If $m<1$, then
\begin{align}
\lim_{x\to +\infty} x^{\alpha}\mathbb{P}\left(M\geq x\right) =\frac{c_+}{(1-m)\alpha}.
\end{align}
ii) If $m=1$ and {\bf(H)} holds, then
$$
\lim_{x\rightarrow+\infty}x^{\alpha/\gamma} \mathbb{P}\left(M \geq x\right)=\left(\frac{c_+}{\alpha C_2(\gamma)}\right)^{\frac{1}{\gamma}},
$$
where $C_2(\gamma)$ is defined in \eqref{Def-of-Lambda-gamma}.
\end{theorem}
\begin{remark}
Theorem \ref{t1} (ii) is also contained in Theorem 3 of the recent preprint \cite{Profeta2025}.
Since there are some differences between the proofs of Theorem \ref{t1} (ii) and \cite[Theorem 3]{Profeta2025}, we include it here for
completeness.
The proof of Theorem \ref{t1} is an adaptation of that of the corresponding result in \cite{profeta}
below. {\bf(H)} only changes the behaviors of $\Phi_0$ and $\Phi_R$, defined in \eqref{main}.
In \cite[(1.5)]{profeta}, the third moment condition on the offspring distribution is used to estimate $\Phi_R$ directly while in the $\gamma$-stable branching case, we have to do a more careful analysis.
\end{remark}
Our second main result is
on the case for critical branching spectrally negative $\alpha$-stable process with $\alpha\in (1,2)$, and it generalizes and refines \eqref{Profeta-bound}.
\begin{theorem}\label{t2}
Suppose $c_+=0$ and $ \alpha\in (1,2)$
or $\alpha=1$ and $\eta <0$.
If $m=1$ and {\bf(H)} holds,
then there exists a constant $C_3(\alpha,\beta,\gamma)\in (0,\infty)$ such that
\begin{align}
\lim_{x\to\infty} x^{\frac{\alpha}{\gamma-1}} \mathbb{P}\left(M \geq x\right) = C_3(\alpha, \beta, \gamma).
\end{align}
\end{theorem}
\begin{remark}
The constant $C_3(\alpha,\beta,\gamma)$ has a probabilistic representation via a super $\alpha$-stable process, see the proof of Proposition \ref{prop1}.
In the proof of Proposition \ref{prop1}, we use the fact that a superprocess is an appropriate scaling limit of branching Markov processes.
\end{remark}
Our last main result is on subcritical branching spectrally negative $\alpha$-stable process with $\alpha\in (1,2)$.
\begin{theorem}\label{t3}
Suppose that $c_+=0$, and that either $\alpha\in (1,2) $
or $\alpha=1$ and $\eta<0$.
If $m<1$ and $\sum_{k=0}^\infty k(\log k)p_k<\infty$,
then there exists a positive constant $C_4(\alpha)$ such that
\begin{align}
\lim_{x\to\infty} e^{((1-m)/C_1(\alpha))^{1/\alpha}x} \mathbb{P}\left(M\geq x\right) = C_4(\alpha).
\end{align}
\end{theorem}
\begin{remark}
Our proof of Theorem \ref{t3} can be easily adapted to a widely class of subcritical branching spectrally negative L\'{e}vy processes.
\end{remark}
\section{An integral equation for $\mathbb{P}\left(M\geq x\right)$}
Let
\begin{align}
u(x):=\mathbb{P}\left(M \geq x\right)\quad \mbox{and}\quad S_t:=\sup _{s \in[0, t]} \xi_s
\end{align}
be the tail probability of $M$ and the supremum of $\xi$ up to time $t$. Let $\mathbf{e}$ be an exponential random variable with parameter $1$ independent of $\xi$.
Define
\begin{align}\label{Def-of-G}
G(x):=
\sum_{k=0}^\infty p_k(1-x)^k - 1+mx,\quad x\in [0,1].
\end{align}
It is easy to see that $G$ is a
non-negative function in $[0,1]$.
\begin{proposition}\label{Proposition-inte}
The function $u$ is a solution of the integral
equation:
\begin{equation}\label{integral equation}
u(x)=\mathbf{P}_0\left(S_{\mathbf{e}} \geq x\right)+\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]-\Phi_0(x)-\Phi_R(x),
\end{equation}
where
\begin{equation}\label{main}
\Phi_0(x)=(1-m)\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]\quad \mbox{and}\quad \Phi_R(x)=\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} G(u\left(x-\xi_{\mathbf{e}}\right))\right].
\end{equation}
(i) If $m=1$ and {\bf(H)} holds,
then for any $\varepsilon>0$, there exists $\delta>0$ such that
\begin{equation}\label{remainder1.1}
\Phi_R(x) \geq (1-\varepsilon)C_2(\gamma)\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma}\left(x-\xi_{\mathbf{e}}\right)\right]-\frac{ (1-\varepsilon)C_2(\gamma)}{\delta}\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma+1}\left(x-\xi_{\mathbf{e}}\right)\right]
\end{equation}
and
\begin{equation}\label{remainder1.2}
\Phi_R(x)\leq (1+\varepsilon)C_2(\gamma)\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma}\left(x-\xi_{\mathbf{e}}\right)\right]+\frac{1}{\delta^{\gamma+1}}\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma+1}\left(x-\xi_{\mathbf{e}}\right)\right],
\end{equation}
where $C_2(\gamma)$ is defined in \eqref{Def-of-Lambda-gamma}.
(ii) If $m<1$ and $c_+>0$, then
\begin{equation}\label{remainder2}
\lim_{x\to\infty}\frac{\Phi_R(x)}{\Phi_0(x)}=0.
\end{equation}
\end{proposition}
\begin{proof}
Applying the Markov property at the first branching time, we get
$$
\mathbb{P}\left(M<x\right)=p_0 \mathbf{P}\left(S_{\mathbf{e}}<x\right)+\sum_{n=1}^{+\infty} p_n \mathbb{P}\left(S_{\mathbf{e}}<x, \xi_{\mathbf{e}}+M^{(1)}<x, \ldots, \xi_{\mathbf{e}}+M^{(n)}<x\right),
$$
where $\left(M^{(n)}\right)_{n \in \mathbb{N}}$ are independent copies of $M$, which are also independent of $\left(\xi_{\mathbf{e}}, S_{\mathbf{e}}\right)$. Hence,
\begin{align*}
1-u(x)&=p_0 \mathbf{P}_0\left(S_{\mathbf{e}}<x\right)+\sum_{n=1}^{+\infty} p_n \mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}}\left(1-u\left(x-\xi_{\mathbf{e}}\right)\right)^n\right] \\
&=p_0 \mathbf{P}_0\left(S_{\mathbf{e}}<x\right)-m\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]\\
&\quad+ \mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}}\left(\sum_{n=1}^{+\infty} p_n\left(1-u\left(x-\xi_{\mathbf{e}}\right)\right)^n+mu\left(x-\xi_{\mathbf{e}}\right)\right)\right].
\end{align*}
Therefore,
\begin{align*}
u(x)=&\mathbf{P}_0\left(S_{\mathbf{e}} \geq x\right)+\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]-(1-m)\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]\\
&-\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}}\left(\sum_{n=1}^{+\infty} p_n\left(1-u\left(x-\xi_{\mathbf{e}}\right)\right)^n+mu\left(x-\xi_{\mathbf{e}}\right)-(1-p_0)\right)\right]\\
=&\mathbf{P}_0\left(S_{\mathbf{e}} \geq x\right)+\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]-\Phi_0(x)-\Phi_R(x),
\end{align*}
where $\Phi_0$ and $\Phi_R$ are given in \eqref{main}.
We prove (i) first.
When $m=1$ and {\bf(H)} holds,
from \cite[Lemma 3.1]{HJRS2025} (for $\gamma\in (1, 2)$) and L'Hopital's rule (for $\gamma=2$), we get that
\begin{align}\label{step_2}
\lim_{u\downarrow 0}
\frac{G(u)}{u^{\gamma}} = C_2(\gamma).
\end{align}
Therefore, for any $\varepsilon>0$, there exists $\delta>0$ such that for all $u\leq\delta$,
\begin{equation}\label{generating}
(1-\varepsilon)C_2(\gamma)\leq \frac{G(u)}{u^{\gamma}}\leq (1+\varepsilon)C_2(\gamma).
\end{equation}
Plugging \eqref{generating} into the definition of $\Phi_R$ in \eqref{main}, we get
\begin{align*}
\Phi_R(x)&\geq \mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} G(u\left(x-\xi_{\mathbf{e}}\right))1_{\left\{u\left(x-\xi_{\mathbf{e}}\right)<\delta\right\}}\right]\\
&\geq (1-\varepsilon)C_2(\gamma)\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma}\left(x-\xi_{\mathbf{e}}\right)1_{\left\{u\left(x-\xi_{\mathbf{e}}\right)<\delta\right\}}\right]\\
&\geq (1-\varepsilon)C_2(\gamma)\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma}\left(x-\xi_{\mathbf{e}}\right)\right]-\frac{(1-\varepsilon)C_2(\gamma)}{\delta}\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma+1}\left(x-\xi_{\mathbf{e}}\right)\right],
\end{align*}
where in the last inequality we used $1_{\left\{u<\delta\right\}}=1-1_{\left\{u\geq\delta\right\}}$ and $1_{\left\{u\geq\delta\right\}}\leq u/\delta$.
Thus \eqref{remainder1.1} is valid.
On the other hand, since $G(x)\leq 1$ for $x\in [0,1]$, we have
\begin{align}\label{upper-of-phi-r}
\Phi_R(x)&= \mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} G(u\left(x-\xi_{\mathbf{e}}\right))1_{\left\{u\left(x-\xi_{\mathbf{e}}\right)<\delta\right\}}\right]+\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} G(u\left(x-\xi_{\mathbf{e}}\right))1_{\left\{u\left(x-\xi_{\mathbf{e}}\right)\geq\delta\right\}}\right]\nonumber \\
&\leq (1+\varepsilon)C_2(\gamma)\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma}\left(x-\xi_{\mathbf{e}}\right)\right]+\frac{1}{\delta^{\gamma+1}}\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{\gamma+1}\left(x-\xi_{\mathbf{e}}\right)\right],
\end{align}
which implies \eqref{remainder1.2}.
We now consider the case $m<1$ and prove (ii) under the assumption $c_+>0$.
Note that for any $\varepsilon>0$, there exists $\delta>0$ such that
\[
G(u)\leq \varepsilon u, \qquad \mbox{ for all } u\leq \delta.
\]
Similar to \eqref{upper-of-phi-r}, using the fact that $G(x)\leq 1$, we get that
\begin{align}
\Phi_R(x)&= \mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} G(u\left(x-\xi_{\mathbf{e}}\right))1_{\left\{u\left(x-\xi_{\mathbf{e}}\right)<\delta\right\}}\right]+\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} G(u\left(x-\xi_{\mathbf{e}}\right))1_{\left\{u\left(x-\xi_{\mathbf{e}}\right)\geq\delta\right\}}\right]\\
&\leq \varepsilon \mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]+\frac{1}{\delta^{2}}\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{2}\left(x-\xi_{\mathbf{e}}\right)\right].
\end{align}
Therefore, to prove \eqref{remainder2}, it suffices to show that
\begin{equation}\label{suffice}
\lim_{x\to\infty}\frac{\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{2}\left(x-\xi_{\mathbf{e}}\right)\right]}{\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]}
=0.
\end{equation}
Considering the case where the initial particle does not split before time 1 and using \eqref{Tail-of-xi}, we get
$$
u(x)\geq e^{-1}\mathbf{P}_0(S_1>x)\geq e^{-1}\mathbf{P}_0(\xi_1>x)\geq \frac{c_+}{2e\alpha}x^{-\alpha}.
$$
when $x$ is large enough. Therefore, when $x$ is large enough,
\begin{align}\label{de}
\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]
&\geq \mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)1_{\left\{|\xi_{\mathbf{e}}|<\frac{1}{2}x\right\}}\right] \nonumber\geq \frac{c_+}{2e\alpha} \mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}}\left(x-\xi_{\mathbf{e}}\right)^{-\alpha
}1_{\left\{|\xi_{\mathbf{e}}|<\frac{1}{2}x\right\}}\right]\\
&\geq \frac{c_+}{2e\alpha} \left(\frac{3}{2}x\right)^{-\alpha}\mathbf{P}_0\left(S_{\mathbf{e}}<x, |\xi_{\mathbf{e}}|<\frac{1}{2}x\right) \geq c_1x^{-\alpha},
\end{align}
for some positive constant $c_1$, where we used the fact that $\lim_{x\to\infty} \mathbf{P}_0\left(S_{\mathbf{e}}<x, |\xi_{\mathbf{e}}|<\frac{1}{2}x\right)=1$.
Now we consider the numerator. Using the fact that $u\leq 1$, we get that that for $\delta^{\prime}>0$ sufficiently small,
\begin{align}\label{upp-u-2}
&\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{2}\left(x-\xi_{\mathbf{e}}\right)\right]\nonumber\\
&=\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^2\left(x-\xi_{\mathbf{e}}\right) 1_{\left\{\xi_{\mathbf{e}}<\left(1-\delta^{\prime}\right) x\right\}}\right]+\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^2\left(x-\xi_{\mathbf{e}}\right) 1_{\left\{\left(1-\delta^{\prime}\right) x<\xi_{\mathbf{e}}<x\right\}}\right] \nonumber\\
&\leq u(\delta^{\prime} x)\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]+\mathbf{P}_0\left((1-\delta^{\prime})x<\xi_{\mathbf{e}}<x\right).
\end{align}
Combining \eqref{de} and \eqref{upp-u-2}, we obtain that
\begin{align}
&\limsup_{x\to\infty}
\frac{\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{2}\left(x-\xi_{\mathbf{e}}\right)\right]}{\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]} \leq \lim_{x\to\infty} u(\delta^{\prime} x) + c_1\limsup_{x\to\infty}x^\alpha \mathbf{P}_0\left((1-\delta^{\prime})x<\xi_{\mathbf{e}}<x\right)\nonumber\\
& = c_1\limsup_{x\to\infty}x^\alpha \int_0^\infty e^{-z} \mathbf{P}_0\left((1-\delta^{\prime})x<\xi_{z}<x\right)\mathrm{d}z.
\end{align}
From \cite[Lemma 2.2]{Ren_Song_Zhang_2023}, when $x$ is large enough,
\[
x^\alpha \mathbf{P}_0\left((1-\delta^{\prime})x<\xi_{z}<x\right)\leq x^\alpha\mathbf{P}_0\left(|\xi_{z}| >(1-\delta^{\prime})x\right)
\leq
c_2z, \quad \mbox{ for all } z>0
\]
for some positive constant $c_2$.
Therefore, combining the dominated convergence theorem and \eqref{Tail-of-xi},
we get that
\begin{align}
&\limsup_{x\to\infty}
\frac{\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u^{2}\left(x-\xi_{\mathbf{e}}\right)\right]}{\mathbf{E}_0\left[1_{\left\{S_{\mathbf{e}}<x\right\}} u\left(x-\xi_{\mathbf{e}}\right)\right]} \leq \frac{c_2c_+}{\alpha} \int_0^\infty ze^{-z} \mathrm{d}z\frac{\delta'}{(1-\delta')^{1+\alpha}} \stackrel{\delta'\downarrow 0}{\longrightarrow}0,
\end{align}
which implies \eqref{suffice}.
\end{proof}
\section{Proof of Theorem \ref{t1}}
\subsection{The case $0<\alpha <1$}
Define
\begin{equation}\label{def-eta}
\eta_\alpha(\lambda):=\Gamma(1-\alpha) \lambda^{\alpha-1}
\end{equation}
and $\xi_{\mathbf{e}}^{+}=\max \left(0, \xi_{\mathbf{e}}\right)$.
It follows from \cite[(2.1)]{profeta} that
\begin{equation}\label{asymptotics}
\lim_{\lambda \downarrow 0}\frac{1-\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]}{\lambda\cdot \eta_\alpha(\lambda)} =\lim_{\lambda \downarrow 0} \frac{1-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]}{\lambda\cdot \eta_\alpha(\lambda)} =
\frac{c_+}{\alpha}.
\end{equation}
We will use $\mathcal{L}[f]$ to denote the Laplace transform of a positive function $f$:
$$
\mathcal{L}[f](\lambda):=\int_{0}^{+\infty}e^{-\lambda x}f(x) \mathrm{d} x, \quad \lambda>0.
$$
The following lemma is given in \cite[Lemma 2.1]{profeta}
\begin{lemma}\label{Important-lemma}
Assume that $\alpha < 1$
and that $f: [0,\infty)\to [0,\infty)$ is a positive and decreasing function.
\noindent
(i)
For any $\lambda>0$, it holds that
\[
\int_0^\infty e^{-\lambda x} \mathbf{E}_0\left[1_{\{S_{\mathbf{e}}<x\}}f(x-\xi_{\mathbf{e}})\right]\mathrm{d}x \leq \mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]\mathcal{L}[f](\lambda).
\]
\noindent
(ii)
For any $\lambda>0$, it holds that
\begin{align}
& \int_0^\infty e^{-\lambda x} \mathbf{E}_0\left[1_{\{S_{\mathbf{e}}<x\}}f(x-L_{\mathbf{e}})\right]\mathrm{d}x\nonumber\\
&\geq \mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^+}\right]\mathcal{L}[f](\lambda) +f(0)\frac{\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^+}\right]}{\lambda} - \mathbf{E}_0\left[1_{\left\{\xi_{\mathbf{e}}<0\right\}} \int_0^{-\xi_{\mathbf{e}}} e^{-\lambda z} f(z) \mathrm{d} z\right] .
\end{align}
\noindent
(iii) If in addition that $\lim_{x\to\infty} f(x)=0$, then
\begin{align}
\lim_{\lambda\downarrow 0} \frac{1}{\eta_{\alpha}(\lambda)} \mathbf{E}_0\left[1_{\left\{\xi_{\mathbf{e}}<0\right\}} \int_0^{-\xi_{\mathbf{e}}} e^{-\lambda z} f(z) \mathrm{d} z\right] = 0.
\end{align}
\end{lemma}
The next lemma can be found in \cite[Lemma 2.2]{profeta}.
\begin{lemma}\label{l3}
For any $\lambda>0$, it holds that
$$
\frac{\lambda}{1-\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]} \mathcal{L}\left[\Phi_0+\Phi_R\right](\lambda) \leq 1
$$
and
$$
\frac{\lambda}{1-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]} \mathcal{L}\left[\Phi_0+\Phi_R\right](\lambda) \geq 1-\lambda \mathcal{L}[u](\lambda)-\frac{\lambda}{1-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]} \mathbf{E}_0\left[1_{\left\{\xi_{\mathbf{e}}<0\right\}} \int_0^{-\xi_{\mathbf{e}}} e^{-\lambda z} u(z) \mathrm{d} z\right] .
$$
\end{lemma}
\bigskip
\begin{proof}[Proof of Theorem \ref{t1} for $\alpha < 1$]
Recall that $u(x)=\mathbb{P}\left(M \geq x\right)$. Using a change of variables and the monotone convergence theorem, we get
$$
\lambda \mathcal{L}[u](\lambda)=\int_0^{+\infty} e^{-z} u\left(\frac{z}{\lambda}\right) \mathrm{d} z \underset{\lambda \downarrow 0}{\longrightarrow} 0 .
$$
Combining \eqref{asymptotics} and Lemma \ref{Important-lemma} (iii) with $f=u$ and Lemma \ref{l3}, we get
\begin{equation}\label{upper bdd}
\lim_{\lambda \downarrow 0}\frac{\mathcal{L}\left[\Phi_0+\Phi_R\right](\lambda)}{\eta_\alpha(\lambda)} =
\frac{c_+}{\alpha}.
\end{equation}
We first consider the case $m=1$. In this case, $\Phi_0(x)=0$.
Combining \eqref{remainder1.2} and Lemma \ref{Important-lemma} (i) with $f=u^{\gamma}$ and $f=u^{\gamma+1}$, we get
\begin{equation}\label{PR}
\mathcal{L}\left[\Phi_R\right](\lambda) \leq
C_2(\gamma)
(1+\varepsilon)\mathcal{L}\left[u^{\gamma}\right](\lambda)+\frac{1}{\delta^{\gamma+1}} \mathcal{L}\left[u^{\gamma+1}\right](\lambda).
\end{equation}
Since $\lim _{x \rightarrow+\infty} u(x)=0$, for any
$\varepsilon_1>0$, there exists $A_1>0$ such that $u(x) \leq \varepsilon_1$ for $x \geq A_1$.
Hence,
\begin{align}\label{u^{gamma+1}}
\mathcal{L}\left[u^{\gamma+1}\right](\lambda)&=\int_{0}^{A_1} e^{-\lambda x}u^{\gamma+1}(x)\mathrm{d} x+\int_{A_1}^{+\infty} e^{-\lambda x}u^{\gamma+1}(x)\mathrm{d} x\nonumber \\
&\leq A_1
+\varepsilon_1\int_{A_1}^{+\infty}e^{-\lambda x}u^{\gamma}(x)\mathrm{d} x\leq A_{1}+\varepsilon_1 \mathcal{L}\left[u^{\gamma}\right](\lambda),
\end{align}
where in the first inequality we used the fact that $u \leq 1$.
Thus,
combining
\eqref{upper bdd}, \eqref{PR} and \eqref{u^{gamma+1}}, we have
\begin{align*}
\frac{c_+}{\alpha}=\lim_{\lambda \downarrow 0}\frac{\mathcal{L}\left[\Phi_R\right](\lambda)}{\eta_\alpha(\lambda)}
\leq & \liminf _{\lambda \downarrow 0}
\frac{\left(C_2(\gamma)(1+\varepsilon)+\varepsilon_1
/\delta^{\gamma+1}\right)\mathcal{L}\left[u^{\gamma}\right](\lambda)+A_1
/\delta^{\gamma+1}}{\eta_\alpha(\lambda)}\\
=& \liminf _{\lambda \downarrow 0}
\frac{\left(C_2(\gamma)(1+\varepsilon)+\varepsilon_1
/\delta^{\gamma+1}\right)\mathcal{L}\left[u^{\gamma}\right](\lambda)}{\eta_\alpha(\lambda)}.
\end{align*}
Letting $\varepsilon_1\to 0$ first and then $\varepsilon\to 0$, we get that
\begin{align} \label{proof-liminf}
\frac{c_+}{\alpha}\leq \liminf _{\lambda \downarrow 0} \frac{C_2(\gamma)\mathcal{L}\left[u^{\gamma}\right](\lambda)}{\eta_\alpha(\lambda)}.
\end{align}
On the other hand, combining \eqref{remainder1.1}, \eqref{u^{gamma+1}},
Lemma \ref{Important-lemma} (ii) with $f=u^{\gamma}$ and Lemma \ref{Important-lemma} (i) with $f=u^{\gamma+1}$, we see that
\begin{align}\label{3}
& \mathcal{L}\left[\Phi_R\right](\lambda) \nonumber\\
&\geq
C_2(\gamma)(1-\varepsilon)\left( \mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right] \mathcal{L}[u^{\gamma}](\lambda)+ \frac{\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]}{\lambda}-\mathbf{E}_0\left[1_{\left\{\xi_{\mathbf{e}}<0\right\}} \int_0^{-\xi_{\mathbf{e}}} e^{-\lambda z} u^{\gamma}(z) \mathrm{d} z\right]\right)\nonumber\\
&-\frac{C_2(\gamma)(1-\varepsilon)}{\delta}\left( A_1+\varepsilon_1 \mathcal{L}\left[u^{\gamma}\right](\lambda)\right).
\end{align}
Dividing both sides by $\eta_\alpha(\lambda)$ and
using Lemma \ref{Important-lemma} (iii) with $f=u^{\gamma}$, we obtain
$$
\frac{c_+}{\alpha} \geq \limsup _{\lambda \downarrow 0} \frac{C_2(\gamma)(1-\varepsilon)\left(1-\varepsilon_1
/\delta\right)\mathcal{L}[u^{\gamma}](\lambda)}{\eta_\alpha(\lambda)}.
$$
Letting $\varepsilon_1\to 0$ first and then $\varepsilon\to0$, we conclude that
\begin{align} \label{proof-limsup}
\frac{c_+}{\alpha} \geq \limsup _{\lambda \downarrow 0} \frac{C_2(\gamma)\mathcal{L}\left[u^{\gamma}\right](\lambda)}{\eta_\alpha(\lambda)}.
\end{align}
Combining \eqref{proof-liminf} and \eqref{proof-limsup}, we conclude that
$$
\lim_{\lambda \downarrow 0}
\frac{\mathcal{L}\left[u^{\gamma}\right](\lambda)}{\eta_\alpha(\lambda)}=\frac{c_+}{\alpha C_2(\gamma)}.
$$
Hence, by the Tauberian theorem, the above limit is equivalent to
$$
\lim_{x \to +\infty}
\frac{1}{\eta_\alpha\left(\frac{1}{x}\right)}\int_0^{x}u^{\gamma}(z) \mathrm{d} z= \frac{c_+}{\alpha\Gamma(2-\alpha)C_2(\gamma)}.
$$
Applying Karamata's
monotone density theorem \cite[Theorem 1.7.2]{bingham_goldie_teugels_1987}, we get the desired result for $m=1$.
Now we deal with the subcritical case $m<1$. For any $\varepsilon^{\prime}>0$,
by \eqref{remainder2}, we see that there exists a constant $A'$ such that for all $x\geq A'$,
\begin{equation}\label{varepsilon_prime}
0\leq \Phi_R(x)\leq \varepsilon^{\prime}\Phi_0(x).
\end{equation}
Similar to \eqref{u^{gamma+1}}, using \eqref{varepsilon_prime}, we get that for all $\lambda>0$,
\begin{align}\label{Bounds-for-Phi-R}
0\leq \mathcal{L}{[\Phi_R]}(\lambda) \leq A'+ \varepsilon^{\prime} \int_{A'}^\infty e^{-\lambda x} \Phi_0(x)\mathrm{d}x \leq A'+ \varepsilon^{\prime} \mathcal{L}{[\Phi_0]}(\lambda),
\end{align}
which together with \eqref{upper bdd} implies that
\begin{align}
& \limsup_{\lambda\downarrow 0} \frac{\mathcal{L}[\Phi_0](\lambda)}{\eta_\alpha(\lambda)} \leq \frac{c_+}{\alpha}\leq \liminf_{\lambda\downarrow 0} \frac{(1+\varepsilon')\mathcal{L}[\Phi_0](\lambda)+ A'}{\eta_\alpha(\lambda)} = (1+\varepsilon')\liminf_{\lambda\downarrow 0} \frac{\mathcal{L}[\Phi_0](\lambda)}{\eta_\alpha(\lambda)} .
\end{align}
Letting $\varepsilon'\downarrow 0$, we get
$$
\lim_{\lambda \downarrow 0}
\frac{1}{\eta_\alpha(\lambda)}\mathcal{L}[u](\lambda)=\frac{c_+}{\alpha(1-m)}.
$$
Hence, by the Tauberian theorem, we have
$$
\lim_{x \rightarrow +\infty}
\frac{1}{\eta_\alpha\left(\frac{1}{x}\right)} \int_0^{x}u(z) \mathrm{d} z =\frac{c_+}{\alpha\Gamma(2-\alpha)(1-m)} .
$$
Applying Karamata's
monotone density theorem \cite[Theorem 1.7.2]{bingham_goldie_teugels_1987}, we get the desired result.
\end{proof}
\subsection{Proof of Theorem \ref{t1} for $1\leq \alpha<2$}
It follows from \cite[(3.3)] {profeta} that for $\alpha\in (1,2)$,
\begin{equation}\label{a2}
\lim_{\lambda \downarrow 0}\frac{\int_0^{+\infty} e^{-\lambda x} x \mathbf{P}_0\left(S_{\mathbf{e}} \geq x\right) \mathrm{d} x}{\lambda^{\alpha-2}}
=\lim_{\lambda \downarrow 0}\frac{1-\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]-\lambda \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]}{\lambda^2\cdot\lambda^{\alpha-2}} = \frac{c_+\Gamma(2-\alpha)}{\alpha}.
\end{equation}
For $\alpha=1$, combining \eqref{Tail-of-xi}, \cite[Lemma 2.2]{Ren_Song_Zhang_2023} and the dominated convergence theorem, \eqref{a2} remains true for $\alpha=1$ since $\Gamma(1)=1$. Moreover, \eqref{a2} also holds with $S_{\mathbf{e}}$ replaced by $\xi_{\mathbf{e}}^+$.
\begin{lemma}\label{Important-lemma2}
Assume that $\alpha\in [1,2)$
and that $f:[0,\infty)\to [0,\infty)$ is a positive and decreasing function.
(i) We have the following upper bound
\[
\int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{S_{\mathbf{e}}<x\}}f(x-\xi_{\mathbf{e}})\right]\mathrm{d}x \leq
\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]\mathcal{L}[xf(x)](\lambda)+\mathbf{E}_0\left[S_{\mathbf{e}}e^{-\lambda S_{\mathbf{e}}}\right]\mathcal{L}[f](\lambda).
\]
(ii) We have the following lower bound
\begin{align}
& \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{S_{\mathbf{e}}<x\}}f(x-\xi_{\mathbf{e}})\right]\mathrm{d}x \nonumber\\
&\geq f(0)\int_0^{+\infty} e^{-\lambda x} x\left(\mathbf{P}_0\left(\xi_{\mathbf{e}} \geq x\right)-\mathbf{P}_0\left(S_{\mathbf{e}} \geq x\right)\right) \mathrm{d} x + \mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}}1_{\{\xi_{\mathbf{e}}\geq 0 \}}\right]\mathcal{L}[xf(x)](\lambda)\nonumber\\
&\qquad +\int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}<0\}}f(x-\xi_{\mathbf{e}})\right]\mathrm{d}x.
\end{align}
\end{lemma}
\begin{proof}
For (i), see \cite[Lemma 3.1]{profeta} for the proof of $\alpha\in (1,2)$ and the proof for $\alpha =1$ is the same.
Now we prove (ii). Combining the inequalities $1_{\{S_{\mathbf{e}}\geq x \}} \geq 1_{\{\xi_{\mathbf{e}}\geq x \}} $ and $f(x)\leq f(0)$ for all $x\geq 0$, we have
\begin{align}\label{eq1}
& \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{S_{\mathbf{e}}<x\}}f(x-\xi_{\mathbf{e}})\right]\mathrm{d}x \nonumber\\
&\geq f(0)\int_0^{+\infty} e^{-\lambda x} x\left(\mathbf{P}_0\left(\xi_{\mathbf{e}} \geq x\right)-\mathbf{P}_0\left(S_{\mathbf{e}} \geq x\right)\right) \mathrm{d} x + \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{\xi_{\mathbf{e}}<x\}}f(x-\xi_{\mathbf{e}})\right]\mathrm{d}x .
\end{align}
By Fubini's theorem, we have
\begin{align}\label{eq2}
& \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{0\leq \xi_{\mathbf{e}}<x\}}f(x-\xi_{\mathbf{e}})\right]\mathrm{d}x = \int_0^\infty e^{-\lambda x}x \int_0^x f(x-y)\mathbf{P}_0(\xi_{\mathbf{e}}\in \mathrm{d} y)\mathrm{d}x \nonumber\\
& = \int_0^\infty \mathbf{P}_0(\xi_{\mathbf{e}}\in \mathrm{d} y) \int_y^\infty e^{-\lambda x}x f(x-y)\mathrm{d}x \geq \int_0^\infty \mathbf{P}_0(\xi_{\mathbf{e}}\in \mathrm{d} y) \int_y^\infty e^{-\lambda x}(x-y) f(x-y)\mathrm{d}x \nonumber\\
& =\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}}1_{\{\xi_{\mathbf{e}}\geq 0 \}}\right]\mathcal{L}[xf(x)](\lambda).
\end{align}
Now (ii) follows from \eqref{eq1} and \eqref{eq2}
\end{proof}
\begin{lemma}\label{l5}
For any $\lambda>0$, it holds that
(i)
$$
\frac{\lambda^2}{1-\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]-\lambda \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]} \mathcal{L}\left[x(\Phi_0(x)+\Phi_R(x))\right](\lambda) \leq 1+\frac{\lambda^2 \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]}{1-\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]-\lambda \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]} \mathcal{L}[u](\lambda),
$$
(ii)
\begin{align}
&\frac{\lambda^2}{1-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]-\lambda \mathbf{E}_0\left[\xi_{\mathbf{e}}^{+} e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]} \mathcal{L}\left[x(\Phi_0(x)+\Phi_R(x))\right](\lambda) \nonumber\\
& \geq 1- \frac{\lambda^2 \mathbf{E}_0\left[ 1- e^{-\lambda \xi_{\mathbf{e}}^+}\right]\mathcal{L}[xu(x)](\lambda)}{1-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]-\lambda \mathbf{E}_0\left[\xi_{\mathbf{e}}^{+} e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]} \nonumber\\
& \quad - \frac{\lambda^2\left( \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda)
-
\int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}<0\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x\right)}{1-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]-\lambda \mathbf{E}_0\left[\xi_{\mathbf{e}}^{+} e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]}.
\end{align}
\end{lemma}
\begin{proof}
The proof of (i) for $\alpha\in (1,2)$ can be found in \cite[Lemma 3.2]{profeta} and the case $\alpha=1$ can be treated similarly. Now we prove (ii). Combining
\eqref{integral equation}
and Lemma \ref{Important-lemma2} (ii) with $f=u$, we see
that
\begin{align}
& \mathcal{L}\left[x(\Phi_0(x)+\Phi_R(x))\right](\lambda) + \mathcal{L}(xu(x))(\lambda) \nonumber\\
& = \int_0^{+\infty} e^{-\lambda x} x \mathbf{P}_0\left(S_{\mathbf{e}} \geq x\right) \mathrm{d} x+ \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{S_{\mathbf{e}}<x\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x \nonumber\\
& \geq \int_0^{+\infty} e^{-\lambda x} x \mathbf{P}_0\left(\xi_{\mathbf{e}} \geq x\right) \mathrm{d} x+ \mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}}1_{\{\xi_{\mathbf{e}}\geq 0 \}}\right]\mathcal{L}[xu(x)](\lambda)\nonumber\\
&\qquad +
\int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{\xi_{\mathbf{e}}<x\}}f(x-\xi_{\mathbf{e}})\right]\mathrm{d}x .
\end{align}
From the argument of \eqref{a2} and
the fact that $\xi_{\mathbf{e}}= \xi_{\mathbf{e}}^+$ on the set $\{\xi_{\mathbf{e}}\geq x\}$ for $x\geq 0$, we get the lower bound
\begin{align}
& \mathcal{L}\left[x(\Phi_0(x)+\Phi_R(x))\right](\lambda) + \mathcal{L}(xu(x))(\lambda)\nonumber\\
&\geq \int_0^{+\infty} e^{-\lambda x} x \mathbf{P}_0\left(\xi_{\mathbf{e}}^+ \geq x\right) \mathrm{d} x-\mathbf{E}_0\left[(1-e^{-\lambda\xi_{\mathbf{e}}^+})1_{\left\{\xi_{\mathbf{e}}\geq 0\right\}}\right]\mathcal{L}(xu(x))(\lambda) \nonumber\\
&\quad - \left( \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda)
- \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}<0\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x\right)\nonumber\\
& = \frac{1-\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]-\lambda \mathbf{E}_0\left[\xi_{\mathbf{e}}^{+} e^{-\lambda \xi_{\mathbf{e}}^{+}}\right]}{\lambda^2} -\mathbf{E}_0\left[1-e^{-\lambda\xi_{\mathbf{e}}^+}\right]\mathcal{L}(xu(x))(\lambda) \nonumber\\
&\quad - \left( \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda)
- \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}<0\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x\right),
\end{align}
which implies the desired result.
\end{proof}
In the critical case $m=1$, we will need the following
a priori upper bound of $u$.
\begin{lemma}\label{l6}
If $\alpha\in (1,2)$ and $m=1$,
then there exists a constant $A>0$ such that,
$$
u(x)\leq A x^{-\alpha / \gamma}, \qquad \mbox{ for all } x\geq 0.
$$
\end{lemma}
\begin{proof}
The proof is similar to that of \cite[Lemma 3.3]{profeta}.
The main difference is that we have \eqref{222} for our branching mechanism. Denote by $\underline{M}^{(t)}$ the maximum of $(Z_s: s\geq 0)$ on $[0,t]$ and by $\overline{M}^{(t)}$ the maximum of $(Z_s: s\geq 0)$ on $[t,+\infty]$.
Since $\alpha \in (1,2)$, we have $\rho:=\mathbf{P}_0(\xi_1 \geq 0)\in [1-\frac{1}{\alpha},\frac{1}{\alpha}]$.
It follows that
\begin{align}\label{domi-u}
u(x)&=\mathbb P(M\geq x)
\leq \mathbb P(\underline{M}^{(t)} \geq x)+\mathbb P(\overline{M}^{(t)} \geq x)\nonumber\\
&\leq \mathbb P(\underline{M}^{(t)} \geq x)+\mathbb P\left(N_{t} \geq 1\right).
\end{align}
Define $T_x= \inf \left\{t \geq 0: \exists u \in N_{t}, X_{u}(t)\geq x\right\}$. Then $\left\{T_x \leq t\right\}=\left\{\underline{M}^{(t)} \geq x\right\}$. Applying strong Markov property at $T_x$, we get
$$
\mathbb{E}\left[
\sum_{u\in N_t} 1_{\{X_u(t)\geq x\}}
\mid \underline{M}^{(t)} \geq x\right] \geq \rho,
$$
which implies that
\begin{equation}\label{111}
\mathbb{P}\left(\underline{M}^{(t)} \geq x\right) \leq \rho^{-1} \mathbb{E}\left[
\sum_{u\in N_t} 1_{\{X_u(t)\geq x\}}
\right] \leq \rho^{-1}\mathbb{E}[N_{t}] \mathbf{P}_0\left(\xi_t \geq x\right)=\rho^{-1} \mathbf{P}_0\left(t^{1 / \alpha} \xi_1 \geq x\right),
\end{equation}
where we used the facts that $N_t$ is independent of the spatial positions, and $\mathbb{E}[N_{t}]=1$ for all $t \geq 0$.
It follows from \eqref{generating} that the function
$\sum_{k=0}^\infty p_k(1-x)^k-(1-x)$
is a regularly varying at 0 with index $\gamma$.
Hence by \cite{Zolotarev} [Theorem 4] we know that
$Q(t):=\mathbb P(N_{t} \geq 1)$ satisfies the following equation for some positive constant $c$:
$$
\lim_{t \to \infty}\frac{Q(t)}{t(\gamma-1) c Q(t)^{\gamma}}=1,
$$
which implies that
\begin{equation}\label{222}
\mathbb P(N_{t} \geq 1) \sim \left(\frac{1}
{c (\gamma-1)t}\right)^{\frac{1}{\gamma-1}}.
\end{equation}
Taking $t=x^{\alpha(1-\frac{1}{\gamma})}$ in \eqref{111} and \eqref{222},
using \eqref{domi-u}, we get that there is some constant $A>0$ such that
$$
u(x) \leq A x^{-\alpha / \gamma}.
$$
This completes the proof of the Lemma.
\end{proof}
\begin{proof}[Proof of Theorem \ref{t1} for $\alpha\in [1,2)$]
We first prove that
\begin{equation}\label{toprove-Laplace-limit}
\lim_{\lambda \downarrow 0}\frac{\mathcal{L}\left[x (\Phi_0(x)+\Phi_R(x))\right](\lambda)}{\lambda^{\alpha-2}}
= \frac{c_+\Gamma(2-\alpha)}{\alpha} .
\end{equation}
For the upper bound, combining \eqref{a2} and Lemma \ref{l5}(i), we have
\begin{align}\label{e4}
& \frac{\alpha}{c_+\Gamma(2-\alpha)} \limsup_{\lambda \downarrow 0}\frac{\mathcal{L}\left[x (\Phi_0(x)+\Phi_R(x))\right](\lambda)}{\lambda^{\alpha-2}} \leq 1+ \limsup_{\lambda \downarrow 0} \frac{\lambda^2 \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]}{1-\mathbf{E}_0\left[e^{-\lambda S_{\mathbf{e}}}\right]-\lambda \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]} \mathcal{L}[u](\lambda)\nonumber\\
& = 1+\frac{\alpha}{c_+\Gamma(2-\alpha)} \limsup_{\lambda \downarrow 0} \lambda^{2-\alpha} \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]\mathcal{L}[u](\lambda).
\end{align}
When $\alpha\in (\gamma,2)$, combining Lemma \ref{l6} and the fact that $\mathbf{E}_0(S_{\mathbf{e}})<\infty$, we see that
\begin{align}\label{eq4}
\lambda^{2-\alpha} \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]\mathcal{L}[u](\lambda)\leq \mathbf{E}_0\left[S_{\mathbf{e}} \right] \lambda^{2-\alpha} \left(1+ A\int_1^\infty x^{-\alpha/\gamma}\mathrm{d}x\right)\stackrel{\lambda\downarrow 0}{\longrightarrow} 0.
\end{align}
When $\alpha\in (1, \gamma]$,
using $\alpha\in (1,2)$ and $\gamma\in (1,2]$, we have $1+\alpha \gamma^{-1} = \alpha + \gamma^{-1} \left(1-(\alpha-1)(\gamma-1)\right)>\alpha$. Thus there exists $\delta\in (0,1)$ such that $1+\alpha\delta /\gamma >\alpha$.
Therefore, combining Lemma \ref{l6} and $\mathbf{E}_0(S_{\mathbf{e}})<\infty$,
\begin{align}\label{eq5}
&\lambda^{2-\alpha} \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]\mathcal{L}[u](\lambda)\leq A \mathbf{E}_0\left[S_{\mathbf{e}} \right] \lambda^{2-\alpha} \int_0^\infty e^{-\lambda x} x^{-\alpha\delta/\gamma}\mathrm{d}x\nonumber\\
&= A \mathbf{E}_0\left[S_{\mathbf{e}} \right] \lambda^{1+\alpha\delta/\gamma-\alpha} \int_0^\infty e^{-x} x^{-\alpha\delta/\gamma}\mathrm{d}x\stackrel{\lambda\downarrow0}{\longrightarrow}0.
\end{align}
When $\alpha=1$,
fix a constant $\delta\in (1-1/\gamma,1)$.
Combining \cite[(2.1)]{profeta} and Markov's inequality, we have
\[
y^{\delta}\mathbf{P}_0(S_{\mathbf{e}}\geq y) \leq
\frac{y^{\delta}}{(1-e^{-1})}
\mathbf{E}_0(1-e^{-y^{-1}S_{\mathbf{e}}}) =\frac{y^{\delta-1} \ln y}{(1-e^{-1})} \frac{\mathbf{E}_0(1-e^{-y^{-1}S_{\mathbf{e}}})}{y^{-1}\ln y} \stackrel{y\to\infty}{\longrightarrow} 0.
\]
Thus there exists $c_\delta>0$ such that
$\mathbf{P}_0(S_{\mathbf{e}}\geq y)\leq c_\delta y^{-\delta}$ for any $y>0$,
which implies that
\begin{align}\label{eq6}
\lambda \mathbf{E}_0\left[S_{\mathbf{e}} e^{-\lambda S_{\mathbf{e}}}\right]\mathcal{L}[u](\lambda) & =
\lambda\int_0^\infty (1-\lambda y)e^{-\lambda y} \mathbf{P}_0(S_{\mathbf{e}}>y)\mathrm{d}y \int_0^\infty e^{-\lambda x} u(x)\mathrm{d}x \nonumber\\&
\leq A \lambda \int_0^\infty e^{-\lambda y}\mathbf{P}_0(S_{\mathbf{e}}\geq y)\mathrm{d}y \int_0^\infty e^{-\lambda x} x^{-1/\gamma}\mathrm{d} x\nonumber\\
& \leq Ac_\delta \lambda \int_0^\infty e^{-\lambda y} y^{-\delta}\mathrm{d}y \int_0^\infty e^{-\lambda x} x^{-1/\gamma}\mathrm{d} x\nonumber\\
& = Ac_\delta \lambda^{\delta+ \gamma^{-1}-1} \int_0^\infty e^{- y} y^{-\delta}\mathrm{d}y \int_0^\infty e^{- x} x^{-1/\gamma}\mathrm{d} x\stackrel{\lambda\downarrow 0}{\longrightarrow}0.
\end{align}
Combining \eqref{e4}, \eqref{eq4}, \eqref{eq5} and \eqref{eq6}, we get
\begin{align}\label{to-prove-upper}
\limsup_{\lambda \downarrow 0}\frac{\mathcal{L}\left[x (\Phi_0(x)+\Phi_R(x))\right](\lambda)}{\lambda^{\alpha-2}}
\leq \frac{c_+\Gamma(2-\alpha)}{\alpha} .
\end{align}
Now we prove the lower bound. Similarly, combining \eqref{a2} and Lemma \ref{l5}(ii), we see that for any $\varepsilon>0$,
\begin{align}\label{eq8}
& \frac{\alpha}{c_+\Gamma(2-\alpha)} \liminf_{\lambda \downarrow 0}\frac{\mathcal{L}\left[x (\Phi_0(x)+\Phi_R(x))\right](\lambda)}{\lambda^{\alpha-2}}
\nonumber\\
& \geq 1- \frac{\alpha}{c_+\Gamma(2-\alpha)} \limsup_{\lambda \downarrow 0} \lambda^{2-\alpha} \mathbf{E}_0\left[ 1- e^{-\lambda \xi_{\mathbf{e}}^+}\right]\mathcal{L}[xu(x)](\lambda) - \frac{\alpha}{c_+\Gamma(2-\alpha)} \nonumber\\
&\quad \times \limsup_{\lambda \downarrow 0} \lambda^{2-\alpha} \left( \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda) - \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}<0\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x\right).
\end{align}
From Lemma \ref{l6}, it holds that
\begin{align}\label{eq10}
\mathcal{L}[xu(x)](\lambda) \leq A\int_0^\infty e^{-\lambda x} x^{1- \gamma^{-1}\alpha}\mathrm{d}x = A \lambda^{\alpha/\gamma -2}\int_0^\infty e^{- x} x^{1- \gamma^{-1}\alpha}\mathrm{d}x.
\end{align}
Since $1-\gamma^{-1}\alpha > 1- 2=-1$, we conclude from the above inequality that for any $\alpha\in [1,2)$, there exists $A'$ such that
\begin{align}
\lambda^{2-\alpha} \mathbf{E}_0\left[ 1- e^{-\lambda \xi_{\mathbf{e}}^+}\right]\mathcal{L}[xu(x)](\lambda) \leq A' \lambda^{\alpha/\gamma-\alpha} \mathbf{E}_0\left[ 1- e^{-\lambda \xi_{\mathbf{e}}^+}\right].
\end{align}
Note that when $\alpha=1$, by \cite[(2.1)]{profeta}, $\lim_{\lambda\downarrow 0} \frac{1}{\lambda \ln(\lambda^{-1})} \mathbf{E}_0\left[ 1- e^{-\lambda \xi_{\mathbf{e}}^+}\right] = \frac{c_+}{\alpha}$, and when $\alpha\in (1,2)$, $\lim_{\lambda\downarrow 0} \frac{1}{\lambda } \mathbf{E}_0\left[ 1- e^{-\lambda \xi_{\mathbf{e}}^+}\right] =\mathbf{E}_0(\xi_{\mathbf{e}}^+ ) $. Using the fact that $\alpha/\gamma -\alpha +1 = \gamma^{-1}(1-(\alpha-1)(\gamma-1))>0$, we have
\begin{align}\label{eq7}
\limsup_{\lambda\downarrow 0} \lambda^{2-\alpha} \mathbf{E}_0\left[ 1- e^{-\lambda \xi_{\mathbf{e}}^+}\right]\mathcal{L}[xu(x)](\lambda) =0.
\end{align}
Plugging \eqref{eq7} into \eqref{eq8} yields that
\begin{align}\label{eq9}
& \frac{\alpha}{c_+\Gamma(2-\alpha)} \liminf_{\lambda \downarrow 0}\frac{\mathcal{L}\left[x (\Phi_0(x)+\Phi_R(x))\right](\lambda)}{\lambda^{\alpha-2}}
\geq 1 - \frac{\alpha}{c_+\Gamma(2-\alpha)} \nonumber\\
&\quad \times \limsup_{\lambda \downarrow 0} \lambda^{2-\alpha} \left( \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda) - \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}<0\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x\right).
\end{align}
When $\alpha\in (1,2)$, using the fact that $\mathbf{E}_0(|\xi_{\mathbf{e}}|)<\infty$, we see that
\begin{align}\label{eq11}
& \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[
1_{\{ \xi_{\mathbf{e}}\leq 0\}}
u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x = \mathbf{E}_0\left(
1_{\left\{\xi_{\mathbf{e}}\leq 0\right\}}
\int_{-\xi_{\mathbf{e}}}^\infty e^{-\lambda(x+\xi_{\mathbf{e}})}(x+\xi_{\mathbf{e}})u(x)\mathrm{d}x\right)\nonumber\\
&\geq \mathbf{E}_0\left(
1_{\left\{\xi_{\mathbf{e}}\leq 0\right\}}
\int_{-\xi_{\mathbf{e}}}^\infty e^{-\lambda x}(x+\xi_{\mathbf{e}})u(x)\mathrm{d}x\right) \geq \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda) \nonumber\\
&\quad - \mathbf{E}_0\left(1_{\left\{
\xi_{\mathbf{e}}\leq 0\right\}}
\int_0^{-\xi_{\mathbf{e}}} e^{-\lambda x}xu(x)\mathrm{d}x\right) - \mathbf{E}_0(|\xi_{\mathbf{e}}|) \mathcal{L}(u)(\lambda)\nonumber\\
&\geq \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda) -2 \mathbf{E}_0(|\xi_{\mathbf{e}}|) \mathcal{L}(u)(\lambda).
\end{align}
Therefore, combining \eqref{eq4} and \eqref{eq5},
we conclude from \eqref{eq11} that
\begin{align}
&\limsup_{\lambda \downarrow 0} \lambda^{2-\alpha} \left( \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda) - \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}<0\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x\right)\nonumber\\
&\leq 2\mathbf{E}_0(|\xi_{\mathbf{e}}|) \limsup_{\lambda \downarrow 0} \lambda^{2-\alpha} \mathcal{L}(u)(\lambda) =
0.
\end{align}
Combining the above inequality with
\eqref{to-prove-upper} and \eqref{eq9}, we get \eqref{toprove-Laplace-limit} holds for $\alpha\in (1,2)$.
Now we consider the case $\alpha=1$.
For any fixed $\delta\in (0,1)$,
noticing that $\mathbf{E}_0(|\xi_{\mathbf{e}}|^{1-\delta})<\infty$, we get that
\begin{align}\label{eq11'}
& \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}\leq 0\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x = \mathbf{E}_0\left(1_{\left\{\xi_{\mathbf{e}}\leq 0\right\}}\int_{-\xi_{\mathbf{e}}}^\infty e^{-\lambda(x+\xi_{\mathbf{e}})}(x+\xi_{\mathbf{e}})u(x)\mathrm{d}x\right)\nonumber\\
&\geq \mathbf{E}_0\left(1_{\left\{\xi_{\mathbf{e}}\leq 0\right\}}\int_{-\xi_{\mathbf{e}}}^\infty e^{-\lambda x}(x+\xi_{\mathbf{e}})u(x)\mathrm{d}x\right) = \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda) \nonumber\\
&\quad - \mathbf{E}_0\left(1_{\left\{\xi_{\mathbf{e}}<0\right\}}\int_0^{-\xi_{\mathbf{e}}} e^{-\lambda x}xu(x)\mathrm{d}x\right) - \mathbf{E}_0\left(1_{\left\{\xi_{\mathbf{e}}<0\right\}}\int_{-\xi_{\mathbf{e}}}^\infty e^{-\lambda x} |\xi_{\mathbf{e}}| u(x)\mathrm{d}x\right) \nonumber\\
&\geq \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda) - 2\mathbf{E}_0(|\xi_{\mathbf{e}}|^{1-\delta}) \mathcal{L}(x^\delta u(x))(\lambda).
\end{align}
By taking $\delta<\gamma^{-1}$, we get from Lemma \ref{l6} and \eqref{eq11'} that
\begin{align}\label{eq12}
&\limsup_{\lambda \downarrow 0} \lambda \left( \mathbf{P}_0 \left(\xi_{\mathbf{e}}\leq 0\right) \mathcal{L}(xu(x))(\lambda) - \int_0^\infty e^{-\lambda x}x \mathbf{E}_0\left[1_{\{ \xi_{\mathbf{e}}<0\}}u(x-\xi_{\mathbf{e}})\right]\mathrm{d}x\right)\nonumber\\
&\leq 2\mathbf{E}_0(|\xi_{\mathbf{e}}|^{1-\delta}) \limsup_{\lambda \downarrow 0} \lambda \mathcal{L}(x^\delta u(x))(\lambda) \leq 2A\mathbf{E}_0(|\xi_{\mathbf{e}}|^{1-\delta}) \limsup_{\lambda \downarrow 0} \lambda \int_0^\infty e^{-\lambda y} y^{\delta-1/\gamma}\mathrm{d} y\nonumber\\
& = 2A\mathbf{E}_0(|\xi_{\mathbf{e}}|^{1-\delta}) \limsup_{\lambda \downarrow 0} \lambda^{1/\gamma -\delta} \int_0^\infty e^{-y} y^{\delta-1/\gamma}\mathrm{d} y=0.
\end{align}
Combining \eqref{to-prove-upper}, \eqref{eq9} and \eqref{eq12}, we get that \eqref{toprove-Laplace-limit} holds when $\alpha=1$.
The rest of the proof is now similar to the case
$\alpha \in(0,1)$.
We first consider the case $m=1$. In this case, $\Phi_0(x)=0$.
Combining
\eqref{remainder1.2} and
Lemma \ref{Important-lemma2} (i)
with $f=u^{\gamma}$ and $f=u^{\gamma+1}$, we get that
\begin{align}\label{upper-phi-r-2}
\mathcal{L}\left[x\Phi_R\right](\lambda) &\leq (1+\varepsilon)C_2(\gamma)\left( \mathcal{L}\left[xu^{\gamma}(x)\right](\lambda)+\mathbf{E}_0\left[S_{\mathbf{e}}\right]\mathcal{L}\left[u^{\gamma}\right](\lambda)\right)\nonumber \\
&\quad +\frac{1}{\delta^{\gamma+1}}\left( \mathcal{L}\left[xu^{\gamma+1}(x)\right](\lambda)+ \mathbf{E}_0\left[S_{\mathbf{e}}\right]\mathcal{L}\left[u^{\gamma+1}\right](\lambda) \right) .
\end{align}
For any $\varepsilon_1>0$, since $\lim _{x \rightarrow+\infty} u(x)=0$,
there exists
$A_2>0$
such that $u(x) \leq \varepsilon_1$ for $x \geq A_2$.
Similar to \eqref{u^{gamma+1}}, we have
\begin{equation}\label{xu^{gamma+1}}
\mathcal{L}\left[xu^{\gamma+1}(x)\right](\lambda)\leq A_{2}^2+\varepsilon_1 \mathcal{L}\left[xu^{\gamma}(x)\right](\lambda).
\end{equation}
Combining \eqref{toprove-Laplace-limit}, \eqref{upper-phi-r-2} and \eqref{xu^{gamma+1}}, we get
\begin{align}
\frac{c_+\Gamma(2-\alpha)}{\alpha} \leq
\left((1+\varepsilon)C_2(\gamma)+\varepsilon_1/\delta^{\gamma+1}\right)
\liminf _{\lambda \downarrow 0}\lambda^{2-\alpha} \mathcal{L}\left[xu^{\gamma}(x)\right](\lambda).
\end{align}
Letting $\varepsilon_1\to 0$ first and then $\varepsilon\to 0$, we get that
\begin{align}\label{lowerbound2}
\frac{c_+\Gamma(2-\alpha)}{\alpha} \leq
C_2(\gamma) \liminf _{\lambda \downarrow 0}\lambda^{2-\alpha} \mathcal{L}\left[xu^{\gamma}(x)\right](\lambda).
\end{align}
Next, combining \eqref{xu^{gamma+1}}, Lemma \ref{Important-lemma2} (ii) with $f=u^{\gamma}$ and Lemma \ref{Important-lemma2} (i) with $f=u^{\gamma+1}$, it holds that
\begin{align*}
& \mathcal{L}\left[x\Phi_R(x)\right](\lambda) \geq C_2(\gamma)(1-\varepsilon)\int_0^{+\infty} e^{-\lambda x} x\left(\mathbf{P}_0\left(\xi_{\mathbf{e}} \geq x\right)-\mathbf{P}_0\left(S_{\mathbf{e}} \geq x\right)\right) \mathrm{d} x\\
& +C_2(\gamma)(1-\varepsilon)\left\{\mathbf{E}_0\left[e^{-\lambda \xi_{\mathbf{e}}^{+}}\right] \mathcal{L}\left[x u^{\gamma}(x)\right](\lambda)-\mathbf{E}_0\left[1_{\left\{\xi_{\mathbf{e}}<0\right\}} \int_0^{-\xi_{\mathbf{e}}} e^{-\lambda z} z u^{\gamma}(z) \mathrm{d} z\right]+ \mathbf{E}_0\left[\xi_{\mathbf{e}} e^{-\lambda \xi_{\mathbf{e}}^{+}}\right] \mathcal{L}\left[u^{\gamma}\right](\lambda)\right\}\\
& -\frac{1}{\delta^{\gamma+1}} \left\{A_{1}^2+\varepsilon_1 \mathcal{L}\left[xu^{\gamma}(x)\right](\lambda)+ \mathbf{E}_0\left[S_{\mathbf{e}}\right]\mathcal{L}\left[u^{\gamma+1}\right](\lambda)\right\}.
\end{align*}
Applying Lemma \ref{Important-lemma2} (iii) with $f=u^{\gamma}$, we obtain
$$
\frac{c_+\Gamma(2-\alpha)}{\alpha} \geq
\left(C_2(\gamma)(1-\varepsilon)-\varepsilon_1 /\delta^{\gamma+1}\right) \limsup _{\lambda \downarrow 0}\lambda^{2-\alpha}\mathcal{L}\left[xu^{\gamma}(x)\right](\lambda).
$$
Letting $\varepsilon_1\to 0$ first and then $\varepsilon\to 0$, we get that
\begin{align}\label{upperbound2}
\frac{c_+\Gamma(2-\alpha)}{\alpha}\geq C_2(\gamma) \limsup _{\lambda \downarrow 0}\lambda^{2-\alpha} \mathcal{L}\left[xu^{\gamma}(x)\right](\lambda).
\end{align}
Combining \eqref{lowerbound2} and \eqref{upperbound2}, we conclude that
$$
\lim_{\lambda \downarrow 0}\frac{\mathcal{L}\left[x u^{\gamma}(x)\right](\lambda)}{\lambda^{\alpha-2}}=
\frac{c_+\Gamma(2-\alpha)}{C_2(\gamma)\alpha}.
$$
Finally, by the Tauberian theorem, we get
$$
\lim_{x \rightarrow+\infty}\frac{1}{x^{2-\alpha}}
\int_0^x z u^{\gamma}(z) \mathrm{d} z= \frac{c_+}{\alpha(2-\alpha)C_2(\gamma)}.
$$
The desired result now follows from Karamata's monotone density theorem \cite[Theorem 1.7.2]{bingham_goldie_teugels_1987}.
Next we consider the case $m<1$. Combining $\Phi_R\geq 0$, \eqref{varepsilon_prime}
and Lemma \ref{Important-lemma2} (i) with $f=u$,
we get that for any $\varepsilon^{\prime}>0$, there exists a constant $A_3=A_3(\varepsilon^{\prime})$, such that for all small $\lambda$,
$$
\mathcal{L}\left[x\Phi_0(x)\right](\lambda) \leq \mathcal{L}\left[x(\Phi_0+\Phi_R)\right](\lambda) \leq (1+\varepsilon^{\prime})(1-m)\left\{\mathcal{L}[xu(x)](\lambda)+ \mathbf{E}_0\left[S_{\mathbf{e}}\right]\mathcal{L}[u](\lambda)\right\}+A_3,
$$
which, by \eqref{toprove-Laplace-limit}, implies that
\begin{equation}\label{lowerbound}
(1-m) \limsup _{\lambda \downarrow 0}\lambda^{2-\alpha} \mathcal{L}[xu(x)](\lambda)\leq \frac{c_+\Gamma(2-\alpha)}{\alpha} \leq
(1-m)(1+\varepsilon') \liminf _{\lambda \downarrow 0}\lambda^{2-\alpha} \mathcal{L}[xu(x)](\lambda) .
\end{equation}
Letting $\varepsilon'\downarrow 0$, we get that
$$
\lim_{\lambda \downarrow 0}
\frac{1}{\lambda^{\alpha-2}}\mathcal{L}\left[xu(x)\right](\lambda)=\frac{c_+\Gamma(2-\alpha)}{\alpha(1-m)}.
$$
Finally, by the Tauberian theorem,
$$
\lim_{x \rightarrow+\infty}
\frac{1}{x^{2-\alpha}}\int_0^x z u(z) \mathrm{d} z= \frac{c_+\Gamma(2-\alpha)}{\alpha(1-m)}.
$$
The desired result now follows from Karamata's monotone density theorem \cite[Theorem 1.7.2]{bingham_goldie_teugels_1987}.
\end{proof}
\section{Proof of Theorem \ref{t2}}
Define $\widetilde{\xi}=-\xi$, $\widetilde{\tau}_y:= \inf\left\{t: \widetilde{\xi}_t\leq y\right\}$ and
\[
f(u) := \frac{G(u)}{u}= \frac{\sum_{k=0}^\infty p_k (1-u)^k -(1-u)}{u}, \quad u\in [0, 1].
\]
According to \cite[Lemma 2.3]{HJRS2025}, $u$ has the following representation:
\begin{align}\label{FK-formula}
u(x)= \mathbf{E}_x\left(\exp\left\{-\int_0^{\widetilde{\tau}_y}f\left(u\left(\widetilde{\xi}_s\right)\right)\mathrm{d}s\right\}\right) u(y), \quad 0\leq y <x.
\end{align}
Similar to \cite{HJRS2025}, we consider the function
\[
[0,\infty) \ni x\mapsto \frac{u\left(x+ y u(x)^{-\frac{\gamma-1}{\alpha}}\right)}{u(x)},
\]
which is bounded between $0$ and $1$. Therefore, using a diagonalization argument, we can find a subsequence
$\{x_k\in [0,\infty)\}$ with $\lim_{k\to\infty} x_k =+\infty$ such that for all $y\geq 0, y\in \mathbb{Q}$, the following limits exist:
\begin{align}\label{step_7}
\phi(y):= \lim_{k\to\infty} \frac{u\left(x_k+ y u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)}{u(x_k)}.
\end{align}
Since $u(x)$ is decreasing, we see that $\phi(0)=1$ and $\phi(y)\in [0,1]$ for any $y\in \mathbb{Q}\cap [0,\infty)$. Moreover, $\phi$ is decreasing in $\mathbb{Q}\cap [0,\infty)$. Therefore, for any $y\geq 0$, we can define
\begin{align}\label{step_8}
\phi(y):= \sup_{z\in \mathbb{Q}, z\geq y} \phi(z) = \lim_{z\in \mathbb{Q}, z\downarrow y} \phi(y).
\end{align}
\begin{lemma}\label{l:lemma4}
The limit \eqref{step_7} holds for all $y\geq 0$.
Also, it holds that for any $K>0$,
\begin{align}\label{step_5}
\lim_{k\to\infty} \sup_{y\in [0,K]} \left| \frac{u\left(x_k+ y u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)}{ \phi (y)u(x_k)} -1\right| =0.
\end{align}
Moreover, $\phi$ satisfies the equation
\begin{align}\label{PDE}
\phi(y) = \mathbf{E}_{y}\left(\exp\left\{-C_2(\gamma)\int_0^{\widetilde{\tau}_0}\left(\phi(\widetilde{\xi}_s)\right)^{\gamma-1}\mathrm{d}s\right\} \right),\quad y\geq 0,
\end{align}
where $C_2(\gamma)$ is defined in \eqref{step_2}.
\end{lemma}
\begin{proof}
Fix two arbitrary non-negative rational numbers $y_1<y_2$ and set $z_i(k)= y_i u(x_k)^{-\frac{\gamma-1}{\alpha}}$, $i=1, 2$. Combining the definition of $\phi$ and \eqref{FK-formula}, we get that
\begin{align}\label{step_1}
\phi(y_1) & \geq \phi(y_2) = \lim_{k\to\infty} \frac{u\left(x_k+ z_2(k)\right)}{u(x_k)} \nonumber\\
& = \lim_{k\to\infty} \mathbf{E}_{x_k+ z_2(k)}\left(\exp\left\{-\int_0^{\widetilde{\tau}_{x_k+z_1(k)}}f\left(u\left(\widetilde{\xi}_s\right)\right)\mathrm{d}s\right\}\right) \frac{u\left(x_k +z_1(k)\right) }{u(x_k)}\nonumber\\
& = \phi(y_1) \lim_{k\to\infty} \mathbf{E}_{z_2(k)}\left(\exp\left\{-\int_0^{\widetilde{\tau}_{z_1(k)}}f\left(u\left(x_k +\widetilde{\xi}_s\right)\right)\mathrm{d}s\right\}\right).
\end{align}
Combining the scaling property of $\widetilde{\xi}$, \eqref{step_2} and \eqref{step_1}, we get that there exists $c>0$ such that
\begin{align}
\phi(y_1) & \geq \phi(y_2) \geq \phi(y_1) \lim_{k\to\infty} \mathbf{E}_{z_2(k)}\left(\exp\left\{- c\int_0^{\widetilde{\tau}_{z_1(k)}}\left(u\left(x_k +\widetilde{\xi}_s\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right)\nonumber\\
& = \phi(y_1)\lim_{k\to\infty} \mathbf{E}_{y_2 }\left(\exp\left\{- c\int_0^{u(x_k)^{-(\gamma-1)}\widetilde{\tau}_{y_1}}\left(u\left(x_k +\widetilde{\xi}_{s u(x_k)^{\gamma-1}} u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right)\nonumber\\
& = \phi(y_1)\lim_{k\to\infty} \mathbf{E}_{y_2 }\left(\exp\left\{- c\int_0^{\widetilde{\tau}_{y_1}}\left(\frac{u\left(x_k +\widetilde{\xi}_{s } u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)}{u(x_k)}\right)^{\gamma-1}\mathrm{d}s\right\}\right).
\end{align}
Since $u\left(x_k +\widetilde{\xi}_{s } u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)\leq u(x_k)$ for all $s\leq \widetilde{\tau}_{y_1}$, we get from the above inequality that
\begin{align}\label{step_3}
\phi(y_1) & \geq \phi(y_2) \geq \phi(y_1) \mathbf{E}_{y_2 }\left(\exp\left\{- c\widetilde{\tau}_{y_1}\right\}\right) = \phi(y_1) \mathbf{E}_{1}\left(\exp\left\{- c(y_2-y_1)^\alpha\widetilde{\tau}_{0}\right\}\right),
\end{align}
where in the last equality we also used the scaling property of $\widetilde{\xi}$.
Therefore, for any $y> 0$ and any positive rational number $y_1 \leq y<y_2$, we have
\begin{align}\label{step_4}
\phi(y_2) & =\lim_{k\to\infty} \frac{u\left(x_k+ y_2 u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)}{u(x_k)}\leq \liminf_{k\to\infty} \frac{u\left(x_k+ y u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)}{u(x_k)}\nonumber\\
& \leq \limsup_{k\to\infty} \frac{u\left(x_k+ y u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)}{u(x_k)}\leq \lim_{k\to\infty} \frac{u\left(x_k+ y_1 u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)}{u(x_k)}=\phi(y_1).
\end{align}
Combining \eqref{step_3} and \eqref{step_4}, we see that \eqref{step_7} holds for all $y\geq 0$.
Taking $y_1=0$ in \eqref{step_3}, we see that $\inf_{y\in [0,K]} \phi (y)>0$.
Therefore, using an argument similar to that leading to \cite[(3.9)]{HJRS2025}, we can get \eqref{step_5}.
Now we prove \eqref{PDE}. Since $u(x_k +\widetilde{\xi}_s)\leq u(x_k)$ for any $s\leq \widetilde{\tau}_{0}$, combining \eqref{step_2} and \eqref{FK-formula}, we see that for any $\varepsilon >0$,
\begin{align}
\phi(y) &= \lim_{k\to\infty} \frac{u\left(x_k+ y u(x_k)^{-\frac{\gamma-1}{\alpha}}\right)}{u(x_k)} = \lim_{k\to\infty} \mathbf{E}_{y u(x_k)^{-\frac{\gamma-1}{\alpha}} }\left(\exp\left\{-\int_0^{\widetilde{\tau}_0}f\left(u\left(x_k +\widetilde{\xi}_s\right)\right)\mathrm{d}s\right\}\right)\nonumber\\
& \geq \lim_{k\to\infty} \mathbf{E}_{y u(x_k)^{-\frac{\gamma-1}{\alpha}} }\left(\exp\left\{-C_2(\gamma)(1+\varepsilon)\int_0^{\widetilde{\tau}_0}\left(u\left(x_k + \widetilde{\xi}_s\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right).
\end{align}
Applying the scaling property again, we get from the above inequality that
\begin{align}
\phi(y) &\geq \lim_{k\to\infty} \mathbf{E}_{y}\left(\exp\left\{-C_2(\gamma)(1+\varepsilon)\int_0^{\widetilde{\tau}_0}\left(\frac{u\left(x_k + \widetilde{\xi}_su(x_k)^{-\frac{\gamma-1}{\alpha}} \right)}{u(x_k)}\right)^{\gamma-1}\mathrm{d}s\right\}\right)\nonumber\\
& = \mathbf{E}_{y}\left(\exp\left\{-C_2(\gamma)(1+\varepsilon)\int_0^{\widetilde{\tau}_0}\left(\phi(\widetilde{\xi}_s)\right)^{\gamma-1}\mathrm{d}s\right\} \right),
\end{align}
where in the last equality we used the dominated convergence theorem. Letting $\varepsilon\downarrow 0$, we get the lower bound. The proof of the upper bound is similar and we omit the details.
\end{proof}
As a consequence of Lemma \ref{l:lemma4}, we see that $\phi(y)\in (0, 1)$ for all $y>0$.
\begin{proposition}\label{prop1}
(i) $\phi$ is the unique solution of of \eqref{PDE}.
(ii) The equation
\begin{align}\label{Another-PDE}
U(z) = \mathbf{E}_{z}\left(\exp\left\{-C_2(\gamma)\int_0^{\widetilde{\tau}_y}\left(U(\widetilde{\xi}_s)\right)^{\gamma-1}\mathrm{d}s\right\} \right)U(y),\quad z> y\geq 0
\end{align}
has a unique solution satisfying the boundary conditions $\lim_{y\downarrow 0} U(y)=+\infty$ and $\lim_{y\to\infty} U(y)=0$.
\end{proposition}
The proof of Proposition \ref{prop1} is postponed to Section \ref{s:5}.
\begin{lemma}\label{test-lemma}
It holds that
\[
\limsup_{x\to\infty} x^{\frac{\alpha}{\gamma-1}} u(x) = \limsup_{x\to\infty} x^{\frac{\alpha}{\gamma-1}} \mathbb{P}\left(M\geq x\right) <\infty.
\]
and that
\[
\liminf_{x\to\infty} x^{\frac{\alpha}{\gamma-1}} u(x) = \liminf_{x\to\infty} x^{\frac{\alpha}{\gamma-1}} \mathbb{P}\left(M \geq x\right) >0.
\]
\end{lemma}
\begin{proof}
Define $w(x)= x^{\frac{\alpha}{\gamma-1}} u(x)$ and set $A= \liminf_{x\to +\infty} w(x)$ and $B= \limsup_{x\to\infty} w(x)$. It suffices to show that $0<A \leq B<\infty$.
Combining \eqref{step_5} and Proposition \ref{prop1}, we get that, for any $K>0$,
\begin{align}\label{step_6}
0 & = \lim_{x\to+\infty} \sup_{y\in [0,K]} \left| \frac{u\left(x+ y u(x)^{-\frac{\gamma-1}{\alpha}}\right)}{ \phi (y)u(x)} -1\right| \nonumber\\
&=\lim_{x\to\infty} \sup_{y\in [0,K]} \left| \frac{w\left(x \left(1+ y w(x)^{-\frac{\gamma-1}{\alpha}}\right)\right)}{ \phi (y)w(x) \left(1+ yw(x)^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}}} -1\right| .
\end{align}
First we show that $A<\infty$. If $A=\infty$, we define $b_k:= \sup\left\{x: w(x)< k\right\}$. Then $b_k\to\infty$ as $k\to\infty$.
Using the the definition of $b_k$ and the left-continuity of $w$, we get that $w(b_k)\leq k\leq \inf_{z>b_k} w(z)$.
Therefore, taking $x=b_k$ in \eqref{step_6}, we get that for any $y>0$,
\begin{align}
1& = \lim_{k\to\infty} \frac{w\left(b_k \left(1+ y w(x)^{-\frac{\gamma-1}{\alpha}}\right)\right)}{ \phi (y)w(b_k) \left(1+ yw(b_k)^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}}} \geq \lim_{k\to\infty} \frac{k}{ \phi (y) \left(w(b_k)^{\frac{\gamma-1}{\alpha}}+ y\right)^{\frac{\alpha}{\gamma-1}}}\nonumber\\
& \geq \lim_{k\to\infty} \frac{k}{ \phi (y) \left(k^{\frac{\gamma-1}{\alpha}}+ y\right)^{\frac{\alpha}{\gamma-1}}} = \frac{1}{\phi(y)},
\end{align}
which is a contraction to
Lemma \ref{l:lemma4}.
Hence $A<\infty$. Similarly we can show that $B>0$.
Now we show that $B<\infty$. Assume that $A<B=\infty$.
Note that for any $K>0$,
\begin{align}
\lim_{A_1 \to\infty} \phi(K) \left(1+ K A_1^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}} = \phi(K)<1.
\end{align}
Therefore, we may fix an $A_1>A$ and an $\varepsilon>0$ such that
\begin{align}\label{step_10}
(1+\varepsilon) \phi(K) \left(1+ K A_1^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}} <1.
\end{align}
Fix another $B_1>A_1$. Define
\begin{align}
a_1& := \inf\left\{x>0: w(x) <A_1\right\},\quad d_1:= \inf\left\{x>a_1: w(x) >B_1+1\right\},\nonumber\\
a_k &:= \inf\left\{x>d_{k-1}: w(x)< A_1 \right\},\quad d_k:=\inf\left\{x>a_k: w(x)> B_1+k\right\},\nonumber\\
a_k^*&:= \sup\left\{x\in [a_k, d_k]: w(x)< A_1\right\}.
\end{align}
By \eqref{step_6}, for any $\varepsilon, K>0$ satisfying \eqref{step_10}, there exists $N>0$ such that when $x>N$.
\begin{align}\label{step_9}
\sup_{y\in [0,K]} \left| \frac{w\left(x \left(1+ y w(x)^{-\frac{\gamma-1}{\alpha}}\right)\right)}{ \phi (y)w(x) \left(1+ yw(x)^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}}} -1\right| <\varepsilon.
\end{align}
By the left continuity of $w$, taking $x=a_k^*$ in \eqref{step_9}, we see that when $k$ is large enough, for all $y\in [0, K]$,
\begin{align}\label{step_11}
w\left(a_k^* \left(1+ y w(a_k^*)^{-\frac{\gamma-1}{\alpha}}\right)\right) & \leq (1+\varepsilon) \phi (y)w(a_k^*) \left(1+ yw(a_k^*)^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}} \nonumber\\
& = (1+\varepsilon) \phi (y) \left(w(a_k^*)^{\frac{\gamma-1}{\alpha}}+ y\right)^{\frac{\alpha}{\gamma-1}} \leq (1+\varepsilon) \phi(y) \left(A_1^{\frac{\gamma-1}{\alpha}}+ K\right)^{\frac{\alpha}{\gamma-1}} \nonumber\\
&<B_1+k.
\end{align}
Therefore, we see that when $k$ is large enough,
\begin{align}\label{step_12}
\left\{ a_k^* \left(1+ y w(a_k^*)^{-\frac{\gamma-1}{\alpha}}\right): y\in [0, K]\right\} \subset [a_k^*, d_k] \quad \Longrightarrow \quad w\left(a_k^* \left(1+ K w(a_k^*)^{-\frac{\gamma-1}{\alpha}}\right)\right) \geq A_1.
\end{align}
However, combining \eqref{step_10} and \eqref{step_11}, we have
\begin{align}
w\left(a_k^* \left(1+ K w(a_k^*)^{-\frac{\gamma-1}{\alpha}}\right)\right) \leq (1+\varepsilon) \phi(K) \left(A_1^{\frac{\gamma-1}{\alpha}}+ K\right)^{\frac{\alpha}{\gamma-1}} < A_1,
\end{align}
which contradicts \eqref{step_12}. Therefore, $B<\infty$.
Now we prove $A>0$. If $A=0<B$,
then
combining \eqref{step_3} and \eqref{Laplace-of-stopping-time}, we see that for any $y>0$,
\begin{align}
\phi(y)\geq \mathbf{E}_0\left( \exp\left\{-c y^\alpha \widetilde{\tau}_{-1} \right\}\right)= e^{-c_1 y}
\end{align}
for some constant $c_1>0$, where $c$ is the constant in \eqref{step_3}. Note that for any $B_2>0$,
\begin{align}
\phi(y)^{\frac{\gamma-1}{\alpha}} \left(1+y B_2^{-\frac{\gamma-1}{\alpha}}\right) & \geq e^{-c_1 \frac{\gamma-1}{\alpha} y} \left(1+y B_2^{-\frac{\gamma-1}{\alpha}}\right) \geq \left(1-c_1 \frac{\gamma-1}{\alpha} y \right)\left(1+y B_2^{-\frac{\gamma-1}{\alpha}}\right) \nonumber\\
& = 1+ \left( B_2^{-\frac{\gamma-1}{\alpha}} -
c_1
\frac{\gamma-1}{\alpha} \right) y -c_1 \frac{\gamma-1}{\alpha}B_2^{-\frac{\gamma-1}{\alpha}}y^2.
\end{align}
Thus we can take $B_2$ and $K$ sufficiently small so that for all $y\in (0, K]$,
\[
\phi(y)^{\frac{\gamma-1}{\alpha}} \left(1+y B_2^{-\frac{\gamma-1}{\alpha}}\right)>1.
\]
Let $B_2$ and $K$ be chosen as above and fix $y\in (0, K]$. Then there exists $\varepsilon>0$ such that
\begin{align}\label{step_10'}
(1-\varepsilon) \phi (y)\left(1+ yB_2^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}} >1.
\end{align}
Take $N$ large enough so that $N^{-1}< B_2$ and define
\begin{align}
h_1& := \inf\left\{x>0: w(x) >B_2 \right\},\quad j_1:= \inf\left\{x>h_1: w(x) < \frac{1}{N+1} \right\},\nonumber\\
h_k &:= \inf\left\{x>j_{k-1}: w(x)> B_2 \right\},\quad j_k:=\inf\left\{x>h_k: w(x)< \frac{1}{N+k}\right\},\nonumber\\
h_k^*&:= \sup\left\{x\in [h_k, j_k]: w(x)>B_2\right\}.
\end{align}
Combining \eqref{step_9} and the left continuity of $w$, we see that $w(h_k^*)\geq B_2$ and that, when $k$ is large enough, for any $y\in [0, K]$,
\begin{align}\label{step_13}
w\left(h_k^* \left(1+ y w(h_k^*)^{-\frac{\gamma-1}{\alpha}}\right)\right) & \geq (1-\varepsilon) \phi (y)w(h_k^*) \left(1+ yw(h_k^*)^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}} \nonumber\\
& = (1-\varepsilon) \phi (y) \left(w(h_k^*)^{\frac{\gamma-1}{\alpha}}+ y\right)^{\frac{\alpha}{\gamma-1}} \geq (1-\varepsilon) \phi(K)B_2>\frac{1}{k+N},
\end{align}
which implies that
\begin{align}\label{step_12'}
\left\{ h_k^* \left(1+ y w(h_k^*)^{-\frac{\gamma-1}{\alpha}}\right): y\in [0, K]\right\} \subset [h_k^*, j_k] \quad \Longrightarrow \quad w\left(h_k^* \left(1+ K w(h_k^*)^{-\frac{\gamma-1}{\alpha}}\right)\right) \leq B_2.
\end{align}
However, combining \eqref{step_10'} and \eqref{step_13}, we get
\begin{align}\label{step_14}
w\left(h_k^* \left(1+ K w(h_k^*)^{-\frac{\gamma-1}{\alpha}}\right)\right) & \geq (1-\varepsilon) \phi (K)B_2\left(1+ KB_2^{-\frac{\gamma-1}{\alpha}}\right)^{\frac{\alpha}{\gamma-1}} >B_2,
\end{align}
which contradicts \eqref{step_12'}. Therefore, $A>0$ and the proof is copmplete.
\end{proof}
Now we are ready to prove Theorem \ref{t2}.
\begin{proof}[Proof of Theorem \ref{t2}]
Define $U^{(x)}(y):= x^{\frac{\alpha}{\gamma-1}}u(xy)$, then it follows
from Lemma \ref{test-lemma} that for some constant $\gamma_1, \gamma_2$ and $A$, it holds that
\begin{align}\label{Rough-bound}
\frac{\gamma_1}{y^{\frac{\alpha}{\gamma-1}}}\leq U^{(x)}(y)\leq \frac{\gamma_2}{y^{\frac{\alpha}{\gamma-1}}},\quad xy\geq A.
\end{align}
It follows from \eqref{step_2} that for any $u_0\in (0, 1)$,
$f(u)/u^{\gamma-1}\leq c$ for all $u\in [0, u_0]$ for some positive constant $c$.
Fixing a $y_0>0$. Then by \eqref{FK-formula}, when $x> A/y_0$, we see that for any $z>y>y_0$, under $\mathbf{P}_{xz}$, we have $u(\widetilde{\xi}_s)\leq \frac{\gamma_2}{\widetilde{\xi}_s^{\frac{\alpha}{\gamma-1}}}\leq \frac{\gamma_2}{(xy_0)^{\frac{\alpha}{\gamma-1}}}$ on the set $\{s< \widetilde{\tau}_{xy}\}$. Therefore, by \eqref{Laplace-of-stopping-time},
for any $y_0<y<z$,
\begin{align}\label{step_16}
U^{(x)}(z) & = \mathbf{E}_{xz}\left(\exp\left\{-\int_0^{\widetilde{\tau}_{xy}}f\left(u\left(\widetilde{\xi}_s\right)\right)\mathrm{d}s\right\}\right) U^{(x)}(y)\nonumber\\
& \geq \mathbf{E}_{xz}\left(\exp\left\{- c\int_0^{\widetilde{\tau}_{xy}}\left(u\left(\widetilde{\xi}_s\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right) U^{(x)}(y)\nonumber\\
&\geq \mathbf{E}_{xz}\left(\exp\left\{- \frac{c\gamma_2^{\gamma-1}}{(xy_0)^{\alpha}}\widetilde{\tau}_{xy} \right\}\right) U^{(x)}(y)= \mathbf{E}_{1}\left(\exp\left\{- \frac{c\gamma_2^{\gamma-1}}{(y_0)^{\alpha}}(z-y)^\alpha\widetilde{\tau}_{0} \right\}\right) U^{(x)}(y)\nonumber\\
& = e^{-c_1 (z-y)} U^{(x)}(y),
\end{align}
for some positive constant $c_1$.
Therefore, combining \eqref{Rough-bound} and \eqref{step_16}, we see that when $x>A/y_0$, for any $y_0<y<z$,
\begin{align}\label{step_15}
\left| U^{(x)}(z) -U^{(x)}(y) \right| = U^{(x)}(y) -U^{(x)}(z)\leq U^{(x)}(y)\left( 1- e^{-c_1(z-y)}\right) \leq \frac{\gamma_2}{y_0^{\frac{\alpha}{\gamma-1}}} c_1 |z-y|.
\end{align}
Therefore, combining \eqref{Rough-bound} and a diagonalization argument, we can find sequence from $\{t_k: k\geq1\}\subset (0,\infty)$ with $t_k\to\infty$ such that
\begin{align}\label{Limit2}
U(y)= \lim_{k\to\infty} U^{(t_k)}(y),\quad \mbox{for all}\quad y\in \mathbb{Q}\cap (0,\infty).
\end{align}
Moreover, using a standard argument (for example, see \cite[Lemma 3.1]{HRS2024}) and with the help of \eqref{step_15}, one can show that \eqref{Limit2} holds for all $y>0$.
taking $x=t_k$ in \eqref{Rough-bound} and letting $k\to\infty$, we see that
$U(y)$ is comparable to $y^{-\frac{\alpha}{\gamma-1}}$, which implies that $\lim_{y\downarrow 0} U(y)=\infty$ and that $\lim_{y\to\infty} U(y)=0$.
Let $z>y\geq y_0$. Since $u(\widetilde{\xi}) \leq u(xy)$ for all $s\leq \widetilde{\tau}_{xy}$ under $\mathbf{P}_{xz}$, we get from \eqref{step_2} that for any $\varepsilon, y_0>0$, when $x$ is large enough, we have,
\begin{align}
U^{(x)}(z) & \geq \mathbf{E}_{xz}\left(\exp\left\{-C_2(\gamma)(1+\varepsilon)\int_0^{\widetilde{\tau}_{xy}}\left(u\left(\widetilde{\xi}_s\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right) U^{(x)}(y)\nonumber\\
& = \mathbf{E}_{z}\left(\exp\left\{-C_2(\gamma)(1+\varepsilon)\int_0^{x^\alpha\widetilde{\tau}_{y}}\left(u\left(x\widetilde{\xi}_{s/x^\alpha}\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right) U^{(x)}(y)\nonumber\\
& = \mathbf{E}_{z}\left(\exp\left\{-C_2(\gamma)(1+\varepsilon)\int_0^{\widetilde{\tau}_{y}} x^\alpha\left(u\left(x\widetilde{\xi}_{s}\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right) U^{(x)}(y) \nonumber\\
& = \mathbf{E}_{z}\left(\exp\left\{-C_2(\gamma)(1+\varepsilon)\int_0^{\widetilde{\tau}_{y}} \left(U^{(x)}\left(\widetilde{\xi}_{s}\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right) U^{(x)}(y) .
\end{align}
Taking $x=t_k$ in the above inequality and then letting $k\to\infty$, we get that
\begin{align}
U(z) &\geq \mathbf{E}_{z}\left(\exp\left\{-C_2(\gamma)(1+\varepsilon)\int_0^{\widetilde{\tau}_{y}} \left(U\left(\widetilde{\xi}_{s}\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right) U(y) \nonumber\\
& \stackrel{\varepsilon\to 0}{\longrightarrow}\mathbf{E}_{z}\left(\exp\left\{-C_2(\gamma)\int_0^{\widetilde{\tau}_{y}} \left(U\left(\widetilde{\xi}_{s}\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right) U(y).
\end{align}
Similarly, we also have for any $y_0\leq y<z$,
\begin{align}
U(z) &\leq \mathbf{E}_{z}\left(\exp\left\{-C_2(\gamma)\int_0^{\widetilde{\tau}_{y}} \left(U\left(\widetilde{\xi}_{s}\right)\right)^{\gamma-1}\mathrm{d}s\right\}\right) U(y).
\end{align}
Therefore, $U$ is the solution of the equation in Proposition \ref{prop1}(ii). Now the constant $C_3(\alpha, \beta,\gamma)= U(1)$ follows immediately, we are done.
\end{proof}
\section{Proof of Theorem \ref{t3}}\label{s:5}
Repeating the proof of \cite[Lemma 2.3]{HJRS2025}, we can get that, in the case $m<1$,
$u$ has the following representation:
\begin{align}\label{FK-formula-sub}
u(x)= \mathbf{E}_x\left(\exp\left\{- (1-m)\widetilde{\tau}_y-\int_0^{\widetilde{\tau}_y}f_{sub}\left(u\left(\widetilde{\xi}_s\right)\right)\mathrm{d}s\right\}\right) u(y),\quad x>y\geq 0
\end{align}
where
\[
f_{sub}(u) = \frac{\sum_{k=0}^\infty p_k (1-u)^k -(1-m)(1-u)}{u}
\]
is a continuous function of $u\in (0,1]$ with $\lim_{u\to 0+}f_{sub}(u)=0$. Moreover, by \cite[Lemma 2.7]{HRSZ2025}, $f_{sub}$ is increasing in $u$, and
that if $\sum_{k=0}^\infty k(\log k)p_k<\infty$, then for any $c>0$, $\int_0^\infty f_{sub} \left(e^{-ct} \right)\mathrm{d} t<\infty$, which is equivalent to
\begin{align}\label{Inte-of-f-sub}
\sum_{n=1}^\infty f_{sub} \left(e^{-cn} \right)<\infty.
\end{align}
\begin{proof}[Proof of Theorem \ref{t3}]
For simplicity, define $a_0:= \left((1-m)/C_1(\alpha)\right)^{1/\alpha}.$ According to \eqref{Laplace-of-stopping-time}, we have for any $x>y\geq 0$,
\[
u(x)\leq \mathbf{E}_x\left(\exp\left\{- (1-m)\widetilde{\tau}_y\right\}\right) u(y) = e^{-a_0(x-y)}u(y).
\]
Therefore, we see that $e^{a_0x} u(x)$ is decreasing in $x$ and that $\lim_{x\to\infty} e^{a_0x} u(x)\in [0,1].$ Thus, it remains to show that the limit is positive.
Taking $x=n+1, y=n$ in \eqref{FK-formula-sub}, we see that
\begin{align}
u(n+1) = \mathbf{E}_{n+1}\left(\exp\left\{- (1-m)\widetilde{\tau}_n-\int_0^{\widetilde{\tau}_n}f_{sub}\left(u\left(\widetilde{\xi}_s\right)\right)\mathrm{d}s\right\}\right) u(n).
\end{align}
Since under $\mathbf{P}_{n+1}$, on the set $s<\widetilde{\tau}_n$, we have $\widetilde{\xi}_s \geq n$. Therefore, by the monotonicities of $f_{sub}$ and $u$, we conclude from the above identity that
\begin{align}\label{lower-of-u-n}
u(n+1) & \geq \mathbf{E}_{n+1}\left(\exp\left\{- (1-m)\widetilde{\tau}_n-\int_0^{\widetilde{\tau}_n}f_{sub}\left(u\left(n\right)\right)\mathrm{d}s\right\}\right) u(n)\nonumber\\
& \geq \mathbf{E}_{n+1}\left(\exp\left\{- \left(1-m+f_{sub}\left(e^{-a_0n}\right)\right)\widetilde{\tau}_n\right\}\right) u(n)\nonumber\\
& = \exp\left\{- H\left( 1-m+f_{sub}\left(e^{-a_0n}\right)\right)\right\} u(n),
\end{align}
where $H(a):= (a/C_1(\alpha))^{1/\alpha}$.
Noticing that for $\alpha\in (1,2)$, by Taylor's expansion, we have
\[
H(1-m+v)= a_0+ H'(1-m) v+ O(v^2), \quad v\to0.
\]
Therefore, there exists $C>0$ such that for all $v\in (0,1)$, $ H(1-m+v)\leq a_0+C v$.
Combining this inequality with \eqref{Inte-of-f-sub} and \eqref{lower-of-u-n},
we conclude that
\begin{align}
e^{a_0(n+1)} u(n+1) & \geq e^{a_0n} u(n) \exp\left\{- Cf_{sub}\left(e^{-a_0n}\right)\right\} \nonumber\\
&\geq \cdots \geq u(0) \exp\left\{-C\sum_{k=0}^n f_{sub}
\left(e^{-a_0k}\right) \right\}\nonumber\\
&\geq \exp\left\{-C\sum_{k=0}^\infty f_{sub}\left(e^{-a_0n}\right) \right\}>0,
\end{align}
which implies the desired result.
\end{proof}
\section{Proof of Proposition \ref{prop1}}\label{s:6}
The proof of Proposition \ref{prop1} relies heavily on another important Markov process: super $\alpha$-stable process. We will briefly introduce this process and some known results.
Let $\mathcal{M}_F(\mathbb R)$ be the families of finite Borel measures on $\mathbb R$. We will use $\mathbf{0}$ to denote the null measure on $\mathbb R$. Let $B_b(\mathbb R)$ and $B_b^+(\mathbb R)$ be the spaces of bounded Borel functions and non-negative bounded Borel functions on $\mathbb R$ respectively.
For any $f\in B_b(\mathbb R)$ and $\mu\in \mathcal{M}_F(\mathbb R)$,
we use $\langle f, \mu\rangle$ to denote the integral of $f$ with respect to $\mu$. For any $\alpha\in (1, 2]$, the function
\begin{align}\label{Stable-Branching-mechanism}
\varphi(\lambda) := C_2(\gamma) \lambda^\gamma
\end{align}
is a branching mechanism.
For any $\mu \in \mathcal{M}_F(\mathbb R)$, we use
$X=\{(X_t)_{t\geq 0}; \mathbb P_\mu\}$
to denote a super $\alpha$-stable process with spatial motion $\widetilde{\xi}$ and branching mechanism $\varphi$, that is, an $\mathcal{M}_F(\mathbb R)$-valued Markov process such that for any $f\in B_b^+(\mathbb R)$,
$$
-\log \mathbb{E}_{\mu} \left(\exp\left\{ -\langle f, X_{t} \rangle\right\}\right)=
\langle v_f(t, \cdot), \mu\rangle,
$$
where $(t, x)\mapsto v_f(t, x)$
is the unique locally bounded non-negative solution to
\begin{align}\label{Evolution-cumulant-semigroup}
v_f(t, x)&=\mathbf{E}_x\left(f(\widetilde{\xi}_t) \right)- \mathbf{E}_y\Big(\int_0^t \varphi\left( v_f(t-s,\widetilde{\xi}_s)\right)\mathrm{d}s\Big).
\end{align}
According to Dynkin \cite{E.B1.}, for any open set $Q$ of $\mathbb R$, there corresponds a random measure $X_Q$ such that,
$\mu\in \mathcal{M}_F(\mathbb R)$ with $\textup{supp}\ \mu \subset Q$, and any $f\in B_b^+(\mathbb R)$,
$$
\mathbb{E}_\mu \left(\exp\left\{-\langle f, X_Q \rangle\right\}\right) = \exp\big\{-\langle v_f^{Q} , \mu \rangle \big\},
$$
where
$v^{Q}_f(x)$
is the unique positive solution of the equation
\begin{align}\label{Equation-exit-measure}
v_f^{Q} (x) & = \mathbf{E}_{x} \left(f(\widetilde{\xi}_\tau)\right)- \mathbf{E}_{x} \int_0^{\tau_Q} \varphi\big(v_f^{Q}(\widetilde{\xi}_r)\big)\mathrm{d} r,
\end{align}
with $\tau_Q:= \inf\left\{ r: \widetilde\xi_r \notin Q\right\}$.
\begin{proof}[Proof of Proposition \ref{prop1}]
Taking $Q= (0, \infty)$
and $f=1$ in \eqref{Equation-exit-measure},
we see that $\phi(x)$ is the unique bounded solution of the following equation:
\begin{align}
\phi(x)& = 1- \mathbf{E}_{x} \int_0^{\widetilde{\tau}_0}\varphi\big(\phi( \widetilde{\xi}_r)\big)\mathrm{d} r.
\end{align}
Since the above equation is equivalent to \eqref{PDE}, we complete the proof of (i).
Now we turn to the proof of (ii). Similarly, let $U$ be an solution to \eqref{Another-PDE} with boundary condition $U(0+)=\infty$ and $U(\infty)=0$. Noticing that for each fixed $y$, \eqref{Another-PDE} is equivalent to
\begin{align}
U(z)& = U(y)- \mathbf{E}_{z} \int_0^{\widetilde{\tau}_y}\varphi\big(U( \widetilde{\xi}_r)\big)\mathrm{d} r, \quad z>y>0.
\end{align}
Therefore, since $\widetilde{\xi}$ is spectrally positive, we see that for $Q= (y,\infty)$,
is supported on $\{y\}$.
Therefore, from \eqref{Equation-exit-measure}, we conclude that
\begin{align}
U(z) = -\log \mathbb{E}_{\delta_z}\left(\exp\left\{-U(y) X_{(y,\infty)}(\left\{y\right\})\right\}\right) = -\log \mathbb{E}_{\delta_{z-y}}\left(\exp\left\{-U(y) X_{(0,\infty)}(\left\{0\right\})\right\}\right) ,
\end{align}
where in the last equality we used the spatial homogeneous property of
super $\alpha$-stable process. Therefore, replacing $z$ by $z+y$ first and then letting $y\to 0+$, we conclude that
\begin{align}
U(z)= \lim_{y\to0} U(z+y)= -\lim_{y\to0}\log \mathbb{E}_{\delta_{z}}\left(\exp\left\{-U(y) X_{(0,\infty)}(\left\{0\right\})\right\}\right) = -\log \mathbb{P}_{\delta_z}\left(X_{(0,\infty)}(\{0\})=0\right),
\end{align}
which implies the desired result.
\end{proof}
\noindent
{\bf Acknowledgements:}
We thank Yichao Huang for helpful discussions.
\bigskip
\small
\begin{thebibliography}{99}
\bibitem{bertoin} Bertoin, J.: L\'{e}vy processes. Cambridge University Press, Cambridge, 1996.
\bibitem{bingham_goldie_teugels_1987} Bingham, N. H., Goldie C. M. and Teugels, J. L.: Regular variation, Encyclopedia Math. Appl., 27.
Cambridge University Press, Cambridge, 1989.
\bibitem {E.B1.}
Dynkin, E. B.: Branching exit Markov systems and superprocesses. \emph{Ann. Probab.} \textbf{29} (2001), 1833--1858.
\bibitem{HJRS2025} Hou, H., Jiang, Y., Ren, Y.-X. and Song, R.: Tail probability of maximal displacement in critical branching L\'{e}vy process with stable
branching. \emph{Bernoulli} \textbf{31} (1) (2025) 630--648.
\bibitem{HRS2024}
Hou, H., Ren, Y.-X. and Song, R.:
Tails of extinction time and maximal displacement for critical branching killed L\'evy process.
arXiv:2405.09019.
\bibitem{HRSZ2025} Hou, H., Ren, Y.-X., Song, R. and Zhu, Y.: Asymptotic behaviours of subcritical branching killed Brownian motions with drift. To appear in \emph{Adv. in Appl. Probab.}
\bibitem{Tauberian} Korevaar, J.: Tauberian theory: a century of developments. Berlin: Springer, 2004.
\bibitem{kyprianou2014fluctuations} Kyprianou, A. E.: Fluctuations of L\'{e}vy processes with applications: Introductory Lectures. Springer Science \& Business Media, 2014.
\bibitem{lalleywalk} Lalley, S. P. and Shao, Y.: On the maximal displacement of critical branching random walk. \emph{Probab. Theory Relat. Fields.} \textbf{162}(1--2) (2015) 71--96.
\bibitem{lalley} Lalley, S. P. and Shao, Y.: Maximal displacement of critical branching symmetric stable processes. \emph{Ann. Inst. Henri Poincar\'{e} Probab. Stat. } \textbf{52} (3) (2016) 1161--1177.
\bibitem{neuman2017maximal} Neuman, E. and Zheng, X.-H. : On the maximal displacement of subcritical branching random walks. \emph{Probab. Theory Related Fields} \textbf{167} (3--4) (2017) 1137--1164.
\bibitem{profeta} Profeta, C.: Extreme values of critical and subcritical branching stable processes with positive jumps. \emph{ALEA, Lat. Am. J. Probab. Math. Stat.} \textbf{19}(2) (2022) 1421--1433.
\bibitem{profeta:hal-03736526} Profeta, C.: Maximal displacement of spectrally negative branching L\'{e}vy processes. \emph{Bernoulli }\textbf {30} (2)(2024) 961--982.
\bibitem{Profeta2025} Profeta, C.: On the maximal displacement of some critical branching L\'{e}vy processes with stable offspring distribution. arXiv: 2503.18451.
\bibitem{Ren_Song_Zhang_2023} Ren,Y.-X., Song, R. and Zhang, R.: Weak convergence of the extremes of branching L\'{e}vy processes with regularly varying tails. \emph{J. Appl. Probab.} \textbf {61} (2)(2024) 622--643.
\bibitem{Sato} Sato, K.: L\'{e}vy processes and infinitely divisible distributions. Cambridge Stud. Adv. Math., 68. Cambridge University Press, Cambridge, 1999.
\bibitem{Sawyer} Sawyer, S. and Fleischman, J.: Maximum geographic range of a mutant allele considered as a subtype of a Brownian branching random field. \emph{Proceedings of the National Academy of Sciences.} \textbf{76}(2) (1979) 872--875.
\bibitem{Zolotarev} Zolotarev,V.-M.: More exact statements of several theorems in the theory of branching processes. \emph{Teor. Veroyatnost. i Primenen.}
\textbf {2} (1957) 256--266.
\end{thebibliography}
\vskip 0.2truein
\vskip 0.2truein
\noindent{\bf Haojie Hou:} School of Mathematics and Statistics, Beijing Institute of Technology, Beijing 100081, P. R. China.
Email: {\texttt
houhaojie@bit.edu.cn}
\smallskip
\noindent{\bf Yiyang Jiang:} School of Mathematical Sciences, Peking
University, Beijing, 100871, P.R. China. Email: {\texttt
jyy.0916@stu.pku.edu.cn}
\smallskip
\noindent{\bf Yan-Xia Ren:} LMAM School of Mathematical Sciences \& Center for
Statistical Science, Peking
University, Beijing, 100871, P.R. China. Email: {\texttt
yxren@math.pku.edu.cn}
\smallskip
\noindent {\bf Renming Song:} Department of Mathematics,
University of Illinois Urbana-Champaign,
Urbana, IL 61801, U.S.A.
Email: {\texttt rsong@illinois.edu}
\end{document}