Piotr Dyszewski, Samuel G. G. Johnston, Sandra Palau, Joscha Prochno
{"title":"The largest fragment in self-similar fragmentation processes of positive index","authors":"Piotr Dyszewski, Samuel G. G. Johnston, Sandra Palau, Joscha Prochno","doi":"arxiv-2409.11795","DOIUrl":null,"url":null,"abstract":"Take a self-similar fragmentation process with dislocation measure $\\nu$ and\nindex of self-similarity $\\alpha > 0$. Let $e^{-m_t}$ denote the size of the\nlargest fragment in the system at time $t\\geq 0$. We prove fine results for the\nasymptotics of the stochastic process $(s_{t \\geq 0}$ for a broad class of\ndislocation measures. In the case where the process has finite activity (i.e.\\\n$\\nu$ is a finite measure with total mass $\\lambda>0$), we show that setting\n\\begin{equation*} g(t) :=\\frac{1}{\\alpha}\\left(\\log t - \\log \\log t +\n\\log(\\alpha \\lambda)\\right), \\qquad t\\geq 0, \\end{equation*} we have $\\lim_{t\n\\to \\infty} (m_t - g(t)) = 0$ almost-surely. In the case where the process has\ninfinite activity, we impose the mild regularity condition that the dislocation\nmeasure satisfies \\begin{equation*} \\nu(1-s_1 > \\delta ) = \\delta^{-\\theta}\n\\ell(1/\\delta), \\end{equation*} for some $\\theta \\in (0,1)$ and\n$\\ell:(0,\\infty) \\to (0,\\infty)$ slowly varying at infinity. Under this\nregularity condition, we find that if \\begin{equation*} g(t)\n:=\\frac{1}{\\alpha}\\left( \\log t - (1-\\theta) \\log \\log t - \\log \\ell \\left(\n\\log t ~\\ell\\left( \\log t \\right)^{\\frac{1}{1-\\theta}} \\right) +\nc(\\alpha,\\theta) \\right), \\qquad t\\geq 0, \\end{equation*} then $\\lim_{t \\to\n\\infty} (m_t - g(t)) = 0$ almost-surely. Here $c(\\alpha,\\theta) := \\log \\alpha\n-(1-\\theta)\\log(1-\\theta) - \\log \\Gamma(1-\\theta)$. Our results sharpen\nsignificantly the best prior result on general self-similar fragmentation\nprocesses, due to Bertoin, which states that $m_t = (1+o(1)) \\frac{1}{\\alpha}\n\\log t$.","PeriodicalId":501245,"journal":{"name":"arXiv - MATH - Probability","volume":"3 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - MATH - Probability","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.11795","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Take a self-similar fragmentation process with dislocation measure $\nu$ and
index of self-similarity $\alpha > 0$. Let $e^{-m_t}$ denote the size of the
largest fragment in the system at time $t\geq 0$. We prove fine results for the
asymptotics of the stochastic process $(s_{t \geq 0}$ for a broad class of
dislocation measures. In the case where the process has finite activity (i.e.\
$\nu$ is a finite measure with total mass $\lambda>0$), we show that setting
\begin{equation*} g(t) :=\frac{1}{\alpha}\left(\log t - \log \log t +
\log(\alpha \lambda)\right), \qquad t\geq 0, \end{equation*} we have $\lim_{t
\to \infty} (m_t - g(t)) = 0$ almost-surely. In the case where the process has
infinite activity, we impose the mild regularity condition that the dislocation
measure satisfies \begin{equation*} \nu(1-s_1 > \delta ) = \delta^{-\theta}
\ell(1/\delta), \end{equation*} for some $\theta \in (0,1)$ and
$\ell:(0,\infty) \to (0,\infty)$ slowly varying at infinity. Under this
regularity condition, we find that if \begin{equation*} g(t)
:=\frac{1}{\alpha}\left( \log t - (1-\theta) \log \log t - \log \ell \left(
\log t ~\ell\left( \log t \right)^{\frac{1}{1-\theta}} \right) +
c(\alpha,\theta) \right), \qquad t\geq 0, \end{equation*} then $\lim_{t \to
\infty} (m_t - g(t)) = 0$ almost-surely. Here $c(\alpha,\theta) := \log \alpha
-(1-\theta)\log(1-\theta) - \log \Gamma(1-\theta)$. Our results sharpen
significantly the best prior result on general self-similar fragmentation
processes, due to Bertoin, which states that $m_t = (1+o(1)) \frac{1}{\alpha}
\log t$.