The Boundary Principle of a Single Big Jump: Refined Asymptotics for Branching Processes with Immigration

Yunfan Zhao · 2025

Read The Boundary Principle of a Single Big Jump: Refined Asymptotics for Branching Processes with Immigration by Yunfan Zhao (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-05650.txt.

Introduction

A moving surface that drifts at constant speed, called a translator, cannot end like an upright wall unless it is one, Yunfan Zhao proves in a solo 2025 geometry paper. Zhao studies complete translators set in Euclidean space of dimension at least four, with finite topology. Their ends are assumed to approach upright hyperplanes in a smooth-once way. The only such translator, the theorem states, is the vertical hyperplane itself.

Catalog record

CREATORYunfan Zhao
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-05650

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The Boundary Principle of a Single Big Jump: Refined Asymptotics for Branching Processes with Immigration Yunfan Zhao Abstract We study a fixed-point equation arising in branching processes with state-independent immigration, where both immigration and offspring distributions exhibit heavy tails with boundary index one. Under the assumptions P ⁡ ( A > x ) ∼ ( 1 + x ) − 1 , P ⁡ ( B > x ) = L ⁡ ( x ) 1 + x , L ⁡ ( x ) ∼ ( log ⁡ x ) − 1 − ε , P(A>x)\sim(1+x)^{-1},\quad P(B>x)=\frac{L(x)}{1+x},\quad L(x)\sim(\log x)^{-1-\varepsilon}, we establish a boundary version of the principle of a single big jump . Our main result shows that the stationary solution X X satisfies P ⁡ ( X > x ) ∼ 1 ( 1 − b ) ​ ( 1 + x ) , x → ∞ , P(X>x)\sim\frac{1}{(1-b)(1+x)},\quad x\to\infty, with an explicit two-scale refinement that identifies logarithmic corrections of order L ⁡ ( x ) ​ log ⁡ x 1 + x \tfrac{L(x)\log x}{1+x} , which are asymptotically negligible. The proof combines tools from subexponential distribution theory, cluster expansions of the fixed-point solution, and generation-level tail asymptotics. In particular, we develop a countable closure principle for subexponential summations and a structural cluster expansion representation of X X , enabling us to disentangle immigration-driven and branching-driven extremal events. This approach yields a transparent probabilistic decomposition of the stationary tail, extending and refining earlier work on heavy-tailed fixed-point problems. 1 Introduction 1.1 Motivation and Literature Review Heavy-tailed phenomena arise naturally in diverse contexts ranging from insurance and risk management to network traffic, epidemics, and financial systems. In such systems, extreme events are not merely rare anomalies but often dominate the long-term behavior. Classical branching models, dating back to Galton and Watson, provide a natural framework for studying growth and extinction, yet their asymptotic analysis under heavy-tailed inputs remains subtle. The Principle of a Single Big Jump (PSBJ) captures the idea that the tail of a distribution is typically governed by a single extreme contribution. Understanding how this principle operates in recursive fixed-point equations is essential for both theoretical development and practical applications. [ 1 ] investigated the existence and properties of invariant measures for simple branching processes, providing a mathematical characterization of conditions under which such measures arise. The study established that invariant measures can be identified through limit distributions associated with normalized population sizes, which shed light on the long-term behavior of branching systems. In particular, Seneta demonstrated how these invariant measures depend on the criticality of the process, distinguishing between subcritical, critical, and supercritical regimes. His results extended the understanding of stationary and quasi-stationary distributions in branching processes, offering rigorous tools for analyzing persistence and extinction phenomena. This work provided a foundational contribution to the probabilistic theory of branching processes, influencing subsequent research in stochastic population models and applied probability. [ 2 ] established limit theorems for Galton–Watson branching processes with time-dependent immigration, extending classical results to more general stochastic population models. Their analysis demonstrated how the long-term distributional behavior of the process is shaped by variations in immigration rates over time. They proved conditions under which the process converges to non-degenerate limiting distributions, clarifying the roles of both reproduction mechanisms and immigration dynamics. Importantly, the study highlighted how time-varying immigration can stabilize or destabilize population trajectories, thereby enriching the theoretical framework for branching processes with external inputs. [ 3 ] introduced a generalization of slowly varying functions, expanding the classical framework of regular variation. Their work defined a broader class of functions that retain essential asymptotic properties while relaxing the strict conditions of Karamata’s theory. This generalization enabled a more flexible analysis of asymptotic growth rates, particularly in settings where classical slow variation proved too restrictive. The authors demonstrated that the extended class preserves key limiting behaviors necessary for applications in probability theory and analysis, thereby providing a stronger foundation for studying heavy-tailed distributions and branching process asymptotics. [ 7 ] investigated the behavior of the maximum of a random walk with long-tailed increments and negative drift over a random time horizon. They showed that, despite the negative drift ensuring that the walk tends to decrease over time, the heavy-tailed nature of increments can produce large upward excursions, making the distribution of the maximum essentially governed by the tail of the increment distribution. The authors derived precise asymptotics for the maximum under various random time intervals, including geometric stopping times, thereby extending classical results from ruin theory and queueing models. Their work highlighted that in systems with heavy-tailed risks, rare but large jumps dominate maximum behavior, which has direct implications for applied probability areas such as insurance, finance, and queueing networks. [ 8 ] analyzed the probability that a random walk with heavy-tailed increments exceeds a high boundary during a random time interval. They established precise asymptotic results for this exceedance probability, showing that rare but very large jumps dominate the event of boundary crossing, even in the presence of negative drift. By considering a variety of random time horizons, they extended earlier work on maxima of heavy-tailed random walks and highlighted the robustness of the “big jump principle” in determining extremal behavior. Their findings are particularly relevant in applied settings such as insurance risk, finance, and queueing theory, where extreme events over random horizons play a critical role in system performance and stability. [ 9 ] developed an implicit renewal theory framework to study the emergence of power-law tails in distributions defined on random trees. Their analysis established general conditions under which recursive distributional equations, often arising in branching processes and random tree models, yield heavy-tailed solutions. By extending renewal-theoretic techniques to a multivariate and tree-based setting, the authors provided asymptotic characterizations that unify and generalize earlier results in branching random walks, random recursions, and network models. This work demonstrated how power-law behavior naturally arises in hierarchical stochastic structures, offering broad applications to areas such as algorithms on random trees, complex networks, and heavy-tailed phenomena in applied probability. In recent years, [ 10 ] studied the distribution of customer sojourn times in a GI/GI/1 feedback queue when service times exhibit heavy-tailed behavior. They demonstrated that feedback significantly amplifies the impact of heavy tails, leading to extremely long sojourn times dominated by rare but very large service requirements. Using probabilistic and renewal-theoretic methods, they derived asymptotic characterizations showing that the tail of the sojourn-time distribution is heavier than that of the service times themselves. This work highlights the critical role of feedback mechanisms in queueing systems with heavy-tailed inputs, with implications for performance analysis in communication networks, computer systems, and service operations where feedback loops and bursty workloads are common. [ 11 ] investigated fixed-point equations arising in single- and multi-class branching processes and queueing systems, focusing on conditions that lead to solutions with regularly varying tails. They established general results showing how heavy-tailed behavior naturally emerges in such stochastic fixed-point problems, unifying phenomena observed in branching structures, recursive distributional equations, and service systems. Their analysis demonstrated that regular variation is a robust property in these models, providing sharp asymptotics for tail probabilities that are crucial for understanding extremes in complex stochastic networks. [ 12 ] examined

Read the complete text