Table of Contents
Fetching ...

Decompositions of CSBPs via Poissonian Intertwining

Clément Foucart, Olivier Hénard

TL;DR

This work unifies skeletal decompositions of CSBPs within an intertwining framework. It constructs a two-type $\mathbb{R}_+\times\mathbb{Z}_+$-valued process $(X^\lambda,L^\lambda)$ whose first coordinate preserves the CSBP law while the second is Poisson-poissonized with parameter $\lambda X^\lambda_t$, linked to the CSBP generator via a Poisson kernel $K$. By analyzing the position of $\lambda$ relative to $\rho$, the authors recover classical skeleton results for $\lambda\ge \rho$ and reveal a novel birth-immigration mechanism when $\lambda<\rho$; they also prove explosion occurs simultaneously along the skeleton and the CSBP when killing is absent. Furthermore, they establish a scaling limit: as $\lambda\to\infty$, the rescaled skeleton $L^\lambda/\lambda$ converges to the original CSBP in the Skorokhod sense. Collectively, these results provide a unified, operator-based perspective on skeletal decompositions and genealogies in CSBPs, with implications for branching structure and scaling behavior.

Abstract

We revisit certain decompositions of continuous-state branching processes (CSBPs), commonly referred to as skeletal decompositions, through the lens of intertwining of semi-groups. Precisely, we associate to a CSBP $X$ with branching mechanism $ψ$ a family of $\mathbb{R}_+\times \mathbb{Z}_+$-valued branching processes $(X^λ, L^λ)$, indexed by a parameter $λ\in (0, \infty)$, that satisfies an intertwining relationship with $X$ through the Poisson kernel with parameter $λx$. The continuous component $X^λ$ has the same law as $X$, while the discrete component $L^λ$, conditionally on $X^λ_t$, has a Poisson distribution with parameter $λX^λ_t$. The law of $(X^λ, L^λ)$ depends on the position of $λ$ within $[0, \infty) = [0, ρ) \cup [ρ, \infty)$, where $ρ$ is the largest positive root of $ψ$. When $λ\geq ρ$, various well-known results concerning skeleton decompositions are recovered. In the supercritical case ($ρ> 0$), when $λ<ρ$, a novel phenomenon arises: a birth term appears in the skeleton, corresponding to a one-unit proportional immigration from the continuous to the discrete component. Along the way, the class of continuous-time branching processes taking values in $\mathbb{R}_+ \times \mathbb{Z}_+$ is constructed.

Decompositions of CSBPs via Poissonian Intertwining

TL;DR

This work unifies skeletal decompositions of CSBPs within an intertwining framework. It constructs a two-type -valued process whose first coordinate preserves the CSBP law while the second is Poisson-poissonized with parameter , linked to the CSBP generator via a Poisson kernel . By analyzing the position of relative to , the authors recover classical skeleton results for and reveal a novel birth-immigration mechanism when ; they also prove explosion occurs simultaneously along the skeleton and the CSBP when killing is absent. Furthermore, they establish a scaling limit: as , the rescaled skeleton converges to the original CSBP in the Skorokhod sense. Collectively, these results provide a unified, operator-based perspective on skeletal decompositions and genealogies in CSBPs, with implications for branching structure and scaling behavior.

Abstract

We revisit certain decompositions of continuous-state branching processes (CSBPs), commonly referred to as skeletal decompositions, through the lens of intertwining of semi-groups. Precisely, we associate to a CSBP with branching mechanism a family of -valued branching processes , indexed by a parameter , that satisfies an intertwining relationship with through the Poisson kernel with parameter . The continuous component has the same law as , while the discrete component , conditionally on , has a Poisson distribution with parameter . The law of depends on the position of within , where is the largest positive root of . When , various well-known results concerning skeleton decompositions are recovered. In the supercritical case (), when , a novel phenomenon arises: a birth term appears in the skeleton, corresponding to a one-unit proportional immigration from the continuous to the discrete component. Along the way, the class of continuous-time branching processes taking values in is constructed.

Paper Structure

This paper contains 17 sections, 20 theorems, 151 equations, 2 figures.

Key Result

Theorem 2.1

For any admissible parameters $(\gamma,b,\sigma,\kappa, \mathrm{k},\pi, \rho)$, there exists a unique $E^{\Delta}$-valued càdlàg strong Markov process $\mathbf{X}=(X,L)$, with cemetery state $\Delta$, solution to the martingale problem Moreover, the semigroup of $\mathbf{X}$ satisfies for any $(x,n)$, $t\geq 0$ and $(q,r)\in (0,\infty)\times (0,1)$ with $t\mapsto F_t(q,r):=\left( u_t(q,r), f_t(q,

Figures (2)

  • Figure 1: A supercritical mechanism and its Esscher transforms
  • Figure 2: Schematic representation of the intertwined two-type branching process

Theorems & Definitions (48)

  • Theorem 2.1
  • Remark 2.2
  • Proposition 2.3
  • Remark 2.4
  • Proposition 2.5: Explosion and autonomous coordinates
  • Theorem 3.1: Algebraic intertwining
  • Corollary 3.2: Joint branching mechanism
  • Remark 3.3
  • Theorem 3.4: Intertwined semigroups and skeleton decomposition
  • Remark 3.5: Conditional law of $X^\lambda$ given $L^{\lambda}$
  • ...and 38 more