Paperlane

no. 15 · Monday 16 November 2026

Domains of attraction of quasi-stationary distributions for an absorbed random walk

Phil. Pollett

math.PR

Abstract

We study a continuous-time random walk on the positive integers, killed at zero, where the downward rate $a$ and upward rate $b$ satisfy $a>b>0$. Its quasi-stationary distributions form a one-parameter family. We show that if the initial distribution $(m_n)_{n\geq 1}$ is strictly positive and satisfies $m_{n+1}/m_n\to r$, where $(b/a)^{1/2}<r<1$, then its conditional distribution given survival converges in total variation to the corresponding nonminimal quasi-stationary distribution. We obtain asymptotic equivalents for the joint probabilities of position and survival, and for the survival probability itself. We also give a moment condition under which the minimal quasi-stationary distribution is selected. Finally, we construct an initial distribution whose successive-term ratios do not converge but whose conditional distribution nevertheless converges to a nonminimal quasi-stationary distribution. Thus the ratio condition is sufficient but not necessary.

fields
Probability
pdf ↗ page arXiv:2611.17791 ↗