The flux of particles in a one-dimensional Fleming-Viot process
Éric Brunet, Bernard Derrida
TL;DR
This work analyzes the Fleming-Viot process with absorbing origin, showing that in the large-$N$ limit the particle density becomes deterministic and akin to a quasi-stationary distribution, while providing an exact finite-time solution on the positive integers for arbitrary initial data. It reveals a Fisher-KPP–like one-parameter family of steady states, a Bramson-type logarithmic shift in long-time behavior, and a pulled–pushed transition induced by a near-origin modification of the diffusion, extended to the continuum half-line. A cut-off approximation is employed to predict leading large-$N$ corrections to the absorbed-flux, yielding $(\ln N)^{-2}$ scaling in the pulled regime and power-law scaling in the pushed regime, with a detailed crossover in the critical regime. The results connect explicit solvability with front-formation phenomena, offering precise predictions for finite-size effects and potential extensions to simulations and large-deviation analyses. The findings have broad relevance for understanding interacting particle systems with absorption and for exploring genealogies and fluctuations in quasi-stationary settings.
Abstract
The Fleming-Viot process describes a system of $N$ particles diffusing on a graph with an absorbing site. Whenever one of the particles is absorbed, it is replaced by a new particle at the position of one of the $N-1$ remaining particles. Here we consider the case where the particles lie on the semi-infinite line with a biased diffusion towards the origin which is the absorbing site. In the large $N$ limit, the evolution of the density becomes deterministic and has a number of characteristics similar to the Fisher-KPP equation: a one-parameter family of steady state solutions, dependence of the long time asymptotics on the initial conditions, Bramson logarithmic shift, etc. One noticeable difference, however, is that in the Fleming-Viot case, the solution can be computed explicitly for arbitrary initial conditions and at an arbitrary time. By modifying the diffusion rule near the origin, one can produce a transition in the flux of absorbed particles, very similar to the pushed-pulled transition in travelling waves. Lastly, using a cut-off approximation (which is known to be correct in the theory of travelling waves), we derive a number of predictions for the leading large $N$ correction of the flux of absorbed particles.
