A Gaussian particle distribution for branching Brownian motion with an inhomogeneous branching rate (Q2243930)

A branching Brownian motion with inhomogeneous branching rate is a particle system on the real line in which each particle move independently according to Brownian motions with drift. Additionally, each particle may either split into two children or die, with the splitting rate of a particle at position \(x\) given by \(b(x)\) and the death rate by \(d(x)\). In this model, it is assumed that \(b(x)-d(x) = \beta x\) for some \(\beta > 0\), i.e. that the average reproduction rate of a particle grows linearly with \(x\). This particle system models the evolution of a population with a fitness playing a role on the reproduction of the individuals. Namely, a particle at position \(x\) represents an individual with fitness \(x\), which it transmits to its children. The question of the rate of increase of the fitness of a population is an important problem in evolutionary biology. It has been observed in the physics and biology literature that in a large population, the fitness distribution in the population tends to be Gaussian and to grow linearly on average. In other words, the fitness of a population evolves as a Gaussian travelling wave over time. The present article gives the first rigorous analysis of a mathematical model with the apparition of a Gaussian travelling wave as the fitness distribution. More precisely, consider a family of branching Brownian motions with inhomogeneous branching rate such that the drift \(-\rho\) and \(\beta\) both converge to \(0\) with \(\rho^3 \ll \beta\). Then, provided some assumptions on the initial distribution of this family of branching Brownian motions, the distribution of particles remains, on a large time-scale close to a Gaussian distribution. The distribution of the most fit particles is also studied. Under a well-chosen normalization, it is proved to converge to the probability distribution \(\nu\) on \((-\infty,0)\) with density proportional to \(\mathrm{Ai}(\gamma_1 - y)\), with \(\mathrm{Ai}\) the Airy function of the first kind, and \(\gamma_1\) its largest \(0\).