arXiv2020
We study the dynamics of random walks hopping on homogeneous hyper-cubic lattices and multiplying at a fertile site. In one and two dimensions, the total number $\mathcal{N}(t)$ of walkers grows exponentially at a Malthusian rate depending on the dimensionality and the multiplication rate $μ$ at the fertile site. When $d>d_c=2$, the number of walkers may remain finite forever for any $μ$; it surely remains finite when $μ\leq μ_d$. We determine $μ_d$ and show that $\langle\mathcal{N}(t)\rangle$ grows exponentially if $μ>μ_d$. The distribution of the total number of walkers remains broad when $d\leq 2$, and also when $d>2$ and $μ>μ_d$. We compute $\langle \mathcal{N}^m\rangle$ explicitly for small $m$, and show how to determine higher moments. In the critical regime, $\langle \mathcal{N}\rangle$ grows as $\sqrt{t}$ for $d=3$, $t/\ln t$ for $d=4$, and $t$ for $d>4$. Higher moments grow anomalously, $\langle \mathcal{N}^m\rangle\sim \langle \mathcal{N}\rangle^{2m-1}$, in the critical regime; the growth is normal, $\langle \mathcal{N}^m\rangle\sim \langle \mathcal{N}\rangle^{m}$, in the exponential phase. The distribution of the number of walkers in the critical regime is asymptotically stationary and universal, viz. it is independent of the spatial dimension. Interactions between walkers may drastically change the behavior. For random walks with exclusion, if $d>2$, there is again a critical multiplication rate, above which $\langle\mathcal{N}(t)\rangle$ grows linearly (not exponentially) in time; when $d\leq d_c=2$, the leading behavior is independent on $μ$ and $\langle\mathcal{N}(t)\rangle$ exhibits a sub-linear growth.