Log-normal distribution in growing systems with weighted multiplicative interactions

Many-body stochastic processes with weighted multiplicative interactions are investigated analytically and numerically. An interaction rate between particles with quantities $x, y$ is controlled by a homogeneous symmetric kernel $K(x, y) \propto x^{w} y^{w}$ with a weight parameter $w$. When $w<0$, a method of moment inequalities is used to derive log-normal type tails in probability distribution functions. The variance of log-normal distributions is expressed in terms of the weight $w$ and interaction parameters. When interactions are weak and a growth rate of systems is small, in particular, the variance is in proportion to the growth rate. This behavior is totally different from that of one-body stochastic processes, where the variance is independent of the growth rate. At $w>0$, Monte Carlo simulations show that the processes end up with a winner-take-all state.