no. 14 · Friday 13 November 2026
Metriplectic Lotka–Volterra systems with noise:interactions invisible at stationarity, visible in time
Abstract
We show that a Lotka–Volterra system with an interior equilibrium admits a metriplectic structure, with a Volterra function $V_C$ as energy and an entropy variable for the environment, if and only if Volterra's dissipativity condition $CA+A^{\top}C\le0$ holds, and we construct it explicitly with a log-canonical Poisson bivector. Adding noise through the fluctuation–dissipation relation of stochastic GENERIC yields multiplicative environmental noise, with covariance fixed by the symmetric part of $AC^{-1}$, whose stationary law is an explicit Gibbs measure: a product of independent Gamma distributions. Unlike known Gibbs measures for stochastic Lotka–Volterra systems, this holds for nonsymmetric interactions, predator–prey included. The stationary law determines only the equilibrium and $C/k$: every interaction matrix in an open cone of dimension $n^2$, in particular any Poisson part, produces with its compatible noise the same law as uncoupled stochastic logistic equations. Gamma-distributed, uncorrelated abundance fluctuations, as observed in microbial communities, are therefore no evidence of absence of interactions. The interactions are, however, visible in time: the zero-lag derivative of the log-abundance covariance equals $A(C/k)^{-1}$, and its antisymmetric part is the Poisson part, which carries all the stationary entropy production, computed in closed form. Finally, for general noise we prove an exact Lyapunov-type identity $AK+K^\top A^\top+Q=0$ for $K=\mathrm{Cov}(x,\log x)$ at any stationary solution. It implies that the populations are independent if and only if the noise is metriplectically compatible, and gives an explicit lower bound on interspecies cross-covariances in terms of the distance of $Q$ to the compatible family. Simulations of a predator–prey model show that the bound is nearly sharp.
- fields
- Mathematical Physics