Feasibility of sparse large Lotka-Volterra ecosystems (Q2105799)

This paper discusses the feasibility of sparse large Lotka-Volterra ecosystems modeled by the ODE system \[ \frac{d\mathbf{x}_n}{dt}=\mathbf{x}_n(\mathbf{1}_n-x_n+M_n\mathbf{x}_n). \] The interaction matrix \(M_n\) is a sparse random matrix of the form \(M_n=\frac{1}{\alpha_n\sqrt{d_n}}\Delta_n\circ A_n\), where \(\Delta_n\) is a patterned adjacency matrix of a directed \(d_n\)-regular graph with \(n\) vertices and \(A_n\) is a \(n\times n\) matrix with independent Gaussian \(\mathcal{N}(0,1)\) entries, \(\circ\) being the Hadamard matrix product. Also, \((\alpha_n)_{n\geq 1}\) is a positive sequence that can be tuned to obtain a feasible solution, while \(d_n\) parametrizes the connectivity of the food web. It is shown, via the large random matrix theory for sparse matrices and Gaussian concentration of measures that for a given range of \(d_n\) and a certain block structure of \(\Delta_n\) there is a threshold guaranteering the existence of a positive equilibrium as \(n\) gets large. Results regarding the global stability of the said positive equilibrium are also provided, along with information about convergence rates.