A fitness-driven cross-diffusion system from population dynamics as a gradient flow (Q294216)

The existence of non-negative weak solutions is established to the parabolic system: \(\partial_t u_i = -\mathrm{div} ( u_i \nabla f_i)+u_if_i\) defined in \(\Omega\times ]0,\infty[\subset \mathbb{R}^{n+1}\), accomplished with the boundary conditions \(u_i\nabla f_i \cdot \nu=0\), under initial conditions \(u_i^0\geq 0\) for \(i=1,\cdots, N\in\mathbb{N}\). Here, \(\mathbf{f}=\mathbf{m}- A \mathbf{u}\) represents the fitness of \(N\) interacting populations. The authors prove that the population densities \(\mathbf{u}\) satisfy the entropy-dissipation inequality \(\mathcal{E}(\mathbf{u}(t_1)) +\sum_{i}\int_{t_0}^{t_1} \int_{\Omega} u_i (|\nabla f_i|^2+ |f_i|^2)\leq \mathcal{E}(\mathbf{u}(t_0))\), for a.e. \(0\leq t_0\leq t_1\). The proof of existence of weak solutions involves three sucessive approximations, indexed by \(\varepsilon\rightarrow 0\), \(M\rightarrow\infty\) and \(\delta\rightarrow 0\), via the initial-boundary value problem to the regularized system: \(\partial_t u_i +\varepsilon (\mathcal{A} \mathbf{u})_i=- \mathrm{ div} ( \tilde u_i \nabla f_i)+ \tilde u_if_i + \delta\Delta u_i \), where \(\mathcal{A}\) is a suitable elliptic operator of higher order and \(\tilde u=\max (0,\min (M,u))\). Under additional structural conditions on \(A\) and \(\mathbf{m}\), a long-time convergence of \(\mathbf{u}(t)\) to the ideal free distribution with survival of all the species, \(\mathbf{u}^\infty: = A^{-1} \mathbf{m}\), is obtained.