Existence and regularity for a system of porous medium equations with small cross-diffusion and nonlocal drifts (Q2157312)

This paper investigates a degenerate parabolic drift-diffusion model for a mixture of \(M \geq 1\) components of the form \[ \partial_t u_i = \operatorname{div}\Big(u_i (\nabla u_i + \nabla L_i(x,t,u_i) + \delta F_i(x,t, u, \nabla u)) \Big) \,\, \text{ for } i = 1,\ldots,M ,\tag{1} \] where \(u = (u_1, \ldots,u_M)\) are the densities of the species. The problem is investigated in cylinder domains \((0,T) \times \Omega\) with elements \((t,x)\), where the time \(T >0\) is finite and \(\Omega \subset \mathbb{R}^d\) is bounded. The non-local drift-terms \(L\) is given by a spacial convolution of the form \[ L_i(x,t, u_i) = V_i(x,t) + (W_i \star u_i(\cdot,t))(x) , \] in which the smooth functions \(V_i\) and the smooth convolution kernel \(W_i\) are fixed. The cross-diffusion term \(F\) is affine in the gradient variable and of the special form \[ F_i(t,x, u, p) = G_i^0(t,x, u) + \sum_{j=1}^M \sum_{\ell = 1}^d G^1_{ij\ell}(t,x, u) p_{j\ell} , \] in which the data \(G^0\) and \(G^1\) are smooth in all variables and globally bounded on the respective domain, and \(p \in \mathbb{R}^{M\times d}\) is a place-holder for \(\nabla u\). The paper studies the initial-boundary value problem generated by the PDEs (1), vanishing fluxes on \((0,T) \times \partial \Omega\), and initial conditions \(u_i(\cdot,0) = u^0_i\) in \(\Omega\), with \(u^0\) non-negative in \((L^p(\Omega))^M\), \(p > 1\). The main result asserts for fixed \(T\) the existence of \(\delta_0 > 0\) depending on \(\Omega\), \(T\) and the data \(L,F\) in the respective norms such that under the condition \(\delta < \delta_0\), a weak solution to the boundary-value-problem exists globally in the class \[ u \in L^2(0,T; H^1(\Omega))^M \,\, \text{ subject to } \partial_t u \in L^{r^{\prime}}(0,T; (W^{1,r}(\Omega) )^{\prime})^M, \, r = 2(d+1) . \] Certain numerical illustrations are also given. As a remark of the reviewer, it seems extremely restrictive that the critical parameter \(\delta\) be inverse proportional to the time \(T\) and the diameter of the domain \(\Omega\). This is probably not necessary, in view of the inequality (33) and the assumption (5).