Quasilinear nonuniformly parabolic system modelling chemotaxis (Q865358)

The global existence of non-negative classical solutions \((u,v)\) to the parabolic system \[ \begin{aligned} \partial_t u & = \nabla\cdot (\alpha(u) \nabla u - u \beta(u) \nabla v) + f(u) \quad\text{in } (0,\infty)\times U,\\ \partial_t v & = \Delta v - v + g(u) \quad\text{in } (0,\infty)\times U, \end{aligned} \] is investigated when the diffusion coefficient \(\alpha\) is positive but vanishes at infinity, that is, \(\alpha(r)\to 0 \) as \(r\to\infty\). This system is connected to the Keller-Segel model for chemotaxis which corresponds to the choices \(\alpha(u)=g(u)=u\), \(\beta(u)=1\), and \(f=0\). Here, \(U\) is a bounded open subset of \({\mathbb R}^n\) with smooth boundary, \(f\) is a smooth function with linear growth and \(f(0)=0\), \(\alpha\) and \(\beta\) are smooth and bounded positive functions, \(g\) is a smooth and bounded non-negative function, and \((u,v)\) satisfy homogeneous Neumann boundary conditions. Assuming that either \(\beta/\alpha\) is bounded or that \(\alpha(r) r^{1/n}\to\infty\) as \(r\to\infty\), the existence of a global classical solution \((u,v)\) is shown. The main point is to prove that, despite the degeneracy of the diffusion for large values of \(u\), the function \(u\) stays bounded on any compact time interval \([0,T]\). To this end, de Giorgi's method is used in the first case and Moser's iteration technique in the second case. (There is a misprint in equation~(17) which should be \(\nu(\eta+\varepsilon)>C \eta^{-1/n}\)).