Blow-up of solutions of parabolic equations with nonlinear memory (Q1105142)

This paper deals with the initial-boundary value problem \[ u_ t=\Delta u+\int^{t}_{0}m(t-\tau)f(u(x,\tau))d\tau +g(x) \] in \(Q_{\infty}\), \(u(x,0)=u_ 0(x)\) in \(\Omega\), \(u(x,t)=0\) in \(\Gamma_{\infty}\); here \(\Omega\) is a bounded domain in \(R^ n\) with \(C^ 2\) boundary, \(Q_ t=\Omega \times (0,t)\), and \(\Gamma_ t=\partial \Omega \times (0,t)\), where the functions f, m, g, and \(u_ 0\) satisfy certain smoothness and monotonicity conditions. Under these conditions and specifically \(\int^{\infty}_{0}(F(s))^{-}ds<\infty\), where F is an antiderivative of f, the solution u blows up in finite time. This paper aims at characterizing the blow-up set; i.e., the set \(\{\) \(x\in \Omega:\exists (x_ n,t_ n)\to (x,T)\ni u(x_ n,t_ n)\to \infty \}\). In much of this paper \(f(u)=(u+\lambda)^ p\), \(p>1\), \(\lambda >0\). A basic method for dealing with the delay aspect of this equation is to consider the t-derivative of both sides and to use rather strong monotone properties of m'(t) in the hypotheses. [A number of misprints was noted by the reviewer, which contributed to difficulties he experienced following details of some of the proofs.]