Two-phase entropy solutions of forward-backward parabolic problems with unstable phase (Q899689)

The author considers a two-phase problem of nonlinear diffusion. The equation is \(u_t=\phi(u)_{xx}\), the function \(\phi\) is non-monotone, so the problem is called ``forward-backward'' and is ill-posed. Also \(\phi\) is assumed to look like a cubic parabola: it has one minimum and one maximum and tends to \(\pm\infty\) at \(\pm\infty\). A piece-wise linear \(\phi\) is used as an example. The problem is spatially one-dimensional; initial distribution \(u_0\) is \(L^\infty\) with \(\phi(u_0)\) from \(H^1_{\mathrm{loc}}\). The problem is regularized by replacing \(\phi(u)\) by \(\phi(u)+\epsilon u_t\). One of the phases is assumed to be unstable. An entropy solution is constructed for a large class of initial data.