A Galerkin method for mixed parabolic-elliptic partial differential equations (Q707579)

The author considers, for \(x\in \mathbb R^n\), a problem which is parabolic in a strictly inner part of the solution domain, and elliptic (along with homogeneous boundary values) else. Referring often to \textit{R. Dautray} and \textit{J.-L. Lions} [Mathematical analysis and numerical methods for science and technology. Vol. 5. Berlin: Springer (2000; Zbl 0956.35003)], he proves uniqueness of the weak solution of the problem, regularizes it by introducing a penalty term to make it parabolic everywhere (the penalty parameter not affecting the ellipticity of the resulting operator, only its continuity constant), and defines a fully discrete Galerkin-Euler-backward discretization. For this, uniqueness and a-priori estimates are shown, finally, error estimates are obtained. For the a-priori estimates, some additional regularity is needed which is proved to be present in the two-dimensional case when the domain ha no slits. The error estimates practically mean that a choice of the time step as the square of the maximal spatial step is optimal.