Computation of sharp phase boundaries by spreading: The planar and spherically symmetric cases (Q808670)

The purpose of this important paper is to study the problem of numerical computation of a moving boundary. The authors focus the attention mainly on (sharp) free boundaries which arise from phase transitions. The (sharp interface) free boundary problems are described as follows. For a region \(\Omega \subset R^ n\) and \(t\in R^+\) the problem is to determine a function u(x,t) and a surface \(\Gamma\) (t)\(\subset \Omega\) such that the following are satisfied: (1) \(u_ t=K\Delta u\) in \(\Omega \setminus \Gamma (t)\), (2) \(\ell v=-K[\nabla u]^+_ -\) on \(\Gamma\) (t), and (3) \(u=-\frac{\sigma}{\Delta s}k-\alpha \frac{\sigma}{\Delta s}v\) on \(\Gamma\) (t), where \(\ell\) is the latent heat, K the diffusivity, k the sum of principal curvatures at a point of the interface, \(\sigma\) the interfacial (or surface) tension and \(\alpha\) is a constant related to a relaxation time, all in dimensionless units, \([\nabla u]^+_ -\) is the jump in the gradient of u across \(\Gamma\) and v is the signed magnitude of the normal velocity, the quantity \(\Delta\) s is the entropy difference between the phases. Main result: Using the phase field model proposed by the first author as a tool for approximating these problems, the authors demonstrate that the sharp interface can be spread dramatically without significant error. In a test case for which an analytical solution is available a computed phase field interface easily approximates the exact one to within three to four digits.