For the phase change process, non-physical “jump” in the predicted properties across the interfaces, separating the single and the two-phase regions, was reported.
In phase change problems the physical properties of the phase-changing medium change significantly. If I understand your question correctly, what you refer to as a "jump in the diffusion coefficient" is most likely caused by the latent heat. However, not the diffusion coefficient itself, but the physical properties like specific heat capacity, density and thermal conductivity change significantly during phase change - as a result the diffusion coefficient changes. There are approximations for the specific heat capacity, like the apparent heat capacity method, for example. One of my papers explains this in more detail:
Many Thanks Christoph for your answer. Actually, I was developed a new general smoothing algorithm in order to deal with the discontinuities in the effective diffusion coefficient across the interfaces (in all my publications was explained this method in details) . This method was successful for avoiding the jumps in the predicted temperature distribution. Now, I am looking forward whether if there is another new methods that can can be used for the same problems.