Thank you for your reply again! I understand that stafanmelting2d is a benchmark case that can be compared with non-LBM approaches to demonstrate the capabilities of the current LBM method, and the results match very well.
However, when I developed another self-coded program concerning phase change and used the harmonic mean value for “cp_ref,” the simulation diverged. On the other hand, when I used “cp_s” for “cp_ref,” the simulation converged. Interestingly, the simulation diverged regardless of whether “cp_ref” was larger or smaller than “cp_s.”
Therefore, if “cp_ref” affects the accuracy, I may need to conduct a verification experiment. This is why I want to figure this out.