Spurious void generation in solidifying flow
I've found a small reproducer for the anomalous void generation issue that we've been investigating in the Campbell Al casting benchmark problem. Using the same material properties and problem dimensions (of the plate) it is setup like the natural convection problem with a hot left side and cold right side, totally filled with liquid, and with temperatures that keep things above the liquidus. I run the problem for a while to generate a rotating flow in the domain. I then generate a restart file where I put a bit of void in the lower-right corner cell (the coldest corner) and restart with an added htc BC on the bottom to start to freeze the bottom. Here the only void remains in the corner cell but it increases significantly in volume fraction over time.
Here are the input files: add-solid-void.py