The current approach for the vacancy test, and probably others (other point defects, surfaces?) is to read bulk_supercell.xyz from the test directory, rescale it to match the lattice constant of the corresponding bulk structure (indicated by its bulk_struct_test info field), then relax the atoms without the cell DOFs. This was originally conceived to make it possible for the user to do complicated things like impose a symmetry breaking (like the Si vacancy Jahn-Teller), or add some initial relaxation manually for interstitials.
However, this is incorrect for lower symmetry crystal lattices, because it does not relax the cell shape. For all intents and purposes the test must use a supercell constructed from the relaxed bulk cell (which can be extract from the bulk test). If additional manual adjustments must be made there needs to be some other mechanism to specify them.
The current approach for the vacancy test, and probably others (other point defects, surfaces?) is to read
bulk_supercell.xyzfrom the test directory, rescale it to match the lattice constant of the corresponding bulk structure (indicated by itsbulk_struct_testinfo field), then relax the atoms without the cell DOFs. This was originally conceived to make it possible for the user to do complicated things like impose a symmetry breaking (like the Si vacancy Jahn-Teller), or add some initial relaxation manually for interstitials.However, this is incorrect for lower symmetry crystal lattices, because it does not relax the cell shape. For all intents and purposes the test must use a supercell constructed from the relaxed bulk cell (which can be extract from the bulk test). If additional manual adjustments must be made there needs to be some other mechanism to specify them.