Charged phases that cannot charge balance raise errors

[bocklund]

Discussed in #502

^{Originally posted by zliang1990 December 13, 2023}
Hi Richard (@richardotis),

My name is Zhi Liang, Materials Design Engineer in QuesTek. When I recently attempted to run the attached V-O description (GitHub does not support .TDB so I zipped it) with Pycalphad, I noticed that the following error message will pop up if I only choose oxygen (and vacancy) as elements.

Upon debugging, it seems to do with the HALITE phase, whose sublattice model is V,V+2,V+3,VA : O-2,VA. When only choosing oxygen, first sublattice only allows VA while second sublattice allows O-2 and VA, which seemingly triggers the following error message. This issue can be a blocker if in a multicomponent system, certain phase only has sublattice model A,VA : O-2,VA, but the element selection only has element B and oxygen, and this error message will pop up as well (Sorry that I cannot provide our multicomponent database since it is proprietary).

Could you please help to see how we can resolve this?

Thanks!

Best,
Zhi

15YANG_V-O description.zip

[bocklund]

From #502:

What's happening here is that when only vacancies and oxygen are entered as elements PyCalphad is treating CORUNDUM and HALITE as active phases (because there are active species in each sublattice).

For both phases, it's impossible to charge balance given these components (pure vacancy endmembers don't count, such as in the case for HALITE) and that gives array size errors because there are no feasible points. I'm creating an issue for this, since it's a bug. We should probably prevent phases that cannot charge balance from being included in the active set of phases.

[richardotis]

This should be easy to fix with the new sampler, as it only samples the feasible subspace of points. For these cases it would find no solution, so it would be easy to raise an error (or return zero points) so the phase is automatically excluded from the active set.

[richardotis]

This is now fixed in #432