Error in equilibrium with pycalphad 0.10.2 about pycalphad HOT 4 CLOSED

As a workaround, you can try altering the sampling point density:
eq = equilibrium(dbf, ['FE','C','VA'], phases, {v.N:1, v.P:1e5, v.T:T, v.X('C'):0.0053}, verbose=True, calc_opts={'pdens': 500})

from pycalphad.

Hi @richardotis,

Thanks for having a look at this issue!
Yes sure, you can use the TDB in your test suite. (Just be aware that the current parameters are not the 'correct' ones to give the correct phase boundaries of the Fe-C system.)

Cheers,
Tobias

from pycalphad.

I'm able to reproduce this error, which is a bug. What's happening is that the minimizer is returning a non-physical solution of four stable phases, when only two are allowed by the Gibbs phase rule for the given conditions. (pycalphad will tolerate one extra stable phase (which is where the 3 comes from), to catch invariant reactions.)

The change in behavior is (probably) because of the bugfix in #428 which is causing the minimizer to return a different starting point than in 0.10.1. That fix has apparently uncovered another bug here. It should be impossible for the minimizer to return a solution that violates the Gibbs phase rule. I'm investigating it now.

from pycalphad.

@tobiasspt I've put together a solution to fix this issue in the minimizer. Is it alright to put the example TDB into the test suite so we can ensure this bug stays fixed?

from pycalphad.