A Problem about Phase Missing

[MWW726]

I want to know How to solve the problem of phase missing in low or some middle temperature.

[bocklund]

For low temperatures, pycalphad will not automatically extrapolate outside of the Piecewise temperature limits like you may be used to in other software. That is, if the TDB defines

FUNCTION GHSERAL  298.15 <expression>; 2000 N !

the Gibbs energy would be strictly defined between 298.15 K and 2000 K and the Gibbs energy will be zero otherwise. Please double check that all the temperatures are within the temperature limits of the TDB.

[MWW726]

Thank you. Got it. But, I still can`t understand phase missing in 500K while no phase missing in 450K and 550K. In fact, I found phase suddenly miss sometimes in Piecewise temperature limits. do you have some suggestions?

[bocklund]

Can you share the database file? I can try to reproduce the issue and find a solution.

[MWW726]

al-cu.txt

[bocklund]

A small update here. I am able to reproduce the issue. At this composition, ALCU_ETA becomes the only stable phase and there's a discontinuity in the chemical potentials at this point, which causes pycalphad to get stuck in a loop of trying to add ALCU_THETA, ALCU_ZETA, ALCU_DELTA phase, depending on the chemical potentials/driving forces.

This is a bug in the minimizer that may have been introduced when we changed some solver internals. Depending on your application, a possible workaround is to nudge the composition to be slightly Cu-rich or Cu-poor, e.g.

eq = equilibrium(dbf, comps, phases, {v.P: 101325, v.T: 450, v.X("CU"): 0.50 + 1e-6, v.N: 1}, output='HM', verbose=False)

This will give you two phases, but the phase fraction of the second phase will be small.

[Wuwei-Mo]

Hello, recently, I have been performing equilibrium calculations for aluminum alloys with multiple components. I have encountered the same issue of missing phases that we discussed before. However, this time, I have more elements, and the values are constrained to 5-6 decimal places, yet the problem persists. Additionally, since the TDB file for this calculation belongs to our research group, I cannot share it. Is there a solution to this issue?