What to do when the calculation of eutectic point is inaccurate?

[HaihuiZhang]

Dear Administrator,
The eutectic point calculated by ESPEI is not consistent with the experimental and T-C phase diagrams. I'm not sure whether there is a problem with the data or an error in the optimization. How do I check my data?

[HaihuiZhang]

input-data.zip

[HaihuiZhang]

[bocklund]

If you are certain about the eutectic point, one thing you can do is to add the three phase equilibria for that eutectic and increase the weight of that dataset to push the MCMC in that direction.

Something like the following should work

{
    "components": ["AG", "CU", "VA"],
    "phases": ["FCC_A1", "LIQUID"],
    "conditions": {
	"P": 101325,
	"T": [1052]
    },
    "broadcast_conditions": false,
    "output": "ZPF",
    "values":   [
        [["FCC_A1", ["CU"], [0.141]], ["FCC_A1", ["CU"],  [0.951]], ["LIQUID", ["CU"],  [0.38]]]
    ],
    "reference": "Fictitious",
    "weight": 100.0
}

[HaihuiZhang]

If there is the monotectic transition in the phase diagram, L-L1+L2, two liquids phase, can we input it as LIQUID#1 and LIQUID#2?

[bocklund]

1. I don’t think that three phase equilibrium would be useful. See [1] for an explanation of how the phase boundary data are optimized. On the other hand, tie-lines in the miscibility gap that has the two liquids with distinct compositions would likely be useful.
2. The #1 notation that is used to distinguish composition sets of the same phase in Thermo-Calc is not used in pycalphad or ESPEI, similar to how the fcc phase boundary data are described for your Ag-Cu datasets

[1] B. Bocklund, R. Otis, A. Egorov, A. Obaied, I. Roslyakova, Z.-K. Liu, ESPEI for efficient thermodynamic database development, modification, and uncertainty quantification: application to Cu–Mg, MRS Commun. (2019) 1–10. doi:10.1557/mrc.2019.59.

[HaihuiZhang]

I think the tie-line in the phase diagram maybe make up of two liquid phases, [["LIQUID", ["PB"], [0.25]], ["LIQUID", ["PB"], [NULL]]], when a monotectic transition occurs (L-L1+fcc). Is that right?

[HaihuiZhang]

/(ㄒoㄒ)/~~Another error occurs when I add the three-phase equilibria data to the input database refer to your answer. It can work in the parameter generate step, and mcmc optimization also done, but the plot step went error. Would you mind to help me find out what went wrong about my data, please?

• Attached files contain the input database, the tdb files, and the mcmc trace files.

AG_CU_20191007.zip