Flat Miscibility Gap Tops

[maxposchmann]

Description
Near the tops of miscibility gap regions, Thermochimica pretty consistently converges incorrectly. These are difficult cases, as the energies of systems with 1 and 2 copies of the miscibility gap phase approach one another. Since Thermochimica doesn't cache any phase compositions, it will change the composition of the first copy while comparing one assemblage versus another.

Two distinct cases have been observed:

1. More commonly, the end of the miscibility gap is at approximately the correct temperature, but the phase compositions are more different from each other than they ought to be. See example provided.
2. Sometimes the miscibility gap ends at too low a temperature. In this case,Thermochimica correctly finds a negative driving force for the second copy of the miscibility gap phase, but that value is smaller than the relevant tolerance (dTolerance(4)).

To Reproduce
Look at the miscibility gap in the IONIC_LIQ phase on the O-Sn binary phase diagram at 10000 atm using the ZIRC_no_liq.dat database. The top should be around 2450 K.

Alternatively, consider the input provided below. The last calculation resulting in a miscibility gap has a 7% difference in the composition of the two liquid phases.

Input
temperature = 2450.0:2452.0:1.0
pressure = 10000.0
mass(8) = 0.33333
mass(50) = 0.66667
temperature unit = 'K'
pressure unit = 'atm'
mass unit = 'moles'
step together = .FALSE.
data file = data/ZIRC_no_liq.dat
print mode = 2
write json = .TRUE.