Chemical composition mixtures and not equilibrium transformations

[gplanu]

Dears,
after a very fruitful discussion also with Mr. Richard Otis, I am facing to these two problems next explained:

1. Evaluation of complex multi component chemical composition: how could be the fastest way to extract for a defined composition the resulted Aem, A1 equilibrium curves using calculate and equilibrium pycalphad methods? Is it possible to obtain the resulted binary diagram maintaining constant i.e. Mn, V, Nb and varying only Fe and C (with vacancies of course)?
2. Regarding martensite and bainite transformation working with the same complex mixtures I have convergence problems adding components. I am working using an second order explicit Euler algorithm (I attach a pic below). Red points are using a simplified algorithm in Matlab, blue ones are from pycalphad.
   
   
   
   

Looking forward for your suggestions.
Thanks a lot in advance for your help and attention.
Best regards!

[richardotis]

1. At the moment, a binary iterative search is probably the quickest way to find liquidus or solidus temperatures. For binary phase diagrams, it's possible to programmatically access the phase boundaries directly, but a binary search is going to be the fastest general way for now. For isopleth diagrams, pycalphad does not offer any support yet. We have something in progress for it, but it will probably not be ready before the new year.
2. We have some bug fixes ready to be merged which may address this issue.

[gplanu]

The excel screens above show that there is the phase formation. Moreover, I tried to launch an equilibrium calculation of the two phases and the result is in the screen below:

So is not easy to understand which phase has to be considered... that is the reason why I came back to the theory to fit the quadratic equation..

If you see the composition is really a steel with low amount of manganese and vanadium, so I expect that the calculation performs without any issue.

[gplanu]

Dear Richard, regarding our conversation attached below in pycalphad/pycalphad Gitter: which is the main difference between your suggestion and the tentative that I show you above considering equilibrium method giving as input a single phase?
Could you explain me please that from calphad and numerical point of view? Thanks a lot!

☝️ Edit: @gplalab_gitlab: I have some new features in progress which should address this concern. For T0, I have some code using my development branch which uses the entropy differences between the phases to find it efficiently for a given composition. In the next version you'll also be able to access the CompositionSet objects and control the energy minimizer directly, without automated miscibility gap detection interfering.

The current concept (subject to change) looks like:

Get a starting point from a previous calculation, called 'wks' (not shown)

def find_first_compset(phase_name):
for index, compsets in wks.enumerate_composition_sets():
for compset in compsets:
if compset.phase_record.phase_name == phase_name:
return compset
return None
bcc_compset = find_first_compset('BCC_A2')
hcp_compset = find_first_compset('HCP_A3')

[ some code omitted for brevity ]

bcc_result = Solver().solve([bcc_compset], initial_conditions)
hcp_result = Solver().solve([hcp_compset], initial_conditions)
The solver won't add any phases when used this way, so you can manipulate single-phase properties.

[gplanu]

Hello @richardotis, I have another question for you. I am solving an equilibrium for a steel mixture with 10 components using equilibrium method. I am trying to solve equilibrium only for the BCC_A2 phase with a range of 300-2000 K (temperature step 50K) and carbon range 0-0.25 molar fraction (carbon step 0.01).
During phase elaboration I have this screen which reports to solver.py (attached). Any suggestion?

Thanks a lot for your help.
Giulio

[richardotis]

That looks like a bug. If you can put together a reproducible example, you can open a new issue so we can address it.

[gplanu]

Ok, I will share in a different issue. In any case, I think is a bug because working with single temperatures it solve (of course with NaN case) but few points could be extracted.