I am wondering how could I generate the energy convex hull of a pseudo-ternary, say BaZrO3-Ba2Y2O5.
Currently, in ref_energy.out, I only see the energy of each individual atoms, i.e. Ba, Zr, Y, O, Va. However, how those energy can be determined? The smallest chemical unit with calculated energy is either BaZrO3 or Ba2Y2O5. There is no formation energy with pure element as reference states.
I am only interested in the "charge-neutrality" plane of Ba-Zr-Y-O-Va system, so after mmaps with concentration constraint, how could I generate the energy w.r.t the pseudo-ternary end member? The fit.out seems giving formation energy information w.r.t pure elements…
You are correct that some energy reference may not be unique when concentrations span a space of lower dimension that the number of species-1. In such case, maps picks the solution with the smallest L2 norm. (This is analogous to Moore-Penrose generalized inverses of matrices).
Note that it may be tricky to construct a cluster expansion with only structure on the charge-neutrality plane. There is nothing to prevent artificial ground states off of that plane. Even if running MC under composition constraints, the system could phase-separate into two non-neutral fictitious phases.
You can always specify your own ref_energy.in (NOTE ‘.in’ not ‘.out’) the get the formation energies wrt the reference you want. It’s not a very difficult math problem so find the elemental energies that give a zero formation energy for your endmemebers (the solution is not unique, though, but that’s ok).
BTW, I wouldn’t be surprised if the default that maps selects would already give you what you want.
So you mean, only doing CE in neutral-plane is not sufficient? But that seems too complex, if we include O non-stoichiometry.
How to deal with the situation, where a pseudo-binary BaZrO3-Ba2Y2O5 becomes a pseudo-quarternary Va-BaZrO3-Ba2Y2O5-O2?
You mentioned the concern of lacking sufficient reference state off the neutral-plane to prevent the MC structures from becoming phase separated and charged. Will that be taken into consideration by the ECIs of pairs, triplets,…?