Generating SQS with multiple substitution simultaneously at different Wyckoff Sites

icet
1

Hi,

I was trying to generate an SQS structure, but I get the following error message:

    raise ValueError('Symbols {} found on multiple active sublattices'.format(bad_symbols))
ValueError: Symbols {'Ti'} found on multiple active sublattices

Isn’t it possible to generate an SQS with a few active/wyckoff sites containing a particular element and few other containing another element, simultaneously?

for example I used this ClusterSpace for mixing of Ti-Ta and Ti-Mo

cs = ClusterSpace(primitive_structure, [10.0, 6.0], [['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'], ['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ta', 'Ti'],['Ti', 'Mo'],['Ti', 'Mo'],['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O'], ['O']])

If this is possible, how can I achieve it?

Thanks,
Manish

2

Is your only goal to generate a SQS structure or you want to construct a ClusterExpansion?

3

My only goal is to generate an SQS structure.

4

You could try setting Ti atoms on one sublattice to another atom type, generate the SQS structure and then change back.

The resulting structure might have slight correlations but could still work, depending on what you want to use the structure for.

5

Is it better to generate the SQS in two steps—first by fixing one atom, say Mo, and then using that SQS structure to allow Ti to occupy the Mo sublattice? Or is it preferable to use the method you suggested, for example by temporarily replacing Ti with Zr and then switching back to Ti?

I will study the diffusion of the Li atom in that SQS structure.

6

I dont know which will be better, or how to quantify “better”.
I’d suggest try both and see how they look and maybe if they give any differences in diffusivity.

Additionally, if you’re using a big supercell for diffusivity analysis, then just randomizing the atoms in the supercell (rather than SQS) probably is pretty good.