Converting/Understanding ECIs with vacancies

Hello,

I naively defined my BCC lattice with interstitial sites as: [‘Nb’, ‘Ti’, ‘Zr’] (sublattice A), [‘O’, ‘X’] (sublattice B) . Now, I’m trying to understand my ECIs and it’s a bit difficult. If I understand this correctly, a multi-component vector value of 0 for sublattice B indicates the site is a vacancy. If so, my ECIs are parameterized for metal-vacant site interactions when it would be much easier to understand if it was metal-oxygen site interactions. Is there a way to convert the cluster expansion? My MC simulations are already done so refitting isn’t an option.

index | order |  radius  | multiplicity | orbit_index | multi_component_vector | sublattices | parameter |    ECI   
--------------------------------------------------------------------------------------------------------------------
   0  |   0   |   0.0000 |        1     |      -1     |           .            |      .      |    -0.782 |    -0.782
   1  |   1   |   0.0000 |        1     |       0     |          [0]           |      A      |     0.117 |     0.117
   2  |   1   |   0.0000 |        1     |       0     |          [1]           |      A      |     0.243 |     0.243
   3  |   1   |   0.0000 |        3     |       1     |          [0]           |      B      |    -0.127 |   -0.0423
   4  |   2   |   0.8493 |        6     |       2     |         [0, 0]         |     A-B     |    0.0353 |   0.00589
   5  |   2   |   0.8493 |        6     |       2     |         [1, 0]         |     A-B     |    0.0262 |   0.00436
...

Best,
-Mike

Okay, I found a solution. I fit a new CE to the old CE using structure enumeration to create 1000’s of training structures.