I am trying to generate an SQS structure for a binary alloy system. The input structure is a 2x2x5 FCC slab oriented in the (111) direction. Given that the structure is composed of 20 atoms, I expected to be able to specify a composition of X=0.95,Y=0.05 at each site and produce a structure with 1 Y atom and 19 X atoms. However, when I attempt this operation the job fails with the following error:
Impossible to match point correlations due to incompatible supercell size.
However, when I specify a composition of X=0.75,Y=0.25 the operation succeeds. I have tested all valid compositions between this range and they all fail in the same way.
If anyone has experienced this issue and found a workaround I would greatly appreciate some advice.
The sqs-finding procedure requires that sites be placed in groups that are symmetrically equivalent. The supercell must be big enough that it is possible to match the target composition within any one group of symmetrically equivalent sites. For instance, if your 20 sites/cell has 5 groups of sites, each of which has 4 sites that are symmetrically equivalent to each other, then, one 20 atom unit cell could at best match composition in multiple of 1/4.
Note that the file rndstrgrp.out (generated by mcsqs) groups the sites by groups that are symmetrically equivalent.
A possible work-around, if you value a small supercell over a strictly correct sqs, is to specify a different composition for the sites in each group, so that the overall composition is the one you want. In this case, it may make sense to generate multiple sqs with the compositions assigned differently. The average of the properties of these sqs would be more representative of the true disordered than one arbitrarily chosen site composition choice.