Orthogonal cell - mcsqs

Hi,

I am trying to generate a disordered structure with mcsqs.
As a test, I used the following simple rndstr.in:

2.0 0.0 0.0
0.0 2.0 0.0
0.0 0.0 2.0
1 0 0
0 1 0
0 0 1
0.0 0.0 0.0 O=0.5, Al=0.5

After running corrdump, I ran mcsqs -n=32. However, the structure I get is not always orthogonal, which makes it more difficult to handle:

2.000000 0.000000 0.000000
0.000000 2.000000 0.000000
0.000000 0.000000 2.000000
2.000000 -4.000000 4.000000
0.000000 -5.000000 3.000000
2.000000 -1.000000 -1.000000
2.000000 -2.000000 0.000000 O
1.000000 -1.000000 0.000000 Al

Am I doing something wrong here, or is there a way to force the lattice vectors to remain orthogonal?
I also would prefer the supercell to be as cubic as possible. Can I assign its dimensions beforehand, and let mcsqs populate it?

Thank you for your help.

Well, -rc and listing the supercell dimensions in sqscell.out did the job.

Using the -rc option is discussed for other purposes in the mcsqs paper. Is there any problem with proceeding in such a way?

Thanks.

Thanks for self-answering :wink: it makes my job easier!
It’s a perfectly valid way to go if you have reasons to prefer orthogonal lattice vectors.
But of course, imposing a constraint always reduces the optimal point you can reach, so you would end up being able to match fewer correlations to the disordered state values for a given number of atoms in the cell. Most good SQS have "weird" lattice vectors precisely because it helps in simulating a disordered phase.

Thanks for the reply, and glad I could save you some time!

It’s just that orthogonal cells are much more convenient for visualization. It’s also easier to handle in Lammps for instance, or to apply uniform/uniaxial compression.

Anyway, I’d have thought that the results do not depend on the shape of the cell. Or are the PBC not taken into account during the supercell building?

Have a good one

I understand about the advantages of orthogonal cells.
The PBC are fully taken into account during the SQS construction. It’s just that you get better SQS when no cell constraints are imposed. A constrained optimization always yields a worse optimum than an unconstrained optimization.

I see. Thanks a lot for the clarifications.

Have a good one