The sqscell.out does contain the number of cell shapes, each of which is described by a 3x3 matrix.
You do want to use the primitive cell, otherwise, you will be missing supercell of the primitive cell that are not supercell of the conventional cell.
Supercells of a given structure do not necessarily have the same symmetry as the given structure - this is perfectly normal.
Sorry, the file attachment feature of the board is currently not working - please paste files in the box instead.
The issue is not that number of cell is incorrect, it is that the code takes very long to write cells to disk, so the number you see in the header is larger than the structure written so far.
I’ve looked into this some more and the code is stuck in the routine that finds the most "symmetric" version of the cells generated. I am working on fixing this.
In the meantime, I’ve noticed that your concentration is 1/5 for each species but your requested cell size in 54. You would not be able to generate a cell with the right composition: you need the cell size to be multiple of 5, so -n=55 or -n=50 would work in your case. (And actually the bug you experience does not happen at those sizes).
This should get you going for now. But I am still working of fixing the freezing issue.
Yes, sorry you’re right, "scell" should be just "cell":
if (near_zero(norm(b)-lc)) { // line added
lp[2].init(cell,1);
while (norm((rVector3d)lp[2])<lc-zero_tolerance) {lp[2]++;}
} // line added
From mcsqs -h:
Creating a file called sqsparam.in (or as specified by the -pf option)
with 4 numbers in it allows you to set the -wr, -wn, -wd and -T options during runtime.
(for backward compatibility, if 2 numbers are specified, they set the -wr and -T options
while the remaining options are set to their defaults.)
The definition of "Q" depends on the number of clusters that were included during the optimization. So if that changed between the runs, you can’t compare the numbers. If you did keep the same clusters, then ignore the previous caveat.
Now, it is possible that the closeness to the random state correlation does not improve that much with cell size. It is also useful to pick cell sizes that are a composite number with many factors (especially factors related to the number of clusters) to maximize the chances that correlations can be matched exactly.
Finally, It’s more informative to reader to report the number of correlations matched exactly rather than "Q". Q is useful for optimization purposes but is not very easy to interpret.