SQS question

[huangyc]

Dear Prof. Dr. A. van de Walle,

Recently I am using ATAT code to produce SQS structure with 441 supercell. The corrdump' was correctly done within several seconds, and then the command of mcsqs -n = 32 -rc’ was executed. The code indeed correctly ran, but nearly one week past, the `mcsqs’ is still running, and no files are updated. I am wondering what the matter of this case is. How can I monitor the computing process? Thanks in advance.
This is my rndstr.in file.
3.81 3.81 20.00 90 90 120
1.00 0.00 0.00
0.00 1.00 0.00
0.00 0.00 1.00
0.00 0.00 0.00 Sn
0.33333 0.6666667 0.25000 S=0.75,Se=0.25
0.6666667 0.333333 0.75000 S=0.75,Se=0.25

And this is the sqscell.out.
1
4.000 0.00000 0.00000
-2.0000 3.4641 0.000000
0.00000 0.0000000 1.000

Other files like clusters.out, mcsqs.log, bestsqs.out, bestcorr.out were correctly generated after `mcsqs’ task was submitted but no updating up to now. This is one of the content of bestsqs.out file.
3.810000 0.000000 0.000000
-1.905000 3.299557 0.000000
0.000000 0.000000 20.000000
4.000000 0.000000 0.000000
-2.000000 3.464100 0.000000
0.000000 0.000000 1.000000
0.333330 0.666667 0.250000 S
1.333330 0.666667 0.250000 S
2.333330 0.666667 0.250000 S
3.333330 0.666667 0.250000 Se
0.333330 1.666667 0.250000 S
1.333330 1.666667 0.250000 S
2.333330 1.666667 0.250000 Se
-0.666670 1.666667 0.250000 S
-0.666670 2.666667 0.250000 S
0.333330 2.666667 0.250000 S
1.333330 2.666667 0.250000 S
2.333330 2.666667 0.250000 Se
1.333330 0.202567 0.250000 Se
2.333330 0.202567 0.250000 S
3.333330 0.202567 0.250000 S
0.333330 0.202567 0.250000 S
3.666670 0.333337 0.750000 Se
0.666670 0.333337 0.750000 S
1.666670 0.333337 0.750000 S
2.666670 0.333337 0.750000 S
-0.333330 1.333337 0.750000 S
0.666670 1.333337 0.750000 S
1.666670 1.333337 0.750000 Se
2.666670 1.333337 0.750000 S
-1.333330 2.333337 0.750000 S
-0.333330 2.333337 0.750000 Se
0.666670 2.333337 0.750000 S
1.666670 2.333337 0.750000 S
-1.333330 3.333337 0.750000 S
-0.333330 3.333337 0.750000 S
0.666670 3.333337 0.750000 Se
1.666670 3.333337 0.750000 S
2.000000 3.464100 1.000000 S
-1.000000 3.464100 1.000000 S
-0.000000 3.464100 1.000000 S
1.000000 3.464100 1.000000 S
-0.000000 1.000000 1.000000 Sn
1.000000 1.000000 1.000000 Sn
2.000000 1.000000 1.000000 Sn
3.000000 1.000000 1.000000 Sn
-1.000000 2.000000 1.000000 Sn
-0.000000 2.000000 1.000000 Sn
Need I still wait?

[terencelz]

What version of atat are you using? The syntax of sqscell.out is different since 3.07.
The hexagonal lattice in your bestsqs.out seems wrong.

The -rc option lets you specify supercells to use (via a sqscell.out file).
Useful if brute force enumeration takes too long.
The sqscell.out usually comes from a previous mcsqs run but you can create it yourself.
Note that the sqscell.out file must start with the number supercells you will provide
and that supercells must be expressed in multiple of the axes defined in the rndstr.in file
(in particular, all numbers entered should be integers).
[If your are using a version prior to 3.07, these should be entered in cartesian coordinates.]

So if you are using atat 3.07+, sqscell.out should be

1
4 0 0
0 4 0
0 0 1

I haven’t tried it myself yet, but think it’ll work…

[huangyc]

Thank you for your reply. The code I used is the stable version (3_14).
I had followed your advice, but things do not improve. Still, the files cannot be updated after two days past. The "mcsqs" process is still running. Anyboby can give me some suggestions?
Thanks in advance!

[terencelz]

The official webpage (https://www.brown.edu/Departments/Engine … avdw/atat/) shows the stable version is 3.04, rather than 3.14. In this case, it matters because the input syntax is different.

Try cellcvrt -pbv < bestsqs.out to see if your bestsqs.out so far is hP (hexagonal) or not to make sure the right sqscell.out is given.

I’m using 3.16, and when sqscell.out is [4,4,1], the objective function can only go as low as -2.916667. When it is [4,4,2], or larger, Perfect match is immediately found.

The reason is, when a supercell you specified is small, the stochastic search can find it hard, or impossible (even after enumeration) to get the Perfect match.

[huangyc]

Thank you for your prompt reply! I downloaded 3.14, maybe not the stable version. The "cellcvrt" command shows it is a hP.
When I used 442, indeed a perfect match is immediately found. But the structures become planar, in shrap contrast to the really configuration (Sn sandwiched by two Se layers). Also 442 produces 96 atoms, my understanding is that it only increases the height of the vacuum layer, the atom number should equall to the 441 of 48 atoms.
Can you give me a detailed computional process? Many thanks. I only need 441.

[terencelz]

I got the same outcome as you did. The program only knows to expand the supercell periodically and fill it with atoms. It doesn’t create vacuum.

So if you want a perfectly randomly matching supercell, go with 442. 441 doesn’t appear to perfectly match the random correlation function.

I’ll wait for other replies too.

[avdw]

Thanks to those who offered some help. Just to add: if mcsqs does not update the files, it’s just that there is likely no better solution. Because it uses a stochastic algorithm it has no way to know when it’s done. The question is more: "is the SQS in bestsqs.out good enough for your application"?