How to choose/create rndstr.in file?

Alloy Theoretic Automated Toolkit (ATAT)
1

Hi,

I am a beginner to the ATAT and trying to create SQS using it. I have a system of PbTe conventional unit cell with its POSCAR as:
Te4 Pb4
1.00000000000000
6.5591489608095106 -0.0000000000000000 -0.0000000000000000
-0.0000000000000000 6.5591489608095106 -0.0000000000000000
0.0000000000000000 0.0000000000000000 6.5591489608095106
Te Pb
4 4
Direct
0.5000000000000000 0.0000000000000000 -0.0000000000000000
0.5000000000000000 0.5000000000000000 0.5000000000000000
-0.0000000000000000 -0.0000000000000000 0.5000000000000000
0.0000000000000000 0.5000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 0.5000000000000000 0.5000000000000000
0.5000000000000000 0.0000000000000000 0.5000000000000000
0.5000000000000000 0.5000000000000000 -0.0000000000000000
I want to know, how would a basic rndstr.in (an example) file look in this case?

And, in the command: corrdump -l=rndstr.in -ro -noe -nop -clus -2=4; I wanted to make sure, is distance 4 in Angstrom?

Best,
Abhiyan Pandit

2

Further,

I created rndstr.in as:6.559 6.559 6.559 90 90 90
1 0 0
0 1 0
0 0 1
0.5000000000000000 0.0000000000000000 -0.0000000000000000 Te
0.5000000000000000 0.5000000000000000 0.5000000000000000 Te
-0.0000000000000000 -0.0000000000000000 0.5000000000000000 Te
0.0000000000000000 0.5000000000000000 0.0000000000000000 Te
0.0000000000000000 0.0000000000000000 0.0000000000000000 Pb
0.0000000000000000 0.5000000000000000 0.5000000000000000 Pb
0.5000000000000000 0.0000000000000000 0.5000000000000000 Pb
0.5000000000000000 0.5000000000000000 -0.0000000000000000 Pb=0.5, Vac=0.5

Then I used: corrdump -l=rndstr.in -ro -noe -nop -clus -2 = 7
and then: mcsqs -n=16
I got the bestsqs.out as:6.559000 0.000000 0.000000
0.000000 6.559000 0.000000
0.000000 0.000000 6.559000
0.000000 -1.000000 0.000000
0.000000 0.000000 -1.000000
2.000000 0.000000 0.000000
0.500000 -0.500000 -1.000000 Vac
1.500000 -0.500000 -1.000000 Pb
2.000000 -0.500000 -1.000000 Te
1.000000 -0.500000 -1.000000 Te
0.500000 -1.000000 -1.000000 Te
1.500000 -1.000000 -1.000000 Te
0.500000 -0.500000 -0.500000 Te
1.500000 -0.500000 -0.500000 Te
2.000000 -1.000000 -0.500000 Te
1.000000 -1.000000 -0.500000 Te
2.000000 -1.000000 -1.000000 Pb
1.000000 -1.000000 -1.000000 Pb
2.000000 -0.500000 -0.500000 Pb
1.000000 -0.500000 -0.500000 Pb
0.500000 -1.000000 -0.500000 Pb
1.500000 -1.000000 -0.500000 Pb

I want to make sure if I am following correctly.

Any help/suggestions would be highly appreciated.

3

Continuous with the above post, in specific, how would I know the number of atoms to choose for the best SQS? (I know from some documentation that higher negative Objective_function value (or Perfect_match) in bestcorr.out file tells the best number of atoms in SQS, is it?)
Also, in general, I think we need the structurally relaxed unit cell before using it in the SQS generation. Isn’t it? What is the procedure if not so?

Best,
Abhiyan

4

There is no absolute answer here. The larger the SQS, the better you can make it (in terms of matching random state correlations). It’s just a matter of how much precision you need and/or how large of cell you can reasonably handle with ab initio methods.
You should look at the correlations (rather than the objective function) and see if they match well enough for your application.
You could also compute the quantity of interest (e.g. energy) as a function of SQS size and see that it converges.

No need to structurally relax before generating an SQS.
The unrelaxed structure just has to be "close enough" so that the neighbors are roughly in the same ordering of distance than in a fully relaxed structure.