Mcsqs monoclinic supercell

Alloy Theoretic Automated Toolkit (ATAT)
1

Dear all,
I’m using mcsqs and trying to create a 96-atom monoclinic supercell (Space Group P21/c) for Yttria-Stabilized-Zirconia (YSZ, with 10mol% of Y2O3: 96 atoms, 26 Zr-, 6 Y-, 61 O-atoms and 3 O-vacancies). It nicely worked for the cubic cell and I was able to create some good structures with this tool, but I really don’t know what could be wrong when trying the monoclinic unit cell. I tried different ways to reach the goal, but always ended up with the following error:

Impossible to match point correlations due to incompatible supercell size.

What I did:
1.) To check if it works for the monoclinic unit cell I changed the rndstr.in cell parameters to the following values:

5.140000 5.200000 5.310000 90.00000 99.20000 90.00000
1.000000 0.000000 0.000000
0.000000 1.000000 0.000000
0.000000 0.000000 1.000000
0.070300 0.335900 0.340600 O=0.5,Vac=0.5
0.929700 0.664100 0.659400 O=0.5,Vac=0.5
0.929700 0.835900 0.159400 O=0.5,Vac=0.5
0.070300 0.164100 0.840600 O=0.5,Vac=0.5
0.557700 0.254900 0.021100 O=0.5,Vac=0.5
0.442300 0.745100 0.978900 O=0.5,Vac=0.5
0.442300 0.754900 0.478900 O=0.5,Vac=0.5
0.557700 0.245100 0.521100 O=0.5,Vac=0.5
0.275800 0.041100 0.208200 Zr=0.5,Y=0.5
0.724200 0.958900 0.791800 Zr=0.5,Y=0.5
0.724200 0.541100 0.291800 Zr=0.5,Y=0.5
0.275800 0.458900 0.708200 Zr=0.5,Y=0.5

Then I ran

corrdump -l=rndstr.in -ro -noe -nop -clus -2=4
mcsqs -n 96

changed the sqscell.out to:

1

2. 0. 0.
0. 2. 0.
0. 0. 2.

and after running "mcsqs -rc" I got my final SQS.

2.) Afterwards I changed the probabilities/occupancies for Zr and Y in the rndstr.in to the desired values for the 96-atom cell,

0.275800 0.041100 0.208200 Zr=0.8125,Y=0.1875
0.724200 0.958900 0.791800 Zr=0.8125,Y=0.1875
0.724200 0.541100 0.291800 Zr=0.8125,Y=0.1875
0.275800 0.458900 0.708200 Zr=0.8125,Y=0.1875

did all steps from above again and it worked quite well.

3.) As final step I changed the values for O and O-vacancies:

0.070300 0.335900 0.340600 O=0.953125, Vac=0.046875
0.929700 0.664100 0.659400 O=0.953125, Vac=0.046875
0.929700 0.835900 0.159400 O=0.953125, Vac=0.046875
0.070300 0.164100 0.840600 O=0.953125, Vac=0.046875
0.557700 0.254900 0.021100 O=0.953125, Vac=0.046875
0.442300 0.745100 0.978900 O=0.953125, Vac=0.046875
0.442300 0.754900 0.478900 O=0.953125, Vac=0.046875
0.557700 0.245100 0.521100 O=0.953125, Vac=0.046875

Running "mcsqs -n 96" again, delivers the error from the top of the page. :?:

I also tried the steps mentioned in https://www.brown.edu/Departments/Engineering/Labs/avdw/forum/viewtopic.php?f=4&t=306&p=1544&hilit=incompatible#p1544, but nothing seems to help solving my problem.
I found out, that it works for 192=2*96 atoms. I think this works due to the fact that 192 times the probabilities from the rndstr.in results in an integer and the calculation with 96 doesn’t. On the other hand the O-sublattice in the 96-atom cell should contain 64 positions, which will lead to integer values too concerning the probabilities. That’s what I thought was right and what should be fulfilled when calculating the probabilites for the sublattice. So I’m not sure what exactly could lead to this error and would appreciate any help, thank you.

2

Hello,
you can find a tutorial on https://cniu.me/2017/08/05/SQS.html for creating your rndstr.in and doing the simulations using mcsqs. I used an optimized structure obtained via VASP, but for bcc-HEA of VNbMoTaW I would recommend you to look for literature and create the cells via VESTA or a similar program visualizing your input. You need the lattice constants, angles should be 90M-BM-0 within bcc-cell and you need the atomic positions with probabilities for each sublattice. Hope it helps you solving your problem.
Regards
Tom

3

Hi,
I wanted to ask how did you take into account the change in the volume of the cell upon substitution with vacancies since you fix the sqscell.out to 2x2x2? I guess, this fixes the supercell lattice vectors.

Any help would be highly appreciated.

Thanks
Ankit Sharma
University at Buffalo

4

When generating sqs, lattice vectors (and site coordinates) are kept fixed. This is harmless because they will be relaxed when you actually perform an ab initio calculation with the generated structure as a starting point.

5

Hi Thomayto,
How did you prepared the rdnstr.in? and how can I get these parameters for a given system(e.g, bcc-HEA of VNbMoTaW).
Thanks a lot.