General

Alloy Theoretic Automated Toolkit (ATAT)
21

Hi Axel,
In vasp.wrap, if I use something other than GAMMA (i.e., KSCHEME = GAMMA) what is it called?
And, in that case what should I specify in KPPRA =???

Thanks,
Gautam

22

Hi Axel,
I forgot to ask a couple of questions:
(i) If I want to run CE of ternary system (including pseudobinary with a spectator specie), should use maps or mmaps?
(ii) If I want to multiple runs of CE in the same machine, how do I set up the system? Should I make multiple copies of vasp executable? What else?

Thanks,
Gautam

23

(i) For a pseudobinary with a spectator specie you can use maps or mmaps.

I don’t see any problem, just make sure they are run in different subdirectories.

24

Hi Axel,
Is it possible to include more than three species (i.e., A, B, C, D and E for example) in genSQS?
If not, what is the theoretical limit as to max number of species one could use in gensqs?

Another question regarding CE:
When one should use "constituent strain formalism" to do CE? Related questions: (i) what is the name of the binary to run "constituent strain formalism"? (ii) are (is) the input file for it the same as regular CE? (iii) how do run mixed basis CE?

Thanks,
Gautam

25

Hi Axel,
If I have an output file of Monte Carlo simulation, how do I calculate "configurational entropy"?
Could you give me command line arguments to calculate it?

I discussed this issue with Marcelo Sluiter as he has a CVM code! But, his code is not generalized to consider an arbitrary no of pairs, triplet and so on as in ATAT! Without re-writing his CVM code he can not output configurational entropy!

I also discussed the same with Tetsuo Mohri, but he has the problem to equilibrate the system!

After these discussions, I came to the conclusion that ATAT followed by MCS is the solution for me!

So, could you give me necessary instructions.

Thanks,
Gautam

26

There is no fundamental limit, other than your patience :wink:
But I would suggest using mcsqs: you will get at least a fairly good answer much more quickly (especially for large sqs).

The constituent strain (CS) formalism is usefull when there are large atomic size mismatches.
csfit is the executable to fit the CS expression (see csfit -h for some help). Next, when you run maps or emc2, you include the -ks=cs option.
Note that I’ve only implemented this for binaries with cubic symmetry.

Sorry, I have not implemented the CVM, which is what you would need. I am quite sure Marcel Sluiter’s code can handle any multibody interaction. But the problem is more that you need that all your interactions fit with a give "maximal cluster" and then the computational cost scales as 2^(number of site in maximal cluster), which can quickly become too large.
You could just use published expressions for CVM in the tetrahedron and/or octahedron approximation (assuming you system is fcc) and just ignore the longer pairs or triplets, etc. You would already see if it gives you significant corrections relative to the simple x lnx mean field term.
After these discussions, I came to the conclusion that ATAT followed by MCS is the solution for me!

27

That is totally correct :slight_smile:

28

Hi Axel,
I am doing sub-lattice CE, and in a specific case it has generated 16 structures but no fit to the energies yet! If I do, more maps.log, it says,

Maps version 3.04
Not enough known energies to fit CE

So, a natural question is how do I force ATAT to fit energies so that I get eci.out, clusters.out, gs.out etc to facilitate MC run? Please, let me know.

BTW, now I could run ATAT in a big machine. I figured it out by myself. It was easy!

Thanks,
Gautam