I am interested to model binary ordered compound A-B with small concentration of impurities X. I want to generate representation A-B-X configurations (for A50B50 case) to be able to average some property of interest at a finite temperature.
I did A-B-X ternary alloy CE calculation within ATAT and now I am struggling with ATAT manual section "Monte-Carlo simulations " . I need to indentify proper choise of parameters for memc2. Do I need to make memc2 for all groundstates and average them or I need just to do that for the ground state A-B and choose somehow amount of X?
The best option for solving original problem would be to be able to generate CE and be able to take those cluster structures for calcualtions of the property and weight those according to some fixed order parameter in A-B matrix (order 0.8-0.9) and also over configurations represntative for impurities.
I cannot find any information in the manual how to do that.
if you are ok with neglecting short-range order, than you could just generate a Special Quasirandom Structure with mcsqs that mimics the random state at a given composition in a cell of a given size.
Otherwise, you should take snapshots out of a (small) monte carlo simulation. Of course this assumes you have obtained a CE already (with mmaps). I don’t know exactly where you have trouble with memc2 - I would need more specifics to help you here.
Axel,
the best option for me is 2, so it’d be great to get snapshots for me in order to take into account the short ragne order. Now the questions are following.
What conditions does mcsnapshot.out correspond to? Starting? Ending? Average?
Is it possible to use canonical ensemble? How can I specify its initial composition? I see conccons.in file, but its purpose if not fully clear. "1.0Al -1.0Li +0.1*Co = 0.5" - looks like a linear equation. Can I use this file to get atomic ordering with the fixed composition?
Whithin the Semi-GC ensemble do I need to run over all possible values of chemical potentials to get the desired compositions? I need a composition AB*nX (n varies from 0 to 0.2).
As far as I understand, mu in memc2 is in eV now in contrast to emc2?
ending. The option -opss allows you to do periodic snapshots.
Yes: just put more than one linear constraint: e.g.
1.0Al=0
1.0Li=0
1.0*Co=0
The "0" on the right are irrelevant - all composition will be fixed at their starting values.
If you use the previous trick you shouldn’t have to.
Actually emc2 can do both conventions (eV or dimensionless).
But in multidimension there was no easy definition for a dimensionless mu.
So for memc2 mu is in eV