Accurate fit of reference energies

Alloy Theoretic Automated Toolkit (ATAT)
1

Dear Axel,

I am unable to get an accurate fit of the reference energies via the cluster expansion, even though the CV score is low (0.020 eV) and the CE converges wrt cluster diameter. I have tried re-starting maps with the weight.in file that has a single column containing the weights in the same order as fit.out, but with a larger weight of 10 for the structures in the ref-dirs (i.e 0/ and 1/ for my case). But this file doesn’t seem to be read, as the resulting CV and fit.out shows a weight of 1 for all structures. I have also tried changing the weights to 0.1 instead of 10, but it doesn’t change the weights in fit.out.

Any suggestions? I am using MAPS v 2.86.

Thanks,
Ganesh

2

Exactly. Just make sure you do that after maps is done, otherwise your work will be overwritten.

3

Download the "Beta version of the latest update" at https://alum.mit.edu/www/avdw/atat/.
(2.86 did not have this feature).
You should see "weights.in file detected, reading user-specified weights." when you run.

BTW, you can always tweak the empty and point cluster after the fact to reproduce the exact end member energies.

4

Do you mean tweak the values in eci.out corresponding to empty and point clusters to get exact reference energies, right?