I tried to generate SQS supercell using ICET using generate_sqs_from_supercells (the 8x8x8 supercell of cubic 2-atomic system). However, in different runs (using the same input) I get different supercells. I guess the reason is too small parameters (number of MC steps and cuts).
So my question is how to judge if the parameters (and thus the supercell) are ok?
At the output I got parameters such as temperature and best score.
As far as I understand, the temperature must be close to zero. But what about the best score? What does this parameter say?
I think it is probably expected to get different output SQS structures when using a large supercell since its a random Monte Carlo search for the best SQS structure.
Maybe the documentation provides some more help.
In general it is probably difficult to say if the output SQS structure is “random enough” or “good enough”. My suggestion would be to generate a few different SQS structures and use these for your application, and see if the using one or the other changes the result you are interested significantly or not.
thank you for your kind reply.
The last thing I can’t find in the documentation - do you know what “best score” (parameter printed at the output) means? Is there any expected value of this parameter?
I dont know what typically “good” values for best score is.
The score, as I understand it, is the measure the similarity between the cluster vector of the SQS cell and the random alloy and the best_score is the structure that had the lowest score so far during the MC simulation.