Vibrational free energy of single substitutional atom

Hi, dear everyone,

I’m working on the free energies of alloys in the dilute concentration. I want to ask would it be a good or bad idea to use fitsvsl code to study the vibrational free energy of configurations where substitutional solutes are isolated or slightly clustered (several solute atoms binded together, say small clusters of 2-5 solute atoms)?

I understand that the CE usually is not ideal for the study of very dilute configurations, because the CE energy is validated as a per atom quantity. The problem here for vibrational contribution in dilute configurations is that can we get somehow reasonable transferable force constants from a small training set of dilute configurations?

For example, can I rely on the following procedure? 1) build several configurations of dilute concentration c<0.02 (and that of pure reference c=0), 2) apply perturbations, 3) and use fitsvsl to study the relative vibrational energy against pure elemental reference. 4) try to obtain the transferable force constants, and apply it in some slightly cluster configurations not in the training set.
fitsvsl came into my mind because it seems to be a rather cheap way to evaluate the vibrational contribution.

btw: is there any references from which we can understand the effect of the option "-er"? and any suggestion on its value?

Many thanks.

This is a good and valid idea to get the vibrational free energy contribution of dilute defects. But make sure to include some configurations with neighboring impurities in your "training" data set.

The -er is a way to specify the supercell: it is simply the minimum distance between periodic images of the same atom. The code will generate a supercell that satisfies this constraint. Typically, -er must be at least twice the longest pair you are considering.

Dear Axel,

Thank you for your answering!