Dear avdw:
In old version of SQS, by using CORRDUMP we can obtain correlation functions and initial SQS structure.
If the initial SQS structure is relaxed in VASP, we obtain the relaxed structure. How can I obtain the corresponding correlation functions related to the new relaxed structure? These correlation functions can be different from the original ones.
In other words, if I have a SQS structure of a certain compositon, how can I obtain the correlation functions?
Thank you very much!
Although the radial distribution function for the relax SQS may differ from the one of the unrelaxed structure, the correlations (that are used in a cluster expansion) should not. If they did, it would mean that the SQS has relaxed to an entirely different symmetry, in which case it would not be a valid SQS.
If you still want to compute the correlation of a relaxed SQS, corrdump does that. If allows for distortions in the structure is calculates the correlations of. But you have to provide an unrelaxed lattice file (lat.in). (Actually, the lat.in could be relaxed too, but it must be a unit cell and it must list the possible species on each site - i.e. it can’t be the relaxed SQS!)
dear avdw:
A special structure was relaxed,then the correlation function seems still not changed? I tried to change the value of the atom positions of a sturcture,the corresponding correlation function is still the same from the ATAT, Is that right?
As long as the "distorted" structure is still in the basin of attraction of the ideal undistorted structure it is logical the assign it the same correlations. This is different from the concept of radial distribution function, which is sensitive to distortion.
Thank you very much! When one do the mcsqs, the value of objective function should be more closer to zero? The perfect match means the value is zero?
Hi,
Assuming I’ve been following the conversation, regarding your last question, the objective function need not be zero for perfect match. It depends on the user parameters as well as the number of different atoms in your supercell.
The mcsqs.log lets you know the value after each iteration it changes.
Indeed, for the objective function, lower is better, but there is no special meaning to "0". If all correlation match exactly the ones of the disordered state, the code will output:
Objective_function= Perfect_match