Dear all,
given a unit cell and a corresponding ideal supercell, is there a way to obtain the list of the unique elements of the force constants which are effectively calculated and the map between these unique elements and those that are equivalent by symmetry?
Thanks a lot in advance for your reply.
Antonio
Could you explain a bit more what you are trying to do?
I think all this information should technically be available in e.g. the ClusterSpace and ForceConstantModel objects but it is probably be a bit tricky to extract.
Maybe a starting point is to dig into the information available in cluster_space.orbit_data?
@freeriks Any suggestions?
I have a trajectory from a dynamics simulation on a periodic system and I would like to test how the number of unique elements effectively calculated affects the phonon-derived properties, then I would like to compare the same properties obtained without symmetries.
How do you want to vary the number of unique elements?
A drastic way is to turn off the search for the space group symmetries, that is, imposing space group 1, maybe still maintaining the acoustic sum rule and the rotational invariance. Another way is to set the tolerance on the symmetry finder lower and lower until it detects space group 1. It is just a test: no matter the symmetry that the software exploits, if the system owns those symmetries intrinsically, they will appear anyway in the equivalence of the elements of the force constants, of course within a numerical tolerance.
Ok I think that should be doable, changing symprec in the ClusterSpace would be one way. To induce P1 you can also test introducing small displacements to the input structure, e.g. supercell.rattle(0.0001) and ClusterSpace(supercell).
Yes, given that you have enough training data.