Closed MBartkowiakSTFC closed 2 weeks ago
Looks great. I generated some cif files of some averaged structures and they loaded up in mercury successfully.
It won't make much (if any) difference but what do you think about running average over the atomic positions in fractional coordinates first then converting back to cartesians with the averaged unit cell rather than the current approach which runs the average over cartesian coordinates?
I think the two approaches would be the same if the unit cell is cubic and stays cubic during the simulation but I wonder if it would be different if it is not cubic and the unit cell angles change.
I thought that using the Cartesian coordinates would result in a slightly lower risk of getting wrong results. The averaging of the unit cell, on the other hand, is something that I would worry about in general, independent of how the coordinates were handled.
I cannot really predict how bad the worst case scenario would be, since generally the expectation we have that the atoms stay around their equilibrium positions normally should correspond to a reasonably stable unit cell shape as well.
Description of work A new analysis type has been added in this PR. It is a tool mentioned in #446 allowing the user to output the time-averaged crystal structure of the simulated system. The intention is to use it with an external tool like phonopy or spglib to determine the symmetry of the system and find the single unit cell of the supercell that was being simulated.
Fixes
To test Tests must pass. You can run the AverageStructure analysis on a crystal(like) system; the LAMMPS CuSbS trajectory can be used if you exclude the Cu atoms.