What is the best strategy of extracting 3rd force constants?

Should I extract the 2nd FC and 3rd FC separately or extract them at one time? I remember there was some discussion on it, but I can’t find it now. Is it good to include the 4th cutoff while fitting the 3rd FC?


It depends on your system (low or high symmetry?) training structures (small or large displacements, many or few structures?) and what you want to use the force constants for.

Fitting 2nd and 3rd separately is something we’ve done when extracting effective FCs at a given temperature when training structures are from MD simulations. This makes the FC2/FC3 ratio larger and thus may be more suitable in perturbation theory.

Including 4th order will make the FC2 and FC3 agree better with the zero-Kelvin force constants since rather than higher order anharmonicity being renormalized into FC2 or FC3 it will now end up in FC4.

If you are unsure, maybe try the different approaches and see if it leads to any difference in your property of interest.