[BUG] DCSF assumes f(q,t)_AB = f(q,t)_BA is this correct? about mdanse HOT 3 CLOSED

I think things are calculated correctly, The structure factor shouldn't be symmetric in general, which you can see from the definition of the van Hove function:

https://en.wikipedia.org/wiki/Dynamic_structure_factor#The_van_Hove_function

If you exchange the r, and r+r' terms you don't have the same equation, you kind of get one going backwards in time. I guess at equilibrium we might expect the two to match, but that should be something that happens in the data, not the calculation.

from mdanse.

Lucas is right: F_ab(Q,t) = F_ba(Q,t). As every time you run a calculation the number of vectors (if > Nmax) is chosen randomly, nothing ensures that you will get exactly the same result. I suppose this could explain why you got two different curves. Otherwise, some checking would be needed.

from mdanse.

I used the grid qvector generator which is not random. I think the difference is due to the small number of configurations that are average over at large t, see #435.

from mdanse.