[Umbrella sampling] ACF calculation for diffusivity

[vtlim]

Data collection for the ACF calculation should be pretty frequent (every 2 fs? every 100 fs?) for the overdamped Langevin approach. This is due to calculation of the autocorrelation function (whether of positions or velocities). Perform a test on a simple system in which the frequency of data collection is varied. See how this frequency affects (a) the output of the autocorrelation function, and (b) the speed of the calculation.

From DLM:

[You might not have to] store data EVERY step to get an adequate estimate of the aucorrelation function, since normally the autocorrelation time is very long compared to a timestep, and that writing every step will slow things down — so you’d probably get better answers faster if you store less frequently.