I am dealing with a 64-window frwd FEP simulation, in the lambda range 0.2-1.0, for Unbound ligand - in the context of receptor- ligand FEP. I am observing, in duplicated runs from scratch, a sudden drop of /\A near lambda 1.0. preequil 175,000, numSteps 750,000. Overlap with back satisfactory.

Specifically, lambda//\A 0.9500/56.1717 0.9625/54.8583 0.9750/523438 0.9875/48.1643 1.0000/41.3893. From lambda 0.5 to 0.95, /\A is constant around 56.

I wonder whether such a steep change of /\A from 48 to 41 fits common experience. With a smaller number of windows, /\A = 56 from lambda 0.5 to 1.0.

thanks for advice (this question had no answer in NAMD forum)