Re: Convergence of ABF calculation: n_samples

From: Jerome Henin (
Date: Wed Dec 05 2007 - 10:14:37 CST

Dear Vaithee,

On Dec 5, 2007 9:39 AM, Subramanian Vaitheeswaran <> wrote:
> Dear Chris and Jerome,
> I have a few questions regarding the optimization of an ABF calculation. When would it be a good idea to break the >reaction coordinate into multiple windows? - when the free energy changes along this coordinate are large?

What determines the speed at which the system evolves is a combination
of the free energy and diffusion properties. ABF is pretty good at
erasing features of the free energy landscape, but then one is left
with diffusion in a flat free energy landscape, and that can still be
very slow, particularly for a broad range of the RC. Using different
windows is mostly a way of reducing the impact of slow diffusion.

> If I do have multiple windows, I can't use the uniformity of n_samples to judge the convergence of the pmf. Would it be >enough if sampling is uniform (or approximately so) *within* each window?

That should happen, but there are other criteria:
* reversible diffusion throughout each window
* continuity of the free energy derivative (average force) between
adjacent windows
* if you can afford it, add a final run covering the whole range and
using data from the windowed calculations, and see what happens.


This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:45:39 CST