Re: Choosing of windows and width for Umbrella Sampling

From: Giacomo Fiorin (
Date: Mon Dec 08 2014 - 08:29:49 CST

Hello Kevin, for (1) you should have overlapping histograms, otherwise you
won't be able to use wham or any related method.

For (2), the force constant is rescaled so that the unit if the colvar is
the width parameter, so that multidimensional restraints over very
different variables can share the same force constant. Because in US
you're calculating the histogram yourself, leave the width equal to 1, and
change the force constant so that you have an overlapping distribution.


On Dec 8, 2014 7:49 AM, "Kevin C Chan" <> wrote:

> Dear Users,
> I am currently doing Umbrella Sampling using colvars in NAMD. I am
> moving a protein towards a membrane and sampling a vertical distance
> of 30A in 21 windows and with width of 0.25 in colvars. The result
> histogram shows very "concentrated" distribution at distances
> representing the centers of each window. This actually makes sense as
> the system should be confined to be near the center defined in colvars
> by some harmonic restraints.
> My question is
> 1. Is such histogram agrees with the principle of Umbrella Sampling in
> which a "overlapping" histogram is required? As in fact the
> distribution does not even "touch" any of the adjacent windows.
> 2. Can anyone explain more about the width option in colvars? The
> manual sometimes confuses me: it says that width helps to rescale the
> "value" for harmonic restraints. So what should I do if I would like
> to broaden the distribution of histogram? Increase the width or
> decrease the force constant, or they are actually the same thing?
> Thanks in advance, I will greatly appreciate it if anyone could share
> your experience.
> Kevin
> City University of Hong Kong

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:23:06 CST