Re: [AMBER] restraint in umbrella sampling

From: Jason Swails <>
Date: Wed, 2 Jun 2010 14:25:55 -0400


r1 and r4 are the values at which the potential becomes linear. Between r1
and r2 and between r3 and r4, the potential is harmonic, and between r2 and
r3 it's flat (0 potential). What you should aim for is to choose r1 and r4
such that the simulation will never sample these distances, because you want
the potential to always be parabolic. r1 near 0 is reasonable, since the
system will never sample this area (due to vdw clashes at the very least).
Making r4 200 or 2000 or 20000 makes no difference to the simulation as long
as the distance doesn't attain a value higher than any of those values.

Of course, each simulation will eventually differ due to round-offs and the
fact that MD on a many-body system is chaotic, but this is not due to the
restraint potential.

All the best,

On Wed, Jun 2, 2010 at 2:17 PM, dhacademic <> wrote:

> Dear Amber users,
> I have two questions about the restraint in umbrella sampling.
> (1) According to the Amber9 tutorial (p159), the torsion angle restraint at
> the minimum of 170 degree is as follow: r1=0, r2=170, r3=170, r4=360, where
> r2=r3, and r1<<r2, r3<<r4. Similarly, the restraint at the minimum of 10
> degree can be r1=-160, r2=10, r3=10, r4=200. I think the "r1=-160" here is
> reasonable (minus degree of angle has a physical meaning). However, if I
> want to impose a bond length restraint at the minimum of 1.2 angstrom, how
> can I choose the r1 and r4? If I use r1=0.2, r2=1.2, r3=1.2, r4=2.2, it
> seems that r1 and r4 do not satisfy the requirement that "r1 and r4 should
> far away from r2 (or r3), so that the potential is essentially harmonic
> everywhere". However, if I use r1=-8.8, r2=1.2, r3=1.2, r4=11.2, there still
> has a problem: what does the r1=-8.8 mean?
> (2) For a given step size between different windows, is there any general
> rule to choose the force constant? In my calculations, I do some test
> calculations to make sure the overlapping of sampling between neighbor
> windows is good enough. In principle, PMF results should not depend on the
> force constant used in each window. However, from my test calculations, I
> find that different force constants may have some different PMF results
> after long time samplings. So I am confused about the force constant.
> Many thanks for any suggestion.
> Best,
> Hao Dong
> 2010-06-02
> dhacademic
> _______________________________________________
> AMBER mailing list

Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
AMBER mailing list
Received on Wed Jun 02 2010 - 11:30:04 PDT
Custom Search