RE: AMBER: nscm in simulation annealing

From: Ross Walker <>
Date: Tue, 17 Apr 2007 09:25:35 -0700

Dear Shaowen,

> Last week you wrote "Basically, center of mass motion should never be
> removed in Langevin simulations. In amber9, center-of-mass velocity
> removal is never done when ntt=3 (the system is just recentered every
> nscm steps, but the velocities are not altered)." I have thought that
> nscm should set to 0 whenever ntt=3 is used. However, after checking
> several tutorials using ntt=3
> (, I
> found nscm is
> set to 1000, the default value. Could you please clarify the usage of
> this keyword again? I am thinking whether I need to redo my
> calculations. Your kind assistance is highly appreciated.

The issue with removing center of mass in langevin simulations is a
theoretical one. If you regularly remove center of mass motion you will not
get the correct ensemble and things like diffusion coefficients etc will be
incorrect. You should read up on the langevin method in the literature to be
sure you understand the theory.

With regards to the keyword nscm the behavior is as follows:

1) When ntt /= 3
   default nscm=1000
   Center of mass translation and rotation is removed from the system every
nscm steps. After the first initial correction each subsequent correction
should be small simply coming from net translation introduced by the
thermostat. If you are running NVE the corrections should be VERY small
representing simply the errors inherent in the integration method.

2) When ntt == 3
   default nscm=1000
   Every nscm steps the coordinate system is recentered. Thus your system
can have center of mass motion but since in a very long simulation the
system could drift many hundreds of angstroms we don't want it to diffuse so
far that we overflow what the trajectory file can handle. This also makes
visualization of the trajectory file easier. In the output file the amount
by which the origin of the coordinates had to be shifted to recenter the
system is printed. This therefore correctly implements the langevin method.
You could later postprocess the trajectory file to remove the origin
recentering if you so wished. Alternatively in calculating properties such
as diffusion coefficients since the amount of movement is written to the
output file so you can correct for this in your calculations.

  Alternatively if you want you can set nscm=0 in which case no recentering
will be done. However with a very long simulation you would see your system
translate a long way from the starting coordinates.

  With ntt=3 if you take the nscm=1000 simulation and post process the
trajectory to account for the recentering of the origin as specified in the
output file you will end up with the exact same trajectory as if you set

I hope that clarifies things.

|\oss Walker

| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- |
| | PGP Key available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Wed Apr 18 2007 - 06:07:35 PDT
Custom Search