Hi,
Ross (the principled physicists at heart) is right here, insomuch that a
single frequency too fast to integrate will destroy your simulation. It is
up to you to show that not to be the case in your particular system.
However, I (the sloppy chemist at heart) still think that a single
unshaked bond will not cause trouble in any typical MD application.
BOE estimation:
A 'fast' bond will vibrate at ca. 3000 cm^-1, giving an 11 fs period.
Integrating such an oscillator numerically, either with a 1fs timestep or
a 2fs timestep propagation will cause some error by introducing (or
removing) energy from the system. Since the potential is quadratic, the
error from doubling your timestep should be four times bigger than before.
So one unshaked bond at 2fs timestep will cause 4x the inaccuracy that an
unshaked bond at 1fs would. Since we would accept 10.000s of such
inaccurate oscillator propagations for the 1fs timestep, completely
unshaked system, I still think that one bond will not cause problems.
Unless it causes your simulation to fail, something I have never observed
yet.
Length and rms for that bond might come out inaccurate, but wouldn't that
be even worse if SHAKE is turned on and the bond behaves completely
unnatural?
> following a protocol with a few unshaken Hs and dt=0.002 (to my
> understanding of e.g. Steinbrecher et al J Comp Chem 2011 3253-3263). If
I think this one disappears complete molecules, so all SHAKE constraints
would have remained intact.
Kind Regards,
Thomas
Dr. Thomas Steinbrecher
formerly at the
BioMaps Institute
Rutgers University
610 Taylor Rd.
Piscataway, NJ 08854
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Nov 02 2011 - 09:00:04 PDT
