SHAKE does, in theory, conserve energy, but the Devil is the details and
one detail is the tolerance. If you set tol = 1.0e-8, this is generally
something that the double precision codes can handle and will greatly
improve your energy conservation situation. SETTLE, SHAKE for rigid
waters, is analytic and therefore has no meaningful tol setting, so it's
fine. Note, however, that SHAKE is an outlet to energy in the system.
SHAKE errors, if I recall, tend to cause the energy to drift downward. If
there are other things in your conditions that make the energy trend up,
plugging the SHAKE hole will unmask them. Another thing that will really
make your total energy nosedive is the nscm setting, resetting net
translational momentum every so often. There should be conservation of net
forces, but the Ewald grid doesn't really do that unless you have good
precision there, which generally means a grid density 30-50% larger than
Amber's default settings will produce. Beyond that, there can be things
about the particular system that even I would have to spend time figuring
out. Energy conservation is an important and strenuous test of an
algorithm, but with the complexity of modern codes and periodic simulations
of arbitrary topologies, it is somewhat harder to measure than most people
realize.
Dave
On Sat, May 9, 2020 at 11:53 AM Adrian Roitberg <roitberg.ufl.edu> wrote:
> To add to what Dave Case said, you need to be careful with shake
> setting. In principle, using shake for X-H bonds will make NVE not
> really conserve energy.
>
> To check this, set ntc and ntf not to be used, and try very small dt,
> maybe 0.5 fs or potentially lower.
>
> Adrian
>
> On 5/9/20 12:38 AM, Liao wrote:
> > [External Email]
> >
> > Dear Amber users,
> >
> > I realized my simulation temperature, and total energy, keeps on
> dropping with NVE ensemble. (ntt=0, ntb=1 periodic boundaries on). Isn’t it
> suppose to be constant total energy? What’s more mysterious, is that this
> only happens when I started using an Hydrogens-mass-repartitioned (HMR)
> parmtop file.
> > Has anyone else ran into such a situation?
> > Thank you.
> >
> >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.ambermd.org_mailman_listinfo_amber&d=DwIGaQ&c=sJ6xIWYx-zLMB3EPkvcnVg&r=dl7Zd5Rzbdvo14I2ndQf4w&m=hYRPmvhsZiJmX3GNTkop39O2LjBCyoV6wN3tqtWOdY8&s=5oc9jKoLFZ4IZf1h9FAHEd8fNZlw9ZiLrglVsPecVfU&e=
>
> --
> Dr. Adrian E. Roitberg
> V.T. and Louise Jackson Professor in Chemistry
> Department of Chemistry
> University of Florida
> roitberg.ufl.edu
> 352-392-6972
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun May 10 2020 - 22:30:02 PDT
