Re: amber-developers: energy conservation

From: David A. Case <>
Date: Tue, 15 Jan 2008 00:00:58 -0800

On Mon, Jan 14, 2008, Jun Wang wrote:
> recently I am checking energy conservation of the program sander and
> pmemd. I use the topology and restart file in amber9/benchmark/jac. I
> turned off the temperature and pressure coupling. Shake is also turned
> off. But I found the total energy of the isolated system drops gradually.
> Who can help me to explain why this happened?
> the following is the input file I used:
> &cntrl
> ntx=7, irest=1,
> ntc=1, ntf=1, tol=0.0000001,
> nstlim=25000,
> ntpr=1, ntwr=10000,
> dt=0.0005, vlimit=0.0,
> cut=9.,
> ntt=0, temp0=300.,
> /
> &ewald
> nfft1=64,nfft2=64,nfft3=64,
> skinnb=2.,
> /

Even with a half fs time step, it will be hard to get a good simulation using
TIP3P water with SHAKE turned off. First, the TIP3P potential is designed
for SHAKE, and will give very unphysical results if SHAKE is removed.
Secondly, would have to re-equilibrate everything (for more than 10ps I think)
to get a good test of energy conservation.

You say the "total energy of the isolated system", but you are not running an
isolated system (the jac benchamrk is a periodic system). How much does the
energy (or temperature) change?

Amber does have small energy drifts, but these are generally acceptably small.
For example, I have run 100 ns of NVE on a protein-water system (with SHAKE!)
and seen the system heat up by less that 0.5 K. This is certainly not
perfect, and there are ways to probably get better conservation, but that is
not the only desirable feature of simulations.

...hope this helps....dac
Received on Wed Jan 16 2008 - 06:07:23 PST
Custom Search