Re: AMBER: NVE energy and temperature drift

From: Thomas Cheatham III <>
Date: Thu, 9 Aug 2007 16:27:50 -0600 (Mountain Daylight Time)

> I have run a couple of 20 ns NVE simulations of a protein in tip3p water
> ntt=ntp=0, ntc=ntf=2, ntb=1 and cut=10.
> Over 20 ns Etot rose absolutely linearly and with barely any random
> fluctuations from ~-76660 to ~-75880. Over the same period the

This is 1% drift in total energy over 20 ns and is of some concern...
Have you altered the PME parameters or are you running with a
non-conservative integration method like RESPA?

I would try to lower the SHAKE tolerance (0.000001) and to increase the
accuracy of the PME by reducing DSUM_TOL by an order of magnitude
(0.000001). If the NFFT values are close to the box size, I would up them
to the next product of powers of 2,3 or 5.

Of additional concern is whether there is any drift at restrt due to the
finite precision of the restrt file (however you report nearly linear
increase with no obvious periodicity related to restart). To avoid error
on restrt requires switching to binary restrts.

Finally, are the potential energies also changing significantly? Note
that if I were going to play with this, I would also turn on constant
pressure but set the coupling time to be longer than the simulation
(TAUP=99999999.9) so I could monitor changes in pressure too while still
effectively running NVE.

-- tom
The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Sun Aug 12 2007 - 06:07:42 PDT
Custom Search