Re: NVE simulations in NAMD - the right choice of timesteps and parameters

From: Brian Radak (
Date: Tue Nov 29 2016 - 10:04:53 CST

Depending on how you are checking energy drift, a "gotcha" is that
"COMmotion off" is the default, so that the kinetic energy due to the
center of mass is removed at step 0 - therefore the energy value that is
conserved is /not/ the same as that of the initial configuration unless
you previously used zeroMomentum on (which is not the default).

That being said, Felipe is almost certainly correct that multiple
time-stepping is the first thing to check here.


On 11/29/2016 07:03 AM, Felipe Merino wrote:
> Hi,
> Using multi-timesteps can be quite aggressive. Usually you don't see
> this change in energy because of the themostat, but in the NVE
> simulations they appear immediately. Why do you want to run NVE anyways?
> In any case, try for example using 2fs for everything and you'll see
> that the simulation becomes better behaved. Whether or not this is
> advisable for you depends on what you are trying to obtain from the
> simulation in the first place.
> Best
> Felipe
> On 29/11/16 11:54, Harish Srinivasan wrote:
>> I am trying simulate micellar systems like SDS, SDBS micelle in
>> water. I equilibrated the system for 20 ns using NPT Langevin piston.
>> Even after 20 ns run, the average temperature remained at 295 K for
>> the system (I am still pondering why this happened), so I used
>> rescale velocity (with rescalefreq 2 timesteps) for 500 ps.
>> After this I ran a NVE simulation (meaning, i switched off all the
>> thermostats and barostats) for about another 20 ns, the total energy
>> of the system seemed to change about 20000 units in the span of 20 ns.
>> I am using 1 fs timestep, with 2fs for non-bonded forces and 4fs for
>> full electrostatics. No rigidity imposed.
>> Thank you,
>> Harish

Brian Radak
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory
9700 South Cass Avenue, Bldg. 240
Argonne, IL 60439-4854
(630) 252-8643

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:22:39 CST