Re: drop in Temp from NVT to NVE

From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Wed Jul 02 2008 - 10:49:15 CDT

If you're going to turn off rigidbonds, you should at least equilibrate
the system with a thermostat for a little bit before you start your nve
run. Why? You're adding a bunch of degrees of freedom to the system by
removing a bunch of constraints (the temperature is defined in terms of
the number of degrees of freedom, which changes when you turn off
rigidbonds, and the total kinetic energy, which doesn't), and you need
to allow the system to adjust to this. There's no law that says you
can't use rigidbonds with 1 fs timesteps, but you should in either case
finish equilibration your system using the same potential energy surface
and constraints that you intend to use for the production run.

Best,
Peter

S.K. Ghosh wrote:
> Hi NAMD users,
>
> I had run an NVT with 2 fs timestep (and rigidbonds) and the temp of
> the last step came to 308K. Then I ran an NVE with 1 fs timestep (and
> so removed rigidbonds), using the velocities of the last step of NVT
> run as the initializing temp. Interestingly, I saw that the velocity
> of 0th step of this NVE run was computed as 213K. Could you please
> give me an insight into why exactly this could have happened.
>
> I guess it has got to do with how kinetic energy is calculated in the
> case of rigidbonds. Does this also have something to do with reduction
> in timestep of integration?
>
> Thanks,
> Sourav

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:48:01 CST