From: Mark Abraham (Mark.Abraham_at_anu.edu.au)
Date: Mon Mar 06 2006 - 05:37:10 CST
EPF (Esben Peter Friis) wrote:
> Hi again
>
> Even though they use the same force field, the simulation engines of
> NAMD and Charmm are not entirely identical (?). Even if you could read
> the velocities from the charmm run into NAMD, I believe it would still
> lead to a different result than continuing in Charmm. It is probably ok
> to do the equilibration in Charmm, and then reassign the velocities and
> do the production in NAMD. But, as Michel points out, it would probably
> be better to start from scratch, if it is not an enourmous amount of CPU
> time.
I've got lots of money that says you will not be able to give CHARMM and
NAMD the same starting positions+velocities and have them produce the
same simulation. The differences in implementation are a subtly
different model physics... even the order in which summation is done can
be enough to produce a small difference in one quantity that describes
the system. MD systems are chaotic, and over time, a small difference
will render the two simulations unrecognizable from each other. The
model physics are different - whether the difference is significant,
nobody has studied, I believe.
Also, much of the point of equilibration is to get a sensible ensemble
of starting velocities for the relevant conditions - the stochastic
sampling that is done when you generate them just can't be realistic. It
is self-defeating to equilibrate, and then throw away your velocities.
Mark
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:41:42 CST