AMBER Archive (2008)

Subject: RE: AMBER: QM/MM Heating

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Tue Jan 29 2008 - 08:49:48 CST


Hi Steve,

> > One key thing to remember
> > is that MD simulations will only conserve energy for
> equilibrated systems
> > which yours was not, the density was way too low so this
> may well have
> > contributed to the problems you see.
>
> I was under the impression that, as long as the system
> doesn't have any
> 'hot-spots' (overlapping atoms/massive forces) and the timestep is
> chosen reasonably well (small enough to sample the quickest
> oscillations
> well but not so small that large roundoff errors begin to
> creep in) then
> an NVE simulation should approximately conserve the system's total
> energy, i.e. there will be fluctuations but no long-term drift (since
> the numerical approximation is the exact solution to a
> 'nearby' Hamiltonian)

I believe that this is true but my understanding was that generally the
further a system is from equilibrium the shorter the time step you require.
However, this of course is related to the size of the forces being seen.
However, I have never tested this to destruction to find out where the
limits are so perhaps others with longer experience running MD simulations
can comment here. It is certainly possible that QM/MM is more sensitive than
purely classical although I have never looked for such evidence either
theoretically or in practice.

> I ran a pure MM simulation on the input files I sent before,
> using SHAKE
> and a large 2fs time step, which from what other have said is on the
> verge of instability, and there seemed to be no long term
> energy drift.
> This seems to suggest that my system was equilibrated enough not to
> cause major problems.
 
> Or am I missing something? If not, then should a system be 'even more
> equilibrated' for a QM/MM run, i.e. are QM/MM runs more sensitive than
> pure MM to non-equilibrated conditions?

To be honest I don't know the answer to this question. At first glance one
would assume not, however, it is very possible that QM/MM is more sensitive
but I think this would qualify as a research question.

One other thing to note is that I have been playing around with QM/MM for
Replica Exchange and the langevin thermostat does a much much better of
temperature distribution than Berendsen in cases where you have a small QM
region that is not covalently bound to the MM region.

All the best
Ross

/\
\/
|\oss Walker

| Assistant Research Professor |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
| http://www.rosswalker.co.uk | PGP Key available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu