[gmx-users] "Justifying" 4fs production runs after 1fs equilibrations?

[Tsjerk Wassenaar]

Hi,

Highly unlikely. The point of equilibration is to move from some point near
> the ensemble, to some point in the ensemble, perhaps without perturbing the
> starting structure significantly. The last phase of equilibration should
> match the production ensemble (or simply be subtracted from the production
> run), but this is not likely to be critical, either.
>

This may be splitting hairs, but there is no outside of the ensemble. A
structure can't be far from or near to the ensemble. There are regions with
low probability and there will be regions which cannot be reached normally,
because there are insurmountable barriers. Only if you've sampled, you can
end up with an ensemble that doesn't match the proper one.

In relation to the issue at hand, any level of simulation will be a trade
off between speed and accuracy. Yes, there is the Hamiltonian of your
system, if it were 'real'. But no simulation will sample that 'real
ensemble', not even with QM. Using a 1 fs time step and double precision
will give a good classical approximation, but you'll miss some details
because you want to stretch the time limit. The same is true for using a 2
fs time step, a 4 fs time step or even coarse graining. You can sample
things on time scales and system sizes you can't do with a lower level, but
you'll miss out on some details.

Nice to have a discussion on MD principles once in a while :)

Cheers,

Tsjerk