From: Jérôme Hénin (jerome.henin_at_uhp-nancy.fr)
Date: Mon Jun 27 2005 - 03:31:34 CDT
On Sunday 26 June 2005 14:41, Floris Buelens wrote:
> My approach now takes a configuration file parameter
> vdw_cap, and any troublesome energy value exceeding
> this vdw_cap (I'm arbitrarily using 20) is simply cut
> off and set to vdw_cap, thereby discarding the worst
> of the clashes.
> That's the plus side, but is this all a bit crude? Am
> I losing too much information by just sweeping
> shifty-looking vdW numbers under the carpet?
Well, the very reason for the end-point issue is that the very low (large
negative) delta_E terms bring a significant contribution to the FEP average,
while they are also rarely sampled.
What your modified code does amounts more or less to truncating the
distribution of delta_E, thereby discarding meaningful terms from the
average. So yes, you are losing some information.
Additionally, for the FEP equation to be valid, you have to do the calculation
with the same Hamiltonian you use to run the dynamics, otherwise you're not
sampling the right ensemble.
In short, I can't see how we could possibly get away without having the actual
soft-core potential properly coded. But lo! thou needst not despair, for some
shall work on this. Hopefully, a time will come when no such hacks are
necessary anymore. :)
-- Jérôme Hénin Equipe de Dynamique des Assemblages Membranaires Université Henri Poincaré / CNRS Tel : (33) 3 83 68 43 95 Fax : (33) 3 83 68 43 87 http://www.edam.uhp-nancy.fr/
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:39:37 CST