[gmx-users] Re: problem with free energy calculations

Michael Brunsteiner mbx0009 at yahoo.com
Tue Feb 6 21:55:06 CET 2007

Thanks for your comments!

Berk Hess wrote:

> One "obvious" problem is that using two thermostats is incorrect.
> In this way the com of the water with respect to the com of the rest
> will not be sampled correctly (both are fixed).

I am not sure I understand what you mean, is the COM reset 
for each temperature group seperately or not ?? 
(I used nstcomm = 100) ... when the systems are still 
interacting, i.e., when lambe < 1 I wouldn't expext 
this to be a problem, while when the water is completely 
de-coupled (lambda=1) then I actually NEED to use two 
thermostates plus a finite nstcomm, since otherwise I 
might end up with a water thats either much too "hot" 
or too "cold", or a tumbling ice-cube ...

> The solution is using the sd integrator.

There must be a way to do that with a normal verlet algorithm
or the like ... the sd integrator, i.e. the Langevin dynamics, was
meant to be used with implicit solvent, after all. There is
this friction term that's meant to mimic the effect of the
explicit water, and even if you let the friction constant zeta
be very large, i.e., if you do away with the friction term, you
just end up with a primitive kind of thermostate, and i can't
see a reason why this should be any better than Nose Hoover !?

> Another issue is that since water is strongly hydrogen bonding
> and the hydrogen bond interaction is rapidly soft-cored,
> you probably need much closer spaced lambda point at lambda=0
> (i.e. use an inhomogeneous spacing).

that would mean to forgo one of the advantages of softcore
potentials altogether, since this is exactly why we use 
them: to spare ourselves the extensive sampling required
when approaching lambda=1 (where the interactions between
particle and the rest approach zero, and dA/dlambda 

I've simulated the very same system in CHARMM, with
soft core potentials plus NH thermostate, and 21 equally
spaced windows, and got perfectly reasonable results.
can it be that the small LJ interaction parameters
that CHARMM gives the TIP3P water hydrogens (for reasons
related to a completely different problem, and in fact
irrelevant for the resulting (free) energies) makes
all the difference ??

> There could be other issues, but one would only find those
> after studying the results in detail.

that's what i am doing at the moment.


Need a quick answer? Get one in minutes from people who know.
Ask your question on www.Answers.yahoo.com

More information about the gromacs.org_gmx-users mailing list