[gmx-users] RE: RE : FEP : separating components of dgdl

[Berk Hess]

----------------------------------------
> Date: Tue, 29 Jan 2008 10:24:57 +0100
> From: michael.bon at cea.fr
> To: gmx-users at gromacs.org
> Subject: RE : FEP : separating components of dgdl
> 
> 
> Thanks for the quick reply.
> 
> I have an other question : I ran a quick simulation on a single node using mdrun -sepdvdl and the components are written in
> the .log file. But the final value of dVpot/dlambda doesn't seem to be the sum of the components above. Example for some step:
> 
> VdW and Coulomb SR : dVdl -2.72859e+02
> Angle :              dVdl  3.56537e+00
> Proper Dih. :        dVdl -7.31260e-03
> Ryckaert-Bell. :     dVdl -1.13654e-01
> LJ-14 + Coulomb-14 : dVdl  6.39499e+02
> Dispersion correc. : dVdl  0.00000e+00
> 
> And I read : dVpot/dlambda = 3.62310e+02, instead of 3.70e+02.
> 
> There is no PME in this sim.
> 
> What did I forget ?

I assume you are running single processor.
The terms are per processor.

I think that the only contributions that are no added are:
constraints, only when you change constraint lengths,
kinetic energy, only when you change masses.

So do you change constraint lengths or masses?

Berk.

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/