[gmx-users] Re: REMD ... again a problem with analysis (on the fly)

[Berk Hess]

>From: Luca Mollica <YLU at zurich.ibm.com>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: gmx-users at gromacs.org
>Subject: [gmx-users] Re: REMD ... again a problem with analysis (on the 
>fly)
>Date: Tue, 25 Oct 2005 10:50:08 +0200
>
>Hi,
>
>the temperature distribution for my system is the following:
>
>250.0 254.0 258.3 263.0 267.9 273.2 278.8 284.7 291.0 297.5 304.4 311.6
>319.1 327.0 335.1 343.6 352.4 361.5 371.0 380.7 390.8 401.2 411.9 423.0
>434.3 446.0
>
>But I would like to point out a thing about my simulation that could help
>to find a solution to my problems: my system contains 17.000 (more or
>less) water molecule: could this fact influence in any way  the proper
>work of REMD ?? I am trying to set up a simulation with a denser replica
>space and a less temperature difference: also the closest replica have a
>difference of about 20 kJmol-1 .....

Ah, that explains it.
In the new manual, which is unfortunately not yet online, I have written
a section on REMD.
For watery systems you should increase the temperature by a fraction
of roughly 1/sqrt(Ndf), where Ndf is the total number of degrees of freedom.
Taking only your 17000 waters you get: 1/sqrt(6*17000)~0.003
So from 250 K you should not go to more than 251 K.

Berk.