[gmx-users] Questions regarding REMD simulation

[Mark Abraham]

On 18/09/2011 9:35 PM, michael zhenin wrote:
>
> Hi All,
>
> I have few questions regarding REMD simulation.
>
> I assume REMD works in the following way: Each replica starts with a 
> given temperature and the temperature of each replica changes 
> along the simulation (as described in Sugita and Okamoto publication, 
> 1999).
>

As Sebatian noted, in GROMACS the replicas move between simulations and 
the temperature of each simulation is constant. You will see this in the 
temperatures in the .log files. This approach is conceptually identical, 
of course, but there are advantages for doing one or the other when 
writing the code.

> Then demux.pl <http://demux.pl/> and trjcat collect the trajectories 
> of the lowest temperature from each replica.
>

They are already present in the lowest-numbered replica trajectory. 
Using demux.pl will be counter-productive - it generates geometrically 
continuous trajectories from ensemble-continuous trajectories (and vice 
versa, I think).

> My problem is that while analyzing the results, I found in the edr 
> files that the temperatures along each replica are relatively
>
> constant (+- 2K from the starting temperature) and the average 
> temperature is equal to the initial temperature.
>

That's consistent with GROMACS REMD simulating a constant ensemble. You 
can't do demux on the energy files in the current implementation.

Mark

>
> How can this be explained? Does the temperature of each replica change 
> along the simulation or maybe the temperature of each
>
> simulation stays close to its initial value and the coordinates are 
> changed between the different replica? (if so, why do i need to use
>
> the post-processing tools such as the demux.pl <http://demux.pl/> script?)
>
>
> Thanks,
>
>
> Michael
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20110919/9c627f89/attachment.html>