[gmx-users] exchanged pairs in the log file of a remd simulation

[Berk Hess]

> Date: Sun, 3 Feb 2008 10:46:28 +0100
> From: spoel at xray.bmc.uu.se
> To: gmx-users at gromacs.org
> Subject: Re: [gmx-users] exchanged pairs in the log file of a remd simulation
> 
> OZGE ENGIN wrote:
> > Hi all,
> > 
> > I am performing replica exchange MD simulations. Here, I attach the part of a log file which gives info about the exchanges made at step 46000 and 47000. It is just a test run. As you can see, on the line 2, it is indicated that the exchange is made between the rep 1 and rep 2 or between the repl2 and repl3. The case is the same for other time steps.
> > There are no other types of pairs. 
> > 
> > Why?
> > 
> Other exchanges are not attempted, although this is not entirely 
> correct. However since the probability for exchanges is ver small when 
> the energy overlap is small this is not a serious approximation. In your 
> case it seem like you have quite high probabilities however, indicating 
> that your temperature spacing is quite small. In that case exchanges 1-3 
> could be possible, but these are not attempted in gromacs.

This is not a matter of correctness.
The scheme in Gromacs in completely correct.
The pairs that are chosen for attempt are not chosen based on their energy or exchange probability,
so there is no bias.
It is only a matter of efficiency.
It would be slightly more efficient to try to exchange all pairs.
But then you do get into matters of correctness.
When I first try to exchange replicas 1 and 2, the result depends on the energies of replicas 1 and 2,
an exchange of 2 and 3 will therefore not only depend on the energies of 2 and 3, but of 1, 2 and 3.
Maybe this is not incorrect, but to avoid this issue and to simplify the practicalities of exchanging the states,
I decided to alternate the odd and even pair attempts.

Berk.


_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20080203/2e1d1dc5/attachment.html>