Re: [AMBER] acceptance probabilities remd

From: Adrian Roitberg <>
Date: Tue, 08 Mar 2011 12:17:58 +0100

On 3/8/11 11:02 AM, Jorgen Simonsen wrote:
> Hi
> I am monitoring my remd simulation where I have used exponential
> scaling of the temperatures. I do not think I have gain detailed
> balance in my system for the acceptance probabilities. From the
> rem.log file I am count the same state for example my starting
> temperature 300 K becomes 70% as well as my end state temperature 500
> K 70 % - the temperature in between has probabilities of staying in
> the same state 40 % of the times now they have two neighbors so their
> exchange is the same ~ 30 % but to higher and lower temperature hence
> only 40% in its own state but should this be the same for all?

Short answer is 'no'. Let me elaborate. The exponentially distributed
temperatures should give you roughly an equal exchange rate between
neighbor temperatures under the assumption that the specific heat is
linear constant vs temperature. This is sometimes a correct assumption,
but rarely completely true, and it depends on the system, the size, etc.

So, variations for exchange rates are not wrong. Now, if they are very
different, take it as a chance to add a replica between two with low
exchange rates, or to take one out if two replicas exchange 'too much'.

I hope this helps.

> Any help appreciated,
> Thanks
> _______________________________________________
> AMBER mailing list

                            Dr. Adrian E. Roitberg
                              Associate Professor
                Quantum Theory Project, Department of Chemistry
                            University of Florida
on Sabbatical in Barcelona until August 2011.
AMBER mailing list
Received on Tue Mar 08 2011 - 03:30:02 PST
Custom Search