[gmx-users] Replica exchange temp. distribution

[Mark Abraham]

sarbani chattopadhyay wrote:
>   Hi,
>     I do understand that reducing the number of replicas will limit my 
> temperature space,
> but gromacs 3.3.2 can run 1 replica per node. Thus having 4 nodes , I 
> can have only 4
> replicas. Thus in case I want to span a temperature space of 313K to 
> 333K using only 4
> replicas, is there any possible method by which I can choose 4 optimal 
> temperature values,
> maintaining a decent exchange-probability value? 

The GROMACS manual and the website you cited both deal with this issue. 
An exponential distribution of T is normal for NVT, so find x such that 
313 * x^3 = 333.

Mark