[gmx-users] Nose-hoover T-coupling in REMD

[ms]

On 24/12/10 10:36, David van der Spoel wrote:
> On 2010-12-24 11.04, Mark Abraham wrote:
>> On 22/12/2010 5:44 PM, Qin Qiao wrote:
>>> Dear all,
>>>
>>> I posted yesterday but didn't get answer..I guess it's due to my wrong
>>> approach to ask. I would like to explain more and hope I would get
>>> your help.
>>>
>>> I'm doing an REMD, and want to have a strong temperature coupling to
>>> make sure every replica is at its target temperature. I guess I could
>>> do it by setting tau_t small in the Nose-hoover T coupling, but there
>>> will come the warning in Gromacs.
>>>
>>> I maxwarned the warning to proceed, and find the temperature is ok,
>>> but I am not sure whether it will introduce some artifacts in the
>>> simulation and what kind of artifacts. Could you tell me? Thanks a lot.
>>
>> The temperature coupling algorithms are discussed in the literature,
>> when they were derived (citations in GROMACS manual) and subsequently. I
>> suggest you do your own homework there. The use of REMD is more or less
>> irrelevant to the trade-offs when deciding what temperature coupling
>> scheme to use.
>>
> That is correct, but do not use Berendsen! There is a recent paper from
> the Garcia group (IIRC) about the effect of Berendsen coupling on REMD
> sampling. Not good.
>> Mark

I'm not an expert, but isn't Berendsen usually not used because it 
doesn't give a correct ensemble? I may be partial because I personally 
know Giovanni Bussi, but it seems from what I've heard that v-rescale is 
the best choice usually.

-- 
Massimo Sandal, Ph.D.
http://devicerandom.org