[gmx-users] Re: Langevin thermostat implementation in GROMACS vs AMBER

[Julian Garrec]

Le 08/05/13 19:33, Christoph Junghans a écrit :
>> Date: Wed, 08 May 2013 16:34:46 +0200
>> From: Julian Garrec <julian.garrec at epfl.ch>
>> Subject: Re: [gmx-users] Langevin thermostat implementation in GROMACS
>>          vs AMBER
>> To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>> Message-ID: <518A6286.8020504 at epfl.ch>
>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>>
>> No suggestion ?? :)
>>
>> In my post I forgot to say that my system contains explicit water.
>>
>> I'm still very curious about the difference in implementation in AMBER
>> and GROMACS. And I still don't really understand the meanning of tau-t
>> in LD and how  it is related to other (well defined) quantities such as
>> the friction coeficient in equation 3.114 of the manual (v4.6.1).
> I am not really sure what your question is, section 7.3.3 is pretty
> clear about that:
> "the inverse friction constant for each group is set with tau-t", so
> gamma_i=1/tau-t, which also makes sense unit-wise as gamma has units
> 1/ps and tau-t ps
>
> bd-fric is only used for the Brownian dynamics integrator and if
> bd-fric=0, the friction is calculated as mass/tau-t. It might be a bit
> misleading that bd-fric is listed under section 7.3.4 (Langevin
> dynamics), but that stems from the fact that Brownian dynamics is the
> overdamped case of stochastic dynamics and hence it is Langevin
> Dynamics, too, just in a different limit.
>
> Christoph
Dear Christoph,

Thank you very much for your reply. Actually I am not so much confused 
about the theory :) , but more about how it is implemented, and how this 
implementation differs in GROMACS vs AMBER.

Yes, BD is a limiting case of LD, but still both are implemented with 
different integrators in GROMACS (integrator=sd/bd). From the manual, it 
is clear that both {integrator=bd} and {integrator=bd} can be combined 
with the option bd-fric=0, so I am not so sure that the usage of bd-fric 
is restricted to BD...

Anyway, you are right that it make sense that "gamma_i=1/tau-t", so I 
will keep using {integrator=sd ; bd-fric=0 ; tau-t=XX}.

Many thanks again,

Julian


>> Please help,
>>
>> Sincerely,
>>
>> Julian
>>
>>
>> Le 04/05/13 13:42, Julian Garrec a 閏rit :
>>> Dear all,
>>>
>>> I am trying to use stochastic dynamics in GROMACS to reproduce some
>>> previous results that I got with the AMBER software using the Langevin
>>> thermostat.
>>>
>>> In AMBER, one typically sets what they call the "collision frequency"
>>> (noted gamma_ln in the AMBER manual). I typically set gamma_ln to 2
>>> collisions per ps (significantly smaller than the internal friction of
>>> water, since I use Langevin dynamics as a thermostat).
>>>
>>> 1) In the mdp options manual, it is recommended to use {integrator=sd
>>> ; bd-fric=0 ; tau-t=2}. My understanding is that bd-fric corresponds
>>> to 'm_i*gamma_i' in equation 3.114 of the theoretical part of the
>>> manual (version 4.6.1). Am I right ?
>>>
>>> 2) Then, is {integrator=sd ; bd-fric=0 ; tau-t=2} equivalent to
>>> {integrator=sd ; bd-fric=0.5} ? (In the theoretical part of the
>>> manual, it is said that gamma_i should be set to 0.5 to follow the
>>> same spirit).
>>>
>>> 3) Is 'bd-fric' of GROMACS equivalent to 'gamm_ln' of AMBER ?
>>>
>>> Thanks for your help,
>>>
>>> Julian
>>>
>>
>> --
>> Julian Garrec, research associate
>>
>> Web: https://sites.google.com/site/juliangarrec/
>>
>> 蓂uipe : Chimie et Biochimie Th閛riques
>> Unit�Mixte de Recherches CNRS UHP 7565
>> Universit�de Lorraine, Nancy I BP 239,
>> 54506 Vandoeuvre-l鑣-Nancy, cedex France
> --
> Christoph Junghans
> Web: http://www.compphys.de


-- 
Julian Garrec, research associate

Web: https://sites.google.com/site/juliangarrec/

Équipe : Chimie et Biochimie Théoriques
Unité Mixte de Recherches CNRS UHP 7565
Université de Lorraine, Nancy I BP 239,
54506 Vandoeuvre-lès-Nancy, cedex France