[gmx-users] Temperature in Brownian Dynamics simulation

[Suman Chakrabarty]

Dear developers,

could any of you please confirm whether one has to use certain general
thermostat in brownian dynamics (or stochastic dynamics) simulation or
"tcoupl = no" is justified? The manual says for stochastic dynamics
"the parameter tcoupl is ignored" (manual for ver. 4, page 141,
section 7.3.3). I would guess the same would be true for bd as well.

If "tcoupl = no" should be alright, then why the temperature obtained
from g_energy is almost double of the ref_t parameter? I have checked
this for a system consisting of a single polymer chain in vacuum.

If one has to use a general thermostat indeed, then please advise
which thermostat would be best appropriate for BD/SD simulations.


Thanks,
Suman.


On Fri, Apr 17, 2009 at 4:55 PM, Suman Chakrabarty
<suman at sscu.iisc.ernet.in> wrote:
> On Fri, Apr 17, 2009 at 5:32 AM, Mark Abraham <Mark.Abraham at anu.edu.au> wrote:
>> Suman Chakrabarty wrote:
>>>
>>> Hi,
>>>
>>> it was pointed out before that for Brownian Dynamics simulation, the
>>> temperature calculated by g_energy does not match with the ref_t in
>>> parameter file and it is almost double:
>>> http://www.gromacs.org/pipermail/gmx-users/2007-June/028321.html ...
>>> but unfortunately I find no follow-up to that post.
>>>
>>> So I want to ask the question again here. I am doing Brownian Dynamics
>>> simulation of a single polymer chain consisting of 100 atoms. While I
>>> have set 300K as my simulation temperature,
>>
>> I don't think you have. You've generated velocities at that temperature, but
>> you're not thermostatting to that temperature, since you have "Tcoupl = no"
>
>
> Are you sure? The manual says for stochastic dynamics "the parameter
> tcoupl is ignored" (manual for ver. 4, page 141, section 7.3.3). I
> would guess the same would be true for bd as well. For both sd and bd
> there should be a different Langevin thermostat applicable, where the
> temperature will be incorporated through the Gaussian distributed
> random force term.
>
> So I shall still need further clarification regarding this. Do I
> really need to mention a general MD thermostat? Won't it be ignored as
> mentioned for sd? If not, which thermostat is best for doing bd
> simulations?
>
>
> Regards,
> Suman.
>
>
>>> when I try to use g_energy
>>> to check the temperature it reports the following:
>>>
>>> Energy                      Average       RMSD     Fluct.      Drift
>>>  Tot-Drift
>>>
>>> -------------------------------------------------------------------------------
>>> Temperature                 624.188    44.0862    44.0862 -1.52693e-06
>>>  -0.152693
>>> T-Protein                   624.188    44.0862    44.0862 -1.52693e-06
>>>  -0.152693
>>>
>>>
>>> Here is the content of my mdp file:
>>>
>>> =============================================================================
>>> integrator          =  bd
>>> dt                  =  0.002    ; ps !
>>> nsteps              =  50000000 ; total 100 ns.
>>>
>>> nstxout             =  1000
>>> nstvout             =  0
>>> nstfout             =  0
>>> nstlist             =  10
>>> nstlog              =  100
>>>
>>> rlist               =  1.0
>>> rvdw                =  1.0
>>>
>>> bd_fric             =  0
>>> ld_seed             =  -1
>>>
>>> constraints         = all-bonds
>>> lincs_order         = 8
>>> lincs_iter          = 8
>>>
>>> Tcoupl              =  no
>>> tau_t               =  0.01
>>> tc-grps             =  Protein
>>> ref_t               =  300
>>>
>>> Pcoupl              =  no
>>>
>>> gen_vel             =  yes
>>> gen_temp            =  300
>>> gen_seed            =  173529
>>> =============================================
>>>
>>> Please help me troubleshoot this problem. Thanks.
>>>
>>>
>>> Best regards,
>>> Suman Chakrabarty.