[gmx-users] using one thermostat for entire system ??

[Mark Abraham]

On 8/03/2011 2:50 AM, Jennifer Williams wrote:
> Hi,
>
> I simulated the diffusion of small gases (CO2, N2) in a framework 
> structure which was mostly frozen with some mobile surface groups. I 
> applied a temperature thermostat to the entire system (i.e I didn't 
> couple the gas molecules and framework separately). I have now been 
> asked the following by a reviewer:
>
> "Please comment on any artefacts that might arise as a result of 
> non-equipartition of energies.  For example, what is the calculated 
> temperature of each of the gas species and mobile species?"
>
> I have tried to explore the temperature of each species separately but 
> g_energy will only give me the energy of the system as a whole. Are 
> there any other tools within gromacs to look at the temperature of 
> each species in turn from the md runs I already have?
>
> Does anyone have a suggestion as to how I can address the reviewer's 
> comment and proove that using one thermostat for the whole system is 
> OK (that is what I am hoping!).  Is there anything in particular that 
> I should be looking closely at?

g_energy can only report the global temperature as saved in the .edr 
file, computed during the simulation. The only way of decomposing the 
temperature is to save velocities to the .trr file with nstvout. (Some 
colleagues had to do this recently.) Then, judicious use of trjconv to 
take subsets and matching topology hacking (tpbconv doesn't quite work) 
mean you can use mdrun -rerun on the subset .trr and matching .tpr to 
get a group-wise temperature. That can demonstrate whether temperature 
differentials (ie. heat flows) exist.

If you didn't save velocities, you're out of luck.

Mark