[gmx-developers] rerun thermostated

[Berk Hess]

Michel Cuendet wrote:

>
> Hi Berk,
>
> Berk Hess wrote:
>
>> The rerun should be thermostatted, but of course in the proper way.
>
>
> No. I mean rerun with "mdrun -rerun traj.trr". Then the velocities are 
> read at sparse timesteps, and they are not recalculated, and there is 
> thus no reason to scale them. Whatever way the velocities in traj.trr 
> were obtained, mdrun -rerun should take them as they are, and 
> calculate the corresponding temperature, pressure, etc...
>
I don't agree completely.
You might want to rerun using frames from every step to recalculate the 
virial or such.
You then also want the proper temperature and pressure coupling.
Anyhow, if the coupling is done properly with the reread frames this is 
not an issue.
When no velocities are present for certain frames the pressure is 
meaningless anyway.

>
>> Are both the run and the rerun done with 3.3?
>
>
> No, but that was precisely the point : see what temperature 3.3 would 
> give for velocities which 3.2 believed were at 300K. (answer: 303.47 
> K, see the other thread). If I leave the NH thermostat on in the mdp 
> file for the rerun, the rerun gives me 303-something in the beginning 
> of the trajectory, and as the thermostat momentum builds up, starts to 
> scale the velocities back to 300 towards the end of the trajectory. 
> These scaled velocites don't make any sense.

Ah, I see your point.
For Nose-Hoover this goes wrong as you do not have the proper thermostat 
momentum.
This would be solved by reading an energy file or by storing the momentum in
the trajectory.

Berk.