[gmx-users] Re: Re: temperature and pressure values

[Mark Abraham]

Marino Convertino wrote:
> Dear Mark and gmxers,
> thanks a lot for your reply.
> I am checking md.log file after 5-days-computing-time of the same 
> simulations.

5 more days? Or just 5 days starting from an unequilibrated starting 
point? Since we don't know how good your computers are, please report MD 
simulation time, not wall time.

Have you tried g_energy like I suggested last time?

> About temperature and pressure I find these values:
>  
> Temperature    Pressure (bar)
>  3.54285e+02    6.54609e+01
>  
> Temperature    Pressure (bar)
>  3.52770e+02   -3.39396e+01
>  
> Temperature    Pressure (bar)
>  3.54570e+02   -4.24080e+01
>  
> Temperature    Pressure (bar)
>  3.54669e+02    4.29965e+01
>  
> Temperature    Pressure (bar)
> 3.51791e+02   -1.49698e+01
>  
> And so on...
> Are these variations ok?
> I wonder if temperature value ( constantly 2-4K over the fixed value) 
> can be considered good.

I would think not.

> And what about pressure variation?

The range of average values (if they are average values) looks too large 
for an equilibrated system.

> I use these options in my NPT simulations:
>  
> Tcoupl              =  Berendsen
> tc-grps              =  Protein  SOL
> tau_t                =  0.1      0.1    ;time in ps
> ref_t                =  350      350
> Pcoupl                =  Berendsen
> Pcoupltype          =  isotropic
> tau_p                  =  1.0
> compressibility    =  4.5e-5
> ref_p                   =  1.0

That's fine.

> Should I modify any option in order to obtain a better convergence?
> Should I use a different electrostatic treatment (PME)?
> Sorry for the long message.

I think you are not using an appropriate regime for equilibration. 
Something along the lines of energy minimization to relieve bad 
contacts, equilibration at constant volume to get the temperature and 
velocity distribution roughly correct, and then switch to constant 
pressure for some more equilibration. Use g_energy and visual inspection 
with n_gmx to check that these stages appear to be working reasonably.

Mark