[gmx-users] Pressure annealing in Gromacs?

[Mark Abraham]

Jussi Lehtola wrote:
> On Thu, 2009-12-03 at 02:30 +1100, Mark Abraham wrote: 
>> Jussi Lehtola wrote:
>>> I'm experiencing trouble converging the density of some heavy liquid
>>> alcohols (after 10 ns of simulation the density is still changing
>>> linearly). Is there any way to run pressure annealing in Gromacs?
>>> Running the system through a high pressure and temperature might give an
>>> equilibrium structure quicker.
>> There's nothing native that I'm aware of, but it should be 
>> straightforward to use sed or perl in a script to do the annealing "by 
>> hand" in a series of mdrun invocations.
> 
> Yup, that came to mind but then one ends up with N trajectories and log
> files. And one has to run a big pile of grompp commands, that waste some
> parallel run time, so I really would like something integrated.

One of the hidden beauties of checkpoint files might be that you can 
prepare all the .tpr off-line, and just match them up while running, and 
use mdrun -append to concatenate files. I'd give that idea a solid test 
before using on something I cared about though! If it worked, the only 
inferiority would be repetitious parallel setup.

> Besides, a thing like changing the reference pressure should be a
> no-brainer in the source code - assuming the code is well structured and
> one knows it well beforehand.

Agreed. For all I know, there are some state variables in the above 
checkpoint solution that shouldn't be carried over when the reference 
pressure changes, anyway.

> Anyway, the problem was solved for now already. Still, pressure
> annealing would be a nice addon to gromacs. (I'll open a bug on bugzilla
> about it.)

Mark