[gmx-users] Vritual Sites and simulation time-step

[Mark Abraham]

On Fri, May 31, 2013 at 5:19 PM, James Starlight <jmsstarlight at gmail.com>wrote:

> 3) Also I've already performed small simulation with the time step 5 fs
> (defining virtual sites by means of pdb2gmx -vsites  and introducing heavy
> hydrogens. I've not observed any errors during my simulation. When I've
> tried to increase dt up to 7 fs I've obtained warnings according to that I
> should increase tau_p (from 2 ps with Parinello coupling). How tau_p is
> relevant to the timestep and when I could obtain more information about
> such constants ? E.g if i increase tau_p doest it means that I provide
> weaker coupling to the system with pressure bath ? (higher constant- higher
> relaxation times )
>

You're integrating forces calculated at one instant and assuming they are
applicable over a finite time period. Obviously this is an approximation,
and it gets worse with longer time step. The worse approximation means the
sampling is from a shadow Hamiltonian that is even further from the real
one. P-R is not very stable if the sampling is not from equilibrium, so
probably the change suggested in the warning helps keep P-R working.

The absence of the simulation crashing is not a suitable proxy for assuming
there are no problems. You should compare observables with those from
conservative parameter sets and see whether differences exist.

Mark