[gmx-users] angular momentum.

[Erik Lindahl]

Nguyen Hoang Phuong wrote:
>>>>Dear All,
>>>>
>>>>the simulations of a peptide in vacuum are performed using NVE
>>>>ensemble. In principal, the angular momentum of the molecule should be
>>>>conserved but in my simulations, its magnitude  increases as a function of
>>>>time. Does anyone know why?
>>>
>>>
>>>Yes, it is caused by a build-up of integration and/or round-off
>>>errors, which unfortunately is inevitable in a computersimulation.
>>>Try David's suggestion...
>>>
>>
>>Actually, it's probably due to your neighborlist not being longer than 
>>the cutoff; I'm adding some code to improve this for 3.1.5 (or 3.2, we
>>might have changed enough things to bump the release).
> 
> 
> the simulations run with: 
> rlist = 1.
> rcoulonb =1.
> rvdw = 1.
> PME was used.
> I wonder if these values cause the problem? After 10ps, the magnitude of
> the angular momentum is increased by a factor of 5, the timestep is
> 0.5fs and double precision is switched on. Within this short
> simulation time, does the numerical error of intergrations is big enough?
> 

For strict energy conservation you will need to use switch,shift or PME, 
and have rlist about 0.2 nm longer than the cutoff. For some reason we 
checked against this in grompp, but I've fixed it in CVS, and it we will
have a new release with it soon.

I'll run a couple of angular momentum tests; for now I've only been 
checking energy conservation, and that works perfect over at least 10 ns.

Cheers,

Erik