[gmx-users] Bad energy conservation.

[David]

On Wed, 2002-12-04 at 20:33, Ram'on Garc'ia Fern'andez wrote:
> On Wed, Dec 04, 2002 at 07:22:21PM +0100, David wrote:
> > rlist is the radius for neighboursearching, however rvdw is the max
> > distance for vanderwaals calculations and rcoulomb for coulomb.
> 
> Then, I understand that rlist should be higher than both rcoulomb
> and rvdw, so that when the forces are evaluated atoms outside these
> cutoffs can be discarded. By contrast, if rlist is lower than rcoulomb
> then some atoms will no be taken into account when they should.
if rlist < rcoul or rlist < rvdw, then a twin range setup is used, where
long range forces (rlist < r < rvdw) are only computed when you do
neighboursearching.

Note that doing neighboursearching every step kills your performance.
But sometimes you need it. Our attitude is always that when you do
temperature coupling or pressure coupling all other inaccuracies are
negligable. However, for water when using PME one does get very good
energy conservation, although I suspect that there is some systematic
compensation of errors.  nstlist  > 1 gives cooling in LJ systems,
whereas other error give heating. Nevertheless, at nstlist = 5 with PME
and 0.9 nm cut-off for water the total energy is very well conserved,
and since this is what most people use for e.g. protein simulations, all
is well...

-- 
Groeten, David.
________________________________________________________________________
Dr. David van der Spoel, 	Biomedical center, Dept. of Biochemistry
Husargatan 3, Box 576,  	75123 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://xray.bmc.uu.se/~spoel
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++