Re: [ESPResSo-devel] Nsquare cellsystem and folded positions

[Axel Arnold]

On Monday 13 August 2007 16:26, Olaf Lenz wrote:
> Axel Arnold wrote:
> > Why not????? For forces, energy etc. the distances are correctly folded,
> > as far as I can remember.
>
> Hmm, apparently you are not of the same opinion as Hanjo in that
> respect. However, my tests showed that at least for constraints, the
> force computation does not work correctly. I'll provide an example
> script soon.

Ok, to clarify that: The pair forces use the minimum image distance, both in 
domdec and nsquare. For the constraint forces, for some reason we use the 
internally folded coordinates. That can indeed cause a problem, but actually 
for both cellsystems. E.g. consider an excluded sphere close to the boundary. 
Again, the example:

setmd box_l 5 5 5
part 0 pos 0 0 0 v -0.1 0 0

inter 0 0 lennard-jones 1.0 1.0 1.12246 0.25 0

constraint sphere center 4 0 0 radius 1 direction outside type 0

setmd time_step 0.01
thermostat off
setmd skin 1
for {set i 0} {$i < 100} {incr i} {
    integrate 10
    puts [part 0 pr folded]
}

The particles goes unperturbed right into the sphere, then gets folded, and 
Espresso crashes :). And that is domaindecomposition.

So, we have a bug treating constraints, but that actually applies to all 
cellsystems.

Axel

-- 
Dr. Axel Arnold
FOM Institute for Atomic and Molecular Physics
Kruislaan 407                                   Phone: +31 20 6081 275
1098 SJ Amsterdam, The Netherlands              E-mail: address@hidden