Re: [ESPResSo] MMM2D method

[Axel Arnold]

On Tuesday 16 February 2010 00:48:00 Salvador Herrera Velarde wrote:
> Hi All,
>
> I am using the MMM2D method in a charged system with two wall constraints:
> constraint wall normal 0 0 1 dist 2.5 type 9
> constraint wall normal 0 0 -1 dist 34  type 9
>
> Although the particles remain between the wall-constraints, I am receiving
> the next message: some particles left the assumed slab, precision might be
> lost

The constraints are not the problem, MMM2D basically doesn't see them. What 
the error message means, is that during the near formula calcuation, two 
particles were further apart in z than L_y/2, which should not happen. While 
tuning, MMM2D actually checks that your layers are defined such that cannot 
occur. So, somehow the box geometry for MMM2D after tuning changes, which it 
should be able to handle, but apparently isn't. Can you send me your script? 
Then I can try to figure out why.

Regarding how big the errors can be: that depends highly on |z|. There is a 
small gap until which the losses are negligible, but very quickly the 
Bernoulli sum in the near formula will be outside its convergence radius, and 
then the errors are disastrous (that is, can be in the order of 10^5 easily). 

Axel

-- 
Dr. Axel Arnold Tel: +49 711 685 63593
ICP, Universität Stuttgart      Email: address@hidden
Pfaffenwaldring 27
70569 Stuttgart, Germany