[Wien] optimization of Bader program

[Javier Daniel Fuhr]

Dear Jorissen,

> I'm wondering whether you feel there is still room for optimization of 
> the efficiency of the Bader program.  For SURF/IRHO, aim is taking 
> many, many hours.
>
>  
>
There some optimizations that could be done in the Boder program:
1) the first I can think of is the algorithm to follow the gradient 
field ("follow" and "onesptep" routines). Although I used an adaptive 
method that varies the step length, some other methods could be more 
efficient. One could also think of using the Hessian of the density (if 
the cost of calculating it, is compensated by less steps). Also, more 
tweaking of the fixed parameters (that I put by hand from my experience) 
could improve the efficiency. The problem is, as always, to not lower 
the robustness while changing the parameters.
2) another more general optimization I can think of is that if one wants 
to calculate the charge of all atoms in the unit celd, one could use the 
information of the already calculated atomic surfaces, for the 
calculation of the rest of the surfaces. One should device an 
intelligent way of using this information.

> Eg., I think that often the program searches in which MT a point 
> belongs, while it ought to know perfectly, since the point was 
> designed to be in a particular sphere.  But I am not 100% sure of this 
> observation.
>
>  
>
I am not sure if I have understood the observation, but the method used 
for the surface calculation can be resume as:
For each spherical angle (theta,phi) search for the radius R at which
1) starting from a smaller radius r<R and following the gradient we 
arrived at the corresponding atom
2) starting from a larger radius r>R and following the gradient we 
arrived at another atom
This function R(theta,phi) defines the atomic surface.
When we follow the gradient we never now at which atom we will arrive, 
and also in some cases a portion of the Bader surface is inside the MT.

> Please share your views on this; I'm sure that (especially the 
> authors) are best placed to judge where optimization could be achieved.
>
> Thank you,
>
> Kevin Jorissen.
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>  
>
Regards,
Javier

----------------------------------------------------------
Javier Daniel FUHR
CRMCN-CNRS
Campus de Luminy, Case 913
13288 Marseille Cedex 9 - FRANCE

e-mail: fuhr at crmcn.univ-mrs.fr