Splines in CP2K

Matthias Krack matthia... at psi.ch
Thu May 5 08:08:03 UTC 2016

Hi Christopher

if I am not mistaken, all nonbonded pair potentials are tabulated using 
splines in the MM pre-processing phase and these are then used as lookup 
tables during the simulation. I think this is a quite common approach which 
keeps the computational effort at the same level independent of the 
algebraic complexity of the actual pair potential. A control of the mapping 
(splining) accuracy is possible by CP2K input parameters (e.g. EPS_SPLINE). 
That might answer your first two questions at least partially. As far as I 
know, it is not possible to avoid that tabulation (splining) of the 
nonbonded pair potentials currently without a code change. The Poisson 
solver follows a different approaches depending on the chosen method.

It would be helpful if you would describe your problems with the splines 
more precisely. The splining of the pre-defined pair potentials works 
usually fine. You can activate SPLINE_DATA and SPLINE_INFO in the section 
CP2K_INPUT <https://manual.cp2k.org/cp2k-3_0-branch/CP2K_INPUT.html> / 
<https://manual.cp2k.org/cp2k-3_0-branch/CP2K_INPUT/FORCE_EVAL.html> / MM 
<https://manual.cp2k.org/cp2k-3_0-branch/CP2K_INPUT/FORCE_EVAL/MM.html> / 
This will print a data file for each nonbonded pair potential interaction 
(atomic kind pair) using splines.



On Wednesday, 4 May 2016 23:07:34 UTC+2, Christopher Mauney wrote:
> Hello,
> I'm running into a lot of problems with splines.  The documentation on 
> these is frustratingly imprecise and vague.
> Why does CP2K use splines by default in MM calculations?
> Are we modelling potentials as a spline? (That is, are we fitting a spline 
> based on finding the values of the potential on some underlying grid?)
> Can we turn off this behavior and just calculate the potential values 
> directly?
> What kind of spline is being used? Can we modify this?
> Does the Poisson solver use splines the same way?
> Any help here is appreciated.  I do have specific problems with input 
> files, but this is an aspect of CP2K I've had trouble with on several 
> projects and would prefer a general response.
> -Christopher Mauney
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20160505/689f5af4/attachment.htm>

More information about the CP2K-user mailing list