spline problem with GENPOT

Sal sria... at uci.edu
Sun Oct 11 18:53:42 CEST 2015


I was having kind of similar problem before. What I did was to do few steps 
of minimization on FIST or other MD packages.
Or see if having "EMAX_SPLINE 1.0E8" under your SPLINE directive is 
helpful. 
I've seen in the old posts people have also tried  NEIGHBOR_LIST option, 
under : FORCE_EVAL 
<http://manual.cp2k.org/cp2k-2_5-branch/CP2K_INPUT/FORCE_EVAL.html> / MM 
<http://manual.cp2k.org/cp2k-2_5-branch/CP2K_INPUT/FORCE_EVAL/MM.html> / 
NEIGHBOR_LISTS 
<http://manual.cp2k.org/cp2k-2_5-branch/CP2K_INPUT/FORCE_EVAL/MM/NEIGHBOR_LISTS.html>


On Friday, September 4, 2015 at 6:17:54 AM UTC-7, danie... at gmail.com 
wrote:
>
> I am doing alchemical transformations in FIST using GENPOT, and run into 
> an error when the potential that I specify in GENPOT becomes too “exotic”:
>
> ****************************************************************** 
> *** ERROR in build_neighbor_lists (MODULE fist_neighbor_lists) *** 
> ****************************************************************** 
>
> *** GEOMETRY wrong or EMAX_SPLINE too small! *** 
>
> Typically, this is because atoms are “too close” together, but given the 
> potential that I specified (which contains partially annihilated 
> particles), there *should* occasionally be atoms very close together. The 
> only solution that I have found is to add the following to the input 
> section:
>
> &SPLINE 
> R0_NB 0.00001 
> NPOINTS 500000 
> &END SPLINE 
>
> The problem is that, even at very high values of "NPOINTS", there are 
> still convergence problems. Is there a better way to do this? 
>
> Thanks, 
>
> Daniel Sadowsky 
> Ecole Polytechnique Fédérale de Lausanne 
> Environmental Chemistry Modeling Laboratory 
>
> EPFL ENAC IIE LMCO 
> GR C2 555 (Bâtiment GR) 
> Station 2 
> CH- 1015 Lausanne 
> +41 21 69 38074 
> danie... at epfl.ch <javascript:>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20151011/39fb35bc/attachment.html>


More information about the CP2K-user mailing list