[CP2K:1887] Geometry optimization convergence problem

[Teodoro Laino]

Dear Peter,
an additional remark on your input file:

>     &POISSON
>       PERIODIC NONE
>       PSOLVER WAVELET
>       &EWALD
>         EWALD_TYPE NONE
>       &END EWALD
>     &END POISSON
>   

The above section (POISSON) makes no sense since you don't ask for any 
specific kind of EWALD (so this
means that the Coulomb interactions are computed by pair 1/r using the 
cutoff of the NB interactions) in addition with
a WAVELET solver.
If you want full electrostatic you need to choose a proper EWALD_TYPE.
Keep in mind that none of the decoupling techniques (either MT or 
WAVELET or others) work in conjuction
with MM FF. With classical run you must do everything fully periodic.

Cheers
Teo