Re: [CP2K:4452] Re: Error in charges increases every OUTER_SCF cycle.

[John J. Low]

Flo,

I still has this issue with a cutoff of 800 Rydbergs in a 4x4x4 box (which converges the energy to 1 mh).

I solved my problem by using a different preconditioner (FULL_S_INVERSE) and increasing the cutoff to 1600 Rydbergs (or 800 Hartrees, which might have been what you meant). The error in the charge density does not increase every OUTER_SCF cycle after these changes.

It seems like the preconditioners based on diagonalization have this issue. I am not sure if this should have been expected from the theory or is a implementation issue.

Thanks for your help. Getting the OUTER_SCF working will be a big help to me.

John J. Low

[Florian Schiffmann]

Hi John,

there is another parameter to tune. FULL_ALL uses the ENERGY_GAP option, which is used as a scaling factor (usually improves convergence). Setting this to a higher value (0.5) could help there as well.

Flo