[CP2K:955] Re: SCF goes wrong in QMMM after 37000 steps

[Teodoro Laino]

> &FORCE_EVAL
>   METHOD QMMM

First of all in the DFT section there is not decoupling section.. in  
this case your QM periodic images will be coupled with the
same periodicity of the QMMM cell. To be really consistent you may  
want in this case have EXACTLY the same cell between MM and QM.

>   &DFT
>       RESTART_FILE_NAME H2O-32-qmmm-1_32000.restart
>     BASIS_SET_FILE_NAME GTH_BASIS_SETS
>     POTENTIAL_FILE_NAME POTENTIAL
>     &MGRID
>       COMMENSURATE
>       CUTOFF 300
>     &END MGRID
>     &QS
>     &END QS
>     &SCF
>       SCF_GUESS atomic
>     &END SCF
>     &XC
>       &XC_FUNCTIONAL pade
>       &END XC_FUNCTIONAL
>     &END XC
>   &END DFT
>   &MM
>     &FORCEFIELD
>       parmtype CHM
>       parm_file_name charmm.pot
>       &CHARGE
>         ATOM OT
>         charge -0.8340
>       &END CHARGE
>       &CHARGE
>         ATOM HT
>         charge 0.4170
>       &END CHARGE
>     &END FORCEFIELD
>     &POISSON
>       &EWALD
>         EWALD_TYPE ewald
>         ALPHA .44
>         GMAX 25
> #        O_SPLINE 6
>       &END EWALD
>     &END POISSON
>   &END MM

Second.. your QMMM calculation does not specify any periodicity for  
the QMMM coupling. This means that the electrostatic coupling is  
evaluated
always according the PBC between QM and MM atoms. If an atoms is  
going out of the cell and is wrapped back in the cell (due to the  
PBC) than OF COURSE your system cannot conserve the total energy.
If you want a fully energy conserving QMMM-MD you've to specify that  
you want a QMMM periodic coupling (treating the electrostatic through  
an Ewald sum): have a look at the few templates in the tests  
directory (of course the number have been properly adapted, those are  
silly).

>   &QMMM
>     &CELL
>       ABC 10.0 10.0 10.0
>       UNIT ANGSTROM
>     &END CELL
>     ECOUPL GAUSS
>     &MM_KIND H
>       RADIUS 0.44
>     &END MM_KIND
>     &MM_KIND O
>       RADIUS 0.78
>     &END MM_KIND
>     &QM_KIND H
>       MM_INDEX 2 3
>     &END QM_KIND
>     &QM_KIND O
>       MM_INDEX 1
>     &END QM_KIND
>   &END QMMM
>   &SUBSYS
>     &CELL
>       ABC 9.865 9.865 9.865
>       UNIT ANGSTROM
>     &END CELL

Teo