Energy of atomic oxgyen and smearing

[Matthias Krack]

Dear Natalie

for a non-periodic calculation, you have to specify "PERIODIC none" in the 
&CELL *and* in the &DFT section using for instance MT as the Poisson solver 
for the electrostatic decoupling:

    &POISSON
     PERIODIC none
     POISSON_SOLVER MT
    &END POISSON

which is needed for the calculation of isolated systems.
That won't solve, however, your convergence problem for the O atom. The 
DIIS method employed with diagonalisation is a very efficient for SCF 
convergence acceleration, but it is also very sensible to numerical noise. 
That issue can be fixed by increasing the CUTOFF, e.g. from 400 to 600 Ry. 
I have attached the output of such a run.
You should also keep in mind that the convergence criterion applied for 
diagonalisation is different to the OT criterion. OT checks the energy 
convergence whereas the largest change in a density matrix element between 
two consecutive SCF steps is checked when using diagonalisation. Thus 
EPS_SCF can be 10-100 times larger with diagonalisation compared to OT. An 
EPS_SCF value of 1.0E-7 results in an energy convergence almost within 
machine precision as you can see in the attached output.

Best regards

Matthias

On Thursday, 14 July 2016 04:05:20 UTC+2, Natalie Austin wrote:
>
> Hello Matthias,
>
> The input file is below:
>
> &FORCE_EVAL
>   METHOD QS
>   &DFT
>     BASIS_SET_FILE_NAME BASIS_MOLOPT
>     POTENTIAL_FILE_NAME GTH_POTENTIALS
>     &MGRID
>       CUTOFF 400
>       REL_CUTOFF 60
>       NGRIDS 4
>     &END MGRID
>     &QS
>        EPS_DEFAULT 1.0E-14
>        MAP_CONSISTENT
>     &END QS
>   &SCF
>      SCF_GUESS  ATOMIC
>      EPS_SCF 1.0E-7
>      MAX_SCF 500
>      CHOLESKY INVERSE
>      &DIAGONALIZATION
>           ALGORITHM STANDARD
>      &END DIAGONALIZATION
>      &MIXING
>           METHOD DIRECT_P_MIXING   
>           ALPHA   0.1      *# I also ran a case where I increased the 
> value to 0.4 as you suggested. *
>           BETA    1.5
>           NBROYDEN  8
>      &END MIXING
>   &END SCF
>     &XC
>       &XC_FUNCTIONAL PBE
>       &PBE
>       PARAMETRIZATION REVPBE
>       &END PBE
>       &END XC_FUNCTIONAL
>       &VDW_POTENTIAL
>          DISPERSION_FUNCTIONAL PAIR_POTENTIAL
>          &PAIR_POTENTIAL
>            TYPE DFTD3
>             REFERENCE_FUNCTIONAL PBE
>             CALCULATE_C9_TERM .TRUE.
>             PARAMETER_FILE_NAME dftd3.dat
>             R_CUTOFF 15.0
>             VERBOSE_OUTPUT .TRUE.
>          &END PAIR_POTENTIAL
>       &END vdw_POTENTIAL
>     &END XC
> MULTIPLICITY 3
> UKS .TRUE.
>   &END DFT
>   &SUBSYS
>     &CELL
>       ABC  20.00000  20.00000  20.00000
>       ALPHA_BETA_GAMMA    90.00    90.00    90.00
>       PERIODIC NONE
>     &END CELL
>     &TOPOLOGY
>       COORD_FILE_NAME o.xyz
>       COORDINATE XYZ
>     &END TOPOLOGY
>     &KIND O
>       ELEMENT   O
>       BASIS_SET DZVP-MOLOPT-SR-GTH-q6
>       POTENTIAL GTH-PBE-q6
>     &END KIND
>   &END SUBSYS
> &END FORCE_EVAL
> &GLOBAL
>   PROJECT o
>   RUN_TYPE ENERGY
>   PRINT_LEVEL MEDIUM
> &END GLOBAL
> &MOTION
>   &GEO_OPT
>     MAX_FORCE 0.0004
>     MAX_ITER 2000
>     OPTIMIZER BFGS
>     TYPE MINIMIZATION
>   &END GEO_OPT
> &END MOTION
>
>
> and the coordinate file:
>        1
>  o
> O         7.46880        7.39185        7.47705
>
>
>
> Natalie
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20160714/4e719634/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: O.out
Type: application/octet-stream
Size: 67816 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20160714/4e719634/attachment.obj>