SCF does not converge for a single atom of cerium

[Jonathan Campeggio]

Dear cp2k developers,

 I am trying to test the calculation using a single atom of cerium. The coordinates should be ok, I am using a single atom at the origin of a box of 5 A of side length. Also both basis set and pseudopotential should be ok, I have found them in the data directory. I do not understand why the SCF does not converge. Surely, there is an error of mine but I cannot see it. I have put the muliplicity to 3, because in its electronic configuration Ce has 2 unpaired. electrons. This is my input.

&GLOBAL
  PROJECT Ce
  RUN_TYPE ENERGY
  PRINT_LEVEL LOW
&END GLOBAL
&FORCE_EVAL
  METHOD QS
  &DFT
    BASIS_SET_FILE_NAME BASIS_MOLOPT_UCL
    POTENTIAL_FILE_NAME GTH_POTENTIALS
    CHARGE 0
    MULTIPLICITY 3
    &MGRID
      CUTOFF 550
      REL_CUTOFF 80
    &END MGRID
    &QS
      EPS_DEFAULT 1.0E-12
      EXTRAPOLATION PS
      EXTRAPOLATION_ORDER 3   
    &END QS
   &SCF
      SCF_GUESS RESTART
      EPS_SCF 1.0E-7
      MAX_SCF 500
      &MIXING T
        METHOD BROYDEN_MIXING
        ALPHA 0.4
        NBROYDEN 8
      &END MIXING
      
      &DIAGONALIZATION
        ALGORITHM STANDARD
      &END DIAGONALIZATION
    &END SCF
   &XC
       DENSITY_CUTOFF     1.0000000000000000E-10
       GRADIENT_CUTOFF     1.0000000000000000E-10
       TAU_CUTOFF     1.0000000000000000E-10
       &XC_FUNCTIONAL  NO_SHORTCUT
        &PBE
            PARAMETRIZATION ORIG
        &END PBE
       &END XC_FUNCTIONAL
     &END XC  
    &POISSON
      PERIODIC NONE
      POISSON_SOLVER MT
    &END POISSON
  &END DFT
  &SUBSYS
    &CELL
      ABC 5. 5. 5.
      PERIODIC NONE
    &END CELL
    &COORD
    Ce 0.00 0.00 0.00
    &END COORD
&KIND Ce
      ELEMENT   Ce
      BASIS_SET DZVP-MOLOPT-SR-GTH-q30
      POTENTIAL GTH-PBE-q30
    &END KIND
&END SUBSYS
&END FORCE_EVAL

Have a nice day,

Jonathan

[Matt W]

Couple of comments. This is actually a rather difficult calculation to get correct - the full solid state calcs will probably me much easier to converge.

(i) your grid cuttoff is probably far to low for a q30 pseudo. You could also try uing GAPW method (METHOD GAPW in the QS section)

(ii) Your cell is probably too small for the MT poisson solver try increasing the size of the cell. WAVELET solver may be more efficient for atomic calcs.

(iii) ALPHA might be too big - try reducing it. Maybe then restart and increase a little, or try OT method.

Matt

[Gabriel Reynald]

Hi Jonathan,

Did you try to use smearing? It's very useful for metallic systems. See the final of the tutorial of Energy and Force here https://www.cp2k.org/howto:static_calculation. There's shows how to use it.