Problem converging SCF for a system containing Fe2+

[salome.ll...@gmail.com]

Dear all, 

I am quite new to CP2K and working with transition metals. 

I am trying to reproduce some results of an O2 molecule dissociating from an heme group (Fe2+). I am struggling to converge the SCF wavefunction even in a singlepoint calculation and I am not sure what I am doing wrong here. 

As an initial guess, I am using PBE with the following combination of basis sets and pseudopotentials: 

- Fe: BASIS_SET DZVP-MOLOPT-SR-GTH & GTH-PBE-q16. 

- All the other atoms: DZVP-MOLOPT-GTH with GTH-PBE-q* . 

I have attached the input file and the initial geometry.

Can you help me out? 

Thank you very much in advance, 

Salome

[Pierre-André Cazade]

Hi Salome,

 

Two things come to my mind, first you should specify the charge and multiplicity of your system. Then, the cutoff and rel_cutoff might be too small.

 

Actually, there’s a third: why mixing MOLOPT and MOLOPT-SR? Just go with the latter, they usually work fine.

 

Regards,

Pierre

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/f894e8c6-c46a-417b-83a6-6343cf0aff64n%40googlegroups.com.

[salome.ll...@gmail.com]

Hi Pierre, 

Thank you so much for your answer, 

I have added all your suggestions:

- I am using DZVP-MOLOPT-SR-GTH for all atoms. 

- I have added CHARGE 0 and MULTIPLICITY 2 ( I am not completely sure about this but I have 55 atoms.).

- I have expanded the Cutoff as follows:

 &MGRID

      CUTOFF 400

      REL_CUTOFF 60

      NGRIDS 5

    &END MGRID

I still have the same problem. I am worried because the Convergence of the SCF optimisation up instead of decreasing. 

 Step     Update method      Time    Convergence         Total energy    Change

  ------------------------------------------------------------------------------

     1 P_Mix/Diag. 0.40E+00   25.1     1.97116897      -352.7972578538 -3.53E+02

     2 P_Mix/Diag. 0.40E+00   22.1     1.15920278      -353.8713026582 -1.07E+00

     3 P_Mix/Diag. 0.40E+00   22.5     0.70468277      -354.5020638040 -6.31E-01

     4 P_Mix/Diag. 0.40E+00   22.5     1.81519648      -354.7042785765 -2.02E-01

     5 P_Mix/Diag. 0.40E+00   22.6     1.40667909      -353.4830224789  1.22E+00

     6 P_Mix/Diag. 0.40E+00   22.6     1.61367789      -352.6169777702  8.66E-01

     7 P_Mix/Diag. 0.40E+00   22.8     9.77791081      -349.7222190083  2.89E+00

     8 P_Mix/Diag. 0.40E+00   22.7     3.04606642      -338.8207963019  1.09E+01

     9 P_Mix/Diag. 0.40E+00   22.9    10.99358257      -288.8580020934  5.00E+01

    10 P_Mix/Diag. 0.40E+00   22.9     5.98455238      -271.1871548265  1.77E+01

...

And after this remains between 4 and 5 for a long time. 

Thanks, 

Salome

[Matt W]

Reduce the ALPHA value in &SCF / &MIXING

https://manual.cp2k.org/cp2k-8_1-branch/CP2K_INPUT/FORCE_EVAL/DFT/SCF/MIXING.html

the default is much to large and causes the oscillations.

OT is often a faster / more reliable solver for non-metals.

Matt

[Pierre-André Cazade]

Hi Salome,

I agree with Matt. Try OT:

and if you still prefer to use the default solver, use Broyden Mixing instead of P-mixing.

    &SCF
      SCF_GUESS  ATOMIC
      &MIXING
        METHOD BROYDEN_MIXING
        ALPHA 0.4
        BETA 0.5
        NBROYDEN 8
        NMIXING 2
      &END MIXING
    &END SCF

It should be CHARGE 2 and MULTIPLICITY 0.

     &SCF
       MAX_SCF  20
       SCF_GUESS  ATOMIC
       &OT  T
         MINIMIZER  CG
         PRECONDITIONER  FULL_ALL
       &END OT
       &OUTER_SCF  T
         MAX_SCF  100
       &END OUTER_SCF
     &END SCF

Try increasing the CUTOFF to 500 or 600.

Regards,

Pierre

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/b43fc558-0c90-4f56-b932-3e9f71f880f1n%40googlegroups.com.

[Pierre-André Cazade]

Hi Salome,

Sorry, basd copy/paste. Here is a much better email.

I agree with Matt. Try OT:

[salome.ll...@gmail.com]

Dear Matt and Pierre

I have changed to OT and it works! :) 

Thanks a lot, 

Best wishes, 

Salome