Incorrect optimized geometry using CELL_OPT

[Diego López]

Dear CP2K users,

I am kind of new using cp2k but I have been using the software fo a couple of months. I am trying to optimize both geometry and cell parameters of different semiconductor material but the obtained geometries are distorted compared with experimental data or relaxed geometries obtained using Quantum Espresso or VASP. 

My optimized geometries have a lack of symmetry compared with experimental data, i.e. the atoms that should be aligned in an especific axis tend to alternate the positions in the crystal, so my results are no compatible with the expected ones.

Which parameters should I change in my calculations? I tried lowering the EPS_SCF to increase accuracy, increasing the supercell and using larger basis sets but the result does not change. I attatch an example of input that I used for these calculations.

&GLOBAL
  PROJECT
  RUN_TYPE CELL_OPT
  PRINT_LEVEL MEDIUM
&END GLOBAL

&FORCE_EVAL
  METHOD Quickstep              ! Electronic structure method (DFT,...)
  STRESS_TENSOR ANALYTICAL
  &DFT
    BASIS_SET_FILE_NAME  BASIS_MOLOPT
    POTENTIAL_FILE_NAME  GTH_POTENTIALS
    MULTIPLICITY 301
    UKS
    &POISSON                    ! Solver requested for non periodic calculations
      PERIODIC XYZ
    &END POISSON
     &XC
      &XC_FUNCTIONAL PBE
       &PBE
        PARAMETRIZATION PBESOL
       &END
      &END XC_FUNCTIONAL
     &END XC
    &SCF
      SCF_GUESS ATOMIC
      MAX_SCF 40
      EPS_SCF 3E-06
      &OT
      PRECONDITIONER FULL_KINETIC
      MINIMIZER CG
      &END OT
      &OUTER_SCF
       MAX_SCF 320
       EPS_SCF 3E-06
      &END
    &END SCF
    &MGRID
      NGRIDS 4
      CUTOFF 550
      REL_CUTOFF 90
    &END MGRID
    &QS
      METHOD GPW
      EXTRAPOLATION ASPC
      EPS_DEFAULT 1E-014
    &END
  &END DFT
  &SUBSYS
    &CELL
      ABC 3.601367151 4.818258107 25.0
      ALPHA_BETA_GAMMA 90 90 90
      PERIODIC XYZ
      SYMMETRY ORTHORHOMBIC
      MULTIPLE_UNIT_CELL 5 5 1
    &END CELL
    &TOPOLOGY                    ! Section used to center the atomic coordinates in the given box. Useful for big molecules
      COORD_FILE_FORMAT xyz
      COORD_FILE_NAME  ./crsbr.xyz
      MULTIPLE_UNIT_CELL 5 5 1
    &END
    &KIND Cr
      ELEMENT Cr
      BASIS_SET SZV-MOLOPT-SR-GTH
      POTENTIAL GTH-PBE-q14
      MAGNETIZATION 3.0
    &END KIND
    &KIND S
      ELEMENT S
      BASIS_SET SZV-MOLOPT-GTH
      POTENTIAL GTH-PBE-q6
    &END KIND
    &KIND Br
      ELEMENT Br
      BASIS_SET SZV-MOLOPT-SR-GTH
      POTENTIAL GTH-PBE-q7
    &END KIND
  &END SUBSYS
&END FORCE_EVAL
&MOTION
  &CELL_OPT
    TYPE DIRECT_CELL_OPT
    MAX_ITER 200
    OPTIMIZER LBFGS
    KEEP_SYMMETRY
  &END CELL_OPT
&END MOTION

Thanks in advanced

[Marcella Iannuzzi]

Hi, 

I would start by using better basis sets (e.g., DZVP, TZVP) and increasing the PW cutoff 

Regards

Marcella 

[Diego López]

Hi  Marcella,

Thanks for your answer. I tried with PW cutoff of 1000 and with TZVP basis set but the result is the same

Regards

Diego

[Lobna Saeed]

The minimizer DIIS is not reliable. This OT method is not always converging with me and when converges doesnt give accurate enough numbers. Therefore, please try to switch to diagonalization.

--
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/fb6271ad-4513-41a8-a08c-f20b5c23e849n%40googlegroups.com.

[Lobna Saeed]

And also please do energy convergence tests. Dont try to increase your pw cutoff values randomly. 

[Fabian Ducry]

Hi Diego,

are you sure that MULTIPLICITY 301 is correct?

Cheers,
Fabian

[Diego López]

Dear Fabian,

I am using this specific multiplicity because experimental data suggests that this is the magnetic ground state, but I am not obtaining the proper magnetic moments on the Fe atoms. Should I try using RELAX_MULTIPLICITY?

Regards,

Diego

[fabia...@gmail.com]

Hi Diego,

I have not used RELAX_MULTIPLICITY myself, you can give it a try. You could also use DIAGONALIZATION together with SMEAR then the spins are not fixed by default. What spin population did you obtain with vasp and QE?

Have you also checked that all the individual scf between the geometry optimization steps are converged?

Cheers,

Fabian

[Diego López]

Hi Fabian,

Thanks for your answer. My system is a semiconductor so I think that smearing is not a good choice for my calculations. Using QE I obtained a good approximation of the magnetic moments of the system and also a reasonable geometry optimization. What do you mean with convergence of individual scf steps?

Best,

Diego 

[fabia...@gmail.com]

Hi Diego,

In QE you have (almost certainly) used semearing too. It can be used with semiconductors just as well as with metals. Between each geometry optimization step there is an energy minimization, did these all converge?

Cheers

[Diego López]

Hi Fabian

I tried using smearing and now the magnetic moments and the PDOS look really nice! I will try to optimize the system with this set up. Thanks a lot for your comments.

Best,

Diego