Convergence problem transition metal oxide

Diego López

unread,

Jul 3, 2023, 4:15:37 AM7/3/23

to cp2k

Dear CP2K users,

I am currently working with a series of transition metal oxides and CP2K is working fine but when I try to converge a calculation on the mixed valence Cu oxide (Cu3O4) my calculations are not converging. I have read that this oxide is a tricky system for DFT, so I would be really grateful if you could give me some suggestions to solve this problem. I attatch the input file that I am using. I tried to use different mixing, diagonalization and smearing methods with the same problem.

&GLOBAL
PROJECT Co3O4
RUN_TYPE ENERGY
PRINT_LEVEL MEDIUM
&END GLOBAL

&FORCE_EVAL
METHOD Quickstep
&DFT
BASIS_SET_FILE_NAME BASIS_MOLOPT_UZH
POTENTIAL_FILE_NAME POTENTIAL_UZH
UKS
MULTIPLICITY 1
&POISSON
PERIODIC XYZ
&END POISSON
&XC
&XC_FUNCTIONAL PBE
&PBE
PARAMETRIZATION PBESOL
&END
&END XC_FUNCTIONAL
&END XC
&MGRID
NGRIDS 4
CUTOFF 800
REL_CUTOFF 50
&END MGRID
&QS
METHOD GPW
EXTRAPOLATION ASPC
EPS_DEFAULT 1E-014
&END QS
&SCF
SCF_GUESS ATOMIC
EPS_SCF 1.0E-7
ADDED_MOS 20
MAX_SCF 20
&OUTER_SCF
MAX_SCF 300
EPS_SCF 1.0E-7
&END
&SMEAR ON
METHOD FERMI_DIRAC
ELECTRONIC_TEMPERATURE [K] 3000
&END SMEAR
&DIAGONALIZATION
ALGORITHM STANDARD
&END DIAGONALIZATION
&MIXING
METHOD BROYDEN_MIXING
ALPHA 0.02
BETA 0.01
&END MIXING
&END SCF
&END DFT
&SUBSYS
&CELL
ABC 5.938 5.945 5.938
ALPHA_BETA_GAMMA 60 60 60
PERIODIC XYZ
&END CELL
&TOPOLOGY
COORD_FILE_FORMAT xyz
COORD_FILE_NAME ./Co3O4-2.xyz
&END
&KIND O
ELEMENT O
BASIS_SET DZVP-MOLOPT-PBE-GTH-q6
POTENTIAL GTH-PBE-q6
&BS
&ALPHA
N 2
L 1
NEL +2
&END ALPHA
&BETA
N 2
L 1
NEL +2
&END BETA
&END BS
&END KIND
&KIND Co
ELEMENT Co
BASIS_SET DZVP-MOLOPT-PBE-GTH-q17
POTENTIAL GTH-PBE-q17
&BS
&ALPHA
N 4 3
L 0 2
NEL -2 -1
&END ALPHA
&BETA
N 4 3
L 0 2
NEL -2 -1
&END BETA
&END BS
&DFT_PLUS_U
L 2
U_MINUS_J [eV] 2.0
&END DFT_PLUS_U
&END KIND
&KIND Co1
ELEMENT Co
BASIS_SET DZVP-MOLOPT-PBE-GTH-q17
POTENTIAL GTH-PBE-q17
&BS
&ALPHA
N 4 3
L 0 2
NEL -2 -3
&END ALPHA
&BETA
N 4 3
L 0 2
NEL -2 +3
&END BETA
&END BS
&DFT_PLUS_U
L 2
U_MINUS_J [eV] 2.0
&END DFT_PLUS_U
&END KIND
&KIND Co2
ELEMENT Co
BASIS_SET DZVP-MOLOPT-PBE-GTH-q17
POTENTIAL GTH-PBE-q17
&BS
&BETA
N 4 3
L 0 2
NEL -2 -3
&END BETA
&ALPHA
N 4 3
L 0 2
NEL -2 +3
&END ALPHA
&END BS
&DFT_PLUS_U
L 2
U_MINUS_J [eV] 2.0
&END DFT_PLUS_U
&END KIND
&END SUBSYS
&END FORCE_EVAL

Thank you in advanced.
Best regards,
Diego.

Marcella Iannuzzi

unread,

Jul 3, 2023, 5:36:33 AM7/3/23

to cp2k

Dear Diego

A box of 5.938 5.945 5.938 is very small and not properly representing the bulk system, if only Gamma point is considered.

Often for transition metal oxides DFT+U turns out to help, probably you can find in the literature whether this is the case of the oxides you are interested in.

Regards

Marcella

Diego López

unread,

Jul 3, 2023, 6:46:13 AM7/3/23

to cp2k

Hello Marcella,

Thank you for your answer. I have tried also with bigger supercells (3x3x3 and 5x5x5) but there is not convergence. Also I tried to use different sets of U - J values (2 eV seems to be wide used in bibliography) but the result is the same. Any further recommendations?

Regards,

Diego

Marcella Iannuzzi

unread,

Jul 3, 2023, 6:55:27 AM7/3/23

to cp2k

Hi Diego

It is hard to guess what is going on without more information.

Probably you need to explore more values of the U+J parameters or a different way to initialise the magnetization.

Regards

Marcella

Diego López

unread,

Jul 3, 2023, 9:54:39 AM7/3/23

to cp2k

Hi Marcella,

Thanks for your answer. I think you are right because non spin-polarized calculations are converging but the spin-polarized does not, so I assume there is a problem with magnetization. I am using &BS section to describe the valence electrons of Co. Do you think I should move to MAGNETIZATION keyword?