Cholesky decomposition failed for oxide surface geometry optimization

341 views

Skip to first unread message

Nisa Ulumuddin

unread,

Jul 22, 2020, 8:14:07 PM7/22/20

to cp2k

Hi,

I am quite new to CP2K and I do not know why I am failing to complete an scf cycle for a Cu2O(111) surface.

The error I am getting is :

Number of orbital functions: 4536

Number of independent orbital functions: 4536

Extrapolation method: initial_guess

*******************************************************************************

* ___ *

* / \ *

* [ABORT] *

* \___/ Cholesky decompose failed: the matrix is not positive definite or *

* | ill-conditioned. *

* O/| *

* /| | *

* / \ fm/cp_fm_cholesky.F:94 *

*******************************************************************************

===== Routine Calling Stack =====

8 cp_fm_cholesky_decompose

7 make_basis_sm

6 calculate_first_density_matrix

5 scf_env_initial_rho_setup

4 init_scf_run

3 qs_energies

2 qs_forces

1 CP2K

And my input file is:

&FORCE_EVAL
METHOD Quickstep
&DFT
UKS T
!PLUS_U_METHOD MULLIKEN #_CHARGES
MULTIPLICITY 1
!CHARGE 1
SURFACE_DIPOLE_CORRECTION
SURF_DIP_DIR Z
POTENTIAL_FILE_NAME /opt/apps/cp2k/6.1/gcc/7.3.0/openmpi/3.0.0/cp2k-6.1.0/data/POTENTIAL
BASIS_SET_FILE_NAME /opt/apps/cp2k/6.1/gcc/7.3.0/openmpi/3.0.0/cp2k-6.1.0/data/BASIS_MOLOPT
WFN_RESTART_FILE_NAME nu-RESTART.wfn
&MGRID
CUTOFF 400
NGRIDS 4
&INTERPOLATOR
!KIND SPLINE_3
&END INTERPOLATOR
&END MGRID
&QS
EPS_DEFAULT 1.0E-10
EXTRAPOLATION PS
!EXTRAPOLATION_ORDER 2
EXTRAPOLATION_ORDER 2
EPS_PGF_ORB 1E-10
&END QS
&SCF
!CHOLESKY OFF
SCF_GUESS RANDOM
! SCF_GUESS RESTART
EPS_SCF 1.0e-6
MAX_SCF 400
&OUTER_SCF
OPTIMIZER SD
EPS_SCF 1.0E-6
MAX_SCF 10
STEP_SIZE 0.1
&END
&OT T
!CHOLESKY OFF
MINIMIZER CG
PRECONDITIONER FULL_SINGLE
LINESEARCH 3PNT
&END
&PRINT
!&RESTART
! FILENAME =RESTART
!&END
&END PRINT
&END SCF
&XC
&XC_GRID
XC_SMOOTH_RHO NN50
&END XC_GRID
&XC_FUNCTIONAL PBE
&PBE
PARAMETRIZATION PBESOL #optimized for solids and surfaces
&END PBE
&END XC_FUNCTIONAL
!&vdW_POTENTIAL
!DISPERSION_FUNCTIONAL PAIR_POTENTIAL
!&PAIR_POTENTIAL
! TYPE DFTD3
! REFERENCE_FUNCTIONAL PBE
! PARAMETER_FILE_NAME /global/homes/g/gbcoll/dftd3.dat
! R_CUTOFF [angstrom] 15.0
! VERBOSE_OUTPUT T
!&END PAIR_POTENTIAL
!&END vdW_POTENTIAL
&END XC
&POISSON
&EWALD
EWALD_ACCURACY 1.00e-06
&END EWALD
POISSON_SOLVER ANALYTIC
PERIODIC XY
&END POISSON
&PRINT
&END PRINT
&END DFT
&SUBSYS
&CELL
A 18.3111991882 0.0000000000 0.0000000000
B -9.1555995941 15.8579636708 0.0000000000
C 0.0000000000 0.0000000000 28.7215003967
PERIODIC XY
&END CELL
&COORD
@include 'supercell.vasp.xyz'
&END COORD
&KIND Cu
BASIS_SET DZVP-MOLOPT-SR-GTH-q11
POTENTIAL GTH-PBE-q11
!&DFT_PLUS_U T
! L 2
! U_MINUS_J [eV] 8.0
!&END DFT_PLUS_U
!MASS 8.0
&END KIND
&KIND O
BASIS_SET DZVP-MOLOPT-SR-GTH
POTENTIAL GTH-PBE
!MASS 8.0
&END KIND
&END SUBSYS
&END FORCE_EVAL
&GLOBAL
PROJECT nu
RUN_TYPE ENERGY_FORCE
PRINT_LEVEL LOW
WALLTIME 01:00:00
&END GLOBAL

err.tar

Sam Broderick

unread,

May 11, 2022, 4:15:22 AM5/11/22

to cp2k

OT may be running into trouble with Cu. For metals, it is not recommended. However, this is a compound...

I ran into a similar issue with an Au cluster and addressed it by using Davidson diagonalization.