SCF Convergence Failure
125 views
Skip to first unread message
Pavlo Kostetskyy
Oct 16, 2019, 2:22:16 PM10/16/19
to cp2k
Dear Colleagues,
Thank you in advance for any input!
I am looking to converge a geometry optimization for Ni-exchanged zeolite system. The unit cell has been optimized without issues. Below is the input and relevant output. All following SCF steps only oscillate between positive and negative energies
Thank you in advance for any input!
-Pavlo
&GLOBAL
PROJECT AFI
RUN_TYPE GEO_OPT
PRINT_LEVEL MEDIUM
&END GLOBAL
&FORCE_EVAL
METHOD QS
&DFT
BASIS_SET_FILE_NAME GTH_BASIS_SETS
POTENTIAL_FILE_NAME POTENTIAL
CHARGE 0
MULTIPLICITY 1
&MGRID
CUTOFF 500
REL_CUTOFF 60
NGRIDS 4
&END MGRID
&QS
EPS_DEFAULT 1.0E-14
MAP_CONSISTENT
&END QS
&SCF
SCF_GUESS ATOMIC
EPS_SCF 1.0E-8
MAX_SCF 500
ADDED_MOS 200
CHOLESKY INVERSE
&SMEAR ON
METHOD FERMI_DIRAC
ELECTRONIC_TEMPERATURE [K] 300
&END SMEAR
&DIAGONALIZATION
ALGORITHM STANDARD
&END DIAGONALIZATION
&END SCF
&XC
&XC_FUNCTIONAL PBE
&PBE
PARAMETRIZATION ORIG
&END PBE
&END XC_FUNCTIONAL
&VDW_POTENTIAL
DISPERSION_FUNCTIONAL PAIR_POTENTIAL
&PAIR_POTENTIAL
TYPE DFTD3
REFERENCE_FUNCTIONAL PBE
CALCULATE_C9_TERM .TRUE.
PARAMETER_FILE_NAME dftd3.dat
R_CUTOFF 15.0
VERBOSE_OUTPUT .TRUE.
&END PAIR_POTENTIAL
&END vdw_POTENTIAL
&END XC
&POISSON
POISSON_SOLVER PERIODIC
&END POISSON
&END DFT
&SUBSYS
&CELL
A 14.05700000 0.00000000 0.00000000
B -7.02874500 12.17414345 0.00000000
C 0.00000000 0.00000000 17.46000000
PERIODIC XYZ
&END CELL
&TOPOLOGY
COORD_FILE_NAME afi_supercell_2Al_anti_higherer.xyz
COORDINATE XYZ
&END TOPOLOGY
&KIND Al
ELEMENT Al
BASIS_SET TZVP-GTH
POTENTIAL GTH-PBE-q3
&END KIND
&KIND Si
ELEMENT Si
BASIS_SET TZVP-GTH
POTENTIAL GTH-PBE-q4
&END KIND
&KIND O
ELEMENT O
BASIS_SET TZVP-GTH
POTENTIAL GTH-PBE-q6
&END KIND
&KIND Ni
ELEMENT Ni
BASIS_SET DZVP-MOLOPT-SR-GTH-q18
POTENTIAL GTH-PBE-q18
&END KIND
&END SUBSYS
&END FORCE_EVAL
&MOTION
&GEO_OPT
MAX_FORCE 0.0004
MAX_ITER 2000
OPTIMIZER BFGS
TYPE MINIMIZATION
&END GEO_OPT
&END MOTION
***********************************************************
SCF WAVEFUNCTION OPTIMIZATION
Step Update method Time Convergence Total energy Change
------------------------------------------------------------------------------
Trace(PS): 784.0000000000
Electronic density on regular grids: -783.9999999995 0.0000000005
Core density on regular grids: 784.0000000000 -0.0000000000
Total charge density on r-space grids: 0.0000000005
Total charge density g-space grids: 0.0000000005
1 P_Mix/Diag. 0.40E+00 2.1 8.22481733 -1901.7313217467 -1.90E+03
Trace(PS): 784.0000000000
Electronic density on regular grids: -783.9999999995 0.0000000005
Core density on regular grids: 784.0000000000 -0.0000000000
Total charge density on r-space grids: 0.0000000005
Total charge density g-space grids: 0.0000000005
2 P_Mix/Diag. 0.40E+00 3.0 7.21038392 -1900.7645680671 9.67E-01
Trace(PS): 784.0000000000
Electronic density on regular grids: -783.9999999995 0.0000000005
Core density on regular grids: 784.0000000000 -0.0000000000
Total charge density on r-space grids: 0.0000000005
Total charge density g-space grids: 0.0000000005
3 P_Mix/Diag. 0.40E+00 2.9 5.21542654 -1899.9828419152 7.82E-01
Trace(PS): 784.0000000000
Electronic density on regular grids: -783.9999999995 0.0000000005
Core density on regular grids: 784.0000000000 -0.0000000000
Total charge density on r-space grids: 0.0000000005
Total charge density g-space grids: 0.0000000005
4 P_Mix/Diag. 0.40E+00 2.9 402.76710057 -1895.0291997541 4.95E+00
Trace(PS): 783.9999999999
Electronic density on regular grids: -783.9999999995 0.0000000005
Core density on regular grids: 784.0000000000 -0.0000000000
Total charge density on r-space grids: 0.0000000005
Total charge density g-space grids: 0.0000000005
5 P_Mix/Diag. 0.40E+00 3.0 1907.73544022 -1877.8977759403 1.71E+01
Trace(PS): 783.9999999999
Electronic density on regular grids: -783.9999999995 0.0000000005
Core density on regular grids: 784.0000000000 -0.0000000000
Total charge density on r-space grids: 0.0000000005
Total charge density g-space grids: 0.0000000005
6 P_Mix/Diag. 0.40E+00 3.7 2931.12581846 -1401.4996955435 4.76E+02
Trace(PS): 784.0000000000
Electronic density on regular grids: -783.9999999994 0.0000000006
Core density on regular grids: 784.0000000000 -0.0000000000
Total charge density on r-space grids: 0.0000000006
Total charge density g-space grids: 0.0000000006
Thomas Kühne
Oct 16, 2019, 3:02:09 PM10/16/19
to cp...@googlegroups.com
Dear Pavlo,
try to reduce MIXING@DFT/SCF and potentially also the mixing scheme.
BTW, EPS_SCF 1.0E-8, which corresponds to the maximum norm within
\Delta P when using diagonalization, is very tight.
Best,
Thomas
--
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/03a16edc-76ef-4432-8944-285e98f5b92f%40googlegroups.com.
==============================
Thomas D. Kühne
Dynamics of Condensed Matter
Chair of Theoretical Chemistry
University of Paderborn
Warburger Str. 100
D-33098 Paderborn
Germany
Pavlo Kostetskyy
Oct 16, 2019, 9:43:49 PM10/16/19
to cp2k
Dear Thomas,
Best Regards,
Your suggestion worked and the calculation converged - thank you very much for the speedy reply! A quick follow up question regarding the ALPHA value - is there a systematic way to optimize it for quickest convergence, or is it a binary outcome of converged/not converged? Thank you again.
Best Regards,
Pavlo
To unsubscribe from this group and stop receiving emails from it, send an email to cp...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/03a16edc-76ef-4432-8944-285e98f5b92f%40googlegroups.com.
Xin Wu
Oct 17, 2019, 1:46:03 AM10/17/19
to cp2k
Let Thomas' postdoc answer this question...
There is no way to determine the *best* ALPHA value. ALPHA mixes the electron density in the current iteration
with that from the preceding SCF iteration and generates the new density for the next iteration, so that the variation of
electron density during the SCF procedure becomes smooth and therefore
it leads convergence even for some *difficult* systems.
Xin
Reply all
Reply to author
Forward
0 new messages