High cutoff for Force Converage

[Hedley Dong]

Hi everyone,

I am test energy and force converage with different cutoff, and the energy converged at a reasonable cutoff (~500 Ry). However I found the Force did not converge until I increased the cutoff to 800 Ry. I am not sure such a high cutoff is reasonable, or can I imporve the force converage with other parameters? I upload the input and output, please help!

Thank you!

[Hedley Dong]

&GLOBAL

PROJECT DPGEN

&END GLOBAL

&FORCE_EVAL

METHOD QS

STRESS_TENSOR ANALYTICAL

&DFT

BASIS_SET_FILE_NAME BASIS_MOLOPT

POTENTIAL_FILE_NAME GTH_POTENTIALS

CHARGE 0

UKS T

MULTIPLICITY 1

&MGRID

CUTOFF 350

REL_CUTOFF 50

NGRIDS 4

&END MGRID

&QS

EPS_DEFAULT 1.0E-12

&END QS

&SCF

SCF_GUESS ATOMIC

EPS_SCF 5.0E-6

MAX_SCF 300

&OT

PRECONDITIONER FULL_SINGLE_INVERSE

MINIMIZER DIIS

&END OT

&OUTER_SCF ON

MAX_SCF 4

&END OUTER_SCF

&END SCF

&XC

&XC_FUNCTIONAL PBE

&END XC_FUNCTIONAL

&vdw_POTENTIAL

DISPERSION_FUNCTIONAL PAIR_POTENTIAL

&PAIR_POTENTIAL

TYPE DFTD3

PARAMETER_FILE_NAME dftd3.dat

REFERENCE_FUNCTIONAL PBE

R_CUTOFF 15.

&END PAIR_POTENTIAL

&END vdw_POTENTIAL

&END XC

&END DFT

&SUBSYS

&CELL

A 7.552 0.    0.   

B 0.    7.552 0.   

C  0.     0.    18.972

&END CELL

&COORD

@include coord.xyz

&END COORD

&KIND O

BASIS_SET DZVP-MOLOPT-SR-GTH

POTENTIAL GTH-PBE-q6

&END

&KIND Ti

POTENTIAL GTH-PBE-q12

BASIS_SET DZVP-MOLOPT-SR-GTH

&DFT_PLUS_U

EPS_U_RAMPING 1.0E-3

U_RAMPING [eV] 1

L 2

U_MINUS_J [eV] 5

&END DFT_PLUS_U

&END

&END SUBSYS

&PRINT

&FORCES ON

&END FORCES

&STRESS_TENSOR ON

&END STRESS_TENSOR

&END PRINT

&END FORCE_EVAL

[Hedley Dong]

[Lucas Lodeiro]

Hi Hedley,

Typically forces are more difficult to converge. Forces are more sensitive to convergence parameters, and are affected by "noise" in the derivation. Have you tried to decrease the EPS_SCF? Maybe 5E-6 is not sufficient to ensure a free noise force... 1E-7 or 1E-8 could be useful to smooth the force profile respect to cutoff. And the XC_GRID subsection could be useful.

Regards - Lucas Lodeiro

--
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/8485d638-b1d6-4586-b761-fe0193384ac4n%40googlegroups.com.

[Hedley Dong]

Thanks for your advice, Lucas. I have tried decrease EPS_SCF to 3.0E-7, and the force did not change compared to EPS_SCF=5.0E-6 at the same cutoff. I will try a lower EPS_SCF and take a look of XC_GRID.