Convergence issue with smear tag


Click here for full thread
Clicked A Few Times
echo
title "cluster"
geometry autosym
.
.
.
.
.
.
end

basis "ao basis" cartesian
Pd library lanl2dz_ecp file /usr/local/NWChem/data/libraries/
end

ecp
Pd library lanl2dz_ecp file /usr/local/NWChem/data/libraries/
end

dft
vectors input cluster-3.movecs output cluster-smear.movecs
iterations 3000
direct
mult 1
XC xpbe96 cpbe96
convergence ncyds 3000
grid fine
tolerances accqrad 20
smear 0.001

end

DRIVER
XYZ a.xyz
MAXITER 5000
END


task dft energy

I already have an optimized geometry which did not include any smearing. However when I use that optimized geometry in the above input file, the the scf cycles don't converge at all. Part of the output is shown below whose deltaE is always b/w D-01 and D+01 (even for ~ 100 cycles).

d= 0,ls=0.0,diis 1 -2563.1088837131 -1.07D+04 1.71D-02 1.02D+01 207.6
Grid integrated density:     367.999970645462
Requested integration accuracy: 0.10E-05
sigma= .10D-02 eFermi= -0.07880 T*S= -4.45E-05 nel(1)= 368.00 nel(
d= 0,ls=0.0,diis 2 -2562.7372814973 3.72D-01 5.69D-02 5.25D+00 389.4
Grid integrated density: 367.999952201265
Requested integration accuracy: 0.10E-05
sigma= .10D-02 eFermi= -0.14160 T*S= -1.62E-04 nel(1)= 368.00 nel(
d= 0,ls=0.0,diis 3 -2483.8130328019 7.89D+01 4.61D-02 3.72D+02 576.9
Grid integrated density: 367.999968493405
Requested integration accuracy: 0.10E-05
sigma= .10D-02 eFermi= -0.14870 T*S= -2.08E-04 nel(1)= 368.00 nel(
d= 0,ls=0.0,diis 4 -2550.5940778012 -6.68D+01 3.07D-02 8.44D+01 762.3
Grid integrated density: 367.999971195797
Requested integration accuracy: 0.10E-05
sigma= .10D-02 eFermi= -0.14698 T*S= -1.90E-04 nel(1)= 368.00 nel(
d= 0,ls=0.0,diis 5 -2545.8670568312 4.73D+00 2.81D-02 1.09D+02 947.5

Thanks for any guidance.