Hi Salvador,
first of all pseudo for gold should be relativistic - it's a heavy atom
and relativistic effects change significantly the eigenenergies.
2. You should use 5d10, 6s1 configuration.
3. What basis set do you use? An optimized basis can give much better
results. There was a good basis sets in the mailing list (as well as
pseudopotential).
4. 4x4x4 k-grid is definitely not enough for metal. Check the
convergence of total energy s k-grid. You'd better use kgrid_cutoff
(around 15 Ang is needed) rather than %block kgrid_Monkhorst_Pack and use
5. Check also the convergence of total energy vs. MeshCutoff (I think
that for NLCC potential 300Ry should be used at least).
6. Are you sure that you get convergence of your SCF cycles with
> MaxSCFIterations 100
> DM.MixingWeight 0.2
also ElectronicTemperature 300K is too high for Fermi-Dirac occupation
function.
Try this:
DM.MixingWeight 0.01 # New DM amount for next SCF cycle
DM.NumberPulay 3
DM.PulayOnFile false
DM.NumberKick 20
DM.KickMixingWeight 0.05
DM.Tolerance 0.00001
OccupationFunction MP
ElectronicTemperature 300 K
7. Some additional splitting may occur due to spin-orbit interaction
which you don't have in SIESTA.
P.S. Here are the basis and the pseudo from the mailing list by
Junquera. But I believe you have them, since those are the only that I
saw in the list.
pg Gold with 5d in the valence.
tm2 2.60
n=Au c=car
0.0 0.0 0.0 0.0 0.0 0.0
12 4
6 0 1.00 0.00
6 1 0.00 0.00
5 2 10.00 0.00
5 3 0.00 0.00
2.32 2.32 2.32 2.32
#23456789012345678901234567890123456789012345678901234567890
%block PAO.Basis
Au 3 0.02
n=6 0 2 E 10.00 6.00
6.500 5.500
1.000 1.000
n=6 1 1 E 10.42 3.69
5.850
1.000
n=5 2 2 E 102.50 5.62
7.200 5.570
1.000 1.000
%endblock PAO.Basis
