how to obtain accurate result for non-magenetic free atom energy

[TING QIN]

I have one problem to calculate the free atom energy without magnetism for tungsten. I just find the
electronic relaxation can reach 60 steps(which is the default value for NELM) and stop. If I set NELM larger, it
can run NELM steps. However, does this mean that the relaxation process converges or how can we get convergence.

I also decreased SIGMA and it still ran NELM steps.

Thanks a lot. The files are listed below:
INCAR:
System = W, single atom
ENCUT = 330
ISMEAR = 0
SIGMA = 0.1
EDIFF=0.0001
ISPIN=1
NELM=120
POSCAR:
W in single atom
25
1 0 0
0 1 0
0 0 1
1
direct
0.0 0.0 0.0
KPOINTS:
automatic_generate
0
Monkhorst
1 1 1
0 0 0

Part of the OSIZCAR:
RMM: 117 -0.187057037789E+01 0.18275E+00 -0.42782E-02 14 0.209E+00 0.576E+00
RMM: 118 -0.205330092095E+01 -0.18273E+00 -0.26786E-02 14 0.158E+00 0.521E+00
RMM: 119 -0.187051176102E+01 0.18279E+00 -0.42779E-02 14 0.209E+00 0.576E+00
RMM: 120 -0.205331395580E+01 -0.18280E+00 -0.26784E-02 14 0.158E+00
1 F= -.20533140E+01 E0= -.20318639E+01 d E =-.643501E-01

[Vlad]

Try to set SIGMA to zero. I find this often helps to improve convergence for single atoms.

[admin]

1) please mind that for free atoms Hund's rule applies!
2) SIGMA gas to be chosen much smaller for free atoms (discrete levels),
set it to 0.001 or even less