Problem calculating Nitrogen atom

[matlgen]

Hello. I needed to calculate the energy of a Nitrogen atom, but I ran into some serious difficulties. Using NSPIN=2 and NUPDOWN=-1, it successfully calculated the doublet. However, the ground state is a quartet.

When I ran the calculation with NUPDOWN=3 I had problems. Starting at SCF step 12, the number of electrons would switch from 5.0 to 4.4999990. Once the SCF cycles converged, it would show the occupancy as a total of 4.5.

I played with the different mixing parameters, but it always came out the same.

I was able to get around the problem by putting smearing on. In fact, using smearing as small as 0.00001 let it calculate perfectly.

Does anyone have an explanation for this?

David


P.S. I've put the relevant info from my input files below in case anyone spots anything that I did that may cause the problem, or in case anyone wants to test it out. I am using VASP 4.6.3 with the climbing nudged elastic band method incorporated (but not used in this calculation). I'm happy to supply any additional info that people need to determine this bug, if it is indeed a bug.

=====INCAR====
PREC = NORMAL
GGA = 91
VOSKOWN = 1

ISPIN = 2
NUPDOWN = 3

ISIF = 2
LREAL = T real-space projection
IALGO = 48 algorithm
EDIFF = 1E-04 stopping-criterion for ELM
ISYM = 0
NELM = 400;

EDIFFG = -.5E-01 stopping-criterion for IOM
NSW = 200 number of steps for IOM
IBRION = 1 ionic relax: 0-MD 1-quasi-New 2-CG
POTIM = 0.5

ISMEAR = 2
SIGMA = 0.0 ## 0.00001 will let the calc run correctly

NSIM = 1
NPAR = 1
LPLANE = .TRUE.
LSCALU = .FALSE.

LCORR = T Harris-correction to forces
LDIAG = T sub-space diagonalisation

====KPOINTS====
K-Point Grid
0
Monkhorst Pack
1 1 1
0 0 0

====POTCAR (synopsis from the outcar)====
POTCAR: US N
VRHFIN =N: s2p3
LEXCH = 91
EATOM = 262.6548 eV, 19.3046 Ry
TITEL = US N
LULTRA = T use ultrasoft PP ?
IUNSCR = 0 unscreen: 0-lin 1-nonlin 2-no
RPACOR = .000 partial core radius
POMASS = 14.001; ZVAL = 5.000 mass and valenz
RCORE = 1.650 outmost cutoff radius
RWIGS = 1.480; RWIGS = .783 wigner-seitz radius (au A)
ENMAX = 348.097; ENMIN = 261.072 eV
EAUG = 700.000

ICORE = 2 local potential
LCOR = T correct aug charges
RMAX = 2.471 core radius for proj-oper
QCUT = -5.058; QGAM = 10.116 optimization parameters


====POSCAR====
N
1.0
15.5 0.0 0.0
0.0 15.5 0.0
0.0 0.0 15.5
1
direct
0.5 0.5 0.45

[admin]

in general, the PPs were generated for the BULK electronic configuration,
therefore the calculation of free atoms may require some additional input
to enforce the atom to obey Hund's rule:
1) either set NUPDOWN (3 for N) or
FERWE and FERDO and occupy the states according to Hund's rule
2) set ISMEAR = -2 to keep the occupancies fixed
3) free atoms have discrete energy levels therefore the smearing has to be
set to very small values.
4) set LDIAG = .False. to keep the ordering of the eigenstates fixed

for further details I recommend to have a look at the handsonI.pdf
VASP-workshop tutorial, which can be downloaded from our site,
http://cms.mpi.univie.ac.at/vasp-workshop
(goto session talks and tutorials)