Calculation converges with Al potential but not Al_h?

[bmeredig]

Hi all,

I'm having a strange issue with the Al_h potential. It seems to work fine for me for an Al primitive cell, but when I make a 21-atom supercell for a phonon calculation (with an atomic displacement), the static calculation never achieves electronic convergence(or takes an exceedingly long time). The exact same cell/INCAR converges readily (within ~10 iterations) with the Al potential. Here's what the problem looks like:

N E dE d eps ncg rms rms(c)
DAV: 1 0.374051838104E+02 0.37405E+02 -0.20948E+04 41600 0.653E+02
DAV: 2 -0.720920574329E+02 -0.10950E+03 -0.10211E+03 51424 0.978E+01
DAV: 3 -0.828871407507E+02 -0.10795E+02 -0.49359E+01 64424 0.157E+01
DAV: 4 -0.884014716849E+02 -0.55143E+01 -0.41535E+01 77168 0.879E+00
DAV: 5 -0.891304911787E+02 -0.72902E+00 -0.53189E+00 77624 0.350E+00 0.402E+01
RMM: 6 -0.903540610229E+02 -0.12236E+01 -0.11274E+01 45188 0.749E+00 0.157E+01
RMM: 7 -0.851326636699E+02 0.52214E+01 -0.20752E+00 69055 0.210E+00 0.114E+01
RMM: 8 -0.851361744099E+02 -0.35107E-02 -0.15393E+00 80436 0.123E+00 0.110E+01
RMM: 9 -0.849155508710E+02 0.22062E+00 -0.11966E+00 73328 0.845E-01 0.220E+01
RMM: 10 -0.844545724129E+02 0.46098E+00 -0.83624E-01 72062 0.103E+00 0.257E+01
RMM: 11 -0.842020493176E+02 0.25252E+00 -0.72195E-01 71092 0.129E+00 0.420E+01
RMM: 12 -0.835434131866E+02 0.65864E+00 -0.36844E-01 70940 0.133E+00 0.227E+01
RMM: 13 -0.836435196747E+02 -0.10011E+00 -0.33743E-01 71291 0.116E+00 0.238E+01
RMM: 14 -0.835223335723E+02 0.12119E+00 -0.27707E-01 75836 0.475E-01 0.200E+01

...continues...

RMM: 98 -0.832985417586E+02 -0.47287E-04 -0.11253E-03 55738 0.105E-02 0.113E+00
RMM: 99 -0.832985637559E+02 -0.21997E-04 -0.10820E-03 54172 0.898E-03 0.112E+00
RMM: 100 -0.832985761994E+02 -0.12444E-04 -0.10257E-03 52408 0.767E-03 0.108E+00
RMM: 101 -0.832983182635E+02 0.25794E-03 -0.79886E-04 58954 0.112E-02 0.840E-01
RMM: 102 -0.832986411315E+02 -0.32287E-03 -0.88647E-04 58619 0.156E-02 0.824E-01
RMM: 103 -0.832986223809E+02 0.18751E-04 -0.76587E-04 56663 0.911E-03
1 F= -.83298622E+02 E0= -.83356513E+02 d E =0.173671E+00

Here's my INCAR:

NSIM = 4
NPAR = 1
ISTART = 0
ICHARG = 1
ISMEAR = 1; SIGMA = 0.2
ISIF = 3
PREC = ACCURATE
!LREAL = AUTO
ALGO = FAST
IBRION = -1
NSW = 0
EDIFF = 1E-4
EDIFFG = -5E-3
NELMIN = 6
NELM = 120
ADDGRID = .TRUE.
ENCUT = 550
ENAUG = 1100
VOSKOWN = 1
LORBIT = 11
NBANDS = 80

I've tried all the obvious stuff like changing KPOINTS, M-P vs. gamma, smearing, width of smearing, sticking with DAV, adding bands (the default # was too small), etc. I haven't played with the mixer but I've never had to do that in VASP before, and I've calculated quite a few different systems. Any help would be much appreciated!

Bryce

[admin]

the charge density does not converge properly, please reduce the mixing parameters and increase NELMDL to get better pre-convergence.

[job]

The Al_h potential suffers from ghost state problems and should not be used.

[bmeredig]

Thanks to admin and job for the input. I might play around with the mixing, but it's good to know the potential may have issues.