B3LYP calculation of single O atom does not converge.

[ynwu]

Hello, all,

I am doing some tests of B3LYP xc function in VASP. My test case is O single atom in a box (10A*10A*10A). Only Gamma point is used. Here is my INCAR file.

=====================================
ENCUT = 600.0
NELM = 40
NELMDL= 5
EDIFF = 1E-5
NSW = 0
ISMEAR = 0
SIGMA = 0.1
PREC = High
ISPIN = 2
ISYM = 1
IBRION = -1
#LWAVE = .F.
#LCHARG = .F.
LREAL= .False.
LPLANE = .TRUE.
ISIF = 2

VOSKOWN = 1
GGA=B3
LHFCALC = .TRUE.
AEXX=0.2
AGGAX=0.72
AGGAC=0.81
ALDAC=0.19
ALGO = D
TIME = 0.4
PRECFOCK = Normal

ADDGRID = .TRUE.
==============================

The energy fluctuates largely, as shown in the OSZICAR file:

================================
DAV: 1 0.315610019315E+02 0.31561E+02 -0.30072E+03 32 0.201E+02
DAV: 2 0.763401275504E+01 -0.23927E+02 -0.23257E+02 32 0.181E+01
DAV: 3 0.273989843681E+01 -0.48941E+01 -0.48656E+01 32 0.150E+01
DAV: 4 0.239947561561E+01 -0.34042E+00 -0.32481E+00 32 0.395E+00
DAV: 5 0.237233240282E+01 -0.27143E-01 -0.26949E-01 32 0.114E+00
SDA: 6 -0.119153246457E+02 -0.14288E+02 0.64865E+02 32 -0.162E+03 0.000E+00
DMP: 7 0.292985041860E+03 0.30490E+03 -0.21288E+02 32 0.531E+02 0.400E+00
DMP: 8 0.254716317768E+03 -0.38269E+02 -0.37059E+02 32 0.367E+02 0.146E+03
DMP: 9 0.153444569699E+03 -0.10127E+03 -0.10415E+03 32 0.508E+02 0.549E+03
DMP: 10 0.297744055250E+01 -0.15047E+03 -0.94096E+01 32 0.509E+02-0.717E+02
DMP: 11 0.105640726887E+02 0.75866E+01 -0.64291E+01 32 0.207E+02-0.121E+02
DMP: 12 0.372720400043E+02 0.26708E+02 0.93915E+02 32 -0.231E+03-0.933E+01
DMP: 13 0.353541112136E+03 0.31627E+03 -0.56257E+01 32 0.158E+02-0.459E+01
DMP: 14 0.348385123724E+03 -0.51560E+01 -0.44508E+01 32 0.718E+01 0.103E+02
DMP: 15 0.320700575598E+03 -0.27685E+02 -0.60360E+02 32 0.411E+02 0.287E+03
DMP: 16 0.997880725446E+02 -0.22091E+03 -0.68896E+02 32 0.198E+03-0.687E+02
============================
It ends here with an error message:

ERROR in subspace rotation PSSYEVX: I2,IFAIL= 1 40


I have tried AMIX BMIX with small number, but it doesn't work.
If I do spin-unpolarized calculation, it is converging well. I guess this is possibly related to spin.
I am wondering if anybody has this issue before? Thanks in advance.




<span class='smallblacktext'>[ Edited ]</span>

[admin]

please try to use the Davidson algorithm and start with the correct spin multiplet configuration (NUPDOWN = 2)

[ynwu]

I tried IALGO=38 and NUPDOWN =2. It works, though convergence is slow.

But it does not work with a prec=Medium. Following is the OSZICAR. I am wondering if B3LYP is that sensitive to prec setting. I have tried HSE06 and PBE0 functional, and they don't have this problem.


==========================
N E dE d eps ncg rms rms(c)
DAV: 1 0.285851701248E+02 0.28585E+02 -0.96315E+02 64 0.177E+02

DAV: 2 0.199928448312E+01 -0.26586E+02 -0.26564E+02 128 0.212E+01

DAV: 3 0.321160311665E+00 -0.16781E+01 -0.16476E+01 64 0.102E+01

DAV: 4 0.298092536617E+00 -0.23068E-01 -0.23048E-01 96 0.138E+00

DAV: 5 -0.149167095703E+02 -0.15215E+02 -0.46488E+00 64 0.553E+00 0.993E-01
DAV: 6 -0.150281531433E+02 -0.11144E+00 -0.18193E-01 64 0.125E+00 0.128E+00
DAV: 7 -0.153127537253E+02 -0.28460E+00 -0.16718E-02 80 0.440E-01 0.576E+00
DAV: 8 -0.164264644560E+02 -0.11137E+01 -0.36374E-01 64 0.166E+00 0.491E+00
DAV: 9 -0.164797183830E+02 -0.53254E-01 -0.67111E-02 80 0.831E-01 0.412E+00
DAV: 10 -0.164847792291E+02 -0.50608E-02 -0.39417E-02 96 0.574E-01 0.351E+00
DAV: 11 -0.164735870102E+02 0.11192E-01 -0.25664E-02 64 0.431E-01 0.304E+00
DAV: 12 -0.164575372974E+02 0.16050E-01 -0.16290E-02 64 0.331E-01 0.266E+00
DAV: 13 -0.164414276576E+02 0.16110E-01 -0.10116E-02 64 0.267E-01 0.234E+00
DAV: 14 -0.164270017621E+02 0.14426E-01 -0.63849E-03 64 0.220E-01 0.207E+00
DAV: 15 -0.164148163114E+02 0.12185E-01 -0.42587E-03 64 0.209E-01 0.183E+00
DAV: 16 -0.164047145947E+02 0.10102E-01 -0.34907E-03 64 0.232E-01 0.162E+00
DAV: 17 -0.164068709083E+02 -0.21563E-02 -0.27494E-03 64 0.220E-01 0.183E+00
DAV: 18 -0.166366224149E+04 -0.16473E+04 -0.16376E+04 96 0.200E+00 0.315E+01
DAV: 19 -0.210591773995E+06 -0.20893E+06 -0.19841E+06 128 0.560E+01 0.925E+01
DAV: 20 -0.126283551993E+07 -0.10522E+07 -0.11528E+07 64 0.161E+04 0.118E+02
=============================
<span class='smallblacktext'>[ Edited Wed Apr 25 2012, 05:56PM ]</span>

[alex]

Check your occupation. SIGMA seems to be too large in your case. Try sth. like 0.001 or 0.01 ...

Hth

alex