Convergence problem

Message

dan · #1 Post by **dan** » Tue Jan 15, 2008 4:13 pm

Dear All,

I have recently calculated successfully the oxalic acid molecule (COOH-COOH) in vacuum. However, I have met serious difficulties when trying to calculate the anions of this molecule in vacuum (i.e. when trying to calculate the COOH-COO and COO-COO structures in vacuum). More precisely, instead of obtaining the monoanion COOH-COO I obtain two independent molecules: COOH and COO separated by aproximative 3A. The same happens, also, with the dianion which get splitted in COO and COO groups. Both the monoanion and the dianion has been previously obtained theorethically by several groups who used GAUSSIAN 92/94/98 program (see for example the papers of C.Chen, S.F. Shyu, International Journal of Quantum Chemistry, Vol.76, 541-551, 2000; M.Remko, K.R.Liedl, B.M.Rode, J.Chem.Soc., Perkin Trans. 2, 1996). Nevertheless I can not reproduce these results. Even worse, I get the same problem when trying to calculate the anions for other molecular systems. As I am not a VASP expert my input settings might be wrong, so here is bellow how I typically run these jobs:

INCAR
-------------------------------------------------------------
SYSTEM = Oxalate (oxalic dianion COO-COO)
#! first 20 MD steps

ISTART = 0 ! no WAVECAR available
ICHARG = 2 ! superposition of at charge densit
ENCUT = 550 ! cut off energy

PREC = High ! precision high
ISPIN = 2 ! spin polarized calculation
ISIF = 2 ! calculate stress tensor
ISYM = 2 ! sym ON save memory
#! IDIPOL = 4 ! dipol moment in all directions

EDIFF = 5e-7 ! stopping crit for electr min
NELM = 300 !
NELMIN = 5 ! minimum 5 electronic steps
IALGO = 48 ! RMMDIIS for electrons

EDIFFG = -0.025711 ! stopping crit for ions 1 mRy/a.u.
NSW = 20 ! 20 ionic steps
POTIM = 0.0052918 !
IBRION = 2 ! conjugate gradient for ions

ISMEAR = 0 ! Gaussian smearing
SIGMA = 0.11 ! broadening should be ok

LCHARG = TRUE ! write CHGCAR CHG files
LWAVE = TRUE ! write WAVECAR file
LREAL = Auto ! proj op eval in real sp

LPLANE = TRUE !
#! check if NGZ>3*NPROC/NPAR; OPT NGZ=n*NPAR
#! check also max/min on nodes

NPAR = 32 !
#! nr of nodes working on a band = NPROC/NPAR
#! NBANDS=NELECT/2+NIONS/2

NSIM = 4 !
#! nr of bands optimized in the same time

-------------------------------------------------------------

POSCAR
-------------------------------------------------------------
Oxalate (oxalic dianion COO-COO) M.Remko coords
1.00000000000000000
14.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 14.0000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 14.0000000000000000
2 4
Selective dynamics
Cartesian
0.000000 -0.775000 0.000000 F F F
0.000000 0.795945 0.098026 T T T
0.000000 -1.415079 1.092271 T T T
0.000000 -1.274327 -1.163369 T T T
1.130014 1.365647 0.133576 T T T
-1.130014 1.365647 0.133576 T T T
-------------------------------------------------------------

I use only 1 kpoint (the Gamma point) and the PAW_PBE C_h 20Dec200, PAW_PBE O_h 20Dec200 pseudopotentials (for the monoanion also the PAW_PBE H_h 07Sep2000 psp).

Trying to obtain the convergence for the oxalic anions I have done the following tests:
-tryed the non-spin polarized calculation -> FAILED
-run for only 10-20 MD and then restart only from WAVECAR -> FAILED
-run for only 10-20 MD and then use only the CONTCAR for a new start -> FAILED
-make constrained relaxation (i.e. fix certain bonds) -> FAILED
-check for IBRION=1 -> FAILED
-tryed different types of mixing: Broyden, Kerker, Pulay, Tchebycheff -> FAILED
-mix slowly (AMIX, BMIX) -> FAILED
-use the linear mixing (BMIX = 0.0001) -> FAILED
-vary MAXMIX and NFREE -> FAILED
-vary EDIFF -> FAILED
-follow the advices from the manual related to efficient relaxation starting from reasonable/unreasonable configurations -> FAILED
-various combination of the above ones -> FAILED

Now, I do not know what I am doing wrong. If anyone can give me a hint, it would be really great.

With all the best wishes,
dan

#2 Post by **admin** » Mon Jan 28, 2008 10:01 am

please do use the dipole corrections.. However, you will have to upgrade vasp to the latest release (4.6.34), because in all former versions there had been a bug in calculating the forces for ionic species when dipole corrections were used. Please ask the colleague who is responsible for the maintainance of VASP at your site to retrieve vasp.4.6.34 for you from our ftp server.

#3 Post by **admin** » Mon Jan 28, 2008 10:03 am

please do use the dipole corrections.. However, you will have to upgrade vasp to the latest release (4.6.34), because in all former versions there had been a bug in calcutating the forces for ionic species when dipole corrections were used. Please ask the colleague who is responsible for the maintainance of VASP at your site to retrieve vasp.4.6.34 for you from our ftp server.