VASP cannot solve single atom: Zn, Be, Ti

Message

rpf · #1 Post by **rpf** » Thu Apr 12, 2012 12:43 pm

I perform very simple calculations to obtain the ground state of single atoms. However I see that VASP gives weird results for Zn Be and Ti atoms.

So far I've tried a lot of different methods (that were suggested by VASP official website) but none of them works. I also see that some other people asked similar questions and there has been no response for their questions so far.

Dear VASP administrators, can you explain the recipe for Zn Ti and Be isolated atoms ?

#2 Post by **admin** » Thu Apr 12, 2012 4:04 pm

VASP is a numerical method. Therefore some numerical problems are to be expected caused mainly by symmetry. E.g. single atom sitting at the position 0 0 0 in a cubic cell can be the problematic case. Try the following:
1) Usea non cubic cell.
2) Move the atom slightly away from its position.

rpf · #3 Post by **rpf** » Fri Apr 13, 2012 2:33 pm

Dear Admin thank you for your rapid response but the solution is not that easy.

I have tried:

1) Cubic Unitcell
2) Hexagonal Unitcell
3) Rectangular Prism Unitcell
4) Shifted positions of single atom

but none of them works. I am sorry

I think we need more than that. Could you help?

#4 Post by **admin** » Mon Apr 16, 2012 11:17 am

I tried the most difficult case. Inserted single atom (Zn) in a cubic cell (10x10x10) to the most critical site (0 0 0) and it works (PAW_GGA Zn)

Danny · #5 Post by **Danny** » Mon Apr 16, 2012 1:15 pm

Are you using PBE? spin-polarized calculations?
Do you use smearing? anions?

rpf · #6 Post by **rpf** » Mon Apr 16, 2012 10:09 pm

[quote="blue"]Could you put your INCAR here ?[/color][/b]

#7 Post by **admin** » Tue Apr 17, 2012 11:07 am

There is nothing special about INCAR. The total energy of the Zn atom calculated with the GGA functional is close to zero. Because GGA potential is asymmetric a change of symmetry of the atomic position in the cell can cause such a change of the total energy that it can become positive.
E.g. Zn in the 10x10x10 cell gives:
position 0 0 0: energy +0.00013 eV
position 0.1 0.1 0.1: energy -0.000084 eV.
LDA and PBE functionals are not that sensitive and provide safe solutions. No problems are observed with Be and Ti atoms.

[ Edited Tue Apr 17 2012, 11:12AM ]

rpf · #8 Post by **rpf** » Thu Apr 19, 2012 4:31 pm

[quote="blue"]You say that it is solved but there is no solution here. Could you share your simple INCAR file with us pleas[/color]e ?

I am using LDA. Cell is 10x10x10. Here are my two INCAR files.

=============================

SYSTEM = Geometry-Relaxation

Startparameter for this Run:
NWRITE = 2;
LPETIM = F
ISTART = 0
LPLANE = T
NPAR = 2
NSIM = 1
LORBIT = 11

Electronic Relaxation 1
ENCUT = 500
NELM = 150
EDIFF = 1E-05

NSW = 0
IBRION = -1
NBLOCK = 1
KBLOCK = 1
ISIF = 2
ISYM = 0
IWAVPR = 1
ISPIN = 2
NUPDOWN = 0

ISMEAR = 0
SIGMA = 0.01

ALGO = Normal algorithm
LDIAG = T sub-space diagonalisation
LREAL = F real-space projection
LELF = T

================================

and the other INCAR file

==========================

ISYM = 0 ! no symmetry
ISPIN = 2 ! allow for spin polarisation
VOSKOWN = 1 ! this is important, in particular for GGA
ISMEAR = 0 ! Gaussian smearing, otherwise negative occupancies
SIGMA = 0.01 ! intermid. smearing width
AMIX = 0.2 ! mixing set manually
BMIX = 0.0001
NELM = 30 ! 30 electronic steps
ICHARG = 1

===============================

You say that it is solved but there is no solution here. Could you share your simple INCAR file with us please ?

Best Regards

[ Edited Thu Apr 19 2012, 04:57PM ]

#9 Post by **admin** » Fri Apr 20, 2012 10:52 am

SYSTEM = Zn atom
ISTART = 0 ! start from 0: scratch, 1: previous WAVECAR
ENCUT = 500 ! ENCUT for O
EDIFF = 1.0E-7
PREC = High
ISYM = 0 ! no symmetry
ISPIN = 2 ! allow for spinpolarisation
VOSKOWN = 1 ! this is important, in particular for GGA
ISMEAR = 0 ! Gaussian smearing, otherwise negative occupancies
SIGMA = 0.10 ! intermid. smearing width
AMIX = 0.2 ! set mixing manually
BMIX = 0.0001
NELM = 40 ! 40 electronic steps
IBRION = -1 ! -1: Fix atoms; 0: MD; 2: ConjGrad relax
NSW = 0 ! Number electronic steps
NPAR = 1

rpf · #10 Post by **rpf** » Sat Apr 21, 2012 10:58 am

...

[ Edited Tue Apr 24 2012, 03:34PM ]

rpf · #11 Post by **rpf** » Tue Apr 24, 2012 3:35 pm

Dear admin, thank you for INCAR file.

You can see the total energies obtained by using following INCAR as follows. We have still serious problems.

Be/OSZICAR-gga E0= -.42098286E-01 d E =-.188426E-10 mag= 0.0000

Be/OSZICAR-lda E0= 0.66660070E-02 d E =-.139457E-10 mag= 0.0000

Be/OSZICAR-pbe E0= -.27492673E-01 d E =-.236046E-10 mag= 0.0000
Magnetic moment of 0.000 it is not the ground state. LDA is totally wrong. There is no solution for Be.

Ti/OSZICAR-gga E0= -.21992887E+01 d E =-.162687E+00 mag= 3.7405

Ti/OSZICAR-lda E0= -.20341525E+01 d E =-.165710E+00 mag= 3.9891

Ti/OSZICAR-pbe E0= -.21364338E+01 d E =-.135796E+00 mag= 3.9860
Mag moment should be 4.0000. Anyway, we can say potentials are OK for Ti.

Zn/OSZICAR-gga E0= -.64871379E-03 d E =-.239919E-11 mag= 0.0000

Zn/OSZICAR-lda E0= 0.96192741E-02 d E =-.750283E-11 mag= 0.0000

Zn/OSZICAR-pbe E0= -.16610653E+00 d E =-.212742E-10 mag= 0.0000

We have still problems with Zn. There is no solution for Zn with LDA.

We need a better recipe for isolated atom calculations

Best Regards

Danny · #12 Post by **Danny** » Thu Apr 26, 2012 10:39 am

rpf, have you tried starting with the "required magnetic moment set? SOmetimes the 0 magnetic moment is a local minimum preventing VASP from escaping to the global minimum. If that is the case you should also see lower energies.

*Note: did the calculations converge in 40 steps? if not run a second (and third ) time continuing from previous CHGCAR untill you reach convergence. Normally the mag should be much better than what you present
[ Edited Thu Apr 26 2012, 10:42AM ]

rpf · #13 Post by **rpf** » Thu Apr 26, 2012 4:52 pm

yes I've tried various moments by using NUPDOWN tag. Unfortunately the INCAR put here by admin doesn't work.
Results ar as follows. I am still waiting for a solution.

Be/OSZICAR-gga E0= -.42098286E-01 d E =-.188426E-10 mag= 0.0000

Be/OSZICAR-lda E0= 0.66660070E-02 d E =-.139457E-10 mag= 0.0000

Be/OSZICAR-pbe E0= -.27492673E-01 d E =-.236046E-10 mag= 0.0000
Magnetic moment of 0.000 it is not the ground state. LDA is totally wrong. There is no solution for Be.

Ti/OSZICAR-gga E0= -.21992887E+01 d E =-.162687E+00 mag= 3.7405

Ti/OSZICAR-lda E0= -.20341525E+01 d E =-.165710E+00 mag= 3.9891

Ti/OSZICAR-pbe E0= -.21364338E+01 d E =-.135796E+00 mag= 3.9860
Mag moment should be 4.0000. Anyway, we can say potentials are OK for Ti.

Zn/OSZICAR-gga E0= -.64871379E-03 d E =-.239919E-11 mag= 0.0000

Zn/OSZICAR-lda E0= 0.96192741E-02 d E =-.750283E-11 mag= 0.0000

Zn/OSZICAR-pbe E0= -.16610653E+00 d E =-.212742E-10 mag= 0.0000

We have still problems with Zn. There is no solution for Zn with LDA.

We need a better recipe for isolated atom calculations

Danny · #14 Post by **Danny** » Fri Apr 27, 2012 9:07 am

your Ti results show me you have not used the NUPDOWN tag, since then the mag would be exactly 4.

At this point I am also starting to wonder if it might not be a problem with your installation....

btw note that the "energy" is not neccessarily negative!! The negative only indicates if it is more stable that the ground state configuration used to generate the potential.

For Zn I find:
0.010323 for E0 in LDA
-0.164848 for E0 in PBE
[ Edited Fri Apr 27 2012, 09:18AM ]

#15 Post by **admin** » Fri Apr 27, 2012 12:25 pm

which VASP version do you use, and what is your k-point sampling?
VOSKOWN is NOT needed for LDA Potentials
check the last columns of OSZICAR whether the charge and the spin densities are converged (values should drop to about 1e-3). If they are not, please decrease the mixing parameters for the spin density mixing.