Forces problem with single point energy calculation

[gaetan]

Hi Dear all,

I am new user of VASP and I have a problem when I do some single point energy calculations.

I have done structure optimization on organic crystals and I set EDIFFG to -0.001. And it works. But after doing a SPE calculation, when I look at the OUTCAR file, some atoms have their forces upper 0.001 eV/Ã….

I do not understand how forces can increase so far after a SPE calculation.

Thanks

[sankh]

Hi,
This is surprising. Have you

1. Done the SPE on relaxed geomtrey moving the CONTCAR of relaxation to POSCAR of SPE?
2. Done SPE in SCF (ICHARG = 0 or 2 or 1) or non-SCF (ICHARG = 11 or 12) fashion?
3. Changed the BZ integration scheme (for ex from Gaussian to Tetrahedron+Blöchl or MP) without IBRION = -1, although of course NSW=0 without IBRION=-1 setting would do fine?
4. Changed the K-mesh density?
5. Changed energy cutoff?

Perhaps you could quot here your INCAR fiels for relaxation and SPE.

Cheers,
Sankh

[gaetan]

Hi,

This is the INCAR file from the relaxation:

NWRITE = 1
EMIN = -15
EMAX = 15
IALGO = 38
NELM = 30
EDIFF = 0.0001
EDIFFG = -0.001
IBRION = 1
ISMEAR = 4
SIGMA = 0.2
ENCUT = 520
ISIF = 3
ISYM = 0
NSW = 800
NELMIN = 4
LREAL = .FALSE.
PREC = High

And this is the INCAR file from SPE:

NWRITE = 1
EMIN = -15
EMAX = 15
IALGO = 38
NELM = 30
EDIFF = 0.00001
IBRION = -1
ISMEAR = 4
SIGMA = 0.2
ENCUT = 520
NELMIN = 4
LREAL = .FALSE.
PREC = High

I have used the CONTCAR file from relaxed calculation and turned it into POSCAR for SPE calculation.
And both calculations have the same K-mesh.

Thanks,
Gaëtan

<span class='smallblacktext'>[ Edited Wed Mar 09 2011, 09:40AM ]</span>

[a.mavr]

I noticed that you changed EDIFF from 1E-4 to 1E-5. Did you already check that 30 steps (NELM=30) are enough to achieve convergence for the SCF ?
Did you compare the energies computed from your optimization and your SPE ? Do they differ?
In your optimization, the symmetry is turned off (ISYM=0), whereas in the SPE not. Does VASP detect any symmetry, when you do the SPE? Check the outcar file.


Andreas

[gaetan]

I always check if my calculation converge or not and in this case, the converge is ok.

I get -435.173052 eV for the optimization and -435.172817 eV for the SPE.

And yes, VASP detect the symmetry I expected.

Thanks,

Gaëtan

[a.mavr]

If VASP detects a symmetry in the SPE, then, the two calculations are performed under different symmetry restrictions. (Which might lead to different electronic states). Try taking a look at your occupied orbitals and check the occupation numbers.

Try setting ISYM=0 also for the SPE and then compare the energy with the optimization job.
I would also try to perform a symmetry restricted (ISYM=1 or 2) optimization and compare the energy with the ISYM=0 job.

Hope that helps.

Andreas

[gaetan]

Hi,

So, I have tried two thnings:

- Perform a SPE with ISYM = 0 tag (E= -435.172817 eV, same as ISYM = 2 SPE)


-Perform a symmetry restricted optimization (E= -435.172821 eV and then a symmetry restricted SPE (E= -435.172752 eV).

Both attempt lead to an increase in forces for the SPE.

[admin]

most probably, the following has occured:
I see that you chose ISIF=3 for the first run. This changes the shape and the volume of the cell during the run, without updating the PW-meshes.
If volume and cell shape are changed considerably, the PW-mesh used throughout the run becomes worse in quality as compared to the input geometry.
For the follow-up SPE calculation, the PW-meshes are re-calculated, in order to give the desired accuracy for ENCUT and the new cell dimensions. ---> with the change of the basis set, the forces will change as well.
please compare the NGX, NGY and NGZ parameters of runs 1 and 2 to see if this may be the reason of the inconsistency of the forces