VERY BAD NEWS! internal error in subroutine SGRCON: Found some non-integer element in rotation matrix 3

[jortegar]

Hi everyone,

As stated in the subject, I am getting the following messages:

VERY BAD NEWS! internal error in subroutine IBZKPT:
Reciprocal lattice and k-lattice belong to different class of lattices. Often results are still useful... 32

VERY BAD NEWS! internal error in subroutine SGRCON:
Found some non-integer element in rotation matrix 3

when I try to do a geometric optimization of a Cu2O surface with an OH on top of one of the copper atoms.
Based on previous posts with similar errors (https://vrruwv.vasp.at/forum/viewtopic.php?t=789, https://www.vasp.at/forum/viewtopic.php?t=6582), I have systematically tried:

1. Set the parameter ISYM = 0

2. Set SYMPREC = 1e-6

3. Rattle the atoms before the calculation using ASE method atoms.rattle()

However, I keep getting the same error messages. I have attached the INCAR, KPOINTS, POSCAR, POTCAR, OUTCAR, the output file (output.log), and the Python code I use to generate the input files and run the calculation through ASE (vrelax.py) inside a zip file called 'VASP_files'.
Your help is deeply appreciated.

Best regards,
Jessica.

[michael_wolloch]

Dear Jessica,

Your reciprocal lattice is hexagonal, and your k-lattice is orthorhombic. You get a warning regarding this in your output, but it is not very detailed due to the old version of VASP that you are using. This is a result of the ASE script you are using to create the calculation. You are enforcing different k-point densities for the x and y directions of your reciprocal lattice. Since these lattice vectors are of equal size this results in the mismatch between the k-lattice and the reciprocal lattice of your cell.

I changed the KPOINTS file to be Gamma centered (recommended for hexagonal cells!) and 2x2x1:

Code: Select all

 modified KPOINTS
 0 
 Gamma
  2 2 1 
 0 0 0

This removes the problem, and I could converge the first ionic step of your calculation without any issues using the current release. I did simplify your INCAR however and lowered some settings to speed up the calculation.
Here is my INCAR for completeness:

Code: Select all

 ENCUT = 350.000000
 SIGMA = 0.10000
 EDIFF = 1.00e-04
 ISMEAR = 0
 ISPIN = 2
 KPAR = 2
 LORBIT = 11
 NELM = 250
 NCORE = 5
 MAGMOM = 72*0.0000 2*0.5000
 LREAL = Auto

Note that the current release version, VASP 6.4.3, also works with the original mesh! It prints a warning and I would not recommend running the calculation that way, but it would work.

Of course, you could also use only the Gamma point (1x1x1 mesh) and the gamma-only version of the code if that gives you enough accuracy for your quantity of interest.

Please use a KPOINTS mesh that fits your simulation cell and report back if your error disappears for version 5.4.4, so I can lock the topic.

Note that you get a couple of other warnings that you should address: NCORE and LREAL. When using a MetaGGA like SCAN, it is also a good idea to start from a converged WAVECAR obtained with PBE. If you still have convergence troubles, it might be a good idea to also deactivate the dipole tags for that initial run.

Cheers, Michael

[jortegar]

Dear Michael,

Thanks for the suggestions. The calculation ran successfully with the modifications in the k-points.

[michael_wolloch]

Thanks for reporting back. Is there any more you want to ask, or can I lock the topic?
Cheers, Michael