VASP Forum

Dear experts,

I am having a problem relaxing the slab with a single O atom as the adsorbate. I am using the VASP 6.5.1 version for the calculation, and I see the following error
| Error EDDDAV: Call to ZHEGV failed. Returncode = 10 2 16 |
| |
| ----> I REFUSE TO CONTINUE WITH THIS SICK JOB ... BYE!!! <---- |

Kindly go through the attached input files and provide the solution.

Hi,

If you still have the OUTCAR and OSZICAR files could you please upload them?

Dear expert,

Please find the attached OUTCAR and OSZICAR.

It seems that you did not provide the correct files: the INCAR is not readable since binary and the OSZICAR corresponds to a calculation that finished properly.

Please find the corrected input files attached. I have also attached OUTCAR and OSZICAR.
I sincerely apologize for having sent a corrupted INCAR file earlier — I didn’t notice the issue at that time.

Hi,

I noticed that ISPIN=2 in INCAR, which means a spin-polarized calculation. For such calculations one should usually specify reasonable initial magnetic moments (value and sign) on the atoms (with MAGMOM), otherwise the initial magnetic moments are by default set to 1 for all atoms, which is probably not physically meaningful for your system. Such non-physical solutions may lead to technical problems and possibly to a crash of the calculation. Thus, I suggest that you restart your calculation with MAGMOM set to some meaningful choice (and do not forget to delete the old WAVECAR file before restarting the calculation).

Dear Expert,

I have performed the calculation, including the MAGMOM tag. This time, I encountered the error "Error EDDDAV: Call to ZHEGV failed. Returncode = 18 2 20".
Here, I have attached the input and output files for your reference. Please have a look.

Hi,

The calculation crashed at the beginning of the second ionic step. In the OUTCAR, we can see that the forces at the first ionic step are large on atoms 20 and 53 (search for TOTAL-FORCE in OUTCAR) . These two atoms form a bond and it may possibly be that the distance between these two atoms is unrealistic. Do you still have the CONTCAR file that was generated at the end of the first ionic step (and used for the second ionic step)?

Please find the attached CONTCAR generated after the first iteration.

I had a closer look at the geometry, and it seems that the crash is a consequence of the bond length between the O atom (atom no. 53) and one of the Bi atoms (no. 20) that is unrealistically too small (1.4 Å). Attached, is a modified POSCAR file, where the O atom is at a higher \(z\)-position such that the Bi-O bond length is about 2.5 Å. With this POSCAR no crash occurs at the second ionic step.