Page 1 of 1
convergence problem
Posted: Tue Jul 22, 2025 5:02 pm
by serenity_3
Hello,
I am running calculations involving CO +O on a metal surface. Most of my calculations do not converge and I receive various error messages such as this one:
ERROR FEXCP: supplied Exchange-correletion table |
| is too small, maximal index : 4175
or that one:
Error EDDDAV: Call to ZHEGV failed. Returncode = 15 1 16 .
Any suggestions what could be wrong? I am attaching one of my jobs to this post.
Thank you in advance for your help!
Re: convergence problem
Posted: Thu Jul 24, 2025 1:24 pm
by christopher_sheldon1
Dear Serenity,
Thank you for your question. I have repeated your calculation and replicated the error. In my calculation, I find that the energy (EDIFF) is not converged within the 60 steps (defined in NELM) after the first step. On further investigation, increasing NELM and using a better algorithm, e.g., ALGO = All, does not improve things. Looking at your OUTCAR, I notice that there is are huge forces in the second step (grep "TOTAL-FORCE" OUTCAR), indicating an issue with the structure.
In the second step, the C-O bond becomes very small, around 0.5 A, before the molecule dissociates in the third step. Two O atoms adsorb to the surface and the C atom in the vacuum above the surface. I found the same issue when using a PBE functional, so I think this indicates an issue with your initial structure. The CO and the O are on neighbouring sites in quite a small cell (2x2). Usually, I would expect the coverage to be at most 1 molecule per cell for a (2x2) cell, i.e. per 4 surface atoms, so 1/4 ML coverage. I recommend trying a (3x3) cell with the C and O spaced further apart.
Let me know if this helps.
Best wishes,
Chris