WF convergence issue

[aldo_romero1]

I have been trying to converge the wavefunction of a system I am dealing with. The energy goes down for few iterations as:
N E dE d eps ncg rms rms(c)
DAV: 1 -0.217186945372E+04 -0.21719E+04 0.97968E+13 18496 0.158E+12
DAV: 2 -0.979649664299E+02 0.20739E+04 -0.40086E+03 20968 0.447E+02
DAV: 3 -0.166678657388E+03 -0.68714E+02 -0.63448E+02 26664 0.969E+01
DAV: 4 -0.171380884559E+03 -0.47022E+01 -0.46862E+01 27940 0.209E+01
DAV: 5 -0.171568694410E+03 -0.18781E+00 -0.18775E+00 26256 0.411E+00 0.348E+03
RMM: 6 0.700459931069E+04 0.71762E+04 -0.35924E+08 29412 0.569E+06 0.425E+05

But then I get the error:

| ERROR FEXCF: supplied exchange-correlation table |
| is too small, maximal index : 6025 |
| |
| ----> I REFUSE TO CONTINUE WITH THIS SICK JOB ... BYE!!! <---- |

Interatomic distances are fine, cutoff is fine, K-mesh grid seems OK, though I have tried different values.

Any suggestions on how to tackle this problem?

Thanks!

PS Let me know if you need any further information.

[marie-therese.huebsch]

Hi, thanks for raising the issue.

Can you please share the files for a minimal reproducible example in accordance with the forum guidelines?

Cheers,
Marie-Therese

[aldo_romero1]

I am attaching a tgz file with all necessary information to reproduce my problem.

Thanks

-aldo.

[marie-therese.huebsch]

Hi, thank you for sharing the files. I see that you have chosen NCORE=32. Though likely inefficient, that should be fine, since you are running on 32 ranks. Still some issue occurs due to the large value of NCORE. You can see already the first electronic step complains about a nonhermitian subspace matrix. I will look into the reason.

As a fix for your calculation, I suggest running a few steps (NELM=10) with NCORE= 2, 4, 8, 16 and select the one most efficient. For this test, particularly the number of ranks, PREC, ENCUT should remain the one you will use in the final calculation.

Could you please let me know if this helps?

Best regards,
Marie-Therese