ERROR FEXCP: supplied Exchange-correletion table

[KWH]

Hello:

Part way through a calculation I am getting the error:
ERROR FEXCP: supplied Exchange-correlation table is too small, maximal index : 5325

I don't know what it means and I don't understand why it is happening after one complete electronic optimization. I have attached all the relevent files. Your help will be appreciated.

[henrique_miranda]

Two things seem to be happening:

1. The SCF part of the calculation is slow to converge but it seems to stop because you have have a relatively high tolerance for the convergence of the energy: EDIFF = 0.001

2. This might be inssuficient to get accurate enough forces. Indeed looking at them in the OUTCAR I see that some of them are quite large.
   This might make the next ionic step into a very different structure and then you get this error.

The first thing you should probably try to do is see if the electronic minimization converges with a minimum version of the INCAR you have:

Code: Select all

 ISYM = 0
 NCORE = 64

# Accuracy controls:
 PREC = ACCURATE         ## want ACCURATE eventually
 ENCUT = 500
 SIGMA = 0.1
 EDIFF = 1e-6 # this needs to be lower than your setting
 
## opt PBE + vdW
GGA = OR
AGGAC = 0.0
LUSE_VDW = .TRUE.
LASPH = .TRUE.

If this works, then you can try using IBRION=3 in the first steps of the ionic minimization.
Let me know if this helps!

[KWH]

Henrique

It changed the error but still crashed (files attached). This time it failed to converge even at the first electronic optimization. Should I next try the ALGO=ALL I used in my previous attempt? It did get through the first electronic optimizaation?

[KWH]

I did try to change ALGO to ALL, but the electronic cycle still diverged.

[henrique_miranda]

Ok, it looks like it is oscillating around the solution.
You tried ALGO=ALL that is a good idea.

Can you try the following steps:

1. ALGO=All, with TIME=0.05, NELM=120 and with low ENCUT (if you don't set it then VASP will use the MAX ENMAX of the POTCARS).
2. If step 1 converges then try the ionic relaxation with IBRION=3 and NSW=10 or so.
3. Once you reach a more stable ionic convergence you can try to increase TIME. This will speed up the electronic convergence allowing you to do faster ionic steps.
4. Increase ENCUT

Let me know if this helps.