SC steps converging with Zn but not Cu in supercell

[hubertc]

Dear Admin, dear VASP'users

I'm working on 2 materials, one based on Zn2+ and the second based on Cu2+. Both solids have the same structure with only a small difference in the lattice parameters arising from the different atom size of Zn2+ and Cu2+.

I'm trying to calculate the ground state of both solid in a supercell 112 but only the Zn-based solid is converging to EDIFF = 10E-4 in the electronic SC steps while the Cu-based solid reachs the "default" 60 SC steps wihtout converging, the dE in the OSZICAR changes between + and - 0.5E-1.

I tried to calculate the ground state in the unit cell first for both solid, it is converging in less than 30 SC steps for both solids, then from the relaxed structures, I created the supercell 112, again the Zn-based solid converges but not the Cu-based solid. I've also tried difference calculations set up such as NELM to 500, use the relaxed lattice parameters and the experimental lattice parameters (at 5K), use the pseudopotential Cu_pv, IBRION 1 or 2, POTIM to 0.1.

my only success has been to increase SIGMA from 0.05 to 0.2 (with ISMEAR=0) but 0.05 is working fine with the Zn solid and normally 0.2 is for metals, and those solids contains some organic parts.

i'd really appreciate your help.

best regards
Hub

[admin]

LMAXMIX=4 can help

[hubertc]

thank you. Indeed it helped, it's converging now. Nonetheless, the first time I tested the LMAXMIX flag, it did not make any difference, I was using IBRION=1 then now with IBRION=2, it is working fine, any particular reason behind this? also, i'm a bit surprised that the calculations are working fine with Zn but not Cu, is there any reason? could it be due to the pseudopotentials?
regards
Hub

[admin]

Pseudopotentials are developed in a way to be able to reproduce properties of solids.
Because the solution (ground state) is searched for using numerical tools
there are always possible situations when the numerical approach does not work.