SCF Convergence to Metallic State in Inorganic Systems

similt

Dear CRYSTAL Support Team,

I hope this message finds you well.

I have been using the CRYSTAL software for some time and have generally found it to perform very well, particularly for organic molecular systems. However, I have recently encountered persistent SCF convergence issues when working with inorganic materials.

Specifically, I observe that during the self-consistent field (SCF) cycles, the calculations often converge to a metallic solution instead of the expected insulating state. This behavior occurs for several inorganic systems I have tested. For instance, in my current study of a CdS slab (input file attached), calculations using various functionals converge to a metallic state, whereas bulk CdS correctly yields an insulating band gap of approximately 3 eV at the HSE06 level.

For comparison, my VASP calculations for the same systems converge reliably to the correct insulating solution. This suggests that the issue may be related to the SCF procedure in CRYSTAL, particularly for slab or defect systems.

Could you please advise on strategies or recommended settings to avoid this incorrect metallic convergence? Any suggestions or example input configurations would be greatly appreciated.slab-m06.d12.txt

eascrizzi

Hi Similt!

You may want to try the LEVSHIFT option, see page 118 of the CYSTAL User's Manual. With this keyword you can better separate occupied and unoccupied states.

Let me know if it helps,
Eleonora

GiacomoAmbrogio

Hi similt,

The convergence difficulties you're facing are likely related to the more complex electronic structure of metallic states. These issues arise when the system is metallic, but they can also appear in insulating systems if the SCF goes through some metallic states, and usually it can be stuck there (as seems to be happening in your case).

The general suggestion is to use the SMEAR keyword , that can help a lot when dealing with metallic states.
Also I would suggest to remove the BROYDEN convergence accelerator; and use the default option DIIS.
For metaGGA functionals such as M06, it’s also a good idea to increase the integration grid size (by defayult it should be raised to XXXLGRID, but you can push it further using HUGEGRID)

Following these points, I tried running your calculation on our cluster, and it converged nicely in 12 cycles, giving a bandgap of 3.29 eV with PBE0.
Indeed, during cycle 1 and 2, the system shows metallic behaviour, but from cycle 3 onward it becomes insulating.

Here there is the input file I used: inputPBE0.d12

Let me know if you need further help.

similt

Thanks. SMEAR helped me to converge scf.