SCF convergence issue for Fe2O3/Co3O4 heterojunction using diagonalization

[milk milk]

Dear colleagues ,

I am trying to optimize an Fe₂O₃/Co₃O₄ heterojunction structure using diagonalization, but I have been unable to achieve SCF convergence despite many attempts.

I have tested several different MIXING METHOD settings and also increased the electronic smearing temperature from 300 K to 1000 K.  I have set &BS for Fe, Co and O. And I set DFT+U for Fe and Co. However, the SCF residual still only decreases to around 0.1 and does not converge further.

Apart from switching to the OT method, are there any other approaches or SCF settings that I could try to improve the convergence for this system?

I have attached my input file for reference. Any suggestions or advice would be greatly appreciated.

Best regards,

Milk

[milk milk]

This is my input file, Thanks for any suggestions!

[Shiyang Wang (W0728)]

Hi,

Currently in develop branch of CP2K, OT with k-points has been available.

For diagonalization, you can try Kerker mixing with common settings:

```

      &MIXING
        METHOD KERKER_MIXING
        ALPHA 0.4
        BETA 1.6
        ALPHA_MAG 1.0
        BETA_MAG 0.0
        NBROYDEN 8
      &END MIXING

```

Alternatively you can try ADIIS but it usually spends much more steps than common density mixing schemes.

Regards,

SY

在2026年8月31日星期一 UTC+8 17:09:22<milk milk> 写道：