LDA+U in ATK

[yasheng]

Hi,

I have some problems with setting up calculations for Cu/Cu2O interface. I want to disable Hubbard U for metallic Cu while I want to use Hubbard U for the Cu atoms in Cu2O.  Can you please give some advice on how to do it?

Regards,
Yasheng

[Jess Wellendorff]

You need to be able to define 2 different basis sets for the copper atoms with and without a +U term. Use tags to do this:
1) In the Builder, set some tag, e.g. "U", on the Cu atoms that should have +U enabled.
2) When setting up the calculator exchange-correlation, click "Activate tags", and tick your tag.
3) Then turn Hubbard-U on for the tagged atoms but off for the others. Remember to set the value of U.
See attached PNG.

[yasheng]

Hi Jess,

Thank you for your advice. I prepared the input files for Cu/Cu2O interface according to your post. However, calculation stopped with error "Exception: 'tuple' object has no attribute 'element'".

I attached my input files, can you please check and if there is anything wrong in the input? Thanks.

Regards,
Yasheng


 

[Jess Wellendorff]

Hi Yasheng.

You have unfortunately stumbled upon a bug in ATK. We will fix it for the upcoming 2016 release.
In the meantime, I have implemented a temporary workaround in your script. New script attached. Instead of using the tagging mechanism, I have simply replaced all the Cu atoms in the right-hand copper part with uranium, and then defined a fake copper basis+pseudopotential to be used for the uranium atoms. This makes it possible to attach a basis+pseudopotential including Hubbard U to the copper atoms in the oxide, while the other copper atoms do not have the U. Do not worry about the use of uranium atoms, it looks peculiar in the Viewer and Builder, but the calculation will really be for ordinary copper.

Hope this helps you.

[Jess Wellendorff]

Forgot to mention a few general comments on you script:
1) Your k-point sampling looks insufficient in the (periodic) A and B directions.
2) The last few lines, where you set an initial spin, do not really have an effect. The SCF calculation is executed before that.

[yasheng]

Hi Jess,

This trick works fine for me. I look forward to the new version of ATK.

Thanks for your help and regards,
Yasheng