When to use Hubbard U and in what form when soling for BCC Fe

[abc123def]

I am trying to optimize the shape and size for bcc Fe. In the basis set, I am using spin polarized Tight Tier 1 basis set with FHI (Z=8) pseudopotential. Do I need to turn on the Hubbard U for energy minimization/geometry optimization? If yes, should I use onsite mode or dual mode?

[Jess Wellendorff]

I have 2 comments for this:
1) No, you should not use a Hubbard U for geometry optimization. The +U method is in general not intended for first principles forces.
2) The Tight Tier 1 basis has a very large cutoff radius, which can be computationally intense. I suggest you use the DZP basis for the FHI pseudopotential, at least for an initial geometry optimization. Should be much faster.

[abc123def]

Jess Wellendorff: Thank you. Can you please explain when we need to include +U in general?

[Jess Wellendorff]

There are probably many different attitudes towards that question, but the +U method provides in general a pertubation to the electronic eigenvalue spectrum (hence: modified band structures), and is sometimes also useful for energetics, e.g. transition-metal oxide formation energies. But no forces from the +U term are available, so it cannot be used for geometry optimization...

[Jess Wellendorff]

Update: It looks like you can actually use the Hubbard U for relaxing forces and stress with ATK-DFT.  Sorry for the confusion.

[abc123def]

Thanks again !