Limitations of extremes for the exchange-correlation functional - DFT

Question

In DFT the exchange-correlation function is critical for the correct description of the system. I have a question about the applications and shortcoming of some functionals. LDA delocalizes electrons so are good for conductors. And pure HF localizes electrons so it's good for molecules/atoms.

So, is it safe to say that if one doesn't include correlation and only uses HF exchange energy that crystals, or any material with long range ordering, couldn't be properly described?

And the same around, if one only uses correlation, conducting crystals are properly described but no molecules/atoms could be studied correctly?

I want to understand the applications and limitations of both extremes and I am having a hard time finding that kind of analysis online.

Thanks!

Answer 1

One way to begin to gain some insight in this is to consider the work done on benchmarking density functionals. I shamelessly plug the work done by my old group, which can be found here. The GMTKN30 data include results for Hartree-Fock and LDA functionals with large basis sets. The description of the several benchmark sets involved and the analysis in the actual paper will illustrate many of the shortcomings and strengths of these approaches (and why better, more pragmatical DFT approaches are more widely used). While I have no access to the GMTKN55 paper, I trust that it contains references to other benchmark papers that may actually deal with crystals.