11
$\begingroup$

I have heard people talk about DFT functionals built starting from MP2 corrections. Do you happen to have a reference I read about that from?

$\endgroup$
12
$\begingroup$

These are called double hybrid functionals. You add a fraction of MP2 correlation on top of the DFT energy.

A good reference is e.g. the review by Goerigk and Grimme from 2014 published in WIREs 4, 576 (2014). Note that this is already 7 years old; many double hybrids have appeared since then. However, it should give you a good overview.

$\endgroup$
4
  • $\begingroup$ Thank you very much @Susi $\endgroup$
    – Yepman
    Jul 23 at 19:46
  • $\begingroup$ @Yepman you've very welcome! $\endgroup$ Jul 23 at 21:58
  • 1
    $\begingroup$ +1. @Yepman I noticed you "accepted" the answer but didn't cast any votes. You don't have to, but generally if an answer is helpful it's a good idea to upvote (it looked a bit bizarre here when there was a green checkmark but 0 upvotes!). $\endgroup$ Jul 23 at 22:16
  • $\begingroup$ Thank you very much for noticing that @Nike Dattani $\endgroup$
    – Yepman
    Jul 24 at 7:11
11
$\begingroup$

Susi is right that these are called double hybrids, I'll just add a bit more to complement that correct answer. My answer to "What are some recent developments in density functional theory?" was about double hybrids, and it looks like the community here considered that to be one of the most important "recent" developments in DFT.

About the MP2 contributions, single contributions are ignored (this is already the case when starting from Hartree-Fock because Brillouin's theorem says they don't exist, but when starting from a DFT calculation single excitations can contribute, and early developers of double hybrid functionals decided not to include them), see this: Why are single excitations ignored in the MP2 component of double hybrid functional calculations?

Another interesting discussion about double hybrids can be found here: Physically motivated double hybrid DFT?, and here: Using double-hybrid functional instead of MP2 to avoid basis set dependence, and if you're interested in going even further than MP2, there's been some papers that used QCISD or RPA: Extended Hybrid Methods.

Finally, double hybrids have been recommended or their accuracy has been discussed, in a lot of answers on this site, for example:

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.