9
$\begingroup$

I have seen that some articles calculate Hubbard $U$ value self-consistently using Quantum Espresso code. Then, they use it in VASP calculations. My question is: In which conditions can this value be transferred from a code to another?

$\endgroup$
3
  • 2
    $\begingroup$ Its a bit too short to be an answer, but I fear the answer is never. I hope I am wrong though since this would be useful to me. $\endgroup$ Nov 18 '21 at 13:20
  • 2
    $\begingroup$ It might help if you link to the article where you noticed that it was being done this way . $\endgroup$ Nov 19 '21 at 14:26
  • 1
    $\begingroup$ @RashidRafeek Reference to the article has been added. $\endgroup$
    – Chi Kou
    Nov 20 '21 at 12:18
7
$\begingroup$

I have looked into this a bit, I think in general it is ill-advised to mix codes like this unless it is needed, but VASP does not support self-consistent hubbard U (no perturbations at non-zero U). As long as you can prove the two codes produce the same results, it is probably okay. Specifically, if you can prove the property of interest is the same at U=0 and U=Uscf, then I would not be too skeptical though.

The main issue here is the U parameter is dependent on everything about the system. Energy cutoff, grids, psuedopotentials, etc. could all make a difference so it should probably not be directly used without checking that band gap is the same for example.

$\endgroup$
4
  • 1
    $\begingroup$ There are also several different "Hubbard U" methods and basis sets used -- indeed, VASP alone has several different Hubbard methods implemented. As an example, are you using the method as originally proposed, or a "rotationally invariant" one? Are you computing the occupation matrix by projecting onto a pseudo-state basis, or a hydorgenic basis, or a Gaussian basis...? $\endgroup$ Nov 22 '21 at 0:32
  • 1
    $\begingroup$ @PhilHasnip Yeah this is exactly the reason I hesitate to say its okay, but maybe if everything aligns then its appropriate but only in the sense that you can get two models to agree $\endgroup$ Nov 22 '21 at 0:59
  • 1
    $\begingroup$ 🧐I think there is a hidden feature allowing users to calculate U self-consistently with VASP. See chengcheng-xiao.github.io/post/2019/07/06/Hubbard_U.html. $\endgroup$ Nov 22 '21 at 14:39
  • 1
    $\begingroup$ This is for the SCF perturbation calculation. self consistent U requires the application of U with a perturbation, not just unfixing the charge density. $\endgroup$ Nov 22 '21 at 16:14

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.