I am a beginning PhD student studying topological materials and strongly-correlated systems, such as high Tc superconductivity. I am using density functional theory in my work. I am wondering what I need to know in order to master DFT calculations using vasp (and also QE). I currently have some level of understanding of the theoretical basis of DFT (the Kohn-Sham equations, the Hohenberg-Kohn theorem, the Levy-Lieb functional etc.), but I do not have a lot of understanding on the pseudo-potential formalism (e.g. PAW), and the algorithms used in VASP to calculate the electronic ground state (e.g. the blocked Davidson algorithm and the RMM-DIIS algorithms).

I am wondering if it pays to spend time trying to understand how the DFT code actually works and if so, what is the best way to do that. My principal goal is to do DFT calculations to understand topological systems and strongly-correlated material systems. This involves building tight-binding models etc.

  • $\begingroup$ This might be a helpful start: mattermodeling.stackexchange.com/questions/7182/… $\endgroup$ Aug 31, 2022 at 23:54
  • $\begingroup$ I also wrote some information about optimisation methods here: mattermodeling.stackexchange.com/questions/8795/… $\endgroup$ Sep 1, 2022 at 0:02
  • $\begingroup$ That's great. Thanks. I haven't seen that answer. I spent a few days trying to understand PAW, both the blocked Davidson algorithm and the RMM-DIIS algorithms. I got some general idea, but it is still not clear to me since I don't have a background in numerical linear algebra. $\endgroup$
    – dbrane
    Sep 1, 2022 at 0:05
  • 1
    $\begingroup$ If you'd like to post a more specific question (e.g. "how does density mixing work?") you'd probably get more answers (I could answer that one, for example). $\endgroup$ Sep 1, 2022 at 0:25
  • $\begingroup$ Some information on the Davidson method: mattermodeling.stackexchange.com/questions/6563/… $\endgroup$ Sep 1, 2022 at 0:26

1 Answer 1


You should start off with running DFT calculations and then over time you can learn about theoretical and implementation details that interest you or are important for what you are doing. Based on what you've said, you should have the requisite knowledge to understand what is going on initially; once you've done calculations, you will have a better sense for what additional theoretical topics you will need to cover.

Different people with different research aims will need to spend more or less time on different subtopics. For example, many people don't need to learn much more about PAW than having some vague understanding of the "PAW formalism" section of the VASP wiki, whereas other people will have to go deep into the details and even implement code that must take into account the PAW formalism.

If you will be working on tight binding models and DFT, then Wannier90 likely will end up being useful. Perhaps you could do some DFT tutorials in VASP or QE and then also follow some of the Wannier90 tutorials. Along the way, you'll probably run into questions or things that interest you that can help inform what topics you will spend more time working with.

  • $\begingroup$ Great, thanks. I think I can do the calculations in vasp without having a very good understanding of how the code actually works, but I thought I might be able to avoid the convergence issues that I tend to get if I understand how it actually works. For example, in magnetic systems, one has to change the mixing parameters sometimes to get a convergence. I do not understand how this works at all $\endgroup$
    – dbrane
    Aug 31, 2022 at 23:47
  • 1
    $\begingroup$ This is a really general question, it's going to be difficult to answer properly and cover PAW, blocked Davidson (and alternatives?), density mixing etc. I suggest you ask a specific question, e.g. about how density mixing works. $\endgroup$ Aug 31, 2022 at 23:53

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .