I am looking for an electronic structure method code (e.g. tight binding DFTB) that supports cheap calculations of NMR shieldings and has support for periodical systems. I would be grateful for any help!

There exist many codes using GIPAW and GIAO that use expensive DFT calculations.

However, the only papers using DFTB I have found appear not to publish their code or they do not support periodic systems (see e.g. https://pubs.acs.org/doi/abs/10.1021/jp9923062)

  • 1
    $\begingroup$ I guess the answer will depend heavily on the elements present in your system. For example some (hypothetical) program may support tight-binding-based NMR calculations for H,O,Al,Si only, while another program may support H,C,N,O,F instead. $\endgroup$
    – wzkchem5
    Feb 22, 2022 at 20:35
  • 1
    $\begingroup$ CENSO can do NMR, currently with implicit solvent. It is based on xTB, and they are implementing periodic boundary conditions, so I would keep an eye on it. xtb-docs.readthedocs.io/en/latest/CENSO_docs/censo_nmr.html $\endgroup$
    – B. Kelly
    Feb 23, 2022 at 10:24
  • 1
    $\begingroup$ It's possible that Phil didn't see your answer to his question, because you didn't use the "ping" feature. We only get notifications when you use the @ character before the name. I'm pinging @PhilHasnip now in case he didn't see the answer to your question :) It seems that you're looking for a code that can calculate NMR shifts in periodic systems at DFTB speed but with not much requirement for the accuracy. Does that help to write an answer Phil? $\endgroup$ Sep 2, 2022 at 21:28
  • 3
    $\begingroup$ Thanks @NikeDattani , I hadn't spotted it. I'll have to think about this, usually NMR calculations are very sensitive to the accuracy of the calculation, but I'll have a look and see if anything crops up. $\endgroup$ Sep 2, 2022 at 23:12
  • 2
    $\begingroup$ OK, it appears that there are some efforts to use machine learning to give quick estimates of NMR chemical shifts, e.g. nature.com/articles/s41467-018-06972-x . Initial results from one of my collaborators were not amazing, but not terrible either - perhaps 5-10x the difference from experiment, as compared to DFT (with the usual caveats: DFT structure, T=0 K, classical nuclei etc) $\endgroup$ Sep 21, 2022 at 23:20