I would like to know for which DF is useless to add dispersion corrections. For example, I suppose that for the nonlocal density-based functionals such as wB97M-V it is unnecessary because the dispersion effect is included inside the VV10 nonlocal correlation part. I am wrong? Can you list functionals (at least among the most used) for which is useless to add dispersion correction such as D3, D4 and so on? Thank you!

  • 3
    $\begingroup$ Some functionals like M06-2X are so heavily parameterised (i.e. their training data contain systems with vdW interactions), that they can reproduce energies of vdW systems in many cases without the need for D3. However, this would only work with the type of systems that were used to parameterise. I remember reading a study that showed that adding D3 correction to M06-2X or similar functionals still improved results, i.e. the default parameterisation is not enough. $\endgroup$
    – S R Maiti
    Apr 3, 2022 at 9:16
  • $\begingroup$ mattermodeling.stackexchange.com/questions/8863/… $\endgroup$
    – S R Maiti
    Apr 3, 2022 at 9:24
  • 1
    $\begingroup$ Thank you. I read the answers to this question and I also read some paper that include dispersion correction also to the M06-2X functional. As you say, for this Minnesota functional in theory the correction should not be necessary, but someone used it. My question is: for which functional can be considered wrong in principle to apply empirical dispersion correction? $\endgroup$
    – NickZ
    Apr 3, 2022 at 11:35

1 Answer 1


You should not apply empirical dispersion corrections to functionals that are overbinding already without dispersion. LDA functionals are a prominent example: even LDA exchange is overbinding by itself!


You must log in to answer this question.

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