When performing electronic structure calculations on a molecule one can surround it with point charges to mimic a solvent environment and lend polarization effects. There is a bit of a caveat unfortunately.

If diffuse functions are used, there is a risk that the point charges surrounding the "cavity" that the molecule of interest is in, will lead to distortion of the electron density. Another way of saying this is that the electron density will(may) migrate to towards the point charges in an unphysical manner.

This is unfortunate, but, not too big of a deal for neutral and cation species which are not reliant on diffuse functions (just don't use them). However, anions really should be done with diffuse functions to account for the loose electron. This applies to anything where electron density will not be as tight around nuclei, i.e., excited states.

Is there a ready solution to this problem of charge migration?

This also applies when using implicit solvents as well, the tessellated surface of the cavity can lead to unphysical migration of the solute's electron density towards the cavity surface.

  • $\begingroup$ The edits I made should bump this unanswered question up (and I do believe all edits were necessary anyway). Now that I've thought about your question more, I have some questions: How specifically can you say that diffuse functions are the problem, when in fact aug-cc-pVDZ has "diffuse functions" but not always as diffuse as the most diffuse function in cc-pVQZ (which does not have "diffuse" functions in the aug- sense, but does have smaller exponents than the "diffuse" functions of aug-cc-pVDZ)? Next: we use diffuse functions to bring us closer to the CBS limit, your concern is that you $\endgroup$ – Nike Dattani Jun 12 '20 at 22:29
  • $\begingroup$ don't want the functions of the basis set to come in contact with any of the point charges? If you just want to increase the accuracy of the electronic structure calculations on the solute without using basis functions that extend all the way to the point charges, I suppose you could just add a lot more non-diffuse functions, which will still bring you close to the CBS limit without using diffuse functions. Unfortunately I'm not familiar enough with your specific problem to answer it properly, or even appreciate it! Are you using COSMO? $\endgroup$ – Nike Dattani Jun 12 '20 at 22:32
  • $\begingroup$ When using anions you must use diffuse functions, since the loose electron will correspond to higher electron density further from the core. You raise a good point about quadruple zeta vs triple zeta with diffuse. Quadruple is also a problem. I never thought of the problem as being due the functions in the pVQZ, but that would make perfect sense. Actually, this may be quite significant. I need to ponder this, but it all fits with the general problems I am having with polarized electron densities. $\endgroup$ – B. Kelly Jun 12 '20 at 23:59
  • $\begingroup$ I find that people use the word "diffuse functions" as if they are something more special than just the basis functions with the smallest exponents in the "augmented" basis sets. If you take cc-pVDZ and you add one more S, one more P, one more D, etc., such that the exponent is smaller than all the others, you get aug-cc-pVTZ (which has "diffuse" functions). But if you add several more S, several more P, several more D, you get cc-pVTZ which does not "have diffuse functions" in that sense, but it has diffuser funcitons than cc-pVDZ. Does it make sense? What's wrong when you use QZ? $\endgroup$ – Nike Dattani Jun 13 '20 at 0:03
  • 1
    $\begingroup$ I agree, it is all gaussian functions with exponents, but I never actually compared the diffuse exponents to those in larger basis sets which is a noob mistake ;) $\endgroup$ – B. Kelly Jun 13 '20 at 0:16

I have some limited experience with implicit solvents and attempting to polarize the system. I would suggest that if you cannot afford to model the real solvation environment, try using a shell of solvent which can be polarized correctly. This is also quite tricky to get working correctly, but may improve the situation since you are no longer using point charges directly to get polarization of the molecule.

  • 1
    $\begingroup$ +1. Wow this question was asked on 26 May and you answered on 26 July. It's a close call, but let's see if you get the silver necromancer badge. $\endgroup$ – Nike Dattani Jul 26 '20 at 18:57
  • 2
    $\begingroup$ This is actually quite a common problem, while I don't know much about the reason it happens exactly (The person asking has some experience with what basis sets actually cause the problem), this problem shows up quite often when doing this sort of work. $\endgroup$ – Tristan Maxson Jul 26 '20 at 20:20

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.