Choosing between 6-311++G(3df,3pd) and def2-TZVPPD for certain type of diffuse system

Question

I need to use a diffuse basis set for my highly delocalised molecule, consisting of C, N, O, Cl, S, Mg, Al, B and Si only, since it has many non-covalent interactions inside it. After reading some more papers, regarding its predictable properties(sorry, but there are things called personal research secrets. They are irrelevant to the core of this question anyway), I decided that 6-311++G(3df,3pd) and def2-TZVPPD are the only choices I could use.

However, neither basis set is truly completely diffuse- 6-311++G(3df,3pd) augments only with p & d functions, while def2-TZVPPD does it only with s & d functions. Checked the BSSE website- same there.

What I need to know right now are the IP, EA, band gap and dipole moment- which basis set of the two should I use for these properties?

P.S. the functional I'll be using is M11-L.

Answer 1

Do not use Pople basis sets.

1. They are not systematical, as evidenced by the plethora of possible augmentations
2. They are only available for a small fraction of the periodic table (the first 3 periods)
3. 6-311G in specific is not a triple zeta basis set
4. The augmented basis sets lead to qualitatively wrong geometries

Instead, use systematic basis set families. Good choices include

• the Karlsruhe def2 basis sets, available for most of the periodic table, suitable for both Hartree-Fock / density functional theory as well as post-Hartree-Fock calculations (when augmented with the second set of polarization functions)
• for post-Hartree-Fock calculations, Dunning's correlation consistent family (cc-pVXZ)
• for density functional calculations, Jensen's polarization consistent family (pc-n or pcseg-n)

These modern basis set families come in systematic sequences of levels of accuracy, ranging from split-valence (def2 and pc families) or valence polarized double zeta (cc family) to fully converged (polarized quadruple-zeta for def2, quintuple-zeta for pc-n, hextuple-zeta for cc family).

Accordingly, you should be able to converge whatever property you are interested in to the basis set limit, and then trace back what basis set is the smallest to give you a value that is close enough.

Note also that M11-L is a pathological functional and as a semilocal functional lacks the ability to describe long-range dispersion. A recent review should be of help to pick a suitable functional and basis set combination.