I want to do some DFT calculations for structures of protonated alkaloid dimers. I know that a method of choice when vdW interactions are involved is the use of dispersion corrections. But now I am asking myself if there is a rule of thumb for choosing the right basis set for describing long-range interactions accurately. How relevant is it to consider the different types of basis sets (correlation-consistent, split-valence, etc.) ? Or is it sufficient just to choose a basis set with diffuse functions ?
If you use DFT-D3, vdW interactions are already included in the empirical dispersion correction.
The general guidelines for choosing basis sets apply: use modern families of consistent basis sets. For DFT studies you should look at the Karlsruhe def2 series or Jensen's pc(seg)-n series; the correlation consistent basis sets are not cost efficient for DFT. Examine the sequence of basis sets, and examine your property's convergence in the cardinal number, e.g. def2-SVP $\to$ def2-TZVP $\to$ def2-QZVP, and the sensitivity of diffuse functions, e.g. def2-SVP $\to$ def2-SVPD.
If you want a single answer, def2-TZVPPD is probably practically at the basis set limit for DFT-D3.