Basis Sets

• The use of minimal basis sets (e.g., STO-3G) is highly discouraged.
• Polarized double-zeta basis sets (e.g., cc-pVDZ, DZP, 6-31G*) are the smallest basis sets which can be counted on to give semiquantitative predictions. Try to use 6-31G** [also called 6-31G(d,p)] when you can instead of 6-31G* [also called 6-31G(d)].
• Pure Angular Momentum vs. Cartesian Functions: The Pople 6-31G* basis is defined for 6 Cartesian $d$-type functions, whereas the Dunning correlation-consistent basis sets are defined for pure angular momentum (5d, 7f, etc.) functions. Please find out whether your basis set should use pure angular momentum or Cartesian functions! Q-Chem will automatically pick the appropriate setting if it knows the basis. PSI3 and ACES II do not automatically know what to do; use the puream keyword if necessary.
• Balance your choice of basis with the theoretical method. Usually errors in basis set will largely cancel errors in the treatment of correlation if you pick a proper pair. For Hartree-Fock, use polarized double zeta basis sets. For MP2, CISD, CCSD, use either polarized double zeta or polarized triple zeta basis sets (e.g., TZ2P, cc-pVTZ). For CCSD(T), MRCI, etc, try to use polarized triple or quadruple zeta basis sets (or even cc-pV5Z if possible for tiny molecules).
• Diffuse functions: If you have an anion or want to study a Rydberg excited electronic state (or a mixed valence/Rydberg state), you must have diffuse functions. Must. Really. Diffuse functions will be present in Pople basis sets with + signs in them, e.g., 6-31+G*, 6-31++G**, etc. They are also found in the so-called ``augmented'' basis sets of Dunning and co-workers, e.g., aug-cc-pVDZ. Diffuse functions may also be useful for electronegative atoms, and should be important in van der Waals complexes.
• Pseudopotentials: Once you get to the second transition row or beyond (and maybe even for the first?) it is important to use pseudopotentials to approximately account for relativistic effects. I have little experience with these and can't recommend one set over another.
• General quality: In my experience, cc-pVDZ is bad. 6-31G** and DZP seem reasonable. Christian Ochsenfeld often argues that the Ahlrichs VDZP is better, but I have no experience with it. cc-pVTZ is probably slightly worse than TZ(2df,2pd) (also called TZ2Pf) but is easier to find built into packages. Atomic Natural Orbitals (ANO's) should be good, compact basis sets, but there is at least one case (Ca) where they act really odd (some old project of Fritz Schaefer's I forget, CaOH maybe with Steve Wesolowski).