Computational search for nonlinear optical materials: are polarization functions important in the hyperpolarizability predictions of molecules and aggregates?

Hyperpolarizability of conjugated molecules with relatively long pi-system (three and more double bonds in the pi-bridge) and molecular aggregates (composed of the same molecules) can be predicted at the Density Functional Theory level with the 6-31G basis set to within 5-10% (compared to 6-31+G* results), while 6-31G* and especially MIDI! basis sets are much less accurate.