Basis Sets for Organic Molecules: Single-zeta vs. Double-zeta

A few tests have been done on small (less than 10 atoms) and medium sized (20-30 atoms) organic molecules (not containing transition metals) to compare double-zeta with single-zeta (minimal) basis sets. The two procedures were: a) a straightforward geometry optimization with the double-zeta basis set, and b) a geometry optimization with a single zeta basis, followed by a double zeta single-point calculation in the optimized geometry to evaluate the bonding energy and other properties. The results differed very little as regards the final geometry and therefore also as regards the energy etc.: from less than 0.01eV for small molecules to 0.25eV for a 26-atom case (debrisoquine CHN).

The additional single-point double-zeta calculation, required to obtain the bonding energy, makes that the computational costs are not automatically lower for the single-zeta optimization procedure. For the smaller molecules they take indeed some 25% more time. For the larger molecules this seems to get reversed however, the single-zeta approach being less than half as expensive as the double zeta.

Results for optimized bond-lengths in single-zeta bases are found to be very inaccurate when Sulphur atoms occur in chains. In such molecules these atoms need a 3d-polarization function, as provided in adf's standard type-DZP (old name III) (double-zeta plus polarization) basis sets. Results so far suggest that this particular problem with single-zeta bases does not occur when such atoms occur in a ring rather than in a chain within the molecule. Probably this is related to the empty d-shell in the atom being rather close to the valence p-shell. It can be expected that the same behavior will be displayed by Phosporus. Obviously one should not rely on such generalizations too strongly; it is sensible to always run a few tests and verify whether it can be applied to the case at hand.

Although the remarks above suggest a promising time-saving approach for the optimization of larger organic molecules, the conclusions must be considered with very great care since these investigations have been carried out only for a very small number of molecules. Also one should be extremely careful to extend the conclusions to transition metal complexes.