Gaussian-n theory has been used for a long time to assess the accuracy of density functionals, and for some reason still appears to be used in the literature even though much more accurate alternatives like the Weizmann ab initio datasets, e.g. W4-11 and W4-17, are available.
The Gaussian-n test sets include at least
- G2-1 test set J. Chem. Phys. 94, 7221 (1991)
- G2 neutral test set a.k.a. G2/97, adding more molecules (G2-2 test set) to the G2-1 test set J. Chem. Phys. 106, 1063 (1997)
- G3/99 test set J. Chem. Phys. 112, 7374 (2000)
- G3/05 test set J. Chem. Phys. 123, 124107 (2005)
but none of these works contains the actual molecular geometries or the inputs that are available as supplementary information in the Weizmann sets.
I have found geometries for G2/97 in J. Chem. Phys. 136, 164102 (2012) that suggests new, more accurate reference energies for the database; but the geometries are in a PDF file which will be painstaking to parse, and the other databases are still missing.
Does anyone know where to find the databases in machine readable format? If I missed some Gn test sets, please let me know as well.