This “consensus” PACO functional may serve as a measure of how well the computational chemistry community is doing if compared to state-of-the-art reference data.
It will change over time and hopefully will result in gradual improvements. To track the performance of the PACO functionals, it is checked for a variety of diverse interactions: atomization energies of six representative molecules (AE6), barrier heights of hydrogen-transfer reactions (BH6), π-π stacking of anti-parallel cytosine dimer, spin-state splittings of FeFHOH and Ni(EDT)22–, excitation energies (singlet, triplet) of CO, and hydrogen-bonding in four dimers (ammonia, water, formic acid, formamide). These small systems are studied using basis sets close to the basis set limit, and compared to either coupled cluster results or experimental data.
A surprising result from the poll was that the widely-used B3LYP is in fact not the most popular density functional, but came only second after PBE0 (also known as PBE1PBE).