Nuclear-electronic calculations need uncontracted basis sets on the quantum protons

An attractive way to calculate nuclear quantum effects is to describe select nuclei quantum mechanically at the same level as the electrons, requiring the solution of coupled Schr"odinger equations for the electrons and the quantum nuclei. This is commonly known as the nuclear-electronic orbital (NEO) method, but it also has many other names. Two types of basis sets are required: a nuclear basis set is required in addition to the usual electronic basis set. In this work, we demonstrate that while existing nuclear basis sets are sufficient for NEO density-functional calculations, many sets producing proton affinities converged within 0.1 kcal/mol of the complete basis set limit, NEO calculations should always use uncontracted electronic basis sets on the quantum protons, since the contraction coefficients in typical electronic basis sets have been derived for point nuclear charge distributions. Uncontracting the basis sets on the quantized protons leads to significantly faster convergence to the basis set limit, leading to improvements of 18 kcal/mol and 10 kcal/mol in proton affinities employing double-$\zeta$ aug-pc-1 and triple-$\zeta$ aug-pc-2 electronic basis sets, respectively, with little effect on the computational effort. The partially uncontracted aug-pc-3 electronic basis set already affords proton affinities converged beyond 0.1 kcal/mol from the complete basis set limit. Similar results are also obtained with Dunning's correlation-consistent cc-pVXZ basis sets, as well as the Karlsruhe def2-XZP basis sets, albeit at a somewhat slower rate of convergence. As the protonic basis sets yield fully converged values, we find the protonic basis sets to be unnecessarily large for ground state density functional calculations, as the error in the protonic basis set is not balanced with that for typical electronic basis sets.