NENCI-2020: A Large Benchmark Non-Equilibrium Non-Covalent Interaction Database with Emphasis on the Repulsive Wall

In this work, we present NENCI-2020: a benchmark database of non-equilibrium non-covalent interaction energies for a large and diverse set of intermolecular complexes. NENCI-2020 contains ~8,000 interaction energies computed with CCSD(T)/CBS, and includes 7 non-equilibrium intermolecular distances (spanning 0.7-1.1x the equilibrium separation) and 8 non-equilibrium angles (per distance) for ~150 complexes. Using SAPT2+, we demonstrate that NENCI-2020 contains a diverse array of intermolecular binding motifs, making this database well suited for testing and developing next-generation force fields, density functional theory (DFT) approximations, quantum chemical (QC) methods, and machine-learning based approaches. This is followed by a critical assessment of ~75 dispersion/van der Waals-corrected DFT and QC methods, in which we find that most approaches can describe interaction energies for equilibrium and farther-than-equilibrium configurations with chemical accuracy (i.e., to within 1 kcal/mol). More importantly, we find that nearly all methods suffer from a rapid and systematic increase in error as the intermolecular distances become small, thereby suggesting that more work will be needed to describe intermolecular potential energy surfaces with uniform accuracy.