Benchmarking B3LYP, PBE0 and M06L for nuclear spin–spin couplings against CC3: influence of geometry optimization

Abstract

We benchmark the DFT functionals B3LYP, PBE0, and M06L for calculating 45 indirect nuclear spin–spin coupling constants (SSCCs) in 13 electron correlation-rich small molecules, using high-level CC3 reference values from a previous study. The dataset spans a wide variety of one-, two-, and three-bond couplings, enabling a systematic assessment of functional performance and the impact of geometry optimization on SSCC calculations. Overall, PBE0 combined with PBE0-optimized geometries provides the closest match to CC3 results, followed by M06L at B3LYP or M06L geometries. A strong functional dependence is observed for fluorine-containing couplings: M06L markedly outperforms PBE0 and B3LYP for these, especially for one-bond C–F couplings, whereas PBE0 is clearly superior for couplings not involving fluorine. Across all molecules, the geometry-optimization method has a significant influence on the resulting SSCCs, and the best performance is consistently achieved when the same method is used for both geometry optimization and SSCC computation. In contrast, CCSD(T) geometries systematically yield the poorest agreement with CC3 SSCCs when combined with any of the three DFT functionals.