W1-S_N2-BH: A Large-Scale CCSD(T)/CBS Kinetic Database

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Theoretical Chemistry

Background:

• Accurate prediction of reaction barrier heights is crucial for understanding chemical kinetics.
• Existing benchmark databases for nucleophilic substitution (SN2) reactions often lack sufficient diversity or high-level theoretical accuracy.
• Density Functional Theory (DFT) methods require rigorous validation against accurate reference data.

Purpose of the Study:

• To establish the W1-SN2-BH database, a comprehensive kinetic benchmark set for SN2 reactions.
• To rigorously evaluate the performance of various Density Functional Theory (DFT) and double-hybrid DFT methods.
• To assess the impact of empirical dispersion corrections and explore the potential of machine-learned functionals.

Main Methods:

• Calculation of all-electron CCSD(T)/CBS reference barrier heights for 1881 SN2 reactions using high-level W1w theory.
• Systematic benchmarking of 40 DFT and double-hybrid DFT methods, including 19 D4-corrected variants.
• Evaluation of a deep-learning functional (Skala) and analysis of the effect of empirical dispersion corrections.

Main Results:

• The W1-SN2-BH database covers diverse chemical space with barrier heights spanning a broad energetic range.
• Range-separated double-hybrid ωB97M-2 showed exceptional performance (MAD = 1.19 kcal mol⁻¹), outperforming other tested DFT methods.
• The deep-learning functional Skala outperformed conventional local functionals, and empirical dispersion corrections were found to systematically deteriorate performance for some methods.

Conclusions:

• The W1-SN2-BH database provides a rigorous and diverse benchmark for developing and validating next-generation computational methods.
• Specific DFT functionals, particularly range-separated double-hybrids, demonstrate high accuracy for SN2 reaction barrier heights.
• Machine-learned functionals show promise, and careful consideration of dispersion corrections is needed for accurate barrier height predictions.