Enhancing the Efficiency and Flexibility of AutoMeKin: Integrating ORCA and Machine-Learning Potentials

This repository contains the reaction-network data and level-of-theory benchmark for five automated reaction discovery (AutoMeKin) studies: cBD-CCH, cBD-CN, cBD-OH, Tz2-HA, and MEA. For each system it includes:- LL (low-level) exploration data, obtained with the semiempirical PM7 method, used to automatically generate candidate reaction networks (minima, transition states and products).- HL (high-level) refined networks at four independent levels of theory: DFT (ωB97X-D3/def2-TZVP), two machine-learning interatomic potentials (UMA-M and MACE-OMol), and a Δ-ML-corrected composite method (r2SCAN-3c plus a machine-learned correction toward the CC level, labeled DELTA in this repository).- CC single-point reference energies, computed on the DFT-optimized geometries (and, for a few method-specific channels, on the ML-potential geometries), used as the benchmark against which every other level is compared.- Node-by-node correspondence tables linking the internal numbering used by each level of theory to the labeling used in the corresponding published figures, together with the raw CC output files used to extract the reference energies. For further details on the models and computational methodology, please visit: https://github.com/ComputationalChem-USC/AutoMeKin-X.git