Theory Benchmarking Set Generation Scheme

Contributors: @Hyesu Jang @Lee-Ping Wang

Goal: Design a QM benchmark set to find a level of theory that provides the best compromise between accuracy (in torsion barriers and optimized geometries) and computational cost.

Reference: A preliminary selection of molecules provided by L-P.

https://docs.google.com/presentation/d/1eb5GGutg1AWdapTCcLMBtDVCDcAOgMHXe74Peh0YKlQ/edit?usp=sharing

Selection Scheme:

1. Consideration of the coverage of chemical diversity

  • Variations in central bonds, formal charges, element compositions, intramolecular interactions;

  • Inclusion of molecules (1) w/ non-zero formal charges, (2) w/ strong internal interactions, (3) w/ central bond conjugated( < 10 kcal/mol rotational barrier) or (4) w/ halogen

2. Consistency in molecular size

Selected torsions:

  • Starts with torsions selected by L-P ( torsions in slide 16), Consider coverage of central bonds defined in the ff, and select one torsion per each central bond in a way to minimize the overlap of functional groups between molecules;

  • 36 1-D torsions were selected from Roche and Coverage set.