An Imbalance in the Force: The Need for Standardized Benchmarks for Molecular Simulation

Area of Science:

• Computational Chemistry
• Molecular Simulation
• Cheminformatics

Background:

• Force fields (FFs) are crucial for molecular simulation but lack standardized benchmarks for rigorous testing.
• Current FF development often relies on proprietary training and test sets, hindering model comparison.
• Existing benchmarks are common in quantum chemistry but less so for empirical FFs.

Purpose of the Study:

• To provide an overview of available tests and benchmarks for computational chemistry, focusing on organic compounds, halogens, and ions.
• To assess the reusability of quantum chemistry benchmark datasets for evaluating FFs.
• To identify data needs for improving FF development and validation.

Main Methods:

• Literature review of existing computational chemistry benchmarks.
• Analysis of applicability of quantum chemistry datasets for FF evaluation.
• Identification of data gaps for specific chemical elements and properties.

Main Results:

• Many quantum chemistry benchmarks can be adapted for FF evaluation.
• New gas-phase data are needed for phosphorus and sulfur compounds in various valence states.
• Enhanced data on nonequilibrium interactions and molecular properties like electrostatic potentials are beneficial.

Conclusions:

• Standardized benchmarks are essential for comparing and improving molecular simulation force fields.
• Reusing existing quantum chemistry data and generating new data for specific elements can advance FF development.
• Adoption of common datasets by FF developers and AI researchers will facilitate model comparison and progress.