VIB5 database with accurate ab initio quantum chemical molecular potential energy surfaces

Area of Science:

• Computational Chemistry
• Astrophysics
• Machine Learning

Background:

• Accurate molecular potential energy surfaces (PESs) are essential for simulating rotation-vibration spectra.
• High-level ab initio quantum chemical (QC) calculations are computationally expensive for generating extensive PES data.
• Machine learning (ML) offers a cost-effective approach to constructing PESs but requires large, high-quality datasets.

Purpose of the Study:

• To introduce the VIB5 database, a structured collection of ab initio potential energy data for small, astrophysically relevant molecules.
• To provide a publicly available, high-quality dataset suitable for machine learning applications in molecular spectroscopy.
• To enhance existing PES data with new QC calculations and ensure consistency for ML model development.

Main Methods:

• Generation of high-density grids of nuclear geometries for five small polyatomic molecules (CH3Cl, CH4, SiH4, CH3F, NaOH).
• High-level ab initio quantum chemical (QC) calculations to determine potential energies and energy gradients.
• Compilation and structuring of data, including theoretical best estimates and energy correction terms, into the VIB5 database.

Main Results:

• The VIB5 database contains tens of thousands of grid points for each of the five molecules.
• Provides theoretical best estimates of potential energies and their components.
• Includes a data-extraction script and complementary QC calculations for gradient-based ML methods.

Conclusions:

• The VIB5 database addresses the need for accessible, high-quality ab initio PES data for ML-driven molecular simulations.
• Facilitates accurate simulation of molecular rotation-vibration spectra for molecules of astrophysical significance.
• Enables the development and application of advanced ML techniques for PES construction and spectral analysis.