Assessing the persistence of chalcogen bonds in solution with neural network potentials

Area of Science:

• Computational chemistry
• Supramolecular chemistry
• Materials science

Background:

• Non-covalent interactions are key in catalysis and materials design.
• Computational models often overlook temperature and solvent effects, altering molecular properties.
• Neural network potentials (NNPs) offer accurate simulations in condensed phases.

Purpose of the Study:

• To investigate chalcogen bond persistence and strength in condensed phase.
• To analyze competing intermolecular interactions, especially with solvents.
• To evaluate NNPs for modeling non-covalent interactions in solution.

Main Methods:

• Trained direct and baselined NNPs using density functional theory (DFT) data.
• Performed simulations of a solute-Cl^--THF mixture in explicit solvent.
• Compared NNP results with other potentials (AMOEBA, continuum solvent models).

Main Results:

• NNPs accurately reproduced DFT energies and forces.
• Explicit solvent simulations revealed competition between chalcogen bonds and solvent interactions.
• Chalcogen bonds exhibit short-range directionality impacting solution properties.

Conclusions:

• Baselined NNPs provide a reliable method for studying non-covalent interactions in solution.
• Solvent interactions significantly compete with and modify chalcogen bonds.
• Accurate modeling of condensed-phase non-covalent interactions is essential for understanding molecular behavior.