Solvation strategies for free-energy calculations in a halogen-bonded complex: implicit, explicit, and machine...

Area of Science:

• Computational chemistry
• Supramolecular chemistry
• Physical chemistry

Background:

• Halogen bonding is crucial in various chemical systems.
• Accurate solvation models are needed for predicting binding energies.
• Molecular iodine and tetramethylthiourea form a relevant model complex.

Purpose of the Study:

• To evaluate different solvent models for halogen-bonded complexes.
• To identify efficient and accurate computational approaches for solvation.
• To reproduce experimental trends in binding free energies.

Main Methods:

• Implicit and explicit solvent models were compared.
• Micro-solvation and ONIOM methods were explored.
• Periodic metadynamics simulations were performed.
• Machine learning perturbation theory was applied.

Main Results:

• Implicit models are accurate in low/moderate polarity solvents but fail in high-polarity ones.
• Explicit solvent representations are necessary for polar media.
• Periodic metadynamics offers accuracy but is computationally intensive and limited to GGA functionals.
• Machine learning perturbation theory improved accuracy beyond GGA.

Conclusions:

• Accurate modeling of halogen bonding in polar solvents requires explicit solvation.
• Machine learning perturbation theory provides a computationally feasible route to high-accuracy free energy calculations.
• This work advances the understanding of solvation effects in non-covalent interactions.