Quantifying Energetic and Entropic Pathways in Molecular Systems.

Area of Science:

• Physical Chemistry
• Computational Chemistry
• Machine Learning

Background:

• Activated barrier crossing events at nonzero temperatures require considering both energy and entropy.
• Effective reaction coordinates are crucial for describing metastable states and transition mechanisms.

Purpose of the Study:

• To introduce and validate the state predictive information bottleneck (SPIB) as a physics-based machine learning method.
• To identify nonlinear reaction coordinates for systems of varying complexity.
• To demonstrate SPIB's capability in uncovering entropic and energetic barriers.

Main Methods:

• Utilized the state predictive information bottleneck (SPIB), a physics-based machine learning approach.
• Applied SPIB to an analytical flat-energy double-well system.
• Applied SPIB to an analytical four-well system.
• Applied SPIB to a simulation of benzoic acid permeation through a lipid bilayer.

Main Results:

• SPIB correctly predicted an entropic bottleneck in a double-well system.
• SPIB identified entropy- and energy-dominated pathways in a four-well system.
• SPIB discovered entropic and energetic barriers for benzoic acid permeation.

Conclusions:

• SPIB is a robust method for identifying critical entropy, energy, and enthalpy barriers in physical systems.
• The identified barriers enhance the understanding and sampling of activated mechanisms.
• SPIB offers a powerful tool for studying complex chemical dynamics.