Collective Variables for Free Energy Surface Tailoring: Understanding and Modifying Functionality in Systems...

Area of Science:

• Computational Chemistry
• Statistical Mechanics
• Machine Learning

Background:

• Systems with rare events are challenging to study due to infrequent occurrences.
• Understanding and controlling system functionality often requires modifying underlying free energy landscapes.

Purpose of the Study:

• To develop a method for elucidating and modifying the functionality of systems dominated by rare events.
• To construct collective variables (CVs) that capture essential information about rare events.
• To enable semiautomated tuning of free energy surfaces.

Main Methods:

• Harmonic Linear Discriminant Analysis (HLDA), a machine learning technique, is used to identify relevant CVs.
• HLDA is trained on data from short simulations in metastable states.
• Identified CVs are used to pinpoint critical interaction potentials for modification.

Main Results:

• The method successfully identifies collective variables that encode rare event dynamics.
• Critical interaction potentials were identified and modified to tailor free energy surfaces.
• Demonstrated tractability in modifying thermodynamic and kinetic properties across three diverse systems.

Conclusions:

• The proposed approach offers a powerful tool for understanding and controlling systems with rare events.
• Semiautomated tuning of free energy surfaces can be achieved with minimal prior knowledge.
• This method facilitates the targeted alteration of system functionality by modifying key interactions.