Machine Learning Framework for Modeling Exciton Polaritons in Molecular Materials.

Area of Science:

• Quantum Chemistry
• Materials Science
• Computational Chemistry

Background:

• Polaritons, light-matter quasiparticles, emerge from strong coupling between molecules and optical cavities.
• Polariton chemistry demonstrates the ability to manipulate chemical reactions through collective molecular effects within cavities.
• Simulating ensembles of excited molecules coupled to cavity modes presents significant theoretical and computational challenges.

Purpose of the Study:

• To develop a general computational protocol for predicting excited-state properties of molecules in strong coupling regimes.
• To apply machine learning techniques for modeling the complex interactions in polariton chemistry.
• To enable accurate simulations of collective light-matter interactions for understanding polariton chemistry.

Main Methods:

• Utilized a hierarchically interacting particle neural network to predict excited-state properties.
• Developed machine learning models capable of handling complex chemical systems.
• Applied ML predictions to compute potential energy surfaces and electronic spectra.

Main Results:

• Successfully predicted excited-state properties including energies, transition dipoles, and nonadiabatic coupling vectors.
• Computed potential energy surfaces and electronic spectra for azomethane in a collective coupling scenario.
• Demonstrated the efficacy of ML in tackling computationally demanding simulations.

Conclusions:

• The developed ML protocol provides a robust framework for simulating excited-state polariton chemistry.
• This work facilitates a deeper understanding of collective light-matter interactions in chemical reactions.
• The computational tools pave the way for exploring diverse molecular systems in polariton chemistry.