Machine learning photodynamics reveal intersystem-crossing-driven ladderdiene ring opening

Area of Science:

• Photochemistry
• Computational Chemistry
• Organic Synthesis

Background:

• Photochemical ring-opening reactions are vital for mild and atom-economical chemical synthesis.
• Developing efficient photochemical reactions requires understanding excited-state dynamics.

Purpose of the Study:

• To propose and investigate a near-visible light-induced electrocyclic ring-opening reaction of carbonyl-functionalized tricyclooctadiene.
• To utilize machine learning-accelerated photodynamics simulations for predicting reaction efficiency.

Main Methods:

• Density Functional Theory (DFT) calculations (CAM-B3LYP/cc-pVDZ) to determine excited state energies.
• Multiconfigurational CASSCF calculations to analyze energy gaps.
• Machine learning-accelerated photodynamics simulations using 1000 FSSH trajectories.

Main Results:

• Carbonyl group significantly lowers the S1 energy of the substrate to 3.65 eV (340 nm), enabling near-visible light absorption.
• Small S1 and T1 energy gaps were observed near the S1-minimum.
• Simulations revealed a stepwise ring-opening mechanism from S1, with rapid S1/T1 intersystem crossings within 20 ps.

Conclusions:

• The proposed reaction is highly efficient, with a predicted quantum yield of 89% for carbonyl-functionalized cyclooctatetraene (COT).
• This study demonstrates the predictive power of machine learning-photodynamics for designing photochemical reactions.