Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO₂ Reduction Performance Using Machine...

Area of Science:

• Photocatalysis
• Renewable Energy
• Circular Economy

Background:

• Sunlight-driven CO2 reduction is key for a circular economy.
• Traditional optimization methods struggle with multivariable photocatalytic systems, optimizing one metric at the expense of others.
• This limits overall system performance and hinders progress.

Purpose of the Study:

• To develop a holistic performance metric for multivariable photocatalytic systems.
• To employ machine learning to efficiently navigate complex parameter spaces for optimization.
• To make holistic optimization accessible for experimentalists in catalysis.

Main Methods:

• Defined a novel metric for holistic system performance considering multiple figures of merit.
• Utilized a machine learning algorithm to guide experimental optimization.
• Employed a five-component system forming photocatalytic micelles for CO2-to-CO reduction.

Main Results:

• Achieved simultaneous optimization of yield, quantum yield, turnover number, and frequency with high selectivity.
• Machine learning quantified parameter effects, revealing buffer concentration as the dominant factor (4x more important than catalyst concentration).
• Demonstrated efficient, holistic optimization of a complex photocatalytic system.

Conclusions:

• The developed methodology enables holistic optimization of photocatalytic systems, overcoming limitations of traditional approaches.
• Standardization of this approach will accelerate progress in catalysis by providing deeper insights and enhancing comparability.
• This work moves beyond subjective figure-of-merit comparisons towards objective, comprehensive performance evaluation.