Hydrogen Evolution Prediction for Alternating Conjugated Copolymers Enabled by Machine Learning with Multidimension...

Area of Science:

• Materials Science
• Photocatalysis
• Computational Chemistry

Background:

• Alternating conjugated copolymers are promising for hydrogen evolution reactions.
• Developing efficient and predictive models for these materials is crucial.

Purpose of the Study:

• To develop novel data-driven strategies for studying alternating conjugated copolymers.
• To create machine learning models for predicting photocatalytic performance.
• To explore the underlying mechanisms of photocatalysis in these polymers.

Main Methods:

• Development of two multidimension fragmentation descriptors (MDFD): structure-based (SMDFD) and electronic property-based (EPMDFD).
• Application of machine learning (ML) algorithms for property prediction and reaction rate assessment.
• Integration of explainable ML with first-principles calculations to analyze photocatalytic dynamics.

Main Results:

• SMDFD models demonstrated high accuracy in predicting electronic properties.
• EPMDFD-based ML models achieved excellent accuracy (0.91 real-test accuracy) for hydrogen evolution reaction prediction without experimental parameters.
• Analysis revealed the critical role of electron delocalization in excited states for photocatalysis.

Conclusions:

• Novel MDFDs and ML models offer a powerful data-driven approach for designing polymeric photocatalysts.
• The developed methods enable accurate prediction and virtual design of high-performance materials for hydrogen evolution.
• This work highlights the significant potential of ML in advancing photocatalysis research.