Machine learning to accelerate screening for Marcus reorganization energies

Area of Science:

• Organic electronics
• Materials science
• Computational chemistry

Background:

• Predicting charge transport in π-conjugated materials is crucial for organic electronics.
• Marcus reorganization energy is a key factor in charge transfer efficiency.
• Current density functional methods are computationally expensive.

Purpose of the Study:

• To develop active machine learning (ML) methods for predicting intramolecular reorganization energies.
• To screen polythiophene compounds with low internal reorganization energies for efficient charge transport.

Main Methods:

• Utilized active machine learning algorithms.
• Trained models on a diverse set of polythiophene compounds.
• Validated model performance on a dedicated screening set.

Main Results:

• Achieved an overall root mean square error (RMSE) of ±0.113 eV.
• Demonstrated a significantly lower RMSE of ±0.036 eV on the screening set.
• Identified that higher errors were associated with high-reorganization energy compounds.

Conclusions:

• Active ML methods provide a computationally efficient approach to predict reorganization energies.
• The developed ML models are effective in screening for organic materials with desirable charge transport properties.
• This work facilitates the design of new organic electronic devices with improved performance.