Learning Intermolecular Electronic Coupling with Molecular-Orbital-Based Descriptors.

Area of Science:

• Computational chemistry
• Quantum chemistry
• Molecular modeling

Background:

• Electronic coupling governs intermolecular electron transfer (ET) and energy transfer (EET) rates.
• Excited-state couplings are state-specific, arising from molecular orbital (MO) interactions.

Purpose of the Study:

• To develop a graphic MO-based descriptor for predicting intermolecular electronic couplings.
• To accurately model state-specific couplings in excited molecular systems.

Main Methods:

• Transforming MOs into feature vectors capturing quantum and spatial information.
• Developing MO-pair descriptors by multiplying paired MO vectors.
• Utilizing a deep neural network (DNN) model to learn couplings.

Main Results:

• High accuracy achieved in learning hole transfer (HT), electron transfer (ET), and Dexter energy transfer (DET) couplings for naphthalene dimers.
• The MO-pair descriptor effectively characterizes MO interactions.
• Successful prediction with a small training dataset compared to quantum chemical (QC) calculations.

Conclusions:

• The proposed MO-pair descriptor accurately predicts intermolecular electronic couplings.
• This descriptor shows potential for modeling other state-specific properties in excited molecular systems.
• The strategy offers a high-performance approach for electronic coupling prediction.