Physics-Guided Inverse Design of Nonfullerene Acceptors via a Deep-Learning-Accelerated Genetic Algorithm.

Area of Science:

• Materials Science
• Computational Chemistry
• Organic Electronics

Background:

• Discovering high-performing nonfullerene acceptors (NFAs) for organic solar cells (OSCs) is complex due to vast chemical possibilities and strict electronic requirements.
• Current methods often rely on empirical efficiency surrogates, limiting targeted molecular design.

Purpose of the Study:

• To present a physics-informed generative framework for inverse molecular design of NFAs.
• To enable direct optimization of quantum-relevant descriptors governing charge generation efficiency.

Main Methods:

• Integration of an evidential message-passing neural network (MPNN) with a constraint-encoded genetic algorithm (GA).
• Optimization of oscillator strength (f), exciton binding energy (Eb), and LUMO-LUMO+1 energy gap (ΔELUMO) using quantum-relevant descriptors.
• Enforcement of structural validity and chemical realism during molecular evolution.

Main Results:

• The MPNN-GA workflow efficiently explored chemical space, identifying synthetically plausible NFAs meeting multiobjective constraints.
• Predicted properties strongly correlated with quantum chemical benchmarks, validating the surrogate model's reliability.
• Pareto analyses demonstrated the framework's ability to navigate quantum-chemical trade-offs, extending the design frontier for NFAs with high f, low Eb, and suppressed ΔELUMO.

Conclusions:

• The developed framework offers a scalable and interpretable approach for physics-driven inverse design of next-generation NFAs.
• This methodology provides a generalizable strategy for accelerated molecular discovery in organic electronics.
• The study highlights the potential of AI and quantum-chemical principles in advancing organic solar cell technology.