ChemNavigator: Agentic AI Discovery of Design Rules for Organic Photocatalysts

2026-01-23

AI Summary

This study addresses the challenge of efficiently designing organic photocatalysts, a task hindered by the vast chemical space and reliance on empirical knowledge. The authors propose a multi-agent AI system that integrates large language model reasoning with density functional tight binding (DFTB) calculations to autonomously emulate scientific discovery and explore structure–property relationships. Notably, without explicit programming, the system derives chemically consistent and interpretable design rules for the first time, uncovering non-additive effects among molecular strategies. From a dataset of 200 molecules, it identifies six statistically significant design principles—such as the incorporation of ether linkages, carbonyl groups, and extended conjugation—that collectively outperform conventional machine learning approaches in predictive accuracy and mechanistic insight.

Technology Category

• Multiagent Systems: Mechanism Design
• Search and Optimization: Learning to Search
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions

Abstract

The discovery of high-performance organic photocatalysts for hydrogen evolution remains limited by the vastness of chemical space and the reliance on human intuition for molecular design. Here we present ChemNavigator, an agentic AI system that autonomously derives structure-property relationships through hypothesis-driven exploration of organic photocatalyst candidates. The system integrates large language model reasoning with density functional tight binding calculations in a multi-agent architecture that mirrors the scientific method: formulating hypotheses, designing experiments, executing calculations, and validating findings through rigorous statistical analysis. Through iterative discovery cycles encompassing 200 molecules, ChemNavigator autonomously identified six statistically significant design rules governing frontier orbital energies, including the effects of ether linkages, carbonyl groups, extended conjugation, cyano groups, halogen substituents, and amine groups. Importantly, these rules correspond to established principles of organic electronic structure (resonance donation, inductive withdrawal, $\pi$-delocalization), demonstrating that the system can independently derive chemical knowledge without explicit programming. Notably, autonomous agentic reasoning extracted these six validated rules from a molecular library where previous ML approaches identified only carbonyl effects. Furthermore, the quantified effect sizes provide a prioritized ranking for synthetic chemists, while feature interaction analysis revealed diminishing returns when combining strategies, challenging additive assumptions in molecular design. This work demonstrates that agentic AI systems can autonomously derive interpretable, chemically grounded design principles, establishing a framework for AI-assisted materials discovery that complements rather than replaces chemical intuition.

Problem

• organic photocatalysts
• design rules
• chemical space
• hydrogen evolution
• molecular design

Innovation

• agentic AI
• structure-property relationships
• hypothesis-driven exploration
• interpretable design rules
• multi-agent architecture