Autonomous Experimentation Enables Rapid Optimization of Photochemical Continuous-Flow Synthesis of Diazaxanthilidene Fluorophores

Photoconvertible diazaxanthilidene dyes enable optical labeling and tracking in live cells, but the batch photocyclization that generates them requires days of irradiation and analog-specific optimization. Here we report a closed-loop continuous-flow platform that uses Bayesian optimization to tune light intensity, temperature, flow rate, and concentration. In campaigns on the parent diazaxanthilidene and four analogs, the platform improved throughput for four of five compounds. For the parent diazaxanthilidene, photoconversion yield rose from approximately 69% in batch to 79%, while throughput increased from 0.52 to 87 nmol min-1. The platform autonomously performs >40 optimization experiments to turn a multiday, analog-specific batch synthesis into a same-day flow synthesis.