AlphaFlow: autonomous discovery and optimization of multi-step chemistry using a self-driven fluidic lab guided by reinforcement learning

Closed-loop, autonomous experimentation accelerates and improves material-efficient exploration of large reaction spaces without user intervention. This work presents AlphaFlow, a self-driven fluidic lab for autonomous discovery of complex multi-step chemistries. AlphaFlow uses reinforcement learning with a modular microdroplet reactor to perform reaction steps with variable sequence, phase separation, washing, and continuous in-situ spectral monitoring. It successfully identified and optimized a novel multi-step reaction route (up to 40 parameters) for core-shell semiconductor nanoparticle shell-growth, exceeding conventional colloidal atomic layer deposition (CALD) sequences. AlphaFlow demonstrates the capabilities of closed-loop, reinforcement learning-guided systems in exploring and solving challenges in multi-step nanoparticle syntheses, relying solely on in-house generated data from a miniaturized microfluidic platform.

Amanda A. Volk, Robert W. Epps, Daniel T. Yonemoto, Benjamin S. Masters, Felix N. Castellano, Kristofer G. Reyes, Milad Abolhasani