A Modular and Affordable Self-Driving Laboratory for Vision-Guided Optimization of Metal Electrodeposition

Metal electrodeposition is a widely used materials synthesis technique; however, industrial applications often require the optimization of complex precursor formulations and electrodeposition parameters, which is typically performed in a slow, inefficient, and empirical manner. Self-driving laboratories (SDLs) that combine automated electrodeposition with machine-learning-guided decision-making could accelerate discovery in these complex, high-dimensional parameter spaces, but they face practical challenges. Notably, automated electrodeposition platforms need a continuous supply of pristine electrode surfaces, preservation and logging of electrodeposited samples, and the ability to easily incorporate nonelectrochemical downstream characterization. To overcome these challenges, we developed a modular, affordable, and accessible automated electrochemistry platform that uses a roll-to-roll design to enable continuous electrodeposition experiments. The roll-to-roll architecture enables easy integration with downstream characterization techniques in a conveyor belt fashion, and electrodeposited samples are stored by rolling up the spent electrode material. We performed 300 identical Cu electrodeposition experiments over 35 h, demonstrating high repeatability and continuous 24/7 experimentation. We demonstrate the use of our platform for an autonomous campaign, employing a vision-guided Bayesian optimization (BO) workflow to tune electrodeposition parameters and additive concentrations to achieve a visible-light-absorbing Cu deposit. An interpretable machine-learning approach was used to identify benzotriazole as a key additive for controlling the brightness of deposited Cu films. The roll-to-roll design enabled the preservation of samples and offline analysis with scanning electron microscopy, providing insight into the impact of benzotriazole on the morphology of the deposited films.