Machine Learning Guided Interrogation of Gas Evolving Electrode Catalyst Activity

Being able to effectively manage gas evolution at electrochemical electrodes is a major challenge in a wide variety of industrial applications. Perhaps the most important activities where electrochemical bubble generation is critical to control are the chlor-alkali (chlorine gas production) and Hall-Héroult (electrochemical aluminum smelting) processes. Significant effort goes towards mitigating the “anode effect” of the Hall-Héroult process, where carbon dioxide gas inactivates the carbon anode, causing catastrophic increases in overpotential. For the chlor-alkali process, the attachment of bubbles to electrode surfaces accounts for approximately 20% of the overpotential at industrially relevant current densities. Similar inefficiencies and problems are caused during water electrolysis when bubbles adhere to or flood electrodes. As a result, the limiting effects of bubbles restrict the adoption of traditional, low-temperature electrochemical methods for hydrogen generation from renewable electricity. In this way, both understanding and mitigating the negative impacts that bubbles have on electrode performance offer a unique path to enable high performance gas evolving electrochemical technologies, which will play a major industrial role in a renewably electrified future.<br/><br/>These problems motivate this work as we investigate the way that bubble inefficiencies manifest on electrodes by fabricating novel electrodes for studying the effects that adhered bubbles have for electrochemical performance using both hydrogen and oxygen evolution reactions. Despite extensive knowledge of the negative impacts adhered gas bubbles cause at a system level, there is limited fundamental understanding of the extent to which specific mechanisms inactivate active electrode materials. By using surface engineering gas evolving electrode for testing, we can manipulate the micro- and nanoscale electrode morphologies to control bubbles nucleation and bubble/electrode interactions. By developing machine vision algorithms for the detection and tracking of bubbles from electrode surface imaging, we obtain empirical results which oppose the prevailing view held in regard to how bubbles inactivate electrodes. The utility of tracking nucleating and growing bubbles on electrodes’ surfaces can also provide important spatiotemporal information about the activity of the electrocatalyst materials used, not only regarding the inactivation that the bubbles cause. As a result, the machine vision methods developed are also proposed and demonstrated as a useful electrocatalyst screening tool for the evaluation of catalyst materials for gas evolving electrochemical reactions. As a result of these findings and the methods developed, we provide a novel framework for designing gas-evolving electrochemical electrodes and for the discovery of highly active electrocatalyst materials for gas evolving reactions.