Generalizable Electrocatalyst Design Framework Combining Multi-Modal Data and Artificial Intelligence

C55-05a-270716Deep decarbonization is urgently needed across the board to solve the climate crisis. While carbon emissions associated with electricity production have reduced significantly over the past decade due to a precipitous decline in solar, wind, and battery costs, several sectors including chemicals and industrial transportation remain hard- to-abate. Specifically, by shifting chemical and fuel production away from fossil-based high-temperature processes towards the use of renewable feedstocks and electricity (i.e., “electrifying” the industry), we can substantially reduce carbon emissions from this sector. However, the electrochemical production of chemicals and fuels from renewable feedstocks is still an emerging industry, and for many processes is not economically viable. To deploy numerous electrochemical processes at the rate at which the climate crisis demands, a new approach is needed for the design, discovery, and optimization of novel electrocatalysts and systems. Here we propose a conceptually new approach for uncovering the operating configuration of active sites on the surface of catalysts based on our proprietary ultrahigh-throughput experimental characterization of catalysts, computational modeling, and multi-modal machine learning. We will invert the existing paradigm – which uses a simulation of a material’s structure to predict its function – and instead start with catalyst’s functional data and reconstruct the surface structure of a catalyst from a pool of possible unit cells. This approach eliminates the need for prohibitively slow and expensive experimental structural analysis, enabling us to rapidly discover novel catalyst materials optimized to device architecture and electrochemical environment. Ultimately, the proposed technology will result in dramatically accelerated development and deployment of novel electrochemical processes that address the climate crisis. In Phase I, we will focus on electrocatalytic activity as the main function input complemented by density functional theory modeling and optical microscopy for nanoparticles within a combinatorial 3-element design space. We will explore systems with well-understood phase diagrams and simple-pathway reactions – such as oxygen reduction and hydrogen evolution reactions utilized in fuel cells – to demonstrate the potential of this approach. Electrocatalytic activity data will be collected using Stoicheia’s proprietary high-throughput synthesis and screening technology, which will be complemented by computational modeling and artificial intelligence. In Phase I, we will demonstrate that large-scale, high-quality functional data in combination with computational modeling can be utilized to build a multi-modal artificial intelligence framework for accurately predicting structural characteristics of electrocatalysts. The development of better electrocatalysts is critical for achieving a sustainable future, and at a fundamental level, deep understanding of the ways in which molecules interact with catalyst materials will enable the efficient design of electrochemical systems for accelerated time to market. The proposed work enables novel electrocatalysts to be discovered at a significantly faster pace, higher accuracy, and drastically reduced compute power – without the burden of characterizing each material’s structure – representing a paradigm shift in the way new materials are designed and commercialized. Indeed, this foundational understanding will have significant economic impact. For instance, high-performance, earth-abundant catalysts for oxygen reduction in fuel cells, oxygen evolution in water-splitting electrolyzers, and electroreduction of CO2 will provide significant climate benefits as well as unlock hundreds of billions of dollars of value. The proposed work will empower Stoicheia to effect significant improvements broadly across electrochemistry, each of which will bring the global chemical and energy industries closer to decarbonization. Finally, the deployment of earth-abundant catalyst materials (versus scarce materials with few mining locations in the world) will help ensure the geopolitical security of the United States.