AI speeds discovery of materials for cleaner industrial chemical production
Researchers used machine learning to identify new oxygen-carrier materials that could make chemical looping—a process for producing chemicals and separating oxygen—more efficient and cost-effective. The approach cuts discovery time dramatically, potentially accelerating deployment of greener alternatives to conventional industrial processes.
Originaltitel: Discovery of high-entropy perovskite oxygen carriers for chemical looping applications via an autonomous active learning protocol
The discovery and design of new materials are paramount for advancing green technologies. High-entropy oxides represent one such group that has been only tentatively explored, mainly due to the inherent problem of navigating vast compositional spaces. Here, oxygen carriers for chemical looping processes have been identified using active learning-based strategies and first-principles-informed calculations. The proposed approaches were validated using an established computational framework for identifying high-entropy perovskites suitable for chemical looping air separation and dry reforming. The central insight gained was the identification of effective strategies, including greedy and Thompson-based sampling, informed by uncertainty estimates from Gaussian processes. Building on this knowledge, the concept was applied to the challenge of discovering high-entropy oxygen carriers for chemical-looping oxygen uncoupling. This resulted in both qualitative and quantitative outcomes, including lists of materials with high oxygen transfer capacities and configurational entropies. The top candidates were based on the known oxygen carrier CaMnO 3 and included expected elements such as titanium, cobalt, and copper, as well as unexpected ones such as yttrium and samarium. The results suggest that adopting active learning approaches is critical for materials discovery, as these methods are already reshaping research practice and will soon become the norm. • Demonstration of an active-learning-based strategy for material discovery based on first-principles-informed calculations. • The approach is proven to be efficient for finding oxygen carriers for chemical looping applications. • The process is validated and thoroughly tested for dry reforming and chemical looping air separation, while also yielding a wide range of potential candidates. • More detailed calculations lead to the identification of a multitude of high-entropy oxygen carriers for chemical oxygen uncoupling.