Three Grand Challenges for the Computational Design of Heterogeneous Catalysts

Over the past decades, first-principles-based modeling has transformed heterogeneous catalysis. Initially, computational modeling has provided insights into the nature of the active sites and catalytic mechanisms, and soon the potential to predict materials with desirable catalytic properties was recognized. As of today, the computational design of heterogeneous catalysts is a mature field. At the same time continuous developments in computational methodology and increased availability of computational resources indicate a significant potential for growth in the future. In this Perspective, I outline three potential directions for the development of the computational design of heterogeneous catalysts that directly address the shortcomings of today's catalyst design paradigms. These three grand challenges are (i) machine learning and AI for the automated discovery of heterogeneous catalysts, (ii) the selective modification of potential energy surfaces, and (iii) the holistic design of heterogeneous catalysts. The motivation for progress in each area is given; before the challenges, the current state of the art and a long-term goal to aspire to are discussed. If these long-term goals are achieved, the three grand challenges described in this contribution have the potential to transform materials discovery for heterogeneous catalysis and will lead to a shift from an experimentally based process today to a computationally driven future.