Autonomous high-throughput computations in catalysis

Autonomous atomistic computations are excellent tools to accelerate the development of heterogeneous (electro-)catalysts. In this perspective, we critically review achieved progress high-throughput screening aimed at identifying promising catalyst materials via databases, workflow managers, and machine-learning techniques. Outstanding challenges also discussed extensively: modification stability surfaces under realistic reaction conditions is key for meaningful predictions. Furthermore, adequately accounting solvent effects remains a topic active research particularly relevant biomass transformations electrocatalysis. Finally, efficient, autonomous workflows investigating sites amorphous catalysts remain underdeveloped. The can be supplemented with laboratories, which allow performance sophisticated experiments driven by artificial intelligence-augmented design experiments, reducing human-time investment optimizing synthesis as well characterizations. combination laboratories promise power dearly needed transition sustainable chemical industry. Challenges opportunities: • atomic-scale significantly development, but necessary (software) infrastructure not yet widely adopted or even known in community. particular, reinforcement learning adapted needs targeted reduced human cost. Accounting working challenging achievable machine learning. should complemented experimentation optimization through require large investments from facilities. speed up new discovery. surrogate models, learning, managers crucial achieving vision. Investigating one major overcome. These augmented develop active, stable, low-cost materials.