[Abstract]

The development of new active, stable and cost-effective alternative catalysts has been of utmost focus in renewable and sustainable future energy technologies such as hydrogen peroxide production, ammonia production, fuel cells and electrolyzers. Unfortunately, the current bruteforce approach is nearly impossible to investigate all possible combinations in broad chemical space. To overcome the current scientific challenges, atomistic simulations have demonstrated the potential to accelerate a discovery of new catalysts. For the past decades, a prediction of catalytic properties from the atomistic simulations has become reasonably accurate, and a descriptor-based screening of catalysts has become more feasible due to the exponential increase in computing power. More excitingly, recent advances in machine-learning have opened up the possibility of high-throughput catalyst screening with the minimal number of expensive density functional theory (DFT) calculations.

In this talk, I will discuss how atomistic simulations and machine learning could be of significant help to discover new catalysts. I will present my recent results on automated DFT calculations to predict catalytic activities of oxide materials, and convolutional neural network model to reduce the number of DFT jobs for the future high-throughput catalyst screening.