(477e) Machine Learning Predictions of Novel Ammonia Synthesis Catalysts Using Experimental and Literature Data

Ammonia is emerging as a carbon-free energy carrier for hydrogen generated via distributed and intermittent renewable electricity. In order to synthesize ammonia on-site on a distributed scale, a potential solution is to operate small-scale catalytic membrane reactors at milder conditions (300-400 °C and 1-3 MPa) using promoted ruthenium or cobalt catalysts compared to the conventional Haber-Bosch process using iron catalysts in large-scale facilities operated at harsher conditions (400-600 °C and 20-40 MPa). However, current ruthenium and cobalt catalysts suffer from hydrogen poisoning which limits ammonia synthesis activity at higher pressures and lower nitrogen conversion. Although most research is focused on developing new supports to alleviate hydrogen poisoning, recent research has shown that combinations of active metals and promoter formulations can also alleviate hydrogen poisoning.

In this work, the authors extract experimental data from the published literature for thermochemical ammonia synthesis. A machine learning model is developed to establish complex correlations that exist between catalyst formulations, elemental properties, support properties, synthesis parameters, reaction conditions, and ammonia synthesis rates. Unsupervised machine learning methods reveal the gaps in the catalyst search space that have not been explored in the literature for ammonia synthesis. A supervised machine learning model is developed to predict the ammonia synthesis activity of catalysts from literature data and extract knowledge which shows the correlations of features and activity. The activity of unknown catalyst formulations is predicted for certain search spaces, and the top predicted catalyst candidates are experimentally synthesized and tested to validate the predictions. Consequently, this work led to the discovery of new catalyst formulations that were originally unknown with higher activity compared to some state-of-the-art catalysts in the literature.