Finding heterogeneous nucleating agents for ice using a data-driven approach

Abstract

We present a high-throughput data-driven workflow to identify potential heterogeneous nucleating agents from structural databases for phase change materials, such as ice. Our model evaluates the fit between ice I_h and nucleator docked slabs, considering Miller index planes up to (333), thus addressing some of the structural complexities in nucleation by examining crystal morphology features. Bulk water immersion experiments on a set of ten known nucleators set a delineating temperature to distinguish between good and poor nucleation behaviour, which helped derive numerical tolerance limits to allow reliable differentiation on the basis of the number of predicted matching interface models. We then used our algorithm to screen 3500 simple metal oxides and halides taken from the inorganic chemistry structural database (ICSD), and show that just 7% of the former and 3% of the latter were predicted to nucleate ice on the basis of geometric slab matching alone. Subsequent experimental testing of 22 compounds suggested a 64% correct prediction rate, and identified four new ice nucleators (CeO₂, WO₃, Bi₂O₃, Ti₂O₃). Inspired by the ice-nucleating efficiency of copper oxides, we also tested copper tubing with local tap water, and observed sub-cooling suppression, most likely due to copper oxide buildup. Although based on a simple geometric interface matching model, this approach offers an efficient route as a first stage high throughput screen for potential heterogeneous nucleating agents.