Machine learning and quantum-guided modeling of metal oxide thermodynamic properties

Investigating and optimizing processes at high temperatures is experimentally challenging, and first principles modeling is computationally demanding and typically too approximate.

We apply physics-based machine-learning models to the prediction of metal oxide reduction temperatures in high-temperature smelting processes and to the prediction of metal oxide melting temperatures in thermal recycling that are commonly used for the extraction of metals from their ores and from electronics waste and have a significant impact on the global energy economy and greenhouse gas emissions. Here, we introduce a hybrid approach combining zero-Kelvin first-principles calculations with both unsupervised learning and regression models. The hybrid models predict accurate reduction and melting temperatures of unseen oxides, are computationally efficient, and surpass in accuracy computationally much more demanding first-principles simulations that explicitly include temperature effects. This approach provides a general framework for learning high-temperature thermodynamic properties of metal oxides.