Chemistry Informed Machine Learning-Based Heat Capacity Prediction of Solid Mixed Oxides.

Area of Science:

• Materials Science
• Computational Chemistry
• Thermodynamics

Background:

• Accurate heat capacity prediction is vital for thermal energy storage and modeling temperature changes in oxide mixtures.
• Existing methods like ab initio simulations and computational thermodynamics are often computationally intensive, lack generalizability, or are inaccurate.
• Machine learning (ML) offers a promising avenue for fast, accurate, and generalizable property predictions, yet its application to mixed oxide heat capacity is limited.

Purpose of the Study:

• To develop a generalizable machine learning (ML) method for predicting the heat capacity of solid oxide pseudobinary mixtures.
• To address the limitations of current prediction techniques in terms of computational cost, accuracy, and scope.
• To establish a workflow that leverages existing data and computational tools for efficient materials property prediction.

Main Methods:

• Utilized heat capacity data from computational thermodynamics.
• Employed descriptors derived from ab initio databases.
• Trained machine learning models on these datasets to predict heat capacity.
• Validated model performance against experimental uncertainty.

Main Results:

• Achieved a mean absolute error (MAE) of 0.43 J mol^-1 K^-1 in heat capacity predictions.
• The developed ML models demonstrated an error lower than the uncertainty of differential scanning calorimetry measurements.
• The workflow proved to be generalizable for predicting properties of mixed oxides.

Conclusions:

• The proposed ML workflow provides a fast, accurate, and generalizable method for predicting the heat capacity of solid oxide pseudobinary mixtures.
• This approach surpasses the accuracy and efficiency of traditional computational methods.
• The methodology can be extended to predict other Gibbs free energy-derived properties and for higher-order oxide mixtures, broadening its applicability in materials science and energy applications.