Physics-Based Model for Accessible High-Throughput Heat Capacity Estimation

The ability to characterize and predict thermal properties of materials is fundamental to enable engineering design of temperature profiles in energy conversion and storage devices, especially in systems undergoing phase transformations as in batteries and some thermoelectrics. In particular, knowledge of the heat capacity is crucial to characterizing thermal evolution in working devices, as it is directly related to the enthalpy change within the material and is intimately related to thermal transport. The ability to accurately model heat capacity from accessible materials information can enable high-throughput engineering design and serve as a basis for advanced materials characterization. In this work, we describe the process for building up a physical model of heat capacity in solids, based in thermodynamics and materials physics, and specifically show how to utilize realistic approximations for the constituent contributions to heat capacity from parameters that are widely accessible (e.g. from Materials Project). This includes using a pre-trained machine-learned model for the vibrational density of states, instead of the widely-used Debye model, in the phonon contribution to heat capacity. Approximating the density of states in this way provides more accurate estimates of heat capacity over a wide temperature range, as demonstrated for 38 diverse materials. This model for single-phase materials thus provides a starting point for more sophisticated analysis that may consider composite materials and phase transformation kinetics.