Search for: All records

Many thermodynamic calculations and engineering applications require the temperature-dependent heat capacity (Cp) of a material to be known a priori. First-principle calculations of heat capacities can stand in place of experimental information, but these calculations are costly and expensive. Here, we report on our creation of a high-throughput supervised machine learning-based tool to predict temperature-dependent heat capacity. We demonstrate that material heat capacity can be correlated to a number of elemental and atomic properties. The machine learning method predicts heat capacity for thousands of compounds in seconds, suggesting facile implementation into integrated computational materials engineering (ICME) processes. In this context, we consider its use to replace Neumann-Kopp predictions as a high-throughput screening tool to help identify new materials as candidates for engineering processes. Also promising is the enhanced speed and performance compared to cation/anion contribution methods at elevated temperatures as well as the ability to improve future predictions as more data are made available. This machine learning method only requires formula inputs when calculating heat capacity and can be completely automated. This is an improvement to common best-practice methods such as cation/anion contributions or mixed-oxide approaches which are limited in application to specific materials and require case-by-case considerations.

Prompted by limited available data, we explore data-aggregation strategies for material datasets, aiming to boost machine learning performance. Our findings suggest that intuitive aggregation schemes are ineffective in enhancing predictive accuracy.

Advancements in materials discovery tend to rely disproportionately on happenstance and luck rather than employing a systematic approach. Recently, advances in computational power have allowed researchers to build computer models to predict the material properties of any chemical formula. From energy minimization techniques to machine learning-based models, these algorithms have unique strengths and weaknesses. However, a computational model is only as good as its accuracy when compared to real-world measurements. In this work, we take two recommendations from a thermoelectric machine learning model, TaVO[Formula: see text] and GdTaO[Formula: see text], and measure their thermoelectric properties of Seebeck coefficient, thermal conductivity, and electrical conductivity. We see that the predictions are mixed; thermal conductivities are correctly predicted, while electrical conductivities and Seebeck coefficients are not. Furthermore, we explore TaVO[Formula: see text]’s unusually low thermal conductivity of 1.2 Wm[Formula: see text]K[Formula: see text], and we discover a possible new avenue of research of a low thermal conductivity oxide family.

Diffusion Models outperform Generative Adversarial Networks (GANs) and Wasserstein GANs in material discovery.

Low-cost self-driving labs (SDLs) offer faster prototyping, low-risk hands-on experience, and a test bed for sophisticated experimental planning software which helps us develop state-of-the-art SDLs.