CALPHAD-based thermodynamic modeling is now routinely used to predict and understand the observed properties of multi-component alloy systems. While these thermodynamic predictions are often in qualitatively good agreement with experiments, we show that there are significant limitations in the commercially available thermodynamic databases. These limitations are due to the cumbersome task of maintaining complex multi-component databases.
This work describes new software architecture, with a particular focus on uncertainty quantification, for improving these databases. This approach requires little input from the user, making it possible to be integrated into automated, high-throughput modeling infrastructure. With a more robust and efficient method for generating multi-component thermodynamic databases based on the latest available data, thermodynamic studies of non-conventional materials systems such as bulk metallic glasses, high-entropy alloys, and compositionally-graded alloys will become more predictive and accurate.