Advancing toward a sustainable society relies on intensive use of critical metals such as indium, germanium, and antimony. This requires improving their recovery rates in metal production, finding value in waste materials and creating new recycling methods. However, there are gaps in our understanding of critical metal behavior in metallurgical processes. A promising approach to understanding the distribution behavior of these target metals is integrating systematic thermodynamic modeling with key experiments. This work aims to develop a comprehensive thermodynamic database for oxide systems, via adding critical element oxides like GeO2, In2O3, and Sb2O3 to major slag components such as PbO, ZnO, CaO, SiO₂, and FeO. Our methodology involves creating physics-based machine learning models to estimate the thermodynamic properties (\(\Delta H_{298.15K}^{0} , S_{298.15K}^{0} , C_{P}\)) of critical metal-containing oxides, which are often scarce in current literature. This model will support thermodynamic modeling of phase equilibria and diagrams based on the CALPHAD method, validated through experiments. The resulting oxide database can be linked with existing metal and gas databases, allowing for predictive modeling of current processes and exploration of new recovery methods for critical metals from waste streams, including by-products from copper, lead, and zinc production, as well as electronic waste. This integration enables reasonable calculations of how critical metals distribute across different phases (metal, slag, and gas), helping to identify optimal conditions, like composition, temperature, and oxygen levels, within the complex parameter space of multi-component systems. This work summarizes the challenges, advancements, and findings achieved so far.
