A Machine Learning Model for Vickers Hardness of Sn-In-X Low-Temperature Solder

Low-temperature solders are increasingly essential for flexible electronic devices. Tin-indium (Sn-In) alloys have emerged as promising candidates; however, concerns remain regarding their low strength, resulting in a poor fatigue performance. To enhance its mechanical strength, third element (e.g., Cu, Ag, Zn) addition to Sn-In alloys is typically introduced. Nevertheless, optimizing alloy composition via experimental approach is time-consuming with high costs. In this study, we developed a Vickers hardness database for Sn-In-based alloy and employed a machine learning method to construct a hardness model for predicting new alloy recipe. The results revealed Cu addition to Sn-In alloy would significant improvement its hardness due to the presence of inter metallic compounds (IMCs). The machine learning model informed with the alloy composition showed RMSE of 1.61 and R-squared of 0.989. The five-fold CV results showed RMSE of 6.22 and R-squared of 0.840. The predicted new composition and their hardness values were found to be highly consistent with the actual measurements. The result shows its potential as a powerful tool for exploring alloy properties in complex systems.