Accelerated discovery of lead-free solder alloys with enhanced creep resistance via complementary machine learning strategy

Minor alloying is an effective method to improve the performance of lead-free solder alloys. In this study, we propose a complementary Machine Learning (ML) strategy for minor alloying to design solder alloys with enhanced creep resistance. Two ML models, leveraging compositional and knowledge-aware features, respectively, were constructed to predict the creep stress exponent of Sn-Ag-Cu (SAC)-based solder alloys. Five new solder alloys were designed and experimentally evaluated by screening the virtual sample space consisting of the critical elements, including Bi, In, and Ni, affecting the creep resistance of SAC solder alloys with the ML model. The creep resistance of the designed alloys was characterized using nanoindentation tests. Notably, the creep stress exponent of SAC387-3 wt % Bi-0.4 wt% Ni was determined to be 12.82 at room temperature, indicating a significant 33.9% decrease in creep strain rate compared to the SAC387 alloy. This study demonstrates the potential of the ML approach to design solder alloys with enhanced creep resistance.