Accelerated discovery of high-performance Cu-Ni-Co-Si alloys through machine learning

Abstract Cu-Ni-Co-Si alloys have been regarded as a candidate for the next-generation integrated circuits. Nevertheless, using the trial and error method to design high-performance copper alloys requires a lot of effort and time. Thus, the material design method based on machine learning is used to accelerate the exploitation of alloys. In this study, a composition-process-property database of Cu-Ni-Co-Si alloys was established, and a new strategy that could simultaneously realize the prediction of properties and the optimization of compositions and process parameters was proposed. Four groups were chosen from 38,880 candidates by the multi-performance screening method; good agreements existed between the prediction and the test. The Cu-2.3Ni-0.7Co-0.7Si alloy had the best performance among the designed alloys, and this alloy was studied in depth. The influence of the dissolution of Co in Ni2Si was analyzed from a novel perspective. Interestingly, the trace amount of Co replacing Ni to form (Ni, Co)2Si increased the phase dissolution temperature dramatically and shortened the coarsening rate. Affected by Co, the over-aging process was slowed down, which broadened the use range of alloys greatly. Therefore, the developed Cu-2.3Ni-0.7Co-0.7Si alloy can prove to be promising materials that meet different working conditions, and its performance was better than C70350 alloy.