Accelerating the development of multi-component Cu-Al-based shape memory alloys with high elastocaloric property by machine learning

Exploring elastocaloric materials with high transformation entropy change (ΔS) is a key mission for the development of elastocaloric refrigeration technology. Here, we show an adaptive design strategy, tightly coupled a machine learning (ML) with theoretical calculations to accelerate the discovery process of multi-component Cu-Al-based shape memory alloys (SMAs) with high ΔS. Based on a linear regression model, Al, Co, Fe, Ni are the elements that are beneficial to the significant promotion of ΔS in the Cu-Al-based alloys. In our results, Cu72.2Al20.2Ni6.2Co0.7B0.7 is discovered with the highest ΔS of 1.88 J/mol K from a potential space of ~500,000 compositions, which is higher than the highest ones found in ternary Cu-Al-Mn and reported experimental value by 9.9% and 17.5%.