Exploring novel Ni-Mn-X based magnetocaloric materials via machine learning with physical descriptors

Hongxing Li, Xiaohua Tian, Wenbin Zhao, Jie Yang, Xiaochuan Wang, Kun Zhang, Jian Li, Changlong Tan,
Exploring novel Ni-Mn-X based magnetocaloric materials via machine learning with physical descriptors,
Applied Materials Today,
Volume 47,
2025,
102923,
ISSN 2352-9407,
https://doi.org/10.1016/j.apmt.2025.102923.
(https://www.sciencedirect.com/science/article/pii/S2352940725003415)
Abstract: Exploring novel Ni-Mn-X (X= In, Sn, Sb) magnetic shape memory alloys exhibiting large magnetic entropy change is crucial for magnetocaloric refrigeration. However, current research methods relying on experimental trial and error are inefficient. Although machine learning has demonstrated considerable potential in material design, its application in the Ni-Mn-X system remains challenging due to the scarcity of datasets and the complexity of phase transformation behaviors. In this study, we present a machine learning approach utilizing physical descriptors to discover novel Ni-Mn-X based magnetic shape memory alloys. The incorporation of physical descriptor features, specifically the magnetic moment difference (∆m), energy difference (∆E), and volume difference (∆V), significantly enhances the performance of the machine learning model. This strategy, which leverages physical descriptors to aid machine learning, enables the magnetic entropy change model to achieve high fitting accuracy and excellent predictive capabilities. The coefficient of determination (R2) reaches 0.988 for the training set and 0.951 for the test set. A novel composition (Ni49Mn35In15Fe0.5Ga0.5) has been designed and synthesized by machine learning prediction, which exhibits a high magnetic entropy change of 27.16 J kg⁻¹ K⁻¹ at 302 K under a 7 T magnetic field change. This work holds promise for accelerating the discovery of novel Ni-Mn-X-based magnetic shape memory alloys with substantial magnetic entropy changes. In addition, the method of incorporating physical descriptors can serve as a reference for addressing other challenging machine learning tasks.
Keywords: Magnetic shape memory alloys; Magnetic entropy change; Machine learning; Physical descriptors