Presentation Information
[18a-PA1-15]Development of a Prediction Model for Magnetization Compensation in Impurity-Doped Mn4N Thin Films
〇Soushi Akita1, Takamitsu Ishiyama1 (1.Tsukuba Univ.)
Keywords:
spintronics,materials informatics,nitrides
We focus on Mn4N-based thin films as promising candidates for next-generation spintronic materials. Identifying the specific compositions that exhibit magnetization compensation—a prerequisite for efficient domain wall motion—remains a bottleneck due to the reliance on experimental verification. To address this, we constructed a machine learning model that predicts magnetization compensation based on the fundamental physical properties of dopant elements. A Random Forest classifier evaluated via Leave-One-Out Cross-Validation (LOOCV) successfully achieved an accuracy of 0.77. Feature importance analysis highlighted the critical role of structural and electronic factors, demonstrating the efficacy of this data-driven approach in accelerating material design.