Machine Learning Assisted Statistical Variation Analysis of Voltage Gated Spin Orbit Torque Magnetic Tunnel Junction

In the semiconductor industry, spintronic technology has gained significant attention in the new era due to its compatibility with CMOS design. Among the promising spintronic devices, the voltage-gated spin-orbit torque magnetic tunnel junction (VGSOT-MTJ) stands out, offering the potential to overcome limitations found in other spintronic devices like spin-transfer torque (STT) and spin-orbit torque (SOT) magnetic tunnel junctions. However, the performance and reliability of VGSOT-MTJ can be influenced by variations in critical device parameters. This work investigates the variability of device characteristics in VGSOT-MTJ and proposes an innovative framework assisted by machine learning (ML) to streamline the technological pathfinding process. The proposed framework leverages three regression models based on machine learning techniques: K-nearest neighbors (KNN), random forest, and neural network regressors. Each model's effectiveness is assessed by evaluating metrics such as mean error, R2 score, root mean square error (RMSE), and inference time. It is observed that the random forest regressor outperforms neural network regressor and K-nearest regressor model in terms of mean error, R2 score, and RMSE. Our ML assisted approach provided a more accurate and efficient analysis, with R2 scores of 0.99. Furthermore, significant improvement in prediction time is observed as compared to the SPICE simulation time.