Design of behavior prediction model of molybdenum disulfide magnetic tunnel junctions using deep networks

Abstract

Abstract Magnetic tunnel junctions (MTJ) are widely used in spintronics development owing to their high scalability and minimal power consumption. However, analyzing the electrical and magnetic behaviors of MTJ in real-time applications is challenging. In this study, an MTJ based on molybdenum disulfide (MoS2) is designed, and a novel deep Elman neural behavior prediction model is developed to analyze its behavior. MoS2 acts as a tunnel barrier in the proposed model, whereas iron oxide (Fe3O4) acts as a ferromagnetic electrode. The interface between Fe3O4 and MoS2 in the MTJ improves the spin polarization and tunnel magnetoresistance ratio. Herein, the performance parameters of the MTJ are used as inputs for the developed prediction model, which analyzes the magnetic and electrical properties of the MTJ using prediction parameters. The spin currents in the parallel and antiparallel configurations are also determined. The designed model is implemented using MATLAB and validated by comparing simulation and experimental results. Moreover, a maximum resistivity of 91 Ω is attained at a temperature of 300 K for the proposed model. At 120 K, under a positive bias, the proposed model achieves a TMR ratio of 0.936. Under negative bias, the maximum TMR ratio attained by the proposed model is 0.817.