Size dependence of the properties of synthetic-antiferromagnet-based stochastic magnetic tunnel junctions for probabilistic computing

Stochastic magnetic tunnel junctions (s-MTJs) are core components for spintronics-based probabilistic computing, offering promising routes toward energy-efficient unconventional computing. A synthetic antiferromagnetic (SAF) free layer configuration has been proposed to enhance the robustness of these devices against external magnetic field disturbance. Here, we systematically investigate the stochastic properties of SAF-based s-MTJs, with a particular focus on their size dependence. We demonstrate that decreasing the junction size from approximately 100 nm to around 40 nm reduces the relaxation time. Importantly, smaller junctions show markedly enhanced robustness against external magnetic fields, achieving susceptibility values below 0.1 mT−1 for sizes under 80 nm. Furthermore, we observe that smaller junctions exhibit enhanced insensitivity to bias voltage, with sensitivity values of around 0.5 V−1 at device sizes of approximately 40 nm, which is 100 times lower than that of conventional s-MTJs. These quantitative results provide clear guidelines for designing robust and scalable s-MTJ devices, essential for advancing spintronic probabilistic computing toward practical implementation.