High-TMR SHNO-Based Ising Machines for CMOS-Compatible Unconventional Computing
Tzu-Yen Hung, Jui-Yu Hsieh, Wei-Bang Liao, Chun‐Yi Lin, Chiao‐Yun Lo, Shih-Ching Chiu, Yu-Chen Hsin, Jeng−Hua Wei, Chi‐Feng Pai · IEEE Electron Device Letters · 2025
Ising machines are promising unconventional computing solvers, leveraging quantum spin systems to find optimal solutions. In this research, Ta/CoFeB magnetic tunnel junctions (MTJs) are utilized as spin Hall nano-oscillators (SHNOs) to construct Ising machines. A type-Y field-free Ta-based SHNO with high tunneling magnetoresistance (TMR) ratio (174.9%) and tunable resonance frequencies (1.9–2.5 GHz) via spin-orbit torque (SOT) is developed, enabling CMOS-compatible Ising circuit design. Second harmonic injection locking (SHIL) is modeled, showcasing SHNOs as strong candidates for Ising machines. This approach solves the 16-node max-cut problem with 100% accuracy and produces reliable solutions for larger node sets. The system solves the Burma-14 cities Traveling Salesman Problem within 2 μs, outperforming traditional GPU solvers by several orders of magnitude. This research paves the way for implementing fast and energy-efficient Ising machine solvers in future edge-computing and IoT systems.