Abstract
Ising machines are attractive for efficiently solving NP-hard combinatorial optimization problems (COPs). In this work, a scalable monolithic-3D (M3D) oscillatory Ising machine (OIM) is proposed using ferroelectric field-effect transistors (FeFETs) serving as an in-memory routing switch (RS) and bi-stable resistor (biristor)-based oscillators for the first time. The M3D OIM achieves low static power consumption while offering high reconfigurability. Through careful control of FeFET routing switches, weights of the Ising model are embedded in coupled biristors. The performance is validated through simulations and experiments, in successfully solving King's graph sub-problems and the MaxCUT problem. By leveraging the intrinsic OIM features of parallel computing together with M3D integration, it is reported that the M3D OIM outperforms reported OIMs in scalability and speed. Such an approach provides new insights and significant potential for solving COPs.