I. Introduction
Driven by the burgeoning needs of artificial intelligence and deep learning, the computing demand is escalating at an unprecedented pace. However, the performance boost from technology scaling is stagnant. Increased leakage currents restrict reductions in threshold voltage (), while heightened dynamic power constrains the elevation of the supply voltage (). Consequently, these factors collectively limit performance improvements. Furthermore, the threshold voltage () cannot be scaled down aggressively, complicating efforts to reduce the power-supply voltage () for power conservation. Moreover, enhancing performance by increasing the clock frequency is becoming increasingly challenging as the dynamic power consumption rises accordingly [2]. Thus, innovative approaches are needed to advance modern transistors and interconnect to fulfill the exponentially growing computing demand.