Advanced Next-generation Semiconductor Lab, SNU

Bio-inspired computing (neuromorphic computing) can perform AI tasks with ultra-low power by mimicking the human brain that consumes only 20 W. To create neuromorphic hardware, it is essential to employ artificial neurons and synapses that mimic the functionalities of biological neural networks. We develop various neuron and synaptic devices, as well as perform system-level studies including algorithms. Furthermore, we develop diverse neuromorphic sensors that have both neuromorphic and sensory functions, enabling in-sensor computing for ultra-low power Internet of Things (IoT) sensors. Quantum-inspired computing can efficiently solve combinatorial optimization problems that challenge conventional processors. Probabilistic computing uses p-bits, stochastic elements that fluctuate between 0 and 1, to efficiently explore solution spaces at room temperature, offering a practical alternative to quantum hardware. Similarly, Ising machines reformulate hard combinatorial tasks into spin models, where the collective dynamics naturally converge toward near-optimal solutions. By combining device-level innovations with system-level architectures, we create scalable, low-power platforms for quantum-inspired computing hardware. As semiconductor devices are aggressively scaled and stacked into three dimensions, ensuring their reliability throughout the operating lifetime becomes as important as their raw performance. A variety of degradation mechanisms — including the floating-body effect (FBE), self-heating effect (SHE), and dielectric/interface degradation such as bias-temperature instability and time-dependent dielectric breakdown — induce threshold-voltage shifts, increased leakage, and eventual device failure. We investigate the physical origins of these mechanisms through pulse