Comparative analysis of two-terminal memristors, three-terminal transistors, and their integration in neuromorphic computing

The convergence of artificial intelligence and neuromorphic hardware has accelerated the development of synaptic devices capable of brain-like computation. This review presents a comparative analysis of two-terminal memristors and three-terminal synaptic transistors, with emphasis on their material innovations, plasticity emulation, energy efficiency, and integration strategies. While memristors provide dense, low-power, nonvolatile memory suitable for in-memory computing, three-terminal transistors offer greater precision in synaptic weight modulation and decoupled read/write operations. This review further explores hybrid memtransistor platforms and multi-terminal architectures that replicate complex neuronal behaviors such as heterosynaptic and anti-Hebbian plasticity, paving the way for multimodal and autonomous learning. In parallel, heterogeneous and three-dimensional (3D) monolithic integration strategies enable the development of ultra-compact, CMOS-compatible neuromorphic systems. By benchmarking key performance metrics and highlighting recent breakthroughs, this review provides critical insights and design guidelines to support future innovations in energy-efficient, scalable, and intelligent neuromorphic computing.