Answers are grounded in this paper and cite their sources.
Review discusses oxide transistors enabling synaptic devices for neuromorphic computing, suggesting practical integration with memory and processing.
Data‐intensive workloads demand energy‐efficient platforms that seamlessly integrate memory, processing, and spatiotemporal responses. Oxide transistors offer a compelling route toward this goal due to their low‐temperature processability, large‐area uniformity, low off‐state leakage, and defect‐mediated temporal dynamics. In this review, we discuss oxide‐transistor‐based neuromorphic devices from the perspectives of materials physics, device operation, and hardware implementation. We first outline how wide‐bandgap electronic structures, oxygen‐related defects, cation chemistry, carrier transport, and stress‐induced instabilities govern the electrical and optoelectronic responses of oxide semiconductors. We then examine synaptic devices based on charge trapping, ferroelectric polarization, and light‐induced conductance modulation, highlighting how those mechanisms enable multilevel weight storage, short‐ and long‐term plasticity, timing‐dependent learning, and artificial vision functions. Finally, we discuss the transition from device‐level synaptic emulation to practical neuromorphic hardware, emphasizing conductance linearity, retention, variability, array addressability, in‐memory computing, and in‐sensor processing. Oxide neuromorphic devices require that defects and transient dynamics be controlled as functional resources, not eliminated as reliability concerns. This perspective positions oxide transistors not merely as artificial synapse elements, but as multifunctional device platforms in which memory, sensing, and computation can be combined through controlled material dynamics.
No takes yet. Share an insight, caveat, or question.
Kim et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: