Rare earth oxide-based memristors for neuromorphic applications

Abstract The escalating demands from artificial intelligence, the Internet of Things, and wearable electronics are accelerating the rapid exploration of hardware that can deliver low-power, low-latency, and highly integrated information processing that transcends the limitations of conventional von Neumann architecture. To achieve this objective, memristor-based neuromorphic systems are significantly promising to enable the physical convergence of memory and computation. In particular, rare earth-based oxides (REOs) are particularly distinguished by their unique combination of multivalent states, versatile redox behavior, and highly tunable defect kinetics advanced by their specific 4f electronic configuration. These distinctive characteristics render REOs exceptionally promising for regulating resistive switching, stabilizing analog conductance modulation, and realizing synaptic plasticity with enhanced functional versatility. In this work, we review the remarkable progress of REOs-based memristors covering material design, switching mechanisms, synaptic emulation, array integration, and chip-oriented neuromorphic architectures. Further, we also systematically elaborate on their emerging applications in artificial intelligence hardware, neuroscience interfaces, and intelligent sensing systems. Finally, the key bottlenecks of REOs-based memristors constraining large-scale implementation are critically assessed in terms of variability, analog precision, endurance, and cross-level co-optimization. Therefore, REO memristors serve as a highly promising material system to bridge device-level innovation and next-generation brain-inspired hardware.