Ionic Landscape Engineering via Perovskite Quantum Dots for Reliable and Energy-Efficient Perovskite Memristors

Abstract

The rapid growth of data-intensive artificial intelligence workloads has exposed data movement in conventional von Neumann architectures as a critical bottleneck to both enhanced energy efficiency and reduced latency. Among the materials investigated for resistive random-access memory, halide perovskites have garnered extensive attention owing to their tunable electronic properties and low-power operation, making them suitable for high-density memory applications. Although all-inorganic CsPbI₃ offers high thermal stability, its reliability is compromised by mobile iodide vacancies causing stochastic switching. To address this limitation, this paper introduces a halide exchange-driven interface engineering strategy using CsPbBr₃ quantum dots (QDs). Unlike conventional passivation, this approach enables spontaneous Br^– diffusion into the CsPbI₃ layer, passivating interfacial defects and promoting structural reorganization at the interface. The optimized device exhibits highly uniform switching and a considerable reduction in SET power consumption from 37.57 to 2.52 μW. Object-detection simulations demonstrate that while the control device suffers an accuracy loss of 17.9%, the QD-incorporated device maintains robust performance with only a 2.8% loss in accuracy over 2,000 cycles. These results establish QD-driven defect engineering as a robust pathway for developing reliable components for future neuromorphic systems.