Bifacially Engineered Perovskite-Based Synaptic Memristors Achieve High Linearity and Symmetricity for Accurate and Robust Neuromorphic Computing.

Achieving both high linearity and symmetricity in metal halide perovskite (MHP)-based memristors remains challenging, primarily due to their abrupt switching behaviors and irregular conductive filament (CF) pathways. Here, bifacially engineered MHP memristors exhibiting simultaneous high linearity, symmetricity, and reliability are reported. Top-surface passivation using phenylethylammonium iodide (PEAI) facilitates the formation of an ultrathin 2D perovskite layer (PEAPbI), promoting gradual switching and effectively suppressing ion migration during CF formation, thereby significantly enhancing the linearity of long-term potentiation. Meanwhile, bottom-side PEAI treatment alleviates tensile strain and enhances perovskite grain uniformity, leading to stable CF rupture and improved linearity in long-term depression as well as symmetricity. The resulting bifacially engineered memristor device achieves an exceptionally high I/I ratio of 3.67 × 10, remarkable endurance exceeding 11 000 cycles, and robust data retention time over 10 s. Moreover, these bifacially engineered synaptic memristors demonstrate superior classification accuracies of 92.60% and 94.53% in Canadian Institute for Advanced Research 10 (CIFAR-10) and Modified National Institute of Standards and Technology (MNIST) simulations, respectively. This study provides an effective engineering strategy for overcoming persistent challenges in MHP-based memristors, thus advancing their potential for next-generation hardware-based neuromorphic computing applications.