Vertically Aligned Two-Dimensional Halide Perovskite as Artificial Synapses Toward Neuromorphic Computing

Neuromorphic computing, inspired by the human brain which offers a high-performance processing capability with low power consumption (~fJ), has recently emerged as a promising avenue to overcome the limitation of existing von-Neumann computing systems. However, emulation of synaptic properties with conventional silicon circuits inevitably demands extra analog converters, causing problems with power consumption and scalability. Therefore, analog-type artificial synapses, mimicking the biological properties, are needed to implement the brain-inspired computing system.<br/>Thanks to low ion migration energy and fascinating memristive characteristics owing to mixed ionic-electronic conductivity, which are the core of the neuromorphic computation, 3D organo-lead halide perovskites (HPs), have been attracting great attention as artificial synapses recently. They have ABX₃ crystal structure, A = methylammonium (MA), or formamidinium (FA), B = Pb, and X = I, Br, or Cl. However, thin films of 3D HPs typically are formed in polycrystalline nature and thus contain numerous grain boundaries. Since the transport of charge carriers strongly depends on the conduction paths in the insulating layer of a MIM structure, reliable memristive behavior is hard to achieve in polycrystalline HPs. Furthermore, 3D HPs are vulnerable to heat and humid conditions, causing the degradation of memristive characteristics. To address these issues, recent attempts have been made to find novel halide perovskite compositions or to fabricate air-stable dual-phase halide perovskites. Nonetheless, because multiple grains are arranged in numerous orientations, 3D HPs with vertically stacked MIM structure makes it difficult to change the conductance linearly and reliably.<br/>Here, we successfully fabricated vertically aligned 2D halide perovskite films (V-HPs) for active layers of artificial synapses, showing moisture stability for several months. The large organic cations in layer-structured 2D perovskites (formed as R₂A_n-1B_nX_3n+1, R is bulky ammonium cation) can restrict the transport of charge carriers within 2D due to anisotropy in the in-plane and out-of-plane conductivity. Therefore, the arrangement of the PbI6 octahedra framework perpendicular to the substrate can facilitate charge transport in the vertical direction across the electrodes, resulting in reliable switching with ion migration. Unlike random-oriented HPs, which exhibit negligible current hysteresis, the V-HPs possess multilevel analog memristive characteristics, programmable potentiation and depression with distinguished multi-states, long-short-term plasticity, paired-pulse facilitation, and even spike-timing-dependent plasticity. Furthermore, high classification accuracy is obtained with implementation in deep neural networks. These remarkable improvements are attributed to the vertically well-aligned lead iodide octahedra acting as the ion transport channel, confirmed by first-principles calculations. This study paves the way for understanding HPs nanophysics and demonstrating their potential utility in neuromorphic computing systems.