Brain‐Inspired CMOS‐Compatible ZnSnO Synaptic Array with Ultra‐High PPF for Versatile Neuromorphic Computing

Neuromorphic computing demands artificial synaptic arrays that combine low power consumption, scalability, and full compatibility with semiconductor manufacturing. However, the development of electrolyte-gated transistors (EGTs) into wafer-scale synaptic systems has been hindered by the poor stability and CMOS incompatibility of conventional liquid or polymer electrolytes. Here, we demonstrate a brain-inspired all-inorganic synaptic transistor array integrating low-cost ZnSnO (ZTO) channels with LiPON solid-state electrolytes, both deposited in a CMOS-compatible process. The 6 × 6 crossbar array exhibits exceptional uniformity, high endurance (12 500 switching cycles), and dynamic synaptic plasticity, including excitatory postsynaptic current (EPSC), paired-pulse facilitation (PPF), and long-term potentiation/depression (LTP/LTD). Notably, the devices achieve an ultra-high PPF index of 305%, symmetric conductance modulation, and negligible drift after repeated cycling. System-level validation demonstrates the functional relevance of the array: offline training combined with hardware-aware inference achieves 97.0% and 86.7% accuracy on the MNIST and Fashion-MNIST datasets, respectively. 96.0% accuracy in convolutional neural network (CNN) simulations for human action recognition. Furthermore, electromyography (EMG) signal classification improves from 88.4% to 96.5%, highlighting its practical potential in neuromorphic sensing interfaces. By combining a ZTO channel with an inorganic electrolyte, this work establishes a CMOS-compatible and scalable materials platform, providing a practical pathway toward system-level neuromorphic applications.