Ferroelectric memristor crossbar arrays for highly integrated neuromorphic computing system

Abstract

This study suggests a ferroelectric memristor array device optimized for neuromorphic computing systems, leveraging a TiN/HAO/SiO2/n+ Si structure. The proposed 24 × 24 crossbar array demonstrates scalable device characteristics through varying cell sizes (10 × 10–70 × 70 µm²), highlighting improved tunneling electroresistance (TER) ratios and switching speed in smaller cells due to reduced domain counts. The device exhibits short-term memory (STM) and long-term memory (LTM) properties, enabling the emulation of biological synaptic behaviors such as paired-pulse facilitation (PPF) and spike-duration/spike-number-dependent plasticity (SDDP, SNDP). Furthermore, the ferroelectric memristor array functions as a reservoir layer in a reservoir computing (RC) system, achieving high accuracy in MNIST and Fashion MNIST pattern recognition (98.78 % and 88.78 %, respectively). Experimental results confirm its capability to mimic Pavlovian associative learning and nociceptor functions, reflecting both volatile and non-volatile memory characteristics. The uniformity of the fabricated array is validated through device-to-device and cycle-to-cycle switching variations, ensuring its feasibility for high-density memory applications. This work underscores the potential of ferroelectric memristor devices as key components in future neuromorphic computing architectures, offering energy efficiency, scalability, and functional versatility. © 2025 Elsevier Ltd