Improved Performance of Yttrium Oxide-Based Memristor Through TiN Electrodes and Device Scaling for Neuromorphic and Pattern Recognition

Herein, we present a CMOS-compatible fabrication process, in-depth materials, and electrical analysis of yttrium oxide (Y₂O₃)-based memristive devices having a device size of 100 μm². The fabricated devices exhibit improved performance by incorporating TiN electrodes and device scaling and efficiently emulate the various low-power neuromorphic and pattern recognition tasks. The fabricated memristive devices exhibit stable bipolar resistive switching behavior with an excellent endurance beyond 50,000 cycles and retention properties exceeding 10⁶ s by maintaining a very high ON/OFF ratio of 10⁴. Additionally, the fabricated devices show remarkable stability in the device switching voltages under cycle-to-cycle (C2C) and device-to-device (D2D) wherein, the coefficient of variability (C_V) in the device switching voltages in C2C and D2D is 4.95% and 11.39%, respectively. Moreover, the fabricated devices efficiently emulate the synaptic response by emulating potentiation, depression, paired-pulse facilitation (PPF), and paired-pulse depression (PPD) and also exhibit the device conductance tunability under the variations in the pulse width as similar to the biological synapse counterpart. Furthermore, the fabricated devices efficiently show the pattern recognition task by achieving an accuracy of 88.2% for the handwriting MNIST dataset. Therefore, the present work opens a new horizon in the field of miniaturized artificial synapses and neuromorphic computing to perform various operations.