Flexible Active Crossbar Arrays Using Amorphous Oxide Semiconductor Technology toward Artificial Neural Networks Hardware

Memristor crossbar arrays can compose the efficient hardware for artificial intelligent applications. However, requirements a linear and symmetric synaptic weight update low cycle-to-cycle (C2C) device-to-device variability as well sneak-path current issue have been delaying its further development. This study reports on thin-film amorphous oxide-based 4×4 1-transistor 1-memristor (1T1M) crossbar. The a-IGZO is built flexible polyimide substrate, enabling IoT wearable In novel framework, transistor memristor are fabricated at same level, with processing steps sharing materials all layers. 1T1M cells show symmetrical plasticity characteristic C2C variability. performs like an analog dot product engine vector–matrix multiplications in crossbars demonstrated experimentally, which successfully suppressed, resulting proof-of-concept cost-effective, neural networks hardware.