A low-power Hf0.5Zr0.5O2-based thin-film memristor enabling boolean logic operations and brain-inspired computing

Complex metal oxide materials-based memristors have attracted increasing attention in emerging computing paradigms because of their inherent data-storage capability and application potential in information processing. In this work, we demonstrate a low-power memristor based on an Hf 0.5 Zr 0.5 O 2 (HZO) film that performs Boolean logic operations and supports brain-inspired computing. The proposed Ag/HZO/Si-based memristor can achieve stable bipolar and non-volatile resistance switching behavior over 10 2 cycles with low programming voltage (−1.5 V to 1.5 V), low power as low as 3 μW/1.8 μW, a coefficient of variation as low as 0.17 %/2.5 %, endurance of 2 × 10 3 cycles, retention of 10 4 s, and excellent weight modulation characteristics. Based on the ECM model, the physical mechanism of the device is attributed to the formation and rupture of metal conductive filaments. On this basis, programmable current characteristic modulation of the device is demonstrated to realize a class of Boolean logic operations without peripheral circuits. Furthermore, the iterative training of convolutional neural networks and facial recognition functions are successfully demonstrated based on long-term potentiation (LTP) and long-term depression (LTD) characteristics. This work provides new technical support for developing next-generation storage technologies.