Neuromorphic computing, inspired by the brain, has emerged as a potential approach to overcome the current von Neumann bottleneck. Herein, a transparent and integrable ZnO/In2O3 heterojunction crossbar artificial synapse was fabricated using magnetron sputtering for neuromorphic computing. The resistive switching behavior of the device originates from the regulated migration of oxygen vacancies within the heterojunction under an applied electric field. Moreover, the fabricated artificial synapse exhibited exceptional stability, with its resistive switching characteristics remaining unchanged after six months of exposure to ambient conditions. The device demonstrates typical synaptic behavior, including paired-pulse facilitation (PPF), short-term to long-term memory transitions, spike-timing-dependent plasticity (STDP), potentiation/depression, and learning–forgetting processes. Furthermore, the fabricated device-based reservoir computing enabled efficient temporal information processing, and an artificial neural network (ANN) incorporating the device’s synaptic achieved 92.37% recognition accuracy of the MINIST database. Besides, in combination with MATLAB, basic digital logic circuits such as OR, AND, XOR, and XNOR were successfully implemented. Based on this, a half-adder and a parity checker were further constructed to achieve the encryption and decryption of information. This work demonstrates the feasibility of ZnO-based heterojunction memristors to develop multifunctional electronic devices for neuromorphic computing and logical operation applications.
Cao et al. (Thu,) studied this question.