A physics-oriented memristor model with the coexistence of NDR effect and RS memory behavior for bio-inspired computing

Bio-inspired computing promises fundamentally different ways to advances in artificial intelligence with extreme energy efficiency. Memristive technologies due to the non-volatility, high density, low-power, and synaptic bionic properties can help in realizing bio-inspired architecture and its hardware implementation. This paper proposes a novel physics-oriented memristor model with coexistence of negative differential resistance (NDR) effect and resistive switching (RS) memory behavior for bio-inspired computing. Firstly, an Ag/TiOx/FTO memristor is fabricated using sol-gel and magnetron sputtering method, and its performance test demonstrates that the coexistence of NDR effect and RS memory behavior can be modulated by the moisture. Then, a physical-oriented memristor model is constructed, which provides the possibility to explore the dynamics of the coexistence of NDR effect and RS memory behavior in simulation. Furthermore, a memristor-based affective computing circuit emulating the process of human affective associative learning is designed. The experiment demonstrates that the coexistence of NDR effect and RS memory behavior can change the memory time without additional circuit and cost, which is expected to realize the automatic conversion from short-term memory to long-term memory in bio-inspired computing.