Anlysis And Application Research Based On Threshold Adaptive Memristor Model

With the Very Large Scale Integration(VLSI) circuits approaching the area of deep-submicron, the scale of electronic devices integration is becoming massively. At the same time, transistor’s scale is close to the bottleneck, which leads to the unstable and unreliable performance in the applications of logical operation and information storage with more challenges. In recent years, although the artificial intelligence and neural network have made dramatic advances, they are still confined by the bulk of transistor so that are unable to implement on a large scale. Memristor is a next-generation high-powered RRAM device with the features of nano-scale and non-volatile, which is expected to extend the “Moore’s Law” and also meets the requirements of neuromorphic implement massively. With the theoretical foundation of memristor growing fast and mature, the Tr Es hold Adaptive Memristor(TEAM) model has received more and more researchers’ concern. On the one hand, the TEAM model defines the switching hop thresholds to the internal of model, which revises the lack of threshold definition in the two-terminal HP memristor model. On the other hand, based on the physical theory of Simmons Tunnel model, the TEAM model is approaching to the memristor’s actual situation. Moreover, the flexibility of parameters’ adjustable makes TEAM model being ability of becoming the universal memristor model. Utilizing the TEAM model in logical operation, crossbar array, neuromorphic can not only decrease the bulk of circuit, also be beneficial and meaningful in memristive system and memristive large-scale integration.In this paper, the TEAM model is analyzed in the areas of inner characters and composite circuits. On this basis, the TEAM model is applied in the logical operation, multi-value storage, wavelet neural network. Specific content as follows:(1) Introduce the basic memristor model and few classical window functions to simulate the ion drift in the boundary. Meanwhile, the Simmons tunnel model and two types of TEAM model(Current Threshold Adaptive Model and Voltage Threshold Adaptive Model) are programmed in Matlab. From the simulation, the high fitting between Simmons tunnel model and TEAM model is demonstrated.(2) The multi-parameters in TEAM model are simulated by Monte Carlo simulation in pairs. The upper or lower boundary of adjusting unilaterally are verified. The result also reveals the mechanism of switching asymmetry in the TEAM model. What’s more, the TEAM model composite circuits are proposed to verify the influence to whole memristance of different polarity or different structure.(3) Aimed on the different threshold voltage in VTEAM model, a series of VTEAM logical operation circuits are designed to realize the logical operations of OR, AND, NOT, NOR, NAND. Different conditions of external voltage source are deduced. At the same time, a TEAM crossbar array scheme is p roposed. Compared to former schemes, a novel multi-voltage control approach is presented to improve the efficiency of parallel processing in nonvolatile memristive image sto rage.(4) A novel wavelet neural network(WNN) neuromorphic system is proposed combining WNN and VTEAM model. The VTEAM’s memristance variation is deduced due to different periodic of pulse function. The results reveal the possibility of VTEAM acting as electronic synapse to achieve dynamic updating weights. Furthermore, we utilize the VTEAM WNN model in the short-term traffic prediction, which verify the correctness of whole framework design.