Programming Exploration of Memristor Crossbar

Memristor crossbar is prevailing as one of the most promising candidates to construct the neural network because of their similarity to biological synapses, favorable programmability, simple structure and high performance regarding area efficiency and power consumption. However, the performance of the memristor crossbar is limited by unideal programming and sensing process.\ud \ud In this thesis, the most preferred cell structure which is known as “one-transistor-one-memristor” is investigated. Different factors that may have impacts on programming, such as the structure, the parameters and the conductance of a crossbar cell are studied using both theoretical analysis and simulation. \ud \ud Based on previous analysis, the programming process of the memristor crossbar deserves a deep exploration. For programming, the primary objective is to find out the relationship between the programmability of the memristor crossbar and its characteristics, such as the IR-drop and the crossbar size. The results are expected to be useful references for researchers designing the memristor crossbar. \u