Research On Memristor Based Nonvalatile Memory

With the size of transistor approaching the sub 100 nm barrier, a few serious hinders have started to emerge, including difficulty in fabrication and fundamental performance defects of transistors such as unmanageable electron motion and so on. Transistor based traditional memory has encountered a development bottleneck. It is believed that Flash memory will approach the end of scaling down within a decade. As a result, novel devices and technology need to be developed to meet the ever increasing demands for nonvolatile, high-density, low-power, and high-speed primary and secondary storage devices. The memristor is a novel type of nonlinear two-terminal nanoscale element and has promising applications in the areas of nonvolatile memory, very large scale integration (VLSI), artificial neural networks, artificial intelligence and so on. The development of memristor is expected to improve the theoretical foundation and applications of entire electronic circuits. With nano-scale size, memory capability, continuously variable resistance, fast switching (<10 ns), low energy (-1 pJ/operation), high write endurance (1010), scalability, stack-ability, and CMOS compatibility, memristor brings new supports for the development of semiconductor memory technology. Memristor based memory is expected to be one of the most important technologies for new generation nonviable memory.In this paper, the theoretical foundation and working principle of memristor have been studied in depth and several kinds of mathematical formulas have been derived. The results of numerical analysis and experimental simulations verify the basic characteristics of memristor. Following that, a memristive crossbar array is proposed by using the switching characteristic and multilevel resistance state of memristor with applications in binary, gray-scale and color image processing. Meanwhile, considering the combination with current digital computers and the compatibility with CMOS circuit for memristor, we propose the schemes for binary and multilevel memristor based resistance RAM (MRRAM). They have a similar structure to SRAM (static RAM) but possess non-volatile property, higher storage density and capacity. A series of computer simulations prove that they can successfully perform binary data and multilevel information storage and output. Finally, on the basis of the analog storage ability of memristor, an implementation scheme for analog record/play system with memristor based analog memory is designed. Systematically, experimental simulations show the process from the inputting to storage to recovering for analog audio signal. The research results of this paper will provide significant theoretical supports and experimental basis for developing a new generation of nonvolatile memory, which is expected to be used in future biomimetic intelligent devices, including the novel intelligent computer system with the ability of information processing that is similar to that of human brain.