Memristor based chaotic circuit and neuromorphic systems

Abstract

The field of electronics has had three fundamental passive components for the past few centuries. In 1971, Prof. Leon O. Chua mathematically postulated the possibility of a two-terminal component named memristor which is passive with properties that cannot be replicated by any combination of the existing passive components. The experimental investigation by HewlettPackard in 2008 has led to the discovery of the first evidence of a physical device which has the properties of the proposed device. The device resistance is dependent on the past value and the current applied input. It is the nonvolatile property, resistance does not change until the next input, that has generated immense interest. The memristor requires two equations, the port and state equation, to define the device whereas the other fundamental devices require only a single equation. It has applications in memory technology, programmable circuits, neuromorphic systems and computing architecture. The programmability and non-volatility properties have opened the route for analog implementation of neural networks. In this thesis, the author has conducted research work in model development, chaos generation using memristors and the study of the application of memristors in neuromorphic systems. There have been several memristor models developed. Most models are based on the linear drift model which models the port equation of the memristor as a combination of a slider with low and high resistors. Other models which are based on this linear drift model, add a window function to the state equation to include non-linearity. The models developed on the linear drift model produces the required fingerprints that characterise the memristor but are not based on any physical properties. The objective of the model development section of this research work is to develop a memristor model that is based on the physics of the devices and also satisfy the fingerprint requirements. Both the port and the state equation have to be based on physical characteristics thatmo