Introduction
Cryogenic processors are gaining interest due to the rising demand for cryogenic and space electronics.
However, today’s processors are based on CMOS technology that is prone to freeze attack and magnetic interference at low temperatures.
In this project, you will invent a novel building block of processors based on memristor technology to achieve error-tolerant AI chips. Memristor is an emerging memory technology that can mimic a biological neural network, rendering low-powered neuromorphic computing. Memristor can also be programmed with optical stimulation, thus called optomemristor, demonstrating its unique feasibility to be integrated with photonics for ultra-fast operation.
Von Ardenne GmbH, Germany (VA) will be the industrial partner of this project, and you will have the opportunity to visit VA facilities and collaborate with VA engineers who can provide support on materials processing and characterisations.
The Course
In the 1st year, you will build your first optomemristor chip prototype utilising Southampton’s and Von Ardenne’s state-of-the-art cleanroom facility. You will learn nanofabrication (thin film engineering and lithography technique) to fabricate a wafer-scale massive array of optomemristor and surface/interface analysis (advanced microscopy and spectroscopy tools) to evaluate the quality of your prototype.
In the 2nd year, you will develop a protocol to encode data via laser pulses under a magnetic field at cryogenic temperatures. You will learn electrical characterisation techniques to evaluate the synaptic capability and performance of your prototype.
In the 3rd year, you will design a neural network based on the experimental results and evaluate the computational accuracy of your chip. You will learn how to inject noises into the algorithm to emulate the interference induced by magnetic field and cryogenic temperatures, and develop error correction mitigation strategies.