Memristive devices for neuromorphic computing hardware

Project ID: 2531bd1664

(You will need this ID for your application)

Research Theme: Information and Communication Technologies

Research Area(s): Advanced Materials
AI Technologies
Microelectronic Device Technologies

UCL Lead department: Electronic and Electrical Engineering (EEE)

Department Website

Lead Supervisor: Tony Kenyon

Project Summary:

In this project you will develop new memristive materials and devices for brain-inspired, neuromorphic, computing systems. Memristors are a new class of electronic device whose electrical behaviour mimics some of the properties of biological neurons and synapses, and are something that we have been studying for several years. We know that the human brain requires around a million times less power than equivalent digital computers, so taking inspiration from biology opens exciting possibilities to drive down the power consumption of computing and perhaps provide a better understanding of some of the ways the brain functions.

In more detail, you will:

  1. Characterise the electrical response of multilayer oxide nanolayers in novel memristive devices.

  2. Develop physical and compact models for the electrical response of such systems.

  3. Investigate the use of such systems in the new class devices whose data retention characteristics can be tailored to specific tasks and traded off against the number of times the memory element can be rewritten. This emerging class of memory has huge potential to drive down the energy consumption of cloud-based AI systems.

  4. Explore novel device designs incorporating the above systems, paying particular attention to their manufacturability using standard CMOS processing.

  5. Explore how such devices can enhance the operation of standard memristive devices when combined into small systems.

The project will involve a mix of sample preparation, electrical and physical characterisation, and development of appropriate physical and electrical models. There will be opportunities through the Neuroware IKC to collaborate with other academic partners and to attend training courses, including those run by our foundry partner, Imec. We anticipate that some device samples will be provided by our spin-out company, Intrinsic. We will also provide access to entrepreneurial training through the IKC and other routes.