Photo-chemical adaptive integrated circuits for next generation neuromorphic computing

Neuromorphic computing, i.e., an approach to information processing inspired in basic morphology and working principles of the brain, is a rapidly growing area of research due to the availability of high-speed hardware and need for artificial intelligence (AI) systems across a wide range of applications. In the last decade, artificial neural networks (ANNs) have achieved the highest success in machine learning tasks and have driven leading advances in AI. Currently, ANNs rely on classical electronics-based computing architectures that scale inefficiently with the system size and challenge energy consumption.

 Photonics presents fundamental advantages for large-scale hardware neural networks. A simple photonic link (i.e., optical fibres and on-chip waveguides) can simultaneously transmit multiple independent signals at very high bandwidths and with low attenuation and heat generation. Thus, photonic integrated circuits are strong candidates to overcome important bottlenecks to neuromorphic hardware scalability, by enabling the required large-scale parallel connectivity at considerably higher energy efficiencies and operation bandwidths. A key outstanding challenge in photonic systems is the realisation of plasticity. A network is plastic when its input can modify its response in a non-volatile way, which means that some network parameters depend on the past input history. This so-called synaptic plasticity will have a fundamental role in the future development of neuromorphic computing. Indeed, the biological brain learns and memorizes by plastically adapting to its input, so that its network parameters do not need to be tuned by an external algorithm. The implementation of plasticity in a photonic ANN requires, at the same time, the interaction of the light with itself via the network (i.e., nonlinear effects) and the permanent storage of the effects of this interaction. Both requirements are notoriously difficult to meet in optics, especially when energy efficiency and scalability are key.

 The main goal of this project is to implement photonic integrated devices whose state can be tuned with light and stored as in-device memory. Those devices will be used to build hardware-based photonic neural networks demonstrating synaptic plasticity.

Light-based tunability will be achieved via hybridization of the semiconductor chips with photochemical polymers and hydrogels capable to change their state when exposed to light. Because of the high sensitivity of such reactions, we expect to obtain energy efficient and scalable plasticity in the network connections, therefore enabling large-scale photonic interconnects that can adapt to input data. Moreover, such hybrid devices will exhibit emerging dynamics allowing us to explore the rich boundaries between the fields of complex systems and machine learning.

 In this project, the student will first gain expertise in design and fabrication of photonic integrated devices in the TIC cleanroom facility. Then, the student will develop strategies to hybridize those semiconductor devices with photochemical materials for self-tuning behaviour. Finally, those devices will be used to implement photonic neural networks capable of synaptic plasticity. The student will be part of a larger research group with the opportunity to work with others in a collegiate and enthusiastic team. Noteworthy, this project is part of a funded, international collaboration with two other groups from Germany (Johannes Gutenberg University Mainz) and from Italy (University of Trento). The student will benefit from this partnership, gaining expertise on photochemical materials from Uni. Mainz and on advanced machine learning algorithms from Uni. Trento. The student will have the opportunity to visit those two groups during the PhD period. Research findings will be published in high impact journals with the opportunity to present at international conferences.

Institute of Photonics: The Institute of Photonics (IoP), part of the Department of Physics, is a centre of excellence in applications-oriented research at the University of Strathclyde. The Institute’s key objective is to bridge the gap between academic research and industrial applications and development in the area of photonics. The IoP is located in the £100M Technology and Innovation Centre on Strathclyde’s Glasgow city centre campus, at the heart of Glasgow’s Innovation District, where it is co-located with the UK’s first Fraunhofer Research Centre. Researchers at the IoP are active in a broad range of photonics fields under the areas of Photonic Devices, Advanced Lasers and Neurophotonics, please see:

http://www.strath.ac.uk/science/physics/instituteofphotonics/ourresearch/.

Strathclyde Physics is a member of SUPA, the Scottish Universities Physics Alliance.

The University of Strathclyde has, in recent years, been the recipient of the following awards: The Queen’s Anniversary Prizes for Higher and Further Education 2019, 2021 & 2023; Times Higher Education University of the Year 2012 & 2019; Daily Mail University of the Year 2024 Runner-Up; Daily Mail Scottish University of the Year 2024; Triple E European Entrepreneurial University of the Year 2023.

Student eligibility:

To enter our PhD programme applicants require an upper-second or first class BSc Honours degree, or a Masters qualification of equal or higher standard, in Physics, Engineering or a related discipline. Full funding, covering fees and stipend, is available for applicants who are UK Nationals (meeting residency requirements) or have settled status (meeting residency requirements), pre-settled status or otherwise have indefinite leave to remain or enter.