Thesis
Photonic technologies for ultrafast neuromorphic information processing
- Creator
- Rights statement
- Awarding institution
- University of Strathclyde
- Date of award
- 2026
- Thesis identifier
- T18072
- Person Identifier (Local)
- 202188699
- Qualification Level
- Qualification Name
- Department, School or Faculty
- Abstract
- The rapid growth of artificial intelligence has placed increasing demands on digital computing hardware, motivating the exploration of alternative computing paradigms that offer improved efficiency, bandwidth, and scalability. Neuromorphic computing, inspired by the event-based and energy-efficient operation of biological neural systems, represents one such paradigm. In parallel, photonic technologies offer unique advantages for neuromorphic processing, including ultrafast signal propagation, high bandwidth, low communication losses, and intrinsic parallelism. This thesis investigates the use of two photonic and optoelectronic devices as photonic spiking neurons for neuromorphic computing. The first part of the thesis focuses on Vertical-Cavity Surface-Emitting Lasers (VCSELs) operated under optical injection. It is shown experimentally that VCSELs can exhibit rich nonlinear dynamics that closely resemble key neuron-like behaviours, including excitability, thresholding, temporal integration, refractoriness, and spike generation on sub-nanosecond timescales. These properties are exploited to realise photonic spiking neural networks using time-division multiplexing, enabling networks with large effective neuron counts operating at gigahertz rates. The computational capabilities of these systems are demonstrated through tasks such as nonlinear classification and chaotic time-series prediction, supported by training strategies adapted to the binary and temporal nature of photonic spikes. The second part of the thesis investigates Resonant Tunnelling Diodes (RTDs) as compact optoelectronic spiking neurons. A physics-based circuit model is used to describe the excitable dynamics arising from the negative differential resistance of RTDs, and numerical simulations are used to explore their spiking behaviour under optical and electrical perturbations. Experimental demonstrations validate these models and show that RTD neurons can operate as ultrafast, multi-modal spiking neurons. The applicability of RTD-based neuromorphic systems is illustrated through several tasks, including photonic spiking extreme learning machines for classification, recurrent architectures for tunable photonic memory, and event-based time-series processing. In particular, a dual optical–electrical modulation scheme is shown to enable efficient rising-edge detection in complex signals, bridging sensing and computation at the device level.
- Advisor / supervisor
- Hurtado, Antonio
- Resource Type
- DOI
Relations
Items
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
PDF of Thesis T18072 | 2026-08-17 | Public | Download |