Multi-Plane Light Converter based on metasurface and Machine Learning to understand the mode sorter’s applications

Angelucci, Sara (2025) Multi-Plane Light Converter based on metasurface and Machine Learning to understand the mode sorter’s applications. PhD thesis, University of Glasgow.

Abstract

This scientific study delves into the realm of metasurfaces, offering an exhaustive investigation into their underlying principles, practical applications, and the fabrication methods imperative for their realisation. A focal point of this exploration is the detailed exposition of the fabrication process for the Multi-Plane Light Convertor (MPLC) device, supported by captured images validating the precision of each critical step. Initial results indicate the satisfactory functioning of the MPLC, yet further analyses and optimisations are deemed essential to unlock its full potential.

The MPLC device demonstrates versatile applications across telecommunications, energy-related fiber sensing, medical imaging, and biological tomoholography. However, at present, no physical devices based on metasurfaces are available that can fully implement these functions.

In parallel, a novel and robust fibre bend sensor has been developed, showcasing the capability to precisely locate bends through inter-modal coupling. Modal decomposition reduces sensitivity to relative phase, revealing features providing accurate information about the shape or position of bends within the fiber. The simplicity and cost-effectiveness of this approach offer potential applications in wearable technology, motion sensors and aircraft wing shape sensing.

Both experiments revolve around the concept of a mode sorter. The first experiment focuses on creating a novel device not yet available on the market, specifically the Multi-Plane Light Converter (MPLC). The second set of experiments, on the other hand, is centered around the practical application of the mode sorter as an instrumental component. The combination of mode de-multiplexing with machine learning holds promise for powerful applications, particularly in scenarios where constant variations in relative phase can be treated as noise, such as monitoring atmospheric conditions or extracting information from environments with dense scattering.

Practical deployment considerations include the need for retraining in cases of significant system or fiber type changes. Once fully trained, retraining intervals are typically weeks to months under normal temperature fluctuations, necessitating further research into extreme temperature variations encountered in applications like aviation. The use of multi-core fibers is recommended to enhance sensitivity to multiple directions.

In summary, the study demonstrates the feasibility of utilizing machine learning for accurate millimetric-scale curvature detection by incorporating a mode sorter into the optical setup. While exhibiting robust performance, limitations exist in detecting bends or movements not introducing changes in inter-modal coupling and relative phase shifts. The consistent alignment and outcomes observed in experiments underscore the stability and reliability of the experimental setup, instilling confidence in the algorithm’s performance.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
• Colleges/Schools: College of Science and Engineering > School of Engineering
• Supervisor's Name: Clark, Professor Alasdair
• Date of Award: 2025
• Depositing User: Theses Team
• Unique ID: glathesis:2025-84862
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 06 Feb 2025 14:17
• Last Modified: 06 Feb 2025 14:17
• Thesis DOI: 10.5525/gla.thesis.84862
• URI: https://theses.gla.ac.uk/id/eprint/84862