Development of Novel Nano Platforms and Machine Learning Approaches for Raman Spectroscopy

de Carvalho Gomes, Paulo Alexandre ORCID: 0000-0003-2986-1157 (2023). Development of Novel Nano Platforms and Machine Learning Approaches for Raman Spectroscopy. University of Birmingham. Ph.D.

Preview

Gomes2023PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (14MB) | Preview

Abstract

In Raman spectroscopy, data analysis occupies a large amount of time and effort; thus, it is paramount to have the proper tools to extract the most meaning from the Raman analysis. This thesis explores improved ways to analyse Raman data mostly by using machine learning techniques available in Python. The substrate used throughout this thesis has been patterned through an electrohydrodynamic process that patterns micrometric pillars onto the substrate, which, after being gold coated, can generate surface-enhanced Raman scattering. An initial theoretical background was laid for the electrohydrodynamic process and additional observations regarding the fluid mechanics. Furthermore, when the structures are fabricated, and Raman measurements are taken, we show that it is possible to create an effective convolutional neural networks that systematically evaluate these patterns’ surface morphology and extracts features responsible for the surface-enhanced Raman scattering phenomenon. Being able to predict 90% of the time from optical microscope images and 99% of the time with atomic force microscopy images Additionally, a thorough machine learning analysis of the Raman literature was done. The best machine learning algorithms were put together into a script combined with a graphical user Interface that can run multiple commands such as principal component analysis and self-organizing maps, all in a centralised way. This way, we managed to consistently extract information from Raman and surface-enhanced Raman scattering spectra to open possibilities for precise peak analysis methods using a multi-Lorentzian fit algorithm.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):