Spontaneous Compressive Raman technology: developments and applications

Spontaneous Raman scattering is a physical process that provides a unique knowledge of materialsat the molecular level. Its high chemical specificity with no labels motivates its use in manydifferent fields, ranging from biomedical research to industrial quality control. Nevertheless, theefficiency of this simple process is limited by its extremely weak cross-section.Typically, the Raman scattered light is dispersed and collected onto an array detector, for severalspatial positions of the sample, resulting in a hyperspectral image. Yet, this leads to thegeneration of overwhelmingly large data sets and to lengthy acquisitions. In situations where hyperspectralmeasurements simply aim to map the spatial distribution of molecules, the spectraldata is unmixed in a postprocessing step, in order to detect molecular species and/or estimatetheir concentrations. In those cases, acquiring a complete vibrational spectrum per spatial pixelmay be inefficient, and a massive speed-up can be achieved by encompassing compressive techniquesin the acquisition process. Some strategies, including compressive Raman technology(CRT), use spectral a priori information to integrate chemometric analysis directly into thespectrometer hardware: the measurement is designed to directly probe quantities of interest tobe estimated (e.g., molecular concentrations), rather than deducing them from complete hyperspectralmeasurements. In CRT, this is made possible by replacing the array detector by asingle-pixel-detector, combined with a programmable optical filter. Based on the a priori knownspectra of pure molecular species contained in the sample, these filters select accurately chosenspectral components and combine them into the detector.This thesis develops some theoretical and technological aspects of CRT and applies it to severalconcrete applications. In a first part of the work, we investigate the estimation precision achievedby CRT, show that our method of estimation is efficient, and experimentally validate thisanalysis. In a second part of the work, we compare CRT, to some extent, to commercial state-of-the-art instrumentation. We find some clear advantages in terms of acquisition speed and limitof detection. We also show some preliminary results that suggest its usefulness for fields relatedto biomedical imaging, pharmaceutical industry and the environment. Last, we take furtheradvantage of the single-pixel architecture of CRT to perform multiplexed line-scan imaging.We quantify the potential gain of this approach in terms of signal-to-noise ratio, when themeasurements are shot-noise limited.