Multivariate Statistical Methods that Enable Fast Raman Spectroscopy
Raman spectroscopy is a useful tool in investigating inter- and intra-molecular interactions as well as classifying and quantifying chemical species in a sample. Many materials of societal interest, such as proteins and pharmaceuticals, have distinctive Raman spectra with sharp features. However, the adoption of Raman spectra has been hindered by the low rate of Raman scattering, interference from fluorescence, and high spectrometer costs. This work demonstrates multivariate stastical methods that enable fast Raman measurements, despite the low rate of Raman scattering. These methods include a novel type of spectrometer, that uses computer-controlled optical filters to efficiently capture Raman photons and multiplex them onto either one or two photon counting detector(s). This method, referred to as optimal-binary compressive detection (OB-CD), allows for the collection of chemical information in 10’s of microseconds, rather than milliseconds as might be common for Raman spectroscopy performed using a multichannel detector. A method for orthogonalizing moderate amounts of fluorescence from Raman signal in OB-CD is presented. Fast imaging, with speeds as high as 2.5 frames-per-second, is demonstrated and algorithms for image denoising are discussed. Lastly, methods that enables Raman classification using minimal computation time and a technique for accurately processing Raman thermometry data are presented.
Ben-Amotz, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our