Raman Microscopy Based On Low Resolution Spectroscopy And Its Application

Raman microscopy has attracted more and more attention in the field of chemical analysis due to its properties of being non-destructive,non-labeling and not requiring sample pretreatment.However,due to the fundamentally weak nature of the Raman signal,it is usually necessary to use an expensive,frequency-stabilized laser to excite the Raman scattering,and a costly,deep-cooled detector with high quantum efficiency to collect the Raman signal generated by the sample under low-noise,long-exposuretime conditions in practical applications.This not only leads to a very low detection throughput,but also necessitates a high system cost(due to the precision instrumentation)and a large instument volume,which are both not conducive to the promotion of Raman microscopy as pervasive analytical method.Recently,some studies have shown that reducing spectral resolution can increase the throughput of systems without sacrificing analytical performance,and the idea of "low resolution Raman" has led to the production of low-cost,portable Raman equipment,which has begun to promote the popularity of Raman detection technology outside of laboratory settings.However,there is a lack of a quantitative description of the relationship between spectral resolution and spectral signal intensity,and there is a lack of systematic research on the costs as well as the benefits of low resolution Raman spectroscopy in practical applications.In addition,the current low-cost,portable Raman equipment is limited to macroscopic"point-sampling",and can only be applied to the detection of large volume of liquid,powder and other mesoscopic and macroscopic samples using a manual operator,without the possibility of automated microscopic analysis.This thesis aims to solve these problems,and the specific research content is as follows:1.The "spectral squeezing" theory was proposed based on the Spectral Point Spread Function(SPSF),which fully characterizes the Raman spectral signal as the spectral resolution and spectral sampling are independently varied.The relationship between spectral signal intensity and spectral signal-to-noise ratio is described quantitatively.Our work based on numerical simulation and practical experiments has shown that the decrease in spectral resolution does not necessarily lead to the increase in spectral signal intensity.The width of the original peak would also affect the intensity.The wider the peak width,the greater the intensity increase after the decrease of spectral resolution.Contrariwise,due to the presence of the SPSF,for the narrower peak width,the decrease of spectral resolution may lead to a decrease in signal intensity.2.Taking three practical applications:the relationship between spectral resolution and the detection limit of the system,the relationship between spectral resolution and prediction accuracy,and the relationship between spectral resolution and hyperspectral image quality as examples,we explore the performance of low resolution Raman and analyze the cost and benefit of the method.First,for most of the compounds and biological samples,reducing the spectral resolution can improve the signal intensity,thus improving the detection limit of the system and improving the detection sensitivity.Secondly,the reduction of resolution will lead to both the enhancement of signal strength and the loss of spectral information,which compete with each other to determine the accuracy of substance classification.In our experiments,we found that for most(but not all)samples,reducing the spectral resolution improved the accuracy of the prediction(shortening the exposure time needed to achieve the same prediction accuracy).Thirdly,the reduction of spectral resolution can help us to obtain high-contrast hyperspectral images in a shorter time,and can help us to overcome the camera readout noise,which provides a new idea for the realization of low-cost Raman.3.Using the theoretical and practical knowledge base built up in the preceeding work,we have built a set of portable,low-cost,low-resolution Raman microscope,which has reduced the cost and volume of a Raman microscope by 10 times and 5 times,respectively,compared with the traditional laboratory version.In addition,the automated acquisition capability of the system enables the researchers to easily collect thousands of samples.We used the system to accurately predict the proportion of rare species from mixed samples with a population ratio of more than 1:200.Through the automated and quantitative study of thousands of milk globules from cow and goat milk,combining Raman spectroscopy and bright field imaging,we found that saturated fatty acids were more easily enriched in smaller fat globules.The analysis of milk fat globules can provide insight into milk quality and the health of the individual livestock producing the milk.In this thesis,the relationship between spectral resolution and spectral signal intensity is investigated theoretically.For most applications,reducing the spectral resolution can improve the detection sensitivity of the system,further making it possible to achieve low-cost Raman microscopy devices.