Automated identification of components in a chemical mixture utilizing multi‐wavelength resonant‐Raman spectroscopy and a Pearson correlation algorithm

Abstract

In complex environments, the ability to identify the constituent chemicals within a mixture is extremely important. By utilizing a Pearson correlation algorithm to compare sets of multi-wavelength resonance-Raman signatures, we demonstrate the automated identification of chemicals within a mixture. Applying a linear mixture model, we are also able to estimate the fractional volumetric abundances contained therein. The multi-wavelength resonance-Raman signature used for identification is obtained by illuminating the unknown mixture with a series of 21 sequential laser wavelengths. This signature is then compared with the signatures of a set of known chemicals. By maximizing the Pearson correlation coefficient between the signature of the mixture and a weighted superposition of the signatures of the pure chemicals, we are able to determine the mixture components with 100% accuracy. The linear superposition of the selected chemicals, which minimizes the least squares distance between the signatures of the mixture, and its mathematical recreation determines the corresponding fraction, by volume, of each chemical within the mixture.

Multiwavelength Resonance Raman Characterization of the Effect of Growth Phase and Culture Medium on Bacteria

“…We have previously demonstrated the use of two-dimensional resonance Raman spectroscopy in identifying constituent components of chemical mixtures with the use of Pearson correlation coefficient-based algorithms. 14 We employed a similar technique here to distinguish between bacterial species. The 2D Raman signatures were grouped into 12 classes based on growth condition, three classes for each of the four species, and assembled into a signature library.…”

“…Multiple-wavelength resonance Raman spectroscopy has been demonstrated to offer significant advantages over conventional Raman spectroscopy in distinguishing between different types of bacteria and chemicals, especially when they are present in complex mixtures. 13,14…”

Recent advances in linear and nonlinear Raman spectroscopy. Part VII

“…Lau and co‐workers developed a Bayesian Whittaker–Henderson smoother for general‐purpose and sample‐based spectral baseline estimation and peak extraction . Lunsford and co‐workers described a Pearson correlation algorithm for the automated identification of components in a chemical mixture utilizing multi‐wavelength resonant‐Raman spectroscopy . Widjaja and Garland presented a detailed spectroscopic analysis of complex multicomponent materials using a combination of Raman mapping with band‐target entropy minimization …”

Isotope Identification Mechanisms Enabled by Swept-Wavelength Raman Spectroscopy

“…This type of additional information has already been shown to aid in the detection of materials in complex environments. 14–16 The shifted correlations illustrate the potential of swept wavelength techniques by showing the isolated effect of variations in Raman band intensity at different wavelengths. Measurements in the visible range suggested that the identification mechanisms are primarily evident in the UV, or resonance Raman, region.…”

“…A swept-wavelength spectrum provides a linearly independent set of Raman spectra, further discussed in Appendix 1, which enables improved signal processing by measuring not only the wavenumber of Raman peaks but also the shape of the peaks as a function of wavelength. This has previously been shown to increases the accuracy of sample identification in complex mixtures with overlapping Raman peaks, 14 correctly identifying each component in a set of 31 unique mixtures, while single wavelength measurements were insufficient in up to 25% of the cases. Increased specificity has also been demonstrated 15 by calculating the ratio of Raman band intensities with as few as two wavelengths.…”