Development of a method for analyzing two-component mixtures using raman spectra

Abstract

In this paper, the urgent problem of quantitative analysis of two-component mixtures by raman spectroscopy is solved. Existing chemometric approaches, such as the latent structure projection (PLS) method, require extensive calibration samples, which limits their use in rapid analysis. An alternative method based on the physical principle of spectrum additivity and calibrated using a single averaged spectrum of a mixture of known composition is proposed.
To identify the components of the mixture, an iterative algorithm based on the Pearson correlation coefficient is implemented, followed by subtraction using the non-negative least squares (NNLS) method. The baseline correction by the method of asymmetric least squares (AsLS) was applied to remove the fluorescent signal. A mathematical relationship is proposed that relates the ratio of component contributions to their concentration.
A modification of the developed method using derivative spectrophotometry is proposed, which automatically eliminates the fluorescent signal.
The experimental test was carried out on 9 types of two-component mixtures with a mass content of components from 10/90% to 90/10% in increments of 10%. The training sample for the PLS reference method was 216 spectra (24 spectra for each percentage). For the developed method, calibration was performed using only one averaged spectrum of a 50/50% mixture. Validation of both approaches was carried out on an independent test sample of 54 spectra (6 measurements for each concentration point). All spectra were recorded with three exposure time values — 30, 100, and 300 ms, which provided a variation in the signal-to-noise ratio in the SNR range of 20-100.

Speaker

Gylka Roman
Bauman Moscow State Technical University
Russia

Discussion

Ask question