Predicting Silicate Glass Geochemistry Using Raman Spectroscopy and Supervised Machine Learning: Partial Least Square...

Area of Science:

• Geochemistry
• Spectroscopy
• Materials Science

Background:

• Accurate geochemical analysis of silicate glasses is crucial for various scientific disciplines.
• Traditional methods can be time-consuming and destructive.
• Raman spectroscopy offers a rapid, non-destructive analytical technique.

Purpose of the Study:

• To develop a quantitative geochemical analysis method for silicate glasses using Raman spectroscopy and partial least squares (PLS) calibration.
• To establish oxide-specific models for predicting major element concentrations (wt%) in synthetic and natural silicate glasses.
• To investigate the efficacy of multivariate analysis for Raman spectroscopic data in geochemistry.

Main Methods:

• Utilized Raman spectroscopy to collect spectral data from synthetic and natural silicate glass samples.
• Developed univariate partial least squares (PLS) calibration models for quantifying eight major oxides (SiO2, Na2O, K2O, CaO, TiO2, Al2O3, FeOT, MgO).
• Trained PLS models on 48 samples and validated on 21 held-out samples, also testing on an external set of 11 natural basaltic glasses.

Main Results:

• Achieved accurate predictions for major oxides with average root mean square errors for calibration, cross-validation, and prediction around 2.4-2.9 wt%.
• Identified key Raman spectral channels crucial for the regression of specific oxides.
• Demonstrated that Raman band shifts correlate with oxidation state and silica concentration, influenced by Fe3+/Fe2+ ratios and SiO2 content.

Conclusions:

• Partial least squares (PLS) calibration combined with Raman spectroscopy provides a robust method for quantitative geochemical analysis of silicate glasses.
• The developed models show potential for analyzing amorphous silicates, with preliminary validation on natural samples indicating suitability for specific applications.
• This study establishes a foundational approach for future advancements in Raman spectroscopy for geochemical analysis of silicate materials.