Minerals, Vol. 15, Pages 1316: A Raman Measurement and Pre-Processing Method for the Fast In Situ Identification of Minerals

Minerals doi: 10.3390/min15121316

Authors:
Dhiraj Gokuladas
Julia Sohr
Andreas Siegfried Braeuer
Daniela Freyer

Through this work, an experimental setup and pre-processing method for obtaining fluorescence and quasi-noise-free Raman spectra of minerals for in situ mineral identification in an underground environment is proposed. It uses a combination of methodologies like dual excitation wavelengths, Shifted Excitation Raman Difference Spectroscopy (SERDS), and deep learning-based U-Net model for background and noise correction. The dual excitation wavelengths technique employs a near-infrared SERDS laser for the fingerprint and a red laser for the large Raman shift region. The SERDS laser operates at two excitation wavelengths and is tuneable in the vicinity of 785 nm. The red laser uses 671 nm excitation wavelength. The obtained fingerprint and large Raman shift Raman spectra are then fed to a pre-processing method containing the trained U-Net model for obtaining a background-corrected and quasi-noise-free Raman spectrum. The proposed method addresses issues of existing handheld Raman systems in terms of spectrometer sensitivity, spectrum acquisition speed, pre-processing time, fluorescence effects, and other interferences due to surrounding light or vibration. The obtained final processed Raman spectrum is then deconstructed into pseudo-Voigt peaks. The identification of the minerals can be based on the number and the positions of the pseudo-Voigt peaks. Samples of gypsum (CaSO4·2H2O) and anhydrite (CaSO4) were used for evaluating the performance of the proposed method. The influence of measurement time on the reproducibility and precision of the peak identification and, thus, mineral identification is also analyzed.

Source link

Dhiraj Gokuladas www.mdpi.com