Extended Multiplicative Signal Correction-Assisted ResNet1D for Laser-Induced Breakdown Spectroscopy (LIBS)...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Geochemistry

Background:

• Laser-induced breakdown spectroscopy (LIBS) is crucial for elemental analysis.
• Rough surfaces in uranium polymetallic ores cause complex matrix effects and nonchemical variability, hindering accurate rare earth quantification.
• Existing methods struggle with the combined challenges of matrix effects and surface morphology in ore analysis.

Purpose of the Study:

• To investigate the impact of surface roughness on LIBS spectral quality in uranium polymetallic ores.
• To develop and evaluate a novel method for improving rare earth element (neodymium, gadolinium, samarium) quantification accuracy on rough ore samples.
• To compare the performance of a deep learning approach against traditional chemometric methods.

Main Methods:

• Utilized rough, unpolished pelletized uranium polymetallic ore samples.
• Implemented a within-sample median-based extended multiplicative signal correction (EMSC) strategy to mitigate roughness-induced variability.
• Developed a one-dimensional residual network (ResNet1D) for quantitative regression, compared with Partial Least Squares Regression (PLSR) and peak-area normalization.
• Employed a leave-one-sample-out cross-validation (LOSO-CV) framework for robust performance evaluation.

Main Results:

• The proposed EMSC_median correction followed by ResNet1D significantly improved quantification performance for neodymium, gadolinium, and samarium.
• R-squared values increased, and Root Mean Square Error (RMSE) values decreased for all target elements after applying the EMSC_median-ResNet1D method compared to raw data.
• Attribution analysis confirmed that ResNet1D effectively utilizes spectral information from relevant rare earth element regions.

Conclusions:

• The combined EMSC_median and ResNet1D approach demonstrates high effectiveness in reducing nonchemical variability and enhancing rare earth quantification accuracy in rough uranium polymetallic ores.
• This method offers a significant methodological advancement for in situ, rapid LIBS analysis of complex geological samples.
• The findings provide a valuable reference for improving LIBS applications in challenging ore analysis scenarios.