A LIBS spectrum baseline correction method based on the non-parametric prior penalized least squares algorithm.

Area of Science:

• Spectroscopy
• Analytical Chemistry
• Data Science

Background:

• Laser-induced breakdown spectroscopy (LIBS) offers real-time, non-destructive multi-element analysis.
• A significant challenge in LIBS is the presence of continuous background noise in spectra, hindering accurate analysis.
• Existing baseline correction methods often require prior parameter knowledge and can lack robustness.

Purpose of the Study:

• To develop an advanced baseline correction method for LIBS spectra.
• To improve the accuracy of quantitative analysis in LIBS by addressing spectral background interference.
• To create a robust and adaptive algorithm that does not require prior parameter settings.

Main Methods:

• Proposed a novel non-parametric prior penalized least squares (NPPPLS) algorithm for LIBS spectral baseline correction.
• Introduced a new weighting method for faster convergence and combined the Adam algorithm for adaptive parameter updates.
• Validated the method using both simulated data and experimental LIBS spectra, followed by univariate and multivariate analyses.

Main Results:

• The NPPPLS algorithm demonstrated excellent baseline correction performance on simulated data, even without parametric priors.
• The method showed improved stability and robustness, unaffected by the initial balance parameter value.
• Baseline correction significantly enhanced quantitative analysis accuracy, with multivariate analysis achieving an R² of 0.99 for element detection.

Conclusions:

• The proposed NPPPLS algorithm effectively corrects spectral baselines in LIBS, leading to improved quantitative analysis accuracy.
• The adaptive nature and robustness of NPPPLS make it a superior alternative to traditional methods.
• This method holds potential for baseline correction in other spectroscopic techniques like Raman and near-infrared spectroscopy.