Machine learning efficiently corrects LIBS spectrum variation due to change of laser fluence

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning Applications

Background:

• Laser-induced breakdown spectroscopy (LIBS) is sensitive to laser pulse energy fluctuations.
• Variations in laser pulse energy (7.9–71.1 mJ) significantly impact spectral intensity.
• Accurate elemental analysis requires robust correction methods for these variations.

Purpose of the Study:

• To demonstrate the efficacy of machine learning for correcting LIBS spectral intensity variations.
• To develop a multivariate correction model for LIBS data affected by laser pulse energy changes.
• To compare the developed method against classical univariate correction techniques.

Main Methods:

• Development of a multivariate machine learning model for spectral correction.
• Application of the model to LIBS spectra of aluminum alloys.
• Experimental variation of laser pulse energy over a wide range (7.9–71.1 mJ).
• Comparison with univariate methods using parameters like laser pulse energy, total spectral intensity, ablation crater volume, and plasma temperature.

Main Results:

• The machine learning model precisely corrected spectral intensity variations.
• Accurate determination of magnesium concentration in aluminum alloys with 6.3% relative standard deviation (RSD).
• The multivariate approach demonstrated superior performance compared to univariate correction methods.

Conclusions:

• Machine learning offers an efficient and powerful solution for correcting laser pulse energy-induced variations in LIBS.
• The developed multivariate model significantly improves the accuracy and precision of elemental analysis in alloys.
• This approach enhances the reliability of LIBS for quantitative analysis in challenging conditions.