Intrinsic Performance of Monte Carlo Calibration-Free Algorithm for Laser-Induced Breakdown Spectroscopy

Area of Science:

• Analytical Chemistry
• Plasma Physics
• Computational Science

Background:

• Laser-Induced Breakdown Spectroscopy (LIBS) is a powerful analytical technique for elemental analysis.
• Calibration-free LIBS methods aim to determine elemental concentrations without the need for physical standards.
• The Monte Carlo (MC) algorithm offers a computational approach to complex modeling problems.

Purpose of the Study:

• To evaluate the performance and accuracy of the Monte Carlo (MC) algorithm for calibration-free LIBS.
• To assess the algorithm's efficiency on simulated metallurgical slag spectra.
• To identify factors influencing the MC method's accuracy in elemental analysis.

Main Methods:

• A simulated spectrum mimicking metallurgical slag was generated based on a uniform, isothermal, and stationary plasma model in local thermodynamical equilibrium.
• The MC algorithm iteratively generated millions of plasma parameter configurations (temperature, size, species concentrations) and corresponding spectra.
• A cost function minimized the difference between synthetic and simulated spectra, with species concentrations extracted upon convergence.
• The algorithm was parallelized on a GPU (NVIDIA Tesla K40) to accelerate computation.

Main Results:

• The MC calibration-free LIBS method demonstrated an intrinsic accuracy of approximately 1% for eight tested elements on simulated data.
• Minimization of the cost function on the GPU took several minutes, dependent on the number of iterated elements.
• The study highlighted that real-world experimental spectra might yield lower efficiency due to model idealizations and suboptimal experimental conditions.

Conclusions:

• The MC algorithm shows promising accuracy for calibration-free LIBS analysis of complex samples like metallurgical slag.
• Computational efficiency can be significantly improved using GPU parallelization.
• Further research is needed to address the gap between simulated performance and real-world applicability, focusing on model refinement and experimental parameter optimization.