In-situ dynamic correction of progressive ablation fluctuations in laser-induced breakdown spectroscopy (LIBS) using Raman spectroscopy and deep learning

Correcting dynamic fluctuations in Laser-induced breakdown spectroscopy (LIBS) due to the interaction between a continuously pulsed laser and the sample remains a critical challenge in enhancing LIBS as a high-precision analytical tool. In this study, we developed an in-situ dynamic correction method based on Raman spectroscopy to optimize and correct fluctuation characteristics during the LIBS dynamic ablation process. Initially, a continuous LIBS ablation model of a metallic sample was constructed, revealing that the plasma temperature follows a Gaussian distribution, and the dynamic ablation mechanism of sample under continuous laser pulses was analyzed. Next, by integrating Raman spectroscopy with deep learning modeling, an in-situ online feedback correction system was designed to iteratively correct LIBS plasma temperature. Finally, using a deep convolutional neural network (CNN) architecture and feature-level data fusion, the Accuracy, Precision, Recall, and F1-Score of the classification model on the corrected dataset were significantly improved to over 99.3 %, compared to values below 82 % before correction. This notable improvement verifies the effectiveness of the dynamic correction method, which provides strong support for advancing LIBS technology toward mature and practical applications.