In this study, the applicability of laser-induced breakdown spectroscopy (LIBS) for quantitative analysis of inorganic constituents and origin discrimination of fermented and powdered food products was investigated. In the first part of the study, a l...
In this study, the applicability of laser-induced breakdown spectroscopy (LIBS) for quantitative analysis of inorganic constituents and origin discrimination of fermented and powdered food products was investigated. In the first part of the study, a low-power, low-resolution LIBS system was employed to quantify magnesium (Mg) content in fermented soybean paste samples. Nine commercial fermented soybean paste products were analyzed using a low-power diode-pumped solid-state laser coupled with a compact, low-resolution spectrometer. The Mg concentrations were independently determined by inductively coupled plasma–optical emission spectroscopy (ICP-OES) and used as reference values. Univariate calibration models were constructed based on the Mg II (279.8 nm) and Mg I (285.2 nm) emission lines observed in the LIBS spectra. In addition, a partial least squares regression (PLS-R) model exhibited higher accuracy than the univariate calibration models. The Mg content showed a strong correlation with the type of salt used in the products, and the results demonstrated that real commercial products could be effectively utilized as calibration standards for LIBS analysis. In the second part of the study, LIBS was combined with machine learning algorithms to discriminate the geographical origin of 79 red chili powder samples produced in South Korea and China. Based on interclass distance analysis of the LIBS spectra, nine representative emission lines corresponding to Mg, Li, Rb, Ca, K, Na, H, C, and O were selected. Among these, combinations of Mg, Li, and Rb variables were used to construct one-, two-, and three-dimensional classification models using the k-nearest neighbors (k-NN) algorithm. The three-dimensional model achieved the highest classification accuracy of 97.5% based on leave-one-out cross-validation (LOOCV). These results suggest that differences in agricultural environments and geological characteristics of the production regions are reflected in the LIBS spectral features. Overall, this study demonstrates that LIBS can be reliably applied to both quantitative mineral analysis of fermented soybean paste and machine learning–based origin discrimination of red chili powder using real food samples. Owing to its minimal sample preparation requirements and capability for simultaneous multi-element analysis, LIBS was shown to be an effective analytical technique for both quantitative and classification analyses. The findings indicate that LIBS has strong potential as a rapid, practical, and reproducible analytical tool for food quality control and raw material authentication.