Reducing matrix effects on the quantification of Ca, Mg, and Fe in soybean leaf samples using calibration-free LIBS and one-point calibration

Matrix effects are one of the most difficult analytical challenges in laser-induced breakdown spectroscopy (LIBS), since signal intensity depends on both element concentration and matrix. For instance, the measurement of nutrients in plant leaves using LIBS is notably impacted by matrix effects because of differences in fertilizer application, weather, soil properties, and health levels. Thus, to overcome matrix effects on the quantification of nutrients in plant leaves, we propose the use of CF-LIBS, in which the intensity of emission lines is corrected using the one-point calibration (OPC) method. In this study, we analyzed thirty soybean leaf samples of cultivars from various locations and health levels with 0.5 wt% TiO2 (internal standard), for which the matrix effects were visible in the calibration curves. Normalization using a Ti II line increased the average r2 of the calibration curves for Ca II, Mg II, and Fe II from 0.24 to 0.73, but this was insufficient to adjust for the variations in plasma temperature and for the degree to which the elements were ionized. With the proposed models and using just one sample to calibrate the OPC factors, we obtained an accuracy >92% and an average r2 of 0.87 for the quantification of Ca, Mg, and Fe in relation to the atomic absorption spectroscopy (AAS). We then concluded that CF-LIBS with the OPC corrections might mitigate matrix effects in LIBS analysis of soybean leaf samples. More studies are needed to assess whether the OPC factors calibrated with soybean leaf samples can be applied to other species. [Display omitted] •The Ca, Fe, and Mg content of soybean leaf samples was measured using LIBS.•The leaf samples differed in terms of color, texture, composition, and cultivar.•The calibration curves' lack of correlation revealed the presence of matrix effects.•CF-LIBS with OPC corrections was used to reduce matrix effects.•The model achieved >92% accuracy and a relative error of