ИСТИНА |
Войти в систему Регистрация |
|
ИПМех РАН |
||
Laser-induced breakdown spectroscopy (LIBS) is a type of atomic emission spectroscopy which employs laser emission for sampling via ablation while the resulting laser-induced plasma acts as both excitation source and analytical signal source. Ecological monitoring of lead in soils (for example, near roads or metallurgical/glass factories) and geochemical prospecting require fast elemental analysis of solids suitable for being conducted in the field. LIBS is a promising technique for such purposes. However, spectral noises caused by interfering elements worsen accuracy of optical atomic emission spectroscopy techniques. Since multivariate calibration allows improving accuracy of analysis when signals overlap, we checked whether the reduction of uncertainty was possible for low lead concentrations (32 ppm is MPC according to Russian norms / 16 ppm is the mean content in Earth crust) in soils of different types (black, red, gray soil and others). A set of 15 samples was involved into construction of calibration models. A spectral range containing lead line at 405.78 nm was selected. Leave-one-out cross-validation was employed to validate the model. Different baseline correction methods were tried; subtraction of minimal value within spectral range for each spectrum provided the best accuracy for PCR. A comparison of PCR results with univariate regression model results was performed. The use of PCR has lowered the RMSEC, thus the determination of lead in soils down to 15 ppm became available. The uncertainties of predicted lead content at the required concentration range were in agreement with the confidence band given for certified values.