Выпуск: 1 , Том: 543 , Год издания: 2005

We report a method to predict absolute concentrations of elements in natural wet sediments (cores) from their SR XRF scan measurements. This prediction is problematic because: (i) relationship between fluorescent intensity and element concentrations is strongly biased by variability of pore water content in thick samples; (ii) thickness and mass of fluorescent layers are likewise variable; and (iii) spectral responses are perturbed by interference of elements. It is suggested to overcome these difficulties using an adapted algorithm of fundamental parameters. The mathematical model based on this algorithm accounts for two events of photon/matter interaction. Element concentrations are estimated using reference samples and the internal standard procedure, both with correction of interference effects. The pore water content is inferred from its correlation with the Rayleigh (coherent)/Compton (incoherent) scatter intensity ratio. Sediment density is found from water content by a reliable sedimentological law. Normalization to Compton scattering accounts for the irradiated mass of wet sediment controlled by porosity and sampling-related core disturbance. The mathematical model also includes these scattering variations related to lithology, water content, and density of samples. The new method was applied to SR XRF scans of cores from Lake Baikal and Lake Teletskoye to predict concentrations of 20 elements (K-Ba, REE, Th, U). Tests against ICP-MS and conventional SR XRF (with traditional pretreatment of samples) show good agreement and stability of SR XRF scanning. Neglect of the disturbing effects may cause errors up to 30-150%