Development of the Monte Carlo-Library Least-Squares approach for energy dispersive X-ray fluorescence analysis

A new analysis principle called Monte Carlo-Library Least-Squares (MCLLS) approach has been identified and investigated for quantitative energy dispersive X-ray fluorescence analysis. This new approach consists of: (1) generating the complete spectral response for a sample of known (assumed) composition by Monte Carlo simulation, (2) keeping track of the individual elemental response within the Monte Carlo simulation for use as library spectra, (3) use of the linear library least-squares spectrum, and (4) iterating these steps if the unknown amounts are too far from the assumed composition originally used.;In order to apply the MCLLS approach for quantitative energy dispersive X-ray analysis, a general geometry Monte Carlo simulation code (called NCSXRF) for simulating EDXRF analytical systems and an improved Si(Li) detector response function to convert Monte Carlo simulated X-ray spectra into pulse-height spectra has been developed. Compared to a previous EDXRF Monte Carlo simulation code, the NCSXRF code has the following improvements: (1) extended simulation ability to general geometry problems by writing the code to include an existing general geometry subroutine HERMETOR; (2) simulation of the entire EDXRF system including collimator, source shielding, sample and detector shielding etc.; (3) use of the correlated sampling variance reduction technique to make the code more efficient; (4) user options which allow either more accuracy and completeness or greater speed with some approximation; and (5) code optimization with respect to run-time, memory and round-off error so that the code can be efficiently run on several different types of computers. Compared to the previous Si(Li) detector response function the newly developed response function has the two improvements: (1) extended response function energy range from 5-20 Kev to 5-60 Kev; (2) elimination of the discrepancy in parameter values for K;The feasibility of the MCLLS principle has been investigated with a radioisotope source excited EDXRF analytical system for several standard samples of known composition. With good initial elemental composition guesses, the MCLLS results showed excellent conformance to elemental compositions in the standard samples. Furthermore, by taking the ratio of the photoelectric peak counts in the sample experimental spectra as approximate initial guesses, the proposed iterative MCLLS algorithm works quite well converging to the correct elemental compositions. Only one iteration was needed for major element analysis. For trace elements a second iteration was needed in some cases. Other possible applications of the MCLLS approach and future work are also discussed.