DEVELOPMENT OF THE MONTE-CARLO LIBRARY LEAST-SQUARES METHOD OF ANALYSIS FOR NEUTRON-CAPTURE PROMPT GAMMA-RAY ANALYZERS

A new analysis principle called the Monte Carlo-Library Least-Squares (MCLLS) principle has been identified and investigated in a preliminar y way for neutron capture prompt gamma-ray (NCPGR) analyzers. In order to apply the MCLLS method to obtain sample elemental concentrations, the elemental library spectra of the sample elements are required. To accomplish this, an existing Monte Carlo model which simulates the ana lysis of coal on a conveyor belt using Cf-252 and a Ge detector has be en modified to predict these elemental library spectra. This Monte Car lo model includes several very efficient variance reduction techniques such as: (1) use of either expected value technique or discrete impor tance function to determine the interaction type, (2) forcing all prom pt gamma-rays to be emitted after an (n, gamma) event has occurred, (3 ) use of the direction biasing technique to bias the emitting or scatt ering direction of primary and secondary photons towards the detector, (4) use of both deterministic estimator and statistical estimator to compute the unscattered detection probabilities of the primary photons and the subsequent secondary photons, (5) use of the correlated sampl ing method in order to study the sensitivity of the library spectrum s hape to sample composition, and (6) use of the detailed detector respo nse function to transform the detected photon spectra to pulse-height spectra. The pulse-height spectra for each element in the sample is st ored separately to provide the library spectrum for each element. By s tudying with the simulated spectra, the MCLLS method appears to be fea sible for a good initial guess of the sample composition; furthermore, the proposed iterative MCLLS algorithm works for poor initial guesses of the major element amounts. Testing of the proposed algorithm on th e experimental spectrum showed that for the practical application of t he iterative MCLLS method certain guidelines should be used. These are provided in the paper. The major problems with the application of MCL LS and their possible solutions are also discussed.