摘要

Partial least squares, principal component regression and support vector machine multivariate methods were used for overlap correction of sodium-zinc (Na(K alpha)-Zn(L alpha)) spectral lines generated by means of wavelength dispersion X-ray fluorescence (WDXRF) combined with standard-less software (IQ+) technique for the analyses of mineral samples. This methodology uses one scan channel using PX1 analyzer crystal, 550-mu m collimator, flow detector (Ar + CH4), and rhodium (Rh) tube for determination of Na and Zn in mineral compositions in minimum time. The calibration matrix was made up of 35 samples containing different amounts of Na2O and ZnO. The considered concentration ranges were 0-5% for both Na2O and ZnO. The values for 2 theta angle were recorded between 25 degrees and 29.9 degrees at every 0.1 degrees. Variable tube powers (kV - mA) were used to investigate the effect of tube power on the analyses of elements. The validation of the multivariate methods was realized by analyzing soil samples. Atomic absorption and flame photometry methods were used as reference methods for analyzing Zn and Na in the soil samples, respectively. The results of using chemometric methods, WDXRF (standard-less software) and reference method determined partial least squares and support vector machine models obtained more acceptable results for Na2O in presence of ZnO than those of WDXRF (standard-less software).

  • 出版日期2014-5

全文