Precision in estimating mineral proportions from chemical analyses

Abstract

The precision of estimating rock mineralogy by CIPW norm calculation and the least-squares analysis method were examined in this study. Five hundred granitoid rocks having varying compositions of quartz, K-feldspar, plagioclase, hornblende, and biotite were randomly generated by a computer program. Random errors of up to 5% were added to both rock and mineral compositions to determine the applicability of these two methods. Results indicate that the least-squares analysis method gave better-estimated rock mineralogy than did the CIPW norm calculations. Precision of estimating mineral proportions are 6.5% for quartz, 7.5% for K-feldspar, and 8.1% for plagioclase in the least-squares analysis, but for the CIPW norm calculation they are 14.0% for quartz, 13.1% for K-feldspar, and 17.4% for plagioclase. Correlation coefficients of true mineral mass proportion versus estimated mineral mass proportion range from 0.995 to 0.997 in the least-squares analysis, whereas they range from 0.896 to 0.936 in the norm calculation. These results show that the least-squares analysis method consistently yields better estimates of mineral mass proportions from chemical analyses.