Previous   |    Next

doi:10.2204/iodp.proc.314315316.201.2011

Results

Tabulation of the results for standard mineral mixtures appears in Table T1. Regression analysis shows that a power function yields the best correlation coefficient (r) for the ratio of PA to TA versus the known abundance of the mineral (Fig. F4). Linear regression, in comparison, shifts the intercept away from the origin, which is nonsensical. Two equations were constructed. For kaolinite-dominant specimens (kaolinite wt% = 100 × (PA/TA)^1.613, where r = 0.9870) and for chlorite-dominant specimens (chlorite wt% = 100 × (PA/TA)^1.433, where r = 0.9974). This pair of equations allows researchers a choice depending on which of the two minerals is dominant in the natural mix.

Table T1 also compares the calculated mineral abundances using three approaches (Fig. F5). The doubled half-peak area method yielded the best agreement between the calculated abundance and the actual abundance (Fig. F5). The maximum error is 7.7%, and the average error is only 2.6%. Using the intensity-ratio method, the average error increases to 7.4%, but those results become more accurate with well-balanced proportions (chlorite to kaolinite ratio close to 50:50). Using the fitted-area-ratio method, the average error is 5.1%, and those results are more reliable when kaolinite contents are between 20% and 80%. The maximum errors increase to 17.6% and 12.3%, respectively, using intensity-ratio and fitted-area-ratio (Table T1). Thus, for use with a broad range of potential mixtures in natural samples, the doubled half-peak area method is judged to be the most accurate.

Quantitative accuracy of XRD results is known to be affected by a long list of natural variables and laboratory artifacts, including the type of diffractometer, sample disaggregation technique, chemical pretreatment, particle size separation, size and shape distribution within a selected size fraction, chemical composition of clay minerals (e.g., content of iron in chlorite), structural ordering and crystallinity, the degree of preferred orientation of crystallites on the scanned surface, and peak-fitting algorithms (Reynolds, 1989; Moore and Reynolds, 1997; Ottner et al., 2000; Środoń, 2002). Besides those variables, the thickness of aggregates on a glass slide will have some effect on the peak intensities and peak areas. This inconsistency in thickness helps explain the scatter of our results across similar ranges weight percentages for a given mineral (Fig. F3). Finally, we assumed ideal Pearson II peak shapes when computing the half-peak areas, but the instrumental peak shape of a Bragg-Brentano diffractometer is usually somewhat asymmetric.

Unquestionably, the two CMS source clays KGa-1 (kaolinite) and CCa-2 (chlorite) are not representative of all possible clay mineral assemblages in natural marine sediments. The refinement probably could be improved even more by analyzing additional mixtures with a broader range of chemical compositions and crystallinities, particularly for chlorite, but then extra attention would be required to match by iteration each natural assemblage with its best-fit standard mixture. Errors can also increase in a more unpredictable way if the natural sediment contains minerals in addition to kaolinite and chlorite whose reflections interfere with the 3.5 Å double peaks (Biscaye, 1964). Although imperfect, we have demonstrated that the doubled half-peak area approach is consistently more accurate than estimates from peak-intensity ratio and fitted-peak-area ratio.

Top of page   |    Previous   |    Next