Previous   |    Next

doi:10.2204/iodp.proc.302.203.2009

Methods and materials

Sediment samples were frozen, freeze-dried to remove water, and ground by hand with an agate mortar and pestle (see the “Sites M0001–M0004” chapter) or agate mills (new ZEKAM, AWI, and ICBM Oldenburg sample sets) (see Stein et al., 2006; Sluijs et al., 2008). A few onshore party samples were replicated in the new sample sets. No significant differences in X-ray diffraction (XRD) data from differently ground samples were found. The high number of 1570 investigated samples was only possible through cooperation with R. Stein (Alfred Wegener Institute Bremerhaven, AWI) and H.-J. Brumsack (Institute for Chemistry and Biology of the Sea, University of Oldenburg, ICBM), who provided their Paleogene samples of the ACEX cores (basically between 180 and 430 meters composite depth [mcd]). The upper 60 mcd of the ACEX cores was sampled by the author.

X-ray diffraction

Instrument parameters

All 1570 XRD measurements were performed at the Crystallography Department of Geosciences, University of Bremen, on a Philips X’Pert Pro MD X-ray diffractometer equipped with a Cu tube (Kα, λ 1.541), a fixed divergence slit (¼°2θ), a 15-sample changer, a secondary monochromator, and the X’Celerator detector system. Samples were prepared with the standardized Philips/Panalytical backloading system, which provides nearly random distribution of the particles. Measurements were made from 3° to 85°2θ with a calculated step size of 0.016°2θ. The calculated time per step was 100 s. The 1570 samples needed ~1700 h measuring time on our XRD instrument. This time would increase to at least ~5100 h (~212 days/24 h running) on a standard XRD instrument with scintillation detector, and thus consume very much of the mean lifetime of a standard X-ray tube.

Peak identification was done graphically through the Apple MacIntosh program MacDiff (version 4.5)
(servermac.geologie.uni-frankfurt.de/​Staff/​Homepages/​Petschick/​RainerE.html) (Petschick et al., 1996). To minimize subjective influences, the baseline has been determined automatically with MacDiff defaults. The program also provides a mechanism to automatically estimate the integrated peak area intensities with given defaults (for example, by deconvolution of nearby peaks), as many other XRD data programs do. I do not recommend these “black boxes” for any kind of single peak-based investigations, as natural mixtures of mixed elemental silicates like clay minerals, feldspars, and so on, are too variable to be accurately identified by these automatic routines. In particular, the peak position and ranges of a certain peak or group of peaks are too variable in different samples even from the same region or core (for example, between the Neogene and the Paleogene section of the ACEX cores). This means that the recognition of peaks and the calculation of integrated peak areas is a matter of a certain amount of subjectivity, though that can be minimized, and that is why I emphasize here that all 1570 samples have been investigated by one person.

Mineral identification and semiquantification

Integrated peak area intensities for the 38 investigated mineral peaks were calculated by MacDiff. Based on these areas, ratios were calculated versus each other and versus the sum of all peak area intensities. To provide an easy comparison to published data on surface samples of the potential source regions (Andersen et al., 1996; Vogt, 1997; Vogt et al., 2001), the fixed divergence was changed to automatic divergence using an algorithm integrated in MacDiff. This point is important to note, as a fixed divergence slit leads to a stronger radiation of the lowest angles (Fischer, 1996). If single peaks are investigated at these low angles, the higher radiation leads to higher peaks and to higher content percentages (see Krylov et al., 2008, for a comparison of fixed divergence and automatic divergence slit measurements and the content of the smectite group of the clay fraction). Whereas mineral content trends stay the same over the core, absolute values differ, and a comparison of these absolute numbers between an XRD data set measured with fixed and another measured with automatic divergence slit would lead to difficulties in comparing results.

I also did not multiply the peak area data by any kind of factor (e.g., Schultz, 1964; Griffin, 1971; Ramm, 1991), as (1) the chemistry and mineralogy of Arctic Ocean sediments are different from the rocks investigated by these authors and the factors might be completely wrong, (2) a comparative work similar to Schultz (1964) for Arctic Ocean source region materials is pending (see Forsberg et al., 1999), and (3) the use of different factors in various publications leads to content percentages that cannot be compared to each other at all. In general, single peak data produce the highest quantification error, as can be seen by interlaboratory comparisons (cf. Omotoso et al., 2006, and references therein).

The latest point is illustrated by Figure F3. Vogt (1997) used the calculation factors of the four different popular ways of calculating the clay mineral assemblage composition in Arctic Ocean sediments. Massive differences appear, particularly in the content of kaolinite and chlorite, because the peak intensities selected for chlorite depend directly on the Fe content in the chlorites (illustrated by the peak intensity ratio 4.72/3.54 Å in Fig. F3) (see Brown and Brindley, 1980). At this point even the trends in clay mineral group contents differ. Comparison of such different initial data could lead to a misinterpretation, particularly in overview articles that do not check for the initial semiquantification technique (e.g., Dethleff, 2005).

Absolute intensities also depend on the general XRD instrument configuration, the radiation source, and the sample preparation methodology. Therefore, d-values and ratios of intensities for the investigated peaks versus the total intensity of the investigated samples are given in the results table in the PANGAEA/WDC Mare database (doi.pangaea.de/​10.1594/​PANGAEA.705057). I propose that these ratios are easier to compare between different XRD instruments. I recommend Kahle et al. (2002) along with Moore and Reynolds (1997) for further reading about the use of single peak intensities for semiquantification.

In our laboratory we also use the Philips/Panalytical software X’Pert HighScore for fast identification of mineral phases. Based on a vast reference database, the software also gives a semiquantitative estimate for each identified mineral on the basis of the relative intensity ratio (R.I.R.) values. These R.I.R. values are calculated as intensity ratio of the most intense reflex of a specific mineral phase to the intensity of the most intense reflex of pure corundum (I/I_c) referring to the “matrix-flushing method” after Chung (1974). Therefore, the availability of suitable references in the database with correct R.I.R. is crucial. Hence, only two examples are shown here (Figs. F4, F5), although most of the samples have already been investigated with HighScore.

Top of page   |    Previous   |    Next