Si, Al, Ti, Fe, Mn, Ca, Mg, P, Ba, and Sr were analyzed. Although this element menu is less extensive than the common igneous suite (Murray et al., 2000), this specific suite was constructed to provide critical information on terrigenous abundance (based on Al and Ti), the oxide metalliferous component (Fe and Mn), and biogenic input (Si, Ca, P, Ba, and Sr). Although additional elements could have been analyzed (e.g., Zr), we limited the protocol to 10 elements to maximize sample throughput. The measured wavelengths for the targeted elements are recorded in Table T1.
Bulk-sediment samples were taken at a frequency of one 2.5-cm3 sample per section of core (approximately every 1.5 m) for Sites 1215-1217. To conserve shipboard argon, used in record quantity during the leg, samples were taken at a frequency of three samples per core (Sections 2, 4, and 6) for Sites 1218-1222. All samples were taken adjacent to physical properties samples to provide the most complete data set for the calculation of chemical accumulation rates that are based in part upon dry bulk density values measured on the physical properties samples.
Bulk samples were prepared according to the method of Murray et al. (2000). Although the procedure was originally developed and optimized for analysis of igneous rocks, we found that the sample masses used (0.1 g), the flux mass (0.4 g), and other aspects of the preparation protocol worked well for the diverse sedimentary lithologies recovered during Leg 199 (see "Calibration Standards"). It will be apparent from the discussions below of in-run, in-site, and long-term precision that procedural differences were not a significant factor in the outcome of the analyses. Refer to Murray et al. (2000) and the "Explanatory Notes" chapter for preparation and procedure.
Compared with ICP-AES analyses of igneous rocks and interstitial waters, sediment analyses require a strict control on analytical drift because of the variability and low concentrations of specific elements in the sediment. A drift solution must be analyzed multiple times throughout an analytical run in order to account and correct for instrumental drift during data reduction. Typical drift should be between ±1% and ±5%, which is ~1% per hour of analyses. The drift solution must be matrix matched for the most accurate and precise results (Murray et al., 2000). This presented a particular challenge for the sedimentary analyses during this leg given the variety of lithologies encountered. We found, however, that overall drift for each element (except P, which is discussed later) was <5%. We used the igneous rock standard K1919 (Kilauea basalt) for drift corrections. Although not a sedimentary rock, we found no matrix-induced effects in using this material. Using this drift solution also helped to assess long-term (interleg) behavior of the ICP-AES instrument, as this is the same drift solution used in igneous rock protocols.
During data reduction, one must account for instrumental drift as measured by the drift solution and background counts as measured by the analytical blank solution as well as differences in sample weight for each analysis. In order to account for differences in sample and acid weight during data reduction, flux and sample must be weighed to a precision of 0.5% of the measured value, which can be accomplished by weighing within a range of 0.3995-0.4005 g for the flux and 0.0995-0.1005 g for the sample. To correct for the weight differences during data reduction, the values are recorded and the weight factor is determined based on a value of one (1.00) being idealized 4000x dilution as indicated by Murray et al. (2000). This value is then multiplied by the drift-corrected counts before being used in the calibration regression.
ICP-AES is a comparative analytical technique. In order to convert the counts as provided by the instrument into concentrations, standard reference materials (SRMs) are prepared in the same manner as the samples and analyzed within each separate run. For Leg 199, SRMs were prepared once at the beginning of the leg in one batch following the procedure outlined above and in Murray et al. (2000). The integrity and stability of the solutions remained strong throughout the cruise, as indicated by the consistent and highly linear calibrations. The drift- and blank-corrected counts were then regressed against the known concentrations for each of the elements for the standards (Fig. F1). SRMs were run as separate samples in each analytical run. This calibration must be done for each element for each analytical run. Although we used an SRM for the drift correction (K1919), we did not use this SRM in the calibration.
The four or five SRMs selected must cover the anticipated high and low concentrations for all of the elements that will be analyzed. Again, the diversity of recovered lithologies of Leg 199 provided a particular challenge, both in terms of covering the range of elemental concentrations and being aware of possible matrix affects. We chose Cody shale (SCo-1), Pahoehoe basalt lava flow (BHVO-2), and marine mud (MAG-1) from the U.S. Geological Survey and argillaceous limestone (NIST-1C) from the U.S. National Institute of Standards and Technology. This particular suite was selected in order to accommodate the chert, radiolarite, carbonate, and clay-rich sediments. We modified the procedures of Murray et al. (2000) to optimize the results specifically for bulk sediments. The concentration of the analytical blanks is considered zero and was used in the calibration regression. An alternative would be to perform a blank subtraction and force the regression through the origin (0,0); however, we found that using the blank as a default zero yielded the best results. Detection limits for each of the elements analyzed are also based on the blank analyses (Table T1).
The strong linearity (e.g., 0.999-1.000) of the calibrations (Fig. F1) indicates that the above suite of chosen SRMs provides an excellent calibration protocol appropriate for the majority of the range of sediment lithologies. This suite of SRMs is likely to be appropriate for other legs, as well. However, for legs that may encounter carbonate-rich lithologies (see "Difficulties and Challenges"), it is recommended that a pure CaCO3 powder be added to the SRM menu.
Precision is measured both short term within a single analytical run, between several runs at one site, and long term from site to site throughout the leg (Table T1). The ICP-AES analyzes each element in a given solution between one and five (this is a parameter set by the analyst) and is most commonly selected to analyze in triplicate. This provides a measurement of precision for each element for each analyzed sample, and is reported as a percent relative standard deviation (%RSD). The %RSD improves over time within a run as the instrument continues to warm up and equilibrate with its surroundings. Typical shipboard %RSD is between 1% and 2%.
Within-run precision is measured by repeated, or replicate, analyses of one particular solution. For the shipboard Leg 199 analysis, the SRM Japanese chert (JCh-1) was used as the replicate solution in order to conserve precious sample, argon, and preparation time (see "Difficulties and Challenges" regarding argon usage). Typical precision values for individual analyses were between 1% and 8% (except for P, as discussed below). For shore-based analytical runs, it is standard practice to prepare two or more identical samples using the same procedure and then analyze them all in one run, which quantifies both precision of the instrument and of the preparation procedure. Because of logistical concerns (e.g., consumption of Ar), we did not use that strategy. We also used the drift solution to calculate the precision between several analytical runs performed at a single site, in addition to the individual run precision. This helps to further constrain the data and detect problems. Drift solution (K1919) was not included in the calibration and was treated as a sample in the run, which allowed us to use its values to determine site precision. Typical site precision values for Leg 199 were between 0.7% and 4% (except for P, as discussed below). Long-term precision is determined by calculating the average and standard deviation (on a per element basis) of all the data on this solution (made from an identical powder) throughout the leg.
As the Al analyses from each site show (Fig. F2), precision does not necessarily improve with increased number of analytical runs. The overall (site) precision of analyses for all elements, except P, is within 0.5%-15% of the measured value, which is quite acceptable given the difficult sample preparation and analytical environment at sea (compared to shore-based precision, which is typically 0.5%-2%). Phosphorous has greater variability in its precision (between 8% and 67% of the measured value) as a result of nitrogen flow; this is discussed in greater detail later in this paper (see "Difficulties and Challenges"). Overall, long-term precision ranges between 3% and 21% for all elements excluding P, which is ~33%.