Previous   |    Next

doi:10.2204/iodp.proc.320321.216.2014

Methods and materials

The spliced sediment columns from Sites U1335–U1337 were scanned using the third-generation Avaatech XRF scanner at the IODP Gulf Coast Repository in College Station, Texas (USA) (odases.tamu.edu/research-facilities/xrf-request/). This XRF scanner is equipped with a Canberra X-PIPS SDD, model SXD 15C-150-500 150 eV resolution X-ray detector and uses an X-ray tube with a Rh-anode to generate the incident X-rays that illuminate the sample. The XRF scanner is configured to analyze split sediment core halves for elements between Al and U on the periodic table. The X-ray tube and detector are mounted on a moving track so that multiple spots at different depths can be analyzed on a split core during the scanning run, and multiple scans with different settings can be programmed to run automatically (Richter et al., 2006). There are, however, parameters controlled by the operator, such as X-ray tube current, voltage, measurement time (live time), filters, and area of X-ray illumination. The downcore position step is precise to 0.1 mm.

XRF scanning

Sample spacing along each of the Site U1335 and U1336 core sections was set at 2.5 cm. For Site U1337, sample spacing was set at 2.5 cm in the lower spliced section and 5 cm in the 0–110 mcd section overlapping with K. Iijima’s (unpubl. data) work. Two separate scans at different voltages were used on each section to gather all chemical data. The first scan was performed at 10 kV with no filter for Al, Si, S, Cl, K, Ca, Ti, Mn, and Fe. The measurement (live) time was set at 30 s, and the X-ray tube current was set to at 500 µA. A second scan for Ba was performed at 50 kV with a Cu filter, a measurement time of 10 s, and the X-ray tube current set to 1000 µA. We lowered the X-ray tube current from 2000 µA used for the Site U1338 scanning (Lyle et al., 2012) in order to preserve tube life and reduce the possibility of peak overlap problems. The core face illuminated by X-rays was set at 1.0 cm in the downcore direction and 1.2 cm in the cross-core direction.

For the Site U1335 and U1337 cores, the scans were run down the center of the split core half (6.8 cm total diameter). For the Site U1336 cores, the majority of the scans were run down the left side of the split core half to avoid the U-channel gap left by sampling for magnetostratigraphic analyses on the archive core prior to the XRF scanner runs. On the U-channeled sections, the track in the XRF core scanner was offset ~3 cm from the core’s centerline. Additionally, cracks and holes from sampling in any of the cores were skipped if they corresponded to XRF sample spacing.

Each core was removed from refrigeration a minimum of 8 h prior to scanning to allow it to warm to room temperature. Each section was then covered ~15 min prior to scanning with 4 µm thick Ultralene plastic film (SPEX Centriprep, Inc.) to protect the detector face from becoming contaminated during the scan. Following the methods in Lyle et al. (2012), the Ultralene film was placed over the core sections only once they had warmed to room temperature in order to avoid condensation building up on the film while the core sections warmed. Condensation buildup can lead to severely reduced light element (e.g., Al and Si) XRF peak areas because of the condensation absorbing the low-energy X-ray fluorescence coming back from the sediment to the detector (Tjallingii et al., 2007).

There were a few unexplained data dropouts that we found after plotting the data, one and occasionally two data points that were significantly lower than the data trends of the scan. Where we had overlapping data, we found these “events” on only one of the overlapping sections. We attribute the dropouts to an intermittent instrument failure and removed them from the data sets, including those shown in Tables T1, T2, and T3. We found 15 bad data points between 109 and 114 mcd in the Site U1336 core sections. Of the 15 bad data points found in the Site U1336 splice, 7 were outside of the splice, meaning they were in the overlapped section and would not have been included in the final splice even if they were kept, and 8 were within the splice, meaning that they would have been included in the final splice. All 15 bad data points were removed. Two additional bad data points were found in the Site U1336 splice at ~32 and ~83 mcd. Both of these points were removed as well.

Sediment splices

XRF scanning was done to cores along the sediment splices, not to all sections recovered from the different holes at each site. Only core sections along the continuous spliced sections of Sites U1335 and U1336 were analyzed. We scanned every archive core half included on the Site U1336 and U1337 splice tables in 2011 and 2012 and completed the scanning of the Site U1335 spliced sections in January 2013. If the splice transferred from one hole to the next in the middle of a section, both sections were run in their entirety to provide enough overlap to allow for the record to be adjusted, if needed. In the Site U1337 splice, substantial changes have been made to the original continuous splice table, so additional sections were run to aid in revising that record.

Tables T1, T2, and T3 contain all of the data collected, not just the data that were used to construct the final splice (also see XRF in “Supplementary material”). Data incorporated into the splice are marked for easy extraction. Each table includes raw XRF peak area data and NMS reduced data for all three cores. Depths are indicated in meters below seafloor (mbsf), which is equivalent to core depth below seafloor (CSF-A), as well as converted to mcd, which is equivalent to CCSF-A. References to depth throughout this manuscript are in mcd. Medians of shipboard geochemical data and calculated area medians, both used in scaling and normalizing the raw XRF area data, are presented in Table T4. For each element in Tables T1, T2, and T3, we have listed the raw XRF peak area, the median-scaled total, and the NMS data as described below.

XRF data reduction

Data reduction was accomplished by a three-step method generally following the methods in Lyle et al. (2012). In the first step, which is an addition to the two-step reduction method developed in Lyle et al. (2012), negative peak area values were set to zero and removed from further calculations. Negative peak areas can appear where there are overlapping peaks and the elemental concentration is low. In this case, the measured data are actually below the modeled energy curve at 0 concentration.

Second, data were scaled by the median shipboard-measured bulk sediment elemental composition in order to scale elemental peak areas into typical ranges of sediment composition. Finally, the scaled components were summed and normalized to 100% in order to account for changes in porosity and cracks throughout the core sections. Only sections used in the final splices were used in calculating the area medians, so changes to the final splices may require recalculation of those medians and therefore NMS values.

To scale the XRF data, the following equation from Lyle et al. (2012), was used:

where Se is the elemental scaling, Med%e is the median weight percent of a sedimentary component, which was the oxide for each element (e.g., for Al, we used the oxide Al₂O₃, and for Ca, CaCO₃) from shipboard geochemical data, PeakAreae is the measured elemental peak area in the sample, and PeakAreae,med is the median peak area over the data set.

To normalize the scaled XRF data, thereby removing the volume effect, each component (C) was multiplied by 100/(raw sum) to bring the total sum of components to 100%, rather than the lower value that results from high porosity or cracks:

where NMSc is the normalized median-scaled value for the component and C is the median-scaled value of the component. The sedimentary components in these data sets are Al₂O₃, SiO₂, K₂O, CaCO₃, TiO₂, MnO, Fe₂O₃, and BaSO₄. Normalization minimizes the high-frequency noise produced from greater porosity in the shallower portions of the splice and increased cracks and uneven surfaces in the lower sections of the splice.

Through the scaling and normalization process, we developed a way to quantitatively estimate sediment concentration based on XRF scans. XRF NMS estimates, however, can have significant errors if the model of sediment composition (the “type” of sediment composition) used is very different from the actual sediment composition. In other words, errors will result if the model sediment components do not match those within the actual sediment (e.g., if Ca is found in a clay and thus would be better represented by a CaO component) or if a major element found in the sediment is not included in the model. Despite these issues, NMS data are significantly better to study the changes in sediment composition than the raw XRF peak area because the porosity effect is removed. Raw peak area data can have larger relative errors because of the differences in porosity between the sediment layers and from technical problems landing the detector on a flat sediment surface. Very dry and cracked sediment can cause the detector to not land flat against the sediment while collecting peak area data, or to have small cracks included in the scan area, lowering the raw counts. Figure F5 shows the raw median-scaled CaCO₃, the NMS CaCO₃, and discrete CaCO₃ for Sites U1335–U1337. Scaling and normalization reduced apparent noise in the XRF data and made the total range more similar to the variability seen in the shipboard data.

The NMS data can be easily calibrated to estimated sediment composition by analyzing a small set of check data that spans the typical range of sediment values and correlating the NMS data to the discrete measurements (Lyle and Backman, 2013).

Top of page   |    Previous   |    Next