A key element of ODP Leg 187 was the use of onboard geochemical analysis in support of the primary leg objective—to locate and characterize the boundary between isotopically defined Indian-type and Pacific-type mid-ocean-ridge basalt (MORB) mantle provinces between Australia and Antarctica. First recognized along the Southeast Indian Ridge (SEIR) (Klein et al., 1988), the isotopic boundary is now known to have migrated westward in the last 4-5 m.y., based on analyses of 0- to 7-Ma off-axis dredge samples (Pyle et al., 1992; Christie et al., 1998). Although the boundary is defined by the isotopic signatures of seafloor lavas (assumed to represent the mantle source), examination of SEIR MORB glass data from this region has shown that the great majority of 0- to 7-Ma lavas can be correctly identified as Indian or Pacific type based on variations in Ba and Zr contents (Fig. F7) (Pyle et al., 1995; D.G. Pyle and D.M. Christie, unpubl. data).
From 0 to 4 Ma, the isotopically defined mantle boundary coincides with a distinct morphological change in seafloor fabric. The seafloor to the east (Pacific side) of the boundary is characterized by smooth, ridge-parallel abyssal hills, whereas rough, highly tectonized, chaotic terrain occurs to the west (Indian side) (Christie et al., 1998). The change between smooth and rough seafloor topography is inferred to reflect contrasts in melting conditions and magma supply within a segment. Differentiation trends of lavas dredged from these two types of seafloor terrain are distinct enough to be recognized on simple Mg-oxide variation diagrams (Figs. F8, F9). We therefore have two means of delineating the isotopic boundary: (1) by a mantle source geochemical signature and (2) by compositional indicators of mantle melting conditions that also influence axial morphology and, consequently, seafloor fabric. For shipboard use, we chose the diagram of each type that best discriminated 0- to 7-Ma Indian from Pacific lavas. These are Zr/Ba values vs. Ba content and Na2O/TiO2 vs. MgO.
For every site, we have prepared a geochemical summary that includes
The glass data are then compared with dredged 0- to 7-Ma lavas from the SEIR segment in which the site is located. Figures F8 and F9 show major and trace element MgO variation diagrams for 0- to 7-Ma MORB from the SEIR between 123°E and 133°E. These MORB glass data provide the basis for site-by-site comparison, allowing an evaluation of whether the mantle source and/or melting characteristics within individual segments have changed through time.
Finally, the mantle province for basalts at each site is assessed based on variations in Zr/Ba vs. Ba and Na2O/TiO2 vs. MgO (Fig. F7A, F7B). The basis for using Zr/Ba vs. Ba to discriminate between Pacific-type and Indian-type isotopic mantle provinces is shown in Figure F10A and F10B. Isotopically, 206Pb/204Pb clearly defines a sharp boundary between Indian-type (<18.3 206Pb/204Pb) and Pacific-type (>18.5 206Pb/204Pb) MORB mantle sources along the SEIR at ~126°E (Pyle et al., 1992; Pyle et al., 1995). The Zr/Ba values show a positive correlation with 206Pb/204Pb and separate the Indian type from the Pacific type with little overlap. Most Indian-type MORBs have Zr/Ba values <14, and most Pacific-type MORBs have Zr/Ba values >14. A few samples cannot be distinguished by Zr/Ba values alone; for these, Indian- and Pacific-type MORBs can usually be distinguished, as they have slightly different Ba concentrations for a given Zr/Ba ratio (Fig. F7A).
Although Ba seems an unlikely geochemical discriminant because of its susceptibility to contamination by seawater, particularly in older seafloor basalts, it is the only choice, given the analytical limitations of seagoing instrumentation. We relied on fresh basaltic glass, free of alteration, for this assessment throughout Leg 187 because whole-rock crystalline interiors are clearly prone to seawater circulation and alteration. This can be seen in many of the whole-rock analyses, which tend to have higher Ba contents than associated basaltic glasses. Nevertheless, as data accumulated through the leg, it was apparent that whole-rock data formed coherent trends with their associated glasses at most holes. For Indian samples, these glass-whole-rock trends lie at lower Ba values for a given Zr/Ba than for Pacific samples (see Fig. 23A in the "Site 1156" chapter and the individual site chapters for details).
The data set for each site includes major and trace element data from both standard ODP XRF and newly developed (for ODP) ICP-AES analytical procedures. Elements analyzed by XRF include Si, Ti, Al, Fe, Mn, Mg, Ca, Na, K, P, Nb, Zr, Y, Sr, Rb, Zn, Cu, Ni, Cr, V, and Ce. Elements analyzed by ICP-AES include Si, Ti, Al, Fe, Mn, Mg, Ca, Na, K, P, Zr, Y, Sr, Ba, Ni, Cr, and Sc. The Zn, Cu, and V contents of typical MORBs are well within the analytical capabilities of the ICP-AES, but these trace elements were omitted from the analytical routine to minimize Ar usage, an important consideration at sea. Data tables for each site report include
Unless otherwise indicated, ICP-AES analyses presented in duplicate represent separate analyses of a single sample dissolution. Where duplicate analyses are presented in the tables, the geochemical variation diagrams show only the averaged values.
Our first analytical priority was to quickly determine the probable mantle source (Indian or Pacific type) by analyzing picked, fresh glass in order to facilitate drilling target selection. The ability to accurately analyze Ba and Zr (e.g., ±1 ppm Ba in the range of <5-10 ppm) enabled us to select the next site based on rapid analyses of samples recovered at an occupied site. This was critical to our responsive planning and enabled a much more precise tracing of the mantle boundary than would have been possible otherwise. Our second priority was to characterize the compositional variability of each hole by analysis of representative samples from lithologic units identified in hand sample and thin section.
Representative samples of major lithologic units were selected for shipboard XRF analysis. Large pieces (~20 cm3) were reduced to fragments <1 cm in diameter by crushing between two disks of Delrin plastic in a hydraulic press. The sample was then ground for ~5 min in a Spex 8510 shatterbox with a tungsten carbide barrel. Contamination of the samples with Nb during grinding was investigated prior to the start of Leg 152, and none was detected (Larsen, Saunders, Clift, et al., 1994).
Major elements were analyzed using fused lithium tetraborate glass disks doped with lanthanum oxide as a heavy absorber (Norrish and Hutton, 1969). The disks were prepared from 600 mg of rock powder that had been ignited for 2 hr at about 1025°C and mixed with 7.2 g of dry flux consisting of 80% lithium tetraborate and 20% lanthanum oxide. This mixture, with 20 mL of lithium bromide (8.6 M) added to prevent adhesion to the Pt-Au crucible, was melted in air at 1150°C for ~4 min with constant agitation to ensure thorough mixing and then cooled. Trace elements were analyzed using pressed-powder pellets. These were made by mixing 5 g of rock powder with 30 drops of a solution of Chemplex polymer in methylene chloride (100 mg/cm3) and then pressing the mixture into an aluminum cap under a load of 8 T. We measured loss on ignition from weighed powders heated for 4 hr at 1025°C.
A fully automated wavelength-dispersive ARL 8420 XRF system (equipped with a 3-kW generator and a Rh-anode X-ray tube) was used to determine the major and trace element concentrations in the samples. With the 12:1 flux:sample ratio and the use of the heavy absorber, matrix effects within the fused glass disks are insignificant over the normal range of igneous rock compositions, and the relationship between X-ray intensity and element concentration is linear. A pressed pellet made with 5 g of basalt powder should be infinitely thick at the shortest wavelengths used in the analysis. X-ray intensities were corrected for line overlap and interelement absorption effects. The latter corrections were based on the relationship between mass absorption coefficient and the intensity of the Rh-Ka Compton scatter line (Reynolds, 1963, 1967; Walker, 1973). Analytical conditions are given in Table T3. The spectrometer was calibrated using a suite of 30 well-analyzed reference standards. Standard values recommended by Govindaraju (1989) were used for all elements. Precision estimates, based on replicate shipboard analyses of reference standards AII-92, MGR-1, and BIR-1, are given in Table T4 (major elements) and Table T5 (trace elements).
Major and trace element concentrations of both glass and whole-rock samples were determined with the JY2000 ULTRACE ICP-AES that was installed aboard the JOIDES Resolution immediately prior to Leg 187. ICP-AES protocols were developed by R. Murray (unpubl. data) and optimized by Leg 187 shipboard technical staff and scientists for the rapid dissolution and analysis of basaltic glass chips and whole-rock powders. The ICP-AES methods and procedures that we tested and adopted are described below in detail. The ICP-AES offers several advantages over routine XRF analysis in that
Most samples analyzed by XRF during Leg 187 were duplicated by ICP-AES as a quality control check because the JY2000 ICP-AES is a new seagoing analytical tool. Overall, the major and trace element data agree within the error of each analytical method.
Because the ICP-AES was a new analytical instrument, considerable technique development occurred in the early weeks of Leg 187. During this time, samples were analyzed under a variety of instrument conditions. The specific analytical conditions for each sample run are provided in Table T6. ICP-AES analyses of samples from the first three sites are of questionable quality; glass samples were reanalyzed later in the leg. Sample runs containing data of questionable quality are noted below and in the site chapters.
The JY2000 is a sequential atomic emission spectrometer that measures the intensities of characteristic emission wavelengths between ~100 and ~800 nm, one peak at a time. Therefore, analysis time depends on the number of elements determined, the number of emission lines per element, the mode of analysis, the number of replicates, and the counting time. The 18 elements analyzed during Leg 187 and the emission lines used are listed in Table T7. All ICP-AES data presented in the site chapter reports were analyzed using Mode 5 (see "Mode of Analysis") of the JY software. With this analytical mode, the intensity at the peak of an emission line is measured and averaged over a 1-s (2 s for Ni, Cr, Na, and K) counting interval repeated three times (see "Mode of Analysis"). Each sample solution was repeated at least once during a single run, except in sample runs Leg187B and Leg187C. Major and trace elements were measured during a single run using the autoattenuation feature of the JY2000 software, which adjusts the photomultiplier voltage to optimize the peak to background ratios for each emission line.
Our analytical and sample preparation techniques were significantly improved by the time we arrived at Site 1155 (sample run Leg187D; Table T6), resulting in greater plasma stability, thus greatly reducing signal noise. Both V-groove and concentric nebulizers were tested. Filtering of solutions and higher sample dilutions are required for the concentric nebulizer. We preferred the concentric nebulizer since it delivers a finer, more stable aerosol to the plasma, resulting in a more stable signal (see Table T6). Greater sample dilution also reduces salt buildup on the torch glassware.
The mechanical step position of each emission line was initially calibrated using 10-ppm single-element standard solutions prepared at Boston University. We did not have a single-element Sc solution aboard, so we calibrated the emission peak closest to the theoretical step position of Sc (361.84 nm) using a solution of the Lamont-Doherty Earth Observatory rock standard K-1919. Multielement standards approximating typical basalt concentrations were prepared in 10% HNO3 and used during the first 2 weeks of the leg to check the initial-emission-line calibration by autosearching a small (0.002 nm) window across each emission peak. A second tuning of emission line positions was then performed by again autosearching a small (0.002 nm) window using the K-1919 basalt-rock standard. A peak profile was collected for each emission line during the initial setup by using the K-1919 standard to determine peak-to-background intensities and to set the locations of background points for each element. To optimize peak to background conditions, the photomultiplier was set for each element by autoattenuating on the K-1919 standard.
At sites where rapid decisions were needed, the first recovery of basaltic glass was sampled for immediate processing. Additional samples were selected based on availability of glass, position in the core, and number of units defined. Glass samples were chipped from whole-rock pieces by hammering the sample while it was wrapped in plastic-coated freezer paper or, alternatively, by chiseling. The resulting mixed whole-rock and glass-chip product was then carefully crushed in an alumina mortar and pestle, sieved to a 1- to 2-mm fraction, washed and insonified in nanopure H2O for 20-40 min (depending on the extent of palagonite alteration), and dried in an oven at 80°-100°C. Fresh shiny black basalt glass was then hand-separated from all altered material, minerals, whole-rock chips, and spherulitic microcrystalline material, under binocular magnification.
We weighed 0.100 ± 0.002 g of glass chips or whole-rock powder (prepared for XRF) and mixed it with 0.400 ± 0.0004 g of Li-metaborate (LiBO2) flux that was preweighed onshore. Standard rock powders and full procedural blanks were included with the glass and whole-rock unknowns for each sample run. For Sites 1152 and 1153, standard powders were preignited at 1025°C for 4 hr before weighing and fusing. After Site 1153 (sample run Leg187C) only unignited standards were used since glass and whole-rock unknowns were never preignited prior to weighing for ICP-AES analysis. Samples and standards were weighed on a Cahn automatic electrobalance attached to a gimballed table. Weighing errors at a 99.5% confidence are conservatively estimated to be ~±0.0001 g. Weighing a single sample can take as long as 15 min on station in moderate seas, even though the low inertial forces applied to the balance by small sample weights help to speed the weighing process. This and handpicking of difficult glass samples were the time-limiting steps in the analytical process. Minimum sample processing time for ICP-AES (from on deck to dissolution) is ~5-12 hr.
Rock powder/flux mixtures were fused in Pt crucibles heated by the same induction furnace apparatus used for XRF glass-bead preparation. A lithium bromide wetting agent was used to prevent the cooled bead from sticking to the crucible. Cooled beads were transferred to 60-mL wide-mouth Nalgene polypropylene bottles and dissolved in 50 mL of 5% HNO3 by shaking with a Burrell Wrist Action bottle shaker for ~1.5 hr. After digestion of the glass bead, 10 mL of this solution was passed through a 0.45-µm filter and diluted with 30 mL of 5% HNO3. Each sample and standard solution was spiked with 5 ppm Ge as an instrument drift monitor. The final solution-to-sample dilution factor for this procedure is ~2000.
The JY2000 plasma was ignited 30 min before each run to warm up and stabilize the instrument. After the warm-up period, the following steps are required: (1) a zero-order search required by the software is carried out to check the mechanical zero of the diffraction grating; (2) the mechanical step positions of the emission lines are tuned by autosearching a small (0.002 nm) window across each emission peak using a K-1919 standard solution; (3) autoattenuation is performed using the K-1919 standard if there have been significant changes in operating conditions, such as changing the nebulizer.
A typical sample run (see "Sample Run Format and Data Reduction") lasts from 4 to 6 hr, depending on the number of samples and the number of sample repeats. Each run used 50%-75% of a high-pressure Ar bottle.
A typical ICP-AES run consisted of
Instrument stability and short-term drift was monitored with the ~5-ppm Ge internal spike added to each solution. Lithium is also useful as an instrument drift monitor since each sample was fused with approximately the same amount of LiBO2 (e.g., Ramsey et al., 1995).
Following each sample run, the raw intensities were transferred to a data file, and data reduction was completed by spreadsheet to assure control over standardization and drift correction (we did not test the instrument's data reduction software): (1) intensities for all samples were corrected by subtracting the procedural blank; (2) drift correction was accomplished by interpolating between two consecutive drift-correcting solutions and normalizing the intensities of the intervening samples; (3) normalized intensities were corrected for sample weight; and (4) concentrations were determined using calibration curves created from replicate measurements of the USGS and in-house standards (Table T8).
Estimates of accuracy and precision for major and trace element analyses are based on replicate analysis of the USGS reference standard, BHVO-2, and the Zone A SEIR whole-rock basalt sample MW8801-17-26 (Tables T9, T10). In general, replicate analyses of BHVO-2 are accurate to within 3% for major elements and 5% for trace elements. For most elements, shipboard analyses of BHVO-2 by ICP-AES are slightly lower than published values but well within the analytical error reported by Plumlee (1998a, 1998b) (Tables T9, T10). Within-run precision of replicates improved greatly during the course of Leg 187, to much better than 3% for most elements. Run-to-run precision was <3% for the major elements, with the exception of K, Na, and P. Na and K have poorer precision because of the instability of the sheath gas flow (see "Plasma Instability"). Run-to-run precision for trace elements was <5%, depending in part on the concentration of the element in the sample. For example, compare the relative standard deviation (RSD) and Ba values for MW8801-17-26 and BHVO-2 in Table T10.
In general, the ICP-AES and XRF analyses compare very well (see the "Geochemistry" sections in the individual site chapters), except for Ni and Cr, which were consistently lower in the ICP-AES analyses. XRF Cr results are consistently high relative to published standards values, and Ni tends to be low. The ICP-AES Ni and Cr data are problematic as well (see Fig. F9).
Before developing the successful technique described above, several problems compromised the quality of the data from early sample runs.
The concentration of the digestion acid was based on tests with 1%, 2%, 5%, and 10% HNO3. We determined that a cooled, fused bead dissolved in 50 mL of 5% HNO3 in a heated ultrasonic bath within ~60 min. Sample solutions were not diluted further for our initial ICP-AES runs because many trace elements were at or near their detection limits. We encountered significant problems with Si gel formation when dissolving samples in bottles in a heated ultrasonic bath. The gel problem was clearly observed by using clear 50-mL polypropylene Corning centrifuge tubes. These tubes are highly recommended because they allow a visual check on undissolved particulates and/or the formation of Si gel. Si gel formation was the dominant cause of our initial poor results, affecting runs Leg187A and B. Gel formation is promoted if a compositional gradient is allowed to develop in the dissolution vessel; this is a severe problem with stationary bottles in an ultrasonic bath but is readily avoided by using a mechanical shaker. We used a shaker for the remainder of Leg 187 with no further gel-formation problems.
By sample run Leg187D, we determined that plasma stability and analytical accuracy improved if samples were filtered and significantly more diluted, at least ~2000 times. The additional dilution did not compromise detection limits as it allowed the use of the concentric nebulizer, increasing the efficiency of sample aspiration.
Five modes of analysis are available in the JY software. During the first week of Leg 187, the total analysis time and counting statistics of three of these modes were briefly compared: Mode 1 measures intensities at a variable number of points (3-11) across a peak and then averages the intensities of a subset (1-9) of those points with the largest intensities. For example, Mode 1 will average the intensities of the three highest contiguous "points" within a set window that can include as many as 11 preset positions across a peak. Mode 2 fits a Gaussian curve to a user-determined number of points across a peak and then integrates to determine the area under the curve. Mode 5 determines a single peak intensity at a preset wavelength position.
In selecting an analytical mode, the user must balance improvements in signal reproducibility against increased analysis time. The latter is an important consideration as Ar supplies are limited aboard the JOIDES Resolution. Modes 1 and 2 theoretically compensate for instability in the peak position during a run that might result from the constant movement and vibration of the ship. These two modes require significantly greater counting times and consequently use more Ar. Analysis by Mode 5 is quicker by a factor of ~3 but may be more susceptible to signal instability if the peak position is not relocated exactly by the spectrometer throughout the entire run. We adopted Mode 5 for all Leg 187 ICP-AES analyses because of our need to limit Ar consumption. Because of the Ar limitation, we were unable to fully evaluate the other modes.
Before Site 1155 (sample run Leg187D), plasma instability led to signal noise and significantly reduced data quality. By increasing the sample dilution factor, filtering sample solutions, and replacing the V groove with the concentric nebulizer, we significantly improved plasma stability. Improper operation of the Ar humidifier may also have contributed to plasma instability and signal noise. Humidified Ar is used to inhibit the precipitation of salts around the capillary orifice within the nebulizer. We observed, however, that condensed water from the humidifier would fill the Ar flow line at the connection to the nebulizer after relatively short periods (<1 hr) of instrument operation. The Ar stream would then bubble through the water into the nebulizer, causing erratic sample aspiration, resulting in random negative spikes in the emission signal.
Following sample run Leg 187D the Ar humidifier was turned off, and there were no further significant negative spikes in the emission signal. With the Ar humidifier off, however, salt buildup around the capillary orifice within the nebulizer gradually decreased the Ar flow throughout each run (see the "Nebulizer flow" values for sample run Leg187E in Table T6). The salt buildup does not entirely block the capillary orifice, and the flow of sample solution through the nebulizer remains constant. The combined effect of decreasing Ar flow and constant sample flow results in a steady increase in peak and background intensities. As this drift is steady and smooth through the sample run, we accounted for it by applying a linear drift correction (see "Sample Run Format and Data Reduction").
Another source of plasma instability specific to the alkali elements Na and K and the JY instrument results from variability in the sheath gas flow. The light emission from these elements originates lower within the plasma; the JY2000 compensates by increasing the Ar gas flow through the torch, which effectively lifts the plasma so that the Na and K emissions are positioned at an optimal viewing height. We found that the sheath Ar gas flow in this instrument is subject to both short- and long-term variations. These variations in Ar gas flow cause short-term noise and long-term drift in Na and K signals. Decreases in the measured intensity for Na and K ranged up to 80% in sample run Leg187E. Several times during Leg 187 the sheath gas flow had to be adjusted. The ideal Ar gas flow of 0.8 L/min was difficult to obtain because the adjustment screw is very sensitive. In general, percent RSD and signal strength improved greatly after the sheath gas flow was adjusted to an ideal rate of 0.8 L/min. The source of both short-and long-term variations of the sheath gas flow is suspected to be related to a faulty Ar gas flow valve.