LITHOSTRATIGRAPHY

We followed normal ODP procedures for recording sedimentologic information on visual core description (VCD) forms on a section-by-section basis (Mazzullo and Graham, 1988). VCD data were transferred to core-scale "barrel sheets" using AppleCORE software (Shipboard Scientific Party, 1997). Textural subdivisions for siliciclastic sediments and the classification scheme for siliciclastic lithologies follow Mazzullo et al. (1988). The "Graphic Lithology" column on each barrel sheet shows to scale all intervals that are at least 10 cm thick. Combined graphic patterns are used to indicate interbeds less than 10 cm thick. Differences between the silty clay and clayey silt lithologies can be subtle and not easily separated without further analysis, so we grouped this entire range of sediment textures into the category "clayey silt" on all barrel sheets. Figure F2 displays the graphic patterns for all lithologies encountered during Leg 190. Separate patterns have not been used for more heavily indurated examples of the same lithologies (e.g., sand vs. sandstone). Also shown are symbols for internal sedimentary structures, deformation structures, and symbols for core disturbance in both soft sediment and indurated sedimentary rock. The "Samples" column on the barrel sheets indicates the positions and types of samples taken from each core for shipboard analyses. The abbreviations for these samples are as follows: IW = interstitial water; SS = smear slide; WRP = whole round for physical properties; WRB = whole round for microbiology; WHC = whole-core sample. Using the working half of each core, sediment color was measured at 10-cm intervals with a Minolta CM-2002 Spectrophotometer mounted on an automated scanning table. The specifics of this spectrophotometer were described by Schneider et al. (1995).

The results of semiquantitative smear-slide analyses are tabulated with visual percentage estimates for each constituent grouped into the following categories: D = dominant (>50%); A = abundant (>20%-50%); C = common (>5%-20%); P = present (>1%-5%); R = rare (0.1%-1%); T = trace (<0.1%). In the "Lithology" column of the data tables, "D" indicates that the sample is representative of the dominant lithology of the core and "M" indicates a minor component. Data were entered into the Sliders computer program.

Thin sections of sand, volcanic ash, tuff, and basalt were prepared for petrographic analysis, with visual percentage estimates for each constituent grouped into the following categories: D = dominant (>50%); A = abundant (>20%-50%); C = common (>5%-20%); P = present (>1%-5%); R = rare (0.1%-1%); T = trace (<0.1%). Unconsolidated sands were washed of fine-grained matrix using a 63-µm sieve, dried, and impregnated in epoxy. Data were entered into the STP computer program.

Routine samples for shipboard X-ray diffraction (XRD) analysis were taken from intervals adjacent to whole-round samples, and most are part of sampling clusters adjacent to physical properties and carbonate samples. Additional samples were collected periodically from such unusual lithologies as carbonate-cemented claystone and volcanic ash. Samples were freeze-dried, crushed either by hand or with a ball mill, and mounted as random bulk powders. The XRD laboratory aboard the JOIDES Resolution is equipped with a Philips PW-1729 X-ray generator and a Philips PW-1710/00 diffraction control unit with a PW-1775 35 port automatic sample changer. Machine settings for all standards were as follows: generator = 40 kV and 35 mA; tube anode = Cu; wavelength = 1.54184 Å (CuK); intensity ratio = 0.5; focus = fine; irradiated length = 12 mm; divergence slit = fixed at 1°; receiving slit = 0.2 mm; step size = 0.02°2; count time per step = 1 s; scanning rate = 2°2/min; rate meter constant = 0.2 s; spinner = off; monochronometer = on; scan = step; scanning range = 2°2 to 40°2.

The software used for XRD data reduction is MacDiff (versions 4.0.4 and 4.1.1). This shareware application for Macintosh computers supports digital data processing and measurement of peak geometry. Peak intensity (counts per step) and peak area (total counts) were recorded after creating a baseline (200 iterations for all 2- values) and smoothing the counts (17-term filter of standard weighted means).

The method of Fisher and Underwood (1995) was employed to determine semiquantitative relative abundances of minerals. This mathematical technique uses matrix singular value decomposition (SVD) to solve for reliable normalization factors. Calibration depends upon the analysis of known weight percent mixtures of mineral standards that are appropriate matches for the natural sediments encountered in the Nankai Trough. We emphasize that the normalization factors are specific to the combination of XRD hardware and software utilized during Leg 190. Seven reference minerals were chosen for the standard mixtures: smectite (Wyoming montmorillonite), illite (2M1 polytype), chlorite, kaolinite, quartz (St. Peter sandstone), plagioclase feldspar (Ca-rich albite), and calcite (Cyprus chalk).

Figure F3 shows the positions of all peaks used in semiquantitative analysis of mineral abundance. The numerical technique of Fisher and Underwood (1995) allows one to assign either positive or negative normalization factors to relate each indicator mineral to each target phase (Table T1). Sixteen mixtures of mineral standards were tested; each mixture was split and analyzed three times, and the average peak-area value for each mineral was input into the SVD calculations for each mixture. All XRD-determined abundances were normalized to 100%. Differences between the measured weight percentage for each mineral in the standard mixture and its respective average of XRD-determined weight percentage are generally <5% (Table T2). Average errors range from ±0.8% to ±3.2%. Errors are larger for chlorite + kaolinite and quartz.

As an independent check on the accuracy of SVD-based calculations, we also completed linear regression analysis on the relation between weight percent and peak area for each mineral in the standard mixtures. The smallest correlation coefficients are for smectite (r = 0.90), chlorite + kaolinite (r = 0.92), and quartz (r = 0.95). Errors increase using this technique as the absolute abundance of a particular mineral increases (Table T2). Nevertheless, when applied to data from natural bulk-powder specimens, regression equations for each mineral (Table T3) yielded results that are very similar to those based on SVD. Small systematic shifts occur in quartz (~2% higher using SVD) and plagioclase (~2% lower using SVD). Finally, we also tested the reproducibility of data reduction technique by having multiple operators determine peak areas using MacDiff software. These deviations, on average, are less than 1% and are caused by variations in where peaks intersect with baseline counts.

Limitations of X-ray diffraction analysis, as applied to specifically to sediments from the Nankai Trough, include (1) peak interference between (001) smectite and (001) chlorite; (2) peak interference between (002) chlorite and (001) kaolinite; (3) contamination of the Wyoming montmorillonite standard by quartz, which Fisher and Underwood (1995) estimated to be 39%; heterogeneity of the montmorillonite standard means that this value may not be accurate for all subsamples; and (4) modest contamination of the illite standard by small amounts of quartz. Given these caveats and the inherent problems of analyzing mixtures of highly crystalline minerals (quartz and plagioclase) and randomly oriented phyllosilicates (clay minerals), the bulk-powder data should not be used to characterize abundance of individual clay minerals. Moreover, relative values for quartz, plagioclase, and calcite should not be confused with absolute percentages of the total bulk solids.

Beyond routine semiqualitative assessment of relative mineral abundance, we also used XRD to characterize representative samples of volcanic ash. Because of the variability of crystal content, amorphous glass content, and alteration products in such samples, we simply recorded the intensity and area of representative peaks generated by common minerals (e.g., smectite, illite, chlorite, zeolites, quartz, plagioclase, cristobalite, calcite, amphibole, pyroxene, pyrite, and halite). A final parameter to measure is the ratio of peak areas for (100) quartz (d-value = 4.257 Å) to (101) cristobalite (d-value = 4.0397 Å). The accuracy of this ratio suffers from interference between the highest-intensity cristobalite peak and a secondary plagioclase peak at ~22°2.