IODP Proceedings    Volume contents     Search


Physical properties

Shipboard measurements of physical properties were undertaken to characterize recovered core material and other rocks. These data are used to link the geological observations made on the core to the results of downhole logging and regional geophysical survey results (seismic profiles, magnetic surveys, and so on).

Prior to physical property measurements, whole-round cores were allowed to thermally equilibrate for 4 h to ambient room temperature.

All core sections were run through the Whole-Round Multisensor Logger (WRMSL) to measure gamma ray attenuation (GRA) density and magnetic susceptibility. The Natural Gamma Radiation Logger (NGRL) was used to measure gamma radiation for whole-round sections. We also applied a filter to remove data acquired from gaps and cracks between core pieces (e.g., magnetic susceptibility data plot in the VCDs), the process of which is described in detail below (see “Data filtering”). We did not use the P-wave logger on the WRMSL, as these measurements require full-diameter core and good coupling to the liner. This is generally not the case for hard rock cores.

Following whole-round measurements and core splitting, the archive half of the core was passed through the Section Half Multisensor Logger (SHMSL) for measurement of magnetic susceptibility with a Bartington MS2E1 contact sensor probe and color reflectance with an Ocean Optics photospectrometer. The SHMSL uses a laser to record piece heights, which can locate gaps and cracks between pieces of the core. These laser profile data, together with GRA density data, were used to filter whole-round and split-half measurements (see “Data filtering”).

Thermal conductivity was measured on pieces from either the archive or working half of the split core sections, depending on the availability of suitable material. Discrete samples (~8 cm3 cubes) were taken from Expedition 312 and 335 working section halves for physical properties and paleomagnetic measurements. Sampling was intentionally limited to preserve material for postcruise analyses. Shipboard samples were located close to where shipboard geochemistry and thin section samples were taken. Discrete samples were used for compressional wave (P-wave) velocity measurements in three orthogonal directions following the standard IODP convention (Fig. F2), and moisture and density (MAD) measurements were used to determine bulk density and porosity.

A comprehensive discussion of methodologies and calculations used in the JOIDES Resolution physical properties laboratory is presented in Blum (1997). Throughout the cruise, we uploaded raw and “processed” data to the LIMS database.

Whole-Round Multisensor Logger measurements

GRA bulk density and magnetic susceptibility were measured nondestructively with the WRMSL. The sampling interval for WRMSL measurements was set at 1 cm with an integration time of 5 s for each data point to allow both instruments to acquire values from the same location downcore. Calibration was verified after each core measurement by passing a freshwater-filled calibration core through the WRSML. The nominal accuracy of the calibrated instruments is 1%–2%.

Gamma ray attenuation bulk density

The GRA densiometer on the WRMSL operates by passing gamma rays from a 137Cs source through a whole-round core into a 75 mm × 75 mm sodium iodide detector located directly below the core. The input gamma ray peak has a principal energy of 0.662 MeV and is attenuated as it passes through the core. Attenuation of gamma rays, mainly by Compton scattering, is related to electron density, which is related to material bulk density by

ρb = ρew/2ΣN,


  • ρb = bulk density,

  • ρe = electron density,

  • w = molecular weight, and

  • N = atomic number of elements in the material.

For the majority of elements and for rock forming minerals, 2ΣN/w is ~1, whereas for hydrogen 2ΣN/w = 1.9841. Therefore, for a known thickness of sample the gamma ray count is proportional to density. Calibration of the GRA densiometer was performed using a core liner filled with freshwater and aluminum density standards. Recalibration was performed if the measured density of the freshwater standard was not 1.00 ± 0.02 g/cm3. The spatial resolution of the GRA densitometer is <1 cm.

Magnetic susceptibility

Magnetic susceptibility, κ, is a dimensionless measure of the degree to which a material can be magnetized by an external magnetic field:

κ = M/H (SI),

where M is the magnetization induced in the material by an external field with a strength H. Magnetic susceptibility varies in response to the type and concentration of magnetic grains, making it useful for the identification of compositional variations.

The WRMSL measures volume magnetic susceptibility using a Bartington Instruments MS2 meter coupled to a MS2C sensor coil with a diameter of 8.8 cm and operating at a frequency of 0.565 kHz. During Expedition 335, the instrument was set to record SI units with an integration period of ~1 s, to give a sensitivity of 1 × 10–5 SI. The core diameter is smaller than the sensor coil aperture. The instrument output (κMEAS) depends on the diameter of the core (d) passing through the coil diameter (D), so a correction factor (κREL) is necessary to convert the instrument output to true volume susceptibility (κ in SI), where κREL = 3.45(d/D)3 (Bartington Instruments Ltd., 2011). The sensor is nominally calibrated for core where d/D = 0.66, so for the sensor in the WRMSL (where D = 8.8 cm), the correction factor is 1.0 and κ = κMEAS if d = 5.8 cm. During Expedition 335, where the core diameter varied in an unpredictable manner both within and between core pieces, a single correction factor was not justified, and therefore no corrections were applied to WRSML magnetic susceptibility measurements; instead, raw data are reported in instrument units.

The along-core response curve of the MS2C coil has a full width of half maximum (FWHM) of ~4 cm (Blum, 1997) and is consistent with the decay in magnetic intensity with distance from a dipole (Fig. F20). Therefore, measurements of susceptibility from core pieces <8 cm long will significantly underestimate magnetic susceptibility by more than 10%.

The Bartington sensor has a maximum output threshold of 9,999 instrument units, so any readings ≥10,000 lose the most significant digit and are “wrapped” around to lower values. Observations that we judged to be in error, based on high susceptibility values of neighboring analyses and examination of the core, were corrected by adding 10,000. The few corrected magnetic susceptibility data are the only unfiltered data ≥10,000. The corrected data are archived as “processed” data to the LIMS database.

Natural Gamma Radiation Logger

Gamma rays are emitted from rocks primarily as a result of the radioactive decay of 40K and the decay of isotopes in the decay series of 238U and 232Th. Measurement of NGR from the recovered core provides an indication of the concentration of these elements and can also be used to correlate core with the downhole gamma ray logs (e.g., Révillon et al., 2002).

The NGRL installed on the JOIDES Resolution was designed and built at the IODP-USIO at Texas A&M University (USA; Vasilyev et al., 2011). The main NGR detector unit consists of eight sodium iodide (NaI) scintillator detectors (~500 in3 each), seven plastic scintillation detectors, 22 photomultipliers, and passive lead shielding. The eight NaI detectors are spaced every 20 cm in the detector; the detectors themselves are semicylindrical annuli around the lower half of the core (each crystal is ~13 cm wide along the core). Detectors are shielded by lead to reduce the measurement of external gamma radiation, and the NGRL also employs seven plastic scintillation detectors that detect and actively suppress the effect of high-energy gamma and muon components of cosmic radiation. The NGRL was calibrated using 137Cs and 60Co sources to identify peaks at 662 and 1330 keV, respectively.

Background measurements of an empty core liner counted for 20,000 s (>5 h) were made upon arrival at Site 1256 (19 April 2011). Over the 100–3000 keV integration range, background counts averaged 4–5 cps and contributed ~80% of the uncorrected total counts for the low-radioactivity mafic rocks measured during Expedition 335.

A single measurement run with the NGRL provides a total of 16 measurements at 10 cm intervals over a 150 cm section of core. To achieve a 10 cm interval using the NGRL’s eight sensors spaced every 20 cm, the NGRL records two sets of measurements offset by 10 cm. Readings at the ends of core sections are automatically corrected for the reduced signal because of an inferred “missing” volume of material, assuming the core is truncated perpendicular to the z-direction at 0 cm and at a user-specified distance along the section. Total counts are routinely summed over the range of 100–3000 keV.

The quality of the energy spectrum measured depends on the concentration of radionuclides in the sample and on the counting time, with longer counting times providing better counting statistics. A live counting time of 1800 s (30 min) was set in each position (total live count time of 1 h per section). Improved counting statistics allow qualitative identification of the main contributors to the energy spectra (i.e., 40K and intermediate isotopes in the 232Th and 238U decay chains). To this end, we extracted counts for the energy windows recommended by the International Atomic Energy Agency ( for K, effective Th, and effective U (Table T8). Results provide a qualitative indication of the relative concentrations of each element. However, interference of energy spectra from each element’s decay series means that counts in each energy window include counts due to the decay of other radiogenic elements. For example, counts in the U window will include counts from the Th decay chain (and to a much lesser extent K), and counts in the K window will include counts from the decay chains of both Th and U. During Expedition 335, comparisons made between the counts in each window are thus qualitative. By building a database of well-resolved background-corrected spectra, the separation of 40K, 232Th, and 238U contributions from the measured spectra will be straightforward once well-defined standards become available and the sensitivity matrix and/or stripping ratios are defined.

Thermal conductivity

Thermal conductivity (k) is a material property that describes how heat is transported through a material. At steady state, it is the constant of proportionality between the spatial temperature (T) gradient and the heat-flux (q) down-gradient, such that

q = –kT.

Thermal conductivities of rocks depend on many factors, including temperature, pressure, porosity, type of saturating fluid, and the composition, distribution, and alignment of mineral phases. Thermal conductivity was measured on split core pieces under ambient conditions using a Teka TK04 system. All measurements were made at room temperature and pressure and were not corrected for in situ conditions.

This system measures thermal conductivity by transient heating of the sample with a known heating power and geometry. Changes in temperature with time during heating are recorded and used to calculate thermal conductivity. Heating power can be adjusted for each sample; as a rule of thumb, heating power in watts per meter is set to be approximately two times the expected thermal conductivity in watts per meter degree Kelvin. The temperature of the superconductive needle probe has a quasi-linear relationship with the natural logarithm of the time after the initiation of heating (Blum, 1997). The TeKa TK04 device uses a complex “special approximation method” (SAM) to calculate conductivity and to assess the fit of the heating curve. This method fits discrete windows of the heating curve to the theoretical temperature (T) with time (t) function:

T(t) = A1 + A2 ln(t) + A3 [ln(t)/t] + (A4/t),

where A1–4 are constants that are calculated by linear regression. A1 is the initial temperature, whereas A2, A3, and A4 are related to geometry and material properties surrounding the needle probe. Having defined these constants (and how well they fit the data), the apparent conductivity (ka) for the fitted curve is time dependent and given by

ka(t) = q/4π{A2 + A3[1 – ln(t)/t] – (A4/t)},

where q is the input heat flux. The maximum value of ka and the time, tmax, at which it occurs on the fitted curve are used to assess the validity of that time window for calculating the thermal conductivity. The best solutions are those where tmax is greatest, and these solutions are selected for output. Fits are considered good if ka has a maximum value, tmax is large, and the standard deviation of the least-squares fit is low. For each heating cycle, several values are output that can be used to assess the quality of the data, including “logarithm of extreme time” (LET), which should be large, the number of solutions (N), which should also be large, and the contact value, which assesses contact resistance between the probe and the sample and should be small (or uniform for repeated measurements).

Half-space determinations of thermal conductivity were made with a needle probe embedded in the bottom of a Plexiglas block with a thermal conductivity of 0.184 W/(m·K). Heat is assumed to be transferred through the sample, and the TK04 documentation indicates that heat flow through the Plexiglas block itself is only significant for sample thermal conductivities < 1 W/(m·K). Good thermal contact with the heating needle is required, so the split face of the samples was polished with 120-gauge silicon carbide powder. Tests conducted on Section 312-1256D-223R-3A (Piece 2) during Expedition 335 showed negligible improvement using finer polishing grit (2.201 ± 0.015 W/[m·K] before polishing, 2.240 ± 0.028 W/[m·K] after polishing with 120-gauge silicon carbide powder, and 2.237 ± 0.035 W/[m·K] after polishing with 320-gauge silicon carbide powder).

We avoided the use of silicone thermal contact gel that could contaminate the core. Empirical tests with both a certified MACOR ceramic standard (k = 1.637 ± 0.033 W/[m·K]) and a gabbro sample suggested that deionized (DI) water or seawater (k = ~0.6 W/[m·K]) can be used to improve sample/needle contact and provide consistent analyses within the analytical error (2%). Measurements without contact fluid yield more variable results, with most analyses >2% different from the expected conductivity (Table T9; Fig. F21).

Samples were saturated and left to equilibrate to room temperature in a seawater vacuum saturator for >24 h, and then the sample and sensor needle were equilibrated at room temperature in an isolated Styrofoam-covered seawater bath for at least 15 min prior to measurement. Isolation of the sample and sensor needle eliminated the effect of small but rapid temperature changes introduced by air currents in the laboratory, as well as the ship’s motion. The instrument internally measures temperature drift and does not begin a heating run until sufficient thermal equilibrium is attained.

Core pieces were measured at irregular intervals downhole (approximately one sample per section) depending on the availability of homogeneous and relatively vein/crack-free pieces long enough to be measured without edge effects (pieces >7 cm long; i.e., longer than the instrument needle). The MACOR ceramic standard was analyzed frequently between measurements to check the calibration and consistency of results (Table T9).

Section Half Multisensor Logger measurements

The SHMSL was used to measure spectral reflectance and magnetic susceptibility on archive section halves. An electronic platform moves along a track above the section half, recording the sample height using a laser sensor. The laser establishes the location of the bottom of the section; then the platform reverses the direction of movement, moving from bottom to top taking measurements of point magnetic susceptibility and spectral reflectance data at 1 cm intervals.

Color reflectance spectrometry

Reflectance from the section half was measured using an Ocean Optics Inc. system for ultraviolet through visible to near-infrared light (171–1100 nm wavelength at 2 nm intervals). Each measurement takes ~5 s to acquire. Spectral data are routinely reduced to the L*a*b* color system for output and presentation, in which L* is luminescence, a* is the blue–yellow value, and b* is the red–green value. The color reflectance spectrometer calibrates on two spectra, pure white (reference) and pure black (dark). Color calibration was conducted approximately every 12 h.

Point magnetic susceptibility

Point magnetic susceptibility is measured using a Bartington MS2E contact probe with a flat 15 mm diameter sensor operating at a frequency of 0.580 kHz. The sensor takes and averages three measurements at 1 s intervals to an accuracy of 5%. The area of response of the MS2E sensor is 3.8 mm × 10.5 mm, with a depth response of 50% at 1 mm and 10% at 3.5 mm, providing higher resolution measurements than the whole-round magnetic susceptibility instrument (Bartington Instruments, 2011). Units are reported in dimensionless SI units on a volume basis. The point magnetic susceptibility meter was calibrated by the manufacturer before installation on the ship. The probe is zeroed in air before each measurement point, and a background magnetic field is measured and removed from the data before being output.

As with the Bartington MS2C sensor on the WRMSL, the MS2 recorder attached to the SHMSL has an output threshold of 9999 SI and truncates the most significant digit for measurements >9999 SI. Following processing to remove data from within 1 cm of piece edges (see “Data filtering”), data were manually corrected and archived. We manually checked the data points, compared them to the data from the magnetic susceptibility coil on the WRMSL, and found a few spurious points that were corrected from the upper part of Section 335-1256D-214R-2 (see “Physical properties” in the “Site 1256” chapter).

Discrete sample measurements

Cubes (~8 cm3) were cut from working section halves and non-core samples for discrete measurements of P-wave velocity and MAD. Physical property measurements were conducted following the thermal demagnetization temperature step of 100°C. The samples were then passed on to the paleomagnetists for further measurements.

Moisture and density

Mass and volume measurements on discrete samples were made to determine water content, bulk and dry density, porosity, and void ratio. MAD data are also used for comparison with GRA bulk density data from the WRMSL. The shipboard MAD facility for hard rock samples consists of a vacuum water saturator, a dual balance system, and a hexapycnometer.

Vacuum water saturator

We used a vacuum pump system to ensure complete saturation of discrete samples. The system consists of a plastic chamber filled with seawater. A vacuum pump then removes air from the chamber, essentially sucking air from pore spaces. Samples were kept under vacuum for at least 24 h. During this time, pressure in the chamber is monitored periodically, via a gauge attached to the vacuum pump, to ensure a stable vacuum. After removal from the saturator, cubes are stored in sample containers filled with seawater to maintain saturation.

Dual balance system

A dual balance system was used to measure both wet and dry masses. Two analytical balances (Mettler-Toledo XS204) compensate for ship motion; one acts as a reference and the other measures the unknown (i.e., a sample). A standard mass of similar value to that of the sample was placed on the reference balance to increase accuracy. Using a reference mass within ~10% of the sample mass, an accuracy of 0.005 g is readily attainable. After wet mass determinations and P-wave measurements and prior to the determination of dry masses, samples were placed in an oven at 105° ± 5°C for 24 h and then allowed to cool in a desiccator.

Hexapycnometer system

The hexapycnometer is an IODP custom-built system using six Micromeritics cell units, custom electronics, and custom control programs. The system measures dry sample volume using pressurized helium-filled chambers with a precision of 0.02 cm3. At the start of the expedition, and whenever the helium gas tank is changed, shipboard technicians perform a calibration using stainless steel spheres of known volume. For a measurement five cells were run that contain unknowns and one cell that contains two stainless steel calibration spheres (3 and 7 cm3) with a total volume of ~10 cm3. Calibration spheres were cycled through the cells to identify any systematic error and/or instrument drift. Spheres are assumed to be known to within 1% of their total volume. Individual volume measurements were preceded by three purges of the sample chambers with research grade (99.995% or better) helium heated to 28°C, followed by three data acquisition cycles.

Moisture and density calculations

For density calculations, both mass and volume are first corrected for the salt content of the pore fluid:

Ms = [s(MwMd)]/(1 – s),


  • s = pore water salinity,

  • Ms = mass of salt,

  • Md = dry mass of the sample, and

  • Mw = wet mass of the sample.

Grain density (ρg) is determined from the dry mass (Md) and dry volume (vd) measurements:

ρg = (MdMs)/[VdMss)],

where ρs = density of salt (2.257 g/cm3).

The salt-corrected mass of pore water (Mpw) is calculated as

Mpw = (MwMd)/(1 – s).

Then, the volume of pore water (Vpw) is

Vpw = Mpwpw = (MwMd)/[(1 – s)ρpw],

where we assume the density of the pore fluid (ρpw) = 1.024 g/cm3 (from calculations of fluid sampled at ~713 m in the hole; Teagle, Alt, Umino, Miyashita, Banerjee, Wilson, and the Expedition 309/312 Scientists, 2006).

To calculate sample bulk density (ρb), first compute bulk volume:

Vb = Vd + Vpw.


ρb = Mw/Vb.

Porosity calculation

Porosity (ϕ) is calculated from the two volume parameters above:

ϕ = Vpw/Vb.

Compressional wave velocity

P-wave velocity (VP) measurements of hard rock samples were performed on the same discrete cube samples that were used for MAD and paleomagnetic determinations. VP measurements were performed on seawater-saturated samples immediately after wet mass determinations were made. Measurements were made using the x-axis caliper contact transducers on the VP gantry. Samples were oriented following standard IODP conventions, and measurements were made in the x-, y-, and z-directions for each cube (Fig. F2). The system uses Panametrics-NDT Microscan delay line transducers, which transmit at 0.5 MHz. The peak of the first arrival was identified automatically by the installed IODP software. The complete waveform is stored with the data if reanalysis is deemed necessary. Shipboard visual checks of the picks appeared satisfactory. The distance between transducers was measured with a built-in linear voltage displacement transformer (LDVT).

Measurements on standards were conducted as frequently as necessary. Calibration was made with a series of acrylic cylinders of differing thicknesses and a known VP of 2750 ± 20 m/s. In a set of 25 calibration measurements, there was no systematic drift over time with the mean value of 2753 ± 22 m/s, which is well within a 2% error margin and is very close to certified acrylic velocity of 2750 m/s. Individual results deviate from the “true” value by up to 2%. Further extensive test runs on standards confirmed the accuracy of 1%–2%. Instrument performance is worse on copper cylinders (which are more suitable for the hard rock calibration because of high rigidity and velocity). In a set of 12 measurements on copper cylinders, the mean was 4798 ± 80 m/s (1σ error), which is within the expected 2% error margin, but individual measurements may deviate by as much as 4% from the mean. This result most likely reflects the errors in the measurement of the shorter traveltimes through the faster velocity medium.

At the beginning of Expedition 335, >300 VP measurements were taken on 11 discrete gabbro samples from Expedition 312 cores to constrain the repeatability and uncertainty of VP measurements, and thereby to define the optimum protocol for shipboard VP measurements.

First, we conducted VP measurements following Expedition 312 protocol. The surfaces of seawater-saturated cubes were wiped to remove excess water. The cubes were weighed for MAD analysis, and then VP measurements were performed in three orthogonal directions, adding a drop of deionized water to the cube surface (according to the standard procedure) to improve coupling between the sample and the transducers. Each measurement was repeated three times to provide an estimate of the precision (Table T10).

One concern identified during Expedition 335 was that the majority of samples show a systematic increase in VP values for each repeated measurement (Fig. F22). Large variations in VP measured in orthogonal directions within each sample were also observed.

To investigate the causes of these increasing VP values, we conducted 10 sets of repeat measurements on the same cube (Sample 335(312)-231R-3W, 52–54 cm; PP/PMAG), adding deionized water to cube surfaces for each measurement on each of three directions (Fig. F23; Table T11). A similar systematic increase in VP was observed for each direction, until after approximately seven measurements when the results stabilized. The average increase in velocity before reaching a plateau value is ~600 m/s. Caliper separation was constant for each repeat reading, and measurements of standards before and after sample measurements allowed instrumental drift to be ruled out. So the systematic increase in velocity is produced by a decrease in the measured transit time, which we hypothesize reflected the resaturation of surfaces after they had been dried during wet mass measurements.

To test this hypothesis, we made repeat measurements of a sample under different conditions of surface saturation: without water added, with deionized water added to surfaces, and fully submerged in seawater bath during measurement. Measured VP values decreased as the cube surface dried out (drying was qualitatively assessed as the sample surface color became lighter during the measurements) (Fig. F24; Table T12), suggesting that the saturation of the sample surfaces has a large effect on measured VP .

To improve measurement precision, the saturation of sample surfaces needs to be constant for all measurements. We therefore constructed a seawater bath to maintain a constant surface saturation during VP measurements of discrete samples (Fig. F25). The seawater bath has an open top to allow direct contact between the minicube and the upper transducer and a thin flexible rubber membrane at the bottom to provide good coupling with the lower transducer. The container was filled with seawater to ~1 mm above the sample to provide complete saturation and to avoid flooding the upper transducer. Measurements conducted using the seawater bath were more consistent than those conducted without the bath. The mean of standard deviations is 25 m/s with the bath, 135 m/s for wet surfaces without the bath, and 157 m/s for dry surfaces.

The presence of the membrane changed the elastic properties of the system, lowering velocity measurements compared to the plateau for saturated measurements without the seawater bath. To correct the effect of this rubber membrane, we made measurements on standards with and without the seawater bath. Repeat measurements under both saturated and dry conditions for acrylic, copper, and aluminum standards indicate that the membrane is 0.298 ± 0.033 mm thick and adds 0.333 ± 0.030 µs to the recorded traveltime (See PWAVE_CALIBRATION _BATH in PHYSPROP in “Supplementary material”). Measurements made with the seawater bath can thus be corrected using

VP = (s – 0.298 mm)/(t – 0.333 µs),

where s is the measured caliper separation and t is the measured sonic traveltime. This correction increases the measured velocity of gabbroic samples by ~600 m/s.

Data filtering

Data obtained from the WRMSL, NGRL, and SHMSL are used to determine natural variations in physical properties, but a large degree of measured variation can be attributed to systematic errors associated with the geometry of the recovered core, specifically because of measurements that were taken (1) close to the ends of sections and pieces, (2) over cracks within pieces, or (3) on pieces that were not ideal 66 mm diameter cylindrical cores for which the system is calibrated. For bulk volumetric measurements (magnetic susceptibility on the WRMSL and NGR on the NGRL), corrections can be made that account for the reduced volume of material in the vicinity of the sensor. For point measurements (GRA, reflection spectroscopy color [RSC], and point magnetic susceptibility [MSPOINT]), gaps and irregular core surfaces can lead to poor data, and these should be removed before the data are interpreted. We therefore filtered the data to remove and/or correct measurements that are affected by geometrical departures from a continuous cylindrical core 66 mm in diameter.

A filtering algorithm was developed to remove data points from gaps and cracks using GRA density data from the WRSML and the laser profile height from the SHMSL to identify gaps. GRA density data are sensitive to the bulk density of the rock in an area of ~1 cm2. Gaps between pieces or fractures can generate anomalously low densities, and as the densities of rocks vary by less than an order of magnitude and are greater than water (1 g/cm3), it is relatively easy to define a threshold filter that uses density to identify gaps or sections of the core with a reduced volume of material. This method of filtering is particularly effective for core recovered during Expedition 335 and core reanalyzed from Expedition 312, where porosity is low and bulk density is relatively constant. Laser profile height data from the SHMSL provides an independent method to identify core pieces.

Data culling

Using GRA density and laser profile data, we defined gaps in the core on the basis of a threshold value below which gaps were defined (1 g/cm3 for whole-round GRA density and <2 cm for laser profile data). In addition, gaps were defined where significant horizontal gradients in the data occurred (>0.2 g/cm3 per centimeter for GRA density and >1 cm per centimeter for laser profile height). Visual inspection suggested that these two parameters defined all significant piece edges and gaps in the sections. Once gaps were defined, data were removed (culled) from gaps and for a distance adjacent to the gap that was specified by the instrument’s resolution or sensor size (Table T13). Given the low resolution of the NGRL, no gap filtering was attempted for NGR data.

Predictive filtering of magnetic susceptibility data

The simple culling of data points within a certain distance of piece edges is a useful first-order filter. However, for the magnetic susceptibility meter attached to the WRMSL the response curve indicates that the minimum length of pieces that can be analyzed to give >90% of the true magnetic susceptibility is ~8 cm. During Expeditions 312 and 335, very few pieces longer than 8 cm were recovered; thus, only ~10% of data reanalyzed from Expedition 312 is retained, whereas <1% of data from Expedition 335 core is retained after a simple culling filter is applied.

Simple culling of the data therefore retains little data for interpretation. Consequently, we developed and applied a filter defined by the response curve of the MS2C meter (Fig. F20). Using the distribution of known gaps to define a gap function g(x), where g(x) = 1 for core and g(x) = 0 for gaps between core, g(x) can be convolved with the instrument response curve to specify the magnitude of correction required (Fig. F26). Clearly, this method requires the location of the gaps to be accurately known, and errors associated with the correction are greater closer to piece edges. Despite this caveat, synthetic tests show that, theoretically, this filter can be used to determine the magnetic susceptibility of pieces <8 cm in length (e.g., the interval from 10 to 30 cm).

In practice, we defined and applied this filter in three separate ways:

  1. Using gaps defined from GRA density data to make a simple binary input that was convolved with the response function,

  2. Using gaps defined by the laser profile height data to make a simple binary input filter that was convolved with the response function, and

  3. Using normalized GRA density to define a normalized function of material volume (assuming constant grain density) that was convolved with the MS2C response curve.

The results of these three filters are averaged to define a single output value (Fig. F27); we take this approach (1) so that the different data used to define the filters (GRA density and laser profile data) are treated independently and (2) to provide an independent assessment of the error in the filters based on the standard deviation of the three filtered results, in addition to errors defined by uncertainty in the filter definition.

Scaling archive-half data measured by WRMSL

For WRMSL magnetic susceptibility of Expedition 312 archive section half measurements, results were compared with SHMSL point magnetic susceptibility data. Despite scatter due to the different resolutions of the two measurements, a linear trend was observed, indicating a best-fit scaling factor of ~1.8. This scaling factor accounts most likely for both the decreased volume of measuring section halves and for the fact that the Expedition 312 core is not 58 mm in diameter.

GRA density measurements of section halves are significantly lower than full core section measurements (~1–1.2 g/cm3). Calculated density is dependent on the volume of analyzed material according to attenuation of waves and generally obeys an exponential decay function with distance; it follows that the scaling factor for GRA density measured for a core of reduced diameter can be approximated by

ρscale = ρobs(D/d),


  • D = diameter of core for which the instrument is calibrated,

  • d = diameter of the measured core,

  • ρobs = output GRA density, and

  • ρscale = corrected GRA density.

For section half measurements, d = ~28 mm and we assume D = 66 mm, so measurements are scaled by ~2.3. Following this correction, archive-half measurements are directly comparable to Expedition 312 GRA density measurements and discrete sample density measurements.

Magnetic susceptibility measurements on non-core samples

We used a Bartington MS2F handheld susceptometer to measure susceptibility of non-core samples recovered during Expedition 335. After zeroing the probe in air, we measured rock samples by placing the probe at the surface of the samples. The instrument has an operating frequency of 0.580 kHz (Bartington Instruments Ltd., 2011). Measured values are approximately the same as true volume susceptibility of the material.