Magnetostratigraphic importance of secondary chemical remanent magnetizations carried by greigite (Fe₃S₄) in Miocene sediments, New Jersey shelf (IODP Expedition 313)

The ferrimagnetic iron-sulfide greigite (Fe₃S₄) is frequently found in lacustrine (e.g., Dell, 1972; Frank et al., 2007; Roberts et al., 1996; Skinner et al., 1964; Snowball and Thompson, 1988, 1990) and rapidly deposited marine sediments (e.g., Florindo and Sagnotti, 1995; Horng et al., 1998; Kao et al., 2004; Oda and Torii, 2004; Roberts and Turner, 1993; Tric et al., 1991). Greigite can form authigenically in anoxic sediments as a precursor to pyrite (FeS₂), associated with bacterial degradation of organic matter (Berner, 1984), or through biomineralization by magnetotactic bacteria (Bazylizinki et al., 1993; Mann et al., 1990; Vasiliev et al., 2008). Kao et al. (2004) suggested that greigite preservation is favored by conditions with high concentrations of reactive iron in fine-grained sediments with limited organic carbon supply. In such cases, the reactive iron exhausts supply of dissolved sulfide so effectively that the pyritization process is arrested.

It has been argued that the iron-sulfide formation in continental shelf sediments is restricted to the upper anoxic ∼1 m, where sulfate reduction occurs, and that chemical remanent magnetizations (CRMs) acquired by greigite will accurately record the geomagnetic field with only a relatively small time lag (Roberts and Turner, 1993; Tric et al., 1991). However, as numerous later studies have shown (e.g., Florindo and Sagnotti, 1995; Horng et al., 1998; Roberts and Weaver, 2005), iron-sulfide diagenesis may occur at different times after the sediments have been deposited, making it difficult to ascertain the time at which the CRM was acquired. These studies highlight the need for detailed mineral magnetic and/or petrographic analyses in order to establish the timing of greigite formation and to evaluate the reliability of the paleomagnetic signal.

In 2009, coring associated with Integrated Ocean Drilling Program (IODP) Expedition 313 was conducted on the New Jersey shallow shelf to unravel the sedimentation history controlled by eustatic sea-level change. Initial paleomagnetic results from Miocene clay-rich horizons identified several reversal boundaries with partly stable polarity in between, which suggests that magnetostratigraphic interpretations could potentially be used to develop age models for the sediments (Mountain et al., 2010). However, similar to a previous study from the New Jersey shelf (Oda and Torii, 2004), some samples contain evidence for gyroremanent magnetization (GRM) acquisition, which can be indicative of the presence of greigite (Snowball, 1997).

In this study, we attempted to identify the main magnetic minerals in the Miocene clay-rich horizons from Expedition 313. Through detailed rock-magnetic analyses, we investigated the importance of magnetic iron-sulfide diagenesis and assessed the fidelity of the magnetic polarity stratigraphy at the studied site.

The New Jersey shelf is a classic passive margin, with the main tectonics for the last ∼165 m.y. dominated by simple thermal subsidence, sediment loading, and flexure (Reynolds et al., 1991; Watts and Steckler, 1979). Holes M0027A, M0028A, and M0029A were drilled in shallow water depths (35–40 m) along a seaward transect (Fig. 1). The recovered sediments ranged from coarse sand to clay, which were deposited in cyclic nearshore to offshore successions from the Late Oligocene to present (Mountain et al., 2010).

In this study, we concentrated on the clay-dominated upper part of unit II in Hole M0027A (core 64–70; 192–208 m) and Hole M0028A (core 2–8; 224–244 m) directly overlying the m4.1 maximum flooding surface (Mountain et al., 2010). The sequences are interpreted as being deposited during the middle Miocene in an offshore environment at water depths ranging from 50 to 100 m (Mountain et al., 2010).

Mountain et al. (2010) described the upper unit II sediments in M0027A as dark- and pale-gray clay with occasional color banding and silty laminae. There is a general increase in bioturbation from bottom to top, culminating in a darker-gray–colored, relatively homogeneous clay sequence between 199 and 195 m. Above 207 m, the bedding is contorted at a range of scales, which indicates postdepositional disruption of the successions. The lower section of the clay-dominated sediments in M0028A (between 244 and 239 m) consists of similar dark- to pale-gray clays as in M0027A, with contorted bedding and high degrees of bioturbation. The overlying sediments consist of gray-yellow-brown color-banded clay with chondrite trace fossils and pyritic silty laminae.

All coring was carried out between 30 April and 17 July 2009 during the offshore phase of the expedition. M0027A core 64 was drilled using a hydraulic piston corer (HPC), and M0027A cores 65–70 were drilled using an extended nose corer (EXN). M0028A cores 2–8 were drilled using a standard rotary corer (ALN). All cores, collected in 3-m-long IODP-standard liners (∼62 mm in diameter), were cut into sections (maximum 1.5 m length), sealed, and stored in refrigerated conditions.

The cores were split lengthwise into working and archive halves during the IODP Onshore Science Party, held between 4 November and 4 December 2009 in Bremen, Germany. For this study, 277 paleomagnetic samples were collected from the working halves of the cores. Oriented plastic sample boxes (6.2 cm³) were carefully pushed directly into the sediments. For harder sediments, a custom-made stainless steel “square shaped” tube was first used to extract the sediments, which were then transferred into the sample boxes. Sediments with evidence of postdepositional deformation, e.g., soft-sediment folding structures, were avoided. The samples were stored in refrigerated conditions.

All paleomagnetic measurements were carried out using a 2G Enterprises pass-through direct-current (DC) cryogenic magnetometer (model 755 R, horizontal orientation) equipped with an automated alternating field (AF) demagnetization system and an automated sample handler capable of processing 96 discrete samples in one batch. Measurements were carried out in the rock and paleomagnetic laboratory at the Department of Geosciences, Bremen University.

During the sampling in 2009, the natural remanent magnetization (NRM) was measured for all samples, as well as the remanent magneti after AF demagnetization at 5 mT increments up to 60 mT. Samples in the first batch were further demagnetized with increments of 10 mT up to a maximum field of 100 mT. However, due to the limited duration of the visit to Bremen, the remaining samples were only demagnetized up to 60 mT.

In August 2010, the remaining samples were further demagnetized at 10 mT increments to a maximum field of 100 mT. Anhysteretic remanent magnetizations (ARMs) were induced in a peak AF of 100 mT and a DC bias field of 50 µT. This is also expressed as susceptibility of ARM (K_ARM), after normalizing by the 50-µT DC bias field. Isothermal remanent magnetizations (IRMs) were induced in a DC field of 700 mT. After each treatment, the samples were AF demagnetized using the same steps as for the NRM.

All mineral magnetic analyses were carried out after the paleomagnetic measurements were completed using facilities in both the Paleomagnetic and Mineral Magnetic Laboratory at the Department of Geology, Lund University, and the Geomagnetism Laboratory at the School of Environmental Sciences, University of Liverpool. However, for comparison, we also show magnetic susceptibility (κ_initial) measured immediately after the cores were split in 2009 using a Geotek multisensor core logger (MSCL) at the IODP Bremen Core Repository, Bremen.

The IRM induced in a field of 700 mT was assumed to represent the saturation IRM (SIRM). A few samples exhibited SIRMs in excess of the dynamic range of the cryogenic magnetometer, and therefore an SIRM was also induced on selected samples using a Redcliffe 700 BSM pulse magnetizer and measured with a Molspin Minispin magnetometer. Volume-specific magnetic susceptibility (κ) was measured using a Geofyzica Brno Kappabridge (KLY-2) magnetic susceptibility meter.

The anisotropy of magnetic susceptibility (AMS) was measured on 67 samples using an AGICO Multi-Function Kappabridge (MFK1-FA) equipped with an automated rotation system. Using the same equipment, finely ground ∼300 mg extracts of six samples were selected for measurements of low- and high-temperature dependence of magnetic susceptibility. To minimize oxidization during heating, the high-temperature experiments were conducted in an argon atmosphere.

A variable field translation balance (MM VFTB) was used to make hysteresis (peak fields of 800 mT) and strong-field thermomagnetic measurements on finely ground ∼200 mg extracts from 34 representative samples. A Princeton Measurements Corporation alternating gradient magnetometer (AGM M2900–2) was used to complement the hysteresis measurements (peak fields of 1 T) for three weakly magnetized samples and to measure first-order reversal curves (FORCs) for six representative samples (222 FORCs per sample, averaging time = 100 ms). The FORC distributions were calculated using the MATLAB-based software UNIFORC (Winklhofer and Zimanyi, 2006) and the algorithm of Pike et al. (1999). All measurements with the AGM were carried out on finely ground ∼50 mg extracts of samples dispersed within a diamagnetic epoxy resin.

Down-core variations of room-temperature mineral magnetic parameters as well as paleomagnetic results from both Hole M0027A and Hole M0028A are summarized in Figure 2. Five mineral magnetic assemblages (A1 to A5) were identified based on changes in κ, SIRM, and the median destructive field of ARM (MDF_ARM). We note similarities between the magnetic mineral assemblages and lithology. The same color code adopted for A1 to A5 in Figure 2 is used in all subsequent figures.

Initial AF demagnetization results provided evidence of GRM acquisition, where an AF demagnetization-induced magnetic remanence is acquired more or less perpendicular to the applied AF in strongly anisotropic magnetic material (Stephenson, 1980). In light of this finding, we changed the demagnetization procedure following the technique proposed by Dankers and Zijderveld (1981), which we refer to as the anti-GRM procedure (AGY), where the remanence in each axis is measured directly after AF demagnetization of the same axis. The GRM effect can be quantitatively described by the ΔGRM/ΔNRM ratio (Fu et al., 2008), where ΔGRM = NRM_{100 mT} – NRM_minimum and ΔNRM = NRM_{0 mT} – NRM_minimum. This ratio was calculated using the final magnetometer reading after each step (i.e., not using the AGY data). Hole M0027A (Fig. 3A) has ΔGRM values that are often more than 10 times higher than ΔNRM in the upper 4 m of the studied interval. The ΔGRM values are slightly lower for the rest of the studied interval, mostly due to a loss of remanence during storage between the initial AF demagnetization (0–60 mT) in November 2009 and the final AF demagnetization (60–100 mT) in August 2010. The percentage of NRM_{60 mT} lost between 2009 and 2010 for M0027A is shown in Figure 3B. Loss of remanence during storage has been attributed to oxidation of greigite and/or to “storage diagenesis” of other magnetic minerals (Oldfield et al., 1992; Snowball and Thompson, 1988).

A viscous overprint was removed by AF demagnetization up to 10–30 mT, and the characteristic remanent magnetization (ChRM) was isolated using steps within a range of 10–50 mT. Samples from A1 to A4 have high NRM intensity in the range of 10–300 mAm^–1 with maximum angular deviation (MAD) values typically below 10° (Fig. 2) and demagnetization vectors directed toward the origin of vector component diagrams. The NRM is generally strongest in A2, but spikes in intensity coincide with dark laminae (e.g., at 192.2, 201.7, and 203.2 m). Samples from A5 have distinctly different AF demagnetization behavior, with little or no GRM, up to two orders of magnitude lower NRM intensities, and, except for the upper parts, generally high MAD values (Figs. 2 and 3).

Declination data are not shown because the cores were not oriented and would have rotated during drilling. Inclination for well-defined directions (MAD <5°) varies around that predicted for a geocentric axial dipole (GAD) field at the location (±58.3°), which suggests that the data faithfully record different polarity states. However, in M0027A, we observe inclinations that are consistently ∼10° too steep (compared to the GAD prediction) for both polarity states, which indicates that either the whole sequence was tilted tectonically after the sediments were magnetized or that the cores were drilled at an oblique angle. Samples from the top of M0027A (A1 and upper A2) reveal a partially overprinted reversed polarity component, but the ChRM is difficult to isolate due to GRM acquisition at higher fields. In the lower part of M0028A (A3), we observe a mixed polarity interval that coincides with changes in NRM intensity (samples with low NRM intensity and reversed polarity are highlighted with open circles). The preliminary polarity states estimated during initial measurements are shown in Figure 2 (samples with low NRM intensity and reversed polarity are highlighted with open circles).

The magnetic susceptibility values measured after core splitting (κ_initial, surface scan) and after the paleomagnetic measurements (κ, volume specific) are not directly comparable, but they can be used to infer relative changes from oxidation/diagenesis that may have occurred during sediment storage. The two sets of measurements are compared in Figure 2. Here, κ varies in the range of 10^–4–10^–3 SI units, with peak values observed in A3. Note, κ_initial has similar variations for A3 to A5; however, A1 to A2 have higher values relative to κ, which suggests a higher degree of alteration in these sections during storage. The lack of similar trend in Figure 3 suggests that the enhanced loss of susceptibility was due to selective oxidation of the finest, superparamagnetic (SP), particles. Drying of the samples after the paleomagnetic analyses appears to have halted the loss of remanence/susceptibility (cf. Roberts et al., 2011), which is evident in the reproducibility of κ and SIRM measurements thereafter.

SIRM varies from 0.01 to 0.1 Am^–1 for A5 to more than 100 Am^–1 for A1 to A4, with the latter assemblage having similar relative variation to κ (Fig. 2). SIRM/κ ratios for A1 to A4 are high, on the order of 30–100 kAm^–1 (see Figs. 4A and 4D), which is indicative of greigite (e.g., Reynolds et al., 1994; Roberts, 1995; Roberts and Turner, 1993; Snowball and Thompson, 1990). SIRM/κ and κ values (Fig. 4A) for A1 to A4 are similar to those used by Larrasoaña et al. (2007) to identify greigite and pyrrhotite formed in gas-hydrate–bearing sediments. κ_ARM, which is sensitive to finer grain sizes, has more similar variations to κ_initial, with the highest values (up to 3 × 10^–2 SI units) observed in A2 (Fig. 2). The κ_ARM/κ ratios (Fig. 4B) illustrate this difference, which indicates the presence of finer magnetic grain sizes in A2. MDF_ARM varies between 20 and 50 mT but is relatively stable within each assemblage (Fig. 2). MDF_ARM reflects the resistance to AF demagnetization, which depends on the coercivity of the ARM fraction but potentially also the countering effect of simultaneous GRM acquisition (Roberts et al., 2011; Stephenson and Snowball, 2001). High κ_ARM and low MDF_ARM in A2 could therefore partly be due to less GRM acquisition in these sediments (Fig. 3).

Biplots of the S-ratio (IRM_{–100 mT}/SIRM) versus ARM_{40 mT} divided by the initial ARM (SARM) and SIRM/κ versus the coercivity of remanence (H_cr), which were used by Peters and Thompson (1998) to distinguish different magnetic minerals at room temperature, are shown in Figures 4C and 4D. The S-ratio and the ARM_40mT/SARM ratio are both easily measured parameters that provide similar information to H_cr (lower S-ratio = higher H_cr) and MDF_ARM, respectively. Based on the observations of Peters and Thompson (1998), we conclude that the ferri assemblage in A3 to A4 is dominated by greigite. A5 plots mainly in the range for (titano-) magnetite, and low ARM_{40 mT}/SARM in A1 to A2 (Fig. 4C) suggests that these samples may contain mixtures of (titano-) magnetite and greigite. However, as noted by Roberts et al. (2011), the analyses of Peters and Thompson (1998) were carried out on single-domain (SD) greigite and may not apply to non-SD greigite particles. In addition, the high ARM_{40 mT}/SARM ratios used by Peters and Thompson (1998) as one indicator of greigite could potentially also underestimate the influence of the fraction of SD greigite that does not acquire a GRM during AF demagnetization (e.g., Sagnotti et al., 2005).

IRM acquisition curves of A1 to A4 all have a sigmoidal shape typical of SD particles in that most of the IRM acquisition is confined to a narrow field range, in this case between 40 and 100 mT, reaching saturation at around 400 mT (Fig. 5). Our results, in particular A1, A2, and A4, plot close to the range observed for greigite by Peters and Thompson (1998). Greigite usually saturates at applied fields below 300 mT, but higher values have been recorded in partially oxidized greigite-bearing samples (Roberts et al., 2011). IRM acquisition curves can in theory be used to “unmix” a multicomponent specimen with different coercivity contributions (Heslop et al., 2002). The non-Gaussian shapes of the IRM acquisition derivatives (Fig. 5B), which are slightly skewed toward lower field values, suggest minor influences from one or more (additional) lower-coercivity components. We note, however, that the IRM unmixing approach should only be used for noninteracting magnetic assemblages, which as shown in the next section, does not apply to the samples from this study.

The AMS results for A3 to A4 and the lower part of A2 are characterized by high degrees of anisotropy (P′ = 1.15–1.5) and normal (oblate) magnetic fabric, with the minimum susceptibility direction perpendicular to the bedding plane (Fig. 6). Above 198 m in M0027A, the sediment becomes less anisotropic (P′ = 1.01–1.05) with both normal and inverse fabrics observed. A5 has a similar pattern, with a normal fabric in the lower part that changes progressively toward lower degrees of anisotropy and an inverse fabric at around 228 m in M0028A (Fig. 6). The observed inverse fabric is most likely related to physical disturbance of the sediments (Rosenbaum et al., 2000), as suggested by the lithostratigraphic description. Given that extensive deformation was observed in the A3 sediments, it is surprising that these sediments have high degrees of anisotropy and normal fabric. We note that inverse AMS fabric can also be caused by iron carbonates such as siderite (FeCO₃) (Rochette, 1988), which has been detected in X-ray diffraction (XRD) analyses of the studied clays (Mountain et al., 2010; van Geldern et al., 2013).

Hysteresis loops from A1 to A4 indicate largely SD behavior with high M_rs/M_s ranging from 0.35 to 0.55 and low H_cr/H_c around 1.5 (Fig. 7A). Coercivities (H_c) are generally high (30–55 mT), with an increasing trend from A1 to A4 (Fig. 7B). FORC distributions (Fig. 7C) all have a negative region in the lower left-hand corner, which is indicative of a SD response (Newell, 2005). The concentric contours are also indicative of SD particles, where the vertical spreading is due to strong magnetostatic interactions characteristic of authigenic greigite that grows in clusters (Roberts et al., 2006; Rowan et al., 2009). Several samples, particularly in A2, also have slightly elongated contours along the H_c axis. This “central ridge” feature is caused by the presence of noninteracting SD particles, often associated with intact magnetofossil chains (Egli et al., 2010). The presence of the noninteracting SD particles in A2 gives rise to a bimodal coercivity distribution. The coercivity distribution along the H_c axis (Fig. 8) has two local maxima: (1) one around 20 mT associated with the central narrow ridge and another (2) around 50 mT associated with the concentric distribution.

Hysteresis loops from A5 are slightly wasp-waisted, which can be diagnostic of a larger SP contribution (Roberts et al., 1995; Tauxe et al., 1996). The hysteresis data from A5 plot in the pseudo–single domain (PSD) region of the Day plot (Fig. 7A), with M_rs/M_s ranging between 0.1 and 0.2 and H_cr/H_c between 1.5 and 3. The FORC diagram contains nondiverging and concentric contours close to the H_u axis, which are indicative of PSD particles (Muxworthy and Dunlop, 2002; Roberts et al., 2000).

The hysteresis ratios for all samples appear to follow the theoretical single domain–multi domain (SD-MD) mixing relationships (Fig. 7A) determined for (titano-) magnetite (Dunlop, 2002a, 2002b), but which may also be applica to greigite (Chang et al., 2007; Roberts et al., 2011). Contrary to most studies involving greigite, which tend to plot along a SD-SP mixing line (Roberts, 1995; Roberts et al., 2011), H_cr/H_c ratios are relatively low, which suggests only minor contributions from SP particles.

Strong-field thermomagnetic curves from all samples contain two main breaks in slope, associated with enhanced loss of magnetization, around 150 °C and/or 280 °C (Figs. 9A–9C). Heating cycling experiments show irreversible behavior, which indicates thermal alteration of the initial magnetic mineral (Figs. 9D–9F). Around 500 °C, there is a secondary peak in magnetization, which decays at 560–590 °C within the Curie temperature range of magnetite. This secondary peak, which is observed in all samples but is generally more pronounced in A1 to A2 and A5, is most likely due to oxidation of different iron-bearing minerals, e.g., siderite (Pan et al., 2000), pyrite (Passier et al., 2001), or greigite (e.g., Roberts et al., 2011), and conversion into magnetite. Temperature cycling experiments reveal several increases in magnetization on heating above 180 °C, which indicates growth of magnetite, which masks any further loss of the initial magnetic phase.

Irreversible loss of magnetization due to decomposition upon heating to temperatures above 200–300 °C is typical of greigite (e.g., Dekkers et al., 2000; Krs et al., 1992; Reynolds et al., 1994; Roberts et al., 1995, 2011; Roberts and Turner, 1993; Snowball and Thompson, 1988, 1990; Tric et al., 1991). Alteration in greigite-bearing samples starting at 150 °C (Fig. 9A) is not well documented in the literature, but the thermomagnetic behavior of the studied samples bears close resemblance to hydrothermally synthesized greigite from Dekkers et al. (2000). This alteration at low heating temperatures, which appears to be more important in samples with high coercivity from A4 and A3, might be due to some form of impurity or defect/tension in the greigite crystal lattice.

The temperature dependence of mass-specific susceptibility (χ) for typical samples is illustrated in Figure 10. At low temperatures, χ decreases with warming, which is typical of paramagnetic material (Fig. 10A). In an attempt to isolate the ferrimagnetic component, we subtracted a hyperbolic function from the data, fitted between –140 °C and 0 °C, following the method of Hrouda (1994) (Fig. 10B). We do not observe any low-temperature transitions and generally only small χ variations within the measured temperature range. This relatively stable behavior at low temperatures is consistent with that documented for SD-MD greigite particles (Chang et al., 2009). Lack of a Verwey transition within the measured temperature interval suggests that the sediments do not contain significant amounts of stoichiometric (unoxidized) magnetite (Moskowitz et al., 1993; Özdemir et al., 1993); however, we note that it does not exclude contributions from titanomagnetite (Moskowitz et al., 1998). The χ-T curves are characterized by an irreversible increase at around 240 °C and another strong increase at around 300 °C (Fig. 10C) due to the same secondary alteration observed in the strong-field thermomagnetic curves.

Previous magnetostratigraphic studies from the New Jersey continental shelf (Oda and Torii, 2004; Van Fossen and Urbat, 1996) and the New Jersey coastal plain (Van Fossen, 1997) have all proven problematic either due to weak magnetizations resulting from dissolution of the primary magnetic mineral (Urbat, 1996) and/or to multi magnetizations involving a later chemical overprint from authigenic greigite (Oda and Torii, 2004). Our results are largely consistent with these interpretations. The mineral magnetic data suggest that the remanent magnetization in the sediments is carried by greigite, with possible traces of titanomagnetite.

We identified five mineral magnetic assemblages (A1 to A5) in the studied sediments. The most convincing evidence for greigite is found in sediments from A3 to A4, which is dominated by SD magnetic behavior. The hysteresis data (Fig. 7) suggest a progressive trend from SD to more PSD behavior with decreasing depth in Hole M0027A (A4 to A1). We interpret this to reflect a change in grain-size distribution toward progressively coarser greigite particles. However, the low MDF_ARM and the presence of noninteracting SD particles in A2 suggest that a substantial part of the magnetization in these sediments could be carried by greigite magneto or possibly ultrafine authigenically precipitated greigite. Based on the mineral magnetic results, we cannot exclude minor contributions from titanomagnetite. The A5 samples are characterized by PSD behavior (greigite or titanomagnetite) and could possibly represent the coarser-grained end member within the magnetic mineral assemblage.

Several of the identified reversal boundaries in the studied sequence coincide with changes in mineral magnetic properties (e.g., grain size and/or concentration) (Fig. 2). These observations suggest that the sediments have been remagnetized at one or several occasions through authigenic greigite growth. The presence of greigite magnetofossils and/or titanomagnetite grains in A2 (and possibly also A1) suggests that part of the paleomagnetic signal in these sediments could be primary or acquired shortly after deposition. Most likely, the dual component magnetization observed in A1 and the upper part of A2 is due to an initial low-coercivity, normal-polarity component that is overprinted by a higher-coercivity, authigenically grown greigite component that carries a reversed polarity signal.

Bulk sediment XRD analyses reveal the presence of siderite, but only small amounts of pyrite in the clay-dominated sediments, except for the section associated with A5, which contains both siderite and pyrite (Mountain et al., 2010). Pyrite concentrations, which are strongly correlated with total sulfur, are generally high in other finer-grained sediments in both M0027A and M0028A (Mountain et al., 2010). The relatively low concentration of pyrite in the studied clays suggests that the pyritization process was inhibited or arrested shortly after burial, possibly due to an excess of reactive iron compared to available organic carbon, or to the low permeability of the clays, which resulted in consumption of all available H₂S to form greigite before pyrite could form (Kao et al., 2004; Roberts and Turner, 1993). The formation of siderite indicates continued iron reduction in a carbonate-rich nonsulfidic (Berner, 1981) or sulfide-poor (Pye et al., 1990) anoxic environment. This observation supports the suggestion of Kao et al. (2004) that greigite is preserved when dissolved iron content is high but dissolved sulfide content is low.

Stable carbon isotope and gas analyses have revealed the existence of methane production at depths below 350 m below seafloor (van Geldern et al., 2013). Methanogenesis was also suspected at ODP Leg 174A, Sites 1071 and 1072, based on analyses of carbonate concretions (Malone et al., 2002). However, methane was largely absent in the recovered sediments and must have either oxidized or vented, perhaps related to sea-level fluctuations. Anaerobic oxidization of such upward-migrating methane, trapped in the upper unit II clay-rich horizon, could perhaps provide a mechanism for late authigenic greigite formation (see Larrasoaña et al., 2007).

Pore-water chemistry measurements indicate that the studied clay sediments are located at the upper (M0027A) and lower (M0028A) edges of large distinct freshwater lenses (Mountain et al., 2010; van Geldern et al., 2013). The freshwater is preferentially stored in fine-grained marine sediments, indicating that the lenses are postdepositional features (Lofi et al., 2010). Stable isotope ratios are identical to those from modern precipitation in New Jersey (van Geldern et al., 2013) and suggest that the water originated from onshore meteoric water, either through recharge via aquifers on mainland New Jersey or emplacement at a time with climatic and hydrologic conditions similar to today.

There are several mechanisms through which sediments might be remagnetized due to the greigite growth triggered by changes in pore-water chemistry (Roberts and Weaver, 2005). For example, Oda and Torii (2004) observed an increase in greigite diagenesis on the New Jersey continental shelf associated with sea-level lowstands. This was interpreted to be due to oxidation of pyrite by downward-percolating freshwater. Sagnotti et al. (2005) in turn were able to explain a zone of mixed polarity (similar to the one in M0028A) by greigite growth from siderite cement, which provided reactive iron to form greigite when dissolved sulfide became available due to late diagenetic pore-water migration. It is likely that the sediments from this study may have experienced similar diagenetic greigite growth phases related to variable influx of freshwater during large-amplitude sea-level fluctuations. Lateral interactions between fresh groundwater and saline pore water will create sulfate concentration gradients that could promote greigite growth (Roberts and Weaver, 2005). Small but detectable amounts of SO₄^2– (∼5 mmol/L) in the pore water associated with the clay layers suggest that the conditions for greigite formation, and therefore remagnetization, still exist. Greigite growth would most likely also be mediated by high microbial activity (such as sulfate reduction) along the fresh-saline gradient at the edges of the lenses. It is possible that movement of these fronts over time could be responsible for episodes of greigite growth, as revealed by the different mineral magnetic assemblages. However, we also note that the identified mineral magnetic assemblages could be a result of variations in the initial composition of the sediments, e.g., the availability of reactive iron and organic matter.

The magnetic mineral assemblages of Miocene clay sediments in upper unit II at IODP Holes M0027A and M0028A are dominated by greigite. Although we find some evidence for a primary magnetic phase (A2), the paleomagnetic signal is most likely a record of one or several remagnetization events. It is not possible to establish the timing of these events (and remagnetization may still be ongoing), but we conclude that they are likely associated with sea-level fluctuations. The occurrence of freshwater lenses in the sediments provides evidence for major variations in the pore-water chemistry that could have triggered episodes of authigenic greigite growth. Anaerobic oxidation of upward-migrating methane, produced in deeper sediments, may represent an alternative mechanism for late greigite formation. Based on these observations, we conclude that the initial magnetic polarity stratigraphy (Fig. 2) should not be used in the construction of an age-depth model for the sediments.

This work was partially funded by the Natural Environment Research Council, UK. Samples were provided through the Integrated Ocean Drilling Program. We thank Thomas Frederichs for assistance with facilities at the University of Bremen paleomagnetic laboratory; Jennifer Inwood for helpful discussions; and the drillers, platform staff, and scientists of IODP Expedition 313 for great collaboration. This work was partially funded by the Natural Environment Research Council, UK, grant number NE/I013873/1. We also thank Andrew P. Roberts and an anonymous reviewer for their constructive reviews, which significantly improved the manuscript.