Sediment core GH98-1232 from the northern part of the Japan Sea at 44°48.09’N, 139°41.97’E was used for this study (Fig. 1). The water depth of the coring site is 838 m, and the present-day Fe-redox boundary occurs a few centimeters below the sediment-water interface, which is recognized from a brown-to-gray change in sediment color. A rock-magnetic study by Yamazaki et al. (2003) revealed that reductive dissolution of magnetic minerals starts at depths of about 1.15 m in this core and is mostly completed within an interval of about 0.35 m (Fig. 2). Magnetic concentration-dependent parameters decrease rapidly downcore within this interval with the anhysteretic remanent magnetization (ARM) decreasing first, followed by the saturation isothermal remanent magnetization (SIRM), and then the low-field magnetic susceptibility. Average magnetic grain size estimated from the ratio of ARM susceptibility (kARM) to SIRM rapidly increases downcore from 1.18 to 1.35 m, which indicates that finer magnetic grains are lost before larger grains. Average magnetic grain size then decreas a little between 1.35 and 1.6 m. Based on S-ratio and low-temperature measurements, magnetite dominates the magnetic mineral assemblage of surface sediments to the depth which dissolution initiates, and the relative proportion of high-coercivity minerals (e.g., hematite) increases with depth as magnetite dissolves preferentially in this interval (Yamazaki et al. 2003).
FORC diagrams
FORC diagrams are used widely in rock magnetism to elucidate the distribution of coercivities (Hc) and magnetostatic interactions (Hu) within magnetic particle assemblages, from which information on domain states, grain sizes and shapes, mineralogy, and spatial distribution of constituents can be obtained (e.g. Pike et al. 1999; Roberts et al. 2000, 2014). FORC diagrams are particularly useful for detecting biogenic magnetite in sediments; intact chains of biogenic magnetite behave as isolated non-interacting single-domain (SD) grains, which produce a characteristic narrow peak that extends along the coercivity axis at zero interaction field on a FORC diagram, called the “central ridge” (Egli et al. 2010; Li et al. 2012; Roberts et al. 2012; Yamazaki and Ikehara 2012; Chang et al. 2014).
FORC measurements were conducted at ~5 cm stratigraphic intervals within and near the magnetic-mineral dissolution zone using an alternating-gradient magnetometer (Princeton MicroMag 2900) at the Geological Survey of Japan (GSJ), National Institute of Advanced Industrial Science and Technology (AIST). A total of 167 FORCs were measured for each sample, with Hc between 0 and 100 mT, Hu between -50 and 50 mT, and field spacing of approximately 1.3 mT. The maximum applied field was 1.0 T, and the averaging time for each measurement point was 200 ms. The FORCinel software (Harrison and Feinberg 2008, version 3.05 in 2019) was used to produce diagrams, and the VARIFORC algorithm of Egli (2013) was used to smooth the data with smoothing parameters of Sc0=4, Sb0=3, Sc1=Sb1=7.
FORC diagrams above the magnetic-mineral dissolution zone (1.00 and 1.14 m) have a sharp central ridge feature along the Hu=0 axis, which indicates the contribution of non-interacting SD grains (Fig. 3a). This SD signature is interpreted to be of biogenic origin. The FORC diagrams also have broad vertical spread with an elliptical peak at a coercivity of 10–15 mT, and outer contours diverge from the Hu=0 line toward the Hu axis. This component is considered to be carried by a mixture of interacting SD, vortex state, and multi-domain (MD) grains (Roberts et al. 2000, 2014, 2017; Lascu et al. 2018), which are interpreted to be of terrigenous origin. Collapsed chains of biogenic magnetite may also be included in this component (Li et al. 2012; Harrison and Lascu 2014). The contribution of the central ridge component decreases with depth within the dissolution zone, which is evident as a diminishing extracted central ridge signal with depth (Fig. 3b) and a downcore decrease of the proportion of the central ridge component to the broader, more vertically spread components on profiles along Hc=40 mT (Fig. 3c) and along Hu=0 (Fig. 3d). These observations indicate selective dissolution of biogenic magnetite. Coercivity distributions of both the biogenic and terrigenous components shift to lower values along with progressive reductive diagenesis (Fig. 3d); the peak coercivity of the biogenic component, initially ~45 mT, decreases to ~25 mT. This is probably due to decreasing grain volumes with dissolution. For the sample from the base of the dissolution zone at 1.49 m, vertical spread becomes somewhat narrower and the proportion of the central ridge component increases a little compared with the sample from 1.36 m.
TEM observations
TEM observations provide a direct way to investigate biogenic magnetite in sediments, which is complementary to FORC measurements. Biogenic magnetite can be identified directly in TEM images from its characteristic morphologies (bullet shaped, hexagonal prisms, and octahedra) and uniform SD grain sizes (Bazylinski et al. 1994; Kopp and Kirschvink 2008). Particle morphology is not known from FORC distributions, although particle elongation may be estimated from coercivity distributions (Egli 2004; Yamazaki and Ikehara 2012; Usui et al. 2017; Yamazaki et al. 2020). The limitation of TEM observations is, on the other hand, that particle population under a TEM may not faithfully represent the original magnetic-mineral assemblage in a sample because magnetic-mineral extraction is required for observations. In this context, FORC distributions provide more quantitative information.
Magnetic minerals extracted from seven different depths in the core, 1.00, 1.14, 1.18, 1.23, 1.27, 1.36, and 1.49 m, were observed using a TEM. To extract magnetic minerals, sediments were first dispersed in distilled water with sodium hexametaphosphate using an ultrasonic bath, and magnetic minerals were then collected by circulating the dispersed sediments through a high magnetic-field gradient. The magnetic extracts were dispersed in ethanol, and a small drop of the suspension was subsequently dried on a carbon-coated copper grid. A TEM (JEOL JEM-1400) at the Atmosphere and Ocean Research Institute, The University of Tokyo, operated at 120 kV, was used for the observations.
From morphologies seen in TEM images, magnetofossils were classified into three groups: bullet shaped, elongated, and equant (Fig. 4), although some ambiguity remains in estimating three-dimensional morphology from an image projected on a plane. The elongated group consists of hexagonal prisms and elongated octahedra. The equant group is represented by cubo-octahedra, although short hexagonal prisms with length/width ratios close to one are also included in this group. Bullet-shaped magnetofossils can be identified easily from their morphology. Relative abundances of biogenic magnetite from each morphological group and their grain-size distributions were obtained by counting several hundred grains in about 100 images for each sample. It should be noted that this kind of analysis remains semi-quantitative.
The sediments before reductive dissolution starts contain abundant magnetofossils of all three morpho-types (Fig. 4a), as reflected by the wide coercivity distribution along the central ridge of the FORC diagrams, from about 20 to 80 mT (Fig. 3). At the beginning of the reductive dissolution interval at 1.14 m, most of the biogenic magnetite has sharp crystal edges with little sign of dissolution in TEM images (Fig. 4b). However, some bullet-shaped magnetofossils have wavy crystal surfaces (Fig. 4c, arrows), which are indicative of partial dissolution. For the samples from 1.18 and 1.23 m, at which dissolution is underway (Fig. 2), many crystals are partially etched (Figs. 4d–4g). It is often observed that the caps of hexagonal prisms, which are constituted by {111} faces (Mann 1985; Meldrum et al. 1993), are etched while side faces are almost intact (Figs. 4d, and 4f). The wavy corrosion also affects bullet-shaped magnetofossils (Fig. 4e). Corroded equant magnetofossils have rounded shapes in general, which suggests that octahedral edges are corroded (Figs. 4g and 4h). In the lower part of the dissolution interval at 1.27 m, the characteristic morphologies of biogenic magnetite are almost lost (Figs. 4i and 4j), although some particles are inferred to be of biogenic origin from their sizes within or near the SD range. At 1.36 and 1.49 m near the base of the dissolution interval, no magnetofossil remains. Some grains that resemble silicate-hosted magnetic inclusions (Chang et al. 2016a; Zhang et al. 2018) are observed (Figs. 4k and 4l).
Length/width distributions for each morpho-type of magnetofossils determined in TEM images at four depths above and within the dissolution zone are shown in Fig. 5. Bullet-shaped magnetofossils are lost before others, and magnetofossils of the elongated group are most resistant to reductive dissolution. The proportion of the bullet-shaped group is ~40% before dissolution starts, and decreases to 14% at 1.23 m. On the other hand, the proportion of the elongated group increases from 38 to 63%.