Widespread occurrence of silicate-hosted magnetic mineral inclusions in marine

sediments and their contribution to paleomagnetic recording

Liao Chang1,2, Andrew P. Roberts2, David Heslop2, Akira Hayashida3, Jinhua Li4, Xiang

Zhao2, Wei Tian1, and Qinghua Huang1

1. School of Earth and Space Sciences, Peking University, Beijing 100871, P. R. China

2. Research School of Earth Sciences, The Australian National University, Canberra,

ACT 2601, Australia

3. Department of Environmental Systems Science, Doshisha University, Kyotanabe,

Kyoto 610–0321, Japan

4. Paleomagnetism and Geochronology Laboratory, Key Laboratory of Earth’s Deep

Interior, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing

100029, P. R. China

Abstract Magnetic mineral inclusions occur commonly within other larger mineral

phases in igneous rocks and have been demonstrated to preserve important paleomagnetic

signals. While the usefulness of magnetic inclusions in igneous rocks have been explored

extensively, their presence in sediments has only been speculated upon. The contribution of

magnetic inclusions to the magnetization of sediments, therefore, has been elusive. In this

study, we use transmission electron microscope (TEM) and magnetic methods to demonstrate

the widespread preservation of silicate-hosted magnetic inclusions in marine sedimentary

settings. TEM analysis reveals detailed information about the microstructure, chemical

composition, grain size, and spatial arrangement of nanoscale magnetic mineral inclusions
This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process which may
lead to differences between this version and the Version of Record. Please cite this article as
doi: 10.1002/2016JB013109

© 2016 American Geophysical Union. All rights reserved.

within larger silicate particles. Our results confirm the expectation that silicate minerals can

protect magnetic mineral inclusions from sulfate-reducing diagenesis, and increase

significantly the preservation potential of iron oxides in inclusions. Magnetic inclusions

should, therefore, be considered as a potentially important source of fine-grained magnetic

mineral assemblages, and represent a missing link in a wide range of sedimentary

paleomagnetic and environmental magnetic studies. In addition, we present depositional

remanent magnetization (DRM) modeling results to assess the paleomagnetic recording

capability of magnetic inclusions. Our simulation demonstrates that deposition of larger

silicate particles with magnetic inclusions will be controlled by gravitational and

hydrodynamic forces rather than by geomagnetic torques. Thus, even though these large

silicates may contain ideal single domain particles, they can not contribute meaningfully to

paleomagnetic recording. However, smaller silicate grains (e.g., silt- and clay-sized) silicates

with unidirectionally magnetized magnetic inclusions can potentially record a reliable DRM.

Keywords: magnetic mineral inclusions, silicates, marine sediments, detrital remanent

magnetization, transmission electron microscope

1. Introduction

Magnetic iron-titanium oxide mineral inclusions hosted within silicate minerals, e.g.,

plagioclase (NaAlSi3O8 to CaAl2Si2O8) and clinopyroxene ((Ca, Mg, Fe)2Si2O6), occur

widely in igneous and metamorphic rocks. The importance of magnetic inclusions for

paleomagnetic studies has been recognized since the discovery of ultrafine-grained iron

oxides in silicates in the Modipe Gabbro [Evans et al., 1968; Evans and Wayman, 1970].

More recently, there has been renewed interest in magnetic mineral inclusions due to the

development of single-crystal based paleomagnetic analysis and the much increased

© 2016 American Geophysical Union. All rights reserved.

sensitivity of modern cryogenic magnetometers [e.g., Cottrell and Tarduno, 1999; Tarduno et

al., 2001, 2006, 2010; Muxworthy and Evans, 2013; Sato et al., 2015]. Paleomagnetic signals

from magnetic mineral inclusions have been investigated extensively to expand knowledge of

past magnetic field behavior of the Earth and other bodies within the Solar System [e.g.,

Tarduno et al., 2001, 2010, 2015; Feinberg et al., 2005, 2006; Lappe et al., 2011, 2013;

Muxworthy et al., 2013; Usui et al., 2015]. Unlike magnetically unstable multi-domain (MD)

particles, magnetic mineral inclusions often occur as fine-grained stable single-domain (SSD)

or small pseudo-single-domain (PSD) particles [e.g., Harrison et al., 2002] that are capable

of carrying stable remanences over billions of years [Evans et al., 1968]. Moreover, silicate

host minerals can protect magnetic inclusions against changes in the local environment that

can give rise to chemical alteration. These characteristics make silicate hosted magnetic

mineral inclusions a promising candidate for retaining reliable paleomagnetic signals over

long geological timescales [e.g., Tarduno et al., 2006, 2010].

The presence of magnetic mineral inclusions in igneous rocks and their important

contributions to paleomagnetic records are well established. But knowledge of the presence

of magnetic inclusions within detrital particles in sediments and understanding their

contribution to sedimentary magnetic signals have been elusive. In this study, we investigated

marine sediment samples to search for magnetic mineral inclusions using transmission

electron microscope (TEM) and magnetic analyses. We also modeled the depositional

remanent magnetization (DRM) of magnetic inclusions to assess their paleomagnetic

recording capability. Potential implications of magnetic inclusions for sedimentary

paleomagnetic studies are discussed.

2. Materials and methods

© 2016 American Geophysical Union. All rights reserved.

 Marine sediment samples from a range of deep-sea sediment cores and sediment

outcrops were investigated in this study (Fig. 1). Core MD01-2421 (36°01.4′N; 141°46.8′E;

2,224 m water depth; 45.82 m long) was recovered from the North Pacific Ocean ~100 km

offshore of central Japan [Oba et al., 2006; Chang et al., 2016b]. Sediments in this core are

homogenous olive-gray silty clays with calcareous and siliceous microfossils, with high total

organic carbon (TOC) content (0.5-2.1 wt.%; Ueshima et al. [2006]). Sample “MD01-2421-

7-110” from a depth of 110 cm in core section 7 (at a depth of 10.06 m) was analyzed. The

studied sample was selected from bulk sediment with no evident disseminated volcanic ash.

Core CD143-55705 (22°22.4′ N, 60°08.0′ E; 2,193 m water depth; 10.63 m long) was

recovered on the continental margin of Oman, northwestern Arabian Sea [Rowan et al.,

2009]. The recovered sediments are homogeneous, light green-brown to gray-green

hemipelagic clays, and are high in TOC (1-2%) [Rowan et al., 2009]. Sample “CD143-

55705-7-82” from a depth of 82 cm in core section 7 (at 7.49 m depth) was analyzed. Central

equatorial Pacific Ocean sediment core RR0603-03JC (2°33′ N, 117°55′ E; 4195 m water

depth) was recovered during the AMAT03 site survey cruise for Integrated Ocean Drilling

Program Proposal 626. The lithology of this core is mainly diatom nannofossil ooze. The

studied sample “RR0603-03JC-2-60” is from a depth of 60-62 cm in core section 2 (at 1.04

m depth). Dust may be an important component of the studied samples from the equatorial

Pacific Ocean and Arabian Sea. Marine sediment samples (magnetic separate sample

“BL37,38,39” and bulk sediment sample “BR49D” that are close to each other

stratigraphically) were collected from tectonically uplifted Upper Miocene marine sediments

exposed in Blind River, Lower Awatere Valley, northeastern South Island, New Zealand

[Roberts and Turner, 1993]. The succession contains siliciclastic marine sediments of the

Awatere Group and are probably derived from greywacke basement rocks and igneous

© 2016 American Geophysical Union. All rights reserved.

sources in central Marlborough. Key information for the studied samples is summarized in

Table 1.

Hysteresis parameters (Table 1) and first-order reversal curve (FORC) measurements

were made with a MicroMag vibrating sample magnetometer (model 3900) at the Australian

National University (ANU). FORC measurements [Roberts et al., 2000, 2014] were made

with a field step of 1.5 mT, maximum applied fields of 1 T, and averaging times of 200-400

ms. For some magnetically weak samples, we followed the protocol of Zhao et al. [2015],

in which 120-160 FORCs with irregular measurement grids were measured with averaging

times of 200-400 ms. FORC data were processed using the software package of Zhao et al.

[2015]. No data pre-treatments, i.e., removal of first-point artefact and subtraction of lower

branch [Egli, 2013], were applied. Low-temperature (LT) magnetic properties were measured

with a Quantum Design Magnetic Property Measurement System (MPMS; model XL7) at

ANU. For warming of a saturation isothermal remanent magnetization (SIRM), samples were

first cooled to 10 K in either zero field (zero-field cooled; ZFC) or in a 5 T field (field-

cooled; FC). At 10 K, a 5 T field was applied and was then switched off to impart a LT SIRM,

and the MPMS magnet was reset. ZFC and FC curves were measured during zero-field

warming in a sweep mode at 5 K/min.

Magnetic minerals were separated from bulk sediments following Chang et al. [2012]

using a Frantz isodynamic magnetic separator. TEM observations were carried out with a

JEOL 2100F field-emission (FE) TEM and a Philips CM300 TEM at the Centre for Advanced

Microscopy (CAM), ANU, and with a JEOL 2100 TEM at the Institute of Geology and

Geophysics, Chinese Academy of Sciences (CAS). The JEOL 2100F at CAM is equipped

with a FE gun and scanning transmission electron microscope (STEM) detectors, and is

operated at 200 kV. STEM observations were performed in the high-angle annular dark field

(HAADF) mode. Energy dispersive X-ray spectroscopy (EDS) analysis was performed using

© 2016 American Geophysical Union. All rights reserved.

a Silicon Drift Detector (SDD) with an ultrathin Be window. EDS maps were acquired in the

STEM HAADF mode, with a focused electron beam of a few nm. The Philips CM300 TEM

at CAM is equipped with an EDAS Phoenix retractable X-ray detector and a Gatan CCD

camera, and is operated at 300 kV. The JEOL 2100 TEM at CAS was operated at 200 kV.

To model DRM, magnetic and hydrodynamic torques that act on a settling detrital

particle were compared to assess the ability of magnetic inclusion-bearing particles to align

with the geomagnetic field. Detrital sediment particles that contain inclusions are assumed to

be prolate ellipsoids. The aspect-ratio (ratio of the semi-major axis to the semi-minor axis) of

such particles plays a key role in controling their orientation as they settle through the water

column. Spherical particles will experience no shape-induced hydrodynamic torque; however,

as the aspect-ratio of a particle increases, the hydrodynamic torque also increases and tends to

rotate a settling ellipsoid so that its long axis is horizontal. The approximation provided by

Heslop [2007] was employed to find the maximum hydrodynamic torque, τH, that acts on a

settling particle with a given volume and aspect-ratio. Magnetic nano-inclusions within

detrital particles are assumed to be SSD magnetite particles with diameters of 100 nm that are

aligned along a single preferred crystallographic direction. To represent the magnetization of

the magnetite assemblage, we assume that the particles carry a weak-field thermoremanent

magnetization (TRM). Dunlop [1990] demonstrated that weak-field TRM in magnetite varies

as a function of particle size, with an assemblage of randomly oriented 100-nm particles

acquiring a TRM of ~10 kA/m in a 100-µT field. We assume that the TRM intensity is

proportional to field strength, so we scale this empirical value for a typical geomagnetic field

strength of 50 µT. Finally, we multiply the resulting TRM by a factor of 2 to remove the

partial cancelation that occurs over a collection of randomly oriented particles. This process

yields an estimated TRM of 10 kA/m for aligned particles in a 50-µT field. Assuming that

TRM-bearing crystallographically aligned SSD magnetite particles make up a given volume

© 2016 American Geophysical Union. All rights reserved.

percentage of the host sediment particle, it is possible to estimate the maximum magnetic

torque, τM, experienced by a sediment particle as it settles through a water column in an

ambient 50-µT field.

3. Results

3.1. Magnetic properties

Magnetic measurements, including FORC and LT magnetic measurements (Fig. 2),

were made on bulk sediment samples (sister samples of the studied TEM samples) and a

magnetic separate sample “BL37,38,39” to characterize the constituent magnetic minerals

and to constrain results from TEM observations. The FORC diagram for sample “MD01-

2421-7-110” indicates a dominantly SSD signature with moderate magnetostatic interactions

(Fig. 2a). LT warming of SIRM reveals a weak double Verwey transition (Tv) signature (Fig.

2b) that is indicative of the presence of small amounts of both biogenic and inorganic

magnetite [e.g., Chang et al., 2016a], where the more pronounced Tv at ~120 K is mostly

likely to be associated with the presence of detrital magnetite. It should be noted that the

double Tv signature is not observed commonly for other sediment samples from core MD01-

2421 [Chang et al., 2016b]. The FORC diagram for sample “RR0603-03JC-2-60” from the

central equatorial Pacific Ocean contains a dominant central-ridge signature associated with

non-interacting SD particles, superposed on a weak background SD signal with stronger

interactions (Fig. 2c). LT magnetic measurements for this sample did not reveal a clear Tv

signal (Fig. 2d). The FORC diagram for sample “CD143-55705-7-82” from the Oman

continental margin contains a SD component with weak to moderate magnetostatic

interactions (Fig. 2e). This sample has a pronounced Tv at ~120 K (Fig. 2f). Sample “BR49D”

from the Lower Awatere Valley, New Zealand, has two major FORC components: a SD

distribution with weak to moderate magnetostatic interactions and a MD component with

© 2016 American Geophysical Union. All rights reserved.

vertical spreading along the Bu axis at low coercivities (Fig. 2g). LT data for this sample

reveal a pronounced Tv at ~120 K (Fig. 2h). The FORC diagram for magnetic separate sample

“BL37,38,39”, which is from an outcrop close to that from which sample BR49D was taken,

contains two major FORC distributions: a SD component with moderate vertical spread and a

low-coercivity component (Fig. 2i). We plot the coercivity profiles from FORC diagrams

(Fig. 2j) for the studied samples and compare them with published results for some biogenic

magnetite samples [Roberts et al., 2013], and magnetic inclusion-bearing igneous rocks

[Muxworthy and Evans, 2013; Usui et al., 2015]. The Bc profile for sample “MD01-2421-7-

110” is broader and extends to larger fields compared to other samples, while profiles for

other samples containing magnetic nano-inclusions appear to be similar to those for biogenic

magnetite.

3.2. TEM observations of magnetic mineral inclusions from core MD01-2421

3.2.1. Microstructures and crystallographic orientations of magnetic mineral inclusions

TEM analysis of magnetic extracts reveals abundant detrital magnetic particles with

variable grain sizes, which are probably from igneous lithic fragments sourced from Japan.

Large particles are too thick for electron transmission and appear dark under bright field

TEM observations. We selected thinner edge areas from large particles and small particles for

our detailed TEM and TEM-EDS analysis. This approach does not enable observations of

nanoparticles that occur deeper within large silicate grains. TEM observations reveal

abundant nano-sized magnetic mineral inclusions (Fig. 3) that were difficult to be observed

from scanning electron microscope observations. The nano-sized magnetic minerals must be

embedded within host minerals, rather than being attached to particle surfaces, which were

clearly visualized by HAADF-STEM imaging due to their different chemical contrast to host

minerals (Fig. 3a-d). The nanoparticle inclusions could also be observed by bright-field TEM

© 2016 American Geophysical Union. All rights reserved.

imaging in the cases of relatively thin or small host minerals (Fig. 3e-i). In both cases, there is

a clear contrast shift under TEM between inclusions and hosts from the edge to interior due to

variable inclusion depths within the respective host particles. Observed nanoparticle

inclusions have sizes that range from a few nm to several hundred nm with variable

morphologies. We observed three main types of inclusion microstructure: nanoparticle

clusters (Fig. 3a-c, h), dendrites (Fig. 3d-f), and crystallographically oriented nanoparticles

(Fig. 3g). The nanoparticle clusters consist of euhedral octahedral, sub-rounded, and

irregularly shaped crystals (Fig. 3a-c, h). Many of the nanoparticles are nearly isotropic or are

slightly elongated, some of which are closely packed (Fig. 3b, h). We observed less abundant

large magnetic mineral inclusions (i.e. ~1 m; Fig. 3i). The observed dendrites have complex

microstructures with variable one-, two-, and three-dimensional structures (Fig. 3d-f). Some

nanoparticles are assembled along specific crystallographic directions (double-headed arrows

in Fig. 4a, e, f). It is possible that some of the oriented nanoparticles represent arrested

dendritic growth. In contrast, some nanoparticle clusters appear to be more randomly

distributed within host crystals (Fig. 4h, i). We carried out selected area electron diffraction

(SAED) analysis with the TEM stage tilted at different angles (Fig. 4a-d) to determine the

crystallographic orientation of aligned inclusions. During tilting, collective diffraction of

nanoparticles (i.e., appearance (Fig. 4a-c, black) and extinction (Fig. 4d) of nanoparticles)

occurs simultaneously. This behavior confirms a preferential alignment of nanoparticle

inclusions within the host crystal, which is further demonstrated by spot-like SAED patterns

(insets in Fig. 4a-c). In contrast, some nanoparticle clusters have ring-like diffraction patterns

(inset in Fig. 4g), which indicate a more random distribution of inclusion orientations.

3.2.2. High-resolution TEM (HRTEM) analysis of magnetic mineral inclusions and host

minerals

© 2016 American Geophysical Union. All rights reserved.

 Further HRTEM and SAED analyses (Fig. 5) indicate that all analyzed magnetic

nanoparticle inclusions, including clusters and dendrites, have clear lattice fringes (Fig. 5d,

h, j, l, n, p) and strong diffraction patterns (inset in Fig. 5b), which indicate good crystallinity.

The observed d-spacing values and diffraction patterns for the inclusions match well the

crystal structure (Fd3m space group) of magnetite and titanomagnetite. The observed lattice

fringes from magnetic mineral inclusions do not reveal signs of crystal defects (Fig. 5d, 5h,

5j, 5l, 5n, 5p, 6c, 6d, 6h). However, we occasionally observed titanomagnetite nanoparticle

inclusions with crystal twinning (arrows in Fig. 6g). HRTEM and SAED analyses indicate

that the host minerals are also crystalline, as evidenced by HRTEM lattice images and SAED

patterns (Fig. 6e, f, i-l). However, the host minerals generally have weaker diffraction and

less clear lattice patterns compared to the inclusions (Fig. 6f, j). This may be attributed to

variable extents of destruction of crystal structures that were observed after a few seconds of

electron beam radiation (Fig. 6j).

3.2.3. Chemical compositions

EDS mapping (Fig. 7a-l) and point analyses (Fig. 7m-u) were carried out to determine

the chemical composition of inclusions and host crystals. EDS mapping for one area (Fig.

7a) indicates that the magnetic mineral inclusions are rich in Fe (Fig. 7b) with a much smaller

Ti concentration (Fig. 7c). The host mineral is rich in O (Fig. 7d), and Si (Fig. 7e), and also

contains a small concentration of Al (Fig. 7f) and Ca (data not shown). EDS mapping of

another area (Fig. 7g) indicates similar characteristics, where the inclusions are Fe rich (Fig.

7h), and lacking in Ti (Fig. 7i), and the host contains O, Si, and Al (Fig. 7j-l), and Ca (data

not shown). EDS spectra of host minerals (Fig. 7m, p, s) indicate the presence of Si, O, Al,

and Ca peaks. EDS spectra of inclusions also contain these elements because they are

embedded in host mineral grains, but the EDS spectra of inclusions (Fig. 7n, o, q, r, t, u)

© 2016 American Geophysical Union. All rights reserved.

contain much higher concentrations of Fe and minor Ti compared to those of the hosts. The

relative intensities of Fe and Ti peaks are variable (ratios are indicated in Fig. 7n, q, r, t, u),

but Ti contents are small. EDS mapping and point analyses, therefore, consistently indicate

that the magnetic mineral inclusions have chemical compositions that are consistent with

those in the magnetite-ulvöspinel solid solution series (mainly Ti-poor titanomagnetite). Most

of the analyzed host silicate minerals (containing O, Si, Al, and Ca; Fig. 7) are plagioclase

feldspar. Occasionally, we observed silicate host minerals with O, Si, Ca, Mg, and Fe peaks

(data not shown), which are likely to be clinopyroxene.

3.3. TEM observations of marine sediment samples from other localities

We investigated samples from other localities to test whether magnetic mineral

inclusions are commonly present in marine sediments. TEM observations of a marine

sediment sample “CD143-55705-7-82” from the Oman continental margin reveal the

presence of crystallographic orientated small needle-like magnetic minerals within silicates

(Fig. 8a). This microstructure is consistent with that of exsolved magnetite in igneous rocks.

The presence of magnetic mineral inclusions in samples from core CD143-55705 from the

Arabian Sea confirms expectations from magnetic analyses (Fig. 2e, f; Chang et al. [2016a]).

TEM analysis on sample “RR0603-03JC-2-60” indicates the presence of abundant biogenic

magnetite crystals, as evidenced by apparently intact magnetosome chain structures and well-

defined magnetosome crystal morphologies (Fig. 8b). The TEM results are consistent with a

FORC diagram from the same bulk sediment (Fig. 2c), which has a strong central-ridge

signature [e.g., Egli et al., 2010; Roberts et al., 2012; Chang et al., 2014]. Despite the

occurrence of biogenic magnetite in sample “RR0603-03JC-2-60”, we also observed silicate

hosted titanomagnetite nanoparticles in this sample (Fig. 8c). The magnetic separate sample

“BL37,38,39” from outcrops in New Zealand contains abundant silicate particles. Detailed

© 2016 American Geophysical Union. All rights reserved.

TEM analysis of the silicate particles often reveals the presence of Fe-Ti oxide nanocrystals

(magnetite or titanomagnetite). Most of the magnetic inclusions appear to be randomly

distributed within the silicate hosts (Fig. 8d), although possible dendritic titanomagnetite

structures are observed (Fig. 8e). The host silicate minerals within sample “BL37,38,39”

often have rough surfaces, with chemical compositions that are consistent with silicates with

major Si and O peaks, and minor Al, Ca, Na, Fe, or Mg peaks in the EDS spectra.

3.4. Numerical DRM modeling of silicate-hosted magnetic inclusions

To make a first-order assessment of the ability of an inclusion-bearing detrital

sediment particle to align with the ambient geomagnetic field, we compare τH and τM for a

range of equivalent particle diameters (the diameter of a sphere that has the same volume as

the ellipsoid under consideration). To illustrate the relationship between the competing

hydrodynamic and magnetic torques, we calculate the aspect-ratio for a sediment particle

with a given effective diameter and magnetite volume percentage at τH = τM (Fig. 9). Our

calculation indicates that it would require unrealistically high aspect-ratios to achieve τH = τM

in smaller particles (i.e., the shape-induced hydrodynamic torque must be increased to

infeasibly high levels to achieve parity with the magnetic torque). Thus, sediment particles in

this size range will be dominated by magnetic torques and could contribute to a reliable

sedimentary paleomagnetic signal. In contrast, larger sediment particles must have aspect-

ratios close to 1 to achieve τH = τM (i.e., the shape-induced hydrodynamic torque must be

suppressed substantially to achieve parity with the magnetic torque), which is again

unrealistic for natural particles. Therefore, larger particles will be dominated by

hydrodynamic torques, which will restrict their ability to record reliably the ambient

geomagnetic field. The shift from dominance of a magnetic to a hydrodynamic torque in our

© 2016 American Geophysical Union. All rights reserved.

simple model occurs over a narrow size window within the fine to medium silt size range.

The importance of our model results are discussed further below.

4. Discussion

4.1. Widespread occurrence of silicate-hosted magnetic inclusions in marine sediments

Detrital minerals sourced from continents are important constituents of marine

sediments. Our detailed TEM observations indicate that magnetic nanoparticle inclusions are

widely present in marine sediments. Magnetic inclusions can even dominate the magnetic

signal (Fig. 2a) [Chang et al., 2016b]. The observed magnetic mineral microstructures have

two main origins due to exsolution and inclusion (see Tarduno et al. [2006] for a discussion).

Dendrites and crystallographically oriented nanoparticles are exsolved microstructures in

silicates, which form due to phase separation in an originally homogenous solid solution

during initial cooling of igneous rocks. Inclusions, such as euhedral crystals, in contrast, form

prior to the host silicate minerals and are incorporated into the host mineral during its

subsequent crystallization. Preservation of magnetic mineral inclusions in marine sediments

is not surprising. Silicate minerals that host magnetic mineral inclusions occur widely in

igneous rocks [e.g., Evans et al., 1968; Evans and Wayman, 1970; Haggerty, 1991; Feinberg

et al., 2006; Wakabayashi et al., 2006], so it is to be expected that such particles will occur as

detrital grains in sedimentary strata. However, magnetite is a mixed valence iron oxide

mineral that is unstable in both oxidizing and reducing sedimentary environments [Roberts,

2015]. In particular, magnetic iron oxide minerals will undergo dissolution during sulfate-

reducing diagenesis that results in significant depletion of these minerals and formation of

iron sulfide minerals [Roberts, 2015]. Unprotected iron oxides, such as coarse-grained

magnetic minerals and fine-grained biogenic magnetite, are prone to rapid dissolution in

sulfate-reducing marine environments [Karlin and Levi, 1983; Canfield and Berner, 1987;

© 2016 American Geophysical Union. All rights reserved.

Rowan et al., 2009; Chang et al., 2016a, b]. The studied sediments from New Zealand and

the Japan and Oman margins have undergone extensive sulfidic diagenesis that has removed

much of the magnetite signal [Roberts and Turner, 1993; Rowan and Roberts, 2006; Rowan

et al., 2009; Chang et al., 2016a, b]. In contrast, silicate minerals are relatively stable against

reductive diagenesis in marine sedimentary environments [Canfield and Raiswell, 1991;

Poulton et al., 2004; Roberts, 2015]. Protection from diagenesis by host silicate crystals will

increase the preservation potential of magnetic minerals. Such a protection mechanism is

likely to explain the preservation of detrital magnetic minerals in diagenetically altered

marine sediments, where magnetite dissolution is expected to be pervasive [Roberts, 2015].

4.2. Identification of magnetic mineral inclusions in sediments

Magnetic mineral inclusions in sediments are important for a wide range of

paleomagnetic and environmental magnetic studies. Therefore, robust and efficient methods

are needed to identify their presence within sediments. However, this is not straightforward

because inclusions are fine-grained (often in the nanometer size range). Also, sediment

samples often contain mixed magnetic mineral assemblages. The most robust way to identify

magnetic inclusions is by direct TEM observations, as has been demonstrated in this study.

But this requires time-consuming sample preparation and analysis, which makes analysis of

large sample sets impossible. We, therefore, explore whether magnetic screening of bulk

sediment samples can provide useful indications about the possible presence of magnetic

mineral inclusions.

Our detailed TEM and magnetic analyses reveal important properties of magnetic

mineral inclusions in sediments that provide clues about their presence. Magnetic mineral

inclusions are fine-grained, and often have SSD-like magnetic properties (Fig. 2). But the

grain size distributions of magnetic inclusions can overlap with those of other type of fine

© 2016 American Geophysical Union. All rights reserved.

particles, such as biogenic magnetite crystals, which can complicate their discrimination.

Some magnetic mineral crystals hosted in silicates occur in clusters or as complex dendrites

that produce some degrees of three-dimensional magnetostatic interactions to varying extents.

Such microstructures differ from those of intact biogenic magnetite chains in sediments,

which often produce a non-interacting uniaxial SSD signature [Egli et al., 2010]. Such

contrasting properties produce detectable rock magnetic signatures that enable discrimination

between these two important types of magnetic minerals in sediments. For example, FORC

diagrams with a SSD component and moderate vertical spread (Fig. 2a, e, g) are a useful

indication of the presence of magnetic mineral inclusions [Lappe et al., 2011; Muxworthy and

Evans, 2013] that contrast with the non-interacting central-ridge FORC signature observed

for biogenic magnetite [e.g., Egli et al., 2010; Roberts et al., 2012; Chang et al., 2014].

However, samples that contain dispersed magnetic nanoparticles in silicates can also give rise

to FORC signatures with weak magnetostatic interactions [e.g., Usui et al., 2015]. The size

distribution of magnetic inclusions is often broad, ranging from just a few nanometers to a

few microns (Fig. 3-8). Such size distributions can be detected magnetically. For example,

decomposition of isothermal remanent magnetization (IRM) acquisition curves produces a

component with large dispersion parameter (DP) values (i.e., >~0.3) for the studied sample

from core MD01-2421. This may also be reflected in the FORC coercivity profiles (Fig. 2j).

For example, the Bc profile for samples with magnetic inclusions is broad and extends up to

higher fields, i.e., 200 mT for sample “MD01-2421-7-110” from the North Pacific Ocean.

This may be because titanomagnetite is magnetically harder than pure magnetite. Low-

temperature magnetometry is also useful for detecting magnetic mineral inclusions in

sediments (Fig. 2). For example, it was demonstrated recently that biogenic and inorganic

magnetite in marine sediment samples have two distinct Tv temperatures clustered at ~100

and 120 K, respectively [Chang et al., 2016a]. Thus, combined magnetic analyses, such as a

© 2016 American Geophysical Union. All rights reserved.

Tv signature at 120 K together with a SSD FORC signature with weak or moderate

magnetostatic interactions, provides a practically useful way to discriminate magnetite

inclusions from biogenic magnetite within sediment samples (Fig. 2). Nevertheless, definite

rock magnetic identification of magnetic mineral inclusions is difficult to achieve because

sediment samples often contain mixed magnetic mineral assemblages and also because

similar coercivity distributions and magnetostatic interactions can be observed for both

lithogenic and biogenic magnetite (Fig. 2).

4.3. Implications for sedimentary magnetism and paleomagnetism

Identification of magnetic mineral inclusions within detrital particles has important

implications for understanding the magnetization of marine sediments and paleomagnetism.

First, our results demonstrate that silicate-hosted magnetic nanoparticles are an important

source of fine-grained SSD particles in marine sediments, in addition to biogenic magnetite

[e.g., Roberts et al., 2012]. SSD magnetic minerals are important for paleomagnetic studies

because they are ideal magnetic recorders that can carry stable remanences over long periods

of geological time [Dunlop and Özdemir, 1997]. Second, compared to unprotected magnetic

mineral particles, silicate-hosted magnetic inclusions have a much higher preservation

potential against sulfate-reducing diagenesis. This preservation potential will have a

potentially important influence on environmental magnetic records. For example, an

integrated study of marine sediment core MD01-2421 from the continental margin of Japan

demonstrates that monsoon-controled changes in non-steady state diagenetic conditions can

drive preferential dissolution of different populations of magnetic mineral grains. Such

diagenetic processes produced a periodically varying sedimentary magnetic pattern

throughout the core, where monsoon events gave rise to an enhanced environmental magnetic

signal from magnetic inclusions that would otherwise have been destroyed by reductive

© 2016 American Geophysical Union. All rights reserved.

diagenesis [Chang et al., 2016b]. The possible presence of magnetic nano-inclusions within

detrital particles is, therefore, likely to be important for interpreting diagenetically altered

marine sediment records in a wide range of settings. Third, igneous formation of silicate-

hosted magnetic nanoparticles is related to a range of factors, such as oxygen fugacity, cation

content, temperature, and pressure. The transportation pathway of erosional detritus from

igneous rocks into marine environments is also sensitive to environmental conditions.

Therefore, detection and characterization of silicate-hosted magnetic inclusions preserved in

marine sediments could also be useful for tracking geological provenance.

Characterization of silicate-hosted magnetic mineral inclusions is also potentially

important for paleomagnetic studies. Significant questions exist about their potential

paleomagnetic recording capability. For igneous rocks that contain such magnetic inclusions,

how do they acquire a TRM and is the anisotropy of elongated particles important for

interpreting paleomagnetic signals? How do magnetostatic interactions among such

nanoparticles affect paleomagnetic recording fidelity [e.g., Feinberg et al., 2006]? Our

characterization of microstructures of magnetic mineral inclusions and hosts indicate that

these are important questions to address when subjecting such materials to paleomagnetic

analysis. For example, three-dimensional micromagnetic models can be constructed to

simulate the magnetic properties of magnetic mineral inclusions with complex morphologies

and their paleomagnetic recording fidelity can be assessed quantitatively [e.g., Williams et al.,

2010; Muxworthy and Evans, 2013]. Moreover, single plagioclase crystals that contain

magnetic mineral inclusions have been used for absolute paleointensity determinations [e.g.,

Tarduno et al., 2001]. Some of our silicates differ from those documented in prior

paleointensity and paleomagnetic studies of single silicate crystals, particularly in the density

of inclusions; this may be partially due to selection criteria in those studies that excludes

crystals with visible inclusions (at low magnification) that are aimed at avoiding MD

© 2016 American Geophysical Union. All rights reserved.

magnetic carriers [e.g., Tarduno et al., 2006]. However, the presence of silicates with high

inclusion density in our sediments highlights the continued need to test for the possibility of

interactions by nanoscale imaging [e.g., Feinberg et al., 2006; Bono and Tarduno, 2015],

FORC analyses [e.g., Tarduno and Cottrell, 2005], and the application of paleointensity

selection criteria (the latter can suggest the presence of interactions if natural remanent

magnetization (NRM)/TRM plots are non-linear).

An important question to consider in relation to sedimentary paleomagnetic recording

is how efficiently do magnetic mineral inclusions in silicates produce a DRM in sediments?

The large size of some host silicate particles (ranging from microns to hundreds of microns)

means that hydrodynamic forces will be important during deposition and that large particles

are unlikely to be aligned by a geomagnetic torque (Fig. 9). It is, therefore, to be expected

that such large particles will contribute to randomization of sedimentary paleomagnetic

signals. But how do such particles compare with the particle size distributions of sediments

that are subjected to paleomagnetic investigations? Sandstones are rarely used for

paleomagnetic analysis because even if magnetic particles occur in the finest possible sand

category (very fine sand), they will have sizes of at least 50 µm. Such magnetic particles will

have MD properties that will not enable recording of a stable paleomagnetic signal. Likewise,

sand-sized host particles with magnetic nano-inclusions will be dominated by hydrodynamic

rather than by magnetic torques and will not record a stable paleomagnetic signal (Fig. 9). In

contrast, clay-rich sediments (< 2 µm) are often considered ideal for paleomagnetic analysis

because fine particles are more likely to give rise to stable paleomagnetic recording.

However, clay minerals are products of weathering rather than being primary detrital

minerals that have been abraded to ultra-fine sizes, so that much of the clay size fraction in a

sediment will be due to clay minerals. Nevertheless, some part of the clay size fraction of

sediments could represent particles that have been abraded to ultra-fine sizes.

© 2016 American Geophysical Union. All rights reserved.

 Windblown sediment particles are likely to have greater roughness than other types of

sediments that have been rounded extensively through abrasion in fluvial and other aquatic

systems, and provide a worthwhile end-member for considering particle aspect ratios and the

effects of hydrodynamic versus magnetic torques. Okada et al. [2001] analyzed particle

shapes for atmospherically transported mineral particles from three Chinese arid regions.

These fine silt- to clay-sized detrital mineral particles (0.1 to 6 µm) have irregular shapes as

expected, with aspect-ratios that are size independent and that range from values of 1 to ~3

(shaded region in Fig. 9), with skewed distributions and median aspect-ratios of 1.3 to 1.4.

Virtually 100% of their analyzed mineral particles have aspect ratios <5. The results of Okada

et al. [2001] place useful constraints on the region of Figure 9 that is likely to be meaningful

for paleomagnetic recording of host particles with magnetic nano-inclusions. From the results

of our simple models, it appears that host silicate particles that contain magnetite nano-

inclusions should be capable of contributing to sedimentary paleomagnetic records at

equivalent particle diameters below 12 µm for 1% magnetite concentrations (Fig. 9). The

effective diameter of particles that can be aligned by geomagnetic torques will increase for

larger magnetite nano-inclusion concentrations.

Following the above arguments, silicate-hosted magnetic inclusions could be

important for paleomagnetic recording in silt- and clay-sized sediments. It is unlikely that

magnetite-rich host silicates will contribute to the magnetization of medium silts with particle

sizes above ~18-20 µm (Fig. 9). Sediments always contain a distribution of particle sizes.

Size distributions that cross the τH = τM line, which from Fig. 9 is likely to occur in silt- and

clay-sized sediments, will have some capacity for reliable paleomagnetic recording with

considerable partial cancelation due to both particle types. The resulting magnetization will

not be efficient, which is consistent with the low efficiency of sedimentary magnetizations

[e.g., Tauxe et al., 2006; Mitra and Tauxe, 2009; Heslop et al., 2014]. Regardless, the simple

© 2016 American Geophysical Union. All rights reserved.

model results presented in Fig. 9 should be considered a best-case scenario for paleomagnetic

recording. Some silicate inclusions will have two or more preferred crystallographic

orientations of inclusions [e.g., Feinberg et al., 2006] for which variable extents of magnetic

moment cancelation would be expected.

5. Conclusions

TEM observations of magnetic mineral extracts from geographically widely

distributed samples confirm the abundant occurrence of magnetic nanoparticle inclusions

hosted within silicate crystals in marine sediments. We document variable inclusion

morphologies, including isolated nanoparticles (i.e., octahedra, sub-rounded, and irregular

shapes), nanoparticle clusters, and dendrites. EDS analysis indicates that the magnetic

mineral inclusions consist of magnetite to titanomagnetite (with low but variable Ti contents),

while the hosts are silicate minerals (mostly plagioclase feldspar and clinopyroxene). Some

magnetic nanoparticles occur with crystallographically preferred orientations within the host

silicates. The inclusion density in some of the silicates isolated here differs from those

documented in rock magnetic and paleomagnetic studies [e.g., Feinberg et al., 2005; Bono

and Tarduno, 2015]; while such particles may have been excluded in prior paleomagnetic and

paleointensity studies by the selection criteria used [e.g., Tarduno et al., 2006], they may be

magnetically important in sedimentary paleomagnetism. Silicate minerals are relatively stable

against diagenetic alteration in sulfate-reducing diagenetic environments, which can,

therefore, protect the embedded mineral inclusions from dissolution. Our results demonstrate

that silicate-hosted magnetic mineral inclusions are an important source of fine-grained

magnetic minerals in sediments, and provide important constraints on understanding

sedimentary paleomagnetic and environmental magnetic records. The DRM of magnetic

inclusions was modeled to assess their paleomagnetic recording capability. Calculations

© 2016 American Geophysical Union. All rights reserved.

indicate that deposition of large silicate particles will be controlled by hydrodynamic forces

rather than by geomagnetic torques, so that even if large particles may contain ideal SSD

inclusions, they are unlikely to contribute meaningfully to paleomagnetic recording.

Nevertheless, deposition of smaller silicate particles with magnetic mineral inclusions could

give rise to a reliable paleomagnetic record.

Acknowledgement We are grateful to Felipe Kremer and Frank Brink at the Centre for

Advanced Microscopy, ANU, for helping with TEM analysis, and Penelope King and Andrew

Berry for useful discussions. Adrian Muxworthy and Yoichi Usui are thanked for providing

published FORC data of silicate crystals extracted from igneous rocks. We thank Richard

Harrison and John Tarduno for helpful review comments, and André Revil and an Associate

Editor for efficient editorial handling. The data in this paper can be obtained by contacting

the corresponding author (liao.chang@pku.edu.cn). The data can be found at the RMAG

portal (rock magnetic database) of the Magnetics Information Consortium

(http://earthref.org/MAGIC/). This study was supported by the “1000 Talents Plan” program

of China, the National Natural Science Foundation of China (grant 41574060), and the

Australian Research Council (grants DP120103952, DP140104544, and LE0882854).

References

Bono, R. K., and J. A. Tarduno (2015), A stable Ediacaran Earth recorded by single silicate

r sta s th a 565 Ma t- les intrusion, Geology, 43, 131–134,

doi:10.1130/G36247.1.

Canfield, D. E., and R. A. Berner (1987), Dissolution and pyritization of magnetite in anoxic

marine sediments, Geochim. Cosmochim. Acta, 51, 645–659, doi:10.1016/0016-

7037(87)90076-7.

© 2016 American Geophysical Union. All rights reserved.

Canfield, D. E. and R. Raiswell (1991), Pyrite formation and fossil preservation, in

Taphonomy: Releasing the Data Locked in the Fossil Record, Allison P. A., and D. E.

G. Briggs (Eds.), 337–387, Plenum Press, New York.

Chang, L., A. P. Roberts, W. Williams, J. D. Fitz Gerald, J. C. Larrasoaña, L. Jovane, and A.

R. Muxworthy (2012), Giant magnetofossils and hyperthermal events, Earth Planet.

Sci. Lett., 351–352, 258–269, doi: 10.1016/j.epsl.2012.07.031.

Chang, L., A. P. Roberts, M. Winklhofer, D. Heslop, M. J. Dekkers, W. Krijgsman, J. D. Fitz

Gerald, and P. Smith (2014), Magnetic detection and characterization of biogenic

magnetic minerals: A comparison of ferromagnetic resonance and first-order reversal

curve diagrams, J. Geophys. Res., 119, 6136–6158, doi:10.1002/2014JB011213.

Chang, L., D. Heslop, A. P. Roberts, D. Rey, and K. J. Mohamed (2016a), Discrimination of

biogenic and detrital magnetite through a double Verwey transition temperature, J.

Geophys. Res., 121, 3–14, doi: 10.1002/2015JB012485.

Chang, L., C. T. Bolton, M. J. Dekkers, A. Hayashida, D. Heslop, W. Krijgsman, K. Kodama,

G. A. Paterson, A. P. Roberts, E. J. Rohling, Y. Yamamoto, and X. Zhao (2016b),

Asian monsoon modulation of nonsteady state diagenesis in hemipelagic marine

sediments offshore of Japan, Geochem. Geophys. Geosyst., 17, doi:10.1002/

2016GC006344.

Cottrell, R. D., and J. A. Tarduno (1999), Geomagnetic paleointensity derived from single

plagioclase crystals, Earth Planet. Sci. Lett., 169, 1–5, doi:10.1016/S0012-

821X(99)00068-0.

Dunlop, D. J. (1990), Developments in rock magnetism, Repts. Progr. Phys., 53, 707–792.

Dunlop, D. J., and Ö. Özdemir (1997), Rock Magnetism: Fundamentals and Frontiers,

Cambridge Uni. Press, 573 pp.

Egli, R. (2013), VARIFORC: An optimized protocol for calculating non-regular first-order

© 2016 American Geophysical Union. All rights reserved.

 reversal curve (FORC) diagrams, Global Planet. Change, 110, 302–320,

doi:10.1016/j.gloplacha.2013.08.003.

Egli, R., A. P. Chen, M. Winklhofer, K. P. Kodama, and C.-S. Horng (2010), Detection of

noninteracting single domain particles using first-order reversal curve diagrams,

Geochem. Geophys. Geosyst., 11, Q01Z11, doi:10.1029/2009GC002916.

Evans, M. E., and M. L. Wayman (1970), Investigation of small magnetic particles by means

of electron microscopy, Earth Planet. Sci. Lett., 9, 365–370.

Evans, M. E., M. W. McElhinny, and A. C. Gifford (1968), Single domain magnetite and high

coercivities in a gabbroic intrusion, Earth Planet. Sci. Lett., 4, 142–146.

Feinberg, J. M., G. R. Scott, P. R. Renne, and H. R. Wenk (2005), Exsolved magnetite

inclusions in silicates: Features determining their remanence behavior, Geology, 33,

513–516.

Feinberg, J. M., R. J. Harrison, T. Kasama, R. E. Dunin-Borkowski, G. R. Scott, and P. R.

Renne (2006), Effects of internal mineral structures on the magnetic remanence of

silicate-hosted titanomagnetite inclusions: An electron holography study, J. Geophys.

Res., 111, B12S15, doi:10.1029/ 2006JB004498.

Haggerty, S. E. (1991), Oxide mineralogy of the upper mantle, edited by D. H. Lindsley,

Oxide Minerals: Petrologic and Magnetic Significance, Reviews in Mineralogy, 25,

355–416.

Harrison, R. J., R. E. Dunin-Borkowski, and A. Putnis (2002), Direct imaging of nanoscale

magnetic interactions in minerals, Proc. Natl. Acad. Sci. U.S.A., 99, 16,556–16,561.

Heslop, D. (2007), Are hydrodynamic shape effects important when modelling the formation

of depositional remanent magnetization? Geophys. J. Int., 171, 1029–1035,

doi:10.1111/j.1365-246X.2007.03588.x.

Heslop, D., and A. P. Roberts (2012), Estimation of significance levels and confidence

© 2016 American Geophysical Union. All rights reserved.

 intervals for first-order reversal curve distributions, Geochem. Geophys. Geosyst., 13,

Q12Z40, doi:10.1029/2012GC004115.

Heslop, D., A. P. Roberts, and R. Hawkins (2014), A statistical simulation of magnetic

particle alignment in sediments, Geophys. J. Int., 197, 828–837.

Karlin, R., and S. Levi (1983), Diagenesis of magnetic minerals in recent haemipelagic

sediments, Nature, 303, 327–330.

Lappe, S.-C. L. L., N. Church, T. Kasama, A. Fanta, G. Bromiley, R. E. Dunin-Borkowski, J.

M. Feinberg, S. Russell, and R. J. Harrison (2011), Mineral magnetism of dusty olivine:

A credible recorder of pre-accretionary remanence, Geochem. Geophys. Geosyst., 12,

Q12Z35, doi:10.1029/2011GC003811.

Lappe, S.-C. L. L., J. M. Feinberg, A. Muxworthy, and R. J. Harrison (2013), Comparison

and calibration of nonheating paleointensity methods: A case study using dusty olivine,

Geochem. Geophys. Geosyst., 14, 2143–2158, doi:10.1002/ggge.20141.

Mitra, R., and L. Tauxe (2009), Full vector model for magnetization in sediments, Earth

Planet. Sci. Lett., 286, 535–545.

Muxworthy, A. R., and M. E. Evans (2013), Micromagnetics and magnetomineralogy of

ultrafine magnetite inclusions in the Modipe Gabbro, Geochem. Geophys. Geosyst., 14,

921–928, doi:10.1029/2012GC004445.

Muxworthy, A. R., M. E. Evans, S. J. Scourfield, and J. G. King (2013), Paleointensity results

from the late-Archaean Modipe Gabbro of Botswana, Geochem. Geophys. Geosyst., 14,

2198–2205, doi:10.1002/ggge.20142.

Oba, T., T. Irino, M. Yamamoto, M. Murayama, A. Takamura, and K. Aoki (2006),

Paleoceanographic change off central Japan since the last 144,000 years based on high-

resolution oxygen and carbon isotope records, Global Planet. Change, 53, 5–20.

Okada, K., J. Heintzenberg, K. Kai, and Y. Qin (2001), Shape of atmospheric mineral

© 2016 American Geophysical Union. All rights reserved.

 particles collected in three Chinese arid-regions, Geophys. Res. Lett., 28, 3123–3126.

Poulton, S. W., M. D. Krom, and R. Raiswell (2004), A revised scheme for the reactivity of

iron (oxyhydr)oxide minerals towards dissolved sulfide, Geochim. Cosmochim. Acta,

68, 3703–3715.

Roberts, A. P. (2015), Magnetic mineral diagenesis, Earth-Sci. Rev., 151, 1–47.

Roberts, A. P., and G. M. Turner (1993), Diagenetic formation of ferrimagnetic iron sulphide

minerals in rapidly deposited marine sediments, South Island, New Zealand, Earth

Planet. Sci. Lett., 115, 257–273.

Roberts, A. P., C. R. Pike, and K. L. Verosub (2000), First-order reversal curve diagrams: A

new tool for characterizing the magnetic properties of natural samples, J. Geophys.

Res., 105, 28,461–28,475, doi:10.1029/2000JB900326.

Roberts, A. P., L. Chang, D. Heslop, F. Florindo, and J. C. Larrasoaña (2012), Searching for

single domain magnetite in the “pseudo-single-domain” sedimentary haystack:

Implications of biogenic magnetite preservation for sediment magnetism and relative

paleointensity determinations, J. Geophys. Res., 117, B08104,

doi:10.1029/2012JB009412.

Roberts, A. P., F. Florindo, L. Chang, D. Heslop, L. Jovane, and J.C. Larrasoaña (2013),

Magnetic properties of pelagic marine carbonates, Earth-Sci. Rev., 127, 111–139.

Roberts, A. P., D. Heslop, X. Zhao, and C. R. Pike (2014), Understanding fine magnetic

particle systems through use of first-order reversal curve diagrams, Rev. Geophys., 52,

557–602, doi:10.1002/2014RG000462.

Rowan, C. J., and A. P. Roberts (2006), Magnetite dissolution, diachronous greigite

formation, and magnetizations arising from pyrite oxidation: Unravelling complex

magnetizations in Neogene marine sediments from New Zealand, Earth Planet. Sci.

Lett., 241, 119–137.

© 2016 American Geophysical Union. All rights reserved.

Rowan, C. J., A. P. Roberts, and T. Broadbent (2009), Reductive diagenesis, magnetite

dissolution, greigite growth and paleomagnetic smoothing in marine sediments: A new

view, Earth Planet. Sci. Lett., 277, 223–235.

Sato, M., S. Yamamoto, Y. Yamamoto, Y. Okada, M. Ohno, H. Tsunakawa, and S. Maruyama

(2015), Rock-magnetic properties of single zircon crystals sampled from the Tanzawa

tonalitic pluton, central Japan, Earth Planets Space, 67, 150, doi:10.1186/s40623-015-

0317-9.

Tarduno, J. A., and R. D. Cottrell (2005), Dipole strength and variation of the time-averaged

reversing and nonreversing geodynamo based on Thellier analyses of single plagioclase

crystals, J. Geophys. Res., 110, B11101, doi:10.1029/ 2005JB003970.

Tarduno, J. A., R. D. Cottrell, and A. V. Smirnov (2001), High geomagnetic field intensity

during the mid-Cretaceous from Thellier analyses of single plagioclase crystals,

Science, 291, 1779–1783, doi:10.1126/science.1057519.

Tarduno, J. A., R. D. Cottrell, and A. V. Smirnov (2006), The paleomagnetism of single

silicate crystals: Recording geomagnetic field strength during mixed polarity intervals,

superchrons, and inner core growth, Rev. Geophys., 44, RG1002,

doi:10.1029/2005RG000189.

Tarduno, J. A., R. D. Cottrell, M. Watkeys, A. Hofmann, P. V. Doubrovine, E. E. Mamajek, D.

Liu, D. G.Sibeck, L. P. Neukirch, and Y. Usui (2010), Geodynamo, solar wind, and

magnetopause 3.4 to 3.45 billion years ago, Science, 327, 1238–1240.

Tarduno, J. A., R. D. Cottrell, W. J. Davis, F. Nimmo, and R. K. Bono (2015), A Hadean to

Paleoarchean geodynamo recorded by single zircon crystals, Science, 349, 521–524.

Tauxe, L., J. L. Steindorf, and A. Harris (2006), Depositional remanent magnetization:

Toward an improved theoretical and experimental foundation, Earth Planet. Sci. Lett.,

244, 515–529.

© 2016 American Geophysical Union. All rights reserved.

Ueshima, T., M. Yamamoto, T. Irino, T. Oba, M. Minagawa, H. Narita, and M. Murayama

(2006), Paleo-productivity record from core MD01-2421, the mid-latitude western

North Pacific, during the last 145,000 years, Global Planet. Change, 53, 21–28.

Usui, Y., T. Shibuya, Y. Sawaki, and T. Komiya (2015), Rock magnetism of tiny exsolved

magnetite in plagioclase from a Paleoarchean granitoid in the Pilbara craton, Geochem.

Geophys. Geosyst., 16, 112–125, doi:10.1002/2014GC005508.

Wakabayashi, K., H. Tsunakawa, N. Mochizuki, Y. Yamamoto, and Y. Takigami (2006),

Paleomagnetism of the middle Cretaceous Iritono granite in the Abukuma region,

northeast Japan, Tectonophysics, 421, 161–171.

Williams, W., M. E. Evans, and D. Krása (2010), Micromagnetics of paleomagnetically

significant mineral grains with complex morphology, Geochem. Geophys. Geosyst., 11,

Q02Z14, doi:10.1029/2009GC002828.

Zhao, X., D. Heslop, and A. P. Roberts (2015), A protocol for variable resolution first-order

reversal curve measurements, Geochem. Geophys. Geosyst., 16, 1364–1377,

doi:10.1002/2014GC005680.

© 2016 American Geophysical Union. All rights reserved.

Figure captions

Figure 1 Locations of marine sediment core MD01-2421 from the north Pacific off the

east coast of Japan, core CD143-055705 on the continental margin of Oman, core

RR0603-03JC from the eastern equatorial Pacific Ocean, and marine sediment outcrop

from the Lower Awatere Valley, northeastern South Island, New Zealand.

© 2016 American Geophysical Union. All rights reserved.

© 2016 American Geophysical Union. All rights reserved.
Figure 2 FORC diagrams (a, c, e, g, i) and low-temperature SIRM warming curves (b,

d, f, h) for: (a, b) sample “MD01-2421-7-110” from the North Pacific Ocean, (c, d)

sample “RR0603-03JC-2-60” from the eastern equatorial Pacific Ocean, (e, f) sample

“CD143-55705-7-82” from the Oman margin, Arabian Sea, and (g, h) samples

“BR49D” and (i) “BL37,38,39” from the Lower Awatere Valley, New Zealand. (j)

Coercivity distributions (horizontal profiles at Bu = 0) extracted from FORC diagrams.

Published data are shown in (j) for several marine sediment samples with biogenic

magnetite as the dominant magnetic mineral (samples “ODP-738B-4H-6-130”, “ODP-

738C-11R-1-28”, “ODP-689D-8H4-71”, and “ODP-690C-9H6-76”, Figure 10 in

Roberts et al. [2013]), and igneous rocks containing magnetic inclusions (a handpicked

sample “B4HP” containing pure pyroxene crystals (Figure 3d; Muxworthy and Evans

[2013]) and a handpicked sample containing six plagioclase crystals (Figure 4b; Usui et

al. [2015]). “BL37,38,39” is a magnetic separate from Roberts and Turner [1993],

while all other studied samples are bulk marine sediments. Note that the studied

samples also contain other magnetic assemblages in addition to magnetic inclusions

(see text for discussion). FORC diagrams in (c, e, g) were measured with variable field

steps following the protocol of Zhao et al. [2015]. All FORC diagrams were processed

using the algorithm of Zhao et al. [2015]. Thicker black lines correspond to the 0.05

significance level [Heslop and Roberts, 2012]. Dashed black lines in the FORC

diagrams correspond to the profile of Bu = 0.

© 2016 American Geophysical Union. All rights reserved.

Figure 3 (a-d) STEM and (e-i) bright-field TEM images of microstructures of magnetic
nanoparticle inclusions within silicate crystals for sample “MD01-2421-7-110” from
the North Pacific Ocean, offshore of central Japan. In the STEM images (a-d), bright
particles are mineral inclusions. In the bright-field TEM images (e-i), the mineral
inclusions appear dark. Observed morphologies of magnetic mineral inclusions include
(a-c, h) nanoparticle clusters, (d-f) dendrites, (g) crystallographically oriented
nanoparticles, and (i) a large titanomagnetite inclusion.

© 2016 American Geophysical Union. All rights reserved.

Figure 4 Bright-field TEM images and associated SAED patterns for magnetic mineral

inclusions for sample “MD01-2421-7-110”. The microstructures and electron

diffraction patterns (a-g) indicate that some nanoparticle inclusions have a prefered

crystallographic orientation (double headed arrows) within silicate host crystals, while

(h, i) other nanoparticles appear to be more randomly oriented (see text for discussion).

© 2016 American Geophysical Union. All rights reserved.

Figure 5 Bright-field TEM images at progressively higher magnifications (from left to
right for each row of images) that reveal lattice fringes of magnetic mineral inclusions
for sample “MD01-2421-7-110”. In (b), the insert image (1) is a bright-field TEM
image of the diffracting area (when a diffraction aperture was in the electron beam).
The insert image (2) is the corresponding SAED pattern of the area circled in (1). The
cross sign in (b) indicates the location of the EDS spot. All analyzed nanoparticles have
clear lattice fringes. The host minerals also have clear lattice fringes (p). Values of
lattice spacings and the corresponding Miller indices (hkl) are indicated along the
lattice fringes. All measured lattice spacings of mineral inclusions are consistent with
those of titanomagnetite or magnetite.

© 2016 American Geophysical Union. All rights reserved.

Figure 6 High-resolution TEM and SAED analyses of magnetic mineral inclusions and

host minerals for two areas (a-f, and g-l) for sample “MD01-2421-7-110”. For one area

© 2016 American Geophysical Union. All rights reserved.

(a-f), images in (b-d) and (e, f) correspond to magnetic mineral inclusions and host

minerals, respectively, for areas indicated in (a). The SAED pattern in (b) is from the

whole area in (a). For another area (g-l), images in (h) and (i-l) correspond to magnetic

mineral inclusions and host minerals, respectively, for areas indicated in (g). Clear

lattice fringes for the inclusions and host minerals are observed. Arrows in (g) indicate

crystal twinning of magnetic nanoparticle inclusions. The mineral inclusions and

silicate host minerals are identified to be titanomagnetite and plagioclase, respectively.

© 2016 American Geophysical Union. All rights reserved.

Figure 7 EDS analyses of magnetic mineral inclusions and host minerals. (a, g) STEM

images and (b-f, h-l) corresponding elemental maps of two areas within silicate crystals

with magnetic nanoparticle inclusions, and (m-u) EDS spectra for magnetic

nanoparticle inclusions and their host minerals for three analyzed areas in sample

“MD01-2421-7-110”. The nanoparticle inclusions are rich in Fe, but only have small Ti

concentrations. The Ti map in (i) is not as clear as that in (c), which appears to be due

to a low Ti content. The EDS spectra of the host mineral contain mainly Si and O, with

smaller concentrations of Al and Ca. The host mineral grains (a, g) are rich in O, and Si

© 2016 American Geophysical Union. All rights reserved.

and contain small concentrations of Ca and Al. The host silicate minerals here are likely

to be plagioclase. The nanoparticle inclusions (b, c, e, f, h, i) are rich in Fe and O, and

also contain a small Ti peak. The (*) symbol indicates Cu peaks, which originate from

the TEM grid and are present in all spectra. Fe/Ti ratios for titanomagnetite inclusions

are indicated for relevant EDS spectra.

© 2016 American Geophysical Union. All rights reserved.

Figure 8 (a) Bright-field TEM image of acicular magnetite inclusions for sample

“CD143-55705-7-82” from marine sediment core CD143-55705 from the Oman

margin, Arabian Sea. The host mineral is rich in O and Si. The exsolved acicular

inclusions appear to have a preferential alignment along their elongation direction

(double headed arrow). (b) A bright-field TEM image of biogenic magnetite crystals

and (c) a STEM image of titanomagnetite nanoparticle inclusions hosted in silicates for

marine sediment sample “RR0603-03JC-2-60” from eastern equatorial Pacific Ocean

core RR0603-03JC. (d) A bright-field TEM and (e, f) STEM images of magnetic

mineral inclusions hosted in silicates from magnetic separate sample “BL37,38,39”

from the Lower Awatere Valley, New Zealand [Roberts and Turner, 1993]. Small black

holes in (f) are ablation pits left after EDS point analyses.

© 2016 American Geophysical Union. All rights reserved.

Figure 9 Model results for a DRM carried by magnetic mineral inclusions within

silicates. Estimated aspect ratios are shown at which τH = τM for host particles that

contain different volume percentages of magnetite nanoparticles (colored lines). To

achieve τH = τM, smaller particles would require unrealistically high aspect-ratios,

which indicates that particle orientations are expected to be dominated by magnetic

torques. In contrast, larger particles require aspect ratios close to 1, which indicates that

the orientation of such particles is expected to be dominated by hydrodynamic torques

(see methods for details of the numerical simulations). The dashed lines denote

boundaries between very fine, fine, and medium silt. Schematic illustration of prolate

ellipsoids alongside the calculated curves, which represent modeled silicate particles,

© 2016 American Geophysical Union. All rights reserved.

which contain embedded magnetite inclusions (black spheres). The light gray area

highlights the aspect-ratio range of 1-3 in which most detrital particles are expected to

fall [Okada et al., 2001]. The dark gray area indicates silicates with a volumetric

magnetite content of 10%. According to the numerical model, particles in this region

should not able to acquire a significant DRM. Arrows indicate the trends of the

respective τH = τM lines where the magnetic and hydrodynamic forces balance.

© 2016 American Geophysical Union. All rights reserved.