Earth and Planetary Science Letters 305 (2011) 171–184

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

j ourna l homepage: www.e lsev ie r.com/ locate /eps l

29

Self-reversed magnetization held by martite in basalt flows from the1.1-billion-year-old Keweenawan rift, Canada

Nicholas L. Swanson-Hysell a,⁎, Joshua M. Feinberg b, Thelma S. Berquó b, Adam C. Maloof a

a Department of Geosciences, Princeton University, Guyot Hall, Washington Road, Princeton, NJ 08544, USAb Institute for Rock Magnetism, Department of Geology and Geophysics, University of Minnesota, Minneapolis, MN 55455, USA

⁎ Corresponding author. Tel.: +1 609 258 0836; fax:E-mail address: nswanson@princeton.edu (N.L. Swa

0012-821X/$ – see front matter © 2011 Elsevier B.V. Adoi:10.1016/j.epsl.2011.02.053

a b s t r a c t

a r t i c l e i n f o

Article history:Received 9 November 2010Received in revised form 28 February 2011Accepted 28 February 2011Available online 29 March 2011

Editor: P. DeMenocal

Keywords:self-reversalantiparallel remanencemartitehematiterock magnetismpaleomagnetismexchange couplingKeweenawan Rift

In some basalt flows of the 1.1-billion-year-old Keweenawan Rift exposed atMamainse Point, Ontario, thereis a magnetic phase that holds a remanence antiparallel to the populations of magnetite and hematite thatare typical of flows in the succession. The paleomagnetic and geological context of this componentdemonstrates that it is not a chemical overprint whose remanence is attributable to a subsequently reversedgeomagnetic field, but that the component is a self-reversal of the primary magnetization. Here we use rockmagnetic experiments and Mössbauer spectroscopy to show that this phase occurs in the most oxidizedflows and is held by a fine-grained population of hematite that acquired its self-reversed remanence throughinteractions with a phase of lower blocking temperature. We propose that this self-reversed hematite is amanifestation of the self-reversal phenomena that has been observed to occur experimentally during thetransformation ofmaghemite to hematite and that has been attributed to negative exchange coupling acrossthe crystal lattices. We suggest that hematite formed in association with iron-silicates in the basalt flowscarries a remanence reflective of the field in which it formed, but that martite (hematite pseudomorphedafter magnetite) which formed through the progressive oxidation of magnetite to maghemite withsubsequent inversion to hematite carries the self-reversed remanence. This study marks the second timethat a naturally occurring self-reversedmagnetization has been attributed to this mechanism andmarks theoldest reported instance of any type of self-reversed remanence. Themartite that forms through this processhas the ability to retain a record of the negative exchange coupling from the time of its formation forhundreds of millions of years.

+1 609 258 1274.nson-Hysell).

ll rights reserved.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Rapidly cooled volcanic rocks contain abundant small titanomag-netite grains that serve as excellent magnetic recorders. Given theirhigh unblocking temperatures, strong magnetization and paleomag-netic stability over billion-year time-scales, the titanomagnetite inbasalt flows serves as an important repository of geomagneticinformation. Much of what we know about the past latitude ofcontinents and the past intensity of the geomagnetic field comes frompaleomagnetic studies of extrusive basalt flows. Fully understandingthe magnetization of such flows is of paramount importance so thatworkers can be confident in interpretations of past geomagnetic andplate tectonic behavior.

The oxidation state of high-temperature magmatic environmentscontrasts greatly from those at Earth's surface. When FeTi-oxidegrains are erupted onto the surface, they are out of equilibrium with

the newly established redox conditions due to falling temperature andincreasing oxygen fugacity. This disequilibrium can result in deutericoxidation as the lava flow cools. A commonly observed effect of theincrease in the ratio of Fe2+ to Fe3+, that accompanies deutericoxidation, is the exsolution of ilmenite lamellae within titanomagne-tite grains (Aragon et al., 1984; Buddington and Lindsley, 1964). Thisprocess subdivides the titanomagnetite into many smaller grains thatare Ti-poor and in the single to pseudo-single domain size range(Evans and Wayman, 1974). These grains become excellent recordersof magnetic information as they cool through their blockingtemperatures.

During and after initial cooling, the magnetic mineralogy of basaltsremains out of equilibriumwith the oxic surface environment and canbe further altered by low-temperature oxidation. Low-temperatureoxidation commonly alters cubic magnetite grains to cubic maghe-mite (Fe3O4→γFe2O3). Remanent magnetization in ferromagneticminerals arises from the interaction of unfilled 3d orbitals betweeniron cations known as exchange coupling. In iron oxides, exchangecoupling occurs due to electron exchange between the iron cationsand the oxygen anions that link them (this interaction is referred to assuperexchange coupling). During the oxidation of magnetite to

172 N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

maghemite, maghemite retains the inverse spinel lattice of theprimary magnetite (Dunlop and Özdemir, 1997). During maghemi-tization, superexchange coupling overwhelms the influence of theexternal field and maghemite preserves the remanent magnetizationdirection of the original magnetite (Johnson and Merrill, 1974;Özdemir and Dunlop, 1985). Maghemite is a metastable mineral andcan invert to the stable polymorph of rhombohedral hematite(αFe2O3). The thermal activation necessary for this inversion can beachieved on laboratory time-scales at temperatures as low as 250 °C(Bernal et al., 1957) and can occur at lower temperatures ongeological time scales (Hargraves and Banerjee, 1973). In additionto magnetite oxidation, low-temperature oxidation can result inhematite formation as an alteration product of iron-silicate minerals.Hematite formation also occurs during high-temperature deutericoxidation of olivine during initial magma cooling (Haggerty andBaker, 1967). If hematite formation occurs during and/or soon afterthe emplacement of a basalt flow, the hematite would be expected torecord a remanence direction similar to that held by the primarytitanomagnetite grains.

The direction of remanent magnetization held by these Fe-oxidephases can be determined through thermal demagnetization of basaltsamples. Typically, magnetite and hematite have a distinct range ofunblocking temperatures given that the Curie temperature ofmagnetite (580 °C) is significantly lower than the Néel temperatureof hematite (675 °C) and that the thermal demagnetization spectra ofhematite is often quite “blocky,” such that most of the hematiteremanence is lost close to the Néel temperature (Dunlop andÖzdemir, 1997). If there is a notable decrease in the derivative ofmagnetic moment versus time at the magnetite Curie temperature,the direction of magnetization held by magnetite and hematite can beevaluated independently (e.g. Fig. 1A,E).

The magnetizations of extrusive basalt flows from the 1.1 GaKeweenawan mid-continent rift are generally quite simple, withremanence held by unidirectional components of magnetite andhematite of very similar direction (e.g. Fig. 1A,E; Swanson-Hysellet al., 2009). However, an additional component of magnetizationwas discovered in 14 of the 72 studied flows that has anintermediate unblocking temperature (580–650 °C) and holds aremanence that is antiparallel to that of the components thatunblock between 500 and 575 °C and between 650 and 675 °C(Fig. 1). Unlike these components, which can be confidentlyinterpreted as being carried by magnetite and hematite, respec-tively, on the basis of thermal demagnetization data, the magneticphase that carries the 590–650 °C component requires additionalrock magnetic experiments to determine its carrier. We will referto this phase throughout the text as the “antiparallel component.”The presence of the antiparallel component results in a “zig-zag” invector component plots of thermal demagnetization data (Fig. 1I).The stratigraphic context of the flows and the progression inpaleolatitude due to fast plate motion during their emplacement(Fig. 2) indicate that this “antiparallel” component is a self-reversal. In this contribution, we utilize paleo- and rock magneticdata paired with optical and electron microscopy to determine thecarrier of this self-reversed component and evaluate its origin.

1.1. Geological and paleomagnetic background

Keweenawan rocks of the 1.1 Ga mid-continent rift were the firstunits of Precambrian age to be studied paleomagnetically in NorthAmerica (Dubois, 1955; Graham, 1953). The development ofpaleomagnetic records from the rift has led to the best resolvedapparent polar wander path (APWP) in the Precambrian (the “LoganLoop”, Dubois, 1962; Halls and Pesonen, 1982). This APWP forms thebasis for much of our understanding of late Mesoproterozoicpaleogeography during the assembly of the supercontinent Rodinia(e.g. Buchan et al., 2001; Evans, 2009; Weil et al., 1998). The rock

magnetic behavior of Keweenawan lavas and specifically those fromMamainse Point were important in the development of the theoreticalblocking curves for magnetite and hematite (Pullaiah et al., 1975).There have been efforts to obtain records of the intensity of Earth'sgeomagnetic field from Keweenawan volcanics and intrusives(Pesonen and Halls, 1983), and to use the distribution of paleomag-netic directions from individual lava flows to characterize paleose-cular variation (Tauxe and Kodama, 2009).

A long standing enigma of paleomagnetic records obtained fromthe Keweenawan rift was the fact that compiled reversed and normalpaleomagnetic directions are not antiparallel to one another, with theolder reversed directions being steeper in their absolute inclination.This feature of the database has alternatively been interpreted asreversal asymmetry resulting from significant non-dipole contribu-tions to the surface geomagnetic field (Pesonen and Nevanlinna,1981), or as resulting from a change in paleolatitude associated withplate motions that occurred while volcanism was ongoing (Davis andGreen, 1997). In an effort to solve this problem, a recent studyreported a high-resolution paleomagnetic record from the mostcomplete section of extrusive lava flows in the rift at Mamainse Point,Ontario (Swanson-Hysell et al., 2009). These data spanned threereversals at high resolution, revealed a progressive decrease ininclination upwards through the stratigraphy and showed eachreversal to be symmetric. These results demonstrate that the periodof Keweenawan volcanism was a time of significant motion of theNorth American continent and that there is no need to invokesignificant non-dipole components to Earth's geomagnetic field toexplain the data.

The magnetizations carried by magnetite and hematite areconsistently of similar direction such that in 195 specimens whereleast squares fits could confidently bemade to distinct magnetite andhematite components, the average of the absolute value of thedifference between the inclinations is 3.6 (one standard deviation of2.8°). Given the 152° variation in inclination through the section, thissimilarity within individual specimens is striking. Directions fromthemagnetitefits show a progressive decrease in absolute inclinationup section associatedwith the equatorwardmotion of North Americafrom ~80° to ~30° (Fig. 2; Swanson-Hysell et al., 2009). Hematite thatformed due to late-stage hydrothermal activity should haveshallower inclinations than that isolated for the magnetite in agiven flow—unless its magnetization was completely dominated byinternal coupling. In contrast to the expected direction for late-stagehematite formation, the mean difference in the absolute value ofinclination between the magnetite and hematite fits (rather than theabsolute value of the difference) for the same set of samples showhematite fits to be 1.5° steeper on average than the magnetite fits.These results suggest that hematite formed during initial cooling, orduring low-temperature oxidation that occurred soon after flowemplacement, thereby recording a similar snapshot of the geomag-netic field as the magnetite.

In flows without the antiparallel component, little remanence islost during thermal demagnetization between 590 and 650 °C(Fig. 1A,D). Where the antiparallel component is present, itsrelative strength varies from marginally detectable (Fig. 1E), tooverwhelming any remanence held by magnetite (Fig. 1M). Thepresence of the antiparallel component is variable within thestratigraphy at Mamainse Point as it can be found in one flow andnot in the flow that is immediately above or below it stratigraphi-cally (Fig. 2). This variable presence suggests that its formation istied to the unique eruptive and oxidative history of each individualflow. On average, the flows we have studied paleomagnetically thatcontain the antiparallel component are thinner (mean thickness of3.4 m) than those where no such component is expressed (meanthickness of 7.5 m). The directions of the “antiparallel component”are close to truly antiparallel with angular differences between theFisher mean for that component in a given flow and that of the

hematite

magnetite

F

G

HThermal Demag (°C)

0.6

0.8

1.0

1.2

0.4

0.2

100

250 004

J0 =4.56e-7 Am2

500

575

670

650

NRM (J0)

Thermal (0-675°C)

NR

M

Z,N

W

InclinationDeclination

TILT-CORRECTED

E

675

675

650

650

575

575

500

J (1

0-4

A m

2)

MP

105-

20.4

MP

111-

149.

2

hematite

magnetite

E

MP

209-

5.4

hem_sr

J

K

TT

TT

Z,N

InclinationDeclination

TILT-CORRECTED

565

450

620

660

660565

450

620

I

1.0LThermal Demag (°C)

1.0

0.75

0.5

0.25

100

200

300 004

J0=1.90e-5 Am2

J0=1.14e-7 Am2

500

565

650

605

NRM (J0)

Thermal (0-675°C)

NR

M

hematitemagnetite

C

B

MP

105-

29.4

5

MN

Z,N

InclinationDeclination

E

GEOGRAPHIC

Thermal Demag (°C)

1

2

3

4

5

100

175

250

325

450

500004

J0=5.45e-7 Am2

575

590

640

NRM (J0)

Thermal (0-675°C)

NR

M

1.0P

hematite

hem_sr

O

550

550

500

500

565

565

575

575E

Z,N

A

InclinationDeclination

GEOGRAPHIC

1.0DThermal Demag (°C)

1.0

0.75

0.5

0.25

J/J 0

J/J 0

J/J 0

J/J 0

100

175

250

325 004

500

575 62

0

670

550

NRM (J0)

Thermal (0-675°C)

NR

M

Fig. 1. Representative vector component diagrams (Zijderveld, 1967) (A, E, I and M), tilt-corrected equal-area projections (B, F, J and N), least-squares fits (C, G, K and O), andmagnetic intensity (J/J0) plots (D, H, L and P) showing demagnetization behavior for Mamainse Point basalts with varying contributions from the antiparallel hematite component. Inthe vector component diagrams, the high-temperature components are separated into the magnetite component (traced with a blue arrow), the highest-temperature hematitecomponent (traced with a red arrow) and the antiparallel hematite component (traced with a green arrow and labeled “hem_sr”). Least-squares fits of these components aresummarized in equal-area projections (C, G, K and O). In all plots, green-filled shapes are natural remanent magnetization (NRM) directions and red-filled shapes reflect thermaldemagnetization. In the vector component diagrams, closed-squares/open-squares represent declination/inclination, while in the equal-area nets color-filled/color-rimmed circlesrepresent lower/upper-hemisphere directions.

173N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

ChRMinclination ( )

antiparallel componentinclination ( )

angle between ChRMand antiparallel vectors ( )

meters

reve

rsed

reve

rsed

norm

al

0

500

1500

1000

2000

2500

3000

3500

norm

al

reversednormal

0 30 60 90

reversednormal

0 30 60 90

hematiteChRM

mixedChRM

magnetiteChRM

90 120 150 180

Fig. 2. a) SimplifiedMamainse Point stratigraphy and inclination for the Fisher mean of ChRM directions of each flow (~6 samples per flow) and of the antiparallel component wherepresent in multiple samples. The data are shown as the absolute value of inclinations to allow for easy comparison between reversed and normal directions. Also shown are theangular differences between the Fisher mean of the ChRM and of the antiparallel component. When a magnetite component was resolvable, the angular difference was calculatedbetween that component and the antiparallel component (open triangles). Otherwise the difference was calculated between the high-temperature unblocking hematite and theantiparallel component (closed triangles) or a ChRM mean that contained both magnetite and hematite components (half-closed triangles).

174 N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

characteristic remanence direction ranging from 171.0 to 178.0(with a mean of 175.2, Fig. 2). The large decrease in inclination seenin the magnetite remanences is mirrored by that of the antiparallelcomponent. This context strongly suggests that the antiparallelcomponent is not the result of a phase that formed and acquired itsmagnetization from an external field that was subsequentlyreversed from the field in which the magnetite thermal remanencemagnetization formed. Rather, the antiparallel component is a self-reversed magnetization that acquired a remanence antiparallel tothe other ferromagnetic minerals.

1.2. Background on self-reversed magnetizations

While the list of examples of self-reversed behavior is notparticularly long, such behavior has an important place in thehistory of paleomagnetism. Following the discovery of the self-reversed Mount Haruna dacite (Nagata, 1953), a debate ensuedconcerning whether reversed magnetizations result from reversalsof the geomagnetic field or from self-reversal during continuousnormal polarity. This controversy was settled by the discovery ofseafloor magnetic anomalies (Vine and Matthews, 1963) and theconsistency of magnetostratigraphic sections (Cox, 1969)—pavingthe way for the plate tectonic revolution and the development ofthe geomagnetic polarity time-scale. Nevertheless, self-reversedcontributions to natural remanence continue to be identified andtheir occurrence and carriers need to be fully understood in orderfor the origins of paleomagnetic directions and intensity to beconfidently interpreted.

Self-reversed magnetizations can arise either through N-typeferrimagnetism, when a single magnetic phase behaves in thisfashion, or through coupling between two distinct phases. There arefew examples of N-type ferrimagnetism present in the naturalremanence of samples, as most minerals with N-type behavior onlyexhibit it at temperatures below those typical for Earth's surface.Recently, Doubrovine and Tarduno (2004) discovered a phase withN-type ferrimagnetic behavior in ocean floor basalts whose self-reversal was reproducible in the lab. This phase unblocks bytemperatures of 300 to 400 °C and was demonstrated to be held bytitanomaghemite that formed through low-temperature oxidation ofprimary titanomagnetites. Self-reversal of the primary thermalremanent magnetization (TRM) was proposed to occur throughionic reordering that led to the antiparallel A sublattice contributingmore substantially to the magnetic moment than the usuallydominant B sublattice.

The most well-documented self-reversing mineral phases aretitanium-rich titanohematites (Fe2−yTiyO3) with a y value between0.5 and 0.8 (e.g. Haag et al., 1990; Lawson et al., 1987; Nagata, 1953).These self-reversed titanohematites consist of exsolved Ti-rich and Ti-poor phases whose interaction gives rise to self-reversal (Dunlop andÖzdemir, 1997). The magnetization of antiferromagnetic Ti-poortitanohematites is weaker than that of ferrimagnetic Ti-rich titano-hematite and is blocked at higher temperatures. During cooling, moreweakly magnetized Ti-poor regions first become magnetized in adirection parallel to the external field. The crystallographically-controlled contacts between intergrowths result in exchange couplingbetween the phases. As temperatures cool, the Ti-rich regions acquire

175N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

their remanent magnetization under the influence of negativeexchange coupling with the already magnetized Ti-poor regionsoverwhelming any influence from the external field. As a result, these

0 500

0.5

1

1.5

0 500

0.5

1

1.5

2

0 100 200 300 400 500 6000

0.2

0.4

0.6

0.8

1

0 100 200 300 400 500 6000

0.2

0.4

0.6

0.8

1

x 10-7 x 10-7

x 10-7 x 10-7

Z,N

MP209-5.4Thermal Demag of NRM

MP209-5.4AF Demag of NRM

MP209-5.8AF Demag of NRM

MP209-5.8Thermal Demag of NRM

25 mT

7

NRM

NRMNRM

NRM

NRM

NRM

25 mT

450°C

565°C

660°C

605°C

NRM

400°C

400°C

565°C

550°C

550°C635°C

575°C

565°C

NRM450°C

565°C

620°C

E

Z,N

E

ALTERNATINTEMPERATURE (°C)

ALTERNATINTEMPERATURE (°C)

inclination

declination

inclination

declination

inclination

declination

inclination

declination

geographic coordinates geographic coordinates

geographic coordinates geographic coordinates

MA

GN

ET

IZAT

ION

(A

m2 )

MA

GN

ET

IZAT

ION

(A

m2 )

MA

GN

ET

IZAT

ION

(A

m2 )

MA

GN

ET

IZAT

ION

(A

m2 )

Fig. 3. Vector component diagrams (Zijderveld, 1967) and plots of magnetic moment forremoved by 565 °C (magnetite), the component removed between ~620 °C and ~675 °C (h~565 °C and ~620 °C. Demagnetization behavior is shown for thermal demagnetization of theand for thermal demagnetization of a laboratory induced thermoremanent magnetization (170 mT are, for both samples, very similar to the point reached by the 565 °C demagnetsignificantly affected by the AF demagnetization and that the coercivity of the other two co

Ti-rich regions become magnetized in a direction antiparallel to theexternal field swamping the signal from the Ti-poor regions andresulting in a net self-reversed magnetization. Since titanohematites

100 150

100 150

0 100 200 300 400 500 6000

2

4

6

8

0 100 200 300 400 500 6000

2

4

6

x 10-6

x 10-6

MP209-5.4Thermal Demag of TRM

MP209-5.8Thermal Demag of TRM

170 mT

170 mT

40 mT

40 mT

70 mT

170 mT

580°C

680°C

450°C

300°C

NRM

0 mT

70 mT

70 mT

170 mT

580°C

680°C

450°C

300°C

NRM

Z,N

E

E

Z,N

Z,NZ,N

E

E

G FIELD (mT) TEMPERATURE ( °C)

G FIELD (mT) TEMPERATURE ( °C)

inclination

declination

inclination

declination

core coordinates

core coordinates

MA

GN

ET

IZAT

ION

(A

m2 )

MA

GN

ET

IZAT

ION

(A

m2 )

two samples with at least three components to their magnetization: the componentematite parallel to the magnetite) and the antiparallel component removed betweennatural remanent magnetization (NRM), alternating field demagnetization of the NRMTRM) along the positive z-axis. Note that the end points of the AF demagnetization atization step. This result indicates that the magnetite was the only component to bemponents is higher than 170 mT.

176 N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

of this composition have Curie temperatures between 200 °C and−27 °C, they cannot be the carrier responsible for the antiparallelcomponent in the Mamainse Point flows.

Self-reversal also is possible in decoupled magnetic phases if oneof the phases forms in close proximity to the other under the influenceof an interaction field that locally can overwhelm Earth's field (Néel,1951). This coupling is referred to as magnetostatic interaction andcan be either positive or negative—in which case it theoretically couldresult in a self-reversal (Dunlop and Özdemir, 1997; Krása et al.,2005). This type of interaction has been invoked to explain self-reversed titanomaghemites in basalt flows from the Neogene andQuaternary (Krása et al., 2005).

In regard to partial self-reversal in the Keweenawan rift, a 1968study of the Portage Lake volcanics found a rise in magnetic momentduring the first thermal demagnetization step to 200 °C in six of thesamples studied (Vincenz, 1968). The author inferred the presence ofa partial self-reversal of magnetization from this rise in moment, aswell as from pTRM experiments, and ascribed it to titanohematite.This claim is difficult to evaluate without vector component plots.Regardless of the origin of this potentially self-reversed component,its low unblocking temperature renders it quite distinct from theantiparallel component investigated herein.

2. Results

2.1. The presence of the antiparallel component in the natural remanentmagnetization

Alternating field (AF) demagnetization of the natural remanenceshows that the end point reached by levels of 170 mT for sampleswith the self-reversed component is the same as that reached bythermal demagnetization to 565 °C (Fig. 3). At this level of AF andthermal demagnetization, the magnetite component has beenremoved, but the antiparallel component is still present. This resultshows the coercivity of the antiparallel component to be greaterthan 170 mT and demonstrates that the antiparallel component isfully present in the natural remanence of unheated samples. As aresult, we can be confident that the antiparallel component is not aresult of alteration occurring in the lab during thermaldemagnetization.

2.2. Decrease in NRM associated with hematite formation

The strength of the NRM is lower for flows with significanthematite than for flows with negligible hematite. The NRMmoment is even lower for flows that have the antiparallel phase(although this difference is associated in part with the vectorsubtraction of the antiparallel phase from the overall moment;Fig. 4). The decrease in moment associated with progressivehematite formation indicates that some of the iron used in the

0 0.005 0.01 0.015 0.02 0.0

0.2

0.4

0.6

0.8

1

NRM mome

cum

ulat

ive

frac

tion

Fig. 4. Empirical cumulative distribution functions for the moment of the NRM for samplesremoved by 590 °C but with no antiparallel component and samples with the antiparallel c

formation of post-eruptive hematite was obtained at the expenseof primary magmatic magnetite.

2.3. pTRM and TRM acquisition and demagnetization experiments

Partial thermoremanent magnetization (pTRM) experimentswere conducted on specimens taken from samples from whichother specimens had been used for thermal demagnetization of theNRM. The experimental protocol for the pTRMwas designed to targetthe magnetic minerals with laboratory blocking temperaturesbetween 620 °C and 600 °C such that a portion of the mineralsholding the antiparallel remanence would be magnetized withoutaffecting the magnetite and minimally magnetizing the highunblocking temperature hematite. The protocol proceeded asfollows: After measuring the NRM of the samples, the sampleswere heated to 670 °C in zero field (typical background field levels inthe furnace used are b5 nT). The samples were cooled to 620 °Cwhilestill in zero field. When the samples had equilibrated to temperatureat 620 °C, a 35 μTfieldwas applied along the z-axis of the samples andthe samples were cooled in this field to 600 °C. The samples werethen cooled to room temperature in a zero field. The remanencesacquired during this pTRM experiment were measured on a 2 Gsuperconducting rock magnetometer and were found to be in thesame direction as the applied field. This result eliminates thepossibility that the antiparallel phase behaves as an N-typeferrimagnet and instead suggests that its formation is related tointeraction between two phases.

A full thermoremanent magnetization (TRM) was applied to asubset of specimens to investigate whether interactions with the highunblocking temperature hematite could be causing a self-reversal ofthe antiparallel phase. The samples were heated to 670 °C in zerofield. Once equilibrated at 670 °C, a 35 μT field was applied along thez-axis of the samples and the samples were cooled in this field downto room temperature (~25 °C). The resulting magnetization, asrevealed through the thermal demagnetization of the laboratory-induced TRM, was a single component remanence parallel to theapplied field (Fig. 3). This result indicates that the antiparallelcomponent does not acquire a self-reversed TRM due to interactionwith the high unblocking temperature hematite phase on laboratorytime-scales.

2.3.1. Mössbauer spectroscopyMössbauer spectra were obtained from three samples of crushed

bulk rock. The magnetic hyperfine parameters and the relativeproportions of each component determined from the spectra arelisted in Table 1. Samples MP101-36.9 and MP111-143.8 contain theantiparallel phase while sample MP101-9.6 does not and itsremanence is interpreted to be dominated by magnetite on thebasis of thermal demagnetization data. The Mössbauer spectrum ofthe MP101-9.6 sample was fitted by two sextets, corresponding to

025 0.03 0.035 0.04 0.045 0.05

nt (10-3 Am2)

>95% of remanenceremoved by 590 C

<20% of remanenceremoved by 590 C

samples with theantiparallel component

with 95% of their remanence removed by 590 °C, samples with 20% of their remanenceomponent.

Table 1Magnetic hyperfine parameters at 300 K determined through Mössbauer spectroscopy.

Phase Parameter SampleMP101-9.6

SampleMP101-36.9

SampleMP111-143.8

Hematite Bhf (T) 51.2(1) 51.5(1)QS (mm/s) −0.21(1) −0.21(1)IS (mm/s) 0.44(1) 0.36(1)% 64 74

Fe2+ QS (mm/s) 2.61(2) 2.59(1) 2.04(1)IS (mm/s) 1.15(1) 1.18(1) 1.13(1)% 42 29 26

Fe3+ QS (mm/s) 0.97(4) 0.74(4)IS (mm/s) 0.33(2) 0.49(2)% 9 7

Magnetite Bhf (T) 49.2(2)Site A QS (mm/s) −0.07(1)

IS (mm/s) 0.30(1)% 25

Magnetite Bhf (T) 46.0(1)Site B QS (mm/s) 0.01(1)

IS (mm/s) 0.68(1)% 24

Notes: All of the spectra were obtained at room-temperature using a conventionalconstant-acceleration spectrometer in transmission geometry. The spectrometer uses a57Co/Rh source and α-Fe calibrate isomer shifts and the velocity scale. The magnetichyperfine field (Bhf), isomer shift (IS) and quadrupole splitting (QS) and relative area(%) were determined from the spectra using the NORMOS program (Brand, 1987).Components identified through thermal demagnetization spectra for each samplewere: MP111-9.6 (magnetite), MP101-36.9 (magnetite, antiparallel component,hematite), MP111-143.8 (antiparallel component, hematite).

Velocity (mm/s)

Tran

smis

sion

(a.

u.)

MP111-143.8

MP101-9.6 magnetite

magnetite hematite hematite (self-rev)

hematite hematite (self-rev)

MP101-36.9

1.00

1.00

1.00

0.99

0.99

0.99

0.98

0.98

0.98

0.97

0.97

0.97-12 -8 -4 12840

Fig. 5. Mössbauer spectra at room temperature for whole rock powder of three basaltsamples. Open circles are experimental points, the colored lines correspond to thespectrum best fit (red), hematite (blue), site A of magnetite (orange), site B ofmagnetite (cyan), Fe2+ (green) and Fe3+ (magenta). Each specimen is labeled with themagnetic phases (magnetite, hematite, self-reversed hematite) that are resolvable inthe thermal demagnetization data for that sample.

177N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

sites A and B of magnetite, and two doublets of paramagnetic Fe2+

and Fe3+ associated with Fe-silicates (Fig. 5). The relative proportionof sites A and B was higher than the value of 0.5 that is expected forstoichiometric magnetite indicating that some degree of oxidation hasdecreased the contribution of site B. No hematite was detected in thespectrum for MP101-9.6.

The Mössbauer spectra show the samples with the self-reversedphase to be dominated by hematite (74% for MP111-143.8.; 64% forMP101-36.9) with the rest of the Fe residing as paramagnetic Fe2+

and/or Fe3+ in association with Fe-silicates (Fig. 5, Table 1). Themagnetic hyperfine parameters that were fitted for the hematite inboth samples are close to the magnetic hyperfine parameters of thestoichiometric phase (Murad and Cashion, 2003). Given that theantiparallel component comprises a significant portion of thesesamples' magnetization, it is expected that it would contribute tothe Mössbauer spectroscopy data which are sensitive to componentsthat comprise N2% of the sample's total iron. For example, maghemitewould have a unique signature on the spectra which is not observed ineither MP101-36.9 or MP111-143.8. Therefore, the Mössbauerspectroscopy data strongly suggest that the antiparallel remanenceresides in a population of stoichiometric hematite.

2.4. Low-temperature cycling

The Mamainse Point basalts can be grouped into 4 categories:

(1) Flows with remanence dominated by magnetite (Fig. 1A).(2) Flows with remanence held by both magnetite and hematite

(Fig. 1E).(3) Flows with remanence held by magnetite, self-reversed

hematite and hematite (Fig. 1I).(4) Flows with remanence held by self-reversed hematite and

hematite with no appreciable magnetite present in thermaldemagnetization data (Fig. 1M).

Low-temperature experiments were performed using a QuantumDesign Magnetic Properties Measurement System. Specimens fromsamples in each of the above categories were pulsed with a 2.5 T fieldat room temperature before being cooled to 10 K and then warmed

back to room temperature in near zero field with frequent measure-ments of the isothermal remanent magnetization (Figs. 6, S1-S2).

The Verwey transition at ~120 K is very pronounced in flowswhose remanence is dominated by magnetite (Fig. 6). Oxidation ofmagnetite, and the resulting deviation of stoichiometry, has the effectof spreading out and then suppressing the Verwey transition(Özdemir et al., 1993). The excellent expression of the Verweytransition in the flows whose remanence is dominated by magnetitedemonstrates that there are flows in the succession that have beenvery minimally affected by post-eruptive oxidation. There is acontinuum in the suppression of the Verwey transition (Fig. 6) fromsamples of type 1 (remanence dominated by magnetite) to flows oftype 4 (remanence held by self-reversed hematite and hematite).Thiscontinuum likely represents a progression in oxidation from type 1 totype 4 samples. The Verwey transition is still pronounced in flows oftype 2 (magnetite+hematite) indicating that despite hematiteformation magnetite remains well-preserved (see MP105-20.4 inFig. 6). In flows with self-reversed remanence, there begins to be asignificant suppression of the Verwey transition (see MP111-11.0 inFig. 6) indicating that the self-reversed hematite has formed at theexpense of magnetite. These results suggest a connection betweenprogressive oxidation, the destruction of stoichiometric magnetite,and the formation of the self-reversed component.

0 50 100 150 200 250 3000.020

0.025

0.030

0.035

MP105-20.4

Temperature (K)

1

2

3

0 50 100 150 200 250 3000.15

0.20

0.25

MP209-64.4

Temperature (K)

0

1

2

3

4

0 50 100 150 200 250 300

2.8

3.0

3.2

3.4

x 10-3 MP111-143.8

Temperature (K)

0

1

2

3

4

5

6

0 50 100 150 200 250 300

0.0095

0.0100

0.0105

0.0110

0.0115

MP111-11.0

Temperature (K)

0.0

0.5

1.0

1.5

2.0

2.5

Mom

ent (

Am

2 /kg)

Mom

ent (

Am

2 /kg)

Mom

ent (

Am

2 /kg)

Mom

ent (

Am

2 /kg)

Mom

ent (

Am

2 /kg)

Mom

ent (

Am

2 /kg)

0 50 100 150 200 250 300

0.10

0.15

0.20

0.25

0.30MP101-9.6

Temperature (K)

dM/d

T (

Am

2 K-1)

dM/d

T (

Am

2 K-1)

dM/d

T (

Am

2 K-1)

dM/d

T (

Am

2 K-1)

dM/d

T (

Am

2 K-1)

dM/d

T (

Am

2 K-1)

0

2

4

6

x 10-3

x 10-4

x 10-5

x 10-3

x 10-5

x 10-6

0 50 100 150 200 250 300

0.018

0.019

0.020

0.021

MP101-36.85

Temperature (K)

2

4

6

8

10

v

v

v v

v

coolingwarming

hematitemagnetite

magnetite

hematitehem_self-rev

magnetite

magnetite

hematitehem_self-rev

magnetite

v

v

hematitehem_self-rev

Fig. 6. Low-temperature cycling experiments on Mamainse Point basalts where the samples were exposed to a 2.5 T field at room temperature and then cycled in zero field to 10 Kbefore returning to room temperature. Points in blueweremeasured while the samples were cooling, while points in redweremeasured duringwarming. The derivative of the curveis shown for the cooling trajectory with a thin black line. Each specimen is labeled with the magnetic phases (magnetite, hematite, self-reversed hematite) that are resolvable in thethermal demagnetization data for that sample. The Verwey transition of magnetite at 120 K is labeled on each plot with a tick mark and a “V”.

178 N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

The Morin transition of hematite is not expressed as a sharp dropin moment confined to 260–250 K in any of the samples studied. Insamples where the presence of hematite is inferred from thermaldemagnetization data, there is evidence for a broad Morin transitioninitiating at ~250 K and continuing below 200 K (Fig. 6). Suppressionof theMorin transition occurswith very small hematite grains as thereis a strong dependence of the Morin transition temperature onparticle size below 200 nm until ~30 nm when the transition is nolonger observed (Jacob and Abdul Khadar, 2010; Zysler et al., 2003).This interpretation of Morin transition suppression implies that thereis a range of hematite grain sizes present in the samples, some ofwhich are quite fine-grained (b200 nm). It also has been hypothe-sized that lattice defects could similarly lower the temperature of theMorin transition by anchoring regions of surface spins (Özdemir et al.,2008). Since most hematite grains formed at low-temperatures havemagnetizations dominated by defect moments (Tauxe, 2010), thisbehavior could be affecting the expression of the Morin transition inthe Mamainse Point basalts.

2.5. Microscopy

Considering their 1.1 Ga age, the basalts at Mamainse Point areexceptionally well-preserved. This preservation is evidenced on the

outcrop scale by the excellent preservation of original eruptivefeatures, such as pahoehoe textures on flow tops (Swanson-Hysellet al., 2009), and on the microscope scale by the preservation oforiginal mineral phases. The final mineral assemblage in each flow hasbeen influenced by three important steps: (1) original emplacementand cooling of the flow that was accompanied by high-temperatureoxidation, (2) oxidation related to subsequent subaerial weathering,and (3) low-grade zeolite to lower-most prehnite-pumpellyitemetamorphism associated with burial (Massey, 1983; Smith, 1974).Individual flows exhibit a range of porosities and experienced varyingdurations of subaerial alteration prior to being buried by subsequentflows. Consequently, the alteration of primary silicates and oxidesvaries heterogeneously throughout the Mamainse Point basalts. Thisheterogeneity could have been accentuated by the presence ofhydrous phases in some flows during reactions associated with low-grade burial metamorphism (Smith, 1974). Following this initialburial, theMamainse Point successionwas uplifted atop rift sedimentsalong a reverse fault thought to be related to far-field stresses of the~1.0 GaGrenville Orogeny (MansonandHalls, 1994). In the time since,there have been no significant thermal events in the region. Belowwedescribe the range of alteration textures associated with the silicateand titanomagnetite components of the rock, which in turn, exert astrong influence on the paleomagnetic remanence of the basalt flows.

179N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

2.5.1. Silicate mineralsThin sections of minimally altered basalts samples show sub-

ophitic textures with calcic plagioclase laths (up to 2 mm) and moreequant clinopyroxene (up to 0.5 mm) making up the majority of thesilicate mineralogy and hosted in a fine-grained groundmass.Plagioclase laths contain minor amounts of sericite. Clinopyroxenegrains are well-preserved and range from small grains in the

Fig. 7. Photomicrographs and SEM images of hematite forming in altered silicate minerals inMP101-12.4 (G, H, I) and MP111-143.8 (J, K, L). MP111-143.8 contains the self-reversed cowere shown through EDS and EBSD to be titanomagnetite while the skeletal magnetite morphematite (see Fig. 8). The photomicrographs (A, D, G, J) were taken in plane-polarized light. Tbeam at 15 kV acceleration. The field of view in C, F, I and L are shown with boxes on the B

groundmass to larger grains that sometimes enclose plagioclaselaths. In addition to small pyroxenes, the fine-grained groundmassalso contains smaller grains and skeletal titanomagnetite (describedbelow). The lower part of the stratigraphy is dominated by high-Mgolivine-rich picritic and basaltic lavas (Klewin and Berg, 1991). Wherepresent, olivine grains are typically altered to hydrous silicates in theinterior and have rims of hematite (Fig. 7). Vesicles occur in many

samples MP101-41.1 (A, B, C) and MP105-29.3 (D, E, F) and of iron oxides in samplesmponent while MP101-12.4 does not. The large iron oxides in the MP101-12.4 imageshologies in MP111-143.8 were shown through EBSD analysis to have been replaced byhe secondary electron SEM images (B, C, E, F, H, I, K, L) were collected with the electron, E, H and K image respectively.

180 N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

samples and are commonly filled with amygdules consisting of anoutside ring of ~100 μm quartz crystals which are then in-filled withfine-grained zeolite, chlorite or clay minerals. In more oxidizedsamples, plagioclase becomes progressively more albitized, as notedby earlier workers who observed that albite is more prevalent withinthin flows and on flow tops and bases (Smith, 1974), and the extent ofsericitization within the laths increases. The groundmass can also bealtered to clay minerals and hematite.

From a paleomagnetic perspective, the most important silicatealteration feature is the oxidation of iron-rich silicate phases (e.g.olivine). The resulting oxidation texture typically consists of core clayminerals surrounded by a 50–100 μm thick, optically opaque rim ofhematite grains (Fig. 7A and D). The hematite rims often preserve aeuhedral outline of the precursor silicate mineral and hematite septacan cross the clay-rich interiors. Scanning electron microscope (SEM)images of these hematite rims show them to be composed of regularlyaligned, tightly packed, submicrometer hematite platelets (Fig. 7C).Energy dispersive spectroscopy (EDS) shows the platelets to becomposed of stoichiometric hematite. Electron backscatter diffraction(EBSD) of these oxide rims shows that the crystallographic orienta-tions of the platelets are predominantly aligned in a single orientation.Because the clay–mineral aggregates themselves are randomlyoriented throughout a sample, no remanence anisotropy is expectedand their remanence is anticipated to be a chemical remanentmagnetization acquired under the influence of the external magneticfield as the hematite platelets grow through their blocking volumes.

2.5.2. Iron oxidesThe occurrence of the antiparallel component within the

Mamainse Point basalts is closely linked to the degree of preservationof titanomagnetite grains. In flowswith no evidence of the antiparallelcomponent, titanomagnetite grains are beautifully preserved andoccur as skeletal grains internally subdivided into sub-micrometer, Ti-poor subgrains by ilmenite lamellae exsolved along the {111}magnetite

(Fig. 7G–I). This type of intergrowth texture (stages C3 and C4 ofHaggerty, 1991) occurs during initial cooling when disequilibriumbetween the primary titanomagnetite and the oxygen fugacity of thebasalt results in oxidation of Fe, diffusion of Ti to the resultingvacancies in the octahedral sites of the {111} crystal planes, and theformation of ilmenite lamellae (Aragon et al., 1984; Buddington andLindsley, 1964).

In more oxidized samples, titanomagnetite grains have undergonemartitization (the transformation of magnetite into hematite). Insome grains, this process did not reach completion and relict grains ofmagnetite are observed within the secondary hematite (Figs. 8 andS3). In the most oxidized samples, martitization caused the completereplacement of titanomagnetite, leaving behind hematite grains thatare quite small, but whose collective morphology is a pseudomorphafter skeletal titanomagnetite. When examined under high magnifi-cation, the secondary hematite occurs as platelets generally ranging insize from 150 to 300 nm. In addition to secondary hematite,pseudomorphs of skeletal titanomagnetite sometimes contain ilmen-ite lamellae, whichare assumed to be primary, and secondary titanitewith grain diameters b100 nm. In some samples with both martitizedtitanomagnetites and well-preserved titanomagnetites, it was ob-served that martitization preferentially occurred in grains that wereclose to vesicles.

In order to understand the crystallographic relationship of thesecondary hematite to the primary titanomagnetite, we measured theorientations of the two phases in partially martitized grains usingEBSD; a technique that is readily capable of distinguishing betweencubic and trigonal systems. The magnetite and hematite are orientedsuch that the {0001}hematite are always parallel to {111}magnetite, whilethe {110}magnetite is sometimes parallel to the {112̄0}hematite (Fig. 8).This epitaxial relationship between martite and the host magnetitehas long been recognized on the basis of reflection microscopy and X-

ray diffraction data (see Davis et al., 1968 and references therein) andrecently has been identified in oxidized banded iron formation usingEBSD data (Barbosa and Lagoerio, 2010). This crystallographicrelationship between parent magnetite and the daughter hematiteresults because both the low temperature oxidation of magnetite tomaghemite and the subsequent inversion of maghemite to hematiteare topotactic transformations (Davis et al., 1968; Dunlop andÖzdemir, 1997).

The transformation of magnetite to hematite in rocks is mostcommonly attributed to an oxidation reaction such as:

4Fe3O4 magnetiteð Þ + O2→6Fe2O3 hematiteð Þ: ð1Þ

Recent work on banded-iron formations has shown that similartextures may also form by non-redox mechanisms (Ohmoto, 2003;Otake et al., 2007), such that:

Fe3O4 magnetiteð Þ + 2Hþ→Fe2O3 hematiteð Þ + Fe2 + + H2O: ð2Þ

Eq. (1) is associated with a volume increase of 1.66%, while Eq. (2)is associated with a volume decrease of 32.22% (Mücke and Cabral,2005). The samples examined in this study do not show evidence for adramatic 32% volume loss, and consequently, we interpret thereplacement textures in the Mamainse Point basalts to be due tooxidation. This interpretation of oxidation is further supported by theobservation of altered Fe-silicates with thick rims of hematite withinthe same samples that contain the martitized magnetite. Takentogether, these results demonstrate that the magnetite in samplescontaining the antiparallel component has been largely martitized tohematite (Figs. 7 and 8).

3. Discussion

3.1. Composition of the antiparallel remanence

The high coercivity of the antiparallel component revealedthrough AF demagnetization combined with the Mössbauer spectros-copy data provides strong evidence that the self-reversed remanenceis naturally present in the samples and is held by hematite. Thecontinued unblocking of the antiparallel component up to tempera-tures of ~650 °C demonstrates that the component is not held bymagnetite and is unlikely to be held by maghemite due to the typicalinversion of maghemite to hematite at these temperatures.

Although hematite is known to have a Néel temperature of~675 °C, its unblocking temperature is also a function of grain volume(v), saturation magnetization (Ms), and microscopic coercivity (Hc,Pullaiah et al., 1975). If hematite is responsible for the antiparallelremanence, then the grains holding this remanence must bemagnetically stable on a geological timescale and unblock between550 and 640 °C. According to Néel (1949), the relaxation time (τ)magnetization held by an assemblage of non-interacting single-domain grains can be calculated as:

τ = C−1evμoMsHc

2kT ð3Þ

where C is the characteristic frequency of thermal fluctuation (in Hz),μo is the permeability of free space, k is Boltzmann's constant, and T istemperature. Both Ms and Hc are themselves a function of temper-ature, and we use the empirical temperature-dependent behavior ofMs reported by Özdemir and Dunlop (2005) in this analysis. Themicroscopic coercivity of hematite is dominated by magnetoelasticeffects and the temperature dependence of these effects remainspoorly defined. Following the example of Pullaiah et al., (1975), wecalculate the temperature dependence of Hc as proportional to Ms(T)3

normalizing to the room-temperature coercivity of a population of

Fig. 8. SEM images, EBSD mineral identification maps, and pole figures of hematite and magnetite orientations from EBSD analysis of hematite pseudomorphed after skeletalmagnetite. The pole figures are lower hemisphere equal-area projects that have been contoured using a half-width of 10° and clustering of 5°. White arrows in MP111-143.8 pointout the prismatic {112̄0} planes of hematite that correspond with the {110} magnetite planes. The parallel orientation of the basal {0001} hematite planes with the octahedral {111}magnetite planes is quite clear in the pole figures.

181N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

naturally-occurring 500–3000 nm diameter hematite grains (260 mT;Dunlop and Kletetschka, 2001).

Modeled unblocking temperatures for grains of hematite as afunction of diameter are depicted in Fig. 9. Results for relaxation timesof 5 min and 4.5 billion yr are shown in order to demonstrate thetemperatures that would unblock given grain sizes on laboratory andgeological timescales. This theoretical calculation approximates thegrains as spheres for the purpose of calculating volume and, in effect,imparts the uniaxial shape anisotropy of the empirically determined

room-temperature Hc value. Hematite grains with greater anisotropywould have higher values of Hc, and therefore longer relaxation timesand slightly higher unblocking temperatures (Özdemir and Dunlop,2002). The results of the calculations indicate that hematite grainswitha volume of 4.8×10−21 m3 (~210 nm diameter) will be magneticallystable on geological timescales and have an unblocking temperaturesimilar to that of pure magnetite on laboratory timescales. At roomtemperature, hematite grains with volumes below 3.8×10−22 m3

(90 nm in diameter) will behave superparamagnetically. At hematite

(Un)

Blo

ckin

g Te

mpe

ratu

re (

°C)

TN hematite

TC magnetite

Grain Diameter (nm)0 100 200 300 400 500 600

0

100

200

300

400

500

600

700

τ = 5 minutes

τ = 4.5 billion yrs

antiparallelhematite

component

Fig. 9. Calculated blocking temperature of idealized spherical hematite grains.Calculations using thermal fluctuation rates of 109, 1010 and 1011 s−1 are shown inblue (light grey in print), green (medium grey in print) and red (dark grey in print),respectively. For reference, the Curie temperature for magnetite and the Néeltemperature for hematite are shown. The hematite grains holding antiparallelremanence are estimated to have diameters between 190 and 340 nm.

182 N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

grain volumes above 3.4×10−20 m3 (~400 nm diameter), unblockingtemperatures become nearly indistinguishable from that of the largesthematite grains and approach the ideal Néel temperature for hematite(675 °C). These calculations are consistentwith empirical studies on theinfluence of grain-size on hematite's rock magnetic properties. Thethreshold diameter between superparamagnetic and single-domainbehavior in hematite at room temperature is known to be between 40and 160 nm (Raming et al., 2002), while the transition from SD to MDbehavior occurs at grain sizes near100 μm(KletetschkaandWasilewski,2002). This analysis indicates that an assemblage of hematite grainswith diameters between 190 and 340 nm could have unblocked onlaboratory timescales in the temperature range that unblocked theantiparallel component in the Mamainse Point basalts.

The grain size predictions from these calculations are consistentwith the hematite grain sizes estimated for martitized former skeletalmagnetite crystals. This grain size analysis is possible through high-resolution EBSD mapping that estimates the extent of crystals basedon the regions of constant orientation. For a martitized magnetitecrystal in sample MP111-143.8, the grain size of these hematite grainsis estimated by this method to be 194±119 nm. SEM images showthe hematite grains associatedwith Fe-silicates to be larger—generallybetween 1 and 5 μm (Fig. 7). This larger grain size for the hematiteplatelets associated with Fe-silicates is consistent with such grainsholding the magnetite-parallel remanence that unblocks close to theideal Néel temperature.

3.2. Formation of the antiparallel remanence

In all samples with the antiparallel component there is alsosignificant remanence associated with hematite that unblocks close tothe Néel temperature. This result from thermal demagnetization isconfirmed by Mössbauer spectroscopy as well as optical and electronmicroscopy. Low-temperature cycling experiments show that thepresence of the self-reversed component is associated with progres-sive oxidation of magnetite grains. Taken together, these data point toa positive relationship between the degree of low-temperatureoxidation of a basalt flow and the presence of the antiparallel phase.

A low-temperature oxidation pathway is consistent with the pTRM/TRM results that suggest a chemical, rather than thermal, origin of theself-reversed remanence due to the needed interaction between theself-reversed phase and a phase with a lower blocking temperatureduring its formation.

The most likely mechanism for the formation of the self-reversal isthe transformation of metastable maghemite to hematite. Since theremanence of maghemite is blocked during low-temperature oxida-tion, its internal magnetic field strongly influences the formation ofthe daughter hematite. Laboratory experiments have demonstratedthat the hematite formed via the inversion of maghemite to hematitecan hold a remanence antiparallel to that of the parent maghemite(Hedley, 1968; McClelland and Goss, 1993). In these experiments,maghemite is produced by heating lepidocrocite (γFeOOH) leading tothe creation of maghemite followed by hematite all in the presence ofan applied field. The maghemite acquired a magnetization parallel tothe applied field while the hematite acquired a remanence that wasantiparallel to the field. In the results of McClelland and Goss (1993),the hematite self-reversal only occurred when the magnetization ofthe parent maghemite was still blocked at the time of its transfor-mation to hematite. McClelland and Goss (1993) explain this behavioras the result of exchange interactions between the parent maghemiteand the daughter hematite. The formation of a self-reversal during themaghemite–hematite transition had previously been predicted byBanerjee (1963). Exchange interactions between the two phases arepossible given that hematite and maghemite can share O2− anionsacross reaction surfaces as the result of the crystallographicallycontrolled orientation of the two phases during the inversion. Asdiscussed above, exchange coupling between maghemite and themagnetite from which it forms results in the retention of the originalmagnetite remanence direction. During subsequent inversion of themaghemite, negative exchange coupling across the phase boundarywould pass through the crystals, causing the daughter hematite tobecome magnetized in a direction antiparallel to the remanencecarried by the maghemite. This mechanism has been suggested as theexplanation for a self-reversed phase that is present in samples ofSiluro–Devonian andesites from the Lorne plateau of Scotland(McClelland, 1987).

The formation of hematite in association with silicates in thebasalts would not result in a magnetization of those grains that wassignificantly influenced by the ferromagnetic minerals in the rock.Therefore, we posit that the more ubiquitous hematite that formed inassociation with Fe-silicate phenocrysts, and the ground-mass of thebasalt, has a parallel remanence to the magnetite. This hematiteformationwould have begunwith deuteric oxidation of olivine duringinitial magma cooling (Haggerty and Baker, 1967) and could havecontinued within Fe-silicate phenocrysts and throughout the ground-mass as a result of subaerial weathering before the emplacement ofsubsequent flows. As low-temperature oxidation of the flow contin-ued, the maghemitization of the primary magnetite progressedfurther, particularly in the thin vesiculated flows that contain theantiparallel component. Maghemite formation occurs in weatheredbasalts in both temperate (Soubrand-Colin et al., 2009) and tropicalclimates (Garcia de Oliveira et al., 2002). While maghemitizationcould have been associated with later hydrothermal activity, the highlevel of similarity between magnetite and high-unblocking temper-ature hematite magnetizations (see section 1.1) suggests that suchlate-stage oxidation was not associated with significant hematiteproduction. These data may point towards the importance of deutericmetasomatism (autohydrothermal alteration) in addition to low-temperature oxidation. Regardless of when the maghemitizationoccurred, it subsequently inverted to fine-grained hematite that,through the exchange-coupling described above, acquired a rema-nence that was self-reversed with respect to the parent maghemite.The timing of this inversion is difficult to constrain, but is likely tohave occurred between 1.1 and 1.0 Ga during burial under post-rift

183N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

sediments prior to Grenville-aged uplift of the succession along theMamainse Point fault (Manson and Halls, 1994).

Prior to this work we are aware of only two other studies that haveattributed an antiparallel natural remanence removed above 500 °C toself-reversed hematite: the Lorne plateau andesites (McClelland,1987) and a study of diorite of the Buck's batholith in northernCalifornia (Merrill andGrommé, 1969). However, thismechanismmaybemore ubiquitous than thepaucity of published examples suggests asthe magnetite→maghemite→hematite pathway should be quitecommon. In order for the component to be recognized, it needs to becomparable in magnitude to remanence held by primary magnetiteand/or parallel hematite, and thermal demagnetization must proceedat high resolution (b20 °C). Antiparallel directions held by hematitehave been observed in many studies of hematite-rich oxidizedsediments (“red beds”; e.g. Robinson and McClelland, 1987; Rösleret al., 1997; Roy and Park, 1972; Schmidt and Williams, 1995; Turner,1981; Turner et al., 1984). The explanation given for these antiparallelremanences is that the samples contain hematite that formedsignificantly after deposition following a geomagnetic reversal.While this is a tenable explanation in the absence of significant platemotion, we suggest that antiparallel hematite in red beds could formthrough the self-reversalmechanismdescribed above asfirst proposedby McClelland and Goss (1993). In this scenario, hematite formed inassociationwith silicateminerals in the sediment acquires the polarityof the current geomagnetic fieldwhile hematite forming in associationwith maghemitized magnetite grains acquire a polarity antiparallel tothe detrital remanent magnetization of the sediment. In fact, somestudies of oxidized sediments with antiparallel magnetizations haveobserved the presence of martite (e.g. Schmidt and Williams, 1995;Turner, 1981; Turner et al., 1984). Future studies that reveal hematitewith an antiparallelmagnetization to themagnetite remanence, or theremanence held by another population of hematite, in any lithologyshould consider that instead of resulting fromprolonged acquisition ofchemical remanence across polarity reversals, this behavior may beattributable to martite self-reversal.

4. Conclusions

Some flows in the volcanic stratigraphy of the mid-continent riftatMamainse Point contain a phase that unblocks between 590 °C and650 °C and is antiparallel to the magnetite and high-unblockingtemperature hematite remanences. The geological and paleomag-netic context of these basalt flows provides robust evidence that thephase is self-reversed, rather than having been acquired duringsubsequent polarity intervals. The high coercivity of this antiparallelphase, its high unblocking temperatures and results fromMössbauerspectroscopy show the phase to be carried by hematite. Itssuppressed unblocking temperature suggests that the grains ofhematite carrying this magnetization are finer than the hematitepopulation carrying a magnetization parallel to the magnetiteremanence, which unblocks closer to the ideal Néel temperaturefor hematite. Electron microscopy shows that samples with the self-reversed component contain hematite pseudomorphs after magne-tite (martite) that have formed within skeletal titanomagnetitecrystals and that there is a preferential alignment of the basal plane ofthe hematite with the {111} plane of the magnetite. These hematitegrains are in a size range consistentwith their suppressed unblockingtemperatures. The probable mechanism for the self-reversal isexchange coupling with the parent maghemite that formed duringlow-temperature oxidation. Variable porosities, permeabilities andexposure time between individual flows provide a ready explanationfor why the antiparallel phase is found heterogeneously throughoutthe succession. The variable occurrence of this phase throughout thesuccession emphasizes the need for high-resolution thermal demag-netization protocols when studying volcanic and sedimentarylithologies that contain hematite.

Acknowledgments

The experimental work was possible through a visiting fellowshipto the Institute for Rock Magnetism (IRM) awarded to N.L.S.-H. TheIRM is funded by the Keck Foundation, the National ScienceFoundation and the University of Minnesota. Excellent support atthe IRM was provided by M. Jackson and J. Bowles. Fieldwork atMamainse Point was funded through a Sigma Xi Grant-In-Aidawarded to N.L.S.-H. and with support from Princeton University.Fieldwork at Mamainse Point was conducted in collaboration with B.Weiss, D. Jones, C. Rose and N. Eichelberger. We thank C. Bayne of BayNiche Conservancy for land access. Some of the paleomagnetic datapresented herein was obtained at the Yale Paleomagnetic Laboratorywhich is run by D. Evans and funded by NSF and the David and LucilePackard Foundation. The electron microscopy was funded by a SigmaXi Grant-In-Aid to N.L.S.-H. and benefitted from the technicalexpertise of N. Seaton. This work has benefitted from discussionswith S. Banerjee, J. Bowles, S. Bowring, M. Brown, S. Burgess, D. Evans,M. Jackson, D. Kent, T. Kilian, J. Kirschvink, R. Mitchell, B. Moskowitz,T. Raub, J. Tarduno and B.Weiss. We thank H. Halls and an anonymousEPSL reviewer for comments that improved themanuscript. This workis contribution number 1006 of the Institute for Rock Magnetism.

Appendix A. Supplementary data

Supplementary data to this article can be found online atdoi:10.1016/j.epsl.2011.02.053.

References

Aragon, R., McCallister, R.H., Harrison, H.R., 1984. Cation diffusion in titanomagnetites.Contrib. Mineralog. Petrol. 85, 174–185.

Banerjee, S., 1963. Magnetic properties of rocks andminerals. Ph.D. thesis, University ofCambridge.

Barbosa, P., Lagoerio, L., 2010. Crystallographic texture of the magnetite–hematitetransformation: evidence for topotactic relationships in natural samples fromQuadrilátero Ferrífero, Brazil. Am. Mineralog. 95, 118–125.

Bernal, J.D., Dasgupta, D.R., Mackay, A.L., 1957. Oriented transformations in iron oxidesand hydroxides. Nature 180, 645–647.

Brand, R.A., 1987. Improving the validity of hyperfine field distributions from magneticalloys, part I: unpolarized source. Nucl. Instrum. Meth. Phys. Res. B 28, 398–416.

Buchan, K., Ernst, R., Hamilton, M., Mertanen, S., Pesonen, L., Elming, S., 2001. Rodinia:the evidence from integrated palaeomagnetism and U-Pb geochronology. Precam-brian Res. 110, 9–32.

Buddington, A.F., Lindsley, D.H., 1964. Iron-titanium oxide minerals and syntheticequivalents. J. Petrol. 5, 310–357.

Cox, A., 1969. Geomagnetic reversals. Science 163, 237–245.Davis, D., Green, J., 1997. Geochronology of the North American Midcontinent rift in

western Lake Superior and implications for its geodynamic evolution. Can. J. EarthSci. 34, 476–488.

Davis, B.L., Rapp, G., Walawender, M.J., 1968. Fabric and structural characteristics of themartitization process. Am. J. Sci. 266, 482–496.

Doubrovine, P.V., Tarduno, J.A., 2004. Self-reversed magnetization carried by titano-maghemite in oceanic basalts. Earth Planet. Sci. Lett. 222, 959–969.

Dubois, P., 1955. Paleomagnetic measurements of the Keweenawan. Nature 176,506–507.

Dubois, P., 1962. Paleomagnetism and correlation of Keweenawan rocks. Bull. Geol.Surv. Can. 71, 1–75.

Dunlop, D., Kletetschka, G., 2001. Multidomain hematite: a source of planetarymagnetic anomalies? Geophys. Res. Lett. 28, 3345–3348.

Dunlop, D., Özdemir, Ö., 1997. Rock Magnetism: Fundamentals and Frontiers.Cambridge University Press, Cambridge, UK.

Evans, D., 2009. The palaeomagnetically viable, long-lived and all-inclusive Rodiniasupercontinent reconstruction. In: Murphy, J., Keppie, J., Hynes, A. (Eds.), AncientOrogens and Modern Analogues: Geological Society of London Special Publication,Vol. 327, pp. 371–404.

Evans, M., Wayman, M., 1974. An investigation of the role of ultra-fine titanomagnetiteintergrowths in paleomagnetism. R. Astron. Soc. Geophys. J. 36, 1–10.

Garcia de Oliveira, M.T., L., F. M. L., Indio da Costa Jr., M., Meunier, A., 2002. Thetitanomagnetite to titanomaghemite conversion in a weathered basalt profile fromsouthern Paraná Basin, Brazil. Clays Clay Miner. 50, 478–493.

Graham, J., 1953. Changes of ferromagnetic minerals and their bearing on magneticproperties of rocks. J. Geophys. Res. 58, 243–260.

Haag, M., Heller, F., Allenspach, R., Roche, K., 1990. Self-reversal of natural remanentmagnetization in andesitic pumice. Phys. Earth Planet. Inter. 65, 104–108.

Haggerty, S.E., 1991. Oxide textures: a mini-atlas. Rev. Mineral. Geochem. 25, 129–219.

184 N.L. Swanson-Hysell et al. / Earth and Planetary Science Letters 305 (2011) 171–18429

Haggerty, S.E., Baker, I., 1967. The alteration of olivine in basaltic and associated lavas.Contrib. Mineralog. Petrol. 16, 233–257.

Halls, H., Pesonen, L., 1982. Paleomagnetism of Keweenawan rocks. Geol. Soc. Am.Mem.156, 173–201.

Hargraves, R.B., Banerjee, S.K., 1973. Theory and nature of magnetism in rocks. Annu.Rev. Earth Planet. Sci. 1, 269–296.

Hedley, I., 1968. Chemical remanent magnetization of the FeOOH, Fe2O3 system. Phys.Earth Planet. Inter. 1, 103–121.

Jacob, J., Abdul Khadar, M., 2010. VSM and Mössbauer study of nanostructuredhematite. J. Magn. Magn. Mater. 322, 614–621.

Johnson, H.P., Merrill, R.T., 1974. Low-temperature oxidation of a single-domainmagnetite. J. Geophys. Res. 79, 5533–5534.

Kletetschka, G., Wasilewski, P.J., 2002. Grain size limit for SD hematite. Phys. EarthPlanet. Inter. 129, 173–179.

Klewin, K.W., Berg, J.H., 1991. Petrology of the Keweenawan Mamainse Point lavas,Ontario: petrogenesis and continental rift evolution. J. Geophys. Res. 96, 457–474.

Krása, D., Shcherbakov, V., Kunzmann, T., Petersen, N., 2005. Self-reversal of remanentmagnetization in basalts due to partially oxidized titanomagnetites. Geophys. J. Int.162, 115–136.

Lawson, C.A., Nord, G.L., Champion, D.E., 1987. Fe–Ti oxide mineralogy and the origin ofnormal and reverse remanent magnetization in dacitic pumice blocks from Mt.Shasta, California. Phys. Earth Planet. Inter. 46, 270–288.

Manson, M.L., Halls, H.C., 1994. Post-Keweenawan compressional faults in the easternLake Superior region and their tectonic significance. Can. J. Earth Sci. 31, 640–651.

Massey, N.W.D., 1983. Magma genesis in a late Proterozoic proto-oceanic rift: REE andother trace-element data from the Keweenawan Mamainse Point Formation,Ontario, Canada. Precambrian Res. 21, 81–100.

McClelland, E., 1987. Self-reversal of chemical remanent magnetization: a palaeomag-netic example. Geophys. J. Int. 90, 615–625.

McClelland, E., Goss, C., 1993. Self-reversal of chemical remanent magnetization on thetransformation of maghemite to haematite. Geophys. J. Int. 112, 517–532.

Merrill, R.T., Grommé, C.S., 1969. Nonreproducible self-reversal of magnetization indiorite. J. Geophys. Res. 74, 2014–2024.

Mücke, A., Cabral, A., 2005. Redox and nonredox reactions of magnetite and hematite inrocks. Chem. Erde 65, 271–278.

Murad, E., Cashion, J., 2003. Mössbauer Spectroscopy of Environmental Materials andTheir Industrial Utilization. Springer.

Nagata, T., 1953. Self-reversal of thermo-remanent magnetization of igneous rocks.Nature 172, 850–852.

Néel, L., 1949. Théorie due traînage magnétique des ferromagnétiques en grains finsavec applications aux terres cuites. Ann. Géophys. 5, 99–136.

Néel, L., 1951. L'inversion de l'animantation permanente des roches. Ann. Geophys. 7,90–102.

Ohmoto, H., 2003. Nonredox transformations of magnetite–hematite in hydrothermalsystems. Econ. Geol. 98, 157–161.

Otake, T., Wesolowski, D., Anovitz, L., Allard, L., Ohmoto, H., 2007. Experimentalevidence for non-redox transformations between magnetite and hematite underH2-rich hydrothermal conditions. Earth Planet. Sci. Lett. 257, 60–70.

Özdemir, Ö., Dunlop, D.J., 1985. An experimental study of chemical remanentmagnetizations of synthetic monodomain titanomaghemites with initial thermo-remanent magnetizations. J. Geophys. Res. 90, 11513–11523.

Özdemir, Ö., Dunlop, D., 2002. Thermoremanence and stable memory of single-domainhematites. Geophys. Res. Lett. 29. doi:10.1029/2002GL015597.

Özdemir, Ö., Dunlop, D.J., 2005. Thermoremanent magnetization of multidomainhematite. J. Geophys. Res. 110. doi:10.1029/2005JB003820.

Özdemir, Ö., Dunlop, D.J., Moskowitz, B.M., 1993. The effect of oxidation on the Verweytransition in magnetite. Geophys. Res. Lett. 20, 1671–1674.

Özdemir, Ö., Dunlop, D.J., Berquo, T.S., 2008. Morin transition in hematite: sizedependence and thermal hysteresis. Geochem. Geophys. Geosyst. 9. doi:10.1029/2008GC002110.

Pesonen, L.J., Halls, H., 1983. Geomagnetic field intensity and reversal asymmetry in latePrecambrian Keweenawan rocks. Geophys. J. R. Astron. Soc. 73, 241–270.

Pesonen, L., Nevanlinna, H., 1981. Late Precambrian Keweenawan asymmetricreversals. Nature 294, 436–439.

Pullaiah, G., Irving, E., Buchan, K.L., Dunlop, D.J., 1975. Magnetization changes caused byburial and uplift. Earth Planet. Sci. Lett. 28, 133–143.

Raming, T.P., Winnubst, A.J.A., van Kats, C.M., Philipse, A.P., 2002. The synthesis andmagnetic properties of nanosized hematite (α-Fe2O3) particles. J. Colloid InterfaceSci. 249, 346–350.

Robinson, M.A., McClelland, E.A., 1987. Palaeomagnetism of the Torridonian of Rhum,Scotland: evidence for limited uplift of the central intrusive complex. Earth Planet.Sci. Lett. 85, 473–487.

Rösler,W., Metzler, W., Appel, E., 1997. Neogenemagnetic polarity stratigraphy of somefluviatile Siwalik sections, Nepal. Geophys. J. Int. 130, 89–111.

Roy, J.L., Park, J.K., 1972. Red beds: DRM or CRM? Earth Planet. Sci. Lett. 17, 211–216.Schmidt, P.W., Williams, G.E., 1995. The Neoproterozoic climatic paradox: equatorial

palaeolatitude for Marinoan glaciation near sea level in South Australia. EarthPlanet. Sci. Lett. 134, 107–124.

Smith, R., 1974. The production of spilitic lithologies by burial metamorphism of floodbasalts from the Canadian Keweenawan, Lake Superior. In: Amstutuz, G. (Ed.),Spilites and Spilitic Rocks. Springer-Verlag, pp. 403–416.

Soubrand-Colin, M., Horen, H., Courtin-Nomade, A., 2009. Mineralogical and magneticcharacterisation of iron titanium oxides in soils developed on two various basalticrocks under temperate climate. Geoderma 149, 27–32.

Swanson-Hysell, N.L., Maloof, A.C., Weiss, B.P., Evans, D.A.D., 2009. No asymmetry ingeomagnetic reversals recorded by 1.1-billion-year-old Keweenawan basalts. Nat.Geosci. 2, 713–717.

Tauxe, L., 2010. Essentials of Paleomagnetism. University of California Press.Tauxe, L., Kodama, K., 2009. Paleosecular variation models for ancient times: clues from

keweenawan lava flows. Phys. Earth Planet. Inter. 177, 31–45.Turner, P., 1981. Relationship betweenmagnetic components and diagenetic features in

reddened Triassic alluvium (St Bees Sandstone, Cumbria, UK). Geophys. J. Int. 67,395–413.

Turner, P., Clemmey, H., Flint, S., 1984. Palaeomagnetic studies of a Cretaceous molassesequence in the central Andes (Coloso Formation, Northern Chile). J. Geol. Soc. 141,869–876.

Vincenz, S.A., 1968. Phenomenon of partial self-reversal in Keweenawan rocks 1.Magnetization of Portage Lake lavas. J. Geophys. Res. 73, 2729–2752.

Vine, F.J., Matthews, D.H., 1963. Magnetic anomalies over oceanic ridges. Nature 199,947–949.

Weil, A., Van der Voo, R., Mac Niocaill, C., Meert, J., 1998. The Proterozoic supercontinentRodinia: paleomagnetically derived reconstructions for 1100 to 800 Ma. EarthPlanet. Sci. Lett. 154, 13–24.

Zijderveld, J., 1967. A.C. demagnetization of rocks. In: Collinson, D., Creer, K., Runcorn, S.(Eds.), Methods in Paleomagnetism. Elsevier, New York, pp. 256–286.

Zysler, R.D., Fiorani, D., Testa, A.M., Suber, L., Agostinelli, E., Godinho, M., 2003. Sizedependence of the spin-flop transition in hematite nanoparticles. Phys. Rev. B 68,212408.