Plate motion and a dipolar geomagnetic field at 3.25 Ga

Stepwise alternating field and thermal demagnetization up to 590 °C on all specimens revealed four components of ancient magnetization with unblocking temperatures between 75 and 580 °C (Fig. 2 A–C and SI Appendix, Figs. S2A and S3; see SI Appendix, Table S1 for component directions and statistics and SI Appendix, Appendix S1 for detailed interpretations). The first three thermal demagnetization components (“L2,” “M,” and “M2”) can be tied to resetting events at ∼0.54, ∼2.2–1.7, and ∼2.78 Ga, respectively, based on paleopole comparisons (10–12) and failing fold tests (L2 and M components only; SI Appendix, Table S1).

The final component, termed “H”, is present in 44 of 117 samples as an origin-trending, site-coherent magnetization unblocking above ∼500 °C in 9 out of 11 sites (Fig. 2 and SI Appendix, Table S1). Site-mean directions of this component converge following correction for local structural tilt and rotation, passing a fold test (13) and indicating that the H magnetization predates 2.93 Ga folding of the SVS (Fig. 3 and SI Appendix, Fig. S2B; see SI Appendix, Appendix S3 for details of regional structure). Additionally, H components from sites KUA5-9 (“Upper KUA”) are antipodal within 20° to those from all other sites, statistically consistent with antipodality. While the number of sites limits the statistical power of the reversal test [“indeterminate” in the classification of McFadden et al. (14); see SI Appendix, Appendix S3.2], the proximity to antipodal directions, similarity in thermal demagnetization behavior, and lithostratigraphic, petrographic, and geochronological context of the H component (see below) demonstrate that a 3.25-Ga reversal is the only probable explanation for the polarity groups, implying that the H magnetization was acquired shortly after eruption onto the seafloor.

Micromagnetic examination independently verifies these paleomagnetic field tests and identifies the mechanism of magnetization (SI Appendix, Figs. S7–11 and Appendix S4). Optical, Raman, and quantum diamond microscopy (QDM) demonstrate that major ferromagnetic grain populations—nearly all magnetites—are ubiquitously derived from replacement of primary minerals and precipitation of new minerals via hydration and Fe mobilization during multistage fluid flow. These reactions included recrystallization of primary titanomagnetite grains and dendrites (SI Appendix, Figs. S8 and S12 B and C), Fe exchange with magmatic Cr-spinels (SI Appendix, Fig. S12D), and hydration/mobile element exchange with primary ferromagnesian silicates (e.g., olivine, SI Appendix, Fig. S12E; orthopyroxene, SI Appendix, Fig. S12F; and volcanic glass). In void spaces (vesicles and fractures), magnetite is closely associated with infill phases, such as calcite after Ca-zeolites and anorthite, which occasionally preserve a complex prograde infill and replacement sequence (SI Appendix, Figs. S9–S12). Notably, magnetite grains associated with one such vesicular infill stage exhibit thermal unblocking behavior identical to that of the H component in the bulk sample, robustly indicating that these grains contribute to the H magnetization (Fig. 2C and SI Appendix, Fig. S10). Later stages in the hydrothermal alteration sequence precipitated many secondary minerals, including the titanites targeted for geochronology (see below) and carbonaceous matter that records peak maturation temperatures between 300 and 330 °C (SI Appendix, Fig. S7 K and L).

Titanite is a suitable mineral for U-Pb petrochronology, so titanite postdating the magnetite crystallization thus offers an opportunity to establish a minimum age for the H component. We performed in situ laser ablation U-Pb analyses of 120 titanite grains formed during late albitization reactions, 93 of which passed selection criteria, defining an array of U-Pb compositions (SI Appendix, Appendix S4.5 and Fig. S11). Excluding analyses affected by minor postcrystallization Pb loss, the remaining data define a lower-intercept age of 3.223 ± 0.023 Ga (2σ).

These analyses directly associate the H component remanence and its carrier minerals to early hydrothermal alteration bracketed by 3.275–3.249 Ga Kunagunarrina eruption and late alteration at 3.223 ± 0.023 Ga. Given the history of mineralization in the East Pilbara, the emplacement of nearby granitic intrusions of the Cleland Supersuite at 3.257–3.235 Ga are most likely responsible for this alteration (15, 16). This voluminous plutonism resulted in extensive metamorphism and subseafloor hydrothermal circulation, which notably formed major volcanic-hosted massive sulfide (VHMS) deposits dated to 3.265–3.235 Ga [Fig. 3 and SI Appendix, Fig. S1 and Appendix S4 (15–17)]. These deposits are hosted within “exhalite” horizons, representing the seafloor expressions of pulses of hydrothermal circulation beneath the paleo-seafloor and above the driving intrusions (18). These pulses have durations comparable to the cooling timescale of the intrusions that drive them, driving mineral alteration for up to thousands to hundreds of thousands of years (19, 20). This is long enough that data from a suite of geographically separated sites that were altered (and thus magnetized) at different times by the same protracted VHMS circulation event can be averaged together to sample paleosecular variation (PSV) and record the time-averaged field during alteration. This is supported by the low within-site but higher between-site scatter of our directional data, both within each polarity group of sites and in the compilation of all sites (SI Appendix, Appendix S3.3). Two stratigraphic horizons in our study area contain deposits of hydrothermal quartz typical of this style of hydrothermal activity, one just above our lowermost “R” polarity sites, and another overlying our uppermost “N” polarity sites (Fig. 1 and SI Appendix, Fig. S1). The latter is the 3.241–3.235 Ga “Marker Chert” (Figs. 1 and 3 and SI Appendix, Fig. S1), a regionally traceable seafloor exhalative chert horizon associated with intrusion of the Strelley Laccolith north of our study area (21, 22). Since each of these two hydrothermal events cap a lava package hosting a magnetization antipodal to that of the other package, these events, separated by up to ∼20 My given the timespan of nearby Cleland Supersuite plutonism, were likely each responsible for imparting one polarity of the reversing thermochemical remanent magnetization that we document.

In brief, the combination of a passing fold test, a probable reversal, magnetic microscopy, and titanite petrochronology shows that the H component is a TCRM acquired during 3.265–3.235 Ga sea-floor alteration shortly after lava emplacement (16, 17, 21). The H magnetization thus defines an ∼3.25-Ga paleopole from the East Pilbara (λ, ϕ = 65.6°S, 143.7°E; α₉₅ = 16.5°; n = 9; paleolatitude λ_p = 42.7 ± 13.1°), which we term “KUH-R” for “Kunagunarrina high-temperature – reversed” (Fig. 4 and SI Appendix, Table S1). Direct comparison to other pre-2.8 Ga poles from the Pilbara craton requires correction for an ∼2.93-Ga vertical-axis rotation of the SVS block relative to the rest of the East Pilbara (SI Appendix, Fig. S5 Table S1, and Appendix S2.2). This rotation is constrained to between 0° (no rotation) and 70° clockwise by the curvature of the fault bounding the SVS to the west (SI Appendix, Fig. S5). We consider the full permissible range of this rotation below, noting that it does not affect the relative positions of poles HEBh and KUH-R, since both were sampled from within the SVS.

Taken together, poles EBm, KUH-R, and HEBh resolve the East Pilbara APW path at ≤95-My intervals, substantially constraining the range of possible motions between 3.34 and 3.18 Ga and representing the oldest such record available. Due to reversal-state ambiguities between the poles and possible block rotation of the SVS shortly preceding folding at 2.93 Ga, several APW paths and corresponding motion reconstructions are possible. Below, we construct the APW path and reconstruction that minimizes the intervening motion of the East Pilbara (Fig. 4), noting that all other permissible cases imply faster motions, in most cases exceeding 1°/My. For detailed treatment of these less-likely scenarios, see SI Appendix, Appendix S2 and Fig. S6.

In our conservative, preferred scenario, pole EBm samples the opposing reversal state relative to HEBh and KUH-R. Since the positions of SVS poles KUH-R and HEBh relative to pole EBm also depend on how much structural block rotation (between 0° and 70° clockwise) the SVS area experienced, we prescribe the magnitude of this rotation (20° clockwise) that minimizes the distance (and therefore motion) between poles EBm and KUH-R (Fig. 4A). These conservative assumptions yield a two-stage reconstruction: first, from 3.34 to 3.25 Ga (∼95 My), the East Pilbara drifted latitudinally at an average rate of ${0.55}_{-0.16}^{+0.19}$ °/My with no resolvable vertical-axis rotation (Fig. 4B). Second, from 3.25 to 3.18 Ga, the East Pilbara rotated counterclockwise by ${0.55}_{-0.38}^{+0.46}$ °/My with no resolvable latitudinal motion (Fig. 4B). We emphasize that, while this reconstruction and its motion rates are not unique—since others are possible due to uncertain reversal states and structure-forming events—the rates are the slowest allowable given these uncertainties, meaning the East Pilbara experienced horizontal motions at least as rapid as those presented above.

Reconciling these ∼0.55°/My rates with permissible motions in a stagnant-lid regime is challenging, especially over a 95-My interval. Lithospheric net rotation has not exceeded 0.75°/My for the last 200 My and averages much less over tens of My timescales [Fig. 5 and SI Appendix, Appendix S5 (23)]. The net rotation speed limit of an Archean stagnant lid is expected to have been less than or comparable to modern rates due to less-efficient lithospheric coupling to mantle flow (SI Appendix, Appendix S5). However, achieving this speed limit would have simultaneously required a “modern-like” lithospheric structure, a very hot Archean mantle, and a contrived mantle flow pattern that would not have persisted for tens of My, making net rotation an unlikely candidate to drive the observed 0.55°/My motions (Fig. 5 and SI Appendix, Fig. S13 and Appendix S5). An alternative type of non-plate-tectonic surface motion called TPW requires a Euler pole on the paleoequator, 90° away from the paleopole, thus resulting in paleopoles spread along a great circle (24). The most likely reconstructions all suggest the East Pilbara rotated in place from 3.25 to 3.18 Ga, implying a midlatitude Euler pole. That said, it is technically possible to ascribe all observed motions to two successive TPW events, one during each of the intervals between poles. However, all permissible APW paths would require that the successive TPW rotations were about axes 48° to 90° apart in longitude. In contrast, previously documented examples of interpreted successive TPW events have all involved Euler poles separated by <25° of longitude due to persistence in the geoid shape (25). Therefore, even though TPW may have contributed to some of the motion of the East Pilbara, it is highly unlikely that TPW drove all observed motions.

On the other hand, all observed motions are fully consistent with mobile-lid tectonic processes. For instance, plate rotations often arise today within oroclines (26, 27) and on microplates captured between complex convergent boundaries and associated back-arc rifts (3). One such system contains the Woodlark and Manus Basins on the northeastern Australian plate margin, featuring 0.6°/My latitudinal motion and up to 9°/My rotations (3). These similarities may suggest a tectonically complex convergent margin setting for the East Pilbara around 3.2 Ga, consistent with the craton’s stratigraphy and previous assertions of rift (28) and arc (29) settings during this time, although other tectonic settings can host similar motions.

In summary, the East Pilbara’s paleomagnetic record demonstrates that, by ∼3.25 Ga, geodynamic processes were driving rapid (≥0.55°/My or ≥6.1 cm/y) lithospheric motions that were sustained for tens of My. Differential motion within a mobile lid is the only mechanism that remains compatible with this record without invoking exceptional circumstances. That said, our observations do not require that modern-like plate tectonics sensu stricto were operating by 3.3 Ga, as sluggish lid, episodically mobile lid, and a variety of transitional regimes may all produce differential surface motions (1). We therefore pose these observations as a challenge to nonuniformitarian geodynamic models of the early Earth, which should seek to explain motions of this rate and duration.

The apparent polarity reversal provides further insights into early Earth’s deep interior. The polarities depart from antipodality by 19.9° with 10.1° of inclination mismatch. However, these “asymmetries” are not resolvably greater than zero within uncertainty and thus remain compatible with a reversal of a geocentric axial dipole (GAD) paleofield. Further, we can test the degree to which departures from this dipolar paleofield are statistically compatible with our polarity observations. This test consists of two constraints on field geometry. First, because stronger nonreversing axial quadrupole ($G2$) and octupole ($G3$) field components relative to the axial dipole result in more reversal asymmetry, the observed inclination asymmetry between polarities can constrain the relative moment of these higher-order geomagnetic field components (SI Appendix, Fig. S4A). Second, the PSV dispersion of our paleomagnetic directions can be used to estimate the dipolarity of the geomagnetic field when coupled with the empirical dipolarity-dispersion relationship identified by Biggin et al. (30) (SI Appendix, Fig. S4C). Combining these two constraints permits the calculation of confidence regions in $G2$–$G3$ space that delineate statistically permissible deviations from a pure GAD field given our dataset (Fig. 6 and SI Appendix, Fig. S4D).

Reproducing the observed polarity inclinations exactly without including their uncertainties would require only a nonreversing axial octupole with 15% of the dipole moment, within the range of estimated field dipolarity since 2.7 Ga [Fig. 6 and SI Appendix, Fig. S4 and Appendix S1 (30–33)]. Additionally, the data strongly argue that the dipole moment exceeds that of all other field components (Fig. 6 and SI Appendix, Fig. S4), since a dominantly nondipolar field would fail to reproduce both the high degree of antipodality and low among-flow PSV dispersion of our observed paleofield directions. The lack of observed excursion directions among the 20 sites from poles HEBh and KUH-R also sets a 2σ confidence upper bound of 1-(0.05)^1/20 ∼14% on the fraction of time the paleofield spent in a transitional state, similar to geologically recent periods of relatively unstable yet still strongly dipolar fields (34).

This reversal, predating the oldest high-fidelity example by 480 My [SI Appendix, Appendix S2.1 (35)], constitutes the oldest direct test of the GAD field geometry. The 3.25-Ga geodynamo was dominantly a reversible yet directionally stable dipole, the hallmark of stable self-alignment to Earth’s rotation axis. This stands in contrast to thin-shell geodynamo mechanisms [e.g., a basal silicate magma ocean (9)] with high aspect ratios that would generate strongly nondipolar and unstable surface fields that do not undergo antipodal reversals, similar to those of the ice-giant planets (36). This scenario involving a nondipolar dynamo would require the following unlikely sequence of events to explain the reversal in our samples: the local paleofield happened to persist over many PSV timescales in one direction and then coincidentally changed to the antipodal direction and persisted there.

In contrast, a stable dipole is in agreement with models of the Archean geodynamo that generate a field throughout the volume of the core (8), as well as with previous observations suggesting a stably dipolar Archean field (30, 37, 38), reaffirming the basic principles of paleomagnetism-based paleogeography up to 3.25 Ga. Additionally, the solar wind standoff distance is several times greater for dipolar fields than for multipolar fields of similar strength (39), so dipole-dominant fields can suppress [or enhance (40)] atmospheric modification via nonthermal escape and better shield the surface from cosmic radiation. This suggests a stably dipolar geodynamo may have contributed to a stable and habitable surface environment for the nascent biosphere since at least 3.25 Ga.

More-detailed materials and methods are presented in the supplement in SI Appendix, Appendix S1). We extracted 117 oriented 2.5-cm cores of Kunagunarrina Formation lavas from 11 total sites (cooling units; 7–12 cores each) representing three localities (KUA, KUB, and KUT) in the southern SVS (Fig. 1 and SI Appendix, Fig. S1). We drilled and field oriented all core samples using magnetic and solar compasses. We performed stepwise thermal demagnetization on all core samples, measuring their magnetic moments using a 2G Enterprises DC-SQuID Superconducting Rock Magnetometer at the Harvard Paleomagnetics Laboratory. We subjected all samples to alternating field (AF) demagnetization up to 10 mT in steps of 1 mT followed by thermal demagnetization up to 590 °C in steps of 10–40 °C (e.g., Fig. 2 A–C). We also performed a hybrid thermal demagnetization up to 430–540 °C followed by AF demagnetization to 110 mT (SI Appendix, Fig. S3) on selected samples and merged this dataset with our thermal demagnetization data (SI Appendix, Appendix S1.3).

We quantified the directions of all magnetization components using principal-component analysis (41). To compute directions in tilt-corrected (bedding) coordinates, we first rotated directions and bedding attitudes from KUA 40° clockwise and KUB 52° clockwise about the vertical axis to correct for drag folding of the SVS hinge plane (SI Appendix, Fig. S5 C and D). We then rotated all directions and bedding attitudes to correct for the plunging axis of the SVS (plunging 41° toward 69° E of N) and then corrected all directions to restore local paleohorizontal based on lava flow bedding attitudes (Fig. 2E and SI Appendix, Fig. S5 B and C). Over the last tilt-correction step, we performed the paleomagnetic fold test of Tauxe and Watson (13) (SI Appendix, Fig. S2B). On fully tilt-corrected directions, we performed the reversal test of McFadden and McElhinny (14). When comparing with other paleopoles, we performed additional rotations about a vertical axis on data from the SVS (pole HEBh and pole KUH-R from the present study) to correct for ∼2.93 Ga 0–70° clockwise rotation of the SVS structural block (SI Appendix, Fig. S5 A and B; see SI Appendix, Appendix S2; ref. 22).

To identify the remanence-carrying magnetic grains, we observed oriented polished sections of several cores with the QDM (42) at the Harvard Paleomagnetics Laboratory, as in Brenner et al. (4) (SI Appendix, Figs. S7–S12). For one sample (KUB2), we also conducted an abbreviated demagnetization routine (NRM, 10 mT AF, and thermal cycles to 330 °C, 490 °C, 520 °C, and 560 °C) and imaged the sample throughout with the QDM. Following the methods of Volk et al. (43), changes in surface field patterns during demagnetization captured the unblocking of the phases hosting the imaged magnetizations (SI Appendix, Fig. S10).

We used a Horiba Scientific XploRA Plus Raman microscope in the Harvard Laboratory for Mineral Physics to identify mineral phases. Raman spectra from reduced carbonaceous matter (SI Appendix, Fig. S7 K and L) allowed for estimates of peak alteration temperature, following the methods of Kouketsu et al. (44). We also imaged selected samples with a JEOL JSM-7900F Schottky field emission scanning electron microscope.

We used the laser-ablation split-stream inductively coupled plasma mass spectrometry facility at the University of California, Santa Barbara to collect U-Pb petrochronology and trace-element abundance data from 120 titanites in situ on selected polished sections from samples KUB1 and KUB2 (SI Appendix, Fig. S11; see below SI Appendix, Appendix S4.5). The spot size of all analyses was 30 μm. All uncertainties reported herein are 2σ, and all reported ages are lower intercept ages anchored to the common Pb composition of ²⁰⁷Pb/²⁰⁶Pb = 1.13 expected for ∼3.2-Ga rocks (45). We filtered the analyses based on several criteria (SI Appendix, Appendix S1.7) to a compilation of 93 spots. To isolate analyses unaffected by minor postcrystallization Pb loss, we performed Gaussian deconvolution of their age distribution (46).

We thank Arthur Hickman and Martin Van Kranendonk for their insights on the rock units, preservation potential, and structural elements of the East Pilbara and Soanesville Syncline. We thank Rebecca Fischer for the use of the Raman microscope in the Harvard Laboratory for Mineral Physics. We thank Timothy Cavanaugh for imaging our samples with the SEM at the Harvard Center for Nanoscale Systems. We thank Franklin Wolfe and John Shaw for advice on the treatment of structural corrections. We thank Hairuo Fu for assistance with sampling. Finally, we acknowledge the Nyamal and Kariyarra peoples, Traditional Custodians of the land on which we conducted this study and of Kunanganaranga Pool on the Turner River, from which the Kunagunarrina Formation derives its name. This work was supported by grants from the NSF (EAR-1847042 and EAR-1723023).