Introduction

Plate tectonics is presently the major mechanism for heat dissipation and material circulation between Earth’s interior and surface. How far back in time plate tectonics has operated, and how it may have initiated, are major questions in the Earth sciences1,2,3,4,5. Existing estimates for the onset of plate tectonics range from the Hadean to Neoproterozoic (Supplementary Fig. S1), based on diverse constraints from geology1,4,6,7, geochemistry8,9,10,11, paleomagnetism4,12,13,14, and geodynamic models15,16, many of which are complicated by terminological ambiguities and imprecision of empirical constraints3.

In principle, one of the most robust arguments for Archean plate tectonics can come from paleomagnetism. However, the paleomagnetic argument for plate tectonics is weakened if relative motion is not demonstrated. For example, 3.45–3.18 Ga paleomagnetic directions of volcanic rocks from Pilbara craton have been interpreted to represent lateral motions by plate-tectonic process13,14,17; however, data from a single craton can also be explained by true polar wander (TPW)18. Documenting significant relative motion between two or more cratons, which is necessary to resolve such ambiguities, requires differing paleomagnetic apparent polar wander paths between the blocks over the same time interval12. Robust paths are established for Pilbara and Kaapvaal cratons from 2.8 to 2.7 Ga, but the data permit their common motion as a single Archean block19. Until now, the oldest robust documentation of significant differential motions between two or more cratons is constrained to a broad age range between 2.71 and 2.17 Ga based on their two pairs of paleomagnetic pole ages20,21. In this study, we report a new U-Pb age and paleomagnetic pole from the 2.48 Ga Elbow Creek mafic dikes in the Wyoming craton. Aided by the ca. 2.48 Ga poles from Superior craton and 2.71 Ga poles from both blocks, the data provide quantitative kinematic constraints for the assembly of Wyoming and Superior cratons between 2.71 and 2.48 Ga, within the Neoarchean portion of the previously broadly constrained time interval. Our results document the oldest substantial relative motion between crustal blocks, providing a potential minimum age for the initiation of plate tectonic mobilism.

Wyoming Craton has a rich Archean history extending back to 3.5 Ga, and its northern and central regions were stabilized by 2.71 Ga22,23,24. The southern accreted terrane, which consists of various fragments of arcs and micro-continents, was sutured to the craton along the Oregon Trail structural belt at ca. 2.65–2.63 Ga23,25,26. After stabilization of the Wyoming craton, multiple generations of Neoarchean-Paleoproterozoic mafic intrusions were emplaced into the basement, including the ca. 2.71 Ga Stillwater Complex and numerous swarms of mafic dikes21,27,28,29. The Wyoming craton arrived at its current position relative to Superior during the amalgamation of Laurentia at ca. 1.8–1.7 Ga28,30. Laramide orogenic uplift during late Cretaceous to Cenozoic time denuded and exposed the basement rocks23.

The ca. 2.71 Ga Stillwater Complex is exposed within the Beartooth Mountains in southern Montana29. After the emplacement of the Stillwater Complex, mafic dikes intruded into the Complex and its basement footwall (Fig. 1). The dikes are NW-trending or NE-trending diabase or gabbro and are usually 5–10 m wide; their petrography and geochemistry at the type locality of Elbow Creek, near the Stillwater Mine, is described by Rogers et al.31. The complex and the dikes are overlain by Paleozoic and Mesozoic sedimentary rocks, and the entire succession was structurally imbricated with northerly vergence during the Cretaceous Laramide orogeny32. Despite steep dips of the Stillwater layered series, the cross-cutting dikes are generally vertical or moderately south-dipping, and shallow dips of the nonconformably overlying Paleozoic strata indicate only minor Laramide tilting of the central zone of uplift where the majority of paleomagnetic sites are located33. In this study, samples were collected from 26 dikes. Conducting baked-contact tests to constrain the age of the magnetization, we also collected samples from their host rocks including anorthosites from the ca. 2.71 Ga Stillwater Complex and its ca. 2.8 Ga footwall Mouat quartz monzonite (Fig. 1). Block samples of the dikes were collected for zircon and baddeleyite geochronology.

Fig. 1: Location of the sampling dikes.
figure 1

A Schematic map showing the location of Wyoming in Laurentia. B Simplified geological maps of the Stillwater regions and the spatial distribution of the studied dikes.

Full size image

Results

Isotope dilution thermal ionization mass spectrometry (ID-TIMS) on four baddeleyite fractions of one Elbow Creek dike near the Stillwater Mine (J20S24, Fig. 1) produced U-Pb data that are 1–2% discordant. Linear regression yields upper and lower intercept dates of 2480 ± 4 Ma and 717 ± 500 Ma. The 2480 ± 4 Ma date is interpreted as the crystallization age of the dike (Fig. 2).

Fig. 2: Geochronological and paleomagnetic results for the ca.2480 Ma Elbow Creek dikes.
figure 2

A Concordia plot of U–Pb isotope dilution thermal ionization mass spectrometer (ID-TIMS) baddeleyite dating results for one Elbow Creek dike J20S24 (Supplementary Table S7). B Equal-area projection of the mean high-temparature component (HTC) direction for ~2480 Ma Elbow Creek mafic dikes in the Stillwater region; filled symbols represent directions plotted onto lower hemisphere.

Full size image

Rock magnetic results including the K-T curves, hysteresis loops, isothermal remanent magnetization (IRM) and back-field demagnetization of IRM show that the main magnetic carriers are single-domain or pseudo-single domain magnetite (Supplementary Figs. S2 and S3).

Stepwise thermal demagnetization was conducted on all paleomagnetic samples. A low-temperature component (LTC) was removed before 500 °C. LTCs are either similar to the present-day magnetic field direction or scattered (Supplementary Fig. S4). A middle-temperature component (MTC) was isolated between 500 and 560 °C from three dikes and their adjacent host rocks (Supplementary Fig. S4). After the removal of LTC and MTC, the high-temperature components (HTCs) were isolated. Thirteen dikes yielded an HTC (hereafter referred to as component A), which is directed south-southeast and down with a moderate inclination (Fig. 2; Supplementary Fig. S5). Less abundant dike sites with other stable HTC directions are listed in Supplementary Table S1 and described in the Supplementary Section A3.2. They are interpreted as intrusions of various Proterozoic ages, including five dikes with north or south shallow remanence (component B), one dike with southeast and moderate upward direction (component C), and one dike with a shallow east remanence (component D).

Component A is clearly distinct from the HTCs of host rocks in the Stillwater Complex, which are directed south-southeast and upward34; and a suite of baked-contact tests, with one particularly definitive result, demonstrate that component A can be interpreted as primary in origin (Supplementary Section A2.3; Supplementary Figs. S7–S9). Additionally, one positive regional stratigraphic consistency test demonstrates that component C could also be primary in origin, which would further rule out widespread regional remagnetization across the study area (Supplementary Section A2.3; Supplementary Fig. S9). After applying a uniform Laramide tilt correction using Paleozoic strata, which are locally preserved atop the central outcrop region of the Stillwater Complex with variable but low northerly dips32,33, the mean direction of the thirteen dikes with Component A is: D = 145.8°, I = 59.6°, k = 31.5, α95 = 7.5°, and the mean pole for these thirteen dikes is: 2.0 °N, 275.3 °E (A95 = 10.2°). A paleosecular variation test shows that our paleomagnetic results likely provide a sufficient time average of geomagnetic variations, therefore the pole can be used for robust paleogeographic comparisons (Supplementary Section A2.5). Our new pole fulfills six of seven quality criteria for paleomagnetic poles35, lacking only dual polarity of remanence, and can be treated as a key paleomagnetic pole for the Wyoming craton at 2480 Ma.

Assembly of Wyoming and Superior during ca. 2.7–2.5 Ga

Striking stratigraphic and sedimentological similarities between the Huronian and the Snowy Pass Supergroups led Roscoe and Card36 to propose that Wyoming and Superior cratons were a conjugate rift pair, connected along their present southern margins between ca. 2.4 and 2.1 Ga. The matching of large igneous province temporal “barcodes” between the two cratons also suggests their proximal relationship throughout the early Paleoproterozoic (Supplementary Fig. S17)37. The connection finds additional support from the two cratons’ ca. 2.17 Ga paleomagnetic poles21,38, as well as two ca. 2.22 Ga virtual geomagnetic poles (VGPs) from Wyoming and the ca. 2.22 Ga poles from Superior20,39,40. The present study extends the connection back to ca. 2.5 Ga based on our new ca. 2.48 Ga Elbow Creek pole from Wyoming craton and the poles from the ca. 2.50 Ga Ptarmigan-Mistassini dikes and the 2.47–2.45 Ga Matachewan dikes from Superior (Fig. 3B)38. In the common reference frame of the conjoined assembly, however, the ca. 2.71 Ga poles from the Stillwater Complex of the Wyoming craton and the rocks from the Abitibi subprovince of Superior are substantially separated from each other (Fig. 3B); this requires cratonic assembly with relative motions on the order of thousands of km (minimum ~30° of arc distance; Fig. 3C, D, Supplementary Fig. S16A) during ca. 2.71–2.48 Ga. Within that time interval, the ca. 2.65–2.63 Ga Oregon Trail structural belt is the most likely manifestation of their assembly23,25,26,36. Using the Oregon Trail age estimate for the amalgamation, the apparent polar wander path convergence of ~3000 km in 70 million years indicates an average rate of about 4–5 cm/yr. These calculations are intentionally approximate due to the inherent paleolongitudinal uncertainties of paleomagnetic reconstructions, but they are consistent with present-day plate motion rates14.

Fig. 3: Lithospheric rigidity, relative motion of cratons, and proxy evidence for subduction suggest that plate mobilism in operation prior to 2.5 Ga.
figure 3

A The rigidity of the lithosphere indicated by a time-space plot of major felsic and mafic/ultramafic igneous activity, deformational events, intrusion of mafic dikes for the Pilbara, Kaapvaal, Superior, Wyoming, Congo-SF, Karelia, Yilgarn, Slave, Zimbabwe, Dhawar and North China cratons for the period 3.5–2.4 Ga. B Comparison of the ~2710–2160 Ma poles of the Wyoming and Superior (in present East Superior coordinates). Wyoming and its poles are rotated to East Superior using Euler rotation (48.0°, 265°, 125°). C, D Paleogeographic reconstructions of Wyoming and Superior at ~2710 Ma and ~2500 Ma (see the Supplemental Material) suggest they had relative motion during 2.7–2.5 Ga. Poles are listed in Supplementary Table S8, and the Euler rotation parameters are listed in Supplementary Table S9. E Proxy evidence for geothermal gradients through Earth history, including the appearance of eclogite-facies rocks and a step-change of metamorphic thermobaric ratios in late Archean time6. Blue, yellow and red dots represents low, intermediate and high temperature/pressure metamorphism.

Full size image

Paleomagnetism constrains the oldest recognized relative cratonic motion

Our present study is the first to document a paleomagnetically robust Archean record of assembly between cratons in a geologically compelling reconstruction. The ca. 2.71–2.17 Ga apparent polar wander paths for Wyoming and Superior previously suggested they experienced relative motion prior to ca. 2.17 Ga, but without a precise temporal constraint20. Before this study, the minimum age of indisputable relative motion between any two cratons was constrained to Rhyacian time: Buchan et al.41 and Mitchell et al.42 documented relative motion between Slave and Superior cratons at ca. 2.2–2.0 Ga. Although Condie et al43. suggested changes of relative positions between Superior, Kaapvaal and southern Indian cratons through ca. 2.4–2.1 Ga, the method they used is not definitive (Supplementary Fig. S18); such methodological ambiguity also compromises an earlier suggestion of Neoarchean relative cratonic motions4.

Various models of global-scale Neoarchean-Paleoproterozoic paleogeographic reconstruction are similarly ambiguous. Liu et al.44 presented two paleogeographic options for the interval 2.62–2.41 Ga: one with substantial relative motion between two craton groups, and the other with relative rigidity yet a sizable gap between the groups. Salminen et al.45 concluded that separately drifting cratons were more likely than a supercontinent through this time interval, but relative motions could not be conclusively demonstrated.

Late Archean plate mobilism

Modern-style plate tectonics has at least three important features: large lateral mobility, rigid lithosphere, and temporally continuous seafloor spreading balanced by subduction1,5. The present study documents Neoarchean relative lateral motion between cratons on the order of thousands of km, demonstrably between 2.71 and 2.48 Ga, and likely occurring prior to 2.63 Ga. Based on the preserved global rock record (Fig. 3A), rigidity of lithospheric plates developed between 3.2 and 2.5 Ga5, or perhaps even earlier3. In particular, changes in the bulk composition of continental crust10, the emplacement of dike swarms5, and replacement of dome and keel structures by linear thrust-imbricated belts46 all suggest that many lithospheric plates were rigid since at least late Archean.

Whether modern-style seafloor spreading and subduction existed in Archean time is more difficult to demonstrate. Oceanic lithosphere of such antiquity is not preserved in situ, thus proxy records must be sought within continental lithosphere that would have been overriding or advancing toward the convergence zones. Large-scale lithospheric mobility can be accommodated either continuously or episodically, depending on the planetary mode of convection3. Episodic mantle overturning has been proposed for Neoarchean time11,15, but such models predict global-scale tectonic quiescence on the order of ca. 100 million years between the putative overturn events that are contradicted by a continuity of aggregate records for convergence through the 2.8–2.5 Ga interval (Fig. 3A).

Geodynamic modeling, isotopic data from mantle xenoliths, kimberlitic diamonds, and inherited or detrital zircons indicate that materials that formed near or close to Earth’s surface were recycled into the sub-lithospheric mantle prior to 2.5 Ga (Supplementary section A1)47,48,49,50. Robust petrologic and metamorphic lines of evidence for subduction (Fig. 3E) are high-P/low-T metamorphic rocks like blueschists, jadeitites, eclogites, and lawsonite-bearing rocks6, but those proxies generally appear in the geologic record only during various stages of Proterozoic time51,52, with rare and debated instances in the Neoarchean7,53. Archean lithospheric recycling may have differed in style from modern subduction11,54,55. But in the context of these broader datasets, the 2.7–2.5 Ga paleomagnetic data from Wyoming and Superior cratons now provide a quantitative kinematic demonstration of Archean plate mobilism.

Methods

U-Pb ID-TIMS geochronology

Approximately 200 g of each sample was processed at the department of Geology, Lund University, following the procedure described by Söderlund and Johansson56. Approximately 40 brown baddeleyite grains of excellent quality were recovered from one sample from dike J20S24 near the Stillwater Mine (45°23'36.91“N, 109°52'58.26“W).

The best-quality baddeleyite grains were transferred to Teflon© dissolution capsules and rinsed repeatedly in ultrapure 7 M HNO3 and H2O in order to dilute the Pb blank. An ultrapure HF-HNO3 mixture was added to the dissolution capsules together with a 205Pb-233-236U tracer solution. Capsules were then placed in an oven at 190 °C for 3 days until the grains were fully dissolved. After dissolution, the samples were dried down on a hotplate and re-dissolved in ultrapure 6.2 M HCl with the addition of ultrapure 0.25 M H3PO4 before being dried down again on a hotplate. The samples were dissolved in 2 μl of silica gel and loaded on outgassed Re filaments. The U and Pb isotopic compositions were measured at the Department of Geosciences at the Swedish Museum of Natural History in Stockholm using a Thermo Finnigan Triton Mass Spectrometer equipped with Faraday cups and a Secondary Electron Multiplier (SEM). The intensities of 208Pb, 207Pb, 206Pb and 205Pb were measured in static mode using the Faraday detectors. An ETP SEM equipped with a RPQ filter was used to measure the intensity of 204Pb. The SEM-Faraday gain was controlled by measuring a ~ 5–10 mV signal alternating between Faraday and SEM detectors prior to each Pb analysis. The 233U, 236U and 238U isotopic measurements were made in dynamic (peak-switching) mode at filament temperatures greater than 1350 °C. Procedural blanks are typically 0.3 pg Pb and 0.03 pg U at the Swedish Museum of Natural History. Mass fractionation is 0.1% per mass unit for Pb, determined by replicate analyses of NBS standards SRM 981 and SRM 983. U fractionation was determined directly from the measured 233U/236U isotopic ratio. Initial Pb compositions were taken from the model common Pb evolution curve57. The decay constants used were 1.55125 × 10−11 for 238U and 9.8485 × 10−10 for 235U58, with the isotopic composition of U being 238U/235U = 137.8859. The listed uncertainties in Pb/U ratios were calculated by propagating the within-run error for measured isotopic ratios, with the uncertainties in fractionation (±0.04 for Pb, absolute uncertainties), Pb and U blank concentration (±50%), and Pb blank composition (2.0% for 206Pb/204Pb and 0.2% for 207Pb/204Pb). Analytical results were calculated and plotted using the Microsoft Excel Macro, Isoplot60.

Rock magnetic and paleomagnetic analyses

In this study, a portable drill was used in sample collection for most sites; oriented block samples were collected when drilling was not feasible. Orientation for the core samples was achieved using a magnetic compass with clinometer, usually supplemented by a solar compass. No significant difference was observed between the results using the two orientation devices. All the samples were cut into 1–3 cylindrical paleomagnetic specimens (2.5 cm in diameter, and approximately 1–1.5 cm in height) in the laboratory.

All samples were subjected to stepwise thermal demagnetization in 15–25 steps up to 580 or 590 °C in a nitrogen-gas environment using an ASC-Scientific TD-48 furnace in which internal residual fields is ~10 nT. Sister samples for sites J20S24-26 were also selected for alternating-field demagnetization using a Molspin tumbler AF demagnetizer. Measurements of the natural remanent magnetization (NRM) and all subsequent demagnetization steps were conducted on an automated 2G-Enterprises DC SQuID magnetometer. The demagnetizations and remanence measurements were conducted in a shielded chamber, which has an ambient field typically less than 300 nT. To characterize the properties of magnetic carriers in the samples, rock magnetic studies were conducted on representative samples. Thermomagnetic curves of bulk susceptibility versus temperature were achieved by measuring the bulk susceptibility from –192 to 700 °C followed by cooling back to room temperature in argon atmosphere. The measurements were conducted using an AGICO KLY-4S Kappabridge equipped with a high-temperature apparatus and a cryostat. All these experiments were conducted in the Paleomagnetic Laboratory at Yale University. In addition, hysteresis loops and backfield curves were obtained between –500 mT and 500 mT using a Princeton Measurement Corporation MicroMag 2900 Series alternating gradient magnetometer (AGM) at the Yale Archaeomagnetism Laboratory.

The paleomagnetic data were analyzed using the computer packages developed by Enkin61 and Cogné62. Directional data were fit using principal component analysis63, and the mean directions were calculated using Fisher statistics64. The paleogeographic reconstructions were made using the GPlates freeware package65.