2.7 Ga high-pressure granulites of the Teton Range: Record of Neoarchean continent collision and exhumation

The nature of tectonic processes affecting the early Earth is an enduring and controversial topic (Van Kranendonk, 2010; Hamilton, 2011; O’Neill et al., 2016; Roberts et al., 2015; Smart et al., 2016; Tang et al., 2016). One of the most significant obstacles to unraveling the early history of tectonic processes on Earth is the scarcity in both time and space of representative rock samples; the situation becomes more limiting as time before present increases. Given this fundamental fact, studies of Archean rocks produced by tectonic processes of that era and minimally affected by later tectonism and metamorphism are of primary interest in the study of these phenomena.

Because tonalite-trondhjemite-granodiorite (TTG) and granite-greenstone terrains are distinctive Archean assemblages, they have been the subject of considerable study. However, their tectonic interpretation remains unsettled, with some workers concluding that they may form in a static lid or plume-dominated regime whereas others call on plate tectonics to supply the water needed for their formation (i.e., Condie, 2016). In this study, we look instead at Archean high-pressure granulites as indicators of tectonic regime. In the Phanerozoic, granulite-facies crustal rocks that preserve a record of transport from the surface to depth along a clockwise P-T-t path are diagnostic of collisional plate tectonic processes that double the thickness of the crust. However, high-pressure granulites are uncommon in the Archean. A compilation by Brown and Johnson (2018) of more than 450 localities with well-determined metamorphic conditions and ages identifies very few occurrences of Archean rocks that record pressures of 10 kbar or greater (Table 1). Most are from relatively small outcrops and commonly they have been overprinted by subsequent metamorphic events, making their tectonic significance difficult to establish. All but two are younger than 3 Ga; in one of these two, the Barberton terrane, temperatures do not reach granulite facies.

In this contribution we evaluate the geological record of an Archean high-pressure granulite terrain in the Wyoming Province, North America. No Proterozoic or Phanerozoic metamorphism or deformation has affected these rocks, making the Teton Range a suitable place to identify Archean tectonic processes. Archean rocks of the northern Teton Range include metapelitic gneisses that record 2.70 Ga high-pressure granulite facies metamorphism followed by tectonic juxtaposition with gneisses of different protoliths and geologic histories, and then by intrusion of leucogranites. This evidence is best explained by doubling of crustal thickness through continent-continent collision followed by exhumation, decompression melting, and intrusion of 2.68 Ga leucogranitic sills.

Field and structural relations, petrology and geochemistry, Nd isotopic compositions, and U-Pb zircon geochronology allow us to describe the origin and evolution of this well-preserved assemblage. This terrain preserves evidence of both crustal thickening (based on high-pressure granulite facies metamorphic conditions) and exhumation as evidenced by later, amphibolite facies metamorphism. Zircon geochronology brackets the time for the whole cycle to occur. Although the terrain is only 50 × 15 km in size, it is an excellent place to study Archean tectonic processes because there are rocks that are sensitive recorders of metamorphic conditions and there are no significant younger tectonic overprints to complicate the interpretations.

The Teton Range, located in northwestern Wyoming, USA, exposes some of the westernmost outcrops of the Archean Wyoming Province (Fig. 1). The northern portion of the Teton Range is underlain by a series of variably deformed paragneisses and orthogneisses (Reed, 1973) that are intruded by dikes of undeformed 2547 ± 3 Ma peraluminous leucogranite known as the Mt. Owen Quartz Monzonite (Zartman and Reed, 1998). The metamorphism and deformation described in this contribution are thus required to be older than the 2547 Ma Mt. Owen Quartz Monzonite. The geologic map of the northern Teton Range (Fig. 2) is derived from the map of Reed (1973). Reed recognized three gneiss units: the Layered Gneiss, the leucogranitic Webb Canyon Gneiss, and Augen Gneiss. A portion of the Layered Gneiss of Reed consists of mafic gneiss and pelitic gneiss, both of which show evidence of high-pressure granulite metamorphism. Because these rocks are so distinctive, we separate them out as the Moose Basin gneiss. (We do not capitalize “gneiss” here because this unit has not been formally recognized on U.S. Geological Survey maps [Frost et al., 2006]).

A characteristic feature of the Moose Basin gneiss is large kyanite blades in the pelitic gneiss, which are commonly more than a cm in length (Fig. 3A). Structurally beneath the pelitic gneiss is a mafic gneiss that locally retains granulite assemblages (Fig. 3B). In a few places, the mafic gneiss is cut by garnet-bearing leucosomes (Fig. 3C). The areal extent of the Moose Basin gneiss is reasonably well constrained. Kyanite is found as far south as where upper Bitch Creek crosses the Teton gneisses near the southern margin of Figure 2. Highly retrograded pelitic gneisses occur along the Teton divide at the head of Leigh Creek, ∼5 km south of the southern margin of Figure 2. These occurrences also may be part of the Moose Basin gneiss. Isolated, highly retrograded pelites occur in Osprey Canyon, on the eastern slope of the range (sample 07T32). The eastern margin of the Moose Basin gneiss is a complex contact zone ∼100 m wide consisting of interlayered amphibolite and Webb Canyon Gneiss. Following Reed (1973), we include this contact zone within the Moose Basin gneiss on Figure 2.

The Layered Gneiss consists of pelitic and psammitic paragneisses that are interlayered with quartzofeldspathic orthogneiss, amphibolite, metagabbro, and metaperidotite (Reed, 1963; Miller et al., 1986); pelitic gneisses are rare. Locally the paragneiss is migmatitic (Fig. 3D).

Intrusive into the Moose Basin and Layered gneisses is foliated leucogranitic Webb Canyon Gneiss of Reed (1973). On the basis of major element geochemistry, Frost et al. (2016) subdivided this unit into the Webb Canyon Gneiss and the Bitch Creek gneiss. The Webb Canyon Gneiss is the major leucogranite gneiss in the area. It forms the large tabular exposures shown on Fig. 2. In contrast, the Bitch Creek gneiss occurs in small sills, dikes, and plutons that cannot be delineated on Figure 2.

Reed (1973) mapped Augen Gneiss near the western limits of Precambrian outcrop. Because Miller et al. (1986) and Frost et al. (2016) have interpreted this unit as a porphyritic portion of the Webb Canyon Gneiss, we have shown it as part of the Webb Canyon Gneiss on Figure 2. The major ferromagnesian mineral is biotite with accessory hornblende. It is moderately well-foliated and lineated, with little or no compositional layering.

Titanite-bearing amphibolites with well-developed fabric occur as generally concordant but locally cross-cutting bodies within both the Moose Basin gneiss and the Layered Gneiss. These rocks are described in more detail below under “Thermobarometry.”

Structural measurements made in the field were plotted and interpreted using Stereonet 9 software described by Cardozo and Allmendinger (2013). A compilation of the structural measurements is available as Table S4 (^{footnote 1}).

U-Pb zircon geochronology was undertaken using the sensitive high-resolution ion microprobe with reverse geometry (SHRIMP-RG) at Stanford University. Zircon grains were mounted in epoxy. After polishing, cathodoluminescence (CL) scanning electron microscope (SEM) images were taken for all zircon grains. Isotopic ratios and U, Th, and Pb concentrations were standardized against VP-10 zircons. The data were reduced following the methods of Williams (1998) and using the SQUID Excel macro of Ludwig (2000). Uncertainties given for individual analyses (ratios and ages) are at the 1σ level, with correction for common lead made using the ²⁰⁴Pb lead and the model Pb evolution curve of Stacey and Kramers (1975). Concordia plots and linear discordia regression fits were carried out using Isoplot (Ludwig, 2003) and uncertainties are reported at the 95% confidence level. U-Pb data are reported in Supplemental Tables S1 and S2¹.

Sm-Nd isotopic data were obtained at the University of Wyoming on a VG Sector 54 thermal ionization multiple collector mass spectrometer. Samples were dissolved in HF and HNO₃, then converted to chlorides. One aliquot of the sample was spiked with ¹⁴⁹Sm and ¹⁴⁶Nd. Rare-earth elements (REE) were separated using conventional cation exchange chromatography, and Sm and Nd were further separated by eluting HCl through di(2-ethylhexyl) orthophosphoric acid columns. An average ¹⁴³Nd/¹⁴⁴Nd ratio of 0.511846 ± 11 (2σ) was measured for the La Jolla Nd standard, normalized to ¹⁴⁶Nd/¹⁴⁴Nd = 0.7219. Uncertainties in Nd isotopic ratio measurements are ± 0.00001 (2 standard errors). Uncertainties on Sm and Nd concentrations are ± 2% of the measured values. Sm-Nd isotopic data are given in Table 2.

Mineral compositional data were used in conjunction with whole-rock chemical composition data for thermobarometric analysis and for construction of pseudosections. Whole-rock geochemical analyses used for pseudosection calculations were obtained either from Frost et al. (2018) or by X-ray fluorescence (XRF) using a Panalytical Axios 4 KW wavelength dispersive XRF instrument at the University of Wyoming. Whole-rock compositional data used in calculation of pseudosections are given in Table 3.

Electron probe microanalysis of rutile, garnet, pyroxene, and hornblende were conducted at the University of Wyoming using the JEOL 8900 microprobe (Tables 4A–4C).

Pseudosections were constructed using Perple_X_6.7.7 (Connolly 2009) and fluid equation of state for H₂O and thermodynamic data set from Holland and Powell (1991, 1998, 2011). In all samples, it was assumed that there was a saturated component (SiO₂). The thermodynamic components selected were Na₂O, MgO, Al₂O₃, K₂O, CaO, TiO₂, FeO^t and H₂O. Solution models employed in the calculation of both pseudosections include: Gt(W), Chl(W), St(W), Crd(W), Ctd(W), Bi(W), Mica(W), Ilm(W), and melt(W) (White et al., 2014) and feldspar (Fuhrman and Lindsley, 1988). Assumptions for fluid composition are given in the text for each pseudosection.

Mineral abbreviations used in the text are after Whitney and Evans (2010). Mineral abbreviations used in the pseudosection plots are those generated by the Perplex program, correlate with the solution models listed above, and adhere to the following conventions. If an abbreviation begins with an upper case letter, a solution model was used for that mineral in generating the pseudosection. If an abbreviation begins with a lower case letter, a pure endmember was used in generating the pseudosection. Mineral abbreviations beginning with upper case letters and shown in italics are subsets of the solution model used in the calculation of the pseudosection. For example, Pl refers to plagioclase and Afs refers to alkali feldspar produced when the feldspar (ternary feldspars) solution model was employed.

Locations and brief descriptions for all samples referenced in this contribution are given in Table S3 (^{footnote 1}).

The older gneisses in the northern Teton Range have been affected by three phases of folding. The first, F₁, is manifested by rarely preserved refolded isoclinal fold hinges within the pelitic gneiss in the Moose Basin gneiss and in the Layered Gneiss (Fig. 4). The second, F₂, produced decameter-scale isoclinal folds in the Moose Basin gneiss, the limbs of which define the major foliation in the region (A in Fig. 4). In Moose Basin the mafic gneiss forms the cores of the F₂ antiforms. Adjacent to the pelitic gneiss the mafic gneiss is amphibolite, locally with garnet. Distal from the contact with the pelitic gneiss the mafic gneiss contains granulite facies assemblages. The contact between the Moose Basin gneiss and the leucogranites is conformable to the limbs of the F₂ folds (B in Fig. 4).

The Webb Canyon and Bitch Creek gneisses do not contain F₁ elements, but they have a lineation that is parallel to the trend and plunge of the F₂ fold axes of the Webb Canyon Gneiss (C in Fig. 4); F₂ fold hinges are seen only locally in the leucogranites (D in Fig. 4), and F2 folds occur in some of the migmatites formed during peak metamorphism. The leucogranites are intruded by amphibolite dikes that locally show extensional sense-of-shear (Frost et al., 2016).

The F₃ event is seen as a broad open fold that exhibits no axial planar schistosity and involves all metamorphic rocks in the area (Fig. 4). The Moose Basin gneiss occurs in the core of a major F₃ synform and hence lies structurally above the Layered Gneiss, although the direct contact is nowhere preserved. Poles to the F₂ foliations measured in a small region around Moose Basin show a strong maximum in the southeast quadrant (Fig. 5A). Lineations in the same area lie ∼90° away in the northeast quadrant (Fig. 5B). The poles of the foliation for all the gneisses across the northern Teton Range define a fold, the axis of which trends 20° and plunges 24° (Fig. 5C). We interpret this as the axial orientation of the F₃ folding. Lineations across the northern part of the range form a strong maximum that lies in the northeast quadrant, with the same orientation as the axial plunge of the F₃ fold. The fact that the lineations in the Moose Basin area (Fig. 5B) have the same trend as all the lineations across the northern part of the range (Fig. 5D) indicates that the F₂ and F₃ folds were nearly coaxial and suggests that they formed in the same deformation event.

Two leucosome rocks contain zircon that formed during crystallization of the partial melts, and U-Pb dating of zircon from these rocks constrains the timing of metamorphism in the northern Teton Range (Tables S1 and S2 [^{footnote 1}]). One is a garnet-bearing leucosome that formed from its host mafic granulite in the Moose Basin gneiss (06T16, Fig. 3B). The zircon grains in this sample range in shape from rounded to euhedral and contain inherited cores that are resorbed, zoned, and relatively bright in cathodoluminescence (Fig. 6A). The second sample is a leucosome from partially melted Layered Gneiss (06T9, Fig. 3C). The zircon grains from this sample are euhedral, show primarily oscillatory growth zoning under cathodoluminescence, and lack recognizable older cores (Fig. 6B).

There are two age populations of zircon from 06T16 (Fig. 7). The cores of the zircon grains, some of which are relatively bright under cathodoluminescence, give ages around 3.1 Ga (Fig. 7A). We interpret this as reflecting the age of the protolith. The rims of the grains, which are generally dark under cathodoluminescence, yield a U-Pb age of 2694.8 ± 6.7 Ma (MSWD = 1.0) (Fig. 7B; one highly discordant analysis was excluded from the calculation). We interpret these data as indicating the age of peak metamorphism and the partial melting of the mafic rocks. The zircon from 06T9 yields U-Pb ages of 2684.8 ± 4.9 Ma (MSWD = 0.3), which we interpret to represent the age of zircon growth from partial melts associated with the F₂ deformation event and the M2 metamorphic event (Fig. 7C). The formation of leucosome in 06T9 was coeval with the intrusion of Webb Canyon and Bitch Creek gneisses between 2686 and 2675 Ma (Frost et al., 2016).

Sm-Nd isotopic data were obtained to help constrain the protoliths of the Moose Basin and Layered gneisses. We report data for 20 felsic samples (8 samples of the Moose Basin gneiss, 12 samples of the Layered Gneiss), and 4 samples of mafic rocks within the Layered Gneiss (Table 2; Fig. 8). In general, Sm and Nd contents and Sm/Nd ratios are higher in the Layered Gneiss than in the Moose Basin gneiss. These characteristics reflect the high REE contents and lack of light REE (LREE)-enrichment typical of the Layered Gneiss compared to the generally lower REE contents and LREE enrichment of the Moose Basin gneiss (Frost et al., 2018). Mafic rocks interlayered within the Layered Gneiss have variable Sm and Nd contents and Sm/Nd that reflect flat to LREE-enriched REE patterns (Frost et al., 2018). Amphibolite dike samples (07T9 and 07T27) have higher Sm/Nd and positive initial ε_Nd, whereas gabbro samples (07T13 and T35E) have lower Sm/Nd and slightly negative initial ε_Nd.

Both the Moose Basin and Layered gneisses are characterized by variation in initial ε_Nd, but there are important differences between the two units. With one exception, the Moose Basin gneisses have initial ε_Nd less than –0.8, whereas with one exception the felsic and mafic Layered Gneiss samples all have initial ε_Nd –0.8 or greater (Fig. 8). These differences in initial Nd isotopic composition of the Moose Basin and Layered gneisses indicate that the two paragneiss units were composed of clastic detritus that was derived from different sources: those of the Moose Basin gneiss are on average older crustal sources, and those of the Layered Gneiss are more juvenile. This conclusion is also supported by the presence of >3 Ga zircon cores in the Moose Basin gneiss and the absence of these ancient cores in all zircons analyzed from the Layered Gneiss.

Rocks with low variance assemblages were selected for thermobarometric analysis. Mineral and whole-rock chemical data were evaluated in the context of petrographic observations described above to deduce the P-T-t path inferred below.

The pelitic rocks in the Moose Basin gneiss are interlayered with leucosomes, indicating significant melting at peak metamorphic conditions (Fig. 3A). The least retrograded Moose Basin gneiss pelites preserve evidence of an early, high-pressure granulite facies metamorphic event predating or coeval with penetrative deformation and a later amphibolite facies event that is largely unaffected by penetrative deformation. Garnets with rare kyanite, sillimanite, zircon, pyrite, and rutile inclusions record the peak metamorphic conditions experienced by these rocks. While sparse biotite, plagioclase, and ilmenite after rutile occur along fractures within the garnets and along the garnet margins, none of these phases was observed to occur as inclusions within homogeneous portions of the garnets suggesting they were not stable during early garnet growth at peak metamorphic conditions (Figs. 9A and 9B).

The later amphibolite facies reaction assemblage (in the same rocks) is Gt–St–Crd–Bt–Sil–Pl–Ilm–Zrn–Qz. Retrograde chlorite after biotite and garnet and intergranular sericite are also present. The garnets are rimmed by cordierite, are typically irregular in shape and have likely experienced significant resorption, consistent with enrichment of Mn and Fe around current grain boundaries (Table 4A). Because of extensive resorption and generally anhedral shapes of garnets, relative timing of crystallization of sillimanite and kyanite is not constrained by spatial distribution of the aluminosilicate inclusions in garnets, but cm-sized kyanite blades in the matrix are partially replaced by sillimanite, indicating that the kyanite predates sillimanite in these samples (Fig. 9C). Staurolite is locally reacting to cordierite + sillimanite (Fig. 9D). Ilmenite occurs as mantles on rutile within the matrix (Fig. 9E), but is absent on rutile inclusions in unfractured garnet (Fig. 9B). Textural relations clearly indicate that the Gt–St–Crd–Bt–Sil–Pl–Ilm–Zrn–Qz assemblage post-dates the Gt–Ky–Sil–Zrn–Qz assemblage, and the delicate reaction textures lack evidence of significant deformation after reaction. Finally, alkali feldspar is restricted to the leucosomes and is completely absent in the pelitic rocks.

Temperatures for both assemblages are constrained by the Zr-in-rutile thermometer (Tomkins et al., 2007). Rutile occurring as isolated inclusions in garnet has Zr as high as 3800 ppm, whereas rutile in the matrix and near fractures in the garnet ranges from 260 to 400 ppm Zr (Table 4B). The Zr-in-rutile thermometer yields temperatures of 893 °C (8 kbar) to 920 °C (12 kbar) for the rutile inclusions in garnet, and temperatures of 630 °C to 690 °C (depending on the SiO₂ polymorph and the pressure) for the rutile in the matrix. Because rutile loses Zr during re-equilibration at lower temperature, the maximum temperature was at least 900 °C.

A pseudosection was calculated for one sample (04T30) assuming SiO₂ saturation, fixed 4% H₂O by weight, and components Na₂O, MgO, Al₂O₃, K₂O, CaO, TiO₂, and FeO (Fig. 10). Details for solution models used in the calculation are given in the “Methods” section. We did not include MnO in this calculation because of its low concentration in the whole-rock chemical analysis. We did not include Fe₂O₃ because the ilmenite is very nearly stoichiometric and pyrite is present in both the matrix and as inclusions in the garnets in this sample. We based our estimate of H₂O on loss on ignition from preparation of the XRF borosilicate beads; this estimate is supported by the fact that significantly lower estimates for H₂O (2.5%) yield large alkali feldspar fields near the independently estimated T and P, and significantly higher estimates for H₂O (6% and H₂O saturation) truncate the kyanite/sillimanite reaction curve at a melt field above 700 °C.

The 900 °C Zr-in-rutile temperature intersects the kyanite/sillimanite phase boundary at 12 kbar, constraining the peak granulite facies conditions. These results are independent of assumptions about H₂O, MnO and Fe₂O₃ abundances or activities. These conditions are well above the wet melting curve for biotite, consistent with the presence of significant amounts of leucocratic material in the rocks. Biotite, plagioclase, and alkali feldspar are not stable in this sample at these conditions, consistent with their absence as inclusions in the garnet. Interestingly, garnet core compositions (Table 4A) combined with 900 °C estimate for temperature also yield a minimum GRAIL (garnet–rutile–Al₂SiO₅ polymorph–ilmenite–quartz) pressure of equilibration for these rocks of 12 kbar (Bohlen et al., 1983). The Gt–St–Crd–Bt–Sil–Pl–Ilm–Zrn–Qz amphibolite facies assemblage occurs at 675 °C < T < 710 °C and 6.4 kbar < P < 7.5 kbar. This result is completely independent of but consistent with the Zr-in-rutile temperature based on rutile occurring outside of garnets (and rimmed by ilmenite) in this sample.

Like the pelites, the least retrograded Moose Basin gneiss mafic granulites preserve evidence of an early, granulite facies metamorphic event and a later amphibolite facies event. Mafic rocks in the cores of F2 folds preserve the granulite facies assemblage Opx–Cpx–Pl–Ilm–(Gt)±Qz, and garnet-bearing leucosomes from partial melting of these rocks occur locally (Fig 3B, C). The presence of garnet in the pre-amphibolite facies mafic granulites is inferred based on occurrence of likely pseudomorphs of cummingtonite/gedrite/plagioclase after garnet in the rocks and on the presence of garnet in the associated leucosomes; no garnet has been identified in the mafic granulites. The later amphibolite-facies assemblage (in the same rocks) is Hbl–Pl–Ilm–Qz, where hornblende occurs exclusively as mantles on ilmenite and relict orthopyroxene and clinopyroxene formed during the earlier granulite facies metamorphism. These rocks are generally massive with only weakly developed fabric. The absence of surviving garnet and the strong dependence of P, T estimates for these rocks on P_H2O, however, limits their value in establishing quantitative estimates of metamorphic conditions for either the granulite facies or the amphibolite facies metamorphic events.

The two-pyroxene mafic granulite assemblage does provide temperature constraints for these rocks but not pressure constraints. Both pyroxenes are moderately exsolved and have outer rims depleted in exsolution lamellae, indicating that some re-equilibration of peak compositions has occurred. An estimate of maximum T can be obtained using cation exchange thermometry for the two-pyroxene assemblage, which is not dependent on P_H2O. Using integrated compositions from cores of orthopyroxene and clinopyroxene grains (Table 4C), the QUILF program of Andersen et al. (1993) yields temperatures in the range of 816 °C to 832 °C. The presence of garnet-bearing leucosomes in the mafic granulites suggests temperatures closer to 900 °C (Fig. 3C; Pattison et al., 2003). Without preserved garnet these rocks provide no constraints on pressure of equilibration. However, they yield temperatures comparable to the more reliable pelite-based temperatures described above, and their granulite-facies assemblage and close proximity to those high-pressure pelites suggests that they probably equilibrated at pressures comparable to those above.

The Layered Gneiss includes interlayered quartzofeldspathic, migmatitic, and amphibolitic rocks. The most common assemblage in the Layered Gneiss is Qz–Bt–Pl±Afs±Gt, and rocks with assemblages suitable for thermobarometry are rare. We did not recognize high-pressure granulite facies assemblages in the Layered Gneiss, but rare migmatitic pelitic samples yield amphibolite facies metamorphic conditions (Fig. 3D). Sample 06T10 is a pelitic enclave in a leucosome (06T9) from which the zircons yielded an age of 2685 Ma (Fig. 7). This rock contains Gt–St–Bt–Sil–Pl–Rt–Zrn–Qz. The biotite is partially replaced by chlorite and the plagioclase is partially replaced by sericite. Plagioclase laths >5 cm in length also occur, suggesting the presence of melt zones. The garnets are subhedral, fractured, and contain abundant inclusions of quartz, plagioclase, biotite, rutile, zircon, and rare, minute grains of ilmenite (Fig. 9F). Patches of sericite and quartz with myrmekitic texture are common (Fig.9G), indicating significant hydrothermal alteration and possibly potassium metasomatism. The rutile invariably occurs intergrown with chlorite and Zr concentration in rutile occurring within the matrix and as inclusions in garnets is generally below detection limits (Figs. 9H and 9I). For these reasons, we interpret the rutile as a retrograde product forming after ilmenite, and we believe the principal Ti phase at peak metamorphic conditions was ilmenite.

Because of evidence for late re-equilibration of mineral compositions, we restrict our analysis of thermobarometric conditions of equilibration for this sample to inferences based on assemblage alone (and not on individual mineral compositions). Garnet and staurolite are texturally early, suggesting that 06T10 records a melting event in which the pelitic rocks contained staurolite. We calculated a pseudosection for this sample assuming 4% H₂O, SiO₂ saturation, and no Fe₂O₃ (Fig.11). This pseudosection requires the Gt–St–Bt–Sil–Pl–Ilm–Zrn–Qz assemblage to coexist along a reaction curve from 665 °C, 6 kbar to 690 °C and 7.3 kbar; the absence of kyanite and primary rutile restricts this assemblage to P < 7.3 kbar (Fig. 11). Allowing some Fe₂O₃ would lower the activity of ilmenite and extend its stability to higher pressure. Based on this calculation, coexistence of this assemblage and melt restricts the assemblage to P > ∼7.2 kbar. The assemblage Gt–St–Sil–Pl–Qz has been shown experimentally to coexist with a water-saturated melt in the limited P-T range 665 °C < T < 680 °C and 6.6 kbar < P < 7.5 kbar (shaded area in Fig. 11; Garcia-Casco et al., 2003). The results of the pseudosection calculation are within 10 °C and 100 bars of the experimental results for the same assemblage. These relations suggest that the melting of 06T10 occurred at ∼680–700 °C and 6.5–7.5 kbar.

Two compositionally distinct groups of amphibolites occur in the northern Teton Range with the assemblage Hbl–Pl–Bt–Ilm–Qz±Gt±Ttn±Cum. One set has high Fe/(Fe+Mg) whole-rock and mineral compositions, comparatively sodic plagioclase, and garnet is present in virtually all of these samples. The second set has lower Fe/(Fe+Mg) ratios and more calcic plagioclase and only a few samples in this category contain garnet (Fitz-Gerald, 2008). Both varieties occur in both the Moose Basin gneiss and in the Layered Gneiss. The amphibolites form generally concordant but locally cross-cutting bodies within the host gneisses and leucogranites; they all have well-developed fabrics. Fitz-Gerald evaluated pressures and temperatures for these rocks using the barometer of Kohn and Spear (1990) and the thermometer of Dale et al. (2000). Mineral chemistry appropriate for this barometer occurred only in the lower Fe/(Fe+Mg) amphibolites. Two garnet-bearing samples within the prescribed composition range for the barometer yield 540 °C to 600 °C and 4.0 ± 1 kbar and there is no systematic difference between amphibolites from the Layered Gneiss and the Moose Basin gneiss (Fitz-Gerald, 2008). These results are consistent with the presence of titanite in the assemblage and the absence of garnet in all but the most iron-rich amphibolites in the region.

Based on the information presented above, the following observations must be taken into account in a tectonic interpretation of the Archean rocks of the northern Teton Range.

• 1. Two contrasting gneiss units are present in the northern Teton Range:

  • a. Moose Basin gneiss, composed of metapelitic and mafic rocks. These have reached granulite facies and have partially melted. Peak pressures and temperatures are in excess of 12 kbar and 900 °C. The pelitic rocks contain zircon cores ca. 3.1 Ga and their Nd isotopic compositions are consistent with derivation from older continental crustal sources. The Moose Basin gneiss preserves evidence of both F1 and F2 deformation.

  • b. The Layered Gneiss, composed of quartzofeldspathic paragneiss, interlayered amphibolite and areas of peridotite and gabbro. These quartzofeldspathic gneisses are migmatitic but apparently did not exceed amphibolite facies conditions of ∼7.5 kbar and 700 °C. The Layered Gneiss has relatively radiogenic Nd isotopic compositions consistent with derivation from juvenile sources. The Layered Gneiss preserves evidence of both F1 and F2 deformation.

  • c. Amphibolites, some of which contain garnet, occur in both the Moose Basin gneiss and in the Layered Gneiss. Amphibolite dikes are generally concordant but locally cut the leucogranites and migmatites, suggesting that they were emplaced either during or after the emplacement of the leucogranites. Titanite-bearing garnet amphibolites in both gneisses record pressures and temperatures of 4 ± 1 kbar and 540 °C to 600 °C.

• 2. These units are intruded by the Webb Canyon and Bitch Creek leucogranites. These leucogranites did not experience the earliest folding event, F₁, which is preserved in both the Moose Basin gneiss and Layered Gneiss.

• 3. Peak granulite facies metamorphism is dated at 2695 ± 7 Ma by zircon growth in leucosome associated with mafic granulite in the Moose Basin gneiss. The date of partial melting of the Layered Gneiss is 2685 ± 5 Ma, within the range of crystallization ages for the Webb Canyon and Bitch Creek leucogranites (2675–2686 Ma).

The thermobarometry shows that the rocks from the northern Teton Range record three distinct thermal regimes (Fig. 12). The Moose Basin gneiss records a high-pressure granulite-facies event of T ∼ 900 °C and P > 12 kbar (M1) with a cooling trend to 690 °C and P ∼7 kbar (M2). The Layered Gneiss records peak conditions of only amphibolite-facies, with a pressure of ∼7.3 kbar and temperature of ∼690 °C. Significant melting accompanied both metamorphic events, and some unknown amount of melt was probably lost from the affected units. Garnet amphibolites that occur within the Moose Basin gneiss and the Layered Gneiss record pressures lower than 5 kbar and temperatures less than 600 °C (M3, significantly lower grade conditions than M2).

From the P-T relations and the geochronology we infer that the Moose Basin gneiss underwent high-pressure granulite metamorphism and melting at 2694.8 ± 6.7 Ma; these rocks were then tectonically juxtaposed with the Layered Gneiss at conditions near those of Barrovian metamorphism at 2684.8 ± 4.9 Ma (Fig. 12). A final lower grade amphibolite facies event is recorded in both the Moose Basin gneiss and the Layered Gneiss garnet amphibolites at 540 °C < T < 600 °C and P 4 ± 1 kbar (M3). This event also produced titanite in amphibolite and gabbro throughout the Teton Range. This event is interpreted as the time of accretion of crustal blocks, including the northern Teton Range, to the Wyoming craton (Frost et al., 2018) and is unrelated to the 2.70–2.68 Ga collisional orogeny described in this paper. This event may also be responsible for some of the extensive retrogression of the higher grade assemblages observed in the Layered Gneiss and in the Moose Basin gneiss.

These observations require a clockwise P-T path for the Teton Range (Fig. 12). In Moose Basin gneiss pelites, both kyanite and sillimanite occur as inclusions in garnets, but sillimanite clearly replaced kyanite as metamorphism and deformation progressed. Rutile inclusions in garnet record temperatures of 900 °C and garnet-bearing leucosomes formed in mafic granulites (M1). Approximately 10 m.y. later, amphibolite facies assemblages replace the granulite facies assemblages in Moose Basin gneiss pelites and mafic granulites, and rutile everywhere except as inclusions in garnet lost most of its Zr in response to falling temperatures. Migmatitic Layered Gneiss lacks evidence of the early granulite facies metamorphic event (M1), but clearly records evidence of the later amphibolite facies event (M2). This clockwise P-T path is likely the product of rapid burial by thrusting, heating, uplift, and exhumation, all characteristic of continent-continent collisions in Phanerozoic time (England and Thompson, 1984; Thompson and England, 1984).

The boundary between the Moose Basin and Layered Gneiss was intruded by leucogranites between 2674 ± 5 Ma and 2686 ± 5 Ma; the dates for the Webb Canyon and Bitch Creek leucogranites are indistinguishable (Fig. 13; Frost et al., 2016). These dates indicate that the high-pressure metamorphism occurred around 10 m.y. before tectonic assembly, although within error it could have been nearly coeval. Extensive emplacement of leucogranites could have followed the 2685 Ma tectonic assembly for up to 10 m.y. The ages of the leucogranites fall within error of each other, but because some leucogranites show F₂ folding and some do not, Frost et al. (2016) proposed that the emplacement must have occurred in at least two stages.

The Webb Canyon and Bitch Creek leucogranites, although both trondhjemitic, have distinct geochemical characteristics. The Webb Canyon Gneiss is strongly ferroan, comparatively low in alumina, and is characterized by high Zr and Y, low Sr, and high REE contents that define “seagull” shaped patterns. The Bitch Creek gneiss is lower in Zr, Y, and REE and alumina and Sr are higher than in the Webb Canyon Gneiss. Frost et al. (2016) concluded that these differences reflect two different petrogenetic processes, both of which are typical of collisional environments. The geochemical characteristics of the Bitch Creek gneiss are consistent with water-excess melting in a collision-related overthrust, where a relatively cool, hydrous lower plate releases water into a hotter upper plate. The Webb Canyon magmas formed by dehydration melting caused when dramatically thickened crust undergoes gravitational collapse and tectonic extension. Because the slope of the dehydration reaction is positive, melts formed by this process can migrate to shallower levels without intersecting their solidus. The sheet-like form and F₂ fabrics of the Webb Canyon Gneiss are consistent with layer-parallel magma migration during orogenic collapse. Both water-excess and dehydration melting have been called upon to explain leucogranite of the Himalaya (Le Fort et al., 1987; Reichardt and Weinberg, 2012; Searle et al., 2009; Visona’ et al., 2012).

Five features in the gneisses of the northern Teton Range are characteristic of modern continent-continent collisional orogens.

• (1) High-pressure granulite-facies metamorphism is followed by clockwise cooling.

• (2) High-pressure metamorphism precedes the tectonic assembly of the gneisses by ∼10 m.y. This is typical of high-pressure rocks from the Himalaya and Alps, where rocks subducted to great depths are exhumed as the subduction zone migrates prior to the final collision (Rubatto et al., 1999; Babist et al., 2006).

• (3) Leucogranites are emplaced in an extensional environment immediately following tectonic assembly (Frost et al., 2016). Such a time scale is typical of the Himalayan orogeny (Searle et al., 2003). Our results suggest that leucogranite emplacement accompanied tectonic assembly at 2685 Ma, and continued for as long as 10 m.y.

• (4) Initial ε_Nd of the metasedimentary rocks in Moose Basin gneiss and those in the Layered Gneiss are distinct, which suggests the two sedimentary packages were derived from different sources and were deposited in separate basins. The Nd isotopic evidence thus supports the interpretation that the Moose Basin gneiss represents detritus derived from a different, more ancient continental block than the Layered Gneiss, which is derived from relatively juvenile crustal sources, and that these were juxtaposed during collision at high-pressure granulite facies conditions. Ultramafic rocks within the Moose Basin and Layered gneisses may represent remnants of oceanic crust caught up with the two sedimentary packages during collision.

• (5) The association of metaperidotite, metagabbro, and juvenile metasedimentary rocks in the Layered Gneiss is an assemblage that is found in Phanerozoic mountain belts where an ocean basin has been consumed during continental collision. An example is the Tectonic Accretion Channel in the central Alps (Engi et al., 2001).

Korenaga (2013) observes that for high-pressure granulites to be observed at the Earth’s surface, two processes must occur. First, crustal rocks must be taken down to depth where they are metamorphosed under high pressure and temperatures conditions. Second, they must then be exhumed rather rapidly in order to preserve peak metamorphic conditions. One might predict that in a hotter Archean Earth, hotter slabs would be weaker and less likely to subduct. Korenaga’s analysis of the strength of Archean slabs suggests that to the contrary, they are likely to be stronger. A hotter mantle early in Earth history is likely to have experienced a greater degree of partial melting and partial melting also may have taken place at greater depths. Because water in the source mantle is transferred almost completely into the melt phase, a thicker dehydrated lithosphere is formed, one which is stiffer and stronger than hydrated mantle. As these thick plates become more dense with age they will reach neutral buoyancy, and then subduct.

If Archean lithosphere is strong enough to subduct, why are Archean high-pressure granulites so rare? Korenaga speculates that the paucity of high-pressure rocks early in Earth’s history may reflect some difficulty in the exhumation stage. Thermomechanical numerical models indicate that the ultimate driver is the ratio of the intrinsic buoyancy of the subducted continental crust to side traction forces in the conduit (Butler et al., 2014). Models by Husson et al. (2009) show that slab rollback within subduction zones induces conditions where material in the subduction channel flows upwards. Moreover, when rollback is associated with a decrease in slab dip, the exhumation process becomes more efficient. These studies suggest that exhumation is not expected to accompany vertical tectonic processes that have been called on to explain many Archean terrains. High-pressure granulites may instead be restricted to those areas where subduction enables exhumation rather than where sagduction or crustal overturn processes occur.

We have argued above that the Archean rocks of the northern Teton Range are best interpreted in terms of a collisional orogeny in which fine-grained sediments and ultramafic rocks are transported along a clockwise P-T path to depths of 35 km or greater, then exhumed during orogenic collapse and intrusion of leucogranites. This distinctly plate tectonic interpretation has not been universally invoked for the other Archean high-pressure granulite terrains shown on Table 1 and Figure 14. In many cases, incomplete preservation or a subsequent metamorphic overprint precludes definitive identification of tectonic process. For the oldest high-pressure rocks, both plate tectonic and pre-plate vertical tectonic interpretations have been proposed. The oldest high-pressure granulites occur within the 3.8 Ga Itsaq gneiss. Brown and Johnson (2018) estimate that the assemblage Gt–Cpx–Hbl–Qz formed at ∼12 kbar and 870 °C. Some authors have interpreted these rocks to record a collisional orogeny (Nutman et al., 2013, 2015), others propose that stagnant-lid tectonics were dominant at this time and that the Itsaq rock record reflects local, episodic subduction (e.g., Brown, 2016). Likewise, the high-pressure amphibolites of the Inyoni shear zone, Barberton granite-greenstone terrane have yielded multiple tectonic interpretations. Moyen et al. (2006) suggested that the high-grade rocks were subducted to depths of at least 35 km before being exhumed along the shear zone in a cycle that took place between 3.23 and 3.22 Ga. They interpreted synkinematic trondhjemites as the result of decompression melting during return flow. Alternatively, Van Kranendonk et al. (2014) propose that the greenstones foundered to depths of 12–15 kbar during partial convective overturn at around 3.23 Ga. They suggest that the greenstones were exhumed along a mylonite zone associated with an intrusive event at 3.11 Ga (Van Kranendonk et al., 2014). If so, then exhumation was much later than expected by analogy with modern continent-continent collisions.

Most Archean high-pressure granulites are Neoarchean, and one of the best studied also has disparate tectonic interpretations. The granulites of the Assynt block, Lewisian Gneiss Complex, record multiple periods of deformation and metamorphism, making the Archean high-pressure event difficult to interpret. Johnson et al. (2016) conclude that a subduction-related origin is possible in part because many of the rocks have trace element signatures characteristics of modern arcs. However, the authors propose that a process of sagduction is also possible, in which mafic and ultramafic rocks sank into the deep crust due to their greater density compared to underlying, partially molten felsic orthogneisses. The authors suggest that downward flow would be arrested by increasing stiffness of the orthogneiss residua as partial melt was extracted.

In contrast to Archean terranes that have been affected by Proterozoic metamorphism and deformation, the last such events in the Wyoming province were Neoarchean. The lack of subsequent tectonism greatly simplifies the interpretation of Archean events in the northern Teton Range, allowing the peak conditions and clockwise P-T-t path to be identified and the 20 million year cycle of subduction and exhumation to be quantified.

We suggest that the Archean rocks of the northern Teton Range formed by a process involving lateral, collisional orogeny. The metapelitic rocks of the Moose Basin gneiss derived from a ca. 3.1 Ga continental source were tectonically buried to depths corresponding to >12 kbar at 2695 Ma. This material was exhumed and juxtaposed with the Layered Gneiss, composed of juvenile metasediments and interlayered amphibolites, gabbros, and peridotite, at depths equivalent to ∼7 kbar at 2685 Ma. The mafic and ultramafic rocks are interpreted as the remnants of ocean crust consumed prior to collision. Shortly thereafter and for as long as 10 million years following, leucogranites formed by water excess melting and dehydration melting were intruded. This history is analogous to Cenozoic continental collisions such as the Alps (Engi et al., 2001; Babist et al., 2006; Searle et al., 2009).

A number of authors, including Voice et al. (2011), Condie and Aster (2010), and Condie et al. (2011) have identified major zircon U-Pb age peaks at 2.7–2.5 Ga, 2.0–1.7 Ga, 1.6 Ga, 1.2–1.0 Ga, and 0.7–0.5 Ga. These age peaks correlate with times of supercontinent formation. Brown and Johnson (2018) noted that ages of peak metamorphism also cluster at these times, and suggest that metamorphism records the amalgamation of continental fragments into supercontinents. Neoarchean continent assembly includes the construction of Superior, Sclavia, and Vaalbara (Bleeker, 2003). Collisional orogeny recorded in the Teton Range supports the hypothesis that plate tectonic processes were involved in the amagalmation of crustal blocks into larger continental masses by the Neoarchean.

The authors acknowledge financial support provided by National Science Foundation grant EAR 0537670 to B.R. Frost, C.D. Frost, and S.M. Swapp and by University of Wyoming/Grand Teton National Park grant 49182 to B.R. Frost and C.D. Frost. We extend sincere thanks to Dave Mogk and to one anonymous reviewer for extremely thorough and very helpful reviews; the paper is greatly improved thanks to their comments and suggestions. We also thank Al McGrew for his help as the managing editor for this submission. Dr. Joe Wooden is thanked for hosting our analytical sessions on the SHRIMP-RG instruments at Stanford. This manuscript was completed while C.D. Frost was serving at the National Science Foundation.