In situ Pb–Pb garnet geochronology as a tool for investigating polymetamorphism: a case for Paleoarchean lateral tectonic thickening Open Access

Supplementary material: Description of CT imaging, analytical settings for Pb-Pb age dating and supplemental P-T diagrams, garnet mineral chemistry, whole rock XRF data, garnet trace element data and garnet and standard Pb-Pb age data are available at https://doi.org/10.6084/m9.figshare.c.6724419

The Barberton Granite–Greenstone Belt (BGGB) is the oldest and southeasternmost in a series of NE–SW-trending greenstone belts in the northeastern Kaapvaal Craton (Fig. 1a). It constitutes a key locality in the debate concerning the onset of plate tectonics (Anhaeusser et al. 1983; Kisters et al. 2003; Moyen et al. 2006; Van Kranendonk et al. 2009, 2014; Cutts et al. 2014, 2015; Wang et al. 2019). This debate has largely concerned the nature of the c. 3.23 Ga event in the BGGB with numerous studies favouring a vertical tectonic process producing this event (Van Kranendonk et al. 2009, 2014; Van Kranendonk 2011; Wang et al. 2019) and a similarly large number of studies advocating for a horizontal tectonic process (De Wit et al. 1992; De Ronde and De Wit 1994; Kisters et al. 2003; Dziggel et al. 2006; Moyen et al. 2006; Cutts et al. 2014, 2015; Diener and Dziggel 2021). Several of these studies have revealed the presence of an earlier event at c. 3.45 Ga that has previously been interpreted as a contact or regional metamorphic event related to intrusion of the early trondhjemite–tonalite–granodiorite rocks (TTGs) (Cutts et al. 2014), duplication of rocks in the southern part of the BGGB (Kisters et al. 2010) or a major tectonic uplift and accretion event at 3.43 Ga (Grosch et al. 2011). In the southern BGGB, the high grade of metamorphism related to the subsequent c. 3.23 Ga event has largely obscured the nature of the earlier event (Armstrong et al. 1990; Dziggel et al. 2002; Van Kranendonk et al. 2009; Cutts et al. 2014; Wang et al. 2019). The two most recent studies identifying the c. 3.45 Ga event have attempted to discern P–T conditions. Cutts et al. (2014) investigated two samples from the same locality, one producing a monazite U–Pb age of 3436 ± 18 Ma, and another with garnet cores containing a relict foliation that indicated peak P–T conditions of 4–5 kbar and 550°C based on P–T pseudosection modelling (Fig. 1b, point 4). Wang et al. (2019) found metamorphic zircon ages of 3443 ± 13 Ma and 3419 ± 8 Ma from samples that give P–T conditions of 550–625°C and 8.5–9 kbar, and 650–725°C and 6–8 kbar, respectively, obtained using conventional thermobarometry. Cutts et al. (2014) interpreted this early event as potentially a result of thermal metamorphism due to intrusion of the nearby c. 3.44 Ga Theespruit Pluton, although the garnet cores that they interpret to be associated with this event contain preserved foliations indicating they are syn- or post-tectonic. Wang et al. (2019) interpret both the 3.4 Ga and the later 3.2 Ga event to be associated with magmatism and result from partial convective overturn. Both previous studies produced P–T estimates using samples from directly adjacent to the Theespruit and Stolzburg plutons, which are also dated at c. 3.45 Ga (Fig. 1b; Kamo and Davis 1994; Cutts et al. 2014; Wang et al. 2019). This study targets a sample obtained from the Theespruit Formation occurring adjacent to the c. 3.51 Ga Steynsdorp Pluton (Fig. 1b). If the older metamorphism is related to intrusion of the TTGs then a sample from next to the Steynsdorp Pluton should produce an older metamorphic age.

The objective of this study is to resolve the nature of this early event by directly dating the garnet using the Pb–Pb system and using the same garnet grain to determine P–T conditions for both the 3.4 and 3.2 Ga events. There is presently a resurgence of interest in garnet dating with the advent of increasing sensitivity and sophistication of instrumentation (i.e. Seman et al. 2017; Millonig et al. 2020; Schannor et al. 2021; Simpson et al. 2021, 2023; Tamblyn et al. 2022). The utility of garnet ages is clear. As the ultimate petrochronometer (Baxter et al. 2017), it is possible to infer precise metamorphic conditions from the same garnet crystal that it is now possible to date with a variety of methods (U–Pb and Lu–Hf; Seman et al. 2017; Simpson et al. 2021). This study adds to the growing number that utilize garnets as a geochronometer and highlights the utility of obtaining age and P–T conditions from one crystal in order to unravel Archean polymetamorphism and gain insight into Archean tectonism.

The Paleo–Mesoarchean BGGB is situated in the Mpumalanga Province of South Africa and consists of a sequence of 3.55–3.22 Ga volcano-sedimentary successions known as the Barberton Supergroup (Fig. 1; De Wit et al. 1992; De Ronde and de Wit 1994; Kamo and Davis 1994; Dziggel et al. 2002, 2006; Kisters et al. 2003, 2010; Drabon et al. 2019). The BGGB is oriented NE–SW and is one in a sequence of NE–SW-oriented greenstone belts that developed on the northeastern edge of the Kaapvaal Craton (Fig. 1a). The Barberton Supergroup is broadly divided into three groups: (1) the Onverwacht Group (ultramafic, mafic and felsic volcanic rocks; c. 3.55–3.3 Ga); (2) the Fig Tree Group (clastic to volcaniclastic sediments, c. 3.26–3.225 Ga); and (3) the Moodies Group (sandstones and conglomerates, c. 3.225–3.215 Ga; Anhaeusser 1976; Sanchez-Garrido et al. 2011). The structurally lowermost units of the Onverwacht Group are the Sandspruit and Theespruit formations. These two units occur in the Stolzburg Domain in the southern BGGB (Fig. 1b) and are intruded by trondhjemite–tonalite–granodiorite (TTG) plutons: Steynsdorp (c. 3.51 Ga), Theespruit (c. 3.45 Ga) and Stolzburg (c. 3.45 Ga). The Stolzburg Domain is located in the region south of the Komati Fault and east of the Inyoni Shear Zone (Fig. 1b). Rocks in the Stolzburg Domain reached amphibolite facies conditions (Dziggel et al. 2002; Stevens et al. 2002; Kisters et al. 2003; Diener et al. 2005). A study by Cutts et al. (2014) utilized phase diagram modelling to indicate peak conditions of c. 8.5 kbar and 640°C during the c. 3.2 Ga event. Moyen et al. (2006) investigated rocks from the Inyoni Shear Zone using the average P–T method of THERMOCALC and indicated peak conditions of 12–15 kbar and 600–650°C, which they attributed to the 3.23 Ga event. Later P–T phase diagram modelling by Cutts et al. (2021) suggested more modest conditions of c. 8–10 kbar and 650°C, combined with garnet Sm–Nd ages of 3202–3200 Ma, indicating that the metamorphism in the Inyoni Shear Zone may be slightly younger than the regional metamorphism. Diener and Dziggel (2021) used P–T modelling on samples from the Inyoni Shear Zone, the central Stolzburg Domain and the Tjakastad Schist Belt, which occurs between the Stolzburg and Theespruit plutons (Fig. 1b). Their results indicated that the whole of the Stolzburg Domain experienced similar peak P–T conditions of c. 10 kbar and 650–700°C for the c. 3.2 Ga event. The upper units of the Onverwacht Group (Komati, Hooggenoeg, Kromberg and Mendon formations) north of the Komati Fault experienced only greenschist facies metamorphism (Kisters et al. 2003). The Komati Fault is a 1 km-wide ductile–brittle extensional detachment (Dziggel et al. 2002, 2006; Kisters et al. 2003).

The Steynsdorp area occurs in the southernmost part of the BGGB (dashed box in Fig. 1b) and is composed of 3511–3502 Ma TTG gneisses (Kröner et al. 1996) and the overlying 3540–3530 Ma Theespruit Formation (Kamo and Davis 1994; Kröner et al. 1996). In the south of the Steynsdorp area, the gneisses are intruded by the 3.1 Ga Mpuluzi potassic granite batholith (Fig. 1b; Lana et al. 2010). Lana et al. (2010) present a detailed study of the Steynsdorp area, indicating that the supracrustal rocks reached peak metamorphic conditions of 10–13 kbar and 640–660°C, which they interpret to have occurred during the regional metamorphic peak at 3.23 Ga. These are similar to the conditions obtained from the Inyoni Shear Zone by Moyen et al. (2006). The Steynsdorp Pluton and Theespruit Formation contain a unidirectional northeastward-directed lineation (see Lana et al. 2010, Fig. 2). This, together with consistent granitoid-up–greenstone-down kinematics, point to extrusion of the TTG gneisses along an extensional detachment. The contact between the amphibolite facies Theespruit Formation and the greenschist facies Komati Formation is a 20–50 m-wide zone where altered pillow lavas of the Komati Formation become elongated and highly sheared rods approach the contact with the Theespruit Formation (Lana et al. 2010). This boundary represents a metamorphic break with P–T estimates of 2.5 kbar and 350°C for the Komati Formation (Cloete 1999), indicating 6–8 kbar of difference between these units. Lana et al. (2010) presented no metamorphic ages but indicate the presence of large garnet porphyroblasts with Mn zonation indicative of two metamorphic cycles. They interpret that the cores of these garnets record a pre-3.2 Ga metamorphic event.

Sample BKC-3 was sampled from the Theespruit Formation in a river exposure of the Steynsdorp area (26° 09′ 18.5″ S, 030° 57′ 07.5″ E; WGS84; Fig. 1b). The outcrop consists of a strongly foliated grey gneiss with fine (mm–cm scale) dark and light banding, as well as thin quartzose layers (Fig. 2a). The largely fine-grained matrix contains large garnet porphyroblasts (Fig. 2b). The matrix consists of biotite, chlorite, epidote, quartz, plagioclase and amphibole with minor ilmenite, apatite and titanite. In the matrix of the sample, ilmenite is rimmed by titanite, there are also symplectites of quartz and epidote in the matrix, and plagioclase has albitic compositions on the rims where these symplectites occur (Fig. 2c). The foliation defining minerals are amphibole, biotite and ilmenite (Fig. 2d). Biotite in the matrix occurs as small, elongate, rounded grains with amphibole oriented parallel to the matrix foliation (Fig. 2c), but can also occur as large euhedral grains on garnet rims or in fractures in garnet grains. Where quartz and plagioclase are elongate, they are aligned parallel to the foliation (Fig. 2d). Some garnet grains are surrounded by a thin film of quartz, which increases in size in the pressure shadows of the garnet and displays a dextral shear sense (Fig. 2b). The thin quartzose layers are associated with thin dark bands consisting of biotite and amphibole, which can also wrap garnet grains (Fig. 2b). Garnet grains are large (up to 3 cm in diameter; Fig. 2b), commonly fractured and contain numerous inclusions. The larger inclusions occur dominantly as a discrete band, which is a focus of large fractures in the grain (Fig. 3a–c). This band occurs between the largely inclusion-free core and rim (Fig. 3a–c). Minerals included in garnet are quartz, plagioclase, ilmenite, apatite and biotite. Titanite, epidote, allanite and chlorite occur exclusively on the large fractures in the garnet. The garnet core has fine inclusions (<50 µm, usually much less) of ilmenite, apatite, biotite and quartz. There does not seem to be a preferred orientation for these, although in some places the ilmenite appears to occur in clumps. Generally, the matrix foliation wraps the garnet grains; however, in some places it is possibly continuous with ilmenite grains in the garnet that seem to have a preferred orientation.

Following CT imaging (see Supplementary Material for details), a single large garnet was cut out of sample BKC-3 and the weathered surface ground off before being mounted in epoxy and polished (Fig. 3d). Major element X-ray maps were conducted for Fe, Mg, Mn and Ca using a JEOL SuperProbe JXA-8200 microprobe hosted at the Steinmann-Institut, Bonn. X-ray maps were collected in wavelength dispersive mode with an accelerating voltage of 15 kV, a 50 nA beam current and a dwell time of 150 ms. Mineral major element compositions were analysed using a Zeiss EVO MA15 Scanning Electron Microscope in the Central Analytical Facility at Stellenbosch University. Textures were studied in backscattered electron (BSE) mode and mineral compositions quantified by EDX (Energy Dispersive X-ray) analysis using an X-Max 20 mm² ED X-ray detector and Oxford INCA software. Beam conditions were 20 kV accelerating voltage and 1.5 nA probe current, with a working distance of 8.5 mm and a specimen beam current of −19.0 to −20 nA. X-ray counts were typically c. 7000 cps and the counting time was 10 s live-time. Analyses were quantified using natural mineral standards. Comparisons between measured and accepted compositions of control standards analysed within this laboratory have been published by Diener et al. (2005) and Moyen et al. (2006) as a reflection of the accuracy of the analytical technique. The amount of Fe₂O₃ was calculated for garnet and ilmenite using the method of Droop (1987), although it was found to be low in both minerals (<1 wt%, usually less than 0.5 wt%). All garnet analyses are included in Supplementary Table 1.

The sample bulk composition was obtained via whole-rock XRF analysis carried out at the Central Analytical Facility at Stellenbosch University (Supplementary Table 2). The sample size was approximately 15 cm × 5 cm × 3 cm.

Garnet trace elements analyses were undertaken by laser-ablation inductively-coupled-plasma mass spectrometry (LA-ICP-MS) at the Central Analytical Facility, Stellenbosch University. The trace and major element traverses were carried out along the same line. A 40 µm-diameter ablation spot was generated by a New Wave 213 nm Nd-YAG Laser coupled to an Agilent 7500ce ICP-MS using a mixture of Ar–He as the carrier gas. Operating conditions for the laser were 5 Hz frequency and 6 J cm⁻². Data were acquired in time-resolved mode (Longerich et al. 1996), which allowed potential contamination from mineral inclusions or fractures to be identified and excluded from the analysis. NIST-610 glass was used as a primary reference standard and SiO₂ (39 wt%) was used for internal normalization. Accuracy and reproducibility of multiple analyses was established from the analysis of NIST 612. Results were better than 5% relative for most elements. Data were processed using Glitter (v 4.4.2) software and 1σ errors are reported as defined by the software (Supplementary Table 3). Chondrite-normalized trace element values were corrected using the normalization values of McDonough and Sun (1995).

Garnet Pb–Pb dating was conducted on the same garnet grain mounted in epoxy resin. Analyses were conducted at the Universidade Federal de Ouro Preto using a Thermo-Finnigan Neptune multi-collector ICP-MS coupled with a Photon-Machines 193 nm excimer laser system (LA-MC-ICP-MS). For Pb–Pb data acquisition, the magnet was settled on a virtual mass (c. 223.2) in the centre of the array and all relevant masses were measured simultaneously (²⁰²Hg, ²⁰⁴(²⁰⁴Pb + ²⁰⁴Hg), ²⁰⁶Pb, ²⁰⁷Pb, ²⁰⁸Pb, ²³²Th and ²³⁸U). Data collection consisted of 550 cycles (or sweeps of c. 0.1 s each) of all the relevant masses and ratios (the method of Lana et al. (2017) was followed), and a table detailing all analytical conditions for the Pb–Pb dating is included in the Supplementary Material). A 15 s background was collected prior to 40 s of sample acquisition, with a typical analysis time of 55 s. A spot size of 130 µm was used for all garnet analyses, with a 6 Hz repetition rate and laser fluence of 1–2 J cm⁻². Soda-lime glass SRM-NIST 614 was used as the primary reference material (with a spot size of 85 µm), and BHVO (Woodhead and Hergt 2007) and a garnet grain of known age (internal standard Zeek) were analysed in the same analytical session. Raw data were corrected offline using Saturn software (Silva et al. 2022). The ²⁰⁷Pb/²⁰⁶Pb ratio was corrected for mass bias (0.08%) and the ²⁰⁶Pb/²³⁸U ratio for inter-element fractionation (c. 25%), including drift over the 2 h of sequence time, using SRM-NIST 614 (n = 10).

Plots and ages were calculated using IsoplotR (Vermeesch 2018). All uncertainties are reported at the 2σ level. Analyses of unknowns and standards are included in Supplementary Table 4.

A garnet from a skarn in the Bushveld contact aureole known as Zeek was analysed in order to test the setup using a matrix matched material of known age, and to test whether there is a difference between Pb–Pb and U–Pb ages in garnet (i.e. as a result of Pb loss). Zeek produced a Pb–Pb isochron age of 2031 ± 127 Ma (n = 19, MSWD = 0.48). This garnet is usually used as a U–Pb internal standard, previously producing a U–Pb intercept age of 2058 ± 3.5 Ma (n = 9, MSWD = 1.09; Marques et al. 2023). The Bushveld intrusion age is 2059 ± 1 Ma (Buick et al. 2001), although more recent work indicates an emplacement interval of c. 5 myr from 2060 to 2055 Ma (Scoates et al. 2021).

The garnet preserves complicated major element zonation patterns (Fig. 4a, b). Cores are rich in X_Sps ( = Mn/(Mn + Ca + Mg + Fe²⁺)) with a maximum value of 0.16, but this drops to 0.03 moving towards the inclusion-rich fracture zone between the core and rim (Fig. 4a, b). The core of the garnet is defined as the region with the highest X_Sps content and extends towards the first low in X_Sps (0.03, seen on the left-hand side of the garnet traverse at 2.5 mm on the x-axis of Fig. 4a). In the rim, X_Sps rises abruptly to 0.1 before gradually dropping towards the grain edge (0.02; Fig. 4a). X_Pyr ( = Mg/(Mn + Ca + Mg + Fe²⁺)) rises steadily from core until the inclusion-rich fracture zone (0.04 to 0.07). In the rim, X_Pyr abruptly drops to 0.05 before rising again towards the grain edge to 0.07. In the core, X_Grs ( = Ca/(Mn + Ca + Mg + Fe²⁺)) gradually decreases from 0.15 to 0.11 before rising sharply within the inclusion-rich fracture zone from 0.19 to 0.22. After the inclusion-rich fracture zone, at the start of the rim, X_Grs abruptly drops before rising steadily in the rim (0.15 to 0.25). X_Alm ( = Fe²⁺/(Mn + Ca + Mg + Fe²⁺)) rises gradually towards the inclusion-rich domain from 0.64 to 0.73. The composition changes abruptly at the start of the inclusion-rich domain to 0.7 and then gradually drops to 0.65. On the core–rim boundary, X_Alm abruptly increases to 0.72, then gradually decreases towards the grain edge (0.65).

The garnet trace element traverse indicates that garnet cores are slightly enriched in heavy rare earth elements (HREEs; Gd_N/Lu_N = 0.05–0.48, with an average of 0.17) relative to garnet rims (Gd_N/Lu_N = 0.07–4.7, with an average of 1.21; Fig. 5a). The cores also have slightly elevated Eu anomalies (Eu/Eu* = Eu_N/(√(Sm_N*Gd_N))) of 1.18 to 3.61, whereas rims give values of 0.8 to 1.8. Zoning profiles show elevated Ti and V in the cores (Fig. 5b, c). On the edge of the core domain, there is an increase in Lu and Y (Fig. 5b).

In sample BKC-3, 20 uncontaminated analyses were obtained from the core and 28 from the rim (Fig. 6a). All analyses that have significant amounts of U or Th were removed. Generally, inclusions were easy to detect during analysis due to spikes in the signal. Garnet cores from BKC-3 have U contents below 1 ppm (generally 0.04 to 0.85 ppm). The data form a ²⁰⁶Pb/²⁰⁴Pb v. ²⁰⁷Pb/²⁰⁴Pb isochron with an age of 3435 ± 45 Ma (MSWD: 1.3; n = 20; Fig. 6b). Garnet rims have generally overlapping to higher U contents of 0.02 to 2.19 ppm. Two analyses plotted off the isochron, potentially due to contamination, and are removed from age calculations. The rest of the data form a ²⁰⁶Pb/²⁰⁴Pb v. ²⁰⁷Pb/²⁰⁴Pb isochron with an age of 3245 ± 41 Ma (MSWD: 1.1; n = 26; Fig. 6c).

Pressure–temperature pseudosections were calculated for sample BKC-3 using the software package Theriak-Domino (De Capitani and Petrakakis 2010) and the database of Holland and Powell (2011; ds62) for the geologically realistic system MnNCKFMASHTO (MnO–Na₂O–CaO–K₂O–FeO–MgO–Al₂O₃–SiO₂–H₂O–TiO₂–Fe₂O₃).

The ‘metabasite set’ of models from Green et al. (2016), converted to Theriak-Domino format by Doug Tinkham (see Jorgensen et al. 2019), were applied. These are: White et al. (2014) for orthopyroxene, garnet, biotite, muscovite and chlorite; Green et al. (2016) for clinoamphibole, augite and metabasite melt; Holland and Powell (2011) for olivine and epidote; Holland and Powell (2003) for plagioclase; White et al. (2002) for spinel and magnetite; and White et al. (2000) for ilmenite. Due to the large amount of Mn present in the garnet, MnO was included in the system despite a lack of Mn end-members in clinoamphibole, augite and ilmenite. The effect of this will be a larger stability field for garnet and potentially the presence of higher MnO contents in the modelled garnet than what appears in the sample. A diagram without MnO was also calculated using the same Fe³⁺ content as the final diagram (90% Fe²⁺) and garnet only occurred in an extremely limited region of P–T space (high P and T – see Supplementary Figure 1). Given the size of garnet grains in this sample, including MnO in the modelling seems to reflect reality more accurately.

As the mineral assemblages and field observations indicate peak metamorphic conditions to be sub-solidus or not significantly melt-bearing, H₂O was set in excess for the sub-solidus parts of the diagram. A H₂O value allowing for a wet solidus was selected for the supra-solidus parts of the diagram. Compositional isopleths for garnet were calculated to aid with interpretation of the P–T path.

Determining the effective bulk composition seen by samples containing large porphyroblasts can be problematic because as the porphyroblasts grow, their composition is removed from the effective bulk composition, along with any inclusions that they may contain (i.e. Marmo et al. 2002). In the case of sample BKC-3, the garnet contains numerous inclusions of apatite and ilmenite (with 5 wt% MnO), so along with determining the optimum Fe³⁺ content to use, we also explored how changes in the MnO and CaO contents would affect the topology of diagrams and calculated chemical compositions.

Initially, a temperature-composition (⁠$T−MFe2O3$⁠) diagram was calculated in order to determine the most suitable Fe³⁺ value to set in order to represent the mineralogy that we observe in the sample (Fig. 7a). It was found that if Fe³⁺ content is too high, magnetite becomes stable. Since no magnetite is observed in the samples, an Fe³⁺ content of 10% was used (Fig. 7a).

Following this, the effect of reducing the CaO content was investigated. Based on an estimated apatite content of 2–3%, CaO was reduced by 1.5 wt% and a T–M_CaO diagram was calculated (Fig. 7b). The reduction of CaO reduces the stability of epidote and increases the stability of white mica, neither of which are interpreted as peak minerals. This diagram was also contoured for the X_Grs content of garnet. The left-hand side of the diagram has X_Grs values of 0.28–0.32, whereas garnet in the sample has X_Grs ranging from 0.1 to 0.25. In order to better match the X_Grs values observed in the sample, the CaO content was reduced to 0.8 of this diagram (a reduction of approximately 1.2 wt%; Fig. 7b). Based on the presence of ilmenite in garnet (estimated at about 2% mode and containing 5 wt% MnO), the amount of MnO was reduced by 0.1 wt%.

As indicated above, the growth of large porphyroblasts can result in fractionation of their composition from the bulk rock composition (i.e. Marmo et al. 2002). Thus, when modelling the second metamorphic event (garnet rim growth), the composition of the garnet core was removed from the whole rock composition. Since the garnet changes composition from core to rim, the removed composition was determined using a stepwise approach. Each step represented a small domain of similar composition. In each step the composition was averaged and then weighted by volume and removed from the whole rock composition. An estimated 2% (of the rock volume) of the inner garnet core composition was removed, followed by an additional 3% of the garnet outer core composition. The garnet mode was estimated based on field and hand sample observations (while garnet porphyroblasts are large, they are sparse, making up less than 10% of the rock).

The removal of the garnet core from the starting composition resulted in the calculation of two P–T diagrams (Figs 8 & 9). The first utilizes the composition of the whole rock with the appropriate Fe³⁺ content and a slightly reduced CaO and MnO. Based on the inclusion assemblage of garnet, the mineral assemblage stable during initial garnet growth is garnet + plagioclase + biotite + ilmenite + amphibole + quartz. The mineral assemblage occurs in a large field at 2–8 kbar and 500–700°C (Fig. 8a). The prograde P–T evolution is defined by an increase in garnet mode (Fig. 8b). The major element composition of the garnet core was used to attempt to constrain the P–T conditions of garnet core growth with the X_Grs, X_Sps and X_Alm values of the core (0.15, 0.16 and 0.64, respectively) overlapping at P–T conditions of 4–5 kbar and 560–630°C (Fig. 8c; full isopleth figures are included as Supplementary Figure 2). The interpreted prograde P–T path is consistent with a decrease in plagioclase mode (from 5.5 to 0.8%; Fig. 8c). This may account for the positive Eu anomaly of the garnet core (Fig. 8c). The peak conditions for the garnet core zone are interpreted to occur in the garnet–amphibole–plagioclase–quartz–ilmenite–melt field. It is impossible to know for sure whether melt was present during the growth of the garnet core. However, the final compositions obtainable from the garnet core just before the edge of the rim are consistent with peak conditions in this field. This P–T evolution is also consistent with the change in garnet composition during core growth (increase in X_Grs, X_Alm and X_Pyr and decrease in X_Sps).

A subsequent diagram was made with the garnet core removed from the composition modelled in Figure 8 (see Figs 4 & 9a). Since there is not a large amount of garnet present in the sample, and the core is volumetrically inferior, this did not result in a significantly different diagram (compare Figs 8a & 9a). The first rim composition (highest X_Sps values for the rim) is used to define the start of the P–T evolution for the rim (X_Sps of 0.10, X_Grs of 0.15). The final rim composition indicates peak conditions, and the interpreted P–T path runs between these points and is based on an increase in garnet mode (Fig. 9b, c). Maximum pressure conditions are constrained by the absence of rutile (which occurs on the P–T pseudosection above 9 kbar; Fig. 9a). The peak field for the rim is the garnet + amphibole + plagioclase + ilmenite + quartz + melt field with peak conditions of 690–750°C and 8–9.5 kbar (Fig. 9a). The rocks have been strongly deformed; however, there are clear quartz-rich bands that increase in size in the pressure shadows of garnet (Fig. 2b), which are commonly bordered by thin mafic bands.

The method used here for determining the P–T evolution of these rocks by removing garnet from the whole-rock composition to model subsequent garnet growth events relies on several interpretations which will be outlined in detail here. The main interpretations that we have made are the selection of the core–rim location, whether the profile sees the centre of the garnet and whether the outer core was preserved (resorption prior to rim growth).

The core–rim boundary was selected based on the discontinuity of X_Sps zoning (see Fig. 4); this is most clearly seen on the left-hand side of the profile, and can also be seen in the Mn elemental map (Fig. 4b), where a dashed line shows this as the boundary between the blue and green zones. The core zone has a continuous bell-shaped X_Sps profile, interpreted as a result of fractional crystallization during continuous garnet growth (i.e. Hollister 1966; Ikeda 1993). However, there is another compositional break in the garnet profile at around 4 mm (see Fig. 4a), this time with an abrupt increase in X_Grs and decrease in X_Alm. This also corresponds to the zone with many inclusions occurring in the final part of the core zone. It is plausible that this represents a separate growth zone which has continuous X_Sps with the garnet core (i.e. Argles et al. 1999). It was not possible to get an age for this zone due to the abundance of large inclusions that were present. Since it is not possible to date this zone, we have chosen to interpret the core and this inclusion-rich zone to be continuous with the break in X_Grs and X_Alm resulting from changes to garnet growth speed (and major element supply). Vielzeuf et al. (2021) also see enrichment in Ca content in garnet rims as well as inclusion-rich rims, which they associate with melting or influx of melt into the system. Further study is required to ascertain whether this is the case for sample BKC-3.

An additional issue is whether the investigated compositions are from the actual core of the garnet. The CT images (Fig. 3a–c) show that the BKC-3 garnet is not spherical, but rather an ellipsoid, making it difficult to discern the location of the actual core. If we did not hit the actual core then it is possible that our maximum X_Sps values are not the highest values present in garnet in sample BKC3. Generally, in trace element profiles for growth zoning, garnet cores are enriched in Lu (i.e. Raimondo et al. 2017; Rubatto et al. 2020). This is not the case for BKC-3 (see Fig. 5), which has Lu enrichment only in the outer core and middle rare earth element (MREE) depletion in the core. The implication for not sampling the true core would be that the amount of Mn in the core is underestimated, which means that too little Mn may have been removed from the whole rock composition for the modelling of the rim growth. This would put the low X_Sps contents we observe in the garnet rim at higher P–T conditions in the modelled P–T pseudosection, resulting in an overestimate of the peak P–T conditions. The mismatch between the observed peak assemblage (with biotite) and the modelled peak assemblage (without biotite) already suggests that the modelled X_Sps contents for garnet are too high. We suggest that peak conditions of 700°C and 9 kbar are more suitable for the garnet rim.

The final interpretation relates to whether the outer core was preserved prior to rim growth. The increase in Lu and Y in the final part of the core may be a result of resorption of the outer core followed by later garnet growth (i.e. Skora et al. 2006; Cruz-Uribe et al. 2015) or breakdown of Y and HREE accessory minerals such as allanite (i.e. Pyle and Spear 1999; Gieré et al. 2011).

Breakdown of the garnet core prior to rim growth (or garnet resorption following the first metamorphic event) would also result in Mn resorption into garnet (i.e. Carlson 2002), which has not been observed in our samples. In BKC-3, the X_Sps profile has a distinct break between core and rim with no indication of an increase in X_Sps on the outer part of the core profile (Fig. 4a). This break is also observed in the Mn elemental map (Fig. 4b). This suggests that for the retrograde evolution of the garnet core, garnet was not resorbed. This may be due to a lack of fluid in the rock preventing the alteration of garnet to chlorite (Guiraud et al. 2001).

Despite the lack of evidence for retrogression of the sample following the growth of the garnet core, we think it is likely that the sample was exhumed prior to the growth of the garnet rim. The peak temperatures of the core are 700°C and 7 kbar. If the garnet remained at these conditions for 200 million years, even with slow diffusion rates and large grain sizes, some diffusion of the core zoning in the garnet would be observed (Carlson 2006; Caddick et al. 2010).

Also, in order to grow a garnet rim, the rock must experience an increase in garnet mode. Even with the garnet core removed, the bulk rock composition for the rim growth would likely increase garnet mode starting from low to moderate P–T (compare Figs 8b & 9b). This suggests that the hiatus in garnet growth between the garnet core and rim occurred due to exhumation.

In this study we have applied Pb–Pb dating due to the age of the sample. With an age >3.0 Ga, a garnet with initially low U does not have much U remaining (generally much less than 1 ppm (see Supplementary Table 4), making the U–Pb method less reliable. The Pb–Pb isochron method can produce a statistically robust age for the garnet core of BKC-3, producing an isochron age of 3436 ± 45 Ma (Fig. 6b). This age is similar to that of the Theespruit Pluton (c. 3443 ± 4 Ma; Lana et al. 2010) but slightly younger than the directly adjacent Steynsdorp Pluton (c. 3510 Ma; Lana et al. 2010). Plausibly this difference in age between sample BKC-3 and the Steynsdorp Pluton could be a result of Pb loss. The U–Pb method allows for the evaluation of lead loss, but it cannot be detected when using the Pb–Pb method alone. During the analytical session, an internal standard from the Bushveld metamorphic aureole was analysed and produced a Pb–Pb isochron age of 2031 ± 127 Ma (n = 19, MSWD = 0.48). This is within error of the LA-MC-ICP-MS U–Pb age produced for the same standard material (2058 ± 3.5 Ma; Marques et al. 2023) and the Bushveld intrusion age (2060–2055 Ma; Scoates et al. 2021). Both U and Pb diffusion rates in garnet are proposed to be extremely slow to negligible, with estimates of U–Pb closure temperatures of >800°C for garnet grains larger than 1 mm in diameter (Mezger et al. 1989, 1991; Burton et al. 1995; Zhu et al. 1997). Considering that the garnet grain investigated in this study has a grain diameter of 3 cm (with a core of 1 cm) and the c. 3.2 Ga metamorphic event is estimated to have reached peak temperature conditions of 600–650°C, it is unlikely that the garnet core age is a cooling age or has been affected by diffusional re-equilibration.

Figure 10a presents a compilation of all ages and P–T conditions obtained for the c. 3.45 Ga event. A U–Pb monazite age of c. 3436 Ma was obtained from the Onverwacht Group in the Stolzburg Domain (point 4, Fig. 1b; Cutts et al. 2014), interpreted to correspond to P–T conditions of 4–5 kbar and up to 600°C. Wang et al. (2019) found metamorphic zircon ages of 3443 ± 13 Ma and 3419 ± 8 Ma from samples that give P–T conditions of 550–625°C and 8–9 kbar and 650–725°C and 6–8 kbar, respectively (point 6, Fig. 1b). Even with a large error (3436 ± 45 Ma), the age obtained from BKC-3 is not within error of the emplacement age of the Steynsdorp Pluton (3511–3502 Ma; Kröner et al. 1996). This would suggest that the event resulting in garnet growth and metamorphism reaching conditions of 6–7 kbar and 700°C as determined in this study was in fact not related to the TTG emplacement. Additionally, the samples of Wang et al. (2019) and Cutts et al. (2014) were obtained from the Stolzburg Domain to the NW of the Steynsdorp area. Wang et al. (2019) presented only peak P–T estimates, but Cutts et al. (2014) suggested a clockwise P–T evolution for the c. 3.45 Ga event. This resulted in lower peak P–T conditions (c. 4.5 kbar and 550°C) but occurred on a similar geothermal gradient (Fig. 10a). One of the samples investigated by Wang et al. (2019) from the Stolzburg Domain (A2 in Fig. 10a) produced similar peak P–T conditions to sample BKC-3. Thus, the present data suggest that this event produced similar peak P–T conditions and P–T evolutions over the whole southern BGGB (Fig. 10a).

The garnet rim age of 3245 ± 41 Ma is within error of the age of the major metamorphic event in Barberton at 3.23 Ga (Dziggel et al. 2002; Stevens et al. 2002; Kisters et al. 2003; Diener et al. 2005; Cutts et al. 2014; Wang et al. 2019; Fig. 10b). Wang et al. (2019) obtained peak P–T conditions of 600–700°C and 9–10 kbar at 3222 ± 8 Ma (D1, Fig. 10b) and 8.4–12.7 kbar and 630–700°C at 3249 ± 6 Ma (D2, Fig. 10b) using conventional thermobarometry. Cutts et al. (2014) presented clockwise P–T paths with peak conditions of 7 kbar and 560°C and 8–9 kbar and 600–650°C (C1-C3, Fig. 10b). Lana et al. (2010) inferred that their obtained P–T conditions for the Steynsdorp area relate to the 3.23 Ga event (they used the rim zone and matrix minerals to calculate average P–T estimates) and indicated peak conditions of 10–13 kbar and 640–660°C. Diener and Dziggel (2021) investigated samples from the Inyoni Shear Zone, the central Stolzburg Domain and the Tjakastad Schist Belt which all produce similar peak conditions of c. 10 kbar and 650–700°C (I1-I3, Fig. 10b). As also indicated by Diener and Dziggel (2021), these results indicate similar maximum P–T conditions and clockwise P–T evolutions for the whole of the southern BGGB (Stolzburg Domain and Steynsdorp area), suggesting this region was a coherent block, metamorphosed as a whole at similar P–T conditions. Observed P–T paths usually involve a component of heating at depth (i.e. C2 and C3, Fig. 10b) consistent with a collisional setting (Diener and Dziggel 2021).

Therefore, the results of this study, in combination with previous work, suggest that the whole southern BGGB (Stolzburg Domain and Steynsdorp area) was affected by two moderate- to high-pressure metamorphic events at 3.45 and 3.23 Ga, with both having clockwise P–T evolutions consistent with crustal thickening, an increase in temperature and subsequent exhumation and cooling (Fig. 10).

The results of this study show two moderate- to high-pressure metamorphic events preserved within a single garnet crystal. The peak pressure conditions of up to 7 kbar and the presence of an earlier foliation (Cutts et al. 2014) together with the clockwise P–T evolution suggest the c. 3.45 Ga event may have been the result of a collision between early continental fragments, perhaps followed by orogenic collapse that caused basement doming. The paucity of evidence for the c. 3.45 Ga event makes assigning a tectonic setting difficult. It is possible that the c. 3.45 Ga event results from subsidence of dense greenstones into hot TTG crust, as suggested by Wang et al. (2019) for both the 3.45 and 3.23 Ga events. However, while models of this process have shown multiple P–T evolution shapes (Francois et al. 2014), the mechanism for exhuming the greenstone rocks is unclear. Moreover, if BKC-3 was at high P–T conditions for 200 million years, the major element zoning in the garnet would not be so well preserved. The rocks of the Steynsdorp area must have been exhumed prior to burial in the 3.23 Ga event. Also, the partial convective overturn model needs to be reconciled with the older Steynsdorp Pluton, meaning some delay between TTG intrusion at 3.5 Ga and peak metamorphism at 3.45 Ga.

At 3.23 Ga, the main collisional event was preserved in the BGGB, producing higher P–T conditions and elevated temperatures at depth over the whole southern BGGB (Diener and Dziggel 2021). This was followed by orogenic collapse with lateral extrusion (directed to the SW, Fig. 1; Kisters et al. 2003) of the amphibolite facies lower crust (the Stolzburg Domain and Steynsdorp area; Fig. 10). Wang et al. (2019) also indicate partial convective overturn as the cause for the 3.23 Ga metamorphic event. This model is inconsistent with the similarity of all investigated samples in the southern BGGB, with all presenting clockwise P–T evolution and peak conditions ranging from 7–11 kbar and 550–700°C (Fig. 10b).

P–T evidence alone can only indicate the possibility of Archean plate tectonics. We also require supporting tectonic and geochemical evidence from multiple locations. Besides what is presented here, the evidence for horizontal plate tectonic processes in the northeastern Kaapvaal Craton before 3.0 Ga is overwhelming, and increasingly studies utilizing novel isotope methods are indicating the presence of surface derived isotopic signatures in the Archean mantle (i.e. Lewis et al. 2023).

The BGGB is the oldest of a series of NE–SW-trending greenstone belts, with greenstone belts towards the north becoming younger towards the edge of the Kaapvaal Craton (Fig. 1a). Such a series of belts could be produced by accretion of microplates.

Isotopic studies investigating the Bushveld Complex (c. 2.05 Ga) have found sulfur isotope signatures indicating the presence of recycled Archean surface material (Magalhães et al. 2019). Surface sulfur isotope signatures have also been found in the Itsaq Gneiss Complex in southern West Greenland, indicating mantle domains that were fertilized by surface-derived material (Lewis et al. 2023).

Zircon Hf isotopes from the Bushveld Complex, which indicate an enriched mantle source, are suggested to be the result of prolonged subduction on the northern margin of the Kaapvaal Craton prior to collision with the Zimbabwe Craton from 3.2 to 2.6 Ga (Zeh et al. 2009; Zirakparvar et al. 2014).

Smart et al. (2016) show that diamonds from the Kaapvaal Craton have nitrogen and oxygen isotopic compositions consistent with the inclusion of isotopically light surface material into the mantle source. All of the isotopic investigations outlined above indicate recycling of crustal material into the mantle during the Archean by some process that also results in collision and medium- to high-pressure metamorphism.

Isotopic studies can also present contrasting points of view, with Rollinson (2021) presenting trace element, Nd and Hf data to suggest that mixed isotopic signatures are the result of mixing between the lower and upper mantle facilitated by a mantle plume or heat pump. Johnson et al. (2017) presented a P–T modelling and trace element study of basalts from the Pilbara Terrane, Western Australia to suggest that Archean TTG formation was a protracted, multi-stage process that likely occurred near the base of thick, plateau-like basaltic crust. This would indicate that production of early Archean TTGs did not require subduction.

Lateral plate motion does not necessarily require modern-style subduction, with several contributions suggesting alternate mechanisms for lateral plate motion which are not driven by subduction. Chowdhury et al. (2017, 2020) suggested that convergence may have happened in the Archean as a result of lithospheric peeling. This process involves delamination or peeling off of the subcontinental lithospheric mantle and the lower continental crust. It is proposed to occur due to higher mantle temperatures; it allows for crustal recycling into the mantle and can account of the appearance of paired metamorphism, MP-HP TTGs and K-granites (Chowdhury et al. 2017, 2020).

Alternatively, Strong et al. (2023) used an isotopic study to present an accordion tectonic model for the Superior Province involving cratonic growth interspersed with periods of disaggregation with formation of rifts and volcanosedimentary successions with later collisions reassembling the fragments.

For the Isua area of West Greenland, surface isotope signatures for oxygen (Gauthiez-Putallaz et al. 2020) are recorded in metasedimentary garnet with a multistage metamorphic history. They propose that the source of these rocks was exposed to surficial alteration then buried to mid-crustal levels to form high δ¹⁸O garnet within a period of 10–50 myr and suggest that the tectonic setting for this was flat subduction followed by collisional orogeny at 3.69–3.66 Ga (Gauthiez-Putallaz et al. 2020).

Thus, if we place our P–T results for Barberton in this broader context, prograde metamorphism achieving P–T conditions of 7 kbar at 3.45 Ga and then 9–10 kbar at 3.23 Ga could feasibly be the result of lateral collision of microcontinents and tectonic thickening.

Kathryn Cutts acknowledges CNPq for her Science without Borders, Jovem Talento Scholarship. Thorsten Nagel is thanked for assistance in collecting the garnet elemental maps. We thank reviewers Gautier Nicoli, Annika Dziggel and Craig Storey for their comments, which significantly improved the manuscript. We also thank Valby van Schijndel for the editorial handling.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

KAC: conceptualization (lead), formal analysis (lead), investigation (lead), methodology (equal), writing – original draft (lead), writing – review and editing (lead); CL: formal analysis (supporting), investigation (supporting), methodology (equal), writing – original draft (supporting), writing – review and editing (supporting); GS: conceptualization (supporting), methodology (supporting), writing – original draft (supporting), writing – review and editing (supporting); ISB: methodology (supporting), writing – review and editing (supporting).

Cristiano Lana acknowledges funding from CNPQ_PQ Processo 307353/2019-2.

All data generated or analysed during this study are included in this published article (and, if present, its supplementary information files).