Knowing which geodynamic regimes characterised the early Earth is a fundamental question. This implies to determine when and how modern plate tectonics began. Today, the tectonic regime is dominated by mobile-lid tectonics including deep and cold subduction. However, in the early Earth (4.5 to 2 Ga) stagnant-lid tectonics may also have occurred. The study of high pressure–low temperature (HP–LT) metamorphic rocks is important, because these rocks are only produced in present-day subduction settings. Here, we characterize the oldest known HP–LT eclogite worldwide (2089 ± 13 Ma; 17–23 kbar/500–550 °C), discovered in the Democratic Republic of the Congo. We provide evidence that the mafic protolith of the eclogite formed at 2216 ± 26 Ma in a rift-type basin, and was then subducted to mantle depths (>55 km) before being exhumed during a complete Wilson cycle lasting ca. 130 Ma. Our results indicate the operation of modern mobile-lid plate tectonics at 2.2–2.1 Ga.
Eclogites are high-pressure metamorphic rocks mainly composed of omphacite and garnet. Their pressure–temperature conditions of formation are characteristic of modern subduction zones and, as such, they have been considered as representative of subduction processes in the geological record1,2. Few occurrences of true eclogites with precise ages have been described from the Archean to Paleoproterozoic rock record, thus the pattern and timing of early Earth tectonics are still heavily debated. The oldest currently proposed eclogites from an orogenic belt are recorded from the Belomorian Belt in Russia, which are dated between 1.9 and 2.8 Ga3,4,5. Other relicts of eclogites are found in Paleoproterozoic orogens, in the 1.9 Ga Snowbird zone from the Canadian Shield6 and in the 1.9–2.0 Ga Ubendian-Usagaran Belt of Tanzania7,8,9 (Fig. 1a). The oldest known high temperature eclogites (18–20 kbar and 800 °C; 2093 ± 45 Ma) with a MORB-like chemistry occur at the northwestern margin of the Congo Craton in the Nyong Complex of Cameroon10 (Fig. 1a).
The apparent geothermal gradient recorded by rocks may be used to discriminate geodynamical processes in the Early Earth, and more particularly to infer whether or not deep and cold subduction, i.e. ‘modern-style’ plate tectonics was in operation11,12. However, relying only on the apparent geothermal gradient might be misleading. Indeed, Archean metamorphic rocks not only record high apparent geothermal gradients, but also a large range of other possible apparent geothermal gradients (i.e. 15–30 °C/km), including low values similar to modern subduction13. The maximum pressure attained will be limited in the case of sagduction, which is a partial convective overturn due to density contrast between dense (ultra)mafic covers into their granitoid crustal basement coupled to partial melting in the lower crust13. As such, it is an intracrustal process and the maximum pressure recorded depends on the crustal thickness (greenstones and crustal basement13,14,15,16). Thus, a very high pressure (i.e. >15–20 kbar ≈ 50–65 km) seems difficult to reconcile in the case of stagnant-lid and sagduction tectonics. Therefore, the study of (U)HP-LT rocks, including eclogites, seems to be a robust tool to evidence modern-style (deep and cold subduction) tectonics. Here, we focus our study on eclogites discovered in Democratic Republic of the Congo (DRC) in the Archean to Paleoproterozoic Congo Craton.
The studied sample is conserved at the Royal Museum for Central Africa, Tervuren, Belgium (RG-45977 serial number collection, Fig. 2a) and was collected in 1946 by Pierre Schnock. This rock comes from the South East of Gandajika town (value in decimal degrees: S6.5-S7/E23.9–24.5 close to Kayemba Ngombe town), in the northern part of the Archean to Paleoproterozoic Kasai Block (Fig. 1) within the Congo Craton. The northern part of the Kasai Block is composed of the Musefu Granulitic Complex (2.6–3.1 Ga17; Fig. 1b) and the Dibaya migmatitic Complex (2.6–2.8 Ga17,18). This Archean Block was marked by the Eburnian–Transamazonian (2.2–1.98 Ga) orogeny, which resulted from the accretion of the Congo Craton and the Brazilian São Francisco Craton19. The associated metamorphism has been dated at 2.05 Ga in Cameroon10,20 (Fig. 1a) and at 2.10–2.07 Ga in Brazil21. The area records the emplacement of the Lueta gabbronoritic Complex and the Lusanza Supergroup (2.2–1.9 Ga17,19) during the Paleoproterozoic. Some enclaves of the upper Lusanza Supergroup have been described within the Musefu granulitic Complex close to Mwene Ditu town (Fig. 1b). No evidence of HP rocks (blueschist or eclogite) has previously been described from the area, other than the ca. 70 Ma22 eclogite xenoliths from kimberlites found close to Mbuji-Mayi town.
The chemical composition of the eclogite is basaltic (Table 1 and see Supplementary Fig. S1) with SiO2 = 51.6 wt.%, Na2O = 2.7 wt.% and TiO2 = 1.2 wt.%. It contains low K2O (<0.5 wt.%), but high CaO (11.3 wt.%). A primitive mantle-normalised trace element diagram (see Supplementary Fig. S2) shows an enriched-MORB signature.
The sample is a retrogressed eclogite and consists mainly of garnet, clinopyroxene, amphibole, rutile, feldspar, ilmenite, hematite, quartz and pyrite (Fig. 2, see Supplementary Table S1 and Fig. S3). Garnets are a solid solution between almandine (53–60%), grossular (21–29%), pyrope (11–19%) and spessartine (1–7%; see Supplementary Table S1 and Fig. S4a). They present a zoning pattern with a Fe- and Mn-rich core, and Ca- and Mg-rich rims. Rutile, clinopyroxene, amphibole and quartz are present in inclusions in garnet (Fig. 2f) and form the first paragenesis. No coesite was found. Corona textures around garnet are retrograde (Fig. 2c–f). Some garnets display atoll-shaped microstructures (Fig. 2e). A similar garnet shape was observed in other eclogitic rocks23,24,25. Clinopyroxenes have a pale greenish colour and constitute the major part of the matrix, often associated with albite-rich plagioclases in symplectites, which grew during the decompression (Fig. 2b). They are Ca- and Na-rich (see Supplementary Table S1) and have a composition of aegirine-augite. The XMg content varies from 0.016 to 0.08, the XCa between 0.88 to 0.98 and the XFe between 0.05 and 0.40. The jadeite amount (XJadeite) is between 2.0 and 4.0. However, as this eclogite is retrogressed, the initial composition of clinopyroxene was close to omphacite. The composition of omphacite was estimated by adding the oxide wt.% of clinopyroxene and the oxide wt.% of albite-rich plagioclase (SiO2 (Cpx) + SiO2 (Pl); TiO2 (Cpx) + TiO2 (Pl); …) analysed by electron microprobe (see Supplementary Table S1). The estimated compositions of XJadeite in omphacite were close to 24–28 wt.% and probably below 30 wt.%. Amphiboles have an intense greenish colour. They are mainly calcic, with Mg content ranging between 0.5 and 0.65, the Na + K content <0.05 and the Si content <7.1 (hornblende: pargasite to ferro-edenite; see Supplementary Table S1 and Fig. S4b). They occurred between garnet and pyroxene and sometimes within garnet (Fig. 2c,d). Feldspars are rich in Na and Ca when close to clinopyroxene and amphibole, and richer in K close to garnet (Fig. 2f). Rutile occurs in the matrix and mainly as inclusions within garnet (Fig. 2c,d). Ilmenite, hematite and titanite commonly replace rutile in the matrix (Fig. 2c,e). Apatite and zircon occur as accessory minerals. Kyanite is absent and, except amphibole, no hydrated mineral is present.
In order to constrain the P–T conditions, we performed thermodynamic modelling (see Supplementary Fig. S5) using the phase-diagram calculation software Perple_X26 (version 6.8.3) and the self-consistent thermodynamic database and mineral solution models (solution_model_682; upgrade 2018). Bulk-rock compositions were calculated in the TiMnNaCaFMASH system from modal phase proportions. Mineral solution models used are Grt(WPH)27, Opx(HP)28, Cpx(HP)28, Omph(GHP)29, Pl(h)30, Chl(HP)31 and cAmph(DP)32. Water content was estimated at 0.4 wt.% using a xH2O vs. temperature diagram (at 17 kbar). Considering the pseudosection, the first paragenesis of garnet, omphacite, amphibole, quartz and rutile is stable over a large range of P–T between 400 and 550 °C for a pressure exceeding 10 kbar but lower than 24 kbar because no coesite was present. Adding the isopleths modelled for this assemblage for garnet: XPyrope (11–15 wt.%) and for XJadeite in the estimated omphacite (<30 wt.%), P–T conditions for the first paragenesis are estimated between 17 to 23 ± 1 kbar and 500–550 ± 50 °C (Fig. 3d and Supplementary Fig. S5, see33,34 for the associated errors). The exhumation is characterized by higher content of XPyrope (15–19 wt.%), a low content of XJadeite in the clinopyroxene (2–4 wt.%), the appearance of plagioclase by the substitution of omphacite in augite and albite (XAlbite: 61–81 wt.%) and the appearance of ilmenite and hematite (Fig. 3d and Supplementary Fig. S5) around 7.5–9.5 ± 1 kbar and 450–575 ± 50 °C.
Zircons (20–100 µm) show subhedral to oval shapes, some grains displaying irregular and poorly visible zoning with locally preserved thin overgrowths (see Supplementary Fig. S6). Their morphology is very similar to that of zircons observed in some Variscan eclogite-facies meta-gabbros35. The rutiles (50–500 µm) generally appear homogeneous in BSE and reflected light images (see Supplementary Fig. S7). U-Pb ages were determined by Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICP-MS) at the Laboratoire Magmas et Volcans, Clermont-Ferrand, France (Fig. 3a,b and Table 2). All the dated rutile grains contain a low U content of about 20 ppm and no Th, as often observed in rutiles36. A total of 22 spots was performed on 18 rutile crystals (see Supplementary Fig. S7) and yields a discordia line with an upper intercept at 2089 ± 13 Ma (MSWD 0.66, Fig. 3a), the lower intercept being at the origin of the concordia diagram within uncertainties. However, diffusion-induced resetting is unresolvable in LT eclogites36 and rutiles in eclogites provide similar ages than eclogitic zircons37. Consequently, the obtained 2089 Ma upper intercept can be confidently interpreted as the age of the eclogite-facies event. These results evidence the highest pressure so far reported for eclogitic facies metamorphism in the Paleoproterozoic. Fifteen zircons crystals were analysed (sorted and in thin section; see Supplementary Fig. S6) and yielded a discordia line with an upper intercept age of 2216 ± 26 Ma (MSWD 0.2, Fig. 3b), the lower intercept being, as for rutiles, at the origin of the diagram within uncertainties. These sub-euhedral zircons display a weak and irregular zoning and a mean Th/U ratio of 0.4 regardless of the variable amounts of U, rather favouring a magmatic protolith38. A single zircon grain is concordant at 2087 ± 45 Ma (Table 2) and is characterized by a lower Th/U ratio of 0.05. It provides a similar age than the rutiles and may have been dissolved and recrystallized during the eclogite facies event. On the contrary, all the other zircons are interpreted as dating the magmatic stage of the mafic protolith at 2216 Ma.
The 147Sm-143Nd systematic gives a εNdi of +2.04 at 2216 Ma (Fig. 3c and see Supplementary Table S2), i.e. slightly more enriched than the evolution of the depleted MORB mantle (DMM) at ~2 Ga (e.g.39). The TDMM model age for the sample is relatively close, at 2350 Ma (Fig. 3c). The trace element pattern also indicates a source slightly more enriched in incompatible trace elements than the DMM (Table 1 and see Supplementary Fig. S2). This slight enrichment in incompatible elements, combined to enrichment in robust and immobile elements such a Zr, Nb and Y, can be ascribed as T(transitional)-MORB rather than E(enriched)-MORB. Because T-MORB are considered characteristic of a transitional geodynamic tectonic setting between oceanic and continental lithospheres, i.e. rifting and continental breakup40, crustal contamination is expected and could explain the decrease of the εNd from ~+3–4 for the DMM39 to +2.04 as measured in the sample at 2216 Ma. Because of the crustal contamination, the model age at 2350 Ma is a maximum, the metamorphic age at 2089 Ma being the minimum. This is coherent with the age obtained on the zircon discordia line (2216 ± 26 Ma), that is thus interpreted as the true crystallization age.
The 2.09 Ga eclogites of the Nyong complex of Cameroon and the 2.0 Ga eclogites of the Usagaran Belt of Tanzania have a geochemical affinity to oceanic crust and are interpreted to represent the relics of subducted Paleoproterozoic oceanic crust at the margins of the Congo Craton7,8,10. These eclogite occurrences with MORB-like compositions in a continental setting support the hypothesis that plate tectonics operated on Earth in the Paleoproterozoic Era, apparently in a similar fashion as in the modern Earth, since production of the eclogite facies MORB requires the subduction of an old, cold and dense lithosphere (e.g.9,41). Moreover, the RDC eclogite presented here is the first evidence of an entire Wilson cycle in the Paleoproterozoic comprising HP-LT subduction. These eclogites derive from a mafic protolith, with a T-MORB signature, formed at 2216 ± 26 Ma in a intra-cratonic rift-type basin inside the Congo Craton, then buried at high pressure and low temperature (17–23 ± 1 kbar and 500–550 ± 50 °C) and exhumed during a cycle of ca. 130 Ma. These observations evidence a modern-style plate tectonics at 2.2–2.1 Ga. We thus show here that modern-style plate tectonics, as evidenced by cold and deep subduction (>55 km), operated at least since the Paleoproterozoic. Because it certainly took some time of a transient regime from stagnant-lid tectonic42 to mobile-lid subduction, this result is compatible with a major change in Earth’s tectonic regime between 2.5 and 3.0 Ga43. On the other hand, it is difficult to envision how mobile-lid plate tectonics could have started since ~4.5 Ga and left no older compelling imprint, even when considering the incompleteness and preservation bias of the Archean rock record.
The sample was crushed for dating and rutiles and zircons were separated using standard heavy liquids and magnetic techniques. Rutiles (50–500 µm) and zircons (20–100 µm) were hand-picked and mounted in a 1 inch epoxy disc, which was polished to expose the mid-section of grains. Zircon crystals were also found in situ in petrographic thin section (20–50 µm), located in the matrix as well as inside garnets. The internal structures of zircons and rutiles were investigated with backscattered electron (BSE) and cathodoluminescence (CL) images at the University Pierre & Marie Curie, Paris (France).
U-Th-Pb isotope data were measured by laser ablation inductively coupled mass spectrometry (LA-ICP-MS) at LMV (Clermont-Ferrand, France). Zircons were ablated using a Resonetics Resolution M-50 equipped with a 193 nm Excimer laser system coupled to a Thermo Element XR high resolution ICP-MS. Helium carrier gas was supplemented with N2 prior to mixing with Ar for sensitivity enhancement44. The laser was operated with a repetition rate of 3 Hz, a fluence of 3.5 J/cm2 and spot diameters of 15 and 33 µm for zircon and rutile, respectively. The signals of 204Pb (+Hg), 206Pb, 207Pb, 208Pb, 232Th, and 238U were acquired during each analysis45. Background levels were measured on-peak with the laser off for ~30 seconds, followed by ~60 seconds of measurement with the laser firing and then ~30 seconds of washout time. Reduction of raw data was carried out using the GLITTER® software package of Macquarie Research Ltd46. Isotope ratios were corrected for laser-induced and instrumental mass fractionation via sample-standard bracketing using the GJ-1 zircon47 and Sugluk-4 rutile48 reference materials. Concentrations of U, Th, and Pb were calculated by normalization to the certified composition of GJ-146 and 9150049. Data were not corrected for common Pb. Concordia diagrams were generated for each sample using the Isoplot/Ex v. 2.49 software of 50. Error ellipses for each point are shown at the 2σ level and incorporate both internal and external uncertainties. The 91500 zircon and PCA-207 rutile48 were analysed along with the samples, to independently monitor the external precision and accuracy of the measurements. The pooled ages for 38 analyses of 91500 and 27 analyses of PCA-S207 conducted over the course of the study were 1064.9 ± 4.5 Ma and 1862.2 ± 6.9 Ma, respectively.
Around 50 mg of powdered sample was mixed with 1 g of ultrapure lithium metaborate and tetraborate (4:1). After heating at 1000 °C, the bead was re-dissolved in 50 ml of HNO3 5% plus traces of HF. After ad-hoc dilution, the sample was measured on the Agilent 7700 ICP-MS at ULB, Belgium. BHVO standard was used to ensure the precision and reproducibility of the measurements, which was better than 5% for the elements presented here.
After crushing, ~200 mg of powder have been dissolved in a mixture of ultrapure HF:HNO3 (1:3). After removing the supernatant, the solid residue has been re-dissolved by the same but fresh mixture in high-pressure vessels to ensure a complete dissolution of refractory phases such as zircon. The two fractions were recombined and after evaporation, HCl 6 N was added. Once the solution was clear, a small aliquot was taken and spiked with a mixed 150Sm-148Nd spike and the mixture was equilibrated 24 h on hotplate. Both unspiked and spiked aliquots were purified on a cationic resin by rinsing in 1.5 N HCl and collecting REE in 6 N HCl. Then, REE were purified from each other using home-made HDEHP resins. Both spiked and unspiked cuts have been measured on the HR-MC-ICP-MS Nu-Plasma 1 at ULB, Belgium. For the unspiked cut, the value was corrected for mass fractionation by using the ratio 146Nd/144Nd = 0.7219, and then for the accepted Rennes Nd standard value 143Nd/144Nd of 0.511963. The internal total reproducibility (n = 9) was better than 22 ppm. The measurement was replicated and values are well-within errors. For the spiked cut, mass fractionation was calculated by iterative calculation as in51.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank the Royal Museum for Central Africa (RMCA, Tervuren/Belgium) for access to the sample, Omar Boudouma (UPMC, Paris) for SEM and CL imagery, Michel Fialin (Camparis, Paris) for Electron Microprobe analyses, Sabrina Cauchies (ULB) for Nd data, and Marcella Giraldo (ULiege) and Bernard Charlier (ULiege) for help with minerals sorting. This paper has benefited from discussions with G. Godard (IPGParis), J.P. Liégeois (RMCA), M. Fernandez-Alonso (RMCA) and B.K. Baludikay (ULiege). Research funding came from the European Research Council Stg ELITE FP7/308074, and the Francqui Foundation (EJJ, CF), the BELSPO IAP PLANET TOPERS (EJJ, CF, VD) and the FNRS-FRS and the ERC StG “ISoSyC” FP7/336718 (VD).