No evidence of supracrustal recycling in Si-O isotopes of Earth’s oldest rocks 4 Ga ago

Abstract

Identifying the oldest evidence for the recycling of hydrated crust into magma on Earth is important because it is most effectively achieved by subduction. However, given the sparse geological record of early Earth, the timing of first supracrustal recycling is controversial. Silicon and oxygen isotopes have been used as indicators of crustal evolution on Archean igneous rocks and minerals to trace supracrustal recycling but with variable results. We present Si-O isotopes of Earth’s oldest rocks [4.0 billion years ago (Ga)] from the Acasta Gneiss Complex, northwest Canada, obtained using multiple techniques applied to zircon, quartz, and whole rock samples. Undisturbed zircon is considered the most reliable recorder of primary Si signatures. By combining reliable Si isotope data from the Acasta samples with filtered data from Archean rocks globally, we observe that widespread evidence for a heavy Si signature is recorded since 3.8 Ga, marking the earliest record of surface silicon recycling.

No surficial Si is detected in the oldest rocks until ca. 3.8 Ga; thus, supracrustal recycling was not evident before this time.

INTRODUCTION

The recycling of surface materials via crustal dynamics, i.e., supracrustal recycling, depends on a planet or moon’s geodynamic mode. As one end-member, a stagnant-lid regime does not involve supracrustal materials in convection (1). At the other end of the spectrum, a mobile-lid regime, i.e., plate tectonics, involves continuous surface recycling in the form of subduction (2). The operation of plate tectonics makes planet Earth unique within the solar system (3). However, the key questions of whether early Earth was uniformly characterized by a mobile-lid regime (4, 5) or whether its geodynamics might have changed from a stagnant-lid to a mobile-lid regime (6–9) remain hotly debated. Isotope analysis is an effective means of detecting the transportation of surface materials into regions of melting; thus, testing Earth’s oldest known rocks should provide clues about early geodynamic modes.

Silicon has three stable isotopes, ²⁸Si, ²⁹Si, and ³⁰Si, and different isotopic ratios of juvenile and evolved rocks enable the detection of surface materials in magmatic sources (10, 11). Although oceans are argued to have been present as far back as 4.4 Ga (12), the lack of large-scale skeletal biosilicification on early Earth suggests that the concentration of dissolved Si in Archean seawater was at least one order of magnitude higher than in Phanerozoic oceans (13, 14). Thus, seafloor silicification would have been a common phenomenon in the Archean (15, 16). As a consequence, Archean silicified seafloor would have had a Si isotopic composition {expressed as δ³⁰Si (‰) = [(³⁰Si/²⁸Si_sample)/(³⁰Si/²⁸Si_{NBS28 standard}) − 1] × 1000} considerably heavier than the modern (Phanerozoic) oceanic crust (17). Incorporation of such materials into source regions where partial melting occurs should produce Archean granitoids with heavy Si signatures (11, 17, 18). More information on Si isotope fractionation is provided in the Supplementary Materials. O isotopes [δ¹⁸O (per mil, ‰)] as another tracer of crustal evolution have also been widely applied in Earth science (19), but limited paired Si-O isotopic studies have been conducted on Archean granitic rocks. Combined Si and O isotope analyses may provide further insight into the origin of early Archean rocks. For example, the addition of chemical weathering products (desilication) and silica precipitation to the magma sources may result in similar elevated O isotope signatures but divergent Si signatures (11).

Improvements in analytical techniques over recent years have seen increasing use of Si isotopes as a tracer of crustal evolution (11, 17, 18, 20). By analyzing whole rock samples and quartz grains, previous studies have reported Archean granitic rocks [3.8 to 2.7 billion years ago (Ga)] with systematically heavy Si isotopic compositions and no light Si signatures being detected, requiring the presence of surface-derived materials in the source regions of their parent magmas (17, 18). However, Earth’s oldest rocks are highly metamorphosed. Although Si isotopes are likely resistant to metamorphism (21), it is difficult to determine whether the whole rock samples remained a closed Si system and whether the Si isotope composition of quartz grains is primary or was modified by subsequent fluid-rock interaction. Zircon has a high resistance to weathering (22), and a few studies have reported zircon Si isotopes of samples from both Archean granitoids (ca. 4.0 to 2.7 Ga) (20)) and Jack Hills detrital zircon (ca. 4.2 to 3.3 Ga) (11). These zircon samples display highly scattered Si signatures from light to heavy with no apparent variation through time. Because igneous rocks exhibit very limited Si isotope fractionation (23), any possible temporal transitions may be hidden by the relatively large analytical errors of these pioneering in situ analyses. As a result, the degree of recycling of fractionated surface materials into the Earth’s interior through time is poorly understood.

Trondhjemite-tonalite-granodiorite rocks (TTGs), as the dominant components of the Archean crust, are critical to understanding the evolution of early Earth (24). To help resolve the aforementioned problems, this study focuses on TTGs (4.02 to 3.57 Ga) from the Acasta Gneiss Complex, Northwest Territories, Canada, including samples from the oldest rock unit (4.0 Ga), the Idiwhaa tonalite (fig. S1) (25, 26). The oldest Idiwhaa unit of the Acasta (ca. 4.0 Ga) displays bimodal zircon O isotopic values that trend both slightly higher and lower than the mantle zircon field (25). It likely formed in an Iceland-like setting associated with supracrustal materials (25). However, inferring geodynamic settings from any one isotopic system alone may provide biased information. To further investigate the magma sources of the oldest known TTGs, we present paired Si-O isotopic data of these Acasta samples. By using multiple techniques, a variety of proxies, and careful evaluation of the data (see Materials and Methods), we conduct ultrahigh-precision in situ secondary ion mass spectrometry (SIMS) and solution multicollection inductively coupled MS (S-MC-ICP-MS) analyses of zircon, quartz, and whole rock samples to constrain the primary Si signature of the protoliths. We then use these data to trace the origin of the oldest known evolved rocks.

RESULTS

Whole rock geochemistry and zircon U-Pb data

In terms of their normative proportions of quartz and feldspar, all six Acasta samples classify as granodiorite or tonalite, with SiO₂ ranging from 61 to 70 wt % (figs. S2 and S3 and data file S1). However, the four ca. 4.0-Ga samples are not typical TTGs, as shown by their high Fe and Zr contents and high Fe/Mg ratio (fig. S3 and data file S1). The four ca. 4.0-Ga samples have similar weighted mean zircon ²⁰⁷Pb/²⁰⁶Pb ages of 3955 ± 25, 3988 ± 8, 4003 ± 23, and 3982 ± 6 million years ago (Ma; uncertainty at 95% confidence; fig. S4 and data file S2). The two other samples, with significantly lower Fe and Zr contents and Fe/Mg ratios, have much younger ages at 3574 ± 17 and 3723 ± 6 Ma (figs. S3 and S4). Representative cathodoluminescence images of the analyzed zircon grains are shown in fig. S5.

In situ zircon Si isotopes

The zircon grains selected for silicon isotopic analysis have a U-Pb age discordance ratio between −8.0 and 0.5%, U abundances of 118 to 824 parts per million, and Th/U ratio of 0.10 to 0.94. Zircon δ³⁰Si_zrc values of the ca. 4.0-Ga samples range from −0.43 ± 0.07 to −0.59 ± 0.09‰ (2 SE), and the younger Acasta samples (3.7 to 3.6 Ga) are from −0.20 ± 0.07 to −0.35 ± 0.05‰ (2 SE) (Fig. 1A, table S1, and data file S3). On the basis of the statistics of this study, 84% of less representative zircon grains have lower δ³⁰Si_zrc than well-crystallized grains; when U-Pb age discordance is greater than 10%, metamictization is intermediate or strong or cracks are present (more details are provided in table S1 and Materials and Methods).

Fig. 1. Zircon U-Pb ages versus δ³⁰Si on early Earth.

(A) Zircon ages versus Si isotopes. Literature data are from Canada (Acasta, Slave craton), Siberia, Antarctica, Greenland, and Western Australia (Jack Hills) (11, 20). (B) Sample ages versus quartz Si isotopes. Literature data (17) are from Siberia, Norway, Finland, Canada (Slave and Superior cratons), and South Africa. (C) Sample ages versus whole-rock Si isotopes. Literature data in (B) and (C) are from South Africa (Kaapvaal craton), Siberia (Onot terrane), Norway (Varanger Complex), Canada (Slave and Superior cratons), Finland (Naavala), and Greenland (southern part of the Isukasia terrane) (17, 18, 21). Except for the Jack Hills detrital zircon, all the data plotted here are from granitic rocks, including trondhjemite-tonalite-granodiorite rock (TTG), granite, and syenite (details are provided in data file S4). Qtz, quartz; WR, whole rock; MC, solution multicollection inductively coupled mass spectrometry (S-MC-ICP-MS) method; ‰, per mil. The gray field represents the time period of the statistically significant Si isotopic shift. Average isotopic values (solid lines) with 2σ uncertainties (dashed lines) consistently display a transition at 3.7 to 3.8 billion years ago (Ga; see Materials and Methods for statistical information). Jack Hills zircon data were not included in the statistical analysis of (A) (see text for details), and outliers of Greenland samples were not included in (C), although the change point remains significant if they are included. Because zircon, quartz, and whole rock data are from different samples and cratons at 3.9 to 3.8 Ga, the step change varies among the panels at ca. 3.8 Ga (the gray field).

Quartz and whole-rock Si isotopes

In situ quartz δ³⁰Si_qtz of the ca. 4.0-Ga samples is between −0.09 ± 0.1 and −0.14 ± 0.09‰ (2 SD). The younger samples have higher δ³⁰Si_qtz ranging from 0.06 ± 0.09 to 0.07 ± 0.05‰ (2 SD) (Fig. 1B, table S2, and data file S3). S-MC-ICP-MS analysis of quartz grains from the ca. 4.0-Ga samples gives δ³⁰Si_qtz values ranging from −0.06 ± 0.06 to −0.19 ± 0.03‰ (2 SD), consistent with in situ SIMS analyses (fig. S6). The whole-rock δ³⁰Si_wr compositions of the ca. 4.0-Ga samples range from −0.21 ± 0.05 to −0.28 ± 0.05‰ (2 SD) (Fig. 1C, table S3, and data file S3).

In situ zircon and quartz O isotopes

Zircon δ¹⁸O_zrc values of ca. 4.0-Ga samples range from 4.75 ± 0.18 to 6.31 ± 0.18‰ (2 SE), with zircon cores at 4.93 to 6.31‰ and rims at 4.75 to 5.25‰ (Fig. 2, table S1, and data file S3). Younger samples have zircon δ¹⁸O_zrc ranging from 5.68 ± 0.19 to 7.04 ± 0.15‰ (2 SE). In situ quartz δ¹⁸O_qtz values of the ca. 4.0-Ga samples are from 9.77 ± 0.31 to 10.35 ± 0.32‰ (2 SD), and the younger samples are from 9.54 ± 0.24 to 9.57 ± 0.36‰ (2 SD) (table S2 and data file S3).

Fig. 2. Zircon Si-O isotopes on early Earth.

Acasta granitoid data are from this study. Data in red are the most representative, and data in pink may show weak Si mobilization. Jack Hills zircon data, mantle zircon Si isotope data (−0.38 ± 0.04‰, 2 SD), and the three isotopic trajectories are from (11). Mantle zircon O isotope value (5.3 ± 0.6‰, 2 SD) is from (19). The solid lines represent isotopic values for mantle zircon, and the dashed lines are the corresponding errors. The silicified seafloor source was primarily proposed from (18).

Zircon Raman spectra

Zircon grains 19AC08@21; 19AC10@10, AC10@49, AC10@64, AC10@15, and AC10@34; 19AC33@29 and AC33@54; and 19AC23@31 and AC23@37 exhibit Raman spectra with a full width at half maximum (FWHM) of <15 cm⁻¹ and a primary peak of >997 cm⁻¹, which demonstrates weak to no metamictization in the sections where analyzed (fig. S7 and table S1). Grains such as 19AC10@4 and AC10@24, 19AC12@20, 19AC13@48, and 19AC33@49 with an FWHM of >15 cm⁻¹ or a primary peak of <997 cm⁻¹ were considered to be partly metamict zircon. Grains with the primary peak barely visible were interpreted as strongly metamict zircon, which was used as a guide for interpreting the zircon Si isotopes.

DISCUSSION

Zircon Si isotopes

If the U-Pb system in zircon has not been disturbed, then zircon Si likely remains a closed system, as constrained by the relative diffusivities (D) of these elements in zircon (D_Th,U << D_Si << D_Pb), with differences of approximately two orders of magnitude (27, 28). Thus, only grains with close to concordant U-Pb ages (discordance of <10%) are regarded as providing a reliable Si isotope signature of the magmatic system in which they crystallized. The detailed screening criteria (e.g., metamictization and cracks) are discussed in Materials and Methods, and the filtered data are summarized in table S1.

Filtered zircon Si isotopes define two distinct groups. The most representative zircon grains of ca. 4.0-Ga samples have δ³⁰Si_zrc values (−0.43 to −0.46‰) within error of mantle zircon [−0.38‰ ± 0.04 (11)] (Figs. 1A and 2), indicating minimal involvement of surface-derived silica in their petrogenesis. The result is consistent with the Acasta zircon (ca. 4.0-Ga) Si isotopes analyzed in (20), when using the most concordant grains only, but that study also revealed a few grains at ca. 3.95 Ga with slightly lower δ³⁰Si_zrc of around −0.5‰ (Fig. 1A). The younger samples (3.7 to 3.6 Ga) show higher δ³⁰Si_zrc values (−0.24 to −0.26‰) above mantle zircon field (Figs. 1A and 2). These different δ³⁰Si_zrc signatures correspond to the change in the whole rock composition from the older to younger samples, as noted above. Furthermore, the Acasta Ti isotopes also display a shift from tholeiitic- to calc-alkaline–style magmatism at ca. 3.8 Ga (29) that is coincident with the overall Si isotope shift that we report. Thus, the heavy zircon Si isotopic signature of the younger rocks likely indicates the involvement of silicified basaltic crust and/or chemical sediments (e.g., chert) in the petrogenesis of their protoliths (11, 17, 18), where both processes can yield heavy Si isotopes.

When compiling the filtered granitic zircon Si isotope data from other Archean cratons (Napier, Siberia, and Greenland; data file S4) (20), we observe that the earliest robust evidence of heavy Si isotope signature from granitic rock samples, recording the recycling of surface materials into magmatic sources, occurs around 3.8 to 3.7 Ga (Fig. 1A). The Jack Hills, Western Australia, detrital zircon of Hadean-Eoarchean age has scattered Si isotope signatures (11), but with large analytical errors, so it is unclear whether the scatter represents the real range of values (Figs. 1A and 2). The Jack Hills data overlap the more precise data presented here, and consequently, we do not discuss those previous data further. A comparison of analytical errors between this study and the literature data is provided in data file S3.

Paired zircon Si-O isotopes

As a paired constraint with the Si isotopic composition, O isotopes in zircon display a small increase from the older (ca. 4.0 Ga) to younger samples, with δ¹⁸O_zrc rising from 4.75 to 6.31 to 5.68 to 7.04‰ (Fig. 2 and table S1). On the basis of the different sensitivities of Si and O isotopes to various weathering and alteration conditions, three trends are evident for paired Si-O isotopes (Fig. 2) (11), according to which the ca. 4.0-Ga samples plot mostly within error of the “mantle” signature (Fig. 2). The ca. 4.0-Ga samples show a slight increase in δ¹⁸O_zrc to above mantle values, with a constant mantle Si isotopic signature (Fig. 2), which may indicate that the source magmas involved slightly serpentinized materials [i.e., altered but not by surface processes/fluids (30); see more information in the Supplementary Materials]. The paired zircon Si-O isotope data indicate that the ca. 4.0-Ga samples incorporated no obvious supracrustal materials. Zircon O isotopes of the younger samples coincide with the trajectory of Si precipitation (Fig. 2). This observation demonstrates that the distinct increase of Si-O isotope values in the younger samples is likely due to the input of silica initially precipitated from seawater (11) and/or silicified basalt (17, 18).

A previous study reported zircon rims with ages indistinguishable from the cores at ca. 4.0 Ga where the rims have slightly lower average δ¹⁸O values (4.7‰) than the cores (5.6‰), which are both within error of the mantle zircon field (25). It was suggested that the lighter O isotopic signature of the zircon rims was caused by high-temperature hydrothermal alteration interacting with surface waters in the magma source (25). Our study also yielded δ¹⁸O values on both zircon rims and cores consistent with those of the previous study (table S1), but we provide an alternative interpretation of the lower δ¹⁸O values of the rims. Because the ca. 4.0-Ga Acasta samples have high Fe and Zr contents, we suggest that a ferrogabbro-type fractional crystallization process may account for ∼0.5‰ δ¹⁸O depletion (31, 32), which is broadly consistent with the offset of the slightly low δ¹⁸O rim values from those of the cores. For example, the Fe-rich plagiogranites reported in (31) have the same slightly lighter O isotopes on zircon rims compared to their cores. Another example of Fe-rich granite with nearly the same temperature as the Acasta samples at ~800°C also has light O isotopes but no evidence of hydrothermal alteration (32). Thus, it is possible that the light O isotopes on 4.0-Ga Acasta zircon rims are caused by the Fe-rich magmatic affinity, which does not require the involvement of surface waters in the magma. The Si isotopes of zircon rims also do not show a significant supracrustal material signature. Although δ³⁰Si_zrc of rims have been affected by metamictization and cracks with slightly lower values (Fig. 2 and table S1), zircon rims likely have a mantle Si signature similar to the cores according to the statistics of this study, with ~84% of zircon displaying lower δ³⁰Si_zrc values after secondary processes such as metamictization (table S1). This explanation is also supported by the quartz and whole-rock Si isotopes discussed in the following sections.

Quartz Si and O isotopes

The in situ Si isotope compositions of quartz from the six samples also exhibit two distinct groups, and the δ³⁰Si_qrz exhibits a significant increase from the ca. 4.0-Ga samples (−0.14 to −0.09‰) to the younger samples (0.07 to 0.06‰) (Fig. 1B). The in situ O isotope compositions of quartz slightly decrease from the ca. 4.0-Ga (10.35‰) to the younger samples (9.54‰; table S2). However, photomicrographs reveal that all of the Acasta gneisses exhibit recrystallized quartz grains (fig. S2); thus, there is no compelling evidence that the quartz Si isotope data represent primary signatures. Because the zircon Si isotope data have been carefully evaluated, we applied Si fractionation between quartz and zircon to test the credibility of the quartz Si isotope data. Si and O isotope fractionation in the quartz-zircon system can be applied to calculate equilibrium temperatures, which are reported as T_Si and T_O, respectively. The temperatures are determined through first-principle calculations based on the fractionation of Δ³⁰Si(qtz-zrc) and Δ¹⁸O(qtz-zrc) according to (33). Calculations using other methods are also presented in the Supplementary Materials. The T_Si of the most representative samples ranges from 737° to 855°C, which is consistent with their zircon saturation temperatures (T_zrc) of 753° to 889°C (Fig. 3A, fig. S8, and table S4). The T_Si is also within error of the zircon Ti temperature (T_Ti) of 799° to 933°C obtained from (20). In contrast, the O equilibrium temperature (T_O) is 419° to 687°C, which is considerably lower than T_Si, T_zrc, and T_Ti. Thus, we conclude that the quartz was likely in a closed Si system and the Si isotope values represent the primary Si isotope signature, whereas the quartz O isotopes might have been disturbed during metamorphism. When we combine the quartz data from granitic samples from other Archean cratons, including samples from the Nuvvuagittuq belt of the Superior craton that fill the gap at 3.8 Ga (17), the Si isotope values in quartz appear to be largely coupled with the Si isotope variation in zircon, and they also display a shift to heavy isotope signatures at ca. 3.8 Ga (Fig. 1B).

Fig. 3. Si-O isotope fractionation.

(A) Comparison of zircon saturation temperature (T_zrc) with Si-O isotope equilibrium temperature in the quartz-zircon system. The equilibrium temperature equation is y = 6.5516x + 0.0278, summarized after first-principle calculations from (33). (B) Si isotope fractionations among whole rock, quartz, and zircon. Acasta samples are from this study (red, pink, and blue data). The Jindabyne granite (Lachlan orogen) data are from (11). Numbers marked in blue are fractionations between quartz and whole rock, measured on S-MC-ICP-MS. Dashed lines are used to visually separate quartz, whole rock, and zircon data to better understand the Si fractionations between the minerals and whole rock. Qtz, quartz; WR, whole rock; Zrc, zircon.

Whole-rock Si isotope data

Si isotope fractionation between quartz/zircon and whole rock samples can be applied to evaluate the fidelity of whole-rock Si isotope compositions, because both filtered zircon and quartz grains analyzed from the Acasta samples record the primary Si isotope signature, as discussed above. Two samples (19AC08 and 19AC13) have larger fractionations with ΔSi(qtz-wr) at 0.18 and 0.14, respectively, compared with the theoretical value of 0.05 (Fig. 3B and table S5). Other samples have fractionations within error of their theoretical calculated values similar to the Phanerozoic Jindabyne granite (11). The higher degree of fractionation might be caused by the alteration of the samples evidenced by slightly altered feldspar (fig. S2). Alternatively, the whole-rock Si isotope system has been disturbed during metamorphism. Consequently, the whole-rock δ³⁰Si_wr of the two samples is slightly lower than the primary signature according to their zircon and quartz Si isotopes. Thus, whole-rock Si isotopes should also be carefully assessed before any further interpretation. Filtered whole-rock Si isotope values from the ca. 4.0-Ga Acasta samples, together with younger granitoids (≤3.8 Ga) from other cratons, give a similar trend to those of quartz and zircon, although samples from the Isua region, Greenland have distinctly high δ³⁰Si_wr at 3.8 Ga (Fig. 1C). The Si isotope shift at 3.8 Ga is also consistent with sulfur isotopes of samples from Isua that suggest the recycling of sedimentary and hydrothermal materials into magmatic sources by this time (34).

Earliest Si isotopic shift and implications

During magmatic differentiation, δ³⁰Si increases linearly from mafic to felsic end-members, expressed as an “igneous array” (Fig. 4A) (35). When the filtered zircon, quartz, and whole-rock Si isotope data of extant Hadean-Archean granitoids from six cratons are plotted with the igneous array (Fig. 4A) or simply plotted together (fig. S9), they display a clear shift to heavy Si isotopes at ca. 3.8 Ga. Although the age of the step change varies among the zircon, quartz, and whole-rock Si isotope datasets due to the limited data during the 3.9- to 3.8-Ga interval, an overall shift at this time is clearly demonstrated (the gray field in Fig. 1). It is consistent with the O isotopic shift presented in this study and in (36).

Fig. 4. Whole-rock δ³⁰Si_wr versus SiO₂ and models of geodynamic evolution.

(A) Zircon and quartz δ³⁰Si values were converted to the whole rock system based on their zircon saturation temperatures, SiO₂ contents, and the corresponding fractionation factors (table S5). The sources of literature data are provided in Fig. 1, except that a few samples with no available SiO₂ data are not included here. The equation for the igneous array is δ³⁰Si_wr (‰) = 0.0056 × SiO₂ (wt %) − 0.567 (23). Samples with Si mobilization (pink marks in other figures) are not included here. (B) The ca. 4.0-Ga Acasta suite could be formed by a process with minimal involvement of surficial silicon, whereas all other Archean samples (≤3.8 Ga) reflect the involvement of surficial silicon, most likely pointing to convergent plate boundary processes capable of surface recycling in their genesis.

Recent geodynamic models suggest that efficient crustal recycling was already operational in the Hadean (5, 37). Although an impact-induced melt origin for the Idiwhaa tonalite cannot be definitively rule out (38), it is arguably inconsistent with the absence of a surface Si-O isotope signature in these rocks. A plume setting proposed from (25, 29) works consistently with both the Si-O isotopes and the petrogenesis of the ca. 4.0-Ga granitoids. These samples have higher total Fe, Zr, Ba, and Fe/Mg than the younger Acasta samples and show more similarities with ca. 3.64-Ga Fe-rich granitoids with associated ferrodiorites from the Itsaq Gneiss Complex of southwest Greenland (fig. S3), which do not require a subduction setting (39). Our results and literature data show that the granitoids with lower Fe/Mg and heavier Si isotopes are first recognized at ca. 3.8 Ga, indicating that by then, surface materials were being transported back into the magmatic source at this time (Fig. 4B). Although the small variation of seawater Si isotopic composition that is observed in Eoarchean marine sedimentary rocks (40) may also contribute to the isotopic shift documented here, geodynamic changes coupled with petrogenesis are more likely to account for the shift.

Globally, diverse proxies applied in early Earth studies, including both geological observations (41–44) and robust stable isotope/trace element data (e.g., Ca, Hf, Ti, S, Cl, and Al compositions), consistently indicate a geodynamic change took place before ca. 3.8 to 3.7 Ga (table S6) (9, 29, 34, 45–47) or even earlier at ca. 3.9 Ga (48). For example, by combining the Acasta and global Hf isotope data, zircon ¹⁷⁶Hf/¹⁷⁷Hf shifted to higher values in each craton at ca. 3.8 Ga, requiring the input of a juvenile source on a global scale at this time, which is suggested to most likely have been achieved by a transition of tectonic regime from stagnant-lid to mobile-lid (9). The Ti isotopes of Acasta samples also record a magmatic contrast of 4.0-Ga tholeiitic to 3.8-Ga calc-alkaline, which is likely due to a contrasting plume-like versus subduction-like or sagduction settings (29). The Al composition of the Jack Hills zircon displays a notable increase at ca. 3.8 Ga, indicating that more zircon grains younger than this age were likely generated from peraluminous crustal magma and/or during a moderate high-pressure fractionation process under horizontal tectonics (46). Placing a geodynamic transition before the earliest Archean is also supported by other relevant events such as the emergence of weathering and crustal rejuvenation by this time (49–52). Si and O isotopes are particularly sensitive to surface material recycling (11). Thus, the Si-O isotope shift also supports a fundamental change in geodynamic setting. Nevertheless, although subduction at 3.8 Ga is the most effective way to transport the surficial materials back to the magma sources and to rapidly generate new crust (9), other nonsubduction models have been proposed for the generation of Archean TTGs, such as dripping of buried surface materials or continuous crustal delamination [e.g., (53)].

We emphasize that the pre–3.8-Ga data come from a single Acasta locality, and hence the absence of a heavy Si isotopic signature detected in an individual locality of ca. 4.0-Ga granitoids cannot preclude magmatic rocks with a different Si isotopic signature formed elsewhere in the Hadean. Nevertheless, on the basis of our current database, we conclude that the Si-O isotope data indicate no evidence of supracrustal recycling in the oldest known rocks at ca. 4.0 Ga and the oldest evidence for the recycling of supracrustal materials into magmas on Earth is at least from ca. 3.8 Ga.

MATERIALS AND METHODS

Samples

The Acasta Gneiss Complex is a Hadean-Eoarchean terrane, which hosts the oldest rocks (ca. 4.03 to 3.40 Ga) on Earth, and is exposed in the western margin of the Slave craton in the Northwest Territories, Canada (fig. S1) (54, 55). The local region has been divided into two simplified lithological assemblages, with high-strain banded gneisses in the west and low-strain felsic to intermediate gneisses in the east (36, 55). Because this study is designed to detect early surface material recycling, we focus on the granitic gneisses of the eastern part with ages ranging from 4.03 to 3.59 Ga (55), including the oldest known rock unit, the Idiwhaa tonalite (25). Six representative samples were selected for whole-rock geochemistry and zircon geochronology analyses to confirm the general properties and to select the most concordant zircon grains for Si isotope analysis. Raman spectra were used to filter out the strongly metamict zircon samples. Undisturbed zircon and quartz grains were each analyzed for Si-O isotopes using high-precision in situ analysis. Subsequently, Si-O isotope equilibrium temperature in the zircon-quartz system can be evaluated to assess whether the recrystallized quartz Si-O isotopes represent the initial geological conditions or have been affected by recrystallization. Four representative samples were also selected for both quartz and whole-rock solution analyses to assess the consistency between different methods and evaluate the whole-rock Si isotopic data by considering Si fractionation. Detailed information for the sample sites can be found in fig. S1C and data file S1, and sample descriptions are provided in the Supplementary Materials.

Zircon Si-O isotope screening criteria

As discussed here, the likelihood of Si isotope mobility in zircon can be assessed using the concordance of the U-Pb ages, because the relative diffusivities in zircon are D_Th,U << D_Si << D_Pb in zircon (27, 28). If the zircon U-Pb system has not been disturbed (e.g., no evidence of Pb loss), then zircon Si isotopes are more likely to be immobile and represent primary signatures. Thus, as the primary screening criterion, we use a discordance of <10% in the U-Pb age to define representative grains. Considering the long geological history of Hadean-Eoarchean zircon, radiation damage may result in varying degrees of metamictization (56, 57), which potentially could cause disturbance of zircon Si isotope signatures by open-system behavior of Si (20, 58). Compared with well-crystallized zircon grains, the zircon Raman spectra and Si-O isotope data show that a low degree of metamictization has no obvious influence on zircon Si-O isotope signatures, but that intermediate to strong metamictization causes most zircon δ³⁰Si_zrc to decrease, but without any apparent effect on zircon O isotopes (fig. S7). Therefore, as a second filter, we select grains with weak to no metamictization. Last, abnormal silicon yield may be caused by microinclusions of other silicates in zircon. To avoid this problem, we only use grains with Si yields within the error of standard zircon yields (fig. S10A). Other potential causes for Si mobilization, such as high U content, abnormal Th/U, and cracks are all considered in this study (fig. S10, B and C).

On the basis of zircon U content and experimental data for radiation damage, most Archean zircon grains would have been metamict (56, 57). In this study, the most representative zircon samples have weak to no metamictization evidenced by the Raman spectra (fig. S7), indicating that these grains were probably annealed at some point. Because the Acasta gneisses were metamorphosed up to amphibolite facies (650° to 750°C; fig. S2), indicating no higher degree of annealing, and the selected zircon grains have concordant U-Pb ages, these grains might have been annealed before strong metamictization (57). Thus, the zircon O isotopes with a relatively small range of δ¹⁸O_zrc values very likely represent the primary signature of the gneisses (Fig. 2).

Whole-rock major and trace element analysis

Samples were crushed into fine powder for x-ray fluorescence major element analysis at Wuhan Sample Solution Analytical Technology Co. Ltd. (WSSAT). Trace element analyses were performed on an ICP-MS (Agilent 7700e) also at WSSAT. Further details are provided in the Supplementary Materials.

SIMS zircon U-Pb analysis

Zircon U-Th-Pb analyses were undertaken on a CAMECA instrument mass fractionation factor (IMS) 1280-HR SIMS at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS) following the analytical protocols from (59). The primary oxygen ion beam intensity was set at ~12 nA with an analysis spot size of ~20 by 30 μm. To separate Pb⁺ peaks from isobaric interferences, the secondary ion beam optics had a mass resolution of ~7000 at 50% peak height. Measured ²⁰⁶Pb/²³⁸U data and U-Th abundances of unknown zircon were calibrated using the reference material zircon 91500 (1065 Ma) (60). A secondary reference material Temora 2 (417 Ma) (61) was used to monitor data quality. The results for Temora 2 are within the error of accepted values (data file S2). Common Pb was corrected using measured ²⁰⁴Pb and model common Pb composition (62). According to a long-term observation of measured ²⁰⁶Pb/²³⁸U in the laboratory, 1.5% (1 relative SD) uncertainty of standards was propagated in quadrature to the ²⁰⁶Pb/²³⁸U uncertainties in the unknowns (63). Data were assessed using Isoplot 4.16 (64). Individual analyses are presented at 1σ uncertainty.

SIMS in situ Si isotope analysis of zircon and quartz grains

Zircon and quartz Si isotope analyses were conducted on a CAMECA IMS 1280 SIMS at IGGCAS following the analytical protocol from (65). Analytical conditions are given in data file S3. Both zircon and quartz mounts were coated with pure gold at a thickness of ~50 nm to reduce the influence of charge accumulation. Reference materials for zircon Si isotopes were Mud Tank (δ³⁰Si_zrc = −0.34‰) (11), 91500 (δ³⁰Si_zrc = −0.36‰) (66), and Penglai zircon (δ³⁰Si_zrc = −0.34‰) (67). Uncertainties of individual zircon Si isotope data are presented as 2 SE. The quartz Si isotopic data were calibrated using a newly developed reference material GLASS (δ³⁰Si_glass = −0.10‰) (67). NBS28 (δ³⁰Si_NBS28 = 0‰) (68), and Qinghu quartz grains (δ³⁰Si_qtz = −0.03‰) (67) were used as secondary reference materials. The results of all standard reference materials agree well with the recommended values within analytical errors (data file S5). Uncertainties of quartz δ³⁰Si_qtz are presented as 2 SD for each sample.

SIMS in situ O isotope analysis of zircon and quartz grains

Zircon and quartz O isotopes were measured on a CAMECA IMS 1280 SIMS at IGGCAS following standard protocols (69). The Cs⁺ primary ion beam intensity is set as ~2 to 3.5 nA with an analysis spot size of ~25 μm in diameter (≤1 nA, ~10 μm for zircon rims). The IMS was calibrated using primary reference zircon Penglai (δ¹⁸O_zrc = 5.3‰) (70). Secondary reference zircon 91500, Temora, and Qinghu were applied to test the accuracy. The reference materials for quartz O isotopes are Qinghu (δ¹⁸O_qtz = 8.49‰) (69), GLASS (δ¹⁸O_glass = 1.68‰) (69), and NBS28 (δ¹⁸O_NBS28 = 9.6‰) (71). Results of reference materials are within error of accepted values (data file S5).

S-MC-ICP-MS Si isotope analysis of quartz and whole rock samples

Pure quartz grains and whole rock powders were digested using the alkali fusion method, and Si was purified through cation exchange resin (200 to 400 mesh; AG50W-X12, Bio-Rad) at the University of Science and Technology of China following the protocol from (72). Silicon isotopes were analyzed using an MC-ICP-MS (Neptune 3 Plus, Thermo Fisher Scientific). Reference material NBS28 was used for instrument mass bias correction, and Hawaiian Volcano Observatory Basalt (BHVO-2) (73) and Guano Valley Andesite (AGV-2) (35) were analyzed as secondary standards. The results of the reference materials are within the error of the recommended values (data file S3) (35, 73).

Zircon Raman spectra

Zircon Raman spectra were recorded on a confocal Raman microscope WITec alpha 300R at IGGCAS. The Raman data were acquired using WITec Project 5.3 plus software. The laser power was set at 5 mW with an excitation wavelength of 532 nm. The grating setting is 300 g/mm with a blaze wavelength of 500 nm. The laser beam was focused to <1 μm or <500 nm using a 50× ZEISS objective with a numerical aperture (NA) of 0.75 or a 100×/0.9 (NA) objective, respectively. The integration time was 1 to 3 s, and the number of accumulations was 20 to 30.

Statistical change-point analysis

To statistically evaluate the possibility of shifts in the variability of the Si isotope data, we use a Bayesian change-point algorithm (conjugate partitioned recursion) (74, 75) that reveals a consistent transition at 3.8 to 3.7 Ga (Fig. 1). The conjugate partitioned recursion algorithm uses a strategy of binary partitioning by marginal likelihood, combined with conjugate priors. The algorithm first calculates the marginal likelihood for both a no-change model and a multiple change-point model to identify whether a change point is appropriate. If the marginal likelihood favors multiple change-point models, then the algorithm defines the change points and 2σ uncertainty bounds (74).

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S10

Legends for tables S1 to S6

Legends for data S1 to S5

References

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S6

Data S1 to S5