A CO₂ greenhouse efficiently warmed the early Earth and decreased seawater ¹⁸O/¹⁶O before the onset of plate tectonics

Significance

Due to the lower luminosity of the young Sun, climate modelers struggle to explain why the climate on early Earth was not freezing cold. This “faint young Sun paradox” is in conflict with apparently hot Archean ocean temperatures (∼70 °C) that can be estimated from the ¹⁸O/¹⁶O stable isotope ratio of chemical sediments. We show that the later temperatures had been overestimated because the ¹⁸O/¹⁶O of seawater also changed over time due to intense carbonatization and silicification of the oceanic crust, which consumes heavy ¹⁸O. Because these processes require high fluxes of CO₂, greenhouse warming by a CO₂-rich atmosphere appears most feasible to explain all observations.

Abstract

The low ¹⁸O/¹⁶O stable isotope ratios (δ¹⁸O) of ancient chemical sediments imply ∼70 °C Archean oceans if the oxygen isotopic composition of seawater (sw) was similar to modern values. Models suggesting lower δ¹⁸O_sw of Archean seawater due to intense continental weathering and/or low degrees of hydrothermal alteration are inconsistent with the triple oxygen isotope composition (Δ’¹⁷O) of Precambrian cherts. We show that high CO₂ sequestration fluxes into the oceanic crust, associated with extensive silicification, lowered the δ¹⁸O_sw of seawater on the early Earth without affecting the Δ’¹⁷O. Hence, the controversial long-term trend of increasing δ¹⁸O in chemical sediments over Earth’s history partly reflects increasing δ¹⁸O_sw due to decreasing atmospheric pCO₂. We suggest that δ¹⁸O_sw increased from about −5‰ at 3.2 Ga to a new steady-state value close to −2‰ at 2.6 Ga, coinciding with a profound drop in pCO₂ that has been suggested for this time interval. Using the moderately low δ¹⁸O_sw values, a warm but not hot climate can be inferred from the δ¹⁸O of the most pristine chemical sediments. Our results are most consistent with a model in which the “faint young Sun” was efficiently counterbalanced by a high-pCO₂ greenhouse atmosphere before 3 Ga.

The amount of carbon that degassed from a solidifying magma ocean on the infant Earth 4.5 Ga ago was probably similar to the amount of CO₂ that is now present in the atmosphere of Venus (pCO₂ ∼90 bar) (1). Subsequently, dynamic carbon cycling between the early Earth’s atmosphere (atmospheric reservoir [R_A]), the ocean (R_O), and the oceanic crust (R_OC) reservoirs stabilized a primordial Earth–atmospheric pCO₂ to about 1.5 bar (2). Decreasing pCO₂ over the Earth’s history reflects the net transfer of large masses of carbon out of the atmosphere-ocean-oceanic crust (R_A+O+OC) system into the mantle and the emerging continental crust reservoir (R_CC), the latter providing a long-term sink for carbon in the form of inorganic carbon (carbonate rocks) and organic material (organic matter–rich shales, coal, oil, and gas) (1, 2).

Early carbon-cycle models predicted high Archean pCO₂ to account for the faint young Sun paradox (2, 3), whereas direct pCO₂ estimates from the end of the Archean now imply much lower atmospheric pCO₂ [∼0.01 to 0.1 bar (4)]. However, a compilation of evidence from the sedimentary record implies much higher pCO₂ between 3.2 to 3.0 Ga compared to the 2.9 to 2.7 Ga interval (5). These authors state that even several bars pCO₂ are feasible between 3.2 to 3.0 Ga, which is not in conflict with much lower Neoarchean pCO₂ estimates (Fig. 1) or paleo-atmospheric pressure estimates at 2.7 Ga [<2 bar (6) and <0.5 bar (7)]. A fundamental drop in atmospheric CO₂ mixing ratio is also reflected in the observation that >3 Ga, Archean mafic crust (greenstones) is commonly characterized by very intense carbonatization and silicification that is unparalleled in their modern analogs (8–12). Such observations provide evidence for deep-time paleo-pCO₂ fluctuations with a drastic pCO₂ drop starting around 3 Ga ago (5, 13).

Fig. 1.

Constraints on the R_A+O+OC size over time with implications for pCO₂. A illustrates the size of the R_A+O+OC carbon reservoir and its distribution between the oceanic crust (R_OC in gray), ocean water (R_O in dark blue), and the atmosphere (R_A in light blue). Transition of this carbon to the long-term R_CC and the mantle (green) decreases the size of the R_A+O+OC reservoir, which was still large at 3.2 Ga (see SI Appendix) and minimal at the onset of the global glaciations at 2.4 Ga. B shows two recent pH curves over Earth’s history to illustrate how the pH-dependent distribution of carbon between R_A and R_O may translate into pCO₂ at a given time. (C) The panel summarizes pCO₂ estimates [adopted from Catling and Zahnle (4)] and proposed pCO₂ evolution curves illustrated as dashed lines: a (2), b (18), and c (proposed here). Qualitative evidence to construct curve c comes from rare evidence for glaciers (44) and lower degrees of carbonatization of oceanic crust at 3.5 Ga compared to 3.2 Ga (13), suggesting a transient interval of somewhat low pCO₂ in the Paleoarchean. Low pCO₂ is indicated prior to the onset of cold climates later in the rock record. The black bar at 4.5 Ga is derived from carbon-flux arguments and the primordial carbon reservoir (1, 2). The black bar at 3.2 Ga applies the same carbon-flux arguments to translate the high carbonate content observed in the oceanic crust in Pilbara into tentative pCO₂ estimates (SI Appendix), which are most consistent with curve a (2).

Today, carbonatization, which is the formation of carbonates during alteration of the oceanic crust, mainly occurs in relatively cool, off-axis hydrothermal systems over the first 20 Mya after crust formation at midocean ridges (14–16). Elevated degrees of carbonatization in oceanic crust from the Cretaceous and Jurassic are assigned to higher dissolved inorganic carbon at the time (14). Hence, higher pCO₂ (i.e., a larger R_A+O+OC) directly translates into higher degrees of carbonatization. While carbonate is mainly observed as vein fillings in the upper 300 m of oceanic crust today (14, 15), it extensively replaces glass and igneous minerals in Archean greenstones down to depths of 2 km below the ancient seafloor (8–10, 17). The amount of CO₂ fixed in 3.2-Ga-old oceanic crust from Pilbara, Australia is estimated at 1.2 × 10⁷ mol ⋅ m⁻² (±10%) (17)—a remarkable figure that is about two orders of magnitude more compared to today (SI Appendix). Much lower degrees of carbonatization in oceanic crust are already observed at 2.6 Ga, suggesting a drastic Mesoarchean drop in pCO₂ (13) (Fig. 1A).

The qualitative evidence from the sediment record (5) and from the degree of carbonatization and silicification of oceanic crust (13) has not been included in proposed pCO₂ curves (4, 18) because the quantitative conversions into absolute pCO₂ require some assumptions (SI Appendix). Nevertheless, pCO₂ during the early Archean may have been as high as initially predicted by Kasting (2), followed by a pronounced Mesoarchean drop (13) to levels consistent with available paleo-pCO₂ estimates toward the Neoarchean (4) (stippled line “c” in Fig. 1C). Further decreasing pCO₂ toward modern levels partly reflects increasing ocean pH (Fig. 1) rather than a shrinking R_A+O+OC (SI Appendix). Hence, only small effects of carbonatization on the δ¹⁸O_sw value are expected for post-Archean seawater (16). Here, we focus on the very high carbonatization (8–10, 13) and silicification (11, 12) fluxes before the Mesoarchean pCO₂ drop, and we model the respective effects on ancient δ¹⁸O_sw.

The Effect of an Increased CO₂ Flux on the δ¹⁸O_sw of Seawater

The δ¹⁸O_sw of seawater is controlled by the fluxes F_i of oxygen between rocks and the hydrosphere (10). Low-temperature (low-T) weathering products such as clays have ∼20 to 25‰ higher δ¹⁸O values compared to pristine silicate crust (∼5.5 to 12‰). Hence, low-T weathering on the continents (expressed as flux F_cw) and on the seafloor (F_sfw) lowers the δ¹⁸O_sw value of seawater by extracting heavy ¹⁸O from the hydrosphere. Alteration of oceanic crust at hydrothermal temperatures (∼250 to 350 °C) produces rocks with ∼1 to 2‰ lower δ¹⁸O (∼4.5‰) than the pristine oceanic crust basalt (∼5.8‰), thereby adding ¹⁸O to the oceans (F_sp). Classic mass-balance models also account for minor effects from water recycling through the mantle (F_r) and the influence of continental growth (F_cg) (19, 20).

In order to explain the enigmatic shift in δ¹⁸O of ∼15‰ of chemical sediments through time (21–28), extreme and difficult-to-reconcile changes of oxygen fluxes F_i are required (20, 29). As shown in Fig. 2, even a moderate shift of ocean water δ¹⁸O_sw by −5‰ would necessitate very high continental weathering rates (F_cw >10 times present) or very low hydrothermal alteration fluxes (F_sp <0.15 times present). Besides the substantial implications for Precambrian paleoenvironments (21, 30), such flux manipulations would induce much higher ∆’¹⁷O (SI Appendix, Definitions) in Precambrian cherts than documented (Fig. 2). This is presently acknowledged as a strong argument against low δ¹⁸O_sw as an explanation for the low δ¹⁸O of Precambrian chemical sediments (24–27, 31). Most of the ∆’¹⁷O_chert data do not fall on the black equilibrium curve in Fig. 2B (24–27), suggesting that these samples are affected by postdepositional alteration processes. Alteration lowers both δ¹⁸O and ∆’¹⁷O, hence most triple oxygen isotope data seem to suggest postdepositional alteration as the most viable hypothesis for the long-term shift in δ¹⁸O (24–27). Due to sample-size requirements, the least altered cherts measured for ∆’¹⁷O_chert comprise somewhat lower δ¹⁸O compared to the most pristine in situ δ¹⁸O analyses (27, 31). Nevertheless, triple oxygen isotope compositions of well-preserved cherts from the Barberton Greenstone Belt, South Africa fall close to the equilibrium line consistent with hot Archean seawater (31). By combining data from the two techniques, Lowe et al. (31) conclude that “…Archean surface temperatures were well above those of the present day, perhaps as high as 66 to 76 °C.”

Fig. 2.

Plots of ∆’¹⁷O_sw versus δ¹⁸O for the various models and published chert compositions. A shows how seawater composition changes when individual fluxes are manipulated. The δ¹⁸O_sw of seawater decreases for substantially lower F_sp (green) or higher F_cw (gray) but comes along with a ∆’¹⁷O_sw increase. Tick marks refer to multiples of the fluxes used in the base scenario (SI Appendix). Accounting for Archean carbonatization and silicification (F_CO2+SiO2) also decreases the δ¹⁸O_sw of seawater (brown) but leaves ∆’¹⁷O_sw largely unchanged (a). Temperatures in A indicate average precipitation temperatures (T_precip) for carbonates and silicates. B shows how measured triple oxygen isotope compositions of cherts compare to the model. The data are derived from refs. 25–27 and 31. Silica precipitated in equilibrium with modern ocean water must fall on the black equilibrium curve at the respective precipitation temperature. The greenish field illustrates ∆’¹⁷O compositions expected for cherts that precipitated from a low δ¹⁸O_sw due to increasing F_cw or decreasing F_sp as modeled by Sengupta and Pack (24). When accounting for F_CO2+SiO2, however, the δ¹⁸O_sw of seawater decreases, and ∆’¹⁷O_sw remains broadly similar, consistent with Precambrian chert ∆’¹⁷O (brown field). The solid and stippled gray curves indicate precipitation from scenarios a and b, respectively. The light brown field encompasses all possible solutions. The stippled brown line at δ¹⁸O = 23‰ indicates the highest Archean δ¹⁸O analyzed in situ by microanalytical techniques (27, 31). The red arrow depicts an expected slope for alteration.

This interpretation hinges on the assumption that δ¹⁸O_sw was always close to its present-day value. The classic and apparently strong argument for a constant δ¹⁸O_sw of the oceans is derived from the roughly constant δ¹⁸O composition of oceanic crust profiles through time (29). It has been proposed that low δ¹⁸O_sw ocean water would drive altered oceanic crust to low values (29) as observed for rocks altered by isotopically light meteoric water (32–34). However, while local crust alteration with meteoric water constitutes the case of an open system, alteration of the oceanic crust by the ocean-water reservoir as a whole should be regarded as a closed system. In order to extract ¹⁸O from seawater, a complementary reservoir must incorporate ¹⁸O. Interestingly, Archean greenstones tend to be slightly enriched in δ¹⁸O (29) and thus provided a sink for ¹⁸O. A typical feature of Archean terranes >3 Ga is their high degree of silicification, which may well explain the overall elevated δ¹⁸O observed in Archean greenstones (11, 12, 29), which act as the complementary reservoir for depleted ocean water δ¹⁸O in the scenario suggested herein.

We propose that the significance of CO₂ sequestration through carbonatization along with intense silicification, both associated with high Archean pCO₂, have been overlooked and that they offer an explanation for the long-standing oxygen isotope conundrum of chemical sediments deep in time.

Methods and Results

To explore how variable CO₂ sequestration through carbonatization and silicification affects the oxygen isotope composition of the oceans at variable temperatures, we use Muehlenbachs’ model (20), which was recently extended for δ¹⁷O_sw (24) (Fig. 2A). After slight modifications (SI Appendix), the base scenario for modern-day fluxes gives a seawater steady state δ¹⁸O_sw-ss = −0.53‰, intermediate between the modern ocean and the presumed value for an ice-free world (i.e., δ¹⁸O_sw ∼−1‰). The respective ∆’¹⁷O_sw-ss = −9 ppm (i.e., −0.009‰) matches the measured value of −5 ppm for modern seawater within uncertainty (35) (Fig. 2A).

Volcanic CO₂ equilibrates with silicates in the magma chamber before it is outgassed with near-mantle–like δ¹⁸O ∼5.5‰ (SI Appendix). It is subject to various fractionation processes but is eventually immobilized as carbonate with higher δ¹⁸O. Such carbonates, if they become subducted, are mostly volatilized, and the respective CO₂ is reintroduced into the atmosphere with mantle-like δ¹⁸O. Because of the practically infinite mantle oxygen reservoir, CO₂ sequestration and recycling (defined here as F_CO2) extracts ¹⁸O from the hydrosphere.

Heavily silicified greenstones are common in the Archean (11, 12) simply because CO₂ weathering of silicates produces both carbonate (CaCO₃) and silica (SiO₂). For each mole of consumed CO₂ and precipitated CaCO₃, 1 mol SiO₂ forms (CaSiO₃ + CO₂ → CaCO₃ + SiO₂, Urey reaction) so that F_SiO2 equals F_CO2. As for carbonates, quartz comprises high δ¹⁸O even at elevated precipitation temperatures and removes heavy ¹⁸O from the oceans (SI Appendix). Therefore, enhanced CO₂-driven carbonatization and silicification of oceanic crust provides a means of lowering the δ¹⁸O value of Archean seawater (Fig. 2A). These CO₂-related fluxes, F_SiO2 and F_CO2, are not considered in existing models (19–21, 24, 29) and could dramatically differ from the present day owing to the potentially very different atmospheric composition.

The CO₂-sequestration flux F_CO2 by carbonatization of oceanic crust at 3.2 Ga is estimated as 1.5 × 10¹⁴ mol ⋅ yr⁻¹ (17), which is two orders of magnitude larger than the modern carbonatization rate of only 0.45 to 2.4 × 10¹² mol ⋅ yr⁻¹ (14, 15). This high estimate is based on the assumption that spreading rates at the time were three times as high as those at present (36, 37), which is inconsistent with suggestions that plate tectonics only started around 3 Ga (38, 39) but is consistent with proposals that the mode of plate tectonics simply evolved over time (40). The various competing tectonic models proposed for the Archean impose large uncertainties not only on newly introduced F_CO2+SiO2 but for all fluxes. Respective modifications, especially for the classic F_cw and F_sp fluxes, have previously been invoked for the Archean (21, 30) to argue for very low δ¹⁸O_sw, down to −13.3‰. These particular flux modifications can be ruled out, as they induce high ∆’¹⁷O_sw, which is inconsistent with the observed low ∆’¹⁷O_chert (24–27). By adding the CO₂-sequestration flux of Shibuya et al. (17) to our base model, we can explore how it affects the triple oxygen isotopic composition of seawater. Moderate modification of all fluxes then allows identification of feasible flux combinations that most closely resemble the measured ∆’¹⁷O_chert data.

Accounting for the respective F_CO2+SiO2 yields δ¹⁸O_sw-ss of seawater between −7.5 and −2‰ (Fig. 2) and mainly depends on the average temperature for carbonation and silicification within oceanic crust (T_precip). Heavy ¹⁸O is extracted more efficiently from the hydrosphere at low T_precip. Most notably, consideration of carbonatization and silicification fluxes F_CO2+SiO2 and precipitation of the silica and carbonate at 100 to 200 °C lowers the ocean’s δ¹⁸O_sw-ss but does not lead to the increase in ∆’¹⁷O_sw-ss as seen in (24) (Model S1 and Fig. 2A).

Accounting for CO₂ sequestration and silicification provides a much better fit to observed ∆’¹⁷O of Precambrian cherts compared to previous suggestions to lower δ¹⁸O_sw without F_CO2+SiO2 (Fig. 2B). Simply using the modern-day fluxes of Muehlenbachs (20) in combination with the estimated CO₂ flux at 3.2 Ga (17) and T_precip of 150 °C yields δ¹⁸O_sw-ss of −3‰ (scenario “a” in Fig. 2A). The δ¹⁸O_sw-ss can be forced to very low values, but the ∆’¹⁷O_sw increases for such end-members (Fig. 2), making respective scenarios incompatible with chert data. A range of more realistic flux combinations cluster around a resultant δ¹⁸O_sw-ss of seawater of −5‰ with near zero ∆’¹⁷O (e.g., scenario “b” in Fig. 2A; see Model S1 and SI Appendix for a sensitivity analysis). Such a scenario is best compatible with triple oxygen isotope ratios of Archean cherts (25–27, 41) and shales (42).

A direct seawater estimate [δ¹⁸O_sw = −1.7 ± 1.1‰ (43)] at 2.4 Ga corresponds to a time when the degree of carbonatization of oceanic crust was even lower than today (13), suggesting that slightly lower δ¹⁸O_sw persisted in the Paleoproterozoic even without significant CO₂ sequestration and silicification. We suggest that δ¹⁸O_sw further decreased to −5 ± 2‰ when pCO₂ was high at 3.2 Ga. We propose a slightly higher δ¹⁸O_sw of ∼−3 ± 2‰ for the Paleoarchean (at 3.5 Ga) to account for apparently lower degrees of carbonatization (13), observations of cold climate (44), and high δ¹⁸O_carbonate and δ¹⁸O_phosphate (45) at the time (Fig. 3). Both approximations are used in combination with direct δ¹⁸O_sw estimates (43, 46, 47) to construct a seawater curve over time (Fig. 3). Shifts in δ¹⁸O_sw on these timescales are viable because δ¹⁸O_sw can adapt to a new steady state within a few tens of millions of years (SI Appendix).

Fig. 3.

Compiled oxygen-isotope record over Earth’s history with suggested δ¹⁸O_sw evolution. The black curve from Jaffres et al. (21) is constructed through the database for δ¹⁸O_carbonate (calcite = gray points; dolomite and other = gray diamonds) and slightly extended (stippled). Chert data compiled by Knauth (28) (green circles) is extended by data from refs. 22, 25–27, and 55 (green dots). Ranges for modern and ancient δ¹⁸O_phosphate (45) are indicated by brownish boxes. The proposed seawater curve is constructed from published δ¹⁸O_sw (43, 46, 47) and Archean δ¹⁸O_sw proposed herein (see Methods and Results). High δ¹⁸O_sw estimates (56, 57) are not included (SI Appendix). The curve and its blue margin of 2‰ are not based on statistics. Arrows indicate warming (red) and cooling (blue) trends expected from the changing size of the R_A+O+OC reservoir. Temperature calibrations are anchored at δ¹⁸O_sw = −5‰ for cherts and calcite and at δ¹⁸O_sw = −3‰ for phosphates. Apparently, too-high δ¹⁸O_carbonate are dolomites that generally comprise ∼2‰ higher δ¹⁸O compared to calcite.

Discussion

Even with the moderately low δ¹⁸O_sw of seawater and warm precipitation temperatures proposed here, alteration is still required to explain the triple oxygen isotopic compositions of most Precambrian cherts (Fig. 2). Bedded cherts are frequently interpreted to reflect precipitation from warm/hot bottom water brines or hydrothermal plumes (25–27). Cherts of superior-type banded iron formations, however, certainly originate from primary precipitates from ambient seawater temperatures on continental shelves. Surface waters within epicontinental seaways may have been low in δ¹⁸O due to a partly meteoric origin (29), and respective ∆’¹⁷O would have been only slightly elevated. Alternatively, isotopic exchange between iron phases (with low δ¹⁸O_FeOx) and silica bands (with high δ¹⁸O_opalA) decreased δ¹⁸O_chert and generated ∆’¹⁷O_chert below the equilibrium curve (41). Alteration with (meteoric) water resembles a similar mixing mechanism that has been suggested to lower ∆’¹⁷O_chert (25–27). The variable combinations of these processes induce the considerable scatter in the oxygen-isotopic composition of chemical sediments through time (Fig. 3). Collectively, the Archean δ¹⁸O record can be rationalized by a combination of the classic explanations: 1) a moderately low δ¹⁸O_sw, 2) warm ocean temperatures, and 3) alteration.

Although even the most pristine Archean cherts are altered to some extent, their oxygen isotope composition reflects lower δ¹⁸O of Archean ocean water as a result of high degrees of carbonatization (8–10, 13) and silicification at >3 Ga ago (11, 12). This hypothesis needs further verification from high-precision triple oxygen isotope analyses of other types of chemical sediments. Eventually, concepts to “see through” diagenesis (48) can reveal the most viable triple oxygen isotope composition of Archean seawater. Analyzing the triple oxygen isotopic composition of apparently unaltered phosphates (45) (but see ref. 31) may even put direct constraints both on Archean seawater ∆’¹⁷O_sw and precipitation temperatures. Knowledge of the exact ∆’¹⁷O_sw would further restrict feasible Archean oxygen fluxes including the relative proportion of the CO₂-sequestration flux. Respective constraints may also provide information on the tectonic regime (e.g., sagduction versus subduction, high versus low spreading rates, etc.) that is active at a given time (36–40).

From the model presented here, lowering the seawater δ¹⁸O requires high CO₂ fluxes, which are most viable for a high-pCO₂ atmosphere. The exact level of pCO₂ in the Archean most probably fluctuated quite a bit as it has done for the past billion years. Large fluctuations, such as the proposed Mesoarchean drop in pCO₂ (5, 13), should induce a significant shift in δ¹⁸O_sw. Indeed, a respective shift is broadly recorded in the oxygen isotope record (Fig. 3), implying that carbonatization and silicification processes had indeed been large enough to significantly decrease the δ¹⁸O_sw.

Even with the moderately low δ¹⁸O_sw proposed herein, the temperatures estimated from chemical sediments imply a very warm climate on the early Earth, which is consistent with the rare evidence for ice before 3 Ga ago (44). With respect to the vigorously debated faint young Sun paradox (3, 49, 50), this implies generally high concentrations of greenhouse gases for most of the Archean. Our study supports the original idea that CO₂ and not CH₄ counterbalanced the lower luminosity of the faint young Sun (2, 3), at least prior to the stark Mesoarchean drop in pCO₂ (5, 13). The apparently warmer temperatures compared to the Phanerozoic leads to the question of how pCO₂ can be maintained at high levels in a constantly warm climate, where it accelerates CO₂ weathering, which leads to a decrease in pCO₂ and keeps Earth’s surface in a temperature window that is “just right” (51).

Carbon transfer from the R_A+O+OC reservoirs to the R_CC and the mantle (i.e., to the long-term storage reservoirs) was probably inefficient on the early Earth. Continental crust was less abundant, and thus the continental reservoir was small. The residence times for carbon on early continents was probably shorter than today due to efficient recycling of thin continental-crust margins and acidic rain (due to high pCO₂). Sagduction tectonics, proposed for the early Earth (39), is associated with high thermal gradients leading to efficient CO₂ recycling and thus inefficient carbon storage in the mantle.

Modern-style plate tectonics provides a mechanism to accrete land masses and to thicken continental crust (52), which generates topography and subaerial land surfaces. The two negative feedbacks for pCO₂ are the following: 1) the establishment of stable, long-term storage reservoirs for carbon and 2) an increased supply of weatherable rocks, enhancing continental CO₂-weathering fluxes. Subaerial land mass is assumed to be small before the Neoarchean, with the large-scale emergence of continents above sea level proposed to occur at 2.5 Ga (53). The R_A+O+OC carbon reservoir drastically decreased in size in the preceding several hundred million years, suggesting that carbon was already relocated to the long-term storage reservoir on evolving continental platforms (38, 40, 52) and that carbon transfer to the mantle became more efficient within proto-Phanerozoic–type subduction zones. Hence, sharply decreasing pCO₂ could be a consequence of a rapid or gradual switch from vertical sagduction–type tectonics to a horizontal subduction regime (39), generating thick, silica-rich continental crust from about 3 Ga onwards (52).

Conclusions

In this contribution, we propose that the intense carbonatization and silica precipitation due to high CO₂ fluxes in the Archean led to lower δ¹⁸O_sw values of seawater without affecting the Δ’¹⁷O. The shift in δ¹⁸O_sw is mirrored in the lower δ¹⁸O of early Archean chemical sediments. The decreasing intensity of carbonatization and silica precipitation around ∼3 Ga shifted the isotope composition of the ocean to similar values as seen in the oceans today. On a smaller scale, however, variable CO₂ fluxes in the Proterozoic and Phanerozoic also induce variations in F_CO2+SiO2 and thus in seawater δ¹⁸O_sw. Therefore, long-term fluctuations in the δ¹⁸O of chemical sediments do not only reflect temperature but also changing δ¹⁸O_sw. The apparent temperature change estimated for long-term climate fluctuations (e.g., ref. 54) are thus systematically overestimated.

Data Availability

All data are included in the article and/or supporting information.

Change History

June 3, 2021: The supporting information has been updated.