Great Oxidation and Lomagundi events linked by deep cycling and enhanced degassing of carbon

Abstract

For approximately the first 2 billion years of Earth history, atmospheric oxygen levels were extremely low. It wasn’t until at least half a billion years after the evolution of oxygenic photosynthesis, perhaps as early as 3 billion years ago, that oxygen rose to appreciable levels during the Great Oxidation event. Shortly after, marine carbonates experienced a large positive spike in carbon isotope ratios known as the Lomagundi event. The mechanisms responsible for the Great Oxidation and Lomagundi events remain debated. Using a carbon-oxygen box model which tracks surface and interior C fluxes and reservoirs while also tracking C isotopes and atmospheric oxygen levels we demonstrate that about 2.5 billion years ago a tectonic transition resulting in increased volcanic CO₂ emissions could have led to increased deposition of both carbonates and organic carbon via enhanced weathering and nutrient delivery to oceans. Increased burial of carbonates and organic carbon would have allowed accumulation of atmospheric oxygen while also increasing delivery of carbon to subduction zones. Coupled with preferential release of carbonates at arc volcanoes and deep recycling of organic C to ocean island volcanoes we find such a tectonic transition can simultaneously explain the Great Oxidation and Lomagundi events without any change in the fraction of carbon buried as organic carbon relative to carbonate, which is often invoked to explain carbon isotope excursions.

Carbon dioxide and oxygen, two gases critical for Earth’s habitability, are linked via photosynthesis:

$${\text{CO}}_2+{\text{H}}_2\text{O}={\text{CH}}_2\text{O}+{\text{O}}_2$$ (1)

The link between CO₂ and O₂ is evident in the geologic record from the association of the most dramatic events in the histories of both gases. The Great Oxidation Event (GOE), during which Earth’s atmospheric oxygen levels increased by several orders of magnitude, and the Lomagundi Event (LE), during which carbon isotope ratios of marine carbonates increased by up to ~10‰, both occurred around 2.3–2.0 Ga^1–3 (Fig. 1). The temporal association of the two events suggests both may be attributed to a single mechanism, such as enhanced organic carbon burial⁴. This is because increasing the fraction of carbon buried as organic carbon (f_org) relative to carbonate will increase the amount of oxygen produced and allowed to accumulate via equation (1). Increased f_org also increases δ¹³C of carbonates because organic matter preferentially incorporates ¹²C, leaving the carbonate-forming C reservoir enriched in ¹³C (ref. ⁴).

Figure 1. Variations in atmospheric O₂, δ¹³C of carbonates, and tracers of tectonic processes through geologic time.

a, Qualitative model of atmospheric O₂ (solid blue curve)¹ and estimation of atmospheric O₂ from Zn/Fe ratios in carbonates (dashed blue line with shaded region for uncertainties)⁴⁶. b, Carbonate δ¹³C through time⁴³. c, Zircon δ¹⁸O-εHf covariance*slope from Keller, et al.²⁴, where positive values indicate increases in crustal contribution to magmatism. Orange band is proposed timing for rapid emergence of subaerial continents inferred from triple-oxygen isotopes¹⁵, blue band is timing large-scale glaciations²⁴.d, U/Pb age distribution of detrital and orogenic zircons⁴⁷, with orange band showing the proposed timing of transitional tectonics between stagnant lid and plate tectonics¹².

Despite the apparent temporal association of the GOE and LE and the conceptual consistency of explaining both events via increased f_org, its role as the driver of the GOE and LE has been questioned. There is evidence that oxygenic photosynthesis evolved at least half a billion years before the GOE⁵, so why would atmospheric oxygen have remained low for so long?⁶. In addition, other studies have suggested that the GOE may have preceded the LE by ~100 million years^2,3,7,8. If the onset of the GOE and LE were indeed temporally disparate, it becomes difficult to explain both events via a single mechanism. This has led researchers to propose mechanisms invoking decreased sinks of atmospheric oxygen as the main driver of the GOE^9–11. While decreased oxygen sinks would allow atmospheric O₂ to accumulate, it is unclear how they would drive the associated positive C isotope excursion.

Here, we propose a single mechanism for the GOE and LE. Our modeling shows that the proposed mechanism can simultaneously explain increased oxygen production as well as a positive C isotope excursion in marine carbonates. In addition, we show that the proposed mechanism is consistent with a delay between the evolution of photosynthesis and increased atmospheric oxygen levels, as well as a delay between the build-up of oxygen in the atmosphere and the positive C isotope excursion observed in marine carbonates.

Enhanced volcanic CO₂ driven by tectonic transitions

We propose that around 2.5 billion years ago a major tectonic transition increased volcanic CO₂ emissions, driving increased rates carbonate and organic C burial via enhanced weathering and nutrient delivery to oceans. A transition from stagnant/sluggish lid to plate tectonics has been proposed around 2.5 Ga¹², which would have caused increased volcanic CO₂ emissions^13,14. Also proposed around 2.5 billion years ago are the rapid emergence of subaerial continents¹⁵ and major glaciations¹⁶. These events would have increased the supply of hydrated sediments to trenches, lubricating the subduction interface and enhancing plate tectonic activity, which would have led to enhanced volcanic CO₂ emissions¹⁷. Moussallam et al.¹⁸ propose that mantle cooling resulted in lower eruptive temperatures for volcanic gases, which caused the gases to become more oxidized, resulting in CO₂ emissions to increase relative to emissions of more reduced C gases.

A GOE resulting from tectonically-driven increases in volcanic CO₂ emissions is consistent with oxygenic photosynthesis evolving prior to the GOE⁵ because the production rate of oxygen is controlled by the flux of organic C burial:

$$F_{\mathrm{org}}=F_{\mathrm{tot}}\times f_{\mathrm{org}}$$ (2)

where F_org is the flux of organic carbon burial, and F_tot is the flux of total carbon burial (carbonate + organic C). Therefore, F_org, and the oxygen production flux, can increase by increasing F_tot, instead of f_org. At the onset of oxygenic photosynthesis, oxygen production could have been low because Earth’s tectonic mode promoted low weathering fluxes (i.e., stagnant-lid, un-emerged continents), resulting in low F_tot and thus low F_org. Then a tectonic transition could have increased F_tot and thus F_org, resulting in a delay between the evolution of photosynthesis and the GOE^5,19. This could have all been achieved without any significant changes to f_org.

Carbonate release at arcs drive positive spike in δ¹³C

Positive δ¹³C excursions in carbonates are commonly attributed to increased f_org:

$${\mathrm{\delta }}^{13}{\mathrm{C}}_{\mathrm{in}}={\mathrm{\delta }}^{13}{\mathrm{C}}_{\mathrm{carb}}(1-f_{\mathrm{org}})+{\mathrm{\delta }}^{13}{\mathrm{C}}_{\mathrm{org}}{f}_{\mathrm{org}}$$ (3)

where δ¹³C_in is δ¹³C of atmospheric inputs of CO₂ and δ¹³C_org is δ¹³C of organic C. If fractionation of C isotopes between carbonates and organic C (Δ¹³C_carb-org ≡ δ¹³C_carb- δ¹³C_org ≈ 25‰), and δ¹³C_in remain constant, increased f_org increases δ¹³C_carb. δ¹³C_carb also increases if δ¹³C_in increases at constant f_org and Δ¹³C_carb-org. We propose that the increased δ¹³C_carb of the LE could have been caused by increased δ¹³C_in rather than increased f_org.

δ¹³C_in depends on the balance of CO₂ being emitted at different volcanic settings:

$${\mathrm{\delta }}^{13}{\text{C}}_{\text{in}}=\frac{F_{\text{OIB}}}{F_{\text{tot}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{OIB}}+\frac{F_{\text{MORB}}}{F_{\text{tot}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{MORB}}+\frac{F_{\text{arc}}}{F_{\text{tot}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{arc}}$$ (4)

where F_i is CO₂ flux at volcanic setting i and δ¹³C_i is δ¹³C of CO₂ emitted at volcanic setting i. δ¹³C_MORB ≈ −5‰^20,21, CO₂ flux-weighted δ¹³C_arc ≈ −3‰²², and sparse studies on δ¹³C_OIB suggest it could be similar to MORB or more negative^21,23. Therefore, increasing $\frac{F_{\text{arc}}}{F_{\text{tot}}}$ increases δ¹³C_in by shifting the balance of δ¹³C_in towards δ¹³C_arc.

The tectonic transitions discussed above can increase carbonate and organic C burial leading to increased subduction of C. Upon reaching subarc depths the increased flux of subducted C would be released from the slab at an increased rate resulting in increased $\frac{F_{\text{arc}}}{F_{\text{tot}}}$, shifting the balance of δ¹³C_in towards δ¹³C_arc. This mechanism is consistent with oxygenation preceding increased δ¹³C_carb⁸ because oxygen accumulates immediately after burial of organic C. Contrastingly, δ¹³C_carb increases after the increased C flux is buried, subducted, released from the slab, emitted at arcs as CO_2, and incorporated into marine carbonates. Evidence for increased delivery of subducted sediments to arc volcanic sources is provided by the record of covariance of δ¹⁸O and εHf in zircons - both of which quantify the contribution of subducted weathering products to magma generation²⁴. Fig 1c shows that the peak in δ¹³C_carb closely correlates with the peak in δ¹⁸O-εHf_zircon covariance, supporting the idea that δ¹³C_carb may be controlled by the balance of δ¹³C of CO₂ emitted at different volcanic centers.

Increasing δ¹³C_in by increasing $\frac{F_{\text{arc}}}{F_{\text{tot}}}$ is possible if δ¹³C_arc is high, which is most easily explained by a heavy contribution of carbonates to arc CO₂ emissions. A recent study determined that the flux-weighted, average δ¹³C_arc≈−3‰²². Whether high δ¹³C_arc is due to contribution of carbonates from the subducted slab or upper crust remains debated^22,25. Figure 2a shows δ¹³C_arc versus f_org of the sediments entering the respective trench. As f_org of subducted C becomes greater, δ¹³C_arc becomes more negative, implying that f_org of subducted sediments is the dominant controller of δ¹³C_arc²⁶. The dashed line in Fig. 2a shows what δ¹³C_arc would be if carbonates and organic C were released with equal efficiency. Data in Fig. 2a suggest preferential carbonate release, supporting the idea that δ¹³C_arc is high due to preferential release of carbonates relative to organic C from the subducted slab. Below, we test whether ancient arcs may have also had high δ¹³C.

Figure 2. Natural data and petrologic calculations suggesting preferential release of carbonate over organic carbon in present-day and ancient subduction zones.

a, δ¹³C_arc²² versus f_org of sediments at modern subduction zones⁴⁸. Orange symbols represent volcanoes erupting through carbonates, increasing likelihood of crustal carbonate dominating δ¹³C_arc⁴⁹. Dashed line shows equal efficiency of release for carbonates and organic C. Error bars are 1σ. b, Ancient slab P-T paths²⁷ (Dashed black lines are cold and hot P-T paths, solid black line is average of the two) compared to metamorphic decarbonation (blue lines labeled with wt% CO₂ retained in slab), and melting out of carbonates²⁹ (red line). c, Percentage of C retained in slab versus depth. Blue lines show how much C remains in slab (solid and dashed lines calculated along respective P-T paths of b) during metamorphic decarbonation. Red lines calculated for carbonates released by carbonated melting (red line in b). Black lines for organic C release by graphite dissolution in silicate melts.

Ancient subduction zones were likely hotter than present day due to secular mantle cooling²⁷. Figure 2b,c demonstrates that along P-T paths of subducting slabs adjusted for secular mantle cooling²⁷, carbonate will be removed from the slab more efficiently than organic C, which metamorphoses to graphite during subduction²⁸. In a carbonated pelitic sediment, metamorphic decarbonation and partial melting²⁹ can release up to 100% of subducted carbonates (Fig. 2c). Graphite removal via CO₂ dissolution in pelitic sediment partial melts can only extract up to ~50% of the subducted organic C. This demonstrates that like present-day arcs and perhaps even more so, ancient δ¹³C_arc was more strongly influenced by δ¹³C_carb than δ¹³C_org.

Considering the results provided above, we propose enhanced subduction of C can increase δ¹³C_carb by increasing δ¹³C_in due to preferential release of carbonates relative to organic C at arcs. This explains increased δ¹³C_carb observed in the LE, however it doesn’t explain why δ¹³C_carb returned to around 0‰ (Fig. 1). We propose that the return to “normal” δ¹³C_carb (~ 0‰) can be explained by delayed release of organic C at intraplate ocean island volcanoes.

Return to normal δ¹³C_carb via deep recycling of organic C

After releasing carbonates, the slab becomes enriched in organic C, which remains in the slab as refractory graphite/diamond^27,30. Subducted slabs have been proposed to descend to the core-mantle boundary and reside in “slab graveyards” until being entrained in upwelling mantle plumes, fueling volcanism at ocean islands^31,32. Upon ascent through the mantle, reduced organic C oxidizes and initiates redox melting³³, allowing organic C to be emitted at ocean islands as CO₂. The large reservoir of Fe³⁺ relative to organic C in the mantle can effectively oxidize organic C without experiencing major changes over geologic time. Unlike arc δ¹³C_arc, δ¹³C_OIB has received relatively little attention. However, several studies report δ¹³C_OIB, which may support the hypothesis of ocean islands being enriched in organic C.

Figure 3 shows a compilation of δ¹³C_OIB. δ¹³C_OIB measurements at Kilauea were made on high-temperature volcanic CO₂³⁴, making them the most appropriate to compare with the arc dataset, which are largely fumarolic measurements. Figure 3 shows that δ¹³C at Kilauea is lower than δ¹³C_arc. Measurements made at Koolau³⁵, Society²³, Loihi³⁶, and Pitcairn²¹ were made on CO₂ from vesicles, glass, or melt inclusions and are subject to isotope fractionation during degassing³⁷. For this reason, we identify the least degassed sample from each dataset. Except for Society, which has been shown to have similar noble gas ratios to MORB²³, the least degassed samples from ocean islands have lower δ¹³C than MORBs and arcs. These measurements support the possibility of organic C enrichment in ocean island basalts. We propose that delayed release of organic C at ocean islands can explain the return to “normal” δ¹³C_carb (about 0‰) at the tail-end of the LE. This mechanism requires that subducted organic C retains its low δ¹³C, which is supported by the observation of light δ¹³C values in eclogitic diamonds³⁸. Additionally, organic C hosted in subducted oceanic crust may remain physically and chemically isolated from C hosted in ambient mantle precluding any alteration of δ¹³C values. Other elements, such as nitrogen, have also been argued to show an enrichment in subducted organic components at ocean island volcanoes³⁹. A schematic diagram showing preferential release of organic C at ocean islands and carbonates at arcs shown in Fig. 4.

Figure 3. Comparison of carbon isotope compositions of CO₂ emissions from different volcanic settings.

Measurements made on samples from Koolau³⁵, Society²³, Loihi³⁶, and Pitcairn²¹ were made on CO₂ from vesicles, glass, or melt inclusions and are therefore subject to isotope fractionation during degassing³⁷. Least degassed samples are marked for each location with a green star. Measurements at Kilauea³⁴ are made on high-T volcanic gas emissions and are therefore the most directly applicable for comparison to the global flux-weighted average of arc CO₂ emissions²². Available data suggests that ocean islands may emit CO₂ with lower δ¹³C than MORB and arcs.

Figure 4. Schematic diagram showing preferential release of carbonate C at arc volcanoes and organic C at ocean island volcanoes.

Both carbonates and organic carbon are deposited on the seafloor and subducted into Earth’s interior. Upon subduction, organic C metamorphoses to graphite, which is more refractory than carbonate. At subarc conditions carbonates are preferentially released from the slab (Fig. 2) relative to graphitized organic C, increasing arc volcano δ¹³C to values greater than −5‰. Graphitized organic C remains in slab and is transported deep into the mantle, where it may become entrained in upwelling mantle plumes feeding ocean island volcanoes. Upwelling, reduced organic C likely undergoes redox melting, decreasing ocean island volcano δ¹³C to <−5‰.

The proposed mechanism combined with the observation that δ¹³C_carb has always returned to ~0‰ after large excursions (Fig. 1), leads us to infer that C fluxes into and out of the mantle are at near steady-state with response times to perturbations on the time-scale of mantle transit times, which are on the order of hundreds of million years⁴⁰. If C fluxes into and out of the mantle were out of balance, then due to the more refractory nature of graphitized organic C, we would expect δ¹³C_carb to increase with time due to the preferential sequestration of light organic C in the mantle. However, δ¹³C_carb is not observed to increase with time, rather it experiences excursions then returns to ~0‰ (Fig. 1).

Modeling oxygen levels and δ¹³C_carb

Here, we present a carbon-oxygen box model to test whether increased volcanic CO₂ emissions can drive oxygenation and a positive δ¹³C_carb excursion. The key model assumptions are: 1) Atmospheric oxygen levels are proportional to the mass of organic C in the system. 2) The mass of organic C buried depends on the flux of carbon deposition, which is proportional to the concentration of atmospheric CO₂:

$$F_{\mathrm{w}}=k\times [{\mathrm{CO}}_{2,\mathrm{atm}}]$$ (5)

where F_w is the flux of CO₂ drawdown from the atmosphere, k is the strength of the weathering feedback, and [CO_2,atm] is the concentration of atmospheric CO₂. 3) δ¹³C_arc is strongly influenced by δ¹³C_carb and δ¹³C_OIB by δ¹³C_org. 4) F_arc responds to surficial changes in CO₂ fluxes on the order of 10 Myr, while F_OIB responds on the order of 100 Myr, corresponding to one mantle transit time for subducted crust^32,40. These timescales are equivalent to residence times of subducted carbonates and organic C in their respective mantle reservoirs. These residence times appear short compared to those estimated in previous studies (e.g.⁴¹). However, the > 1 billion years residence time of carbon in the mantle is calculated assuming the mantle as a single reservoir for C, which is an oversimplification of the natural C cycle because the mantle contains both recycled and primordial carbon. We assume the residence time of subducted carbonates in the ancient mantle is on the order of 10 million years because they are rapidly released to arc source mantle during subduction⁴². We relate the residence time of organic C to the timescales of mantle convection because organic C is hosted as refractory/graphite diamond in subducted oceanic crust and is unlikely to efficiently mix with surrounding mantle. Therefore, as soon as the subducted oceanic crust completes a convective cycle and undergoes partial melting in the oxidized upper mantle, organic C will be efficiently degassed as CO₂.

Figure 5 was generated by increasing mid-ocean ridge CO₂ emissions by 1000-fold instantaneously at 2.5 Ga. Figure 5 demonstrates that in response to increased CO₂ emissions, the production flux of carbonates and organic C increases, causing atmospheric oxygen to increase immediately due to the increased reservoir of buried organic C. Shortly after oxygen increases, δ¹³C_carb increases as $\frac{F_{\text{arc}}}{F_{\text{tot}}}$ increases, with the delay being dictated by the time required for carbonate to be deposited, subducted, and released at arcs as CO₂ (Fig. 5). After a longer delay, dictated by the time required for organic C to be buried and later released as CO₂ at ocean islands, δ¹³C_carb decreases in response to increased $\frac{F_{\text{OIB}}}{F_{\text{tot}}}$. Note that model results were generated keeping f_org constant at 0.20 throughout the model run.

Figure 5. Model results showing how tectonically-driven increased CO₂ emissions and deep recycling of organic C can drive GOE and LE.

a. δ¹³C_carb versus time. Orange curve is model result, blue symbols are natural data. When increases to MOR CO₂ emissions are large, magnitude of C isotope excursion becomes insensitive to magnitude of changes to MOR CO₂ emissions (See SI). b. Atmospheric O₂ versus time. Orange curve shows results for 1000-fold increase in MOR CO₂ emissions, while bands show results when increases range from 100-fold to 10,000-fold. Blue curves from Fig. 1. Vertical dashed line emphasizes that oxygen increases before C isotope excursion. c. CO₂ fluxes versus time. Shaded regions as in b. d. C reservoir size versus time. Shaded regions as in b.

A notable result of the model is that simply increasing volcanic CO₂ emissions can match the magnitude of oxygen increase while nearly matching the magnitude of δ¹³C excursions observed for the GOE and LE, which can’t be achieved by solely changing f_org^43,44_. Another notable result of the model is that it closely reproduces the topology of the Lomagundi event, which displays a gradual increase in δ¹³C and terminates abruptly with δ¹³C rapidly decreasing to ~0‰. Supplementary Fig. 2 shows how varying different model parameters affects the shape and magnitude of the C isotope excursion. Furthermore, the present model shows that a permanent transition to increased volcanic CO₂ emissions produces a permanent transition to higher atmospheric O₂ but causes a transient excursion in δ¹³C as observed in natural data. We note that our model is consistent with any event that increases CO₂ emissions. Therefore, the proposed mechanism is not only applicable to the Great Oxidation and Lomagundi event (see Supp. Fig. 3), it can also explain other large oxidation and C isotope excursions such as the events observed at the end of the Proterozoic¹, which have also been proposed to result from large emissions of CO₂⁴⁵. This is in contrast to the model of Duncan and Dasgupta²⁷, which relies on initiation of plate tectonics to coincide with the GOE.

In summary, we propose that the GOE and LE can be explained by a single mechanism – increased carbonate and organic C deposition driven by enhanced volcanic CO₂ degassing, possibly driven by a transition from stagnant/sluggish lid to plate tectonics¹². Alternatively, the rapid emergence of subaerial continents¹⁵ or global glaciation^17,24, proposed to have occurred ~2.5–2.3 Ga, have been argued to facilitate enhanced volcanic CO₂ emissions by supplying sediments to convergent margins thereby facilitating more efficient subduction and related volcanism¹⁷. The presented hypothesis is consistent with natural observations such as contemporaneous increases in continental weathering²⁴, heavy δ¹³C_arc²², and possible light δ¹³C_OIB²¹. We present a carbon-oxygen model which demonstrates increased volcanic CO₂ emissions coupled with preferential release of carbonate-derived CO₂ at arc volcanoes and the delayed release of organic C at ocean island volcanoes can drive atmospheric oxygenation and positive C isotope excursions without requiring any changes to f_org

The present study raises several important implications. 1) Large-scale, long-lived positive δ¹³C isotope excursions such as the LE do not require changes in f_org, they may be the result of fractionation of carbonates and organic C in Earth’s interior. 2) Oxygen and biomass production may be controlled by tectonic conditions dictating volcanic CO₂ emissions. 3) Carbon fluxes into and out of the mantle may be close to steady state, with response times to perturbations on the order of mantle transit times. These conclusions provide an alternate interpretation of the evolution of CO₂ and O₂ on Earth and can inform our thinking on how Earth, the only known inhabited planet in our Solar System and beyond, became and remained oxygenated, allowing for the proliferation of complex life on the surface.

Methods

Quantifying C release from the subducting slab

To quantify carbon release from the subducting slab via metamorphic decarbonation, we performed thermodynamic calculations with the software package Perple_X ⁵⁰. We calculate the stable mineral assemblage of a model carbonated pelite as a function of pressure and temperature, with a bulk composition from Tsuno et al.²⁹. We use the calculated bulk composition of the residual solid lithology to determine the amount of CO₂ that can be retained in the solid lithology in the form of carbonates. The CO₂ wt.% of the bulk rock decreases as P and T increase (Fig. 2a). We then calculate how much of the original carbon is retained in the solid lithology along subduction zone P-T paths relevant to ancient subduction zones. Following the arguments of Duncan and Dasgupta²⁷ we assume that ancient subduction zones were hotter than those in the present day by about 70 °C/Gyr. Therefore, to investigate conditions relevant to the GOE and LE, we increased the temperatures of slab top P-T paths from Syracuse et al.⁵¹ by 175 °C. Assuming a starting CO₂ of 5 wt.% as mineral carbonates, the amount of original subducted CO₂ retained in the slab as carbonate was calculated along P-T paths according using the following equation:

$$\text{Percent}{\text{CO}}_2\text{retained}=\frac{5-{\text{CO}}_2\text{wt}\%\text{of}\text{bulk}\text{rock}}{5}\times 100$$ (6)

Where 5 is the initial amount of CO₂ in the bulk rock composition in wt%, and ‘CO₂ wt% of bulk rock’ is the CO₂ content of the bulk calculated as a function of pressure and temperature using Perple_X.

We note that the extent of metamorphic decarbonation calculated in this fashion provides an underestimate if extraneous fluid source is available. Incorporating open system, fluid infiltration induced metamorphic decarbonation more efficiently breaks down carbonates in the slab⁵². Similarly, the availability of extraneous fluid allows carbonate dissolution to be an important process of carbonate removal⁴². Furthermore, for warm subduction zones of ancient past, complete carbonate removal from the downgoing slab likely took place via carbonated melting⁵³. Hence there are several mechanisms that could have led to complete exhaustion of carbonates from the slab in ancient subduction zones.

To quantify graphitized organic carbon release from subducting slabs we followed the methods of Duncan and Dasgupta²⁷, in which the authors predict dissolved CO₂ concentrations in hydrous rhyolitic melts in equilibrium with graphite/diamond as a function of P, T, and fO₂. In our calculations of graphite/diamond dissolution in silicate melts we use the same P-T paths described above and fO₂ was calculated along the CCO buffer⁵⁴. Organic C removal efficiency would be even lower if more reduced slab conditions are used, i.e., the fO₂ conditions used give the highest CO₂ dissolution capacity and graphite/diamond saturation. The results shown in Fig. 2 were generated assuming 1 wt% graphite in the initial bulk composition. See Duncan and Dasgupta²⁷ for more details on calculations.

Modeling atmospheric oxygen levels and δ¹³C of carbonates

We developed a new coupled carbon-oxygen cycle model which tracks δ¹³C. The model tracks how a set of carbon reservoirs (C_i) respond to changes in carbon fluxes (F_i) between these reservoirs. The surface reservoirs are the atmosphere-ocean (C_atm), solid carbonates (C_carb), and organic carbon (C_org), which are generated from the atmospheric reservoir (in this model we do not treat any fractionations between the ocean-atmosphere reservoirs). The transfer of carbon from the atmosphere-ocean reservoir into the carbonate and organic carbon reservoirs is controlled by a weathering flux (F_w, which is divided into F_carb and F_org) which is proportional to the weathering constant (k) and the concentration of CO₂ in the atmosphere (C_atm). Carbon that is fluxed out of the atmosphere can be deposited as either carbonates or organic carbon. In Fig. 5 we prescribe that 40% of carbonates and organic C fluxed out of the atmosphere are added to a crustal reservoir (C_carb and C_org), while the other 60% are subducted (χ =0.60). To demonstrate that a carbon isotope excursion is possible without changing the fraction of organic C buried relative to carbonate, we hold f_org constant at 0.20 throughout the model.

While treating carbon in the Earth’s interior, we divide the mantle into three different carbon reservoirs. The first reservoir holds primitive carbon C_prm, which we treat as having existed in the mantle since its formation, additionally it receives no contribution from subducted components throughout the model evolution. The other two mantle carbon reservoirs consist of carbon hosted in subducted carbonate (C_mcarb), and carbon hosted in organic carbon (C_morg), which have influxes from surfaces reservoirs controlled by the flux of subducted carbonate (F_subc) and organic carbon (F_subo) respectively. To relate oxygen levels to the carbon cycle we assume the atmospheric oxygen levels are directly proportional to the mass of organic carbon in the surface and mantle reservoirs.

We prescribe that subducted carbon can leave the mantle reservoirs via arc volcanoes (F_arc) and ocean island volcanoes (F_OIB). In order to keep track of how differential release of carbonates versus organics at different volcanic settings affects C isotopes of CO₂ emissions through time, we treat volcanic CO₂ fluxes at arc volcanoes and ocean island volcanoes as the sum of CO₂ fluxes derived from carbonates (F_arcc & F_OIBc), organic carbon (F_arco & F_OIBo), and primitive mantle carbon (F_arcp & F_OIBp), each having their own distinct δ¹³C. To account for the differential release of carbonates and organics at arcs and OIBs we prescribe that a certain fraction of subducted carbonates (α_carb) and organics (α_org) are released at arcs, with the remaining fraction released at OIBs (1- α_carb, and 1- α_org). In accordance with our calculations in the main text (Fig. 2), we treat carbonates as being completely released at arcs (α_carb = 1), and organics as being completed released at OIBs (α_org = 0) when generating Fig. 5. Supplementary Figs. 2 and 3 show how probable variations in α_carb and α_org would influence the modeled evolution of δ¹³C_carb.

Subducted carbon cycled through the mantle and released at arc volcanoes travel on the order of hundreds of kilometers, while carbon cycled through ocean island volcanoes may travel to the core-mantle boundary, traveling distances on the order of thousands of kilometers. Assuming mantle convection occurs on the order of 1 cm/yr the residence time of mantle carbon released at arcs (τ_arc) is on the order of 10 Myr, while the residence time of mantle carbon released at ocean island volcanoes (τ_OIB) is on the order of 100 Myr. Therefore, perturbations to subduction carbon fluxes will manifest themselves in arc volcanoes before ocean island volcanoes. The total volcanic CO₂ outgassing flux (F_out) is a sum of the fluxes at arcs, OIBs, and MORB.

The system of equations representing this coupled surface and interior carbon cycle are:

$$\frac{{d\text{C}}_{\text{atm}}}{\mathit{dt}}=F_{\text{out}}-F_{\text{w}}$$ (7)

$$\frac{{d\text{C}}_{\text{carb}}}{\mathit{dt}}=(1-f_{\text{org}})F_{\text{w}}-F_{\text{subc}}$$ (8)

$$\frac{{d\text{C}}_{\text{org}}}{\mathit{dt}}=f_{\text{org}}{F}_{\text{w}}-F_{\text{subo}}$$ (9)

$$\frac{{d\text{C}}_{\text{mcarb}}}{\mathit{dt}}=F_{\text{subc}}-F_{\text{arcc}}-F_{\text{oibc}}$$ (10)

$$\frac{{d\text{C}}_{\text{morg}}}{\mathit{dt}}=F_{\text{subo}}-F_{\text{arco}}-F_{\text{oibo}}$$ (11)

$$\frac{{d\text{C}}_{\text{prm}}}{\mathit{dt}}={-(F}_{\text{OIBp}}+F_{\text{MORB}})$$ (12)

$${\text{O}}_{\text{atm}}={\text{C}}_{\text{org}}+{\text{C}}_{\text{morg}}$$ (13)

$$F_{\text{w}}=k{\text{C}}_{\text{atm}}$$ (14)

$$F_{carb}=F_w(1-f_{\mathrm{org}})$$ (15)

$$F_{org}=F_{\mathrm{w}}(f_{\mathrm{org}})$$ (16)

$$F_{\text{subc}}=\chi F_{\text{carb}}$$ (17)

$$F_{\text{subo}}=\chi F_{\text{org}}$$ (18)

$$F_{\text{arc}}=F_{\text{arcc}}+F_{\text{arco}}$$ (19)

$$F_{\text{arcc}}={\mathrm{\alpha }}_{\text{carb}}{F}_{\text{subc}}$$ (20)

$$F_{\text{arco}}={\mathrm{\alpha }}_{\text{org}}{F}_{\text{subo}}$$ (21)

$$F_{\text{OIB}}=F_{\text{OIBc}}+F_{\text{OIBo}}+F_{\text{OIBp}}$$ (22)

$$F_{\text{OIBc}}=(1-{\mathrm{\alpha }}_{\text{carb}})F_{\text{subo}}$$ (23)

$$F_{\text{OIBo}}=(1-{\mathrm{\alpha }}_{\text{org}})F_{\text{subo}}$$ (24)

$$F_{\text{OIBp}}={\kappa }_1$$ (25)

$$F_{\text{MORB}}={\kappa }_2$$ (26)

$$F_{\text{out}}=F_{\text{OIB}}+F_{\text{MORB}}+F_{\text{arc}}$$ (27)

A main focus of the model is to demonstrate how δ¹³C_carb evolves in response to perturbations to carbon fluxes. Therefore, we track the δ¹³C evolution of different carbon reservoirs and fluxes with the following set of equations:

$${\mathrm{\delta }}^{13}{\text{C}}_{\text{atm}}(t)=\frac{F_{\text{OIB}}}{F_{\text{out}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{OIB}}(t)+\frac{F_{\text{arc}}}{F_{\text{out}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{arc}}(t)+\frac{F_{\text{MORB}}}{F_{\text{out}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{MORB}}(t)$$ (28)

$${\mathrm{\delta }}^{13}{\text{C}}_{\text{OIB}}(t)=\frac{F_{\text{OIBc}}}{F_{\text{OIB}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{carb}}(t-{\tau }_{\text{OIB}})+\frac{F_{\text{OIBo}}}{F_{\text{OIB}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{org}}(t-{\tau }_{\text{OIB}})+\frac{F_{\text{OIBp}}}{F_{\text{OIB}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{prm}}$$ (29)

$${\mathrm{\delta }}^{13}{\text{C}}_{\text{arc}}(t)=\frac{F_{\text{arcc}}}{F_{\text{arc}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{carb}}(t-{\tau }_{\text{arc}})+\frac{F_{\text{arco}}}{F_{\text{arc}}}{\mathrm{\delta }}^{13}{\text{C}}_{\text{org}}(t-{\tau }_{\text{arc}})$$ (30)

$${\mathrm{\delta }}^{13}{\text{C}}_{\text{carb}}(t)={\mathrm{\delta }}^{13}{\text{C}}_{\text{atm}}(t)+5$$ (31)

$${\mathrm{\delta }}^{13}{\text{C}}_{\text{org}}(t)={\mathrm{\delta }}^{13}{\text{C}}_{\text{atm}}(t)-20$$ (32)

where ${\mathrm{\delta }}^{13}{\text{C}}_{\text{i}}(t\text{)}$ is the ${\mathrm{\delta }}^{13}{\text{C}}_{\text{i}}$ value for carbon reservoir or flux i at time t, with atm = atmospheric reservoir, carb = carbonate reservoir, org = organic carbon reservoir, prm = primitive mantle reservoir, arc = arc flux, and OIB = ocean island flux, and MORB = mid-ocean ridge flux. We set ${\mathrm{\delta }}^{13}{\text{C}}_{\text{prm}}$ to a constant value of −5‰ for the duration of the model, as we have assumed it has no influx of carbon from the surface reservoirs. The equations for ${\mathrm{\delta }}^{13}{\text{C}}_{\text{crb}}(t)$ and ${\mathrm{\delta }}^{13}{\text{C}}_{\text{org}}(t)$ are offset from ${\mathrm{\delta }}^{13}{\text{C}}_{\text{atm}}(t)$ by +5 and −20 respectively because we assume that Δ¹³C_crb-org = 25‰ and f_org = 0.20.

The model run to generate Fig. 5 was designed to simulate increased CO₂ emissions. To simulate this scenario, the model was run with the initial conditions given in supplementary Table 1. After evolving with no perturbations for 1 billion years, we prescribed that MORB emissions increase from 10¹³ to 10¹⁶ g C/Myr. In order to check whether the model is an acceptable representation of the global C cycle model we increased MORB emissions a second time to 10¹⁹ g C/Myr (similar to present day estimates⁴¹), and model-dependent reservoir sizes and fluxes are in line with present day estimates (Supp. Fig. 3). The increased CO₂ emissions cause increased deposition of carbonates and organic C, which drives increased subduction fluxes of both carbonates and organic C. The increased subduction flux of C coupled with the differential release of carbonates and organics at different volcanic settings described above results in increased oxygen levels and the positive C isotope excursion shown in Fig. 5.