10.2.1 Sources to Sinks and Back

Subduction zones act both as a source and a sink of carbon for the exosphere. Over geological timescales, the source is the volcanic flux of carbon degassed in continental and oceanic arcs, while the sink is the carbon subducted in both reduced (organic C) and oxidized (carbonate) forms. But subduction fluxes do not depend on tectonic processes and mantle convection only; they are also controlled by surface processes driven by solar energy and tied to the water cycle.Reference Campbell and Taylor²⁰ The overwhelming majority of carbon subducted is allochthonous,Reference Kelemen and Manning²¹ which means that most of it is added to marine sediments and the oceanic lithosphere from external sources – mainly the oceans and continents. The fate of carbon in subduction zones is therefore under tight oceanographic, biological, and geomorphic control (Figure 10.1), and it is especially sensitive to the partitioning of carbon between deep (pelagic) and shallow (e.g. shelves, oceanic plateaus) oceanic domains.Reference Caldeira²²

There are three main carbon sinks, or pumps, that mediate the transfer of allochotonous carbon from oceans to sediment and the oceanic lithosphere: the hydrothermal carbon pump, which stores CO₂ within the oceanic lithosphere during seafloor weathering; the soft-tissue pump, which leads to accumulation of soft organic tissues (of terrestrial or marine origins) in seafloor sediments; and the carbonate pump, which controls the topology of the carbonate compensation depth above which carbonate sedimentation on the seafloor may occur and below which carbonates are undersaturated and may lead to dissolution of old CaCO₃ already deposited on the seafloor.Reference Archer^23, Reference Zachos²⁴

10.2.2 Heterogeneity of Sedimentary Carbonate Subduction (Carbonate Pump)

Knowledge of carbonate heterogeneities on the seafloor and of carbon subduction rates mostly comes from recent decades of deep-sea drilling efforts.Reference Plank and Langmuir²⁵ Continuous core sampling of sedimentary covers and upper oceanic crust offers a direct window onto what rocks may eventually subduct and degas. Today, most marine carbonates form and accumulate in the Atlantic and Indian Oceans, where few subduction zones have formed yet, and to a lesser degree on top of the ridges of the southeast Pacific. Estimates of total annual carbonate subduction range between ~66 and 103 MtC/yr (Table 10.1), with about 60% present in sediments (~40‒60 MtC/yrReference Clift and Vannucchi^10, Reference Plank and Langmuir^25, Reference Clift²⁶) and 40% in the altered oceanic lithosphere (~26‒43 MtC/yr in basalts and serpentinitesReference Kelemen and Manning²¹). Carbonate subduction fluxes have been less than total carbonate deposition rates over the Cenozoic (Table 10.1), which suggests net accumulation of pelagic carbonate.

While efforts are underway to reduce these large uncertainties, the most important limitation for past and future predictions of carbon cycle dynamics is the heterogeneity of the flux (Figure 10.3): 25% of the global carbon subduction flux of the last 10 MyrReference Clift²⁶ has occurred in the Sunda trench. The importance of the Indonesian hot spot, as well as those of the Aegean and Makran (Figure 10.3), contrasts with the western Pacific, where subduction and degassing fluxes are generally low. But the southern Pacific New Hebrides–Solomon–Vanuatu segment does not conform to this rule. Unlike most settings of the western Ring of Fire, the slab is young and plunges to the east. The system is long and so is rich in carbonate,Reference Clift²⁶ which it is a sizable contributor to the global subducted carbon inventory (Figure 10.2). Allard et al.Reference Allard¹² showed that the Ambrym volcano in the Vanuatu islands pours out slab carbon at a record rate (~5–9% of the global volcanic carbon flux), providing a unique window onto the unusual CO₂ productivity of this young arc system.Reference Johnston, Turchyn and Edmonds²⁷

Figure 10.3

Contribution of selected subduction zones to global sedimentary C subduction flux. Fluxes are from Ref. Reference Clift26, where length of subduction zones and carbon (organic and inorganic) concentrations in trench sediments are compiled (see also Ref. Reference Plank and Langmuir25). Note that the computation of flux includes corrections for sediment porosity used but not reported in Ref. Reference Clift26. Rate of carbon subduction are provided in t/km/yr for each subduction zone. Organic carbon is in gray and black, and black circles denote OC hot spots. Inorganic carbon is in pale and dark blue, with dark blue denoting hot spots. The fractional contribution of each subduction to the global sedimentary C flux is indicated as a percentage of total inorganic (blue) or organic (black) carbon subduction, respectively, in the histogram of the right panel (cf. Table 10.1). Note that the flux associated with basalt, gabbro, and ultramafic carbonate is not included. This corresponds to an additional subduction rate of ~850 ± 200 tC/km/yr.

10.2.3 Hot Spot of Organic Carbon Subduction in the Sub-Arctic Pacific Rim (Soft Tissue Pump)

A heterogeneous distribution and diversity characterize the soft tissue pump,Reference Arndt²⁸ as well as the subduction of organic carbon (OC). The burial of subductable OC in pelagic environments represents a flux of up to 10–30 MtC/yr over the last 150 kyr (Figure 10.2),Reference Cartapanis, Bianchi, Jaccard and Galbraith²⁹ and perhaps during most of the Phanerozoic.Reference Hayes and Waldbauer² Though modest compared to net primary productivity (~100 Gt/yr),Reference del Giorgio and Duarte³⁰ this weak but continuous isolation of organic reductants in sediments and rocks is essential for the persistence of atmospheric O₂ levels that are so high and so far from thermodynamic equilibriumReference Krissansen-Totton, Bergsman and Catling³¹ over geological timescales.

Just like carbonate, OC subduction is geographically heterogeneous,Reference Clift²⁶ but its deposition hot spots are elsewhere (Figure 10.3). The trenches of Alaska, the Aleutians, and British Columbia contribute ~23% of global OC subduction, while many other trenches, such as those of the western Pacific and Central America, are comparatively marginal. This is either because the trenches are small or because sediments do not contain much OC (Figure 10.2). This heterogeneity seems primarily controlled by continental erosion, weathering, and marine sedimentation rates. The northern Pacific hot spot is dominated by subduction of thick sequences of terrigenous sediments.Reference Cui, Bianchi, Jaeger and Smith³² Surprisingly, all OC subduction hot spots are associated with the subduction of deep-sea turbiditic sequences, such as Surveyor fan (Alaska) and Indus fan (Makran) (cf. Figure 10.3). The OC fraction derived from terrestrial sources (petrogenic and biologicReference Galy, Peucker-Ehrenbrink and Eglinton³³) can locally reach more than 70% in active margin sediments.Reference Blair and Aller³⁴ This means that the geomorphic and climatic conditions that influence the export of organic-rich tropical, Arctic, and mountainous sedimentsReference Hilton^35, Reference Hilton³⁶ to the marine environmentReference Husson and Peters³⁷ are the same as those that control the pattern of OC subduction, too. Hence further research is needed to elucidate the link between continental configuration in the Wilson cycle, erosion/sedimentation patterns, and OC burial, with special attention paid to active margins.Reference Bao³⁸

Unlike carbonates, which conserve their structure across a broad range of pressure and temperature conditions of Earth’s interior, aging biomolecules display stunning plasticity in structure and compositionReference Beyssac, Rouzaud, Goffé, Brunet and Chopin^39–Reference Adam, Schneckenburger, Schaeffer and Albrecht⁴¹ acquired during transport and transformation through the lithosphere (Figure 10.2). It is often assumed that graphite is the main form of OC in rocks. However, organic macromolecules (e.g. kerogens) may not become proper crystalline graphite before the late stages of the subduction process (i.e. >600°C; Figure 10.1). The transformation of organic materials to their stable forms (e.g. graphite) involves a succession of metastable macromolecular intermediates called kerogens.Reference Helgeson, Richard, McKenzie, Norton and Schmitt⁴² Increasing temperature and the presence of water accelerates this process and drives the sequential release of CO₂ (i.e. reduction of the kerogen residue) and then hydrocarbons (i.e. oxidation of the kerogens) from the kerogen residue in shallow, unconsolidated sedimentsReference Berner⁴³ and in accretionary wedges (Figure 10.1).Reference Petrenko⁴⁴ This so-called carbonizationReference Helgeson, Richard, McKenzie, Norton and Schmitt⁴² may decrease the amount of OC effectively subducted by 20–40% or more depending on whether or not hydrocarbons are trapped within the descending slabs. This is still an open question.

$\begin{array}{ccccc}& {\mathrm{CO}}_2\uparrow ,{\mathrm{H}}_2\mathrm{O}\uparrow \mathrm{to}\text{atmosphere}& & {\mathrm{CH}}_4\uparrow ,{\mathrm{H}}_2\mathrm{O}\uparrow \mathrm{to}\text{atmosphere}& \\ \left[{\mathrm{CH}}_2\mathrm{O}\right]& \to & \text{kerogen}& \to & \text{graphite}/\text{diamond}\\ & \mathit{\text{reduction}}\left(\mathit{\text{abiotic}}/{\mathit{\text{biological}}}^{44}\right)& & \mathit{\text{oxidation}}& \end{array}$(10.1)

Subsequent collapse of the kerogen structure involves 3D ordering of the carbon-rich aromatic backbone and is favored by pressure and shear.Reference Oohashi, Hirose and Shimamoto⁴⁵ This structural ordering is called graphitization (Figure 10.1). First-principles approaches are now shedding light on the microscopic pathways involved in this process.Reference Berthonneau^46, Reference Weck⁴⁷ Because the transformations modeled by (10.1) are so slow in nature, it is even conceivable that the coldest slabs (e.g. the Mariana and the Aegean OC subduction hot spots) could reach the diamond stability field before completion of the carbonization and graphitization process. The graphitization process itself is irreversible, and field observations indicate that slabs may not lose much OC beyond 350–400°C.Reference Connolly and Galvez^48, Reference Zhang, Ague and Brovarone⁴⁹

The low reactivity of compact graphitic materials and their intricate weaving within poorly soluble mineral matrices explain why a graphitic fraction dislodged from rocks during erosion and weathering (18–104 MtC/yrReference Galy, Peucker-Ehrenbrink and Eglinton³³) is continuously exported to the ocean (Figure 10.1). How much of this fossil petrogenic material subducts is not yet known, but it may represent 50% of OC in the trenches of South America.Reference Blair and Aller³⁴

10.2.4 An Ancient Hydrothermal Carbon Sink

A last major contributor to allochthonous carbon incorporation to the oceanic lithosphere is hydrothermal carbonatization (Figure 10.1 and Table 10.1).Reference Alt⁵⁰ More details are provided in a companion chapter in this book (Chapter 15). In short, the circulation of cold and carbon-rich seawater through the oceanic crust and mantle dissolves cations from the oceanic lithosphere. Elevated concentrations of cations and rising temperature have been responsible for cycles of carbonate deposition in oceanic hydrothermal systemsReference Alt⁵¹ today and in the past. Occurrences of carbonatized basalts have been reported as early as in the early Archean CratonReference Nakamura and Kato⁵² of eastern Pilbara (Australia) and in Archean greenstone belts.Reference Shibuya, Komiya, Nakamura, Takai and Maruyama⁵³ They suggest that: (1) carbonatization of the lithosphere may have been the means by which the atmosphere lost its primordial CO₂Reference Ueda, Sawaki and Maruyama⁵⁴; and (2) carbonatization of the lithosphere may have been the only flux of global importance capable of balancing outputs from the deep Earth during the Archean, Proterozoic, and large swaths of the early Phanerozoic (Figure 10.1). Therefore, surface biogeochemical processes involving the water cycle may have controlled the magnitude and longevity of carbon inputs to subduction zones since their initiation, more than 3 Ga.Reference Dhuime, Hawkesworth, Cawood and Storey⁵⁵ The above shows that the pelagic reservoir is not at steady state over the Cenozoic. What about the subduction zone itself?

10.4.1 Dissolution by Rising Pressure and Temperature: Which Silicate Is in Charge?

An essential but implicit meaning of (10.2) is that the release of C from slabs is linked to that of water – the solvent and catalyst for the reaction. Carbon release tracks the main sequence of fore-arc and sub-arc mineral dehydration reactions.Reference McGary, Evans, Wannamaker, Elsenbeck and Rondenay⁶⁴ The most important of these reactions are the destabilization of lawsonite/epidote during eclogitization of the altered oceanic crustReference Poli, Franzolin, Fumagalli and Crottini⁶⁵ at ~450–700°C and the dehydration of antigorite/chloriteReference Ulmer and Trommsdorff⁶⁶ in the serpentinized oceanic mantle. Both take place at around the same temperature, but the second pathway occurs deeper in the subduction zone because deserpentinization and dechloritization reactions occur in the altered ultramafic rocks of the lower and usually colder section of the descending slab (Figures 10.1 and 10.4).

Figure 10.4

Devolatilization pattern (H₂O and C) in a subducting slab and its link to subduction efficiency. The latter is the fraction of subducted carbon released at fore-arc and sub-arcs depths (cf. Ref. Reference Volk3). The results are only qualitative and are based on ongoing studies investigating the coupling between C, H, Na, K, Si, and Al cycles in open subduction-zone systems.Reference Galvez, Connolly and Manning⁶⁷ The shallow output flux depends on the hydration structure of the oceanic lithosphere, and conceivably, on the degree of partial melting (red tones). Overall, this flux varies between ~15 and 50 MtC/yr (i.e. in the order of magnitude of arc emissionsReference Kelemen and Manning^21, Reference Johnston, Turchyn and Edmonds²⁷), between 0.1 and 0.6 of the incoming flux, and more likely ~0.4–0.5 in the Cenozoic (correspond to σ of ~0.5–0.6; cf. Figures 10.6 and 10.7). This ratio may have approached 0 in the Paleozoic and Mesozoic (i.e. before the rise of pelagic calcifiers). This should be considered in long-term models of Phanerozoic carbon cycle evolution.

Another insight from (10.2) is that the solubility of carbonates in hydrous fluids is primarily controlled by the relative stability of the various silicates (e.g. wollastonite, garnetReference Galvez⁴⁰) in which they transform; in other words, the rapid rise of C solubility at the onset of garnet formation in most sediment and mafic lithologiesReference Galvez, Connolly and Manning⁶⁷ reflects the increasing relative stability of calc-silicates with rising temperature:

$\begin{array}{ccccccccc}M-\text{carbonate}& +& \left(\mathrm{Al}\right)\text{silicate}& +& \text{solvent}& \to & M-\left(\mathrm{Al}\right){\text{silicate}}^{\mathrm{ss}}& +& \text{mixed carbonic fluid}\\ \left(\mathit{Ca},\mathit{Mg},\mathit{Mn},\mathit{Fe}\right){\mathit{CO}}_3& & \mathit{\text{quartz}}/\mathit{\text{lawsonite}}& & H_{2}O& & {\mathit{\text{CaSiO}}}_3/\mathit{\text{garnet}}& & \mathit{\text{dissolved}}{\mathit{CO}}_2\end{array}$(10.3)

This process is an incongruent dissolution,Reference Galvez, Connolly and Manning⁶⁷ where M is a divalent metal Ca, Mg, Fe, or Mn.

In practice, the congruent or incongruent nature of the dissolution depends on the kinetics of dissolution, the thermodynamic properties of calc–silicates, the kinetics of calc–silicate precipitation, and the rate of fluid transport.Reference Galvez^40, Reference Galvez, Connolly and Manning⁶⁷ Recent thermodynamic analysis, however, suggest that congruent pathways may be restricted only to very high pressures and low temperaturesReference Galvez, Connolly and Manning⁶⁷; regimes where fluid fluxes tend to be low. In addition, the ubiquitous presence of calc–silicates such as lawsonite down to 350°C in most silicic and aluminous lithologiesReference Tsujimori and Ernst⁶⁸ supports the idea that their formation is not kinetically limited in the subduction zone. Hence, field and thermodynamic evidence suggests that incongruent pathways have been the most important in driving the dissolution of carbon along most subduction geotherms, today and in the geological past).

This general rule may not hold in the case of dolostone and limestone lithologies. These rocks usually lack the alumina and, in the most extreme case, silica that promote vigorous dissolution (10.2 and 10.3). In addition, marble lithologies are also rather dry and also notably impermeable to fluid flow, and therefore not prone to dissolution. Yet dissolution reactions may also be sustained by chemical disequilibrium between contrasted lithologies even when pressure and temperature are invariant. For example, diffusion of volatiles (H₂OReference Ague⁶⁹ or H₂Reference Galvez⁴⁰) and nonvolatile elements (Al and SiReference Ague and Nicolescu⁷⁰) into the marbles may modify a(CO₂), or the stability of calc–silicates (10.2 and 10.3) and drive reactions (10.2) and (10.3). In fact, field evidence shows that when dissolution of marbles does occur, it tends to be restricted to the vicinity of fracturesReference Ague and Nicolescu⁷⁰ and tectonic discontinuities,Reference Ague^69,7Reference Sleep¹ precisely where fluids rich in Al and Si can circulate and boost carbon dissolution processes. Therefore, it is likely that incongruent pathways play the dominant role in carbon mobilization from slabs at subsolidus conditions. The growing importance of deep-sea carbonates (carbonate oozeReference Pimm⁷²) subduction in the next 100 million years (Table 10.1) makes them obvious targets for future research focusing on their peculiar mechanical properties and open-system behavior at elevated pressures and temperatures.

10.4.2 Carbonate Melts from Hot Slabs and Diapirs?

Partial melting of the slab may start at >900°C in the presence of water. It can affect the top section of hot slabs (e.g. Aleutians/Mexico) when it is infiltrated by water, and it may also occur in unusually hot and buoyant fragments of slab sediments in the mantle wedge.Reference Gerya and Yuen⁷³ The process forms ionic melts that are rich in carbonate components (i.e. hydrous carbonatites),Reference Skora⁷⁴ but the thermodynamic,Reference Kang, Schmidt, Poli, Franzolin and Connolly⁷⁵ structural,Reference Foustoukos and Mysen⁷⁶ and rheological properties of those compositionally heterogeneous melts remain largely unknown. Assessing their properties is important because carbonate melts are highly mobile, highly reactive,Reference Russell, Porritt, Lavallée and Dingwell⁷⁷ and important carriers of carbon beyond sub-arc depth anywhere between ~100 km and 400–600 km within the slab and in the mantle wedge. A few implications follow:

1. (1) In carbonate bearing lithologies, formation of metamorphic garnet (e.g. grossular), the ubiquitous silicate mineral of high-pressure metamorphism, may be the most practical and important indicator of both dehydrationReference Baxter and Caddick⁷⁸ and C redistribution (e.g. carbonate dissolution) between rocks and fluids via Urey-type processes (Figure 10.4).

2. (2) Carbon cycling in the slab is a problem of transportReference Frezzotti, Selverstone, Sharp and Compagnoni⁷⁹ where dehydration, decarbonation, desilicification, and dealuminification of rocks are parts of a single overarching problem.Reference Galvez, Connolly and Manning⁶⁷ Work is underway to link the microscopic properties of geo-fluids (aqueous, carbonate, silicate) and minerals with macroscopic patterns of element transport across shallow lithospheric reservoirs.Reference Connolly and Galvez^48, Reference Galvez, Connolly and Manning^67, Reference Galvez, Manning, Connolly and Rumble⁸⁰

3. (3) There is a growing consensus, informed by thermodynamic models, that pulses of carbon release from slabs by subsolidus dissolution do occur around 500–800°C, representing an overall flux of between ~10 and 40 MtC/yr between ~80 and 140 km depth below the volcanic arc.Reference Connolly and Galvez^48, Reference Gorman, Kerrick and Connolly^81, Reference Connolly⁸² This flux is uncertain, but overall it is consistent with volcanic emissions.Reference Kelemen and Manning²¹ The pulsatile character (Figure 10.4) of the flux, however, is primarily due to the timing and volume of H₂O infiltration rather than to the thermodynamics, or kinetics, of the Urey-type reactions. This is important because it means that the subduction efficiency σ, as opposed to total carbon release, is controlled by where H₂O is concentrated within the hydrated slab, rather than by the absolute amount of H₂O contained in it. Only the water contained in the uppermost 10 km of slab mediates the fast cycling of carbon in the fore-arc and sub-arc regions.

4. (4) The parameter σ may not be a mere parameter at all over geological timescales. It is more likely to be a variable that is dependent on the nature of carbon inputs. Indeed, we have shown that limestones and carbonatized lithosphere rocks behave differently, both mechanically and chemically. For example, while σ of ~0.4–0.8 in the Cenozoic (Table 10.1) is characterized by important subduction of limestones (Table 10.1), this value may have vanished during most of the Precambrian, Paleozoic, and early Mesozoic, stranding carbon in shallow surface reservoirs. Interestingly, this idea is qualitatively consistent with recent isotopic and geochemical proxies,Reference Hirschmann⁸³ which suggest a net growth of the continental and pelagic carbonate reservoir (Table 10.1) over the last 2 Gyr.Reference Hayes and Waldbauer^2, Reference Hirschmann⁸³

10.4.3 Where Is the Barrier to Deep Carbon Subduction?

Global budgets are sensitive to thermal models and the permeability/rheology of rocks within the slab. Therefore, slabs that are cold and/or covered by thick and impermeable carbonate layers should retain a significant fraction of their carbon cargo to beyond sub-arc depths. What is the fate of carbon beyond this limit (Figure 10.1)?

If carbon makes it beyond sub-arc depths (~150 km), it may not push much further beyond the transition zone (~410–660 km). The reason for this is the existence of a deep depression in the temperature (1000–1100°C) at which anhydrous carbonated oceanic crust melts at uppermost transition-zone conditions.Reference Thomson, Walter, Kohn and Brooker⁸⁴ Thomson et al. showed that the curvature of the solidus of carbonated eclogite is also controlled by the partitioning of nonvolatile elements, chief among them calcium and sodium, between mineral phasesReference Thomson, Walter, Kohn and Brooker⁸⁴ in high-pressure eclogites. Most geotherms should intersect this melting barrier.

To the best of our knowledge, there is as yet no experiment that quantifies the effect of water on the topology of the anhydrous solidus of carbonated eclogites in the vicinity of the deep curvature (i.e. at the conditions of the transition zone around ~10–20 GPa). Qualitatively at least, it is now undisputed that increasing activity of water a(H₂O) dampens the solidus of carbonate rocks,Reference Skora^74, Reference Wyllie and Tuttle⁸⁵ bringing their temperature of melting closer to 900–1000°C, promoting the formation of carbonate melts. This principle applies to carbonatized basaltic lithologiesReference Poli⁸⁶ in particular and to all carbonate-bearing systems in general, typically those rich in iron and manganeseReference Kang and Schmidt⁸⁷ that are relevant for the fate of subducted banded iron formation in the Archean Earth. This should limit even more the possibility of carbon subduction through the transition zone (Figure 10.1).

10.4.4 Are Thermal Anomalies the Norm?

There is a growing consensus that the relevant pressure–temperature trajectories of slab materials may be hotter than most canonical models assume.Reference Gorman, Kerrick and Connolly⁸¹ First of all, there is an apparent incompatibility between slab thermal models predicted by theoryReference Syracuse, van Keken and Abers⁸⁸ – usually cold – versus those inferred from phase assemblage relations and mineral chemistryReference Penniston-Dorland, Kohn and Manning⁸⁹ – generally warmer. Second, the mantle wedge may be populated by detached fragments of slab sediments (e.g. diapirsReference Gerya and Yuen⁷³) that are thought to follow hotter pressure–temperature trajectories than the rest of the slab (Figure 10.1). It is important because it implies that the production of hydrous carbonatite melts may be more common than expected, particularly for cold slabs (e.g. Aegean) covered by kilometer-thick buoyant sedimentary piles. Overall, research into the thermal regime of slab materials may eventually redefine what thermal normality is in the subduction environment.

Taken together, the possibility of paths that are hotter than normal for most subducted sediments and the existence of a chemical barrier to deep carbonate subduction (Figure 10.1) means that most subducted carbon must have been stranded above the transition zone through most of Earth’s history.Reference Kelemen and Manning²¹ However, there remain important uncertainties on the subduction efficiency, in particular, its variability in space and time. This uncertainty hinders quantitative understanding of the response dynamics of the carbon cycle to perturbations.

10.5.1 Pressure and Temperature

For the same reason that rising pressure and temperature enhances the solubility of carbonates in subducting slabs, downward temperature paths, too, may lead to carbonate precipitation out of fluids ascending along and across the slab. This chromatographic process is predicted thermodynamicallyReference Ague⁶⁹ and is observed in the field.Reference Piccoli, Brovarone, Beyssac, Martinez, Ague and Chaduteau⁹² This illustrates that there is a delay between the time carbon is dissolved from the slab and the time it eventually reaches the surface.

10.5.2 Low- and High-Temperature Redox Processes

The redox state, too, is important for the fate of both carbonates and organic materials, and it is usually measured by the thermodynamic activity of oxygen. Carbon can bond to both oxygen and hydrogen, and the balance between oxygenated species (e.g. CO₂) and hydrogenated carbon species (e.g. CH₄) in a fluid–rock system depends on the nature and abundance of other “redox” elements in the rock, such as Fe (Fe²⁺, Fe³⁺), Mn, and multiple S species susceptible to exchanging electrons with carbon.Reference Tumiati, Godard, Martin, Malaspina and Poli⁹³ But there is a limit to the amount of C a fluid can contain at typical subduction-zone conditions. This limit is fixed by thermodynamics, and varies with pressure, temperature, and rock composition.Reference Galvez, Connolly and Manning^67, Reference Connolly and Cesare⁹⁴ If graphitic materials are stable – as is usually the case in subducted sediment (Figure 10.4) – then the so-called graphite-saturated COH system imposes a minimum in carbon solubility midway between fluids dominated by CO₂ and fluids dominated by CH₄.Reference Connolly and Cesare⁹⁴ The existence of such a minimum is important because it implies that quasi-static changes in pressure, temperature, a(H₂O), or fO₂ of a graphite- or diamond-saturated fluid could involve both carbon dissolution and/or precipitation of the fluid. Therefore, C may be locked in the form of graphite during fluid ascent through the slabReference Galvez^40, Reference Brovarone⁹⁵ or through the continental crust,Reference Weis, Friedman and Gleason^96, Reference Rumble, Duke and Hoering⁹⁷ creating complex locking and unlocking pathways for deep carbon (Figure 10.5).

Figure 10.5

Redox pathways in the subduction zone. Graphite precipitation from carbonate at a lithological interface. (a) Field image of the outcrop in Alpine Corsica.Reference Galvez⁴⁰ (b) Representative COH diagram showing the curvature of the C-saturation surface at the elevated pressures and temperatures typical of subduction zones, illustrating various pathways leading to elemental carbon precipitation.

Fluctuations in redox conditions may occur anywhere from the shallow to the deep subduction zone. For example, it has been shown that serpentinite assemblages maintain their reducing power in the shallow subduction zone at conditions lower than 10 kbar and 500°C.Reference Malvoisin, Chopin, Brunet and Galvez⁹⁸ Those conditions may cause the spontaneous formation of refractory graphite from carbonate according to:

$\begin{array}{ccccccc}{\mathrm{CO}}_2& +& 4\mathrm{FeO}& =& 2{\mathrm{Fe}}_2{\mathrm{O}}_3& +& {\mathrm{C}}^0\\ \left(\mathit{\text{carb}}\right)& & \left(\mathit{mag}/\mathit{\text{serp}}\right)& & \left(\mathit{\text{garnet}}\right)& & \left(\mathit{\text{graphite}}\right)\end{array}.$(10.4)

There is now evidence for such a mechanism occurring at temperatures of no more than 450°C (Figure 10.4).Reference Galvez^40, Reference Galvez⁷¹ A redox mechanism formally analogous to (10.4) is thought to occur below 250 km, too, where carbonatite melts expelled from the slab re-equilibrate with mantle lithologies where metallic iron is stable (Figure 10.1).Reference Rohrbach and Schmidt⁹⁹ Mutual re-equilibration between carbonatite melts and mantle rocks involves the spontaneous production of diamond according to:

$\begin{array}{ccccccc}{\text{MgCO}}_3& +& 2{\mathrm{Fe}}^0& =& 3\left({\mathrm{Mg}}_{0.33},{{\mathrm{Fe}}^{2+}}_{0.67}\right)\mathrm{O}& +& {\mathrm{C}}^0\\ \left(\mathit{\text{melt}}\right)& & \left(\mathit{\text{metallic iron}}\right)& & \left(\mathit{\text{iron oxide}}\right)& & \left(\mathit{\text{diamond}}\right)\end{array}.$(10.5)

10.5.3 SiO₂ Activity

Just like a(SiO₂) influences the dissolution of C in hydrous fluids of slabs via Urey-type reactions (cf. (10.2) and (10.3)), fluctuations of a(SiO₂) impact the behavior of C in all types of subduction-zone liquid in which it may dissolve. This is particularly true of carbonate and other alkaline melts. Experimental worksReference Russell, Porritt, Lavallée and Dingwell^77, Reference Dalton and Wood¹⁰⁰ have shown that dolomitic melts formed in the descending slab degas at the contact of peridotitic mineral assemblages characterized by comparatively higher a(SiO₂) (orthopyroxene and clinopyroxene).

$\begin{array}{lllllll}{{2\mathrm{CO}}_3}^{2-\text{melt}}& +& {{\mathrm{SiO}}_2}^{\text{rock}}& =& {{\mathrm{SiO}}_4}^{4-,\text{melt}}& +& {{2\mathrm{CO}}_2}^{\mathrm{gas}}\\ \text{base}\left(\text{carbonatitc melt}\right)& & \text{acid}\left(\text{mantle rock}\right)& & \text{base}\left(\text{carbonatitic melt}\right)& & \text{acid}\left(\text{mantle}\mathrm{gas}\right)\end{array}.$(10.6)

This mechanism may supply large fluxes of diffuse CO₂ outgassing with decompression, as the melt dissolves more silica or any other acidic component. Therefore, cataclysmic eruptions (kimberlitesReference Jones, Genge and Carmody¹⁰¹) may ensue if those melts stall for long periods of timeReference Sleep¹ in environments such as the base of the continental crust.

10.5.4 Water

Just like surface processes, many processes of the slab–mantle wedge interface involve repeated cycles of fluid desiccation, whereby fast and near-quantitative redistribution of H₂O from fluids toward solid (e.g. mantle wedge serpentinization) or melt phases occurs. Desiccation is a possibly dominant mode of elemental carbon sequestration in the crust and upper mantle, although this mechanism has received little attention so far. Because the solubility of water in silicate melts that are poor in alkalis is greater than that of carbonic species,Reference Stolper, Fine, Johnson and Newman^102, Reference Mysen¹⁰³ rehydration (below 600°CReference Pirard and Hermann¹⁰⁴) and flux melting of the mantle wedge by infiltrating COH fluids may form restitic graphite/diamond (Figure 10.5b) via:

This mechanism is supported by theory (Figure 10.5); evidence for it in nature may be found in the Kokshetav massif (north Kazakhstan)Reference Korsakov and Hermann¹⁰⁵ or in Lianoning (northeast China),Reference Zhang, Yu and Jiang¹⁰⁶ where the association of graphite with jadeites is intriguing.

Overall, a combination of physicochemical processes drive not only carbon loss from slabs, but also its reactivity and fate in the heterogeneous chromatographic columns that separate slab liquids from the surface. Therefore, a carbon flux at the surface of Earth (or of any planet) does not mean an active release is operating.

10.7.1 Biological Evolution Influences Carbon Subduction: First Milestone

Genomic studiesReference Soo, Hemp, Parks, Fischer and Hugenholtz¹²³ and isotopic proxiesReference Krissansen-Totton, Buick and Catling¹⁹ place the origin of autotrophic life and respiration – the main carbon source and sink today – at some time in the Archean, but life alone did not affect the geological cycle of carbon as much as mortality did. Viral lysis of prokaryotesReference Bidle and Falkowski^124, Reference Fuhrman¹²⁵ and, possibly, genetically encoded death pathwaysReference Bidle and Falkowski¹²⁴ may have played a disproportionate role in driving prokaryote mortality and, therefore, a primordial soft-tissue pump that helped the early biological necromass to spread across all marine habitats, including trenches and subduction zones. The ancestral origin of virusesReference Forterre and Prangishvili¹²⁶ is at least qualitatively consistent with the idea that a sustained flux of dead organic materials from the sunlit upper ocean to the marine sediments may have been established at some time in the Archean. While it is conceivable that carbonatization of the oceanic crust, and its subduction, had been in place ever since oceans and the oceanic crust coexisted, it is the marine accumulation, isolation, accretion, and subductionReference Duncan and Dasgupta^127, Reference Godderis and Veizer¹²⁸ of organic reductants that sparked an OC cycle that was truly geological in scale.

10.7.2 Biological Evolution Influences the Dioxygen Cycle: Second Milestone

About 2.4 Gyr ago, the accumulation of OC in sediments, continents, and the mantle through subduction zonesReference Godderis and Veizer¹²⁸ became, for the first time, tied to the production and eventual accumulation (about 2.33 Ga)Reference Hayes and Waldbauer² of dioxygen gas in the lithosphere (in the form of iron oxides) and then in the exosphere.Reference Knoll¹²⁹ The evolution of the water-splitting systemReference Fischer, Hemp and Johnson¹³⁰ exploited, for the first time, the abundance of the water molecule as an almost unlimited source of protons and electrons for photosynthesis, releasing O₂ as a highly reactive waste product.Reference Falkowski and Godfrey¹³¹

This redox pathway enhanced the stability of atmospheric CO₂ (causing major glaciation, Figure 10.7), acidified riverine waters,Reference Halevy and Bachan¹³² and therefore may have boosted the weathering flux of essential nutrients (e.g. trace elements, phosphorus, etc.Reference Anbar¹³³) from proto-continents to oceans. Life, therefore, exerted its first indirect control of the cycle of alkalinity and caused large-scale perturbations in the carbon cycle (Figure 10.7). A more subtle effect of this gradual redox shift was that it contributed to locking water on Earth, instead of water slowly escaping into spaceReference Catling, Zahnle and McKay¹³⁴ as it did on Venus.Reference Driscoll and Bercovici¹³⁵ This is important because water drives and lubricates the subduction-zone system and, through it, the entire carbon cycle. Therefore, the rise of redox disequilibrium between the fluid and solid compartments of Earth may have laid the ground for the exceptional longevity and vigor of Earth’s subduction carbon cycle.

10.7.3 Biological Evolution Influences the Cycle of Alkalinity: Third Milestone

About 1.5 billion years later, the biosphere enhanced its control on the geological cycle of alkalinity. The invasion of continents by algae and plants with roots in the early Paleozoic enhanced erosion and continental weathering.Reference Schwartzman and Volk^136, Reference Hemingway¹³⁷ Enhanced sediment export to the ocean boosted marine sedimentation of OC (Figure 10.7),Reference Husson and Peters^37, Reference Peters and Gaines^138, Reference Berner¹³⁹ yet another potential process feeding back into the subduction carbon cycle.

10.7.4 Alkalinity Feeding Back into Biological Evolution: Fourth Milestone

The rising flux of dissolved weathering products (e.g. Ca) may have promoted further biological innovations. The geological record suggests algae evolved new ways to exploit the carbonate oversaturation state of the oceans over the Mesozoic.Reference Monteiro¹⁴⁰ By accelerating chemical reactions, these organisms were able to reroute excess oceanic alkalinity to the carbonate skeleton and ultimately to rocks. The expansion of calcifiers to the distal parts of the oceans over the last 150 Myr boosted deep seafloor carbonate deposition and subduction. Therefore, the idea that a clear-cut dichotomy separates Earth’s biological processes (dominated by redox processes) and abiotic geochemical reactions (dominated by proton transfer chemistry)Reference Falkowski, Fenchel and Delong¹⁴¹ does not hold in the context of geological time (Figure 10.2). Fascinatingly, much of the biotic carbonate of the Atlantic and Indian oceans has yet to be subducted; when will the Atlantic Ring of Fire form? We do not know.

10.7.5 Response of Climate to the Enhanced Subduction of Pelagic Carbonates

The geological response to pelagic calcifiers is still very much ahead of us. The models of Figures 10.6 and 10.7 illustrate potential scenarios and the timescales involved. They illustrate how subduction zones are components of a larger dynamic system that includes continents, oceans, and biosphere. Figure 10.8 is simplistic, and it builds on much previous workReference Volk^3, Reference Caldeira⁸; details are provided in the Supplementary Online Materials.

Figure 10.8

Non-steady-state response of the carbonate cycle to geodynamic and biological forcing: model design. The pelagic (M_p) and continental (M_c) carbonate reservoirs evolve by exchange of C (i.e. fluxes F_i) with the atmosphere and mantle (Figure 10.6). Mantle is assumed to be a reservoir of infinite residence time. The silicate weathering flux F_sw is linked to atmospheric temperature by a simple polynomial relation from Ref. Reference Berner, Lasaga and Garrels61. The C fluxes F_i are linked to M_i via rate constants k_i (e.g. F_cm = k_cmM_c, and F_sub = k_subM_p). The initial steady state is obtained with starting conditions chosen and updated from Ref. Reference Volk3 to be consistent with present-day values for C fluxes and reservoir sizes: F_i = 3 Tmol/yr, F_sw,i = 10 Tmol/yr (Ref. Reference Caves, Jost, Lau and Maher152); F_cw,i = 15 Tmol/yr (Ref. Reference Caves, Jost, Lau and Maher152); k_cm = 0.00035, k_sub = 0.0072, M_c,i = 7.583 × 10²¹ mol,Reference Hirschmann⁸³ M_p,i = 0.917 × 10²¹ mol,Reference Hirschmann⁸³ σ = 0.6 (see above), β = 0.25,Reference Volk³ ζ = 0. The system is solved analytically to find its steady state, which is close to present-day condition (i.e. M_c = 7.909 × 10²¹ mol (residence time τ_c ≈ 2800 Myr); M_p = 0.810 × 10²¹ mol (residence time τ_p ≈ 140 Myr); cf. Supplementary Online Materials). This steady state is then used for perturbation analysis (Figure 10.7).

We focus on the geological response of the carbonate cycle and atmospheric temperature to modifications in key geodynamic or biological parameters: β, the partitioning of carbonate between the pelagic and continental (accretion) environment; σ, the fraction of subducted carbon transported beyond sub-arc depths (Figure 10.3); and ζ, a self-amplification factor corresponding to the fraction of continental carbonate degassing that is controlled by arc magmatism. The pelagic (M_p) and continental (M_c) carbonate reservoirs evolve by exchange of C (fluxes F_i) with the atmosphere and mantle (Figure 10.6). The silicate weathering flux F_sw is linked to atmospheric temperature by a simple polynomial relation.Reference Berner, Lasaga and Garrels⁶¹ Starting from a set of initial conditions (steady state), the evolution of the system described in Figure 10.8 obeys a set of coupled differential equations:

For the sake of simplicity, we also impose that the surface reservoir evolves quasi-statically (i.e. F_sw responds instantaneously to any change in input flux; e.g. mid-ocean ridge degassing (F_j), the fraction of subducted carbon degassed in the fore-arc and sub-arc ((1 – σ)F_sub), or the flux of continental carbon degassing (F_cm)), which gives:

Perturbation analysis reveals a few important patterns (Figure 10.6). First, the shape and relaxation time varies over two orders of magnitudes. It is shorter when perturbation mostly affects the pelagic reservoir where residence time is short (e.g. σ) and it is longer when it affects both pelagic and continental reservoirs (e.g. β, ζ). Modification of the rate of subduction (k_sub) may also induce very long relaxation times because it affects the degassing of continental carbonates by arc magmatism (rate k_am) via ζ and thus M_c. Second, the non-steady-state response to augmentation (e.g. in β) involves an initial stage of atmospheric heating (β = 0.25–0.37 leads to ~2°C warming over the next 500 Myr), followed much later by relaxation to an atmospheric temperature that is cooler than the initial condition.

The important point is that paleoclimates most likely reflect a non-steady-state dynamic of the carbon cycle. This behavior prevails over timescales that can reach hundreds of millions of years (Figure 10.9). For example, we find that the signal corresponding to the time evolution of continental OC burial from Husson and PetersReference Husson and Peters³⁷ contains at least two salient modes at ~0.08 and ~0.4 fHz, equivalent to periods of 360 and 74 Myr, respectively (Figure 10.9d), which are typical of tectonic events; the former is most likely linked to supercontinent cycles. Surprisingly, biological evolution may cause long-term non-steady-state responses from the geological carbon cycle, too (Figure 10.9c). In this case, the timescale of the response is not controlled by the biosphere, but by the large residence time of carbon in lithospheric reservoirs, particularly continents.

Figure 10.9

Non-steady-state response of the carbonate cycle to geodynamic and biological forcing: perturbation analysis. (a) Average atmospheric temperature obtained by solving the differential system of equations analytically (cf. Supplementary Online Materials) for values of β ranging from 0.1 (continental mode) to 0.9 (pelagic mode). (b) Size of continental (M_c) and pelagic (M_p) reservoirs as a function of time after initial perturbation of the steady state (Figure 10.6). At time 0, β is subjected to a step rise from 0.25 to 0.37 and the system is left to evolve. The transient values of M_c and M_p are tracked for 5 Gyr. (c) (Upper Panel) The shape and response time of the system (average temperature) for perturbations in β, σ, k_cm, and k_sub are compared. It is often the non-steady-state response of the system that matters when studying the long-term climatic (and isotopicReference Caves, Jost, Lau and Maher¹⁵²) impact of geodynamic transitions and/or biological innovations. (Lower Panel) Magnification of the 400 Myr following the perturbation, with the approximate location of the present time with respect to the mid-Mesozoic revolution.Reference Ridgwell⁶ (d) Pulse of the OC cycle on Earth over geological timescales. Fourier transform of the OC burial flux of Ref. Reference Husson and Peters37 showing two dominant main modes. Typical frequencies are in femtohertz. Surprisingly, the signature of the ’supercontinental cycle at around 0.08 fHz seems to be significant, despite the important noise of the signal. FFT = fast Fourier transform.