CO₂ outgassing during collisional orogeny is facilitated by the generation of immiscible fluids

Abstract

Orogenic degassing is emerging as a potentially relevant source of carbon dioxide (CO₂) from the continental crust. However, the processes of carbon mobilization are still poorly explored. Here, we use thermodynamic modeling to investigate the decarbonation of sediments metamorphosed under high geothermal gradients. Our modeling shows that immiscible CO₂-rich vapors and hydrosaline brines are generated at these conditions, with different properties and mobility through the crust. The CO₂-rich fluid fraction could rapidly rise toward the surface without interacting with the host rocks by carbo-fracturing the host rocks or through deep faults. The denser hydrosaline brines likely permeate the source rocks. When applied to the active Himalayan orogen, these observations reconcile measured CO₂ fluxes at the surface and positive conductivity anomalies associated with micro-seismicity at depth. Our modeling shows that the continental crust represents a relevant reservoir of CO₂ that can be efficiently degassed during hot collisions.

Introduction

Although the global Earth’s natural CO₂ degassing is considered irrelevant compared to anthropogenic emissions, much effort is devoted to assessing our planet contribution to the global carbon cycle. Knowing the rate of the Earth’s natural CO₂ emissions is, in fact, crucial for quantifying and predicting the influence of anthropogenic perturbations on climate processes¹. It is well known that metamorphism of crustal rocks in collisional orogens produces carbon-bearing fluids over geologic time scales, primarily through decarbonation reactions^2,3,4,5,6,7. However, the potential contribution of orogenic processes to the global Earth’s carbon cycle is a long-standing open question^{5,8,9,10,11,12}. The efficiency of metamorphic CO₂ degassing in orogenic settings is primarily controlled by the ease at which the CO₂-bearing aqueous fluids produced at depth are transported upward without interacting with the host rocks. Fluid-rock interactions following prograde metamorphic devolatilization reactions can lead to carbon re-precipitation typically as carbonate^{13,14,15,16,17,18,19,20,21} or graphite^22,23, thus hampering significant CO₂ outgassing at the surface. The ability (or inability) of deep CO₂ to migrate upward and reach the Earth’s surface opens up two possible scenarios. If retrograde metamorphism represents a sink for CO₂¹², collisional orogens’ contribution to the global Earth’s CO₂ degassing would be minimal. Conversely, if metamorphic CO₂-bearing fluids do not react with the host rocks, orogenic decarbonation would represent an important source of CO₂ at the global scale. Recent research⁷ balancing estimates of decarbonation rates from deeply exhumed rocks^24,25,26 and CO₂ fluxes measured at the surface^9,27,28, support the second hypothesis. However, neither were the physical and chemical characteristics of the released metamorphic fluids studied in detail nor was the transport path of deep CO₂ fully elucidated.

Here, we study decarbonation reactions occurring in carbonate-bearing sediments along high geothermal gradients to define the physical and chemical nature of fluids that transport carbon in large hot orogens. We find that the productivity of CO₂-rich fluids is maximum for carbonate-bearing sediments originally containing low to moderate amounts of calcite (10–30 vol%) and metamorphosed at medium- to high-temperature (T) (i.e., T > 600 °C) and at medium pressure (P) (i.e., P > 8 kbar) conditions. We present petrological evidence that most dehydration and decarbonation reactions in these lithologies occur at P–T conditions in the two-phase field of the H₂O–CO₂–salt ternary fluid systems, generating immiscible CO₂-rich vapors and hydrosaline brines. Our results demonstrate a crucial role of fluid immiscibility in driving CO₂ transport from the deep crust, explaining how and why significant amounts of CO₂ could be effectively degassed at the surface from orogenic belts. This study reconciles geophysical and geochemical observations from active collisional orogens such as the Himalaya, where intense CO₂ degassing is currently measured at the surface. Results further highlight the role of hydrosaline brines as metasomatizing and/or granulitizing agents in the lower crust.

Results

Fluid production and evolution in large hot orogens

In large hot orogens (i.e., mountain belts characterized by high crustal temperatures and extreme crustal thickening^29,30), CO₂ sources are represented mainly by carbonate-bearing sediments ranging from calcareous pelites to marls and impure limestones, important constituents of thick sedimentary sequences deposited along passive margins³¹. During prograde metamorphism along medium to high dT/dP gradients, these lithologies are transformed into calcic metapelites, calc-silicate rocks, and silicate-bearing marbles and generally show an internally buffered behavior^24,31,32,33. Phase equilibria modeling of carbonate-bearing sediments obtained as “synthetic” mixtures of variable proportions of calcite (i.e., 10, 30, 50, 70 vol%, corresponding to initial bulk compositions Cal10, Cal30, Cal50, and Cal70; Table 1) with an average pelite³⁴ allows predicting the fluid evolution and compositions along a dT/dP geothermal gradient typical of large hot orogens (i.e., 750 °C GPa⁻¹, ca. 25 °C km⁻¹ (see ref. ³⁵)). The results of the modeling are illustrated in Figs. 1 and 2.

Table 1 Modeled bulk compositions (mol%) and approximate protolith’s mineral compositions (vol%).

Fig. 1: Modal variations and wt% of CO₂ and H₂O released during prograde metamorphism of Cal10 and Cal30 bulk compositions.

a, c Mode box changes in calculated mineral proportions (vol%, fluid excluded) during prograde metamorphism of sediments initially containing (a) 10 vol% of calcite (Cal10) and (c) 30 vol% of calcite (Cal30) along the modeled internally buffered P/T–X(CO₂) paths. b, d Loss (wt%) of CO₂ (black line) and H₂O (gray line) relative to the initial CO₂ and H₂O content in the solids, and X(CO₂) of the fluid (blue lines) released during prograde metamorphism of Cal10 (b) and Cal30 (d), along the same internal buffered P/T-X(CO₂) path as in (a) and (c). Events I (yellow) and II (red) refer to the main events of CO₂ production, as discussed in the text. Dashed lines highlight the main dehydration and decarbonation reactions, corresponding to abrupt consumption and/or complete disappearance of hydrous and CO₂-bearing minerals. Reactions corresponding to the event I (R1) and event II (R2a, R2b) are specified.

Fig. 2: Modal variations and wt% of CO₂ and H₂O released during prograde metamorphism of Cal50 and Cal70 bulk compositions.

a, c Mode box changes in calculated mineral proportions (vol%, fluid excluded) during prograde metamorphism of sediments initially containing (a) 50 vol% of calcite (Cal50) and (c) 70 vol% of calcite (Cal70) along the modeled internally buffered P/T–X(CO₂) paths. b, d Loss (wt%) of CO₂ (black line) and H₂O (gray line) relative to the initial CO₂ and H₂O content in the solids, and X(CO₂) of the fluid (blue lines) released during prograde metamorphism of Cal50 (b) and Cal70 (d), along the same internal buffered P/T–X(CO₂) path as in (a) and (c). Events I (yellow) and II (red) refer to the main events of CO₂ production, as discussed in the text. Dashed lines highlight the main dehydration and decarbonation reactions, corresponding to abrupt consumption and/or complete disappearance of hydrous and CO₂-bearing minerals. Reactions corresponding to the event I (R1) and event II (R2b) are specified.

In carbonate-bearing metasediments, volatiles (H₂O and CO₂) are primarily hosted in phyllosilicates and carbonates; scapolite can represent an additional, often neglected, CO₂ reservoir^24,25,31,36. Prograde metamorphism generates C–O–H fluids dominated by H₂O and CO₂ components, the amounts of the other molecular species (CO, CH₄, and H₂) being negligible (i.e., <0.005 mol%) for all the investigated bulk compositions (Supplementary Data Files S1–S4). The aqueous component of the released C–O–H fluids is generated chiefly by the breakdown of chlorite, muscovite, biotite and/or epidote, and the CO₂ component by calcite and/or scapolite consumption. The interplay between formation and consumption of these phases controls the relative amounts of H₂O and CO₂ that can be released at increasing P–T conditions (i.e., if biotite forms during the muscovite breakdown, part of the H₂O is transferred from muscovite to biotite and, similarly, if scapolite formation coincides with calcite consumption, part of the CO₂ produced through calcite breakdown is fixed in scapolite). The results of the modeling show that, independently from the original amount of calcite in the sedimentary protolith, the production of C–O–H fluids in internally buffered carbonate-bearing metasediments mainly occurs in pulses (Figs. 1b, d and  2b, d), i.e., through nearly discontinuous reactions operating in narrow P–T intervals (see also refs. ^24,25,31).

Two main events of CO₂ production are predicted at 575–605 °C (event I) and 630–670 °C (event II), respectively. For any of the modeled initial bulk compositions, event I corresponds to the muscovite-out reaction R1: Cal + Qz + Mu + Pl = Kfs + Scp + F, nearly coinciding with the total consumption of calcite for Cal10 bulk composition. Event II corresponds to the scapolite-out reaction R2a: Bt + Qz + Scp = Kfs + Cpx + Grt + Pl + F for Cal10 bulk composition; for Cal30, Cal50 and Cal70 bulk compositions, event II coincides with the biotite-out reaction R2b: Cal + Bt + Qz + Scp = Kfs + Cpx + Pl + F (Figs. 1 and  2). At each event, the fluid composition (X(CO₂) = molar fraction of CO₂) varies depending on the initial bulk composition (Figs. 1 and  2). The X(CO₂) generally increases at increasing temperatures (i.e., from event I to event II), except for Cal10 where the maximum X(CO₂) is reached at event I (Fig. 1b). The X(CO₂) of the fluids released at event II is generally higher than 0.5, with maximum values of X(CO₂) = 0.64–0.74 (Cal30; Fig. 1d), whereas it is more variable for fluids released at event I, ranging from X(CO₂) = 0.62–0.65 (Cal10; Fig. 1b) to X(CO₂) = 0.30–0.33 (Cal70; Fig. 2d). The amount of CO₂ (i.e., wt% loss of CO₂ relative to the initial CO₂ content in the solids) produced at each event increases up-temperature, except for Cal10, for which event I releases slightly more CO₂ than event II (Fig. 1b). The total CO₂ productivity increases from Cal10 (4.7 wt%) to Cal30 (7.0 wt%) and then decreases for Cal50 (3.7 wt%) and Cal70 (1.9 wt%). The same trend is observed when considering the CO₂ productivity related to event II only, maximum for Cal30 (3.6 wt%; Fig. 1d) and minimum for Cal70 (0.5 wt%: Fig. 2d).

Our results show that CO₂ is an essential component of C–O–H fluids released at relatively high temperatures (630–670 °C) and medium pressures (8.4–8.9 kbar or 25–30 km depth) during the formation of large hot orogens. The CO₂ productivity is maximum for the carbonate-bearing sediments originally containing low to moderate modal amounts of calcite (i.e., Cal10 and Cal30; Fig. 1) (see also refs. ^24,25,31), whereas it is considerably lower in impure limestones (Cal70)³⁷. Also, the model predictions are in excellent agreement with those discussed in recent papers focused on thermodynamic modeling of natural rock samples^24,25,31,38 and with fluid inclusion studies providing direct measurements of the fluid composition in mixed pelitic–carbonatic metamorphic sequences from collisional orogens^39,40,41. As a step forward, determining the physical and chemical properties of the fluids generated at these P–T conditions would be essential for understanding their behavior and, ultimately, the connection between the deep metamorphic production of CO₂ in the crust and its release at the surface.

Fluid immiscibility in the H₂O–CO₂–salt systems

Traditionally, metamorphic fluids have been considered as binary CO₂–H₂O mixtures in most thermodynamic modeling studies^42,43,44,45. However, fluids derived from metasedimentary rocks originally deposited at continental passive margins such as those involved in large hot orogens are generally characterized by variable, relatively high salinity (i.e., from 5 to 50–60 wt% NaCl eq.) even at high metamorphic grades, reflecting the presence of highly saline pore water and/or evaporites in the initial sedimentary sequence^{37,46,47,48,49}. Thick layers of evaporites are well known to occur along present-day passive margins, where they control the formation and storage of oil and gas^50,51, but they are also common in ancient sedimentary sequences since the Proterozoic ages^52,53. Direct evidence for saline fluids in metamorphosed passive margin sequences is represented by rare halide minerals, fluid inclusions, and/or high Cl contents of key high-grade metamorphic minerals^54,55,56,57. Evidence for salinity dilution at high temperatures due to progressive dehydration is feeble, suggesting that even advanced metamorphic dehydration cannot flush away the salinity inherited from the protoliths^46,47,52.

In this framework, salts components are particularly relevant because they strongly influence the topology of fluid phase equilibria, expanding the two-phase immiscibility fields^{37,46,47,58,59,60,61}. Fluid(s) produced by decarbonation reactions of calcite-bearing sedimentary protoliths can be reasonably described by the H₂O–CO₂–NaCl system^{40,57,62,63,64} and the H₂O–CO₂–CaCl₂ system^41,65. These ternary systems have been studied experimentally from 1 to 9 kbar and from 400 to 1000 °C^{64,66,67,68,69,70}, i.e., at P–T conditions compatible with those attained in large hot orogens. In order to describe phase relations of fluids produced during the main events of CO₂ production (i.e., event I: 575–605 °C, 7.6–8.0 kbar; event II: 630–670 °C, 8.3–9.0 kbar), we have interpolated the H₂O–CO₂–salts ternary diagrams obtained experimentally by ref. ⁶⁹ at 800 °C, 9 kbar and 500 °C, 5 kbar. The phase diagrams resulting from these interpolations (Figs. 3 and 4) are valid at ca. 575 °C, 6 kbar (i.e., for event I) and at ca. 650 °C, 7 kbar (i.e., for event II).

Fig. 3: Phase equilibria in the H₂O–CO₂–NaCl system.

Phase equilibria in the H₂O–CO₂–NaCl system at ca. 575 °C, 5 kbar and at ca. 650 °C, 7 kbar (interpolated from experimental data by ref. ⁶⁹), showing phase relations for fluids generated by the modeled sediments (a, b Cal10; c, d Cal30; e, f Cal50) at the event I (a, c, e) and event II (b, d, f), in the hypothesis of salinity of 10 wt% NaCl (corresponding to NaCl ≈ 6–7 mol%). Dashed lines outline the V–L tie-lines within the two-phase fields. Depending on the bulk composition, fluids released during events I and II plot either within the solvus (b, c, e, f) or within the V + NaCl two-phase field (a, d). In the first case, two immiscible fluids coexist: a CO₂-rich vapor and a hydrosaline brine (boxes on the solvus), whereas in the second case, a CO₂-rich vapor coexists with halite. F one-phase fluid, V vapor, L liquid. Phase relations for Cal70 are reported in Supplementary Fig. S1a, b.

Fig. 4: Phase equilibria in the H₂O–CO₂–CaCl₂ system.

Phase equilibria in the H₂O–CO₂–CaCl₂ system at ca. 575 °C, 5 kbar and at ca. 650 °C, 7 kbar (interpolated from experimental data by ref. ⁶⁹), showing phase relations for fluids generated by the modeled sediments (a, b Cal10; c, d Cal30; e, f Cal50) at the event I (a, c, e) and event II (b, d, f), in the hypothesis of salinity of 10 wt% CaCl₂ (corresponding to CaCl₂ ≈ 3–4 mol%). Dashed lines outline the V–L tie-lines within the two-phase fields. Fluids released during events I and II systematically plot within the solvus. Two immiscible fluids thus coexist at every P–T conditions: a CO₂-rich vapor and a hydrosaline brine (boxes on the solvus). F one-phase fluid, V vapor, L liquid. Phase relations for Cal70 are reported in Supplementary Fig. S1c, d.

In the H₂O–CO₂–NaCl system, the evolution of the fluids produced during the most CO₂-productive event II proceeds as a function of the initial bulk-rock composition and the initial salinity. Fluids generated during event II have an X(CO₂) systematically higher than 0.4 (Cal10: X(CO₂) = 0.57–0.64; Cal30: X(CO₂) = 0.64–0.74; Cal50: X(CO₂) = 0.61–0.68; Cal70: X(CO₂) = 0.45–0.51; Figs. 1b, d and  2b, d). In the hypothesis of a bulk salinity of 10 wt% NaCl, these fluids plot in the miscibility gap (Fig. 3b, d, f and Supplementary Fig. S1b), either in the vapor + liquid (V + L) two-phase field (Cal10, Cal50, and Cal70), or in the vapor + salt (V + NaCl) two-phase field (Cal30). In the first case, once liberated, fluids would inevitably split into a CO₂-rich phase of a lesser density (i.e., vapor⁶²) (Cal10: CO₂ = 60–66 mol%; Cal50: CO₂ = 62–66 mol%; Cal70: CO₂ = 49–53 mol%; Table 2) and a highly saline aqueous phase (i.e., liquid⁶²) (Cal10: NaCl = 38–42 mol%; Cal50: NaCl = 40–42 mol%; Cal70: NaCl = 23–30 mol %; Table 2), with the CO₂-rich vapor being the dominant fluid type (i.e. >85 mol%; Fig. 3b, f and Supplementary Fig. S1b). For Cal10 and Cal50 bulk compositions, the segregation of low amounts of halite from the two immiscible fluids (i.e., three-phase field V + L + NaCl; Fig. 3b, f) is predicted for the highest X(CO₂) values of the produced fluid. In the second case, a single vapor-like CO₂-rich fluid (Cal30: CO₂ = 66–73 mol%; Table 2) would coexist with solid halite. The immiscible nature of the fluids produced at the P–T conditions of event II is further amplified for higher initial fluid salinity, resulting in a more enriched CO₂ vapor and a hydrosaline brine with higher salinity; the fraction of the brine increases with initial salinity.

Table 2 Compositional ranges (mol%) of immiscible fluids generated by each modeled bulk composition (Cal10, Cal30, Cal50, Cal70) during event I and event II in the H₂O–CO₂–NaCl system.

A similar evolution is predicted for the fluid produced at event I in the H₂O-CO₂-NaCl system (Fig. 3a, c, e and Supplementary Fig. S1a). At T = 595–605 °C and P = 7–8.0 kbar, Cal10 bulk composition generates a fluid whose CO₂-rich composition (X(CO₂) = 0.62–0.65; Fig. 1b) plots within the V + NaCl two-phase field or in the nearby V + L + NaCl three-phase field. At these P–T conditions, a CO₂-rich vapor (CO₂ = 64–66 mol%; Table 2) coexists with halite, ± a very subordinate fraction of a hydrosaline brine (Fig. 3a). On the contrary, fluids produced by Cal30, Cal50, and Cal70 bulk compositions have lower X(CO₂) values (Cal30: X(CO₂) = 0.49–0.51; Cal50: X(CO₂) = 0.37–0.41; Cal70: X(CO₂) = 0.30–0.33; Figs. 1d and  2b, d) and thus plot within the V + L two-phase field (Fig. 3c, e and Supplementary Fig. S1a), occasionally close to the boundary of the solvus (Cal50 and Cal70). Two immiscible fluids thus coexist also at these P–T conditions. The composition of conjugate fluids change as a function of the bulk-rock composition: specifically, the amount of CO₂ in the CO₂-rich vapor progressively decreases from Cal30 (CO₂ = 53–57 mol%) to Cal50 (CO₂ = 39–42 mol%) and Cal70 (CO₂ = 32–38 mol%) and the salinity of the brine decreases from Cal30 (NaCl = 24–27 mol%) to Cal50 (NaCl = 13–15 mol%) and Cal70 (NaCl = 11–12 mol%) (Table 2).

The topology of the ternary diagrams in the H₂O–CO₂–CaCl₂ system shows an even more extensive miscibility gap⁶⁹ at any P–T conditions (Fig. 4, Supplementary Fig. S1, and Table 3). Thus, fluids produced during events I and II constantly plot within the solvus, independent of the initial bulk-rock composition and even for extremely low C–O–H fluid salinities. Immiscible CO₂-rich vapor and a hydrosaline CaCl₂-rich brine are predicted for both events.

Table 3 Compositional ranges (mol%) of immiscible fluids generated by each modeled bulk composition (Cal10, Cal30, Cal50, Cal70) during event I and event II in the H₂O–CO₂–CaCl₂ system.

Therefore, carbonate-bearing sediments undergoing prograde metamorphic decarbonation reactions at relatively high temperatures and medium pressures release a conjugate fluid pair “born this way” (immiscible at the source; i.e., at a crustal depth of ∼25–30 km), even for relatively low-salinity C–O–H initial fluid compositions (i.e., ≥5 wt% NaCl eq). In most cases, the CO₂-rich vapor is the dominant phase, coexisting with subordinate hydrosaline brines. Also, in the H₂O–CO₂–NaCl system, the coexistence of a vapor-like phase strongly enriched in CO₂ coexisting with halite is predicted. The immiscible CO₂-rich vapors produced during the most productive event II generally consist of more than 60 mol% CO₂ (up to 73 and 76 mol% in the NaCl and CaCl₂ systems, respectively) and are moderately saline, while hydrosaline brines generally contain > 60 wt% dissolved salt (i.e., close to saline melts; up to 69 wt% NaCl and 85 wt% CaCl₂).

Discussion

Implications for the CO₂ transport from the deep metamorphic source to the surface

Metamorphic C–O–H–salt fluids involved in dehydration and decarbonation reactions are generally modeled as a single phase. Fluid immiscibility has seldom been considered, although predicted by phase equilibria and documented by fluid inclusions, over different metamorphic environments and P–T–X conditions^{58,71,72,73,74}. The present results illustrate how decarbonation occurs within the solvus, generating immiscible CO₂-rich vapors and hydrosaline brines in the deep crust of large hot orogens, at peak metamorphic conditions. Immiscibility favors fluid segregation allowing transport of carbon in the crust. Because of the density and viscosity contrast between coexisting fluids, significantly less dense CO₂-rich fluids (∼1.06–1.10 g cm^−3 75,76) effectively separate from denser hydrosaline brines (∼1.80 g cm⁻³), where most solutes concentrate, and migrate upwards. The contrasting wetting behavior and reactivity of CO₂ and brines further influence fluid migration from the deep crust. In silicate-dominated rocks, CO₂-rich fluids have considerably higher dihedral angles (θ >65°, up to 90°) compared to aqueous fluids, in which lower values are favored by the presence of alkali or alkaline-earth halides (θ < 40 °C)^77,78. In carbonate-dominated rocks, the behavior of CO₂-rich fluids is less obvious; low solid–fluid dihedral angles (θ < 60°), and therefore wetting properties, are observed for H₂O–CO₂ fluids with intermediate X(CO₂) compositions (0.2 < X(CO₂) < 0.6), whereas H₂O-rich fluids (X(CO₂) < 0.2) and CO₂-rich fluids with X(CO₂) > 0.6 have high dihedral angles (θ > 65°, up to 90°) and therefore they are non-wetting fluids⁷⁹. Therefore, CO₂-rich fluids generated during events I and II mostly have a non-wetting behavior (except in carbonate-dominated lithologies, i.e., Cal70, whose CO₂ productivity is so low that they will not be further considered in the discussion). Thus, it is conceivable that the less dense CO₂-rich fluids are highly buoyant and unreactive, whereas the volumetrically minor and denser hydrosaline brines are highly reactive and could migrate by porous flow through thin interconnected films along grain boundaries⁷⁷ (Fig. 5). Our results also suggest that a solid salt instead of a brine could be segregated from a single CO₂-rich vapor-like phase for specific rock and fluid compositions.

Fig. 5: Sketch illustrating fluid production and evolution at the proceeding of a decarbonation reaction (Bt + Cal + Qz = Cpx + Kfs + fluid), which occurs at a depth of 25–30 km (i.e., at P–T conditions compatible with event II).

Two immiscible fluids are generated in the source rocks: the CO₂-rich fluid is shown in yellow-red, with colors changing from yellow to red according to the progressive increase of the fluid pressure within the pores, whereas the hydrosaline brine is reported in blue. The short to long black arrows radiating from the CO₂-filled pores indicate the progressive increase of the fluid pressure in the pores. The dihedral angles of the two fluids (CO₂-rich fluid: θ > 60°; hydrosaline brine: θ < 60°) imply that the CO₂-rich fluid accumulates within isolated fluid pockets, whereas the hydrosaline brine forms thin interconnected films along grain edges. As far as the decarbonation reaction proceeds, fluid pressure within the CO₂-rich pockets increases until fluid overpressure is reached, carbo-fracturing the host rocks. The CO₂-rich fluid thus rapidly escapes toward the surface along the newly created fracture network, whereas the denser, hydrosaline brine remains trapped in the source rock.

Thus, in a two-fluid flow regime, CO₂ and brines migrate separately. CO₂-rich fluids generated at depths of 25–30 km (8–10 kbar) would accumulate, eventually forming large reservoirs. With decarbonation reactions proceeding, confined CO₂-rich fluids would cause fluid overpressure, inducing carbo-fracturing of the host rocks and migrating upward. Such processes could trigger crustal permeability, possibly resulting in earthquake nucleation and seismicity^{28,80,81,82,83,84,85}). An alternative scenario of a passive ascent of CO₂-rich fluids, favored by brittle fracturing of the crust during earthquakes or by ductile deformation, could be equally possible to enhance fluid mobility through channelization along preferential pathways. In both cases, fracturing and faulting would allow fast CO₂-rich fluids migration toward the surface^{28,81,82,85,86,87,88}. As CO₂-rich fluids are removed from the reacting sites, the remaining hypersaline brines could represent important metasomatic agents in the lower crust^48,58,89. Moreover, their low water activity could promote granulitization of the deep crust, delaying its partial melting.

Fluid-rock interaction during ascent is unlikely because the relatively high transport velocities and the nonpolar nature of CO₂-rich fluids do not allow chemical equilibration with most metamorphic rock compositions. To our knowledge, pervasive carbonation by fluid-rock interactions at P–T conditions compatible with fluid ascent in large hot orogenic settings is limited to soapstones, listvenites and sagvandites (i.e., talc + magnesite, quartz + magnesite and enstatite + magnesite rocks derived from ultramafic lithologies by reaction with a CO₂-rich fluid^{90,91,92,93,94,95,96,97}). However, ultramafic lithologies are rare in large hot orogens and are regarded as exceptions rather than rules. Also, precipitation of epigenetic graphite should be negligible due to the low water activity in CO₂-rich fluids coupled with the high temperatures at which decarbonation reactions occur⁹⁸.

Fluid immiscibility supports geochemical and geophysical observations from active large hot orogens

Segregation of CO₂-rich fluids by phase separation at birth (i.e., directly in the deep crustal source) reconciles geochemical and geophysical evidence from active collisional orogens, e.g., the Himalaya, considered as the archetype of large hot orogens and characterized by present-day intense CO₂ surface degassing.

First, immiscibility, with the generation of CO₂-dominated fluids (i.e., X(H₂O) = 26–38 mol%), could explain the gaseous CO₂ emissions measured at the surface. Diffuse degassing occurs over extensive areas in Himalaya, testified by the widespread occurrence of CO₂-rich hot springs and gaseous CO₂ ground discharges along the main tectonic discontinuities and distributed along the entire length of the orogenic belt^{9,27,28,85,99,100}. Geochemical and isotopic analyses revealed that CO₂ released at the surface has a crustal metamorphic signature, produced at >5 km depth^{9,27,28,85,100}. Girault & Perrier⁸⁴ and Girault et al.⁸⁸ suggested that CO₂ is not dissolved in aqueous fluids but, instead, rapidly outgas along “dry” faults toward the surface. Estimated transport velocities are in the order of 0.1 to 1 m sec⁻¹ to account for the observed radon and CO₂ fluxes.

Second, beneath the Himalayan metamorphic core, a major conductive zone is revealed at a depth of ∼20–30 km¹⁰¹, immediately below a zone of intense microseismicity^102,103,104. This highly conductive and seismically active zone is located along a mid-crustal ramp below the superficial emergence of the Main Central Thrust (Fig. 6), i.e., the main tectonic discontinuity along which most of the CO₂-rich hot springs are concentrated. Lemonnier et al.¹⁰¹ first interpreted the conductivity anomaly as related to highly connected aqueous fluids released by metamorphic dehydration reactions, percolating upward into the brittle portion of the crust, where microseismic activity is observed. In aqueous fluids, conductivity is low but readily increases with salinity¹⁰⁵. Thus, this deep conductive zone could result from the stagnation of hydrosaline brines⁴⁸ left behind by the CO₂-rich fluids once they rise toward the surface. Although volumetrically minor, hydrosaline brines have an extremely wetting behavior¹⁰⁶. It is, therefore, likely that they are highly connected in the source rocks (Fig. 5). Like Lemonnier et al.¹⁰¹ original interpretation, the zone of intense microseismicity immediately above the conductive anomaly could be related to periodic carbo-fracturing events in the brittle crust induced by the accumulation of CO₂-rich fluids at the lower crust–upper crust boundary. Finally, given the striking similarities between actual CO₂ fluxes measured at the surface and those predicted by modeling of prograde metamorphism of carbonate-bearing lithologies in Himalaya at depths of 25–30 km (i.e., at 650–750 °C, 8–10 kbar^24,25,31,38), it seems likely that CO₂-producing processes and transport mechanisms similar to those occurring in Himalaya at present were also active in the past.

Fig. 6: Schematic geologic section across central Nepal Himalaya at the longitude of Kathmandu.

The thick red line shows the Main Himalayan Thrust, MHT, which reaches the surface at the front of the Siwalik Hills, coinciding with the Main Frontal Thrust (MFT). The Main Boundary Thrust (MBT) separates the metasediments of the Lesser Himalayan Sequence (LHS) from the molasse deposits of the sub-Himalaya (the Siwaliks Hills). The Main Central Thrust (MCT) places the higher-grade metamorphic rocks of the Greater Himalayan Sequence (GHS) over the LHS metasediments. The GHS is divided into two tectono-metamorphic units (Lower-GHS and Upper-GHS) by the High Himalayan Discontinuity (HHD); geologic cross-section modified from refs. ^104,122. The red to yellow area highlights the zone of high conductivity located at depths >15 km beneath the superficial emergence of the MCT, as obtained from the magnetotelluric experiment of ref. ¹⁰¹. The conductivity anomaly is associated with intense microseismic activity, evidenced by the schematic distribution of earthquake hypocentres (small circles). The white diamond indicates the CO₂-rich hot springs and gaseous CO₂ discharges from the ground, located along the MCT.

In conclusion, we demonstrate that the most productive CO₂-source rocks in large hot orogens are calcareous pelites and carbonate-poor marls (i.e., Cal10 and Cal30), which generate immiscible CO₂-rich fluids and hydrosaline brines when metamorphosed at T > 590 °C. We suggest that fluid immiscibility provides a physical mechanism to transport carbon liberated during prograde metamorphism. Segregation due to density contrast and different chemical properties between CO₂ and brines allows the rapid migration of CO₂ from the deep source to the surface. Such a model could explain the CO₂ fluxes currently measured at the surface in active collisional orogenic settings such as the Himalaya, which are impossible to explain in the absence of fluid phase separation. Finally, our study contributes to solving the debate on the role of orogenic settings for the global Earth’s CO₂ tectonic outgassing. An array of independent observations consistently indicates that active collisional orogens could represent an important (and so far under-considered) source of CO₂ on a global scale. Since the Himalayan orogen is a modern analog of ancient large hot orogens, such as the Mesoproterozoic Grenville and Sveconorwegian orogens and the Late Palaeozoic Variscan orogen, we advocate that the same process currently active in the Himalaya should have been repeatedly active at different times, during the main orogenic events, possibly causing periodic perturbations of the atmospheric CO₂ levels.

Methods

Model bulk-rock compositions

Model bulk-rock compositions used to calculate the P/T–X(CO₂) buffering paths were obtained by adding 10, 30, 50, and 70 vol% of calcite to the average pelite composition of ref. ³⁴ (“shale and slate” group; his Table 2) (Table 1). The modal proportions of the protolith’s minerals have been obtained by applying the least square method (freeware application available on demand¹⁰⁷) and using end-member compositions and molar volumes for kaolinite, illite, clinochlore, daphnite, albite, anorthite, quartz, and K-feldspar. The result is considered satisfactory if the residuals (i.e., molar bulk composition of the protolith’s minerals—molar bulk-rock composition) is close to zero. The following assumptions were additionally made: (a) CaO from silicate fraction is equivalent to Na₂O^108,109 and is incorporated in anorthite; the remaining CaO is incorporated in calcite; (b) chlorite is the only Fe-Mg mineral in the protolith and (c) albite incorporates all the Na₂O.

Thermodynamic modeling

Thermodynamic modeling was performed using PerpleX 6.9.0^110,111 (version June 2020), the internally consistent thermodynamic dataset (version ds55), and the equation of state for H₂O-CO₂ binary fluid of ref. ¹¹². P/T–X(CO₂) buffering paths were calculated along a P/T geothermal gradient representative of large hot orogens (dT/dP = 750 °C GPa⁻¹ (see ref. ³⁵)). The following solution models were considered: carbonate (i.e., Ca–Mg–Mn–Fe carbonate with calcite structure¹¹³), dolomite¹¹², scapolite¹¹⁴, chlorite¹¹⁵, white mica^116,117, biotite¹¹⁸, plagioclase¹¹⁹, K-feldspar¹²⁰, Ca-amphibole (ideal tremolite solution model), clinopyroxene¹¹², garnet¹¹², epidote¹¹², chloritoid¹¹², and staurolite¹¹². A generic hybrid molecular fluid-equation-of-state solution model was used for fluid (COH-Fluid model available in PerpleX), including H₂O, CO₂, CO, CH₄, and H₂ molecular species. This model allows considering redox processes, such as the possible reduction of carbon to graphite due to the exchange of oxygen between the C–O–H fluid and the Fe³⁺-bearing solid phases (white mica, biotite, clinopyroxene, garnet, epidote).

Calculation of P/T–X(CO₂) buffering paths requires fixing a P–T–X(CO₂) starting condition; this was fixed at 350 °C, 4 kbar, X(CO₂) = 0.005, such as the predicted mineral modes are as close as possible to the calculated protolith’s mineral modes (with muscovite replacing illite). The Werami routine of Perple_X was used to calculate the amount of H₂O, C, and O₂ stored in solid phases at the starting P–T–X(CO₂) fluid-saturated conditions (Cal10_0, Cal30_0, Cal50_0, and Cal70_0 compositions in Supplementary Tables S1–4). To start calculating the buffering paths, 1 mol% of the equilibrium fluid with X(CO₂) = 0.005 was added to each model bulk-rock composition (Cal10_1, Cal30_1, Cal50_1, Cal70_1 compositions in Supplementary Tables S1–4). The addition of a small amount of fluid is required because, at fluid saturation conditions, the Werami routine returns the amounts of H₂O, C, and O₂ stored in minerals, but not the amount of free fluid in equilibrium with solid phases, which in turn depends on the porosity of the system (see ref. ¹²¹ for further details). P/T–X(CO₂) buffering paths were calculated following the constant porosity model of ref. ¹²¹. According to this model, the fluid is allowed to accumulate within the rock until it reaches a specified molar proportion threshold; each time this threshold is exceeded, fluid loss occurs through a stepwise process. Starting from the initial fluid mole proportion of 0.01 (1 mol%, relative to the solids + fluid assemblage), we have fixed the threshold at 0.02; this implies that when the fluid proportion reaches this value, its amount is reduced to 0.01. Bulk compositions after each fractionation step are reported in Supplementary Tables S1–4. Supplementary Tables S1–4 also include the modeled amounts of fluid (vol%, relative to the solids + fluid assemblage) immediately before and immediately after each fractionation step, and show that fractionation of fluid allows maintaining a porosity <2 vol%. The two extreme cases of an internally buffered, completely closed system (i.e., no fluid loss is allowed) and of an internally buffered completely open system (i.e., all the fluid that is produced is immediately lost) were further considered, which provide the minimum and maximum X(CO₂) values, respectively, for the same starting conditions.

Figures 1 and 2 show the variations in mineral proportions (vol%, fluid excluded), the loss (wt%) of CO₂ and H₂O relative to the initial CO₂ and H₂O content in the solids, and the composition (X(CO₂) = molar fraction of CO₂) of the fluid released along the calculated buffering paths. The complete speciation of the fluid along the modeled buffering paths is reported in Supplementary Data Files S1–4.

Estimate of the amounts of CO₂ and H₂O released along the modeled prograde P/ T–X (CO₂) buffering paths

The Werami routine of Perple_X was used to infer: (i) the abundances (wt%) of both C-bearing minerals (carbonates and scapolite) and hydrous minerals (chlorite, muscovite, biotite, epidote) along the calculated P/T–X(CO₂) buffering paths, (ii) the amounts (wt%) of C, O₂, and H₂O hosted in these minerals at each P–T conditions. Although scapolite can incorporate significant amounts of Cl in compositions close to the marialite end-member, it is to be noted that in the available scapolite solution model¹¹⁴ marialite is treated as a hypothetical CO₃ end-member. This would potentially result in an overestimation of the CO₂ content in scapolite. However, this should not significantly influence the final estimate of CO₂ production because available data on natural scapolite from large hot orogenic contexts suggest that most scapolite contain negligible (if not null) amounts of Cl^24,25. The amounts (wt%) of CO₂ and H₂O released during prograde metamorphism from each model bulk composition were then calculated by adding carbonates’ contribution with that of scapolite and chlorite’s contribution with that of muscovite, biotite, and epidote, respectively. Supplementary Data Files S1–4 summarize the variations in mineral proportions (vol% and wt%) and the amounts (wt%) of C, O₂, and H₂O released along the calculated buffering paths and used for Figs. 1 and 2.