Introduction

Oceanic sediments constitute a substantial carbon source within subduction zones, encompassing both inorganic and organic carbon components that are transported into the subduction factory. Understanding the fate of these carbon inputs in subduction zones is crucial for gaining insights into the global carbon cycling throughout Earth’s history. While the significance of carbonate in the subduction-zone carbon cycle has been extensively investigated in recent decades1,2,3, the importance of organic carbon within subduction-zone lithologies and deep fluids has been relatively overlooked4,5. The deep subduction of reduced organic carbon has impacts on various geological events, including the Great Oxygenation Event6, the Lomagundi event7, arc carbon isotopic signature8, and the formation of deep diamonds with light carbon isotopic features9.

Diamond provides the principal data source for investigating mantle carbon. Low δ13C values (ranges from −40 to +3‰, with peak values of −6‰) in eclogitic diamonds, along with higher-than-mantle δ18O values in their silicate inclusions, suggest the potential contributions of deeply subducted organic carbon into the mantle10,11. Similarly, the relatively light carbon isotope composition (δ13C ranges from −29 to −18‰) of ophiolite-hosted diamonds indicates the possible contribution of subducted biogenic carbon9. Kimberlite, a common host lithology of diamonds, also offers insights into mantle carbon. The lower δ13C ratios of carbonate-rich hypabyssal kimberlites from the Mesozoic onwards (<250 Ma) suggest that subduction of organic carbon into the deep mantle, following the Cambrian Explosion when organic carbon deposition in marine sediments increased significantly12. While these findings suggest the potential subduction of surface-derived organic carbon into the deep mantle, they lack petrological evidence to confirm the deep subduction of organic carbon.

Here, we present robust petrological evidence to support the subduction of surficial organic carbon to mantle depths, which may contribute to the formation of (ultra-) deep diamonds with light δ13C values and potentially offer an explanation for the Lomagundi event through the increased subduction of organic carbon.

Results and discussion

Petrography and mineral chemistry of graphite-bearing pelitic schists

The studied Tianshan graphite-bearing pelitic schists (Supplementary Note 1 and Supplementary Fig. 1) are well-foliated, with phengite defining the foliation. These schists consist of white mica (phengite 25% and paragonite 10%), garnet (10%), quartz (25%), retrograde albite (25%), and graphite ( < 5%); rutile, epidote/zoisite, apatite, tourmaline, chlorite, carbonate and titanite/ilmenite occur in minor amounts. Modal abundances were estimated through visual observation.

Garnet occurs as subhedral to euhedral crystal, measuring around 0.1−0.5 mm in size (Fig. 1a–d), and some garnets are fractured. Obvious compositional zonation can be observed in these garnets (Fig. 1e, f and Supplementary Data 1). From core to rim, XGrs (mineral abbreviations refer to Supplementary Note 2) increases from 0.08 to 0.24, XPrp remains constant at about 0.14, and XSps and XAlm decrease from 0.03 to 0.01 and 0.77 to 0.62, respectively. Graphite, rutile, and occasionally coesite are included in these garnet crystals (Figs. 1b–d and 2a–d). There are also some atoll-shaped garnets in the samples, whose irregular shell structure encircles cores partially or entirely replaced by phengite, epidote, and quartz. The “shell” of atoll-shaped garnets shows a composition similar to the rim of complete garnet crystals. No fluid inclusions were observed in the garnet crystals of the studied pelitic schists.

Fig. 1: Representative photomicrographs and mineral compositions of the studied graphite-bearing pelitic schists from Tianshan, China.
figure 1

a and b Representative photomicrographs (plane-polarized light) of graphite (Gph)-bearing pelitic schists, which are mainly composed of phengite (Phn), quartz (Qz), and garnet (Grt), with minute graphite in phengite, quartz, and garnet. c and d Representative photomicrographs (plane-polarized light) of graphite inclusion, which trails along crystal faces highlighting growth zones of garnet, and rutile (Rt) is included in the matrix quartz. e and f Backscattered electron images and chemical composition profiles of garnets in c and d.

Full size image
Fig. 2: Coesite and graphite inclusions in garnet.
figure 2

a and b Two coesite (Coe) crystals were found in the garnet (Grt), which contains quartz (Qz) with radial fractures. Graphite (Gph) and rutile (Rt) are contained in garnet. a and b are plane- and cross-polarized light, respectively. c and d Coesite inclusions in garnet. Plane- and cross-polarized light. e and f Raman spectra of Coe-1 and Coe-2 in garnet. g and h Raman spectra of graphite in garnet and quartz.

Full size image

Phengite grains in the matrix measure around 50−150 μm in width (Fig. 1a, c) and have a Si content of 3.34−3.42 atoms per formula unit (apfu; Supplementary Data 1). The phengite filling in the core of the atoll-shaped garnet shows a Si content of 3.39−3.45 apfu. Paragonite has a grain size of 100−300 μm in width, coexisting with phengite or quartz. Paragonite shows steady Na content (0.67−0.75 apfu) and K content (0.04−0.08 apfu).

Graphite grains are dispersed throughout the investigated pelitic schists (Fig. 1a, b) and are included in garnet, phengite, and quartz (Fig. 1c, d), as well as paragonite, albite, titanite, and tourmaline in some cases. Notably, minute graphite inclusion commonly trails extend parallel to the crystal faces of garnet at discrete growth stages (Fig. 1b–d), and rutile inclusions are also included in these graphite trails.

UHP metamorphism of graphite-bearing pelitic schists

The identification of coesite inclusions within garnet (Fig. 2a–f) in the studied pelitic schists provides compelling evidence for UHP metamorphism of the metasedimentary rocks in the Tianshan orogenic belt. This finding is in agreement with previously published literature13,14. Two coesite crystals were identified in the garnet, which contains quartz with radial fractures (Fig. 2a, b). The monomineralic coesite (Coe-1) displays a well-characterized strong coesite band at 521 cm−1, with no quartz bands observed (Fig. 2a–e). The good preservation of this small coesite grain can be explained by its size, which was too small to cause remarkable fractures in the rigid host garnet during decompression. This is further evidenced by the weak radial fractures surrounding Coe-1 (Fig. 2c and insert figure in Fig. 2e). Another coesite crystal (Coe-2) shows a strong band at 521 cm−1. Additionally, relatively weak quartz bands at 467 and 205 cm–1 are visible (Fig. 2a, b, d, f), potentially indicating the partial transformation of coesite to quartz during the decompression process.

The P–T pseudosection of the graphite-bearing pelitic schist H1813B reveals a roughly isothermal decompression path (Fig. 3a). Garnet core composition plots in the field of Grt + Omp + Phn + Cld + Amph + Lws + Qz/Coe + Rt, highlighting a peak pressure stage at 28 kbar and 540 °C (M1 stage). These values may imply UHP metamorphism and concur with the measured and projected phengite compositions (Si = 3.34–3.45 apfu; Fig. 3b), matching with the well-preserved small coesite inclusions in garnet.

Fig. 3: P–T phase diagram together with Si in phengite isopleths for bulk H1813B.
figure 3

a P–T pseudosection for the investigated coesite- and graphite-bearing UHP pelitic schist (sample H1813B) from the Chinese southwestern Tianshan HP−UHP metamorphic belt, system NCKFMASHTi+CO2 (+ quartz/coesite + rutile + F) based on the XRF bulk composition from Supplementary Data 3. XPrp, XGrs, and XAlm isopleths in garnet were plotted in green, pink, and purple dashed lines, respectively. The gray curved line is the constrained P–T path for the exhumation of H1813B. M1, M2 and M3 stages are constrained by the composition of garnet core, mantle and rim, respectively. b Phengite Si-content isopleths (apfu).

Full size image

Carbon source of graphite

Previous kinetic studies15,16 have indicated that the carbon diffusion in graphite is very sluggish, implying that well-crystalized graphite is isotopically inert and does not undergo isotopic alteration during subsequent hydrothermal alteration, even under high P–T conditions. Metamorphic graphite derived from biogenic carbon commonly shows light isotopic signatures with a mean δ13C of around −25‰17. This signature is quite homogeneous for a set of grains derived from the same organic precursor and formed under the same metamorphic grade18. In contrast, fluid-deposited graphite shows a wide range of carbon isotopic signatures19, depending on both the origin of carbon (e.g., <–20‰ for unmetamorphosed marine organic matter, –5 ± 2‰ for mantle carbon18 and 0 ± 2‰ for marine carbonates20) and the genetic mechanisms involved in its precipitation, such as the evolution of the carbonic fluid with time21.

The total organic carbon (TOC) concentration in the studied coesite- and graphite-bearing pelitic schists ranges from 0.31 to 1.22 wt.% (Supplementary Data 2). These pelitic schists have homogeneous δ13CTOC values of −24.7 to −22.5‰, which fall in the range of typical marine organic matter (Fig. 4; Supplementary Data 2). The δ13CTOC of picked graphite-bearing garnet crystals and picked graphite-bearing whitemica (phengite and paragonite) are in agreement with that of the bulk rock (Fig. 4 and Supplementary Data 2). This consistency suggests that graphite serves as the primary organic carbon phase in these samples. These isotopic data are close to the δ13CTOC values of organic matter found in iron deposits from the western Tianshan (−25.2‰; Fig. 4)22, suggesting a surface-derived organic carbon source for the studied graphite (Fig. 4 and Supplementary Data 2).

Fig. 4: δ13C values (versus Pee Dee Belemnite standard) of bulk rock organic carbon and graphite contained in garnet and whitemica crystals from the coesite- and graphite-bearing pelitic schists in the Chinese southwestern Tianshan UHP metamorphic belt.
figure 4

For comparison, δ13CTOC (TOC, total organic carbon) values of abiotic graphite formed at HP conditions in subduction zones24, 25, mantle-derived carbon18, typical marine organic carbon20, organic carbon from iron deposits from western Tianshan22, and ophiolitic diamond33 are also plotted.

Full size image

Formation of graphite

In subduction zones, graphite can be formed by the transformation of organic matter during metamorphism (metamorphic graphite) or by precipitation from carbon-bearing fluids (fluid-deposited graphite)19,23.

The studied graphite occurs as minute solid inclusion tracks along crystal faces of garnet at each growth stage (Fig. 1b–d). There are two hypotheses for the formation of this graphite: (i) graphite precipitated from a carbonic fluid during garnet crystallization, and (ii) pre-existing graphite was included by the garnet during its crystallization.

Regarding the first hypothesis for the formation of graphite in the studied sample, it is plausible that the graphite precipitated from either CO2- or CH4-bearing metamorphic fluids. The former can be ruled out because the graphite precipitated from sedimentary carbonate-sourced COH fluids generally shows high δ13C values of around 0‰24 in Alpine Corsica or around −13‰25 in the Chinese Tianshan (Fig. 4). Conversely, biotic (thermogenic) CH4-bearing fluids could potentially be the source of graphite in these samples, given the low δ13C value (<−20‰) associated with biotic CH426. Although carbonaceous matter in subducted sediments can release biotic CH4, forming carbonic fluids with low δ13C during the thermal maturation stage at temperatures between 200 and 300 °C, these slab-released CH4 fluids would tend to go upwards rather than go down with the slab to the deep subarc depths (as indicated by coesite inclusions representing UHP conditions). This makes the CH4 fluid deposition hypothesis less plausible. Furthermore, the graphite trails observed in our study do not correspond to chemical zones in garnet (Fig. 1c–f), further diminishing the likelihood of the fluid deposition hypothesis. Hence, we think the first hypothesis (fluid deposition) is less possible in our study.

According to the published literature, the degree of metamorphic graphite crystallization linearly correlates with the peak temperature experienced during metamorphism27. Raman spectroscopy is an effective tool for evaluating the crystallization degree of graphite and its relationship to the metamorphic degree28. In this study, graphite shows a sharp peak at the G band (around 1580 cm−1) with a small peak at the D1 band (around 1350 cm−1) in the first-order region, and D2 (around 1620 cm−1) is inseparable (Fig. 2g, h). In the second-order region, S1 (~2450 cm−1), S2 (~2700 cm−1), and S4 (~3200 cm−1) peaks were observed. The R2 ratio29 of graphite, defined as the D1/(G + D1 + D2) peak area ratio, varies from 0.16 to 0.37, with the majority falling between 0.2 and 0.3. The Raman spectrum of the studied graphite is comparable to that of metamorphic graphite formed at ~540 °C in the European Alps27. Since the Tmax in this study is around 550 °C (M1 stage in Fig. 3a), it suggests that graphitization might be a possible mechanism for graphite formation in this study. Additionally, the similar R2 ratio of graphite crystals in garnet at each growth stage indicates that these graphite trails underwent a consistent peak metamorphic temperature. It is suggested that the graphite was transformed from disordered organic carbon via solid-state transformation (graphitization) in sediments with increasing subduction temperature. The graphitized organic carbon grains were then contained by the crystallized garnet at the M1 stage. These graphite trails are likely the result of the interplay between surface energy and growth kinetics30, whereby a growing garnet porphyroblast may displace the graphite instead of incorporating it until the graphite trail becomes too large to displace, and only then is the trail included. In conclusion, the second hypothesis (graphitization) is considered to be the most feasible one for the formation of graphite in the studied pelitic schists.

Subduction of sedimentary organic carbon to more than 90 km depths

The transport and storage of organic carbon during subduction processes can substantially impact the global carbon cycle and climate regulation6,8. In this work, the characteristic of the light organic carbon isotopic composition (Fig. 4 and Supplementary Data 2) implies that the material source of pelitic schists-hosted graphite is surface-derived organic matter. The coexistence of coesite and graphite in garnet crystal (Fig. 2a–d) indicates the presence of graphite under the coesite stability field, which is at pressures exceeding 2.8 GPa when temperatures higher than 500 °C (see the Qz-to-Coe phase transition in Fig. 3a), but lower than the diamond stability field. Therefore, it can be proposed that the surface-derived organic matter was transported to subarc depths (>90 km) along with the oceanic slab.

Implication for diamond genesis

Diamond, a high-pressure polyphase of graphite, is estimated to be formed in the lithospheric mantle at depths exceeding 140 km or under pressures greater than 4.5 GPa31. An intriguing debate related to crustal recycling into the deep mantle regards the involvement of subduction-derived C in the genesis of diamond32. According to the literatures, some diamonds exhibit lighter carbon isotope characteristics (i.e., lower than mantle range)10,11, particularly ophiolitic diamonds, which show δ13C ranging from −29‰ to −18‰ with a peak value of −25‰9,33 (Fig. 4). Various theories have been posited to explain the light δ13C compositions of these diamonds, including the primordial carbon reservoir, intra-mantle isotopic fractionation, or the incorporation of subducted biogenic/organic carbon32. However, mantle Sr, Nd, and O isotopic geochemistry data rule out the preservation of primordial carbon isotope heterogeneity in Earth’s mantle. Moreover, the narrow range of δ13C values from individual diamonds does not support isotope fractionation from a normal mantle carbon source with δ13C of –5 ± 2‰32. Therefore, the light carbon recorded by ophiolitic diamonds possibly reflects the composition of subducted organic carbon. However, the lack of robust petrological evidence makes this hypothesis arguable.

The results of this investigation have important implications for our understanding of the organic carbon cycle in the deep Earth and may inform ongoing debates concerning the origin and formation of diamonds with lower δ13C values. Our study provides the petrological evidence for the subduction of surface-derived organic carbon to the mantle depths of approximately 90 km, with the formation of graphitized organic carbon (graphite). Because diamonds can be formed by direct conversion from graphite within the diamond stability field34, we propose that the continued subduction of this graphitized organic carbon may contribute to the formation of diamonds with light carbon isotope composition. This possible process can be inferred from the consistent C isotopic signatures between the ophiolitic diamond and organic carbon-sourced graphite (Fig. 4).

We conclude that, with the increasing subduction depth, surface-derived disordered organic carbon transformed to crystalized graphite at subarc depth (around 90 km) and subsequently to diamonds at lithospheric mantle depth (140 km), as illustrated in Fig. 5. Further studies are needed to fully elucidate the mechanisms and processes involved in the transition from surface organic carbon to graphite and then to diamonds.

Fig. 5: Cartoon illustrating the fate of organic matter in subduction zones.
figure 5

Disordered organic matter (OM) (δ13COM < −20‰) contained in marine sediments undergoes partial to complete graphitization with increasing temperature in subduction zones. The results presented in this study suggest that these disordered organic matter from marine sediments subducted to subarc depths (around 90 km, coesite-stable field), forming graphite with δ13CGph of −24‰. This subduction of surface organic carbon may contribute to the formation of ophiolitic diamonds with light δ13C values (peak value of −25‰).

Full size image

Implication for the Lomagundi Event

The Lomagundi event, occurring around 2.3–2.0 billion years ago (Ga), was marked by a notable positive C isotope excursion of marine carbonates, with an increase of ~10‰35. This event has been attributed to a variety of factors, including changes in the biogeochemical cycling of carbon36, enhanced organic carbon burial6,7,37,38,39,40, and increased volcanic outgassing7,37.

In recent years, the deep subduction of organic carbon has emerged as a potential driver of the Lomagundi event, drawing considerable attention. Experimental and thermodynamic studies, along with the Carbon-oxygen box model6,7 all support this mechanism. This proposed process emphasizes the deep cycling of subducted organic carbon, which is deposited on the seafloor and subsequently subducted into the Earth’s interior along with carbonates. Following subduction, organic carbon can be metamorphosed to graphite, which is more refractory than carbonate. At subarc conditions, carbonates are preferentially released from the slab over graphitized organic carbon, leading to an increase in arc volcano δ13C values above −5‰. Meanwhile, graphitized organic carbon remains in the slab and is transported deep into the mantle, where it may become entrained in upwelling mantle plumes feeding ocean island volcanoes. Reduced organic carbon in the upwelling plume is likely to undergo redox melting, resulting in a decrease in ocean island volcano δ13C values to below −5‰7. In comparison to the subduction process (especially the carbonate-releasing process), the delayed eruption of the mantle plume resulted in a decrease in atmospheric δ13C values. This decline subsequently contributed to a return of the δ13C values of deposited marine carbonates to normal levels (∼0‰) after undergoing a large positive spike in carbon isotope ratios.

Our study offers petrological evidence to support the subduction of organic carbon to mantle depths, implying a plausible contribution of this process to the Lomagundi event’s positive shift in carbon isotopic composition. These findings underscore the importance of considering the deep cycling of subducted organic carbon for understanding the mechanisms driving notable changes in biogeochemical cycles of carbon over geological timescales.

Methods

Mineral chemistry

Mineral major element compositions were analyzed with a JEOL 8230 electron probe microanalysis (EPMA) at the School of Earth and Space Sciences (SESS), Peking University, China. Synthetic silica (Si) and spessartine (Mn), natural sanidine (K), pyrope (Mg), andradite (Fe and Ca), albite (Na and Al), and rutile (Ti) served as standards for EPMA calibration. The acceleration voltage and beam current were 15 kV and 10 nA, respectively. The beam diameter was 2 μm for silicates, and the counting times were 10–15 s. The PRZ correction was performed at the final calibration stage.

Raman spectroscopy

Raman spectra of coesite and graphite were acquired using a HORIBA Jobin Yvon confocal LabRAM HR Evolution micro-Raman system at the SESS, Peking University, China. The green laser (wavelength of 532 nm) was focused on the polished thin section with a 100× objective. The laser spot size was focused to 2 μm, and the laser power was 100 mW at the source, 1% (1 mW) for graphite, and 100% (100 mW) for silicate inclusions. Before each session, the spectrometer was calibrated using synthetic silicon (520.7 cm−1). We selected graphite grains below the transparent adjacent/including mineral. Acquisition times varied from 10 to 20 s for coesite and from 20 to 40 s for graphite.

Organic carbon isotope and organic carbon concentration

For each analyzed sample, three types of powders (bulk-rock powder, garnet powder, and muscovite powder) were prepared for the carbon isotope ratio and concentration of graphite. The selected samples were washed with deionized water after removing the weathered surfaces. After crushing the samples into small fragments, clean pieces free of cracks and veins were handpicked. Then these pieces were powdered in an agate mortar to make the bulk-rock powder (<200 mesh). Graphite-bearing garnet crystals and graphite-bearing muscovite crystals were handpicked respectively from the bulk-rock powder and were powdered to make the garnet powder (<200 mesh) and the muscovite powder (<200 mesh). All Powders were treated with 1 mol/L HCl to remove carbonate, and the acid-insoluble fraction was then rinsed with distilled water and dried at 60 °C in an oven for 24 h.

Total organic carbon (TOC) concentrations and isotopic composition (δ13C) measurements were carried out by using the Thermo Scientific EA IsoLink Isotope Ratio Mass Spectrometer (IRMS) System at the SESS, Peking University, China. Dried, carbonate-free samples were wrapped in tin capsules and placed in the EA autosampler. Then the samples were dropped into the cleaned quartz tubes and sealed under vacuum for combustion and oxidation with CuO (1100 °C) for carbonaceous matter (CM) extraction. The generated gases (N2 and CO2) were separated on a chromatographic column, and the carbon isotopic composition of the separated CO2 was then measured with a Thermo Scientific Delta V Advantage continuous flow isotope ratio mass spectrometer. Empty muffled tin capsules were run for contamination before analysis, and no carbon was detected in these procedural blank tin capsules. To minimize external and internal uncertainties, calibration standards urea (IVA3380217, δ13C = –40.8‰), GBW04406 (CaCO3, δ13C = –10.85‰), USGS24 (graphite, δ13C = –16.05‰), USGS44 (CaCO3, δ13C = –42.21‰), IAEA-CO-8 (CaCO3, δ13C = –5.76‰), IAEA-603 (CaCO3, δ13C = +2.465‰), and NBS18 (CaCO3 with a δ13C value of –5.01‰) were systematically run during the measurements. δ13C values were normalized to the V-PDB (versus Pee Dee Belemnite) standard. Based on duplicate standards, the δ13C analytical precision was better than 0.2‰, and the carbon concentration analytical precision was better than 0.05%.

Phase equilibrium modeling

Phase equilibrium modeling was conducted in the system of Na2O–CaO–K2O–FeO–MgO–Al2O3–SiO2–H2O–TiO2–CO2 (NCKFMASHTi+CO2) using the XRF bulk composition (Supplementary Data 3) with the measured CO2 content, and H2O and SiO2 in excess for sample H1813B. It is generally accepted that subducted sediments underwent restricted decarbonation during subduction metamorphism along low-temperature geotherms41. Therefore, the measured CO2 content is the total CO2 content of the modeling system. We modified the bulk composition to assume apatite saturation by removing P2O5 and the corresponding CaO using a P/Ca molar ratio of 3/5. The P–T diagrams were calculated with Perple_X42,43 and the thermodynamic database (hp62ver.dat)44. Mineral solid solution models were: (i) Gt(W) for garnet, Mica(W) for white mica, Chl(W) for chlorite, Ctd(W) for chloritoid45, (ii) cAmph(G) for amphibole, Omph(GHP) for omphacite46, (iii) feldspar for feldspar47, (iv) Carp(M) with two endmembers (Mg-Fe) for carpholite48, (v) F for H2O–CO2 fluid solution49, (vi) oCcM(EF) for carbonate50. Pure endmember phases included zoisite, lawsonite, quartz/coesite, ankerite, aragonite, rutile, ilmenite, and titanite.