Introduction

Black carbon (BC) is highly condensed carbonaceous residue mainly from incomplete combustion processes1,2. The modern annual emissions from combustion of fossil fuel and burning of terrestrial biomass are 6.6–7.2 Tg (1012 g)3 and 196–340 Tg, respectively4. Due to its refractory aromatic structures, BC has a long environmental lifetime5, providing sufficient time for BC to reach remote oceans6,7. The BC burial flux in ocean sediments was estimated to be 10.3 Tg yr−1, of which only 0.6 Tg yr−1 was accumulated in deep-sea sediments8. However, the BC to total organic carbon (TOC) ratio (BC/TOC) in deep-sea sediments is usually higher than that in continental shelf sediments9. Once deposited in deep sea, BC escapes from photo-degradation and is relatively resistant to microbial degradation, thus representing an important slow-cycling carbon retention10. Previous studies have reported the BC in sublittoral, bathyal and abyssal zones (water depth <6000 m)6,8,10,11,12, but no data exist for sediments in the hadal zone (water depth of 6000–11,000 m). A recent study13 examined dissolved BC in waters from Mariana Trench (MT) and found that it is cycled and aged on the same time scales as bulk dissolved organic carbon (OC) in the ocean. This finding, however, is not consistent with the prevailing view that the BC was thought to be a refractory pool of carbon with much old 14C ages (~20,000 yr BP)10,14.

The hadal zone is mainly composed of trenches formed at the subduction zones. Despite representing only ~1% of total seafloor area, the hadal zone accounts for 45% of the ocean’s depth range15,16. Hadal trench ecosystems are one of the least explored habitats on Earth due to extreme challenges of sampling15. Recent studies have suggested that elevated OC accumulation rates17,18, benthic abundance and biomass19,20, and microbial activities21,22,23 are prevalent in some hadal trenches. These characteristics have been attributed to the funneling effect of hadal trenches due to unique V-shaped topography, further facilitated by occasional seismic activities and high-frequency fluid dynamics18,24,25. As an example, >7 Tg of OC has been remobilized from surficial slope sediments and exported to the hadal realm of the Japan Trench by earthquakes in the past 2000 years26, confirming the importance of hadal trenches as accumulation regions for OC in the deep sea. Given a close relationship between BC and TOC contents in marine sediments7, we hypothesize that large amount of BC has been accumulated in the hadal zone.

Here, we collected sediment samples from the six hadal trench regions (hadal sites and surrounding non-hadal sites), including MT, Mussau Trench (MST), New Britain Trench (NBT), Bougainville Trench (BT), Kermadec Trench (KT) and Atacama Trench (AT) (Fig. 1; Supplementary Table 1). BC was quantified using the chemothermal oxidation-375 (CTO-375) method. We also measured stable and radioactive carbon isotopes (δ13C and Δ14C) of BC and TOC using isotope ratio mass spectrometry and accelerator mass spectrometry (AMS), respectively. The δ13C enables distinguishing OC from terrestrial C3 plants, C4 plants and marine phytoplankton, which have an average δ13C signature of −27.1 ± 2.0‰, −13.1 ± 1.2‰ and −20 ± 1.0‰, respectively27,28. Since there is only minor isotopic fractionation between the product of vegetation burning and original vegetation biomass (1–2‰)29, the δ13C signature of BC is a reliable source indicator. In addition, the Δ14C is indicative of the contributions of biomass burning versus fossil carbon combustion to the BC pool because fresh biomass has a modern radiocarbon age (Δ14C ≥ 0), whereas fossil carbon is completely devoid of radiocarbon (Δ14C = − 1000‰)30,31. Based on the contents, δ13C and Δ14C signatures of both BC and TOC, and 210Pb-derived sedimentation rates, we hereby attempt to determine the concentration, source, and distribution of BC in the hadal trenches, and to assess the role of the hadal zone for BC burial in the global ocean.

Fig. 1: Map showing sampling sites of six trench regions in the Pacific Ocean.
figure 1

The site information is summarized in Supplementary Table 1. The numbers in bracket denote water depth range at each trench. MT Mariana Trench, MST Mussau Trench, NBT New Britain Trench, BT Bougainville Trench, KT Kermadec Trench, AT Atacama Trench. Base map was drawn using the Ocean Data View software package63.

Full size image

Results

Content, δ13C and Δ14C of TOC in trench surface sediments

The content, δ13C and Δ14C of BC and TOC in trench surface sediments are summarized in Supplementary Table 2. The TOC content (expressed as weight percentage of TOC to sediment; same hereafter) ranged from 0.15% to 2.39%, with an average of 0.31% in MT, 0.57% in MST, 0.45 ± 0.18% in NBT, 0.26 ± 0.08% in BT, 0.24 ± 0.09% in KT, and 0.87 ± 0.58% in AT. The δ13C values of sedimentary TOC varied between −24.4‰ and −19.9‰, with an average of −20.0‰ in MT, −20.0‰ in MST, −22.0 ± 1.6‰ in NBT, −20.7 ± 1.0‰ in BT, −23.1 ± 0.4‰ in KT, and −22.0 ± 1.0‰ in AT. The Δ14C values of sedimentary TOC ranged from −891‰ to −82‰, corresponding to 14C ages of 17,790 to 690 yr BP. For each trench, the Δ14C of sedimentary TOC was −391‰ in MT, −418‰ in MST, −172 ± 65‰ in NBT, −224 ± 126‰ in BT, −683 ± 157‰ in KT, and −446 ± 151‰ in AT.

Content, δ13C and Δ14C of BC in trench surface sediments

The BC content ranged from 0.011% to 0.18%, with an average of 0.013% in MT, 0.071% in MST, 0.024 ± 0.009% in NBT, 0.014 ± 0.002% in BT, 0.036 ± 0.007% in KT, and 0.077 ± 0.046% in AT. The δ13C of BC varied between −27.7‰ and −22.3‰, with an average of −24.3‰ in MT, −24.0‰ in MST, −23.7 ± 0.9‰ in NBT, −25.9 ± 0.9‰ in KT, and −25.7 ± 1.0‰ in AT. Note the lack of BC-δ13C data for BT due to insufficient samples. The Δ14C of BC ranged from −893‰ to −346‰, corresponding to 14C ages of 17,870 to 3340 yr BP. For each trench, the Δ14C of sedimentary TOC was −626‰ in MT, −661‰ in MST, −473 ± 77‰ in NBT, −415 ± 49‰ in BT, −729 ± 105‰ in KT and −694 ± 164‰ in AT.

Depth profiles of TOC and BC in trench core sediments

More extensive investigations were conducted for two cores from the landward abyssal site (A9) and hadal axis site (A10). The TOC, BC, δ13C and Δ14C of these cores were measured down to a depth of 15 and 35 cm, respectively, and the results are summarized in Supplementary Tables 3 & 4. In the core from site A9, the TOC and BC contents varied from 0.90% to 1.56% and 0.11% to 0.20%, respectively, whereas the δ13C fell in a narrow range for TOC (−21.7‰ to −21.1‰) and BC (−26.6‰ to −25.9‰). The sediments from surface (0–1 cm), middle (5–6 cm) and deep (9–10 cm) layers, respectively, had Δ14C values of −319‰, −434‰ and −760‰ for TOC, and −457‰, −453‰ and −767‰ for BC. In the core from site A10, the TOC and BC content ranged from 1.01% to 2.07%, and 0.09% to 0.18%, respectively. Like the core from site A9, one from site A10 has a relatively narrow range of δ13C values for TOC (−22.2‰ to −21.2‰) and BC (−26.4‰ to −25.6‰). The sediments at 0–1 cm, 7–8 cm, 17.5–20 cm and 30–35 cm layers, respectively, had Δ14C values of −311‰, −332‰, −341‰ and −403‰ for TOC, and −430‰, −386‰, −411‰ and −502‰ for BC.

Discussion

A weak but still significant correlation was observed between the BC and TOC contents (r = 0.43, p < 0.05; Fig. 2a). The inter-trench comparison revealed that the BC content was the highest in AT (0.077 ± 0.046%) compared to KT (0.036 ± 0.007%), NBT (0.024 ± 0.009%) and BT (0.014 ± 0.002%) (Fig. 3a). Correspondingly, the AT had the highest TOC content (0.87 ± 0.58%), followed by NBT (0.45 ± 0.18%), BT (0.26 ± 0.01%) and KT (0.24 ± 0.09%) (Fig. 3b). The fact that we only have one sample from MST and MT prevented a robust inter-trench comparison including these two trenches, and thus they have been omitted for this part of the discussion. The highest BC/TOC ratio occurred in KT (0.17 ± 0.05), followed by AT (0.11 ± 0.08), NBT (0.074 ± 0.049) and BT (0.058 ± 0.017) (Fig. 3c). The KT underlies mesotrophic surface waters (Net primary productivity: 135–152 g C m−2 yr−1) that are expected to export moderate amount of marine OC to the seafloor32. The closest countries to the KT are industrialized countries that use fossil fuel primarily for transport (New Zealand) and for the production of electricity (Australia). Thus, the KT may receive substantial amounts of BC from fossil fuel combustion, a notion that is supported by relatively old radiocarbon age of the BC in KT (Fig. 4a). In AT, strong upwelling has induced high net primary productivity (306–449 g C m−2 yr−1) that is suspected to export large amount of marine OC (non-BC) to the seafloor, while Atacama Desert, one of the most arid region of the world, supplied little fresh terrestrial OC33. In addition, most AT sites are close to Antofagasta, a harbor city in northern Chile (Fig. 1). This city, plus mining activities in Atacama Desert34, could produce BC from fossil fuel combustion and rock weathering that was transported to AT by prevailing westerly wind. Taken together, high marine primary productivity, the lack of vegetation cover and important emission of fossil BC result in an intermediate BC/TOC content in AT sediments. The lowest BC/TOC ratio occurring in the BT and NBT regions suggests that sedimentary TOC is mainly derived from non-BC OC such as phototrophic marine products from phytoplankton and terrestrial materials from tropical forests in adjacent Papua New Guinea35. This finding is consistent with less BC emission from fossil fuel combustion in less industrialized Papua New Guinea as compared to more industrialized countries like New Zealand and Australia13.

Fig. 2: Linear correlations for different types of carbon in surface sediments from six trench regions.
figure 2

a Total organic carbon (TOC) and black carbon (BC) contents. b TOC radioactive carbon isotopic composition (Δ14C) and BC/TOC ratio. c BC contents with and without hydrofluoric acid (HF)-treatments. The dotted lines are 95% credible intervals.

Full size image
Fig. 3: Summary results of black carbon (BC) content, total organic carbon (TOC) content, and BC/TOC ratio in surface sediments from six trench regions.
figure 3

a BC content. b TOC content. c BC/TOC ratio. AT Atacama Trench, BT Bougainville Trench, KT Kermadec Trench, MST Mussau Trench, MT Mariana Trench, NBT New Britain Trench. The dash lines denote average values of the parameters in each trench.

Full size image
Fig. 4: Offset of carbon isotopic compositions between black carbon (BC) and total organic carbon (TOC) in surface sediments from six trench regions.
figure 4

a radioactive carbon isotopic composition (Δ14C). b stable carbon isotopic composition (δ13C). AT Atacama Trench, BT Bougainville Trench, KT Kermadec Trench, MST Mussau Trench, MT Mariana Trench, NBT New Britain Trench. Note the δ13C data of the BC at sites B1, B2 and B3 (BT) are not available.

Full size image

Along each trench axis, the BC content and BC/TOC ratio, respectively, varied in a range of 0.04–0.12% (0.07 ± 0.03%) and 0.078–0.29 (0.12 ± 0.08) in AT (A2–A6, A10), and 0.03–0.05% (0.04 ± 0.01%) and 0.13–0.23 (0.17 ± 0.05) in KT (K3–K6) (Fig. 3c). Previous studies have reported high heterogeneity of sedimentary TOC in Yap Trench36, NBT37,38, MT21, KT and AT39. However, to the best of our knowledge, our study is the first report for the heterogenous distribution of BC in hadal trenches. This feature is likely attributed to complex environmental processes due to turbidity currents40, internal tides24 and complex bathymetry within trench interiors41. The site-to-site comparisons of the BC content between hadal and non-hadal surface sediments (i.e., A1 vs. A5; A3 vs. A7; A9 vs. A10) tend to decrease from landward bathyal/abyssal sites (A1, A9) to hadal axis sites (A3, A5, A10) and then to oceanward abyssal site (A7) (Fig. 3a). This seaward decrease suggests that the BC preserved in the AT region is primarily from its adjacent land on the east (i.e., Chile). In contrast, in the NBT and BT regions, the BC content does not show a decreasing trend from landward site to hadal axis site and then to oceanward site (i.e., N4–N1–N3–N2, and B2–B1–B3; Fig. 3a), suggesting that the BC input is not only from New Britain Island and Bougainville Island on the north, but also likely from the mainland of Papua New Guinea on the south.

The δ13C of BC in surface sediments in AT (−27.7‰ to −24.2‰) and KT (−27.3‰ to −25.0‰) is more negative than that in NBT (−24.7‰ to −22.3‰) (p < 0.01), MT (−24.3‰) and MST (−24.0‰) (Fig. 4b), presumably reflecting inter-trench variability in the BC sources (e.g., biomass burning from different type plants or combustion of different fossil fuels)13,28,42. While within each trench (i.e., AT), BC δ13C values do not show significant difference between hadal and non-hadal sediments (−26.3‰ to −24.2‰ vs. −27.7‰ to −24.7‰) (p > 0.05). Taking all surface sediments together, the average BC δ13C value (−25.1 ± 1.2‰) is significantly more negative than average TOC δ13C value (−21.9 ± 1.4‰; Fig. 4b) (p < 0.01), suggesting different sources between BC (predominant terrestrial source) and co-deposited TOC (predominant marine source). Such δ13C offset between BC and TOC was also found in marine sediments from South Atlantic (−23‰ to −27‰ vs. −18‰ to −24‰)6 and the Santa Monica Basin (−24.2‰ to −25.0‰ vs. −21.5‰ to −23.5‰)43.

Interestingly, oceanic dissolved BC, unlike sedimentary BC, presents less negative δ13C signature similar to that of marine phytoplankton. For example, the δ13C of BC in MT sediment is −24.3‰ (this study), whereas that of dissolved BC in MT waters (8000–10,000 m) varies between −20.4‰ and −18.9‰13. At the ocean basin scale, the δ13C of sedimentary BC in six trench regions is −25.1 ± 1.2‰, while that of dissolved BC in Pacific and Atlantic oceans ranges from −22.9‰ to −24.3‰44. This difference may reflect different sources and/or biogeochemical processes between dissolved and sedimentary BC. The result based on the benzene polycarboxylic acid method revealed that oceanic dissolved BC was 6–8‰ enriched in 13C compared to riverine dissolved BC44, while that based on the CTO-375 method also showed that dissolved BC became enriched in 13C from rivers (−26.4‰ to −22.6‰) to ocean (−20.9‰ to −18.4‰)13. This positive shift in δ13C suggests that oceanic dissolved BC either is predominantly derived from a non-riverine source or undergoes isotopic fractionation potentially associated with photo- and bio-degradation. Currently, there is no direct evidence supporting conservative or nonconservative behaviors of BC δ13C during transport and mixing into the ocean, but a large fraction of riverine dissolved BC could be degraded before it reaches open ocean13. Note that the methods like the benzene polycarboxylic acid and CTO-375 have different analytical windows for BC (non-aromatic, condensed aromatic and graphitic fractions)45, and their results are not completely overlapping. However, both Qi et al.13 and our study used the CTO-375 method for the BC in MT and the results still reveal distinct δ13C values between dissolved and sedimentary BC, suggesting that it is not the analytical methods that cause the δ13C offset between oceanic dissolved and sedimentary BC. Considering an importanf mineral matrix on the fate of particulate organic matter in the ocean46 and soil47, we propose that the mineral protection may minimize the isotopic fractionation for particulate BC along the river to ocean continuum, whereas such protection mechanism is absent for dissolved BC. This hypothesis, however, needs to be tested in further work.

In 20 out of the 22 surface sediments, the BC Δ14C values (−893‰ to −346‰) were more negative than the TOC Δ14C values (−891‰ to −82‰) (Fig. 4a), corresponding to an average age offset of 3720 years. A negative correlation between the BC/TOC ratio and Δ14C age of the TOC (r = −0.59; p < 0.01; Fig. 2b) suggests an important influence of the BC content on the bulk radiocarbon ages of sedimentary TOC and a negligible influence of charring during our pretreatment processes for BC analysis. The older age of BC relative to the co-deposited TOC was also found in Arctic sediments11,48, Northeastern Pacific and Southern Ocean sediments10. Thus, the BC preserved in deep-sea sediments generally appeared to be of an older refractory carbon source compared to non-BC OC.

The distribution of the δ13C versus Δ14C of BC from the trench regions (this study) and other marine settings (previous studies) is depicted in Fig. 5. Overall, our BC data from hadal trenches are characterized by more depleted δ13C compared to those from Arctic Ocean11 and Washington coast (North Pacific Ocean)42, and more negative Δ14C compared to those from the East China Sea12 and Santa Monica Basin (North Pacific Ocean)43. The application of Δ14C-based two endmember mixing model shows apparent inter-trench variability in the BC source (Fig. 6). The proportion of fossil carbon (ffossil) and biomass burning (fbiomass) ranges from 38% to 90% (64 ± 16%) for ffossil and 10% to 62% (36 ± 16%) for fbiomass. The mean fbiomass is the highest in the NBT and BT regions (50%, 56%), intermediate in the MST and MT regions (32%, 36%), and the lowest in the KT and AT regions (26%, 29%). This distribution pattern is consistent with the global 14C dataset for atmospheric BC that shows larger biomass contributions in rural and less developed regions49, such as the less industrialized Papua New Guiana close to NBT and BT, and larger fossil carbon contributions in more industrialized regions such as New Zealand and Australia as the nearest lands to KT. The low fbiomass in AT can be explained by the unique feature of its BC-source region (Atacama Desert) where the sparse vegetation contributed to little biomass burning derived BC, but the anthropogenic activities (e.g., mining and related industries) could produce the BC from rock weathering and fossil fuel combustion50.

Fig. 5: Cross-plot of stable carbon isotopic composition (δ13C) and radioactive carbon isotopic composition (Δ14C) for the Black carbon (BC) in sediments.
figure 5

The data are from trench regions (this study) and Arctic Ocean11, East China Sea12, Washington coast (North Pacific Ocean)42, and Santa Monica Basin (North Pacific Ocean)43. Error bars denote standard deviation.

Full size image
Fig. 6: Radioactive carbon isotopic composition (Δ14C)-based estimate for proportion (%) of fossil carbon and biomass burning to the Black carbon (BC) pool in surface sediments from six trench regions.
figure 6

AT Atacama Trench, BT Bougainville Trench, KT Kermadec Trench, MST Mussau Trench, MT Mariana Trench, NBT New Britain Trench. The vertical dash lines denote 50% contribution.

Full size image

Both the hadal axis core site (A10) and landward abyssal core site (A9) display similar and stable δ13C signals of the BC (−26.1 ± 0.3‰ and −26.2 ± 0.3‰, respectively), suggesting that BC is predominantly derived from C3 terrestrial plant material. The BC content and BC/TOC ratio, however, show some differences between the two cores. The core from site A9 is characterized by higher levels for both BC content (0.15 ± 0.03%) and BC/TOC ratio (0.14 ± 0.02) compared to that from site A10 (0.12 ± 0.02% and 0.084 ± 0.013) (p < 0.01) (Fig. 7a). Such differences are likely caused by different distance to the BC source regions with the core from site 9 being closer to the shoreline of Chile. In this core, the BC content exhibits a decrease first and then an increase post-1940s (3–0 cm), whereas the BC/TOC ratio remains at a stable level until 1880 and increased afterwards that is coeval with increasing human activities in the 20th century. This profile, however, is not observed in the core from site A10 that does not show a clear trend except for a peak at ca. 1700 AD (9 cm) for the BC content and a drop during ca. 1730–1810 AD (8–6 cm) for the BC/TOC ratio (Fig. 7b)16. These abrupt changes may be caused by mass-wasting events, such as submarine landslides, because of the high seismicity17,50. Nevertheless, the BC content in surface sediments (0.18% in A9 and 0.11% in A10) is similar to an average level of respective core (0.15% in A9 and 0.12% in A10), suggesting that the BC content in surface sediments may represent a long-term average at this site. Using this rationale, we below estimate BC accumulation rates based on BC values of the surface sediments.

Fig. 7: Depth (Age) profiles of the Black carbon (BC) content and BC over total organic carbon (TOC) ratio in two sediment cores in the Atacama Trench (AT).
figure 7

a non-hadal site A9. b adjacent hadal site A10.

Full size image

The BC preserved in marine sediments represents a long-term carbon sink for atmospheric CO2. The sedimentation rate obtained from regression of 210Pbex (excess 210Pb activity) versus core depth ranges from 0.026 to 0.20 cm yr-1 (Supplementary Table 5)16. Based on the data of sedimentation rate, BC content and dry sediment bulk density, we calculated the accumulation rate of BC in five trench regions, i.e., MT, NBT, BT, KT, and AT. The detailed calculation procedures are explicated in Supplementary Note 4. The results range from 0.07 to 0.48 g m−2 yr−1 with an average value of 0.29 ± 0.15 g m−2 yr−1 for the BC accumulation rate (Table 1; Supplementary Table 6), which is comparable with that for the China continental shelf sea51,52, but 2 order of magnitude greater than that for the open ocean sediments of the Pacific Ocean53 and South Atlantic Ocean6 (Supplementary Table 7). Our finding of higher BC accumulation rates in the deepest trenches is consistent with previous studies that showed higher OC accumulation rates38 and higher concentrations of pollutants54 in the hadal zone relative to surrounding abyssal/bathyal zones. This characteristic is attributed to the hadal zone’s V-shaped topography with steep slopes that can create gravity-driven downward transport, and even mass-wasting events occurred frequently in the hadal zone25,38.25

Table 1 Black carbon (BC) and total organic carbon (TOC) burial estimates in the hadal zone.
Full size table

The quantification of the global BC burial in the hadal zone is still confounded by limited sampling and the heterogeneity of BC content among and within trenches. Nevertheless, our target trenches span a variety of geographically settings, i.e., MT and MST (low primary productivity and remote from lands), NBT/BT (moderate primary productivity and close to less developed country), KT (moderate primary productivity and remote from lands), and AT (high primary productivity and close to the developing country). We extrapolated the estimate from five trenches to the global hadal zone. MST is not included due to the unavailability of sedimentation rate data. Multiplying the number of five trenches by global hadal zone surface area (ca. 3.4 × 106 km2)16 resulted in the accumulation rate of 1.0 ± 0.5 Tg yr−1 (Table 1). This number is greater than previous estimate for the deep sea (0.6 Tg yr−1) by about 70%8. Our result has updated the global marine BC accumulation rate from 10.2 Tg yr−1 to 11.2 Tg yr−1, of which the hadal zone is responsible for 8.7 ± 4.5%. Per unit area, the hadal zone, being ~1% of total seafloor area, has a BC burial rate greater than global ocean average by sevenfold. Note that our estimate does not consider episodic events such as earthquakes that could trigger mass-wasting events and transport enormous amounts of sediment and associated carbon into the hadal zone in a short time17,55, suggesting that our estimate is conservative and likely lower than the actual BC burial flux in the hadal zone. Nevertheless, our estimate for the burial rate of TOC in the hadal zone (17.7 ± 14.1 Tg yr−1; Table 1) is consistent with the previous estimate for the TOC subducted into the trenches based on global sediment cores from non-hadal sites adjacent to the subduction trenches (12 Tg yr−1)56. This consistency supports that our hadal sites-based estimate for TOC and BC burial rates is reasonable. Thus, the hadal zone is an important, but previously unrecognized sink for the BC in the deep ocean.

Methods

Sample collections

A series of short sediment cores (up to 40 cm) were collected from hadal and non-hadal sites of MT, MST, NBT, BT, KT and AT during three cruises (2016–2018) (Fig. 1). A part of cores was sliced, with representative subsamples used in this study. The investigated trenches are in different biogeochemical provinces of the Pacific Ocean, characterized by different levels of net primary productivity (48–449 g C m−2 yr−1)32, water depth (1560–10,840 m) and terrestrial influences. The detailed sampling and site information is summarized in Supplementary Note 1 and Supplementary Table 1.

Pretreatment for TOC and BC measurements

Sediments were freeze-dried and homogenized by an agate mortar and pestle. 2–3 g of dry sediments was weighted into a 50-ml centrifuge tube and was mixed with excess 1 M hydrochloric acid (HCl) to remove inorganic carbon. After reaction for 24 h during which the samples were stirred for several times by a vortex stirrer, the sediments were centrifuged (3800 rpm, 10 min) and the supernatants were carefully removed by pipettes. The residue was further mixed with 20 ml ultrapure water (18.2 MΩ.cm). After centrifuged (3800 rpm, 10 min), the supernatant was again carefully removed. Repeat this step for three times until the pH was close to 7. The decarbonated sediments were freeze-dried and homogenized. Aliquots of decarbonated sediments (ca. 1 g) were used for analyses of content, δ13C and Δ14C of TOC and BC. However, some samples from the BT, KT and MT had extremely low BC contents, and were unable to be directly quantified for BC content. Thus, the decarbonated surface sediments (ca. 1 g) were further concentrated by reaction with excess hydrofluoric acid (HF) for 24 h. In order to avoid the loss of OC and BC, the HF-treated sediments were not rinsed with ultrapure water, and instead were dried at 60 °C prior to analyses of BC content.

Currently, there is no universally accepted method for quantifying BC content. The chemothermal oxidation method at 375 °C (CTO-375) is suitable for soot/graphitic BC in sediments and soils57,58, which was used in our study. About 1 g (dry weight) of acid treated sediment was placed in a thin even layer (1–2 mm) in a crucible and heated at 375 °C for 24 h in an oven with continuous air supply. The residual carbon was operationally defined as BC57. In order to test if the BC content was altered during HF treatment, we chose eight surface sediment samples that have relatively high BC contents, and compared their BC contents with and without HF-treatment (w/o HF). The results show a strong positive correlation: BC content with HCl + HF (%) = 1.112 × BC content with HCl (%) −0.026 (r = 0.93, p < 0.01; Fig. 2c). Therefore, the HF treatment did not significantly change the original BC content.

Measurement for contents and carbon isotopes of TOC and BC

A total of 22 surface sediments (top 1–5 cm) and 2 sediment cores (29 samples) were analyzed. The contents and δ13C of TOC and BC were measured using a Vario Pyro Cube elemental analyzer and an Isoprime 100 continuous flow isotope ratio mass spectrometer, respectively. Acetanilide (IAEA) was used as an intra-lab reference standard for both carbon content (71.09%) and δ13C (−29.53‰). The TOC and BC contents were reported as weight percent of TOC (or BC) to dry weight sediment (wt.%), whereas δ13C was reported in per mil (‰) using δ notation relative to the Vienna Peedee Belemnite standard (VPDB). The average standard deviation of each measurement, determined by replicate analyses, was ±0.03% for TOC or BC contents and ±0.2‰ for δ13C.

The Δ14C of sedimentary TOC and BC was determined using AMS in the Center for Isotope Geochemistry and Geochronology (CIGC) at Qingdao National Laboratory for Marine Science and Technology, China (QNLM) and the Ocean University of China radioCarbon Accelerator Mass Spectrometer Center (OUC-CAMS). Part of samples containing >50 μg C were mixed with pre-combusted copper oxide (CuO) as an oxygen source and combusted into CO2 at 900 °C for 8 h using a vacuum line. The CO2 was purified, quantified, graphitized using a zinc reduction method and then analyzed by AMS at the CIGC-QNLM. While for other samples containing carbon as low as 28 μg, they were combusted in an elemental analyzer (EA) and the product CO2 was directly injected into the ion source of MICADAS through the Gas Interface System (GIS) at OUC-CAMS59. All 14C measurements were reported as fraction modern (Fm) and the conventional radiocarbon ages (year before present, BP) were calculated using the Libby half-life as previously described60. Samples were normalized using oxalic acid II (NIST SRM4990C) and phthalic anhydride as a blank. The Fm precision of CO2 gas measurements for the international standard oxalic acid II was better than ±1%, whereas for the PhA, the Fm precision was <10% and Fm value was <0.005 (~42 ka B.P.). Radiocarbon values were corrected for procedural blanks with modern and dead standards known Fm values. Procedural blanks yield 1.50 ± 0.38 μg C with an Fm value of 0.47 ± 0.12 (n = 8).

Sedimentation rates in hadal trench regions

Because episodic mass wasting events can transport large quantities of sediments from the surrounding non-hadal area (e.g., trench slope) into the trench interior, steady-state accumulation in hadal trenches may have been interrupted by episodic events. Here, we used constant linear sedimentation rates previously reported16,21,37 that are based on excess 210Pb activity (210Pbex) in sediment cores by excluding mass wasting events (if any). The sedimentation rates vary from 0.08 to 0.16 cm yr−1 in NBT, 0.13 to 0.20 cm yr−1 in BT, 0.026 to 0.085 cm yr−1 in AT, 0.030 to 0.042 cm yr−1 in KT, and 0.038 cm yr−1 in MT (Supplementary Table 5).

Estimating contributions of biomass (fbiomass) and fossil carbon (ffossil)

The BC in marine sediments can be derived from fossil fuel combustion, biomass burning and rock weathering10,42. Since fossil fuel and rock weathering derived BC have overlapping δ13C and Δ14C signals42, we combined them as one endmember (fossil carbon) in the isotope mass-balance mixing model. The relative contribution to the BC from biomass burning (fbiomass) and fossil carbon (ffossil) was calculated as following:

$${\varDelta }^{14}{{{{{{\rm{C}}}}}}}_{{{{{{\rm{sed}}}}}}}={\varDelta }^{14}{{{{{{\rm{C}}}}}}}_{{{{{{\rm{biomass}}}}}}}\times {{{{{{\rm{f}}}}}}}_{{{{{{\rm{biomass}}}}}}}+{\varDelta }^{14}{{{{{{\rm{C}}}}}}}_{{{{{{\rm{fossil}}}}}}}\times {{{{{{\rm{f}}}}}}}_{{{{{{\rm{fossil}}}}}}}$$
(1)
$$1\,{{\mbox{=}}}\,{{{\mbox{f}}}}_{{{\mbox{biomass}}}}\,{{\mbox{+}}}\,{{{\mbox{f}}}}_{{{\mbox{fossil}}}}$$
(2)

Here, \({\triangle }^{14}{{{\mbox{C}}}}_{{{\mbox{sediment}}}}\), \({\triangle }^{14}{{{\mbox{C}}}}_{{{\mbox{biomass}}}}\) and \({\triangle }^{14}{{{\mbox{C}}}}_{{{\mbox{fossil}}}}\) represent the radiocarbon activity of the BC in trench sediments, fresh biomass and fossil carbon, respectively. \({\triangle }^{14}{{{\mbox{C}}}}_{{{\mbox{biomass}}}}\) is similar to Δ14C signal of modern atmospheric CO2 that was assigned as +50‰, whereas \({\triangle }^{14}{{{\mbox{C}}}}_{{{\mbox{fossil}}}}\) was designated as −1000‰ since fossil carbon from the weathered rock and fossil fuel combustion is completely devoid of radiocarbon61,62. Given high sedimentation rates in hadal trenches (0.026–0.20 cm yr−1), the surface sediments (most top 5 cm) were deposited within the past <200 years that is much shorter than the half-life of 14C age (5730 yr). Thus, the 14C ages were not corrected for the decay itself. Note that the pre-aged BC from soils and pre-deposited sediments characterized by intermediate 14C ages between modern biomass and fossil carbon was not considered in our two-endmember mixing model due to the lack of endmember values, which may underestimate the contribution of biomass (fbiomass) to some extent.