Introduction

Given the important role of abiotic N2 reduction (ANR) in early Earth’s geological nitrogen (N) cycle, particularly in providing the essential compound of NH3 (or its dissolved equivalent NH4+) for the origin of life in submarine hydrothermal vents1,2,3 and for establishing a warm and habitable environment under dimmer solar luminosity (i.e., the faint young Sun paradox)4,5 on the N2-dominant early Earth surface6, this process has been intensively tested by laboratory experiments. The experimental results demonstrated that ANR could be catalyzed by a variety of naturally occurring minerals and rocks (e.g., FeS, Fe-Ni alloy, green rust, magnetite, peridotite) under submarine hydrothermal conditions7,8,9,10,11,12,13,14,15,16,17,18. However, ANR has not been convincingly detected in field samples, particularly modern submarine hydrothermal vent fluids that have been collected for study. One of the major responsible factors is the overprinting by NH4+ generated in shallow hydrothermal systems from biologically processed N sources such as organic matter, dissolved NO3− and NH4+ in seawater, and dissolved organic N and NH4+ in pore water of sediment. For example, when a sediment cover exists near hydrothermal systems, elevated temperature (T) condition can induce decomposition of organic matter and/or NH4+ desorption from clays19. In addition, dissolved NO3− in seawater can be effectively reduced (even abiotically) in shallow hydrothermal systems as low as 24 °C17,18. These processes can contribute remarkable amounts of NH4+ to increase the NH4+ concentrations of shallow hydrothermal fluids to more than an order of magnitude higher than the ambient seawater NH4+ concentration (~ 1 μM)20. Because these NH4+ components are all derived from surface N sources in shallow localities, they are referred to as surface NH4+ hereafter. Due to the similar N isotopic signatures of these surface N sources, e.g., +3‰ to +8‰ for dissolved NO3− in seawater21, +2‰ to +10‰ for marine organic matter/sediments22 and similar range (with high value up to +17‰) for dissolved organic N and NH4+ in interstitial water23, it is difficult to distinguish between NH4+ derived from these surface sources. Regardless, such 15N-enriched surface NH4+ can effectively overprint the 15N-depleted signal from deep source (e.g., –5‰ for the upper mantle)24,25 in shallow hydrothermal fluids.

A better geological proxy for detecting an ANR signal would be the overlooked hydrothermal veins deposited from focused flow of deep hydrothermal fluids, for two reasons. Firstly, although deep hydrothermal fluids could be derived from deeply circulated seawater, surface NH4+ could be progressively consumed (by NH4+ assimilation into alteration minerals in oceanic crust26,27,28,29,30,31) along seawater circulation pathway into depths. Consequently, the deep hydrothermal fluids should undergo minimal impact from surface NH4+ and have the best chance to expose NH4+ produced by ANR. Secondly, studies on field samples32,33, laboratory experiments34 and theoretical calculations35 suggest that NH4+ can substitute K+ and Na+ in silicate minerals. Thus NH4+ in deep fluids can partially partition into K+- and Na+-bearing vein minerals (e.g., plagioclase, epidote, chlorite) upon their deposition. Once fixed in the structure of vein minerals, NH4+ can be well protected from subsequent low-T disturbance. As a result, vein minerals deposited in the focused flow channel of deep fluids can best reveal the deep-fluid NH4+ signature and have the best chance to disclose the ANR signal (if there is any).

Despite recent advance in characterizing the NH4+ signature of hydrothermally altered oceanic crust (i.e., seafloor basalts, sheeted dikes and gabbros, and serpentinized peridotites)26,27,28,29,30,31, NH4+ in hydrothermal veins in oceanic crust has been rarely examined by far. International Ocean Discovery Program (IODP) Expeditions 367 and 368 drilled into the 16–32 million-year-old oceanic crusts in the South China Sea basin36 (Fig. 1; Supplementary Information). Recovered mid-ocean ridge basalts (MORB) from Hole U1502B show an E-MORB affinity (Supplementary Fig. 2). All these rocks have been altered at various degrees and contain abundant hydrothermal veins with thickness from submillimeter to a few millimeters (Supplementary Fig. 3). The mineral assemblage of hydrothermal veins is dominated by albite and quartz with variable amounts of Fe-Mg-Ca carbonates, chlorite, epidote, pyrite, and Fe–Mn hydroxides (“Methods”; Supplementary Data 1 and Supplementary Fig. 4), which were precipitated from relatively high-T fluids (200−300 °C; Supplementary Information)36,37,38. The hydrothermal fluids, as determined from trace elements and radiogenic isotopes of vein minerals, were a mixture of modified seawater (after reaction with MORB) and deep magmatic fluid38. Here we report the N concentrations and isotope compositions of these vein samples, in comparison with those of their hosting altered MORB, to constrain the N cycle pathways and estimate the NH4+ flux in the deep oceanic hydrothermal systems.

Fig. 1: IODP Hole U1502B location and downhole lithology, N concentration, molar N/K and N/Na ratios, and δ15N values of altered basalts and hydrothermal veins.

In the location map, dashed lines represent the magnetic anomaly lineations. Note the scale change in the N/K and N/Na ratios marked by the dashed lines.

Results and discussion

N enrichment in altered MORB and veins

The altered MORB from Hole U1502B have bulk-rock N concentrations from 18 to 40 μg/g (Fig. 1), which are much higher than that of fresh MORB (<2 μg/g) but still fall in the upper end of the N concentration range of global altered MORB (2–48 μg/g)29, indicating that the alteration-induced N enrichment in these basalts is similar to those in global altered MORB. The δ15N values of the U1502B basalts (–7.6‰ to +0.2‰; Supplementary Data 2) are lower than the values of the surface source21,22,23 and thus suggest that the secondary N came from not only seawater/sedimentary source but also a 15N-depleted source. However, similar to the low-δ15N hydrothermally altered basalts from ODP Sites 801 and 114927 and DSDP Site 41729, no good correlation was observed between the bulk-rock concentrations of N and any other elements (Supplementary Data 2; Supplementary Fig. 5). This can be attributed to the variable hydrothermal conditions over their alteration history, e.g., NH4+ content, temperature, and secondary mineral assemblage (see detailed discussion in Yu et al.19). For example, high-T (>~300 °C) alteration minerals (e.g., amphibole and albite) have lower NH4+-hosting capabilities than low-T (<~300 °C) alteration minerals (e.g., clays)19,30. Some minerals (e.g., clays) even have multiple NH4+-hosting sites with high NH4+-hosting capacity in interlayer sites and low NH4+-hosting capacity in surface and edge sites19. Furthermore, oceanic basalts might experience low-T microbial alteration by lithochemotrophic organisms39 which could also contribute some biological N in rocks, but has not been quantified so far. All these complexities prevent further identification and quantification of the secondary N sources and its connection to ANR in altered MORB samples.

Veins from U1502B contain more abundant N (14–180 μg/g) with extremely low δ15N values from –3.8‰ to –20.4‰ (Fig. 1; Supplementary Data 2). Because the vein minerals (particularly the NH4+-bearing minerals, e.g., albite, epidote, and chlorite) were precipitated at relatively high temperatures (>200 ⁰C) that do not favor microbial activity, N contribution from microbial biomass to the samples can be excluded. The N concentrations of U1502B veins show good correlations with not only the modes of N-bearing secondary silicate minerals (Fig. 2), but also the concentrations of alkali elements Na and K and trace elements Rb and Cs (Fig. 2, Supplementary Fig. 5). These correlations verify that the N in the vein samples exists mainly in the form of NH4+ that substitutes Na+ and/or K+ in the vein minerals. The signatures of very low K2O concentrations (0.06 ± 0.03 wt.%; 1 SD; Supplementary Data 2) and high N/K molar ratios (0.51 ± 0.25; 1 SD), but high Na2O concentrations (0.55 ± 0.58 wt.%; 1 SD) and low Na/K molar ratios (0.05 ± 0.03; 1 SD) of the vein samples, which are similar to those of high-T altered oceanic crust (e.g., sheeted dikes and gabbros) in global oceans, further suggest that the secondary NH4+ in the U1502B veins is mainly hosted in the Na+ site rather than the K+ site30,40. Using N/Na molar ratio to remove the modal heterogeneity of NH4+-bearing minerals across samples, the U1502B vein samples still have higher N/Na molar ratios than not only their hosting altered basalts (Fig. 1) but also the high-T altered gabbroic sections in global oceanic basement (N/Na = 0.0006–0.0022)30. This difference suggests that the NH4+ concentration was higher in the focused-flow hydrothermal fluids that formed the U1502B veins than in the diffusive fluids that dominantly altered the gabbros and basalts in global oceanic basements30,40.

Fig. 2: Comparison of nitrogen concentrations with δ15N values and the fractions of nitrogen-bearing secondary silicate minerals in altered basalts and hydrothermal veins.

Data of global altered basalts in panel (a) are from Li and Li29 and reference therein. Nitrogen-bearing secondary minerals in panel (b) include albite, chlorite, augite, and/or epidote.

Nitrogen sources

It is interesting to observe that the veins display a persistent uphole increase in δ15N from –20.3‰ to –3.8‰ (Fig. 1) and a decrease in N/Na ratio from 0.084 to 0.018. Because the mineral assemblage and formation T of the veins did not change significantly throughout the core36,37,38, these gradual changes in N/Na ratio and δ15N value cannot be explained by temperature effect on elemental partitioning34 and N isotope fractionation35. Neither these changes can be explained by low-T organic contamination which would otherwise give higher N/Na ratios for shallower samples. These extremely low δ15N values can neither be attributed to a microbial source because the δ15N values of living microbial biomass in hydrothermal systems mostly fall in a range of –4‰ to +7‰41. Instead, these uphole changes are best explained by a two-component mixing model (Fig. 3a; “Methods”). The shallow component is characterized by low N/Na ratios (or low NH4+ concentrations) and positive δ15N values, consistent with a seawater-dominated fluid source containing 15N-enriched surface NH4+. The deep component is characterized by high N/Na ratios (or high NH4+ concentrations) and extreme 15N depletions (–12‰ to –21‰). Taking the isotope fractionation factor between mineral (e.g., albite) and aqueous NH4+ (~2‰ to 3‰ in the range of 200–300 °C)35 into consideration, the δ15N value of NH4+ in the deep fluids is expected to vary from <–14‰ to –23‰.

Fig. 3: Nitrogen isotope modeling.

a, b Two-endmember mixing model for hydrothermal veins and altered basalts, respectively. c The isotopic values of potential endmember NH4+ from partial reduction of N2 (red curves) or NO3− (blue curves) based on Rayleigh distillation model. In panel a: the mixing curves along the upper and lower data boundaries of vein sample data converge at a confined shallow endmember with positive δ15N values and low N/Na ratios, which are consistent with seawater. The lower endmember displays variable δ15N values from –12‰ to –21‰, likely due to the variable extent of ANR (see text). In panel b: reference mixing curves illustrate addition of secondary NH4+ into fresh basalts from hydrothermal fluids with δ15N values of +6‰ (a typical seawater value), –12‰, and –21‰. Ticks on the mixing curve mark the proportions of secondary N in total N. Note the U1502B data sit far away from the fresh basalt endmember and suggest that >99% of the N originated from secondary sources involving both seawater and 15N-depleted hydrothermal fluids. For comparison, DSDP/ODP/IODP basalts from global oceans are also plotted, with each point and associated error bars representing the average and 1 SD of the data from an individual site (see ref. 29 for the full dataset). In panel (c): F refers to the fraction of remaining N after the reaction, from no reduction (F = 1) to complete reduction (F = 0). See text for the determination of the isotope enrichment factors (ε) for the NH4+ − NO3- pair and the NH4+ − N2 pair. The results illustrate that, depending on the extent of the reaction, the δ15N value of the accumulated NH4+ product may increase from extremely negative values at low degree of N2 reduction to close to its source value at high degrees of reduction.

Known N reservoirs that possibly contributed to the deep fluids, i.e., the mantle source of MORB (~ –5‰ for both N-MORB and E-MORB)24,25, dissolved atmospheric N2 (~0‰) and NO3− (+3‰ to +8‰)21 in seawater, and marine organic matter/sediments including dissolved organic N and NH4+ in interstitial water (mostly in the range of +2‰ to +10‰22, up to 17‰23), are all much more 15N-enriched. Thus, the extremely low δ15N values of NH4+ in the deep fluids have to be attributed to isotope fractionation associated with abiotic reactions in the deep hydrothermal system. In light of the N isotope fractionation factors determined by laboratory experiments17,18 and theoretical calculations35,42, only two abiotic processes can produce remarkably 15N-depleted NH4+ in submarine hydrothermal environments. One is partial reduction of NO3−. However, given that NO3− in bottom seawater has a δ15N value of about +5 ± 2‰21, the lowest δ15N value of NH4+ product can be –10‰ (at 200 °C) to –6‰ (at 300 °C)18,35 (Fig. 3c), which are much less negative than the observed δ15N values of NH4+ in U1502B deep fluids. The other is partial reduction of N2. Although NH4+ is more enriched in 15N than N2 at isotopic equilibration42, experimental studies reveal that N2 is extremely difficult to reach isotope equilibration with other N species due to its strong triple bond as an energy barrier for isotope exchange even at high temperatures of 300 – 800 °C43,44. As a result, abiotic reactions involving N2 generally produce large kinetic isotope fractionations, e.g., –18‰ to –16‰ at 600–800 °C during NH3 decomposition43, a reverse process of ANR. From these values and the equilibrium isotope fractionation factors42, kinetic isotope enrichment factors of –13‰ to –16‰ can be deduced for ANR at 600–800 °C (“Methods”). Accordingly, if the initial N2 is from the upper mantle source (–5‰), ANR-produced NH4+ at 600–800 °C can have δ15N values as low as –21‰ (Fig. 3c). If ANR occurs at lower temperatures, which is highly likely for a submarine hydrothermal system, the magnitude of kinetic isotope fractionation should be larger (although not quantitatively constrained)43 and thus result in more negative δ15N values in the NH4+ product (Fig. 3c). N2 is commonly exsolved during mantle upwelling and partial melting beneath mid-ocean ridges24,25. The N2 outflux at global mid-ocean ridges (1.6 × 1010 mol⋅yr–1)45 is at the same order of magnitude with the N2 outflux at global arc volcanoes (2.0−6.7 × 1010 mol⋅yr–1)46, and thus can provide a sustainable source for ANR. Therefore, the extremely 15N-depleted NH4+ in the deep fluids, as recorded by the vein minerals, can be best explained by abiotic reduction of mantle N2 in the deep oceanic basement through the interaction between deep fluids and Fe2+-bearing minerals (Fig. 4). The relatively large δ15N range of the deep-fluid endmember in Fig. 3a, b may be attributed to different extents of ANR (Fig. 3c) and/or heterogeneous mixing with surface NH4+.

Fig. 4: Schematic diagram (not to scale) showing abiotic N2 reduction in deep fluids, mixing of seawater into deep fluids, and alteration of oceanic crust in mid-ocean ridge hydrothermal systems.

a-NH4+ denotes the NH4+ produced from abiotic N2 reduction by Fe2+-bearing minerals; b- NH4+ denotes biogenic NH4+ in seawater. Along the circulation pathway of deep seawater, b- NH4+ was progressively consumed via NH4+ assimilation into secondary silicate minerals formed during seawater-rock interaction. Thus, the impact of b-NH4+ on the N signature of hydrothermal fluids was less in depths but more and more prominent toward the seafloor.

Given that the hydrothermal system below U1502B was sustained by heat flux from intrusive mafic dykes38, the most likely Fe2+-bearing minerals for ANR are pyroxene (ferrosilite as the Fe2+-rich endmember) and possibly olivine (fayalite as the Fe2+-rich endmember) in these rocks. The ANR can be expressed by Reactions (1) − (2) for ferrosilite or Reactions (3) − (4) for fayalite22:

$${9{{{\rm{FeSiO}}}}}_{3}+{{{\rm{N}}}}_{2}+{3{{{\rm{H}}}}}_{2}{{{\rm{O}}}}+{2{{{\rm{H}}}}}^{{+},\to ,{3{{{\rm{Fe}}}}_{3}{{{\rm{O}}}}}_{4}+9{{{\rm{SiO}}}}_{2}+{2{{{\rm{NH}}}}}_{4}}{+}$$

(1)

$${6{{{\rm{FeSiO}}}}}_{3}+{{{\rm{N}}}}_{2}+{3{{{\rm{H}}}}}_{2}{{{\rm{O}}}}+{2{{{\rm{H}}}}}^{{+},\to ,{3{{{\rm{Fe}}}}_{2}{{{\rm{O}}}}}_{3}+6{{{\rm{SiO}}}}_{2}+{2{{{\rm{NH}}}}}_{4}}{+}$$

(2)

$$9{{{{\rm{Fe}}}}}_{2}{{{{\rm{SiO}}}}}_{4}+{2{{{\rm{N}}}}}_{2}+{6{{{\rm{H}}}}}_{2}{{{\rm{O}}}}+{4{{{\rm{H}}}}}^{{+}\to ,{6{{{\rm{Fe}}}}}_{3}{{{{\rm{O}}}}}_{4}+{9{{{\rm{SiO}}}}}_{2}+{4{{{\rm{NH}}}}}_{4}}{+}$$

(3)

$${3{{{\rm{Fe}}}}}_{2}{{{{\rm{SiO}}}}}_{4}+{{{{\rm{N}}}}}_{2}+{3{{{\rm{H}}}}}_{2}{{{\rm{O}}}}+{2{{{\rm{H}}}}}^{{+}\to {3{{{\rm{Fe}}}}}_{2}{{{{\rm{O}}}}}_{3}+{3{{{\rm{SiO}}}}}_{2}+{2{{{\rm{NH}}}}}_{4}}{+}$$

(4)

The extreme 15N depletion in the vein samples progressively diminishes toward the surface, associated with a decrease in N/Na ratio (Fig. 1). This indicates the ANR-produced NH4+ in the deep fluid below U1502B has been progressively overprinted by surface NH4+ following enhanced mixing with seawater-derived fluids towards seafloor (Fig. 4). This mixing effect can also explain why the ANR signal is rarely recorded in global altered oceanic basalts, because the alteration fluids in the highly porous upper section of oceanic crust are dominated by relatively less evolved seawater that contains sediment-derived NH4+29. In contrast, in more evolved deep fluids, 15N-enriched surface NH4+ from seawater-dominated fluid could have been stripped away by oceanic basement rocks along the pathway that the shallow fluid was pumped down. Thus, the deep fluid can preserve the 15N-depleted signature of ANR, as manifested by the minerals deposited from the focused flow channel of these deep fluids. This mixing discrepancy is clearly shown by the difference in N concentration and isotope composition between veins and their direct wall-rock basalts in Hole U1502B (Figs. 1–2).

Nevertheless, the δ15N values of U1502B altered basalts are clearly lower than the range of surface NH4+ (+2‰ to +10‰) and extend towards the negative deep fluid values (Fig. 3b). This implies that the ANR signal could still have been partially implanted into altered basalts, but significantly weakened due to surface NH4+ overprinting. Significant 15N depletion (as low as –20‰) has also been observed in the hydrothermally altered oceanic basalts from the western Pacific (ODP Sites 1149 and 801 with ages of 130–170 Ma and fast half-spreading rates of 51 − 80 mm·yr–1)27 and the mid-Atlantic (DSDP Site 417 with an age of 120 Ma and a slow half-spreading rate of ~12.5 mm·yr–1)29, as well as their blueschist- to eclogite-facies metamorphic equivalents22,47. This may imply that ANR has occurred more commonly in worldwide submarine hydrothermal systems.

Implications to Earth’s N cycling

To assess the ANR contribution to the seawater NH4+ reservoir, we employed a commonly used model (“Methods”) to estimate the NH4+ concentration of the deep fluid, which gave a ballpark range of 12.5–15.0 mM. These values are higher than the few available NH4+ concentration data of high-T (300–380 °C) hydrothermal fluids without sedimentary N contribution (e.g., <0.01 mM; Supplementary Data 4) but close to the NH4+ concentrations of the high-T (300–315 °C) hydrothermal fluids in the Guaymas Basin (12.9–15.2 mM), which were speculated to originate from thermal degradation of organic matter[48](https://www.nature.com/articles/s41467-025-65711-1#ref-CR48 “Von Damm,