Stable Abiotic Production of Ammonia from Nitrate in Komatiite-Hosted Hydrothermal Systems in the Hadean and Archean Oceans

: Abiotic ﬁxation of atmospheric dinitrogen to ammonia is important in prebiotic chemistry and biological evolution in the Hadean and Archean oceans. Though it is widely accepted that nitrate (NO 3 − ) was generated in the early atmospheres, the stable pathways of ammonia production from nitrate deposited in the early oceans remain unknown. This paper reports results of the ﬁrst experiments simulating high-temperature, high-pressure reactions between nitrate and komatiite to ﬁnd probable chemical pathways to deliver ammonia to the vent–ocean interface of komatiite-hosted hydrothermal systems and the global ocean on geological timescales. The ﬂuid chemistry and mineralogy of the komatiite–H 2 O–NO 3 − system show iron-mediated production of ammonia from nitrate with yields of 10% at 250 ◦ C and 350 ◦ C, 500 bars. The komatiite–H 2 O–NO 3– system also generated H 2 -rich and alkaline ﬂuids, well-known prerequisites for prebiotic and primordial metabolisms, at lower temperatures than the komatiite–H 2 O–CO 2 system. We estimate the ammonia ﬂux from the komatiite-hosted systems to be 10 5 –10 10 mol/y in the early oceans. If the nitrate concentration in the early oceans was greater than 10 µ mol/kg, the long-term production of ammonia through thermochemical nitrate reduction for the ﬁrst billion years might have allowed the subsequent development of an early biosphere in the global surface ocean. Our results imply that komatiite-hosted systems might have impacted not only H 2 -based chemosynthetic ecosystems at the vent-ocean interface but also photosynthetic ecosystems on the early Earth. This reports the ﬁrst experimental results of a simulation of high-temperature, high-pressure reactions between nitrate and komatiite to ﬁnd chemical pathways that may have delivered ammonia to the vent–ocean interface and the global ocean on geological timescales. Komatiite is an ultramaﬁc lava (MgO produced by exceptionally hot melting in mantle Ga


Introduction
The reduction of abundant dinitrogen (N 2 ) to ammonia (NH 3 ) is important for prebiotic chemical evolution and the early development of life on Earth. Ammonia participates in the synthesis of nitrogen-bearing organic compounds that are essential for cellular functions and self-reproduction (e.g., amino acids and nucleotides). Thus, the production of ammonia is assumed to be a prerequisite for origin-of-life scenarios, e.g., [1,2], and the development of microbial ecosystems in the Hadean and Archean oceans, e.g., [3][4][5]. Theoretical and observational assessments indicate that the main components of the early atmosphere are N 2 and CO 2 , e.g., [6][7][8][9][10][11]. Thus, high energy reactions are required to reduce chemically stable N 2 . Previous studies have suggested that nitric oxide (NO) was produced by electrical discharges in the early atmosphere [3,12]. The resulting NO was likely converted into nitric acid (HNO 3 ) or nitrous acid (HNO 2 ) due to photochemical and aqueous-phase reactions, and it was eventually deposited in the ocean through rain [13]. This suggests ammonia was produced by the abiotic reduction of nitrate (NO 3 − ) and nitrite (NO 2 − ) in the early oceans because the oldest traces of early life were reported from 3.9-3.8 Ga sedimentary rocks of marine origin, e.g., [14][15][16][17][18][19][20], c.f. [21][22][23].
Iron-bearing materials are potential reductants for ammonia production from nitrate and nitrite in the early anoxic ocean. Accordingly, chemical reactions of nitrate and nitrite in the water column or marine sediments have been tested experimentally using ferrous ions and sulfate green rust (Fe II 4 Fe III 2 (OH) 12 SO 4 ·yH 2 O) at around 25 • C [24][25][26][27]. These experiments have provided the following three findings. (1) At 22 • C, Nitrite reduction by ferrous ions produces ammonia at pH ≥ 7.5, but ammonia production is suppressed in CO 2 -saturated water with respect to FeCO 3 . (2) Nitrate reduction by ferrous ions produces ammonia at pH > 8 in the presence of Cu 2+ . (3) Sulfate green rust reduces nitrite and nitrate to ammonia at pH 7-8. 3. These findings suggest that the reduction of nitrite and nitrate at around 25 • C depends highly on the composition of the aqueous solution and hardly occurs at acidic pH. Because the early oceans are thought to be CO 2 -rich and sulfate-poor, with an acidic to slightly alkaline pH (ca. 5.5-7.5) [8,9,11,[28][29][30], the reduction of nitrite and nitrate may have been inhibited in the water column or marine sediments in these oceans at around 25 • C.
Seafloor hydrothermal environments are another potential site of pre-biotic ammonia production. Abiotic reduction of nitrate and nitrite has been examined under simulated hydrothermal conditions with transition metal sulfides (Fe 1−x S, FeS 2 , NiS, Cu 2 S, CuFeS 2 ), oxides (Fe 3 O 4 , NiO, Cu 2 O), and iron-nickel alloys [31][32][33][34]. The results of these experiments have shown that some minerals participate in the reduction of nitrate and/or nitrite to ammonia, while others participate in the reduction of nitrate and/or nitrite to other forms of nitrogen (putatively N 2 or N 2 O). Major anions (e.g., chloride, phosphate, sulfate, and dissolved carbonate species) further suppress nitrate reduction by Fe 1−x S at room temperature due to competing adsorption onto Fe 1−x S [31]. Given that concentrations of transition metal sulfides are typically found in hydrothermal mounds where high-temperature fluids mix with cold seawater, the reactions examined in previous studies [31][32][33][34] occur in nature under fluctuations and steep gradients of temperature, pH, and redox potential. These fluctuations and the associated unstable environment in hydrothermal mounds should result in a lower conversion to ammonia than in the laboratory experiments, with the exception of the relatively robust reduction of nitrite to ammonia catalyzed by MoS with geoelectrical current [35]. These results imply that other mechanisms for the sustainable supply of ammonia to the early oceans are favorable both for the pre-biotic chemical evolution in hydrothermal vent environments and the accumulation of ammonia in the global ocean.
This study focused on the reaction of nitrate with oceanic crust under high temperature and high-pressure conditions. Such reactions probably occur in the recharge and reaction zones of hydrothermal systems (i.e., sub-seafloor environments) and have affected the composition of vent fluids and the budget of nitrogen species in the early oceans. At present, the ammonia concentration in high-temperature hydrothermal fluids in sedimentfree systems is typically similar to that of ambient deep-sea water (1 µm or less), although occasionally it is as high as 15 µm at certain locations [36,37]. These observations suggest the possibility of abiotic reduction of nitrate in seawater (40 µm) within oceanic crust. However, the simultaneous complex processing of nitrogen compounds by thermophilic microorganisms within hot oceanic crusts can obscure the signal of abiotic nitrate chemistry, as observed for nitrogen dynamics in geothermal environments [38]. Thus, an experimental approach is required to understand the abiotic chemistry of nitrate within hot oceanic crusts and the resultant vent fluid composition. This paper reports the first experimental results of a simulation of high-temperature, high-pressure reactions between nitrate and komatiite to find chemical pathways that may have delivered ammonia to the vent-ocean interface and the global ocean on geological timescales. Komatiite is an ultramafic lava (MgO > 18%) produced by exceptionally hot melting in mantle upwellings, e.g., [39]. Most komatiite eruptions occurred between 3.5 and 1.5 Ga [40], and komatiite is believed to be the predominant ultramafic rock on the Hadean-Archean seafloor. The young Earth was hotter, and a thick layer of oceanic crust (more than three times greater than the present-day equivalents) probably restricted the eruption of mantle peridotites to the seafloor [41,42]. Therefore, rather than peridotitehosted hydrothermal systems at slow-spreading mid-oceanic ridges, komatiite-hosted hydrothermal systems on oceanic islands and/or plateaus has been hypothesized to generate H 2 -rich and alkaline fluids [43], which are prerequisites for pre-biotic and primordial metabolisms at the vent-ocean interface, e.g., [43][44][45][46]. On the basis of the findings of the present study, the roles of komatiite-hosted hydrothermal systems on pre-biotic chemistry and implications for biological nitrogen metabolism and the global nitrogen cycle on the early Earth were explored.

Materials and Methods
Synthesis of komatiite, hydrothermal reaction experiments and chemical/mineralogical analyses of the reactants and products were conducted at Japan Agency for Marine-Earth Science and Technology (JAMSTEC). Details were provided in previous publications [47][48][49][50][51][52].

Starting Materials
Komatiite used in the experiments was synthesized from a mixture of 12 reagents (i.e., SiO 2 , TiO 2 , Al 2 O 3 , Fe 2 O 3 , MnO, MgO, CaCO 3 , Na 2 CO 3 , K 2 CO 3 , P 2 O 5 , NiO, and Cr 2 O 3 ) by stepwise heating and cooling at oxygen fugacity of the quartz-fayalite-magnetite buffer. Mixing ratios of the reagents were adjusted to make Al-depleted (Barberton-type) komatiite more representative of ultramafic volcanism in Hadean oceanic islands and/or plateaus with a hotter mantle upwelling (plume) compared with Al-undepleted (Munrotype) komatiite [49]. The synthetic komatiite was primarily composed of olivine and glass. The komatiite was crushed using a tungsten mortar and an agate mill and then sieved to obtain particles smaller than 90 µm. Approximately 30 g of the powdered komatiite was ultrasonically washed with acetone and pure water (18.4 MΩ cm at 25 • C) several times and freeze-dried in a vacuum to remove possible contamination by organic matter during sample preparation. Specifically, carbon and nitrogen contents of the resultant komatiite sample were 210 ppm and 0.5 ppm, respectively [52].
Simulated aqueous solutions of Hadean seawater containing nitrate were prepared with the following compositions: 1000 mmol/kg NaCl and 10 mmol/kg NaNO 3 for the experiment at 250 • C; and 1000 mmol/kg NaCl and 0.94 mmol/kg KNO 3 for the experiment at 350 • C. Prior to use, the NaCl reagent was baked at 450 • C for 5 h under atmosphere to combust any organic contaminants.
We also conducted hydrothermal experiments without komatiite to test nitrate reduction in the H 2 O-NO 3 − system at 250 • C and 350 • C under 500 bars (experiments N250C and N350C). The initial aqueous solutions were prepared with the following composition: 1000 mmol/kg NaCl and 10 mmol/kg KNO 3 .

Hydrothermal Experiments
The hydrothermal reaction experiments were conducted in a flexible-cell hydrothermal apparatus consisting of a flexible gold bag with a titanium head (reaction cell) enclosed in an Inconel ® alloy autoclave (modified after [53]). Prior to the experiments, the titanium head attached to the gold bag was completely oxidized to TiO 2 to eliminate possible H 2 generation by a reaction between metallic titanium and water at the experimental conditions. All materials in contact with the reacting fluid (i.e., gold bag, titanium head, and gold-lined sampling tube) were baked at 450 • C for 5 h under atmosphere in a muffle furnace to remove any organic contamination.
Approximately 10-12 g of the komatiite powder was initially reacted with 64-72 g of the simulated Hadean seawater in the reaction cell at 250 • C and 350 • C, under 500 bars (experiments N250 and N350). Changes in fluid composition were monitored by collecting 3-5 g of fluid samples multiple times during the ongoing experiment from the reaction cell through the gold-lined sampling tube while maintaining the temperature at 250 • C or 350 • C and the pressure at 500 bars. Consequently, the water-rock mass ratios decreased from 7 to 4 during experiment N250 (4870 h) and from 5 to 3 during experiment N350 (2515 h). These experimental conditions mimic the high-temperature regions in sub-seafloor hydrothermal systems where the water-rock mass ratios are generally below 5 [54]. After the experiments, solid reaction products were freeze-dried in a vacuum and then stored in a vacuum desiccator until mineral compositions could be analyzed.

Aqueous Fluids
For pH analysis, 1.0 mL of fluid was collected in a vial and the pH was measured by a pH meter (LAQUAtwin; HORIBA Ltd., Kyoto, Japan) at a room temperature under atmospheric conditions. The estimated precision of pH measurement was ±0.2 units of the reported value. For H 2 analysis, 0.5 mL of fluid was introduced to an Ar-purged vial. The vial was left for 30 min to be reached in equilibrium at a room temperature. The H 2 and N 2 O concentrations in the headspace of the vial was measured by gas chromatography with dielectric-barrier discharge ionization detector (GC-BID) (Nexis GC-2030, SHIMADZU Corp., Kyoto, Japan). The analytical precision (1SD) of replicate measurements was typically better than 5% for H 2 and N 2 O at 100 ppm.
For cation analyses, 0.2 mL of fluid was sampled in a vial and diluted with 10 mL of pure water. The final pH of the diluted solution was adjusted to 2 by adding 0.2 mL of 1N HNO 3 to avoid mineral precipitation. For anion analyses, 0.2 mL of fluid was collected and diluted with 10 mL of pure water. The final pH of the diluted solution was adjusted to 12 by adding 0.2 mL of 1N NaOH was added to avoid the precipitation of amorphous silica. The concentrations of Cl, and Na were analyzed by ion chromatography (Dionex ICS-1600/2100; Thermo Fisher Scientific, Waltham, MA, USA), while those of K, Ca, Mg, Fe, Mn, and Si were analyzed by inductively coupled plasma optical emission spectrometry (ICP-OES) (Spectro ARCOS; AMETEK, Berwyn, PA, USA). The analytical precision (1SD) of replicate measurements was typically better than 2% for Cl, Na, and K, and typically better than 5-10% for the other elements depending on their concentrations.
The nitrate and nitrite concentrations were measured using ion chromatography (IC-Pac A25S column; Thermo Fisher Scientific, Waltham, MA, USA) with UV detection at 220 nm (GL-7451; GL Science, Tokyo, Japan) (the detection limit was 0.1 µmol/kg and the reproducibility was better than 3%) [51]. The ammonium concentration in the reacted fluid was determined with fluorescence photometry after reacting at 50 • C with a 20 mmol/kg orthophthaldialdehyde solution and a 200 mmol/kg sodium borate solution containing 2 mmol/kg of sodium sulfite [52]. The analytical precision (1SD) of replicate measurements was typically better than 1%, 4%, and 3% for nitrate, nitrite and ammonium, respectively.

Solids
Initial characterization of the morphology and chemical composition of the solid reaction products was performed after the experiments using scanning electron microscopes (SEM) equipped with energy dispersive X-ray spectrometers (EDX) (a Hitachi High-Tech Miniscope ® TM3000 and a FEI Quanta 450 FEG). Data on the mineral assemblages of reaction products were obtained by X-ray diffraction (XRD) and Raman spectroscopy using a Rigaku MiniFlex II and a Nanophoton RAMANtouch. Quantitative compositional data of the reaction products were obtained through electron probe micro analysis (EPMA) with wavelength dispersive X-ray spectroscopy using a JEOL JXA-8500F. The whole-rock compositions of 10 major elements (i.e., Si, Ti, Al, Fe, Mn, Mg, Ca, Na, K, and P) of the synthetic komatiite were analyzed by X-ray Fluorescence (XRF) using a Rigaku ZSX Primus II. Three types of samples were prepared from the solid reaction products for the morphological, mineralogical and compositional analyses. The first group comprises fragments mounted in epoxy resin and polished by silicon carbide powder and diamond paste to determine the cross sections of the solid products. Samples in the second group were crushed using a stainless spatula, and the resulting sub-millimeter size grains were mounted on a glass slide with carbon tape or double-sided tape to examine 3D structures. Half of the second group samples were washed further with ethanol and purified water in an ultrasonic cleaner to remove the suspended fraction. The procedure was repeated until the ethanol and water were no longer cloudy. The washed samples were dried in a vacuum at 40 • C and then mounted on a glass slide. After examination using optical microscope and by Raman spectroscopy, the polished and crushed samples were coated with osmium or carbon particles for the analyses by SEM/EDX or EPMA, respectively. The third group of fragment samples was ground with agar mortar for XRD and XRF analyses. The resulting powder samples were processed further for XRF analysis to prepare glass beads using Li 2 B 4 O 7 reagent.

Thermodynamic Calculation
The pH in-situ of the sampled fluids was calculated using the Geochemists' Workbench computer code [55], based on pH at room temperature and the concentrations of dissolved species. Saturation index (SI) was calculated from the relation SI = log(Q/K), where Q is the activity quotient of the species involved in the reaction and K is the equilibrium constant. The required thermodynamic database was generated using the SUPCRT92 computer program [56] with thermodynamic data for aqueous species and complexes obtained from [57][58][59][60]. The B-dot activity model was used in the calculation of sampled fluids [61,62], and the activity coefficients of neutral species were assumed to be unity. Table 1 and Figure 1 present the concentrations of selected species measured for aqueous fluids in the komatiite-H 2 O-NO 3 − system as a function of time at 250 • C and 350 • C under 500 bars. The pH values of the initial solutions of experiments N250 and N350 after the pressurization to 500 bars at room temperature were approximately 9. During the high temperature water-rock reactions, the pH measured at room temperature (pH 25 • C ) decreased from the initial values of 9.6 and 7.2 to the steady-state values of 8.6 (alkaline) and 5.7 (acidic) at 250 • C and 350 • C, respectively. "Steady state" here refers to the condition under which the concentration of a given dissolved species changes little with time [49,50].  Table 1).  Nitrate concentrations decreased from 11 mmol/kg to approximately 0.01 mmol/kg after 264 h at 250 • C, and decreased from 1 mmol/kg to approximately 0.005 mmol/kg after 71 h at 350 • C. Over the same time interval, ammonium concentration increased from 0.02 mmol/kg to 1.14 mmol/kg at 250 • C and 0.08 mmol/kg to 0.18 mmol/kg at 350 • C. Nitrite concentration increased from less than 0.002 mmol/kg to 0.011 mmol/kg in the first 18 h and then decreased to less than 0.002 mmol/kg at 250 • C, whereas it was less than 0.002 mmol/kg throughout the experiment at 350 • C.

Fluid Chemistry
The time-dependent changes of nitrogen compounds suggest that a multi-step pathway exists for the reduction of nitrate to ammonia via nitrite at 250 • C. This is comparable to the biological process referred to as dissimilatory nitrate reduction to ammonia. The mol ratio of ammonia produced to nitrate reduced (i.e., the product yield) was estimated to be 10% at both 250 • C and 350 • C. Thus, the other 90% of nitrate initially present in the fluid is expected to be reduced to other nitrogen species, most likely N 2 via nitrous oxide (Section 4.2). The decrease in nitrate concentration appeared to be coupled with the increase in calcium concentration and with the decrease in pH in-situ at both 250 • C and 350 • C.
By comparison, experiments without komatiite show less reduction of nitrate at both 250 • C and 350 • C. In these experiments, nitrate concentrations were constant at 10 mmol/kg within the range of analytical uncertainty (±2%) and decreased slightly from 10 mmol/kg to 9 mmol/kg within 860 h and 264 h under 500 bars at 250 • C and 350 • C, respectively (Table 2, Figure 2). Over the same time intervals, nitrite concentration increased to 1.1 mmol/kg at 350 • C, whereas it was below detection limit at 250 • C. Nitrous oxide concentrations increased up to 0.01 mmol/kg and 0.05 mmol/kg at 250 • C and 350 • C, respectively. Ammonium concentrations were below the detection limit at both 250 • C and 350 • C. This indicates nitrate reduction to nitrite in the H 2 O-NO 3 system at 350 • C. However, the rate of nitrate reduction is more than 40 times slower than that in the komatiite-H 2 O-NO 3 system at 350 • C. Hence, komatiite causes or stimulates nitrate reduction and ammonia production in aqueous fluids at temperature above 250 • C and pressure of 500 bars.  The dissolved H 2 concentration increased to steady-state values of 2.3 and 1.7 mmol/kg at 250 • C and 350 • C, respectively. By comparison, steady-state concentrations of the dissolved H 2 concentration in the komatiite-H 2 O-CO 2 system were reported to be 0.02 mmol/kg and 2 mmol/kg at 250 • C and 350 • C, respectively ( Figure 3) [50]. The steady-state concentration of dissolved H 2 in the komatiite-H 2 O system was reported to be 23 mmol/kg at 300 • C [49]. Table 3 compares the steady-state composition of fluids in the komatiite-H 2 O-NO 3 − system with that in the komatiite-H 2 O-CO 2 system, as reported by [50]. Besides H 2 , the remarkable differences between the aqueous fluids in the two systems are the steady-state concentrations of Mg, and Fe and fluid pH at 250 • C. The dissolved Mg and Fe concentrations in the komatiite-H 2 O-NO 3 − system are 2-3 orders of magnitude lower than those in the komatiite-H 2 O-CO 2 system. The pH measured at room temperature in the komatiite-H 2 O-NO 3 − system is more alkaline than that in the komatiite-H 2 O-CO 2 system.  This study This study [50]. [50] Concentrations are in mmol/kg [m]. n.m.-not measured. a Water-rock mass ratio. b Initial aqueous solutions were prepared with the following chemical composition: 1000 mmol/kg Cl, 1000 mmol/kg Na, and 400 mmol/kg ΣCO 2 . The pH values of the initial solutions were~4.9 at room temperature.

Alteration Products
The characterization of the solid products of the komatiite alteration experiments at 250 • C and 350 • C by X-ray diffraction, Raman spectroscopy, EPMA and SEM/EDX indicates that the products were dominated by a mixture of smectite, chlorite, and serpentine, and/or mixed layer clays such as chlorite/smectite and chlorite/serpentine (Figures 4 and 6 and Tables 4 and 5). Andradite and clinopyroxene were observed as characteristic major components at 250 • C and 350 • C, respectively. The reaction products exhibited the following characteristics: 1.
The reaction products at 250 • C and 350 • C were mainly composed of 10-50-µm-size clusters of mineral plates/flakes. The plates/flakes developed an edge-to-face pattern and/or a rosette-like pattern (Figures 5e and 7d,g). Such patterns are observed for the plates/flakes of chlorite, smectite and corrensite (mixed layered chlorite-smectite mineral) coated on a variety of silicate minerals in natural and experimental systems, e.g., [63][64][65]. Raman spectra of the clusters separated as the yellowish green or greenish grains from the bulk products showed peaks attributable to smectite, chlorite, serpentine and vermiculite (mixed layered chlorite-smectite mineral) (Figure 6h-1,h-3).
The clusters also showed intermediate compositions between smectite (saponite), chlorite (clinochlore), and serpentine (Tables 4 and 5). Collectively, we interpreted that the mineralogy of the clusters was a mixture of smectite, chlorite, and serpentine, and/or mixed layer clays such as chlorite/smectite and chlorite/serpentine (Table 4).

2.
The clusters at both 250 • C and 350 • C had overgrowth rims extending outward from the center (Figure 5a,b and Figure 7a,b). The rims were composed of slightly curved flakes that developed honeycomb patterns with pore of 1-2 µm and 2-3 µm in size at 250 • C and 350 • C, respectively (Figures 5c and 7b). Moreover, overgrowth rims with frost column textures partly covered the surface of the cluster at 250 • C (Figure 5e). There were not marked differences in elemental ratios between the rims and the phyllosilicate clusters at both 250 • C and 350 • C, suggesting that the rims had the same mineral composition as the clusters (Table 5).

3.
Spherical (polygonal) minerals (ca. 1-3 µm in size) were found in the reaction products at 250 • C (Figures 5e and 7a,b,f). They existed both in the cluster and the overgrowth rims. These minerals were identified as andradite (Ca 3 Fe III 2 Si 3 O 12 ) based on their Raman spectra (Figure 6h-2). Results of the semi-quantitative analysis by SEM/EDX of the spherical minerals indicate elemental ratios that were consistent with those of andradite (Table 5). 4.
Needle-shaped minerals occurred with the overgrowth rims in the reaction products at 350 • C (Figure 7a,b,f,h). The needles stuck together and formed an irregularshaped, opaque grain that was separated as an unsuspended fraction after sonication (Figure 6d,g and Figure 7f). Raman spectra of these opaque grains indicated that the needles were clinopyroxene (Figure 6h-4). Semi-quantitative analysis by SEM/EDX of the needles showed elemental ratios that were consistent with clinopyroxene (Table 5).

5.
Fibrous minerals were observed in the reaction products at 350 • C (Figure 7c,d).
Regarding their morphology, they could be chrysotile fibers. 6.
Needle-shaped calcium silicate occasionally occurred with andradite in the reaction products at 250 • C (Figure 5h). SEM/EDX analysis shows Ca/Si over 1 with Mg + Fe/Si smaller than 0.5 (Table 5). Clinopyroxene was also found among the reaction products at 250 • C as a minor component. Abbreviations are the same as in Table A1.  Table A1. show peaks attributable to smectite, chlorite, actinolite, and serpentine that have nearly the same wavenumber. Raman spectra of the small spherical grains of N250 (number 2) show peaks attributable to andradite. Raman spectra of the random-shaped opaque grains of N350 (number 4) show the peaks attributable to clinopyroxene. A large peak between 650 cm -1 and 700 cm -1 is common in inosilicate and phyllosilicate candidates. Abbreviations in Figure 6h are the same as in Table A1.  Table A1.  Based on the petrographic observations described above, clinopyroxene deposited after the formation of smectite-serpentine-chlorite core assemblage at 350 • C.

Water-Rock Interactions in the Komatiite-H 2 O-NO 3 − System
Fluid composition and secondary mineral assemblages provide information on the water-rock interactions that include hydrogen generation and nitrate reduction. To aid the interpretation of reactions occurring during the experiments, saturation indexes of solid materials were calculated based on the measured fluid composition (Table 6). A list of potential water-rock reactions is also shown in Table 7.
In experiment N250, the concentration of dissolved Si (SiO 2(aq) ) increased to the maximum value (6 mmol/kg) during the first 8 h at 250 • C and then, remained at the maximum value from 8 h to 44 h ( Figure A1). The maximum concentration of SiO 2(aq) was 15% of the SiO 2(aq) concentration predicted from the solubility of amorphous silica (Table 6), but it was~100 times higher than the SiO 2(aq) concentration under olivine dissolution at temperatures from 230 • C to 320 • C and apH in-situ of 8 [66]. Moreover, the positive SI values of forsterite and fayalite after the first 44 h (i.e., supersaturation with respect to forsterite) indicates that olivine dissolution had been inhibited in the first 44 h (Table 6). Therefore, the water-rock interactions that mainly occurred from 8 h to 44 h are predicted to be the dissolution of komatiite glass and the precipitation of secondary silicates. Given the low solubility of Al 2 O 3 [67] and the fact that the sampled fluids had low concentrations of Mg and Fe, the precipitation of secondary silicates may have largely occurred through reactions involving solid phases of komatiite glass (rxns. 1-3 in Table 7). This inference is supported by the stable pH in-situ value in the first 44 h because reactions 1 to 3 did not affect pH in-situ much as compared with silicate formation from dissolved species (rxns. [6][7][8].
The time interval between 44 h to 264 h was characterized by the decrease in SiO 2(aq) concentrations from the maximum value to the steady state value and the decrease in pH in-situ from 7.7 to 6.2. The decrease in the SiO 2(aq) concentration suggests the completion of the dissolution of the SiO 2 component in komatiite glass and the precipitation of secondary silicates. In particular, the decrease in pH in-situ requires the precipitation of dissolved Ca as Ca-bearing silicates (saponite, andradite, Ca-pyroxene) (rxns. 1, 4, 5) because the dissolved Ca concentration was high at the milli-molal level, while the dissolved concentrations of Mg and Fe were low over the same time interval. The precipitation of Ca-bearing silicates was consistent with the positive SI values of andradite and diopside (CaMgSi 2 O 6 ) over the same time interval (Table 6). On the other hand, this was apparently inconsistent with the increase in the dissolved Ca concentration from 6 to 36 mmol/kg. This discrepancy can be resolved if the decrease in pH in-situ promoted the dissolution of the CaO component in glass (rxn. 15). Because andradite precipitation involves iron oxidation and H 2 generation (rxn. 4), this reaction probably induced nitrate reduction between 44 h to 264 h (Section 4.2). The oxidation of the FeO component in komatiite glass also possibly contributed to H 2 generation and nitrate reduction between 44 h to 264 h (rxns. [13][14].
The last interval between 264 h to 4871 h was characterized by the decrease in the Ca concentration to the steady state value (30 mol/kg; 266 h to 1054 h) and the increase in the H 2 concentration to 2 mmol/g. The decrease in the Ca concentration was attributed to the formation of andradite because the SI value of andradite was positive until 1054 h, whereas the SI value of diopside became negative after 264 h ( Table 6). This suggests that andradite formation had produced H 2 until 1054 h. On the other hand, dissolution of fayalite, as inferred from the saturation of fluid with respect to fayalite since 1054 h (SI = 0), may have produced H 2 since 1054 h by the oxidation of the FeO component in olivine (FeO ol ) with water to form magnetite and/or Fe III -bearing serpentine (rxns. [13][14] [68].
In experiment N350, the SiO 2(aq) concentration in the first 17 h at 350 • C had the maximum value (20 mmol/kg), that is 63% of the SiO 2(aq) concentration predicted from the solubility of amorphous silica ( Figure A1). Over the same time interval, pH in-situ decreased from 7.1 to 5.8, whereas the Ca concentration increased to 10 mmol/kg. As discussed above, this suggests the dissolution of komatiite glass coupled with the formation of secondary silicates, which includes the precipitation of dissolved Ca as Ca-bearing silicates (saponite, Ca-pyroxene) (rxns. [1][2][3][4][5]. Subsequently, the SiO 2(aq) concentration decreased from the maximum value to the steady state value (1 mmol/kg) from 17 h to 71 h. The decrease in the SiO 2(aq) concentration was accompanied by a decrease in pH in-situ . Thus, this time interval was probably characterized by the complete consumption of the SiO 2 component in komatiite glass and the precipitation of secondary silicates, including Ca-bearing silicates, which is consistent with the positive SI value of diopside at 17 h.
A main feature of the time interval between 71 h to 2515 h was the increase in the H 2 concentration to 3 mmol/kg. The SI values of secondary minerals predicted that H 2 generation during the entire interval of N350 can be driven by the oxidation of dissolved and solid phases of ferrous iron (Fe 2+ (aq) , FeO rock = FeO glass , FeO ol ) to form magnetite and/or Fe III -bearing serpentine (rxns. [13][14]. Because nitrate reduction to ammonia was completed by 71 h, H 2 produced from iron oxidation may have been completely and/or partially consumed to reduce nitrate before 71 h.

Mechanisms of Nitrate Reduction to Ammonia and Denitrification in the
This study strongly suggests that reactions in komatiite cause the complete reduction of nitrate to N 2 and NH 3 with a ratio of N 2(aq) /NH 3(aq) of 4.5 at both 250 • C and 350 • C, under pressures of 500 bars (Figure 8). The thermodynamic equilibrium of dissolved nitrogen species in the N-H-O system was evaluated first as a possible explanation for the N 2(aq) /NH 3(aq) ratio to characterize the mechanism of nitrate reduction in hydrothermal environments. thought to be another end product of abiotic nitrate reduction via nitrous oxide ("chemodenitrification"). Nitrogen species detected by experiments are shown in square. Experiment IDs in which a given nitrogen species was detected are shown. The fact that nitrite remained undetected in N250C and N350 suggests that the rate of nitrite reduction is much faster than that of nitrate reduction to nitrite. n.m.-not measured. Figure 9a,b show the pH in-situ and aH 2(aq) (or aO 2(aq) ) states of water with theoretically predicted stability fields of dissolved nitrogen species in thermodynamic equilibrium at 250 • C and 350 • C, 500 bars, respectively. The stability fields are drawn based on the activity of NH 4 + in the final sampled fluid (aNH 3(aq) = 10 −3.228 and 10 −3.98 at 250 • C and 350 • C, respectively). Bold red lines in the diagrams indicate the pH in-situ and aH 2(aq) values at which the activity ratios of two nitrogen species (NH 3(aq) -NH 4 + , NH 3(aq) -N 2(aq) , NH 4 + -N 2(aq) , and N 2(aq) -NO 3 − ) equal one in thermodynamic equilibrium. The pH in-situ and aH 2(aq) values were plotted for all sampled fluids in the komatiite-H 2 O-NO 3 − and H 2 O-NO 3 − systems in these diagrams to compare the dominant nitrogen species in the experimental products with those from the thermodynamic prediction. Given that the very low H 2 concentrations of sampled fluids in the H 2 O-NO 3 − systems were difficult to measure, the presumed aH 2(aq) values were plotted in the diagrams assuming that sub-milli molal O 2 in the initial solutions was not consumed during the experiments at 250 • C and 350 • C (e.g., dissolved O 2 concentration is 0.2 mmol/kg in an air saturated seawater). In the H 2 O-NO 3 − system, sampled fluids at 250 • C and 350 • C were predicted to locate the stability fields of nitrate and N 2 , respectively (purple and black boxes in Figure 9). In these experiments, nitrate at 250 • C was unreacted for 862 h, whereas nitrate at 350 • C was partially reduced for 264 h to nitrite and N 2 O, which were expected to be reduced to N 2 via chemo-denitrification. The nitrate chemistry observed in the experiments is consistent with that predicted from the thermodynamic equilibrium of dissolved nitrogen species in the N-H-O system.
In the komatiite-H 2 O-NO 3 − systems, sampled fluids at 250 • C and 350 • C evolved to have higher aH 2(aq) values than that in the H 2 O-NO 3 − systems, which are more favorable for ammonium production (green and yellow circles in Figure 9). The N 2(aq) /NH 3(aq) ratios in thermodynamic equilibrium at the experimental aH 2(aq) and pH in-situ conditions are estimated to be less than 50 for fluid samples taken during the interval of nitrate reduction at 250 • C (the first 264 h; N250-1-N250-6), whereas they increase from less than 5,000,000 to 50 for fluid samples taken during the first 264 h at 350 • C (N350-1-N350-3). Therefore, theoretical prediction assuming thermodynamic equilibrium of nitrogen species in N-H-O system underestimates the amount of NH 3(aq) produced from nitrate in the komatiite-H 2 O-NO 3 − systems by a factor of 10 and 10-1,000,000. This suggests that nitrate reduction proceeds at the site of H 2 generation where local aH 2(aq) is higher than aH 2(aq) of bulk fluids (e.g., heterogeneous reaction at the solid surface). It is also possible that nitrate reduction is mainly driven by a reductant other than H 2 .
Ferrous iron components in komatiite (FeO rock ) are potential candidates for reaction sites to reduce nitrate. A general H 2 generation reaction during serpentinization is written as iron oxidation reactions 14 and 15 in Table 7 [68]. In the komatiite-H 2 O-NO 3 − systems at 250 • C, the H 2 generation reaction also occurs during andradite formation as discussed in Section 4.1 (rxn. 4). As discussed in Section 4.1, nitrate reduction in experiments N250 and N350 occurred during the time interval that was thermodynamically favorable to andradite formation (at 250 • C) and iron oxidation to form magnetite and/or Fe III -bearing serpentine (at both 250 • C and 350 • C). Thus, we suggest the following reactions as nitrate reduction in the komatiite-H 2 O-NO 3 − system (see Appendix A for details).
• Nitrate reduction during serpentinization (T = 250 • C and 350 • C, P= 500 bars): The reactions 19 to 24 represent nitrate reduction by both H 2 and Fe II at the site of serpentinization and andradite formation. They probably contribute to a lower N 2(aq) /NH 3(aq) ratio in the komatiite-H 2 O-NO 3 system than expected due to the thermodynamic equilibrium of dissolved nitrogen species in N-H-O system at 250 • C and 350 • C, 500 bars.

H 2 Generation Ability of Komatiite-Hosted Hydrothermal System in Early Earth
Previous simulations of komatiite hydrothermal alteration in the early ocean [49,50] typically considered two types of experiments, i.e., those under CO 2 -bearing and CO 2free conditions. These contrasting conditions not only reflect geological records (such as strongly carbonated komatiites and carbonate-free, serpentinized komatiites) [69], but also the two types of fluid circulation paths. The first type addresses the direct reactions with Hadean-Archean seawater containing high concentration of ΣCO 2 [50]; this is similar to sulfate in modern seawater introduced directly through faults into reaction zones at the base of seafloor hydrothermal systems, forming sulfate minerals that strongly affect the redox state of hydrothermal fluid, e.g., [70,71].
The second type assumes reactions between komatiite and CO 2 -free fluid in a reaction zone where CO 2 in downwelling seawater has been mostly precipitated as carbonate minerals in the surrounding rocks (e.g., recharge zone) before the fluid reaches reaction zones [49]. Previous studies revealed that the formation of carbonate minerals in the reaction zone suppressed H 2 generation given that most of iron originally contained in igneous phases was incorporated directly into carbonate minerals as Fe(II) without going through oxidation and reduction processes in water that generate H 2 [50]. The experiments in this study were a simulation of the CO 2 -free hydrothermal alteration of komatiite in a reaction zone, which has higher reduction abilities than those in CO 2 -bearing systems.
In this case, a certain amount of H 2 generation in the fluids was expected in the CO 2 -free experiments. However, the H 2 concentrations in fluid in N350 is 1.7 mmol/kg, which is almost identical to or lower than that in CO 2 -bearing experiments at 350 • C. On the contrary, the H 2 concentration in N250 is 100 times higher than that of the CO 2 -bearing experiments at 250 • C (Table 3). This indicates that the reduction ability of komatiite was partially used for the reduction of nitrate during the experiments, as discussed in Section 4.2. Accordingly, the potential H 2 concentrations were calculated for the case where nitrate was not included in the initial solutions to assess the maximum H 2 generation ability of komatiite during its alteration at 250 • C and 350 • C. Using the N 2 (aq)/NH 3 (aq) ratio of 4.5 in experiments N250 and N350, the presumed mass balance reaction of nitrate reduction by H 2 (aq) that is generated from the reaction of ferrous iron with water can be expressed as This reaction implies that the reduction of nitrate in the initial fluid requires 2.6 times H 2(aq) , suggesting that an additional 26 mmol/kg and 2.6 mmol/kg of H 2(aq) were potentially generated in N250 and N350, respectively. Certainly, these values should be upper limits for H 2 concentrations given than nitrate can be reduced by ferrous iron (Section 4.2.). However, the potential H 2 concentrations in N250 and N350 and the results of the previous CO 2 -free experiment at 300 • C (H 2 = 23 mmol/kg) [49] suggests that the optimal temperature range for H 2 generation is 250-300 • C during the serpentinization of Al-depleted komatiite in CO 2 -free conditions. This temperature range is comparable to the optimal temperature range for H 2 generation during the serpentinization of peridotites inferred from thermodynamic constraints [68,72].

Roles of Komatiite-Hosted Hydrothermal System on Pre-Biotic Chemistry
Komatiite-hosted hydrothermal systems have influenced pre-biotic chemistry and affected the sustainability of primitive microbial communities through the production of H 2 , alkalinization of fluids, and ammonia production in the komatiite-H 2 O-NO 3 − system. H 2 is a key molecule for pre-biotic chemistry and energy metabolism. Theoretical predictions for the thermodynamic state of pre-biotic chemical evolution have highlighted that an H 2 -rich hydrothermal environment was energetically advantageous for the synthesis and preservation of amino and fatty acids [73]. Further, the most plausible energy metabolisms to support the emergence and early evolution of life are considered to be H 2 -driven hydrogenotrophic methanogenesis/acetogenesis and/or methanotrophic acetogenesis, e.g., [43,45,46]. On the basis of fluid chemistry and mineralogy in the komatiite-H 2 O-CO 2 system, Ueda et al. (2016) have suggested that H 2 -rich fluid might not have been produced by the hydrothermal alteration of komatiite under CO 2 -bearing conditions at temperatures ≤ 250 • C [50]. This implies that hydrothermal systems on the deep seafloor (more than 1000 m below the sea surface) were favorable for the sustainability of primitive microbial communities as high hydrostatic pressure was necessary to create high-temperature fluids. In contrast, the steady-state H 2 concentrations in the komatiite-H 2 O-NO 3 system at 250 • C were 2.3 mmol/kg for experiment N250 and could be as high as 26 mmol/kg if the initial nitrate concentration is lower than 10 mmol/kg (Section 4.3). Such hydrogen concentrations are comparable to those in high-temperature hydrothermal fluids venting from a modern, peridotite-hosted system that harbors H 2 -based chemosynthetic microbial communities, including methanogens (H 2 = 3-16 mmol/kg) (e.g., Kairei and Rainbow Fields) [74,75]. Thus, H 2 -based, primitive microbial communities in the Hadean ocean may have developed in komatiite-hosted hydrothermal systems of lower fluid temperature (down to 250 • C) than previously thought.
Alkaline fluids were necessary for primordial metabolisms driven by chemiosmosis and electrochemistry at the vent-ocean interface (i.e., hydrothermal mounds), where hydrothermal fluids mixed with acidic and/or neutral, CO 2 -rich Hadean seawater [76][77][78]. For example, the selective synthesis of alanine from pyruvate and ammonia (>30% in yield) occurs in water with pH of 9-11 at 70 • C in the presence of iron oxyhydroxides [76]. Certain steps in the reductive tricarboxylic acid cycle are promoted by FeS-Fe 0 assemblages that can be produced electrochemically in alkaline water with pH of 10 at 100 • C in the presence of 1 mmol/kg H 2 [78]. This study shows that the pH of fluid in the komatiite-H 2 O-NO 3 − system becomes more alkaline than that in the komatiite-H 2 O-CO 2 system at 250 • C (pH 25 • C = 8.7 and 6.0, respectively) ( Table 3). Thus, hydrothermal fluids generated in the komatiite-H 2 O-NO 3 system may have been more favorable to certain types of primordial metabolisms at the vent-ocean interface where the reaction zone temperature was around 250 • C. In contrast, when the temperature of the reaction zone rises over 350 • C, the hydrothermal fluids in the komatiite-H 2 O-CO 2 system become more favorable for primordial metabolisms as the destabilization of carbonate minerals above 350 • C stimulates the alkalinization effect by ΣCO 2 in the aqueous fluid, producing an alkaline pH in the CO 2 -bearing fluid [48,50,79].
Ammonia is produced from nitrate at a high yield (10%) in the komatiite-H 2 O-NO 3 system both at 250 • C and 350 • C (Table 1). Ammonia is a key reactant for synthesizing amino and nucleic acids, and previous research successfully produced organic nitrogen compounds (amides, amino acids, and nitriles) by reacting ammonia with a variety of organic compounds (e.g., methane, formic acid, oxalic acid, and pyruvate) under limited hydrothermal conditions with and without minerals [76,[80][81][82][83]. As the fluid residence time from the onset of the high-temperature (>200 • C) reaction to venting at the seafloor (several years for present-day hydrothermal systems) [84,85] is much longer than the time required for complete reduction of nitrate at ≥250 • C (<300 h) (Table 1), a stable supply of ammonia to hydrothermal environments could be expected for the komatiite-H 2 O-NO 3 − system in the Hadean ocean. Although the actual concentration of nitrate in the early oceans (thus, ammonia concentration in hydrothermal fluids) may have been lower than that used in our experiments, the stable supply of ammonia may have allowed abiotic synthesis of organic nitrogen compounds at least in localized environments that can concentrate ammonia such as pore spaces of clay mineral clusters in altered komatiite crusts, and interiors of iron oxyhydroxides in hydrothermal mounds (e.g., Figure 5g). However, further studies are necessary to explore pathways for the synthesis of organic nitrogen compounds that are resistant to the fluctuation of environmental conditions (pH, temperature and coexisting mineral type) that the hydrothermal vent environments probably encountered.

Implications for Biological Nitrogen Metabolism, Phytosynthetic Activity, and the Global Nitrogen Cycle in the Early Earth
The stable and high-yield production of ammonia in the komatiite-H 2 O-NO 3 − system is an important process not only for pre-biotic chemistry in the hydrothermal environments but also for controlling ammonia concentrations in the early global ocean that probably influenced biological nitrogen metabolism, the development of photosynthetic ecosystems, and global nitrogen cycle in the early Earth. Approximations of the ammonia production rate in the komatiite-hosted hydrothermal systems in the early ocean, F NH3,ko (mol/y), suggest that biological activity had started in hydrothermally influenced local environments, but was not likely widespread throughout the global surface ocean. The F NH3,ko (mol/y) value can be estimated from the following relation: where Y NH3,ko is the yield of ammonia by thermochemical nitrate reduction in the komatiitehosted hydrothermal systems, F H2O is the high-temperature (≥250 • C) fluid flux from the komatiite-hosted hydrothermal systems (kg/y), and C NO3 is the nitrate concentration in seawater (mol/kg). Our experiments suggest that Y NH3,ko value is 0.1, whereas numerical modeling by Laneuville et al. (2018) indicated that C NO3 in early seawater was dependent upon atmospheric nitrate production rate and had a range from 10 −6 to 10 −2 mol/kg [86], cf. [87,88]. The F H2O value is initially approximated below, followed by the F NH3 value.
In the modern ocean, hydrothermal fluid flux is considered proportional to the annual production rate of oceanic crust at fast-spreading ridges [89]. Numerical modeling also suggests that plate velocity has not changed much throughout Earth's history [90,91]. Thus, the high-temperature fluid flux from the komatiite-hosted hydrothermal systems can be approximated as follows: F H2O,ko = k P ko , where P ko is the annual komatiite production rate (km 3 /y) and k is a coefficient that relates high-temperature fluid flux to P ko (kg/km 3 ). Field observations suggest that the thickness of the Archean oceanic crust was approximately three times greater than modern equivalents owing to the higher degree of partial melting [41,42]. Further, numerical modeling suggests that the global spreading rate of Archean oceanic crust (the product of plate velocity and plate boundary length: km 2 /year) was approximately three times higher than that at present [90]. Thus, the annual production rate of Archean oceanic crust was estimated to be 180 km 3 /y, that is, nine times greater than at present [92]. If the production ratio of oceanic plateau to oceanic crust at mid-oceanic ridges in the Archean was identical to that at present (0.05), the P ko value was estimated to be 10 km 3 /y in the Archean, which is nine times greater than that of the present oceanic plateau (1.1 km 3 /year) [93]. Given that high-temperature (>250 • C) fluid flux in the modern ocean is estimated to be 0.2-3 × 10 13 kg/y, e.g., [94][95][96], k is calculated to be 10 11 -10 12 kg/km 3 . Collectively, the F H2O,ko and F NH3,ko values of the Archean ocean are estimated to be 10 12 -10 13 kg/y and 10 5 -10 10 mol/y, respectively (Table 8). Given higher heat flux from Earth's interior in the Hadean than Archean, the F H2O,ko and F NH3,ko values of the Hadean ocean are considered to be greater than the Archean ocean. Table 8. Estimates of ammonia and dinitrogen production rates by thermochemical nitrate reduction in the early oceans.
The F NH3,ko value provide insights into the evolution of biological nitrogen metabolism. It has to be noted that our estimate of the F NH3,ko value is based on simple equations and several assumptions, thus allowing the F NH3,ko value to have a wide range from 10 5 to 10 10 mol/y. Nonetheless, our estimate indicates that the ammonia concentration in the early oceans had increased to 0.01-1000 µmol/kg from 4.4 Ga to 3.4 Ga if ammonia consumption rate in the ocean (e.g., biological assimilation) was much lower than the ammonia production rate. Microorganism can assimilate ammonia at the micro-to-milli molal levels for growth. For example, a half saturation constant for ammonia uptake has been reported to be 4-74 µmol/kg for bacteria isolated from different habitats [97]. Thus, if nitrate concentration in the early oceans was greater than 10 µmol/kg, the long-term production of ammonia through thermochemical nitrate reduction for the first billion years might have allowed the subsequent development of an early biosphere in the global surface ocean. In contrast, if nitrate concentration in the early oceans was less than 1 µmol/kg, komatiitehosted hydrothermal activity would not have contributed to the accumulation of ammonia in the global ocean. In this case, the onset of biological nitrogen fixation (N 2 → NH 3 ) was required for the development of early biosphere, consistent with genomic, physiologic, and geological records that are compatible with nitrogen fixation by the last universal common ancestor or methanogen in hydrothermal environments at 3.5 Ga [4,5,98,99], cf. [100].
To determine which scenario is more appropriate for the evolutional history of nitrogen metabolism, future work is required to estimate nitrate concentration in early seawater from geological records. Nitrogen isotopic characterization of abiotically produced ammonia is also important to determine the source of nitrogen isotopic compositions in Archean sedimentary rocks of marine origin, which have been interpreted as evidence for biological nitrogen fixation in surface oceans at 3.2 Ga [101] on the basis of nitrogen isotopic similarity to N 2 in the Archean atmosphere (ca. 0‰) [10,[102][103][104].
Results of our experiments also suggest that 90% of nitrate nitrogen introduced to the komatiite-hosted hydrothermal system had probably been converted to N 2 , resulting in a larger input of N 2 to the early ocean than that of ammonia. The higher input of N 2 may have stabilized the partial pressure of N 2 (pN 2 ) in the atmosphere and climate in an era of low solar luminosity as the increase in pN 2 would have increased the warming effect of existing greenhouse gases in an anoxic atmosphere (CO 2 , CH 4 ) [105]. The approximate global N 2 flux from thermochemical nitrate reduction in early Archean seafloor hydrothermal systems, including both komatiite-and basalt-hosted systems, is F N2 (mol/y), which can be compared to the N 2 fluxes from other processes to develop a quantitative assessment of the contribution of thermochemical nitrate reduction to pN 2 in the early Archean.
The estimated F N2 value in the early Archean includes the estimate of the N 2 production rate in the komatiite-hosted hydrothermal system in the early Archean, F N2,ko (mol/y) using the following equation: where Y N2,ko is the yield of N 2 by thermochemical nitrate reduction in the komatiitehosted hydrothermal systems. By substituting 0.45, 10 12 -10 13 kg/y, and 10 −6 -10 -2 mol/kg for Y N2,ko , F H2O,ko , and C NO3 in Equation (28), respectively, F N2,ko is calculated to be 10 6 -10 11 mol/y. If the production ratio of oceanic plateau to oceanic crust at mid-oceanic ridges in the Archean was identical to that at present (0.05), high-temperature fluid flux from the basalt-hosted hydrothermal system at Archean mid-oceanic ridges is one order magnitude larger than F H2O,ko . Accordingly, F N2 is estimated to be 10 7 -10 12 mol/y (i.e., 10 8 -10 13 gN/y) in the early Archean, if the yield of N 2 by thermochemical nitrate reduction in the basalt-hosted system Y N2,mo had been equal or greater than Y N2,ko , which is an appropriate assumption due to lower potentials of H 2 and NH 3 generation in basalt-hosted system than komatiite-hosted system. When comparing N 2 fluxes from major processes, the F N2 value in the early Archean is quantitatively important. Major processes that supply N 2 to the present atmosphere are biological denitrification in the global ocean (NO 3 − → N 2 ) (1-3 × 10 14 gN/y) [106], degassing from the MORB source mantle (3 × 10 10 gN/y) [107] and degassing by arc volcanism (3 × 10 11 gN/y) [108]. In the early Archean, geological records suggest that biological denitrification was not yet underway, whereas degassing fluxes of N 2 from the MORB source mantle and via arc volcanism could have been one order of magnitude greater than today as they change proportionally with the annual production rate of oceanic crust. Therefore, N 2 flux by thermochemical nitrate reduction in the global hydrothermal systems would have been greater than that from the Earth's interior if nitrate concentration in Archean seawater had been greater than 1000 µmol/kg.
In summary, the approximate N 2 fluxes to the atmosphere in the early Archean from hydrothermal systems by thermochemical nitrate reduction and from the Earth's interior were 10 8 -10 13 gN/y and 10 12 gN/y, respectively. Meanwhile, fluid inclusions in the Archean rock samples indicate that the abundance of N 2 in the Archean atmosphere was 2-12 × 10 21 gN, which is 0.5-3 times the present atmospheric N 2 [10,103,104,109]. Therefore, the residence time of atmospheric N 2 in the early Archean is calculated to be 10 8 y or greater. For comparison, the residence time of atmospheric N 2 at present is 10 7 years due to the high rates of biological nitrogen fixation and biological denitrification in the global ocean (1-3 × 10 14 gN/y) [106]. Although pN 2 and its secular variation during Earth history have not yet been estimated precisely, our simple estimate implies that pN 2 is stable for ≥10 8 years in the early Archean if the global nitrogen cycle was driven primarily by abiotic nitrogen processes in the atmosphere and seafloor hydrothermal systems. Together with the emission of hydrogen and methane gases during serpentinization, this point should be considered when evaluating the role of hydrothermal activity in the long-term climate of the early Earth.

Conclusions
We conducted hydrothermal experiments simulating the high-temperature, highpressure reactions between nitrate in seawater to find probable chemical pathways to deliver ammonia to the vent-ocean interface of komatiite-hosted hydrothermal systems and the global ocean on geological timescales. Our experiments provide the following key results and conclusions.

•
In the komatiite-H 2 O-NO 3 − system, ammonia was produced from nitrate with yields of 10% at 250 • C and 350 • C in a CO 2 -free condition. The fluid chemistry and mineralogy suggest that nitrate reduction co-occurred with serpentinization and andradite formation at 250 • C, whereas it co-occurred with serpentinization at 350 • C. This represents a case where CO 2 in downwelling fluid precipitated as carbonate minerals in surrounding rocks before reaching the high-temperature reaction zones. • Nitrate reduction by both H 2 and Fe III at the site of serpentinization and andradite formation probably contributed to a lower N 2(aq) /NH 3(aq) ratio (4.5) than expected due to the thermodynamic equilibrium of dissolved nitrogen species in N-H-O system at 250 • C and 350 • C, 500 bars. • H 2 -rich and alkaline fluids, which are prerequisites for both prebiotic and primordial metabolisms, were generated in the komatiite-H 2 O-NO 3 system at lower temperatures (down to 250 • C) than previously thought under CO 2 -rich conditions. The komatiite-H 2 O-NO 3 − system also provided a stable supply of ammonia to the hydrothermal environments for biomolecule synthesis. Given that the temperature of hydrothermal fluids depends on hydrostatic pressure, the komatiite-H 2 O-NO 3 − system probably reduces the minimum water depth of seafloor hydro-thermal systems to about 1000 m, which may have favored the start of pre-biotic chemistry and the sustainable development of H 2 -based microbial ecosystems at the vent-ocean interface of the Hadean ocean.

•
The ammonia production rate in the komatiite-hosted hydrothermal systems depends on nitrate concentration in seawater and is estimated to be 10 5 -10 10 mol/y in the Archean ocean. The rate in the Hadean ocean was likely to be greater than the Archean ocean given higher heat flux in the Hadean ocean. If nitrate concentrations in the early oceans were greater than 10 µmol/kg, the long-term production of ammonia by thermochemical nitrate reduction for the first billion years might have allowed the subsequent development of an early biosphere in the global surface ocean through biological ammonia assimilation.

•
The N 2 production rate in the komatiite-hosted hydrothermal systems is estimated to be 10 6 -10 11 mol/y in the Archean ocean. The estimate implies that pN 2 in the Archean atmosphere is stable at least for 10 8 years if abiotic nitrogen processes in the atmosphere and seafloor hydrothermal systems mainly drove the global nitrogen cycle.
Collectively, these findings imply that komatiite-hosted systems might have impacted not only H 2 -based chemosynthetic ecosystems at the vent-ocean interface but also photosynthetic ecosystems on the early Earth.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

•
The accelerating voltages for the secondary electron imaging and semi-quantitative chemical analyses by SEM/EDX were 5 kV and 15 kV, respectively. • EPMA parameters for the quantitative analyses were 15 kV of accelerating voltage, and 12 nA of specimen current. • XRD analyses were conducted at 30 kV and 15 mA using Cu-Kα radiation. XRD peaks were assigned with the aid of PDXL software using the Inorganic Crystal Structure Database. • XRF analyses were conducted at 50 kV and 60 mA using an Rh target. • Raman spectra of the solid reaction products were obtained with 532 nm excitation. The incident laser power density was typically 100-700 kW/cm 2 for 100× (NA, 0.90) objective lenses. The final laser power at the sample surface was estimated to be 1-3 mW, and the spatial resolution of the Raman analysis was finer than 1 µm. Each point typically was scanned three times for 10-90 s from 100 to 2600 cm -1 at a spectral resolution of 2.3 cm -1 with a grating containing 600 grooves/mm. The Raman shift was calibrated using a silicon wafer prior to the analyses.