Introduction

Banded Iron Formations (BIFs) are ancient marine chemical sedimentary deposits rich in iron (Fe) and silica (Si), with peak deposition 4000–1850 Million years (Ma) ago and briefly during the 720–635 Ma Neoproterozoic Snowball Earth glaciations1,2,3. They provide key information about the chemistry of the ancient seawater from which they were formed and about the iron cycling in the early oceans. Algoma-type BIFs, often interlayered with Archean-Paleoproterozoic-aged volcano-sedimentary sequences, are interpreted to have been deposited near submarine volcanic centres via exhalative-hydrothermal processes1,2. Their distinct sedimentary and geochemical features suggest an open or partially restricted deep sea depositional setting. Algoma-type BIFs tend to be widespread in Archean cratons and sporadically appear in Proterozoic marine sedimentary environments. The deposition of Quaternary Algoma-type BIF analogue rocks on the coast of Milos, Greece4,5 suggests Algoma-type BIFs may also be associated with near continental margin shallow hydrothermal vent fields, driven by microbial iron cycling6,7. In contrast, Superior-type BIFs are typically interbedded with Paleoproterozoic carbonates and black shales, reflecting deposition in nearshore shelf environments1,2. Algoma-type BIFs are much thinner and smaller in lateral extent than Superior-type BIFs, although the former can be more abundant in individual occurrences8. Rapitan-type BIFs, linked to the Neoproterozoic Snowball Earth glaciations, were formed when thick ice sheets resulted in the deoxygenation of the seawater, enabling ferruginous conditions to develop in the global ocean through hydrothermal exhalation2.

Generally, BIFs contain at least 15 wt.% bulk Fe and often show spectacular alternating millimetric (microbands) to metric silica and Fe(III) oxide or Fe(II) silicate layers1,2,3. According to a consensual model, the large-scale accumulation of Fe required that dissolved Fe(II) sourced from hydrothermal vents was oxidized by various mechanisms to insoluble Fe(III) in association with abiotic precipitation of amorphous silica1. Existing geochemical models propose that the microbands in Algoma-type BIFs likely formed through seasonal variations or microbial activity9. Similar microbial mechanisms have been suggested for the deposition of Quaternary shallow submarine hydrothermal vent field analogues4,5. However, there is a notable lack of in situ studies that detail the pathways of Fe(II) oxidation leading to the formation of primary amorphous Fe(III) minerals in BIFs and the processes involved in the sequential transformation of these minerals into highly crystalline Fe(III) phases with low organic carbon content, remain understudied. Here, we provide a mechanism from a modern environment that could support a potential biological origin for Algoma-type BIF-like microbands in a modern deep-sea hydrothermal setting.

Planktonic (i) O2-producing photosynthetic cyanobacteria in the Proterozoic Ocean, coupled with abiotic Fe oxidation, or chemolithoautotrophic Fe(II)-oxidizers, and (ii) Archean anoxygenic photoautotrophic Fe(II)-oxidizing bacteria (photoferrotrophy)6,10,11,12,13, are among some of the most prominent mechanisms proposed for Precambrian BIF deposition. For example, microaerophilic Fe(II)-oxidizing microorganisms (FeOM), including Zetaproteobacteria, that likely evolved to dominate pelagic microaerobic conditions following the Great Oxidation Event (GOE), ca. 2500–2100 Ga14,15, are suggested to have produced a “ferrihydrite rain” to the Precambrian seafloor2. In contrast, a consortium of benthic microbial communities may have formed discrete microbands of iron oxides in stromatolitic facies16. However, it remains unclear how these processes and benthic microbial communities contributed to the generation of BIF bands.

Although present in varying amounts, iron oxides in Archean and Proterozoic BIFs are dominated by magnetite (Fe(III)2Fe(II)O4) and haematite (Fe(III)2O3), respectively17. Both are interpreted as having formed from an initial poorly crystalline Fe(III)-oxyhydroxide precursor phase, such as ferrihydrite1,3 (Fh – Fe(III)(OH)3), suggesting the iron oxide mineralogy of Precambrian BIFs is likely secondary rather than primary. Alternatively, greenalite (Fe(II)3Si2O5(OH)4) formed under oxidant-depleted environmental conditions, has been suggested as an alternate primary Fe precursor in BIFs18,19,20 prior to the GOE. Based on this theory, iron oxides in Archean BIFs are secondary products formed following post-depositional oxidation of the primary ferrous silicates21,22. In addition, green rust, a ferrous–ferric hydroxy salt, which would have precipitated near the Fe redoxcline and transformed into stable mineral phases (e.g., magnetite) within the water column and sediments, has been suggested as a precursor to BIF mineral assemblages23. Consequently, there is no consensus on the dominant Fe mineral precursors to BIFs. One reason for these disagreements arises from the scarcity of modern analogues of banded Fe-Si-rich precipitates from diverse marine settings mimicking the Precambrian Ocean to enable the appropriate testing and validation of these various hypotheses.

Moreover, the low organic matter content of BIFs has been used to dismiss the potential direct involvement of biological activity in the precipitation of BIF Fe(III)-rich minerals. However, near-complete respiration/oxidation of organic matter coupled to Fe(III) reduction by various bacteria and archaea are frequently hypothesized to explain the low organic C content (<0.8 wt.%) in BIFs1,12,24. This view is anchored on intimately associated magnetite and 13C-depleted Fe(II)-bearing carbonates (e.g., siderite), and the general absence of definitive microfossils preserved in the non-silicified Fe-rich bands1,24. However, direct (i.e., textural) evidence for biological activity provided by microfossils and stromatolitic oncoidal structures associated with granular IF deposition has been recently documented25,26,27. Therefore, understanding the complex interplay between BIFs, ocean chemistry, and the biosphere to unravel the mechanisms responsible for the extensive generation of organic C-lean BIFs, remains an ongoing endeavour as are the pathways involved in the transformation of primary Fe-bearing mineral precursors into crystalline phases.

To provide new insights into these primary and secondary BIF formation pathways, we focus on present-day active submarine hydrothermal ecosystems displaying high dissolved Fe fluxes associated with chemolithoautotrophic microaerophilic FeOMs, typical of the mid-ocean ridges, back-arc spreading centres, and submarine arc volcanoes28. There, hydrothermal fluid discharge produces chemical and redox gradients that enable the development of a variety of microbial mats along cracks and fractures on various seafloor lithologies29,30,31,32,33,34,35. Although deep-sea hydrothermal environments have been widely studied for their microbial composition and environmental conditions, there is a dearth of knowledge on their potential association with the biological development of Precambrian BIFs35,36,37,38,39. We therefore collected in situ modern Fe-rich microbial mats from the Lucky Strike Hydrothermal Field (LSHF; 37°17’N & 32°16’W; Mid-Atlantic Ridge40) (Fig. 1; Table 1; Supplementary note 1), to investigate a potential linkage and association. At the seafloor, white and orange microbial mats, linked to sulphur oxidation30 and Fe oxidation29,41, respectively, flourish on a heterogeneous substratum in close vicinity of diffuse vents at temperatures <100 °C42,43 from different well-known sites, including Capelinhos (CAP), North Tour Eiffel (NTE), South Isabel (SI), West Sintra (WS), and Y3.

Fig. 1: Bathymetric map of the Lucky Strike hydrothermal field and images of the sampled iron-rich microbial mats, modified from Chavagnac et al.60 and Astorch-Cardona et al.29.
figure 1

The studied mats and their associated active hydrothermal vents are shown. No hydrothermal vent was detected at LL site. Chlorinity increases from CAP to LL sites. Scale bars: 5 cm. [Capelinhos (CAP); North Tour Eiffel (NTE); South Isabel (SI); West Sintra (WS); Lava Lake (LL)].

Table 1 Specific features of the studied Fe(II)-oxidizing microbial mats from Lucky Strike Hydrothermal Field. For further details, see Supplementary discussion concerning the high positive δ13Ccarb values

Given the relatively high Fe(II) concentrations in Archean oceans (µM then versus nM today)44, it is commonly accepted that deep-marine hydrothermal vents in the Archean delivered tremendous amounts of Fe(II) and other reduced compounds that modulated nutrient cycling (e.g., C, N, S, Fe) and contributed to BIF deposition1,45. The LSHF mats host abundant microbial communities, including autotrophic Fe(II)-oxidizing Zetaproteobacteria coexisting with N, S, and methane-cycling microorganisms29,41 (see Supplementary note 1). Zetaproteobacteria are the presumed primary producers in iron-rich mats flourishing in diverse hydrothermal environments41,46 as they are obligate chemoautotrophs, oxidizing Fe(II) to gain energy. Thus, they are essential for the development of ecological niches within the mats produced in Fe(II)-rich hydrothermal vent fields.

Considering that metabolic pathways within such nutrient cycles (i.e., C, N, S, Fe) emerged and diversified in the Archean45, including microaerophilic microbial Fe oxidation47, we couple petrographic and mineralogical observations with C-N systematics in the LSFH mats to validate primary and secondary microbial processes potentially involved in BIF microbands. We discuss the possibility that similar mat-forming FeOMs could have contributed to the generation of microbands in BIFs that formed close to hydrothermal vent field settings, for example, Algoma-type BIFs, under constrained microaerobic environmental depositional conditions and hydrothermal exhalation of Fe(II)-rich fluids.

Results

Texture and mineralogy of microbial mats

Except for the Lava Lake (LL) site where no diffuse venting was observed, two types of microbial mats, dominated by crumbly and flocculent structures, flourished in the vicinity of diffuse hydrothermal fluid venting at CAP, NTE, SI, WS, and Y3 (Fig. 1)48,49,50. Coherent and porous deposits with variable thickness of <20 cm, together with occasional dark-coloured laminated structures typify the crumby microbial mats located at the CAP, NTE, and Y3 sites (Figs. 1, 2a−i, 3a−g, and 4a−i; Table 1). The mats at the SI, WS, and LL sites are composed of a mixture of flocculent material associated with microorganisms and inorganic particles (Fig. 1; Table 1). All the sampled Fe-rich mats have been previously studied for their Zetaproteobacterial diversity and microbial filament morphology29.

Fig. 2: Morphological, textural, and elemental image of Capelinhos (CAP A) mat.
figure 2

a Brown to orange Fe(II)-oxidizing microbial mats. b Thin section. The Fe mineralogy exclusively consists of ferrihydrite (Fh), as demonstrated by XRD results of our study. c, d Transmitted light photomicrographs (plane-polarized light). Filamentous microbial structures. e–h SEM images in BSE mode showing the aggregation of a large number of filaments (mainly stalks here). Filaments crossing with each other to form a mesh structure. i EDS analysis of stalks displaying a predominantly Fe- and Si-rich composition.

Fig. 3: Morphological, textural, and elemental image of NTE (NTE B) mat.
figure 3

a Brown to yellowish Fe(II)-oxidizing microbial mats with mm-thick dark laminae. The sample is extremely crumbly. b Thin section. Note the absence of specific structures. c Transmitted light photomicrograph (plane-polarized light). d–f SEM images in BSE mode showing the presence of numerous minerals of various shapes and sizes. The Fe mineralogy exclusively consists of ferrihydrite (Fh), as suggested by the Mössbauer spectroscopy. The identification of Mn oxides and silicates was carried out with EPMA. g EDS analysis of Fh displaying an Fe- and Si-rich composition.

Fig. 4: Morphological, textural, and elemental image of Y3 (Y3 B) mat.
figure 4

a Brown to orange Fe(II)-oxidizing microbial mats with darker bands (arrows). b Thin section. Note the micrometric laminated structures with darker layers towards the bottom of the mat. c, d Reflected light photomicrograph (plane-polarized light) of the dark layers and related EDS analysis. Close association of magnetite, a grey-coloured mineral with a higher reflectance than other Fe(III)-rich minerals, and siderite. e–i SEM images in BSE mode and related EDS analysis showing crystal shapes of crystalline Fe oxides and oxyhydroxides (here referred to as Fe oxides for simplicity). Highly aggregated octahedral crystals in (e) colloform texture with radiating fibrous morphologies in (f) and (g) and lath-like crystals in (h). Crystalline Fe(III)-bearing minerals mainly consist of goethite and lepidocrocite in (fh), as suggested by EPMA.

Based on petrographic observations, the CAP mats dominantly exhibit mm-scale outward-convex structures with internal concentric bands consisting of a more or less porous network of filamentous structures dominated by stalks (Fig. 2b−h), a typical feature linked to FeOMs51. The twisted stalks, among other morphologies of filaments in the CAP mats29, are thought to reflect dynamic microbial mat growth features13,52. X-ray diffraction (XRD) patterns further show two broad peaks centred at ~2.70–2.90 Å and ~1.51 Å (Fig. 5; Supplementary Fig. 1), which deviate from the pure 2-line Fh peaks characterized by d-spacings at 2.54 and 1.47 Å53,54. This shift, together with the poorly resolved peak at ~1.51 Å, the energy-dispersive spectra (EDS) analysis (Fig. 2i), and infrared-based spectroscopy (Supplementary Figs. 2, 3 & Supplementary note 2), point to poorly crystalline Si-bearing Fh55,56. The combined datasets highlight that the Fe-rich microstructures are heavily mineralized by Si-bound Fh.

Fig. 5: Representative XRD patterns of randomly oriented powders and Mössbauer spectra from the LSHF iron-rich microbial mats.
figure 5

Diagnostic XRD spacings (Å) are shown. Although the presence of crystalline minerals makes amorphous to poorly crystalline phases difficult to detect using XRD, the isomer shift (IS) and quadrupole splitting (QS) are consistent with reported values for Fh at SI, NTE, and Y3 sites (see Supplementary Note 2). The light and dark red shaded regions indicate the center of broad peak areas of Si-Fh and Fh, respectively. Note that a diffractometer using MoKα radiation was used to analyse Fe- and Mn-containing mats at the NTE site. [Ferrihydrite (Fh); room temperature (RT); diffraction peak assigned to the sample holder (*)].

Contrary to the CAP site, mat growth patterns are not visible at the NTE site. Instead, micronodule structures of <120 µm in diameter and angular minerals from the altered substratum compose the mats (Fig. 3b–f). Results from XRD, Mössbauer and infrared-based spectroscopies, and EDS analyses highlight the widespread occurrence of poorly crystalline Mn oxides (todorokite, birnessite, and pyrolusite), anorthite, and amphibole (edenite), and Si-bound Fh as the main Fe(III)-rich mineral (Figs. 3g, 5; Supplementary Figs. 13; Supplementary Table 1).

The Y3 mats comprise two Fe-rich microbands composed of multiple laminated and upward-convex layers (Fig. 4a, b), probably reflecting successive steps of mat growth52. Petrographic observation of the bottom most part of the mats is distinctively represented by a mm-thick black band consisting mainly of siderite and magnetite that appear as isotropic octahedral crystals showing a higher reflectance relative to other Fe(III)-rich minerals (iron oxides and well-crystallized Fe(III)-oxyhydroxides are referred to as FeOx) (Figs. 4c−e, 5; Supplementary Fig. 1). Towards the top of the mats, the black laminae are succeeded by alternating dark and light brown laminated layers composed of well-crystallized FeOx of various crystal shapes, including goethite [α-Fe(III)OOH], lepidocrocite [γ-Fe(III)OOH], and akageneite [β-Fe(III)OOH·Cl] (Figs. 4f−h, 5; Supplementary Figs. 13 & Supplementary note 2). In contrast to the other LSHF mats, little Si is bound to these FeOx in all laminae as suggested by the lack of an EDS Si peak, high EDS Fe/Si ratios, and the absence of Fe-O-Si stretching (Fig. 4i; Supplementary Figs. 3, 4). In addition, Mössbauer spectroscopy was employed to confirm the presence of Fh in the Y3 mats (Fig. 5; Supplementary Table 1 & Supplementary note 2). Similar to CAP mats and other natural biofilms57, microbial filaments predominantly composed of stalks, coral-like, and Y-shaped structures at Y3 associated with Zetaproteobacteria29, are thickly coated with poorly crystalline Fh. From top to bottom, the Y3 mats contain Fh-encrusted microbial filaments that disappear at the expense of increasing enrichment of well-crystallized FeOx minerals and ferrous carbonates (Fig. 6).

Fig. 6: Microtextures of Y3 (Y3 B) mat.
figure 6

SEM-BSE images of microbial filaments showing an in situ model of increasing enrichment and crystallization of sheaths and stalks with FeOx from the upper (a, b), middle (c–g), to the bottom (h), parts of the Y3 B mat. As the mat aged, the microbial filaments are usually encrusted by poorly crystalline ferrihydrite (Fh)57 first act as nucleation sites, followed by gradual transformation into well-crystallized FeOx that cover them with successive steps of FeOx crystallization (bf), and eventually disappear (gh). The orange arrow reported in (d) shows the decreasing proportion of filamentous structures with increasing depth.

The Fe mineralogy of the flocculent mats at the WS, LL, and SI sites is dominated by both Si-Fh and Fh phases (Fig. 5; Supplementary Figs. 2, 3; Supplementary Table 1). Neither Si-rich bands nor amorphous silica phases like opal-A, common in modern marine sediments and defined by a broad peak centred at ~4.10 Å (~25.2° 2θ CoKα)31,32,34,58,59 were evidenced by XRD and petrographic observations in the LSFH mats. Finally, unidentified organic molecules are associated with the Fh-dominated mats (Supplementary Fig. 3).

Bulk geochemistry

Bulk mat geochemistry at LSHF shows heterogeneous chemical major element composition, with Fe2O3 content varying between 10.8 and 99.6 % (Supplementary Table 2 & Supplementary note 2). At the NTE site, elevated concentrations of Al2O3 and CaO ranging from 9.4–13.3 % and 4.2–6.2 %, respectively, compared to other sites with <3.3% Al2O3 and <1.6% CaO, are related to basaltic-derived alteration fraction composed of anorthite and amphibole. P concentrations vary considerably in function of Fh/FeOx mineralogy, where very low (<0.04 %) P content is associated with crystalline FeOx phases and high (0.15-0.72 %) P concentrations with poorly crystalline Fh (Supplementary Fig. 5), reflecting a 3- to 18-fold decrease from crystalline FeOx to poorly crystalline Fh.

X-ray fluorescence (XRF) analysis of the LSHF mats reveals SiO2 concentrations in the range of 3.9-38.1% (Supplementary Table 3), the lowest in the Y3 mats, moderate in the CAP, WS, and LL mats, and highest in the SI and NTE mats. XRF findings agree with the above mineralogical and petrographic results and support the potential occurrence of opal-A at the SI site. Moreover, the observations indicate a limited presence of Si-bound FeOx minerals at Y3, the presence of silicate minerals at NTE, and the general lack of amorphous silica phases at LSHF.

Spatial chemical concentration of Fe-bearing mineral phases

Electron microprobe analyses (EPMA) align with the XRD mineralogy (Supplementary Table 4 & Supplementary note 2). Moreover, EPMA chemical compositions of Fh-encrusted filamentous microstructures and crystalline FeOx show three distinct groups in the crumbly mats (Supplementary Table 5). Considering the Si concentrations of end-member hydrothermal fluids are of the same range of concentration (12.7 to 18 mM) at LSHF41,60, Fh in CAP and NTE sites show an average Fe2O3/SiO2 ratio of 3.5 ± 1.2 (n = 16; 1σ), and 6.7–7.2 (n = 2) in the filaments of the Y3 mats, whereas a considerably higher ratio of 90.6 ± 46.6 (n = 17; 1σ) is recorded in the crystalline FeOx phases at Y3. The latter observation clearly aligns with electron distribution maps and infrared-based spectroscopy, illustrating stalks enriched with Si-Fh phases in the CAP mats containing quantitatively higher SiO2 content and to a lesser extent P2O5 compared to the filamentous microfossils and the more crystalline FeOx minerals in Y3 mats (Fig. 7).

Fig. 7: SEM-BSE images showing features of the Fe(II)-oxidizing mats from CAP (CAP A; left panels) and Y3 (top of the Y3 B mat; right panels) sites, and corresponding EPMA element mapping (viridis colour map).
figure 7

Fe2O3, SiO2, and P2O5 compositional variations revealed by EMPA maps. SiO2 and P2O5 concentrations in Fh-dominated CAP mat show higher proportions than those in crystalline FeOx from the Y3 mat. Red arrows denote Si-poor lath-like minerals, likely goethite/lepidocrocite, compared to Fh-encrusted microbial filaments (white arrows). Scale bars are 30 μm.

Carbon and nitrogen systematics

The C isotopic compositions of carbonates (δ13Ccarb) and organic matter (δ13Corg) and N isotope (δ15N) values in the LSHF Fe-rich mats differ considerably between and within sites and samples (Table 2). These mats contain low 0.18-0.83 wt.% total organic C (TOC) and 0.02-0.07 wt.% total N (Table 2). δ13Ccarb, δ13Corg, and δ15N values vary from -6.0 to +11.5‰ (n = 11), -35.1 to -18.9‰ (n = 11), and -1.4 to +3.1‰ (n = 4), respectively. The siderite-bearing Y3 mats show negative to slightly positive δ13Ccarb values, while siderite-free mats (SI, LL, and CAP) display both negative and highly positive δ13Ccarb values. Besides, TOC contents, δ13Ccarb, δ13Corg, and δ15N values decrease from top to bottom with increasing siderite abundance in the Y3 B mat (Table 2). Finally, P is moderately correlated with TOC (R2 = 0.50, n = 6, p < 0.12) (Supplementary Fig. 6a), whereas a weak covariation between δ13Corg and δ13Ccarb (R2 = 0.26, n = 11, p < 0.109) and a strong one between δ13Corg and δ15N (R2 = 0.74, n = 5, p < 0.054) are observed (Supplementary Fig. 6b, c).

Table 2 Carbon and nitrogen data

Discussion

Microbial activity

Carbon fixation and oxidation: Microaerophilic Fe(II)-oxidizing Zetaproteobacteria, grow at low O2 concentrations (5–20 µM61) where they compete with abiotic oxidation of Fe(II) due to the slower oxidation rate in such conditions62,63. They use the Calvin–Benson–Bassham (CBB) pathway to autotrophically build biomass64 according to the equation65:

$$6{\mbox{Fe}}^{2+} + 0.5{\mbox{O}}_{2}+ {\mbox{CO}}_{2}+16{\mbox{H}}_{2} {\mbox{O}} \to [{\mbox{CH}}_{2} {\mbox{O}}]+6{\mbox{Fe}}({\mbox{OH}})_{3}+12{\mbox{H}}^{+}$$

The variable δ13Corg values of -35.1 to -18.9‰ in the LSHF mats likely point to multiple autotrophic C fixation pathways66,67, and potential respiration of generated organic matter by Fe(III)-reducing bacteria within the mats (Fig. 8a), connecting the Fe cycle with carbon and nitrogen cycling (see Supplementary discussion). In this regard, most of our δ13Corg values are consistent with C fixation through the CBB cycle pathway that fractionates Corg isotopes by 20 to 30‰66, considering a δ13C value of the dissolved inorganic C close to 0‰. In support of this observation, the Lōʻihi seamount mats in the Pacific Ocean, containing the RubisCO gene (key enzyme in the CBB cycle)64 and Zetaproteobacteria, have shown δ13Corg values ranging from -28 to -23‰68, which further strengthens the suggestion that Zetaproteobacteria contributed significantly to the δ13Corg values bracketing between -20 and -30‰ in our samples and that they are foundational to the mat building process. In contrast, the reductive citric acid and 3-hydroxypropionate C fixing cycles produce δ13Corg values > -20‰66. Except for one sample, our δ13Corg values < -20‰ suggest that the latter autotrophic mechanisms are not dominant in the LSHF mats. Further, the anaerobic C fixation via the reductive acetyl-coenzyme A or Wood–Ljungdahl pathways produces δ13Corg values <-30‰46,66, similar to those observed in some Y3 mats. Although the extent of methanotrophic microorganisms is limited in the LSHF mats29,41, they could have also contributed to the lowest δ13Corg values due to the 13C-depleted methane hydrothermal source69 and large isotopic fractionation of up ~30‰ during methanotrophy70.

Fig. 8: Proposed conceptual models of the microbial composition and evolution of the LSHF mats.
figure 8

a Mat trophic structure (Y3 site) and their effect on C isotope compositions. b Formation of the Fe-rich microbanding patterns and their related mineralogical composition and biotic/abiotic processes in the LSHF Fe-rich mats. (I-III) Successive steps of the generation of the microbands at Y3. Temperature of the diffuse fluids controls Fh particle size, which, in turn, affects surface area reactivity and adsorption capacity for chemical elements. Note that diffuse hydrothermal fluids were not observed during mat collection at WS, SI, and LL sites. For simplicity, stalk-formers are only represented in these mats. Vertical and horizontal scales are not represented. Not all microbial metabolisms described at LSHF29,41 and seawater are shown. LMWOC, low molecular weight organic compounds.

Mechanism of microbial Fh precipitation

Zetaproteobacteria oxidize Fe(II) on mat surfaces, leading to even greater O2 depletion resulting in the development of anaerobic heterotrophic microbial communities deeper in the mats. This is consistent with increasing oxidation of organic C by dissimilatory Fe(III)-reducing bacteria, resulting in a much lower residual biomass content and formation of siderite and magnetite deeper in the Y3 mat, together with more negative δ13Corg and δ13Ccarb values71 (Fig. 8a).

Based on mineralogical and textural observations, the ubiquitous occurrence of Fh, together with abundant biogenic filaments in the studied mats, indicate initiation of poorly crystalline Fh precipitation by Zetaproteobacteria, with the filaments acting as nucleation sites72. Moreover, the stalk-/sheath-forming Zetaproterobacteria have been specifically linked with the LSHF mats29, similar to many Fe-Si microbial deposits reported near modern hydrothermal vents31,32,34,58,59,73, which are believed to reflect a response to changing Fe(II) and O2 redox gradients52. Besides, the widespread association of filamentous morphologies in the LSHF mats has been suggested to represent a growth strategy that enables Zetaproteobacteria to escape permanent entombment and asphyxiation by Fh/FeOx precipitates29,52. Consequently, the resulting encrustation of abandoned filaments by Fh leads to exquisite preservation of mat growth patterns (except for the NTE site where substratum alteration occurs in this area43). Moreover, nucleated Fh triggers additional autocatalytic oxidation of Fe(II) in the absence of biomass, further templating and scavenging abiotic Fh from seawater3,74.

Sequential deposition of iron oxyhydroxide phases

Fh reactivity and environmental conditions

Because of their high specific surface area and sorption capacity, poorly crystalline nm-size Fe(III)-oxyhydroxide (Fh) particles are highly reactive compared to other crystalline FeOx (e.g., goethite and magnetite)53. However, surface coating of Fh by organic molecules, silicate, phosphate, and trace metals may slow down or inhibit transformation into crystalline FeOx3,54,74,75, thereby enhancing its preservation54,76. In this regard, our data show an abundance of organic molecules (infrared-based analyses; Supplementary Fig. 3), SiO2, and P2O5 in Fh-dominated mats compared to the crystalline FeOx phases in the Y3 mats. In agreement with the accepted consensus that dissolved aqueous Si favours direct adsorption to Fh even at extremely low Si concentrations55,77, the indurated Fh-encrusted microbial filaments at LSHF host high Si concentrations sourced from active hydrothermal vents. Further, our mineralogical observations indicate that Si tends to associate intimately with poorly crystalline Fh rather than form separate Fh and Si minerals at the surface of the mats. Similar to Si-Fh mineral associations reported in other deep-sea hydrothermal vent settingse.g.,34, environmental factors such as fluid temperature and pH likely influenced the binding capacity of Fh for Si and various elements78. For example, an increase in Fh crystallite size lowers the proportion of atoms present at the mineral surface, resulting in a proportionate reduction in the mineral’s specific surface area available for adsorption and/or scavenging of dissolved trace elements54,79. No correlations were observed between the pH of the diffuse hydrothermal fluids and Fh/FeOx mineralogy at LSHF (pH ranges from 5.1 to 6.1 at CAP and 5.6 to 5.8 at Y3)42. In contrast, persistent 90 °C diffuse fluids at Y3 compared to temperatures of ~50 °C at the CAP and NTE sites42 would have promoted nucleation of slightly larger Fh particles at the former78 (Fig. 8b). This phenomenon likely decreased Fh surface area reactivity, adsorption capacity for Si, P, and other elements, and likely enhanced transformation to secondary FeOx mineral phases solely in the Y3 mats.

Transformation of primary Fh to secondary Fe(III) and Fe(II) mineral phases

Different environmental conditions (e.g., temperature of diffuse fluids) at LSHF would be expected to produce different Fh transformation products. For example at the Y3 site, stalk-forming Zetaproteobacteria and organic molecules facilitate the nucleation of biogenic Fh57. As the mats aged, thermodynamically unstable Fh eventually transformed into more crystalline FeOx phases associated with residual filamentous microbial structures, as evidenced by petrographic textures (Fig. 6). Haematite formation from Fh occurs rapidly at temperatures >50 °C or after extended reaction times80, while goethite/lepidocrocite is an intermediate mineral formed as a function of temperature, pH, and Fe(II) concentrations during the transformation of Fh to haematite74,80. As a consequence, the LSHF elevated Fe(II) concentrations41,60 would have preferentially transformed Si-free Fh into goethite and lepidocrocite3,74 (Fig. 8b), which explains the lack of haematite. This is supported by the fact that goethite may require several years/centuries (or even more) to form haematite under suitable environmental conditions, as observed in Quaternary shallow marine hydrothermal deposits associated with microbial filaments4,5,6.

We observe a unique transformation of Fh/FeOx to siderite and magnetite in the bottommost part of the Y3 mat, which could result from both biotic and abiotic processes3. Siderite and magnetite do not appear to have replaced or overgrown an earlier Fe-rich phase. Magnetite could have formed from the conversion of Fe(III)-oxyhydroxides with the addition of high amounts of hydrothermally derived Fe(II)74. However, petrographic observations and all isotopic evidence suggest that they are a product of early diagenetic Fh/FeOx transformation rather than primary growth (Fig. 8b). We use C isotope systematics to decipher the origin of both minerals24, considering the variable δ13Ccarb values could have originated from dissolved inorganic C in hydrothermal-seawater fluid mixtures because δ13C values of hydrothermal vent fluids are systematically < −1‰69. However, the lack of correlation between Eu/Eu* anomaly as a tracer of hydrothermal contribution81,82 and δ13Ccarb values (R2 = 0.03, n = 6, p < 0.75) (Supplementary Fig. 7; Supplementary Table 2) indicates that alternative mechanisms controlled the δ13Ccarb values in the mats. Furthermore, hydrothermal fluid influence cannot explain the unique presence of siderite in the bottom layer of the Y3 mats, seeing that our samples covered several sites associated with hydrothermal activity (except for the LL site).

Instead, the transition to lower δ13Ccarb, with δ13Corg depletion and decreasing TOC content deeper in the Y3 mat (Supplementary Fig. 6b; Table 2), as shown by the low δ13Ccarb and δ13Corg values of +2.0‰ and -21.4‰ in the bottom part, respectively, compared to +11.5‰ and -19.0‰ at the mat surface, respectively, point to a biotic origin71 (Fig. 8). This largely supports a coupling of early diagenetic microbial reduction of Fh/FeOx to the oxidation of residual organic C1,12,24 to produce 13C-depleted biomass and siderite, and double valence magnetite. This process is also consistent with the 3- to 18-fold reduction of P concentration in the FeOx-rich layer compared to the Fh-rich mats via the reductive release of organic-bound P into seawater and sediment pore water to sustain local primary productivity. Furthermore, the weak correlation between δ13Ccarb and δ13Corg values in the LSHF mats implies that biologically mediated organic C oxidation promoted the release of CO2 into pore waters, during which some combined with Fe(II) to form siderite and the remainder was lost to the overlying water column. The above mechanism for siderite and magnetite formation also aligns with the presence of Geospsychobacter Fe(III)-reducing bacteria and other archaeal Fe(III) reducers in the LSHF mats29.

Implication for microband formation in BIFs

Role of benthic microbial contribution to Fe-rich microband formation

Our approach provides a detailed modern understanding of interactions between benthic microbial activity, environmental conditions, and Fe-rich minerals where Fe-rich microbands reminiscent of BIFs formed solely in the Y3 mats. It has been suggested that microorganisms played a significant role in the formation of microbands in BIFs9, a proposition challenged by the low organic C content of <0.8 wt.% in BIFs1,12,24. While a mean TOC of ~5 wt.% has been extrapolated from C isotopes, magnetite and siderite mass balance studies of BIF deposits1, and suggestions made that this organic matter could have been consumed by post-depositional microbial oxidation, direct evidence remains lacking. Our data suggest that microbial-driven siderite and magnetite formation can occur at even much lower organic C content in BIFs. While it is unclear to what extent benthic mats may have contributed to BIF deposition, our observations suggest that the low TOC content in such BIFs, particularly Algoma-type BIFs, may be explained by three main mechanisms from the perspective of biological Fe cycling (Fig. 8b):

(1) Zetaproteobacteria that dominate Fe(II) oxidation in benthic marine ecosystems consistently escape encrustation to the surface of the mats to continue the mat-building process, leaving residual mineralized structures to form the framework on which the mat layers are built. The paucity of C in stalks of mat-forming Zetaproteobacteria from the Jan Mayen Vent Fields corroborates our interpretation35. Regardless of FeOM lifestyles, the residual organic C could originate from a smooth coating of polysaccharides bound to the stalks during cell growth51.

(2) The remaining residual organic matter is rapidly oxidized by a consortium of microorganisms, a mechanism supported by genome-resolved metagenomics and metatranscriptomics from the Lōʻihi mats46. Besides, Fe(III)-reducing microorganisms usually rely on fermenting microbes to produce low molecular weight organic compounds that can be more readily assimilated83 (Fig. 8a). Fermenters also produce H2 to facilitate Fe(III) reduction and oxidation of organic molecules for Fe(III) reducers84. These microbial communities tend to form a relatively little biomass compared to the large volume of Fe(III)-oxyhydroxides needed for growth because of the much lower amount of energy extractable from Fe(III)-oxyhydroxides compared to O2 as the terminal electron acceptor. We note that such processes potentially operated in all BIF types if they were associated with enough organic matter to support the development of a similar microbial consortium.

(3) Finally, as evidenced by Fig. 6, the Fe biomineralized filaments provide a large adsorptive surface area for continuous scavenging of considerable amounts of hydrothermally-derived Fe(II) in the absence of direct microbial biomass, collectively forming organic C-poor Fh precursors to FeOx. The resulting substantial increase in Fe2O3/SiO2 ratios from the filamentous microbial structures to the more crystalline FeOx, together with the unique absence of bulk Si-O-Fe elemental bonds, suggest that the gradual Fh transformation dislodges Si during the initial stage of diagenesis (Fig. 8b), offering new insights into formation mechanisms of Fe-rich/Si-poor microbands in BIFs.

Role of environmental contribution to the formation of Si-rich microbands

The exact timing and depositional mechanism of Si-rich layers in BIFs is uncertain. Si enrichment in Precambrian IFs may have originated through the precipitation of Fe-Si gels before reaching the seafloor and their subsequent dewatering and Si release during diagenesis1, while metamorphic reactions have been proposed for the BIF meso/macrobands9. It has also been suggested that the formation of Si-rich layers in Quaternary BIF analogue rocks4,5,6 was a syn-depositional feature and controlled mainly by abiotic processes involving for example, cycles of intense and paused hydrothermal/volcanic venting. It was concluded that their deposition with exquisite alternating Fe and Si bands was enabled by basin restriction from the open ocean, resulting in Si saturation and abiotic precipitation from the water column5,6. Other studies infer that biannual temperature cycles resulting in episodic abiotic silica saturation could have caused the prominent micro- to mesoband layering in BIFs85,86. Alternatively, Si precipitation could occur at, or immediately below, the sediment-water interface in mats along a saturation front with respect to amorphous silica due to the conductive cooling of the local hydrothermal environment32,38. Critical explanations about the timing of silica precipitation could come from Si isotope compositions of Si-rich minerals in mats1,4,32.

As previously discussed, at LSHF, the Fe-rich mats host Si-containing Fh marked by the absence of amorphous Si mineral phases and Si-rich bands, contrasting with other Fe-Si mat-related deposits worldwide39,59. Si concentrations of the diffuse hydrothermal fluids are >2 mM at LSHF (Supplementary Table 6), similar to the concentrations in the global oceans that deposited BIFs but much higher than the 0.07 mM of modern seawater87. However, diffuse fluid temperatures of ~50–90 °C likely inhibit massive abiotic Si precipitation in the mats, as suggested by PHREEQC thermodynamic calculations (Supplementary Table 6). It is also probable that rapid diffusive dispersal of dissolved Si to the low Si-containing modern ocean compared to Precambrian times, dilutes aqueous Si, thereby inhibiting seawater saturation with regard to silica minerals and precipitation of Si-rich bands. We, therefore, hypothesize that the subseafloor architecture beneath hydrothermal fields, which controls the fluid flow temperature toward the surface, is critical to facilitating abiotic silica precipitation and formation of Si-rich microbands in modern mats during the initial stage of diagenesis. Finally, we suggest that early diagenetic Si-rich microbands would have formed in parallel with the seawater Si precipitation responsible for BIF meso/macrobands.

Contribution of benthic mats to Algoma-type BIFs

We pose the question of whether Fe-metabolizing bacteria as part of microbial mats were associated with the deposition of Algoma-type BIFs. Our observations point to a relatively rapid and progressive disappearance of Fh-based microbial filaments and possible generation of Fe-rich BIF microbands without phototrophic involvement, although carbon from the lysed planktonic phototroph biomass can fuel Fe(III) reducers and methanogens84. Furthermore, our model highlights that magnetite crystallization predates haematite formation as the mats aged and that benthic microbial communities may be a key ingredient for (episodic) siderite-/magnetite-rich microbands in BIFs, which does not preclude existing models of magnetite precipitation in Archean BIFs (see Supplementary discussion).

Overall, we hypothesize that the diffusion efficiency of hydrothermal plumes and the relative location of hydrothermal/volcanic centres to suboxic surface waters would have strongly controlled the development of primitive benthic mats because of the FeOM need for microaerobic conditions. In the late Archean-Paleoproterozoic, these environmental conditions were to be expected where the Fe(II)/O2 chemocline intersected with the shelf13,88. Estimations for O2 concentrations suggest that dissolved [O2] of <1 µM in late Archean shallow seawater oases would have jumped to 5–30 µM in the surface ocean at the Archean-Proterozoic transition45,88. Considering that the [O2] in the range of 5-20 µM corresponds to the optimum growth of microaerophilic FeOMs, their proliferation was likely plausible after the GOE. By contrast, the growth of modern FeOMs is limited at 1-5 µM [O2], although cells are still active61. Evolving Archean-Proterozoic seawater [O2] however likely helped shape the pace of FeOM evolution47, suggesting that the primitive FeOMs could have evolved and flourished in the presence of nanomolar [O2] and, thus, contributed to the deposition of BIFs near shallower submarine volcanoes and hydrothermal vent fields to form Algoma-type BIFs. Finally, microbanding patterns formed in the LSHF mat material may be comparable to the Fe-rich microbands of the ~2.5 Ga-old Superior-type BIFs89,90. It is, thus, conceivable that similar microband-forming processes could have been operating on the continental shelf if microaerobic Fe(II)-oxidizing conditions were rife, coupled with pulsed delivery of Fe(II)-rich deep-sea hydrothermal vent fluids to shelf environments where Superior-type BIFs formed.

Conclusion

We follow multiple lines of petrographic, mineralogical, and isotopic evidence to show that the development of C-poor Fe-rich BIF microbands could have been controlled by benthic mat-forming microbial processes. Our model shows that these processes begin with microbial oxidation of dissolved hydrothermal Fe(II) by Zetaproteobacteria, where microbial filaments provide a large surface of extensive nucleation and removal of dissolved Fe(II) from seawater in the presence or absence of microbial cells. This process is then followed by gradual biotic and abiotic transformation of poorly crystalline Fh into crystalline FeOx minerals as mats aged, during which the Fh-encrusted microbial filaments systematically disappear. Particularly, during early diagenesis, the coupling of sequential microbial reduction of Fh/FeOx to the oxidation of residual organic C by dissimilatory Fe(III)-reducing microorganisms leads to the transformation of Fh/FeOx to secondary crystalline siderite and magnetite. It is suggested that this process points to a key mechanism explaining the low organic C content of BIFs and helps to constrain similar isotopic signatures in BIFs as biomarkers. We suggest that these mechanisms may have been essential to the formation of BIF microbands, particularly in deposits formed in submarine hydrothermal vent fields and continental shelf environments where conducive Fe(II)/O2 gradients developed. We further propose that the striking absence of Si-rich bands at LSHF can be related to the enduring dominance of high temperature ~50-90 °C diffuse fluids and dispersal of Si to the low Si-containing open ocean, which inhibits the supersaturation state required to drive Si band formation. Importantly, the present study sheds new light on the biotic and abiotic mechanisms by which Fe-rich marine rocks with organic C-poor and Fe-rich microbanding form in the modern ocean and helps explain deep-time observations in which biological processes are difficult to reconstruct.

Methods

Sampling and sample preparation

Mat morphology and their location are described in Table 148,49,50. The Fe-bearing microbial mats were collected from Capelinhos (CAP), North Tour Eiffel (NTE), South Isabel (SI), West Sintra (WS), Y3, and Lava Lake (LL) sites during the annual MoMARSAT EMSO-Azores Observatory maintenance cruises. Two types of microbial mats were sampled, including flocculent mats at WS, SI, and LL and crumbly mats at CAP, NTE, and Y3 sites between 2021 and 202348,49,50. The flocculent mats were easily disturbed and resuspended during sample collection. In situ sample collection and on-board sample processing were conducted, as previously described29,41. A fraction of each Fe-rich mat was milled into a fine powder in agate mortar for whole-rock mineralogical and geochemical analyses. Thin sections of CAP, NTE, and Y3 mats were made using standard polishing procedures at GET laboratory, University of Toulouse (France). Thin sections were investigated under transmitted and reflected lights using an Olympus BX51 microscope combined with a Olympus SC30 microscope camera.

Diffuse hydrothermal fluids sampling and analysis

Diffuse fluids are limited to the first few dozen metres away from the high temperature chimneys. The diffuse hydrothermal fluid samples of CAP, NTE, and Y3 sites were collected during the MoMARSAT cruise in June 202242,49. Prior to fluid sampling using titanium gas-tight syringe with a fennel-like snorkel, the temperature of diffuse hydrothermal fluids was measured in situ using the temperature sensor of the HOV Nautile submersible. The full procedure for diffuse fluid sampling is described in detail in Wheeler et al.43. Aboard the research vessel, hydrothermal fluids were extracted from the fluid sampler, filtered with 0.22-μm Millipore filters and processed on board as described in Chavagnac et al.43,60. Aliquots were transferred in acid-cleaned HDPE bottles for a first set of analyses, i.e. density (refractometry), dissolved Fe concentrations (photometry), pH, Eh, salinity, total dissolved salts, dissolved H2S concentrations, and conductivity (electrodes).

Mineralogical analysis

X-ray diffraction analysis of Mn-poor bulk powder was undertaken with a Panalytical Empyrean diffractometer (CoKα radiation) operating at 35 kV and 45 mA. To determine the mineralogy of Fe- and Mn-rich samples, XRD patterns were acquired with a WAXS Panalytical Empyrean diffractometer (MoKα radiation) operating at 60 kV and 40 mA. In the latter case, a small fraction of bulk powders was mounted in a glass capillary tube of 0.5 mm diameter. Analysis of whole-rock powder samples was performed over angular ranges of 6–76° (CoKα) and 2–30° (MoKα) 2θ and a step size of 0.022° 2θ per 1 s at LPCNO, University of Toulouse III. Peak indexing was performed using Bruker Eva software and mineral identification was carried out by comparing XRD patterns with reference data53,54.

Short wavelength infrared (SWIR) and Fourier transformed infrared (FTIR) spectroscopies were carried out at IC2MP, University of Poitiers (France), to investigate further mineralogy, structure, and bonding of Fe oxides. SWIR reflectance spectra of bulk rock powders were obtained with a ASD TerraSpec 4 Standards-Resfield spectrometer within the 500–2500 nm range. The spectrometer is equipped with a contact probe including a white light source and three detectors (one Si detector for the 350–1000 nm range with 3 nm resolution and two InGaAs detector for the 1000-2500 nm range with 10 nm resolution). An average of 50 scans was measured on each sample. The Indico Pro software was used to collect the SWIR spectra. FTIR spectra were collected in transmission mode using a Nicolet iS50 FTIR spectrometer equipped with an Ever-Glo source, a potassium bromide (KBr) beamsplitter, and a DTGS KBr detector. Prior to FTIR analyses, KBr pellets, consisting of a mixture of 1 mg of sample and 149 mg of KBr, were prepared. The preparation was pressed for 5 min at 8 kbar to produce pellets and dried overnight in an oven at 120 °C to remove adsorbed water. The FTIR spectra were recorded in the 4000 to 400 cm−1 region (middle infrared) and acquired from a mean of 100 scans at a resolution of 4 cm−1. The study of stretching and bending vibrations using FTIR spectroscopy is of deep interest for characterizing adsorbed Si onto Fe oxide surface55,56.

Because it was difficult to detect poorly crystalline ferrihydrite in the presence of crystalline mineral phases by XRD, the SI, NTE, and Y3 microbial mats were investigated by 57Fe Mössbauer spectroscopy at the Charles Gerhardt Institute, University of Montpellier (France) to characterize Fe-bearing minerals, but also the distribution and redox state of Fe. Analyses were carried out on ca. 300 mg of rock powder loaded in a 30-mm diameter plastic holder irradiated by a 57Co source in rhodium metal. The Mössbauer spectra were acquired using a spectrometer in the constant acceleration mode over the velocity range of −4 to 4 mm/s at room temperature ( ~ 22 °C). The velocity scale was calibrated using an α-Fe foil. The spectrometer was operated using a triangular velocity waveform for transmission geometry. Gamma rays were detected with a NaI scintillation detector (detection limit of 2%). The Mössbauer spectra were fitted by Lorentzian lines to determine the spectral parameters (IS: isomer shift; QS: electric quadrupole splitting; LW: full line width at half maximum; A: relative resonance areas).

The thin sections were carbon-coated and imaged in BSE mode with a Tescan Vega 4 LMU scanning electron microscope (SEM) equipped with an energy dispersive spectroscopy (EDS) system at GET laboratory. Detailed observations of filamentous structures were conducted on platinum-coated thin sections using a JEOL JSM-6700F field emission SEM at Centre Castaing, University of Toulouse. The SEM-EDS analyses were run at a 10-30 kV accelerating voltage and working distances of 7.9 and 14.7 mm.

Bulk elemental analysis

Whole-rock geochemical analyses of major and trace elements were performed at the clean lab of the GET laboratory. Approximatly 100 mg of bulk rock powders were dissolved using double-distilled reagent acids in closed screw-top Teflon vessels (Savillex). First, samples were digested with a concentrated HNO3-HCl mixture using a MARS 6 microwave digestion system at 180 °C for 15 min. After evaporation of the solution to dryness on hot plate at 70 °C, samples were dissolved in suprapur HF at 70 °C for 2 h to break down silicates. Major (except Si due to the evaporation of Si-F complexes) and trace elements were measured by ICP-MS using a Thermo Scientific iCAP triple quadrupole instrument, with a precision better than 2%. Calibrations were made with multielemental solutions diluted to different concentrations. The reference material CRPG Mica-Fe was used as a control.

Furthermore, the Si concentration of bulk rock samples were determined by X-Ray Fluorescence (XRF) on glass fusion discs using a PANalytical AXIOS Max spectrometer equipped with an SST-max X-ray tube at RRXG platform, University of Montpellier. The glass fusion discs were prepared by mixing 0.6 g of rock powder with 6 g of lithium tetraborate. We analysed 9 reference geological standards for internal calibration and one external reference standard (IF-G). Loss of ignition (LOI) was determined by heating the rock powders in a muffle furnace at 900 °C for 4 h.

Electron probe micro-analyser

Electron probe micro-analyses (EPMA) were carried out on carbon-coated thin sections using a CAMECA SXFive electron microprobe equipped with five wavelength dispersive spectrometers at Centre Castaing, University of Toulouse III. Analytical conditions of EPMA were performed using a ~1 µm focused beam at 15 kV and 20 nA. The counting time was set to 10 s for all elements. Si, P, and Fe were measured with a focused electron beam and step size of 0.5 μm during quantitative mapping. For silicate EPMA, the instrument was calibrated using reference material standards, including albite for Na, MgO for Mg, Al2O3 for Al, wollastonite for Si and Ca, graftonite for P, sanidine for K, Cr2O3 for Cr, MnTiO3 for Mn and Ti, haematite for Fe, and SrSO4 for S. For sulfur EPMA, we used the following standards: chalcopyrite for S, Fe, and Cu, mispickel for As, and native Mo for Mo.

Stable isotope analysis

Representative microbial mat samples were analysed for their C isotope ratios of carbonates (δ13Ccarb) and organic matter (δ13Corg) and for their N isotope compositions (δ15N) at Cardiff University. The δ13Ccarb analysis was conducted on a Thermo Delta V Advantage mass spectrometer, coupled to a Thermo gasbench II for head-space sampling of carbonate samples. The powdered samples were first weighed into septum vials, which were flushed with helium and then acidified manually with 99% orthophosphoric acid. Samples were then left to react for 24 h at 60 °C to ensure complete reaction of dolomite and siderite. The precision of an in-house Carrara marble standard calibrated against NBS19, expressed as the standard deviation of repeated analyses, resulted in a value 0.07‰ for δ13Ccarb. For total organic content (TOC), total N (TN), associated δ13Corg and δ15N compositions, powdered samples acidified in 25% HCl were left to react for two days to eliminate inorganic C. Residues were rinsed three times with ultrapure deionize water to remove remaining acid, dried and weighed into tin capsules. Samples were then measured on a Thermo Delta V Advantage mass spectrometer with Thermo Flash EA and Conflo III. Samples were analysed in combination with three in-house standards composed of a lab-grade caffeine with δ13C = -33.30‰ and two commercial collagen food supplements (MarCol) with a δ13C of -16.20‰ and an MCF with a δ13C value of -22.36‰. The in-house standards were calibrated against the international standards IAEA-CH6 and IAEA-600. Data precision, expressed as the standard deviation for repeated analyses of the independent standard MCF, was 0.2‰ (1 SD) for δ13Corg and 0.1‰ (1 SD) for δ15N. The results are reported in the standard δ-notation in ‰ on the Vienna-Pee Dee Belemnite (VPDB) scale for C and air for N.

PHREEQC geochemical modelling

Thermodynamic calculation of the amorphous silica solubility as a function of temperature was performed with the batch reaction option of PHREEQC software package developed by USGS91. In addition, the effect of pH was considered and modelled. The calculation of the maximum concentration of dissolved Si required to reach thermodynamic equilibrium between amorphous silica and a standard seawater solution92 was calculated using the llnl.dat data base93. Temperatures were set between 25 °C and 90 °C to represent environmental temperature bathing the microbial mats and the effect of salinity on activity coefficients was taken into account. According to the thermodynamic calculation, the main factor controlling the solubility of Si is temperature rather than pH.

Authors contributions

J.A., E.C.F., and V.C. designed the study. V.C., C.R., C.D., and A.C. collected the samples and, analysed diffuse hydrothermal fluids. J.A., T.D., and F.B. conducted mineralogical analyses. F.P., M.H, A.N., conducted geochemical analyses. C.D. performed PHREEQC modelling. C.R. and V.C. carried out funding acquisition and supervised the project. J.A. drafted the original manuscript and made the figures. The draft was reviewed before submission by E.C.F., C.D., T.D., F.P, F.B, C.R., and V.C.