Banded Iron Travertines at the Ilia Hot Spring (Greece): An interplay of biotic and abiotic factors leading to a modern Banded Iron Formation analogue?

A hot spring at Ilia in the Greek Island of Euboea precipitates iron‐rich travertine at an ore‐grade concentration (up to 35.3 wt% Fe). This hydrothermal chemical sediment system deposits bands of iron oxyhydroxides (ferrihydrite), millimetres to centimetres thick, alternating with calcium carbonate‐dominated layers, creating “Banded Iron Travertine” (BIT). The ferrihydrite laminae display a dendritic texture formed of spherical nodules often covering filaments identified as bacterial stalks of Zetaproteobacteria. These microaerophilic iron‐oxidizing bacteria were identified by their 16S rRNA gene sequences in ferrihydrite‐enriched samples from areas under high water flow. They were missing in the aragonite/calcite‐dominated samples exhibiting features of aerial exposure and cyanobacteria instead. These characteristics, and the relative depletion in Fe‐rich layers of redox‐sensitive elements like Mn and Ce, as well as the presence of halite in Ca‐rich layers, suggest that the bands form by successive changes in hydrothermal flow. This allowed microaerophilic iron oxidation to form Fe‐rich layers, while Ca‐rich bands precipitated when the hydrothermal water had time to equilibrate with the atmosphere. This sea water‐dominated hydrothermal system is enriched in reduced iron and rapidly precipitating carbonates and ferrihydrite in the form of bands, having similarities to “Banded Iron Formation” (BIF). BIF represents archives of Earth's primitive biogeochemistry although the combined abiotic and biotic processes that have likely led to their formation are not fully resolved. Diagenesis and metamorphism have a strong imprint on BIF. Thus, continuous efforts are pursued to identify modern analogues that could help unravel their formation. Although carbonate is not a common feature of BIFs, Ilia system provides an interesting analogue for their depositional processes and potential microbial–mineral associations they may have hosted. It also presents pre‐diagenesis facies association and mineralogy that could bring new clues for unravelling BIF modes of formation and the salient biogeochemical conditions characteristic of their original depositional environment.


| INTRODUCTION
Among the most important iron ore deposits on Earth are the spectacular and enigmatic Banded Iron Formations (BIFs). Despite extensive studies, their modes of depositions have not been fully resolved; although one of the prevalent models is that BIFs are thought to be the result of the reaction between hydrothermal reduced iron, and photosynthetically produced oxygen. Microbes are generally recognized as important contributors to the specific sedimentological conditions of these deposits, both regarding the processes of primary iron oxidation and secondary iron reduction (Posth, Konhauser, & Kappler, 2013). Since BIFs likely hold information on some of the earliest forms of life on Earth, they are extensively investigated and analogues for their mode of formation are thoroughly searched for.
Active hydrothermal iron-rich deposits are among the targets of this search. They can be found in terrestrial environments, that is, iron-rich thermogenic travertines (Kanellopoulos, Mitropoulos, Valsami-Jones, & Voudouris, 2017;Takashima & Kano, 2005;Takashima, Okumura, Nishida, Koike, & Kano, 2011) and in submarine hydrothermal environments (Bortoluzzi et al., 2017;Kilias et al., 2013;Pichler & Veizer, 1999). In these systems, the presence of microbial activity is most often identified at redox interfaces and has been suggested to play an active role in the formation of iron minerals such as iron oxides (Kappler & Newman, 2004;Konhauser, 1998;Widdel et al., 1993). Hot springs are ideal environments for iron oxidation, since the hydrothermal fluid is often enriched in Fe(II) and free of oxygen. In such conditions, a steep redox gradient develops in the water immediately adjacent to the spring vent through uptake of atmospheric O 2 (Takashima, Kano, Naganuma, & Tazaki, 2008). In near neutral pH hydrothermal springs, the Fe-rich deposits consist primarily of ferric hydroxides (commonly as ferrihydrite). These ferric hydroxides precipitate readily abiotically in oxygenated environments or their precipitation can be microbially mediated and influenced (Chan, Fakra, Emerson, Fleming, & Edwards, 2011;Konhauser, 1998). When microbes are involved, rates of iron mineralization can increase four-fold (Kasama & Murakami, 2001). These processes generally occur at redox interfaces, where iron-oxidizing bacteria are commonly found and where microaerophilic conditions dominate (Konhauser, 1998). Such conditions have been suggested in the formation of BIFs, and iron-oxidizing bacteria as associated with the formation of iron bands in Precambrian BIFs (Chaudhuri, Lack, & Coates, 2001;Holm, 1987;Konhauser et al., 2002). However, the study of iron mineralization in hydrothermal systems is often hampered by the poor accessibility of active sites, and the instable characteristics of ferrihydrite therein.
This study characterizes the iron-rich thermogenic travertines of Ilia, on the Greek island of Euboea. The depositional system hosts almost pure ferrihydrite lamina which, together with fast precipitation rates, high iron enrichment and specific hydrothermal chemistry (Kanellopoulos et al., 2017) provides a unique setting for the study of laminated iron chemical sediments. The purpose of the work presented in this paper is to understand the depositional processes leading to the formation of laminated iron-rich deposits in Ilia and to further link them to BIF depositional models. This study documents the geochemistry of the Fe-rich hydrothermal fluids from the Ilia area, the macrofacies and microfacies, the petrographical characteristics, mineralogical and geochemical composition of the newly formed iron-rich thermogenic travertine, and the microbial communities associated with the different facies of this deposit. The Fe-rich travertine is compared with active hydrothermal systems elsewhere and all of the available data synthesized to propose a model for Banded Iron Travertine (BIT) deposition, with possible analogy to the formation of BIFs. The goal is to provide a depositional model associated to an easily accessible and dynamic system precipitating banded iron-rich chemical sediments.

| GEOLOGICAL SETTING
Ilia is located in the north-western part of Euboea Island, Greece ( Figure 1). The geological setting of the region is unique since it is located at the western extremity of the North Anatolian Fault and is in a back-arc position with respect to the South Aegean Active Volcanic Arc (Pe-Piper & Piper, 2002;Ring, Glodny, Will, & Thomson, 2010;Vött, 2007). It is one of the most neo-tectonically active areas in Greece dominated by extensional tectonics similarly to the rest of the Aegean Sea.
The hydrothermal fluids of northern Euboea's hot springs have near neutral pH and a sodium-chloride composition. They are similar in their chemical characteristics and are classified as deep geothermal fluids, with volcanic origin affinities (Kanellopoulos et al., 2017). Based on geochemical and isotopic studies (Dotsika, 2015;Kanellopoulos, 2011;Kanellopoulos et al., 2017;Shimizu, Sumino, Nagao, Notsu, & Mitropoulos, 2005), their chemistry is controlled mainly by contribution of three components: (a) high sea water participation, (b) deeper magmatic fluids, and (c) the local bedrock. The chemical composition of the thermogenic travertine, together with drilling data (Gkioni-Stavropoulou, 1998;Hatzis, Kavouridis, Bakalopoulos, & Xenakis, 2008;Kanellopoulos et al., 2017), strongly suggests that the source of calcium for the extensive travertine deposits of Euboea is the dissolution by geothermal fluids of marble/limestone located below the metamorphic sequence of Aedipsos and Ilia.

| Hot water samples
Hydrothermal fluid samples from the hot spring/artesian borehole were collected from the mouth of the pipe . Physicochemical parameters such as temperature, pH, electrical conductivity (E.C.), and total dissolved solids (T.D.S.) were measured in situ, using portable probes. The pH-meter was calibrated with standard buffer solutions of pH 4.0 and 7.0 before any measurements were taken. The pH measurement error, including accuracy and reproducibility, is better than ±0.05 pH units. Temperature was measured with the probe connected to the pH-meter and the error is estimated to be less than ±0.3°C. Moreover, a thermal imaging camera was used in order to present through thermal images the temperature changes at the sampling site. Thermal imaging camera is a type of thermographic camera, which is reading infrared radiation as visible light and allows us to see areas of heat through water or solid rock.
Water was collected in polyethylene bottles for laboratory analyses. Aliquots of 250 ml intended for the determination of trace elements were acidified to pH < 2 with analytical grade HNO 3 (Suprapur 65%). All samples were stored in a portable cooler containing ice packs, transported to the laboratory and refrigerated at 4°C until analysis. One-litre aliquots (non-acidified) were filtered in the laboratory before major ion determinations.
The major ion concentrations were measured using spectrophotometry, titration, and Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES; Perkin

| Travertine samples
The hydrothermal spring of Ilia vents from an artesian borehole accessible from the road that run along the beach. The water then flows along a 3-5 m long pipe and out onto the beach (Figure 2). Samples of thermogenic travertine were collected from the top part of the borehole, from the venting point of the pipe and from the deposits at the beach ( Figure 2). The mineralogical composition of the main mineral phases in these travertines was identified with X-Ray Diffraction (XRD) using a Siemens Model 5005 Xray diffractometer, Cu Ka radiation at 40 kV, 40 nA, 0.020°step size, and 1.0 sec step time. The XRD patterns were evaluated using the EVA program of the Siemens DIFFRACplus and the D5005 software package. The minor mineral phases were identified by Scanning Electron Microscopy (SEM) and Energy Dispersive Spectroscopy (EDS). The SEM-EDS analyses were carried out using a Jeol JSM 5600 SEM instrument, equipped with an Oxford ISIS 300 micro analytical device.
The travertine samples were analysed for whole rock geochemistry after drying and pulverizing in an agate mortar and mill to less than 0.075 mm. The samples were digested with a mixture of HCl-HNO 3 -HF acids and analysed with Inductively Coupled Plasma-Optical Emission Spectroscopy method (ICP-OES; Perkin Elmer Optima 5300 DV) for Ca, Na, P, S, Si, and by Inductively Coupled Plasma-Mass Spectrometry method (ICP-MS; Perkin Elmer Sciex Elan 6100) for a series of trace elements, including rare earth elements (REE). Analytical data quality was tested with internal standards, certified and in-house reference materials in random positions within the analytical batch and by blank analysis and duplicate analysis of a proportion of the samples (Ramsey, Thompson, & Banerjee, 1987).

| Molecular biology
Samples for microbiological studies were collected at the vent of the artesian borehole (Ilia Drill for ID-labelled samples, Figure 3a,b) and at the pipe outlet (Ilia Pipe for IP-labelled samples, Figure 3c,d). They experienced different environmental conditions at time of sampling, which are detailed in Table 1. Hard rock samples were taken using a chisel and a hammer cleaned and sterilized with ethanol. Soft material was sampled using a sterile scalpel. All samples were stored in 50 mL falcon tubes or zip-lock bags and kept at −20°C until further processing. Hard rock samples were then crushed under clean conditions in a sterilized agate mortar. Two grams of processed sample was used for DNA extraction using the MOBIO power soil kit. Triplicate DNA amplification was realized with a total of~10 ng of DNA per triplicate using universal primer 515F (5′-GTGYCAGCMGCCGCGGTA-3′) and 909R (5′-CCCCGYCAATTCMTTTRAGT-3′) for the V4-V5 hypervariable region of the 16S rRNA gene (Wang & Qian, 2009), with indexes integrated following the dual-indexing procedure described by Kozich, Westcott, Baxter, Highlander, and Schloss (2013). Product triplicates were then pooled, quantified using Picogreen assay (Life Technologies) and pooled equimolarly. The final pool was concentrated using a SpeedVac Plus SC110A Savant and purified with CleanNA beads (Moka science).
Sequencing was performed by Fasteris (Geneva, Switzerland) on an Illumina Miseq with 2 × 250 cycles, with settings of 7.5 Gb yield (including PhiX), an error rate of 2.5% and Q30 at 75%. The analysis yielded 5.3 Gb of sequences with an error rate within quality specifications. Adapters were removed using trimmomatic (Bolger, Lohse, & Usadel, 2014), paired-ends reads were joined using eautils (Aronesty, 2011), quality checks were performed using FastQC, and samples were demultiplexed using the Fasteris in-house script. Fasteris also did all the downstream work described above.
DNA-derived 16S rRNA gene sequences were processed using Mothur (Schloss et al., 2009). Samples were dereplicated, aligned, and filtered by length. Chimeras were removed using uchime (Edgar, Haas, Clemente, Quince, & Knight, 2011), and taxonomic affiliation was then realized using the method of Wang, Garrity, Tiedje, and Cole (2007) at a cutoff of 80% against the Silva SSU database 123 (Quast et al., 2013). Operational Taxonomic Units (OTU) were then defined at a 97% similarity and used for similarity analysis. Random subsampling was realized based on the smallest number of obtained sequences in one sample after singleton removal (see Tables S1 and S2 for more details). A matrix was built at the phylum level (excluding phyla that accounted for less than 1% of the total community) and Principal Component Analysis (PCA) was performed using Past (Hammer, Harper, & Ryan, 2001). Sequences were submitted to NCBI and can be found under the accession number MH385389 -MH388018.

| Water chemistry
Based on repeated measurements conducted over the sampling period and the previous years (Kanellopoulos et al., 2017), the physicochemical parameters of the hydrothermal fluid have only limited variation, with temperature varying from 60.9 to 63.7°C and acidic pH values varying from 5.88 to 6.45 (Table 2, Figure 2). Table 2 shows the physicochemical and chemical parameters analysed in situ and in the laboratory. Chloride is the dominant anion of the Ilia fluid (1.28% Cl). Based  (Table 2), the fluid is characterized by Na-Cl hydrochemical type. Specific enrichment exists in major elements such as Fe (13.3 mg/L) and Ca (1,570 mg/L) and in trace elements such as As (up to 250 μg/L), Mn (500 μg/L), Ni (up to 80 μg/L), and ammonium cation (up to 1.9 mg/L ΝΗ 4 + , Table 2). It also shows concentrations of Sr of 24 mg/L, B of 9.7 mg/L, F of 3.8 mg/L, and Li of 3.4 mg/L.

| Morphology of the travertine deposit
Ilia's hot water artesian borehole deposits thermogenic travertine at the discharge point, along the pipe, at the mouth of the pipe and at its outflow on the beach, building up the sides of a keel that slowly rises in height with time (

| Mineralogical study and microfacies of the travertine deposit
Samples from Fe-rich and Ca-rich components were separated and analysed by XRD ( Figure 5a). Based on these analyses, the Fe-rich component is composed mainly of two-line ferrihydrite and to a lesser extent of aragonite and calcite (Figure 5b), as previously described by Kanellopoulos (2012) and Kanellopoulos et al. (2017). The Ca-rich component is composed mainly of aragonite and calcite with a minor contribution from ferrihydrite and halite ( Figure 5c).
Under the optical microscope, the Fe-rich and Ca-rich bands are markedly distinct ( Figure 6a). The Ca-rich bands are mainly laminated (Figure 6b), consisting of needle-like crystals of aragonite, growing normal to the substrate (Figure 6c). The crystals are elongated parallel to the c-axis, exhibiting straight extinction. The substrate is usually a thin lamina of silt-sized micritic aggregates ( Figure 6c) or another Fe-rich lamina. In several cases, parallelogramshaped areas of aragonite, with rounded edges, are observed developing within Fe-rich masses (Figure 6d), suggesting the complex 3-D spatial spreading of the two main components (i.e., Fe-rich and Ca-rich components).
Under the optical microscope, the Fe-rich components mainly consist of dark red to brown ferrihydrite. Combined optical microscope observations and SEM images show two ferrihydrite facies. The first type consists of laterally aligned shrub-like structures extending and branching upwards and forming laminae (Figures 6e-g and 7a,b).
Br (μg/L) 40,300 38,100 --100,000 -6,800 87,300 Fe (μg/L) 11,300 13,300 12,100 6,800 314,500 24.2 (%) 1,720 52 Typically, they are elongated, dense masses consisting of ferrihydrite micritic clumps, usually with spherical nodules creating "branches" and dendritic fabric (Figures 6e,f and 7f-h). Based on the classification scheme of Chafetz and Guidry (1999), they could be characterized as woody shrub or bush (i.e., bacterial shrubs). These shrubs are usually developed above aragonitic laminae (Figure 6e). In some cases, dark-tinted axes (filament-like), potentially consisting of organic matter, can be observed encased within the ferrihydrite shrubs (Figure 6g). The filaments extend and branch upwards, over a distance corresponding to the thickness of the lamina. Revealed under SEM, the iron-rich components of the studied samples are commonly composed of fine-scale laminae of the iron-rich mineral phase (ferrihydrite) alternating with calcium-rich (aragonite) laminae over a few tens of micrometres (Figure 7a).  Ferrihydrite has been observed in several cases to fill empty space between ferrihydrite shrubs, or between aragonite crystals (Figure 7e) creating very dense laminae. However, in several cases empty spaces between the dendrite aggregate were observed inside the laminae (Figure 7b). Additionally, ferrihydrite was identified in a second type of facies consisting of dense laminae, with very limited porosity (Figures 6d,h and 7c-e). In this case, no crystalline habit or microfacies of the ferrihydrite could be identified. Although, the ferrihydrite laminae followed the geometry dictated by the substrate (i.e., Ca-rich laminae or bedrock).
Minor amounts of sulphides such as pyrite, As-rich pyrite, arsenopyrite, stibnite, galena, Fe-rich sphalerite, chalcopyrite and oxides, and non-metallic mineral phases such as barite, fluorite, and some REE-bearing phases, can be found in the studied samples. All of these minerals occur as very small crystals (few μm in size), in accordance to Kanellopoulos et al. (2017).

| Travertine geochemistry
Samples from Fe-rich and Ca-rich components were separated to analyse their bulk geochemical composition ( Table 3). The Fe-rich component of the Ilia travertine is enriched compared to the Ca-component (and typical travertines) in Fe (35.3%), As (2.6%), and Na (4%) for major elements and in Y (323 mg/kg), Nd (96 mg/kg), Dy (63 mg/kg) for trace and rare earth elements (REE). In contrast, the Ca-rich components of the Ilia travertine are less enriched in all of these elements (11% Fe, 931 mg/kg As, 0.08% Na, inc. all the REE), but have higher concentrations of certain elements like Mn (330 mg/kg), S (0.27%), and other metals such as Co (0.24%), Cr (2.68%), and Ni (1.58%). Both components are presenting comparable concentrations to elements such as Cu (Fe-rich component: Normalization of the whole rock chemical data to the average composition of Upper Continental Crust (UCC, Rudnick & Gao, 2003) and Post-Archaean Australian Shale (PAAS, Taylor & McLennan, 1985;Figure 8) indicates that both laminae are enriched in elements such as As, Fe, and S. The Fe-rich component in particular is enriched 5,440 times in As, compared to UCC abundances and almost 10 times in Fe, compared to UCC and PAAS abundances. Additionally, the Fe-rich component shows a clear enrichment in REE compared to UCC and PAAS, by more than 10 times (Figure 8). The two studied components show similarities with respect to light REE depletion, but the Ca-rich component has small positive anomalies in Ce and Y, which are not present in the Fe-rich component.

| Analysis of microbial communities based on 16S rRNA gene sequences
Microbial communities differ from each other based on sample lithology and exposure to water/air (Figure 9a). Bacteria-associated reads are dominant in all samples, but samples ID1 and ID4 have 19% and 26% of archaeal reads respectively, and IP2 and IP3 reach 44% and 28% (Figure 9a).
Clades involved in iron reduction and oxidation are abundantly represented in the samples of the Ilia travertine. Members of the iron-oxidizing Zetaproteobacteria represent 3% of the community in ID2 and ID3. This clade is even more abundant in the Ilia pipe samples, where they correspond to 10% and 7% of IP2 and IP3 microbial communities, respectively. The presence of unclassified Rhodobacteracea may also be related to phototrophic iron oxidation, but this is not certain.

| Hydrothermal fluid
The hot spring/artesian borehole of Ilia degases mainly CO 2 (D'Alessandro, Brusca, Kyriakopoulos, Bellomo, & Calabrese, 2014) and deposits Fe±As -rich thermogenic travertine (Kanellopoulos, 2011(Kanellopoulos, , 2012Kanellopoulos et al., 2017). It is part of a large single hydrothermal system, elements in their native form, such as Fe, Pb, Cu, (b) the presence of native alloys such as Au±Cu-Ag and the enrichment of metalloids such as As, Sb, and (c) the abundance of REE, suggested that the system has a magmatic contribution. As a result, metals and metalloids were mainly derived from magmatic fluids, which after mixing with heated sea waters deposited native elements, sulphide mineralization at depth, and As-Fe-rich travertines in the surface (Kanellopoulos et al., 2017(Kanellopoulos et al., , 2018. The repeated measurements conducted during the last years (including wet and dry seasons, Kanellopoulos et al., 2017) have shown that the physicochemical parameters of the hydrothermal fluid are relatively stable with temperatures varying only from 60.9 to 63.7°C and pH values from 5.9 to 6.45 (Table 1, Figure 2). The results reported here are in line with previously published values. Terrestrial thermogenic Fe-rich travertines with ferrihydrite have been reported in Shionaha (Takashima & Kano, 2005) and Okuoku-Hachikurou, Japan (Takashima et al., 2008(Takashima et al., , 2011. Additionally, the presence of ferrihydrite in hydrothermal environments has also commonly been reported in submarine vents (chimneys and hydrothermal sediments) such as the Basiluzzo Islet, Italy (Bortoluzzi et al., 2017), Kolumbo, Greece (Kilias et al., 2013), and Tutum Bay in Papua New Guinea (Pichler & Veizer, 1999). By comparing Ilia's hydrothermal fluid with these occurrences (Table 2), the studied fluids show similar pH values (slightly acidic: 5-6.6). The temperatures of the other known terrestrial ferrihydrite depositing hot springs in Japan are 20-35°C lower, potentially allowing the development of non-thermophilic organisms. Regarding iron, Ilia's hydrothermal fluid shows higher concentrations than the terrestrial hot springs of Japan (Takashima & Kano, 2005;Takashima et al., 2008) and the hydrothermal vent from Tutum Bay (Papua New Guinea; Pichler & Veizer, 2004). However, iron concentrations are lower when compared to the Basiluzzo Islet (Italy; Bortoluzzi et al., 2017) and Kolumbo (Greece; Kilias et al., 2013) submarine vents. Hence, Ilia hot spring shares some similar physicochemical characteristics with the fluids of the above-mentioned areas. Additionally, based on these characteristics Ilia can be ranked among the most iron enriched terrestrial hot springs worldwide, with near neutral pH.

| Iron-rich travertine composition
Based on their strong differences such as their colour and also their facies and chemical composition, it is clear that the Fe-rich and Ca-rich laminae of the travertine are formed under very different conditions. In particular, spatial discrimination of Fe-deposits from Ca-deposits is observed, with a dark colour (representative of iron oxides of the Ferich laminae) only present when there is hot water flowing at a relatively high speed. When the flow rate decreases (away from the vent) or where there is no water flow (on the sides of the pipe for example), the orange colour, representative of the Ca-rich component, dominates. These observations strongly suggest that the differential deposits of the Fe-rich and Ca-rich component are controlled by variations in the flow rate of hot water venting from the spring. As a result, it is expected that the laminae are controlled by these changes.
The elements As, Sc, and Y are clearly enriched in the Fe-rich component compared to the Ca-rich component. The strong binding potential of metal oxides towards arsenic could explain this difference in the ferrihydrite-rich component (Smedley & Kinniburgh, 2002). In contrast, the Ca-rich component of the Ilia travertine clearly has higher concentrations of Mn (330 mg/kg) compared to the Fe-rich component. This may be related to the fast oxidation of Mn 2+ in changing redox conditions, which may have varied between the two different phases. Metals such as Co and Ni follow the same trend. Similarly, the positive Ce anomaly observed in the Ca-rich component could be linked to the sensitivity of Ce to oxidative conditions (Bau & Koschinsky, 2009). Hence, enrichment is observed in elements that become insoluble when oxidized in the Carich component, potentially pointing towards changes in redox conditions during the precipitation of this phase.

| Microbial communities of the travertine deposits of Ilia
Communities of the thermogenic travertines of Ilia vary with changes in water flow, distance to the vent, aerial exposure, and natural redox conditions (Table 1, Figure 9a). The two samples that experienced the highest water flow (ID2 and IP2) are marked by an abundance of thermophilic organism of the Aquificae phylum. Sequences of this phylum are uniquely attributed to the family Hydrogenothermaceae, which consists of chemolithoautotrophic microaerophilic thermophiles (Takai & Nakagawa, 2014). The other phyla consist mainly of unclassified sequences, except for Zetaproteobacteria, which are attributed to the Fe-oxidizers of Mariprofundaceae family (Emerson, Fleming, & McBeth, 2010). Interestingly, these Fe-oxidizers are characteristic of marine environments like the Loihi seamount (Emerson & Moyer, 2012), Santorini marine hydrothermal springs (Handley, Boothman, Mills, Pancost, & Lloyd, 2010), or of Basiluzzo Islet in the Aeolian islands (Bortoluzzi et al., 2017). Communities obtained from IP3 and ID3 share similarities allowing a continuum to be drawn between the environments of ID2 and IP2. Aquificae and Zetaproteobacteria sequences are still present (3.3% and 7.2%). Additionally, species richness is higher, with photosynthetic T A B L E 3 Whole-rock travertine chemistry of the studied travertines as determined by ICP-MS and ICP-AES analysis and comparison with Fe-rich travertines elsewhere and submarine Fe-rich hydrothermal depositions  (2008); 4 = hydrothermal aragonite, Pichler and Veizer (2004); 5 = sediment cores, Price and Pichler (2005); 6 = hydrothermal Fe III. oxyhydroxide precipitates, Pichler and Veizer (1999) 1,640 28,000 4.2 32,000 27,000 33,000 33,000 500 4,000 35,000 2,000 8.2 1,000 5,000 110,800 118,000 28,000 86,000 Cyanobacteria in IP3 and a large number of Bacteroidetes and Alphaproteobacteria. These are widespread clades for which metabolic styles are hard to distinguish at the phylum to family level. However, the coexistence of Cytophagia, Erythrobacteraceae, Rhodospirillaceae, and Rhodobacteraceae clades suggests a dynamic environment in photic zone (Table 1), with different sources of carbon vis autotrophy and heterotrophy under microaerophilic to aerobic conditions. Sequences obtained from samples ID5 and IP1 show a clear dominance of aerobic metabolisms. No iron-related clade was obtained and visual examination clearly showed photosynthetic activity (green-pigmented travertine interpreted as containing Cyanobacteria) for these two samples (Figure 3a,c), as well as in other areas F I G U R E 8 Bulk geochemical analysis of travertine normalized against average Upper Continental Crust (Rudnick & Gao, 2003) and Post-Archaean Australian Shale (PAAS, Taylor & McLennan, 1985). of the Ilia system where water flow was slow or non-existent (because of daily to monthly flow variations). It is suggested that the high water temperature prevents the development of Cyanobacteria in other light exposed systems. This would differentiate this system from the Okuoku-Hachikurou hot spring (Takashima et al., 2011). But as soon as the water flow rate drops or diverges, Cyanobacteria take over and dominate the travertine assemblage. Samples ID1 and ID4 show completely different communities (Figure 9b) marked by anaerobic clades. Sequences are attributed to members of the Planctomycetes, Chloroflexi unclassified or attributed to the Anaerolinae class, which consists only of strict anaerobes commonly found in hot spring microbial mats (Hanada & Pierson, 2006), Defferibacter and Deltaproteobacteria. From the latter class, sequences that could be assigned to an order were either related to Desulfobacterales, Desulfoarculales, or Syntrophobacterales, which all point towards organisms involved in sulphate reduction. Defferibacter-associated sequences could only be assigned to the thermophilic anaerobic Caldithrix genus that consists of fermenters potentially using nitrate as an electron acceptor (Alauzet & Jumas-Bilak, 2014). Overall, this environment is likely anaerobic and its microbial assemblage relies mainly on the anaerobic degradation of organic carbon, potentially involving sulphate and nitrate reduction.
The 16S rRNA gene analysis revealed the presence of diverse microbial communities, which could be associated with the deposition of ferrihydrite. The compositional variability of the microbial community is indicative of the heterogeneity and dynamic nature of the studied system. Microaerophilic conditions dominate at the vent, where water starts to equilibrate with the ambient oxygen. When the water flow decreases, oxygen content increases and microaerophilic organisms are replaced by aerobic autotrophs and heterotrophs. In buried settings, fermenters and other sulphate reducers degrade the organic matter previously deposited by runoff or past microbial production.

| Biologically induced ferrihydrite precipitation
Ferrihydrite precipitates from oxidation of Fe 2+ to Fe 3+ and from rapid hydrolysis of Fe 3+ in abiotic or biologically aided processes (Jambor & Dutrizac, 1998;Konhauser, 1998). In the case of the Ilia travertine, several lines of evidence suggest that ferrihydrite is mainly precipitated by bacterial activity in microaerophilic conditions. This conclusion is supported by the presence of fabrics showing a close association of ferrihydrite with filamentous textures resembling sheaths formed by Zetaproteobacteria (Figures 6g and 7g,h). They are covered with ferrihydrite particles, and are strikingly similar to the structures described in other settings exhibiting the same organisms (Bortoluzzi et al., 2017;Handley et al., 2010;Takashima et al., 2008). In submarine or continental hydrothermal systems, ferrihydrite is interpreted as precipitated by sheathed or stalked iron oxidizers. These sheathed or stalked iron oxidizers are generally encountered in loose sediments without any textural characteristics. Until now, only Takashima et al. (2008) have identified well-preserved erected filaments encased in dendritic structures, in their case in the Shionoha iron-rich travertine deposits although these were preserved in calcite. Encased-filaments from Ilia are visually similar to the Shionoha structures and could possibly be formed by analogous microbe-mineral interactions. In Shionoha spring (Takashima et al., 2011), the central, darker part of the dendritic fabric of the ferrihydrite spherical aggregates has been interpreted as microbially encrusted filaments. Zetaproteobacteria Mariprofundaceae, identified in the Ilia travertine 16S rRNA gene library, is known to form these ferrihydrite-covered stalks (Emerson et al., 2007;Handley et al., 2010). In Ilia deposits, they were found in environments of high water flow, which experienced moderate oxygenation that could have allowed their development under microaerophilic conditions (Konhauser, 1998;Takashima et al., 2008). All of the Ilia system data thus suggest microbially aided deposition of ferrihydrite by Zetaproteobacteria under microaerophilic conditions.
The second type of ferrihydrite facies, described as dense laminae with limited porosity and no crystalline habit, has been attributed to abiotic precipitation by Kanellopoulos et al. (2017). The findings presented here cannot rule out this hypothesis, and it is possible that ferrihydrite precipitation continues to occur once it has been initiated by the iron-oxidizing bacteria community, potentially evolving from the shrub ferrihydrite facies to porosity-filling by small ferrihydrite spherulites. A geomicrobiological investigation at the lamina scale will be necessary to untangle these processes.

| Proposed model for the precipitation of Banded Iron Travertine (BIT)
The two types of bands (i.e., Fe-rich and Ca-rich) observed in the BIT display different mineralogical, morphological, chemical, and microbiological features, all related to the distinct processes of deposition. Fe-rich bands show relative depletion in elements that are the most sensitive to redox conditions (Mn and Ce), implying more reduced conditions during their precipitation than for the Ca-rich bands. Microbiological analyses of organisms living under these conditions support this kind of redox process, all displaying a significant fraction of iron-oxidizing bacteria from the Zetaproteobacteria class. Hence, it is suggested that these microorganisms only develop where microaerophilic conditions prevail, and that they are responsible for most of the primary ferrihydrite precipitation within the iron bands. Due to the significant variations in water flux at the spring, these microaerophilic environments may develop at different levels of the borehole and pipe system. During periods of higher flow, the upcoming reduced spring water will be oxidized further away from the vent, and the site where microaerophilic Fe-oxidizers can develop and precipitate ferrihydrite will move away as well. In conditions of lower flow, fully oxidized conditions may develop in areas that used to be microaerophilic. In such conditions, biologically induced ferrihydrite precipitation would not occur and the CaCO 3 precipitation takes over as exposure to the atmosphere allows degassing and travertine formation (Figure 10). Potentially, the development of Cyanobacteria in such settings would enhance this CaCO 3 formation. Areas of standing water will most likely develop such conditions and evaporative processes accompanying the travertine deposition will allow co-precipitation of halite, as observed in the XRD data ( Figure 5).
In the case of the Okuoku-Hachikurou hot spring, a symbiotic relationship between Cyanobacteria and Fe-OB was suggested to explain the laminated texture of iron-rich deposits. These laminations were thought to reflect changes in photosynthetic intensity (Takashima et al., 2011). In the Ilia travertines, co-occurrence of Zetaproteobacteria and Cyanobacteria was not observed, likely because of the more dynamic with higher temperature. The interlayering change of Fe-rich and Ca-rich bands is explained by changes in water flow coming from the vent. It allowed the biologically driven precipitation of the iron bands to alternate with mainly abiotically driven precipitation of the CaCO 3 bands in the same location. These changes may be linked to tide cycles as sea water fuels the hydrothermal spring system, or by magmatic pulses. Pulses have been observed on a daily basis in the Euboea hydrothermal system (unpublished results). While cyanobacteria may contribute to the carbonate precipitation in this system, the high precipitation rates suggest that it is unlikely that they play the major role in carbonate precipitation.

| Iron-rich travertines as analogues to BIF
The Ilia travertines are mainly the result of sea water circulation within a hydrothermal system that subsequently leads to the formation of iron-rich deposits, when interacting with oxidative environments, at a geologically fast rate. It leads to deposition of almost pure ferrihydrite laminae alternating with another chemically precipitated facies (here calcium carbonate) that resemble BIF characteristics. In particular, the Ilia travertines have iron enrichment of up to 35 wt%, only comparable with systems like Tutum Bay in Papua New Guinea (Pichler & Veizer, 1999) and the Endeavour Segment Main Field (Hannington, De Ronde, & Petersen, 2005). Even the Ca-rich bands exhibit higher iron concentrations than many active submarine hydrothermal deposits (Table 3), which emphasizes the enormous iron enrichment exhibited by the Ilia travertine and corroborates its resemblance to BIF. As an example, Algoma-type BIF ore grade varies from 15 to 45 wt% (Bekker et al., 2010(Bekker et al., , 2014Posth, Konhauser, & Kappler, 2010). Other similarities with the general geological setting of Algoma-type BIFs (Bekker et al., 2010(Bekker et al., , 2014Cannon, 1986) are apparent in the Ilia case. The Ilia travertine forms in a back-arc region, after circulation within deeper bedrock including ophiolitic-ultramaphic rocks. Algomatype BIFs were likely deposited in island arc/back-arc basins (Bekker et al., 2010(Bekker et al., , 2014Veizer, 1983) associated with similar (ultra)mafic volcanics (Harnmeijer, 2010;Gourcerol, Thurston, Kontak, Côté-Mantha, & Biczok, 2016).
The entangling of biogenic and abiotic processes is also a salient feature of BIF deposition. Several hypotheses have been suggested concerning the biological influence on BIF deposition (Bekker et al., 2010(Bekker et al., , 2014: the anoxygenic F I G U R E 1 0 Model for the formation of banded iron travertine of Ilia. Switching from one mode to another might be controlled by variations in the flow of venting water, potentially controlled by volcanic pulses and whatever controls the fluid circulation in the hydrothermal system phototrophic iron-oxidizing bacteria, which can couple Fe (II) oxidation to the reduction of CO 2 (Ehrenreich & Widdel, 1994;Widdel et al., 1993), the cyanobacterial model for BIF formation (Cloud, 1973) or the chemolithoautotrophic iron-oxidizing bacteria living in a neutral pH environment (Holm, 1989). In the case of the BIT of Ilia, the latter model likely explains the process of iron precipitation in the Fe-rich bands. This process, however, relies on the existence of suboxic conditions (incomplete oxygenation of the spring water at the Ilia vent). Similar suboxic conditions could have been encountered during the time of BIF deposition, when oxygen production fuelled by ancestral cyanobacterial activity was only being initiated.
Although, there are some BIF with carbonate bands such as the Cauê Iron Formation in Brazil (Spier, de Oliveira, Sial, & Rios, 2007;Teixeira et al., 2017), BIFs do not typically contain carbonate bands. The mineralogy of BIF generally consists of silica as chert, magnetite, hematite, Fe-rich silicate minerals such as stilpnomelane, minnesotaite, greenalite, and riebeckite, carbonate minerals such as siderite, ankerite, calcite, and dolomite, and in some cases sulphides such as pyrite and pyrrhotite (Bekker et al., 2010(Bekker et al., , 2014Koeksoy, Halama, Konhauser, & Kappler, 2016;Gourcerol et al., 2018). The silica considered to have been originated either from the sediment-water interface in colloidal form, by absorption on iron oxyhydroxides from sea water and scavenged with organic matter (Bekker et al., 2010Fischer & Knoll, 2009Grenne & Slack, 2005;Krapez, Barley, & Pickard, 2003), or from replacement of the original sediment, formed beneath the sediment-water interface (Bekker et al., 2014;Krapez et al., 2003). However, all BIFs have undertaken significant modifications either during diagenesis, or during metamorphism. So, their mineralogy is the result of a series of factors, including the bulk composition of the original sediment, the diagenetic conditions, the metamorphic conditions, and the post-depositional fluid flow. Especially, the effects of temperature and pressure have led to mineralogy replacement, recrystallization, and from change to deletion of primary textures (Klein, 2005;Bekker et al., 2014).
The BIT of Ilia interestingly present poorly diagenetically transformed structures and chemical components that could help untangle the primary features of BIFs. For example, the presence of both, ferric and ferrous minerals in BIFs suggests an average oxidation state of Fe 2.4+ (Klein & Beukes, 1992), which means that 60% of the Fe in BIFs is made up of Fe(II). The BIF mineralogical composition reflects significant post-depositional alteration under diagenetic and metamorphic conditions. The Fe-rich bands, which consist of magnetite and hematite, are most likely to have formed from an initial Fe(III) oxyhydroxide phase such as ferrihydrite (Bekker et al., 2010(Bekker et al., , 2014Gauger, Konhauser, & Kappler, 2015;Klein, 2005;Posth, Hegler, Konhauser, & Kappler, 2008). The BIT of Ilia have an almost pure ferrihydrite composition in their Fe-rich bands, which could represent the primary composition of Fe-rich bands in BIFs and therefore may provide a model of BIF depositional characteristics.
Moreover, the BIT of Ilia also procures a model for potential diagenetic processes in precursor bands. The existence of fermenters and sulphate reducers for example, in these Fe-rich reducing environments (Figure 9) attests of organic matter recycling already taking place within sites of actively forming iron bands. The Ilia spring is therefore an interesting and easily accessible laboratory to investigate the initiation of iron mineral precipitation, micro to macrobanding formation and iron and carbon early syn-sedimentary recycling. Overall, the Ilia BIT shares many textural, mineralogical, chemical, and depositional similarities with Algoma-type BIFs. Given its intense precipitation rates, strong iron enrichment and easy access, it is suggested that a thorough investigation of the biotic and abiotic processes involved in this dynamic small-scale system will help in understanding the suite of complex processes that have led to BIFs as we know them today.