A fresh decimetre-sized sample of a Dresser chert vein was obtained from a recent cut wall of the abandoned Dresser Mine in the Pilbara Craton, Western Australia (GPS: 21∘09′04.13′′ S; 119∘26′15.21′′ E; Figs. 1, S1d; for geological maps, see Hickman, 1983; Van Kranendonk, 1999; Hickman and Van Kranendonk, 2012). The external surfaces (ca. 1–2 cm) of the sample block were removed using an acetone-cleaned rock saw and then used for the preparation of thin sections. The surfaces of the resulting inner block were extensively rinsed with acetone and then removed (ca. 1–2 cm) with an acetone-cleaned high precision saw (Buehler; Isomet 1000, Germany). The surfaces of the resulting sample were again rinsed with acetone and then crushed and powdered using a carefully acetone-cleaned pebble mill (Retsch MM 301, Germany).

Petrographic analysis was conducted using a Zeiss SteREO Discovery.V8 stereomicroscope (transmitted and reflected light) linked to an AxioCam MRc5 5-megapixel camera.

Raman spectra were recorded using a Horiba Jobin Yvon LabRam-HR 800 UV spectrometer (focal length of 800 mm) attached to an Olympus BX41 microscope. For excitation an Argon ion laser (Melles Griot IMA 106020B0S) with a laser strength of 20 mW was used. The laser beam was focused onto the sample using an Olympus MPlane 100× objective with a numerical aperture of 0.9 and dispersed by a 600 Lmm-1 grating on a charge-couple device (CCD) detector with 1024×256 pixels. This yielded a spectral resolution of < 2 cm-1 per pixel. Data were acquired over 10 to 30 s for a spectral range of 100–4000 cm-1. The spectrometer was calibrated by using a silicon standard with a major peak at 520.4 cm-1. All spectra were recorded and processed using the LabSpec^™ database (version 5.19.17; Jobin Yvon, Villeneuve d'Ascq, France).

H / C values were calculated based on integrated peak intensities of Raman spectra using the formula H / C = 0.871 ⋅ ID5/(IG+D2)-0.0508 (Ferralis et al., 2016). The peaks were fitted in the LabSpec^™ software (see Sect. 2.2) using the Gauss/Lorentz function.

All materials used for preparation were heated to 500 ∘C for 3 h and/or extensively rinsed with acetone prior to sample contact. A laboratory blank was prepared and analysed in parallel to monitor laboratory contaminations.

We applied catalytic hydropyrolysis (HyPy) to release molecules from the Dresser kerogen and pre-extracted biomass of the heterocystous cyanobacterium Anabaena cylindrica SAG 1403-2 following previously published protocols (Snape et al., 1989; Love et al., 1995, 2005). HyPy allows the breaking of covalent bonds by progressive heating under high hydrogen pressure (150 bar). The released products are immediately removed from the hot zone by a constant hydrogen flow and trapped downstream on clean, combusted silica powder cooled with dry ice (Meredith et al., 2004). All hydropyrolysates were eluted from the silica trap with high-purity dichloromethane (DCM), desulfurized overnight using activated copper and subjected to gas chromatography – mass spectrometry (GC-MS). Before all experiments, the empty HyPy system was heated (ambient temperature to 520 ∘C, held for 30 min) to remove any residual molecules.

Isolation of the Dresser kerogen followed standard procedures (cf. Brocks et al., 2003b). 57 g of powdered sample was first demineralized with hydrochloric acid (24 h) and then hydrofluoric acid (11 days). The purified organic matter was then exhaustively extracted using three excess volumes of DCM and n-hexane, respectively (×3), ultrasonic swelling in pyridine (2×20 min at 80 ∘C), and ultrasonic extraction with methanol, DCM (×3) and DCM / methanol (1/1; v/v). The kerogen was subsequently extracted with n-hexane until no more compounds were detected via GC-MS. The pure kerogen was used for HyPy following an experimental protocol optimized for the analysis of Archaean kerogens (Brocks et al., 2003b; Marshall et al., 2007).

In order to monitor potential contamination, we applied HyPy to blanks before and after the kerogen run. The Dresser kerogen (131.06 mg) and blanks were sequentially heated in the presence of a sulfided molybdenum catalyst (10 wt%) and under a constant hydrogen flow (5 dm3min-1) using a two-step approach (cf. Brocks et al., 2003b; Marshall et al., 2007; Fig. S2). The low-temperature step included heating from ambient temperature to 250 ∘C (300 ∘Cmin-1) and then to 330 ∘C (8 ∘Cmin-1) to release any molecules that were strongly adsorbed to the kerogen and not accessible to solvent extraction. The silica powder was subsequently recovered from the product trap, and the trap was refilled with clean, combusted silica powder for the following high-temperature step. This step included heating from ambient temperature to 520 ∘C (8 ∘Cmin-1) to release molecules covalently bound to the kerogen.

We used cell material of the heterocystous cyanobacterium Anabaena cylindrica strain SAG 1403-2 as it fulfils all criteria for reference material (availability, well characterized, etc.). The material was obtained from the Culture Collection of Algae at the Georg-August-Universität Göttingen (Germany) and grown at the Christian-Albrechts-Universität Kiel (Germany). The axenic batch culture was maintained in 250 mL of BG110 medium free of combined nitrogen sources (Rippka et al., 1979) at 29 ∘C. A light : dark regime of 14 h : 10 h with a photon flux density of 135 µmolphotonsm-2s-1 was provided by a white fluorescent light bulb. At the end of the logarithmic growth phase, cells were harvested by centrifugation, and they were lyophilized thereafter. Subsequently, an aliquot of the freeze-dried biomass was exhaustively extracted following the methodology described by Bauersachs et al. (2014).

HyPy was performed following an experimental protocol optimized for biomass applications (Love et al., 2005). Briefly, the material was heated in the presence of a sulfided molybdenum catalyst (1.5 times the weight of the bacterial extraction residue) and under a constant hydrogen flow (6 dm3min-1). The HyPy program included heating from ambient temperature to 260 ∘C (300 ∘Cmin-1) and then to 500 ∘C (8 ∘Cmin-1).

Fischer–Tropsch-type reactions were carried out based on McCollom et al. (1999). A mixture of 2.5 g oxalic acid (dihydrate, suprapur^®, Merck KGaA), 1 g montmorillonite (K10, Sigma Aldrich; pre-extracted with DCM; ×4) and ca. 11 mL ultrapure water was heated to 175 ∘C for 66 to 74 h in a sealed Morey-type stainless steel autoclave. The autoclave was rapidly (≤ 15 min) cooled to room temperature with compressed air.

Fluid and solid phases were collected and extracted with DCM (×4). The montmorillonite was removed by centrifugation and filtration with silica powder and sea sand (which were combusted prior to use). The obtained extracts were then concentrated by rotary evaporation (40 ∘C, 670 mbar) and subjected to GC-MS analysis. Analytical blank experiments were carried out with all reactants to keep track of contamination.

GC-MS analysis of HyPy products was carried out with a Thermo Scientific Trace 1310 GC coupled to a Thermo Scientific Quantum XLS Ultra MS. The GC instrument was equipped with a capillary column (Phenomenex Zebron ZB-5, 30 m, 0.25 µm film thickness, 0.25 mm inner diameter). Fractions were injected into a splitless injector and transferred to the GC column at 270 ∘C. He was used as carrier gas with a constant flow rate of 1.5 mLmin-1. The GC oven temperature was held isothermal at 80 ∘C for 1 min and then ramped to 310 ∘C at 5 ∘Cmin-1, at which it was kept for 20 min. Electron ionization mass spectra were recorded in full-scan mode at an electron energy of 70 eV with a mass range of m/z 50–600 and scan time of 0.42 s.

Polyaromatic hydrocarbons (PAHs) are organic compounds consisting of multiple fused benzene rings. Methylation and isomerization characteristics of various GC-amenable PAHs are altered during maturation and thus provide a measure for assessing the thermal maturity of organic matter (Killops and Killops, 2005; Peters et al., 2005). The following PAH maturity parameters were used in this study:

methylnaphthalene ratio (MNR) = 2-MN / 1-MN (Radke et al., 1984),

TOC, δ13CTOC and compound-specific δ13C analyses were conducted at the Centre for Stable Isotope Research and Analysis (KOSI) at the Georg-August-Universität Göttingen, Germany. Stable carbon isotope data are expressed as delta values relative to the Vienna Pee Dee Belemnite (VPDB) reference standard.

The TOC content and δ13CTOC values were determined in duplicate using an elemental analyser (NA-2500 CE-Instruments) coupled to an isotope ratio mass spectrometer (Finnigan MAT Delta plus). Ca. 100 mg of powdered and homogenized whole rock material were analysed in each run. For internal calibration an acetanilide standard was used (δ13C = -29.6 ‰; SD = 0.1 ‰). TOC content measurements showed a mean deviation of 0.1 wt%. The average δ13CTOC value had a standard deviation of 0.3 ‰.

Compound-specific δ13C analyses were conducted with a Trace GC coupled to a Delta Plus isotope ratio mass spectrometer (IRMS) via a combustion interface (all Thermo Scientific). The combustion reactor contained CuO, Ni and Pt and was operated at 940 ∘C. The GC was equipped with two serially coupled capillary columns (Agilent DB-5 and DB-1; each 30 m, 0.25 µm film thickness, 0.25 mm inner diameter). Fractions were injected into a splitless injector and transferred to the GC column at 290 ∘C. The carrier gas was He at a flow rate of 1.2 mLmin-1. The GC oven temperature program was identical to the one used for GC-MS analysis (see above). CO2 with a known δ13C value was used for internal calibration. Instrument precision was checked using laboratory standards. Standard deviations of duplicate measurements were better than 1.7 ‰.

The organic record of Archaean rocks is commonly affected by thermal maturation (Brocks et al., 2008; Brocks, 2011; French et al., 2015). The kerogen within the analysed Dresser sample is thermally mature and structurally disordered, as evidenced by relatively wide D and G bands in the Raman spectra (see Fig. 2c for a representative Raman spectrum). This is also supported by the absence of S1 bands at 2450 cm-1, high R1 ratios (0.98–1.05) and low FWHM-D1 values (68.64–76.12), consistent with prehnite–pumpellyite to lower greenschist metamorphism at a temperature of ca. 300 ∘C (cf. Yui et al., 1996; Delarue et al., 2016). All of these observations are well in line with published Raman spectra of Archaean organic matter from the same region (Ueno et al., 2001; Marshall et al., 2007; Delarue et al., 2016) and the general thermal history of the host rock (regional prehnite–pumpellyite to lower greenschist metamorphism; Hickman, 1975, 1983, 2012; Terabayashi et al., 2003).

Fitting of D5 peaks can be difficult for Raman spectra of highly mature organic matter (Ferralis et al., 2016). Therefore, the H / C ratios calculated herein (0.03–0.14) should be treated with caution. However, the Raman-based H / C ratios and the methylphenanthrene / phenanthrene value (MP / P = 0.33; Table 2) of the Dresser kerogen are in good accordance with data reported from more mature kerogens from the same region (ca. 3.4 Ga Strelley Pool Formation: 0.08–0.14 and 0.24–0.37, respectively; Marshall et al., 2007). The low methylphenanthrene index (MPI-I = 0.23) and high phenanthrene / methylphenanthrene index (P / MP = 3.01) result in a computed vitrinite reflectance (Rc (MPI-I)) of 2.87 (Table 2), indicating a thermal maturity far beyond the oil-generative stage (Radke and Welte, 1983; Boreham et al., 1988). However, it has to be considered that the MPI-I is potentially affected by (de-)methylation reactions (Brocks et al., 2003a). The calculated methylnaphthalene ratio (MNR = 2.52; Table 2) corresponds to a somewhat lower mean vitrinite reflectance of ca. 1.5, which is again in line with a post-oil window maturity (cf. Radke et al., 1984). The mismatches between single aromatic maturity parameters are negligible, as these indices have limited application for highly mature Archaean organic matter (Brocks et al., 2003a). The putative offset between the indices and the metamorphic overprint indicated by Raman data is most likely due to the protection of kerogen-bound moieties even under elevated thermal stress (Love et al., 1995; Lockhart et al., 2008). Furthermore, it has been shown that kerogen isolated from the Strelley Pool Formation also contains larger PAH clusters which are not GC-amenable (> 10–15 rings; Marshall et al., 2007). Consequently, it can be anticipated that the compounds detected in the HyPy pyrolysate of the Dresser kerogen represent only a small fraction of the bulk macromolecular organic matter.

The distribution of monomethylalkanes ≤n-C18 (Figs. S3, S4) released from the Dresser kerogen resembles high-temperature HyPy products from the Strelley Pool kerogen (Marshall et al., 2007; their Fig. 15). Such isomeric mixtures are typically formed during thermal cracking of alkyl moieties (Kissin, 1987) and are in good agreement with the estimated maturity range. Methylated aromatics such as methylnaphthalenes and methylphenanthrenes have also been observed in other hydropyrolysates from Archaean kerogens that experienced low-grade metamorphism (Brocks et al., 2003b; Marshall et al., 2007; French et al., 2015). In all of these cases, the degree of alkylation varied with the exact thermal alteration of the respective kerogens (Marshall et al., 2007; French et al., 2015).

The kerogen of the Dresser Formation exclusively occurs in the form of fluffy aggregates and clots embedded within a very dense chert matrix that is, once solidified, highly impermeable to fluids. The depositional age of the formation is constrained to 3481±3.6 Ma (Van Kranendonk et al., 2008), and the investigated chert vein shows no evidence for disruption by post-depositional hydrothermal fluids. This has also been described for other hydrothermal chert veins of the Dresser Formation, where the kerogen has been interpreted as being syngenetic (i.e. formed prior to host rock lithification; Ueno et al., 2001, 2004; Morag et al., 2016). Furthermore, the maturity of the embedded kerogen is in good accordance with the thermal history of the host rock. An introduction of solid macromolecular organic matter from stratigraphically younger units in this region during later fluid-flow phases, as proposed for the younger Apex chert (Olcott-Marshall et al., 2014), can therefore be excluded.

As the bitumen fractions of Precambrian rocks are easily biased by the incorporation of contaminants during later stages of rock history (Brocks et al., 2008; Brocks, 2011; French et al., 2015), studies have increasingly focussed on kerogen-bound compounds that are more likely to be syngenetic to the host rock (Love et al., 1995; Brocks et al., 2003b; Marshall et al., 2007; French et al., 2015). Potential volatile organic contaminants adhering to the kerogen are removed through excessive extraction and a thermal desorption step (∼350 ∘C) prior to high-temperature HyPy (550 ∘C; cf. Brocks et al., 2003b; Marshall et al., 2007). The recurrence of few n-alkanes in the range of n-C14 to n-C24 (maximum at n-C15) in high-temperature HyPy blank runs obtained immediately before and after the actual sample run (Figs. 3a, S2, S3) indicates minor background contamination during pyrolysis, with a source most likely within the HyPy system. However, these contaminants do not significantly affect the n-alkane pattern yielded by high-temperature HyPy of the Dresser kerogen, as evidenced by blank subtraction (Fig. 3c).

Contamination of the sample can be further deciphered by the presence of polar additives or hydrocarbons that are not consistent with the thermal history of the host rock. Plastic-derived branched alkanes with quaternary carbon centres (BAQCs), a common contaminant in Precambrian rock samples (Brocks et al., 2008), have not been detected (Fig. S7). Traces of functionalized plasticizers (phenols and phthalic acid; Figs. S2, S6, S8) are unlikely to survive (or result from) HyPy treatment. These compounds are therefore considered as background contamination introduced during sample preparation and analysis after HyPy. The observed siloxanes (Fig. 3, S2) probably originate from the GC column or septum bleeding and are unlikely to be contained in the sample. All of these compounds occur only in low or trace abundances and can be clearly distinguished from the ancient aliphatic and aromatic hydrocarbons contained in the Dresser kerogen.

Contamination by (sub-)recent endoliths can be excluded as sample surfaces have been carefully removed and there is no petrographic indication for borings or fissures containing recent organic material (Fig. 2). HyPy of untreated or extracted biomass would yield a variety of acyclic and cyclic biomarkers (cf. Love et al., 2005). However, high-temperature HyPy of the Dresser kerogen almost exclusively yielded n-alkanes, minor amounts of monomethylalkanes and various aromatic hydrocarbons (Figs. 3b, c, S2–S6), while hopanoids or steroids were absent (Figs. S8, S9). This is in good agreement with the maturity of the Dresser kerogen and previous HyPy studies of Archaean rocks (Brocks et al., 2003b; Marshall et al., 2007; French et al., 2015).

Accidental contamination of the kerogen by mono-, di- and triglycerides (e.g. dust, skin surface lipids) can also be ruled out as HyPy treatment of these compounds typically results in n-alkane distributions with a distinct predominance of n-C16 and n-C18 homologues, corresponding to the n-C16 and n-C18 fatty acid precursors (Craig et al., 2004; Love et al., 2005; unpublished data from our own observations). At the same time, the observed distribution of kerogen-derived n-alkanes, with a distinct decrease beyond n-C18 (Fig. 3b, c), is notably similar to high-temperature HyPy products of kerogens isolated from the ca. 3.4 Ga Strelley Pool Formation of the Pilbara Craton (Marshall et al., 2007; their Fig. 14). Marshall and co-workers considered these n-alkanes unlikely to be contaminants as they were released only in the high-temperature HyPy step. Furthermore, the stable carbon isotopic composition of n-alkanes in the Dresser high-temperature pyrolysate (-29.4 to -33.3 ‰; mean -31.4 ± 1.2 ‰) is very similar to the δ13CTOC signature of the sample (-32.8 ± 0.3 ‰; Figs. 3, S10; Table 1), indicating that these compounds were generated from the kerogen. Hence, we consider the compounds released from the Dresser kerogen during high-temperature HyPy (Fig. 3b, c) as syngenetic.

High-temperature HyPy of the Dresser kerogen yielded a variety of aromatic hydrocarbons, which are orders of magnitudes lower or absent in all other pyrolysates (Figs. 3, S2, S4–S6). It also produced significantly higher amounts of n-alkanes than the low-temperature step, and these further showed a distinct distribution pattern (i.e. a step ≤n-C18; Figs. 3, S2, S3). The lack of even low quantities of n-alkenes that typically accompany bond cleavage in kerogen pyrolysis was also observed by Marshall et al. (2007) in their HyPy analysis of cherts from the Strelley Pool Formation. As a possible explanation for the lack of n-alkenes, these authors suggested that the n-alkanes were not cracked from the kerogen but were rather trapped in closed micropores until the organic host matrix was pyrolytically disrupted. Alternatively, however, unsaturated cleavage products may have been immediately reduced during HyPy by the steadily available hydrogen and catalysts. Indeed, it has been demonstrated that double bonds in linear alkyl chains are efficiently hydrogenated during HyPy, even in the case of pre-extracted microbial biomass (e.g. Love et al., 2005; our Fig. 3d). We consider both as plausible scenarios to explain the absence of n-alkenes in the Dresser chert hydropyrolysate.

The δ13CTOC value of -32.8 ± 0.3 ‰ is consistent with carbon fixation by photo- or chemoautotrophs (cf. Schidlowski, 2001). However, organic compounds exhibiting similar 13C depletions (partly down to ca. -36 ‰ relative to the initial substrate) could also be formed abiotically during the serpentinization of ultramafic rocks (Fischer–Tropsch-type synthesis; McCollom et al., 1999; McCollom and Seewald, 2006; Proskurowski et al., 2008). A further constraint on the origin of the Dresser kerogen is provided by the distinct decrease in n-alkane abundance beyond n-C18 observed in the high-temperature HyPy pyrolysate (Fig. 3b, c). This distribution resembles HyPy products of pre-extracted recent cyanobacterial biomass, which also shows a very pronounced restriction in carbon number to homologues ≤n-C18 (Fig. 3d).

The pre-extracted cyanobacterial cell material and the abiotically produced n-alkanes studied as reference samples experienced no thermal maturation. It can nevertheless be expected that burial, and thus heating, of Anabaena biomass (Fig. 3d) would initially liberate lower n-alkane homologues from their predominant n-alkyl moieties while, up to a certain point, retaining the distinct step at n-C18. Experimental maturation of an immature kerogen revealed the preservation of distinct alkyl-chain length preferences even after 100 days at 300 ∘C (Mißbach et al., 2016; e.g. a step beyond n-C31 in their Fig. 2). Further maturation would ultimately lead to a unimodal distribution of short-chain n-alkanes and erase the biologically inherited pattern (cf. Mißbach et al., 2016). In contrast, abiotically synthesized extractable organic compounds show a unimodal homologue distribution from the beginning (Fig. 3e) and will retain it, while thermal maturation would gradually shift the n-alkane pattern towards shorter homologues. It can therefore be expected that organic compounds cleaved from an abiotic “Fischer–Tropsch-kerogen” – whose existence has not been proven yet – would also exhibit a unimodal distribution. Consequently, the distinctive distribution of n-alkanes released from the Dresser kerogen (i.e. the step ≤n-C18; Figs. 3b, c) can be regarded as a molecular fingerprint relating to a biosynthetic origin of the organic matter.

Highly 13C-depleted methane in primary fluid inclusions in hydrothermal chert veins of the Dresser Formation (δ13C < 56 ‰) was taken as evidence for biological methanogenesis and thus the presence of Archaea (Ueno et al., 2006). Our δ13CTOC value (-32.8 ± 0.3 ‰; Table 1; Fig. S10) would generally be consistent with both bacterial and archaeal sources (cf. Schidlowski, 2001). Archaea only synthesize isoprene-based compounds (Koga and Morii, 2007; Matsumi et al., 2011). Straight-chain (acetyl-based) hydrocarbon moieties such as fatty acids – the potential precursors of the kerogen-derived n-alkanes – are to current knowledge being formed only by Bacteria and Eukarya, where they typically function as constituents of membranes or storage lipids (cf. Erwin, 1973; Kaneda, 1991). The formation of these lipids is tightly controlled by different biosynthetic pathways resulting in characteristic chain-length distributions. In bacterial lipids, carbon chain lengths typically do not extend above n-C18 (cf. Kaneda, 1991). Consequently, given the lack of convincing evidence for the presence of Eukarya as early as 3.5 Ga (cf. Parfrey et al., 2011; Knoll, 2014; French et al., 2015), the most plausible source of kerogen-occluded n-alkanes in the Dresser hydrothermal chert is Bacteria. Strongly reducing conditions during the deposition of the Dresser Formation are indicated by widespread pyrite, Fe-rich carbonates and trace element signatures (Van Kranendonk et al., 2003, 2008). Potential microbial sources for the Dresser kerogen therefore may have included anoxygenic photoautotrophic, chemoautotrophic and heterotrophic microorganisms.

Our results strongly support a biological origin of the kerogen found in the early Archaean hydrothermal chert veins of the Dresser Formation. We explain this finding by the redistribution and sequestration of microbial organic matter that may have formed in a variety of Dresser environments through hydrothermal circulation (proposed herein as the “hydrothermal pump hypothesis”; Fig. 4). Higher geothermal gradients prior to the onset of modern-type plate tectonics at ca. 3.2–3.0 Ga (Smithies et al., 2005; Shirey and Richardson, 2011) were possibly important drivers of early Archaean hydrothermal systems. In fact, the Dresser Formation was formed in a volcanic caldera environment affected by strong hydrothermal circulation, where voluminous fluid circulation locally caused intense acid–sulfate alteration of basalts and the formation of a dense hydrothermal vein swarm (Nijman et al., 1999; Van Kranendonk and Pirajno, 2004; Van Kranendonk, 2006; Van Kranendonk et al., 2008; Harris et al., 2009). In addition to the high crustal heat flow, the absence of thick sedimentary cover may have facilitated the intrusion of seawater into the hydrothermal system. The associated large-scale assimilation of particulate and dissolved organic matter and its transport and alteration by hydrothermal fluids (Fig. 4) therefore appears to be a plausible mechanism that may, at least partly, explain the high amounts of kerogen in early Archaean hydrothermal veins.

The hydrothermal pump hypothesis requires a source of organic matter during the deposition of the Dresser Formation (Fig. 4). Whereas contributions from extraterrestrial sources, as well as from Fischer–Tropsch-type synthesis linked to the serpentinization of ultramafic rocks, cannot be excluded, our results indicate a primarily biological origin for the kerogen contained in the chert veins (Fig. 4). The inferred biogenicity is also in line with the consistent δ13C offset between bulk kerogens (ca. -20 to -30 ‰) and carbonate (ca. ±2 ‰) in Archaean rocks (Hayes, 1983; Schidlowski, 2001). Prokaryotic primary producers and heterotrophs may have flourished in microbial mats (Dresser stromatolites; Walter et al., 1980; Van Kranendonk, 2006, 2011; Philippot et al., 2007; Van Kranendonk et al., 2008), the water column (planktic “marine snow”; Brasier et al., 2006; Blake et al., 2010) and even hot springs on land (Djokic et al., 2017). Another biological source for the ancient organic matter could have been chemoautotrophs and heterotrophs thriving in subsurface environments, such as basalts (Banerjee et al., 2007; Furnes et al., 2008) and hydrothermal vent systems (Shen et al., 2001; Ueno et al., 2001, 2004, 2006; Pinti et al., 2009; Morag et al., 2016) (Fig. 4). All of these systems are not mutually exclusive and the largely anoxic conditions would have encouraged a high steady-state abundance of organic matter in the aquatic environment (Fig. 4).

Dissolved organic matter (DOM) in modern seawater may resist decomposition over millennial timescales (Druffel and Griffin, 2015). In recent hydrothermal fields, however, organic matter becomes thermally altered and redistributed (Simoneit, 1993; Delacour et al., 2008; Konn et al., 2009). Laboratory experiments using marine DOM indicate that thermal alteration already occurs at temperatures >68–100 ∘C and efficient removal of organic molecules at 212–401 ∘C (Hawkes et al., 2015, 2016). It has been argued, however, that such DOM removal may also be due to transformation into immiscible material through, for example, condensation (Castello et al., 2014) and/or defunctionalization reactions (Hawkes et al., 2016). These processes, however, are as yet poorly understood. In the Dresser Formation, hydrothermal temperatures ranged from ca. 300 ∘C at depth to 120 ∘C near the palaeosurface, causing propylitic (ca. 250–350 ∘C) and argillic (including advanced argillic: ca. 100–200 ∘C) alteration of the host rocks (Van Kranendonk and Pirajno, 2004; Van Kranendonk et al., 2008; Harris et al., 2009). Given this variety of thermal regimes, and the generally anoxic nature of early Archaean seawater (e.g. Van Kranendonk et al., 2003, 2008; Li et al., 2013), it is likely that some of the organic substances underwent in situ alteration, but not complete oxidation, during hydrothermal circulation. The entrained organics would have been trapped in the chert that instantaneously precipitated from the ascending hydrothermal fluids due to subsurface cooling (cf. Van Kranendonk, 2006).

In summary, the hydrothermal pump hypothesis proposed here (Fig. 4) includes (i) a net build-up of organic matter in different Dresser environments under largely anoxic conditions, (ii) a large-scale assimilation of particulate and dissolved organic matter from various biological sources and its subsurface transport and alteration by hydrothermal fluids, and (iii) its sequestration within hydrothermal chert veins as kerogen. This model explains the presence of abundant organic carbon in early Archaean hydrothermal veins, as well as its morphological, structural and isotopic variability observed in the Dresser hydrothermal chert veins (Ueno et al., 2001, 2004; Pinti et al., 2009; Morag et al., 2016). It does not, however, help to pinpoint the formation pathways of distinct carbonaceous structures, as for instance those preserved in the Dresser Formation (Glikson et al., 2008) or the younger Apex chert (e.g. Schopf, 1993; Brasier et al., 2002, 2005; Schopf et al., 2002; Bower et al., 2016). Further work is necessary to test whether consistent molecular and compound-specific isotopic patterns can be generated from a larger set of Archaean kerogens.