CO₂ reduction driven by a pH gradient

Significance

Biology is built of organic molecules, which originate primarily from the reduction of CO₂ through several carbon-fixation pathways. Only one of these—the Wood–Ljungdahl acetyl-CoA pathway—is energetically profitable overall and present in both Archaea and Bacteria, making it relevant to studies of the origin of life. We used geologically pertinent, life-like microfluidic pH gradients across freshly deposited Fe(Ni)S precipitates to demonstrate the first step of this pathway: the otherwise unfavorable production of formate (HCOO^–) from CO₂ and H₂. By separating CO₂ and H₂ into acidic and alkaline conditions—as they would have been in early-Earth alkaline hydrothermal vents—we demonstrate a mild indirect electrochemical mechanism of pH-driven carbon fixation relevant to life’s emergence, industry, and environmental chemistry.

Abstract

All life on Earth is built of organic molecules, so the primordial sources of reduced carbon remain a major open question in studies of the origin of life. A variant of the alkaline-hydrothermal-vent theory for life’s emergence suggests that organics could have been produced by the reduction of CO₂ via H₂ oxidation, facilitated by geologically sustained pH gradients. The process would be an abiotic analog—and proposed evolutionary predecessor—of the Wood–Ljungdahl acetyl-CoA pathway of modern archaea and bacteria. The first energetic bottleneck of the pathway involves the endergonic reduction of CO₂ with H₂ to formate (HCOO^–), which has proven elusive in mild abiotic settings. Here we show the reduction of CO₂ with H₂ at room temperature under moderate pressures (1.5 bar), driven by microfluidic pH gradients across inorganic Fe(Ni)S precipitates. Isotopic labeling with ¹³C confirmed formate production. Separately, deuterium (²H) labeling indicated that electron transfer to CO₂ does not occur via direct hydrogenation with H₂ but instead, freshly deposited Fe(Ni)S precipitates appear to facilitate electron transfer in an electrochemical-cell mechanism with two distinct half-reactions. Decreasing the pH gradient significantly, removing H₂, or eliminating the precipitate yielded no detectable product. Our work demonstrates the feasibility of spatially separated yet electrically coupled geochemical reactions as drivers of otherwise endergonic processes. Beyond corroborating the ability of early-Earth alkaline hydrothermal systems to couple carbon reduction to hydrogen oxidation through biologically relevant mechanisms, these results may also be of significance for industrial and environmental applications, where other redox reactions could be facilitated using similarly mild approaches.

A dependable supply of reduced organic molecules is essential for any scenario of the origin of life. On the early Earth, one way in which this supply could have been attained was by reduction of CO₂ with H₂ (1–6). Geological studies indicate that CO₂ was at comparatively high concentrations in the ocean during the Hadean eon, whereas H₂ was the product of multiple processes in the Earth’s crust and would have emanated as part of the efflux of alkaline hydrothermal vents (3, 7–11). The model suggests that on meeting at the vent–ocean interface, the reaction between the two dissolved gases would have produced hydrocarbons, which would in turn take roles in the transition from geochemistry to biochemistry (2, 12). Under standard conditions (1 atm, 25 °C, pH 7), the reaction between CO₂ and H₂ to produce formate (HCOO^–) is thermodynamically disfavored, with ΔG⁰’ = +3.5 kJ mol^–1 (13, 14). In ancient alkaline vents, however (Fig. 1A), H₂ was present in the OH^–-rich effluent of the vent, which would have favored its oxidation to water (1). Conversely, CO₂ would have been dissolved in the relatively acidic ocean, which facilitated protonation in its reduction to HCOO^– (1). Assisted by Fe(Ni)S minerals precipitated at the interface, a pH gradient of more than three units should have been enough to increase the viability of the reaction by ∼180 mV, thereby rendering it favorable (15, 16).

Fig. 1.

Proposed mechanism of pH-driven CO₂ reduction with H₂ across a conducting Fe(Ni)S barrier, and schematic of the reactor. (A) Under alkaline-vent conditions, the oxidation of H₂ (Left) is favored by the alkaline pH due to the presence of hydroxide ions (OH^–) that react exergonically in the production of water. Released electrons would travel across the micrometers- to centimeters-thick Fe(Ni)S network (53) (Center) to the more oxidizing acidic solution on the ocean side. There they meet dissolved CO₂ and a relatively high concentration of protons (H⁺), favoring the production of formic acid (HCOOH) or formate (HCOO^–). This electrochemical system enables the overall reaction between H₂ and CO₂, which is not observed under standard reaction conditions. (B) Diagram of the reactor, with embedded micrograph of a reaction run with precipitate at the interface. Further details are provided in the main text and SI Appendix, Fig. S2.

Once produced, and depending on local conditions, formate would have had ample abiotic chemical potential. For example, formyl groups have been shown to yield intermediates of both the reverse citric acid (Krebs) cycle (17) and the Wood–Ljungdahl (WL) reductive acetyl-CoA pathway (18), offering a possible entry route into early biological metabolism. Alternatively, on heating in the presence of ammonia—itself also a predicted component of alkaline-vent effluents—formate yields formamide [HC(O)NH₂], a highly reactive molecule central to a separate major theory for the emergence of life (19). Further reaction of this latter mixture yields hydrogen cyanide (HCN), which is also at the core of a major alternative origin-of-life scenario (20). In turn, dehydration of formate yields carbon monoxide (CO), which features prominently in a separate hydrothermal-vent model for the origin of life (21).

While there were multiple sources of reduced carbon on the early Earth and multiple plausible environments that would have hosted rich chemistries (17–25), the alkaline hydrothermal scenario described above can be especially appealing because of its resemblance to the WL pathway of carbon fixation (SI Appendix, Fig. S1) (2, 26). Highlighting its potential relevance to studies of the emergence of life, the WL process is the only one of six known biological carbon-fixation pathways that releases energy overall instead of consuming it, and variations of it are present in extant members of both Archaea (methanogens) and Bacteria (acetogens) (2, 4, 6, 12, 13, 27). The first step in the pathway is the reduction of CO₂ with H₂ to produce formate (HCOO^–, or its dehydrated electronic equivalent, CO). This reaction is endergonic, so multiple members of both Archaea and Bacteria use either electron bifurcation or a chemiosmotic pH gradient across the cell membrane to power the otherwise unfavorable step (14, 28, 29). However, in the absence of cellular coupling mechanisms such as electron bifurcation or chemiosmosis, this first endergonic step is a key energetic bottleneck in the WL pathway and remains a major open question in studies of the origin of biological carbon fixation.

Here we show the abiotic indirect reduction of CO₂ to HCOO^– with H₂, driven by a microfluidic pH gradient across freshly deposited Fe(Ni)S precipitates, via a mechanism that resembles—and may have been the evolutionary predecessor of—the decoupled electron flow of the WL pathway.

Results

Precipitation of Fe(Ni)S at the Interface in a Gas-Driven Y-Shaped Microfluidic Reactor.

Aiming to coincide with fluid compositions in previous work (5, 30, 31), and as detailed in SI Appendix, Table S1, we prepared an alkaline vent-simulant fluid containing Na₂S (100 mM), K₂HPO₄ (10 mM), and Na₂Si₃O₇ (10 mM) in deaerated water. The counterpart ocean analog fluid contained FeCl₂ (50 mM) and NiCl₂ (5 mM). The two fluids were introduced at the inlets of a Y-shaped borosilicate microfluidic reactor (Fig. 1B). Ambient pressures of H₂ and CO₂ have proven insufficient to drive CO₂ reduction (31, 32), so instead of attempting to dissolve either gas by bubbling before the reaction, here we used gas pressure-driven microfluidic pumps. The alkaline fluid was pushed with H₂ at 1.5 bar, and the ocean analog was pushed with CO₂ at the same pressure (see reactor schematic in Fig. 1B and photograph in SI Appendix, Fig. S2). We split each reactor run into two consecutive stages, the first for deposition of the Fe(Ni)S precipitates at the interface between the two fluids and the second (referred to as “postprecipitation” below) for attempting the reaction between CO₂ and H₂ (or other reagents, as detailed where relevant). Confluence of the ocean and vent analogs over 15 to 60 s in the precipitation stage yielded a 30- to 60-µm-wide precipitate at the interface between the two fluids, visible under a digital 200× optical microscope (Fig. 1B, Center). Removing metals from the ocean side following precipitation prevented the precipitate from growing to the point of occluding the reactor channel.

Detection of Formate and Confirmation with ¹³C Isotopic Labeling.

Following formation of the precipitate, and to prevent further precipitation from clogging the microfluidic channels (32), in the second (postprecipitation) stage we switched the ocean fluid to pure deaerated H₂O, pushed by CO₂ (Fig. 1B, Right). The vent simulant was as before, with Na₂S, K₂HPO₄, and Na₂Si₃O₇, pushed by H₂. Using in-line pH meters, we determined the pH of the ingoing fluids at the point of entry (shortly before entering the reactor; Fig. 1B), finding the ocean simulant at pH 3.9 and the vent simulant at pH 12.3. At a flow rate of 5 µL/min for each input, the residence time of fluids within the central channel was ∼3.3 s, so we allowed the system to flow for at least 2 min before collecting output. We then collected a single (i.e., mixed) output of the reactor over the subsequent 50 min and later analyzed it via NMR spectroscopy. Under these conditions, we detected HCOO^– by both ¹H and ¹³C NMR (Table 1, experiment 1), at an average concentration of 1.5 µM. Singlet peaks in the ¹H (8.42 ppm; Fig. 2A) and ¹³C (165.8 ppm; SI Appendix, Fig. S3) NMR spectra matched samples of pure (>98%) formic acid (SI Appendix, Fig. S4). Keeping the ocean solution free of metal salts from the start—i.e., in the absence of the precipitate—gave no detectable product (SI Appendix). Running both the precipitation and reaction stages with isotopically enriched (99% ¹³C) ¹³CO₂ (Table 1, experiment 2) gave a stronger singlet in the ¹³C spectrum (165.8 ppm; Fig. 2B) and yielded the expected splitting of the formyl singlet into a doublet in the ¹H spectrum (J = 195 Hz) due to ¹H–¹³C coupling in the formyl group (Fig. 2C).

Table 1.

Mechanistic analysis of CO₂ reduction

Experiment	Vent driving gas	Ocean driving gas	Ocean solvent	Product detected
1	H2	CO2	H2O	HCOO–
2	H2	13CO2	H2O	H13COO–
3	N2	CO2	H2O	n.d.
4	D2	CO2	H2O	HCOO–
5	H2	CO2	D2O	DCOO–

n.d., none detected. For precipitation, the ocean fluid contained FeCl₂ (50 mM) and NiCl₂ (5 mM) in deaerated H₂O or D₂O (as shown); no solutes other than the driving gas (CO₂) were included in the ocean fluid postprecipitation. Vent fluid contained Na₂S (100 mM), K₂HPO₄ (10 mM), and Na₂Si₃O₇ (10 mM) both during and following precipitation. The ocean and vent fluid pumps were driven by the gases shown, each at 1.5 bar and a flow rate of 5 µL/min. For relevant experiments from this table, quantification is listed in Table 2 and was achieved through ¹H NMR by integrating the formyl singlet compared with a known concentration of internal standard. Most entries in this table could not be quantified in this fashion, because deuterium labeling resulted in loss of the ¹H signal (experiment 5), the formyl singlet was split into a doublet by the ¹³C labeling (experiment 2), or no formate was detected (experiment 3).

Fig. 2.

¹H and ¹³C NMR spectra of produced formate. (A and B) Singlets in the ¹H (A) and ¹³C (B) NMR spectra demonstrate the production of formate in our system. (C) With isotopically labeled ¹³CO₂, coupling between ¹H and ¹³C produced an expected doublet in the ¹H spectrum. (D) Replacing regular water with deuterated ²H₂O on the ocean side yielded a triplet in the ¹³C spectrum due to coupling between the ²H and ¹³C nuclei. (E and F) No formate was detected on replacing H₂ with N₂ as the vent-driving gas. Corresponding entries in Table 1 are A:1, B:2, C:2, D:5. E:3, and F:3.

H₂ appears to be necessary for CO₂ reduction. With the vent-side fluid driven by N₂ instead of H₂ (i.e., in the absence of H₂ both during and following precipitation), we observed no reduction products (experiment 3 and Fig. 2 E and F).

Evidence for an Indirect pH-Dependent Electrochemical Mechanism of CO₂ Reduction Driven by H₂ Oxidation.

In search of mechanistic insight, we conducted deuterium (²H, or D)-labeling experiments (Table 1, experiments 4 and 5), using the isotopic variants throughout the entire experiment. Regardless of whether we used unlabeled H₂ (experiment 1) or D₂ (experiment 4) to drive the vent-side pump, we observed only non-isotopically labeled HCOO^– in the efflux, suggesting that CO₂ reduction may be occurring exclusively on the ocean side. Conversely, with D₂O used in place of regular H₂O on the ocean side, and with unlabeled H₂ driving the vent-side pump (experiment 5), we detected only deuterated formate (DCOO^–), as evidenced by a triplet in ¹³C NMR, J = 33 Hz, and the lack of any other discernible peaks (Fig. 2D). This further confirms that the CO₂ reduction here matches the isotopic composition of the ocean side, not that of the vent side.

We next explored the role of the inherent pH gradient of the simulated submarine alkaline hydrothermal system. The successful reductions reported in Table 1 proceeded with an initial ocean analog pH of 3.9 and a vent analog pH of 12.3. On mixing, this initial ΔpH of 8.4 units would have inevitably dropped, but, as we have previously shown, pH gradients of multiple units hold successfully over time across microfluidic scales, particularly in the presence of a precipitate at the interface (30). We aimed to determine whether such a pH gradient was required in our reduction system to facilitate the oxidation of H₂ on the alkaline side and the reduction of CO₂ on the acidic side (Fig. 1A). Following precipitation under the same conditions as above for experiment 1 in Table 1 (repeated in Table 2), we evaluated the effects of various changes to the pH and composition of each of the two fluids (Table 2). Replacing the vent simulant with pure H₂O driven by H₂ afforded no product (Table 2, experiment 6). Likewise, leaving Na₂S, K₂HPO₄, and Na₂Si₃O₇ in the vent solution and H₂ as the driver gas, but acidifying the vent solution with HCl to pH 3.9 and pH 7.0, resulted in no detectable formate production (Table 2, experiments 7 and 8, respectively). Adding 100 mM Na₂CO₃ to the ocean fluid while still using CO₂ as the driving gas (Table 2, experiment 9) raised the ocean-side pH to 9.8, and no product was detected under these conditions. Removing silicate from the vent side after precipitation still yielded formate (Table 2, experiment 10), as did removing both silicate and phosphate to have only Na₂S (Table 2, experiment 11). With only K₂HPO₄ in the vent side postprecipitation, we detected only trace amounts of formate (below our limit of quantification of 0.37 µM; see SI Appendix, Methods and Table S2), possibly due to the insufficiently alkaline pH of 9.1 (Table 2, experiment 12). The more alkaline K₃PO₄ raised the pH to 12.1 and led to production of considerably more formate (Table 2, experiment 13).

Table 2.

Exploration of the role of the microfluidic pH gradient across the mineral precipitate in CO₂ reduction

Experiment	Vent-side pH	Ocean-side pH	Vent-side postprecipitation solute*	Ocean-side postprecipitation solute/gas†	Formate concentration, μM‡
1§	12.3	3.9	Na2S/K2HPO4/Na2Si3O7	None/CO2	1.5 (1.66; 1.40)
6	7.0	3.9	None	None/CO2	n.d.
7¶	7.0	3.9	Na2S/K2HPO4/Na2Si3O7	None/CO2	n.d.
8¶	3.9	3.9	Na2S/K2HPO4/Na2Si3O7	None/CO2	n.d.
9	12.3	9.8	Na2S/K2HPO4/Na2Si3O7	Na2CO3/CO2	n.d.
10	12.3	3.9	Na2S/K2HPO4	None/CO2	1.5 (1.33; 1.69)
11	12.6	3.9	Na2S	None/CO2	1.8 (2.12; 1.44)
12	9.1	3.9	K2HPO4	None/CO2	(<0.37)#
13	12.1	3.9	K3PO4	None/CO2	0.9 (1.04; 0.76)

n.d., none detected.

^*Vent fluid concentrations for precipitation in all reactions were as follows: Na₂S, 100 mM; K₂HPO₄, 10 mM; Na₂Si₃O₇, 10 mM. The same concentrations were used postprecipitation, as relevant. K₃PO₄ (100 mM) was used postprecipitation in experiment 14. In all reactions, both during and after precipitation, vent fluids were driven by H₂ at 1.5 bar and a flow rate of 5 µL/min.

^†Ocean fluid for precipitation was composed of FeCl₂ (50 mM) and NiCl₂ (5 mM) in H₂O, pushed by CO₂ (1.5 bar) at a flow rate of 5 µL/min. Following precipitation, the ocean fluid for reaction was changed to contain no solutes besides dissolved CO₂ from the driving gas, except for experiment 9 with Na₂CO₃ (100 mM).

^‡Average calculated concentrations listed, with results from duplicate samples in parentheses.

^§Same as in Table 1.

^¶Vent fluid titrated with 1 M HCl to pH 7.0 (experiment 7) or pH 3.9 (experiment 8).

^#Peak was observable, but concentration was below the limit of quantification (0.37 µM; SI Appendix, Fig. S16).

While we cannot preclude the possibility of precipitate-bound sulfide acting as a reductant in addition to H₂, these results simultaneously confirm the role of the pH gradient and show that a continuous supply of aqueous sulfide is not required in our system; this contrasts with the continued requirement for H₂ (Table 1, experiment 3). Removing Ni from the ocean precipitation fluid (experiment 14) produced only small amounts of formate (below our limit of quantification of 0.37 µM). Conversely, replacing Fe to leave Ni as the only metal in the ocean precipitation fluid (NiCl₂, 55 mM; experiment 15) yielded 1.4 µM formate, pointing toward a crucial role for Ni within the precipitates. Removing both FeCl₂ and NiCl₂ from the ocean fluid expectedly did not produce a precipitate, and did not result in the production of detectable formate (experiment 16).

NMR spectra and further analyses for all experiments discussed here are presented in SI Appendix, Figs. S3–S20. Before-and-after images of the precipitates in a number of the reactions are provided in SI Appendix, Fig. S21.

Discussion

Here we report the abiotic pH-driven reduction of CO₂ to formate with H₂ in a minimalistic microfluidic replicant of an alkaline hydrothermal system. Removing the pH gradient entirely or significantly reducing it—either by acidifying the vent fluid or alkalinizing the ocean fluid—yielded no detectable formate. With the acidic ocean side fixed at pH 3.9, we observed greater formate yields with increasing vent-side pH. In experiments 7, 12, 13, 1, and 11 in Table 2, as the vent-side pH rose (7.0 < 9.1 < 12.1 < 12.3 < 12.6, respectively), so did the average formate concentration measured (undetected < below 0.37 µM < 0.9 µM < 1.5 µM < 1.8 µM).

Removing the precipitate entirely also yielded no product. Without Ni in the ocean precipitation fluid, the yield dropped below our limit of quantification, suggesting that Ni is a crucial part of the precipitate for the reduction mechanism operating here. Given that Ni can typically serve as a more efficient classical and ionic hydrogenation catalyst than Fe (33), the requirement for Ni within the precipitate is not surprising. Although we do not invoke such a direct hydrogenation mechanism here, the indirect electrochemical mechanism that we propose in Fig. 1A would rely on the metal center for similar catalytic functions as those required during classical or ionic hydrogenations.

Removing H₂ entirely by replacing it with N₂ as the vent-driving gas, while still keeping Na₂S in the vent fluid, also led to no detectable product (experiment 3). Conversely, we did observe successful formate production with Na₂S absent from the vent fluid postprecipitation and H₂ as the vent-driving gas (experiment 14). Taken together, these results suggest that H₂ is a necessary reagent in our mechanism and that sulfide is insufficient as a reducing agent. It is also noteworthy that experiment 11, which lacked K₂HPO₄ and Na₂Si₃O₇ in the vent solution at the postprecipitation stage, yielded a higher concentration of formate than experiment 1 despite containing the same concentration of sulfide; we attribute this result to the higher vent-side pH in the absence of K₂HPO₄ and Na₂Si₃O₇. While we cannot presently affirm that the precipitates are acting only as catalysts, overall our results suggest that H₂ is the main electron donor, that a large pH gradient is necessary for its oxidation, and that sulfide is insufficient (and might not be required) as an electron donor.

We used 1.5 bar pressures of both H₂ and CO₂, which are nominally not too dissimilar to the ambient-pressure bubbling used in similar earlier work that did not yield any detectable reduced carbon (31, 32). In those experiments, the two gases were dissolved into the fluids before the reactions, with inevitable loss by volatilization, particularly of H₂. Recent calculations suggest that atmospheric pressures of H₂ and CO₂ are only marginally insufficient to drive the reaction to formate (32). Therefore, we infer that the positive results reported in this work are most plausibly due to the advantages of the gas-driven system used here, which likely resulted in higher dissolved gas concentrations than the atmospheric pressure bubbling performed previously (31, 32). Chiefly, the ability to keep the gases consistently within the system, as well as to more reliably keep the fluids anoxic, may suffice as an explanation for the positive results observed here compared with previous attempts.

Alternative Mechanisms for the Reduction of CO₂ Linked to H₂ Oxidation.

Several CO₂-reduction mechanisms are possible for a laminar-flow system such as the one that we have used. We briefly explore a number of them below and discuss why we deem them less likely than the electrochemical mechanism proposed in Fig. 1A. Further details are presented in SI Appendix, Figs. S22–S27.

Perhaps the most chemically intuitive—albeit least biochemically homologous—mechanisms of carbon reduction possible here would be direct hydrogenations (Fig. 3 A–C), in which hydrogen from H₂ would be transferred directly to CO₂ (34–38) either as atomic hydrogen (classical hydrogenation) or as hydride (ionic hydrogenation). Most simply, the product in such a mechanism should match the isotopic signature of the H₂/D₂ vent gas. Instead, the formate produced here matches only the isotopic makeup of the ocean-side water, regardless of the composition of the vent-side gas or water. In these direct-hydrogenation mechanisms, adsorbed hydrogen species could in principle exchange with the surrounding fluid, such that the original isotopic signature is lost (39). However, any such process would imply the migration of significant amounts of fluid across the precipitate. The substantial mixing of fluids involved should have produced a scrambled H/D formyl signal, which we did not observe, all but ruling out any direct hydrogenation mechanisms.

Fig. 3.

Alternative CO₂ reduction mechanisms. (A–C) Classical hydrogenations with CO₂ permeability (A), H₂ permeability (B), or passage of dissociatively adsorbed atomic H (C) are all unlikely: isotopic labeling (Table 1) indicates that the formyl H derives from ocean-side water rather than H₂ from the vent side. (D) Localized redox cycling with CO₂ and permeability of H⁺ (either through a pore or anhydrously through the precipitate) are both unlikely, because in our fully hydrated system, H⁺/D⁺ would exchange with local vent water, giving a mixed or exchanged isotopic signal that we do not observe. (E) H₂ permeability is also an unlikely mechanism here, because H₂ oxidation is much less favored on the acidic (ocean) side compared with the alkaline (vent) side, as demonstrated by the pH exploration data in Table 2. More details are provided in the text and SI Appendix, Figs. S22–S27.

Alternatively, the hydrogen atoms in the resultant formate might not derive directly from H₂. Instead, the mechanism may proceed via redox cycling of an edge or corner Fe or Ni atom (M²⁺ ⇄ M⁰), wherein a metal is first reduced by H₂ (leaving two protons to dilute away) and then the metal transfers the acquired electrons on to CO₂, with accompanying proton abstraction from the local aqueous environment (Fig. 3 C–E). Noteworthy for any mechanisms that rely on proton transfer, none of our experiments contained acids added to the ocean side. The acidic pH of 3.9 that we report was achieved solely by dissolution of CO₂ in water. Thus, any ocean-side protons must derive from the dissociation of carbonic acid via:

$${\text{H}}_2\text{O}+{\text{CO}}_2\rightleftharpoons {\text{H}}_2{\text{CO}}_3\rightleftharpoons {\text{H}}^{+}+{\text{HCO}}_3^{–}.$$

When our reaction was conducted with D₂O (experiment 5) as the ocean-side solvent, we found only DCOO^– in the efflux, suggesting that CO₂ reduction did not occur on the vent side, which had normal H₂ as the feed gas and normal H₂O as the solvent and thus would have yielded unlabeled HCOO^– as the product. This was confirmed in experiment 4 with D₂ as the vent-driving gas, which yielded only unlabeled HCOO^–.

Several scenarios for such a localized redox cycling are possible (Fig. 3 D and E), but because all require colocation, none could feasibly offer the exclusively ocean-side isotopic signature that we observe. Specifically, if H⁺ were to cross the precipitate (Fig. 3D; dotted line through pore), it would likely exchange with the isotopic signature of the vent-side water, and we did not observe such a fully exchanged or even mixed signal in the isotopic labeling experiments. Passage of H⁺ by anhydrous proton conductivity through the mineral lattice (Fig. 3D; dotted line through precipitate) could potentially explain the lack of H⁺/D⁺ signal exchange, but such anhydrous proton conductivity would be highly unlikely in an otherwise fully aqueous system. Alternatively, this possible redox cycling could instead be driven by H₂ crossing to the ocean side (Fig. 3E); however, this is also unlikely, given that H₂ oxidation is considerably less favored on the acidic (ocean) side compared with the alkaline (vent) side, as demonstrated by the pH dependence of our reactions.

As an alternative to the foregoing catalytic mechanisms (classical and ionic hydrogenations, or localized redox cycling), a noncatalytic (i.e., stoichiometric) precipitate oxidation could be driving CO₂ reduction, with electrons provided by either free or bound metals or sulfide. However, the lack of reaction in the aqueous sulfide-containing but H₂-free experiment 3 suggests that there is a requirement for a continuous supply of electrons from H₂, not from dissolved sulfide or the precipitate itself.

Together with the strong pH dependence of the reactions shown in Table 2, these observations suggest that the reduction of CO₂ here proceeds via an indirect electrochemical mechanism in which electrons from H₂ oxidation on the alkaline vent side are shuttled across the Fe(Ni)S precipitates toward CO₂ on the acidic ocean side (Fig. 1A). There the reacting CO₂ picks up a proton from the local water, yielding formate with the isotopic signature of the ocean side.

Venturi Pull Inside Hydrothermal Pores Would Bring Ocean Fluids into the Vent System.

We show that production of organics could have occurred on the ocean side of an ancient alkaline hydrothermal-vent system. This raises the question of whether these organics would simply dilute away into the ocean before they could assume any biochemical role (40). It is also conspicuously unlike the WL pathway (32), in which carbon reduction occurs inside the cell. Because the microscopic structure of hydrothermal vents is typically highly porous and reticulated (9, 41), we hypothesize that ocean fluids could have been actively pulled into the microfluidic system due to Venturi-effect suction. Once inside, the ocean’s carbonic fluids could react with electrons being transferred across the vent’s catalytic minerals, and fresh precipitation could also occur further in as the two fluids meet. We have developed a computational finite-elements simulation to explore this prediction (SI Appendix, Fig. S28 and accompanying text in SI Appendix). Our results show that a system of size similar to our 300-µm-wide reactor would indeed lead to the type of microfluidic confluence of reagents that we have shown here and elsewhere (30, 31, 42). Notably, such an effect is not limited to submarine alkaline vents and will likely occur within porous hydrothermal systems at any location and depth, lending itself to multiple geochemical scenarios for the emergence of life.

Batch vs. Flow Reactors.

An alternative one-pot batch (rather than microfluidic) system for the reduction of CO₂ with H₂ has been reported recently (43), elaborating on previous results (18, 44, 45). Using more highly reduced minerals (Fe₃Ni), higher pressures (10 bar H₂), and higher temperatures (100 °C) than those used here, the batch system generates significantly higher concentrations of formate, as well as multiple further reduction products (including acetate, methanol, and pyruvate), all at a formate production rate (5.21 × 10⁻⁹ mol/s) four orders of magnitude higher than that reported here (2.5 × 10⁻¹³ mol/s). Several of these reduction products (e.g., acetate, pyruvate) represent an as-yet unachieved benchmark for microfluidic systems: the formation of new C–C bonds.

In a closer comparison to our system, when the mineral Fe₃S₄ was used (43) at a pressure of 1.6 bar H₂ and temperature of 20 °C, the production rate (1.35 × 10⁻¹¹ mol/s) was two orders of magnitude higher than that for our system. The differences between the two systems are crucial. The pH-driven CO₂ reduction demonstrated here, along with the possibility of thermal-driven concentration increases in small pores (46–48), make microfluidic reactors attractive despite the presently lower yields. The two types of systems thus provide complementary compelling evidence for organics arising under anoxic alkaline hydrothermal vent conditions, and both must be explored further as potential sources for life’s first molecules.

Fates of Formate beyond the WL Pathway.

While we speculatively favor the continuity between geochemistry and biochemistry offered by the WL pathway of carbon fixation, it is worth noting that formate can readily yield formamide, hydrogen cyanide, and carbon monoxide, each of which is at the core of major alternative scenarios for the origin of life (19–21). Therefore, our results may extend beyond the alkaline hydrothermal vent theory or may link it to other scenarios, a set of possibilities that remain to be explored.

Conclusions

We report the pH-driven reduction of CO₂ with H₂ in the first step of a geologically plausible analog—and proposed evolutionary predecessor (2)—of the WL pathway of carbon fixation. These results tie in with previous findings that alkaline vent-like pH gradients can be kept at a microscale across Fe(Ni)S precipitates (30), that these pH gradients can provide the necessary overpotentials to drive the reaction between H₂ and CO₂ (15), that low pressures of the two gases are insufficient for reaction (31, 32), that heat gradients in hydrothermal systems can drive concentration increases by thermophoresis and capillary flow at gas–liquid interfaces (46–48), and that redox and pH gradients can drive early bioenergetics (49, 50) and amino acid synthesis (51).

Both dissolved hydrogen gas and a significant pH gradient of the type and polarity found at alkaline hydrothermal vents appear to be necessary driving forces in our system. The exact sources and sinks of H₂—and corresponding concentrations of the dissolved gas—on present-day and early-Earth hydrothermal systems remain areas of intensive research (11, 52). Our results suggest that compartmentalization and pH gradients may become more important at lower H₂ concentrations (52). Furthermore, our findings suggest an electrochemical mechanism whereby dissolved H₂ oxidizes on the alkaline vent side and its electrons travel through the catalytic precipitate network toward the acidic ocean side, where they reduce dissolved CO₂ to formate. Experimental confirmation of this hypothesis has proven elusive to date (5, 31, 32). Since this is the first endergonic hurdle in the reduction of CO₂ via the sole energetically profitable CO₂ fixation pathway known, our findings provide a geologically credible route to a plausible origin of carbon fixation at the emergence of life on the early Earth and potentially on water-rock planetary bodies elsewhere.

Methods

Minimal simulants of alkaline-vent and oceanic fluids were driven into respective tips of a Y-shaped borosilicate reactor. To maximize the amount of gases dissolved, we used gas pressure-driven microfluidic pumps (Mitos P-Pump; Dolomite Microfluidics). The ocean fluid was pushed with CO₂, and the vent fluid was pushed with H₂, both at 1.5 bar. The reaction was undertaken in an anaerobic N₂-purged glovebox. Single (mixed) effluxes were collected and analyzed using ¹H and ¹³C NMR. Quantification was achieved with ¹H NMR. Further descriptions of experimental, analytical, and computational methods used in this study are provided in SI Appendix.

Data Availability.

All data necessary for the replication of this work, including descriptions of the reactor and fluid compositions, are available in the main text and SI Appendix.