Introduction

Phosphorus-bearing organic compounds are essential for cellular structure (phospholipids), metabolic energy transfer (ATP), and information storage (DNA, RNA) in all known forms of life. However, phosphorus is less abundant than other bioessential major elements (C, H, O, N, S)1 and is mostly locked in sparsely-soluble phosphate (P(V)) minerals (e.g., hydroxylapatite (Ca10(PO4)6(OH)2))2. Phosphorus is therefore a limiting nutrient for marine bioproductivity across geologic timescales3 and may have hindered the prebiotic formation of phosphorylated organic compounds that are necessary for the origin of life4,5,6. Low solubility of phosphate in water and its low reactivity toward the organic compounds (i.e. phosphorylation) are collectively known as “Phosphate Problem” for the origin of life.

Several hypotheses have been developed to resolve the solubility issue with phosphate including its presence in carbonate-rich alkaline lakes7, or in urea-ammonium formate-rich solutions with low water activity that enhance the solubility of natural phosphate mineral including apatite8. It has also been proposed that the early ocean contained higher levels of dissolved phosphate than previously thought due to new constraints on the solubility of vivianite9. However, these hypotheses have several limitations. For example, in alkaline lakes concentrations of metal micronutrients such as Fe, Cu, and Zn required for the origin and survival of life are limited, and natural environments with millimolar level urea and ammonium formate have yet to be discovered. Similarly, for the open ocean, changes in pH and Ca concentration or conditions that favor greenalite or ferrihydrite precipitation may have reduced phosphate availability below experimental predictions.

An alternative solution to the low solubility of phosphate (Pi) may be the conversion to reduced P such as phosphite (HPO32−, P(III)) that is ca. 1,000 times more soluble than phosphate in natural fluids including seawater in the presence of bivalent metals such as Ca, Mg, and Fe(II)2,10,11. Phosphite has been detected in Archean metasedimentary rocks12 and has been utilized by life since at least the Neoarchean-Paleoproterozoic13. Phosphite has also been shown to be more efficient than phosphate in the formation of organophosphorus compounds (phosphonylation vs. phosphorylation)14.

Phosphorylation, the second aspect of the “phosphate problem”, is key to the origin of life as phosphorylated compounds, in particular nucleotides, are a prerequisite for the RNA-world15. Two major abiotic pathways are known to make them. The first pathway involves the formation of nucleosides from ribose and nucleobases followed by phosphorylation of nucleosides16,17,18,19. In the second pathway, ribose phosphate is formed from ribose and a phosphate precursor (dissolved phosphate or apatite), followed by addition of a nucleobase20,21. Although the latter mimics modern biology and is possible in the presence of low dissolved phosphate concentrations, these pathways require borate to stabilize ribose and execute the reactions. The availability of borate may have been restricted to certain parts of the prebiotic Earth as evident from rare reports of elevated boron levels in Archean sedimentary rocks22. In the absence of borate, the first pathway of nucleotide formation is more likely; in that case, polyphosphates would be better phosphorylating agents than orthophosphate as they are more reactive towards organic matter6,14,23.

Polyphosphates are also known as condensed phosphates that include long-chain polyphosphate such as pyrophosphate (PPi), triphosphate (PPPi), tetra- and other higher-order phosphates (PPPPi+), and cyclophosphates such as trimetaphosphate (PPPc) and tetrametaphosphate (PPPPc). Polyphosphates are also considered a possible predecessor of ATP, i.e. the energy currency of all life on Earth24. Cyclophosphates in particular may act both as polymerizing and phosphorylating agents, e.g., they can be precursors of amidophosphate and are capable of phosphorylating glyceric acid, sugars, amino acids, and nucleosides6,23,25. Furthermore, PPPc may react with ammonium to produce diamidophosphate26, which can react with diols, forming cyclic phosphates including ribose 1-,2-cyclic phosphate, a phosphorylated compound that can form canonical ribonucleotides with the direct reaction to nucleobases27,28. Given the importance of phosphite and the polyphosphates, it is imperative to determine if and where these alternative P species could have been produced in significant abundances on the early Earth29.

Three main mechanisms have been proposed for phosphite formation on the prebiotic Earth including (1) water-induced oxidation of schreibersite (i.e. (Fe,Ni)3P)30,31, which was either delivered by meteorites32 or produced in lightning-struck soil33; (2) reduction of apatite during lightning in volcanic ash clouds34; and (3) reduction of phosphate coupled to Fe2+ oxidation under anhydrous conditions at elevated temperature (180 °C)11. Among these mechanisms, Fe2+-induced phosphate reduction has the potential to be the most widespread. A natural example of this mechanism includes the diagenesis of ferruginous sediments11 and serpentinization of ultramafic rocks containing olivine35. There is, however, only one experiment that explored Fe(II)-induced phosphate reduction so far, and it was limited to 180 °C. Little is known about the yields of phosphite formation by Fe(II)-induced reduction under a wide range of geologic conditions relevant to prebiotic Earth, and the role of other reducing agents such as H2 and mineral catalysts, particularly Fe-, Ni-, and Cr-bearing minerals, are unknown.

The formation of polyphosphates, on the other hand, has been demonstrated (1) from pure phosphates (e.g., NaH2PO4/NH4H2PO4) at 80–1200 °C36,37; (2) in high-temperature (>1200 °C) volcanic systems38; and (3) via metallic phosphide dissolution at ambient temperatures39. However, these scenarios would likely have been rare on early Earth and Mars. For example, the pure sodium or ammonium phosphate salt systems may not exist in nature, and temperatures above 1200 °C where they might exist are relatively rare5. More widespread near the surface would have been ferruginous sedimentary environments that underwent thermal effects (diagenesis and metamorphism), similar to those found to yield reduced P11,35. One study conducted at 180 °C under ferruginous conditions demonstrated PPi formation with a yield of 15%11. Hence metamorphic sedimentary settings may be conducive to P polymerization in addition to reduction and may have provided the reactive P species required for the origin of life. However, the formation of higher-order polyphosphates in such settings has so far not been demonstrated, and it is unknown how polymerization yields vary as a function of temperature. Furthermore, it has been proposed that amorphous silica may facilitate polymerization in alkali phosphate systems40,41; however, the effects of other prebiotically relevant minerals on phosphate polymerization in ferruginous environments has not yet been explored.

Here we address this knowledge gap about phosphite and polyphosphate formation in ferruginous settings on early Earth by greatly expanding on the parameter space explored previously11. To mimic these settings, we carried out thermodynamic modeling and experiments and measured P speciation between 85 and 700 °C in acidic ferruginous conditions and in the presence of different mineral substrates and N2 and/or H2 gas (see Supplementary Fig. S1 for the experimental set-ups). We prepared solutions of sodium phosphate and iron(II) chloride (Fe:P = 3:1 molar ratio) and let these evaporate to dryness at 60 °C under vacuum or constant flow of N2 gas, followed by pulverizing and further heating at 170–700 °C under static N2 or H2 or under a flow-through N2 or H2 (5%) atmosphere for up to one week, maintaining an anoxic environment throughout. Magnetite, chromite, pyrite, Ni, Fe, and Ni-sulfide were added to the solution before evaporation (see “Methods” for details). In two static experiments, the gas phase was briefly vented after 3 h and 24 h, respectively, to further enhance water removal. The final solids were analysed by XRD. Phosphorus species were extracted with EDTA-NaOH solutions and analysed by NMR and IC-ICPMS. Our data shed new light on the speciation of phosphorus on the early Earth.

Results

Mineralogy of evaporated and heated residues

The experimental conditions and the major results are summarized in Tables 1, 2, respectively. In general, the evaporated solid was composed of halite (NaCl), rokuenite (FeCl2·2H2O), and an amorphous Fe-PO4 phase, irrespective of evaporation method (vacuum or N2 flow-through) (Fig. 1, Table 2). Crystalline vivianite (Fe3(PO4)2·8H2O) is the most stable phase in the Fe-PO4 system, whose crystallization takes place through an intermediate water-poor amorphous phase (Fe3(PO4)2·4.7H2O)42. In our case, vivianite was not observed, but we detected an amorphous Fe(II)-phosphate phase. We tentatively identify this phase as Fe3(PO4)2·xH2O where x is between 0 to ≤8. The mineralogy of the Fe-/Ni-/Cr-bearing phases remained the same, and they did not affect the composition of the evaporated residues.

Table 1 Summary of the experiments
Full size table
Table 2 Summary of the experimental results
Full size table
Fig. 1: XRD data of selected experimental products.
figure 1

The broad peak from 27 to 35 o (2θ) is shown in the dotted boxes and some representative broad peaks are enlarged in the inset graphs, which represent amorphous Fe3(PO4)2·xH2O (and possibly other amorphous polyphosphate phases for the 170 °C and 200 °C N2 flow-through experiments). At 350 °C, all phases are crystalline; and amorphous Fe3(PO4)2·xH2O appears to be unstable at this temperature. Some phases in the 200 and 350 °C residues could not be unambiguously identified (e.g., the presence of NaFePO4 and Fe2OPO4 is speculative).

Full size image

The mineralogy of the heated residues at 200 °C under static N2 or H2 was broadly similar to that of the evaporation stage (Fig. 1). In contrast, the N2 flow-through experiments at 170 and 200 °C contained anhydrous FeCl2 instead of rokuenite (Fig. 1). The 350 °C heated residues from static N2/H2 and N2 flow-through experiments lacked amorphous phases and were instead dominated by Fe-hydroxide (Fe(OH)2,), Fe-chlorides (FeCl2 and FeCl2·2H2O), Fe-phosphates (possibly NaFePO4 and an unidentified Fe-PO4 phase), NaCl, and other unidentified phases (Fig. 1). The added Fe-/Ni-/Cr-bearing phases remained mostly unchanged. In H2-flow experiments, Fe, NaCl, and Fe-phosphates (most likely NaFePO4 i.e. maricite and an unidentified Fe-PO4 phase), and other identified phases were produced at 600 °C, while Fe and Fe3(PO4)2 are observed at 700 °C (Supplementary Fig. S2).

P speciation in solution

After the evaporation step, most P remained in the form of phosphate. Less than 2% PPi was generated, and no other polyphosphates or phosphite were detected (detection limits are 0.05% and 0.001% of initial P, respectively). In contrast, further heating yielded variable proportions of polymerized P and small but detectable amounts of reduced P species (Fig. 2, Table 2).

Fig. 2: Proton decoupled 31P NMR data of EDTA-NaOH extracts.
figure 2

Box A and B are enlarged in inset graphs. Chemical shifts are presented relative to an 85% H3PO4 standard having a chemical shift of 0 ppm. Pi and PPi have chemical shifts of 5 and –6.2, respectively. PPPi appears as doublet in between –5.3 and –5.4 ppm and as a triplet in between –19.7 and –20.0 ppm. PPPPi+ has multiple peaks in between –5.1 to –5.3 ppm and between –20.1 and –21.2 ppm. PPPc and PPPPc have single peaks at –21.5 ppm and –23.7 ppm, respectively. Slight variation in the chemical shift is due to the pH of the solution (we noticed that the dependency of Pi on pH is more intense compared to other condensed P species). All analyses were performed within 1–2 days after the experiments, except for the sample marked with an asterisk, which was stored at room temperature for 50 days before analysis.

Full size image

Phosphate polymerization and speciation were found to be highly dependent on temperature. The highest yield of polymerization (70%) and the largest number of polymerized P species were detected at 170 °C with 19% PPi, 24% PPPi, and 26% PPPPi+ (Figs. 2, 3A, B). This is the only experiment where penta- and possibly hexapolyphosphate were formed (Fig. 2). A second analysis of this solution, after 50 days of storage at room temperature, showed near identical total yields (70%); however, the speciation had changed to 61% PPi, 5% PPPi, 4% PPPc, and most likely 0.3% PPPPc38 (Figs. 2, 3B). Polymerization yields and diversity decreased in experiments above 170 °C (Figs. 2, 3A). For example, at 200 °C, the average yield was 44% with 38% PPi, 5% PPPi, and 1% PPPPi+ while at 350 °C, the yield decreased to only 0.2% PPi. Polymerization was undetectable at 600 °C and 700 °C (Figs. 2, 3C).

Fig. 3: Experimental polymerization and reduction yields relative to the initial phosphate input.
figure 3

Red horizontal line = detection limit of the IC-ICPMS method. Typical error of repeat experiments or in measurements is given where P species are detected. 600 and 700 °C experiments were conducted in a tube furnace unlike other experiments conducted in hydrothermal reactors and marked with gray column. A Polymerization yield as a function of temperature under gas flow-through conditions. B Change in the proportions of differing condensed species formed at 170 °C after 50 days of storage at room temperature (labeled 170 °C*). Proportions of PPi, PPPc, and PPPPc have increased in concomitant reduction of PPPi and PPPPi + . C Polymerization yield as a function of catalysts (200 °C, H2-static experiments). D Reduction yield in 200 °C H2-static experiments. E Reduction yield in 300 °C H2-static experiments. F Phosphite yields as a function of temperature in the absence of catalysts with N2 flow (175–350 °C) and H2 flow (600–700 °C).

Full size image

Polymerization yields were higher when water could escape more efficiently during heating. N2 flow-through experiments where water vapor was constantly flushed out showed higher yields compared to static N2 experiments (e.g., at 200 °C, yields were 44% and 15% under flow-through and static conditions, respectively). Similarly, the 200 °C static H2 experiment, where the gas was vented briefly after 24 h, shows a higher polymerization yield (56% with 44% PPi, 9% PPPi, and 1% PPPPi + ) compared to when venting occurred after 3 h (total 19.3% with 19% PPi and 0.3% PPPi).

Mineral substrates impacted the polymerization yield. The presence of chromite and Fe/Ni or their sulfides increased yields up to 10 times at 200 °C, while magnetite supressed polymerization (Fig. 3C; Supplementary Fig. S3). The effect of these minerals became negligible at 350 °C.

Regarding phosphate reduction to phosphite, the type of reducing agents (Fe2+, and H2) as well as temperature exerted strong controls. Under N2 flow-through conditions, where Fe2+ was the reducing agent, phosphite was not detected at 80–200 °C, but it became detectable at 350 °C (yield 0.075%). When H2 was added, phosphite was detectable at 200 °C, particularly in the presence of other minerals, reflecting the ability of H2 to act as an additional reducing agent (Fig. 3D, Supplementary Fig. S4). On the other hand, a direct comparison between 200 and 300 °C N2 flow, between 200 °C and 350 °C static H2, and between 600 and 700 °C H2-flow experiments shows a phosphite yield increase of 75, 15, and 100 times, respectively, revealing enhanced reduction at elevated temperature in each pair (Fig. 3D–F). We stress that the 600 and 700 °C experiments were conducted in a different setup and were therefore not directly comparable to those at ≤350 °C, but the temperature dependence was confirmed.

The Fe-/Ni-/Cr-bearing minerals also impacted the phosphite yield, and the effect of these minerals is dependent on temperature. At 200 °C, the highest phosphite yield of 0.004% was obtained in the presence of chromite (Fig. 3D). Lower yields of 0.002% were obtained in the presence of Ni and Ni3S2. Phosphite was not detected in experiments without these minerals and when magnetite or pyrite were used (Fig. 3D). At 350 °C, the maximum yield of 0.058% was obtained when Ni and Ni+Fe were used as catalyst (Fig. 3E). Ni3S2- and chromite-bearing samples showed moderate yields of 0.015% and 0.008%, respectively, while magnetite had no effect (Fig. 3E).

For the thermodynamic modeling, we note that specific thermodynamic data for Fe-P phases (polyphosphates and reduced P compounds) are unavailable, and thus the modeling is constructed for the simpler H-P-O system (Fig. 4). Therefore, the graphs are illustrative but not exactly reflective of the experiments. For example, the calculations suggest the stability of PPi up to 700 °C in the H-P-O system, which is not the case in the experiments (Figs. 2, 4A). However, the thermodynamic model predicts that polymerization occurs best at low water activity and reduction is best facilitated by high H2, consistent with the experiments (Fig. 4A). The modeling also suggests that formation of cyclophosphates may be feasible at lower temperatures if the water activity is low (Fig. 4B).

Fig. 4: Equilibrium calculations for phosphate polymerization and reduction.
figure 4

Pressures are in atm. The calculations are based on the data from Pasek69 at 25 °C and from 0 to 700 °C using data from HSC Chemistry70. In all calculations, the activity of the P species on a given line was assumed to be equal. A Using HSC Chemistry, the predicted domains where phosphate (as H3PO4) polymerizes into pyrophosphate (H4P2O7) and phosphite (as H3PO3) is a function of H2O and H2 activity. It is also changed by the temperature of the system, which in this graph changes from 0 to 700°C. These species were all considered as acids, to avoid pH effects. B Using data from Pasek69 on prebiotic phosphorylating agents, the regions where various polyphosphates, phosphite, and pyrophosphate would be expected to dominate at pH 7 is likewise a function of H2O and H2 activity. Notably, when linear polyphosphates become larger than 3–4 phosphate units, cyclization is spontaneous.

Full size image

Discussion

New constraints on phosphate reduction and polymerization

Our results provide several new insights into the formation of reactive P species that may have fueled prebiotic chemistry on the early Earth:

(1) Polymerization of phosphate in P-rich ferruginous conditions maximises at moderate temperature and becomes limited at lower and high temperatures. Our results disprove the hypothesis of increasing polymerization yield with increasing temperatures (above 200 °C)11 and suggest that moderate temperatures (80–200 °C) are optimal for polymerization in the Fe-PO4 system. The reason for the temperature dependence is likely linked to Fe-mineralogy. Previous studies noted that polyphosphates tend to crystallize as amorphous phases36,43, but we noted the formation of crystalline phosphates at temperatures ≥350 °C (Fig. 1), which may explain lower polymerization yields. Also, the Fe-phosphate precursor likely has an impact. A previous study found that vivianite loses water sequentially at 100–500 °C but does not form any polymers44. Likewise, polymerization was not observed when we conducted experiments with vivianite at 200 °C (not shown). Instead, a meta-stable amorphous Fe-phosphate phase (Fe3(PO4)2·xH2O), as generated during evaporation in our experiments, may be a prerequisite for polymerization.

(2) Polymerization yields increase in the presence of Fe-/Ni-/Cr-bearing minerals except for magnetite, which decreases polymerization. It has been previously shown that silica stabilize phosphate on its surface, thereby facilitating polymerization upon subsequent heating40,41. Most likely, amorphous Fe-phosphate precipitated on the surface of these minerals (except for magnetite), which brought the phosphate molecules closer together, enhancing Pi-Pi interaction and polymerization. In the case of magnetite, which contains Fe3+ and is known to strongly absorb phosphate45, the interaction between the surface and Pi might have overpowered the Pi-Pi interaction, thereby reducing the Pi-Pi interaction and the polymerization yield. Alternatively, magnetite might have enhanced the hydrolysis of polyphosphates in the EDTA-NaOH solution, as has been described for other Fe-oxides46, lowering polymerization yields.

(3) Our data provide the first evidence that cyclophosphates and long-chain polyphosphates (up to pentaphosphate) can form in naturally relevant ferruginous conditions. The long-chain phosphates converted into cyclophosphates (PPPc and PPPPc) at room temperature in contrast to previous reports of the same happening at 100 °C in a P4O10-rich solution38. Higher thermodynamic stability of cyclophosphates compared to linear phosphates suggests that cyclization can occur spontaneously (Fig. 4B). Our PPPc yield is similar to that reported in high-temperature volcanic environments38, in phosphide oxidation reactions39, or in the ammonium phosphate ± urea system37, suggesting an additional and perhaps more naturally widespread route for PPPc formation on the early Earth. PPPPc minerals have been hypothesized to be the product of pyrolytic oxidation of metallic phosphides at >300–400 °C, and the mechanism might have formed cyclophosphates on the early Earth47. Our data reveal an additional route at lower temperatures that does not require precursor phosphide minerals.

(4) Iron(II)-induced phosphate reduction to phosphite is more difficult than previously reported11, but on the other hand, some degree of Fe(II)- and H2-induced reduction can occur over a wider temperature range (200–700 °C) than previously explored. Herschy et al.11 obtained a Fe(II)-induced reduction yield of 4% at 180 °C while we did not see any detectable phosphite at 170–200 °C although experimental setups were broadly similar. Our maximum yield was 0.1% at 700 °C, which probably is a combined effect of Fe(II)- and H2-induced reduction. Herschy et al.‘s (2018)11 thermodynamic calculations predicted that phosphite yields increase with temperature and Fe2+ is a more effective reductant than H2 at 100–500 °C11. Our data support the former conclusion and show that the latter occurs at temperatures around 350 °C. However, at 200 °C, H2 appeared to be a better reducing agent compared to Fe2+, particularly if Ni, Ni-sulfide, and chromite are present. These minerals are known to catalyze the reduction of carbon and nitrogen48 and may have a similar catalytic effect on phosphate reduction. If so, it would indicate kinetic inhibition of phosphate reduction in the absence of these catalysts. Such a kinetic effect may also explain why elevated temperatures results in higher reduction yields, as this may help overcome the kinetic barrier. One possible reason for absence of a catalytic effect in magnetite may be the presence of Fe3+, which perhaps changes the redox state of the Fe-P system in favor of phosphate11.

(5) Lastly, the formation of phosphite is not precluded by hydrous mineral phases. Phosphite is thermodynamically stable under highly reducing conditions below the stability field of water, that favor H2 over H2O10. This likely explains the temperature-dependence of phosphite formation, as H2O is lost with increasing temperature. However, some Fe(OH)2 and FeCl2·2H2O were still present in experiments where phosphite formed, indicating that phosphite can form as long as residual water is locked into minerals.

Implications for early Earth

The conditions we explored here are relevant to a range of settings on the early Earth and can therefore help to constrain the availability of reactive P species for the origin and early evolution of life. The formation of phosphite induced by concomitant Fe(II) oxidation can take place during burial or contact metamorphism at 350 °C or higher in P-bearing ferruginous sediments, which would have been widespread in the Archean, when the ocean was ferruginous with possibly millimolar concentrations of Fe2+ 49,50. Herschy et al.11 reported phosphite from Archean Isua and Akilia rocks in Greenland (3.7-3.8Ga) and provided several possibilities for its origin, including diagenesis and metamorphism. As we see a strong correlation between the phosphite yield and temperature in our experiments, and as the Isua and Akilia rocks went through amphibolite and granulite facies metamorphism, respectively, we suggest that a significant part of the phosphite might have been produced during metamorphism, i.e., at higher temperatures than that explored experimentally by Herschy et al.11. Importantly, the phosphite/phosphate ratio is higher in granulite facies Akilia rocks (3.24) compared to amphibolite facies Isua rocks (0.31–0.84), which further supports the temperature dependence of phosphite formation. In these localities, phosphate may have co-precipitated with iron3 and formed Fe-phosphate phase during diagenesis50,51,52. Subsequently, Fe(II)-induced reduction of phosphate may have happened during metamorphic heating. We note that our and the previously reported experimental phosphite yield are significantly lower than that in natural rocks, which further attests to the kinetic barrier of the reduction reaction. This barrier is likely overcome with time. While our experiments were very P-rich, high P concentrations are probably not required to drive phosphate reduction, as long as conditions are sufficiently hot and anhydrous and reductants such as Fe(II) or H2 are present. In fact, the formation of amorphous Fe-phosphate, most likely the type that formed in these experiments, is preferred during diagenesis in ferruginous conditions with low P concentrations52. Hence this reaction could have occurred in sediments deposited in marine environments, which were perhaps P-depleted53, but Fe2+-enriched49 on the early Earth.

Our results provide an explanation for the detection of phosphite in serpentinite35 and suggest that it may be a common occurrence. Serpentinizing environments are highly reducing environments that produce a significant amount of H2 as well as metallic nickel and iron as awaruite along with nickel sulfide and magnetite54,55. Our results provide experimental evidence that in the presence of Ni, Fe, nickel sulfide, and chromite (usually present as a primary phase in ultramafic rocks), H2 may act as the main reducing agent for phosphate during serpentinization, particularly at temperatures below 350 °C in which H2 is a better reducing agent compared to Fe2+. H2 production during serpentinization can be variable but it has been shown to reduce C into CH4 via Fisher-Tropsch-Type reaction during serpentinization56, and may thus be sufficient for phosphate reduction as well. Dry conditions are required for phosphate reduction, consistent with serpentinization settings where H2O is removed during conversion of olivine into serpentine minerals55,57. We speculate that such dry conditions can be achieved in deeper part of the crust where the serpentinization may happen at high temperatures at very low water:rock ratios, particularly at micron scale. At comparatively lower depths, where the water:rock ratio may be higher, the reduction of phosphate into phosphite is perhaps suppressed. We further note that magnetite may interrupt the formation of phosphite while chromite, metallic Ni, and Ni-sulfide may accelerate it, which may suggest that H2-induced reduction during serpentinization can be heterogeneous. The presence of water (i.e., high water:solid ratio) or magnetite may explain the absence of phosphite in certain serpentinite muds reported in Pasek et al.35.

Regarding phosphate polymerization on the prebiotic Earth, relatively high initial phosphate concentrations are probably required. Dissolved phosphate concentrations in the Archean ocean are debated with estimates varying from sub-micromolar levels3 to up to 1 mM, i.e. 1000 times higher than today9. If seawater phosphate levels were low, then volcanic lakes and hot-spring pools may have offered an alternative setting with high dissolved phosphate levels, like their modern analogs such as Lake Magadi in Africa and thermal springs on the Kamchatka Peninsula7,58. Here, volcanically-heated subaerial hot-spring environments with lakes and hydrothermal pools59,60 may have offered ideal conditions for phosphate polymerization as identified in our experiments (Fig. 5). As noted earlier, the solubility of bioessential micronutrients such as Cu and Zn may be low in high-pH lakes, but once polyphosphates are formed, they may be exported to other environments with a higher overall metal inventory. The abundance and extent of these subaerial hot-spring environments on the early Earth are unknown; however, their occurrence is recorded in 3.5 Ga rocks on the Pilbara Craton59, indicating the natural relevance of this type of environment. The concentration of dissolved phosphate in such settings is dependent on physicochemical conditions, particularly pH, bivalent cations and carbonate concentrations, with alkaline conditions favouring phosphate enrichment7. However, acidic hot spring pools, such as those on the modern Kamchatka Peninsula, also show high phosphate levels reaching up to 1 mM along with high concentration of Fe (reaching an Fe:P ratio up to 8)58, a condition that mimics our experiment. Modern-day volcanic lakes and hydrothermal pools experience wetting-drying cycles and variability in temperature and pH61, which would likely also have happened on the early Earth. Evaporation enriches dissolved phosphate7,9,58 and potentially precipitates Fe-phosphate50,52, similar to what we observed in the evaporation stage (Fig. 5B). These lakes or pools likely contained other bivalent metals such as Ca, Mg, and other anions such as carbonate and silica, which might have impacted the mineralogy of the evaporated residue. Although exploring these parameters was beyond the scope of the study, formation of Fe-PO4 phases is likely, particularly because volcanic environments are typically Fe-enriched and formation of amorphous Fe-PO4 may take place in low-P ferruginous conditions conducive to greenalite formation50,52. Moreover, the formation of Fe-phosphate may happen after deposition and during diagenesis as observed in ferruginous rocks51,52. Subsequently, magmatic intrusions or lava flows could have triggered thermal metamorphism of these sediments and evaporites29, producing the wealth of polyphosphate species that we observed in the experiments. Mineral catalysts including magnetite, chromite, pyrite, Ni, Fe, and Ni-sulfide derived from the (un-)altered bedrocks could have accumulated in these lakes or pools62, which likely enhanced the polymerization process during the thermal heating stage. Importantly, phosphite formation may also occur in the same environment along with polyphosphates if temperatures reached above 200 °C and would have enhanced if H2 produced by hydrothermal alteration of (ultra-)mafic rocks55 was available. Furthermore, in saline lake environments, solar heating alone may achieve temperatures as high as 95 °C63. In such cases, evaporation of phosphate-rich solution may lead to some degree of polymerization, enhanced by the catalyst minerals that we used in our experiments. In summary, temperature gradients, reducing agents (Fe(II) and possibly H2), and potential catalysts make hot-spring settings and saline lakes attractive for phosphate reduction and polymerization on the early Earth.

Fig. 5: Schematic of the proposed mechanism of phosphate reduction and polymerization on early Earth.
figure 5

A Possible routes of phosphate, Fe2+, and catalyst mineral accumulation in wet-dry pools by hydrothermal circulation and rock weathering29,32,58,59,62. B Formation of evaporites or sediments containing Fe3(PO4)2.xH2O in these pools. Thermal metamorphism may ensue as a consequence of magmatic heating, driving phosphate polymerization and/or reduction, depending on temperature, reductants, and catalyst availability. C Subsequent re-dissolution of the products and release of polymerized and reduced P species back into the environment11,43.

Full size image

Liberation of reduced and polymerized P species into water is the next important step for phosphorylation reactions for the origin of life. Phosphite is ca. 1000 times more soluble than phosphate11, and hence hydrothermal rehydration of metamorphosed ferruginous sediments that have undergone phosphate reduction would have created a plausible pathway for dissolved phosphite back into the environment (Fig. 5C). Similarly, phosphite produced during serpentinization may be released into water as evident by the lower P content in altered ultramafic rocks compared to its unaltered precursor35. On the other hand, the solubility of the various polyphosphate species in natural fluids is not well known, especially under ferruginous conditions. However, the presence of dissolved H2S, which may be abundant in volcanic settings, can enhance polyphosphate solubility by scavenging Fe2+ as iron sulfides43. Additionally, Fe-chelating organics, such as citrate, which were possibly present on the prebiotic Earth64, could have enhanced the release of polyphosphates from meta-sediments or meta-evaporites (Fig. 5C). Upon re-dissolution, straight-chain polyphosphates undergo hydrolysis and transform into shorter chain (poly)phosphates and cyclophosphates, which subsequently hydrolyze into phosphate36,43,65,66. Previous studies suggest that the half-life of these poly- and cyclophosphates in water may vary from days to thousands of years, depending on pH, temperature, and the presence of bivalent cations36,65,66; and this time may be sufficient for some phosphorylation reactions5,6,23. If a sufficient amount of the straight-chain polyphosphates and cyclophosphates accumulated in a pool or lake that contained organic compounds including urea and/or fomamide, subsequent wet-dry cycles may produce a wealth of phosphorylated compounds via different reaction routes. For example, if borate were also present along with ribose in these settings, then nucleotide formation via reactions between ribose-phosphate and nucleobase is likely20,21. In the absence of borate, nucleotide formation via reactions between nucleosides and reactive P-species, including PP, PPP, PPPc, PPPPc, and phosphite, could have occurred6,16,17,18,19. Furthermore, PPPc may transform into amidotriphosphate or monoamidophosphate upon reaction with dissolved ammonium, which may subsequently facilitate nucleoside phosphorylation25,27,28. To summarize, the alternative P species that were formed in our experiments can be useful for the prebiotic phosphorylation of organic molecules, which is key step for the origin of life.

In conclusion, we find that thermal metamorphic processes on ferruginous sediments and evaporites precipitated in phosphate-rich volcanic environments on the early Earth and possibly on Mars could have created an important route for the formation of reactive P species, including polyphosphates and phosphite. Our findings include a naturally relevant pathway for the formation of cyclophosphates, which may have been particularly reactive towards organic matter in prebiotic reactions23. Phosphite formation may have occurred also in other P-depleted metamorphic settings and during serpentinization, and it may have been accelerated in the presence of catalytic minerals. Our identified polymerization pathways may have been more common than previously proposed alternatives and may therefore represent a new environmental solution to the P-problem in the origin of life.

Materials and methods

Experiments

Acid-washed and baked (500 °C) glass containers were used in all the stages of the experiments and subsequent sampling. FeCl2.4H2O, NaH2PO4, and catalysts (Ni, Fe, Ni3S2, Ni+Fe (3:1), FeS2, Fe3O4, FeCr2O4) were added in a molar ratio of 3:1:0.5 (or 1) in deoxygenated, deionized water with a pH of 4 to produce weakly acidic solution representative of hydrothermal pools (this design was initially adopted from Herschy et al.11). Control experiments without catalysts were also performed. Among these chemicals, FeCl2.4H2O, NaH2PO4, Ni, Fe, Ni3S2, and Fe3O4 were synthetic and bought from Fisher Scientific while FeS2 and FeCr2O4 are natural samples. All the catalysts were used in powdered form and analyzed using XRD before experiments. We did not see any impurity in them except for chromite, which contains minor olivine. More importantly, we did not see any detectable amount of phosphite or polyphosphate in any of the starting materials including the catalysts. In a typical experiment, we took 400 ml of water and added 1 M HCl dropwise to make a pH of 4. The pH was measured using the Thermo Scientific pH paper strips. Then, we added 2.3857 g FeCl2.4H2O, 0.47992 gm, NaH2PO4 and the known amount of catalyst (weight equivalent of 5 or 10 mM) to make a solution containing 30 mM Fe(II) and 10 mM phosphate. In a few experiment, we took 300 mM Fe(II) and 100 mM phosphate at this stage; however, we did not see any difference in the products (tested in XRD) data after evaporation. The solution was stirred for 40 minutes in anoxic conditions (purged with N2 gas) that led to the dissolution of FeCl2 and phosphate salts. The Fe-/Ni-/Cr-bearing minerals remained the same before and after the evaporation stage as evident from XRD data. We cannot completely rule out the possibility that there could be minor dissolution of these catalysts, but the bulk of the mass remained in a solid state based on visual inspection. In one experiment (700 °C, described below), we used NH4H2PO4 (Fisher Scientific) instead of NaH2PO4 as the phosphate source.

The solution was then loaded into a Berghof reactor equipped with a heating mantle, a K-type thermocouple for temperature monitoring, gas lines (H2 and N2), and a vacuum pump (Supplementary Fig. S1A). After loading the solution, an anoxic environment was created inside the reactor using the vacuum pump and by flushing the system with N2 followed by evacuation (5 cycles of N2 flushing and evacuation). The heat source was turned on and the solution was evaporated at 60–65 °C either under vacuum or constant N2-flow to maintain anoxic conditions to replicate the evaporation of those hydrothermal pools under an anoxic Archean atmosphere. The evaporated residue was then cooled down to room temperature and then quickly crushed in an agate mortar and pestle (see Herschy et al.11) and reloaded into the reactor (Berghof or Parr) to be dry-heated for up to 7 days at temperatures of 170, 200, and 350 °C, either in static or flow-through N2/H2 environments (see Table 1 for the initial conditions of the experiments). In two gas-locked experiments, the gas was briefly vented from the reactor after 3 h or 24 h, respectively. This step was performed to get rid of water that would have built up within the headspace of the reactor within those first few hours. This experiment was done to further explore the effect of water on phosphate reduction and polymerization11,43.

Two high temperature (600 and 700 °C), H2-flow (5% H2 + 95% N2; flow rate of 20 ml/min) experiments were performed in a tube furnace where the evaporated residue was loaded into an alumina boat, a K-type thermocouple was used to control temperature, and a ramp of 5 °C/min was used to reach the desired temperature, which was maintained for either 1 hour (700 °C) or 2 h (600 °C), followed at 5 °C/min (Supplementary Fig. S1B). Heated residues were cooled down and stored in N2-environments to avoid oxidation before subsequent analysis.

Solid characterization using powder X-ray diffraction (PXRD)

The evaporated and heated solids were powdered with pestle & mortar and part of them was loaded into 0.5 mm capillary tubes for XRD analysis. The PXRD patterns were recorded on a STOE STADIP diffractometer using Mo Kα1 radiation at room temperature from 2.5o to 35° (2θ) with a scan rate of 2.5–3.5o (2θ)/step in capillary Debye-Scherrer mode. The PXRD data were compared to solids in the Inorganic Crystal Structure Database (ICSD) for phase identification using the Crystal Diffract software (version 6.9.3).

Quantification of P species in the solution

Part of the solid was treated with Ethylenediaminetetraacetic acid-sodium hydroxide (0.05 M EDTA and 0.25 M NaOH solutions were prepared separately and mixed in 1:4 ratio) solution35 maintaining a solid:solution ratio of 1:200 or 1:50 for 3–15 h. Higher solution volume was used for experiments where Ni-/Fe-/Cr-bearing minerals were not used to make sure to extract maximum amount of P and get rid of all the dissolved Fe2+. After this treatment, an additional 0 to 30% EDTA-NAOH solution with respect to the initial volume was added to the solutions to precipitate all the remaining Fe2+. This step is essential because iron may precipitate as oxides in the separation column of the ion chromatograph (IC) and bind phosphate by adsorption, thereby impacting analytical quality. It also hampers the magnetic resonance spectrum in nuclear magnetic resonance (NMR). The solution was centrifuged several times to obtain a clear solution where all iron had precipitated to the bottom of the centrifuge tube.

For the phosphite measurements, we used the IC-ICPMS set-up of Baidya and Stüeken67 modified from Ivey et al.68. A Thermo Scientific Dionex ICS-6000 ion chromatograph equipped with a Dionex AS-AP autosampler, a 25 mm Dionex IonPac AS17-C separation column (2 mm bore), a 25 mm Dionex IonPac AG17-G guard column (2 mm bore), and a Dionex ADRS 600 2 mm Suppressor Module was used to separate the phosphite fraction from phosphate and other P species in the solution. The flow rate was held constant at 0.5 ml/min while the concentration of the KOH eluent solution was ramped up from 1 mM to 40 mM over 20 min and then this maximum KOH concentration was held constant for another 20 min. This IC configuration was capable of separating phosphite from phosphate and pyrophosphate. Suppressor outlet of the IC was physically connected to 1 ml/min nebulizer attached to the spray chamber (Scott model; quartz glass) of a Thermo Scientific Element2 ICP-MS operated in medium resolution mode to separate 31P from HNO interferences with the same mass. Data were collected in the ICP-MS for 3 minutes with a chromatographic method centered on the m/z of 31P (one minute for monitoring the pre-phosphite background, one minute for the phosphite peak, and one minute for monitoring post-phosphite background). The chromatographic data from the ICP-MS were smoothened with the Origin Lab software, using the fast furrier transform filter with a points-of-window value of 5–6, and the peak area under the curve was used for quantification of phosphorus. A calibration curve was generated using phosphite standards (Na2HPO3.5H2O; Fisher Scientific®) ranging from 0.1 ppb to 50 ppb P and used to calculate phosphite concentration in the experimental samples. The detection limit of the IC-ICPMS was <0.1 ppb. This corresponds to ca. 0.001% of the total phosphorus present in the experiment.

Other polyphosphate species including pyrophosphate (PPi), triphosphate (PPPi), trimetaphosphate (PPPc)38, tetrametaphosphate (PPPPc)38, tetra- and other higher order polyphosphates (PPPPi+) were measured on a Bruker AVIII 500 MHz nuclear magnetic resonance (NMR) instrument equipped with a nitrogen-cooled broadband cryoprobe. Sample solutions and standards were added with 10-20% D2O to make a total volume of 0.6 ml and analyzed in proton-decoupled mode with 3000-7000 scans. The 31P chemical shifts are referenced to phosphoric acid having a chemical shift of 0δ. Standards of known concentrations (0.1 ppm to 800 ppm phosphorus, generated from; K4P2O7, Na5P3O10, and Na5P3O9; Fisher Scientific) were analysed to determine the detection limit and build calibration curves for each P species. We were not able to create calibration curves for PPPPc and PPPPi+ because relevant standards could not be purchased. Instead, we tentatively used the calibration curve of PPPi. The detection limits of the various polyphosphate species was 0.05% relative to the total phosphorus present in the experiment.

Thermodynamic modeling

Thermodynamic relationships between various species of H-P-O compounds were determined using data from Pasek 69 and from data in the HSC Chemistry program70. For the latter, the balanced reactions involving H3PO4(aq), H4P2O7(aq), and H3PO3(aq) were constructed. From these, the partial pressures of H2O(g) and H2(g) were determined for the line of equivalence of each as a function of temperature (e.g., where [H3PO4] = [H4P2O7]), from 0 to 700 °C. All species were considered to be aqueous solvated acids even at high temperature, despite the fact that the system was investigated under dry conditions, mainly because the data was most available for these species. Because all three species were solvated, error introduced by operating outside the range of temperature of liquid water stability should be smaller than if a species were present as a anhydrous acid (e.g., H3PO4).

Using the shape of this diagram, data from Pasek 69, which also includes triphosphate, cyclic triphosphate, polyphosphates, and the dimer of phosphite called pyrophosphite, was used to determine stability regions for various P species. Calculations were at 25 °C and a pH of 7. As before, the lines of equivalent concentrations for each species were determined as a function of the partial pressure of both H2O(g) and H2(g). No counter ion (e.g., Fe2+) was considered in these calculations, as the data was not available for those species. Data for the triphosphate, tetraphosphate, pyrophosphite, and cyclic trimetaphosphate were also not available as a function of temperature, hence all calculations were performed at 25 °C.