## Introduction

The mineralogy of the element phosphorus (P) is generally perceived as being limited to phosphates under terrestrial conditions. Cosmically, P-mineralogy includes phosphide minerals in which phosphorus mainly forms binary compounds with a metal such as Fe or Ni. The mineral schreibersite, (Fe,Ni)3P, is an example of such a binary solid and is cosmically the most abundant phosphide. Terrestrial occurrences of schreibersite and other phosphides have since been discovered within pyrometamorphic rocks1,2, and especially within fulgurites3,4, which are glasses formed by cloud-to-ground lightning. The presence of phosphides within fulgurites parallels the existence of phosphides within impact rocks such as tektites5, with impact cratering also being the likely route to forming phosphides in lunar rocks6, and in impact melts on other meteorites7,8.

Phosphorus has thus far been found in mineral form in only two oxidation states: as +5 in phosphates, and as ~−1 in phosphides. This latter oxidation state is based on the binding energy of inner electrons of the P atom in phosphides9, which is confirmed by ab initio calculations10. Notably absent from known P-mineralogy are any other oxidation states between these, wherein the P anion is a reduced phosphorus oxyacid. Salts with P in a + 1 oxidation state (hypophosphite), +3 oxidation state (phosphite), and +4 oxidation state (hypophosphate) have been synthesized but are not known from naturally-occurring geologic samples. Despite the apparent lack of reduced phosphorus oxyacids in mineral form, reduced phosphorus compounds are not rare in the geo-environment. For example, the gas phosphine (PH3) has also been recognized to be a trace gas in many environments11,12,13, and has even been implicated as a biosignature gas on Venus14. More specifically to the case of potentially mineral-forming P-compounds, between 1 and 10% of all microbially lineages can use phosphite (and often hypophosphite) as sole P sources15, demonstrating these ions have sources in the environment16. In addition, reduced phosphorus oxyacids have been found in geothermal spring water and anaerobic water in Florida where their origin is likely microbial17,18. Within rock samples phosphite has been found in extracts of fulgurites19, in serpentinites20, and in extracts of Archean rock samples21,22, though the mineral source of such a phase is unclear. Therefore, reduced P oxyacids must exist in solid form of some sort in the environment, either as a pure compound such as CaHPO3, or with the phosphite exchanging for other ions such as phosphate or sulfate. The failure to identify a specific mineral form of reduced phosphorus oxyacid in the environment is problematic given the widespread evidence for phosphorus redox in biogeochemistry16, but may be due instead to an incomplete search for such material.

Due to their rapid formation which results in lithologies that may vary significantly with respect to temperature, pressure, and fO2 over a few millimeters of distance, fulgurites are an ideal material in which to search for reduced P-compounds. Phosphides such as schreibersite have been identified in multiple fulgurites: from Michigan3, Pennsylvania19, and Illinois4. In addition, even more reduced material such as iron silicides are recognized in multiple fulgurites3,23,24,25,26,27,28,29. Such material is presumed to form from the intensely reducing environment possibly generated by a combustion reaction of in situ organics3,30. Pasek and Block19 also reported phosphite in extracts of fulgurites, which were attributed in one fulgurite to phosphides, which react with water to release phosphite as a metastable solute31, but in others the source of phosphite could not be wholly identified.

Here we identify (1) a novel form of crystalline P within a fulgurite from New Port Richey, Florida, USA; and (2) propose that the reduction of phosphate to more reduced forms takes place through at least two separate pathways and that both may be relevant to the planetary distribution of phosphorus32,33. The New Port Richey (hereafter NPR) fulgurite was obtained from private sellers who found the fulgurite on their property in a residential neighborhood in New Port Richey, Florida (28.248 N, 82.718 W) the summer of 2012 after a thunderstorm. We then purchased the fulgurite in late 2012. The fulgurite formed in a partially drained sandy soil, composed primarily of quartz sand and clay34, and was about 500 g in total mass. In addition, as the area receives heavy rain, roots are often coated with an iron oxide plaque35,36,37, which is composed of a 1–2-mm-thick rind of iron oxide that cements quartz sand particles (Supplementary Fig. 1).

The fulgurite is a type II fulgurite25 with a massive, glassy wall that surrounds the central void (Fig. 1a). The fulgurite’s color ranges from white to dark gray. The most unusual feature of the fulgurite is the presence of large (~1 cm diameter) spherules (Fig. 1b) of a material with a dull gray metallic luster that are embedded in the glassy matrix of the fulgurite. These spherules have a density of 4.9–5 g/cm3 and are not attracted to a magnet.

## Results

The New Port Richey fulgurite bears both silicate and reduced (silicide) compositional regions. The silicate fraction of the New Port Richey fulgurite is volumetrically more common and is composed of glass with two major compositions (Fig. 1c): one exclusively SiO2, likely relict quartz sand grains, and the other composed mostly of Ca, Al, and Si oxides (Supplementary Table 1). Within the silicate lithologies are small droplets (<15 µm) composed of P-rich iron metal (Fig. 1d). Such droplets have been reported previously in other fulgurites3,4,19.

The large metal spherules within the NPR fulgurite consist of silicides, specifically with the composition of FeSi and FeSi2, in addition to a material that is darker in contrast in back-scattered electron (BSE) imaging that bears Ca and P and is amorphous to X-ray diffraction (Fig. 1e, f). The silicides also bear 0.45–1.25 wt.% P. The groundmass of the spherules is composed of FeSi2 (84% by volume), in which euhedral crystals of FeSi (11%) grow, often together with the CaP material (5%). The nominal cationic composition of the spherule is thus Si63Fe33Ca2P2. The CaP material bears vermiform FeSi with a width of ~1 µm. The CaP material is most closely associated with the FeSi material, which grows as euhedral crystals (30–70 µm) that are larger than the average FeSi crystals in the FeSi2 groundmass, or more compact (with less extreme maximum compared to minimum dimensions). Though the CaP material is in contact with the FeSi2 groundmass, crystal growth of FeSi2 within the CaP material is less common. Inside the CaP material there are very rare (less than 20 µm sized) crystal fragments that turned out to be a new crystalline phosphite (see the dashed circle in Supplementary Fig. 2).

Given the similarity in chemical composition and unit-cell values, the structure of the new phosphite was refined starting from the atom coordinates given by Phillips and Harrison38 for the P43212 structural model of synthetic CaHPO3. Phillips and Harrison38 describe the three-dimensional structure of synthetic CaHPO3 as having CaO7 capped distorted trigonal prisms, with pseudopyramids of HPO3 sharing corners and edges. The formula of the phosphite phase in the NPR fulgurite is (Ca0.91Fe0.09)HPO3. Although there is a thin boundary between anthropogenic and natural fulgurites (see below), here we first report a phosphite compound which represents the reduced version of monetite (CaPO3OH), as the phosphite (HPO32−) anion bears a + 3 oxidation state for P. To the best of our knowledge, CaHPO3 is also the first fulgurite-occurring crystalline material wherein P is not a phosphate or a phosphide.

We attempted to replicate the CaHPO3 mineral formation by adding CaHPO4 to a mixture of Fe and Si (0.1:1:2 molar ratio) and heating to 1000 °C for 8 h. The phosphate was demonstrated to form reduced P-compounds upon extraction with Na4EDTA solution by NMR (nuclear magnetic resonance) spectroscopy (Fig. 2) that demonstrates the formation of P3+ and P4+ species. However, such chemistry naturally arises from the extraction of P-compounds from Fe3P (or schreibersite), and the ratio of these species (1:5:5:1 of hypophosphate: phosphate: phosphite: pyrophosphate, see ref. 39) implicates that—rather than reducing CaHPO4 to CaHPO3—reduction of CaHPO4 to Fe3P occurred instead. If we assume that the reaction velocity doubles with every 30 °C increase in temperature (corresponding to an activation energy of 20 kJ/mol), then at 1220 °C—where FeSi2 would have melted—we would expect to see the reduction of CaHPO4 to Fe3P over the course of 5 min. Given that this did not occur in the NPR fulgurite, that implies the reduction timescale was less than 5 min.

The NPR fulgurite bears a crystalline CaHPO3 material that may be the first mineral bearing a phosphite (HPO32–, P3+) group to the best of our knowledge. Notably, this fulgurite is the second to have formed from New Port Richey, FL, USA (contrasted with that reported in ref. 40). However, fulgurites can occur in material that is also influenced by either (1) artificial target material such as pavement or conductor metal, or (2) formed through downed powerlines that introduce energy to the environment slowly and over time40. Notably, the NPR fulgurite shows no indication of the former as there is no aluminum or copper in a metallic material, and the groundmass of the fulgurite is consistent with local soil compositions34. The source of the discharge to be an artificial source may be unlikely due to the incomplete reduction of phosphate to phosphide within the FeSi/FeSi2 droplets, as reduction should be complete if the system attained equilibrium. The reduction experiments described above show CaHPO4 to Fe3P would likely have occurred as temperatures exceeded 1220 °C (the melting point of FeSi2) under relatively short timescales (~5 min) for this fulgurite, presuming a reaction rate doubling every increase of temperature of 30 °C. That the reduction of CaHPO4 to Fe3P did not occur implicates a shorter (<100 s) heating timescale, consistent with a lightning strike.

A metallic silicide spherule belonging to the NPR fulgurite was studied by high-resolution multiscale nanotomography. Fig. 3 reports the color-coded 3D thickness distribution of the pore network inside the cement sample (see Supplementary Video 1) together with the thickness distribution plot of the analysis (inset in Fig. 3). The porosity also a directionality to it, which may be the same as the orientation of the flow of gas out of the fulgurite. Figure 4 shows a virtual cut of the spherule evidencing melting and recrystallization structures.

The reported presence of much larger iron silicides41,42 in geologic specimens suggests that large spherules are formed by lightning and may be the source of these silicides that have been reported in various locations. For these reasons, we argue the CaHPO3 could be a natural material and therefore represent a member of a new mineral group, one predicted to occur in the environment from biology and biochemistry15.

## Discussion

The NPR fulgurite presents an unusual mineralogical paradox: the more reduced phosphorus mineral (P1− as a metal phosphide) is in contact with a silicate glass, but the more oxidized phosphorus material—the calcium phosphite (P3+)—is in direct contact with extremely reduced iron silicides. Such a difference is especially exemplified by the fO2 mineral buffer diagram (Fig. 5) that demonstrates that the formation of CaHPO3 should not occur in contact with FeSi or FeSi2. We examine the cause of this disparity in P chemical oxidation state and the implications of its formation.

The presence of silicides within fulgurites has been reported previously29. The size of the crystals of the silicides is comparable to the Winans Lake fulgurite3, wherein crystals of silicides ~10–100 µm wide were found within metal spherules. Compositionally, these silicides also match those reported in the Houghton Lake fulgurite28, in which FeSi and FeSi2 formed from a eutectic melt within spherules. The size of the silicide-bearing spherules is much larger than those found previously25 by about a factor of 10×, which were previously the largest spherules reported in the literature from fulgurites. The size of these spherules (1 cm), however, is comparable to those reported as purported distal impact ejecta41.

The formation of the iron silicide spherules in the NPR fulgurite likely occurred during the combustion of roots coated with iron oxide plaque (see Supplementary Fig. 3 for details on the reduction of Fe2O3 and SiO2 by organics). Iron oxide plaques form around roots, and cement sand as lithotrophic bacteria oxidize Fe(II) in wetlands environments35, and such plaques are known to concentrate phosphate43, including as calcium phosphates such as brushite44. During the lightning strike, the combustion of organic root matter likely drove extreme reduction of the iron oxide plaque, melting it as FeSi2 formed (T > 1220 °C), still well below the melting point of calcium phosphate (>1600 °C45). Calcium phosphate, entrained within the iron silicide, was subjected to the extreme reduction, and locally formed the CaHPO3 material. Given that the CaHPO3 is in contact with FeSi, and FeSi is formed under somewhat more oxidizing conditions than FeSi2 (Fig. 5), perhaps the reduction of CaHPO4 to CaHPO3 was accompanied by the oxidation of FeSi2 to FeSi and SiO2. In addition, as Fe2+ is ~9% of the divalent cation of the CaHPO3 material (Ca0.91Fe0.09HPO3) iron as FeSi may have served as a reductant as well. Elemental mapping of the CaHPO3 material (Fig. 6) demonstrates that both Si and O are present within this substance. Experimentally heating CaHPO4 with Fe and Si at 1000 °C did not produce CaHPO3 but instead produced Fe3P after 8 h, suggesting that rapid heating and cooling of this material—such as what occurs in a lightning strike on the timescale of 60 seconds or less—is necessary to make CaHPO3.

The NPR fulgurite bears two distinct, reduced P phases- the CaHPO3 in the FeSi/FeSi2 spherule, and P dissolved within Fe as much smaller droplets within the silicate glass. These likely represent two separate reduction pathways for phosphorus. The P-rich iron spherules in silicatic glass (Fig. 1c, d) are smaller (15 µm) and their spherical shape and higher P concentration along the edges of the spheres46 indicate that molten iron reacted with P at high temperature (T > 1538 °C), probably as volatilized P, through phosphidation of the metal. The spherical shape (Fig. 1d) also indicates that iron melted. Metal phosphidation has been proposed to be a kinetically favored route to phosphate reduction that may be the source of metal phosphides on the Moon47. This route involves volatilization of P from phosphates to form P and P2 gases under reducing (but not extremely reducing) conditions, followed by reaction of P gas with metal to form metal phosphides, a process that occurs rapidly48.

In contrast, the CaHPO3 mineral likely formed from the reduction of calcium phosphate, possibly as either amorphous calcium phosphate, or brushite (CaHPO4·2H2O). Brushite can dehydrate to monetite (CaHPO4) with heating and does so rapidly (timescale of <1 min for similar Mg-phosphates49) at temperatures >200 °C49,50. Brushite reduction and dehydration, as:

$${{{{{{\rm{CaHPO}}}}}}}_{4}\cdot 2{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}={{{{{{\rm{CaHPO}}}}}}}_{3}+2{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}+1/2\,{{{{{{\rm{O}}}}}}}_{2}$$

was approximated using limited thermodynamic data19 but should occur under more oxidizing conditions than at the wustite-magnetite fO2 buffer. In contrast, the FeSi or FeSi2 oxidation buffer3 is well below the iron–wustite buffer, just slightly above the Si–SiO2 buffer. This suggests that the reduction of phosphate to phosphide proceeds through a phosphite intermediate, which did not complete due to the relatively short duration of heating (<1 min) during the lightning strike fusion of the soil. The CaHPO3 compound is thus out of equilibrium with its surrounding assemblage and should have been reduced to phosphide.

To this end, the reduction of phosphate proceeds via two routes (Fig. 7): a high temperature (>1700 °C) reduction wherein Fe metal forms and is rapidly phosphidized by P gas at moderately low fO2, and a lower temperature (>1220 °C) reduction wherein P as calcium phosphates is entrained within extremely reduced material and is rapidly reduced to phosphites but kinetically inhibited from forming the ultimate reduced materials, the phosphides.

Although the production of CaHPO3 from CaHPO4 in contact with FeSi likely has few direct counterparts in natural geologic samples (as silicides are rare), the identification of a likely natural, crystalline form of a P3+ material is important to understanding the role of redox in the phosphorus cycle. Hess et al.4 specifically call out the lightning reduction of phosphate to have been a widespread phenomenon on the early Earth potentially transforming thousands of kg of phosphate to phosphides and phosphites annually, which would have rivaled meteoritic sources. These materials may have been prebiotically important31,51. Pasek and Block19 argue lightning would have been a less important (~3000 kg/yr) source of reduced P on the modern Earth, but the relatively facile reduction of phosphate to phosphite (Fig. 5) indicates that phosphites may be present in other geologic or cosmic samples subjected to reduction but that do not reach very high temperatures. For instance, phosphates within pallasites or iron meteorites may equilibrate with iron metal to transform into phosphite minerals. Phosphite is also found within serpentinites20, and phosphites such as CaHPO3 (specifically the Mg/Fe2+ equivalents) may be the mineral carrier of phosphite that yield phosphite during the extraction of these rocks. The chemistry of P within reduced igneous rocks (at or below the WM buffer) may include reduced P material such as CaHPO3, but few speciation-specific studies have been done of P chemistry in these types of rocks.

Modern metabolic pathways implicate P3+ as being a non-negligible portion of the total phosphorus inventory, but the absence of solid P3+ materials as sources of this ion for life has drawn questions as to the environmental phosphite reservoir15. Thus a careful investigation of the anionic character of P within reduced material—though not necessarily as reduced as investigated here—might reveal other forms of crystalline reduced P minerals, many of which may have been important in the development of life on the Earth52, or in the supply of nutrients to modern environments18.

## Methods

### SEM

The instrument used was a Zeiss–EVO MA15 Scanning Electron Microscope coupled with an Oxford INCA250 energy-dispersive spectrometer, operating at 25 kV accelerating potential, 500 pA probe current, 2500 cps as average count rate on the whole spectrum, and a counting time of 500 s. Samples were sputter-coated with 30-nm-thick carbon film. X-ray maps were collected with acquisition times of 10 ms per pixel.

### X-ray diffraction

Studies were performed at the CRIST, Centro di Studi per la Cristallografia Strutturale, Department of Chemistry, Università di Firenze, Italy. A small fragment (size about 14 × 13 × 11 μm) was extracted from the polished section (Supplementary Fig. 2) under a reflected light microscope and mounted on a 5- μm-diameter carbon fiber, which was, in turn, attached to a glass rod. Single-crystal X-ray diffraction data were collected by means of a Bruker D8 Venture equipped with a Photon II CCD detector, using graphite-monochromatized MoKα radiation (λ = 0.71073 Å), exposure time of 30 s per frame and a detector-to-sample distance of 7 cm. Intensity data were integrated and corrected for Lorentz-polarization and absorption with the Bruker software packages. A full Ewald sphere was collected up to 2θ = 70° at room temperature.

### LA-ICPMS

Laser Ablation Inductively Coupled Plasma Mass Spectrometry was performed using a New Wave Research (esi) laser (NWR-213) attached to a Perkin Elmer Nexion 2000P Quadrupole ICPMS. Two samples of groundmass glass were analyzed by LA-ICPMS. Glass samples were chosen as fragments of the NPR fulgurite, then sequentially ground and polished to a 0.05-µm alumina grit and were mounted using putty. Four points on each glass were ablated and measured for the abundance of 29Si, 48Ti, 27Al, 23Na, 57Fe, 55Mn, 43Ca, 39K, 31P, and 11B (each chosen to minimize the effect of other isobars), which were referenced to USGS glasses BCR-2G, BHVO-2G, and BIR-1G53 mounted in epoxy to constrain major element abundances that were normalized to 100% by weight.

### Thermodynamic calculations

Data for the Gibbs free energy of various species that determine oxygen fugacity as buffers (Fe2O3, Fe3O4, FeO, Fe, FeSi, FeSi2, CaSiO3, Ca3(PO4)2, Fe3P, SiO2, C6H12O6, CO2(g), and O2(g)) were retrieved from HSC Chemistry (Outokompu Research Oy, v. 7.1, see details in refs. 54,55). Data for the CaHPO3/CaHPO4 are from Pasek and Block19 and were calculated by determining the difference in solubility between CaHPO3 and CaHPO4 (monetite), then assuming the CPs of both were similar, allowing extension of the buffer to high temperature. While such a method is likely to be less accurate at high temperature, it provides an approximation of the fO2 buffer that may be accurate enough on the logarithmic scale to make conclusions about the relative importance of various reactions.

### CaHPO4 reduction experiments

CaHPO4 (99% from Fisher Scientific, 0.031 g) was mixed with Fe and Si powder (0.1 g each) to give a 0.1:1:2 mixture of CaHPO4:Fe:Si by mole. This mixture was heated in an alumina crucible under N2 gas at 1000 °C for 8 h. The resulting powder was extracted with a solution of Na4EDTA in water (10 mL) for 1 week, and analyzed by 31P NMR (details in49) for 10,000 scans in 1H-coupled mode.

### Nanotomography

High-resolution multiscale nanotomography was carried out by means of a Bruker Skyscan 2214. The sample was fixed on the sample holder with dental wax, and the entire volume was scanned in one field of view to visualize pores and different density phases inside the sample. The Paganin Phase Retrieval algorithm was applied to the projection images to enhance the different density feature inside the sample.