An evaporite sequence from ancient brine recorded in Bennu samples

Brines (over 3.5 wt% dissolved solids) are environments in which life could have evolved or might persist in the Solar System⁶, and are targets for spacecraft exploration. Evaporation or freezing can lead to the formation of brines from which a variety of minerals (for example, carbonates, sulfates and halides) precipitate. On Earth, such mineral deposits are a major source of technologically critical elements⁷. On Mars, brine freezing points extend to approximately −20 °C, prolonging the liquid state of water⁸. Icy outer Solar System bodies contain subsurface brines, sometimes as oceans. Evidence of subsurface brines is found on Saturn’s moon Enceladus⁹ and the dwarf planet Ceres³, the largest body in the asteroid belt.

Our knowledge of brines beyond Earth is hampered by a lack of samples. Remote-sensing observations of Mars, Ceres and Enceladus limit our ability to determine precipitated phases in minor to trace abundances, unravel the age and timing of fluid evolution and precipitation and determine the compositions of the associated fluids. Evaporite phases known from meteorites are extremely limited. These include sulfates and halides in Martian nakhlites¹⁰, intrusive igneous rocks that experienced secondary alteration, and potentially indigenous halite in ordinary chondrites¹¹.

Early analyses of samples from Bennu recorded evidence of pervasive aqueous alteration—including hydrated phyllosilicate clay minerals, carbonates, magnetite and sulfides⁵—similar to samples from asteroid (162173) Ryugu returned by the Hayabusa2 mission and the most extensively aqueously altered carbonaceous chondrites (for example, Ivuna-type (CIs)). Pervasive aqueous alteration did not occur on Bennu itself, which is a second- or later-generation rubble pile formed within the past 65 million years¹², but on Bennu’s larger parent asteroid in the early Solar System¹³. Remote sensing of Bennu by the OSIRIS-REx mission showed distinct boulder populations that probably formed at different depths within the parent body^14,15 and were probably sampled by the spacecraft^5,16,17.

The Bennu samples contain an array of phases formed by precipitation from a fluid. The most abundant of these phases include sulfides, magnetite and carbonates (10, 5 and 3 vol.%, respectively, by X-ray diffraction of a bulk sample⁵), although concentrations differ between individual millimetre-sized asteroidal particles. These phases often occur in association. Sulfides, primarily pyrrhotite, occur as pseudohexagonal mineral grains tens to hundreds of microns in size, micron-size anhedral grains⁵ and vug-filling laths, the last of these being associated with magnesian phyllosilicates and apatite⁵ (Extended Data Fig. 1a). Carbonate includes grains of calcite (CaCO₃), dolomite ((Mg,Ca)CO₃) and Fe,Mn-bearing magnesite (MgCO₃). Using scanning electron microscopy (SEM), we observed subhedral to rounded calcite occurring with magnetite (Fe₃O₄), the latter as framboids (Fig. 1a), plaquettes and spherules. These phases also appear separately. Calcite occurs as ovoid shapes with aspect ratio of roughly 2:3 (Extended Data Fig. 1b), and magnetite framboids occur in veins. SEM and electron microprobe analysis (EMPA) show that the calcite has near-end member composition (4 mol.% or less MgCO₃ and FeCO₃; 0.1 mol.% or less MnCO₃; Extended Data Fig. 2).

a, Multiphase particle exhibiting successive growth of calcite (c), magnetite (mt) and zoned Mg,Na phosphate (ph), set in a matrix of phyllosilicates (multielement X-ray map; sample no. OREX-803065-0). b, HR-CL image of zoned carbonate. A core of Fe,Mn-bearing magnesite (M) appears dark (non-luminescent) and is surrounded by multiple dolomite overgrowths (D) exhibiting a range of luminescence intensity (dull to bright) around 650 nm due to Mn variations. Phase identification confirmed by SEM (sample no. OREX-800045-106). c, X-ray map of molecular carbonate acquired using scanning transmission X-ray microscopy by integration of the area under the 290.5 eV peak (yellow) overlaid on an inverted bright-field TEM image of sample no. OREX-803030-100, demonstrating that molecular carbonate is heterogeneous and ubiquitous in phyllosilicates. The spectrum of the outlined area is shown in Extended Data Fig. 3. Scale bars, 20 μm (a), 5 μm (b) and 2 μm (c).

Intergrowths of Ca,Mg,Fe,Mn-bearing carbonates, including dolomite and Fe,Mn-bearing magnesite, are observed using SEM and high-resolution cathodoluminescence (HR-CL). Composite grains exhibit mixed cores of Fe,Mn-bearing magnesite and dolomite, sometimes rimmed by subsequently precipitated dolomite (Fig. 1b). Dolomite also occurs as veins cross-cutting the phyllosilicate matrix and 200–500-µm grains as the dominant phase at millimetre scale. SEM and EMPA indicate dolomite of subequal molar Ca and Mg, with minor Fe and Mn (under 11 mol.% FeCO₃, under 8 mol.% MnCO₃; Extended Data Fig. 2). Fe,Mn-bearing magnesites are Mg dominant, with under 33 mol.% FeCO₃, under 18 mol.% MnCO₃ and under 10 mol.% CaCO₃ (Extended Data Fig. 2).

In addition, we observed a continuum of submicron amorphous to molecular carbonate occurring with phyllosilicates (Fig. 1c) using transmission electron microscopy (TEM) and scanning transmission X-ray microscopy. This carbonate exhibits the 290.5-eV peak characteristic of the carbonate functional group (Extended Data Fig. 3), but does not diffract as crystals on TEM. Similar amorphous carbonate was observed in Ryugu samples^18,19. This carbonate could represent an important reservoir of total carbonate ions in Bennu samples, given the abundance of phyllosilicates (about 80 vol.%)⁵.

Nanometre- to micron-sized Na,Ca carbonate grains were studied using TEM (Fig. 2a), including selected-area electron diffraction (SAED). The phase is sensitive to beam damage. Co-occurring phases include calcite, TiO₂, pyrrhotite, pentlandite and phyllosilicates. Na,Ca carbonate studied by SAED around 1 month following sample allocation yielded d-spacings consistent with hydrated Na,Ca carbonates gaylussite (Na₂Ca(CO₃)₂·5H₂O) or pirssonite (Na₂Ca(CO₃)₂·2H₂O) (Extended Data Table 1). The 2.84-Å d-spacing is inconsistent with anhydrous Na,Ca carbonate, nyerereite (Na₂Ca(CO₃)₂), suggesting that the Bennu Na,Ca carbonate is hydrated. The same grain examined roughly 5 months later following continuous storage in an air-filled desiccator showed extensive alteration, including NaCl growth on the surface (Fig. 2b), suggesting that the phase is highly reactive in air.

a, Na,Ca carbonate grain, studied about 1 month following removal from nitrogen storage in the curatorial facility, exhibits SAED d-spacing consistent with the hydrated phases gaylussite or pirssonite. The mineral is sensitive to beam damage (dark spot within grain), consistent with a hydrated phase (multielement Kα X-ray map; sample no. OREX-800045-102). b, NaCl particles grown on the same grain after 5 months (multielement Kα X-ray map). Scale bars, 200 nm.

Magnesium sodium phosphate forms bright coatings on some Bennu particles⁵ (Extended Data Fig. 4a). It occurs as blocky grains with textures indicative of desiccation. Using X-ray diffraction (XRD), we show that Mg,Na phosphate in Bennu samples is amorphous (Extended Data Fig. 5), probably the result of desiccation⁵. Mg,Na phosphate occurs as a late-stage phase in calcite–magnetite-rich vugs (Fig. 1a); on irregular broken surfaces, suggesting complex three-dimensional vein shapes (Extended Data Fig. 4b); and associated with veins of Mg phyllosilicates lacking micrometre-sized sulfide inclusions that rim large dolomites (Fig. 3). Using SEM energy-dispersive X-ray spectroscopy (EDS), we observed up to 10 wt% Na, although typically 0–5 wt%. Some grains exhibit approximate Na zonation of 2 wt% (Fig. 1a). Minor concentrations of K (0.06 wt% or less), Cl (0.13 wt% or less) and F (0.60 wt% or less) are detected by EMPA.

Mg,Na phosphate occurs with Mg phyllosilicates largely free of micron-sized sulfide inclusions. The Mg,Na phosphate–Mg phyllosilicate veins rim large dolomite (Dol) grains that include magnetite (multielement Kα X-ray map; sample no. OREX-803085-100). Scale bar, 200 μm.

Late-stage, sodium-rich phases typically occur as isolated veins or individual grains, rather than being associated with other salts. Veins of Na carbonate up to 20 µm in width and 200 µm in length form segments, possibly exposures of a continuous sheet, bounding a clast (Fig. 4a and Extended Data Fig. 4). Within these veins, Na carbonate occurs as randomly oriented needles of width below 1 µm and length 5–10 µm (Fig. 4b). Both SEM and TEM show that this phase is beam sensitive, probably reflecting the hydrated nature of the phase, and is composed of Na, C and O (Extended Data Fig. 6a). In TEM bright-field imaging, Na carbonate has a highly porous and lacy texture (Extended Data Fig. 6d), possibly reflecting, in part, damage due to focused ion beam (FIB) milling. Scanning TEM (STEM)-based electron energy-loss spectroscopy (EELS) mapping shows a sharp peak at around 290 eV in the carbon K edge (Extended Data Fig. 6b), consistent with carbonate groups. SAED (Extended Data Fig. 6c and Extended Data Table 2) from several locations shows that the Na carbonate is poorly crystalline. Matches for carbonates with allowable Na, C, O and H and measured d-spacings include the hydrous phase trona (Na₃H(CO₃)2·2H₂O) and the H-bearing phases wegscheiderite (Na₅H₃(CO₃)₄) and nahcolite (NaHCO₃). Wegschiederite is the poorest fit to d-spacings and trona the best, although the nanocrystalline nature increases uncertainty. We favour identification as trona but cannot exclude nahcolite or a mixture of the two.

a, Na carbonates (orange) underlie a flat surface, which is itself overlain by dolomite-bearing (teal), phyllosilicate-rich (purple) host rock containing magnetite and sulfide (red) (multielement Kα X-ray map; sample no. OREX-803088-0). b, Na carbonate, occuring as needles of less than 1-µm wide and 5–10-µm long, in a vein or sheet that underlies a clast (backscattered electron (BSE) image; sample no. OREX-803088-0). Scale bars, 500 μm (a) and 20 μm (b).

Sodium sulfate occurs as distinct grains of roughly 5 µm within a single particle dominated by phyllosilicates (Fig. 5a). The composition of sodium sulfate by EDS is 32.8 wt% Na, 19.2 wt% S and 48 wt% O, yielding a formula of Na_1.9S_0.8O₄, a reasonable approximation for thénardite (Na₂SO₄).

a, Na sulfate, probably thénardite, occurs as multiple grains (roughly 5 µm) in a phyllosilicate-dominated clast (BSE image; sample no. OREX-803172-0). b, Halite and sylvite occur in a particle with calcic pyroxene (Cpx) and plagiolcase (Pl) (ToF–SIMS ion map; sample no. OREX-501070-0). c, Sodium fluoride grains of approximately 1 µm, possibly villiaumite, occur with phyllosilicates and Mg,Na phosphate (secondary electron image; sample no. OREX-803174-0). Scale bars, 10 μm (a), 5 μm (b) and 1 μm (c).

Halite (NaCl) and sylvite (KCl) were observed as angular to rounded grains smaller than 5 μm (mostly under 1 μm) in size, co-existing with calcic pyroxene and plagioclase in a Na-rich phyllosilicate matrix (Fig. 5b and Extended Data Fig. 7). Time-of-flight–secondary ion mass spectrometry (ToF–SIMS) indicates variable Na and K between grains. Both phases contain detectable, although variable, F. Reimaging of the particle after 8 months in an air-filled autodesiccator showed that halite-sylvite grains had disappeared, leaving voids (Extended Data Fig. 7e).

Sodium fluoride occurs as rare grains roughly 1 µm in size, co-existing with phyllosilicate matrix and Mg,Na phosphate (Fig. 5c). The composition of sodium fluoride by EDS is 40.7 wt% F and 59.3 wt% Na, corresponding to a formula of Na_1.22F. No Cl was detected. Given the uncertainties in analysis of small, irregular grains, this is a reasonable approximation for villiaumite (NaF).

The occurrence of diverse salts, particularly Na-rich salts, distinguishes Bennu samples from meteorites, in which evaporite minerals are rare. Extraterrestrial halite and sylvite have been reported in both ordinary chondrites¹¹ and aqueously altered Mighei-type chondrites²⁰. Hydrated Na sulfate (for example, bloedite, MgSO₄·Na₂SO₄·4H₂O (ref. ²¹)) in the CI chondrite Ivuna was attributed to terrestrial alteration. Surficial sulfates and halite formed in the Mighei-type chondrite Winchcombe days to months after its fall²². These observations support the idea that extensive sulfate veining in CI chondrites, once attributed to asteroidal aqueous alteration, formed by terrestrial alteration²³—as does the alteration of Na,Ca carbonate (Fig. 2b) and complete loss of NaCl and KCl (Extended Data Fig. 7e) in Bennu samples stored in air-filled desiccators over 5–8 months at relative humidity of around 10–40%.

Volatile element depletions observed in chondrites²⁴ that have had residence times on Earth from years to centuries could be explained in part by loss of evaporites rich in Na, Cl, F and S due to prolonged contact with air and attendant water vapour. By contrast, Bennu samples stored in nitrogen (much less than 1% relative humidity) retained their Na-rich evaporite phases, including Na sulfate, carbonate and fluoride. In the absence of sample return, curation and storage during analysis under carefully controlled conditions (in nitrogen), this full array of salts would not have been discovered.

The array of salts and associated phases in Bennu samples suggests that a protracted fluid evolution occurred on Bennu’s parent asteroid, notably forming veins that cross-cut the host rock. On Earth, these salts form primarily by precipitation from fluids. Closed basins serve as the optimal terrestrial analogue for Bennu’s parent body, which would have undergone limited or no input of fluid during evaporation or freezing. In terrestrial closed basins, dilute solutions can produce brines through evaporation or freezing¹, and the nature of those brines is controlled largely by mineral precipitation events, termed chemical divides. Carbonate precipitation has been suggested to produce fluids rich in Na, SO₄ and Cl, and poor in Mg and Ca^25,26.

Among the best-studied terrestrial basins with early Ca,Mg carbonate precipitation is Searles Lake in California, USA^4,27. Calcite and dolomite formed there by precipitation primarily because of initial saturation of the fluid, rather than evaporation. Na,Ca carbonate (gaylussite/pirssonite) at Searles Lake, formed during syndepositional back-reaction between previously formed carbonate and Na-rich brine, was the last phase to precipitate before evaporation dominated. Later phases in the stratigraphic sequence at Searles Lake, including Na sulfate, Na carbonate and Na chloride, form an evaporite sequence. Similar studies have been conducted for freezing of terrestrial lakes and proposed brines on outer planet moons²⁸.

Bennu samples contain six of the minerals observed at Searles Lake: calcite, dolomite, gaylussite/pirssonite, thénardite, trona and halite. In the absence of petrographic constraints on the timing of the latter four phases, we adopt the Searles Lake order of formation (Fig. 6). Four salt or associated phases in Bennu samples are not observed at Searles Lake. Abundant magnetite was an early-formed precipitate, as evidenced by magnetite included in dolomite (Fig. 3). Magnetite plaquette and framboid shapes may form by interaction of iron sulfides with an alkaline fluid²⁹, evidenced by surficial Na coating magnetite framboids in the Tagish Lake ungrouped carbonaceous chondrite³⁰. Coprecipitation of sulfide and apatite, consistent with their co-occurrence in vugs, may have occurred in neutral fluids before calcite precipitation. Together, calcite, dolomite, magnesite, magnetite, apatite and gaylussite/pirssonite were formed during the early stages of water–rock interactions on the parent body.

Black bars indicate the appearance of phases formed by precipitation and evaporation as a function of approximate percentage of brine evaporation, determined by the salinity at which the phase would first appear (Methods). The first and final appearances of a variety of elements, along with evolution from neutral to alkaline, Na-rich fluids, are shown at the top.

By contrast, Na-bearing phosphate, Na carbonate, sulfate, chloride, fluoride and K chloride form a late-stage evaporite sequence. Magnesium sodium phosphates probably precipitated in the middle of the mineral sequence, based on relevant terrestrial analogues and their spatial relationships (Figs. 1a and 6). The low-temperature secondary mineral apexite (NaMg(PO₄)·9H₂O (ref. ³¹)), from Nevada, USA, occurs with calcite and gaylussite, suggesting a relationship. In Bennu’s parent body, phosphate abundance was probably controlled by the apatite-carbonate buffer³², and evaporation increased Na concentrations to the point at which Mg,Na phosphate precipitated. Following precipitation of Na sulfate and Na carbonate, sylvite coprecipitated with halite. Na fluoride probably precipitated after halite when Cl in the fluid was controlled by the solubility of halite. Bennu halides contain tenths of a weight percentage of F, consistent with high F concentration in the fluid. Other classes of minerals observed in closed basins on Earth, including borates and lithium-rich fluorides and brines, are not observed in Bennu samples, consistent with the parts-per-million concentrations of B and Li in their bulk composition⁵.

The salt sequence recorded in the Bennu samples (Fig. 6) may partially constrain the temperature within the parent body, and the relative roles of evaporation and freezing, which could occur in a liquid brine as low as the H₂O-NH₄ eutectic (176 K). Modelling of the Searles Lake evaporites constrains temperatures to 20–29 °C (ref. ⁴). Although we have far fewer constraints for Bennu’s parent body (for example, pCO₂, co-occurrence of minerals, initial fluid composition), this temperature range is consistent with maximum formation temperatures for gaylussite (below 40 °C (ref. ³³)) and pirssonite (below 55 °C (ref. ³⁴)) and minimum equilibrium formation temperature for trona (above 20 °C (ref. ³⁵)); it is also compatible with temperatures inferred from co-existing sulfide compositions in Bennu samples. We recognize that trona can form by flash-freezing of brines occuring with NaCl³⁶, an association not observed in Bennu samples. Together, these geothermometer temperatures favour evaporation, rather than freezing, as the dominant control on fluid composition, at least through the formation of Na carbonate. Fluid freezing following Na carbonate formation is possible.

The setting and relative timing of salt formation can be constrained by the evaporation required. Single-phase evaporite veins suggest precipitation in cracks within Bennu’s parent body. Whether these are small-scale offshoots of a larger brine body is unclear. Preliminary calculations (Methods) suggest formation of Na carbonate and Na chloride at a minimum of 60–70% evaporation. The physical plausibility of extensive evaporation of a fluid moving solely through cracks is unconstrained. By contrast, larger volumes of water, from which evaporating brine could be periodically injected into small fractures, would be consistent with findings from spacecraft observations of Bennu. Sedimentary deposition of large clasts within layered boulders on Bennu would require water several metres deep inside the parent body³⁷. Metre-scale bodies of water on Bennu’s parent body are evidenced by veins within boulders on Bennu³⁸; spectral features suggest that these veins are primarily carbonates (calcite, dolomite, magnesite) and associated organics. Smaller-scale carbonate veins have been observed in a Bennu sample (Fig. 4 and Extended Data Fig. 4), albeit as Na carbonate, which was not detected on Bennu during proximity operations. This is, perhaps, not surprising given the dominance of Ca,Mg carbonates in Bennu samples. We suggest that carbonate veins on Bennu are metre-scale brine pockets that produced the salts observed and were trapped following global alteration.

Brines are a subject of intense interest to the planetary science community³⁹ in regard to understanding the internal structure of ice-rich worlds, and to the formation and preservation of prebiotic organics and their astrobiological potential. Samples returned from the carbonaceous asteroid Ryugu contain Ca,Mg carbonates, Mg,Na phosphate⁴⁰ and Na carbonate⁴¹. Extracts produced with hot water, termed salts, contain (Na + K/Na + K + Mg + Ca) of about 0.9 and, for formic acid extracts, termed carbonates, this value is roughly 0.10–0.15 (ref. ⁴²), suggesting that additional Na-rich evaporite phases may be present in Ryugu samples.

Sodium carbonate (possibly hydrated) and hydrated sodium chloride were spectrally identified in high-albedo areas on Ceres^3,43. These deposits probably result from recent emplacement of deep-seated brines, perhaps by impact processes^44,45. Hydrated Na sulfates and carbonates exposed on the surface of an airless body dehydrate in 10³–10⁸ years⁴⁶, dependent in part on temperature. The occurrence of Na carbonate on Bennu’s parent body suggests that its ancient brine, at about 55% evaporation, produced evaporites similar to those emplaced recently on Ceres. Furthermore, earlier stages of water–rock interaction and evaporation may have produced abundant Ca,Mg carbonates, magnetite and phosphates on Ceres by analogy with Bennu. Ammonium chloride or ammonium carbonates³ observed in bright deposits on Ceres suggest brine temperatures much lower than those for H₂O/NaCl mixtures. These phases have not yet been observed in Bennu material, but comparatively high concentrations of ammonium (13.61 ± 0.36 µmol g⁻¹) occur in a hot-water extract of bulk sample⁴⁷.

Active plumes have been observed on Enceladus, with ejected material forming the faint E ring around Saturn⁴⁸. Salt-rich ice grains from these plumes, analysed by Cassini’s Cosmic Dust Analyzer, showed ions of Na, K, Cl, CO₃ and PO₄, including Na carbonate² and Na phosphate³². On Enceladus, active subsurface brine oceans cryovolcanically erupt materials comparable to those of the ancient evaporites from Bennu’s parent body. Strong scientific justification exists for future missions to Ceres and Enceladus.

The evolving brine and phyllosilicate-rich host rock on Bennu’s parent body pose an intriguing, but untested, environment for prebiotic organic syntheses in a single location. Among the most interesting challenges is the origin of prebiotic organic molecules, particularly nucleotides (nucleobase + sugar + phosphate). Bennu samples contain nucleobases⁴⁷, including the five canonical nucleobases that comprise nucleic acids in terrestrial biology. Minerals that catalyse the sugar ribose (C₅H₁₀O₅) include hydroxyapatite⁴⁹, and salts can catalyse the formation of nucleosides (nucleobase + sugar)⁵⁰. Although both phosphate and salts are observed in Bennu samples, experimental studies are needed to determine whether nucleotides could have formed in the chemical conditions on Bennu’s parent body. If nucleotides form, polymers of 30–60 monomers can be catalysed by montmorillonite⁵¹, although such a process has not been demonstrated for the phyllosilicate phases in Bennu samples. Finally, as polymers increase in length, adsorption strength to the mineral surface increases⁵². These polymers could be released by concentrated salt solutions. Such processes have yet to be studied or their products observed in Bennu samples; nevertheless, the presence of clays, phosphates and evaporites indicative of alkaline fluids makes Bennu a promising target for investigation of potential prebiotic organic synthesis.