Discovery of primitive CO₂-bearing fluid in an aqueously altered carbonaceous chondrite

Discovery of CO₂-bearing aqueous fluid in primitive meteorite gives the direct evidence of dynamic evolution of the solar system.

Abstract

Water is abundant as solid ice in the solar system and plays important roles in its evolution. Water is preserved in carbonaceous chondrites as hydroxyl and/or H₂O molecules in hydrous minerals, but has not been found as liquid. To uncover such liquid, we performed synchrotron-based x-ray computed nanotomography and transmission electron microscopy with a cryo-stage of the aqueously altered carbonaceous chondrite Sutter’s Mill. We discovered CO₂-bearing fluid (CO₂/H₂O > ~0.15) in a nanosized inclusion incorporated into a calcite crystal, appearing as CO₂ ice and/or CO₂ hydrate at 173 K. This is direct evidence of dynamic evolution of the solar system, requiring the Sutter’s Mill’s parent body to have formed outside the CO₂ snow line and later transportation to the inner solar system because of Jupiter’s orbital instability.

INTRODUCTION

Water is abundant as solid ice in the solar system, providing evidence that water exists in the solar system, consistent with observed aqueous alteration in primitive meteorites and formation of the Earth’s ocean, and liquid water closely relates to the origin of life. Not only H₂O ice but also CO₂ and CO ice define the particular distances from the Sun in the early solar system called snow lines (SLs) beyond which it was cold enough for these volatile components to condense into ice grains. If we know the composition of such ice (e.g., CO₂/H₂O ratio) recorded in primitive meteorites, we can distinguish between formation inside or outside of the SL (1).

Most meteorites originate from asteroids that formed as planetesimals in the early solar system (2). Minerals and ice were accreted together outside the SL to form planetesimals. Aqueous alteration by reaction of melted ice with anhydrous minerals occurred and was recorded in some primitive meteorites called carbonaceous chondrites (CCs) [e.g., (3)]. Water is mostly present as hydroxyl and/or H₂O molecules of hydrous minerals. It is expected that liquid water should remain as fluid inclusions in minerals (e.g., calcite: CaCO₃) that precipitated in aqueous fluid. However, completely verified fluid inclusions in CCs have not been recognized so far (4). The only fluid inclusions from the early solar system were found in halite (NaCl) associated with CC clasts in less primitive ordinary chondrites (5, 6). Here, we report the presence of CO₂-bearing fluid inclusion in a calcite grain of an aqueously altered CC [the Sutter’s Mill meteorite; Mighei-type (CM) chondrite] and use the results to infer the origin of the meteorite parent body.

RESULTS

Micrometer- and nanosized inclusions in calcite grains

We conducted a specific analysis protocol (fig. S1) to search for fluid inclusions, systematically combining scanning electron microscopy (SEM), focused-ion beam (FIB) microsampling, synchrotron radiation–based x-ray computed nanotomography (SR-XCT), scanning transmission electron microscopy with energy-dispersive spectroscopy (STEM/EDS), and TEM with cryo-stage (cryo-TEM).

We examined polished sections of Sutter’s Mill and searched for calcite (Fig. 1A and fig. S1) appropriate for the present analysis. Seven grains ~30 to 50 μm in size were selected and extracted using FIB to make equant samples with side length of ~30 μm (Fig. 1B and fig. S1). The samples were imaged using SR-XCT to obtain their three-dimensional (3D) images with absorption and phase-shift contrasts using two different methods: dual-energy tomography (DET) (7) and scanning-imaging x-ray microtomography (SIXM) (8) with ~70 and 100 nm in voxel size, respectively (3). A large number of inclusions in the interior of calcite grains were recognized in 3D (Fig. 1, B and C). They are divided into two types: micrometer- and nanosized inclusions typically about or more than a few micrometers and about or less than a few hundred nanometers in size, respectively. Micrometer-sized inclusions are usually irregular (Fig. 1E) or subspherical (Fig. 1F) in shape and more or less have facets. Some are hexagonal platelets (negative crystals) with basal faces of probably {001} (Fig. 1G). The values of refractive index decrement (RID) in phase-shift images are closely proportional to material density (RID = 8.88 × 10⁻⁶, 3.62 × 10⁻⁶, 2.51 × 10⁻⁶, and 4.2 × 10⁻⁹ at 8 keV for calcite, water, liquid CO₂, and air, respectively). The RID values of micrometer-sized inclusions are less than that of water and liquid CO₂, indicating no trapped fluid inside (m-inc in A-A′: Fig. 1D).

Fig. 1. SEM and XCT slice images of calcite grains, line profiles of RID values, and bird’s-eye views of micrometer-sized inclusions.

(A) A backscattered electron SEM image shows a typical calcite grain (CC-7; see Materials and Methods for the sample number) in matrix. Micrometer-sized inclusions and a large number of nanosized inclusions are observed in an absorption XCT image at 7 keV of CC-7 (B) and a phase-shift image of CC-25 (C). (D) Line profiles of RID values along A-A’ and B-B′ in (C). The RID values larger than zero are due to size effects (see Materials and Methods). Bird’s-eye views of micrometer-sized inclusions obtained from XCT images have a variety of morphologies, from irregular [(E): CC-7], subspherical with facets [(F): CC-27], and tabular hexagonal (negative crystal) [(G): m-inc in (C)]. cc, calcite; m-inc, micrometer-sized inclusion; n-inc, nanosized inclusions; gl, slide glass.

Nanosized inclusions are located in specific regions (Fig. 1, B and C), and they are too small to confirm the presence of fluid by SR-XCT. We extracted ultrathin sections of nanosized inclusion-rich regions of two XCT samples using FIB and observed them by TEM. We also extracted TEM sections from five calcite grains without performing SR-XCT (fig. S1). The typical inclusion sizes range from <10 to a few hundred nanometers (Fig. 2). The thickness of the section is ~300 nm to prevent the inclusions from decrepitating (bursting). Some inclusions are distributed as band-like growth zoning or healed fractures (Fig. 2A), and some are located along twin boundaries (Fig. 2B). Most of them have ovoidal shapes with facets (Fig. 2D) probably of calcite {001} plane (fig. S2). Some inclusions contain trapped or daughter crystals (Fig. 2D), which seem to be Na-Mn and Fe sulfates, FeOOH (or Fe₂O₃) (fig. S3), and apatite (fig. S4) based on their EDS spectra.

Fig. 2. TEM images of nanosized inclusions in calcite.

Some nanosized inclusions are distributed as bands [(A): bright-field (BF) image of CC-15] and along twin boundaries [(B): arrows, BF image of calcite CC-6]. One inclusion [(C): arrow, STEM image of C7K] has extra diffraction spots in cryo-TEM observation (Fig. 3). Nanosized inclusions have ovoidal shapes with facets [(D): STEM image of CC-7]. Tiny faceted objects as trapped or daughter crystals are observed [(C) and arrows in (D); see also fig. S3].

Discovery of CO₂-bearing fluid inclusion

Then, we made TEM observations using a cryo-TEM on relatively large inclusions that may preserve fluid, <100 nm in size, apart from twin boundaries, and having relatively dark contrast in TEM bright-field images. In one inclusion (arrow in Fig. 2C), we found three selected-area electron diffraction (SAED) spots at 173 K that were not observed at 293 K (yellow and red arrows in Fig. 3A). When heated to 293 K, the three spots disappeared (Fig. 3B). After keeping the sample at room temperature for 33 days, we found two new spots (red arrows in Fig. 3C). After cooling to 173 K again, the two spots disappeared and three new spots appeared (dashed red and yellow arrows, respectively, in Fig. 3D). When heated to 293 K again, the three spots disappeared (Fig. 3E). The extra spots appearing at 173 K cannot be indexed by H₂O ice (cubic and hexagonal) or CaCO₃ polymorphs (calcite, aragonite, etc.) but are successfully indexed as CO₂ [0$\overline{1}$1] ice or CO₂ hydrate (CO₂·5.75H₂O) indicating freezing of H₂O-CO₂ fluid. This is consistent with the CO₂ melting point (~218 K) or the CO₂ ice–CO₂ hydrate eutectic point (~216 K) (9). The possibility of CH₄ and CO ice can be excluded because of their low melting points, 90.5 K (10) and 70 K (11), respectively. The peaks appearing at 293 K can be explained by a metastable CaCO₃ phase (aragonite [$\overline{4}\overline{2}$3] or vaterite) or monohydrocalcite (CaCO₃·H₂O). This phenomenon can be explained by dissolution and reprecipitation of a CaCO₃ phase with nucleation delay during cooling and heating, respectively, and the negative temperature dependence of CaCO₃ phase solubilities (12) (see “Process occurring in the cryo-TEM observation” in Supplementary Text, fig. S5, and table S1). No elements other than those detected in the host calcite were detected in this inclusion by STEM/EDS except for a very small amount of P. We observed a tiny grain, probably of trapped apatite, in this P-bearing region (Fig. 2C). However, no diffraction spots corresponding to this grain were recognized because we did not observe any steady spots at 297 and 173 K.

Fig. 3. Change in SAED patterns of a nanosized inclusion in cryo-TEM.

SAED patterns of an area including a nanosized inclusion in Fig. 2C (arrow) cooled at 173 K (A), raised to 293 K (B), kept for 33 days at 293 K (C), cooled again at 173 K (D), and raised to 293 K (E) are shown. Yellow and red arrows show that spots appeared only at 173 K and after keeping the sample at 293 K, respectively.

DISCUSSION

Origin of CO₂-bearing fluid

Micrometer- and nanoinclusions with facets (Figs. 1, F and G, and 2) strongly suggest the presence of original fluid; facets formed from irregular interface (Fig. 1E) by dissolution and precipitation of calcite in fluid make the interfacial energy smaller, and finally, negative crystals (Fig. 1G) are formed. The calcite-water interface with the minimum energy is {104} based on atomistic simulations (13) and is different from {001} observed in the present inclusions (fig. S2). This could be due to the CO₂-bearing fluid, which may change the interfacial energies. Sulfate grains (fig. S3) might precipitate in such a fluid. Fluid easily escaped from inclusions along cleavages and/or twin boundaries by mechanical disturbances during the long history of the meteorite (~4.5 billion years). Large faceted pores along calcite-matrix boundaries (P in B-B′; Fig. 1D) might also have originally been fluid-filled. Nanosized inclusions have a good chance to preserve fluids because of their small size and large numbers, although the probability is not large because we found fluid only in one inclusion among the seven examined samples.

The origin of CO₂ in the CO₂-bearing fluid is probably (1) oxidation of organics and (2) CO₂ ice (1). The former contribution should be very small because strong oxidants are required and the maximum amount of CO₂ is expected to be only ~0.1 weight % (wt %) in CCs (1). In addition, the degree of thermal metamorphism of the present sample is so low that the formation of a sufficient amount of CO₂ is not expected from organics. In contrast, CO₂ ice was abundant in the early solar system and is one of the major C-bearing molecules in comets (CO₂/H₂O ~0.03 to 0.3) (14).

The observed CO₂ ice and/or CO₂ hydrate assemblage is reasonable at temperatures <~218 K with the molecular CO₂/H₂O ratio >~0.15 based on the phase diagram of CO₂-H₂O (9), indicating that the CO₂/H₂O ratio of the fluid is ~0.15 or larger. Selective escape of H₂O and enrichment of CO₂ in fluid inclusions in terrestrial quartz have been reported (15). However, this possibility can be eliminated because we did not observe any void in the inclusion (Fig. 2C). The mean CO₂/H₂O ratio of ice originally present in CCs was estimated to be 0.012 for 45 CM chondrites (Sutter’s Mill not included) from the ¹³C/¹²C ratio of carbonates (calcite and dolomite) (1). However, this ratio is dependent on the assumption of the isotope ratios of CO₂ ice and an unknown C-source. The large CO₂/H₂O ratio (>~0.15) is appropriate at least for Sutter’s Mill ice because the ratio is based on the phase diagram (9).

CO₂-bearing fluid was plausibly present during aqueous alteration in the interior of an asteroid. According to the phase diagram (9), at aqueous alteration temperatures of CMs [~270 to 300 K (16)], CO₂-bearing water (W: ~5 mol % CO₂) coexists with nearly pure CO₂ liquid (L) or fluid (F: as a liquid-gas mixture) at pressures >~50 bars or >~200 bars, which correspond to asteroids >~100 km and >~200 km in diameter, respectively (see “Conditions for aqueous alteration and the characteristics of “fluid” in the nano-inclusion” in Supplementary Text and table S2). Calcite could precipitate from W or F, but the incorporation of W or F alone cannot explain the CO₂/H₂O ratio of the nanoinclusion or the carbonate appearance by heating (Fig. 3), respectively. Accordingly, a W + L or W + F mixture should be incorporated during calcite precipitation. Because F is very mobile, W + F mixture is favorable, suggesting a parent body >~200 km in diameter.

Formation of SMPB beyond the CO₂ SL and its astronomical significance

CO₂-rich ice in Sutter’s Mill indicates that its parent body formed outside the CO₂ SL in the early solar system. We may exclude the possibility of direct origin from a comet (or mixture of comet and silicate body) because we do not have any evidence of CO, which is another major C-bearing molecule in comets (14). The H₂O, CO₂, and CO SLs are roughly determined by the condensation temperatures (~160, ~75, and ~22 K, respectively) (17). Assuming that the temperature of the solar nebula was determined by internal viscous heating in addition to solar irradiation (see “Calculation of H₂O, CO₂ and CO snow lines” in Supplementary Text), their locations moved toward the Sun as the nebula accretion rate Ṁ decreased with time (Fig. 4). H₂O (and CO₂) ice should be absent for Ṁ < ~5 × 10⁻⁹ mass of the Sun (M_☉)/year because Earth does not have abundant H₂O (and CO₂). Thus, Sutter’s Mill’s parent body (SMPB) should have formed at >3 to 4 astronomical units (au) from the Sun but at <~50 au as CO is not in the fluid (Fig. 4).

Fig. 4. H₂O, CO₂, and CO SLs and SMPB formation.

During the evolution of the early solar system, nebular accretion rate, Ṁ, decreased with time and the distances of the SLs from the Sun decreased. A possible region of SMPB formation is bounded by CO₂ and CO SLs and the ice-depleted region.

Recent dynamic models of solar system evolution have proposed that volatile-rich asteroids formed in the outer solar system beyond Jupiter’s orbit and were subsequently transported into the inner region, where the asteroids are currently located in the main asteroid belt, because of an orbital instability of the giant planets (18, 19). A dichotomy in isotopic compositions of rock-forming elements between CCs and non-CCs also suggested formation regions outside and inside of Jupiter, respectively (20, 21). The present result gives direct evidence of such a dynamic model. However, this does not necessarily imply that SMPB formed at >~10 au like a trans-Neptunian object (21). If SMPB formed beyond Jupiter’s core that formed earlier, SMPB could collect CO₂ ice. It should be noted that CO₂ (and H₂O) ice could not migrate inward across Jupiter’s core. Assumption of Jupiter’s core formation in the early stage at ~3 au is based on the distribution of refractory elements and inclusions in meteorites, which requires that CM parent bodies formed at 3.76 au within CO₂ SL (22).

Carbonates are commonly present in some CCs—CM, Ivuna-type (CI), and the Tagish Lake meteorite (C ungrouped)—but are rare in Renazzo-type (CR) and Vigarano-type (CV) or absent in Ornans-type (CO) and the Acfer 094 meteorite (C ungrouped) (16). If carbonates precipitated from CO₂-bearing fluid originating from CO₂-bearing ice, CI and Tagish Lake might have also formed outside the CO₂ SL. The CO₂/H₂O ratio of Tagish Lake ice estimated from the carbon isotopic ratio is large (~0.24) (1).

MATERIALS AND METHODS

Samples

Two polished thin sections of a CM2 clast of Sutter’s Mill (CM2) were used in this study. One is small (~1 mm by ~1 mm) and the other [SM51 in (23)] is large (~18 mm by ~22 mm). This particular sample was collected immediately after the fall, and terrestrial formation of carbonates is precluded.

Experimental design

To search for inclusions in calcite grains of the Sutter’s Mill meteorite and obtain their mineralogical characteristics, we performed systematic analyses combining field-emission SEM (FE-SEM), FIB microsampling, SR-XCT, and field-emission (scanning) TEM (FE-S/TEM) using a stage at room temperature (S/TEM) and a cryo-stage (cryo-S/TEM) (fig. S1). This method was originally developed by Miyake et al. (24). We adopted two types of analytical flow: one is (1) FE-SEM, (2) FIB microsampling, (3) SR-XCT, (4) FIB thinning, and (5) S/TEM and cryo-S/TEM and the other is one that skips processes (2) and (3). The former is essentially similar as described in (3). The details for each analysis are as follows.

FE-SEM observation

We observed the polished sections using an FE-SEM (JEOL, JSM-7001F, at Kyoto University) equipped with energy-dispersive x-ray spectroscopy (EDX) (Oxford Instruments, X-Max^N 150 mm²) operated at 10 kV and selected 12 calcite grains belonging to type 1 (25) ~30 to 50 μm in size (CC-3, CC-6, CC-7, CC-10, CC-25, CC-27, CC-36, C5K, C7K, C10K, C15K, and C16K), which are less cracked and seem to have inclusions [process (1) in fig. S1].

FIB microsampling

Seven calcite grains (CC-3, CC-6, CC-7, CC-10, CC-25, CC-27, and CC-36) were extracted from the polished sections together with surrounding matrix for SR-XCT analyses using FIBs (Thermo Fisher Scientific, Helios NanoLab 3G CX, and Quanta 200 3DS, at Kyoto University) [process (2) in fig. S1]. Specific areas (~30 μm by 30 μm) were cut out to a depth of ~30 μm using a Ga⁺ ion beam at 30 kV [e.g., an area bounded by a red box in fig. S1 (1)]. Subsequently, the blocks were lifted from the polished section and mounted on W-needles. The mounted samples were observed by SR-XCT in 3D [process (3) in fig. S1].

After the SR-XCT observation, two samples (CC-6 and CC-7) were sliced into some pieces of regions of interest (ROIs) that have many nanosized inclusions for FE-TEM using the FIB [process (4) in fig. S1]. Slices for FE-S/TEM were also extracted from the five different calcite grains (CK5, CK7, CK10, CK15, and CK16) without performing SR-XCT. The sliced pieces were mounted on TEM grids and thinned to a thickness of ~300 nm using a Ga⁺ ion beam at 30 kV. The thickness is thicker than conventional TEM sections (~100 nm or thinner) to prevent inclusion decrepitation (keep fluid in inclusions as far as possible, if present). The damaged layers formed on the ultrathin sections were removed using a Ga⁺ ion beam at 2 kV.

We assessed the effect of heat damage by the FIB sample preparation. The temperature rise was calculated by using (26) with the average thermal conductivity of calcite (~4 W m⁻¹ K⁻¹). We obtained ~<90- and ~27-K rise at the sample surface during sample thinning at <0.77 nA and final thinning at 0.083 nA, respectively. As the temperature rise in the central portion of the TEM sample is much smaller than that at the surface, we can rule out the possibility of the heat damage to the fluid.

SR-XCT experiment

The SR-XCT imaging experiment was performed for the seven samples at BL47XU of SPring-8, a synchrotron facility in Hyogo, Japan. SR-XCT enables us to obtain nondestructive 3D structure with high signal to noise ratios and high spatial resolution. We used two different methods: DET and SIXM. Details are described in (3).

Dual-energy tomography

The DET method is a tomography using x-ray absorption contrast for imaging (7), which gives us a spatial distribution of x-ray linear attenuation coefficients (LACs) of materials as a digital image. The LAC value, which is a function of the density, the elemental compositions of an object, and x-ray energy enable us to identify some mineral phases and their rough chemical compositions. The DET method was originally developed for iron-bearing samples to discriminate mineral phases with two different energies below and above the K-adsorption edge energy of iron (7.11 keV). Here, we used 7 and 8 keV. An imaging optical system with a Fresnel zone plate (FZP) was used to accomplish the submicrometer spatial resolution with the voxel size of 60 to 70 nm, which gives the effective resolution of ~200 nm. In the experiment, multiple projection images (1800 projections) were taken by rotating the sample in small intervals through a total of 180° (0.1° per projection). 3D CT images were reconstructed from the projection images using a convolution back-projection algorithm.

We easily recognized calcite and Fe-bearing minerals (serpentine, tochilinite, and cronstedtite) in the matrix. Inclusions in calcite and pores in the matrix were also easily recognized by their LAC values of almost zero at both 7 and 8 keV. Even if the inclusions are small, we can recognize these by edge-enhanced contrast, which is seen as a pair of bright and dark lines along mineral inclusion boundaries because of the refraction of x-rays. However, we cannot discriminate low-LAC materials such as water and organics from pore (air) by the absorption contrasts alone.

Scanning imaging x-ray microscopy

SIXM is a hybrid tomography system consisting of scanning x-ray microscope optics with 1D focusing (line-focusing) device and imaging x-ray microscope optics with a 1D objective (8). In the experiment, samples were illuminated with a line-focused beam and the transmitted and deflected x-ray beams were recorded in a 2D image detector. By rotating and scanning the samples with the line-focused beam, we can obtain both CT images of x-ray differential phase contrasts and x-ray absorption contrasts at the same time. We only used the phase-contrast images because absorption contrast images with higher resolution and less noise than SIXM were obtained by DET.

The phase-contrast images are a spatial distribution of x-ray RID, δ, of an object, which is the difference of x-ray refractive index, n, from unity

$$n=1-\mathrm{\delta }$$ (1)

The RID value is a function of x-ray energy almost irrespective of the element species and is closely proportional to the density of the object. Furthermore, the phase contrast has a higher sensitivity than the absorption contrast in particular for light elements, where the RID/LAC sensitivity ratio exceeds a thousand. Therefore, the RID values are useful for discriminating materials particularly formed by light elements such as water and organics from air (Fig. 1D), which cannot be discriminated by the LAC values in x-ray absorption images. Here, the SIXM experiment was performed at 8 keV. Multiple projection images (750 projections) were taken by rotating the sample in small intervals through a total of 180° (0.24° per projection). Horizontal scan pitch was set to ~100 nm. Both phase and absorption CT images in 3D were reconstructed by a convolution back-projection algorithm. The voxel size of the CT images is ~100 nm, which results in an effective spatial resolution of about a few hundreds of nanometers.

In the phase-contrast images, nanosized inclusions are more poorly recognized than in the absorption images because the edge-enhanced contrast seen in DET images is not produced in the SIXM (Fig. 1, B and C). Nevertheless, we can determine whether water (or organics) is present inside micrometer-sized inclusions by using the RID values (Fig. 1D). If an inclusion is small, the RID values become larger than the real RID values because of overlapping of the two point spread functions from the both calcite walls as shown in a micrometer-sized inclusion (m-inc) in the A-A′ section of Fig. 1D. However, a RID value smaller than the RID value of water and liquid CO₂ shows that the inclusion is fluid free.

Image processing and analysis

As the LAC and RID values obtained by the reconstruction are slightly smaller than the real values, we corrected the LAC and RID values by calibration using standard materials as performed in (3). We searched for inclusions in the CT images by both DET and SIXM, and their locations in the samples in 3D were recognized using an open source image processing package “Fiji-ImageJ.” Inclusions in calcite grains were digitally extracted by binarization of 3D-CT images, and their bird’s-eye views were made (Fig. 1, E to G) using the software Slice (27).

FE-S/TEM experiments

FE-S/TEM-EDS observation and elemental analysis at room temperature

The ultrathin sections made from the two XCT samples and directly from the five grains were analyzed using a FE-S/TEM (JEOL, JEM-2100F, at Kyoto University) with an EDS (JEOL, JED-2300T) operated at 200 kV [process (5) in fig. S1]. TEM images were recorded using charge-coupled device cameras (Gatan, Orius 200D and Orius 1000A) and analyzed using the software package Gatan Digital Micrograph. Crystal structural identification was based on SAED patterns (Fig. 3 and fig. S2). The SAED patterns were analyzed using the software package ReciPro (28). The crystallographic orientation of the facets was estimated from the TEM image and SAED (fig. S2). We also acquired annular dark-field STEM (ADF-STEM) images and x-ray spectra and their elemental maps using the STEM-EDS system (figs. S3 and S4). Elemental maps containing data of x-ray intensities in individual pixels were acquired by scanning a focused electron beam with beam current (10 to 100 pA) at a high speed (0.1 ms per point) to avoid the x-ray count decrement caused by electron beam irradiation damage. No quantitative analysis was made. The identities of trapped or daughter crystals in some inclusions, such as Na-Mn and Fe sulfates, FeOOH (or Fe₂O₃) (fig. S3), and apatite (Fig. 2C and fig. S5), were estimated by their EDS elemental maps and spectrum, but are too small to be defined by SAED.

FE-S/TEM-EDS observation using a cryo-stage

After the TEM observation and analysis at room temperature, we transferred the TEM samples from a conventional stage to a cryo-stage (Gatan, 636 double tilt liquid nitrogen cooling holder) and observed fluid inclusion candidates, which are relatively large and may preserve fluid based on the room temperature observation (<100 nm in size, apart from twin boundaries, and having a relatively dark contrast in TEM bright field), at 173 and 293 K using the same FE-TEM with the same observation conditions except for the temperature. It took about less than 30 and 90 min to decrease and increase the temperature between 293 and 173 K, respectively, but we waited for about ~60 and 180 min, respectively, to stabilize the temperatures and the stage drift before measurements were made.

We obtained SAED patterns of the fluid inclusion candidates at 293 and 173 K to detect any extra spots by frozen product(s) of possible fluid at 173 K (Fig. 3). Only one inclusion showed a definitely clear SAED pattern of extra spots of ice. This may suggest that the probability of the presence of fluid is originally very small. Another possibility is the loss of fluids (decrepitation) from many inclusions during the long history of the meteorite, preparation of the samples by FIB microsampling, handling of the TEM samples by mechanical disturbance including thermal stress, and/or observation under SEM and TEM by electron beam irradiation.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/7/17/eabg9707/DC1