Impact-driven oxidation of organics explains chondrite shock metamorphism dichotomy

Abstract

Shocked meteorites can be used to probe the dynamics of the early Solar system. Carbonaceous chondrites are less shocked than ordinary chondrites, regardless of the degree of aqueous alteration. Here, we show that this shock metamorphic dichotomy is a consequence of impact-driven oxidation of organics that are abundant in carbonaceous but not ordinary chondrites. Impact experiments at 3–7 km s⁻¹ using analogs of chondrite matrices reveal evidence of local heating in the matrix up to ~2000 K. Impacts on carbonaceous asteroids cause explosive release of CO and/or CO₂, which can efficiently remove evidence of shock. We show that highly shocked materials are lost to space from typical-sized chondrite parent bodies (~100 km in diameter), but are retained on the largest known carbonaceous asteroid, namely, (1) Ceres, due to its stronger gravity. Ceres’ surface is thus a witness plate for the ancient impact environment of carbonaceous chondrite parent bodies.

Due to its strong gravity, Ceres’ surface still retains evidence of past collisions, which have been lost via impact-driven oxidation of organic material from typical C-rich asteroids, according to impact experiments

Introduction

Shocked meteorites are Rosetta stones to decode the dynamical environment in the early Solar system¹ because a single impact event produces a variety of shock metamorphism in different minerals². Progressive shock stages have been constructed^2,3 to classify shocked meteorites^3,4. Figure 1 shows the number fraction of ordinary (NC)⁴ and carbonaceous (CC) chondrites³ as a function of shock degree. There is a clear dichotomy between NCs and CCs in terms of shock metamorphism. The fraction of highly shocked CCs ranked $\ge$S3 ( > 15 GPa) is only ~15%, whereas $\ge$63% of NCs are classified into $\ge$S3. Samples returned from carbonaceous asteroid Ryugu did not contain obvious shocked features⁵, except for a micro fault^6,7, a high-pressure mineral⁶, and small crystallographic misorientations in an olivine grain⁸. Since the shock pressure on Ryugu was estimated to be 2–10 GPa^6–8, Ryugu samples were classified as S1–S2 ($\le$10 GPa). Impact-induced dehydration of phyllosilicates has been proposed qualitatively as a mechanism to explain the virtually unshocked nature of hydrous CCs including CI and CM chondrites⁹; however, anhydrous CCs, such as CO and CV types, are also less shocked than NCs. In addition, a recent reevaluation of dehydration from phyllosilicates during impacts demonstrated that water vapor production is much lower than previously expected¹⁰.

Fig. 1. Shock metamorphic dichotomy in chondrites.

The number fraction of ordinary (NC, red thick) and carbonaceous (CC, blue thin) chondrites as a function of shock degree. The data are taken from the previous studies^3,4. The corresponding approximate shock pressure of each shock degree is also shown on the top x-axis.

It has been known that ~80% of meteorites, including NCs and CCs, come from a small number of parent bodies that have young asteroid families^11–15. These asteroid families were formed due to catastrophic disruption recently (10–1000 Ma) with respect to the age of the Sun^11,12 and their associated meteorites were likely launched from the families at 1–100 Ma by impact events¹⁶. Consequently, the parent bodies are expected to witness impact bombardments for up to ~4.5 Gyrs. For example, the Chelyabinsk meteorite recorded at least eight impact events¹⁷. Despite the limited number of parent bodies, shocked meteorites provide a rich record of impact.

The straightforward interpretation of the less-shocked nature of CCs is that impact velocities on the parent bodies of CCs were lower than on those of NCs. It is widely known that an isotopic dichotomy between NCs and CCs exists on a ε⁵⁴Cr–ε⁵⁰Ti diagram¹⁸ and hence that the birth places of NC and CC parent bodies were spatially divided in the early solar nebula¹⁹. According to this model, the parent bodies of CCs form beyond the orbit of the proto-Jupiter while those of NCs form in the inner Solar System. Since the impact velocity distribution is expected to be proportional to the Keplarian velocity v_K around the Sun (v_K$\propto$a^−1/2, where a is the semi-major axis), impact velocities onto the CC parent bodies could be as low as ~1 km s⁻¹ in the Kuiper Belt Object (KBO) region^20,21 compared with the typical impact velocity of ~5 km s⁻¹ in the main belt²². This scenario seems to be consistent with the shock metamorphic dichotomy. However, the parent bodies of CCs must migrate from their original orbit (>5 AU) to the Main Belt Asteroid (MBA) region (2–5 AU) at some stage²³. The most plausible driving force of such migration is gravitational scattering during the early orbital instability of the gas giants, including Jupiter, Saturn, Uranus, and Neptune^24–26. Moreover, the strong remanent magnetization of Ryugu particles⁵ suggests that the parent body of carbonaceous asteroid Ryugu was injected into the MBA region before the dissipation of the solar nebula, that is, within several Myrs after the formation of the Sun²⁷. If most CC parent bodies also migrated at this time, the parent bodies of both NCs and CCs must have spent ~4.5 Gyr in the MBA region under strong gravitational perturbation from Jupiter. Consequently, the sample evidence from Ryugu and recent dynamical theories imply that the impact conditions of NC and CC parent bodies were likely similar to each other during past 4.5 Gyrs.

Here, we propose a hypothesis that the shock metamorphic dichotomy is a consequence of the difference in response to impact shock between NC and CC parent bodies rather than different impact conditions for NC and CC bodies. Since CC bodies contain abundant organics in their matrices, the mechanical and thermochemical properties of CC materials differ substantially from those of NC materials. We conducted impact experiments with analog materials of chondrite matrices to investigate the effects of organics on the samples’ response to impact. The impact velocity v_imp was varied from 3 to 7 km s⁻¹, which spans the typical v_imp in the main asteroid belt²². The methodology is summarized in the Methods section.

Results

Roles of carbon in shock-induced devolatilization

Our impact experiments used four porous target pellets, named Fe–O, Fe–O–C, Si–O, and Si–O–C. We used commercial powder reagents of magnetite (Fe₃O₄), quartz (SiO₂), and graphite to make analogs of the chondrite matrices of aqueous-altered and anhydrous carbonaceous chondrites. We simplified the target compositions to clearly isolate the effects of the presence of the carbon on the total vapor production. The other reasons for our choice of powders and the justification for the choice in detail are described in the Methods section. The porosity of the pellets was adjusted to ~60%, which is close to the typical value of the chondrite matrices²⁸, except for a few cases to investigate the effects of porosity on the outcomes (Supplementary Note, S1 and Fig. S1). The atomic ratio of C to O for the target was set to 0.1, which is close to the C/O ratio of CI chondrites²⁹ as a fiducial case. The corresponding carbon content f_C of the Fe–O–C and Si–O–C targets was 2 wt.% and 4 wt.%, respectively. Since the carbon content of the Si–O–C targets, which are analogs of the matrices of CO and CV chondrites, was higher than the actual chondritic value, we investigated the effects of carbon content of the Si–O–C targets on the impact outcomes (Supplementary Note, S2). We confirmed that the dependence of the total production of CO and CO₂ on f_C was not significant in the case where the targets contain carbon down to 0.3 wt.% of f_C. The total vapor production is proportional to f_C^0.6 (Supplementary Fig. S2). In this sensitivity test, we also used ¹³C amorphous carbon for the Si–O–C targets to increase the signal-to-noise ratio in mass spectrometry especially at relatively low impact velocities. Figure 2 shows the results of the mass spectrometry at an impact velocity v_imp of 4.2–4.8 km s⁻¹. This impact velocity is close to the typical impact velocity in the main asteroid belt²². Hereafter, we denote the ion current for the mass number M/Z = i as I_i. We clearly detected a rise in all or some of I₂₈(CO⁺), I₃₂(O₂⁺), and I₄₄(CO₂⁺) from the targets except for the Si–O target, showing that the vaporization of magnetite and quartz and/or the oxidation of the organics occurs. Although vaporization of the Si–O target was not detected at v_imp = 4.5 km s⁻¹, a rise in I₃₂ for the Si–O target was observed at velocities higher than 5.8 km s⁻¹. Since the experimental chamber was buffered by an argon gas flow up to 500 Pa to minimize chemical contamination by the gases inevitably produced by the gun operation³⁰, the ion current ratio of I_i to I₄₀(Ar⁺) corresponds to the partial pressure of the gas species having M/Z = i in the experimental chamber. Although the I₄₀ level was slightly different from shot–to–shot, it was stable during the measurements even at the time of the impact (t = 0 s, where t is the time after impact). We can quantify the total production of CO, O₂ and CO₂ based on the ion current ratios. The procedures for the calibration of the measuring system and data reduction are described in detail in the Supplementary Information (Supplementary Note S3 and Figs. S3 and S4). Figure 3 shows the total vapor production M_vap as a function of impact velocity. The sum of the masses of CO, O₂, and CO₂ at each shot is shown here. The comparison of M_vap between Fe–O and Fe–O–C, and Si–O and Si–O–C, shows the presence or absence of carbon greatly affects the total vapor production, and the presence of carbon substantially enhances the degree of vaporization of the matrix minerals up to two orders of magnitude. The gas production of each gas can be found in Supplementary Note, S4 and Fig. S5. The effects of the incidence angle of projectiles onto the target on the total vapor production should be noted. It is widely known that all natural impacts occur at an oblique incidence angle³¹, while here we only performed vertical impacts. According to oblique impact experiments³², the heat generated during an impact is the sum of shock heating and subsequent shear heating. The heat generated due to the former and latter mechanisms exhibits an inverse correlation with impact angle at least down to 15 degrees measured from the horizontal³². The enhancement of shear heating in a more oblique impact tends to compensate for the weaker shock heating. Consequently, we suggest that the total gas production shown in Fig. 3 is a first-order estimation in the case of oblique impacts.

Fig. 2. Mass spectrometry data.

Time variation of the ion currents for gases with the mass numbers I_i. The target compositions are shown on the top of the panels. The impact velocities v_imp of the shots (indicated in the panels) were similar. Also shown are results with ¹³C amorphous carbon (Si–O–¹³C) and a blank experiment using an aluminum plate for comparison. The carbon content of the Si–O–¹³C target was 1.6 wt.%, which is 40% of the Si–O–C target (See Supplementary Note S2).

Fig. 3. The enhancement of the total vapor production due to the presence of carbon.

The total vapor production normalized by the projectile mass as a function of impact velocity for hydrous (A) and anhydrous (B) chondrite analogs. Vapor production from the Si–O target was not detected at impact velocities lower than 5.8 km s⁻¹. The atomic ratio of carbon to oxygen for the targets with carbon was 0.1 (See text). The error bars are come from the fitting errors in the calibration experiments (Supplementary Note S3 and Supplementary Fig. S4).

Discussion

Evidence for local high temperature

The impact experiments reveal that local energy concentration occurs in the porous targets in cases both with and without carbon. First, we discuss the C-free targets, that is, Fe–O and Si–O. We conducted shock physics and thermochemical modeling to investigate the degree of heating in the porous magnetite (Fe–O) and quartz (Si–O) targets. As equations of state were available or could be easily constructed for these single-mineral materials, we calculated the mass, momentum, and energy transfer from the impact point to the inside of the target with the iSALE shock physics code^33–36. Since the matrix particle size of 1–10 μm is much smaller than the projectile diameter of 2 mm, we treated the targets as continuous microporous media with uniform initial porosity in the shock physics code. Further details are described in the Methods section (See also Supplementary Notes, S5 and, S6, Supplementary Fig. S6, and Supplementary Table, S1).

Figure 4 shows a comparison of the decomposed target mass between the experimental and numerical results. We found that the actual decomposed mass in the experiments is only 5–40% of the numerical predictions. In contrast, we detected the decomposition of quartz and magnetite in the experiments at a velocity where decomposition does not occur in the modeling. We attribute both differences to localized concentrations of heat caused by a non-uniform distribution of porosity. Our simulations assume uniform initial porosity and hence a uniform degree of pore crushing and shock heating. Local variations in porosity in the real samples would create zones hotter and cooler than the average as demonstrated by shock physics modeling at the meso-scale^28,37, which would lower the bulk threshold of the onset of incipient devolatilization and shield some of the sample from decomposing.

Fig. 4. Comparison between the experimental and numerical results.

The red circles show the results for impact velocities from 3.2 to 7.0 km s⁻¹. The green squares include data where porosity ranged from 51% to 79%. The error bars on the experimental data have come from the fitting errors in the calibration experiments as well as Fig. 3. The gray dashed lines show 5%, 40% and 100% of the numerical results.

Second, we discuss our results for the targets with carbon, which are Fe–O–C and Si–O–C. In this case, the absence of available equation of state models precluded investigation by shock physics simulations. Instead, the measured molar ratio of CO to CO₂ can be used as a thermometer because it is controlled solely by the temperature-dependent oxygen fugacity¹⁰. We obtained theoretical curves of temperature versus CO/CO₂ ratio for the Fe–O–C and Si–O–C targets with the NASA CEA code³⁸. Figure 5A shows the measured CO/CO₂ ratio as a function of impact velocity. The CO/CO₂ ratio pertaining to the Fe–O–C targets increases as v_imp increases and the CO/CO₂ ratio for the Si–O–C targets is nearly constant. Figure 5B shows the predicted CO/CO₂ ratio as a function of temperature assuming thermal equilibrium. We can estimate the temperatures at the locations where the oxidation of carbon occurred by comparing the observed CO/CO₂ ratios in the experiments with the thermochemical model curves. We found that the temperature of the reaction field exceeds 2000 K even at v_imp ~ 4 km s⁻¹. Further results of the thermochemical calculations are described in Supplementary Note S5. The implied gas temperature is much higher than the temperature required for decomposition of magnetite and quartz under the condition where carbon coexists (See Supplementary Fig. S6). The unexpected high temperature is further evidence for localization of heat in the porous target. The energy localization results in a heterogeneous distribution of temperature during shocks in chondrite matrices, which is consistent with the observation of a polished thin section of CR2 chondrite GRA 06100²⁸.

Fig. 5. Gas temperature estimation.

A shows the molar ratio of CO to CO₂ measured in the experiments as a function of impact velocity. The error bars are the uncertainties in the molar ratio due to the error propagation from the CO and CO₂ productions. B shows the equilibrium composition of the molar ratio as a function of temperature. In this calculation, the elemental abundances were set to the same as the targets. The shaded regions correspond to the experimental results.

Explosive dispersal makes an “apparent” dichotomy in shock metamorphism

Here, we discuss the sequence of events after hypervelocity impact on a carbonaceous asteroid. Local heating of the matrix is expected to cause the generation of CO and CO₂ with the temperatures above 2000 K at least in the velocity range performed in this study (from 3 km s⁻¹ to 7 km s⁻¹). The volume of highly shocked materials, classified as ≥S3, will be limited to the immediate vicinity of the impact point, and form a thin high-temperature layer at the surface of the growing crater¹⁰. The generation of high-temperature gas due to energy localization will lead to pulverization and explosive dispersal of the thin, high-temperature, highly shocked material layer. Highly shocked particles would hence be accelerated into space due to the ram pressure from the expanding CO and CO₂ gases. The internal energy of impact-generated vapor is converted into kinetic energy of a spherically expanding gas flow directed into space³⁹. We can estimate the mass escaping from the host body M_esc due to the explosion by energy conservation as follows:

$$M_{\mathrm{vap}}{C}_{\mathrm{V}}T=\frac{1}{2}{M}_{\mathrm{esc}}{v_{\mathrm{esc}}}^2$$ 1

where M_vap, C_v, T, and v_esc are the total vapor production defined by the experiments (see Fig. 3), the isochoric specific heat of the major gas species, the temperature of the gas generated by the local energy concentration, and the escape velocity from the host body, respectively. Our experiments show M_vap normalized by the projectile mass is described as a power law of v_imp; hence, we obtain

$$\frac{M_{\mathrm{esc}}}{M_{\mathrm{proj}}}=2C_{\mathrm{V}}T{v_{\mathrm{esc}}}^{-2}\alpha {v_{\mathrm{imp}}}^{\beta },$$ 2

where M_proj is the projectile mass, and α, β are fitting constants to the experimental results (Supplementary Note S7, Supplementary Table S2). Figure 6 shows M_esc normalized by M_proj as a function of v_imp. The C_v values of CO and CO₂ were taken from the NIST–JANAF thermochemical table⁴⁰. Kurosawa et al. (2021)¹⁰ conducted a series of impact experiments with a CI chondrite simulant, which includes phyllosilicates, magnetite, sulfides, and organics, and measured the generated gases with the same method used in this study. The major gas was CO₂ regardless of impact velocity as well as the Fe–O–C targets used in this study. The data of the CI chondrite simulant can be also used to investigate the escaping mass from the parent bodies of hydrous CCs. Further details are described in the Methods section. Since the escaping materials are pulverized, they should be removed from the Solar system due to the interaction with the solar wind, which has been known as Pointing–Robertson effect⁴¹, within a very short time with respect to the history of the Solar system.

Fig. 6. Escaping mass from carbonaceous asteroids.

The graph shows the mass escaping from the host bodies M_esc as a function of impact velocity. The colored lines are calculated with the Eq. (2). The upper and lower lines of the Si–O–C correspond to the carbon content of 4 wt.% and 0.45 wt.%. The latter carbon content is an average of CO chondrites²⁹. The upper line was obtained the experimental results and the Eq. (2). We converted the upper line to the lower one with the dependence of the carbon content on the total vapor production (Supplementary Note S2 and Fig. S2). The diameter of the host bodies and the escape velocities from their surface are indicated on the right side of the graph. The mass experiencing the compression higher than 15 GPa M_sh in the case of the collision between two basaltic bodies⁴³ is shown as the black dashed line. We also show the results for CI chondrite simulant as the green line based on the previous result¹⁰ (See “Methods”).

To explore the efficiency of dispersal of shocked materials as a function of parent body size we consider two cases where impact occurs on a typical carbonaceous asteroid with a diameter of 100 km⁴² and the asteroid (1) Ceres. To serve as a reference, we define the shocked mass M_sh as material that would be classified as ≥S3 ( > 15 GPa). We estimate M_sh (also normalized by M_proj) by considering a vertical impact of nonporous basalt projectiles on basalt plates⁴³, which is also plotted in Fig. 6 as a black dashed line. The basalt line can be considered an upper limit of the shocked mass classified to ≥S3 for typical collisions between rocky bodies because it assumes vertical impact and neglects porosity. The comparison between M_esc and M_sh shows that for typical 100-km diameter carbonaceous asteroids highly shocked rocks are efficiently ejected to space. In contrast, rocks of the same shock state would be retained on the dwarf planet (1) Ceres, which is the largest known body in the asteroid belt, due to the stronger surface gravity.

Some primitive ordinary chondrites that are classified as petrologic type 3 also contain carbon up to 0.6 wt.%⁴⁴. This carbon content does not differ substantially from the values of CO and CV chondrites²⁹. According to extensive observations of the shock stage classification of ordinary chondrites⁴, the degree of shock metamorphism depends on petrologic type, and highly shocked ordinary chondrites of petrologic type 3 are relatively rare. The variation in the number fraction of meteorites with different shock stages as a function of petrologic type therefore also supports our hypothesis.

Methods

Impact experiments

We used a two-stage light gas gun at the Planetary Exploration Research Center of Chiba Institute of Technology, Japan⁴⁵. Helium and hydrogen gases were used for projectile acceleration. An Al₂O₃ sphere with a diameter of 2 mm was used as a projectile. A nylon-slit sabot⁴⁶ was used to accelerate the projectile.

We used a quadrupole mass spectrometer (QMS, Pfeiffer Vacuum, Prisma plus QMG220) with a technique referred to as the two-valve method³⁰ to prevent contamination of mass spectra from the gases inevitably produced by the gun operation. This method allows direct measurement of impact-generated gases in a fully open system with the same geometry as natural impact events.

Target preparation

We used magnetite (Fe₃O₄) and quartz (SiO₂) powders to produce target pellets. Magnetite is the second major constituent of CI^47,48 and Ryugu particles⁵ and is abundant in CM chondrites⁴⁹. Although phyllosilicate is the most abundant component in hydrous chondrites⁴⁷ and Ryugu particles⁵, we did not include this material in our target because a recent re-evaluation revealed that the degree of dehydration of phyllosilicates during impacts is much lower than previously expected¹⁰. Note that Ryugu particles, CI and CM chondrites also contain remarkable amounts of sulfides^5,48,49. Kurosawa et al. (2021)¹⁰ investigated the gas production and species from pellets made from a CI chondrite simulant⁴⁷, which contains sulfides. They confirmed that the impact-produced volumes of sulfur-bearing gases, such as H₂S and SO₂, were much smaller than carbon-bearing ones. Thus, we did not consider any sulfides in our samples in this study. Quartz was chosen as an analog of the silicates in the matrices of CO and CV chondrites, which are anhydrous carbonaceous chondrites, although silica is not abundant in any chondrites in general. However, the difference between silica and relevant silicates, such as olivine, pyroxene, and feldspar, does not substantially affect our experiment because of the similarity of their physical and thermoelastic properties, resulting in similar shock responses to quartz. The thermochemical behavior of quartz at high temperatures is also similar to that of olivine (Supplementary Note S5 and Fig. S6). Graphite and ¹³C amorphous carbon powders were chosen as the simplest insoluble organic materials (IOMs) because IOMs account for the majority of the total organic carbon in both Ryugu and CCs⁵⁰. Homogeneous powder reagents of magnetite, quartz, graphite and ¹³C amorphous carbon with a stable quality are also easy to obtain. This allows us to conduct systematic experiments with good reproducibility. We confirmed the identity of these reagents with X-ray powder diffraction analysis (Supplementary Note S8 and Figs. S8–S11). We excluded hydrogen from the matrix and the organics in our analogs. The presence of hydrogen does not substantially affect the carbon chemistry because water vapor is chemically stable under a wide range of temperatures and redox conditions⁵¹. This simplification is advantageous in mass spectrometry because it allows accurate determination of the molar ratio of CO and CO₂ of the impact vapor without the contamination of the signal by hydrocarbons.

Four types of porous pellets with a diameter of 54 mm and a length of 54 mm were made. Among them are (1) magnetite only, (2) a mixture between magnetite and carbon, (3) quartz only, and (4) a mixture of quartz and carbon. In this study, they are referred to as Fe–O, Fe–O–C, Si–O, and Si–O–C, respectively. The porosity of the pellets was adjusted to ~60% as a fiducial case, which is close to the porosity of chondrite matrices²⁸. We also conducted blank experiments at the velocity ranged from 3.3 km s⁻¹ to 7.1 km s⁻¹ with Al6061 plates. We confirmed that the generated CO, O₂, and CO₂ detected in the mass spectrometry mainly come from the target pellets, not from various sources of contamination, such as the intrusion of the combustion gases and the desorption from the chamber wall.

Shock physics modeling

We conducted a series of numerical simulations to investigate the degree of decomposition of magnetite and quartz in the non-transparent target pellets. We used the two-dimensional version of the iSALE shock physics code, iSALE-Dellen^33–36 to model the impact experiments conducted in this study. A cylindrical coordinate system was employed. The computational domain was divided into 500 × 1500 cells in the radial and vertical directions. We also set extension zones with 200 cells on the outside of the domain. The size of the computational cells in the extension zone was enlarged as a sequence of equal ratios with an extension factor of 1.02. The projectile was divided into 50 cells per projectile radius (CPPR), which corresponds to a spatial resolution of 0.02 mm cell⁻¹. Impact velocities were set to the same values as the experiments. We used the Tillotson equations of state (EOS)⁵² with the parameter set for Al₂O₃⁵³ to represent the projectile and used the ANalytical Equations Of State (ANEOS)⁵⁴ for quartz⁵⁵ and magnetite to represent the targets. The parameter set for magnetite was constructed in this study based on the shock Hugoniot data⁵⁶ and thermoelastic properties of magnetite^57,58 (See Supplementary Note S9 and Figs. S12 and S13). Lagrangian tracers were inserted into all the computational grid cells in the high-resolution zone. The time variation of entropy during the simulations was stored for each tracer. We calculated the decomposed mass of magnetite and quartz with the same method used in Kurosawa, Genda et al. (2021)⁵³. Since the target pellets were made from powders, the cohesion, which is defined as yield strength at zero pressure, is expected to be much lower than consolidated rocks. Although the dry friction between particles in the target pellets contributes to the yield strength of the target pellets, it would be weakened or diminished due to temperature rise⁵⁹ caused by work done during porosity compaction and the local energy concentration. Thus, we neglected any material strength in this study. The input files used in the iSALE simulations are included in Supplementary Data 1 and 2. Note that the absence of appropriate equations of state model for the mixtures of minerals and carbon precluded shock physics modeling for the target pellets with carbon as mentioned in the main text.

Explosive dispersal

The main gas products from the Fe–O–C and the Si–O–C were CO₂ and CO, respectively. We considered only the main gas product in Fig. 6. Since isochoric specific heat C_v depends on temperature, a vapor temperature of 2000 K was taken for the Fe–O–C and the Si–O–C based on Fig. 5. The C_v values of CO and CO₂ at 2000 K are 0.996 and 1.182 kJ K⁻¹ kg^{−1 40}, respectively, implying that the specific internal energy E of the generated gas reaches ~2 MJ kg⁻¹. The thermal energy is converted to kinetic energy of the gas flow directed to space. The gas flow accelerates the shocked materials within the thin high-temperature layer as mentioned in the main text. Note that we also show the results for CI chondrite simulant based on the previous result¹⁰. The main gas released from the CI simulant was CO₂ as well as the Fe–O–C target. A vapor temperature of 1200 K¹⁰ and C_v of 1.092 kJ K⁻¹ kg^{−1 40} were used to calculate the line for the CI simulant in Fig. 6.

Author contributions

K.K. carried out the impact experiments, analyses, and numerical simulations, and wrote the initial draft of the manuscript. K.K., G.S.C., T.M.D., and T.M. conceived the initial research idea. K.K., G.S.C., T.M.D. constructed the numerical model suites. T.O. and K.I. assisted the impact experiments. All the authors, unfortunately except for late T.M., discussed the results and their implications.

Peer review

Peer review information

Nature Communications thanks Kieren Torres Howard, Boris Ivanov and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the figures in the main text and Supplementary Figs. can be found in Source Data. The title of each source data contains the title of the figure produced used that data. The input files related to the iSALE simulations conducted in this study can be found in Supplementary Data 1 and 2. Source data are provided with this paper.

Code availability

The iSALE shock physics code and the NASA CEA code were used in this study. The iSALE is not fully open-source, and it is distributed on a case-by-case basis to academic users in the impact community for non-commercial use only. A description of the application requirements can be found at the iSALE website (https://isale-code.github.io). Any recent stable release can be used to reproduce the data. The NASA CEA can be used via the website (https://www1.grc.nasa.gov/research-and-engineering/ceaweb/).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-58474-2.