Direct observation of shock-induced formation of davemaoite, Ca-perovskite

Abstract

Perovskite-structured silicates are key minerals of Earth’s lower mantle and have also been identified in shocked meteorites as the (Mg,Fe)SiO₃-perovskite phase, bridgmanite. In contrast, CaSiO₃-perovskite (davemaoite) has never been found in meteorites and is known only from rare inclusions in deep-mantle diamonds. This contrast highlights a fundamental gap: while natural samples provide indirect evidence for silicate perovskites under shock conditions, laboratory dynamic compression experiments have yet to directly capture perovskite transitions in either MgSiO₃ or CaSiO₃, likely due to kinetic limitations imposed by nanosecond timescales. Here we present the in situ, atomic-scale observations of CaSiO₃-perovskite formation under dynamic compression. Using nanosecond laser-driven shocks and ultrafast X-ray diffraction (XRD) at SACLA, we tracked the transformation of CaSiO₃ from low-pressure phases to an amorphous state, and ultimately to davemaoite above ~91(10) GPa, suggesting a disorder–order transformation pathway toward reconstructive crystallization on nanosecond timescales. Upon release, davemaoite undergoes amorphization via a transient low-symmetry phase. These observations provide direct structural evidence for silicate perovskite formation under dynamic loading, resolving a long-standing experimental problem and establishing dynamic compression as an effective route to probe deep-mantle minerals relevant to Earth and Ca-rich exoplanet interiors.

Shock compression is an effective approach for studying materials under extreme conditions. Using X-ray diffraction, Lee et al. observe shock-induced davemaoite formation, demonstrating that silicate perovskite can crystallize on nanosecond timescales.

Introduction

Perovskite-structured silicates dominate Earth’s lower mantle. CaSiO₃-perovskite (davemaoite), the third most abundant mineral in this region, accounts for ~10 vol% of a pyrolitic mantle and up to ~30 vol% in subducted basaltic lithologies^1,2. Unlike bridgmanite, however, davemaoite cannot be quenched to ambient conditions and reverts to lower-pressure phases upon decompression^3–5. Natural occurrences show the same contrast: bridgmanite has been recovered from shock meteorites^6,7, whereas davemaoite is known only from rare inclusions in super-deep diamonds^6,8, suggesting fundamental differences in their shock formation and preservation mechanisms. These observations highlight long-standing questions about the mechanisms that govern the formation, stability, and preservation of silicate perovskites under extreme conditions.

Despite its scarcity, davemaoite plays a central role in planetary evolution. It hosts heat-producing elements (U, Th, K)⁹, contributes to seismic heterogeneity detectable in the lower mantle^10–12, and may incorporate trace amounts of H₂O^13,14, making it important for Earth’s thermal evolution, mantle geochemistry, and convection. Its relevance extends beyond Earth: davemaoite is expected to remain stable in the perovskite structure at pressures where bridgmanite transforms to post-perovskite (pPv)¹⁵, emphasizing its importance to the interiors of super-Earths. Both astrophysical observations and condensation models suggest that Ca- and Al-rich rocky exoplanets are plausible: stars with unusually high Ca abundances¹⁶ and high-temperature condensation processes of planetary building blocks^17,18 point to the possibility that davemaoite may dominate the mantles of Ca-enriched super-Earths. Constraining the properties of davemaoite is therefore essential not only for understanding Earth’s mantle, but also for a broader class of rocky exoplanets.

However, despite its geophysical importance, direct in situ structural evidence for davemaoite formation under dynamic compression has remained elusive. This knowledge gap persists because of experimental challenges. Over the past two decades, numerous dynamic compression experiments on silicates have been performed^19–34. While recent work reported crystallization of shock-compressed CaSiO₃ glass to perovskite³⁵, in situ XRD of crystalline silicates has consistently revealed amorphous or metastable phases rather than the crystalline perovskite^19–21,36. These results underscore the formidable kinetic barriers that inhibit silicate crystallization on nanosecond timescales and the absence of direct structural evidence for perovskite formation under shock.

Here, we demonstrate that these kinetic barriers can be overcome for CaSiO₃. Using nanosecond laser shocks synchronized with femtosecond X-ray pulses, we present the in situ, atomic-scale measurements of the shock-induced transformation of CaSiO₃ into a crystalline phase, consistent with davemaoite. We resolve the structural pathway—from low-pressure polymorphs through an amorphous intermediate to the perovskite structure, and its amorphization upon release. These observations provide the direct structural evidence for davemaoite under dynamic compression, resolving a long-standing challenge in high-pressure mineral physics and establishing dynamic compression as an effective tool for probing deep-mantle minerals and rocky planetary interiors.

Results & discussion

The experiments were carried out at the BL3-EH5 endstation of the SPring-8 Ångstrom Compact Free-Electron Laser (SACLA)³⁷. An optical laser was used to drive shock waves into the targets, and femtosecond X-ray pulses probed the structural response (Fig. 1). A line-imaging velocity interferometer system for any reflector (VISAR) monitored shock planarity and interface or free-surface velocities for pressure determination. The delay time of shock arrival at the LiF window or free surface was defined as t = 0, such that t > 0 corresponds to compressed and partially (or fully) released states, whereas t < 0 represents compressed and ambient regions (see Methods for details). All the shot conditions and results are summarized in Supplementary Table 1.

Fig. 1. Schematic of the experimental configuration.

A 532-nm optical laser generated shock waves into a target consisting of a 27 µm polyimide (Kapton), a 10–20 µm Al-coated CaSiO₃, and a 100 µm [100]-oriented LiF window (inset). Diffracted X-rays were recorded using flat panel detector. A 515-nm line-imaging VISAR system monitored wave arrival times and velocities at the sample and window interface or at the free surface. The VISAR system consisted of two channels with field-of-view diameters of 210 and 320 μm, respectively. The target was tilted at a 45° angle relative to both the X-ray and drive-laser axes. The diffraction image shown on the detector corresponds to shot #1477544, identified as davemaoite.

Polycrystalline wollastonite CaSiO₃ was compressed to pressures of 33(2)–127(9) GPa (Table 1). Figure 2 shows representative XRD images (left) and the corresponding one-dimensional integrated patterns (right). Between 33(2) and 59(6) GPa, diffraction patterns consisted of spotty reflections rather than continuous rings. Due to limited data quality and the structural complexity expected in this pressure regime, we collectively designated these features as the low-symmetry phase (Supplementary Fig. 1). Above 59(4) GPa, these features were replaced by two broad peaks (the first and second sharp diffraction peaks, FSDP and SSDP), indicating an amorphous state. Such broad scattering features closely resemble amorphization observed in other silicates under shock loading^19–21. We verified the amorphous assignment by comparing the peak positions and widths with high-pressure data on other silicate glasses (Supplementary Figs. 1 and 2), given the lack of experimental data for amorphous CaSiO₃ above 20 GPa. The distinctly broader FWHM values compared to davemaoite further support the disordered nature of this phase. Importantly, this state appeared at P–T conditions below the melting line (Fig. 3), confirming that it represents a metastable disordered solid rather than a liquid-like state. Above 91(10) GPa, sharp reflections emerge with the amorphous background, providing the direct structural evidence for a crystallization pathway involving a transient amorphous intermediate under dynamic loading – a disorder–order pathway previously predicted by theory or static studies^38–41. Static-compression studies place davemaoite within its equilibrium stability field during compression at substantially lower pressures than those required for its first appearance in our shock-compression experiments (Fig. 3 and Supplementary Fig. 3)⁴². The delayed appearance of davemaoite under shock loading therefore indicates a substantial kinetic overpressure associated with phase transformation on nanosecond shock-compression timescales (see Methods for details of phase identification).

Table 1.

Pressure range of shock-compressed phases

Phase	Pressure range (GPa)
Low-symmetry	33(2)–59(6)
Amorphous	59(4)–66(4)
Mixed	87(6)–99(13)
Davemaoite	91(10)–127(9)

Phase determination from XRD. Pressure uncertainties represent 1σ.

Fig. 2. XRD results from shock-compressed polycrystalline wollastonite.

Two-dimensional XRD images in 2θ–φ space (left) and the corresponding integrated one-dimensional diffraction patterns (right). Peak pressure and X-ray probe time offset (Δt = t_X-ray – t_breakout) are shown for each shot. Asterisks indicate diffraction from ambient wollastonite. For compression data, patterns are ordered by increasing pressure from bottom to top. For release data, patterns are ordered by increasing X-ray probe time from bottom to top, showing the temporal evolution during decompression. Calculated peak positions for davemaoite are marked by blue ticks, and those for ambient wollastonite (with relative intensities >2%) are marked by thick black ticks. Colors correspond to different structural states: yellow, low-symmetry phase; green, amorphous phase; purple, mixed phase; and blue, davemaoite. Pressure uncertainties represent 1σ.

Fig. 3. Pressure–density and pressure–temperature relations of CaSiO₃.

a Pressure–density data of davemaoite from this study compared with the calculated Hugoniot. Shock data are shown for davemaoite (dark blue, this study), wollastonite starting materials (dark gray, gas-gun), and glass starting materials (light gray, gas-gun)^49,81,82. Davemaoite observed upon release is indicated by open symbols. The calculated Hugoniot is shown from the assumed transition pressure; the range above 90 GPa is shaded in light blue, representing the maximum and minimum bounds (see Methods). Our davemaoite data points fall within the calculated Hugoniot range, confirming the phase identification. The highest-pressure point (127(9) GPa) lies slightly below the Hugoniot, likely due to partial pressure release during the X-ray probe window. Equations of state are plotted as colored lines^11,86,90,91. Error bars on davemaoite data represent 1σ uncertainties. (wo, wollastonite; Pwo, pseudowollastonite; Byi, breyite; Lrn, larnite; Ttn, titanite). b Pressure–temperature relation for the calculated davemaoite Hugoniot compared with previous experimental and theoretical results. The onset of davemaoite formation observed in this study (~91 GPa) is indicated by dark blue arrow. Details of phase diagram are shown in Supplementary Fig. 3.

This mechanism stands in contrast to other silicates. In Mg-silicates, dynamic compression consistently yields dense disordered structures, and bridgmanite crystallization has never been observed in situ from either crystalline or glass starting materials^19,21,24. Previous studies showed that shock-compressed enstatite (MgSiO₃) and forsterite (Mg₂SiO₄) collapse into a metastable disordered solid below the melting line, suggesting that bridgmanite crystallization kinetics is too sluggish to operate on nanosecond timescales and requires microseconds for nucleation and growth¹⁹. In Mg-silicates, disorder represents a kinetic dead end on nanosecond timescales. Comparable behavior was observed in SiO₂, where the shock response depends strongly on the starting polymorph: fused silica transforms into the stable high-pressure phase stishovite on nanosecond or sub-microsecond^43–45, whereas quartz and coesite generally persist as metastable or disordered states^46,47. Recent molecular dynamics simulations of shock-compressed quartz demonstrate that the amorphous phase represents an intermediate state preceding crystallization into an ordered high-pressure phase, rather than a terminal disordered state⁴⁸.

The observed phase evolution, including amorphization at moderate pressures, the appearance of davemaoite at higher pressures, and their coexistence over a finite pressure range, is compatible with a similar amorphous-intermediate pathway under shock loading. However, the present data cannot establish whether davemaoite necessarily nucleates from the amorphous phase or whether part of the material bypasses the amorphous state during compression. Nevertheless, the observed phase evolution suggests that amorphization may play an important role in facilitating subsequent crystallization under highly non-equilibrium conditions. This disorder–order pathway represents a reconstructive phase transformation involving changes in Si–O coordination (from fourfold tetrahedral to sixfold octahedral), rather than a simple displacive transition. Such reconstructive character may favor the formation of an amorphous intermediate, while the combined effects of overpressure, shock-induced heating, and the larger size and chemistry of Ca²⁺ may contribute to a lower effective barrier for crystallization, potentially enabling CaSiO₃ to crystallize while Mg-silicates remain trapped in amorphous or metastable states.

Our findings complement recent gas-gun shock experiments on CaSiO₃ glass⁴⁹. The densities derived from the diffraction data are consistent with the calculated Hugoniot of davemaoite (Fig. 3a), confirming that the observed reflections correspond to the crystalline perovskite phase. This agreement validates the phase assignment and shows that crystalline wollastonite follows a similar thermodynamic compression path toward the perovskite stability field, as observed in gas-gun experiments with glass starting materials⁴⁹. Gas-gun measurements inferred complete davemaoite crystallization above ~120 GPa from sound-velocity measurements, whereas our in situ diffraction measurements reveal the onset of davemaoite near ~90 GPa, with sharp perovskite reflections emerging from an amorphous background (Supplementary Fig. 4). This difference reflects both the distinct sensitivities of the two techniques and differences in experimental conditions. Sound velocity responds to the bulk elastic properties of a substantial crystalline fraction, whereas XRD directly detects the initial emergence of long-range order. Additionally, different starting materials (crystalline wollastonite vs. glass) and shock-temperature paths (with glass expected to reach higher temperatures at comparable pressures) further influence transformation kinetics. These results demonstrate that davemaoite formation is not a sharp threshold event, but a process spanning a range governed by starting structure, nucleation pathway, shock-temperature path, and the detection method.

Following davemaoite crystallization, we observed its amorphization during decompression. The release pathway proceeds through a transient low-symmetry phase prior to amorphization (Fig. 2). Based on diffraction peak positions and inferred densities, the low-symmetry phase is broadly consistent with expanded breyite and/or wollastonite rather than pseudowollastonite (Supplementary Fig. 5). Hydrodynamic simulations indicate that at +3.69 ns, the sample has undergone nearly complete pressure release while retaining elevated temperatures from residual shock heating (Supplementary Fig. 6). However, due to the limited number of diffraction peaks and their overlap with broad amorphous features, we conservatively designate it as an unresolved low-symmetry phase (see Methods).

The amorphization observed upon release in our experiments is consistent with the behavior of CaSiO₃ in shocked meteorites, though contrasting outcomes are observed in other natural settings. In shocked meteorites, CaSiO₃ is typically found as glass in impact-melt veins rather than as crystalline phases^50–52, consistent with static compression studies showing that davemaoite transforms into an amorphous state upon release to ambient conditions⁵³. In super-deep diamonds, by contrast, high-pressure phases such as breyite, titanite-structured CaSi₂O₅ (Ttn), larnite (β-Ca₂SiO₄; Lrn), and more recently davemaoite have been reported as preserved^6,54–56. Breyite is widely interpreted as the retrogressed product of davemaoite formed during slow exhumation or cooling⁵⁷, although alternative formation mechanisms (e.g., CaCO₃ + SiO₂ breakdown) have been proposed⁵⁸.

These contrasting outcomes reflect differences in decompression rate and thermal path⁵⁹. In our laser-shock experiments, nanosecond-scale release under elevated temperatures favors amorphization via a low-symmetry intermediate. This behavior is similar to natural impact events where CaSiO₃ is predominantly preserved as glass^50–52,60. In contrast, in diamonds, the slow exhumation and near-equilibrium cooling stabilize breyite or, in rare cases, davemaoite itself^50–52. Such kinetically controlled behavior during decompression is also observed in bridgmanite^60,61, highlighting the general importance of decompression rate and thermal evolution in governing silicate phase transitions.

The successful synthesis of davemaoite under dynamic loading demonstrates its kinetic accessibility on nanosecond timescales, opening a direct pathway to experimentally constrain its equation of state at the multi-megabar pressures relevant to super-Earth interiors. Current mass–radius models for rocky exoplanets rely largely on extrapolations of low-pressure data or theoretical equations of state⁶², introducing substantial uncertainties in estimates of interior structure and composition. Direct experimental access to davemaoite under such extreme conditions would eliminate this dependence on extrapolation, providing the empirical constraints on Ca-rich planetary mantles. Achieving this requires experimental platforms capable of reaching extreme pressures while maintaining solid-state conditions and minimizing shock heating. Ramp (quasi-isentropic) compression is well suited to meet these requirements and has already been successfully applied to probe phase transitions and equation of state in oxides and metals to terapascal pressures^63–68. Davemaoite is particularly well-suited for such studies: its Ca-rich composition (Ca:Z = 20) enhances X-ray scattering intensity, while its high-symmetry cubic perovskite structure produces simple and well-resolved diffraction patterns that facilitate phase identification and quantitative analysis. Such capabilities would provide benchmark constraints for models of Earth’s deep mantle and the interiors of Ca-rich exoplanets, refining our understanding of silicate behavior and planetary structure across the terrestrial regime.

Methods

Starting materials and sample preparation

CaSiO₃ targets were synthesized in a piston-cylinder apparatus at the Carnegie Institution for Science by sintering wollastonite powder (Alfa Aesar, 99% purity) at 1.5 GPa and 1100 °C for 24 h. The XRD measurements under ambient conditions at the SACLA facility confirmed the starting phase as wollastonite. A Le Bail fit of the XRD pattern yielded a crystallographic density of 2.9379(1) g/cm³. The bulk density, measured by the Archimedes method, was 2.91(1) g/cm³, corresponding to a porosity of ~1%; these effects were therefore neglected. The thickness of the wollastonite layer was optimized using one-dimensional Lagrangian hydrodynamic simulations with HYADES to ensure that the probed region was under uniform pressure conditions during compression⁶⁹. These simulations also provided estimates of the pressure–time history during compression and release, but not pressure determination (Supplementary Fig. 6).

Laser-shock platform and target configuration

Experiments were conducted at the BL3-EH5 endstation of the SACLA. A 532-nm optical laser (5-ns duration, 3–19 J energy) was focused onto a 260 µm² focal spot to drive shock waves into the target assembly at oblique incidence (Supplementary Fig. 7). The target assembly consisted of a 27-µm-thick polyimide (Kapton) ablator glued to a 10–20-µm-thick polycrystalline CaSiO₃, as shown in Fig. 1. A ~ 200-nm-thick aluminum coating was deposited on the rear surface of the CaSiO₃ to provide reflectivity for the VISAR, which was used to monitor shock breakout. A subset of targets included a ~ 100-µm-thick [100]-oriented LiF window. All layers were attached with ∼1-μm-thick epoxy layers (Fig. 1).

Pressure determination

Peak pressure states were controlled by adjusting the incident laser energy. A line-imaging VISAR system operating at 515 nm monitored shock planarity and recorded velocity histories at the sample–LiF interface or at the free surface (Supplementary Fig. 8)⁷⁰. Two VISAR legs with different velocity sensitivities (2.874 and 1.077 km/s per fringe) were used to resolve fringe-jump ambiguities. Although high-sensitivity legs are generally preferred in VISAR analysis, under our experimental conditions, the low-sensitivity VISAR1 (2.874 km/s per fringe) yielded superior data quality and was therefore used for direct analysis (Supplementary Fig. 9).

The measured surface velocities at the LiF interface were corrected for shock-induced changes in the refractive index of LiF. Although the refractive index is not strictly linear under shock loading⁷⁰, a linear fit is required for conventional VISAR analysis. We therefore compiled recent shock-compression data for LiF at 514 and 532 nm and derived a linear calibration (Supplementary Fig. 10)^71–79.

Peak pressures were determined for all samples. For shots with LiF windows, standard impedance matching was applied (Supplementary Fig. 11)⁸⁰, with interface velocities obtained from the flat-top regions of the velocity profiles using Monte Carlo sampling (Supplementary Fig. 9). The equations of state for CaSiO₃ (wollastonite) and LiF were taken from previous studies, with the CaSiO₃ Hugoniot refined by segmented least-squares fitting to our results (Supplementary Fig. 12)^75,81,82. Due to the sparse nature of previous shock data for wollastonite, the Hugoniot was divided into two regimes separated at u_p = 2.5 km/s: a low-pressure phase (LPP, u_p < 2.5 km/s) and a high-pressure phase (HPP, u_p > 2.5 km/s).

For samples without LiF windows, peak pressures were inferred from an empirical relationship between laser energy and peak sample pressure derived from shots with LiF windows (Supplementary Fig. 13):

$$P\left(\mathrm{GPa}\right)=11.8\left(9\right){E\left(\mathrm{J}\right)}^{0.77\left(4\right)}.$$ 1

Uncertainties in peak pressure were quantified using Monte Carlo error propagation, accounting for errors in VISAR fringe analysis, equations of state, and the empirical energy fitting.

XRD measurement and analysis

XRD measurements were performed using 10.01(1) keV, a bandwidth of ΔE/E ~ 0.2%, and a pulse duration of ~7 fs from SACLA, focused to a 30 × 20 µm² (2024) or 40 × 10 μm² (2025) spot centered on the optical laser drive. An individual XRD pattern was collected for each shot (Supplementary Fig. 14). The X-ray probe was synchronized with the shock breakout time so that diffraction patterns were recorded during steady shock compression. Throughout this article, the X-ray probe timing is expressed as Δt = t_X-ray – t_breakout. The systematic error in Δt is estimated to be 0.2 ns, implying that even at Δt = 0, the effective probe time may deviate by up to ±0.2 ns⁸³. As a result, some XRD data may include ambient conditions (Δt <0) or partial release features (Δt > 0). For selected targets, the probe was intentionally delayed beyond breakout to capture release behavior. In these cases, changes in the distance between the sample and the detector were taken into account.

Diffracted X-rays were collected with a flat panel detector covering 14°–72° in 2θ. Detector calibration used CeO₂ or Au standards, and azimuthal integration was performed with Dioptas⁸⁴. For targets with LiF windows, single-crystal reflections from uncompressed and compressed LiF were masked prior to 1D integration. Lattice parameters of compressed davemaoite and wollastonite were determined by Le Bail fitting of the 1D patterns using GSAS-II (Supplementary Fig. 15)⁸⁵.

Phase identification under compression

The starting material was wollastonite (space group P$\overline{1}$), and both the low-pressure phase of davemaoite and its metastable structures are characterized by triclinic or monoclinic symmetries⁸⁶. The low-symmetry diffraction patterns observed in this study were spotty rather than ring-shaped, which further complicated the analysis (Supplementary Fig. 1). Given the inherently low data quality of dynamic compression experiments, specific phase assignments were not attempted. Instead, we collectively designate these features as the “low-symmetry phase,” since most candidate structures in this pressure regime exhibit low symmetry. Operationally, patterns were classified as low-symmetry when large, distinct spots appeared in the 2D diffraction images that were clearly distinguishable from ambient features (Supplementary Fig. 1). The low-symmetry phase was observed in the 33(2)–59(6) GPa range. Although additional low-symmetry data were collected, only the shot with an LiF window (used to constrain the pressure–energy relation) and the boundary case leading into the amorphous regime are reported here.

Above 87(6) GPa, the peak widths decreased, and new reflections appeared at positions where no peaks had been observed in the previous amorphous phase (Supplementary Fig. 15). These peak positions are consistent with cubic davemaoite. However, in the 87(6)–99(13) GPa range, we observed a transitional regime where weak (200) and (211) reflections appeared as isolated spots rather than continuous rings. We classified these patterns as “mixed phase,” as reliable lattice fitting was not possible (Supplementary Fig. 4). Well-defined davemaoite diffraction patterns were observed in the 91(10)–127(9) GPa range.

Davemaoite structure assignment

CaSiO₃ perovskite is known to adopt both tetragonal and cubic structures. The tetragonal phase was first identified in 1996 and later confirmed in various studies^11,87,88. Distinguishing tetragonal from cubic symmetry requires diffraction resolution sufficient to resolve subtle peak splitting^88,89. In our study, the perovskite phase forms within the pressure–temperature stability field of the cubic phase, and we did not observe any resolvable splitting indicative of tetragonal symmetry. Therefore, we interpret our results in terms of the cubic davemaoite phase. We define the davemaoite phase as present when the FWHM value of the FSDP is less than 3.0 (Supplementary Fig. 2), indicating sufficient crystallinity to resolve the (200) peak from the amorphous background and to clearly identify the (211) reflection in 2D images. Le Bail fitting of the davemaoite diffraction profiles confirmed that peak positions are consistent with the expected cubic perovskite structure, although intensity mismatches suggest mixing with residual amorphous components (Supplementary Fig. 15).

Phase identification during decompression

The release data were collected at a peak pressure of ~110 GPa. The X-ray calibration was performed with reference to the pre-shock configuration. However, as illustrated in the experimental schematic (Fig. 1), after the breakout time, the sample moves at the free-surface velocity, leading to a mismatch with this calibration. While this effect is negligible up to Δt = +16.5 ns based on free-surface velocity estimation, it becomes significant at longer delays. The dataset with an X-ray probe time offset (Δt) of +3.69 ns exhibited a diffraction pattern composed of rings. However, since only two distinct sharp diffraction features were observed, we compared these two peaks with the calculated diffraction patterns of the candidate structures by isotropically adjusting their lattice densities based on the ambient structure (Supplementary Fig. 5). None of the candidates fully matched the observed features. Notably, pseudowollastonite required a structure more compressed than its ambient form to reproduce the observed patterns and was therefore excluded from consideration. Consequently, the phase at this stage was identified as breyite and/or expanded wollastonite, collectively referred to as the “low-symmetry phase.” The +16.5 ns dataset showed no sharp diffraction features. Although we cannot definitively determine whether this state was liquid or glassy, we infer that it was an amorphous phase.

Hugoniot calculations

A theoretical Hugoniot for davemaoite was calculated using an experimentally constrained isentropic compression curve and the Mie-Grüneisen equation of state²⁰. The transition energy E_tr was estimated using:

$$E_{tr}\cong P_{tr}\mathrm{\Delta }V.$$ 2

where P_tr and ΔV are the transition pressure and the volume change, respectively. We constrained P_tr between 20 GPa, where the slope of shock data changes rapidly, and 90 GPa, where davemaoite was first observed, and used this range to estimate E_tr (Supplementary Fig. 3). ΔV was determined from the 300 K compression data at 20 and 90 GPa. At the lower bound (20 GPa), ΔV was taken as the volume difference between wollastonite and breyite, whereas at the upper bound of E_tr (90 GPa), ΔV was taken as the difference between wollastonite and cubic davemaoite^11,86. Although cubic davemaoite is not stable at ambient conditions and wollastonite becomes unstable above ~9 GPa, we nevertheless applied bulk moduli reported in previous studies to the Birch–Murnaghan equations of state to calculate ΔV^11,86. We also calculated the release isentrope path for davemaoite (Supplementary Fig. 16). The thermodynamic parameters used are listed in Supplementary Table 2.

Author contributions

D.K. conceived and supervised the project. M.L., S.T., M.J., G.L., H.L., N.T., J.C., K.M., T.Y., N.O., T.P., H.N., A.A., H.H., and D.K. performed shock experiments at SACLA. M.L., M.J., G.L., Y.L., D.S., C.P., Y.F., S.J.T., and H.H. prepared the samples. M.L. and D.K. performed data analysis. M.L. and D.K. wrote the manuscript with support from all authors. H.H. and D.K. supervised the project.

Peer review

Peer review information

Nature Communications thanks Takehiko Yagi, Jinping Hu, and Youjun Zhang for their contribution to the peer review of this work. A peer review file is available.

Funding

D.K. discloses support for the research of this work from the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT [RS-2023-00284250 and RS-2024-00342193], and a Korea University grant. C.P. discloses support for the research of this work from the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education [RS-2024-00464096]. Y.L. discloses support for the research of this work from the National Research Foundation of Korea (NRF) funded by the Ministry of Education [RS-2023-00301938] through the Global-Learning & Academic research institution for Master’s and PhD students, and Postdocs (LAMP) Program. N.O. discloses support for the research of this work from the Japan Society for the Promotion of Science (JSPS) KAKENHI [22K14128, 23K20038, and 25H00618] and the MEXT Quantum Leap Flagship Program [JPMX0118067246].

Data availability

The raw X-ray diffraction and VISAR data generated in this study have been deposited in the Figshare database under accession code [https://doi.org/10.6084/m9.figshare.31932522].

Code availability

Igor code [AnalyzeVISAR] developed by Lawrence Livermore National Laboratory (LLNL) and used to analyze the VISAR data described here is available from the corresponding author upon request, with permission from LLNL.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76624-y.