Introduction

To meet the rapidly evolving technology needs, the materials research community has been striving to seek new methodologies for advanced materials design. Increasing the number of components could effectively expand the configurational space for the exploration of novel materials. Recently, multiple principal components (>5) alloys, so-called high-entropy alloys (HEAs) have been successfully developed in metallic systems with solid solution phases. These HEAs are realized based on an idea with deliberately incorporating five or more elemental species into a single solid solution lattice with random site occupancy, which maximizes the configurational entropy to stabilize the system1,2,3,4,5. Surprisingly, this high-entropy strategy also works for oxides with ionic bonds. Rost et al. first synthesized the transition metal (Mg0.2Co0.2Ni0.2Cu0.2Zn0.2)O high-entropy oxide (HEO) with a rocksalt-type structure6. Subsequently, Djenadic et al. synthesized single CaF2-type phase rare-earth HEOs (RE-HEOs) (Ce0.2La0.2Pr0.2Sm0.2Y0.2)O2−δ7. More recently, complex HEOs with multiple cation Wyckoff positions, like perovskites8,9 or spinels10, have also been reported. HEOs provide great flexibility for element-based property design and engineering, therefore, have attracted extensive attention owing to their tremendous potential in various applications. Many remarkable properties of HEOs have been reported, including huge dielectric constant11 and high room temperature Li ion conductivity12, which have advanced several potential applications like defect dipoles13, epitaxial interface engineering14, and electrode materials in Li-ion battery15,16.

High-entropy materials are also expected to possess ultrahigh structural stability due to their maximized configurational entropy in design (high-entropy effect), which could dramatically lower down the Gibbs free energy of the system. On the other hand, the high chemical disorder could induce severe lattice distortion which could raise the energy of the system. Therefore, it is compelling to experimentally clarify the stability and tunability of these novel materials under various conditions, which is crucial for our fundamental understanding of those new materials and also their potential applications17,18. High-temperature is usually involved during the formation of single phase HEOs6,19. Besides temperature, pressure is another fundamental parameter controlling the structure and properties of materials20,21,22,23. Especially, for HEOs, the Gibbs energy (G = PV − ST), the local lattice distortion, the high-level oxygen vacancies, and even electronic states (valency) are all expected to be susceptible to pressure tuning, which, however, have not been explored before. Very recently, the seemingly ultra-stable HEAs have already been found to show rich polymorphism under high-pressure at room temperature. For example, pressure-induced irreversible polymorphic transitions from the fcc to hcp phases were observed in the prototype CoCrFeMnNi HEA24,25, and a medium-entropy NiCoCrFe alloy26; In another rare-earth system, an hcp-structured HoDyYGdTb HEA, more complex transitions following the sequence of hcp → Sm-type → dhcp → dfcc were reported27; Even five polymorphs were discovered in the bcc Al0.6CoCrFeNi HEA18,28. These results demonstrate that the HEAs are far more complex than expected as ultrastable solid solutions. Inspired by these recent results of HEAs, an intriguing question is raised: are the structure and properties tunable in HEOs?

In this work, we use in situ high-pressure synchrotron radiation X-ray diffraction (XRD), pair distribution function (PDF), Raman spectroscopy, and ex situ high-resolution transmission electron microscopy (HRTEM) to investigate the structural evolution of the rare earth (Ce0.2La0.2Pr0.2Sm0.2Y0.2)O2−δ RE-HEO ((Ce,La,Pr,Sm,Y)O will be used hereafter for simplicity). Lattice distortion in the initial structure is confirmed by the slight deviation of bond angles from the ideal values. The bond angles are found to be extremely susceptible to pressure, which continuously bend during compression and eventually causes a pressure-induced amorphization (PIA) in HEOs. The optical properties and band gap are also investigated through the transition by in situ high-pressure ultraviolet–vis (UV–vis) absorption spectroscopy, which shows interesting tunable behavior as well accompanying the structural change.

Results

Structure of the initial sample

(Ce,La,Pr,Sm,Y)O HEO powder samples were synthesized using nebulized spray pyrolysis (NSP) method7. The initial structure of the sample was characterized by chromatic-aberration corrected HRTEM and synchrotron radiation high-energy XRD. The experimental details are shown in the “Methods” section. HRTEM results (Supplementary Fig. 1) show that the sample has a particle size of ~0.5 µm, a crystalline grain size of ~7 nm, and a CaF2-type (space group: Fm-3m) structure with its lattice parameter a = 5.6 (1) Å. The structure factor S(q) obtained from the high-energy XRD with the moment transfer q coverage up to 30 Å−1 is shown in Fig. 1a, the structure can be well indexed into a CaF2-type structure with a refined lattice parameter a = 5.507(1) Å. These results of the initial structure of the (Ce,La,Pr,Sm,Y)O HEO are consistent with the previous reports7. Fig. 1b shows the PDF data, G(r), which was obtained by a Fourier transformation of the S(q) data. Bond lengths can be accurately determined by fitting the peaks in G(r), then, bond angles can be calculated. The initial average bond angles between oxygen and rare-earth elements, θRE–O–RE (~108.6°) and θO–RE–O (~68.3°), both are slightly smaller than the ideal values of a standard CaF2-type crystal structure (θRE–O–RE = 109.5° and θO–RE–O = 70.5°), this is also consistent with the observation of smaller bond angles in the HRTEM results (Supplementary Fig. 1c), which may be attributed to the local lattice distortion in the (Ce,La,Pr,Sm,Y)O HEO. The lattice distortion could be caused by atomic size mismatch between neighboring RE elements, owing to the fact that cationic radii of the constituent rare-earth elements vary considerably, for example, from 0.97 Å (for Ce4+) to 1.116 Å (for La3+). Additionally, the high level of oxygen vacancies could also play an unneglectable role resulting in the observed lattice distortion29.

Fig. 1
figure 1

High-energy XRD of the initial sample. The structure factor a and pair distribution function pattern b of the (Ce,La,Pr,Sm,Y)O HEO at ambient conditions obtained by high-energy X-ray diffraction. The X-ray wavelength was 0.1173 Å. A transmission mode was used for the sample without any container

In situ high-pressure XRD

To address the structural stability of the (Ce,La,Pr,Sm,Y)O HEO, its structural evolution up to ~31 GPa was monitored by in situ high-pressure synchrotron radiation XRD in a symmetric diamond anvil cell (DAC). A schematic illustration of a DAC is shown in Fig. 2a. Figure 2b presents the XRD patterns during compression and decompression. During compression, the initial CaF2-type structure loses its crystal coherence above ~16 GPa indicated by the disappearance of the initial sharp Bragg peaks and the emerging of amorphous-like broad peaks. The amorphous-like phase exists up to the highest pressure of ~31 GPa and seems recoverable to ambient pressure during decompression without the reappearance of the sharp Bragg peaks. The two-dimensional (2D) XRD images are shown in Supplementary Fig. 2.

Fig. 2
figure 2

In situ high-pressure XRD. a A schematic illustration of the DAC used to generate high-pressure. b Representative XRD patterns of (Ce,La,Pr,Sm,Y)O during compression and decompression at room temperature. Inset is an image of the sample loaded inside a DAC, the scale bar is 100 μm. The peaks of the amorphous phase are marked by asterisk symbols. The X-ray wavelength is 0.3218 Å. c Pressure dependence of d/d0 of the CaF2-type cubic phase upon compression. Pressure dependence of the peak height d and peak width e of the (111) peaks (at ~5.8°) upon compression

Below 16 GPa, with increasing pressure, all the Bragg peaks of the initial CaF2-type structure shift to the larger two-theta values below ~9 GPa simply due to the pressure-induced lattice shrinkage (Fig. 2c). At ~4 GPa, there is a slight slope change, and the peak position stops shifting above ~9 GPa. Below ~9 GPa, the intensity of all XRD peaks gradually but significantly weakens (Fig. 2d), and all the peak widths continuously increase as well (Fig. 2e). Both peak intensity and peak width show a slope change at ~4 GPa and become almost constant above 9 GPa. The peak intensity weakening and peak width increasing start at the very beginning of compression. This result indicates that the initial CaF2-type structure of the (Ce,La,Pr,Sm,Y)O HEO is susceptible to pressure and continuously becomes more and more disordered (topologically) during compression. Below ~16 GPa, the structure evolution has three stages (one till ~4 GPa, another between ~4 and ~9 GPa, and lastly above ~9 GPa) according to the peak position, intensity, and peak width changes as a function of pressure.

In situ high-pressure PDF

Since structural disordering occurs in the (Ce,La,Pr,Sm,Y)O HEO during compression, to derive more detailed information of the structural evolution, another experiment with larger Q coverage (up to 12 Å) was performed (more details are shown in the “Methods” section). The in situ high-pressure structure factor S(q) was obtained after careful background subtracting and normalization (Fig. 3a). The peak position, intensity, and width of the S(q) exhibit consistent behavior with the raw XRD data shown in Fig. 2. The peak intensity (Fig. 3b) and width (Fig. 3c) significantly change their slopes at ~16 GPa. Above ~30 GPa, smooth amorphous patterns are obtained. Meanwhile, the peak intensity and width do not change with pressure anymore, suggesting the completion of the amorphization process.

Fig. 3
figure 3

Structure evolution in reciprocal space. a Selected structure factor of (Ce,La,Pr,Sm,Y)O HEO during compression and decompression at room temperature. b Pressure dependence of the peak height b and peak width c of the first sharp diffraction peak (~2.0 Å−1) upon compression

Structural information in real space is presented by PDF (Fig. 4a). Below ~20 GPa, the PDF patterns show strong oscillations without noticeable damping up to 25 Å, which is a typical feature of crystalline materials with long-range translational symmetry. In contrast, above ~20 GPa, all the oscillations are significantly weakened and gradually smeared out above ~10 Å, which is in line with the amorphization process indicated by S(q) data in reciprocal space. All the peaks of PDF which correspond to the distances between different atom pairs shift to smaller r values with increasing pressure below ~16 GPa. For convenience, the first nine peaks (illustrated in Fig. 4b) of the initial CaF2 structure are denoted as P1–P9 in Fig. 4a. Some peaks gradually weaken and merge into one new broad peak during compression, e.g., P1 and P2, P4 and P5. Above ~20 GPa, only four broad and weak peaks remain due to amorphization. The changes of the nine peak positions, the calculated atomic pair distances, and bond angles are shown in Fig. 5. The atomic pair distances shrink quite consistently below ~4 GPa, and then diverge at higher pressures, which could be the reason for the slight slope change at ~4 GPa observed in Fig. 2c–e. It should be noted that although the q coverage in the in situ high-pressure PDF measurement is limited (qmax ≈ 12 Å−1) due to the restricted opening of DACs, the S(q) and G(r) are reliable according to their consistent feature with the high-energy PDF of the ambient samples (qmax ≈ 30 Å−1, see Fig. 1).

Fig. 4
figure 4

Structure evolution in real space. a Selected pair distribution function (PDF) of the (Ce,La,Pr,Sm,Y)O HEO during compression and decompression at room temperature. The numbers denote the peaks for different atomic pairs. b The structure model for the cubic structure of (Ce,La,Pr,Sm,Y)O HEO. The distances between different atom pairs are shown with arrows, which correspond to peaks marked by numbers in a. Different colors of the cubes represent different atomic layers

Fig. 5
figure 5

Pressure dependence of atomic pair distances and bond angles. a The pressure dependence of the atomic pair distances associated with the peaks in Fig. 4a. The error bars are smaller than the symbol sizes. b Normalized atomic pair distance of the major peaks in Fig. 4a before amorphization. c The pressure dependence of the two representative bond angles θO–RE–O (in one cube) and θRE–O–RE (between the nearest-neighbor cubes)

The CaF2-type structure of the (Ce,La,Pr,Sm,Y)O HEO consists of REO8 hexahedrons (cubes) with a cubic close-packed arrangement, in which each RE atom is surrounded by eight O atoms forming eight RE–O bonds. All the REO8 cubes connect with each other by sharing one edge. The RE–O bonds inside REO8 cubes are quite rigid and remain almost constant during compression up to ~30 GPa, while all the other longer atomic distances between different REO8 cubes considerably decrease with increasing pressure (Fig. 5a, b). With accurately determined bond lengths by fitting the peaks in G(r), the pressure dependence of bond angles inside one cube θO–RE–O and between the nearest cubes θRE–O–RE can be calculated as shown in Fig. 5c. During compression, both of the average θO–RE–O and θRE–O–RE decrease continuously. The sample volume compression is achieved mainly by bending the bond angles. Eventually, the long-range coherence between REO8 cubes breaks down when the pressure is above the mechanical instability limit, hence, leading to amorphization.

Upon decompression, both S(q) and G(r) seem to retain their amorphous feature. However, regardless of the peak shifting due to elastic recovery, the S(q) pattern is slightly different from the high-pressure pattern at 34.6 GPa especially in the range from 2.5 to 4.5 Å−1. This result suggests that the low-pressure recovered state is different from the high-pressure amorphous state. G(r) also looks different after decompression. To clarify the structure change after the pressure release, HRTEM was performed on the recovered samples (Supplementary Fig. 3). The recovered sample mostly remains in a highly disordered amorphous state, but some fine nano-grains of the initial CaF2 crystalline structure show up in the amorphous matrix. Therefore, a composite HEO with amorphous matrix plus some super fine nanocrystals has been synthesized via compression and the following decompression processes of the initial HEO.

The structural stability of the prototype (Ce,La,Pr,Sm,Y)O HEO has been studied from room temperature to ~1023 K at ambient pressure, but no amorphization or any other phase transition has ever been observed7. A symmetry lowering phase transition from Fm3m to Ia-3 space groups was observed only after calcination at 1273 K for 1 h7. Its phase stability under high-pressure was not explored before. In contrast, the phase transitions under high-pressure have been extensively studied in all of its constituent rare-earth oxides in nano or bulk forms, for example, Ce2O330, CeO230,31,32, La2O333, Pr2O334, PrO235, Sm2O336, and Y2O337,38,39. Among them, however, the PIA process was only observed in the Y2O3 nanoparticles with sizes <16 nm39. The Y2O3 nanoparticles have a cubic structure (space group: Ia-3) at ambient condition, which consists of YO6 octahedra as the building block. The YO6 octahedra are less compressible than the global Y2O3 lattice. Therefore, during compression, the densification is mainly carried by bending the polyhedral-linking angles to squeeze the space between the octahedra. When the bending is beyond a critical mechanical limit, the structure collapses and transforms into an amorphous state at ~25 GPa. Smaller grain size usually results in higher transition pressures for the PIA39. This mechanism was also proposed for the PIA in many other polyhedral-linked materials such as the prototype oxide, SiO240. PIA refers to an unusual phenomenon with direct transformation of a crystal into an amorphous solid, which was firstly reported in ice Ih, afterwards, has attracted intense interest and observed in a wide range of substances41. Our results in this work demonstrate that a similar mechanism proposed for PIA in conventional oxides may also hold for the case of the chemically much more complex HEO according to the bonding length and angle changes during compression.

The (Ce,La,Pr,Sm,Y)O HEO consists of five cations with different ionic radii, and more importantly different oxidation states (valence), e.g., 3+ oxidation for La, Sm, Y; 4+ for Ce, and mixed multiple oxidation states (3+ and 4+) for Pr29,42. CeO2 is believed to be the critical component which stabilizes the single-phase of the HEO and provides a CaF2-type host lattice for the other cations. However, except Ce4+ and Pr4+, the other rare-earth cations including partial Pr are present in the 3+ oxidation states. To keep the charge balance of the system, high-level oxygen vacancies are naturally required and created in the (Ce,La,Pr,Sm,Y)O HEO29. Therefore, the oxygen vacancies and cations valence are closely associated in the HEO. The different ionic radii and a high level of oxygen vacancies could both contribute to the lattice distortion in the initial (Ce,La,Pr,Sm,Y)O HEO sample which may be the reason for the slight deviation of the bond angels from ideal values. However, according to the G(r) data, the first atomic shell consists of RE–O bonding is rigid without obvious bond shrinkage during compression. Therefore, the size of different cations may not change much as well during the entire compression. On the other hand, the atoms (both the RE cations and other oxygen inions) surrounding a given oxygen vacancy site are expected to displace from their original positions in a perfect lattice, which could also result in bond angle deviations. Therefore, the behavior of oxygen vacancies during compression could dramatically affect the structure evolution under high-pressure.

In situ high-pressure Raman spectroscopy

Oxygen vacancies of the (Ce,La,Pr,Sm,Y)O HEO can be readily detected by Raman spectroscopy29. Thus, we performed in situ high-pressure Raman spectroscopy (Fig. 6) on the (Ce,La,Pr,Sm,Y)O HEO in a DAC (details can be found in the “Methods” section), which could provide another indicator of the development of the lattice distortion of (Ce,La,Pr,Sm,Y)O HEO during compression. The F2g symmetric vibration mode of the eight-fold RE–O bond in the (Ce,La,Pr,Sm,Y)O which is typically located at ~450 cm−1 is barely visible. The Raman peak weakening for F2g can be explained by the symmetry breaking (the severe local lattice distortion) in the (Ce,La,Pr,Sm,Y)O caused by chemical disorder29. The Raman peak at ~600 cm−1 (VO band) at ambient conditions was confirmed to be the characteristic peak of oxygen vacancies of the (Ce,La,Pr,Sm,Y)O HEO which is relatively strong at ambient conditions. During compression, the VO band shifts to higher wavenumbers, and its intensity decreases significantly until ~20 GPa. This observation indicates that the oxygen vacancies may change their distribution and concentration during compression below 20 GPa, which could be associated with the relatively large compressibility of the vacancy spot and also the tunable valences of the Pr3+ to Pr4+ during compression43,44. Above ~20 GPa, the amorphization is completed, therefore, it is reasonable that the intensity of VO band remained almost constant. Naturally, the bond angles and local lattice distortion will be altered by the collapse of oxygen vacancies and oxidation state-induced change in the cationic radii of Pr cation, which is consistent with the results obtained by in situ high-pressure XRD. During decompression, the intensity and peak position of the Vo band gradually recovered to the initial values. This result suggests that the collapse of oxygen vacancies may be caused by elastic deformation and the change of cations valence during compression, which is reversible upon decompression and likely independent of the lattice disorder45. To clarify the role of each constitutent element of this HEO in its structural evolution under high-pressure, element-specific technique, such as x-ray absorption spectroscopy (XAS) is needed. However, in situ high-pressure XAS expecially extended x-ray absorption fine structure (EXAFS) is challenging to the (Ce,La,Pr,Sm,Y)O HEO in a DAC due to the either too high X-ray energy (K-edge) for the thin samples (<20 μm) or too low X-ray energy (L-III edge) for the thick diamond anvils (~2.4 mm), and too close absorption edges of some constitutent elements.

Fig. 6
figure 6

In situ high-pressure Raman spectra. In situ high-pressure Raman spectra of the (Ce,La,Pr,Sm,Y)O HEO in a DAC during compression a and decompression b at room temperature. The dotted line is the spectrum of the initial sample collected outside of the DAC at ambient conditions (the intensity was rescaled by a factor of 0.53). The peak intensity and peak width of the recovered sample almost return back to the initial values before compression. The Raman peak position c and intensity d as a function of pressure upon compression. A dramatic slope change can be found at ~20 GPa in both c and d

During decompression, most of the amorphous state could be kinetically retained by the possible large barrier required for the recovery of the initial cubic crystalline structure. During heating at ambient pressure, the (Ce,La,Pr,Sm,Y)O HEO will have a phase transition from the fluorite-type to the bixbyite-type phase transition with lower symmetry7. However, we did not observe this polymorphic crystal-to-crystal transition under high-pressure and room temperature. Instead, an amorphization takes place. We speculate that the crystal-to-crystal transitions may be hindered by a large energy barrier in this system, therefore, we also performed in situ laser heating XRD measurements at high-pressure (more experimental details can be found in ref. 46) on the (Ce,La,Pr,Sm,Y)O HEO (Supplementary Fig. 4). Thermodynamically more stable crystal phases (do not match the bixbyite-type phase) do exist at pressures before the amorphization and after the amorphization if the sample is heated above ~1100 K. A possible phase separation also cannot be neglected under high-pressure and high-temperature. More efforts are needed to solve the structure of the high-pressure and high-temperature crystalline phase in the future.

Pressure tuning of band gap

Since the structure of (Ce,La,Pr,Sm,Y)O HEO is extremely susceptible to pressure, we expected tunable properties during compression as well. In this work, we studied its band gap as a function of pressure using in situ high-pressure UV–vis spectroscopy. The indirect and direct gap values were obtained by fitting the absorption spectra using the Tauc relation (Fig. 7)29, i.e., [F(Rhv]1/n = A·(hv − Eg), where F(R) is the Kubelka–Mubelka–Munk function, h is the Planck’s constant, v is the frequency, A is a constant, and Eg is the band gap energy. Details of the fitting of Eg can be found in Supplementary Fig. 5. Before amorphization (<~16 GPa), the sample exhibits enhanced band gap with increasing pressure, this is unusual compared to the typical band gap decreasing in traditional oxides during compression47,48. It may be caused by the accumulated lattice distortion energy. Above 16 GPa, once the sample starts amorphization, the band gap turns to decrease sharply until the amorphization completes at ~22 GPa. Above 22 GPa, the band gap increases with pressure again. The complex pressure dependence of the band gap observed in the (Ce,La,Pr,Sm,Y)O HEO is consistent with its structure evolution during compression revealed by XRD. The enhanced band gap below 16 GPa and above 22 GPa during compression seems to be an intrinsic feature in both the cubic crystalline and amorphous states during compression. The drop of band gap between 16 and 22 GPa is likely associated with the amorphization process, which may be attributed to the distortion (strain) release due to the amorphization (bond breaking and atom rearrangement). Although there is no reliable data collected during decompression in this work due to our experimental issue, we expected a partially reversible band gap change during decompression according to its atomic structure evolution. More tunable functional properties of the HEOs under high-pressure are promising and worth exploring in the future. The band gap of the initial (Ce,La,Pr,Sm,Y)O HEO sample (direct band gap: 2.06 eV, indirect band gap: 1.52 eV) is much narrower than the parent CeO2 sample (direct band gap: 3.17 eV, indirect band gap: 2.76 eV). The origin for the band gap narrowing in the (Ce,La,Pr,Sm,Y)O HEO is not very clear. The tunable behavior of the band gap over a relatively broad range (changes by 9% for direct band gap and 24% for the indirect band gap from ~1 to ~10 GPa) achieved under high-pressure could help to understand its underlying mechanism and even achieve band gap engineering via high-pressure treatment49,50.

Fig. 7
figure 7

In situ UV–vis absorption spectra. a In situ high-pressure UV–vis absorption spectra of the (Ce,La,Pr,Sm,Y)O HEO upon compression. b Fitted direct (blue balls) and indirect (red balls) band gap as a function of pressure. The error bars are from the linear fitting of the absorption spectra according to the Tauc plots

Discussion

The structural stability of the (Ce,La,Pr,Sm,Y)O HEO is studied during compression up to ~35 GPa using in situ high-pressure synchrotron radiation XRD. A crystal to amorphous phase transition is observed which starts at ~16 GPa and completes at ~22 GPa. PDF analysis reveals the microscopic mechanism of the amorphization, which is attributed to the breakdown of long-range connectivity between REO8 cubes caused by continuous bending/tilting of bond angles inside and between the REO8 cubes. The atomic distortion caused by chemical disorder and high-level oxygen vacancies is suggested to exist even in the initial crystalline HEO (the bond angles deviate from ideal values) and is extremely sensitive to pressure. The development of lattice distortion plays a vital role in the amorphization and band gap tuning. These results suggest that although the HEOs are very stable at high temperatures and ambient pressure, they are susceptible to pressure loading. Their properties could be highly tunable under high-pressure as well. Pressure-induced transformation directly from crystalline states to the amorphous solid states at temperatures below the melting point or glass transition temperature is well-known as the “PIA” phenomenon since the 1980s in various materials51. However, it has never been observed in the multicomponent HEO systems before. For HEO, the PIA also means the direct transition from ceramics to glasses, which could lead to the development of useful nano-ceramic-glass composite materials with tunable properties. Since the HEOs may commonly involve severe lattice distortion due to cation size mismatch and/or a high level of oxygen vacancies, we expect interesting pressure-induced transitions including PIA in other HEOs.

Methods

In situ high-pressure synchrotron XRD measurements

In situ high-pressure angle-dispersive XRD experiments were mainly performed at the GSECARS beamline 13-ID-D, Advanced Photon Source (APS), Argonne National Laboratory (ANL). A monochromatic X-ray beam with a wavelength of 0.3218 Å was focused by a Kirkpatrick–Baez (KB) mirror system down to ~2.5 × 3.5 μm2 at the rotation center of the sample stage. The detector was a Mar165 charge-coupled device (CCD). The position and orientation of the detector to the X-ray beam were calibrated using the diffraction pattern of a LaB6 standard. High-pressure was generated using a symmetric DAC with a culet size of ~400 μm in diameter. The gasket was T301 stainless steel that was pre-indented to ~20 GPa. A hole with a diameter of ~150 μm was drilled to serve as the sample chamber by a laser drilling system at the center of the gasket indent. The powder was pre-pressed into a thin pellet and cut into flakes with proper sizes for DACs. Then, the tiny flake was loaded into a DAC along with a tiny ruby ball and Au foil beside the samples to calibrate pressure. Methonal:ethonal = 4:1(volume ratio) mixture was used as the pressure-transmitting medium. The background scattering was collected at each pressure point by shining the X-ray beam on the empty area inside the sample chamber, which only passed through the pressure medium and two diamond anvils. The exposure time was set at 10 s for normal XRD. For PDF measurements, the exposure time for each pattern was set to be 30 s. One-dimensional (1D) XRD patterns were obtained by integrating the 2D patterns along the azimuth angle from 0° to 360° with the Dioptas software52. The resulting diffraction patterns were refined using the GSAS package.

In situ high-pressure Raman spectroscopy measurements

Raman spectroscopy measurements were conducted using a Renishaw inVia spectrometer with a 532 nm laser excitation. High-pressure was generated using a symmetric DAC with a culet size of ~400 μm in diameter. The powder sample was pre-pressed into a pellet and then loaded into a DAC, along with two tiny ruby balls beside the samples to calibrate pressure. Silicone oil was used as the pressure-transmitting medium. The exposure time for each pattern was ~50 s. The signal from background was obtained by shining a laser beam on the spot inside the sample chamber but away from the sample.

In situ high-pressure UV–vis spectroscopy measurements

UV–vis spectroscopy measurements were conducted using an Ocean Optics DH-2000-BAL spectrometer. High-pressure was generated using a symmetric DAC with a culet of ~300 μm in diameter. The powder sample was pre-pressed into thin flakes with right thickness and then loaded into a DAC, along with two tiny ruby balls beside the sample to calibrate pressure. Silicone oil was used as the pressure-transmitting medium. Each spectrum was collected for 1 s with wavelength ranged from 180 to 860 nm. The background signal was collected at the empty area (with pressure medium only) of the sample chamber.

HRTEM

The powder samples were dispersed in ethyl alcohol and picked up using a holey carbon grid. HRTEM experiments were carried out using the chromatic-aberration-corrected TEM (ACAT) at ANL with an image corrector to correct both spherical and chromatic aberrations. An accelerating voltage of 80 kV was used to minimize the possible electron irradiation effect on the sample structure.