Stable magnesium peroxide at high pressure

Rocky planets are thought to comprise compounds of Mg and O as these are among the most abundant elements, but knowledge of their stable phases may be incomplete. MgO is known to be remarkably stable to very high pressure and chemically inert under reduced condition of the Earth’s lower mantle. However, in exoplanets oxygen may be a more abundant constituent. Here, using synchrotron x-ray diffraction in laser-heated diamond anvil cells, we show that MgO and oxygen react at pressures above 96 GPa and T = 2150 K with the formation of I4/mcm MgO2. Raman spectroscopy detects the presence of a peroxide ion (O22−) in the synthesized material as well as in the recovered specimen. Likewise, energy-dispersive x-ray spectroscopy confirms that the recovered sample has higher oxygen content than pure MgO. Our finding suggests that MgO2 may be present together or instead of MgO in rocky mantles and rocky planetary cores under highly oxidized conditions.

and T = 2150 K with the formation of the theoretically predicted I4/mcm MgO 2 (Ref. 3 ). Raman spectroscopy detects the presence of a peroxide ion (O 2 2-) in the synthesized material as well as in the recovered specimen. Likewise, energy-dispersive x-ray spectroscopy confirms that the recovered sample has higher oxygen content than pure MgO. Our finding suggests that MgO 2 may substitute MgO in rocky mantles and rocky planetary cores under highly oxidizing conditions.

Main Text
Oxygen and magnesium are the first and second most abundant elements in the Earth's mantle 4 ; thus knowledge of stable phase relations in the Mg-O system as a function of thermodynamic parameters is necessary input information for reconstructing Earth-like planetary interiors. For example, ferropericlase (MgO with a relatively low Fe content) is the second most abundant mineral on Earth owing to its remarkable thermodynamic stability in the Fm3m crystal structure (up to 500 GPa and at least 5000 K for pure MgO) 5,6 . This is why ferropericlase has been assumed in gas giant cores 7,8 as well as in extrasolar terrestrial mantles 9,10 . However, planet-harboring stars vary in chemical composition 11 , which likely affects the composition of planetary building blocks and exoplanet mineralogy 2 . Therefore, Earth-like mantle mineralogy should not be assumed for terrestrial exoplanets. Elevated oxygen contents have been observed in planet-host stars 1 , which may affect the stability of MgO and favor other solid phases in the Mg-O system 3,12 . For example, magnesium peroxide (MgO 2 ) may be synthesized at nearambient conditions and at high oxygen fugacities in the pyrite-type (Pa3) structure 12 . However, Pa3 MgO 2 is thermodynamically unstable and readily decomposes to MgO and O 2 upon heating to 650 K at ambient pressure 12 . The intrinsic instability of MgO 2 is attributed to the strong polarizing effect of the Mg 2+ ion possessing high charge density in a relatively small ionic radius 13 . This is why the stability of Group II peroxides increases down the Group: beryllium peroxides are not known 13 , while Ca, Sr and Ba form increasingly more stable peroxides at ambient conditions 14,15 . Therefore, using empirical considerations on chemical pressure 16,17 MgO 2 may be expected to become stable under high pressure conditions. Indeed, ab initio simulations found that I4/mcm MgO 2 becomes stable at P > 116 GPa (Ref. 3 ) and 0 K. Here, we report on the synthesis of I4/mcm MgO 2 in laser-heated diamond anvil cell (DAC). MgO 2 may be an abundant mineral in the interiors of Uranus, Neptune, and highly oxidized terrestrial exoplanets. Our finding also suggests that the Mg-Fe-Si-O system likely has more unexpected chemistry at high pressure.
Two types of chemical precursors were loaded in DAC to study the MgO-O 2 phase diagram in the 0-154 GPa pressure range (see Table 1 and Methods). In type-A experiments we put two 4 µm thick MgO disks in the sample cavity which was subsequently filled with liquefied oxygen (Fig. 1, inset). In type-B runs we used commercially available magnesium peroxide complex (24-28% Pa3 MgO 2 , 42-46% MgO, ~30% Mg) mixed with submicron Au powder serving as a laser absorber. The mixture was loaded without pressure medium.   Fig. 1). Rietveld refinement of the new phase was not possible due to its spotty diffraction texture and because low intensity peaks could not be resolved.
In the experiments with type-B precursors MgO, ε-O 2 , and Au were the only phases observed in XRD patterns after it was heated to T > 2000 K in the pressure range of 5-70 GPa.
Bragg peaks that can be assigned to Pa3 MgO 2 were completely absent in the reaction products suggesting that the precursor had decomposed to MgO and O 2 . Indeed, the presence of pure oxygen in the quenched sample was confirmed with Raman spectroscopy. Noteworthy, we did not observe elemental Mg (neither hcp at P < 50 GPa nor bcc at P > 50 GPa) in the reaction products. Magnesium likely reacts with oxygen as the latter gets liberated upon Pa3 MgO 2 decomposition at high temperature.  Table S1) were used to construct the I4/mcm MgO 2 P-V equation of state (EOS) (Fig. 3). We also computed the MgO 2 volume in the 70-150 GPa pressure range (Supplementary Table S2).
The EOS parameters are reported in the Supplementary Table S3.   19 . The x-ray wavelength is 0.3344 Å.   Figure 4A shows   Raman spectra of I4/mcm MgO 2 were followed on A2 decompression run. In Figure 4B the pressure-frequency dependence of the A 1g band of I4/mcm MgO 2 is compared with that in BaO 2 (Ref. 24 ) and Pa3 MgO 2 (this study and Ref. 21 ). We could only trace the high-frequency band down to 50 GPa, and then at 0-10 GPa because of the overlap with the overtone of oxygen L2 peak (2υ L2 ) 22 21 suggesting that the recovered product is likely Pa3 MgO 2 . Overall, our data provide spectroscopic evidence for the peroxide ion in the synthesized material and that the material containing peroxide ion is preserved to ambient conditions.

Stability of I4/mcm MgO 2 and planetary implications
I4/mcm MgO 2 can be synthesized in the mixture of MgO with O 2 at 94GPa indicating a thermodynamic stability of MgO 2 at this pressure, which is close to the theoretical predicted pressure of 116 GPa (Ref. 3 ). We therefore conclude that I4/mcm MgO 2 is a thermodynamically stable phase in the high pressure phase diagram of the Mg-O system (Fig. 3, inset). On decompression, we could only follow the new phase in XRD down to 74 GPa, while Raman spectroscopy shows no discontinuities in the position of the high-frequency band down to at least 50 GPa. It remains unclear what physicochemical transformations occurred in the synthesized phase at P < 50 GPa, but at 1 bar the laser-heated area of the recovered sample (A2) ( Supplementary Fig. S3) shows a Raman signature of P a3 MgO2. Mapping the extracted sample with an energy-dispersive x-ray spectroscopy (EDS) revealed that the laser-heated area has higher oxygen content than the area that was not subjected to high temperatures ( Supplementary   Fig. S4). Detailed chemical characterization, however, was not possible because unreacted MgO is mixed with the oxygen-rich phase in the laser-heated area. Nevertheless, EDS analysis provides independent evidence for MgO 2 in the recovered sample.  Fig. S5 B, C).

Materials and samples
Diamond anvils with culets of 200, 300/100, and 300/80 µm were used to access the 100-150 GPa pressure range. Rhenium foils (200 µm thick) were indented to a thickness of 30-40 µm and then laser-drilled to create holes (30-100 µm in diameter) serving as sample chambers.
Magnesium oxide (99.85%) available from Alfa-Aesar was used for the type-A experiments.
Before sample loadings magnesia was annealed at 1293 K for 12 hours to get rid of any adsorbed water. Two MgO disks were made by compressing the magnesia powder to a thickness of 4-5 µm and were stacked in the gasket hole. The remaining volume of the sample chamber was filled with liquefied zero-grade oxygen (99.8%, Matheson Gas Products) at approximately 77 K. In type-B experiments we used magnesium peroxide complex available from Sigma-Aldrich (24-28% Pa3 MgO 2 , 42-46% MgO, ~30% Mg). The magnesium peroxide complex was mixed with submicron gold powder and loaded in the sample chambers with no pressure medium.

Synthesis and characterization
All XRD experiments were performed at the undulator beamline at 13ID-D GeoSoilEnviroCARS, APS, using the online double-sided laser-heating system 29 . The laserheating radiation was coupled to oxygen in type-A, and to the gold powder in type-B runs.
Synchrotron XRD was collected in-situ at high temperature and high pressure in the diamond anvil cells to determine the onset of chemical and physical transformations with the x-ray beam  GPa pressure range, which is within the computational uncertainty.

X-ray diffraction analysis
2D XRD patterns were integrated using the Dioptas software (written by C. Prescher).
Manual background subtraction was done in Fityk 2 . Preliminary Bragg peaks indexing was performed with Dicvol06 (Ref. 3 ). GSAS/EXPGUI 4, 5 was used for Rietveld refinement in accordance with the guidelines provided in Ref. 6 . Oxygen spotty reflections overlapping with the continuous lines produced by the new phase were masked. Also, we did not use the region of 2θ > 13° where the background scattering is not uniformly distributed in the azimuth range of 0 to 360°. Scaling factors and unit cell parameters were refined first. Subsequently, peak profiles (x in the 8h position). Crystal structures were visualized with the use of VESTA 3 package 7 . The 300 K third-order Birch-Murnaghan equation of state (EOS) was obtained using a (sigma)volume-weighted fitting procedure was performed as implemented in the EoSFit7GUI (R. Angel).

Computational Methods
Density functional theory (DFT) within the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) 8 as implemented in the VASP code 9 , was used for structural and vibrational analysis. For the structural relaxation, we used the all-electron projector-augmented wave (PAW) method 10 and the plane wave basis set with the 600 eV kinetic energy cutoff; the Brillouin zone was sampled by Г-centered meshes with the resolution 2π × 0.06 A -1 . The phonon frequencies were calculated using the finite displacement approach as implemented in the Phonopy code 11 . The Raman intensities were obtained by computing the derivative of the macroscopic dielectric tensor with respect to the normal mode coordinate 12 .