# Pressure-stabilized divalent ozonide CaO3 and its impact on Earth’s oxygen cycles

Sep 17, 2020

### Stable Ca–O compounds at high pressure

For insights to help find new calcium oxide compounds, we have employed unbiased crystal structure search techniques as implemented in CALYPSO code18,19, which has been successful in resolving crystal structures of a large number and variety of materials at high pressure20. Here, we explore calcium oxides in the oxygen-rich regime, seeking compounds that do not exist under ambient conditions. Studies of mantle rocks have shown that oxygen fugacity of the upper mantle is relatively high21, thus connecting the present work to prominent geological topics concerning oxidation states of minerals and oxygen storage and cycles inside Earth. We have performed structure searches on CamOn (m = 1, 2 and n = 2, 3, 4) with maximum simulation cells up to four formula units (f.u.) at each composition, and this procedure identifies two stable Ca–O compounds, a CaO2 phase at 30 GPa and an unusual stoichiometric CaO3 phase at 50 GPa. This result distinguishes CaO2 and CaO3 as two viable oxygen-rich calcium oxides. Furthermore, the CaO4-containing superoxide group (O2) is found to stay above but close to the convex hull, making it energetically more favorable with respect to the dissociation route into CaO + O2 above 34 GPa and thus may be experimentally synthesized at high temperature [see Supplementary Note and Supplementary Figs. 1–3 for details on the structure search results].

We characterize the newly identified calcium ozonide by examining its synthesis routes and structural, bonding, and electronic properties. The CaO3 phase crystalizes22 in a tetragonal BaS3-type structure23 (space group P-421m, 2 f.u. per cell) in a wide range of pressures and exhibits a distinct configuration containing isolated V-shaped O3 units and edge-sharing CaO8 cuboid (Fig. 1a). We compare to some well-established compounds on key structural and bonding characters of the crystalline CaO3 at 30 GPa, which is inside its stability field. The O–O bond length and O–O–O bond angle of the ozonide anion in CaO3 are 1.44 Å and 114.57°, respectively, larger than the corresponding values in KO3 (1.34 Å, 109.33°)16, indicating a weaker O–O bonding and more ionic nature due to additional electrons in the antibonding molecular orbitals (MO) of ozonide anion. The same pattern of relative bond lengths and angles for CaO3 and KO3 persists in a wide pressure range of 20–50 GPa (see Supplementary Fig. 4). Moreover, the shortest Ca–O distance in CaO3 (2.31 Å) is comparable to that in the prototype ionic compound CaO (2.27 Å)12, suggesting an ionic bonding between Ca and O3 in CaO3.

Formation enthalpy calculations reveal that CaO3 is energetically favorable relative to decomposition into CaO12 and solid O2 (ref. 24) above 27.2 GPa via the reaction:

$${mathrm{CaO}} + {mathrm{O}}_{mathrm{2}} = {mathrm{CaO}}_{mathrm{3}}$$

(1)

To account for thermal effects, we further examine vibrational contributions and entropic effects for the relevant phases25, and construct the finite-temperature phase diagram of CaO3. Calculated zero-point energy values at 30 GPa for CaO, solid O2, and CaO3 are 0.13, 0.22, and 0.32 eV/f.u., respectively, resulting in a minimal difference between the reactants and products for the above reaction of only −0.03 eV/f.u., which has only a minor impact on the threshold pressure for CaO3 decomposing into CaO and O2, reducing it from 27.2 to 25.3 GPa (Fig. 1b). The threshold pressure for the stability of CaO3 increases with rising temperature, going from 25.3 GPa at 0 K to 26.5 GPa at 2000 K. We further checked phonon dispersions of CaO3 at 20 and 50 GPa, and the results (see Supplementary Fig. 5) show no imaginary frequencies, indicating that CaO3 is dynamically stable in this wide pressure range, making it a metastable phase at lower pressures like diamond versus graphite.

### Experimental synthesis of Ca–O compounds

Our systematic assessment of energetic, dynamic, and thermodynamic stability of CaO3 under pressure suggests its likely synthesis through the reaction indicated by Eq. (1). We have performed HPHT experiments employing a laser-heated diamond anvil cell (DAC). CaO or Ca powder and liquefied O2 were loaded into an Re or Fe gasket hole, then compressed gradually to 35–40 GPa, and heated up to a temperature of approximately 3100 K with an off-line laser heating technique (see “Methods”). The Raman spectra were collected during the laser heating as shown in Fig. 2a, where new features are presented in the Raman spectra of the laser heated CaO or Ca and O2, unlike those of the pure solid oxygen appeared within the sample chamber22. Distinct new Raman modes are observed at about 767 and 1140 cm−1, suggesting the formation of new phases. The low-frequency modes appeared at about 200 cm−1 also indicate the presence of potential new bonding structure distinct from the starting CaO and O2. The calculated Raman spectrum of CaO3 at 38 GPa is found to be close to those modes observed in the experiment. In particular, the frequency mode near 767 cm−1 observed in experiment can be assigned to the vibration mode of O32−, whereas the high-frequency mode of 1140 cm−1 is attributed to the vibration mode of intramolecular O–O in CaO4, whose bond length of 1.31 Å is within the length range (1.3–1.4 Å) of the superoxide O2.

The powder X-ray diffraction (PXRD) patterns around the heating spot were collected are shown in Fig. 2b. We observe two distinct Bragg peaks at 10.6° and 11.4° and several small peaks from the raw 2D diffraction images and integrated PXRD patterns that do not correspond to CaO, CaO2, or any known calcium oxides. Meanwhile, the measured XRD pattern can be indexed by the predicted tetragonal BaS3-type structure of CaO3, together with CaO4, unreacted CaO, and oxygen, due to the mixed feature of obtained phases. The observed peaks at 10.6° and 11.4° in the XRD pattern correspond to the (200) and (111) crystal planes of tetragonal CaO3. The obtained lattice parameters of the synthesized CaO3, a = 4.67 Å, and c = 2.92 Å, are close to theoretical data of a = 4.87 Å and c = 2.98 Å. A decompression run was performed to assess volume change versus pressure, and the resulting pressure–volume data (Supplementary Fig. 6) are fitted by equation of state with B0 = 103(9) GPa, B0ʹ = 3.9 for CaO and B0 = 114(11) GPa, B0ʹ = 2.7 for CaO3, in good agreement with theoretical B0 values, 113.6 GPa, B0ʹ = 4.0 for CaO, and 99.8 GPa, B0ʹ = 4.0 for CaO3, which are obtained by fitting calculated total energies versus volume to the Birch–Murnaghan equation26. The CaO3 signals persist to at least 20.0 GPa as shown in Supplementary Fig. 7a.

### Electronic properties

To decipher the nature of bonding and charge states in CaO3, we have examined electron localization function27 in this compound. The results (Fig. 3a, upper panel) show clear covalent O–O bonding evidenced by the strong charge localization between the nearest-neighbor O atoms in the O3 units. Meanwhile, a less localized charge distribution is seen on the asymmetric Ca–O bonds (Fig. 3a, lower panel), indicating a significant degree of ionicity between the O3 anions and Ca cations. From a Bader charge analysis28, charge values on Ca and O are calculated at 30 GPa, and the results are listed in Table 1. There is a charge transfer of 1.51e from Ca to O3 unit in CaO3, comparable to those in CaO (1.43e) and CaO2 (1.48e), but much greater than that in KO3 (0.86e). This result highlights a crucial distinction of the O3 unit in CaO3 compared to the [O3]−1 anion in KO3. It is also seen that the two terminal O atoms carry more negative partial charges (0.65e per O) than the bridge O atom (0.21e) within each O3 anion in CaO3, because the central O interacts much less with Ca cations than with the terminal O anions. To the best of our knowledge, the presently identified [O3]−2 ionic charge state has never been seen in other ozonides17.

We examine MO schemes29,30 to elucidate electronic configurations in O3 and O3−2. Results (Fig. 3b) show that electrons in the antibonding 2b1 orbital dictate properties of O3 anions. Alkali-metal ozonides29 containing [O3] belong to a small group of chemical species hosting unpaired p-electrons that produce a paramagnetic state. In stark contrast, divalent ozonide anion [O3]−2 has a closed-shell configuration (Fig. 3b) with each O3 unit containing 20 electrons in 10 orbitals with no unpaired p-electron, leading to non-magnetic characteristics (Fig. 3c). To illustrate this point, we have constructed a model system of hypothetical Ca0O3 where all Ca atoms were removed from the BaS3-type structure, and this model system exhibits partially unoccupied bonding states of the O 2p orbital (Fig. 3d), which become fully occupied once Ca was incorporated into the crystal lattice due to charge transfer from Ca to O, leading to the non-magnetic insulating state in CaO3.

### Implications for geoscience

The reactivity of CaO and O2 is strongly driven by the denser structural packing of CaO3. Our calculations show that the reaction indicated in Eq. (1) at 30 GPa supports volumes of 23.48 Å3 for CaO in Fm-3m structure, 15.86 Å3 for O2 in Cmcm structure, and 36.20 Å3 for CaO3, with a large volume shrinkage of ∆V/V = −7.96%. Consequently, the PV term in Gibbs free energy strongly favors the formation of CaO3 at high pressures. For comparison, we also have explored the possibility of forming MgO3 at pressures up to 50 GPa, but the associated positive formation enthalpy (1.45 eV/f.u.) and smaller volume shrinkage of −3.76% render MgO3 unstable against decomposition into MgO and solid O2 in C2/m symmetry, in agreement with previous reports31.

The newly discovered calcium ozonide is expected to have major implications for geoscience. In this context, we have examined additional viable routes producing CaO3 involving several minerals abundant in Earth’s mantle as reactants:

$${mathrm{Ca}}left( {{mathrm{OH}}} right)_2 + {mathrm{O}}_2 = {mathrm{H}}_2{mathrm{O}} + {mathrm{CaO}}_{mathrm{3}},$$

(2)

$${mathrm{Ca}}left( {{mathrm{OH}}} right)_2 + {mathrm{Al}}_2{mathrm{O}}_3 + {mathrm{O}}_2 = 2{mathrm{AlOOH}} + {mathrm{CaO}}_3,$$

(3)

$$4{mathrm{FeO}}_2 + {mathrm{CaO}} + {mathrm{H}}_2{mathrm{O}} = 4{mathrm{FeOOH}} + {mathrm{CaO}}_3$$

(4)

with the structures of pertinent materials employed in calculating the reaction enthalpies are presented in Supplementary Table 1. Similar to the reaction shown in Eq. (1), the reactions in Eqs. (2) and (3) occur in oxygen-saturated environments and the reactants and products attain equilibrium at pressures of 20 GPa (Fig. 4a) and 40 GPa (Fig. 4b), respectively, corresponding to conditions near the top of the lower mantle, where previous studies reveal that oxygen fugacity is likely inhomogeneous with some regions containing relatively high oxygen content21, thus conducive to these reactions in forming CaO3.

Our calculations show that the reactions described in Eqs. (1) and (2) produce CaO3 at ~20 GPa, which corresponds to pressures at the boundary of Earth’s upper and lower mantle. Previous studies revealed that several minerals such as CaCO3 (ref. 32), MgCO3 (ref. 33), and CO2 (ref. 34) can dissociate and produce oxygen in this pressure range, offering an abundant source of O2 for these proposed reactions inside Earth’s mantle. The resulting compound CaO3, which was not previously considered, provides an alternative mechanism to explain seismic anomalies near 660 km depth in Earth’s mantle where pressure is ~20 GPa35,36.

The reaction route indicated in Eq. (4) describes the formation of CaO3 and FeOOH by FeO2 and CaO under H2O-saturated conditions and the equilibrium pressure of this reaction is about 90 GPa (Fig. 4c), corresponding to deep lower mantle conditions. Recently, a pyrite FeO2 phase stabilized at high pressure (76 GPa) and temperature (1800 K) was proposed8 to exist in Earth’s lower mantle below 1800 km. Our calculations show that once CaO and H2O are thrusted to deeper than 1800 km, they can react with FeO2 and produce CaO3 + FeOOH. Due to the higher density of FeOOH compared to that of the mantle37 (Fig. 4d), FeOOH would sink towards the core38, while the lighter CaO3 would ascend by mantle dynamic processes. Once reaching the transition zone at depths of less than 500 km, CaO3 would decompose to provide a sporadic source of extra O2 that would work its way up toward the surface of Earth to complete the oxygen cycle.

Oxygen fugacity and oxidation states of minerals in geological environments play pivotal roles in deciding planetary chemical and physical dynamics, and such key information can be determined through mineral equilibria39. Quantification of oxygen fugacity depends sensitively on the content and stability of mineral assemblages at the pressures and temperatures in Earth’s interior. Our discovery of divalent ozonide CaO3 introduces a new ingredient to buffer oxygen fugacity and influence redox equilibria of Earth’s mantle, providing crucial insights into the redox state of the largely inaccessible deeper mantle. Furthermore, our results highlight CaO as a reducing agent to react with free oxygen at high pressures, suggesting a natural reservoir for O2 storage in Earth’s mantle and providing a possible resolution to the missing O2 paradox before the Great Oxidation Event40. The present findings also raise exciting prospects of synthesizing CaO3 via additional avenues, such as those listed in Eqs. (2)–(4), in the laboratory setting for a more in-depth understanding of these reactions and their roles in influencing important geological events. The discovery of crystalline divalent calcium ozonide is expected to stimulate further experimental and theoretical exploration for further insights into this compound and the associated intriguing bonding characters that hold great promise for probing exotic properties that have great fundamental significance and implications for practical processes in chemistry and geoscience.

We have conducted a joint computational and experimental exploration of calcium oxides at high pressure, aiming to probe unusual stoichiometry, structural form, and oxidation states. Our study leads to a discovery of CaO3, expanding both the calcium oxide family and ionic ozonide family of compounds. This rare crystalline ozonide is computationally predicted and then experimentally synthesized via reaction of solid CaO and O2 at HPHT conditions in a DAC assisted by laser heating. Remarkably, a charge analysis indicates that the O3 unit in CaO3 carries a formal oxidation state of −2. These findings enrich fundamental understanding of bonding interactions between calcium and oxygen, highlighting novel ozonide chemistry at high pressure, and the reported results have major implications for elucidating prominent seismic anomalies and oxygen cycle processes in Earth’s mantle.