Introduction

The structures, properties and high-pressure behavior of corundum-type oxides have been extensively investigated because of their wide variety of elastic, electrical and magnetic properties and importance in earth sciences and technology1,2,3. High-pressure studies of hematite, α-Fe2O3 (Fig. 1a), have attracted special attention due to their geophysical interest and the unclear role of Fe3+ in the nature and dynamics of the earth’s lower mantle1,2. Particular attention has been focused on elucidating the nature of phase transition(s) and the structure of the high-pressure phase of hematite observed above 50 GPa (refs. 1,4,5,6,7,8,9,10,11,12). For this phase two structures have been proposed by different groups: Rh2O3-II-type (space group Pbcn, no. #60) and GdFeO3-perovskite-type (space group Pbnm, no. #62) structures4,7. While Mössbauer spectroscopic and resistivity measurements clearly demonstrate the importance of electronic changes in Fe3+ and seem to support the Rh2O3-II-type structure5, powder diffraction data collected by various groups over several decades did not allow an unambiguous assignment of the structural type (see refs 4, 5, 7, 8 and references therein). Only recent single-crystal high-P,T diffraction data12 were able to solve this challenge; they demonstrated that the Rh2O3-II-type phase of Fe2O3 (which we refer to below as , Fig. 1b) forms upon laser heating at pressures above 40 GPa; whereas, compression of hematite at ambient temperature to over 50 GPa results in the formation of a phase with distorted GdFeO3-perovskite-type, dPv ζ-Fe2O3, structure (Fig. 1c). Experiments in laser-heated diamond anvil cells (DACs) revealed the formation of a CaIrO3-type phase (‘post-perovskite’ (PPv) η-Fe2O3, Fig. 1d) at pressures above 60 GPa (refs. 1,9,12,13). However, the behaviour of this phase under compression is not well studied. The phase diagram of Fe2O3 in the megabar pressure range is incomplete and the data are often conflicting1,9,10,13.

Figure 1: Crystal structures of iron oxide phases studied in the present work.
figure 1

Building blocks are octahedra (brown) and trigonal prisms (blue). The prisms in Fe5O7, Fe25O32 and η-Fe2O3 have one or two additional apices. Hematite (a) consists of FeO6 octahedra connected in a corundum-like motif, namely each octahedron connects with three neighbours via edges in honeycomb layers, and layers are interconnected through common triangular faces of octahedra. The structure (b) is built of only FeO6 octahedra but each two octahedra are connected through a common triangular face; such units pack in a herringbone pattern and layers pack with a shift along the c-direction having common edges. In distorted perovskite ζ-Fe2O3 (c) octahedra connect through common vertices and prisms share only common edges. θ-Fe2O3 (e) adopts the packing motif from but instead of octahedra it consists of FeO6 prisms. Post-perovskite (d) and Fe5O7 (f) are members of the homologous series nFeO·mFe2O3 (see also Fig. 4), where prisms are connected through common triangular faces, while octahedra connect only via shared edges. In addition to triangular face-shared prisms and edge-shared octahedra, Fe25O32 (g) has edge-shared one-capped prisms; therefore it belongs neither to the homologous series nor adopts any other known structural motif.

In order to study the high-pressure high-temperature (HPHT) behaviour of ferric iron (Fe3+) oxide we apply the complementary methods of single crystal X-ray diffraction in laser-heated DACs and synchrotron Mössbauer source (SMS) spectroscopy (see Methods section). We observe hitherto unknown Fe–O phases, show the results of their structure solution and refinement, and characterize the pressure–temperature conditions, at which different Fe2O3 polymorphs occur. Crystal chemical analysis of the new structures and known Fe(II, III) oxides reveals their fundamental relationships as members of the homologous series nFeO·mFe2O3. We observe that at pressures above 60 GPa and at high temperatures (that is, at conditions of the earth’s lower mantle), Fe2O3 decomposes with release of oxygen. The same phenomenon is observed for Fe3O4. Our results indicate that mixed-valence iron oxides may play a significant role in oxygen cycling between the earth’s atmosphere and mantle.

Results

Structural transformations in Fe2O3

In agreement with previous studies4,5,7,8,12, our cold compression experiments on hematite single crystals to 54(1) GPa result in a transition to the ζ-Fe2O3 phase manifested by a 8.4% volume discontinuity (Supplementary Fig. 1). Although in earlier work we indexed the diffraction pattern of ζ-Fe2O3 in a monoclinic unit cell12, the new extended dataset acquired in the present work showed that the structure is in fact triclinic (see Supplementary Note 1 for details), similar to Mn2O3 (ref. 14). An insufficient number of independent reflections prevented structural refinement of ζ-Fe2O3 in triclinic symmetry, so we used a monoclinic model12 to qualitatively constrain the atomic arrangement in ζ-Fe2O3. Upon further pressure increase from 54(1) to 67(1) GPa, we observed a reduction in the splitting of reflections, indicating an increase in symmetry. The structure of ζ-Fe2O3 thus becomes closer to that of GdFeO3-type-perovskite (Supplementary Fig. 2b). At 67(1) GPa a small drop in the unit cell volume (1.7%) manifests the next transformation to the θ-Fe2O3 phase (Fig. 1e) with orthorhombic symmetry (space group Aba2, no. #41, a=4.608(7), b=4.730(4), c=6.682(18) Å (Supplementary Table 1)). On compression at ambient temperature θ-Fe2O3 can be observed to at least 100 GPa (Supplementary Fig. 1). The transformational PT diagram for Fe2O3 is given in Fig. 2.

Figure 2: Transformational phase diagram of Fe2O3.
figure 2

(a) - hematite (α-Fe2O3), Δ- distorted perovskite (ζ-Fe2O3), -Aba2 (θ-Fe2O3, probably metastable), -Cmcm post-perovskite (η-Fe2O3) and × -Rh2O3-II type phase (). The boundary between hematite α-Fe2O3 and is defined according to ref. 10. The geotherm is defined according to refs 39, 40.

During in situ laser heating of θ-Fe2O3 between 1,000 and 1,550(50) K at 78(2) GPa, we observed no evidence of a phase transformation. The absence of transformations may either be evidence that θ-Fe2O3 is stable at these conditions or an indication that higher temperatures are required to overcome kinetic barriers to further structural transitions. Indeed, heating at 1,600(50) K results in the formation of post-perovskite type η-Fe2O3 coexisting with θ-Fe2O3. Both phases (θ-Fe2O3 and η-Fe2O3) were observed in situ simultaneously upon heating to 1,850(50) K at pressures up to 113(1) GPa. However, temperature-quenched products contained only η-Fe2O3 (Fig. 2). Once synthesized, η-Fe2O3 may be preserved at ambient temperature down to at least 26 GPa. At lower pressures it transforms back to hematite (see Figs 1 and 2 for structures and phase relations). Moderate heating to 2,000 K at pressures of about 50 GPa provokes a transition to the dPv ζ-Fe2O3 phase. Decompression of ζ-Fe2O3 or η-Fe2O3 to 41(1) GPa with heating at 1,800(100) K results in growth of Rh2O3-II type (Supplementary Fig. 1, Supplementary Table 1). Interestingly, was synthesized earlier10,11 from hematite, thus bracketing the possible P–T stability field of the phase (Fig. 2).

Electronic transformations in Fe2O3

The sequence of phase transitions in Fe2O3 in the megabar pressure range and temperatures up to about 2,500 K (Fig. 2) can be neatly rationalized through the variation of molar volumes of the phases observed as a function of pressure (Supplementary Fig. 1), complemented by the corresponding SMS spectroscopy data (Fig. 3). The bulk modulus of hematite, 219(7) GPa, is in good agreement with previous studies15 and at 67 GPa it reaches 392(10) GPa, whereas the bulk modulus of ζ-Fe2O3 at 54 GPa is substantially lower, 320(18) GPa. Such a large drop of bulk modulus (18%) associated with a large reduction of molar volume (8.4%) is very unusual and is likely caused by changes in the electronic state of Fe3+. The Mössbauer spectrum of ζ-Fe2O3 collected immediately after the transition at 50 GPa shows two components (Fig. 3b), a magnetic sextet having centre shift (CS) of 0.424(7) mm s−1 corresponding to the high-spin (HS) state of Fe3+, and a doublet (CS=0.074(5) mm s−1) with hyperfine parameters characteristic for low-spin (LS) Fe3+ in an octahedral oxygen environment16. The relative abundance of the components is 1:1, as expected for the perovskite-type structure of ζ-Fe2O3 with HS-Fe3+ located in large bipolar prisms and LS-Fe3+ in smaller octahedra (Fig. 1c). Upon further compression of ζ-Fe2O3 the amount of HS-Fe3+ decreases (Fig. 3c), which explains the anomalously high compressibility of this phase.

Figure 3: Evolution of SMS spectra of Fe2O3.
figure 3

Spectra collected during compression (ad) and after heating (e). In hematite (a) iron atoms have a HS state (at 24 GPa CS=0.306(4) mm s-1), and spectra are split due to magnetic ordering (M). After the first transition at 49 GPa (b) a new non-magnetic (NM) component appears with CS of 0.074(5) mm s-1 corresponding to a LS state. During further compression a fraction of the magnetic component decreases (c) and it disappears completely after the second transition to the θ-Fe2O3 phase (d) that has only one non-magnetic position of LS iron atoms in the crystal structure. After heating above 1,600(50) K (e) a transformation to η-Fe2O3 occurs. The crystal structure has two HS-iron positions (both CS are 0.45 mm s-1), where one position is magnetically ordered and the other is non-magnetic.

Transformation to θ-Fe2O3 is associated with a small decrease of molar volume (1.7%) and an increase of bulk modulus as expected (418(11) GPa for θ-Fe2O3 at 67 GPa compared with 371(20) GPa for ζ-Fe2O3 at 70 GPa) (Supplementary Fig. 1). The Mössbauer spectrum of θ-Fe2O3 (Fig. 3d) shows that all Fe3+ is in the LS state and there is only one type of iron atom in the crystal structure in accordance with the single crystal X-ray diffraction data (Fig. 1e).

Heating of θ-Fe2O3 above 1,600 K at pressures above 70 GPa resulted in partial or complete transformation into CaIrO3-PPv-type η-Fe2O3 (Fig. 2). The Mössbauer spectrum of pure η-Fe2O3 at 91(2) GPa (Fig. 3e) contains two components (a magnetically ordered sextet and a paramagnetic doublet) with equal abundances and almost equal CS (0.45 mm/s) corresponding to HS-Fe3+. Within the accuracy of our X-ray diffraction data the molar volumes of θ-Fe2O3 and as-synthesized η-Fe2O3 are indistinguishable (Supplementary Fig. 1), suggesting that the atomic packing density increase in the CaIrO3-PPv-type η-Fe2O3 structure compensates the difference in ionic radii of HS and LS Fe3+ ions in the ζ-Fe2O3 structure. Note that Shim et al.1 also reported magnetic ordering in η-Fe2O3 based on nuclear forward scattering measurements. One of the magnetic sites described by the authors1 has hyperfine parameters close to those that we observed; however, the second non-magnetic component in the nuclear forward scattering spectra was not identified in ref. 1.

Thermal stability of Fe2O3

The behaviour of η-Fe2O3 under heating is rather remarkable. First, we noted that its unit cell volume increases by up to 1% upon laser heating to about 2,000 K at 56 and 64 GPa. (Supplementary Fig. 3). Second, after heating for a few seconds to 2,700–3,000 K and 71 GPa we observed the immediate appearance of new sharp spots in the diffraction pattern. The peaks were indexed in the C2/m space group and the structure solution using direct methods identified the phase as a novel mixed-valence iron oxide with stoichiometry Fe5O7 (FeO·2Fe2O3) (Fig. 1f, Supplementary Table 2). Visual observations (particularly preservation of the shape of the samples upon heating) and careful analysis of diffraction patterns (absence of diffuse scattering) verify that samples were not melted in experiments where Fe5O7 was synthesized. The phase is preserved on decompression down to at least 41(1) GPa. Thus, we explain our observations as a continuous loss of oxygen by η-Fe2O3 upon heating at moderate temperatures and pressures above 60 GPa according to the reaction η-Fe2O3→η-Fe2O3−δ+0.5δ·O2. Note that a similar process is well known for perovskites17 and other oxides18. The reaction is accompanied by a partial reduction of Fe3+ to larger-sized Fe2+ that consequently increases the unit cell volume. Upon heating at sufficiently high temperature (above 2,700 K), the oxygen deficiency in η-Fe2O3 reaches a critical limit and provokes a reconstructive phase transition resulting in the formation of the mixed-valence iron oxide Fe5O7: 5η-Fe2O3→2Fe5O7+0.5O2. From both X-ray diffraction and Raman spectroscopy we did not find any evidence to suggest involvement of carbon from the diamond anvils in the chemical reactions. Indeed this was not expected, because at the HPHT conditions of our experiments carbon and oxygen do not react19. Mössbauer experiments show that laser heating of Fe2O3 at pressures above 80 GPa leads to formation of phases containing iron with hyperfine parameters characteristic of a mixed valence state (Supplementary Fig. 4).

Similarities in the crystal structures of η-Fe2O3, Fe5O7, high-pressure polymorph of Fe3O4 (HP-Fe3O4, space group Bbmm, no. #63, Supplementary Tables 1 and 2), and the recently discovered Fe4O5 (ref. 20) and Fe5O6 (ref. 21) (Fig. 4) demonstrate22 that iron oxide phases form a homologous series nFeO·mFe2O3 (with wüstite, FeO and η-Fe2O3 as the end-members) and indicate that a mixed-valence state of iron may become crystal chemically important at high pressures and temperatures.

Figure 4: Homologous series of iron oxides described by the common formula nFeO·mFe2O3.
figure 4

The structures may be described as assembled from two building blocks, FeO6 octahedra and trigonal prisms (prisms could be two-capped but they are not shown for simplicity). Prisms connect to each other through triangular faces, while octahedra share edges, so that they form parallel columns of face-shared prisms and edge-shared octahedra arranged in different motifs as seen in the figure with structures viewed from the top of the columns. Increasing Fe2+ content favours octahedral packing over mixed octahedral and prismatic packing. This requires denser packing of FeO6 octahedra and as a result columns of octahedra condense in slabs by sharing common edges. In particular, η-Fe2O3 has ordinary columns of prisms and octahedra with a chequerboard-like arrangement; Fe5O7 has ordinary and doubled columns of octahedra; the HP-Fe3O4 (ref. 41) possesses only doubled columns; Fe7O9 (ICSD reference number CSD-430601)42 has doubled columns and tripled columns organized in zigzag slabs; Fe4O5 (ref. 20) possesses only tripled and Fe5O6 (ref. 21) only quadruple zigzag slabs. The end-member of the homologous series wüstite (FeO) consists of octahedra with a maximum (12) number of edge-shared neighbours.

Discussion

Our results demonstrate clearly the complex behaviour of iron oxide subjected to high pressures and temperatures and may have significant consequences for modelling of the earth’s interior. Hematite is one of the major components of banded iron formations (BIFs) and ironstones, and these huge sedimentary rock formations occurring on all continents may reach up to several hundred meters in thickness and hundreds of kilometres in length. Deposited in the world’s oceans, BIFs as part of the ocean floor are recycled into the earth’s interior by subduction2,23 to depths extending possibly to the core–mantle boundary region2. Available experimental data2,13,24 suggest that iron oxides melt above the geotherm in the entire mantle and thus remain solid in slabs that are colder than the surrounding mantle. Even assuming a slow subduction rate of 1 cm per year with slabs reaching a depth of about 2,000 km in 200 Ma, this geological time is sufficient to influence only a few tens of meters of rocks beneath the BIF’s surface. Thus the fate of iron oxides, a major component of subducted BIFs, depends on the pressures and temperatures (P–T), to which they are exposed. Upon subduction of BIFs into the lower mantle, hematite undergoes numerous phase transformations. At pressures above 60 GPa the HP phase η-Fe2O3 starts to decompose, producing oxygen. Moreover, experiments on Fe3O4, the second major component of BIFs, show that it also decomposes upon heating at pressures above 70 GPa, forming in particular the phase Fe25O32 (Fig. 1g, see also Supplementary Note 2). Based on estimates of the amount of BIFs subducted into the earth’s mantle2 and that BIFs may consist of 50% Fe2O3 by volume, the amount of oxygen produced by the formation of Fe5O7 alone can be as high as 8–10 times the mass of oxygen in the modern atmosphere. Extrapolation of available data25 indicates that oxygen would be in the liquid state at geotherm temperatures. Since the oxygen fugacity of the lower mantle is expected to be low through equilibrium with metallic iron, an oxygen-rich fluid could locally oxidize surrounding material (particularly Fe2+ in ferropericlase as well as bridgmanite, and metallic iron in a (Fe, Ni)-metal phase26). Seismic tomography reveals pronounced complex heterogeneities in the lower mantle at depths of 1,500–2,000 km associated with subducted slabs27,28,29 and the presence of oxidized material may be a reason for these observations30. On the other hand, a low oxygen chemical activity at high pressure19,31,32 could prevent the immediate reaction of oxygen in the lower mantle or even in the transition zone, and instead allow an oxygen-rich fluid to pass to the upper mantle, thus shifting Fe2+/Fe3+ equilibria in silicate minerals and greatly raising the oxygen fugacity in this region. In any case, our study suggests the presence of an oxygen-rich fluid in the deep earth’s interior that can significantly affect geochemical processes by changing oxidation states and mobilizing trace elements.

Methods

Sample preparation

Single crystals of α-Fe2O3 enriched with 57Fe (57Fe2O3) were grown by means of HPHT technique at 7 GPa and 800 °C in a 1,200-t Sumitomo press at Bayerisches Geoinstitut (Bayreuth, Germany). As a precursor, a 1:1 mixture of a powder of non-enriched hematite (α-Fe2O3) of 99.998% purity and a pure powder of 57Fe2O3 (96.64%-enriched) was used. Magnetite synthesis was performed in the same way at 9.5 GPa and 1,100 °C as described in ref. 33. Synthesis of non-enriched hematite single crystals was described in ref. 15.

Single crystals with an average size of 0.03 × 0.03 × 0.005 mm3 were pre-selected on a three-circle Bruker diffractometer equipped with a SMART APEX CCD detector and a high-brilliance Rigaku rotating anode (Rotor Flex FR-D, Mo-Kα radiation) with Osmic focusing X-ray optics.

Selected crystals together with small ruby chips (for pressure estimation) were loaded into BX90-type DACs34. Neon was used as a pressure transmitting medium loaded at Bayerisches Geoinstitut.

X-ray diffraction

The single-crystal X-ray diffraction experiments were conducted on the ID09A beamline at the European synchrotron radiation facility (ESRF), Grenoble, France (MAR555 detector, λ=0.4126–0.4130 Å); on the 13-IDD beamline at the advanced photon source (APS), Chicago, USA (MAR165 CCD detector, λ=0.3344 Å); and on the extreme conditions beamline P02.2 at PETRA III, Hamburg, Germany (PerkinElmer XRD1621 flat panel detector, λ=0.2898–0.2902 Å). The X-ray spot size depended on the beamline settings and varied from 4 to 30 μm, where typically a smaller beam was used for laser heating experiments. A portable double-sided laser heating35 system was used for experiments on ID09A (ESRF) to collect in situ single-crystal X-ray diffraction. State-of-the art stationary double-side laser-heating set-up at IDD-13 (APS) has been used for temperature-quenched single-crystal X-ray diffraction. Crystals were completely ‘surrounded’ by laser light and there were no measurable temperature gradients within the samples. In the case of prolonged heating experiments the temperature variation during the heating did not exceed ±100 K. Pressures were calculated from the positions of the X-ray diffraction lines of Ne (http://kantor.50webs.com/diffraction.htm). X-ray diffraction images were collected during continuous rotation of DACs typically from –40 to +40 on ω; while data collection experiments were performed by narrow 0.5–1° scanning of the same ω range. The crystallographic information is also available as Supplementary Data 1-7.

Data analysis

Integration of the reflection intensities and absorption corrections were performed using CrysAlisPro software36. The structures were solved by the direct method and refined in the isotropic approximation by full matrix least-squares refinements using SHELXS and SHELXL software37, respectively.

SMS spectroscopy

Energy-domain Mössbauer measurements were carried out at the nuclear resonance beamline ID18 at ESRF (see ref. 38 for more details).

Additional information

Accession codes: The X-ray crystallographic coordinates for structures reported in this article have been deposited at the Inorganic Crystal Structure Database (ICSD) under deposition number CSD 430557–430563. These data can be obtained free of charge from FIZ Karlsruhe, 76344 Eggenstein-Leopoldshafen, Germany (fax: (+49)7247-808-666; e-mail: crysdata@fiz-karlsruhe.de) through the hyperlink ‘ https://www.fiz-karlsruhe.de/en/leistungen/kristallographie/kristallstrukturdepot/order-form-request-for-deposited-data.html’.

How to cite this article: Bykova, E. et al. Structural complexity of simple Fe2O3 at high pressures and temperatures. Nat. Commun. 7:10661 doi: 10.1038/ncomms10661 (2016).