Iron oxides have great importance for all natural sciences as well as numerous industrial applications. Considering the high abundance of oxygen and iron in the Earth’s crust and the mantle, binary iron oxides and their derivatives are important endmembers of phases that make a significant contribution to properties of the Earth. Many studies have been devoted to investigations of various properties of iron oxides at conditions relevant to the Earth’s interior, i.e., at high pressure or at high pressure combined with high temperature (HP-HT). These studies reported a number of remarkable findings for the three basic iron oxides, Fe1-xO wüstite1,2, α-Fe2O3 hematite3,4,5,6,7,8, and Fe3O4 magnetite9,10,11. Meanwhile, several experimental and theoretical studies indicated that the chemistry of iron oxides at extreme conditions of high pressure and high temperature may extend to intriguing behavior beyond these three well-known oxides12,13,14,15. For instance, recent experimental HP-HT investigations in the pressure range of 10–20 GPa reported two new orthorhombic Cmcm polymorphs with Fe4O513 and Fe5O614 stoichiometry, i.e., between Fe3O4 and Fe1-xO, and hence, with mixed Fe2+ and Fe3+ oxidation states. More interestingly, using modern theoretical methods an opposite tendency in the stoichiometry changes in Fe-O systems was found, suggesting an extended stability of ferric hematite and a gradual shift to Fe4+–bearing oxides at megabar pressures15. In particular, these studies predicted the existence of a new exotic FeO2 oxide with highly charged Fe4+ ions that should have an extended range of stability against decomposition under megabar pressures15. Recently, a number of experimental studies have shown the important role of the iron oxides under HP-HT conditions in the Earth’s deep interior16,17,18,19. Decomposition reactions of iron oxides and oxyhydroxides may induce the release of oxygen or hydrogen in the Earth’s lower mantle17,18.

In addition to their obvious and primary importance for geosciences, iron oxides play crucial roles in many technological processes and applications, and remain among a handful of key materials with significant impact on the fundamental behavior of materials, including charge carrier transfer and interactions, spin dynamics of electrons, and other central topics. In other words, iron oxides are important prototype materials. For instance, the oldest-known natural magnet, magnetite, demonstrates a fascinating ‘metal-insulator’-type transition near 120–125 K20, named later as the ‘Verwey transition’ after its discovery in 1939. This Verwey transition was believed to be related to charge ordering on octahedral sites in the spinel structure, and has been hotly debated for decades. Only recently it was revealed that this transition is linked to formation of hitherto unknown ‘quasiparticles’ consisting of three bonded Fe ions called ‘trimerons’21. In hematite another landmark transition was discovered near 255 K with a related abrupt and drastic reorientation of spins of Fe3+ ions, named afterwards as a ‘spin-flop’ or Morin transition22. Also iron-deficient wüstite, Fe1-xO, displays puzzling complexities with regard to stoichiometry, defect structure, and elastic and physical properties23,24,25,26,27,28, and serves as a prototype for systems with non-stoichiometry. Recently it was experimentally demonstrated that, similar to magnetite, the newly-discovered Fe4O513 also undergoes an unprecedented ‘metal-insulator’–type transition upon cooling below 150 K with competing dimeric and trimeric ordering in the Fe chains, leading to strong structural modulations29. Therefore, synthesis of new iron oxides with both mixed Fe2+ and Fe3+ valences and bearing highly charged Fe4+ ions is of significant interest for many scientific fields.

In the present work we investigated the phase stability of iron oxides at HP-HT conditions (see Methods) and discovered in samples recovered at ambient conditions the presence of crystals of a new iron oxide with unusual Fe7O9 stoichiometry. Fe7O9 with its ratio of Fe/O ~ 0.777 lies between Fe3O4 and newly-discovered Fe4O513, but in contrast to both, it is non-magnetic at ambient conditions and adopts a monoclinic crystal structure with four sites for Fe cations. These observations combined with the difference in Fe3+/Fe2+ ratios (4/3 in Fe7O9 versus 2 in Fe3O4 and 1 in Fe4O5) suggest that the physical properties of Fe7O9 may be remarkably different. In addition, we also synthesized Fe7O9 containing a significant amount of Mg and discuss possible geological implications for this new polymorph.

Results and Discussion

Single crystal XRD measurements

The single crystals of pure Fe7O9 and Mg-doped (Mg,Fe2+)3Fe3+4O9 had sizes less than 50 μm in their linear dimensions that restricted detailed investigations of their properties. The chemical composition of the samples was determined using conventional electron microprobe methods and from single crystal X-ray diffraction data. We also collected Mössbauer spectra from these samples to determine the oxidation states of the Fe ions.

By means of single crystal X-ray diffraction on the crystals (Fig. 1 and Supplementary Figure 1) we solved and refined the crystal structures of Fe7O9 and (Mg,Fe2+)3Fe3+4O9 at ambient conditions and confirmed their stoichiometry. Crystals of Fe7O9 and (Mg, Fe2+)3Fe3+4O9 had no pronounced asymmetry in their shape, but were rather small (0.03 × 0.02 × 0.01 and 0.03 × 0.02 × 0.02 mm3), so it was not possible to perform an analytical absorption correction based on crystal shape. Crystals of Fe7O9 appeared to be twinned, and due to the high degree of peak overlap (>50%), we had to integrate both twin domains simultaneously to perform a refinement of the crystal structure against HKLF5 data (BASF value was about 49.8%). The twinning could also influence the quality of determination of the anisotropic parameters. For consistency, we also refined the structure in an isotropic approximation and demonstrated a negligible influence on the atomic positions (CIF-files in the Supplementary Materials). Technical details of the structure determinations are summarized in Tables S1 and S2 in the Supplementary Materials. We established that both compounds adopt the same monoclinic structure of the C2/m space group. The unit cell parameters in Fe7O9 are as follows: a = 9.696(2) Å, b = 2.8947(6) Å, c = 11.428(3) Å, β = 101.69(2)°, V = 314.10(12) Å3, and Z = 2 (Fig. 2, and Tables S1 and S2 and CIF-files in the Supplementary Materials). The crystal structure of Fe7O9 has four different crystallographic sites for cations, three (Fe1, Fe2, Fe3) are octahedrally–coordinated and connected in a 3D network, while the fourth, Fe4, has a trigonal-prismatic arrangement (Fig. 2 and Table S2 in the Supplementary Materials). By analyzing the Fe-O bond distances in this polymorph using a bond valence sums (BVS) method30, we established the average oxidation state for Fe ions occupying the Fe1, Fe2, Fe3, and Fe4 sites to be +2.74, +2.72, +2.82, and +2.10, respectively. Thus, we conclude that the Fe4 sites are occupied almost exclusively by Fe2+ ions (Fig. 2). The other octahedral sites participate in electronic exchange between Fe2+ and Fe3+ ions through the polaron hopping mechanism, similar to the octahedral network of magnetite20 and other Fe-bearing oxides31. Thus, the BVS method shows the average charge for each of the Fe1-Fe3 sites. We note that various considerations lead to octahedral sites in the recently-discovered Fe4O5 phase having different charges29.

Figure 1
figure 1

X-ray diffraction images.

(a) Example of X-ray diffraction image of microscopic single crystal of (Mg,Fe)3Fe4O9 collected at ambient conditions under rotation of a sample over 360 degrees in a beam. This image contains several hundreds of small well-resolved reflections. Insets show selected magnified areas of this image that better show indistinct reflections. (b,c) Projections of these X-ray diffraction data in reciprocal space, for two selected planes, 1kl (b) and h1l (c).

Figure 2
figure 2

The crystal structure of Fe7O9.

This structure corresponds to ambient conditions and is shown in a projection along the b axis.

Mg-doped Fe7O9 crystals adopt the same crystal structure as Fe7O9 with similar unit cell parameters: a = 9.6901(12) Å, b = 2.8943(5) Å, c = 11.4397(15) Å, β = 102.045(14)°, V = 313.77(8) Å3, and Z = 2 (Fig. 2 and Table S1 in the Supplementary Materials). Electron microprobe analysis established their chemical composition to be Mg1.06Fe5.94O9, i.e., nearly 15% of Fe ions are substituted by Mg. We used this chemical composition in the crystal structure refinement and found that Mg2+ ions occupy all four Fe sites, but with a noticeable preference for (i) the spacious Fe4 sites that are occupied by the larger Fe2+ ions in Fe7O9, and (ii) the Fe1 sites, located between the Fe4 sites (Fig. 2, Table S2 in the Supplementary Materials). In the case of Mg doping of Fe4O5, Mg ions were also found to occupy all Fe sites in the structure32. Repeating the same BVS analysis30 as above for Fe7O9 taking into account the Mg atom distribution determined by single crystal X-ray diffraction (Table S2 in the Supplementary Materials), we confirmed the ferrous nature of Fe4 ions and detected a minor increase in the BVS of all other octahedrally coordinated Fe ions as +2.81, +2.78, and +2.87 for the Fe1, Fe2, and Fe3 sites, respectively (Fig. 2). Thus, the results of BVS analysis suggest the persistence of charge transfer in the 3D octahedral network with the distributed Mg ions. However, these Mg impurities should dramatically lower the mobility of hopping polarons, and hence, the bulk electrical conductivity of Mg-doped Fe7O9 is expected to be much lower than that of Fe7O9.

Mössbauer spectroscopy

Both Fe7O9 and (Mg,Fe)7O9 samples were analyzed by Mössbauer spectroscopy using a synchrotron Mössbauer source that gave excellent signal to noise ratios (Fig. 3 and Table S3 in the Supplementary Materials). We did not observe any magnetic component in these spectra, and hence conclude that these materials are non-magnetic at ambient conditions. The spectrum of Fe7O9 (Fig. 3a) could be fitted by a superposition of two basic components, including (i) Fe2+ ions at the prismatically-coordinated Fe4 sites in the crystal structure (Fig. 2) and (ii) a merged component related to octahedrally-coordinated Fe1, Fe2, and Fe3 ions with an average oxidation state of Fe2.8+ (Fig. 3a and Table S3 in the Supplementary Materials). We note that this Fe2.8+ component is an average because of the above-mentioned rapid charge exchange between Fe2+ and Fe3+ ions at the octahedral sites, similarly to Fe3O433. This finding is in excellent agreement with the above BVS oxidation states of the Fe ions obtained from the single crystal XRD data. It should be noted here that the three slightly structurally-inequivalent Fe1, Fe2 and Fe3 sites (Fig. 2) give similar contributions to the Mössbauer spectrum because of the very similar environment (edge-sharing FeO6 polyhedron) of the Fe ions, so it is not possible to separate their individual components in the spectrum (Fig. 3a). Spectra obtained from (Mg,Fe)7O9 were quite similar to Fe7O9 (Fig. 3b), while the merged component in the Mg-bearing sample spectrum shows a slightly higher average oxidation state of Fe2.9+ due to incorporation of Mg.

Figure 3
figure 3

Mössbauer spectra of Fe7O9 samples.

(a) Single crystal of Fe7O9. (b) Single crystal of (Mg,Fe)3Fe4O9. Both spectra were collected at ambient conditions. Black open circles, experimental spectrum; lines, fitted spectra; broken line, residual.


In our work we synthesized Fe7O9 and (Mg,Fe2+)3Fe3+4O9 crystals at high pressures around 24–26 GPa. In previous studies the orthorhombic polymorphs of Fe4O513 and Fe5O614 were prepared at substantially lower pressures, between 10 and 20 GPa. It is interesting to note that the first pressure-driven structural phase transitions in the known iron oxides were detected at similar pressures around 20–25 GPa. For instance, at room temperature cubic Fe1-xO wüstite with the rocksalt structure transforms to a rhombohedral lattice above 20 GPa1, while at high temperatures the rocksalt structure of Fe1-xO is stable to at least 60 GPa1,2. High temperature-assisted phase transitions in Fe3O4 from cubic spinel to an orthorhombic phase34,35, and from corundum-type α-Fe2O3 to an orthorhombic Rh2O3(II)-type or to other phases4,5,6,7,8 were also observed in some studies already above 20–25 GPa, although these phase transitions are still hotly debated. These observations suggest that all conventional iron oxides (α-Fe2O3, Fe3O4, and Fe1-xO) become unstable with respect to structural transformations in approximately similar pressure ranges that might be related to similar shortening of Fe-O bond lengths in their structures. Hence, the resultant high-pressure polymorph of a compressed and heated iron oxide could depend on its stoichiometry, thereby suggesting chemical tuning as a route to new structural phases. In general, one could expect that a minor tuning in stoichiometry could lead either to structures with vacancies (ordered or disordered) or to modified, Fe3O4-like or Fe2O3-like high-pressure phases in new oxides. Likewise, significant shifts from known stoichiometry could potentially lead to hitherto unknown structures. For instance, the newly-discovered orthorhombic Cmcm polymorphs of Fe4O513 and Fe5O614 crystalize in structures that are linked to the high-pressure orthorhombic polymorph of Fe3O434,35 (Fig. 4). By analogy with the existing family of calcium ferrites, 36, it was proposed that iron oxides with this Cmcm structure could also form such a family as 37, which includes Fe3O4, Fe4O513 and Fe5O614. However, the present discovery of Fe7O9 that does not belong to this family on the one hand, but having a certain structural similarity with the above oxides on the other hand (Fig. 4), suggests that the family of iron oxides that are structurally linked to the high-pressure polymorph of Fe3O4 may be more broad, e.g., like , thereby suggesting a potentially greater diversity than in the calcium ferrite oxides. For instance, Fe9O11 (n = 2) or Fe11O13 (n = 3) might be hypothetically stable under certain HP-HT conditions.

Figure 4
figure 4

Comparison of unit cells of crystal structures of iron oxides.

(a) High-pressure orthorhombic polymorph of Fe3O434, (b) Monoclinic Fe7O9 polymorph discovered in the present work. (c) Orthorhombic Cmcm polymorph of Fe4O5 discovered in ref. 13. (d) Orthorhombic Cmcm polymorph of Fe5O6 discovered in ref. 14. Different colors of the octahedra denote different crystallographic sites for Fe ions.

At the moment the stability field of this new Fe7O9 polymorph is not well defined, although it appears to lie at pressures higher than those of Fe4O513 and Fe5O614. The chemical compositions and the structural phases of iron-magnesium oxides in the Earth’s mantle remain a disputed issue that requires in situ investigations at HP-HT conditions under different oxygen fugacities. In this regard, the unexpected discovery of a Fe7O9 polymorph provides new insight into possible compositions of mantle phases, and provides a new type compound that may play a key role in determining physical and chemical properties.

The new oxide, Fe7O9, is a compound with a ratio of Fe3+/Fe2+ intermediate between those of Fe3O4 and Fe4O5 (Fig. 4). Both Fe3O4 and Fe4O5 are model systems for investigations of Fe2+–Fe3+ interactions in solids, demonstrating enigmatic low-temperature phase transitions of ‘metal-insulator’-type that lead to the formation of exotic ‘trimeron quasiparticles’ in Fe3O421 or to even more intricate ordering patterns in Fe4O529. We note that the low-temperature Verwey transition in Fe3O4 has had a strong impact on solid state physics and chemistry for decades. Thus, Fe7O9 presents an exciting compound that promises important implications for geosciences, solid state physics and chemistry with potential for industrial applications.

The recently discovered iron oxides may play important roles in the cycling of volatiles in the Earth’s deep interior16,17,18,19. For instance, oxygen could be released in the deeper part of the lower mantle via decomposition reactions of Fe2O3 and Fe3O4 into Fe5O7 and Fe25O32 above 60 GPa17. Further, it was recently reported that FeO2 could be formed in the lower mantle as a product of FeOOH goethite decomposition18. Moreover, this reaction could supply hydrogen to the surrounding mantle18. Physical and chemical properties of the newly-discovered iron oxides can therefore provide novel insights into the chemical evolution of the Earth’s interior.

Additional Information

How to cite this article: Sinmyo, R. et al. Discovery of Fe7O9: a new iron oxide with a complex monoclinic structure. Sci. Rep. 6, 32852; doi: 10.1038/srep32852 (2016).