Hematite α‐Fe2O3(0001) in Top and Side View: Resolving Long‐Standing Controversies about Its Surface Structure

Hematite is a common iron oxide found in nature, and the α‐Fe2O3(0001) plane is prevalent on the nanomaterial utilized in photo‐ and electrocatalytic applications. The atomic‐scale structure of the surface remains controversial despite decades of study, partly because it depends on sample history as well as the preparation conditions. Here, a comprehensive study is performed using an arsenal of surface techniques (non‐contact atomic force microscopy, scanning tunneling microscopy, low‐energy electron diffraction, and X‐ray photoemission spectroscopy) complemented by analyses of the near surface region by high‐resolution transmission electron microscopy and electron energy loss spectroscopy. The results show that the so‐called “bi‐phase” termination forms even under highly oxidizing conditions; a (1 × 1) surface is only observed in the presence of impurities. Furthermore, it is shown that the biphase is actually a continuous layer distorted due to a mismatch with the subsurface layers, and thus not the proposed mixture of FeO(111) and α‐Fe2O3(0001) phases. Overall, the results show how combining surface and cross‐sectional imaging provides a full view that can be essential for understanding the role of the near‐surface region on oxide surface properties.


MAIN TEXT.
Iron oxides are extensively studied thanks to their plethora of electronic, magnetic, and catalytic properties conferred by the different crystal structures and stoichiometries of iron and oxygen in various compounds. 1Hematite (α-Fe2O3), the most abundant iron oxide on the Earth, has attracted much attention in green chemistry due to its large availability and low toxicity. 24][5] Its bandgap (1.9-2.2 eV) permits electron excitations by visible light, but the actual efficiency of the PWS on hematite is well below the predicted (and industrially required) 15% solar-to-hydrogen conversion. 6Sixty years of PWS research has led to the consensus that the reaction is promoted by electronic states formed at the interface between the oxide and the electrolyte. 7However, the origin, amount, and role of these states are still a matter of controversy. 8,9A comprehensive understanding of the surface atomic structure is crucial for the identification of active surface sites.Reaction-limiting factors such as light-depth penetration and charge-carrier trapping and recombination are inherently linked to the subsurface and bulk properties. 10,11In recent years, attention has been paid to the characterization of bulk phenomena such as polaron formation and electron-hole recombination [12][13][14] and the effect of the bulk morphology on the final reaction yield. 15,16Due to the importance of hematite in catalysis, the fundamental investigation of its redox behavior is the first step to designing improved industrial processes and chemical pathways.
The two leading models are the 'biphase reconstruction', which consists of alternating domains with FeO and Fe2O3 stoichiometries 23 , and the 'honeycomb model', an O-M-O-trilayer surface structure proposed as a possible polarity-compensation mechanism of corundum oxides. 24,25Above µo = −1.8eV, the bulk fully recovers its Fe2O3 stoichiometry, and the surface is thought to terminate in a bulk-like (1×1) structure. 26The O3-, O-H-Fe-, and half-metal Fe terminations have been proposed for the (1×1) surface phase of Fe2O3(0001). 27,28e coexistence of different bulk stoichiometries and surface phases on single crystals in ultrahigh vacuum (UHV) hampers obtaining well-defined samples, presenting thus a "minefield" 1 in basic research of model hematite systems.Mixed bulk and surface iron oxide phases can be misidentified due to similarities in their spectral signatures, such as the Fe 2p Fe 3+ satellite structures of Fe2O3 and γ-Fe2O3 in X-ray photoemission spectroscopy (XPS) 29 or their pre-peak feature at the Fe 2p L3 edge in X-ray absorption spectroscopy. 30Thus, a combination of realand reciprocal-space imaging is required to ensure the existence of a single surface phase. 24reover, the presence of natural dopants (Na, Mg, K, Ca, Ti, V, Cr) in commercially available natural hematite crystals can shift the thermodynamic stability window of bulk hematite phases. 31Surfaces make no exception: Foreign impurities can stabilize new surface phases.For example, K and Ti doping promotes (3×2) and (2×1) surface reconstructions on Fe2O3(11 ̅ 02) for oxygen chemical potentials at which the (1×1) bulk reconstruction occurs in pristine hematite. 32,33Sample impurities have also been linked to the (1×1) termination of Fe2O3 34 and α-Al2O3(0001). 35e rich chemistry of hematite requires analyzing the bulk and surface of iron oxide not as separated entities but as closely interrelated systems. 6This work aims at solving the current challenges using α-Fe2O3(0001) as a model system for UHV investigations 36 , and settle the controversy over the nature of the honeycomb/biphase reconstruction.The structural and chemical evolution of the bulk and surface of Fe2O3 single crystals was monitored at different redox conditions, also emphasizing the role of natural impurities.The surface of Fe2O3 was characterized by scanning tunneling microscopy (STM), non-contact atomic force microscopy (nc-AFM), low-energy electron diffraction (LEED), and XPS.Investigating the evolution of surface reconstructions on natural and synthetic Fe2O3 crystals allows disentangling the effect of sample impurities on their stability.High-resolution transmission electron microscopy (HRTEM) and electron energy loss spectroscopy (EELS) provide an atomically resolved view of the near-surface bulk phases of Fe2O3 single crystals.nc-AFM shows that the honeycomb/biphase Fe2O3 surface phase is formed by a compact 2D layer that is structurally perturbed locally.This local perturbation gives rise to the areas of distinct contrast identified as a 'biphase' in previous works. 23,37

Hematite bulk oxidation
The near-surface bulk of Fe2O3 is significant for on-surface catalytic reactions as it can store defects, host electron-hole pair creation and their transport to the surface or determine the penetration depth of light.Model hematite systems used for UHV investigations lack information on the extent to which the bulk transforms when equilibrating at different oxygen chemical potentials.Mixed bulk phases and the interface between bulk hematite and surface phases cannot be probed by surface-sensitive techniques.To fully describe the surface, one needs information about the near-surface bulk structure.
Bulk phases of hematite thin films grown on polar oxides can be probed by HRTEM. 38Here, HRTEM is used to access cross-sectional information on the sub-surface region of Fe2O3(0001) single crystals.Two limiting cases are presented: one sample featuring the Fe3O4(111) phase (µo < −2.5 eV), and one fully Fe2O3 stoichiometrically recovered (µo > −1.8 eV).As a technical note, the study of the sub-surface microstructure by TEM requires an electron-transparent specimen cut from the sample (a lamella).The fabrication of lamellas and the procedure to avoid surface contamination in the process is detailed in the methodology section.
The HRTEM images of the reduced sample (Figs.1a, b) show the near-surface structure of the Fe3O4(111)-terminated Fe2O3.Two distinct regions are divided by a sharp interface.The first region is a surface layer that can be identified as a Fe3O4 lattice oriented to the zonal axis (i.e., parallel to the electron beam) [1 ̅ 12], Fig. 1c, by evaluation of FFT and crystallographic standard. 39Note that HRTEM cannot unambiguously distinguish the lattice of Fe3O4 from γ-Fe2O3 due to their similar inverse cubic spinel crystal structure and lattice parameter.The second region is a single-crystalline bulk with a dumbbell atomic pattern identified as the rhombohedral lattice of Fe2O3 hematite oriented to the zonal axis [112 ̅ 0], Fig. 1d, by evaluation of FFT and crystallographic standard. 40Along the whole length of the lamella, the surface layer reveals a defect-free single crystal structure without fragmentation to misoriented sub-grains.
The thickness of the surface layer is about 20-25 nm; however.This thickness generally depends on the degree of reduction of the sample, i.e., by the ion beam energy, total sputtering dose, and the duration of UHV-annealing treatments performed prior to cutting the lamellae.
The HRTEM analysis of the fully reoxidized sample given in Figs.1e and 1f reveals that Fe2O3, again identified by FFT evaluation and crystallographic standard, is present up to the sample surface, forming an atomically sharp interface with the Ag protective layer.This indicates that a full oxidation of the sample can be achieved at µo > −1.8 eV even in a UHV chamber.EELS mapping performed in scanning-TEM mode (STEM-EELS) provides spatially resolved information about the chemical composition, oxidation, and valence states of the investigated iron oxides. 41Fig. 1g shows the EELS atomic concentration maps associated with Fe-L and O-K edges, obtained from the quantification of a spectrum image at the Ag-buried interface.The relative Fe and O concentration as a function of depth (along the arrow in Fig. 1g) is shown in Fig. 1h.The deeper bulk part shows a 40:60 Fe:O ratio, while a 43:57 ratio is measured in the upper surface layer.These values correspond well to the nominal stoichiometries expected for Fe2O3 (40:60) and Fe3O4 (42.9:57.1),respectively.Fig. 1i shows monochromated STEM-EELS spectra of the Fe-L3,2 fine-edge structure measured in both bulk (cyan curve) and reduced (blue curve) regions of the same sample.The bulk spectrum shows a strong pre-peak L3 component and two L2 main contributions, whereas the upper layer shows no L3 pre-peak and three L2 contributions.These spectra are consistent with those expected for hematite and magnetite (not maghemite), respectively. 42The combination of TEM and EELS structural and chemical analyses reveal the formation of a surface magnetite layer with a well-defined interface on the hematite bulk structure under the conditions of controlled UHV experiments.

Overview of the previously reported surface reconstructions
In combination with the bulk configuration, the initial structure and evolution of the surface of hematite determines the performance of a photoelectrocatalytic reaction (e.g., via the amount and type of surface states).Figure 2 shows an overview of individual surface reconstructions typically found on (0001)-oriented Fe2O3 natural single crystals under different oxygen chemical potentials.They are commonly referred in the literature as (a) (1×1)-Fe3O4(111), (b) honeycomb/biphase, and (c) (1×1)-Fe2O3(0001) phases.This nomenclature refers to the magnetite and hematite periodicities relative to the periodicity of the iron layers.Instead, sometimes the periodicities are specified relative to the oxygen basal planes in the literature; namely (2×2)-Fe3O4( 111) and (√3×√3)R30°-Fe2O3(0001).The (1×1)-Fe3O4(111) phase, Fig. 2a, is obtained when samples are treated under reducing conditions.It results from preferential sputtering of O atoms and subsequent annealing in UHV (µo <−2.5 eV). 1 Fe3O4(111) may terminate at up to 6 different possible layer cuts, of which the Fetet bulk termination displayed in Fig. 2a is the most stable as the single phase at µo = −2.5 eV. 43LEED shows the corresponding diffraction pattern.The Fe 2p XPS spectrum lacks the Fe 3+ satellite peaks characteristic of Fe2O3, 32 and shows a distinctive broad 2p3/2 component due to a mixture of Fe 2+ and Fe 3+ multiplet peaks.Fig. 2b shows the honeycomb/biphase reconstruction, which results from annealing in oxygen between µo = −2.5 eV and −1.8 eV.This surface displays complex tip-dependent STM contrasts and a floretted LEED pattern, arising from the Moiré superstructure formed by the Fe2O3 substrate and an FeO2 overlayer. 24The associated Fe 2p XPS signal develops Fe 3+ satellites and a sharper Fe 2p3/2 component.At higher µo, Fe2O3 natural crystals exhibit instead a (1×1) periodicity, Fig. 2c.The LEED pattern of the (1×1) phase shows only the main spots.The Fe 2p lineshape from the (1×1) phase shows Fe 3+ satellites typical of stoichiometric hematite.The controversy around the biphase vs. honeycomb models has its foundation in the complex STM contrast shown in Fig. 2b, which has often been interpreted as a coexistence of multiple distinct structural phases (FeO and Fe2O3).To address this controversy, nc-AFM measurements at 78 K were performed.Figure 3a shows an atomically resolved, constantheight image of the honeycomb phase.Individual atoms, imaged as dark circles, are arranged with a periodicity of 0.30 ± 0.01 nm (yellow rhombus).This basic structural motif is modulated on a large scale forming the 4 ± 0.1 nm honeycomb superstructure (green rhombus).Within the superstructure, three areas have a distinct appearance.These are marked as (√3×√3), 'bright' (1×1), and 'dark' (1×1).The fast Fourier transform (FFT) shown in Fig. 3b gives an additional 0.52 ± 0.03 nm periodicity, which is most apparent within the (√3×√3) region (blue rhombus).The unit cell vector with length of 0.52 nm corresponds to the hematite substrate.The LEED pattern in Fig. 2b and the FFT in Fig. 3b are strikingly reminiscent of a Moiré structure.It arises from the existence of a large periodicity, as revealed by the nc-AFM contrast in Fig. 3a.Previous work 24 attributed the presence of one (√3×√3) and two (1×1) domains in the STM images to differences in atomic heights of the last O layer of FeO2, according to the overlayer placement on the substrate. 24However, the different domains could also arise from electronic effects due to different tunneling between tip and sample at different sites. 37The nc-AFM images support the former scenario.They were obtained in constant height mode (high sensitivity to sub-nanometer atomic-height variations) at the contact potential difference between tip and sample.At these conditions, Pauli and van der Waals interactions (electrostatic contributions) are minimized.Moreover, since hematite is insulating at the acquisition temperature of 78 K, there is no tunneling current between tip and sample.Thus, the measured contrast should largely arise from slight atomic height differences along the honeycomb superstructure, which strongly supports the topographic origin of the AFM (and STM) contrast in favor of the overlayer (honeycomb) model.

Forming the honeycomb or (1×1)-Fe2O3(0001): The role of impurities
The (1×1)-Fe2O3(0001) atomic termination at high µo, Fig. 2c, is also subject to some controversy.The [0001] direction of hematite has a polar alternation of Fe and O planes.Based on the autocompensation mechanism, only a stoichiometric Fe-terminated surface would consistent with non-polarity under UHV conditions. 25,45However, (1×1) O-, Fe-, H-O-and mixed terminations have been reported for samples prepared in UHV and near-ambient pressure (NAP). 1,46It has been hinted that their stability might be linked to the metallic support used to grow Fe2O3 films. 27Hematite single crystals can be considered as a quasi-infinite system on which autocompensation must happen at a vacuum-exposed surface.On supported oxide thin films, this can occur at the metal-oxide interface, providing more degrees of freedom for surface terminations.Another possibility is that the surface structure is altered by natural or incorporated impurities. 34The (1×1) termination has been reported to form on both natural and synthetic single crystals. 47However, it cannot be ruled out that the (1×1) forms due to impurities (e.g., adventitious carbon and water) adsorbed during air transport to the measuring instruments.Natural impurities or added dopants such as K and Ti are known to induce surface restructuring on α-Fe2O3(11 ̅ 02). 32,33 disentangle the effects of sample impurities from other effects on given surface reconstructions, natural and synthetic Fe2O3 sample were prepared and investigated in UHV, using typical ranges of pressure and temperature found in the literature for the preparation of the honeycomb and (1×1)-Fe2O3(0001) phases.The amount and type of impurities in natural crystals vary with each sample.Common impurities detected during XPS investigations were alkali metals such as Ca, K, and Na, and transition metals such as Ti, Mn and Cr.Clean, epitaxial thin films of ≈100 nm thickness were grown by pulsed laser deposition (PLD) on natural hematite single crystals, 33 henceforth referred to as 'synthetic' hematite.These films are free from impurities within the resolution limit of the XPS setup.Table 1 shows the range of chemical potentials probed in this work when oxygen-annealing natural and synthetic α-Fe2O3(0001) samples.Natural crystals exhibit the honeycomb surface reconstruction when treated under reducing conditions and the (1×1)-Fe2O3(0001) termination under oxidizing conditions.The phase transition between honeycomb and (1×1)-Fe2O3(0001) on natural crystals fits well with the µo values reported for these phases on metal-supported hematite and natural crystals.On synthetic samples, however, only the honeycomb reconstruction was obtained; the (1×1)-Fe2O3(0001) termination could not be reproduced within the pressure and temperature ranges that are commonly applied.What is more, the growth kinetics and time used for equilibration of the honeycomb phase depends on µo.To obtain the honeycomb phase from the Fe3O4-terminated hematite within UHV-compatible oxygen pressures (p <≈ 10 −4 mbar), annealing between 2 and 8 hours is typically required (the exact duration depends on the history of the sample).However, merely 10−30 min are required under high oxygen pressure (≈ 1 mbar).

Table 1. Summary of surface structures obtained when equilibrating natural and synthetic
Fe2O3 at different oxygen chemical potentials.The honeycomb-to-(1×1) transition occurs at a µo ~ −1.8 eV only on natural samples; however, the precise value depends on the specific sample.

Contamination segregation during hematite oxidation.
The effect of the intrinsic impurities of natural crystals not only affect the stability of surface reconstructions of α-Fe2O3(0001), but also the bulk transformations and the fully stoichiometric recovery of hematite.Fig. S1a in the Supplementary Information shows a HRTEM image of a partially reoxidized magnetite-like inclusion within the fully recovered hematite.FFT analysis of the inclusion and the surrounding area, Figs.S1b and c the inclusion has the same magnetite/maghemite structure as shown in Fig. 3c, whereas the rest of the bulk has the characteristic hematite structure.The EELS mapping in Figs.S1d and S1e reveals an inhomogeneous presence of Mn in the 6 -12% range.The Fe concentration is locally decreased in an equivalent amount.O concentration is decreased only to a degree expected for magnetite.The Fe-L2,3 edge structure of the inclusion, Fig. S2, resembles that of magnetite reported in Fig. 1i.This evidence suggests the formation of Fe3-xMnxO4 ferrite.Outside the inclusion, the Mn concentration is below the detection limit of EELS.The interplanar-spacing vector lengths of the Mn-rich area is shown in Table ST1 in the Supplementary Information.
Mn locally hinders the fully reoxidation of the sample at oxygen chemical potentials at which a fully stoichiometric hematite is obtained on clean samples.Moreover, Mn single dopants alter the hole mobility in hematite 48 and, depending on the Mn concentration and sample preparation parameters, Mn possibly could lead to ferrite or Mn oxide formation.It is also likely that minute amounts of impurities such as Mn can distribute all over the hematite surface, resulting in a (1×1) bulk-like termination.
These results clearly show that the natural impurities play a crucial role in changing the thermodynamic balance of the hematite surface and near-surface structure and chemistry.
During the reoxidation, the impurities are pushed out, thus changing the surface and nearsurface composition and chemistry.The concentration and elemental distribution of contaminants vary within each natural crystal and preparation.Hence, synthetic films are essential for an adequate characterization of hematite in catalysis.

Conclusion
This work addressed the structure of the hematite α-Fe2O3(0001) surface and near-surface.
High-resolution nc-AFM images acquired in UHV strongly support that the oxidized surface of hematite is natively composed of a 2D oxide layer.The layer shows a Moiré pattern with three different areas due to the distinct attachment to the underlying fully oxidized bulk; these areas are responsible for the 'biphase' contrast observed in STM.A series of controlled experiments compared the oxidation of a synthetic films with natural crystals, the former clean within the resolution limit of XPS, and the latter contaminated by intrinsic impurities.The surface termination is strongly influenced by the presence of impurities, forcing the formation of surface phases that are not thermodynamically preferred on a clean sample.Specifically, the presence of impurities (e.g., alkali metals) enforces the formation of a (1×1)-Fe2O3(0001) termination instead of the honeycomb one in a wide range of oxygen chemical potentials.
Finally, subsurface cross-sectional imaging by TEM reveals the spatial extent of changes introduced by surface preparation methods and the spatial localization of impurities that can be associated only with a particular ferrite phase.
The atomic structure and transformations of α-Fe2O3(0001) during actual catalytic reactions are generally poorly understood. 36Impurities significantly influence the structure and chemistry of both the hematite surface and its bulk.This is of relevance in photoelectrocatalysis (water splitting, Fenton process, Fischer-Tropsch reaction), which typically uses alkalicontaining electrolytes (e.g., KOH, NaOH, and carbonates) and metal-doped hematite (e.g., Ti, Zn, Mn, Ni) to improve conductivity and charge separation.This study suggests that the surface and bulk of polar hematite facets will be stabilized by the photoelectrochemical conditions away from the initial state.The fact that the surface structure depends on the environment (both in terms of oxygen chemical potential and presence of alkali metal atoms) provides the essential clue to interpret the experimental results on the catalytic activity of hematite surfaces.Metallic dopants, surface (oxy)hydroxylation and/or cation adsorption will play a crucial role in achieving a non-polar, stable surface structure of α-Fe2O3(0001), i.e., the actual catalytically active phase.This should be considered when aiming to rationalize structure-function relationships.

Sample preparation:
Natural α-Fe2O3(0001) single crystals were acquired from SurfaceNet GmbH.A total of ten crystals have been used to carry out the experiments.Each crystal had a different concentration and elemental composition of impurities; the ones found were: Na, Ca, K, Mn, Zn, In, Ti, Mo, Cr, Al, Sr.
To reduce the samples into the magnetite phase, α-Fe2O3( 0001 (1 × 10 −6 mbar, 600 °C) to remove adventitious carbon contamination.Two crystals were heavily contaminated by bulk dopants, and it was not possible to reoxidize the magnetite phase.
Before the growth, the natural samples were cleaned by repeated sputtering-annealing cycles

Data processing:
XPS data was processed using the KolXPD software.The binding energy positions were calibrated using the fermi edge measured on a Ta sample plate.LEED images were acquired by averaging for 10 s with a camera in a dark receptacle.Dark-field images were acquired by turning off the LEED-screen acceleration voltage.They were subtracted from the original data to remove stray light and filament reflections.The contrast of the images was inverted to enhance the diffraction pattern.The STM and nc-AFM images were processed using custom ImageJ plug-ins.Microscope noise frequencies were filtered out.The lattice parameters of the honeycomb phase were obtained by dividing the distance between two spots containing 10-15 atomic positions by the number of unit cells.This was repeated in several locations in the three crystallographic directions to account for scanning distortions.The error bars correspond to the standard deviation of these measurements.The figures were prepared using ImageJ, Gimp and Inkscape.

TEM lamella fabrication and measurements:
The samples were covered by an Ag layer in the same UHV system were the samples were prepared.This protective cap prevents contamination of the topmost surface layers as a result of exposure to ambient conditions.Additionally, it prevents the surface from coming in direct contact with the reactive layers that are deposited during the lamella fabrication.Ag was deposited by thermal evaporation from a Knudsen-type effusion cell (crucible at 850 °C).The sample was held at room temperature in front of the effusion cell for 78 min, resulting in a ~30 nm-thick Ag film.The pressure in the chamber during deposition was 4 × 10 −10 mbar.
Afterwards, the sample was removed from UHV.
As the first step of the lamella fabrication, a protective cap (~ 350 nm thick) was deposited using electron-beam-induced deposition (EBID), followed by ion-beam-induced deposition (IBID) of the same element.Several gas injection system (GIS) chemistries for the deposition were tested (C, Pt, W): W or Pt did not introduce unwanted species into the layers of interest.
The deposition of C makes it difficult to restore the sample for UHV experiments and causes extensive deposition of carbonaceous layers during TEM/EELS measurements, which are detrimental to the quality of results.The lamella was then liberated from the bulk by FIB milling with 30 keV Ga ions and transferred onto a Cu support grid.In order to suppress carbon contamination of the lamella as much as possible, the FIB-SEM chamber was plasma-cleaned before inserting the bulk sample.Two final thinning and polishing steps were conducted at 5 and 2 keV beam energies, respectively.In this study, the TEM lamella surface is normal w.r.t.
The TEM measurements were performed at an accelerating voltage of 300 kV with a microscope TITAN Themis 60-300 (Thermofisher Scientific) equipped with a monochromator and a spherical aberration (CS) corrector of objective lens.The HRTEM images were acquired with CS ~ 0 m and with an appropriate defocus in range of few nm to observe atomic columns with minimum delocalization.Velox software v.2.12 was used for the image acquisition and processing of corresponding FFT patterns used for crystallographic evaluation.
The STEM-EELS measurements were performed with a Quantum ERS spectrometer (Gatan), a probe convergence semi-angle of 10 mrad, collection semi-angle of 28.The interplanar-spacing vector lengths d (Table ST1) obtained from the FFT of the Mn-rich phase (Fig. S1b) reveal a unit cell close within the experimental error to the structure of magnetite observed in Fig. 3b of the main text and reported elsewhere. 1The relative atomic

Figure 1 .
Figure 1.Reduced and oxidized hematite in side view.(a) and (b) HRTEM images of the

Figure 3 .
Figure 3. Analysis of the honeycomb phase with constant-height nc-AFM.(a) Atomically- Figs. 3c and 3d show a decomposition of the nc-AFM image in the adlayer (yellow) and Moiré (green) components of the FFT of Fig. 3b.The combination of Moiré and adlayer contributions in Fig. 3e produces an image typical for supported 2D materials. 44Adding the substrate contribution to Moiré plus adlayer composition produces the (√3×√3) and two (1×1) fine modulations, given in Fig. 3f.The (√3×√3) and (1×1) areas can be, in principle, identified as 'domains' with distinct local structure and chemistry (the local differences in AFM contrast indicate different tip-sample chemical forces), but not in the extent considered within the biphase model (alternating islands of different stoichiometry, FeO and Fe2O3).

(− 1 . 5
10 min, 1 × 10 −6 mbar Ar + , 1 keV; 30min, 600 °C, 1 × 10 −6 mbar O2) till no change in the contaminant signals was visible in XPS.Then, the samples were annealed at 850 °C for 1 h at 2 × 10 −2 mbar O2, to promote the flattening of the surface morphology and ensure complete oxidation of the crystals.The cleanliness of the films was checked by LEED and XPS.XPS, LEED, STM and nc-AFM characterization:Normal emission XPS measurements were carried out using a laboratory-based system (SPECS Surface Nano Analysis GmbH, monochromatized Al-Kα source) with a base pressure 5 × 10 −10 mbar.The core-level spectra were recorded with a pass energy of 20 eV, step size of 0.05 eV, and dwell time of 200 ms.No charging was observed.LEED patterns were acquired with SPECS ErLEED 150 setups.The LEED apparatus was under operation parameters for at least 1 h before transferring the samples to avoid contamination (C, F).High-energy (150−300 eV) electrons were used to probe traces of the magnetite phase below the surface layers.Roomtemperature STM images were acquired in several STM apparatuses with electrochemically etched W tips.The tips were treated on Au(111) before measuring hematite.Non-contact AFM images were acquired in an Omicron POLAR-SPM microscope at 78 K using Qplus® sensors (resonance frequency of ca.47 kHz, Q factor of ca.5000) with an electrochemically etched W tip. The tip was prepared on Cu(100) until a change in the resonance frequency smaller than Hz at 0.1 V bias was obtained while approaching the tip, and the contact potential difference between tip and sample was < 0.2 V.

2 or 56. 4 mrad
and entrance aperture of 2.5 or 5 mm, respectively.The EELS datasets were obtained with GMS v.3.3 software with enabled Dual-EELS mode allowing a simultaneous collection of a low-loss spectrum image and a high-loss spectrum image containing a zero-loss peak and edges of elements of interest, respectively, in each pixel.The used pixel size (i.e., the spatial resolution) and pixel time of the high-loss spectrum images was 0.3 -0.5 nm and 0.02 -0.08 s, respectively.The EELS data of Fe-L3,2 fine-edge structure was acquired in monochromated STEM and with electron energy dispersion of 0.025 eV per channel giving an energy resolution approximately 0.12 eV.The EELS data for chemical concentration measurements and elemental mapping were performed with dispersion of 0.25 or 0.5 eV per channel.Relative chemical concentrations were calculated in At. % by using a model-based EELS quantification function included in the GMS 3, where the following settings were chosen with emphasis forMn-rich clusters on natural αFe2O3(0001)Intrinsic impurities on natural Fe2O3 crystals can be pushed to the surface and near-surface regions at the temperatures and oxygen partial pressures that are used to obtain stoichiometric hematite under UHV.FigureS1shows a representative case where Mn impurities form clusters with stoichiometry and structure close to inverse spinel Fe3O4.These clusters are embedded within stoichiometric hematite and are not oxidized to Fe2O3 within the temperature and pressure range achievable under UHV.

Figure S1 .
Figure S1.(a) HRTEM image of a Mn-rich cluster within hematite.(b) and (c) FFT analysis , reveals that