Adsorption and Incorporation of Transition Metals at the Magnetite Fe 3 O 4 (001) Surface

The adsorption of Ni, Co, Mn, Ti, and Zr at the (√ 2 × √ 2)R45 •-reconstructed Fe 3 O 4 (001) surface was studied by scanning tunneling microscopy, x-ray and ultraviolet photoelectron spectroscopy, low-energy electron diffraction (LEED), and density functional theory (DFT). Following deposition at room temperature, metals are either adsorbed as isolated adatoms or fill the subsurface cation vacancy sites responsible for the (√ 2 × √ 2)R45 • reconstruction. Both configurations coexist, but the ratio of adatoms to incorporated atoms depends on the metal; Ni prefers the adatom configuration, Co and Mn form adatoms and incorporated atoms in similar numbers, and Ti and Zr are almost fully incorporated. With mild annealing, all adatoms transition to the incorporated cation configuration. At high coverage, the (√ 2 × √ 2)R45 • reconstruction is lifted because all subsurface cation vacancies become occupied with metal atoms, and a (1 × 1) LEED pattern is observed. DFT+U calculations for the extreme cases, Ni and Ti, confirm the energetic preference for incorporation, with calculated oxidation states in good agreement with photoemission experiments. Because the site preference is analogous to bulk ferrite (XFe 2 O 4) compounds, similar behavior is likely to be typical for elements forming a solid solution with Fe 3 O 4 .


I. INTRODUCTION
Iron oxide surfaces play a central role in environmental chemistry and corrosion processes and are important in technologies such as microelectronics, biomedicine, and heterogeneous catalysis [1,2].Due to the common usage of iron oxides as a catalyst support, the adsorption of catalytically active late transition metals has been extensively studied [3][4][5][6][7][8][9].As is normal in metal-on-metal-oxide systems, threedimensional (3D) nanoparticles typically form because the surface energy of the metal oxide is lower than the metal [10][11][12].Considerably less effort has been invested studying adsorption of 3d transition and alkaline metals, despite the fact that they are known to promote catalytic reactions such as water-gas shift and ethylbenzene dehydrogenation, in which iron oxides are the active phase [13][14][15][16][17].The role of the promoter, or even its location within the iron-oxide surface, is rarely known.Given their propensity for oxidation, 3d transition metals have been reported to react with metal oxide surfaces, usually with concomitant reduction of the host cations at the interface [18][19][20].If the metal is soluble in the bulk of the oxide, diffusion into the bulk [21] competes with cluster formation at high temperature.
Here we study the adsorption of Ni, Co, Mn, Ti, and Zr on the (001) surface of Fe 3 O 4 using a variety of surface science techniques.Adsorption can occur as cationic adatoms twofold coordinated to oxygen, as observed previously [12,22,23] for Au, Ag, and Pd, but incorporation into the Fe 3 O 4 surface lattice is favored.The occupation of octahedral lattice sites is consistent with the structure of well-known bulk M 1−x Fe 2+x O 4 spinel ferrite phases, suggesting similar interfacial ternary oxide formation can be expected for other metals forming solid solution with Fe 3 O 4 .The formation of spinel ferrite phases on the conducting Fe 3 O 4 (001) substrate makes this class of exciting materials accessible to surface science studies [24].

II. METHODS
The experiments were performed on a synthetic Fe 3 O 4 single crystal grown using the floating zone method [25] and a natural Fe 3 O 4 single crystal.No differences between the samples have been observed.Scanning tunneling microscopy (STM), low-energy electron diffraction (LEED), and x-ray photoelectron spectroscopy (XPS) data were acquired in a two-vessel ultrahigh vacuum (UHV) setup consisting of a preparation chamber (base pressure <10 −10 mbar) and an analysis chamber (base pressure <7 × 10 −11 mbar) equipped with an Omicron μ-STM operated in constant current mode.The analysis chamber includes a commercial LEED setup, a VG Microtech dual anode x-ray source and a SPECS EA 10 Plus analyzer.The samples were prepared by cycles of sputtering with Ar + ions (E Ar = 1 keV, ion current density j sample ≈ 6.5 μA/cm 2 , 10 min) and annealing in O 2 at 870 K (p O 2 = 10 −6 mbar, 10 min).Temperatures were measured with a K-type thermocouple attached to the sample holder.Because the thermocouple is not attached directly to the sample, the temperature readout can be up to 50 K too low at high temperatures.Up to 300 • C, a systematic uncertainty of the given temperatures of approximately ±20 K is possible.Additional XPS data for Ni, Co, and Ti were acquired in a separate UHV chamber equipped with a He flow cryostat, direct current sample heating, a twin anode (Al/Ag) x-ray source with a Focus 500 monochromator, a SPECS Phoibos 150 analyzer, and a LEED setup.The analyzer has been calibrated (±0.1 eV) using a Au reference.In this system the sample was prepared by cycles of sputtering with Ne + ions (E Ne = 1 keV, j sample ≈ 2.3 μA/cm 2 , 10 min) and annealing in O 2 (using a directional doser) at 870 K (chamber pressure p O 2 = 8 × 10 −8 mbar, 15 min).The valence band and the Fe 2p photoemission experiments were performed at the Synchrotron Beamline I311 at the Max IV Laboratory in Lund, Sweden.After each injection the Au Fermi edge was measured as an energy reference.In all setups, Ni, Co, Ti, and Mn were deposited using Omicron electron-beam evaporators.Zr was deposited using a home-built Zr sputter source, similar in design to the one described by Mayr et al. [26].The deposition rates (≈1 ML/min, 1 ML = 1 adatom per unit cell = 1.42 × 10 14 atoms/cm 2 ) were calibrated using watercooled or temperature-stabilized quartz crystal microbalances (QCM).The coverages given in the description of the data are determined by the QCM calibration, which is in line with estimates from STM images where counting of adsorbed atoms is possible.
The density functional theory (DFT) calculations were performed using the full-potential augmented plane wave+local orbital (APW+lo) method as implemented in WIEN2K [27].We use the generalized gradient approximation with a Hubbard U (U eff = 3.8 eV) to treat the highly correlated Fe, Ni, and Ti 3d electrons.Adsorption on magnetite was modeled on a force-relaxed 17-layer slab with inversion symmetry and a vacuum layer of 25 bohr (13.2 Å), built from a ( √ 2 × √ 2) supercell of bulk cubic magnetite.The total energies were calculated with lattice parameters of bulk magnetite optimized for U eff = 3.8 eV.Atomic sphere sizes of 0.98, 0.79, 1.0, and 1.0 Å were used for Fe, O, Ni, and Ti, respectively.The energies of the different configurations are compared to the sum of the total energies of two free metal atoms in an otherwise empty cell (same size as the 17-layer magnetite slab with vacuum layer) and a pristine magnetite slab.A plane-wave cutoff of Rk max = 7.0, corresponding to 296.6 eV, was used, and the Brillouin zones of the surface models were sampled with a 3 × 3 × 1 k mesh.The surface models were relaxed until all forces were below 1 mRy/bohr (26 meV/ Å).A Fermi broadening of 0.08 eV was employed.STM simulations were performed in constant current mode using the Tersoff-Hamann approximation [28].

A. Experimental results
Magnetite crystallizes in the inverse spinel structure, based on an fcc oxygen lattice with interstitial iron atoms in octahedral (Fe oct ) and tetrahedral (Fe tet ) coordination to oxygen.The (001) surface forms a ( √ 2 × √ 2)R45 • reconstruction, in which two subsurface Fe oct vacancies (V Fe ) in the third (S-2) layer are correlated with an interstitial Fe int with tetrahedral coordination in the second (S-1) layer [Fig. 1  feature corresponding to a hydrogen atom bonding to surface oxygen (surface hydroxyl group) [31], a common adsorbate on this surface following UHV preparation.OH groups exhibit a characteristic hopping between two surface oxygen atoms and are thus easily distinguished from other surface species in STM movies [31,32].The inset in Fig. 1 001) surface reveal element-specific differences in the adsorption behavior.For Ni, the main features are bright protrusions between the Fe rows.An example is highlighted by the red circle in Fig. 2(a).The appearance is similar to that of Au, Ag, and Pd adatoms [12,22,23] adsorbed on the surface twofold coordinated to oxygen, and since the number of such features corresponds approximately to the nominal coverage determined using the QCM, we assign this species to Ni adatoms.For Mn, the number of adatom features is clearly less than the deposited amount of 0.2 ML, and many new protrusions appear that are located directly at the Fe rows (yellow arrows).The protrusions appear similar to OH groups at 1 V but have a different apparent height depending on bias voltage and do not exhibit the characteristic hopping of OH groups [32] in STM movies.Although the deposited coverages of Ti and Zr are higher than the ones for Ni and Mn, almost no Figure 3 shows an STM image sequence acquired for Co in which adatoms transition from the adatom to the on-the-row configuration at room temperature.The initial state after deposition resembles Mn/Fe 3 O 4 (001) [Fig.2(b)], i.e., a mixture of adatoms and on-the-row features.Three Co adatoms are marked by red circles in Fig. 3   In LEED, the diffraction spots associated with the ( 2)R45 • reconstruction become weaker as the coverage is increased, and at 2.15 ML they are completely absent.In both cases, the density of Ni adatoms is significantly less than deposited; for the nominal 1 ML coverage it is 0.44 ML, suggesting the remainder forms on-the-row features.However, individual on-the-row features are difficult to discern at such a high coverage.Interestingly there are areas in which two Ni adatom protrusions are observed per unit cell (nearest neighbor distance of 6 Å, red outline) rather than one, as observed at low coverage (yellow outline).Given that the subsurface rearrangement is directly responsible for the ( √ 2 × √ 2)R45 • periodicity, these data suggest that Ni modifies the subsurface restoring a bulklike spinel structure.
Photoelectron spectroscopy was used to determine the chemical state of the different metals.The XPS spectra in Fig. 6 show the 2p regions of Ni, Ti, and Co, as well as the Zr 3d peaks recorded at normal emission.In Fig. 6(a), Ni 2p spectra are shown for three different preparations: deposited at room temperature (red line), deposited at 200 K (blue line), and annealed to 465 K (black line).The room-temperature Ni 2p 3/2 spectrum shows a peak at ≈855 eV with a low-bindingenergy shoulder at ≈853.7 eV.This shoulder at ≈853.7 eV is significantly more pronounced when the Ni is deposited at low temperature.After annealing to 465 K, only the high-binding-energy component remains.Thus, two different species of Ni coexist in the surface region before annealing, in agreement with the results of the thermal stability study (Fig. 4).Both species have higher binding energy than metallic Ni, indicating that they bind to the lattice oxygen.The peak at 855 eV, dominant at RT and enhanced at high temperature, is associated with on-the-row features.The intensity shift to higher binding energy after annealing indicates that these onthe-row features contain Ni that is more oxidized than the Ni adatoms.
The Ti 2p spectrum in Fig. 6(b) shows symmetric peak shapes at binding energies indicating strongly oxidized Ti, similar to TiO 2 spectra [33].The binding energies of the Co 2p 3/2 peak, its satellite, and the satellite intensity are in good agreement with literature values [34] for CoO, indicating Co atoms with octahedral coordination to oxygen [Fig.6(c)].The Zr 3d 5/2 peak at ≈181.4 eV [Fig.6(d)] corresponds to Zr in mixed oxides [35], considerably shifted from the strongly oxidized state in ZrO 2 .
Figure 7 shows the effect of metal deposition on the Fe 3 O 4 substrate.The photoemission spectra were acquired at grazing angle (55 • from the surface normal) and at low kinetic energies.The red curve in (a) shows that the Fe 2p 3/2 spectrum of the pristine ( 001) surface is dominated by Fe 3+ (710.6 eV) [36].Ni or Ti deposition (black and blue curves) causes the emergence of a clear Fe 2+ shoulder (708.3 eV) [36].Interestingly, although the coverage of Ni is significantly higher (1 ML) than that of Ti, the Fe 2+ shoulder is more pronounced for the deposition of 0.4 ML Ti.This indicates that Ti deposition leads to a strong reduction of the surface Fe.Similar conclusions can be drawn from valence band photoemission.shift from the shoulder at ≈2.2 eV, which originates from Fe 3+ , to ≈1 eV, attributed [37] to Fe 2+ .

B. DFT results
To ascertain the adsorption geometry of Ni and Ti atoms we performed DFT+U calculations.Figure 8(a) shows the optimized structural model for a Ni adatom, twofold coordinated to oxygen.For Ti, such a twofold coordination is not the ground state as the Ti adatom relaxes closer to the surface striving for higher coordination to oxygen, eventually finding a local force minimum involving distortions of the lattice at the surface.Since the experimental data show that metal adsorption induces a (1 × 1) periodicity, configurations in which the metal atoms replace one Fe atom within a bulklike  Literature values [40] of the cohesive energy of TiO 2 indicate that the high energy differences for the Ti configurations result from the choice of free Ti atoms as an energy reference.
The valence state of the atoms can be determined by comparing the calculated total magnetic moments of a configuration to that of the pristine surface in order to estimate how much charge has been transferred to the surface.Additional information can be obtained from the magnetic moments and partial charges inside the atomic spheres.For the Ni adatom, the total magnetic moment of the slab changes by less than 0.1 μ B , whereas the magnetic moment of Ni inside its atomic sphere is |μ| = 1.0 μ B and aligned with the Fe oct spin orientation (spin up).The partial charges of the Ni adatom indicate a 4s 0 3d 9 -like state with one d electron in linear bond coordination to the surface oxygen atoms.Hence the Ni adatom donates one electron to the surface and assumes a +1 state.In the case of incorporated metal atoms, the major difference in the total magnetic moment of the slab with respect to the subsurface cation vacancy structure is caused by the rearrangement of Fe atoms in the S-1 and S-2 layers [ |μ| ≈ 20 μ B for the symmetric slab of Ni oct (S), Ni oct (S-2)].If the contributions of the rearranged Fe and the magnetic moment of the Ni atom are subtracted from the total |μ|, the remaining difference gives a good estimate of the Ni charge donated to the slab.In the case of Ni oct (S) and Ni oct (S-2), the remaining magnetic moment is ≈2μ B , corresponding to Ni 2+ .The partial charges inside the atomic sphere are in good agreement with a Ni 2+ state, showing a 4s 0 3d 8 -like configuration with unoccupied minority spin e g -like orbitals for octahedrally coordinated Ni.
In all cases Ti is found to have small magnetic moments (|μ| < 0.12 μ B inside the atomic sphere) and negligible charge in the orbitals of 4s and 3d character.This allows the immediate conclusion that it assumes a Ti 4+ state in all configurations.All these valence states agree well with the peak positions observed in the XPS spectra.
Figures 8(c)-8(f) show the electronic density of states (DOS) of the Ni/Ti atoms and of the surface layer and the first two subsurface layers of Fe oct (normalized to the number of Fe atoms) for different configurations, i.e., the subsurface cation vacancy structure before metal deposition (c) and three energetically favorable configurations of Ni/Ti atoms [Ni adatom (d), Ni oct (S-2) (e), and Ti oct (S-2) (f)].The position of the Ni electronic states for the Ni adatom indicates that the increased intensity at ≈1 eV below E F in the photoemission experiments partially originates from Ni 3d occupied states close to the Fermi energy.For incorporated Ni and Ti, however, these states shift to higher binding energies or to unoccupied states.The comparison of the Fe oct DOS shows that the increase in Fe 2+ following metal deposition observed in photoemission is reflected in the DFT calculations by the emergence of an occupied spin-down electronic state close to the Fermi level in the first three Fe oct layers.The Fe oct occupied DOS in the surface layer remains largely unaltered.In the unoccupied states, however, a sharp increase of the unoccupied surface Fe oct DOS close to the Fermi level is observed for the incorporation of Ni and Ti.In the first two subsurface layers, the Fe oct atoms of the clean reconstructed Fe 3 O 4 (001) surface exhibit hardly any occupied density of states close to the Fermi level.The adsorption of Ni adatoms causes a distinct peak at ≈E F − 0.3 eV to appear in the Fe oct DOS of both subsurface layers.The incorporation of Ni in the first subsurface octahedral layer (S-2) leads to a significant enhancement at ≈E F − 0.4 eV in the second subsurface layer (S-4).Ti occupying an octahedrally coordinated subsurface Fe site leads to a considerable peak in the occupied subsurface Fe oct DOS in the region between E F − 0.3 and E F − 0.6 eV.
All these results are consistent with an increase in Fe 2+ observed in the valence band and Fe 2p photoemission spectra.The change from Fe 3+ to Fe 2+ is a direct result of the incorporation of additional metal atoms donating electrons to the slab.While in the case of Ni incorporation, the charge state of the Fe atoms in the (S-2) layer remains unaffected and Fe 2+ is only found in deeper layers, the incorporation of Ti leads to one Fe 2+ in the (S-2) layer and the emergence of a charge-ordered state from the (S-4) layer inwards, similar to that observed in the bulk-terminated Fe 3 O 4 (001) structure [41,42].
In Figs.8(g)-8(k) we show simulated STM images based on the minimum-energy DFT+U configurations described above.The clean Fe 3 O 4 (001) surface [Fig.8(g)] exhibits the undulating rows of surface Fe oct atoms, as shown previously, and the addition of a Ni adatom results in a bright protrusion between the Fe rows [Fig.8(h)], as expected simply on the basis of topography.The simulations of the incorporated Ni and Ti both exhibit a pair of particularly bright surface Fe oct atoms, similar in appearance to the "on-the-row" features observed in experiment.Interestingly though, in the case of Ni, the bright surface Fe oct atom pair is not located directly above the subsurface Ni but rather above the opposite row where the relocation of Fe int into the octahedral site locally restores the bulk Fe 3 O 4 structure.For Ti, a similar contrast enhancement occurs above the Fe oct (S-2), but the STM contrast is dominated by a brighter Fe oct pair located directly above the incorporated 075440-7 Ti oct (S-2) atom.A comparison of simulated STM images of the Ti oct (S-2) and Ti oct (S) configurations is shown in Fig. S1 (see the Supplemental Material) [43].Note that for Ti oct (S-2) the STM simulation is based on a 2 × 2 unit cell rather than the ( √ 2 × √ 2)R45 • cell utilized previously.This choice ensures an area of the surface remains reconstructed, allowing easier comparison of the modified contrast to the clean Fe 3 O 4 (001) surface.

IV. DISCUSSION
Based on the experimental data and DFT calculations, a full picture of the adsorption behavior of metals at the Fe 3 O 4 (001) surface can be developed.Two distinct adsorption configurations exist following deposition at room temperature; adatoms and "on-the-row" features related to incorporation of metal atoms in the surface lattice.The adatom configuration is twofold coordinated to surface oxygen and similar to that reported previously for Au, Ag, and Pd [12,22,23].Although the adatoms occupy a bulk-continuation tetrahedral site, it is important to note that they are in no way part of the spinel compound; while a bulk Fe atom should be 3+ and spin down in this site, the Ni adatom, for example, is close to 1+ and has a spin-up orientation.
For all metals studied here the incorporated configuration is clearly more stable.Upon deposition the barriers for incorporation are surpassed by the thermal energy of the evaporated atoms and the energy released by initial bond formation.The remaining adatoms are fully incorporated following mild annealing.Photoemission experiments show that the incorporated metal atoms are more oxidized and exhibit peak positions and line shapes that compare well with relevant bulk metal oxide compounds (e.g., the Co 2p core level shift and peak shape resemble that of Co 2+ in CoO, and Ti 2p resembles that of Ti 4+ in TiO 2 ).Our DFT calculations find that the lowest-energy configurations occur when the foreign metal atom occupies a subsurface octahedral site, and STM simulations of this structure compare very well with the "on-the-row" features observed in experiment.Note that the deposited metal fills one of the Fe oct vacancies of the ( √ 2 × √ 2)R45 • reconstruction, which induces Fe int to occupy the other, thereby restoring the spinel structure in the surface.This is consistent with the (1 × 1) LEED pattern observed at high metal coverage as well as the emergence of a second stable adsorption site for adatoms.
The incorporation of the foreign metal atoms into the lattice modifies the electronic structure of the surface.Fe 2p XPS spectra clearly show an enhancement in Fe 2+ in the surface region for all metals considered; the effect is strongest for Ti.The DFT+U calculations for incorporated Ni oct (S-2) reveal that Fe oct atoms in the surface Fe oct -O layer remain Fe 3+ , but that one Fe oct in the subsurface (S-4) layer obtains a magnetic moment of ≈3.6 μ B , indicative of an Fe 2+ oxidation state.For Ti this effect is stronger, resulting in Fe 2+ -like magnetic moments (≈3.5μB , ≈3.7μ B ) for two Fe oct in the subsurface (S-2) layer.The enrichment in Fe 2+ leads to an increase in the Fe oct DOS near E F [Fig. 7(b)], which DFT calculations find to be 100% spin polarized.These results suggest that doping the Fe 3 O 4 (001) surface with metals may be a viable route to tailor the spin polarization of the interface in Fe 3 O 4 based spintronics devices.
While the general trend for incorporation is common to all five metals in the present study, the energy scale of this process is element specific.Ni forms mostly adatoms (Ni 1+ ) at room temperature, whereas Ti is almost exclusively incorporated into the surface (as Ti 4+ ).Co and Mn are somewhat intermediate, which raises the question whether there is an underlying trend.A simple approach to classify the affinity for oxidation of each metal is to consider the heat of formation of the most stable respective oxide, which decreases as one follows the sequence of elements from Ti to Ni across the periodic table [10].This is in good agreement with the high DFT energy gain of incorporated Ti compared to Ni and reflected in experiment in the tendency to incorporate into the surface rather than forming adatoms with lower coordination to oxygen.An increased heat of oxide formation might also facilitate the incorporation into the lattice by providing the necessary energy to overcome the diffusion barrier for incorporation.
The incorporation of transition metal atoms within the Fe 3 O 4 lattice invites comparison to the transition metal ferrites, ternary spinel compounds with the formula MFe 2 O 4 or FeM 2 O 4 (M = metal) [44].The ferrites can be highly nonstoichiometric and are usually described in terms of a solid solution with Fe 3 O 4 .The M cation replaces Fe in either octahedral or tetrahedral sites depending on the metal, with the site preference depending on several factors, including relative cation size, oxidation state, and crystal field stabilization energy [44].Our DFT+U prediction that Ni 2+ and Ti 4+ occupy octahedral sites is in good agreement with the site preference for the respective bulk ferrite phases [38,39].As such, the incorporation of metals within the Fe 3 O 4 (001) surface can be viewed as the formation of a dilute ferrite phase localized at the surface or, alternatively, as an ultrathin ferrite film supported by a (conductive, lattice-matched) Fe 3 O 4 substrate.This system can thus be used to study the surface properties of insulating ternary compounds such as NiFe 2 O 4 (Fe 3 O 4 will provide sufficient conductivity for STM and electron spectroscopies) and can serve as a template for the growth of thicker stoichiometric ferrite phases.
Considering the ferrite analogy, we suggest that the formation of ternary surface phases is likely not restricted to the small set of transition metals studied here but most likely extends to any metal that can form solid solutions with Fe 3 O 4 , much of the periodic table.Recently, there has been much interest in synthesizing novel spinel phases for energy applications [24,45] and incorporation at the Fe 3 O 4 (001) surface could be used to quickly investigate the potential stability and properties of new and exciting spinel phases predicted by theory.Our observation of Zr incorporation in the Fe 3 O 4 lattice, for which there is no reported bulk ferrite analog, represents an early step in this direction.

V. CONCLUSION
The transition metals Ti, Mn, Co, Ni, and Zr were deposited onto the Fe 3 O 4 (001) surface.These transition metals are incorporated into the spinel lattice, filling subsurface Fe vacancies and locally lifting the ( √ 2 × √ 2)R45 • reconstruction.We suggest that such behavior is not exclusive to the metals studied but likely occurs for other metals that form solid solution with Fe 3 O 4 .
FIG. 1. (Color online) (a) DFT+U-optimized structural model of the ( √ 2 × √ 2)R45 • -Fe 3 O 4 (001) surface.The outermost layer consists of a bulklike layer containing Fe in octahedral sites (Fe oct , dark blue) and O (red).The first subsurface layer consists of tetrahedrally coordinated Fe (light blue): two bulklike Fe tet and one additional Fe int per ( √ 2 × √ 2)R45 • unit cell.The black triangle indicates the region where the subsurface differs from the regular spinel structure.Two octahedral Fe atoms are missing from the subsurface Fe oct -O layer (S-2) beneath Fe int (dashed blue circles).(b) STM image of a clean Fe 3 O 4 (001) surface, in which the yellow square illustrates the ( √ 2 × √ 2)R45 • unit cell.The bright double protrusions on the Fe rows (example highlighted by the cyan arrow) are due to surface hydroxyl groups, a common adsorbate at this surface.The inset shows a LEED pattern of the reconstructed Fe 3 O 4 (001) surface.The red and yellow squares highlight the (1 × 1) and ( √ 2 × √ 2)R45 • periodicities, respectively.

FIG. 2 .
FIG. 2. (Color online) Overview of the adsorption of different metals at the Fe 3 O 4 (001) surface at room temperature.STM images show that the deposited metals either adsorb as adatoms (red circles) or form bright surface features on the Fe oct rows (yellow arrows).The ratio of these two configurations varies for the different metals.(a) 0.2 ML Ni: almost exclusively adatoms.(b) 0.2 ML Mn: few adatoms, many on-the-row features (c) 0.4 ML Ti: hardly any adatoms, high coverage of bright on-the-row features.(d) 0.4 ML Zr: very few adatoms, mainly on-the-row features.

FIG. 3 .
FIG. 3. (Color online) Co adatoms transition into on-the-row features at room temperature.STM image sequence (same area) of 0.4 ML Co/Fe 3 O 4 (001) taken at room temperature.Three Co adatoms (red circles) are transformed into on-the-row features (encompassed by the yellow rectangles).
FIG. 4. (Color online) Thermal stability study of Ni/Fe 3 O 4 (001) (different area in each image): Examples of adatoms and on-the-row features are highlighted by red circles and yellow rectangles.(a) At room temperature almost exclusively adatoms are observed.(b) After annealing to 423 K the number of adatoms is diminished.They coexist with an increased number of on-the-row features.(c) Almost all adatoms are converted to on-the-row features.it was not possible to image the same area after heating.The as-deposited surface exhibits Ni adatoms only (examples highlighted by red circles), but after annealing adatoms and bright on-the-row features coexist [Fig.4(b)].After annealing to 448 K [Fig.4(c)] the adatom phase is almost completely converted into the bright on-the-row features.Increasing the annealing temperature above 448 K causes the Ni to diffuse into the bulk and a clean Fe 3 O 4 surface is recovered.STM images show neither adatoms nor on-the-row features.A ( √ 2 × √ 2)R45 • LEED pattern is observed.In XPS measured after annealing to 870 K at a photon energy of 1050 eV the Ni 2p peak in XPS vanishes, indicating diffusion into the bulk of the crystal (not shown).At metal coverages in excess of 1 ML, the ( √ 2 × √ 2)R45 • reconstruction is lifted in favor of a (1 × 1) periodicity.Figure5shows STM images of the Fe 3 O 4 (001) surface following deposition of 1 ML and 2.15 ML Ni at room temperature, together with the respective LEED patterns.
Figure 5 shows STM images of the Fe 3 O 4 (001) surface following deposition of 1 ML and 2.15 ML Ni at room temperature, together with the respective LEED patterns.

FIG. 6 .
FIG. 6. (Color online) Normal-emission XPS spectra showing regions characteristic of the chemical state of the deposited metals.The red curves were acquired following deposition at room temperature.For Ni (a), additional spectra were acquired following deposition at low temperature (blue curve) and annealing to 465 K (black curve).Two different Ni species contribute to the 200 K and 300 K spectra.The peak positions in (b)-(d) indicate strongly oxidized metal species, in agreement with incorporation into the Fe 3 O 4 lattice.
FIG. 7. (Color online) Comparison of photoemission spectra of clean Fe 3 O 4 (001), 1 ML Ni/Fe 3 O 4 (001), and 0.4 ML Ti/Fe 3 O 4 (001), both as deposited at room temperature.(a) In the Fe 2p 3/2 region, an increase of Fe 2+ after metal deposition is observed.(b) In the valence band spectrum, the metal deposition causes a significant increase of intensity at ≈1 eV.

FIG. 8 . 2 × √ 2 )
FIG. 8. (Color online) Summary of DFT calculations.(a) DFT+U optimized structural model of the reconstructed Fe 3 O 4 (001) surface with a Ni adatom (side view).(b) Bulk-terminated structural model with a Ti atom substituting Fe in the energetically most favorable site (DFT+U optimized) for incorporated Ti and Ni atoms.[(c)-(f)] Density of states (DOS) of Ni/Ti and Fe oct in the surface and first two subsurface layers for the clean surface and the favored configurations of Ni and Ti.[(g)-(k)] Simulated STM images of the four configurations based on a ( √ 2 × √ 2)R45 • unit cell for the pristine surface (g), the Ni adatom (h), and the Ni oct (S-2) configuration (i), and a (2 × 2) supercell for the Ti oct (S-2) configuration (k).
termination were also considered.The surface octahedral site [Fe oct (S)], a tetrahedral site in the first subsurface layer [Fe tet (S-1)], and an octahedral site in the second subsurface layer [Fe oct (S-2)] were selected as candidate sites.In these configurations the two Fe vacancies in the S-2 layer are filled by Fe int and one Fe atom [for the Ni/Ti oct (S) and Ni/Ti tet (S-1) configurations] or one Ni/Ti atom [for the Ni/Ti oct (S-2) configuration].This occupation of the cation sites corresponds to a bulk-terminated Fe 3 O 4 (001) surface with one Fe substituted by Ni/Ti.The structural model in Fig. 8(b) shows the force-relaxed Fe 3 O 4 (001) slab with a Ti atom in an Fe oct (S-2) site.This configuration is found to be energetically favorable for both Ni and Ti.Table I gives an overview of the energies of all configurations.The Ni oct (S-2) geometry ( E = −3.39eV) is clearly preferred compared to the

TABLE I .
Calculated energy differences for all studied configurations of Ni and Ti with respect to reconstructed Fe 3 O 4 (001) and free atoms, site preference of the metals in a bulk spinel ferrite compound, calculated oxidation states, and experimental XPS peak energies.The comparison of DFT+U energies shows that Ti oct (S-2) and Ni oct (S-2) are the preferred configurations.