An Ordered Mixed Oxide Monolayer Formed by Iron Segregation on

Nov 1, 2016 - Iron impurities in the rutile-TiO2(011) surface can result in the formation of an ordered, ternary iron–titanium oxide monolayer. Here...
0 downloads 8 Views 2MB Size
Download PDF
Article pubs.acs.org/JPCC

An Ordered Mixed Oxide Monolayer Formed by Iron Segregation on Rutile-TiO2(011): Structural Determination by X‑ray Photoelectron Diffraction Sandamali Halpegamage,† Luca Bignardi,‡ Paolo Lacovig,‡ Alan Kramer,† Zhan-Hui Wen,§ Xue-Qing Gong,§ Silvano Lizzit,‡ and Matthias Batzill*,† †

Department of Physics, University of South Florida, Tampa, Florida 33620, United States Elettra-Sincrotrone Trieste S.C.p.A., Strada Statale 14 Km 163.5, I-34149 Trieste, Italy § Key Laboratory for Advanced Materials, Centre for Computational Chemistry and Research Institute of Industrial Catalysis, East China University of Science and Technology, Shanghai 200237, P. R. China ‡

S Supporting Information *

ABSTRACT: Iron impurities in the rutile-TiO2(011) surface can result in the formation of an ordered, ternary iron− titanium oxide monolayer. Here the FeTi2O5 mixed oxide monolayer predicted by DFT simulations is confirmed by synchrotron based angle scanned X-ray photoelectron diffraction (XPD) studies. The ternary oxide monolayer has been synthesized on rutile-TiO2(011) substrate via two different experimental pathways: first, by segregation of Fe impurities from the bulk by annealing a clean TiO2(011) substrate in 1 × 10−7 mbar of oxygen at ∼450 °C and, second, by physical vapor deposition of Fe on clean TiO2(011) in 1 × 10−7 mbar of oxygen at ∼450 °C. For both preparation procedures an intermixed surface oxide is formed with Fe and Ti in 2+ and 4+ charge states, respectively. The surface is characterized by high-resolution and fast X-ray photoelectron spectroscopy (XPS), XPD, and low energy electron diffraction (LEED). Multiple electron scattering simulations implemented in the electron diffraction in atomic clusters (EDAC) package were performed for comparing experimental XPD patterns with structural models. The reliability of the approach was tested on XPD pattern of reconstructed clean TiO2(011)-2 × 1 surface, for which a structural model exists. The XPD pattern of the monolayer ternary iron titanium oxide is compared to structural models proposed by density functional theory (DFT)-based models. model, and “brookite (001)-like” model.5 By now, the latter is the accepted model since it is energetically more favorable than the other models and agrees well with surface X-ray diffraction6 and LEED I−V studies.7 Furthermore, the calculated STM images for the “brookite (001)-like” model is consistent with the zigzag row structure observed in experimental STM images of TiO2(011)-2 × 1 surface under appropriate tunneling conditions. We recently reported that deposition of certain transition metal oxides can result in restructuring of 2 × 1 reconstruction of TiO2(011) surface. Under controlled UHV preparation conditions these metal atoms intermix with the atoms in the surface of the substrate and form mixed oxide monolayers.8 Interestingly, we observed that the transition metals selected in our study formed an universal structure on the TiO2(011)

1. INTRODUCTION Titanium dioxide has been of great interest to materials scientists for decades due to its outstanding properties that can be utilized in a wide range of applications such as photocatalysis, heterogeneous catalysis, as a white pigment, and solid state gas sensor. The fact that TiO2 is chemically stable, abundant, inexpensive, and nontoxic has contributed to its technological importance. Ti minerals are mainly found in the form of rutile-TiO2, ilmenite-FeTiO3, and CaTiSiO5.1 Natural rutile commonly contains trace impurities such as Fe, Nb, Ta, Ca, Mg, etc.,1,2 which give rise to the color in the mineral samples. The (011) orientation of rutile TiO2 is the second most stable surface orientation of rutile and is reported to exhibit the highest photocatalytic activity for the rutile polymorph of TiO2.2,3 It is well established that TiO2(011) surface undergoes a (2 × 1) reconstruction in ultrahigh vacuum.4,5 Several possible structural models have been proposed in the past, namely the “titanyl” model, “microfaceting missing row type” © 2016 American Chemical Society

Received: September 23, 2016 Revised: November 1, 2016 Published: November 1, 2016 26414

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

for the FeTi2O5 model gives the similar R-factor (R ∼ 0.2) when compared to experimentally prepared mixed Fe−Ti-oxide monolayer, suggesting that the structure of this mixed oxide monolayer is in good agreement with the model. The reliability factors for Fe 3p diffraction pattern are, however, higher (less good). Nevertheless, the agreement is satisfactory, indicating that the DFT-based structural model is in good agreement with the actual surface structure.

surface that based on DFT simulations was identified to exhibit a MTi2O5 composition for all M investigated, i.e., M = Fe, V, Ni, and Cr. There are two main aspects that motivate the study of monolayer oxides on bulk oxide substrates in the field of chemical surface properties: (i) All technological materials possess some bulk impurities (sometimes bulk dopants, including iron, are added on purpose), and these impurities/ dopants may under certain thermodynamic conditions segregate to the surface. Thus, impurities may, sometimes unknowingly, modify the surface structure and properties. (ii) Monolayer oxides are used for creation of catalytic active phases. For instance, well-known monolayer oxide catalysts with an oxide monolayer supported on bulk oxide are V2O5/ TiO2, MoO3/Al2O3, CrO3/Al2O3, WO3/Al2O3, Re2O7/Al2O3, TiO2/SiO2, Fe2O3/SiO2, NiO/ZrO2, and WO3/ZrO2.9−11 Intermixed monolayer oxides where an oxidized transition metal intermixes with the surface of an oxide support to form a surface confined novel ternary oxide phase are less common but have also been reported.12 Thus, the segregation behavior and the formation of unique intermixed oxide surface phases are critical to understanding the surface chemistry of common oxides, like TiO2. In our previous studies we showed that many oxidized transition metals form intermixed oxide surfaces with the rutile TiO2(011) surface, and we speculated that the relative instability of the (011) surface compared to the more frequently studied (110) surface facilitates such an intermixing. Especially for iron oxide, a well-ordered and single phase with a FeTi2O5 composition was identified by scanning tunneling microscopy (STM), and a structural model was suggested based on density functional theory (DFT). For small amounts of Fe in the surface layer the surface structure segregated into two phases. One phase was the new intermixed FeTi2O5 phase, and the other phase remained as the pure TiO2(011)-2 × 1 phase. This phase segregation indicates that the two surface compositions formed what is known as line phases in bulk phase diagrams, suggesting that these two compositions form particularly stable, low energy surface phases. The details of our previous experimental studies can be found in ref 13 and for the DFT simulations in ref 14. In this paper we present experimental verifications of the most stable structural model proposed for the aforementioned FeTi2O5 monolayer. Determination of unique (sub)monolayer structures on bulk materials is a challenging task, and studies on single crystal model surfaces15−19 are generally needed that employ surface sensitive probes. Here we utilize angle-scanned X-ray photoelectron diffraction (XPD) studies,20 a method that uses the diffraction of photoemitted electrons by neighboring atoms to probe the atomic structure surrounding the emitting atoms. Using synchrotron radiation enables us to tune the photoelectron energy to extreme surface sensitivity and thus probing the structure of a monolayer. We test the expected agreement between experimental and simulated XPD patterns by comparing the results for the established “brookite(001)like” model for clean TiO2(011)-2 × 1 surface. We obtain similar reliability factors for the XPD pattern for the proposed FeTi2 O 5 structure for Ti 3s diffraction pattern. The brookite(001)-like model is well established for clean TiO2(011)-2 × 1 structure, and the reliability factor obtained (R ∼ 0.2) for this structure is an indication of the R-factor that should be expected when a model is in good agreement with the experimental XPD pattern. The R-factor obtained for Ti 3s

2. EXPERIMENTAL AND COMPUTATIONAL METHODS As we discussed in our earlier work, stabilizing a surfaceconfined single atomic layer oxide is sensitive to preparation conditions in UHV. Previously we synthesized the FeTi2O5 monolayer in a system equipped with STM. Although STM can characterize surface morphology and identify the presence of new surface phases, it is not well suited as a detailed structural probe. This makes confirmation of DFT-based theoretical structures uncertain despite the good agreement we obtained between simulated STM images and atomic-resolution images. XPD, on the other hand, is a structural probe that if combined with soft-X-rays can be extremely surface sensitive and thus powerful probe of surface structures. Multiple electron scattering of photoemitted electrons probes the local environment of the emitting atom. Monitoring the photoemission intensity as a function of emission angle thus gives information on the structure surrounding emitters of a specific element. To interpret the photoelectron intensity variation in XPD, the measurements need to be compared to multiple electron scattering simulations of a model structure. The measured intensity variation as a function of angle is the sum of all emitters of a specific element with different local environments, and thus the simulations need to take into account all possible local configurations in a specific structural model. The agreement between measurement and simulated XPD for a model structure is judged by using Pendry R-factor analysis.21 Below we describe the experimental and simulation methods in detail. 2.1. Experimental Methods: Sample Preparation, XPS, XPD, and LEED. All the photoemission and LEED experiments were performed in a UHV end station of the SuperESCA beamline at the Elettra synchrotron radiation facility, Trieste, Italy.22 The base pressure of the UHV system is ∼1 × 10−10 mbar. The beamline provides photons with energy 85−1500 eV with a resolving power, E/ΔE, 1 × 104 at 400 eV and 5 × 103 at 900 eV. The kinetic energy of the emitted electrons was detected with a Phoibos (SPECS GmBH) electron energy analyzer equipped with a homemade delay-line detector. The chamber is also equipped with LEED and a mass spectrometer. The commercially available TiO2(011) single crystal substrate (MTI Corporation) was mounted on a Ta plate to which a thermocouple was spot-welded very close to the sample. The Ta plate and the sample were heated with a filament placed behind them. The sample was mounted on a 5°-of-freedom manipulator. The sample was cleaned by multiple cycles of Ar+ sputtering at room temperature (1 kV, 6 μA, 20 min) followed by UHV annealing at 620 °C for 10 min. To ensure uniform cleaning, the sample was rotated to different combinations of polar/azimuthal angles during sputter and anneal cycles. Physical vapor deposition of Fe was done by heating a 2 mm diameter Fe rod in a water-cooled mini e-beam evaporator. To attain a radially uniform coverage, the sample was azimuthally rotated during deposition. 26415

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

between the emitters and the surface.25 The input parameters for the EDAC code for Ti 3s in clean TiO2(011)-2 × 1 model structure and for Ti 3s and Fe 3p in FeTi2O5 model structure are summarized in Table 1. For Fe-EDAC simulations of the FeTi2O5 structure only two Fe atoms per surface unit cell were considered as the emitters. In contrast, for Ti XPD simulations, Ti-emitters distributed down to 10 or 12 layers (∼10−12 Å) below the surface were considered in order to ensure that all the diffraction features were detected. A reasonable value for the Debye temperature θD (1000 K)26 was used in order to take into account for the thermal vibrations of the atoms. Polarization p-LP (which refers to linearly polarized light with the polarization vector contained in the plane of incident beam and the surface normal) and an emission angle window of 5° were chosen in accordance with the actual experimental conditions. Rmax, lmax, IMFP, θD, and V0 are not a priori exactly known and thus were optimized to get better theory−experiment agreement; however, it should also be pointed out that small changes (±∼20%) of these values do not significantly change the simulated XPD patterns. The quantitative comparison of experimental and theoretical photoelectron diffraction data is carried out by calculating the reliability factor (R-factor):

For high-resolution XPS the sample surface was normal to the analyzer. In the course of an XPD measurement the sample was rotated over polar θ and azimuthal Φ angles so that the polar angle runs from 0° (normal emission) to 70° out of normal (normal incidence conditions in the present setup), and the azimuthal angle spans 90° for each θ. Figure 1 shows a

R=

∑i (χexp, i − χth, i )2 ∑i (χexp, i 2 + χth, i 2 )

where χ represents the modulation function defined as χ=

Figure 1. Schematic reprecentation of the actual experimental XPD set up showing the surface orientaion of TiO2(011)-2 × 1 sample with respect to the analyzer and the photon beam.

I(Θ, Φ) − I0(Θ) I0(Θ)

(I0 is the average intensity for each azimuthal scan at fixed polar angle Θ), and the sum runs over all the available data points for the different angles. The lower the R-factor, the better is the agreement.

schematic diagram of the beamline geometry and the sample orientation. XPS and XPD measurements were taken for the Ti 3s and Fe 3p peaks with a photon energy of 175 eV and energy resolution below 50 meV. In the high resolution spectra the binding energy has been calibrated with the Fermi level of a Taclip used for mounting the sample. For determining the XPD patterns, the core level photoemission peaks were fitted with a Doniach−Sunijc function convoluted with a Gaussian broadening.23 2.2. Computational Methods: Simulation of XPD within EDAC Code. XPD patterns were simulated by multiple electron scattering simulations implemented in the EDAC package.24 The atom clusters were constructed from the atom coordinates determined in DFT simulations for various surface models.14 The size and the scattering volume of the cluster are defined by Rmax which is constrained by the computation time. A parabolic cluster is defined such that all the points that satisfy Rmax is greater than or equal to the sum of the distance to the surface and the distance to the reference point are included in the cluster. Parabolic shape for the cluster was favored since it increases the contribution of scattering of atoms that lie

3. RESULTS AND DISCUSSION In the following subsections we briefly review previously published STM and DFT results for clean TiO2(011)-2 × 1 structure and for the FeTi2O5 mixed oxide structure, followed by the experimental and computational results of the current work. 3.1. STM and DFT Structural Models of Clean TiO2(011)-2 × 1 and FeTi2O5 Mixed Oxide Monolayer. Previously, we reported the formation of FeTi2O5 monolayer on TiO2(011)-2 × 1 substrate by depositing Fe in 5 × 10−8 Torr of O2 at room temperature and then annealing the sample to ∼350 °C in the same oxygen background. Figure 2a,b shows atomically resolved STM images for the clean TiO2(011)-2 × 1 surface and for the FeTi2O5 mixed oxide surface. The insets show the rectangular unit cells for the two structures. It is apparent that both surfaces have the same unit cell symmetry and dimensions; i.e., the FeTi2O5 mixed monolayer forms also a (2 × 1) superstructure with respect to the rutile TiO2(011)

Table 1. Input Parameters Used in the EDAC Code for Different Elements in the Two Structures initial state of element

Rmax (Å)

emission energy KE (eV)

inner potential V0 (eV)

inelastic mean free path (Å)

no. of emitters

max orbital quan num lmax

Ti 3s for TiO2(011)-2 × 1 Ti 3s for FeTi2O5 Fe 3p for FeTi2O5

15 12 11

112 112 120

10 10 10

5.8 5.5 5.7

20 16 2

8.1 8.1 8.3

26416

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

Figure 2. STM images of (a) clean rutile-TiO2(011)-2 × 1 surface. The inset shows the high resolution image of zigzag rows run along [01−1] direction and the rectangular unit cell dimentions. Bright features are oxygen atoms located in the topmost layer. (b) FeTi2O5 monolayer on rutileTiO2(011)-2 × 1 obtained by depositing Fe at room temperature in 5 × 10−8 Torr of oxygen and annealing at ∼350 °C in the same O2 pressure.The inset shows the high resolution image of c(2 × 1) superstructure with rectagular unit cell. The tunneling current and the bias voltage used for the image are ∼400 pA and ∼1.2 V.

Figure 3. (a) DFT-based “brookite(001)-like” model proposed for TiO2(011)-2 × 1 structure, and (b) shows the top view of the (011) plane. (c) Lowest energy DFT-based structural model proposed for FeTi2O5 mixed oxide monolayer formed on TiO2(011)-2 × 1 surface, and (d) is the top view.

from the topmost O2c atoms in the proposed structure. In contrast, simulated STM images of the FeTi2O5 surface structure suggest that the Fe atoms on the surface layer give rise to the brightest features in empty state STM images. While the model shown in Figure 3c for the FeTi2O5 surface is the most stable and has the best match to the STM studies, other stable surface models have been obtained from DFT.12 These models and their corresponding simulated XPD patterns are shown in the Supporting Information, but neither of these structures exhibit any close resemblance to the measured XPD patterns and are therefore not further discussed. 3.2. XPD and XPS of Clean TiO2(011)-2 × 1 and FeTi2O5 Mixed Oxide Monolayer. The XPS for a freshly

substrate. The side view and the top view of the corresponding DFT structural models for these two surfaces are shown in Figure 3a,c and in Figure 3b,d, respectively.5,12 The energetically most stable model of the clean TiO2(011)-2 × 1 surface, that is, the “brookite(001)-like” model, contains 5-fold coordinated Ti atoms uniformly surrounded by neighboring O atoms.5 On the other hand, the FeTi2O5 surface consists of uniformly distributed Fe and Ti atoms in 1:2 ratio.12 It should also be noted that the formation of an intermixed monolayer has resulted in complete reordering of the originally reconstructed surface of clean TiO2(011) surface. On the pure TiO2(011)-2 × 1 surface, the features imaged in STM as bright oval-shaped protrusions, arranged in zigzag rows come 26417

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

Figure 4. (a) X-ray photoemission spectra for clean TiO2(011)-2 × 1 sample after sputtering at room temperature (bottom) and after annealing in UHV at 600 °C (top). (b) Growth of Fe 3p peak while annealing in 1 × 10−7 mbar of oxygen at ∼450 °C. (c) High resolution spectra and peak fitting of Ti 3s and Fe 3p measured at photon energy of 175 eV, after annealing in oxygen. (d) High resolution spectra for Fe 2p measured at photon energy of 850 eV. The peak shape and position suggest that Fe is in 2+ charge state.

sputtered clean TiO2(011)-2 × 1 sample is shown in Figure 4a. The Ti 3s peak is broad due to sputter-induced reduction and the presence of Ti species (for e.g. Ti3+ and Ti2+) overlapped with Ti4+. After annealing to temperatures higher than 600 °C, the peak became narrower and shifted to higher binding energy, consistent with the recovery of the stoichiometric TiO2 surface. The sample was annealed for ∼10 min at this temperature in UHV. Then the temperature was lowered to 450 °C, and 1 × 10−7 mbar of O2 was introduced in the chamber. In the presence of oxygen the Fe 3p signal started to grow (see Figure 4b) until saturation, then the oxygen was turned off. The Fe 3p/Ti 3s peak ratios for this process are plotted in Figure 5a. It is apparent that the rate of Fe diffusion to the surface slows down with time, which may be a consequence of limited iron impurities in the bulk and may also be due to reaching an equilibrium. Iron only segregates to the surface in the presence of an oxygen background, while annealing in UHV (or strongly reducing environment) allows iron to diffuse into the bulk. Thus, the surface structure and composition are strongly dependent on the chemical environment of the crystal. The high resolution Ti 3s, Fe 3p, and Fe 2p core level spectra for the sample with iron segregated from the bulk are shown in Figure 4c,d. The peak shapes and binding energies indicate that Fe is in a 2+ and Ti remains in a 4+ charge state for the ironsegregated surface. The measured Fe 3p/Ti 3s ratio of 0.23 can be used to estimate the amount of iron in the surface layer. If we assume that Fe atoms only reside in the topmost surface layer while the Ti signal originates from surface as well as subsurface atoms, we can find an expression for the ratio of the core-level intensity as follows:

Figure 5. Fe 3p/Ti 3s peak ratios plotted vs time for (a) the sample annealed in 1 × 10−7 mbar of oxygen at 450 °C and for (b) Fe deposition in 1 × 10−7 mbar of oxygen at 450 °C.

26418

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

Figure 6. (a) Experimental XPD pattern for Ti 3s of clean TiO2(011)-2 × 1 surface. (b) Simulated XPD pattern of Ti 3s for “brookite(001)-like” model of TiO2(011)-2 × 1 proposed by DFT-based calculations. (c) Experimental XPD overlapped with simulated XPD. The two patterns are in good agreement qualitatively and quantitatively (R = 0.22). (d) Line profiles and R-factors along azimuthal scans for selected polar angles. The polar angles selected are marked by dashed circles in XPD plots given in (a) and (b). ∞

(1 − α)A Ti + ∑n = 1 A Ti e−nΔx / λ Ti ITi = αAFe IFe

approximately constant, it is apparent that the ratio saturates at a value of 0.49. Such saturation is explained if Fe only accumulates in the surface layer until the surface is completely covered with the FeTi2O5 phase and excess Fe diffuses into the bulk. This would further support the notion that the FeTi2O5 phase is a thermodynamically stable surface phase. In this scenario, the measured Fe 3p:Ti 3s peak intensity ratio is that for a complete FeTi2O5 monolayer on a TiO2 substrate. Using the above expression, we obtain α = 0.4, which is close but slightly higher than the value of α = 1/3 we would expect from the 1:2 ratio of Fe:Ti in the mixed oxide. Thus, it appears that our chosen values for λTi, ATi, and AFe slightly overestimate the iron concentration in the surface layer. This also suggests that the fraction of the mixed oxide just after oxygen annealing is closer to 50% rather than 66%. A roughly 50% coverage of the surface with FeTi2O5 phase is also obtained from XPD analysis, which is discussed below. XPD data were collected for the three samples discussed above, i.e., (i) a clean TiO2(011)-2 × 1, (ii) low coverage (∼1/ 2 ML) of Fe−Ti-oxide prepared by segregating Fe from the bulk, and (iii) high coverage (∼1 ML) of Fe−Ti-oxide prepared by depositing Fe in 1 × 10−7 mbar at ∼400−450 °C. XPD patterns were recorded for Ti 3s of the clean TiO2(011)-2 × 1 and for Ti 3s and Fe 3p of the low coverage and the high coverage Fe−Ti-oxide sample. Figure 6a shows the experimental XPD for Ti 3s of the clean TiO2(011)-2 × 1 sample. In Figure 6b the simulated XPD of Ti 3s for the TiO2(011)-2 × 1

where α denotes the fraction of iron atoms in the surface layer, Δx is the separation of atomic layers for a stratified crystal model, ATi and AFe are the atomic sensitivity factors for iron and titanium, respectively, and λTi is the inelastic mean free path length for the Ti 3s photoelectrons. The atomic sensitivity factors at the photoelectron energies of 120 and 112 eV are estimated to ATi = 0.3 and AFe = 0.9 for Fe and Ti, respectively,27,28 and the inelastic mean free path (λTi) for the Ti 3s photoelectrons at 112 eV kinetic energy is estimated to be 0.58 nm.23 The sample after annealing in oxygen shows an XPS ratio ITi/IFe = 4.0, hence α = 0.2. Based on our previously reported STM observation, the surface segregates into a pure TiO2 phase and a FeTi2O5 phase. Thus, all the iron signal originates from a phase with a Fe:Ti ratio of 1:2, and it is easily shown that α = 0.2 corresponds to 2/3 of the surface covered with the mixed oxide. However, there are large uncertainties in the values for atomic sensitivity factors and the inelastic mean free path of the photoelectrons, and thus an independent verification of the surface coverage is desirable. Therefore, next we deposited Fe on this sample in 1 × 10−7 mbar of O2 with the substrate held at ∼450 °C in an attempt to form a complete monolayer of the mixed oxide surface. The Ti 3s and Fe 3p signals were monitored while deposition and the change in Fe 3p:Ti 3s ratio is shown in Figure 5b. Although the Fe flux was 26419

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

Figure 7. (a) Experimental XPD pattern of Ti 3s for high coverage Fe−Ti−O sample prepared by Fe deposition in 1 × 10−7 mbar of O2 and at ∼450 °C. (b) Simulated XPD pattern of Ti 3s for DFT-based structural model of FeTi2O5 mixed oxide monolayer. (c) Experimental XPD overlapped with simulated XPD. The two patterns are in good agreement qualitatively and quantitatively (R = 0.23). (d) Line profiles and R-factors for azimuthal scans at some selected polar angles. The polar angles selected are marked by dashed circles in XPD plots given in (a) and (b).

No other structural model suggested by DFT simulations resulted in XPD patterns that were close to the experimental data. The simulated XPD patterns for these structures and the corresponding R-factors are given in the Supporting Information. Figure 7d shows line profiles over azimuthal angles for selected polar angles indicated by dashed circles in Figure 7a,b. The intensity variations of the theoretical and experimental line profiles follow in very good agreement, which is also shown by the low R-factors for the line profiles. The Ti 3s XPD for the sample prepared by Fe segregation from the bulk is shown in Figure 8d. The XPD pattern obtained for a not completely covered surface should be the sum of the XPD patterns of the FeTi2O5 and the TiO2(011)-2 × 1 surface phases. Therefore, we calculated mixed XPD patterns with different ratios of the two phases and used those mixed XPD patterns to calculate the R-factor with the experimentally observed XPD pattern. The change in the R-factor as a function of surface coverage with the FeTi2O5 phase is shown in Figure 8e for both a mixing of the EDAC results as well as the experimentally determined XPD patterns of the clean TiO2(011)-2 × 1 surface and a surface fully covered with a monolayer of FeTi2O5. For both approaches a minimum in the R-factor close to a 50% surface coverage is observed, which is in excellent agreement of the amount of Fe in the surface layer determined by the analysis of the Fe 3p:Ti 3s core level ratios. The “mixed” EDAC simulation with a 1:1 ratio is shown in Figure 8c. This result confirms that

model is shown, and in Figure 6c the experimental XPD is overlapped with the simulated XPD. As can be seen visually, there is an excellent agreement between the theory and the experiment. The R-factor calculated for the entire set of data is 0.22. Since the structural model for the TiO2(011)-2 × 1 reconstruction was previously confirmed by surface X-ray diffraction, our XPD study is an additional confirmation of this structure. The R-factor we obtain for this known structure is an indication of the kind of agreement we can expect to obtain between the experimental and simulated XPD patterns for the correct structural model. For clarity, Figure 6d shows azimuthal plots for few of the selected polar angles marked in Figure 6a,b by dashed circles. We see that the photoelectron intensity variations of respective profiles of theory and experiment follow very well, and the R-factors for the individual profiles are indicated in the figure. Turning to the structure of the iron−titanium surface oxide, Figure 7a shows the experimental XPD for Ti 3s for the sample prepared by Fe deposition and is assumed to cover almost the entire surface with the mixed ternary monolayer oxide. Figure 7b shows the simulated XPD of Ti 3s for the DFT-model of FeTi2O5. Figure 7c is experimental XPD overlapped with the simulated XPD. Once again, there is a very good agreement between the theory and the experiment. The R-factor for this plot was calculated as 0.23, which is almost identical to the agreement we obtained for the clean TiO2(011)-2 × 1 surface. 26420

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

Figure 8. (a) Simulated XPD pattern of Ti 3s for DFT structural model proposed for TiO2(011)-2 × 1. (b) Simulated XPD pattern of Ti 3s for DFT structural model proposed for FeTi2O5 mixed oxide monolayer. (c) The 1:1 mixture of two simulated XPD patterns. (d) Experimental XPD of Ti 3s for low coverage Fe−Ti−O sample prepared by annealing the clean TiO2(011) substrate in 1 × 10−7 mbar of O2 at ∼450 °C. It is clear that the experimental XPD best matches with the weighted mixture of the two simulated patterns. The R-factors as a function of a weighted mixture of the simulated XPD patterns as well as by mixing the experimental XPD patterns for a pure TiO2(011)-2 × 1 as well as what is assumed a complete FeTi2O5 mixed oxide monolayer is shown in (e). For both approaches a minimum close to a 1:1 ratio are observed, indicating that the surface is described by the coexistence of these two surface structures and that in this particular case the surface is covered by ∼50% with the FeTi2O5 phase.

consequence of the lower Fe concentration in the sample, but also occurrence of disordered clusters and/or defects (e.g., antiphase domain boundaries) in the monolayer phase cannot be excluded which will adversely affect the XPD measurements. Several attempts have been made to improve the R-factor by (i) introducing subsurface Fe, i.e., by replacing subsurface Ti atoms by Fe atoms, (ii) by introducing line defects due to the formation of antiphase domain boundaries, or (iii) relaxing the Fe layer normal to the surface. Relaxing the Fe layer normal to the surface did not improve the XPD agreement visually or quantitatively. However, superimposing the XPD intensities for the structures with line defects and with subsurface Fe, the pattern visually improved and the R-factor was also lowered slightly to R = 0.46 (see Supporting Information Figure S8). 3.3. LEED Patterns of Clean TiO2(011)-2 × 1 and FeTi2O5 Mixed Oxide Monolayer. LEED patterns were recorded for several different primary electron energies, and the patterns for 71 eV are shown in Figure 10a,b. The unit cells for clean TiO2(011)-2 × 1 and for the new Fe−Ti oxide monolayer have the same symmetry and size. This is in agreement with the proposed structural model. In STM images of the FeTi2O5 mixed oxide monolayer we observed a high density of antiphase domain boundaries with 1/2 a unit cell offset along the

the Fe in surface layer segregates into a separate structural phase at the surface and thus for low amounts of iron the surface consists of two structural and compositional phases, namely, a pure TiO2(011)-2 × 1 phase and an ordered mixed oxide with a FeTi2O5 composition. While the Ti 3s XPD shows good agreement with the FeTi2O5 DFT model, the agreement with the Fe 3p XPD is not as good. Figure 9a shows the simulated XPD of Fe 3p for the FeTi2O5 model, and Figures 9b and 9c show the experimental XPD data for Fe 3p for low coverage and for high coverage Fe− Ti-oxide sample, respectively. The same general XPD pattern for low and high Fe coverage is in agreement with the observed phase segregation and iron only being present in FeTi2O5 domains at the surface. Although visual inspection of the simulated XPD pattern with that of experimental results shows close similarities, the calculated R-factor is only 0.59. The disagreements between the simulations and the XPD data can also be seen from the line scans, although better R-factors are obtained from the line scans. The reasons for the relative worse agreement between the simulated and measured XPD data can have various origins. The much noisier signal for Fe 3p and the resulting weaker intensity modulations in XPD compared to Ti 3s affects the calculated R-factors. The weaker Fe signal is a 26421

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

Figure 9. (a) Simulated XPD of Fe 3p for DFT-based structural model proposed for FeTi2O5 mixed oxide monolayer. (b) Experimental XPD of Fe 3p for Fe−Ti−O sample prepared by annealing the clean TiO2(011) substrate in 1 × 10−7 mbar of O2 at ∼450 °C. (c) Experimental XPD pattern of Fe 3p for high coverage Fe−Ti−O sample prepared by Fe deposition in 1 × 10−7 mbar of O2 and at ∼450 °C. (d) Line profiles and R-factors of azimuthal scans for selected polar angles.

Figure 10. Low energy electron diffraction (LEED) pictures taken for clean TiO2(011)-2 × 1 sample and for the high coverage Fe−Ti−O sample prepared by Fe deposition in 1 × 10−7 mbar of O2 and at ∼450 °C, at electron energy of 71 eV. The size and the symmetry of the surface unit cell seem to be the same for both samples. Both patterns have missing spots, suggesting the glide plane symmetry in the two structures.

4. CONCLUSIONS

direction. These domain boundaries may give rise to the streakiness of the LEED patterns after forming the Fe−Ti-oxide layer. The LEED patterns for both the TiO2(011)-2 × 1 surface2 as well as the FeTi2O5 surface both exhibit “missing” (2n − 1, 0) spots. This indicates a glide plane symmetry of the unit cell of both structures in agreement with the proposed structural models.

Impurity segregation in TiO2 and other oxides is well-known. Here we show that on the rutile TiO2(011) surface iron impurities segregate to the surface only in the presence of an oxygen background pressure. Furthermore, by vapor deposition of iron, we observe that the iron content in the surface layer at elevated temperature is limited to the formation of a stable 26422

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C

Functional Theory, X-Ray Diffraction, and Scanning Tunneling Microscopy Study. Surf. Sci. 2009, 603, 138−144. (6) Torrelles, X.; Cabailh, G.; Lindsay, R.; Bikondoa, O.; Roy, J.; Zegenhagen, J.; Teobaldi, G.; Hofer, W. A.; Thornton, G. Geometric Structure of TiO2(011)(2 × 1). Phys. Rev. Lett. 2008, 101, 185501. (7) Chamberlin, S. E.; Hirschmugl, C. J.; Poon, H. C.; Saldin, D. K. Geometric Structure of TiO2(011)(2 × 1) Surface by Low Energy Electron Diffraction (LEED). Surf. Sci. 2009, 603, 3367−3373. (8) Halpegamage, S.; Wen, Z.-H.; Gong, X.-Q.; Batzill, M. Monolayer Intermixed Oxide Surfaces: Fe, Ni, Cr, and V Oxides on Rutile TiO2(011). J. Phys. Chem. C 2016, 120, 14782−14794. (9) Wachs, I. E.; Briand, L. E.; Jehng, J.-M.; Burcham, L.; Gao, X. Molecular Structure and Reactivity of the Group V Metal Oxides. Catal. Today 2000, 57, 323−330. (10) Xie, Y.-C.; Tang, Y.-Q. Spontaneous Monolayer Dispersion of Oxides and Salts onto Surfaces of Supports: Applications to Heterogeneous Catalysis. In Advances in Catalysis; Eley, D. D., Pines, H., Weisz, P. B., Eds.; Academic Press: 1990; Vol. 37, pp 1−43. (11) Wachs, I. E. Molecular Engineering of Supported Metal Oxide Catalysts: Oxidation Reactions over Supported Vanadia Catalysts. In Catalysis; Spivey, J. J., Ed.; The Royal Society of Chemistry: 1997; Vol. 13, pp 37−54. (12) Colussi, S.; Gayen, A.; Farnesi Camellone, M.; Boaro, M.; Llorca, J.; Fabris, S.; Trovarelli, A. Nanofaceted Pd-O Sites in Pd-Ce Surface Superstructures: Enhanced Activity in Catalytic Combustion of Methane. Angew. Chem., Int. Ed. 2009, 48, 8481−8484. (13) Halpegamage, S.; Ding, P.; Gong, X.-Q.; Batzill, M. Ordered Fe(II)Ti(IV)O3 Mixed Monolayer Oxide on Rutile TiO2(011). ACS Nano 2015, 9, 8627−8636. (14) Wen, Z.-H.; Halpegamage, S.; Gong, X.-Q.; Batzill, M. Fe(II)Ti(IV)O3 Mixed Oxide Monolayer at Rutile TiO2(011): Structures and Reactivities. Surf. Sci. 2016, 653, 34−40. (15) Denk, M.; Kuhness, D.; Wagner, M.; Surnev, S.; Negreiros, F. R.; Sementa, L.; Barcaro, G.; Vobornik, I.; Fortunelli, A.; Netzer, F. P. Metal Tungstates at the Ultimate Two-Dimensional Limit: Fabrication of a CuWO4 Nanophase. ACS Nano 2014, 8, 3947−3954. (16) Ma, L. Y.; Doudin, N.; Surnev, S.; Barcaro, G.; Sementa, L.; Fortunelli, A.; Netzer, F. P. Lattice Strain Defects in a Ceria Nanolayer. J. Phys. Chem. Lett. 2016, 7, 1303−1309. (17) Artiglia, L.; Agnoli, S.; Vittadini, A.; Verdini, A.; Cossaro, A.; Floreano, L.; Granozzi, G. Atomic Structure and Special Reactivity toward Methanol Oxidation of Vanadia Nanoclusters on TiO2(110). J. Am. Chem. Soc. 2013, 135, 17331−17338. (18) Granozzi, G.; Rizzi, G. A.; Sambi, M. Structural Studies of Epitaxial Ultrathin Oxide Films and Nanoclusters by Means of AngleScanned Photoelectron Diffraction (XPD). J. Phys.: Condens. Matter 2002, 14, 4101−4117. (19) Uhlrich, J. J.; Sainio, J.; Lei, Y.; Edwards, D.; Davies, R.; Bowker, M.; Shaikhutdinov, S.; Freund, H. J. Preparation and Characterization of Iron-Molybdate Thin Films. Surf. Sci. 2011, 605, 1550−1555. (20) Granozzi, G.; Sambi, M. Angle Scanned Photoelectron Diffraction: Probing Crystalline Ultrathin Films. Adv. Mater. 1996, 8, 315−326. (21) Pendry, J. B. Reliability Factors for LEED Calculations. J. Phys. C: Solid State Phys. 1980, 13, 937−944. (22) Baraldi, A.; Barnaba, M.; Brena, B.; Cocco, D.; Comelli, G.; Lizzit, S.; Paolucci, G.; Rosei, R. Proceedings of the Sixth International Conference on Electron Spectroscopy Time Resolved Core Level Photoemission Experiments with Synchrotron Radiation. J. Electron Spectrosc. Relat. Phenom. 1995, 76, 145−149. (23) Baraldi, A.; Lizzit, S.; Comelli, G.; Paolucci, G. Oxygen Adsorption and Ordering on Ru(101−0). Phys. Rev. B: Condens. Matter Mater. Phys. 2001, 63, 115410. (24) García de Abajo, F. J.; Van Hove, M. A.; Fadley, C. S. Multiple Scattering of Electrons in Solids and Molecules: A Cluster-Model Approach. Phys. Rev. B: Condens. Matter Mater. Phys. 2001, 63, 075404. (25) Despont, L.; Naumović, D.; Clerc, F.; Koitzsch, C.; Garnier, M. G.; Garcia de Abajo, F. J.; Van Hove, M. A.; Aebi, P. X-Ray

mixed oxide phase with a FeTi2O5 structure, while excess iron diffuses into the bulk. This suggests that this iron−titanium oxide phase is a particularly stable surface phase. The preferred formation of this ordered phase is also apparent from the observed phase separation into a pure TiO2(011)-2 × 1 and this iron-containing phase, in cases where there is not sufficient iron to complete a monolayer. The composition of the surface layer is the same as of the mineral ferro-pseudobookite, but the proposed structure does not exhibit any clear resemblance to this known bulk phase. Therefore, the monolayer structure is a unique surface phase. The structural model of the FeTi2O5 surface oxide proposed by DFT was confirmed in the study presented here by synchrotron XPD measurements. On a general note, many oxide (photo)catalysts are intentionally or unintentionally doped by cations. This study illustrates that “bulk” dopants may give rise to unique surface phases that confined to a single atomic layer and may remain largely undetected but will affect the surface properties.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcc.6b09651. (i) EDAC simulations of XPD patterns for various other possible mixed iron−titanium oxide structures and their respective R-factors compared to the experimental results, (ii) LEED for different electron energies, (iii) illustration of the glide plane symmetry of the surface unit cells, and (iv) Fe 3p simulated XPD taking line defects into account (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail [email protected] (M.B.). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The USF group acknowledges financial support from the National Science Foundation under CHE-1505609. The ECUST group acknowledges the financial support from National Natural Science Foundation of China (21573067, 21421004, and 21322307) and the Program of Introducing Talents of Discipline to Universities (B16017). We also acknowledge the computing time in the National Super Computing Center in Jinan.

■

REFERENCES

(1) Meinhold, G. Rutile and Its Applications in Earth Sciences. EarthSci. Rev. 2010, 102, 1−28. (2) Dulub, O.; Di Valentin, C.; Selloni, A.; Diebold, U. Structure, Defects, and Impurities at the Rutile TiO2(011)-(2 × 1) Surface: A Scanning Tunneling Microscopy Study. Surf. Sci. 2006, 600, 4407− 4417. (3) Luttrell, T.; Halpegamage, S.; Tao, J.; Kramer, A.; Sutter, E.; Batzill, M. Why Is Anatase a Better Photocatalyst Than Rutile? Model Studies on Epitaxial TiO2 Films. Sci. Rep. 2014, 4, 4043. (4) Beck, T. J.; Klust, A.; Batzill, M.; Diebold, U.; Di Valentin, C.; Selloni, A. Surface Structure of TiO2(011)-(2 × 1). Phys. Rev. Lett. 2004, 93, 036104. (5) Gong, X.-Q.; Khorshidi, N.; Stierle, A.; Vonk, V.; Ellinger, C.; Dosch, H.; Cheng, H.; Selloni, A.; He, Y.; Dulub, O.; et al. The 2 × 1 Reconstruction of the Rutile TiO2(011) Surface: A Combined Density 26423

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424

Article

The Journal of Physical Chemistry C Photoelectron Diffraction Study of Cu(111): Multiple Scattering Investigation. Surf. Sci. 2006, 600, 380−385. (26) Wang, Y.-J.; Chang, J.; Tan, L.-N.; Chen, X.-R. Elastic Properties of Rutile TiO2 at High Temperature. Chin. Phys. Lett. 2007, 24, 2642− 2645. (27) Yeh, J. J. Atomic Calculation of Photoionization Cross-Sections and Asymmetry Parameters; Gordon and Breach Science Publishers: Langhorne, PA, 1993. (28) Yeh, J. J.; Lindau, I. Atomic Subshell Photoionization Cross Sections and Asymmetry Parameters: 1 ⩽ Z ⩽ 103. At. Data Nucl. Data Tables 1985, 32, 1−155.

26424

DOI: 10.1021/acs.jpcc.6b09651 J. Phys. Chem. C 2016, 120, 26414−26424