Direct Evidence of Surface Reduction in ... - ACS Publications

Article pubs.acs.org/cm

Direct Evidence of Surface Reduction in Monoclinic BiVO4 Marta D. Rossell,*,† Piyush Agrawal,†,‡ Andreas Borgschulte,§ Cécile Hébert,∥ Daniele Passerone,‡ and Rolf Erni†

Downloaded via WESTERN SYDNEY UNIV on January 11, 2019 at 18:42:37 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

†

Electron Microscopy Center, Empa, Swiss Federal Laboratories for Materials Science and Technology, CH-8600 Dübendorf, Switzerland ‡ Nanotech@surfaces Laboratory, Empa, Swiss Federal Laboratories for Materials Science and Technology, CH-8600 Dübendorf, Switzerland § Advanced Analytical Technologies, Empa, Swiss Federal Laboratories for Materials Science and Technology, CH-8600 Dübendorf, Switzerland ∥ Centre Interdisciplinaire de Microscopie Electronique, École Polytechnique Federale de Lausanne, CH-1015, Lausanne, Switzerland ABSTRACT: Local measurements of the oxidation state of vanadium in monoclinic BiVO4 particles by means of electron energy-loss spectroscopy in scanning transmission electron microscopy reveal a pronounced surface reduction: within a 5nm-thick shell, the oxidation state of vanadium is reduced from +5 to about +4. Thus, charge neutrality near the surface demands for ∼15% oxygen vacancies. Our results provide direct evidence for the segregation of oxygen vacancies at the surface of BiVO4. This observation is confirmed by X-ray photoelectron spectroscopy. The experimental findings are complemented with all-electron density functional theory based WIEN2k calculations of the density of electron states and of the electron energyloss near-edge structure. The theoretical results provide further information on the electronic changes induced by the experimentally verified oxygen vacancies.

■

INTRODUCTION Since its identification in 1998 as an active photocatalyst,1 BiVO4 (BVO) has emerged as a promising candidate for photoelectrochemical water splitting and wastewater treatment.2−5 As opposed to the most widely used and investigated TiO2 photocatalyst,6 the main advantage of BVO is its ability to use visible light rather than ultraviolet light irradiation to split water and to break down organic pollutants in sewage water. Additionally, the abundance and nontoxicity of its constituent elements make bismuth vanadate a low-cost and environmentally friendly photocatalyst. BVO is known to exist in three natural crystalline polymorphs: orthorhombic pucherite,7,8 tetragonal dreyerite,9 and monoclinic clinobisvanite.9 However, BVO synthesized in the laboratory normally does not adopt the pucherite structure. Among the three phases, the monoclinic clinobisvanite BVO (m-BVO) phase exhibits a much higher photocatalytical activity due to its favorable band gap (2.4−2.5 eV) in the visible region of the electromagnetic spectrum, and a valence band position suitable for driving water oxidation under illumination.2,10,11 The conduction band, however, is reported to be positioned too low relative to the proton reduction potential for hydrogen evolution.1 Nonetheless, the overall water splitting reaction can be accomplished by using a separate hydrogen evolution catalyst in tandem, commonly a noble- or transition-metal catalyst.5,12,13 Yet, due to its poor charge carrier mobility and separation of photogenerated electron−hole pairs,14 pure m© 2015 American Chemical Society

BVO shows a very low photoconversion efficiency for practical applications. In order to enhance the photoconversion efficiency of bismuth vanadate, several doping strategies have been investigated.12,15 While some of the incorporated impurities improve the charge transport, others simply affect the crystal growth and alter the formation of defects.16 Interestingly, oxygen vacancies, the dominant intrinsic defects in m-BVO,17 have been recognized to play a key role in the performance of clinobisvanite. According to Kho et al.,4 oxygen vacancies act as harmful photogenerated carrier traps. However, several groups have reported that the photoactivity of bismuth vanadate can be considerably enhanced by the presence of a high density of oxygen vacancies at the surface of m-BVO.18,19 Thus, Wang et al.19 argued that these defects increase the donor densities of BVO without introducing deep trap states and hence facilitate the charge transport and collection. Such contradictory opinions on the role of oxygen vacancies in the photoconversion efficiency of m-BVO highlight the importance of gaining a deeper understanding of the concentration and the spatial distribution of oxygen vacancies in m-BVO on the nanoscale. In this work, we have examined the structural properties and the electronic behavior of bismuth vanadate using a Received: November 19, 2014 Revised: April 14, 2015 Published: April 14, 2015 3593

DOI: 10.1021/cm504248d Chem. Mater. 2015, 27, 3593−3600

Chemistry of Materials

Article

combination of scanning transmission electron microscopy (STEM), electron energy-loss spectroscopy (EELS), and X-ray photoelectron spectroscopy (XPS). Vanadium reference spectra for V3+, V4+, and V5+ were obtained from bulk V2O3, VO2, and V2O5 reference samples and used for the characterization of a commercial bismuth vanadate powder. The experimental results are interpreted with the help of WIEN2k-based density functional theory (DFT) calculations of the density of electron states (DOS) and the energy-loss near-edge structure (ELNES).

■

EXPERIMENTAL SECTION

Materials. The studied m-BVO sample (99.9% purity) was purchased from Alfa Aesar. The V2O3, VO2, and V2O5 reference samples (≥99.9% purity) were supplied by Sigma-Aldrich, and the Au powder (max particle size 53 μm) was supplied by Goodfellow. Sample Characterization. X-ray diffraction (XRD) patterns were collected at room temperature using a Stoe Stadi P powder X-ray diffractometer in transmission mode (Cu Kα1 radiation, Gemonochromator, Dectris Mythen silicon strip detector). Samples for transmission electron microscopy were prepared by dispersing the m-BVO powder in ethanol and dropping the dispersion onto a lacey carbon coated copper grid, followed by a very gentle oxidizing plasma treatment to remove hydrocarbon contamination. In order to preserve the surface properties of the m-BVO particles, a very short plasma exposure (10 s) was applied using a Fishione shielded holder port which reduces the plasma density by reducing the crosssectional area available to the plasma near the holder tip. The m-BVO particles were studied by high-angle annular dark field (HAADF) STEM using a double spherical aberration-corrected JEOL JEMARM200F microscope operated at 200 kV. A probe semiconvergence angle of 25.3 mrad was set yielding a calculated probe size of about 78 pm. The annular semidetection range of the annular dark-field detector was calibrated at 68−280 mrad. The EELS data in this study were obtained using a JEOL 2200FS TEM/STEM microscope operated at 200 kV and equipped with an in-column Omega-type energy filter and a Gatan DigiScan STEM control system. Under typical operating conditions for the experiments described in this paper (spherical aberration CS of 1 mm, probe semiconvergence angle of 10.8 mrad, inner semidetection angle of the annular dark field detector calibrated at 100 mrad), the microscope provides a spatial resolution of about 1.6 Å. For the EELS data acquisition, the convergence and collection semiangles were set to 10.8 and 10 mrad, respectively. For these values, the energy resolution measured as the full width at half maximum of the zero-loss peak is ∼1.2 eV and the dispersion was set to 0.08 eV/channel. XPS surface analysis was performed in a modified VG ESCALab spectrometer with a base pressure below 10−9 mbar. The powder samples were mixed with Au powder and pressed to 0.2-mm-thick pellets (of 10 nm diameter). The samples were inserted via an Ar glovebox (O2 < 1 ppm and H2O < 0.1 ppm) directly connected to the spectrometer and transferred without exposure to air. XPS spectra were collected with a SPECS PHOIBOS 100 analyzer using a nonmonochromated X-ray source (Mg Kα: 1253.6 eV). The binding energy was recalibrated using the Au 4f7/2 peak set to 83.8 eV. First-Principles Calculations. To investigate the change in the density of states and in the ELNES spectra due to the presence of oxygen vacancies in BVO, we employed the full-potential linear augmented plane wave method within density functional theory via the WIEN2k code20,21 equipped with the TELNES 3.0 program.22,23 The exchange and correlation potential was treated by the generalized gradient approximation using the Perdew−Burke−Ernzerhof (PBE) parametrization. As input structure for the calculations, we used the clinobisvanite structure9 with space group I2/b; lattice parameters a = 5.1935 Å, b = 5.0898 Å, c = 11.6972 Å, and γ = 90.387°; and atomic positions given by Bi (0, 1/4, 0.6335), V (0, 1/4, 0.130), O1 (0.1465, 0.5077, 0.2082), and O2 (0.2606, 0.3810, 0.4493); see Figure 1b. Three cases were studied: (i) BiVO4 with no vacancies, (ii) 12.5%

Figure 1. (a) XRD pattern of the sample showing the characteristic split of the monoclinic clinobisvanite phase. (b) Rendering of the mBVO structure along [100] with (blue) Bi, (gray) V, and (red) O columns. The V is at the center of a distorted tetrahedron. The blue arrows show the displacement of the Bi3+ and V5+ cations from the centrosymmetric positions. (c) Atomic resolution HAADF-STEM image of the m-BVO crystal structure imaged along the [100] zone axis. Inset: corresponding electron diffraction pattern. oxygen vacancies simulated with one O1 atom removed out of eight oxygen atoms, and (iii) 12.5% oxygen vacancies simulated with one O2 atom removed out of eight oxygen atoms. The lattice parameters of all structures were relaxed internally using the minimization program min_lapw. The following parameters were used for the simulations: the muffin-tin radii Rmt were selected as 2.43, 1.67, and 1.51 au for Bi, V, and O atoms, respectively; the valence and core states were separated by 6.0 Ry of energy, and the plane-wave cutoff parameter Rmt × kmax was set to 7.00. We sampled the Brillouin zone with a uniform 5 × 7 × 5 mesh of k points. An electron beam energy of 200 kV, semicollection angle of 5.0 mrad, and semiconvergence angle of 1.8 mrad were used for the EELS calculations in the TELNES module. A broadening of 1.2 eV was used to account for the instrumental energy resolution.

■

RESULTS AND DISCUSSION A commercial BiVO4 powder with the monoclinic clinobisvanite (m-BVO) structure was employed in this study. In order to confirm the crystallographic structure of the as-received sample, powder XRD patterns were collected at room temperature. The XRD pattern of Figure 1a proves that the sample consists entirely of the monoclinic phase (JCPDS file no. 14-0688) with no traces of other phases. The obvious splitting of the (101) and (011̅ ) peaks around 18.5° and of the (200) and (020̅ ) peaks around 35° provides evidence for distinguishing the monoclinic from the tetragonal phase. The studied m-BVO particles have a random morphology with sizes between a few nanometers up to several micrometers. An atomic resolution HAADF-STEM image of the edge of one m-BVO particle imaged along the [100] zone axis is shown in Figure 1c. A model of the monoclinic BVO structure along the 3594



Article

vanadium edge onset and subsequently normalized to the intensity maxima of the L2 edge. The extraction of the O-K edge is however hindered by the proximity of the V-L2,3 edge. On the surface of the particle, a shift of the V-L2,3 edge of ∼1.2 eV to lower energies is observed. Additionally, the small shoulders appearing at the lower energy side of the L2 and L3 white lines of the “bulk” (black) spectrum are not observable at the “edge” (red) spectrum due to the broadening of the peaks. These observations are indicative of the presence of V4+ and V3+ species, as reported in previous EELS24,25 and X-ray absorption spectroscopy26 studies of LaVO3 and LaVO4. The change of the oxidation state of the near-surface vanadium ions could also be related to the presence of structural defects or impurities. Yet, our measurements did not reveal the presence of defects in any of the studied particles. Thus, the oxidation state of the surface vanadium ions is found to change owing to oxygen deficiency at the surface. To explore the extent of the reduction shell, we have measured V-L2,3 spectra across a line perpendicular to the edge of a m-BVO particle as schematically indicated by the solid dots in Figure 3b. The interval between two neighboring spectra is

[100] crystallographic direction is given in Figure 1b; it consists of rows of isolated VO4 tetrahedra separated by Bi atoms coordinated with eight O atoms to form BiO8 dodecahedra. Both Bi3+ and V5+ cations are displaced from the centrosymmetric sites along the same c direction, with the direction of the displacement alternating up and down along the c axis. This results in a layered-like structure of VO4 and BiO8 groups separated in nonequidistant planes. The layered structure is readily visible in the HAADF-STEM image of Figure 1c; due to the strong atomic-number contrast, the brighter dots correspond to Bi columns and the weaker ones are V columns. HAADF-STEM imaging further proves that the sample consists of monoclinic BVO, because in the tetragonal dreyerite phase the Bi3+ and V5+ cations are located in centrosymmetric sites in equidistant planes along c. Additionally, no structural modifications or reconstructions were detected at the edge of the displayed m-BVO particle or in any other studied particle. EELS is used to investigate the oxidation state of the vanadium ions in m-BVO by probing the vanadium L2,3 edge. Significant changes of the V-L2,3 and the O-K edge fine structures are observed at the surface of all scanned particles. Figure 2b shows the spectra acquired at the edge of the particle and at approximately 25 nm from the edge. In order to reduce beam damage, during acquisition of the experimental spectra the probe was scanned in a small area of about 4 nm2 around the red and white dots, respectively; see Figure 2a. For clarity, the spectra were background subtracted by fitting a decaying power-law function to an energy window just in front of the

Figure 3. (a) Vanadium L edge showing the L3 and the L2 white lines acquired across a line perpendicular to the surface of a m-BVO particle. For better comparison, the spectra were normalized to the intensity maxima of the L2 line. (b) HAADF-STEM image with the probing path indicated by the dotted line. (c) Energy shift of the V-L2,3 white lines across the probing path. Vertical bars indicate standard errors.

∼2.5 nm. Figure 3a shows the result of the scan across the first 24 nm from the surface. The energy shift of the V-L2,3 edge was obtained by fitting each white line to a Gaussian peak using a nonlinear least-squares routine. The result of this fit is depicted in Figure 3c. Interestingly, the energy loss of the V-L2,3 peaks is approximately the same for the first three surface spectra. An energy shift of ∼1.2 eV is observed between these surface spectra and the “bulk” (black) spectrum. On the other hand, the energy shift for the intermediate spectra is progressively reduced toward the center of the particle, as these spectra are a linear combination of surface and bulk contributions. Hence, the thickness of the reduction shell is estimated to be approximately 5 nm. The reproducibility of this result was

Figure 2. (a) HAADF-STEM image of the edge of a m-BVO particle imaged along [100] showing no structural modifications at the surface. (b) Vanadium L2,3 edge and oxygen K edge spectra taken from the areas indicated in panel a. For better comparison, the spectra were normalized to the intensity maxima of the L2 line. 3595



Article

BiV3.8+O4−δ□δ with δ = 0.6 (where □ represents an oxygen vacancy). Thus, the m-BVO structure incorporates ∼15% of oxygen vacancies at the surface of the particles. This is not surprising considering that the clinobisvanite phase adopts the scheelite structure, which is known to easily accommodate cations with various oxidation states27 and to tolerate a significant number of cation and oxygen vacancies.28,29 Moreover, recent calculations have predicted that oxygen vacancies in m-BVO have very low formation energies.16,19 We have compared our experimental spectra with theoretical ELNES to get a better insight into the energy shift of the V-L2,3 edge with the presence of oxygen vacancies. Here, the “surface” shell comprises the first 5 nm of the material, as evidenced by the experimentally measured 1.2 eV energy shift shown in Figure 3. In this context, it is acceptable to model the system by using a simplified bulk model structure and neglecting the effects of the broken symmetry at the surface. Figure 5 shows

confirmed in all studied grains regardless of the direction of the line scan. The oxidation state of the surface vanadium atoms is determined by using reference V-L2,3 spectra for V3+, V4+, and V5+. The reference spectra obtained from bulk V2O3, VO2, and V2O5 are shown together with the m-BVO spectra in Figure 4.

Figure 5. Calculated vanadium L2,3 spectra for m-BVO with and without oxygen vacancies. The V-L peak positions of the structures containing oxygen vacancies (represented by □) at the O1 and O2 positions are found at lower energies than those of the unmodified mBVO structure.

Figure 4. Comparison of the m-BVO spectra with the V2O3, VO2, and V2O5 reference spectra revealing that vanadium exists in multiple valence states on the surface of the m-BVO particles (red BVOedge spectrum).

the calculated V-L2,3 ELNES of the unmodified m-BVO structure (black spectrum) and of the oxygen vacancy models (red and blue spectra). In particular, either one O1 or one O2 atom was removed out of eight oxygen atoms (for details see the Experimental Section). Thus, both modified m-BVO structures contain 12.5% oxygen vacancies, very close to the ∼15% oxygen deficiency required to balance the vanadium reduction observed experimentally at the surface of the m-BVO particles. The calculated spectra of both reduced structures are almost identical and very similar to that of the unmodified mBVO (Figure 5). The main difference is found in the position of the V-L2,3 edges; the edges of the reduced structures are shifted to lower energies by ∼1.6 eV. Thus, considering the inherent limited precision of DFT calculations in predicting absolute values for energy losses, the theoretical calculations qualitatively reproduce the experimental data. To provide a basis for the interpretation of the EELS spectra, the projected density of states (pDOS) for the m-BVO structure with and without oxygen vacancies were calculated via DFT. Shown in Figure 6 are the pDOS for O-2p, V-3d, Bi-6s, and Bi-6p. The results of the unmodified m-BVO structure are in excellent agreement with previously reported theoretical results.30−32 Thus, the valence band is dominated by O-2p states between 0 and −5 eV. Hybridization with V-3d states is observed at −3.2 eV, while the top of the valence band is mainly composed of O-2p and Bi-6s states. Additionally, two additional peaks of primarily Bi-6s character are found between

It is found that by decreasing the vanadium oxidation state, the V-L3 peak position shifts toward lower energies by ∼0.8 eV (V5+ → V4+) and by ∼1.7 eV (V5+ → V3+). These results are in excellent agreement with previously reported values.24 Additionally, the position of the L2 and L3 peaks for the BVO bulk spectrum and the V2O5 spectrum are almost identical, indicating that the oxidation state of the bulk vanadium atoms is +5. On the other hand, the L3 and L2 edges of the BVO edge spectrum appear at 515.2 and 521.8 eV, respectively. These values are found to be between those of V2O3 and VO2, implying that vanadium atoms are in the form of V3+ and V4+ at the surface of the m-BVO particles. The oxidation state of the surface vanadium atoms is estimated by fitting a linear regression to the experimental peak positions of all three reference spectra over the vanadium oxidation state. The linear regression produced the following equation of the least-squares line: EV−L3 = 0.85x + 511.97 (where x is the vanadium oxidation state and EV−L3 is the peak position), with a regression coefficient R2 of 0.9988. Thus, the average oxidation state x of the surface vanadium atoms is +3.8, and the fractional contributions of V3+ and V4+ to the surface spectra are determined as 0.2 and 0.8, respectively. The oxygen vacancy concentration δ is directly related to the vanadium oxidation state x by δ = (5 − x) /2. This results in a surface composition 3596



Article

Figure 6. Projected electron density of states for all three atoms without oxygen vacancies (top) and with 12.5% O1 oxygen vacancies (bottom). In both panels, the vertical dashed line at zero-energy marks the Fermi level. The inset shows an enlarged view of the donor states formed near the Fermi level. V1 and Bi1 indicate the vanadium and bismuth atoms located closer to the O1 vacancy, while V2 indicates the vanadium atom located far from the O1 vacancy. The different atom contributions are indicated by distinct colors.

Figure 7. Spatial representation of the charge density plot constructed from the donor band corresponding to the electronic states induced by an O1 vacancy in m-BVO. Regions with an excess of charge and O1 vacancies are shown as yellow lobes and spheres, respectively.

6p states. More importantly, the present results indicate a charge localization of the two defect electrons on an ideal bond joining the vanadium (V1) neighboring the O1 vacancy and the bismuth (B1) atom at the other side (see Figure 7). Additionally, by comparing the bond distances of the pristine bulk structure with those of the reduced structure, we observe that the rearrangement following the O1 removal is shown by a significant displacement of the V1 neighbor away from the vacancy site with a shortening of the V1−O1 distance and an increase of the V1−O2 distances (see Table 1).

−9 and −10 eV. The conduction band is found to be dominated by V-3d states, with significant contributions from O-2p and Bi-6p. When comparing the pDOS for the structure with no vacancies and with 12.5% vacancies, it is evident that the V-3d conduction states have shifted by ∼1.5 eV toward the Fermi level. A shift of ∼2 eV is also observed for all occupied DOS away from the Fermi level. Additionally, in the unmodified m-BVO, a prominent resonance between the O2p and V-3d bands is evident at E − EF = −3.2 eV. Such resonance is still present in the BVO structure with vacancies (E − EF = −4.5 eV) but is depleted due to the oxygen vacancy and subsequent electric distribution. More interestingly, oxygen vacancies give rise to new localized donor states close to the Fermi level, about 0.64 eV below the conduction band. These defects are shallow donors as determined by previous transition-energy calculations.16 They are dominated by V-3d states, implying that vanadium is reduced and contains more occupied d states as compared to the unmodified structure. We decomposed the states near the Fermi level (see inset in Figure 6) in order to find out the contribution of each atom. The largest contribution to the new localized donor states arises from the vanadium atom (V1) located closer to the O1 vacancy, and also from Bi1−6p states (where Bi1 is the bismuth atom located closer to the O1 vacancy). The other states (V2− 3d and O1−2p) have a negligible impact. Thus, our calculations support the reduction of vanadium to a lower oxidation state. To explicitly address the nature of the localized states observed in the density of states, we have computed the charge density restricted to an energy window containing these states. These valence charge density maps are used to generate a visual representation of a given selection of bands and are computed using the lapw5 package of WIEN2k. This package is used to obtain 2D slices from the electronic density file calculated by WIEN2k. The 3D potential can then be extracted by combining the calculated 2D slices using the wien2venus script written by Masao Arai.33 In Figure 7, we depict the charge density plot constructed from the donor band corresponding to the electronic states induced by a O1 vacancy in m-BVO. The localized states corresponding to this particular valence band mostly have V1−3d characteristics, with participation of Bi1−

Table 1. Selected Interatomic Distances for the BVO Structures with and without Vacancies no vacancy interatomic distance (Å) V···O1 1.767 V···O2 1.691

12.5% vacancies interatomic distance (Å) V1···O1 1.728 V1···O2 1.722 1.719 V2···O1 1.767 1.786 V2···O2 1.747 1.685

In order to find out the formal charge state of the vanadium ion close to the oxygen vacancyand to compare it with the vanadium charge state in the bulk m-BVO structurewe used the topological analysis of the electron density introduced by Bader.34,35 One should be aware that Bader charges do not match the formal charge state. For example, the Bader charge of vanadium in the bulk m-BVO structure is calculated to be +2.186 |e| (see Table 2). Thus, a more meaningful result is obtained by using reference states. For this reason, Bader charges for V3+, V4+, and V5+ were calculated by using the AIMS module of WIEN2k from three reference vanadium oxides, namely V2O3, VO2, and V2O5. Then, the same analysis was carried out for the unmodified bulk m-BVO structure and for BVO with vacancies (see Table 2). Surprisingly, although the structural environment of the V1 and V2 ions in the BVO structure with vacancies is very different, the values for the Bader charge are almost the same for V1 and V2 (+2.07 and 3597



Article

Table 2. Bader Charges of Vanadium Atoms in Three Reference Vanadium Oxides, and in BVO with and without Vacancies vanadium atom

Bader charge (|e|)

reference vanadium oxide structures V3+ (V2O3) V4+ (VO2) V5+ (V2O5) bulk m-BVO structure V5+ BVO structure with 12.5% oxygen (O1) V1 V2 BVO structure with 12.5% oxygen (O2) V1 V2

+1.81 +2.05 +2.21 +2.186 vacancies +2.07 +2.09 vacancies +2.09 +2.07

Figure 9. O 1s and V 2p spectra of pure BVO powder (top) and powder mixed with Au nanopowder (bottom) after energy scale correction using the oxygen peak (top) and the internal Au reference (bottom), respectively.

+2.09 |e|), and very close to the VO2 reference (+2.05 |e|), thus supporting the experimentally measured +3.8 vanadium oxidation state. The surface chemical composition and oxidation state of each element were also determined by XPS. Figure 8 shows

156.9 eV37] and lattice O2−. However, an oxidation state less than +5 is observed for vanadium [E(V2O5 2p3/2) = 517.4 eV,38 E(VO2 2p3/2) = 516.4 eV,39 E(V2O3 2p3/2) = 515.7 eV,40 and E(V 2p3/2) = 512.16 eV37]. Thus, the XPS measurements unambiguously confirm the surface reduction in m-BVO observed by STEM-EELS ruling out electron beam induced radiation effects for this electron sensitive material. Additionally, by comparing the concentrations derived from the peak intensities, a pronounced excess of bismuth is detected at the surface of the particles. A concentration of 23.1 atom. %, 14.4 atom. %, and 62.5 atom. % is found for Bi, V, and O, respectively. Similar to the energy positions, these concentrations do not change upon heating. The measured O/V ratio of around 4.3:1 is close to the expected bulk ratio, but the Bi concentration is nearly twice as high as expected. In some Bicontaining oxides, similar segregation phenomena were previously observed. For example, bismuth segregation with formation of Bi2O3 has been detected at the surface of bismuth−tin pyrochlores,41 Bi2MoO6,42 and BiFeO3 epitaxial films.43 In analogy to these systems, one can explain the depletion of oxygen in the subsurface via the segregation of O to the surface, which leads to a termination of the surface by bismuth and oxygen:

Figure 8. High resolution XPS spectra of pure BVO powder (inset) and powder mixed with Au nanopowder (black and red overview spectra). The energy scales of the spectra are uncorrected. The inset shows the splitting of the Bi 4f peaks of the pure BVO powder. This splitting does not occur in the overview spectra of BVO with the conductive medium (Au) demonstrating that it originates not from a change in oxidation states but from a charge shift, which depends on the sample preparation, i.e., heating.

BiVO4 ... BiVO4 − δ ... Bi−O

Such an arrangement explains the measured intensity ratios, as XPS is limited to the first atomic layers only. This effect is even more enhanced on powder samples. Furthermore, such a termination is stable in an oxidizing atmosphere, because the topmost layer is oxygen, while the subsurface vanadium ions are reduced. The surface termination by bismuth oxide might also contribute to the relatively low catalytic activity of BiVO4 as an oxygen evolution catalyst.44

high resolution XPS overview spectra of BiVO4 mixed with Au nanopowder as prepared and heated in UHV at 160 °C. The spectra reveal huge shifts equal for all peaks, which are thus attributed to charging of the sample. The powder sample without the conductive medium even displays various charge shifts on the same sample (see inset in Figure 8). Clearly, these shifts are measurement artifacts and do not hint toward a change of the oxidation state, of, here, e.g., Bi. When the energy scale is recalibrated using the internal Au reference, it becomes evident that, there is no significant chemical shift (Figure 9) upon heating, revealing the following binding energies: E(Bi 4f7/2) = 159.1 eV, E(V 2p3/2) = 516.7 eV, and E(O 1s) = 529.9 eV, in good agreement with earlier work.36 The energies are indicative for Bi3+ [E(Bi2O3 4f7/2) = 159.0 eV,36 E(Bi 4f7/2) =

■

CONCLUSIONS In summary, the presence of a 5-nm-thick reduction shell in mBVO particles was revealed using STEM-EELS. A shift of the V-L2,3 edge to lower energies was observed at the surface of the particles. On the basis of reference measurements of three VnOm compounds, we showed that the surface vanadium ions are reduced from V5+ to V3.8+, demanding for the presence of a 3598



■

ACKNOWLEDGMENTS We are grateful to K. Jorissen for assistance and fruitful discussions about WIEN2k. We thank M. Walter and M. V. Kovalenko for the XRD measurements. This research was supported by the Swiss National Science Foundation under the project number 200021_147105 and by the University Research Priority Program (URPP) for solar light to chemical energy conversion (LightChEC).

significant amount of oxygen vacancies. However, no structural modifications associated with the electronic reconstructions at the surface of the particles were detected by our HRSTEM observations. From the WIEN2k-based density functional theory calculations, we have shown that the presence of oxygen vacancies has a direct impact in the electronic structure of m-BVO, resulting in clear differences in the ELNES and pDOS. In particular, our calculations show that oxygen vacancies in mBVO give rise to localized donor states about 0.64 eV below the conduction band. These defects act as shallow donors and are responsible for the n-type conductivity, as previously suggested by Yin et al.16 Thus, the presence of a reduction shell in the studied m-BVO particles creates an n+−n homojunction, which is known to enhance the charge-separation efficiency.44,45 However, as opposed to previous results on homojunctions in m-BVO,44,45 in the present case the homojunction is reversed, i.e. the bulk (core) m-BVO is a n-type semiconductor and the reduction shell is an n+-type semiconductor (Figure 10). Thus,

■

REFERENCES

(1) Kudo, A.; Ueda, K.; Kato, H.; Mikami, I. Catal. Lett. 1998, 53, 229. (2) Kudo, A.; Omori, K.; Kato, H. J. Am. Chem. Soc. 1999, 121, 11459. (3) Zhou, Y.; Vuille, K.; Heel, A.; Probst, B.; Kontic, R.; Patzke, G. R. Appl. Catal., A 2010, 375, 140. (4) Kho, Y. K.; Teoh, W. Y.; Iwase, A.; Mdler, L.; Kudo, A.; Amal, R. ACS Appl. Mater. Interfaces 2011, 3, 1997. (5) Kohtani, S.; Hiro, J.; Yamamoto, N.; Kudo, A.; Tokumura, K.; Nakagaki, R. Catal. Commun. 2005, 6, 185. (6) Fujishima, A.; Honda, K. Nature 1972, 238, 37. (7) Qurashi, M. M.; Barnes, W. H. Am. Mineral. 1953, 38, 489. (8) Granzin, J.; Pohl, D. Z. Kristallogr 1984, 169, 289. (9) Sleight, A. W.; Chen, H. Y.; Ferretti, A. Mater. Res. Bull. 1979, 14, 1571. (10) Tokunaga, S.; Kato, H.; Kudo, A. Chem. Mater. 2001, 13, 4624. (11) Payne, D. J.; Robinson, M. D. M.; Egdell, R. G.; Walsh, A.; McNulty, J.; Smith, K. E.; Piper, L. F. J. Appl. Phys. Lett. 2011, 98, 212110. (12) Luo, W.; Yang, Z.; Li, Z.; Zhang, J.; Liu, J.; Zhao, Z.; Wang, Z.; Yan, S.; Yu, T.; Zou, Z. Energy Environ. Sci. 2011, 4, 4046. (13) Zhou, B.; Zhao, X.; Liu, H.; Qu, J.; Huang, C. P. Appl. Catal., B 2010, 99, 214. (14) Sayama, K.; Nomura, A.; Arai, T.; Sugita, T.; Abe, R.; Yanagida, M.; Oi, T.; Iwasaki, Y.; Abe, Y.; Sugihara, H. J. Phys. Chem. B 2006, 110, 11352. (15) Zhang, K.; Shi, X. J.; Kim, J. K.; Park, J. H. Phys. Chem. Chem. Phys. 2012, 14, 11119. (16) Yin, W. J.; Wei, S. H.; Al-Jassim, M. M.; Turner, J.; Yan, Y. Phys. Rev. B 2011, 83, 155102. (17) Abdi, F. F.; Savenije, T. J.; May, M. M.; Dam, B.; van de Krol, R. J. Phys. Chem. Lett. 2013, 4, 2752. (18) Jiang, H.; Dai, H.; Meng, X.; Ji, K.; Zhang, L.; Deng, J. Appl. Catal., B 2011, 105, 326. (19) Wang, G.; Ling, Y.; Lu, X.; Qian, F.; Tong, Y.; Zhang, J. Z.; Lordi, V.; Rocha Leao, C.; Li, Y. J. Phys. Chem. C 2013, 117, 10957. (20) Blaha, P.; Schwarz, K.; Madsen, G.; Kavasnicka, D.; Luitz, J. WIEN2k, An Augmented Plane Wave + Local Orbitals Program for Calculating Crystal Properties; Schwarz, K., Ed.; Technische Universität: Wien, Austria, 2001. (21) Hébert, C. Micron 2007, 38, 12. (22) Nelhiebel, M.; Louf, P.-H.; Schattschneider, P.; Blaha, P.; Schwarz, K.; Jouffrey, B. Phys. Rev. B 1999, 59, 12807. (23) Hébert-Souche, C.; Louf, P. H.; Blaha, P.; Nelhiebel, M.; Luitz, J.; Schattschneider, P.; Schwarz, K.; Jouffrey, B. Ultramicroscopy 2000, 83, 9. (24) Fitting Kourkoutis, L.; Hotta, Y.; Susaki, T.; Hwang, H. Y.; Muller, D. A. Phys. Rev. Lett. 2006, 97, 256803. (25) Chi, M.; Mizoguchi, T.; Martin, L. W.; Bradley, J. P.; Ikeno, H.; Ramesh, R.; Tanaka, I.; Browning, N. J. Appl. Phys. 2011, 110, 04604. (26) Wadati, H.; Hawthorn, D. G.; Geck, J.; Regier, T. Z.; Blyth, R. I. R.; Higuchi, T.; Hotta, Y.; Hikita, Y.; Hwang, H. Y.; Sawatzki, G. A. Appl. Phys. Lett. 2009, 995, 023115. (27) Morozov, V. A.; Arackcheeva, A. V.; Chapuis, G.; Guiblin, N.; Rossell, M. D.; Van Tendeloo, G. Chem. Mater. 2006, 18, 4075. (28) Morozov, V. A.; Mironov, A. V.; Lazoryak, B. I.; Khaikina, E. G.; Basovich, O. M.; Rossell, M. D.; Van Tendeloo, G. J. Solid State Chem. 2006, 179, 1183.

Figure 10. Schematic energy bands of the separate phases (left) and after the formation of the n+−n homojunction (right). The valence band, conduction band, and Fermi level are indicated as VB, CB, and EF, respectively. When BiVO4 and BiVO4−δ are brought into contact, the Fermi levels of both phases align, causing a bending of the energy bands.

in this situation an inner electrical field is established in the direction of the n+-type to the n-type semiconductor (from the shell to the core of the BVO particles). As a result, the energy bands of the core BVO would shift upward while those of the reduced BVO shell shift downward in order to align the Fermi levels of both semiconductors. At this band edge position, the photogenerated electrons of the conduction band of the n-type core transfer to that of the n+-type shell, and simultaneously, holes on the valence band of the n+-type shell move to that of the n-type core under the influence of the potential setup by the band energy difference. As a consequence, the oxygen evolution reaction at the surface of the m-BVO particles is expected to be greatly reduced. Thus, our findings suggest that the photochemical behavior of “pristine” m-BVO needs to be interpreted in terms of a distinct surface shell of different properties than the actual bulk BVO. This result emphasizes the importance of surface and grain boundary effects in this new candidate for photoelectrochemical energy conversion and wastewater purification.

■

Article

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest. 3599



Article

(29) Babin, V.; Bohacek, P.; Krasnikov, A.; Nikl, M.; Stolovits, A.; Zazubovich, S. J. Lumin. 2007, 124, 113. (30) Zhao, Z.; Li, Z.; Zou, Z. Phys. Chem. Chem. Phys. 2011, 13, 4746. (31) Ding, K.; Chen, B.; Fang, Z.; Zhang, Y. Theor. Chem. Acc. 2013, 132, 1352. (32) Cooper, J. K.; Gul, S.; Toma, F. M.; Chen, L.; Glans, P.-A.; Guo, J.; Ager, J. W.; Yano, J.; Sharp, I. D. Chem. Mater. 2014, 26, 5365. (33) Arai, M. wien2venus.py. Available from: http://www.nims.go.jp/ cmsc/staff/arai/wien/venus.html. (34) Bader, R. F. W. Atoms in Molecules: A Quantum Theory; Oxford University Press: New York, 1990. (35) Henkelman, G.; Arnaldsson, A.; Jonsson, H. Comput. Mater. Sci. 2006, 36, 354. (36) Schuhl, Y.; Baussart, H.; Delobel, R.; Bras, M. L.; Leroy, J.; Gengembre, L. G.; Rimblot, J. J. Chem. Soc., Faraday Trans. I 1983, 79, 2055. (37) Powell, C. J. J. Electron Spectrosc. Relat. Phenom. 2012, 185, 1. (38) Bond, G. C.; Flamerz, S. Appl. Catal. 1989, 46, 89. (39) Kasperkiewicz, J.; Kovacich, J. A.; Lichtman, D. J. Electron Spectrosc. Relat. Phenom. 1983, 32, 128. (40) Colton, R. J.; Guzman, A. M.; Rabalais, J. W. J. Appl. Phys. 1978, 49, 409. (41) Bhattacharya, A. K.; Forster, S. F.; Pyke, D. R.; Mallick, K. K.; Reynolds, R. J. Mater. Chem. 1997, 7, 837. (42) Galván, D. H.; Castillón, F. F.; Gómez, L. A.; Avalos-Borja, M.; Cota, L.; Fuentes, S.; Bartolo-Pérez, P.; Maple, M. B. React. Kinet. Catal. Lett. 1999, 67, 205. (43) Béa, H.; Bibes, M.; Barthélémy, A.; Bouzehouane, K.; Jacquet, E.; Khodan, A.; Contour, J.-P.; Fusil, S.; Wyczisk, F.; Forget, A.; Lebeugle, D.; Colson, D.; Viret, M. Appl. Phys. Lett. 2005, 87, 072508. (44) Abdi, F. F.; Han, L.; Smets, A. H. M.; Zeman, M.; Dam, B.; van de Krol, R. Nat. Commun. 2013, 4, 2195. (45) Han, M.; Sun, T.; Tan, P. Y.; Chen, X.; Tan, O. K.; Tse, M. S. RSC Adv. 2013, 3, 24964.

3600


Direct Evidence of Surface Reduction in ... - ACS Publications

Recommend Documents