Electronic Structure of Pure and N-Doped TiO2 Nanocrystals by ...

ARTICLE pubs.acs.org/JPCC

Electronic Structure of Pure and N-Doped TiO2 Nanocrystals by Electrochemical Experiments and First Principles Calculations Francesca Spadavecchia,* Giuseppe Cappelletti, Silvia Ardizzone, Michele Ceotto, and Luigi Falciola Dipartimento di Chimica Fisica ed Elettrochimica, Universita degli Studi di Milano, Via Golgi 19, 20133 Milano, Italy ABSTRACT: Optical and electrochemical characterizations are carried out in conjunction with first-principles calculations on pure and N-doped titania nanocrystals. These are prepared in laboratory with initial doping concentrations of triethylamine in the range of 0.1-0.5 N/Ti molar ratio. Diffuse reflectance UV-vis spectra of N-doped samples present a significant absorption in the visible region. The flatband potential (Efb) of pure and nitrogen-doped TiO2 (-0.6 ( 0.2 V vs NHE) is determined by impedance spectroscopy (Mott-Schottky plots) and the quasi-Fermi level, nEF* (-0.67V vs NHE) by photovoltage measurements as a function of the suspension pH in the presence of an electrochemical probe (methylviologen, MV2þ). Theoretical density of electronic states calculations, where several N doping versus vacancy combinations are taken into consideration, together with the optical and electrochemical experiments allow us to draw a detailed picture of the electronic features of the doped samples.

1. INTRODUCTION Titanium dioxide is considered one of the most promising photoactive semiconductors, exhibiting a strong oxidation as well as a moderate reduction potential. Thanks to its chemical stability, low-cost, and nontoxicity, it has been extensively used in photovoltaic solar cells, photoelectrochemical devices, and photocatalytic decontamination of aqueous and atmospheric environment.1 Along with this, the search for semiconducting materials catalyzing efficient visible-light photoredox reactions has become a pressing issue. Doping titania with non metals has attracted considerable attention due to a red-shift of the light absorption edge. The N-doped TiO2, obtained by addition of different nitrogen sources (ammonia, ammonium salts, urea, amine, etc.),2 seems to be very promising among the so-called second generation photocatalysts, which are designed to overcome the intrinsic limitations of pure TiO2 of having a quite large band gap (3.0-3.2 eV). A lively debate on the causes that provoke the shift of the absorption onset has arisen and grown. The main point of discussion concerns the eventual narrowing of the semiconductor band gap as a consequence of the doping or the creation of intra gap states.3 Also, the chemical nature and the location in the solid of the nitrogen species responsible for visible light activity is still controversial. N atoms or even other complex species such as NO, NO2, or NHx may occupy substitutional and/or interstitial sites in the TiO2 lattice, but it has not been easy to unequivocally identify them.4 In this work, we report on TiO2 samples obtained by a sol-gel synthesis using triethylamine as nitrogen organic source. This specific route turned out to be successful in producing samples with increased photocatalytic activity and significant presence of bulk nitrogen paramagnetic Nb• species.5 A few other authors6 have reported the fortunate doping by triethylamine. r 2011 American Chemical Society

According to literature theoretical calculations, the reasons for such an improved photoactivity should be attributed to the fact that nitrogen doping leads to a significant reduction of the energy cost to form oxygen vacancies in bulk titanium dioxide.7 This suggests that such doping is very likely to be accompanied by the presence of oxygen vacancy formation, which can be, however, already present in the undoped oxide in a certain amount, and even more if nanosized. In which ratio and how oxygen vacancies interact with nitrogen doping centers is still unclear, even if some possible scenarios have been put forward.8 The main open question is about the role of oxygen vacancies in the visible light sensitivity of both doped and undoped TiO2. However, the improved photocatalytic activity cannot be ascribed only to the presence of oxygen vacancies, because it is a complex balance among several factors, including electron-hole recombination rate and morphological/structural properties of the material. More generally, the removal of one neutral lattice oxygen atom leaves two extra electrons (assuming a formal -2 oxidation state of O in TiO2) which fill empty states of Ti ions. Then, there are several possibilities and the actual situation is quite complex because a point defect and three undercoordinated (5-fold) Ti ions are formed. The two electrons associated to the defect may be both delocalized on several Ti ions or localized on single (different) Ti ions; of course, intermediate situations where one electron is localized and the second is not are also possible. However, general experimental agreement has been reached on interpreting a band gap feature at about 1 eV below the conduction band edge in terms of occupied Ti3þ 3d states.9 The fate of these two extra electrons in the presence of either substitutional or Received: January 14, 2011 Revised: February 15, 2011 Published: March 10, 2011 6381

dx.doi.org/10.1021/jp2003968 | J. Phys. Chem. C 2011, 115, 6381–6391

The Journal of Physical Chemistry C interstitial N doping is even more unclear, because one should consider several vacancies and N doping scenarios. In order to understand these electronic aspects, optical characterizations of TiO2 features alone are not enough because limited to “apparent” band gap evaluation. Other electronic aspects, such as absolute scale conduction/valence band location, excited electronic states, and Fermi levels go undetected under such kinds of experiments. For these reasons, in this work we propose to combine optical measurements with direct electrochemical characterizations of the electronic structure of the semiconductor,10 such as impedance and photovoltage measurements. At the same time, theoretical DFT (Density Functional Theory) calculations of DOS (Density of electronic States) are performed to shed light on the electrochemical picture and provide a comprehensive electronic framework. This work is not intended for a direct comparison between the absolute values derived from theoretical calculations and those obtained from electrochemical measurements but for a trend-comparison between the flatband potential of the samples and the relative first-principles Fermi levels at varying initial nitrogen content. As far as the above-mentioned experimental applications are concerned, a proper understanding of the surface chemistry at the oxide/liquid interface and a better clarification of the nitrogen role in the physical features of the semiconductor are needed. Aiming at these goals, the specific electrochemical characterizations were carried out, at the same time, on commercial and homemade titania nanopowders in order to evaluate both their flatband potential (the potential at which no band bending is present) and their quasi-Fermi (excited electronic states) potential of electrons. A pictorial representation of these different electrochemical techniques is represented in Figure 1, on the grounds of literature data.11 On panel (a), the ordinary electronic structure of the TiO2 is reported. This is an n-type semiconductor because the vacancies act as electron donors. When a Mott-Schottky measurement is performed, the semiconductor is in contact with a solution where a redox couple is present. Then, an exchange of electrons can occur in order to match the two Fermi energies (Ef and Eredox) as depicted on panel (b) of the same figure. This electron flow induces a band bending due to the pinning of the lower edge of the conduction band at the pristine value. Now, the flatband potential, Efb, that can be measured by an MS plot is the counter-potential that flats the potential back again by compensating the migrated charges capacity effect. Thus, the flatband potential represents the lower edge of the conduction band, which is approximately the value of the Fermi energy, on an electrochemical scale. On panel (c) of Figure 1, the photovoltage technique is schematically represented. Here, an initial light irradiation induces the separation of electron-hole pairs. The associated energy states are called quasi-Fermi energy levels and are located just below the conduction band for the electrons and just above the valence band for the holes. Because of the protolitic equilibrium between the semiconductor and the aqueous solution taking place at the titanium oxide interface, the electronic levels of the semiconductor can be rigidly shifted by varying the pH of the solution. Now, once the pH is high enough to guarantee that the electron quasi-Fermi energy level is more negative (on the electrochemical scale) than the one of a pHindependent redox couple in solution, the electrons will flow from the semiconductor to the solution. Thus, a sudden voltage change versus the pH will be observed in a titration curve fashion. From the pH value at the inflection point (pH0) one can calculate

ARTICLE

Figure 1. Pictorial representation of the electronic energy levels for titania: (a) for the gas phase titania where CB is the conduction band energy, VB the valence band energy, Ef the Fermi energy and ECB the conduction band edge; (b) for titania in contact with a solution where a redox couple is dissolved. Ox stands for oxidated form, Red for reduced form and Eredox is the redox potential. The depletion region is the one where the band is bended. (c) The photoexcitation process and the electron-hole pair energy levels are represented. Ef (e-) indicates the electron quasi-Fermi energy levels, while Ef (hþ) the hole one. Three different pH scenarios are reported (see the following).

the quasi-Fermi electron energy levels as it will be reported in eq 2. In the following, we will use both the “quasi-Fermi level” and “flatband potential” terms according to the techniques used, though the former would be more accurate to describe both determinations for small crystal size (about 5 nm, in the case of our homemade titania, both pure and N-doped) which presumably does not produce a relevant band-bending. Only in conjunction with the theoretical DOS calculations, the electrochemical measurements will allow us to verify whether a shift of the quasi-Fermi states versus the conduction band edge or a shift of the Fermi energy can unveil a deeper insight into the improved photocatalytic properties of N-doped titania samples.

2. EXPERIMENTAL SECTION Pure and N-doped titania samples were synthesized by a sol-gel route as described elsewhere;5 titanium(IV) isopropoxide was used as a TiO2 precursor and triethylamine as a nitrogen source, with different initial N/Ti molar ratios in the range 0.1-0.5. The samples were finally calcined at 400 C under O2 stream. Two different commercial samples were used for the sake of comparison: HOMBIKAT UV100 (Sachtleben), characterized by large surface area and P25 (Degussa), with surface area much lower than the homemade samples. XRPD spectra were collected on the ID31 high-resolution powder diffraction beamline at the European Synchrotron Radiation Facility (ESRF). The experimental conditions are reported in our previous work.5 Specific surface areas were determined by 6382

dx.doi.org/10.1021/jp2003968 |J. Phys. Chem. C 2011, 115, 6381–6391

The Journal of Physical Chemistry C the classical BET procedure using a Coulter SA 3100 apparatus. Diffuse reflectance spectra of the powders were measured on a Perkin-Elmer, Lambda 35 spectrophotometer, equipped with a diffuse reflectance accessory. HRTEM measurements were performed on a JEOL JEM-2100URP microscope equipped with a Gatan Ultrascan 1000 CCD camera (2048 2048 pixels), operating at 200 kV. X-ray photoelectron spectra (XPS) were performed with a Perkin-Elmer PHI ESCA 5500 equipped with a Al KR X-ray source (1486.6 eV). The energy resolution of the spectrometer was set at 0.1250 eV. 2.1. Electrochemical Setup. The flat-band potentials were measured by impedance spectroscopy using the Mott-Schottky plots. The measurements were performed in a conventional thermostatted (25 C) three-electrode cell, with a 2 cm2 platinum flag as counter electrode and a saturated calomel electrode as reference. The working electrode was a TiO2 thin film prepared as follows: the TiO2 powder was well dispersed with isopropanol and the suspension was applied on an indium-doped tin oxide conducting glass (ITO), 2 1 cm2, by drop casting. After being dried, the film was finally annealed at 400 C for 1 h. The experiment was performed in aqueous 0.5 M Na2SO4 solution at pH ≈ 6. The potential was systematically varied between þ1.7 and -1.3 V (vs NHE) with the frequency range being modulated between 500 to 2000 Hz by an ECO-CHEMIE Autolab PGStat 30 - Potentiostat Galvanostat equipped with Frequency Response Analyzer (FRA).12 The quasi-Fermi level of electrons was also measured according to the literature13 using methylviologen dichloride ((MV)Cl2, Ered (MV2þ/MVþ) = -0.4421 V vs NHE) as a pH-independent redox system. Thus the observation of a pH effect on the measured photovoltage can be considered as related only to changes in the bulk or at the surface of the oxide. A two-electrode cell using a 2 cm2 platinum flag and a saturated calomel electrode (SCE) as working and reference electrodes, respectively, was adopted. A combined glass electrode was also employed for pH measurements. Thirty milligrams of semiconductor powder were suspended and sonicated in 50 mL of 0.l M KNO3, and then placed in a thermostatted (25 C) cell. After being degassed with N2 for 0.5 h, 6 mg of methylviologen dichloride was added to the suspension, and again degassed for about 15 min. The pH of the suspension was adjusted to pH 1 using 1 M HNO3 and then it was raised by adding NaOH solutions. Stable photovoltages were recorded about 30 min after changing the pH value. Magnetic stirring and nitrogen flow were kept constant during the measurement. The light source was a 500 W UV halogen lamp (Jelosil HG 500, iron halides, 315-400 nm). The cell potential differences between the working electrode and the reference SCE were recorded with a KEITHLEY 619 differential Electrometer/ Multimeter, with an input impedance greater than 1014 Ω. The precision of potential difference measurements was (0.01 mV. For pH measurements, an AMEL 338 pH-meter was used, after appropriate calibration. 2.2. Computational Setup. All of the spin-polarized calculations were performed using the projector augmented wave (PAW) pseudopotentials to treat the valence-core interactions. The core of Ar for Ti, He for O, and He for N were employed, as implemented in the VASP code.14 The Perdew-Burke-Ernzerhof parametrization15 of the generalized gradient approximation16 was adopted for the exchange-correlation potential. The valence electron wave function was expanded in plane waves basis set up to a cutoff energy of 400 eV. Forces on the ions were calculated

ARTICLE

through the Hellmann-Feyman theorem as the partial derivatives of free energy with respect to the atomic position, including the Harris-Foulkes correction to forces.17 This calculation of the forces allows geometry optimization using the conjugategradient scheme.18 Iterative relaxation of atomic positions was stopped when the change in total energy between successive steps was less than 0.001 eV. Electronic property calculations were carried out using the block Davidson scheme.19 The supercell and atomic relaxations were carried out until the residual forces were below 0.01 eV/Å. Since the excess charge localization and the strongly correlated nature of the occupied Ti 3d states seriously challenges DFT calculations, the electronic structure calculations were conducted using both the GGA and the GGA þ U method,20 which can lead to a good description for the TiO2, as shown in previous works.21 The DFT þ U approach introduces an on-site correction in order to describe systems with localized d and f electrons, which can produce better band gaps and midgap states energetics in comparison with experimental results. Here, effective on-site Coulombic interactions U (U = U0 - J) for Ti 3d were used. U0 and J represent the energy cost of adding an extra electron at a particular site and the screened exchange energy, respectively. A value of U = 5 eV was used, which has previously been shown to properly account for the electronic structure of the Ti 3d states.21 The bulk doped systems were constructed from the relaxed 3 3 3 162-atom anatase TiO2 supercell. The optimized supercell lattice parameters were a = 11.392 Å and c = 28.606 Å (a = 3.797 Å and c = 9.535 Å for a primitive cell), in good agreement with experimental results.22 Both of these results indicate that our computational approach is reasonable. A variety of positions of N atoms in the TiO2 lattice were considered, such as substitutional N (N@O) and several interstitial N (Nint) geometries. Reciprocal space sampling was restricted to the Γpoint, which is justified due to the rather large size of the used simulation supercells. In our theoretical calculations, no border effects are taken into account. Then, such theoretical calculations are better suited to describe a minimized surface/volume situation, as spherical particles.

3. RESULTS AND DISCUSSION In order to describe the relations between electronic structure and photochemical properties, different experimental approaches are presented. The experimental characterizations were performed on both commercial and homemade samples. The latter were prepared by sol-gel route adopting triethylamine as a nitrogen source. A final calcination step at 400 C was introduced to promote crystallinity and lattice incorporation of N species. The specific surface areas and crystallite sizes are reported in Table 1 (3rd and fourth columns): the surface areas progressively decrease by increasing the N content in the synthesis, while the crystallite sizes are almost invariant for all homemade samples (5-6 nm). These results are confirmed by the high-resolution transmission electron micrographs presented in Figure 2 for the undoped (T) and two different N-doped samples (TN_0.10 and TN_0.50). In each case, the various domains are well crystallized, and several dislocations can be observed. At increasing the nitrogen content, the shape of the particles changes from spherical to prismatic type. All of the samples are characterized by a presence of anatase/ brookite mixture; the anatase percentage progressively increases 6383


The Journal of Physical Chemistry C

ARTICLE

Table 1. Specific Surface Areas (BET Analyses), Average Diameter of the Anatase Crystallites (XRPD), Surface N/Ti Atomic Ratios (XPS) and “Apparent” Band-Gap Values (DRS) for Commercial and Homemade Samples ÆDanatæ (nm)

initial N/Ti molar ratio

SBET (m2g-1) 171

4.8 ( 0.1

TN_0.10 TN_0.20

0.10 0.20

120 98

4.9 ( 0.1 5.7 ( 0.1

0.022 0.022

2.98 2.94

TN_0.40

0.40

90

5.8 ( 0.1

0.018

2.88

TN_0.50

0.50

87

5.9 ( 0.1

0.019

P25

50

30.0 ( 0.3

3.30

HOMBIKAT UV100

354

10.0 ( 0.3

3.34

sample T

from 75% in the case of undoped T sample to around 90% for the highest initial nitrogen amount.5 Figure 3a,b reports, respectively, the XPS Ti 2p and N 1s peaks for a representative sample, TN_0.40. The region of Ti 2p shows the classical doublet typical of Ti4þ in the oxide.23 In the case of N 1s the single component at 399.5 eV can be attributed to interstitial N in the oxide.4 N/Ti atomic ratios evaluated by the fitting of the XPS spectra are apparently invariant for the different samples (see Table 1, fifth column). This lack of variation can be the result, especially for the low N content samples, of the high noise to signal ratio of the N 1s peak. Further, XPS data are to be considered as only representative of the outer shell of the nanoparticles; however, the decrease of the specific surface area at increasing N content confirms the presence of the dopant, in different amounts, acting as a sintering center in the nanoparticle. 3.1. Optical Measurements. Experimental data of diffuse reflectance (DR) UV-vis spectra are elaborated by integrating the absorbance values—I(A) —to better highlight the different optical features of the catalysts (Figure 4). It is clear that all curves of the N-doped samples lie above the one of the pristine oxide when λ > 400 nm and hν > ∼2.5 eV; this fact indicates a significant absorption in the visible region, which is the typical feature of doped materials. Note that the TN_0.40 and TN_0.50 curves totally overlap. The higher the starting N amount in the samples, the more the absorption in the visible range. In the inset of Figure 4 the experimental curves (%R vs λ) are converted to absorption coefficient values F(R) according to the Kubelka-Munk equation24 and plotted as [F(R)]0.5 against the photon energy. The marked tails in the low energy (visible) region of the N-doped sample curves (Figure 4, inset) suggest the presence of midgap states,5 located just above the valence band edge. The [F(R)]0.5 curves were elaborated in order to obtain the “apparent” band gap values, as described elsewhere.5 These band gap data are reported in the sixth column of Table 1 along with the initial N/Ti molar ratios (2nd column). By increasing the N content of the samples, the values show a progressive decrease of the “apparent” band gap. 3.2. Electrochemical Measurements. Pristine TiO2 can be considered as an n-type semiconductor, with its Fermi level (EF) located quite close to the conduction band (ECB). When TiO2 particles are irradiated with UV-visible light, a nonequilibrium population of electrons and holes is generated, thereby splitting the Fermi level into two quasi-Fermi levels, one for the holes and the other for the electrons. The latter (nEF*), which almost merges into the TiO2 conduction band, is to be considered since the electrons are the majority charge carriers in a material such as (N-)TiO2. In the presence of an electrolyte, nEF* will equilibrate with the potential of the redox couple in solution. Among the variety of techniques for the electrochemical semiconductor features determination (i.e., impedance spectroscopy,25

N/Ti (XPS)

Ebg (eV) 3.21

2.87

photocurrent measurements,26 flash photolysis,27 modulation spectroscopy28 and spectroelectrochemistry29), two intentionally different kinds of measurements were employed in the present work and applied to selected samples: impedance spectroscopy and photovoltage method, leading to the evaluation of the flatband potential and quasi-Fermi level, respectively. A precise determination of the flatband potential can be often complicated by the presence of surface states and of defects unavoidably connected with the nanometer size of the oxide. Assuming that both techniques represent a direct measure of the lower edge of the conduction band, it is noteworthy to underline that the impedance method is performed in the dark and the photovoltage test under illuminated conditions, thus capturing a picture of two different physical situations but for the same oxide. Recent works are not abundant on this topic and an accurate assessment of the flatband potential/quasi-Fermi level for polycrystalline semiconductor particles is still problematic and not straightforward. Literature data concerning the evaluation of these electronic properties for N-TiO2 nanoparticles in particular are very scarce. The only two recent, sound data show opposite trends of the quasi Fermi energy levels: while Kisch et al.30 reported a variation toward more positive values upon N-doping, Dai et al.31 found a slight shift toward the negative direction. Actually, as shown in Table 2, there is also a significant variation in the literature values concerning undoped TiO2 powders. Comments on the comparability of the data reported in Table 2 are not possible, since the available metrology protocol details are sometimes poor and quite often the uncertainty of the measure is not given. Moreover, one should consider that different synthetic and doping procedures may lead to materials with largely different properties. Since not only the metrology methods but also the types of electrode deposition vary considerably among the reported values, it is definitely difficult to ascribe the wide range of results to precise factors and to compare them unambiguously. For instance, Bolts et al.26 found by Mott-Schottky measurements that Efb values differ by up to -0.3 V for the various determinations. Instead, Beranek and co-workers44 affirmed that they determined nEF* from the dependence of the electrode open-circuit potential on the illumination intensity, since capacitance measurements did not give reliable data due to a high frequency dispersion of the resulting Mott-Schottky plots. Still, Hirai et al.40 declared that their experimental results do not show an ideal MS-behavior, so the flatband potential could not be determined by the impedance method, though a value is reported. Then, in Table 2 only the values of flatband potential/quasiFermi level related to powders are considered, thus excluding single crystals, nanotubes, array electrodes, and so on. Indeed, the crystallinity degree of the oxide, related to its specific surface area, 6384



ARTICLE

Figure 3. XPS spectra of (a) Ti 2p and (b) N 1s region for TN_0.40 sample.

Figure 4. Integral representation of DRS for all the homemade samples, both undoped (T) and N-doped (TN_0.10, TN_0.20, TN_0.40, TN_0.50). Inset: Kubelka-Munk elaboration of the experimental data.

Figure 2. HRTEM images of (a) T, (b) TN_ 0.10, and (c) TN_0.50.

should be taken into account: the defective nature of nanoparticles can introduce large differences from one sample to another. For this reason, in this work, we have chosen both homemade and commercial samples by purpose with very dissimilar specific surface areas and average crystallite diameters (Table 1). It should be noted that all the N/Ti XPS atomic ratios5 are lower than the starting molar ratios adopted in the synthetic route. As reported by us in a previous work,54 an appreciable loss of N species occurs during calcination.

3.2.1. Mott-Schottky Measurements. Flatband potential values from impedance experiments were obtained, according to the literature,26 from the extrapolation of Mott-Schottky plots (C-2 vs E, electrode potential) using the following equation: 1 2 kB T ¼ E - Ef b ð1Þ C2 ND 3 ε0 3 εTiO2 3 e0 e0 where C is the space charge capacitance, E the externally applied potential, Efb the flatband potential at semiconductor/electrolyte junction, ND the donor density, ε0 the permittivity of the free space, εTiO2 the permittivity of the semiconductor electrode, e0 the elementary charge, kB the Boltzmann’s constant, and T the operation temperature. As a rule of thumb, a judicious selection of electrodes (surface/ interface) preparation methods is necessary, in order to optimize titania as working electrode and to ensure that the electrochemical parameter is not influenced by spurious contributions. 6385



ARTICLE

Table 2. Literature Values of Flatband Potential (Efb) and Quasi-Fermi Level (nEF*) Obtained by Different Metrology Methods for Undoped and N-Doped TiO2 Powders Mott-Schottky plot Efb (V vs NHE) undoped TiO2

pH

photovoltage/photocurrent method/spectroelectrochemistry n

ref

EF* (V vs NHE)

pH

-0.01 V

5

12

-0.08

0

41

-0.41 V

7

32

-0.71 V

7

32

-0.9 V -0.9 V

10 13

33 26

-0.52 V -0.68 V

7 7

42 43

-0.2 V

0

34

-0.55 V

7

44

-0.40 V

7.5

35

-0.56 V

7

45

-0.40 V

6.5

36

-0.82 V

11

46

-0.42 V

8

37

-0.05

0

47

-0.16 V

0

38

-0.14 V

0

48

-0.10 V

7

39

-0.28 V

6.6

49

-0.53 V -0.52 V -2.00 V

6 n.r.a

50 31

n.r.

51

n.r.

52

-0.20 V, -0.22 V

7

40

-0.47 V -0.48 V -0.49 V

7

42

-0.59 V, -0.64 V

7

43

-0.35 V

7

44

-0.16 Vb -0.42 V, -0.48 Vc

7 7

44 45

-2.04 V, -2.10 V

n.r.

53

∼ -1.2 V -0.50 V N-doped TiO2

ref

-0.52 V

7

40

a

n.r. = Value not reported in the reference. b Value derived from the dependence of the electrode open-circuit potential (EOC) on the illumination intensity. c N,C-TiO2 sample.

For the sake of comparison, the same range (0.2-1.2 V vs NHE) was adopted for the extrapolation for all samples. A flatband potential of (-0.6 ( 0.2) V vs NHE can be provided as average datum deriving from all impedance measurements. Thus no significant effects are introduced by the doping. 3.2.2. Photovoltage Measurements. The photovoltage method has the important advantage of directly employing TiO2 as a powder in suspension, thus overcoming the possible problems connected with the deposition of the powder onto the solid support. This is a crucial issue, especially when heterogeneous and polydisperse samples are concerned. This technique is based on the linear pH-dependence of the quasi-Fermi level of TiO2 according to the equation: n EF ðpHÞ ¼ n EF ðpH ¼ 0Þ - kpH ð2Þ

Figure 5. Mott-Schottky plot obtained at different frequencies for undoped TiO2 (T).

Examples of the Mott-Schottky plots obtained for bare TiO2 layers on ITO (indium-doped tin oxide) for different frequencies are shown in Figure 5. The experimental points were fitted by linear extrapolation in the range (0.2-1.2 V vs NHE) and the final value of Efb was obtained by the intersection with the potential axis (Efb = -0.5 V vs NHE). While in the case of this sample and of undoped commercial ones, the reproducibility was pretty good, the N-doped samples reproducibility is less satisfactory. This is more probably due to the formation of heterogeneous films ensuing the increasing polydispersity of the doped powders.

wherein the factor k is a constant which is specific for every material and can be assumed to be equal to 59 mV (Nernstian behavior) for TiO2, as reported by other workers26 for metal oxide semiconductors. This behavior can be associated with the protonation/deprotonation equilibrium of the oxide surface. Upon recording the photovoltage as a function of pH, a curve with a shape similar to a titration curve is obtained; the inflection point is the pH value (pH0) at which the redox potential of a selected redox couple, methylviologen in the present case, and the quasi-Fermi level of the oxide are equal. Figure 6 shows the experimental curves obtained for selected samples. The pH0 values are converted to the quasi-Fermi level by the equation: n EF ðpHÞ ¼ Ered ðMV 2þ =MV þ Þ - kðpH - pH0 Þ ð3Þ Provided that the value of k is constant, it is possible to calculate the quasi-Fermi level at any pH. 6386



ARTICLE

Figure 6. Photovoltage vs pH for selected TiO2 suspensions in the presence of (MV)Cl2.

Table 3. Quasi-Fermi Levels Evaluated at pH 6 by Photovoltage Method for Homemade and Commercial Samples sample

n

EF* (V vs NHE)

T

-0.65 ( 0.07

TN_0.10

-0.67 ( 0.07

TN_0.20 TN_0.40

-0.71 ( 0.06 -0.67 ( 0.06

TN_0.50

-0.66 ( 0.01

P25

-0.64 ( 0.04

HOMBIKAT UV100

-0.62 ( 0.04

Table 3 reports the obtained values at pH 6, referred to the NHE scale. There is no significant variation among all the N-doped samples, which range around -0.7 V, nor between doped and undoped ones. Overall, we found the measurement of the photovoltage to be very simple and reliable, even more accurate and reproducible with respect to many other techniques, such as the Mott-Schottky plot by impedance spectroscopy, though the latter is absolutely well-known in the field of semiconductors and validated when applied to single crystals. 3.3. Density Functional Theory (DFT) Calculations. To study N-doping in the presence of oxygen vacancies, DFT calculations have been performed on a bulk anatase supercell containing either substitutional or interstitial nitrogen doping, plus one or two oxygen vacancies located as far as possible from the N-centers, so as to avoid any direct defect-impurity interaction. Theoretical calculations were performed without considering the presence of the small brookite percentage. In the presence of N impurities, oxygen vacancies excess electrons are transferred from the higher energy Ti3þ states to the empty N mid-gap states. Even if this internal charge transfer occurs independently of the exchange-correlation functional used, depending on the stoichiometric ratio between vacancies and impurities, this transfer can involve only some of the Ti3þ 3d electrons, leaving others on the Ti ions sites. Thus, one needs to go beyond plain DFT because of the strongly correlated nature of the d-electrons in titania.55 There exist several schemes by which one can do so. The two most common are the hybrid functionals56 and the so-called DFT þ U.57 Both these methods suffer from dependence on a tunable parameter. In hybrid-DFT, a degree of exact exchange is introduced by mixing DFT and Hartree-Fock. In DFT þ U, the method employed in this article, one tries to better describe the electron correlation effects by adding a Hubbard-U term in the

Figure 7. Density of electronic States (DOS) at the level of GGA for different combinations of oxygen vacancies (V@O) and N-doping locations (N@O for substitutional and Nint for interstitial) and their stoichiometric ratio; spin polarized states are in continuous black and dashed lines. Fermi energy levels are indicated by the vertical dashed lines.

functional, representing an on-site Coulomb repulsion among selected orbitals associated with the given atomic sites. Thus, to correct the self-interaction error and the resulting bias toward noninteger orbital occupations in DFT,58 the U parameter has been tested. Thus, the results can depend on the value of the interaction parameter U in the DFT þ U scheme. Although schemes have been derived for calculating the U parameter from DFT, the resulting U can be quite different, dependent on the scheme used.59 The extent to which the excess electrons were found to spread over the system differed considerably.60 The present Density of States calculations are reported on Figures 7 and 8. As usual for periodic PBE calculations, the band gap is underestimated around 2.2 eV instead of 3.2 eV, as showed on panel (a) of Figure 7.61 As far as we know, among DFT-based approaches, only the hybrid functional gives a better agreement to the band gap energy.55 Once a oxygen vacancy is generated, examination of DOS on panel (b) of Figure 7 reveals that for U = 0 the Ti 3d-like state lays just below the conduction band, as previously reported.62 At the PBE level, the two extra electrons are fully delocalized on all of the Ti ions in the supercell, and consequently, the singlet and triplet spin solutions are degenerate. The structural deformation of the lattice is very small and symmetric, with the three undercoordinated Ti ions around the vacancy showing a slight outward relaxation with respect to their equilibrium position. 6387



Figure 8. The same as Figure5, but at the level of GGA þ U calculations. U has been set equal to 5 eV.

Instead, on the same panel, but on Figure 8 where DOS were calculated for U = 5 eV, two distinct peaks are observed in the density of states. By inspecting the energy levels of the KS (Kohn and Sham) orbitals for the GGA þ U optimized geometry, one is located at 2.14 eV and the other at 2.36 eV from the top of the valence band, where for GGA þ U the original GGA band gap has been opened to about 2.5 eV, in agreement with previous DFT þ U calculations.21 Since the TiO2 is more of a chargetransfer type semiconductor than a Mott-Hubbard insulator, one should not expect to be able to open up the band gap to its experimental values using a physically reasonable value of U.21e Once the substitutional nitrogen is introduced concomitant with an oxygen vacancy, the localized N states just above the valence band behave as excellent electron traps. These shallow gap states just above the valence band—reported in panels (c)-(e) of Figures 7 and 8—are originated from the combination of the substitutional N 2p orbitals with the oxygen ones and they were detected by the Optical Measurements described above. If interstitial N doping is considered, the N orbitals are deeper into the band gap as shown in panels (f) and (g) of Figures 7 and 8. It has been suggested that these states may act as recombination ones, annihilating the electron-hole photogenerated pair. However, from an optical point of view, also these states explain the apparent band gap narrowing observed in the DR measurements. Either N-substitutional shallow or N-interstitial deeper electronic states play a crucial role as acceptors promoting the conversion of the Ti3þ species into the Ti4þ ones.63 However, the three N p states originally host five electrons

ARTICLE

Figure 9. Pictorial representation of the electron transfer process during nitrogen substitutional doping: (a) an oxygen vacancy and a substitutional N is present and the Fermi energy is invariant; (b) an oxygen vacancy and two substitutional N are present and the Fermi energy level is shifted at the top of the valence band; (c) two oxygen vacancies and a single substitutional N are present and the Fermi energy is invariant. The same reasoning can be applied to interstitial doping energy levels.

and can accept only one more. For this reason, when an equal number either of substitutional or interstitial N doping is present as it is the case in panel (c) and (f) of Figures 7 and 8, one Ti3þ electron is still present out of the two originally created by a single oxygen vacancy, and the Fermi energy is still pinned at the bottom of the conduction band. This is pictorially represented on panel (a) of Figure 9. Similar conclusions have been reached by Graciani et al.64,8b but in the case of the rutile surface and subsurface oxygen vacancies in the presence of nitrogen doping. Only when enough N states are introduced, as showed on panel (d) of Figures 7 and 8, the charge transfer from the Ti3þ states is complete and the Fermi energy is lowered to just above the valence band. Each nitrogen impurity traps one Ti 3d electron as shown on panel (b) of Figure 9. Another possible setup is the one reported on panel (e) and (g) of Figure 7 and 8, where vacancies stoichiometric coefficient is double respect to the N doping one. In these cases, the Fermi energy is even more pinned at the bottom of the conduction band, since an extra number of Ti3þ states are present. The orbitals representation of this electronic arrangements is the one on panel (c) of Figure 9. Finally, by comparing panels (b), (e), and (g) of Figure 7 versus the same panels of Figure 8, one can appreciate how DFT þ U was necessary to describe the Ti3þ states. Localized states arise on panels (b) of Figure 8 and clearly increment their populations when the number of vacancies is doubled, as is evident from the DOS of panels (e) and (g) of Figure 8. However, the same conclusions can be reached in terms of Fermi energy location both using DFT or DFT þ U approach, as indicated by vertical lines in Figures 7 and 8. 6388



4. DISCUSSION Several speculative explanations about the improved photocatalitic effects of N-doped titania have been put forward in recent years, due to the crucial role that this compound has for photocatalisis. The present DFT calculations, in agreement with previous literature results, clearly indicate that the apparent band gap narrowing observed in the DR measurements is due to the presence of intragap states induced by the N p orbitals mixing with the oxygen ones.63a However, from this partial point of view, the photogenerated electrons fate is still unclear and opened to the possibility of complete or partial transfer into the empty N orbitals generated by doping. Thus, only partial conclusions can be drawn by comparison between optical experiments and DFT calculations. Then, considering that the visible light response could be induced not directly by the doping, but by oxygen vacancies, stabilized by the presence of nitrogen as a result of charge compensation, and acting as color centers,65,4 the motivation for other electronic experiments in conjunction with DFT calculations is clear and the theoretical findings allow one to have a comprehensive interpretation of the impedance and photovoltage experimental results. On the one hand, by inspection of the Ti3þ 3d states pictorially reported in Figure 9 just below the conduction band, we can safely conclude that these states are originated by the oxygen vacancies and are invariant under N doping. On the other hand, the quasi-Fermi energy levels determined by photovoltage experiments are the electronic orbitals populated under light irradiation and they should be distinguished from the Fermi energy level which, instead, indicates the half population energy. Thus, a direct comparison between the quasi-Fermi energy levels and the Ti3þ 3d states can be made, showing the agreement with the photovoltage results. Instead, the Mott-Schottky plots, which are obtained under dark conditions, allow one to measure Efb. This value indicates the average occupation number, i.e., the Fermi level. For an n-type semiconductor, it represents also the conduction band edge. A comparison between the Mott-Schottky and the theoretical results can be done by looking at the vertical dashed lines (Fermi energy) in Figures 7 and 8. This excludes panel (d) of Figures 7 and 8 to be a realistic doping set up. Besides, the Mott-Schottky flatband and the photovoltage quasi Fermi energy values show that these are located at the same levels on an electrochemical scale. Taking into consideration that the Mott-Schottky measurements are under dark and the photovoltage ones under light irradiation, one can univocally identify the location of the conduction band and conclude that this is invariant under N-doping for all our samples. At the light of these considerations and by comparison with the several different N-doping and O vacancy ratio scenarios simulated at the level of DFT þ U calculations (Figures 7 and 8), one can safely conclude that in the case of the present samples, oxygen vacancies are more numerous than N-doping centers, since doping is not significantly changing the Fermi energy location. Thus, on one side, partial accommodation of oxygen vacancy electrons into N 2p states occurs and leads to the formation of charged diamagnetic N impurities and reoxidized Ti ions. On the other side, this electron transfer process induces the formation of extra oxygen vacancies, which has been proven to be favored in the presence of N-doping.7 Eventually, reoxidation of Ti ions is somewhat compensated by the formation of Ti

ARTICLE

3d states and the Fermi energy is left invariant, i.e., pinned at the bottom of the conduction band. Thus, a likely comprehensive picture is that represented by panel (c) of Figure 9. In conclusion, the effects of N-doping sites may be multiple, since they act not only as electron scavengers, but also as stabilizers of the color centers.5

5. CONCLUSIONS Titanium dioxide is one of the most commonly adopted semiconductors in photocatalysis. However, univocal conclusions on the Fermi energy levels and related charge transfer processes versus N-doping concentration were still missing. In this work, the electronic effects induced by the presence of N impurities in the TiO2 nanocrystals were investigated. Both photovoltage and impedance experiments in conjunction with theoretical DFT calculations allowed us to outline a comprehensive picture of the electronic structure of N-doped nanoparticles. While a band gap estimate value is easily obtained from diffuse reflectance measurements, the flatband potential and the quasiFermi level evaluations generally require more effort. Among several approaches described in the literature, the photovoltage method turned out to be reliable and simple, probably being the most suitable technique for nanosized semiconductor powders for quasi-Fermi level experimental evaluation. From both photovoltage and impedance measurements, one can appreciate that, in the present case, quasi Fermi energy levels and flatband potentials are invariant under N-doping. These findings are rationalized by assuming that the position of the conduction band is not affected by the doping, and these considerations can be extended to the Fermi energy level. Then, comparisons with theoretical calculations allow us to assert that vacancies are more numerous than N doping centers in our samples and that the electron transfer from Ti3þ 3d orbitals occurs only in part, keeping the Fermi energy pinned at the bottom of the conduction band. ’ AUTHOR INFORMATION Corresponding Author

*Phone: þ390250314219; Fax: þ390250314228; E-mail: [email protected].

’ ACKNOWLEDGMENT The authors acknowledge the Lumilab group of the University of Ghent (Belgium) for the XPS characterization. L.F. and M.C. acknowledge the “Universita’ degli Studi di Milano” for a uma generous starting grant (“5 per mille”). M.C. thanks M. C and J. J. Plata for useful discussions. ’ REFERENCES (1) (a) Hardin, B. E.; Hoke, E. T.; Armstrong, P. B.; Yum, J.-H.; Comte, P.; Torres, T.; Frechet, J. M. J.; Nazeeruddin, M. K.; Graetzel, M.; McGehee, M. D. Nat. Photonics 2009, 3, 406–411. (b) Roy, P.; Albu, S. P.; Schmuki, P. Electrochem. Commun. 2010, 12, 949–951. (c) Roy, P.; Dey, T.; Lee, K.; Kim, D.; Fabry, B.; Schmuki, P. J. Am. Chem. Soc. 2010, 132, 7893–7895. (d) Ardizzone, S.; Bianchi, C. L.; Cappelletti, G.; Naldoni, A.; Pirola, C. Environ. Sci. Technol. 2008, 42, 6671–6676. (e) Milanesi, F.; Cappelletti, G.; Annunziata, R.; Bianchi, C. L.; Meroni, D.; Ardizzone, S. J. Phys. Chem. C 2010, 114, 8287–8293. (2) (a) Peng, F.; Cai, L.; Yu, H.; Wang, H.; Yang, J. J. Solid State Chem. 2008, 181, 130–136. (b) Xing, M. Y.; Zhang, J. L.; Chen, F. Appl. 6389


The Journal of Physical Chemistry C Catal. B: Environ. 2009, 89, 563–569. (c) Wang, Y.; Zhou, G.; Li, T.; Qiao, W.; Li, Y. Catal. Commun. 2009, 10, 412–415. (d) Mitoraj, D.; Kisch, H. Chem.—Eur. J. 2010, 16, 261–269. (e) Zhao, Y.; Qiu, X.; Burda, C. Chem. Mater. 2008, 20, 2629–2636. (3) (a) Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, 269–271. (b) Sato, S.; Nakamura, R.; Abe, S. Appl. Catal., B 2005, 284, 131–137. (4) Emeline, A. V.; Kuznetsov, V. N.; Rybchuk, V. K.; Serpone, N. Int. J. Photoenergy 2008, 2008, 1–19. (5) Spadavecchia, F.; Cappelletti, G.; Ardizzone, S.; Bianchi, C. L.; Cappelli, S.; Oliva, C.; Scardi, P.; Leoni, M.; Fermo, P. Appl. Catal., B 2010, 96, 314–322. (6) (a) Gole, J. L.; Stout, J. D.; Burda, C.; Lou, Y.; Chen, X. J. Phys. Chem. B 2004, 108, 1230–1240. (b) Huang, D.; Liao, S.; Quan, S.; Liu, L.; He, Z.; Wan, J.; Zhou, W. J. Non-Cryst. Solids 2008, 354, 3965–3972. (c) Ananpattarachaia, J.; Kajitvichyanukulb, P.; Seraphin, S. J. Hazard. Mater. 2009, 168, 253–261. (7) Di Valentin, C.; Pacchioni, G.; Selloni, A.; Livraghi, S.; Giamello, E. J. Phys. Chem. B 2005, 109, 11414–11419. (8) (a) Di Valentin, C.; Finazzi, E.; Pacchioni, G.; Selloni, A.; Livraghi, S.; Paganini, M. C.; Giamello, E. Chem. Phys. 2007, 339, 44–56. (b) Graciani, J.; Nambu, A.; Evans, J.; Rodriguez, J. A.; Sanz, J. F. J. Am. Chem. Soc. 2008, 130, 12056–12063. (9) (a) Henrich, V. E.; Kurtz, R. L. Phys. Rev. B 1981, 23, 6280–9287. (b) Aiura, Y.; Nishihara, Y.; Haruyama, Y.; Komeda, T.; Kodaira, S.; Sakisaka, Y.; Maruyama, T.; Kato, H. Physica B 1994, 194-196, 1215–1216. (c) Zhang, Z.; Jeng, S.-P.; Henrich, V. E. Phys. Rev. B 1991, 43, 12004–12011. (d) Heise, R.; Courths, R; Witzel, S. Solid State Commun. 1992, 84, 599–602. (e) Nerlov, J.; Ge, Q.; Møller, P. J. Surf. Sci. 1996, 348, 28–38. (f) Thomas, A. G.; Flavell, W. R.; Kumarasinghe, A. R.; Mallick, A. K.; Tsoutsou, D.; Smith, G. C.; Stockbauer, R.; Patel, S.; Graetzel, M.; Hengerer, R. Phys. Rev. B 2003, 67, 035110/1–035110/7. (g) Diebold, U. Surf. Sci. Rep 2003, 48, 53–229. (10) (a) Baumanis, C.; Bahnemann, D. W. J. Phys. Chem. C 2008, 112, 19097–19101. (b) Spagnol, V.; Sutter, E.; Debiemme-Chouvy, C.; Cachet, H.; Baroux, B. Electrochim. Acta 2009, 54, 1228–1232. (c) Lisowska-Oleksiaka, A.; Szybowskaa, K.; Jasulaitiene, V. Electrochim. Acta 2010, 55, 5881–5885. (11) (a) Bard, A.; Faulkner, L. R. Electrochemical Methods: Fundamentals and Applications, 2nd ed.; John Wiley & Sons, Inc.: New York, 2001. (b) Ward, M. D.; White, J. R.; Bard, A. J. J. Am. Chem. Soc. 1983, 105, 27–31. (c) Mitoraj, D.; Kisch, H. J. Phys. Chem. C 2009, 113, 20890–20895. (12) Bandara, J.; Pradeep, U. W. Thin Solid Films 2008, 517, 952–956. (13) Roy, A. M.; De, G. C.; Sasmal, N.; Bhattacharyya, S. S. Int. J. Hydrogen Energy 1995, 20, 627–630. (14) (a) Kresse, G.; Hafner, J. J. Phys. Rev. B 1993, 47, 558–561. (b) Kresse, G.; Furthm€uller, J. Phys. Rev. B 1996, 54, 11169–11186. (15) Perdew, J. P.; Burk, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865–3868. (16) Perdew, J. P.; Wang, Y. Phys. ReV. B 1992, 45, 13244–13249. (17) Harris, J. Phys. Rev. B 1985, 31, 1770–1779. Foulkes, W. M. C.; Haydock, T. Phys. Rev. B 1989, 39, 12520–12536. (18) (a) Teter, M. P.; Payne, M. C.; Allan, D. C. Phys. Rev. B 1989, 40, 12255–12263. (b) Bylander, D. M.; Kleinman, L.; Lee, S. Phys Rev. B 1990, 42, 1394–1403. (19) Davidson, E. R. In Methods in Computational Molecular Physics; Dierksen, G. H. F., Ed.; NATO Advanced Study Institute, Series C; Plenum: New York, NY, 1983; Vol. 113, p 95. (20) Dudarev, S. L.; Botton, C. A.; Savarsov, S. Y.; Hunphreys, C. J.; Sutton, A. P. Phys. Rev. B 1998, 57, 1505–1509. (21) (a) Finazzi, E.; Di Valentin, C.; Pacchioni, G.; Selloni, A. J. Chem. Phys. 2008, 129, 154113/1–154113/9. (b) Mattioli, G.; Filippone, F.; Alippi, P.; Amore Bonapasta, A. Phys. Rev. B 2008, 78, 241201/1–241201/4. (c) Morgan, B. J.; Watson, G. W. Surf. Sci. 2007, 601, 5034–5041. (d) Cheng, H.; Selloni, A. J. Chem. Phys. 2009, 131, 0547031–1/0547031-10. (e) Stausholm-Moeller, J.; Kristoffersen,

ARTICLE

H. H.; Hinnemann, B.; Madsen, G. K. H.; Hammer, B. J. Chem. Phys. 2010, 133, 144708/1–144708/8. (22) Burdett, J. K.; Hughbandks, T.; Miller, G. J.; Richardson, J. W.; Smith, J. V. J. Am. Chem. Soc. 1987, 10, 3639–3646. (23) Ardizzone, S.; Bianchi, C. L.; Cappelletti, G.; Naldoni, A.; Pirola, C.; Ragaini, V.; Gialanella, S. Nano. Res. Lett. 2009, 4, 97–105. (24) (a) Burgeth, G.; Kisch, H. Coord. Chem. Rev. 2002, 230, 41–47. (b) Cappelletti, G.; Bianchi, C. L.; Ardizzone, S. Appl. Catal., B 2008, 78, 193–201. (25) Fabregat-Santiago, F.; Garcia-Belmonte, G.; Bisquert, J.; Bogdanoff, P.; Zaban, A. J. Electrochem. Soc. 2003, 150, E293–E298. (26) Bolts, J. M.; Wrighton, M. S. J. Phys. Chem. 1976, 80, 2641–2645. (27) Duonghung, D.; Ramsden, J.; Gratzel, M. J. Am. Chem. Soc. 1982, 104, 2977–2985. (28) Chaparro, A. M. J. Electroanal. Chem. 1999, 462, 251–258. (29) Boschloo, G.; Fitzmaurice, D. J. Phys. Chem. B 1999, 103, 2228–2231. (30) Kisch, H.; Sakthivel, S.; Janczarek, M.; Mitoraj, D. J. Phys. Chem. C 2007, 111, 11445–11449. (31) Tian, H.; Hu, L.; Zhang, C.; Liu, W.; Huang, Y.; Mo, L.; Guo, L.; Sheng, J.; Dai, S. J. Phys. Chem. C 2010, 114, 1627–1632. (32) Kandiel, T. A.; Feldhoff, A.; Robben, L.; Dillert, R.; Bahnemann, D. W. Chem. Mater. 2010, 22, 2050–2060. (33) Kea, T.-Y.; Lee, C.-Y.; Chiu, H.-T. Appl. Catal., A 2010, 381, 109–113. (34) Palmas, S.; Polcaro, A. M.; Rodriguez Ruiz, J.; Da Pozzo, A.; Vacca, A.; Mascia, M.; Delogu, F.; Ricci, P. C. Int. J. Hydrogen Energy 2009, 34, 9662–9670. (35) Wang, G.; Wang, Q.; Lu, W.; Li, J. J. Phys. Chem. B 2006, 110, 22029–22034. (36) Chettah, H.; Abdi, D.; Amardjia, H.; Haffar, H. Ionics 2009, 15, 169–176. (37) Radecka, M.; Rekas, M.; Trenczek-Zajac, A.; Zakrzewska, K. J. Power Sources 2008, 181, 46–55. (38) Kalayanansundaram, K.; Graetzel, M. Coordin. Chem. Rev. 1998, 77, 347–414. (39) Hattori, A.; Tokihisa, Y.; Tada, H.; Tohge, N.; Ito, S.; Hongo, K.; Shiratsuchi, R.; Nogami, G. J. Sol-Gel Sci. Technol. 2001, 22, 53–61. (40) Hirai, T.; Tari, I.; Ohzuku, T. Bull. Chem. Soc. Jpn. 1981, 54, 509–512. (41) Duonghong, D.; Ramsden, J.; Graetzel, M. J. Am. Chem. Soc. 1982, 104, 2977–2985. (42) Sakthivel, S.; Kisch, H. ChemPhysChem 2003, 4, 487–490. (43) Sakthivel, S.; Janczarek, M.; Kisch, H. J. Phys. Chem. B 2004, 108, 19384–19387. (44) Beranek, R.; Kisch, H. Electrochem. Commun. 2007, 9, 761–766. (45) Mitoraj, D.; Kisch, H. J. Phys. Chem. C 2009, 113, 20890–20895. (46) Redmond, G.; Fitzmaurice, D. J. Phys. Chem. 1993, 97, 1426–1430. (47) Ward, M. D.; White, J. R.; Bard, A. J. J. Am. Chem. Soc. 1983, 105, 27–31. (48) Duonhong, D.; Ramsden, J.; Gratzel, M. J. Am. Chem. Soc. 1982, 101, 2977–2985. (49) Wodka, D.; Bielanska, E.; Socha, R. P.; Wodka, M. E.; Gurgul, J.; Nowak, P.; Warszynski, P.; Kumakiri, I. Appl. Mater. Inter. 2010, 2, 1945–1953. (50) Di Iorio, Y.; Rodriguez, H. B.; San Roman, E.; Grela, M. A. J. Phys. Chem. C 2010, 114, 11515–11521. (51) Huang, F. Q.; L€u, X. J.; Mou, X. L.; Wu, J. J. Adv. Funct. Mater. 2010, 20, 509–515. (52) Premkumar, J. Chem. Mater. 2004, 16, 3980–3981. (53) Huang, Y.; Mo, L.; Guo, L.; Sheng, J.; Dai, S. J. Phys. Chem. C 2010, 114, 1627–1632. (54) Bianchi, C. L.; Cappelletti, G.; Ardizzone, S.; Gialanella, S.; Naldoni, A.; Oliva, C.; Pirola, C. Catal. Today 2009, 144, 31–36. (55) Janotti, A.; Varley, J. B.; Rinke, P.; Umezawa, N.; Kresse, G.; Van de Walle, C. G. Phys. Rev. B 2010, 81, 085212/1–085212/7. (56) Becke, A. D. J. Chem. Phys. 1993, 98, 1372–1377. 6390



ARTICLE

(57) Anisimov, V. I.; Solovyev, I. V.; Korotin, M. A.; Czy_zyk, M. T.; Sawatzky, G. A. Phys. Rev. B 1993, 48, 16929–16934. (58) K€ummel, S.; Kronik, L. Rev. Mod. Phys. 2008, 80, 3–60. (59) Cococcioni, M.; De Gironcoli, S. Phys. Rev. B 2005, 71, 035105/1–035105/16. (60) (a) Vijay, A.; Mills, G.; Metiu, H. J. Chem. Phys. 2003, 118, 6536–6551. (b) Wu, X. Y.; Selloni, A.; Nayak, S. K. J. Chem. Phys. 2004, 120, 4512–4516. (c) Rasmussen, M. D.; Molina, L. M.; Hammer, B. J. Chem. Phys. 2004, 120, 988–997. (61) Long, R.; English, N. J. Appl. Phys. Lett. 2009, 94, 132102/ 1–132102/3. (62) (a) Ramamoorthy, M.; Vanderbilt, D.; King-Smith, R. D. Phys. Rev. B 1994, 49, 16721–16727. (b) Paxton, A. T.; Thiên-Nga, L. Phys. Rev. B 1998, 57, 1579–1584. (63) (a) Batzill, M.; Morales, E. H.; Diebold, U. Phys. Rev. Lett. 2006, 96, 026103/1–026103/4. (b) Batzill, M.; Morales, E. H.; Diebold, U. Chem. Phys. 2007, 339, 36–43. (64) Graciani, J.; Alvarez, L. J.; Rodriguez, J. A.; Sanz, J. F. J. Phys. Chem. C 2008, 112, 2624–2631. (65) Ihara, T.; Miyoshi, M.; Iriyama, Y.; Matsumoto, O.; Sugihara, S. Appl. Catal., B 2003, 42, 403–409.

6391
