Article pubs.acs.org/JPCC
Fluorine- and Niobium-Doped TiO2: Chemical and Spectroscopic Properties of Polycrystalline n‑Type-Doped Anatase Jakub Biedrzycki,† Stefano Livraghi,† Elio Giamello,*,† Stefano Agnoli,‡ and Gaetano Granozzi‡ †
Dipartimento di Chimica and NIS, Università di Torino, Via P. Giuria 7, 10125 Torino, Italy Dipartimento di Scienze Chimiche, Università di Padova, Via Marzolo 1, 35131 Padova, Italy
‡
Downloaded via UNIV OF SOUTH DAKOTA on September 5, 2018 at 06:52:32 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
S Supporting Information *
ABSTRACT: Doping titanium dioxide (anatase) with elements carrying an extra electron, such as Nb and F, or with their mixtures leads to n-type materials showing peculiar properties with respect to the pristine oxide. Niobium and fluorine are present in the lattice in the form of Nb5+ and F− ions (detected by XPS), and the extra electrons carried by the dopants are stabilized on titanium ions, which become EPR-visible as Ti3+ ions homogeneously dispersed in the bulk of the crystals. Under such conditions, the optical band gap transition is slightly red-shifted (by a few tenths of an electronvolt) for all samples containing fluorine, and the Fermi level lies, depending on the material, at the boundary or even in the lower region of the conduction band. The typical Ti3+(I) centers generated by valence induction are responsible for the already reported conductivity properties of the system. The presence of these centers also influences the process of electron injection in the solid, favoring the dilution of additional reduced centers in the bulk, thereby leading to a homogeneously reduced material with optoelectronic properties differing from those of reduced anatase.
1. INTRODUCTION Titanium dioxide is one of the most investigated metal oxides. The importance of this abundant, inexpensive, and nontoxic material is certainly due to its large-scale applications as a white pigment. However, a great deal of the scientific investigations concerning this oxide are essentially due to the exciting applications related to its photochemical and photophysical properties. This explains the exponential growth in studies on both bare and modified TiO2 appearing in the literature in the past 20 years.1,2 Titanium dioxide (or titania) is the most important photocatalyst, active in several reactions for pollutant abatement,3 and since the discovery of water photosplitting in 1972,4 it can also be considered as a system of interest for photochemical and photoelectrochemical applications in energetics (artificial photosynthesis). Furthermore, titanium dioxide is employed as a biocide,5 in odor control,6 and in the preparation of self-cleaning surfaces and superhydrophilic antifogging layers.7 The use of titanium dioxide in photovoltaics is also extremely important. In dye-sensitized solar cells (DSSCs),8 in fact, TiO2 plays the roles of acceptor of electrons from the dye excited state and carrier of electrons toward the electric contact. The main polymorphs of titanium dioxide are rutile and anatase. The former phase is more stable than the latter, whereas the latter is more effective in terms of photochemical applications. In parallel with the growth of titanium dioxide applications, intense activity began to focus on the preparation of modified forms of the solid in an effort to tune some of its specific physical properties. This is the case for the inclusion of either transition-metal or nonmetal impurities in the lattice to © 2014 American Chemical Society
modify the optical absorption of the system. One of the goals of this activity, although not the only one, is to reduce the optical absorption threshold of the oxide, whose band gap is around 3.2 eV for the anatase polymorph (corresponding to a UV photon), and to make it photosensitive to visible light.1,9−11 Among the various types of nonmetal doping of TiO2, that with fluoride ions assumes a particular role. About 10 years ago, it was reported that the inclusion of fluoride ions in the oxide matrix improves the photocatalytic performance of the bare oxide in the mineralization of various organic pollutants using either UV12 or visible13 light. To obtain fluorine-doped titania, fluoride ions are usually introduced in the liquid medium for the preparation of the oxide through hydrolysis (for instance, by the frequently used sol−gel technique). A fraction of the F− ions is included in the lattice such that they substitute O2− anions. An investigation by some of us, based on electron paramagnetic resonance (EPR) spectroscopy14 of F-doped anatase TiO2 (hereafter F−TiO2) prepared by the sol−gel method, clarified that paramagnetic Ti3+ ions are present in the solid. This occurs because at least a fraction of the excess electrons introduced by fluorine are localized by lattice Ti4+cations. The presence of reduced Ti3+ centers in the asprepared, fully oxidized solid indicates that the F−TiO2 system can be described in terms of an n-type semiconductor. It is worth noting that pristine titanium dioxide (belonging to the class of reducible oxides because it easily loses oxygen) also Received: February 3, 2014 Revised: April 1, 2014 Published: April 1, 2014 8462
dx.doi.org/10.1021/jp501203h | J. Phys. Chem. C 2014, 118, 8462−8473
The Journal of Physical Chemistry C
Article
Table 1. Abbreviations Used for the Samples in the Present Work and Corresponding Compositional, Structural, and Optical Properties phasea,b (w/w %) sample abbreviation
Nb/Ti nominal
Nb/Ti XRF
F/Ti nominal
calcination temp (K)
A
R
B
crystal sizec (nm)
band gap, Eg (eV)
TiO2 F2 Nb1 NbF25 NbF35 NbF45 NbF25a
0 0 0.05 0.05 0.05 0.05 0.05
0 − 0.05 0.05 0.05 0.06 0.05
0 0.90 0 0.25 0.35 0.45 0.25
770 770 770 970 970 970 970
100 100 55 98 100 100 100
0 0 10 2 0 0 0
0 0 35 0 0 0 0
30 30 12 21 24 32 23
3.20 3.20 2.91 3.03 3.08 3.11 3.18
a
A = anatase, R = rutile, B = brookite. bData obtained by Rietveld refinement using the MAUD program. cData apply to the anatase polymorph only.
2. EXPERIMENTAL SECTION 2.1. Sample Preparation. The TiO2 sample was prepared by the sol−gel process using titanium isopropoxide as the Ti precursor, as reported in a previous work.20 Doped samples were prepared as described in the following subsections. 2.1.1. F−TiO2. This set of samples was prepared by the sol− gel method using hydrofluoric acid solution as the hydrolyzing agent. In a typical preparation, 7.5 mL of titanium isopropoxide was diluted in 7.5 mL of 2-propanol, and the obtained mixture was hydrolyzed with 4 mL of HF solution of different concentrations to obtain nominal F/Ti ratios of 0.01, 0.90, 1.80, 3.50, 5.50, and 7.00. The formed solid was dried at 340 K and finally calcined at 770 K (heating rate of 10 K/min) for 1 h to obtain a fully oxidized material and to eliminate all contaminants left from the synthesis. Higher temperatures were not used to limit fluorine loss.30 This set of samples is labeled F1−F6 in order of increasing F/Ti ratio. For the sake of brevity, only data related to F2 are reported here. The data related to the remaining samples are provided in the Supporting Information. 2.1.2. Nb−TiO2. This sample was prepared by the sol−gel process. An appropriate amount of NbCl5, to obtain a nominal doping of 5% (Nb/Ti = 0.05), was added to 7.5 mL of 2propanol, and the mixture was stirred at room temperature until the NbCl5 was completely dissolved. Then, 7.5 mL of titanium isopropoxide was added. The final solution was hydrolyzed with 4 mL of H2O. The obtained solid was dried at 340 K and finally calcined at 770 K (heating rate of 10 K/min) for 1 h to obtain a fully oxidized material and to eliminate all contaminants left from the synthesis. Higher temperature were not used to avoid anatase-to-rutile phase conversion. This sample is denoted as Nb1. 2.1.3. F−Nb−TiO2 (Codoped Samples). This set of samples was prepared as follows: A solution containing the cations (Nb, Ti) was prepared as in the previous case and then hydrolyzed with solutions of hydrofluoric acid having different concentrations to obtain nominal F/Ti ratios of 0.05, 0.15, 0.25, 0.35, and 0.45. The obtained solids were dried at 340 K and finally calcined at 970 K (heating rate of 10 K/min) for 1 h. In this case, a higher calcination temperature was employed because fluorine (although partially expelled from the solid under these conditions) stabilizes the anatase phase, reducing the conversion to rutile and allowing materials with high crystallinity to be obtained. These samples are labeled NbF05, NbF15, NbF25, NbF35, and NbF45 according to their F/Ti ratios (Table 1). An additional sample labeled NbF25a was prepared using H2O as a hydrolyzing agent and NbF5 as a source of both Nb and F. For
becomes an n-type system when annealed under a vacuum, as a result of oxygen depletion, formation of anion vacancies, and excess electron stabilization (Ti3+). Fully oxidized, undefective TiO2 and F−TiO2 thus differ because the latter is, intrinsically, an n-type semiconducting system, as indicated by early investigations pointing to the electronic and photoelectric properties of F-doped rutile TiO2.15,16 The same electronic effect described for F−TiO2 is observed when the tetravalent Ti of anatase is substituted by a pentavalent element such as Nb or Sb. Also in this case, the extra electron introduced by the dopant is expected17 to localize on the titanium ions, producing Ti3+. The formation of lattice trivalent titanium ions has been monitored directly by EPR spectroscopy for Nb−TiO2 (anatase),18−20 The interest in Nbdoped titania has recently increased because of the search for new transparent conducting oxides (TCOs) for use in optoelectronic devices21−23 and, in particular, in DSSCs to promote charge transport within the transparent oxide layer usually constituted by bare TiO2.24−26 It is worth mentioning that, in the case of Nb−TiO2, the described valence induction effect (i.e., formation of Nb5+ and Ti3+) holds for the anatase polymorph, whereas in rutile doped with niobium, the extra electron remains on the dopant atom, as indicated by the presence of Nb4+ ions, which were detected both by EPR spectroscopy at low temperature for Nb−TiO2 rutile single crystals27,28 and by X-ray photoelectron spectroscopy (XPS) measurements.29 In the present article, we report a detailed study of the chemical and spectroscopic properties of n-type doped polycrystalline anatase performed with the joint use of optical spectroscopy (DR UV−vis−NIR), photoelectron spectroscopies, and electron paramagnetic resonance (EPR) spectroscopy. The materials under investigation included both singly doped (F−TiO2 and Nb−TiO2) and codoped (F,Nb) systems. Our investigation started with two main goals. The first one was to explore band gap modifications of TiO2 depending on the type of dopant. The second goal was that of examining the behavior of the doped materials upon addition of further excess electrons (by direct injection or reduction) in comparison with the behavior of pristine TiO2. The presence and mobility of excess electrons are essential in photochemical applications of solids such as photocatalysis and dye-sensitized solar cells. A similar approach induced Wold and co-workers,15 more than 30 years ago, to investigate the photoconduction of fluorinated rutile crystals, which showed significantly higher conductivity than reduced rutile. More recently, studies on Nb-doped anatase have shown similar conductivity increments.24,25 8463
dx.doi.org/10.1021/jp501203h | J. Phys. Chem. C 2014, 118, 8462−8473
The Journal of Physical Chemistry C
Article
Figure 1. XRD patterns of TiO2-based materials: (a) TiO2, (b) F2, (c) Nb1, (d) NbF25, (e) NbF35, (f) NbF45, (g) NbF25a. The bars corresponding to TiO2-A, TiO2-R, and TiO2-B indicate the diffraction peaks of the anatase, rutile, and brookite polymorphs, respectively.
the sake of brevity, only data related to samples NbF25, NbF35, NbF45, and NbF25a are reported herein. The data related to the remaining samples are provided in the Supporting Information. Table 1 summarizes the main features and abbreviations for the samples reported in this work. Because of the loss of fluorine that occurs progressively during the various reductive treatments, the evaluation by chemical analysis of the initial F/ Ti ratio is substantially insignificant and was therefore not performed. 2.2. Structural and Spectroscopic Characterization of the Materials. X-ray diffraction (XRD) patterns were collected on a Phillips PW1830 diffractometer using Co Kα radiation. Diffraction peaks were indexed according to the Inorganic Crystal Structure Database (ICSD). The crystallite sizes of the various investigated materials were obtained by employing the Debye−Scherrer equation. Phase composition was determined by refining the obtained data by the Rietveld method using the MAUD program.31 Photoemission data were acquired in a custom-designed ultrahigh-vacuum (UHV) system equipped with a VG MK II Escalab electron analyzer, working at a base pressure of 10−10 mbar. Core-level photoemission spectra were recorded at room temperature in normal emission using a nonmonochromatized Al/Mg twin-anode X-ray source, whereas for valence-band data, we used a He discharge lamp (Omicron HIS 15). Powder samples were suspended in bidistilled water and drop casted on high-purity copper foils. After being allowed to dry in air, the obtained films were introduced into the ultrahigh-vacuum system, outgassed overnight, and finally annealed in UHV for 20 min at 650 K. The charging observed during measurements of some samples was corrected using the adventitious carbon as an internal reference. The energy was calibrated using a gold sample (Au 4f at 84 eV) Elemental analysis was performed by means of X-ray fluorescence (XRF) spectroscopy. The samples were analyzed using an EDAX Eagle III energy-dispersive micro-XRF (μXRF) spectrometer equipped with a Rh X-ray tube and a polycapillary with a circular area of nominally 30-μm diameter. Data collection occurred at each point for a 250-s detector live time, with X-ray tube settings adjusted for 30% dead time.
Diffuse-reflectance ultraviolet−visible−near-infrared (DR UV−vis−NIR) spectroscopy was employed to characterize the absorption features over a range of wavelengths (250−2500 nm). These measurements were performed on fine powders of the samples in a cell with optical quartz walls. The spectra were collected in reflectance mode with a Perkin-Elmer Cary 5000 instrument equipped with an integrating sphere and then reported as an absorbance-like pattern by means of the Kubelka−Munk function. Optical band gap absorption was obtained by means of a Tauc plot of the Kubelka−Munk absorption as a function of the photon energy. Electron paramagnetic resonance (EPR) spectra were obtained using an X-band continuous-wave- (CW-) EPR Bruker EMX spectrometer equipped with a cylindrical cavity operating at 100 kHz field modulation. The measurements were carried out at 77 K in cells that can be connected to a conventional high-vacuum apparatus (residual pressure < 10−4 mbar).
3. RESULTS AND DISCUSSION 3.1. Composition and Structural Features of Fluorineand Niobium-Doped Titanium Dioxide. In the present article, we report results concerning two singly doped (with F and Nb) titania samples and four codoped materials. A sample of bare anatase (TiO2) prepared by the same procedure (section 2) was used for comparison. Results on a wider collection of doped and codoped samples are available in the Supporting Information. Table 1 presents some basic features of the samples examined here. The atomic ratios between the anionic dopant (F−) and titanium are relative to the composition of the liquid phase prior to gelification and do not represent the actual composition of the solid,14,30 because it is known that a fraction of fluorine is eliminated from the solid during the final calcination of the materials. The four codoped solids were prepared using a constant Nb concentration (Nb/ Ti = 0.05) and a varying concentration of fluorine. For the cationic dopant, the μXRF analysis showed a good correspondence between the nominal and final Nb/Ti contents for all Nb-containing samples (see Table 1). Figure 1 shows the XRD plot of all of the samples described in Table 1. Anatase is the unique phase in all samples except for Nb1 (Nb−TiO2, Figure 1c), which shows the presence of rutile 8464
dx.doi.org/10.1021/jp501203h | J. Phys. Chem. C 2014, 118, 8462−8473
The Journal of Physical Chemistry C
Article
Nb-doped titania reported in the literature, however, a blue shift of the band gap edge was observed, which was explained in terms of the Burstein−Moss effect.18,24,25,34−36 This effect is based on the filling of the conduction band by extra electrons that suppresses absorption at the band edge. Nevertheless, in some other cases, a red shift, as in the present work, has been reported for Nb−TiO2.37−41 There are also examples in which no change in the optical spectra was observed.42,43 An unambiguous interpretation of the red shift is not present in the literature. Some authors have explained this experimental evidence by invoking the presence of Nb2O5 microaggregates in the TiO2 matrix.38 Even though there is no trace of niobium oxides in XRD patterns of our samples (Figure 1) (which could be due to quantitative limits of the technique or to the presence of amorphous fractions), the fact remains that the Burstein− Moss effect does not occur in Nb-doped and Nb−F-codoped materials prepared by the sol−gel method, for which a red shift of the absorption edge was systematically observed. In the spectroscopic range between 450 and 650 nm, the spectral lines are flat for all samples, and there is no evidence of absorption in the visible range. The same occurs in the IR region until 2500 nm (vide infra). This result differs from what has been reported in some case for Nb-doped titania, such as in the recent article by De Trizio et al.,26 who obtained bluecolored materials. This considerable difference is very probably due to the partially reduced state of their colloidal samples (see section 3.6 on the optical properties of reduced materials). A similar situation was described by Hopper et al.,25 who prepared Nb−TiO2 suspensions by a hydrothermal reaction and also obtained blue materials that, however, definitely became white after treatment at 770 K or long air exposure. In the case of the materials described in the present article (having a relatively low Nb loading), despite the presence of Ti3+ reduced centers revealed by EPR spectroscopy, no optical absorption in the visible range occurred, and the samples, except for the pale yellow Nb1 (whose color is due to the band gap transition tail), were white. 3.3. EPR Spectroscopy of Doped and Codoped Materials. As recalled in the Introduction, aliovalent substitution (F or Nb) implies the presence of excess electrons in the system, which leads to the formation of Ti3+ centers by a mechanism of valence induction. These ions are paramagnetic, and the corresponding EPR spectra recorded at 77 K (Figure 3) were conclusively assigned in our previous work to reduced Ti3+ centers in regular lattice sites in the bulk of the anatase
and brookite, and the low-fluorine-content NbF25, which exhibits traces of rutile (Figure 1d). It is known, in fact, that the presence of fluorine allows the formation of a highly crystalline anatase phase. This state is achieved because F, inhibiting the anatase-to-rutile transition, allows relatively high temperatures of calcination to be reached. At these temperatures, a better quality of anatase crystals is obtained. This latter factor has been invoked as the reason for the good photocatalytic performances of fluorine-doped anatase powders.30 The solubility of fluorine in TiO2 is limited by two factors, however. The first is the mentioned elimination of this element during calcination, and the second is the formation of an extra phase amenable to TiOF2 occurring for high fluorine concentrations13 (see Supporting Information, Figure SI-1). The behavior of niobium is different from that of fluorine because Nb ions can be incorporated into TiO2 at a wider range of concentrations.32 The presence of niobium, however, does not completely suppress the formation of rutile even at relatively low calcination temperatures. The Nb−TiO2 sample (Nb1, Table 1 and Figure 1c) in fact contains about 10% w/w of rutile after calcination at 773 K. To obtain single-phase (anatase) materials, the drawbacks of the two single-doping procedures can be overcome using codoping with the two elements. Niobium is not removed at high temperature, ensuring a certain degree of n-doping, and the presence of fluorine stabilizes the anatase polymorph. 3.2. Optical Properties of (Oxidized) n-Type Doped TiO2. The optical properties of the various doped and codoped TiO2 samples are illustrated in Figure 2 by DR UV−vis spectra.
Figure 2. DR UV−vis spectra: (a) TiO2, (b) Nb1, (c) NbF25, (d) NbF35, (e) NbF45, (f) NbF25a.
All of the spectra are dominated by a strong absorption in the UV region due to the transition from the valence band to the conduction band, although the absorption threshold changes in the case of the doped systems. Except for NbF25a, which was prepared by a different procedure and showed an absorption edge that was basically coincident with that observed for pure TiO2, all samples showed a red shift of a few tenths of an electronvolt with respect to the absorption threshold of bare TiO2. The Nb− TiO2 (Nb1) sample exhibited a larger red shift, and the material showed a yellowish color (Figure 2, black line). This sample, however, was the only one containing appreciable amounts of the other two TiO2 polymorphs (Table 1), namely, brookite (with a band gap similar to that of anatase33) and rutile, which has a smaller band gap than anatase. In the majority of cases of
Figure 3. Normalized EPR spectra recorded at 77 K of n-type TiO2: (a) F2, (b) Nb1, and (c) NbF45. 8465
dx.doi.org/10.1021/jp501203h | J. Phys. Chem. C 2014, 118, 8462−8473
The Journal of Physical Chemistry C
Article
matrix.44 For both dopants, the signal is axial and characterized by g⊥ = 1.992, g∥ = 1.962. The same signal [hereafter Ti3+(I)] was observed in the case of the codoped materials (Figure 3). It is worth recalling that the samples whose spectra are reported in Figure 3 were fully oxidized materials (see Experimental Section) and that the presence of Ti3+ ions is uniquely due to the valence induction effect and not to a partial reduction of the solid (vide infra). These reduced titanium ions were welldispersed in the solid matrix (relatively narrow line width), and there was no evidence of appreciable amounts of surface reduced ions, which have different EPR parameters.20 Summarizing, both F−TiO2 and Nb−TiO2 can be described as n-type doped oxides (electrically neutral systems containing, with respect to the bare oxide, extra electrons conveyed by the presence of diluted aliovalent elements). Their formulas, which can be written, as a first approximation and on the basis of the valence induction effect, as Ti4+(1−x)Ti3+x F−x O2−(2−x) and Ti4+(1−2x)Nb5+xTi3+xO2−2, respectively, point to the stabilization of the extra electrons on lattice titanium ions as indicated by the EPR spectra, which are the same in the two cases. The codoped systems can be described by a combination of the two previous formulas. Another compensation mechanism based on the formation of Ti4+ vacancies, which has been proposed based on conductivity measurements,24 cannot in principle be excluded and will be discussed in section 4. The features of the EPR spectra, however, unambiguously indicate that at least a fraction of the dopant (F or Nb) releases excess electrons into the solid producing donor Ti3+ centers that modify therefore the electronic structure of the solid itself. 3.4. XPS Characterization. To investigate the compositional features and the electronic structure of the doped solids a detailed investigation based on photoelectron spectroscopy was carried out. In Figure 4 we report the core-level spectra of Ti 2p and Nb 3d: in all samples the Ti 2p3/2 peak maximum is centered at 459 eV, it is highly symmetric and, despite the EPR evidence reported above, does not show any significant presence of the shoulder connected to Ti3+ reduced species at 457.5 eV. Similarly, the Nb 3d photoemission line shows a highly symmetric doublet with the maximum centered at 207.5 eV, which nicely corresponds to Nb5+ species.26,45 This dopant is therefore present in the systems in its maximum oxidation state. A small shift at slightly lower binding energy was observed for sample Nb1 only. Fluoride dopants were observed in relatively small concentrations, and no impurities other than C (
