Article pubs.acs.org/IC
Unusual Changes in Electronic Band-Edge Energies of the Nanostructured Transparent n‑Type Semiconductor Zr-Doped Anatase TiO2 (Ti1−xZrxO2; x < 0.3) Daniel G. Mieritz, Adèle Renaud, and Dong-Kyun Seo* School of Molecular Sciences, Arizona State University, Tempe, Arizona 85287-1604, United States S Supporting Information *
ABSTRACT: By the establishment of highly controllable synthetic routes, electronic band-edge energies of the n-type transparent semiconductor Zr-doped anatase TiO2 have been studied holistically for the first time up to 30 atom % Zr, employing powder X-ray diffraction, scanning electron microscopy, transmission electron microscopy, nitrogen gas sorption measurements, UV/vis spectroscopies, and Mott−Schottky measurements. The materials were produced through a sol−gel synthetic procedure that ensures good compositional homogeneity of the materials, while introducing nanoporosity in the structure, by achieving a mild calcination condition. Vegard’s law was discovered among the homogeneous samples, and correlations were established between the chemical compositions and optical and electronic properties of the materials. Up to 20% Zr doping, the optical energy gap increases to 3.29 eV (vs 3.19 eV for TiO2), and the absolute conduction band-edge energy increases to −3.90 eV (vs −4.14 eV). The energy changes of the conduction band edge are more drastic than what is expected from the average electronegativities of the compounds, which may be due to the unnatural coordination environment around Zr in the anatase phase.
1. INTRODUCTION Because of its abundance, stability, and desirable electronic structure, TiO2 has been explored for various photochemical and photoelectrochemical applications including solar water splitting and as electrodes in devices such as dye-sensitized solar cells (DSSCs). In light of these promising traits, transition-metal dopants can be chosen to beneficially modify the optical and morphological properties of TiO2 for the applications. Among the various dopants, isovalent Zr4+ has been desirable because its vacant 4d orbitals with high energy may lead to a conduction band with a higher energy in Zrdoped TiO 2 (ZTO, or more precisely solid solution Ti1−xZrxO2) than that of TiO2.1 Supported films (coatings) of nanoporous ZTO have been explored as electrodes in devices such as DSSCs and UV detectors,2 owing to their higher conduction band and open-circuit voltage and improved responsivity to UV light. In DSSC applications, it has been indicated that Zr doping effectively raises the conduction band energy and thus increases the open-circuit potential of the devices.2a © XXXX American Chemical Society
Despite the importance of the material, systematic studies on how Zr doping affects the band energies of ZTO have been missing in the literature. More fundamentally, there have been inconsistencies in the correlations between the Zr-doping level and lattice parameters among different reports and also in the maximum level of Zr doping (or Zr substitution) in ZTO.2d,3 For example, the highest doping limits (∼30 atom %) have been reported based on powder X-ray diffraction (PXRD) studies on thin films of ZTO.3i However, a much lower maximum doping amount (7.5 atom %) has been reported for bulk samples through a systematic examination of the correlation between the Zr-doping level and the unit cell constants. Namely, the unit cell constants changed more or less linearly up to 7.5 atom % within the experimental error limit, following Vegard’s law.3h Despite this trend, the results were unusual because the c axis showed a decrease upon doping, while the a axis increased without an apparent reason or Received: March 26, 2016
A
DOI: 10.1021/acs.inorgchem.6b00712 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
2. EXPERIMENTAL SECTION
explanation. Moreover, the corresponding unit cell volume change was negligible and did not show Vegard’s behavior, unlike the unit cell parameters, even below the suggested maximum doping limit. These incoherencies in the literature reflect the lack of a congruous trend among the reported optical band gaps2d,3b,c,i,4 and peculiarly large2d,3i,4b,d or small4f band gaps even for undoped samples. Further anomalous and inconsistent reports of the Brunauer−Emmett−Teller (BET) surface area and pore characteristics exacerbate the difficulty of comparing the results from the literature in terms of the influence of the nanostructures from various synthetic procedures.3b,f,4a−c,e,f,5 One possible reason for the inconsistencies could be related to the fact that different works employed different synthetic methods, although they can all be categorized as sol−gel methods. Sol−gel methods are advantageous for the synthesis of nanostructured metal oxides because they allow relatively low calcination temperatures to retain the gel nanostructures without significant sintering and consolidation of gel network nanoparticles. Furthermore, as far as the synthesis of ZTO is concerned, sol−gel methods appear to allow a high degree of doping and maintain the anatase structure, unlike other methods that rely on coprecipitation or high-temperature calcination conditions.6 However, high-valent metal ions such as Ti4+ and Zr4+ are known to undergo rather uncontrollable hydrolysis and polycondensation in typical sol−gel synthetic conditions, undermining the intended advantages of sol−gel processes.7 Chelation of the metal ions with organic ligands may help in alleviating the problem, and yet it is not clear if the aqueous chemistries of Zr4+ and Ti4+ complexes are similar enough to provide the required homogeneity of metal-ion distribution after the gelation. The inhomogeneity at various levels will lead to inconsistent results in terms of the lattice parameters, particle sizes, and phase impurities in the products as well as their physicochemical properties. The epoxide method reported by Baumann et al. provides a convenient means to increase the pH at a controlled rate throughout the body of acidic metal precursor solutions in order to form a uniform gel.8 By initiation of the gelation with epoxides, the dissolved metal species with disparate reactivities gelate with a uniform composition at the atomic scale.9 The method has been adapted to incorporate (dope) Sb into SnO2 in the form of dense thin films with a relatively uniform distribution of the doping ions, resulting in the desired electrical properties.10 More recently, we reported the synthesis of highly porous Sb-doped SnO2 films, through a modified epoxide method where inorganic/polymer dual-gel-network structures are utilized to preserve the porosity of the gel to the greatest extent. The interpenetrating gel structures in the dual gel support each other during solvent removal and reduce the shrinkage while retaining 3D connectivity. By adapting the inorganic/polymer dual-gel method, herein we report systematic preparation and characterization studies of nanoporous ZTO both in bulk and in thin-film forms, with a doping content from 0 to 30 atom %. Various synthetic parameters including the calcination temperature have been examined to establish the possible doping concentration range while preserving the porosity of the materials. After firmly establishing the relationship between the doping levels and phase behavior, we observed a strong violation of the scissor relationship in the electronic band-edge energies of the ZTO with different doping levels based on optical and electrochemical impedance spectroscopy studies.
2.1. Materials Synthesis. Samples were prepared by varying the synthetic parameters of the amount of dopant and calcination temperature, which is denoted in Tables 1 and S1. In the sample
Table 1. Surface Areas and Pore Characteristics of the Productsa
sample
BET surface area (m2/g)
pore volume (cm3/g)
average pore sizeb (nm)
porosityc (%)
crystallite size (nm)d
0-500 10-500 20-500 30-500 0-600 10-600 20-600 30-600 0-500-noRF 10-500-noRF 20-500-noRF 30-500-noRF
64.3 136 148 136 44.9 113 78.7 84.4 2.1 140 128 149
0.12 0.26 0.23 0.22 0.10 0.24 0.20 0.22 0.003 0.15 0.10 0.05
7.2 7.5 6.1 6.3 8.5 8.4 10.3 10.5 3.2 3.5 3.2 2.8
33 48 49 45 29 49 43 48 1.1 38 29 17
18.9 ± 4.9 8.43 ± 1.2 14.3 ± 2.5 29.4 ± 3.8 60 21 24.3 ± 4.3 36 6.4 8.2 7.9
a
See the Experimental Section for the sample naming convention. The samples in bold are extensively discussed throughout the text. b4(total pore volume)/(BET surface area). cBased on porosity = pore volume/ (pore volume + 1/ρZTO), where ρTiO2 = 3.90 g/cm3 is assumed to be the density of TiO2, ρZTO1 = 4.06 g/cm3 is assumed to be the density of 10% ZTO, ρZTO2 = 4.19 g/cm3 is assumed to be the density of 20% ZTO, and ρZTO3 = 4.34 g/cm3 is assumed to be the density of 30% ZTO. dThe values with standard deviations were an average from three to five samples from different batches. nomenclature in the tables, the first number represents the atom % of Zr doping, which was varied from 0 to 30%, and the second number stands for the temperature of calcination. Various calcination temperatures were used up to 850 °C. Further detailed characterizations and data analysis are given for the samples from the lowest calcination temperatures that could produce pure phases because those samples have the highest porosity for our purpose. The concentrations of the organic components including resorcinol (R), formaldehyde (F), and poly(ethylene glycol) (PEG) were fixed to [R] = 0.45 M, [F] = 0.90 M, and [PEG] = 6.6 wt %, as was the total concentration of the metal ions ([Ti] + [Zr] = 0.50 M). The precursor amounts for the samples are given in Table S1. For a typical procedure, using sample 10-500 as an example, 6.24 g (18.3 mmol) of titanium n-butoxide [Ti(O-Bu)4; Mw = 340.32 g/mol; Sigma-Aldrich, 97%) was added to 14.4 g of absolute ethanol (Koptec, 200 proof) in a 70 mL glass bottle, and the solution was acidified with 2.49 g (27.5 mmol of HNO3) of concentrated nitric acid (EMD, 15.7 M). In a separate bottle, 0.48 g (2.0 mol) of ZrCl4 (233.04 g/mol; Alfa-Aesar, 99.5+%) was hydrolyzed by 3.20 g of deionized water (Caution! exothermic reaction). Once the solution cooled, it was added to the Ti(O-Bu)4/ethanol solution with swirling. A total of 2.98 g of PEG (Sigma-Aldrich, 15−20 kDa) and 2.00 g (18.2 mmol) of resorcinol (110.1 g/mol; Sigma-Aldrich, 99%) were dissolved with magnetic stirring for up to 1 h to yield an orange, clear solution. The solution bottle was placed in an ice bath for about 15 min before 10.51 g (113.6 mmol) of ±-epichlorohydrin (92.52 g/mol; Fluka, ≥99%) and subsequently 3.04 g (37.4 mmol of HCHO) of formaldehyde solution (Sigma-Aldrich, 37 wt % in water with 10−15 wt % methanol) were weighed into the solution. The solution was then put in the ice bath for another 10 min. It is noted that epichlorohydrin must be added prior to the formaldehyde solution. Otherwise, only precipitates would form instead of the desired monolithic gel. When the solution was taken out of the ice bath and gradually warmed to room temperature, the pH of the solution slowly increased B
DOI: 10.1021/acs.inorgchem.6b00712 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry from 0 to 4, which was accompanied by an increase in the viscosity and by a change from transparent orange to translucent deep red. After about 4 h, the solution formed a soft gel, which was then aged for 1 day to yield a translucent, maroon monolith. The color change suggests some degree of R−F polymerization. This gel was heated to 70 °C for 3 days to produce a hard, red-brown composite gel. The monolith was then broken up and dried ambiently for at least 1 day. Subsequently, this composite xerogel was heated in an ashing furnace (Carbolite, AAF1100) at a rate of 100 °C/h to 500 °C, where it was calcined for 10 h. To prepare films of the same materials, the precursor solution was deposited on a fluorine-doped tin oxide glass (FTO) (Hartford Glass, Tec 7) substrate. To prepare the substrate, FTO slides were washed with soapy water, rinsed well, and sonicated in an acetone bath for 5 min. A scotch tape mask was applied to cover three edges of an FTO slide to define both the coverage area and the thickness of the final nanoporous films. The doctor blading technique was employed to apply the film. A small amount of the precursor solution was dropped onto the edge opposite to the unmasked edge and drawn smoothly and quickly across the slide with a Pasteur pipet. The wet films were immersed in mineral oil (Spectrum, extra heavy) within about 10 s after the deposition and heated at 70 °C in a laboratory oven. After the films were taken out of the oil bath, the tape was removed, and residual mineral oil was rinsed off with hexanes before calcination. 2.2. Materials Characterization. Powder X-ray diffraction (PXRD) was conducted on a Bruker D8 diffractometer using Cu Kα radiation. Both primary and secondary goniometer radii were 205 mm, and the Si-strip linear position-sensitive detector had a forward dispersion slit angle of 1° and a beam spill of 12 mm. The crushed bulk sample was spread on a quartz plate, and the data presented were acquired by scanning 2θ from 5 to 90°, with a step size of ∼0.016° and a scan time of 60 min. The samples were premixed carefully with a Si powder internal standard (Deane K. Smith; X-ray diffraction accessories). The Si (111) reflection at 2θ = 28.443° was used as a reference for sample peak positions. The full-width half-maximum of the microcrystalline Si standard was measured to be 2θ = 0.1°, which was subtracted from the full-width half-maximum of the measured Bragg peak before application of the Scherrer equation for particle-size estimation. The unit cell constants of the anatase-type products were determined from all of the indexed Bragg reflections in 2θ = 15−90° by using the TOPAS 4.2 software program (Bruker). Elemental analysis was performed by inductively coupled plasma optical emission spectrometry (ICP-OES; Thermo iCAP6300). Nitrogen sorption isotherms were collected on a Micromeritics TriStar II 3020 surface area and porosity analyzer at 77 K. Samples were degassed under flowing nitrogen at 150 °C for 8 h. For the surface area calculation, the Brunauer−Emmett−Teller (BET) model was applied to the adsorption branch in the partial pressure range of 0.05−0.2 and all samples showed the C value in the range from 80 to 110. The Barrett−Joyner−Halenda (BJH) model was applied to the desorption branch to calculate the pore-size distribution using the Halsey thickness curve, heterogeneous surfaces, and Faass correction to account for multilayer desorption in estimating the thickness of the adsorbed nitrogen.11 The total pore volume was approximated by the total quantity of gas adsorbed at the data point closest to P/P0 = 0.98 on the desorption branch. Scanning electron microscopy (SEM) studies were performed on films on FTO and on dry-ground samples on C tape using an FEI XL30 environmental scanning electron microscope and 10 keV electrons. For transmission electron microscopy (TEM) and high-resolution TEM (HR-TEM) studies, samples were prepared by dry grinding and dusting onto TEM grids. HR-TEM and scanning TEM (STEM) images, as well as small-angle electron diffraction (SAED) patterns and electron-dispersive X-ray spectrometry (EDS) spectra, were collected on a JEOL 2010F microscope at an accelerating voltage of 200 kV. UV/vis absorption spectra of the transparent thin films were obtained on a Shimadzu UV-1601 UV/vis spectrophotometer using quartz cuvettes as sample holders, and the corresponding reflectance spectra were obtained on a PerkinElmer Lambda 18 spectrophotometer equipped with a Spectralon reflectance sphere accessory.
Electrochemical impedance spectroscopy (EIS) measurements were carried out on thin films on FTO glass and also on dense pellet electrodes immersed in a LiClO4 (1 M) aqueous solution with a Pt counter electrode and an Ag/AgCl reference electrode. Pellets were produced as working electrodes by pressing dried and well-ground powder samples (0-500, 10-500, and 20-600) around 6000 psi and sintering at 450 °C for 2 h in air. Then, each pellet was connected to a Cu wire with C paste, mounted in a chemical epoxy resin (CaldoFix-2 kit, Struers), and polished (SiC paper, grid 1200 and 4000) to give a smooth electrode surface. A potentiostat/galvanostat (CH Instruments, model 600 D) was used in the frequency range of 10 Hz to 100 kHz for the measurements, in which an alternating-current voltage (5 mV in amplitude and a frequency range of 1−10 kHz) was imposed in a potential range of −0.8 to +0.2 V vs Ag/AgCl.12 Subsequently, the interface semiconductor/electrolyte capacitance (C) was determined according to the imposed potential (E) using a simplified Randles equivalent circuit represented in Figure S1,13 where the diffusion (the Warburg) in the faradaic part was neglected at high frequencies (0.8− 10 kHz) because it was not visible in the Nyquist diagram. Thus, it consists of a series resistances (RS), representing mainly the faradaic and contact resistances and the resistance of the electrolyte, in series with a constant phase element (CPE), namely, a pseudocapacitance (Q), which is parallel with the charge-transfer resistance (RCT) associated with the electron transfer between the semiconductor and electrolyte. The impedance of the CPE is given by ZCPE = 1/Q(jω)α, where j, ω, and α represent the imaginary unit, the pulsation frequency, and the difference from an ideal capacitance (when α = 1), respectively. The flat-band potential (Ufb) and charge-carrier density (NA) were then deduced from the Mott−Schottky plot (CSC−2 vs E), which utilizes the Mott−Schottky equation for an n-type semiconductor: 1/CSC2 = 2/εε0eA2ND(E − Ufb − kT/e), where CSC is the capacitance in the space charge region of the semiconductor, A the interfacial surface area between the semiconductor electrode and the electrolyte, k the Boltzmann constant, T the temperature, e the electron charge, ε0 the vacuum permittivity, and ε the relative permittivity of the semiconductor. For calculations, CSC−2 was approximated to be C−2 because of the large capacitance of the Helmholtz layer, at the semiconductor surface in the electrolyte, in comparison to CSC. Semiconductor capacitance (C) was then calculated according to the following formula derived from Brug et al., and Mott−Schottky plots were obtained.14 (α − 1)/ α ⎛1 1 ⎞ + C = Q1/ α ⎜ ⎟ R CT ⎠ ⎝ RS
3. RESULTS AND DISCUSSION 3.1. Sol−Gel Synthetic Procedure. The overall synthetic procedure to produce nanoporous TiO2 and ZTO is depicted in Figure 1, which is adapted from Volosin et al.15 This procedure has been essential in the production of highly nanoporous metal oxides with uniform chemical compositions despite high-temperature calcination in a later step. Subsequent inorganic gelation and organic polymerization create a 3D network of interpenetrating hydrous metal oxide and R−F polymer gels throughout the solution. Drying and calcination of this composite gel produces the nanoporous metal oxides, during which the R−F polymer acts as a sacrificial buttress, keeping the porosity of the gel. The Ti(O-Bu)4 and ZrCl4 precursors are hydrolyzed in the acidified water/ethanol mixture solvent, and the initial pH of the precursor solution was adjusted to be about 0.15 based on the estimation from the concentrations of the precursors. When the pH was higher than 0.18, the precursor solution turned into an opaque gel within a minute, which needs to be avoided. This pH-dependent stability of the solution is consistent with a previous work by Chong et al., in which the acidified precursor solutions of C
DOI: 10.1021/acs.inorgchem.6b00712 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
OES results on two sets of independent sample batches. Although not detailed here, the TEM−EDS results were consistent with these results. 3.2. Effect of the Calcination Temperature on Phase Formation. The phase formation of TiO2 and 10%, 20%, and 30% ZTO samples was examined at various calcination temperatures between 500 and 850 °C by using PXRD studies (Figure S3). Diffraction patterns in Figure S3a show that the Bragg peaks and intensities of the undoped sample match the simulated pattern of anatase, shown with vertical red lines, and that the peaks become suddenly sharp above 600 °C. This nanocrystalline TiO2 exhibits typical behavior for sol−gelproduced TiO2, in that the phase transition to rutile (diffraction peaks indicated by R) becomes visible in PXRD with calcination at 600 °C and that the nanoparticles begin to sinter into a bulk material more significantly when held above 600 °C.21 The coinciding phase transition and coarsening are dependent, as described by Navrotsky et al., and difficult to pinpoint because the synthetic method strongly affects the thermal phase behavior of the TiO2 and ZTO materials.22 In our synthesis, the phase transition does not occur below 600 °C, and coarsening of the anatase product is well suppressed like in previous sol−gel syntheses.3b,c,5f,23 In contrast to TiO2, 10% ZTO undergoes almost no transformation into rutile TiO2 (diffraction peaks indicated by R),2d,3a−h,4c,d,f,5,23a,24 while a small average crystallite size (
