Photocatalytic hydrogen production with non-stoichiometric pyrochlore bismuth titanate

Catalysis Today 225 (2014) 102–110

Contents lists available at ScienceDirect

Catalysis Today journal homepage: www.elsevier.com/locate/cattod

Photocatalytic hydrogen production with non-stoichiometric pyrochlore bismuth titanate Oliver Merka a , Detlef W. Bahnemann b , Michael Wark a,c,∗ a b c

Lehrstuhl für Technische Chemie, Ruhr-Universität Bochum, Universitätsstr. 150, 44801 Bochum, Germany Institut für Technische Chemie, Leibniz Universität Hannover, Callinstr. 9, 30167 Hannover, Germany Institut für Chemie, Lehrstuhl Technische Chemie, Carl-von-Ossietzky Universität Oldenburg, Carl-von-Ossietzky-Str. 9-11, 26129 Oldenburg, Germany

a r t i c l e

i n f o

Article history: Received 21 June 2013 Received in revised form 1 September 2013 Accepted 5 September 2013 Available online 22 October 2013 Keywords: Photocatalytic hydrogen production Non-stoichiometry Bismuth titanate Pyrochlore Stability

a b s t r a c t Pyrochlore bismuth titanate (Bi2−x Ti0.75x )Ti2 O7 was prepared by an aqueous sol–gel method and annealed at different temperatures. Non-stoichiometry was obtained by adjusting the Ti/Bi ratio during synthesis. Additionally the two bismuth titanate phases Bi4 Ti3 O12 and Bi2 Ti4 O11 were synthesized. The materials were characterized by X-ray diffraction and UV–vis reflectance spectroscopy. Photocatalytic hydrogen production was tested in the presence of methanol as sacrificial agent after loading the bismuth titanates with nanoparticles of platinum acting as co-catalysts. It was found that stoichiometric Bi2 Ti2 O7 being completely inactive could be turned into a good photocatalyst by increasing the Ti/Bi ratio to 1.50 corresponding to (Bi1.55 Ti0.33 )Ti2 O7 still crystallizing in a pure pyrochlore phase. The increase in activity is supposed to derive from an optimization of the TiO6 -octahedral geometry due to the generation of bismuth and oxygen vacancies in the lattice. If platinum is applied as co-catalyst on (Bi2−x Ti0.75x )Ti2 O7 samples, stability issues occur during photocatalysis, which are suppressed by coating the platinum particles with a Cr2 O3 shell. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Bismuth titanate is a polymorphic material, which crystallizes in a variety of different phases. Because of being promising for electrooptic devices [1] and non-volatile memory materials, the most intensively studied compound in the series of bismuth titanates is the ferroelectric Aurivillius phase of Bi4 Ti3 O12 . The other stable phases of dielectric pyrochlore Bi2 Ti2 O7 [2] and monoclinic Bi2 Ti4 O11 [3] being antiferroelectric (␣-modification) or paraelectric (␤-modification), respectively, have attracted less attention. The above listed compounds as well as the bismuth rich Bi12 TiO20 [4] are the well-known phases of bismuth titanate. The existence of another compound, namely Bi2 Ti3 O9 , is controversially discussed. Yordanov et al. reported on its existence as they measured an Xray diffraction pattern being significantly different from those of a proposed physical mixture composed of Bi2 Ti2 O7 and Bi2 Ti4 O11 [5]. However, a suitable crystal structure was not discussed and by Zaremba such a Bi2 Ti3 O9 stoichiometry could not be confirmed [6]. As photocatalysis is supposed to be a key technology for solving environmental problems [7], all of the above mentioned bismuth

∗ Corresponding author at: Institut für Chemie, Lehrstuhl Technische Chemie, Carl-von-Ossietzky Universität Oldenburg, Carl-von-Ossietzky-Str. 9-11, 26129 Oldenburg, Germany. Tel.: +49 441 798 3675. E-mail address: [email protected] (M. Wark). 0920-5861/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.cattod.2013.09.009

titanate phases have been tested for photocatalytic activity, but in these tests mainly dye degradation reactions have been considered [8–11]. Regarding literature, only one study on photocatalytic hydrogen production using bismuth titanate exists: Kudo and Hijii [12] reported that the pyrochlore Bi2 Ti2 O7 as well as the Aurivillius phase Bi4 Ti3 O12 were almost inactive yielding either no or only an insignificant hydrogen production rate of 0.6 ␮mol H2 /h, respectively. Contrary to that, Murugesan et al. [13] conducted a DFT study on cubic pyrochlore Bi2 Ti2 O7 proving that all requirements are met for being a good photocatalyst for hydrogen production and even water splitting. To be a suitable photocatalyst in hydrogen production, the conduction band potential of the semi-conductor must be more negative than the potential of the redox couple H+ /½ H2 to provide photogenerated electrons with a suitable power to reduce protons to molecular hydrogen. In most cases, the photocatalyst itself does not provide suitable active sites for the generation of hydrogen and a co-catalyst has to be loaded on the surface. Precious metals like platinum [14], rhodium or silver [15] creating electronic sinks on the surface inhibit the recombination of photogenerated electron–hole pairs in the bulk effectively. Organic sacrificial reagents like methanol are widely used to determine the photocatalytic activity in hydrogen production. The methanol is reformed to carbon dioxide via the considerably stable intermediates of formaldehyde and formic acid [16]. This paper deals with the preparation of bismuth depleted bismuth titanates to generate vacancies in the cationic

O. Merka et al. / Catalysis Today 225 (2014) 102–110

sub-lattice. Non-stoichiometric chemical compositions are obtained by the variation of the Ti/Bi ratio in the preparation process. Although the samples are depleted in bismuth, the band gap energy is situated appropriately to absorb in the visible light region. The vacancy structure in non-stoichiometric bismuth titanate pyrochlore “Bi2 Ti2 O7 ” is analyzed by Rietveld refinement. The samples are tested for activity and stability in the hydrogen producing test reaction. The two bismuth titanate phases Bi4 Ti3 O12 and Bi2 Ti4 O11 providing Ti/Bi ratios, which are not obtainable as pyrochlore structure, were tested for comparison as well. 2. Experimental 2.1. Catalyst preparation Samples were prepared by a modified aqueous Pechini sol–gel route [17] with citric acid monohydrate and ethylenediaminetetra-acetic acid (EDTA) acting as complexing reagents. Non-stoichiometric compositions of (Bi2−x Ti0.75x )Ti2 O7 were achieved by adjusting the Ti/Bi ratio in the range from x = 1.00 to 1.50. Stoichiometric samples of the other bismuth titanates Bi4 Ti3 O12 and Bi2 Ti4 O11 were prepared by using Ti/Bi ratios of 0.75 and 2.00, respectively. A high accuracy of the chemical catalyst composition is ensured by using a diluted Bi(NO3 )3 ·6H2 O precursor solution. The ratio of overall metal cations to EDTA and to citric acid was adjusted to be 1:1.05:1.5. In a typical synthesis of stoichiometric Bi2 Ti2 O7 , 7.84 g EDTA and 8.06 g citric acid monohydrate were dissolved in a solution containing 550 mL distilled water and 18.5 mL of an aqueous 30% ammonia solution. To that 4.5 mL of 65% nitric acid and 50 mL of the Bi-precursor solution containing 12.78 mmol Bi were added. For the non-stoichiometric samples the contents of titanium, EDTA and citric acid were increased accordingly. Pre-dilution of Ti(Oi Pr)4 in absolute ethanol was inevitable to dissolve the titanium source successfully in the aqueous medium. 150 mL of the titanium precursor solution containing 12.78 mmol (3.63 g) of Ti(Oi Pr)4 were added to the mixture. The clear solution was evaporated at 373 K for 18 h to ca. 50 mL of a light yellowish gel, which was transferred into a calcination basin from Haldenwanger (type 130). Complete decomposition into black amorphous powder was achieved in a heating mantle at ca. 573 K. Due to the highly exothermal decomposition reaction the temperature in the heating mantle could not be maintained and increased to about 673 K. Finally, the amorphous samples were annealed in static air at elevated temperatures between 773 K and 1073 K for 4 h (heating ramp: 100 K/h). 2.2. Characterization The crystalline phases were determined by X-ray powder diffraction at room temperature with Cu K␣ radiation on a Bruker D8 Advance instrument in the 2 range of 10–80◦ using a step size of 0.02◦ and 0.5 s per step. The Rietveld refinement was performed using Bruker TOPAS 4.2. The octahedral geometry was analyzed by Diamond 3.1f. Average crystallite sizes were calculated within the Rietveld refinement applying the Scherrer equation [18]. UV–vis diffuse reflectance spectra were measured on a Varian Cary 4000 UV-vis spectrophotometer. The band gap energies were calculated by analysis of the Tauc plots resulting from Kubelka–Munk transformation of the diffuse reflectance spectra. 2.3. Co-catalyst loading and photocatalytic tests The photocatalytic activities of the bismuth titanates were investigated by detecting the hydrogen production rate from 75 mL of an aqueous solution containing 10 vol% of methanol (2.47 mol/L) as sacrificial agent. As the blank materials were inactive, platinum

103

[19] was chosen as co-catalyst; a quite high co-catalyst loading of 3.0 wt% had to be used, as activities were found to decline at lower loadings. The reactions were performed in a 100 mL flow-through quartz reactor (continuous argon flow rate of 20 mL/min) equipped with a cooling mantle maintaining the temperature of the reaction mixture at 288 K. An LOT Oriel 300 W Xenon arc lamp equipped with a 10 cm water filter bath was used for irradiation. In a typical run, 75 mg of the photocatalyst was added to the methanol/water mixture containing 6.0 mg of H2 PtCl6 ·6H2 O previously prepared as highly diluted solution. The suspension was left at its natural pH value of 3.45 and was de-aerated by bubbling argon gas over a period of 30 min. A control measurement proving oxygen free conditions was performed after de-aeration. The gas stream leaving the reactor was analyzed at intervals of 30 min by gas-chromatography. A Shimadzu GC-2014 is applied as gas-chromatograph equipped with a special valve setup normalizing the gas pressure in the sample loop (500 ␮L) prior to the sample injection into the column. A micro-packed column (ShinCarbon ST 100/120, Restek) was used. The photodeposition of the Pt/Cr2 O3 core–shell co-catalyst system was performed in the same 100 mL quartz reactor used for the hydrogen producing test reaction. After de-aerating the solution for 30 min and performing a control measurement, the solution is irradiated for 30 min to deposit the platinum core (3.0 wt%) followed by a measurement of the hydrogen production rate for comparison. The illumination is intercepted and 3.36 mL of an aqueous solution containing 16.8 mg K2 CrO4 were added to provide a maximum Crloading of 4.5 wt%. After completed Pt or Pt/Cr2 O3 photodeposition, the photocatalyst is filtrated, thoroughly washed with distilled water and dried overnight. The residual chromium solution is analyzed by UV–vis spectroscopy at a wavelength of 370 nm to determine the actual amount of chromium photodeposited. The hydrogen production test reaction was performed in the same way described above, using 75 mg of the core–shell modified catalyst. To check the reproducibility of the photocatalytic test reaction, of the synthesis and finally, the photodeposition method, the (Bi1.55 Ti0.33 )Ti2 O7 773 K, 8 h sample was synthesized in two different batches. The first batch was tested once and the second batch twice. The obtained results were found to differ by around 10–15%. These deviations in activity are supposed to derive mainly from the in situ photodeposition method. 2.4. Actinometry and photonic efficiencies The photon flux of the LOT Oriel 300 W Xenon arc lamp was quantified by using the standard potassium ferrioxalate actinometer as proposed by Hatchard and Parker [20]. The 0.02 mol/L actinometer solution was capable of absorbing photons up to a wavelength of 485 nm completely, if applied in the quartz reactor used during photocatalysis featuring an optical path length of 80 mm. By using a quantum efficiency of 1.1, a photon flux of 1.48 × 10−6 mol/s is measured at full arc irradiation. For TiO2 anatase a current-doubling mechanism is reported to occur in the photocatalytic hydrogen production in presence of methanol while working under oxygen free conditions. Methanol is most likely oxidized to a CH2 OH radical intermediate, which is able to inject an electron into the conduction band of the TiO2 to form HCHO [21]. As the conduction bands of the investigated bismuth titanates consist of unoccupied Ti 3d orbitals as well, electron injection should be also possible here. Formaldehyde is further oxidized to HCOOH and finally to CO2 [22,23]. Both of these processes may involve currentdoubling once again. In principle, current-doubling facilitates the formation of one molecule H2 , which involves a two electron reduction, by the absorption of just a single photon. Current-doubling

104

O. Merka et al. / Catalysis Today 225 (2014) 102–110

Fig. 2. XRD patterns and UV–vis diffuse reflectance spectra (inset) of “Bi2 Ti2 O7 ” samples (Ti/Bi = 1.50) annealed at 773 K for 4 h (a), 6 h (b) and 8 h (c). The degree of crystallinity is improved with prolonged annealing.

Fig. 1. XRD patterns of various bismuth titanates Bi4 Ti3 O12 (a), non-stoichiometric “Bi2 Ti2 O7 ” (Ti/Bi = 1.40) (b) and Bi2 Ti4 O11 (c) annealed at 873 K, 773 K and 1073 K for 4 h, respectively.

does not change the definition of the photonic efficiency, but leads to the uncertainty, whether one or two photons are needed to produce one molecule of H2 . As radical intermediates might be oxidized by photogenerated holes in the normal photocatalytic manner as well, the actual amount of needed photons will most likely lie in between one and two. The photonic efficiency is calculated by assuming one photon per H2 and thus, is regarded as a lower limit: =

H2 production rate in␮mol/h × 100. photon flux in␮mol/h

3. Results and discussion 3.1. Sample characterization Several phases of bismuth titanate with different Ti/Bi ratios were synthesized. A Ti/Bi ratio of 0.75 and annealing at 873 K leads to the perovskite structure of Bi4 Ti3 O12 . The corresponding X-ray diffraction pattern is shown in Fig. 1(a). However, the paraelectric tetragonal phase (symmetry group (SG): I4/mmm) and the ferroelectric orthorhombic phase (SG: B2cb) of Bi4 Ti3 O12 [25] generate almost the same X-ray pattern making a decision about the present modification difficult. Fig. 1(c) shows the diffraction pattern of Bi2 Ti4 O11 , which is obtained by adjusting the Ti/Bi ratio to 2.00 at annealing temperatures as high as 1073 K. Lattice parame˚ b = 3.81 A, ˚ c = 14.96 A˚ and ˇ = 93.18◦ were found ters of a = 14.61 A, for monoclinic ␣-Bi2 Ti4 O11 (SG: C2/c), which are in good accordance to the literature [3]. At lower temperatures a mixture of phases is obtained as shown in Fig. S1. At 873 K, non-stoichiometric pyrochlore “Bi2 Ti2 O7 ” is formed together with TiO2 rutile, and at 973 K the Bi2 Ti4 O11 phase is spiked with a small amount of the pyrochlore phase. Kidchob et al. [26] found a similar behavior.

According to reference [2], a pure pyrochlore phase of Bi2 Ti2 O7 is difficult to achieve at a Ti/Bi ratio of 1.00. Actually, we also did not succeed in preparing stoichiometric Bi2 Ti2 O7 . The sample was highly amorphous at an annealing temperature of 673 K, and at higher temperatures of 773 K and 873 K the byproduct Bi4 Ti3 O12 appeared as shown in Fig. S2. Nevertheless, non-stoichiometric pure pyrochlore phases were found to form in a wide range of Ti/Bi ratios from 1.20 to 1.50. The cubic pyrochlore structure (SG: Fd3m) has the ideal composition of A2 B2 O6 O with a seventh oxygen occupying the O -site. Bismuth is placed on the A-site arranged in the center of a puckered hexagonal bipyramid of oxygen [27]. The titanium ions in the B-site are sixfold coordinated by oxygen forming a network of corner-sharing octahedra. Due to the titanium excess, vacancies are formed at the Bi-site and potentially at the O -site, which may have a stabilizing effect on the structure. Exemplarily for the non-stoichiometric “Bi2 Ti2 O7 ” series, the X-ray diffraction pattern of the “Bi2 Ti2 O7 ” (Ti/Bi = 1.40) sample annealed at 773 K for 4 h is shown in Fig. 1(b). An annealing time of 4 h leads to a complete crystallization of non-stoichiometric “Bi2 Ti2 O7 ” samples featuring Ti/Bi ratios of up to 1.40. Higher titanium excesses of 45% or 50% results in partially amorphous samples. Fig. 2 displays the diffraction data of the Ti/Bi 1.50 sample of “Bi2 Ti2 O7 ” annealed for different periods of time. An increased background is obvious, if annealing is conducted for only 4 h. A period of 8 h is needed for entire crystallization. The UV–vis diffuse reflectance spectra presented in the inset of Fig. 2 show a slight shift of the band gap energy into the visible region giving a pale yellow color, if the sample is annealed for only 4 h. Whereas entirely crystalline samples were almost white, such a pale yellow color was also obtained for largely amorphous samples calcined at only 673 K. In conclusion, a phase of Bi2 Ti3 O9 as predicted by Yordanov et al. [5] for a Ti/Bi ratio of 1.50 was not found for annealing temperatures ranging from 673 K to 873 K. In fact, a pure pyrochlore phase of non-stoichiometric “Bi2 Ti2 O7 ” is obtained after specific heat treatment at 773 K for 8 h. The lattice parameter of non-stoichiometric “Bi2 Ti2 O7 ” is found to be a function of the Ti/Bi ratio. As shown for the series of samples annealed at 773 K for 4 h the lattice parameter is reduced from 10.37 A˚ for a stoichiometric composition to 10.33 A˚ for “Bi2 Ti2 O7 ” featuring a Ti/Bi ratio of 1.50 (cf. to Fig. 3). The increasing degree of vacancies is considered to be responsible for the lattice contraction. This behavior is found for samples annealed at a higher temperature of 873 K only up to a Ti/Bi ratio of 1.30. Higher titanium excess does not longer lead to a pyrochlore being more enriched in titanium and vacancies, but to the formation of a TiO2 rutile byproduct.

O. Merka et al. / Catalysis Today 225 (2014) 102–110

Fig. 3. The lattice parameter of non-stoichiometric “Bi2 Ti2 O7 ” samples annealed at 773 K and 873 K as a function of the Ti/Bi ratio.

The lattice parameters of the “stoichiometric” Bi2 Ti2 O7 samples are lower than expected, which is easily explained by the presence of a Bi4 Ti3 O12 byproduct (as shown in Fig. S2), so that the pyrochlore structure is in fact not stoichiometric, but depleted in bismuth. The high amount of Bi4 Ti3 O12 in the stoichiometric composition annealed at 873 K leads to a pyrochlore being even more depleted in bismuth than the sample with a 20% excess of titanium. Incorporation of bismuth in mixed oxide photocatalysts is a promising way to achieve visible light response. The hybridization of fully occupied Bi 6s and O 2p orbitals [28] is known to shift the top of the valence band to more negative values (vs. NHE) lowering the band gap energy in comparison to, e.g., yttrium titanates [29]. The conduction band is formed by Ti 3d orbitals. Fig. 4 shows the Kubelka–Munk transformation of the UV–vis diffuse reflectance spectra of the herein investigated bismuth titanates. Close to the absorption edge, the band gap energy Ebg is directly proportional to the reciprocal optical absorption coefficient ˛. This empirical relation is known as the Tauc plot [30–32]: a · hv ∝ const(hv − Ebg )

n

The reciprocal optical absorption coefficient ˛ itself is proportional to the Kubelka–Munk function F(R∞ ), so that the following equation results: (F(R) · hv)

1/n

∝ const(hv − Ebg )

Fig. 4. Tauc-plots (for indirect bad gap transition) calculated from the UV–vis diffuse reflectance spectra for band gap energy calculations of Bi4 Ti3 O12 (a), “Bi2 Ti2 O7 ” (Ti/Bi = 1.40) (b) and Bi2 Ti4 O11 (c) annealed at 873 K, 773 K and 1073 K for 4 h, respectively.

105

For semiconductors providing an indirect band gap transition, n is defined to be 2, while it values n = 1/2 for direct band gap semiconductors. Samples of Bi2 Ti4 O11 are supposed to reveal an indirect band gap transition. Wei et al. reported about an indirect band gap transition for Bi2 Ti2 O7 [24]. The tetragonal phase of Bi4 Ti3 O12 provides an indirect transition as well, while the orthorhombic phase was calculated to yield a direct band gap transition [25]. As both phases are difficult to differentiate, band gap transitions were assumed to be indirect for all samples. Thus, the band gap energies are obtained from the intersection point of the linear part of the plot (F(R)·h)1/2 vs. h with the energy axis. The non-stoichiometric “Bi2 Ti2 O7 ” samples feature almost the same band gap energies if entirely crystallized. Energy band gaps of 3.2 eV, 3.0 eV and 3.1 eV were obtained for Bi4 Ti3 O12 , “Bi2 Ti2 O7 ” and Bi2 Ti4 O11 featuring Ti/Bi ratios of 0.75, 1.00–1.50 and 2.00, respectively. Expectably, a clear dependency between the Ti/Bi ratio and the band gap energy is not found, as the crystal structure has an important influence, too. Wei et al. [24] calculated the band gap energy for Bi2 Ti2 O7 by DFT studies to 2.46 eV, but explained that this value is underestimated and should be around 3 eV. This value corresponds well to our experimental results.

3.2. The vacancy structure in non-stoichiometric “Bi2 Ti2 O7 ” In order to check the influence of variation in the Ti/Bi ratio on the structural properties and thus, on the photocatalytic activity, a Rietveld refinement including the refinement of atomic positions was conducted on the non-stoichiometric series of “Bi2 Ti2 O7 ” samples annealed at 773 K. The Rietveld refinement unveils the location of vacancies in the structure, which have to emerge as a consequence of non-stoichiometry. Moreover, an increasing amount of vacancies in the structure is supposed to change atomic distances and angles. In the pyrochlore structure A2 B2 O6 O with space group Fd3m (No. 227) the crystallographic origin is usually placed on the Wyckoff site 16c (with coordinates of 0, 0, 0). Beside the 16c site also the Wyckoff sites of 16d (1/2, 1/2, 1/2) and 8a (1/8, 1/8, 1/8) are highly symmetric, i.e., these sites do not have any positional parameters. For undistorted pyrochlores like Y2 Ti2 O7 , the A cations are placed on the 16c site, the B cations on 16d and the O anion on 8a. Only the O anions occupy sites with a positional parameter, namely 48f (x, 1/8, 1/8) [33]. Four structural models for non-stoichiometric “Bi2 Ti2 O7 ” are suggested and refined by the Rietveld method. The models differ in terms of vacancy structure and atomistic distortion caused by bismuth cations as shown in Table 1. In the first structure suggestion (structure 1) an undistorted structure model was utilized for the bismuth titanate series. Moreover, the different occupation numbers for bismuth and the O oxygen in non-stoichiometric compositions were not taken into account, i.e., Bi and O sites are considered to be fully occupied and no vacancies are allowed in the structure. This structure model was only chosen as an initial approach, as the disregard of non-stoichiometric compositions cannot lead to realistic results. The second suggestion (structure 2) considers the stoichiometry changes due to different Ti/Bi ratios. This model predicts the generation of vacancies at the A site combined with vacancies at the O site due to charge neutrality. The chemical composition is (Bi2−x x )Ti2 O7−1.5x 1.5x . Such a stoichiometry is often proposed for non-stoichiometric pyrochlores featuring an A-cation with inert electron pair [34,35]. All ions are placed on the highly symmetric atomic positions used in structure 1 as well. In structure suggestion 3 the vacancy model was kept as it is in structure 2, i.e., showing A, O -site vacancies, but bismuth as well as O oxygen are symmetrically displaced from their ideal positions. Large bismuth ions providing an inert electron lone pair are well known to displace off

106

O. Merka et al. / Catalysis Today 225 (2014) 102–110

Table 1 Structure models for the Rietveld refinement of the “Bi2 Ti2 O7 ” series annealed at 773 K. The structures are different in the predicted stoichiometry and vacancy distribution.

1 2 3 4

Stoichiometry

Vacancies ()

Stoichiometric Non-stoichiometric Non-stoichiometric Non-stoichiometric

No A and O A and O A

A site

B site

O site

O site

Bi



Ti

Ti

O

O



16c 16c 96g 96g

No Yes Yes Yes

– – – 16c

16d 16d 16d 16d

48f 48f 48f 48f

8a 8a 48f 8a, 48f

No Yes Yes No

Bi2 Ti2 O7 (Bi1.33 0.67 )Ti2 O6 1 (Bi1.33 0.67 )Ti2 O6 1 (Bi1.55 Ti0.33 0.11 )Ti2 O7

The Wyckoff positions are characterized by the following coordinates: 16c (0, 0, 0), 16d (1/2, 1/2, 1/2), 48f (x, 1/8, 1/8), 8a (1/8, 1/8, 1/8) and 96g (x, x, z).

high symmetry positions. Esquivel-Elizondo et al. [2] performed DFT calculations on Bi2 Ti2 O7 predicting the ideal displacement of bismuth on the 96g site to be (0.015, 0.015, 0.964). Each bismuth ion is located at one of six possible equivalent positions around the 16c site. As a consequence of bismuth displacement the O anions are distorted as well leading to 6 possible positions around the 8a site. Coordinates of (0.136, 1/8, 1/8) are utilized [2]. The chemical structure stays the same as in structure 2. In the structure suggestion 4 the distorted sites for bismuth and O oxygen are used, but it is predicted that the charge neutralization due to the bismuth depletion is ensured by titanium partially occupying the bismuth position. Thus, vacancies are only formed on the Asite, but no longer on the O site, which leads to a significantly reduced amount of overall vacancies. The general chemical structure is (Bi2−x Ti0.75x 0.25x )Ti2 O7 . By DFT calculations Stanek et al. proposed such a vacancy structure as being energetically favored for pyrochlores with BO2 excess [36]. The A-site titanium is placed on the 16c site and the corresponding degree of O oxygen occupies the undistorted 8a site, as no distortion is expected for titanium being a very small cation. The occupancies used in the refinements for the structure suggestions 2–4 are determined in order to fulfill the aforementioned restrictions in non-stoichiometric bismuth titanate and are listed in Table S1. For the refinement of structure 1, all occupancies were set to one. The Rietveld refinement provides the evolution of the x positional parameter of oxygen as a function of the Ti/Bi ratio for each structure suggestion. Considering the facts that titanium ions are rigid in the lattice and that the TiO6 octahedral network is build up only from O anions, a change in the x positional parameter of the O anion will change the octahedral geometry. The results are summarized in Table 2 also showing the corresponding goodness of fit (GOF). The GOF value is given as the quotient of the weighted profile R-factor Rwp and the expected R-factor Rexp and values to 1, if the applied model is in complete accordance with the measured X-ray diffraction pattern. Measurement parameters like step size and time per step may have a significant influence on the Rexp . In general, it can be seen that the x positional parameter of oxygen decreases with increasing Ti/Bi ratio. This trend maintains irrespectively of the applied structure model. By means of the high GOF values it is obvious that structure 1 is, as expected, not an applicable approach for describing non-stoichiometric bismuth titanate. However, the results obtained for structure 2 do not lead to a much better fit, as the structure is not sufficiently described by just including the non-stoichiometry. In fact, atomistic distortion of bismuth and O oxygen has to be taken into consideration which is proven by the results achieved for structure 3, indicating a significantly improved fit. Finally, the vacancy structure in non-stoichiometric pyrochlores is analyzed. Structures 3 and 4 differentiate between the compositions of (Bi2−x x )Ti2 O7−1.5x 1.5x and (Bi2−x Ti0.75x 0.25x )Ti2 O7 with vacancies on the A,O -sites or solely on the A site, respectively. As the GOF values obtained for structure 4 are significantly lower than those for structure 3, vacancies only at the A site are clearly favored. Especially at high Ti/Bi ratios of ≥1.35 corresponding to samples possessing a high amount of vacancies, the GOF values are greatly improved.

One should not pay too much attention to the poor GOF values obtained for the Bi2 Ti2 O7 sample with a Ti/Bi ratio of 1.00, as this sample is not really stoichiometric, but contaminated with a significant amount of Bi4 Ti3 O12 byproduct making the fitting more deficient. Exemplarily, the data of the Rietveld refinement of nonstoichiometric “Bi2 Ti2 O7 ” featuring a Ti/Bi ratio of 1.50 annealed at 773 K for 8 h is shown in Fig. 5. In conclusion, the structure of non-stoichiometric pyrochlore bismuth titanate is described in the best way by considering atomistic distortions of bismuth and O oxygen and vacancies solely on the A site leading to a composition of (Bi2−x Ti0.75x 0.25x )Ti2 O7 (structure suggestion 4). 3.3. Photocatalytic hydrogen production The samples were tested in photocatalytic hydrogen production with methanol as sacrificial reagent involving the in situ photodeposition of 3.0 wt% of platinum. The maximum photocatalytic activities of the (Bi2−x Ti0.75x )Ti2 O7 samples annealed at 773 K for 4 h in the Ti/Bi range of 1.00 to 1.50 are presented in Fig. 6. Maximum hydrogen production rates are achieved after complete deposition of the platinum co-catalyst. The stoichiometric composition consisting of (Bi2−x Ti0.75x )Ti2 O7 and a small amount of Bi4 Ti3 O12 byproduct was found to be almost completely inactive. The same counts for samples up to a Ti/Bi ratio of 1.25. Initially, the pure pyrochlore phase of (Bi1.71 Ti0.22 )Ti2 O7 (Ti/Bi of 1.30) shows a measurable hydrogen production rate. The hydrogen production rate is enhanced considerably by stepwise increase of the Ti/Bi ratio

Fig. 5. Rietveld refinement of non-stoichiometric “Bi2 Ti2 O7 ” featuring a Ti/Bi ratio of 1.50 annealed at 773 K for 8 h. The main part shows the raw data and the calculated fit for structure 4. The difference plots for all four structures are given in the lower part each providing the same scale ranging from −3000 to +3000 counts.

O. Merka et al. / Catalysis Today 225 (2014) 102–110

107

Table 2 Results of the Rietveld refinement for structures 1–4 each describing a slightly different structure of “Bi2 Ti2 O7 ” by systematically combining non-stoichiometry (structures 1–3), atomistic distortion of bismuth and O oxygen (structures 3–4), and different vacancy distributions (structures 2–4). Structure

1

Bi2 Ti2 O7

2

(Bi2−x x )Ti2 O6 1−1.5x

3

(Bi2−x x )Ti2 O6 1−1.5x

4

(Bi2−x Ti0.75x 0.25x )Ti2 O7

Ti/Bi ratio

x GOF x GOF x GOF x GOF

1.00

1.20

1.25

1.30

1.35

1.40

1.45*

1.50*

0.4692 2.58 0.4692 2.58 0.4621 2.77 0.4689 2.56

0.4520 3.20 0.4254 2.97 0.4336 2.15 0.4499 2.19

0.4462 3.51 0.4173 3.31 0.4291 2.08 0.4444 2.10

0.4502 3.31 0.4168 2.95 0.4279 2.12 0.4416 2.08

0.4350 3.92 0.4005 3.84 0.4210 2.51 0.4363 1.89

0.4290 2.90 0.3901 2.95 0.4157 2.20 0.4325 1.60

0.4260 2.89 0.3813 3.00 0.4142 2.38 0.4305 1.58

0.4262 2.84 0.3719 3.03 0.4131 2.56 0.4296 1.62

The x stands for the x positional parameter of O oxygen and GOF for the goodness of fit. Samples with Ti/Bi ratios ≤1.40 were annealed at 773 K for 4 h, the samples with Ti/Bi ratios of 1.45* and 1.50* at 773 K for 6 h and 8 h, respectively.

to 1.40. Interestingly, further increase of the titanium content to Ti/Bi ratios of 1.45 and 1.50 leads to pure pyrochlore phases, but decreases the photocatalytic activity, if annealed for only 4 h. Both samples were proven to be partially amorphous (cf. to Fig. 2). For a complete crystallization, (Bi1.59 Ti0.31 )Ti2 O7 and (Bi1.55 Ti0.33 )Ti2 O7 featuring Ti/Bi ratios of 1.45 and 1.50 had to be annealed for 6 h and 8 h, respectively. The influence of the annealing time on the photocatalytic activity of (Bi1.55 Ti0.33 )Ti2 O7 is presented in Fig. S3. The photocatalytic activity is improved with increasing degree of crystallinity. An optimum annealing time was found for every Ti/Bi ratio, as the photocatalytic activity is again decreased by even longer annealing times (cf. to inset of Fig. 6). All pyrochlore bismuth titanate phases provide average crystallite sizes of about 70 nm calculated from X-ray diffraction by Scherrer equation. Thus, differences in the crystallite size or in the surface area are not responsible for the significant variation in photocatalytic activity. The right axis of Fig. 6 is assigned to the corresponding photonic efficiencies. A maximum hydrogen production rate of 28.4 ␮mol H2 /h and a photonic efficiency of about 0.55% are obtained for (Bi1.55 Ti0.33 )Ti2 O7 (Ti/Bi = 1.50) after heat treatment at 773 K for 8 h. Furthermore, the results for samples annealed at a higher temperature of 873 K for 4 h are also included in Fig. 6. For comparison, the very active TiO2 catalyst Hombicat UV100 (Sachtleben Chemie GmbH) provides a photonic efficiency of ca. 4% in the same photocatalysis setup when loaded with 0.75 wt% of platinum. As can be deducted from the insets of Fig. 7 and Fig. S3, the hydrogen production rates of (Bi2−x Ti0.75x )Ti2 O7 samples are not

Fig. 6. Maximum photocatalytic activities and photonic efficiencies regarding hydrogen production over (Bi2−x Ti0.75x )Ti2 O7 annealed at 773 K for 4 h as a function of the Ti/Bi ratio. The arrows (6 h and 8 h) refer to the photocatalytic activities obtained, if the samples are annealed for 6 h or 8 h instead of 4 h. The inset shows the maximum hydrogen production rates over (Bi2−x Ti0.75x )Ti2 O7 with Ti/Bi ratios of 1.40–1.50 annealed at 773 K as a function of the annealing time.

perfectly stable with time, although a high co-catalyst loading of 3.0 wt% was applied. However, plotting the produced hydrogen cumulatively (cf. to Fig. 7 and S3), which is the common way in literature, covers the stability issue to some extent. The drawback of decreasing activities over time makes the platinum loaded bismuth titanate not yet applicable in hydrogen production. In organic degradation reactions bismuth titanate does not suffer from this drawback, as oxygen acts as an excellent electron scavenger. Thus, the presented stoichiometry optimization is supposed to give a substantial progress for these reactions using (Bi2−x Ti0.75x )Ti2 O7 being excitable by visible light irradiation. To check whether stability issues are restricted to pyrochlore (Bi2−x Ti0.75x )Ti2 O7 samples, bismuth titanate structures with other Ti/Bi ratios of 0.75 (Bi4 Ti3 O12 ) and 2.00 (Bi2 Ti4 O11 ) were synthesized and tested in photocatalytic hydrogen production as well. The results are presented in Fig. 8. A clear correlation is found between the stability during photocatalysis and the bismuth content in the structure. Very low activities were provided by samples of perovskite Bi4 Ti3 O12 (Ti/Bi of 0.75), which were also highly instable losing almost their entire activity within 3 h of illumination. Bi2 Ti4 O11 featuring the lowest bismuth content is less active than the non-stoichiometric pyrochlore (Bi1.55 Ti0.33 )Ti2 O7 (Ti/Bi of 1.50)

Fig. 7. Photocatalytic activity regarding (Bi2−x Ti0.75x )Ti2 O7 calcined at 773 K for 4 h.

hydrogen

production

over

108

O. Merka et al. / Catalysis Today 225 (2014) 102–110

Fig. 8. Photocatalytic activity regarding hydrogen production over Bi4 Ti3 O12 (Ti/Bi = 0.75) annealed at 773 K and 873 K for 4 h, (Bi1.55 Ti0.33 )Ti2 O7 (Ti/Bi = 1.50) calcined at 773 K for 8 h, and Bi2 Ti4 O11 (Ti/Bi = 2.00) annealed at 873 K, 973 K and 1073 K for 4 h.

in the initial stage, but the enhanced stability during photocatalysis leads to a higher activity at longer irradiation times. 3.4. Structure–photocatalysis relation In order to correlate the photocatalytic activities with the structural properties of the pyrochlore (Bi2−x Ti0.75x )Ti2 O7 (cf. to Section 3.2, structure suggestion 4), the TiO6 octahedral geometry is taken into consideration. The geometry of the TiO6 octahedra is determined by the x positional parameter of O oxygen and the lattice parameter a. Table 2 shows that the x positional parameter decreases with increasing Ti/Bi ratio. Knop et al. [33] claimed an x positional parameter of 0.4201 and a lattice parameter of 10.095 A˚ for Y2 Ti2 O7 being the most active material in the group of pyrochlores, which yielded a hydrogen production rate of 71 ␮mol H2 /h [29]. Increasing the Ti/Bi ratio in (Bi2−x Ti0.75x )Ti2 O7 leads to a convergence of the x positional parameter toward the value given for Y2 Ti2 O7 . As those samples featuring a high Ti/Bi ratio are also

highly active in photocatalysis, a correlation between activity and changes in the TiO6 octahedral geometry is suggested. In the pyrochlore structure, a total of 15 O Ti O angles are provided in the structure. Three of them are predetermined to be 180◦ . The other six pairs of O Ti O angles have the same value with each pair adding up to a sum of 180◦ . Fig. 9 presents the characteristics of the TiO6 octahedral geometry, i.e., the Ti O Ti and O Ti O angles as well as the Ti O distance, as a function of the Ti/Bi ratio. The stoichiometric sample possesses the highest octahedral distortion with O Ti O angles of 76.0◦ /104.0◦ . For the sample featuring a Ti/Bi ratio of 1.35 an almost perfect octahedral geometry with angles of 89.5◦ /90.5◦ is obtained. But it is questionable, if values of 90◦ /90◦ are ideal, as the O Ti O angles in the very active Y2 Ti2 O7 material amount to 83◦ /97◦ . On the other hand, similar angles of 85◦ /95◦ are present at a Ti/Bi ratio of 1.20, but this sample is almost inactive. Therefore, the O Ti O angles cannot be responsible for a high photocatalytic activity and attention should be paid to the Ti O Ti angle in the TiO6 octahedral network (Fig. 9b). In the cubic perovskite system, a study on luminescence properties revealed that the closer the M O M angle in the octahedra is to 180◦ , the more the excitation energy is delocalized, which may result in a higher photocatalytic activity [38]. In the pyrochlore structure, a Ti O Ti angle significantly lower than 180◦ is considered favorable, as the TiO6 octahedra are arranged as a sixmembered ring around the bismuth cation. Highly active Y2 Ti2 O7 provides a low Ti O Ti angle of 131.4◦ . The Ti O Ti angle in (Bi2−x Ti0.75x )Ti2 O7 decreases significantly by increasing the Ti/Bi ratio. Thus, there is strong evidence that the photocatalytic activity of pyrochlores is enhanced by reducing the M O M angle. Beside the octahedral angles also the Ti O distance influencing the orbital overlap is examined. The Ti O distance derives not only from the O oxygen x positional parameter, but also from the lattice parameter a. Due to an increasing degree of vacancies in the structure, the lattice parameter a was found to decrease as a function of the Ti/Bi ratio (cf. to Fig. 3), which should lead to a decrease of the Ti O distance. However, despite this lattice shrinkage, the Ti O distance is actually even increased as shown in Fig. 9c. Additional space provided by the progressive generation of vacancies in the structure might be used by oxygen anions moving slightly away from titanium cations and thus, might lead to an increase ˚ The Ti O distance in the Ti O distance from 1.861 A˚ to 1.965 A. ˚ of active Y2 Ti2 O7 is again very close to those distances (1.960 A) obtained for highly active (Bi2−x Ti0.75x )Ti2 O7 samples. In titanate pyrochlores empty Ti 3d orbitals are available for ␲-bonding with filled O 2p orbitals of 48f oxygen [37]. A larger distance between titanium and oxygen usually leads to a decrease in the orbital overlap reducing the charge delocalization. The reason why this

Fig. 9. Evolution of the O Ti O angle, (a) the Ti O Ti angle (b) and the Ti O distance (c) in the TiO6 octahedra as a function of the Ti/Bi ratio in (Bi2−x Ti0.75x )Ti2 O7 . For annealing times refer also to Fig. 6. The values of Y2 Ti2 O7 are given for comparison.

O. Merka et al. / Catalysis Today 225 (2014) 102–110

Fig. 10. Activity and stability of photocatalytic hydrogen production over (Bi1.55 Ti0.33 )Ti2 O7 (Ti/Bi = 1.50) annealed at 773 K for 8 h loaded with bare Pt or core–shell Pt/Cr2 O3 as a function of the co-catalyst loading. All measurements were performed in the presence of 10 vol% of methanol acting as sacrificial reagent.

is advantageous for photocatalysis is not clear yet. Radosavljevic et al. [34] reported about the synthesis and structural properties of a non-stoichiometric composition Bi1.74 Ti2 O6.84 by using excess titanium. The sample was contaminated with TiO2 and Bi2 Ti4 O11 byproducts. The O Ti O angle was found to be 92.5◦ /87.5◦ and the ˚ The Ti O Ti angle was not evaluated. Ti O distance to be 1.96 A. These values correspond quite well with our results for a Ti/Bi ratio of 1.20–1.40. 3.5. Pt/Cr2 O3 core–shell co-catalyst Usually, the co-catalyst/chromium core–shell system is effectively used in the process of photocatalytic overall water splitting to inhibit the back reaction between just formed hydrogen and oxygen to water, which is thermodynamically favored on the surface of precious metals like platinum or rhodium [39]. Unfortunately, neither the pyrochlore nor one of the other bismuth titanates was found to be active in overall water splitting. In order to evaluate the ability to promote hydrogen evolution, the core–shell co-catalyst system was tested in hydrogen production as well. Interestingly, the Pt/Cr2 O3 core–shell system applied on the pyrochlore (Bi1.55 Ti0.33 )Ti2 O7 (Ti/Bi = 1.50) provided not only adequate hydrogen production, but a stable activity over time as well. The photodeposition process of the Pt/Cr2 O3 core–shell structure was performed in presence of methanol, but was difficult to control, as the hydrogen production rate decreases markedly with prolonging irradiation time, if Pt acts as co-catalyst. Thus, the Cr(VI) solution was added already after 30 min of irradiation to reduce the period of platinum core deposition as much as possible even if the Pt deposition might not be completed. An impregnation method was not considered for the chromium shell deposition, as the photochemical deposition is thought to be more specific toward platinum active sites compared to a chemical method. The two single data points in Fig. 10 are obtained after 30 min of irradiation prior to addition of the chromium source and are referred to as a test of reproducibility, which is quite good as being closely to the data points achieved in the normal test runs applying Pt, which

109

are displayed in Fig. 10 as well. After addition of the Cr(VI) solution, the hydrogen production rate decreased to zero and hydrogen could not be detected within the 300 min of Cr2 O3 photodeposition any more. Since residual Cr6+ in solution inhibits any hydrogen formation, filtration was performed before starting the test reaction. The (Bi1.55 Ti0.33 )Ti2 O7 773 K 8 h sample was loaded with 2.0 wt% Pt/1.55 wt% Cr (in Cr2 O3 ) and tested for photocatalytic hydrogen production with methanol as sacrificial reagent. The results are shown in Fig. 10. The actual Cr loading was determined by UV–vis spectroscopy. Although the initial hydrogen production rate was lower than obtained with the platinum sample, a stable hydrogen production rate of 8 ␮mol H2 /h persisted over the whole irradiation period of 5 h. Compared to the platinum sample, the activity even becomes superior after 2 h of irradiation. Additionally, a Pt/Cr2 O3 loading of 3.0 wt% Pt/2.31 wt% Cr (in Cr2 O3 ) was tested, as 3.0 wt% was found to be the optimum Pt loading. A significant improvement in activity is obtained for the Pt sample, whereas the Pt/Cr2 O3 loaded sample does not benefit from a higher co-catalyst loading. However, the activities of both samples converge with time. Precious metals like platinum are good hydrogen evolution sites, because they act as electron sinks and accumulate photogenerated electrons at their surface sites. The loss in activity occurring for samples loaded with blank Pt is supposed to be attributed to the photo-reduction of Bi3+ at the interface between the pyrochlore oxide and platinum forming bismuth suboxides or elemental bismuth. However, crystalline byproducts could not be detected by X-ray diffraction even after 24 h of illumination. The Cr2 O3 shell helps to prevent this photodecomposition, the mechanism, however, needs further investigations.

4. Conclusions In addition to earlier studies with the pyrochlores Y2 Ti2 O7 [29] and (Y1.5 Bi0.5 )1−x Ti2 O7−3x and (YBi)1−x Ti2 O7−3x [40] it is now also shown for the pyrochlore (Bi2−x Ti0.75x )Ti2 O7 that a stringent structural tune-up improves the photocatalytic activity, whereas the vacancy structure in the pyrochlore is carefully analyzed by Rietveld refinement. The very large effective ionic radius of bismuth makes it hard to achieve stoichiometric Bi2 Ti2 O7 , but enables the formation of a very high degree of Bi-vacancies in the lattice. A largely extended stability range of titanium rich stoichiometries is proven for non-stoichiometric (Bi2−x Ti0.75x )Ti2 O7 ranging at least up to a Ti/Bi ratio of 1.50, which corresponds to a chemical composition of (Bi1.55 Ti0.33 )Ti2 O7 . Vacancies solely formed on the A-site are supposed to be required for high photocatalytic activity by reducing stress in the TiO6 octahedral geometry, which leads to an improved charge carrier mobility and thus, to a better charge separation. Thus, structure optimized (Bi2−x Ti0.75x )Ti2 O7 provides a photocatalytic activity in hydrogen production of 28.4 ␮mol H2 /h, although stoichiometric Bi2 Ti2 O7 is completely inactive. Other bismuth titanate phases of Bi4 Ti3 O12 and Bi2 Ti4 O11 were less active compared to the stoichiometry optimized pyrochlore (Bi1.55 Ti0.33 )Ti2 O7 . Due to the augmented vacancy generation at higher Ti/Bi ratios, the oxygen ions undergo a slight change in their lattice position leading to an enlarged Ti O distance and a reduced Ti O Ti angle. These changes in the TiO6 octahedral geometry are considered being responsible for the significant improve in photocatalytic activity. Probably due to the formation of bismuth suboxides, samples of bismuth titanate were found to be instable within the process of photocatalytic hydrogen production leading to decreasing activities with time. However, this major drawback can be overcome by coating the platinum co-catalyst particles with a Cr2 O3 shell.

110

O. Merka et al. / Catalysis Today 225 (2014) 102–110

Acknowledgements The authors thank Prof. Dr. J. Caro (Institut für Physikalische Chemie und Elektrochemie, Leibniz Universität Hannover) for providing laboratory space and the X-ray diffractometer. Financial support by the Deutsche Forschungsgemeinschaft (DFG) (WA 1116/18-1) is gratefully acknowledged. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j.cattod. 2013.09.009. References [1] S.Y. Wu, W.J. Takei, M.H. Francombe, Ferroelectrics 10 (1976) 209–213. [2] J.R. Esquivel-Elizondo, B.B. Hinojosa, J.C. Nino, Chem. Mater. 23 (2011) 4965–4974. [3] V. Kahlenberg, H. Bohm, Acta Crystallogr. B: Struct. 51 (1995) 11–18. [4] A.D. Morrison, Ferroelectrics 2 (1971) 59–62. [5] S.P. Yordanov, C.P. Carapanov, I.S. Ivanov, P.T. Cholakov, Ferroelectrics 209 (1998) 541–552. [6] T. Zaremba, J. Therm. Anal. Calorim. 93 (2008) 829–832. [7] M.R. Hoffmann, S.T. Martin, W.Y. Choi, D.W. Bahnemann, Chem. Rev. 95 (1995) 69–96. [8] W.F. Yao, H. Wang, X.H. Xu, X.F. Cheng, J. Huang, S.X. Shang, X.N. Yang, M. Wang, Appl. Catal. A: Gen. 243 (2003) 185–190. [9] X.H. Xu, W.F. Yao, Y. Zhang, A.Q. Zhou, Y. Hou, M. Wang, Acta Chim. Sinica 63 (2005) 5–10. [10] W.F. Yao, X.H. Xu, H. Wang, J.T. Zhou, X.N. Yang, Y. Zhang, S.X. Shang, B.B. Huang, Appl. Catal. B: Environ. 52 (2004) 109–116. [11] J. Wang, L. Jing, L. Xue, Y. Qu, H. Fu, J. Hazard. Mater. 160 (2008) 208–212. [12] A. Kudo, S. Hijii, Chem. Lett. 28 (1999) 1103–1104. [13] S. Murugesan, M.N. Huda, Y.F. Yan, M.M. Al-Jassim, V. Subramanian, J. Phys. Chem. C 114 (2010) 10598–10605.

[14] D.W. Bahnemann, A. Henglein, L. Spanhel, Faraday Discuss. 78 (1984) 151–163. [15] A. Sclafani, M.N. Mozzanega, P. Pichat, J. Photochem. Photobiol. A 59 (1991) 181–189. [16] T.A. Kandiel, R. Dillert, D.W. Bahnemann, Photochem. Photobiol. Sci. 8 (2009) 683–690. [17] Z.P. Shao, W.S. Yang, Y. Cong, H. Dong, J.H. Tong, G.X. Xiong, J. Membr. Sci. 172 (2000) 177–188. [18] J.I. Langford, A.J.C. Wilson, J. Appl. Crystallogr. 11 (1978) 102–113. [19] A. Sobczynski, A.J. Bard, A. Campion, M.A. Fox, T. Mallouk, S.E. Webber, J.M. White, J. Phys. Chem. 91 (1987) 3316–3320. [20] C.G. Hatchard, C.A. Parker, Proc. Roy. Soc. Lond. Ser.: A 235 (1956) 518–536. [21] C.Y. Wang, R. Pagel, D.W. Bahnemann, J.K. Dohrmann, J. Phys. Chem. B 108 (2004) 14082–14092. [22] Y. Maeda, A. Fujishima, K. Honda, J. Electrochem. Soc. 128 (1981) 1731–1734. [23] T. Lana Villarreal, R. Gomez, M. Neumann-Spallart, N. Alonso-Vante, P. Salvador, J. Phys. Chem. B 108 (2004) 15172–15181. [24] W. Wei, Y. Dai, B. Huang, J. Phys. Chem. C 113 (2009) 5658–5663. [25] T. Goto, Y. Noguchi, M. Soga, M. Miyayama, Mater. Res. Bull. 40 (2005) 1044–1051. [26] T. Kidchob, L. Malfatti, D. Marongiu, S. Enzo, P. Innocenzi, J. Am. Ceram. Soc. 93 (2010) 2897–2902. [27] H. Nyman, S. Andersson, B.G. Hyde, M. O‘Keeffe, J. Solid State Chem. 26 (1978) 123–131. [28] J. Zhou, Z. Zou, A.K. Ray, X.S. Zhao, Ind. Eng. Chem. 46 (2007) 745–749. [29] O. Merka, D.W. Bahnemann, M. Wark, ChemCatChem 4 (2012) 1819–1827. [30] J. Tauc, R. Grigorovic, A. Vancu, Phys. Stat. Sol. 15 (1966) 627–637. [31] L. Bi, J.Y. Feng, J. Lumin. 121 (2006) 95–101. [32] M.K. Jayaraj, A. Antony, M. Ramachandran, Bull. Mater. Sci. 25 (2002) 227–230. [33] O. Knop, F. Brisse, L. Castelliz, Can. J. Chem. 47 (1969) 971–990. [34] I. Radosavljevic, J.S.O. Evans, A.W. Sleight, J. Solid State Chem. 136 (1998) 63–66. [35] O. Merka, V. Yarovyi, D.W. Bahnemann, M. Wark, J. Phys. Chem. C 115 (2011) 8014–8023. [36] C.R. Stanek, L. Minervini, R.W. Grimes, J. Am. Ceram. Soc. 85 (2002) 2792–2798. [37] A.M. Srivastava, J. Lumin. 129 (2009) 1000–1002. [38] G. Blasse, J. Solid State Chem. 72 (1988) 72–79. [39] M. Yoshida, K. Takanabe, K. Maeda, A. Ishikawa, J. Kubota, Y. Sakata, Y. Ikezawa, K. Domen, J. Phys. Chem. C 113 (2009) 10151–10157. [40] O. Merka, O. Raisch, F. Steinbach, D.W. Bahnemann, M. Wark, J. Am. Ceram. Soc. 96 (2013) 634–642.