WO3-modified TiO2 nanotubes for photocatalytic elimination of methylethylketone under UVA and solar light irradiation

Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

Contents lists available at SciVerse ScienceDirect

Journal of Photochemistry and Photobiology A: Chemistry journal homepage: www.elsevier.com/locate/jphotochem

WO3 -modified TiO2 nanotubes for photocatalytic elimination of methylethylketone under UVA and solar light irradiation Yas Yamin, Nicolas Keller, Valérie Keller ∗ Laboratoire des Matériaux, Surfaces et Procédés pour la Catalyse (LMSPC), UMR 7515, CNRS-University of Strasbourg, 25 rue Becquerel, 67087 Strasbourg Cedex, France

a r t i c l e

i n f o

Article history: Received 3 February 2012 Received in revised form 22 June 2012 Accepted 27 June 2012 Available online 8 July 2012 Keywords: Photocatalysis TiO2 nanotubes WO3 modification UV-A activation Solar light activation Methylethylketone elimination Semiconductor coupling

a b s t r a c t We report on the synthesis, the characterizations and the photocatalytic activity of WO3 /TiO2 nanotubes for methylethylketone (MEK) elimination under both UVA and simulated-solar light irradiations. TiO2 nanotubes (TiNTs) were synthetized using the hydrothermal method in a concentrated 10 M NaOH solution heated at 130 ◦ C during 48 h, post-impregnated with tungstate salt and finally calcined. The photocatalytic efficiency for both UVA and solar light irradiation strongly depends on the WO3 content, ranging from 4 to 51 wt.%. The different photocatalytic activities were analysed, in terms of initial and stabilized activities and deactivation phenomena. Complementary post-tests characterizations (TGA and XPS) aiming at the establishment of correlations between WO3 content, surface composition, surface species and the photocatalytic activity and deactivation are discussed. © 2012 Elsevier B.V. All rights reserved.

1. Introduction TiO2 -based nanomaterials are well-studied and commonly used materials for liquid and gas-phase photocatalytic applications [1,2]. Since the discovery of carbon nanotubes [3], intensive researches of one-dimensional nanostructures such as nanotubes, nanorods, nanowires, nanobelts, etc. of inorganic materials have attracted great attention because they offer interesting potential for use in various applications [4]. Amongst them, there have been significant efforts to elaborate controlled TiO2 -based nanotubes. Currently developed methods for elaborating TiO2 nanotubes include electrochemical anodic oxidation [5–8], assisted-template methods [9,10], and sol–gel processes [11,12]. However, these nanotubes generally have large diameters and their walls are composed of nanoparticles. Kasuga et al. [13] revealed that via a simple hydrothermal treatment of crystalline TiO2 particles with NaOH aqueous solution, high quality TiO2 nanotubes (TiNTs) with uniform diameters around 10 nm could be obtained. These nanotubes are not constituted of nanoparticles but are grown through the bending and rolling of titanate nanosheets [14–16]. However, the exact formation mechanism and the phase nature of TiO2 nanotubes prepared by hydrothermal method is still under debate. The nanoscale 1D-layered titanate structures obtained by this simple hydrothermal synthesis have

∗ Corresponding author. Tel.: +33 3 68 85 27 36; fax: +33 3 68 85 27 61. E-mail address: [email protected] (V. Keller). 1010-6030/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jphotochem.2012.06.021

attracted much attention since they offer a larger surface area available in comparison to nanoparticles and to TiO2 nanotubes obtained from the template and electrochemical anodization methods. Moreover, they provide channels for enhanced electron transfer thus limiting photogenerated charge recombinations [13,17]. The combination of these properties should have determinant roles in photocatalytic reactions, in addition to the fact that the hydrothermal technique is a simple and cost-effective method for large scale production of TiO2 nanotubes. Nowadays, one of the challenges in photocatalysis is to build up photocatalytic processes working with solar light as a clean and renewable energy source. However, the large band gap value (around 3.1–3.2 eV) of TiO2 nanotubes, resulting in low absorption of solar light, is a serious drawback for these applications. Thus, to keep advantages of the unique properties of this 1D-layered TiO2 morphology for visible light activation, several strategies could be explored. Up to now, amongst them, one can note fluor, nitrogen and carbon doping [18–20] but also coupling with a narrow gap semiconductor like CdS, ZnS and WO3 , yielding heterojunction formation [21–27]. In the present study, WO3 -modified TiNT samples with different WO3 contents were prepared by post-impregnation of TiNT obtained via hydrothermal synthesis, followed by a final calcination step. Their photocatalytic properties were evaluated towards UV-A and simulated-solar light degradation of methylethylketone (MEK). Electronic microscopies, surface area and porosity measurements, UV–vis, TGA, zetametry and XPS characterizations were performed

44

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

on the as-prepared samples, whereas complementary TGA and XPS analyses were carried out after photocatalytic tests. The contribution of WO3 in terms of visible light activation (in addition to UV-A activation) and of heterojunction formation was discussed and on-stream deactivation behaviour was also explained. 2. Experimental 2.1. Synthesis of WO3 /TiNT catalysts Typically, high aspect ratio TiNTs were synthetized by hydrothermal treatment of 3 g of TiO2 powder in 180 ml of concentrated 10 M NaOH solution at 130 ◦ C for 48 h. After hydrothermal treatment, the white powder obtained was vacuum-filtered, washed with 1 M HCl solution and distilled water until neutral pH, followed by overnight drying at 100 ◦ C. Post-synthesis modifications with WO3 were performed by classical impregnation with ammonium paratungstate salt. Starting from 1 g of the prepared TiNT catalyst, 0.047 mg, 0.098 mg, 0.214 mg and 0.530 mg of (NH4 )10 W12 O40 ·5H2 O salt was added and completed with 10 ml of distilled water, magnetically stirred until evaporation of the solvent and dried overnight at 100 ◦ C. A final calcination step for 3 h at 350 ◦ C was performed to yield theoretically 4, 9, 18 and 51 wt.% of WO3 /TiNT, respectively. 2.2. Characterizations of WO3 /TiNT catalysts The surface area measurements were performed on ASAP2010 Micromeritics porosimeter using N2 as an adsorbant at liquid N2 temperature. Before each measurement, the sample was outgassed overnight at 250 ◦ C. Surface areas were calculated from the N2 adsorption isotherms using the BET method (SBET ). The micropore surface areas were calculated using the t-plot method and the pore size distribution was obtained using the B.J.H. method during the desorption branch [28]. A detailed study of the method, concerning the correctness of the different parameters used, has been published elsewhere [29]. Structural characterization was done by powder X-ray diffraction (XRD) measurements, carried out with a D8 Advance Bruker diffractometer, using a Cu K␣ radiation source in a /2 mode, with a duration scan of 0.5 s and a small step scan (0.04◦ 2). Thermogravimetric analysis (TGA) was performed using a Q500 TA Instrument thermoanalyzer. Each sample was placed in a platinum crucible and heated from room temperature to 900 ◦ C at a rate of 10 ◦ C/min, using a 20% (v/v) O2 /N2 mixture at a flow rate of 25 ml/min. X-ray photoelectron spectroscopy (XPS) surface characterizations were performed using a Multilab 2000 (Thermo) apparatus equipped with an Al K␣ (1487) source (pass energy of 20 eV). All of the spectra were decomposed assuming several contributions, each having a Doniach–Sunjic shape [30] and a Shirley background subtraction [31]. The tungsten-to-titanium (W/Ti) surface atomic ratios were calculated using the sensitivity factors, as determined by Scofield [32]. The subtraction of the energy shift due to electrostatic charging was determined using the contamination carbon C1s band at 284.6 eV as a reference. Scanning electron microscopy (SEM) was performed using a JEOL XL 30 FEG microscope working at 10–20 kV. Transmission electron microscopy (TEM) was performed on a Topcon 002B microscope working with 200 kV and a point-to-point resolution of 0.17 nm. The sample was sonically dispersed in an ethanol solution before a drop of this solution was deposited onto a copper grid covered by a holey carbon membrane for observation. Isoelectrical point (IEP) measurements were carried out on a Malvern ZetaSizer with automatic titration.

UV–vis absorption spectra of the materials were recorded on a Cary 100 Scan UV/vis spectrophotometer from Varian equipped with a DRA-CA-301 Labsphere diffuse reflectance cell. 2.3. Experimental photocatalytic device and procedure The photocatalytic reaction was carried out in a 300-mm long cylindrical concentric tubular Pyrex reactor composed of two coaxial tubes 4 mm apart, between which the reactant mixture passes through. Detailed descriptions of the photocatalytic reactor and device can be found elsewhere [33,34]. Photocatalytic material (400 mg) was evenly coated on the internal side of the 35-mm diameter external tube by evaporating a catalyst containing aqueous slurry to dryness. The catalyst-coated reactor was then dried at 110 ◦ C for 1 h in air. MEK (Sigma–Aldrich, 99%) and water were fed at ambient temperature and atmospheric pressure by bubbling air through two saturators, then mixed with additional air (using Brooks 5850 massflow controllers) to obtain the required MEK–water–air ratios with a constant total air flow of 350 cm3 /min. The MEK content was set at 400 ppm in flowing air. The relative humidity was set at 50%, with 100% of relative humidity defined as the saturated vapour pressure of water at 25 ◦ C, corresponding to about 24 Torr, that is, about 3% relative to the total atmospheric pressure. No pretreatment of the photocatalysts has been carried out prior to photocatalytic experiments. Before the photocatalytic reaction, the catalyst was first exposed to the polluted air stream with no illumination until dark-adsorption equilibrium was reached. Afterward, illumination was switched on. Illumination was provided by a commercially available 8-W UV-A (45.3 W/m2 ) black light tube (Philips) with a spectral peak centred around 365 nm (Fig. 1a) or by a 8-W day light tube (Philips) (Fig. 1b) located inside the inner tube of the reactor. One may mention that the daylight, considered here as being simulated-solar light, was composed in terms of irradiance power of 42.5 W/m2 of visible light and of 0.7 W/m2 of UV-A (i.e. of 1.6%). The reaction products were analysed on-line every 2 min by a thermal conductivity detector on a micro-gas chromatograph (Agilant microCG R3000), allowing detection and quantification of MEK, water, CO2 , and organic byproducts on Stabilwax, PLOT Q, OV1 and MS-5A columns. The efficiency of the depollution process was expressed in terms of MEK conversion, of CO2 selectivity and yield, as well as of carbon deficit, and calculated as follows: Conversion (%) =

SCO2 (%) =



[MEK]in − [MEK]out × 100 [MEK]in



1 [CO2 ]out × × 100 4 [MEK]in − [MEK]out

CO2 yield (%) is defined as : (SCO2 × Conversion)/100. The carbon deficit (expressed in ppm) is defined as the amount of carbon remaining adsorbed or deposited on the catalyst surface and is expressed as C deficit (ppm) = [MEK]in − [MEK]out −

 1 ˛j

[Products]j,

gas phase

where [Products]j is the concentration of each j carbonaceous compound in the gas phase and ˛j is the stoichiometry factor of the j compound. Carbon deficit yield (%) is defined as : (1 − SCO2 × Conversion) × 100.

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

50000

3. Results

a

3.1. Characterizations of the fresh photocatalysts

Intensity, A.U.

40000

3.1.1. Elementary analysis From elementary analysis performed on WO3 -modified TiNT samples, it could be determined that the true WO3 contents are 3.8, 8.7, 17.8 and 51 wt.% for the samples further designated as 4, 9, 18 and 51 wt.%, respectively, meaning that the obtained contents of WO3 are not far from theoretical values.

30000

20000

10000

0 200

300

400

500

600

700

800

Wavelenght, nm 80000

b

70000 60000

Intensity, A.U.

45

50000 40000 30000 20000 10000 0 200

300

400

500

600

700

800

Wavelenghts, nm Fig. 1. Emission spectra of commercially available (a) 8-W UV-A black light lamp (Philips) and (b) 8-W day light lamp (Philips).

3.1.2. Structural characterizations XRD (Fig. 2) characterization showed that the WO3 -modified and -free TiNT samples (calcined at 350 ◦ C) exhibited the same TiO2 (B) structure, with the main peaks at 24.4 and 48.5◦ corresponding to B(1 1 0) and B(0 2 0) crystal planes, respectively [35,36]. Even if these two peaks are very close to those of anatase TiO2 having the main diffraction peaks at 24.8◦ and 48.3◦ and corresponding to (1 1 0) and (0 2 0) crystal planes [37], the discrimination could be confirmed looking at the broad and small peak at 29.6◦ , exclusively assigned to B(1 1 1) planes and absent in the anatase structure. Although this diffraction region is broadened, it matched the pattern for TiO2 (B). However, at this stage, the presence of TiO2 anatase cannot be excluded. The XRD patterns of WO3 /TiNT catalysts did not evidence any diffraction peaks corresponding to WO3 crystallites, except for 51 wt.% WO3 /TiNT material, which displayed the characteristic profiles with three peaks at 23.6◦ , 24.4◦ and 48.3◦ , indexed to (0 1 1), (2 0 0) and (2 0 1) planes of hexagonal WO3 particles [38,39], respectively. The average particle size (D) of WO3 crystallites, estimated from the full width at half-maximum (FWHM), using the Debye Sherrer equation (Eq. (1)) was estimated to be about 26 nm: D=

K ˇ cos 

(1)

in which  is the diffraction angle of the main peak, ˇ is the FWHM and K is a constant equal to 0.9.

Fig. 2. XRD pattern of TiNT calcined at 350 ◦ C and WO3 -modified TiNT materials.

46

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

not transparent to the electron beam. However, for the highest WO3 content some particles could be observed (Fig. 4d).

Fig. 3. SEM images of the calcined TiNT samples.

3.1.3. SEM characterizations From Fig. 3a showing SEM images of calcined TiNT nanotubes, it is worthy to note that they are highly agglomerated into fibrous clusters of some microns. At higher magnification (Fig. 3b), one can observe the presence of one-dimensional structures with a length of few hundred of nm. According to literature, this agglomeration phenomenon is more marked in the case of short nanotubes because of their highest surface energy. It must be mentioned that WO3 -modified TiNT materials displayed the same morphology and features than unmodified ones. 3.1.4. TEM observations Fig. 4a and b shows TEM images of the TiO2 nanotubes after drying and calcination at 350 ◦ C, respectively. Dried nanotubes were open-ended (Fig. 4a) with lengths of few hundreds of nm and diameters around 10 nm. Calcination at 350 ◦ C led to the crystallization of the multilayered walls of the nanotubes without modifying the tubular morphology and with wall thickness of about 4 nm (inset Fig. 4b) and internal diameter of ca. 5 nm. This was in agreement with the results reported in the literature [40]. This also confirmed the below mentioned porosity distribution of the calcined nanotubes. The very difficult deagglomeration of closely entangled powdered TiO2 nanotubes (Fig. 3a), even after sonication, required to record TEM images only near the membrane border on the thinner parts of the sample. In these analysis conditions, no WO3 particles were observed below 51 wt.% WO3 (Fig. 4c) although they have been detected with EDX analysis, meaning that they may be mainly located in the more agglomerated regions of the samples,

3.1.5. Surface area (BET) and porosimetry determinations Fig. 5a exhibited the nitrogen adsorption–desorption isotherms of dried, calcined at 350 ◦ C and WO3 -modified-TiNT samples, corresponding to a mainly mesoporous solid of type IV with pore sizes between 2 and 50 nm, according to the well-accepted classification [41]. After drying, the material showed a high surface specific area of 349 m2 /g (Table 1) with no microporosity. This surface specific area remained unchanged after the calcination step and the impregnation with 4 wt.% of WO3 did not affect the surface specific area of the TiNT. Nevertheless, increasing further the amount of WO3 resulted in a decrease of this value, to reach 164 m2 /g on 51 wt.% WO3 /TiNT (Table 1). In the calculation of a theoretical monolayer of WO3 , it was assumed that one WO3 molecule occupies 23 A˚ 2 [42], so with the assumption that the whole surface of TiNT is accessible to WO3 particles, a monolayer of WO3 nanoparticles should correspond to ca. 38 wt.% of WO3 . The quantitative pore size distribution given in Fig. 5b revealed a bimodal and mesoporous pore size distribution. The first contribution, centred around 4 nm for all of the calcined samples corresponds to the internal diameter of the nanotubes, confirming the abovementioned TEM observations. Comparing with dried TiNT, it seems that calcination led to a slight increase of the internal diameter. Concerning the corresponding pore volume, it remained quite unchanged in the 4–18 wt.% range, meaning that WO3 particles were not present inside of the nanotubes. However, a drastic decrease was observed for the 51 wt.% WO3 /TiNT loading. In this last case, it could be supposed that the accessibility of the gas inside of the tubes has drastically decreased. Indeed, as the mean crystallite size of these particles were estimated previously from DRX analysis to be about 26 nm, much more important than the internal diameter of the tubes, their important number and accumulation outside of the nanotubes could lead to the blocking of their entrance. It is worthy to note, that the second contribution, centred around 15 nm, remained unchanged whatever thermal treatment or WO3 content. Consequently, this porosity could be attributed to that resulting from the aggregation and entangling of nanotubes, delimitated by their external surfaces. As the corresponding pore volume linearly decreased with increasing WO3 content, it could be confirmed that WO3 particles occupy mainly this second kind of porosity network and are thus deposited on the external surfaces of the nanotubes. As a result, Fig. 5a and b highlighted that the surface specific area mainly resulted from two kinds of mesoporosity, the first one corresponding to internal geometrical surfaces developed by the 1D-nanostructures and the second one issued from their entangled external surfaces, being mainly affected by WO3 modification. From these results pointing out the non-accessibility of internal surfaces of TiNT by WO3 particles, it could be concluded that the usable surface area for WO3 deposition, corresponding in a larger part to external surfaces of TiNT, is lower than 359 m2 /g. Consequently, a theoretical monolayer of WO3 may correspond to much less than 38 wt.%. From TEM observations, it could be deduced that the external diameter is about 1.5 times larger than the internal one, meaning that the ratio between external and internal surface was also 1.5. Consequently, if we consider that the total surface area mainly resulted from the contribution of both the external and internal surfaces of the nanotubes, it can be deduced that the value of the external surface, which is the major one accessible to WO3 particles, is 3/5 times the value of the total surface. In conclusion, the corrected theoretical monolayer of WO3 should be of ca. 22 wt.% of WO3 .

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

47

Fig. 4. TEM images of (a) the dried TiNT, (b) the calcined TiNT, (c) the 18 wt.% WO3 /TiNT and (d) the 51 wt.% WO3 /TiNT samples.

3.1.6. Zetametry measurements Fig. 6 showed the zeta potential as a function of pH for the determination of IEP from the dried, calcined and WO3 -modified-TiNT samples, compared with commercially available TiO2 P25 (Evonik) sample as a reference. All the TiNT-based materials exhibited lower IEP, compared to the quasi-neutral surface of TiO2 P25 (PIE = 6.8). More precisely, the IEP of WO3 -free TiNT was around 3–3.5, indicating the presence of an acidic surface resulting from the surface composition of nanotubes. Addition of WO3 decreased strongly this value down to 2.1 for 51 wt.% WO3 , approaching the IEP of the bare WO3 oxide evaluated between 0.2 and 0.5, thus being considered as an acid oxide [43]. Measurements of isoelectrical point could also provide complementary and useful informations, taking into account that the isoelectrical point could also be considered as the pH at which the stability of nanoparticles in suspension were the weakest, due to a more pronounced aggregation phenomena resulting mainly from Van der Waals interactions existing between nanoparticles in solution. Consequently and according to the theory of electrostatic stabilization in solution [44], putting forwards that a high absolute value of zeta potential yield increased electrostatic repulsions between particles and thus a better stability of the suspension [45], it is important for homogeneously deposition of TiNT-based materials to work with impregnation solutions having a pH value far away from the 2–3.5 range to ensure deposition of

non agglomerated materials. These considerations may be taken with attention, especially for some specific substrates [46]. 3.1.7. XPS analysis Deconvoluted XPS characterizations performed on dried, calcined and WO3 -modified-TINT samples are reported in Fig. 7a and b for the Ti2p and W4f region, respectively. The Ti 2p XPS spectra reported in Fig. 7a showed a signal that could be fitted with only two components, related to the Ti 2p spin–orbit components of Ti4+ surface species [47]. Comparison between the different samples revealed a consistent binding energy shift of 0.9 eV towards higher binding energies between dried TiNT and WO3 -modified TiNT material. A less important shift of 0.5 eV was also observed between dried TiNT and WO3 -free but calcined TiNT photocatalysts. Nevertheless, no shift in binding energies was observed depending on the content of WO3 . The Ti 2p3/2 –Ti 2p1/2 spin–orbit components appeared at 458.0–463.7, 458.4–463.9, and 458.9–464.6 eV for dried TiNT, calcined TiNT and WO3 /TiNT samples, respectively. These shifts to higher binding energies observed on calcined TiNT and in a larger extend on WO3 /TiNT materials resulted from a different chemical environment of Ti4+ and more precisely from an increase in effective positive charges around Ti4+ . No contributions corresponding to Ti lower oxidation states have been detected.

Table 1 BET surface areas of dried, calcined and WO3 -modified TiNT samples. TiNT dried

TiNT cal. 350 ◦ C

4% WO3 /TiNT

9% WO3 /TiNT

18% WO3 /TiNT

51% WO3 /TiNT

349 m2 /g

359 m2 /g

359 m2 /g

340 m2 /g

267 m2 /g

164 m2 /g

48

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

1000

a

Adsorbed quanty, cm3/g

4%WO3/TiNT ads 4%WO3/TiNT des

800

18% WO3/TiNT ads 18% WO3/TiNT des 51% WO3/TiNT ads

600

51% WO3/TiNT des TiNT cal 350 ads TiNT cal 350 des

400

200

0 0

0.2

0.4

0.6

0.8

1

and the second around 37 eV is assigned to a secondary Ti3p peak of Ti4+ . From the evolution of the W/Ti surface atomic ratio as a function of the real WO3 content determined from elementary analysis (Fig. 7c), two zones can be distinguished: (i) a first quasi-linear one and (ii) a second one for higher loadings, exhibiting a less pronounced W/Ti increase with WO3 content. The break between these two domains, identified around 20 wt.% of WO3 , clearly suggest that the deposition of the tungstate surface species is not the same between these two regions. This difference of behaviour could take place at near-saturation monolayer coverage, thus estimated previously as being close to 20 wt.% WO3 . One may mention that this experimental value is close to the estimated theoretical monolayer, assuming that mainly the external surface of the nanotubes were accessible to tungstate species, further transformed into WO3 particles during calcination step.

Relave pressure, p/p° 0.006

b

4% WO3/TINT 9% WO3/TiNT

Pore volume, cm3/g/Å

0.005

18% WO3/TiNT 51% WO3/TiNT

0.004

TiNT dried TiNT cal 350°C

0.003 0.002 0.001 0 1

10

100

1000

10000

Pore diameter, Å Fig. 5. (a) N2 adsorption–desorption isotherms and (b) pore size distribution of dried, calcined and WO3 -modified TiNT samples.

Fig. 7b displayed theW4f XPS spectra of the catalyst samples after calcination at 350 ◦ C as a function of tungsten content. It can be observed that, whatever the tungsten content, XPS signal of W4f after calcination can only be fitted with two components. The first one, at 35.5 and 37.7 eV for theW4f spin–orbit components of the W4f5/2 –W4f7/2 doublet, is attributed to W6+ surface species [48]

P25 Dried TiNT TiNT cal 350°C 4% WO3/TiNT

50 40

9% WO3/TiNT

Zeta potential (mV)

30

18% WO3/TiNT

20

WO3

10

pH

0 0

1

2

3

4

5

6

7

8

9

-10 -20 -30 -40 -50

Fig. 6. Zeta potential measurements of dried, calcined and WO3 -modified TiNT samples.

3.1.8. UV–vis absorption Fig. 8 showed the UV–vis absorption spectra recorded on the WO3 /TiNT samples and both WO3 -free dried and calcined TiNT materials, as well as for the TiO2 P25 reference. First, it is worthy to note that the reference TiO2 P25 sample displayed a first absorption part in the UV-A region, followed by a shoulder extended to the beginning of the visible light spectrum. This shoulder may be the result of the contribution of the rutile phase (ca. 20%), characterized by a lower band gap value, thus absorbing in the beginning of the visible light region. Dried and calcined TiNT mainly absorbed in the UV-A region, even if a very small part of the absorption tail may be located at the very beginning of the visible light spectra. Generally, a shift towards the visible light part of the spectrum is observed on WO3 -modified samples, with an enlargement when increasing the WO3 content. The absorption of the WO3 -modified TiNT occurring in the UV-A as well as in the visible light region of the spectrum, may suggest the potential use of these nanocomposites under solar light irradiation (UV-A and visible light wavelengths conditions). Nevertheless, one may mention that on the WO3 -based catalysts, the absorption spectra did not come back to the baseline in the visible part. This phenomenon increased when increasing the amount of WO3 and was certainly due to diffusion overall the whole visible light region, induced by WO3 nanoparticles. 3.2. Photocatalytic activity towards MEK elimination 3.2.1. Under UV-A irradiation Optimization of TiO2 content towards photocatalytic degradation of MEK has been previously carried out on the commercially available TiO2 P25 (Evonik) photocatalyst at 200 cm3 /min at a relative humidity ratio of 50% (considered elsewhere as belonging to the optimal range of humidity [49]) and at reactant concentration of 900 ppm. Fig. 9 represents, in these conditions, the evolution of total conversion as a function of TiO2 content coated on the inside wall of photoreactor. It can be observed that the maximum photoactivity for the oxidation of MEK was obtained for a surface concentration of TiO2 of 0.25 mg/cm2 , corresponding to ca. 440 mg of TiO2 coated inside of the reactor. Below this value, it can be supposed that the reactor is not completely covered by the photocatalyst. On the other hand, for higher values, no more increase in conversion is obtained. The amount of 440 mg of photocatalysts was thus choosen for the following photocatalytic tests, taking into account in first approximation only the TiO2 surface concentration of the photoreactor. Photocatalytic activities towards MEK elimination are reported in Fig. 10 for the TiNT-based catalysts, in terms of conversion (Fig. 10a), CO2 selectivity (SCO2 ) (Fig. 10b) and carbon deficit (Fig. 10c). Depending on WO3 content, the different samples

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

49

Fig. 7. XPS spectra of dried, calcined and WO3 -modified TiNT samples. (a) Ti 2p region, (b) W4f region and (c) W/Ti surface atomic ratios vs. real WO3 content.

revealed different photocatalytic behaviours. From the comparison between dried and calcined TiNT photocatalysts (Fig. 10a), it can be concluded that calcination is required to enhance the activity. Indeed, dried TiNT led to an initial conversion around 80%, followed by a rapid deactivation under stream to reach a near zero value within 100 min. Calcination of the sample at 350 ◦ C resulted in a drastic enhancement of the photocatalytic performances with a total (100%) MEK conversion during 45 min, followed by a smooth deactivation to reach a stabilized conversion of ca. 80%. After modifications with WO3 , all the samples exhibited the same general trend: an initial maximum conversion, followed by deactivation and stabilization of the photocatalytic activity. The value and duration of the maximum conversion as well as the value of the stabilized conversion (reported in Table 2) strongly depended on the WO3 content. On all the WO3 -modified materials, the initial maximum conversion is correlated with the stabilized conversion, i.e. the sample exhibiting the best initial conversion also displayed the best stabilized activity. Generally, increasing WO3 content resulted in a decrease in both the maximum and

the stabilized MEK conversion, at the exception of 18 wt.% WO3 content. More precisely, it could be observed than 18 wt.% WO3 yielded the best performances with a maximum in conversion of 100% and with only a slight deactivation to reach a stabilized conversion of 90%. Lower WO3 contents (4 and 9 wt.%) also resulted in high maximum conversions (between 90 and 94%), but suffered from a more pronounced deactivation to reach stabilized values between 60 and 67%. The most loaded sample displayed the worst activity, both in terms of maximum conversion and of stabilized activity. For the two lowest WO3 contents (4 and 9 wt.%), the CO2 selectivity evolution (Fig. 10b) showed the same general trend, i.e. a first zone of variation made of a rapid increase, followed by a diminution and by a second slower increase before stabilization. The first region showing a maximum in CO2 selectivity was also associated with the maximum in MEK conversion (Fig. 10a). After a few tens of minutes, SCO2 reached a stabilized value. For the other TiNT-based materials, there was a continuous increase in SCO2 to reach its stabilized value. It must also be mentioned that the only calcined sample

50

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

1,2

Absorbance (a.u.)

1,0

1,2

P25 TiNT dried TiNT cal 350°C 4% WO3/TiNT

1,0

9% WO3/TiNT 18% WO3/TiNT

0,8

51% WO3/TiNT

0,6

0,8

0,4 0,2

0,6

0,0 300

320

340

360

380

400

420

440

0,4

0,2

0,0 300

400

500

600

700

800

Wavelength (nm) Fig. 8. UV–vis absorption spectra of dried, calcined and WO3 -modified TiNT samples.

3.2.2. Under solar light illumination Photocatalytic performances under simulated-solar light irradiation showed the same general trend (Fig. 11), whatever the sample tested: an initial maximum conversion followed by a more or less pronounced deactivation to reach a stabilized activity. The value of the maximum conversion, its duration, as well as the level of stabilized activity clearly depended on the WO3 content

50

40

MEK conversion, %

without any WO3 modifications showed the best selectivity towards CO2 formation. Amongst the WO3 -modified samples, it must be underlined that the photocatalysts displaying the beststabilized activity (Fig. 10a) also showed the best stabilized CO2 selectivity, corresponding to the highest stabilized CO2 yields. It is worthy to note that no other gaseous components have been detected on our analytical system at the exception of low amounts of acetaldehyde and traces of unidentified compound. However, both intermediates did not account for more than 1.5% and 0.3% in terms of selectivity, respectively, whatever the TiNT-based samples tested. Anyway, as on WO3 -modified materials, CO2 selectivity did not exceed 57% in the best case, we can assume that a large part of the reaction products or intermediates remained on the surface of the catalyst. Looking at the carbon deficit evolution (Fig. 10c), corresponding to the amount of carbon remaining adsorbed or deposited on the photocatalyst surface, and since on WO3 -modified TiNT materials, the lowest carbon deficit is related to the less pronounced deactivation, it can be deduced that deactivation phenomena is correlated with any adsorption or deposition on the surface.

30

20

10

0

0

0.05

0.1

0.15

0.2

0.25

0.3

TiO2 surface coverage, mg/cm2 Fig. 9. Evolution of total conversion (%) vs. TiO2 surface concentration of the reactor.

(Table 3). The highest initial maximum conversion was obtained on the 4 wt.% WO3 /TiNT photocatalyst, which was unfortunately followed by a strong deactivation to reach a very low stabilized conversion. This important deactivation phenomena also occurred on the 9 wt.% WO3 loaded sample. More generally, except for the highest WO3 loading, the highest maximum activity was followed by the strongest deactivation, at the opposite to the results obtained under UV-A activation. On the other hand, the lowest the maximum activity, the less important the deactivation and the highest the stabilized activity. Even if modification of TiNT with 18 wt.% WO3 did not lead to the best initial activity, it resulted in the best

Table 2 Maximum MEK conversion (%), duration at maximum conversion (min), stabilized conversion (%), stabilized CO2 selectivity (%) and stabilized CO2 yield (%) during UV-Aactivated MEK elimination (400 ppm, 350 cm3 /min, 50% RH) as a function of WO3 content. WO3 content (wt.%) 0 4 9 18 51

Maximum MEK conversion (%) 100 94 90 100 64

Duration at max. conv. (min) 45 15 7 30 4

Stabilized conversion (%)

Stabilized CO2 selectivity (%)

Stabilized CO2 yield (%)

80 67 60 90 40

68 46 38 57 28

54 31 23 51 11

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

51

Table 3 Maximum MEK conversion (%), duration at maximum conversion (min), stabilized conversion (%), stabilized CO2 selectivity (%) and stabilized CO2 yield (%) during solar light-activated MEK elimination (400 ppm, 350 cm3 /min, 50% RH) as a function of WO3 content. WO3 content (wt.%)

Maximum MEK conversion (%)

Duration at max. MEK conversion (min)

Stabilized conversion (%)

Stabilized CO2 selectivity (%)

0 4 9 18 51

65 84 74 65 50

5 9 6 6 4

∼0 5 ∼0 20 15

2 0 0 0 0

a

100

18% WO3/TiNT TiNT calcined 350°C

80

Conversion, %

4% WO3 /TiNT 60 9 % WO3 /TiNT 40

51% WO3 /TiNT

20

performances in term of stabilized activity (20%). At this stage, it is also worthy to note that CO2 was only detected in the gas phase in very small amounts on the WO3 -free sample, while no CO2 was detected on the WO3 -modified TiNT photocatalysts. Taking into account that only traces (<0.5% in selectivity) of an unidentified gaseous product was observed, it could be considered that all the carbon present in the incoming MEK flow was remained on the catalyst surface. This hypothesis may explain the strong deactivation of the photocatalysts. 3.3. Characterizations of the photocatalysts after MEK photooxidation

Dried TiNT

0 0

20

40

60

80

100

120

140

Time on stream under UV-A illumination, min

b

80 TiNT calcined 350°C

18% WO3/TiNT

60

SCO2, %

4% WO3/TiNT 40

9% WO3/TiNT

20

51% WO3/TiNT

0 0

20

40

60

80

100

120

Time on stream under UV-A illumination, min

c

3.3.1. TGA analysis As the most pronounced deactivation phenomena occurred on the solar light activated photocatalysts, TGA analyses were first performed on this series of samples, after MEK photooxidation. Fig. 12 displayed TGA results of TiNT-based samples (a) before and (b) after solar light tests, respectively. From Fig. 12a, it can be observed that, whatever the catalyst, the same general trend is observed, with two distinguished regions, (i) the first one showing the major weight loss before 250 ◦ C and (ii) the second and minor one between 280 and 450 ◦ C. Comparing first the WO3 -free TiNT samples, it can be observed that the calcined one exhibited a most important total weight loss than the dried one, mainly due to a different behaviour observed in the first region, below 250 ◦ C. In the case of the dried TiNT material, the important weight loss displayed in the first region can be associated to water elimination from two regions (i) either adsorbed on the internal/external surface of the nanotubes or (ii) present in intra or inter-sheet position of the nanotube layered titanate structure before crystallization [50,51], following the equation: H2 Ti3 O7 → 3TiO2 + H2 O. The calcined and crystallized nanotubes being mainly present in the TiO2 (B) structure may no more be

500

51% WO3/TiNT

C deficit, ppm

400

9% WO3/TiNT

300

18% WO3/TiNT

200

TiNT calcined 350°C 100

0 0

20

40

60

80

100

120

140

Time on stream under UV-A illumination, min Fig. 10. UV-A-activated elimination of 400 ppm of MEK at 350 cm3 /min with 50% RH on dried, calcined and WO3 -modified TiNT samples. (a) Photocatalytic activities (%), (b) CO2 selectivities and (c) carbon deficit (ppm).

Fig. 11. Photocatalytic activities (%) towards simulated solar light-activated elimination of 400 ppm of MEK at 350 cm3 /min with 50% RH on dried, calcined and WO3 -modified TiNT samples.

52

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

Temperature, °C

a 0

200

400

0

800

9% WO3/TiNT

-4

Weight loss, %

600 Dried TiNT TiNT cal 350°C 4% WO3/TiNT 18% WO3/TiNT 51% WO3/TiNT

-8

-12

-16

Temperature, °C

b

0

200

400

600

800

0 TiNT cal 350°C 4% WO3/TiNT 9% WO3/TiNT

Weight loss, %

-4

18% WO3/TiNT 51% WO3/TiNT

-8

-12

-16

Fig. 12. TGA analyses (a) on dried, calcined and WO3 -modified TiNT samples before test and (b) on dried, calcined and WO3 -modified TiNT samples after solar lightactivated photocatalytic test.

affected by water elimination coming from the layered structure, but only adsorbed on the surface of the tubes, in addition to water elimination coming from dehydroxylation, as it has already been mentioned on other TiO2 crystallized materials [52]. This phenomenon of dehydroxylation, up to 400 ◦ C, could explain the more pronounced weight loss observed on the calcined WO3 -free material. Amongst all the WO3 -modified TiNT materials, addition of 51 wt.% WO3 yielded the less important weight loss to trend to the behaviour of bulk WO3 (not represented here), meaning that the weight loss coming from TiNT is larger than the one attributed to WO3 . Moreover, the weight loss increased when decreasing the content of WO3 , at the exception of 18 wt.% WO3 . After photocatalytic degradation of MEK and whatever the catalyst sample, an additional weight loss was detected between 200 ◦ C and 400–450 ◦ C, corresponding to accumulation of reaction products or reaction intermediates on the catalyst surface. Quantification of the difference of weight loss before and after solar light activated photocatalytic tests, representative of the amount of intermediates or products deposited or adsorbed on the surface, is reported in Table 4. It revealed that the weight loss associated to the photocatalytic test decreased with increasing WO3 content. More generally, when comparing with solar light activities presented previously, one can deduce that, at the exception of the WO3 -free catalyst, the samples showing the most important maximum activity and deactivation (4 and 9 wt.%) under solar light activation, also exhibited the most pronounced weight loss. On the other hand, the samples displaying the highest stabilized activity also showed the lowest weight loss (18 and 51 wt.%).

Fig. 13. Comparison of TGA analyses on 18 wt.% WO3 /TiNT samples before and after UV-A and solar light-activated photocatalytic elimination of MEK.

Further, to get more information about the differences in terms of residues deposited or adsorbed on the surface between UV-A and solar light activated photocatalyst, TGA analyses were also performed on the 18% WO3 /TiNT sample both after UV-A and solar light illuminated test. Fig. 13 revealed that the UV-A activated material exhibited a more marked weight lost. These observations are in accordance with the yield of carbon deficit (Table 5) that could be deduced from results given in Tables 2 and 3, related respectively to UV-A and solar light activity of WO3 -modified TiNT materials. Indeed, from Table 5, two phenomena could be observed, (i) a maximum carbon deficit yield, generally higher under solar light irradiation (except for 51% WO3 /TiNT), but for very short durations, followed by (ii) a longer duration stabilized carbon deficit, largely higher under UV-A irradiation. Thus, it is easy to understand that the total carbon deficit yield is more pronounced under UV-A activation even if, whatever the WO3 content, the corresponding activities were higher and the deactivations less pronounced. 3.3.2. XPS analysis XPS studies on used catalysts were only focused on the WO3 modified samples after solar light-activated tests, because they exhibited a more important deactivation than under UV-A activation. From Ti 2p XPS spectra (not represented here), it could be observed that the photocatalytic test did not induce any changes in terms of binding energy positions or of the appearance of other Ti surface oxidation states. In the same manner, W4f spectra (Fig. 14) also revealed that the photocatalytic tests did not affect the oxidation state of superficial tungsten species, as the binding energies assigned to W6+ 4f5/2− 4f7/2 at 35.5 and 37.7 eV did not change. These results meant that the photocatalytic tests did not modify the chemical environment of both Ti or W surface atoms by the formation of any chemical bond. Unfortunately, the C1s spectra could not be analysed due to the presence of contamination carbon inside of the XPS vacuum chamber, so that even the fresh TiNTbased materials showed superficial carbon species. As a result, it Table 4 Differences in weight loss before and after solar light-activated MEK elimination (400 ppm, 350 cm3 /min, 50% RH) as a function of WO3 content. WO3 content (wt.%)

Absolute weight loss (%)

0 4 9 18 51

0.90 3.95 2.85 2.3 1.3

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

53

Table 5 Comparison of the maximum carbon deficit yield (%) (and duration (min)) and the carbon deficit yield at stabilization (%) between UV-A and solar light-activated MEK elimination (400 ppm, 350 cm3 /min, 50% RH) as a function of WO3 content. WO3 content (wt.%)

4 9 18 51

Max. carbon deficit yield (%) (duration (min))

Carbon deficit yield at stabilization (%)

UV-A

Solar light

UV-A

Solar light

84 (9) 74 (6) 65 (6) 50 (4)

69 77 49 89

5 0 20 0

57 (15) 66 (7) 43 (30) 82 (4)

Table 6 W/Ti surface atomic ratios on fresh and used after simulated solar light-activated MEK elimination (400 ppm, 350 cm3 /min, 50% RH) WO3 -modified TiNT samples as a function of WO3 content. WO3 real content (%)

4 10 18 51

W/Ti surface atomic ratio Before photocatalytic test

After photocatalytic test

Difference (%)

0.054 0.078 0.151 0.185

0.052 0.073 0.124 0.170

1.9 6.4 17.2 8.6

was not possible to analyse C/Ti and C/W atomic ratios. Looking at W/Ti surface atomic ratios, the differences between fresh and used photocatalysts revealed complementary information (Table 6): it seemed that the difference in W/Ti surface atomic ratios increased with WO3 content and in a higher extend on 18% WO3 /TiNT catalysts. Thus, it can be supposed that there is a preferential deposition on W surface species in comparison to Ti ones, and especially in the case of 18 wt.% WO3 loading. It is worthy to recall here that the 18% WO3 /TiNT photocatalyst, also identified as the material exhibiting the best stabilized activity, corresponds to a WO3 content close to the theoretical monolayer coverage, assuming previously that only the external surface of the nanotubes were accessible for WO3 nanoparticles deposition.

4. Discussion and conclusion The studies related to UV-A and solar light photocatalytic removal of MEK on WO3 -modified TiO2 nanotubes varying in their WO3 content, as well as complementary and detailed characterizations performed on fresh and used photocatalysts allowed us to get more insights about correlations between WO3 content, surface composition and surface species responsible for photocatalytic activity but also about the causes of deactivation. It has been observed in this study that TiO2 nanotubes catalysts required a calcination step resulting in the formation of TiO2 (B) as the main crystallographic phase (although the presence of anatase TiO2 could not be excluded) to achieve photocatalytic activities. All

Fig. 14. XPS spectra of W4f region of dried, calcined and WO3 -modified TiNT samples after solar light-activated photocatalytic elimination of MEK.

54

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

the photocatalysts tested exhibited the same general behaviour, i.e. a initial maximum activity followed by a deactivation to reach a stabilized activity. Nevertheless, the level of the initial activity, of the deactivation and of the stabilized efficiency was strongly dependant on WO3 content. In addition, it is also worthy to underline that the impact in terms of photocatalytic efficiency, CO2 selectivity, carbon deficit and deactivation was also linked to the activation wavelengths, either UV-A or simulated-solar light. Under UV-A irradiation, the highest the initial activity, the less pronounced the deactivation and thus the greater the stabilized activity. Thus, in these activation conditions, the initial maximum activity decreased with increasing the amount of WO3 , at the exception of the 18 wt.% WO3 /TiO2 nanotube sample. In the same manner, the deactivation was lower for the lowest WO3 concentrations to reach stabilized activities thus increasing with lowering WO3 loadings (at the same exception of 18% WO3 /TiO2 ). It has also been shown that the selectivity towards gaseous CO2 is inversely proportional to the extent of the deactivation phenomenon, this later one being directly correlated with the carbon deficit. From TGA analyses, it could be concluded that this carbon deficit, responsible for deactivation, was due to adsorbed or deposited species resulting from the photocatalytic oxidation of MEK. To summarize, under UV-A activation, the best photocatalytic performance for MEK elimination was obtained on 18 wt.% WO3 /TiNT material. For other WO3 contents (lower or higher), the performances were worse than on the calcined sample, without WO3 modification. Under solar light activation, the photocatalytic behaviours are inversely correlated with the WO3 content and the calcined WO3 free catalysts did not exhibit the best initial activity. In this case, the highest the WO3 content, the lowest the initial activity, but the less pronounced the deactivation and the highest the stabilized activity. It must also be underlined that one of the main difference between UV-A and solar light illumination lies in the deactivation phenomenon. For solar light activation, it increases when decreasing WO3 concentration, at the contrary of UV-A activation. It is also worthy to mention that gaseous CO2 formation was never detected in the case of solar light illumination, meaning that all the incoming carbon remained on the catalyst surface. This is certainly the reason for the highest deactivation. Furthermore, from XPS studies, it was observed that, even if the surface chemical environments of Ti4+ and W6+ were not affected during the photocatalytic test, there was a preferential deposition on W surface species which increased with the content of WO3 (at the exception of 18 wt.% WO3 ). In contrast to UV-A activation, WO3 modification globally induced enhancement of the activity, compared to the only calcined sample. The best stabilized activity was observed on 18 wt.% WO3 /TiNT sample. These different behaviours observed on WO3 -modified TiO2 nanotubes and their significant differences between UV-A and solar light activation could result from five main causes: (i) the potential of WO3 semiconductor to absorb a part of visible light, (ii) the formation of heterojunctions and resulting charge transfers when coupling TiO2 and WO3 semiconductors, (iii) the contact between WO3 particles and TiO2 nanotubes, (iv) the electron trapping effect of WO3 , and (iV) the surface acidity of WO3. (i) It was shown from UV–vis absorption (Fig. 8) that the association of TiO2 nanotubes with WO3 particles shifted the light absorption towards visible light wavelengths, in accordance with the theoretical band gap value of WO3 (∼2.8 eV). This shift seemed to increase with the WO3 content. (ii) Coupling TiO2 with another semiconductor having its valence and its conduction bands suitably disposed, is well known to enhance the photogenerated charge separation through electron and hole transfer, thus limiting their recombination [53,54]. However, the extent of these charge transfers is not the same under UV-A, pure visible and solar light (small

amount of UV-A + visible) light activation, as represented in Fig. 15a–c, respectively. It must also be mentioned that, because of the more favourable energetic difference between conduction bands of TiO2 and WO3 than between their valence bands, electron transfer from the conduction band of TiO2 to the conduction band of WO3 are favoured in comparison to hole transfer from valence band of WO3 to that of TiO2 (in dotted line). Of course, in these cases, the improvement of photooxidation may be strongly influenced by the contact between the two semiconductors, which should be the more intimate to ensure optimal charge transfer and separation. In addition, we can observe from Fig. 15 that vectorial (twoways) charge transfer occurs only in the case of UV-A and solar light activation. In the case of pure visible light irradiation, only a less favourable one-way hole transfer can take place from the valence band of WO3 to that of TiO2 . (iii) The contact between WO3 and TiO2 nanotubes is crucial, taking also into account the one-dimensionality of TiO2 . A schematic representation of WO3 particles on the external surface of TiO2 nanotubes is given in Fig. 16, as it was deduced from the present study, that the major part of WO3 particles, whatever their content, are located on the external surface of the TiNT. It was also estimated that the theoretical monolayer coverage is about 20 wt.%. In this case, the concentration of 18 wt.% WO3 may exhibit the best contact between WO3 particles and TiO2 nanotubes. (iv) The beneficial effect of WO3 on photocatalytic activity can be explained by the loading of electron-accepting species on the TiO2 surface, which is a good way of slowing down electron–hole recombination. So, according to some authors [55,56] and since W(VI) can be easily reduced in W(V) [57], it can be deduced that the photoexcited electrons in the conduction band of TiO2 can be easily accepted by WO3 , following the scheme W6+ + e− → W5+ . In this case, WO3 helps in the (TiO )cb 2

trapping of photogenerated electrons. This has also been confirmed in literature [58] by charge transfer measurements using vis–UV diffuse reflectance spectroscopy of photoproduced electrons from TiO2 to W6+ ; this charge transfer increased as the WO3 content increased. But, on the other hand, it must be taken into account that too high-loaded WO3 /TiO2 catalysts could act as recombination centres for electron–h+ pairs, according to the 6+ scheme W6+ + e− → W5+ but W5+ + h+ (TiO )vb → W . So, (TiO )cb 2

2

as observed and confirmed in this paper, an optimum in WO3 content is necessary to optimize its role of photogenerated electron trapping. (v) Zeta potential measurements (Fig. 6) showed that increasing WO3 content led to an increase of surface acidity. Thus, the WO3 /TiNT acidic surfaces have higher affinity for species with unpaired electrons. That is why it may adsorb a greater amount of OH− or H2 O, which is considered to be necessary for the generation of OH radicals. At the same time, the catalyst is able to more easily adsorb organic reactants with polarized functional groups, like MEK, having high affinity with acidic surfaces. Nevertheless, it seems obvious that this affinity towards organics should not be too important due to the risk of limiting their degradation by a too strong adsorption.

It is evident that these different consequences of the modification of TiO2 nanotubes when loading with WO3 particles have not the same impact on photocatalytic activity towards MEK photooxidation. We suppose, regarding our results, that the two most important causes lay on the capacity of WO3 to absorb a part of visible light wavelengths (in the blue region) and on the photogenerated charge transfers induced by the coupling of WO3 particles with TiO2 nanotubes. In this last case, the amount of WO3

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

55

Fig. 15. Concept of TiO2 and WO3 semiconductor coupling: (a) under UV-A irradiation, resulting in e− and holes charge transfers; (b) under visible light irradiation, resulting in e− and holes charge transfers; (c) under solar light (UV-A + visible light) irradiation, resulting in e− and holes charge transfers.

determining the quality of the contact, between these two semiconductors, which was optimized at near monolayer coverage, is crucial. Consequently, the different behaviours can be explained as follows. 4.1. Under UV-A irradiation Considering the level of decrease in initial activity, it has been observed that increasing the content of WO3 yielded a decrease in performances, except for 18 wt.% WO3 content. Indeed, even if in addition to TiO2 , WO3 can also be activated by UV-A irradiation and photooxidize MEK, its quantum yield is much lower. Consequently, increasing WO3 content resulting in the proportional diminution of TiO2 sites that have more important photooxidation activities, was detrimental for the overall performance of WO3 -modified material. We can assume that the exception observed for 18 wt.% WO3 is specific at the near monolayer coverage obtained, resulting in an optimized contact between the two semi-conductive materials, thus allowing better vectorial photogenerated charges transfer and

separation (Fig. 15a). In this case, the benefit in terms of limitation of charges recombination may largely compensate the detrimental effect due to the increase of WO3 and decrease of TiO2 sites. Looking at the CO2 selectivity, the deficit of carbon in the gas phase on TiO2 nanotubes could be explained, pointing out that a part of CO2 produced by the photoxidation of MEK may react with reactive oxygen species like superoxide radicals (resulting from the reaction of photogenerated electrons with adsorbed oxygen) to produce carbonate radicals [59] (Fig. 17, top). Those carbonates radicals yield further deposition of carbonates species, leading to the poisoning of TiO2 active sites. We can also assume that the increase in deactivation when increasing WO3 content is due to:

(i) the additional formation of carbonates on WO3 particles, coming from the formation, on the same particles, of CO2 resulting from the photooxidation of MEK. Even if these reactions are less favourable on WO3 than on TiO2 because of lower quantum yield, their probability proportionally increases with WO3 concentration,

Fig. 16. Schematic representation of WO3 particles deposited on the external surface of TiNT-based samples.

56

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

Fig. 17. Schematic representation of the evolution of reactants and products during UV-A-activated elimination of MEK. On TiO2 (top) and WO3 -modified TiO2 (bottom). The probability of reaction decreases going from full dotted arrows.

(ii) the migration of carbonates most likely produced on TiO2 sites on WO3 particles. In this case, the driving force may certainly be the more acid WO3 surfaces, as observed with zetametry measurements. It must be remembered here that neither TiO2 nor WO3 surfaces can promote photogenerated electron to directly react with CO2 . 4.2. Under solar light irradiation (UV-A + visible light) It must be recall from Fig. 1b that the as-considered solar light irradiation contained 1.6% of UV-A light, in terms of irradiance. Considering first the good initial activity of non-modified WO3 –TiO2 nanotubes that do not absorb visible light (UV–vis absorption in Fig. 8), one may deduce that the contribution of UV-A irradiation is non-negligible in term of activation, even if its proportion is very low. Addition of WO3 at 4 and 9 wt.% yields an increase of initial activity compared to non-modified TiO2 nanotubes. This may certainly be the result of the addition of WO3 particles able to absorb a part of visible light and thus to both (i) directly photo-oxidize MEK and to (ii) photo-oxidize MEK on sensitized TiO2 resulting from the photogenerated holes transfer from WO3 valence band to TiO2 valence band (Fig. 18). However, as in the case of WO3 -free TiO2 nanotubes, the contribution of UV-A irradiation through the direct photooxidation of MEK on TiO2 , in addition to the photooxidation on visible light-sensitized TiO2 , must also be considered as non-negligible (Fig. 19). This may also explain, as it was assumed previously in the case of pure UV-A irradiation, the decrease in initial activity when increasing further the amount of WO3 , due to the proportional decrease of TiO2 sites. Furthermore, as visible light activation of WO3 remains all the same the major phenomena, the impact of UV-A activated photogenerated electron transfer

from conduction band of TiO2 to conduction band of WO3 is minor, thus limiting the favourable effect of semi-conductor coupling on vectorial photogenerated charge separation. This reason may also explain why 18% WO3 /TiO2 nanotubes, which were supposed to exhibit the best contact between WO3 and TiO2 do not display better initial activity than the other WO3 -modified samples. Considering CO2 selectivity, the original feature is the absence of CO2 formation on WO3 -modified TiO2 nanotubes and its low value (∼2%) on unmodified ones. As in the case of pure UV-A activation, the deficit of carbon in the gas phase, associated with the low CO2 selectivity can be the result from the reaction of CO2 with reactive oxygen species to produce carbonate radicals. But in contrast with pure UV-A activation, photogeneration of electrons on the conduction band of TiO2 may happen in a lower extend in comparison to photogeneration of electrons on the conduction band of WO3 , especially with the increase in WO3 concentration. This may be the reason why there is a preferential carbonates formation on WO3 particles in our solar light illumination conditions, in addition with the lowest probability of CO2 formation on TiO2 surfaces. Even

Fig. 18. Schematic representation of the evolution of reactants and products during simulated visible light-activated elimination of MEK. The probability of reactions decreases going from full to dotted arrows.

Y. Yamin et al. / Journal of Photochemistry and Photobiology A: Chemistry 245 (2012) 43–57

57

Fig. 19. Schematic representation of the evolution of reactants and products during simulated solar light-activated elimination of MEK. The probability of reactions decreases going from full to empty and to dotted arrows.

considering some minor CO2 formation on TiO2 , one may assume that the one-dimensionality of the nanotubes should facilitate CO2 migration to WO3 particles, due to good contact between the nanotubes and WO3 particles. Furthermore, this contact and transfer should have been the best on the 18 wt.% near monolayer coverage, explaining the low deactivation and thus best stabilized conversion observed on this sample. This assumption is confirmed by XPS studies revealing, after test, an increase of deposited surface species preferentially on WO3 surfaces when increasing WO3 content, with the highest amount on 18 wt.% WO3 content. This specific accumulation of carbonates (rather than the accumulation and poisoning by polymerization of organic compounds derived from the photooxidation of MEK) on WO3 particles may thus result in the specific poisoning of WO3 sites, leaving TiO2 surfaces free for further MEK adsorption and photooxidation. This phenomenon can explain the evolution of deactivation with WO3 content and the obtained stabilized activities. In these stabilized conditions under solar light irradiation, WO3 helps for accumulation of carbonates poisons, allowing the minor photooxidation phenomena on TiO2 surfaces to be maintained.

References [1] D.F. Ollis, E. Pelizetti, N. Serpone, Environmental Science and Technology 25 (1991) 1522. [2] D.F. Ollis, H. Al-Ekabi (Eds.), Photocatalytic Purification and Treatment of Water and Air, Elsevier, Amsterdam, 1993. [3] S. Ijima, Nature 354 (1991) 56. [4] M. Remskar, Advanced Materials 16 (2004) 1497. [5] G.K. Mor, K. Shankar, M. Paulose, O.K. Varghese, C.A. Grimes, Nano letters 5 (2005) 191. [6] A. Ghicov, H. Tsuchiya, J.M. Macak, P. Schmuki, Electrochemistry Communications 7 (2005) 576. [7] J.R. Jennings, A. Ghicov, L.M. Peter, P. Schmuki, A.B. Walker, Journal of the American Chemical Society 130 (2008) 13364. [8] P. Roy, S. Berger, P. Schmuki, Angewandte Chemie International Edition 50 (2011) 2904. [9] S.M. Liu, L.M. Gan, L.H. Liu, W.D. Zhang, H.C. Zeng, Chemistry of Materials 14 (2002) 1391. [10] J.H. Jung, H. Kobayashi, K.J.C. van Bommel, S. Shinkai, T. Shimizu, Chemistry of Materials 14 (2002) 1445. [11] T. Kasuga, M. Hiramatsu, A. Hoson, T. Sekino, K. Niihara, Langmuir 14 (1998) 3160. [12] X. Sun, Y.D. Li, Chemistry-A European Journal 9 (2003) 2229. [13] T. Kasuga, M. Hiramatsu, A. Hoson, T. Sekino, K. Niihara, Advanced Materials 11 (1999) 1307. [14] B.D. Yao, Y.F. Chan, X.Y. Zhang, W.F. Zhang, Z.Y. Yang, N. Wang, Applied Physics Letters 82 (2) (2003) 281. [15] D.V. Bavykin, J.M. Friedrich, F.C. Walsh, Advanced Materials 18 (2006) 2807. [16] Y. Guo, N.-H. Lee, H.-J. Oh, C.-R. Yoon, K.-S. Park, W.-H. Lee, Y. Li, H.-G. Lee, K.-S. Lee, S.-J. Kim, Thin Solid Films 516 (2008) 8363. [17] T. Tachikawa, S. Tojo, M. Fujitsuka, T. Sekino, T. Majima, Journal of Physical Chemistry B-Letters 110 (2006) 14055. [18] H. Tokudome, M. Miyauchi, Chemistry Letters 9 (2004) 1108. [19] F. Dong, W. Zhao, Z. Wu, Nanotechnology 19 (2008) 365607. [20] Z. Wu, F. Dong, W. Zhao, H. Wang, Y. Liu, B. Guen, Nanotechnology 20 (23) (2009) 235701.

[21] M. Hodos, E. Horvath, H. Haspel, A. Kukovecz, Z. Konya, I. Kiricsi, Chemical Physics Letters 399 (2004) 512. [22] H. Li, B. Zhu, Y. Feng, S. Wang, S. Zhang, W. Huang, Journal of Solid State Chemistry 180 (2007) 2136. [23] M.W. Xiao, L.S. Wang, X.J. Huang, Y.D. Wu, Z. Dang, Journal of Alloys and Compounds 470 (2009) 486. [24] Q. Li, T. Kako, J. Ye, Chemical Communications 46 (2010) 5352. [25] Y.-C. Nah, A. Ghicov, D. Kim, S. Berger, P. Schmuki, Journal of the American Chemical Society 130 (48) (2008) 16154. [26] A. Benoit, I. Paramasivam, Y.-C. Nah, P. Roy, P. Schmuki, Electrochemistry Communications 11 (2009) 728. [27] I. Paramasivam, Y.-C. Nah, C. Das, N.K. Shrestha, P. Schmuki, Chemistry-A European Journal 16 (2010) 8993. [28] E.P. Barrett, L.G. Joyner, P.H. Halenda, Journal of the American Chemical Society 73 (1951) 373. [29] R.S.H. Mikhail, S. Brunauer, E.E. Bodor, Journal of Colloid and Interface Science 26 (1968) 45. [30] S. Doniach, M. Sunjic, J.C. Phys, Solid State Physics 3 (2) (1970) 285. [31] D.A. Shirley, Physical Review B 5 (1972) 4709. [32] J.H. Scofield, Journal of Electron Spectroscopy and Related Phenomena 8 (1976) 129. [33] V. Keller, P. Bernhardt, F. Garin, Journal of Catalysis 215 (2003) 129. [34] E. Barraud, F. Bosc, D. Edwards, N. Keller, V. Keller, Journal of Catalysis 235 (2) (2005) 318. [35] H.-L. Kuo, C.-Y. Kuo, C.-H. Liu, J.-H. Chao, C.-H. Lin, Catalysis Letters 113 (2007) 7. [36] A.R. Armstrong, G. Armstrong, J. Canales, P. Bruce, Angewandte Chemie International Edition 43 (2004) 2286. [37] J.G. Yu, H.G. Yu, B. Cheng, X.J. Zhao, Q.J. Zhang, Journal of Photochemistry and Photobiology A 182 (2006) 121. [38] V. Puddu, R. Mokaya, G.L. Puma, Chemical Communications 45 (2007) 4749. [39] Z.J. Gu, H.Q. Li, T.Y. Zhai, W.S. Yang, Y.Y. Xia, Y. Ma, J.N. Yao, Journal of Solid State Chemistry 180 (2007) 98. [40] D. Bavykin, J.M. Friedrich, Advanced Materials 18 (2006) 2807. [41] S.J. Gregg, K.S.V. Sing, Adsorption, Surface Area and Porosity, Academic Press, 1982. [42] Y. Xie, Y. Tang, Advances in Catalysis 37 (1990) 1. [43] K. Kosmulski, Chemical Properties of Material Surfaces, Marcel Dekker, 2001. [44] J. Ren, J. Shen, S.C. Lu, Science and Technology of Particles Dispersion, Chemical Industry Press, Beijing, 2005, pp. 74–85. [45] H. Zhu, C. Zhang, Y. Tang, J. Wang, B. Ren, Y. Yin, Carbon 45 (2007) 203. [46] M. Grandcolas, A. Louvet, N. Keller, V. Keller, Angewandte Chemie International Edition 48 (1) (2009) 161. [47] M. Grandcolas, L. Sinault, F. Mosset, A. Louvet, N. Keller, V. Keller, Applied Catalysis A: General 391 (1–2) (2011) 455. [48] L.F. Cheng, X.T. Zhang, B. Lin, H.Z. Wang, Y.C. Li, Y.B. Huang, Z.L. Du, Nanotechnology 16 (2005) 1341. [49] V. Keller, F. Garin, Catalysis Communications 4 (2003) 377. [50] S. Kato, Y. Hirano, M. Iwata, T. Sano, K. Takeuchi, S. Matsuzawa, Applied Catalysis B: Environmental 57 (2005) 109. [51] Y. Suzuki, S. Yoshikawa, Journal of Materials Research 19 (2004) 982. [52] M.A. Fox, M.T. Dulay, Chemical Reviews 93 (1993) 341. [53] A. Di Paola, L. Palmisano, V. Augugliaro, Catalysis Today 58 (2000) 141. [54] Y. Bessekhouad, D. Robert, J.V. Weber, Catalysis Today (3–4) (2004) 315. [55] G. Marci, L. Palmisano, A. Sclafani, A.M. Venezia, R. Campostini, G. Carturan, C. Martin, V. Rives, G. Solana, Journal of the Chemical Society, Faraday Transactions 92 (1996) 819. [56] Y.R. Do, W. Lee, K. Dwight, A. Wold, Journal of Solid State Chemistry 108 (1994) 198. [57] K. Tennakone, O.A. Heperuma, J.M.S. Bandara, W.C.B. Kiridena, Semiconductor Science and Technology 7 (1992) 423. [58] C. Martin, L. Palmisano, V. Rives, A. Sclafani, Catalysis Letters 49 (1997) 235. [59] B. Ezraty, M. Chabalier, A. Ducret, E.T. Maisonneuve, S. Dukan, EMBO Reports 12 (2011) 321.