Structural and photocatalytic properties of Pd-deposited semiconductors with different morphology

In this work, we studied the effect of Pd deposition (0.3 wt%) on the structural and photocatalytic properties of TiO2 anatase with different morphologies (nanoparticles, nanotubes and nanofibers). The Pd-deposited semiconductor samples were synthesized by a simple deposition–precipitation method and used as catalysts for photocatalytic degradation of p-nitrophenol under visible light illumination. Pd-deposited TiO2 nanofibers showed higher photocatalytic activity than Pd-deposited TiO2 nanoparticles or nanotubes. Pd-deposited (0.3 wt%) CeO2 and WO3 nanofibers were also synthesized to investigate whether the advantages of nanofiber morphology extend to other semiconductors, with lower and higher band gap energies. The performance of Pd-deposited CeO2 nanofibers was superior to that of all other Pd-deposited samples. The samples were characterized by elemental analysis, XRD, SEM, TEM, DR UV-vis, N2 physisorption and XPS to investigate the role of structural properties on photocatalytic performance. The XRD data indicated that the crystal structures of TiO2, WO3 and CeO2 were not modified after Pd deposition and TEM indicated that the Pd was well dispersed on the semiconductor surfaces, with particle sizes <20 nm. Pd deposition changed the morphology of the CeO2 nanofibers to a unique ‘mat-like’ structure, with small (4 nm) Pd nanoparticles. This latter sample had the highest photocatalytic activity, which was attributed largely to high photocurrent density. Conditions for photocatalysis were investigated and Pd-deposited CeO2 nanofibres were shown to be reusable for at least five cycles, without significant loss of photocatalytic activity. Our study suggests that Pd-deposited CeO2 nanofibres could be used industrially to degrade p-nitrophenol.


Introduction
Photocatalysis has emerged as a viable alternative for many applications, including utilization of solar energy, 1 CO 2 reduction 2 and removal of organic contaminants. 3 The development of suitable photocatalytic processes for the purication of contaminated waters and air is of particular interest since photocatalysts can completely degrade organic contaminants into CO 2 and H 2 O. p-Nitrophenol (p-NP) is commonly used as an intermediate in the manufacture of medicines, pesticides, dyes and ne chemicals 4 and is oen discharged in waste water. This creates environmental problems since p-NP is toxic and carcinogenic and has very low biodegradability. It is, therefore, very important to remove p-NP from effluents before they are discharged into the ecosystem. 5 Since its discovery by Fujishima and Honda, 1 TiO 2 has attracted considerable interest as a wide band gap semiconductor. TiO 2 has frequently been used as a heterogeneous photocatalyst because it is inexpensive and has low toxicity. The main drawbacks in effective commercial application of TiO 2 are its large band gap energy (3.2-3.4 eV), which precludes the use of sunlight, and agglomeration of particles aer use, which has a negative effect on photocatalytic activity. 6 A number of strategies have been adopted to shi the TiO 2 absorption spectrum towards the visible light region to enable its use as an effective photocatalyst in sunlight. The incorporation of precious and non-precious metals 7 into the TiO 2 lattice is one way to tune the band gap energy. 8 The precious metals have been suggested to enhance photocatalytic activity by (i) enhancing the adsorption of organic molecules onto the photocatalytic surface; (ii) increasing electron excitation and electron-hole separation through surface plasmon resonance of precious metal nanoparticles on the semiconductor and (iii) acting as electron traps to hinder electron/hole recombination. 9 The presence of Pd on the surface of ZnO has been shown to have a large effect on photocatalytic activity, in both the UV and visible regions of the spectrum. Liqiang et al. 10 attributed the increased photocatalytic efficiency in the presence of Pd to increased amounts of adsorbed oxygen on the surface. TiO 2 nanotubes incorporating Pd nanoparticles have been used for the photocatalytic decomposition of dyes under sunlight. 11 Pd/ TiO 2 nanotubes completely degraded the dye more quickly than TiO 2 nanotubes (150 min versus 250 min). Incorporation of Pd into TiO 2 also improved hydrogen production compared with pure TiO 2 . 12 Au-TiO 2 and Pt-TiO 2 catalysts have been reported to provide 90-100% degradation of p-NP in 300 min using visible light irradiation. 13 Degradation of p-NP in visible light has also been reported using Fe 3 O 4 -N-doped TiO 2 with continuous air bubbling 14 and using TiO 2 -graphene-Pd nanowires. 15 The different photocatalytic performances observed for different samples are likely inuenced by factors such as composition, crystallinity, surface area and morphology of the active particles. 16 Different materials investigated include macro-mesoporous TiO 2 , 17 WO 3 microspheres 18 and a Pd-CeO 2 nanocomposite, which had enhanced photocatalytic activity under visible light. 19 Semiconductor materials with different shapes (spheres, cubes, tubes, wires, rods, sheets and akes), sizes and dimensions can be synthesized by varying the method of preparation and calcination temperature. 20 Suzuki et al. 21 suggested that there are still many possibilities for improving the photocatalytic activity of semiconductors, especially by altering their morphology. Campelo et al. 22 showed that there are some benets in using nanosized Pd-containing photocatalysts since the Pd nanostructures have enhanced lightharvesting ability. Pd nanostructures were reported to show better photocatalytic activity compared with Ag and Au nanostructured catalysts.
In the present work, we have studied the effect of Pd deposition (0.3 wt%) on the structural and photocatalytic properties of nanostructured semiconductors with different morphologies (particles, tubes and bers). The Pd-deposited samples were synthesized by a simple deposition-precipitation method and the synthesized materials were used as catalysts for photocatalytic degradation of p-NP under visible light. Pd-deposited TiO 2 nanobers (Pd-Ti-NF) showed higher photocatalytic activity than Pd-deposited TiO 2 nanoparticles (Pd-Ti-NP) or nanotubes (Pd-Ti-NT). Because of the superiority of the nano-ber morphology, we also wished to synthesize Pd-deposited WO 3 and CeO 2 nanobers to investigate whether the effect of morphology extended to other semiconductors with lower and higher band gap energies. The synthesized samples were characterized structurally and morphologically using different techniques and their efficiency as catalysts for photocatalytic degradation of p-NP under visible light at room temperature was compared.

Materials
All the commercial reagents were analytical grade and used as received without any purication.  (Ti-NP). TiO 2 -NPs have been synthesized by using modied sol-gel method as described in our previous publication. 23 The synthesized TiO 2 -NPs were calcined in air at 400 C for 5 h.
2.2.2 TiO 2 nanotubes (Ti-NT). TiO 2 -NTs were synthesized by alkaline hydrothermal synthesis method reported by Kasuga et al. 24 The synthesized TiO 2 nanotubes sample was dried in air at 100 C for 12 h and calcined in air at 400 C for 5 h.
2.2.3 TiO 2 nanobers (Ti-NF), CeO 2 nanobers (Ce-NF) and WO 3 nanobers (W-NF). Ti-NF, Ce-NF and W-NF materials were purchased from Pardam nanotechnology, Czech Republic. The purchased materials were calcined in air at 400 C for 5 h before their use as supports to deposit the palladium.
2.2.4 Preparation of Pd deposited semiconductors. Deposition-precipitation method was used to deposit the 0.3 wt% of Pd over the TiO 2 , CeO 2 and WO 3 semiconductors. In this method, palladium chloride was hydrolyzed by the addition of 1.0 M Na 2 CO 3 on the semiconductor from the uniformly dispersed palladium chloride solution. 25 The pH of total contents was adjusted to 10.5 and it was maintained for 1 h. Then the solid was ltered, washed with deionized water several times (until all the chloride ions were not detected by silver nitrate solution). Finally, the ltered cake was dried in an electric oven at 110 C for 2 h, subsequently it was calcined at 400 C for 4 h in air.

Characterization of synthesized samples
The powder XRD patterns are collected by using Philips PW1700 diffractometer at room temperature. The XRD patterns were obtained using Cu Ka radiation and graphite monochromator with automatic divergent slit. The collected patterns were matched with standard JCPDS les. The crystallite size was calculated from (101) reection of anatase phase using the Scherrer formula.
where 'D' is the average crystallite size of the phase under investigation, B is the Scherer constant (0.89), 'l' is the wavelength of the X-ray beam used (1.54056Å), 'b 1/2 ' is the full width at half maximum (FWHM) of the diffraction peak and 'q' is the diffraction angle. A Philips CM200FEG microscope operated at 200 kV, equipped with a eld emission gun was used for TEM analysis. The coefficient of spherical aberration was C s ¼ 1.35 mm. Highresolution images with a pixel size of 0.044 nm were taken with a CCD camera. Field emission scanning electron microscopy (FESEM) images of the samples were collected using on a FEI Nova NANOSEM 230 spectrophotometer. N 2 -physisorption measurements were carried at À196 C as per the recommendations of the IUPAC using ASAP 2010 Micromeritics Instrument. Specic surface area (S BET ) values were calculated by applying the BET equation. The average pore width was calculated by the BJH method as well as the average pore width method (employing the relation 4 V p /S BET , where V p is the specic pore volume).
DR UV-vis spectra for synthesized samples were collected using U-4100, Hitachi spectrophotometer equipped with an integrating sphere in the wavelength range 200-800 nm to measure the reectance spectra of each sample. Band gap of all the synthesized samples was determined using Kubelka-Munk method. The Kubelka-Munk transformation (K) was estimated using the equation; where R is the reectance. The wavelengths (nm) were translated into energies (eV)*, and a plot of (K Â hn) 0.5 vs. E was made. The bandgap energy (eV) was estimated as the intersection of the two slopes in the curve obtained. The bulk composition of synthesized samples was determined using inductively coupled plasma-atomic emission spectroscopy (ICP-AES; IRIS Advantage). The XPS measurements were carried out using ESCA Lab 220i-XL spectrometer using Al Ka X-rays. Peak shi due to charge compensation was corrected using the binding energy of C 1s peak. The data was acquired using pass energy of 100 eV, dwell time 200 ms with step size of 0.1 eV and 30 scans.

Photocatalytic degradation of p-nitrophenol (p-NP)
The photocatalytic degradation of p-NP was performed in a Pyrex glass reactor using synthesized catalysts under the visible light illumination for various time intervals. The p-NP (100 ppm) solution was prepared in 100 mL deionized water and 0.4 g L À1 of catalyst was added to the solution. The resulting suspension was equilibrated by stirring for 45 min to stabilize the adsorption of p-NP over the surface of the catalyst in the dark. The photocatalytic degradation of p-NP was monitored by measuring the absorbance of p-NP at regular time interval using a SmartSpec 3000, Bio-Rad Laboratories Inc., USA. The degradation percentage was calculated using the expression Where 'C 0 ' is the concentration of p-NP before illumination and 'C' is the concentration of p-NP aer a certain irradiation time. The stability of the photocatalysts was studied by measuring reusability of the samples. Aer the rst cycle of the activity measurement, the catalyst was ltered from the reactor and the aliquots by centrifugation. The obtained catalyst was thoroughly washed with distilled water and acetone. The catalyst was dried at 80 C for 2 hours and then reused for the next cycle of the photocatalysis measurements. Similarly, the experiment was repeated for ve cycles to study the stability of the catalyst for repetitive use.

Powder X-ray diffraction
Powder X-ray diffraction (XRD) measurements were used to investigate the phase structure and particle size of the catalysts.
The XRD patterns ( Fig. 1) showed that the Pd-deposited TiO 2 nanostructures were composed of anatase, as a single phase [JCPDS no. . The diffraction peaks at 2q ¼ 25.3 , 37.8 , 48.0 , 55.0 and 62.7 could be assigned to (101), (004), (200), (211) and (204) planes, respectively, of anatase. The patterns obtained also conrmed the absence of rutile and brookite phases. The XRD pattern of Pd-Ti-NP showed sharp diffraction peaks whereas the Pd-Ti-NF sample showed broad diffraction peaks. This indicates that Pd-Ti-NF has a smaller crystallite size than Pd-Ti-NT or Pd-Ti-NP. The Pd-Ce-NF sample showed reections attributable to (111), (200), (220) and (311) planes, corresponding to the cubic uorite structure of CeO 2 [JCPDS no. 75-0076], while the Pd-W-NF sample showed diffraction peaks corresponding to the monoclinic WO 3 phase. Reections due to Pd metal or PdO were not detected by XRD in the 2q region from 20 to 80 for any of the Pd-deposited semiconductor samples. It is possible that the deposited Pd is highly dispersed on the surface of the semiconductors or the size of the crystallites could be smaller than the detection range of the XRD technique.
The average crystallite sizes of the semiconductor phases (TiO 2 , CeO 2 and WO 3 ) were calculated using the Scherrer equation. The crystallite sizes of TiO 2 nanoparticles, TiO 2 nanotubes and TiO 2 nanobers were 180 nm, 83 nm and 25 nm, respectively ( Table 1). The crystallite size of none of the nanostructures was changed aer depositing Pd, indicating that the Pd was deposited onto the surface of the semiconductors and not incorporated into the lattices.

Morphology studies
Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) were used to investigate the size and morphology of the semiconductors and Pd-deposited semiconductors in order to determine the importance of these factors for photocatalytic ability. TEM images of the Pd-deposited semiconductors aer calcination at 400 C for 4 h are shown in Fig. 2 (insets show SEM images before deposition). The bers in TiO 2 -NF were $3 mm in length and $9 nm in diameter. The length of the bers in Pd-Ti-NF was reduced from mm to nm. It is clear that Pd deposition and thermal treatment led to the formation of Pd nanoparticles and that some of the particles were coordinated with the surface of the nanobers. The SEM image of TiO 2 -NP particles showed irregularly shaped, spherical particles and the TEM image of Pd-Ti-NP showed very similar morphology, with dark shaded TiO 2 -NP particles attributable mostly to the presence of Pd nanoparticles. This sample consisted of agglomerated TiO 2 particles with diameters in the range 100-180 nm. The TiO 2 -NT consisted of anatase nanotubes with an average size of $50 nm. The Pd-Ti-NT sample also showed tube-like morphology, but with shorter NTs. The morphologies of Pd-Ce-NF and Pd-W-NF were remarkably different from those of bulk samples of CeO 2 -NF and WO 3 -NF. The TEM image of Pd-Ce-NF showed a unique 'mat-like' structure, with the deposited Pd visible as small black dots on the surface of the CeO 2 . Zhang et al. 26 also observed that CeO 2 adopts a readily distinguishable 'mat-like' morphology, and it appears that this unique structure inhibits the aggregation of metal nanoparticles. The TEM image of Pd-W-NF showed that the WO 3 had formed nanorods (length 50-150 nm), with the deposited Pd visible as small particles on the surface of the nanorods. These results clearly indicate that the deposition of Pd has a signicant effect on the morphology of CeO 2 and WO 3 nanobers.
All of the Pd-deposited semiconductor samples showed very small Pd nanoparticles, which were dispersed on the surface of semiconductor. Pd-Ti-NF and Pd-Ce-NF, particularly, had small uniform Pd nanoparticles, with average sizes in the range 3-5 nm. In contrast, Pd-Ti-NT, Pd-Ti-NP and Pd-W-NF showed slightly larger Pd particles, with average sizes of 8-15 nm. No agglomerated particles were observed, demonstrating that the deposition-precipitation method used in this study hindered aggregation of the Pd nanoparticles. Particle sizes measured by TEM analysis and crystallite sizes of the Pd-deposited semiconductor samples calculated using the Scherrer equation are compared in Table 1. Particle sizes of the semiconductor samples determined from TEM are smaller than those determined from XRD measurements, probably because of slight changes in morphology and agglomeration of the semiconductor samples. The particles sizes of Pd species dispersed on Pd-Ti-NP and Pd-W-NF were larger than those of Pd particles dispersed on Pd-Ce-NF and Pd-Ti-NF, probably because the interaction of Pd with nanobers is different in semiconductors with different morphologies. It has been reported that small Pd nanoparticles can dissociate into Pd atoms at high temperatures (800 C), 27 but since the Pd-deposited samples synthesized in the present study were heated to only 500 C, dissociation of Pd would not have occurred.

Diffuse reectance UV-vis spectroscopy
Diffuse reectance (DR) UV-vis spectra of bare and Pd-deposited semiconductor samples showed marked differences (Fig. 3). The bare TiO 2 nanostructured samples showed a UV-vis absorption peak centered at $250 nm, which can be attributed to electron transfer from the valence band to the conduction band of TiO 2 . 28 The DR UV-vis spectra showed that Pd-deposited semiconductor samples absorb more visible light than bare semiconductor samples. The Pd-deposited samples also showed a slight red shi in the absorption edge, indicating that Pd deposition is responsible for the absorption of light in the visible region. Additionally, a wide absorption peak, centered at 480 nm, appeared in all Pd-deposited semiconductor samples, indicating the presence of plasmonic Pd nanoparticles on the semiconductor surface. 29 This demonstrates the existence of surface plasmon resonance (SPR) in the Pd-deposited semiconductor samples.
Band gap energy values for all the synthesized samples were calculated by plotting the Kubelka-Munk function (K) against energy (eV). The calculated values of band gap energy for Pd-Ti-NF, Pd-Ti-NT and Pd-Ti-NP were 2.96 eV, 2.98 eV and 3.01 eV, respectively, which are all lower than that of bare anatase (3.40 eV). The band gap energy values for Pd-Ce-NF and Pd-W-NF samples were calculated as 2.92 eV and 2.88 eV, respectively, which are likewise smaller than values for bare CeO 2 and WO 3 samples. It can be concluded, therefore, that deposition of Pd had an effect on the optical properties of the different semiconductor samples.

N 2 physisorption
N 2 adsorption-desorption isotherms for bare and Pd-deposited semiconductor samples are shown in Fig. 4. The samples of bare TiO 2 -NP and Pd-Ti-NP clearly showed type-IV isotherms with H2-type hysteresis loops as per the IUPAC classication, demonstrating that both samples contain mesoporous pores. A relatively smaller hysteresis loop was observed aer Pd deposition on TiO 2 -NP. Samples of bare TiO 2 -NT and Pd-Ti-NT also showed type-IV isotherms. 30 Nanotube samples showed H1-type hysteresis loops, indicating the presence of uniform cylindrical pore structures in these samples. The isotherms of bare and Pd-deposited TiO 2 -NF, CeO 2 -NF and WO 3 -NF were identied as type-III. Nano-brous samples (specically WO 3 -NF) thus showed weak adsorption of N 2 molecules, indicating that these samples possess very few micropores or mesopores. An increase in the volume of adsorbed N 2 was, however, observed at high pressures (0.89-0.99), revealing the presence of macropores in the hollow structures of the nanobers. The specic surface area (S BET ), average pore diameter and pore volume were obtained from the N 2 physisorption measurements ( Table 2). The S BET was calculated using the BET equation. S BET values for bare TiO 2 -NP, TiO 2 -NT and TiO 2 -NF were 70 m 2 g À1 , 53 m 2 g À1 and 45 m 2 g À1 , respectively. S BET is thus dependent on the surface morphology of the sample; anatase particles had a higher surface area than tubes or bers. An increase in S BET was observed for the three Pd-deposited TiO 2 nanostructure samples ( Table 2). The highest increase in surface area was observed for Pd-Ti-NT (from 53 m 2 g À1 to 62 m 2 g À1 ) because of the mesoporous structure of nanotubes. S BET values of bare CeO 2 and WO 3 nanober samples were $30 m 2 g À1 and 7 m 2 g À1 , respectively, in agreement with a literature report. 31 S BET values of Pd-Ce-NF and Pd-W-NF were slightly increased (to 35 m 2 g À1 and 10 m 2 g À1 , respectively), due to modication of the morphology of the nanobers aer Pd deposition. Variations in the morphology of the semiconductors could be the reason for the changes in the textural properties of the samples. The pore volumes of bare TiO 2 -NP, TiO 2 -NT, TiO 2 -NF, CeO 2 -NF and WO 3 -NF were 0.077 cm 3 g À1 , 0.376 cm 3 g À1 , 0.062 cm 3 g À1 , 0.053 cm 3 g À1 and 0.021 cm 3 g À1 , respectively. Aer deposition of Pd, a small increase in pore volume was observed for all Pd-deposited semiconductor samples, irrespective of morphology. It was previously noted that, when the amount of Pd was too low to cover the support with a monolayer, formation of crystalline particles of Pd or PdO did not occur, which possibly explains the increased pore volume. The bare TiO 2 -NP sample showed a single peak for pore size distribution (PSD) at 37Å, corresponding to mesopores (ESI, Fig. S1 †), whereas a broad PSD peak (centered at 195Å) was observed for the bare TiO 2 -NT sample. Interestingly, the nano-ber samples (TiO 2 -NF, CeO 2 -NF and WO 3 -NF) showed a narrow range of PSD peaks (17-23Å). The maximum pore size observed in the TiO 2 -NT sample was in the mesoporous region, indicating that this sample has large spaces between the nanotubes. The data clearly show that the average pore diameter did not change signicantly aer Pd deposition.

X-ray photoelectron spectroscopy
The deconvoluted X-ray photoelectron (XP) spectra for all Pddeposited semiconductor samples are shown in Fig. 5. For the Pd-deposited TiO 2 , CeO 2 and WO 3 semiconductor samples, the Pd 3d doublet was deconvoluted into two components. The Pd 3d XP peaks at 335.4 eV and 340.7 eV are known to correspond to Pd 3d 5/2 and Pd 3d 3/2 of Pd 0 , respectively. 32 Two weaker features were also observed at 336.7 eV and 342.0 eV, which were assigned to Pd 2+ species (Pd 3d 5/2 and Pd 3d 3/2 , respectively). The percentage of Pd 0 and Pd 2+ species in all Pddeposited semiconductor samples was determined from the peak areas of the Pd 3d states ( Table 2). All the Pd-deposited semiconductor samples were found to contain both Pd 2+ and Pd 0 species, although the contribution of Pd 0 species was higher in the Pd-Ti-NF sample than in the other samples. Taking into account the general mechanism for depositionprecipitation on semiconductors, in which the Pd precursor is rst hydrolyzed and subsequently deposited on the semiconductor surface, it is likely that some of the hydrolyzed Pd species were reduced to Pd 0 species during the thermal treatment, while the rest remained as Pd 2+ species.
It has been reported that a pure sample of TiO 2 anatase shows mainly a Ti 2p doublet; Ti 2p 3/2 BE at 459.2 eV and Ti 2p 1/ 2 BE at 464.3 eV with a BE difference of 5.7 eV. The Ti 2p XP spectra of Pd-Ti-NP, Pd-Ti-NT and Pd-Ti-NT samples showed doublets consisting of two XP peaks. For the Pd-deposited TiO 2 nanostructured samples, the Ti 2p XP peaks can be tted to two components. The Ti 2p 3/2 peak at 459 eV was assigned to Ti 4+ species and the peak at 457.5 eV was assigned to Ti 3+ species. 33 These results demonstrate that different proportions of Ti 4+ and Ti 3+ species are present on the surfaces of the three samples.
The Ce 3d 5/2 XP spectrum of the Pd-Ce-NF sample showed XP peaks in the range 877-894 eV. 34 The peaks were tted to four different components at 880.3 eV, 882.4 eV, 884.5 eV and 888.9 eV. The peaks at 882.4 eV and 888.9 eV correspond to Ce(IV) species and the peaks at 880.3 eV and 884.5 eV correspond to Ce(III) species. 35 It is very clear from the spectrum that the peak due to Ce(IV) species is more dominant than peaks due other species. It has been reported that W 4f 7/2 peaks appear at 37.2 eV, 36.2 eV and 35.2 eV and that W 4f 5/2 peaks appear at 35.1 eV, 33.6 eV and 32.9 eV, for W 6+ , W 5+ and W 4+ species, respectively. 36 The XP spectrum of Pd-W-NF showed W 4f 7/2 and W 4f 5/2 peaks at 37.5 eV and 35.5 eV and at 35.8 eV and 33.5 eV, respectively. The presence of these peaks conrms the coexistence of W 6+ and W 5+ species in this sample.
The O 1s spectra of the Pd-deposited semiconductor samples showed three contributions at 527.6-528.6 eV, 529-530 eV and 531-532.5 eV. It has been reported that surface O 2À , OH À and H 2 O species show XP peaks at 530.9 eV, 531.6 eV and 533.5 eV, respectively. In the O 1s spectra, peaks around 529-530 eV and 531-532.5 eV were attributed to lattice oxygen in the semiconductor and surface hydroxyl species, respectively. The minor peak appearing at 527.6-528.6 eV could be due to the oxygen in the interactive Pd-O-M species. As was reported previously, electron exchange between CeO 2 and a few Pd atoms at the Pd-CeO 2 interface could lead to the formation of ionic Pd species. 37 Yu et al. 38   Detailed quantitative results from peak-tting of Pd 3d, Ti 2p, Ce 3d, W 4f and O 1s are presented in Table 3. The bulk and surface elemental composition of the Pd-deposited semiconductor samples was determined by inductively coupled plasma-atomic emission spectroscopy (ICP-AES) and XPS. Pddeposited samples showed a Pd content of $0.25 wt% (slightly lower than the theoretical Pd loading of 0.3 wt%). The  (Table 3).

Photocatalytic degradation of p-NP
The photocatalytic performance of the Pd-deposited semiconductors was evaluated by monitoring the degradation of p-NP over several minutes. Blank experiments were conducted to conrm that reaction did not take place in the absence of catalyst or visible light. UV-vis absorbance spectra of p-NP solutions (100 ppm) with different catalysts and different reaction times are shown in Fig. 6. As described above, the catalysts were equilibrated with aqueous p-NP solution to investigate the adsorption of p-NP onto the powdered catalysts. The spectra depicted in Fig. 6 were obtained aer equilibration in the dark. Negligible degradation of p-NP was observed under visible light in the absence of catalyst. The UV-vis spectrum of a standard p-NP solution showed a strong absorption band at 317 nm and a weak absorption band at 405 nm. The intensity of these absorption bands weakened and the bands eventually disappeared with increasing reaction time. The activity measurements clearly indicated that degradation of p-NP occurred over all the catalysts. The reduced absorbance of the aqueous p-NP solution aer the reaction is attributed to degradation of the aromatic ring present in p-NP. Pd-Ce-NF gave almost 99% degradation of p-NP within 60 min, whilst Pd-Ti-NF and Pd-W-NF catalysts gave 90% and 82% degradation aer the same reaction time. Pd-Ti-NT and Pd-Ti-NP required 90 and 120 min, respectively, for complete degradation of adsorbed p-NP molecules under similar reaction conditions. The photocatalytic degradation efficiency of the investigated samples was determined using eqn (3). Pddeposited semiconductors with nanobrous morphology offered better performance than Pd-deposited TiO 2 nanotubes and particles. Out of all the synthesized samples, Pd-Ce-NF was shown to be the best catalyst.
In order to understand the inuence of catalyst concentration on photocatalytic activity, the amount of each catalyst was systematically varied. The relationships between degradation efficiency and amount of catalyst are shown in Fig. 7(a). The optimum amount of all catalysts for degradation of p-NP was 0.4 g L À1 , except for Pd-W-NF where the optimum amount was 0.5 g L À1 . Degradation efficiency increased as the amount of catalyst was increased from 0.2 g L À1 to 0.4 g L À1 . As the amount of catalyst in the reactant solution was increased, the number of adsorption sites on the surface increased, allowing greater adsorption of p-NP onto the catalyst surface and increased degradation efficiency. Absorption increased with increasing amounts of catalyst in the reactant solution, likely until the surface was saturated, but degradation efficiency then began to decrease. The probable reason for the decrease in degradation efficiency is that visible light was unable to penetrate the solution and reach the surface in the presence of larger amounts of catalyst.
Together with the physicochemical properties of the catalyst, the pH of the organic pollutant solution is known to play a crucial role in aqueous phase semiconductor-mediated photocatalytic degradation of p-NP. 39 The inuence of the pH of the p-NP aqueous solution on the photocatalytic performance of the most active catalyst (Pd-Ce-NF) was studied over the pH range 2.0-12.0. It can be observed that all the catalysts offered better efficiency at lower pH values; maximum efficiency was observed at pH 4.0 ( Fig. 7(b)), when almost 98% of the p-NP was degraded aer 60 min under visible light. Highly acidic or alkaline solutions were not favorable for complete p-NP degradation because formation of hydroxyl radicals is inuenced by the pH of the reactant solution. The pH of the solution also inuences the ionic character of the semiconductor surface. For instance, the isoelectric point of CeO 2 is 4.43, 40 which means that the surface of CeO 2 could easily be positively charged at low pH but would be negatively charged at pH > 4.43. The hydroxyl group of p-NP would also be protonated in acidic solution, since the pK a of p-NP is 7.15. The observed results clearly show that protonation of p-NP favors degradation.
To optimize the reaction conditions, the effect of p-NP concentration on degradation was also studied. The p-NP concentration was varied over the range 0.005-0.025 M in the presence of Pd-Ce-NF (0.4 g L À1 ) at pH 7.0 (Fig. 8). The percentage degradation clearly decreased as the p-NP concentration increased. It was previously observed that, as the p-NP concentration increases, the interaction between HO* species and p-NP becomes weaker and penetration of visible light is decreased. Only a small number of HO* species are then formed, which contributes to the lowered photodegradation efficiency of the catalysts. This behavior can be explained by the fact that photocatalytic reactions generally occur on the surface of the catalyst and the morphology and surface state of semiconductor nanostructures have crucial roles in their photocatalytic activity.
Many requirements, including proper band alignment and electronic coupling between the metal nanoparticles and the semiconductor, should be satised for electron injection between the metal nanoparticles and the semiconductors. Plasmon-induced resonance energy transfer, however, is not constrained by such requirements. 41 It is widely acknowledged that the phenomenon of localized surface plasmon resonance (LSPR) occurs because of interaction between localized electrons of the metal particles (Pd, in this study) with incident visible light. 42 Because of LSPR, Pd-deposited semiconductor samples absorbed more light (DR UV-vis spectral data). It appears that the LSPR phenomenon is dependent on the size, local environment and morphology of the particles. 43 Leong et al. 29 reported that small size Pd nanoparticles (20-30 nm) showed LSPR in the visible region ($500 nm), which is in agreement with the DR UV-vis spectral results for our synthesized samples. Plasmon-induced resonance energy transfer depends on near-eld enhancement and overlap between the LSPR absorption and the optical transitions in semiconductors. This energy transfer process also localizes the generation of electron-hole pairs mainly near to the metal nanoparticles, which signicantly decreases the carrier recombination rate. 41 Furthermore, this energy transfer only occurs if the plasmonic metal and semiconductor are in direct contact. In our study, TEM analysis showed that the Pd and semiconductors are in intimate contact with each other. The electrons normally become excited aer the semiconductor is exposed to visible light irradiation (Scheme 1). Electrons in the Pd nanoparticles that are excited by LSPR are transferred to the conduction band of the semiconductor, leaving holes in the metal nanoparticles. 44 When the Pd nanoparticles have close interactions with the semiconductor, an internal electric eld close to the metal-semiconductor interface, known as a Schottky barrier, is generated. 45 The internal electric eld drives the excited electrons and holes in different directions and the Schottky barrier acts as an electron trap to prevent the backow of electrons to the semiconductor. The Pd nanoparticles serve as a pool for electrons, thereby catalyzing the reduction of O 2 . As a result, larger numbers of *OH species are generated and become available on the semiconductor surface where they participate in the degradation of p-NP. Zhang et al. proposed that electron collisions at the metal nanoparticle-semiconductor interface excite electrons from the valence band of the semiconductor to the conduction band, thus generating electron-hole pairs. This group also described the process as an LSPR-powered bandgap breaking effect. 46 Clavero observed that photocatalysts with an LSPR-powered bandgap effect absorbed light very efficiently over a wide spectral range and that generating electron-hole pairs oen leads to a low carrier recombination rate. These types of photocatalyst thus have the greatest potential for improving the intrinsic limitations of conventional semiconductor photocatalysts. 47 It was expected that Pd-W-NF would be a more effective photocatalyst than Pd-Ti-NF or Pd-Ce-NF because the smaller band gap would allow better photo absorption in the visible light spectrum by WO 3 , compared with TiO 2 and CeO 2 . 48 However, WO 3 has a lower light energy conversion efficiency than TiO 2 or CeO 2 because the reduction potential is relatively low. 49 The surface area of the sample is one of the factors that affects photoactivity. As shown by the N 2 physisorption results, the surface area of Pd-Ti-NP ($74 m 2 g À1 ) is much higher than those of Pd-Ti-NF, Pd-Ce-NF and Pd-W-NF (50 m 2 g À1 , 35 m 2 g À1 and 10 m 2 g À1 , respectively). Although higher surface area is known to result in a larger capacity to adsorb p-NP, it is important to note that the photocatalytic activity of Pd-Ti-NP was lower than that of Pd-Ce-NF or Pd-Ti-NF, despite its larger surface area. Tuning of photocatalytic activity for a specic semiconductor depends on both structure and morphology and this has been used as a general versatile rationale to design appropriate photocatalysts for different types of reaction. 50 The Pd-deposited CeO 2 nanobrous sample had an increased surface-area-to-volume ratio compared with nano-bers of TiO 2 and WO 3 or nanoparticles and nanotubes of TiO 2 . This is mainly due to transformation of the nanobrous CeO 2 morphology to a unique 'mat-like' morphology (TEM analysis). It appears that this morphology of CeO 2 reduces agglomeration of the Pd nanoparticles. The theoretical photocurrent density of the samples was investigated using theoretical calculations and  DR UV-vis spectral analysis (ESI †) since the samples displayed absorption proles at different absorption edges. The photocurrent density is most strongly inuenced by the lightharvesting ability and photo charge separation efficiency of the catalyst. The theoretical photocurrent density of Pd-Ce-NF (0.658 mA cm À2 ) is higher than those of Pd-W-NF (0.559 mA cm À2 ), Pd-Ti-NF (0.478 mA cm À2 ), Pd-Ti-NT (0.425 mA cm À2 ) and Pd-Ti-NP (0.401 mA cm À2 ), demonstrating the benecial role of the unique morphology of CeO 2 and the size of the Pd particles in the photoactivity.

Reuse of Pd-CeO 2 -NF catalyst
Reusability is an important requirement for photocatalysts because use of catalysts for long periods of time leads to signicant loss of activity due to agglomeration of particles and photo corrosion. To investigate the reusability of the synthesized samples, the Pd-Ce-NF sample was recycled ve times. The degradation efficiency of the sample was 93%, even aer ve cycles, showing that Pd-Ce-NF could be used as an effective industrial photo catalyst to degrade p-NP. The slight reduction (3%) in degradation efficiency aer ve cycles could be due to loss of catalyst incurred during the ltration and drying processes.

Conclusions
A simple deposition-precipitation method was used to synthesize different Pd-deposited semiconductors (TiO 2 , CeO 2 and WO 3 ), with different morphologies (nanoparticles, nanotubes and nanobers). The obtained materials were used as catalysts for photocatalytic degradation of p-NP under visible light irradiation. Among the Pd-deposited TiO 2 samples, TiO 2 nanobers showed better photocatalytic activity than Pd-deposited TiO 2 nanoparticles or nanotubes. To explore the advantages of nanober morphology, Pd-deposited CeO 2 and WO 3 nanober samples were also tested for photocatalytic activity. Pddeposited CeO 2 nanobers showed the best performance among all the Pd-deposited semiconductor samples. Elemental analysis, XRD, SEM, TEM, DR UV-vis, N 2 physisorption and XPS techniques were used to investigate the role of structural properties on photocatalytic performance. XRD and TEM showed that Pd deposition did not alter the crystal structure of TiO 2 , WO 3 and CeO 2 and that Pd was well dispersed on the semiconductor surfaces, with particle sizes < 20 nm. However, Pd deposition resulted in a change in morphology of the CeO 2 nanobers to a unique 'mat-like' structure. The presence of small Pd nanoparticles (4 nm) and a high photocurrent density in Pd-Ce-NF is mainly responsible for the high photocatalytic activity. The Pd-deposited samples can be reused at least for ve cycles without signicant loss of photocatalytic activity.

Conflicts of interest
There are no conicts to declare.