Chlorinated phenoxyacetic acid (denoted MCPA-Na) is one of the oldest and most readily available herbicides in the world, and is widely used as a weed killer on cereal crops, pasture, and orchards. However, MCPA-Na is considered potentially dangerous to both animals and humans and is a well-known endocrine disruptor [1]. The biodegradability of MCPA-Na is extremely low and it has been detected as a major contaminant in effluents released into both subterranean and superficial water bodies [2].

Semiconductor-assisted photocatalysis can complement conventional approaches to degrade or transform hazardous chemical waste [3, 4]. Zinc oxide (ZnO) is one of the most extensively studied semiconductor photocatalysts because it is environmentally friendly, cheap, and has a large exciton binding energy [4-7]. Various morphologies of ZnO including nanorods [8, 9], nanotubes [10], hollow nanospheres [11], nanoplates [12], and nanoflower-like structures [13, 14] have been prepared. In particular, the "one-dimensional" tubular morphology of ZnO has attracted interest because of its special hollow structure and large surface area, which are crucial for optimizing the performance of dye-sensitized solar cells, sensors, and hydrogen devices [15-18]. Moreover, the "one-dimensional" morphology can facilitate transport of charge carriers and minimize the loss of charge carriers at grain boundaries [19, 20].

Most reported ZnO nanotubes are wide-band-gap semiconductors and can only absorb ultraviolet (UV) light. Therefore, it is necessary to functionalize ZnO nanotubes so that they absorb the visible component of sunlight. Coupling a narrow-band-gap semiconductor with a wide-band-gap semiconductor is a useful approach to capture visible light because the coupling of different semiconductor oxides can narrow the band gap of the resulting composite. Moreover, higher photocatalytic activity may be achieved for semiconductor composites because electron–hole pairs maybe separated efficiently under irradiation. It has been reported that the construction of heterostructures is an effective strategy to improve photocatalytic performance [21-23].

Tungsten trioxide (WO₃) is a promising photocatalyst that has an appropriate band gap of 2.8 eV to enable absorption in the visible region, stable physicochemical properties, and high resistance to photocorrosion [24, 25]. It has been reported that the coupling of ZnO nanoparticles with WO₃ resulted in composites with higher photocatalytic activity than those of pristine ZnO and pristine WO₃[26-30]. However, the photocatalytic properties of ZnO nanotubes loaded with WO₃ have not been reported yet. Therefore, we decided to couple WO₃ nanoparticles with ZnO nanotube arrays in the present work. ZnO nanotube arrays are first synthesized by electrodeposition. Composites with different contents of WO₃ are obtained by the addition of ammonium metatungstate hydrate ((NH₄)₆H₂W₁₂O₄₀· XH₂O) to the prepared ZnO nanotube arrays and annealing at 450 ℃ in air for 2 h. The photocatalytic properties of the prepared ZnO–WO₃ nanotube arrays with different contents of WO₃ nanoparticles are evaluated by their ability to degrade MCPA-Na. Based upon the results of activity evaluation and characterization, a photocatalytic mechanism for the composites is proposed.

Electrodeposition of ZnO nanotubes was carried out in a conventional three-electrode cell using a Pt electrode with an area of about 3.0 cm² as the auxiliary electrode. The working electrode was indium tin oxide (ITO)-coated glass with a sheet resistance of 6–8 Ω/cm². A saturated Ag/AgCl electrode was used as the reference electrode. The ITO glass substrate with dimensions of 3 × 4 cm was cleaned ultrasonically in distilled water, ethanol, acetone, and 6 vol% hydrochloric acid in sequence for 5 min each, and then rinsed in distilled water before electrodeposition. ZnO crystal seeds were grown on the ITO substrate in solution containing 0.02 mol/L zinc nitrate (Zn(NO₃)₂), 0.013 mol/L ammonium acetate (CH₃CO₂NH₄), and 0.01 mol/L hexamethylenetetramine (C₆H₁₂N₄) under a cathodic voltage of –1.8 V (vs. Ag/AgCl) for 300 s at 90 ℃. The cathodic voltage was switched to -1.2 V to prepare the ZnO nanorod arrays. After electrodeposition for 60 min at this applied potential, a uniform white film was obtained. The film was washed with distilled water and then used as the working electrode in 0.07 mol/L C₂H₄(NH₂)₂ at −0.2 V for 60 min at 70 ℃ to prepare pristine ZnO nanotube arrays.

To prepare ZnO–WO₃ nanotube arrays, the required amount of (NH₄)₆H₂W₁₂O₄₀·XH₂O was first dispersed in acetone. The solution was added dropwise onto the prepared ZnO nanotube surface. The substrate was annealed in a muffle furnace at 450 ℃ in air for 2 h. The composite fabrication process is illustrated in Fig. 1. All reagents used in this study were of analytical grade and used directly without any purification. The obtained ZnO–WO₃ samples are denoted as ZnO–WO₃ (2%, 3%, or 5%) according to their W/Zn molar ratio of 2%, 3%, and 5%, respectively, in the precursor solution. For comparison, annealed pristine ZnO nanotube arrays were prepared under the same conditions without the addition of (NH₄)₆H₂W₁₂O₄₀·XH₂O. Pristine WO₃ powder was also prepared by calcination of (NH₄)₆H₂W₁₂O₄₀·XH₂O at 450 ℃ in air for 2 h in a muffle furnace.

The crystal structures of the prepared ZnO and ZnO–WO₃ nanotube arrays were characterized by X-ray diffraction (XRD) using a PANalytical B.V. MPDDY2094 X-ray diffractometer (Almelo, the Netherlands) with Cu K_α radiation (λ = 1.5406 ). Scanning electron microscopy (SEM) images and quantitative standard microanalyses were obtained using a Zeiss ultra plus field-emission scanning electron microscope (Germany) (FE-SEM) equipped with energy-dispersive X-ray spectrometer (EDS) analysis apparatus. High-resolution transmission electron microscopy (HRTEM) images were captured using a Tecnai G² 20 microscope (FEI, USA) at an accelerating voltage of 200 kV. UV-vis diffuse reflectance spectroscopy (DRS) was recorded on a Perkin Elmer Lambda 35 UV-vis spectrophotometer (Wellesley, USA) using BaSO₄ as a reference over the scan range of 200–800 nm. Photoluminescence (PL) measurements were performed on a Hitachi F7000 fluorescence spectrophotometer (Japan) with an excitation wavelength of 325 nm at room temperature. Mott–Schottky (MS) measurements were performed in 0.5 mol/L Na₂SO₄ solution using an electrochemical workstation (CHI660E, China) with a three-electrode system at room temperature. The obtained ZnO, WO₃, and ZnO–WO₃ nanotube array samples served as the working electrode. A Pt plate and saturated Ag/AgCl were used as the counter and reference electrodes, respectively. Fourier transform infrared (FTIR) spectra were measured using a Thermo Scientific Nicolet IS10 spectrometer (USA) over the scan range from 400 to 4000 cm^‒1 using KBr disks. X-ray photoelectron spectroscopy (XPS) was performed on a Thermo Scientific ESCALAB 250Xi spectrometer (USA) with an Al K_α excitation source.

The photocatalytic activities of the samples were evaluated by the degradation of MCPA-Na aqueous solution under simulated sunlight irradiation. In each experiment, the sample was put into water (100 mL) containing MCPA-Na (20 mg). A 300-W xenon lamp with a power density of 20.15 mW/cm² (Beijing Science and Technology Co, Ltd., Park Philae) was set inside a cylindrical reactor and surrounded by a circulating water jacket to cool the lamp and minimize infrared radiation. UV–Vis spectroscopy was used to determine the solution concentration of residual MCPA-Na at room temperature (TU-1900, Beijing Purkinje General Instrument Co, Ltd.). The absorbance accuracy was ±0.002 Abs (0–0.5 Abs) and ±0.004 Abs (0.5–1.0 Abs). The total organic carbon (TOC) remaining in the solution after the photodegradation process was measured using a Jena Multi C/N 3100 analyzer.

Fig. 2 shows the microstructure of the obtained ZnO and ZnO–WO₃ (3%) nanotube arrays. The pristine ZnO nanotubes (Fig. 2(a)) possess a hollow hexagonal structure with an internal diameter of 400 nm and thickness of 80 nm. The pristine ZnO nanotubes have a smooth surface. The ZnO nanotubes after loading WO₃ (Fig. 2(b)) have a smaller internal diameter than that of the pristine ZnO nanotubes. The elemental maps of ZnO–WO₃ (3%) recorded by EDS are shown in Fig. 2(c–e). These elements are all distributed uniformly on the surface. Element contents of the samples are listed in Table 1. Only Zn and O were detected in the pristine ZnO nanotubes. However, the atomic ratio of Zn to O was not exactly equal to 1; the atomic ratio of Zn was slightly higher than that of O. Table 1 reveals that the W contents in ZnO–WO₃ (2%), ZnO–WO₃ (3%), and ZnO–WO₃ (5%) are 12 at%, 11 at%, and 20 at%, respectively. Higher W contents were found in the ZnO–WO₃ samples than present in the precursor solution, indicating that most of the WO₃ in solution was loaded on the surface of the ZnO nanotubes. Moreover, the W content of ZnO–WO₃ (3%) was slightly lower than that detected for ZnO–WO₃ (2%), indicating more WO₃ was inside the ZnO nanotubes in the ZnO–WO₃ (3%) sample than in ZnO–WO₃ (2%).

Fig. 2(f) displays a typical TEM image of the ZnO–WO₃ (3%) sample, which clearly indicates that it has a tubular structure with some substance in the tube and on its surface. Fig. 2(g) shows a HRTEM image of the same sample, in which two single-crystalline phases can distinguished. The insets are the fast Fourier transformation (FFT) of the marked areas. The FFT spots corresponded to the (1 0 3) crystalline plane of ZnO with an interlayer spacing of around 0.15 nm as well as the (0 1 1) and (2 0 2) crystalline planes of WO₃ with interlayer spacings of around 0.54 and 0.26 nm, respectively.

Fig. 3 shows the XRD patterns of the pristine ZnO, WO₃, and ZnO–WO₃ (3%) samples. The peaks at 34.422° (002), 36.253° (101), 47.539° (102), 56.603° (110), 62.864° (103), 67.963° (112), 69.100° (201), and 72.562° (004) were indexed to ZnO with a hexagonal wurtzite structure (JCPDS Card No. 00-036-1451). The peaks observed at 34.730° (020), 23.172° (002), 24.104° (110), 28.341° (102), 36.292° (221), 50.726° (114), and 56.785° (412) (JCPDS Card No. 01-083-0950 and 01-087-2401) were consistent with monoclinic WO₃, indicating that WO₃ can be obtained by annealing (NH₄)₆H₂W₁₂O₄₀·XH₂O. No obvious diffraction peaks from WO₃ were observed for the ZnO–WO₃ (3%) sample, which is probably because of its low content of WO₃. Because the samples were fabricated on ITO glass substrates, In₂O₃ (JCPDS Card No. 03-065-3170), SnO₂ (JCPDS Card No. 00-046-1242), and SiO₂ (JCPDS Card No. 00-046-1242) were also detected for these samples.

XPS measurements were conducted to obtain insight into the composition of the ZnO–WO₃ (3%) sample; the results are shown in Fig. 4. The survey spectrum (Fig. 4(a)) revealed the presence of W, O, and Zn. As illustrated in Fig. 4(b), the W 4f spectrum contains typical W⁶⁺ peaks at binding energies of 35.4 and 37.5 eV, which are very close to the values reported for WO₃ [31]. In Fig. 4(c), the Zn 2p peaks at 1022.0 and 1045.2 eV indicated that Zn existed as Zn²⁺, which was consistent with the values reported for ZnO [10]. Figure 4(d) shows that the O 1s peak can be deconvoluted into two individual peaks at the binding energies of 530.3 and 531.6 eV, which correspond to the lattice oxygen of ZnO and hydroxide (OH^–), respectively [10].

Fig. 5 presents the FTIR spectra of pristine ZnO and ZnO–WO₃ samples. The peak at 491 cm^–1 is ascribed to the stretching of Zn–O bonds [32]. The peak at 1636 cm^–1 is assigned to aromatic and –OH bending vibrations, and the peak at 3428 cm^–1 corresponds to the stretching of surface OH^- groups [33]. Compared with that of pristine ZnO, the FTIR spectra of the ZnO–WO₃ samples contained new peaks at around 870 and 989 cm^–1, which belonged to W–O–W stretching, bending, or lattice modes [34]. Moreover, the peak at 3428 cm^–1 for the ZnO–WO₃ composites is stronger than that for pristine ZnO, indicating more OH^- groups were present in the ZnO–WO₃ composites than in ZnO. It has been reported that a higher amount of OH^- groups in the catalyst may result in higher photocatalytic activity because OH^- can capture photogenerated holes and transform into reactive hydroxyl radicals [31].

UV-vis DRS is an effective technique to determine the light absorption ability of a semiconductor at different wavelengths. Fig. 6(a) shows the UV-vis DRS obtained for all the samples. Pristine ZnO exhibited strong light absorption from 200 to 420 nm and little absorption in the visible light range. In comparison, the absorption spectra of the ZnO–WO₃ (3%) and ZnO–WO₃ (5%) composites displayed an apparent, red shift of the band edge, which may be caused by the aggregation of WO₃ particles [35]. Fig. 6(a) reveals that the absorption intensity in the visible region increases with the content of WO₃ in the ZnO–WO₃ samples, which indicates that the coupling of WO₃ with ZnO can enhance the absorption of ZnO in the visible-light region. This increased intensity may be attributed to the interaction between W⁶⁺ and crystal lattice oxygen resulting in an unsaturated bond and surface vacancy on ZnO [28].

The optical bandgap of the composites is calculated using the fundamental absorption, which corresponds to electron excitation from the valance band to conduction band. The absorption coefficient (α) is calculated using the Eq. (1):

(1)

where A is the absorbance, and d is the film thickness. The absorption coefficient (α) and the incident photon energy (hν) are related by Eq. (2):

(2)

where h, ν, C and E_g are plank's constant, light frequency, proportionality constant, and the optical band gap, respectively. From the plot of (αhν)² vs. hν (Fig. 6(b)), the band gap energy is obtained by extrapolating the linear portion of the curve to the hν axis. The pristine ZnO and WO₃ exhibited the E_g of 3.15 and 2.8 eV, respectively, while the ZnO–WO₃ (2%, 3%, and 5%) composites displayed narrower E_g of about 2.96, 2.85, and 3.10 eV, respectively. E_g of the ZnO–WO₃samples initially decreased and then increased with the WO₃ content. ZnO–WO₃ (3%) possesses the narrowest E_g, which indicates that it can absorb more visible light than the other samples.

PL spectroscopy is an effective technique to evaluate the efficiency of charge carrier trapping, transfer, and separation in semiconductor materials and surface defects of samples [36-39]. Fig. 7 shows the room-temperature PL spectra of the ITO-supported ZnO and ZnO-WO₃ nanotube arrays obtained with an excitation wavelength of 325 nm. The emission at 380 nm is caused by the recombination of photogenerated electrons and holes [28, 40, 41]. High PL intensity at 380 nm is observed for the pristine ZnO nanotubes sample, indicating the rapid recombination of electrons and holes. The intensities of this emission peak for the ZnO–WO₃ composites are dramatically decreased compared with that of the pristine ZnO nanotubes. Very weak emission intensity is observed for the ZnO–WO₃ (2% and 3%) photocatalysts, which implies that a large number of photogenerated electrons and holes were trapped and their recombination was prohibited for these samples. Moreover, there is an extra emission peak at 600 nm for the ZnO–WO₃ (2% and 3%) composites. It has been reported that the visible emission peaks at around 420–620 nm observed for ZnO are related to defects and doubly charged oxygen vacancy states in the nanostructures [28, 42, 43]. The stronger the intensity of this luminescence, the more intrinsic defects (oxygen and zinc vacancies or interstitials) present in the sample [44]. As shown in Fig. 7, the visible emission peak at 600 nm is much more intense for the ZnO–WO₃ (2% and 3%) samples compared to the case for ZnO–WO₃ (5%) and pristine ZnO samples, which indicates there are more defects in the former two samples than in the latter two samples. It has been reported that surface defects can improve the photocatalytic performance of metal oxide semiconductors because defects can act as charge carrier traps, where the charges are transferred to adsorbed species, preventing the recombination of photogenerated electrons and holes [42].

Fig. 8(a) depicts the relationship between the MCPA-Na degradation efficiency and irradiation time of the prepared ZnO nanotubes and ZnO–WO₃ composites. After 200 min of simulated sunlight irradiation, about 58% of MCPA-Na was degraded by pristine ZnO nanotubes, while about 92%, 98.5%, and 49% of MCPA-Na was degraded by ZnO–WO₃ (2%), ZnO–WO₃ (3%), and ZnO–WO₃ (5%), respectively. Increasing the content of WO₃ in ZnO from 2% to 3% effectively enhanced the photodegradation efficiency of the ZnO nanotubes. However, further increasing the WO₃ content had an adverse effect. The TOC contents of MCPA-Na solution were also measured, as shown in Fig. 8(b). The TOC content decreased to almost zero after 200 min of photocatalytic degradation, which means the herbicide pollutant was almost totally removed.

The pseudo-first-order model expressed by the equation −ln(C/C₀) = kt was used to quantitatively explain the reaction kinetics of MCPA-Na degradation. Here, C₀ and C are the initial pollutant concentration and pollutant concentration at any time, respectively, k is the pseudo-first-order rate constant, and t is the reaction time. According to this equation, k can be obtained from a linear plot of −ln(C/C₀) against t. The kinetic plots for photocatalytic degradation of MCPA-Na over composites with different WO₃ loading are presented in Fig. 8(c) and the calculated k values are shown in Fig. 8(d). The highest k was 0.0148 min^–1 for ZnO–WO₃ (3%), and it was 0.0145, 0.003, and 0.007 min^–1 for ZnO–WO₃ (2%), ZnO–WO₃ (5%), and ZnO, respectively.

Cycling stability is an important factor to evaluate the performance of a photocatalyst. Herein, the cycling stability of the prepared ZnO and ZnO–WO₃ (3%) thin films was evaluated. As shown in Fig. 8(e), the pristine ZnO sample exhibited poor cycling stability. The ZnO photocatalyst achieved 58% MCPA-Na degradation in the first run, which decreased markedly to 7% and 6% in the second and third runs, respectively. This poor cycling stability of the ZnO photocatalyst is mainly caused by photoinduced corrosion [45]. In contrast, the ZnO–WO₃ (3%) composite degraded 98%, 63%, and 41% of MCPA-Na in the first, second, and third runs, respectively. Thus, photocatalytic stability was improved by the coupling of WO₃ with ZnO. However, it still needs to be further improved for practical use of this composite catalyst.

The electronic properties of a semiconductor can be investigated by MS analysis of its capacitance measurements. The type of semiconductor can be determined from the plot of 1/C_SC² versus E, where C_SC² is the space–charge capacitance and E is the scanning potential. A negative slope indicates a p-type semiconductor, while a positive slope indicates an n-type semiconductor. Moreover, the intercept of a linear plot at 1/C_SC² = 0 gives the flat band potential (E_FB). Fig. 9(a) shows the MS plots for the ZnO–WO₃ (3%) sample at different frequencies. The derived E_FB were almost the same at different frequencies. Therefore, the capacitance measurements for MS analysis of all samples were collected at 1000 Hz; the results are presented in Fig. 9(b). The slopes of the MS plots are positive for all the samples, indicating their n-type nature. To estimate E_FB, the linear part of each plot was extrapolated to 1/C_SC² = 0. The E_FB values for ZnO, ZnO–WO₃ (2%), ZnO–WO₃ (3%), ZnO–WO₃ (5%), and WO₃ samples estimated from their MS plots are -0.23, -0.25, -0.46, -0.19, and -0.52 V (vs. Ag/AgCl), respectively. Because the Fermi level of NHE at 25 ℃ is -4.5 eV with respect to the vacuum level [46], the energy positions of band edges on the electrochemical scale can be converted from the values on the absolute vacuum energy scale using Eq. (3):

(3)

The calculated E_FB on the absolute vacuum energy scale for ZnO, ZnO–WO₃ (2%), ZnO–WO₃ (3%), ZnO–WO₃ (5%), and WO₃ are -4.47, -4.45, -4.24, -4.50, and -4.18 eV, respectively. Because E_FB lies just below the conduction band (CB) edge energies (E_c) for n-type semiconductors and just above the valence band (VB) edge energies (E_v) for p-type semiconductors, it could be assumed that E_C is equal to E_FB for n-type semiconductors. Combined with the results of UV-vis DRS, the E_v values of ZnO and the ZnO–WO₃ composites were calculated using Eq. (4); the results are summarized in Table 2.

(4)

Based on the positions of the CB, VB, and E_g of WO₃ and ZnO, we speculate that the possible energy storage mechanism of the ZnO–WO₃ composites is as follows (Fig. 9(c)). When WO₃ and ZnO are irradiated with solar light, electrons are excited to the CB, leaving holes in the VB. Because both the bottom of the CB and top of the VB of ZnO are lower than those of WO₃, ZnO can act as a sink for the photogenerated electrons in the coupled oxides. In addition, the photogenerated holes in ZnO might be trapped within the WO₃ particle, making charge separation more efficient. Therefore, the recombination of the photogenerated carriers is suppressed, leading to an increase in photo-oxidation efficiency. However, if WO₃ content of the composite is too high, the dispersion of WO₃ becomes poor and the aggregated WO₃ can behave as recombination centers, resulting in a decrease in photocatalytic activity.

WO₃ was successfully loaded onto ZnO nanotube arrays by chemical bath deposition. The addition of WO₃ changed the band structure of the ZnO nanotube arrays, allowing the recombination of photogenerated electrons and holes to be suppressed at a proper loading content of WO₃. About 98.5% of MCPA-Na in solution was degraded by ZnO–WO₃ (3%) under simulated sunlight irradiation, and the degradation rate of MCPA-Na by ZnO–WO₃ (3%) was about twice that of the pristine ZnO nanotube array sample. The high relative content of surface defects on the ZnO–WO₃ composite increased its photocatalytic activity in MCPA-Na degradation.