Enhanced Photocatalytic CO2 Reduction in Defect-Engineered Z-Scheme WO3-x /g-C3N4 Heterostructures.

Oxygen vacancy-modified WO3-x nanorods composited with g-C3N4 have been synthesized via the chemisorption method. The crystalline structure, morphology, composition, band structure, and charge separation mechanism for WO3-x /g-C3N4 heterostructures are studied in detail. The g-C3N4 nanosheets are attached on the surface of WO3-x nanorods. The Z-scheme separation is confirmed by the analysis of generated hydroxyl radicals. The electrons in the lowest unoccupied molecular orbital of g-C3N4 and the holes in the valence band of WO3 can participate in the photocatalytic reaction to reduce CO2 into CO. New energy levels of oxygen vacancies are formed in the band gap of WO3, further extending the visible-light response, separating the charge carriers in Z-scheme and prolonging the lifetime of electrons. Therefore, the WO3-x /g-C3N4 heterostructures exhibit much higher photocatalytic activity than the pristine g-C3N4.

Introduction

Photocatalytic CO2 reduction is a promising way for clean, low-cost, and environmentally friendly conversion of CO2 into fuels by utilizing the solar energy.[1] Graphitic carbon nitride (g-C3N4) has gained great attention recently as an easy-to-obtain and visible-light-responsive nonmetal semiconductor photocatalyst that shows great stability and nontoxicity.[2−6] WO3 is also considered as a high-performance visible photocatalyst with a band gap of 2.6–2.9 eV.[6−8] During the last decades, compositing WO3 with g-C3N4 has been investigated for efficient photocatalytic organic degradation, water splitting, and CO2 reduction. Chen et al. investigated the separation mechanisms of charge carriers for C3N4/WO3 composites.[9] Huang et al. synthesized WO3/g-C3N4 composites with effective photocatalytic activity on the degradation of organic pollutants.[10] Liang et al. reported a three-dimensional WO3 superstructure coupled with g-C3N4, which exhibits an enhanced photocatalytic performance.[11] Moreover, WO3 composited with g-C3N4 Z-scheme photocatalyst has been reported by several researchers for enhanced photocatalytic activity.[11−14] Chai et al. synthesized WO3 nanoplate anchored on g-C3N4 Z-scheme photocatalyst with enhanced photocatalytic activity.[12] Lu et al. constructed g-C3N4/RGO/WO3 Z-scheme photocatalyst by using graphene oxide as the electron mediator.[15] However, the photocatalytic performance is still limited and challenge remains to further improve the photocatalytic performance of C3N4/WO3 heterostructures.

Recently, defect engineering in semiconductors has been widely considered for improving the photocatalytic performance.[16−24] The introduction of oxygen vacancy in WO3 can narrow the band gap and separate the photoinduced carriers, resulting in enhanced photocatalytic activities. Liu et al. found that the oxygen vacancies in WO3 can promote the separation of charge carriers and enhance the photocatalytic performance.[19] Li et al. prepared defect-modified WO3 with enhanced visible photocatalytic and electrochemical performance.[18] Meng et al. induced oxygen vacancies on specific surface of WO3 with enhanced visible-light photocatalytic activity.[23] However, most of the reports have focused on the influence of oxygen vacancies on the WO3 photocatalyst itself. The effect of oxygen vacancies on the interface of heterostructures, such as visible-light response, separation behaviors of charge carriers, and lifetime of photogenerated electrons, is rarely reported.

Herein, defect-engineered WO3– nanorods are prepared by vacuum thermal treatment and WO3–/g-C3N4 heterostructures are synthesized through a chemisorption method. The morphology structure, band structure, and separation of electrons and holes are investigated systematically. These WO3–/g-C3N4 heterostructures exhibit a significantly improved catalytic activity for reduction of CO2 into CO compared to pristine g-C3N4, WO3, and WO3/g-C3N4. The introduced oxygen vacancies would create defect energy levels below the conduction band of WO3 and facilitate the separation of charge carriers in Z-scheme pathway. This work provides insights into practical applications of Z-scheme heterostructures toward efficient photocatalytic CO2 reduction.

Results and Discussion

Figure shows the X-ray diffraction (XRD) patterns of pure WO3, WO3–, and WC3 (the abbreviations of WO3–/g-C3N4, and the nominal weight ratio of g-C3N4 to WO3– is 3%) samples and g-C3N4. Two diffraction peaks at 13.1 and 27.4° are found for g-C3N4 samples, corresponding to the (001) and (002) planes (JCPDS No. 87-1526). For pure WO3, only two obvious diffraction peaks at 23.0 and 46.9° are observed, ascribed to the (001) and (002) planes of WO3, respectively (JCPDS No. 20-1324). Compared to the standard diffraction data of WO3, no other diffraction peaks are observed. The peak of (001) plane is significantly sharp, suggesting that the WO3 crystals grow preferentially along the c axis direction. The XRD pattern of WO3– is almost the same as that of pure WO3, suggesting that the vacuum thermal treatment does not change the crystal structure of WO3– samples. After the compositing process, the XRD patterns of WC3 are almost the same as those of pure WO3 and WO3– samples. No additional peaks related to g-C3N4 are detected, which is probably due to the relative low weight ratio of g-C3N4.

Figure 1

XRD patterns of pure WO3, WO3–, and WC3 samples and g-C3N4.

The morphologies of the as-prepared nanostructures (WC3 (the abbreviations of WO3/g-C3N4, and the nominal weight ratio of g-C3N4 to WO3 is 3%) and WC3) are shown in Figure . As shown in the transmission electron microscopy (TEM) image in Figure a, the WC3 samples are composed of crystallized nanorods with a uniform diameter of approximately 10 nm and length of 50–100 nm. It is observed from the high-resolution transmission electron microscopy (HRTEM) image in Figure b that some WO3 nanorods are composited with g-C3N4 (TEM of pristine g-C3N4 shown in Figure S1). In addition, the lattice fringe is 0.39 nm corresponding to the (001) plane of WO3 (JCPDS No. 20-1324), which suggests that the growth of the WO3 nanorods is along the [001] direction. After the vacuum thermal treatment (Figure c,d), the morphology of the nanorods remains almost unchanged and the WO3– nanorods are still composited with g-C3N4. These TEM results are in good agreement with the XRD results.

Figure 2

TEM and HRTEM images of WC3 (a, b) and WC3 (c, d) samples.

X-ray photoelectron spectroscopy (XPS) images are provided in Figure to investigate the chemical states of W, O, C, and N atoms. Figure a shows the XPS C 1s spectra of the pure WO3, WC3, and g-C3N4 samples. For all samples, the peaks centered at 284.8 eV are ascribed to the surface adsorbed C atoms that can be used for calibration.[25] For the g-C3N4 sample, in addition to the C 1s peak at 284.8 eV, a new peak at approximately 288.1 eV is detected, corresponding to the C atoms in −N–C–N– coordination for g-C3N4. For WC3 samples, a weak peak centered at approximately 288.6 eV is also found, suggesting the existence of g-C3N4. For the XPS N 1s spectra (Figure b), the major peaks at approximately 398.7 eV for g-C3N4 and 399.1 eV for WC3 are ascribed to the sp2 bonding N atoms in −C=N–C– structures.[26] The small difference for the C 1s and N 1s spectra of WC3 and g-C3N4, respectively, is caused by the heterojunction between g-C3N4 and WO3. The chemical interaction between g-C3N4 and WO3–, such as the Mo–N bonding, improves the electron transfer process in the WC3 heterojunction. For the XPS W 4f spectrum (Figure c), the double peaks at approximately 35.9 and 38.1 eV are attributed to the W 4f7/2 and W 4f5/2 of W6+ ions in WO3, respectively.[6,7] Moreover, two different valence states of W ions can be observed for WC3 samples. In addition to the peaks of W6+ ions, the new double peak at 35.5 and 37.7 eV can be observed, ascribed to the W 4f7/2 and W 4f5/2 of W5+ ions, respectively.[19,23] The existence of W5+ is usually considered as a direct evidence for the existence of oxygen vacancies in the WO3 system.[18,19] The XPS O 1s spectra of WO3 and WC3 are deconvoluted into two peaks, as shown in Figure d. One peak at 530.9 eV is ascribed to the lattice O in WO3, and the other peak at 531.9 eV is ascribed to the surface adsorbed O species. Compared to WO3, the XPS O 1s peak for the WC3 sample shifts to a lower binding energy by 0.1 eV, which is due to the change of chemical states from W6+ to W5+.[27] This XPS analysis confirms that the oxygen vacancies are successfully introduced into WO3 via the vacuum thermal treatment and the WO3–/g-C3N4 heterostructure is formed. Moreover, the peak intensity of the electron spin resonance (ESR) signal at g = 2.002 increases in WC3, compared to that in WC3 (Figure S2), which also confirms that oxygen vacancies are introduced into WO3 nanorods after the vacuum thermal treatment.

Figure 3

XPS C 1s (a), N 1s (b), W 4f (c), and O 1s (d) spectra of WO3, WC3, and g-C3N4 samples.

To evaluate the photocatalytic performance of the as-prepared photocatalyst, the photoreduction of CO2 with H2O into CO is shown in Figure . Both WO3 and g-C3N4 samples exhibit poor photocatalytic activities, with only 1.11 and 0.92 μmol of CO detected after 8 h irradiation, respectively, which are almost the same as the blank experiment (0.49 μmol). The WO3– samples also exhibit a limited photocatalytic activity (1.39 μmol). A series of WCX samples all exhibit enhanced photocatalytic activities. The WC3 sample exhibits the best photocatalytic performance among all samples (Figure S3). There is 5.17 μmol of CO produced after 8 h irradiation. The photocatalytic performance is further enhanced for WC3 samples, and 6.64 μmol CO is detected for 8 h irradiation. These results suggest that the introduction of oxygen vacancies into the interface between g-C3N4 and WO3 is an effective method to improve the photocatalytic activity.

Figure 4

Photocatalytic activity for reduction of CO2 into CO of pure g-C3N4, WO3, WO3–, WC3, and WC3 samples under a xenon lamp for 8 h.

Figure a shows the photoluminescence (PL) spectra of g-C3N4, WC3, and WC3 samples to investigate the influence of oxygen vacancies at the interface of WO3– and g-C3N4 on the behaviors of photogenerated charge carriers. The g-C3N4 sample exhibits the highest emission intensity among all samples, indicating the fast recombination of photogenerated electrons and holes. After the g-C3N4 is composited with WO3 nanorods (WC3 and WC3), the PL intensity is quenched efficiently, suggesting an efficient separation of charge carriers at the interface of g-C3N4 and WO3. The electrons in the conduction band of WO3 might combine with the holes in highest occupied molecular orbital (HOMO) levels of g-C3N4, suppressing the recombination of charge carriers. Furthermore, it is noted that the PL intensity for WC3 slightly decreases compared to that of WC3, implying that the oxygen vacancies in WO3 can further promote the separation of charge carriers. The electrons in the energy levels of oxygen vacancies might recombine with the holes in HOMO levels of g-C3N4, further separating the charge carriers.

Figure 5

Photoluminescence spectra (a) and time-resolved PL decay curves (b) for WC3 and WC3 samples.

Time-resolved PL decay curve is a sensitive and powerful method to further study the separation behaviors of charge carriers. As shown in Figure b, the fast decay lifetime (τ1) is usually related to the nonradiative relaxation process and the longer decay lifetime (τ2) is related to the radiative process, arising from the recombination of photogenerated electrons and holes. The values of τ1 and τ2 for WC3 and WC3 are calculated by double exponential decay fitting. The τ2 value for WC3 is 6.29 ns, which is much longer than that for WC3 (3.91 ns). This result suggests that the introduction of oxygen vacancies on the surface of WO3 can prolong the lifetime of charge carriers in the WO3–/g-C3N4 heterojunction. The photogenerated electrons in WO3 would be trapped in the energy levels of oxygen vacancies and further combine with the holes in HOMO levels of g-C3N4. It becomes difficult for the electrons to stay in the lowest unoccupied molecular orbital (LUMO) levels of g-C3N4 to recombine with the holes. A similar reason can be applied to the holes in the valence band of WO3. Therefore, the lifetime of the charge carriers for WC3 is prolonged, which is in favor of the photocatalytic performance eventually.

Moreover, the photocurrent responses of WC3 and WC3 are illustrated in Figure S4. The photocurrent densities for WC3 are much higher than those of WC3, which is in good agreement with the photocatalytic results. This observation confirms that the introduction of the oxygen vacancies on the surface of WO3 can enhance the visible-light response and facilitate the separation of charge carriers of the WO3–/g-C3N4 heterojunction. The detailed photocatalytic mechanism will be discussed in the following section.

The diffuse reflectance UV–visible (UV–vis) absorption spectra of g-C3N4, WO3, WO3–, and WC3 nanorods are plotted in Figure a. Pure WO3 samples show strong absorption in the UV light region and also exhibit some absorption from 400 to 550 nm. For WO3– samples, the absorption edge is almost the same as that of pure WO3, suggesting that the vacuum thermal treatment can hardly change the band gap of WO3. However, a broad peak from 400 to 800 nm is observed for WO3– samples, whose absorption maximum is at approximately 523 nm. It has been reported that the oxygen vacancies would induce defect energy levels below the conduction band of WO3.[18,19,28] Thus, the enhancement of visible-light response can be ascribed to the electron transition from the valence band to energy levels of oxygen vacancies. On the other hand, g-C3N4 shows a strong absorption, ascribed to the electron transition from the HOMO levels to the LUMO levels. The maximum of absorption peak of g-C3N4 is also at approximately 420 nm. After WO3– is composited with g-C3N4, a red shift is observed for the absorption edge and the absorption around 420 nm is enhanced for the WC3 samples.

Figure 6

(a) Diffuse reflectance UV–vis absorption spectra and (b) corresponding Tauc plots of g-C3N4, WO3, WO3–, and WC3 samples.

The band gap of semiconductors can be determined by the following Tauc equationwhere α is the absorption coefficient, h is the Planck’s constant, v is the light frequency, A is the proportionality, and Eg is the band gap. n is 1/2 for direct band gap semiconductors and 2 for indirect band gap semiconductors. In this work, WO3 is a direct band gap semiconductor and g-C3N4 is an indirect band gap semiconductor.[9−11] Hence, n is 2 for g-C3N4 and 1/2 for WO3. As shown in Figure b, the corresponding band gaps of g-C3N4 and WO3 are estimated to be 2.62 and 3.17 eV. The band gap of WO3 in this work is larger than that from previous reports,[10,12,29−31] which may be caused by the ultrathin diameter of nanorods. After the introduction of oxygen vacancies, the band gap of WO3– nanorods remains as 3.17 eV. The direct band gap for WC3 is narrowed to 3.05 eV, corresponding to the red shift of the absorption edge. As the absorption maximum of the tailed peak for WO3– samples originating from oxygen vacancies is around 523 nm, the energy level of oxygen vacancies is deduced to be approximately 0.8 eV below the conduction band minimum (CBM) of WO3.

The XPS valence band spectra of WO3, WO3–, WC3, and g-C3N4 are plotted in Figure . The valence band maximum of WO3 is approximately 3.20 eV (+2.80 eV, vs normal hydrogen electrode (NHE)), and the HOMO level for g-C3N4 is at approximately 2.05 eV (+1.65 eV, vs NHE). Moreover, the valence bands of WO3– and WC3 samples are almost the same as that of pure WO3. According to the discussion above, the corresponding band gaps of WO3 and g-C3N4 are estimated to be 3.17 and 2.62 eV. Therefore, the LUMO levels and the conduction band minimum for g-C3N4 and WO3 are estimated to be at −0.97 and −0.37 eV (vs NHE), respectively.

Figure 7

XPS valence band spectra of pure WO3, WO3–, WC3, and g-C3N4.

Based on the discussion above, the schematic diagram of the band structure for WO3/g-C3N4 heterostructures can be plotted in Figure S5. For WO3/g-C3N4 heterostructures, there are two possible pathways to separate charge carriers, namely, typical heterostructure pathway and Z-scheme pathway. In the typical heterostructure pathway, the holes would migrate from the valence band of WO3 to the HOMO levels of g-C3N4; the electrons in the LUMO levels of g-C3N4 would transfer to the conduction band of WO3. In the Z-scheme pathway, the excited electrons in the conduction band of WO3 can recombine with the holes in the HOMO levels of g-C3N4 directly, leaving the electrons enriched in the LUMO levels of g-C3N4. As shown in Figure S5, the conduction band minimum of WO3 (−0.37 eV, vs NHE) is less negative than the reduction potential of CO/CO2 (−0.52 eV, vs NHE). As a result, the electrons in the conduction band of WO3 cannot reduce the CO2 molecules to produce CO effectively. Therefore, it is necessary for the photoinduced carriers to be separated in the Z-scheme pathway to reduce CO2 into CO for the WO3/g-C3N4 heterostructures. The generation of hydroxyl radicals •OH can be a strong and direct evidence to confirm that g-C3N4/WO3 is a direct solid-state Z-scheme heterostructure.[32,33] According to Figure S6, it can be confirmed that the charge carriers are separated in the Z-scheme way for WO3/g-C3N4 samples in this experiment.

The detailed improved photocatalytic mechanism of WO3–/g-C3N4, illustrated in Figure , can be explained as follows. The photocatalytic reduction mechanism of CO2 into CO can be explained by eqs –3[34−38]To achieve efficient photoreduction of CO2 into CO, the potential of electrons must be more negative than the reduction potential (Ered = −0.52 V vs NHE) and the potential of holes must be more positive than the oxidation potential (Eox = +0.82 V vs NHE). Although the electrons and holes for g-C3N4 can participate in the reaction directly, the photocatalytic activity is still limited because of the rapid recombination of charge carriers (Figure a). Both pure WO3 and WO3– nanorods exhibit a poor photocatalytic activity because of the relative low potential of conduction band minimum, as the photogenerated electrons can hardly participate in the reaction in eq . After WO3– nanorods are composited with g-C3N4 nanosheets, the holes in the HOMO levels of g-C3N4 would combine with the electrons in the conduction band of WO3–, separating the carriers efficiently in the Z-scheme pathway. As the LUMO levels of g-C3N4 and the valence band maximum of WO3 match the redox potential of CO2 reduction (eqs –3), photogenerated carriers could react with the surface adsorbed CO2 molecules to participate in the photoreduction reaction. It is noted that the photocatalytic activity can be further enhanced for WO3–/g-C3N4 samples by introducing the oxygen vacancies into the WO3 surface. The energy levels of the oxygen vacancies locate at approximately 0.8 eV below the CBM of WO3–. The electrons can be excited from the valence band to the energy levels of oxygen vacancies directly. The electrons in the conduction band of WO3– would also be trapped by the energy levels of oxygen vacancies. As the energy levels of oxygen vacancies are close to the HOMO levels of g-C3N4, the electrons in the energy levels of oxygen vacancies can combine with holes in the HOMO levels of g-C3N4 more easily, promoting the separation of photogenerated electrons and holes in the Z-scheme pathway. The holes left in the valence band of WO3– can oxidize the H2O into H+ (eq ), and the electrons left in the LUMO levels of g-C3N4 can react with CO2 and H+ to yield CO and H2O (eq ). Therefore, the oxygen vacancies at the interface of WO3– and g-C3N4 could extend the visible-light response, facilitate the separation of charge carriers in the Z-scheme pathway, and prolong the lifetime of electrons and holes, resulting in a significantly enhanced photocatalytic activity on the reduction of CO2 into CO.

Figure 8

Schematic illustration of CO2 photoreduction into CO for WO3–/g-C3N4.

Conclusions

In summary, we have constructed oxygen vacancy-modified Z-scheme heterostructures by compositing WO3– nanorods with g-C3N4 nanosheets. The photocatalytic activity on the photoreduction of CO2 into CO is remarkably enhanced for WC3 samples (WO3– nanorods composited with 3 wt % of g-C3N4), compared to WO3, g-C3N4, and WC3 samples. It is revealed that the introduced oxygen vacancies result in defect energy levels below the conduction band of WO3–, which extend the visible-light response, promote the separation of charge carriers in the Z-scheme, and prolong the lifetime of photogenerated electrons. This work may provide physical insights for designing and fabricating oxygen vacancy-modified novel heterostructures with highly efficient performance.

Experimental Section

Synthesis of WO3–/g-C3N4

WO3 nanorods were synthesized via the typical hydrothermal method: 3 mL of concentrated hydrochloric acid (12 mol/L) and 3.3 g of Na2WO4·H2O were added into 70 mL of deionized water (18.2 MΩ·cm) and a yellow floc formed immediately. This mixture was added with 20 g of K2SO4 and then transferred into a 100 mL Teflon-lined stainless autoclave and heated at 180 °C for 24 h. After cooling down to room temperature, the obtained participates were washed with deionized water six times and then dried at 60 °C for 10 h. The WO3– nanorods with oxygen vacancies were prepared by vacuum thermal treatment. WO3 powders (0.3 g) were placed in an electric vacuum drying oven. The vacuum pressure was 5–20 mTorr, and the temperature was elevated to 200 °C with a ramp rate of 10 °C/min. After vacuum thermal treatment for 5 h, the powders were placed in air and allowed to cool down to room temperature naturally and were labeled as WO3–. The g-C3N4 was prepared by directly calcining 20 g of urea in muffle with a heating rate of 2.5 °C min–1 and then kept at 550 °C for 3 h.

The WO3–/g-C3N4 heterostructures were prepared as follows: WO3– powders (0.2 g) were dispersed into 50 mL of deionized water and ultrasonicated in 40 KHz for 3 h, which was designated as sample A. At the same time, g-C3N4 powders (6 mg) were dispersed in 50 mL of 0.02 mol/L hydrochloric acid and ultrasonicated in 40 kHz for 3 h, which was designated as sample B. Sample A was mixed with sample B under vigorous stirring for 12 h and dried in an oven for 10 h. The obtained powders were designated as WCX, where X% indicated the nominal weight ratio of g-C3N4 to WO3–. Similarly, WO3/g-C3N4 was also prepared by the same procedure just by replacing the WO3– powders with the WO3 powders. The obtained samples were designated as WC3 (0.2 g of WO3 and 6 mg of g-C3N4).

Characterization

XRD patterns were collected on a Rigaku D/max 2500 X-ray diffraction spectrometer (Cu Kα, λ = 1.54056 Å). HRTEM images were obtained by a JEOL 3010, for which the samples were prepared by applying a drop of ethanol suspension onto an amorphous carbon-coated copper grid and dried naturally. XPS measurements were carried out with a VG Scientific ESCA Lab 250i-XL spectrometer by using an unmonochromated Al Kα (1486.6 eV) X-ray source. All of the XPS images were calibrated with respect to the binding energy of the adventitious C 1s peak at 284.8 eV. Diffuse reflectance UV–vis absorption spectra were recorded on a UV–vis spectrometer (U-4100, Hitachi). PL spectra were measured by fluorescence spectrophotometer (Edinburgh Instruments, FLS920) using the 340 nm line of a Xe light as the excitation source. The ESR spectra were measured on a Bruker a300 spectrometer at room temperature in air. The time-resolved PL decay curves were measured on FLS 980 (Edinburgh Instruments, FLS920) excited at 340 nm and monitored at 450 nm.

Evaluation of Photocatalytic Activity

The photocatalytic reduction of carbon dioxide (CO2) into carbon oxide (CO) was performed in a sealed 275 mL Pyrex glass reactor. Photocatalysts (100 mg) were uniformly dispersed on the round bottom of the reactor (about 18 cm2). This reactor was inflated with CO2 gas (99.999%) at a flux of 0.3 L/min for 45 min consistently to exclude other gases (for example, N2 and O2). Then, 0.4 mL of deionized water was injected into the reactor. The whole reactor was sealed and placed at 27 cm right below a 500W Xe lamp. At 2 h intervals, 0.4 mL of gas was extracted to detect the amount of yield CO by a gas chromatograph (GC 2010 plus, SHIMADZU).