Regulating electron transfer pathway in Au/W18O49 heterostructures by structural design for revealing the photocatalytic mechanism of metal/semiconductor heterostructures

The construction of metal/semiconductor heterostructures is a useful technique for improving the electron-hole separation of Semiconductor Photocatalysts. However, there only are a few studies on the mechanism of electron transfer between metal/semiconductor heterostructures. Therefore, through the intelligent design of the material structure, two metal/semiconductor heterostructures (Au/W18O49 heterostructure) were fabricated with identical composition but different structures by different preparation methods: (1) The heterostructure of Au nanoparticles at the tip of sea urchin W18O49 was achieved through photoreduction method; (2) the heterostructure of Au nanoparticles at the root of sea urchin W18O49 was achieved via chemical reduction method, and their electron transfer paths were studied. The results demonstrated that the two materials went through completely different electron transfer paths, and these different electron transfer path also leads to an opposite set of photocatalytic properties. The heterostructure achieved through photoreduction has the best photocatalytic performance. Nevertheless, the photocatalytic performance of the heterostructure prepared by chemical reduction is far inferior to that of the heterostructure prepared by photoreduction, and even inferior to the original W18O49 sample. Therefore, we believe that the structural characteristics of metal/semiconductor heterostructure have a great influence on the electron transfer path. Our work renders useful information that may facilitate the design of heterostructure photocatalyst based on metal/semiconductors.


Introduction
As a typical n-type semiconductor material, tungsten oxide has the advantages of abundance, low cost, ecofriendliness, non-toxicity, stability, and efficiency. In particular, the highly adjustable crystal structure usually leads to a unique set of physical and chemical properties. Therefore, it demonstrates great potential in photocatalysis [1,2], electrochemistry [3,4], and photothermal therapy [5,6]. The lattice structure of tungsten oxide can withstand the existence of a large number of oxygen vacancies, hence it can exist in the form of substoichiometric (WO 3−x ), such as: W 25 O 73 , W 20 O 58 , W 18 O 49 , etc. The generation of oxygen vacancies generally leads to the formation of new defect levels in the gap band and consequently increases the free carrier density in tungsten oxide [7]. High oxygen vacancy concentration will lead to high free carrier concentration, producing a local surface plasmon resonance effect (LSPR) [8], which can significantly enhance the light absorption of WO 3−x in the visible or near-infrared region [7]. Surprisingly, in addition to improving semiconductor photoresponse, WO 3−x 's LSPR can also excite hot carrier production, which has been shown to play a significant role in driving the catalytic reaction [9][10][11], hence WO 3−x is widely used in the field of photocatalysis. However, due to their rapid recombination (∼100 fs), these active hot carriers have a low photocatalysis utilization efficiency [12,13]. Meantime, the intrinsically excited electrons and holes of WO 3−x also have a high recombination rate in the photocatalytic process [14][15][16]. All these limit the application of WO 3−x in the field of photocatalysis.
Based on the above problems, many researchers have put forth several solutions, Among these, the construction of metal/semiconductor heterostructures is an effective strategy to improve hot-electron retention and electron-hole separation [17][18][19]. Until now, several research studies have focused on developing and investigating Tungsten oxide/metal heterostructures. By assembling Ag nanorices (NRs) onto W 18 O 49 nanowires, Yang Liu et al [20] create a new form of plasmonic coupling het-erostructures film (NWs). The results show that photoinduced electrons are transferred at an ultrafast speed from W 18 O 49 to Ag NRs in their heterostructures, thereby preventing 'hot electron' recombination. However, in the work by Wei Wei et al [21]. Ag/W 18 O 49 heterostructures were also prepared, but there were completely opposite electron transfer paths, photoinduced electrons are transferred at an ultrafast speed from Ag to W 18 O 49 in their heterostructures. Meanwhile, in a study directed at enhancing the separation of intrinsic excited electrons and holes, Yechen Wang et al [22], developed an efficient WO 3 /Au photocatalyst that can generate H 2 O 2 under a visible wavelength range. The experiment indicates that there are two electron transfer pathways: (1) the transfer of an electron from the excited WO 3 to Au under UV irradiation, and (2) the transfer of an electron from the excited Au to WO 3 under visible irradiation. The above studies all report heterostructures formed by tungsten oxide and precious metals. However, the electron transfer paths between tungsten oxide and precious metals are different. As already known, the determination of the electron transfer path is a very important problem in the design of metal/semiconductor heterostructure photocatalysts. However, there are relatively few studies investigating the various factors influencing the electron transfer path. We believe that this is a problem worthy of study for the design of photocatalysts based on metal/semiconductor heterostructures.
In this work, we prepared sea urchin-like W 18 O 49 nanocrystals by a simple solvothermal method, and subsequently, gold nanoparticles were grown at different positions within W 18 O 49 nanocrystals by employing various reduction methods to prepare W 18 O 49 /Au heterostructures. It has been found that the Au nanoparticles on the heterostructure prepared by photoreduction grow at the tip of the urchin-like a W 18 O 49 rod, and the gold nanoparticles on the heterostructure prepared by chemical reduction grow at its roots like a W 18 O 49 rod. In addition, we confirmed through the photoelectric performance test and photocatalytic performance test that the two photocatalysts with the same composition and different structures differ in terms of their electron transfer paths. This study confirmed that the structure of photocatalyst has a great influence on its electron transfer path and photocatalytic performance. Furthermore, we believe that this work could be of advantage to researchers attempting the design of efficient metal/semiconductor heterostructure photocatalysts. composite heterostructures were obtained through a photo-reduction process (abbreviated to WAPR in following text). In this process, 0.2 g of the as-prepared W 18 O 49 were dispersed into 50 ml deionized water, stir evenly, and then add 0.3 ml HAuCl 4 (50 mmol l −1 ) solution under the irradiation of 500 W xenon lamp (λ = 300 ∼ 800 nm). After illumination for 1 h, then wash and dry the obtained sample with ethanol and deionized water for later use.

Experimental
In addition, The Au/W 18 O 49 composite heterostructures were prepared by chemical reduction method (abbreviated to WACR in following text). 0.2 g of the as-prepared W 18 O 49 were dispersed into 50 ml deionized water. Thereafter, 0.3 ml HAuCl 4 (50 mmol l −1 ) pour the solution into the stirred suspension. Subsequently, the W18O49/Au composite heterostructure was obtained through a reducing agent (NaBH 4 ). It can be observed that the blue suspension changes to a light purple color. Subsequently, continue stirring and complete the reduction while ensuring the stability of the material structure. Finally, wash and dry with ethanol and deionized water for later use.

Characterization
The crystal structures of the samples were characterized by x-ray diffraction (XRD) (Bruker D8 Advance diffractometer using Cu Kα (λ = 1.5406 Å) radiation). Scanning electron microscopy (SEM) images of the samples were measured using Nova Nano SEM 450.Transmission electron microscope (TEM) images of sample were measured by JEM-2100 at 300 kV. The ultraviolet visible near-infrared absorption spectrum of the sample was characterized by a spectrophotometer (Hitachi UV-4100), The reflection sample is BaSO 4 . The photoluminescence spectrum of the sample was characterized by an FLSP-980 spectrophotometer (Edinburgh Instruments, Edinburg, UK). The photocurrent analysis of the sample was measured at room temperature using an electrochemical workstation (Bio-Logic SP-300).

Photocatalytic experiments
The photocatalytic activities of the prepared hybrid nanocomposites were evaluated by the degradation of rhodamine B (RhB) and photocatalytic H 2 evolution from ammonia borane (AB) aqueous solution. The photocatalytic efficiency of the sample for RhB decomposition was tested using a 500 W xenon lamp, accompanied by UV-vis (λ = 300-800 nm) illumination. To ensure adsorption desorption equilibrium, 20 mg of photocatalyst was added to the stirred RhB aqueous solution (60 ml, 20 mg l −1 ) in a dark environment. During the reaction process, take out 5 ml of suspension at regular intervals and centrifuge wash the sample. The residual organic pollutants were characterized by a Hitachi UV-4100 spectrophotometer. Further, in order to detect the generated active species in the photocatalysis. In this work, we use EDTA, IPA, and BQN acid are used as scavengers for trapping photogenerated holes (h + ), hydroxyl radicals (·OH), and superoxide radicals (·O 2− ), respectively.
Photocatalytic H 2 evolution from ammonia borane (AB) aqueous solution of samples was tested under UVvis (λ = 300 ∼ 800 nm) and full spectra (λ = 300-1100 nm) light irradiation with a 300 W Xe lamp (MC-PF300C, Beijing Merry Change Technology Co., Ltd). 20 mg of the as-fabricated samples were dispersed into 48 ml of deionized water through magnetic stirring. Then, degassing the sealed reactor with argon gas for 30 min. Subsequently, 2 ml of solution containing 10 mg of NH 3 BH 3 was added to the photoreactor. The generated H 2 was periodically analyzed by a Gas chromatograph (GC) equipped with a thermal conductivity detector (TCD) (FuLI 9790II). To maintain the catalytic reaction temperature, place the quartz reactor in a water tank connected to the reflux condenser.

Materials characterization
To look at the effect of the structure of metal/semiconductor heterostructure photocatalyst on its electron transfer path and photocatalytic performance, we fabricated photocatalysts with identical composition and differing structures by a simple two-step method (figure 1): (1) the W 18 O 49 were synthesized through a mild solvothermal method. (2) the Au NRs were randomly assembled onto the surface of W 18 O 49 by photoreduction and chemical reduction methods respectively to form composite heterostructures. First, to verify the successful preparation of the samples, we carried out the x-ray diffraction (XRD) analysis of the prepared specimens. The results have been shown in figure 2(a). It is evident from the analysis that the as-prepared W 18 O 49 samples possess a monoclinic structure according to the JCPDS card (002-4731). In particular, the sharp narrow diffraction peak at 23.4°corresponds to the (010) plane, and the weak reflection at 47.9°aligns with the (020) plane. In addition to the signals from W 18 O 49 , the diffraction peaks located at 38.2°and 44.4°in the XRD pattern of two W 18 O 49 /Au composite heterostructures could be assigned to the (111) and (200) crystal planes of metallic Au with the cubic crystal structure (JCPDS card no.46-1043). XRD patterns thus confirmed the successful preparation of the samples.
To prove that the two heterostructures prepared by different methods have different structures, the morphologies of the prepared samples were investigated using transmission electron microscopy (TEM) and scanning electron microscopy (SEM). From the SEM images in figure 2(b), it is obvious that the as-synthesized W 18 O 49 possesses an urchin-like structure assembled between several nanowires having lengths in the range of 500-600 nm and a diameter less than 20 nm. In addition, SEM tests were also conducted on WAPR and WACR, and the results are shown in  49 . This is because in the process of heterostructure preparation by the photoreduction method, the photogenerated electrons of W 18 O 49 were transmitted to the tip of sea urchin, and HAuCl 4 was reduced by the photogenerated electrons, leading to the preferential nucleation and growth of Au NRs at the tip. Figures 1(e) and (f) shows the W 18 O 49 /Au composite heterostructures prepared by the chemical reduction method. Au nanoparticles preferentially nucleate and grow between sea urchin rods with a large surface area and are randomly distributed between sea urchin rods. Meanwhile, the size of Au nanoparticles is adjusted by controlling the reduction time, hence, the size of gold nanoparticles in the sample prepared by the chemical reduction method is also 50 nm, which is identical to that prepared by the light reduction method. Therefore, it has been found that W 18 O 49 /Au heterostructure photocatalysts prepared by different methods have different morphologies and structures as depicted by the TEM micrographs of the samples.  To confirm whether the electron transfer paths of the two samples with different structures are different, we first tested the light absorption properties of the samples. As shown in figure 5(a), the W 18 O 49 sample give a pair of absorption bands: the peak at 200 ∼ 400 nm can be ascribed to the inter band absorption of the semiconductor feature, whereas the LSPR band is responsible for another peak ranging from 500 nm to the IR region, and arises from the collective oscillations of excess W 18 O 49 surface electrons due to the presence of abundant oxygen vacancies [23][24][25]. Moreover, in comparison with the original W 18 O 49 sample, The W 18 O 49 /Au composite heterostructure prepared by different methods has an absorption peak at 450 ∼ 620 nm, which is attributed to the LSPR band of Au nanoparticles [26]. However, it can be seen from the figure that the LSPR of the three samples is different. In theory, the electron density is responsible for the plasmon resonance wavelength of nanostructures, and their relationship is expressed by the following function: Where λ denotes the plasmonic resonance wavelength; m e represents the effective mass and Ne represents the electron density of metal nanostructures. As shown in figure 5( figure 5(a) and figure 5(b) that the light absorption of heterostructure photocatalysts prepared by different methods is also markedly different. Firstly, the LSPR band of W 18 O 49 on the heterostructure prepared by the photoreduction method undergoes a red shift with the heterostructure prepared by the chemical reduction method, which indicates that more interfacial electron transfer on the heterostructure prepared by photoreduction method leads to a lower concentration of electrons W 18 O 49 . Secondly, in comparison with the specimens obtained via the chemical reduction method, the LSPR band of Au nanoparticles on the samples prepared by the light reduction method manifests a blue shift. It is well known that the electron concentration on Au nanoparticles is markedly affected by its morphology [27], however, it is evident from the transmission electron micrographs that the morphology of Au nanoparticles on the samples prepared by the two methods is similar. Therefore, we believe that the interfacial electron transfer from W 18 O 49 to Au nanoparticles will occur following the formation of heterostructures. However, the interfacial electron transfer of Au nanoparticles on the sample at the tip is more intense. The above outcome primarily verified our understanding that the photocatalyst structure has a great influence on the electron transfer path of metal/semiconductor composite photocatalyst.
To further verify our conjecture, we tested the photoluminescence spectra of the prepared samples. Photoluminescence (PL) was also utilized to investigate the produced photocatalysts' capacity to separate photogenerated electron-hole (e _ h + ) pairs, typically, a lower PL intensity indicates a slower recombination rate [28]. As illustrated in figure 5(c), the luminescence peak intensity of W 18 O 49 /Au composite heterostructures prepared by the photoreduction method is lower than that of pristine W 18 O 49 sample. This confirms our presumption that the interfacial electron transfer to Au nanoparticles following the formation of heterostructures leads to the reduction of the recombination rate of photoinduced electron hole pairs. However, in comparison with the pristine W 18 O 49 sample, the W 18 O 49 /Au composite heterostructures prepared by the chemical reduction method demonstrate a stronger photoluminescence intensity. This shows that the formation of composite heterostructures on the samples prepared by the chemical reduction method does not reduce the rate of recombination of photoinduced electron-hole pairs it increases the recombination rate, which is an abnormal phenomenon. Mei Li et al [29], irradiated sea urchin-like W 18 O 49 by a xenon lamp resulting in the generation of e − -h + pairs by the W 18 O 49 material. Subsequently, Photogenerated electrons can quickly move along radial nanorods of a sea urchin-like structure, effectively separating electrons and holes. Meanwhile, the nanotip effect enriched the electrons at the tips of rods and the holes at the roots of rods. As a result, we believe that gold nanoparticles on light-reduced samples are at the tip of electron enrichment, whereas gold nanoparticles on chemical-reduced samples are at the root of hole enrichment, resulting in distinct electron transport routes under xenon lamp irradiation. In the samples prepared by photoreduction, the interface electrons are easier to transfer from W 18 O 49 to Au nanoparticles because the gold nanoparticles are at the tip of electron enrichment whereas, in the samples prepared by the chemical reduction method, the transfer of interface electrons from W 18 O 49 to Au nanoparticles is rather difficult. After all, Au nanoparticles are at the root of hole enrichment. In addition, under the influence of light, the transferred interface electrons will be transformed into high-energy hot electrons and transferred to W 18 O 49 again to combine with the holes at the root. This will render the luminescence intensity of the sample prepared by the chemical reduction method greater than that of the original sample. In addition, if the above speculation is correct, the concentration of photogenerated electrons on the sample prepared by the chemical reduction method will be much less than that of the original sample, and the concentration of photogenerated electrons on the sample prepared by the photoreduction method will be far greater than that of the original sample.
Therefore, to further evaluate the photogenerated electron concentration of the prepared samples, we also tested the photocurrent of the samples. The photocurrent was measured in a standard three-electrode system with a bare glassy carbon electrode as the working electrode, a Ag/AgCl electrode as the reference electrode and a carbon rod as the counter electrode in a 0.5 M Na 2 SO 4 electrolyte solution in the frequency range of 0.01-10 5 Hz with a voltage amplitude of 0.01 Vat and an open circuit potential of 1 V. As depicted in figure 5(d), The photocurrent of the sample prepared by photoreduction is the largest, followed by the original sample, and the smallest is the sample prepared by chemical reduction. This test result further verifies our presumption.

Photocatalytic activities
Next, we verified our presumption through photocatalytic experiments. First, to verify that gold nanoparticles in the root of sea urchin like W 18 O 49 rod will reduce the concentration of photogenerated holes, we selected rhodamine B as the target of photocatalytic degradation. Because rhodamine B is often used to test the catalytic performance of photocatalysts [30][31][32]. The time-dependent degradation percentage of RhB under visible light irradiation was calculated by (C/C 0 ), where C 0 represents the original concentration of the dye (mg * l −1 ) and C represents the concentration of the dye following Sunlight irradiation for a certain time (mg * l −1 ). The results have been shown in figure 6(a). It is noteworthy that for the first 90 min following the addition of material to the dye, the reaction was carried out in the dark, and the decrease of dye concentration was attributed to the adsorption of the material. From the 60th minute onwards to the 90th minute, the dye concentration remained basically unchanged, indicating that the material reached adsorption-desorption equilibrium. It can be seen clearly from figure 6(a) where C 0 is the original concentration of RhB and C is the concentration of RhB at different reaction times, and k is the rate constant. The rates of the reaction k of the WO, WACR, and WAPR are 0.0155 min −1 ,0.0107 min −1 and 0.0589 min −1 ( figure 6(b)). In addition, the samples prepared by photoreduction only fit the first three points because they degrade the dye completely only in the first 40 min of illumination, however, there is little change after 40 min of illumination. The above photocatalytic dye degradation experiments show that the samples prepared by the photoreduction method demonstrate the best performance in terms of photocatalytic degradation of dyes, whereas the samples prepared by chemical reduction method have the worst performance, which is consistent with our previous speculation. It is more difficult to transfer the interfacial electrons in the samples prepared by the chemical reduction method. Simultaneously the transferred interfacial electrons will be affected by the high concentration of photogenerated holes on W 18 O 49 , which will transfer from Au nanoparticles to W 18 O 49 again and combine with the holes, resulting in a great decrease in the concentration of photogenerated holes on W 18 O 49 , and finally, the photocatalytic degradation efficiency of dyes is greatly reduced.
Additionally, we carried out free radical elimination experiments on the samples prepared by photoreduction. Figure 7 shows the rate of degradation of a typical catalyst with different scavengers. With the addition of EDTA in the dye solute ion, the degradation efficiency reduced slightly from 93.52% to 77.84%. In the presence of BQN and IPA, the degradation efficiency was dramatically reduced to 42.40% and 25.18%. Therefore, we believe that the degradation mechanism of RhB dye is as follows: The above free radical elimination experiments show that the photogenerated holes of the sample play the most important role in the photocatalytic degradation of RhB dyes. On one hand, they have direct participation in the photocatalytic degradation of dyes. On the other hand, they combine with water molecules to form ·OH and subsequently participate in the photocatalytic degradation of dyes. A large number of photogenerated holes are left within the samples prepared by the photoreduction method following interfacial electron transfer, which leads to a great improvement in photocatalytic performance. However, due to the different positions of Au nanoparticles in the samples prepared by the chemical reduction method, the recombination of photogenerated holes leads to a significant decline in photocatalytic performance.
Secondly, to verify the transfer path of hot electrons produced by the local surface plasmon resonance effect, we carried out a photocatalytic hydrogen production experiment of aminoborane (AB) aqueous solution. Given the photocatalytic hydrogen production in AB aqueous solution, the current study suggests that a single noble metal nano-structure can drive photocatalytic H 2 evolution from multiple hydrogen-carrier molecules [34,35]. and the hot electrons generated by local surface plasmon resonance play an important role in the photocatalytic hydrogen production process of AB aqueous solution [36,37]. As shown in figure 6(d) depicts the rate of hydrogen production of the three samples under xenon lamp illumination with a wavelength range of 300-800 nm and figure 6(e) shows the hydrogen production rates of the three samples under xenon lamp illumination at a wavelength of 300-1100 nm. The hydrogen production rate of the samples prepared by any method is the highest under 300-1100 nm xenon lamp irradiation. Simultaneously, whether the cut-off wavelength of xenon lamp is 800 nm or 1100 nm, the hydrogen production rate of W 18 O 49 /Au composite heterostructure prepared by photoreduction is significantly higher than that of the pristine W 18 O 49 sample. However, the hydrogen production rate of W 18 O 49 /Au composite heterostructures prepared by the chemical method is the lowest. Among them, the hydrogen production rate of the sample prepared by the photoreduction method reaches a value of 9 mmol/g/h under 300-1100 nm xenon lamp irradiation, which is 1.58 times in comparison to the original sample and 4.11 times in comparison to the sample prepared by chemical reduction method. The hydrogen production rate of the sample prepared by photoreduction method reaches 3.72 mmol/ g/h under a 300-800 nm xenon lamp irradiation, which was 2.08 times that of the original sample and 4.18 times that of the sample prepared by the chemical reduction method. The catalytic hydrolysis of NH 3 BH 3 for H 2 evolution over the suitable catalysts is according to the following equation [38][39][40][41]:  [43,44]. The photo-thermal effect, in addition to plasmon-induced hot electrons, contributes to the improvement of catalytic activity for H 2 evolution on the W 18 O 49 /Au heterostructures by hydrolyzing NH 3 BH 3 under xenon lamp irradiation by hydrolyzing NH 3 BH 3 [45]. Because the localized temperature created by LSPR excitation can overcome a minor energy barrier in the hydrolysis reaction, the H 2 evolution reaction can be accelerated. This leads to the best photocatalytic hydrogen production performance of the samples prepared by photoreduction. On the contrary, in the samples prepared by chemical reduction, the transferred interfacial electrons will recombine with holes, resulting in a significant reduction in the concentration of hot electrons within the materials, resulting in the worst photocatalytic hydrogen production performance. The above experimental results confirm our conjecture.

Feasible mechanism
First, in order to determine the energy band structure of W 18 O 49 , its band gap is calculated as 2.86 eV from the absorption spectrum of W 18 O 49 ( figure 8(a)), and the valence band position of W 18 O 49 is measured by ultraviolet photoelectron spectroscopy (UPS) ( figure 8(b) 3.46 eV versus E f ). In addition, the work function of W 18 O 49 ( figure 8(c)) is also tested. The work function is the difference between the Fermi level and the vacuum level [46][47][48], so the Fermi level position of W 18 O 49 (−4.59 eV versus E VAC ) can be calculated through the work function. According to the above test results, the relative position of the energy band structure of W 18 O 49 and Au is obtained ( figure 8(d)), where the Fermi energy level position of Au is the standard value (−5.2 eV versus E VAC ).
As shown in figure 8(d), the Fermi energy level of W 18 O 49 is 0.61 eV higher than that of Au. This means that when they touch in the dark, the electrons will transfer from W 18 O 49 to Au until they reach thermodynamic equilibrium (figure 9), that is, their Fermi energy levels will align. At this time, negative charges will be enriched on the side near Au at both ends of the heterostructure interface, and positive charges will be enriched on the side near W 18 O 49 . When W 18 O 49 is positively charged and Au is negatively charged, a Hermann von Helmholtz double layer can be established at the W 18 O 49 /Au interface to form a built-in electric field. The built-in electric field around the space charge region of W 18 O 49 will cause the energy band of W 18 O 49 to bend upwards [49].
Based on the above results and previous research, we propose a reasonable mechanism for interfacial charge transfer on Au/W 18 O 49 heterostructures. As shown in figure 9, the sea urchin shaped W 18 O 49 is illuminated by xenon lamps to generate e − -h + pairs. Subsequently, photogenerated electrons can quickly transfer along the radial nanorods of the sea urchin like structure, effectively separating electrons and holes. Meanwhile, due to the nano tip effect, electrons are enriched at the tip of the rod and holes are enriched at the root of the rod. At this point, at the tip of the sea urchin like W 18 O 49 , the built-in electric field at the interface between W 18 O 49 and Au disappears due to electron enrichment, while hole enrichment at the root does not cause the built-in electric field at the interface between W 18 O 49 and Au to disappear.
The disappearance of the built-in electric field at the interface between W 18 O 49 and Au in the WAPR (figure 10(a)) leads to the transfer of photo generated charge carriers and LSPR hot electrons in the conduction band of W 18 O 49 to lower energy Au nanoparticles, which is conducive to the separation of hot electrons and the separation of electrons and holes. However, it is difficult for electrons on W 18 O 49 to transfer to Au nanoparticles due to the presence of built-in electric field in the WACR ( figure 10(b)). In addition, the electrons on Au nanoparticles are excited to high-energy positions by the LSPR effect. Under the built-in electric field, the interface electrons will transfer to W 18 O 49 and recombine with the holes enriched at the root, which increases the cooling of hot electrons and the recombination of electron holes (figures 9 and 10).

Conclusion
In conclusion, Au nanoparticles were anchored at different positions of W 18 O 49 by two different methods to prepare two metal/semiconductor heterostructure photocatalysts with identical compositions and different structures. The study shows that there are different electron transfer routes in the two materials: (1) In the samples prepared by photoreduction, the transfer of interface electrons from excited state W 18 O 49 to Au nanoparticles, since Au nanoparticles are located at the tip of electron enrichment; (2) In the samples prepared by chemical reduction method, the transfer of interface electrons from excited state W 18 O 49 to Au nanoparticles is more difficult because Au nanoparticles are located at the root of hole enrichment. At the same time, the transferred interface electrons will be excited again, transferred to W 18 O 49 and recombined with holes. Therefore, we believe that the structural attributes have a great influence on the electron transfer of metal/ semiconductor hetero-structures. This work thus furnishes useful data for the design and development of metal/semiconductor heterostructure photocatalysts.