**3. Results**

The crystallographic structures of the as-prepared ZnO nanorods, ZnO–WO3 nanorods with and without thermal annealing at 400 ◦C were identified using XRD measurements (Figure 1). Figure 1a shows the XRD pattern of the ZnO nanorods which were used as a template for preparing various composite nanorods. The Bragg reflections in Figure 1a demonstrate that the ZnO nanorods has a hexagonal wurtzite structure and exhibits preferred (002) orientation (JCPDS no. 005-0664). The XRD pattern of the ZnO nanorods in situ sputtering coated with WO3 thin films was exhibited in Figure 1b. The Bragg reflections centered approximately 24.5◦ and 34.2◦ are ascribed to (001) and (201) crystallographic planes of orthorhombic WO3 (JCPDS no. 20-1324), respectively. Well crystalline ZnO–WO3 composite nanorods were successfully formed via sputtering WO3 thin films onto the surfaces of the ZnO nanorods. Figure 1c shows the XRD patterns of the ZnO–WO3 nanorods annealed at 400 ◦C. Figure 1c shows that the WO3 peaks are disappeared completely after the annealing procedure and several new diffraction peaks are observed at approximately 19.6◦, 24.5◦, 25.3◦ and 31.3◦ which are assigned to (100), (011), (110) and (111) of monoclinic ZnWO4 (JCPDS no. 15-0774). The XRD result transformation of ultra-thin WO3 thin film with the ZnO into the ternary ZnWO4 phase occurred after the thermal annealing procedure. M. Bonanni et al. reported that the solid-state reaction between ZnO and WO3 starts to develop at 350 ◦C [17]. The annealing temperature of 400 ◦C herein might have a sufficient thermal energy to activate the phase transformation between the WO3 and ZnO of the ZnO–WO3 composite nanorods. The XRD results revealed the ZnWO4 layers with a polycrystalline feature were formed on the surfaces of residual ZnO nanorods after the annealing procedure.

**Figure 1.** X-ray diffraction (XRD) patterns: (**a**) ZnO nanorods, (**b**) ZnO–WO3 composite nanorods, (**c**) ZnO–WO3 composite nanorods annealed at 400 ◦C.

Figure 2 shows SEM images of the ZnO, ZnO–WO3 nanorods with and without annealing at 400 ◦C. Figure 2a shows a typical SEM image of as-synthesized ZnO nanorods, revealing a hexagonal crystal featured cross-section of the ZnO nanorods with a diameter in the range of 85–100 nm. The surface of the ZnO nanorods was smooth. Figure 2b presents the SEM image of ZnO nanorods after sputtering coated with ultra-thin WO3 thin film. Compared to the bare ZnO nanorods, the surface feature of the ZnO–WO3 composite nanorods became more rugged when the WO3 crystallites were decorated onto the surfaces of the ZnO nanorods. Notably, the hexagonal cross-sectioned morphology of the ZnO nanorods was maintained after coating the WO3 thin film, revealing the deposition of the ultra-thin WO3 layer on the ZnO nanorods' surfaces. Figure 2c shows the ZnO–WO3 composite nanorods annealed at 400 ◦C. The ZnO–WO3 composite nanorods annealed at 400 ◦C did not exhibit substantial morphology change. The composite nanorods maintained a visible hexagonal cross-sectional crystal feature. By contrast, a high solid-state reaction temperature above 600 ◦C involves the marked surface roughening process of oxide composite nanorods; this was observed in phase transformation of other oxide nanocomposite systems [16].

**Figure 2.** Scanning electron microscopy (SEM) micrographs: (**a**) ZnO nanorods, (**b**) ZnO–WO3 composite nanorods, (**c**) ZnO–WO3 composite nanorods annealed at 400 ◦C.

Figure 3a shows a low-magnification TEM image of a ZnO nanorod coated with a thin WO3 layer. The surface of the composite nanorod exhibited an uneven feature. Figure 3b,c demonstrate high-resolution TEM (HRTEM) images of a ZnO–WO3 composite nanorod taken from the different positions at the WO3/ZnO interface. Figure 3b reveals an ultra-thin WO3 layer covered on the surface of the nanorod and the interface between the ZnO nanorod and WO3 layer is abrupt. By contrast in Figure 3c, tiny, nanoscaled surface bumps appeared on the surface of the composite nanorod. This might engender the uneven surface feature of the composite nanorod. The sputtering growth of binary oxides at an elevated temperature is likely to form island- or bump-like crystals on the hetero-substrates [4,7]. The in situ sputtering growth of WO3 crystals onto the surfaces of the ZnO nanorods at 375 ◦C herein might cause locally inhomogeneous crystal growth and formed WO3 bumps on the ZnO nanorods. The ordered lattice fringes in the outer region of the composite nanorod revealed the coverage of well-crystallized WO3 crystals on the surface of the nanorod. The lattice fringe spacing of approximately 0.38 nm corresponds to the interplanar distance of orthorhombic WO3 (001). Furthermore, the lattice fringe spacing of approximately 0.26 nm in the figures demonstrated the interplanar distance of hexagonal ZnO (002). The selected area electron di ffraction (SAED) pattern taken from several ZnO–WO3 composite nanorods revealed the crystalline feature and a composite structure of the hexagonal ZnO nanorods sputtering coated with the orthorhombic WO3 thin film. Furthermore, elemental line-scan profiles of Zn, W, and O elements across the ZnO–WO3 composite nanorod were displayed in Figure 3e. The line-scan profiles revealed that the W element was well distributed on the surface of the ZnO nanorod, revealing a formation of the compositionally defined composite structure of core-ZnO and shell-WO3.

**Figure 3.** Transmission electron microscopy (TEM) analysis of the ZnO–WO3 composite nanorod: (**a**) low-magnification image, (**b**,**<sup>c</sup>**) high-resolution images from local regions, (**d**) selected area electron di ffraction (SAED) pattern, (**e**) line-scan profiling spectra across the composite nanorod.

Figure 4a shows a low-magnification TEM image of a ZnO–WO3 composite nanorod with a thermal annealing at 400 ◦C. HRTEM images of the composite taken from various interfacial regions are demonstrated in Figure 4b,c. In Figure 4b, well-ordered and long-range arrangemen<sup>t</sup> of lattice fringes appear at the outer region of the composite nanorod. Moreover, the ordered lattice fringes arranged in the other orientation were found in the inner region of the composite nanorod. The lattice fringes with a spacing of approximately 0.36 nm in the outer region of the composite nanorod were attributed to the interplanar distance of monoclinic ZnWO4 (110). By contrast, the lattice fringes with a spacing of 0.26 nm in the inner region of the composite nanorod were assigned to the interplanar distance of hexagonal ZnO (002). Figure 4b reveals that the WO3 phase in the outer region of the ZnO–WO3 composite nanorod transforms into a ternary phase of ZnWO4 through a solid-state reaction process during the post-annealing procedure in this study. The ZnWO4 crystals exhibited a good crystalline feature on the outer region of the composite nanorod and the interface of the ZnWO4 and ZnO phase was sharp. However, the mixed lattice fringes arrangements were observed in the outer region of the composite nanorod in Figure 4c. In addition to the ordered lattice fringes which originated from the ZnWO4 (110) as indexed in the figure, some local region demonstrated that the lattice fringes were

arranged in a slightly chaotic state. This revealed the presence of crystalline ZnWO4 and deteriorated WO3 crystals in the outer region of the composite nanorod. The observation of the TEM analysis demonstrated that most WO3 crystals transformed into crystalline ZnWO4 after annealing; whereas, partial WO3 crystals did not yield the phase transformation with the ZnO due to the insu fficient reaction condition. Moreover, the residual WO3 phase region demonstrated the deteriorated crystallinity after annealing because of the presence of hydrogen in the annealing atmosphere. In conclusion, the composite nanorod is mainly composed of crystalline ZnWO4 phase with a larger range and spatially distributed residual WO3 phase in a smaller content in the outer layer. The SAED pattern taken from the several composite nanorods in Figure 4d indicate the various groups of di ffraction rings, suggesting the presence of crystalline ZnO and ZnWO4 in the composite nanorods. The elemental mapping images were further used to analyze the distribution of Zn, W, and O in a single ZnO–WO3 composite nanorod treated with a thermal annealing at 400 ◦C (Figure 4e). The EDS mapping images clearly identified the spatial distributions of Zn, W, and O in the composite structure. The Zn and O elements existed the entire area of the nanorod. In particular, W is the main element distributed in the outer shell of the composite nanorod. TEM results indicated that a solid-state reaction occurred at the interface of WO3/ZnO and formed a new shell layer phase of ZnWO4 on the surface of the residual ZnO core in the composite nanorod. Similar solid-state reaction between the di fferent binary oxides in low-dimensional systems to form a ternary phase in the outer region of the composite nanorods has also been reported in ZnO–SnO2 and ZnO–TiO2 [16,18].

**Figure 4.** TEM analysis of the ZnO–WO3 composite nanorod annealed at 400 ◦C: (**a**) low-magnification image, (**b**,**<sup>c</sup>**) high-resolution images from local regions, (**d**) SAED pattern, (**e**) elemental mapping images of the nanorod.

The elemental binding states of the ZnO–WO3 composite with and without a thermal annealing at 400 ◦C were investigated by XPS. The narrow spectra of Zn 2p, W 4f, and O 1s for the ZnO–WO3 composite nanorods are recorded in Figure 5a–c. Zn 2p spectrum of the composite nanorods in Figure 5a shows two peaks at approximately 1044.8 eV and 1021.8 eV which are respectively attributed to Zn 2p1/2 and Zn 2p3/2 and sugges<sup>t</sup> the presence of Zn2<sup>+</sup> ions in the oxide [16]. Moreover, the W4 f spectrum (Figure 5b) of pristine ZnO–WO3 nanorods consisted of two spin-orbit doublets corresponding to the di fferent valence states of tungsten. The bigger doublet located at 37.4 eV and 35.3 eV is assigned to W 4f5/2 and W 4f7/2 of <sup>W</sup>6+, respectively and the smaller one is allocated to

W5<sup>+</sup> [11]. No metallic W component was detected from the sample. The asymmetric O 1s spectrum of ZnO–WO3 composite nanorods was displayed in Figure 5c and that spectrum was deconvoluted into three subpeaks at approximately 530.1 eV, 530.9 eV, and 531.9 eV, matching the oxygen coordination in lattice oxygen, vacancy oxygen, and surface chemisorbed oxygen, respectively [18,19]. By contrast, in Figure 5d, the Zn 2p spectrum of the ZnO–WO3 composite nanorods with a thermal annealing at 400 ◦C exhibited a similar spectrum feature as exhibited in Figure 5a, revealing the divalent state of the zinc in the nanorods. Figure 5e shows the W4f spectrum of the composite nanorods annealed in a hydrogen-contained atmosphere. Notably, even annealed in oxygen deficient atmosphere, the metallic W component was not detected on the surfaces of the composite nanorods under the given annealing condition. Comparatively, the W 4f spectrum with the deconvoluted peaks in Figure 5e exhibited that the ZnO–WO3 composite nanorods with the thermal annealing procedure demonstrated the area ratio of W5<sup>+</sup> spin-orbit doublet becomes larger, resulting from the existence of the crystal deterioration region in the shell oxide layer. The O1s spectrum of the corresponding sample was shown in Figure 5f and was further used to explain the W4f XPS result. The O 1s spectrum from the composite nanorods with a thermal annealing process showed a marked intensity decrease in the lattice oxygen subpeak and a relative intensity rise in the subpeaks associated with oxygen vacancy and chemisorbed oxygen, compared with those from the ZnO–WO3 composite nanorods without a thermal annealing. The concentration of oxygen vacancies has a direct relationship with the state of the oxide's crystallinity. A similar phenomenon of increased oxygen vacancies in oxides annealed in hydrogen-contained atmosphere has been proposed in previous works [20]. A higher degree of oxygen deficiency in the outer region of the composite nanorods engendered a larger content of W5<sup>+</sup> in the tungsten-based oxides of the composite nanorods. The XPS results herein demonstrated that the annealing temperature of 400 ◦C for the ZnO–WO3 composite nanorods engendered more oxygen vacancies in the surfaces; the tungsten was still in an oxide binding status without reducing to the metallic binding form.

**Figure 5.** X-ray photoelectron spectroscope (XPS) analysis of the ZnO–WO3 composite nanorods: (**a**) Zn 2p, (**b**) W 4f, (**c**) O1s. XPS analysis of the ZnO–WO3 composite nanorods after annealing: (**d**) Zn 2p, (**e**) W 4f, (**f**) O1s.

The optimal operating temperature with the highest gas-sensing response for the various nanorods was determined. The gas-sensing responses of all samples on exposure to 50 ppm ethanol vapor were measured at the temperature range of 200–325 ◦C (Figure 6a). For the ZnO nanorods, the gas-sensing response to 50 ppm ethanol vapor varied from 1.1 to 2.8 corresponding with the temperature from 200

to 325 ◦C (black curve in Figure 6a). The gas-sensing responses of ZnO–WO3 composite nanorods were from 1.3 to 7.3 (red curve in Figure 6a), which was substantially higher than that of the ZnO nanorods at all tested temperatures. Moreover, the gas-sensing responses of ZnO–WO3 composite nanorods annealed at 400 ◦C ranged from 2.1 to 16.2 (blue curve in Figure 6a), which demonstrated the highest response among the samples at the tested temperatures. Significantly, the nanorod sensors herein exhibited the maximum gas-sensing responses to ethanol at 300 ◦C, suggesting that a resultant equilibrium between surface reaction with ethanol vapor molecules and the di ffusion of ethanol vapor molecules to the nanorods' surfaces occurred at 300 ◦C [9]. Figure 6b–d show the dynamic response curves of ZnO nanorods, ZnO–WO3 nanorods, and ZnO–WO3 nanorods annealed at 400 ◦C, respectively, exposed to 25–500 ppm ethanol vapor at the operating temperature of 300 ◦C. For comparison, gas-sensing tests of the ZnO–WO3 composite nanorods annealed at 500 ◦C were also conducted to evaluate whether the higher annealing temperature improves the gas-sensing performance of the initially-synthesized ZnO–WO3 composite nanorods (Figure 6e). All nanorod samples exhibited reversible and stable response and recovery behaviors during gas-sensing tests. The nanorod samples herein showed a typical n-type sensing behavior because of the n-type conduction nature of the constituent oxides. The gas-sensing responses of the sensors made from various nanorods on exposure to various ethanol concentrations were summarized in Figure 6f. The WO3-decorated ZnO nanorod sensor exhibited much higher responses than the pristine ZnO. Furthermore, the ZnO–WO3 composite nanorods annealed at 400 ◦C showed the highest response in all ethanol vapor concentrations. Notably, the sensing ability of the ZnO–WO3 composite nanorods annealed at the higher temperature of 500 ◦C was substantially weakened. To confirm the possible reason for the deterioration of the gas-sensing ability, XPS measurements were conducted. Figure 6g demonstrates that the W4f spectrum include not only W6<sup>+</sup> and W5<sup>+</sup> but also W4<sup>+</sup> and W<sup>0</sup> in the ZnO–WO3 composite nanorods annealed at 500 ◦C. The subpeaks located at 33.4 eV and 35.5 eV are ascribed to tetravalent bond of tungsten and that at 31.1 eV was associated with the contribution of metallic tungsten [21,22]. The existence of mixed binding states of tungsten implied that substantial deoxidization of the WO3 shell layer occurred during the high-temperature annealing in the hydrogen-contained atmosphere. The appearance of metallic W component revealed that the WO3 is not in a pure oxide phase and this might deteriorate the gas-sensing performance of the n-type WO3 oxide. A similar deoxidization of metal oxides annealed in the hydrogen-contained atmosphere at a high temperature has been proposed and resultant deteriorated electric properties of the metal oxides are involved [20,23]. The reproducibility and stability of the sensor made from the ZnO–WO3 nanorods annealed at 400 ◦C were further examined at its optimum operating temperature of 300 ◦C to 50 ppm ethanol vapor concentration (Figure 6h). The di fference obtained in the gas response values of the sensor after cycling tests was very small and hence negligible. This suggests that the proposed composite nanorod sensor showed good reproducibility to detect ethanol vapor. Figure 6i shows the selectivity of the gas sensor based on the ZnO–WO3 nanorods annealed at 400 ◦C. The sensor was exposed to ammonia gas, ethanol vapor, nitrogen dioxide gas, and hydrogen gas of the appropriate concentrations at 300 ◦C, respectively. It can be seen that the sensor exhibited the substantially highest response to ethanol vapor, revealing its suitability for detecting ethanol vapor in the test environment. Table 1 compares the gas-sensing responses of various ZnO-based composites exposed to appropriate ethanol vapor concentrations at 300 ◦C [24–26]. The ZnO–WO3 composite nanorods annealed at 400 ◦C in this study presented superior ethanol vapor detecting ability among the various reference works.

**Figure 6.** (**a**) Gas-sensing response vs. temperature curves of various nanorod sensors on exposure to 50 ppm ethanol vapor. Dynamic gas-sensing response curves of sensor on exposure to 25–500 ppm ethanol vapor: (**b**) ZnO nanorods, (**c**) ZnO–WO3 composite nanorods, (**d**), ZnO–WO3 composite nanorods annealed at 400 ◦C, (**e**) ZnO–WO3 composite nanorods annealed 500 ◦C. (**f**) Gas-sensing response vs. ethanol vapor concentration curves. (**g**) XPS W 4f spectrum of the composite nanorods annealed at 500 ◦C. (**h**) Cycling testes of the composite nanorods annealed at 400 ◦C exposed to 50 ppm ethanol vapor. (**i**) Selectivity tests of the composite nanorods annealed at 400 ◦C.

**Table 1.** Comparison of ethanol gas-sensing responses of various ZnO-based heterogeneous sensors at 300 ◦C [24–26].


## **4. Discussion**

The possible reasons caused various gas-sensing responses of the ZnO nanorods and various composite nanorods were further explained with the space-charge layer model [1,27]. In ambient air, oxygen molecules were absorbed on surfaces of the composite nanorods. These oxygen molecules become surface absorbed oxygen species (such as <sup>O</sup><sup>−</sup>(ads) and <sup>O</sup>2−(ads)) by capturing free electrons from the conducting bands of the oxides at the elevated sensor operating temperature of 300 ◦C. The reactions are described as follows:

$$\text{O}\_2\text{ (ambient)} \rightarrow \text{O}\_{2(ads)}\text{ (n-type oxides)}\tag{1}$$

$$\text{2O}\_{2\text{ (ads)}} + 2\text{e}^- \rightarrow 2\text{O}^-\_{\text{(ads)}} \tag{2}$$

$$\text{O}^{-}\text{ (ads)} + \text{e}^{-} \rightarrow \text{O}^{2-}\text{ (ads)}\tag{3}$$

In this process, an electron depletion layer will be formed on the surfaces of the composite nanorods, resulting in a decrease of carrier concentration and an increase of sensor resistance. When the oxide nanorod sensor was exposed to ethanol vapor, the absorbed oxygen species will react with ethanol molecules according to the following possible reactions:

$$\rm C\_2H\_5OH + 6O^- \rightarrow 2CO\_2 + 3H\_2O + 6e^- \tag{4}$$

$$\rm C\_2H\_5OH + 6O^{2-} \rightarrow 2CO\_2 + 3H\_2O + 12e^- \tag{5}$$

As a result, the electrons trapped in the oxygen species are released back into the conduction band, leading to a decrease of the thickness of the depletion layer and the resistance of the oxides. In addition to the surface depletion layer, the contact of the di fferent oxides in a one-dimensional heterostructure system engenders formation of interfacial depletion regions because of di fferent work functions of the adjacent oxides. A proposed energy band structure diagram of the ZnO/WO3 and ZnO/ZnWO4 heterojunction herein are shown in Figure 7a [28,29]. The electrons will flow from ZnO nanorod to outer WO3 (or ZnWO4) crystals in the ZnO/WO3 (or ZnO/ZnWO4) heterostructures until their Fermi levels are equalized. Therefore, the exposure of the composite nanorods in air ambient result in formation of surface depletion regions in the surface WO3 (or ZnWO4) crystals and interfacial depletion regions inside the ZnO nanorod for the proposed various composite nanorods. This process creates an electron depletion layer on the surface of the ZnO core material and further bended the energy band and lead to a higher resistance of the composite nanorods with and without thermal annealing. The formation of additional interfacial depletion regions increased the potential barrier number in the composite nanorods than in the pristine ZnO nanorods; therefore, a larger resistance variation degree of the composite nanorods than that of the ZnO nanorods on exposure to the ethanol vapor is observed. This explained that the higher ethanol vapor sensing responses of the composite nanorods than those of the ZnO nanorods under the given gas-sensing tests herein. Similarly, a heterogeneous structure that improved the gas-sensing behavior of one-dimensional n-type oxide nanorods was demonstrated in ZnO–SnO2, ZnO–Zn2SnO4, and TiO2–CdO on exposure to test gases [16,19]. Comparatively, the ZnO–WO3 nanorods with a thermal annealing at 400 ◦C substantially enhanced their ethanol gas-sensing responses at the given test conditions. This improvement of the gas-sensing response of the composite nanorods annealed at 400 ◦C might be attributed to the random thickness of the depletion layers of the surface and interface regions of the composite nanorods resulting from the local phase transformation of the ZnO–WO3 composite system during the annealing process herein (Figure 7b). Compared to the chemically homogeneous shell layer of ZnO–WO3 nanorods, the ZnO–WO3 nanorods with a thermal annealing at 400 ◦C had a composite shell layer structure consisted of crystalline ZnWO4 and deteriorated WO3 as revealed in the structural analysis results. It is expected that the potential barrier number of the ZnO–WO3 composite nanorods annealed at 400 ◦C was higher than that of the composite nanorods without a thermal annealing. Three types of depletion regions included surface depletions in the deteriorated WO3 and crystalline ZnWO4, depletion regions at WO3/ZnWO4 boundaries of the shell layer and interfacial depletion regions at the ZnO/WO3 and ZnO/ZnWO4 are expected to exist in the ZnO–WO3 composite nanorods annealed at 400 ◦C. The local phase transformation of the ZnO/WO3 after thermal annealing at 400 ◦C for the ZnO–WO3 composite nanorod system in this study created a higher number of the potential barriers in the composite system as exhibited in Figure 7b. An increased potential barrier number in the composite systems has shown a substantial drop degree of the sensor resistance on exposure to the reducing gases and therefore this resulted in an enhanced gas-sensing response [3,9]. The cross-sectional potential barrier height variation alone the guided arrow red line in Figure 7b demonstrated that a more complex potential barrier height variation before and after introducing the ethanol vapor will be expected for the ZnO–WO3 composite nanorods with an annealing procedure. Comparatively, further introducing the reducing vapor of ethanol into the test chamber, the injection of electrons from the adsorbed oxygen ions into the conduction bands of the constituent oxides of the composite nanorods caused a larger degree of resistance variation of the composite nanorods with thermal annealing because of their diverse microstructures in the composite system. The substantial microstructural differences in the ZnO–WO3 composite nanorods with and without thermal annealing at 400 ◦C herein supported the different ethanol vapor sensing responses of various composite nanorods.

**Figure 7.** (**a**) Energy band alignments of ZnO/WO3 and ZnO/ZnWO4. (**b**) Possible ethanol gas-sensing mechanisms of the ZnO–WO3 composite nanorods with and without a thermal annealing at 400 ◦C. The corresponding cross-sectional views for the possible potential barrier height variation was also shown in the bottom region of the plot.
