Porous Au-embedded WO3 Nanowire Structure for Efficient Detection of CH4 and H2S

We developed a facile method to fabricate highly porous Au-embedded WO3 nanowire structures for efficient sensing of CH4 and H2S gases. Highly porous single-wall carbon nanotubes were used as template to fabricate WO3 nanowire structures with high porosity. Gold nanoparticles were decorated on the tungsten nanowires by dipping in HAuCl4 solution, followed by oxidation. The surface morphology, structure, and electrical properties of the fabricated WO3 and Au-embedded WO3 nanowire structures were examined by scanning electron microscopy, transmission electron microscopy, X-ray diffraction, X-ray photoelectron spectroscopy, and current–voltage measurements. Formation of a nanowire structure resulted in significant enhancement in sensing response to H2S and CH4 gases. Furthermore, Au embedment into the WO3 nanowire structures remarkably improved the performance of the sensors. The increase in response performance of sensors and adsorption–desorption kinetic processes on the sensing layers were discussed in relation with the role of Au embedment.

improvement in sensing performance towards CH 4 and H 2 S gases. In particular, WO 3 -based materials for CH 4 sensors have not been reported. However, the additional embedment of Au further enhanced the sensing performance. We found that the Au-embedded WO 3 sensor showed high response to CH 4 . Embedding of Au nanoparticles on WO 3 surface also showed interesting response properties toward H 2 S gas. The adsorption-desorption kinetic processes on the surface of Au nanoparticles and WO 3 nanowires surface were also discussed.

Results
Structural Properties. Porous nanowires of tungsten were obtained by sputter deposition of tungsten metal on porous SWCNTs template (Fig. 1a). The characteristic nature of the open spaces of the SWCNT template was stable during the metal sputter deposition and the following oxidation processes, as previously reported 12 . The morphology was negligibly changed after immersion in Au-containing solution ( Fig. 1b and Fig. S3). As seen in Fig. 1c, WO 3 nanowire structure was formed by interconnection of WO 3 grains during the oxidation process at 700 °C in air. Gold atoms on the surface also agglomerated to Au nanoparticles during the oxidation process. The distribution of Au nanoparticles was non-continuous and scattered on the surface, as revealed by the low-magnification TEM images in Fig. 2a,b. The diameter of the gold nanoparticles ranges from 8 nm to 20 nm and the average diameter of WO 3 nanowires is ~120 nm. The HR-TEM images revealed a high crystallinity of the synthesized WO 3 3 nanowire. The images were obtained by STEM-energy-dispersive X-ray (EDX) spectroscopy, which suggest that gold atoms not only form nanoparticles on the WO 3 nanowire surface, but also diffuse inside of the WO 3 crystalline lattice.
To investigate the surface composition and chemical states of the elements in the sample, XPS spectra of the nanowire structures of WO 3 and Au50-WO 3 were obtained. The spectrum in Fig. 4a indicates that the surface of the samples is composed of tungsten and oxygen elements with carbon contamination. In addition, Au distribution after annealing of Au-embedded WO 3 sample is evident. The peak at 285.7 eV is attributed to CO 2 , which is commonly adsorbed on the surface of the samples 13 . The high-resolution spectra of W4f (Fig. 4b) showed peaks at 35.85 and 38.02 eV, which are assigned to W4f 7/2 and W4f 5/2 , respectively. These peaks corresponded to oxidation state + 6 of tungsten atoms (WO 3 ) 14,15 . The shift of these peaks toward lower binding energy in Au50-WO 3 can be attributed to photoelectrons emitted from the lower oxidation states of tungsten (sub-stoichiometric WO 3−x ) 14 . The Au4f core level spectrum (Fig. 4c) recorded from the Au50-WO 3 sample showed peaks at 83.6 and 87.3 eV, which were assigned to Au4f 7/2 and Au4f 5/2 , respectively. These data confirmed that the Au nanoparticles are in metallic state 15 . The O1s peak in Fig. 4d showed a shift to lower binding energy with Au-embedment suggesting increased defective or incomplete W-O binding via Au addition. The O1s peak could be resolved to two Gaussian function peaks (1) and (2) having binding energies of 531 and 531.8 eV for WO 3 nanowire structure, and 530.7 and 532.2 eV for Au50-WO 3 nanowires, respectively. Peak (1) may be related to oxygen in the lattice and O 2− ions, and peak (2) might be due to the ions of O 2− , O − , and OH − in the oxygen-deficient regions 16 . A shift of the peak (1) to lower binding energy and a shift of the peak (2) to higher binding energy were observed with Au embedment. The shift of the major peak (1) to lower binding energy with Au embedment may be attributed to the defective WO 3 or sub-stoichiometry of WO 3−x , whereas the shift of peak (2) may suggest an increased binding energy of the ionosorbed oxygen via Au addition. Details of the causes for the latter observation or the stronger binding of the oxygen ions on the surface are unknown, but it is consistent with the delayed surface reaction kinetics in Au50-WO 3 , which will be discussed. DC conductivity measurements. The current-voltage (I-V) characteristics of the fabricated nanowires devices, as fundamental properties of electronic materials, were measured with varying temperatures in argon and dry air. Prior to the I-V measurements, the devices were heated to 350 °C in the given ambient conditions to remove the effect of adsorbed water molecules on the surface of the nanowires. The I-V curves of both WO 3 and Au50-WO 3 are linear, as shown in Fig. 5a-d, indicating the formation of ohmic contacts between the nanowires and Au electrodes. The resistances of nanowires calculated from the I-V measurements for pure WO 3 , Au10-WO 3 , and Au50-WO 3 nanowire structures are shown in Fig. 5e. The measurements in Ar ambient condition are not affected by oxygen ionosorption effect; results revealed monotonic decrease of the resistance with increasing temperature for all sensors. This relationship indicates a dominant semiconducting behavior. Therefore, the metallic nature of Au did not alter the semiconducting resistance-temperature behavior of the WO 3 nanowires probably because of the discreteness of the distribution of Au particles on the surface (Fig. 2). Nevertheless, a finite Au-embedment effect was observed, in which the resistance decreased with increasing content of embedded Au.
The increase in conduction with Au embedment can be explained by the Au-doping effect in the WO 3 lattice. This possibility could be postulated based on the EDS data in Fig. 2, which shows a uniform distribution of Au solid through the WO 3 nanowire body. Although most of the Au atoms agglomerated on the surface as nano-crystallites, some atoms are incorporated into the WO 3 nanowires during the annealing process and might form small Au clusters in theWO 3 lattice. If the conductivity of a transition metal oxide is established by hopping of electrons via dopant impurity levels, then the resistance of the oxide can be expressed by the following equation 17,18 : where R(T) is the resistance of a semiconductor at the absolute temperature T, ΔE a is the thermal activation energy for hopping, and k B is the Boltzmann constant. The temperature-dependent resistances of the pure WO 3 and Au-embedded WO 3 nanowire structures are shown in Fig. 5f, which were obtained by using the logarithm of equation (  due to creation of additional impurities in the lattice. The distribution of Au clusters in the WO 3 lattice could somehow increase the oxygen vacancies. Some Au atoms may have occupied the lattice sites to play directly as a dopant. Au atoms can create impurities in the bandgap, but the observation is in agreement with the cases of In-doped CdS 19 and Al-and Sb-doped CdTe 20 , which showed a decrease in activation energy with increasing dopant concentrations. In summary, the increase in conductance with Au embedment was analyzed, but a concrete conclusion cannot be elucidated at present and separate systematic study is necessary.
Interestingly, Fig. 5e shows that the temperature-dependent resistance behaviors of the Au-embedded WO 3 nanowires in air deviated from the semiconducting behavior at temperatures >150 °C; the resistances increased with increasing temperature. The increase in resistance is due to an enhanced oxygen ionosorption rate and corresponding increase of surface depletion depth of WO 3 nanowires. This phenomenon is the catalytic effect of the embedded Au nanoparticles because the pure WO 3 did not show such deviation even in air environment. Activation of oxygen adsorption on the surface of tin oxide decorated with various novel metals has been observed previously 21 . We found that the catalytic effect of Au critically enhances the dissociation of oxygen molecules at >150 °C, at which the adsorption ionic form of oxygen changes from O 2 − to O −22 .
Gas sensing properties. CH 4 gas sensing. The development of resistive sensors based on metal oxide semiconductors for CH 4 detection has been given much research interests in recent years. Among the metal oxides, tin oxide (SnO 2 ) semiconductors have attracted particular attention 2 , but sensing by WO 3 have been seldom reported. The CH 4 gas sensing properties of pure WO 3 and Au-embedded WO 3 nanowire sensors were examined and compared. The sensors exhibited a typical gas-sensing behavior of an n-type semiconductor, as shown by the decrease in sensor resistance with exposure to the reducing gas CH 4 . The response and recovery behaviours of the Au50-WO 3 nanowire sensor upon exposure to 100 ppm CH 4 diluted in dry air was measured at different operating temperatures of 200 °C, 250 °C, and 300 °C, as shown in Fig. 6a. The response behaviours of WO 3 , Au10-WO 3 , and Au50-WO 3 sensors measured at the optimum working temperature of 250 °C are compared in Fig. 6b. The trends of temperature dependence of the sensor signal were the same in all the sensors, showing the highest response at 250 °C.
The increasing Au content revealed increasing response signals, as shown in Fig. 6c. The response with pure WO 3 nanowire sensor was very low, which may be the reason for the limited reports on CH 4 gas sensing by WO 3 -based sensor. The response enhancement was from 3% with pure WO 3 to 37% with Au50-WO 3 sensor at 100 ppm CH 4 and 250 °C. The sensing behaviors of Au50-WO 3 sensor for different CH 4 gas concentrations measured at 250 °C are shown in Fig. 6d. The concentration dependence of the responses for Au10-WO 3 and Au50-WO 3 are shown as the inset. Notably, the Au50-WO 3 sensor showed a reduced operating temperature in comparison with the SnO 2 -based methane sensors, as shown in our previous report 23 .
The enhanced response with Au embedment is ascribed to the catalytic effect of Au on the sensing process. We described the gas adsorption and desorption kinetics in relation to gas sensing of semiconducting WO 3 nanowires, and found that the response level and the gas selectivity of an oxide sensor is derived from the ratio of the forward (adsorption) reaction rate over the reverse (desorption) reaction rate of the relevant gas species on the oxide surface 24,25 . The higher response in the Au-embedded structure indicates that the reaction rate of CH 4 with ionosorbed oxygen was enhanced with respect to the recovery rate, which is determined by the oxygen adsorption rate in air. We observed that Au has a catalytic effect on oxygen ionosorption and enhances the adsorption rate at >150 °C, which will enhance the recovery rate above. However, given that the enhanced response with Au-embedded sensor dictates far higher enhancement in the reaction CH 4 + 4O − → CO 2 + 2H 2 O+4e − , gold has to further accelerate the CH 4 decomposition reaction. For example, for methane dissociated to a methyl group and a hydrogen adatom on Au surface 26 , the dissociated molecular species can more actively react with the adsorbed atomic oxygen than the covalently bonded CH 4 molecules. In conclusion, Au showed catalytic effects on CH 4 dissociation and oxygen adsorption, but the rate-enhancement level was greater with CH 4 than with O 2 . The net consequence was the enhanced sensor response for CH 4 with Au-embedded WO 3 sensor structures.
We also reported that a rush of reducing gas, such as H 2 and NH 3 , on oxygen-ionosorbed oxide surface can produce an instant pile up of H 2 O molecules on the surface, leading to an overshoot in the response cycle curves 24,25 . The similar overshoots in the response cycles were observed with CH 4 on Au50-WO 3 sensor at 250-300 °C, as shown in Fig. 6a. We believe the overshoots were also derived from the combined effect of the enhanced reaction rate at high temperature and the high impingement rate of C, H, and other fractions of CH 4 catalytically dissociated by Au.
H 2 S gas sensing. The operation temperature effect of H 2 S gas sensing was examined for pure WO 3 and Au50-WO 3 . Fig. 7a shows the response-and-recovery curves of Au50-WO 3 nanowire sensor upon exposure to 10 ppm H 2 S at different operating temperatures (171-362 °C). The response levels are summarized in Fig. 7b. Remarkable improvement in the responses by Au-embedment was observed at all the operation temperatures up to 362 °C. The sudden increase of the response times at 362 °C in both sensors was interesting (Fig. 8a). This phenomenon may be due to the catalytic decomposition of H 2 S occurring in temperatures between 300 °C and 400 °C 27,28 . Thus, the species binding on the sensing layer changes at ~300 °C. Therefore, the operating temperature was set at 291 °C for the sensing of H 2 S to distinguish it from the different sensing mechanisms at > 300 °C. The response behaviors of Au50-WO 3 and WO 3 nanowire sensors to H 2 S were measured at various concentrations of 5, 10, 25, 50, and 100 ppm diluted in dry air at operating temperature of 291 °C, as shown in Figs. 7c, d. The responses of these sensors at different gas concentrations were also summarized in Fig. 7d, as shown in the inset. A response of 700 was obtained with the Au50-WO 3 nanowire sensor at 100 ppm H 2 S (Fig. 7c), which is much higher than the response of 100 obtained from the pure WO 3 nanowire sensor (Fig. 7d). This result can definitely be explained by the catalytic effect of gold. The catalytically fractionized H 2 S species could more actively react with ionosorbed oxygen. The WO 3 and Au50-WO 3 nanowire sensors also showed excellent selectivity to H 2 S, as is evident from Fig. S4. The response to H 2 S was more than two orders of magnitude higher than the response to other interfering gases, including H 2 , NH 3 , CO and CH 4 the same concentration. The selectivity towards H 2 S was further enhanced by the Au embedding (Fig. S4).
However, the Au50-WO 3 nanowires showed much slower response and recovery rates compared with the pure WO 3 nanowires through the whole working temperatures < 300 °C and the tested H 2 S gas concentrations, as summarized in Fig. 8. Herein, the response and recovery times were measured assuming exponential rise and decay of the curves based on the first-order surface reaction kinetics for adsorption and desorption 24,25 . Therefore, they are the characteristic average times of the processes and are the times required for completion of approximately 63% (1-1/e) of the response and recovery processes. The decrease in response and recovery times with increasing working temperatures (Figs. 8a,b) is basically due to the enhanced surface chemical reaction rates in the response and recovery cycles. An exception is the abnormal increase of the response time at 362 °C with both pure WO 3 and Au50-WO 3 nanowire structures caused by the change in the formula of the species adsorbing on the Au and WO 3 surfaces. Its effect on WO 3 conductance will be discussed.
The response and recovery times measured with varying H 2 S concentrations are shown in Figs. 8c,d. The decreasing response time with increasing gas concentrations observed with the pure WO 3 nanowires is the general trend in the sensing of reducing gases when the sensing behavior is based on a simple chemical reaction on the surface 24,25 . However, interestingly, the response time abnormally changes with H 2 S concentration for the Au-embedded sample, as shown in Fig. 8c. The result reveals a difference in H 2 S gas-sensing mechanism between pure WO 3 and Au-embedded WO 3 structures.
Generally, the response levels were enhanced but the response-recovery kinetics slowed down with Au-embedment. The improved gas-sensing response levels were definitely assisted by the Au nanoparticles. As we systematically examined in Refs. 25 and 29, the response and recovery processes of the WO 3 sensing layer towards reducing gases are the results of thermally activated chemical reaction processes on the surface. H 2 S gas molecules continuously react with the pre-absorbed oxygen ions (say O − ) 22 via and consequent reduction of W 6+ to W 4+ to release electrons into the WO 3 nanowires was also proposed to be the reason for the response. However, a study of the detailed chemical route for sensing is beyond the scope of the present research. Therefore, the enhanced response in Au50-WO 3 nanowire sensor with respect to the pure WO 3 (Fig. 7b) can be attributed to the highly catalytic dissociation of H 2 S by Au nanoparticles. As can be recall  Fig. 5c, the Au nanoparticles also enhanced the dissociation rate for oxygen adsorption. Therefore, the higher response observed with Au nanoparticles than with the pure WO 3 could be derived from the increase in desorption rate of the adsorbed oxygen ions in equation (2), which was propelled by both of the enhanced dissociation reactions of H 2 S and O 2 .
A probable explanation for the enhanced catalytic role of Au in increasing the response of the sensor may be derived from the literature. H 2 S is known to adsorb strongly onto Au because of the high chemical affinity of S on Au 30  The binding of S on Au can lower the surface work function of Au as much as 1 eV 32 , so this lowering of Au work function would decrease the band bending at the interface of Au and WO 3 nanowires, leading to a further resistance decrease and an improved response in the Au-embedded WO 3 sensor. Both effects enhance the sensor response and explain the Au catalytic effect. Now, the issues that need to be addressed are as follows: (i) the longer response and recovery times of Au-embedded WO 3 than WO 3 (Fig. 8), (ii) the abnormal increase of the response time at 362 °C (Fig. 8a), and (iii) the sudden rise of response time for higher H 2 S concentration in Au-embedded WO 3 (Fig. 8c). The slower response and recovery kinetics with Au-embedment should be caused by some delayed reactions in relation to Au nanoparticles. A simple reaction of (2) may explain the sensing with pure WO 3 . The decreasing response time as the H 2 S concentration increased in Fig. 8c for WO 3 reflects the concentration-dependent surface reaction rate (or 1/τ ∝ C H2S ), which is observed when the surface reaction kinetics controls the sensing 24 may occur with Au at relatively low temperatures. The locally generated H 2 gas molecules are added to the WO 3 surface for additional reactions. While this added surface reaction increased the sensor response level, the dissociation reaction on Au and/or the successive reaction on WO 3 surfaces delayed the overall response and recovery processes. As the H 2 S concentration increased, further delayed responses were observed due to the absolute increase of the reactions, but the contribution of the reactions occurring directly on the WO 3 surface [or equation (2)] became relatively dominant at far higher H 2 S concentrations, resulting in reduced response time (Fig. 8c).
For the recoveries, the simple re-adsorption of oxygen atoms on the WO 3 surface explains the recovery cycle for WO 3 . However, for Au-embedded WO 3 , the surfaces not only of WO 3 , but also of Au particles, need to be recovered to the standby condition. In other words, chemical reactions, such as 2SH (ads) + O 2 → H 2 S (g) + SO 2(g) and/or S (ads) + O 2 → SO 2(g) , are required to recover to the original Au surface condition. The reactions themselves require time and may be slow, and also can influence the processes occurring on nearby WO 3 surface and finally determine the response time. For example, the produced H 2 S gas from the Au surface can impinge on the WO 3 surface to play the role of analyte gas molecule and hinder the oxygen ionosorption. The delayed response and recovery rates are clear for different working temperatures and H 2 S gas concentrations, as shown in Fig. 8.
Another observation is that the further slowing down of the response rate at higher temperature of 362 °C was observed in both pure WO 3 and Au-WO 3 sensor structures (Fig. 8a). Therefore, it is a temperature-related reaction kinetics that occurs on WO 3 . A new reaction causing the apparently slower kinetics suddenly emerged at the above-mentioned temperature. The new reaction could be caused by the sudden change in (i) the formula of analyte gas provided and/or (ii) the ionosorbed oxygen formulas, resulting in a change in the reaction equation on the WO 3 surface. If the impinging H 2 S gas molecules directly decompose to S and H 2 and adsorb on the WO 3 surface via equation (3), the high temperature further accelerates the decomposition reaction. The delayed escape of the highly increased product gas of H 2 O out of the surface could increase the apparent response time. Otherwise, the direct decomposition of H 2 S on the WO 3 22 . The binding energy of O 2− is expected to be greater than that of O -; thus, the desorption of the former is more difficult. This situation led to the longer response time.

Conclusions
Highly porous Au-embedded WO 3 nanowire structures with diameters of ~120 nm were fabricated by tungsten deposition on porous SWCNT template, followed by dipping in HAuCl 4 solution and oxidation. The Au atoms coated on the W nanowire surface agglomerated to form Au crystallites on the surface and in the WO 3 lattice. The effects of Au, particularly of the embedded Au crystallites on the surface, on the gas-sensing properties were investigated based on the surface reactions. A comparison of the foregoing findings with the results obtained from pure WO 3 nanowires is useful to discuss the Au embedment effect. For CH 4 gas-sensing performance, the WO 3 nanowire structure showed a significant lowering of working temperature with respect to SnO 2 structure, whereas the Au-embedded structure demonstrated enhanced gas-sensing response because of improved dissociation of CH 4 gas molecules. A different sensing performance of Au-embedded WO 3 nanowires was observed for H 2 S gas. Although the gas-sensing signals were similarly improved, the response and recovery kinetics were slowed down by Au. Such complication seems to have originated from the diverse reaction routes of H 2 S gas on the Au and WO 3 surfaces. The intermediate gas products on the surface may be impinged on the surface and play the role of analyte gas (such as H 2 ) and/or temporarily adsorbed on the surface and show delayed detachment (such as H 2 O). The Au particles embedded on the surface catalyzed the dissociation reactions of O 2 , CH 4 , and H 2 S gases, thereby enhancing the oxygen ionosorption reaction and the gas-sensing reactions.

Methods
Fabrication of sensor structures. The Au patterned substrates were installed on the inside wall of the arc-discharge chamber; the substrate were then coated with SWCNTs, as described in Ref. 12  Synthesis was performed at an arc current density of 40 A/cm 2 in H 2 gas at 400 Torr for 4 min. The carbon source used was a graphite rod that contained catalyst wires of iron, nickel, and molybdenum 12,33,34 . The SWCNT template substrates were heat-treated at 400 °C in air for 2 h to remove amorphous carbon. The SWCNT bundles seated on the substrate revealed large open spaces among the SWCNTs because of steric hindrance 12 . Tungsten metal layers were deposited on the SWCNT templates using a DC magnetron sputtering system. The deposition was performed at room temperature under a constant Ar pressure of 3.8 × 10 −2 Torr and input power of 100 W for 120 s. During deposition, the substrates were rotated to attain uniform thickness. The conformal deposition of W along the porous SWCNT bundles also revealed highly porous W/SWCNT composite nanowires. Gold salt solutions were prepared by dissolving different amounts of HAuCl 4 salt in the mixture of distilled water and ethanol at 1:3 ratio. The W/ SWCNT samples were dip-coated with Au by immersing in gold salt solutions of 0, 10, and 50 mM of HAuCl 4 ; the samples were identified as WO 3 , Au10-WO 3 , and Au50-WO 3 , respectively. Oxidation of the samples at 700 °C for 2 h resulted in conversion of W to WO 3 and formation of Au nanoparticles while burning out the SWCNTs, leaving a highly porous tungsten oxide nanowire network. Fig.S1 shows the flowchart of the fabrication process for open-space ensemble Au-embedded WO 3 nanowires structure using a highly porous SWCNT template.
Characterization. Surface morphology of the nanowire structures was investigated by field emission scanning electron microscopy (FESEM; JEOL, JSM-700F) and transmission electron microscopy (TEM), scanning TEM (STEM), and high-resolution TEM (HR-TEM). The structural properties were investigated by X-ray diffraction (XRD; Rigaku D/MAX-RC) using Cu Kα radiation with a Ni filter. The electronic structure of the surface of samples was elucidated by X-ray photoelectron spectroscopy (XPS; VGMultilab 2000; Thermo VG Scientific, UK).
Gas sensing property measurement. Gas sensing properties were also measured using 6487 Keithley in a chamber with controls for temperature and gas flow 12,35 . The response of a sensor, S, was defined by R i /R f or (R i -R f )/R f for reducing gases, where R i is the standby resistance in air, and R f is the resistance upon exposure to reducing gas. We used 1000 ppm CH 4 and H 2 S gases diluted in nitrogen as the source gas, which was further diluted in dry air at varying concentrations in the ppm range. The structure of gas sensor is shown in Figs. S2a-c. The setup for the gas sensing measurements is shown in Fig. S2d.