Preparation and Hydrogen-Sensitive Property of WO 3 /Graphene/Pd Ternary Composite

: Hydrogen (H 2 ) is a renewable energy source that has the potential to reduce greenhouse gas emissions. However, H 2 is also highly ﬂammable and explosive, requiring sensitive and safe sensors for its detection. This work presents the synthesis and characterization of WO 3 /graphene binary and WO 3 /graphene/Pd (WG-Pd) ternary nanocomposites with varying graphene and Pd contents using the microwave-assisted hydrothermal method. The excellent catalytic efﬁcacy of Pd nanoparticles facilitated the disintegration of hydrogen molecules into hydrogen atoms with heightened activity, consequently improving the gas-sensing properties of the material. Furthermore, the incorporation of graphene, possessing high conductivity, serves to augment the mobility of charge carriers within the ternary materials, thereby expediting the response/recovery rates of gas sensors. Both graphene and Pd nanoparticles, with work functions distinct from WO 3 , engender the formation of a heterojunction at the interface of these diverse materials. This enhances the efﬁcacy of electron–hole pair separation and further ampliﬁes the gas-sensing performance of the ternary materials. Consequently, the WG-Pd based sensors exhibited the best gas-sensing performance when compared to anther materials, such as a wide range of hydrogen concentrations (0.05–4 vol.%), a short response time and a good selectivity below 100 ◦ C, even at room temperature. This result indicates that WG-Pd ternary materials are a promising room-temperature hydrogen-sensing materials for H 2 detection.


Introduction
Hydrogen, an eco-friendly and renewable energy source, has gained considerable interest in its wide range of applications, including automobile, fuel cell and space rocket applications [1,2]. However, hydrogen exhibits a colorless, odorless and explosive nature with a low ignition energy (0.02 mJ) and a wide flammable range of 4-74 vol.% in air, leading to many safety issues [3,4]. Therefore, rapid and accurate detection of hydrogen leakage is necessary for large-scale utilization of hydrogen. In recent years, chemiresistive hydrogen sensors based on metal oxide semiconductors (MOS), such as ZnO, SnO 2 , MoO 3 , WO 3 , etc., have been widely applied to hydrogen detection due to their low cost, high sensitivity and short recovery time [5][6][7][8]. Among these MOS, WO 3 , with its unique morphological structure, gasochromic properties, and high diffusion coefficient of oxygen vacancies, has proven to be one of the most attractive sensing materials [9]. However, the pure WO 3 sensors need a high working temperature (200-400 • C) to detect hydrogen, which may lead to power consumption and source of ignition [10,11].

Fabrication of WO 3 /Graphene/Pd Ternary Materials
A certain amount of GO was dispersed in 190 mL deionized water followed by ultrasonication. Next, 1.06 g of Na 2 WO 3 was added into GO dispersion, and 50 mL of HCl (2 M) was added drop-by-drop into the mixture under vigorous stirring for 30 min. The suspension was poured into autoclaves and subjected to microwave heating at 180 • C for 1 h using a Multiwave PRO oven (Anton Paar, Graz, Austrian). The precipitate was then collected and washed by centrifugation in deionized water and absolute ethanol, followed by freeze-drying under vacuum at room temperature. The WO 3 /graphene nanomaterials were synthesized with different GO/WO 3 mass ratios of 0.5, 1, 2 and 3:50, and denoted as W0.5G, WG, W2G and W3G, respectively. Pure WO 3 was also prepared in the same Chemosensors 2023, 11, 410 3 of 12 way without GO for comparison. A quantity of 50 mg of WG and a specific amount of PdCl 2 were dispersed in 5 mL of absolute ethanol using ultrasonication for a duration of 30 min. The resulting mixture was then dried in an oven at a temperature of 80 • C for a period of 12 h, followed by thermal annealing in a nitrogen atmosphere at a temperature of 300 • C for 2 h. The samples obtained from this process, with PdCl 2 masses of 1, 5 and 10 mg, respectively, were designated as WG-1Pd, WG-5Pd and WG-10Pd.

Characterization
Scanning electron microscopy (SEM, JEOL JSM-7610F, Tokyo, Japan) and transmission electron microscopy (TEM, FEI Talos, Columbia, SC, USA) were used to observe the morphologies of ternary nanomaterials. The phases of samples were analyzed by X-ray diffraction (XRD, Bruker D8 Advance, Billerica, MA, USA). The chemical compositions of ternary nanomaterials were measured using Thermo Fisher (Waltham, MA, USA) ESCALAB 250 XI X-ray photoelectron spectroscopy (XPS).

Fabrication and Test of Gas Sensors
The WG and WG-Pd powders were blended with terpineol in a 1:2 mass ratio, while finely grinding them in a mortar. The resultant paste was uniformly coated on an Al 2 O 3 tube with a pair of Pt wires. To heat the gas sensor, a Ni-Cr heating wire was inserted into the tube. The sensors were pre-treated at 100 • C for 5 days to enhance their stability before the tests. The gas-sensing measurements were carried out using a CGS-8 Gas Sensing Measurement System (Beijing Elite Tech Company Limited, Beijing, China) with a 500 mL test chamber. The sensors' resistance was stabilized at the desired temperature before a known volume of gas was injected into the chamber. The tests were performed under ambient conditions of 25 ± 5 • C and 40 ± 5% relative humidity.
The gas response was calculated as (R air − R gas )/R air , where R air and R gas are the sensor resistance in air and target gas, separately. The response time and recovery time are the time intervals required for the sensor response to reach 90% of its total change upon exposure to the target gas or air, respectively.

Morphology and Structure
In order to characterize the surface morphology and distribution of WO 3 , WG and WG-Pd, SEM and EDS were utilized. As shown in Figure 1a, WO 3 obtained by the microwave-assisted hydrothermal method showed a nanoribbon-like structure with a length of about 200 nm and a thickness of about 30 nm. The two-dimensional structure had a high surface-to-volume ratio and abundant surface functional groups, which offer a large number of adsorption sites for the target gas and oxygen molecules. This leads to effective physisorption and chemisorption of gas molecules at low temperature. Hence, two-dimensional materials have some advantages in providing fast and complete diffusion of H 2 gas throughout the structure and are widely used as gas-sensing materials [22][23][24]. Following the incorporation of GO, WG was interconnected by delicate folded RGO layers amid multiple WO 3 nanoribbons. The folded configuration of RGO primarily resulted from lamellar distortion induced by partial reduction of GO during the microwave-assisted hydrothermal process (Figure 1b-e) [25,26]. In the case of W0.5G with a lower graphene content, only a limited number of WO 3 nanoribbons were linked by graphene ( Figure S1a). Conversely, with a higher graphene content in W2G, a significantly larger proportion of WO 3 nanoribbons were connected by graphene, albeit some were enveloped by it ( Figure S1b). As the graphene content increased, porous structures gradually formed within the composite due to graphene stacking, enveloping the WO 3 nanoribbons and leaving only a few exposed ( Figure S1c). Subsequent to Pd loading onto the surface of WG, a substantial amount of Pd was uniformly distributed, as evidenced by the EDS image (Figure 1f-h). it ( Figure S1b). As the graphene content increased, porous structures gradually for within the composite due to graphene stacking, enveloping the WO3 nanoribbons leaving only a few exposed ( Figure S1c). Subsequent to Pd loading onto the surface of a substantial amount of Pd was uniformly distributed, as evidenced by the EDS im (Figure 1f-h). The hierarchical structures of WG and WG-Pd were characterized by TEM. As sh in Figure 2a-c, in addition to large nanoribbons, there were a few nanosheets of W around 10-20 nm of WG, which should be grown on the surface of graphene. There w a few nanosheets of WO3 of around 10-20 nm on the surface of WG, indicating that were grown on the graphene surface. After loading Pd onto WG, spherical Pd nanop cles appeared on the surface of graphene (Figure 2d-f). With a Pd content of 1%, Pd noparticles were sporadically distributed on graphene nanosheets with a size of 3 When increasing the Pd content, the number and size of Pd nanoparticles also increa When the Pd content increased to 10%, Pd nanoparticles were more densely distrib and had a larger size of 10 nm. Moreover, even for the WG-Pd with the highest conte Pd, no significant aggregation of Pd nanoparticles was observed. This phenome stemmed from the mutual interaction between the abundant oxygen-containing f tional groups on the surface of graphene oxide and Pd ions, which led to their adsorp and uniform growth on the surface of graphene nanoflakes and prevented the aggrega of Pd nanoparticles. This phenomenon can be attributed to the interaction between abundant oxygen-containing functional groups on the surface of graphene oxide and ions. This interaction leads to the adsorption and uniform growth of Pd nanoparticle the surface of graphene nanoflakes, preventing their aggregation. The hierarchical structures of WG and WG-Pd were characterized by TEM. As shown in Figure 2a-c, in addition to large nanoribbons, there were a few nanosheets of WO 3 around 10-20 nm of WG, which should be grown on the surface of graphene. There were a few nanosheets of WO 3 of around 10-20 nm on the surface of WG, indicating that they were grown on the graphene surface. After loading Pd onto WG, spherical Pd nanoparticles appeared on the surface of graphene (Figure 2d-f). With a Pd content of 1%, Pd nanoparticles were sporadically distributed on graphene nanosheets with a size of 3 nm. When increasing the Pd content, the number and size of Pd nanoparticles also increased. When the Pd content increased to 10%, Pd nanoparticles were more densely distributed and had a larger size of 10 nm. Moreover, even for the WG-Pd with the highest content of Pd, no significant aggregation of Pd nanoparticles was observed. This phenomenon stemmed from the mutual interaction between the abundant oxygen-containing functional groups on the surface of graphene oxide and Pd ions, which led to their adsorption and uniform growth on the surface of graphene nanoflakes and prevented the aggregation of Pd nanoparticles. This phenomenon can be attributed to the interaction between the abundant oxygen-containing functional groups on the surface of graphene oxide and Pd ions. This interaction leads to the adsorption and uniform growth of Pd nanoparticles on the surface of graphene nanoflakes, preventing their aggregation. Figure 3 shows XRD patterns of WO 3 and WG-Pd. WO 3 exhibited distinct diffraction peaks of orthorhombic phase of WO 3 ·0.33H 2 O (JCPDS 72-199) [27,28], indicating the presence of crystal H 2 O before thermal annealing. For the WO 3 /graphene composite nanomaterials, the characteristic peaks at 22.7 • , 28.1 • , 36.5 • , 49.9 • and 55.3 • were slightly different from those of pure-phase WO 3 . This was attributed to the interaction between the functional groups (hydroxyl, carbonyl and carboxyl groups) on the graphene surface and Na 2 WO 3 , which influenced the WO 3 growth. Moreover, slight deviation from stoichiome-try (WO 3−x ) could also affect the peak position. They also exhibited a broad characteristic peak around 20 • , which was caused by the imperfect restacking due to the curvature of graphene sheets [29] ( Figure S2). After loading with Pd and thermal annealing, WG-1Pd, WG-5Pd and WG-10Pd had different XRD patterns with WG. As shown in Figure 3b-d, there were diffraction features matching hexagonal phase of WO 3 due to the dehydration after annealed [30]. Furthermore, characteristic peaks of Pd (JCPDS 46-1043) and PdO (JCPDS 43-1024) appeared in WG-Pd, applying the presence of Pd and PdO attributed to the particle oxidization of Pd nanoparticles in air [31].   [27,28], indicating the presence of crystal H2O before thermal annealing. For the WO3/graphene composite nanomaterials, the characteristic peaks at 22.7°, 28.1°, 36.5°, 49.9° and 55.3° were slightly different from those of pure-phase WO3. This was attributed to the interaction between the functional groups (hydroxyl, carbonyl and carboxyl groups) on the graphene surface and Na2WO3, which influenced the WO3 growth. Moreover, slight deviation from stoichiometry (WO3−x) could also affect the peak position. They also exhibited a broad characteristic peak around 20°, which was caused by the imperfect restacking due to the curvature of graphene sheets [29] ( Figure S2). After loading with Pd and thermal annealing, WG-1Pd, WG-5Pd and WG-10Pd had different XRD patterns with WG. As shown in Figure 3c-e, there were diffraction features matching hexagonal phase of WO3 due to the dehydration after annealed [30]. Furthermore, characteristic peaks of Pd (JCPDS 46-1043) and PdO (JCPDS 43-1024) appeared in WG-Pd, applying the presence of Pd and PdO attributed to the particle oxidization of Pd nanoparticles in air [31].   Figure 3 shows XRD patterns of WO3 and WG-Pd. WO3 exhibited distinct diff peaks of orthorhombic phase of WO3·0.33H2O (JCPDS 72-199) [27,28], indicating th ence of crystal H2O before thermal annealing. For the WO3/graphene composite materials, the characteristic peaks at 22.7°, 28.1°, 36.5°, 49.9° and 55.3° were slightly ent from those of pure-phase WO3. This was attributed to the interaction between th tional groups (hydroxyl, carbonyl and carboxyl groups) on the graphene surfa Na2WO3, which influenced the WO3 growth. Moreover, slight deviation from stoich try (WO3−x) could also affect the peak position. They also exhibited a broad charac peak around 20°, which was caused by the imperfect restacking due to the curva graphene sheets [29] ( Figure S2). After loading with Pd and thermal annealing, W WG-5Pd and WG-10Pd had different XRD patterns with WG. As shown in Figur there were diffraction features matching hexagonal phase of WO3 due to the dehyd after annealed [30]. Furthermore, characteristic peaks of Pd (JCPDS 46-1043) an (JCPDS 43-1024) appeared in WG-Pd, applying the presence of Pd and PdO attribu the particle oxidization of Pd nanoparticles in air [31].  In order to accurately analyze the content and valence state of elements in WG-10Pd, XPS analysis was utilized ( Figure 4). As shown in Figure 4a, the material was mainly composed of W, C, O and Pd. Figure 4b demonstrates the deconvoluted Pd (3d) XPS spectra of WG-10Pd. It can be found that Pd element was mainly metallic Pd 0 located at 335.3 and 340.55 eV. Additionally, there were other peaks of oxidized Pd 2+ located at 336.6 and 341.85 eV and Pd 4+ located at 338.2 and 343.45 eV [32]. This also corresponds to the XRD spectra. A small amount of oxidized Pd came from the partial oxidation of surface atoms of Pd nanoparticles by air due to their high activity.
osensors 2023, 11, x FOR PEER REVIEW 6 of In order to accurately analyze the content and valence state of elements in WG-10P XPS analysis was utilized ( Figure 4). As shown in Figure 4a, the material was mainly com posed of W, C, O and Pd. Figure 4b demonstrates the deconvoluted Pd (3d) XPS spect of WG-10Pd. It can be found that Pd element was mainly metallic Pd 0 located at 335.3 an 340.55 eV. Additionally, there were other peaks of oxidized Pd 2+ located at 336.6 and 341.8 eV and Pd 4+ located at 338.2 and 343.45 eV [32]. This also corresponds to the XRD spectr A small amount of oxidized Pd came from the partial oxidation of surface atoms of P nanoparticles by air due to their high activity.

Hydrogen Gas Sensing
The effects of graphene content and working temperature on the hydrogen respons performance of WO3/graphene nanomaterials are shown in Figure 5. As can be seen fro Figure 5a, at 50 °C, WO3, W0.5G and W3G had no response to hydrogen. Among them WO3 and W0.5G had large resistances (7130 MΩ and 2500 MΩ, respectively) that exceede the instrument resistance range (500 MΩ) and could not be measured. This was becau WO3, as a metal oxide semiconductor, had a large resistance at low temperatures, an even with a small amount of graphene added, it still could not meet the measureme requirements. WG had the best response performance, but due to the low working tem perature, it only responded to hydrogen concentrations above 0.4%. Although W2G an W3G had higher graphene contents than WG, their response performance was reduce This was attributed to the excess graphene that could cover or even wrap around WO and provide an additional path for current flow, resulting in a significant increase of th conductivity of the sample, which made resistance variation in reducing environment le noticeable. As the working temperature increased to 100 °C and 150 °C, the response pe formance of WG, W2G and W3G were all improved. The response of WG, W2G and W3 to 1% H2 at 100 °C was 30.8, 28.9 and 2.5%, respectively. After heating at 150 °C, the co responding responses increased to 90.5, 56.3 and 7.9%, respectively. This result indicate that WG had the highest response performance. At the same time, as the working tempe ature increased from 50 °C to 150 °C, WG's response time to 1% hydrogen was shortene from 223 s to 67 s (Figure 5d-e). Meanwhile, WG showed p-n switching phenomenon different temperatures, which was mainly caused by the combination of different sem conductor materials and the change of charge transfer path at different working tempe atures. Graphene (p-type semiconductor) and WO3 (n-type semiconductor) formed composite, in which the carriers of both graphene and WO3 participated in the conductio

Hydrogen Gas Sensing
The effects of graphene content and working temperature on the hydrogen response performance of WO 3 /graphene nanomaterials are shown in Figure 5. As can be seen from Figure 5a, at 50 • C, WO 3 , W0.5G and W3G had no response to hydrogen. Among them, WO 3 and W0.5G had large resistances (7130 MΩ and 2500 MΩ, respectively) that exceeded the instrument resistance range (500 MΩ) and could not be measured. This was because WO 3 , as a metal oxide semiconductor, had a large resistance at low temperatures, and even with a small amount of graphene added, it still could not meet the measurement requirements. WG had the best response performance, but due to the low working temperature, it only responded to hydrogen concentrations above 0.4%. Although W2G and W3G had higher graphene contents than WG, their response performance was reduced. This was attributed to the excess graphene that could cover or even wrap around WO 3 and provide an additional path for current flow, resulting in a significant increase of the conductivity of the sample, which made resistance variation in reducing environment less noticeable. As the working temperature increased to 100 • C and 150 • C, the response performance of WG, W2G and W3G were all improved. The response of WG, W2G and W3G to 1% H 2 at 100 • C was 30.8, 28.9 and 2.5%, respectively. After heating at 150 • C, the corresponding responses increased to 90.5, 56.3 and 7.9%, respectively. This result indicated that WG had the highest response performance. At the same time, as the working temperature increased from 50 • C to 150 • C, WG's response time to 1% hydrogen was shortened from 223 s to 67 s (Figure 5d,e). Meanwhile, WG showed p-n switching phenomenon at different temperatures, which was mainly caused by the combination of different semiconductor materials and the change of charge transfer path at different working temperatures. Graphene (p-type semiconductor) and WO 3 (n-type semiconductor) formed a composite, in which the carriers of both graphene and WO 3 participated in the conduction path, showing a mixed p-n characteristic. At low temperature, WO 3 transferred electrons to GO, showing p-type behavior. At high temperature, WO 3 received electrons from GO, showing n-type behavior. However, WG still had no response to hydrogen at room temperature and needed to be composited with Pd to further improve its response performance. Therefore, we used WG as the base material and composite different contents of Pd nanoparticles to study the effects of Pd content and working temperature on the hydrogen response performance. perature and needed to be composited with Pd to further improve its response performance. Therefore, we used WG as the base material and composite different contents of Pd nanoparticles to study the effects of Pd content and working temperature on the hydrogen response performance. The effect of Pd content and working temperature was studied by testing these gas sensors at different temperature ranging from 25 to 100 °C as shown in Figure 6. For WG-1Pd and WG-5Pd, there was no evident response to 500 ppm of H2 at 25 °C, applying low content of Pd was not sufficient for H2 sensing at room temperature. However, WG-10Pd with larger content of Pd showed response to 500 ppm of H2 at 25 °C with value of 2.3%. WG-1Pd, WG-5Pd and WG-10Pd were more sensitive to H2 at higher working temperature. The response to 500 ppm H2 of WG-1Pd, WG-5Pd and WG-10Pd increased to 86.1%, 92.6% and 94.6% at a working temperature of 100 °C, respectively. This phenomenon stemmed from the lower gas activation energy of adsorption/desorption at higher temperature. Meanwhile, WG-10Pd still showed a higher response to H2 than the others. Compared with the WO3/graphene binary materials without Pd, the hydrogen-sensing performance of the WO3/graphene/Pd ternary composite improved significantly. WG exhibited no noticeable response to 500 ppm hydrogen, while the response value increased to 86.1% with the addition of 1% Pd nanoparticles, indicating that Pd had a remarkable effect on enhancing the hydrogen response. The effect of Pd content and working temperature was studied by testing these gas sensors at different temperature ranging from 25 to 100 • C as shown in Figure 6. For WG-1Pd and WG-5Pd, there was no evident response to 500 ppm of H 2 at 25 • C, applying low content of Pd was not sufficient for H 2 sensing at room temperature. However, WG-10Pd with larger content of Pd showed response to 500 ppm of H 2 at 25 • C with value of 2.3%. WG-1Pd, WG-5Pd and WG-10Pd were more sensitive to H 2 at higher working temperature. The response to 500 ppm H 2 of WG-1Pd, WG-5Pd and WG-10Pd increased to 86.1%, 92.6% and 94.6% at a working temperature of 100 • C, respectively. This phenomenon stemmed from the lower gas activation energy of adsorption/desorption at higher temperature. Meanwhile, WG-10Pd still showed a higher response to H 2 than the others. Compared with the WO 3 /graphene binary materials without Pd, the hydrogen-sensing performance of the WO 3 /graphene/Pd ternary composite improved significantly. WG exhibited no noticeable response to 500 ppm hydrogen, while the response value increased to 86.1% with the addition of 1% Pd nanoparticles, indicating that Pd had a remarkable effect on enhancing the hydrogen response.
The dynamic response of WG-10Pd was studied at various H 2 concentrations (ranging from 0.05-4%) and at temperatures of 25, 50, 75 and 100 • C. As shown in Figure 7a, WG-10Pd showed excellent response to hydrogen in the range of 0.05-4% at room temperature. The response value increased gradually with the increase of the concentration of H 2 . At 4% of H 2 , the response value reached 96.1%. On the contrary, the response time was gradually shortened. The response time was reduced to within 40 s at 4% of H 2 . After heating to 50, 75 and 100 • C, the response value was further improved and the response time was shorter. At 100 • C, the response value of 0.05% H 2 could reach 94.6%, and the response time was shortened to 6 s. However, the response curves reached a plateau at relatively low concentrations, proving its insensitivity to high concentration of H 2 . Therefore, WG-10Pd was more suitable for the measurement of low-concentration hydrogen below 0.05% after heating. The saturation phenomenon was ascribed to the increased catalytic effect of Pd at high temperature, which generated abundant chemisorbed water and occupied the active sites for oxygen adsorption. The dynamic response of WG-10Pd was studied at various H2 concentrations (ranging from 0.05-4%) and at temperatures of 25, 50, 75 and 100 °C. As shown in Figure 7a, WG-10Pd showed excellent response to hydrogen in the range of 0.05-4% at room temperature. The response value increased gradually with the increase of the concentration of H2. At 4% of H2, the response value reached 96.1%. On the contrary, the response time was gradually shortened. The response time was reduced to within 40 s at 4% of H2. After heating to 50, 75 and 100 °C, the response value was further improved and the response time was shorter. At 100 °C, the response value of 0.05% H2 could reach 94.6%, and the response time was shortened to 6 s. However, the response curves reached a plateau at relatively low concentrations, proving its insensitivity to high concentration of H2. Therefore, WG-10Pd was more suitable for the measurement of low-concentration hydrogen below 0.05% after heating. The saturation phenomenon was ascribed to the increased catalytic effect of Pd at high temperature, which generated abundant chemisorbed water and occupied the active sites for oxygen adsorption.  The dynamic response of WG-10Pd was studied at various H2 concentrations (ranging from 0.05-4%) and at temperatures of 25, 50, 75 and 100 °C. As shown in Figure 7a, WG-10Pd showed excellent response to hydrogen in the range of 0.05-4% at room temperature. The response value increased gradually with the increase of the concentration of H2. At 4% of H2, the response value reached 96.1%. On the contrary, the response time was gradually shortened. The response time was reduced to within 40 s at 4% of H2. After heating to 50, 75 and 100 °C, the response value was further improved and the response time was shorter. At 100 °C, the response value of 0.05% H2 could reach 94.6%, and the response time was shortened to 6 s. However, the response curves reached a plateau at relatively low concentrations, proving its insensitivity to high concentration of H2. Therefore, WG-10Pd was more suitable for the measurement of low-concentration hydrogen below 0.05% after heating. The saturation phenomenon was ascribed to the increased catalytic effect of Pd at high temperature, which generated abundant chemisorbed water and occupied the active sites for oxygen adsorption. WG-10Pd hydrogen-sensing material had excellent selectivity and stability. As demonstrated in Figure 7c, the response value of WG-10Pd to 1% H 2 at room temperature was significantly higher than that of CH 4 , CO and NH 3 with the same concentration, showing high selectivity of H 2 , due to the catalytic properties of Pd to hydrogen dissociation. After 30 days of uninterrupted testing, the change of response value was still maintained within 5%, implying the high stability of WG-10Pd.
The effect of relative humidity (RH) on the sensing performance of WG-10Pd at room temperature was also investigated. Figure 7e,f shows that the response of WG-10Pd decreased with increasing RH. This phenomenon was caused by the water vapor occupying the adsorption sites and hindering the oxygen adsorption. The response values of WG-10Pd to 1% H 2 were 77.8%, 73.1%, 69.4%, 65.9% and 60.6% under 24%, 43%, 55%, 71% and 89% RH, respectively. The response variation (calculated as 1 − Response (89RH%)/Response (24RH%)) was 22%, indicating that RH had a significant influence on the hydrogen sensing. Meanwhile, the resistance in air decreased from 42.5 to 35.4 MΩ as RH changed from 24% to 89%. Therefore, the resistance variation was about 16.7%.
Similarly, the response of WG-1Pd and WG-5Pd at 50-100 • C were studied (Figure 8). The variation trend of response value and response time was basically the same as that of WG-10Pd, but the response value was slightly reduced and the response time was extended. The responses to 0.05% H 2 at 100 • C of WG-1Pd and WG-5Pd were 86.1% and 92.6%, respectively, while the response times were 25 s and 9 s, respectively.
After 30 days of uninterrupted testing, the change of response value was still maintained within 5%, implying the high stability of WG-10Pd.
The effect of relative humidity (RH) on the sensing performance of WG-10Pd at room temperature was also investigated. Figure 7e,f shows that the response of WG-10Pd decreased with increasing RH. This phenomenon was caused by the water vapor occupying the adsorption sites and hindering the oxygen adsorption. The response values of WG-10Pd to 1% H2 were 77.8%, 73.1%, 69.4%, 65.9% and 60.6% under 24%, 43%, 55%, 71% and 89% RH, respectively. The response variation (calculated as 1 − Response (89RH%)/Response (24RH%)) was 22%, indicating that RH had a significant influence on the hydrogen sensing. Meanwhile, the resistance in air decreased from 42.5 to 35.4 MΩ as RH changed from 24% to 89%. Therefore, the resistance variation was about 16.7%.
Similarly, the response of WG-1Pd and WG-5Pd at 50-100 °C were studied ( Figure  8). The variation trend of response value and response time was basically the same as that of WG-10Pd, but the response value was slightly reduced and the response time was extended. The responses to 0.05% H2 at 100 °C of WG-1Pd and WG-5Pd were 86.1% and 92.6%, respectively, while the response times were 25 s and 9 s, respectively.

Discussion
The WG-Pd ternary materials had a good response to H 2 from room temperature to 100 • C due to their chemical composition and morphological structure (Figure 9). According to the surface depletion layer model, when the material contacts oxygen in the air, it forms oxygen negative ions (O 2 − , O − , O 2− , etc.) on the surface by adsorbing electrons [33]. This increases electrical resistance. At different temperatures, oxygen anions have different valence states. Below 100 • C, it is mainly O 2 − ions. Between 100-300 • C, O − dominates. Above 300 • C, oxygen anions are mostly O 2− [34,35]. Thus, the adsorbed oxygen ions on the surface of WG-Pd material are mainly O 2 − , as the operating temperature is below 100 • C. When it contacts hydrogen, O 2 − reacts with hydrogen to form water vapor and releases electrons to the material, causing a drop in resistance [36]. The resistance decrease is proportional to the hydrogen concentration, allowing measurement of hydrogen concentration through resistance change. The reaction process is as follows: portional to the hydrogen concentration, allowing measurement of hydrogen concentration through resistance change. The reaction process is as follows: 2H2 + O2 − → 2H2O + e − (1) Figure 9. The schematic illustration of WG-Pd during exposure to H2.
In the WG-Pd ternary composites, the three components showed different functions. As the material with the highest content, WO3 played the main role in gas sensing. The abundant oxygen defects on its surface provided active sites for oxygen adsorption [37]. Pd nanoparticles could dissociate H2 molecules into highly active hydrogen atoms, thus accelerating the reaction of O 2− with H2. H2 could also be directly dissolved into metal Pd to form Pd-H complex, thus significantly reducing the Schottky barrier on WO3-Pd interface, enabling more electrons to enter the conduction band of the material and improving the electrical conductivity of the material [6]. However, electron transfer between different WO3/Pd nanosheets was still hindered by the Schottky barrier between interfaces, while graphene sheets could significantly reduce that, providing multiple paths for electron transfer, thus further reducing the resistance [38]. Graphene itself had a high specific surface area and oxygen-containing functional groups, which improved response performance [39]. Moreover, the hydrogen-sensing performance could also be enhanced by the heterogeneous structure formed at the interface between graphene and WO3. WO3 is an n-type semiconductor material with more electrons than holes as the main carriers. Graphene sheets are p-type semiconductor materials with more holes than electrons as the main carriers, due to the intrinsic properties of graphene. Thus, there are many heterojunctions (p-n junctions) at the interface between p-type graphene and n-type WO3. Because of the different work functions of graphene and WO3, the holes transfer from graphene to WO3 and the electrons transfer from WO3 to graphene at the p-n junction. After the carrier exchange, the holes and electrons with opposite charges recombine. This leads to the decrease of concentration of effective carriers and formation of electron depletion In the WG-Pd ternary composites, the three components showed different functions. As the material with the highest content, WO 3 played the main role in gas sensing. The abundant oxygen defects on its surface provided active sites for oxygen adsorption [37]. Pd nanoparticles could dissociate H 2 molecules into highly active hydrogen atoms, thus accelerating the reaction of O 2− with H 2 . H 2 could also be directly dissolved into metal Pd to form Pd-H complex, thus significantly reducing the Schottky barrier on WO 3 -Pd interface, enabling more electrons to enter the conduction band of the material and improving the electrical conductivity of the material [6]. However, electron transfer between different WO 3 /Pd nanosheets was still hindered by the Schottky barrier between interfaces, while graphene sheets could significantly reduce that, providing multiple paths for electron transfer, thus further reducing the resistance [38]. Graphene itself had a high specific surface area and oxygen-containing functional groups, which improved response performance [39]. Moreover, the hydrogen-sensing performance could also be enhanced by the heterogeneous structure formed at the interface between graphene and WO 3 . WO 3 is an n-type semiconductor material with more electrons than holes as the main carriers. Graphene sheets are p-type semiconductor materials with more holes than electrons as the main carriers, due to the intrinsic properties of graphene. Thus, there are many heterojunctions (p-n junctions) at the interface between p-type graphene and n-type WO 3 . Because of the different work functions of graphene and WO 3 , the holes transfer from graphene to WO 3 and the electrons transfer from WO 3 to graphene at the p-n junction. After the carrier exchange, the holes and electrons with opposite charges recombine. This leads to the decrease of concentration of effective carriers and formation of electron depletion layer and hole depletion layer at the interface of WO 3 and graphene, respectively. The depletion layer can increase the resistance of WG-Pd ternary composites in air and facilitate the adsorption or resistance modulation of H 2 gas molecules, which can improve the hydrogen-sensing properties at room temperature. Through the synergistic effect among the three components of WG-Pd, a sensitive response to hydrogen is achieved at temperatures lower than 100 • C.

Conclusions
In this paper, we fabricated WO 3 /graphene binary materials and WG-Pd ternary materials with different contents of graphene and Pd using the microwave-assisted hydrothermal method. The sensing results showed that WG with 1% graphene exhibited the best hydrogen performance at 100 and 150 • C compared to other WO 3 /graphene binary materials with different contents of graphene. However, WG still exhibited no noticeable response to hydrogen below 100 • C. For WG-Pd ternary materials, WG-10Pd showed the highest response to H 2 at room temperature. Moreover, it showed a wide range of hydrogen concentrations (0.05-4 vol.%), a short response time (40 s) and a good selectivity. Its sensing properties were further enhanced after heating to 50, 75 and 100 • C. The excellent sensing properties resulted from the synergistic effect among the three components of WG-Pd. In summary, WG-Pd ternary materials are promising sensing materials for detection of H 2 leakage at temperatures lower than 100 • C, even at room temperature.

Conflicts of Interest:
The authors declare no conflict of interest.