Highly Dispersive Palladium Loading on ZnO by Galvanic Replacements with Improved Methane Sensing Performance

Methane detection is important for the safety of production and life. Metal oxide semiconductor (MOS) methane detection is a mature and widely used technology but still experiences problems such as unsatisfying low-temperature sensing performances. In this study, ZnO/Pd with Pd nanoparticles of different diameters was prepared to study the influence of Pd dispersion on CH4 sensing properties. Results showed that CH4 sensing enhancements were positively correlated with the dispersity of Pd. Moreover, by galvanic replacement using Ag as the sacrificial template, a highly dispersive loading of Pd on ZnO was realized, and the CH4 sensing performance was further enhanced while the amount of Pd reduced from 1.35 wt% to 0.26 wt%. Experiments and DFT calculation indicated that improved CH4 sensing performance resulted from abundant catalytic sites induced by highly dispersed Pd NPs and the enhanced CH4 adsorption on positively charged Pds caused by electrons transferred from Pd to Ag. This study provides a strategy to achieve high dispersion of Pd to maximize the utilization of noble metal, which is promising for lowering the cost of the MOS-based CH4 sensors.


Introduction
Methane is a flammable and explosive gas with an explosion limit of 5-15% at room temperature [1], which has potential safety hazards. Therefore, the real-time and rapid detection of methane is the key to minimizing the loss of life and property. Catalytic combustion type sensors are widely used for methane detection in coal mines at this stage due to their simplicity, low cost, and fast detection. However, their high operating temperature (600-800 • C) consumes a high amount of energy and leads to a short lifetime due to the catalyst's degradation [2][3][4]. Metal oxide semiconductor (MOS) based gas sensors are promising for methane detection due to their relatively low operating temperatures (200-500 • C), the ease of fabrication and low costs [5][6][7][8], such as SnO 2 [9], ZnO [10], TiO 2 [11] and WO 3 [12]. To further reduce energy consumption, there is still room to reduce the operating temperature of MOS gas sensors. Previous studies have shown that loading noble metals (Pd, Pt, Au, Ag, etc.) on the surface of metal oxides is beneficial for lowering the operating temperature due to their excellent catalytic properties that promote methane dissociation and reduce the activation energy of methane oxidation [13][14][15][16]. Among the noble metals, Pd is the most concentrated for its promotion to C-H bond cleavage and shows the best performance in CH 4 sensing.
However, the current utilization of Pd on the surface of MOS is usually low due to their relatively large size, which means that only the surface atoms is functional while the bulk atoms are not employed. One of the effective methods for improving the utilization is to reduce the size of Pd nanoparticles (NPs). Luo et al. realized highly dispersed Pd NPs with diameters of 2-5 nm taking advantage of the confinement of ZIF-8 cavity, which limited the growth of Pd NPs [17]. Sun et al. synthesized highly dispersed Pd NPs with diameters of 3-6 nm by introducing positive potential sites such as an SBA-15 molecular sieve on the surface of CeO 2 , which generates strong electrostatic adsorption with the anion PdCl 4 2− . Therefore the size of Pd is limited by the size of positive potential sites [18]. Nevertheless, the methods that introduce extra substances usually isolated the contact of MOS with Pd, and the sensing enhancement of Pd is not functional. In addition, the introduction of other substances would also affect CH 4 diffusion.
To better improve the Pd utilization in the MOS CH 4 sensing system, galvanic replacement (GR) is a promising method [19,20]. The galvanic replacement reaction is a thermodynamically spontaneous redox process driven by the standard electrode potentials difference of the two metals. The method requires a second metal as the sacrificial template, on which Pd is highly dispersed through GR reaction [21,22]. Usually Ag (+0.799 V vs. SHE) is considered as the most suitable choice for Pd (+0.915 V vs. SHE). Moreover, the price of Ag is approximately 1/100 of Pd. Therefore, the cost of the MOS-based CH 4 sensor could be significantly decreased by using the GR method.
In this study, ZnO/Pd material with controlled diameters of Pd NPs was prepared firstly to investigate the influence of the amount and dispersion of Pd on the CH 4 sensing properties. Then, in order to further improve the dispersion of Pd on ZnO, the GR method is applied using Ag as the sacrificial template to synthesize the ZnO/Ag@Pd. The Pd dispersion is remarkably improved by galvanic replacements with enhanced CH 4 sensing performances, and the amount of Pd greatly reduced to 0.23 wt% compared with ZnO/Pd (1.37 wt%). Finally, the synergistic effect of Ag@Pd in CH 4 sensing is investigated by density functional theory calculations.

Synthesis
For the preparation of ZnO nanorod, 1.756 g ZnAc 2 and 0.200 g polyvinylpyrrolidone (PVP) were added into 200 mL of ethanol, and 8.000 g NaOH was added into 200 mL of ethanol. Both solutions were heated and stirred at 60 • C to fully dissolve them in order to prepare the ZnAc 2 alcohol solution and the NaOH alcohol solution. After cooling to room temperature, the ZnAc 2 alcohol solution was added dropwise to the NaOH alcohol solution at a constant rate of 5.85 mL/min using a peristaltic pump by thoroughly stirring throughout this procedure. After complete dropwise addition, the solution was centrifuged, the supernatant was discarded, and the precipitate was used as ZnO seeds.
ZnO nanorods as the starting materials were synthesized using ε-Zn(OH) 2 as the precursor [23]. The precursor solution was obtained by dissolving 1.2 g ε-Zn(OH) 2 in 40 mL of 2 M NaOH solution and stirring at room temperature for 20 min to form a suspension. The suspension was repeatedly flushed with a dropper to mix with the pre-prepared ZnO seeds and the suspension was heated and stirred for 30 min at 80 • C. The precipitate was filtered, washed and dried to obtain ZnO nanorods.
For the preparation of ZnO/Pd, 0.4 g of ZnO nanorods, 0.01 g of PdCl 2 and a certain amount of PVP were dispersed into 80 mL of ethylene glycol to form a suspension. Then the suspension was heated at 100 • C for 3 h. Finally, the precipitate was washed and dried at 60 • C overnight, and ZnO/Pd was obtained.
For the preparation of ZnO/Ag, 0.1 g of ZnO nanorods, 0.01 g of AgNO 3 and 0.10 g PVP were dispersed into 20 mL of ethylene glycol to form a suspension. Then the suspension was heated at 50 • C for 1 h. Finally, the precipitate was washed and dried at 60 • C overnight, and ZnO/Ag was obtained.
ZnO/Ag@Pd was prepared by a facile galvanic replacement method. Firstly, 0.01 g PdCl 2 was dissolved into 10 mL 0.04 M hydrochloric acid to prepare a PdCl 2 solution. Then, 0.05 g of ZnO/Ag composite and 1 mL of PdCl 2 solution were added into 20 mL of deionized water, and then the mixture was kept for 30 min at room temperature. The precipitate was washed and dried at 60 • C overnight, and then the ZnO/Ag@Pd composite was obtained.

Characterization
The morphology and microstructure of the samples were examined by field emission scanning electron microscopy (FESEM, JSM 7401F, JEOL, Tokyo, Japan) and high-resolution transmission electron microscopy ((HRTEM, JEM-2010, JEOL, Tokyo, Japan), respectively. The surface composition of the samples was characterized by X-ray photoelectron spectrometry (XPS, PHI-5300, PHI, Chanhassen, USA). Powder X-ray diffraction (XRD) analysis was performed using an X-ray diffractometer (D8 Advance, Bruker, Karlsruhe, Germany) with Cu Kα (λ = 0.154178 nm) radiation for phase identification. An inductively coupled plasma (ICP) spectrometer (Thermo IRIS Intrepid II type) was used to detect the content of specific elements in the samples. A series of standard solutions with a concentration gradient was prepared, and the sample to be tested was dissolved with acid. Temperature-programmed desorption (TPD) of adsorbed oxygen was carried out using Quantachrome ChemBET Pulsar TPR/TPD. In a typical CO-TPD measurement, the sample was initially treated with He at 200 • C for 30 min and then cooled down to 90 • C. Afterwards, the sample was exposed to a mixture of CO and He at 90 • C for 45 min. Then the sample was cooled down to 50 • C and flushed with He. CO-TPD analysis was performed with a heating rate of 10 K min −1 up to 600 • C.

Gas Sensor Fabrication and Gas Sensing Tests
The sample to be tested was grinded thoroughly with deionized water with a concentration of 3.0 g/L to form a slurry. A small amount of the grinded slurry was dipped with a brush and applied evenly on the alumina ceramic tube. The ceramic tube was heated at 250 • C for 3 h to stabilize the powdered composite material bonded to the alumina surface. Using an AS-20 sensor aging table (Beijing Elite Tech. Co., Ltd., Beijing, China), the ceramic tube was aged for 24 h at 160 mA to stabilize its sensing performance. [24] The resistance of the tube was measured by the CGS-8 Intelligent Gas Sensing Analysis System (Beijing Elite Tech. Co., Ltd., Beijing, China). The target concentration of methane (500 ppm-10,000 ppm) was obtained by the injection of certain volume of methane into the fixed chamber with a volume of 20 L, and the variation of the sensor resistance over time was obtained from the analysis system. The response of the sensor is defined as (R a − R g )/R g × 100 (%), where R a and R g are the film resistance in air and methane, respectively.

Computational Methodology
Density function theory (DFT) calculations were performed by the Dmol 3 program [25,26] in Materials Studio, using the Spin Restricted method. All calculations were performed using the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) [27] for the description of the atomic orbitals. The electron truncation energy was taken as 500 eV.
In calculating the structure of the metal surface, the cell model is a 3-layer infinitely extended plane, using a 3 × 3 supercell structure for calculations with a metal surface vacuum layer thickness of 15 Å. Fixing the lower atomic coordinates, the relaxation is performed on the surface atoms. The Brillouin zone is sampled at k points using a 4 × 4 × 1 grid with a grid length of 0.025 Å −1 as the basis for inverse easy space integration.
For the calculations using the DFT method, all possible adsorption sites were selected for the adsorption energy calculations involved, and the adsorption site with the largest adsorption energy was taken as the most stable adsorption site. The adsorption energy calculation is calculated as follows: where E abs represents the adsorption energy; E composite represents the energy of a single gas molecule adsorbed on the blank metal surface to form a composite structure; E CH 4 represents the energy of a single CH 4 molecule; and E metal represents the energy of the blank metal surface.

Effect of Pd Dispersion on the CH 4 Sensing Performance
ZnO/Pd with Pd NPs of different diameter was firstly synthesized to investigate the effect of Pd dispersion on CH 4 sensing performance by adjusting the amount of PVP during the preparation of ZnO/Pd materials (0 g, 0.02 g, 0.05 g, and 0.10 g), and the four samples were labelled as ZnO/Pd-0, 1, 2 and 3. As observed in the SEM images of Figure 1b, the ZnO/Pd material, starting from a center, radially grows many ZnO nanorods in a spiky spherical structure in all directions. The average radius of an individual nanorod is~15 nm, and the length of nanorod is 0.5-2 µm. From the XRD diffraction pattern of Figure 1a, the peaks at 31.77 • , 34.42 • , and 36.35 • correspond to ZnO(100), (001), and (101) crystalline planes (PCPDF 36-1451), respectively, while the peak at 40.20 • corresponds to the Pd(111) crystalline plane (PCPDF 89-4897). As the amount of PVP increases, the peak of Pd(111) becomes wider and lower, indicating that the size of Pd NPs becomes smaller. The HRTEM image in Figure 1c of ZnO/Pd-2 shows the crystalline spacing of 0.522 nm and 0.224 nm, which corresponds to ZnO(001) and Pd(111) crystalline planes, respectively, and showed great correspondence with XRD results. = − 4 − where represents the adsorption energy; represents the energy of gas molecule adsorbed on the blank metal surface to form a composite structu represents the energy of a single CH4 molecule; and represents the energ blank metal surface.

Effect of Pd Dispersion on the CH4 Sensing Performance
ZnO/Pd with Pd NPs of different diameter was firstly synthesized to investi effect of Pd dispersion on CH4 sensing performance by adjusting the amount of P ing the preparation of ZnO/Pd materials (0 g, 0.02 g, 0.05 g, and 0.10 g), and the fo ples were labelled as ZnO/Pd-0, 1, 2 and 3. As observed in the SEM images of Fi the ZnO/Pd material, starting from a center, radially grows many ZnO nanorods in spherical structure in all directions. The average radius of an individual nanoro nm, and the length of nanorod is 0.5-2 μm. From the XRD diffraction pattern of Fi the peaks at 31.77°, 34.42°, and 36.35° correspond to ZnO(100), (001), and (101) cry planes (PCPDF 36-1451), respectively, while the peak at 40.20° corresponds to the crystalline plane (PCPDF 89-4897). As the amount of PVP increases, the peak of becomes wider and lower, indicating that the size of Pd NPs becomes smal HRTEM image in Figure 1c of ZnO/Pd-2 shows the crystalline spacing of 0.522 0.224 nm, which corresponds to ZnO(001) and Pd(111) crystalline planes, respe and showed great correspondence with XRD results. It can be observed from Figure 2 that the average particle size of Pd NPs w 5.78, 2.67, and 2.54 nm for ZnO/Pd-0, 1, 2, and 3, respectively, since PVP limits the and agglomeration of metal NPs due to its capping effect [28]. Pd NPs are un loaded on ZnO nanorods. However, there is no significant difference between th It can be observed from Figure 2 that the average particle size of Pd NPs was 8.63, 5.78, 2.67, and 2.54 nm for ZnO/Pd-0, 1, 2, and 3, respectively, since PVP limits the growth and agglomeration of metal NPs due to its capping effect [28]. Pd NPs are uniformly loaded on ZnO nanorods. However, there is no significant difference between the size of Pd NPs in ZnO/Pd-2 and ZnO/Pd-3, indicating that the size of Pd NPs cannot be further reduced by increasing the amount of PVP.
Chemosensors 2022, 10, x FOR PEER REVIEW Pd NPs in ZnO/Pd-2 and ZnO/Pd-3, indicating that the size of Pd NPs cannot b reduced by increasing the amount of PVP. The sensing properties of ZnO and four ZnO/Pd samples to 500 ppm CH4 at temperatures were tested, and the response results are summarized in Figure  pared with ZnO, four ZnO/Pd samples all showed improved CH4 sensing perf with decreased optimal temperatures (280 °C to 230 °C) and increased responses ing the catalytic promotion by Pd. The CH4 response increased as the size of Pd creased, and ZnO/Pd-2 and ZnO/Pd-3 show the highest CH4 response at about 27 it can be concluded that CH4 responses have a negative correlation to the size of i.e., it is positively correlated with the dispersity of Pd. The smaller the particle s NPs, the better the dispersity of Pd, and the higher the response value of ZnO/Pd to CH4. Pd contents in four ZnO/Pd materials were analyzed using ICP. It was fo Pd contents were 3.26%, 2.49%, 1.35% and 1.41 wt% for ZnO/Pd-0, 1, 2, and 3, resp Combined with TEM observations, both the size and the amount of Pd cannot b reduced by adding more PVP. The sensing properties of ZnO and four ZnO/Pd samples to 500 ppm CH 4 at different temperatures were tested, and the response results are summarized in Figure 3. Compared with ZnO, four ZnO/Pd samples all showed improved CH 4 sensing performance with decreased optimal temperatures (280 • C to 230 • C) and increased responses, indicating the catalytic promotion by Pd. The CH 4 response increased as the size of Pd NPs decreased, and ZnO/Pd-2 and ZnO/Pd-3 show the highest CH 4 response at about 27%. Thus, it can be concluded that CH 4 responses have a negative correlation to the size of Pd NPs; i.e., it is positively correlated with the dispersity of Pd. The smaller the particle size of Pd NPs, the better the dispersity of Pd, and the higher the response value of ZnO/Pd relative to CH 4 . Pd contents in four ZnO/Pd materials were analyzed using ICP. It was found that Pd contents were 3.26%, 2.49%, 1.35% and 1.41 wt% for ZnO/Pd-0, 1, 2, and 3, respectively. Combined with TEM observations, both the size and the amount of Pd cannot be further reduced by adding more PVP.

Highly Dispersive Pd with Enhanced CH4 Sensing Performance via Galvanic Replacement
In order to further reduce the size and the amount of Pd, (i.e., improving dispersity), the GR method was deployed using Ag as a sacrificial template. The reaction of Pd 2+ with Ag NPs on the ZnO surface occurs as follows: Pd 2+ + 2Ag → Pd + 2Ag + . The schematic diagram is shown in Figure 4a.

Highly Dispersive Pd with Enhanced CH 4 Sensing Performance via Galvanic Replacement
In order to further reduce the size and the amount of Pd, (i.e., improving dispersity), the GR method was deployed using Ag as a sacrificial template. The reaction of Pd 2+ with Ag NPs on the ZnO surface occurs as follows: Pd 2+ + 2Ag → Pd + 2Ag + . The schematic diagram is shown in Figure 4a.

Highly Dispersive Pd with Enhanced CH4 Sensing Performance via Galvanic Replacement
In order to further reduce the size and the amount of Pd, (i.e., improving dispersity), the GR method was deployed using Ag as a sacrificial template. The reaction of Pd 2+ with Ag NPs on the ZnO surface occurs as follows: Pd 2+ + 2Ag → Pd + 2Ag + . The schematic diagram is shown in Figure 4a.  Figure 4b, Ag NPs with an average diameter of 8 nm are uniformly loaded on ZnO nanorods. After GR reaction, the ZnO/Ag@Pd was prepared. From the XRD diffraction pattern of Figure 4d, the peak at 38.12 • can be indexed to the Ag(111) crystal plane (PCPDF 89-3722), which is consistent with the HRTEM image with a crystalline spacing of 0.236 nm in Figure 4c. Figure 4f shows that the crystal plane spacing of the particles on ZnO/Ag@Pd is 0.224 nm and 0.236 nm, which should be identified as the Pd(111) and Ag(111) crystal plane spacing, respectively. Compared with the ZnO/Pd (all ZnO/Pd in the following refers to ZnO/Pd-2), Pd NPs on ZnO/Ag@Pd are denser and more uniform, and the particle size is significantly reduced from 2-4 nm to 1-2 nm. From the XRD patterns of ZnO/Ag@Pd in Figure 4d, the Ag(111) peak could not be indexed as the Ag had been sacrificed, and the Pd(111) peak was also absent due to its ultra-small size. The overall XPS spectrum in Figure 4g confirms the coexistence of Ag and Pd. Moreover, by comparing the fine-scanned XPS spectrum of Pd and Ag in ZnO/Pd, ZnO/Ag, and ZnO/Ag@Pd in Figure 5, the 3d 3/2 and 3d 5/2 peaks of Pd shifted from 340.1, 334.9 eV to a higher level of 342.3 and 337.1 eV, while the 3d 3/2 and 3d 5/2 peak of Ag shifted from 373.5 and 367.5 eV to lower levels at 367.5 and 367.1 eV, respectively, indicating that Pd and Ag form an alloy in the ZnO/Ag@Pd with electrons transferring from Pd to Ag. According to ICP results, the Pd content in ZnO/Ag@Pd significantly decreased from 1.35 wt% (ZnO/Pd) to 0.23 wt%. The above results indicated that highly dispersive Pd was successfully achieved by using the GR method. particles on ZnO/Ag@Pd is 0.224 nm and 0.236 nm, which should be identified as the Pd(111) and Ag(111) crystal plane spacing, respectively. Compared with the ZnO/Pd (all ZnO/Pd in the following refers to ZnO/Pd-2), Pd NPs on ZnO/Ag@Pd are denser and more uniform, and the particle size is significantly reduced from 2-4 nm to 1-2 nm. From the XRD patterns of ZnO/Ag@Pd in Figure 4d, the Ag(111) peak could not be indexed as the Ag had been sacrificed, and the Pd(111) peak was also absent due to its ultra-small size. The overall XPS spectrum in Figure 4g confirms the coexistence of Ag and Pd. Moreover, by comparing the fine-scanned XPS spectrum of Pd and Ag in ZnO/Pd, ZnO/Ag, and ZnO/Ag@Pd in Figure 5, the 3d3/2 and 3d5/2 peaks of Pd shifted from 340.1, 334.9 eV to a higher level of 342.3 and 337.1 eV, while the 3d3/2 and 3d5/2 peak of Ag shifted from 373.5 and 367.5 eV to lower levels at 367.5 and 367.1 eV, respectively, indicating that Pd and Ag form an alloy in the ZnO/Ag@Pd with electrons transferring from Pd to Ag. According to ICP results, the Pd content in ZnO/Ag@Pd significantly decreased from 1.35 wt% (ZnO/Pd) to 0.23 wt%. The above results indicated that highly dispersive Pd was successfully achieved by using the GR method. The responses of ZnO/Pd, ZnO/Ag, and ZnO/Ag@Pd to 500 ppm CH4 at 160-280 °C were tested, and the results are shown in Figure 6a. The response value of ZnO/Ag is much weaker than ZnO/Pd and the CH4 response become significant at temperatures higher than 260 °C, while the response of ZnO/Ag@Pd is higher than that of ZnO/Pd at all temperature ranges. The dynamic response curves of three samples at 230 °C are shown in Figure 6b, as all samples show fast responses to CH4, with response/recovery times less than 10 s, and ZnO/Ag@Pd showed the highest response of 32.6%. However, the optimal temperature of ZnO/Ag@Pd was higher than that of ZnO/Pd, and the corresponding response-temperature behavior was more similar to that of ZnO/Ag, which means that Pd can work as the sensing promoter to ZnO/Ag. From the above results, it can be concluded that the GR method using Ag as sacrificial metals can significantly improve the dispersity of Pd and reduce the amount of Pd at the same time. The highly dispersive Pd can provide more active sites for C-H cleavage, and, therefore, accelerate CH4 oxidation to achieve improved CH4 sensing performance. The responses of ZnO/Pd, ZnO/Ag, and ZnO/Ag@Pd to 500 ppm CH 4 at 160-280 • C were tested, and the results are shown in Figure 6a. The response value of ZnO/Ag is much weaker than ZnO/Pd and the CH 4 response become significant at temperatures higher than 260 • C, while the response of ZnO/Ag@Pd is higher than that of ZnO/Pd at all temperature ranges. The dynamic response curves of three samples at 230 • C are shown in Figure 6b, as all samples show fast responses to CH 4 , with response/recovery times less than 10 s, and ZnO/Ag@Pd showed the highest response of 32.6%. However, the optimal temperature of ZnO/Ag@Pd was higher than that of ZnO/Pd, and the corresponding response-temperature behavior was more similar to that of ZnO/Ag, which means that Pd can work as the sensing promoter to ZnO/Ag. From the above results, it can be concluded that the GR method using Ag as sacrificial metals can significantly improve the dispersity of Pd and reduce the amount of Pd at the same time. The highly dispersive Pd can provide more active sites for C-H cleavage, and, therefore, accelerate CH 4 oxidation to achieve improved CH 4 sensing performance.

Mechanism
To further investigate the mechanism, ZnO/Pd, ZnO/Ag, and ZnO/Ag@Pd materials were analyzed by conducting CO-TPD measurement, and the results are shown in Figure  7. It is observed that ZnO/Ag has no signal response, and it is believed that the Ag NPs loaded on ZnO do not provide adsorption sites for CO; on the other hand, the amount of CO adsorbed by the ZnO/Ag@Pd material is much higher than that of ZnO/Pd. The amount of CO adsorbed by ZnO/Ag@Pd is 2.06 times higher than that of the same mass of ZnO/Pd as calculated by the graphic integration of the desorption peaks.
The particle size of Pd can provide more adsorption sites and catalytic sites on the ZnO surface. When sensing reactions occur, these more catalytic sites can enhance the rate of CH4 catalytic oxidation reactions, allowing ZnO to exhibit higher response values and to improve its CH4 sensing performance [15].

Mechanism
To further investigate the mechanism, ZnO/Pd, ZnO/Ag, and ZnO/Ag@Pd materials were analyzed by conducting CO-TPD measurement, and the results are shown in Figure 7. It is observed that ZnO/Ag has no signal response, and it is believed that the Ag NPs loaded on ZnO do not provide adsorption sites for CO; on the other hand, the amount of CO adsorbed by the ZnO/Ag@Pd material is much higher than that of ZnO/Pd. The amount of CO adsorbed by ZnO/Ag@Pd is 2.06 times higher than that of the same mass of ZnO/Pd as calculated by the graphic integration of the desorption peaks.

Mechanism
To further investigate the mechanism, ZnO/Pd, ZnO/Ag, and ZnO/Ag@Pd materials were analyzed by conducting CO-TPD measurement, and the results are shown in Figure  7. It is observed that ZnO/Ag has no signal response, and it is believed that the Ag NPs loaded on ZnO do not provide adsorption sites for CO; on the other hand, the amount o CO adsorbed by the ZnO/Ag@Pd material is much higher than that of ZnO/Pd. The amount of CO adsorbed by ZnO/Ag@Pd is 2.06 times higher than that of the same mass of ZnO/Pd as calculated by the graphic integration of the desorption peaks.
The particle size of Pd can provide more adsorption sites and catalytic sites on the ZnO surface. When sensing reactions occur, these more catalytic sites can enhance the rate of CH4 catalytic oxidation reactions, allowing ZnO to exhibit higher response values and to improve its CH4 sensing performance [15]. The particle size of Pd can provide more adsorption sites and catalytic sites on the ZnO surface. When sensing reactions occur, these more catalytic sites can enhance the rate of CH 4 catalytic oxidation reactions, allowing ZnO to exhibit higher response values and to improve its CH 4 sensing performance [15].
Despite the improvement induced by increased Pd dispersities, the effect of interactions between Pd and Ag should be also taken into consideration. Previous studies have proven that the PdAg showed enhanced catalytic effects than compared to Pd or Ag in several systems. Thus, DFT calculation was conducted to study the influence of PdAg bimetallic structures on CH 4 oxidations. Pd(111) and Ag(111) surfaces were established to study CH 4 adsorption properties on ZnO/Pd and ZnO/Ag according to TEM and XRD results. Ag(111) surfaces doped with Pd atoms (denoted as Ag(111)@Pd surface) were chosen to simulate the structure of Ag@Pd. The adsorption energies of CH 4 on Pd(111), Ag(111) and Ag(111)@Pd surfaces were −0.179, −0.120 and −0.246 eV, respectively, with the order of Ag(111)@Pd > Pd(111) > Ag(111). Furthermore, a differential charge density analysis was performed for the adsorption system with the value of the equivalent surface of 0.001 e/Å 3 , and the spatial configurations of the adsorption structure and the differential charge density diagram are shown in Figure 8a-c. The results showed that there was no obvious charge exchange between Ag and CH 4 in Ag(111), while the charge exchange between Pd and CH 4 was more significant in Ag(111)@Pd compared to Pd(111). To further investigate the reason of charge exchange, differential charge density analyses were performed on the blank Ag(111)@Pd for the doped Pd atoms with the value of the equivalent surface of 0.02 e/Å 3 . As shown in Figure 8e. It was found that the electron on Pd atoms transferred to Ag after Pd doping, in agreement with the previous XPS analysis. From the above results, the mechanism for the CH 4 sensing enhancement by GR method can be concluded in two aspects, as shown in Figure 9. On the one hand, the highly dispersive Pd provided more catalytic sites for CH 4 oxidation with more electrons transferring to ZnO. On the other hand, the bimetallic Ag@Pd structure results in the transfer of electrons from Pd to Ag which makes Pd more positively charged, leading to the stronger adsorption of CH 4 on Pd. Therefore, both effects synergistically contributed to the enhanced CH 4 sensing performance of ZnO/Ag@Pd. method can be concluded in two aspects, as shown in Figure 9. On the one hand, the highly dispersive Pd provided more catalytic sites for CH4 oxidation with more electrons transferring to ZnO. On the other hand, the bimetallic Ag@Pd structure results in the transfer of electrons from Pd to Ag which makes Pd more positively charged, leading to the stronger adsorption of CH4 on Pd. Therefore, both effects synergistically contributed to the enhanced CH4 sensing performance of ZnO/Ag@Pd.

Conclusions
In this study, the CH 4 sensing performance of ZnO/Pd was investigated, and it was demonstrated that the effect of Pd improving the CH 4 sensing performance has a positive correlation with the dispersity of Pd NPs. Based on this, a highly dispersive loading of Pd on the ZnO surface was achieved by using galvanic replacements using Ag as the sacrificial template. The size of Pd NPs was reduced from 2-4 nm to 1-2 nm with the amount of Pd reduced from 1.35 wt% to 0.23 wt%, while the response of ZnO/Ag@Pd to 500 ppm CH 4 at 230 • C increased from 27.1% to 32.6%. The enhanced performance was attributed to two aspects: The GR method significantly improved the dispersity of Pd NPs which provide more catalytic sites for CH 4 oxidation; Ag@Pd structure by the GR method brought positively charged Pd due to the transfer of electrons from Pd to Ag, which strengthened the adsorption of CH 4 . This study provides a strategy for lowering the cost of MOS CH 4 sensors using Pd as a promoter and by reducing the amount of Pd while improving sensing performances, which may provide insights on the reduction in fabrication costs in MOS sensing systems using other noble metals.