Improved TEA Sensitivity and Selectivity of In2O3 Porous Nanospheres by Modification with Ag Nanoparticles

A highly sensitive and selective detection of volatile organic compounds (VOCs) by using gas sensors based on metal oxide semiconductor (MOS) has attracted increasing interest, but still remains a challenge in gas sensitivity and selectivity. In order to improve the sensitivity and selectivity of In2O3 to triethylamine (TEA), herein, a silver (Ag)-modification strategy is proposed. Ag nanoparticles with a size around 25–30 nm were modified on pre-synthesized In2O3 PNSs via a simple room-temperature chemical reduction method by using NaBH4 as a reductant. The results of gas sensing tests indicate that after functionalization with Ag, the gas sensing performance of In2O3 PNSs for VOCs, especially for TEA, was remarkably improved. At a lower optimal working temperature (OWT) of 300 °C (bare In2O3 sensor: 320 °C), the best Ag/In2O3-2 sensor (Ag/In2O3 PNSs with an optimized Ag content of 2.90 wt%) shows a sensitivity of 116.86/ppm to 1–50 ppm TEA, about 170 times higher than that of bare In2O3 sensor (0.69/ppm). Significantly, the Ag/In2O3-2 sensor can provide a response (Ra/Rg) as high as 5697 to 50 ppm TEA, which is superior to most previous TEA sensors. Besides lower OWT and higher sensitivity, the Ag/In2O3-2 sensor also shows a remarkably improved selectivity to TEA, whose selectivity coefficient (STEA/Sethanol) is as high as 5.30, about 3.3 times higher than that of bare In2O3 (1.59). The sensitization mechanism of Ag on In2O3 is discussed in detail.


Introduction
In recent decades, with the sustained growth of environmental issues, fast and timely detection of flammable and harmful gases in our surroundings has attracted increasing attention [1][2][3][4]. Triethylamine (TEA), as a member of volatile organic compounds (VOCs), is a typical organic amine compound with strong pungent smell, which has been widely used in industry as a catalyst, solvent, synthetic fuel, preservative, and so on [5,6]. However, owing to its toxic nature, long-term exposure to excessive TEA (>1 ppm) may cause a series of health problems, such as skin irritation, headache, gastroenteritis, and pulmonary edema [7,8]. In practice, some traditional techniques, such as chromatographic and colorimetric methods, have been applied to detect TEA, but usually suffer from high price, complicated operations, and time-consuming detection processes [9]. Therefore, developing fast and convenient techniques for TEA detection is of great important.
Among different state-of-the-art gas detection techniques, gas sensors based on metal oxide semiconductor (MOS) have attracted extensive attention because of their outstanding merits of low price, small volume, easy fabrication and integration, high sensitivity for diverse gases, and fast response speed [10][11][12][13]. Up to now, several MOSs such as SnO 2 [14], ZnO [15], TiO 2 [16], WO 3 [17], In 2 O 3 [18,19], and Fe 2 O 3 [20] have been studied for TEA sensor application, and significant advances have been achieved. However, there are still some obstacles that need to be overcome from the perspective of practical application, especially in the aspects of sensitivity and selectivity. In previous studies, yearning to improve the gas sensing properties of MOS, researchers have attempted several sensitization strategies. These strategies mainly include fabrication of novel nanostructures [21], doping with foreign element [22,23], modification with noble metal (nanoparticles or single atoms) [24][25][26], and combination with other MOSs to construct heterojunction [27,28]. Among these strategies, modification with nano-sized noble metals was found to be very effective to upgrade the gas sensing performance of MOS, especially in the aspect of gas sensitivity. For example, Sukee et al. [29] prepared Ag-loaded LaFeO 3 by a flame spray pyrolysis (FSP) method and found that the 0.1 wt% Ag-loaded LaFeO 3 sensor showed a response of 60 to 100 ppm acetylene, almost 12 times higher than that of the pure LaFeO 3 sensor. Cai et al. [30] fabricated a highly selective H 2 sensor fabricated with Pd nanoparticle-decorated SnO 2 nanowires, whose response to 100 ppm H 2 was about 12.7 times higher than that of the bare SnO 2 sensor. Yang et al. [31] reported that, benefitting from the increased concentration of oxygen vacancy after modification with Au, the Au-decorated In 2 O 3 hollow nanosphere sensor showed excellent sensitivity (26.3 of 100 ppm) and selectivity toward 1-butylamine at the optimized working temperature of 340 • C. All of the above studies have demonstrated that during the gas sensing process, the noble metal sensitizer can bring a series of positive effects on MOS, such as facilitating the formation of chemisorbed oxygen [32,33], modulating the space charge layer [34][35][36], and catalyzing the gas sensing reaction between gaseous analyte and chemisorbed oxygen [37][38][39].
In 2 O 3 , as an n-type semiconductor (E g = 3.55-3.75 eV) with diverse functions, has been widely studied in the fields of lithium ion battery [40], supercapacitor [41], photocatalyst [42], and gas sensor [43] due to its unique physical and chemical properties, such as low toxicity, strong inoxidizability, and high electric mobility [44]. For gas sensor applications, In 2 O 3 has been found to be sensitive to various gases, including TEA [32], CO [45], methane [46], acetone [47], hydrogen [48], alcohol [49], NO2 [50], etc. To improve the gas sensitivity of In 2 O 3 to TEA, different noble metals were used as sensitizers to functionalize In 2 O 3 . For example, Zheng et al. prepared porous In 2 O 3 microspheres by annealing the In 2 S 3 precursor and then decorated Au nanoparticles on them to obtain novel Au/In 2 O 3 hybrid microspheres. At an operating temperature of 280 • C, the hybrid Au/In 2 O 3 microspheres showed a response of 648.2 to 100 ppm TEA, higher than that of pristine In 2 O 3 [51]. Liu et al. reported a TEA sensor that fabricated with Pd nanoparticles-functionalized In 2 O 3 microspheres, whose response to 50 ppm TEA at 220 • C was 47.56, higher than that of In 2 O 3 counterpart [52]. Compared with other noble metals (Au, Pt, Pd), Ag is more suitable for large-scale practical application owing to its lower price. Although Ag has been proven to be a valid sensitizer for In 2 O 3 to sense NO 2 [53] and H 2 S [54], there are few reports, to the best of our knowledge, on improving the TEA sensing performance of In 2 O 3 by modification with Ag, as well as the detailed sensitization effects of Ag on In 2 O 3 for sensing TEA.
In this study, Ag nanoparticles with a size of around 25-30 nm were evenly decorated on pre-synthesized In 2 O 3 porous nanospheres (PNSs) via a simple room-temperature chemical reduction method. To check the influence of Ag modification on In 2 O 3 and further understand the sensitization effect of Ag nanoparticles, the prepared bare and Agmodified In 2 O 3 samples were characterized by various techniques and their TEA sensing performances were investigated in detail. Research results indicate that after modification with Ag, the In 2 O 3 sensor showed impressive improvements in TEA sensing performance, especially of lower operating temperature, much higher sensitivity, and better selectivity. Specifically, at its optimal working temperature (300 • C), the best Ag/In 2 O 3 sensor can give a response as high as 5697 to 50 ppm TEA, which is about 158 times higher than that of bare In 2 O 3 sensor (36) and is also superior to most previous TEA sensors. The upgraded gas sensing performances of Ag/In 2 O 3 sensor can be ascribed to the spillover and catalytic effects of Ag nanoparticles, as well as the electronic sensitization effect of the Ag-In 2 O 3 Schottky junction. Our research not only provides a promising sensitization strategy to enhance the TEA sensing performance of In 2 O 3 , but also contributes to a deeper understanding on the sensitization mechanism of Ag/In 2 O 3 sensor.

Sample Preparation
All chemicals applied in our experiment were of analytical grade and used as received without further purification. To synthesize Ag-modified In 2 O 3 PNSs, bare In 2 O 3 PNSs were synthesized in advance according to our previous method [55]. Briefly, 0.42 g of In(NO 3 ) 3 ·4.5H 2 O and 0.71 g of sodium citrate were dissolved in 80 mL distilled water under magnetic stirring. After adding 0.20 g of urea into above solution, the obtained reaction solution was sealed in a Teflon-lined autoclave with the capacity of 100 mL and heated at 160 • C for 24 h. After reaction, the product was collected by centrifugation, washed with distilled water and absolute ethanol alternately, and dried at 60 • C in air for 8 h. Finally, bare In 2 O 3 sample was obtained by annealing the collected powder at 500 • C (heating rate: 2 • C/min) in air for 3 h.
To synthesize Ag-modified In 2 O 3 PNSs, a designed amount of as-prepared bare In 2 O 3 powder was ultrasonically dispersed into 150 mL of AgNO 3 aqueous solution (0.6 mmol/L), followed by adding 13 mL of NaBH 4 (7 mmol/L) aqueous solution drop by drop. After the mixed solution was sonicated at room temperature for 3 h, dark brown precipitates were collected and purified by washing with distilled water. Finally, the precipitate was dried at 60 • C for 8 h to obtain the final Ag-modified In 2 O 3 sample. The Ag content in Ag-modified In 2 O 3 sample was controlled by adjusting the using amount of bare In 2 O 3 . When the mass of In 2 O 3 was set as 88.26, 74.68, and 64.72 mg, Ag/In 2 O 3 samples with theoretical Ag contents of 11 wt%, 13 wt%, and 15 wt% were prepared and denoted as Ag/In 2 O 3 -1, Ag/In 2 O 3 -2, and Ag/In 2 O 3 -3, respectively.

Characterizations
The phase structure analysis of the samples was performed by powder X-ray diffraction (XRD, Bruker D8, Cu-Kα1 radiation, λ = 1.5418 Å) in the 2θ range of 10-90 • . The actual Ag contents in Ag/In 2 O 3 samples were measured by using an inductively coupled plasma optical emission spectrometer (ICP-OES, Agilent ICP-OES 725 ES). The morphology and microstructure of the samples were characterized by scanning electron microscope (SEM, Merlin Compact) and transmission electron microscope (TEM, JEOL JEM 2100 F). The chemical composition and surface state of the materials were analyzed by X-ray photoelectron spectroscopy (XPS, Thermo SCIENTIFIC ESCALAB 250Xi). The binding energy of elements is calibrated with the surface adventitious carbon (the C 1s peak at 284.8 eV). The porosity and specific surface area of the samples were measured by N 2 adsorption/desorption instrument (Micromeritics ASAP 2020).

Gas sensor Fabrication and Analysis
The gas sensing performance of the samples was measured on an intelligent gas sensing analysis system of CGS-4TPS (Beijing Elite Technology Co., Ltd., Beijing, China), and the gas sensors were fabricated according to our previous method [56]. In a typical sensor fabrication procedure, approximately 10 mg of as-prepared sample was mixed with 1 mL of distilled water by grinding in an agate mortar to obtain a uniform slurry. The slurry was brushcoated on the surface of a ceramic substrate (Al 2 O 3 , size: 13.4 mm × 7 mm × 0.635 mm) with interdigitated Ag-Pd electrodes (width: 0.2 mm, gap distance: 1 mm) and then dried naturally at room temperature to obtain a resistance-type sensor. Before the test, the fabricated sensors were aged at 280 • C for 10 h to improve their stability. By using ambient air as the diluted and reference gas, a static gas distribution method was applied to achieve a desired concentration of target gas in the test chamber (1.8 L). The response of the bare and Ag-modified In 2 O 3 sensors to the reducing analyte is defined as R a /R g , where R a and R g refer to the resistance of the sensor in air and target gas, respectively. The response and recovery time is measured by recording the time of 90% sensor resistance change after the target gas was injected or released. During the test, the environment temperature was about 25 • C and the relative humidity was 30 ± 8%.

Sample Characterization
The phase structure and composition of the prepared samples were confirmed by XRD analysis. As shown in Figure 1, all diffraction peaks match well with the standard data of cubic Obviously, the actual Ag content of Ag/In 2 O 3 samples is lower than their designed amount. This phenomenon is similar to that of a previous report [25]. Therefore, the absence of Ag peaks in the XRD patterns should be attributed to the low content of Ag and the sufficient dispersion of Ag nanoparticles in the surface of In 2 O 3 . test, the fabricated sensors were aged at 280 °C for 10 h to improve their stability. By using ambient air as the diluted and reference gas, a static gas distribution method was applied to achieve a desired concentration of target gas in the test chamber (1.8 L). The response of the bare and Ag-modified In2O3 sensors to the reducing analyte is defined as Ra/Rg, where Ra and Rg refer to the resistance of the sensor in air and target gas, respectively. The response and recovery time is measured by recording the time of 90% sensor resistance change after the target gas was injected or released. During the test, the environment temperature was about 25 °C and the relative humidity was 30 ± 8%.

Sample Characterization
The phase structure and composition of the prepared samples were confirmed by XRD analysis. As shown in Figure 1, all diffraction peaks match well with the standard data of cubic In2O3 (JCPDS No. 71-2195, a = b = c = 10.117 Å), demonstrating the formation of the cubic In2O3 phase in the four samples. The strong and sharp diffraction peaks reveal their good crystallinity. No obvious peaks arising from Ag metal or its related compounds were detected in the Ag/In2O3 samples. To determine the actual Ag contents in the Ag/In2O3 samples, ICP-OES measurements were conducted. The results show that the Ag contents in Ag/In2O3-1, Ag/In2O3-2, and Ag/In2O3-3 are 2.43, 2.90, and 3.68 wt%, respectively. Obviously, the actual Ag content of Ag/In2O3 samples is lower than their designed amount. This phenomenon is similar to that of a previous report [25]. Therefore, the absence of Ag peaks in the XRD patterns should be attributed to the low content of Ag and the sufficient dispersion of Ag nanoparticles in the surface of In2O3. The SEM and TEM were applied to observe the microstructure and morphology of the prepared samples. Figure 2a,b shows the representative SEM images of bare In2O3 and Ag/In2O3-2, respectively. Both samples show similar sphere-like structure with the diameter about 120 nm, demonstrating that Ag modification has almost no influence on the basic morphology of primary In2O3. Figure 2c shows the TEM image of Ag/In2O3-2, from which the porous structure of the nanospheres is clearly observed. In the high-resolution TEM (HRTEM) image recorded from the edge of a randomly selected nanosphere (Figure 2d), two definite crystalline phases were determined by measuring the lattice fringes. The lattice fringes with the separation distance of 0.287 nm (as displayed in 1) and 0.237 nm (as displayed in 2) correspond to the (222) plane of cubic In2O3 (JCPDS No.71-2195) and (111) plane of Ag metal (JCPDS No.87-0598), respectively. EDS measurement was performed to further confirm the successful decoration of Ag nanoparticles on In2O3 PNS. In the obtained element maps (Figure 2e), besides the well dispersed In and O elements, agglomerated Ag element was also observed on the nano- The SEM and TEM were applied to observe the microstructure and morphology of the prepared samples. Figure 2a,b shows the representative SEM images of bare In 2 O 3 and Ag/In 2 O 3 -2, respectively. Both samples show similar sphere-like structure with the diameter about 120 nm, demonstrating that Ag modification has almost no influence on the basic morphology of primary In 2 O 3 . Figure 2c shows the TEM image of Ag/In 2 O 3 -2, from which the porous structure of the nanospheres is clearly observed. In the highresolution TEM (HRTEM) image recorded from the edge of a randomly selected nanosphere (Figure 2d), two definite crystalline phases were determined by measuring the lattice fringes. The lattice fringes with the separation distance of 0.287 nm (as displayed in (1) and 0.237 nm (as displayed in (2)   The surface chemical state of bare In2O3 and Ag/In2O3-2 was investi ( Figure 3). In their full XPS spectra (Figure 3a), the signals from In, O, and observed. The signals of C in the two samples are ascribed to the contami Compared with bare In2O3, additional signals arising from Ag 3d are Ag/In2O3-2, further confirming the successful introduction of Ag in In2 shows the high-resolution XPS spectra of In 3d. In the bare In2O3 sample, th binding energy of 451.65 and 444.11 eV belong to In 3d3/2 and In 3d5/2, re comparison with bare In2O3, the In 3d3/2 and In 3d5/2 peaks of Ag/In2O3-2 s higher binding energy, which are located at 451.77 and 444.22 eV, respect ferent binding energies of In 3d in the two samples suggest an electronic i tween Ag and In2O3 because of their different work functions [57]. In the h XPS spectrum of Ag 3d of Ag/In2O3-2 (Figure 3c), the Ag 3d5/2 (367.63 eV (373.63 eV) peaks with 6 eV splitting demonstrate the existence of Ag 0 [54,5 sistent with above TEM analysis. Compared with the standard binding ene (368.2 eV) and Ag 3d3/2 (374.2 eV) in metallic Ag, such two peaks in Ag slightly to the lower binding energy side, which can be also ascribed to the Ag with In2O3. Figure 3d shows the high-resolution XPS spectra of O1s. A lution, the signals of O1s can be divided into three peaks, including lattice at about 529.60 eV, surface adsorbed oxygen (Oads) at about 531.24 eV, an groups (OOH) at about 532.08eV [33,59], and their relative percentages ar inset. The estimated percentage of Oads in Ag/In2O3-2 is 48.48%, which is hi in bare In2O3 (34.79%). It was generally believed that during the gas sensin surface chemisorbed oxygen species, including O2 − , O − , and O 2− , can play t dant to react with reducing gas. Therefore, the higher Oads content is usually  Figure 3b shows the high-resolution XPS spectra of In 3d. In the bare In 2 O 3 sample, the peaks at the binding energy of 451.65 and 444.11 eV belong to In 3d 3/2 and In 3d 5/2 , respectively. In comparison with bare In 2 O 3 , the In 3d 3/2 and In 3d 5/2 peaks of Ag/In 2 O 3 -2 shift slightly to higher binding energy, which are located at 451.77 and 444.22 eV, respectively. The different binding energies of In 3d in the two samples suggest an electronic interaction between Ag and In 2 O 3 because of their different work functions [57]. In the high-resolution XPS spectrum of Ag 3d of Ag/In 2 O 3 -2 (Figure 3c), the Ag 3d 5/2 (367.63 eV) and Ag 3d 3/2 (373.63 eV) peaks with 6 eV splitting demonstrate the existence of Ag 0 [54,58], being consistent with above TEM analysis. Compared with the standard binding energy of Ag 3d 5/2 (368.2 eV) and Ag 3d 3/2 (374.2 eV) in metallic Ag, such two peaks in Ag/In 2 O 3 -2 shift slightly to the lower binding energy side, which can be also ascribed to the interaction of Ag with In 2 O 3 . Figure 3d shows the high-resolution XPS spectra of O1s. After deconvolution, the signals of O1s can be divided into three peaks, including lattice oxygen (O Latt ) at about 529.60 eV, surface adsorbed oxygen (O ads ) at about 531.24 eV, and surface OH groups (O OH ) at about 532.08eV [33,59], and their relative percentages are listed in the inset. The estimated percentage of O ads in Ag/In 2 O 3 -2 is 48.48%, which is higher than that in bare In 2 O 3 (34.79%). It was generally believed that during the gas sensing process, the surface chemisorbed oxygen species, including O 2 − , O − , and O 2− , can play the role of oxidant to react with reducing gas. Therefore, the higher O ads content is usually favorable for enhanced gas sensitivity.

Gas Sensing Performance
In order to realize an insight into the effects of Ag modification on the performance of In2O3, gas sensing tests were carried out on the bare and In2O3 sensors. In view of the fact that the gas sensing characteristics of a

Gas Sensing Performance
In order to realize an insight into the effects of Ag modification on the performance of In2O3, gas sensing tests were carried out on the bare and In2O3 sensors. In view of the fact that the gas sensing characteristics of a greatly rely on its working temperature, the temperature-dependent resi sensors as well as their responses to 20 ppm TEA were first measured. As

Gas Sensing Performance
In order to realize an insight into the effects of Ag modification on the TEA sensing performance of In 2 O 3 , gas sensing tests were carried out on the bare and Ag-modified In 2 O 3 sensors. In view of the fact that the gas sensing characteristics of a MOS sensor greatly rely on its working temperature, the temperature-dependent resistances of the sensors as well as their responses to 20 ppm TEA were first measured. As shown in Figure 5a, with the temperature increasing from 200 to 360 • C, the bare and Ag-modified In 2 O 3 sensors show similar response variation trends of "increase-maximum-decrease" and reach their maximum responses at 320 and 300 • C, respectively. At different operating temperatures, the Ag/In 2 O 3 sensors always show higher responses than the In 2 O 3 sensor, demonstrating the sensitization effect of Ag nanoparticles on In 2 O 3 PNSs. From Figure 5b, one can see that the sensors' resistance variation is similar to their response variation. Moreover, in the whole temperature range, the baseline resistances (R a ) of Ag/In 2 O 3 sensors are always higher than that of bare In 2 O 3 sensor. Based on the widely accepted oxygen adsorption theory [15,60], the temperature-dependent variation in sensor resistance and response may be associated with the adsorption and desorption behavior of oxygen molecules. When the operating temperature is relatively low (such as 200 • C), less oxygen molecules can adsorb on In 2 O 3 and Ag/In 2 O 3 to form chemisorbed oxygen due to their lower activity, which will lead to their lower resistance and response. With the increase in operating temperature, more and more oxygen molecules can acquire enough energy to adsorb on the materials and then transform into chemisorbed oxygen species, resulting in the gradually increased resistance and response of In 2 O 3 and Ag/In 2 O 3 . The adsorption and desorption of oxygen molecules on the In 2 O 3 and Ag/In 2 O 3 PNSs may reach their balance at 280 and 300 • C, respectively. At this moment, the amounts of chemisorbed oxygen species will reach their maximum values, leading to the maximum resistance and response, correspondingly. However, as the temperature exceeds 280 • C for In 2 O 3 and 300 • C for Ag/In 2 O 3 , the adsorption-desorption balance of oxygen will remove to the desorption side, which will decay the resistance and response due to the decreased amount of chemisorbed oxygen species. In Figure 5b, the higher resistance of Ag/In 2 O 3 sensors than the bare In 2 O 3 sensor can be understood from two aspects. Firstly, Ag nanoparticles can facilitate the formation of chemisorbed oxygen on In 2 O 3 through the spillover effect [61]. Thus, at different operating temperatures, more chemisorbed oxygen was created on Ag/In 2 O 3 than on bare In 2 O 3 . As is well known, the formation of chemisorbed oxygen on n-type MOS usually accompanies increased sensor resistance because of the decreased carrier (electron) concentration. Thus, more chemisorbed oxygen species can endow Ag/In 2 O 3 sensors with higher resistance. Secondly, at the interface of Ag and In 2 O 3 , electrons will migrate from In 2 O 3 to Ag on account of the lower work function of In 2 O 3 (4.3 eV) than that of Ag (4.6 eV), which will also lead to the higher resistance of Ag/In 2 O 3 sensors [57]. In addition, from Figure 5a,b, one can clearly observe that the temperature-dependent response variation of all Ag/In 2 O 3 sensors is strictly in accordance with their resistance variation, and both their response and resistance reach the maximum values at 300 • C, while, in contrast to the Ag/In 2 O 3 sensors, the bare In 2 O 3 sensor gives its maximum resistance at 280 • C, which is lower than its maximum response temperature (320 • C). Such a phenomenon is supposed to be relevant to the catalytic effect of Ag nanoparticles. In general, the response of a MOS sensor is mainly controlled by the gas sensing reaction occurring between chemisorbed oxygen and gaseous analyte. As the gas concentration of analyte is fixed, more chemisorbed oxygen species participating in the reaction usually result in higher sensor response. In the Ag/In 2 O 3 sensors, Ag nanoparticles can catalyze the reaction that occurred between chemisorbed oxygen and TEA, which means that at the temperature before 300 • C, most reactants (chemisorbed oxygen species and TEA molecules) can be activated to complete the gas sensing reaction. In this case, the amount of chemisorbed oxygen species on Ag/In 2 O 3 becomes the decisive factor of sensor response. Since the amount of chemisorbed oxygen of Ag/In 2 O 3 sensors reaches the maximum value at 300 • C (Figure 5b), their maximum response to TEA also appears at 300 • C, accordingly. While, for bare In 2 O 3 sensor, perhaps due to the lack of the catalytic effect of Ag nanoparticles, most of the chemisorbed oxygen species formed at 280 • C could be not active enough to react with TEA until the temperature increases to 320 • C. Therefore, it realizes the highest response at 320 • C but not at 280 • C. From Figure 5b, we can also observe that with the Ag content increases from 2.43 wt% (Ag/In 2 O 3 -1) to 2.90 wt% (Ag/In 2 O 3 -2) and 3.68 wt% (Ag/In 2 O 3 -3), the R a values of the sensor first increase and then decease. This phenomenon may be explained in that, when the Ag content is lower than 2.90 wt%, the Ag nanoparticles decorated on In 2 O 3 have good dispersion. In this case, the existence of more well-dispersed Ag nanoparticles will enhance the spillover effect, resulting in the higher Ra values of Ag/In 2 O 3 -2 than Ag/In 2 O 3 -1. Whereas, with the Ag content further increasing to 3.68 wt% (Ag/In 2 O 3 -3), aggregation between Ag nanoparticles can occur, which will weaken the spillover effect and lead to a lower R a of Ag/In 2 O 3 -3 than Ag/In 2 O 3 -2. Since the Ag/In 2 O 3 -2 sensor shows the highest response among different Ag-modified In 2 O 3 sensors (Figure 5a), in the following tests, it is chosen as the representative sensor to evaluate the effect of Ag modification on the gas sensing performance of In 2 O 3 by comparing with bare In 2 O 3 sensor.
In2O3 sensors (Figure 5a), in the following tests, it is chosen as the representative sensor to evaluate the effect of Ag modification on the gas sensing performance of In2O3 by comparing with bare In2O3 sensor. Figure 5c shows the responses of In2O3 and Ag/In2O3-2 sensors to 50 ppm various gases at 300 °C. Apparently, the responses of Ag/In2O3-2 for different gases are obviously higher than that of bare In2O3, further demonstrating the positive effect of Ag nanoparticles on the gas sensing performance of In2O3 PNSs. Additionally, both sensors exhibit much higher response to TEA than to other gases, revealing their good selectivity to TEA. To quantitatively evaluate the influence of Ag modification on the TEA selectivity of In2O3, the TEA selectivity coefficients of In2O3 and Ag/In2O3-2 were calculated by using STEA/Sinterference gas, where STEA and Sinterference gas refer to the responses of the sensors to TEA and interference gas, respectively. As depicted in Figure 5d, when using different gases as interference gas (ammonia, formaldehyde, methanol, ethanol, acetone, and methylbenzene), the Ag/In2O3-2 sensor always gives much higher selectivity coefficients than the bare In2O3 sensor. For example, the values of STEA/Sethanol and STEA/Sammonia of Ag/In2O3-2 are 5.30 and 932.43, which are about 3.33 and 45.55 times higher than that of In2O3, respectively. The higher response and better selectivity of Ag/In2O3-2 to TEA make it more suitable for practical application.  Figure 6a shows the dynamic resistance changes in In2O3 and Ag/In2O3-2 as they were exposed to 1-50 ppm of TEA at 300 °C. Both sensors show a rapid decay of resistance once exposure to TEA vapor, exhibiting the characteristic response of n-type MOS. With the increase in TEA concentration, the response amplitudes of the two sensors enlarge gradually (Figure 6b,c), while, for different concentrations of TEA, the Ag/In2O3-2 sensor gives much higher responses than the bare In2O3 sensor, demonstrating its superior ability for sensing TEA. For instance, the responses of Ag/In2O3-2 to 50  Figure 5c shows the responses of In 2 O 3 and Ag/In 2 O 3 -2 sensors to 50 ppm various gases at 300 • C. Apparently, the responses of Ag/In 2 O 3 -2 for different gases are obviously higher than that of bare In 2 O 3 , further demonstrating the positive effect of Ag nanoparticles on the gas sensing performance of In 2 O 3 PNSs. Additionally, both sensors exhibit much higher response to TEA than to other gases, revealing their good selectivity to TEA. To quantitatively evaluate the influence of Ag modification on the TEA selectivity of In 2 O 3 , the TEA selectivity coefficients of In 2 O 3 and Ag/In 2 O 3 -2 were calculated by using S TEA /S interference gas , where S TEA and S interference gas refer to the responses of the sensors to TEA and interference gas, respectively. As depicted in Figure 5d, when using different gases as interference gas (ammonia, formaldehyde, methanol, ethanol, acetone, and methylbenzene), the Ag/In 2 O 3 -2 sensor always gives much higher selectivity coefficients than the bare In 2 O 3 sensor. For example, the values of S TEA /S ethanol and S TEA /S ammonia of Ag/In 2 O 3 -2 are 5.30 and 932. 43, which are about 3.33 and 45.55 times higher than that of In 2 O 3 , respectively. The higher response and better selectivity of Ag/In 2 O 3 -2 to TEA make it more suitable for practical application. Figure 6a shows the dynamic resistance changes in In 2 O 3 and Ag/In 2 O 3 -2 as they were exposed to 1-50 ppm of TEA at 300 • C. Both sensors show a rapid decay of resistance once exposure to TEA vapor, exhibiting the characteristic response of n-type MOS. With the increase in TEA concentration, the response amplitudes of the two sensors enlarge gradually (Figure 6b,c), while, for different concentrations of TEA, the Ag/In 2 O 3 -2 sensor gives much higher responses than the bare In 2 O 3 sensor, demonstrating its superior ability for sensing TEA. For instance, the responses of Ag/In 2 O 3 -2 to 50 ppm TEA are as high as 5697, about 158 times higher than that of bare In 2 O 3 (36). In the tested TEA concentration range, both the bare In 2 O 3 and Ag/In 2 O 3 -2 sensors show good response linearity (the insets in Figure 6b,c). Based on the slop of their fitting line, the sensitivity of In 2 O 3 and Ag/In 2 O 3 -2 to 1-50 ppm TEA is determined as 0.69 and 116.86/ppm, respectively. Apparently, after modification with Ag, the sensitivity of our In 2 O 3 sensor to TEA is dramatically increased (about 169.36 times). Since the Ag/In 2 O 3 -2 sensor can give a high response (24.3), even to 1 ppm TEA, its response to ppb level TEA was further tested. As shown in Figure.6d, with increasing TEA concertation from 50 to 600 ppb, the sensor shows a gradually increased response. The response to 50 ppb TEA is as high as 3.4, demonstrating its strong sensing ability for TEA at ppb level. The response and recovery speed of bare In 2 O 3 and Ag/In 2 O 3 -2 were also measured from their transient resistance response-recovery curves to 50 ppm TEA (Figure 6e,f), whose results show that the response/recovery time (τ res. /τ rec. ) of In 2 O 3 and Ag/In 2 O 3 -2 are 5/770s and 6/350s, respectively. Perhaps because of their relatively higher operating temperature, both sensors can give fast response to TEA, while, compared with bare In 2 O 3 sensor, the Ag/In 2 O 3 -2 sensor shows much shorter recovery time. Theoretically, the recovery time of a MOS sensor to garget gas is mainly controlled by two factors, including the desorption speed of the product of gas sensing reaction and the regeneration speed of chemisorbed oxygen on the surface of MOS. In our case, the faster recovery speed of Ag/In 2 O 3 -2 may mean that Ag nanoparticle can not only accelerate the regeneration of chemisorbed oxygen through its strong spillover effect, but also can catalyze the gas sensing reaction to produce some products that can easily desorb from In 2 O 3 PNS. To further evaluate the quality of Ag/In 2 O 3 -2, its TEA sensing performances were compared with that of the previously reported sensors. As displayed in Table 1, the present Ag/In 2 O 3 -2 sensor shows much higher response and better selectivity to TEA than most of the reported sensors, demonstrating its superiority in TEA detection.  The repeatability, long-term stability, and humidity resistance of the Ag/In2O3-2 sensor were also tested to further evaluate its quality. As shown in Figure 7a,b, in five continues response-recovery cycles test results, the sensor can give almost the same response amplitudes to 20 ppm TEA; moreover, within a period of 30 days, there is only a slight decay in its response, revealing its good repeatability and stability. Figure 7c presents the response values of Ag/In2O3-2 sensor toward 20 ppm TEA under different relative humidity (RH). With increasing RH from 25 to 80%, the response of the sensor decreases seriously. Under humidity conditions, water molecules can absorb on the surface of Ag/In2O3 PNSs and then compete with TEA molecules to react with chemisorbed oxygen [25], resulting in a decreased response of TEA under high humidity. According to the research results of Lee's group, doping our In2O3 PNSs with Pr before modification The repeatability, long-term stability, and humidity resistance of the Ag/In 2 O 3 -2 sensor were also tested to further evaluate its quality. As shown in Figure 7a,b, in five continues response-recovery cycles test results, the sensor can give almost the same response amplitudes to 20 ppm TEA; moreover, within a period of 30 days, there is only a slight decay in its response, revealing its good repeatability and stability. Figure 7c presents the response values of Ag/In 2 O 3 -2 sensor toward 20 ppm TEA under different relative humidity (RH). With increasing RH from 25 to 80%, the response of the sensor decreases seriously. Under humidity conditions, water molecules can absorb on the surface of Ag/In 2 O 3 PNSs and then compete with TEA molecules to react with chemisorbed oxygen [25], resulting in a decreased response of TEA under high humidity. According to the research results of Lee's group, doping our In 2 O 3 PNSs with Pr before modification with Ag nanoparticles may be a possible strategy to improve the humidity resistance of the present Ag/In 2 O 3 sensor [65], which will be tried in our future work. with Ag nanoparticles may be a possible strategy to improve the humidity resistance o the present Ag/In2O3 sensor [65], which will be tried in our future work.

Gas Sensing Mechanism
In2O3 is a typical n-type MOS, and its gas sensing phenomenon can be illustrated by widely accepted oxygen adsorption theory [66,67]. As schematically illustrated in Figure  8, when our In2O3 PNS sensor is exposed in air atmosphere, oxygen molecules can adsorb on the surface of In2O3 PNS and then capture electrons from its conduction band to form chemisorbed oxygen species (Equations (1)-(4)). As a result, a thick electron depletion layer (EDL) with higher resistance will form on the surface of In2O3 PNS, leading to its higher sensor resistance in air (Ra) (Figure 8a). Meanwhile, as the sensor is exposed to reducing gases, such as TEA, a redox reaction (gas sensing reaction) between TEA and chemisorbed oxygen species will occur, after which the electrons captured by chemi sorbed oxygen will be released back to In2O3, resulting in a thinner EDL and a decreased sensor resistance (Rg), accordingly (Figure 8c). The varied resistance of the In2O3 PNS sensor in air and target gas endows it with the gas sensing ability. In our experiment, the Ag/In2O3-2 sensor, like the bare In2O3 sensor, shows the characteristic response of an n-type MOS, but its gas sensing performance to TEA was remarkably boosted, especially in terms of lower operating temperature, higher sensitiv ity, and better selectivity. These improvements in gas sensing performance can be mainly attributed to the sensitization effects of Ag nanoparticles on In2O3 PNSs. Firstly, Ag na noparticle can promote the dissociation of oxygen molecules to form chemisorbed oxy gen through its strong "spillover effect" [53] (Figure 8c). That is to say, in air atmos pheres, more chemisorbed oxygen species (O 2− ) will be created on Ag-decorated In2O PNSs than on the bare In2O3 counterpart (the XPS spectra of O 1s in Figure 3d indicate that Ag/In2O3-2 owns higher Oads content than bare In2O3, supporting this speculation) On one hand, the increase in chemisorbed oxygen can endow the Ag/In2O3-2 sensor with higher Ra value (Figure 5b), which is favorable for higher response because the response of the Ag/In2O3 sensor to TEA is defined as Ra/Rg. On the other hand, it will promote the surface sensing reaction due to the increased concentration of O 2− reactant, which also contributes to the enhancement in TEA response. Additionally, the faster recovery speed of Ag/In2O3-2 ( Figure 6f) may also be relevant to the "spillover effect" of Ag nanoparti cles, which can accelerate the regeneration of chemisorbed oxygen after switching the sensor from TEA to air ambient. Secondly, Ag has good catalytic property. When deco

Gas Sensing Mechanism
In 2 O 3 is a typical n-type MOS, and its gas sensing phenomenon can be illustrated by widely accepted oxygen adsorption theory [66,67]. As schematically illustrated in Figure 8, when our In 2 O 3 PNS sensor is exposed in air atmosphere, oxygen molecules can adsorb on the surface of In 2 O 3 PNS and then capture electrons from its conduction band to form chemisorbed oxygen species (Equations (1)-(4)). As a result, a thick electron depletion layer (EDL) with higher resistance will form on the surface of In 2 O 3 PNS, leading to its higher sensor resistance in air (R a ) (Figure 8a). Meanwhile, as the sensor is exposed to reducing gases, such as TEA, a redox reaction (gas sensing reaction) between TEA and chemisorbed oxygen species will occur, after which the electrons captured by chemisorbed oxygen will be released back to In 2 O 3 , resulting in a thinner EDL and a decreased sensor resistance (R g ), accordingly (Figure 8c). The varied resistance of the In 2 O 3 PNS sensor in air and target gas endows it with the gas sensing ability.
In our experiment, the Ag/In 2 O 3 -2 sensor, like the bare In 2 O 3 sensor, shows the characteristic response of an n-type MOS, but its gas sensing performance to TEA was remarkably boosted, especially in terms of lower operating temperature, higher sensitivity, and better selectivity. These improvements in gas sensing performance can be mainly attributed to the sensitization effects of Ag nanoparticles on In 2 O 3 PNSs. Firstly, Ag nanoparticle can promote the dissociation of oxygen molecules to form chemisorbed oxygen through its strong "spillover effect" [53] (Figure 8c). That is to say, in air atmospheres, more chemisorbed oxygen species (O 2− ) will be created on Ag-decorated In 2 O 3 PNSs than on the bare In 2 O 3 counterpart (the XPS spectra of O 1s in Figure 3d indicate that Ag/In 2 O 3 -2 owns higher O ads content than bare In 2 O 3 , supporting this speculation). On one hand, the increase in chemisorbed oxygen can endow the Ag/In 2 O 3 -2 sensor with higher R a value (Figure 5b), which is favorable for higher response because the response of the Ag/In 2 O 3 sensor to TEA is defined as R a /R g . On the other hand, it will promote the surface sensing reaction due to the increased concentration of O 2− reactant, which also contributes to the enhancement in TEA response. Additionally, the faster recovery speed of Ag/In 2 O 3 -2 ( Figure 6f) may also be relevant to the "spillover effect" of Ag nanoparticles, which can accelerate the regeneration of chemisorbed oxygen after switching the sensor from TEA to air ambient. Secondly, Ag has good catalytic property. When decorating Ag nanoparticles on In 2 O 3 PNS, they can catalytic the gas sensing reaction that occurs between TEA and chemisorbed oxygen species (Figure 8d), leading to a lower reaction barrier and a decreased operating temperature, accordingly. As in our discussion on Figure 4a,b, due to the catalytic effect of Ag nanoparticles, the Ag/In 2 O 3 sensors show the maximum resistance and response at the same temperature (300 • C); in contrast, the bare In 2 O 3 sensor shows higher temperature of maximum response (320 • C) than that of its maximum resistance (280 • C). Such a difference between bare In 2 O 3 and Ag/In 2 O 3 sensors supports, to some extent, the existence of catalytic effect of Ag nanoparticles in the Ag/In 2 O 3 sensors. To prove above speculation, the apparent activation energies (E a ) of bare In 2 O 3 and Ag/In 2 O 3 -2 were estimated by their Arrhenius-type plots (Figure 9c,d) that derived from Figure 9a,b [17,61]. The results show that the E a values of bare In 2 O 3 and Ag/In 2 O 3 -2 are 11.819 and 7.404 kJ/mol, respectively. The lower E a of Ag/In 2 O 3 -2 than that of bare In 2 O 3 convinces the existence of catalytic effect of Ag, which not only endows the Ag/In 2 O 3 -2 sensor with lower OWT, but also endows it with higher response for sensing TEA. Thirdly, when decorating Ag nanoparticles on the surface of In 2 O 3 PNSs, the Schottky junction between Ag and In 2 O 3 will be created. Since the work function of In 2 O 3 (4.3 eV) is smaller than that of Ag (4.6 eV) [57], at the interface of Ag nanoparticles and In 2 O 3 PNSs, electrons will flow from In 2 O 3 to Ag to equilibrate their Fermi level ( Figure 8b). Consequently, on the In 2 O 3 side, besides the EDL caused by chemisorbed oxygen, an additional EDL will be also formed, leading to a thicker EDL on Ag/In 2 O 3 and a higher response than bare In 2 O 3 accordingly.
In addition, both our In 2 O 3 and Ag/In 2 O 3 sensors show good selectivity to TEA. Generally, the gas selectivity of a MOS sensor can be influenced by many factors, such as the adsorption property of target gas on the surface of MOS, the reactivity of reactants (chemisorbed oxygen species and adsorbed target gas), the surface catalytic property of MOS, and so on. In our case, the relatively lower bond energy of C-N should be one of the possible reasons for the good TEA selectivity of the In 2 O 3 and Ag/In 2 O 3 sensors. The bond energy of C-N in TEA is 307 kJ/mol, which is lower than that of C=C (toluene, 610.3 kJ/mol), N-H (ammonia, 386 kJ/mol), H-O (ethanol and methanol, 458 kJ/mol), and C=O (formaldehyde and acetone, 798.9 kJ/mol) [61]. The lowest bond energy of C-N endows TEA molecules with high reaction activity during the gas sensing process, leading to good TEA selectivity. reaction barrier and a decreased operating temperature, accordingly. As in our discussion on Figure 4a,b, due to the catalytic effect of Ag nanoparticles, the Ag/In2O3 sensors show the maximum resistance and response at the same temperature (300 °C); in contrast, the bare In2O3 sensor shows higher temperature of maximum response (320°C) than that of its maximum resistance (280 °C). Such a difference between bare In2O3 and Ag/In2O3 sensors supports, to some extent, the existence of catalytic effect of Ag nanoparticles in the Ag/In2O3 sensors. To prove above speculation, the apparent activation energies (Ea) of bare In2O3 and Ag/In2O3-2 were estimated by their Arrhenius-type plots (Figure 9c,d) that derived from Figure 9a,b [17,61]. The results show that the Ea values of bare In2O3 and Ag/In2O3-2 are 11.819 and 7.404 kJ/mol, respectively. The lower Ea of Ag/In2O3-2 than that of bare In2O3 convinces the existence of catalytic effect of Ag, which not only endows the Ag/In2O3-2 sensor with lower OWT, but also endows it with higher response for sensing TEA. Thirdly, when decorating Ag nanoparticles on the surface of In2O3 PNSs, the Schottky junction between Ag and In2O3 will be created. Since the work function of In2O3 (4.3 eV) is smaller than that of Ag (4.6 eV) [57], at the interface of Ag nanoparticles and In2O3 PNSs, electrons will flow from In2O3 to Ag to equilibrate their Fermi level (Figure 8b). Consequently, on the In2O3 side, besides the EDL caused by chemisorbed oxygen, an additional EDL will be also formed, leading to a thicker EDL on Ag/In2O3 and a higher response than bare In2O3 accordingly. In addition, both our In2O3 and Ag/In2O3 sensors show good selectivity to TEA. Generally, the gas selectivity of a MOS sensor can be influenced by many factors, such as the adsorption property of target gas on the surface of MOS, the reactivity of reactants (chemisorbed oxygen species and adsorbed target gas), the surface catalytic property of MOS, and so on. In our case, the relatively lower bond energy of C-N should be one of the possible reasons for the good TEA selectivity of the In2O3 and Ag/In2O3 sensors. The bond energy of C-N in TEA is 307 kJ/mol, which is lower than that of C=C (toluene, 610.3 kJ/mol), N-H (ammonia, 386 kJ/mol), H-O (ethanol and methanol, 458 kJ/mol), and C=O (formaldehyde and acetone, 798.9 kJ/mol) [61]. The lowest bond energy of C-N endows TEA molecules with high reaction activity during the gas sensing process, leading to good TEA selectivity.

Conclusions
In summary, well-dispersed Ag nanoparticles with a size of about 25-30 n decorated on In2O3 PNSs (~100 nm in diameter) via a room-temperature chemica tion method. After modification with Ag nanoparticles, the In2O3 PNSs sensor remarkable improvements in TEA sensing performance, especially of lower OW creased from 320 to 300°C), higher sensitivity (increased from 0.69/ppm to 116. for 1-50 ppm TEA), and better selectivity (STEA/Sethanol increased from 1.59 to 5.30 improved gas sensing performances of as-prepared Ag/In2O3 were contributed positive effects of Ag, the spillover effect, and electron sensitization of the A Schottky junction, as well as the catalysis effect of Ag on the surface gas sensing between TEA and chemisorbed oxygen. Our research strongly demonstrates tha fication of In2O3 with Ag is a promising strategy to develop an advanced TEA sen

Conclusions
In summary, well-dispersed Ag nanoparticles with a size of about 25-30 nm were decorated on In 2 O 3 PNSs (~100 nm in diameter) via a room-temperature chemical reduction method. After modification with Ag nanoparticles, the In 2 O 3 PNSs sensor showed remarkable improvements in TEA sensing performance, especially of lower OWT (decreased from 320 to 300 • C), higher sensitivity (increased from 0.69/ppm to 116.86/ppm for 1-50 ppm TEA), and better selectivity (S TEA /S ethanol increased from 1.59 to 5.30). These improved gas sensing performances of as-prepared Ag/In 2 O 3 were contributed to the positive effects of Ag, the spillover effect, and electron sensitization of the Ag-In 2 O 3 Schottky junction, as well as the catalysis effect of Ag on the surface gas sensing reaction between TEA and chemisorbed oxygen. Our research strongly demonstrates that modification of In 2 O 3 with Ag is a promising strategy to develop an advanced TEA sensor.