Pd-loaded In₂O₃ nanowire-like network synthesized using carbon nanotube templates for enhancing NO₂ sensing performance

Abstract

Pd-loaded In₂O₃ nanowire (NW)-like networks were synthesized via electroless plating using carbon nanotubes (CNTs) as templates, followed by oxidation and removal of the CNTs at 550 °C. Palladium (Pd) was introduced to activate the surface of the CNTs for subsequent plating. Before calcination, Pd was loaded onto In₂O₃. The as-synthesized Pd-loaded In₂O₃ replicated the structure of the CNTs, forming a porous NW-like network with a very large specific surface area. Furthermore, the NO₂ gas sensing properties of the Pd-loaded In₂O₃ NW-like network, porous Pd–In₂O₃ and porous unloaded-In₂O₃ were investigated. The results demonstrated that the Pd–In₂O₃ NW-like network exhibits superior sensitivity with short response and recovery times, and demonstrates a significant response when exposed to NO₂ at concentrations as low as 5 ppm at a temperature of 110 °C. A synergy of electric and chemical effects has been proposed to explain the gas sensing enhancement.

---

Introduction

Nitrogen dioxide (NO₂), as one of the major constituents of photochemical smog, is harmful to the environment and can also cause health problems. In order to detect and control NO₂, it is crucial to develop sensors with a low detection limit, low operating temperature and rapid response and recovery times. Gas sensors are very useful for many applications, such as detection of toxic and deleterious gases, pollution abatement, public safety, and process control. There is an urgent demand for the development of high quality gas sensitive materials.^1–4 Indium oxide (In₂O₃), as a typical n-type semiconductor with a direct band gap of 3.55–3.75 eV, could potentially be used to detecting H₂S, CO, C₂H₅OH and NO₂, because of its high electrical conductivity, abundant defects and autocatalysis.^5–7 In recent years, In₂O₃ has been widely studied as a NO₂ gas sensor. Although the maximum response has been substantially improved, the response and recovery times are generally too long. Furthermore, the gas sensing mechanism is still unclear and further studies are required.

It is acknowledged that the properties of sensors depend significantly on their morphology and structure. In order to improve the sensitivity and shorten the response and recovery times, In₂O₃ has been fabricated in various morphologies including nanoparticles,⁸ nanowires,⁹ nanotubes,¹⁰ nanorods,¹¹ nanobelts,¹² and hollow spheres.¹³ Considering the current investigations on the above nanomaterials, quasi-1D nanomaterials (including nanotubes, nanowires, and nanobelts) would be expected to be the most promising structures for sensors because of their high surface to volume ratio and porous structure for gas diffusion.^14–17

Until now, quasi-1D In₂O₃ has been prepared using several methods, such as sol–gel chemistry, electrospinning,^18–20 and the solid template method. The template method is commonly used to synthesize quasi-1D metal oxides. CNTs are ideal templates to synthesize quasi-1D metal oxides, because of their excellent acid and alkali resistance. Moreover, they can be removed at high temperatures. In earlier studies, Ning Du et al.²¹ prepared porous In₂O₃ nanotubes with polyelectrolytes using layer-by-layer assembly/deposition on CNTs, followed by calcination. Zhang²² developed a novel strategy for the synthesis of CeO₂ nanotubes. This was achieved by coating the CNTs with a continuous layer of CeO₂ nanoparticles using the solvothermal method in a pyridine solution, followed by the removal of the CNTs. Sheng Yi and his co-workers¹⁶ used screen printing technology and calcination to obtain an In₂O₃ NW-like network using CNT templates. According to these studies, CNTs play a critical role as removable templates in the preparation of quasi-1D metal oxides. However, many organic reagents are necessary for these methods, which restricts the practical applications of CNT templates. In addition, many strategies have been developed to enhance gas sensing properties. One of the most simple and effective methods is to incorporate noble metals into the In₂O₃ structure, such as Ag,²³ Au,²⁴ Ru,²⁵ Pt²⁶ and Pd. In particular, palladium (Pd) has been widely applied to enhance the sensitivity and to shorten the response and recovery times of gas sensors.^27–31

Pd-loaded In₂O₃ may theoretically improve sensing performance. Kim et al.³² fabricated Pd-functionalized networked In₂O₃ nanowires by depositing Pd layers using a sputtering method on bare In₂O₃ nanowires. Tian et al.¹⁶ synthesized In₂O₃ NW-like network using sacrificial CNTs templates and screen printing technology, followed by calcination, which requires large amounts of organic reagents to activate the surface of CNTs. Here, we report a novel approach to synthesize a Pd-loaded In₂O₃ NW-like network. This was conducted by electroless plating using CNTs as templates in combination with subsequent calcination. During the procedure, Pd was introduced to activate the surface of the CNTs for the subsequent plating of indium. Moreover, following the calcination of the CNTs, which were coated with indium, the CNTs were oxidized into CO₂. This was accompanied by the transformation of metal indium into indium oxide, Additionally, Pd was loaded in situ onto In₂O₃. The response and recovery time of the gas sensor was shortened as a result of the spillover mechanism. This approach can be applied in order to synthesize other MOS NW-like networks by plating different metals onto CNTs.

Experimental

Materials

Carboxylic multi-walled CNTs used in this work were supplied from Nanjing xfnano Technology Co. Ltd. The other chemical reagents, stannous chloride hydrate (SnCl₂·2H₂O), palladium chloride (PdCl₂), hydrochloric acid (HCl), indium chloride hydrate (InCl₃·4H₂O), sodium borohydride (NaBH₄), and absolute ethanol (C₂H₅OH), were purchased from Tianjin east Reagent Factory. All of the materials were of analytical grade and used without further purification.

Preparation of the Pd–In₂O₃ NW-like network

The complicated network is comprised of NW-like structures, which are interlaced and connected with each other. The detailed formation of an individual In₂O₃ NW-like structure is schematically illustrated in Fig. 1. The procedure is as follows: (1) carboxylic-MWCNTs were sensitized and activated; (2) a complex of InCl₃ and citric acid was added into the solution of Pd-modified CNTs; (3) NaBH₄ aqueous solution was slowly dropped into this mixed solution to reduce the In³⁺ ions into In on the surface of the CNTs; (4) the CNTs coated with Pd–In were oxidized in a drying oven, and (5) Pd–In₂O₃ NW-like network structures were obtained by calcination as a result of the collapse of the CNTs and the crystallization of In₂O₃. Each specific step in the formation of the as-synthesized Pd–In₂O₃ NW-like network is described in the following section.

Fig. 1 Schematic diagram for the preparation process of the individual NW-like In₂O₃.

A typical procedure to prepare modified CNTs with Pd consists of two steps, of which the first is sensitization. 0.25 g SnCl₂·2H₂O was dissolved in 25 mL 0.5 M HCl solution, followed by stirring at room temperature using a magnetic stirrer. 3 mg CNTs were then added into the solution. After being sonicated for 15 min, the excess SnCl₂ was removed by washing with deionized water. Subsequently, the sensitized CNTs were immersed in 50 mL aqueous mixture of 2.5 mg PdCl₂ and 0.1 M HCl for another 15 min at ambient temperature for activation. The Pd-modified CNTs were then washed several times with water and collected by natural sedimentation after 5 min standing. The collected CNT mixture was dried at 90 °C for 6 h.

The modified CNTs were added to 50 mL water with 0.2 g InCl₃ and 0.1 g citric acid, which was sonicated for 0.5 h. At the same time, 0.3 g NaBH₄ was dissolved in 30 mL DI water by magnetic stirring for 10 min. The NaBH₄ aqueous solution was then slowly dropped into the aforementioned solution at 40 °C. After 30 min, the resulting black solid products were centrifuged, washed with distilled water and ethanol to remove any ions remaining in the final products, and dried at 90 °C for 60 h in air. During this period, the surface of metallic indium was oxidized into trivalent indium. Finally, grayish-yellow solid products were obtained after calcination at 550 °C for 2 h in ambient air.

Material characterization

The crystalline phase of the samples was characterized by X-ray diffraction (XRD, RIGAKU/DMAX) with CuKα1 radiation (λ = 0.15406 nm) at a scanning rate of 2° at 2θ min⁻¹ ranging from 10° to 90°. The Raman spectroscopy measurement was carried out on a Thermo Scientific DXR Raman microscope with a 532 nm laser. The BET surface area measurements were performed by nitrogen adsorption on a Quantachrome NOVA 2000 surface analyzer. XPS analysis was performed using a PHL1600ESCA instrument equipped with an onochromatic Mg Kα X-ray source. EDS analysis was performed using X-Max80 (Oxford, UK). The surface morphologies and structures of samples were investigated by field emission scanning electron microscopy (FE-SEM, Hitachi S-4800) and transmission electron microscopy (TEM, Philips Tecnai G2 F20).

Gas sensing measurements and fabrication of gas sensors

The sensor testing system NS-4003 series (Zhong-Ke Micro-nano IOT (Internet of Things) Ltd, China) was used to evaluate the gas sensing performance. The basic process for the fabrication of a typical Pd–In₂O₃-based gas sensor is as follows. The as-prepared Pd–In₂O₃ powder was slightly ground with absolute ethanol in an agate mortar to form a gas sensing paste. The sensor was fabricated by coating the paste directly onto the ceramic tube with four Pt wires attached to a pair of Au electrodes. A Ni–Cr alloy coil was inserted into the ceramic tube as a heater to control the operating temperature. The sensors were dried at 60 °C and aged at 250 °C for 12 h. The gas sensing test was carried out in the glass test chamber, and the volume of the test chamber is 10 L. The gas concentration was adjusted by changing the amount of the target gases. The response of the sensor to target gas was defined as Response = R_gas/R_air, where R_gas and R_air are the resistance values of the sensor measured in the target gas and air, respectively. The response time was defined as the time required for the variation in resistance to reach 90% of the equilibrium value after a target gas was introduced and the recovery time is the time necessary for a sensor to attain a resistance 10% above its original value when the target gas was replaced by air.

Results and discussion

The morphology of pure CNTs was observed in TEM images, which are shown in Fig. 2a and b. It could be clearly seen that the CNTs have outer diameters (OD) of about 18 nm and a relatively smooth surface without any secondary nanostructures. After sensitization and activation, uniformly dispersed nanoparticles are attached to the surface of the CNTs, as shown in Fig. 2c. Fig. 2d is a typical HRTEM image of an individual modified CNT. There are lots of dispersed Pd nanoparticles on the surface of the CNTs. These particles act as catalysts in the indium reduction reaction,^33,34 which makes the rate of indium deposition on the CNTs surface much faster than that in solution. In other words, indium will be deposited preferentially on the CNTs, rather than in the solution. Once the metallic indium deposition was initiated, the deposited indium will act as a self-catalyst for further deposition.³⁵

Fig. 2 (a) TEM image and (b) HRTEM image of pure CNTs; (c and d) morphological and compositional characterization of CNTs modified with Pd by sensitization and activation: (c) TEM image and (d) HRTEM image of an individual modified CNT.

Fig. 3a (FE-SEM image) shows the general morphology of the Pd-modified CNTs that were coated by indium (Pd–In/CNTs) via electroless plating. It can be observed that the surfaces of the products are obviously coarser than those of the pure CNTs. The CNTs that were coated by indium interlace, forming a specific porous network structure. The morphology of the products was further characterized by TEM (Fig. 3b). It can be observed that the modified CNTs are coated by a continuous layer with a bumpy surface. The HRTEM image (shown in the inset of Fig. 3b) demonstrates that the coating layer is amorphous with an approximate size of 4 nm.

Fig. 3 Morphological characterizations of Pd–In/CNTs: (a) SEM image and (b) TEM image, inset in (b) is HRTEM image.

Subsequently, the Pd–In/CNTs were oxidized in air at a temperature of 90 °C. Raman spectroscopy (Fig. 4a) indicates that the pure MWCNTs mainly exhibit D, G, and G* peaks. As the oxidation time increased from 20 to 60 h, with the exception of D, G, and G* peaks, other Raman signals appear between 200 and 650 cm⁻¹. At 60 h, five Raman scattering peaks located at 130, 304, 360, 490, and 628 cm⁻¹ can be clearly observed.³⁶ These correlate to the reference for In₂O₃. Within the XRD patterns (Fig. 4b), the diffractions can all be readily indexed to metallic In (JCPDS 85-1409), except for the peaks at 2θ = 30.6° and 26.4°. These two peaks can be indexed to the (2 1 1) crystal plane of In₂O₃ (JCPDS 71-2194) and the (0 0 2) plane of the CNTs, respectively. No peak arising from SnO₂ or PdO_x/Pd can be detected, because of their low content and small grain size. Calculating the areas under the In (1 0 1) and In₂O₃ (0 0 2) peaks for the samples that were oxidized for 20, 40, and 60 h at a temperature of 90 °C, it can be speculated that an increase in the oxidation time leads to a gradual increase in the crystallinity of In and In₂O₃. This indicates that amorphous In can crystallize at 90 °C and that In was only partially oxidized into In₂O₃. Specifically, the layer coated onto the CNTs is mainly of metallic In, even though the Pd–In/CNTs were oxidized for 60 h. A possible reason for this phenomenon is that metallic In would be oxidized slowly in air at 90 °C and thus form a layer of ultrathin oxidation film, which would prevent the further oxidation of In.

Fig. 4 (a) Raman spectra and (b) XRD patterns of Pd–In/CNTs treated for different oxidation times.

In order to obtain a Pd–In₂O₃ NW-like network, the CNTs should be removed. Bian³⁷ and his coworkers reported that CNTs could be completely oxidized at 500 °C in an air atmosphere. In this study, calcination was conducted at a temperature of 550 °C for 2 h in order to remove the CNTs. Fig. 5 demonstrates the morphologies of the obtained Pd–In₂O₃ samples, which had various oxidation times (20, 40, and 60 h). It can be observed that as the oxidation time increased to 40 h (Fig. 5c and d) or 60 h (Fig. 5e and f), the as-fabricated Pd–In₂O₃ replicate the structure of the CNTs, resulting in the formation of a porous NW-like network. In comparison with Fig. 5d, Fig. 5f exhibits a smaller grain size and more porosity. As shown in Fig. 5a and b, at an oxidation time of 20 h, the Pd–In₂O₃ also replicates the structure of the CNTs. However, this Pd–In₂O₃ structure is comprised of uniform nanoparticles with a size of 20 nm. As a result of the growth of the nanoparticles, it exhibits a porous structure rather than a porous NW-like network. So the as-obtained product was defined as porous Pd–In₂O₃. Furthermore, as the oxidation time increased from 20 to 60 h, the average size of the nanoparticles clearly decreased. We can clearly observe well-defined lattice fringes in Fig. 5a and the fringe spacings are approximately 0.289 and 0.248 nm, which correspond to (2 2 2) and (4 0 0) crystal planes of cubic In₂O₃. In the inset of Fig. 5e, the image of the diffraction rings demonstrates that the products in this study are composed of In₂O₃. The morphology of the Pd–In₂O₃ NW-like network was further observed in Fig. 5f, which clearly shows that a large quantity of extremely thin NW-like In₂O₃ are interwoven into a porous network. Moreover, within the inset of Fig. 5f, some individual NW-like In₂O₃ with a diameter of 8 nm can be clearly observed. There are many pores throughout the network. These pores will significantly improve the properties of the gas sensors, because of their larger surface-to-volume ratios and because there are more channels for gas diffusion.^16,38 To conclude, an oxidation time of 60 h was chosen to prepare the Pd–In₂O₃ NW-like network.

Fig. 5 TEM images of morphological and compositional characterization of as-obtained Pd–In₂O₃ nanostructures by the oxidization of Pd–In/CNT for different time periods: (a and b) 20 h, (c and d) 40 h and (e and f) 60 h, then followed by calcination at 550 °C in air for 2 h.

It should be noticed that the morphology of In₂O₃ changed from near-spherical nanoparticles into irregular nanoparticles. This may be because metallic In has a rapid diffusion rate when the temperature exceeds its melting point. Due to surface tension, molten In tends to form a near-spherical structure in order to achieve the lowest surface free energy. However, as shown in the Raman spectra and XRD patterns, the amount of In₂O₃ increases with oxidation time. We therefore proposed a possible reason for the formation of the NW-like structure. First, a layer of In₂O₃ film is formed on the In surface at a temperature of 90 °C (in air) and, subsequently, the thickness of this In₂O₃ film increases with oxidation time. At an oxidation time of 60 h, the In₂O₃ film is thick enough to restrict the flow of molten indium and its diffusion at high temperature. The size and structure of the metallic In is thus maintained. Subsequently, the internal In is further oxidized into In₂O₃, which is accompanied by collapse of the CNTs and contraction of the In₂O₃, forming an individual NW-like In₂O₃ structure. This reveals that oxidation time plays an important role in the synthesis of the Pd–In₂O₃ NW-like network.

To determine the existence of the element Pd in the as-obtained Pd–In₂O₃ NW-like network, EDS analysis was carried out. Fig. 6 shows the EDS analysis of the obtained Pd–In₂O₃. The EDS spectrum shown in Fig. 6a indicates that the composite consists of In, O and Pd. Moreover, a trace of Sn was observed due to the sensitization of SnCl₂. The carbon and Cu peaks are attributed to the copper–carbon grid. One point that should be emphasized is that the effects of Sn will not be considered in the following performance testing, because all the samples were sensitized under the same sensitization condition. EDS elemental maps of In, O, and Pd are shown in Fig. 6c–e, respectively. The recorded images of In and O mapping contribute to In₂O₃. In addition, the elemental map of Pd corresponds with the STEM image shown in Fig. 6b, indicating that Pd is uniformly distributed in In₂O₃. In addition, the state of Pd in Pd–In₂O₃ NW-like network was investigated by XPS analysis as shown in Fig. S1.† It reveals that palladium has been partially oxidized at 550 °C, forming PdO and PdO_x/Pd. Only a proportion of Pd can therefore enhance the gas sensing properties.

Fig. 6 (a) EDS analysis spectrum and (b) STEM image of Pd–In₂O₃ NW-like network, (c–e) EDS mapping of In, O and Pd.

The porous unloaded In₂O₃ was produced using the same method as for the porous Pd–In₂O₃, except for the absence of PdCl₂ (displayed in Fig. S2†). It is noteworthy that the morphology and grain size of the porous unloaded In₂O₃ is essentially the same as that of the porous Pd–In₂O₃. Its precursor (In/CNTs) was also studied using a TEM, as shown in Fig. S3.† This indicated Pd also plays a significant role in the preparation of the Pd–In₂O₃ NW-like network.

In order to facilitate subsequent gas sensing measurement, SEM images of as-obtained porous unloaded-In₂O₃, porous Pd–In₂O₃ and the Pd–In₂O₃ NW-like network are shown in Fig. 7a–c and XRD patterns of these are merged in Fig. 7d. All the reflections in the pattern can be easily assigned to a pure cubic In₂O₃ (JCPDS 71-2194) without peaks arising from other phases such as PdO_x/Pd and SnO₂. The peak intensity of porous Pd–In₂O₃ and porous unloaded-In₂O₃ are comparable to, but higher than, that of the Pd–In₂O₃ NW-like network. This effect may be due to the grain size and crystallinity of the obtained products, which are in accordance with the HRTEM results.

Fig. 7 SEM images of (a) porous unloaded-In₂O₃, (b) porous Pd–In₂O₃, (c) Pd–In₂O₃ NW-like network and (d) XRD patterns of as-obtained In₂O₃ samples applied in gas sensing properties.

To understand the gas sensing properties and mechanism of the as-obtained Pd–In₂O₃ NW-like network, gas sensing measurements were carried out. For comparison, the sensing performances of the as-prepared porous unloaded-In₂O₃ and porous Pd–In₂O₃ were also studied. Fig. 8a shows the response curves of the three samples at different operating temperatures under exposure to 5 ppm NO₂. Obviously, the responses of all tested sensors vary with operating temperature in a similar way, first increasing with temperature up to a maximum of 110 °C, then decreasing rapidly as operating temperature increases further. In addition, the maximum response of the Pd–In₂O₃ NW-like network sensor can reach 27, which is about 5 times those of the porous unloaded-In₂O₃ and porous Pd–In₂O₃ sensors (their response values are about 6.2 and 5.5, respectively). As a consequence, an optimum temperature of 110 °C was chosen for further gas sensing analysis. Fig. S4† displays the responses of the three sensors with NO₂ concentrations ranging from 1 to 20 ppm at 110 °C. Similarly, the Pd–In₂O₃ NW-like network sensor presents a much higher response than those of the other two sensors. This is attributed to the higher surface-to-volume ratio and more channels for gas diffusion. In addition, the two sensors with similar morphology, but different components, have the same response value at various concentrations of NO₂. The BET surface areas of the three sensors (porous unloaded-In₂O₃, porous Pd–In₂O₃, and Pd–In₂O₃ NW-like network), calculated from 5-point BET surface area plots (Fig. S5†), are about 548.200, 550.101, 1327.035 m² g⁻¹, respectively, indicating that the NW-like network sensor has a much higher surface-to-volume ratio than the other two sensors.

Fig. 8 (a) Dependence of the gas sensor response at various operating temperatures of as obtained samples toward 5 ppm NO₂. Dynamic response curves of (b) porous unloaded-In₂O₃, (c) Pd–In₂O₃, (d) Pd–In₂O₃ NW-like network sensors at optimum operating temperature (110 °C).

Fig. 8b–d shows the dynamic response curves of the three In₂O₃-based sensors towards 5 ppm NO₂ at a temperature of 110 °C. It can be observed that the electric resistances of the sensors increased abruptly on exposure to NO₂ and subsequently decreased to their initial value following the removal of NO₂. The response and recovery times of the Pd–In₂O₃ NW-like network (shown in Fig. 8d) were approximately 9 and 28 s, respectively, and, in Fig. 8c, the corresponding values were about 7 and 26 s for the porous Pd–In₂O₃. However, the response and recovery times of the porous unloaded-In₂O₃ sensors were the longest (71 and 100 s (in Fig. 8b)). This implies that the response and recovery times can be dramatically shortened by Pd loading. The curves in Fig. 8c and d are obviously different from that shown in Fig. 8b. One or more sharp prominent signals, marked as “pulse signal”, appeared once the sensors were exposed to NO₂. Moreover, as the measured time increased, the resistance of the sensors gradually decreased initially and, subsequently, fluctuated around a certain equilibrium value (this value was applied in response calculation instead of R_gas when the “cyclical fluctuation” appeared), before finally achieving dynamic equilibrium.

Fig. 9 clearly demonstrates the typical dynamic response curves of the Pd–In₂O₃ NW-like network sensor towards 5 ppm NO₂ at various temperatures. It can be observed that the four curves possess different characteristics. The curve shown in Fig. 9a was measured at 50 °C. It demonstrates higher amplitude (approximately 0.45 × 10⁹ Ω) than that of the “pulse signal” (approximately 1.01 × 10⁸ Ω). At 110 °C, the amplitude is only 0.50 × 10⁷ Ω (Fig. 9b) and, at an operating temperature of 190 °C, it even disappears, as shown in Fig. 9c. The “pulse signal” demonstrates the same trend. It decreases as the operating temperature increases and disappears at approximately 210 °C (Fig. 9d). Moreover, the same result was obtained for the response curves of porous Pd–In₂O₃, as shown in Fig. S6.† Until now, few reports on this gas sensing phenomenon could be found in the literature. In order to demonstrate that the two novel characteristics, “pulse signal” and “cyclical fluctuations”, are reproducible, gas sensing measurements were carried out on a Pd–In₂O₃ NW-like network prepared 3 months ago. As show in Fig. S7,† we can see that there are almost no changes in the response curves, which confirms the good stability of the sensor.

Fig. 9 Typical dynamic response curves of the Pd–In₂O₃ NW-like network sensor exposed to 5 ppm NO₂ at different temperatures: (a) 50 °C, (b) 110 °C, (c) 190 °C and (d) 210 °C.

Response and recovery times of a gas sensor are also important for practical application, and their changes can provide an intuitive demonstration of the gas sensing mechanism. Fig. 10 shows the response and recovery times of three sensors exposed to 5 ppm NO₂ at different operating temperatures (110, 170 and 210 °C). Both the response and recovery times of the porous unloaded-In₂O₃ sensor dramatically shortened from 71 to 8 s and 100 to 18 s with an increase in operation temperature. Nevertheless, there is some variation in response and recovery times for sensors with Pd loading at various temperatures. The unloaded sensor is most sensitive to temperature and this is further described in Fig. S8.† This proves that Pd-loaded In₂O₃ and unloaded-In₂O₃ involve two key mechanisms, namely electric effects and chemical effects.³¹

Fig. 10 Histograms of response and recovery times of porous unloaded-In₂O₃, porous Pd–In₂O₃ and Pd–In₂O₃ NW-like network sensors to 5 ppm NO₂ at 110, 170 and 210 °C.

For In₂O₃-based gas sensors, the gas sensing mechanism can be explained by changes in resistance, which are attributed to an increase or decrease in free electron concentration caused by adsorption or desorption of gas molecules on the In₂O₃ surface.² When the sensors were exposed to air, O₂ adsorbed on the surface, and the absorbed oxygen captured electrons from the sensors to generate chemisorbed oxygen species (O₂⁻, O⁻ and O²⁻), as shown in eqn (1)–(4). This will cause an increase in the resistance of the In₂O₃ sensors (R_a).^1,7,8

O_2(gas) → O_2(ads) (1)

O_2(ads) + e⁻ → O_2(ads)⁻ (2)

O_2(ads)⁻ + e⁻ → 2O_(ads)⁻ (3)

O_(ads)⁻ + e⁻ → O_(ads)²⁻ (4)

Amongst the In₂O₃-based gas sensors, the gas sensing of porous unloaded In₂O₃ is attributed to electric effects. When the sensors were exposed to NO₂, the adsorbed NO₂ molecules not only capture electrons from the surface of the In₂O₃, but also directly react with the chemisorbed oxygen species, forming adsorbed NO_2(ads)⁻, as shown in eqn (5) and (6).³⁹

NO_2(gas) + e⁻ → NO_2(ads)⁻ (5)

NO_2(gas) + O_2(ads)⁻ + 2e⁻ → NO_2(ads)⁻ + 2O_(ads)⁻ (6)

In the above reactions, the electrons in the In₂O₃ are transferred to NO_2(gas) or O_2(ads)⁻. This process leads to an increase in resistance due to the decrease in the electron concentration in the In₂O₃. In fact, it is difficult to capture electrons from In₂O₃, because the reactive energy barrier cannot be overcome at low temperature.

However, for Pd-loaded In₂O₃, Pd as a catalyst generally plays an important role in enhancing the response and recovery times. It dramatically activates the dissociation of oxygen molecules on the surface of sensors and increases the adsorbed oxygen species.^28,31 Simultaneously, NO₂ is a very efficient source of oxygen adatoms due to the catalyzation of Pd. When NO₂ is introduced, NO₂ molecules are partially decomposed into adsorbed NO (NO_ads) and oxygen adatoms on the surface of Pd (O_ads(Pd)), as shown in eqn (7). Moreover, the dissociated oxygen atoms (O_ads) are transferred to In₂O₃ by means of the spillover effect and capture electrons from In₂O₃, which leads to an increase in the resistance of the sensor in NO₂, as shown by eqn (8) and (9).³²

NO_2(gas) + Pd → NO_ads + O_ads(Pd) (7)

O_ads(Pd) + In₂O₃ → O_ads(In₂O₃) (8)

O_ads(In₂O₃) + e⁻ → O_ads⁻(In₂O₃) (9)

O_ads⁻(In₂O₃) + NO_ads → NO₂ + e⁻ (10)

O_ads is much more active than NO_2(gas) and O_2(ads)⁻, so the rate of capturing electrons in eqn (9) is faster than that in eqn (5) and (6). Pd-loaded In₂O₃ sensors thus possess shorter response and recovery times. Meanwhile, adsorbed NO will be oxidized into NO₂ on the surface of Pd–In₂O₃ sensors and release electrons into In₂O₃, leading to a decrease in its resistance,⁴⁰ as shown by eqn (10). These chemical equations and chemical effects can be used to explain the appearance of the previously described response curves.

According to the spillover mechanism, the loading of Pd onto In₂O₃ can dramatically increase the chemisorption of NO₂. After exposure to NO₂, the Pd–In₂O₃ sensors dissociate NO₂ into NO and O⁻. This is rapidly accompanied by the capturing of electrons from the In₂O₃ surface. The resistance of the sensor thus increases to a maximum value within a short time. Subsequently, the dissociated NO will react with the chemisorbed oxygen species. According to eqn (10), the electrons trapped by the dissociated oxygen species are released and then re-injected into the In₂O₃, which decreases the resistance of the sensor. The resistance of sensor thus falls after it first ascends, and finally tends towards an equilibrium, thus forming the “pulse signal”. Over time, the NO₂ dissociation and the NO adsorption reach dynamic equilibrium, but the two reactions are asynchronous. Firstly, the NO₂ dissociation dominates, which increases the resistance. When the NO concentration then reaches a certain level, a reaction between NO and the adsorption oxygen species occurs, together with a decrease in resistance. In contrast, as the NO is consumed and decreases to a low concentration, NO₂ dissociation becomes dominant again. These two asynchronous reactions repeat until equilibrium conditions exist, contributing to the appearance of “cyclical fluctuations”. It is worth noting that loading of Pd always enhances the response of the gas sensor. In this study, the response of the porous Pd–In₂O₃ is the same as that of the porous In₂O₃ without Pd loading. Introducing Pd into In₂O₃ did not improve the response of In₂O₃. The possible reason is as follows: on the one hand, when NO₂ was introduced, Pd catalyzed the decomposition of NO₂ through the reaction shown in eqn (7), and the dissociated oxygen atoms (O_ads) captured electrons from In₂O₃, resulting in the increase in resistance of the sensor in NO₂ (marked by R_gas↑). On the other hand, loading of Pd dramatically activated the dissociation of oxygen (in air) on the surface of sensor. As more electrons were captured by oxygen molecules, R_air (the resistance of the intrinsic In₂O₃) increased (marked by R_air↑). The catalysis of Pd not only increased R_gas, but also increased R_air. In conclusion, the response (R = R_gas↑/R_air↑) of porous Pd–In₂O₃ was not enhanced.

The correlation of the “pulse signal” and the “cyclical fluctuations” with operating temperature in the response curve of the Pd–In₂O₃ NW-like network sensor can be explained by the synergy of the chemical and electric effects. At a low temperature (50 °C), the adsorbed gas (NO) molecules are not sufficiently activated to overcome the activation energy barrier to react with the adsorption oxygen species.⁷ The reaction between NO and the adsorption oxygen species cannot easily occur until enough NO₂ is dissociated, which induces a high amplitude of pulse and fluctuation. At a temperature of 110 °C, the operating temperature is at an optimum for chemical effects to occur. However, the rate of the reaction between NO and the adsorption oxygen species increases more rapidly than that of NO₂ decomposition. This induces the decrease of fluctuation amplitude.⁴¹ Subsequently, the reaction in eqn (10) accelerates at a higher temperature (such as 190 °C), causing the generated NO to be absorbed rapidly and oxidized without delay. The “cyclical fluctuation” therefore disappears. As the temperature increases, the reaction between NO and the adsorption oxygen species almost ceases as a result of insufficient compensation for the increased surface reactivity. This is a possible reason for the disappearance of the “pulse signal”.

Conclusions

In summary, a Pd-loaded In₂O₃ NW-like network was first synthesized via electroless plating using CNTs as templates, in combination with subsequent oxidation and calcination. The preparation process was studied and it was determined that the morphology of the samples can be controlled by oxidation time. The activation of palladium (Pd) played an important role in replicating the structure of the CNTs, in order to form a NW-like network. Gas sensing measurements revealed that the response of Pd-loaded In₂O₃ NW-like network was 27, which was approximately 5 times greater than that of porous Pd–In₂O₃ (approximately 5.5), when exposed to NO₂ concentrations as low as 5 ppm at a temperature of 110 °C. In addition, results showed that Pd-loading is an effective approach to shorten the response and recovery times. On exposure to NO₂ at a concentration of 5 ppm at a temperature of 110 °C, the response and recovery times of the Pd–In₂O₃ NW-like network were approximately 9 and 28 s, respectively. For the porous unloaded-In₂O₃, the corresponding values were approximately 71 and 100 s. Moreover, two novel characteristics were discovered: the “pulse signal” and “cyclical fluctuations”. These have the potential to be applied to the preparation of multi-functional gas sensors and to quantitative detection. Finally, the synergy of the chemical and electric effects was introduced to explain these results. Our results and findings suggest that the Pd–In₂O₃ NW-like network is a promising candidate to monitor and control NO₂ in terms of its low detection limit, high response, and short response and recovery times.

Acknowledgements

Financial support by National Natural Science Foundation of China (51402211) is gratefully acknowledged.

References

1. P. Li, H. Fan, Y. Cai and M. Xu, CrystEngComm, 2014, 16, 2715–2722.
2. C. Zhao, G. Zhang, W. Han, J. Fu, Y. He, Z. Zhang and E. Xie, CrystEngComm, 2013, 15, 6491–6497.
3. J. Zai, J. Zhu, R. Qi and X. Qian, J. Mater. Chem. A, 2013, 1, 735–745.
4. Y. F. Sun, S. B. Liu, F. L. Meng, J. Y. Liu, Z. Jin, L. T. Kong and J. H. Liu, Sensors, 2012, 12, 2610–2631.
5. N. Siedl, P. Gugel and O. Diwald, Langmuir, 2013, 29, 6077–6083.
6. F. Gu, L. Zhang, Z. Wang, D. Han and G. Guo, Sens. Actuators, B, 2014, 193, 669–678.
7. P. Song, D. Han, H. H. Zhang, J. Li, Z. X. Yang and Q. Wang, Sens. Actuators, B, 2014, 196, 434–439.
8. S. Elouali, L. G. Bloor, R. Binions, I. P. Parkin, C. J. Carmalt and J. A. Darr, Langmuir, 2012, 28, 1879–1885.
9. N. Singh, R. K. Gupta and P. S. Lee, ACS Appl. Mater. Interfaces, 2011, 3, 2246–2252.
10. J. Liu, W. Li, L. H. Zhu, C. Li, F. D. Qu, W. B. Guo, C. H. Feng and S. P. Ruan, J. Nanosci. Nanotechnol., 2014, 14, 3653–3657.
11. X. Lu and L. Yin, J. Mater. Sci. Technol., 2011, 27, 680–684.
12. Y. Li, J. Xu, J. Chao, D. Chen, S. Ouyang, J. Ye and G. Shen, J. Mater. Chem., 2011, 21, 12852–12857.
13. W. J. Tseng, T. T. Tseng, H. M. Wu, Y. C. Her, T. J. Yang and P. Gouma, J. Am. Chem. Soc., 2013, 96, 719–725.
14. N. Du, H. Zhang, X. Ma and D. Yang, Chem. Commun., 2008, 6182–6184.
15. Z. Zanolli, R. Leghrib, A. Felten, J. J. Pireaux, E. Llobet and J. C. Charlier, ACS Nano, 2011, 5, 4592–4599.
16. S. Yi, S. Tian, D. Zeng, K. Xu, S. Zhang and C. Xie, Sens. Actuators, B, 2013, 185, 345–353.
17. A. Gurlo, Nanoscale, 2011, 3, 154–165.
18. L. Xu, B. Dong, Y. Wang, X. Bai, Q. Liu and H. Song, Sens. Actuators, B, 2010, 147, 531–538.
19. S. K. Lim, S. H. Hwang, D. Chang and S. Kim, Sens. Actuators, B, 2010, 149, 28–33.
20. W. Zheng, X. Lu, W. Wang, Z. Li, H. Zhang, Y. Wang, Z. Wang and C. Wang, Sens. Actuators, B, 2009, 142, 61–65.
21. N. Du, H. Zhang, B. D. Chen, X. Y. Ma, Z. H. Liu, J. B. Wu and D. R. Yang, Adv. Mater., 2007, 19, 1641–1645.
22. D. Zhang, C. Pan, L. Shi, L. Huang, J. Fang and H. Fu, Microporous Mesoporous Mater., 2009, 117, 193–200.
23. Y. Zhang, Z. Zheng and F. Yang, Ind. Eng. Chem. Res., 2010, 49, 3539–3543.
24. C. K. Tagad, K. S. Rajdeo, A. Kulkarni, P. More, R. C. Aiyer and S. Sabharwal, RSC Adv., 2014, 4, 24014.
25. Z. Zhu, R. C. Deka, A. Chutia, R. Sahnoun, H. Tsuboi, M. Koyama, N. Hatakeyama, A. Endou, H. Takaba, C. A. Del Carpio, M. Kubo and A. Miyamoto, J. Phys. Chem. Solids, 2009, 70, 1248–1255.
26. M. H. Zuruzi, Y. H. Nurmawati, S. X. Yeo, P. C. Wu, H. Lai and Z. Chen, RSC Adv., 2013, 3, 19971.
27. Y. Sun and H. H. Wang, Adv. Mater., 2007, 19, 2818–2823.
28. C. Liewhiran, N. Tamaekong, A. Wisitsoraat, A. Tuantranont and S. Phanichphant, Sens. Actuators, B, 2013, 176, 893–905.
29. S. J. Pearton, F. Ren, Y.-L. Wang, B. H. Chu, K. H. Chen, C. Y. Chang, W. Lim, J. Lin and D. P. Norton, Prog. Mater. Sci., 2010, 55, 1–59.
30. A. Shanmugasundaram, B. Ramireddy, P. Basak, S. V. Manorama and S. Srinath, J. Phys. Chem. C, 2014, 118, 6909–6921.
31. H. Y. Lai and C. H. Chen, J. Mater. Chem., 2012, 22, 13204–13208.
32. S. S. Kim, J. Y. Park, S.-W. Choi, H. G. Na, J. C. Yang and H. W. Kim, J. Alloys Compd., 2011, 509, 9171–9177.
33. J. Sittikun, Y. Boonyongmaneerat, P. Weerachawanasak, P. Praserthdam and J. Panpranot, React. Kinet., Mech. Catal., 2014, 111, 123–135.
34. L. S. Zhong, J. S. Hu, Z.-M. Cui, L. J. Wan and W. G. Song, Chem. Mater., 2007, 19, 4557–4562.
35. L. M. Ang, T. S. A. Hor, G. Q. Xu, C. H. Tung, S. P. Zhao and J. L. S. Wang, Carbon, 2000, 38, 363–372.
36. L. N. Jin, Q. Liu and W. Y. Sun, CrystEngComm, 2013, 15, 4779–4784.
37. Z. Bian, J. Zhu, F. Cao, Y. Huo, Y. Lu and H. Li, Chem. Commun., 2010, 46, 8451–8453.
38. E. Comini, C. Baratto, G. Faglia, M. Ferroni, A. Vomiero and G. Sberveglieri, Prog. Mater. Sci., 2009, 54, 1–67.
39. J. Zhao, T. Yang, Y. Liu, Z. Wang, X. Li, Y. Sun, Y. Du, Y. Li and G. Lu, Sens. Actuators, B, 2014, 191, 806–812.
40. B. M. Weiss and E. Iglesia, J. Catal., 2010, 272, 74–81.
41. C. S. Rout, K. Ganesh, A. Govindaraj and C. N. R. Rao, Appl. Phys. A, 2006, 85, 241–246.

---

Footnote

† Electronic supplementary information (ESI) available: Fig. S1–S8. See DOI: 10.1039/c4ra14580a

---