Tunning the Gas Sensing Properties of rGO with In2O3 Nanoparticles

Here, we discuss the effect of In2O3 nanoparticles on the reduced graphene oxide (rGO) gas-sensing potentialities. In2O3 nanoparticles were prepared with the polymer precursors method, while the nanocomposites were prepared by mixing an In2O3 nanoparticle suspension with an rGO suspension in different proportions. The gas-sensing performance of our materials was tested by exposing our materials to known concentrations of a target toxic gas in a dry airflow. Our results demonstrate that In2O3 nanoparticles enhance the rGO sensitivity for strong oxidizing species such as O3 and NO2, while a negative effect on its sensitivity for NH3 sensing is observed. Furthermore, our measurements towards H2S suggest that the concentration of In2O3 nanoparticles can induce an uncommon transition from p-type to n-type semiconductor nature when rGO–In2O3 nanocomposites operate at temperatures close to 160 ◦C.


Introduction
Graphene's remarkable physical properties justify the emergence of several studies investigating this material as a sensing material. Its high electronic mobility, intrinsic surface area, high transparency, and excellent mechanical properties make graphene-based materials and devices great potentials to develop new technologies [1]. Furthermore, the relativistic nature of the electronic conduction in graphene and its high surface area has opened several research opportunities to implement these materials as different sensing elements. For instance, there are significant recent advances in graphene-based sensors for human health monitoring [2], FET-based electrochemical sensors and biosensors [3][4][5], flexible electronics [6,7], strain sensors [8][9][10][11], and printed electronics [12,13], among other applications.
Within this context, graphene-based gas sensors have been extensively studied in the recent past [14]. The pioneering work of Schedin et al. [15] demonstrated that pristine graphene could detect low concentrations of typical pollutants when operating at room temperature. There are significant achievements worth noting in graphene-based gas sensors. For instance, they are especially sensitive for nitrogen-reactive species detection such as nitrogen dioxide (NO 2 ) or ammonia (NH 3 ) [16][17][18][19][20][21]. Flexible and printed gas sensors based on graphene derivatives have also been developed with excellent sensitivity [22][23][24][25]. One form of graphene that has been especially explored as chemical sensors is the so-called reduced graphene oxide (rGO). rGO can be depicted as a few layers of graphene-like sheets piled up with a high density of crystalline defects and a small concentration of oxygenated groups (-OH, -COOH, C-O-C) bonded to its surface [26,27]. This material can have its most fundamental physical properties depending on the concentration of the oxygenated groups remaining from the reduction process of graphene oxide [28]. A redox reaction mediates the phenomenon of gas detection by semiconductors between the gas and the material's Furthermore, we show that the performance of such materials towards the detection of typical pollutants (NH 3 , H 2 S, NO 2, and O 3 ) can be tailored by controlling the ratio between rGO and In 2 O 3 nanoparticles. Our results demonstrate that In 2 O 3 nanoparticles can improve the rGO's sensing performance towards NO 2 , which agrees with previous literature results. Regarding O 3 detection, our results suggest that the strong oxidizing nature of this pollutant promotes a permanent oxidation reaction between the gas and the rGO as the operating temperature increases. Towards NH 3 , the addition of In 2 O 3 nanoparticles decreases the sensors' response and increases its response time. Regarding the performance for H 2 S detection, our results highlight that it is possible to induce a change of the fundamental type of response of the rGO-In 2 O 3 nanocomposites exposed to this gas. Consequently, the nanocomposites layers could interact with H 2 S as a p or n-type semiconductor depending on the operating temperature and the concentration of In 2 O 3 nanoparticles used to fabricate the sensing device.

Materials and Methods
A suspension of reduced graphene oxide 10 mg/mL and indium nitrate (In(NO 3 ) 3 ·H 2 O) were used as raw materials and were purchased from Sigma Aldrich Brazil. Nanostructured In 2 O 3 was prepared through the polymeric precursor method. This method is based upon the formation of metallic complexes using hydrocarboxylic acid followed by a polymerization reaction that is promoted by increasing the temperature of this metallic complex in the presence of a polyalcohol. Here, In 2 O 3 nanoparticles were prepared using indium nitrate as the precursor, citric acid as a complexing agent, and ethylene glycol for the polymerization reaction that forms a resin containing In 3+ ions. For this reaction, the optimal molar ratio between citric acid and indium nitrate was 3:1, while the optimal volume ratio between the citric acid solution and ethylene glycol was 3:2, respectively. The nanoparticles are formed after a heat-treatment of the resin at 500 • C for 1 h. The nanocomposites were prepared by diluting the rGO suspension in deionized water to 0.1 mg/mL. Similarly, 50 mg of the In 2 O 3 powder were dispersed in 20 mL of deionized water (2.5 mg/mL). Then, these two dispersions were sonicated for 30 min and mixed in four different proportions: 0, 5, 15, and 50 v/v% of In 2 O 3 to rGO.
The structural characterization was carried out using conventional θ-2θ X-ray powder diffraction in a Rigaku Ultima IV diffractometer using Cu-K α radiation (λ = 1.5418 Å) and LiF (100) monochromator. The Rietveld refinement was carried out with GSAS-II [70] and the 3D structure representations with VESTA Crystallography [71]. After exciting the samples with an Ion-Ar laser beam, the Raman spectra were collected in a Witec microscope equipped with a Peltier cooled CCD (λ = 514 nm). The sensing materials and films' morphology was evaluated with a Zeiss Sigma Field Emission Scanning Electron Microscope. The SEM images analysis was carried out with ImageJ software [72]. The surface area ratio between rGO and In 2 O 3 was estimated for each image using contrast differences, while the nanoparticle size was estimated by measuring the largest dimension of more than 300 nanoparticles measured from the images.
The chemical state of the surface elements was evaluated using a Scientia Omicron ESCA spectrometer with a monochromatic X-ray source Al-Kα (1486.7 eV, with a power of 280 W and a constant pass energy mode of 50 eV).
For the gas sensing performance characterization, the nanocomposites dispersions were drop-casted onto high-density alumina substrates built with screen-printed Pt interdigitated electrodes on the upper side and a screen-printed Pt heater on the bottom side. The electrical measurements and control were carried out in a lab-built experimental setup that enables complete control of experimental conditions: temperature, flow parameters, and pollutant concentration. Additionally, this system allows one to record the electrical resistance data of up to four samples simultaneously measured by the two-probe method and stored synchronously with the operating conditions. All instruments are controlled by a lab-made virtual instrument (VI) developed in LabVIEW. The gas sensing behavior was assessed by continuously monitoring each devices' electrical resistance upon controlled exposure to the gases of interest. Our typical experiments consisted of stabilizing the devices at constant airflow for 12 h until equilibrium, followed by a step of controlled exposures (1, 2, 5, and 10 ppm) of sensing devices to H 2 S, NH 3 , and NO 2 . For O 3 , the exposure concentrations were 250, 500, 750, and 1000 ppb. We compare datasets obtained with operating temperatures set to 100 and 160 • C. In this work, we have defined the sensor's response as ∆R/R 0 = (R gas − R 0 )/R 0 , where R gas is the resistance when the devices are exposed to the target gas, and R 0 is the resistance under air. After 12 h of the stabilization process, the exposure-recovery cycles started after 50 min. For data visualization purposes, we compare each dataset with the normalized resistance, i.e., the electrical resistance values divided the value before the first target gas exposure R/R 50min . Figure 1 reports a typical Raman spectrum of the rGO used to elaborate the nanocomposites. Carbon-based materials exhibit several characteristic Raman vibration modes. The C-sp 2 lattice vibration results in the well-known G-band (~1580 cm −1 ) present in graphite, carbon nanotubes, graphene, GO, and rGO. The D-band (~1350 cm −1 ) results from point defects vibration and is the most intense peak in GO and rGO-based materials. The D' band is the first-order vibration mode of aromatic rings, while the 2D band is the second-order vibration mode of this structure. In ideal pristine graphene, the free π electrons are all confined in aromatic rings. Consequently, the relative intensity of this band should be twice higher than the G band in an ideal graphene sheet [73]. The intensity ratio between the 2D and G bands is a parameter often used to assess the quality of the rGO-based materials. Our samples exhibit I 2D /I G = 0.28, which is comparable to the quality of rGO according to the study of Kaniyoor and Ramaprabhu [74].

Structural Characterization
setup that enables complete control of experimental conditions: temperature, flow parameters, and pollutant concentration. Additionally, this system allows one to record the electrical resistance data of up to four samples simultaneously measured by the two-probe method and stored synchronously with the operating conditions. All instruments are controlled by a lab-made virtual instrument (VI) developed in LabVIEW. The gas sensing behavior was assessed by continuously monitoring each devices' electrical resistance upon controlled exposure to the gases of interest. Our typical experiments consisted of stabilizing the devices at constant airflow for 12 h until equilibrium, followed by a step of controlled exposures (1, 2, 5, and 10 ppm) of sensing devices to H2S, NH3, and NO2. For O3, the exposure concentrations were 250, 500, 750, and 1000 ppb. We compare datasets obtained with operating temperatures set to 100 and 160 °C. In this work, we have defined the sensor's response as ΔR/R0 = (Rgas − R0)/R0, where Rgas is the resistance when the devices are exposed to the target gas, and R0 is the resistance under air. After 12 h of the stabilization process, the exposure-recovery cycles started after 50 min. For data visualization purposes, we compare each dataset with the normalized resistance, i.e., the electrical resistance values divided the value before the first target gas exposure R/R50min. Figure 1 reports a typical Raman spectrum of the rGO used to elaborate the nanocomposites. Carbon-based materials exhibit several characteristic Raman vibration modes. The C-sp 2 lattice vibration results in the well-known G-band (~1580 cm −1 ) present in graphite, carbon nanotubes, graphene, GO, and rGO. The D-band (~1350 cm −1 ) results from point defects vibration and is the most intense peak in GO and rGO-based materials. The D' band is the first-order vibration mode of aromatic rings, while the 2D band is the second-order vibration mode of this structure. In ideal pristine graphene, the free π electrons are all confined in aromatic rings. Consequently, the relative intensity of this band should be twice higher than the G band in an ideal graphene sheet [73]. The intensity ratio between the 2D and G bands is a parameter often used to assess the quality of the rGObased materials. Our samples exhibit I2D/IG = 0.28, which is comparable to the quality of rGO according to the study of Kaniyoor and Ramaprabhu [74].  presents an X-ray diffraction pattern for an In2O3 sample prepared by the polymer precursor method described in Section 3.1 after heat-treating the resin at 500 °C. This oxide crystallizes in a cubic symmetry (space group Ia3). The lattice parameter was estimated at 1.01 nm through the Rietveld refinement. The refinement converged fast and  presents an X-ray diffraction pattern for an In 2 O 3 sample prepared by the polymer precursor method described in Section 3.1 after heat-treating the resin at 500 • C. This oxide crystallizes in a cubic symmetry (space group Ia3). The lattice parameter was estimated at 1.01 nm through the Rietveld refinement. The refinement converged fast and yielded a goodness-of-fit of 1.83 and crystallite size estimated at 59 nm. The inset of Figure 2 shows a schematic representation of the typical bixbyite structure of In 2 O 3 .

Structural Characterization
Surfaces 2022, 5, FOR PEER REVIEW 5 yielded a goodness-of-fit of 1.83 and crystallite size estimated at 59 nm. The inset of Figure  2 shows a schematic representation of the typical bixbyite structure of In2O3.  Figure 3 depicts SEM images of drop-casted nanocomposites samples. It is possible to observe in these micrographs that the amount of these In2O3 nanoparticles increases with the sample's nominal composition. For example, Figure 3a-d show that the amount of nominal In2O3 nanoparticles increases from 0 to 50 v/v%. The Image J software [72] was used to estimate the surface area ratio between In2O3 nanoparticles and rGO (AIn2O3/ArGO). This ratio was calculated for the images in Figure 3b-d and was found to be 20.1%, 29.2%, and 46.3%, respectively. These results demonstrate that it is possible to tailor the chemical nature of the exposed surface to the atmosphere of such nanocomposites.    [72] was used to estimate the surface area ratio between In 2 O 3 nanoparticles and rGO (A In2O3 /A rGO ). This ratio was calculated for the images in Figure 3b-d and was found to be 20.1%, 29.2%, and 46.3%, respectively. These results demonstrate that it is possible to tailor the chemical nature of the exposed surface to the atmosphere of such nanocomposites.

Surface Characterization and Morphology
Surfaces 2022, 5, FOR PEER REVIEW 5 yielded a goodness-of-fit of 1.83 and crystallite size estimated at 59 nm. The inset of Figure  2 shows a schematic representation of the typical bixbyite structure of In2O3.  Figure 3 depicts SEM images of drop-casted nanocomposites samples. It is possible to observe in these micrographs that the amount of these In2O3 nanoparticles increases with the sample's nominal composition. For example, Figure 3a-d show that the amount of nominal In2O3 nanoparticles increases from 0 to 50 v/v%. The Image J software [72] was used to estimate the surface area ratio between In2O3 nanoparticles and rGO (AIn2O3/ArGO). This ratio was calculated for the images in Figure 3b-d and was found to be 20.1%, 29.2%, and 46.3%, respectively. These results demonstrate that it is possible to tailor the chemical nature of the exposed surface to the atmosphere of such nanocomposites.  Figure 4 presents SEM images with higher magnification to better define the morphology aspects of our nanocomposites. One can observe three main features in these images. For instance, Figure 4a shows an agglomerate of In 2 O 3 nanoparticles that are fully exposed to the atmosphere. Figure 4b shows the formation of regions where In 2 O 3 nanoparticles agglomerates are partially incorporated into the rGO-layers. Finally, Figure 4c shows a region in which the particles are better dispersed into the matrix, exhibiting a great In 2 O 3 area exposed to the atmosphere. The average particle size can be defined as the highest dimension of a particle, and by measuring these values of more than 300 particles from the images shown below, it is possible to estimate that the mean particle size of the In 2 O 3 nanoparticles should lie in the vicinity of 34.5 nm after a Gaussian fit, which is in the same order of magnitude as the crystallite size estimated in the Rietveld refinement. Figure 4 presents SEM images with higher magnification to better define the morphology aspects of our nanocomposites. One can observe three main features in these images. For instance, Figure 4a shows an agglomerate of In2O3 nanoparticles that are fully exposed to the atmosphere. Figure 4b shows the formation of regions where In2O3 nanoparticles agglomerates are partially incorporated into the rGO-layers. Finally, Figure 4c shows a region in which the particles are better dispersed into the matrix, exhibiting a great In2O3 area exposed to the atmosphere. The average particle size can be defined as the highest dimension of a particle, and by measuring these values of more than 300 particles from the images shown below, it is possible to estimate that the mean particle size of the In2O3 nanoparticles should lie in the vicinity of 34.5 nm after a Gaussian fit, which is in the same order of magnitude as the crystallite size estimated in the Rietveld refinement.  Figure 5 shows the XPS spectra evolution as the amount of In2O3 used in the fabrication of the nanocomposites increases. It is possible to observe that the intensity ratio of the peaks related to the In chemical surface states increases as the nominal composition of the samples increase. The surface atomic concentration of the two In 3d degenerate states are 1.5, 6.4, and 11.3% for the samples with 5, 15, and 50 v/v% of In2O3 nanoparticles. XPS results are summarized in Table 1.   Figure 5 shows the XPS spectra evolution as the amount of In 2 O 3 used in the fabrication of the nanocomposites increases. It is possible to observe that the intensity ratio of the peaks related to the In chemical surface states increases as the nominal composition of the samples increase. The surface atomic concentration of the two In 3d degenerate states are 1.5, 6.4, and 11.3% for the samples with 5, 15, and 50 v/v% of In 2 O 3 nanoparticles. XPS results are summarized in Table 1. Figure 6a-f present typical X-ray photoelectron spectra of the rGO and In2O3 used in this work separately. Here, it is possible to observe that the rGO survey spectrum shown in Figure 6a exhibits not only C and O but also signals from Na and S. The presence of Na  Figure 6a-f present typical X-ray photoelectron spectra of the rGO and In 2 O 3 used in this work separately. Here, it is possible to observe that the rGO survey spectrum shown in Figure 6a exhibits not only C and O but also signals from Na and S. The presence of Na and S are related to the chemical route used to prepare this material; S is present due to oxidation of graphite with sulfuric acid (H 2 SO 4 ). At the same time, Na is related to the reducing agents used to reduce the GO, such as sodium borohydride (NaBH 4 ) or sodium citrate (Na 3

Gas Sensing Properties
rGO-based gas sensors are well-known to exhibit high and fast responses towards NO2. Figure 7a,b present a dynamic response-recovery dataset of four sensors prepared with an increasing nominal composition of In2O3 nanoparticles to prepare the nanocomposites, as described in Section 2 of this manuscript. Two features should be noted in these figures. One is that the In2O3 nanoparticles tend to increase the sensors' response towards NO2, which can be explained by the formation of p-n heterojunctions between p-type rGO and n-type In2O3 [69]. Moreover, the second feature is that the devices' responses are higher as the operating temperature increases.
Similar to what was observed for NO2, the addition of In2O3 nanoparticles also increased the sensors' responses towards O3. However, some differences should be underlined. It is possible to observe that the responses of these sensors are independent of the  Figure 6d presents the a typical In 2 O 3 survey spectrum, in which it is possible to observe peaks related to In 3p 1/2 (~702 eV), In 3p 3/2 (~664 eV), O 1s (~532 eV), C 1s (~284.5 eV), In 3d 3/2 (~451 eV), In 3d 5/2 (~443 eV), and In 4d (~16 eV). The In 3d peaks are shown in Figure 6e. The high-resolution spectra in the In 3d region shown in Figure 6e suggest that two chemical states are present in our samples, one assigned to In 3+ and another to metallic In. Figure 6f presents the high-resolution spectra of O 1s peak, exhibiting three major components, assigned to structural oxide ions (O 2− ) at 530.5 eV, to weakly adsorbed species at the surface, such as peroxide ions (O − ) or hydroxide ions (OH − ) at 531.8 eV, and C-O bonds from adventitious carbon at 532.9 eV. These O components account for 51.5, 32.8, and 15.7% of the oxygen states on the surface, which is in agreement with the literature [75,76].

Gas Sensing Properties
rGO-based gas sensors are well-known to exhibit high and fast responses towards NO 2 . Figure 7a,b present a dynamic response-recovery dataset of four sensors prepared with an increasing nominal composition of In 2 O 3 nanoparticles to prepare the nanocomposites, as described in Section 2 of this manuscript. Two features should be noted in these figures. One is that the In 2 O 3 nanoparticles tend to increase the sensors' response towards NO 2 , which can be explained by the formation of p-n heterojunctions between p-type rGO and n-type In 2 O 3 [69]. Moreover, the second feature is that the devices' responses are higher as the operating temperature increases.  Figure 8a,b present a comparison between the sensing behavior of rGO-In2O3 based sensors towards NH3 controlled exposure at 100 and 160 °C. Interestingly, the addition of In2O3 nanoparticles has a negative impact on the sensing properties of rGO decrease as the In2O3 nanoparticles' nominal composition increases. For example, the response ΔR/R0 calculated for pure rGO and rGO-50 v/v%-In2O3 at 10 ppm of NH3 is 15 and 6%, respectively, when operating at 100 °C. Similarly, the response decreases when the operating temperature increases. The response ΔR/R0 for 10 ppm of NH3 is 17% to 9% for pure rGO when operating at 100 and 160 °C, respectively. Figure 8c,d show the response-recovery data obtained after controlled H2S exposures. When operating at 100 °C, the addition of In2O3 nanoparticles has a similar effect to what was observed for NH3 (Figure 8a). The response decreases as the In2O3 concentration increases. An interesting effect can be noted on the gas sensing behavior of the rGO-In2O3 nanocomposites by increasing the operating temperature. Figure 8d shows the dynamic recovery curves for rGO-In2O3 operating at 160 °C. At this temperature, the electrical resistance can be tuned down and even change the nature of the interaction between the material and the gas. In other words, the p-type nature observed for all gases interaction rGO showed in this study undergoes a change to an n-type nature upon interaction with H2S in temperatures in the range of 160 °C. Figure 8a,b report the sensing behavior of rGO-In2O3-based sensors towards NH3 controlled exposure at 100 and 160 °C. Interestingly, the addition of In2O3 nanoparticles has a negative impact on the sensing properties of rGO towards ammonia. A decrease in Similar to what was observed for NO 2 , the addition of In 2 O 3 nanoparticles also increased the sensors' responses towards O 3 . However, some differences should be underlined. It is possible to observe that the responses of these sensors are independent of the O 3 concentration when operating at 100 • C, suggesting that, at this temperature, the kinetics of the O 3 adsorption reaction are low, and the steady-state under our experimental setup would take a long time. Another possibility is the low lifetime of O 3 on the nanocomposite's surface with a fast decomposition into oxygen species. The sensors' responses are higher and faster when operating at higher temperatures, as shown in Figure 7d at 160 • C. However, samples operating at these conditions tend to exhibit a strong baseline drift towards higher resistance after each period of recovery. Such behavior is observed for every In 2 O 3 nanoparticles concentration. It suggests that our nanocomposites react irreversibly with O 3 through an ozonolysis process [77], affecting the rGO matrix. This phenomenon is activated by including In 2 O 3 nanoparticles acting as a reaction catalyst. Preliminary XPS results demonstrate that the oxygen surface states can increase to levels of graphene oxide after exposing an rGO-based nanocomposite to one complete cycle of electrical resistance variations towards five exposure-recovery cycles of O 3 concentrations between 250 and 1000 ppb. Figure 8a,b present a comparison between the sensing behavior of rGO-In 2 O 3 based sensors towards NH 3 controlled exposure at 100 and 160 • C. Interestingly, the addition of In 2 O 3 nanoparticles has a negative impact on the sensing properties of rGO decrease as the In 2 O 3 nanoparticles' nominal composition increases. For example, the response ∆R/R 0 calculated for pure rGO and rGO-50 v/v%-In 2 O 3 at 10 ppm of NH 3 is 15 and 6%, respectively, when operating at 100 • C. Similarly, the response decreases when the operating temperature increases. The response ∆R/R 0 for 10 ppm of NH 3 is 17% to 9% for pure rGO when operating at 100 and 160 • C, respectively.
Surfaces 2022, 5, FOR PEER REVIEW 10 behavior of the rGO-In2O3 nanocomposites by increasing the operating temperature. Figure 8d shows the dynamic recovery curves for rGO-In2O3 operating at 160 °C. At this temperature, the electrical resistance can be tuned down and even change the nature of the interaction between the material and the gas. In other words, the p-type sensing behavior observed for rGO in interaction with all gases investigated in this study changes to an n-type behavior for rGO-50 v/v%-In2O3 interacting with H2S in temperatures in the range of 160 °C. The high density of In2O3 particles at the rGO surface led to this switch in the electrical behavior of the sensing device.

Discussion
The structural characterization by XRD and Raman spectroscopy show that the materials used in the synthesis of the nanocomposites consisted of mixing each component that had already been prepared previously. Our results demonstrate that this simple technique can modify the gas sensing properties of the raw materials with a more straightforward path than the traditional techniques based upon hydrothermal processes. Figures 1  and 2 show these results. Both figures show an excellent agreement between the structural features observed on our materials with results from the literature [43,45,47,48,[65][66][67][68][69]. Furthermore, the morphology and surface characterization with SEM and XPS demonstrate that, by changing the proportions between rGO and In2O3, one can change the nature of the surface exposed to the gas, and consequently, change and tune the type of solid-gas interaction that takes place for these materials.
Since rGO is a p-type semiconductor and In2O3 is an n-type semiconductor, increasing the amount of In2O3 nanoparticles in the rGO suspension provides a simple path to tune its gas sensing properties. Our results agree with previous literature reports regarding the detection of oxidizing gases, demonstrating that In2O3 can be combined with rGO in a synergic manner, increasing the material's sensitivity and decreasing the response time. Additionally, even though samples with high In2O3 concentrations tend to exhibit higher ozone sensitivity, our synthesis method did not prevent irreversible chemisorption of O3 molecules and the rGO layers. This effect is known as ozonolysis, and the kinetics of the degradation of the material is higher as the operating temperature rises.
Several authors have observed a synergic effect when rGO and In2O3 are applied as gas-sensitive nanocomposites for oxidizing species. According to Choi and collaborators   (Figure 8a). The response decreases as the In 2 O 3 concentration increases. An interesting effect can be noted on the gas sensing behavior of the rGO-In 2 O 3 nanocomposites by increasing the operating temperature. Figure 8d shows the dynamic recovery curves for rGO-In 2 O 3 operating at 160 • C. At this temperature, the electrical resistance can be tuned down and even change the nature of the interaction between the material and the gas. In other words, the p-type nature observed for all gases interaction rGO showed in this study undergoes a change to an n-type nature upon interaction with H 2 S in temperatures in the range of 160 • C. Figure 8a,b report the sensing behavior of rGO-In 2 O 3 -based sensors towards NH 3 controlled exposure at 100 and 160 • C. Interestingly, the addition of In 2 O 3 nanoparticles has a negative impact on the sensing properties of rGO towards ammonia. A decrease in the sensor response is observed as the In 2 O 3 nanoparticles' nominal composition increases. Moreover, the response decreases when the operating temperature increases. For example, the response ∆R/R 0 calculated for pure rGO and rGO-50 v/v%-In 2 O 3 at 10 ppm of NH3 is 15 and 6%, respectively, when operating at 100 • C. To compare, the response ∆R/R 0 for 10 ppm of NH 3 is 17% to 9% for pure rGO when operating at 100 and 160 • C, respectively. Figure 8c,d show the response-recovery data obtained after controlled H 2 S exposures. When operating at 100 • C, the addition of In 2 O 3 nanoparticles has a similar effect to what was observed for NH 3 (Figure 8a), i.e., the response decreases as the In 2 O 3 concentration increases. Nevertheless, an interesting effect can be underlined on the gas sensing behavior of the rGO-In 2 O 3 nanocomposites by increasing the operating temperature. Figure 8d shows the dynamic recovery curves for rGO-In 2 O 3 operating at 160 • C. At this temperature, the electrical resistance can be tuned down and even change the nature of the interaction between the material and the gas. In other words, the p-type sensing behavior observed for rGO in interaction with all gases investigated in this study changes to an n-type behavior for rGO-50 v/v%-In 2 O 3 interacting with H 2 S in temperatures in the range of 160 • C. The high density of In 2 O 3 particles at the rGO surface led to this switch in the electrical behavior of the sensing device.

Discussion
The structural characterization by XRD and Raman spectroscopy show that the materials used in the synthesis of the nanocomposites consisted of mixing each component that had already been prepared previously. Our results demonstrate that this simple technique can modify the gas sensing properties of the raw materials with a more straightforward path than the traditional techniques based upon hydrothermal processes. Figures 1 and 2 show these results. Both figures show an excellent agreement between the structural features observed on our materials with results from the literature [43,45,47,48,[65][66][67][68][69]. Furthermore, the morphology and surface characterization with SEM and XPS demonstrate that, by changing the proportions between rGO and In 2 O 3, one can change the nature of the surface exposed to the gas, and consequently, change and tune the type of solid-gas interaction that takes place for these materials.
Since rGO is a p-type semiconductor and In 2 O 3 is an n-type semiconductor, increasing the amount of In 2 O 3 nanoparticles in the rGO suspension provides a simple path to tune its gas sensing properties. Our results agree with previous literature reports regarding the detection of oxidizing gases, demonstrating that In 2 O 3 can be combined with rGO in a synergic manner, increasing the material's sensitivity and decreasing the response time. Additionally, even though samples with high In 2 O 3 concentrations tend to exhibit higher ozone sensitivity, our synthesis method did not prevent irreversible chemisorption of O 3 molecules and the rGO layers. This effect is known as ozonolysis, and the kinetics of the degradation of the material is higher as the operating temperature rises.
Several authors have observed a synergic effect when rGO and In 2 O 3 are applied as gas-sensitive nanocomposites for oxidizing species. According to Choi and collaborators [78], the NO 2 sensing is highly dependent on hydroxyl groups remaining on the rGO's surface. Hence, when rGO is combined with In 2 O 3 nanostructures, NO 2 molecules react with p-type rGO and n-type In 2 O 3 . However, in these hybrid materials, the depletion layer formed on the In 2 O 3 nanoparticles expands and increases the potential barriers of conduction charges in the rGO matrix, leading to higher electrical resistance variations upon interaction with the same gas concentration [68,69]. Even though there are no studies regarding the mechanisms for O 3 sensing of rGO-In 2 O 3 nanocomposites, our results suggest that the mechanism of the reaction is composed of two processes: (1)  Similar to what is observed for NO 2 detection, the interaction between NH 3 and rGO-based materials is strongly dependent on the rGO's remaining surface oxygenated groups. Furthermore, the interaction rate between In 2 O 3 nanoparticles and NH 3 occurs at higher operating temperatures [79]. Consequently, the addition of In 2 O 3 nanoparticles when the devices operate close to 160 • C decreases the active surface area towards NH 3 and decreases the device's response. Therefore, we can interpret that the interactions of NH 3 molecules on rGO-In 2 O 3 nanocomposites are more localized on rGO matrix. Consequently, as the In 2 O 3 nanoparticles reduce the rGO surface area, the available active sites density for NH 3 adsorption decreases. Therefore, towards NH 3 , SMOx does not act as a catalyst. It is also important to note that the NH 3 detection mechanism is strongly dependent on traces of humidity present in the air [80]. Consequently, our devices' responses can decrease as the operating temperature increases because the level of humidity tends to decrease at higher temperatures.
Regarding the H 2 S detection, one of the most important results from our measurements demonstrated that adding an n-type SMOx semiconductor to a p-type rGO matrix can induce a p-to-n-type transition of the rGO-In 2 O 3 nanocomposites when interacting with H 2 S at a higher temperature. This effect is a consequence of both materials being active towards H 2 S detection [81][82][83]. In other words, as the n-type In 2 O 3 active surface increases, the overall sensor's response decreases because the net measured signal is the sum of the rGO p-type interaction and In 2 O 3 n-type interaction with the H 2 S gas. The response decreases up to a threshold value, in which there is more H 2 S molecules interaction with In 2 O 3 than with the original rGO surfaces, which explains the observed transition. This effect is temperature dependent because the number of free carriers in the In 2 O 3 nanoparticles increases more rapidly over temperature than the density of holes in the rGO matrix achieving a global n-type behavior.
While such microsensors are insensitive to gaseous species not involving redox reactions with the sensing nanocomposite (such as BTEX), they remain not selective to one of the target pollutants. Nevertheless, it remains possible to identify the nature of the pollutant present in the probing gaseous phase using two identical microsensors working at two different temperatures. Figure 9 shows the response of two sensors prepared with rGO-50 v/v%-In 2 O 3 . The curves shown in the left panels are related to the signals from a sensor operating at 100 • C, while the curves at the right panels are the signals delivered by the sensor operating at 160 • C. As can be seen in this figure, the interaction with H 2 S produces a positive electrical resistance variation on the sensing element operating at 100 • C (Figure 9a), while a negative electrical resistance variation is observed for the sensing element operating at 160 • C (Figure 9d). On the other hand, a positive variation of the electrical resistance in both sensing elements indicates that the devices interact with NH 3 . If both elements exhibit a decrease in electrical resistance, the target gas must be NO 2 . The recognition method is based on studying the various function of the resistance during a few seconds of sensor exposure to gas. Table 2 summarizes how H 2 S, NH 3, or NO 2 can be discriminated by this method. Once the gas has been identified, the microsensors exhibiting the best sensing performances can be used to quantify gas concentration. Consequently, our measurements open new perspectives for studying rGO-SMOx-based gas sensors in the multi-sensor array or electronics nose context [84], since rGO-In 2 O 3 nanocomposites could readily differentiate these two common cross-reactive species [85]. Additionally, a further experimental investigation is necessary to assess the influence of environmental humidity on the gas sensing properties of rGO-SMOx nanocomposites and to further comprehend the reaction mechanism towards both NH 3 and H 2 S. Table 2. Variation function of microsensor resistance during exposure step to gases for 2 different working temperatures. The gas can be identified from the sign of resistance variations.

R(rGO/In 2 O 3 ) during Gas Exposure
Gas Identified @ 100 • C @ 160 • C H 2 S NH 3 NO 2 ing the best sensing performances can be used to quantify gas concentration. Consequently, our measurements open new perspectives for studying rGO-SMOx-based gas sensors in the multi-sensor array or electronics nose context [84], since rGO-In2O3 nanocomposites could readily differentiate these two common cross-reactive species [85]. Additionally, a further experimental investigation is necessary to assess the influence of environmental humidity on the gas sensing properties of rGO-SMOx nanocomposites and to further comprehend the reaction mechanism towards both NH3 and H2S. Figure 9. Dynamic response-recovery curves from sensors based on rGO-50 v/v%-In2O3 nanocomposites operating at different temperatures towards H2S at 100 °C (a) and 160 °C (d); towards NH3 at 100 °C (b) and 160 °C (e); and towards NO2 at 100 °C (c) and 160 °C (f).

Conclusions
This manuscript described a systematic study on the gas sensing properties of rGO-In2O3 nanocomposites. Our results demonstrated that the electrons from the n-type In2O3 nanoparticles make the rGO matrix more sensitive towards oxidizing gases such as NO2 Figure 9. Dynamic response-recovery curves from sensors based on rGO-50 v/v%-In 2 O 3 nanocomposites operating at different temperatures towards H 2 S at 100 • C (a) and 160 • C (d); towards NH 3 at 100 • C (b) and 160 • C (e); and towards NO 2 at 100 • C (c) and 160 • C (f).

Conclusions
This manuscript described a systematic study on the gas sensing properties of rGO-In 2 O 3 nanocomposites. Our results demonstrated that the electrons from the n-type In 2 O 3 nanoparticles make the rGO matrix more sensitive towards oxidizing gases such as NO 2 and O 3 , which agrees with the literature for these nanocomposites. Regarding the O 3 detection, it is critical to note that the strong oxidative nature of O 3 can permanently react with rGO when the devices are operating at temperatures of 160 • C. Regarding the interaction of the rGO-In 2 O 3 nanocomposites with reducing gases, our results strongly suggest that NH 3 reacts preferably with the rGO layers, since the reactivity of such nitrogenated species (NO 2 and NH 3 ) are strongly dependent on the remaining oxygenated groups on the rGO's surface. On the other hand, the interaction with H 2 S occurs at the p-type layers of rGO and the n-type In 2 O 3 nanoparticles. In other words, there is a competitive phenomenon between the free adsorption sites at the rGO layers and the In 2 O 3 nanoparticles. For instance, at lower temperatures, the H 2 S molecules would find more adsorption sites on the rGO layers than on the In 2 O 3 nanoparticles, because the carrier density of In 2 O 3 at low temperatures is low. On the other hand, as the In 2 O 3 nanoparticles concentration and the operating temperature increase, the number of free electrons available to interact with the gas also increases. Consequently, the measured electrical resistance would be a consequence of the net interaction of the gas and the type of adsorption sites available to be adsorbed. Regarding research perspectives, this work has demonstrated the need for