Silver Enhances Hematite Nanoparticles Based Ethanol Sensor Response and Selectivity at Room Temperature

Gas sensors are fundamental for continuous online monitoring of volatile organic compounds. Gas sensors based on semiconductor materials have demonstrated to be highly competitive, but are generally made of expensive materials and operate at high temperatures, which are drawbacks of these technologies. Herein is described a novel ethanol sensor for room temperature (25 °C) measurements based on hematite (α‑Fe2O3)/silver nanoparticles. The AgNPs were shown to increase the oxide semiconductor charge carrier density, but especially to enhance the ethanol adsorption rate boosting the selectivity and sensitivity, thus allowing quantification of ethanol vapor in 2–35 mg L−1 range with an excellent linear relationship. In addition, the α-Fe2O3/Ag 3.0 wt% nanocomposite is cheap, and easy to make and process, imparting high perspectives for real applications in breath analyzers and/or sensors in food and beverage industries. This work contributes to the advance of gas sensing at ambient temperature as a competitive alternative for quantification of conventional volatile organic compounds.


Introduction
The sensing of alcohol vapor enables in situ analysis and online monitoring, which allows for faster response in case of public security [1][2][3], safety risks associated with hazardous compounds [4][5][6], and food analysis [7][8][9][10]. Conventional techniques for alcohol identification and quantification such as gas chromatography and spectrophotometric analyses (i.e., IR, UV-Vis) require sample pre-treatment and sophisticated instrumentation [11,12]. Metal oxide semiconductor (MOS) based gas sensors working under the Taguchi principle [13] and are an alternative sensing technology that is not limited to online monitoring and real time analysis of analytes, in contrast with conventional analytical techniques [14][15][16][17]. Specific surface chemical transformation of MOS exposed to volatile species produces a shift of the surface oxygen reaction equilibrium state due to the presence of the target analyte. Such interactions result in a change in the amount of oxygen molecules chemisorbed on the surface, inducing a change in the resistance of the sensor material acting as a transducer. Thus, the change on the semiconductor resistance becomes the physico-chemical response related to analyte concentration [18,19].

Chemicals and Synthesis of Ag-Modified Hematite Nanospheres and Gas Sensors Assembling
Silver decorated hematite nanospheres (α-Fe 2 O 3 /Ag) were synthesized following a modified chemical co-precipitation method [38], as described in our previous work [39]. Briefly, a 125 mL solution of 0.1 mmol L −1 of Fe(NO 3 ) 3 (Merck Co.) and AgNO 3 (Merck Co.) ranging from 0 up to 7.4 mmol L −1 was added dropwise (0.5 mL min −1 ) into 250 mL of 0.3 mol L −1 Na 2 CO 3 (Riedel-deHaen Co.) solution, in the presence of 0.5 g of polyethylene glycol 2000 (Merck Co.) as surfactant. The pH of the solution was maintained constant at 10.8 during the co-precipitation process by the dropwise (1.5 mL min −1 ) addition of 0.1 mol L −1 Na 2 CO 3 . A brownish precipitate was formed immediately upon dropwise Sensors 2021, 21,440 3 of 13 addition of the metal precursor. The mixture was kept under continuous magnetic stirring at 80 • C for 60 min, and the precipitate aged for an additional 12 h in static conditions. The solid was separated by centrifugation, washed three times with ethanol, and dried at 80 • C for 4 h. The recovered solid was milled using an agate mortar and uniformed using stainless steel test sieve, ø 100 µm. Then, the samples were calcined at 400 • C to ensure the formation of hematite as a pure metal oxide phase decorated with silver.

Assembling of α-Fe 2 O 3 /Ag Gas Sensors
The sensors were prepared by depositing a thin layer of the nanoparticles on interdigitated gold electrodes (0.5 mm wide and 0.5 mm apart) on glass substrates (25 mm × 25 mm × 2 mm), prepared by photolithography [40]. This design was chosen to fit in printed circuit boards to connect with external instruments, see Figure 1.Previously, electrodes were cleaned with a 0.05 mol L −1 HCl solution, rinsed with water, and then cleaned with acetone to remove impurities and degrease the surface. The coating was deposited by screen printing using a thin brush. Briefly, a mass of 30 mg of the nanoparticles (hematite or Ag-decorated hematite) was mixed with 2 drops of ethylene glycol to form an impregnation paste. Once the paste was deposited, the screen-printed sensors were submitted to a thermal treatment at 200 • C for 8 h to remove the solvent and to stabilize the coating on the glass substrate surface. Na2CO3. A brownish precipitate was formed immediately upon dropwise addition of the metal precursor. The mixture was kept under continuous magnetic stirring at 80 °C for 60 min, and the precipitate aged for an additional 12 h in static conditions. The solid was separated by centrifugation, washed three times with ethanol, and dried at 80 °C for 4 h. The recovered solid was milled using an agate mortar and uniformed using stainless steel test sieve, ø 100 μm. Then, the samples were calcined at 400 °C to ensure the formation of hematite as a pure metal oxide phase decorated with silver.

Assembling of α-Fe2O3/Ag Gas Sensors
The sensors were prepared by depositing a thin layer of the nanoparticles on interdigitated gold electrodes (0.5 mm wide and 0.5 mm apart) on glass substrates (25 mm × 25 mm × 2 mm), prepared by photolithography [40]. This design was chosen to fit in printed circuit boards to connect with external instruments, see Figure 1. Previously, electrodes were cleaned with a 0.05 mol L −1 HCl solution, rinsed with water, and then cleaned with acetone to remove impurities and degrease the surface. The coating was deposited by screen printing using a thin brush. Briefly, a mass of 30 mg of the nanoparticles (hematite or Ag-decorated hematite) was mixed with 2 drops of ethylene glycol to form an impregnation paste. Once the paste was deposited, the screen-printed sensors were submitted to a thermal treatment at 200 °C for 8 h to remove the solvent and to stabilize the coating on the glass substrate surface.

Analytical Procedures and Instruments
Morphological features of nanoparticles were recorded by scanning electron microscopy (SEM) using a JEOL JSM-7401F SEM microscope (JEOL, Tokyo, Japan) at 2 kV. The samples for SEM analyses were prepared, spreading the nanoparticles powder on carbon adhesive on a copper stub. Scanning transmission electron microscopy (STEM) images were acquired with a JEOL JEM-2100F TEM-FEG microscope (JEOL, Tokyo, Japan) operating at 200 kV and micrography analyzed using the imageJ 1.53e software. Samples of as-synthesized α-Fe2O3/Ag nanoparticles were dispersed in 3 μL of isopropyl alcohol and drop-casted onto ultrathin carbon film coated 400 mesh copper grids. Crystalline phase

Analytical Procedures and Instruments
Morphological features of nanoparticles were recorded by scanning electron microscopy (SEM) using a JEOL JSM-7401F SEM microscope (JEOL, Tokyo, Japan) at 2 kV. The samples for SEM analyses were prepared, spreading the nanoparticles powder on carbon adhesive on a copper stub. Scanning transmission electron microscopy (STEM) images were acquired with a JEOL JEM-2100F TEM-FEG microscope (JEOL, Tokyo, Japan) operating at 200 kV and micrography analyzed using the imageJ 1.53e software. Samples of as-synthesized α-Fe 2 O 3 /Ag nanoparticles were dispersed in 3 µL of isopropyl alcohol and drop-casted onto ultrathin carbon film coated 400 mesh copper grids. Crystalline phase analysis was conducted by X-ray diffractometry (XRD) using a Bruker D2 PHASER benchtop XRD (Bruker AXS GmbH, Karlsruhe, Germany) equipped with a Cu Kα (λ = 0.15418 nm) radiation source operating at 30 kV and 15 mA. Diffractograms were recorded with a scanning window of 2θ angles of 20-70 • with a 0.04 • s −1 step size. The diffraction patterns were compared with the International Centre for Diffraction Data (ICDD) using DIFFRAC.EVA V5.2 software. Specific surface area of samples were acquired by Brunauer-Emmett-Teller (BET) adsorption analyses with N 2 gas operating a Micromeritics Gemini VII 2390t analyzer (Micromeritics Instrumental Corporation, Georgia, GA, USA). Silver content in nanocomposite samples was quantified by inductively coupled plasma optical emission spectroscopy (ICP-OES) using an SPECTRO ARCOS ICP-OES analyzer (SPECTRO Analytical Instruments, Kleve, Germany). Samples were prepared by digesting the nanoparticle samples in 50 mL of aqua regia 8% v/v prior to analyses.

Sensor Testing
Sensor testing was conducted in a 10 cm 3 gas chamber with hermetic lock in with different gas compositions feeding system for analyses, and the total flow rate of gas was 2.4 L min −1 that results in an average time of gas replacement in the measuring chamber of 0.25 s. The gas chamber had an inlet duct for the gas feed, and a purging duct to rinse the gas outside the chamber. The gas composition in the chamber was controlled by a dual gas delivery system consisting of (i) dry air and (ii) analyte gas. The feed of dry air was used as dilution gas to obtain working concentrations of ethanol in the range of 0 to 35 mg L −1 . Dry air was also used as a blank and cleaning gas during hermetic chamber purge. The gas in-flow was measured with independent rotameters for each gas and controlled with electromagnetic gas valves. The composition of the gas chamber was monitored by Bruker-450 gas chromatograph (Bruker Corporation, Ontario Canada). The manufactured hematite-based sensors were put inside the gas chamber and coupled to a HP4156A Semiconductor Parameter Analyzer (Hewlett-Packard, Palo Alto, CA, USA) that is continuously registering the electrical resistance as a function of time. The test voltage was 40 V and the current through the sensors was between 9.1 × 10 −7 to 4.8 × 10 −8 A. Previously, all sensors were exposed to 0.5 L min −1 of dry air for 2 h in order to reach steady background. Then, the sensor signal (S) towards different gases was calculated according to Equation (1): where R air is the steady resistance of sensors in air, whereas, R gas is measured in the presence of the analyte gas at different concentrations with a total flow rate of 2.4 L·min −1 .

Hematite and Ag-Decorated Hematite Characterization
The SEM images of pristine hematite and nano-composite α-Fe 2 O 3 /Ag nanoparticles are shown in Figure 2. The effect of polyethylene glycol during the precipitation of hematite can be deduced by comparing the images in Figure 2a,b. Larger agglomerates are formed in the absence of surfactant impacting the final size and homogeneity of the material. In fact, that surfactant prevents agglomeration and controls the particle size producing nanoparticles of 76 nm. The notoriously larger specific surface area of nanoparticles prepared in the presence of surfactant may benefit the sensor response. Therefore, the Ag modification was performed in the presence of surfactant. The images shown in Figure 2c illustrate that the co-precipitation method affects neither the spheroidal shape of the hematite nanoparticles nor the average size of 75 nm. Nanoparticles with similar morphology were obtained independently of the relative amount of Ag in the nanocomposite material, except for the presence of more or less large amounts of smaller silver nanoparticles (Figure 2c). The TEM micrograph of α-Fe2O3/Ag 3 wt% depicts a homogeneous distribution of approximately 76 nm diameter nanoparticles ( Figure S1 in Supplementary Materials and Figure 3a). Moreover, Figure 3a revealed the chemical composition of nanocomposites by EDX analysis. As seen, iron (Fe) and oxygen (O) were the major component of the samples analyzed. Note that low quantities of silver (Ag) were observed, as expected for the low loadings of 3.0 wt.% of the decorated nanocomposite. Copper signal is an artifact generated from the copper grid where the sample is supported for analysis. The bright-field (BF) STEM and high-angle annular dark-field (HAADF) images also revealed the presence of two phases, as expected for the presence of 2 to 5 nm large Ag nanoparticles decorating the hematite nanoparticles (Figure 3c  The TEM micrograph of α-Fe 2 O 3 /Ag 3 wt% depicts a homogeneous distribution of approximately 76 nm diameter nanoparticles ( Figure S1 in Supplementary Materials and Figure 3a). Moreover, Figure 3a revealed the chemical composition of nanocomposites by EDX analysis. As seen, iron (Fe) and oxygen (O) were the major component of the samples analyzed. Note that low quantities of silver (Ag) were observed, as expected for the low loadings of 3.0 wt% of the decorated nanocomposite. Copper signal is an artifact generated from the copper grid where the sample is supported for analysis. The bright-field (BF) STEM and high-angle annular dark-field (HAADF) images also revealed the presence of two phases, as expected for the presence of 2 to 5 nm large Ag nanoparticles decorating the hematite nanoparticles (Figure 3b The TEM micrograph of α-Fe2O3/Ag 3 wt% depicts a homogeneous distribution of approximately 76 nm diameter nanoparticles ( Figure S1 in Supplementary Materials and Figure 3a). Moreover, Figure 3a revealed the chemical composition of nanocomposites by EDX analysis. As seen, iron (Fe) and oxygen (O) were the major component of the samples analyzed. Note that low quantities of silver (Ag) were observed, as expected for the low loadings of 3.0 wt.% of the decorated nanocomposite. Copper signal is an artifact generated from the copper grid where the sample is supported for analysis. The bright-field (BF) STEM and high-angle annular dark-field (HAADF) images also revealed the presence of two phases, as expected for the presence of 2 to 5 nm large Ag nanoparticles decorating the hematite nanoparticles (Figure 3c Figure 4a depicts the X-ray diffractograms of hematite and nanocomposites α-Fe2O3/Ag with silver contents of up to 5 wt%, consistent with the characteristic rhombohedral lattice system of hematite as described in [39]. Peaks associated to its typical   Figure 4a depicts the X-ray diffractograms of hematite and nanocomposites α-Fe 2 O 3 /Ag with silver contents of up to 5 wt%, consistent with the characteristic rhombohedral lattice system of hematite as described in [39]. Peaks associated to its typical crystallographic planes were observed at 2θ of 24 [41]. XRD peaks associated to silver metal were not observed in the diffractograms, which is commonly observed for nanocomposites with a high dispersion and low content of silver. However, these were clearly observed in the STEM images of Figure 3b,c that demonstrated the successful decoration of hematite NPs surface during the coprecipitation process. Conversely, when the content of Ag reaches values as high as 10 wt%, hematite is not formed during the coprecipitation process anymore. Indeed, the diffractograms of Figure 4b indicate the formation of 3R-AgFeO 2 with delafossite structure with hexagonal lattice. Therefore, composites with more than 5 wt% of silver were excluded from our study.  [41]. XRD peaks associated to silver metal were not observed in the diffractograms, which is commonly observed for nanocomposites with a high dispersion and low content of silver. However, these were clearly observed in the STEM images of Figure  3c,d that demonstrated the successful decoration of hematite NPs surface during the coprecipitation process. Conversely, when the content of Ag reaches values as high as 10 wt%, hematite is not formed during the coprecipitation process anymore. Indeed, the diffractograms of Figure 4b indicate the formation of 3R-AgFeO2 with delafossite structure with hexagonal lattice. Therefore, composites with more than 5 wt% of silver were excluded from our study. The actual amounts of silver present in the hematite/silver nanocomposites were evaluated by digesting them in aqua-regia and analyzing them using a ICP-OES. As can be seen in Table 1, the silver content in the nanocomposite samples showed an excellent agreement with the iron and silver content used in the synthesis. These results, along with the STEM (scanning transmission electron microscopy) (and the enhancement of BET (Brunauer-Emmett-Teller) surface area values of Ag-hematite composites related to pure hematite (Table 1), allow us to infer that silver decorated nanoparticles were successfully prepared based on the coprecipitation method. Table 1. Elemental composition and specific surface area of hematite and α-Fe2O3/Ag nanocomposites. The actual amounts of silver present in the hematite/silver nanocomposites were evaluated by digesting them in aqua-regia and analyzing them using a ICP-OES. As can be seen in Table 1, the silver content in the nanocomposite samples showed an excellent agreement with the iron and silver content used in the synthesis. These results, along with the STEM (scanning transmission electron microscopy) (and the enhancement of BET (Brunauer-Emmett-Teller) surface area values of Ag-hematite composites related to pure hematite (Table 1), allow us to infer that silver decorated nanoparticles were successfully prepared based on the coprecipitation method.

−1
The current-potential curves of the assembled sensors have a linear relationship according to Ohm's law as usually observed for ideal electric resistors ( Figure S2) The smaller the slope, the higher is the resistance. Generally, materials with higher resistivity tend to have higher sensitivity for gas sensor application. Thus, the electrical responses suggest that α-Fe 2 O 3 /Ag 3.0 wt% may present higher sensitivity for gaseous analytes.

Comparing Hematite-Based Sensor Responses to Ethanol
Sensitivity to the analyte is one of the major indicatives of sensor competitiveness. The response of the hematite-based sensors for ethanol depends on the semiconductor resistance change when exposed to the analyte according to Equation (1). Figure 5 presents the response of hematite-based sensors with different contents of silver in the nanocomposite material when exposed to 35 mg L −1 ethanol in gas phase, demonstrating the benefit of silver nanoparticles in the composite material. Clearly the relative signal increases in comparison to pristine hematite. Note that α-Fe 2 O 3 /Ag 3.0 wt% has about 22.9% higher signal than bare α-Fe 2 O 3 . This positive effect is explained by the role of silver in the sensing mechanism when compared to pristine hematite. Commonly, hematite as an ntype semiconductor must reduce its electrical resistance interacting with reducing gases, however, Figure 5 suggested that hematite behaves as a p-type semiconductor because the electrical resistance increased under exposition of ethanol, which is a reducing gas. This effect is not clearly described yet, but some studies attributed this behavior to low quantities of bulk impurities (e.g., Na, Mg, C, etc.) or to the annealing process in oxygen atmosphere. Both phenomena can generate holes in the lattice of hematite [42,43]. To this study, ions Na + from Na 2 CO 3 could influence the transition nto pof the prepared nanocomposites. Therefore, oxygen adsorbed on hematite is chemisorbed into O 2 − ions (T = 25 • C) by trapping electrons from the valence band. As a consequence, a thin layer of holes in the surface of the semiconductor is formed with regards to the p-type behavior, and lower resistance is adopted. During ethanol release, chemisorbed O 2 − and ethanol molecules react, producing CO 2, H 2 O, and free electrons, which return to the valence band, and the holes layer diminishes, increasing the electrical resistance of hematite [44,45]. In contrast, Ag nanoparticles (even small quantities) play a dual role as (i) electron donor and (ii) chemical sensitizer. The role of the electron donor is forming an ohmic contact at the α-Fe 2 O 3 /Ag interface, as a result of metal-semiconductor junction. Electrons transfer from the silver to hematite since the work function of hematite (ϕ m = 5.88 eV, [46]) is higher than the silver work function (ϕ m = 4.2 eV, [47]). Therefore, silver increases the electron density of hematite. Hence, the adsorption/desorption of oxygen molecules ions (O 2 − ) is boosted. As chemical sensitizer, Ag promotes the adsorption of ethanol gas. Ethanol may occupy extra active sites of Ag nanoparticles, interact with oxygen ions, and enhance the electron charge transfer rate between the ethanol gas and the sensor surface. These effects make metal oxide sensors more sensitive for ethanol vapor detection, as observed in Figure 5. This higher sensitivity to ethanol concentration can be inferred from the steeper slope observed for α-Fe 2 O 3 /Ag 3.0 wt% of 0.119 L mg −1 than the 0.096 L mg −1 for pristine α-Fe 2 O 3 (see Figure 5b).
In contrast, an excessive loading of silver has a deleterious impact and drastically decreases the sensor response. The sensor containing 5.0 wt% of Ag shows a dramatic drop in signal response down to 1.7, corresponding to a 51.4% loss in signal response in comparison to pristine α-Fe 2 O 3 , explained by the conductive character of metallic silver increasing the semiconductor conductivity while acting as charge carrier recombination sites. These effects decrease the differential resistance and interaction capacity of the surface with the analyte, thus lowering sensitivity. The response time represents the period of time taken for the sensor to react to a given stimulus. Response time is defined as 90% of the time to reach the equilibrium value of a response signal [48]. Conversely, the recovery time represents the length of time taken for the sensor to return to the initial background condition (i.e., sensor in air). By definition, the recovery time is defined as the 90% of time required to return to the original signal value when in the air after removing the target analyte gas from the measuring chamber [48]. Figure 6 presents the response time and recovery times of the α-Fe2O3/Ag 1-5 wt% nanocomposite sensors. All the sensors showed that similar time of response is dependent on the ethanol concentration. Note that the relationship between time of response trends is dependent on analyte concentration. This is related to the diffusion velocity of the analyte and therefore should be dependent on the analyte gas concentration. Thus, shorter response times were observed for higher analyte concentrations as has been previously reported for other gas sensors operating at 25 °C [49,50]. The opposite trend is observed in Figure 6b for the recovery times, in which longer times are required for higher analyte concentrations. This is explained by the fact that during the gas purging with air, a higher number of adsorbed molecules should be displaced by oxygen.
The nanocomposite α-Fe2O3/Ag 3.0 wt% is the sensor that presents faster response and recovery times as expected by the catalytic role of Ag that enables faster adsorption/desorption of oxygen molecules on the sensor surface. All these results identify the nanocomposite with 3.0 wt% loading of Ag nanoparticles on α-Fe2O3 nanoparticle as the optimum ratio to achieve enhanced sensor response for ethanol analysis. Note that 32 s is the response time at an ethanol concentration of 2 mg L −1 , but it is shortened to less than 15 s for concentrations larger than 5 mg L −1 , whereas the recovery time increased from 22 to 27 s. These results evidence the rapid recovery of the sensor surface by purging with represents the period of time taken for the sensor to react to a e time is defined as 90% of the time to reach the equilibrium value Conversely, the recovery time represents the length of time taken o the initial background condition (i.e., sensor in air). By definis defined as the 90% of time required to return to the original e air after removing the target analyte gas from the measuring resents the response time and recovery times of the α-Fe2O3/Ag sensors. All the sensors showed that similar time of response is ol concentration. Note that the relationship between time of reent on analyte concentration. This is related to the diffusion vetherefore should be dependent on the analyte gas concentration. mes were observed for higher analyte concentrations as has been ther gas sensors operating at 25 °C [49,50]. The opposite trend is the recovery times, in which longer times are required for higher his is explained by the fact that during the gas purging with air, rbed molecules should be displaced by oxygen. α-Fe2O3/Ag 3.0 wt% is the sensor that presents faster response xpected by the catalytic role of Ag that enables faster adsorpn molecules on the sensor surface. All these results identify the wt% loading of Ag nanoparticles on α-Fe2O3 nanoparticle as the enhanced sensor response for ethanol analysis. Note that 32 s is thanol concentration of 2 mg L −1 , but it is shortened to less than rger than 5 mg L −1 , whereas the recovery time increased from 22 dence the rapid recovery of the sensor surface by purging with e and recovery times of α-Fe2O3/Ag 3.0 wt% nanocomposite as re very competitive for real applications, especially the monitorreath and in drinks. epresents the period of time taken for the sensor to react to a time is defined as 90% of the time to reach the equilibrium value onversely, the recovery time represents the length of time taken the initial background condition (i.e., sensor in air). By definidefined as the 90% of time required to return to the original air after removing the target analyte gas from the measuring esents the response time and recovery times of the α-Fe2O3/Ag sensors. All the sensors showed that similar time of response is l concentration. Note that the relationship between time of rent on analyte concentration. This is related to the diffusion veherefore should be dependent on the analyte gas concentration. es were observed for higher analyte concentrations as has been ther gas sensors operating at 25 °C [49,50]. The opposite trend is the recovery times, in which longer times are required for higher his is explained by the fact that during the gas purging with air, bed molecules should be displaced by oxygen. α-Fe2O3/Ag 3.0 wt% is the sensor that presents faster response pected by the catalytic role of Ag that enables faster adsorpmolecules on the sensor surface. All these results identify the t% loading of Ag nanoparticles on α-Fe2O3 nanoparticle as the enhanced sensor response for ethanol analysis. Note that 32 s is thanol concentration of 2 mg L −1 , but it is shortened to less than ger than 5 mg L −1 , whereas the recovery time increased from 22 ence the rapid recovery of the sensor surface by purging with and recovery times of α-Fe2O3/Ag 3.0 wt% nanocomposite as e very competitive for real applications, especially the monitoreath and in drinks. The response time represents the period of time taken for the senso given stimulus. Response time is defined as 90% of the time to reach the equ of a response signal [48]. Conversely, the recovery time represents the lengt for the sensor to return to the initial background condition (i.e., sensor in tion, the recovery time is defined as the 90% of time required to return signal value when in the air after removing the target analyte gas from chamber [48]. Figure 6 presents the response time and recovery times of t 1-5 wt% nanocomposite sensors. All the sensors showed that similar time dependent on the ethanol concentration. Note that the relationship betw sponse trends is dependent on analyte concentration. This is related to th locity of the analyte and therefore should be dependent on the analyte gas Thus, shorter response times were observed for higher analyte concentratio previously reported for other gas sensors operating at 25 °C [49,50]. The op observed in Figure 6b for the recovery times, in which longer times are requ analyte concentrations. This is explained by the fact that during the gas pu a higher number of adsorbed molecules should be displaced by oxygen.
The nanocomposite α-Fe2O3/Ag 3.0 wt% is the sensor that presents and recovery times as expected by the catalytic role of Ag that enables tion/desorption of oxygen molecules on the sensor surface. All these resu nanocomposite with 3.0 wt% loading of Ag nanoparticles on α-Fe2O3 nano optimum ratio to achieve enhanced sensor response for ethanol analysis. N the response time at an ethanol concentration of 2 mg L −1 , but it is shorten 15 s for concentrations larger than 5 mg L −1 , whereas the recovery time inc to 27 s. These results evidence the rapid recovery of the sensor surface by air. These short response and recovery times of α-Fe2O3/Ag 3.0 wt% nan ethanol sensor at 25 °C are very competitive for real applications, especial ing of ethanol levels in breath and in drinks. The response time represents the period of time taken for the sens given stimulus. Response time is defined as 90% of the time to reach the eq of a response signal [48]. Conversely, the recovery time represents the leng for the sensor to return to the initial background condition (i.e., sensor in tion, the recovery time is defined as the 90% of time required to return signal value when in the air after removing the target analyte gas from chamber [48]. Figure 6 presents the response time and recovery times of 1-5 wt% nanocomposite sensors. All the sensors showed that similar tim dependent on the ethanol concentration. Note that the relationship betw sponse trends is dependent on analyte concentration. This is related to t locity of the analyte and therefore should be dependent on the analyte ga Thus, shorter response times were observed for higher analyte concentrat previously reported for other gas sensors operating at 25 °C [49,50]. The o observed in Figure 6b for the recovery times, in which longer times are req analyte concentrations. This is explained by the fact that during the gas p a higher number of adsorbed molecules should be displaced by oxygen. The nanocomposite α-Fe2O3/Ag 3.0 wt% is the sensor that presents and recovery times as expected by the catalytic role of Ag that enable tion/desorption of oxygen molecules on the sensor surface. All these res nanocomposite with 3.0 wt% loading of Ag nanoparticles on α-Fe2O3 nan optimum ratio to achieve enhanced sensor response for ethanol analysis.  The response time represents the period of time taken for the sensor to react to a given stimulus. Response time is defined as 90% of the time to reach the equilibrium value of a response signal [48]. Conversely, the recovery time represents the length of time taken for the sensor to return to the initial background condition (i.e., sensor in air). By definition, the recovery time is defined as the 90% of time required to return to the original signal value when in the air after removing the target analyte gas from the measuring chamber [48]. Figure 6 presents the response time and recovery times of the α-Fe 2 O 3 /Ag 1-5 wt% nanocomposite sensors. All the sensors showed that similar time of response is dependent on the ethanol concentration. Note that the relationship between time of response trends is dependent on analyte concentration. This is related to the diffusion velocity of the analyte and therefore should be dependent on the analyte gas concentration. Thus, shorter response times were observed for higher analyte concentrations as has been previously reported for other gas sensors operating at 25 • C [49,50]. The opposite trend is observed in Figure 6b for the recovery times, in which longer times are required for higher analyte concentrations. This is explained by the fact that during the gas purging with air, a higher number of adsorbed molecules should be displaced by oxygen.
The nanocomposite α-Fe 2 O 3 /Ag 3.0 wt% is the sensor that presents faster response and recovery times as expected by the catalytic role of Ag that enables faster adsorption/desorption of oxygen molecules on the sensor surface. All these results identify the nanocomposite with 3.0 wt% loading of Ag nanoparticles on α-Fe 2 O 3 nanoparticle as the optimum ratio to achieve enhanced sensor response for ethanol analysis. Note that 32 s is the response time at an ethanol concentration of 2 mg L −1 , but it is shortened to less than 15 s for concentrations larger than 5 mg L −1 , whereas the recovery time increased from 22 to 27 s. These results evidence the rapid recovery of the sensor surface by purging with air. These short response and recovery times of α-Fe 2 O 3 /Ag 3.0 wt% nanocomposite as ethanol sensor at 25 • C are very competitive for real applications, especially the monitoring of ethanol levels in breath and in drinks. The sensor was submitted to several cycles of analyte exposure and purging, showing excellent repeatability and stability of the response signal, as can be seen in Figure 7. Note that the relative resistance change of the sensor as response to the concentration of ethanol in the gas chamber has a good linearity and low deviation (95% confidence interval calculated to 4 values) enabling precise quantification of that analyte in different products (see Figure 7b). Besides, cyclic measurements (inset) for linear regression showed a stable baseline (R0) through the test. Thus, α-Fe2O3/Ag 3.0 wt% have demonstrated appropriate features for ethanol vapor sensing such as other hematite-based sensors [23,[35][36][37][38]. Higher sensibility at 25 °C, short-time gas response, rapid recovery, and excellent repeatability (ΔR/Ro), even at low concentrations (2 mg L −1 ), is the greater contribution of this work. However, relative humidity may reduce the sensor performance. Water molecules may occupy the active sites of sensors and avoid the effective detection of ethanol. High relative humidity even produces a thicker water layer on the surface and considerably reduces sensors' activities as previously reported [51,52]. Further studies should consider the effects of variable humidities and their impact on sensitivity at ambient temperature. The response time represents the period of time taken for the sensor to react to a en stimulus. Response time is defined as 90% of the time to reach the equilibrium value response signal [48]. Conversely, the recovery time represents the length of time taken the sensor to return to the initial background condition (i.e., sensor in air). By defini-, the recovery time is defined as the 90% of time required to return to the original nal value when in the air after removing the target analyte gas from the measuring mber [48]. Figure 6 presents the response time and recovery times of the α-Fe2O3/Ag wt% nanocomposite sensors. All the sensors showed that similar time of response is endent on the ethanol concentration. Note that the relationship between time of rense trends is dependent on analyte concentration. This is related to the diffusion veity of the analyte and therefore should be dependent on the analyte gas concentration. us, shorter response times were observed for higher analyte concentrations as has been viously reported for other gas sensors operating at 25 °C [49,50]. The opposite trend is erved in Figure 6b for the recovery times, in which longer times are required for higher lyte concentrations. This is explained by the fact that during the gas purging with air, igher number of adsorbed molecules should be displaced by oxygen.
The nanocomposite α-Fe2O3/Ag 3.0 wt% is the sensor that presents faster response recovery times as expected by the catalytic role of Ag that enables faster adsorp-/desorption of oxygen molecules on the sensor surface. All these results identify the ocomposite with 3.0 wt% loading of Ag nanoparticles on α-Fe2O3 nanoparticle as the imum ratio to achieve enhanced sensor response for ethanol analysis. Note that 32 s is response time at an ethanol concentration of 2 mg L −1 , but it is shortened to less than s for concentrations larger than 5 mg L −1 , whereas the recovery time increased from 22 7 s. These results evidence the rapid recovery of the sensor surface by purging with . These short response and recovery times of α-Fe2O3/Ag 3.0 wt% nanocomposite as anol sensor at 25 °C are very competitive for real applications, especially the monitorof ethanol levels in breath and in drinks. The response time represents the period of time taken for the sensor to react to a n stimulus. Response time is defined as 90% of the time to reach the equilibrium value response signal [48]. Conversely, the recovery time represents the length of time taken the sensor to return to the initial background condition (i.e., sensor in air). By defini-, the recovery time is defined as the 90% of time required to return to the original al value when in the air after removing the target analyte gas from the measuring mber [48]. Figure 6 presents the response time and recovery times of the α-Fe2O3/Ag wt% nanocomposite sensors. All the sensors showed that similar time of response is endent on the ethanol concentration. Note that the relationship between time of rense trends is dependent on analyte concentration. This is related to the diffusion vety of the analyte and therefore should be dependent on the analyte gas concentration. s, shorter response times were observed for higher analyte concentrations as has been viously reported for other gas sensors operating at 25 °C [49,50]. The opposite trend is erved in Figure 6b for the recovery times, in which longer times are required for higher lyte concentrations. This is explained by the fact that during the gas purging with air, gher number of adsorbed molecules should be displaced by oxygen. The nanocomposite α-Fe2O3/Ag 3.0 wt% is the sensor that presents faster response recovery times as expected by the catalytic role of Ag that enables faster adsorp-/desorption of oxygen molecules on the sensor surface. All these results identify the ocomposite with 3.0 wt% loading of Ag nanoparticles on α-Fe2O3 nanoparticle as the imum ratio to achieve enhanced sensor response for ethanol analysis. Note that 32 s is response time at an ethanol concentration of 2 mg L −1 , but it is shortened to less than for concentrations larger than 5 mg L −1 , whereas the recovery time increased from 22 7 s. These results evidence the rapid recovery of the sensor surface by purging with These short response and recovery times of α-Fe2O3/Ag 3.0 wt% nanocomposite as nol sensor at 25 °C are very competitive for real applications, especially the monitorof ethanol levels in breath and in drinks. The response time represents the period of tim given stimulus. Response time is defined as 90% of th of a response signal [48]. Conversely, the recovery tim for the sensor to return to the initial background co tion, the recovery time is defined as the 90% of tim signal value when in the air after removing the tar chamber [48]. Figure 6 presents the response time a 1-5 wt% nanocomposite sensors. All the sensors sho dependent on the ethanol concentration. Note that sponse trends is dependent on analyte concentratio locity of the analyte and therefore should be depend Thus, shorter response times were observed for high previously reported for other gas sensors operating observed in Figure 6b for the recovery times, in whic analyte concentrations. This is explained by the fact a higher number of adsorbed molecules should be d The nanocomposite α-Fe2O3/Ag 3.0 wt% is the and recovery times as expected by the catalytic ro tion/desorption of oxygen molecules on the sensor nanocomposite with 3.0 wt% loading of Ag nanopa optimum ratio to achieve enhanced sensor response the response time at an ethanol concentration of 2 m 15 s for concentrations larger than 5 mg L −1 , whereas to 27 s. These results evidence the rapid recovery o air. These short response and recovery times of αethanol sensor at 25 °C are very competitive for real ing of ethanol levels in breath and in drinks. The response time represents the period of t given stimulus. Response time is defined as 90% of of a response signal [48]. Conversely, the recovery t for the sensor to return to the initial background c tion, the recovery time is defined as the 90% of t signal value when in the air after removing the t chamber [48]. Figure 6 presents the response time 1-5 wt% nanocomposite sensors. All the sensors sh dependent on the ethanol concentration. Note tha sponse trends is dependent on analyte concentrati locity of the analyte and therefore should be depen Thus, shorter response times were observed for hig previously reported for other gas sensors operating observed in Figure 6b for the recovery times, in whi analyte concentrations. This is explained by the fac a higher number of adsorbed molecules should be The nanocomposite α-Fe2O3/Ag 3.0 wt% is th and recovery times as expected by the catalytic r tion/desorption of oxygen molecules on the sensor nanocomposite with 3.0 wt% loading of Ag nanop optimum ratio to achieve enhanced sensor respons the response time at an ethanol concentration of 2 15 s for concentrations larger than 5 mg L −1 , wherea to 27 s. These results evidence the rapid recovery air. These short response and recovery times of α ethanol sensor at 25 °C are very competitive for rea ing of ethanol levels in breath and in drinks. The sensor was submitted to several cycles of analyte exposure and purging, showing excellent repeatability and stability of the response signal, as can be seen in Figure 7. Note that the relative resistance change of the sensor as response to the concentration of ethanol in the gas chamber has a good linearity and low deviation (95% confidence interval calculated to 4 values) enabling precise quantification of that analyte in different products (see Figure 7b). Besides, cyclic measurements (inset) for linear regression showed a stable baseline (R 0 ) through the test. Thus, α-Fe 2 O 3 /Ag 3.0 wt% have demonstrated appropriate features for ethanol vapor sensing such as other hematite-based sensors [23,[35][36][37][38]. Higher sensibility at 25 • C, short-time gas response, rapid recovery, and excellent repeatability (∆R/R o ), even at low concentrations (2 mg L −1 ), is the greater contribution of this work. However, relative humidity may reduce the sensor performance. Water molecules may occupy the active sites of sensors and avoid the effective detection of ethanol. High relative humidity even produces a thicker water layer on the surface and considerably reduces sensors' activities as previously reported [51,52]. Further studies should consider the effects of variable humidities and their impact on sensitivity at ambient temperature.
The evaluation of reactor selectivity is relevant when considering the effect of interferent species on the sensor response. The sensor based on α-Fe 2 O 3 /Ag 3.0 wt% was exposed to high concentrations of four common gas analytes (methane, propane, sulfur dioxide, and methyl mercaptan) and the impact on the sensor response depicted in Figure 8. Methane and propane at 100 mg L −1 showed ∆R/R 0 signal responses as low as 0.02. Additionally, the sensor had low response to sulfur dioxide, which at concentrations of 200 mg L −1 gave a small signal of 0.08. The gas methyl mercaptan had a much more relevant signal with an ∆R/R 0 response of 0.7 at 80 mg L −1 , much lower than for ethanol. In fact, even at 4 times lower concentration of 20 mg L −1 , the hematite-based sensor at 25 • C showed a much higher response to ethanol of ∆R/R 0 = 2.4, which is 3.4-fold larger than that of methyl mercaptan. These results demonstrate the high selectivity of α-Fe 2 O 3 /Ag 3.0 wt% as an ethanol sensor, especially considering that methyl mercaptan will hardly be an interference in most application conditions. work. However, relative humidity may reduce the sensor performance. Water molecules may occupy the active sites of sensors and avoid the effective detection of ethanol. High relative humidity even produces a thicker water layer on the surface and considerably reduces sensors' activities as previously reported [51,52]. Further studies should consider the effects of variable humidities and their impact on sensitivity at ambient temperature.   The evaluation of reactor selectivity is relevant when considering the effect of i ferent species on the sensor response. The sensor based on α-Fe2O3/Ag 3.0 wt.% wa posed to high concentrations of four common gas analytes (methane, propane, sulfu oxide, and methyl mercaptan) and the impact on the sensor response depicted in F 8. Methane and propane at 100 mg L −1 showed ΔR/R0 signal responses as low as 0.02 ditionally, the sensor had low response to sulfur dioxide, which at concentrations o mg L −1 gave a small signal of 0.08. The gas methyl mercaptan had a much more rele signal with an ΔR/R0 response of 0.7 at 80 mg L −1 , much lower than for ethanol. In even at 4 times lower concentration of 20 mg L −1 , the hematite-based sensor at 2 showed a much higher response to ethanol of ΔR/R0 = 2.4, which is 3.4-fold larger that of methyl mercaptan. These results demonstrate the high selectivity of α-Fe2O3/A wt% as an ethanol sensor, especially considering that methyl mercaptan will hardly interference in most application conditions.

Conclusions
Pristine hematite (α-Fe2O3) nanoparticle and decorated with silver were prepare

Conclusions
Pristine hematite (α-Fe 2 O 3 ) nanoparticle and decorated with silver were prepared according to a facile co-precipitation method and used to assemble gas sensors by depositing a thin layer on glass supported interdigitated gold electrodes by screen printing. Characterization techniques allowed inferring the successful decoration of hematite nanoparticles with Ag and the silver content quantification. The STEM images revealed the presence of 76 nm large nanoparticles whose films followed Ohm's law. The sensor demonstrated high sensitivity towards ethanol vapor at an ambient temperature (25 • C) with good linearity in the 2-35 mg L −1 ethanol vapor concentration range. The results indicated that decoration of hematite nanoparticles with silver accelerates adsorption/desorption of oxygen leading to shorter sensor response and recovery times. Furthermore, the presence of silver up to 3.0 wt% enhanced the sensor sensitivity, selectivity, and reproducibility, but a larger silver loading showed a deleterious effect on sensor sensitivity as a consequence of the larger conductivity and charge recombination. Possible interferant gases such as methane, ethane, sulfur dioxide, and methyl mercaptan showed significantly lower responses as compared to ethanol, demonstrating its good selectivity. These results suggest a high potentiality of ethanol sensors based on α-Fe 2 O 3 /Ag 3.0 wt% for sensor applications in breath analyzers and/or food and beverage industries given its simple fabrication process, low cost, and sensitivity even at ambient temperature.