ZnO nanomaterials were obtained by two different synthesis procedures namely the sol-gel and electrospinning methods. Sol-gel technique has been largely used to synthesize metal oxides for sensing applications. Here, ZnO nanopowders have been obtained by a new approach based on slow hydrolysis of the precursor using an esterification reaction in the alcoholic medium, followed by a supercritical drying in ethanol [19]. Thus, powders of ZnO with size in the nanometer range can be easily obtained, which could show interesting properties for gas sensing. The nanopowders obtained were annealed at medium (400 °C) and high (700 °C) temperature.

Electrospinning technique also offers numerous advantages for obtaining metal oxide fibers, such as TiO₂, MoO₃, SnO₂ and WO₃, for gas sensors. Electrospun metal oxide fibers are reported to be able in detecting low concentration of gaseous substances with high sensitivity [20]. The one-dimensional nanofibers synthesized and annealed at 400 °C provide high surface-to-volume ratio, then much more adsorption sites for gaseous species are expected to be presents on their surface.

To compare the morphological and microstructural characteristics of samples synthesized, Scanning and Transmission Electron Microscopy (SEM, TEM) and X-ray Diffraction (XRD) analysis have been performed. Figure 1a reports a typical TEM image taken from the sample as-prepared by sol-gel in supercritical conditions. Very small ZnO NPs having size in the nanometer range are observed. The crystallites present a prismatic-like shape with a narrow particle size distribution. The majority of ZnO particles have a size of about 25 nm, in good agreement with the mean particle size, D, deduced from XRD spectrum (see Figure 1b) and calculated using the Scherrer’s formula: (1) where λ is the X-ray wavelength, θ_B is the maximum of the Bragg diffraction peak (in radians) and B is the full width at half maximum (FWHM) of the XRD peak. The spectrum shows only diffraction peaks of ZnO, without any additional diffraction peaks. The indexed peaks are related to the hexagonal wurtzite structure according to JCPDS Data Base [21].

Figure 2 reports a SEM image of electrospun fibers collected directly on the Pt-interdigitated electrode area of alumina sensor support. A network of nanofibers, long several tens of microns can be observed. The electrospun ZnO/PVA fiber mat is composed of irregular fibers having diameter in a large range (100–500 nm), fused among them to form extensive junctions. The distribution of the fibers is fairly random with no distinct preferential alignment.

SEM images showing the morphology of the NPs and NFs after annealing at 400 °C are reported in Figure 3. ZnO NPs particles deposited on the sensor substrate form a porous layer which provides a large number of contacts among the particles (Figure 3a). As a consequence of the elimination of the polymer phase, annealing of the ZnO/PVA fibers (Figure 3b,c), leads to the formation of more homogenous nanofibers with a smaller diameter distribution (50–100 nm).

Sensing tests were carried out by means of the home-made sensing probe shown in Figure 4. The sensor head consists of an alumina substrate with Pt interdigitated electrodes. On the back side of alumina substrate, a Pt heater provides to the heating of the sensing element. The active sensing layer was deposited on the Pt interdigitated electrodes area from an aqueous paste of the samples by screen printing. No binder was necessary to enhance the adhesion of the sensing layer on the alumina substrate.

In order to simplify the functioning of the dual sensor device, the single sensors would operate at the same temperature. Preliminary work has been then devoted to find the optimal operation temperature. The temperature of 350 °C has been chosen on the basis of the better performance obtained in terms of sensitivity, selectivity and fast response/recovery time. Figure 5 shows the dynamic gas sensing characteristics towards CO of the sensor based on ZnO NPs annealed at 400 °C and 700 °C respectively, at the operating temperature of 350 °C. The resistance decreased upon exposure to 5–50 ppm of CO, in agreement with the generally accepted sensing mechanism on n-type oxide semiconductors, such as SnO₂, ZnO, and In₂O₃. The oxidation reaction at the semiconductor surface between the reducing gas and the negatively charged surface-adsorbed oxygen (O⁻ or O²⁻) is at the basis of this mechanism. This process leads to the production of free electrons as follows: CO _(g) + O⁻ _(ads) → CO₂ + e⁻(2)

The injection of produced electrons from the surface to the bulk of semiconducting layer induces consequently a decrease of the sensor resistance proportional to CO concentration. From the obtained results, it appears that the annealing temperature influence the sensor response towards CO. The ZnO NPs sample annealed at 700 °C shows greater response than the annealed at 400 °C one. XRD analysis showed that the average particle size increase with the annealing temperature, from 25 nm for the as prepared ZnO NPs up to about 66 nm for the sample annealed at 700 °C, leading consequently to a decrease of surface area. Then, it can be deduced that the increase of sensitivity observed for the ZnO NPs sample annealed at the higher temperature tested should be attributed to other factors than the increase of surface area. It is well known that the crystallinity systematically increase with the increase of annealing temperature, as well as the formation of a greater number of oxygen vacancies [22]. These factors then could explain the increased ZnO sensor sensitivity for CO at higher annealing temperature. It is also interesting to note that the detection limit of the sensor based on ZnO NPs annealed at 700 °C, can be estimated to be below 5 ppm of CO in air, a value better than those reported under similar operating conditions for ZnO systems in previous papers [23,24].

These factors have a great impact in favoring the response towards CO, while the response to NO₂ is unaffected. Indeed, with NO₂ as target gas at concentrations in the sub-ppm level (≤1 ppm), on ZnO NPs samples annealed at different temperatures, the resistance variations are barely discernible from the baseline. This is clearly observed in the calibration curves, registered at the operating temperature of 350 °C for both CO and NO₂ gases (Figure 6). Thus, the sensor with NPs annealed at 700 °C results strongly selective towards CO.

ZnO NFs based sensor behaves instead in a completely opposite way. Specifically, ZnO NFs sensor displayed negligible response to CO in a very large operating temperature range (see Figure 7).

On the other hand, the ZnO NFs sensor has been found sensitive to NO₂ even at sub-ppm concentrations. Figure 8 shows the dynamic gas sensing characteristics towards NO₂ of the ZnO NFs-based sensor. The resistance increased upon exposure to NO₂, suggesting that the detection mechanism can be associated with the following reaction: NO_2(g) + e⁻ → NO + O⁻_(ads)(3)

Conduction electrons are then consumed and this leads to an increase of the surface resistance, in agreement with results reported for similar one-dimensional ZnO systems [25]. The linear trend of the sensor response vs. the NO₂ concentration is shown in the inset of Figure 8.

The understanding of the exact mechanism is out of the scope of the present paper. However, some brief considerations are here reported. By the results above reported it appears evident that the annealing temperature and shape of ZnO particles are key factors in determining the sensitivity and selectivity of the sensors investigated. Furthermore, they can also influence the electrical characteristics of the sensing layer. It can be noted that the NPs films are more conductive than the NFs films by two orders of magnitude.

In order to explain this finding, we would consider that the electrical conductivity of polycrystalline metal oxides is mainly governed by the potential barrier formed at the grain boundaries and the electrical conduction occurs along percolation paths via grain boundaries contacts [26]. On theis basis, the difference observed in the electrical resistance baseline value between NPs and NFs can be due to many morphological and microstructural factors which influence, consequently, their electrical properties. Further, the direct deposition of electrospun one-dimensional ZnO NFs, leads to a limited number of contact points established between the nanofibers and the underlying electrode, resulting in a poor electrical contact with the electrode surface. This could give the major contribution to the higher resistance of the ZnO NFs sensor compared to ZnO NPs sensor. A schematization of the sensor device is reported in Figure 9. In the case of NPs, there are many contact points which contributes to the electrical conduction. In contrast, the electrical conduction in NFs is limited by the smaller number of contact points present (see also Figure 3c).

Changes in particle shape can also dramatically alter the reactivity of metal oxides when interacting with gaseous species [12], but no clear correlation has been established with sensing properties comparing the data reported in literature. Nanofibers are generally known to favors the NO₂ sensitivity. Baratto et al., demonstrated that the fiber structure produces an increase of relative response towards NO₂ with respect to bulk structure, while they observed no interference from reducing gases such as CO and ethanol [27]. Also Hsueh et al., have shown that ZnO-nanowire based CO sensor presents no response at lower CO concentrations than 500 ppm [28]. This behavior is similar to that reported by us in this work (see the responses to NO₂ and CO of ZnO NFs reported in Figure 9). In contrast, Lee reported that nanofibers of ZnO show a significant response to CO even at very low concentrations [29]. On these bases, it appears clear that more investigations are necessary in order to better clarify this important issue.

From a practical point of view, the sensor stability for prolonged times is an important prerequisite. The sensor devices have been then tested for prolonged times (around one-two months) and no remarkable degradation of the performances has been noted, indicating the good thermo-mechanical and electrical stability of the sensing layer. Therefore, even if the sensitivity to NO₂ is lower with respect to similar systems reported in the literature [25], the easy modulation of the sensitivity and selectivity is very promising, making possible the use of the developed ZnO NPs and ZnO NFs single sensor in a dual device designed for the selective monitoring of CO and NO₂ in ambient air.

The microstructure of the samples was investigated by XRD (AXS D8 Advance; BRUKER, Billerica, MA, USA) using the Cu K_α1 wavelength of 1.5405 Å. The surface morphology of the films was monitored by means of scanning electron microscopy (SEM) tests carried out with a JEOL 5600LV electron microscope (JEOL, Tokyo, Japan). TEM images were recorded on a Technai G20-Stwin transmission electron microscope (FEI, Eindhoven, The Netherlands) using an accelerating voltage of 200 kV.

NPs based sensors were made by printing films (1–10 μm thick) of the nano-powders dispersed in water on alumina substrates (6 mm × 3 mm) with Pt interdigitated electrodes and Pt heater located on the backside. NFs based sensors were made by direct spinning of fibers on alumina substrates. The sensors were then introduced in a stainless steel test chamber for the sensing tests. Electrical measurements were carried out in the temperature range from room temperature to 400 °C, under a synthetic dry air total stream of 100 sccm, collecting the sensors resistance data in the four point mode. Gases coming from certified bottles can be further diluted in air at a given concentration by mass flow controllers. The concentration of CO target gas was varied from 5 to 50 ppm, and the concentration of NO₂ target gas was varied from 0.25 to 2 ppm. A multimeter data acquisition unit Agilent 34970A (Santa Clara, CA, USA) was used for this purpose, while a dual-channel power supplier instrument Agilent E3632A was employed to bias the built-in heater of the sensor to perform measurements at super-ambient temperatures. The gas response, S, is defined as S = R₀/R for CO and S = R/R₀ for NO₂, where R₀ is the baseline resistance in dry synthetic air (20% O₂ in nitrogen) and R is the electrical resistance of the sensor at different CO or NO₂ concentrations, respectively, in dry synthetic air.