3.1. Morphological, Textural, and Structural Characterizations
The common and stable zinc oxide phase at ambient conditions is the hexagonal wurtzite structure, which consists of alternating planes composed of tetrahedrally coordinated O
2− and Zn
2+ ions, stacked alternately along the c-axis (polar axis). Polar surfaces are typical of ZnO, with the basal plane being the most common. The oppositely charged ions produce positively charged Zn-(0001) and negatively charged O-(000-1) surfaces, resulting in a spontaneous polarization along the c-axis as well as a different surface energy. To maintain a stable structure, the polar surfaces generally exhibit facets or surface reconstructions, but in ZnO they are atomically flat, stable, and without reconstruction [
57]. The other two most commonly observed facets for ZnO are {10–10} and {11–20}, which are non-polar surfaces and have lower energy than the {0001} facets [
58]. ZnO different morphologies can be grown by tuning the growth rates along these directions. Macroscopically, a crystal has different kinetic parameters for different crystal planes, which are emphasized under controlled growth conditions. Thus, after an initial period of nucleation and incubation, a crystallite will commonly develop into a three-dimensional object with well-defined, low index crystallographic faces. It has been observed, firstly by Laudise [
59], that the relationship between the velocities of ZnO crystal growth to different directions is:
v [0001] >
v [10–11] >
v [10–10].
The
Figure 2 shows the comparison between the leaves and bisphenoids precursor (samples before calcination process) and the leaves and bisphenoids after calcination. Leaves resulted from the ε-Zn(OH)
2 direct precipitation in the synthesis solution starting from the base addition, while bisphenoids derived from the spontaneous growth of leaves kept in the synthesis solution. Leaves samples are constituted of leaves of rhombic shape, which have the longest diagonal of about 6 μm in size, while the shortest diagonal one is about 3 μm long. ZnO leaves (
Figure 2b) are also characterized by a fine structure made up of nanoparticles of about 50–60 nm size. In the case of bisphenoids samples, SEM micrographs reveal an octahedral shape, which axis is about 10 µm. Again, we observe the well crystallized surfaces of bisphenoids precursor, and ZnO bisphenoids made up of nanoparticles with a size slightly smaller than that of the leaves powders. In literature, this morphology is known as octahedral [
60,
61], and is interpreted by us as an association of two bisphenoids ({111} and {11–1}) of original ε-Zn(OH)
2, as described in the next section.
Concerning FE-SEM analysis on the third powder (
Figure 3a, b), this shows a stars-like morphology in which stars have a diameter of about 1 µm. In
Figure 3b, the stars appear as if they were composed of grains about 30 nm in size. Each star is made by a central ellipsoid from which six tips radiate. This morphology reminds the 2D scheme of ZnO flowers shown in
Figure 1.
In
Figure 4, the micrographs of needles, long needles, prisms, and flowers are collected. As regards needles, they are characterized by a homogeneous distribution of elongated crystals, 200–300 nm wide and about 3–4 μm long. Long needles powders are constituted of crystals about 100 nm wide and 7–8 μm long. Hexagonal prisms about 100 nm wide represent the second last morphology. Finally, the last sample is characterized by micro flowers in which petals are hexagonal prisms, about 300–400 nm wide, and 5–6 μm long. The growth of the flower structures could be explained taking into account a possible twinning plane (80–89) that corresponds to the plane (10–11). The (80–89) twinning plane in zincite was firstly reported in Dana’s Series of Mineralogies [
62]. Needles, long needles, prisms, and petals of flowers are all different crystal morphologies of ZnO. The only presence of hexagonal prisms (prisms and flowers samples) is due to the slower growth promoted by a weak base (HMTA). On the contrary, the presence of two pyramids both on the top and the bottom of the hexagonal prisms (needles and long needles samples) reveals a faster growth promoted by ammonia solution, which is a base stronger than HMTA.
Textural characterization gave specific surface areas (SSA) that can be divided into two ranges of values. Stars, needles, long needles, prisms and flowers showed a SSA under 10 m2/g, where the lowest is that of stars sample (5 m2/g). Leaves and bisphenoids exhibited a highest SSA of about 20 m2/g.
Focusing on the structural characterization, all synthesized materials presented a hexagonal wurtzite structure (space group P6
3mc) irrespective both on the synthesis method and the particle morphology, as shown in the XRD patterns (
Figure 5). In
Table 1, the unit cell parameters, the volumes, the Bragg R factors, and the crystallite sizes, evaluated by the Scherrer’s formula, are summarized. Only the crystallite sizes showed some differences. Leaves, bisphenoids and stars had the smallest crystallites with average dimension under 30 nm; long needles were characterized by an intermediate crystallite size of about 40 nm, while needles, prisms, and flowers exhibited an average crystallite dimension of about 50 nm. Considering the last four morphologies, the average crystallite sizes were apparently in disagreement with SEM images, which showed monocrystals of several hundreds of nanometers. However, the Scherrer’s formula allows to calculate the average length of the almost perfect crystals, Thereby, for needles, long needles, prisms, and flowers, the average crystallite size can be considered as an average distance between the growth defects (such as dislocations, stacking faults, etc.) due to the rapid crystallization involved in the specific synthesis methods.
3.2. Morphology Dependence on ZnO-Precursor
In the case of stars, needles, long needles, prisms, and flowers, the shape is directly related to the ZnO crystalline phase through a rapid crystallization in slightly different synthesis conditions. On the contrary, as clearly observable in the
Figure 2, in the cases of leaves and bisphenoids, a particularly important topic must be highlighted: they exhibit a crystalline solid precursor as an intermediate synthesis product that influences the final ZnO morphology. Leaves and bisphenoids’ solid precursor, as confirmed by SEM (
Figure 2a,c) and XRD patterns (
Figure 6), crystallize as orthorhombic ε-Zn(OH)
2, space group P2
12
12
1, parametric units cell a = 4.9050 (1) Å, b = 5.1457 (1) Å, c = 8.4780 (2) Å and volume of 214.176 (8) Å
3 according to JCPDS Card No. 38–0385. During the calcination process, the solid precursor changes into ZnO nanoparticles through an in situ crystallization [
63], releasing hydroxide ions (OH-), but maintaining the original shape and dimensions. In these cases, ZnO behaves as a pseudomorphic material: it is a crystal of a mineral that, subjected to an alteration (partial addition or partial removal of original material), changes the internal structure or chemical composition, but the external form is preserved. Such a mineral with outward crystal form of another mineral species is known as “pseudomorphism” [
62]. Pseudomorphs usually are further defined according the way in which they were formed, as by substitution, encrustation, or alteration. The ZnO pseudomorphism occurs for alteration as well as for example for CaSO
4 that becomes CaSO
4 × 2H
2O.
3.4. Electrical Characterization
In metal oxide gas sensors, the electrical properties are firstly dependent on their aptitude to modify the electrical conductivity when they come into contact with the gases of the surrounding atmosphere. On the basis of this peculiarity, a great amount of features influence the sensor characteristics. Of course, the functional material is the key constituent of the sensor, but the morphology and the size of its particles are certainly critical factors to determine its detection properties [
64]. In particular, morphology and size influence the specific surface area and consequently the surface charge density, which is a function closely connected with the reception property of the sensor itself [
65]. Besides that, many other factors are crucial for gas sensor performance, being the first one, the fabrication procedure (thin or thick film) and correspondently porosity, particle agglomeration, sintering control, etc. All these features together with the functional material affect the transduction function, which in turn determines the ability of the sensor to modify its conductivity in presence of the test gases.
Apart from the aptitude of the specific morphologies to detect the various gases, it is however of the utmost importance to verify for all samples the electrical stability and repeatability. Indeed, they are essential characteristics in order to obtain efficient and reliable sensors able to operate continuously for long time. In
Figure 8, an experiment on a bisphenoid’s, a star’s, a prism’s, and a flower’s sensors alternately subjected to 0 and 1 ppm of acetone is reported. This test highlights the good repeatability of the sensors, confirming that the screen-printing technology together with a proper powder preparation permit to fabricate thick film gas sensors with optimal electrical properties [
53].
Other features of screen-printing technology, such as the high firing temperature or the chance to fabricate in one batch up to 500 sensors, are conditions that contribute to assess screen-printed MOX gas sensors highly reliable among the various kinds of chemical gas sensors.
Considering the specific electrical characterizations, an Arrhenius plot, which consists of the measurement of the conductance as a function of temperature in different atmospheres, was carried out. In this work, dry air and a mixture of dry air and acetone (10 ppm) were chosen, for comparison. Such measurement highlighted for all samples a n-type semiconductor behavior, the conductivity exhibiting a trend with temperature divided in more ranges, behavior ascribable to different chemical adsorption phenomena. Such particular trend is typical of MOX gas sensors and justified by the formation of inter-grain Schottky barriers that modulate the film conductivity [
24]. Moreover, such measurement gives information on the magnitude of conductivity and a preliminary study of the response toward the tested gas. For the sake of clarity, in
Figure 9, it has been chosen to report the measurements relating to the sensors showing the behavior with the greatest difference of all the samples. Specifically, the conductance as a function of the inverse of temperature of a bisphenoids and a flowers sample are reported, the other samples positioning in between. Two main observations can be made: (i) concerning the conductivity, it resulted higher in the thick films composed of aggregates of nanoparticles with respect to the ones in form of crystals, a phenomenon probably due to the different inter-particles percolation paths; (ii) with respect to the difference between the measurement in air or in 10 ppm of acetone in air, it resulted greater for the bisphenoids sample than for the flower sample. Since acetone is a reducing gas, the result is a clear indication of the greater response capacity toward acetone of the bisphenoids sample than for the flowers one.
Passing to dynamical measurements of conductance in presence of the tested gases (acetaldehyde C
2H
4O, acetone C
3H
6O, ethanol C
2H
6O, toluene C
7H
8, isoprene C
5H
8, and ammonia NH
3), the responses toward each tested gas at the concentration of 10 ppm in dry air, for the sensors based on bisphenoids, stars, prisms, and flowers are reported in
Figure 10. Such response values are calculated as mean values of couples of nominally identical sensors, and they are moreover measured at the operating temperature for which the response resulted maximized, as shown in the figure for each sensor. The results of the
Figure 10 are relative to the sensors that showed the best performance (bisphenoids), the worst (flowers), while the based-prisms sensors are shown as example of crystals with peculiar electronic characteristics, as reported in a previous work of some of the authors [
54]. Finally, the stars-based sensors have been chosen as a case of only apparent nano-particle aggregate. The results of
Figure 10 are in good agreement with the morphological, textural, and structural characterizations. In this context, particular attention shall be paid to the evolution of crystallites as grown until their transformation throughout all the cycle of preparation of thick films. Indeed, in the case of leaves and bisphenoids, the crystallites agglomerate to form nanometric roundish grains proportionally sized to that of crystallites, the grains moderately coalescing under firing. As described above, differently from nano-particle aggregates, needles, long needles, prisms, and flowers resulted monocrystals in which the crystallites must be considered only as the regions of almost perfect crystal periodicity, while in stars morphology, the crystallites constitute a “crystallographic continuum” i.e., the first stage of a monocrystal. Due to such characteristics, the crystallite dimensions of the crystalline samples cannot be directly connected with the sensors performances. For this purpose, it resulted instead as more appropriate to compare the crystalline morphologies as offered by SEM observations.
Going back to the electrical measurements, it can be observed in the
Figure 10 that all the sensors showed the best responses toward few gases: acetone, acetaldehyde, and isoprene, the morphology not giving particular aptitude for a specific gas.
The responses at 10 ppm of acetone, which is certainly the gas offering the largest responses, are summarized in the
Table 2 for all the tested temperatures. It resulted that the sensors prepared with powders constituted of nanoparticle aggregates in form of bisphenoids and leaves showed the highest responses. In particular, bisphenoids exhibited better performance and also smaller grains with respect to leaves, the latter offering larger grains and correspondently worse performance, according to [
32]. On the other hand, all the sensors prepared with nanocrystals resulted less able to detect the reducing tested gases. With respect to the test in acetone, the maximum response values resulted in decreasing order for long needles, needles and flowers, respectively. Even in the case of crystal shaped samples, the value of the gas response is inversely correlated to the size of the crystal, in agreement with the SEM micrographs carried out on the thick films (see
Figure 7). Such difference of performance between sensors based on nanoparticle aggregates and monocrystals was already described and interpreted for few types of crystals [
34,
54]. The same result has been now demonstrated valid for a large number of crystalline morphologies, including the particular case of stars, which was shown to be an apparent case of nano-crystal aggregate.
Moreover, in addition to the other tested gases, ZnO sensors underwent to formaldehyde, a particularly dangerous compound for human health, to verify if they were suitable to detect it at low concentrations. In the
Figure 11, the responses at 10, 5, 2, and 1 ppm of formaldehyde at the operating temperature of 500 °C for the selected ZnO sensors are reported. The result was very attractive, so appearing as possible the formaldehyde sub-ppm detection.
In addition to the magnitude of the gas response, in gas sensor field, response and recovery times are also important parameters. With reference to the
Figure 8, defining the response time as the time necessary to reach 90% of the response value and similarly for the recovery time, the values calculated for bisphenoids sensors resulted 2.3 min for the response time and 4.4 for the recovery one. In the case of prisms, the times resulted as lower: 1 min for the response time and 2.2 for the recovery time. This is reasonable due to the lower number of contacts between the particles, and further, in the case of prisms, constituted by well-defined and smoothed surfaces.
However, differently from what usually believed, the sensor time parameters are mainly driven by the geometry of the test chamber, in particular by the volume, rather than by the physical-chemical characteristics of the sensor itself. Indeed, the smaller the test chamber volume, the shorter the response and recovery times [
66].