2.1. Chemico-Physical Characterization

The fabrication process of the target materials is illustrated in Figure 1. Particular efforts were dedicated to elucidating the interplay between the adopted processing conditions and material chemical, physical and functional properties.

The system structure was investigated by XRD analyses (Figure 2). All the observed reflections located at 2θ = 18.0°, 28.9°, 31.0°, 32.3° and 36.1° could be indexed to the (101), (112), (200), (103) and (211) planes of tetragonal hausmannite (α-Mn₃O₄; a = 5.762 Å and c = 9.470 Å [31,38,59]). The occurrence of relatively weak and broad diffraction peaks suggested the formation of defective nanocrystallites [11], whose average dimensions were close to 25 nm for all the target specimens. A comparison of the signal relative intensities with those of the reference pattern [59] did not reveal any significant orientation/texturing effect, and no appreciable reflections from other Mn oxide polymorphs could be distinguished, highlighting the occurrence of phase-pure systems. Upon functionalization of Mn₃O₄ by RF-sputtering, no net variation in the recorded XRD patterns took place. The absence of noticeable diffraction peaks related to Ag or SnO₂ was traced back to their low content and high dispersion into the Mn₃O₄ systems [19,41,44,46] (see also XPS and SIMS results).

Figure 2. X-ray diffraction (XRD patterns for bare and functionalized Mn₃O₄ nanosystems. Vertical black bars correspond to α-Mn₃O₄ signals [59].

The surface chemical state of the developed materials was analyzed by means of XPS. Figure 3a displays the survey spectra for the target specimens, that revealed the presence of oxygen, manganese and eventually, silver or tin signals, for the functionalized systems, together with a minor carbon contribution (<10 at.%) resulting from adventitious contamination. The detection of manganese signals even after RF-sputtering suggested only a partial coverage of Mn₃O₄ by the deposited silver- and tin-containing species. Accordingly, Ag and Sn molar fractions were evaluated to be 47.0% and 31.0%, respectively. For bare Mn₃O₄, the Mn2p_3/2 component was located at BE = 641.8 eV (spin-orbit splitting (SOS) = 11.5 eV, Figure 3b), in accordance with previous literature data [31,35,38,41]. For the functionalized systems, a lower Mn2p_3/2 BE was observed (641.7 eV, for Mn₃O₄-Ag, and 641.5 eV, for Mn₃O₄-SnO₂). This finding suggested the formation of Schottky and p-n junctions for Mn₃O₄-Ag and Mn₃O₄-SnO₂, respectively [33,37,38,41,42], resulting in an Ag → Mn₃O₄ and SnO₂ → Mn₃O₄ electron transfer. This phenomenon, more pronounced for SnO₂-containing samples, as testified by the higher BE decrease, exerted a favorable influence on the resulting gas sensing performances. As regards silver (Figure 3c), the Ag3d_5/2 position (BE = 368.5 eV, SOS = 6.0 eV), as well as the pertaining Auger parameters (see also Figure S1; α₁ = 719.7 eV and α₂ = 725.6 eV), revealed a partial Ag surface oxidation, i.e., the coexistence of Ag(0) and Ag(I) oxide, as typically observed in similar cases [38,46,49,60]. Finally, the main tin photopeak (Figure 3d; BE(Sn3d_5/2) = 486.9 eV; SOS = 8.4 eV) was located at higher energies than those reported for SnO₂ [14,56,61], in line with the above mentioned charge transfer process. Taken together, these results highlighted the formation of nanocomposites in which the single components maintained their chemical identity, and enabled us to discard the formation of ternary phases, in line with XRD results. The deconvolution of O1s photopeaks (Figure S2) revealed the concurrence of two distinct bands at BE = 530.0 eV, resulting from lattice oxygen in Mn₃O₄, Ag(I) oxide (Mn₃O₄-Ag) or SnO₂ (Mn₃O₄-SnO₂) [14,31,38,60,61], and 531.6 eV, assigned to oxygen species adsorbed on surface O defects [4,41,51,52,55]. The contribution of the latter component to the overall O1s signal increased from ≈36.0%, for bare Mn₃O₄, to ≈58.0%, for the functionalized specimens, indicating a parallel increase of the oxygen defect content. The latter feature had a direct beneficial impact on the resulting gas sensing behavior.

Figure 3. (a) X-ray photoelectron spectroscopy (XPS) wide-scan spectra pertaining to bare Mn₃O₄, Mn₃O₄-Ag and Mn₃O₄-SnO₂ samples. (b) Mn2p, (c) Ag3d and (d) Sn3d photoelectron peaks. The color code is reported in panel (a).

Complementary information on material chemical composition was obtained by SIMS in-depth profiling (Figure 4). Upon functionalization of Mn₃O₄, no significant variations in the overall deposit thickness took place (for all specimens, the average value was (400 ± 50) nm, as determined by cross-sectional FE-SEM analyses (see below and Figure 5)). The almost parallel trends of manganese and oxygen ionic yields suggested their common chemical origin, in line with the formation of phase-pure Mn₃O₄. Silver and tin trends could be described by an erfchian profile, such as in thermal diffusion processes [49]. For the Mn₃O₄-SnO₂ sample, Sn yield underwent a progressive decrease throughout the outer 100 nm, subsequently followed by a plateau, whereas, for the Mn₃O₄-Ag specimen, the silver curve continuously declined even at higher depth values. In spite of these differences, a penetration of both Ag and Sn up to the interface with the alumina substrate was observed, and ascribed to the synergistical combination between the inherent RF-sputtering infiltration power and the Mn₃O₄ deposit open morphology [38,41,44,48] (see also Figure 5). This intimate contact between the system components is indeed an issue of key importance in order to benefit from their mutual electronic interplay, as discussed in detail below.

Figure 4. Secondary ion mass spectrometry (SIMS) depth profiles for Mn₃O₄, Mn₃O₄-Ag and Mn₃O₄-SnO₂ specimens.

Figure 5. Representative plane-view and cross-sectional field emission-scanning electron microscopy (FE-SEM) micrographs (left panels) and atomic force microscopy (AFM) images (right panels) for Mn₃O₄, Mn₃O₄-Ag and Mn₃O₄-SnO₂ samples.

The system morphology was investigated by the complementary use of FE-SEM and AFM. FE-SEM micrographs (see Figure 5, left side) highlighted that bare Mn₃O₄ was characterized by the presence of elongated nanoaggregates (mean size = 100 nm), whose interconnection resulted in the formation of arrays with an open morphology. This feature is indeed favorable in view of gas sensing applications, since a higher area available for the interaction with the surrounding gases has a beneficial effect on the ultimate material functional performances [15,19,25,31,40]. After Ag and SnO₂ introduction, no marked variations involving aggregate coalescence/collapse could be observed, validating the potential of the adopted synthetic route in functionalizing Mn₃O₄ nano-deposits without any undesired morphological alteration. AFM analyses (Figure 5, right side) confirmed the presence of the aforementioned aggregates uniformly protruding from the growth substrate, resulting in a crack-free and homogeneous granular topography, yielding an average RMS surface roughness of 40 nm for all the analyzed specimens.

2.2. Gas Sensing Performances

Figure 6 displays representative dynamical responses of the developed sensors towards square concentration pulses of gaseous hydrogen. All the target materials exhibited a p-type sensing behavior, as indicated by the resistance increase upon H₂ exposure due to the reaction of the analyte with adsorbed oxygen species, resulting in a decrease of the major p-type carrier concentration [2,16,31,35,39]. This phenomenon is in agreement with the fact that Mn₃O₄ is the main system component, as indicated by structural and compositional characterization [44].

Figure 6. Dynamical responses of Mn₃O₄ (a), Mn₃O₄-Ag (b) and Mn₃O₄-SnO₂ (c) nanosystems vs. different H₂ concentrations, at a fixed working temperature of 200 °C.

Remarkably, data in Figure 6 evidence that the on-top deposition of Ag and SnO₂ was an effective mean to increase the electrical property modulation upon H₂ exposure with respect to pure Mn₃O₄. For both composite systems, the measured resistance underwent a relatively sharp rise upon hydrogen exposure, and a subsequent slower increase up to the end of each gas pulse. This phenomenon suggested that the rate-limiting step in the resistance change was the chemisorption of molecular hydrogen on the sensor surface [15,31,41,48]. In spite of an incomplete baseline recovery after switching off hydrogen pulses, the measured resistance variations were almost proportional to the used hydrogen concentrations, enabling us to rule out significant saturation effects, an important starting point for eventual practical applications [15,38,39].

To account for the performance increase yielded by composite systems, it is necessary to consider the mechanism of hydrogen detection by the target p-type materials, which can be described as follows. Upon air exposure prior to contact with the target analyte, oxygen molecules undergo chemisorption processes, yielding the formation of various species [6,9,29,36,43,55], among which O⁻ is the prevailing one in the present working temperature interval [14,17,30,33]:

Figure 7. Schematics of the hydrogen gas sensing mechanism and corresponding hole accumulation layer (HAL) modulation for nanosystems based on: (a) bare Mn₃O₄; (b) functionalized Mn₃O₄. The dashed black and red lines indicate the HAL boundaries in air in case of bare and functionalized Mn₃O₄, respectively. Red spheres, yellow areas, and orange ovals indicate adsorbed oxygen, HAL thickness, and functionalizing agent (Ag or SnO₂), respectively. Blue arrows indicate electron flow due to H₂ oxidation (see reaction 4).

A process which decreases the hole concentration and the HAL thickness, resulting, in turn, in an increase of the measured resistance [5,16,17,18,42,43]. Finally, upon switching off the gas pulse, the sensor surface is again in contact with air, and the original situation is restored, with a recovery of the pristine HAL width [30,31,41].

Since all the investigated systems have almost identical mean crystallite dimensions, grain size and RMS roughness values, a significant influence of these parameters on the different gas sensing performances can be reasonably ruled out. Indeed, the enhanced responses of composite sensors with respect to the pristine Mn₃O₄ can be first explained in terms of electronic effects occurring at the interface between Mn₃O₄ and the functionalizing agents, a key aspect to be considered for a deep understanding of gas sensing phenomena [26].

For Mn₃O₄-Ag sensors, these processes result from the formation of Mn₃O₄/Ag Schottky junctions, whose occurrence produces a Ag → Mn₃O₄ electron flow [38,63] (see also the above XPS data), and a consequent thinning of the HAL width in comparison with bare Mn₃O₄ (compare Figure 7a,b). As a consequence, HAL variations upon contact of the sensor with gaseous H₂ produce higher responses by increasing the registered resistance modulations [38,44]. An analogous phenomenon occurs for Mn₃O₄-SnO₂ systems (Figure 7b; HAL thickness = 12.4 nm, see the Supporting Information), although in this case the SnO₂ → Mn₃O₄ electron flow is triggered by a different phenomenon, i.e., the presence of p-n Mn₃O₄/SnO₂ junctions [32,33,37,41,42,43,55].

The latter effect can, in principle, result in enhanced variations of the HAL extension with respect to the case of Mn₃O₄-Ag sensors, since the occurrence of a partial silver oxidation (as evidenced by XPS analysis, see above) precludes a full exploitation of electron transfer effects resulting from the establishment of Mn₃O₄/Ag Schottky junctions [38,46].

Nonetheless, the enhanced hydrogen detection efficiency of Mn₃O₄-based nanocomposites with respect to bare manganese oxide is likely due to the concurrence of additional cooperative phenomena. For both Mn₃O₄-Ag and Mn₃O₄-SnO₂ systems, the higher content of oxygen defects at the composite surface with respect to bare Mn₃O₄ (see the above XPS data and Figure S2), as well as the exposure of a high density of heterointerfaces, can in fact supply active sites for a more efficient chemisorption of both oxygen and analyte molecules, which, in turn, boosts the resulting gas responses [4,8,16,17,18,30,43]. In addition, the intimate component contact enabled by the adopted preparation route, yielding a good intergranular coupling, enables a proficient exploitation of their chemical interplay [38,44,48], related to the synergistical combination of materials with different catalytic activities [10,27,37,41,47]. Hence, the improved sensing performances of functionalized Mn₃O₄ systems can be related to the concomitance of electronic and catalytic effects.

Taken together, the above observations can account for the improved performances at lower working temperatures of SnO₂-containing systems with respect to Ag-containing ones. This result is exemplified by an inspection of Figure 8, showing that, apart from the appreciable response enhancement, deposition of Ag and SnO₂ onto Mn₃O₄ resulted in different trends as a function of the operating temperature. In the case of Mn₃O₄-Ag sensors, the progressive response rise indicated an enhanced extent of reaction (4) upon increasing the thermal energy supply [2,8,15,31,38,39]. Conversely, as concerns Mn₃O₄-SnO₂, a maximum-like response behavior was observed, the best operating temperature being 200 °C. Such a response trend, in line with previous reports regarding H₂ detection by other metal oxides [7,9,17,19,29,43,64], suggested the occurrence of a steady equilibrium between hydrogen adsorption and desorption at 200 °C, whereas an increase of the working temperature resulted in a predominant analyte desorption [27,33,55,64,65]. The lower value of the optimum operating temperature for Mn₃O₄-SnO₂ in comparison to Mn₃O₄-Ag is in line with the more efficient SnO₂ → Mn₃O₄ electron transfer (see the above XPS data).

Figure 8. Gas responses to a fixed H₂ concentration (200 ppm) for the target Mn₃O₄-based sensors at different working temperatures.

The potential of the present results is highlighted by the fact that the best H₂ responses obtained for Mn₃O₄-Ag (at 300 °C) and Mn₃O₄-SnO₂ (at 200 °C) were higher than those reported for sensors based not only on Mn₃O₄ [35,42], but even on other p-type oxides, including CuO [3,29], NiO [2,30], BiFeO₃ [7], Co₃O₄ [16,19], Ni_xCo_3-xO₄ [16], as well as nanocomposites based on CuO-Pt [4] and CuO-WO₃ [18]. In addition, the same responses compared favorably with those pertaining to various MnO₂-based nanomaterials/thin films in the detection of the same analyte [1,64,65]. A comparison of selected representative data is reported in Table S1. It is also worthwhile highlighting that the optimal operating temperature for H₂ detection by the present materials (200 °C for Mn₃O₄-SnO₂ systems) was lower than the ones reported for Mn₃O₄ [35], MnO₂ [8], CuO [3,17,29], Co₃O₄ [16], NiO [9], NiO-ZnO [43], Ni_xCo_3-xO₄ [16], BiFeO₃ [7] and CuO-WO₃ sensors [18]. This result is of importance, not only to avoid dangerous temperature-triggered explosions, but also to implement sensing devices with a higher service life and a lower power consumption [11,25,38,41].

Gas responses were also analyzed as a function of H₂ concentration (Figure S3). The obtained linear trends in the log-log scale confirmed the absence of appreciable saturation phenomena, an important prerequisite for a quantitative analyte detection [37,38,39,44,48]. The best detection limits obtained by fitting experimental data with equation (2) ((18 ± 1) ppm and (11 ± 1) ppm for Mn₃O₄-Ag and Mn₃O₄-SnO₂ sensors, respectively) were close to those previously reported for MnO₂ [8], CoO [6] and CuO-TiO₂-Au [44] sensors, and inferior than those pertaining to ZnO ones [15]. It is also worth noticing that these values were nearly three orders of magnitude lower than the H₂ lower explosion limit (LEL, 40000 ppm) [2,11,12,23,43], highlighting thus the detection efficiency of the present systems.

Beyond sensitivity, selectivity is an important parameter for the eventual utilization of gas sensing devices [6,11,19,27,30]. The responses towards a specific test gas are in fact required to be higher than those of other potential interferents, in order to avoid false alarms in real-time gas monitoring equipment [36,38,39]. In particular, the choice of CH₄ and CO₂ as potential interferents in real-time hydrogen leak detection is motivated by the fact that: (i) hydrogen and methane are common reducing gases, either stored, or used together [1,13]; (ii) the presence of carbon dioxide may hamper hydrogen recognition [5,21,43] in the case of fuel cells eliminating CO₂ and producing electricity and H₂ [66]. As shown in Figure S4, the present sensors yielded no responses towards CO₂, and only weak signals upon exposure to CH₄. In the latter case, the selectivity was estimated as the ratio between the responses to H₂ and CH₄ [32,41], yielding values of 22 and 24 for Mn₃O₄-Ag and Mn₃O₄-SnO₂ at the best working temperatures (300 and 200 °C, respectively). Though preliminary in nature, the latter results are an attractive starting point for further studies aimed at implementing exclusive H₂ sensors, which are highly required for practical applications [13].