1. Introduction
In recent years, the detection of ultratraces of nitroaromatic compounds (NACs), such as 2,4,6-trinitrotoluene (TNT), has gained considerable attention due to environmental, anti-terrorist security, and health-related problems associated with NACs [
1]. In fact, NACs are toxic compounds able to cause severe health concerns in both animals and humans, leading to anemia, abnormal liver function, cataract development, and skin irritation [
2,
3].
The principle of NACs detection is simple since any explosive emits a rather small, but detectable number of molecules [
4]. Based on the concentration in air at the equilibrium at a certain temperature, explosives can be divided into three groups: high, medium, and low vapor pressure, as illustrated in
Figure 1 [
5].
Triacetone triperoxide (TATP), ethylene glycol dinitrate (EGDN), 2,3-dimethyl-2,3-dinitrobutane (DMNB, a detection taggant for explosives), nitroglycerin (NG), and 2,4-dinitrotoluene (DNT) present a high vapor pressure. These explosives have equilibrium vapor concentrations in air in the ratio of about one part per million (1 ppm) [
5,
6,
7]. Medium vapor pressure explosives have equilibrium vapor concentrations in air near one part per billion (1 ppb), while low vapor pressure explosives have equilibrium vapor concentrations of one part per trillion (1 ppt). The medium vapor pressure group includes TNT (2,4,6-trinitrotoluene) and ammonium nitrate (NH
4NO
3) [
5,
6,
7]. The low vapor pressure group includes HMX (octogen), RDX (Research Department Explosive, hexogen or cyclonite), and PETN (pentaerythritoltetranitrate) [
5,
6,
7]. RDX is 1000 times less volatile than 2,4,6-trinitrotoluene (TNT), corresponding to a parts-per-trillion vapor pressure under ambient conditions. Thus, compared to other explosives, trace detection of RDX is highly challenging [
8]. The vapor pressure value concerns pure materials; for mixtures containing these explosives, the vapor pressure may be even lower [
5].
Apart from nitro-based explosives, usually indicated as military explosives, homemade explosives, also referred to as improvised explosives (IEs), are produced widely by many terrorist groups throughout the world. Ammonium nitrate-fuel oil (ANFO) and urea nitrate (UNi) are two main kinds of IEs [
9]. UNi can be transformed into nitro-urea (NH
2CONH–NO
2), a more powerful explosive, through a dehydration process [
9]. The detection of IEs (ANFO, UNi, and nitro-urea) can be achieved by the indirect detection of their raw materials, for example, AN and urea [
9].
Numerous detection techniques have been developed throughout the years, but their common limitations are rather large sizes and weights, high power consumption, unreliable detection with false alarms, insufficient sensitivity and/or chemical selectivity, and hyper-sensitivity to mechanical influences associated with very high price [
1]. Thus, there is a strong need of cheap, rapid, sensitive, and simple analytical methods for the detection and monitoring of these explosives in air.
Table 1 illustrates that explosives’ melting points are generally below 200 °C, while boiling points are in the range 150–300 °C [
10].
Table 1 also indicates that boiling of explosives often leads to molecular decomposition. Once vaporized, the solids decompose into individual molecules, which then further decompose into smaller fragments. Experimentally, it was found that all explosives produce a cloud of reaction products made of nitrogen dioxide (NO
2), nitric oxide (NO), ammonia (NH
3), hydrogen (H
2), carbon monoxide (CO), carbon dioxide (CO
2), nitrogen (N
2), oxygen (O
2) and methane (CH
4) [
10]. Among those, NO
2 is the compound that is most easily detected at low concentrations and with relative selectivity using solid state gas sensors [
10] because of its high electron affinity. This feature allows it to be set aside from most other gases in the ambient air (
Table 2) and enables it to extract in a very efficient way conduction electrons from semiconductor materials [
10].
The electron affinity (Ea) of an element is the energy liberated when an atom in the gas phase captures an electron. If electron affinity is negative, this means that an external energy is necessary to the atom to capture an electron [
11]. The most abundant molecule in air having a positive electron affinity is oxygen (Ea of about 0.5 eV); thus, if free electrons are available, ions
form spontaneously (Equation (1)) in clean ambient air:
In the case air is contaminated with NO
2 molecules (Ea of about 2.3 eV) or O
3 (Ea of about 2.1 eV), electrons initially captured by O
2 molecules are spontaneously transferred (Equations (2) and (3)):
As most explosives either contain
or
side groups (
Figure 2), these materials also present high electron affinities (
Table 1). Once vaporized, explosives molecules can therefore easily take up negative charge from
ions, thus forming negatively-charged analyte ions, which can then be easily analyzed with regard to their ion mass (this process is extensively used in the IMS (Ion Mobility Spectroscopy) detection of explosives, drugs, and chemical weapons in airports [
10].
Semiconductor metal oxides (SMOs) are able to sense gas based on the variation of their electrical properties. SMOs can be divided into two groups according to the operating temperature which dictates the mechanism by which these materials work: the first group is made by materials which follow surface conductance effects (below 600 °C), while the second one is constituted by materials which follow bulk conductance effects (above 700 °C) [
8]. At lower temperature, the bulk defect effect is slow and conductance change is due to the formation and removal of surface-adsorbed oxygen species. The bulk conductance materials respond to the changes in oxygen partial pressure in the upper temperature range (>700 °C) and show the equilibrium between atmosphere and bulk stoichiometry [
12].
The gas-sensing mechanism of SMO sensors involves two major functions: receptor and transducer. The receptor function involves recognition of a target molecule in gas—solid interface, which leads to an electronic change in the surface of the SMOs [
13,
14]. Chemical properties of the surface oxygen of the SMO itself are responsible for the receptor function in a sensor [
14], which can be modified when the oxide surface is loaded with noble metal or acidic or basic oxide particles. Then, change in receptor function induces a large change in sensitivity. Transducer function is responsible for converting the signal generated by chemical interaction on the oxide surface (change in work function) into an electrical signal. This function is governed by each boundary or neck between grains.
Conductometric (resistive) gas sensors have a very simple construction: they are made of two electrodes, a gas-sensitive layer, a dielectric substrate, and a heater to ensure the working temperature in the active region [
15,
16]. As mentioned above, the operating principle of such sensors is based on the conductivity change of semiconductor materials under the influence of the gas-environment [
15].
Figure 3 depicts receptor and transducer functions as well as physicochemical and material properties involved for a
n-type semiconductor gas sensor. On the surface of the grains, the adsorbed oxygen molecules extract electrons from the conduction band and trap the electrons at the surface as ions, which leads to a band bending. Thus, an electron-depleted layer is formed (also called space-charge layer). The space-charge region, being depleted of electrons, is more resistive than the bulk of the SMO. When the particle size (D) of the sensing film is close to or less than twice the thickness of the space-charge layer (L), the sensitivity of the sensor is increased [
16,
17,
18,
19,
20]. The control of L by impurity doping can give rise to a great change of the sensor response (SR) even when D is the same: Al doping of SnO
2 increases SR, while Sb doping of SnO
2 decreases SR, respect to pristine SnO
2 [
20].
The utility factor shown in
Figure 3 concerns the attenuation of the response due to the diffusion and reaction of reactive target gases through the pores of the sensing film [
19]. Experiments and simulations have shown that in order to achieve high sensitivity and fast response, SMOs should present: high gas permeability (large pore size), large surface area, small grain size, low degree of agglomeration, and optimal thickness [
15].
It is now accepted that below 150 °C, oxygen is adsorbed in ionic form (ionosorbed) as
and in the temperature range from 150 °C to 400 °C (usually, the operating temperature of SMOs gas sensors), it dissociates as
. Above 400 °C, formation of
occurs [
14,
16,
17,
20,
21]. The required electrons are extracted from SMO’sconduction band and trapped at the surface, leading to an electron-depleted layer, as already explained. The presence of negative surface charge leads to band-bending which generates a potential barrier at the surface of the grains (
Figure 3). The height and depth of these barriers depend on the amount of charge on the surface and is correlated to the quantity and type of adsorbed oxygen ions. Neutral oxygen species such as physisorbed oxygen are assumed not to play any role in gas-sensing as well as lattice oxygen ions [
16,
17,
21,
22].
In the case of a
n-type semiconductor, the resistance of the gas sensor decreases when in contact with reducing gases or vapors because surface oxygen species are consumed to oxidize the target gas and release electrons in the conduction band. On the contrary, for a
p-type SMO, the conductivity decreases in presence of a reducing gas. The oxygen adsorption, the formation of electrical core–shell structures (due to holes accumulation layer at the outer surface of grains and the insulating core formed after oxygen ions adsorption), the conduction mechanisms, the catalytic activities, and the interactions with humidity of
p-type oxide semiconductors are far different from those of
n-type oxide semiconductors. In
p-type SMOs, conduction occurs preferentially along the hole-accumulation layer near the surface where oxygen anions are adsorbed [
23]. Then, the concentration of holes in this hole-accumulation layer decreases when
p-type oxide semiconductor gas sensors are exposed to reducing gases [
23]. Generally, the gas response of
p-type-oxide-semiconductor-based gas sensors (i.e., the relative change in sensor conductance/resistance when a sensor is exposed to a gas) is equal to the square root of that of
n-type oxide semiconductor-based gas sensors whose morphologies are identical to the
p-type-oxide-semiconductor-based gas sensors [
24]. However, the responses of
p-type oxide semiconductor gas sensors can be increased by using aliovalent materials (i.e., electronic sensitization), by catalytically promoting gas detection (i.e., chemical sensitization), and by decreasing the size of the nanostructure so that
p-type oxide semiconductor gas sensors can be used for practical applications. Finally, the selectivity of
p-type oxide semiconductor gas sensors can be further tuned by doping or loading of catalysts that can promote the reaction between the sensor and a specific gas [
23].
In 1952, Brattain and Bardeen demonstrated the change of the semiconducting properties of germanium with a variation of the partial pressure of ozone or peroxide vapors in the surrounding atmosphere [
25]. Later, Seiyama demonstrated the gas sensing effect on metal-oxides [
26], and Taguchi brought metal-oxide semiconductor gas sensors to market in 1971, after founding in Osaka, in 1969, the company Figaro Engineering Inc., which is still today the largest manufacturer of semiconductor gas sensors world-wide.
The gas sensors market is expected to be valued at
$1297.6 million USD by 2023, growing at a compound annual growth rate (CAGR) of 6.83% between 2017 and 2023 [
27]. Actually, semiconductor oxide materials contribute more than 20% to market [
28].
Commercial sensors should meet the following requirements:
High response to the target agent,
High selectivity to target gas in the presence of a mixture of gases,
Fast and reversible interaction with analyte,
Low sensitivity of the signal to a change in air humidity,
Absence of long-term drift,
Short time to operational status,
Effective low-cost technology,
High reproducibility,
Uniform and strong binding to the surface of the substrate.
The first SMO gas sensors were based on ZnO and SnO
2. Today, many different oxides have been investigated for manufacturing conductometric gas sensors (
Table 3) [
15,
26].
It was experimentally verified, and it became a rule, that the higher the operating temperature is, the larger should be the band gap (E
g): the optimal band gap must be larger than 2.5 eV for solid-state gas sensors working above 300 °C. This requirement is satisfied by most SMOs. On the contrary, for sensors working at ambient temperature, E
g can be considerably smaller, and this may be even an advantage [
30]. However, it has to be underlined that the ability to operate at higher temperatures is an important advantage of solid-state gas sensors, because this considerably reduces the influence of water vapor on gas-sensing characteristics: the lower the operating temperature is, the greater is the sensitivity of the sensors to air humidity [
31].
As the sensing material in conductometric gas sensors should be relatively conductive, the concentration of point defects in the SMOs should be rather high: it was experimentally demonstrated that the optimum lies in the range of 10
17–10
19 cm
−3, which corresponds to a conductivity of 10
−2–10
1 Sm/cm. On the contrary, a too high concentration of point defects and, as a consequence, a high electroconductivity or low resistivity, reduces the influence of the surface on the bulk concentration of charge carriers in the grains and hinders the effects on the surface. A too low concentration of free charge carriers (
n < 10
16 cm
−3, i.e., σ < 10
−4–10
−5 Sm/cm) is obviously also not acceptable [
15].
As previously illustrated, conductometric sensors are based on processes occurring on the surface of the metal oxide and relating to adsorption–desorption phenomena and catalytic reactions. Unfortunately, in many cases, it is not known which crystallographic plane is optimal for gas-sensitive effects: we only know that these properties may be controlled through a change of the surface stoichiometry of the metal oxides by using various processing methods, or via the surface modification of metal oxides by catalytically active additives, such as noble or transition metals. However, the choice of optimal working conditions is empirical in nature, and often, the conditions found during such optimization are specific for the target gas and the technology selected for the synthesis and modification of the SMO [
15]. However, studies have shown that the crystallographic structure of the metal oxides does not play a fundamental role: SMOs used in gas-sensing applications may have a cubic, hexagonal, orthorhombic, or tetragonal structure [
15].
SMO gas sensors are robust and inexpensive; however, they currently lack sensitivity and selectivity to detect low vapor pressure explosives [
32]. Introduction of additives into base metal oxides can change their features, such as, for example, the concentration of charge carriers, chemical and physical properties of the metal oxide matrix, electronic and physical–chemical properties of the surface (energetic spectra of surface states, energy of adsorption and desorption, sticking coefficients, etc.), surface potential and intercrystallite barriers, phase composition, sizes of crystallites, and others also [
33,
34]. Heterojunctions can increase SMOs’ sensitivity and are based on two different metal oxides admixed or layered together (
p-
n,
n-
n and
p-
p diodes made from a
p- and a
n-type semiconductor) [
32,
35,
36].
The effect of the addition of metallic particles on the gas-sensing properties of SMOs has been widely studied for decades [
33,
37]. The influence of metallic particles addition has been classified as chemical or electronic according to two basic sensitization mechanisms. In the electronic mechanism, the reaction with the gas molecules takes place on the surface of the introduced clusters and not in the SMO. These clusters change their charge state, leading to a variation of the surface barrier height and producing a conductance change in the metal oxide. In this case, the base semiconductor has only a transducer role of the changes induced in the metallic particles by their interaction with the target gas. In the chemical mechanism, the metal oxide itself acts as a chemical catalyst: the role of the additive is to increase the reaction rate of the gas molecules, which are firstly adsorbed on the metallic cluster and, later, moved to the oxide surface. This process is the so-called spill-over process. Among metallic surface additives, silver is typically considered to be related with an electronic mechanism, whereas palladium, gold, and platinum are expected to lead to the chemical one [
18,
33,
37,
38]. There is an optimum concentration for catalytically active modifiers at which their effect is maximal: in the case of nanocrystalline SnO
2 modified with Pd, Pt, and Au nanoparticles, this concentration is 0.2 to 3 wt % [
39]. Upon exceeding the optimum impurity concentration, the effect decreases for various reasons: the catalytic conversion of the targeted gas on the modifier clusters without involving the semiconductor matrix, the decreasing portion of the semiconductor oxide-free surface and concentration of the adsorbed oxygen species on it, or the percolation of the noble metal clusters [
39].
Among the different sensors features that can influence its response, it is now recognized that the lower the crystallite size is, the higher the response is [
40]. On the contrary, the response time (the time needed by a sensor to achieve 90% of the total impedance change in the case of gas adsorption) and the recovery time (the time necessary to reach 90% of the total impedance variation in the case of gas desorption) are shortened by the decrease of the crystallite size [
40]. However, it was also found that nanostructured films having a smaller crystallite size presented a stronger sensitivity to the air humidity of the surrounding atmosphere. This behavior was explained because the change of the shape from crystallites (t > 10 nm) to spherulites (t < 7–10 nm) is associated with an increased number of structural and electronic defects. The presence of a huge number of atomic steps and corners, typical of spherulite-type grains, could also be a reason favoring the interaction with water vapor [
40]. Therefore, if the minimum sensitivity to air humidity and maximum stability of sensor parameters are required, the size of crystallites will be the result of a compromise between these two features [
40]. In SMOs, formed by using thick-film technology (like, for example, screen-printing technique), the pore size is strongly correlated with the crystallite size. Then, the larger the crystallite size is, the larger is the pore size. This means that the requirement to increase the pore size for better gas permeability of the thick-film is in contradiction with the need to reduce the crystallite size for achieving a higher sensitivity [
40].
Finally, a considerable contribution to the sensor signal is made by the acidic/basic and redox reactions that occur on the surface of the sensing material when interacting with the gas phase. The specificity or selectivity of these reactions is mainly determined by the nature of the available active sites on the surface [
41]. Hydroxy groups, chemisorbed oxygen, and coordinatively unsaturated atoms can act as active sites on the oxide surface. The oxidation ability of the oxides is attributed to chemisorbed oxygen and variable valence metal atoms. In the second case, the efficiency of the charge transfer from the matrix to an active cation plays a great role too [
41]. Experiment has shown that the use of metal oxide nanocomposites also allows significant increases in the stability of the grain size during annealing, limiting grain growth. Grain growth due to high working temperatures is one of the reasons for the instability of gas sensor parameters during their service life [
14].
Thus, this paper aims at shortly reviewing the most recent SMOs nanocomposites able to sense explosives.