High-Performance Room-Temperature Conductometric Gas Sensors: Materials and Strategies

Abstract

Chemiresistive sensors have gained increasing interest in recent years due to the necessity of low-cost, effective, high-performance gas sensors to detect volatile organic compounds (VOC) and other harmful pollutants. While most of the gas sensing technologies rely on the use of high operation temperatures, which increase usage cost and decrease efficiency due to high power consumption, a particular subset of gas sensors can operate at room temperature (RT). Current approaches are aimed at the development of high-sensitivity and multiple-selectivity room-temperature sensors, where substantial research efforts have been conducted. However, fewer studies presents the specific mechanism of action on why those particular materials can work at room temperature and how to both enhance and optimize their RT performance. Herein, we present strategies to achieve RT gas sensing for various materials, such as metals and metal oxides (MOs), as well as some of the most promising candidates, such as polymers and hybrid composites. Finally, the future promising outlook on this technology is discussed.

1. Introduction

Under the current global ecological crisis, there is an increasing need to take actions directed to detect, control, and reduce the emission of contaminants coming from a wide variety of sources and in diverse forms. Pollution monitoring is at the basis of any control action, and it is essential to minimize its effect on human health and the environment. Tackling this challenging objective requires the development of a battery of different sensors, each designed to work on a specific environment and target pollutant. In particular, air quality surveillance has become more and more demanding as more stringent limits have been imposed, not only for confined indoor areas, such as industrial or laboratory facilities, but also in ample outdoor spaces, such as urban environments or the countryside, where air pollution has been recognized as an important human health issue, at the same level as unhealthy diet or tobacco smoking [1,2]. Among the most common air pollutants, those posing higher risks for human health are SO₂, NO₂, O₃, CO, as well as volatile organic compounds (VOC), to name a few. Other compounds, such as NH₃, HCl, or H₂S, are also relevant in industrial or waste-treatment environments. Broadly speaking, the World Health Organization (WHO) recognizes two major gaps in the monitoring of air pollution levels, which limit the assessment of its impact on the environment and human health [1]. These gaps are the lack of sufficient monitoring posts in rural areas or outside major cities, and the extreme spatial variability occurring in the concentration of some pollutants (such as NO₂), which render the normal sensor distribution within the cities insufficient to track their real levels. Thus, having a dense enough network of gas sensors capable of monitoring the concentration of these compounds in real time becomes essential to develop strategies for air pollution control, including human health risk prevention, as well as to improve the existing models for pollutant diffusion and propagation [3,4]. Implementing such a network will require the use of a large number of gas sensors, which, therefore, must be, at the same time, reliable and highly sensitive, while presenting low production costs and power consumption. Conductometric (also called chemresistive or resistive) gas sensors meet all these requirements, adding some more advantages, such as simplicity in fabrication, high robustness, and very low volume, making them an excellent choice for this mission [5,6,7]. With the advent and maturity of new thin film technologies and, later on, of nanoparticle films, significant improvements in their performance and capabilities have been achieved, resulting in their widespread use from industrial manufacturing to security or environmental monitoring [5,7], with a global market of USD 2330M in 2020 and an expected growth between 2019 and 2025 of 7.8%, according to the report of Grand View Research [8]. This has also translated into a constant increase in the number of published works related to this topic since the mid-2000s (Figure 1).

The transducing mechanism of any conductometric gas sensor is based on the modulation of their electrical conductivity, regulated by changes in the composition of its surrounding atmosphere. Other relevant technologies in the field of gas sensing are classified by their transducing mechanism: optical sensors, chemical sensors, optochemical sensors, or electrochemical sensors [5,6,9]. Each of these technologies present intrinsic advantages in relation with conductometric sensors, but also specific drawbacks that limit their use in dense monitoring networks, such as high volumes or expensive manufacturing costs in the case of optical or optochemical sensors, lack of reusability in the case of chemical sensors, or stringent operation conditions and easy degradation in the case of electrochemical sensors [5]. A more comprehensive relation of the specific advantages and drawbacks of these technologies can be found in Refs. [5,6,9]. In the particular case of conductometric gas sensors, their major disadvantages are their low specificity compared to some of their counterparts, their tendency to present temporal drifts, requiring periodic calibrations, and their high operation temperature, which imply larger power consumptions [5,10]. These drawbacks limit their practical use and overcoming them has become one of the major goals in the field of conductometric gas sensors.

The vast majority of conductometric gas sensors are based on semiconducting metal oxides (MOs) [6,7,11,12,13]. The reasons for this are their tunable transport properties and the high sensitivity of their surface electronic properties to changes in the composition of the surrounding atmosphere [13], caused by their large stoichiometry variability (i.e., presence of oxygen or cationic vacancies) [14], their relatively large catalytic activity, and, in some cases, the presence of different cationic oxidation states. Despite the great success of MOs as conductometric gas sensors, their employment as active (i.e., sensing) material is weighted down by their usual high operation temperature (>150 °C), which is used to both improve their sensitivity and response/recovery times, and to mitigate the interference caused by their high sensitivity to humidity [13]. High temperature operation imposes a significant energy toll, limiting the number of devices that can be installed as well as the number of viable locations. Besides, higher operation temperature can also induce poor stability. Thus, in order to improve the energy efficiency of conductometric gas sensors, it is essential to find new strategies to reduce their operation temperature down to room temperature (RT). The aim of this work is to summarize the most recent advances on room temperature conductometric gas sensing for environmental monitoring, including the use of new materials beyond conventional MOs and the development of novel strategies to achieve effective RT operation. While several reviews have been published on the topic of conductometric gas sensors [6,13,15,16], most of them mainly focused on recent advances and novel materials for RT operation; herein, we also focus on the most common strategies to obtain RT operation and the physical and chemical explanation of these phenomena. Thus, for some materials, such as MOs, strategies, such as the optimization of their dimension and morphology, could enhance its RT operation. The use of materials, such as polymers, and the formation of inorganic/organic composites is also here reviewed, as it becomes one of the main approaches to reach RT operation.

2. Sensing Principle and Mechanisms

The general operation principle of any conductometric gas sensor is based on the modulation of its majority carrier density through interaction with the analyte, and can be regarded as a charge injection/extraction process caused by the interaction between the analyte and the sensor surface, which, in turn, modifies the overall conductivity of the sensor. There are several different mechanisms that can lead to this charge exchange process, including direct charge transfer between the analyte and the sensor surface (Figure 2A) [17,18], the catalytic decomposition of the analyte and its chemical reaction with other adsorbed species (Figure 2B) [19,20], redox reactions between the analyte and the sensor (Figure 2C) [21,22], or the competition for adsorption sites with other adsorbed species (Figure 2D) [23], among others. This charge exchange process may tune the overall conductivity of the sensor directly by varying the net concentration of available free carriers [13,17,18] (n for electrons, p for holes), according to:

$$\sigma =e\left({\mu }_{n}n+{\mu }_{p}p\right)$$

(1)

Here, µ is the mobility of the corresponding free carrier type, and e the elemental charge. This mechanism and its implication on the overall sensor performance were explored in detail by P. Moseley in Ref [13], who showed that in this case scenario, the response is strongly influenced by the material’s bulk donor density, N_D, with a linear dependence on the concentration of analyte for pure n- (high N_D) or p-type (low N_D) materials and a mixed behavior for intermediate N_D values (Figure 3). Most commonly, however, the conductance is affected by the formation of space charge regions in the form of depletion or accumulation layers due to surface band bending effects [24,25,26]. The impact of these space charge regions depends strongly on the microstructure of the sensor. For polycrystalline materials with grain sizes comparable to their Debye lengths, the appearance of depletion regions induces the formation of conduction channels (Figure 2F, the width of which is regulated by these charge injection/extraction processes [26]. Similarly, accumulation layers would act as preferential conduction paths (Figure 2G) with variable width [27]. For larger crystalline sizes, and in the presence of depletion regions, conduction becomes limited by the appearance of voltage barriers at the grain boundaries (Figure 2H). In this case, the injection/extraction of carriers modifies the height of the barrier, which determines the overall conduction through the grain boundary [28]. Conversely, the formation of accumulation layers at grain boundaries has the effect of providing low-resistivity conduction paths, similar to the small-crystalline-size scenario. This difference in the control of the conductivity for the same microstructure translates into different sensitivities for the same band bending, V_b (in absolute value), caused by the adsorption of the analyte, depending on whether it causes the depletion or the accumulation of free carriers, according to [27]:

$$S_a=\sqrt{S_d}$$

(2)

where S_a is the sensitivity for the case in which accumulation layers are formed, and S_d is the sensitivity for the formation of depletion layers. Thus, control on the carrier density and microstructure of the active material is crucial to optimize the sensor response.

A somewhat different process involves the formation of new chemical species by the reaction of the analyte with the sensor surface (Figure 2E). The new chemical species usually have different conductivity characteristics, which reflect on the overall conductivity of the sensor. This is the case, for instance, with highly active redox reactions in which the whole sensor surface changes its oxidation state, leading to the formation of a highly conducting metallic surface [29], but also has been observed in certain sulfuration reactions, such as those produced during H₂S detection by NiO sensors, in which the reversible NiS phase presents higher conductivities than its oxide counterpart [30]. It should be pointed out that other MOs have also shown great results detecting H₂S, such as CuO [31]. However, this mechanism is not very common, as chemical reactions must be fully reversible at operation conditions and recovery should proceed at a fast enough pace in order to ensure the proper functioning of the sensor.

Thus, each specific sensing mechanism results from the combination of a particular analyte–sensor interaction and its effect on the conductivity of the sensor. Usually, there is more than one sensing mechanism taking place at the same time during the conductometric response of any device [18,32]. For instance, Bartolomé et al. [18] showed the simultaneous operation of two different competing mechanisms during room temperature sensing of ethanol with p-type NiO, consisting on the catalytic decomposition of adsorbed ethanol on the NiO surface, and the direct charge transfer between both compounds, which lead to different responses, depending on the operation conditions. Which process or processes will eventually dominate the response would depend on the specific physical and chemical properties of the sample, as well as operation conditions (i.e., temperature, atmosphere, etc.).

In the case of MO-based sensors, the majority of these sensing mechanisms have been wrapped up in the so-called ionosorption model. According to this model, the overall charge exchange between the analyte and the sensor surface is governed by ionized adsorbed species. These species may be the analyte itself, but most commonly, the whole process is mediated by adsorbed oxygen ions, in the form of O₂⁻, O⁻, and O₂⁻, depending on the operation temperature [6], although at RT, the most relevant species is O₂⁻. Oxygen adsorbs on the surface of any MO sensor, trapping an electron from it and, depending on the operation temperature, further decomposing into O^• free radicals and/or trapping an additional electron. This creates an electron depletion layer on n-type MOs or a hole accumulation layer on p-type Mos, strongly impacting the conductivity of the sensor, through any part of the process described before. Reducing analytes usually react chemically with these ionosorbed oxygen species, which are then released in the form of reaction products, giving back the trapped electrons, while oxidizing analytes interact directly with the surface, trapping an electron from the sensor and, thus, becoming ionized [15,33]. A comprehensive mathematical description of these oxygen ionosorption processes, regarding equilibrium carrier concentration and sensor conductivity dependence on analyte concentration, was given by Barsan and Weimar in Ref [34] for the particular case of SnO₂, which may be extrapolated for other materials as well. The ionosorption (sometimes called oxygen ionosorption) model has been used to successfully describe a wide number of sensor–analyte systems [13,27]; however, this has led to the unfortunate generalization of the terms “n-/p-type sensor” and “oxidizing/reducing analyte” in relation only to the observed response, without any further consideration of the actual physicochemical mechanisms taking place. This oversimplification of the overall sensing process may induce some misconceptions regarding the expected behavior and performance of the sensors towards specific analytes, sometimes leading to unexpected and even contradictory results. However, these results are entirely the consequence of the coexistence of several different sensing mechanisms that, once they are properly understood, can be exploited to develop a myriad of different MO gas sensors [17,18,23,30].

It is noteworthy that while the majority of the sensors can be described in the frame of the aforementioned mechanisms, some particular systems have specific sensing mechanisms that are not included in this overview. This is the case of novel organic sensors, which are discussed in Section 4, humidity sensors based on thick or ceramic films that governed the Grotthuss mechanism of proton hopping [35], or 2D material (2DM) sensors (such as graphene, s transition metal dichalcogenides (TMDs), metal dichalcogenides (MD), graphene oxide (GO), or MXenes), also described in Section 4, which may involve more complex behaviors by virtue of their peculiar physical properties.

3. Characterization Techniques for Studying Conductometric Gas Sensors

Due to the intrinsic surface nature of the analyte–sensor interaction, the study of any conductometric gas-sensing mechanism normally requires the employment of surface and/or chemically sensitive characterization techniques. Usually, the basic characterization of the samples is performed ex situ, recording the electrical response to the analyte separately from the rest of the characterization techniques. Despite this procedure being able to provide plenty of useful information, it hinders the proper identification of the relevant sensing mechanisms involved in the response due to the lack of correlation between the conductivity of the samples immersed in different atmospheres and the rest of their physical properties. Nonetheless, in recent years, this tendency has been progressively reversing with the establishment of new in-situ/in-operando measurements, particularly for MOs but also for organic and 2DM-based sensors, in which physical changes in both the sample (sensor) and the analyte are monitored during the sensing process, employing chemically sensitive techniques, such as Fourier transform infrared (FTIR) spectroscopy, diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), Raman spectroscopy, or X-ray photoelectron spectroscopy (XPS) in near ambient pressure (NAP) conditions [21,32,36,37,38,39]. DRIFT/FTIR and Raman spectroscopies may provide relevant information on the adsorption processes and chemical decomposition of the analyte on the surface of the sensor. Besides, 2DMs are very prone to be doped by adsorbed gases [40,41], so in-operando Raman spectroscopy as well as photoluminescence (PL) spectroscopy are excellent probes to follow the processes occurring during the whole gas-sensing cycle, as both techniques may be used to characterize their doping and strain state, particularly for graphene [42] and transition metal dichalcogenides (TDM) [43,44,45]. A review on the use of Raman spectroscopy for in-situ and in-operando studies of different gas sensors can be found in Ref. [46]. Similarly, XPS is a very powerful characterization tool, as it provides information on the surface chemistry of the sample and its electronic surface states, which strongly determine the kind of interaction that can take place between the sample and the analyte. However, contrary to (DR)FTIR/Raman spectroscopy, XPS requires ultra-high vacuum (UHV) conditions, which hinder the study of the sensing processes at normal operation conditions. Fortunately, the recent development of NAP-XPS has helped to bridge the gap between its normal ultra-high vacuum operation and ambient pressure operation of gas sensors, allowing the observation of analyte–sensor interactions in more realistic conditions, close to their normal operation in ambient conditions [36,37,47]. One major drawback of XPS is the need to use high-intensity, high-energy X-ray beams, which tend to degrade sensitive samples, such as organic-based sensors, limiting its applicability on these systems. In these cases, the use of in-situ (DR)FTIR/Raman spectroscopy becomes essential to understand the processes involved in the sensing response. The combination of several of these in-operando tools for the study of specific analyte–sensor systems has shown great potential in understanding the dynamic evolution of complex systems during the sensing process. For instance, Sänze et al. [21] employed a combination of in-operando FTIR and Raman spectroscopy measurements to unravel the chemical surface modifications of the sensor and the formation of different byproducts during ethanol, ethene, and acetaldehyde gas sensing with In₂O₃ thin films, depending on the sensing temperature and environmental oxygen partial pressure (Figure 4).

4. Strategies for RT Operation

Room temperature operation is usually limited by the kind of sensing mechanism involved in the response and the sensitivity of the sample towards the presence of moisture. The dominant sensing mechanism determines both the sensitivity, S, of the sensor and its kinetics, i.e., the response and recovery times, τ_res and τ_rec. Sensor sensitivity is usually defined as either the ratio or the relative resistance difference between the resistance upon analyte exposure, R_g, and the base resistance R₀:

$$S=\frac{R_g}{R_0}(\%),\hspace{1em}\hspace{1em}\hspace{1em}S=\frac{R_g-R_0}{R_0}=\frac{\Delta R}{R_0}(\%)$$

(3)

When R_g < R_0, usually, the first definition is rewritten as the ratio between R₀ and R_g, in order to allow proper comparison between different response types, sometimes making unclear the actual direction of the resistance change, unless explicitly stated. Conversely, the second definition allows a direct comparison between positive and negative sensitivities without any expression reorganization, facilitating the discussion and also providing a more direct way to compare with the background noise in order to set the low detection limit of the sensor. Response and recovery times are usually defined as the time required to reach 90% of the final resistance value after either exposing the sensor to a stable concentration of the analyte or after completely withdrawing the analyte. According to this definition, it is necessary to let the signal stabilize after each change in analyte concentration in order to properly measure the value of τ_res and τ_rec. However, in many cases, large values of τ_res and τ_rec yield unsaturated response curves, which, in turn, lead to an underestimation of their value, hindering the assessment of the sensor performance. In these cases, extending the measuring time until the curves are fully saturated may not be an option, particularly in the presence of strong temporal drifts. A way to avoid this is to employ an alternative definition, assuming an exponential behavior and fitting the curves to a phenomenological expression [18]:

$$R=R_i+\mathsf{\Delta }R\left(1-e^{-t/\tau }\right)$$

(4)

where τ is either τ_res or τ_rec, R_i is, respectively, R₀ or R_g and ΔR is R₀ ± R_g. This expression allows a direct estimation of the response and recovery times, even if the curves are not fully saturated, although it gives somewhat lower τ values, compared to the 90% definition.

Low-temperature operation, down to RT, tends to reduce the speed at which reactions take place at the surface of the sensors, negatively impacting both the sensitivity and the response/recovery time of the devices. Moreover, lower response and drifts in the sensing performance can also occur at room temperature. However, as explained before, some sensing mechanisms do not rely on chemical reactions at all, such as the case of direct charge transfer. In those cases, reducing the temperature might actually be beneficial [18]. In other cases, the sensitivity of the sensors is so high that room temperature operation is perfectly feasible. Sometimes, the adsorption/desorption speed may become the actual bottleneck for the overall sensor performance, and, hence, a different (more efficient) source of energy, such as ultraviolet (UV) illumination, may be employed in order to enhance these processes. Reducing the impact of humidity interference without heating the sensor usually requires the selection of humidity-impervious materials. Thus, several different strategies have been explored so far in order to achieve room temperature operation with conductometric gas sensors, although very often, two or more of these strategies are combined in order to obtain the best possible performance. Some of the strategies and the previously described parameters alongside with recent results are shown in

Table 1. Summary of different reported strategies to achieve room temperature response for conductometric gas sensors.

Strategy	Type	Material	Structure	Gas	Concentration (ppm)	Sensitivity Equation	S	τres/τrec (s)	Ref.
Light activated		In2O3	NPs film	O3	10	$S=\frac{R_{Ozone}}{R_{UV}}$	105	>1/30	[24]
		TiO2	Fractal carbon + TiO2	Acetone	12.5	$S=\frac{R_0}{R_g}$	100	12/174	[48]
		ZnO		Acetone	0.1–1000	-	1–400	-	[49]
Specific sensing pathways		NiO	Ceramic	Ethanol	200–16,000	$S=\frac{\mathsf{\Delta }R}{R_0}$	2	30.6/86.8	[18]
		In2O3	NWs	H2S	20	$S=\frac{R_0}{R_g}$	141.1	-	[50]
		In2O3	NTs	H2S	20	$S=\frac{R_0}{R_g}$	166.6	-	[50]
Morphology optimization	0D	PbS	QDs	NO2	30	$S=\frac{R_n}{R_g}$	11.8	13 s/14 min	[51]
	1D	Ag	NW	NH3	1–2	-	5	-	[52]
	1D	In2O3	NW	NO2	0.02	-	25	-	[52]
	2D	SnS2	2D layers	NO2	8	$S=\frac{R_g}{R_0}$	10.8	164/236	[53]
Heterojunctions	2D/0D	rGO/CD	-	NO2			0.010–25	100/150	[54]
	2D/0D	SnS2/SnO2	-	NH3			100–500	200/300	[54]
	2D/3D	rGO/n-Si	-	NO2			250–1000	100/200	[54]
		In2O3/SnO2	Nanorods	NOX	0.1–100	$S=\frac{\mathsf{\Delta }R}{R_g}$	0.1–9	4.67–8.98	[55]
Conductive polymer		PANI	-	NH3	50	$S=\frac{R_0}{R_g}$	2.6	290/-	[56]
		PEDOT:PSS/EG	Thin film	ethanol	200	$S=\frac{\mathsf{\Delta }R}{R_0}$	0.2	-	[57,58]
		PPy	Thin film	NH3	4–80	$S=\frac{R_g}{R_0}$	1.12	20 s/15 min	[59]
		PTh	Thin film	NO2	10–100	$S=\frac{R_g}{R_0}$	1.33	220/1603	[59]
Hybrid composite		PEDOT:PSS/AuNps		CH4	0.02–1		8.6	22/43	[60]
		PANI/CeO2		NH3	50	$S=\frac{R_0}{R_g}$	6.5	57.6/-	[56]
		PEDOT:PSS/EG/SnO		Ethanol	200	$S=\frac{\mathsf{\Delta }R}{R_0}$	2.6	-	[57,58]
		PEDOT:PSS/EG/SnO2		Ethanol	200	$S=\frac{\mathsf{\Delta }R}{R_0}$	0.36	-	[58]
		PEDOT:PSS/EG/TiO2		Ethanol	200	$S=\frac{\mathsf{\Delta }R}{R_0}$	0.9	-	[57]

.

4.1. Light-Activated RT Operation

One of the most successful strategies directed to obtain RT gas sensors consists of the employment of light, typically in the range of UV to visible light, as an energy source to enhance some of the processes taking place during the sensing response [61,62]. There are currently several reviews available covering this particular topic from different perspectives [63,64,65,66,67], so here, we will try to summarize some of the most important aspects of this strategy. There are several ways by which UV and visible light can enhance the response of the sensors, but usually, they are all based on the generation of photocarriers that interact with the adsorbed analyte (Figure 5a–c). Depending on the role played by these photocarriers, two photoactivated enhancement mechanisms can be distinguished:

• Analyte adsorption/desorption enhancement: this approach is used on highly sensitive sensors, which show good response times at RT but slow recoveries due to the slow desorption rate of the analyte. Photogenerated carriers may rapidly recombine with any adsorbed ionic species, either ionosorbed oxygen/analyte molecules or any ionized product formed during the decomposition of the analyte, causing them to desorb as neutral species and speeding up the recovery process. A theoretical model of the kinetics of the photo-enhanced desorption of oxygen on MOs was developed by Melnick [68] with ZnO as a case study. An example of such photoactivated RT gas sensing was demonstrated for In₂O₃ thin films with UV back-illumination for ozone detection (Figure 5d) [28]. By periodically switching on and off the UV light, the authors managed to modulate the desorption speed of the decomposed O₃ molecules. The measured resistance is then dependent on the equilibrium between the O₃ adsorption rate, which depends on the concentration of O₃ molecules, and the desorption rate, which depends on UV light illumination, i.e., on the ON/OFF state (Figure 5e). The obtained response, measured as the resistance ratio between the OFF (R_O3) and the ON (R_UV) states for a given O₃ concentration, was found to vary linearly with O₃ concentration (Figure 5f).
• Analyte reaction enhancement: this approach can be employed to enhance the response in gas sensors based on the catalytic decomposition of the analyte [67,69]. Many sensing mechanisms are based on the catalytic decomposition of the analyte on the sensor surface; the obtained subproducts may then either react with ionosorbed oxygen species, releasing trapped electrons, or trap free carriers themselves. These processes usually require high temperatures as a source of energy to proceed at reasonable speeds. Photocatalytic materials use photon energy instead to speed up the chemical decomposition of the analyte and promote their sensitivity at RT. In this case, photogenerated carriers interact with the analyte, breaking chemical bonds and promoting either their oxidation with ionosorbed oxygen species or their chemical reaction with other adsorbed species, such as H₂O or other decomposed products [70]. Many MOs are known to have photocatalytic properties, such as TiO₂ [48], SnO₂ [20], or ZnO [49], but also organic polymers [71] or 2D materials [72,73].

Since sensor photoactivation requires the generation of photocarriers, the energy of the photons must be larger than the energy bandgap of the active material in order to promote electrons to form the valence band (VB) to the conduction band (CB), in the case of inorganic semiconducting sensors, or from the highest occupied molecular orbital (HOMO) to the lowest unoccupied molecular orbital (LUMO) in the case of organic sensors. Since many MOs have energy bandgaps well above 3 eV, within the UV region, the kind of illumination sources that can be used becomes very limited, not to mention their associated health risks. To overcome this problem, some strategies have been used, such as the employment of organic dyes [74] and quantum dot (QD) sensitizers [75], noble metal decoration for localized surface plasmon resonance (LSPR) light absorption enhancement [76], or employing narrow bandgap MOs as sensitizer or directly as the active material [77].

While this method shows good results, it still requires an additional energy source to power the light emitter, which must be integrated in the sensor, increasing its complexity, and humidity interference is not always avoided, even at high UV irradiation intensities [63].

4.2. Specific Sensing Pathways

Other approaches make use of specific sensing mechanisms with enhanced RT responses, such as direct charge transfer between the analyte and the active material or particular chemical reactions that are favored at low temperature [18,50]. For instance, Bartolomé et al. showed that an enhanced ethanol sensing performance could be achieved at RT on NiO sensors, provided that the ethanol decomposition sensing path is blocked and only direct charge transfer is allowed between adsorbed ethanol molecules and the NiO surface [18]. Weakly physisorbed ethanol acts as an efficient electron trap by virtue of its strongly electronegative OH group, increasing the conductivity of p-type NiO. Due to its weak bonding to the surface, physisorbed ethanol can efficiently desorb after it has been removed from the surrounding atmosphere, yielding to short response and recovery times (Figure 6a). Conversely, when ethanol undergoes a dissociative chemisorption, the obtained subproducts can react with ionosorbed oxygen species, releasing their trapped electrons and producing an opposite behavior, which diminishes the overall sensitivity and increases the recovery time of the sensor (Figure 6b). This mechanism is the dominant one at high temperature, but at room temperature, it can be hampered by carefully selecting the microstructure and crystalline orientation of the sensing layer. A different route is exploited by Xu et al. on electrospun In₂O₃ nanowire and nanotube gas sensors for H₂S detection [50]. In this case, In₂O₃ undergoes a sulfuration reaction after being exposed to H₂S, reversibly transforming between In₂O₃ and In₂S_3, following the reactions of Equations (5) and (6):

$${\mathrm{In}}_2{\mathrm{O}}_3\left(\mathrm{s}\right)+3{\mathrm{H}}_2\mathrm{S}\left(\mathrm{g}\right)\to { \mathrm{In}}_2{\mathrm{S}}_3\left(\mathrm{s}\right)+3{\mathrm{H}}_2\mathrm{O}\left(\mathrm{g}\right)$$

(5)

$${ \mathrm{In}}_2{\mathrm{S}}_3\left(\mathrm{s}\right)+\frac{9}{2}{\mathrm{O}}_2\left(\mathrm{g}\right) \to { \mathrm{In}}_2{\mathrm{O}}_3\left(\mathrm{s}\right)+3{\mathrm{SO}}_2\left(\mathrm{g}\right)$$

(6)

These reactions are spontaneous at room temperature, with a negative Gibbs free energy difference, ΔG, which increases with temperature. As the temperature increases, ΔG, also increases, approaching to 0 and, thus, making the transformation less and less favorable. Thus, at room temperature, the sulfuration reaction dominates the response, with high sensitivities and low response/recovery times, while at high temperatures, the reduced speed of sulfuration reaction allows H₂S to also react directly with ionosorbed oxygen species through a more conventional, but also less-sensitive sensing mechanism, thus, having a combination of both mechanisms in the overall response.

One of the major disadvantages of these strategies is that they are very sensor-analyte specific, and their systematization requires a profound knowledge of the existing sensing mechanisms, which is still lacking for most of the materials systems, although some progress has been made in this direction over the last few years.

4.3. Morphology Optimization (0D, 1D, 2D)

Another way to enhance the response of conductometric gas sensors at RT is the use of optimized sensor morphologies. The central idea of this approach is that, on one hand, the higher the surface-to-volume ratio of the active layer, the higher the sensitivity and the better the response/recovery time will be for any given sensing mechanism; on the other hand, reducing the size of the active material close to its Debye length allows the formation of conduction channels that can be completely depleted or even inverted (in terms of majority carrier type) upon analyte adsorption [26,64]. This strategy has led to the development, over the last decade, of a vast amount of 0D and 1D nanostructured materials for gas sensing, including the fabrication of nanoparticle thin films, presenting a large variety of morphologies, as well as nanowires, ribbons, tubes, and rods, either as single structures or in the form of dispersed bundles. For instance, Mitri et al. [51] showed good RT response on PbS-dispersed colloidal QDs towards NO₂ (see Figure 7) with good selectivity, a very low theoretical detection limit of 0.15 ppb, and very good response times, although the recovery is still too long, in the order of several tens of minutes. The overall performance of the device was strongly affected by the thickness of the QD film (Figure 7d), which was controlled by the number of dispersed layers (Figure 7a). This phenomenon was attributed to the fact that sensitivity was expected to increase with thickness layer as more material is able to react to the presence of NO₂ until a certain optimal thickness is reached, from which gas diffusion starts to limit the amount of NO₂ that is capable of reaching the deepest layers. Some works reviewing the room temperature operation of 0D and 1D nanomaterial sensors can be found in Refs [15,16,52,64,78]. Still, most of these devices require high-temperature operation to achieve good sensing performance [33,79].

With the discovery of graphene and the subsequent explosion of novel 2D materials (2DMs), a new pathway has been opened to achieve RT conductometric gas sensors. 2DMs possess the highest possible surface-to-volume ratio, meaning that any change in their surface properties will affect the entire material. The use of different chemical compositions, other than MOs, opens new opportunities in terms of sensor–analyte interaction and sensing mechanisms. Besides, the peculiar properties of this new class of materials, such as their unique electronic structure or their inherent flexibility and optical transparency, considerably broaden their field of applicability [80]. For instance, graphene sensors were shown to provide an extremely low noise signal, intrinsic of this material, capable of distinguishing single-molecule adsorption events, although with negligible recovery at RT, which required either heating at 150 °C or illuminating the sample with UV light [81]. This is why the number of works and reviews devoted to gas sensors based on 2DMs (hereinafter 2D gas sensors) is increasing at a staggering pace, despite being a relatively recent research field [16,54,80,82,83,84,85]. Generally speaking, the transduction, i.e., sensing, mechanism for conductometric 2D gas sensors is simply described as a charge transfer process between the analyte and the 2DM, in contrast to the aforementioned oxygen ionosorption model extensively used to explain the sensing process in MO sensors. Adsorption of analytes is also assumed to be mediated by weak Van der Waals forces, which, in principle, should lead to an easy and fast desorption at RT, facilitating their use as RT sensors with fast response and recovery times [83]. However, this is not always the case, and sometimes, the use of high temperatures or UV light illumination is necessary, even to obtain any recovery at all [81]. In fact, analyte adsorption is strongly affected by the presence of lattice defects, such as vacancies or dopants, as well as some functional groups [41,73,86], which has led to a field of study of its own. For semiconductor 2DMs, such as transition metal dichalcogenides (TMDs), metal dichalcogenides (MD), or graphene oxide (GO), direct charge transfer translates in a shift in their Fermi level, which modifies the concentration of free carriers, the same way as in MOs sensors (Figure 8a,b). For pristine graphene, on the other hand, charge transfer implies a shift in the Fermi level away from Dirac point, which has a minimum in the density of states (DOS), implying an increase in its conductivity [16] (Figure 8c,d). For other metallic 2DMs, such as MXenes, sensing is achieved by the formation of Schottky junctions with a semiconductor (typically a MOs) [84]. In this case, the MXene work function controls the height of the potential barrier and any direct charge transfer from the analyte implies a change in the MXene Fermi level and, therefore, its work function. The use of Schottky junctions as well as van der Waals junctions or 2DM/MOs heterojuntions are also common strategies among 2D gas sensors to enhance the performance of these devices [82,83,84]. These will be briefly discussed in the following section. Similar to what happens with the oxygen ionosorption model for MO sensors, the general description of weak analyte physisorption and charge transfer in 2DMs is a rough simplification of the overall sensing process. Very often, the sensing mechanisms of 2DMs are not well understood, and sometimes, contradictory results are found throughout the literature. For instance, the role played by lattice defects and/or intentional doping/group functionalization in the sensing mechanisms of 2D sensors has been scarcely investigated [86,87,88] and more efforts are required in this direction. Besides, the electronic properties of 2DMs as well as their interaction with adsorbates are strongly influenced by their interaction with the substrate [89,90,91,92], which may introduce important variations in the sensor performance. Properly understanding the specific sensing mechanisms of these particular materials and the effect of the interaction with the substrate, may be at the core of RT-enhanced gas-sensor operation, not only in single 2DM sensors, but also in more complex heterojunction devices.

4.4. Heterojunctions: Schottky, p-n and p-p/n-n Junctions

The use of heterostructures, either by decorating the sensing MOs with metal nanoparticles or by fabricating p-n or n-n/p-p MOs heterojunctions, has also been explored as an alternative to reach room-temperature operation [15,64]. Similarly, the fabrication of 2DM–MOs, 2DM–metal or 2DM–2DM Schottky junctions/heterojunctions is one of the most common strategies to fully exploit the capabilities of 2D materials [54,84,93,94,95]. The underlying physics behind the sensing performance improvement at room temperature of metal–MO heterostructures are usually different form the remaining heterostructures. In this case, the enhancement is usually achieved via a spillover effect [96], thanks to the higher catalytic activity of the metallic nanoparticles. However, in some cases, the formation of a Schottky barrier at the metal–MO heterojunction has also been exploited to obtain highly non-linear responses that can promote the sensing response at room temperature. This high non-linearity is at the core of p-n or n-n/p-p MOs heterojunction response. The potential barrier formed at the heterojunction drastically determines the conductivity properties of these devices, and the height of the barrier and the width of the space charge region is controlled by each material majority carrier density, which is modulated by their response to the analyte [11,12,26,97,98,99]. Thus, it is possible to achieve high enough sensitivities to reduce the operation temperature down to RT. Still, the response and recovery times, as well as the interference of humidity in the sensing process, may be an issue. Depending on the band alignment and Fermi level of the components of the heterojunctions, different band bending, and, thus, different sensing characteristics can be achieved (Figure 9).

Graphene is a particular case as it behaves as a quasi-metallic zero gap material. Thus, direct contacting with other metals should lead to a metal–metal junction. However, owing to its reduced DOS close to the Dirac point, some junction resistance is observed, due to the appearance of a charge transfer region (Figure 9f), the width of which depends on the DOS as $\lambda ={\left[4\pi DOS\left(E_F\right)\right]}^{-1/2}$ [100]. It can also form Schottky contacts with other semiconductors, ranging from conventional MOs (serving either as substrates or as low dimensional structures) [83] to other 2D materials [94].

Control over the Schottky barrier height is essential when dealing with Schottky barrier conductometric sensors. Kim et al. [93] showed that it was possible to properly tune the sensing characteristics of 2D MoS₂ sensors by either changing the electrode material, thus, changing the work function of the electrode, or by changing the number of MoS₂ layers, which determines the bandgap of MoS₂ (Figure 10). They also showed a significant enhancement in the response towards CO and CO₂ by replacing the Au electrode with Ag, going from almost no response at all to 15% sensitivities for 500 and 5000 ppm, respectively.

A different use of a heterojunction barrier formation was demonstrated by Xu et al. [55] on In₂O₃-SnO₂ nanoparticle (NP) composites (Figure 11a) for RT detection of NO_x species. The addition of small portions of In₂O₃ NPs improved the sensitivity of NP SnO₂-based sensors over a factor 11 and, more interestingly, reduced their response times down to a few seconds. Due to their specific band alignment and work function difference, electrons are transferred from In₂O₃ to SnO_2, creating an electron depletion/accumulation region, respectively (Figure 11b,c). The presence of free electrons at the SnO₂ interface enhances the chemisorption and subsequent ionization (i.e., ionosorption) of oxygen species at SnO₂ NP, which, on the one hand, improves its sensitivity towards NO_x species and, on the other hand, facilitates the recovery of the initial state after NO_x has been removed (Figure 11d,e). Thus, in this case, the formation of a heterojunction not only improves the sensitivity of the device but also increases the speed of the sensor response.

4.5. Organic Sensors

An additional approach to design RT conductometric sensors is based on the use of organic materials. In particular, conductive polymers (CP) [59,101,102,103] and phthalocyanines [15,101] are suitable materials, with great performance, sensitivity and fast time response under low-concentration gas influence, even at room-temperature conditions. Inherent properties of polymers related to their great designability, chemical stability, organic nature, and good mechanical properties allow for a simple approach to fabricate light-weight sensing devices, regarded as low-cost effective due to simple synthesis methods, such as polymerization [16,104,105]. Moreover, due to their affinity with water, they can be diluted in aqueous media, allowing an easy device processing to assemble various morphologies, such as thin layers or nanostructures. Various techniques are employed to assemble polymer-based conductometric sensors, such as dip-coating, spin-coating, doctor blade, thermal evaporation, drop-coating, vapor deposition polymerization, or electrochemical deposition [104]. Among the most promising materials, organic conducting polymers, including polypyrrole (PPy) [102,106], polyaniline (PANI) [107], polythiophene (PTh), poly(phenylene vinylene) (PPV), poly(3,4-ethylenedioxythiophene) (PEDOT), and poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) and their derivates [15,59,101,108], are some of the candidates for creating polymeric gas sensing-devices, whose chemical structures are displayed in Figure 12a.

CPs’ gas-sensing working principle is commonly based on the charge transfer between gas molecules and the polymer backbones, which opens the possibility of improving their performance by controlling charge transfer [109]. Their RT performance is mostly related to their electrical conductivity changes due to the affinity with reductive or oxidative gases, which occur even at low temperatures [15]. In fact, unlike MOs, polymer-based sensors could detect many VOCs, such as benzene, which is not chemically reactive with the sensing material at RT, through the measurement of the polymer swelling [110].

Polypyrrole (PPy) was one of the first polymers employed as a gas sensor, most notably for detecting NH₃ gas [59]. The surface of PPy plays a crucial role in the gas-sensing performance as well as its dimensions and morphologies, with thin film [106,111] nanofibers and nanotubes being the most common [112]. Actually, PPy shows great results as a 1D nanostructure due to the large surface area. This allows one, even for low concentrations (25 ppm) at sufficient RT, to obtain a sensor response of 1.12%, with response times as low as 20 s [106]. PPy also presents great selectivity, being investigated for various analytes, such as H₂, NO₂, CO, and HClV [59,111,113].

PEDOT and PEDOT:PSS [103,104,114] are two standpoints on polymeric materials for gas sensing. PEDOT:PSS, which exhibits p-type conductivity, is gaining interest over PEDOT due to its easier processability and higher electrical conductivity. PEDOT:PSS-based sensors have been able to detect different organic analytes, such as NH₃, CO, CO₂, NO₂, as well as ethanol or water vapor [114]. This is particularly interesting for the latter, as PEDOT:PSS films are characterized by a grain-like structure, in which PEDOT grains (conductive) are surrounded by PSS (insulating, hydrophilic), which is linked to the decreasing electrical interconnections among PEDOT chains caused by the water absorption (swelling) of PSS [104]. 1D PEDOT:PSS nanowires can reach high sensitivities of 5.46% and an ultra-fast response of 0.63 s [115].

The sensing mechanism of conducting polymers is directly related to the charge conduction mechanism along the conducting polymer. The conducting polymer structure is composed of conjugated backbones [59], which nominally have a poor conductivity that can be enhanced through appropriate doping [116]. The term doping relates to the concept on inorganic semiconductors, in which the conductivity of a material can be largely increased with the addition of different species in minimum concentrations. These chemical reactants oxidize (or reduce), which modifies the electric conductivity. In a polymer, the overlapping of the π-electron orbitals, composed by delocalized π-electrons alongside the entire chain, contributes to the conductive properties. Doping can enhance the presence of polarons, bipolarons, and solitons (Figure 12b), which create energy states between the HOMO (Highest Occupied Molecular Orbital) and the LUMO (Lowest Unoccupied Molecular Orbital), also affecting the polymer conductivity and, therefore, its sensing properties.

Several sensing mechanisms have been proposed to explain conducting polymer systems, which are often summarized in three groups: (1) redox reaction between the analyte and the chemisorbed oxygen, (2) direct charge transfer between the analyte ant the polymer surface and (3) swelling process from the diffusion of the analyte [117], mechanisms which are schematized in Figure 13. These mechanisms can occur either individually or in a combined manner.

The first proposed mechanism consists of the presence of chemisorbed oxygen molecules, which causes electrons to be removed from the conduction band. Those oxygen species are converted into oxygen ions (single or double), which are then ionosobed. For a p-type material, the hole concentration increases, which causes a decrease in the resistance. When a reducing analyte, such as ethanol or NH₃ (electron-donor), reacts with the ionosorbed O₂^ads, electrons are donated to the conduction band of the p-type material, which reduces the hole concentration and, therefore, the resistance increases. For oxidizing gases, the effect is the contrary and the resistance decreases. For an n-type material, the sensing mechanism becomes the opposite. A classic example is the redox reaction between PPy and ammonia. In the presence of this reducing gas, the p-type PPy suffers a deprotonation, which leads to a decrease in the PPy backbone hole density. As a consequence, the base resistance of the sensor is generally increased [113].

$${\mathrm{PPy}}^{+}+{\mathrm{NH}}_3^0\to {\mathrm{PPy}}^0+{\mathrm{NH}}_3^{+} \left(\mathrm{Adsorption}\right)$$

(7)

$${\mathrm{PPy}}^0+{\mathrm{NH}}_3^{+}\to {\mathrm{PPy}}^{+}+{\mathrm{NH}}_3^0 \left(\mathrm{Desorption}\right)$$

(8)

The direct charge transfer process is also responsible for the sensing ability at room temperature of polymers. When, for instance, NH₃ molecules are absorbed by PPy or PEDOT:PSS by physisorption, their hole concentration will interact with the electrons from the analyte and, thus, the overall hole concentration is reduced, following Equations (7) and (8). This leads to the formation of a neutral polymer backbone and a decrease in their major charge carriers will result in a decrease in the electrical conductivity of the film. This mechanism is also responsible for the changes in conductivity on the doping/dedoping processes [59,117], which increase or decrease the charge carrier concentration.

Finally, the swelling process can also affect the sensing performance of such films. In this case, when the analyte molecules diffuse into the polymer matrix, the electron hopping process is blocked as the inter-chain distance increases because of the swelling, reducing the possible conductive pathways.

Doping/dedoping plays a crucial role in the gas-sensing mechanism of the conducting polymer sensors. Doping levels of conducting polymers can be easily changed by chemical reactions at room temperature, which simplifies the detection of analytes. Electron transferring can cause changes in more than just the resistance of the materials, as its work function can be affected, which also plays a significant role on the sensing properties. The most common approaches for doping CP, often referred to as “secondary doping”, implies either (a) the use of organic solvents (film-treatment or blended with the polymer), (b) surfactant addition, or (c) pH modifications (such as acidic treatments) [116], among others. The specific mechanism for enhanced conductivity is the subject of debate. For instance, enhanced conductivity observed in PEDOT:PSS is believed to be due to the partial removal of PSS from PEDOT:PSS and the conformational change in the PEDOT chain from a benzoid coil-like structure to a quinoid linear-like structure, enhancing hopping transport. Other authors attributed the increase in conductivity to the partial replacement of PSS by SO₄⁻, which increases the bipolaron population, leading to an increase in the doping level and, thus, the conductivity (Figure 12b) [118]. Recently, the combination of CPs with other nanomaterials to assemble hybrid composites has been a subject of debate [108,110].

4.6. Hybrid Composites: Inorganic/Organic Frameworks

While polymer-based sensing behavior shows great sensitivities at RT, its applicability is hindered due to the relatively low conductivity (which is often lower than 10⁻⁵ S/cm) [15,101], prior to any doping or treatment, and affinity to different volatile compounds, such as moisture, which can hinder their long-term performance and stability, suffering degradation and unequal performance. Limits on the surface area and non-porous structure of polymers are often pointed out as the reasons for low sensitivity to certain analytes [104]. A possible strategy to overcome these drawbacks is the design of materials based on hybrid inorganic–inorganic [119], organic–organic [120] and, most notably, inorganic–organic materials [114,121], which are connected either by van der Waals or hydrogen bonding interactions or by strong covalent or ionic bonding. These nanocomposites often show novel or enhanced physico-chemical functionalities, attributed to the synergic interaction between the two phases [15,122,123]. In many cases, this allows one to preserve RT-sensing abilities of the organic matrix, while benefiting from the counterpart properties. The inorganic counterpart can be very diverse [15], from metal oxides [56,109,124,125,126] metal nanoparticles [83,104,127], carbon nanotubes (CNT) [128], or graphene [16,108], among others. From this combination, a gate is opened to achieve an enhanced mechanism for the precisely designed and optimized sensing performance. These mechanisms for hybrid materials are mainly [121], for inorganics-in-organics, by charge transfer [129], charge carrier transport [64,65,66] manipulation/construction of heterojunctions [56,109,125,129].

The combination of polymer with metal oxide semiconductors in various forms, such as nanoparticles [108], thin film or fiber [59,126], or core-shell nanoparticles [56], provides an improvement in sensing response based on the synergistic effect among the polymer and the inorganic counterparts. Different metal and metal oxides have been mixed with PPy, PEDOT:PSS, or PANI, such as Au [60], ZnO [103], SnO₂ [124], WO₃ [125], TiO₂ [126], or CeO₂ [56,109]. The observed change in the resistance is due to physical adsorption of the gaseous molecules onto the surface of the nanocomposite film. The interaction of the gas analyte with a π- electron network of a polymer, which has embedded metal or metal oxide nanoparticles, results in the capture/donation of electrons, depending, as mentioned in Section 2, upon the nature of gaseous molecules, either decreasing/increasing resistance [130].

An n-type metal oxide nanoparticle forms a barrier layer with a p-type polymer matrix, creating a depletion region (p-n junction). The schematic of the formed p-n heterojunction is shown in Figure 14a. The improved sensitivity for hybrid composites could be understood by the modulations in the p-n junction [56,131]. In the most common case, conformed by a p-type polymer and an n-type nanoparticle, the space charge region, also known as the depletion layer, is naturally formed by the union of the p-type polymer and the n-type MOs. When exposed to an electron donor, such as ethanol or NH₃, the concentration of holes in the polymer backbone is reduced and the depletion layer increases in size (marked as w_p and w_p-NH3 on Figure 14a, respectively), which, in turn, results in a conductivity change [125,132]. This modulation of the space-charge region results in enhanced sensitivity for the desired gas. This working principle is similar for n–n, p–n, p–p, and p–n–p heterojunctions.

The creation of heterojunctions between the metal oxide and the polymers, which acts as absorption sites for gas analytes, in combination with a reduction in the signal-to-noise ratio, provides better sensitivity [104]. The sensing mechanism related to charge transfer is different in organic–inorganic frameworks, as charge transfer, in this case, depends not only on intrinsic polymer factors, such as oxidation level or the intra- inter-chain transport, but also on extrinsic factors, which arise because of the nanoparticle–nanoparticle or -metal interaction, playing a significant role [109,114].

In fact, different works have pointed out the importance of the formed heterojunction within the hybrid composite. Wang et al.’s [56] PANI-based sensor using core-shell CeO₂@PANI showed a sensitivity 6.5 to 50 ppm under NH₃ detection and great long-term stability, attributed to the formation of a p-n junction among the nanomaterials through a modification of the space-charge region via the electrons donated by NH₃, which decreased the hole concentration in the PANI, enlarging the depletion layer [56] from w_p to w_p-NH3, as observed in Figure 14. Other effects, such as the conversion of PANI from emeraldine salt (ES) to emeraldine base (EB) due to the NH₃ interaction, can also contribute to the increase in resistance of PANI. As observed in Figure 14b,c, the sensing response of the composite shows a rapid response compared with pristine PANI and the CeO₂ nanoparticles, which is improved with increased NH₃ concentration, the hybrid material being capable of detecting concentrations as low as 2 ppm at RT. Other composites, such as PTh + SnO₂, have shown great sensitivity to NO₂ gas, attributed to the formed p-n junction in addition to the high surface area of the hybrid material [124].

PEDOT:PSS-based composites have also shown an enhancement in the sensitivity against ethanol detection at RT using TiO₂, SnO, or SnO₂ nanoparticles [57,58]. Figure 15 shows a hybrid sensor based on the combination of PEDOT:PSS with ethylene glycol (EG) (as a conductivity enhancer and to enhance nanoparticle dispersion) and with different MO nanoparticles (SnO, SnO, or TiO₂). The performance of the composites under 200 ppm of ethanol gas at RT is shown in the following figures. It must be outlined that SnO nanoparticles exhibit p-type conductivity, while SnO₂ and TiO₂ show n-type conductivity. Figure 15a displays the TEM images of the last two MOs, which average a few nanometers. The polymer was blended with the nanoparticles and deposited by means of spin coating onto a substrate, as shown in Figure 15b,c. Figure 15d shows an example of the resistance changes for the PEDOT:PSS/TiO₂ composite, which is similar for the rest of the composites, as shown in Figure 15f, showing higher sensitivity as compared with the pristine polymer (Figure 15e). However, the highest sensitivity is observed for the composites containing SnO nanoparticles (Figure 15g), attributed to the p-type SnO nature, which could trigger and boost similar sensing mechanisms as for p-type PEDOT:PSS [58].

While we focused this research on polymer/MO composites, not only these materials were considered for hybrid composites. Carbon nanomaterials (CNMs), such as graphene, carbon nanotubes (CNT) or multi-wall carbon nanotubes (MWCNTs) have also been widely employed for chemosensing applications, mainly due to their easily modifiable conductivity, low toxicity, and excellent optoelectronic and mechanical properties, summarized in recent reports [128]. However, by themselves, they present poor results in terms of sensitivity and selectivity and their compounds, with the combination with polymers regarded as some of the most promising materials [128]. In fact, their combination with CPs, such as PEDOT:PSS and polyaniline, allows them to reach sensitivities as high as 28% to NH₃ or NO₂ at 100 ppm [133], as well as a reduction in the recovery time and improved thermal stability [134]. In this case, this phenomenon can be understood due to the physisorption and chemisorption of the analyte molecules by the MWCNT surface. The combination of both inorganic nanoparticles and carbon allotropies has also been considered in recent research. Xiang et al. [135] prepared, by sol–gel and polymerization, a composite containing graphene nanoplatletes (GNs) decorated with TiO₂ nanoparticles (TiO₂@PPy-GN), with a good response to NH₃ gas at RT, attributed to the formation of a p-n junction in the TiO₂ and PPy-GN complex.

5. Future Outlook on Conductometric Gas Sensors: Wearable, Self-Heating, and Flexible Sensors

The fast development of technology has pushed gas-sensing technology into various fields. The current technology trends, such as Internet-of-Things (IoT) and 5G, have created new demand to integrate gas sensors with new devices, such as smartphones, tablets, smart watches, and clothing, to name a few. For those applications, new sensors should fulfill several requirements, such as great flexibility, stability, and low cost, on top of the expected properties of a high-performance device, such as fast response time or sensitivity. In that frame, the achievement of RT operation is considered a favorable condition as well.

The development of flexible gas-sensing devices is gaining increased attention on the market, seeking the possibility of wearable and portable electronic products [80,127]. The main considerations to obtain flexible sensors are selecting the right material and substrate. The material should preserve its sensing abilities despite the strain, stretching, and bending [136]. Currently, substrates can be from plastic, paper, or textile materials [105], while the sensing materials are often polymers, such as polyaniline, polypyrrole, polythiophene (Pth) and poly(3,4-ethylenedioxythiophene) (PEDOT), and PEDOT:PSS, carbon allotropies, such as graphene, transition metal dichalcogenides [80,137], and inorganic/organic hybrid materials [120]. These flexible/wearable sensors have gained substantial interest with the development of technologies, such as e-textiles, in which the sensor can be integrated over/within the textile fabrics [138]. In this field, several works are being conducted regarding the washing, stability, and efficiency of the sensors. For those requirements, inorganic/organic hybrid sensors are among the most promising candidates [120].

Self-heating and energy-saving [138] gas sensors are also a promising trend, which could decrease power consumption [138], especially for MO-based materials. The main idea under self-heating gas sensors is the use of an appropriate voltage, which, via the Joule effect, will generate heat inside the sensor, increasing the temperature and optimizing the sensing working system. Alongside MOs [136], metal–MOs [139] and p-n junctions formed by MOs combined with 2D materials [140] are some of the current candidates. A subset of those self-heating gas sensors can work at RT [139,140].

Finally, machine learning could play a key role in terms of solving some critical issues regarding the sensor operation, such as, for instance, the lack of selectivity from using multiparameter sensors or sensor multi-arrays in combination with deep-learning algorithms for gas discrimination in electronic nose devices [141,142], as well as in future smart sensors. Besides, regulations and safety standards should evolve to undertake the advent of new sensing devices.

6. Conclusions

The advent of a society of ubiquitous sensing in many industries will require gas-sensing systems to be more selective and able to measure lower concentrations of the target gases, including VOC compounds, while involving lower consumption and providing consistent, safe, and reliable performance over longer durations. These challenging tasks, including the development of RT operation sensors, are considered in most of the recent gas-sensing roadmaps. Actually, RT conductometric gas sensors are a trend with expected growth in the near future; however, some challenges must still be faced in order to overcome some of the limitations of these devices in the search for improved stability, selectivity, and sensitivity. This review covers the recent state of the art and progress in the field of gas sensing, with special emphasis on RT operation. In particular, several sensing mechanisms involved in the conductometric sensors’ response are reviewed, as well as materials and characterization techniques. Regarding the RT gas sensors, diverse approaches, including light activation, controlled morphology, heterojunctions, as well as the use of organic and hybrid materials, are considered among the common strategies to pursue RT performance. In particular, conductive polymers have shown outstanding results at RT detection, with great selectivity and sensitivity. Furthermore, the combination of inorganics-in-organic led to improved results, mainly related to the formation of p-n junctions and the depletion zone modification during redox processes, which could also be optimized.

Advances in this technology have led to great sensitivity sensors at RT, which now face new desires and challenges due to new technologies. With the improvement in synthesis techniques, assembly, materials selection, and structure design, high-performance and high-flexibility wearable sensors that can work at RT will be more commonly available in the coming years.