Multilayered and Chemiresistive Thin and Thick Film Gas Sensors for Air Quality Monitoring

Abstract

Selective detection of gases such as nitrogen dioxide (NO2), carbon monoxide (CO), carbon dioxide (CO2), and various volatile organic components (VOCs) is necessary for air quality monitoring. Detection of hydrogen (H2) is equally important as it is a flammable gas and poses serious threat of explosion when exposed to oxygen gas. We have studied the sensing characteristics of these gases using thin film deposited by chemical solution deposition as well as relatively thicker films deposited by atmospheric plasma spray (APS) process. The chapter starts with the sensing mechanism of chemiresistive sensors followed by the definition of gas sensing parameters. Subsequently, we have demonstrated selective NO2 sensing characteristics of zinc oxide-graphene (ZnO-G) multilayered thin film followed by CO and H2 sensing characteristics of ZnO thin film and SnO2 thick film. Cross-sensitivity among CO and H2 gases has been addressed through the analysis of conductance transients with the determination of activation energy, Ea, and heat of adsorption, Q. The concepts of reversible and irreversible sensing have also been discussed in relation to CO and H2 gases. CO2 sensing characteristics of LaFe0.8Co0.2O3 (LFCO)-ZnO thin film have been elucidated. Interference from CO has been addressed with principal component analyses and the ascertaining of Ea and Q values. Additionally, the variation of response with temperature for each gas was simulated to determine distinct parameters for the individual gases. Further, VOC sensing characteristics of copper oxide (CuO) thin film and WO3-SnO2 thick film were investigated. Principal component analysis was performed to discriminate the gases in CuO thin film. The interaction of WO3-SnO2 thick film with various VOCs was found to obey the Freundlich adsorption isotherm based on which Ea and Q values were determined.

1. Introduction

A recent surge in diseases related to poor air quality has made it mandatory to effectively monitor the level of harmful pollutants in the atmosphere. Poor ventilation in indoor spaces as well as vehicular exhausts and emissions from industries poses serious threats to the quality of breathable air. Indoor air quality monitoring has become essential in order to counter the rising incidence of diseases such as sick building syndrome, which occurs in residential and office spaces due to poor ventilation. Indoor air pollutants are mostly volatile organic compounds (VOCs) which include alcohols such as ethanol and methanol and adverse pollutants such as benzene and formaldehyde. Other than this, carbon dioxide (CO₂) and nitrogen dioxide (NO₂) also pollute indoor living space [1].

Outdoor air quality is contaminated mainly due to the presence of carbon monoxide (CO), carbon dioxide, nitrogen oxides (NO_X), sulfur dioxide (SO₂), volatile organic compounds, and particulate matter. Urban sources of pollutants include vehicles and industries. In rural India, many people still use solid fuel such as wood, crop wastes, charcoal, and cow dung in open fires popularly known as “chulhas.” Such cooking practices are highly damaging to human health as they emit huge quantities of CO and NO₂ and cause an array of diseases including but not limited to pneumonia, stroke, heart disease, and lung cancer [2].

Air quality monitoring is thus mandatory to check the levels of various pollutants in the atmosphere. Agencies like Occupational Safety and Health Administration (OSHA) define the permissible exposure limits (PELs) of air pollutants to protect the health of workers. Table 1 shows the permissible limit of various air contaminants.

Pollutant	Averaging time (h)	Level (ppm)
Carbon monoxide	8	35
Nitrogen dioxide	1	100
Carbon dioxide	8	5000
Sulfur dioxide	8	2
Ethanol	8	1000
Acetone	8	1000
Benzene	8	1
Formaldehyde	8	0.75

Table 1.

OSHA permissible limit for various air pollutants [3].

Apart from these pollutants, detection of hydrogen (H₂) gas is also important. It is used as fuel in space craft and rockets. It is reported that almost all CO sensing materials are cross-sensitive toward H₂ as well. It remains a major challenge to address the cross-sensitivity of CO sensor.

To detect these pollutants, one can use chemiresistive type thin or thick film oxide sensors, made from economic precursor materials. The use of oxide sensing materials in thin/thick film form has advantage over commonly used bulk Taguchi-type sensors. Thin/thick film sensors consume less power and also allow miniaturization of gas sensing systems.

The surface of oxide sensors acts as a catalyst to oxidize reducing pollutants (CO, H₂, VOCs, etc.) to relatively benign gases (viz., CO forms CO₂). Limited attempt has so far been made to study the gas sensing performance of oxide ceramics together with catalytic activity to oxidize the gas sensed [4]. For gas sensing performance, thin film sensor is preferred; however, for catalytic conversion, relatively thicker coating of oxide materials over a large area surface (often curved) is required.

Understanding of gas–solid interaction is of utmost importance to optimize the sensing performance of chemiresistive oxide sensors. Typically, Langmuir adsorption isotherm has been utilized to understand the resistance transients recorded upon exposing the oxide sensor to test gas environment. Through these kinetic analyses, interrelation among sensing layer thickness, its morphology, response %, sensitivity, and response/recovery times can be addressed adequately [5, 6, 7]. One can also estimate the adsorption and desorption energies of the test gas and reaction products, respectively. The estimated energies could be an effective tool to differentiate various types of test gases [7]. Base line stability of the sensing materials is also important to fabricate commercial sensing elements. Interrelation among the surface morphology, response time, and test gas concentration needs to be understood to fine-tune the base line stability [6, 8].

In view to the above, the primary goal of the present chapter is to describe a comprehensive approach to model the resistance/conductance transients based on Langmuir and Freundlich adsorption isotherms to address cross-sensitivity. Modeling the resistance/conductance transients allows us to estimate sensing parameters (such as response %, response/recovery time, etc.) and adsorption/desorption energies of test gas and reaction products. The estimated parameters help us to better understand the effect of the receptor (adsorption of gas molecule on sensing surface), transduction (electronic conduction pathway), and utility functions (effect of thickness and morphology of sensing materials) controlling the sensitivity, stability, and selectivity criteria of thin film type chemiresistive sensors.

The chapter is organized as follows: In the following sections, we have elucidated the mechanism of gas sensing followed by defining various sensing parameters. Subsequently we have briefed the dominant material characteristics influencing the sensing performance of thin film chemiresistive type gas sensors. Next, we have outlined the synthesis procedure of thin as well as thick film sensing elements. The synthesized materials were characterized in terms of their phase purity and microstructure evaluation. Finally, we have outlined the details of static as well as dynamic flow gas sensing measurements to investigate the sensing performance of the synthesized materials.

We have thus prepared thin and thick film sensors to detect NO₂, CO, H₂, CO₂, and VOCs (ethanol, acetone, and isopropanol). Gas sensing mechanism has been described for each gas. Cross-sensitivity has been addressed through conductance transient analysis. The gas–solid interaction has been observed to follow either Langmuir or Freundlich adsorption characteristics. Based on this approach, the activation energy, Ea, and heat of adsorption, Q, were calculated to discriminate the gases. Stability of the gas sensors was also studied with respect to thin and thick films.

2. Salient features of semiconducting oxide gas sensor

Semiconducting metal oxide (SMO) sensors are attractive for lower cost, smaller size, simpler operation principle, durability, and ease of fabrication together with their low concentration of gas detection limit [9]. These sensors change their conductivity when exposed to test gases of different concentrations [9]. However, the sensor operating temperature (∼300°C) needs to be lowered, and cross-sensitivity toward multiple gases needs to be minimized for its wide commercial adaptation. The operating principle and salient features of SMO-based chemiresistive sensors are briefly described as follows:

2.1 Principal of operation

For reducing CO gas sensing, Figure 1 schematically illustrates the operative sensing mechanism. At elevated temperature, when the sensing film is exposed to air, the sensor surface provides active sites for the chemiadsorption of oxygen ions. Thus, oxygen adsorbs on to the sensor surface by accepting electrons from the conduction band of the sensing material. In the operating temperature ranging 100–500°C, ${\mathrm{O}}_2^{-},{\mathrm{O}}^{-},\text{and}{\mathrm{O}}^{2-}$ are chemiadsorbed [10] forming an electron depletion layer (EDL). The EDL impedes grain to grain electron migration increasing the surface resistance for an “n”-type sensing material. When the sensor is exposed to a reducing gas such as CO or H₂, the adsorbed oxygen ions react with the test gas to release the electrons back to the conduction band of the sensing material. As a result the EDL width is reduced to decrease the surface resistance of the sensor.

Figure 1.

Gas sensing mechanism for a semiconducting metal oxide sensor.

In the case of a p-type sensor, since the majority carriers are holes, the oxidation reaction of the reducing test gas and the oxygen ions on the sensor surface would decrease the concentration of holes, increasing the sensor resistance. Assuming n-type and p-type sensors have similar morphologies, the response of p-type and n-type sensor is related by the following relation [11].

$$S_p=\sqrt{S_n}$$ E1

where $S_p\text{is the}\mathrm{gas}\text{response of}\mathrm{p}-\text{type sensor}$ and $S_n$ is the gas response of n − type sensor.

2.2 Characteristic sensor parameters

Figure 2 illustrates the characteristic sensor parameters one can estimate from the recorded resistance transient of an n-type sensor used for reducing gas sensing.

Figure 2.

Resistance transient of an n-type gas sensor showing stable resistance in air (R_a), stable resistance in gas (R_g), response time (τ_res), and recovery time (τ_rec).

Response (S) (%): Response is defined by the expression $\left(R_a-R_g\right)/R_a\times 100$ or $R_a/R_g$ where $R_a$ and $R_g$ are the measured resistances of the sensor exposed to air and test gas, respectively.

Response time ( ${\mathit{\tau }}_{\mathit{res}}$ ) (s): Time required for a sensor to reach 90% of the maximum response on exposure to test gas [12].

Recovery time ( ${\mathit{\tau }}_{\mathit{rec}}$ ) (s): Time required for a sensor to reach 90% of the original baseline resistance upon removal of the test gas.

Optimum temperature ( ${\mathit{T}}_{\mathit{opt}}$ ) (°C): The temperature at which the sensor exhibits maximum response.

Selectivity: The ability of a sensor to sense a selective gas preferentially among other test gases.

Stability: Minimum variation of sensor base resistance over a period of time.

Sensitivity: Change of measured signal (resistance, voltage or current) per analyte concentration unit [13].

2.3 Factors affecting gas sensing performance of thin film sensor

The sensing performance of SMO-based sensor is primarily influenced by the receptor, transducer, and utility functions of thin film sensing elements. The receptor function is influenced by the crystallite size of the sensing material, as at elevated temperature oxygen ions are chemiadsorbed on these crystallites forming a charge-depleted layer. The receptor function controls the interaction of test gases with the sensing elements. The transducer function controls the flow of charges between the electrodes. Finally the utility function controls the diffusion of gases. The parameters that regulate the receptor function of the sensor include composition of the material, surface to volume ratio, and size of the pores and grains. The transducer function is influenced by the grain to grain contact and the phase purity of the material. The composition of the material influences the defect chemistry which in turn changes the nature and concentration of the defects in the sensing material. The surface to volume ratio, pore size and grain size, and grain to grain contact of the sensing material are dependent on the growth condition.

Usually, the sintered block as well as thick film sensors consists of loosely sintered secondary particles made of innumerous tiny primary particles. Generally, the dimension of pores between the primary particles is in microns (termed as macropores), whereas those in between the primary particles are in the order of nanometers (<50 nm) (termed as mesopores). When the sensor is aged at elevated temperature, oxygen gas is diffused through both the macro- (corresponds to molecular diffusion) and mesoporous (corresponds to Knudsen diffusion) regions. Sensor particles are charge depleted due to chemiadsorption of oxygen. During gas sensing, the test gas diffuses through macro- and mesoporous regions and reacts with chemiadsorbed oxygen. The diffusion of gases through the mesoporous regions is known to play an important role during reducing gas sensing using semiconducting oxide-based sensing elements. For compact thin films, the gas-solid interaction mostly takes place at the geometric surface and to a limited extent along grain boundaries. In contrast, the active surface of the porous thin film is much larger than the compact thin films. In both these thin film structures, the key parameters that control the receptor function of the sensing element include its composition, nature of metal-oxygen bonding, specific surface area, porosity, and crystallite size. The transducer function is influenced by grain to grain contact and phase purity of the sensing element, and the utility function will be controlled by the gas diffusion either through grain boundaries or pores. For thin film sensing elements with pore size distribution in the range of 2–50 nm, Knudsen-type gas diffusion seems to be operative. Since the Knudsen diffusion coefficient is related to operating temperature, pore radius, molecular weight of the diffusing gas, and diffusion distance are related to film thickness; it is quite obvious that the sensing performance of a thin film sensor is related to the operating temperature, film thickness, and molecular weight of the test gas. Very limited attempt has so far been made to theoretically investigate the gas transport phenomena within the sensing film and correlate it with its sensing performance.

In the subsequent sections, we have outlined the sensing performance of ZnO-graphene multilayered thin films for NO_x sensing. Subsequently, CO and H₂ sensing performance of ZnO thin film and SnO₂ thick film were investigated. This was followed by CO₂ sensing performance of 0.5% LFCO-ZnO thin film and VOC sensing characteristics of copper oxide (CuO) thin film and WO₃-SnO₂ thick film.

At constant sensor operating temperature, we assumed monolayer coverage of the reducing gas. Assuming Langmuir adsorption kinetics, we have theoretically predicted the response and recovery transients during gas sensing using thin and thick film sensing elements. From an application point of view, marginal base resistance drift of these thin film sensing elements is desirable during repeated response and recovery cycles. Often a significant drift in base line resistance is observed due to the partial recovery of thin and thick film sensing elements. For such partial recovery, the sensing is termed irreversible, whereas for reversible sensing, the base resistance is fully recovered. For a wide range of test gas concentrations and operating temperatures, the response transients of thin film sensing elements are modeled using Langmuir-Hinshelwood reaction mechanism. It is predicted that for irreversible-type sensing, the response time is reduced with the increase in test gas concentration, whereas for reversible sensing, response time is independent of the test gas concentration [12]. The WO₃-SnO₂ thick film sensor was modeled with Freundlich adsorption isotherm in which the concentration-dependent conductance was ascertained to be nonlinear.

3. Experimental

An economic wet chemical synthesis route was adopted to synthesize thin film gas sensing elements. Wet chemical synthesis is also energy efficient, and relatively cheaper precursor materials are used in wet chemical synthesis. We have synthesized the ZnO-graphene (G) multilayer to study its NO₂ sensing characteristics. ZnO thin film sol was prepared by dissolving stoichiometric amount of zinc acetate dihydrate (Zn(CH₃COO)₂.2H₂O) in 2-methoxyethanol (2-MOE) solvent with monoethanolamine (MEA) as stabilizer. The final concentration of the precursor solution was ∼0.4 M/L. The precursor solution was heated at ∼80°C for complete dissolution and stabilization. Graphene sheet from crystalline graphite was exfoliated by micromechanical cleavage using facile ultrasonication synthesis [14]. For this purpose, 5 mg of graphite powder was mixed in 5 ml of 2-MOE solution. N-methyl-2-pyrrolidine (NMP) solution was added for stabilization of the graphite sol. This suspension was then ultrasonicated for 3 h at room temperature and aged for 48 h. Multilayered ZnO-G thin films were fabricated by alternate deposition of ZnO and exfoliated graphite sol. Prior to sol deposition, the glass substrates were cleaned in trichloroethylene for 5 min followed by ultrasonic cleaning in the presence of acetone and deionized water. This was followed by drying of the substrates in flowing air. Details of deposition process have been reported elsewhere [14]. The multilayered films were annealed at 500°C for 1 h in air. The sensing performance of ZnO-G multilayered films were characterized using NO₂ (0.5–5.0 ppm), CO (50 and 500 ppm), H₂ (50 and 500 ppm), and i-C₄H₁₀ (50 and 500 ppm) as test gases. The operating temperature was maintained in the range 100–200°C during gas sensing measurements. Additionally, mixed gas sensing behavior for the ZnO-G multilayered films has also been investigated using NO₂/CO (5 ppm/50 ppm), NO₂/H₂ (5 ppm/50 ppm), and NO₂/i-C₄H₁₀ (5 ppm/50 ppm) gas mixtures.

We have also studied the CO and H₂ sensing behavior of ZnO thin film and SnO₂ thick film. ZnO thin film was synthesized following the method described above. SnO₂ thick film was synthesized using a plasma spray deposition process. Commercial SnO₂ powder was used as feedstock material for the plasma spray process. Nitrogen (sans hydrogen) was used as the plasma forming gas and compressed air was used for cooling. Details of the process and equipment can be found elsewhere [7]. Alumina substrates (Ants Ceramic, India) were used for deposition as these are unaffected by the high temperatures generated during the atmospheric plasma spray process.

We have also studied the CO₂ sensing behavior of 0.5% LFCO (LaFe_0.8Co_0.2O₃)-ZnO thin film composite. Lanthanum nitrate hexahydrate (La(NO)₃.6H₂O), iron (III) nitrate nonahydrate (Fe(NO)₃.9H₂O), and cobalt (II) nitrate hexahydrate (Co(NO)₂.6H₂O) were dissolved in 2-MOE to prepare LFCO precursor sol with a final solution of 0.2 M via continuous stirring at room temperature. LFCO and ZnO sol were mixed in appropriate proportions to make composites of the composition xLFCO-(1−x) ZnO. The resulting sol was then spin coated on quartz substrate. Each layer was baked at 300°C with an intermittent heating interval of 10 min between layers to burn out the organic moieties. The coating and baking processes were repeated to yield films about 300nm thick. The resultant films were then annealed at 600°C for 2 h.

VOC sensing characteristics of CuO thin film and WO₃-SnO₂ thick film were measured in a static gas sensing system. CuO sol (0.25 M) was synthesized using copper (II) acetate monohydrate (Cu(COOCH₃).H₂O) and 2-MOE as precursor materials. Glass substrates used for deposition were cleaned using trichloroethylene, acetone, and deionized water as described earlier. CuO sol was spin coated on glass substrates. Each layer was baked at 300°C for 5 min. The resultant film which was about 200 nm thick was annealed at 600°C for 1 hour to achieve complete crystallization. WO₃-SnO₂ thick film composite was deposited on alumina substrates using a plasma spray deposition process. The feedstock material consists of WO₃ and SnO₂ powders in the weight ratio ∼25:75. This ratio was found to be optimum in a series of experiments with bulk pellets with various weight ratios (∼25:75, 50:50, and 75:25) of WO₃ and SnO₂ [15].

The resistance change of the thin film sensing elements was measured using gold interdigitated electrode which was sputter coated onto the thin films using a dc sputter coater (Cressington 108, Ted Pella, Inc., USA). High-temperature silver paste was used as electrodes for the thick coatings. An indigenously developed dynamic gas sensing setup was used to measure the gas sensing characteristics of the thin films and thick plasma sprayed coatings. The gas sensing setup consists of a gas chamber connected to mass flow controllers (PR 4000, MKS Instruments, Germany) which control the flow rate of gases. The flow rate is calculated by the relation [12, 16].

$$C_{\mathit{\text{mixed}}\mathit{gas}}=\left[C_{\mathit{\text{test}}\mathit{gas}}\times \frac{{\mathit{dV}}_{\mathit{\text{test}}\mathit{gas}}}{\mathit{dt}}\right]/\left[\frac{{\mathit{dV}}_{\mathit{\text{test}}\mathit{gas}}}{\mathit{dt}}+\frac{{\mathit{dV}}_{\mathit{\text{carrier}}\mathit{gas}}}{\mathit{dt}}\right]$$ E2

where $C_{\mathit{\text{test}}\mathit{gas}}\text{is the concentration of test}\mathrm{gas}$ , $\frac{{\mathit{dV}}_{\mathit{\text{test}}\mathit{gas}}}{\mathit{dt}}$ is the flow rate of test gas, and $\frac{{\mathit{dV}}_{\mathit{\text{carrier}}\mathit{gas}}}{\mathit{dt}}\text{is the flow rate of carrier}\mathrm{gas}$ .

The temperature of the sensor inside the dynamic chamber is controlled with a temperature controller. An electrometer (6517 A, Keithley, USA) is connected to the sensor via gold-coated silver probes. Necessary voltage for electron migration between the grains of the sensor is controlled through the electrometer. The entire gas sensing setup is interfaced with a PC equipped with an RS-232 interface, GPIB card (National Instruments, USA), and LabVIEW 8.5 (National Instruments, USA) software [17]. A schematic of the mentioned gas sensing setup may be found in [12].

A static gas sensing set was employed to study the VOC (ethanol, acetone, and isopropanol) sensing characteristics of CuO thin film and WO₃-SnO₂ thick film. Such a setup (also indigenous) consists of a rectangular chamber housing the sensor and the heater which is connected to the temperature controller [18]. The sensor is connected to a microcontroller which in turn is connected to the PC. Details of the system schematic may be found elsewhere [15]. Data (in this case, resistance) is collected at the rate of 1 bit/sec with the help of terminal emulation software. The sensor is heated to the temperature of operation and aged at that temperature to obtain a stable resistance ( $R_a$ ) which is followed by injection of gas. The resistance at which the sensor stabilizes in the presence of gas is termed $R_g$ . The response of the sensor to the test gas in measured in terms of response percent which is defined as [19, 20].

$$\mathit{\text{Response}}\left(S\right)\%=\frac{\left|R_a-R_g\right|}{R_a}\times 100$$ E3

4. Investigations on the gas sensing performance for air quality monitoring

Air quality monitoring involves the monitoring and quantification of various gases present in the atmosphere. Such gases include NO_X [21], CO [22], CO₂ [23], and VOCs [24]. In addition, flammable gases such as H₂ also need detection in order to prevent major disasters [25]. Apart from the detection of low concentration test gases, addressing the cross-sensitivity of the sensors also remains a major research issue. In addition to this, cost of air quality monitoring system is high, and efforts are needed to make it economic for wide uses. Keeping these considerations in mind, we have developed sensors to monitor these gases selectively. For cross-sensitive sensing elements, necessary protocols are developed to address their selectivity. The sensing performances of synthesized sensing elements to detect test gases pertinent to air quality monitoring are described as follows.

4.1 NO₂ sensing

The main sources of NO₂ emissions are the incomplete burning of fuel and vehicular exhausts. According to a report in The Hindu [26], one of India’s premier newspapers, winter smog is caused in many parts of northern India due to burning of crop residue. An estimated 240 million tons of NO_X (NO and NO₂) is generated because of this practice. Figure 3 shows percentage of NO_X emissions in India from various sources. Other than this, vehicular exhausts also contribute to NO_X formation. NO_X, along with other pollutants such as SO₂, VOCs, and particulate matter, forms smog as well as acid rain. These are markers of extremely polluted air and can have adverse health effect on humans.

Figure 3.

NO_X emissions in India from various sources. Most of NO_X emissions in India come from the use of chemical fertilizers in agriculture.

Traditionally, semiconducting metal oxides such as SnO₂ [27], WO₃ [28], ZnO [29], etc. have been used for NO₂ sensing. Recently, composites such as Fe₂O₃-ZnO [30] has also gained prominence for selective NO₂ sensing. In addition, metal oxide (MO)-graphene (G) composites such as 0.5 wt%G-WO₃ [31] and SnO₂-RGO [32] have also been reported as these sensors show high response at low operating temperature. These sensors are also highly selective.

Based on literature study (Table 2), we realized that MO-graphene (G) sensor will be an effective candidate for NO₂ sensing study. We have chosen the MO to be ZnO as it is an extremely stable sensor and can be easily modulated with graphene to operate at a low temperature. We have thus prepared a ZnO-G multilayer system for NO₂ detection. We have measured NO₂ sensing with ZnO which is an n-type semiconductor. It shows optimum response of 26% to 5 ppm NO₂ at 150°C when the test temperature range is 100–200°C. Below 150°C, the response to NO₂ is almost negligible. However, with the addition of graphene to ZnO layers, the sensing response significantly increases to 894% [14]. These results are presented in Figure 4.

Sensor material	Form	Gas	Topt (°C)	S%	Selectivity	τres, τrec (s)	Conc. (ppm)	Ref.
SnO2-RGO	Nano powder	NO2	30	4.63 (Rg/Ra)	Yes	177, 510	5	[32]
SnO2	Hollow sphere	NO2	160	—	Yes	90, 25	5	[27]
0.5 wt%G-WO3	Nanopowder	NO2	RT, 250	1.96, 130 (Rg/Ra)	No	25–200	1, 5	[31]
ZnO	Nano powder	NO2	350	1.8 (Rg/Ra)	Yes	180, —	1	[29]
WO3	Thin film	NO2	300	70	No	—	10	[28]
Fe2O3-ZnO	Thin film	NO2	150	—	No	—	2.5	[30]
WO3	Thin film	NO2	300	87	Yes	300, 360	2	[33]
ZnO-G	Thin film	NO2	150	894	Yes	150, 315	5	[14]

Table 2.

Comparison of the NO₂ sensing characteristics of various metal oxides.

Figure 4.

Resistance transients of (a) ZnO and (b) ZnO-G multilayer at 150°C. an enormous increase in response is observed on addition of graphene (Reprinted from [14] with the permission of AIP publishing).

The response of ZnO-G multilayer thin film was also measured for 500 ppm H₂, CO, and i-butane gases in the said temperature range. It was found that the response to 5 ppm NO₂ was far superior to any of the other gases. We have defined the selectivity (κ) as

$$\kappa =\frac{S_{\mathit{NO}2}}{S_{\mathit{gas}}}$$ E4

where $S_{\mathit{NO}2}$ is the response % for NO₂ and $S_{\mathit{gas}}$ is the response % for other gas (H₂, CO, and i-butane).

The selectivity plot for NO₂ with respect to H₂, CO, and i-butane has been shown in Figure 5. For all the gases, the selectivity factor κ has been found to be greater than 1. Thus we could detect NO₂ selectively in the presence of other gases.

Figure 5.

Response % for 500 ppm i-butane, 500 ppm CO, 5 ppm NO₂, and 500 ppm H₂ gases. Figure clearly shows the selective detection of NO₂ at 150°C (Reprinted from [14] with the permission of AIP publishing).

4.1.1 NO₂ sensing mechanism

The significant enhancement in NO₂ sensing may be attributed to the formation of p-n junction between ZnO and graphene. In ZnO-G composite, graphene adheres to ZnO surface to increase the interaction with NO₂. Due to the formation of p-n junction, electron flows from graphene to ZnO resulting in an electron-depleted surface. On exposure to air, majority of ${\mathit{O}}_{\mathbf{2}}^{\mathit{-}}$ (dominant species at 150°C) species adsorb on to the ZnO-G surface, leading to the formation of electron-depleted layers. NO₂ reacts with this surface to form NO which is explained via the following equations:

$${\mathit{NO}}_2\left(\mathit{gas}\right)+e^{-}\leftrightarrow {\mathit{NO}}_2^{-}\left(\mathit{ads}\right)$$ E5

$${\mathit{NO}}_2^{-}\left(\mathit{ads}\right)+O_2^{-}\leftrightarrow 2\mathit{NO}\left(\mathit{gas}\right)\uparrow +2O_2^{-}$$ E6

The interaction of reducing gases (H₂, CO and i-butane) with $O_2^{-}$ is poor at 150°C, thus leading to selective detection of NO₂. A detailed figure of the sensing mechanism may be found in [14].

4.1.2 Mixed gas sensing

We felt it important to sense NO₂ selectively in a mixed gas environment. Thus, the sensing performance of NO₂ was investigated in the presence of 5 ppm NO₂ and 50 ppm CO, 5 ppm NO₂ and 50 ppm i-butane, and 5 ppm NO₂ and 50 ppm H₂. The response % histograms are presented in Figure 6. Principal component analyses (PCA) were used to address the cross-sensitivity. PCA involves the fast Fourier transforms (FFT) of the resistance transients of each gas (accomplished using MATLAB®, Mathworks, USA) which are fed into this unsupervised pattern recognition algorithm using STATISTICA-9, StatSoft, USA.

Figure 6.

Gas sensing responses of individual NO₂, CO, i-C₄H₁₀, and H₂ gases and mixed gas sensing for NO₂/CO, NO₂/i-butane, and NO₂/H₂ (Reprinted from [14] with the permission of AIP publishing).

The plots between PC #1 and PC#2 are plotted separately in Figure 7. As shown in the inset of Figure 7a, distinct cluster forms among individual test gases (H₂, CO, and iC₄H₁₀). Also for NO₂-CO, NO₂-i-butane, and NO₂-H₂ mixtures, distinct clustering is readily identified in Figure 7a. PCA is also performed in mixed gases varying NO₂ contents in the range of 0.5–5 ppm (Figure 7b). Note that distinct clustering is also achieved in mixed gases with variable NO₂ contents, which eventually leads to successful differentiation between them.

Figure 7.

Principal component analysis of (a) NO₂, CO, H₂, i-C₄H₁₀, NO₂-CO, NO₂-H₂, and NO₂-i-C₄H₁₀ gases and (b) mixed NO₂/CO, NO₂-H₂, and NO₂-i-C₄H₁₀ gases and NO₂ gas varying NO₂ contents in the range 0.5–5 ppm measured at 150°C (Reprinted from [14] with the permission of AIP publishing).

4.2 CO and H₂ sensing characteristics of thin and thick films

Carbon monoxide is a major air pollutant. Open burning of wastes as well as fuel and biomass burning mainly in rural India is the main source of carbon monoxide. Additionally, fuel adulteration, especially in auto-rickshaws, also adds to the CO in air. On breathing CO-rich polluted air, one might end up inhaling CO as opposed to oxygen. This CO then easily combines with the hemoglobin in blood forming carboxyhemoglobin, which reduces the oxygen-carrying capacity of blood. Additionally, tobacco smoke and heating appliances add to CO pollution of indoor air.

Hydrogen (H₂) is a colorless, odorless, and tasteless gas. H₂ is also highly flammable with a low ignition energy (0.017 mJ) and low ignition temperature (∼560°C). Hydrogen detection is critical to maintaining the safety of nuclear reactors where the reaction of water with high temperature core or cladding gives rise to evolution of hydrogen in radioactive waste tanks [34]. A small amount of H₂ is also produced as a result of methane explosions in coal mines [35]. H₂ is used as a liquid propellant for rockets as it is a zero emission fuel. It is also used in the reduction of metallic ores and for methanol production. H₂ has a lower explosive limit (LEL) of 4 vol % which makes its detection fairly crucial.

Semiconducting metal oxides such as SnO₂, ZnO, WO₃, etc. have demonstrated huge research potential when it comes to sensing of reducing gases such as hydrogen, carbon monoxide, hydrogen sulfide, ammonia, and VOCs. However, such materials exhibit poor selectivity, and efforts are on to achieve selectivity either by doping metal oxides or synthesis of composites. A brief review of literature (Table 3) suggests that selectivity has not been addressed for CO and H₂ sensing of metal oxides. We have synthesized ZnO thin film via spin coating process and SnO₂ thick film via plasma spray deposition process and studied their gas sensing characteristics with respect to H₂ and CO gases. We have analyzed the issue of selectivity through conductance transient analyses. The analyses have been performed for both thin and thick films for comparison. Additionally, we have discussed how stability varies in thin and thick film gas sensors when the test gases under consideration are the same.

Sensor material	Geometry	Gas	Topt (°C)	S%	Selectivity; (w.r.t H2/CO)	τres, τrec (s)	Conc. (ppm)	Ref.
Sm2O3	Thick film	CO	250	1.4 (Rg/Ra)	No	35, 110	5	[36]
CuO	Nano-tube	CO	175	3.27 (Rg/Ra)	No	29, 37	1000	[37]
In2O3	Thin film	CO	350	45, 85 (Rg/Ra)	No	10– 50	1000	[38]
Pd-ZnO	Thin film	CO	150	—	No	17, 23	500	[39]
ZnO	Thin film	H2	300	21	No	58, 70	200	[40]
In-ZnO	Thin film	H2	300	65	No	35, 64	500	[41]
SnO2	Thin film	H2	350	83	No	23, 139	265	[42]
ZnO	Thin film	H2	400	23	Yes	110	1200	[43]
ZnO	Thin film	H2	300	68	Yes	53, 67	500	[9]
SnO2	Thick film	H2	250	80	Yes	30, 118	500	[7]

Table 3.

Comparison of the CO and H₂ sensing characteristics of various semiconducting metal oxides.

ZnO thin film and SnO₂ thick film were investigated for H₂ and CO sensing characteristics in the temperature range 350–200°C. The concentration of each gas was fixed at 500 ppm. A maximum response of 68% for 500 ppm H₂ and 49% for 500 ppm CO was recorded at 300°C by ZnO thin film. SnO₂ thick film recorded maximum response to 500 ppm H₂ (81%) and 500 ppm CO (73%) at 250°C. Careful inspection of Figure 8 circumstantiates cross-sensitivity of H₂ and CO gases at all temperatures for both ZnO thin film and SnO₂ thick film sensors. This necessitates further analysis of the sensing data obtained. Analyses of conductance transients are a well-proven coherent technique to address the problem of cross-sensitivity. Although cross-sensitivity is shown by the sensors at all temperatures, we have performed mathematical analyses only for the optimum temperature of each kind of film. For H₂ and CO sensing, the recorded conductance transients of both ZnO (thin film) and SnO₂ (thick film) sensors were modeled using Langmuir adsorption kinetics [5].

Figure 8.

Response % of (a) ZnO thin film and (b) SnO₂ thick film in response to 500 ppm H₂ and 500 ppm CO.

4.2.1 Analyses of conductance transients

In a preceding section, we have already outlined the sensing mechanism of n-type semiconducting sensor toward reducing gas sensing. Thus, ambient oxygen physiadsorbs on the sensor surface at ambient temperature (Eq. (7)). At elevated temperatures, the physiadsorbed oxygens are chemiadsorbed (Eq. (8)) by extracting electrons from semiconductor oxide grains. This creates an electron depletion layer (EDL) at the sensor surface [5, 10, 44, 45].

$$\frac{1}{2}{O}_2+{\mathit{\text{sensor}}}_{\mathit{\text{surface}}}\leftrightarrow O_{\mathit{ad}-\mathit{\text{surface}}}\left(\mathit{\text{physi}}-\mathit{\text{adsorption}}\right)$$ E7

$$O_{\mathit{ad}-\mathit{\text{surface}}}+e\leftrightarrow O_{\mathit{ads}}^{-}\left(\mathit{\text{chemi}}-\mathit{\text{adsorption}}\right)$$ E8

On exposing the sensor to a reducing gas, the chemiadsorbed oxygen reacts with the test gas (R) to yield oxidized reaction products. The reaction schemes and desorption of reaction product (RO_gas) is described in Eqs. (9)–(11).

$$R+\left[\mathit{\text{sensor}}\right]\leftrightarrow R_{\mathit{ad}}\left(\mathit{\text{physi}}-\mathit{\text{adsorption}}\right)$$ E9

$$R_{\mathit{ad}}+O_{{\mathit{ad}}^{-}}\leftrightarrow {\mathit{RO}}_{\mathit{ad}}+e$$ E10

$${\mathit{RO}}_{\mathit{ad}}\leftrightarrow {\mathit{RO}}_{\mathit{gas}}\uparrow +{\mathit{\text{sensor}}}_{\mathit{\text{surface}}}$$ E11

Among the cited relations, adsorption of oxygen ions and desorption of the reaction product (Eqs. (10) and (11)) are the rate-determining steps for response and recovery, respectively.

As discussed earlier, when the sensor is exposed to air, oxygen chemisorbs onto the sensor surface which results in the formation of an electron depletion layer (EDL). The EDL formed impedes the transfer of charge carriers from one grain to another. The relation between the height of this potential barrier and the conductance G of the sensor is given by the following Eq. (5):

$$G=G_{0}exp\left(-\frac{e{V}_s}{\mathit{kT}}\right)$$ E12

where eV_s is the Schottky barrier, k is the Boltzmann constant, and T is the absolute temperature of operation of the sensor.

As the sensor is exposed to a reducing gas like CO or H₂, the chemisorbed oxygen ions react with the gas to reduce the potential barrier between the grains, thus allowing the flow of electrons [46]. This process increases the conductance of the sensor.

The adsorption process of gas molecules on a solid surface was studied by Langmuir who developed an empirical model based on the surface coverage θ. The Langmuir model assumes that (i) the surface on which the gas adsorbs is homogeneous, (ii) a specific number of sites are available for gas adsorption, and once all these sites are occupied, further adsorption is not possible. This is called monolayer adsorption, and (iii) all sites on which gas is adsorbed are equivalent in terms of heat of adsorption. Assuming the above, Langmuir formulated the equation below [5, 15].

$$\theta \left(t\right)=\frac{N\left(t\right)}{N^{\ast }}=A\{1-exp\left(-\frac{t}{t_A}\right)\}$$ E13

where N(t) is the number of gas molecules adsorbed at time t and N^* is the total adsorption sites available, and

$$A\equiv \frac{P}{P+b}$$ E14

where P is the gas pressure and b is related to heat of adsorption Q and is defined by the following relation:

$$\mathrm{b}={\mathrm{b}}_{0}exp\left(-\frac{\mathrm{Q}}{\mathrm{RT}}\right)$$ E15

$${\mathrm{b}}_0=\frac{\mathrm{\upsilon }}{{\mathrm{K}}_0.{\mathrm{S}}_{\text{area}}.{\mathrm{N}}_{\mathrm{A}}}\times \sqrt{2\text{πMRT}}$$ E16

where b₀ is a constant, Q is the heat of adsorption (Q = E_d − E_a > 0), E_a and E_d are the adsorption and desorption activation energies of gas molecules, R is the gas constant, and T is the operating temperature, $T_{\mathit{opt}}$ . The pre-exponential constant b_o depends on υ (frequency of oscillation of adsorbed molecules), K_o (fraction of those molecules with energy > E_A, adsorbed on surface), S_area (the surface area of a single adsorbed molecule), N_A (Avogadro number), M (molar mass of the gas), and R (the gas constant).

The amount of gas adsorption is dependent on the nature of the gas sensing surface, the operating temperature, and pressure of the gas. An adsorption isotherm is defined as the study of the amount of gas adsorption as a function of gas pressure. The simplest adsorption isotherm was given by Langmuir in which he assumes no interaction between adsorbed atoms or molecules. At equilibrium, the rate of adsorption is equal to the rate of desorption. Thus, when t → ∞,

$$\mathrm{\theta }=\frac{\mathrm{N}\left(\mathrm{t}\right)}{{\mathrm{N}}^{\ast }}=\mathrm{A}=\frac{\mathrm{P}}{\mathrm{P}+\mathrm{b}}$$ E17

Assuming a linear functional dependence between θ(t) and conductance transient (G(t)), the response behavior of the sensor that follows Langmuir model can be written as [5]

$$G{\left(t\right)}^{\mathit{\text{response}}}=G_0+G_1\left[1-\mathit{exp}\frac{-t}{{\tau }^{\mathit{res}}}\right]$$ E18

where $G_0$ is the base conductance of the sensor at time t = 0, $G_1$ is the fitting parameter, and ${\tau }^{\mathit{res}}$ is the characteristic response time. This equation will only be valid if all three of Langmuir’s criteria are satisfied.

$G_1$ in Eq. (18) is identical to A in Eq. (17); hence, at low gas concentration if we assume P ≈ gas concentration (C), we can then write

$$\frac{{\mathrm{G}}_1}{1-{\mathrm{G}}_1}=\frac{\mathrm{C}}{\mathrm{b}}$$ E19

If Langmuir adsorption isotherm is followed, then variation of $G_1$ /(1- $G_1$ ) vs. C should be a linear fit and pass through the origin. The parameter b can be estimated from the slope of the plot.

Using Eq. (18), conductance transients during response at various concentrations for SnO₂ sensing element for H₂ and CO sensing were fitted to estimate values of $G_0$ , $G_1,$ and ${\tau }^{\mathit{res}}$ . Figure 9a shows the fitting of conductance transient of SnO₂ thick film at 250°C in response to 500 ppm H₂. Figure 9b shows the variation of $G_1$ /(1- $G_1$ ) with test gas concentration (C) at 250°C. Since the linear fits for H₂ and CO pass through the origin, Langmuir adsorption isotherm is satisfied [5, 47, 48]. The parameter b can be obtained from the inverse of the slope of the plot. Accordingly, the estimated ‘b’ values for H₂ (1.5349 × 10¹⁰ Pa) and CO (1.7834 × 10¹¹ Pa) have been determined from these plots [7]. From Eq. (15), expression for heat of adsorption (Q) can be written as

Figure 9.

(a) Conductance transient plot for 500 ppm H₂ sensing. Experimental points are shown by symbols, and the solid line was fitted using Eq. (18), (b) linear variation of G1/(1-G1) with concentration, (c) log?log plot of τ^res with C. Experimental points are shown by symbols, and the solid line was fitted using Eq. (23), (d) variation of E_a with ln C for H₂ and CO sensing. For all these measurements, operating temperature was kept fixed at 250°C (Figure 9a?d, reprinted from [7] with permission from Elsevier).

$$Q=\mathit{RT}\left[\mathit{ln}\left(b_0\right)-\mathit{ln}\left(b\right)\right]$$ E20

The molecular area of nitrogen is approximately 16.2 Å² as given in [49]. We have considered this as the area per site occupied by the test gas molecule (S_area). Nitrogen has a molecular weight of 28 g mole⁻¹ which gives S_area ∼ 16.2 Å². A similar calculation for carbon monoxide with molecular weight 28.01 g mol⁻¹ yields S_area to be ∼16.2 Å². Assuming condensation coefficient K_o = 1 [47], υ = 10¹⁴ Hz [49], and R = 8.314 JK⁻¹; b_o (at T ∼ 523 K) can be estimated for H₂ and CO gases using (Eq. (16)).

From the estimated values of b and b₀, heat of adsorption (Q in J mol⁻¹) can be obtained for H₂ and CO sensing using Eq. (20) [7].

$$Q_{H_2}=23.418\times {10}^3\mathrm{J}{\mathrm{mol}}^{-1}=23.418\mathrm{KJ}{\mathrm{mol}}^{-1}$$ E21

$$Q_{\mathit{CO}}=7.0138\times {10}^3\mathrm{J}{\mathrm{mol}}^{-1}=7.0138\mathrm{KJ}{\mathrm{mol}}^{-1}$$ E22

Figure 9c shows the plot of response time ( ${\tau }^{\mathit{res}}$ ) versus gas concentration (C) on a logarithmic scale. As shown in Figure 9c, variation of ${\tau }^{\mathit{res}}$ with C follows the following relation:

$${\mathrm{\tau }}^{\mathrm{res}}={\mathrm{\tau }}_0\times {\mathrm{C}}^{-\mathrm{\beta }}$$ E23

The ln ${\tau }^{\mathit{res}}$ versus ln C can be fitted linearly to estimate the rate constant and β for H₂ and CO sensing where ${\mathrm{\tau }}_0$ is the fitting constant.

The temperature variation of ${\tau }^{\mathit{res}}$ can be expressed by the following relation (47):

$${\tau }^{\mathit{res}}={\mathrm{\upsilon }}^{-1}\times exp\left(\frac{E_a+Q}{\mathit{RT}}\right)$$ E24

From Eqs. (23) and (24), the value of E_a can be written as

$$E_a=\mathit{RT}\mathit{ln}\left({\tau }_0\times C^{-\beta }\times \mathrm{\upsilon }\right)-Q$$ E25

The variation of E_a with concentration for H₂ and CO gases is shown in Figure 9d [7]. The activation energy decreases from 134.3 kJ/mol to 131.8 kJ/mol for 500 ppm H₂ to 50 ppm H₂, whereas the activation energy decreases from 151.05 kJ/mol to 150.07 kJ/mol and from 500 ppm CO to 50 ppm CO for SnO₂ thick film sensor.

Similar analyses for the ZnO thin film sensor yield activation energy values of 138.62 and 151.87 kJ/mol for 500 ppm H₂ and CO, respectively. Activation energy increases with decreasing concentration for both the gases. The Q values for H₂ and CO are 31.48 and 18.48 kJ/mol, respectively. Thus, the activation energy, E_a, and heat of adsorption, Q, may be used to differentiate H₂ and CO gases for both thin and thick films.

4.2.2 Stability of thin and thick film gas sensors

Figure 10 shows the schematic of Langmuir-Hinshelwood mechanism [8] in which the oxygen as well as reducing test gas is initially adsorbed on the sensor surface shown by “A.” An oxidized reaction product is produced as a result of collision among adsorbed oxygen and reducing test species. For efficient formation of an oxidized reaction product “RO_gas,” reducing test gas “R_gas” (having concentration “Cg”) and oxygen “O” should cross the energy barrier height shown by “TS” (transition state).

Figure 10.

Schematic representation of Langmuir-Hinshelwood mechanism.

During this investigation, the sensing element was exposed to the test gas (during response) and air (during recovery), respectively. After removal of the test gas, it is expected that the sensor should switch back to the initial base resistance in air. Considering this fact, gas sensing can be classified into reversible and irreversible type. In reversible sensing, the base resistance is fully recovered, whereas for irreversible sensing, the base resistance recovers partially. The mechanism of reversible and irreversible sensing was reported elsewhere [8]. The irreversible sensing can be better understood from the gas–solid interaction. During gas sensor response process, CO and H₂ gases are oxidized to CO₂ and H₂O. These oxidized products are entrapped in the small pores of the sensing element. If the desorption of these products does not occur fully, the base resistance of the sensing element recovers partially. Therefore, a marginal drift in the baseline resistance is observed. The mathematical formulations of irreversible and reversible sensing are described in the subsequent sections.

4.2.2.1 Irreversible sensing model

It is assumed that the site fraction θ on the sensor surface is occupied by adsorbed oxygen ions $O_{\mathit{ad}}^{-}$ and test gas (R_gas) adsorption takes place over remaining available sites. For an “n”-type metal oxide semiconductor gas sensor, reducing gas reacts with chemiadsorbed oxygen to form an oxidized product as represented by Eq. (26).

$$R_{\mathit{gas}}+O_{{\mathit{ad}}^{-}}\leftrightarrow {\mathit{RO}}_{\mathit{ad}}$$ E26

At constant operating temperature, it was assumed that the total number of active sites (Fθ) can be expressed by the following relation:

$$\theta \left(\sim \left[O^{\mathit{ad}-}\right]\right)+{\mathit{RO}}_{\mathit{ad}}\left(\mathit{\text{occupied}}\mathit{by}{R}_{\mathit{gas}}\right)=\mathit{F\theta }$$ E27

It has been assumed that adsorption of the reaction product ${\mathit{RO}}_{\mathit{ad}}$ takes place on the same sites over which reducing test gas (R_gas) adsorbs initially. Since the formation of ${\mathit{RO}}_{\mathit{ad}}$ is the rate-determining step for response process, sensor response is directly proportional to the formation of an oxidized reaction product ${\mathit{RO}}_{\mathit{ad}}$ . Therefore, the sensor response during irreversible sensing can be expressed mathematically by the following relation:

$$\frac{d\left[{\mathit{RO}}_{\mathit{ad}}\right]}{\mathit{dt}}=k\left[{O_{\mathit{ad}}}^{-}\right]C_g$$ E28

From Eq. (27), Eq. (28) can be written as

$$\frac{d\left[{\mathit{RO}}_{\mathit{ad}}\right]}{\mathit{dt}}=k\left[\mathit{F\theta }-R{O}_{\mathit{ad}}\right]\ast C_g$$ E29

Integrating Eq. (29), we obtain

$$\int \frac{d\left[R{O}_{\mathit{ad}}\right]}{\mathit{dt}}=\int k\left[\mathit{F\theta }-R{O}_{\mathit{ad}}\right]\ast C_g$$ E30

$$\int \frac{d\left[R{O}_{\mathit{ad}}\right]}{\left[\mathit{F\theta }-R{O}_{\mathit{ad}}\right]}=\int k\ast C_g\ast \mathit{dt}$$ E31

Simplifying Eq. (31), we can write

$$-\mathit{ln}\left[\mathit{F\theta }-R{O}_{\mathit{ad}}\right]=k\ast C_g\ast t+C$$ E32

where C is the constant of integration.

$$\mathit{At}t\to 0,R{O}_{\mathit{ad}}\to 0\mathit{\text{and}}C=-ln\mathit{F\theta }$$

Therefore Eq. (32) can be written as

$$-\mathit{ln}\left[\mathit{F\theta }-R{O}_{\mathit{ad}}\right]=k\ast C_g\ast t-ln\mathit{F\theta }$$ E33

Simplifying Eq. (33), we can write

$$\mathit{ln}\left[1-\frac{R{O}_{\mathit{ad}}}{\mathit{F\theta }}\right]A=-k\ast C_g\ast t$$ E34

Concentration of ${\mathit{RO}}_{\mathit{ad}}$ at time t can be expressed as

$$\left[R{O}_{\mathit{ad}}\right]\left(t\right)=\mathit{F\theta }\left[1-\mathit{exp}\left(-k{C}_{g}t\right)\right]$$ E35

When all active sites (Fθ) are adsorbed by the oxidized reaction product ( ${\mathit{RO}}_{\mathit{ad}}$ ), maximum sensor response is obtained. Therefore, the sensor response at a time t can be expressed by the following relation:

$$S\left(t\right)=S_{\mathit{max}}\left[1-\mathit{exp}\left(1-k{C}_{g}t\right)\right]$$ E36

$$S\left(t\right)=S_{\mathit{max}}\left[1-\mathit{exp}\left(-(t/{\tau }_{\mathit{\text{irrev}}}\right)\right]$$ E37

$\mathit{\text{where}}{\tau }_{\mathit{\text{irrev}}}=\frac{1}{k{C}_g}$ is termed as the characteristic response time (in seconds) during irreversible sensing.

4.2.2.2 Reversible sensing model

In this case, the equilibrium constant for the sensing reaction (Eq. (38)) is termed as K, and the forward and reverse rate constants are k and k/K. The rate of change of ${\mathit{RO}}_{\mathit{ad}}$ at time t can be written as

$$\frac{d\left[{\mathit{RO}}_{\mathit{ad}}\right]}{\mathit{dt}}=k\left[{O_{\mathit{ad}}}^{-}\right]C_g-k/K\left[R{O}_{\mathit{ad}}\right]$$ E38

Rearranging Eq. (38),

$$\frac{d\left[R{O}_{\mathit{ad}}\right]}{k\left[\mathit{F\theta }\right]C_g-k\left[R{O}_{\mathit{ad}}\right]\left(C_g+\frac{1}{K}\right)}=\mathit{dt}$$ E39

Simplifying Eq. (39), we can write

$$\frac{d\left[{\mathit{RO}}_{\mathit{ad}}\right]}{\left\{\frac{\left[\mathit{F\theta }\right]C_g}{C_g+\frac{1}{K}}-\left[R{O}_{\mathit{ad}}\right]\right\}}=k\left(C_g+\frac{1}{K}\right)\mathit{dt}$$ E40

Integrating Eq. (40), one can write

$$\int \frac{d\left[{\mathit{RO}}_{\mathit{ad}}\right]}{\left\{\frac{\left[\mathit{F\theta }\right]C_g}{C_g+\frac{1}{K}}-\left[R{O}_{\mathit{ad}}\right]\right\}}=k\left(C_g+\frac{1}{K}\right)\int \mathit{dt}$$ E41

Simplifying Eq. (41), one can write

$$-\mathit{ln}\left\{\frac{\left(\mathit{F\theta }\right)C_g}{(C_g+\frac{1}{K})}\right\}-\left[R{O}_{\mathit{ad}}\right]=k\left(C_g+\frac{1}{K}\right)t+C$$ E42

$$\mathit{\text{When}}t\to 0,{\mathit{RO}}_{\mathit{ad}}\to 0,\mathit{\text{and}}C=-\mathit{ln}\left\{\frac{\left(\mathit{F\theta }\right)C_g}{(C_g+\frac{1}{K})}\right\}$$ E43

Therefore, Eq. (43) can be simplified as

$$-\mathit{ln}\left\{\frac{\left(\mathit{F\theta }\right)C_g}{(C_g+\frac{1}{K})}\right\}-\left[R{O}_{\mathit{ad}}\right]=k\left(C_g+\frac{1}{K}\right)t-\mathit{ln}\left\{\frac{\left(\mathit{F\theta }\right)C_g}{(C_g+\frac{1}{K})}\right\}$$ E44

Eq. (44) can be simplified to

$$\mathit{ln}\left[\frac{\frac{\left(\mathit{F\theta }\right)C_g}{\left(C_g+\frac{1}{K}\right)}-R{O}_{\mathit{ad}}}{\frac{\left(\mathit{F\theta }\right)C_g}{\left(C_g+\frac{1}{K}\right)}}\right]A=-k\left(C_g+\frac{1}{K}\right)t$$ E45

Eq. (45) can be further simplified to

$$1-\frac{\left[{\mathit{RO}}_{\mathit{ad}}\right]}{\frac{\left(\mathit{F\theta }\right)C_g}{\left(C_g+\frac{1}{K}\right)}}=\mathit{exp}\left(-k\left(C_g+\frac{1}{K}\right)t\right)$$ E46

The rate of change of ${\mathit{RO}}_{\mathit{ad}}$ at time t can be written as

$$\left[R{O}_{\mathit{ad}}\right]\left(t\right)=\left\{\frac{\left(\mathit{F\theta }\right)C_g}{\left(C_g+\frac{1}{K}\right)}\right\}\left\{1-\mathit{exp}\left(-k\left(C_g+\frac{1}{K}\right)t\right)\right\}$$ E47

$$\left[R{O}_{\mathit{ad}}\right]\left(t\right)=\left\{\frac{\left(\mathit{F\theta }\right)C_{g}K}{\left(1+C_{g}K\right)}\right\}\left\{1-\mathit{exp}-\left\{\left(1+C_{g}K\right)/K\right\}\mathit{kt}\right\}$$ E48

Sensor response (S(t)) can be written as

$$S\left(t\right)=S_{\mathit{max}}\left\{\frac{C_{g}K}{\left(1+C_{g}K\right)}\right\}\left\{1-\mathit{exp}-\left\{\left(1+C_{g}K\right)/K\right\}\mathit{kt}\right\}$$ E49

$$\mathit{\text{Assuming}}1+C_{g}K\approx C_{g}K,\mathit{one}\mathit{can}\mathit{\text{write}}$$

$$S\left(t\right)=S_{\mathit{max}}[1-\mathit{exp}\left(-t/{\tau }_{\mathit{rev}}\right)]$$ E50

where τ_rev is the characteristic response time during reversible sensing.

Figure 11a,b shows the concentration-dependent resistance transients for H₂ and CO sensing for ZnO thin film. These measurements were taken at 300°C by varying the test gas concentration. During these measurements, drift in baseline resistance was observed as shown by the dashed line in Figure 11a,b. Therefore, time-dependent response % for H₂ and CO sensing was fitted using irreversible sensing model as given in Eq. (37). Figure 11c shows the typical response % vs. time fitting for 500 ppm H₂ and CO sensing. Inset of Figure 11c represents variation of ${\tau }_{\mathit{\text{irrev}}}$ (in seconds) with concentration (in ppm) in which ${\tau }_{\mathit{\text{irrev}}}$ decreases systematically with increase in test gas concentration which is a typical behavior in case of irreversible gas sensing as reported by [8].

Figure 11.

Concentration-dependent resistance transients recorded at 300°C for (a) H₂, (b) CO, (c) time-dependent response % fitting for 500 ppm H₂ sensing for ZnO thin film. Experimental points (symbols) and the solid line were fitted using Eq. (37). Inset shows the variation of τ_irrev with concentration.

We have also investigated the reversible sensing behavior of plasma sprayed SnO₂ thick film sensor [7]. The SnO₂ thick film sensor shows marginal drift in baseline resistance when exposed to 20–500 ppm CO and H₂. This is shown in Figure 12.

Figure 12.

Concentration-dependent resistance transients recorded at 250°C for (a) H2, (b) CO, and (c) time-dependent response % fitting for 500 ppm H2 sensing. Experimental points (symbols) and the solid line were fitted using Eq. (50). Inset shows the variation of τ_rev with concentration (Figure 12a–c has been reprinted from [7] with permission from Elsevier).

Inset of Figure 12c shows the variation of characteristic response time τ_rev (s) with concentration (ppm). τ_rev was almost invariant of the test gas concentration in the range 500–200 ppm. Interestingly, τ_rev was also invariant of test gas concentration in the range 100–20 ppm. However, the value of τ_rev was found to be higher in the lower concentration range (100–20 ppm) than that of the higher concentration range (500–200 ppm).

Figure 13 shows the surface micrograph for (a) ZnO thin film and (b) SnO₂ thick film sensing elements. ZnO thin film exhibits denser surface morphology than SnO₂ thick film as revealed from Figure 13a. In line with this discussion, irreversible sensing behavior for ZnO thin film can be attributed to denser surface morphology. It may be noted here that our results for ZnO thin film are in accordance with [8]. On the other hand, reversible sensing behavior in case of SnO₂ thick film can be attributed to its porous surface morphology as revealed from Figure 13b.

Figure 13.

Surface micrograph: (a) ZnO thin film and (b) SnO₂ thick film.

4.3 CO₂ sensing characteristics

Carbon dioxide concentrations have increased enormously in the atmosphere in the past hundred years or so. In the preindustrial era, CO₂ concentration was ∼280 ppm. However, current global standards put CO₂ concentrations at an alarming level of 400 ppm mainly arising from deforestation and the burning of fossil fuels. India alone emitted 2299 million tons of CO₂ in 2018 according to an International Energy Agency (IEA) report [50]. Most of India’s emissions come from the burning of coal. Increased CO₂ emissions have led to global warming—a climate phenomenon that is responsible for the increasing number of storms as well as droughts [51]. In addition, increased levels of CO₂ in the atmosphere affect human health severely. Table 4 summarizes the effects on human health with increasing CO₂ concentrations. It is thus incumbent that we take necessary steps to detect CO₂ concentrations in the atmosphere.

Concentration of CO2 (ppm)	Effect on human health
250–300	Normal outdoor air
350–1000	Indoor air
100–2000	Drowsiness (poor air)
2000–5000	Increased heart rate, nausea, headaches
>5000	Oxygen deprivation begins
>40,000	Serious oxygen deprivation leading to brain damage, coma, and death

Table 4.

Effect on human health with increasing concentrations of CO₂.

Research on CO₂ sensing gained prominence in the late 1980s when researchers such as Toshio Maruyama [52] and Yasuhiro Shimizu [53] reported a NASICON (Na₃Zr₂Si₂PO₁₂)-based solid electrolyte gas sensor and an electrochemical K₂CO₃-polyethylene glycol CO₂ sensor, respectively. However, it was soon realized that handling these materials was cumbersome and incorporating these into actual devices posed a serious challenge. Research was thus shifted to resistive-type sensors [54, 55] which were far better in terms of stability and could be easily miniaturized to accommodate in sensing devices. However, such sensors were plagued with poor selectivity and low response to CO₂. This prompted the use of mixed oxide or composite sensors [56, 57, 58], the research potential of which is still being investigated [59, 60].

A comprehensive study of literature (elaborated in Table 5) related to CO₂ sensing leads us to the fact that formation of p-n junctions is an important criterion for CO₂ detection. We have reported CO₂ sensing of ∼36% achieved using a LFCO-ZnO composite thin film.

Sensor material	Form	Gas	Topt (°C)	S%	Selectivity	τres, τrec (s)	Conc. (ppm)	Ref.
ZHS microcubes on ZnO nanorods	Thin film	CO2	150	354	No	10, 24	2500	[61]
LaBaCo2O5 + δ	Bulk	CO2	300	1.04 (Rg/Ra)	No	—	400	[62]
Graphene sheet	Thin film	CO2	60	26	No	8, 10	100	[63]
BaCO3-Co3O4	Bulk	CO2	150	15	No	227, 245	10,000	[59]
Ag/BaTiO3-CuO	Thin film	CO2	250	28	No	900, 600	5000	[64]
Ca/Al-doped ZnO	Bulk	CO2	300	15	No	—	2500	[65]
LaFeO3-SnO2	Bulk	CO2	250	3 (Rg/Ra)	No	<20	4000	[60]
Ca-ZnO	Thin film	CO2	350	53	Yes	—	50,000	[66]
LFCO-ZnO	Thin film	CO2	300	36	Yes	300, 630	2500	This work

Table 5.

Overview of the carbon dioxide sensing characteristics of various thin and bulk materials from literature.

CO₂ sensing characteristics of LFCO-ZnO composite thin film was measured in the temperature range 350–150°C. The highest response of ∼36% was obtained at 300°C to 2500 ppm CO₂, and the response decreases to 12% at 150°C. The response and recovery times recorded at 300°C are 300 and 630 s, respectively. The composite shows n-type sensing behavior which confirms the dominance of ZnO crystallites [67]. The LFCO-ZnO thin film composite is cross-sensitive to CO₂. A response of 65% was recorded for 500 ppm CO at 300°C. The composite shows higher response to both CO and CO₂ than pure LFCO and ZnO thin films. To address the cross-sensitivity of the composite thin film, principal component analysis of the resistance transients was performed. The relevant features of the resistance transients of the test gases are extracted using fast Fourier transform technique. These features are then used as input parameters for PCA. Figure 14d shows low dispersion of CO and CO₂ for the LFCO-ZnO composite thin film after performing PCA. Thus, FFT-PCA is an efficient tool for the differentiation of CO and CO₂ gases for LFCO-ZnO thin film.

Figure 14.

Temperature variation of response % for pure LFCO and pure ZnO to (a) 500 ppm CO and (b) 2500 ppm CO₂. (c) Temperature variation of response % of 0.5% LFCO-ZnO for 500 ppm CO and 2500 ppm CO₂. (d) Principal component analyses for CO and CO₂ at 300°C.

The LFCO-ZnO composite follows the gas sensing behavior of an n-type semiconductor due to the presence of a large number of ZnO crystallites. Therefore, analyses of the conductance transients of CO and CO₂ gases follow Langmuir adsorption mechanism similar to ZnO thin film sensor. Activation energy, E_a, and heat of adsorption, Q, have been calculated in accordance with the theory described in the previous section. Figure 15d represents the variation of E_a with concentration for CO and CO₂ gases. It should be noted that each gas possesses distinct activation energy values. The activation energy decreases from 153.83 kJ/mol to 149.67 kJ/mol for 500 ppm CO to 50 ppm CO, whereas the activation energy decreases from 155.93 kJ/mol to 155.28 kJ/mol for 2500 ppm CO₂ to 1000 ppm CO₂. The variation in E_a is much less in CO₂ than in CO which could pertain to the efficiency of CO₂ sensing of 0.5% LFCO-ZnO thin film.

Figure 15.

(a) Conductance transient plot for 500 ppm CO sensing. Experimental points are shown by symbols, and the solid line was fitted using Eq. (18), (b) linear variation of G1/(1-G1) with concentration, (c) log–log plot of τ^res with C. experimental points are shown by symbols, and the solid line was fitted using Eq. (23), (d) variation of E_a with ln C for CO and CO₂ sensing. For all these measurements, operating temperature was kept fixed at 300°C.

As discussed earlier, when the sensor surface is exposed to air, oxygen ions adsorb on the sensor surface which leads to the formation of an electron-depleted layer and hence a potential barrier between two grains. The conductance G at temperature T is given by Eq. (12).

Now, assuming Langmuir adsorption (ideal gas), when sensor surface is exposed to gas, it is covered with a monolayer of gas molecules/atoms. The conductance transients are fitted according to Eq. (18) for the response transient. The recovery transient, however, represents decay of conductance, and it is fitted by the following equation:

$$G{\left(t\right)}_{\mathit{\text{recovery}}}=G_0^{\prime }+G\left[\mathit{exp}\left(\frac{-t}{{\tau }_{\mathit{\text{recovery}}}}\right)\right]$$ E51

Figure 16 shows the conductance transients of CO and CO₂ fitted with Eqs. (18) and (51). Table 6 summarizes the fitted parameters. Each parameter is unique to both the gases. It may be noted that the response times to both CO and CO₂ are almost similar (∼60 and 68 s, respectively) although the response % is markedly different (64 and 36%, respectively).

Figure 16.

Fitted conductance transients of (a) CO and (b) CO₂ at 300°C in accordance with Eqs. (18) and (51).

Fitted response parameters	Topt	G0 (Ω^−1)	G (Ω^−1)	τres (s)
CO	300 °C	3.45165E-9	7.60875E-9	59.74
CO2	300 °C	2.69149E-9	2.34245E-9	303.00
Fitted recovery parameters	Topt	G0 (Ω^−1)	G (Ω^−1)	τrec (s)
CO	300°C	4.09001E-9	7.6711E-9	68.56
CO2	300°C	2.95542E-9	1.57362E-9	220.29

Table 6.

Fitted parameters for the conductance transients of CO and CO₂ at 300°C.

Assuming nonlinear variation of sheet conductance with test gas concentration, the gas sensing response varies with absolute temperature T according to the following relation:

$$\frac{R_a-R_g}{R_a}=\frac{a_0}{6}\times \mathit{exp}\left(-\frac{E_a}{\mathit{RT}}\right)\left[6+n.m_0^2\mathit{exp}\left(-\frac{E_k}{\mathit{RT}}\right)T^{-0.5}\right]\times {\left[\frac{C_{\mathit{AS}}}{cosh\left(m_0.\mathit{exp}\left(-\frac{E_K}{2\mathit{RT}}\right)T^{-0.25}\right)}\right]}^n$$ E52

where $R_a$ is the resistance in air, $R_g$ is the resistance in gas, $a_0\mathit{\text{is the}}$ pre-exponential constant, $n$ is the sensitivity of the gas, $m_0$ is the Hatta number, $E_k$ is the activation energy of the first order reaction, and $C_{\mathit{AS}}$ is the test gas concentration at the film surface [6].

The response (S) versus absolute temperature (T) follows a “bell-shaped” dynamic and can be fitted in accordance with Eq. (52) to yield parameters specific to CO and CO₂. Figure 17 shows the S versus T fitting for CO and CO₂. Sensitivity of the thin film, “n,” is higher for CO₂ which shows that the sensor surface is more active toward CO₂ adsorption. Addition of 0.5% LFCO to ZnO structure enhances oxygen adsorption on the ZnO surface due to the formation of additional defects. Also, the formation of multiple junctions in the composite leads to the phenomenon of “band bending” which enhances the interaction of CO₂ with oxygen ions. It may be noted that CO interferes with CO₂ sensing in the composite due to the formation of multiple junctions. This problem is eliminated in the ZnO/LFCO bilayer (not shown) which consists of only a single p-n junction and, thus, is selective to CO₂.

Figure 17.

Estimated value of sensitivity “n” for (a) CO and (b) CO₂ from the fitting of response vs. temperature using Eq. (52).

Simulated parameters from Figure 17 are summarized in Table 7 where E_a and E_k have units of J mol⁻¹. Significant differences may be noted for the simulated parameters of each gas. Table 7 and Figure 15d corroborate the fact that the activation energy is lower for CO which explains the higher response to CO. It may also be emphasized that the parameters are valid for the entire range of operations of the sensor. The simulation may provide useful insights regarding the maximum and minimum temperature at which the sensor is operative. Further, a logarithmic plot of sensor response S and concentration C yields a value of ∼0.67 for CO and ∼1.56 for CO₂ (Figure 18) which is well in agreement with the simulated values. Additionally, it is shown that sensor response decreases sequentially with decreasing concentration as can be seen from Figure 18. The limiting concentration for CO is 50 ppm and that for CO₂ is 500 ppm. The detection limit for CO₂ is well above the permissible exposure limit (PEL) which is >5000 ppm. Thus, we may conclude that 0.5% LFCO-ZnO composite thin film is a fair CO₂ sensor and is well suited for practical applications.

Fitting parameters	Carbon monoxide	Carbon dioxide
Ek/R	3611.87	15674.17511
a0/6	3278.17	1.22219E-5
Ea/R	6062.32	2007.15851
m0 = L(√(k/Dk)	1240.78	2.9675E6
n (sensitivity)	0.5654	1.56
CAS (ppm)	500	2500

Table 7.

Fitted parameters from the fitting of response versus temperature of CO and CO₂.

Figure 18.

Concentration variation in 0.5% LFCO-ZnO for (a) CO (500–50 ppm) and (b) CO₂ (2500–500 ppm); inset in both figures shows logarithmic variation of response with concentration which can be fitted linearly to obtain “n,” the sensitivity.

4.3.1 CO₂ sensing mechanism of LFCO-ZnO thin film gas sensor

The electrical resistance change in ZnO, which is an n-type conducting material, is attributed to adsorption and desorption kinetics of oxygen on ZnO surface. Chemisorption of molecular oxygen of the type ${\mathit{O}}_2^{\mathit{-}}$ occurs above room temperature on ZnO surface, leading to the formation of electron depletion layer (EDL). As the temperature increases up to ∼300°C, the molecular oxygen dissociates into atomic form and forms a thicker EDL which pushes inside the ZnO core. Homogeneous particles observed in ZnO thin film microstructure (Figure 13a) contribute to the increased rate of oxygen adsorption on ZnO surface. This increases the sensitivity toward CO₂ gas as can be observed from Figure 14b. Reaction of carbon dioxide molecules with adsorbed oxygen is given below [68]:

$${\mathit{CO}}_2+{\mathit{e}}^{-}\to {\mathit{CO}}_{2\left(\mathit{ads}\right)}^{-}$$ E53

$${\mathit{CO}}_{2\left(\mathit{ads}\right)}^{-}+{\mathit{O}}_{\left(\mathit{ads}\right)}^{-}+2{\mathit{e}}^{-}\to 2\mathit{CO}+2{\mathit{O}}_2^{-}$$ E54

These equations are consistent with the reduction in resistance of ZnO sensor on exposure to CO₂. In principle, ZnO sensor should register an increase in resistance on exposure to CO₂ as it is an oxidizing gas. However, it was found that ZnO demonstrates typical n-type behavior when brought in contact with CO₂. This anomaly can be explained from the defect chemistry and catalytic properties of ZnO. This demands a more detailed study and is currently beyond the scope of this book. Thus, we have confined ourselves to the current discussion of LFCO-ZnO composite thin film sensor.

When n-type ZnO is combined with p-type LFCO (LaFe_0.8Co_0.2O₃) sensor, a p-n junction is created at the interface between the two sensors. On exposure to CO₂, the junction pushes further in to the ZnO core. This leads to an enhanced CO₂ sensing response as compared to pristine LFCO and ZnO sensors.

The general reaction mechanism of LaFeO₃ (LFO) was reported by [69] in which LFO reacts with CO₂ in the presence of moisture to form carbonates. The reaction is given below.

$${\mathit{H}}_2\mathit{O}+\frac{1}{2}{\mathit{O}}_2+2{\mathit{e}}^{-}\to 2{\mathit{OH}}^{-}$$ E55

$${\mathit{O}}_2+{\mathit{e}}^{-}\to {\mathit{O}}_2^{-}$$ E56

$${\mathit{O}}_2^{-}+{\mathit{e}}^{-}\to 2{\mathit{O}}^{-}$$ E57

$${\mathit{La}}^{3+}+3{\mathit{OH}}^{-}+3{\mathit{CO}}_2\to \mathit{La}{\left({\mathit{CO}}_2\mathit{OH}\right)}_3$$ E58

The carbonates thus formed aid in CO₂ sensing by reducing the activation energy required for the reaction to occur. Further, equalization of Fermi levels takes place via migration of carriers leading to the formation of p-n junction. On exposure to CO₂, hydroxycarbonates form at the junction interface leading to the increase of potential barrier. This increases the concentration of holes at LFCO interface. However, as the concentration of LFCO is only 0.5%, hole accumulation at LFCO surface is impeded, and fall in resistance occurs, and the composite as a whole shows n-type sensing behavior [60].

SnO₂ thick film gas sensor did not show any appreciable CO₂ sensing. This is due to the absence of a Schottky contact in between SnO₂ thick film and silver electrodes. Investigations on CO₂ sensing were not performed for any composite thick film sensor.

4.4 VOC sensing characteristics of thin and thick film

Volatile organic compounds such as ethanol, acetone, and isopropanol pollute auto cabins and indoors of buildings. Long time spent indoors in such an environment ultimately leads to sick building syndrome and also has been found to be linked to leukemia and lymphoma. VOCs are also biomarkers of lung cancer and type 2 diabetes. Ethanol, acetone, and isopropanol have a permissible exposure limit of 1000 ppm over an 8 h work shift as documented by OSHA.

Figure 19 demonstrates the necessity of VOC detection. Traditionally, n-type semiconducting metal oxides such as SnO₂, ZnO, WO₃, TiO₂, In₂O₃, Fe₂O₃, etc. have been used for VOC sensing. However, recent research activities show that p-type semiconducting metal oxides such as copper oxide have been gaining popularity in gas sensing mainly due to its ease of fabrication, low cost, high stability, and high response to volatile organic compounds. CuO architectures such as nanorods [70], nanosheets [71], nanoparticles [72], and thin film [19] play an important role in gas sensing behavior. Thin films, in particular, are preferred over powdered CuO variants due to high stability and ease of device fabrication (Table 8). Additionally, composites have also gained prominence in selective VOC sensing. We have synthesized CuO thin films using solgel technique. Quartz substrate was used to deposit the thin films using spin coating. After heat treatment, gold electrodes were deposited on these films using sputter coating following which the gas sensing measurements were performed in a static gas sensing setup. The VOCs were found to be cross-sensitive. Cross-sensitivity was addressed with principal component analysis. Additionally, we have also investigated the VOC sensing characteristics of WO₃-SnO₂ thick film. Investigations into various thick films for VOC sensing has been tabulated in Table 9. The gas–solid interaction in this case was modeled with the Freundlich adsorption isotherm which is nonlinear in nature.

Figure 19.

Permissible exposure limit and harmful effects of volatile organic compounds.

Sensor material	Geometry	Test gas	Topt (°C)	S%	τres, τrec (s)	Mixed gas sensing	Conc. (ppm)	Ref.
CuO	Nanowire	n-Propanol	190	6.2	1.2, 6.6	No	100	[76]
CuO	Nanosheets	Ethanol	320	4.22	2–15, 5–24	No	500	[73]
CuO	Thin film	2-Propanol	225	282	12, 279	No	300	[19]
CuO	Nanorods	Ethanol	350	90	—	No	600	[70]
CuO	Nanoparticles	Ethanol	200	5.6	10, 8	No	500	[74]
CuO	Powder	Ethanol	230	2.37 (Rg/Ra)	—	No	210	[75, 76]
CuO	Nanoparticles	Ethanol	220	2.7 (Rg/Ra)	11, 7	No	0.1	[72]
CuO	Thin film	Ethanol; Acetone	300	144; 168	3, 216; 7, 248	Yes	300	This work

Table 8.

Comparison of VOC sensing performances of CuO powders and thin films reported in recent literature.

The CuO sensors, thus synthesized, were used for detection of ethanol and acetone by varying the operating temperature (200–300°C) and gas concentration (25–300 ppm). It was found that CuO shows high response and excellent stability with respect to both these gases. The response to 300 ppm acetone and ethanol was found to be 168 and 144%, respectively (Figure 12). The response and recovery times were found to be 7 and 248 s for acetone and 3 and 216 s for ethanol, respectively. The static gas sensing setup is similar to a real environment response which explains the delayed recovery times. The response times mentioned are not exact as in a static gas sensing setup, where response to the test gas does not saturate before recovery begins [19]. Mixed gas sensing was also performed with ethanol and acetone (300 ppm/300 ppm) at 300°C. It was found that the response was not affected by the order in which the gases were injected into the static chamber. Principal component analysis was used as a discriminatory method to quantify the gases for mixed gas sensing data (Table 9).

Sensor material	Geometry	Gas	Topt (°C)	S%	τres, τrec (s)	Selectivity	Conc. (ppm)	Ref.
CeOx - SnO2	Thick film	Ethanol	350	95	1.1	Yes	200	[77]
Mg0.5Zn0.5Fe2O4	Nanoparticles	Ethanol	325	50	—	No	200	[78]
TiO2	Nanoparticles	Ethanol	375	65	—	No	200	[79]
Barium hexaferrite	Nanoparticles	Ethanol	375	90	—	Yes	200	[80]
Sr-SnO2	Thick film	Ethanol	300	65	2, 7		100	[81]
SnO2	Thick film	Ethanol	300	84	8, 340	Yes	300	[20]
WO3-SnO2	Thick film	Ethanol	250	73	10, 312	Yes	300	[15]

Table 9.

Comparison of ethanol sensing performances of nanopowders and thick films reported in recent literatures.

4.4.1 Ethanol and acetone sensing mechanism in CuO sensor

CuO is a p-type sensor, and as such the sensing mechanism of a p-type sensor differs slightly to that of an n-type sensor. The major carriers in this case are holes. The concentration of holes changes at the sensor surface due to reaction with the target gas. When the CuO sensor is exposed to air at the operating temperature range of 200–300°C, oxygen takes up electrons from the sensor surface, and the oxygen ions formed chemisorb onto the surface. This increases the density of holes resulting in the decrease of the surface Fermi level. A hole accumulation layer (HAL) at the surface increases the conduction of holes which reduces the baseline resistance of CuO (R_a = resistance in air ambient). The equations involved are listed below [19].

$$O_2\left(\mathit{gas}\right)\to O_2\left(\mathit{ads}\right)$$ E59

$$O_2\left(\mathit{ads}\right)+e^{-}\to O_{\mathit{ads}}^{-}+h$$ E60

$$O_{\mathit{ads}}^{-}+e^{-}\to O_{\mathit{ads}}^{2-}+h$$ E61

On exposure to acetone and ethanol, the oxygen ions react with the test gases to produce carbon dioxide and water with the release of electrons. This process reduces the thickness of the hole accumulation layer which increases the sensor resistance (R_g = resistance in gas ambient) [75].

The probable reaction mechanisms for acetone and ethanol are

$$8O_{\mathit{ads}}^{2-}+{\mathit{CH}}_3\mathit{CO}{\mathit{CH}}_3\to 16e^{-}+3{\mathit{CO}}_2+3H_{2}O$$ E62

$$6O_{\mathit{ads}}^{2-}+C_2{H}_5\mathit{OH}\to {\mathit{CH}}_3\mathit{CHO}+H_{2}O+e^{-}$$ E63

$$e^{-}+h\to \mathit{\text{null}}$$ E64

In the end, electron hole recombination takes place and the HAL diminishes.

4.4.2 Mixed gas sensing in CuO thin film sensor and principal component analyses

We have also investigated mixed gas sensing for 300 ppm acetone and 300 ppm ethanol. We found that the mixed gas sensing response (205%) is much higher than individual sensing response of acetone (168%) and ethanol (144%), but it was not an addition of the individual responses (Figures 20 and 21). Further, it was found that the order in which the gases were injected into the static chamber had no bearing on the mixed gas sensing response.

Figure 20.

Resistance transients showing the response % of (a) ethanol and (b) acetone.

Figure 21.

Resistance transients showing the response % of (a) 300 ppm ethanol and 300 ppm acetone and (b) 300 ppm acetone and 300 ppm ethanol.

This can be attributed to the fact that the gas with the higher reactivity will first interact with the adsorbed O⁻ ions, leaving the next gas to react with less number of the same ions leading to reduced sensing [77].

The gases can be discriminated by principal component analyses, which quantify the gases in a multi coordinate system. Fast Fourier transform was utilized to extract the important features of the resistance transients of both acetone and ethanol measured at 300°C. These parameters were then fed into a linear unsupervised principal component analyses pattern recognition technique. Figure 22a shows the clustering of ethanol and acetone when injected individually. The clustering does not change when all the resistance transients of 300 ppm each of acetone and ethanol are analyzed (Figure 22b). Thus, PCA can be successfully used to identify the gases in a mixture of acetone and ethanol.

Figure 22.

Principal component analyses for (a) individual gases and (b) mixed gases; note the separate clustering enabling us to distinguish the gases.

Additionally, we have explored the VOC sensing characteristics of WO₃-SnO₂ composite thick film sensor [15]. Sensing measurements were made in the range 300–150°C for 300 ppm ethanol, acetone, and isopropanol. The composite thick film was found to be cross-sensitive to each of the gases. To address the cross-sensitivity, the conductance transients of the gases were further analyzed. These conductance transients were obtained at 250°C, and VOC concentration was kept fixed at 300 ppm. Conductance transients for the three VOCs at respective concentration level were fitted using Eq. (18) to estimate the values of parameters such as G₀ (mho), G₁ (mho), and τ^res (s). Figure 23b envisages the nonlinear variation of G1/(1-G1) with VOC concentration (C) (symbols). In this case, Langmuir adsorption isotherm is not validated since the variation of G1/(1-G1) with concentration is nonlinear for all the gases and does not pass through the origin. In fact, the results are much closer to the Freundlich adsorption isotherm. The relation between G₁ and C in this case obeys the following relation:

Figure 23.

(a) Conductance transient plots measured at 250°C for 300 ppm ethanol, acetone, and isopropanol sensing. Experimental points are shown by symbols, and the solid line was fitted using Eq. (18), and (b) nonlinear variation of G1/(1-G1) with concentration (ppm). Experimental points are shown by symbols, and the solid line was fitted using Eq. (65) (Figure 23a,b reprinted from [15] with permission from Elsevier).

$$\frac{{\mathrm{G}}_1}{1-{\mathrm{G}}_1}=\mathrm{\alpha }{\mathrm{C}}^{\mathrm{\delta }}$$ E65

From the nonlinear fit, we estimated the values of α and δ for all VOCs under consideration. These results indicate that the sensor surface is not homogeneous and heat of adsorption (Q) is not uniform unlike the Langmuir model.

In case of Langmuir adsorption isotherm, assuming Z = G₁/(1-G₁), differentiating Eq. (19) we can write

$$\frac{\mathrm{dZ}}{\mathrm{dC}}=\frac{1}{b}$$ E66

Differentiating Eq. (66), we can write

$$\frac{\mathrm{dZ}}{\mathrm{dC}}=\left(\mathrm{\alpha \delta }\right){\mathrm{C}}^{\mathrm{\delta }-1}$$ E67

Assuming Z vs. C slope at a particular point ∼1/b, we can write

$$\frac{1}{b}=\left(\mathit{\alpha \delta }\right)C^{\delta -1}$$ E68

$$\mathrm{b}=\frac{1}{\left(\mathrm{\alpha \delta }\right){\mathrm{C}}^{\mathrm{\delta }-1}}={\mathrm{b}}_{0}exp\left(-\frac{\mathrm{Q}}{\mathrm{RT}}\right)$$ E69

Simplification of Eq. (69) yields

$$\mathrm{Q}=\mathrm{RT}\left[{\mathrm{lnb}}_{\mathrm{o}}–\left(1-\mathrm{\delta }\right)\mathrm{lnC}+ln\mathrm{\alpha \delta }\right]$$ E70

We have estimated b_o from the expression given in Eq. (16). Assuming υ = 10¹³ Hz [49], R = 8.314 JK⁻¹, and T = 523 K, Eq. (16) can be simplified to.

$${\mathrm{b}}_0=2.7443\times {10}^{-9}\left(\frac{1}{{\mathrm{K}}_0.{\mathrm{S}}_{\text{area}}}\right)\times {\mathrm{M}}^{1/2}\mathrm{Pa}$$ E71

As described earlier, considering the surface area, S_area ∼ 16.2 Å² for nitrogen (molecular weight = 28 g mol⁻¹), we can easily calculate the surface area for ethanol which is 26.61 Å² (molecular weight = 46.0 g mol⁻¹). Assuming condensation coefficient K_o = 1 [47], for all test gases, b_o (at T ∼ 523 K) can be estimated using Eq. (71). From Eq. (70) we have estimated Q as a function of VOC concentration at 250°C.

Figure 24a shows the log–log plot of characteristic response time (τ^res) vs. VOC concentration. As shown in Figure 24a, variation of τ^res with C follows the following relation:

Figure 24.

(a) Log-log plot of τ^res with C. The sensing measurements were carried out at 250°C, and (b) variation of E_a and Q with concentration for ethanol, acetone, and isopropanol sensing (Figure 24a,b reprinted from [15] with permission from Elsevier).

$${\mathrm{\tau }}^{\mathrm{res}}={\mathrm{\tau }}_0\times {\mathrm{C}}^{-\mathrm{\beta }}$$ E72

The ln (τ^res) vs. ln (C) plot was fitted linearly in order to estimate value of β for three VOCs under consideration.

In case of Freundlich adsorption isotherm, temperature-dependent variation of τ^res can be written as per the following relation:

$${\tau }^{\mathit{res}}={\mathrm{\upsilon }}^{-1}\times exp\left(\frac{E_a+Q}{\mathit{RT}}\right)$$ E73

Using Eqs. (70) and (72), we can easily simplify Eq. (73). Activation energy, $E_a,$ is expressed via the following relation:

$${\mathrm{E}}_{\mathrm{a}}=\mathrm{RT}\times ln\left(\frac{{\mathrm{\tau }}_0.\mathrm{\upsilon }.{\mathrm{C}}^{\left(1-\mathrm{\delta }-\mathrm{\beta }\right)}}{\mathrm{\alpha }.\mathrm{\delta }.{\mathrm{b}}_0}\right)$$ E74

From Eqs. (70) and (74), we can determine the linearity of Q versus ln C plot as well as the $E_a$ versus ln C plot. Also, we can see that Q decreases with gas concentration, whereas $E_a$ shows the exact opposite behavior. Figure 24b shows the variation of E_a and Q for each VOC concentration (T = 250°C) [15]. From our analysis, we have estimated that $E_a$ for ethanol sensing is lower than that of acetone and isopropanol which justify the larger yield of the composite sensor in terms of ethanol sensing ( $S\%\to \text{ethanol}>\text{acetone}>\text{isopropanol}$ ).

4.4.3 Reproducibility of CuO thin film and WO₃-SnO₂ thick film

The stability and reproducibility of CuO thin film and WO₃-SnO₂ thick film composite sensors were studied with respect to acetone and ethanol gases (Figure 25). Excellent baseline recovery was achieved for both the sensors for an extended period of time with marginal change of initial response %. The response % for CuO sensor was recorded for a period of ∼150 days, whereas that for WO₃-SnO₂ sensor was recorded for ∼500 days. Stable baseline was recorded for CuO thin film sensor, whereas WO₃-SnO₂ thick film sensor was prone to drift in baseline resistance.

Figure 25.

Typical resistance transient plots showing reproducibility toward 300 ppm (a) acetone sensing in CuO thin film and (b) ethanol sensing in WO₃-SnO₂ thick film; inset of each figure shows invariance of response % with time (Figure 25b reprinted from [15] with permission from Elsevier).

5. Conclusions

This chapter throws light on detection of common air pollutants including carbon monoxide, carbon dioxide, nitrogen dioxide, volatile organic compounds, and also hydrogen. The permissible exposure limits for these gases act as benchmark for minimum concentration of gas detection. Gas sensing terminologies such as response %, response time, recovery time, selectivity, and stability have been defined. The interaction of semiconducting metal oxide sensor with reducing gas has been illustrated. Factors that affect the performance of a gas sensor, namely, receptor, transducer, and utility functions, have been explained. It was reported that the receptor function is influenced by the grain size of the sensor. The transducer function controls the flow of charge carriers between the electrodes. It is highly dependent on purity of the sensor material. The utility function controls the diffusion processes of gas during response and recovery processes.

A modified wet chemical route was chosen as the method to synthesize sols required for thin film deposition process. The sols were deposited on quartz and glass substrates and annealed at a high temperature. The materials (ZnO-G, ZnO, SnO₂, LFCO-ZnO, CuO, and WO₃-SnO₂) for gas sensors were chosen after extensive literature review and earlier experiments conducted by us in our laboratory.

High response (894%) to 5 ppm NO₂ sensing was reported for ZnO-G multilayer as compared to pure ZnO (26%) at 150°C. Also, NO₂ detection is selective at 150°C with respect to 500 ppm each of CO, H₂, and i-C₄H₁₀ gases. The high response and selectivity have been attributed to the presence of p-n junction between ZnO and G grains. The formation of a p-n junction increases the amount of available electrons increasing the gas response. In addition, mixed gas sensing for NO₂/CO, NO₂/H₂, and NO₂/i-C₄H₁₀ has also been reported. Principal component analyses have been demonstrated as an effective tool for the quantification of these mixed gases. ZnO thin film and SnO₂ thick film were investigated for CO and H₂ detection. Both materials demonstrated cross-sensitivity to the gases. Cross-sensitivity was addressed through modeling based on Langmuir adsorption mechanism. Activation energy, E_a, and heat of adsorption, Q, were calculated based on the model. E_a and Q values were found to be distinct for each gas and varied as a function of gas concentration. Further, the irreversible nature of ZnO thin film and reversible nature of SnO₂ thick film were demonstrated when exposed to a fixed concentration of CO and H₂ gases. CO₂ sensing characteristics of LFCO-ZnO thin film sensor were investigated. The thin film sensor yields 36% response to 2500 ppm CO₂ at 300°C. LFCO-ZnO thin film is cross-sensitive to CO at this temperature. Principal component analyses have been performed to eliminate cross-sensitivity. Further, Langmuir adsorption mechanism is validated in LFCO-ZnO thin film sensor as the variation of conductance and concentration is a straight line passing through origin (not shown). A model has also been developed for the thin film gas sensor relating response with temperature. Modeling of the response with temperature for CO and CO₂ sensing yielded parameters unique to both gases. Volatile organic compounds have been detected with a CuO thin film sensor. Both acetone and ethanol show high response with a response time of only a few seconds. The gases have been quantified using principal component analyses. Mixed gas (ethanol/acetone) sensing has also been reported for CuO thin film. The mixed gases have been identified on the basis of the PCA cluster for individual gases. WO₃-SnO₂ thick film sensor was investigated for ethanol, acetone, and isopropanol sensing characteristics. The thick film sensor yielded highest response to 300 ppm ethanol. Freundlich adsorption mechanism was dominant as the variation of conductance and concentration is nonlinear. Based on this adsorption mechanism, E_a and Q values were estimated for each gas. Both E_a and Q values were found to vary linearly with gas concentration.

Acknowledgments

The research work was partially supported by research grant obtained from CSIR, Government of India, vide sanction letter no. 03/(1371)/16/EMR-II, dated 10 May 2016 and NNetRa grant vide sanction 5 (1)/2017-NANO dated 28 March 2018, DST/NM/NNetRa/2018 (G)-IIT-KGP dated 21 March 2018.

Conflict of interest

The authors declare no conflict of interest.

C	concentration
CO	carbon monoxide
CO2	carbon dioxide
CuO	copper oxide
C g	reducing gas concentration
EDL	electron depletion layer
eV s	Schottky barrier
E a	activation energy for gas adsorption during response process
E d	activation energy for gas desorption during response process
FFT	fast Fourier transform
Fθ	total number of active sites
G	graphene
G t	conductance transient of the sensor
G t response	conductance transient of the sensor during response process
G 0	conductance of the sensor in air
G 1	conductance of the sensor at site 1
HAL	hole accumulation layer
H2	hydrogen
k	Boltzmann constant
K	equilibrium constant for response process
K 0	fraction of molecules with energy > E a
LFCO	cobalt doped lanthanum ferrite
M	molar mass of gas
MEA	methanolamine
MO	metal oxide
MOE	methoxyethanol
NMP	N-methyl-2-pyrrolidine
NO2	nitrogen dioxide
N A	Avogadro number
N t	number of gas molecules adsorbed at time t
N ∗	total number of available adsorption sites
OSHA	Occupational Safety and Health Administration
O ads −	chemiadsorbed oxygen
O ad − surface	physiadsorbed oxygen
P	gas pressure
PCA	principal component analysis
PEL	permissible exposure limit
Q	heat of adsorption
R	universal gas constant
R a	sensor resistance in air
R ad	physiadsorbed reducing gas
R g	sensor resistance in reducing gas
RO ad	adsorbed oxidizing product of reducing gas
RO gas	desorbed oxidizing product of reducing gas
S	sensor response (%)
S area	surface area of a single adsorbed molecule
S p	sensor response (%) of p-type semiconductor
S n	sensor response (%) of n-type semiconductor
SMO	semiconducting metal oxide
SnO2	tin oxide
T opt	operating temperature of sensor
VOC	volatile inorganic compound
WO3	tungsten oxide
ZnO	zinc oxide
κ	selectivity factor
ν	frequency of oscillation of adsorbed gas
τ irrev	characteristic response time in irreversible response process
τ rev	characteristic response time in reversible response process
τ res	response time of sensor
τ rec	recovery time of sensor