In 2 O 3-and SnO 2-Based Thin Film Ozone Sensors : Fundamentals

The paper considers SnO 2 and In 2 O 3 thin films as materials for the design of solid-state conductometric ozone sensors in depth. In particular, the present review covers the analysis of the fundamentals of SnO 2 and In 2 O 3 -based conductometric ozone sensor operation.Themain focus is on the description of mechanisms of ozone interaction withmetal oxides, the influence of air humidity on sensor response, and processes that control the kinetics of sensor response to ozone.


Introduction
Ozone is a highly reactive form of oxygen.The molecules of ozone contain three oxygen atoms (O 3 ) and are unstable with respect to O 2 .A certain amount of ozone is produced in the troposphere in a chain of chemical reactions involving hydrocarbons and nitrogen-containing gases.Though ozone is a minor atmospheric constituent, with an average concentration of about 3 ppmv (parts per million volume), this gas plays great role in our life.The ozone layer protects life on Earth from ultraviolet solar radiation and is a significant contributor to the radiative energy balance of the atmosphere.Most of the atmospheric ozone (90%) is located in the stratosphere with a maximum concentration between 17 and 25 km.This means that ground concentration of ozone is considerably smaller than in the stratosphere.However, this situation started to change during last decades.The increased level of ozone in the atmosphere is a result from the interaction between sunlight and various chemicals emitted into the environment by industrial means.Automobiles can significantly affect ozone measurements as well.The oxides of nitrogen emissions can be catalyzed by the summer sun, resulting in ozone alerts.Thus, ozone near the ground is a product of industrial and urban pollution.In addition, in the last decade, considerable interest was shown in ecologically clean "ozone" technologies [1][2][3] (see also http://www.io3a.org/and http://www.ioa-ea3g.org/).It was established that ozone can be widely used in medicine, as well as in different technological processes [4][5][6][7][8][9][10].Due to its special characteristics, ozone can be applied in industries such as pharmaceuticals, food, textiles, and chemicals, as well as in water preparation, medical treatment, deodorizing, and the purification of gases (Figure 1).For example, ozonation, which is an effective technique for sanitizing ultrapure water and treatment systems, was introduced in the semiconductor industry over 30 years ago.Unlike other agents, such as chlorine or chromic acid, ozone oxidation is not accompanied by the appearance of toxic waste, since this process generates only oxidation products and oxygen.Ozone can be used for the improvement of organoleptic qualities, bleaching, iron and manganese elimination from water, the reduction of organic substances in suspensions, and the elimination of toxic compounds.Because of its relatively short halflife, ozone is always generated on-site by ozone generators, which usually use UV light and corona discharge for ozone generation.
However, it has been established that high concentrations of this powerful oxidizing gas in the ambient atmosphere are hazardous to human health [11][12][13].A high level of ozone gas in the atmosphere is harmful to the human respiratory system, causing inflammation and congestion of the respiratory tract.An individual who remains in a 0.1 ppm O 3 environment for two hours will sustain a loss of 20% breathing capacity and, after remaining in 1 ppm O 3 for six hours, will suffer an attack of bronchitis.A mouse kept in 10 ppm O 3 will not survive.It is necessary to note that ozone, due to its relatively short half-life (minutes in confined spaces), does not instantly diffuse in air to create a uniform concentration for easy measurement.Therefore, in order to control the ozone concentration in the atmosphere and surrounding environment, it is necessary to develop both effective and economical methods and devices for the continuous monitoring of ozone concentrations at multiple locations, including technological processes with incorporated ozone treatments.These methods are also needed to determine attenuation of the ozone layer by manmade gases, to validate model estimations of changes in ozone layer, and to confirm the efficiency of the Montreal Protocol on Substances that deplete the ozone Layer (http:// ozone.unep.org/pdfs/Montreal-Protocol2000.pdf).As is well known, in recent decades there has been the depletion of the ozone layer, which can negatively affect the life on Earth.The increase in the number of articles devoted to the development and research of ozone sensors (see Figure 2) confirms the importance of this issue.Experiments have shown that analytical techniques developed to solve these problems should be capable of measuring ozone concentrations in the range of 0.01-10 ppm for these purposes.For example, ambient air measuring at the ground requires instruments with little ppb accuracy in a range 20-200 ppb, whereas in industrial and commercial applications the precision is not so stringent and the expected concentrations are higher (up to 10 ppm).At present, a number of conventional analytical methods, such as UV absorbance, optochemical, optical, chemiluminescence, fluorescence, and electrochemical methods, are available for the accurate analysis of ozone concentration in the air [14][15][16][17].However, these sophisticated high accuracy techniques are advisable for application in both metropolitan and industrial urban areas under conditions of relatively high atmospheric pollutant concentration that are close to WHO guidelines.Moreover, as a rule, these methods are based on periodic air sampling and analysis of the samples with rather large and expensive stationary analytical devices.Therefore, it is very difficult, or even impossible, to track the distribution of ozone concentrations within large geographic areas in real time.In addition, the conventional analytical methods do not provide enough sensitivity for the detection of small concentrations of ozone in many cases [14,16].Solidstate conductometric gas sensors discussed in the present paper were designed especially for these purposes.Results of comparative analysis of the methods used for ozone monitoring are shown in Table 1.

Solid-State Conductometric Gas Sensors
The main advantages of solid-state conductometric gas sensors are ease of fabrication by using thin and thick film technologies, simple operation, and low production costs, because the materials and manufacturing techniques easily lend themselves to batch fabrication [18].This means that well-engineered metal oxide conductometric sensors can be mass-produced at a reasonable cost.Reversibility, rapid response, longevity, and robustness are other merits of metal oxide gas sensors.Moreover, these sensors are compact and durable; therefore, they have very good suitability for the design of portable devices [18,19].This compensates for their disadvantages, such as low selectivity and temporal drift, and opens great opportunities for using them in alarm systems and portable instruments, including devices for balloon-borne atmospheric O 3 profile-sounding [20], electronic noises [21,22], and sensor networks for air pollution monitoring with wireless communication [23].Thus, compared to the reference methods discussed in the previous section, the use of low-cost conductometric ozone sensors for monitoring ambient air and technological processes would reduce air pollution monitoring costs and would also allow larger spatial coverage especially in remote areas where monitoring with traditional facilities is cumbersome.One should note that solid-state conductometric gas sensors can be also used for controlling ozone concentration in water.
They can take advantage of the gas permeability of polymer or another suitable membrane to separate the heart of the sensor from the water.This separation enables providing a controlled environment for sensor operation, while allowing ozone to enter from the sample and react with metal oxide surface.General descriptions of conductometric gas sensor fabrication and operation can be found in review papers [18,[24][25][26][27][28][29][30][31][32].

Advantages of In 2 O 3 and SnO 2 for Ozone Sensor Design.
One should note that the first ozone sensors based on the oxides mentioned above were designed by Takada et al. [76,77] at the end of the 80s.SnO 2 and In 2 O 3 are wide-bandgap semiconductors, which typically crystallize in the rutile and cubic structures correspondingly, and they are by far the most studied and most successful semiconductor materials for gas sensors [19], including development of ozone sensors (see Figure 3).These sensing materials have advantages, such as high sensitivity, fast response, and good thermal and temporal stability.Furthermore, In 2 O 3 -and SnO 2 -based ozone sensors also show very low cross sensitivity.An experiment has shown that sensitivity to O 3 , even in the ppb range, was much larger than the response to NO 2 , CO, H 2 , SO 2 , and NH 3 in the tens of ppm range.In 2 O 3 and SnO 2 also provide the opportunity for miniaturization and integration at micromachined substrates [14].
In 2 O 3 and SnO 2 are gas-sensing materials, since they are conductive and possess the ability to donate electron density to the adsorbed oxygen molecules, which arises primarily due to bulk nonstoichiometry firstly within the oxygen lattice.Of course, the reactivity of the metal oxide surface is critically dependent on its doping, the defect structure, and the operating temperature.Possible oxygen anion states of adsorbed oxygen at the surface of SnO 2 and their transformations can be presented by the following scheme: In this scheme, (g) is gas, (ad) is adsorbed, and (lat) is lattice.All these transformations are thermally activated processes and as a rule are accompanied by charge transfer between adsorbed species and the conduction band of metal oxides (see Figure 4) [24,78,79].Thus, the gas-sensing effect strongly points to an intrinsic defect reaction, such as oxygen vacancy (V O ) or the cation interstitial (Sn i , In i ) mechanism [80].As applied to the tin dioxide, these reactions can be described by the following schemes:  Molecular-ionic species Atomic-ionic species Figure 4: Literature survey of oxygen species detected at different temperatures on SnO 2 surfaces with infrared analysis (IR), temperatureprogrammed desorption (TPD), and electron paramagnetic resonance (EPR).Data extracted from [24].
Analysis of the literature indicates that gas sensors are fabricated mainly using either ceramic or thick film technology [18].These sensors typically have a thickness of gas-sensitive layer of more than 10 m.However, experiments have shown that sensitivity to ozone in these sensors usually considerably concedes sensitivity to the ozone of the sensors manufactured on the base of thin film technologies, when the thickness of gas-sensitive layer is not more than 300-500 nm [34,42,60,81].At present, the limited sensitivity of SnO 2 -and In 2 O 3 -based thin film ozone sensors is less than 5-10 ppb.Thin film sensors also have a faster response in comparison with thick film devices, having a thickness exceeding a few micrometers [43,44,82].This is the reason that this review is devoted to ozone sensors with thin gas-sensitive layers.Such sensors can be fabricated using both the thin film and thick film technologies.Studies have shown that after proper optimization of thick film technology this technology allows forming such thin films.
In [16], we began analyzing ozone sensors based on In 2 O 3 and SnO 2 and carried out a detailed comparison of their parameters.In the present paper, we continue considering SnO 2 and In 2 O 3 as materials for ozone sensor design.In particular, the present review covers the analysis of the fundamentals of ozone sensor operation, including a description of the mechanisms of ozone interaction with metal oxides, the influence of air humidity on sensor response, and the processes that control the kinetics of sensor response to ozone.One should note that the results of the analysis conducted in this review can also be used when considering ozone sensors developed on the basis of other metal oxides.

In 2 O 3 -and SnO 2 -Based Conductometric Ozone Sensors.
As indicated previously, the conductometric ozone sensors analyzed in the present paper are the most used and studied type of devices designed for the control of toxic and inflammable gases [83][84][85][86][87][88][89].In 2 O 3 -and SnO 2 -based conductometric ozone sensors are usually planar devices.Planar conductometric sensors have a simple structure [18].They consist of two elements, a sensitive conducting layer and contact electrodes.Contact electrodes are often interdigitated and embedded in the sensitive layer.The sensing layer (in our case In 2 O 3 or SnO 2 films) and electrodes can be fabricated by a number of microelectronic protocols as "single-sided" or "double-sided" designs using thin and thick film technologies.The basis of conductometric sensor operation is the change in resistance under the effect of reactions (adsorption, chemical reactions, diffusion, and catalysis) that take place on the surface of the sensing layer [18,[24][25][26][27][28][29][30][31].Chemical species (ozone, in our case) interact with the sensitive layer and, due to the electron exchange between adsorbed species and the metal oxide, modulate its electrical conductivity.The change in conductivity is correlated to the concentration of the chemical species (ozone in air) and can be measured using DC voltage applied to the device.Sensor response to ozone is, as a rule, calculated as the ratio of the resistances measured in the atmospheres containing and not containing ozone ( =  ozone / air ).
For the activation of surface reactions that take place during gas detection, conductometric ozone sensors operate at a high temperature as a rule [18].For this purpose, gas sensors have incorporated a resistive heater, made usually in the form of a meander, which dissipates the Joule power.Planar sensors are very robust; the temperature homogeneity over the sensing layer is good.It is necessary to note that thick film and thin film technologies are frequently combined to produce planar ozone sensors, especially in laboratory investigations.For example, gas-sensing polycrystalline SnO 2 and In 2 O 3 layers may be prepared via thin film technologies (e.g., spray pyrolysis, magnetron sputtering, laser ablation, or chemical vapor deposition), while the heaters are formed with thick film technology [18].even for 60 ppb of ozone [51].Of course, such sensitivity is excessive for real applications.Sensors with an extremely high response usually have a narrow range of detectable ozone concentration and poor stability due to strong sensitivity to the influence of outside factors, such as air humidity, temperature, interference, and poisoning.Experience has shown that response ∼10 2 -10 3 at 1 ppm O 3 is the sufficient one for market of ozone sensors.

General Characterization of
We admit that there are also some papers that report a maximum response to ozone in the range of 100-200 ∘ C [33,35,42].However, we cannot explain such a strong distinction.There is an assumption that the changing of the grain crystallographic structure with corresponding changes of facets and point defect concentration could affect this distinction [37,97,98].In particular, our experiments have shown that the temperature of the maximum sensor response ( oper (max)) depends strongly on the grain size, deposition temperature [45], and film thickness [57].In other words,  oper (max) depends on the parameters that influence the shape of the crystallites (i.e., crystallographic planes participated in gas-sensing reactions).The results of the XRD and TEM measurements confirm this fact [96,[99][100][101][102][103].In particular, the data presented in Table 4 show that there are many crystallographic planes, which can form facets in crystallites and can therefore participate in gas-sensing effects.Models of several of the most stable SnO 2 and In 2 O 3 crystallographic planes are shown in Figure 6.It can be seen that every crystallographic plane is characterized by a specific concentration and configuration of surface atoms (i.e., every plane has its own specific electronic, thermodynamic, adsorption, and catalytic properties) [80,97,[103][104][105].For example, according to the results of the theoretical simulation carried out by Agoston [80], for In   100), ( 111), ( 110), (001), (001), ( 101), (011), ( 111), ( 111), ( 111), ( 211) [110][111][112] SnO 2 (111), (200), (020), ( 112), ( 120), ( 110), ( 101), (011), (011), ( 101), (010), (001), ( 100), ( 110), (010), (010), ( 110), ( 100), ( 221), ( 221), ( 221) [99,112,113] surfaces with polarity, such as (001), can be significantly off-stoichiometric.The largest stoichiometry variations take place for the (001) surfaces, which partially resemble the chemistry of the (211) surfaces and step edges of the (111) surfaces; (3) the chemistry of In 2 O 3 surfaces depends on orientation.While most stable surfaces (111) and ( 011) are characterized by their chemical inertness up to low oxygen chemical potentials close to the material stability limit, the (001) surfaces exhibit various surface phase transitions, which are sensitive to experimental boundary conditions.In other words, gas surrounding (4) hydroxylation of the polar (001) surface is effective in reducing the surface energy to values comparable to or even lower than those of other faces.This is because water adsorption is relatively weak on the more stable ( 011) and ( 111) surfaces and locally restricted to a small fraction of the total surface area.Nevertheless, dissociative water absorption is still possible on those surfaces also.The same can be said of the various crystallographic planes of SnO 2 [37,97].These results indicate that an explanation of the sensing properties of the oxide nanocrystals cannot be simply provided in terms of the grain size effect [24,[106][107][108][109] but must also refer to the specific transducing properties of the nanocrystal surface [37,51,97,98].However, a large number of crystallographic planes, which can be simultaneously present in the faceting of the crystallites, the strong dependence of crystallite faceting on the crystallite size, and the technological route used for crystallite synthesis or deposition, complicate this task considerably [99,102,[110][111][112][113][114].Perhaps only epitaxial films and nanobelts can be used for designing sensors in which only one crystallographic plane will work [97,104].
A distinctive feature of ozone sensors is the essential difference in response ( res ) and recovery ( rec ) times observed for dry atmosphere (see Figure 7) [43,56].Experiments have shown that the response of In 2 O 3 and SnO 2 thin films deposited by spray pyrolysis is very fast.Even at  oper ∼ 200 ∘ b, the response time of sensors with optimal thickness ( < 60-80 nm) in a dry atmosphere does not exceed 1-2 sec.One should note that, in a wet atmosphere, the response time changes substantially [43,56,115].It was found that the influence of water increases strongly with the decrease in grain size [109], and this influence depends on the operating temperature [56].In low temperature range, the influence of the water vapors on the kinetics of sensor response intensifies [54].
Recovery is much slower ( rec ≫  res ), and  rec is considerably weaker depending on the humidity of the tested gas.For comparison, during the detection of reducing gases, such as CO and H 2 ,  rec ≈  res [116].According to [117], the indicated behavior ( rec ≫  res ) is peculiar for the adsorption/desorption mechanism of gas sensitivity.As is known, in the case of reducing gas detection, the response is controlled mainly by the catalytic reaction of oxidation.The necessity of both the time desorption increase and the time adsorption decrease of chemisorbed particles is one of the important conditions for achieving the maximum response of solidstate gas sensors, predicted in the frame of the adsorption mechanism of gas sensitivity [117].The short adsorption time and long desorption time correspond to the condition for higher surface coverage of the adsorbed species of target gas and hence the maximum change in the electrical properties (i.e., resistance) at a given concentration of the target analyte.In full correspondence with this statement, for both In 2 O 3 and SnO 2 ozone sensors, we observe the fulfilment of the following correlation: where  = 1.0-1.5 [43,45].The bigger the  rec / res ratio is, the greater the sensor response is.However, it is necessary to take into account that the increase in film thickness reduces the difference between response and recovery times indicated above.Additionally, for thick films, we have the trend of  res →  rec [53].This means that the processes that control the kinetics of sensor response depend on the film thickness and can be changed.
Regarding the dependence between sensor signal and ozone concentration, we can say that, as a rule, these dependences (see Figure 8) can be described with the power equation [26]: where  and  are constants;  is the ozone concentration in the air; and  0 and  are stationary sensor signals in the air without ozone and in the presence of ozone, respectively.
Unfortunately, the values of  and  are not constant for various sensors.For example, in the literature on , one can find that values have varied from 0.2 to 1.5 [50,55,123,124].Experimental studies and phenomenological modeling have shown that the power exponent  can, in principle, be derived from the mass action laws and rate constants of the respective surface reactions and the relationship between the sensor resistance and the trapped charge density of chemisorbed species [25][26][27][125][126][127].In practice, however, this derivation is complicated, because the relationship between the resistance and gas concentration depends not only on the surface chemical reaction, but also on the morphology of the sensor, such as a compact (nonporous) thin film or porous thick layer, and on its microstructure (the grain size, the neck area between partially sintered grains, the inhomogeneous distribution of the grain size, the pore size, and other microstructural features) [24,30,[128][129][130].Thus, the present situation with a great range in  values means that, firstly, the metal oxides studied in different laboratories as ozone-sensing materials have a significant difference in structure, and, second, the response of sensors fabricated in different labs can be limited by various processes.

Mechanism of Ozone Interaction with Metal Oxide Surface in Heated Ozone Sensors
One should note that, until now, the sensing mechanism of oxidizing gases, such as O 3 , has been a matter of debate [53,54,58,123,[131][132][133][134].Due to the acceptor character of the surface state induced by the O 3 chemisorption (these states can trap significant amounts of electrons from the conduction band), ozone ionosorption is accompanied by an increase in negative charge at the surface of an n-type semiconductor and an increase in band-bending (Δ  > 0).For an n-type semiconductor, this results in the creation of a depletion layer at the surface of the metal oxide grains and a decrease in the conductance of intercrystallite barriers, which control the conductivity of polycrystalline materials [ 0 ∼ exp(  /)].In this case, the sensor response can be expressed as where  0 and  denote the surface conductance before and during oxygen ionosorption, respectively.The mobility of charge carries changes slightly in comparison with concentration [52].
The analysis of possible chemical reactions of In 2 O 3 and SnO 2 with ozone [55,58,60,123] allows the proposal of the following reactions, which give the clearly expressed acceptor character of ozone interaction with metal oxide: In this reaction, e − is a conduction electron in the oxide film, O − (s) and O 3 − (s) are surface oxygen ions, and O 3 (g) and O 2 (g) are the adsorbing ozone and desorbing oxygen molecules, respectively.
However, by analogy with O 2 , which has a larger dissociation energy than O 3 [123,135] but already adsorbs dissociatively at the surface of SnO 2 at  > 150-180 ∘ C [86,136], we believe that, at  oper > 100-150 ∘ C, reaction (8) (i.e., dissociative O 3 absorption) is preferred [60,123].The same statement can be found in [42,123].It is known that ozone is a highly unstable molecule that readily donates its extra oxygen atom to free radical species.The geometry of the ozone molecule is such that it has a significant dipole moment, and the central oxygen atom has a lone pair of electrons.In other words, the central atom possesses a positive charge, and one of the side atoms possesses a negative charge (see Figure 9).Therefore, it is quite reasonable to assume that this atom interacts with the metal oxide surface and forms an acceptor, whereas the other two atoms form an oxygen molecule that returns to the gas phase, as shown in reaction (8).
This statement finds confirmation in the results reported in [123,136,137], which demonstrate that molecular O 3 was observed at the surface of metal oxides at  < 300 K only.At a higher temperature, O 3 was not stable.As is known, ozone has significantly lower dissociation energy than oxygen; dissociation energies for ozone and oxygen are 1.1-1.3eV and 5.1 eV, respectively [138,139].There is only one investigation [140] that has shown the formation of O 3 − on the metal oxide surface (CeO 2 ) interacting with ozone at temperatures below 65 ∘ C.Moreover, O 3 adsorption in the form of O 3 − should lead to the formation of localized dipoles on the surface [91,141].However, additional signals derived from directed dipole layers adsorbed at surface states were not observed in In 2 O 3 interacting with O 3 , even at 130 ∘ C [91,141].One can assume that, similar to oxygen, ozone dissociatively adsorbs on the reduced surface cations of Sn (Sn 2+ ) and In (In 2+ ) [78,142].As is known [31,143], the active sites of adsorption on a metal oxide surface are coordinatively unsaturated metal cations and oxygen anions.Metal cations as electron-deficient atoms with vacant orbitals exhibit Lewis acid properties, while oxygen anions act as a base.A schematic diagram illustrating this process for SnO 2 is shown in Figure 10.It is important to note that, in addition to Lewis acidity, metal oxides also possess Brønsted acidity, which is defined as the ability to protonate bases (H + -transfer) [31].This type of acidity is the result of the dissociative adsorption of water molecules on the oxide surface for forming protons (H + ) and hydroxyl groups (OH − ).The driving force of these processes is the interaction between a Lewis acid (cation on the metal surface) and its conjugate base (the oxygen atom in a molecule of H 2 O).A proton (Lewis acid) formed as a result of this dissociation is captured and tightly binds the surface lattice oxygen atom of the metal oxide (Lewis base), while a hydroxyl group (base) is fixed to the metal cation (acid).
Of course, the process of oxygen incorporation in the metal oxide lattice can also be involved in the reaction of ozone detection.For example, this reaction can be presented using the Kroger-Vink notation as the pathway: Equation ( 9) represents the exchange reaction of oxygen between the gas phase and the metal oxide lattice, in which the predominant defects are oxygen vacancies.As is known, oxygen vacancies are the main point defects in metal oxides.
In general, as shown in (9), oxygen adsorption and incorporation into the lattice decrease the free carrier (electron) concentration in n-type metal oxides.However, one should take into account that reaction (9) for SnO 2 is valid for the bulk of the metal oxide only.It was found that oxygen incorporation in the surface vacancies of the SnO 2 lattice is not accompanied by the capture of electrons from the conduction band [105,144].This means that, for the indicated electron exchange, the process of oxygen incorporation should be accompanied by bulk diffusion.However, experiments related to the study of metal oxides, such as SnO 2 , TiO 2 , and SrTiO 3 , have shown that, in the temperature range below 500 ∘ C, the incorporation reaction of the adsorbed oxygen atoms into the lattice (i.e., bulk diffusion) is very slow; therefore, the oxygen exchange reaction is effectively blocked [145][146][147][148][149]. As a result, the oxygen vacancies in the bulk are unable to equilibrate with oxygen from the gas phase.Therefore, for most metal oxide conductometric gas sensors that operate in the temperature range between 100 ∘ C and 500 ∘ C, chemisorption is the dominant effect that controls the sensing mechanism as a rule.Only at higher temperatures (typically above 500-700 ∘ C) does the incorporation into the lattice of metal oxides become the dominant mechanism responsible for the change of their resistance, which takes place due to the bulk oxygen diffusion.DFT calculations of the SnO 2 (101) surface support this conclusion, suggesting a self-limiting mechanism for the oxygen exchange reaction at temperatures below 400 ∘ C [150].In this temperature range, SnO 2 sensors show the maximum response to ozone.However, it is important to note that the surface stoichiometry of metal oxides during interaction with a surrounding gas can be changed despite the invariance of the bulk stoichiometry.In particular, Korotcenkov et al. [53,54] have used this approach based on surface reduction/reoxidation processes to gain an understanding of gassensing effects in In 2 O 3 -based ozone sensors.Korotcenkov et al. [53] believe that, at  oper >  th ∼ 120-150 ∘ C, the surface stoichiometry and thus the electronic and surface properties of In 2 O 3 grains can be changed under the influence of the surrounding atmosphere.The threshold temperature ( th ) corresponds to the temperature below which the oxygen can only be adsorbed on the surface of indium oxide without incorporation into the lattice.These results are in agreement with the data reported by Himmerlich [151].Himmerlich [151] did not observe oxygen incorporation in the In 2 O 3 lattice at room temperature either.Moreover, he has shown that the In 2 O 3 band gap in the surface layer can be changed under reduction/reoxidation reactions.
One should note that the maximal sensitivity of SnO 2and In 2 O 3 -based sensors to ozone is observed in a temperature range when the surface of metal oxides is free from molecular water ( > 100 ∘ C) and the concentration of chemisorbed atomic oxygen at the surface of metal oxide is still low [45,56].A diagram illustrating the change of the surface species concentration on the SnO 2 surface is shown in Figure 11.Such a situation seems natural to us, because molecular water apparently blocks the sites for dissociative ozone absorption or transforms ozone into another form.For example, earlier in [152], it was reported that SnO 2 surface dehydroxylation is required to obtain an increase in the surface coverage of charged oxygen species adsorbed at the specific surface sites.Bulanin et al. [136] also found that the adsorption of water, in the case of a totally hydrated CeO 2 surface, completely poisoned the sites of molecular O 3 chemisorption.It is also known that the interaction between the oxygen of water and the central oxygen of ozone produces a stable H 2 O-O 3 complex [153].The necessity of a low initial concentration of chemisorbed oxygen is also understandable, because this value determines the possible range of the surface charge change (i.e., the change of surface potential (  ) during interaction with ozone).As is known, the concentration of chemisorbed oxygen controls the initial band-bending of metal oxides.In addition, the presence of initial chemisorbed oxygen at the surface of metal oxides can initiate reaction (10), which also reduces a sensor response through the consumption of the ozone part involved in surface reactions.Consider the following: We believe that increased operation temperature also resolves the problem of the surface carbon.Certainly, in real conditions, black carbon [the energetic position of the C1s peak corresponds to the highly ordered pyrolytic graphite phase of carbon] is present in high concentrations at the surface of metal oxides.According to [154,155], the surface concentration of carbon at room temperature corresponds to 0.5-0.8 of monolayer.Annealing in a vacuum decreases the surface concentration of carbon.However, in the air, the initial concentration of carbon is restored even at room temperature.The correlation established by Brinzari et al. [154] between carbon surface concentration and peak intensities (Sn3d and O1s), corresponding to SnO 2 lattice atoms, has shown that the intensity of the Sn3d peak has a stronger dependence on the surface concentration of carbon species in comparison with O1s (see Figure 12).Such different behavior allows the assumption that the carbon on the SnO 2 surface is bonded predominantly to Sn Sn , which is also the adsorption site for the dissociative adsorption of ozone.This means that carbon is also a poison for O 3 dissociative adsorption.Ozone exposure removes C-related contamination from the metal oxide surface, including SnO 2 and In 2 O 3 [155,156].Interaction with ozone changes C-related compounds adsorbed onto the metal oxide surface into volatile species, such as CO 2 or CO.However, at RT, this process is comparatively slow.Experiments have shown that even at a high concentration of ozone (up to several vol.%) an equilibrium state can be achieved during several tens of minutes [156,157].Thus, operation at higher temperatures either increases the rate of carbon oxidation (i.e., its removal) or increases the rate of CO 2 catalytic decomposition (i.e., the restoration of initial carbon coverage) and thus reduces the influence of these processes on the kinetics of sensor response.Himmerlich et al. [158] have concluded that desorption of carbon species from the In 2 O 3 surface activates existing defect sites for interaction with gases, such as ozone.They have also found that this effect has a strong influence on the performance and sensitivity of indium oxide-based gas sensors.
Regarding the recovery process during ozone detection, this process is not as simple as it seems at the first glance.Recovery processes during interaction with ozone are usually considered the reaction of associate oxygen desorption: It is clear that this process has the activation nature, because the surface migration of oxygen atoms and electron exchange are required for this reaction.As is known, the ionized chemisorption is limited by the upper band-bending on the material surface [159,160] (often referred to as the Weisz limitation in the sensor literature).Therefore, the concentration of chemisorbed species cannot exceed 1% from the number of sites possible for adsorption.In other words, the chance of finding chemisorbed oxygen at the nearest surface sites is very low.This means that the rate of this reaction should increase with increasing operating temperature and decreasing band-bending.However, as shown in Figure 13, the temperature dependences on recovery time, especially for SnO 2 -based ozone sensors, have at least two sections with specific behavior.At  rec = ( oper ), the dependences have the activation character only at high temperatures.This means that the recovery processes in various temperature ranges have different natures.We assume that, in addition  to the oxygen desorption described by reaction (11), the reactions, such as ( 12) and ( 13)- (15), can also take place and control the recovery time.As is known, the surrounding atmosphere contains ∼0.5 ppm of H 2 and ∼2 ppm of CH 4 : Data presented in [79,136,161,162] demonstrate that reactions ( 12)-( 15) are possible.In particular, Gatin et al. [162] have studied the adsorption of oxygen and hydrogen at the surface of a film consisting of SnO 2 nanoparticles and have found that molecular hydrogen even at room temperature interacted with adsorbed atomic (chemisorbed) oxygen with formation of water molecules and band-bending change.This means that this reaction does not have a strong limitation in the low temperature range.Bulanin et al. [136] have shown that the atomic oxygen at the surface of TiO 2 appeared due to O 3 decomposition being able to interact with CO with forming CO 2 even at 77 K. Tabata et al. [79] have analyzed the interaction of CH 4 with the SnO 2 surface and found that O − was consumed through the exposure of CH 4 at 200 ∘ C. The effect of water, which is present in the atmosphere, cannot also be excluded [45,54,56].According to reaction (16), the initial state (the particle balance) can be recovered through the O  − transformation into hydroxyl groups:

Room Temperature Ozone Sensors Operated under the Influence of UV Illumination
At present, there have been attempts to design SnO 2 -and In 2 O 3 -based ozone sensors for operation at room temperature (RT) [141,[163][164][165][166][167].This type of ozone sensor has recently attracted great interest due to its advantages, such as low energy consumption and the possibilities of integration into plastic packages and portable devices [168].However, one should note that such sensors usually have very low sensitivity, slow response, poor recovery features, and high sensitivity to air humidity.For example, Epifani et al. [165] reported that the response time of In Experiment has shown that indicated improvement of RT ozone sensors takes place due to the strong decrease in the "initial" sensor resistance under the influence of UV illumination.As it is seen in Figure 14, to obtain a maximum value of ozone sensitivity by In 2 O 3 -based sensor a photon energy higher than ∼2.9 eV is required.At photon energy larger than 3.2 eV, the ozone sensitivity remains nearly constant.Thus, the photon energy of UV light from GaN-based LED is acceptable for application in RT ozone sensors, as shown in Figure 14.The similar situation is observed for SnO 2 .For significant photoconductivity effect the wavelength of the illumination should not exceed 400 nm [186].According to general model of photoreduction, UV illumination with ℎ] >   (3.0-3.7 eV) generates electronhole pars in metal oxides.The holes generated in the spacecharge region are attracted to the surface and can be captured by chemisorbed surface species, such as O 2 − .This process can be described by the following schemes: In this scheme, {ℎ]} represents an absorbed photon with energy ℎ].The indicated process transforms oxygen species in the neutral form and thus facilitates their desorption (i.e., the photoreduction of the metal oxide).This process, due to an increase of the bulk charge carriers, because the electron of the hole-electron pair remains in the conduction band, and a decrease of the height of the intergrain potential barrier, is accompanied by a strong decrease of metal oxide resistance [171].The magnitudes of photoreduction are proportional to the UV light intensity.As shown in Figure 15, the reduction at RT in the vacuum is fast.This situation was expected, firstly because the hole injection into the chemisorption states is an exothermic process and, secondly, the oxygen desorption does not require additional reactions.As mentioned above, the change in the resistance of polycrystalline metal oxide films is ascribed to the change in the barrier height at the intergrain interface.Under these circumstances, the ratio of the resistance measured in the dark and after UV illumination can be approximately expressed in the form where ΔB is the barrier height change.Taking into account possible changes in the film resistance, particularly data presented in Figure 15, one can find that the reduction of  Another reason for increasing the RT sensor response under the influence of UV illumination is related to the UV effect on O 3 decomposition.For example, Chen et al. [187] established that the activity of metal oxides toward O 3 decomposition depends greatly on the surface properties of the oxide, the presence of irradiation, and the relative humidity.They have shown that while -Fe 2 O 3 can catalytically decompose O 3 in the dark, the TiO 2 surface is deactivated by O 3 exposure.Under dark conditions, the presence of atmospheric water inhibits O 3 decomposition at room temperature through the competitive adsorption of water molecules and ozone for active surface sites.In the presence of UV radiation under dry conditions, photoenhancement in O 3 decomposition was observed for both -Fe 2 O 3 and TiO 2 .The presence of both water and UV radiation leads to more complicated behavior.In particular, it was found that the UV irradiation of TiO 2 leads to an initial increase and a subsequent decrease in ozone uptake coefficients as the RH is increased.
It should also be taken into account that the joint action of UV illumination and ozone stimulates the removal of carbon species [157,158], which contributes to the activation of the metal oxide surface.As previously described, this process also occurs in the dark but is relatively slow and ineffective as in the presence of UV illumination [156].In an ozone atmosphere without UV light, efficient surface cleaning requires temperatures of approximately 200 ∘ C.
There are also suggestions that the effect of photoreduction/reoxidation on InO  films is related to the formation and annihilation of oxygen vacancies in the film [188].According to this mechanism, UV irradiation transfers oxygen from a bound to the gaseous state.In the resulting vacancies the two electrons of the oxygen ion are left in the vacant site and can contribute to the increase in the density of free carriers.The subsequent oxidation in an ozone atmosphere contributes to the incorporation of oxygen into the film, diminishes the density of oxygen vacancies, and therefore leads to a drastic decrease of the available charge carrier concentration.However, for the implementation of this mechanism in polycrystalline materials it is necessary to assume that fast oxygen lattice diffusion occurs even at room temperature, which is unlikely.Therefore, the majority of researchers are inclined to believe that the photoreduction/reoxidation is a surface effect [170].
For UV illumination, different UV Hg lamps [170] or GaN-based LEDs [163] can be used.A typical photoreduction/oxidation cycle for the In 2 O 3− films is shown in Figures 15 and 16(a).The processes of photoreduction and oxidation were fully reversible.The description of the mechanism of this interaction can be found in [119,131,172].
However, experiments have shown that the sensitivity comparable with that of conventional ozone sensors was achieved only for sensors based on the photoreduction/reoxidation principles, which operated using UV photoreduction in a vacuum or an atmosphere of inert gas.In this case, the photoreduction is carried out in vacuum, but the measurement of sensor response to ozone is conducted For UV illumination, InGaN-LED ( ∼ 375 nm) was used.Data extracted from [169].
in an atmosphere containing ozone at ordinary pressure.Such sensors based on films with thicknesses of less than 100 nm had extremely high sensitivity to ozone of up to 10 5 -10 6 even for the ppb range [15,170,171].It is important to note that the maximum response was observed for the films deposited at low temperatures.Increasing both the deposition temperature (see Figure 17) and the temperature of subsequent annealing significantly reduced this response.Sensor response strongly decreased also with increasing operation temperature (see Figure 18).In all cases, the decrease in sensor response was due to the increase in the conductivity measured in the atmosphere of ozone.It is clear that the cycle of measurement mentioned above is not convenient for real applications.In addition, the response in the low concentration range is especially slow (see Figure 16).One should also note that when using this method it is difficult to distinguish the influence of ozone and oxygen, because reoxidation also occurs in normal atmosphere (see Figure 15), albeit at a slower rate.This approach also does not take into account the change in conductivity due to net electron-hole generation, which can be several orders of magnitude.
For sensors operated in the usual atmosphere the response is subject to the same regularities that have been established for photoreduction in the vacuum, but the magnitude of the response did not exceed 6-10 for the ppm level of ozone concentration [163,166,172,178].Only Wagner et al. [119] reported the extremely high sensitivity of In 2 O 3based ozone sensors at RT ( ∼ 350 at 1.2 ppm O 3 ).A high sensitivity to ozone was observed in both dry and humid atmospheres.Typical characteristics of these sensors are shown in Figure 19.We cannot explain this result, because there are some papers that report an opposite effect.For example, Wang et al. [121] have shown that the sensitivity to ozone of UV-assisted In 2 O 3 sensors strongly decreased in a humid atmosphere at RT.They observed that the increase of air humidity from 0 to 12% RH was accompanied by a drop of sensor response from  ∼ 100 to  ∼ 13.In addition, Munuera et al. [132] have shown in the example of TiO 2 that UV photodesorption from fully hydroxylated surface is a complicated process, because the "photodesorption of water," which can be activated by reaction (20) with forming OH • radicals, can be accompanied by the photoadsorption of oxygen on these "fully hydrated surfaces."Consider the following: Only the thermal removal of water and hydroxyl groups reduced the photoadsorption of oxygen.Munuera et al. [132] assumed that the reduction from Ti 4+ to Ti 3+ under irradiation produced a weakening of the bond to the point of desorbing H 2 O from the surface.During the diffusion of "excitons" from the "porthole sites" to the trapping sites, desorption of water molecules may occur, with each exciton being responsible for desorption of a certain number of water molecules before oxygen adsorption takes place.This means that the UV photoreduction of the metal oxide surface in humid air is not as effective as in dry air.Thus, it is possible that the high RT sensitivity of the sensors designed by Wagner et al. [119] in humid air was achieved due to the specific technology used for the In 2 O 3 synthesis and, therefore, specific surface properties of gassensitive material; mesoporous In 2 O 3 used as gas-sensitive material was synthesized by a nanocasting procedure.In addition, the authors did not use the high temperature treatment ( > 500 ∘ C) to stabilize the properties of the films.Moreover, before each measurement, the samples were subjected to specific pretreatment.The same situation was observed in [182].Chien et al. [182] have found that a sharp increase in the sensitivity of RT ZnO-based sensors to ozone took place only in that case if the photoreduction was carried out at temperatures above 500 ∘ C.
As for the results presented in [176] (see Table 5), there is also some uncertainty about the conditions of the experiment, because the same authors in later articles [121,163,177] for the same films provided data on sensitivity to ozone, which at 4 orders of magnitude were lower than data reported in [176].Apparently, these differences are related to differences in the measurement methodologies used in these articles.In [176], the UV photoreduction was conducted under synthetic air without ozone, but the measurement of sensor response to ozone was performed at off UV light.In this case we have a situation close to that which was observed during UV photoreduction in vacuum, when it was not possible to separate the contributions of ozone and oxygen presented in the atmosphere in the change of film conductivity, while in [121,163,177] UV light was switched on during the whole measurement cycle.
Unfortunately, we have a very limited amount of research in the field of UV assistance RT ozone sensors and very limited information about In 2 O 3 and especially SnO 2 films used for the fabrication of these sensors.Therefore, we do not have the opportunity to conduct a detailed consideration of these devices.We can only say that the main requirements of metal oxides aimed toward application in RT ozone sensors [15,171,173] coincide with the requirements that have been established for conventional sensors.The response to ozone increased with decreasing grain size [163,176] and film thickness [174].For example, Wang et al. [176] reported that an increase in grain size from 7 to 10 nm was accompanied by the decrease in the sensor response from 10 5 to 10.

Water Influence on Operating Characteristics of Heated Ozone Sensors
It is necessary to say that water seems to play an important role in the reactions of ozone detection.As is known, metal oxide surfaces can be characterized by their ability to form strong chemical and hydrogen bonds with water molecules [24].Korotcenkov et al. [53,54,56] have shown that water influenced the magnitude of sensor response to ozone, response and recovery times, and the activation energies of processes that control the kinetics of gas-sensing effects.This means that the presence of humidity in the surrounding atmosphere adds complexity to the sensing process, and possible surface reactions should be more complicated, because water can interact with both lattice atoms of metal oxides and adsorbed species, including oxygen and ozone [53,56,116,166].For example, Cǎldǎraru et al. [189] concluded that the presence of hydroxyl groups at the surface of this oxide limited the oxygen adsorption at temperatures of 320-340 ∘ C even under exposure to pure oxygen.In addition, processes such as water dissociation, the protonation of surface groups, and ion surface complexation should be taken into account.
A schematic illustration of the behavior of water at the surface of SnO 2 is shown in Figure 20.However, admittedly, the role of water is not completely understood yet.For example, we are not able to suggest an experimentally confirmed mechanism, which could explain the growth of response in humid air observed for both In 2 O 3 -[33] and SnO 2 -based ozone sensors [56] operated in the temperature range of 200-400 ∘ C. It is important to note that, for RT In 2 O 3 ozone sensors operated under the influence of UV illumination, the response to ozone decreases with air humidity increases [121].For comparison, the response of the In 2 O 3 and SnO 2 sensors to reducing gases also decreases in humid air as a rule [33,190].We believe that only the mentioned effects are connected to the specific role of water both in the reactions of ozone detection and in the mechanism of ozone dissociation at the hydroxylated surface of SnO 2 .For example, hydroxyl groups are known to catalyze ozone decomposition [55].It is also known that the reaction of hydroxylation has a clear donor-like character [191][192][193] and causes an increase in conductivity and a decrease of band-bending.According to general conceptions [191][192][193], the water molecules interact dissociatively with defect-free tin oxide by forming a terminal OH-group situated at the Sn 5c Lewis acid site, while the proton forms the bridging OH-group with the twofold coordinated O atoms on the next ridge.One of the possible reactions that can explain the above-mentioned behavior of conductivity is presented below: Here, water desorption results from the ozone-OH interaction.We assume that ozone can form intermediates, such as HO 2 or HO 3 [194][195][196], during the interaction with the hydrated surface.In this reaction, we did not take into account molecular oxygen and water, because, as shown in Figure 20, at  > 200 ∘ C, OH-groups and chemisorbed oxygen were dominant at the surface of the metal oxides [56,116,197,198].Molecular water desorbs from the surface of SnO 2 at  ∼ 100 ∘ C, while molecular oxygen desorbs at  > 180-200 ∘ C. As shown in the right part of ( 21), the state of the SnO 2 surface after interaction with ozone is the same as the state of the surface after reaction with ozone in a dry atmosphere in this case.This means that hydroxyl groups participate in the reaction of ozone detection; therefore, dissociative water adsorption/desorption can control the kinetics of sensor response to ozone in humid air (read the next section).
When the operation temperature exceeds 350 ∘ C, the influence of water vapors on the operating characteristics of ozone sensors is being considerably decreased.As shown in Figure 21, at  oper > 350 ∘ C, the influence of RH becomes insignificant.In principle, this effect is expected, because it was observed many times for the SnO 2 and In 2 O 3 sensors during reducing gas detection.Moreover, this observation is consistent with the results of the temperature-programmed desorption (TPD) and Fourier transform infrared (FTIR) spectroscopic studies of OH-desorption from the SnO 2 surface [197,199].These reports show that hydroxyl groups start to desorb at  > 200-250 ∘ C and are virtually absent at  > 400 ∘ C. Thus, if there are no limitations in consumable power and we do not have strict requirements of selectivity (in the high temperature range, sensors are sensitive to reducing gases),  oper ∼ 350-400 ∘ C would be optimal for ozone sensor operation in an environment with variable humidity.Certainly, sensitivity in this temperature range decreases, but this reduction is not critical for ozone detection.

Processes That Control the Kinetics of
Response of Metal Oxide Ozone Sensors Figure 21: Influence of deposition temperature on the sensitivity to air humidity of SnO 2 sensor response to ozone (∼1 ppm).Sensitivity to air humidity was calculated as a ratio of responses measured in dry (1-2% RH) and wet (∼45% RH) atmospheres at  oper corresponding to maximum response in humid atmosphere ( ∼ 30-50 nm).Adapted with permission from [56].Copyright 2007 Elsevier.
but do not answer the question of how this interaction affects the properties of the metal oxide gas-sensing matrix and what processes control the kinetics of these changes.It is clear that the change in conductivity of metal oxides (Δ ∼ Δ  , where   is the concentration of free charge carriers in the metal oxide) induced by their interaction with ozone can be controlled by surface and bulk processes [54].
As shown in [57], In 2 O 3 and SnO 2 films deposited using thin film technologies are compact, and in spite of a considerably smaller thickness they do not have the porosity found in sensors fabricated by thick or ceramic technologies [86].Experiments have shown that films deposited using sputtering technologies had an especially dense structure [86,200].This means that these films are less gas-permeable in comparison with films prepared using thick film technologies.Typical cross sections of such films are shown in Figure 22.These SEM images show a clear interface between crystallites.This fact points to the absence of necks between grains, whose overlap is usually used to explain the high sensitivity of ceramics-based gas sensors.This means that films by thin film technology have larger contact areas between crystallites.addition, due to the columnar structure, in which grains can grow through full film thickness (see Figure 23), the area the indicated contacts is considerably increased with the film growth.For thick film sensors grain and the of intergrain contacts do not depend on the film thickness.Taking into account such a situation, Korotcenkov et al. concluded that intercrystallite (surface) diffusion should be involved in the explanation of the gas-sensing characteristics of thin film gas sensors in addition to commonly used surface and bulk processes.
According to Korotcenkov et al. [54], four processes control the magnitude and kinetics of response to ozone (see Figure 24).They are as follows: (1) adsorption/desorption processes and catalytic reactions at the metal oxide surface (Δ  ∼ Δ  ∼   , where   is the concentration of free charge electrons in the bulk of grains,   is the total concentration of chemisorbed chemical species, and   is the rate of surface chemical reactions; and Δ ∼ Δ  ∼ Δ  ∼ Δ  , where   is the height of the intergrain potential barrier and   is the charge on the interface surface states); ( 2 If all the four processes listed above participate in the gas-sensing to a greater or lesser extent, then the transient characteristics of sensor response should be represented as the superposition of the four components related to these processes [54]; that is, where  1 () is the change of conductivity connected to the adsorption/desorption processes on the surface of the metal oxide,  2 () is the change of conductivity conditioned by the processes of reduction/reoxidation,  3 () is the change of conductivity controlled by the intercrystallite (grain) diffusion of oxygen, and  4 () is the change of conductivity connected to bulk oxygen diffusion.Taking into account the general character of the discussion, to obtain a full picture, we have introduced in ( 22) an additional component,   (), which characterizes the rate of delivery of both analyte gas and oxygen to the place where indicated reactions occur.The last process, as is well known, is controlled by the diffusion of analyte gas and oxygen from the surface region of the gassensing material into its internal region.From (22), one can see that two extreme scenarios could occur in ozone sensors, depending on the correlation between the rate of gas diffusion and the rates of processes taking place on the surface and in the bulk of the metal oxide grains.In the first case, the kinetics of sensor response is controlled by the gas diffusion inside the metal oxide matrix: Such an approach for analyzing the gas-sensing properties of solid-state sensors was developed in [142,[201][202][203][204][205][206][207][208][209][210].
It is referred to as a diffusion-reaction model.According to this model, the gas molecules diffuse inside the sensing layer, while they are consumed at a certain rate during diffusion by the surface reaction.This means that the gas-sensing process is considered a phenomenon in which the diffusion and reaction are coupled.Because of this coupling, the response behavior is strongly influenced by the catalytic and microporous properties (gas penetrability) of the sensor material used.As a rule, the diffusion-reaction model describes good kinetics of response of sensors fabricated by ceramics or thick film technology and uses a gas-sensing layer with thickness in the micrometer range.
In the second case, the kinetics of sensor response has been controlled by the processes taking place either on the surface or in the bulk of the grains [54]: The last statement means that the process of gas diffusion into the metal oxide matrix is fast and does not limit the rate of the film conductivity change during the replacement of the gas surroundings.Such an approach to the analysis of gas-sensing effects was considered in [116,117,[211][212][213].As a rule, such an approach has been used for the description of the gas-sensing characteristics of thin film sensors.
It is significant that, during the measurement of the sensor response rate through the estimation of response and recovery times at the level of 0.5 and even 0.9 from the steadystate value of film conductivity, the time constants of response and recovery could not be controlled by the slowest process under certain conditions.They could be controlled by the most influential process (i.e., the one providing the maximal contribution to the conductivity change of the gas-sensing material).
No doubt, the suggested approach to the consideration of gas response kinetics is simplified enough.While making such a consideration, we did not take into account that the analyzed processes are multistage ones [24,[214][215][216] and, as a rule, comprise electron exchange between adsorbed species and the bulk of the metal oxide.It is known that the adsorption of chemical species without electron exchange (physisorption) cannot be responsible for the change in conductivity of the metal oxide [24,83].This means that, under certain conditions, the time of sensor response and recovery during the process of gas detection could be limited by the rate of electron exchange.However, research on SnO 2and In 2 O 3 -based CO and H 2 sensors conducted in [53,116,211] has shown that, at  oper > 100 ∘ C, the electron transfer of delocalized conduction band electrons to localized surface states took place quite quickly, and this process was not a factor that limited the rate of sensor response.
The experiment and simulation have shown that surface chemical reactions (reduction/reoxidation) may also take place via a precursor state, the occupancy of which is controlled by adsorption/desorption (/) processes [117,[214][215][216].In this case, the change of conductivity Δ 3 () could be presented as follows [116]:  This means that the adsorption/desorption (/) processes could exactly determine the kinetics of those reactions, as the slowest in this sequence of events.Such a conclusion was made in [116,117,211,212] while analyzing the gas sensitivity of SnO 2 -based solid-state sensors to reducing gases.It was established that the kinetics of sensor responses to CO and H 2 was controlled by the processes of oxygen and water adsorption/desorption.
It is significant that if the shape of transient curves () depends on the mechanism of the surface and bulk processes' influence on the metal oxide conductivity [24,212], the absolute value of the time constants, used for the description of observed changes in metal oxide conductivity, and activation energies, determined from the temperature dependences of response and recovery times, are specific parameters for those surface and bulk processes.As is known [28,117,146,201,202,212], processes that control the kinetics of sensor response are activation ones, which have some activation energy specific for each process.
A detailed study of the kinetics of solid-state sensor response to ozone was conducted by Korotcenkov's group for In 2 O 3 -based devices.The results of this research were reported in [54,56,116].It was found that, in a dry atmosphere (1-2% RH), the response time showed a clear dependence on film thickness (Figure 5) and grain size (see Figure 25), which can be described by the equation  res ∼ (, )  with index  close to 2 [43].As is known, such dependence is specific for the diffusion process.
Taking into account the features of the structural properties of thin films [43, 45, 47, 98-102, 109, 114], Korotcenkov et al. [54] have assumed that the above-mentioned diffusion process is intergrain (surface) oxygen diffusion.[54].Analysis of the literature data confirms that such activation energies correspond to the activation energy of oxygen diffusion in polycrystalline metal oxides, which is controlled by surface or intergrain diffusion and does not exceed 1.5-1.8eV [217].For comparison, for bulk oxygen diffusion in metal oxides, including In 2 O 3 , one can find activation energies in the range from 2.6 to 5.0 eV [200,218].As is known, surface diffusion is much faster (∼10 5 ) than bulk oxygen diffusion.At the same time, the recovery time depends weakly on the film thickness (see Figure 25) and the grain size even at relatively low air humidity (1-2% RH).In addition, in a wet atmosphere, the influence of In 2 O 3 grain size on the time constants of gas response to ozone dropped sharply, and the activation energies of  = (/) dependencies were changed (Figure 27(a)).Such behavior of the time constants indicates that the kinetics of sensor response in a wet atmosphere and the kinetics of recovery processes are controlled by several processes, most of which are related to the surface reactions.Analysis of the transient curves of conductivity response to ozone carried out in [53,54] has shown that the kinetics of recovery processes had two characteristic activation energies ( act ) for  = (/) dependencies [54]: (1) 0.25-0.3eV and (2) 0.5-0.6 eV.The  act ∼ 0.5-0.6 eV is peculiar to transient processes of the conductivity response of "porous" films.As a rule, these processes were dominant at  oper > 100 ∘ C when a dry atmosphere was used.Processes with the activation energy 0.25-0.3eV were generally observed during ozone detection in a wet atmosphere.Taking into account that processes with  act ∼ 0.3 eV were dominant in ozone detection in a wet atmosphere and processes with the activation energy of ∼0.6 eV were dominant in a dry atmosphere, it was assumed that the first one is related to water participation, and the second is associated with oxygen participation in these reactions [54].This is a real assumption, since only oxygen and water participate in reactions of ozone detection, and only dissociative oxygen adsorption/desorption in a dry atmosphere should control the kinetics of equilibrium achievement between the stoichiometry of the surface layer and the surrounding atmosphere during the recovery process of ozone detection.Regarding the processes with activation energies between 0.25-0.3and 0.6 eV, which are often observed in experiments, one can say that their appearance is quite ordinary.In a real reaction of ozone detection, both H 2 O and oxygen participate simultaneously; therefore, in many cases the reaction of ozone detection can run through two different but interrelated paths.Temperature programmed desorption (TPD) measurements [219] have shown that both chemisorbed oxygen and water were rather strongly bound to the In 2 O 3 surface.According to Yamaura et al. [219], after exposure to O 2 , significant oxygen desorption from the In As shown in Figures 26 and 27(b), in ultrathin films ( < 50 nm) with a small grain size, processes with activation energies of 0.25-0.3eV are already dominant at RH ∼ 1-3%.This observation is important and indicates that, with a grain size decrease into the nanometer range ( < 10 nm), the change of energetic parameters of adsorption/desorption processes at the In 2 O 3 surface or the increase of the water role in surface reactions occurs.The change of grain shape is confirmation of the first mechanism.As established in [102], at a thickness of less than 40-60 nm, the In 2 O 3 grains lose crystallographic faceting and obtain the shape of spherulite.The increase of water influence both on the gas sensitivity of In 2 O 3 films with a small grain size [53] and on NO  conversion by an In 2 O 3 -Al 2 O 3 catalyst with the smallest In 2 O 3 particles [220] is confirmation of the second mechanism.
It should be noted that the diffusion nature of the kinetics of sensor response in a dry atmosphere and the surface nature of the reactions responsible for kinetics of both the recovery process and sensor response in a humid atmosphere are not contradictory.Korotcenkov et al. [54] believe that diffusion also participates in processes occurring in a humid atmosphere.However, in this case diffusion is controlled by surface reactions.This means that the concentration of oxygen capable of diffusing and affecting the conductivity of metal oxides via influence on the intercrystallite potential barrier is controlled by the rate of surface reactions [221], particularly in a humid atmosphere, by dissociative water adsorption/desorption processes.These processes control the time of equilibrium establishment between the surface stoichiometry of In One should note that the approach used for explaining the kinetics of the In 2 O 3 ozone sensor response can be applied to SnO 2 ozone sensors as well.In particular, Korotcenkov and Cho [57] have shown that diffusion in the gas phase, including Knudsen diffusion [222] (the approach used for "thick film" and ceramic gas sensors [81,206,223]), could not limit the kinetics of response of SnO 2 thin film gas sensors.For films with a thickness of 40-70 nm, it is difficult to imagine the existence of a structure in which the sensitivity is limited by gas diffusion along the pores.Korotcenkov et al. [56] believe that, similarly to In 2 O 3 , the kinetics of SnO 2 based gas sensors is controlled by intergrain (surface) diffusion and surface reactions, such as the adsorption/desorption of water and oxygen in various forms.
Analyzing the kinetics of SnO 2 thin film ozone sensor response, Korotcenkov et al. [56] have established that, for SnO 2 response and recovery processes, there were several characteristic activation energies as well.It was found that transient processes with activation energies of ∼0.6-0.75 eV were typical in a wet atmosphere in the region of maximum sensitivity ( oper ∼ 200-300 ∘ C) (see Figure 28).The temperature dependence of the time constants for the recovery process generally showed activation energies in the range of 0.75-1.1 eV.In a dry atmosphere, the activation energies for dependence  = (1/) increased and attained 0.95-1.1 eV and 1.1-1.6 eV for the response and recovery processes, respectively.For some samples, in  = (1/) dependences, there were regions with activation energies of ∼0.26-0.4eV.As shown in Figure 28, processes with activation energies of 0.25-0.4eV were observed at temperatures of  oper < 200 ∘ C corresponding to the region of low sensitivity, while processes with activation energies of >1.0 eV generally appeared in the region of ( oper ) dependency ( oper > 300 ∘ C), which corresponds to a drop in sensor response to ozone.It is necessary to note that the change of activation energies of  = (1/) dependencies in the wide range from 0.25 to 1.6 eV demonstrates that, even for reactions of ozone detection, one should take into account the presence of various processes and reaction steps in a complex mechanism of response [224].It is interesting that the activation energies reported herein are similar to those obtained during the analysis of the kinetics of SnO 2 sensor response to reducing gases [116,211,212].According to [123], the processes with activation energies of 0.25-0.4eV are the consequence of the adsorption/desorption of molecular oxygen, while the processes with activation energies of ∼0.6-0.7 eV are connected to the adsorption/desorption of water, and processes with activation energies of ∼1.0-1.5 eV are related to the adsorption/desorption of chemisorbed oxygen [141,199].

RT Ozone Sensors Activated by UV Illumination.
Regarding the kinetics of the response of sensors operating at room temperature, the available information is not enough for any reliable conclusions.The differences in film thicknesses, conditions of experiments, and the wavelengths and intensities of the illumination used to activate the low temperature sensitivity to ozone do not allow comparing the results obtained by different teams and making any generalizations.In addition, the presence of water greatly complicates the mechanism of ozone interaction with the surface of metal oxides.We can only refer to the results presented in [176], which showed that the effect of film thickness on the time constants of transient processes is very similar to the regularities previously established for heated sensors working in dry air [43,44].As for the heated sensor the response time increases with increasing film thickness, and the recovery time does not depend on the thickness (see Figure 29).These results suggest that, similar to heated sensors, the response of RT sensors activated by UV radiation is controlled by diffusion processes, whereas some kind of surface reactions, controlled by the rate of charge carrier generation by UV light, is responsible for the rate of recovery processes.In most cases, when the UV illumination has a sufficient power and Power density (mW/cm 2 ) Figure 30: Dependences of the "light on" conductivity response of SnO 2 nanowires on incident power density and wavelength at room temperature.Data extracted from [186].
film is thin, the recovery is fast and the recovery time does not exceed several tens of seconds [169,177].If the rate of charge carrier generation is low, the time for ozone desorption by UV irradiation can reach tens of minutes or more [177].As a rule, this situation occurs when the film is thick or illumination with nonoptimal wavelength ( > 400 nm) is used for activation.In this case to achieve a noticeable effect the intensity of illumination should be much greater (see Figure 30).
According to Wang et al. [176], in case of sensor response, the surface effect becomes dominant only when the film thickness decreases to 10 nm.Wang et al. [176] believe that diffusion process, which was mentioned above, is the diffusion of oxygen molecules into the layer.From our point of view, the intergrain diffusion, which was discussed in the previous section, is a more probable diffusion process.With regard to the reduction of response time with increasing ozone concentration (see Figure 16), the process is quite natural, because the rate of oxidation is directly related to the concentration of the oxidizing agent involved in the process.

Conclusions
This paper presents a critical assessment of the recent literature on ozone measurement with conductometric metal oxide gas sensors.The pertinent literature analysis reveals that SnO 2 and In 2 O 3 metal oxides are promising materials for ozone detection in various applications.It has been shown that SnO 2 -and In 2 O 3 -based ozone sensors can achieve excellent detection limits down to a few ppb of ozone and fast response times of approximately several seconds at optimal operation temperatures.Therefore, we believe that these metal oxides, especially In 2 O 3 , which has a shorter recovery time, are good choices for fabricating low-cost, low-power, highly sensitive, and fast ozone sensors aimed for the sensor market.
This work also demonstrated that ozone sensors based on SnO 2 and In 2 O 3 are complex devices, and understanding their operation requires the consideration of all factors that influence final sensor performance, including surface chemical reactions, various diffusion processes, and the associated charge transfers.In the present review, we discussed these processes.However, it should be recognized that the level of understanding of these processes (i.e., the nature and origin of the mechanism that takes place between SnO 2 and In 2 O 3 surfaces and ozone) is still insufficient and further research in this area is required.
We hope that the present overview of the present state of the SnO 2 and In 2 O 3 response to ozone will be of interest to gas sensor developers and persuade them to carry out much more systematic research in order to better understand SnO 2 and In 2 O 3 chemistry, the adsorption/desorption processes taking place at the surface of SnO 2 and In 2 O 3 films with different crystallographic structures and stoichiometry, and the metal oxide-ozone interaction in atmospheres with different air humidity.Certainly, some precise theoretical calculations and computational modeling of ozone interactions with the SnO 2 and In 2 O 3 surfaces and the subsequent changes in the electronic band structure of gas-adsorbed metal oxide nanostructures are required as well.We believe that deeper knowledge of physical and chemical phenomena that take place at the surface of SnO 2 and In 2 O 3 can improve the sensing capabilities of these sensing materials for real applications.

Figure 1 :
Figure 1: Advantages and possible fields of ozone applications.

Figure 2 :
Figure 2: The growth in the number of publications devoted to the design of ozone sensors.Data extracted from Scopus.

2 3 ZnO SnO 2 In 2 O 3 Figure 3 :
Figure 3: The percentage distribution of the number of articles devoted to the study of metal oxide-based ozone sensors.Data extracted from Scopus.

Figure 9 :
Figure 9: Resonance Lewis structures of the ozone molecule.

3 Figure 10 :
Figure 10: Schematic diagram illustrating process of O 3 dissociative adsorption at the surface of SnO 2 in dry air.Idea from Bulanin et al. [136].

1 )Figure 15 :
Figure 15: Photoreduction in vacuum and oxidation in oxygen of a 60 nm thick InO  film deposited by magnetron sputtering at room temperature.Reprinted with permission from [173].Copyright 2001 American Institute of Physics.

1 )Figure 17 :Figure 18 :
Figure 17: Maximum and minimum conductivities of InO  films prepared at different substrate temperatures during photoreduction/reoxidation process.Films were deposited by DC-magnetron sputtering.Photoreduction was carried out in vacuum using UV Hg lamp.Adapted with permission from [173].Copyright 2001 American Institute of Physics.

Figure 19 :
Figure 19: Response of In 2 O 3 sensor at room temperature to 2.4 ppm ozone in dry atmosphere with and without illumination by a blue LED ( = 460 nm) ( ∼ 5 m).Data extracted from [119].

Figure 20 :
Figure 20: Literature survey of water-related species formed at different temperatures on SnO 2 surfaces.Data extracted from [24].

Figure 22 :
Figure 22: Diagram illustrating difference in structures of the films prepared "thick" and "thin" film technologies.Reprinted with permission from [57].Copyright 2009 Elsevier.

Figure 25 :
Figure 25: Dependencies of response time during ozone detection ( oper ∼ 270 ∘ C) on the grain size of In 2 O 3 films deposited from 0.2 M InCl 3 -water solution.Reprinted with permission from [43].Copyright 2004 Elsevier.

2 O 3
surface begins only at  > 500 ∘ C. In similar experiments, H 2 O began desorbing from the In 2 O 3 surface at  ∼ 210-230 ∘ C, reaching the maximum at  = 400 ∘ C.However, even at  ∼ 600 ∘ C, a sufficient quantity of H 2 O, perhaps in the form of OH-groups, remains on the In 2 O 3 surface.Thus, the surface species (O 2 and H 2 O) in chemisorbed forms (O − and OH) can exert an influence on the gas-sensing characteristics of In 2 O 3 gas sensors over the whole analyzed temperature range.

Figure 28 :
Figure 28: Diagram illustrating interrelation between temperature dependence of SnO 2 response to ozone and activation energies determined from the temperature dependence of time constants for the transient response observations.Reprinted with permission from [56].Copyright 2007 Elsevier.

Figure 29 :
Figure 29: Response and recovery times of RT sensor response to ozone in dependence on the layer thickness.Nanoparticle In 2 O 3 layer was grown at a substrate temperature of 200 ∘ C. Photoreduction was carried out using UV LED (375 nm).Adapted with permission from [176].Copyright 2007 American Institute of Physics.

Table 1 :
Comparison of main types of sensors, which can be used for ozone detection.

Table 3 :
The best parameters of conductometric heated ozone sensors designed on the base of metal oxides.

Table 4 :
Crystallographic planes, which can participate in gas-sensing effects.
[15,[169][170][171][172]rs at room temperature can exceed 50 min.The involvement of UV illumination in the gas-sensing process considerably improves the operating characteristics of RT ozone sensors[15,[169][170][171][172].Typical parameters of SnO 2 -and In 2 O 3 -based ozone sensors operated at room temperature under influence of UV illumination are listed in Table5.For comparison, the same table contains information related to the parameters of RT ozone sensors developed based on other metal oxides.

Table 5 :
Typical parameters of RT ozone sensors operated using UV photoreduction effect.
2 O 3 and the surrounding gas.The abovementioned statement that the surface stoichiometry of In 2 O 3 can control the surface diffusion processes is very important.Such a conception gives us an understanding of the nature of other effects observed during the In 2 O 3 sensor response kinetics study.For example, this conception explains the absence of transient processes controlled by the bulk oxygen diffusion in In 2 O 3 even for the transient sensor response measured at  oper > 350 ∘ C, when diffusion processes should occur.No doubt, further research is required for understanding in detail the role of diffusion and water in reactions of ozone detection.The question regarding oxygen vacancy participation in the reactions of ozone detection also requires an answer.Namely, is there a change in the concentration of surface oxygen vacancies during ozone chemisorption?