Photoactivated Processes on the Surface of Metal Oxides and Gas Sensitivity to Oxygen

Photoactivation by UV and visible radiation is a promising approach for the development of semiconductor gas sensors with reduced power consumption, high sensitivity, and stability. Although many hopeful results were achieved in this direction, the theoretical basis for the processes responsible for the photoactivated gas sensitivity still needs to be clarified. In this work, we investigated the mechanisms of UV-activated processes on the surface of nanocrystalline ZnO, In2O3, and SnO2 by in situ mass spectrometry and compared the obtained results with the gas sensitivity to oxygen in the dark and at UV irradiation. The results revealed a correlation between the photoactivated oxygen isotopic exchange activity and UV-activated oxygen gas sensitivity of the studied metal oxides. To interpret the data obtained, a model was proposed based on the idea of the generation of additional oxygen vacancies under UV irradiation due to the interaction with photoexcited holes.


Introduction
Semiconductor gas sensors, which operate on the principle of a reversible change in the electrical resistance in the presence of a detectable gas, are the most important promising sensors for monitoring the composition of the atmosphere, detecting toxic gases and explosive vapors in everyday life and in chemical production, in medical applications, and in many other situations [1][2][3][4]. The advantages of semiconductor gas sensors usually include the simplicity of their design, high sensitivity, miniaturization, and low cost. The disadvantages of semiconductor sensors can include the high power consumption associated with the need to heat the sensitive metal oxide layer to a temperature of 150-500 • C and the drift of the sensor parameters due to the influence of diffusion and recrystallization processes during operation at high temperatures [5,6]. A promising approach to lowering the operating temperature of semiconductor sensors, which was developed significantly in the last decade, is using UV or visible radiation for the activation of and enhancing the gas sensitivity [7][8][9].
Understanding the mechanisms that underlie the gas sensitivity of metal oxides is essential for the targeted development of sensor materials with desired properties [10]. To describe the mechanism of the gas sensitivity of metal oxide sensors, two main models are known. According to the first, (1) the electrical conductivity of the sensing layer is controlled by chemisorbed oxygen species (O -2 , O 2 -2 , O -, O 2 -, and others) on the surface of metal oxide grains [11,12], and to the second, (2) the electrical conductivity of the sensing layer is controlled by charged oxygen vacancies (V + O and V 2+ O ) near the surface of metal oxide grains [13,14]. Thus, in both models, the interaction of the surface of metal oxides with oxygen occupies a central place in explaining the occurrence of a sensor response to various gases. Surface redox reactions to which involved the molecules of detected gases leads to a change in the concentration of chemisorbed oxygen (oxygen vacancies), whereupon the electrical properties of the sensor layer are changed.

Synthesis on Nanocrystalline Oxides
(1) Nanocrystalline ZnO was synthesized by decomposition of basic zinc carbonate (#96466, Sigma-Aldrich, St. Louis, MO, USA) in air at 300 • C for 24 h according to the following reaction: (2) Nanocrystalline SnO 2 was synthesized at 70 • C by slowly adding concentrated ammonia solution to a SnCl 4 · 5 H 2 O (#244678, Sigma-Aldrich, St. Louis, MO, USA) solution in ethanol with stirring until neutral pH was achieved. The reaction resulted in the formation of α-stannic acid gel: Resulting gel was separated by centrifugation, then the precipitate was redispersed in 3%-NH 4 NO 3 solution and again sedimented by centrifugation to remove the chloride ions. More than 12 washing cycles were performed until a negative test for chloride ions with AgNO 3 solution was achieved. Finally, the SnO 2 · nH 2 O precipitate was washed with ethanol, dried at 70 • C, and annealed at 500 • C in air for 24 h.
(3) Nanocrystalline In 2 O 3 was synthesized at 70 • C by slowly adding concentrated ammonia solution to an In(NO 3 ) 3 hydrate (#326135, Sigma-Aldrich, St. Louis, MO, USA) solution in ethanol with stirring until neutral pH was achieved.

Characterization
Phase composition and crystal structure of synthesized metal oxides was studied by powder X-ray diffraction (XRD) with a Rigaku Max-2500 diffractometer using CuKα radiation (wavelength λ = 1.54059 Å). Average crystallite size D was calculated using the Sherrer equation: where λ is a wavelength of X-ray radiation, nm; β exp is the observed peak width at half height and β app is the instrumental broadening, rad; θ is a diffraction angle; k is a coefficient equal to 0.89 [42]. The specific surface area of nanocrystalline ZnO was measured by low-temperature nitrogen adsorption using the BET model with the Chemisorb 2750 instrument (Micromeritics Instrument Corporation, Norcross, GA, USA).
Raman spectra were registered with a SENTERRA Raman microscope-spectrometer (Bruker, Billerica, MA, USA) using a 50× 0.75 NA microscope objective lens and laser excitation at 785 nm.

Gas-Sensing Measurements
The method for examining the gas-sensitive properties of nanocrystalline oxides and the setup for sensor measurements are described in detail in our previous work [43]. Briefly, the gas-sensitive properties of the synthesized oxides were studied by in situ measurement of the electrical resistance of thick (about 100 µm) conducting polycrystalline porous films. Alumina plates 1.5 × 1.5 mm in size with deposited Pt electrodes with a distance of 200 µm between them were used for measurements. On the reverse side of the plates, a platinum spiral was also deposited to heat the sensors with electric current. A thick conductive layer was formed on the measuring plates over the Pt electrodes by annealing at 300 • C a pasty mixture of metal oxide with α-terpineol. The obtained sensors have a linear current-voltage characteristic at bias up to 4 V [43].
To study the gas sensitivity to oxygen, high-purity nitrogen (99.999%) was used as a carrier gas, which was diluted with high-purity oxygen (99.99%) in the concentration range of 0.2-3.2 vol.%. The measurements were carried out in a flow cell at a flow rate of 100 mL/min. A UV LED (λ max = 365 nm) was used for irradiation; the irradiance of the sensors during measurements was 6 mW/cm 2 . The resistance measurement was carried out on a laboratory-made setup at a bias voltage of 4 V. The sensor signal S both in the dark and under continuous UV irradiation was calculated as the ratio of the increase in sensor resistance in the presence of oxygen to its base resistance in dry nitrogen (background environment):

In Situ Ambient Pressure UV-Assisted Mass Spectrometry
The setup for in situ mass spectrometric measurements under dark conditions and under UV radiation is described in detail in our previous work [43]. Generally, it consists of a flow PTFE cell (internal volume 15 mL) with a quartz window for irradiation of samples, a set of UV LEDs for irradiation (λ max = 365 nm), an electric heater allows to heat samples up to 150 • C, gas and electrical circuits for controlling the cell. The outlet of the cell is connected to the capillary of the mass spectrometer (MS7-200, equipped with RGA-200 analyzer, Stanford Research Systems). During measurements, a carrier gas is passed through the cell, and the composition of the gas phase at the output of the cell is continuously monitored. Mass spectra were recorded in the range of 4-50 a.m.u. (one scan every 9 s) using an electron multiplier with a gain of 500-1000. All measurements were carried out at ambient pressure.
High-purity helium with an oxygen content of about 15 ppm was used as a carrier gas for oxygen photoadsorption experiments and the typical carrier gas flow rate through the cell was 15 mL/min. For POIE experiments, oxygen-18-enriched water (≥98%, "Center of Molecular Research", Moscow, Russia) was used. An electrolyte based on oxygen-18enriched water (saturated Na 2 SO 4 solution) was placed in an electrolyzer with platinum electrodes. During electrolysis, the current was set to 3 mA, and a carrier gas (helium with 15 ppm of naturally occurring oxygen-16) was passed through the anode space of the electrolyzer, then the carrier gas entered the drying column filled with 3A zeolites and, finally, into the measuring cell with samples. The concentrations of 16  For MS studies, the synthesized ZnO, SnO 2 , In 2 O 3 powders were pressed into tablets with a diameter of 20 mm and a thickness of 1 mm at a pressure of 60 kg/cm 2 and 4 tablets were usually placed in the cell; thus, the irradiated surface area of the samples was about 12.5 cm 2 , irradiance of samples at the experiments was 40-60 mW/cm 2 . Irradiance in MS and gas-sensing experiments was measured by Nova II radiometer equipped with photodiode PD300-UV-193 head (Ophir, Jerusalem, Israel).

Structure and Morphology of Synthesized Metal Oxides
Figure 1a-c show the diffraction patterns of the synthesized powders of the metal oxides. As a result of annealing at 300 • C, the ZnO phase is formed with a wurtzite structure (space group P6 3 mc, 36-1451 card, PDF2) and an average crystallite size of 15 nm, estimated from the broadening (101) reflection using the Scherrer equation. The specific surface area of the synthesized ZnO is 4 m 2 /g. The annealing of the alpha-stannic acid at 500 • C leads to a single-phase SnO 2 with a cassiterite structure (space group P4 2 /mnm, 41-1445 card, PDF2) [44]; the average crystallite size estimated from the broadening of the (110) reflection is 11 nm, and the specific surface area is 25 m 2 /g. Upon the annealing of the indium hydroxide at 500 • C, the In 2 O 3 phase with the bixbyite structure is formed (space group Ia3, 6-416 card, PDF2), the average crystallite size estimated from the broadening of the (222) reflection is 25 nm, and the specific surface area is 10 m 2 /g. Figure 1d-f show the Raman spectra of the synthesized metal oxides powders in the range of 100-900 cm −1 . The observed Raman modes confirm the phase composition of the synthesized samples. For ZnO, the most intense E 2 (high) mode is observed at 438 cm −1 , which corresponds to the vibrations of the oxygen sublattice [45]. Less intense modes E 2 (high) − E 2 (low) and 2E 2 (low) at 330 and 204 cm −1 , respectively, are also observed in the spectrum. In 2 O 3 demonstrates the most intense modes at 306 and 130 cm −1 , which correspond to vibrations of the In-O bonds in the InO 6 octahedrons and the bending vibrations of the InO 6 octahedrons. The other two peaks at 494 and 628 cm −1 refer to the stretching vibrations in the InO 6 octahedrons, while the mode at 364 cm −1 refers to the stretching vibrations of the In-O-In and can also depend on the presence of oxygen vacancies in the structure [46,47]. The spectrum of SnO 2 contains the main A 1g mode at 627 cm −1 and a broad Raman peak located around 561 cm −1 . The high ratio between the peak intensities at 561 and 627 cm −1 indicates a high concentration of surface oxygen vacancies on the surface of the SnO 2 . Moreover, a peak with a lower intensity is observed in the spectrum at 309 and 775 cm cm −1 , which correspond to the E u and B 2g Raman modes [48,49]. Figure 2 shows the mass spectra for m/z = 32 obtained from the analysis of a heliumoxygen carrier gas passed through a cell with samples of metal oxides at temperatures of 30, 50, 100, and 150 • C. UV irradiation was carried out at each temperature for 30 min. The tablets were annealed in air at 300 • C immediately before the measurements. For all three studied samples under UV light, a decrease in the oxygen concentration in the carrier gas was noted, which indicates a shift in the adsorption equilibrium (8) under UV radiation toward the formation of an adsorbed form of oxygen, i.e., the photoadsorption process. For ZnO, oxygen photoadsorption is clearly detected at room temperature, and with an increasing temperature, the photoadsorption rate slightly increases. For SnO 2 and In 2 O 3 , the change in the oxygen concentration in the carrier gas under UV light at room temperature is hardly detectable, but at temperatures above 100 • C, oxygen photoadsorption also becomes clearly observed. The oxygen photoadsorption rate decreases during irradiation, which can be explained by the gradual filling of the adsorption sites on the surface of the metal oxides. Figure 3a shows the results of the POIE experiments on the ZnO. As in the previous case, the experiments were carried out on the freshly annealed in air tablets at temperatures of 30, 50, 100, and 150 • C and under UV irradiation at each temperature for 30 min. A carrier gas that consisted of He + 16 O 2 + 18 O 2 was used for the POIE experiments.

Oxygen Photoadsorption and UV-Activated Oxygen Isotopic Exchange
Initially (in the dark), the signal by m/z = 34 in the mass spectrum has a background intensity, because the 16  The dependence of the POIE rate on the temperature for ZnO is clearly non-linear, decreasing upon heating to 50 • C compared to room temperature, increasing upon heating to 100 • C, and decreasing again at 150 • C. Apparently, this behavior is associated with the dependence of the rate of thermally activated oxygen desorption from the surface of ZnO on the temperature. It was also found that for ZnO after a cycle of measurements with heating to 150 • C and subsequent cooling to 30 • C, the POIE rate significantly increases compared to the POIE on the initial (annealed in air) sample at 30 • C (Figure 3a). Assuming that the carrier gas is mainly an inert gas (99.999% He), it can be argued that reductive annealing occurs, which most likely leads to an increase in the density of the surface oxygen vacancies. Figure 4 shows the results of a comparative POIE experiment on ZnO, In 2 O 3 , and SnO 2 carried out under similar conditions and also on samples freshly annealed in air. As can be seen, the studied metal oxides demonstrate the different activities in the POIE experiments. The most intense POIE occurred over the ZnO, which is observed in the temperature range of 30-150 • C, while the maximum rate is observed at a temperature of 100 • C. The POIE of a lower intensity (by 5-10 times in comparison with the ZnO) is observed over the In 2 O 3 , also in the temperature range from room temperature to 150 • C, and the rate is weakly dependent on the temperature. On the contrary, no POIE was observed over the SnO 2 in this experiment. The effect of an increase in the POIE rate at room temperature after the reductive annealing of the samples at 150 • C was also noted for the In 2 O 3 ; however, the POIE was still not detected on the surface of the SnO 2 , even after reductive annealing. It should also be noted that a thermally activated oxygen exchange was not observed in the studied temperature range.

CO 2 Photodesorption
During the experiments, it was found that under UV irradiation and heating from the surface of metal oxides, the photo-and thermal desorption of carbon dioxide occurs. CO 2 photodesorption was observed using both carrier gases containing the natural oxygen isotope 16     The CO 2 photodesorption from the surface of the In 2 O 3 at room temperature is practically not observed; weak photodesorption is observed starting from a temperature of 50 • C. In this case, when heated from 50 to 100 and from 100 to 150 • C, thermal desorption peaks are observed, which have a much higher intensity than the photodesorption peaks. The patterns of the photodesorption of the CO 2 molecules of various isotopic compositions from the surface of the In 2 O 3 are similar, although the intensity of the signal between the naturally and mixed isotope molecules is very low (approximately 1:60 for the photodesorption peak at 100 • C). SnO 2 , like In 2 O 3 , exhibits weak CO 2 photodesorption behavior, but photodesorption is observed in this case at room temperature. As the temperature increases, the photodesorption rate slightly increases, and typical thermal desorption peaks are also observed. The patterns of the photodesorption of the CO 2 molecules with different isotopic compositions from the surface are similar; the intensity of the signal from naturally and mixed isotope molecules is very low (approximately 1:60 for the photodesorption peak at 100 • C).
Thus, among the studied samples, the rate of photodesorption CO 2 at room temperature generally decreases in the series ZnO > SnO 2 > In 2 O 3 .   The dark electrical resistance of the sensors in a nitrogen atmosphere is different, about 10 8 Ω for ZnO, 10 5 Ω for SnO 2 , and 10 3 Ω for In 2 O 3 . With the introduction of oxygen at a concentration of up to 3.2 vol% at room temperature, only a slight increase in the resistance is observed for all three samples, which corresponds to a sensor signal of the order of 0.01-0.1. In addition, the kinetics of the response and recovery of the resistance of the sensors at room temperature are very slow and do not reach saturation during the oxygen supply period. An increase in the temperature leads to both an increase in the sensor signal and an acceleration of the kinetics of the response and decay times of the sensors, while good reversibility is observed only starting from a temperature of 150 • C. At the same time, the metal oxides show close values of the sensor signal to oxygen, for example, in dark conditions at a temperature of 100 • C, the sensor signals to 0.8 vol.% oxygen for oxides of ZnO, SnO 2 , and In 2 O 3 are 0.27, 0.37, and 0.16, respectively. An increase in the temperature to 150 • C leads to an increase in the sensor signal to 2.07, 1.19, and 2.04, but the obtained values are still close to each other.

Gas Sensitivity to Oxygen
The effect of UV irradiation on the electrical response of nanocrystalline ZnO, SnO 2 , and In 2 O 3 to oxygen is shown in Figure 7. It can be seen, firstly, that UV irradiation leads to a significant drop in the baseline resistance of the sensors in a nitrogen atmosphere, to 10 4 Ω for the ZnO and SnO 2 and up to 10 2 Ω for the In 2 O 3 . Moreover, even at room temperature, there is a significant sensor response to the oxygen and a decrease in the response and recovery times. Simultaneous heating to 150 • C and UV irradiation lead to an increase in the sensor signal of the samples compared to dark conditions at the same temperature. Both in the dark and under UV irradiation, the resistance of ZnO-, SnO 2 -, and In 2 O 3 -based sensors increases in the presence of oxygen relative to the base resistance in nitrogen, which implies a typical response of an n-type semiconductor to an oxidizing gas. UV irradiation leads to different effects on the sensor signal for three other metal oxides. Figure 8a shows the temperature dependence of the sensor signal to 0.8% oxygen in the dark (S dark ). It can be seen that the ZnO-, SnO 2 -, and In 2 O 3 -based gas sensors demonstrate a small, close-to-each-other sensor signal in a temperature range of 30-150 • C. With an increasing temperature, the sensor signal increases; in semilogarithmic coordinates, a dependence S dark (T) close to linear is observed, the slope of which is approximately the same for all three oxides. Figure 8b shows the temperature dependence of the sensor signal to 0.8% oxygen under UV irradiation (S UV ). It can be seen that the sensor signal for all the studied oxides increases significantly compared to the S dark and increases with an increasing temperature. For the SnO 2 and In 2 O 3 , the photoactivated sensor signal begins to increase significantly at temperatures above 100 • C. For them, the S UV (T) is close to linear in the semilogarithmic coordinates but with different slopes: for the In 2 O 3 , a steeper slope is observed, and for the SnO 2 , it is flatter. In the case of ZnO, there is an increase in S UV starting from 50 • C; however, at 150 • C, the sensor signal reaches a plateau, so the dependence is no longer linear. According to the absolute value of S UV , ZnO demonstrates the highest sensor signal toward oxygen compared to SnO 2 and In 2 O 3 .  Figure 8c shows the temperature dependence of the ratio S UV /S dark calculated from the data in Figure 8a,b. It can be seen that for the SnO 2 and In 2 O 3 , the S UV /S dark ratio decreases with an increasing temperature, because the S dark values increase faster with an increasing temperature than the S UV . The dependence S UV /S dark on the temperature for the ZnO has a fundamentally different character, for which a hump arises at temperatures of 50-100 • C, caused by an extremely sharp increase in the sensor signal under UV irradiation (about 40 times). It is also clearly seen in Figure 8b,c that the greatest enhancement of the sensor signal under UV irradiation takes place for ZnO; for In 2 O 3 , high ratios of S UV /S dark are also observed; but for SnO 2 , UV irradiation has the least effect on enhancing the sensor signal to oxygen. Figure 8d shows the concentration dependence of the sensor signal to oxygen both in the dark and under UV irradiation at 100 • C. In all the metal oxides, the dependence in the log-log coordinates is close to linear, differing slightly in the slope, i.e., it follows the equation S ∝ C k O 2 . For ZnO, k = 1 is observed in the dark and under UV; for In 2 O 3 , k is also the same in the dark and under UV and equals to 0.8; and for SnO 2 , k equals to 0.85 in the dark and 0.95 under UV.

Discussion
The MS data showed that on the surface of the ZnO, In 2 O 3 , and SnO 2 freshly annealed in air, the photoadsorption of oxygen is observed under UV irradiation. From the data obtained, it follows that these metal oxides are also characterized by the irreversible photoadsorption of oxygen under UV irradiation, while the photoadsorption rate increases with heating. At the same time, the photoadsorption of oxygen on SnO 2 and In 2 O 3 proceeds with a lower rate compared to ZnO. This behavior cannot be explained, for example, by the influence of the specific surface area of the samples, because ZnO, on the contrary, has the smallest specific surface area but the highest photoadsorption rate among the studied samples.
POIE experiments are a continuation of photoadsorption experiments, which allow a more detailed understanding of the mechanism of ongoing processes. The experimental results show that the POIE activity differs significantly for the studied metal oxides and decreases in the ZnO-In 2 O 3 -SnO 2 series. Because the natural 16 O isotope is contained both in the composition of the synthesized metal oxides and in the composition of the carrier gas, it is impossible to state whether the POIE occurs between oxygen from the gaseous and solid phases or whether the exchange occurs between oxygen molecules 16 O 2 and 18 O 2 simultaneously photoadsorbed from the gas phase on the metal oxide surface. However, it can be argued that when the POIE is taking place, photoadsorbed oxygen molecules undergo further breaking of the intermolecular bonds, as a result of which oxygen molecules of a mixed isotopic composition 16 O 18 O appear in the gas phase. In this case, the appearance of 16 O 18 O molecules in the carrier gas indicates that the process of the photodesorption of oxygen also occurs during UV irradiation, although, in general, the adsorption equilibrium is shifted toward the formation of a photoadsorbed form of oxygen.
The comparison of the results of the MS and gas-sensing measurements suggests that there is a correlation between the POIE activity of the metal oxide and the photoactivated sensor response to oxygen (Figure 9). ZnO shows the highest POIE activity and UV irradiation also most evidently enhances its sensor response to oxygen; on the contrary, SnO 2 does not show POIE activity and it is also characterized by the lowest sensor signal to oxygen and has the least effect of UV irradiation on it. In 2 O 3 occupies an intermediate position, demonstrating slight POIE activity and both a moderate sensor response to oxygen and the mean effect of UV irradiation on it.  An attempt to interpret the results obtained in terms of the known models encounters difficulties. In the case of the "ionosorption model", it is assumed that oxygen from the gas phase can be adsorbed on the grains' surface with the capture of a conduction electron: Apparently, in the low-temperature region (below 150 • C), chemisorbed oxygen retains the molecular form, while the existence of other charge forms of chemisorbed oxygen remains questionable. By extending the ionosorption model to the case of photoactivation, additional processes are possible with the participation of photogenerated charge carriers. Photogenerated electrons can cause the adsorption of an additional number of oxygen molecules, i.e., photoadsorption discussed above, while photogenerated holes are responsible for the photodesorption of chemisorbed oxygen molecules.
Thus, the ionosorption model does not provide special ways for the isotopic exchange of chemisorbed oxygen molecules (below 150 • C) and therefore cannot be used to adequately describe the photoactivated processes on the ZnO surface.
On the other hand, a well-developed kinetic model of the photoactivated oxygen exchange on the surface of metal oxides (ZnO, TiO 2 ) is known [29,50], according to which the POIE occurs at active centers, which are surface oxygen atoms that have captured a hole, O 2 - However, such a model developed for photocatalytic materials does not consider the effect of surface processes on the electrical conductivity of metal oxide, i.e., its gas-sensing properties. Moreover, the authors in [50] show that the POIE is much higher on oxidized TiO 2 than on the reduced sample, while we, in the case of ZnO and In 2 O 3 , are faced with the opposite phenomenon. In addition, the abovementioned model does not take into account the role of surface oxygen vacancies in the reactivity of metal oxides.
At present, many works draw attention to the fact that an increase in the density of oxygen vacancies in ZnO enhances its reactivity, which is manifested in the enhancement of the catalytic, gas sensing, and other properties of V O -enriched ZnO [51]. For example, in the work by [52], it was also demonstrated that the photocatalytic activity of ZnO increases significantly when it is thermally treated in a vacuum at 240-260 • C, and the studies also indicated the role of oxygen vacancies in increasing the photocatalytic activity. In a previous article [43], we noted that the reductive annealing of ZnO in He at 150 • C leads to an increase in the oxygen photoadsorption rate. A possible mechanism is an increase in the density of surface oxygen vacancies, which are typical adsorption sites for oxygen molecules, which leads to an increase in the photoadsorption rate. The increase in the concentration of oxygen vacancies in ZnO as a result of reductive annealing was revealed by XPS. In this work, it was found that the reductive annealing of ZnO and In 2 O 3 in an inert gas at 150 • C also leads to an increase in the POIE rate at room temperature, compared with the POIE rate on freshly annealed in air samples. Thus, the key role of oxygen vacancies in the POIE and enhanced photoactivated oxygen-sensing conditions can be assumed.
An attempt to describe the processes of enhancing the oxygen sensor response under UV activation assisted with simultaneous POIE (which is most evident in the case of ZnO) can be based on a model of surface conductivity, which assumes that the conductivity of metal oxide is controlled by the concentration of surface oxygen vacancies. In the low-temperature region without photoactivation, oxygen chemisorption with the participation of doubly charged oxygen vacancies and conduction electrons can occur with the maintenance of its molecular form with forming the O 2 -V O intermediate at the first stage [13,53,54]: Varying the concentration of charged oxygen vacancies on the grains' surface affects the electron density in the electron accumulation layer in the near-surface region, which formed for screening the positive surface charge. Under a change in the partial pressure of oxygen, the equilibrium of the reaction (13) shifts in the corresponding direction, which provides a reversible sensor response to oxygen in the dark by the changing of the surface conductivity. The maintenance of the molecular form of adsorbed oxygen follows from the fact that under dark conditions up to 150 • C, a sensor signal to oxygen is observed for all metal oxides, but a thermally activated oxygen isotopic exchange is not registered. Under photoactivation, an additional number of electrons are generated that can enter into the above process (13) and cause the adsorption of an additional number of oxygen molecules. This path is the "conventional" way of photoactivation available for all the considered metal oxides.
Further, in the case of some metal oxides, a significant increase in the sensors' response can be described by a model based on the formation of an additional number of surface oxygen vacancies during UV irradiation.
For ZnO, the fact is known that UV irradiation can lead to its photolysis, which is a consequence of the oxidation of lattice oxygen by photoexcited holes: The ionization of extra vacancies generated during UV irradiation proceeds by the reaction with photogenerated holes: The formation of V 2+ O will lead to an increase in the concentration of adsorption sites for O 2 molecules and a more evident effect on the change in electrical conductivity under UV irradiation and in the presence of O 2 .
Oxygen vacancies on the surface of ZnO can also be considered as centers for the dissociative adsorption of oxygen molecules: Thus, photoadsorbing oxygen molecules of different isotopic compositions undergo photodissociation, filling the oxygen vacancies, and then the same oxygen atoms undergo photooxidation by holes according to Equation (14) with the formation of various isotopes of oxygen atoms, which, combined with each other during desorption, can release oxygen molecules of a mixed isotopic composition 16 One can also assume a possible process of photooxidation of the adsorbed oxygen molecule in the O 2 -V O intermediate, which will also lead to the breaking of bonds of the chemisorbed oxygen molecule and the formation of a vacancy: Thus, in the proposed model, photogenerated holes lead to the formation of new surface oxygen vacancies, which can be additionally oxidized to a doubly charged V 2+ O by the interaction with photoexcited holes (Equations (15) and (16)). The generation of an additional number of V 2+ O during UV irradiation contributes to a greater effect on the electrical conductivity at the adsorption of oxygen molecules, and in addition, they are the centers for the photoactivated oxygen dissociation according to the reactions (14)- (18). Oxygen vacancies, previously formed as a result of reductive annealing, immediately provide an increased number of sites for oxygen chemisorption, which leads to an increase in the POIE rate on the reduced samples.
Accordingly, a similar (but less pronounced) mechanism for the generation of additional oxygen vacancies during the irradiation period by means of photogenerated holes can be expected in the case of In 2 O 3 , while for SnO 2 , such a mechanism does not occur, which leads to a poor photoactivated sensor signal to oxygen and the absence of POIE.
The proposed assumptions are confirmed by many literature data on experimental and theoretical studies of processes involving oxygen on the surface of metal oxides. Photolytic ZnO decomposition in a vacuum under UV light was intensively studied by mass spectrometry and manometry in the 1970s-1980s of the XX century (see, for example, the references in [55]), and it was found that about 100 surface monolayers undergo photodecomposition. In a recently published work [56], the authors reported an increase in the defective luminescence of ZnO upon prolonged exposure to UV light, which may also correspond to the formation of additional oxygen vacancies due to photodecomposition. Concepts related to the effect of photodecomposition on surface hydroxylation and structural changes in ZnO sol-gel films were also developed by Asakuma et al. [57,58]. The possibility of oxygen adsorption on ZnO proceeding with the filling of an oxygen vacancy was experimentally demonstrated using in situ EPR and PL spectroscopy [59]. The calculations given in the work [60,61] show that on the reduced surface of ZnO, dissociative oxygen adsorption is preferable to molecular adsorption. In addition, extremely low activation barriers for surface oxygen adsorption and bulk diffusion were estimated to be 0.093 and 0.67 eV, respectively [62].
Indium oxide is a much less explored material than zinc oxide. There are only some references in the literature about the possibility of In 2 O 3 photodecomposition under UV light [63,64]. Some calculations also confirm the preference for the dissociative adsorption of oxygen on its surface [65]. No information was found in the literature on the possibility of the photolytic decomposition of SnO 2 under UV irradiation. The calculated data indicate both a dissociative and molecular character of the oxygen adsorption on the surface of SnO 2 [66][67][68].
It is also interesting to note that the reactivity of the studied metal oxides to oxygen under UV activation correlates with the metal-oxygen bond energy. Thus, for ZnO, the Zn -O bond energy is 10.4 eV; for In 2 O 3 , the In -O bond energy is 12.4 eV; and for SnO 2 , the Sn -O bond energy is 15.5 eV. In general, the lower metal-oxygen bond energy implies an easier loss of oxygen from the crystal lattice and the formation of oxygen vacancies. The given bond energy data fit into the sequence ZnO > In 2 O 3 > SnO 2 , which also follows for the POIE activity and UV-enhanced sensor response of the studied oxides. The concept of the relationship between the metal-oxygen bond energy and the surface reactivity and the concentration of active centers was considered in detail by Marikutsa et al. [69], and the method for calculating the metal-oxygen bond energy is also given there. Within these ideas, it can be noted that, for example, WO 3 is characterized by a much higher metaloxygen bond energy (36.6 eV) and is a suitable example for studying and testing the model put forward in this article.
It is worth mentioning that some articles also discuss the mechanism of a photoassisted oxygen release through the oxidation of carbon impurities and the desorption of CO 2 [70]. The studies carried out by the method of mass spectrometry showed that, indeed, the photodesorption of CO 2 from the surface of metal oxides, in particular ZnO, is observed. Taking into account the possibility of the photodesorption of CO 2 also provides a good explanation for why the photoadsorption of oxygen on the surface of oxides is allegedly irreversible. In fact, one of the channels for the release of adsorbed oxygen is its desorption in the form of CO 2 molecules. It is interesting to note that for ZnO, the photodesorption of CO 2 is maximum at temperatures from room temperature to 50 • C, whereas at 100 • C, the CO 2 photodesorption almost ceases, but the POIE rate, on the contrary, increases, which may indicate that the photooxidation of carbon with the formation of CO 2 and the photodesorption of oxygen are competing processes. However, unlike oxygen, mixed isotopic molecules C 16 O 18 O are registered in the photo-and thermal desorption patterns of all three oxides, which may indicate different pathways from their origin. There is also no clear correlation between the oxygen-sensing properties of metal oxides and the photoand thermally stimulated CO 2 release.

Conclusions
As a result, in this work, using nanocrystalline ZnO, In 2 O 3 , and SnO 2 , we have shown that there is a correlation between the POIE activity and the photoactivated sensor response to oxygen under UV irradiation (365 nm). According to the data obtained, in the series ZnO-In 2 O 3 -SnO 2 , these characteristics decrease. ZnO shows outstanding POIE activity and an enhanced gas sensitivity to oxygen under UV irradiation, which can be explained by the unique lability of oxygen in its composition. The data obtained are discussed within the framework of a model that assumes the formation of additional oxygen vacancies during UV irradiation through an interaction with photoexcited holes. Additional oxygen vacancies enhance the sensor response to oxygen and act as centers for the photodissociation of chemisorbed oxygen molecules. The proposed model can be supplemented and modified by expanding the range of studied metal oxides and non-oxide materials, including such promising materials as WO 3 , TiO 2 , Ga 2 O 3 , BiOX (X = Cl, Br, and I), and others.