Copper-Containing Films Obtained by the Simple Citrate Sol–Gel Route for NO2 Detection: Adsorption and Kinetic Study

The citrate sol–gel method was utilized for the fabrication of copper-containing films sensitive to NO2 gas. Effect of annealing temperature on the film phase composition, morphology, and sensor response was studied. X-ray diffraction reveals the formation of Cu2Cl(OH)3 phase at 250 ◦C and the CuO phase at 350 and 500 ◦C. It was found out that the films annealed at 250 ◦C and 350 ◦C showed the best sensor characteristics. The influence of thermal degradation on the probability of percolation effect in films and its connection to a decrease of gas sensitivity was studied. The kinetics of the NO2 gas adsorption on the film’s surface was described following the Elovich model. Activation energy estimated from the ln(S) vs. 1/T plots was 252 and 30 kJ/mol for the films annealed at 250 and 350 ◦C, respectively.


Introduction
Nitrogen oxides are formed in the atmosphere, both natural and by anthropogenic combustion of fossil fuels. Air pollution with nitrogen oxides is generally relatively low. However, in areas with the developed chemical industry, there are local zones of increased NO X content in the air. The main anthropogenic sources of nitrogen oxides entering the atmosphere are the combustion of all types of natural fuels (12 million tons/year), transport (8 million tons/year), and industry (1 million tons/year). So, monitoring of NO 2 level is of considerable interest.
Currently, a kind of gas sensors, such as oxide semiconductors is developed [1][2][3]. Wherein, the oxide semiconductor gas sensor attracted attention due to its high sensitivity, low cost, and simple fabrication [4]. Copper oxide (CuO) as p-type material has attracted significant attention in last decade for its possible application as gas sensors [5][6][7][8][9][10][11][12][13][14][15]. It is evident that when producing gas-sensitive materials, it is necessary to increase and develop a surface adsorption center activity [16][17][18], which significantly depends on the technological conditions of manufacturing, i.e., type of precursors, reaction temperature, time, etc. In addition, the fabrication technique affects the particle size and morphology and electrophysical properties (band gap, resistance, activation energy, etc.). Materials based on copper oxide have promising physicochemical properties for gas sensor applications [15,19,20] due to their capability to form nanostructures with a different particle's shape, such as cubes, sheets, stars, flowers, and others [21][22][23].
The gas-sensing capability of CuO also depends on the operation temperature, because it influences the response and recovery times and also the sensitivity of the sensors [24]. Working temperatures of 2 of 14 copper oxide-based sensors are ranged from 50 to 500 • C. A few works have been reported on the low-temperature gas sensors (<100 • C) [13,[25][26][27][28].
As the principle of the resistive sensors is based on changing conductivity, it is necessary to create a network of conductive particles inside the film's structure. The sol-gel method is suitable for realizing this idea [29,30]. One of the principal advantages of the sol-gel method is the ability to control the structure of the produced materials, but a limited amount of works have been reported on the effect of particle morphology on the gas-sensing properties p-CuO [10,31,32]. Besides, many unanswered questions have been left concerning the gas sensing mechanism of the CuO materials with the non-ordinary shape of particles.
Therefore, herein, we investigate the influence of technological conditions on the value of the band gap, particle shape, and gas sensor response to NO 2 at low operating temperatures. Besides, we studied the adsorption mechanism, depending on the surface particle morphology and conditions.

Characterization
A surface morphological study of the deposited film samples was carried out using scanning electron microscopy (LEO 1560, ZEISS). The SEM operating voltage was employed at 5 kV. X-ray diffraction (XRD) phase analysis was performed with an ARL X'TRA diffractometer equipped with a solid-state Si (Li) detector that registered selectively Cu Ka radiation (λ = 1.5418 Å). Phases were identified using the Powder Diffraction File (PDF-2), release 2006.
The band gap energy was determined from the optical absorption spectra recorded using a commercial spectrophotometer LEKI SS1207 (LEKI Instruments, Helsinki, Finland) and analyzed by the empirical relation Equation (1): where hv is the photon energy; Eg is the band gap energy; n is an exponent determined by the nature of the electron transition during the absorption process, i.e., n = 1/2 for direct transition and n = 2 for an indirect transition; A is constant; and α is the absorption coefficient defined by the Beer-Lambert's Equation (2) law: where d and Abs are the film thickness and film absorbance, respectively.

Gas Sensor Fabrication
Earlier, we have already published the technology of film fabrication [33,34]. In brief, copper-containing thin films were synthesized by a citrate sol-gel method using ethylene glycol as a dispersing agent. The sol was prepared by dissolving 3 wt. % CuCl 2 with a mixture of 10 mL ethanol and distilled water (10 mL). Further, the as-prepared sol was mixed with ethylene glycol in a volumetric ratio of 1:1. Then, citric acid of 10 wt. % concentration was dropped (0.25 mL) into the mixture to reach pH = 4. The resulting solution was stored at room temperature for 24 h. At the next step, the Si/SiO 2 substrate preskimmed in nitric acid (10 min) was immersed in the solution and remained inside it with periodic stirring for 7 days at room temperature. Then, the substrate with the deposited film was pulled up. The prepared films were first, dried at 150 C for 1 h, and then annealed at 250, 350, and 500 • C for 2 h in ambient air. The thickness of the annealed films at 250-500 • C was measured as about 0.2 mm by an interference optical method.
Finally, V/Cu/Ni layers were deposited via an interdigitated shadow mask onto the film surface to serve as the contact electrodes.

Gas Sensor Tests
The gas-sensitive properties of the films were tested to NO 2 inputs to be varied in the concentration range of 10-200 ppm in ambient air. The NO 2 concentration was changed using several cylinders of different concentrations. The experimental setup consisted of a gas block (V = 1 dm 3 ), a data-collection system RL-88AC (an analog-digital and digital-analog converter with the RealLab software), and a gas receiver block [35]. The sensor element was set on the heater inside the gas block. The RL-88AC data-collection system controlled the resistance by a resistance-voltage converter. The sensor can be heated to 350 ± 1 • C.
Gas response (S) vs. NO 2 gas is defined as S = (Ra−Rg)/Ra where Rg and Ra are film resistance measured in NO 2 and air atmosphere, respectively

Modeling
Modeling of the formation of branched structures was carried out in the MATLAB software environment. In a computer experiment, a square four-connected lattice of size n × n was used, with n = 100. So, the maximum number of initial particles was 104. The random walk method was utilized to organize the movement of particles when forming clusters. Both the movement of a single particle and the movement of small clusters are provided. The probability of sticking was calculated as shown elsewhere [34]. Each initial particle was labeled as copper-containing or containing only an organic base. The ratio of such particles was chosen based on experimental data. The marking was necessary for the possibility of removing such particles when simulating the thermal annealing process of the material. The mechanisms of diffusion-limited aggregation and cluster-cluster aggregation were used to model the aggregation process.

SEM and XRD Investigations
SEM micrographs reveal the evolution of surface morphology with increasing the annealing temperature (Figure 1). At 250 • C, the growth of fractal structures leads to the formation of percolation clusters that are destroyed at higher temperatures [34]. So, the S_250 sample exhibits the flower-like morphology. The "flowers" are made up of grain agglomerates with size of 400-500 nm. Agglomeration is the result of the high surface energy of particles [36]. As the temperature increases, the "flowers" destroy and the separate crystallites with the sizes ranging from to 100 to 400 nm form on the film surface (Figure 1a-c).
Chemosensors 2020, 8, x FOR PEER REVIEW 3 of 14 data-collection system RL-88AC (an analog-digital and digital-analog converter with the RealLab software), and a gas receiver block [35]. The sensor element was set on the heater inside the gas block. The RL-88AC data-collection system controlled the resistance by a resistance-voltage converter. The sensor can be heated to 350 ± 1 °C. Gas response (S) vs. NO2 gas is defined as S = (Ra−Rg)/Ra where Rg and Ra are film resistance measured in NO2 and air atmosphere, respectively

Modeling
Modeling of the formation of branched structures was carried out in the MATLAB software environment. In a computer experiment, a square four-connected lattice of size n × n was used, with n = 100. So, the maximum number of initial particles was 104. The random walk method was utilized to organize the movement of particles when forming clusters. Both the movement of a single particle and the movement of small clusters are provided. The probability of sticking was calculated as shown elsewhere [34]. Each initial particle was labeled as copper-containing or containing only an organic base. The ratio of such particles was chosen based on experimental data. The marking was necessary for the possibility of removing such particles when simulating the thermal annealing process of the material. The mechanisms of diffusion-limited aggregation and cluster-cluster aggregation were used to model the aggregation process.

SEM and XRD Investigations
SEM micrographs reveal the evolution of surface morphology with increasing the annealing temperature ( Figure 1). At 250 °C, the growth of fractal structures leads to the formation of percolation clusters that are destroyed at higher temperatures [34]. So, the S_250 sample exhibits the flower-like morphology. The "flowers" are made up of grain agglomerates with size of 400-500 nm. Agglomeration is the result of the high surface energy of particles [36]. As the temperature increases, the "flowers" destroy and the separate crystallites with the sizes ranging from to 100 to 400 nm form on the film surface (Figure 1(a-c)). Elsewhere [34], we have already discussed the phase composition of the fabricated films by the XRD and XANEX studies. So, according to the XRD study ( Figure 2), we showed the influence of annealing temperature on the evolution of the phase composition. Pattern S_250 differs from others and characterizes by the presence of various forms of Cu2Cl(OH)3, indistinguishable at the present level of resolution: atacamite (PDF 01-78-372), paratacamite (PDF 01-70-821), and clinoatacamite (PDF 01-86-1391). Patterns S_350 and S_500 are similar and correspond to the tenorite CuO phase (PDF 00-48-1548).
XANEX study was in a full agreement with XRD patterns and proved the formation of CuO crystallite at 350 and 500 °C. In addition, it was found that the CuK-edge XANES spectrum of Cu2Cl(OH)3 is closer to the S_250 sample [34]. Elsewhere [34], we have already discussed the phase composition of the fabricated films by the XRD and XANEX studies. So, according to the XRD study ( Figure 2), we showed the influence of annealing temperature on the evolution of the phase composition. Pattern S_250 differs from others and characterizes by the presence of various forms of Cu 2 Cl(OH) 3 , indistinguishable at the present level of resolution: atacamite (PDF 01-78-372), paratacamite (PDF 01-70-821), and clinoatacamite (PDF 01-86-1391). Patterns S_350 and S_500 are similar and correspond to the tenorite CuO phase (PDF 00-48-1548).
XANEX study was in a full agreement with XRD patterns and proved the formation of CuO crystallite at 350 and 500 • C. In addition, it was found that the CuK-edge XANES spectrum of Cu 2 Cl(OH) 3 is closer to the S_250 sample [34].

UV-Visible Studies
Because of the crystallinity and better transparency, the films are suitable for optical analysis from which the coefficient of absorption and energy band gap is also determined. Figure 3a shows the combination of optical absorbance spectra with a large series of peaks for all three samples. It is known that many organic compounds give more than one maximum peak when its UV-VIS spectra are analyzed. So, the presence of multiple peaks corresponds to the electron transitions in the organic matrix of the films.
The functional relationship between αhν and photon energy for the fabricated samples is presented in Figure 3b. The band gap energy value can be obtained by extrapolating the linear portion to the photon energy axis (inset Figure 3b).
Because Tauc plots ( Figure 3b) present a large series of peaks at lower energies, estimation of band gap energy was done according to the method described in [37]. So, the linear fit of the fundamental peak is applied. Additionally, a linear fit used as an abscissa is applied for the slope below the fundamental absorption. An intersection of the two fitting lines gives the band gap energy estimation (insets Figure 3b). The direct band gaps of S_250, S_350, and S_500 samples were calculated to be ca. 1.44, 1.55, and 1.47 eV, respectively ( Figure 3b). These values are close to the reported values of bulk CuO (1.4 eV) [38] and lower than the CuO films (1.72 eV) deposited by the chemical route and reported in [39]. This fact may be explained by the agglomeration of coppercontaining particles during the annealing process.

UV-Visible Studies
Because of the crystallinity and better transparency, the films are suitable for optical analysis from which the coefficient of absorption and energy band gap is also determined. Figure 3a shows the combination of optical absorbance spectra with a large series of peaks for all three samples. It is known that many organic compounds give more than one maximum peak when its UV-VIS spectra are analyzed. So, the presence of multiple peaks corresponds to the electron transitions in the organic matrix of the films.
The functional relationship between αhν and photon energy for the fabricated samples is presented in Figure 3b. The band gap energy value can be obtained by extrapolating the linear portion to the photon energy axis (inset Figure 3b).
Because Tauc plots (Figure 3b) present a large series of peaks at lower energies, estimation of band gap energy was done according to the method described in [37]. So, the linear fit of the fundamental peak is applied. Additionally, a linear fit used as an abscissa is applied for the slope below the fundamental absorption. An intersection of the two fitting lines gives the band gap energy estimation (insets Figure 3b). The direct band gaps of S_250, S_350, and S_500 samples were calculated to be ca. 1.44, 1.55, and 1.47 eV, respectively (Figure 3b). These values are close to the reported values of bulk CuO (1.4 eV) [38] and lower than the CuO films (1.72 eV) deposited by the chemical route and reported in [39]. This fact may be explained by the agglomeration of copper-containing particles during the annealing process.

Gas Sensitivity
The film resistance was decreased under NO2 exposure at different operating temperatures. Gas sensing characteristics of the samples at operating temperatures from 20 to 150 °C are shown in Figure  4a-c. As evident from Figure 4d, a maximum response is observed at 45, 75, and 20 °C for the films, annealed at 250, 350, and 500 ℃, respectively. Comparative studies of sensors response to NO2 gas in the concentration range of 10-200 ppm showed that the sample S_250 has the best response to NO2. However, it may be due to the chemical activity of the transition Cu2Cl (OH)3 phase observed by XRD studies and flower-like morphology with high surface energy. Besides, the presence of OH-group allows the creation of additional adsorption centers as is shown earlier in [40].
Also, we declare the difference in gas sensor response between S_350 and S_500 samples.

Gas Sensitivity
The film resistance was decreased under NO 2 exposure at different operating temperatures. Gas sensing characteristics of the samples at operating temperatures from 20 to 150 • C are shown in Figure 4a-c. As evident from Figure 4d, a maximum response is observed at 45, 75, and 20 • C for the films, annealed at 250, 350, and 500°C, respectively. Comparative studies of sensors response to NO 2 gas in the concentration range of 10-200 ppm showed that the sample S_250 has the best response to NO 2 . However, it may be due to the chemical activity of the transition Cu 2 Cl (OH) 3 phase observed by XRD studies and flower-like morphology with high surface energy. Besides, the presence of OH-group allows the creation of additional adsorption centers as is shown earlier in [40]. The explanation of this fact is related to several competitive processes that influence the response of the material to the target gas. On the one hand, at low temperatures (20 °C), the probability of filling oxygen vacancies with water is higher [41]; a monolayer of water is formed on the surface of the film as a result of chemisorption. As shown in the previous work [40], this phenomenon has a positive effect on NO2 adsorption, according to Equation (3). Equation (4) also holds for S_250 sample which differs in the presence of OH-groups. This effect also occurs for the S_500 sample, which shows maximum sensitivity at 20 °C. 4NO2(g) +3ē +2H2O(ads) +O2‾ (ads) →4NO2 − OH‾ (ads) NO2(g) + ē +OH → NO2−OH‾(ads) (4) On the other hand, when the temperature increases, the activation energy of chemosorption becomes less and the number of adsorption sites occupied by the adsorbed oxygen increases [13]. As a result, reaction 5 is difficult to proceed, and the number of sites occupied by nitrogen dioxide decreases. This effect can be useful when creating sensors for reducing gases based on CuOcontaining materials. Not only the number of active chemosorption sites and the value of the activation energy of the process but also the organization of interaction of these sites must effect the conductivity and gas sensor response of the material. Hierarchically organized structures having a small level of sensor material doping doped with metal oxides exhibit percolation effects of conductivity [42][43][44]. As shown in the SEM images (Figure 1), copper-containing compounds represent branched fractal structures. The influence of thermal degradation on the probability of percolation effect in structures was studied using the model described in [34]. So, it is the structure destruction that can be explained by the decrease in sensitivity of S_500 sample compared to S_350 one. In Figure 5, model representations of the evolution of copper-containing structures at the initial moment (S_350) (Figure 5a) and after the final annealing (S_500) are shown (Figure 5b). Also, we declare the difference in gas sensor response between S_350 and S_500 samples. The explanation of this fact is related to several competitive processes that influence the response of the material to the target gas. On the one hand, at low temperatures (20 • C), the probability of filling oxygen vacancies with water is higher [41]; a monolayer of water is formed on the surface of the film as a result of chemisorption. As shown in the previous work [40], this phenomenon has a positive effect on NO 2 adsorption, according to Equation (3). Equation (4) also holds for S_250 sample which differs in the presence of OH-groups. This effect also occurs for the S_500 sample, which shows maximum sensitivity at 20 • C.
On the other hand, when the temperature increases, the activation energy of chemosorption becomes less and the number of adsorption sites occupied by the adsorbed oxygen increases [13]. As a result, reaction 5 is difficult to proceed, and the number of sites occupied by nitrogen dioxide decreases. This effect can be useful when creating sensors for reducing gases based on CuO-containing materials. Not only the number of active chemosorption sites and the value of the activation energy of the process but also the organization of interaction of these sites must effect the conductivity and gas sensor response of the material. Hierarchically organized structures having a small level of sensor material doping doped with metal oxides exhibit percolation effects of conductivity [42][43][44]. As shown in the SEM images (Figure 1), copper-containing compounds represent branched fractal structures. The influence of thermal degradation on the probability of percolation effect in structures was studied using the model described in [34]. So, it is the structure destruction that can be explained by the decrease in sensitivity of S_500 sample compared to S_350 one. In Figure 5, model representations of The dependence of the probability of percolation on the degradation of the organic component of the fractal structure is shown in Figure 6. The probability was calculated as the ratio of found cases of percolation pathways to 1000 simulations of the degradation process at a given percentage of the organic base removed. Dependence has an inflection point. The point "40%" corresponding to the annealing temperature of 350 °C is in the area before the inflection, and the percolation effect has a high probability. At 500 °C, all organic base is removed, and the probability of percolation effect is zero. So, we observe a decrease in gas sensitivity. According to our investigation, we assume that the The dependence of the probability of percolation on the degradation of the organic component of the fractal structure is shown in Figure 6. The probability was calculated as the ratio of found cases of percolation pathways to 1000 simulations of the degradation process at a given percentage of the organic base removed. Dependence has an inflection point. The point "40%" corresponding to the annealing temperature of 350 • C is in the area before the inflection, and the percolation effect has a high probability. At 500 • C, all organic base is removed, and the probability of percolation effect is zero. So, we observe a decrease in gas sensitivity. According to our investigation, we assume that the optimal annealing temperature for fabrication enhanced gas-sensitive films by the citrate sol-gel technique is 250 • C. optimal annealing temperature for fabrication enhanced gas-sensitive films by the citrate sol-gel technique is 250 °C.

Activation Energy and Adsorption Kinetics Study
Adsorption kinetics plays a very important role in devising and developing adsorbent materials for industrial applications. So, understanding the activated chemical adsorption of a gas on sensor surface becomes crucial and such a study will be followed using the Elovich model [14].
where q is the gas amount adsorbed during time t. The ά is the initial adsorption rate constant and α is the constant related to a measure of a potential barrier for successive adsorption. The constants can be obtained from the slope and the intercept of a straight-line plot of Ct-Co vs. ln(t). Co and Ct are the conductances of the films in air and on exposing NO2, respectively. Such plots are shown in Figure 7 for NO2 adsorption on the surfaces of S_350 and S_250 samples at optimal operating temperatures. The extrapolation was done in Excel software using a trendline. The Figure 7 reveals the linear plot in the interval from ln2 to ln4.5 that confirms the validity of the Elovich model (Equation (6)). Elovich model deals with multilayer adsorption, based on a kinetic principle that adsorption sites increase exponentially with adsorption [45]. So, at the initial time, the

Activation Energy and Adsorption Kinetics Study
Adsorption kinetics plays a very important role in devising and developing adsorbent materials for industrial applications. So, understanding the activated chemical adsorption of a gas on sensor surface becomes crucial and such a study will be followed using the Elovich model [14].
where q is the gas amount adsorbed during time t. Theά is the initial adsorption rate constant and α is the constant related to a measure of a potential barrier for successive adsorption. The constants can be obtained from the slope and the intercept of a straight-line plot of C t -C o vs. ln(t). C o and C t are the conductances of the films in air and on exposing NO 2 , respectively. Such plots are shown in Figure 7 for NO 2 adsorption on the surfaces of S_350 and S_250 samples at optimal operating temperatures. The extrapolation was done in Excel software using a trendline. The Figure 7 reveals the linear plot in the interval from ln2 to ln4.5 that confirms the validity of the Elovich model (Equation (6)). Elovich model deals with multilayer adsorption, based on a kinetic principle that adsorption sites increase exponentially with adsorption [45]. So, at the initial time, the adsorption seems to be monomolecular. This is a reason why below ln(t)≈ 2 data plots are not on the fitting line. adsorption seems to be monomolecular. This is a reason why below ln(t)≈ 2 data plots are not on the fitting line. Constant α is responsible for a measure of the extent to which the surface has been screened by the potential barrier for successive adsorption. From Table 1 it is evident that in comparison with S_350 sample, α constant of S_250 sample is less that favored the adsorption on Cu2Cl(OH)3 particles. This fact is in agreement with experimental data (Figure 4d). The constant á depends on the activation energy and is regarded as the initial adsorption rate [14]. We see that á values for the S_250 sample are close and almost independent of concentration. The adsorption rate of the S_350 sample increases with concentration from 3.98•10 -9 to 6.04•10 -8 (Figure 7) Constant α is responsible for a measure of the extent to which the surface has been screened by the potential barrier for successive adsorption. From Table 1 it is evident that in comparison with S_350 sample, α constant of S_250 sample is less that favored the adsorption on Cu 2 Cl(OH) 3 particles. This fact is in agreement with experimental data (Figure 4d). The constant á depends on the activation energy and is regarded as the initial adsorption rate [14]. We see that á values for the S_250 sample are close and almost independent of concentration. The adsorption rate of the S_350 sample increases with concentration from 3.98·10 −9 to 6.04·10 −8 (Figure 7) The surface reactivity depends on the activation energy, which provides the availability of free energy for promoting surface reactions.
The plots of ln(S) vs. 1/T (the linearized form of the Arrhenius-like equation 7 [46]) are used to determine activation energies [47] for S_250 and S_350 samples exposed to 53 ppm of NO 2 ( Figure 8) by: Chemosensors 2020, 8, x FOR PEER REVIEW 11 of 14 The surface reactivity depends on the activation energy, which provides the availability of free energy for promoting surface reactions.
The plots of ln(S) vs. 1/T (the linearized form of the Arrhenius-like equation 7 [46]) are used to determine activation energies [47] for S_250 and S_350 samples exposed to 53 ppm of NO2 ( Figure 8) by: We observe a sudden change in the slope of the Arrhenius-like plots above 75 and 45 °C for S_350 and S_250 samples, respectively. The data for the investigated samples cannot be plotted as a single activated process. So, activation energy decreases from 30 to 9 kJ/mol and from 252 kJ/mol to 80 kJ/mol when the operating temperature increases, for S_350 and S_250 sample, respectively. Accordingly, the observed change in the activation energy can be ascribed to the high surface energy at low temperatures (˂75 °C) [48]. This type of dependence is consistent with the arguments presented above about several competitive processes that occur when gas is adsorbed to the film surface. The minimum point (optimal operating temperature) corresponds to the most favorable ratio of energy parameters of the NO2 adsorption to the surface sites (oxygen vacancies) of the CuO structure, also containing adsorbed water.

Conclusions
We reported on the copper-containing thin films that were synthesized via citrate sol-gel method using ethylene glycol as a dispersing agent. The films were annealed at 250, 350, and 500 °C. Effect on annealing and operating temperatures on the gas sensor response towards NO2 gas was studied. We observe the best response towards NO2 gas by S_250 sample due to the presence of We observe a sudden change in the slope of the Arrhenius-like plots above 75 and 45 • C for S_350 and S_250 samples, respectively. The data for the investigated samples cannot be plotted as a single activated process. So, activation energy decreases from 30 to 9 kJ/mol and from 252 kJ/mol to 80 kJ/mol when the operating temperature increases, for S_350 and S_250 sample, respectively. Accordingly, the observed change in the activation energy can be ascribed to the high surface energy at low temperatures (<75 • C) [48]. This type of dependence is consistent with the arguments presented above about several competitive processes that occur when gas is adsorbed to the film surface. The minimum point (optimal operating temperature) corresponds to the most favorable ratio of energy parameters of the NO 2 adsorption to the surface sites (oxygen vacancies) of the CuO structure, also containing adsorbed water.

Conclusions
We reported on the copper-containing thin films that were synthesized via citrate sol-gel method using ethylene glycol as a dispersing agent. The films were annealed at 250, 350, and 500 • C. Effect on annealing and operating temperatures on the gas sensor response towards NO 2 gas was studied.
We observe the best response towards NO 2 gas by S_250 sample due to the presence of Cu 2 Cl(OH) 3 phase and flower-like morphology. The phenomenon of higher gas sensitivity of S_350 compared to the S_500 sample was declared. We explained this fact by percolation in the films annealed at 350 • C. During a kinetic study, we showed that the NO 2 adsorption on the surface of the fabricated films is described by the Elovich equation.
So, we conclude that the optimal temperature for manufacturing CuO-films sensitive to NO 2 gas by citrate sol-gel route is 250 • C when Cu 2 Cl(OH) 3 phase is formed. The stability of such sensors requires additional research.

Funding:
The research was carried out with the financial support of the Ministry of science and higher education of the Russian Federation in the framework of the state task in the field of scientific activity No. 0852-2020-0015.