Cobalt Oxide Catalysts in the Form of Thin Films Prepared by Magnetron Sputtering on Stainless-Steel Meshes: Performance in Ethanol Oxidation

: Catalytic total oxidation is an e ﬀ ective procedure to minimize emissions of volatile organic compounds (VOC) emissions in industrial gases. Catalysts in the form of meshes are remarkable as they minimize the internal di ﬀ usion of reactants during the reaction as well as the need of expensive active components. In this paper, various conditions of radio frequency magnetron sputtering of cobalt on stainless-steel meshes was applied during catalyst preparation. Properties of the supported Co 3 O 4 catalysts were characterized by SEM, XRD, temperature programmed reduction (H 2 -TPR), FTIR, XPS, and Raman spectroscopy. Catalytic activity was examined in deep oxidation of ethanol chosen as a model VOC. Performance of the catalysts depended on the amount of Co 3 O 4 deposited on the supporting meshes. According to speciﬁc activities (the amounts of ethanol converted per unit weight of Co 3 O 4 ), smaller Co 3 O 4 particle size led to increased catalytic activity. The catalyst prepared by sputtering in an Ar + O 2 atmosphere without calcination showed the highest catalytic activity, which decreased after calcination due to enlargement of Co 3 O 4 particles. However, speciﬁc activity of this catalyst was more than 20 times higher than that of pelletized commercial Co 3 O 4 catalyst used for comparison.


Introduction
Catalytic total oxidation is an effective procedure to minimize emissions of volatile organic compounds (VOC) emissions in industrial gases [1,2]. We have demonstrated a high catalytic activity of cobalt oxide catalysts in oxidation of ethanol [3], which is very often emitted from pharmaceutical plants or printing works. High activity of cobalt oxide catalysts was also observed in total oxidation of other organic compounds, like propane [4][5][6], toluene [5,7], CO [8], or formaldehyde [9]. Formerly we have reported a remarkable oxidation activity of cobalt-containing oxides deposited on anodized aluminum foils [10].
Catalysts intended for fast chemical reactions have to work at short residence times, i.e., at high throughputs of reaction mixture. Under such conditions, pelletized catalysts show high pressure drops, and thus, catalysts bed of high free volume is necessary. Various forms of catalysts are used for this type of reaction: Monoliths, foams, meshes, etc. The well-known application of a metal gauze catalyst is the oxidation of ammonia with air to nitric oxide used in the production of nitric acid and Surface areas of the stainless-steel-supported cobalt oxide catalysts were low, as it turned out by krypton physisorption at −196 • C; the S BET values of 0.0038 and 0.0042 m 2 g −1 were found for the Co-RF-1 and Co-RF-3 samples, respectively. Surface area of the sample prepared by pelletizing of the commercial Co 3 O 4 was also low (2.2 m 2 g −1 ) but more than 500 times higher than those found for the samples prepared on the stainless-steel meshes.
Surface morphology of cobalt oxide layers deposited on stainless-steel meshes is demonstrated in Figure S1 in Supplementary. SEM images of the samples prepared by RF magnetron sputtering of cobalt in an inert argon atmosphere followed by heating in air showed rather compact and homogeneous coatings with small pores; very likely, the formed coatings consist of the interconnected oxidized cobalt particles resembling foams (Figure S1a-c). Increasing amount of cobalt oxides on the meshes due to prolonged magnetron sputtering procedure resulted in deposition of thicker layers but their structure remained identical. The Co-RF-ox/ap sample prepared by RF magnetron sputtering of cobalt in oxidation Ar+O 2 atmosphere showed quite different morphology of the arising cobalt oxide particles as isolated platelets of cobalt oxides can be seen on the surface ( Figure S1d). After calcination of this sample at 500 • C in air, the morphology was changed and globular aggregates of very fine crystals of cobalt oxide can be seen in the SEM image of the Co-RF-ox catalyst ( Figure S1e). SEM image of the Co-I catalyst prepared by impregnation of stainless-steel meshes with cobalt nitrate aqueous solution and subsequent heating at 500 • C in air showed relatively non-homogeneous coating of the wires surface with clusters of globular cobalt oxide particles ( Figure S1f) and, therefore, higher tendency of the coating to peeling can be expected.
Adhesion of deposited cobalt oxide to the stainless-steel support was evaluated by a simple test, when the catalysts were immersed in an ultrasonic bath at room temperature and treated for 10 min in acetone; the weight loss was then determined after drying the samples and related to initial weight of the deposited cobalt oxides (Table 1). The catalysts prepared by magnetron sputtering showed very good adhesion to the support as they exhibited only low weight loss of several percent; the lowest value was found for the Co-RF-ox/ap sample (2.9% related to weight of deposited cobalt oxide). Adhesion of the coatings decreased with increasing content of cobalt oxide in the catalysts. Larger crystals observed in the SEM images indicated higher tendency to be removed from the surface in comparison with the smaller ones. Both catalysts prepared by cobalt sputtering in the oxidation Ar+O 2 atmosphere (the Co-RF-ox/ap and Co-RF-ox catalysts) showed the highest adhesion while the Co-I catalyst prepared by impregnation the lowest one; weight loss of 60% of the deposited oxide was found for this sample.
Powder X-ray diffraction is important technique for the investigation of bulk structure, which is very important, as many of the catalyst characteristics depend on it. Results of the XRD measurements are shown in Figure 1. The diffraction peaks were ascribed to Co 3 O 4 (PDF 04-006-3982); no peaks corresponding to CoO or metallic Co were observed. It indicated practically complete oxidation of the cobalt coatings obtained by magnetron sputtering in inert argon atmosphere after subsequent heating in air, as well as direct oxidation of sputtered cobalt particles in Ar+O 2 atmosphere. The distinct sharp diffraction peaks at approximately 51 and 60 • 2θ corresponded to a Fe0.65Cr0.1Ni0.25 phase (PDF 04-019-2390). They were ascribed to the stainless-steel support. Full width in half maximum (FWHM) values were evaluated from the measured powder XRD patterns and mean coherence lengths of Co 3 O 4 in [311] direction were calculated (Table 1) using Scherrer's formula D = k λ/(β cosθ), where λ is the X-ray wavelength, β is the full width at half maximum (FWHM) of the diffraction peak corresponding to (311) plane of Co 3 O 4 , θ is the corresponding diffraction angle, and k is a constant (k = 0.9 was used in calculation). The smallest D value of 7.8 nm was found for the Co-RF-ox/ap sample prepared by magnetron sputtering of cobalt in Ar+O 2 atmosphere. The mean coherence length increased to 23.5 nm after heating of this sample at 500 • C in air and thus, better structure ordering of the Co-RF-ox catalyst could be expected. The catalysts prepared by magnetron sputtering of cobalt in the inert argon atmosphere followed by heating in air showed very similar powder XRD patterns and only small differences in mean coherence lengths ranging from 32.5 to 38 nm. The largest D value of 75 nm was determined for the pelletized commercial Co 3 O 4 . SEM images of the pellets interior ( Figure S1g) also revealed aggregates of cobalt oxide particles with considerably larger size in comparison with those observed in the examined supported catalysts. Powder XRD pattern of the pelletized commercial Co 3 O 4 catalyst is shown in Figure 1 as well. This catalyst contained, apart from Co 3 O 4 , small amount of CoO (PDF 04-013-5837) admixture. Graphite (PDF 04-016-6288) was also detected, as it was added as a lubricant during pelletizing of the cobalt oxide.
Raman spectroscopy is a non-destructive technique providing detailed information about chemical structure, phase and polymorphy, crystallinity, and molecular interactions. Raman spectra of the examined catalysts are presented in Figure 2. Raman spectra of the catalysts prepared by magnetron sputtering showed five bands at 194, 484, 528, 620, and 692 cm −1 . As found by Hadjiev et al. [22] and Zhang et al. [23], five Raman-active modes and bands were identified at 194 (phonon mode F2g), 482 (Eg), 522 (F2g), 618 (F2g), and 691 (A1g) cm −1 modes of cubic-phase Co 3 O 4 . Similarly, Tang et al. [24] found analogous bands at 482, 519, 621, and 690 cm −1 in the Raman spectrum of Co 3 O 4 . The following Raman bands were reported for CoO: 672 and 468 cm −1 [22] and 675 and 455 cm −1 [24]. There is not a big difference between the spectra of Co-RF-2 and Co-RF-3 samples, but both differ to some extent from the spectrum of Co-RF-1 catalyst, which shows peaks slightly shifted to lower wavenumbers. With respect to the low concentration of cobalt oxides in the Co-RF-1 catalyst, the reason for the shift could be tighter bonds between the cobalt oxides and the support. Bands' intensities in the Raman spectrum of this catalyst are lower than those of the Co-RF-2 and Co-RF-3 catalysts are. Intensity of the bands was increasing with increasing Co 3 O 4 content. The low band intensity and broadening in the spectrum of Co-RF-1 sample could indicate smaller particles of cobalt oxide but SEM images of the Co-RF-x samples (Figure S1a Fourier transform infrared spectroscopy (FTIR) is a valuable non-destructive characterization technique used to identify various oxides including cobalt oxides. The FTIR spectrum of CoO is characterized by a transverse optical (TO) phonon below 400 cm −1 and by a higher frequency longitudinal optical (LO) phonon at 580 cm −1 [25]. Tang et al. [24] and Bao et al. [19]   Fourier transform infrared spectroscopy (FTIR) is a valuable non-destructive characterization technique used to identify various oxides including cobalt oxides. The FTIR spectrum of CoO is characterized by a transverse optical (TO) phonon below 400 cm −1 and by a higher frequency longitudinal optical (LO) phonon at 580 cm −1 [25]. Tang et al. [24] and Bao et al. [19]   Co-RF-2 Co-RF-1 5000 PEL Figure 2. Raman spectra of the examined supported cobalt oxide catalysts.
In our study, FTIR spectra were recorded for the cobalt oxide coatings prepared by magnetron sputtering on stainless-steel meshes, the coating obtained by calcination of the mesh impregnated with cobalt nitrate aqueous solution and for the pelletized Co3O4 ( Figure 3). The bands at 694, 649, 613, 547, and 505 cm −1 were observed in the spectrum of Co-RF-1 sample with the lowest Co3O4 content, obtained by magnetron sputtering in inert atmosphere and subsequent heating in air. In accord with report of Bao et al. [19], the bands at 694 and 613 cm −1 can be ascribed to Co3O4 (LO) and those at 649 and 555 cm −1 to Co3O4 (TO) bands. The band at 505 cm −1 could be assigned to CoO. In the sample with higher cobalt oxide content (Co-RF-3), the intensity of the bands at 510 and 649 cm −1 were slightly decreasing and a new band at 630 cm −1 appeared. Very likely, non-stoichiometric Co3O4 was partly originating during preparation of thicker cobalt oxide coatings [26].
The as-prepared Co-RF-ox/ap sample obtained by magnetron sputtering in oxidation atmosphere showed bands of lower intensity than those previously mentioned. The reason for the observation can be smaller size of particles and a poor structure ordering indicated by powder XRD measurements. The positions of the two main bands (555 and 660 cm −1 ) remained identical as those in the samples prepared by sputtering in inert atmosphere and only the position of the 694 cm −1 band was shifted to lower wavenumbers (688 cm −1 ). After heating the Co-RF-ox/ap sample in air, the positions of the bands in the obtained Co-RF-ox catalyst did not change but their intensities became more pronounced due to improved structure ordering. The Co-I catalyst and pelletized commercial Co3O4 exhibited two very broad bands with maxima at 656 and 544 cm −1 , which originated, very likely, from superimposition of other bands due to high cobalt oxide contents in these catalyst.
Temperature programmed reduction (H2-TPR) is a widely used tool for the characterization of metal oxides, mixed metal oxides, and metal oxides dispersed on a support. The TPR method yields quantitative information of the reducibility of the oxides surface, as well as the heterogeneity of the reducible surface. The TPR profiles of all prepared catalysts are shown in Figure 4. According to former reports [27,28], reduction of Co3O4 proceeds in two steps: Co3O4 is reduced to CoO above 300 °C and CoO is subsequently reduced to cobalt at about 400 °C. Arnoldy et al. [29] observed the reduction of Co3O4 in one peak around 330 °C, and the peaks corresponding to partial reduction were almost indistinguishable. All our examined samples were reduced in one narrow peak with prevailing temperature maximum around 375 °C. The samples were completely reduced before reaching the temperature 500 °C and it means that all cobalt oxide can be utilized in ethanol oxidation [30]. A shoulder at about 400 °C indicates presence of a cobalt compound of lower reducibility, very In our study, FTIR spectra were recorded for the cobalt oxide coatings prepared by magnetron sputtering on stainless-steel meshes, the coating obtained by calcination of the mesh impregnated with cobalt nitrate aqueous solution and for the pelletized Co 3 O 4 ( Figure 3). The bands at 694, 649, 613, 547, and 505 cm −1 were observed in the spectrum of Co-RF-1 sample with the lowest Co 3 O 4 content, obtained by magnetron sputtering in inert atmosphere and subsequent heating in air. In accord with report of Bao et al. [19], the bands at 694 and 613 cm −1 can be ascribed to Co 3 O 4 (LO) and those at 649 and 555 cm −1 to Co 3 O 4 (TO) bands. The band at 505 cm −1 could be assigned to CoO. In the sample with higher cobalt oxide content (Co-RF-3), the intensity of the bands at 510 and 649 cm −1 were slightly decreasing and a new band at 630 cm −1 appeared. Very likely, non-stoichiometric Co 3 O 4 was partly originating during preparation of thicker cobalt oxide coatings [26]. likely CoO [23].  The as-prepared Co-RF-ox/ap sample obtained by magnetron sputtering in oxidation atmosphere showed bands of lower intensity than those previously mentioned. The reason for the observation can be smaller size of particles and a poor structure ordering indicated by powder XRD measurements. The positions of the two main bands (555 and 660 cm −1 ) remained identical as those in the samples prepared by sputtering in inert atmosphere and only the position of the 694 cm −1 band was shifted to lower wavenumbers (688 cm −1 ). After heating the Co-RF-ox/ap sample in air, the positions of the bands in the obtained Co-RF-ox catalyst did not change but their intensities became more pronounced due to improved structure ordering. The Co-I catalyst and pelletized commercial Co 3 O 4 exhibited two very broad bands with maxima at 656 and 544 cm −1 , which originated, very likely, from superimposition of other bands due to high cobalt oxide contents in these catalyst.
Temperature programmed reduction (H 2 -TPR) is a widely used tool for the characterization of metal oxides, mixed metal oxides, and metal oxides dispersed on a support. The TPR method yields quantitative information of the reducibility of the oxides surface, as well as the heterogeneity of the reducible surface. The TPR profiles of all prepared catalysts are shown in Figure 4. According to former reports [27,28], reduction of Co 3 O 4 proceeds in two steps: Co 3 O 4 is reduced to CoO above 300 • C and CoO is subsequently reduced to cobalt at about 400 • C. Arnoldy et al. [29] observed the reduction of Co 3 O 4 in one peak around 330 • C, and the peaks corresponding to partial reduction were almost indistinguishable. All our examined samples were reduced in one narrow peak with prevailing temperature maximum around 375 • C. The samples were completely reduced before reaching the temperature 500 • C and it means that all cobalt oxide can be utilized in ethanol oxidation [30]. A shoulder at about 400 • C indicates presence of a cobalt compound of lower reducibility, very likely CoO [23]. In our TPR experiments, T max temperatures of the supported catalysts varied from 310 to 381 • C, while that one of the pelletized Co 3 O 4 was 426 • C. Nearly identical course of TPR was observed for all Co-RF-x samples and the commercial Co 3 O 4 catalyst. Reduction of the Co-RF-ox/ap sample proceeded at substantially lower temperatures (310 • C) showing small peaks at 275 and 184 • C. After calcination of this sample, the small peaks appearing at low temperatures disappeared.  The finding indicates that small particles are present in the Co-RF-ox/ap sample, which were converted to larger particles upon calcination and were reduced at 367 °C. The Co-I catalyst exhibited similar course of reduction as the Co-RF-ox; this can be explained by similar particle sizes and structure ordering of cobalt oxide particles in these two catalysts (Figure S1e-f, Table 1).
Amounts of hydrogen consumed during H2-TPR experiments in the temperature range from 25 to 900 °C and calculated per gram of cobalt oxides deposited on the meshes are summarized in Table  1. If the cobalt ions in cobalt oxide were only Co 3+ or only Co 2+ and reduced to Co 0 , the H2 consumption would be 16.6 or 13.3 mmol g −1 , respectively. In case of reduction of Co3O4, theoretical consumption of hydrogen for its reduction to metal would be 15.48 mmol H2 g −1 . As revealed in the H2-TPR results, The finding indicates that small particles are present in the Co-RF-ox/ap sample, which were converted to larger particles upon calcination and were reduced at 367 • C. The Co-I catalyst exhibited similar course of reduction as the Co-RF-ox; this can be explained by similar particle sizes and structure ordering of cobalt oxide particles in these two catalysts ( Figure S1e-f for the PEL, Co-RF-1, and Co-RF-2 catalysts, respectively, and from 14.7 to 14.0 mmol H 2 g −1 for the remaining catalysts. Surface analysis of the fresh and used (i.e., after the catalytic testing) Co-RF-ox/ap and Co-RF-ox catalysts was performed using XPS. Metals from the stainless-steel supports (Fe, Cr, Ni, and Mo) were not detected in the spectra, which points to the fact that the meshes were completely covered with cobalt oxides. A carbon tape used for the fixing of the meshes to the holder manifested itself in high concentration of C and O ( Table 2). As the samples used in the XPS measurements were in the form of stainless-steel meshes covered by cobalt oxides, charging of the samples could not be fully compensated by a carbon correction, which would be influenced by underlying carbon tape. Thus, it was necessary to calibrate the spectra according to oxygen (corresponding to binding energy of the metal-oxygen bond in metal oxide at 529.5 eV). Cobalt surface concentrations in the fresh and used catalysts were practically the same (Table 2). However, we found a difference between these values of the Co-RF-ox/ap and Co-RF-ox samples, as surface concentration of Co of the Co-RF-ox/ap was higher than that of Co-RF-ox sample. As the dimensions of the samples applied during XPS measurements were identical, the difference in surface concentrations of Co cannot be explained by variation of the parameter. The phenomenon can be elucidated by larger surface area or lower amount of carbon impurities adsorbed on the catalysts surface due to higher catalytic activity of the Co-RF-ox/ap samples. Surfaces of both samples were covered exclusively with Co 3 O 4 ( Table 3). The finding is documented by deconvolution of the Co 2p3/2 spectra to CoO and Co 3 O 4 contributions according to Biesinger [31]. The fresh Co-RF-ox sample contains a low additional amount of Co 2+ as follows from deconvoluted XPS spectrum ( Figure 5). Deconvolution of oxygen spectrum of the Co-RF-ox/ap sample revealed several peaks ( Table 3). The peak at 529.5 eV can be assigned to metal oxide (lattice oxygen, O 2-), the peak at 531.3 eV can be ascribed to low-coordinated or the adsorbed oxygen bonded to cobalt as O -. The next line with maximum at 532.3 eV can be ascribed to adsorbed OHspecies as well as to simple bond of oxygen to carbon. Bonds of oxygen to carbon can arise as bonds to carbon tape and other adventitious carbon contamination. The fourth peak observed at 533.1 eV can be assigned to adsorbed water. The deconvoluted O 1s spectra indicated that both the Co-RF-ox/ap and Co-RF-ox sample after catalytic tests contained more lattice oxygen than the fresh samples.  Figure 5. Deconvolution of Co 2p3/2 XPS spectra of the Co-RF-ox/ap and Co-RF-ox catalysts to CoO and Co3O4 contributions (their parameters are taken from Biesinger [31]).
The composition determined by XPS shows a higher O:Co ratio than 1.3 corresponding to Co3O4. The origin of this discrepancy is not easy to elucidate. It can be caused either by non-homogeneity or by different surface morphology of the sample, which could alter response factors used for calculation of composition. Either way, the O:Co ratio is similar for all samples except Co-RF-ox/ap. The increase of amount of oxygen bound to cobalt in the Co-RF-ox/ap can be partially explained by higher presence of Co 2+ (additional CoO) in the fresh sample.

Catalytic Activity and Selectivity
Deep oxidation of ethanol was used as a model catalytic reaction to examine catalytic activity and selectivity of the catalysts. Performance of the catalysts was characterized in the terms of T50 temperature, i.e., temperature, at which 50% ethanol conversion was observed.
Ethanol conversions in dependence on reaction temperature are shown in Figure 6. The curves exhibited an S-shape typical of light-off measurements. The reaction temperatures necessary for 50% ethanol conversion (T50) varied in the range from 210 to 291 °C (Table 4). For the Co-RF-x catalysts, the T50 value increased with decreasing amount of the cobalt oxides deposited on the meshes. Among the catalysts prepared by magnetron sputtering, the lowest T50 temperature (205 °C) was found for the non-calcined Co-RF-ox/ap catalyst. The PEL catalyst exhibited the lowest T50 The composition determined by XPS shows a higher O:Co ratio than 1.3 corresponding to Co 3 O 4 . The origin of this discrepancy is not easy to elucidate. It can be caused either by non-homogeneity or by different surface morphology of the sample, which could alter response factors used for calculation of composition. Either way, the O:Co ratio is similar for all samples except Co-RF-ox/ap. The increase of amount of oxygen bound to cobalt in the Co-RF-ox/ap can be partially explained by higher presence of Co 2+ (additional CoO) in the fresh sample.

Catalytic Activity and Selectivity
Deep oxidation of ethanol was used as a model catalytic reaction to examine catalytic activity and selectivity of the catalysts. Performance of the catalysts was characterized in the terms of T 50 temperature, i.e., temperature, at which 50% ethanol conversion was observed.
Ethanol conversions in dependence on reaction temperature are shown in Figure 6. Co-RF-ox/ap Co-RF-ox Intensity, a.u.
Binding energy, eV Figure 5. Deconvolution of Co 2p3/2 XPS spectra of the Co-RF-ox/ap and Co-RF-ox catalysts to CoO and Co3O4 contributions (their parameters are taken from Biesinger [31]).
The composition determined by XPS shows a higher O:Co ratio than 1.3 corresponding to Co3O4. The origin of this discrepancy is not easy to elucidate. It can be caused either by non-homogeneity or by different surface morphology of the sample, which could alter response factors used for calculation of composition. Either way, the O:Co ratio is similar for all samples except Co-RF-ox/ap. The increase of amount of oxygen bound to cobalt in the Co-RF-ox/ap can be partially explained by higher presence of Co 2+ (additional CoO) in the fresh sample.

Catalytic Activity and Selectivity
Deep oxidation of ethanol was used as a model catalytic reaction to examine catalytic activity and selectivity of the catalysts. Performance of the catalysts was characterized in the terms of T50 temperature, i.e., temperature, at which 50% ethanol conversion was observed.
Ethanol conversions in dependence on reaction temperature are shown in Figure 6. The curves exhibited an S-shape typical of light-off measurements. The reaction temperatures necessary for 50% ethanol conversion (T50) varied in the range from 210 to 291 °C (Table 4). For the Co-RF-x catalysts, the T50 value increased with decreasing amount of the cobalt oxides deposited on the meshes. Among the catalysts prepared by magnetron sputtering, the lowest T50 temperature (205 °C) was found for the non-calcined Co-RF-ox/ap catalyst. The PEL catalyst exhibited the lowest T50 The curves exhibited an S-shape typical of light-off measurements. The reaction temperatures necessary for 50% ethanol conversion (T 50 ) varied in the range from 210 to 291 • C (Table 4). For the Co-RF-x catalysts, the T 50 value increased with decreasing amount of the cobalt oxides deposited on the meshes. Among the catalysts prepared by magnetron sputtering, the lowest T 50 temperature (205 • C) was found for the non-calcined Co-RF-ox/ap catalyst. The PEL catalyst exhibited the lowest T 50 temperature in the set of examined catalysts. However, it can be seen from Figure 6 that a substantially higher temperature was necessary to apply to meet complete ethanol conversion to CO 2 with this catalyst in contrast to all other catalysts. Table 4. Physical-chemical and catalytic properties of the examined catalysts (reaction conditions: Space velocity F/W = 21 ± 1 L h g −1 , 760 ± 10 ppm of ethanol in air, temperature ramp 2 • C min −1 ). In case of oxidation catalysts, selectivity is often more important parameter than the T 50 temperature as the by-products formed during the reaction may be more detrimental for human health and environment than the original organic compound. The typical course of reaction components obtained during increasing reaction temperature at ethanol oxidation can be seen in Figure 7. The T 95 CO 2 (temperature necessary for 95% conversion of all organic compounds to CO 2 ) was therefore employed to assess the selectivity of the catalysts. Taking into account the objective of the catalytic reaction, i.e., the complete abatement of volatile organic compound, the T 95 CO 2 temperatures (Table 4) also reveal the order of efficiency of the catalysts. According to this parameter, the Co-RF-ox/ap sample is more efficient than the Co-I catalyst, followed by the Co-RF-ox catalyst. The Co-RF-x catalysts exhibited lower efficiency than their above-mentioned analogues. Finally, the PEL catalyst showed the lowest efficiency of all catalysts as it attained only 93% conversion to CO 2 at 399 • C. Table 4. Physical-chemical and catalytic properties of the examined catalysts (reaction conditions: Space velocity F/W = 21 ± 1 L h g −1 , 760 ± 10 ppm of ethanol in air, temperature ramp 2 °C min −1 ). In case of oxidation catalysts, selectivity is often more important parameter than the T50 temperature as the by-products formed during the reaction may be more detrimental for human health and environment than the original organic compound. The typical course of reaction components obtained during increasing reaction temperature at ethanol oxidation can be seen in Figure 7. The T95CO2 (temperature necessary for 95% conversion of all organic compounds to CO2) was therefore employed to assess the selectivity of the catalysts. Taking into account the objective of the catalytic reaction, i.e., the complete abatement of volatile organic compound, the T95CO2 temperatures (Table 4) also reveal the order of efficiency of the catalysts. According to this parameter, the Co-RF-ox/ap sample is more efficient than the Co-I catalyst, followed by the Co-RF-ox catalyst. The Co-RF-x catalysts exhibited lower efficiency than their above-mentioned analogues. Finally, the PEL catalyst showed the lowest efficiency of all catalysts as it attained only 93% conversion to CO2 at 399 °C.
As the amounts of active components in the catalysts slightly differed, the specific reaction rates at 200 °C were calculated to compare the activity of the catalysts ( Table 4). Performance of the Co3O4 species in the examined catalysts showed that the RF-x catalysts do not differ too much from each other. On the other hand, the RF-Co-ox/ap catalyst prepared by reactive magnetron sputtering in Ar+O2 mixture exhibited twenty times higher activity. After its calcination, the activity of resulting RF-Co-ox catalyst decreased six times, very likely due to increase in structure ordering of cobalt oxide particles (Table 1). The activity of the Co-I catalyst prepared by impregnation of the meshes with cobalt nitrate solution and heating was the highest. However, with respect to the practical application, deposited cobalt oxide layer was not sufficiently adhered to the support (Table 1). As the amounts of active components in the catalysts slightly differed, the specific reaction rates at 200 • C were calculated to compare the activity of the catalysts ( Table 4). Performance of the Co 3 O 4 species in the examined catalysts showed that the RF-x catalysts do not differ too much from each other. On the other hand, the RF-Co-ox/ap catalyst prepared by reactive magnetron sputtering in Ar+O 2 mixture exhibited twenty times higher activity. After its calcination, the activity of resulting RF-Co-ox catalyst decreased six times, very likely due to increase in structure ordering of cobalt oxide particles (Table 1). The activity of the Co-I catalyst prepared by impregnation of the meshes with cobalt nitrate solution and heating was the highest. However, with respect to the practical application, deposited cobalt oxide layer was not sufficiently adhered to the support (Table 1).
The highest specific activity among all examined catalysts was observed with the Co-RF-ox/ap sample (32.4 mmol EtOH g oxides −1 h −1 ), while the lowest one was determined for the pelletized Co 3 O 4 catalyst (PEL sample, 0.3 mmol EtOH g oxides −1 h −1 ). The concentrations of by-products formed during the reaction were similar for all examined catalysts (Table 4); the lowest formation of CO was achieved with the most active Co-RF-ox/ap catalyst.

Discussion
Experimental results carried out in this study demonstrated that the cobalt oxide catalysts prepared by magnetron sputtering exhibit very good adhesion of the Co 3 O 4 layer to the support. They exhibited only low weight loss during treatment in an ultrasonic bath, which was used as a simple test to evaluate the adhesion of deposited oxide; the lowest value was found for the catalyst prepared by Co sputtering in oxidation Ar+O 2 atmosphere (2.9% related to weight of deposited cobalt oxide). Adhesion of the coatings slightly decreased with increasing cobalt oxide content in the catalysts. The oxide coatings consisting of larger crystals observed in the SEM images (Supplementary Figure S1) indicated higher tendency to fall away from the mesh surface in comparison with those consisting of the smaller crystals.
All supported catalysts contain cobalt oxide with average valence of Co corresponding to Co 3 O 4 , whose presence in the catalysts was confirmed by powder XRD, Raman, and FTIR spectroscopy. As mentioned in the Results, consumptions of hydrogen for Co oxides reduction in the examined catalysts were 15.1, 15.2, and 15.6 mmol H 2 g −1 for the Co-RF-1, Co-RF-2, and PEL catalysts, respectively, and from 14.7 to 14.0 mmol H 2 g −1 for the remaining ones. As theoretical consumption of H 2 for reduction of Co 3 O 4 is 15.5 mmol H 2 g −1 , the H 2 consumptions found during TPR experiments, lower than theoretical one, indicate that small amounts of CoO in the supported catalysts cannot be excluded. The presence of CoO was confirmed in the Co-RF-1 sample (Figure 3) as the band with maximum at 505 cm −1 (assigned to CoO) was found in its FTIR spectrum. XRD did not have to record the presence of CoO due to its low concentration in the active phase or too small particles. Similar H 2 consumption during TPR experiments to the theoretical one was observed with the pelletized commercial Co 3 O 4 , despite the fact that CoO admixture was confirmed by powder XRD in this sample. Therefore, precisely differentiating the CoO and Co 3 O 4 contents in the cobalt oxide coatings prepared on the stainless-steel meshes from the H 2 -TPR measurements was not possible, but, in spite of this, formation of non-stoichiometric Co 3 O 4 -type oxides with lower Co valence than corresponding to Co 3 O 4 can be expected in the catalysts prepared by magnetron sputtering.
The highest catalytic activity (lowest T 50 temperature) among the catalysts prepared by magnetron sputtering was found for the non-calcined Co-RF-ox/ap catalyst prepared by Co sputtering in oxidation atmosphere. This catalyst showed the worst structure ordering of Co 3 O 4 (mean coherence length D of 7.1 nm) among all catalysts (Table 1). After calcination at 500 • C, performance of the catalyst decreased (T 50 was 250 • C) as the result of better structure ordering (D = 21.9 nm) and changes in the morphology of the Co 3 O 4 coatings (see Figure S1d,e in Supplementary). Slightly higher catalytic performance was observed with the Co-I catalyst prepared by impregnation (T 50 = 212 • C). However, the weight loss of the Co oxides in the adhesion test of this catalyst was 60% (Table 1), whereas the catalysts prepared by magnetron sputtering exhibited much better adhesion of the active phase: The weight loss was ranging from 3% to 31% in dependence on amount of the deposited Co oxides in the catalysts and the method of their preparation. The highest specific activity of the Co-RF-ox/ap corresponds well with the morphology and size of the cobalt oxide particles: The cobalt oxide coating prepared by reactive magnetron sputtering in Ar+O 2 atmosphere consisted of relatively large, well-developed platelet crystals with diameter of about 200 nm, while the PEL catalyst exhibited nearly 10 times larger particles ( Figure S1). In the presence of this catalyst, no CO was formed. The finding could be caused by the effect of internal diffusion in the pellets, which results in prolonged contact time of reaction components with the active centers.
Based on the catalysts' characteristics, high activity in oxidation of ethanol of the catalysts prepared by magnetron sputtering can be attributed to the excellent redox properties of Co 3 O 4 , the amount of lattice oxygen, and structure morphology. The high catalytic activity can be mainly attributed to low-temperature reducibility caused by the surface oxygen species, bulk oxygen mobility, and active oxygen vacancies [6,32]. Cations in high oxidation states function as oxidizing species while the oxygen species inserted into the oxidation product come from surface oxide ions, i.e., nucleophilic lattice oxygen. According to Finocchio [33], the activity of the p-type semiconducting oxides such as Co 3 O 4 is ascribed to "excess" nucleophilic oxygen species (O 2− ) at the surface, associated with the presence of trivalent cobalt (Co 3+ ). Therefore, the deep oxidation of ethanol over Co 3 O 4 thin film can proceed with Mars-van Krevelen type mechanism. This is confirmed by the fact that the temperature range, in which Co 3 O 4 particles of thin films are reduced (230-410 • C), corresponds well with the temperature range within which ethanol conversion is observed.

Catalysts Preparation
Conditions of catalyst preparation by RF magnetron sputtering are described in detail in [21]. Stainless-steel meshes (composition 71 wt.% Fe, 11 wt.% Ni, 16 wt.% Cr, and 2 wt.% Mo, mesh size 0.40 mm, wire of 0.22 mm in diameter) in the form of circle with outer diameter of 25 mm were used as supports. The meshes were cleaned mechanically before metal deposition using a brush with detergent, then washed in distilled water, and degreased in acetone for 10 min in an ultrasonic bath. Then, the meshes were dried at room temperature in air.
The cobalt deposition by RF magnetron sputtering was performed in a Veeco vacuum chamber evacuated to a base pressure of 2·10 −3 Pa. The magnetron of Kurt J. Lesker Torus 2HV equipped with the high strength magnets was used as a sputtering gun. The planar magnetron was connected to the Advanced Energy ATX 600 RF power supply. Before the cobalt deposition, the standard ion scrubbing process was used for the stainless-steel meshes pretreatment to improve adhesion of the deposited thin films. The chamber was filled with the forming gas (90% N 2 , 10% H 2 ) at the pressure of 20 Pa. The RF discharge was then ignited for 300 s to clean and activate the stainless-steel surface. A holder of six meshes (electrically connected) was placed at axial distance of 70 mm from the target face; this distance ensured homogeneity of film thickness better than 5% for all the meshes placed on the holder. In order to cover both sides of the meshes, the deposition process (except ion scrubbing) was repeated. The cobalt thin films were sputtered in pure argon atmosphere (Ar mass flow rate of 15 sccm) from the cobalt target (99.95% Co, Kurt J. Lesker) of 50 mm in diameter. The working pressure was adjusted by a gate valve throttling the turbomolecular pump to 1 Pa. The RF power delivered into the discharge was set on 200 W. The cobalt films with various thickness (approximately 0.5, 0.75, and 1.0 µm) were prepared after deposition times 10, 15, and 20 min. Then, the samples were calcined for 8 h at 500 • C in air. After calcination, the thickness of the layers increased to about 1.2, 1.8, and 2.4 µm, respectively. The prepared catalysts were denoted as Co-RF-x, where x is equal to 1, 2, or 3 denoting increasing amount of the cobalt layer deposited on the mesh (Table 1).
Analogous procedure was applied for reactive RF magnetron sputtering of cobalt oxide films. The target was reactively sputtered in Ar+O 2 atmosphere to yield cobalt oxide thin films. The films were sputtered from the cobalt target at power of 200 W under pressure of 2.5 Pa in a mixture of argon and oxygen at the Ar and O 2 flow rates of 80 and 5 sccm, respectively. The cobalt oxide films of about 2 µm in thickness were obtained after 75 min deposition. The as-appeared (ap) sample was labeled as Co-RF-ox/ap. After heating this sample at 500 • C for 4 h in air, a different sample was obtained and labelled as Co-RF-ox.

Catalytic Tests
The catalytic activities of the prepared catalysts were examined in the model reaction, total oxidation of ethanol [20]. Eight sieves were placed into the thin stainless-steel cylinder (diameter 25 mm) to simulate monolithic catalysts. Reaction temperature was increased from 100 to 400 • C with a temperature ramp of 2 • C min −1 . Inlet ethanol concentration in the air was 760 ± 10 ppm and space velocity F/W= 21 ± 1 L h g −1 . Reaction products were analyzed using a Hewlett-Packard 6890 gas chromatograph (Santa Clara, CA, U.S.A.) equipped with a FID detector and a capillary column (HP-5 19,091 J-413, 30 m × 0.32 mm × 0.25 mm with 5% phenyl methyl silicone). As measures of the catalysts performance temperatures T 50 and T 95 CO 2 (the temperatures, at which 50% of ethanol conversion and 95% conversion of organic compounds to CO 2 were achieved) were chosen. Ethanol conversions at 200 • C and corresponding reaction rates per gram of active components were determined as well. Selectivity of the catalysts with respect to acetaldehyde and carbon monoxide was characterized as maximum concentrations of acetaldehyde and carbon monoxide and by temperatures, at which they were observed.

Conclusions
The experimental study has shown that RF magnetron sputtering of cobalt followed by calcination in air is a promising tool for the deposition of cobalt oxide thin films on stainless-steel meshes. The catalysts prepared by magnetron sputtering in inert atmosphere exhibited relatively low activity in deep oxidation of ethanol. The catalysts efficiency was considerably improved when cobalt was sputtered in oxidation (Ar+O 2 ) atmosphere and direct oxidation of cobalt particles occurred during the deposition on the support. These catalysts exhibited higher catalytic performance than the pelletized commercial Co 3 O 4 catalyst with 50 times higher content of cobalt oxide. Moreover, the sputtered catalysts showed much better adhesion of the Co 3 O 4 coating to the support than the catalyst prepared by impregnation with aqueous solution of cobalt nitrate and subsequent heating. Comparison of specific activities revealed the size of cobalt oxide particles as an important parameter for oxidation activity: Smaller particle size led to increased catalytic activity. Finally, the sputtered catalysts showed much better adhesion of the Co 3 O 4 coating to the support than the catalyst prepared by impregnation with aqueous solution of cobalt nitrate and subsequent heating.