**1. Introduction**

Global environmental challenges stimulate the development of the carbon-free industry, in which hydrogen is an important energy carrier. Today, H<sup>2</sup> is used as a clean energy source in polymer electrolyte membrane fuel cells. But the high purity of H<sup>2</sup> is required for the more efficient operation of such devices. Even traces of CO (less than 100 ppm) remaining in hydrogen after fuel reforming and the water gas shift reaction should be removed to prevent poisoning of the Pt anode of the fuel cell [1–3]. Preferential CO oxidation (CO-PROX) is considered as one of the simplest and most efficient methods for reducing the CO concentration in the H<sup>2</sup> reach stream to an acceptable level.

The catalysts used at this stage should be highly active and selective in H<sup>2</sup> excess. They should also tolerate large amounts of CO<sup>2</sup> and H2O in the feed stream and be stable and durable in a wide operating temperature range [4].

Although noble metals are very active in CO oxidation [5,6] and show high performance in the CO-PROX [1,7], they have certain disadvantages. The most important ones are their high price, rapid deactivation under high temperature treatment and, in some cases, low selectivity. Copper oxide catalysts supported on cerium dioxide (ceria) are a promising alternative due to their affordable cost, high activity, selectivity and stability [8,9]. Varying the synthesis method, pretreatment conditions, type of promoters and other modifiers allows tuning the catalyst properties in a targeted manner [4,10,11]. Specifically, in the CO-PROX reaction, it is necessary to efficiently convert CO on the catalyst surface while minimizing undesirable hydrogen conversion. The improved performance of the catalyst is achieved via a synergistic action of a nonstoichiometric support that provides oxygen vacancies and an active component that adsorbs and activates CO [12]. The interaction of

**Citation:** Kaplin, I.Y.; Lokteva, E.S.; Tikhonov, A.V.; Maslakov, K.I.; Isaikina, O.Y.; Golubina, E.V. Copper–Cerium–Tin Oxide Catalysts for Preferential Oxidation of CO in Hydrogen: Effects of Synthesis Method and Copper Content. *Catalysts* **2022**, *12*, 1575. https:// doi.org/10.3390/catal12121575

Academic Editors: Yongjun Ji, Liwen Xing and Ke Wu

Received: 31 October 2022 Accepted: 29 November 2022 Published: 3 December 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

copper and cerium oxides [13]—in particular, the extent of their contact [14]—are widely discussed in the literature. Several reasons were proposed to explain their synergistic action in oxidation catalysis. Among them are the redox interplay between Ce3+/Ce4+ and Cu+/Cu2+ redox pairs, ample amount of oxygen vacancies, improvement of the reduction properties of each component, ligand and geometric effects, especially on the CuO–CeO<sup>2</sup> interface [13,15].

A significant number of works have been devoted to thermally stable cerium–zirconium catalysts [16–23]. Doping with tin dioxide was recently reported as a promising way for the modification of cerium-based catalysts [24,25]. It was demonstrated that a solid solution of tin in the crystal lattice of cerium oxide was formed up to a tin oxide content of 5–10 wt.% [26,27]. Earlier, we synthesized for the first time the Cu-Ce-Sn oxide catalysts with the Ce:Sn mole ratio of 9:1 using the cetyltrimethylammonium bromide (CTAB) template and demonstrated its high catalytic activity in the total CO oxidation [28]. It was attributed to the relatively high specific surface area of the catalysts, high dispersion of CuO<sup>x</sup> species and the presence of active copper oxide–ceria interfacial sites.

However, the effects of the copper content and strength of the interaction between copper oxide and ceria on the performance of the CTAB-templated Cu-Ce-Sn oxide catalysts in CO-PROX have not been thoroughly analyzed. Modifying the synthesis method is a promising way to control the strength of interactions between components of ternary oxide catalysts and their surface chemistry.

That is why this work compares two methods for copper incorporation into Ce0.9Sn0.1O<sup>2</sup> oxide (further denoted as CeSn). In the first one, CuO-modified Ce0.9Sn0.1O<sup>2</sup> (further denoted as CuCeSn) was prepared by the one-pot precipitation method using a sacrificial CTAB template and citric acid as a complexing agent. In the second method, copper was supported on Ce0.9Sn0.1O<sup>2</sup> by wet impregnation. The synthesized catalysts were compared with the CuCeZr catalysts produced by the same methods. The copper content was optimized in the CuCeSn catalyst prepared by the more efficient single-stage one-pot method. Different physicochemical methods, including the X-ray diffraction analysis, temperatureprogrammed reduction with hydrogen, low-temperature nitrogen physisorption, X-ray photoelectron and Raman spectroscopies and electron microscopy, were used to elucidate the influence of the preparation method and copper content on the physicochemical and catalytic properties of the synthesized catalysts in the flow-type CO-PROX.

#### **2. Results**

#### *2.1. Catalytic Performance in CO-PROX*

The temperature dependencies of the CO and O<sup>2</sup> conversions and CO<sup>2</sup> selectivity for all the synthesized catalysts (Figure 1) allowed qualitatively compared their catalytic activities, because all the catalysts were tested under the same conditions.

Firstly, we tested the catalysts synthesized by different methods but with the same copper content of 20% calculated as (*n*(Cu)·100%/[*n*(Ce)+*n*(Sn)] that corresponded to the copper loading of 6.1 wt.% relative to the CeSn support (Figure 1a–c). These catalysts were denoted as 20CuCeSn. Their Zr-containing counterparts (20CuCeZr) were also tested in the CO-PROX. For a better comparison, the equal Ce:Zr and Ce:Sn ratios of nine were used in the catalysts. The names of impregnated samples were ended with "-im".

Among the samples with the same composition, the one-pot synthesized catalysts were more active than the impregnated ones. The tin-containing one-pot synthesized 20CuCeSn showed higher CO conversion than its Zr-containing 20CuCeZr counterpart (Figure 1a). However, at low temperatures of 100–150 ◦C, the one-pot synthesized 20CuCeZr catalyst was more efficient than the impregnated 20CuCeSn-im, while, at higher temperatures (at 250–450 ◦C), they showed similar CO conversions.

**Figure 1.** CO (**a**,**d**) and O2 (**b**,**e**) conversions and CO2 selectivity (**c**,**f**) vs. reaction temperature for CuCeSn and CuCeZr catalysts with the same (**a**–**c**) and different (**d**–**f**) copper contents. **Figure 1.** CO (**a**,**d**) and O<sup>2</sup> (**b**,**e**) conversions and CO<sup>2</sup> selectivity (**c**,**f**) vs. reaction temperature for CuCeSn and CuCeZr catalysts with the same (**a**–**c**) and different (**d**–**f**) copper contents.

Firstly, we tested the catalysts synthesized by different methods but with the same copper content of 20% calculated as (*n*(Cu)·100%/[*n*(Ce)+*n*(Sn)] that corresponded to the copper loading of 6.1 wt.% relative to the CeSn support (Figure 1a–c). These catalysts were denoted as 20CuCeSn. Their Zr-containing counterparts (20CuCeZr) were also tested in the CO-PROX. For a better comparison, the equal Ce:Zr and Ce:Sn ratios of nine were used in the catalysts. The names of impregnated samples were ended with "-im". Among the samples with the same composition, the one-pot synthesized catalysts were more active than the impregnated ones. The tin-containing one-pot synthesized 20CuCeSn showed higher CO conversion than its Zr-containing 20CuCeZr counterpart (Figure 1a). However, at low temperatures of 100–150 °C, the one-pot synthesized 20CuCeZr catalyst was more efficient than the impregnated 20CuCeSn-im, while, at higher temperatures (at 250–450 °C), they showed similar CO conversions. Not only CO conversion but also carbon dioxide selectivity are very important for the cost efficiency of the catalyst in the CO-PROX process. As expected, the CO2 selectivity decreased for all the catalysts with increasing the reaction temperature. The conver-Not only CO conversion but also carbon dioxide selectivity are very important for the cost efficiency of the catalyst in the CO-PROX process. As expected, the CO<sup>2</sup> selectivity decreased for all the catalysts with increasing the reaction temperature. The conversion of oxygen at higher temperatures remained close to 100%, while both the CO conversion and CO<sup>2</sup> selectivity decreased (Figure 1b). This fact suggests that a part of the oxygen was consumed by the hydrogen oxidation reaction that produced water. Among all the synthesized catalysts, 20CuCeSn exhibited the highest CO<sup>2</sup> selectivity in a wide temperature range of 200–450 ◦C (Figure 1c). Note that, at 200–250 ◦C, a high CO<sup>2</sup> selectivity was achieved at 100% CO conversion, which provided the target process parameters—complete removal of CO at low degree of hydrogen oxidation, which makes this catalyst cost-efficient. The Zr-containing one-pot synthesized 20CuCeZr catalyst demonstrated higher CO<sup>2</sup> selectivity (98–99%) at 100–150 ◦C than its tin-containing counterpart but at the cost of a lower CO conversion. It can be concluded that templated coprecipitation is the most efficient method for preparation of ternary 20CuCeSn and 20CuCeZr catalysts, but the former catalyst is more efficient. Therefore, it was chosen for the Cu content optimization.

sion of oxygen at higher temperatures remained close to 100%, while both the CO conversion and CO2 selectivity decreased (Figure 1b). This fact suggests that a part of the oxygen was consumed by the hydrogen oxidation reaction that produced water. Among all the synthesized catalysts, 20CuCeSn exhibited the highest CO2 selectivity in a wide temperature range of 200–450 °C (Figure 1c). Note that, at 200–250 °C, a high CO2 selectivity was achieved at 100% CO conversion, which provided the target process parameters—complete removal of CO at low degree of hydrogen oxidation, which makes this The catalytic results for the one-pot synthesized samples with different copper contents are shown in Figure 1d–f. All these catalysts demonstrated a significant CO conversion even at 100 ◦C, but 25CuCeSn and 10CuCeSn were the most active at this temperature, showing CO conversions of 88 and 79%, respectively. At 150 ◦C, all the catalysts demonstrated near 100% CO conversion. Taking into account the uncertainty of the CO conversion measurement, the efficiencies of all the catalysts, except for the somewhat less efficient 10CuCeSn sample, were similar in the temperature range from 150 to 250 ◦C. At higher temperatures, the CO conversion decreased for all the catalysts.

catalyst cost-efficient. The Zr-containing one-pot synthesized 20CuCeZr catalyst demonstrated higher CO2 selectivity (98–99%) at 100–150 °C than its tin-containing counterpart but at the cost of a lower CO conversion. It can be concluded that templated coprecipitation is the most efficient method for preparation of ternary 20CuCeSn and 20CuCeZr catalysts, but the former catalyst is more efficient. Therefore, it was chosen for the Cu content optimization. The catalytic results for the one-pot synthesized samples with different copper con-10CuCeSn demonstrated the highest CO<sup>2</sup> selectivity at 200–400 ◦C (Figure 1f). Despite its lower CO conversion in this temperature range among the catalysts with different copper contents, it was still high enough (over 80%) to provide a good balance between reasonable CO and O<sup>2</sup> conversions and high CO<sup>2</sup> selectivity. At a relatively low temperature of 100 ◦C, the efficiencies of 10CuCeSn and 5CuCeSn were similarly high. However, the widest operating windows were observed for the 20CuCeSn and 25CuCeSn catalysts with high copper loadings.

tents are shown in Figure 1d–f. All these catalysts demonstrated a significant CO conversion even at 100 °C, but 25CuCeSn and 10CuCeSn were the most active at this temperature, showing CO conversions of 88 and 79%, respectively. At 150 °C, all the catalysts A 65% CO conversion and 70% CO<sup>2</sup> selectivity at 100 ◦C were earlier reported for the 5wt.%CuO/CeO<sup>2</sup> catalyst synthesized by the precipitation method [29]. Our 10CuCeSn

catalyst with a lower copper content (3.4 wt.% of Cu according to the AAS data) exhibited at this temperature the CO conversion of about 80% and CO<sup>2</sup> selectivity above 90%. However, the 5wt.%CuO/CeO<sup>2</sup> sample prepared in the above-mentioned work by the hydrothermal method showed a higher catalytic performance than that of our catalyst. However, in addition to the higher copper loading than in our 10CuCeSn catalyst, a stoichiometric CO/O<sup>2</sup> reaction mixture with a lower hydrogen content (50 vs. 80 vol.% H<sup>2</sup> in our work) and more than eight times higher contact time (0.800 vs. 0.095 g·s/cm<sup>3</sup> in our work) were used in [29] for catalytic tests. Under similar test conditions, the performance of our catalysts should be better. catalyst with a lower copper content (3.4 wt.% of Cu according to the AAS data) exhibited at this temperature the CO conversion of about 80% and CO2 selectivity above 90%. However, the 5wt.%CuO/CeO2 sample prepared in the above-mentioned work by the hydrothermal method showed a higher catalytic performance than that of our catalyst. However, in addition to the higher copper loading than in our 10CuCeSn catalyst, a stoichiometric CO/O2 reaction mixture with a lower hydrogen content (50 vs. 80 vol.% H2 in our work) and more than eight times higher contact time (0.800 vs. 0.095 g·s/cm3 in our work) were used in [29] for catalytic tests. Under similar test conditions, the performance of our catalysts should be better.

demonstrated near 100% CO conversion. Taking into account the uncertainty of the CO conversion measurement, the efficiencies of all the catalysts, except for the somewhat less efficient 10CuCeSn sample, were similar in the temperature range from 150 to 250 °C. At

10CuCeSn demonstrated the highest CO2 selectivity at 200–400 °C (Figure 1f). Despite its lower CO conversion in this temperature range among the catalysts with different copper contents, it was still high enough (over 80%) to provide a good balance between reasonable CO and O2 conversions and high CO2 selectivity. At a relatively low temperature of 100 °C, the efficiencies of 10CuCeSn and 5CuCeSn were similarly high. However, the widest operating windows were observed for the 20CuCeSn and

A 65% CO conversion and 70% CO2 selectivity at 100 °C were earlier reported for the 5wt.%CuO/CeO2 catalyst synthesized by the precipitation method [29]. Our 10CuCeSn

*Catalysts* **2022**, *12*, x FOR PEER REVIEW 4 of 20

25CuCeSn catalysts with high copper loadings.

higher temperatures, the CO conversion decreased for all the catalysts.

The stability of 10CuCeSn was tested in the isothermal conditions at 200 ◦C for more than 10 h time-on-stream (Figure 2) at longer contact time than in short-time tests. No loss of activity was observed after more than 10 h of the test at a CO conversion of about 99.8% and CO<sup>2</sup> selectivity of about 92%. The stability of 10CuCeSn was tested in the isothermal conditions at 200 °C for more than 10 h time-on-stream (Figure 2) at longer contact time than in short-time tests. No loss of activity was observed after more than 10 h of the test at a CO conversion of about 99.8% and CO2 selectivity of about 92%.

#### *2.2. Physicochemical Characterization of Catalysts Synthesized by Different Techniques 2.2. Physicochemical Characterization of Catalysts Synthesized by Different Techniques*

The physicochemical parameters of the 20Cu series catalysts prepared by the two methods were measured by N2 physisorption, XRD, XPS and H2-TPR, and the results are summarized in Table 1. The physicochemical parameters of the 20Cu series catalysts prepared by the two methods were measured by N<sup>2</sup> physisorption, XRD, XPS and H2-TPR, and the results are summarized in Table 1.

**Table 1.** Physicochemical parameters of 20CuCeSn and 20CuCeZr catalysts prepared by the **Table 1.** Physicochemical parameters of 20CuCeSn and 20CuCeZr catalysts prepared by the one-pot and wet impregnation methods.


Figure S1 (Supplementary Materials, SM) shows the SEM micrographs of these catalysts and the EDX mappings of the Sn(Zr) and Cu distributions on the catalyst surface. Rather large particles from 50 to 200 µm in size were observed in all the catalysts. The Zr-containing samples looked much less porous, which agreed with their lower specific surface areas. Copper-enriched areas were not detected on the surfaces of all the samples,

even of those synthesized by the impregnation, which can be explained by the relatively low Cu content (6.1 wt.%).

The textural parameters of the catalysts were measured by low-temperature nitrogen physisorption. Though the different synthesis methods were used for 20CuCeZr and 20CuCeZr-im, their physisorption isotherms (Figure S2 in Supplementary Materials) and specific surface areas (Table 1) were similar. The average pore sizes calculated from the NL-DFT pore size distributions were 3.1 and 5.7 nm for 20CuCeSn and 20CuCeSn-im and 7 and 12 nm for 20CuCeZr and 20CuCeZr-im, respectively.

Higher BET specific surface areas of the tin-containing catalysts (Table 1) testify to a much more developed surface of these catalysts than of their Zr-doped counterparts. The impregnated 20CuCeSn-im demonstrated a lower specific surface area than its one-pot synthesized counterpart (20CuCeSn). In contrast to the one-pot synthesized catalyst, CuO<sup>x</sup> particles in the impregnated 20CuCeSn-im sample can block micro- and small mesopores. The specific surface areas of CuCeZr and CuCeZr-im were about the same but much lower than those of their Sn-doped counterparts, which may be the reason for the lower activity of the former catalysts in CO oxidation. *Catalysts* **2022**, *12*, x FOR PEER REVIEW 6 of 20

> To analyze the phase composition of the catalysts, they were studied by XRD and Raman spectroscopy. The diffraction patterns of all the studied samples (Figure 3a) show only broad reflections of the fluorite phase of CeO2. Thus, the samples did not contain wellcrystallized phases of tin/zirconium and copper oxides. We may assume that copper existed in the samples both as ions incorporated into the ceria or double cerium–zirconium/cerium– tin oxide lattice and as highly dispersed X-ray amorphous copper oxide species. defects) [31]. It could be attributed to the LO mode (longitudinal mode, F1u symmetry) appearing due to the local asymmetry caused by the presence of Frenkel-type oxygen defects or non-stoichiometry oxygen vacancies in ceria resulting from the reduction of Ce4+ to Ce3+ [31,34]. Due to a rather small crystallite size in the catalysts, it is difficult to

accurately determine the contribution of this component to the spectra.

**Figure 3.** XRD patterns (**a**) and Raman spectra (**b**) of 20CuCeSn, 20CuCeZr, 20CuCeSn-im and 20CuCeZr-im catalysts before (solid lines) and after catalytic tests (dashed lines). **Figure 3.** XRD patterns (**a**) and Raman spectra (**b**) of 20CuCeSn, 20CuCeZr, 20CuCeSn-im and 20CuCeZr-im catalysts before (solid lines) and after catalytic tests (dashed lines).

The Raman spectra of the 20Cu series samples are similar before and after catalytic tests (Figure 3b). No carbon-related bands appeared in the spectra. The spectra of all the samples, excluding 20CuCeSn, showed a blue shift of the F2g mode. It can be explained by the partial reduction of cerium in the surface layers under the reaction conditions and by the increase in the fraction of Ce3+ ions that are larger than Ce4+ ones. This shift agrees with the equation that describes the changes in the F2g band frequency with varying the crystal lattice parameters of ceria-based oxide [35]: ∆ = −3 ∆ బ , where *γ* is the dimensionless Grüneisen constant (1.44 for CeO2), ∆*a* = (*a* − *a*0) is the difference between the crystal lattice parameters of the doped and pure CeO2 and Δω = (ω − ω0) is the difference between the frequencies of their F2g bands. A weak redshift of the F2g band was observed in the Raman spectrum of the most The Raman spectroscopy data testify to the inclusion of zirconium/tin and copper ions into the crystal structure of cerium oxide (Figure 3b). The main F2g band in the 440–465 cm−<sup>1</sup> region confirms the cubic fluorite structure of CeO<sup>2</sup> [30]. This band is broader for tin-containing catalysts, especially for 20CuCeSn, which could result from the small particle size in these catalysts [30,31]. Indeed, according to the XRD data (Table 1), 20CuCeSn has the smallest crystallite size. The slight asymmetry of the F2g band is clearly visible in the spectrum of 20CuCeSn, along with a red shift of the F2g band that points to the presence of oxygen vacancies [30]. Both the asymmetry of the F2g band and its shift for nanosized particles originate from the phonon confinement effect resulting from the high concentration of structural defects and small average crystallite size [32]. The shift of the F2g band for all the Cu-containing catalysts relative to their positions in CeO<sup>2</sup> (Figure 3b) and CeSn double oxide (see Figure 5b below) can be caused by the change in the Ce−O vibration frequency resulting from the incorporation of Cu2+ ions. A similar

cessing of Ce3*d* and Cu2*p* XPS spectra was detailed in our previous work [36].

active 20CuCeSn sample. This fact indicates the high stability of this system and the high mobility of oxygen, which ensures its fast transport to the surface from deeper structural

The survey XPS spectra are presented in SM (Figure S3a). All the samples contained Cu, O, Ce, Sn or Zr and a relatively high amount of carbon on the surface. The latter can be a template/precursor residue or adventitious carbon resulting from the air exposure of

The Ce/Zr atomic ratios on the surface of zirconium-containing samples calculated from the XPS spectra were close to the target values (Table 1). In contrast, the tin-containing 20CuCeSn and 20CuCeSn-im samples demonstrated different surface Ce/Sn ratios. The surface of 20CuCeSn-im was depleted in tin, while that of 20CuCeSn was enriched with this element. These significant differences in the surface compositions of the samples can be explained by the segregation processes and formation of mixed

The copper content on the surface of the impregnated catalysts is much higher than that for the 20CuCeSn and 20CuCeZr samples (Table 1). This fact confirms the uniform

oxide phases with different compositions.

the catalysts.

effect was observed in [33]. The shift of the band maximum is more pronounced for the CuCeSn samples. All spectra also contain a broad low-intensity band at 560−650 cm−<sup>1</sup> that is usually denoted as the 'O<sup>v</sup> band' (where O<sup>v</sup> means oxygen vacancies) or 'D-band' (D means defects) [31]. It could be attributed to the LO mode (longitudinal mode, F1u symmetry) appearing due to the local asymmetry caused by the presence of Frenkel-type oxygen defects or non-stoichiometry oxygen vacancies in ceria resulting from the reduction of Ce4+ to Ce3+ [31,34]. Due to a rather small crystallite size in the catalysts, it is difficult to accurately determine the contribution of this component to the spectra.

The Raman spectra of the 20Cu series samples are similar before and after catalytic tests (Figure 3b). No carbon-related bands appeared in the spectra. The spectra of all the samples, excluding 20CuCeSn, showed a blue shift of the F2g mode. It can be explained by the partial reduction of cerium in the surface layers under the reaction conditions and by the increase in the fraction of Ce3+ ions that are larger than Ce4+ ones. This shift agrees with the equation that describes the changes in the F2g band frequency with varying the crystal lattice parameters of ceria-based oxide [35]:

$$
\Delta \omega = -3 \,\gamma \,\omega\_0 \frac{\Delta a}{a\_0}
$$

where *γ* is the dimensionless Grüneisen constant (1.44 for CeO2), ∆*a* = (*a* − *a*0) is the difference between the crystal lattice parameters of the doped and pure CeO<sup>2</sup> and ∆ω = (ω − ω0) is the difference between the frequencies of their F2g bands.

A weak redshift of the F2g band was observed in the Raman spectrum of the most active 20CuCeSn sample. This fact indicates the high stability of this system and the high mobility of oxygen, which ensures its fast transport to the surface from deeper structural layers and allows rapid reoxidation of the catalyst surface.

The surface composition of the catalysts was studied by XPS (Table 1). The processing of Ce3*d* and Cu2*p* XPS spectra was detailed in our previous work [36].

The survey XPS spectra are presented in Supplementary Materials (Figure S3a). All the samples contained Cu, O, Ce, Sn or Zr and a relatively high amount of carbon on the surface. The latter can be a template/precursor residue or adventitious carbon resulting from the air exposure of the catalysts.

The Ce/Zr atomic ratios on the surface of zirconium-containing samples calculated from the XPS spectra were close to the target values (Table 1). In contrast, the tin-containing 20CuCeSn and 20CuCeSn-im samples demonstrated different surface Ce/Sn ratios. The surface of 20CuCeSn-im was depleted in tin, while that of 20CuCeSn was enriched with this element. These significant differences in the surface compositions of the samples can be explained by the segregation processes and formation of mixed oxide phases with different compositions.

The copper content on the surface of the impregnated catalysts is much higher than that for the 20CuCeSn and 20CuCeZr samples (Table 1). This fact confirms the uniform distribution of copper in the one-pot synthesized catalysts and less uniformity of the impregnated catalysts, which was expected considering the nature of the synthesis methods.

The highest Cu+/Cu2+ XPS ratio over all samples was found in 20CuCeSn. This fact points to a strong interaction between ceria and CuO<sup>x</sup> and easy electron transfer between Cu2+ and Ce3+ species [28,37]:

$$\text{Cu}^{2+} + \text{Ce}^{3+} \leftrightarrow \text{Cu}^{+} + \text{Ce}^{4+} $$

The ability of the cerium ion to transfer an electron is confirmed by a high surface Ce3+/Ce4+ ratio. The synergistic model proposed in [37,38] supposes a key role of Cu<sup>+</sup> species as surface sites for CO adsorption and activation. The in situ XANES and DRIFTS study demonstrated that the CO-PROX reaction over ceria–copper catalysts mainly proceeded via adsorption of CO on Cu<sup>+</sup> species rather than on Cu2+ and Cu<sup>0</sup> ones [39]. These

Cu<sup>+</sup> sites are stabilized by the interaction between copper oxide species and cerium oxide on their interface [40]. The role of cerium oxide is to create oxygen vacancies. mixture contains an excess of H2, so the reduction of the catalyst components can proceed during the catalytic test.

on their interface [40]. The role of cerium oxide is to create oxygen vacancies.

distribution of copper in the one-pot synthesized catalysts and less uniformity of the impregnated catalysts, which was expected considering the nature of the synthesis

The highest Cu+/Cu2+ XPS ratio over all samples was found in 20CuCeSn. This fact points to a strong interaction between ceria and CuOx and easy electron transfer between

Cu2+ + Ce3+↔ Cu+ + Ce4+

The ability of the cerium ion to transfer an electron is confirmed by a high surface Ce3+/Ce4+ ratio. The synergistic model proposed in [37,38] supposes a key role of Cu+ species as surface sites for CO adsorption and activation. The in situ XANES and DRIFTS study demonstrated that the CO-PROX reaction over ceria–copper catalysts mainly proceeded via adsorption of CO on Cu+ species rather than on Cu2+ and Cu<sup>0</sup> ones [39]. These Cu+ sites are stabilized by the interaction between copper oxide species and cerium oxide

The reducibility of the samples was investigated by H2-TPR. The TPR profiles are presented in Figure 4. This analysis is very important, because the CO-PROX reaction

*Catalysts* **2022**, *12*, x FOR PEER REVIEW 7 of 20

methods.

Cu2+ and Ce3+ species [28,37]:

The reducibility of the samples was investigated by H2-TPR. The TPR profiles are presented in Figure 4. This analysis is very important, because the CO-PROX reaction mixture contains an excess of H2, so the reduction of the catalyst components can proceed during the catalytic test. The main peaks in the TPR profiles can be divided into three groups: low-temperature (80–400 °C), medium-temperature (400–650 °C) and high-temperature (above 650 °C) peaks. The high-temperature peaks are probably associated with the reduction of bulk Ce4+ and Sn4+ [27,41].

**Figure 4.** H2-TPR profiles of 20CuCeSn, 20CuCeZr, 20CuCeSn-im and 20CuCeZr-im catalysts and reference CeSn and CeO2 samples. **Figure 4.** H<sup>2</sup> -TPR profiles of 20CuCeSn, 20CuCeZr, 20CuCeSn-im and 20CuCeZr-im catalysts and reference CeSn and CeO<sup>2</sup> samples.

At medium temperatures, a group of relatively low-intense peaks was observed in all the TPR profiles. These peaks can be associated with the reduction of surface Ce4+, Sn4+ and/or Zr4+ species. The presence of Cu0 species strongly affects their reducibility. The maxima of medium-temperature peaks shift to lower temperatures but to a different ex-The main peaks in the TPR profiles can be divided into three groups: low-temperature (80–400 ◦C), medium-temperature (400–650 ◦C) and high-temperature (above 650 ◦C) peaks. The high-temperature peaks are probably associated with the reduction of bulk Ce4+ and Sn4+ [27,41].

tent for different sample compositions. The low-temperature peaks at 50–400 °C are usually attributed to the reduction of highly dispersed and bulk Cu2+ oxide species and to the sequential reduction of Cu2+ to Cu0 (Cu2+→ Cu+ → Cu0) [11,42]. The reduction of Sn4+ on the surface of the cerium–tin binary sample proceeded at a temperature slightly above 300 °C (Figure 3), but this peak was not visible in the TPR profiles of the copper-containing catalysts. Most likely, it strongly shifted to lower temperatures and over-At medium temperatures, a group of relatively low-intense peaks was observed in all the TPR profiles. These peaks can be associated with the reduction of surface Ce4+, Sn4+ and/or Zr4+ species. The presence of Cu<sup>0</sup> species strongly affects their reducibility. The maxima of medium-temperature peaks shift to lower temperatures but to a different extent for different sample compositions. The low-temperature peaks at 50–400 ◦C are usually attributed to the reduction of highly dispersed and bulk Cu2+ oxide species and to the sequential reduction of Cu2+ to Cu<sup>0</sup> (Cu2+<sup>→</sup> Cu<sup>+</sup> <sup>→</sup> Cu<sup>0</sup> ) [11,42]. The reduction of Sn4+ on the surface of the cerium–tin binary sample proceeded at a temperature slightly above 300 ◦C (Figure 3), but this peak was not visible in the TPR profiles of the copper-containing catalysts. Most likely, it strongly shifted to lower temperatures and overlapped with the CuO reduction peaks, indicating the promoting effect of copper on the Sn4+ reduction. The partial reduction of a fraction of surface Ce4+ species in the low-temperature range also cannot be excluded considering that the intensity of the middle-temperature peak decreased in the TPR profiles of Cu-containing catalysts relative to those of CeSn and CeO2.

The low-temperature peaks were quite similar in the TPR profiles of the impregnated catalysts: both profiles comprised a broad hydrogen consumption peak centered at 222–232 ◦C. However, a slight shift to higher temperatures and a significant increase in the hydrogen consumption (Table 1) were observed for CuCeSn-im.

The hydrogen uptake peaks in the low-temperature region differ markedly for the onepot synthesized 20CuCeSn and 20CuCeZr samples. Several intense peaks in the reduction profile of 20CuCeZr correspond to the reduction of various forms of copper oxide on the surface and in bulk. The TPR profile of 20CuCeSn has only two intense signals with a small shoulder at 119 ◦C, indicating the lower diversity of copper forms in this catalyst. However, the copper reduction in this sample started at lower temperatures, and the hydrogen uptake

was higher than that for 20CuCeZr (Table 1), which may be due to the partial reduction of cerium and tin in the interfacial regions, in which ceria (or Sn-doped ceria) contacted copper oxide.

#### *2.3. Effect of Copper Content on Physicochemical Properties of Catalysts*

This section presents the results of the physicochemical study of the catalysts with different copper contents prepared by the one-pot method (Table 2).

**Table 2.** Textural properties and chemical compositions of one-pot synthesized CuCeSn catalysts with different Cu contents.


The EDX analysis of all the catalysts demonstrated the presence of oxygen, cerium, tin, copper and carbon. A bromine signal was also detected in the EDX spectra of all the samples, but its intensity was near the sensitivity level and similar for all the catalysts. Therefore, it was not taken into account in the sample composition calculations.

According to the EDX element mappings (Figure S4 in Supplementary Materials), all the samples comprised the areas of uniform and nonuniform tin distributions. We could assume either the presence of highly dispersed tin oxide or/and the formation of mixed cerium–tin oxide phases of various compositions. The cerium–copper atomic ratios were calculated from the SEM-EDX data (Table 2). As expected, the Ce/Cu ratio decreased with decreasing the Cu content.

The distribution of copper on the surface of the (5–20) CuCeSn samples could be considered as uniform. However, on the surface of the 25CuCeSn sample with the highest copper content, the copper-enriched areas were detected, which may indicate the presence of copper oxide phases in this sample.

10CuCeSn showed the highest BET-specific surface area. The average pore size calculated from the pore size distributions (NL-DFT method) was 3–4 nm for all the catalysts. A significant contribution of micropores was observed in all the samples, but it was the highest in 10CuCeSn.

No reflections of copper-containing phases were detected in the XRD profiles of the copper-modified samples (Figure 5a). Therefore, all the samples contained only the fluoritetype crystalline phase of cubic ceria. The particle size noticeably increased with the decreasing copper content; the average crystallite sizes were 5 ± 1 nm for 25CuCeSn and 20CuCeSn, 6 ± 1 nm for 10CuCeSn and 9 ± 1 nm for 5CuCeSn, approaching the value of 12 ± 1 nm for double oxide CeSn. Copper could exist in the samples in the form of finely dispersed oxide particles or/and as ions partially incorporated into the ceria/ceria–tin lattice.

A slight shift of the reflections in the diffraction patterns of the cerium–tin–copper samples to higher angles relative to the pattern of reference CeO<sup>2</sup> indicates a decrease in the crystal lattice parameter and can testify to the partial incorporation of tin or/and copper ions into the crystal lattice of cerium oxide. However, it is impossible to accurately calculate the lattice parameter for the catalysts due to the strong broadening of the reflections.

The sizes of Sn4+ (0.81 Å for a coordination number of 8) and Cu2+ (0.71 Å) cations are smaller than those of cerium cations (1.10 Å for Ce3+ and 0.97 Å for Ce4+). Therefore, they can be incorporated into the CeO<sup>2</sup> crystal lattice, and this incorporation will slightly decrease the lattice parameter. At the same time, the incorporation of such ions can promote the Ce4+ to Ce3+ reduction, which increases the crystal lattice parameter. Thus,

it is impossible to draw unambiguous conclusions about the degree of ion incorporation based only on the crystal lattice parameters calculated from the XRD data. rately calculate the lattice parameter for the catalysts due to the strong broadening of the reflections.

ria/ceria–tin lattice.

*Catalysts* **2022**, *12*, x FOR PEER REVIEW 9 of 20

fluorite-type crystalline phase of cubic ceria. The particle size noticeably increased with the decreasing copper content; the average crystallite sizes were 5 ± 1 nm for 25CuCeSn and 20CuCeSn, 6 ± 1 nm for 10CuCeSn and 9 ± 1 nm for 5CuCeSn, approaching the value of 12 ± 1 nm for double oxide CeSn. Copper could exist in the samples in the form of finely dispersed oxide particles or/and as ions partially incorporated into the ce-

A slight shift of the reflections in the diffraction patterns of the cerium–tin–copper samples to higher angles relative to the pattern of reference CeO2 indicates a decrease in the crystal lattice parameter and can testify to the partial incorporation of tin or/and copper ions into the crystal lattice of cerium oxide. However, it is impossible to accu-

**Figure 5.** XRD patterns (**a**) and Raman spectra (**b**) of catalysts with different Cu contents. **Figure 5.** XRD patterns (**a**) and Raman spectra (**b**) of catalysts with different Cu contents.

The sizes of Sn4+ (0.81 Å for a coordination number of 8) and Cu2+ (0.71 Å) cations are smaller than those of cerium cations (1.10 Å for Ce3+ and 0.97 Å for Ce4+). Therefore, they can be incorporated into the CeO2 crystal lattice, and this incorporation will slightly decrease the lattice parameter. At the same time, the incorporation of such ions can promote the Ce4+ to Ce3+ reduction, which increases the crystal lattice parameter. Thus, it is im-The F2g band in the Raman spectra of all the catalysts (Figure 5b) is broadened and shifted by about 20 cm−<sup>1</sup> relative to the spectra of references CeSn and CeO2. As it has been already mentioned, the broadening is caused by the small crystallite size and by the defectiveness of the structure. The reason for the F2g band shift is a partial incorporation of copper into the CeSn lattice.

possible to draw unambiguous conclusions about the degree of ion incorporation based only on the crystal lattice parameters calculated from the XRD data. The F2g band in the Raman spectra of all the catalysts (Figure 5b) is broadened and shifted by about 20 cm−1 relative to the spectra of references CeSn and CeO2. As it has The survey XPS spectra of the catalysts with different Cu contents (Figure S3b) in addition to the expected Ce, Sn and Cu lines showed the presence of C and Br, which agrees with the SEM-EDX results. The calculated Ce/Cu ratios (Table 2) indicated the depletion of the surface in copper, especially for 10CuCeSn, which confirms the SEM-EDX data.

been already mentioned, the broadening is caused by the small crystallite size and by the defectiveness of the structure. The reason for the F2g band shift is a partial incorporation of copper into the CeSn lattice. The survey XPS spectra of the catalysts with different Cu contents (Figure S3b) in addition to the expected Ce, Sn and Cu lines showed the presence of C and Br, which agrees with the SEM-EDX results. The calculated Ce/Cu ratios (Table 2) indicated the depletion of the surface in copper, especially for 10CuCeSn, which confirms the SEM-EDX data. The fraction of Ce3+ decreased with increasing the copper content. It can be assumed that the limit of copper incorporation was reached in 5CuCeSn and 10CuCeSn. At higher copper concentrations, fine copper oxide particles started to form on the surface, which explains the sharp drop in the Ce/Cu and Ce3+/Ce4+ ratios (Table 2). The possibility of partial cerium reduction in the XPS chamber during analysis should be also considered. However, because the spectra acquisition time was about the same for all the catalysts, in the case of the X-ray-induced reduction, we should have expected the same portion of Ce4+ reduced to Ce3+ on the surfaces of all the samples, which was not the case.

The fraction of Ce3+ decreased with increasing the copper content. It can be assumed that the limit of copper incorporation was reached in 5CuCeSn and 10CuCeSn. At higher copper concentrations, fine copper oxide particles started to form on the surface, which explains the sharp drop in the Ce/Cu and Ce3+/Ce4+ ratios (Table 2). The possibility of The surface fractions of copper in +2 and +1 oxidation states were calculated from the areas of the Cu2+ and Cu<sup>+</sup> components in the Cu2*p* spectra (Table 2). The highest Cu<sup>+</sup> content was observed in 10CuCeSn. The important role of Cu<sup>+</sup> sites in CO-PROX was earlier reported [10,39,42].

partial cerium reduction in the XPS chamber during analysis should be also considered. However, because the spectra acquisition time was about the same for all the catalysts, in Figure 6a shows the high resolution O1*s* XPS spectra fitted with several components. The interpretation of these components in the literature is ambiguous, as was shown in our previous work [28]. Many authors have distinguished two major groups of peaks: the components at lower binding energies (denoted as Olat) are associated with lattice oxygen [26,43], while the peaks at higher binding energies (Oads and OOH) are assigned to low-coordinated highly polarized oxygen species (superoxide or peroxide species) adsorbed on the surface or to oxygen in hydroxyl groups [44,45]. These species are believed to play a significant role in catalytic oxidation.

The atomic ratio of lattice and non-lattice oxygen (the latter is the sum of oxygen in hydroxyl and other adsorbed groups) substantially changed with increasing the copper content (Figure 6b). A minimum was observed for the 10CuCeSn sample. The relative contribution of these components to the O1*s* spectrum correlates with the concentration of

the active surface oxygen species [4]. The opposite trend with a maximum for 10CuCeSn was observed for the SBET and Cu+/Cu2+ ratio dependencies on Cu loading. low-coordinated highly polarized oxygen species (superoxide or peroxide species) adsorbed on the surface or to oxygen in hydroxyl groups [44,45]. These species are believed to play a significant role in catalytic oxidation.

earlier reported [10,39,42].

*Catalysts* **2022**, *12*, x FOR PEER REVIEW 10 of 20

the case of the X-ray-induced reduction, we should have expected the same portion of

The surface fractions of copper in +2 and +1 oxidation states were calculated from the areas of the Cu2+ and Cu+ components in the Cu2*p* spectra (Table 2). The highest Cu+ content was observed in 10CuCeSn. The important role of Cu+ sites in CO-PROX was

Figure 6a shows the high resolution O1*s* XPS spectra fitted with several components. The interpretation of these components in the literature is ambiguous, as was shown in our previous work [28]. Many authors have distinguished two major groups of peaks: the components at lower binding energies (denoted as Olat) are associated with lattice oxygen [26,43], while the peaks at higher binding energies (Oads and OOH) are assigned to

Ce4+ reduced to Ce3+ on the surfaces of all the samples, which was not the case.

**Figure 6.** High-resolution O1*s* XPS spectra (**a**) and a correlation between the SBET, ratio of oxygen forms, Cu+/Cu2+ ratio measured by XPS and Cu content in one-pot synthesized CuCeSn catalysts (**b**). **Figure 6.** High-resolution O1*s* XPS spectra (**a**) and a correlation between the SBET, ratio of oxygen forms, Cu+/Cu2+ ratio measured by XPS and Cu content in one-pot synthesized CuCeSn catalysts (**b**). *Catalysts* **2022**, *12*, x FOR PEER REVIEW 11 of 20

The atomic ratio of lattice and non-lattice oxygen (the latter is the sum of oxygen in hydroxyl and other adsorbed groups) substantially changed with increasing the copper content (Figure 6b). A minimum was observed for the 10CuCeSn sample. The relative contribution of these components to the O1*s* spectrum correlates with the concentration of the active surface oxygen species [4]. The opposite trend with a maximum for 10CuCeSn was observed for the SBET and Cu+/Cu2+ ratio dependencies on Cu loading. The H2-TPR experiments were performed to reveal the copper content effect on the catalyst reduction. The H2-TPR profiles of the catalysts with different copper loadings The H2-TPR experiments were performed to reveal the copper content effect on the catalyst reduction. The H2-TPR profiles of the catalysts with different copper loadings (Figure 7) are similar and show a broad reduction signal in the high- and mediumtemperature ranges, in addition to two intense peaks (α and β) in the low-temperature region (80–400 ◦C). The latter peaks can be attributed to the reduction of copper species with H2. The low-temperature peak (α) corresponds to the reduction of copper oxide species finely dispersed on ceria, while the peak at a higher temperature (β) relates to the reduction of well-formed CuO particles or mixed oxide phases (CuCeO<sup>x</sup> or CuCeSnOx) [26,46,47]. decreasing copper content. This indicates the increase in the fraction of copper particles strongly bonded to the oxide support in the catalysts with low copper contents. The observed copper oxide reduction peaks point to the easiest reduction of copper in 10CuCeSn, which can be associated with the fact that the copper content in this catalyst reached the limiting copper "solubility" in the cerium–tin double oxide. A further increase in the copper loading led to the growth of the fraction of CuO particles weakly interacting with the support, which complicates the reduction.

**Figure 7.** H**2-**TPR profiles of one-pot synthesized CuCeSn catalysts with different Cu loadings. **Figure 7.** H<sup>2</sup> -TPR profiles of one-pot synthesized CuCeSn catalysts with different Cu loadings.

An additional shoulder with a maximum at 130 °C was detected near the main signal at 152 °C in the TPR profile of 10CuCeSn. A similar shoulder was observed in the low-temperature peak in the H2-TPR profile of 20CuCeSn, but it strongly overlapped with the reduction signal at 154 °C. This peak can be associated with the hydrogen consumption by oxygen ions weakly bonded with surface copper species [48], probably located at interfacial positions. The quantitative estimation of hydrogen consumption in the low-temperature re-However, the positions of the α and β peaks in the TPR profiles of our samples strongly depended on the Cu loading. With decreasing the copper content (*n*(Cu)·100%/[*n*(Ce)+*n*(M)]) from 25 to 10%, the maxima of both peaks shifted to low temperatures, and the shift of the β peak was more pronounced than that of the α one. As a result, the difference between the positions of their maxima became smaller (Table 3). A further decrease in the copper content shifted the peaks of copper reduction to higher temperatures. One can notice a decrease in the area ratio of the α and β peaks with the decreasing copper content. This indicates

gion is presented in Table 3. The hydrogen uptake increased in this region with increasing the copper content. However, the experimental hydrogen uptakes for all the catalysts exceeded the calculated amounts of hydrogen required for the complete reduction of

ratio in Table 3). The additional hydrogen was apparently consumed by the low-temperature reduction of ceria or tin oxide. This fact indicates that the reduction of these oxides is favored by the presence of dispersed surface CuO species, i.e., the larger the number of CuO–ceria or CuO–SnOx contacts, the easier the reduction of CeO2 and/or SnOx proceeds. The H2/Cu ratio decreased with increasing the copper content probably because of the decrease in the surface fraction of oxide support with Ce4+ and Sn4+ species

Note that the H2-TPR profile of double CeSn oxide showed a noticeable peak at 308 °C. Finer tin oxide particles can be reduced much easier than larger ones [49] due to the improved ability of the former particles to form active oxygen species. In contrast to large SnO2 particles that reduce at temperatures about 677 °C, small nano-sized crystals reduce

available for low-temperature reduction.

in a broad temperature range of 100–500 °C [50].

the increase in the fraction of copper particles strongly bonded to the oxide support in the catalysts with low copper contents.


**Table 3.** H<sup>2</sup> -TPR results for CuCeSn catalysts with different Cu loadings.

<sup>1</sup> Temperature gap between α and β components. <sup>2</sup> Calculated from the Cu content determined by AAS, assuming that H<sup>2</sup> absorbed in the low-temperature region is consumed only for the reduction of CuO to Cu<sup>0</sup> .

The observed copper oxide reduction peaks point to the easiest reduction of copper in 10CuCeSn, which can be associated with the fact that the copper content in this catalyst reached the limiting copper "solubility" in the cerium–tin double oxide. A further increase in the copper loading led to the growth of the fraction of CuO particles weakly interacting with the support, which complicates the reduction.

An additional shoulder with a maximum at 130 ◦C was detected near the main signal at 152 ◦C in the TPR profile of 10CuCeSn. A similar shoulder was observed in the lowtemperature peak in the H2-TPR profile of 20CuCeSn, but it strongly overlapped with the reduction signal at 154 ◦C. This peak can be associated with the hydrogen consumption by oxygen ions weakly bonded with surface copper species [48], probably located at interfacial positions.

The quantitative estimation of hydrogen consumption in the low-temperature region is presented in Table 3. The hydrogen uptake increased in this region with increasing the copper content. However, the experimental hydrogen uptakes for all the catalysts exceeded the calculated amounts of hydrogen required for the complete reduction of CuO, assuming that all the copper determined by AAS existed as CuO (see the H2/Cu ratio in Table 3). The additional hydrogen was apparently consumed by the low-temperature reduction of ceria or tin oxide. This fact indicates that the reduction of these oxides is favored by the presence of dispersed surface CuO species, i.e., the larger the number of CuO–ceria or CuO–SnO<sup>x</sup> contacts, the easier the reduction of CeO<sup>2</sup> and/or SnOx proceeds. The H2/Cu ratio decreased with increasing the copper content probably because of the decrease in the surface fraction of oxide support with Ce4+ and Sn4+ species available for low-temperature reduction.

Note that the H2-TPR profile of double CeSn oxide showed a noticeable peak at 308 ◦C. Finer tin oxide particles can be reduced much easier than larger ones [49] due to the improved ability of the former particles to form active oxygen species. In contrast to large SnO<sup>2</sup> particles that reduce at temperatures about 677 ◦C, small nano-sized crystals reduce in a broad temperature range of 100–500 ◦C [50].

The H<sup>2</sup> uptake in the medium-temperature region (400−650 ◦C) was related to the reduction of the ceria surface, and it depended nonmonotonically on the copper loading. The highest H<sup>2</sup> uptake was observed for the 10CuCeSn sample. The shift of this reduction peak towards higher temperatures relative to other catalysts could deteriorate the catalyst activity at relatively high reaction temperatures.

The TEM images of all the catalysts (Figure S5) demonstrated relatively large particle aggregates and individual CeO<sup>2</sup> particles. The interplanar spacings of *d* = 0.31 and 0.27 in these particles can be attributed to (111) and (200) planes of cubic ceria crystals. Moiré fringes observed in nearly all images can result from the interaction of aggregates of epitaxially interfaced nanocrystallites. Neither metallic copper nor copper and tin oxide crystallites were detected by HR-TEM. TEM-EDX also points to the absence of copperenriched areas. Therefore, the fact that copper oxide particles were not detected in the

catalysts confirms our assumption about the high dispersion or amorphous state of copperand tin-containing phases that was made based on the XRD data.

#### **3. Discussion**

Along with the desired CO oxidation reaction (1), three side reactions (hydrogen oxidation (2), the reverse water gas shift (RWGS) (3) and CO hydrogenation (4)) are involved in the PROX process: [51]:

$$\text{CO}\_{\text{(g)}} + 1/2\text{O}\_{\text{2(g)}} \rightarrow \text{CO}\_{\text{2(g)}}; \qquad \qquad \Delta\_{\text{r}}\text{H}^{\text{o}}\_{\text{298K}} = -283.0 \text{ kJ/mol} \tag{1}$$

$$\text{H}\_{2(g)} + 1/2\text{O}\_{2(g)} \rightarrow \text{H}\_2\text{O}\_{(g)}; \qquad \qquad \Delta\_\text{r}\text{H}^\text{o}\_{298\text{K}} = -241.9 \text{ kJ/mol} \tag{2}$$

$$\rm CO\_{2(g)} + H\_{2(g)} \to CO\_{(g)} + H\_2O\_{(g)}; \qquad \Delta\_r H^0 \, \_{298\, K} = 41.1 \text{ kJ/mol} \tag{3}$$

$$\text{CO}\_{\text{(g)}} + 3\text{H}\_{2\text{(g)}} \rightarrow \text{CH}\_{4\text{(g)}} + \text{H}\_{2}\text{O}\_{\text{(g)}}; \qquad \Delta\_{\text{r}}\text{H}^{\text{O}}\_{\text{-298K}} = -206.2 \text{ kJ/mol} \tag{4}$$

The latter reaction usually shows a significant rate over cobalt-containing [51] and noble metal catalysts [10], while the contributions of reactions (2) and (3) are strong for copper and cerium oxide catalysts.

Our thermodynamic calculations demonstrated that the increase in temperature above 50 ◦C gradually decreased the equilibrium CO conversion and CO<sup>2</sup> yield over the Cu/(CeO2-SiO2) catalysts with the simultaneous increase in the H<sup>2</sup> conversion and water yield [36]. In a real experiment, we should consider both kinetic and thermodynamic factors. Reduced copper species in the CuO/CeO<sup>2</sup> catalyst increase the mobility of lattice oxygen in CeO2, which leads to the surface reduction of CeO<sup>2</sup> and decreases CO conversion in favor of the H<sup>2</sup> oxidation [52]. On the other hand, the presence of Cu<sup>+</sup> ions can increase the activation rate of CO molecules on the catalyst surface. High oxygen mobility and electron exchange between Cu2+/Cu<sup>+</sup> and Ce4+/Ce3+ redox pairs have a crucial effect on the catalytic efficiency. In this way, though CeO<sup>2</sup> reduction can be detrimental to the CO conversion, the improved reducibility of the surface of copper particles to form Cu<sup>+</sup> may play a positive role. The high redox properties of copper species mainly result from their interaction with the cerium-containing oxide support. The template synthesis not only improves the textural properties of mixed oxide catalysts but also contributes to the uniform distribution of elements and strong interaction between components in the catalyst.

In the present work, we studied the influence of the method of copper incorporation into the CeSn and CeZr oxide catalysts and Cu content in the one-pot synthesized CuCeSn catalysts on the physicochemical properties of the catalysts and their efficiencies in CO-PROX.

Both 20CuCeSn and 20CuCeZr were active in CO-PROX. The one-pot templateassisted method of copper incorporation provided a higher performance of the ternary oxide catalysts than the post-impregnation of templated double CeSn and CeZr oxides. The tin-doped 20CuCeSn catalyst prepared by the one-pot method was more efficient than its zirconium-doped counterpart in terms of the CO conversion/CO<sup>2</sup> selectivity ratio; it demonstrated about 70% CO conversion and more than 50% CO<sup>2</sup> selectivity.

We experimentally showed that the one-pot co-precipitation of copper, tin and cerium precursors in the presence of the CTAB template led to the uniform distribution of elements in 20CuCeSn and developed micro- and mesoporosity (average pore size of 3.1 nm) with the increased fraction of micropores and the highest SBET of 149 m2/g. Apparently, this method and these precursors are effective for the formation of micellar structures and complexes with citric acid during synthesis. Moreover, the easier reducibility of tin due to the reversible Sn4+ <sup>↔</sup> Sn2+ transition that can proceed at relatively low temperatures can provide an additional redox pair in contrast to zirconium, which is not prone to such transitions [53]. In addition, the one-pot method provides a high dispersion of copper oxide species on the surface and structure defectiveness, as indicated by the absence of copper oxide reflections in the diffraction pattern and a small average crystallite size (about 5 nm), respectively. The H2-TPR analysis revealed the excellent reducibility of 20CuCeSn, which plays mainly a positive role in catalytic oxidation [4], especially at relatively high

temperatures. However, high reducibility in the low-temperature region can deteriorate the selectivity of the process. Based on the CO-PROX test and operando-DRIFTS and XANES results, some authors [42] believed that the main side process of H<sup>2</sup> oxidation could easily proceed immediately after the onset of Cu2+ reduction to Cu<sup>+</sup> and Cu<sup>0</sup> . This process starts easily in dispersed copper oxide nanoparticles that are in loose contact with the ceria-based support. In this way lower CO<sup>2</sup> selectivity of tin-containing catalysts than of their zirconium-containing counterparts at 100–150 ◦C can be explained.

The change in the copper content in the one-pot synthesized ternary CuCeSn oxide catalyst also strongly affects its catalytic properties. It should be emphasized that, in our work, the copper content was calculated in mole fractions related to the total content of Ce + M (M = Sn or Zr) as (n(Cu)·100/[(n(Ce)+n(M)]). The maximum copper loading did not exceed 7.5 wt.%, i.e., there was not too much copper in all the catalysts.

The CuCeSn samples with different Cu contents prepared by the one-pot method demonstrated the increased defectiveness (broadening and shift of the F2g line in the Raman spectra) and a high SBET (Table 2). For all the samples, a strong interaction between the cerium–tin oxide particles and finely dispersed copper oxide species was observed. This fact was confirmed by the shift of the F2g band in the Raman spectra relative to the spectrum of CeSn and the absence of reflections of crystalline copper oxide phases in XRD and HR-TEM diffraction patterns.

On the other hand, a change in the copper content affected several physicochemical properties of the catalysts, which determined the differences in their catalytic performances. Moreover, some properties changed nonmonotonically (Figure 5b). Thus, the SBET of 10CuCeSn was a bit higher than those of the catalysts with higher and lower Cu contents. The increase in the total copper content led to the growth of the surface copper concentration and improved the copper reducibility, which shifted the copper reduction peaks in the H2-TPR profiles to lower temperatures. This fact was confirmed by SEM-EDX and XPS. However, again, for 10CuCeSn, both the α and β peaks of copper reduction are located at lower temperatures than for the other samples. Nonmonotonic dependences of the fractions of active oxygen and Cu<sup>+</sup> species on the surface on the copper loading in the catalysts were observed. The extrema of these dependencies were also observed for the 10CuCeSn catalyst. Therefore, the minimum Olat/[Oads + OOH] ratio of 1.75, the highest value of the specific surface area of 183 m2/g, the maximum Cu+/Cu2+ ratio of 1.08 and a higher reduction ability at 400–650 ◦C (Table 3) were inherent to the 10CuCeSn sample. However, hydrogen absorption in the low-temperature region grew monotonically with increasing the Cu content (Table 3). It was lower for 10CuCeSn (1213 µmol/g) than for the samples with higher copper contents.

It is important to note that the drop in the ratio of the experimental and theoretical values of hydrogen consumption in the low-temperature region (H2/Cu ratio) and the increase in the temperature gap between the α and β reduction components (∆Tαβ) with increasing the copper content can indicate a decrease in the fraction of easily reducible copper sites. This means that there is a limit of copper "solubility" in CeSn oxide that was approached in the 5CuCeSn and 10CuCeSn samples. This limit can be characterized both by the formation of solid solutions and by high concentrations of contacts between the cerium–tin oxide and fine CuO particles.

A higher reducibility of 10CuCeSn in the medium-temperature interval (400–650 ◦C) is very important. It resulted from the easier reduction of surface Ce4+ ions in direct contact with copper sites in a mixed oxide phase or at interfacial positions. The authors of [42] studied the redox–catalytic correlations in copper–ceria CO-PROX catalysts and concluded that CO predominantly oxidized over interfacial positions of the partially reduced dispersed copper sites. This fact explains the high CO<sup>2</sup> selectivity of 10CuCeSn in the low-temperature region in which there is no noticeable phase segregation, sintering of active phase particles and formation of metallic copper. However, CO conversion over this catalyst was close to 100% only at 150 ◦C. At higher temperatures, it was inferior to the catalysts with lower and higher copper contents.

25CuCeSn with the highest copper content showed the best catalytic performance at 50 ◦C and exhibited excellent CO conversion of 79% and CO<sup>2</sup> selectivity of 97% at 100 ◦C. This catalyst demonstrated a sufficiently easy reducibility of copper, albeit at slightly higher temperatures, and only a slightly lower Cu+/Cu2+ ratio on the surface than in 10CuCeSn. However, the high content of copper in 25CuCeSn provided ample amount of adsorption sites for CO molecules. That is why this catalyst surpassed 10CuCeSn in the catalytic activity at 100 ◦C. However, a large copper fraction in 25CuCeSn weakly interacted with CeSn support, so, at higher temperatures, these copper species were easily reduced with hydrogen from the reaction medium, leading to a sharp drop in the CO<sup>2</sup> selectivity. The surface of the reduced catalyst was active in hydrogen oxidation, which led to hydrogen loss.

The same trend was found for 20CuCeSn. However, because of the lower concentration of the active adsorption sites in 20CuCeSn than in 25CuCeSn, the CO conversion and CO<sup>2</sup> selectivity at 100 ◦C were lower for the former catalyst. It was found early that the increase in the total copper content in the CuO/CeO<sup>2</sup> catalyst significantly improved its reducibility [39]. We can assume that only a limited fraction of copper in the 20CuCeSn catalyst was in tight contact with the CeSn support and could provide interfacial sites active in CO-PROX. The rest part of copper weakly interacted with CeSn particles. Despite their easy reducibility, these species had no influence on the ceria/tin reduction and oxygen vacancies formation, which is consistent with the H2-TPR and XPS data.

An interesting behavior was observed for 5CuCeSn. This catalyst, comprising only a small number of copper-active centers, was inferior to other samples in terms of the CO conversion at low temperatures (only 47% at 100 ◦C). However, at higher temperatures, the CO conversion over this catalyst was comparable with those over 20CuCeSn and 25CuCeSn with much higher copper loadings. The reaction medium possibly affected the properties of this sample to a lesser extent than for 10 CuCeSn, which showed lower CO conversions.

It seems that achieving a high efficiency of the Cu-Ce-Sn ternary oxide catalysts requires not only the developed porosity and highly dispersed active copper oxide species but also the optimal loading of copper that increases the concentration of interfacial centers between copper oxide and cerium–tin oxide and their stability over a wide temperature range.

Thus, in this work, we demonstrated that the one-pot template method is more efficient for the synthesis of the CuCeSn and CuCeZr CO-PROX catalysts than the postimpregnation of double oxides with copper salt. Varying the copper loading in the catalysts allowed tuning the CO conversion/CO<sup>2</sup> selectivity ratio. Nonmonotonic changes in the physicochemical characteristics with increasing Cu loading was explained by the differences in copper distribution in the support; at low loading copper was predominantly distributed in the CeSn phase, while at higher loadings, separate copper oxide phases were formed, which increased the interfacial interactions between the copper-containing and CeSn phases. The high stability of 10CuCeSn, containing only 3.4 wt.% of copper, was demonstrated in the 10-h catalytic test.

#### **4. Materials and Methods**

#### *4.1. Catalysts Preparation*

Double CeO2-SnO<sup>2</sup> (denoted as CeSn) and ternary CuO-CeO2-SnO<sup>2</sup> (denoted as CuCeSn) oxide catalysts were synthesized by the template-assisted evaporation-induced self-assembly method (EISA) described in detail in our previous work [54]. Cetyltrimethylammonium bromide (CTAB, 99%, BioChemica, Billingham, UK) was used as a template. The *n*(CTAB):Σ*n*(Me) ratio was 1:1. (NH4)2[Ce(NO3)6] (puriss., Reachem, Moscow, Russia), SnCl2·2H2O (puriss., Reachem, Moscow, Russia) and Cu(CH3COO)2·H2O (p.a., Vecton, Saint Petersburg, Russia) were used as precursors for the synthesis. The Ce:Sn and CeZr molar ratios were 9 for all the samples. The high cerium to tin ratio in the catalysts was chosen based on the literature data.

Synthesis of CeSn. The required amounts of cerium and tin salts were dissolved in 70 mL of distilled water and added to 30 mL aqueous solution of citric acid. The resulting solution was stirred at a constant rate. Then, the mixture was added dropwise to the solution of CTAB in 15 mL of 96% ethanol and stirred at room temperature for 4 h. After that, the excess of the solvent was slowly evaporated for 2 h. The obtained gel-like mixture was dried at 80 ◦C for 24 h and then heated up to 500 ◦C and calcined in air at this temperature for 3.5 h.

The ternary CuCeSn oxide catalyst was prepared using two different techniques. The first technique was the one-pot synthesis. The aqueous solution of all the precursors was added to the CTAB solution in ethanol, as described above. The copper molar ratios in the catalysts (*n*(Cu)·100%/[*n*(Ce)+*n*(Sn)]) were 5, 10, 20 and 25. These samples were denoted as *y*CuCeSn, where *y* is the copper molar ratio. In the second technique, CeSn oxide was prepared as described above and calcined at 500 ◦C. Then, it was impregnated with the aqueous solution of copper acetate, followed by drying at 120 ◦C for 2 h and calcination in the air at 500 ◦C for 2 h (this sample was denoted as 20CuCeSn-im).

Reference CeO2, 20CuCeZr and 20CuCeZr-im samples with a Ce:Zr ratio of 9 were prepared by the same methods (one-pot and post-impregnation) using (NH4)2[Ce(NO3)6] (puriss., Reachem, Moscow, Russia) and ZrO(NO3)2·H2O(99.5%, Acros Organics, Geel, Belgium) as cerium and zirconium precursors, respectively.

The results of the catalytic test of the unmodified cerium–tin catalyst are not presented. The maximum CO conversion was only 47% at 250 ◦C, which is significantly lower than that for the Cu-containing sample.

#### *4.2. Catalysts Characterization*

The copper content in the catalysts was measured by atomic absorption spectroscopy (AAS) on a Thermo Fisher Scientific series iCE 3000 spectrometer (Waltham, MA, USA) using air–acetylene flame atomization. SOLAAR Data Station software was used for the device control and data processing. The uncertainty of the copper content measurements did not exceed ± 0.2 wt.%.

XRD diffractograms of the catalysts were recorded on a Rigaku Ultima IV powder diffractometer (Rigaku Corporation, Tokyo, Japan) (CuKα radiation, 1.5418 Å) in the 2Ө range of from 5 to 90◦ with a step size of 0.02◦ . The phase composition was analyzed by comparison with the JCPDS PDF1 library data (ICDD database).

X-ray photoelectron spectroscopy (XPS) was used to reveal the composition and oxidation state of elements on the catalyst surface. The spectra were acquired on an Axis Ultra DLD spectrometer (Kratos Analytical, Manchester, UK) with a monochromatic Al*K<sup>α</sup>* radiation source (hν = 1486.7 eV, 150 W). The pass energies of the analyzer were 160 eV for the survey spectra and 40 eV for the high-resolution scans. The binding energy scale of the spectrometer was preliminarily calibrated using the position of the peaks for the Au 4f7/2 (83.96 eV), Ag 3*d*5/2 (368.21 eV) and Cu 2p3/2 (932.62 eV) core levels of pure metallic gold, silver, and copper. The powder samples were mounted on a holder using double-sided nonconductive adhesive tape. The spectra were fitted into CasaXPS software. The Kratos charge neutralizer was used, and the spectra were charge-referenced to the high energy component of the Ce3d spectrum set to 916.5 eV.

Raman spectra were recorded on a Horiba Jobin Yvon Lab RAM HR 800 UV instrument (Horiba ABX S.A., Montpellier, France). An argon ion laser with a wavelength of 514 nm was used for the excitation, and the power on the sample did not exceed 7 mW. For each sample, the spectra were accumulated for 200 s.

Scanning electron microscopy (SEM) of the catalysts was performed on a JCM–6000 Neoscope microscope (JEOL, Tokyo, Japan). The high-resolution transmission electron microscopic images (HR-TEM) were registered using a JEM 2100A instrument (JEOL, Japan) at an accelerating voltage of 200 kV. Both microscopes were equipped with an energy dispersive X-ray spectroscopy (EDX) accessory.

Nitrogen physisorption isotherms were recorded on an Autosorb–1 analyzer (Quantachrome, Boynton Beach, FL, USA). Before measurements, the samples were outgassed for 3 h at 200 ◦C. The specific surface area was calculated by the BET method with an accuracy

of 10%. Desorption branches of the isotherms were used for the calculation of pore size distributions by the NL-DFT method.

Temperature-programmed reduction with hydrogen (H2-TPR) was carried out on a USGA-101 chemisorption analyzer (UNISIT, Moscow, Russia). A 5% H2/Ar mixture was fed to a quartz reactor at a flow rate of 30 mL/min. The sample weight was approximately 50 mg. Before analysis, a catalyst was preliminarily kept at 300 ◦C for 30 min in an argon flow and then cooled to 30 ◦C. The H2-TPR profiles were recorded under heating the sample from 30 to 900 ◦C at a rate of 10 ◦C/min. Hydrogen consumption during analysis was registered with a thermal conductivity detector preliminary calibrated by NiO (99.99%, Sigma-Aldrich, St. Lois, MO, USA) reduction.

#### *4.3. Catalytic Tests*

The catalysts were tested in CO-PROX on a ULCat-1 catalytic unit (UNISIT, Moscow, Russia) equipped with a fixed-bed continuous-flow stainless steel reactor (length—32 cm, inner diameter—12 mm and outer diameter—16 mm) in the temperature range from 50 to 450 ◦C using 150 mg of the catalyst. The reaction mixture comprising 4 vol.% CO, 3 vol.% O2, 13 vol.% N<sup>2</sup> and 80 vol.% H<sup>2</sup> was fed into the reactor at a flow rate of 95 mL·min−<sup>1</sup> (contact time of 0.095 g·s·cm−<sup>3</sup> ). The composition of the effluent was analyzed by a Chromatec-Crystal 5000.2 (Chromatec, Yoshkar-Ola, Mari El, Russia) gas chromatograph (GC) equipped with a Carboxen-1010 PLOT column (Supelco Inc., Bellefonte, Pennsylvania, USA). Air was supplied by an air compressor. He (40 L, grade "A"), N<sup>2</sup> (40 L, high purity grade), CO (40 L, grade 3.0) and H<sup>2</sup> (40 L, grade 3.8) were supplied by the PGS-service company (Moscow, Russia) and used without additional purification. The gases were fed to the reactor using mass flow controllers (Bronkhorst Nederland B.V., Bronkhorst, the Netherlands).

The main catalytic parameters were calculated from the areas of chromatographic peaks (*A*) of each component using the following equations:

$$\text{CO conversion}: \quad X\_{\text{CO}}\ \%= \frac{[A(\text{CO})\_{\text{in}} - A(\text{CO})\_{\text{out}}\Delta\frac{A(\text{N}\_2)\_{\text{in}}}{A(\text{N}\_2)\_{\text{out}}}] \times 100}{A(\text{CO})\_{\text{in}}} \tag{5}$$

$$\text{CO}\_2\text{ conversion}:\quad X\_{\text{O}\_2\prime}\,\%=\frac{[A(\text{O}\_2)\_\text{in} - A(\text{O}\_2)\_{\text{out}}\Delta\frac{A(\text{N}\_2)\_{\text{in}}}{A(\text{N}\_2)\_{\text{out}}}] \times 100}{A(\text{O}\_2)\_{\text{in}}}\tag{6}$$

$$\text{CO2 selectionity}: \quad \text{S}\_{\text{CO}\_2}\text{\%} = \frac{0.5 \Delta f(\text{CO}) \Delta [A(\text{CO})\_{\text{in}} - A(\text{CO})\_{\text{out}} \Delta \frac{A(\text{N}\_2)\_{\text{in}}}{A(\text{N}\_2)\_{\text{out}}}] \times 100}{f(\text{O}\_2) \Delta [A(\text{O}\_2)\_{\text{in}} - A(\text{O}\_2)\_{\text{out}} \Delta \frac{A(\text{N}\_2)\_{\text{in}}}{A(\text{N}\_2)\_{\text{out}}}]} \tag{7}$$

The subscripts "in" and "out" denote the concentrations of components in the feed and effluent of the reactor; *f*(CO) and *f*(O2) are the calibration factors for CO and O2, respectively. The calibration factors were calculated based on the chromatographic analysis of two calibration mixtures (PGS-service, Moscow, Russia): (i) CH<sup>4</sup> (30.20 vol.%), H<sup>2</sup> (4.81 vol.%), CO (4.87 vol.%) and balance N<sup>2</sup> and (ii) O<sup>2</sup> (14.90 vol.%), CO<sup>2</sup> (29.18 vol.%) and balance N2.

The stability of the 10CuCeSn catalyst was tested in the isothermal experiment at 200 ◦C for more than 10 h using the same reaction mixture as in the non-isothermal experiments. However, the increased catalyst loading of 300 mg and a total flow rate of 30 mL·min−<sup>1</sup> were used (contact time of 0.600 g·s·cm−<sup>3</sup> ).

The uncertainties of the calculated conversion and selectivity values did not exceed ±4%.

#### **5. Conclusions**

Thus, in this work, we demonstrated that the one-pot template method allows synthesizing more efficient CuCeSn and CuCeZr catalysts for CO-PROX than the postimpregnation of double CeSn/CeZr oxides with copper salt. A small amount of tin dopant more strongly improves the reducibility of cerium-based catalysts than the same amount of

zirconium. Moreover, the transformation of Sn4+ to Sn2+ can provide an additional redox pair in the catalyst, which positively affects its activity. The one-pot technique promotes the uniform distribution of elements in the 20CuCeSn catalyst and develops its micro- and mesoporosity. Varying the copper loading allows tuning the CO conversion/CO<sup>2</sup> selectivity ratio. Nonmonotonic changes in the physicochemical characteristics of the CuCeSn catalyst with increasing the copper loading can be explained by differences in the copper distribution in the support. At low loading, copper is predominantly distributed in the CeSn phase, while, at higher loadings, separate copper oxide phases are formed, which increases the interfacial interaction between copper and CeSn oxides.

**Supplementary Materials:** The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/catal12121575/s1: Figure S1: SEM images and SEM-EDX elemental mappings of CuCeSn and CuCeZr catalysts. Figure S2: N<sup>2</sup> physisorption isotherms of 20CuCeSn and 20CuCeZr catalysts synthesized by the one-pot method and post-impregnation method. Figure S3: Survey XPS spectra of 20CuCeSn and 20CuCeZr catalysts prepared by different methods (a), and one-pot synthesized CuCeSn catalysts with different copper loadings (b). Figure S4: SEM images and SEM-EDX mappings of one-pot synthesized CuCeSn catalysts with different copper loadings. Figure S5: HR-TEM images of one-pot synthesized CuCeSn catalysts with different copper loadings.

**Author Contributions:** Conceptualization, I.Y.K. and E.S.L.; methodology, I.Y.K.; formal analysis, I.Y.K. and E.S.L.; investigation, I.Y.K., A.V.T., K.I.M. and O.Y.I.; writing—original draft preparation, I.Y.K. and E.S.L.; writing—review and editing, I.Y.K., E.S.L., E.V.G. and K.I.M.; visualization, I.Y.K. and A.V.T. and supervision, E.S.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Russian State assignment "Physical Chemistry of Surface, Adsorption, and Catalysis", registration number AAAA-A21-121011990019-4.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors acknowledge support from the Lomonosov Moscow State University Program of Development for providing access to XPS and TEM instruments.

**Conflicts of Interest:** The authors declare no conflict of interest.

**Sample Availability:** Samples of the catalysts are not available from the authors.

#### **References**

