In recent decades, the number of cars powered by diesel engines has significantly increased owing to their good dynamic performance [1, 2]. However, the soot particulate emitted from diesel engines is one of the most hazardous pollutants that can cause severe environmental and health problems [3-5]. To eliminate soot particle pollution, filtering it with a DPF trap followed by catalytic oxidation is an efficient current strategy [6]. To meet the increasingly stringent emission regulations, it is still a major challenge to develop highly active and stable catalysts for soot particulate combustion, especially a catalyst that can match the low exhaust temperature of diesel engines.

In the late 1970s, Xie et al. [7-9] found that a metal oxide can spontaneously disperse onto the surfaces of different supports to form a monolayer or sub-monolayer, which is a thermodynamically favored process that has been termed by them as "monolayer dispersion". A "close-packed model" has been proposed by them for calculating the ideal and maximum monolayer dispersion capacities of supported metal oxides on different supports. Moreover, X-ray diffraction (XRD), Raman, and X-ray photoelectron spectroscopy (XPS) extrapolation methods have been developed by them to quantify the real monolayer dispersion capacities of different supported systems. It is revealed that in most cases, the real monolayer dispersion capacity of a supported system is lower than the ideal value calculated by the "close-packed model" owing to the interaction between the supported component and the support [9]. Chen et al. [10-13] studied the dispersion of CuO and MoO₃ on various supports such as Al₂O₃, TiO₂, CeO₂, and CeO₂-Al₂O₃ etc. and found the same phenomenon. However, an "incorporation model" was proposed by them to explain the phenomenon. By using different supports, Wachs et al. [14-16] explored the dispersion behaviors of supported MoO₃ and V₂O₅ and discovered a strong interaction between the supports and the supported metal oxides. On the basis of their work, the surface oxide-support interaction theory was proposed. In spite of the different terminologies used, all the researchers have found that when a metal oxide is loaded onto a support, the monolayer dispersion phenomenon is commonly observed, and the monolayer capacity can be quantified by different methods. Furthermore, when a supported system is used as a catalyst for a certain reaction, a monolayer dispersion threshold effect is generally observed. In more detail, a catalyst with a monolayer dispersed active component generally shows some special reaction performance, such as the highest activity and selectivity [14]. Nowadays, monolayer dispersion has been well accepted and has become an important theory in the design of highly functional catalytic materials.

Numerous studies have proved that CuO is very active towards many catalytic oxidation reactions [17-23]. For instance, the famous Hopcalite catalyst, which typically consists of 30%-40% copper oxide and 60%-70% manganese oxide, displays superior activity for CO oxidation at room temperature and has been practically used [17, 18]. For CO oxidation on CuO, Boronin et al. [19] found that O^- species is responsible for its superior CO oxidation activity. Kakuta et al. [20] found that when starting from stoichiometric CuO, the valency of the surface copper cycles as per Cu²⁺ ⇌ Cu⁰ during the reaction process, which might determine the reaction rate. In contrast, Tikhov et al. [21] revealed that Cu⁺ clusters at the outlets of surface defects, which are paths for quick oxygen migration from the CuO bulk, are active sites that account for the CO oxidation activity. Through DFT calculation, Sun et al. [22] investigated the adsorption and dissociation of molecular O₂ on CuO (111) surface. It was disclosed that both the surface species O₂^2- and O₂^- can be effectively formed as active oxygen sites.

Our earlier work also demonstrated that mesoporous copper-tin solid solutions prepared through coprecipitation [24] or hydrothermal method with KIT-6 as the hard template [25] show superior activity towards CO oxidation, owing to the formation of a large amount of surface-active oxygen species, high surface areas, and porous structures. Moreover, by using a colloidal crystal templating method, a 3DOM structured copper-tin solid solution catalyst can be obtained, which exhibits very good soot combustion activity because the coexistence of meso- and macro-pores favors the contact of soot particles with the surface-active oxygen species [26]. Therefore, in the work that continued, the structure-reactivity relationship of pure CuO for soot oxidation was particularly investigated [27]. It was discovered that for CuO prepared by using different copper precursors and precipitants, four kinds of active sites with varied activities were present on the surface, but in different amounts. However, only the two kinds of active sites that are reducible by soot particulates below 500 ℃ contribute to soot oxidation. The mobility and abundance of surface oxygen are critical factors that control the soot oxidation activity of pure CuO catalysts.

However, bare CuO generally has low surface areas and physical stability, which restrict its feasibility as a catalyst for most of the exothermic processes such as soot combustion, where it experiences deactivation [27]. SnO₂ is a typical n-type semiconductor that has a high melting point of 1630 ℃ [28]. Many earlier studies, including those of our group, have proved that it has not only abundant surface-deficient oxygen species, but also good physical stability [29]. Therefore, SnO₂ could be an appropriate support for many active components in the preparation of catalysts for reactions that require good stability more than a high intrinsic activity, e.g., some strong exothermic processes such as soot and VOC combustion. However, its properties as a catalyst support have been rarely investigated. With the objective to understand the interfacial interaction between CuO and SnO₂, and eventually realize more applicable catalysts for real soot combustion processes, in this study, a series of CuO/SnO₂ catalysts have been prepared and used for soot oxidation based on the monolayer dispersion theory. An evident monolayer dispersion threshold effect has been observed, and the catalysts with CuO loadings not below the monolayer dispersion capacity show the best activity towards soot oxidation. By using different characterization methods, the reasons for the observed activity have been elucidated in this work.

All the CuO/SnO₂ catalysts were prepared by the traditional impregnation method with a Cu(NO₃)₂ (Sinopharm Chemical Reagent Corporation, China) solution (0.58 M) acting as the precursor. In detail, a certain amount of CVD SnO₂ (Shanghai Zaiyi New Material Co., Ltd., China) was added into the Cu(NO₃)₂ solution under constant stirring in a water bath at 80 ℃ for 3 h. After evaporating the solvent water, the solids obtained were dried at 110 ℃ overnight for approximately 12 h and then calcined at 550 ℃ in air atmosphere for 4 h, before being heated to the target temperature at the rate of 2 ℃ min^-1. Afterwards, the samples were cooled to room temperature to obtain the final CuO/SnO₂ samples, which are labeled as 1% CuO/SnO₂, 3% CuO/SnO₂, 5% CuO/SnO₂, 7% CuO/SnO₂, 9% CuO/SnO₂, 11% CuO/SnO₂, 13% CuO/SnO₂, and 15% CuO/SnO₂, according to their CuO loadings.

The powder XRD patterns of the samples were recorded on a Bruker AXS D8Focus diffractometer operating at 40 kV and 30 mA with a copper target and K_α radiation. The scans were collected over the 2θ range 20° to 80° in steps of 2° min^-1.

XPS tests were carried out on a PerkinElmer PHI1600 system by using a single Mg-K X-ray source operating at 300 W and 15 kV. The spectra were obtained at ambient temperature and ultrahigh vacuum. The binding energies were calibrated by using the C 1s peak of a standard graphite sample (at 284.8 eV) as the reference.

N₂ adsorption-desorption experiments of the samples were performed at -196 ℃ on an ASAP2020 instrument. The specific surface areas of the catalysts were calculated by using the Brunauer-Emmett-Teller (BET) method in the relative pressure (P/P₀) range 0.05-0.25. The pore size distributions of the samples were calculated through the Barrett-Joyner-Halenda method. Furthermore, the pore sizes were obtained from the peak positions of the distribution curves. The total pore volume was calculated at the relative pressure of P/P₀ = 0.99.

Raman spectra of the catalysts were recorded on a Renishaw inVia Raman spectrometer equipped with an argon laser excitation source (532 nm) and a Renishaw RenCam CCD detector. The scanned Raman shift range varied from 200 to 1100 cm^-1.

Transmission electron microscopy (TEM) images were recorded on a TecnaiTM F30 transmission electron microscope. STEM mapping images of some typical catalysts were also obtained by using the TecnaiTM F30 instrument that was equipped with an Oxford EDX detector operating at 300 keV.

H₂ temperature-programmed reduction (H₂-TPR) experiments were also carried out by using a FINESORB 3010C instrument. In general, 0.05 g of the catalysts was used for the tests. Before the experiments, the catalysts were pretreated in high-purity air that flowed at 120 ℃ for 30 min to remove any possible surface impurities. After purging with ultrahigh purity argon that flowed at room temperature for 30 min, the temperature was then increased from room temperature to 800 ℃, with the ramp rate being 10 ℃ min^-1, in 30 mL min^-1 10% H₂/Ar gas mixture flow. A thermal conductivity detector (TCD) was employed to monitor the H₂ uptake. To quantify the amount of H₂ consumed, a 99.99% CuO sample was used as the calibration standard.

Temperature-programmed oxygen desorption (O₂-TPD) measurements were performed on a DAS-7000 multiple adsorption instrument that was equipped with a TCD detector. Typically, 0.05 g of the sample was placed in a quartz reactor, which was pretreated in 30 mL min^-1 ultrahigh purity argon flow at 300 ℃ for 60 min. Afterwards, the sample was cooled to 50 ℃ and saturated in a 10% O₂+Ar flow at the rate of 30 mL min^-1, which was followed by flushing in 30 mL min^-1 ultrahigh purity argon flow for 60 min to remove any physically adsorbed O₂. The TPD experiments were then carried out from 50 to 700 ℃ at the heating rate of 10 ℃ min^-1 in 30 mL min^-1 ultrahigh purity argon flow.

Soot temperature-programmed reduction (soot-TPR) experiments were conducted by using a TP-5076 multiple adsorption instrument (Xian-quan, China). Typically, 55 mg of the sample, which was prepared by uniform mixing of 5 mg soot with 50 mg catalyst powder, was placed in a quartz reactor. Prior to the test, it was treated in ultrahigh purity argon flow for 60 min at 120 ℃ to remove any physically adsorbed impurities. Afterwards, soot-TPR was carried out from 50 to 800 ℃ at the heating rate of 10 ℃ min^-1 in 30 mL min^-1 ultrahigh purity argon flow. A TCD was employed to monitor the formation of CO₂.

Temperature-programmed oxidation (TPO) was used to evaluate the soot combustion activities of the catalysts. To determine the activity of a catalyst under tight contact condition, typically, 5 mg of soot (Printex-U, diameter 25 nm, purchased from Degussa) was mixed with 50 mg of catalyst powder and ground for 10 min. Before loading into a microreactor with an inner diameter of 6 mm, the mixture was diluted by adding 100 mg of inert silica to ensure a soot/catalyst/silica weight ratio of 1/10/20 to avoid the formation of hot spots during the activity test. A K-type thermocouple was used to accurately monitor the temperature of the catalyst bed, with the thermocouple head point touching the catalyst. To examine the reaction behaviors of the catalysts, all data were collected by increasing the temperature from 120 to 800 ℃ at the rate of 10 ℃ min^-1. The volume composition of the feed gas was 10% O₂, with the balance being high-purity argon; the flow rate was 30 mL min^-1. To accurately quantify the amount of O₂ consumed, prior to entering the TCD, the CO₂ formed was removed thoroughly by using a soda-lime trap.

XRD was used to determine the phase compositions of the CuO/SnO₂ samples, and the patterns are shown in Fig. 1(A). For the pure SnO₂ support, tetragonal rutile SnO₂ is the only crystalline phase detected, as evidenced by the three typical diffraction peaks: (110) = 26.79°, (101) = 34.12°, and (211) = 52.10° (JCPDS no. 41-1445). For pure CuO, monoclinic CuO is the only detected crystalline phase, as indicated by the three typical sharp peaks: (11-1) = 35.68°, (111) = 38.95°, and (20-2) = 49.04° [27, 30]. For all the CuO/SnO₂ samples with different loadings, the rutile SnO₂ phase is obviously detected with very similar peak intensities, proving that the addition of different amounts of CuO has no evident impact on the crystallinity of the SnO₂ support. Interestingly, when the CuO loading is below 5%, no diffraction peak related to it can be observed, which indicates that CuO is dispersed finely on the support surface and is nearly amorphous. However, when the CuO loading is above 5%, the monoclinic CuO phase starts to appear, the intensity of which increases with the increase in the CuO loading.

According to earlier studies, for a metal oxide dispersed on a support, there exists a monolayer dispersion capacity [9]. When the loading is not higher than this capacity, a sub-monolayer or monolayer of the supported metal oxide will be formed. On the other hand, when the loading is above this capacity, the excess supported metal oxide will start forming micro-crystallites, the average size of which increases with loading until the size reaches 5 nm and can be detected by XRD. Therefore, to determine the monolayer capacity accurately, the XRD extrapolation method was developed previously by Xie et al. [9, 31]. Based on this method, the monolayer dispersion capacity of CuO on SnO₂ support was correlated and quantified, as shown in Fig. 1(B). In detail, the intensity of the CuO (11-1) peak of each sample was measured and divided by the intensity of the SnO₂ (110) peak to obtain a series of I_CuO/I_SnO2 ratios, which are then plotted against the CuO loading to obtain a correlation line. The line intersects the x-axis at a point, as displayed in Fig. 1(B). The CuO loading corresponding to this point is considered as the monolayer dispersion capacity of CuO on SnO₂ support, which is 2.09 mmol 100 m^-2 SnO₂ surface. Upon conversion, this monolayer dispersion capacity equals 4.8 wt% loading.

XPS technique was employed to evaluate the surface properties of some typical CuO/SnO₂ samples (Fig. 2). Fig. 2(A) presents the XPS spectra of the SnO₂ support, which displays two Sn 3d peaks at 487.2 and 495.6 eV that are assigned to Sn 3d_5/2 and 3d_3/2 in that order [31]. In contrast, CuO reveals two typical Cu 2p peaks at 934.2 and 953.0 eV, as displayed in Fig. 2(B), which are assigned to Cu 2p_3/2 and 2p_1/2 in that order [32]. In all the CuO/SnO₂ samples, the two typical SnO₂ and CuO peaks can also be observed, but with some slight shifts. In comparison with the binding energies of pure CuO and SnO₂, the binding energy of CuO in the CuO/SnO₂ samples increases, but that of SnO₂ decreases, which indicates that the CuO present on the surface interacts with the SnO₂ support and donates electrons to it.

With the increase in the CuO loading, the intensity of the Sn 3d peak decreases, whereas that of the Cu 2p peak increases, as observed in Fig. 2(A) and 2(B). Therefore, to further confirm the monolayer dispersion capacity of CuO on SnO₂ support that was determined by XRD, the XPS extrapolation method was also used to measure the capacity [9]. In detail, the intensity of the Cu 2p_1/2 peak is divided by that of the Sn 3d_5/2 peak for each sample to obtain a series of I_Sn3d/I_Cu2p, which are then plotted against the CuO loadings of the samples. As exhibited in Fig. 2(C), two linearly correlated lines with different slopes are obtained, which intersect each other at the point corresponding to 2.02 mmol 100 m^-2 SnO₂. According to previous studies [9], the CuO loading corresponding to this point corresponds to the monolayer dispersion capacity of CuO on SnO₂. Within the experimental error range, this value is the same as that of 2.09 mmol 100 m^-2 SnO₂, which is the monolayer dispersion capacity obtained by the XRD extrapolation method.

In brief, the results obtained by using the XRD and XPS extrapolation methods are highly consistent with each other, testifying that CuO exhibits a monolayer dispersion capacity of around 2.09 mmol 100 m^-2 on SnO₂ support, which equals 4.8 wt% CuO loading. Above this capacity, the excess CuO starts to form CuO micro-crystallites, the average grain sizes of which increase until they are detectable by XRD.

The texture properties of the catalysts were evaluated through N₂ adsorption-desorption experiments, and the results are shown in Fig. 3. It is noted here that with the increase in the CuO loading, the surface areas of the CuO/SnO₂ samples decrease slightly, as observed in Table 1. For all the samples containing the SnO₂ support, Fig. 3(A) indicates that a type Ⅳ isotherm and an H3-type hysteresis loop in the relative pressure (P/P₀) range 0.8-1.0 are observed. Fig. 3(B) reveals that the pore size distribution profiles of the samples are similar upon changing the CuO loading, proving that CuO is dispersed finely on the SnO₂ support, without any evident impact on its texture properties.

As quantified by the XRD and XPS extrapolation methods, the monolayer dispersion capacity of CuO on SnO₂ support is 2.09 mmol 100 m^-2, which equals a CuO loading of 4.8%. Therefore, to investigate the changes in the surface properties of the CuO/SnO₂ samples by altering the CuO loading, three typical catalysts with CuO loadings below (1%), close to (5%), and above (9%) the monolayer dispersion capacity were analyzed by TEM. Fig. 4(A, C, and E) reveals that all the three samples are composed of irregular spherical particles, indicating that the change in the CuO loading has no evident impact on the morphology.

Interestingly, the HR-TEM image of 1% CuO/SnO₂ in Fig. 4(B) shows only the (101) and (110) diffraction planes of the SnO₂ support. None of the CuO diffraction peaks are observed, which testify the fact that below the monolayer capacity, CuO is present on the surface in an amorphous state. However, in Fig. 4(D) and (F), which display the HR-TEM images of 5% CuO/SnO₂ and 9% CuO/SnO₂, besides the diffraction planes of SnO₂, the Cu (11-1) plane of crystalline CuO is obviously detected, with the planar distance being 0.260 nm. This demonstrates that when the CuO loading is above the monolayer dispersion capacity, the excess CuO forms micro-crystallites on top of the CuO monolayer.

To obtain more direct information on the CuO dispersion, the three typical samples were further studied by STEM mapping, and the images are exhibited in Fig. 5. For 1% CuO/SnO₂, elemental copper is distributed highly homogeneously along with elemental tin and oxygen, suggesting that it is dispersed very finely on the SnO₂ support surface. However, in the mapping images of 5% CuO/SnO₂, some local CuO aggregation is evidently observed, as indicated by the randomly bright copper mapping spot, which is not accompanied by any elemental tin. This indicates that some excess CuO has already formed in the micro-crystallites. In contrast, in the images of 9% CuO/SnO₂, owing to the formation of a much larger amount of CuO micro-crystallites, the copper mapping signal becomes completely bright, with some of the copper species not being accompanied by elemental tin.

In summary, both the HR-TEM and STEM mapping results reveal that when the CuO loading is below the monolayer dispersion capacity, CuO is present in an amorphous monolayer dispersed state on the support surface. On the other hand, if the CuO loading is above the capacity, crystalline CuO particles will then be formed. These microscopy results provide direct evidence that confirm what was observed through XRD and XPS.

The surface properties of the CuO/SnO₂ samples were further analyzed by Raman spectroscopy in the shift range 200-1100 cm^-1, and the results are displayed in Fig. 6. For comparison purpose, the Raman spectra of pure CuO and SnO₂ support were also collected. Monoclinic CuO displays three typical Raman bands, which are A_g at 275 cm^-1 and B_g at 320 and 605 cm^-1 [27, 33]. The CVD SnO₂ support used in this study exhibits tetragonal rutile structure, which belongs to the D144h space group. According to group theory, its active Raman modes are B_1g, E_g, A_1g, and B_2g, and theoretically four first-order Raman peaks should be observed [34]. However, as shown in Fig. 6, only three Raman peaks at 478, 636, and 780 cm^-1, corresponding to the E_g, A_1g, and B_2g vibration modes, respectively, are detected.

For the CuO/SnO₂ samples, when the CuO loading is below 5%, the three typical Raman peaks of rutile SnO₂ are still evidently observed, but their intensities decrease with the increase in the CuO loading. In addition, the A_1g and B_2g bands shift to lower values gradually, which suggest interaction between CuO and SnO₂ support. In comparison, when the CuO loading is above the monolayer dispersion capacity and reaches 7%, it seems that the E_g band of rutile SnO₂ shifts to 498 cm^-1, while the A_1g and B_2g bands shift further to lower values. More interestingly, two new Raman peaks at about 541 cm^-1 (A₁) and 684 cm^-1 (A₂) are also observed in the spectra of the CuO/SnO₂ samples. According to the literature [34-36], the former peak is ascribed to the interface or surface phonon modes, and corresponds to surface defects such as oxygen vacancies and lattice disorder. The latter peak is believed to correspond to the IR-active A2uLO modes. It is apparent that the difference in the CuO loading has altered the surface properties of the SnO₂ support. Particularly, for the sample with CuO loading close to the monolayer dispersion capacity, an abrupt change in the surface property can be observed. The presence of this turning point indicates that the interactions of monolayer CuO and crystalline CuO with SnO₂ support are different, which testify to the presence of a threshold effect on the surface composition. Indeed, the Raman results are highly consistent with the XRD and XPS results, proving that CuO exhibits a monolayer dispersion capacity of 2.09 mmol 100 m^-2, which equals a CuO loading of around 4.8%.

Soot combustion was used to investigate the reaction performances of the CuO/SnO₂ catalysts, with the TPO profiles depicted in Fig. 7. For easy comparison, the T_i, T_p, and △T_i-p, which are the combustion ignition temperature, peak temperature, and the gap between the two temperatures, respectively, are listed in Table 2. For the thermal combustion of pure soot without any catalyst, the T_i and T_p are 530 and 630 ℃, respectively, and the △T_i-p is as wide as 100 ℃. With the incorporation of CVD SnO₂ as the catalyst, the T_i and T_p drop slightly to 525 and 605 ℃, respectively, and the △T_i-p becomes 20 ℃ narrower, indicating that the presence of SnO₂ accelerates the combustion process. Interestingly, with the addition of even 1% CuO, the T_i and T_p drop significantly to 505 and 560 ℃, respectively, and the △T_i-p becomes 55 ℃, which strongly suggest that the presence of dispersed surface CuO has a tremendous influence on the activity. By increasing the CuO loading to 5%, which is close to the monolayer loading of 4.8%, the T_i and T_p drop further to 465 and 515 ℃, respectively, and the △T_i-p becomes 50 ℃. However, upon further increasing the CuO loading to 7% or higher, no significant impact on the reaction performance of the catalyst is observed. Starting from 5% CuO/SnO₂, the sample with a loading close to the monolayer dispersion capacity, all the catalysts with higher CuO loadings exhibit nearly the same activity relative to each other, which proves that beyond the monolayer capacity, increasing the CuO loading does not have a beneficial effect on the soot combustion activity. Apparently, this substantiates the fact that the CuO monolayer on SnO₂ support plays a crucial role in determining the reactivity of the catalyst. In addition, the presence of the turning point suggests a strong threshold effect for the CuO/SnO₂ catalysts for the soot combustion activity.

Soot combustion over CuO/SnO₂ catalysts should follow the Mars-van Krevlen mechanism, which involves the consumption and regeneration of surface-active oxygen species [36]. With the addition of CuO, active surface oxygen species is formed, as evidenced by the Raman results, which is possibly related to the formation of Cu-O-Sn interface bonds. Many earlier studies have demonstrated that for supported metal oxide systems, surface interface bonds can be effectively generated, and that the interfacial oxygen species are active and selective towards many oxidation reactions [37, 38]. In this study, although the Cu-O-Sn interface bonds were not directly observed, the Raman results in Fig. 6 prove the interaction between CuO and SnO₂ support, which induces the formation of surface-active oxygen species. In the following sections, more experimental details will be provided to support this theory.

Below the monolayer dispersion capacity, CuO keeps the SnO₂ support surface covered, thus producing more Cu-O-Sn interface bonds and improving the activity continuously. On the other hand, above the monolayer dispersion capacity, the excess CuO starts to form micro-crystallites, which remain on top of the CuO monolayer. Since crystalline CuO itself is very active towards oxidation reactions [19, 20, 27, 39], the activities of the catalysts with CuO loadings above the monolayer capacity are not degraded by the formation of CuO crystallites, but rather retained without any evident change.

The stability of a soot combustion catalyst determines its potential for application. Therefore, the typical 5% CuO/SnO₂ catalyst was subjected to five continuous TPO tests. As illustrated in Fig. 8, no temperature increase is observed after five cycles, which indicates that the catalyst displays stable reaction performance and has potential for application in some real exhaust post-treatment processes.

H₂-TPR experiments were performed to evaluate the redox properties of the CuO/SnO₂ catalysts, and the profiles are presented in Fig. 9. Pure SnO₂ shows a major reduction peak at 660 ℃, which is assigned to the reduction of bulk SnO₂ to metallic tin [24, 40]; this is also verified by the quantification results shown in Table 3. For all the CuO/SnO₂ samples, besides the major reduction peak of SnO₂, a group of low-temperature reduction peaks assigned to CuO reduction are observed [27], which is confirmed by the quantified O/Cu atomic ratios listed in Table 3. For the reduction of bare CuO, two peaks are typically observed in the range 200 to 400 ℃, which are attributed to the stepwise reduction of CuO to Cu₂O and eventually to metallic copper [24]. It is apparent that in all the CuO/SnO₂ samples, the reduction of CuO shifts to a significantly lower temperature region (below 300 ℃), due to the fine dispersion of CuO on the support surface. As demonstrated earlier, when a metal oxide disperses finely on a support surface, its reduction will become thermodynamically favorable [9, 11]. However, even in the sub-monolayer or monolayer dispersed state, two peaks corresponding to the two typical reduction steps of CuO can still be clearly observed. With the increase in the CuO loading, the second peak around 200 ℃ becomes gradually larger, which confirms the formation of crystalline CuO.

What is interesting here is the shift in the SnO₂ reduction peak with the addition of CuO. Below the monolayer dispersion capacity, with the increase in the CuO loading up to 5%, the SnO₂ reduction peak temperature drops from 660 to 625 ℃. However, further increasing the CuO loading to 15% results in the reduction peak temperature remaining at 625 ℃. This trend is highly consistent with the reaction performances of the catalysts. Therefore, it is believed that in the presence of gas-phase oxygen, the lattice oxygen of SnO₂ could also be involved in the soot combustion process.

To further understand the oxygen properties of the CuO/SnO₂ samples, O₂-TPD experiments were performed, with the results shown in Fig. 10. For all the samples, two oxygen desorption peaks are observed around 110 and 580 ℃. For the convenience of discussion, the desorption peak around 110 ℃ is named α peak, whereas the desorption peak around 580 ℃ is named β peak. The α peak is assigned to the desorption of loosely bonded surface oxygens, whereas the β peak is assigned to the desorption of mobile surface lattice oxygens [41]. For clarity of information, the integrated areas of the two peaks of each sample have been quantified and included in Table 4. With the addition of CuO, the amount of loosely bonded surface oxygen increases significantly, proving again the generation of surface-active oxygen species due to the interaction between the dispersed CuO and the SnO₂ support, which is evidenced by the Raman results. Although pure SnO₂ contains a slightly larger amount of surface lattice oxygen than the CuO/SnO₂ samples, its peak is much wider in comparison with the β peak of the CuO/SnO₂ catalysts, indicating the surface lattice oxygen of pure SnO₂ is kinetically much less active than that of CuO/SnO₂. Furthermore, with the increase in the CuO loading from 1% to 7%, the β peak temperature decreases, indicating that when the CuO loading is close to the monolayer dispersion capacity, the most active surface lattice oxygen species can be formed. Upon further increasing the CuO loading, the β peak temperature increases slightly due to the formation of surface CuO micro-crystallites. It is noted that the total oxygen desorption amounts of the CuO/SnO₂ catalysts are obviously higher than that of pure SnO₂ support. Indeed, the presence of highly abundant and reactive surface oxygen species on CuO/SnO₂ could account for its much-improved soot combustion activity in comparison with that of pure SnO₂.

To verify what is observed through O₂-TPD experiments, some of the typical CuO/SnO₂ samples were also analyzed by XPS, with attention particularly paid to the surface oxygen property. As shown in Fig. 11, doublet O 1s peaks are observed for all the samples, suggesting the presence of two types of oxygen species with different chemical environments on the catalyst surfaces. According to the literature [42], the peak around 530.6 eV is assigned to the surface lattice oxygen species, and the peak around 531.9 eV can be ascribed to the loosely bounded surface oxygen species. For clarity, the percentages of the surface loosely bounded oxygen species, O_ads/(O_ads+O_latt), are quantified and listed in Table 5. With the addition of 1% CuO, the percentage increases from 22.9% to 28.1%, testifying again that the interaction between the surface sub-monolayer CuO and the SnO₂ support creates more abundant surface-active oxygen species. With the increase in the CuO loading to 5%, the percentage improves to 35.9%. However, further improving the CuO loading has little influence on the percentages of the loosely bounded surface oxygen species, which is in good accordance with the changes in the soot combustion activities of the catalysts.

To determine the relationship between the soot combustion activity and the amount of loosely bounded surface-active oxygen species, the T_i and T_p for soot combustion over the CuO/SnO₂ samples are plotted against their O_ads/(O_ads+O_latt) percentages in Fig. 11(B). Interestingly, below 5% CuO loading, with the increase in the O_ads/(O_ads +O_latt) percentages, both the T_i and T_p of the catalysts follow the same trend in that they decrease. When the CuO loading is above 5% and the O_ads/(O_ads+O_latt) percentages of all the samples become constant, the T_i and T_p of the catalysts also remain unchanged. Therefore, it is rational to propose that the amount of the loosely bounded surface oxygen sites might play a critical role in determining the soot combustion activities of the catalysts.

To examine more directly the interaction of soot particles with the surface-active sites of the CuO/SnO₂ catalysts, they were subjected to soot-TPR tests, and the results are displayed in Fig. 12 and Table 6. The pure SnO₂ support without CuO displays two soot reduction peaks at 200 and 370 ℃, testifying the presence of two kinds of surface sites that can react with soot particulates [36]. With the addition of CuO, both the reduction peaks become larger (Table 6), and the high-temperature peak shifts downward to 340 ℃, suggesting the formation of a large amount of highly active surface oxygen species due to the interaction between the dispersed surface CuO and the SnO₂ support. Our previous results have demonstrated that for bare CuO prepared by using different copper precursors and precipitants, four types of oxygen sites that can react with soot particles are present [27]. However, by dispersing CuO onto SnO₂ support, even with the formation of CuO micro-crystallites above the monolayer dispersion capacity, only two kinds of surface-active oxygen sites are observed in the lower-temperature region, as evidenced by the soot-TPR profiles of the CuO/SnO₂ catalysts. Apparently, owing to the interaction between CuO and SnO₂, the property of the active sites has been changed. In brief, the soot-TPR results are in agreement with the Raman, O₂-TPD, and XPS results, indicating that by dispersing CuO onto SnO₂ support, more abundant and active oxygen species can be formed that benefits the soot combustion activities of the catalysts.

With the objective to design and prepare more applicable catalysts for soot combustion, a series of CuO/SnO₂ catalysts with different CuO loadings have been prepared by impregnation method to investigate the structure-reactivity relationship, which was also characterized by different means.

(1) By using XRD and XPS extrapolation methods, it was discovered that CuO disperses finely on the SnO₂ support surface to form a monolayer with a capacity of 2.09 mmol 100 m^-2, which equals to 4.8 wt% CuO loading. As testified by HR-TEM, STEM mapping, and H₂-TPR results, when the CuO loading is below the monolayer dispersion capacity, CuO is present in a sub-monolayer amorphous state. On the other hand, when the loading is above the monolayer dispersion capacity, CuO micro-crystallites are formed, which coexist with the CuO monolayer.

(2) The soot combustion activity of the catalyst increases with CuO loading until it reaches the monolayer dispersion capacity. Further increase in the CuO loading has no evident influence on the activity. Therefore, an apparent monolayer dispersion threshold effect is observed for soot combustion over CuO/SnO₂ catalysts.

(3) Raman results have testified that with the addition of CuO onto SnO₂ support, surface-active oxygen sites can be formed. O₂-TPD and XPS results have proved that the amount of the surface oxygen species increases significantly with the increase in the CuO loading until it reaches the monolayer capacity. Further increase in the CuO loading has no evident impact on this amount. This is highly consistent with the activities of the catalysts. Therefore, the amount of surface-active oxygen sites is believed to be the main factor influencing the activities of the catalysts.

Acknowledgments

This work is supported by the National Natural Science Foundation of China (21567016, 21666020), the Natural Science Foundation of Jiangxi Province (20181ACB20005, 20171BAB213013, 20181BCD40004, 20181BAB203017), the Innovation Fund Designated for Graduate Students of Jiangxi Province (YC2018-B015), the Education Department Foundation of Jiangxi Province (KJLD14005), and the Opening Fund of Key Laboratory of Process Analysis and Control of Sichuan Universities (2017002), which are greatly acknowledged by the authors.