CeO2-Promoted PtSn/SiO2 as a High-Performance Catalyst for the Oxidative Dehydrogenation of Propane with Carbon Dioxide

The oxidative dehydrogenation of propane with CO2 (CO2-ODP) has been extensively investigated as a promising green technology for the efficient production of propylene, but the lack of a high-performance catalyst is still one of the main challenges for its industrial application. In this work, an efficient catalyst for CO2-ODP was developed by adding CeO2 to PtSn/SiO2 as a promoter via the simple impregnation method. Reaction results indicate that the addition of CeO2 significantly improved the catalytic activity and propylene selectivity of the PtSn/SiO2 catalyst, and the highest space-time yield of 1.75 g(C3H6)·g(catalyst)−1·h−1 was achieved over PtSn/SiO2 with a Ce loading of 6 wt%. The correlation of the reaction results with the characterization data reveals that the introduction of CeO2 into PtSn/SiO2 not only improved the Pt dispersion but also regulated the interaction between Pt and Sn species. Thus, the essential reason for the promotional effect of CeO2 on CO2-ODP performance was rationally ascribed to the enhanced adsorption of propane and CO2 originating from the rich oxygen defects of CeO2. These important understandings are applicable in further screening of promoters for the development of a high-performance Pt-based catalyst for CO2-ODP.


Introduction
Propylene is one of the most important raw materials for the chemical industry [1,2]. It is mainly produced by the steam cracking of naphtha and the byproduct of fluid catalytic cracking (FCC) of heavier oil fractions, which suffer from both low propylene yield and high energy consumption [3]. Moreover, these technologies cannot meet the continuously increased market demand for propylene. As a consequence and owing to the growing alternative supply of propane from the shale gas, the catalytically direct dehydrogenation of propane to propylene (PDH) has been attracting increased attention. However, the currently industrialized PDH process is still challenged by the quick deactivation of commercial PtSn/Al 2 O 3 or CrO x /Al 2 O 3 catalysts, low yields of propylene limited by thermodynamics, and high reaction temperatures [4,5]. To address these issues, the oxidative dehydrogenation of propane to propylene using O 2 , CO 2 , N 2 O, or SO x as an oxidant is proposed as a more efficient route [6,7]. Among these alternative processes, the oxidative dehydrogenation of propane with greenhouse gas of CO 2 (CO 2 -ODP) is the most attractive from an environmental viewpoint. On the one hand, in comparison with PDH, CO 2 -ODP can effectively enhance the equilibrium conversion of propane by removing the produced hydrogen [4,8]. On the other hand, a higher selectivity of propylene for CO 2 -ODP can be achieved in comparison with that achieved by using O 2 as an oxidant, for which deep oxidation is an important issue. Moreover, the CO 2 -ODP process provides an attractively tandem approach via efficient production of propylene with simultaneous conversion of CO 2 to 2.2. Catalyst Characterizations N 2 physical adsorption/desorption isotherms were measured on a Bel-sorp-Max instrument at −196 • C. Before each experiment, the sample was degassed at 300 • C under vacuum for 10 h. Specific surface area and pore-size distribution (PSD) were calculated by the Brunauer-Emmett-Teller equation (BET) and Barrett-Joyner-Halenda method (BJH), respectively.
X-ray diffraction (XRD) patterns were recorded on an X-ray diffractometer (Bruker D8 Advance) equipped with Cu-Kα radiation (40 kV, 40 mA). The sample was scanned from the 2θ of 10 to 80 • with a rate of 0.2 s/step.
Transmission electron microscopy (TEM) images were obtained with a high-resolution transmission electron microscope (Tecnai G2 F20, FEI) operated at 200 kV. Before measurement, the fresh sample was pre-reduced at 500 • C in 10 vol% H 2 /Ar for 1 h. Then, about 2 mg of the reduced sample was ultrasonically dispersed in anhydrous ethanol (2 mL). After 1 h, two drops of the suspension were deposited on a carbon-enhanced copper grid and dried at 60 • C in air for 0.5 h. H 2 -O 2 titration experiments were carried out on a Micromeritics Autochem 2920 instrument to determine the dispersion of Pt. For each test, a total of 150 mg of the sample was pre-reduced at 500 • C for 1 h in 10 vol% H 2 /Ar (30 mL/min). After cooling to 50 • C in Ar, a flow of 3 vol% O 2 /Ar (30 mL/min) was pulsed until the consumption peaks became stable. Subsequently, the sample was purged under an Ar flow for 1 h, and consecutive pulses of 10 vol% H 2 /Ar (0.50 mL) were performed. By assuming that the adsorption stoichiometry factor of Pt/H 2 equals to 2/3, according to references [26,27], the dispersion of Pt was calculated using the following Equation (1).
where V H 2 is the volume of adsorbed H 2 (mL), f is the stoichiometry factor, M Pt is the atomic weight of Pt (g/mol), and W Pt is the weight of the supported Pt on the sample (g). Experiments concerning temperature-programmed reduction of H 2 (H 2 -TPR) were carried out on a Micromeritics Autochem 2920 instrument. About 100 mg of the sample was pre-treated at 350 • C for 0.5 h under an Ar stream. After cooling to 50 • C, H 2 -TPR was performed from 50 to 800 • C at a heating rate of 10 • C/min under a 10 vol% H 2 /Ar flow (30 mL/min). H 2 consumption was monitored and determined by a pre-calibrated thermal conductivity detector (TCD).
X-ray photoelectron spectroscopy (XPS) was conducted on an X-ray photoelectron spectrometer (KRATOS Analytical Ltd., Manchester, UK) equipped with an Al-Kα radiation source (1486.6 eV). Before measurements, all the samples were pre-reduced at 500 • C for 1 h in 10 vol% H 2 /Ar. The C 1s spectrum at 284.6 eV was applied to calibrate the binding energy.
Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) of adsorbed CO (CO-DRIFTS) was carried out on a Nicolet iS50 instrument (Thermo Scientific) equipped with an in situ cell. Firstly, the sample was reduced in situ at 500 • C for 1 h in 10 vol% H 2 /Ar with a flow rate of 30 mL/min. After this, the sample was cooled to 30 • C and purged with Ar. Then, 10 vol% CO/Ar with a flow rate of 30 mL/min was introduced in the cell for 0.5 h. Afterwards, the sample was purged with Ar to remove any physically adsorbed CO on the surface of sample, and DRIFTS spectra were recorded. Prior to each experiment, the background spectra were recorded.
Experiments concerning temperature-programmed desorption of C 3 H 8 /CO 2 /C 3 H 6 (C 3 H 8 /CO 2 /C 3 H 6 -TPD) were performed on a Micromeritics Autochem 2920 instrument. About 100 mg of the sample was pre-reduced at 500 • C for 1 h in 10 vol% H 2 /Ar. After this, the sample was cooled to 70 • C and purged with Ar. Then, the pre-treated sample was saturated with pure C 3 H 8 , CO 2 , or C 3 H 6 with a flow rate of 30 mL/min for 1 h. Afterwards, the sample was purged by an Ar stream for 1 h, and temperature-programmed desorption of C 3 H 8 /CO 2 /C 3 H 6 was performed from 70 to 600 • C at a heating rate of 10 • C/min, respectively. The amount of desorbed C 3 H 8 /CO 2 /C 3 H 6 was monitored and determined by a pre-calibrated thermal conductivity detector (TCD).
Thermogravimetric and differential scanning calorimetry analyses (TG-DSC) of the spent catalysts were carried out on a Q600SDT Thermoanalyzer System (TA Instruments). For each test, about 5 mg of the spent catalyst was heated from 50 to 800 • C with a heating ramp of 10 • C/min −1 in a flow of air.
Raman spectra were obtained on a confocal microprobe laser Raman spectrometer (HORIBA Jobin Yvon) with an excitation laser beam of 532 nm. Spectra in the range of 1000-2000 cm −1 were recorded at room temperature to study the type of deposited coke over the spent catalysts.

Catalytic Tests
Catalytic tests for CO 2 -ODP were carried out in a quartz fixed-bed reactor (6 mm, i.d.) under 550 • C and atmospheric pressure. For each test, 0.25 g of the catalyst (40-60 mesh) diluted with 0.5 g of quartz sand (40-60 mesh) was loaded into the reactor. Firstly, the catalyst was pre-reduced with 20 vol% H 2 /He at 500 • C for 1h. After that, the reactor was heated to 550 • C in a He flow, and the gas mixture of Ar/C 3 H 8 /CO 2 /He in a molar ratio of 1/4/4/16 with a total flow rate of 50 mL/min was introduced into the reactor. The products were analyzed by an online gas chromatograph (GC7920, Peking CEAULIGHT) equipped with FID (Porapak Q column) and TCD (TDX-01) detectors. By using Ar as an internal standard, propane and CO 2 conversion, selectivity of different gas products (CH 4 , C 2 H 4 , C 2 H 6 , and C 3 H 6 ), and propylene yield were calculated by Equations (2)- (5).
where, F C 3 H 8 , F CO 2 , and F CO are the volumetric flow rate (mL/min) of C 3 H 8 , CO 2 , and CO, respectively; i stands for the detected hydrocarbon product, i.e., CH 4 , C 2 H 6 , C 2 H 4 , and C 3 H 6 ; and F i and n i represent the flow rate and carbon number of the hydrocarbon product, respectively.

Catalytic Results of CO 2 -ODP
The time-on-stream (TOS) results of CO 2 -ODP over PtSn/SiO 2 , PtCe/SiO 2 , and PtSnCe/SiO 2 are shown in Figure 1. Indeed, PtSn/SiO 2 showed a very low propane conversion of 4.4% at a TOS of 5 min (Figure 1a). In contrast, a significantly higher propane conversion was achieved over the Ce-containing catalysts, and PtSnCe/SiO 2 showed the highest initial propane conversion of 55.8%. Taking the activity as the propane conversion at a TOS of 5 min, an increased order of PtSn/SiO 2 < PtCe/SiO 2 << PtSnCe/SiO 2 was observed. Moreover, the blank experimental results of CO 2 -ODP indicate a propane conversion of less than 2.5% over SnCe/SiO 2 ( Figure S2). Thus, the Pt species over the catalysts are responsible for converting propane, the activity of which is associated with the added SnO 2 and/or CeO 2 . Where the initial activity indexed by the CO 2 conversion at a TOS of 5 min (Figure 1b) was concerned, PtSn/SiO 2 showed negligible activity. In contrast, a significantly high initial CO 2 conversion of 26.4% over PtCe/SiO 2 and 25.9% over PtSnCe/SiO 2 was achieved. In the case of stability, Figure 1a,b clearly shows that PtSnCe/SiO 2 was the most stabile catalyst at a TOS of 80 min, leading to the concurrently decreased conversions of propane and CO 2 in the same order of PtSnCe/SiO 2 > PtCe/SiO 2 >> PtSn/SiO 2 with increasing TOS.
Concerning product distribution, propylene selectivity was varied to a relatively large extent over these catalysts (Figure 1c). In the case of PtSn/SiO 2 , the lowest propylene selectivity of 30.9% was observed at a TOS of 5 min. On the contrary, a higher propylene selectivity was achieved over the Ce-containing catalysts, and the highest initial propylene selectivity of 89.1% was reached over PtSnCe/SiO 2 . With increasing TOS, propylene selectivity at the end of the reaction was decreased in the order of PtSnCe/SiO 2 (93.4%) >> PtCe/SiO 2 (45.0%) > PtSn/SiO 2 (27.4%). To understand the side reactions, the selectivity of gaseous byproducts originating from C 3 H 8 were examined, and the results are shown in Figure 1d and Figure S3. In the case of PtSnCe/SiO 2 and PtCe/SiO 2 , CO was the predominant byproduct, while the selectivity of C 2 H 6 , C 2 H 4 , and CH 4 was very low. This indicates that CO 2 -RP may be the main side reaction over these catalysts in comparison with cracking [13,28]. The significantly higher CO selectivity of 56.9% at a TOS of 5 min was observed over PtCe/SiO 2 , indicating the more favorable breaking of C-C bonds in propane. In contrast, PtSnCe/SiO 2 showed a very low CO selectivity of about 10%, coinciding with the significantly high propylene selectivity. Concerning PtSn/SiO 2 , although the calculated selectivity of CO was 53.9% at a TOS of 5 min, supporting the occurrence of CO 2 -RP, its error may be large due to the very low CO 2 conversion (Figure 1b). Thus, the CO selectivity of PtSn/SiO 2 is not further discussed with that of PtSnCe/SiO 2 and PtCe/SiO 2 .
SnO2 and/or CeO2. Where the initial activity indexed by the CO2 conversion at a TOS of 5 min (Figure 1b) was concerned, PtSn/SiO2 showed negligible activity. In contrast, a significantly high initial CO2 conversion of 26.4% over PtCe/SiO2 and 25.9% over PtSnCe/SiO2 was achieved. In the case of stability, Figure 1a,b clearly shows that PtSnCe/SiO2 was the most stabile catalyst at a TOS of 80 min, leading to the concurrently decreased conversions of propane and CO2 in the same order of PtSnCe/SiO2 > PtCe/SiO2 >> PtSn/SiO2 with increasing TOS. Concerning product distribution, propylene selectivity was varied to a relatively large extent over these catalysts (Figure 1c). In the case of PtSn/SiO2, the lowest propylene selectivity of 30.9% was observed at a TOS of 5 min. On the contrary, a higher propylene selectivity was achieved over the Ce-containing catalysts, and the highest initial propylene selectivity of 89.1% was reached over PtSnCe/SiO2. With increasing TOS, propylene selectivity at the end of the reaction was decreased in the order of PtSnCe/SiO2 (93.4%) >> PtCe/SiO2 (45.0%) > PtSn/SiO2 (27.4%). To understand the side reactions, the selectivity of gaseous byproducts originating from C3H8 were examined, and the results are shown in Figures 1d and S3. In the case of PtSnCe/SiO2 and PtCe/SiO2, CO was the predominant byproduct, while the selectivity of C2H6, C2H4, and CH4 was very low. This indicates that CO2-RP may be the main side reaction over these catalysts in comparison with cracking [13,28]. The significantly higher CO selectivity of 56.9% at a TOS of 5 min was observed over PtCe/SiO2, indicating the more favorable breaking of C-C bonds in propane. In contrast, PtSnCe/SiO2 showed a very low CO selectivity of about 10%, coinciding with the significantly high propylene selectivity. Concerning PtSn/SiO2, although the calculated selectivity of CO was 53.9% at a TOS of 5 min, supporting the occurrence of CO2-RP, its error may be large due to the very low CO2 conversion (Figure 1b). Thus, the CO selectivity of PtSn/SiO2 is not further discussed with that of PtSnCe/SiO2 and PtCe/SiO2.
To show the superior performance of the PtSnCe/SiO2 catalyst for CO2-ODP, the cal- To show the superior performance of the PtSnCe/SiO 2 catalyst for CO 2 -ODP, the calculated space-time yield of propylene (STY C3H6 ) over PtSnCe/SiO 2 and those over different types of catalysts with the best performance from the representative literature are summarized in Table S1. Among the reported catalysts, including CrO x , VO x , GaO x , Pd, and Pt, the highest STY C3H6 of 0.63 g(C 3 H 6 )·g(catalyst) −1 ·h −1 was observed at a reaction temperature of 600 • C over CrO x -doped mesoporous silica spheres (7.07Cr/MSS-2). In our case, however, the PtSnCe/SiO 2 catalyst showed an initial STY C3H6 as high as 1.75 g(C 3 H 6 )·g(catalyst) −1 ·h −1 at a reaction temperature of 550 • C, which is significantly higher than those over the reported catalysts. Moreover, 1.16 g(C 3 H 6 )·g(catalyst) −1 ·h −1 was still achieved, even at a TOS of 6 h, indicating a superior catalytic stability. The durability of PtSnCe/SiO 2 was further investigated by the reaction/regeneration cycles, the regeneration of which is performed at 500 • C in an air flow for 30 min. As shown in Figure 1e, the CO 2 -ODP performance of the catalyst regenerated for two times was very similar to that of the fresh catalyst, indicating the good durability of PtSnCe/SiO 2 . Thus, the catalytic results clearly reflect the superiority of the PtSnCe/SiO 2 catalyst for CO 2 -ODP.

Textural and Structural Properties
The N 2 adsorption-desorption isotherms of PtSn/SiO 2 , PtCe/SiO 2 , and PtSnCe/SiO 2 are shown in Figure S4a. According to the IUPAC classification, all of the catalysts exhibited a similar type-IV isotherm, indicating the presence of mesopores. Moreover, the appearance of an H1-type hysteresis loop over these catalysts occurred at p/p 0 = 0.4-0.6 characterizing the uniform spherical pores. These observations were more directly reflected from the PSD patterns determined by the BJH method. As indicated by Figure S4b, a very narrow and sharp PSD peak at about 3 nm was observed for all of these catalysts. From the calculated textural parameters summarized in Table 1, the BET specific surface area was slightly decreased from 568.9 to 527.3 m 2 /g in the order of PtSn/SiO 2 > PtCe/SiO 2 > PtSnCe/SiO 2 . In the cases of mean pore size and total pore volume, the changes were also very limited. These results suggest very similar textural properties of the samples due to the minimal loadings of Pt, Sn, and/or Ce species over the same silica support. XRD patterns of the catalysts are shown in Figure 2. All of the catalysts exhibited a broad XRD peak at~22.6 • , corresponding to the amorphous nature of the SiO 2 support [29]. In the case of PtSn/SiO 2 , the characteristic diffractions at 2θ of 39.8, 46.2, and 67.4 • were clearly observed, which were assigned to (111), (200), and (220) crystal planes of the cubic Pt metal, respectively [30,31]. In contrast, when CeO 2 was present, the XRD peaks ascribed to Pt metal disappeared, and only the characteristic diffractions at 2θ of 28.5, 33.1, 47.5, and 56.3 • conclusively attributed to the (111), (200), (220), and (311) crystal planes of the cubic fluorite structure of CeO 2 , respectively [22,32], were clearly observed over PtCe/SiO 2 and PtSnCe/SiO 2 . This indicates that the addition of Ce can significantly improve the dispersion of Pt metal, and the SnO 2 species is present as the amorphous or highly dispersed nature. To directly observe the Pt particles, the catalysts pre-reduced in 10 vol% H2/Ar at 500 °C for 1 h were investigated by TEM analysis. As shown in Figure 3a, Pt particles were clearly observed over all of the catalysts. PtSn/SiO2 showed the largest Pt particles, while the significantly smaller Pt particles were present over PtCe/SiO2 and PtSnCe/SiO2, which supports the XRD results. To make a quantitative comparison, statistics analysis was performed; the Pt particle-size distribution histograms are given in Figure 3b, and the average To directly observe the Pt particles, the catalysts pre-reduced in 10 vol% H 2 /Ar at 500 • C for 1 h were investigated by TEM analysis. As shown in Figure 3a, Pt particles were clearly observed over all of the catalysts. PtSn/SiO 2 showed the largest Pt particles, while the significantly smaller Pt particles were present over PtCe/SiO 2 and PtSnCe/SiO 2 , which supports the XRD results. To make a quantitative comparison, statistics analysis was performed; the Pt particle-size distribution histograms are given in Figure 3b, and the average diameter of Pt is summarized in Table 1. The metallic Pt size was continuously decreased from 8.5 ± 0.2 to 2.1 ± 0.3 nm in the order of PtSn/SiO 2 >> PtCe/SiO 2 > PtSnCe/SiO 2 , indicating that the addition of CeO 2 can significantly improve the dispersion of Pt metal. This is further supported by the H 2 -O 2 titration results, in which the calculated Pt dispersion was continuously increased from 13.4 to 41.3% in the order of PtSn/SiO 2 < PtCe/SiO 2 < PtSnCe/SiO 2 (Table 1). Generally, the addition of SnO 2 can effectively improve the Pt dispersion. However, this is contradictory to the observations over PtSn/SiO 2 , in which a Pt size as high as 8.5 ± 0.2 nm was observed. As for the reason, it has been reported that PtSn/SiO 2 catalysts directly calcined in an oxidative atmosphere during the preparation process show poor Pt dispersion [33] due to the weak interaction between Pt and SnO 2 on the SiO 2 support. This is in agreement with our experimental results, in which all of the catalysts were obtained by calcining in air at 500 • C after impregnation, as described in Section 2.1. In contrast, when CeO 2 was introduced, the Pt size was significantly decreased over PtCe/SiO 2 and PtSnCe/SiO 2 . Taking these results into account, the lower Pt size of 2.1 nm for PtSnCe/SiO 2 , compared to that of PtCe/SiO 2 , indicates that in addition to increasing the Pt dispersion, the presence of CeO 2 can improve the interaction between Pt and SnO 2 , leading to the highest Pt dispersion over PtSnCe/SiO 2 . To directly observe the Pt particles, the catalysts pre-reduced in 10 vol% H2/Ar at 500 °C for 1 h were investigated by TEM analysis. As shown in Figure 3a, Pt particles were clearly observed over all of the catalysts. PtSn/SiO2 showed the largest Pt particles, while the significantly smaller Pt particles were present over PtCe/SiO2 and PtSnCe/SiO2, which supports the XRD results. To make a quantitative comparison, statistics analysis was performed; the Pt particle-size distribution histograms are given in Figure 3b, and the average diameter of Pt is summarized in Table 1. The metallic Pt size was continuously decreased from 8.5 ± 0.2 to 2.1 ± 0.3 nm in the order of PtSn/SiO2 >> PtCe/SiO2 > PtSnCe/SiO2, indicating that the addition of CeO2 can significantly improve the dispersion of Pt metal. This is further supported by the H2-O2 titration results, in which the calculated Pt dispersion was continuously increased from 13.4 to 41.3% in the order of PtSn/SiO2 < PtCe/SiO2 < PtSnCe/SiO2 (Table 1). Generally, the addition of SnO2 can effectively improve the Pt dispersion; however this, is contradictory to the observations over PtSn/SiO2, in which a Pt size as high as 8.5 ± 0.2 nm was observed. As for the reason, it has been reported that PtSn/SiO2 catalysts directly calcined in an oxidative atmosphere during the preparation process show poor Pt dispersion [33] due to the weak interaction between Pt and SnO2 on the SiO2 support. This is in agreement with our experimental results, in which all of the catalysts were obtained by calcining in air at 500 °C after impregnation, as described in Section 2.1. In contrast, when CeO2 was introduced, the Pt size was significantly decreased over PtCe/SiO2 and PtSnCe/SiO2. Taking these results into account, the lower Pt size of 2.1 nm for PtSnCe/SiO2, compared to that of PtCe/SiO2, indicates that in addition to increasing the Pt dispersion, the presence of CeO2 can improve the interaction between Pt and SnO2, leading to the highest Pt dispersion over PtSnCe/SiO2.

Reduction Behavior
The redox properties of the catalysts were analyzed by H 2 -TPR. As shown in Figure 4, for the PtSn/SiO 2 catalyst, a weak and broad reduction peak was observed at~450 • C, which can be can be attributed to the reduction of SnO 2 [34,35]. In the case of PtCe/SiO 2 , two reduction peaks were observed at 279 • C and 733 • C. The first peak at 279 • C was attributed to the reduction of active oxygen species over the CeO 2 surface, while the very weak peak at 733 • C corresponded to the reduction of the lattice oxygen over the bulk CeO 2 [22,36]. Concerning PtSnCe/SiO 2 , only one reduction peak was observed at 233 • C, attributed to the reduction of surface oxygen species. The reduction-peak temperature below 600 • C was continuously decreased in the order of PtSn/SiO 2 >> PtCe/SiO 2 > PtSnCe/SiO 2 , while the amount of the H 2 consumption calculated from the reduction peak during H 2 -TPR below 600 • C ( Table 2) was increased in the order of PtSn/SiO 2 << PtCe/SiO 2 < PtSnCe/SiO 2 , leading to the greatest reducibility of PtSnCe/SiO 2 . It has been reported that the hydrogen spillover effect induced from the interaction between the Pt and SnO 2 /CeO 2 can accelerate the reduction of oxygen over the catalyst, and a higher dispersion of Pt commonly leads to a greater reducibility [37,38]. Based on this explana-tion, the H 2 -TPR results were well understood when the dispersion of Pt, as discussed in Section 3.2, was taken into account. As a result of the highest Pt dispersion over PtSnCe/SiO 2 , the reduction of oxides over the catalysts was enhanced due to the strongest hydrogen spillover effect, leading to the lowest reduction-peak temperature and the greatest amount of H 2 consumption.
attributed to the reduction of surface oxygen species. The reduction-peak temperature be-low 600 °C was continuously decreased in the order of PtSn/SiO2 >> PtCe/SiO2 > PtSnCe/SiO2, while the amount of the H2 consumption calculated from the reduction peak during H2-TPR below 600 °C ( Table 2) was increased in the order of PtSn/SiO2 << PtCe/SiO2 < PtSnCe/SiO2, leading to the greatest reducibility of PtSnCe/SiO2. It has been reported that the hydrogen spillover effect induced from the interaction between the Pt and SnO2/CeO2 can accelerate the reduction of oxygen over the catalyst, and a higher dispersion of Pt commonly leads to a greater reducibility [37,38]. Based on this explanation, the H2-TPR results were well understood when the dispersion of Pt, as discussed in Section 3.2, was taken into account. As a result of the highest Pt dispersion over PtSnCe/SiO2, the reduction of oxides over the catalysts was enhanced due to the strongest hydrogen spillover effect, leading to the lowest reduction-peak temperature and the greatest amount of H2 consumption.

Chemical States
XPS analysis was used to study the surface chemical states of the catalysts. Before measurements, all of the catalysts were pre-reduced in 10 vol% H2/Ar at 500 °C for 1 h. Ce 3d spectra are usually fitted with eight Gaussian-Lorentzian peaks corresponding to two pairs of spin-orbit doubles [32,39]. As shown in Figure S5, the peaks labeled as v, v″, v‴ and u, u″ u‴ were assigned to the ionization of Ce 4+ 3d5/2 and Ce 4+ 3d3/2, respectively, while the peaks marked with vʹ and uʹ were originated from Ce 3+ 3d5/2 and Ce 3+ 3d3/2, respectively. Based on those peak areas, the relative content of Ce 3+ was calculated, defined as the ratio of Ce 3+ /(Ce 3+ + Ce 4+ ). As shown in Table 2, the relative content of Ce 3+ over

Chemical States
XPS analysis was used to study the surface chemical states of the catalysts. Before measurements, all of the catalysts were pre-reduced in 10 vol% H 2 /Ar at 500 • C for 1 h. Ce 3d spectra are usually fitted with eight Gaussian-Lorentzian peaks corresponding to two pairs of spin-orbit doubles [32,39]. As shown in Figure S5, the peaks labeled as v, v , v and u, u u were assigned to the ionization of Ce 4+ 3d 5/2 and Ce 4+ 3d 3/2 , respectively, while the peaks marked with v and u were originated from Ce 3+ 3d 5/2 and Ce 3+ 3d 3/2 , respectively. Based on those peak areas, the relative content of Ce 3+ was calculated, defined as the ratio of Ce 3+ /(Ce 3+ + Ce 4+ ). As shown in Table 2, the relative content of Ce 3+ over PtSnCe/SiO 2 is clearly higher than that of PtCe/SiO 2 , indicating the formation of more oxygen defects after the reduction. This is consistent with the significantly greater amount of H 2 consumption during the H 2 -TPR over the fresh PtSnCe/SiO 2 than that over PtCe/SiO 2 ( Table 2), resulting from the stronger hydrogen spillover effect due to the greater dispersion of Pt (Table 1).
As show in Figure 5a, the binding energies at around 71.4 eV for 4f 7/2 and 74.7 eV for 4f 5/2 were clearly observed over PtSn/SiO 2 , indicating the presence of metallic Pt 0 [40]. In contrast, when CeO 2 was added, the binding energies of Pt 4f were clearly increased, and the XPS peaks at 72.8 eV for 4f 7/2 and 76.1 eV for 4f 5/2 assigned to Pt 2+ species [40] were clearly observed over PtCe/SiO 2 and PtSnCe/SiO 2 . Following the results of deconvolution of the Pt 4f peaks (Figure 5a), the relative content of Pt 0 and Pt 2+ was calculated by respective peak area. As shown in Table 2, PtSn/SiO 2 showed the exclusively metallic Pt species, which is consistent with XRD results (Figure 2). Contrarily, the relative content of Pt 2+ was as high as 69.5% and 66.9% for PtCe/SiO 2 and PtSnCe/SiO 2 , respectively. According to references [41,42], the presence of Pt 2+ over the CeO 2 -containing catalysts originates from the strong interaction between Pt and CeO 2 , which may be the key reason for the greater dispersion of Pt over PtCe/SiO 2 and PtSnCe/SiO 2 than over PtSn/SiO 2 . This is supported by the TEM and H 2 -O 2 titration results ( Figure 3 and Table 1). As shown in Figure 5b, the Sn 3d 5/2 at about 487.0 eV was deconvoluted to analyze the chemical state of Sn. In the case of PtSn/SiO 2 , a symmetric Sn 3d 5/2 XPS peak with a binding energy of 487.2 eV was observed, indicating the presence of only SnO x species, as reported in [28]. However, besides oxide species, a small amount of Sn 0 species located at 485.7 eV was observed over PtSnCe/SiO 2 . As for the reason, it is noteworthy that PtSnCe/SiO 2 showed a slightly higher content of Pt 0 than PtCe/SiO 2 (Table 2), which may originate from the improved interaction between SnO 2 and Pt due to the presence of CeO 2 . This coincides well with the presence of Sn 0 species, indicating the possible formation of Pt-Sn bimetallic nanoparticles [28,43].
of the Pt 4f peaks (Figure 5a), the relative content of Pt and Pt was calculated by respective peak area. As shown in Table 2, PtSn/SiO2 showed the exclusively metallic Pt species, which is consistent with XRD results (Figure 2). Contrarily, the relative content of Pt 2+ was as high as 69.5% and 66.9% for PtCe/SiO2 and PtSnCe/SiO2, respectively. According to references [41,42], the presence of Pt 2+ over the CeO2-containing catalysts originates from the strong interaction between Pt and CeO2, which may be the key reason for the greater dispersion of Pt over PtCe/SiO2 and PtSnCe/SiO2 than over PtSn/SiO2. This is supported by the TEM and H2-O2 titration results ( Figure 3 and Table 1). As shown in Figure 5b, the Sn 3d5/2 at about 487.0 eV was deconvoluted to analyze the chemical state of Sn. In the case of PtSn/SiO2, a symmetric Sn 3d5/2 XPS peak with a binding energy of 487.2 eV was observed, indicating the presence of only SnOx species, as reported in [28]. However, besides oxide species, a small amount of Sn 0 species located at 485.7 eV was observed over PtSnCe/SiO2. As for the reason, it is noteworthy that PtSnCe/SiO2 showed a slightly higher content of Pt 0 than PtCe/SiO2 (Table 2), which may originate from the improved interaction between SnO2 and Pt due to the presence of CeO2. This coincides well with the presence of Sn 0 species, indicating the possible formation of Pt-Sn bimetallic nanoparticles [28,43].

CO-DRIFTS Studies
To further investigate the structural and electronic properties of Pt, DRIFTS experiments were performed by using CO as a probing molecule since its adsorption on Pt surfaces has been well studied. As shown in Figure 6, two overlapping bands were clearly observed in the case of PtSn/SiO2 at 2000 and 2024 cm −1 , assigned to Si-H stretching vibrations in the different SiO2 configuration [44]. Moreover, a very weak peak was detected at about 2074 cm −1 , assigned to the linearly bonded CO on Pt 0 terraces, indicating the presence of large, highly coordinated nanoparticles [45,46]. When PtCe/SiO2 and PtSnCe/SiO2 were considered, significantly weakened and even disappeared peaks were observed for the Si-H stretching vibrations, which may be due to the coverage of CeO2 on the SiO2 surface. Moreover, strong band was observed at ~2060 cm −1 , ascribed to the linearly bonded CO on Pt 0 , with intermediate coordination sites, such as edges or steps sites [47],

CO-DRIFTS Studies
To further investigate the structural and electronic properties of Pt, DRIFTS experiments were performed by using CO as a probing molecule since its adsorption on Pt surfaces has been well studied. As shown in Figure 6, two overlapping bands were clearly observed in the case of PtSn/SiO 2 at 2000 and 2024 cm −1 , assigned to Si-H stretching vibrations in the different SiO 2 configuration [44]. Moreover, a very weak peak was detected at about 2074 cm −1 , assigned to the linearly bonded CO on Pt 0 terraces, indicating the presence of large, highly coordinated nanoparticles [45,46]. When PtCe/SiO 2 and PtSnCe/SiO 2 were considered, significantly weakened and even disappeared peaks were observed for the Si-H stretching vibrations, which may be due to the coverage of CeO 2 on the SiO 2 surface. Moreover, strong band was observed at~2060 cm −1 , ascribed to the linearly bonded CO on Pt 0 , with intermediate coordination sites, such as edges or steps sites [47], indicating the high dispersion of Pt over these two Ce-containing catalysts [48,49]. This is supported by the results of TEM and H 2 -O 2 titration (Figure 3 and Table 1). Noteworthy, besides the band at 2060 cm −1 , a weak adsorption band was detected over PtCe/SiO 2 at 1820 cm −1 , ascribed to the bridge-bonded CO on two neighboring Pt atoms [24]. However, it disappeared in the case of PtSnCe/SiO 2 , accompanying a decreased intensity of the linear adsorption peak at 2060 cm −1 . The disappeared bridge-bonded CO over PtSnCe/SiO 2 suggests that the SnO 2 breaks the ensemble of Pt atoms and forms a checkerboard Pt-Sn surface structure [24,50] because CO does not adsorb at the bridge sites between Sn and Pt. The decreased intensity of the peak at 2060 cm −1 , in comparison with PtCe/SiO 2 , can be explained as the reduced surface coverage of CO due to the presence of SnO 2 [51]. These results indicate that the presence of CeO 2 on a PtSn/SiO 2 catalyst can not only improve the Pt dispersion but also improve the interaction between Pt and SnO 2 .
ear adsorption peak at 2060 cm −1 . The disappeared bridge-bonded CO over PtSnCe/SiO2 suggests that the SnO2 breaks the ensemble of Pt atoms and forms a checkerboard Pt-Sn surface structure [24,50] because CO does not adsorb at the bridge sites between Sn and Pt. The decreased intensity of the peak at 2060 cm −1 , in comparison with PtCe/SiO2, can be explained as the reduced surface coverage of CO due to the presence of SnO2 [51]. These results indicate that the presence of CeO2 on a PtSn/SiO2 catalyst can not only improve the Pt dispersion but also improve the interaction between Pt and SnO2.

Key Factors of Catalytic Activity
As indicated by the results in Sections 3.1 and 3.2, the initial C3H8 conversion at a TOS of 5 min was increased in the order of PtSnCe/SiO2 >> PtCe/SiO2 >> PtSn/SiO2, coinciding well with the dispersion of Pt over the catalysts. This indicates that the amount of active Pt species is the key factor determining the activation of propane in the course of CO2-ODP, which is consistent with the reported results for PDH [52]. When the activation of CO2 was considered, the significant conversion of CO2 was only observed over the Cecontaining catalysts of PtSnCe/SiO2 and PtCe/SiO2, while CO2 conversion for PtSn/SiO2 was negligible. This indicates that the introduced CeO2 plays a key role in the activation of CO2, which is supported by our previous work for oxidative dehydrogenation of ethylbenzene with CO2 [53]. To shed some light on these observations, C3H8-and CO2-TPD experiments were performed over the catalysts. In the case of C3H8-TPD (Figure 7a), a very broad curve was observed for all of the catalysts in the temperature range of 100-400 °C, indicating the varied strength of adsorbed propane [54,55]. For PtSn/SiO2, two overlapping peaks were clearly observed at about 118 °C and 236 °C, respectively. When PtCe/SiO2 and PtSnCe/SiO2 were considered, the peak maxima were shifted toward higher temperatures in comparison with those of PtSn/SiO2. Moreover, the peak areas of desorbed propane were significantly increased. This indicates the intensified adsorption of propane over the Ce-containing catalysts. The amount of desorbed propane was calculated below 400 °C during C3H8-TPD, and the results are given in Table 3. It was increased in the order of PtSn/SiO2 << PtCe/SiO2 < PtSnCe/SiO2, which coincides well with the propane conversion. This clearly reveals that the amount of adsorbed propane plays a key role in determining the activity of these catalysts for CO2-ODP, which can be reasonably associated with Pt dispersion. As for the adsorption of CO2, a broad CO2-TPD pattern similar to that of C3H8-TPD was obtained for all of the catalysts (Figure 7b). For PtSn/SiO2, only a small peak was observed at about 126 °C, indicating the very weak adsorption of

Key Factors of Catalytic Activity
As indicated by the results in Sections 3.1 and 3.2, the initial C 3 H 8 conversion at a TOS of 5 min was increased in the order of PtSnCe/SiO 2 >> PtCe/SiO 2 >> PtSn/SiO 2 , coinciding well with the dispersion of Pt over the catalysts. This indicates that the amount of active Pt species is the key factor determining the activation of propane in the course of CO 2 -ODP, which is consistent with the reported results for PDH [52]. When the activation of CO 2 was considered, the significant conversion of CO 2 was only observed over the Ce-containing catalysts of PtSnCe/SiO 2 and PtCe/SiO 2 , while CO 2 conversion for PtSn/SiO 2 was negligible. This indicates that the introduced CeO 2 plays a key role in the activation of CO 2 , which is supported by our previous work for oxidative dehydrogenation of ethylbenzene with CO 2 [53]. To shed some light on these observations, C 3 H 8and CO 2 -TPD experiments were performed over the catalysts. In the case of C 3 H 8 -TPD (Figure 7a), a very broad curve was observed for all of the catalysts in the temperature range of 100-400 • C, indicating the varied strength of adsorbed propane [54,55]. For PtSn/SiO 2 , two overlapping peaks were clearly observed at about 118 • C and 236 • C, respectively. When PtCe/SiO 2 and PtSnCe/SiO 2 were considered, the peak maxima were shifted toward higher temperatures in comparison with those of PtSn/SiO 2 . Moreover, the peak areas of desorbed propane were significantly increased. This indicates the intensified adsorption of propane over the Ce-containing catalysts. The amount of desorbed propane was calculated below 400 • C during C 3 H 8 -TPD, and the results are given in Table 3. It was increased in the order of PtSn/SiO 2 << PtCe/SiO 2 < PtSnCe/SiO 2 , which coincides well with the propane conversion. This clearly reveals that the amount of adsorbed propane plays a key role in determining the activity of these catalysts for CO 2 -ODP, which can be reasonably associated with Pt dispersion. As for the adsorption of CO 2 , a broad CO 2 -TPD pattern similar to that of C 3 H 8 -TPD was obtained for all of the catalysts (Figure 7b). For PtSn/SiO 2 , only a small peak was observed at about 126 • C, indicating the very weak adsorption of CO 2 , which is consistent with references [56,57]. In contrast, the peak temperature of desorbed CO 2 increased to 141 • C over PtCe/SiO 2 and PtSnCe/SiO 2 . Moreover, a shoulder peak could be observed at a higher temperature of 242 • C, which can be explained by the stronger adsorbed CO 2 on the surface of CeO 2 . This indicates the presence of CeO 2 -enhanced CO 2 adsorption. As shown in Table 3, the amount of desorbed CO 2 was calculated below 400 • C during CO 2 -TPD. It was increased in the order of PtSn/SiO 2 << PtCe/SiO 2 < PtSnCe/SiO 2 , the changing pattern of which coincides well with that of CO 2 conversion at the steady state of TOS (Figure 1b). It has been reported that CeO 2 with richer oxygen defects commonly leads to enhanced adsorption and activation of CO 2 [21,32]. Following this understanding, the greater amount of adsorbed CO 2 in the case of PtSnCe/SiO 2 than that of PtCe/SiO 2 can be reasonably ascribed to the presence of more oxygen defects of CeO 2 , as revealed by the Ce 3d XPS results (Table 2). These results indicate that the amount of adsorbed CO 2 plays a key role in determining the activation of CO 2 over the catalysts, which can be connected with the introduced CeO 2 .
below 400 °C during CO2-TPD. It was increased in the order of PtSn/SiO2 << PtCe/SiO2 < PtSnCe/SiO2, the changing pattern of which coincides well with that of CO2 conversion at the steady state of TOS (Figure 1b). It has been reported that CeO2 with richer oxygen defects commonly leads to enhanced adsorption and activation of CO2 [21,32]. Following this understanding, the greater amount of adsorbed CO2 in the case of PtSnCe/SiO2 than that of PtCe/SiO2 can be reasonably ascribed to the presence of more oxygen defects of CeO2, as revealed by the Ce 3d XPS results (Table 2). These results indicate that the amount of adsorbed CO2 plays a key role in determining the activation of CO2 over the catalysts, which can be connected with the introduced CeO2.

Insights into Product Selectivity
As indicated by the results in Section 3.1, the selectivity of propylene varied to a relatively large extent over the PtSn/SiO2, PtCe/SiO2, and PtSnCe/SiO2 catalysts (Figure 1c). According to the analysis of product distribution (Figure 1d), this is explained by the simultaneous occurrence of CO2-RP in the course of CO2-ODP. By correlating the characterization results of Sections 3.2-3.5, the significant propylene selectivity over PtSnCe/SiO2 can be explained as the Ce promoted interaction between Sn and Pt, which favors the breaking of C-H bonds in propane [23,58]. However, in addition to the side reaction of CO2-RP induced from propane, the selectivity of propylene is also determined by its possible secondary reactions, including successive polymerization (coke deposition) and further cracking, owing to the difficult desorption of propylene from the surface of catalysts [59]. For further insight, C3H6-TPD experiments were performed, and the results are given in Figure 8. A very broad desorption signal was observed in the temperature range of 100 to 450 °C for all of the catalysts, indicating the varied strength of propylene adsorption on the surface of the catalyst [54]. In the case of PtSn/SiO2, two overlapping peaks were

Insights into Product Selectivity
As indicated by the results in Section 3.1, the selectivity of propylene varied to a relatively large extent over the PtSn/SiO 2 , PtCe/SiO 2 , and PtSnCe/SiO 2 catalysts (Figure 1c). According to the analysis of product distribution (Figure 1d), this is explained by the simultaneous occurrence of CO 2 -RP in the course of CO 2 -ODP. By correlating the characterization results of Sections 3.2-3.5, the significant propylene selectivity over PtSnCe/SiO 2 can be explained as the Ce promoted interaction between Sn and Pt, which favors the breaking of C-H bonds in propane [23,58]. However, in addition to the side reaction of CO 2 -RP induced from propane, the selectivity of propylene is also determined by its possible secondary reactions, including successive polymerization (coke deposition) and further cracking, owing to the difficult desorption of propylene from the surface of catalysts [59]. For further insight, C 3 H 6 -TPD experiments were performed, and the results are given in Figure 8. A very broad desorption signal was observed in the temperature range of 100 to 450 • C for all of the catalysts, indicating the varied strength of propylene adsorption on the surface of the catalyst [54]. In the case of PtSn/SiO 2 , two overlapping peaks were observed at about 123 • C and 204 • C. In contrast, both the peak temperature and amount of desorbed propylene over PtCe/SiO 2 and PtSnCe/SiO 2 were clearly higher than over PtSn/SiO 2 , indicating a stronger adsorption of propylene. Furthermore, as given in Table 3, the amount of adsorbed propylene was calculated below 450 • C during C 3 H 6 -TPD and was found to increase in the order of PtSn/SiO 2 << PtCe/SiO 2 < PtSnCe/SiO 2 . When propylene selectivity is compared with the amount of propylene adsorption, exactly the same trend is found, i.e., the greater the amount of propylene adsorption, the higher the propylene selectivity. This result is contradictory to the common expectation. Considering the dominant byproduct of CO in the course of CO 2 -ODP (Figure 1d), it can be concluded that the simultaneous occurrence of CO 2 -RP plays a key role in determining the propylene selectivity in comparison with the secondary reaction of propylene. was found to increase in the order of PtSn/SiO2 << PtCe/SiO2 < PtSnCe/SiO2. When propylene selectivity is compared with the amount of propylene adsorption, exactly the same trend is found, i.e., the greater the amount of propylene adsorption, the higher the propylene selectivity. This result is contradictory to the common expectation. Considering the dominant byproduct of CO in the course of CO2-ODP (Figure 1d), it can be concluded that the simultaneous occurrence of CO2-RP plays a key role in determining the propylene selectivity in comparison with the secondary reaction of propylene. As a matter of fact, coke deposition on the surface of catalysts is a common issue in the course of CO2-ODP, the behavior of which is associated with its catalytic performance [4,54]. TG-DSC was performed to analyze the amount and kind of deposited coke over the spent catalysts after a TOS of 2 h, and the results are shown in Figure S6 and Table 3. For all of the catalysts, a clear weight loss was observed at about 30-200 °C, induced from the physical desorption of water ( Figure S6a), accompanying the clearly endothermic peak of DSC curves at around 74 °C ( Figure S6b). With a further increase in temperature from 200 to 800 °C, the TG signal commonly assigned to the burning of deposited coke was almost steady in the case of PtSn/SiO2, indicating a negligible amount of coke formed on the surface. This coincides well with the significantly low propane conversion (Figure 1a). Contrary to this, a weight loss of 1.71% and 2.53% was clearly observed over PtCe/SiO2 and PtSnCe/SiO2, respectively, at about 300-600 °C, ascribed to the burning of coke. This was further revealed by the exothermic peak of DSC curves. Moreover, the peak temperature of DSC for PtSnCe/SiO2 (400 °C) was clearly lower than that for PtCe/SiO2 (452 °C), suggesting a difference in the degree of graphitization of the deposited coke. To further confirm this, visible Raman characterization was performed. As given in Figure S7, typical Raman shifts were observed over PtCe/SiO2 and PtSnCe/SiO2 at 1340 and 1600 cm −1 , assigned to the disordered (D band) and graphitic carbon (G band), respectively. To quantify the extent of graphitization of the deposited coke, the intensity ratio of the D and G bands, i.e., ID/IG, was calculated. As shown in Table 3, PtSnCe/SiO2 showed a higher value of ID/IG (0.81) than PtCe/SiO2 (0.73), indicating a lesser extent of graphitization of coke species [32]. This is in agreement with the DSC results ( Figure S6b). The difference in the species of deposited coke can be explained by the fact that PtCe/SiO2 is favorable to CO2-RP, while PtSnCe/SiO2 is promising for CO2-ODP. The lesser extent of graphitization of coke species on the surface of PtSnCe/SiO2 mainly originated from the polymerization of the produced C3H6. However, PtCe/SiO2 led to the formation of more graphitic carbon species due to the severe breaking of the C-C bond in propane through CO2-RP. The coking rate (g/mol) of PtCe/SiO2 and PtSnCe/SiO2, defined as grams of deposited coke, was As a matter of fact, coke deposition on the surface of catalysts is a common issue in the course of CO 2 -ODP, the behavior of which is associated with its catalytic performance [4,54]. TG-DSC was performed to analyze the amount and kind of deposited coke over the spent catalysts after a TOS of 2 h, and the results are shown in Figure S6 and Table 3. For all of the catalysts, a clear weight loss was observed at about 30-200 • C, induced from the physical desorption of water ( Figure S6a), accompanying the clearly endothermic peak of DSC curves at around 74 • C ( Figure S6b). With a further increase in temperature from 200 to 800 • C, the TG signal commonly assigned to the burning of deposited coke was almost steady in the case of PtSn/SiO 2 , indicating a negligible amount of coke formed on the surface. This coincides well with the significantly low propane conversion (Figure 1a). Contrary to this, a weight loss of 1.71% and 2.53% was clearly observed over PtCe/SiO 2 and PtSnCe/SiO 2 , respectively, at about 300-600 • C, ascribed to the burning of coke. This was further revealed by the exothermic peak of DSC curves. Moreover, the peak temperature of DSC for PtSnCe/SiO 2 (400 • C) was clearly lower than that for PtCe/SiO 2 (452 • C), suggesting a difference in the degree of graphitization of the deposited coke. To further confirm this, visible Raman characterization was performed. As given in Figure S7, typical Raman shifts were observed over PtCe/SiO 2 and PtSnCe/SiO 2 at 1340 and 1600 cm −1 , assigned to the disordered (D band) and graphitic carbon (G band), respectively. To quantify the extent of graphitization of the deposited coke, the intensity ratio of the D and G bands, i.e., I D /I G , was calculated. As shown in Table 3, PtSnCe/SiO 2 showed a higher value of I D /I G (0.81) than PtCe/SiO 2 (0.73), indicating a lesser extent of graphitization of coke species [32]. This is in agreement with the DSC results ( Figure S6b). The difference in the species of deposited coke can be explained by the fact that PtCe/SiO 2 is favorable to CO 2 -RP, while PtSnCe/SiO 2 is promising for CO 2 -ODP. The lesser extent of graphitization of coke species on the surface of PtSnCe/SiO 2 mainly originated from the polymerization of the produced C 3 H 6 . However, PtCe/SiO 2 led to the formation of more graphitic carbon species due to the severe breaking of the C-C bond in propane through CO 2 -RP. The coking rate (g/mol) of PtCe/SiO 2 and PtSnCe/SiO 2 , defined as grams of deposited coke, was calculated following references [54,60] by converting 1 mole of propane after a TOS of 2 h. In the case of PtSnCe/SiO 2 , the coking rate was 0.07 g/mol, which is clearly lower than that of PtCe/SiO 2 (2.09 g/mol). This indicates that coke deposition over PtSnCe/SiO 2 is significantly inhibited, which may result from the lesser extent of graphitization of coke species for CO 2 -ODP.

Conclusions
In summary, a highly efficient CO 2 -ODP catalyst was developed with STY C3H6 as high as 1.75 g(C 3 H 6 )·g(catalyst) − ·h −1 by simply impregnating Ce (6 wt%) into PtSn/SiO 2 . Moreover, CO 2 -ODP performance was essentially restored after the regeneration of the catalyst at 500 • C for 30 min in an air flow. Additionally, the promotional effect of CeO 2 on PtSn/SiO 2 played a key role in determining the initial CO 2 -ODP performance, leading to the same increased order of PtSn/SiO 2 < PtCe/SiO 2 < PtSnCe/SiO 2 for the initial propane conversion of 4.4%, 20.6%, and 55.8% and propylene selectivity of 31.0%, 39.7%, and 89.1%. Physical, chemical, and spectra characterizations reveal that the addition of CeO 2 led to an increased Pt dispersion of 13.4% for PtSn/SiO 2 < 20.9% for PtCe/SiO 2 < 41.3% for PtSnCe/SiO 2 and strong interactions between Pt and Sn species over the PtSnCe/SiO 2 catalyst, which favors the synchronized activation of C-H bonds in propane and the C=O bonds in CO 2 molecules. This was explained as the enhanced adsorption of propane and CO 2 in the order of PtSn/SiO 2 < PtCe/SiO 2 < PtSnCe/SiO 2 , essentially originated from the rich oxygen defects over the added CeO 2 . With these understandings, the modification of catalysts with improved oxygen defects over oxides, as well as the search for promoters with richer oxygen defects than CeO 2 , is expected to produce a more effective Pt-based catalyst for CO 2 -ODP, with additional studies still in progress in our laboratory.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

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