E ﬀ ect of Reduction Atmosphere on Structure and Catalytic Performance of PtIn / Mg(Al)O / ZnO for Propane Dehydrogenation

: The e ﬀ ect of reduction atmospheres, H 2 / N 2 , C 3 H 8 / H 2 / N 2 , C 3 H 8 and CO, on the structure and propane direct dehydrogenation performance of PtIn / Mg(Al)O / ZnO catalyst derived from ZnO-supported PtIn-hydrotalcite was studied. The physicochemical properties of the as-prepared and used catalytic system were characterized by various characterization methods. The results show that the dehydrogenation performance, especially the stability of the PtIn / Mg(Al)O / ZnO catalyst, was signiﬁcantly improved along with the change in reduction atmosphere. The highest catalytic activity (51% of propane conversion and 97% propylene selectivity), resistance toward coke deposition, and stability for more than 30 h were achieved with the H 2 / N 2 -reduced catalyst. The optimal dehydrogenation performance and coke resistance are mainly related to the high Pt dispersion and In 0 / In 3 + molar ratio, strong Pt–In interaction and small metal particle size, depending on the nature of the reduction atmospheres. The reconstruction of meixnerite favors the stability and coke resistance to some extent.


Introduction
In recent years, the dehydrogenation of low alkanes to the corresponding alkenes has drawn great attention from the world, due to their extensive utilization as raw materials and a versatile class of intermediates in the polymer industry. In this context, it is particularly important to dehydrogenate propane to propylene (PDH) because of the rapidly growing demand for propylene in the production of propylene oxide, acrylonitrile and polypropylene [1][2][3]. The preparation and structure studies of catalysts with high efficiency propylene yield become the key point of PDH.
The support materials, Al 2 O 3 [11], SiO 2 [6], ZrO 2 [12,13] and zeolite [3,14], can also affect catalytic performance on account of the significant differences in textural properties. Especially, the Mg(Al)O Since the activities of C3H8/H2/N2-and H2/N2-reduced catalytic systems were much higher than those of the others, the reaction time was extended to 30 h to compare the stability of the C3H8/H2/N2and H2/N2-reduced catalytic systems. When the reaction time was prolonged, some distinct and significant differences could be found between them. With a reaction of 30 h, the propane conversion and propylene selectivity of H2/N2-reduced samples were maintained at more than 45% and 92% and the by-product selectivity was less than 4%, respectively. However, the C3H8/H2/N2-reduced catalyst showed a rapid decrease of propane conversion and propylene selectivity after 18 h along with the increase of cracking products. In order to evaluate the effect of H2 content, the 14 vol% H2/N2 with the same H2 content as C3H8/H2/N2 was also adopted for the comparison. It is noticeable that the propane conversion and propylene selectivity of the 14 vol% H2/N2-reduced sample was similar to that of the H2/N2-reduced one. This indicates that the reducibility of H2/N2 is quite similar to that of 14 vol% H2/N2 under this reduction temperature. Therefore, the different performance between H2/N2 and C3H8/H2/N2 is arising from the presence of C3H8 in C3H8/H2/N2, which makes the C3H8/H2/N2reduced sample rapidly deactivate along with the increase of cracking products. Meanwhile, the H2/N2-reduced sample had effective catalytic stability and crack resistance.

XRD
According to the XRD patterns of precursors and calcined samples obtained with different preparation processes in Figure S1, it could be confirmed that the HT phase can be formed on the Since the activities of C 3 H 8 /H 2 /N 2 -and H 2 /N 2 -reduced catalytic systems were much higher than those of the others, the reaction time was extended to 30 h to compare the stability of the C 3 H 8 /H 2 /N 2and H 2 /N 2 -reduced catalytic systems. When the reaction time was prolonged, some distinct and significant differences could be found between them. With a reaction of 30 h, the propane conversion and propylene selectivity of H 2 /N 2 -reduced samples were maintained at more than 45% and 92% and the by-product selectivity was less than 4%, respectively. However, the C 3 H 8 /H 2 /N 2 -reduced catalyst showed a rapid decrease of propane conversion and propylene selectivity after 18 h along with the increase of cracking products. In order to evaluate the effect of H 2 content, the 14 vol% H 2 /N 2 with the same H 2 content as C 3 H 8 /H 2 /N 2 was also adopted for the comparison. It is noticeable that the propane conversion and propylene selectivity of the 14 vol% H 2 /N 2 -reduced sample was similar to that of the H 2 /N 2 -reduced one. This indicates that the reducibility of H 2 /N 2 is quite similar to that of 14 vol% H 2 /N 2 under this reduction temperature. Therefore, the different performance between H 2 /N 2 and C 3 H 8 /H 2 /N 2 is arising from the presence of C 3 H 8 in C 3 H 8 /H 2 /N 2 , which makes the C 3 H 8 /H 2 /N 2 -reduced sample rapidly deactivate along with the increase of cracking products. Meanwhile, the H 2 /N 2 -reduced sample had effective catalytic stability and crack resistance.

XRD
According to the XRD patterns of precursors and calcined samples obtained with different preparation processes in Figure S1, it could be confirmed that the HT phase can be formed on the ZnO precursor (hydrozincite phase, JCPDS file No. 19-1458); even the ZnO-supported PtIn-HT can be acquired by impregnation-induced reconstruction. After calcination, only MgO phases (JCPDS file No. 87-0651) could be indexed on ZnO (JCPDS file No. 75-0576). The Al 2 O 3 diffraction peaks could not be found, as it is retained and located in interstitial sites in the MgO framework after calcination [38]. There were no diffraction peaks of Pt-(In-)-based compounds, as a result of their low concentration or uniform distribution [38].
In order to evaluate the crystalline phase of the reduced PtIn/MAZ catalytic system, XRD characterization was conducted, and the results are shown in Figure 2. It can be noted that the XRD patterns of all samples show the same crystalline diffraction peaks. The main crystalline phases were ZnO and MgO and the peaks were same with those of calcined ones ( Figure S1B). This means the change in reduction atmosphere did not cause the obvious change of crystal phases. The diffraction peaks of Pt(In)-species still could not be recognized for the reduced PtIn/MAZ catalytic system. This is owing to the small and uniform distribution of Pt(In) species on the support or the low concentration [19].
ZnO precursor (hydrozincite phase, JCPDS file No. 19-1458); even the ZnO-supported PtIn-HT can be acquired by impregnation-induced reconstruction. After calcination, only MgO phases (JCPDS file No. 87-0651) could be indexed on ZnO (JCPDS file No. 75-0576). The Al2O3 diffraction peaks could not be found, as it is retained and located in interstitial sites in the MgO framework after calcination [38]. There were no diffraction peaks of Pt-(In-)-based compounds, as a result of their low concentration or uniform distribution [38].
In order to evaluate the crystalline phase of the reduced PtIn/MAZ catalytic system, XRD characterization was conducted, and the results are shown in Figure 2. It can be noted that the XRD patterns of all samples show the same crystalline diffraction peaks. The main crystalline phases were ZnO and MgO and the peaks were same with those of calcined ones ( Figure S1B). This means the change in reduction atmosphere did not cause the obvious change of crystal phases. The diffraction peaks of Pt(In)-species still could not be recognized for the reduced PtIn/MAZ catalytic system. This is owing to the small and uniform distribution of Pt(In) species on the support or the low concentration [19].

TEM
The morphology, phase composition and particle size distribution (PSD) of the reduced catalytic system are given in Figure 3. Although it is difficult to distinguish the diffraction of metallic phases from the XRD patterns in Figure 1, the metallic Pt (111) nanoparticles (NPs) can be recognized in the HR-TEM depending on the lattice spacing of 0.226 nm. Meanwhile, PtIn alloy (d = 0.208 nm) can be detected on the H2/N2-reduced sample. From the PSD of all samples, the average particle size of metals was in the range from 0.85-1.20 nm. The small metal NPs can be related to the PtIn-HT reconstruction, as described in Figure S1. Moreover, the average metal particle sizes of H2/N2-and C3H8/H2/N2-reduced samples are smaller than that of C3H8-and CO-reduced ones. This indicates that the H2-containing reduction gas is more efficient to obtain smaller metallic particles. The smaller metal particles are more active for propane dehydrogenation performance and beneficial to resist the coke deposition [39].

TEM
The morphology, phase composition and particle size distribution (PSD) of the reduced catalytic system are given in Figure 3. Although it is difficult to distinguish the diffraction of metallic phases from the XRD patterns in Figure 1, the metallic Pt (111) nanoparticles (NPs) can be recognized in the HR-TEM depending on the lattice spacing of 0.226 nm. Meanwhile, PtIn alloy (d = 0.208 nm) can be detected on the H 2 /N 2 -reduced sample. From the PSD of all samples, the average particle size of metals was in the range from 0.85-1.20 nm. The small metal NPs can be related to the PtIn-HT reconstruction, as described in Figure S1. Moreover, the average metal particle sizes of H 2 /N 2 -and C 3 H 8 /H 2 /N 2 -reduced samples are smaller than that of C 3 H 8 -and CO-reduced ones. This indicates that the H 2 -containing reduction gas is more efficient to obtain smaller metallic particles. The smaller metal particles are more active for propane dehydrogenation performance and beneficial to resist the coke deposition [39]. . The Al2O3 diffraction peaks could not be found, as it is retained and located in interstitial sites in the MgO framework after calcination [38]. There were no diffraction peaks of Pt-(In-)-based compounds, as a result of their low concentration or uniform distribution [38]. In order to evaluate the crystalline phase of the reduced PtIn/MAZ catalytic system, XRD characterization was conducted, and the results are shown in Figure 2. It can be noted that the XRD patterns of all samples show the same crystalline diffraction peaks. The main crystalline phases were ZnO and MgO and the peaks were same with those of calcined ones ( Figure S1B). This means the change in reduction atmosphere did not cause the obvious change of crystal phases. The diffraction peaks of Pt(In)-species still could not be recognized for the reduced PtIn/MAZ catalytic system. This is owing to the small and uniform distribution of Pt(In) species on the support or the low concentration [19].

TEM
The morphology, phase composition and particle size distribution (PSD) of the reduced catalytic system are given in Figure 3. Although it is difficult to distinguish the diffraction of metallic phases from the XRD patterns in Figure 1, the metallic Pt (111) nanoparticles (NPs) can be recognized in the HR-TEM depending on the lattice spacing of 0.226 nm. Meanwhile, PtIn alloy (d = 0.208 nm) can be detected on the H2/N2-reduced sample. From the PSD of all samples, the average particle size of metals was in the range from 0.85-1.20 nm. The small metal NPs can be related to the PtIn-HT reconstruction, as described in Figure S1. Moreover, the average metal particle sizes of H2/N2-and C3H8/H2/N2-reduced samples are smaller than that of C3H8-and CO-reduced ones. This indicates that the H2-containing reduction gas is more efficient to obtain smaller metallic particles. The smaller metal particles are more active for propane dehydrogenation performance and beneficial to resist the coke deposition [39].

XPS
The chemical states of In species over the reduced catalytic system were determined by XPS measurements. The Pt 4f spectra is not shown here due to the difficulty of distinguishing it from the Al 2p peak [40]. The XPS spectra of the In3d5/2 region are shown in Figure 4, and the semi-quantitative results of the corresponding spectra are summarized in Table 1. The In3d5/2 level can be split into two peaks at ~ 444.1 eV and ~ 445.2 eV, which represent In 0 (metallic In or PtIn alloy) and In 3+ , respectively. As shown in Table 1, the In 0 /In 3+ molar ratios for H2/N2,-, C3H8/H2/N2-, C3H8-and CO-reduced samples were 2.50, 1.22, 0.59 and 0.78, respectively. According to the downward trend of the In 0 /In 3+ molar ratio from H2/N2 to C3H8/H2/N2 and to C3H8, it is assumed that the presence of C3H8 can impede the formation of zero-valence In. The deduction is based on the similar reducibility between H2/N2 and 14 vol % H2/N2, which has been proved by the corresponding catalytic performance in Figure 1. In view of this, the In 0 /In 3+ molar ratio in the C3H8/H2/N2-reduced sample should be equal to or higher than that of the H2/N2-reduced one, due to the co-existence of H2 and C3H8. The opposite results indicate that the C3H8 must impede the reduction of In 3+ ions. In this way, it is not difficult to understand why the In 0 /In 3+ molar ratio of CO-reduced sample was higher than C3H8-reduced one.

XPS
The chemical states of In species over the reduced catalytic system were determined by XPS measurements. The Pt 4f spectra is not shown here due to the difficulty of distinguishing it from the Al 2p peak [40]. The XPS spectra of the In3d 5/2 region are shown in Figure 4, and the semi-quantitative results of the corresponding spectra are summarized in Table 1. The In3d 5/2 level can be split into two peaks at~444.1 eV and~445.2 eV, which represent In 0 (metallic In or PtIn alloy) and In 3+ , respectively. As shown in Table 1, the In 0 /In 3+ molar ratios for H 2 /N 2 ,-, C 3 H 8 /H 2 /N 2 -, C 3 H 8 -and CO-reduced samples were 2.50, 1.22, 0.59 and 0.78, respectively. According to the downward trend of the In 0 /In 3+ molar ratio from H 2 /N 2 to C 3 H 8 /H 2 /N 2 and to C 3 H 8 , it is assumed that the presence of C 3 H 8 can impede the formation of zero-valence In. The deduction is based on the similar reducibility between H 2 /N 2 and 14 vol % H 2 /N 2 , which has been proved by the corresponding catalytic performance in Figure 1. In view of this, the In 0 /In 3+ molar ratio in the C 3 H 8 /H 2 /N 2 -reduced sample should be equal to or higher than that of the H 2 /N 2 -reduced one, due to the co-existence of H 2 and C 3 H 8 . The opposite Catalysts 2020, 10, 485 6 of 15 results indicate that the C 3 H 8 must impede the reduction of In 3+ ions. In this way, it is not difficult to understand why the In 0 /In 3+ molar ratio of CO-reduced sample was higher than C 3 H 8 -reduced one. The In 0 /In 3+ molar ratio of the CO-reduced sample was lower than that of H 2 -containing reduced sample, indicating that the reduction ability of CO to In 3+ was weaker than that of H 2 . The In 0 has been proved to promote the transfer of coke from the active site to the support and the stabilization of the active metal species [40,41]. Meanwhile, the presence of C 3 H 8 leads to a decrease of In 0 , thus the propylene selectivity of the H 2 /N 2 -, C 3 H 8 /H 2 /N 2 -and C 3 H 8 -reduced catalytic systems decreases accordingly (see Figure 1).
Catalysts 2020, 10, x FOR PEER REVIEW 6 of 15 The In 0 /In 3+ molar ratio of the CO-reduced sample was lower than that of H2-containing reduced sample, indicating that the reduction ability of CO to In 3+ was weaker than that of H2. The In 0 has been proved to promote the transfer of coke from the active site to the support and the stabilization of the active metal species [40,41]. Meanwhile, the presence of C3H8 leads to a decrease of In 0 , thus the propylene selectivity of the H2/N2-, C3H8/H2/N2-and C3H8-reduced catalytic systems decreases accordingly (see Figure 1).    The binding energy (BE) value of In 0 presented a slightly decreasing trend in accordance with the order of H 2 /N 2 -, C 3 H 8 /H 2 /N 2 -, CO-and C 3 H 8 -reduced samples, signifying the weakening Pt-In interaction and transfer of electrons from metallic In to Pt species [22]. The electron-rich Pt sites are beneficial to hindering the adsorption of alkenes on active sites, consequently decreasing the likelihood of hydrogenolysis and coking [42].
The Pt dispersion of H 2 /N 2 -and C 3 H 8 /H 2 /N 2 -reduced samples was evaluated by CO chemisorption due to their far better propane dehydrogenation performance than the other two samples. According to the results in Table 1, the 48% dispersion of H 2 /N 2 reduced sample is obviously higher than the 34% dispersion of the C 3 H 8 /H 2 /N 2 -reduced one. This indicates that the H 2 /N 2 -reduced catalyst had a higher utilization of Pt. Combined with the high In 0 /In 3+ molar ratio and the formation of PtIn alloy in the H 2 /N 2 -reduced sample, it can be deduced that the Pt-In synergistic effect, high Pt dispersion and large In 0 content result in good activity and high stability to some degree. Figure 5 shows the XRD patterns of the used catalytic system. Compared with the reduced catalytic system, the characteristic diffraction peaks of graphite (JCPDS file No. 75-1621) appear at 2θ of ca. 26 • and 44 • . The peak intensity at 2θ of 26 • cuts down progressively following the order of the catalyst reduced with C 3 H 8 , C 3 H 8 /H 2 /N 2 , CO and H 2 /N 2 , which is correlated with the crystallinity of graphite. It can be declared that the heavily condensed coke species are more difficult to form on the H 2 /N 2 -reduced catalyst than others. At the same time, this also means that the H 2 /N 2 -reduced catalyst is easy to regenerate via oxidation in air [43]. In addition, the Pt and In species still cannot be distinguished, indicating that the metal species are quite stable and highly dispersed on the surface of PtIn/MAZ. The binding energy (BE) value of In 0 presented a slightly decreasing trend in accordance with the order of H2/N2-, C3H8/H2/N2-, CO-and C3H8-reduced samples, signifying the weakening Pt-In interaction and transfer of electrons from metallic In to Pt species [22]. The electron-rich Pt sites are beneficial to hindering the adsorption of alkenes on active sites, consequently decreasing the likelihood of hydrogenolysis and coking [42].

XRD
The Pt dispersion of H2/N2-and C3H8/H2/N2-reduced samples was evaluated by CO chemisorption due to their far better propane dehydrogenation performance than the other two samples. According to the results in Table 1, the 48% dispersion of H2/N2 reduced sample is obviously higher than the 34% dispersion of the C3H8/H2/N2-reduced one. This indicates that the H2/N2-reduced catalyst had a higher utilization of Pt. Combined with the high In 0 /In 3+ molar ratio and the formation of PtIn alloy in the H2/N2-reduced sample, it can be deduced that the Pt-In synergistic effect, high Pt dispersion and large In 0 content result in good activity and high stability to some degree. Figure 5 shows the XRD patterns of the used catalytic system. Compared with the reduced catalytic system, the characteristic diffraction peaks of graphite (JCPDS file No. 75-1621) appear at 2θ of ca. 26° and 44°. The peak intensity at 2θ of 26° cuts down progressively following the order of the catalyst reduced with C3H8, C3H8/H2/N2, CO and H2/N2, which is correlated with the crystallinity of graphite. It can be declared that the heavily condensed coke species are more difficult to form on the H2/N2-reduced catalyst than others. At the same time, this also means that the H2/N2-reduced catalyst is easy to regenerate via oxidation in air [43]. In addition, the Pt and In species still cannot be distinguished, indicating that the metal species are quite stable and highly dispersed on the surface of PtIn/MAZ. It is interesting to notice that the diffraction peaks of meixnerite (JCPDS file No. 38-0478) appeared at 2θ of 35° and 39° in the used catalytic system, instead of the reduced ones reported in [43]. This arises from the high dispersion or low content of meixnerite. On the other hand, the intensity of the corresponding diffraction peaks decreased gradually with the order of H2/N2, C3H8/H2/N2, C3H8 and CO-reduced sample, and even completely disappeared for PtIn/MAZ reduced with CO.

XRD
The formation of meixnerite can be explained by the reconstruction of the calcined HT under the humidified conditions [41,[44][45][46]. As we all know, the calcination can cause H2O and -OH removal from the structure [47,48]. However some highly temperature-stable M-OH2 can be formed on the It is interesting to notice that the diffraction peaks of meixnerite (JCPDS file No. 38-0478) appeared at 2θ of 35 • and 39 • in the used catalytic system, instead of the reduced ones reported in [43]. This arises from the high dispersion or low content of meixnerite. On the other hand, the intensity of the corresponding diffraction peaks decreased gradually with the order of H 2 /N 2 , C 3 H 8 /H 2 /N 2 , C 3 H 8 and CO-reduced sample, and even completely disappeared for PtIn/MAZ reduced with CO.
The formation of meixnerite can be explained by the reconstruction of the calcined HT under the humidified conditions [41,[44][45][46]. As we all know, the calcination can cause H 2 O and -OH removal from the structure [47,48]. However some highly temperature-stable M-OH 2 can be formed on the surface of PtIn/MAZ after reduction with H-containing gas, due to the strong water affinity of calcined HT structure [45]. It is just that the PDH reaction further provides the sufficient -OH 2 concentration level and exposure time on the surface of calcined HT, which makes the calcined HT reconstruct into meixnerite. Therefore, the differences of meixnerite peak intensity among the H 2 /N 2 -, C 3 H 8 /H 2 /N 2and C 3 H 8 -reduced samples can arise from the different levels of -OH 2 , depending on the reducibility of these atmospheres. As a result, the strong diffraction peaks of meixnerite can be obtained by the H 2 /N 2 -reduced sample. The complete disappearance of meixnerite for the CO-reduced sample may have been induced by the formation of bicarbonate species, together with the consumption of oxygen, resulting in no conditions for the formation of meixnerite [36]. As a kind of HT with an interlayer -OH anion, the high crystallinity meixnerite may be helpful to keep the high dispersion and uniform distribution of surface metal active sites during reaction. As a result, the good stability of the H 2 /N 2 -reduced sample can arise from the formation of meixnerite to a certain extent [17].
Additionally, it has been proved that the meixnerite has Brønsted basicity [49], which is different from the Lewis basicity of calcined HT. Thus, the formation of meixnerite can adjust the acid-base properties of the catalyst and improve its coke resistance [50,51]. However, it is difficult to evaluate the acid-base properties by conventional methods because of the influence of deposited coke on the formed meixnerite [50,51]. As compared with the diffraction strength of meixnerite, the only direct evidence is that the diffraction strength of graphite and the amount of deposited coke (see TG analysis below) show an opposite trend in accordance with the order from H 2 /N 2 -to C 3 H 8 /H 2 /N 2 -and C 3 H 8 -reduced catalysts.

TEM
After a reaction of 17 h, the TEM images and the PSD statistics of the used catalytic system reduced with different atmospheres are shown in Figure 6. It was found that the range of PSD became wider, but the average metal particle size only showed a slight increase compared with the corresponding reduced catalytic system. The slight aggregation of metal particles agrees with the fact that the reaction is not in the deactivation stage ( Figure 1). It is worth noting that the metal sintering degree of the C 3 H 8 /H 2 /N 2 -reduced sample was higher than that of the H 2 /N 2 -reduced one, although they had similar catalytic performance during a reaction of 17 h. Therefore, this can be used to explain the stable catalytic performance of the H 2 /N 2 -reduced catalyst and the rapid decline in reactivity of the C 3 H 8 /H 2 /N 2 -reduced sample after a reaction of 17 h. This is attributable to the initial smaller metal particles and high Pt dispersion of the H 2 /N 2 -reduced catalyst. Meanwhile, it indicates strong Pt-In or Pt-support interaction, which results in the superior stability of Pt during the PDH reaction [40]. Coke deposits can be found on the support surface of the used catalytic system instead of the active site, which means that the formed coke should be shifted to the surface of support before a reaction of 17 h [40]. surface of PtIn/MAZ after reduction with H-containing gas, due to the strong water affinity of calcined HT structure [45]. It is just that the PDH reaction further provides the sufficient -OH2 concentration level and exposure time on the surface of calcined HT, which makes the calcined HT reconstruct into meixnerite. Therefore, the differences of meixnerite peak intensity among the H2/N2-, C3H8/H2/N2-and C3H8-reduced samples can arise from the different levels of -OH2, depending on the reducibility of these atmospheres. As a result, the strong diffraction peaks of meixnerite can be obtained by the H2/N2-reduced sample. The complete disappearance of meixnerite for the COreduced sample may have been induced by the formation of bicarbonate species, together with the consumption of oxygen, resulting in no conditions for the formation of meixnerite [36]. As a kind of HT with an interlayer -OH anion, the high crystallinity meixnerite may be helpful to keep the high dispersion and uniform distribution of surface metal active sites during reaction. As a result, the good stability of the H2/N2-reduced sample can arise from the formation of meixnerite to a certain extent [17]. Additionally, it has been proved that the meixnerite has Brønsted basicity [49], which is different from the Lewis basicity of calcined HT. Thus, the formation of meixnerite can adjust the acid-base properties of the catalyst and improve its coke resistance [50,51]. However, it is difficult to evaluate the acid-base properties by conventional methods because of the influence of deposited coke on the formed meixnerite [50,51]. As compared with the diffraction strength of meixnerite, the only direct evidence is that the diffraction strength of graphite and the amount of deposited coke (see TG analysis below) show an opposite trend in accordance with the order from H2/N2-to C3H8/H2/N2-and C3H8reduced catalysts.

TEM
After a reaction of 17 h, the TEM images and the PSD statistics of the used catalytic system reduced with different atmospheres are shown in Figure 6. It was found that the range of PSD became wider, but the average metal particle size only showed a slight increase compared with the corresponding reduced catalytic system. The slight aggregation of metal particles agrees with the fact that the reaction is not in the deactivation stage ( Figure 1). It is worth noting that the metal sintering degree of the C3H8/H2/N2-reduced sample was higher than that of the H2/N2-reduced one, although they had similar catalytic performance during a reaction of 17 h. Therefore, this can be used to explain the stable catalytic performance of the H2/N2-reduced catalyst and the rapid decline in reactivity of the C3H8/H2/N2-reduced sample after a reaction of 17 h. This is attributable to the initial smaller metal particles and high Pt dispersion of the H2/N2-reduced catalyst. Meanwhile, it indicates strong Pt-In or Pt-support interaction, which results in the superior stability of Pt during the PDH reaction [40]. Coke deposits can be found on the support surface of the used catalytic system instead of the active site, which means that the formed coke should be shifted to the surface of support before a reaction of 17 h [40].

TG
In order to evaluate the amounts of deposited coke on the used catalytic system, the TG test was adopted, and the results are shown in Figure 7. The total deposited coke content of the catalyst prereduced with H2/N2 was only 27 wt %, and lower than the others, which coincided with its high propylene selectivity (see Figure 1). The lowest coke formation on H2/N2-reduced catalyst should be related to the small metal particle size, strong Pt-In interaction and suitable acid-base properties caused by the formation of meixnerite. Meanwhile, the more In 0 species also can promote the transfer of coke deposition from the active sites to the support, maintaining the stability of active metal species, ensuring the selectivity of propylene and reducing the occurrence of cracking. The prereduction with C3H8/H2/N2 and C3H8 made the deposited coke content increase to ca. 50 wt %. The severe coke deposition has been proposed to explain the fast loss of activity of the C3H8/H2/N2reduced sample [47]. This is thought to be related to their wide PSD [47]. The mass loss of CO-reduced catalyst is lower than the C3H8/H2/N2 and C3H8-reduced one is, because it is far below propane conversion than them.

TG
In order to evaluate the amounts of deposited coke on the used catalytic system, the TG test was adopted, and the results are shown in Figure 7. The total deposited coke content of the catalyst pre-reduced with H 2 /N 2 was only 27 wt %, and lower than the others, which coincided with its high propylene selectivity (see Figure 1). The lowest coke formation on H 2 /N 2 -reduced catalyst should be related to the small metal particle size, strong Pt-In interaction and suitable acid-base properties caused by the formation of meixnerite. Meanwhile, the more In 0 species also can promote the transfer of coke deposition from the active sites to the support, maintaining the stability of active metal species, ensuring the selectivity of propylene and reducing the occurrence of cracking. The pre-reduction with C 3 H 8 /H 2 /N 2 and C 3 H 8 made the deposited coke content increase to ca. 50 wt %. The severe coke deposition has been proposed to explain the fast loss of activity of the C 3 H 8 /H 2 /N 2 -reduced sample [47]. This is thought to be related to their wide PSD [47]. The mass loss of CO-reduced catalyst is lower than the C 3 H 8 /H 2 /N 2 -and C 3 H 8 -reduced one is, because it is far below propane conversion than them. In comparison with the state-of-the-art Pt-based catalysts in the literature (Table S1), the PtIn/MAZ catalytic system obtained in this work possesses much higher catalytic performance and attractive stability.

Preparation of ZnO Precursor
The ZnO precursor was prepared by the hydrothermal method. Briefly, 0.004 mol of Zn(NO3)2•6H2O and 0.083 mol of urea was fully dissolved in 65 mL deionized water, then moved to a Teflon-lined stainless steel autoclave at 100 °C for 20 h. After cooling to room temperature, the products were filtrated and washed with deionized water till neutrality, then dried at 100 °C overnight to obtain the ZnO precursor.

Preparation of MAZ Support
The Mg(Al)O/ZnO (MAZ) precursor was prepared by the hydrothermal method. The molar ratio of Mg/Al is 2. Firstly, 0.1 g of ZnO precursor was introduced in 65 ml of a mixed solution, including 0.003 mol of Mg(NO3)2•6H2O, 0.0015 mol of Al(NO3)3•9H2O and 0.1 mol of urea under vigorous stirring. After hydrothermal treatment in a Teflon-lined stainless-steel autoclave at 100 °C for 20 h, the mixture was cooled to room temperature. Next, the slurry was filtrated and washed with deionized water till to neutrality, then dried at 100 °C overnight. Finally, the corresponding MAZ support was obtained by calcining the precursor at 600 °C for 4 h in a muffle furnace with a heating rate of 5 °C min −1 . In comparison with the state-of-the-art Pt-based catalysts in the literature (Table S1), the PtIn/MAZ catalytic system obtained in this work possesses much higher catalytic performance and attractive stability.

Materials
The reagents for the preparation of catalyst were analytical grade chemicals. Zn(NO 3

Preparation of ZnO Precursor
The ZnO precursor was prepared by the hydrothermal method. Briefly, 0.004 mol of Zn(NO 3 ) 2 ·6H 2 O and 0.083 mol of urea was fully dissolved in 65 mL deionized water, then moved to a Teflon-lined stainless steel autoclave at 100 • C for 20 h. After cooling to room temperature, the products were filtrated and washed with deionized water till neutrality, then dried at 100 • C overnight to obtain the ZnO precursor.

Preparation of MAZ Support
The Mg(Al)O/ZnO (MAZ) precursor was prepared by the hydrothermal method. The molar ratio of Mg/Al is 2. Firstly, 0.1 g of ZnO precursor was introduced in 65 mL of a mixed solution, including 0.003 mol of Mg(NO 3 ) 2 ·6H 2 O, 0.0015 mol of Al(NO 3 ) 3 ·9H 2 O and 0.1 mol of urea under vigorous stirring. After hydrothermal treatment in a Teflon-lined stainless-steel autoclave at 100 • C for 20 h, the mixture was cooled to room temperature. Next, the slurry was filtrated and washed with deionized water till to neutrality, then dried at 100 • C overnight. Finally, the corresponding MAZ support was obtained by calcining the precursor at 600 • C for 4 h in a muffle furnace with a heating rate of 5 • C min −1 .

Preparation and Reduction of PtIn/MAZ
The PtIn/MAZ precursor was synthesized via successive incipient wetness impregnation method. Firstly, the MAZ support was impregnated with an InNO 3 ·xH 2 O aqueous solution at room temperature for 6 h and dried at 120 • C for 12 h to acquire the In/MAZ precursor. After calcination at 550 • C for 4 h, the In-based calcined product was obtained and labelled as In/MAZ. Next, the same processes were performed on In/MAZ by impregnating with an H 2 PtCl 6 ·6H 2 O aqueous solution to acquire the PtIn/MAZ precursor.
The calcined product PtIn/MAZ was obtained by calcining the corresponding precursor at 550 • C for 4 h with a heating rate of 5 • C min −1 [19,41]. To be consistent with previous research [22], the loading amounts of Pt and In were 0.5 wt % and 1.4 wt %, respectively.
Before the PDH reaction, PtIn/MAZ was reduced with different atmospheres. The reduction atmospheres of H 2 /N 2 (5 vol % H 2 ), C 3 H 8 /H 2 /N 2 (16 vol % C 3 H 8 and 14 vol % H 2 ), C 3 H 8 and CO were used to reduce PtIn/MAZ at 600 • C for 2.5 h respectively. In order to compare with H 2 /N 2 (5 vol % H 2 ) and C 3 H 8 /H 2 /N 2 (16 vol % C 3 H 8 and 14 vol % H 2 ), the reduction gas of 14 vol % H 2 /N 2 was also evaluated in the PDH reaction.

Catalyst Characterizations
The X-ray diffraction (XRD) patterns were recorded on a Bruker D8-Focus X-ray diffractometer (BRUKER AXS GMBH, Karlsruhe, Germany) equipped with Cu Kα radiation (λ = 0.15418 nm). The samples were scanned from the 2θ value of 20 • -75 • , with a scanning speed of 8 • ·min −1 . The X-ray tube was manipulated at 40 kV and 40 mA.
The transmission electron microscopy (TEM) images were taken with a JEM-2100F field-emission transmission electron microscope (JEOL, Tokyo, Japan) operated at 200 kV. The as-synthesized samples were dispersed, sonicated in ethanol, and dropped on carbon-film coated copper grids.
X-ray photoelectron spectroscopy (XPS) was carried out on a Perkin-Elmer PHI 5000C ESCA (PerkinElmer, Waltham, MA, USA) using Al Kα radiation. Prior to the analysis, the samples were reduced under different atmospheres (30 mL min −1 ) at 600 • C for 2.5 h, and the binding energies were calibrated using the C1s peak at 284.8 eV as an internal standard. The whole process from reduction to test was strictly protected with Ar, which can avoid the re-oxidation of In 0 . Thermogravimetric (TG) analysis was carried out with a DTG-50/50H thermal analyzer (PerkinElmer, Waltham, MA, USA) to determine the quantity of coke on the used catalyst, and the TG curves were recorded from room temperature to 900 • C at an increasing rate of 10 • C min −1 .
Dispersion of metals was determined by CO chemisorption, CO chemisorption was conducted on AutoChem II 2920 analyzer (Micromeritics, Norcross, GA, USA). For a fresh catalyst, 100 mg of sample was pre-reduced with different atmospheres at 600 • C for 2.5 h and then purged with Ar flow at 300 • C for 0.5 h. CO chemisorption was carried out at 30 • C by injecting 50 µL CO every 3 min until the peak area became stable. The adsorption capacity of CO is calculated by the following Equation.
CO uptakes (µmol g −1 cat ) = n × (A − B) ×50×10/A × 22.4 (1) where n stands for the number of chemical adsorption peaks, A represents the average value of the peak area when the peak area remains constant, and B stands for the total area of the chemical adsorption peaks.

Catalytic Measurements
The PDH reactions were carried out in a fixed-bed quartz reactor with an 8-mm inner diameter. A quantity of 0.4 g of fresh catalyst (particle size of 40 to 60 meshes) was placed into the center of quartz tube reactor and reduced with different reduction atmospheres, as described in Section 2.3. Afterwards, the mixed gas, including C 3 H 8 , H 2 and N 2 (C 3 H 8 :H 2 :N 2 molar ratio = 8:7:35), was fed at a rate of 62 mL min −1 into the reactor. The PDH reactions were performed at atmospheric pressure, 600 • C and a weight hourly space velocity (WHSV) of 3 h −1 , which was generally used in relevant studies [21,52,53]. A gas chromatograph equipped with an FID detector and Al 2 O 3 column was employed to analyze the gas compositions. The propane conversion and propylene selectivity were calculated by the following Equations: propane, in -propane, out propane, in × 100 (2) Selectivity(%) = propylene, out propane, in -propane, out × 100 (3)

Conclusions
A catalyst, PtIn/Mg(Al)O/ZnO, can be obtained by calcination and reduction of ZnO-supported PtIn-HT. The reduction atmosphere has a profound influence on the physicochemical nature and catalytic performance of the PtIn/Mg(Al)O/ZnO catalyst. The propane conversion and propylene selectivity of the catalyst reduced with H 2 /N 2 can reach 51% and 97%, respectively, and even stabilize at 45% and 92% after running for 30 h. This is mainly attributable to its high Pt dispersion and In 0 /In 3+ molar ratio, strong Pt-In interaction and small metal particle size, which also result in a low content of coke deposits on the surface of the H 2 /N 2 -reduced catalyst. In addition, meixnerite can be formed after the reaction, and the crystallinity is influenced by the reduction atmosphere. At the same time, the formation of the high-crystallinity meixnerite can re-disperse the metal active sites and adjust the acid-base properties of the catalyst, thus improving the stability and coke resistance to some extent.

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