Introduction

Providing the fundamentals for rational design and synthesis of heterogeneous catalysts with a certain kind of supported species or/and bulk structure(s), that are the key descriptors of catalyst activity, product selectivity and on-stream stability, is an insistently pursued task towards sustainable development. The precise synthesis of solid catalysts is, however, challenging due to the heterogeneity of crystal structure(s), crystallinity, and/or chemical composition/distribution from nano to atomic/molecular scales. Moreover, there are great difficulties in identifying of catalytically active sites and/or the mechanism of their formation associated with surface heterogeneity.¹ Indeed, a lot of efforts ranging from first-principle calculations to experimental investigations have been devoted to the development of versatile strategies for purposeful preparation of efficient heterogeneous catalysts.² For example, top-down/bottom-up assembly,³ atomic layer deposition,⁴ flame spray pyrolysis,⁵ solution combustion synthesis,⁶ and plasma-assisted routes⁷ have been applied. Although these strategies have provided important fundamentals for preparation of well-defined structures/phases, their application for large-scale catalyst preparation is commonly failed to meet the economic and/or technological criteria in comparison with the industrially relevant approaches including impregnation, precipitation, and hydrothermal or sol-gel syntheses. The latter methods cannot, however, ensure the creation of controlled structures, resulting in inhomogeneous catalyst structure(s) and composition(s).

Catalysts based on oxides of vanadium, chromium, aluminum, cerium, molybdenum, or zinc are widely used in different industrial reactions involving activation of C−H bonds in various hydrocarbons.⁸ As their preparation typically involves a precipitation process, the inherently unmanageable precipitation/recrystallization rates of hydrous oxides or carbonates in aqueous solutions result in the heterogeneity in structure, crystallinity, and local composition of metal oxide catalysts. In this case, great efforts are practiced to regulate the precipitation rates through adding complex agents,⁹ in addition to a precise control of synthesis parameters such as temperature and pH value. However, the desired structural control during the large-scale production of solid catalysts is still a great issue. This can be exemplarily reflected by widely used γ-alumina, the structure, crystallinity, and textural properties of which are always varied to different extents from batch to batch of the same producer or individual manufacturers.¹⁰ In fact, each precipitation process in an aqueous solution is based on the precipitation-dissolution dynamic equilibrium of metal cations as established and documented in the textbook of heterogeneous catalysis.¹¹ Thus, if a given metal oxide is added to an aqueous solution, its dissolution and re-precipitation may also occur dynamically.¹²

Herein we introduce a simple method with a potential of large-scale applications to tailor the structure of metal-oxide catalysts and to efficiently improve their performance in dehydrogenation reactions of different hydrocarbons. Our working hypothesis was that structural and chemical properties of supported metal oxide species can be controlled through catalyst post-treatment in an NH₃ ⋅ H₂O solution because of dissolution-recrystallization of amorphous or/and low-crystalline supports in the parent catalysts. This methodology was validated on the examples of VO_x−Al₂O₃ and CeO₂−ZrO₂−Al₂O₃ materials tested in the oxidative dehydrogenation of ethylbenzene to styrene with carbon dioxide (CO₂−ODEB) and the non-oxidative propane dehydrogenation to propene (PDH). The activity of isolated VO_x species depends on the strength of V−O−Al bond, which is determined by the content of coordinatively unsaturation of aluminum cations on the support surface. The concentration of such defective sites can be controlled through our preparation method. Redox properties of ceria can also be adjusted and are relevant for the dehydrogenation activity.

Results and Discussion

A VO_x−Al₂O₃ catalyst with a V₂O₅ loading of 3 wt % (abbreviated as s-3VA) synthesized by a sol-gel method¹³ is introduced to demonstrate our approach schematically shown in Figure 1a. The post-treatment of s-3VA in an NH₃ ⋅ H₂O solution with a pH of 13 led to the post-treated catalyst s-3VA-pH 13. As revealed by transmission electron microscopy (TEM), high resolution TEM (HRTEM) and high-angle annular dark field scanning TEM (HAADF-STEM), s-3VA is composed of agglomerated nanosheets (Figure 1b and Figure S1a,b) and Al₂O₃ particles (5–10 nm) with a polycrystalline structure (Figure 1c and Figure S2a,b). The lattice fringes of (400) and (131) with a characteristic acute angle of 72.5° for γ-Al₂O₃ were identified (Figure 1d). In contrast, s-3VA-pH 13 consists of larger particles of Al₂O₃ with relatively regular edges and morphologies (Figure 1e and Figure S1c,d). The crystallinity of γ-Al₂O₃ is obviously improved after the post-treatment; the diffraction spots of FFTs are much sharper and brighter than those of s-3VA (Figure 1f,g and Figure S2c,d). No crystalline V₂O₅ was detected both in s-3VA and s-3VA-pH 13 (Figure 1c,f and Figure S2a,c). Vanadium is well distributed on the surface of catalysts as revealed by the elemental mapping of energy dispersive X-ray spectroscopy (EDX, Figure 1h,i).

To find the reason behind the above-discussed structural modifications of Al₂O₃, s-3VA was additionally treated in an NH₃ ⋅ H₂O solution with pH of 9 or 11 to yield the s-3VA-pH 9 and s-3VA-pH 11 catalysts. Importantly, inductively coupled plasma mass spectrometry (ICP-MS) measurements confirmed the presence of Al³⁺ ions in the supernatant after post-treatment of s-3VA. The concentration of aluminum increased from 6.6 to 32.7 mg/L with an increase in the pH value of NH₃ ⋅ H₂O solution from 9 to 13 (Table S1). X-ray diffraction (XRD) analysis of post-treated s-3VA after drying (not calcined) indicated the presence of hydrous Al₂O₃ species (Figure 2a) mainly in the form of pseudoboehmite (PDF#83-2384) in s-3VA-pH 9-P, gibbsite (PDF#33-0018) in s-3VA-pH 11-P, and nordstrandite (PDF#24-0006) in s-3VA-pH 13-P.¹⁴ Thus, the dissolution and the re-precipitation of Al₂O₃ took place during the s-3VA post-treatment.

As revealed by thermal gravimetric (TG) and differential scanning calorimetry (DSC) analyses (Figure 2b and Figure S3a,b), the amount of hydrous Al₂O₃ increased in the order of s-3VA-pH 9-P<s-3VA-pH 11-P<s-3VA-pH 13-P. This indicates that higher concentration of OH⁻ can accelerate dissolution-precipitation processes, leading to the formation of higher amounts of hydrous Al₂O₃. This compound is known to transform into γ-Al₂O₃ via dehydroxylation/dehydration during calcination at 550 °C.¹⁵ Only diffractions characteristic of γ-Al₂O₃ (PDF#50-0741) were observed in the XRD patterns of the parent and post-treated (calcined) samples (Figure 2c), coinciding with HRTEM observations (Figure 1c,f and Figure S2a,c). The intensity of the main diffractions increased in the order of s-3VA<s-3VA-pH 9<s-3VA-pH 11<s-3VA-pH 13, while the full width at half maximum (FWHM) of the (440) diffraction decreased from 4.5° to 1.8° (Figure S4). These changes indicate the recrystallization of amorphous and low-crystalline Al₂O₃ in the parent s-3VA catalyst during the post-treatment and confirm the importance of the solution pH value for the crystallinity of Al₂O₃.

To check if the structure of hydrous Al₂O₃ affects its transformation into crystalline γ-Al₂O₃, a VO_x−Al₂O₃ catalyst (3 wt % V₂O₅, abbreviated as e-3VA) with an amorphous structure (Figure S5a) was on purpose synthesized by an evaporation-induced self-assembly method.¹⁶ It was also subjected to a post-treatment in an NH₃ ⋅ H₂O solution with pH of 13. In contrast to s-3VA-pH 13-P with the dominant nordstrandite phase (Figure 2a), both the nordstandite and pseudoboehmite phases were formed in the treated but non-calcined e-3VA-pH 13-P (Figure S5a). The amount of hydrous Al₂O₃ determined by TG analysis in this sample is very close to that in s-3VA-pH 13-P (Figure S5b). However, s-3VA-pH 13 possesses a clearly higher amount of crystalline γ-Al₂O₃ than e-3VA-pH 13 after these samples were calcined at 550 °C (Figure S5c). Thus, comparing with pseudoboehmite, the dehydroxylation/dehydration of nordstandite led to a higher crystallinity of γ-Al₂O₃. This is consistent with the higher crystallinity of γ-Al₂O₃ in s-3VA-pH 13 than that in s-3VA-pH 9, because the nordstrandite and pseudoboehmite phases are dominant in s-3VA-pH 13-P and s-3VA-pH 9-P, respectively. Thus, both the amount and the structure of hydrous Al₂O₃, which are determined by the pH value of NH₃ ⋅ H₂O solution during the post-treatment, affect the crystallinity of γ-Al₂O₃ in the post-treated catalysts.

²⁷Al magic-angle-spinning (MAS) NMR measurements were additionally carried out to analyze the chemical environment of Al species in the parent and post-treated samples. Three peaks at about 6, 31 and 70 ppm are present in the obtained spectra (Figure S6a), which can be assigned to Al³⁺ species in octahedral (AlO₆), pentahedral (AlO₅) and tetrahedral (AlO₄) coordination, respectively.¹⁷ The concentration of coordinatively unsaturated Al sites with the AlO₅ coordination (Al_cus) decreases from 19.1 % to 9.3 % in the order s-3VA>s-3VA-pH 9>s-3VA-pH 11>s-3VA-pH 13 (Figure 2d), while the concentration of coordinatively saturated Al sites (Al_cs) with the AlO₆ or AlO₄ coordination shows an opposite order (Figure S6b), which can be attributed to the increased crystallinity of γ-Al₂O₃.

Although a small quantity of vanadium ions was detected by ICP-MS in the supernatant after post-treatment of s-3VA (Table S1), no crystalline V₂O₅ was formed in the post-treated catalysts (Figure 1f and Figure S2c). This suggests that no recrystallization process related to VO_x occurred during the catalyst treatment. To find out the reason behind this observation, a controlled dissolution experiment was performed. The crystalline V₂O₅ sample was completely dissolved in the NH₃ ⋅ H₂O solution within 0.5 h (Figure S7). This indicates that the dissolution-precipitation processes, which take place on the surface of amorphous and low-crystalline Al₂O₃ in the parent s-3VA catalyst during the post-treatment, do not occur in the case of crystalline V₂O₅. Thus, any recrystallization of VO_x species can be ruled out.

The content of vanadium in the parent and post-treated VO_x−Al₂O₃ determined by ICP-MS was same within the experimental error (1.55±0.05 wt %, Table S2). A symmetric V2p_3/2 peak with the binding energy of 517.6 eV and FWHM of 1.8±0.1 eV assigned to V⁵⁺ was observed in the X-ray photoelectron spectra (XPS) of all catalysts (Figure S8). Thus, the post-treatment did not affect the content and the oxidation state of vanadium.

The in situ UV Raman spectrum of the parent s-3VA catalyst contains only two distinct shifts at 901 and 1015 cm⁻¹ (Figure 2e), which are characteristic for the vibrations of V−O−Al and V=O bonds in monomeric VO_x species, respectively.¹⁸ The high dispersion of vanadium is due to its low surface density of 0.44 V atoms/nm² (Table S2), which is far below the theoretical monolayer density of 2.3 V atoms/nm² for monomeric VO₄.¹⁹ Although no additional bands are present in the spectra of the post-treated catalysts, the Raman shift due to the vibration of V−O−Al bonds decreases from 901 to 886 cm⁻¹ in the order s-3VA>s-3VA-pH 9>s-3VA-pH 11>s-3VA-pH 13. The red shift of this characteristic peak was also observed in the spectrum of VO_x/γ-Al₂O₃ after the introduction of sulfur.²⁰ It is commonly explained by weakening the strength of the bonds as reported in previous studies.²¹ Further understanding on this phenomenon can be obtained from in situ UV/Vis spectra of the catalysts under anhydrous conditions (Figure S9). The therefrom calculated edge energy (Eg) of VO_x decreases from 3.75 (s-3VA) to 3.62 (s-3VA-pH 9), 3.53 (s-3VA-pH 11), and 3.50 eV (s-3VA-pH 13) (Figure 2f). Generally, Eg depends on the polymerization degree of VO_x and cationic ligand of the support.²² As only monomeric VO₄ species are present in the parent and post-treated catalysts, the change in the Eg values may originate from the coordination degreed of Al sites in the support surface, which are connected with the VO_x species. Among these four catalysts, the parent s-3VA with amorphous and low-crystalline Al₂O₃ shows the highest Eg values, indicating the strongest VO_x−Al₂O₃ interaction. This coincides with the highest strength of V−O−Al bonds as proved by the Raman results (Figure 2e). The strength and the Eg value decrease with an increase in the pH value of NH₃ ⋅ H₂O solution used for catalyst post-treatment. These facts can be ascribed as the decreased amount of Al sites strongly interacting with VO_x species. In this case, Al_cus typically characterizing the strong interactions with VO_x species should be such sites, which is also supported by the fact that Al_cus content actually decreases with an increase in the crystallinity of γ-Al₂O₃ (Figure 2d and Figure S6a,b).

Regardless of the pH value of the post-treatment solution, the amount of H₂ consumed during temperature-programmed reduction tests with H₂ (H₂-TPR) is same for all catalysts, which corresponds to the reduction of V⁵⁺ to V⁴⁺ (+4.15±0.05, Table S3).^{18b, 23} This can be attributed to the exclusively monomeric VO_x species over these catalysts as revealed by Raman results (Figure 2e). However, the temperature of the maximal rate of H₂ consumption decreases from 520 °C for the parent s-3VA to 514–504 °C for the post-treated s-3VA with an increase in the pH value of the NH₃ ⋅ H₂O solution (Figure S10). This indicates that the reducibility of monomeric VO_x increases in the order of s-3VA<s-3VA-pH 9<s-3VA-pH 11<s-3VA-pH 13. Considering that monomeric VO_x species only contains V−O−Al and V=O bonds, such improvement can be explained by weakened interactions between VO_x and Al₂O₃ as indicated by the decreased Eg value and strength of V−O−Al bonds. Summarizing these results and their discussion, the dissolution-precipitation mechanism and the induced effect of the post-treatment on VO_x−Al₂O₃ catalysts were developed, which are schematically shown in Figure 2g.

We also proved the potential of the developed method for preparation of bulk catalysts. A representative CeO₂, ZrO₂ and Al₂O₃ mixed oxide (28 wt % CeO₂−ZrO₂ with a Ce/Zr molar ratio of 6/4, abbreviated as s-28CZA) was synthesized through a sol-gel method and was post-treated in an NH₃ ⋅ H₂O solution with a pH of 13 (abbreviated as s-28CZA-pH 13). Amorphous Al₂O₃, CeO₂ and ZrO₂ were identified in s-28CZA (Figure 3a and Figure S11a), which may be explained as that Al₂O₃ hinders the formation of crystalline CeO₂−ZrO₂ solid solutions.²⁴ Importantly, the post-treated s-28CZA-pH 13 is composed of well-crystallized γ-Al₂O₃ and cubic Ce-based oxide with oxygen defects (Figure 3a and Figure S11a). To identify the phase properties, pure CeO₂ (abbreviated as s-Ce) and Ce_0.6Zr_0.4O₂ solid solution (abbreviated as s-CZ) also synthesized via the sol-gel method were characterized by visible Raman spectroscopy (Figure S11a) and XRD (Figure S12). In comparison with the XRD pattern of s-Ce, the addition of Zr to CeO₂ (s-CZ) led to a shift of the CeO₂-related reflections to larger 2θ values (Figure 3b, Figure S12) accompanying with decreased lattice parameter (Table S4). This is due to the formation of Ce_0.6Zr_0.4O₂ solid solution via the insertion of Zr⁴⁺ into the CeO₂ lattice.²⁵ Noteworthy, the (111) diffraction and the lattice parameter of s-28CZA-pH 13 are same as those of s-CZ (Figure 3b and Table S4), suggesting the same composition of the solid solution in these two catalysts.^{24a, 26} XRD results of dried precursor, i.e., s-28CZA-pH 13-P clearly show the presence of hydrous CeO₂, nordstrandite and gibbsite (Figure S13a), indicating that the dissolution and the precipitation of amorphous and/or low-crystalline Al₂O₃ and CeO₂ took place simultaneously during the post-treatment of the parent s-28CZA material. Although the hydrous ZrO₂ was not detected by XRD perhaps because of its amorphous nature (Figure S13b), its formation is expected if the solubility product constants of hydrous ZrO₂ (pKsp=48.2) and hydrous CeO₂ (pKsp=47.7) are taken into account. The s-28CZA-pH 13 sample shows a clearly higher reducibility (Figure S14) than the parent s-28CZA. Moreover, in comparison with s-CZ, s-28CZA-pH 13 has a higher concentration of oxygen defects as confirmed by the Raman and Ce 3d XPS patterns (Figures S11b and S15a,b). This is due to smaller crystal size of Ce_0.6Zr_0.4O₂ in s-28CZA-pH 13 (5.2 nm) than that in s-CZ (7.8 nm, Table S4).

The oxidative dehydrogenation of C₂-C₄ alkanes or ethylbenzene (EB) with carbon dioxide is an attractive technology for the efficient production of olefins or styrene in tandem with the CO₂ reduction to CO.^{8a, 8b} V- or Ce-based oxides are promising catalysts for CO₂−ODEB, and their activity is closely connected with their structural and chemical properties.^{18b, 25} Now we demonstrate the potential of the developed catalysts for the CO₂−ODEB reaction. In agreement with our previous findings,^{18b, 27} both the parent and post-treated pure Al₂O₃ and ZrO₂(28 wt %)-Al₂O₃ oxides are essentially inactive (Figure S16). Thus, the below discussion can be safely focused on the role of supported VO_x species or Ce-based phases in this reaction.

The apparent turnover frequency (TOF_EB) over V- and Ce-based oxides is defined and calculated as the number of EB molecules converted per V or Ce atom per hour. The TOF_EB value of VO_x−Al₂O₃ catalysts increases from 22.8 to 41.0 h⁻¹ in the order s-3VA<s-3VA-pH 9<s-3VA-pH 11<s-3VA-pH 13 (Figure 4a), while the selectivity to styrene is above 99.0 % for all catalysts (Figure S17). The enhanced activity of post-treated catalysts can be reasonably associated with the weakened interaction between VO_x and Al₂O₃ because the oxidation state of vanadium (V⁵⁺), the molecular structure of VO_x species (exclusively monomeric VO_x), and the amount of reducible VO_x species are essentially same for all these catalysts.

The TOF_EB value of the s-28CZA-pH 13 catalyst is about 13 times higher than that of s-28CZA (10.6 vs. 0.8 h⁻¹, Figure S18). This result together with the catalyst characterization data (Figure 3a and Figure S11a) proves that amorphous CeO₂ in s-28CZA is almost inert for activating C−H bonds in EB, while the crystalline Ce_0.6Zr_0.4O₂ solid solution in s-28CZA-pH 13 is highly active. A similar conclusion about the effect of crystallinity was also reported for pure ZrO₂-catalyzed PDH reaction.²⁸ Thus, coordinatively unsaturated metal cations (Me_cus) in crystalline CeO₂ or ZrO₂ are required to break C−H bonds in EB or C₃H₈ molecules. This is supported by a higher TOF_EB value of s-28CZA-pH 13 than s-CZ (Figure S18) due to the higher concentration of oxygen defects and Ce³⁺ (direct indicators for Me_cus, Figures S11b and S15a,b) and higher reducibility of Ce_0.6Zr_0.4O₂ in s-28CZA-pH 13 (Figure S14).

The developed s-3VA-pH 13 and s-28CZA-pH 13 catalysts also show high long-term stability. No obvious deactivation was observed during 45 h on EB stream under industrially relevant conditions (Figure 4b and Figure S19a,b). We also benchmarked our catalysts against the state-of-the-art V- or Ce-based catalysts used in the CO₂−ODEB reaction at 550 °C and comparable contact times (Table S5). For this purpose, the space-time conversion of EB (STC) defined as the molar amount of EB converted per mole of V or Ce per hour and relative deactivation rate (R_d) expressed as the ratio of (STC_initial−STC_final)/STC_initial×100 % (STC_initial and STC_final stand for STC at the beginning and at the end of the reaction, respectively) were calculated. The developed s-3VA-pH 13 and s-28CZA-pH 13 show the highest initial STC and the lowest R_d among their previously reported counterparts (Figure 4c). Moreover, the high stability of the developed catalysts is further confirmed by the deactivation rate constant (k_d, Table S6) calculated on the basis of a first-order deactivation model.²⁹

Time-resolved in situ UV/Vis spectroscopy was applied to check if redox kinetics of supported VO_x species and bulk CeO₂ can be a descriptor in interpreting their catalytic activity for CO₂−ODEB. Following previously reported studies,³⁰ H₂ was used as a reducing agent instead of EB to exclude any potential effect of carbon-containing deposits on reaction-induced changes in the UV/Vis spectra. Figure 4d exemplarily shows the spectra (expressed as the Kubelka–Munk function, F(R)) of fully oxidized s-3VA and after different times on H₂ stream (10 %vol H₂/N₂) at 550 °C. The former spectrum is characterized by the oxygen ligand to metal charge transfer (LMCT) bands of V⁵⁺ with the absorption maxima below 400 nm.³¹ The intensity decreases with rising time on H₂ stream due to the reduction of V⁵⁺ to V⁴⁺/V³⁺. The presence of the latter species is supported by the appearance of absorption bands in the range of 600–800 nm, which are assigned to d-d transitions of V⁴⁺/V³⁺.^30a Their intensity remained unchanged after about 10 min in H₂ due to reaching a steady-state concentration of such VO_x species. Similar changes were also observed in the spectra of Ce-based catalysts (Figure S20).

The apparent rate constant for the reduction of oxidized VO_x species in H₂ (k′(H₂)) was calculated through fitting the temporal changes in the F(R) function (Δ F(R)) at 700 nm with a simple kinetic model shown in equation 1. Details for deriving this equation is given in the Supporting note in the Supporting Information.

(1)

where is the concentration of reduced VO_x species, is the total concentration of oxidized VO_x species, which participate in the reduction.

The simulated responses and the experimental ΔF(R) at 700 nm for s-3VA are presented in Figure 4e, while such data for s-3VA-pH 9, s-3VA-pH 11 and s-3VA-pH 13 are given in Figure S21a–c. The model in equation 1 correctly describes the experimental data for all catalysts. The derived k′(H₂) values of s-3VA, s-3VA-pH 9, s-3VA-pH 11 and s-3VA-pH 13 are 0.66, 0.75, 0.83 and 0.96 min⁻¹, respectively. There exists a positive correlation between the TOF_EB and k′(H₂) as seen in Figure 4f. Thus, the post-treatment of the catalyst in the NH₃ ⋅ H₂O solution accelerates the kinetics for the reduction of oxidized VO_x species, which is beneficial for the conversion of EB to styrene. A similar conclusion was also established for s-CZ versus s-28CZA-pH 13 (Figures S18 and S22a,b).

Supported VO_x catalysts are also promising candidates for the non-oxidative dehydrogenation of lower alkanes such as propane.^{8b, 32} V⁴⁺/V³⁺O_x species are considered to be the active sites in the PDH reaction.³³ To check if their intrinsic activity can be regulated through altering the V−O-support interactions, the developed V-containing catalysts were tested in the PDH reaction at 600 °C using an industrially relevant feed consisting of 40 vol %C₃H₈ in N₂. The catalysts were initially reduced in a flow of 50 vol % H₂/N₂ to remove active lattice oxygen species and thus to exclude the oxidative dehydrogenation of propane during a few first minutes on stream.^33c Moreover, coke deposition over VO_x catalysts for PDH is resisted by V−OH species formed during the H₂ pre-treatment.³⁴ As indicated by the results in Figure 5a, the apparent TOF_propane increases in the order s-3VA<s-3VA-pH 9<s-3VA-pH 11<s-3VA-pH 13 while the selectivity to propene is about 95.0 % for all catalysts (Figure S23). Since the molecular structure of exclusively monomeric VO_x is essentially same in all catalysts, the enhanced activity of post-treated catalysts should be attributed to weakened interactions between VO_x and Al₂O₃ as suggested in references.^{33a, 35} For the catalysts with too weak interactions between active sites (e.g., Pt) and support, restructuring may occur during the reaction at high temperatures, leading to the loss of activity.^{16a, 17a, 36} Thus, an additional test consisting of 12 PDH/oxidative regeneration cycles was performed over s-3VA-pH 13. No obvious loss in the propane conversion and propene selectivity could be identified from cycle to cycle (Figure 5b). Thus, VO_x species are not restructured even under severe PDH conditions, which is consistent with the excellent durability of the impregnated VO_x/γ-Al₂O₃ catalyst for the PDH reaction as reported in a very recent study.³⁷ Moreover, as indicated from the apparent TOF_propane of s-28CZA and s-28CZA-pH 13 (4.2 versus 8.4 h⁻¹, Figure S24), the post-treatment of s-28CZA also improves its activity for the PDH reaction.

To derive fundamental insights into the PDH reaction over s-3VA, s-3VA-pH 9, s-3VA-pH 11 and s-3VA-pH 13, we carried out pulse experiments with C₃H₈ or H₂/D₂ in the temporal analysis of products (TAP) reactor at 600 °C in high vacuum. The catalysts were reductively treated to mimic the reduction state of vanadium as in the steady-state PDH tests. Propene and hydrogen were detected after pulsing of a C₃H₈/Ar=1/1 mixture (Figure 5c and Figure S25a–c). The catalytic activity in terms of the propene formation can be ordered as s-3VA<s-3VA-pH 9<s-3VA-pH 11<s-3VA-pH 13 (Figure S25d), which is same to that of the steady-state PDH tests (Figure 5a). This further confirms the improved activity of the post-treated catalysts for activating C−H bond in propane.

The time scale of gas-phase components in Figure 5c and Figure S25a–c was converted into a dimensionless form as suggested in the reference.³⁸ The details can be found in the Supporting Information. Such transformation is required for correct comparing the order of appearance of C₃H₈, C₃H₆ and H₂ since their diffusion velocity is strongly different. As expected, the responses of C₃H₆ and H₂ as products appear after the response of the C₃H₈ reactant. However, the maximum of the dimensionless time of the C₃H₆ response is obviously shorter than that of the H₂ response (Figure 5c and Figure S25a–c). This indicates that the rates for the formation of C₃H₆ and H₂ are not equal under transient conditions, which is different from the steady-state observations. It is commonly accepted that the PDH reaction occurs via the consecutive activation of two C−H bonds in one propane molecule, leading to the formation of adsorbed C₃H₆ and H₂ on the catalyst surface. After desorption, gaseous C₃H₆ and H₂ are released. Thus, based on the position of the maxima of the C₃H₆ and H₂ responses, it can be proposed that the rate-limiting step in the VO_x-catalyzed PDH reaction is the H₂ formation but not the cleavage of C−H bonds in C₃H₈. This statement is further supported by the results of temperature-programmed surface reaction of propane (TPSR-C₃H₈) over s-3VA-pH 13 (Figure S26). The staring temperature of H₂ appearance in the gas phase (350 °C) is clearly higher than that of C₃H₆ (300 °C).

To check if the delayed release of H₂ in the C₃H₈ pulse tests could be caused by its potential reactions with surface OH groups, hydrogen isotopic exchange experiments were carried out using a H₂/D₂/Ar=1/1/1 mixture. As HD was observed in outlet gases (Figure 5d and Figure S27a–c), the breaking of D−D bonds in D₂ molecules and the formation of H−D bonds happened over all of the catalysts. Moreover, the order of catalyst activity for the H/D exchange reaction (Figure S27d) is same to that for the PDH reaction over the investigated catalysts (Figure S25d). Irrespective of the catalysts, the dimensionless time for releasing gaseous HD (Figure 5d and Figure S27a–c) is much shorter than that for the appearance of gaseous H₂ (Figure 5c and Figure S25a–c). This clearly indicates that the delayed appearance of H₂ in comparison with C₃H₆ cannot originate from the secondary transformation and desorption of H₂ in the course of the PDH reaction. Thus, the formation of H₂ after the activation of C−H bonds in propane is reasonably proposed as the rate-determining step in the VO_x catalyzed PDH reaction. Considering this conclusion and the PDH activity in Figure 5a, we put forward that the strength of V−O−Al bond affects the ability of reduced isolated VO_x species to form gas-phase H₂ through recombination of two surface H species originated upon the breaking of C−H bonds in propane.

Conclusion

In summary, we have demonstrated that a simple post-treatment of VO_x−Al₂O₃ and CeO₂−ZrO₂−Al₂O₃ in an aqueous NH₃ solution is an industrially applicable method to tailor structural and redox properties of these catalysts, which are determined by dissolution/precipitation/recrystallization processes of amorphous or less crystalline metal oxides in the treatment solution with varied pH values. The accordingly developed VO_x-containing or Ce-based catalysts outperform the corresponding state-of-the-art catalysts in terms of their activity in the CO₂−ODEB reaction. For VO_x−Al₂O₃ catalysts, such treatment enables control of the concentration of Al_cus sites, which determine the reducibility of isolated V⁵⁺ sites. This catalyst property is a key descriptor affecting the rate of styrene formation as well as the rate of the PDH reaction. The reducibility of bulk Ce-containing catalysts can be controlled by the developed method and governs their activity in the above-mentioned reactions too. The established strategy for tuning redox properties of vanadia and ceria can be used for purposeful preparation of other metal oxides for catalytic or other material-property-related applications.

Conflict of interest

The authors declare the following competing financial interest. A patent (202110110652.0) partially based on the results of this work have been issued to Shaanxi Normal University.