Introduction

Propylene is a vital building block for producing a wide range of high-value-added chemicals. Propane dehydrogenation (PDH) represents an on-purpose technology that produces propylene1,2. However, the reaction is intensely endothermic3,4. Oxidative dehydrogenation of propane (ODHP) using O2 as oxidant has been proposed to overcome equilibrium limitations, but it suffers from overoxidation5,6, giving yields below the PDH equilibrium7,8,9. Carbon dioxide (CO2), a greenhouse gas, can act as a soft oxidant during PDH to lower ΔG for the reaction and make the reaction feasible at lower operating temperature10,11. This not only facilitates the reaction but also presents a promising strategy for mitigating CO2 emissions12. Meanwhile, in the CO2-ODHP system, carbon deposition can be eliminated through the reverse Boudouard reaction (CO2 + C = 2CO). Although desirable, this reaction has proven challenging, as most catalysts for CO2-ODHP tend to undergo severe dry reforming side reactions13.

Previous studies focused on the design of bifunctional catalysts for the coactivation of propane and CO214,15,16,17. Reducible metal ions readily donate abundant lattice O atoms to allow efficient turnover due to their redox and acidic nature. However, the oxides possess limited ability for C-H cleavage and exhibit short life cycles10. Comparatively, Xing et al. reported a Pt-Co-In ternary nanoalloy that demonstrates high C3H6 selectivity and CO2 utilization18. However, the alloy structure’s inability to distinctly segregate the C-H bond activation sites from the C = O bond activation sites results in competitive adsorption and promotes dry reforming of propane (DRP). The similar results were reported by Zhai et al.19 on PtSn/SiO2 catalysts that highlight the metal-oxide interface (i.e., Pt3Sn-SnOx) as active sites for propane dehydrogenation and reverse water gas shift (RWGS) reaction. However, the mechanistic understanding on the diffusion of reactive intermediates between tandem catalytic sites remains challenging20. At the process engineering level, the utilization of membrane reactors and chemical looping is anticipated to bolster propylene selectivity through the physical or temporal segregation of dehydrogenation and CO2 hydrogenation21,22,23,24. However, the implementation of these technologies poses certain challenges in scaling up reactor and processes.

Herein, we develop a coupling process that integrates PDH and RWGS and demonstrate the thermodynamic/kinetic matching of PDH and RWGS steps by regulating the spillover of hydrogen intermediates over the designed tandem catalyst (Fig. 1a). The well-established PDH catalyst, PtSn/Al2O3 provides the dehydrogenation sites, while adjacent CeOx/Al2O3 at the nanoscale acts as the hydrogenation sites for subsequent RWGS reaction. We achieved a high propylene selectivity of 98.8%, accompanied by high CO2 (44.0%) and propane conversion (44.2%) at 550 °C surpassing that via direct PDH. Based on catalytic experiments, spectroscopic characterization, and kinetic studies, we propose a hydrogen spillover-mediated coupling mechanism. The hydrogen species generated at the PtSn/Al2O3 sites migrated to adjacent CeOx/Al2O3 for the RWGS reaction, thereby driving PDH towards propylene production. This mechanism is favored by the proximity between the dehydrogenation and hydrogenation sites.

Fig. 1: Tandem systems for PDH-CO2 hydrogenation process.
Fig. 1: Tandem systems for PDH-CO2 hydrogenation process.
Full size image

a Tandem catalysts in different modes: T1 (dual-bed): Upper bed: Pt1Sn7/Al2O3 (20–40 mesh, 0.35–0.833 mm), lower bed: CeOx/Al2O3 (20–40 mesh); T2 (granule-mix): Physical mixture of Pt1Sn7/Al2O3 and CeOx/Al2O3 (both 20–40 mesh); T3 (powder-mix): Pt1Sn7/Al2O3 and CeOx/Al2O3 separately ground to 0.05–0.075 mm, physically mixed, tableted and re-sieved to 20–40 mesh; T4 (mortar-mix): Pt1Sn7/CeOx-Al2O3 tandem catalyst, sieved to 20–40 mesh. b The representative AC-HAADF-STEM image and (c) HAADF-STEM images of Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4). d Ce 3 d XPS spectra of Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4). and e CO2-TPD profiles of Pt1Sn7/Al2O3, CeOx/Al2O3 and Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4). The a.u. stands for arbitrary units.

Results and Discussion

Design of tandem catalysts

We initially performed thermodynamic calculations for DRP, PDH, direct and indirect pathway of CO2-PDH (Supplementary Fig. 1)25,26. At 550 °C, the equilibrium conversion of CO2 for DRP can reach up to 99.8%, while no C3H6 is formed. The direct pathway of CO2-PDH can reach an equilibrium conversion of C3H8 of 33.0% as opposed to 56.8% for PDH. Comparatively, the tandem reaction of CO2 with C3H8 (indirect pathway) increases the equilibrium conversion of C3H8 to 63.9% by consuming H2 through the RWGS reaction.

Based on the thermodynamic analysis, we investigated the different tandem configurations of dehydrogenation catalysts and RWGS catalysts within a single reactor (Fig. 1a). Given the established advantages of supported PtSn27 and CeOx28 have been in activation of propane and CO2, respectively, these were chosen as the PDH and RWGS catalysts for this study. The tandem models, labeled T1 to T4, correspond to different contact distances between the catalysts, ranging from millimeters to nanometers29.

In tandem mode T1 (dual-bed), PtSn/Al2O3 and CeOx/Al2O3 pellets are loaded into the reactor by means of upper and lower bed filling. In T2 (granule-mix), the two catalyst pellets are mixed and loaded into the reactor. In T3 (powder-mix), the two catalysts are fully milled into a fine powder, pressed, and sieved to produce pellets for reactor loading. In T4 (mortar-mix), nanoscale contact between PtSn nanoparticles and CeOx/Al2O3 is achieved by impregnating a mixed solution of Pt and Sn precursors onto CeOx/Al2O3, resulting in a composite catalyst denoted as PtSn/CeOx-Al2O3.

The PtSn/CeOx-Al2O3 tandem catalyst (T4) was characterized in details. X-ray diffraction (XRD) patterns (Supplementary Fig. 2) and aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) images revealed the presence of oxygen-deficient CeOx within the T4 tandem catalysts (Fig. 1b). This was further corroborated by the formation of a significant density of Ce3+ species, as evidenced by X-ray photoelectron spectroscopy (XPS)30,31 (Fig. 1d), Additionally, highly dispersed PtSn nanoparticles (Fig. 1b, c), measuring approximately 2.0 ± 0.5 nm (Supplementary Fig. 3b–g), adorned the surface32,33. The CO2-TPD profiles (Fig. 1e) exhibited CO2 desorption peaks at approximately 316 °C and 480 °C on CeOx/Al2O3 and PtSn/CeOx-Al2O3, respectively, indicating the presence of additional activation sites for CO2 on the PtSn/CeOx-Al2O3 tandem catalyst compared to the PtSn/Al2O3 catalyst34,35. Furthermore, we confirmed that the CO2 chemical adsorption capacity is significantly enhanced with increasing Ce3+ concentration, which verifies that oxygen vacancies might serve as the primary sites for CO2 activation (Supplementary Fig. 4).

Proximity-dependent propylene selectivity and formation rates

We demonstrated the application of tandem catalysts in a continuous process scheme for PDH followed by CO2 hydrogenation. The outlet gas was analyzed and found to contain CH4, C2H6, C2H4, C3H6, CO, H2O, and H2 (Supplementary Fig. 5). When using PtSn/Al2O3 catalysts, the CO2 conversion remained relatively unchanged as the molar ratio of Sn to Pt was varied (Supplementary Fig. 6a). However, the contribution of dry reforming, in terms of CO production from hydrocarbons, exhibited a reverse volcano peak relationship, reaching a minimum value of 20.3 mmol·gcat−1·h−1 at an optimized Sn to Pt ratio of 7 (Supplementary Fig. 6b).

Inter-particle distances were computationally analyzed using MATLAB with Monte Carlo method (Supplementary Fig. 7 and Supplementary Table 1). A granular mixture (T2) of CeOx/Al2O3 and PtSn/Al2O3 resulted in a decreased CO production from hydrocarbons (10.3 mmol·gcat.−1·h−1) and a higher propylene production rate (24.3 mmol·gcat.−1·h−1) compared to Pt1Sn7/Al2O3 (Fig. 2a, b and Supplementary Fig. 8). For the PtSn/CeOx-Al2O3 tandem catalyst (T4), which exhibited nanoscale intimacy, the propylene formation rate was further enhanced to 31.3 mmol·gcat.−1·h−1, with the lowest CO production from C3H8 (0.87 mmol·gcat.−1·h−1) and the highest propylene selectivity in CxHy (C3H6, CH4, C2H6, C2H4) of 98.8%. The strong correlation between the proximity of PtSn/Al2O3 and CeOx/Al2O3 and the propylene selectivity suggested effective communication between the tandem catalysts. This result aligns with the notion that “the closer-the better” in terms of catalyst intimacy36.

Fig. 2: Catalytic performance of tandem catalysts.
Fig. 2: Catalytic performance of tandem catalysts.
Full size image

a, b Catalytic performance at different distances between PtSn/Al2O3 and CeOx/Al2O3 at 500 °C. c Initial conversion of C3H8 and CO2 at 15 minutes of PDH and CO2-PDH at 500 ⁰C and 550 ⁰C on Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4). Catalytic testing conditions of (a–c): integral reactor, atmospheric pressure, 150 mg catalyst, WHSV of propane = 4.6 h−1, C3H8: CO2: N2 = 5: 3: 27 mL·min−1. d RWGS on Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4) at 550 ⁰C. Catalytic testing conditions: atmospheric pressure, 150 mg catalyst, CO2: H2: N2 = 6: x: (29-x) mL·min−1. e Cyclic stability testing with Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4), regeneration: 45 min regeneration at 500 ⁰C in flowing 20% O2/N2 (35 mL·min−1), and 1 h reduction at 500 ⁰C in flowing 40% H2/N2 (35 mL·min−1).

We further demonstrated the superiority of the tandem catalysts in PDH-CO2 hydrogenation when compared to direct PDH. At 500 °C and WHSV of propane (4.6 h−1), the initial propane conversion was 31.7%, which was higher than that observed for the direct PDH reaction (26.5%) (Fig. 2c). Notably, the CO2 conversion exceeds the equilibrium conversion at the corresponding temperatures (Supplementary Fig. 1: 25.8% at 500 °C, 35.4% at 550 °C). This deviation might be attributed to the additional consumption of CO2 by the DRP, However, as shown in Supplementary Table 2, the experimented ratios of CO/CO2 and CO/C3H8 are closer to the thermodynamic equilibrium values when the PDH-RWGS pathway dominates, and the higher C3H6 to H2 ratio and lower H2 partial pressure at the outlet in CO2-PDH (1.7) versus PDH (0.9) at 500 °C (Supplementary Fig. 9a), meanwhile, the relevant methanation reactions (i.e. CH4 formation rate) are essentially negligible (Supplementary Fig. 9b, c), which further indicated the in-situ consumption of H2 through the PDH-RWGS pathway.

Although the reaction primarily follows the PDH-RWGS pathway, the reaction network exhibits significant temperature dependence, as shown in Supplementary Fig. 10a, b, a systematic evaluation of the temperature ranges from 500 to 600 °C revealed that the conversions of C3H8 and CO2 significantly increase due to endothermic nature of this reaction. However, accelerated catalyst deactivation at higher temperatures leads to a decline in propane conversion. Meanwhile, propylene selectivity gradually decreases with rising temperature, which is attributed to the thermodynamic favorability of the dry reforming reaction (C3H8 + 3CO2 → 6CO + 4H2, ΔH = +620 kJ/mol). As illustrated in Supplementary Fig. 10c, the partial pressure ratio of C3H6 to H2 at the reactor outlet decreases progressively with increasing temperature, confirming that dry reforming or coke formation competitively consumes propane and suppresses propylene yield.

To investigate the catalyst deactivation mechanism, RWGS tests were conducted within the different hydrogen partial pressures observed during the C3H8 and CO2 reaction (Fig. 2d). The PtSn/CeOx-Al2O3 tandem catalysts (T4) exhibited remarkable stability, which suggests that catalyst deactivation is not caused by restructuring induced by CO2. Rather, carbon deposition, likely resulting from the in-situ removal of H2 during PDH, is the primary cause of catalyst deactivation (Supplementary Fig. 11). Notably, the PtSn/CeOx-Al2O3 tandem catalyst fully recovered and maintained its initial C3H8 conversion, C3H6 selectivity (including hydrocarbon and net selectivity), and CO2 conversion after undergoing three consecutive cycles (Fig. 2e, Supplementary Fig. 12). Furthermore, experimental results under different propane concentrations indicate that propane conversion exhibits a non-monotonic decline with increasing propane concentration (Supplementary Fig. 13), while propylene selectivity remains unchanged, even at a higher pre-reduction temperature of 600 °C, the catalyst demonstrates stable performance and regeneration capability (Supplementary Figs. 14, 15). These findings further improve its adaptability potential for industrial applications.

The role of hydrogen spillover in coupling PDH and RWGS

To investigate the promotion effect of CO2 on PDH, catalysis studies focusing on C-H bond activation were carried out. For the C3H8 temperature-programmed surface reaction (TPSR), the similar initial activation temperatures observed on PtSn/Al2O3 (279 °C) and PtSn/CeOx-Al2O3 tandem catalysts (282 °C) indicate that the tandem system does not alter the catalyst’s ability to activate C-H bonds (Supplementary Fig. 16). This similarity is further supported by the comparable geometric and electronic states of Pt in both PtSn/Al2O3 and PtSn/CeOx-Al2O3 tandem catalysts, as evidenced by the similar CO adsorption band at 2083 cm−1 on Pt top sites, as detected by CO-diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) (Supplementary Fig. 17a, b)37. Furthermore, we demonstrated that the CO adsorption band on PtSn/CeOx-Al2O3 catalyst showed negligible changes before and after regeneration across different reactions (PDH, CO2-PDH, and RWGS, Supplementary Fig. 17c).

We noted that when C3H8 and CO2 were co-fed into the reactor, the temperatures for CO2 consumption (306 °C) and CO formation (350 °C) were higher than the initial C-H activation temperature of propane (287 °C) (Fig. 3a). Furthermore, isothermal pulse experiments conducted at 500 °C by continuously switching between 5% C3H8/95% Ar and 5% CO2/5% C3H8/90% Ar atmospheres (Fig. 3b) revealed that upon introduction of CO2, the intensities of C3H8 and H2 decreased, while the intensities of C3H6 increased immediately, accompanied by the formation of CO and H2O. These observations suggest that C-H bond activation precedes the activation of the C=O bond, leading to the generation of CO and H2O. We hypothesize that H*, the product of C-H activation, serves as a crucial intermediate that links the PDH and RWGS reactions through a surface diffusion mechanism.

Fig. 3: Evidence of hydrogen spillover–coupling mechanism of PDH with RWGS on Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4).
Fig. 3: Evidence of hydrogen spillover–coupling mechanism of PDH with RWGS on Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4).
Full size image

a TPSR in CO2-PDH reaction atmosphere on fixed-bed reactor. b Pulse experiment in PDH and CO2-PDH reaction atmosphere. c, d DRIFTS spectra of 10% H2/Ar. e In-situ DRIFTS spectra of CO2-PDH reaction atmosphere. f Schematic illustration of the reaction mechanism of PDH-RWGS on PtSn/CeOx-Al2O3 (T4) in PDH-CO2 hydrogenation process. The a.u. stands for arbitrary units.

We then performed H2 temperature-programmed reduction (H2-TPR) and H2 temperature-programmed desorption (H2-TPD) to investigate the spillover of hydrogen species. PtSn sites can dissociate H2 to generate active hydrogen atoms (H*) (the first reduction peak at 199–232 °C), which spillover to the CeOx surface, significantly reducing the reduction temperature of CeOx. As shown in Supplementary Fig. 18a and Supplementary Table 3, among different configurations, as the PtSn-CeOx interfacial distance progressively decreases (from T1 to T4), the closely contacted mortar-mix (T4) exhibits the lowest reduction temperature (346 °C), whereas the physically separated dual bed (T1) relies on gaseous hydrogen transfer, resulting in a higher reduction temperature (491 °C). Notably, in configurations T1 to T3, the Pt source responsible for the initial H2 dissociation is the same Pt1Sn7/Al2O3 catalyst. In addition, the Pt1Sn7/CeOx-Al2O3 and Pt1Sn7/Al2O3 catalysts exhibit identical particle size distributions (Supplementary Fig. 19) and share the same Pt coordination environment. This ensures consistent morphology of the Pt species when comparing the reduction behaviors across T1 to T4. Therefore, the observed differences in reduction temperatures can be attributed to variations in the spatial configuration between the PtSn active sites and the CeOx reduction sites38. Moreover, as the inter-site distance narrows, hydrogen consumption gradually increases from T1 (0.27 cm3/gcat.) to T4 (1.62 cm3/gcat.), which further demonstrates that the hydrogen spillover effect promotes reduction of the CeOx. For H2-TPD (Supplementary Fig. 18b), on mortar-mix (T4) mode, the first H2 desorption peak (~42 °C) is attributed to the H* on the Pt, and the second H2 desorption peak (~287 °C) is attributed to the spillover hydrogen on the interface between PtSn particles and CeOx/Al2O3, and the other one at higher temperature (>600 °C) is attributable to H* in the subsurface region and spillover hydrogen on CeOx/Al2O3 support39,40,41. Qualitative analysis revealed that as the distance between PtSn particles and CeOx/Al2O3 decreases, the intensity of desorption peaks of spilled hydrogen gradually increases, further demonstrating the importance of site proximity for the hydrogen spillover.

Previous results demonstrated that on pure CeO2, H2 primarily undergoes heterolytic dissociation, with H2O formation limited by converting heterolytic to homolytic products42. In Pt/CeO2, H2 dissociates homolytically on Pt, followed by rapid hydrogen spillover, leading to approximately double the OHads- coverage compared with pure CeO243,44,45,46. The DRIFTS spectra under H2 flow showed a simultaneous increase in [OHads-] (3000–3500 cm−1) and the forbidden 2F5/2-to-2F7/2 electronic transition of Ce3+ (2130 cm–1) (Fig. 3c, d and Supplementary Fig. 20). This correlation indicates that PtSn promotes the spillover of homolytic dissociation hydrogen species onto the CeOx surface, as described by the determined d[OHads-]/dt = k × d[Ce3+]/dt (where k is a positive constant)47,48,49,50. In contrast, without PtSn nanoparticles, the DRIFTS profile of the CeOx/Al2O3 under the same conditions showed that [Ce3+] increased continuously with H2 reduction time (Supplementary Fig. 21), while [OHads-] increased only within 10 minutes and remained essentially unchanged. This provides direct evidence that Pt facilitates the spillover of dissociated hydrogen species onto the CeOx surface50.

In-situ DRIFTS was further performed to monitor the evolution of intermediates during the coupling reaction of PDH and RWGS. As shown in Fig. 3e and Supplementary Figs. 22, 23, for PtSn/CeOx-Al2O3 tandem catalysts, characteristic bands assigned to gaseous C3H8 (2810–3060 cm−1) were detected in the initial stage during PDH reaction51,52. Upon switching the feed from C3H8/Ar to C3H8/CO2/Ar, CO adsorbed on Pt at 2045 cm−1 was detected together with H2O (3470–3580 cm−1), Ce-OH (3625 cm−1) and Al-OH (3665–3770 cm−1)53,54,55, respectively, suggesting the RWGS occurs during the reaction. It is noteworthy that the characteristic peaks of the HCOO* intermediate (1590 cm−1, 1395 cm−1, 1375 cm−1) were observed simultaneously. Previous research showed that alternative intermediates such as *CH3CH2CH2, *CH3CHCH2, and *CH3CH2 could react with *O derived from CO2 to form *CH3CH2CH2O, *CH3CH2CHO, and *H2O instead of involving the formation of formate acting as hydrogen donors56,57,58. Therefore, the formation of formate species should originate from the reaction between CO2 and dissociated hydrogen species (H*) derived from C–H bond cleavage of propane. These results suggest that a coupled PDH-RWGS pathway is dominated, and H2 produced by PDH on PtSn/Al2O3 can spillover to the adjacent CeOx/Al2O3 (Fig. 3f), which results in a subsequent RWGS reaction.

Kinetic studies and KIE experiments

We performed a kinetic study using a differential reactor to understand in depth the functions of tandem catalysis with the exclusion of internal and external diffusion and temperature gradients. More detailed experimental details are given in the supplementary information (Kinetic analysis of CO2-PDH, PDH and RWGS, Supplementary Figs. 24–30, Supplementary Tables 4–8). The initial reaction rates were obtained by correcting the equilibrium and deactivation59.

Across a broad range of partial pressures of C3H8 (0–30 kPa), and H2 (0–40 kPa), the tandem systems achieved the higher propylene formation rates when co-feeding CO2 than that of PDH (Fig. 4a–c). A kinetic analysis was conducted to understand the rate-determining step (RDS) of the PDH and CO2-PDH (see Supplementary Information: Kinetic analysis of PDH and CO2-PDH section). Prior to analyzing the CO2-PDH reaction system, kinetic studies of the RWGS reaction over the Pt1Sn7/CeOx-Al2O3 catalyst, combined with theoretical analyses based on Langmuir-Hinshelwood (L-H) and Mars-van Krevelen (MvK) mechanistic assumptions, indicate that the CO2 + H* reaction in this work follows the L-H mechanism (see Supplementary Information: Kinetic analysis of RWGS section). So we considered a L-H mechanism involving a two-step dehydrogenation of C3H8 to C3H6. Subsequently, CO2 is reduced via hydrogen derived from propane dehydrogenation through the formate pathway to produce CO, with the hydrogen eliminated in the process being oxidized to form water18. We solved the rate equation of each step by applying the quasi-steady-state approximation and site conservation conditions60.

Fig. 4: Kinetics of hydrogen spillover–coupling mechanism of PDH with RWGS.
Fig. 4: Kinetics of hydrogen spillover–coupling mechanism of PDH with RWGS.
Full size image

a–c CO2-PDH and PDH under different C3H8, CO2 and H2 partial pressures. The dash lines are to guide the eyes. Catalytic testing conditions: 500 ⁰C, differential reactor, atmospheric pressure, 50 mg Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4). a C3H8: 7–28 kPa, CO2: 8.6 kPa; (b) C3H8: 9 kPa, CO2: 0–34.4 kPa; (c) C3H8: 14 kPa, CO2: 8.6 kPa, H2: 0–44.4 kPa, 35 mL·min−1 total flow rate balanced with N2 (detail data from Supplementary Figs. 26–30). d Comparison of predicted (from Supplementary Eq. S(63)) and experimental C3H6 formation rates. e KIEs of CO2-PDH on Pt1Sn7/CeOx-Al2O3 tandem catalyst (T4) (detail data from Supplementary Fig. 32). f Proposed reaction processes of hydrogen spillover-coupling mechanism of PDH with RWGS.

The experimental reaction order with respect to C3H8 was found to be 1.24 for CO2-PDH and 1.18 for PDH (Fig. 4a), whereas the reaction order with respect to CO2 was determined to be 0.44 in the range of 0–8.6 kPa, and close to 0 in the range of 8.6–34.4 kPa (Fig. 4b). This suggests that at low CO2 partial pressures (<8.6 kPa), the hydrogen generated from PDH is efficiently consumed by the tandem reaction, resulting in enhanced propylene production. However, since the reaction exhibits zero-order kinetics with respect to CO2 when its partial pressure exceeds 8.6 kPa, further increasing the CO2 partial pressure does not enhance the formation rate of C3H6 reaction rate. Furthermore, the negligible influence of co-feeding CO with C3H8/CO2 rules out the possibility of CO inhibition, likely due to its limited surface coverage at high temperatures (Supplementary Fig. 31). The predicted rates were in agreement with the experimental data (Fig. 4d) only when the second C-H cleavage of propane was considered as the RDS for CO2-PDH (Supplementary Fig. 24).

We further investigated the tandem reaction mechanism with kinetic isotopic experiments (KIE). As shown in Fig. 4e, a normal KIE value of 2.7 (\({r}_{H}\)/\({r}_{D}\)) was observed when switching from C3H8 to C3D8 after reaching the steady state on PtSn/CeOx-Al2O361, while KIE value equals to 0.99 (\({r}_{{12}_{C}}\)/\({r}_{{13}_{C}}\)) by switching 12CO2 to 13CO2, which validated that the kinetic relevance of C-H activation was the RDS62,63. The apparent activation energy (Ea) was estimated for PtSn/Al2O3 and PtSn/CeOx-Al2O3 tandem catalyst by Arrhenius-type plots. The Ea of PDH has essentially the same values (84.4 and 85.4 kJ∙mol−1) for both catalysts (Supplementary Fig. 33). This suggests that the incorporation of CeOx did not alter the activation capacity of PtSn/Al2O3 for C-H bonds, which is also consistent with previously reported results for C3H8-TPSR and CO-DRIFTS. However, the Ea of CO2-PDH was significantly reduced (from 96.7 to 73.6 kJ·mol−1) by the incorporation of CeOx into Pt1Sn7/Al2O3, which validates that the tandem process can promote the dehydrogenation during the co-feeding of CO2 with C3H8.

Related tandem catalysts

To extend the coupling of PDH and RWGS through hydrogen spillover, we investigated tandem catalysts from two perspectives: by modifying either the PDH catalysts (PtSn/Al2O3, GaOx/Al2O3, CrOx/Al2O3) or the CO2 hydrogenation catalysts (CeOx/Al2O3, Cu/Al2O3, TiOx/Al2O3) in intimate tandem configurations, as mentioned earlier (Supplementary Fig. 34). As anticipated, we observed enhanced propylene production rates compared to the parent rates without tandem coupling. These findings support the hydrogen spillover-mediated coupling mechanism in a tandem setup, emphasizing the importance of proximity between the two reaction sites.

Analysis of CO2 footprint

The superiority of PDH-CO2 hydrogenation tandem process to PDH implies the potential application for propylene synthesis25,64,65. To analyze the additional separation processes introduced by CO2 in the PDH-CO2 hydrogenation tandem process compared to conventional PDH, we incorporated two additional PSA units to achieve H2/CO and CO2/C2 separation functionalities. The separated CO2 is recycled back into the feed stream. We then compared the net CO2 emissions and energy consumption for both (Supplementary Tables 9–13 and Supplementary Figs. 35, 36). The PDH-CO2 hydrogenation tandem process for producing C3H6 requires more energy (21.09 GJ/ton of propylene) compared with the Oleflex PDH process (17.78 GJ/ton of propylene). However, due to the increased consumption of CO2 by the PDH-CO2 hydrogenation tandem process, its net CO2 emissions are approximately 0.69 ton/ton of propylene, which is much lower than that of the Oleflex PDH process (1.23 ton/ton of propylene). This indicates that PDH-CO2 hydrogenation tandem process is a promising strategy for carbon utilization to mitigate CO2 emissions. Future advancements might focus on enhancing catalyst efficiency or integrating co-conversion reactions where CO2-PDH products directly feed downstream reactions, which might enable bypassing complex and energy-intensive separation processes.

To conclude, we propose the coupled pathway of PDH-RWGS over tandem catalysts consisting of dehydrogenation sites, i.e. PtSn/Al2O3 and CO2 hydrogenation sites, i.e. CeOx/Al2O3. This tandem configuration effectively mitigates the competitive dry reforming reaction associated with PDH, resulting in a high propylene selectivity of 98.8%. In-situ characterizations and kinetic analysis have revealed that hydrogen (H) serves as a crucial intermediate in coupling PDH and RWGS. The industrial manufacture of tandem catalysts with atomic-level control of different active sites needs substantial development to provide this technology for the olefin industry.

Methods

Materials properties and synthesis

All the catalysts were prepared by incipient wetness impregnation method. Methods of CeOx/Al2O3(10 wt.% CeO2) preparation based on Xie’s work66, In details, the Ce(NO3)3·6H2O (99.0%, Aladdin biological technology Co., Ltd. (Shanghai, China)) solution with determined concentration was added dropwise onto γ-Al2O3 (Adamas-beta reagent Co., Ltd. (Shanghai, China)) under stirring. The obtained wet powder was dried at 120 °C for 1 h and then calcined in air at 550 °C for 2 h. Subsequently, the CeOx/Al2O3 support was obtained by reducing the base support in 10% H2/Ar flow at 750 °C for 2 h with a temperature ramp of 5 °C/min. For the synthesis Pt1Snx/CeOx-Al2O3 (the loading of Pt atoms is fixed at 0.15 wt.% into CeOx/Al2O3, and x = 1, 3, 5, 7, 10, 15. Here x represents the molar ratio of Sn/Pt), H2PtCl6·6H2O (Chemart (Tianjin) Chemical Technology Co., Ltd, 99.9%) and SnCl2·2H2O (Chemart (Tianjin) Chemical Technology Co., Ltd, 99.9%) were added to a 0.1 M HCl solution and used as precursors. The precursors solution with determined concentration of Pt and Sn was dropwise onto CeOx/Al2O3. After impregnation, the catalysts were placed in the atmosphere statically overnight and then dried in the flowing air at 80 °C for 12 h and then calcined at 600 °C for 2 h.

Catalytic testing

The catalytic testing was performed in a quartz fixed-bed reactor with 8 mm inner diameter and 24 cm length at atmosphere pressure. A mixture of 150 mg catalysts and 0.1 mL quartz sand with 20–40 mesh-size distribution was loaded in the quartz tubular reactor. The catalysts were first heated to 600 °C at a rate of 10 °C·min−1 in N2 (35 mL·min−1). After heating to 600 °C, the catalysts were treated in H2/N2 (40% H2/N2, 35 mL·min−1) for 1 h and then purged in N2 (35 mL·min−1) for another 30 min, meanwhile, reduce the reaction temperature to 500/550 °C. Afterwards, the reactant gas [C3H8: N2 = 1: 6 or CO2: C3H8: N2 = x: 1: (6-x), total flow rate is 35 mL·min−1] was fed into the reactor. The gas products were analyzed by an online GC (Shimadzu GC-2014C) what integrated with double detectors: the thermal conductivity detector for H2, N2, CO and CO2, and the flame ionization detector for hydrocarbons, such as CH4, C2H6, C2H4, C3H8, and C3H6, where N2 was used as the internal standard. The C3H8 and CO2 conversions were defined as follows:

$${{\mbox{C}}}_3{{\mbox{H}}}_8\;{{\mbox{conversion}}}:{X}_{{{{ C}}_{{{3}}}{H}}_{{8}}}\left(\%\right)=\frac{{{ F}}_{{{{C}}_{{3}}{ H}}_{{8}}}^{{in}}-{{F}}_{{{{C}}_{{{3}}}{{ H}}}_{{{8}}}}^{{{out}}}}{{{{ F}}}_{{{{ C}}_{{{3}}}{ H}}_{{8}}}^{{{in}}}}\times {{100}}$$
(1)
$${{\mbox{CO}}}_2\;{{\mbox{conversion}}}:{X}_{{{CO}}_{{2}}}(\%)\frac{{{ F}}_{{{CO}}_{{2}}}^{{in}}-{{ F}}_{{{{CO}}}_{{{2}}}}^{{{out}}}}{{{F}}_{{{CO}}_{{2}}}^{{in}}}\times {{100}}$$
(2)
$${{\mbox{C}}}_3{{\mbox{H}}}_6\;{{\mbox{sel. in}}}\;{{{C}}}_{{{x}}}{{H}}_{{{y}}}{:S}_{{{ C}}_{{{3}}}{{H}}_{{6}}}^{{{C}}_{{x}}{{ H}}_{{y}}}\left(\%\right)=\frac{{{F}}_{{{C}}_{{{3}}}{{H}}_{{{6}}}}^{{{o}}{{ut}}}}{{{{F}}}_{{{{C}}}_{{{3}}}{{{H}}}_{{{6}}}}^{{{out}}}+\frac{{{2}}}{{{3}}}{{{F}}}_{{{{C}}}_{{{2}}}{{{H}}}_{{{6}}}}^{{{out}}}+\frac{{{2}}}{{{3}}}{{{F}}}_{{{{C}}}_{{{2}}}{{{ H}}}_{{{4}}}}^{{{out}}}+\frac{{{1}}}{{{3}}}{{{F}}}_{{{C}}{{{ H}}}_{{{4}}}}^{{{out}}}}\times {{100}}$$
(3)
$${{\mbox{Net}}}\;{{{{\rm{C}}}}}_{3}{{{{\rm{H}}}}}_{6}\;{{\mbox{sel.}}}:{S}_{{{{ C}}}_{{{3}}}{{{ H}}}_{{{6}}}}^{{{net}}}\left(\%\right)=\frac{{{{F}}}_{{{{ C}}}_{{{3}}}{{{H}}}_{{{6}}}}^{{{o}}{{ut}}}}{{{{{F}}}_{{{{C}}}_{{{3}}}{{{ H}}}_{{{8}}}}^{{{in}}}-{{F}}}_{{{{C}}}_{{{3}}}{{{H}}}_{{{8}}}}^{{{o}}{{ut}}}}\times {{100}}$$
(4)
$${{{{\rm{CO}}}}\;{{{\rm{formed}}}}\;{{{\rm{from}}}}\;{{{\rm{CO}}}}_{2}:F}_{{CO}}^{{{CO}}_{2}}={ F}_{{CO}_{{2}}}^{in}-{F}_{{CO}_{2}}^{out}({{{\rm{ml}}}}/{\min })$$
(5)
$${{{\rm{{CO}}}}\; {{{\rm{formed}}}}\; {{{\rm{from}}}}\;{{{C}}}_{{{x}}}{{H}}_{{{y}}}:F}_{{CO}}^{{{C}_{{x}}{{H}}}_{{y}}}={{F}}_{{CO}}^{{out}}-{F}_{{CO}}^{{{CO}}_{{2}}}({{{\rm{ml}}}}/{{\min }})$$
(6)
$${{{\rm{Productivity}}}}:{{{{\rm{Rate}}}}}_{{{{{C}}}_{{{3}}}{{H}}}_{{{6}}}}({{{\rm{mmol}}}}/({{{g}}}_{{{cat}}.}{{{\rm{\cdot }}}}{{h}}))=\frac{{{{ F}}}_{{{{{C}}}_{{{3}}}{{H}}}_{{{6}}}}\times {{60}}}{{{{ m}}}_{{{cat}}.}\times {{22.4}}}$$
(7)
$${{{\rm{Carbon}}}}\; {{{\rm{balance}}}}=\frac{{{3}}{{{ F}}}_{{{{ C}}}_{{{3}}}{{{ H}}}_{{{8}}}}^{{{out}}}+{{3}}{{{ F}}}_{{{{ C}}}_{{{3}}}{{{ H}}}_{{{6}}}}^{{{out}}}+{{2}}{{{ F}}}_{{{{ C}}}_{{{2}}}{{{ H}}}_{{{6}}}}^{{{out}}}+{{2}}{{{ F}}}_{{{{ C}}}_{{{2}}}{{{ H}}}_{{{4}}}}^{{{out}}}+{{{ F}}}_{{{ C}}{{{ H}}}_{{{4}}}}^{{{out}}}}{{{3}}{{{ F}}}_{{{{ C}}}_{{{3}}}{{{ H}}}_{{{8}}}}^{{{out}}}}\times {{100}}$$
(8)

Where \({F}_{{{C}_{3}H}_{8}}^{{in}}\), \({F}_{{{CO}}_{2}}^{{in}}\) and \({F}_{{{C}_{3}H}_{8}}^{{out}}\), \({F}_{{{CO}}_{2}}^{{out}}\) mean flow rates of x (mL·min−1) inlet and outlet the reactor and \({m}_{{cat}.}\) means catalyst mass (g). C3H8 conversion, CO2 conversion and C3H6 selectivity were distinguished as Eqs. (1–4). Here, CO can be formed from CxHy by dry reforming as well as from CO2 via RWGS reaction, which were distinguished as Eqs. (5) to (6). C3H6 reaction formation rate was distinguished as Eq. (7). Carbon balance was distinguished as Eq. (8).

When the catalyst is regenerated, it contains 45 min regeneration at 500 °C under air stream. The flow rate of air is 35 mL·min−1.

Characterizations

XRD measurements were performed on a Bruker D8 diffractometer operating at 200 mA and 40 kV, employing the graphite filtered Cu Kα as the radiation source. The data points were collected by step scanning with a rate of 8°·min−1 from 2θ = 10° to 80°.

TEM images were taken using a JEM-2100F transmission electron microscope under a working voltage of 200 kV, equipped with a liquid nitrogen cooled energy dispersive X-ray spectroscopy (EDS) detector for elemental analysis. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) images were obtained by JEM-ARM200F, capable of sub-angstrom resolution. The dried catalyst was firstly reduced at 600 °C for 1 h in a stream of 10% H2/Ar (30 mL·min−1). Then, the sample powder was dispersed in ethanol by ultrasonic and deposited on a cooper grid coated with an ultrathin holey carbon film.

O2-TPO. Temperature-programmed oxidation with O2 was performed on Micromeritics Auto Chem II 2920 chemisorption apparatus equipped with Hiden QIC-20 mass Spectrometer. The 100 mg fresh catalysts that treated 3 h in different reaction atmosphere conditions were pretreated at 300 °C for 1 h under Ar stream, after cooling to 100 °C, 10% O2 in He (20 mL·min−1) mixture was switched to reach a stabilized baseline. Finally, the analysis was heated to 800 °C, at a rate of 10 °C·min−1, and the products (CO2, m/e equal to 44) in the exhaust were measured by the mass spectrometer.

CO2-TPD. CO2 temperature-programmed desorption was also performed on the Micromeritics Auto Chem II 2920 chemisorption apparatus. Typically, 100 mg of sample was first treated at 600 °C for 1 h in 10% H2/Ar (30 mL·min−1) and then purged with Ar (30 mL·min−1) for 30 min and cooled down to 50 °C. CO2 was adsorbed using a flow of 5% CO2/He (20 mL·min−1) for 1 h. Finally, the samples were heated to 800 °C at a rate of 10 °C/min under the flow of He (30 mL·min−1). The output products were measured on Hiden QIC-20 mass spectrometer.

X-ray photoelectron spectroscopy (XPS) was conducted on a Thermo Fisher ESCALAB-Xi equipped with an Al Kα X-ray radiation source (hѵ = 1486.6 eV). The binding energies were calibrated using the C1s peak at 284.8 eV. Avantage software was used to core-level spectra decomposition. The catalyst was firstly reduced at 600 °C for 1 h in a stream of 10% H2/Ar (30 mL·min−1).

H2-TPR. H2 temperature-programmed reduction was performed on the Micromeritics Auto Chem II 2920 chemisorption apparatus, 100 mg of sample was treated at 600 °C for 1 h in H2 atmosphere (30 mL·min−1) and cooled down to 30 °C in Ar atmosphere (30 mL·min−1). After that, the sample was flushed with 10% H2/Ar (20 mL·min−1) for 5 min and heated at a rate of 10 °C/min up to 800 °C.

H2-TPD. H2 temperature-programmed desorption was performed on the Micromeritics Auto Chem II 2920 chemisorption apparatus, 100 mg of sample was treated at 600 °C for 1 h in H2 atmosphere (30 mL·min−1), and then the sample was treated at 600 °C for 30 min and cooled down to −10 °C in Ar atmosphere (30 mL·min−1). Then switch the gas to 10% H2/Ar for 1 h, after that, the sample was flushed with Ar (20 mL·min−1) for 30 min and heated at a rate of 10 °C/min up to 800 °C. The MS signals of H2 (m/e equals to 2) were recorded.

Isothermal and pulse experiments were carried out on the Micromeritics Auto Chem II 2920 chemisorption apparatus, and the output products were also measured on Hiden QIC-20 mass spectrometer). Typically, 100 mg samples were pretreated in U-type reactor at 600 °C for 1 h under 10% H2/Ar (30 mL·min−1), after this, the temperature decreased to 500 °C in Ar (30 mL·min−1). And then, the analysis was carried out using 10 switches back and forth between atmospheres C3H8/CO2/Ar and C3H8/Ar for 1 h. The exhausted gas was analysis by online mass spectrometer (CO2, C3H6, C3H8, CO, H2O and H2, m/e equals to 44, 41, 29, 28, 18 and 2, respectively).

TPSR. Temperature-programmed experiments were all carried out on a fixed-bed reactor equipped with a Mass Spectrometry OmniStar from Pfeiffer Vacuum. For TPSR of C3H8 over the fresh catalyst, 100 mg of sample was treated in a flow rate of 35 mL·min−1 of 40% H2/Ar at 600 °C for 1 h and cooled down to 50 °C in Ar (35 mL·min−1). After switching to C3H8/Ar (5 mL·min−1 C3H8, 30 mL·min−1 Ar) for 30 min, the sample was heated at a rate of 10 °C·min−1 up to 600 °C. For TPSR of C3H8 and CO2 co-feed over the fresh catalyst, the difference is that switching to C3H8/CO2/Ar (5 mL·min−1 C3H8, 3 mL·min−1 CO2, 27 mL·min−1 Ar) for 30 min. The output products (CO2, C3H6, C3H8, CO, H2O and H2, m/e equals to 44, 41, 29, 28, 18 and 2, respectively) were monitored by mass spectrometry (MS).

Laser Raman spectroscopy experiments were carried out using a JEOL NRS-5100 spectrometer at 25 °C to investigate the carbonaceous deposits over the spent (3 h) catalysts, with 2 cm−1 resolution of the apparatus and a 532 nm excitation source.

In-situ DRIFTS. The In-situ Diffuse Reflectance Infrared Fourier-transform Spectroscopy experiments were carried out on a Nicolet IS50 spectrometer, equipped with a Harrick Scientific DRIFTS cell fitted with ZnSe windows. The detailed treatments are given as the follows:

In-situ DRIFTS experiments were performed to monitor the different of coordination of Pt1Sn7/Al2O3 and Pt1Sn7/CeOx-Al2O3. The fresh catalysts were heated from ambient temperature to 600 °C at a rate of 10 °C·min−1 and retained at 600 °C in a flow rate of 35 mL·min−1 of 40% H2/Ar. Then, the catalysts were cooled down to 30 °C and the backgrounds (8 cm−1 resolution, 64 scans) were collected after Ar purging in a flow rate of 35 mL·min−1 for at least 1 h. With the addition of a flow of 3 mL·min−1 of CO, the adsorption of CO molecules on the surface of the catalysts continued for 30 min. After that, the DRIFTS spectra were recorded till no visible change in the absorption band intensities under Ar purging.

In-situ DRIFTS experiments were performed to record the surface species evolution of Pt1Sn7/CeOx-Al2O3 during successive treatments with H2/Ar. The fresh catalyst (50 mg powdered sample) was filled into the reactor. Initially, the fresh catalysts were heated at a rate of 10 °C·min−1 from ambient temperature to 500 °C. The systems were then purged with Ar (35 mL·min−1) at 500 °C for 1 h to remove the adsorbates. After that, the background spectra (a resolution of 8 cm−1, 64 scans) were collected. Subsequently, the sample was treated with 40% H2/Ar (35 mL·min−1) for 1 h, DRIFTS spectra were collected every 1 min. Followed by flushing with Ar (35 mL·min−1) for 10 min.

In-situ DRIFTS experiments were performed to record the surface species evolution of Pt1Sn7/CeOx-Al2O3 under reaction conditions (RWGS, PDH, CO2-PDH). For PDH switch to CO2-PDH DRIFTS, the fresh catalyst (50 mg powdered sample) was filled into the reactor and flushed with Ar (35 mL·min−1) at 300 °C for 1 h to remove the adsorbates. Then, the catalysts were heated from 300 °C to 600 °C at a rate of 10 °C/min and retained at 600 °C in a flow rate of 35 mL·min−1 of 40% H2/Ar. After this, the catalysts were cooled down to 500 °C and the backgrounds (8 cm−1 resolution, 64 scans) were collected after Ar purging in a flow rate of 35 mL·min−1 for at least 1 h. the background spectra (a resolution of 8 cm−1, 64 scans) were collected. Afterwards, a C3H8/Ar mixture with a flow rate of 35 mL·min−1 (5 mL·min−1 C3H8, 30 mL·min−1 Ar) was introduced into the reactor for 32 min, after this, a C3H8/CO2/Ar mixture with a flow rate of 35 mL·min−1 (5 mL·min−1 C3H8, 3 mL·min−1 CO2, 27 mL·min−1 Ar) was introduced into the reactor for 58 min. For RWGS, the different is that a H2/CO2/Ar mixture with a flow rate of 35 mL·min−1 (5 mL·min−1 H2, 5 mL·min−1 CO2, 25 mL·min−1 Ar) was introduced into the reactor after pre-reduced. And for CO2-PDH DRIFTS, the different is that a C3H8/CO2/Ar mixture with a flow rate of 35 mL·min−1 (5 mL·min−1 C3H8, 3 mL·min−1 CO2, 27 mL·min−1 Ar) was introduced into the reactor for 32 min directly after pre-reduced. DRIFTS spectra were collected every 1 min.