Introduction

Propane dehydrogenation (PDH) is an industrially important alternative to oil-based cracking processes1,2. However, the commercial non-oxidative propane dehydrogenation containing CrOx or Pt-based catalysts is endothermic and equilibrium-limited, necessitating much heat to achieve viable propylene yield3,4. Although the oxidative dehydrogenation of propane (ODH) has the potential to improve conversion for favorable thermodynamics, propylene selectivity is hampered by overoxidation to CO25,6. A similar challenge is faced in selective oxidation reactions in the chemical industry7,8.

Chemical looping engineering offers exciting new opportunities for the challenges through the physical or temporal separation of dehydrogenation and oxidation by solid oxygen carrier mediums9,10. Unlike traditional catalysts, the oxygen carriers react with alkanes and undergo reversible changes by donating and replenishing oxygen to close the loop in the reducer and oxidizer reactors. Most oxygen carriers involve the metal centers or oxide composites to modulate lattice oxygen reactivity, using bulk doping11, surface modification12, or confinement in supports13. Recently, vanadia/ceria catalysts have attracted increased attention in the oxidative dehydrogenation of propane with O2 co-feeding. The electronic effects and redox properties were investigated at the molecular level13,14,15,16. Nevertheless, direct experimental and theoretical insights into the lattice oxygen diffusion and the surface dynamics have not been reported yet for the anaerobic oxidative dehydrogenation via chemical looping engineering.

In this work, to unravel the oxygen diffusion and reaction dynamics regarding the active sites, a nanoscale core-shell redox catalyst combining dehydrogenation catalyst and oxygen carrier at one particle is designed. The core-shell redox catalyst is preferably composed of two to three atomic layer-type vanadia coating ceria nanodomains to achieve the synergetic modulation of lattice oxygen bulk diffusion and surface reaction. In the dehydrogenation step (reducer), ceria-vanadia redox catalysts donate lattice oxygen for the dehydrogenation of propane to produce propylene, H2O, and H2, affording a reduced valence state that can be reoxidized in the reoxidation step (oxidizer) by air to close the loop (Fig. 1a). Combining in situ spectroscopies, kinetics, and theoretical calculation, an intrinsically dynamic lattice oxygen “donator-acceptor” process is proposed, which accounts for the synergetic modulation of bulk diffusion and surface reaction in the core-shell redox catalyst. O2− generated from ceria oxygen carrier is boosted to diffuse and transfer to vanadia dehydrogenation sites via a concerted hopping pathway at the interface, stabilizing surface vanadia with moderate oxygen coverage without significant overoxidation or cracking.

Fig. 1: Identification of vanadia layers that coat ceria nanodomains.
figure 1

a Diagram of core-shell redox catalysts in propane dehydrogenation by the chemical looping engineering: dehydrogenation and oxidation in fuel reactor (reducer) and air reactor (oxidizer), respectively. b–d HAADF-STEM images and e–h EELS mappings of core-shell ceria-vanadia redox catalysts (6 V/30CeAl): (f): V; (g): Ce; (h): V + Ce. i Line-scanning EELS. j EELS of the domains ((1), (2), (3)) in (b).

Full size image

Results

Formation of ceria-vanadia core-shell redox catalysts

The core-shell redox catalysts were prepared using a two-step incipient wetness impregnation method. The ceria-vanadia samples were named xV/yCeAl, where x(y) is the percent weight of V(Ce). The vanadia and ceria catalysts were obtained by VOx and CeO2 supported on γ-Al2O3, respectively. At vanadia surface density of 4.3 V/nm2 (Supplementary Table 1)17,18,19 (6 V/30CeAl), atom-resolved high-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) images identified vanadia mainly existed as monolayers and bilayers along ceria surface (Fig. 1b–d). Electron energy loss spectra (EELS) mappings of V L2,3 and Ce M4,5 edges affirmed vanadia sites anchored on widespread ceria nanodomains (Fig. 1e–h). The well-defined core-shell structure was further validated by line-scanning EELS that crossed an individual particle, wherein the outer shell was ~1 nm, corresponding to roughly two to three atomic vanadia layers (Fig. 1i). Judged from variations of V L2,3 and O K edges, vanadia presented dominantly as a mixture of V5+ and V4+, while relative intensity ratios of Ce M4,5 edges (1.11–1.19)20 indicated the presence of Ce3+ and Ce4+ ((1), (2), (3) in Fig. 1b) inside one particle (Fig. 1j, Supplementary Fig. 2, and Supplementary Table 2)21.

Chemical looping oxidative dehydrogenation performance

Application of core-shell redox catalysts was proven in a continuous chemical looping oxidative dehydrogenation scheme (Supplementary Fig. 3). Ceria-vanadia redox catalysts exhibited traceable CO2 (<3%) with high propylene selectivity of 93.5% and formation rate of 42.5 mmol C3H6/gcat/h (at 5th min in one cycle), implying excessive overoxidation or cracking were inhibited. At 600 oC and GHSV of 2500 h−1, an average 90% propylene selectivity at propane conversion of 49% was obtained within 60 mins (Fig. 2b and Supplementary Fig. 4), superior to that of ceria (30CeAl) (78.3%), vanadia (6 V/Al) (71.6%), and state-of-the-art catalysts (Fig. 2c). Industrially relevant K-CrOx/Al2O34,22 was compared under identical reaction conditions. The propylene space-time yield (STY) of ceria-vanadia redox catalysts was 10.3 mmol C3H6/gcat/h, comparable to that of K-CrOx/Al2O3 (10.6 mmol C3H6/gcat/h) (Supplementary Fig. 3g). However, considering different reaction sites in two catalysts, propylene STY normalized by moles of V (13.3 mol C3H6/molV/h) was about five times higher than that normalized by moles of Cr (2.8 mol C3H6/molCr/h). The deactivation rate constant (kd) using a first-order deactivation model was used to determine its life in the dehydrogenation step. Ceria-vanadia redox catalysts exhibited smaller kd (0.04 h−1) than K-CrOx/Al2O3 (0.99 h−1) (Supplementary Figs. 5, 6). When the temperature was increased to 650 oC, propylene selectivity remained at 80%. However, pure vanadia showed quick deactivation (kd = 1.4 h−1) and propylene selectivity decreased to 34%.

Fig. 2: Chemical looping oxidative dehydrogenation performance.
figure 2

a In situ XRD patterns of ceria-vanadia redox catalysts (6 V/30CeAl). b Comparison of ceria (30CeAl), vanadia (6 V/Al), and ceria-vanadia redox catalysts (6 V/30CeAl). Conditions: 600 oC, GHSV = 2500 h−1, C3H8/N2 = 0.25. c Comparing ceria-vanadia redox catalysts (6 V/30CeAl) with established oxide-based and Pt-containing catalysts (see Supplementary Tables 3, 4). Motifs of triangle, rhombus, and sphere represent ODH, PDH, and CL-ODH, respectively. d Cyclic performance over ceria-vanadia redox catalysts (6 V/30CeAl). Dehydrogenation step: 600 oC, GHSV = 2500 h−1, C3H8/N2 = 0.25 for 30 min; Inert purge: 600 oC, N2 = 40 mL/min for 5 min; Oxidation step: 600 oC, 20 vol.% O2/N2 = 20 mL/min for 15 min. e Comparison of energy consumption and CO2 emission of traditional Oleflex technology and chemical looping scheme (see Methods and Supplementary Tables 6–8).

Full size image

The reversible charge-discharge of lattice oxygen in ceria-vanadia redox catalysts was verified by in situ XRD. In the dehydrogenation step at 600 oC, diffract peaks of CeO2 shifted to lower diffraction angles, e.g., the (111) diffract peak shifted from 28.4 o to 28.0 o due to the formation of larger Ce3+ ions. Oxidation with air then recovered its position (Fig. 2a and Supplementary Fig. 7). During 300 long-term chemical looping cycles, structure durability and robust performance with an average 43.6% C3H6 yield and space-time yield of 9.9 mmol C3H6/gcat/h was achieved (Fig. 2d and Supplementary Table 5). When altering either shell or core components in core-shell redox catalysts, comparable C3H6 formation rates were obtained (Supplementary Fig. 8). For chemical looping oxidative dehydrogenation of ethane, the ceria-vanadia redox catalysts also presented 92% ethylene selectivity with 31% ethane conversion at 600 oC (Supplementary Fig. 9), validating its potential application in the dehydrogenation of light alkanes. Compared with the commercialized Oleflex scheme (Supplementary Fig. 10 and Supplementary Tables 6–10), 45% of energy savings can be anticipated from the chemical looping oxidative dehydrogenation system (Fig. 2e), with separation being the main driver for energy consumption.

Evidence of oxygen diffusion and surface reaction

When exposed to propane for 120 mins, peaks of B2 and C in the Ce L3-edge shifted to lower energy (∆2.1 eV). The B0 white line located at 5726 eV was then dominated, a characteristic of Ce3+ (Fig. 3a), indicating the reduction of ceria (Ce4+→Ce3+) in the ceria-vanadia catalysts, in contrast to the negligible formation of Ce3+ in pure ceria (Ce L3-edge shift of ∆0.7 eV). V K pre-edge close to 5467 eV featuring V4+ oxidation state kept nearly unmoved within 30 mins. After that, a decrease of pre-edge peak intensity and shift of edge position to lower energy (~∆1.2 eV) occurred as the reduction of CeO2 stopped (Fig. 3b and Supplementary Fig. 11). This implied that in the ceria-vanadia redox catalysts, vanadia tended to be reduced to lower valence states when oxygen was not timely supplied from ceria. For pure vanadia without CeO2 supporting, V K pre-edge featuring V5+ was easily and quickly reduced to V3+ (V K-edge shift of 2.7 eV)11. Together, the changes of Ce L3-edge and V K-edge indicated that ceria in ceria-vanadia redox catalysts acted as an “oxygen reservoir” that could supply the lattice oxygen to stabilize the surface vanadia, which accords with the previous research that ceria helped to oxidize the reduced vanadia via ceria lattice oxygen23,24,25.

Fig. 3: Experimental evidence of oxygen diffusion and surface reaction.
figure 3

a In situ XANES spectra of Ce L3-edge (CeO2 standards as the references) over ceria-vanadia (6 V/30CeAl) (top) and pure ceria (30CeAl) (bottom) and b V K-edge (V foil, V2O5, VO2, and V2O3 standards as the references) over ceria-vanadia (6 V/30CeAl) (top) and pure vanadia (6 V/Al) (bottom) at 600 °C under the flow of 20% C3H8/N2 (20 mL/min). c In situ Raman spectra of ceria-vanadia (6 V/30CeAl) (top) and vanadia (6 V/Al) (bottom) at 600 °C under the flow of 20% C3H8/N2 (20 mL/min). In situ DRIFTS spectra of temperature-programmed and isothermal propane dehydrogenation over ceria-vanadia (6 V/30CeAl) (d) and vanadia (6 V/Al) (e, f). g The calculated ratios of H2/H2O during the C3H8 transient pulses at 600 oC. h Experimental relaxation curve in the form of fractional weight change as a function of time at 600 oC under the flow of 20% C3H8/He (10 mL/min). i Schematic representation of the concerted oxygen diffusion in the ceria-vanadia redox catalyst. The black, white, and red spheres represent C, H, and O atoms.

Full size image

We further evidenced the dynamic evolution of lattice oxygen in ceria-vanadia redox catalysts. Raman spectra of CeO2 were dominated by the strong F2g mode of fluorite phase at 464 cm−1 with weak bands at 598 cm−1 due to defect-induced (D) mode. With vanadia coating, in addition to V = O and V-O-V stretching, additional bands of V-O-Ce (859 and 720 cm−1) emerged (Supplementary Fig. 1j, k), affirming the construction of vanadia-ceria interface26,27. Upon propane exposure, in situ Raman spectra verified the continuous reduction of CeO2 in ceria-vanadia redox catalysts that the intensity of F2g mode dramatically decreased with the time on stream. It is noted that the band of V-O-Ce kept relatively stable. In contrast, intensity ratios of V-O-Ce band and F2g mode in terms of IV-O-Ce/IF2g increased, validating that Ce-O species in ceria were gradually consumed to supplement and stabilize interfacial and surface V-O species (Fig. 3c and Supplementary Fig. 12)24,25. As much oxygen was depleted after 30 min, the time was also shown in in situ XANES spectra. The D1 band and G band corresponding with coke deposits, were then observed, implying that the cracking and coking of propane occurred on reduced vanadia sites. Comparatively, pure vanadia was readily reduced to V3+, which leads to more coke deposition in the characteristic of the more dominated intensity of D1 and G band27,28,29.

In situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) upon propane exposure identified the co-existence of dehydrogenation and cracking of propane induced by this dynamic oxygen evolution. Peaks ascribed to asymmetric and symmetric CH3 stretching modes (2970 and 2875 cm−1) started at 100–150 oC (Fig. 3d)30. The presence of a band centering at 1645 cm−1 (ν(CH3CH=CH2)) implied that the propyl complex was oxidatively dehydrogenated to propenyl by heterolytically subtracting H to neighboring V-O sites, leading to the occurrence of vanadium hydroxyl band (V-OH, 3660 cm−1)28. However, a peak of ν(C=O) (1680 cm−1) attributed to acetone, the intermediate of overoxidation of propane to COx, was dominated on the pure VOx catalysts when the temperature was higher than 150 oC, along with a significantly negative V = O band induced by the ready reduction of vanadia (V5+ → V3+) (Fig. 3e, f and Supplementary Fig. 11)11,28. This “over quick” oxygen removal would induce the transformation of oxidative to non-oxidative dehydrogenation and the occurrence of propane cracking. At 250–600 oC, two peaks at 1545 and 1460 cm−1 attributed to the unsaturated or aromatic species, the precursors of coke deposits27,28 that lead to fast deactivation were observed, which were also evidenced by the more dominated D1 and G band in situ Raman spectra on pure vanadia catalysts during dehydrogenation step.

As evidenced by the in situ spectroscopies, the release of lattice oxygen would induce the existence and transformation of different reaction periods, including overoxidation, oxidative dehydrogenation, and non-oxidative dehydrogenation. Under differential reactor operation by controlling the propane conversion lower than 10%, the C3H6 formation rate showed a linear relationship with C3H8 pressure, while C3H8 conversion kept identical at different C3H8 pressures, indicating the rate of propene formation is typically related to propane partial pressure, i.e., a first-order reaction with respect to propane (Supplementary Fig. 13). To clarify the contribution of oxidative and non-oxidative dehydrogenation, the formation of H2O and H2 over ceria-vanadia redox catalysts in their dehydrogenation tests were investigated. As shown in Supplementary Fig. 13, the initial ratio of H2O to H2 at the 5th min was 0.44; however, it decreased to about 0.05 after 60 mins. Therefore, oxidative dehydrogenation could be more dominated in the initial period (less than 5 mins) and it was transformed to non-oxidative dehydrogenation with time, accounting for the introduction of the reoxidation step to recover the lattice oxygen after the 30-min dehydrogenation test during the continuous dehydrogenation-reoxidation cycles with the ratios of H2O to H2 of ~0.21.

Transient pulses and oxygen release kinetics

To catch the transient distributions of products, C3H6, COx, H2O, and H2, especially in the initial period, C3H8 transient pulse experiments were employed at 600 oC using online mass spectrometry (MS) (Supplementary Fig. 14). COx was firstly observed due to the overoxidation on the active surface oxygen species (period I), then it would undergo selectively oxidative dehydrogenation to C3H6 and H2O caused by lattice oxygen (period II). After the lattice oxygen were fully released, cracking or hydrogenolysis occurred, which was more dominated for pure vanadia catalysts, leading to the formation of CH4 (period III). Intensity ratios of H2/H2O defined as FH2/H2O, a sign of H combustion, maintained at five over ceria-vanadia redox catalysts (Fig. 3g), implying that O2− diffusion from ceria oxygen carrier leads to a pseudo-steady-state H combustion at surface vanadia sites6,31. In contrast, pure vanadia showed stage-divided products, along with the tenfold increased intensity ratios of H2/H2O with pulses of C3H8, indicating the presence of O2− gradient and transport limitation in the pure vanadia31. Gradually decreased ratios of H2/H2O in pure ceria implied surface H abstraction were sluggish due to the less active C–H dissociation center of CeO2. We then designed a temperature-programmed surface reaction (TPSR) to explore the surface reaction route. Initial dehydrogenation temperatures of C3H8 on pure vanadia and ceria-vanadia redox catalysts were at 338 and 340 oC, respectively, lower than that of pure ceria (412 oC) (Supplementary Fig. 14), indicating the presence of low-temperature C–H dissociation V-O centers ascribed from vanadia catalysts, corresponding to increased medium acidic sites as shown in NH3-TPD profiles (Supplementary Fig. 15).

Oxygen release kinetics provide further quantified rates of oxygen diffusion and surface reaction. 1.83 wt% active oxygen in ceria-vanadia redox catalysts was continuously removed during the dehydrogenation step (Supplementary Figs. 16,  17), much higher than that of the ceria oxygen carrier (0.37 wt%). However, quick oxygen removal (0.16 wt%) and coke accumulation were observed over the vanadia catalyst. Derived from thermogravimetry relaxation and diffusion-reaction equations (Fig. 3h)32, bulk diffusion coefficients (Ddiff) and surface exchange coefficient (kchem) of ceria-vanadia redox catalysts at 550−600 °C were close to one order of magnitude higher than that of ceria oxygen carrier (Supplementary Table 11), validating the dramatic acceleration of oxygen diffusion by ceria–vanadia interaction23,25,33. Derived from Arrhenius plots of Ddiff and kchem at 550–625 oC, ceria-vanadia redox catalysts possessed both lower activation energy (Ea) of oxygen migration (131.7 vs. 179.9 kJ/mol for ceria) and surface exchange barrier (89.8 vs. 176.3 kJ/mol for ceria) (Supplementary Fig. 18), which accounts for the modulation of oxygen diffusion and surface reaction to produce propylene, as shown in transient pulse experiments.

DFT calculations on oxygen diffusion and surface reaction

Atomic-level details of oxygen diffusion and surface reaction over ceria-vanadia redox catalysts were investigated by density functional theory (DFT) calculations. With the elimination of O of VOx (O(V)) in CeO2-VOx (Supplementary Fig. 19), the outermost O of CeO2 (O(Ce)) started to coordinate with V to form V-O-Ce interface that exposes Ce3+ centers (Fig. 4a). Bader charge analysis further proves electrons accumulate on Ce with reduction of VOx (Supplementary Fig. 20). Meantime, VOx preserves its valency state until ML-V2O3 period (Supplementary Table 12). Existence of monolayer-V2O5 (ML-V2O5) on CeO2 is expectedly activated surface oxygen with lower oxygen vacancy (Ovac) formation energies than that of pure CeO2. When surface oxygen was entirely removed, and ML-VO2 formed, oxygen diffusion from CeO2 to VOx turned endothermic with a reaction energy of 0.62 eV. However, it became exothermic with further oxygen release in VOx between ML-VO2 and ML-V2O3, verifying the continuous oxygen transfer from O(Ce) to O(V) sites (Supplementary Fig. 21). Derived from the structure of ML-VO2 on CeO2, a concerted hopping pathway was preferred that O2- diffusion from O(Ce) (1.17 eV) and subsequential transfer from interface V-O-Ce to V=O mediated by bridge V-O-V exhibited the lowest barrier of 0.95 eV (Fig. 4b) as shown in in situ Raman spectra, lower than that of direct hopping pathway (1.45 eV) (Fig. 4c) and isolated hoping pathway (1.47 eV) (Supplementary Fig. 22a). As a result, ML-VO2 on CeO2 could be stabilized by the intrinsically dynamic “donator-acceptor” process. Four V atoms were connected to form a ring-like structure at the interface, leading to an atomic ratio of V/O at 1:2, evidenced by in situ V K-edge QXANES. For each [V4O8] unit (Supplementary Fig. 23), two terminal V=O bonds were kept, which would not exist in crystalline VO2 (Supplementary Fig. 22b, c)11,34.

Fig. 4: DFT calculations on oxygen diffusion and surface reaction.
figure 4

a Models of CeO2 and ML-VO2 (inset [V4O8] units). b Optimal concerted hopping pathway in ML-VO2 contrast to c direct hopping pathway. Ea represents the energy barrier. d The calculated energy profiles on ML-VO2 (blue) and crystalline V2O5 (yellow). Reaction steps include (i) dehydrogenation from propane to absorbed propyl, (ii) dehydrogenation from absorbed propyl to absorbed propene, and (iii) dehydrogenation from absorbed propyl to acetone. TS represents the transition states. V: dark blue; Ce; bright blue; O: red; C: black; H: white.

Full size image

Valence states of V have been believed to manipulate the propylene selectivity35,36. The following calculations on the formation pathways of H2O(g) and H2(g) showed that ML-VO2 has a strong preference to proceed oxidative dehydrogenation compared with ML-V2O3 (Supplementary Fig. 24) and exhibited lower barriers of first and second dehydrogenation of propane to propylene, but higher barriers for acetone formation, a significant intermediate of overoxidation, than crystalline V2O5 and ML-V2O3 (Fig. 4d and Supplementary Fig. 25), according to experimental results that ceria-vanaida redox catalysts showed higher initial propane conversion and propylene selectivity than 6 V/Al and crystalline V2O5. These results firmly support that lattice oxygen transfer from ceria was to stabilize moderate V valence states and oxygen coverage for selective dehydrogenation without significant overoxidation or cracking and coking.

In conclusion, the ceria-vanadia core-shell redox catalysts have been designed for propane dehydrogenation via chemical looping engineering, which exhibited the highest 93.5% propylene selectivity and 43.6% propylene yield during the long-term dehydrogenation-oxidation cycles. An intrinsically dynamic lattice oxygen “donator-acceptor” process in core-shell redox catalyst was proposed by the combination of in situ XAS, Raman, transient pulses and oxygen release kinetic analysis, that the lattice oxygen generated from ceria were boosted to diffuse (close to one order of magnitude higher Ddiff at 550−600 °C that that of ceria oxygen carrier), and transfer to surface vanadia dehydrogenation sites via a concerted hopping pathway at the interface, stabilizing surface vanadia for selective dehydrogenation without significant overoxidation or cracking. Atomic-level details of oxygen diffusion and surface reaction over ceria-vanadia redox catalysts were further revealed by DFT calculations that derived from the structure of ML-VO2 on CeO2, O2- diffusion from O(Ce) and subsequential transfer from interface V-O-Ce to V=O mediated by bridge V-O-V exhibited the lowest barrier lower than that of the direct and isolated hoping pathway. This study opens the possibilities for the design and application of efficient core-shell redox catalysts in selective oxidations and chemical looping systems.

Methods

Material preparation

The VOx-CeO2 catalysts were prepared via two-step-impregnation procedure by dissolving Ce(NO3)4.6H2O (99.5%, Aladdin (China) Chemical Co., Ltd) in deionized water (2 mL/g Al2O3) with γ-Al2O3 (Adamas, 99.99%, SBET = 180 m2/g) as support to well disperse the cerium oxides. Then VOx was impregnated by dissolving NH4VO3 (99.0%, Tianjin Guangfu Technology Development Co. Ltd.) and oxalic acid (99.0%, Aladdin Industrial Corporation) [NH4VO3/oxalic acid = 0.5 (mole ratio)] in deionized water. The samples were dried at 80 oC for 12 h and then calcinated at 600 oC for 3 h. The VOx-CeO2 samples were named xV/yCeAl, where x(y) is the percent weight ratio of V(Ce). The vanadia and ceria catalysts were obtained by VOx, and CeO2 supported on γ-Al2O3 with the same V and Ce loadings with VOx-CeO2, respectively, as reference samples. An industrially relevant K–CrOx/Al2O3 catalyst was carried out (mimic Catofin from Lummus) to compare the VOx-CeO2 catalyst. For K-CrOx/Al2O3 reference catalysts, 20 wt% Cr and 1.0 wt% K were utilized with Cr(NO3)3·9H2O (99.5%, Aladdin (China) Chemical Co., Ltd)37. The catalysts were dried for 2 h at 60 °C and overnight at 120 °C, followed by calcination at 600 °C for 3 h. The amount of Cr and K was about 19.8 wt% and 0.98 wt% based on the weight ratio of M to Al2O3, which were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES),

Reaction test

Reactivity tests were performed in a quartz fixed-bed reactor (8 mm ID) loaded with 0.5 g catalysts (20–40 mesh) mixed with 1 mL of quartz particles at atmospheric pressure. Switching between propane and air flows was employed during tests. The bed temperature was typically 600 °C, and the samples were reduced using propane (4 mL/min) diluted in nitrogen (16 mL/min) at 1.4 atm for 30 min. The samples were then reoxidized using air (20 mL/min) for 15 min. During the reduction and reoxidation reaction period, an inert period (40 mL/min of nitrogen) of about 5 min was inserted to prevent the mixing between propane and air. One redox cycle was thus completed. Afterward, the second reaction cycle was started by switching between propane and air flows. Exhaust streams were analyzed using an online GC (2060) equipped with a flame ionization detector (Chromosorb 102 column) and a thermal conductivity detector (Al2O3 Plot column). In addition, high-time resolution measurements were performed by Agilent 490 Micro GC equipped with three channels (MS 5A Plot column, PoraPlot Q, and PoraPlot Q), which can quickly quantitatively analyze the product compositions about once a minute.

The instantaneous propane conversion and propylene selectivity based on all products (including coking formation) and gas phase products are defined as the instantaneous values at the different time on stream, according to Eq. (1) and Eq. (2) (2a: selectivity including coke formation and 2b: gas selectivity). The propylene yield was calculated based on propane conversion and propylene selectivity (including coke formation):

$${{{{{\rm{Con}}}}}}\,(\%)=100\times ({[{{F}}_{{{{{{\rm{C3H8}}}}}}}]}_{{{{{{\mathrm{inlet}}}}}}}-{[{{F}}_{{{{{{\rm{C3H8}}}}}}}]}_{{{{{{\mathrm{outlet}}}}}}})/{[{{F}}_{{{{{{\rm{C3H8}}}}}}}]}_{{{{{{\mathrm{inlet}}}}}}}$$
(1)
$${{{{{\rm{Sel}}}}}}\,(\%)=100\times {[{{F}}_{{{{{{\rm{C3H6}}}}}}}]}_{{{{{{\mathrm{outlet}}}}}}}/({[{{F}}_{{{{{{\rm{C3H8}}}}}}}]}_{{{{{{\mathrm{inlet}}}}}}}-{[{{F}}_{{{{{{\rm{C3H8}}}}}}}]}_{{{{{{\mathrm{outlet}}}}}}})$$
(2a)
$${{{{{\rm{Sel}}}}}}\,(\%)=100\times 3\times {[{F}_{{{{{{\rm{C}}}}}}3{{{{{\rm{H}}}}}}6}]}_{{{{{{\mathrm{outlet}}}}}}}/ \left({\sum }_{{{{{{\rm{ni}}}}}}}\times {[Fi]}_{{{{{{\mathrm{outlet}}}}}}}\right)$$
(2b)
$${{{{{\rm{Yield}}}}}}\,(\%)={{{{{\rm{Con}}}}}}\,(\%)\times {{{{{\rm{Sel}}}}}}\,(\%)/100$$
(3)

The average conversion and selectivity within t minutes in the dehydrogenation stage were defined as the integral conversion and selectivity in Eq. (4) and Eq. (5) dividing t minutes during the dehydrogenation stage.

$${{{{{{\rm{Con}}}}}}}_{{{{{\mathrm{int}}}}}}\,(\%)=(\int {{{{{\rm{Con}}}}}}\,(\%)\,{{{{{\rm{dt}}}}}})/{{{{{\rm{t}}}}}}$$
(4)
$${{{{{{\rm{Sel}}}}}}}_{{{{{\mathrm{int}}}}}}\,(\%)=\left(\int {{{{{\rm{Sel}}}}}}\,(\%)\,{{{{{\rm{dt}}}}}}\right)/{{{{{\rm{t}}}}}}$$
(5)

The propylene formation rate was defined according to Eq. (6).

$${{{{{\rm{Rate}}}}}}={[{{{{{{\rm{F}}}}}}}_{{{{{{\rm{C}}}}}}3{{{{{\rm{H}}}}}}6}]}_{{{{{{{\mathrm{outlet}}}}}}}}/{{{{{\rm{m}}}}}}$$
(6)

A first-order deactivation model was used to evaluate the catalyst stability:

$${k}_{d}=({{{{{\rm{ln}}}}}}[(1-{{{{{{\rm{X}}}}}}}_{{{{{{{\mathrm{final}}}}}}}})/{{{{{{\rm{X}}}}}}}_{{{{{{{\mathrm{final}}}}}}}}]{{{-}}}{{{{{\rm{ln}}}}}}[(1{{{-}}}{{{{{{\rm{X}}}}}}}_{{{{{{{\mathrm{initial}}}}}}}})\,/{{{{{{\rm{X}}}}}}}_{{{{{{{\mathrm{initial}}}}}}}}])/{{{{{\rm{t}}}}}}$$
(7)

Where i stands for different side products in exhaust gases, ni is the number of carbon atoms of side products i, and Fi is the corresponding molar flow rate (mol/h). [FC3H6]outlet is the flow of propylene out of the reactor (mol/h). [FC3H8]outlet is the flow of propane out of the reactor (mol/h). [FC3H8]inlet is the flow of propane in of reactor (mol/h). t is the time during the dehydrogenation stage (min). m is the weight of catalysts (gcat). Xintial and Xfinal, respectively, represent the conversion measured at the initial and final period of an experiment. t represents the reaction time (h). kd is the deactivation rate constant (h−1). A high kd value means rapid deactivation, that is, low stability.

As internal standard experiments indicated that no significant coking or tar formation occurred under the conditions tested, a mass balance was used to calculate the yields. The selectivity and conversion for carbonaceous species were calculated relative to the carbon mass balance. H2O was calculated from a hydrogen balance. Molar flows of propane and reaction products are determined by mass flow controllers and GC. Flowrates of propane (industrial grade) and nitrogen (UHP) were controlled using two mass flow controllers (Bronkhorst) and calibrated to each individual gas to allow total flow rates of 0–20 mL min−1 and 10–50 mL min−1. Exhaust streams were analyzed using an online GC (2060) equipped with a flame ionization detector (Chromosorb 102 column) and a thermal conductivity detector (Al2O3 Plot column), and Agilent 490 Micro GC equipped with three channels (MS 5A Plot column, PoraPlot Q and PoraPlot Q).

Characterizations

X-ray powder diffraction (XRD) patterns were performed with 2θ values between 10° and 80° by using a Rigaku C/max-2500 diffractometer with the graphite filtered Cu Kα radiation (λ = 1.5406 Å), operated at 40 mA and 40 kV.

Raman spectra were recorded using a Renishaw inVia reflex Raman spectrometer with a 532 nm Ar ion laser beam. Samples were pretreated at 300 oC for 2 h to eliminate the presence of hydrated species.

Transmission electron microscope (TEM) was carried out on a JEM-2100F transmission electron microscope under a working voltage of 200 kV. The aberration-corrected scanning transmission electron microscopy (AC-STEM) images and EELS spectra were characterized on FEI Titan Cubed Themis G2 300 (FEI) 200 kV, capable of sub-angstrom resolution at Tianjin University of Technology. The sample powder was dispersed in deionized water by ultrasonic and deposited on a copper grid coated with an ultrathin holey carbon film.

XPS measurements were taken on a PHI 1600 ESCA instrument (PE Company) equipped with an Al Kα X-ray radiation source (hѵ = 1486.6 eV). Before measurements, all the samples were reduced under a flow of H2 at 600 °C for 1 h. The binding energies were calibrated using the C 1s peak at 284.5 eV as a reference.

H2 temperature-programmed reduction (H2-TPR) was performed on the chemisorption apparatus (Micromeritics AutoChem II 2920). Typically, 100 mg samples were pretreated at 400 °C for 1 h in an Ar stream and then cooled to 80 °C. The analysis was carried out in a mixture of 10 vol% H2 in Ar (30 mL/min), ramping temperature from 80 °C to 800 °C at 10 °C/min.

The C3H8 pulse experiment was measured on a Micromeritics Autochem II 2920 instrument equipped with a Hiden QIC-20 mass spectrometer. About 200 mg samples were pretreated at 400 °C for 1 h and raised up to 600 °C in the Ar stream. The analysis was carried out in a mixture of propane (5 mL/min) at 600 °C for 60 min. The output products (C3H8, C3H6, CO2, CH4, H2, and H2O, m/e equals 29, 41, 44, 16, 2, and 18, respectively) were measured via mass spectrometer.

The thermogravimetric relaxation experiment was performed on Themys. About 30 mg samples were pretreated at 300 oC for 2 h under He (100 mL/min) to eliminate the presence of hydrated species and then reacted in a mixture of 20% C3H8/He (10 mL/min) at 550, 575, 600, and 625 °C, respectively. The general way to model this process is by solving the diffusion coefficient (Ddiff), and surface exchange coefficient (kchem) contained diffusion equation. The mathematical processing can be described as follow:

The diffusion can be seen only in the radius direction of a sphere, and the diffusion coefficient Ddiff remains constant32,38.

$$\,\frac{\,\partial {C}_{{{{{{\mathrm{O}}}}}}}}{\partial t} \,=\,{D}_{{{{{{{\mathrm{diff}}}}}}}}\left(\frac{{\partial }^{2}{C}_{{{{{{\mathrm{O}}}}}}}}{\partial {r}^{2}}+\frac{2}{r}\frac{\partial {C}_{{{{{{\mathrm{O}}}}}}}}{\partial r}\right)$$
(8)

where CO is the oxygen species concentration in the catalysts, t is the reaction time; r is the radius at the specific position of the catalysts powder.

The oxygen exchange happens at the gas/solid interface, and the process is deemed a first-order reaction.

$${D}_{{{{{{{\mathrm{diff}}}}}}}}\frac{\partial {C}_{{{{{{\mathrm{O}}}}}}}}{\partial r}={k}_{{{{{{{\mathrm{chem}}}}}}}}\left[{C}_{{{{{{\mathrm{O}}}}}}}\left({{\infty }}\right)-{C}_{{{{{{\mathrm{O}}}}}}}\left(t\right)\right],\,r=R$$
(9)

where k is the surface exchange coefficient; CO(∞) is the oxygen species concentration at infinite reaction time; CO(t) is the oxygen species concentration at reaction time t; R is the average radius of the catalyst powder.

The reaction proceeds thoroughly under infinite time.

$${C}_{{{{{{\mathrm{O}}}}}}}={C}_{{{{{{\mathrm{O}}}}}}}\left(0\right),\,t=0$$
(10)
$${C}_{{{{{{\mathrm{O}}}}}}}={C}_{{{{{{\mathrm{O}}}}}}}\left({{\infty }}\right),\,t={{\infty }}$$
(11)

By solving PDE under the above conditions, the temporal and spatial distribution of oxygen concentration can be acquired in the following formula:

$$\frac{{C}_{{{{{{\mathrm{o}}}}}}}\left(t,\, r\right)-{C}_{{{{{{\mathrm{o}}}}}}}\left({{\infty }}\right)}{{C}_{{{{{{\mathrm{o}}}}}}}\left(0\right)-{C}_{{{{{{\mathrm{o}}}}}}}\left({{\infty }}\right)}=\frac{2{{{{{{\mathrm{LR}}}}}}}}{r}{\sum }_{n=1}^{{{\infty }}}\frac{{e}^{\left(-{D}_{{{{{{{\mathrm{diff}}}}}}}}\frac{{{\beta }_{n}}^{2}t}{{R}^{2}}\right)}}{\left[{{\beta }_{n}}^{2}+{L}^{2}-L\right]}\cdot \frac{\sin \left({\beta }_{n}\frac{r}{R}\right)}{\sin \left({\beta }_{n}\right)}$$
(12)

with

$${\beta }_{n}\cot \left({\beta }_{n}\right)+L-1=0$$
(13)
$$\,L=\frac{R{k}_{{{{{{\mathrm{chem}}}}}}{}}}{{D}_{{{{{{{\mathrm{diff}}}}}}}}}$$
(14)

where CO(∞) is the oxygen species concentration at infinite reaction time; CO(t, r) is the oxygen species concentration at reaction time t; CO is the oxygen species concentration in the catalysts, t is the reaction time; r is the radius at the specific position of the catalysts powder; R is the average radius of the catalyst powder. \({\beta }_{n}\) is a dimensionless parameter. L is the characteristic length of the solid one; kchem is the surface exchange coefficient; Ddiff is the diffusion coefficient.

The temporal and spatial distribution of oxygen concentration can be transformed to the temporal distribution of catalyst powder weight by a spherical integral of the oxygen concentration distribution.

$$\frac{m\left(t\right)-m\left(0\right)}{m\left({{\infty }}\right)-m\left(0\right)}=1-{\sum }_{n=1}^{{{\infty }}}\frac{6{L}^{2}{e}^{\left(-{D}_{{diff}}\frac{{{\beta }_{n}}^{2}t}{{R}^{2}}\right)}}{{{\beta }_{n}}^{2}\left[{{\beta }_{n}}^{2}+{L}^{2}-L\right]}\,$$
(15)

where m(0) is the weight of the catalysts at reaction time 0; m(t) is the weight of the catalysts at reaction time t; m(∞) is the weight of the catalysts at infinite reaction time.

The TG data can be fitted with the weight distribution formula to acquire the Ddiff and kchem. The fitting is completed with MATLAB® software with the plotting algorithm in the third part toolbox OPTI-master. All data are checked with correlations and confidence intervals to prevent over-parameterization and ensure credibility.

In situ DFIRTS of the catalysts were obtained by Nicolet 6700 spectrometer with a stainless-steel cell connected to a gas-dosing and evacuation system. About 30 mg of the catalysts was compressed into in situ cell equipped with ZnSe windows. The IR spectra can be collected with a resolution of 8 cm−1.

In situ Raman spectra were recorded using a Renishaw inVia reflex Raman spectrometer with a 532 nm Ar ion laser beam. Samples were pretreated at 300  oC for 2 h to eliminate the presence of hydrated species and then reacted at 600 °C under the flow of 20% C3H8/N2 (20 mL/min).

In situ XRD patterns were recorded using SmartLab equipped with an in situ cell and recorded in the 2θ range of 10–80°. The sample was reduced using 20% C3H8/N2 (20 mL/min) at 600 oC for 60 min. The samples were then reoxidized using air (20 mL/min) for 30 min. During the reduction and reoxidation reaction period, an inert period (40 mL/min of nitrogen) of about 10 min was inserted to prevent the mixing between propane and air. One redox cycle was thus completed. Afterward, the second reaction cycle was started by switching between propane and air flows.

In situ X-ray absorption near-edge structure (QXANES) was performed at the XAFS station in the 1W1B beamline of Beijing Synchrotron Radiation Facility (BSRF). Each spectrum was collected under the normal mode and the sampling time was 3 min. During the in situ experiments, the sample was reduced using 20% C3H8/N2 (20 mL/min) at 600 oC, and a V foil was employed for energy calibration. The XANES spectra of metallic V (V foil), V2O5, VO2, and V2O3 were employed as the references. All the data were analyzed by the software of Athena.

Computational details

All the spin-polarized DFT calculation was performed by using the Vienna ab initio simulation package (VASP, 5.4.4 version39), employing the projector augmented-wave model. To depict the exchange-correlation effect, we employed the Perdew, Burke, and Ernzerhof (PBE) functional within the generalized gradient approximation (GGA)40. In order to correct the on-site Coulomb correlation of occupied Ce 4f and V 3d orbitals, the Hubbard U corrections were adopted for both Ce and V elements, using the model proposed by ref. 41. The Ueff (Ueff = Coulomb (U) – exchange (J)) values are 4.5 eV for Ce and 3.2 eV for V42. The valence wave functions were expanded by plane wave with a cutoff energy of 400 eV. J. Sauer et al. investigated the complex VOx-CeO2 system using their global optimization algorithm42. We conducted our calculation based on the optimal structures of 4 × 4 three layers ML - V2O5 and ML - VO2 they found. The Brillouin zone was sampled using only the zone center Γ for geometry optimization and self-consistent calculations. All the structures were optimized until the force on each atom was less than 0.02 eV Å−1. The relaxation of the electronic degrees of freedom will be stopped if the total energy change between two steps is smaller than 10−4 eV.

H2(g) and H2O(g) were used as a reference for the calculation of oxygen vacancy formation energy:

$$\,\triangle {E}_{v}={E}_{{{{{{{\mathrm{surface}}}}}}}\,{{{{{{\mathrm{with}}}}}}}\,{{{{{{\mathrm{Ov}}}}}}}}+{{{{{{\mathrm{H}}}}}}}_{2}{{{{{\mathrm{O}}}}}}\left(g\right)-{E}_{{{{{{{\mathrm{clean}}}}}}}\,{{{{{{\mathrm{surface}}}}}}}}-{{{{{{\mathrm{H}}}}}}}_{2}\left(g\right)$$
(16)

In potential energy diagrams, the energy of C3H8 in the gas phase is taken as a reference, and the energy of intermediates is corrected with H2 in the gas phase. The adsorption energy is defined as:

$$\triangle {E}_{{{{{{{\mathrm{ads}}}}}}}}={E}_{{{{{{{\mathrm{surface}}}}}}}+{{{{{{\mathrm{C}}}}}}}_{3}{{{{{{\mathrm{H}}}}}}}_{x}}+\left(4-0.5x\right){E}_{{{{{{{\mathrm{H}}}}}}}_{2}(g)}-{E}_{{{{{{{\mathrm{clean}}}}}}}\,{{{{{{\mathrm{surface}}}}}}}}-{E}_{{{{{{{\mathrm{C}}}}}}}_{3}{{{{{{\mathrm{H}}}}}}}_{8}(g)}$$
(17)

Moreover, all barrier calculations were performed by using the climbing-image nudged elastic band method (NEB)43 and further improved by the dimer method44. The activation barrier Ea was calculated based on the following equation:

$$\triangle {E}_{a}={E}_{{{{{{{\mathrm{TS}}}}}}}}-{E}_{{{{{{{\mathrm{IS}}}}}}}}$$
(18)

Process simulation

Process simulation was performed with ASPEN Plus. A pair of reactors is applied for the chemical looping oxidative dehydrogenation (CL-ODH) process, of which one is the oxidative dehydrogenation reactor, and the other is the regenerator. Traditional propane dehydrogenation (PDH) process (Oleflex technology) is operated Oleflex process makes use of four adiabatic moving-bed reactors in series to convert paraffin into olefin, which is followed by a continuous catalyst regenerator, and the reaction heat is supplied by interstage reheating furnace1,45. Downstream includes quenching, compression, deep freezing, PSA, Light/heavy hydrocarbon separation, and P-P splitter units. In a chemical looping scheme, the reaction and regeneration of redox catalyst are isolated spatially or temporally. Propane is converted to propane in a dehydrogenation reactor. Meanwhile, the deactivated catalyst circulates between the reactor and regenerator by means of transfer lines9,46. Heat supply from the regeneration and coke combustion, as well as additional fuel gas. Downstream includes CO2 generation for the introduction of lattice oxygen from an additional CO2 removal unit (amine absorption process). The processes are calculated with the feed of 37500 kg/hr (96 wt.% propane, 2 wt% ethane, 2 wt% butane).