Role of CO2 During Oxidative Dehydrogenation of Propane Over Bulk and Activated-Carbon Supported Cerium and Vanadium Based Catalysts

CeO2, V2O5 and CeVO4 were synthesised as bulk oxides, or deposited over activated carbon, characterized by XRD, HRTEM, CO2-TPO, C3H8-TPR, DRIFTS and Raman techniques and tested in propane oxidative dehydrogenation using CO2. Complete oxidation of propane to CO and CO2 is favoured by lattice oxygen of CeO2. The temperature programmed experiments show the ~ 4 nm AC supported CeO2 crystallites become more susceptible to reduction by propane, but less prone to re-oxidation with CO2 compared to bulk CeO2. Catalytic activity of CeVO4/AC catalysts requires a 1–2 nm amorphous CeVO4 layer. During reaction, the amorphous CeVO4 layer crystallises and several atomic layers of carbon cover the CeVO4 surface, resulting in deactivation. During reaction, V2O5 is irreversibly reduced to V2O3. The lattice oxygen in bulk V2O5 favours catalytic activity and propene selectivity. Bulk V2O3 promotes only propane cracking with no propene selectivity. In VOx/AC materials, vanadium carbide is the catalytically active phase. Propane dehydrogenation over VC proceeds via chemisorbed oxygen species originating from the dissociated CO2.


Introduction
Propene is an important commodity chemical in the petrochemical industry and is used for the synthesis of polypropylene, propylene oxide, acrylonitrile, cumene, butyraldehyde, acrylic acid, etc. It is currently produced in the amount of ~ 100 Mt/year, mainly by naphtha steam cracking and FCC (fluidised catalytic cracking) technologies. In recent years, a growing gap between the demand and supply of propene is becoming apparent, which calls for the implementation of additional propene production pathways [1]. Catalytic oxidative dehydrogenation of propane using O 2 or CO 2 emerge as possible solutions. The propane-O 2 ODH reaction is thoroughly researched and supported VO x catalysts have been the focus of many experimental [2,3] and theoretical studies [4]. The support strongly influences the performance of these catalysts and precise active site architecture is required to promote propane dehydrogenation instead of total combustion to CO x [1]. Also, propane conversion and propene selectivity are strongly inversely connected due to facile activation of sp 2 hybridised C=C bond in propene, which promotes its further oxidation.
The propane-CO 2 ODH reaction (reaction 1) has some benefits, as well as drawbacks. It reduces CO 2 to CO and could be implemented for CO 2 valorisation and reduction of anthropogenic CO 2 emissions.
The reaction is endothermic, thermodynamically restricted, and favours carbon formation in a broad range of reaction conditions [5]. Thermodynamic analysis has shown that surplus CO 2 in the feed has the potential to alleviate the restricted equilibrium propane conversion and also lowers the tendency for carbon accumulation [6].
Propene weight time yield (gC 3 H 6 /g cat h) is about two orders of magnitude lower with propane-CO 2 ODH compared to propane-O 2 ODH, which calls for the development of more active catalysts [7,8]. Catalyst optimisation should start with understanding the active sites (requirements for propane and CO 2 activation), and continue by steering the catalyst synthesis towards their highest abundance.
Significantly higher propene selectivity (85 vs 20%) at 50% propane conversion demonstrates CO 2 as a superior oxidant in the propane ODH reaction [7,9]. The high selectivity to the desired alkene originates from a selective formation of nucleophilic monoatomic O 2− ions when CO 2 dissociates on the catalyst's surface. Contrary, dissociation of O 2 leads to the formation of short-lived electrophilic oxygen species (O 2 − and O − ) before they are transformed to O 2− . These electrophilic oxygen species favour deep alkane/ alkene oxidation [10,11]. (1) Several metal oxides (bulk and supported Ga 2 O 3 , Cr 2 O 3 , and Fe 2 O 3 ) were tested by Michorczyk and Ogonowski [12] in the propane-CO 2 ODH reaction. Ga 2 O 3 was identified as the most effective, with propene yield equal to 30% at 600 °C.
Zou et al. [13] investigated the effect of chromium loading in CrO x /SiO 2 catalysts and established the following decreasing order of propane conversion and propene selectivity: 2.5 wt% CrO x /SiO 2 > 5 wt% CrO x /SiO 2 > 10 wt% CrO x /SiO 2 . UV-Vis spectroscopic analysis indicated that ODH activity correlates with highly dispersed chromates, such as Cr 6+ ions in the form of mono, di and polychromates. With increasing chromium loading above 3 wt%, the presence of bulk Cr 2 O 3 was identified. Deactivation of the catalyst with time on stream was observed and related to coke accumulation and changes in chromate speciation from oligomeric to bulk crystalline Cr 2 O 3 .
Martinez Huerta et al. [14] studied catalysts containing 2 and 5 wt. % vanadium dispersed over CeO 2 for ethane-O 2 ODH reaction. The operando Raman-GC analysis showed stepwise dispersion of V 2 O 5 into surface VO x , before the formation of the amorphous CeVO 4 phase. In the CeVO 4 phase, vanadium maintains its 5 + oxidation state and redox activity is related only to Ce 4+ ↔ Ce 3+ . The initially formed amorphous CeVO 4 converts into a well crystalline CeVO 4 phase as the reaction temperature increases. These authors also suggest that the V-O-Ce bond must be the active phase. Increased crystallinity decreases the number of exposed V-O-Ce bonds due to the lower surface-to-volume ratio, which accounts for catalyst deactivation.
Ascoop et al. [15] studied a WO x -VO x /SiO 2 catalyst and using isotopic labelling confirmed that in parallel to an oxidative dehydrogenation pathway (reaction 1), also reverse water gas shift reaction (H 2 + CO 2 ↔ CO + H 2 O) and a direct, non-oxidative propane dehydrogenation reactions (C 3 H 8 → C 3 H 6 + H 2 ) occur over the catalyst at 600 °C. According to their DFT calculations, C-H bond activation in propane is the rate-limiting reaction step, whereas catalyst re-oxidation (V 3+ to V 4+ ) with CO 2 occurs much faster.
Turakulova et al. [16] observed that the interaction between supported VO x and Ce 0.46 Zr 0.54 O 2 results in the formation of CeVO 4 , which is the active phase responsible for propane ODH. The full oxidation of propane to CO x is reported to be catalysed at the CeZrO 2 surface, whereas propene is formed over CeVO 4 . The oxygen exchange properties of the catalyst play an essential role in the ODH reaction. Contrary to Ascoop et al. [15], Turakulova et al. report that re-oxidation of the active sites is the rate-limiting step of the process, which is associated with the redox properties of the VO x /Ce 0.46 Zr 0.54 O 2 catalyst.
Nowicka et al. [17] studied propane-CO 2 ODH reaction over Pd/CeZrAlO x catalysts and showed that over ceriabased catalysts, the reaction proceeds via the Mars van 1 3 Krevelen mechanism in parallel with the RWGS reaction. The high propene selectivity (95% at ~ 3% propane conversion) was attributed to the exclusive formation of selective oxide (O 2− ) ions when CO 2 dissociates at the surface oxygen vacancy sites of CeO 2 . The role of Pd is to enhance the reducibility and to accelerate the re-oxidation of the CeO 2−x surface. Strong catalyst deactivation was attributed to carbon accumulation which blocks the active sites and incomplete catalyst oxidation by CO 2 .
Carbon materials with a high surface area and large pore volume have an increasing number of applications as catalyst supports and adsorbents [18][19][20]. They can be synthesized in a variety of morphologies (nanotubes, ordered mesoporous structures, etc.) and their surfaces can be functionalized with adatoms (V and N) or oxygen containing groups, giving rise to catalytic activity in many reactions, including oxidative dehydrogenation of propane [21][22][23].
Our common observation in many propane-CO 2 ODH studies is that catalytic tests were run at relatively high reaction temperatures (600-700 °C) in order to activate CO 2 and ensure reasonable conversions. Under such conditions, thermal (noncatalytic) reactions strongly contribute to the observed propane conversion and influence the distribution of reaction products [7,15].
This work is focused on the systematic analysis of propane and CO 2 interaction with bulk and activated carbon (AC) supported VO x , CeO 2 and CeVO 4 catalysts. Vanadia based materials are very often investigated catalysts in the oxidative dehydrogenation of parafines [1], whereas ceria is known for its reactivity towards CO 2 [24]. Consequently, these materials, as well as their mixed oxide phase (CeVO 4 ) were the subject of this investigation. Structural, redox, and chemical nature of the active phases were varied by changing their loading on the AC support and were correlated to activity, stability, and olefin selectivity in the propane-CO 2 ODH reaction.

Synthesis of Catalysts
Bulk V 2 O 5 was synthesised by calcination of vanadium acetyl-acetonate (Sigma Aldrich, purity 99%). Bulk CeVO 4 was synthesised by completely dissolving 92 mg of NH 4 VO 3 (Sigma Aldrich, p.a.) in 65 ml of ultrapure water (resistivity of 18.2 MΩ, Elga Purelab Option-Q), while the solution was heated to 80 °C and stirred with a magnetic stirrer. To prevent water evaporation, the glass beaker was covered with a petri dish. Once dissolved, 1 drop of concentrated HNO 3 was added. In a separate glass beaker, 340 mg of Ce(NO 3 ) 3 × 6H 2 O (Sigma Aldrich, purity 99%) was dissolved in 10 ml of ultrapure water. After complete dissolution, the prepared solutions were mixed, resulting in an instantaneous formation of a yellow-brown precipitate. The suspension was stirred at 80 °C for an additional 2 h. After cooling, filtering and washing 3 times with ultrapure water, the suspension was dried overnight at 70 °C in a laboratory drier. Pure CeO 2 was synthesised by dissolving 4.9 g of Ce(NO 3 ) 3 × 6H 2 O (Sigma Aldrich, purity 99%) in 84 ml of ultrapure water. This solution was mixed with 140 ml of 0.1 M aqueous NaOH and stirred for 30 min. Afterwards, it was transferred to Teflon® lined autoclaves (~ 35 ml volume), which were placed into a laboratory drier preheated to 180 °C and kept at this temperature for 24 h. The precipitate was filtered, washed 3 times with ultrapure water, and dried overnight at 70 °C in a laboratory drier. All bulk catalysts were calcined 4 h at 600 °C (Nabertherm, model LT 9/11, heating ramp of 2 °C/min) in static air.
Synthesis of catalysts containing CeVO 4 (10-40 wt% CeVO 4 ) supported on activated carbon (AC) was similar to the one of bulk CeVO 4 with the following exceptions: 0.7 g of activated carbon, finely ground in an agate mortar, was added to the aqueous NH 4 VO 3 solution. After filtration the samples were dried overnight at 70 °C. The VO x (2-12 wt% vanadium) was deposited over AC by dissolving appropriate amounts of NH 4 VO 3 in 10 ml of ultrapure water at 80 °C. Afterwards, 0.7 g of AC was added, stirred for another 2 h at room temperature, filtered and dried overnight at 70 °C in a laboratory drier. CeO 2 (7-17 wt% cerium) was deposited over AC by dissolving appropriate amounts of Ce(NO 3 ) 3 × 6H 2 O in 1 ml ultrapure water. Afterwards, 0.7 g of AC was added, mixed for 2 h and dried overnight at 70 °C. All AC supported catalysts were calcined (4 h at 600 °C, 2 °C/min heating ramp) in a tubular oven (Carbolite, model HST 12/400) in Argon (20 l/h) to prevent AC oxidation. The CeVO 4 content in the CeVO 4 /AC catalysts was selected so the actual content of vanadium or cerium in the CeVO 4 phase covered a similar range as in CeO 2 /AC or VO x /AC catalysts. Actual content of vanadium, cerium and CeVO 4 is shown in Table S1.

Characterisation and Catalytic Testing
XRD analyses were performed on a PANalytical X'pert PRO diffractometer using Cu Kα radiation (λ = 0.15406 nm) between 2theta angles of 10° and 80° with the step size of 0.034°. The BET specific surface area, total pore volume and average pore size were determined using N 2 adsorption/ desorption isotherms at 77 K (Micromeritics, model TriStar II 3020). The samples were degassed before measurements using a SmartPrep degasser (Micromeritics) in a N 2 stream 1 h at 90 °C followed by 4 h at 180 °C. Actual vanadium, cerium and CeVO 4 contents were determined from the mass of oxidic residue remaining after complete oxidation of the AC support (TGA apparatus, model STA6000 by Perkin Elmer) by heating the samples in air (25 ml/min) with a heating ramp of 10 °C/min to 800 °C. The carbon content, accumulated on the bulk catalyst during reaction was analysed by CHNS analytical technique (Series II CHNS analyser, model 2400 from Perkin Elmer).
Prior to TEM analysis, samples were dispersed in absolute ethanol and sonicated for 30 s, then directly transferred to Cu lacey carbon support grids. As prepared samples were analysed by 200 kV transmission electron microscope (TEM, JEM-2010F, Jeol Inc.) and by C s -corrected (TEM, Titan Themis G 3 , FEI Inc., equipped with CEOS CETCOR aberration corrector) operating at 300 kV. Amount of carbon deposited on the bulk unsupported catalysts during the catalytic reaction was analysed using CHNS elemental analyser (Perkin Elmer, model 2400).
Transient interaction of bulk CeVO 4 , CeO 2 and V 2 O 5 with propane and CO 2 was analysed with Diffuse Reflectance Infrared Furrier Transformed Spectroscopy (DRIFTS) analysis (Perkin Elmer, model Frontier) equipped with Dif-fusIR cell from Pike Scientific. Approximately 10 mg of finely powdered sample was pre-treated in 20%O 2 /He stream at 550 °C for 15 min. Afterwards, the sample was exposed to the following sequence of atmospheres: 20% O 2 /He → 20% C 3 H 8 /He → 20% CO 2 /He. Sample spectra were recorded continuously (4 s between scans) in the range between 500 and 4000 cm −1 , 8 accumulations per scan and spectral resolution of 4 cm −1 .
Raman analysis of the powdered catalysts was performed in the spectral range from 70 to 3700 cm −1 using Witec Alpha 300 spectrometer that employed green laser with excitation wavelength of 532 nm and resolution of 4 cm −1 at 30 mW laser power.
Catalytic experiments were performed in a PID Eng&Tech reactor system using a quartz tubular reactor (10 mm I.D.). The powdered catalyst (300 mg) was positioned between two flocks of quartz wool. Reaction temperature was measured with a K-type thermocouple, which was covered by a quartz sleeve, thus minimising the heated volume inside the reactor above the catalyst bed. A quartz rod was used to minimise the reactor volume after the catalyst bed. This reduced the occurrence of thermal, noncatalytic reactions (verified with a blank experiment which showed < 0.5% propane conversion at 550 °C. Before the catalytic reaction, the catalyst was heated using a heating ramp of 10 °C/min in a flow of CO 2 (Linde, purity 5.3) and He (Linde, purity 5.0) with a flowrate of 10 ml/min each, until reaching the reaction temperature of 550 °C. Afterwards, the propane flow (Linde, purity 3.5, 10 ml/min) was added and 15 min were allowed before starting the GC analysis to allow for stabilization of gas concentrations. Agilent 7890A gas chromatograph (equipped with Molesieve 5A and Poraplot Q columns and two TCD detectors) was used for qualitative and quantitative analysis of gas stream. The carbon mass balances were in all cases between 85 and 104%. Operation in kinetic regime was ensured by plotting the propane reaction rate versus progressively increasing C 3 H 8 / CO 2 /He flowrate over a constant mass of catalyst (Fig. S2). Details on calculation of propane conversion and product selectivities are provided in the Supplementary information.

Catalytic Activity of Bulk CeO 2 , V 2 O 5 and CeVO 4
Catalytic performance of all materials during propane-CO 2 ODH reaction is summarized in Tables 1 and 2. Main reaction product over bulk CeO 2 was CO, along with CH 4 and H 2 (71, 10 and 13% selectivity, respectively). The olefin selectivity was low (12% for propene and 4% for ethene).
The bulk V 2 O 5 deactivated quickly: the initial propane conversion of 9.5% stabilised at 3.2% after 240 min TOS (Fig. 1a). Also, C 3 H 6 selectivity decreased from 27 to 0% (Table 1 and Fig. 1b) and CO dropped to zero in 140 min (not shown). Contrary, selectivities for H 2 and CH 4 increased slowly during the experiment from 9 to 25% and 31 to 61%, respectively. This transient behaviour of the V 2 O 5 catalyst indicates drastic changes in contribution of several possible reaction pathways occurring during propane-CO 2 ODH reaction: (i) oxidative dehydrogenation with participation of lattice oxygen producing propene and water, (ii) oxidative dehydrogenation with CO 2 acting as the oxidant, producing propene, CO and water (iii) total oxidation of propane to CO x and water, (iv) nonoxidative dehydrogenation pathway producing propene and hydrogen and (v) propane cracking to CH 4 , H 2 and carbon which deposits over the catalyst's surface (reactions 1-4). The relevance of these reactions will be discussed in detail in the following sections.
Activity of bulk CeVO 4 was stable, as well as selectivity for all reaction products: C 3 H 6 , CH 4 , H 2 and C 2 H 4 at 25, 31, 26 and 18%, respectively. No CO was formed, which indicates the lattice oxygen or CO 2 participation (Reaction 1) and RWGS reaction (CO 2 + H 2 ↔ H 2 O + CO) are not occurring over this catalyst. The non-oxidative dehydrogenation and propane/propene cracking dominate the reaction product distribution over bulk CeVO 4 . Results of a blank experiment (Fig. S3) show that no conversion is taking place at 550 °C in the empty reactor. This confirms the non-oxidative dehydrogenation reaction is a consequence CeVO 4 .

CeO 2 /AC Catalysts
Propane conversions and propene selectivities were notably higher over CeO 2 /AC catalysts compared to bulk CeO 2 ( Table 1). A notable (~ 50%) decrease in propane conversion in the initial 60 min of reaction for 7CeO 2 /AC ( Fig. 2a) was accompanied by an increase in propene selectivity.

VO x /AC Catalysts
Catalytic stability of VO x /AC catalysts is notably improved compared to bulk V 2 O 5 ( Table 1, Figs. 1A and 2A). Propene selectivities ranged between 40 and 51% and were marginally influenced by vanadium content, which was varied between 2 to 12 wt%. Selectivities for CO were between 41 and 50%, for H 2 ranged between 8 and 10%, for CH 4 between 5-6% and for C 2 H 4 between 1-2% (Table 2). Contrary to the complete loss of propene selectivity over bulk V 2 O 5 , the propene selectivity over VO x /AC catalysts was stable during the 240 min of reaction (Fig. 2b). This indicates markedly different catalytic behaviour of supported vanadium species compared to bulk V 2 O 5 .

CeVO 4 /AC Catalysts
Testing of CeVO 4 /AC catalysts revealed a positive correlation between CeVO 4 content and catalytic activity, as well as propene selectivity. The 30CeVO 4 /AC sample achieved 11.4% propane conversion and 57% propene selectivity at 60 min TOS (Table 1 and Fig. 2). This is clearly superior compared to all tested bulk and supported VO x , CeO 2 and CeVO 4 materials. For all CeVO 4 /AC catalysts, a slow continuous deactivation was observed with time on stream. Selectivities for CO ranged between 22 and 59%, H 2 ranged between 8 and 18%, for CH 4 between 5 and 11% and for C 2 H 4 up to 4% (Table 2). Catalytic propane dehydrogenation reaction often suffers from poor stability, which is usually caused by carbon build-up on the catalyst, which blocks the active sites [6]. A long-term catalytic test was performed on the 20CeVO 4 / AC catalyst (Fig. 3). Continuous deactivation was observed in the first 30 h of reaction; the catalyst lost 71% of its initial activity (based on the drop of propane conversion). Catalyst deactivation was accompanied by a slow rise in propene, CH 4 , C 2 H 4 and H 2 selectivities (Fig. 3), as well as decrease of CO selectivity (not shown). This indicates a slow transition from an oxidative propane-CO 2 to a non-oxidative propane dehydrogenation pathway. Finally, conversions of propane and CO 2 stabilised at 3% each. Very similar behaviour was observed over Pd/CeZrAlO x catalysts during propane-CO 2 ODH reaction at 600 °C by Nowicka et al. [17] An important aspect for discussion of catalytic performance of AC supported catalysts is experimental verification that support gasification does not contribute to measured reaction products. Under simulated reaction conditions (4 h at 550 °C in a 30 ml/min total flow consisting of 33% C 3 H 8 , 33% CO 2 and 33% He) inside a thermogravimetric apparatus, the mass of 20CeVO 4 /AC catalyst increased by 0.58 wt. % (Fig. S4A). This experiment confirmed a small amount of carbon was deposited on the catalyst during the 4 h of reaction. Additionally, the TGA-TPO experiment of fresh and spent 20CeVO 4 /AC catalysts after 45 h of reaction showed the weight fraction of carbon in the sample increased by 0.62 wt% (Fig. S4B). These two experiments revealed no gasification of the AC support during catalytic reaction and that carbon accumulation on the catalyst is low and occurs mainly in the initial 4 h of reaction. Spent bulk catalysts after 4 h of propane-CO 2 ODH reaction contain 1.6, 1.4 and 1.4 wt% of carbon for CeVO 4 , V 2 O 5 and CeO 2 , respectively. Propene selectivity at comparable propane conversions is shown in Fig. S5 for all catalysts, which revealed increasing selectivity in the following order: CeVO 4 > V 2 O 5 > CeO 2 .

N 2 Physisorption and XRD Analysis
Morphological properties analysed by N 2 physisorption technique are compiled in Table 2. The V 2 O 5 and CeVO 4 are mesoporous with total low pore volumes and BET specific surfaces, whereas specific surface area of CeO 2 is much higher. After 4 h of reaction, the specific surface area of CeO 2 drops by about 50%, whereas changes for V 2 O 5 and CeVO 4 were much smaller. The specific surface area and pore volume of supported catalysts are dominated by the microporous activated carbon support. With increasing content of the active phase (VO x , CeO 2 or CeVO 4 ), a continuous decrease of specific surface area and pore volume are observed, which is in line with a progressively larger contribution from the deposited oxides.
XRD analysis was performed on the fresh and spent catalysts (Figs. 4, 5 and S6). In bulk CeVO 4 and CeVO 4 /AC samples (Fig. 4a), only a CeVO 4 phase (PDF 00-012-0757) was observed. No diffraction lines of crystalline CeO 2 , V 2 O 5 or V 2 O 3 could be identified. The Scherrer equation was applied and the average scattering domain size of the CeVO 4 crystallites was calculated based on its most intense diffraction line    (Fig. S6) showed no structural changes in the CeVO 4 and negligible sintering since the average crystallite size measured 24, 25 and 26 nm, respectively. The average pore diameter of AC supported catalysts is below 2 nm, which indicates that CeVO 4 crystallites observed by XRD, which are more than an order of magnitude larger, reside in the interparticle voids of the AC support. XRD analysis of bulk V 2 O 5 sample before and after propane-CO 2 ODH reaction (Fig. 4b) shows that the initially present V 2 O 5 phase (PDF 01-089-0612), is reduced and quantitatively transformed to V 2 O 3 (PDF 01-071-0342).
Results for the VO x /AC samples before and after catalytic tests are shown in Fig. 5a. The broad reflection peaks between 20°-30° and 40°-50° are visible on all samples and originate from activated carbon. The diffraction lines at 21.8°, 26.6° and 35.8° can be ascribed to SiO 2 cristobalite*, quartz # and graphite $ (PDF 01-089-3435, 01-085-1780 and 01-075-2078, respectively). These crystalline phases are present in the pristine activated carbon. The amount of SiO 2 in the AC was determined to be 1.1 wt% (thermogravimetric heating of pristine AC sample in air, followed by SEM-EDXS analysis of inorganic, non-combustible residue). No diffraction peaks from vanadium containing crystalline phases could be identified in the 2VO x /AC and 5VO x /AC samples.
In the fresh 12VO x /AC sample weak diffraction lines at characteristic positions for V 2 O 3 became apparent (PDF 01-071-0342). Formation of V 2 O 3 is triggered by sample calcination in argon (see experimental section). The intensity of diffraction lines belonging to V 2 O 3 increased after 4 h of reaction, revealing its sintering. Figure 5b shows XRD results of bulk and supported CeO 2 /AC catalysts. In all samples, only diffraction lines belonging to CeO 2 (PDF 00-034-0394) were identified. Average crystallite size in bulk CeO 2 was 15 nm (calculated by the Scherrer equation).
Broad diffraction lines characteristic of CeO 2 were identified in the CeO 2 /AC catalysts with the average crystallite size of 4 nm which did not change with increasing CeO 2 content from 7 to 17 wt%. Negligible growth of CeO 2 crystallites occurred during the catalytic reaction, as average crystallite size of CeO 2 after reaction increased to 5 nm ( Table 3). The AC support efficiently prevented sintering of the deposited CeO 2 during the reaction.

TEM Analysis
To explore the rapid drop of the catalytic performance, the 20CeVO 4 /AC sample was analysed before and after 45 h of reaction (Fig. 6). In the fresh sample, the individual CeVO 4 crystals are agglomerated in larger clusters, deposited on the outer surface of AC support. The average CeVO 4 particle size measured 24 nm (Fig. S7A), which is in line with XRD estimation (Table 2). Selective area electron diffraction (SAED) pattern analysis confirmed the presence of a crystalline CeVO 4 phase only (Fig. 6, inset)  In order to visualize and analyse the catalytically active phase in the 5VO x /AC catalyst, TEM-SAED analyses were performed (Fig. 7). It was observed that the activated carbon was covered by cubic crystallites measuring 3-10 nm in size. Selective area electron diffraction (SAED) pattern analysis identified these crystals as a cubic vanadium carbide phase (Fig. 7d). No vanadium oxides (crystalline V 2 O 3 or V 2 O 5 ) could be identified.

Temperature Programmed Reduction with Propane (C 3 H 8 -TPR)
Catalyst interaction with propane ( Fig. 8) and CO 2 ( Fig. 9) was tested according to the protocol shown in Fig. S1. These experiments investigated: (a) the ability of lattice oxygen to react with propane, (b) the re-oxidation of the reduced catalyst by CO 2 , and how these processes are influenced by the morphology of the VC, VO x , CeO 2 and CeVO 4 phases. During the C 3 H 8 -TPR experiments, CO x and water were identified in parallel to propane conversion (drop of propane signal in Fig. 8), indicating participation of lattice oxygen in the propane conversion. Propane signal as a function of temperature over CeO 2 containing catalysts is shown in Fig. 8a. Several broad, lowintensity bands are apparent between 150 and 500 °C in CeO 2 /AC samples, which is absent in bulk CeO 2 . These are likely related to propane reacting with coordinatively unsaturated surface lattice oxygen, which are more reactive in 4 nm CeO 2 crystallites (present in all CeO 2 /AC samples), compared to bulk CeO 2 (15 nm). Propane conversion lights-off between 495 and 508 °C for AC supported samples, whereas a higher temperature (530 °C) is required for propane activation over bulk CeO 2 .
Bulk CeVO 4 shows negligible activity for reduction by propane up to 600 °C (black line in Fig. 8b). However, propane is able to reduce the CeVO 4 /AC samples already between 480 and 505 °C, producing propene and water.
Over bulk V 2 O 5 , propane is oxidized extensively in the low temperature region between 350 and 500 °C producing propene and water. These results are consistent with the oxidative dehydrogenation pathway involving lattice oxygen and reduction of V 2 O 5 to V 2 O 3 . The second propane consumption signal between 500 and 600 °C produces methane and especially hydrogen. This is consistent with catalytic tests at TOS ˃ 140 min (Fig. 1). Propane cracking inevitably produces carbon deposition, but this could not be analysed. The VO x /AC catalysts containing either only VC (2VO x /AC and 5VO x /AC) or a combination of VC and V 2 O 3 (12VO x / AC) are starting to decompose propane at about 485 °C. In all cases, propane conversion produces propene and hydrogen (Fig. 8c).
In all analysed samples, higher propane conversion during C 3 H 8 -TPR experiments (Fig. 8) generally correlates with higher catalytic activity (Table 1).

Temperature Programmed Oxidation with CO 2 (CO 2 -TPO)
In situ oxidation experiments with CO 2 were performed on the catalysts previously exposed to the C 3 H 8 -TPR protocol (Fig. S1). During CO 2 -TPO, CO 2 is dissociated over the catalyst to CO and O. The latter oxidizes the catalyst and  Fig. 9.
The re-oxidation of bulk CeO 2-x by CO 2 took place with appearance of two intense CO peaks centred at 460 and 620 °C (Fig. 9a). On the other hand, the re-oxidation of CeO 2-x /AC samples was initiated at between 400 and 450 °C with a slow, continuous rise of the CO signal until the final temperature of 700 °C was reached. This reveals very different dynamics of CeO 2 oxidation which is strongly related to the size of CeO 2 crystallites.
Bulk CeVO 4 does not react with CO 2 up to 700 °C (Fig. 9b), which is in line with inertness of this sample during C 3 H 8 -TPR and catalytic experiment showing no CO among reaction products. Over CeVO 4 /AC catalysts, CO starts to appear at 500 °C and its amount continuously increases with increasing CeVO 4 loading. Oxidation of bulk V 2 O 3 (V 2 O 5 catalyst after C 3 H 8 -TPR) and VO x /AC catalysts with CO 2 was initiated at 570 °C (Fig. 9c), which reveals that at the reaction temperature of 550 °C, V 2 O 3 cannot be re-oxidized with CO 2 . This is in line with results of Ascoop et al. [15] who report that oxidation of WO x -VO x /SiO 2 catalysts with CO 2 at 600 °C can only transform V 3+ to V 4+ . Also, the VC phase is inert towards CO 2 at 550 °C.
The C 3 H 8 -TPR and CO 2 -TPO results revealed that the reduction and re-oxidation of CeO 2 , and CeVO 4 phases are strongly dependent on their morphology. Smaller CeO 2 crystals (4 nm) in CeO 2 /AC are more readily reduced by propane compared to bulk CeO 2 (15 nm), whereas during re-oxidation by CO 2 , the situation is reversed. Both processes are possible at 550 °C, which was also the temperature during propane-CO 2 ODH reaction.
The CeVO 4 /AC samples can be reduced by propane and oxidized by CO 2 at 550 °C. This ability is absent in bulk CeVO 4 , which has negligible activity for propane and CO 2 activation.
In the case of VO x /AC catalysts, the vanadium containing phase depends on the vanadium content (V 2 O 5 in bulk sample, VC in 2VO x /AC and 5VO x /AC and a mixture of V 2 O 3 and vanadium carbide in 12 VO x /AC sample). All VO x /AC and V 2 O 5 samples can dehydrogenate propane: bulk V 2 O 5 already at 350 °C through the oxidative dehydrogenation pathway with lattice oxygen participation, whereas the crystalline VC is active above 485 °C through a non-oxidative pathway. Re-oxidation with CO 2 takes place above 570 °C.

Time Resolved Isothermal DRIFTS Experiments
Transient behaviour of CeO 2 , CeVO 4 and V 2 O 5 during isothermal reduction with propane and re-oxidation with CO 2 was investigated with an in situ DRIFTS analysis at 550 °C (Figs. 10, 11 and S11-S17). Only bulk oxides were analysed due to the black colour and total absorbance of all AC supported catalysts. No interaction of propane or CO 2 with CeVO 4 was observed and these results are consequently not shown.
Exposure of reduced CeO 2-x to CO 2 (Figs. 10b, S13 and S14) leads to a fast increase of two broad polydentate carbonate bands centred at 1454 and 1350 cm −1 (Fig. S14A). These carbonates are observed regularly over ceria and are thermally very stable, as bands remain stable during the reoxidation experiment. Also, instantaneous formation of CO is observed; it goes through a maximum and slowly starts to decline after 300 s of CO 2 exposure. This confirms CO 2 dissociation and oxygen vacancy replenishment, leading to oxidation of CeO 2−x . Existence of the band at 1765 cm −1 (CO adsorbed on reduced Ce 3+ sites [28]) after 10 min of oxidation at 550 °C (Fig. S14A) suggests oxidation of CeO 2-x with CO 2 does not proceed to completion.
Upon exposure of V 2 O 5 to propane (Figs. 11a, S15 and S16), the V=O vibration overtones at 2005 and 1969 cm −1 diminish after 53 s, indicating surface reduction [29]. In parallel to propane introduction, the CO 2 signal (2340 cm −1 ) lights off, goes through a maximum at 135 s is much smaller compared to CO 2 and appears only after the surface V=O overtone signal disappears at 100 s. This time-resolved experiment reveals the oxidation of propane to CO is possible only in the presence of a partly reduced surface, most likely containing V 4+ . On the reduced V 2 O 3 sample, the envelope of signals between 1300 and 1550 cm −1 appears (Fig.  S16A), which belong to gas phase propane; no carbonates are formed. Absence of bands above 1600 cm −1 indicates no carboxylates or bicarbonates are formed. Re-oxidation of V 2 O 3 with CO 2 occurs only marginally, since only about 3% of initial V=O band intensity was achieved after 400 s of oxidation, Fig. 11b, S17).
To summarize, time resolved DRIFTS experiments showed that reduction of CeO 2 with propane at 550 °C is fast, and that instantaneous surface population with polydentate carbonates takes place. Re-oxidation with CO 2 is substantial, but does not proceed to completion. This suggests the working state of the CeO 2 catalyst during the propane-CO 2 reaction is partly reduced.
Reduction of V 2 O 5 is slower (CO 2 peak intensity is reached 60 s after propane introduction, compared to 5 s for CeO 2 ), re-oxidation by CO 2 is negligible. This suggests a slow irreversible transformation of initially present V 2 O 5 into V 2 O 3 . Also, the inability of V 2 O 5 and V 2 O 3 to dissociate CO 2 shows the crucial role of lattice oxygen (nucleophilic O 2− species) for enabling the oxidative dehydrogenation reaction pathway for propene formation. These findings are in line with catalytic tests (stable activity and low propene selectivity over CeO 2 , compared to fast deactivation and total loss of propene selectivity over V 2 O 5 ).

Raman Analysis
In the fresh and spent 2VO x /AC and 5VO x /AC catalysts, no signal below 1000 cm −1 is visible (Fig. 12a). This is due to the absence of V-O bonds in these samples and shows that only vanadium carbide (VC) is present [30]. Also, the VC phase is resistant towards oxidation during propane-CO 2 ODH reaction, thus eliminating this possibility for the observed catalyst deactivation (Table 1). In the fresh 12VO x / AC sample, the bands at 996, 700, 529, 478, 410, 287 and 142 cm −1 are visible, which is consistent with the presence of V 2 O 3 [31]. This is in line with the XRD results (Fig. 5a). In the spent 12VO x /AC sample, four additional bands at 927, 877, 840 and 161 cm −1 are seen, which are characteristic for multi-valent vanadium states as present in V 6 O 13 [32,33]. The formation of newly formed V 6 O 13 phase is likely related to exposure of finely dispersed V 2 O 3 to highly reducing conditions during the propane-CO 2 ODH reaction.
In the CeVO 4 /AC samples (Fig. 12b), only characteristic CeVO 4 Raman bands appear at 217, 257, 369, 463, 775 and 845 cm −1 [14]. In the fresh 30CeVO 4 /AC and 40CeVO 4 / AC samples, weak bands at 998 and 705 cm −1 are visible, which suggests presence of V 2 O 3 in these samples. Their fraction is very likely minor, as they could not be identified through XRD.

Discussion
The initial propene selectivity of CeO 2 /AC catalysts is low (16-27%) and stabilizes after 60 min of reaction at notably higher values (38-50%, Table 1). The re-oxidation of small CeO 2-x crystallites (4 nm), as present in CeO 2 /AC catalysts with CO 2 occurs at higher temperatures compared to bulk CeO 2 (15 nm), revealing it is more difficult to dissociate CO 2 as the CeO 2 size decreases (Fig. 9a). This is consistent with structure sensitivity for CO 2 activation, namely the π-bond in C=O and TOF increases with increasing particle size, as a certain degree of site coordination is required [34,35].
The propene selectivity appears to be influenced by the oxidation degree of CeO 2−x . Lower abundance of reactive surface oxygen sites in partly reduced CeO 2−x crystallites, when supported over AC compared to bulk CeO 2 , provides less active sites for the activation of the C=C bond in propene. Consequently, higher propene selectivities are achieved over nanosized CeO 2-x .
Nowicka et al. [17] ascribed the initial deactivation of Pd/CeAlO x catalysts and a concomitant propene selectivity increase to the consumption of reactive oxygen species stored within the ceria lattice. These authors also suggest that the nonselective oxygen species, which are likely electrophilic, are singly charged interstitial oxygen anions. Once these O − species are consumed, the catalyst can be re-oxidized by CO 2 .
The fast deactivation and complete loss of propene selectivity over bulk V 2 O 5 is related to its quantitative reduction from V 5+ to V 3+ , which leads to a progressive shift from propane ODH to propane cracking reaction. The CO 2 -TPO experiment (Fig. 9c) revealed that bulk V 2 O 3 cannot be reoxidized by CO 2 at 550 °C. This shows the participation of CO 2 in the propane ODH reaction is negligible and the dominant ODH reaction pathway is governed by the availability of lattice oxygen in V 2 O 5 and V 2 O 4 .
The reduction and re-oxidation dynamics of CeO 2 and V 2 O 5 differ considerably, as could be observed by transient isothermal DRIFT spectroscopy analyses (Figs. 10 and 11). Propane interaction with CeO 2 leads to instantaneous and simultaneous CO 2 and CO formation which tail off slowly. Such behaviour is likely connected to the diffusion of lattice oxygen from the bulk to the surface [36], where it participates in the oxidation reactions.
Upon re-oxidation of CeO 2−x with CO 2 , the polydentate carbonate signal stabilizes after 10 s of CO 2 addition, suggesting a kinetic preference for these species. The instantaneous appearance of a CO signal confirms facile CO 2 disproportionation and catalyst re-oxidation at 550 °C.
During V 2 O 5 reduction with propane, the CO 2 signal lights off slowly and reaches a maximum at about 70 s after propane introduction. This suggests higher propane oxidation rates with lattice oxygen of CeO 2 compared to V 2 O 5 , which correlates also with notably higher initial propene selectivity over bulk catalysts (10% over CeO 2 and 27% over V 2 O 5 ). CO appears after V=O overtone vibrations disappear (suggesting the absence of V 5+ on the surface) [37].
At higher vanadium loadings in the VO x /AC catalysts (12 wt%), a separate V 2 O 3 phase is formed which is inert towards CO 2 and does not promote propane dehydrogenation. All VO x /AC catalysts exhibit some deactivation with TOS, but propene selectivity remains constant (Table 1). Also, CO was identified as the reaction product during the entire duration of the catalytic tests and its concentration followed that of propene (Fig. S18), whereas the concentration of all other reaction products remained stable. This suggests that propene and CO are produced by the same oxidative dehydrogenation pathway (reaction 1). Considering that the fresh and spent catalysts contains no lattice oxygen (TEM-SAED, XRD and Raman analyses confirm a VC phase is formed exclusively at vanadium loadings up to  Fig. 12 a Stacked Raman spectra of fresh and spent VO x /AC and b fresh and spent CeVO 4 /AC catalysts. Please refer to online version of this manuscript for colour figure  5 wt. %, whereas VC and V 2 O 3 coexist at 12 wt. % vanadium loading), CO 2 remains as the only possible oxygen source. By combining the C 3 H 8 -TPR, CO 2 -TPO and catalytic results we can postulate that CO 2 activation over the VC crystals is assisted by the adsorbed propane (protons at the methylene and methyl positions). In the absence of these electrophilic species which facilitate CO 2 decomposition [38] and H 2 O formation, CO 2 activation requires temperatures close to 600 °C (Fig. 9c), which is above the reaction temperature used in this work. The deactivation of VO x /AC catalysts is accompanied by a stable propene selectivity, which indicates that the total number of active sites, and not their nature, is changed with TOS. Oxidation of the VC phase during the propane-CO 2 ODH reaction does not occur, due to absence of V-O bonding in Raman spectra of the spent 2VO x /AC and 5VO x /AC catalysts. As a result, deactivation is probably related to blocking of active sites by carbon. Recently, Thakur et al. [39] report of in situ formation of oxycarbide phases (V 2 O 3 and V 8 C 7 ) during exposure of vanadium containing MXene catalysts to CH 4 and CO 2 atmosphere at much higher temperatures (800 °C). The CeVO 4 /AC catalysts showed the highest catalytic performance in terms of propene yields. Upon dispersion of CeVO 4 over activated carbon, a thin (1-2 nm) amorphous phase is formed over the CeVO 4 crystals. Its formation is likely connected to calcination in argon. The amorphous layer is intrinsically defective and the reactivity of oxygen in amorphous mixed metal oxides is higher than in any of its crystalline components [40]. Ruth et al. [41], report that the amorphous part of the multiphase Mo-V-Nb oxide catalyst is particularly important during oxidative dehydrogenation and partial oxidation of ethane.
During oxidative dehydrogenation of ethane, a transformation of VO x /CeO 2 to CeVO 4 was identified by Martinez-Huerta et al. [14]. Based on the operando Raman analysis of the catalyst's structure, they identified the bridging oxygen atom (Ce-O-V) present in the ill-defined CeVO 4 phase as the active site for the rate determining step in the ODH reaction. Exposure to reaction temperatures above 500 °C favours crystallization of the CeVO 4 phase, making it inactive in the ODH reaction [14].
An increase of propane ODH catalytic activity scaled with the CeVO 4 loading. Catalytic activity is in line with the redox ability of the supported CeVO 4 phase, probed by propane and CO 2 (Figs. 8 and 9), which revealed lattice oxygen abstraction and re-oxidation are feasible under the reaction conditions. TEM analysis of the CeVO 4 /AC catalyst after reaction revealed elimination of the amorphous CeVO 4 layer and the deposition of a few layers of carbon on the surface. This change is likely correlated to the observed catalyst deactivation. Since propene selectivity does not alter drastically with TOS (Figs. 2 and 3) despite notable deactivation, the number and not the nature of the active sites is decreased. A recent review of propane ODH with CO 2 by Atanga et al. [5] suggests that indium and especially gallium based catalysts are superior in activity and selectivity compared to chromium, platinum and vanadium based ones. When benchmarking the performance of best performing sample in this work (30CeVO 4 /AC, 15.3% C 3 H 8 conversion, 42% C 3 H 6 selectivity, 6.4% C 3 H 6 yield, C 3 H 6 productivity of 9.5 × 10 -5 mol/(g cat min) at 550 °C, Table 2) with those reported over Ga 2 O 3 /Al 2 O 3 by Michorczyk et al. [42], at 550 °C. A similar initial activity was observed (18% C 3 H 8 conversion), but C 3 H 6 selectivity was much higher (90%), corresponding to C 3 H 6 productivity of 7.2 × 10 -5 mol/ (g cat min). Xu et al. [43] tested several supports (TiO 2 , SiO 2 , ZrO 2 and MgO) for dispersing the active Ga 2 O 3 phase. At 600 °C, the most active was Ga 2 O 3 /Al 2 O 3 , which enabled 26% conversion at 94% C 3 H 6 selectivity, resulting in C 3 H 6 productivity of 11.2 × 10 -5 mol/(g cat min).
Chen et al. [44] investigated how different supports (Al 2 O 3 , SiO 2 and ZrO 2 ) influence the propane ODH activity of 10 wt% In 2 O 3 phase. At 600 °C, the highest propene selectivity of 85% was obtained In 2 O 3 /Al 2 O 3 catalyst at a propane conversion of 20%, resulting in a propene productivity of 5.6 × 10 -5 mol/(g cat min).
Based on the above, the propene selectivity of best catalyst (30CeVO 4 /AC) in this work is about twofold lower compared to those of Ga 2 O 3 or In 2 O 3 catalysts, but propene productivities are comparable.

Conclusions
The re-oxidation of small CeO 2−x crystallites (4 nm) with CO 2 occurs at higher temperatures compared to bulk CeO 2 (15 nm), revealing it is more difficult to dissociate CO 2 as the CeO 2 size decreases. Catalyst reduction by propane and re-oxidation by CO 2 are fast under reaction conditions and the propane dehydrogenation reaction proceeds via lattice oxygen participation. The Main reaction pathway over cerium containing catalysts is the total oxidation of propane.
Under propane-CO 2 ODH reaction conditions, V 2 O 5 is irreversibly reduced to V 2 O 3 , which leads to a progressive shift from propane ODH to the propane cracking reaction. The participation of CO 2 in the ODH reaction is negligible and the reaction pathway is governed by the availability of lattice oxygen in V 2 O 5 and V 2 O 4 . V 2 O 3 preferentially catalyses propane cracking.
In the VO x /AC catalysts, a VC phase is formed exclusively at vanadium loadings up to 5 wt%, whereas VC and V 2 O 3 coexist at 12 wt% vanadium loading. By combining the C 3 H 8 -TPR, CO 2 -TPO, and catalytic results we can postulate that CO 2 activation over the VC crystals is assisted by the adsorbed propane. As a result, the propane-CO 2 ODH reaction, over vanadium carbide, proceeds through the Langmuir-Hinshelwood mechanism. The deactivation of VO x /AC catalysts is accompanied with a stable propene selectivity, which indicates the total number of active sites and not their nature is changed with TOS. Oxidation of the VC phase during the propane-CO 2 ODH reaction does not occur due to the absence of any V-O bonding in spent catalysts. As a result, deactivation is probably related to blocking of the active sites by carbon.
The active site for the propane ODH reaction in the CeVO 4 /AC catalysts is a thin (1-2 nm) amorphous CeVO 4 phase which covers the CeVO 4 crystals. Propane and CO 2 -TPR experiments confirmed that during the propane ODH reaction lattice oxygen abstraction and re-oxidation are feasible, indicating active participation of CO 2 and a Mars van Krevelen reaction mechanism. During reaction, crystallization of the amorphous CeVO 4 layer and surface covering with carbon, which are the causes of catalyst deactivation.