Oxidative dehydrogenation of propane using lattice oxygen of vanadium oxides on silica

Abstract

Oxidative dehydrogenation of propane using lattice oxygen of vanadium oxide was carried out with a fix-bed flow reactor at 450 °C under atmospheric pressure. Vanadium oxide loaded on SiO₂ from V(t-BuO)₃O afforded a high propylene selectivity of 88.3% with a propane conversion of 26.5% based on the lattice oxygen of vanadium oxide. The catalyst prepared with V(t-BuO)₃O exhibited higher propane conversion and propylene selectivity than that prepared with NH₄VO₃. Raman, XPS, and XRD analyses revealed that isolated VO₄³⁻ species seem to be active sites in the ODHP. For the dehydrogenation of propane, the lattice oxygen of the isolated VO₄³⁻ species was used to give propylene and H₂O. The isolated VO₄³⁻ species was reduced to the isolated V³⁺ species and was regenerated by oxidation with oxygen. The activity was maintained for at least 10 repeated cycles, suggesting that VO_x/SiO₂ would be a promising catalyst for ODHP.

---

1. Introduction

Polypropylene is one of the most important general purpose plastics, and currently the demand for it is markedly increasing. Propylene is co-produced with ethylene by the steam cracking of naptha, and with gasoline by the fluidized catalytic cracking of gas oil. However, these processes are carried out at high temperatures and consume a large amount of energy with low selectivities to the desired alkenes.

The development of more selective, simple, and energy-saving processes to produce alkenes starting from less valuable alkanes is of current interest in petrochemical industries. One such process is an oxidative dehydrogenation (ODH) of light alkanes (ODHE, reaction (1); ODHP, reaction (2)). ODH, which is an exothermic reaction, has potential advantages because it consumes smaller amounts of energy and requires little catalyst regeneration.

C₂H₆ + 1/2O₂ → C₂H₄ + H₂O, ΔH = −149 kJ mol⁻¹ (1)

C₃H₈ + 1/2O₂ → C₃H₆ + H₂O, ΔH = −118 kJ mol⁻¹ (2)

From the perspective of space-time yield (STY) in tables summarized by Cavani et al.,¹ ODHP to obtain propylene is more difficult than ODHE to obtain ethylene. In general, in the case of ODHE, the selectivity to ethylene increases with temperature due to the ethyl radical from ethane forming ethylene in the gas phase at higher temperatures (600–700 °C); in addition, the higher stability of ethylene leads to the observed high selectivity.² On the other hand, in the ODHP, the apparent activation energy for propylene combustion is lower than that for propylene formation. This is due to the fact that the bond dissociation energy of allylic hydrogen in propylene is smaller than that for the C–H bonds of propane.¹ For example, in the ODHE, the MoVTeNbO catalyst has been reported to exhibit over 80% ethylene selectivity with ethane conversion higher than 80% (340–400 °C).³ In contrast, in the ODHP, Buyevskaya et al. have reported that VO_x/MCM-48 exhibits a propylene selectivity of 53.0% with 32.8% propane conversion at 500 °C.⁴ This catalyst afforded the highest STY, as indicated in the table summarized by Cavani et al.¹

In general, vanadium-based catalysts are reported to be active and selective for ODHP. As vanadium-based catalysts, V₂O₅/MgO,^5–10 magnesium vanadates (MgV₂O₆, Mg₂V₂O₇, Mg₃V₂O₈),^11–16 V₂O₅/Al₂O₃,^17–19 V₂O₅/SiO₂,^20,21 V₂O₅/ZrO_2,^22,23 and V₂O₅/TiO₂^24,25 have been reported to exhibit a high activity for ODHP. In particular, the V₂O₅/MgO catalyst has been reported to exhibit a propylene selectivity higher than 80% with propane conversion above 10% at 450 °C.¹⁰

The redox properties of these vanadium-based catalysts are of importance for reactions which proceed via the Mars-van-Krevelen mechanism. Hydrocarbon molecules react with oxygen of the vanadium oxides to give alkene and H₂O, and the reduced vanadium oxides are re-oxidized by gas phase molecular oxygen to regenerate the active vanadium oxides. The extent of reduction of the surface V⁵⁺ species also depends on the specific oxide support.^26,27

The selectivity to propylene is not sufficient for industrial application due to parallel and consecutive combustion of propane and propylene to CO₂ and H₂O (reactions (3) and (4)), even though most cases were conducted at a low propylene partial pressure by diluting it with an inert gas. In addition to the lower selectivity to propylene described above, oxygen becomes the limiting reactant, since the higher propane/oxygen feed ratio is usually employed, resulting in oxygen-starved conditions with large oxygen consumption by combustion reactions (3) and (4). Consequently, low propane conversion is obtained.

C₃H₈ + 5O₂ → 3CO₂ + 4H₂O, ΔH = −2220 kJ mol⁻¹ (3)

C₃H₆ + 9/2O₂ → 3CO₂ + 3H₂O, ΔH = −2018 kJ mol⁻¹ (4)

To avoid combustion of propylene, mild oxidants such as N₂O and CO₂ are used in ODHP. Kondratenko et al. have reported that VO_x/γ-Al₂O₃ and VO_x/MCM-41 give high propylene selectivity in the ODHP with N₂O as an oxidant.^28,29 However, industrialization of this process would not be possible because of the high N₂O cost. From an industrial perspective, catalytic oxidative dehydrogenation of propane using oxygen as an oxidant should be exploited without diluting feed gas with an inert gas.

We have reported that the use of CO₂ as an oxidant markedly promotes the dehydrogenation of propane over the oxidized diamond-supported Cr₂O₃ and V₂O₅ catalysts.³⁰ With mild oxidants such as N₂O and CO_2, however, it is difficult to re-oxidize completely reduced VO_x species as compared with oxygen. As a result, propylene yields decrease with increases in the reaction time on stream.

Recently, the V₂O₅–SiO₂ catalyst has been reported to exhibit a propane conversion of 10% with a propylene selectivity of 80% at 550 °C under anaerobic conditions.²⁰ However, after the reaction and regeneration, the catalytic activity is slightly lower than that observed for the first reaction.

Lattice oxygen of Fe₂O₃ and/or NiO has effectively been used to give synthesis gas from CH₄ by using ternary mixed metal oxide catalysts such as Fe₂O₃–Rh₂O₃–Y₂O₃ or NiO–Cr₂O₃–MgO.^31,32 This development has led us to exploit catalysts that use lattice oxygen in the ODHP.

We have reported that CO₂ acts as an oxidant for the selective oxidation of methane and ethane to formaldehyde and acetaldehyde via the lattice oxygen of vanadium oxide loaded on oxidized diamond.^33,34

The objective of the present study was to avoid deep oxidation of propane and to produce propylene by restricting an oxidant to the lattice oxygen of metal oxides having labile oxygen (reaction (5)). Regeneration of the reduced metal oxides with air (reaction (6)) was also studied.

C₃H₈ + MO_x → C₃H₆ + H₂O + MO_x−1 (5)

MO_x−1 + 1/2O₂ (air) → MO_x (6)

2. Experimental

2.1 Materials

The catalyst supports used in this study were SiO₂ (Fuji Silicia Chemical Ltd, Q-3 (629 m² g⁻¹), Q-6 (517 m² g⁻¹), Q-10 (283 m² g⁻¹), Q-30 (118 m² g⁻¹), and Q-50 (63 m² g⁻¹)), γ-Al₂O₃ (Merck, 120 m² g⁻¹), TiO₂ (P25; Japan Aerosil Co., 50 m² g⁻¹), CeO₂ (JRC-CEO-2; Catalysis Society of Japan, 123 m² g⁻¹), and MgO (100A; Ube Industries Ltd, 144 m² g⁻¹). NH₄VO₃, V₂O₅, Co(OAc)₂·4H₂O, Fe(NO₃)₃·9H₂O, Cr(NO₃)₃·9H₂O, Ni(OAc)₂·4H₂O, oxalic acid, tert-butanol, benzene, toluene, and molecular sieves of 5 Å 1/8 were purchased from Wako Pure Chemical Industries, Ltd.

2.2 Catalyst preparation

2.2.1 Preparation by the impregnation method. Supported catalysts were prepared by impregnating the support with an aqueous solution of NH₄VO₃, Co(OAc)₂·4H₂O, Fe(NO₃)₃·9H₂O, Cr(NO₃)₃·9H₂O, or Ni(OAc)₂·4H₂O. Typically, NH₄VO₃ (0.12 g, 1.0 mmol) was dissolved into 10 mL of saturated oxalic acid in water. After impregnating a metal salt onto the support for 24 h, water was evaporated under vacuum. Supported catalyst precursors were calcined at 600 °C for 5 h in air (hereafter denoted as V(NH)).

2.2.2 Preparation of vanadium alkoxide and vanadium alkoxide loaded catalysts. Vanadium alkoxide (V(t-BuO)₃O) was prepared according to a previously reported method.⁷ A 4.0 g portion of V₂O₅ was added to dry tert-butanol (35 mL) in dry benzene (120 mL) in a round-bottom flask fitted with a water separator and a reflux condenser. The mixture was refluxed under stirring for 24 h under an Ar atmosphere. In a water separator, molecular sieves of 5 Å were charged to absorb water produced during reaction (7).

V₂O₅ + 6(CH₃)₃COH ⇄ 2V(OC(CH₃)₃)₃O + 3H₂O (7)

In the progress of the reaction, solid V₂O₅ was dissolved into benzene by changing its color from orange to red to give alkoxide. Unreacted V₂O₅ was removed from the solution by centrifugation. Benzene and unreacted tert-butanol were removed by distillation under an Ar atmosphere, and then under vacuum. After removing excess solvent and tert-butanol, a white crystal was obtained. The product was further heated under vacuum at 313 K overnight to remove trace amounts of solvent.

An SiO₂-supported vanadium oxide catalyst was prepared as follows: the entire procedure was carried out in a glovebox filled with N₂. The SiO₂ support was dried at 200 °C to remove the adsorbed water before impregnation. Onto dried 1.00 g of SiO₂, 0.29 g of V(t-BuO)₃O (1.0 mmol) dissolved in 7 mL toluene was added. The sample was kept still for 24 h and subsequently dried at 120 °C for 1 h under N₂ flow. The supported catalyst precursor was calcined at 600 °C for 5 h in air (hereafter denoted as V(OR)).

2.3 Catalyst test

The ODHP was carried out using a fixed-bed flow-type quartz reactor (10 mm i.d.) at 400–600 °C at atmospheric pressure. After placing 200 mg of the catalyst in the reactor, the ODHP was carried out under 5 mL min⁻¹ of propane and 20 mL min⁻¹ of Ar. Prior to the reaction, the catalyst was treated with air at the reaction temperature for 30 min. Products (CH₄, C₂H₄, C₂H₆, C₃H₆, C₃H₈, H₂, CO, CO₂, and O₂) were analyzed with an on-line gas chromatograph equipped with a TCD detector (PC-Chrom, M2000 Chromato Analyzer) using a molecular sieve of 5 Å and Poraplot Q columns. After the reaction, regeneration of the reduced catalyst was carried out with 5 mL min⁻¹ of O₂ and 20 mL min⁻¹ of Ar at a temperature range of 400–600 °C. During the regeneration stage, formed CO and CO₂ were analyzed by an FID gas chromatograph (Shimadzu GC14B) equipped with a methanizer using an activated carbon column.

2.4 Catalyst characterization

The surface area of the catalyst was measured by the BET method at 77 K using nitrogen as the adsorbate, with a Quantachrome model AUTOSORB-1.

The powder X-ray diffraction (XRD) pattern was obtained with a Shimadzu model XRD-6000 diffractometer with monochromatized Cu Kα radiation. X-Ray photoelectron spectra (XPS) were obtained with a Jeol model JPS-9000MX using Mg Kα radiation as the energy source.

Raman spectra were obtained with a Jasco model NSR-3100 laser Raman spectrometer using 532 nm diode laser excitation with a CCD detector. ESR data were obtained using a Jeol model JES-RE1X operated at the X-band.

3. Results and discussion

3.1 Propane-temperature programmed reaction (TPR)

Selection of active metal oxides was performed using SiO₂ as a support. The following metal oxides were selected: NiO, Co₃O₄, Fe₂O₃, and V₂O₅ from our previous studies on partial oxidation of methane (POM)^31,32 and selective oxidation of methane and ethane to oxygenates with CO₂.^33,34Fig. 1 shows the TPR profiles of NiO, Co₃O₄, Fe₂O₃, and VO_x loaded on SiO₂ using propane as a reactant. On all the metal oxides, sharp rises in H₂ formation were observed starting at different temperatures, indicating that non-oxidative dehydrogenation or cracking of propane (reactions (8) and (9)) proceeded, respectively. The behavior of H₂O formation was much more complicated, and the temperature at which H₂O or H₂ was detected was defined as the formation temperature.

C₃H₈ → C₃H₆ + H₂, ΔH = 124 kJ mol⁻¹ (8)

C₃H₈ → 3C + 4H₂, ΔH = 104 kJ mol⁻¹ (9)

Fig. 1 Temperature-programmed reaction of propane over various metal oxides. Ramping rate: 5 °C min⁻¹, sample: 200 mg, metal oxide: 1.0 mmol g⁻¹ support, flow rate: C₃H₈/Ar = 5/20 (mL mL⁻¹ min⁻¹).

Over NiO/SiO₂, the H₂ formation temperature was 420 °C, and this temperature was lower than that for H₂O at 440 °C. This result indicates that non-oxidative dehydrogenation or cracking of propane proceeded before oxygen transfer from the lattice of NiO to propane. Similarly, over Co₃O₄/SiO₂ and Fe₂O₃/SiO₂ catalysts, H₂ formation was observed below the H₂O formation temperature. On the other hand, over VO_xV(NH)/SiO₂, H₂O formation started at approximately 360 °C, and this temperature was lower than that of H₂ formation at 470 °C. This result indicates that ODHP on VO_x/SiO₂ proceeded at a lower temperature than non-oxidative dehydrogenation or the cracking of propane.

Table 1 summarizes the results of the ODHP over NiO, Co₃O₄, Fe₂O₃, and VO_x/SiO₂. Since the H₂O formation temperature varied according to the oxide species employed, as seen in Fig. 1, the reaction temperature was set at 425 and 600 °C according to the reactivities of the oxides. NiO/SiO₂ (run 1) afforded a high H₂ yield together with a large amount of carbon (selectivity of 78.4%), and a small propane conversion of 2.0%, indicating that the cracking of propane occurred (reaction (9)). Co₃O₄/SiO₂ (run 2) exhibited a high propylene selectivity of 75.9% and a H₂ yield of 0.29 mmol with a higher propane conversion of 9.3%. The results are interpreted based on the progress of non-oxidative dehydrogenation of propane (reaction (8)). Similarly, over Fe₂O₃/SiO₂ (run 3), non-oxidative dehydrogenation again proceeded. VO_x/SiO₂ (run 4) afforded the highest propylene selectivity of 81.5% with a low propane conversion of 2.2% and with the lowest H₂ yield, indicating that ODHP proceeded. The amount of propane fed to the catalyst during the reaction (8 min) and that of lattice oxygen in the VO_x/SiO₂ (200 mg) were 1.63 and 0.46 mmol, respectively. The stoichiometry of the reaction is shown in reaction (10), and evidence for VO₄³⁻ will be discussed later. When V⁵⁺ was reduced to V³⁺, the maximum conversion of propane was estimated to be 11.3%. A propane conversion of 2.2% over VO_x/SiO₂ corresponds to ca. 20% conversion of the lattice oxygen. The propane conversion based on propane feed and the lattice oxygen (V⁵⁺ → V³⁺) is, hereafter, denoted as propane conversion (feed) and propane conversion (VO), respectively. Similarly, propylene yield based on propane feed and the lattice oxygen is denoted as propylene yield (feed) and propylene yield (VO), respectively.

C₃H₈ + VO₄³⁻ → C₃H₆ + H₂O + VO₃³⁻ (10)

Table 1 Effect of metal oxides and vanadium precursors on the oxidative dehydrogenation of propane^a

Run	Catalyst	Precursor	Reaction temp./°C	Conversion/%		Selectivity/%					Yield/μmol		Yield/%
				C3H8^b	C3H8^c	C3H6	CO	CO2	C2 + CH4	Carbon	C3H6	H2	C3H6^d	C3H6^e
a Catalyst: 200 mg, loading as metal: 1.0 mmol g^−1 support, pretreatment: O2/Ar = 5/20 (mL mL^−1 min^−1), 30 min at the reaction temperature. Flow rate: C3H8/Ar = 5/20 (mL mL^−1 min^−1), reaction time: 8 min. b C3H8 conversion based on feed. c C3H8 conversion based on lattice oxygen. d C3H6 yield based on feed. e C3H6 yield based on lattice oxygen. f Vanadium: 1.0 mmol + 1 g SiO2 sand. g Flow rate: C3H8 = 5 mL min^−1 neat.
1	NiO/SiO2	Ni(OAc)2	425	2.0	17.5	14.8	0.0	1.4	5.5	78.4	2.1	103.0	0.3	2.6
2	Co3O4/SiO2	Co(OAc)2	600	9.3	61.6	75.9	1.7	0.3	5.7	16.5	116.2	286.0	7.1	47.0
3	Fe2O3/SiO2	Fe(NO3)3	600	3.1	18.2	64.6	0.0	5.2	20.0	2.1	33.2	22.3	1.9	11.2
4	VOx/SiO2	NH4VO3	450	2.2	19.5	81.5	10.4	5.7	0.1	2.3	29.5	0.3	1.8	15.9
5	V2O5 + SiO2^f	V2O5	450	1.7	15.0	55.7	28.6	13.7	0.2	1.8	15.6	0.2	1.0	8.8
6	VOx/SiO2	V(t-BuO)3O	450	3.0	26.5	88.3	4.0	3.0	0.3	4.4	44.2	0.3	2.7	23.9
7^g	VOx/SiO2	V(t-BuO)3O	450	4.0	35.4	83.6	2.6	6.0	0.1	7.7	55.1	0.3	3.4	30.1

---

At this point, vanadium oxide was selected as the metal oxide in the ODHP with lattice oxygen and used in the following detailed studies.

3.2 Effect of supports and vanadium precursors on the ODHP

Fig. 2 shows the effect of support materials on the ODHP over VO_x-loaded catalysts prepared with V(NH) at 450 °C. VO_x-loaded SiO₂, Al₂O₃, and TiO₂ afforded the same propane conversion (VO) of ca. 20%, but propylene yield (VO) decreased in the following order: SiO₂ > Al₂O₃ > TiO₂. Over VO_x-loaded Al₂O₃ and TiO₂, lower propylene yields (VO) were obtained due to the cracking of propane and oxidation of propane and propylene to CO and CO₂. In contrast, VO_x-loaded CeO₂ and MgO afforded low propane conversions and propylene selectivities. Basic supports seem to be not appropriate for the oxygen transfer from vanadium oxide on these supports. Thus, VO_x/SiO₂ showed the highest performance in the ODHP using lattice oxygen of VO_x.

Fig. 2 Effect of supports on the ODHP. Catalyst: V loading 1 mmol g⁻¹ support, 200 mg; pretreatment: O₂/Ar = 5/20 (mL mL⁻¹ min⁻¹), 30 min, 450 °C; flow rate: C₃H₈/Ar = 5/20 (mL mL⁻¹ min⁻¹); reaction temperature: 450 °C; reaction time: 8 min.

Fig. 3 shows the effect of the surface area of SiO₂ on the ODHP at 450 °C. Five different SiO₂ supports having specific surface areas from 63 to 629 m² g⁻¹ with average pore diameters from 3 to 50 nm and narrow pore size distribution profiles were compared. These catalysts were prepared with V(NH) and V(OR) as vanadium precursors. SiO₂ having lower surface areas of 63, 118, and 283 m² g⁻¹ exhibited higher catalytic activities and slightly increased propane conversion (VO) with an increase in the surface area of SiO₂. Propylene yield (VO) reached a maximum value with SiO₂, having a surface area of 283 m² g⁻¹, and a further increase in the surface area of SiO₂ markedly decreased propane conversion (VO). These tendencies were pronounced when bulkier V(OR) was loaded onto high surface area SiO₂ as seen in Fig. 3b. In contrast, the selectivity to propylene slightly increased, and the selectivities to CO and CO₂ decreased with increasing surface areas of SiO₂. Table 2 summarizes the surface areas of SiO₂ and VO_x/SiO₂ catalysts. When VO_x was loaded onto SiO₂, the surface area decreased for the catalysts except for the lowest surface area of SiO₂ (SA = 63 m² g⁻¹), indicating that VO_x was supported on the inner surface of the mesopores. It should be noted that SiO₂ having higher surface areas of 517 and 629 m² g⁻¹ had very small mesopores and that surface areas significantly decreased to 378 and 462 m² g⁻¹, respectively, by the loading of VO_x. This seems to indicate that certain amounts of vanadium species were loaded on the mouth of narrow pores, and as a result, access of propane into active sites inside pores would be hindered. Therefore, together with the results shown in Fig. 3b, the amount of effective VO_x on the catalyst surface became smaller than that on the lower surface area catalysts. As a result, the higher surface area catalysts exhibited low propane conversions (VO).

Fig. 3 Effect of SiO₂ surface area on the ODHP over VO_x/SiO₂. (a) Propane conversion, propylene yield and selectivity, catalyst prepared with V(NH), (b) propane conversion, propylene yield and selectivity, catalyst prepared with V(OR). Conditions are the same as indicated in the caption of Fig. 2.

Table 2 Surface area of SiO₂ and VO_x/SiO₂ catalysts prepared with NH₄VO₃

Catalyst	SiO2 APD^a/nm	SiO2 surface area/m^2 g^−1	Catalyst surface area/m^2 g^−1
a SiO2 average pore diameter.
VOx/SiO2	50	63	71
	30	118	86
	10	283	228
	6	517	378
	3	629	462

---

In Table 1, runs 4, 5 and 6 provide a comparison of the effects of vanadium precursors on the ODHP. When V(NH) was used as a precursor (run 4), VO_x/SiO₂ exhibited a propylene selectivity of 81.5% with a propane conversion (VO) of 19.5%. A mixture of bulk V₂O₅ and SiO₂ sand (run 5) exhibited the lowest propane conversion (VO) and propylene selectivity of 15.0 and 55.7%, respectively. As seen in run 6, the use of V(t-BuO)₃O (hereafter V(OR)) as a precursor resulted in the highest propylene selectivity of 88.3% with a propane conversion (VO) of 26.5%. Propane conversion based on lattice oxygen and propylene selectivity decreased in the following order: V(OR) > V(NH) > bulk V₂O₅. In addition, over VO_x/SiO₂ prepared with V(NH) and bulk V₂O₅, deep oxidations of propane and propylene to CO and CO₂ proceeded. Pak et al. have reported that the V₂O₅/MgO catalyst prepared with V(OR) shows higher activity than that prepared with V(NH) in the ODHP with molecular oxygen.⁷ Thus, further studies were carried out with SiO₂ having a surface area of 283 m² g⁻¹ using V(OR) as a vanadium precursor for VO_x.

In Table 1, runs 6 and 7 compare the effect of dilution of propane on the ODHP. When neat propane was fed at the same W/F, conversion (VO) increased from 26.5 to 35.4% due to longer contact time, with a slight decrease in propylene selectivities (88.3 to 83.6%). Due to experimental limitation, such as a large dead volume of experimental setup, including an on-line analysis system, accuracy of analyses decreases with a small feed rate, and then further studies were carried out under Ar dilution conditions.

3.3 Effect of VO_x loading level on the ODHP

Fig. 4 shows the effects of VO_x loading level on the ODHP at 450 °C. With an increase in the VO_x loading from 0.5 to 2.0 mmol as V metal on 1 g of SiO₂, propane conversion (VO) reached a maximum of 26.5%, and a further increase in the VO_x loading markedly decreased the propane conversion (VO), together with a decrease in propylene selectivity. In contrast, CO and CO₂ selectivities increased with increasing VO_x loading. The reason for this increase will be discussed later.

Fig. 4 Effect of VO_x loading on the ODHP over VO_x/SiO₂ prepared with V(t-BuO)₃O. (a) Propane conversion and propylene yield, (b) selectivity. Conditions are the same as indicated in the caption of Fig. 2 except for the V loading.

3.4 Effect of reaction temperature on the ODHP

Fig. 5 shows the effects of reaction temperature on the ODHP. With an increase in the reaction temperature from 400 to 475 °C, propane conversion (VO) increased from 7.1 to 42.5%. In contrast, propylene selectivity decreased to some extent (95.1 to 85.6%). At 475 °C, a large amount of H₂ was produced, indicating that non-oxidative dehydrogenation of propane proceeded without consuming lattice oxygen of vanadium oxide. Therefore, 450 °C was considered to be the optimal temperature.

Fig. 5 Effect of the reaction temperature on the ODHP over VO_x/SiO₂ prepared with V(t-BuO)₃O. (a) Propane conversion and propylene yield, (b) selectivity. Conditions are the same as indicated in the caption of Fig. 2 except for the reaction temperature.

3.5 Effect of the propane supply on the ODHP

Fig. 6 shows the effects of the amount of propane supplied to the catalyst on the propane conversion (feed) and propylene selectivity. With an increase in the propane feed from 0.82 to 3.28 mmol against 0.18 mmol of lattice oxygen of VO_x (V⁵⁺ → V³⁺), namely a propane/lattice oxygen ratio between 4.5 and 17.8, the propane conversion (feed) and propylene yield (feed) reached maximum values at a propane/lattice oxygen ratio of 8.9, and further increases in the propane feed decreased the propane conversion (feed) and propylene yield (feed). On the other hand, the selectivity to propylene slightly increased and the selectivities to CO and CO₂ decreased with increases in the propane/lattice oxygen ratio.

Fig. 6 Effect of the amount of C₃H₈ feed on the ODHP over VO_x/SiO₂ prepared with V(t-BuO)₃O. (a) Propane conversion and propylene yield, (b) selectivity. Conditions are the same as indicated in the caption for Fig. 2 except for the propane feed.

Fig. 7 shows the effects of the propane/lattice oxygen ratio, by changing the amount of catalyst, on the conversion and selectivity. With an increase in the propane/lattice oxygen ratio, propane conversion (feed) decreased with a small increase in propylene selectivity. However, even at the lowest propane/lattice oxygen ratio of 4.5, a high propylene selectivity of approximately 85% was maintained. If ODHP was carried out at a lower propane/lattice oxygen ratio, higher propane conversion (feed) was expected. Our present reactor, however, did not allow us to test ODHP below a propane/lattice oxygen ratio of 4.5 due to a limit of the reactor volume.

Fig. 7 Effect of propane/lattice oxygen ratio on the ODHP over VO_x/SiO₂ prepared with V(t-BuO)₃O. (a) Propane conversion and propylene yield, (b) selectivity. Conditions are the same as indicated in the caption of Fig. 2 except for the catalyst weight.

3.6 Repetition of ODHP and catalyst regeneration cycle

In the studies above, a freshly prepared catalyst was used in a once-through reaction mode. In these runs, the catalyst itself was a reactant that supplied oxygen to abstract hydrogen from propane. The amount of lattice oxygen in the vanadium oxide decreased with increases in the amount of propane fed, and finally the catalyst lost its activity. Therefore, regeneration of the catalyst after the ODHP was essential. Fig. 8 shows the results of 10 repeated ODHP and regeneration cycles at 450 °C with propane–Ar (ODHP) and O₂–Ar (regeneration). The reaction times for both were fixed at 8 min. VO_x/SiO₂ prepared with V(OR) exhibited a constant propane conversion (VO) up to the 10th ODHP and a regeneration cycle with a slight decrease in the propylene selectivity.

Fig. 8 Repetition of ODHP and catalyst regeneration cycle. (a) Propane conversion and propylene yield, (b) selectivity. Catalyst: V loading 1 mmol g⁻¹ SiO₂, 200 mg; pretreatment: O₂/Ar = 5/20 (mL mL⁻¹ min⁻¹), 30 min, 450 °C; flow rate: C₃H₈/Ar = 5/20 (mL mL⁻¹ min⁻¹); reaction temperature: 450 °C; reaction time: 8 min; regeneration: O₂/Ar = 5/20 (mL mL⁻¹ min⁻¹), 8 min, 450 °C.

In order to understand the conversion of lattice oxygen in the ODHP, Fig. 9 shows the pulsed oxidation of the used VO_x/SiO₂ catalyst with O₂ at 450 °C. When a series of O₂ pulses (98.2 μmol) were supplied to VO_x/SiO₂ after the ODHP, no response was observed against the first O₂ pulse, with very small responses of CO (m/z = 28) and CO₂ (m/z = 44), presumably due to the combustion of deposited carbon. With the introduction of a second O₂ pulse, a large O₂ response was observed, and nearly the same responses were observed for the successive O₂ pulses. A conversion of lattice oxygen in vanadium oxide was then calculated from the amount of O₂ consumed by correcting for the amounts of CO and CO₂ formed. The amount of O₂ used to re-oxidize vanadium oxide was evaluated as 102 μmol. This value corresponds to 55.2% of lattice oxygen of VO_x (V⁵⁺ to V³⁺) on the fresh catalyst, and 6.3% of propane feed.

Fig. 9 Response to pulsed oxidation of the VO_x/SiO₂ prepared with V(OR) after ODHP. Catalyst: V loading 1 mmol g⁻¹ SiO₂, 200 mg; pretreatment: O₂/Ar = 5/20 (mL mL⁻¹ min⁻¹), 30 min, 450 °C; flow rate: C₃H₈/Ar = 5/20 (mL mL⁻¹ min⁻¹); reaction temperature: 450 °C; reaction time: 8 min; regeneration: Ar = 20 mL min⁻¹, O₂ pulse 98.2 mmol, 450 °C.

These results seem to indicate that the regeneration cycle by oxygen could be shortened from 8 min to less than 30 s. Fig. 10 shows the results of 5 repeated ODHP and regeneration cycles at 450 °C with an 8 min propane–Ar reaction period and then 30 s of O₂–Ar regeneration. Even when the amount of oxygen decreased from 3.28 to 0.20 mmol, VO_x(V(OR))/SiO₂ exhibited a constant propane conversion (VO) with a slight decrease in the propylene selectivity. This result indicates that lattice oxygen consumed by dehydrogenation can be regenerated by only a small excess of molecular oxygen, and that deposited carbon was eliminated from the catalyst.

Fig. 10 Repetition of ODHP and catalyst regeneration cycle. (a) Propane conversion and propylene yield, (b) selectivity. Catalyst: V loading 1 mmol g⁻¹ SiO₂, 200 mg; pretreatment: O₂/Ar = 5/20 (mL mL⁻¹ min⁻¹), 30 min, 450 °C; flow rate: C₃H₈/Ar = 5/20 (mL mL⁻¹ min⁻¹); reaction temperature: 450 °C; reaction time: 8 min; regeneration: O₂/Ar = 5/20 (mL mL⁻¹ min⁻¹), 30 s, 450 °C.

For the application of this system to industrial processes, use of a moving bed or a circulating fluidized bed reactor will be appropriate.

3.7 Catalyst characterization

3.7.1 X-Ray diffraction (XRD) analyses. Fig. 11 shows XRD patterns of bulk V₂O₅ and VO_x/SiO₂ prepared with V(NH) and V(OR). The amount of vanadium species was 1 mmol VO_x/1 g of SiO₂ in all cases. V₂O₅ and VO_x/SiO₂ prepared with V(NH) showed sharp and weak characteristic diffraction peaks corresponding to V₂O₅ crystallites, respectively (Fig. 11a and b). In contrast, no diffraction peaks were observed over VO_x/SiO₂ prepared with V(OR) (Fig. 11c). A high dispersion of vanadium species on SiO₂ was suggested for the catalyst prepared with V(OR). The propane conversion and propylene selectivity among the three catalysts decreased in the following order: V(OR) > V(NH) > V₂O₅, indicating that the activity and selectivity are likely related to the dispersion of vanadium species.

Fig. 11 XRD patterns of VO_x-based catalysts. (a) V₂O₅ + SiO₂ sand, (b) VO_x/SiO₂ prepared with NH₄VO₃, (c) VO_x/SiO₂ prepared with V(t-BuO)₃O.

Fig. 12 shows XRD patterns of VO_x/SiO₂ having different vanadium loading levels from 0.5 to 2.0 mmol g⁻¹ SiO₂ prepared with V(OR). V(0.5)/SiO₂ and V(1.0)/SiO₂ showed no diffraction peaks (Fig. 12a and b). In contrast, V(1.5)/SiO₂ and V(2.0)/SiO₂ showed diffraction peaks assignable to V₂O₅ (Fig. 12c and d). Table 3 summarizes the surface areas of catalysts and the surface densities of vanadium species on these catalysts. Assuming homogeneous dispersion of V species on the SiO₂ support, at a vanadium loading of 1.5 mmol g⁻¹ SiO₂, the surface density of vanadium is calculated to be 3.2 V atom nm⁻² of SiO₂ surface. Gao et al. have estimated the maximum monolayer dispersion to be 2.6 V atom nm⁻².³⁵ Therefore, higher surface densities of 3.2 and 4.3 V atom nm⁻² on SiO₂ led to the appearance of polymeric V₂O₅ crystallites. On the other hand, loading from V(NH) exhibited diffraction peaks of V₂O₅ at a lower loading level of 1 mmol g⁻¹ SiO₂. Therefore, higher propane conversion and propylene selectivity could be achieved with isolated VO_x species on SiO₂.

Fig. 12 XRD patterns of VO_x(0.5–2.0 mmol)/SiO₂ catalysts prepared with V(t-BuO)₃O. (a) V(0.5 mmol), (b) V(1.0 mmol), (c) V(1.5 mmol), (d) V(2.0 mmol).

Table 3 Surface area and surface density of VO_x/SiO₂ catalysts prepared with V(t-BuO)₃O

Vanadium loading on SiO2^a/mmol g^−1 SiO2	V2O5 wt%^b	Catalyst surface area/m^2 g^−1	Surface density of V atom/nm^2
a SiO2 surface area: 283 m^2 g^−1, average pore diameter: 10 nm. b (V2O5)wt/(SiO2 + V2O5)wt × 100.
0.5	4.4	278	1.1
1.0	8.3	242	2.1
1.5	12.0	134	3.2
2.0	15.4	124	4.3

---

3.7.2 Raman spectra. In order to further clarify the effects of vanadium precursors on the ODHP, Raman spectra of VO_x/SiO₂ catalysts prepared with V(NH) and V(OR) are compared in Fig. 13 and 14. Over VO_x/SiO₂ catalysts prepared with V(NH), sharp peaks at 1040, 996, 704, 526, 487, 405, 289, 205, and 147 cm⁻¹ were exhibited before the ODHP (Fig. 13a). The Raman band at 1040 is assigned to the V=O stretching vibration of isolated monovanadate VO₄³⁻,³⁵ and the bands at 996, 704, 526, 487, 405, 289, 205, and 147 cm⁻¹ correspond to V₂O₅ crystallites.³⁶ After the ODHP, the peaks at 704, 526, 487, 405, 289, 205, and 147 cm⁻¹ disappeared, and the intensities of the peaks at 1042 and 996 cm⁻¹ decreased (Fig. 13b).

Fig. 13 Raman spectra of VO_x(1.0 mmol)/SiO₂ catalysts prepared with NH₄VO₃. (a) Before the ODHP, (b) after the ODHP.

Fig. 14 Raman spectra of VO_x(1.0 mmol)/SiO₂ catalysts prepared with V(t-BuO)₃O. (a) Before the ODHP, (b) after the ODHP, (c) after the regeneration, (d) after the 10th regeneration.

On the other hand, VO_x/SiO₂ catalysts prepared with V(OR) (Fig. 14a) exhibited a sharp peak at 1040 cm⁻¹ and weak broad peaks at 920, 796, 488, and 358 cm⁻¹. The Raman band at 920 cm⁻¹ is assigned to the perturbed silica vibration indicative of formation of the V–O–Si bond,³⁵ and the other bands at 796, 488, and 358 cm⁻¹ are ascribed to those of the SiO₂ support.³⁶ After the ODHP, the band at 1040 cm⁻¹ decreased in its intensity (Fig. 14b) and no new peaks were observed. The V=O bond in the isolated VO₄³⁻ seems to have abstracted hydrogen from propane. After regeneration of the once-reacted catalyst and the 10th regeneration cycle (Fig. 14c and d), the intensity of the peak at 1040 cm⁻¹ recovered its original state. The catalyst therefore maintained a constant catalytic performance up to ten ODHP and regeneration cycles.

These results indicate that the reactive lattice oxygen in the ODHP corresponds to the V=O double bond of both isolated VO₄³⁻ and V₂O₅ crystallites. Based on the ODHP results, the reactivity of the V=O double bond in the isolated VO₄³⁻ is higher than that in V₂O₅ crystallites.

Fig. 15 compares Raman spectra of VO_x/SiO₂ catalysts prepared with V(OR) having different vanadium loadings (0.5–2.0 mmol). The band at 1040 cm⁻¹ was observed for all the loadings, and the intensity of this band increased from 0.5 to 1.0 mmol. Above 1.5 mmol, its intensity decreased, and a new band appeared at 997 cm⁻¹. Changes in the intensities of the band at 1040 cm⁻¹ correspond to the growth of crystallite size observed in the XRD shown in Fig. 12d.

Fig. 15 Raman spectra of VO_x(0.5–2.0 mmol)/SiO₂ catalysts prepared with V(t-BuO)₃O. (a) V(0.5 mmol), (b) V(1.0 mmol), (c) V(1.5 mmol), (d) V(2.0 mmol).

3.7.3 X-Ray photoelectron spectra (XPS). Fig. 16 shows XPS patterns of VO_x/SiO₂ catalysts prepared with V(OR). Before the ODHP (Fig. 16b), the fresh catalyst showed overlapping peaks composed of isolated VO₄³⁻, VO_n(OH)_m, or crystalline V₂O₅ and V₂O₃ showing binding energies (BE) at 518.8, 517.4, and 515.9 eV, respectively.³⁷ H₂ reduction of VO_x(0.5)/SiO₂ catalysts was carried out at 550 °C for 30 min. As seen in Fig. 16a, H₂-treated catalysts showed BE at 517.4 eV, indicating that VO₃³⁻ appears here. Deconvolution of three peaks to isolated VO₄³⁻, VO₃³⁻, VO_n(OH)_m, or crystalline V₂O₅ and V₂O₃ for the fresh catalyst led to a relative abundance of 78.1, 19.5 and 2.4%, respectively (Fig. 16b).

Fig. 16 XPS patterns of VO_x/SiO₂ catalysts prepared with V(t-BuO)₃O. (a) After H₂ reduction of VO_x(0.5)/SiO₂, (b) before the ODHP of VO_x(1.0)/SiO₂, (c) after the ODHP of (b), (d) after the 10th regeneration of (b).

To confirm the presence of V⁴⁺ in VO_x/SiO₂ catalysts prepared with V(OR), the same samples shown in Fig. 16 were characterized with ESR spectroscopy (spectra are not shown). The fresh and H₂-treated catalysts showed no ESR signal from V⁴⁺ paramagnetic species (note that V⁵⁺ and V³⁺ are diamagnetic). After the ODHP, the catalyst showed a weak signal with a hyperfine structure (hfs) typical of isolated V⁴⁺ ions. A bulk or V⁴⁺ crystallite exhibits a broad singlet without hfs which is typical to magnetically interacting V⁴⁺ ions.³⁸ From these results, we can safely say the existence of only a small amount of V⁴⁺, which cannot be detected by XPS.

After the ODHP, the fraction of isolated VO₄³⁻ decreased to 39.8%, whereas the fraction of VO₃³⁻, VO_n(OH)_m, or crystalline V₂O₅ and V₂O₃ increased to 35.8 and 24.4% (Fig. 16c), respectively. Even after the 10th ODHP and regeneration cycle, the relative abundance of the three vanadium species was 70.7, 18.2, and 11.1%, which was not a major change (Fig. 16d). This result indicates that the catalyst maintained more than 70% of isolated VO₄³⁻ species, suggesting that isolated VO₄³⁻ species on SiO₂ did not aggregate during repeated redox cycles. Based on the XPS analyses, changes in the relative abundance of VO₄³⁻, VO₃³⁻, V₂O₅, and V₂O₃ were significant before and after ODHP, indicating that ODHP proceeded by the redox cycle between V⁵⁺ (VO₄³⁻, V₂O₅) and V³⁺ (VO₃³⁻, V₂O₃), respectively.

From these findings, we propose the following reaction mechanisms shown in Scheme 1. The first step is H-abstraction from a secondary carbon atom of propane by oxygen of a V=O double bond of isolated VO₄³⁻. The second H-abstraction may occur from a primary carbon atom by the bridging V–O–Si oxygen bond, and propylene is then released from the vanadium active site, together with the production of H₂O. Blasco and Lopez Nieto have reported similar catalytic behavior of supported vanadium oxide catalysts in the dehydrogenation of butenes and n-butane.³⁹ In the regeneration cycle, reduced VO₃³⁻ species is oxidized with molecular oxygen to give the active VO₄³⁻.

Scheme 1 Proposed reaction scheme for the ODHP over VO_x/SiO₂ catalysts prepared with V(OR).

4. Conclusions

ODHP was carried out using the lattice oxygen of metal oxide in the absence of gas-phase oxygen. Vanadium oxide supported on SiO₂ prepared with V(t-BuO)₃O as the vanadium source afforded a high C₃H₆ selectivity of 88.3% with a moderate C₃H₈ conversion of 26.5% based on the lattice oxygen of vanadium oxide. The results of Raman, XPS and XRD analyses suggested that the isolated VO₄³⁻ on SiO₂ was the active site in the ODHP.

The active VO₄³⁻ was reduced to VO₃³⁻, and this VO₃³⁻ on SiO₂ was regenerated with O₂ at 450 °C without deactivation in repeated uses.

Acknowledgements

A part of this work was financially supported by the “High-Tech Research Center (HRC) Project” (2007–2011) by MEXT. KF expresses his thanks for the research assistantship from the HRC project.

References

1. F. Cavani, N. Ballrini and A. Cericola, Catal. Today, 2007, 127, 113–131.
2. F. Cavani and F. Trifiro, Catal. Today, 1995, 24, 307–313.
3. P. Botella, E. Garcia-Gonzalez, A. Dejoz, J. M. Lopez Nieto, M. I. Vazquez and J. Gonzalez-Calbet, J. Catal., 2004, 225, 428–438.
4. O. V. Buyevskaya, A. Bruckner, E. V. Kondratenko, D. Wolf and M. Baerns, Catal. Today, 2001, 67, 369–378.
5. M. A. Chaar, D. Patel and H. H. Kung, J. Catal., 1988, 109, 463–467.
6. A. Corma, J. M. López Nieto and N. Paredes, J. Catal., 1993, 144, 425–438.
7. C. Pak, A. T. Bell and T. D. Tilley, J. Catal., 2002, 206, 49–59.
8. A. Klishinska, K. Samson, I. Gressel and B. Grzybowska, Appl. Catal., A, 2006, 309, 10–16.
9. A. Klishinska, S. Loridant, B. Grzybowska, J. Stoch and I. Gressel, Appl. Catal., A, 2006, 309, 17–27.
10. Z. Chao and E. Ruckenstein, Catal. Lett., 2004, 94, 217–221.
11. D. Siew Hew Sam, V. Soenen and J. C. Volta, J. Catal., 1990, 123, 417–435.
12. V. Soenen, J. M. Herrmann and J. C. Volta, J. Catal., 1996, 159, 410–417.
13. A. Burrows, C. J. Kiely, P. E. Højlund-Nielsen, G. Vorbeck, J. J. Calvino and C. López-Cartes, Catal. Lett., 1999, 57, 121–128.
14. S. Sugiyama, Y. Iizuka, N. Fukuda and H. Hayashi, Catal. Lett., 2001, 73, 137–140.
15. S. Sugiyama, T. Hashimoto, N. Shigemoto and H. Hayashi, Catal. Lett., 2003, 89, 229–233.
16. S. Sugiyama, T. Hashimoto, Y. Morishita, N. Shigemoto and H. Hayashi, Appl. Catal., A, 2004, 270, 253–260.
17. K. Routray, K. R. S. K. Reddy and G. Deo, Appl. Catal., A, 2004, 265, 103–113.
18. B. Mitra, I. E. Wacha and G. Deo, J. Catal., 2006, 240, 151–159.
19. I. Rossetti, L. Fabbrini, N. Ballarini, C. Oliva, F. Cavani, A. Cericola, B. Bonelli, M. Piumetti, E. Garrone, H. Dyrbeck, E. A. Blekkan and L. Forni, Catal. Today, 2009, 141, 271–281.
20. I. Rossetti, L. Fabbrini, N. Ballarini, C. Oliva, F. Cavani, A. Cericola, B. Bonelli, M. Piumetti, E. Garrone, H. Dyrbeck, E. A. Blekkan and L. Forni, J. Catal., 2008, 256, 45–61.
21. S. A. Karakoulia, K. S. Triantafyllidis, G. Tsilomelekis, S. Boghosian and A. A. Lemonidou, Catal. Today, 2009, 141, 245–253.
22. C. L. Pieck, M. A. Banares and J. L. G. Fierro, J. Catal., 2004, 224, 1–7.
23. R. Sasikala, V. Sudarsan, T. Sakuntala, Jagannath, C. Sudakar, R. Naik and S. R. Bharadwaj, Appl. Catal., A, 2008, 350, 252–258.
24. J. B. Stelzer, J. Caro and M. Fait, Catal. Commun., 2005, 6, 1–5.
25. R. P. Singh, M. A. Banares and G. Deo, J. Catal., 2005, 233, 388–398.
26. X. Gao, M. A. Banares and I. E. Wachs, J. Catal., 1999, 188, 325–331.
27. M. A. Banares, M. V. Martinez-Huerta, X. Gao, J. L. G. Fierro and I. E. Wachs, Catal. Today, 2000, 61, 295–301.
28. E. V. Kondratenko and M. Baerns, Appl. Catal., A, 2001, 222, 133–143.
29. E. V. Kondratenko, M. Cherian, M. Baerns, D. Su, R. Schlogl, X. Wang and I. E. Wachs, J. Catal., 2005, 234, 131–142.
30. K. Nakagawa, C. Kajita, N. Ikenaga, M. Nishitani-Gamo, T. Ando and T. Suzuki, Catal. Today, 1998, 84, 149–157.
31. O. Nakayama, N. Ikenaga, T. Miyake, E. Yagasaki and T. Suzuki, Catal. Today, 2008, 138, 141–146.
32. O. Nakayama, N. Ikenaga, T. Miyake, E. Yagasaki and T. Suzuki, Ind. Eng. Chem. Res., 2010, 49, 526–534.
33. K. Okumura, K. Nakagawa, T. Shimamura, N. Ikenaga, M. Nishitani-Gamo, T. Ando, T. Kobayashi and T. Suzuki, J. Phys. Chem. B, 2003, 107, 13419–13424.
34. T. Shimamura, K. Okumura, K. Nakagawa, T. Ando, N. Ikenaga and T. Suzuki, J. Mol. Catal. A: Chem., 2004, 211, 97–102.
35. X. Gao, S. R. Bare, B. M. Weckhuysen and I. E. Wachs, J. Phys. Chem. B, 1998, 102, 10842–10852.
36. S. Xie, E. Iglesia and A. T. Bell, Langmuir, 2000, 16, 7162–7167.
37. C. Hess, G. Tzolova-Müller and R. Herbert, J. Phys. Chem. C, 2007, 111, 9471–9479.
38. A. V. Kucherov, A. V. Ivanov, T. N. Kucherova, V. D. Nissenbaum and L. M. Kustov, Catal. Today, 2003, 81, 297–305.
39. T. Blasco and J. M. Lopez Nieto, Appl. Catal., A, 1997, 157, 117–142.

---