ARTICLE pubs.acs.org/JPCC
Deep Oxidations in the Oxidative Dehydrogenation Reaction of Propane over V2O5(001): Periodic Density Functional Theory Study Guo-Liang Dai,† Zhen-Hua Li,† Jing Lu,† Wen-Ning Wang,*,†,‡ and Kang-Nian Fan*,† †
Department of Chemistry and Shanghai Key Laboratory of Molecular Catalysis and Innovative Materials and ‡Institute of Biomedical Sciences, Fudan University, Shanghai 200433, People’s Republic of China
bS Supporting Information ABSTRACT: The oxidative dehydrogenation (ODH) of propane to propene over a vanadium-based catalyst suffers from side reactions of further and complete oxidations of propene and other intermediates, which limit the yield of propene. These further oxidation reactions are also referred to as deep oxidation reactions of ODH. In this paper, we present a comprehensive study of the deep oxidation reactions in the ODH of propane over the V2O5(001) surface using the periodic density functional theory method. It is shown that the main source of deep oxidation byproducts originates from the dehydrogenation reaction of the surface intermediate isopropoxide, leading to acetone and the following deep oxidation reactions of acetone and propene. Thorough oxidation of acetone is more difficult than that of propene. Noticeably, formation of acetone and deep oxidation of acetone and propene are only feasible on the terminal oxygen site O(1) of the V2O5(001) surface. The bridging site O(2) has similar reactivity for propene formation but is inert for the side reactions, showing its superiority for selectivity of propane ODH.
1. INTRODUCTION The increasing global demand for light alkenes and their shortage have raised new interest in producing them from light alkanes, which are sources of inexpensive raw materials and are generally easily available.13 Currently, the direct dehydrogenation of alkanes is still used in industry for the production of light alkenes. However, this kind of reaction is reversible and suffers from thermodynamic limitations on conversion, low yield, and coke formation. As a new technique of alkene production, oxidative dehydrogenation (ODH) of light alkanes overcomes most of the obstacles mentioned above. Introduction of an oxidant into the reaction mixture allows the oxidation of alkane into alkene and water. The reaction is exothermic and is able to proceed at much lower temperature. This in turn reduces the side reactions, such as cracking of alkanes and coke formation.4 Vanadium-based catalysts are among the best catalysts for the ODH of alkanes to alkenes.510 Supported V2O5 catalysts on oxides such as titania and alumina have better performance due to their high thermal stabilities and large surface areas. However, ODH has the disadvantage of consecutive oxidation from alkenes to carbon oxides, which will reduce the selectivity of alkene.1113 Therefore, the fundamental problem in ODH of light alkanes is how to inhibit the reaction at the desired step by eliminating the subsequent further oxidation (also referred to as deep oxidation) for achieving a high yield of alkenes. During the past 15 years, although extensive study has been carried out on the development of catalysts with high performance in alkane ODH reactions, only a few studies have focused on the catalyst properties related to mechanistic aspects.1418 r 2011 American Chemical Society
These studies have proven that alkene formation occurs via a MarsKrevelen mechanism, but the detailed mechanisms of COx formation remain elusive. Bielanski and Haber19 suggested that, with the participation of adsorbed and lattice oxygen species, alkenes are first activated at the double bond to form saturated aldehydes, which are further oxidized to COx. On the basis of the analysis of the kinetic isotopic effect in the ODH reaction over V2O5/ZrO2, Chen et al.20 proposed that there exist two different lattice oxygen species participating in selective dehydrogenation and nonselective combustion reactions. By means of steady-state and transient isotopic tests, Kondratenko et al.21 investigated the mechanistic aspects of the formation of propene, CO, and CO2 in the ODH of propane over VOx/γ-Al2O3. They found that nonselective consecutive oxidation of propene is initiated with the breaking of the CC bond by the lattice oxygen to form formaldehyde as a side product which is further oxidized to CO and CO2. The reaction pathways of deep oxidations in the propane ODH are more evident in the studies at relatively low temperatures, in which the reaction intermediates and byproducts besides COx can be detected. For example, the oxidation of propane over a V2O5P2O5-based catalyst gave the main products of acrylic acid, acetic acid, and CO at 270360 °C.22 It was found that the selectivities in the oxidation of propane to acrylic acid and to acetic acid are almost the same as those found in the oxidation of Received: September 7, 2011 Revised: November 30, 2011 Published: December 01, 2011 807
dx.doi.org/10.1021/jp208639t | J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
propene.22 Using a MoVTeNbO mixed metal oxide as the catalyst, Lin et al.23 studied the selective oxidation of propane at 391 °C. On the basis of the experimental findings, they proposed that propene is oxidized via two competitive pathways: (a) propene f acrolein f acrylic acid f COx; (b) propene f acetone f acetic acid. Density functional theory (DFT) calculation is a powerful means to understand the detailed mechanisms of catalytic reactions on surfaces. In the past decade, there have been many detailed theoretical studies on the molecular mechanism of propane ODH,2434 but most of those works were focused on the CH bond activation of propane and propene formation, while the detailed study of deep oxidation reactions is rare. In agreement with most of the experimental observations, these theoretical calculations showed that propane ODH on vanadium oxide is initiated by methylene CH bond activation, and the surface vanadyl VdO (O(1)) site exhibits high activity. In a previous work of our group,26 the CH bond activation of propane and the formation of propene were investigated using periodic DFT calculations on V2O5(001), demonstrating that the vanadyl O(1) site is slightly more reactive in CH bond activation than the bridging O(2) site, while the isopropoxide intermediate on O(2) much more easily forms propene. Besides the role of vanadium oxide, some theoretical studies also explored the effect of supporting oxides on the ODH of propane.28 However, without knowing the mechanistic details of the deep oxidation reaction network in the ODH of propane, it is difficult to understand and evaluate the properties of the catalysts that determine the selectivity of propene, as well as the effects of the supporting materials. In this paper, we present a comprehensive periodic DFT study on deep oxidations and side reactions in the propane ODH over the V2O5(001) surface. The detailed reaction pathway of the complete oxidation reaction network of propane ODH has been obtained, which suggests that the two main byproduct channels lowering the selectivity of propene are the formation and oxidation of acetone and the deep oxidation of propene. The deep oxidations only occur on the vanadyl O(1) site, while the bridging O(2) site mainly shows high selectivity for propene production.
All the results reported here were calculated by considering spin polarization. A constrained minimization scheme was employed to search the transition states (TSs) on the potential energy surface.39,40 A TS is identified when (i) the forces on the atoms vanish and (ii) the energy is a maximum along the reaction coordinate, but a minimum with respect to all of the other degrees of freedom. Then the transition states were verified by calculating the numerical harmonic frequencies. A cluster of atoms including the adsorbed molecules and a few surface atoms was isolated and used to numerically calculate the harmonic frequencies. The values of the imaginary frequency in each TS are listed in Table S1 (Supporting Information). The energy barrier is determined as the energy difference between the saddle point and the initial state. The adsorption energy (Eads) is calculated as Eads ¼ Eadsorbatesubstrate ðEadsorbate þ Esubstrate Þ where a negative Eads indicates an exothermic process. To justify the accuracy of the electronic structure method used here, we calculated the energy barrier of methylene CH bond activation of propane on the O(1) site of V2O5(001) using the above method and obtained a value of 24.6 kcal/mol. In our previous study, this energy barrier was calculated to be 27.3 kcal/mol by using GGA-PW91 and the plane wave basis set in the VASP package.26 Another study based on the cluster model and B3LYP/ 6-31G** method gave a value of 23.9 kcal/mol of this energy barrier.32 These results are generally in good agreement with the experimental result of 27 ( 5 kcal/mol on supported vanadium oxides.17 The size of the basis set was also justified through comparison with the calculations using the triple-ξ plus polarization (TZP) basis set. All the comparisons showed that the electronic structure method used in the present calculations has reasonable accuracy to provide reliable relative energetics.
3. RESULTS AND DISCUSSION 3.1. Acetone Formation. On the basis of the previous study,26
it is well established that the first step of propane ODH reaction is the dissociative adsorption of propane on the surface by breaking a methylene CH bond to form a surface isopropoxide and a surface hydroxyl group. Propene is then formed by abstracting a methyl hydrogen atom by a nearby lattice oxygen atom. However, there is another possible reaction channel, in which the only methylene hydrogen of isopropoxide is abstracted by a nearby lattice oxygen atom, leading to adsorbed acetone. As shown in Figure 1, both the methyl hydrogen atom and methylene hydrogen atom of isopropoxide at the O(1) site can be abstracted by a nearby O(1), leading to the formation of propene and acetone with similar energy barriers of 26.7 and 30.5 kcal/mol, respectively. The CO bond in TS1a is almost broken (bond length of 2.67 Å), and the interaction between the propene and surface is very weak; thereby the propene molecule readily desorbs to the gas phase. The surface acetone species is much more stable with a desorption energy barrier of 12.3 kcal/mol. This is not surprising since the desorption of acetone requires the breaking of the VO(1) bond. The energy barrier of acetone formation is 4 kcal/mol higher than that of propene formation, but acetone formation is thermodynamically more favored. Therefore, the reaction pathway leading to acetone is likely competitive. This is in agreement with the experimental results of propane ODH reaction, in which acetone is found to be the primary oxidation product.41,42 In addition to lattice oxygen, the surface
2. COMPUTATIONAL DETAILS All total energy density functional theory calculations were carried out using the SIESTA package with numerical atomic orbital basis sets and TroullierMartins norm-conserving pseudopotentials.35,36 The exchange-correlation functional utilized is the generalized gradient approximation method, known as GGA PBE (PBE = PerdewBurkeErnzerhof).37 A double-ζ plus polarization (DZP) basis set was employed. The orbital-confining cutoff radii were determined from an energy shift of 0.01 eV. The energy cutoff for the real space grid used to represent the density was set as 150 Ry. To further speed up the calculations, the KohnSham equations were solved by an iterative parallel diagonalization method that utilizes the ScaLAPACK subroutine pdsygvx with a two-dimensional block cyclically distributed matrix.38 The Broyden method was employed for geometry relaxation in which the maximal forces on each relaxed atom are less than 0.05 eV/Å. Without specific mentioning, the V2O5(001) surface was routinely modeled by a two-layer slab with all atoms allowed to relax. A (3 1) unit cell (11.54 10.81 Å) was used to model the surface, and the vacuum region was set as 20 Å. Only the Γ point was used to sample the Brillouin zone in our calculations. 808
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
Figure 1. Energy profiles of propene and acetone formation from isopropoxide at the O(1) site on the V2O5(001) surface. Values of relative energies (kcal/mol) and bond lengths (Å) are shown here as well as in the following figures.
hydroxyl group O(1)H can also abstract hydrogen atoms from isopropoxide to form propene/acetone and a molecule of water. As shown in Figure S1 (Supporting Information), the energy profiles are similar to those in Figure 1. The energy barriers leading to propene and acetone are 32.0 and 35.8 kcal/mol, respectively, showing that O(1)H is a slightly weaker oxidant than O(1). However, the reactions are less endothermic due to the water formation. The reactions of propene and acetone formation were also examined on the bridge site O(2). As shown in Figure 2, the most favorable pathway leading to the formation of propene/acetone is through H abstraction by a nearby O(2). The formation of propene is easier on O(2) than that on O(1) site with a CH bond-breaking energy barrier of 25.2 kcal/mol and desorption energy of 6.2 kcal/mol on the O(2) site (Figure 2). The formation of acetone, however, is much more difficult with an energy barrier as high as 46.1 kcal/mol on the O(2) site. As expected, the desorption energy is also very high (36.1 kcal/mol) since depriving an O(2) from the surface significantly disrupts the lattice structure. In summary, acetone formation competes with propene production at the vanadyl O(1) site, while the acetone species is unlikely formed at the bridging O(2) site, implying the higher selectivity of the O(2) site for propene formation.
3.2. Acetone Oxidation. We then examined the possible further oxidation of the byproduct acetone. The desorbed acetone may rehit the surface and react with the lattice oxygen O(1) by activating a CH bond. As shown in Figure 3, through a reaction mechanism similar to that of propane, the energy barrier of acetone CH bond activation is 35.0 kcal/mol, leading to a biradical intermediate, IM3a, that readily rebounds to another O(1) on the surface, giving rise to IM3b. This intermediate species is quite stable. We have examined all the possible oxidation pathways of IM3b, including various CH bond activations (data not shown), and found that the most energetically favorable reaction channel is breaking the CC bond through a TS3c transition state (Figure 3). In TS3c, a surface O(1) atom attacks the carbonyl carbon to form a CO(1) bond with a bond length of 1.58 Å, while the activated CC bond is elongated to 1.93 Å. By overcoming an energy barrier of 31.5 kcal/mol, the breaking of the CC bond leads to an acetic group, CH3COO*, on the surface and a formaldehyde molecule (O* denotes the lattice oxygen). The energy profile demonstrates that the overall energy barrier of this reaction is 38.9 kcal/mol. The further oxidation of the products, i.e., CH3COO* and formaldehyde, will be discussed herein below. 3.3. Oxidation of Propene. The first step of propene deep oxidation on the surface may be the chemisorption of propene 809
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
Figure 2. Energy profiles of propene and acetone formation from isopropoxide at the O(2) site on the V2O5(001) surface.
over a mixed metal oxide catalyst, Lin et al.23 suggested that the byproduct acetone is mainly derived from propene oxidation. Our calculations showed that acetone can be formed from both isopropoxide and propene. On the other hand, the deep oxidation of propene on the surface can start with activation of the allylic CH bond, which is calculated to be very easy (Figure 4). The lattice oxygen atom can abstract a hydrogen atom from propene to form a biradical intermediate, followed by rebinding the radical to a nearby surface oxygen atom. Our calculations showed that O(1) and O(2) have similar ability to activate the allylic CH bond with energy barriers of 21.3 and 19.9 kcal/mol, respectively. We only present the pathway of CH activation on O(2) with lower energy barrier in Figure 4. In the rebinding step, however, the O(1) site is more favorable than O(2) for allyl radical adsorption. The calculated energy barrier for IM4a rebinding is only 0.7 kcal/mol. Therefore, the most feasible allylic CH bond activation pathway over the V2O5(001) surface is that the O(2) atom abstracts a hydrogen atom from propene to form an allyl radical, followed by rebinding of the radical with a nearby VdO(1) (Figure 4).
through a transition state, TSS2a, in which the surface vanadyl O(1) attacks the C(2) of propene to form a C(2)O(1) bond (Figure S2, Supporting Information). The energy barrier of this step is 20.1 kcal/mol. Following the chemisorption, the hydrogen atom on C(2) of IMS2a can be abstracted by a nearby O(1) atom, leading to IMS2b with a low energy barrier of 5.7 kcal/mol (TSS2b). After IMS2b formation, the hydrogen atom on the nearby O(1) site can easily rehit the O(1) atom that bonded with C(2), forming an adsorbed 2-propenol with an energy barrier of only 8.2 kcal/mol. The 2-propenol is not stable on the surface and will desorb easily. Instead of forming 2-propenol, the intermediate IMS2b, however, can be isomerized into a more stable intermediate, IMS2c, via the hydrogen atom migration between two adjacent lattice O(1) atoms. Our calculations showed that this process is barrierless. Then the hydrogen atom on the surface can attack the C(1) carbon to form a new CH bond, leading to the formation of acetone. This process only requires overcoming an energy barrier of 1.1 kcal/mol. Compared with 2-propenol formation, acetone formation is obviously more favorable and 2-propenol may serve as a short-lived intermediate. In the experimental study of the selective oxidation of propane to acetic acid 810
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
Figure 3. Energy profile of acetone oxidation at the O(1) site on V2O5(001).
An additional hydrogen atom of the surface allyl at the O(1) site can be abstracted by a nearby O(1) to form acrolein. The energy barrier of this process is 22.9 kcal/mol, and the whole process is endothermic by 0.2 kcal/mol (Figure 5). Similar to the case of propene formation, the surface hydroxyl group O(1)H can also abstract a hydrogen from allyl to form acrolein and water. This process is exothermic by 9.6 kcal/mol with an energy barrier of 24.3 kcal/mol (Figure 5). Obviously, for acrolein formation, O(1) and the O(1)H group have similar activities. The reaction pathway through hydrogen abstraction by O(2) was also calculated, but was found to be much less favorable energetically (data not shown). As mentioned above, a surface allyl group can also be formed on the O(2) site. The oxidation of allyl at O(2) to acrolein is, however, quite difficult. We have found that the most feasible pathway of hydrogen abstraction is through a nearby O(2) atom, with an energy barrier of 35.9 kcal/mol, which is 13.0 kcal/mol higher than that at the O(1) site (Figure S3, Supporting Information). Moreover, the hydroxyl group at O(2) is found to be more inert for acrolein formation with a high energy barrier of 45.7 kcal/mol. Instead of oxidation at the O(2) site, we examined the transfer of allyl from O(2) to the O(1) site through a ringlike transition
state, TSS4a, as shown in Figure S4 (Supporting Information). The energy barrier is 30.5 kcal/mol, lower than that of H abstraction. After that, the allyl can be oxidized to acrolein at the O(1) site as shown above. Taken together, the two pathways leading to further oxidation of propene at the O(1) site are competitive. The rate-limiting step of propene oxidation to acetone is the chemisorption of propene on the surface O(1) site with an energy barrier of 20.1 kcal/mol, while the rate-limiting step of propene oxidation to acrolein is the initial activation of the allylic CH bond of propene on the O(1) site. 3.4. Oxidation of Acrolein. First, we examined the further oxidation of acrolein before desorption. Without desorption, the activation of both aldehyde and olefin CH bonds requires overcoming an energy barrier of more than 33 kcal/mol (Figure S5, Supporting Information), which is much higher than the desorption energy of acrolein (13.9 kcal/mol, Figure 5). Therefore, acrolein is inclined to desorb from the surface by depriving one lattice O(1) atom after formation. Three possible oxidation pathways of the desorbed acrolein on the surface are shown in Figure S6 (Supporting Information). Along the first pathway, the surface O(1) site can attack the C(2) atom of acrolein through a transition state (TSS6a) that is 811
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
Figure 4. Energy profile of allylic CH bond activation of propene on the V2O5(001) surface.
23.7 kcal/mol higher than the reactant energetically, forming the adsorbed intermediate species IMS6a. Then the abstraction of the aldehyde hydrogen atom by another O(1) results in simultaneous CC bond breaking, yielding a surface vinyl oxide, a surface hydroxyl group, and a molecule of CO. The transition state TSS6b involves simultaneous CH and CC bond elongation and O(1)H bond formation (Figure S6, Supporting Information). The energy barrier of this step from IMS6a is 7.3 kcal/mol. In the first step of the second pathway, acrolein dissociatively adsorbs on the surface through aldehyde CH bond breaking, forming a surface O(1)H group and an adsorbed acrylic group (IMS6b) on another O(1) via transition state TSS6c. This step is exothermic by 12.7 kcal/mol and has a barrier of only 13.8 kcal/mol. In the second step, the H atom of the nearby O(1)H group transfers readily to the O(1) atom of the acrylic group IMS6b, resulting in the formation of acrylic acid by breaking the VO(1) bond. The energy barrier of this step is only 0.2 kcal/mol, and the whole process is exothermic by 15.4 kcal/mol. Compared with the first pathway that leads to CO, it is clear that the second pathway is preferred both thermodynamically and kinetically. The third pathway is a one-step reaction leading to the formation of acrylic acid directly, in which the lattice O(1) inserts into the CH bond of the aldehyde group, with an energy barrier
of only 8.5 kcal/mol. In the transition state TSS6e, the aldehyde CH bond and VO(1) bond are elongated to 1.56 and 2.41 Å, while the O(1)C and O(1)H distances are shortened to 1.34 and 1.31 Å, respectively. Due to the significantly weakened VO(1) bond, the desorption energy of acrylic acid is only 3.2 kcal/mol. The reaction energy of this pathway is 12.2 kcal/mol. Compared with the second pathway, this reaction pathway is largely favored kinetically. Overall, the most favorable pathway for acrolein oxidation over the V 2O5(001) surface is through O(1) insertion into the aldehyde CH bond to form acrylic acid. In other words, acrolein oxidized from propene is readily further oxidized into acrylic acid. 3.5. Oxidation of Acrylic Acid. In this section, we explore acrylic acid oxidation over V2O5(001). The lowest energy pathway is shown in Figure S7 (Supporting Information). At the first step, an O(1) atom attacks the middle carbon of acrylic acid through the transition state TSS7a to form an adsorbed acrylic acid (IMS7a) on the surface. The distance between the middle carbon atom and lattice O(1) is shortened from 1.74 Å in TSS7a to 1.49 Å in IMS7a. This step is endothermic by 18.0 kcal/mol and has a barrier of 21.6 kcal/mol. The subsequent migration of the hydrogen atom of the carboxyl group to a nearby O(1) site can occur easily with an energy barrier of 9.8 kcal/mol. Simultaneously, the CC bond is activated with the bond length elongated 812
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
Figure 5. Energy profiles of acrolein formation through O(1) and O(1)H on V2O5(001).
from 1.56 Å in IMS7a to 1.77 Å in TSS7b. Through the TSS7b pathway, the hydrogen atom of the carboxyl group migrates to the nearby lattice O(1) site to form one hydroxyl group, and the CC bond breaks thoroughly to form the surface vinyl oxide (CH2CHO*) and product CO2 finally. 3.6. Oxidation of Vinyl Oxide CH2CHO*. Both oxidation reactions of acrolein and acrylic acid can give rise to surface vinyl oxide species CH2CHO*. Here we explore the oxidation reaction of CH2CHO* over the surface of V2O5(001). The most favorable reaction channel gave oxidation products of formaldehyde and CO (Figure S8, Supporting Information). In this pathway, a surface O(1) atom attacks the terminal carbon of vinyl oxide to form a ringlike intermediate, IMS8a, and then the aldehyde hydrogen atom is abstracted by another nearby O(1), followed by concurrent splitting of the CC bond, leading to the formation of formaldehyde and CO. The surface formaldehyde desorbs by overcoming an energy barrier of 11.1 kcal/mol. This process is endothermic, and the total reaction energy is 37.4 kcal/mol (Figure S8). 3.7. Formation and Oxidation of Acetic Acid. As mentioned in section 3.2, acetone oxidation over the V2O5(001) surface can produce acetate and formaldehyde on the catalyst surface. In this section, we look into the reaction of acetate over the V2O5(001) surface. As shown in Figure S9 (Supporting Information), a hydrogen
atom of the nearby surface hydroxyl group migrates to the O(1) atom of CH3COO* through TSS9, resulting in the formation of acetic acid. This step is exothermic by 0.6 kcal/mol, and the energy barrier is only 0.2 kcal/mol (Figure S9). The energy cost of acetic acid desorption is 7.1 kcal/mol. Therefore, acetate on the surface can readily transform to gas-phase acetic acid given that the density of the surface hydroxyl is high. The first step in the oxidation of acetic acid is the dissociative adsorption of acetic acid by breaking the methyl CH bond to yield IMS10a (Figure S10, Supporting Information). Starting from IMS10a, the carboxyl hydrogen atom can be abstracted by a nearby lattice O(1) site, accompanying CC bond breaking in TSS10b, yielding CO2 and formaldehyde simultaneously. The rate-limiting step is methyl CH bond breaking, with an energy barrier of 37.1 kcal/mol. It is clear that the oxidation of acetic acid over V2O5(001) is not very easy, so there should be some amount of acetic acid in the propane ODH products over V2O5. This is in good agreement with the experimental data of Lin et al,23 who found that, other than carbon oxides, acetic acid is the major byproduct of propane ODH over the MoVTeNbO catalyst. 3.8. Oxidation of Formaldehyde. As shown in the previous sections, we have confirmed the production of formaldehyde in the process of propane ODH. Figure S11 (Supporting Information) presents the energy profile of formaldehyde oxidation on the 813
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
Figure 6. Lowest energy profiles of deep oxidations of acetone and propene.
surface. The hydrogen atom of formaldehyde is abstracted by O(1) to form a hydroxyl group on the surface and a formyl radical HCO in the gas phase. The energy barrier is calculated to be 21.5 kcal/mol, and the radical intermediate is only 1.9 kcal/mol more stable than the transition state TSS11. The HCO radical is unstable in the gas phase, and it can react with the surface O(1) to break the CH bond readily without an energy barrier, giving CO and a surface hydroxyl. This process is exothermic by 41.7 kcal/mol, and the whole process of formaldehyde oxidation over V2O5(001) is exothermic by 22.1 kcal/mol. Obviously, formaldehyde can react with lattice O(1) easily, which agrees well with the experimentally observed activity of V2O5 toward formaldehyde.21 In the experimental study of the ODH of propane over VOx/γ-Al2O3 materials,21 formaldehyde was identified by isotopically labeled V18Ox species. Formaldehyde appears directly after propene followed by CO and CO2. This observation agrees with our calculations that formaldehyde is an intermediate product of propene oxidation to COx. 3.9. Oxidation of CO. Finally, we examined the oxidation of CO to CO2 by lattice oxygen. As shown in Figure S12 (Supporting Information), the carbon atom of CO interacts with the lattice oxygen atom O(1) to form the second CO bond. In the transition state TSS12a, the distance between C and O(1) is 1.61 Å, and the VO(1) bond is lengthened to 1.71 Å. This reaction with an energy barrier of 23.3 kcal/mol is exothermic by 30.kcal/mol and yields one molecule of CO2. The oxidation of CO by the O(2) site is found to be much less favorable than that by O(1) since the reaction is endothermic by 7.4 kcal/mol and the energy barrier is 2.6 kcal/mol higher.
profiles of these two channels on the basis of the above calculation results in Figure 6. It is worth noting that to fully compare the reaction rates of various reaction channels, we have to estimate the entropy term [exp(ΔS/R)] that affects the pre-exponential factor in the rate equation. However, the current periodic DFT methods cannot compute the entropy term accurately mainly due to the errors in calculating the low-frequency vibrational modes. Nevertheless, it appears that the pre-exponential factors for similar types of reactions are rather similar; i.e., for EleyRideal (ER) mechanism reactions, it is around 109, and for LangmuirHinshelwood (LH) mechanism reactions, it is around 1013. In this work, the first reaction step of the gas-phase reactants belongs to the ER mechanism; therefore, it has a low pre-exponential factor due to the significant loss of entropy on going from the gas-phase molecule to the TS on the surface. By contrast, the reactions of adsorbed species on the surface belong to the LH mechanism. According to our calculations, most of the energy barriers of the first ER mechanism steps are higher than those of the subsequent reactions, except for the case of acrylic acid. Therefore, the pre-exponential factor should not affect the qualitative character of the energy profile shown in Figure 6. It should also be noted that, although the GGA functional PBE used in the calculation has limited accuracy in predicting energy barriers in general,43 the variations among similar systems are usually accurate; i.e., GGA functionals have good performance in predicting trends, which is our main focus.44,45 As shown in Figure 6, the oxidation of acetone on the V2O5(001) surface is relatively difficult. The initial dissociation adsorption by breaking the CH bond has to overcome an energy barrier of 35.0 kcal/mol, which is higher than that of the initial CH bond activation of propane. The following oxidation step leading to the formation of acetate and formaldehyde has an energy barrier of 31.5 kcal/mol. Considering that the first dissociation adsorption step belongs to the ER mechanism reaction, which has a pre-exponential factor of ca. 104 orders, smaller than that of the surface reactions, the rate-limiting step of acetone
4. DISCUSSION On the basis of the calculations shown above, it is obvious that the byproduct in the production of propene through propane ODH reaction over the V2O5 catalyst mainly originates from the further oxidation of propene and the formation and oxidation of acetone from isopropoxide. We summarized the lowest energy 814
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
oxidation to acetate and formaldehyde is the initial CH bond activation. Similarly, in the subsequent oxidation of gas-phase acetic acid, the rate-limiting step is also the methyl CH bond activation with an energy barrier of 37.1 kcal/mol. After the formation of formaldehyde, the following oxidation steps finally leading to CO2 can proceed easily. Therefore, under the typical reaction temperature of propane ODH on the vanadium-based catalyst of 600800 K, both acetone and acetic acid can be oxidized to COx. However, we can anticipate that acetone and acetic acid may be the main byproducts in the ODH reaction of propane over the vanadium catalyst on the basis of the energy profiles at lower temperatures. Experimentally, both acetone and acetic acid have been detected in the selective oxidation of propane over several mixed metal oxide catalysts.4649 Compared with acetone, the deep oxidation of propene is much easier on the V2O5(001) surface (Figure 6). This is mainly due to the high reactivity of the allyl CH bond. The energy barrier of initial allyl CH bond activation on the surface is only 19.9 kcal/mol. Acrolein is even more reactive than propene on the surface. The formation of acrylic acid from acrolein occurs readily through oxo insertion into the aldehyde CH bond. The subsequent oxidation of acrylic acid has a moderate energy barrier and yields formaldehyde and CO through a vinyl oxide intermediate. Compared with the channel of acetone oxidation, the deep oxidation of propene is very favored both kinetically and thermodynamically. Overall, the order of the relative reactivity of these hydrocarbons over V2O5(001) is
Figure 7. Reaction network of propane ODH on V2O5(001).
surface is much more difficult than depriving an O(1) atom by the reactant. Our calculation results show that, after the dehydrogenation of the reactant, the energy barriers of the oxidation step are very high on O(2) sites. Therefore, the deep oxidations of both acetone and propene take place on O(1) sites as shown in Figure 7. In other words, on O(2) sites the CH bond activation is no longer the rate-limiting step of the reaction. Instead, oxygen addition and CC bond breaking are blocked at O(2). This implies that O(2) is an active site with high propene selectivity. Looking at the whole reaction network of propane ODH on the V2O5(001) surface (Figure 7), we can see that the reaction branches at the surface intermediate isopropoxide, from which propene and acetone are formed with similar energy barriers on O(1), while on O(2) the formation of acetone is largely inhibited. It is obvious that the deep oxidations in propane ODH reaction are mostly attributed to the vanadyl oxygen VdO(1) sites on the V2O5(001) surface, and the bridging O(2) site is critical for propene selectivity. This may account for the experimental finding that the highly dispersed vanadium catalyst, such as VOx species, has higher selectivity, since the O(1):O(2) ratio decreases with an increase of the dispersion extent of vanadium oxide. In practice, most vanadium catalysts for propane ODH are supported on various metal oxides, on which new types of two-coordinated oxygen site VOM (M represents the metal atoms of the support) are formed. We can envision that the nature of the VOM sites and dispersion degree will mainly determine the catalyst performance. To get an optimized catalyst, the composition of the metal oxide support needs to be tuned for achieving desirable reactivity and selectivity.
acetic acid ≈ acetone < propane < acrylic acid < propene < acrolein
This order of reactivity is generally consistent with the relative CH bond strengths in these molecules. The calculated and experimental CH bond energies are given in Table S2 (Supporting Information). Notably, the rate-limiting steps in their oxidation reactions are all CH bond activations except for the case of acrylic acid. The methyl CH bonds in acetic acid and acetone have the highest bond energies, followed by the methylene CH bonds of propane. The allyl CH bond in propene and aldehyde CH bond in acrolein are much weaker due to the conjugate effect. The experimental studies of selective oxidation of propane over various mixed oxide catalysts showed a similar order of reactivity. For example, Lin et al.23 found that on the Te/vanadium phosphorus oxide (VPO) catalyst there is almost no acrolein detected, while the selectivity of acrylic acid is as high as 37%. In another propane ODH reaction over TeP/NiMoO, Kaddouri et al.50 found a 13% yield of acrolein, which is still much lower than that of acrylic acid of 23%. Using the K-doped MoVSbO catalysts, Blasco et al.41 investigated the selective oxidation of propane and found that the highest selectivity of acrylic acid is 34.8%. On the contrary, the observed selectivity of acrolein is lower than 0.1%. Actually, in industry, Nippon Kayaku Co. Ltd. has reported an acrylic acid yield of 97.5% using acrolein and MoVCuFeMnMgPO as the raw material and catalyst, respectively, at a low temperature of 210 °C. Even using the propene as the raw material, Nippon Shokubai Co. Ltd. can achieve a 73% acrylic acid yield.51 It is worth noting that the energy profiles in Figure 6 are mostly based on the reactions proceeding on O(1) sites. It is found from both the above calculations and our previous work26 that the bridge lattice oxygen O(2) has a similar ability in activating the CH bonds. However, O(2) is a much weaker oxidant than O(1) in the respect that depriving an O(2) from the catalyst
5. CONCLUSIONS We performed a comprehensive study of all the reaction channels in the deep oxidation of propane ODH over the V2O5(001) surface by using the periodic DFT method. 815
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
Starting from the isopropoxide intermediate on the O(1) site of V2O5(001), there are two competitive reaction pathways leading to the formation of the product propene and byproduct acetone. The deep oxidations of acetone and propene on the surface are two origins of deep oxidation byproducts. The deep oxidation of acetone is more difficult than that of propene, but under the normal reaction temperature of propane ODH both of them can be oxidized all the way to CO2 on O(1). The energy barriers of each surface oxidation step predict the hydrocarbon reactivity order:
(8) Deo, G.; Wachs, I. E.; Haber, J. Crit. Rev. Surf. Chem. 1994, 4, 141. (9) Wachs, I. E.; Wechuysen, B. M. Appl. Catal., A 1997, 157, 67. (10) Eon, J. G.; Olier, R.; Volta, J. C. J. Catal. 1994, 145, 318. (11) Khodakov, A.; Olthof, B.; Bell, A. T.; Iglesia, E. J. Catal. 1999, 181, 205. (12) Khodakov, A.; Yang, J.; Su, S.; Iglesia, E.; Bell, A. T. J. Catal. 1998, 177, 343. (13) Chen, K. D.; Bell, A. T.; Iglesia, E. J. Phys. Chem. B 2000, 104, 1292. (14) Grasselli, R. K. Catal. Today 1999, 45, 141. (15) Chaar, M. A.; Patel, D.; Kung, H. H. J. Catal. 1988, 109, 463. (16) Corma, A.; Nieto, J. M. L.; Paredes, N. J. Catal. 1993, 144, 425. (17) Argyle, M. D.; Chen, K. D.; Iglesia, E.; Bell, A. T. J. Catal. 2002, 208, 139. (18) Kondratenko, E. V.; Cherian, M.; Baerns, M. Catal. Today 2005, 99, 59. (19) Bielanski, A.; Haber, J. Oxygen in Catalysis; Marcel Dekker: New York, 1991; p 320. (20) Chen, K. D.; Iglesia, E.; Bell, A. T. J. Catal. 2000, 192, 197. (21) Kondratenko, E. V.; Steinfeldt, N.; Baerns, M. Phys. Chem. Chem. Phys. 2006, 8, 1624. (22) Ai, M. Catal. Today 1998, 42, 297. (23) Lin, M.; Desai, T. B.; Kaiser, F. W.; Klugherz, P. D. Catal. Today 2000, 61, 223. (24) Redfern, P. C.; Zapol, P.; Sternberg, M.; Adiga, S. P.; Zygmunt, S. A.; Curtiss, L. A. J. Phys. Chem. B 2006, 110, 8363. (25) Klisinska, A.; Haras, A.; Samson, K.; Witko, M.; Grzybowska, B. J. Mol. Catal. A 2004, 210, 87. (26) Fu, H.; Liu, Z. P.; Li, Z. H.; Wang, W. N.; Fan, K. N. J. Am. Chem. Soc. 2006, 128, 11114. (27) Pudar, S.; Oxgaard, J.; Chenoweth, K.; van Duin, A. C. T.; Goddard, W. A., III. J. Phys. Chem. C 2007, 111, 16405. (28) Rozanska, X.; Fortrie, R.; Sauer, J. J. Phys. Chem. C 2007, 111, 6041. (29) Rozanska, X.; Sauer, J. Int. J. Quantum Chem. 2008, 108, 2223. (30) Rozanska, X.; Kondratenko, E. V.; Sauer, J. J. Catal. 2008, 256, 84. (31) Gilardoni, F.; Bell, A. T.; Chakraborty, A.; Boulet, P. J. Phys. Chem. B 2000, 104, 12250. (32) Cheng, M. J.; Chenoweth, K.; Oxgaard, J.; van Duin, A.; Goddard, W. A., III. J. Phys. Chem. C 2007, 111, 5115. (33) Rozanska, X.; Sauer, J. J. Phys. Chem. A 2009, 113, 11586. (34) Nguyen, N. H.; Tran, T. H.; Nguyen, M. T.; Le, M. C. Int. J. Quantum Chem. 2009, 113, 11586. (35) Junquera, J.; Paz, O.; Sanchez-Portal, D.; Artacho, E. Phys. Rev. B 2001, 64, 235111. (36) Troullier, N.; Martins, J. L. Phys. Rev. B 1991, 43, 1993. (37) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865. (38) Blackford, L. S.; Choi, J.; Cleary, A.; D’Azevedo, E.; Demmel, J.; Dhillon, I.; Dongarra, J.; Hammarling, S.; Henry, G.; Petitet, A.; Stanley, K.; Walker, D.; Whaley, R. C. ScaLAPACK Users’ Guide; Society for Industrial and Applied Mathematics: Philadelphia, PA, 1997. (39) Liu, Z. P.; Jenkins, S. J.; King, D. A. J. Am. Chem. Soc. 2004, 126, 10746. (40) Liu, Z. P.; Hu, P. J. Am. Chem. Soc. 2003, 125, 1958. (41) Blasco, T.; Botella, P.; Concepcion, P.; L opez Nieto, J. M.; Martinez-Arias, A.; Prieto, C. J. Catal. 2004, 228, 362. (42) Finocchio, E.; Busca, G.; Lorenzelli, V.; Willey, R. J. J. Catal. 1995, 151, 204. (43) Zhao, Y.; Truhlar, D. G. J. Phys. Chem. A 2005, 109, 5656. (44) Hammer, B.; Hansen, L. B.; Norskov, J. K. Phys. Rev. B 1999, 59, 7413. (45) Hammer, B.; Norskov, J. K. Adv. Catal. 2000, 45, 71. (46) Yi, X. D.; Zhang, X. B.; Weng, W. Z.; Wan, H. L. J. Mol. Catal., A 2007, 277, 202.
acetic acid ≈ acetone < propane < acrylic acid < propene < acrolein
which is in agreement with most of the experimental results. It was found that the O(2) bridging site on V2O5(001) is inert for both acetone formation from isopropoxide and further oxidations of acetone and propene. Although O(2) has a similar ability to activate CH bonds, depriving the O(2) atom from the surface, which is required in the deep oxidation steps, is much more difficult. Therefore, the O(2) site is critical for propene selectivity in propane ODH reaction, implying that a highly dispersed V2O5 species is desirable for the preparation of catalysts with high performance in propane ODH to propene.
’ ASSOCIATED CONTENT Supporting Information. Twelve figures describing the energy profiles of various reaction pathways in the deep oxidation reaction network and two tables containing the imaginary frequency values of the transition states and the CH bond energies. This material is available free of charge via the Internet at http://pubs.acs.org.
bS
’ AUTHOR INFORMATION Corresponding Author
*Fax: +86-21-55665572. E-mail:
[email protected] (W.-N.W.);
[email protected] (K.-N.F.).
’ ACKNOWLEDGMENT This work was supported by the National Science Foundations of China (Grants 20828003, 20973040, and 31070642), National High Technology Research Program of China (Grant 2006AA02A320), National Major Basic Research Program of China (Grants 2009CB623506, 2009CB918600, and 2011CB808505), Zhejiang Provincial Natural Science Foundation of China (Grant Y4090387), Science & Technology Commission of Shanghai Municipality (Grant 08DZ2270500), and Shanghai Leading Academic Discipline Project (Grant B108). We are grateful to the computer center of Fudan University for their allocation of computer time. ’ REFERENCES (1) Cavani, F.; Trifiro, F. Catal. Today 1995, 24, 307. (2) Mamedov, E. A.; Cortes Corberan, V. Appl. Catal., A 1995, 127, 1. (3) Le Bars, J.; Auroux, A.; Forissier, M.; Vedrine, J. C. J. Catal. 1996, 162, 250. (4) Grabowski, R. Catal. Rev. 2006, 48, 199. (5) Chaar, M. A.; Patel, D.; Kung, H. H. J. Catal. 1988, 109, 463. (6) Corma, A.; Nieto, J. M. L.; Paredes, N. J. Catal. 1993, 144, 425. (7) Hardcastle, F. D.; Wachs, I. E. J. Mol. Catal. 1988, 46, 173. 816
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
The Journal of Physical Chemistry C
ARTICLE
(47) Concepcion, P.; Botella, P.; Lopez Nieto, J. M. Appl. Catal. A 2004, 278, 45. (48) Hasan, M. A.; Zaki, M. I.; Pasupulety, L. J. Phys. Chem. B 2002, 106, 12747. (49) Luo, L.; Labinger, J. A.; Davis, M. K. J. Catal. 2001, 200, 222. (50) Kaddouri, A.; Mazzocchia.; Tempesti, E. Appl. Catal., A 1999, 180, 271. (51) Lin, M. M. Appl. Catal., A 2001, 207, 1.
817
dx.doi.org/10.1021/jp208639t |J. Phys. Chem. C 2012, 116, 807–817
