5680
J. Phys. Chem. B 2007, 111, 5680-5683
Probing the Effect of Local Structure on the Thermodynamic Redox Properties of V+5: A Comparison of V2O5 and Mg3(VO4)2 Parag R. Shah, John M. Vohs, and Raymond J. Gorte* Department of Chemical and Biomolecular Engineering, UniVersity of PennsylVania, Philadelphia, PennsylVania 19104 ReceiVed: February 22, 2007; In Final Form: March 20, 2007
Coulometric titration, an electrochemical method for measuring oxidation isotherms, has been used to characterize the redox properties of V2O5 and Mg3(VO4)2 between 823 and 973 K. V2O5 shows distinct regions in the isotherms corresponding to equilibrium with mixtures of V2O3 and V2O4 and of V2O4 and V2O5. From this data, the enthalpies for oxidation of V2O3 to V2O4 and for V2O4 to V2O5 are shown to be -380 ( 10 and -285 ( 20 kJ mol-1 O2, respectively. Oxidation isotherms for Mg3(VO4)2 exhibit a single step between the oxidized sample (all V+5) and a completely reduced sample (all V+3). The enthalpy of oxidation is found to increase with the oxidation state of the sample, from -370 ( 30 kJ mol-1 O2 at an O:V ratio of 1.5 to -460 ( 10 kJ mol-1 O2 at an O:V ratio of 2.5.
Introduction The catalytic properties of vanadia depend strongly on the local composition surrounding the vanadium cations. This is shown by the strong effect that a support can have on the activity and selectivity of supported vanadia1 and by the fact that mixed oxides of vanadia can exhibit significantly different catalytic properties from that of pure vanadia. For example, Mg3(VO4)3 is active and selective for oxidative dehydrogenation of propane, while V2O5 is not.2 Since the Mars van Krevelen mechanism is involved in these oxidation reactions, characterizing the thermodynamics of reduction/oxidation (redox) and the strength of V-O bonds in vanadia catalysts would be informative, but direct measurements of these properties are not available because they are difficult to perform. Characterization of metal-oxygen bond strengths is usually qualitative and indirect. Temperature-programmed reduction (TPR), a measurement of the temperature at which a sample is rapidly reduced in a flow reactor, is sometimes used to determine the relative reducibility of a series of materials,3 but it is very difficult to quantify the energetics of reduction in TPR. Relative bond strengths are also inferred from changes in the metaloxygen vibrational frequencies4 and from UV-visible absorption energies in the solid.5 Energetics of oxidation and reduction are sometimes calculated from calorimetric measurements of O2 adsorption, but we believe these measurements to be unreliable. In addition to difficulties associated with heat losses in microcalorimetry, the measured heats on high-surface-area samples will be, at best, average values if the adsorption enthalpies are a function of stoichioimetry when adsorption is irreversible.6 In the case of oxidation of reduced vanadia, an O2 pulse will almost certainly saturate the external part of the sample before beginning to oxidize internal parts of the sample. Finally, because the entropy of oxidation cannot be determined in the absence of equilibrium data, an accurate and complete thermodynamic description for the redox properties of the catalyst is not possible. * Corresponding author. E-mail:
[email protected]. Fax: 215-5732093.
The most direct way to determine redox thermodynamics is by measuring the equilibrium stoichiometries for the oxides as a function of temperature and oxygen fugacity (P(O2)).7 For example, when the reaction 2V2O4 ) 2V2O3 + O2 is in equilibrium, thermodynamic quantities can be calculated from the following equations:
∆G ) RT ln[P(O2)]
(1)
∆H ) -Rδ ln[P(O2)]/δ(1/T)
(2)
∆S ) (∆H - ∆G) /T
(3)
For most oxides at the temperatures of interest, the range of P(O2)’s relevant for thermodynamic measurements is very low and must be established by equilibrium with a redox couple, such as a mixture of H2 and H2O or CO and CO2. For example, at 10-14 atm of O2 and 1000 K (CO/CO2 ) 6 × 10-4), the flux of O2 molecules to a surface is so small that 24 h would be required to expose the sample to one monolayer of O2, an exposure that is certainly insufficient to establish equilibrium with a bulk sample. Furthermore, at a partial pressure of 10-20 atm and 1000 K, our experimental apparatus would contain less than one molecule. Therefore, it is important that the P(O2) in most equilibrium measurements not be viewed as an actual O2 pressure. In this study, redox isotherms were measured using Coulometric titration.7 In short, the sample is placed in a sealed container, separated from the atmosphere by an ion-conducting membrane, yttria-stabilized zirconia (YSZ) in our case. Electrodes on either side of the membrane are used to measure the potential across the membrane, and the potential can be related to the P(O2) through the Nernst Equation:
V ) -RT/[4F] ln[P(O2)/0.21 atm]
(4)
Oxygen can be added to the sample by passing current through the membrane, with 1 C of charge corresponding to 5.18 µmol of O2-. In this, our first study of the thermodynamic redox properties of vanadia catalysts, we compared bulk vanadia with magnesium
10.1021/jp071498v CCC: $37.00 © 2007 American Chemical Society Published on Web 04/25/2007
Thermodynamic Redox Properties of V+5
J. Phys. Chem. B, Vol. 111, No. 20, 2007 5681
Figure 1. (a) Sheet-like structure of V2O5. (b) Unit cell of Mg3(VO4)2.
orthovanadate, Mg3(VO4)2, in order to gain insights into how the local environment in the solid affects V-O bonding. As shown in Figure 1a, bulk vanadia has a sheet-like structure with V-O-V bonds and reduction of V2O5 can occur sequentially, to V2O4 and then to V2O3. On the other hand Mg3(VO4)2 has an orthorhombic, magnesium-deficient spinel structure (Figure 1b) and all of the V-O bonds link to Mg,8 implying that the oxygen should exist in a very different environment in the two solids. Reduction of Mg3(VO4)2 is known to transform the solid into a cubic Mg3V2O6 which has a “cation-stuffedspinel structure”.8 Therefore, removal of an oxygen from a site in the Mg3(VO4)2 must cause direct reduction of neighboring V+5 to V+3. The redox isotherms presented here support the above picture of reduction in V2O5 and Mg3(VO4)2 and also provide quantitative information on the energetics of the redox processes in both materials. Experimental Magnesium orthovanadate (Mg3(VO4)2) was prepared by the Pechini method.9 Stoichiometric amounts of Mg(NO3)2‚6H2O (Fisher Scientific) and NH4VO3 (Sigma) were dissolved in water acidified with a small quantity of nitric acid. Aqueous citric acid (Aldrich) was added to this solution such that the normality of the citric acid was 1.1 times the total metal-ion valence. The resulting solution was stirred vigorously for an hour and the excess water was removed by evaporation. The remaining solid was dried overnight at 400 K and was then decomposed to the oxide by calcining in air at 973 K for 6 h. The product oxide was then ground using a mortar and pestle and characterized by X-ray diffraction (XRD), using a Rigaku Geigerflex Diffractometer with a Cu KR radiation source (λ ) 1.5405 Å). The diffraction pattern was found to be consistent with published data for the Mg3(VO4)2 phase.8 The V2O5 powder used in this study was obtained from Aldrich Chemical Co. The system used for the Coulometric titration measurements has been described elsewhere.10 The apparatus consists of a YSZ tube (12.7 mm o.d., 9.5 mm i.d., and 15 cm long) with a YSZ pellet plugging one end by a glass seal. The YSZ pellet had conducting electrodes applied to each side (Pt paste on the sample side; a mixture of La0.8, Sr0.2MnO3, and YSZ on the air side), with Pt lead wires pasted onto the electrodes for measuring the electrode potentials. A small vial containing the sample (∼150 mg) was placed inside the YSZ tube, near the YSZ pellet, and this end of the tube was inserted into a furnace. The opposite end of YSZ tube was sealed with a Swagelok UltraTorr fitting that had welded stainless-steel tube for flowing gases over the sample, before being sealed by valves. The leak rate of the apparatus could be monitored by measuring the potential across the YSZ wafer in the absence of a sample and was only considered acceptable if it were significantly less than 1 µmol O2 per day.
Figure 2. Oxidation isotherms for vanadia: 9, 823 K; b, 873 K; 2, 923 K. The absolute amount of oxygen added to the sample from its fully reduced state is plotted on the left, relative to the weight of the fully oxidized sample. The sample is assumed to be fully oxidized at the highest P(O2).
The ability to flow gases over the samples was crucial for the experiments with vanadia because vanadia forms volatile hydroxides in the presence of steam.11 In most of the present experiments, the samples were reduced in CO before beginning the measurements and the P(O2) was then established by equlibrium between CO and CO2 according to eq 5:
P(O2) ) [P(CO2)/P(CO)]2/Keq
(5)
Because volatility was less of a problem with Mg3(VO4)2, some experiments were performed with H2-H2O mixtures on this sample, and the results were indistinguishable. With V2O5, the data were taken between 823 and 923 K, a temperature range low enough to avoid the melting temperature of 963 K but high enough to allow relatively rapid equilibration. For Mg3(VO4)2, the measurement temperatures were kept above 873 K to prevent the formation of carbonates. The criterion we used for establishing equilibrium in Coulometric titration was that the potential of the oxygen sensor changed by less than 0.5 mV/h. After the addition of oxygen, the time required to reach equilibrium was typically one to 2 days. After addition of oxygen, the equilibrium P(O2) was measured at all the desired temperatures, so that equilibrium measurements at different temperatures were at constant O:V ratios. The results were determined to be reversible with respect to the temperature changes, but it was not possible to check for reversibility by removing oxygen from the system because pumping rates were too slow in that direction. The partial molar enthalpies (∆H) were calculated using eq 2 and the measured P(O2) at different temperatures. The uncertainty in the calculated enthalpies was calculated using the standard error of the slope for a linear regression through the isotherm data. Results and Discussion Figure 2 shows oxygen isotherms for V2O5 between 823 and 923 K, for P(O2) ranging between 10-22 and 10-2 atm. The experiments were started with the reduced sample at 10-22 atm, and the absolute amounts of oxygen added to it are plotted on the left side. The calculated O/V ratios are plotted on the right side, assuming that all vanadium was present in the +5 oxidation state at the end of the experiment. As expected, all of the V in the initial state of the sample existed as +3 ions. It is immediately apparent that there are well-defined regions of the isotherm in Figure 2, corresponding to the transition from V2O3 to V2O4 and from V2O4 to V2O5. The vertical lines in the
5682 J. Phys. Chem. B, Vol. 111, No. 20, 2007
Shah et al.
Figure 4. Differential oxidation enthalpies, -∆H, for V2O5 [2] and Mg3(VO4)2 [O] as a function of stoichiometry. Figure 3. Oxidation isotherms for Mg3(VO4)2: 2, 873 K; 9, 923 K, b, 973 K. The absolute amount of oxygen added to the sample from its fully reduced state is plotted on the left, relative to the weight of the fully oxidized sample. The sample is assumed to be fully oxidized at the highest P(O2). The O/V ratio was calculated without the oxygen associated with MgO.
isotherm are consistent with the presence of two phases in equilibrium with each other. When both phases are present, the addition of oxygen simply changes the relative amounts of each phase. The experimental values of P(O2) for which there is an equilibrium between V2O3 and V2O4 and between V2O4 to V2O5 agree well with the reported P(O2), which are 3 × 10-19 atm and 5.5 × 10-6 atm, respectively,12 thus providing confidence in the reliability of the data. These P(O2)’s were obtained from standard handbook values of ∆G for these reactions at 873 K. Finally, the isotherm shows a gradual transition region near O/V ratios of 2, suggesting that there is a single phase of mixed valency. In earlier measurements with Cu,10 there was a sharp transition between the two regions of the oxidation isotherm: for Cu-Cu2O phases followed by Cu2O-CuO phases; showing that the results for vanadia are not likely associated with instrumentation issues. Figure 3 shows the oxygen isotherms for Mg3(VO4)2 between 873 and 973 K, for P(O2) ranging between 10-18 and 10-2 atm. (A higher temperature range was used with Mg3(VO4)2 because of the tendency for Mg+2 to form a carbonate at lower temperatures.) Again, we have plotted both the amounts of oxygen added to the sample (on the left) and the calculated O/V ratio (on the right) assuming that only V cations can be reduced and that all V is present in the +5 oxidation state at the highest P(O2). As in the case of V2O5, there is complete reduction of V to its +3 oxidation state at the lowest P(O2). The most noticeable difference in the data for Mg3(VO4)2 compared to that for V2O5 is that the isotherm exhibits a single vertical feature, suggesting that there is an equilibrium between a phase containing only V+3 (i.e., Mg3V2O6) and a phase containing only V+5 (i.e., Mg3(VO4)2). As pointed out in the Introduction, there are no V-O-V linkages in Mg3(VO4)2, so that the charge associated with removing an oxygen atom from the oxidized sample would have to be accommodated by a single V atom. It is therefore reasonable that thermodynamics would show evidence for only two phases. Wang et al.,8 using data from XRD and IR studies, have also suggested that reduction of Mg3(VO4)2 to Mg3V2O6 occurs with a direct reduction of V+5 to V+3 without an intermediate V+4 state. Furthermore, the structural data suggested that partially reduced samples of Mg3(VO4)2 were a physical mixture of Mg3(VO4)2 and Mg3V2O6, with cell parameters which varied slightly from the respective pure phases.
The differential oxidation enthalpies for V2O5 and Mg3(VO4)2, calculated from the temperature dependence of the isotherms and eq 2, are shown in Figure 4. Based on the isotherms, ∆H for oxidation of V2O3 to V2O4 is between -380 ( 10 kJ mol-1 O2 and for oxidation of V2O4 to V2O5 is -285 ( 20 kJ mol-1 O2. These values are reasonably close to the standard handbook values at 873 K, which are -360 and -270 kJ mol-1 O2, respectively.13 More interesting for this discussion is the comparison of the data for V2O5 with that of Mg3(VO4)2. With Mg3(VO4)2, the magnitude of the oxidation enthalpies increases with O:V ratio, from -370 ( 30 kJ mol-1 O2 at an O:V ratio of 1.5 to -460 ( 10 kJ mol-1 O2 at the point where Mg3(VO4)2 is completely oxidized. The data suggest that the initial oxygen is the most difficult to remove from Mg3(VO4)2, possibly due to a stabilizing effect of the structure. Reduction of a fraction of the V+5 ions to V+3 may destabilize the remaining V+5, causing a decrease in the magnitude of the enthalpy change. Whatever the mechanistic causes behind the differences between V2O5 and Mg3(VO4)2, it is very interesting to consider the influence that this has on the catalytic properties for partial oxidation. For reactions involving hydrocarbons that are relatively difficult to oxidize, most of the V in the catalyst will exist in a +5 oxidation state, given the presence of gas-phase O2. Under these conditions, V2O5 is able to release oxygen much more easily than Mg3(VO4)2 based on the enthalpies measured in this study. This would be expected to increase oxidation rates on V2O5 but at the expense of selectivity. That is indeed what has been observed in at least one study of propane oxidation.2 There may also be other cases where the relative ease with which V2O5 is reduced compared to Mg3(VO4)2 may cause the catalyst to exist in a partially reduced state under reaction conditions, so that the differences in the oxygen binding may be less obvious. Clearly, other factors besides oxygen binding play a role in partial oxidation reactions. The transition-metal cations in the oxide must be involved in binding intermediates and assisting in the making and breaking of C-H and C-C bonds in these reactions. Still, oxygen binding is critical and the type of thermodynamic information provided in this study is crucial for a complete understanding of these reactions. Conclusions A comparison of oxygen binding in V2O5 and Mg3(VO4)2 shows that the local environment has a strong influence on the oxidation enthalpies. While V2O5 can be sequentially reduced, with V+5 going to V+4 and V+4 to V+3, reduction of Mg3(VO4)2 appears to cause the direct conversion of V+5 to V+3. Starting from fully oxidized samples, oxygen binding is much stronger
Thermodynamic Redox Properties of V+5 on Mg3(VO4)2 and this is likely responsible for the increased selectivity observed for some partial oxidation reactions on this catalyst. Acknowledgment. This work was supported by the Department of Energy, Office of Basic Energy Sciences, Chemical Sciences, Geosciences and Biosciences Division, Grant DE-FG02-85ER13350. References and Notes (1) Wachs, I. E.; Deo, G.; Juskelis, M. V.; Weckhuysen, B. M. In Dyn. Surf. React. Kinet. Heterog. Catal. 1997, 109, 305-314. (2) Gao, X. T.; Ruiz, P.; Xin, Q.; Guo, X. X.; Delmon, B. J. Catal. 1994, 148, 56-67. (3) Gorte, R. J. Catal. Today 1996, 28, 405-414.
J. Phys. Chem. B, Vol. 111, No. 20, 2007 5683 (4) Tarama, K.; Teranishi, S.; Yoshida, S.; Tamura, N. Proc. Int. Congr. Catal. 3rd 1964, 282. (5) Yang, S. W.; Iglesia, E.; Bell, A. T. J. Phys. Chem. B 2005, 109, 8987-9000. (6) Parrillo, D. J.; Gorte, R. J. Thermochim. Acta 1998, 312, 125132. (7) Tretyakov, Y. D.; Maiorova, A. F.; Berezovskaya, Y. M. In Electr. Prop. Oxide Mater. 1997, 125, 283-316. (8) Wang, X. D.; Zhang, H.; Sinkler, W.; Poeppelmeier, K. R.; Marks, L. D. J. Alloys Compd. 1998, 270, 88-94. (9) Gao, X. T.; Ruiz, P.; Xin, Q.; Guo, X. X.; Delmon, B. Catal. Lett. 1994, 23, 321-337. (10) Shah, P. R.; Kim, T.; Zhou, G.; Fornasiero, P.; Gorte, R. J. Chem. Mater. 2006, 18, 5363-5369. (11) Yannopoulos, L. N. J. Phys. Chem. 1968, 72, 3293-3296. (12) In Handbook of Chemistry and Physics, 59 ed.; Weast, R. C., Ed.; CRC Press: Boca Raton, FL, 1979; pp D61-66. (13) In Handbook of Chemistry and Physics, 59 ed.; Weast, R. C., Ed.; CRC Press: Boca Raton, FL, 1979; pp D45-50.
