Langmuir 2004, 20, 1189-1193
1189
Novel Dual-Bed Reactors: Utilization of Hydrogen Spillover in Reactor Design John C. Weigle and Jonathan Phillips* Los Alamos National Laboratory, Engineering Sciences and Applications Division, P.O. Box 1663, MS C930, Los Alamos, New Mexico 87545 Received March 7, 2003. In Final Form: December 2, 2003 Hydrogen spillover over macroscopic distances was demonstrated and exploited in the design of two novel catalytic reactors for 1-butene isomerization. A dual-bed reactor containing separate zones of noble metal and bimetallic catalysts yielded activities up to 2.7 times greater than that of the noble metal alone. The noble metal catalyst contained palladium supported on graphitic carbon. The bimetallic catalyst contained a base metal, either iron or cobalt, and a lanthanide metal, either cerium or praseodymium, also supported on graphitic carbon. The bimetallic catalysts by themselves had no measurable activity at the current experimental conditions. Results from a dual-bed, dual-feed reactor using the same catalysts showed dramatic activity increases relative to controls. In this reactor, the hydrocarbon never contacted the noble metal catalyst, yet substantial hydrocarbon conversion was measured. No hydrocarbon conversion was detected when blank support replaced the bimetallic catalyst or when no material at all was placed above the noble metal catalyst. In both reactors, the activity increase was attributed to hydrogen spillover. That is, molecular hydrogen adsorbed and dissociated on the noble metal catalyst. The adsorbed atomic hydrogen was then transported via surface diffusion to the bimetallic catalyst, activating those sites. The results also demonstrated that a catalytic reaction may occur at distinctly different reactive sites and that catalysts may be selected to promote specific steps within the reaction.
Introduction The general objective of heterogeneous catalyst research is to develop catalysts with high activity, selectivity, and stability. This typically means identifying a single catalyst that provides the best combination of these three objectives. An alternative approach is to identify a combination of catalysts that interact to optimize performance. Indeed, the use of physical mixtures to create synergism is an old subject, reviewed by Weisz 40 years ago.1 Recently, Delmon described the “remote control theory”, the use of physical mixtures of different supported metals to achieve synergy in hydrodesulfurization processes.2 Also, Phillips and coworkers used mixtures of FeCe/Grafoil or FePr/Grafoil and either Pt/Grafoil or Pd/Grafoil to yield significant activity and selectivity synergy in the selective isomerization and hydrogenation of 1-butene and 1,3-butadiene.3,4 The observed synergy was attributed to diffusion of atomic hydrogen (hydrogen spillover) from the noble metal catalyst surface to the lanthanide-transition metal bimetallic catalyst surface. The ability of hydrogen spillover to activate catalytic surfaces suggested a novel approach to reactor design. In the present work, the impact of the arrangement of two different supported catalysts in packed bed reactors was tested. The reactors contained a noble metal, palladium, supported on graphitic carbon and a base metal-lanthanide metal pair, also supported on graphitic carbon. In contrast to all previous studies, the catalysts were not mixed; rather, they were present as distinct zones within each reactor. The test reaction was the isomerization of 1-butene to 2-butene. In some cases, more than a 2-fold activity increase over the noble metal alone was observed * To whom correspondence should be addressed. E-mail:
[email protected]. Phone: (505)665-2682. Fax: (505)665-5548. (1) Weisz, P. B. Adv. Catal. 1962, 13, 137-190. (2) Delmon, B. Catal. Lett. 1993, 22, 1-32. (3) Chang, H.; Phillips, J.; Heck, R. Langmuir 1996, 12, 2756-2761. (4) Chang, H.; Phillips, J. Langmuir 1997, 13, 477-482.
when a typical U-tube packed bed reactor was used with two distinct zones of catalyst. Additionally, the bed configuration had a dramatic impact on deactivation rate. The catalyst deactivated rapidly and continuously for at least 8 h when the noble metal was contacted first, but the activity stabilized after about 2 h when the bimetallic catalyst was contacted first. In both configurations, the activity of the dual-bed reactor was far higher than that of the two beds studied independently. This is particularly significant in the case in which the bimetallic catalyst is contacted first. This clearly demonstrates that H-atoms are diffusing upstream. Finally, a microreactor of novel design was built and tested. In this reactor, only hydrogen flowed over the noble metal catalyst; 1-butene was fed separately into the bimetallic portion of the bed. Significant 1-butene conversion occurred, providing direct evidence of hydrogen spillover from the noble metal zone to the bimetallic zone. These results provided significant insight into the nature of hydrogen spillover, and they indicated several ways that spillover can be exploited in reactor design. Experimental Section All catalysts were prepared via incipient wetness, using GTA grade Grafoil as a support. Grafoil is a high purity, highly graphitic carbon with a surface area of approximately 20 m2/g.5 Grafoil sheets were ground to produce a powder with a nominal average diameter of 0.5 mm. Ground Grafoil was then treated in flowing hydrogen for 8 h at 900 °C to remove sulfur impurities.6 The treated Grafoil was then impregnated with aqueous solutions of Pd(NO3)2‚xH2O, Fe(NO3)3‚9H2O, Pr(NO3)3‚6H2O, or Ce(NO3)3‚ 6H2O. The bimetallic catalysts were prepared via coimpregnation. The impregnated support was dried in air overnight, and then the salt was decomposed at 250 °C in flowing 5% hydrogen/ 95% nitrogen for 4 h. All catalysts had a nominal weight loading (5) Bretz, M.; Dash, J. G.; Hickernell, D. C.; McLean, E. O.; Vilches, O. E.; Phys. Rev. A 1973, 8, 1589-1615. (6) Wunder, R. W.; Phillips, J. J. Phys. Chem. 1996, 100, 1443014436.
10.1021/la034400m CCC: $27.50 © 2004 American Chemical Society Published on Web 01/20/2004
1190
Langmuir, Vol. 20, No. 4, 2004
Weigle and Phillips but were not allowed to mix. Hydrogen was fed upward through the noble metal catalyst and subsequently through the bimetallic catalyst. Helium and 1-butene were fed directly into the bimetallic catalyst. Two control studies were conducted to verify that 1-butene conversion was not occurring over the noble metal catalyst. In the first, noble metal catalyst was loaded according to the normal procedure, but no catalyst was loaded into the upper portion of the bed. In the second control study, blank support was loaded into the upper chamber. For either of these control studies, any conversion of 1-butene would indicate that the hydrocarbon was reacting on the noble metal catalyst. The activity and selectivity of a given catalyst bed were calculated based on the measured conversions of 1-butene to 2-butene (cis- and trans-) and butane. Since hydrogen is present in large excess and because 1-butene conversion was kept low, both reactions can be treated as constant volume. Thus, conversion and selectivity can be calculated directly from the mole fractions measured with the GC. Selectivity to 2-butene was defined as the fraction of 1-butene converted that forms 2-butenes; mathematically,
S)
out in out in (ycis-2-butene - ycis-2-butene ) + (ytrans-2-butene - ytrans-2-butene ) in out y1-butene - y1-butene
where yi represents the mole fraction of compound i in the specified stream. Activity was calculated based on the known 1-butene feed rate and its overall conversion in the reactor.
(P)(10 mL/min ) 1 A ) X1-butene M (RT)
Figure 1. Segregated bed reactors. (a) U-tube reactor and (b) dual-bed, dual-feed reactor. of 1% metal; the bimetallic catalysts contained equal atomic fractions of the two metals. Two atmospheric pressure reactors, Figure 1, were used to elucidate the role of hydrogen spillover in the isomerization of 1-butene. They could operate from 0 to 400 °C, but typical operating temperatures ranged from 0 to 40 °C. The first reactor, Figure 1a, was a simple U-tube. Prior to all activity measurements, the catalyst was reduced in flowing hydrogen at 300 °C for 4 h. The activity and selectivity of the catalyst were measured by flowing 500 mL/min ultrahigh purity He, 90 mL/min ultrahigh purity H2, and 10 mL/min 1-butene. Samples of the feed and product streams were injected into an HP 5890 Series II gas chromatograph (GC) equipped with a thermal conductivity detector and a 3 m packed column containing 0.19% picric acid on carbograph (Alltech). Response factors were from Dietz.7 After the noble metal was tested (2 mg Pd/Grafoil plus 18 mg Grafoil), bimetallic catalyst was added in 20-30 mg increments. The catalysts were in contact at the interface between the two zones, but they were not allowed to mix. After each addition of bimetallic catalyst, the dual bed was re-reduced and its activity determined as described above. Incremental additions of the bimetallic catalyst were continued until the total bed weight reached 90 mg. As a control study, increments of blank Grafoil were added instead of the bimetallic catalyst. Additionally, the bimetallic catalysts were tested without noble metal to verify their baseline activity and selectivity at the reaction temperatures. The second reactor was a novel, dual-bed, dual-feed reactor designed to validate the hydrogen spillover hypothesis and to serve as a prototype reactor in which the hydrocarbon stream never contacts the noble metal catalyst. This reactor contains two distinct sections, as shown in Figure 1b. The bottom portion of the reactor contained only Pd/Grafoil (40 mg), and the top portion contained only FeCe/Grafoil (up to 160 mg). As in the U-tube reactor, the two catalysts were in contact at the interface (7) Dietz, W. A. J. Gas. Chromatogr. 1967, 5, 68-71.
where X is the 1-butene conversion and M is the mass of noble metal in the bed. Activity was normalized to the amount of noble metal to provide a consistent basis for comparing the results. Since the amount of noble metal in the bed does not change with successive additions of bimetallic catalyst, expressing activity per gram of palladium is equivalent to the total conversion within the bed. However, it also allows comparison of the results when different amounts of bimetallic catalyst are present. This representation of the activity is preferable to calculating it in terms of total bed weight or residence time. In the absence of noble metal, neither bimetallic catalyst has any activity at these temperatures.8 Expressing the activity normalized to the amount of noble metal in the bed clearly demonstrates the synergism between the two catalysts and suggests that the bimetallic catalyst is indeed being activated.
Results The activity of the catalyst system (two zones) loaded into the U-tube reactor is shown in Figure 2; the synergy is most dramatically exhibited at 40 °C. The activity of the bed containing FeCe/Grafoil, Figure 2a, increases from 0.32 to 0.51 mol/min g-Pd at 40 °C. The activity increases to 0.85 mol/min g-Pd at 40 °C for the bed containing FePr/ Grafoil. The first data point shown, at 20 mg reactor bed weight, contained 2.1 mg Pd/Grafoil and 17.9 mg blank Grafoil. Thus, it represents a baseline activity and selectivity for Pd/Grafoil. The activity increased with each increment of bimetallic catalyst. This is particularly significant because the bimetallic catalysts by themselves have no activity at these temperatures. In the control studies, when blank Grafoil was added instead of bimetallic catalyst, no activity increase was measured. Also, no synergy was observed if a gap existed between the two zones, supporting the supposition that any activating species moves via surface diffusion. The baseline selectivity for Pd/Grafoil was measured to be 68% at 40 °C, and it gradually increased to 75% as FeCe/Grafoil was added. The selectivity of FeCe/Grafoil (8) Lu, W. C.; Chang, H.; Phillips, J. J. Catal. 1994, 146, 608-612.
Reactor Design Using Hydrogen Spillover
Langmuir, Vol. 20, No. 4, 2004 1191
Figure 3. Impact of bed configuration on the deactivation rate in the U-tube reactor [[, reactants contact the bimetallic catalyst first; 9, reactants contact the noble metal catalyst first; 2, control (Grafoil replaces the bimetallic catalyst, and the reactants contact the blank support first)].
Figure 2. 1-Butene conversion in the U-tube reactor. (a) Activity increase as bimetallic catalyst is incrementally added to the reactor and (b) electivity toward 2-butene, cis- + trans-. [9, FeCe at 25 °C; 2, FeCe at 40 °C; [, FePr at 40 °C.]
by itself was measured to be approximately 75%.8 The selectivity when FePr/Grafoil is the second catalyst surprisingly decreases from 68% to 57% with successive additions of bimetallic catalyst. FePr/Grafoil by itself has a selectivity of 92%.8 A model, to be published elsewhere,9 suggests that surface hydrogen concentration decreases with increasing distance from the noble metal catalyst, causing the activity increase per increment to decrease with additional additions of bimetallic catalyst. This is consistent with experimental observations. Characterization work showed that the state of the bimetallic particles was independent of the presence or absence of noble metal during the reduction process.10 Not surprisingly, the palladium catalyst is fully reduced after pretreatment. The two bimetallic catalysts behave in a similar fashion after the pretreatment. The iron is primarily reduced to Fe0, while the lanthanide metals are only partially reduced to nonstoichimetric oxides. Additionally, transmission electron microscopy revealed that all individual particles on the bimetallic catalysts contain both metals as segregated phases, specifically, Fe0 and CeO2-δ. The apparent activation energy for 1-butene conversion over the composite beds is also seen to be lower than that of the bimetallic catalyst alone. The average activation energy for all the composite beds was calculated to be 7 kJ/mol, while the activation energy for Pd/Grafoil alone was calculated to be 8 kJ/mol. The similarity of these energies suggests that the same process, hydrogen dis(9) Weigle, J. C.; Phillips, J. AIChE J., accepted for publication. (10) Weigle, J. C. Ph.D. Thesis, Pennsylvania State University, University Park, PA, 1999.
sociation on Pd/Grafoil, controls the reaction over palladium and over the composite beds. Previous work showed that the activation energy of this reaction over FeCe/ Grafoil by itself was substantially higher, 42 kJ/mol.10 The higher activation energy on the bimetallic catalyst is consistent with the fact that neither metal readily dissociates hydrogen at near-ambient conditions. Figure 3 shows the impact of bed order on deactivation rate. Catalysts were always loaded in the same order: noble metal located at the bottom of the reactor, with bimetallic catalyst on top of it. The U-tube reactor could be installed such that the gases flowed through the bed downward, contacting the bimetallic catalyst first, or upward, contacting the noble metal catalyst first. Perhaps surprisingly, the order in which the gases contacted the bed dramatically influenced catalyst deactivation. For the reactor where the gases contact the noble metal catalyst prior to contacting the bimetallic catalyst, the activity undergoes an initial period (∼2 h) of rapid decline followed by a steady, almost linear decline in activity over the next several hours. When the reactant gases flow over bimetallic catalyst first, there is still an initial period of rapid deactivation. However, the longer-term deactivation occurs at a much slower rate. The contrast of the long-term deactivation rates suggests that a noble metal poison is present in the reactant stream and that the poison is being removed over the bimetallic catalyst prior to the reactant stream contacting the noble metal. Other deactivation mechanisms, for example, sintering, would not depend on the presence or absence of the bimetallic catalyst. 1,3-Butadiene is the most likely candidate as the poison. First, materials in these experiments were identical to those described in the Experimental Section with one difference: the 1-butene stream received from the vendor (Matheson) contained a low concentration of butadiene, approximately 4 ppm. No other contaminants were detected. Second, oxygen/air might be a contaminant, but this possibility can be eliminated. The system contained oxygen traps with visual indicators on the feed line to the reactor. Leakage downstream of the traps can also be discarded as a possible source. The deactivation experiments were repeated several times with comparable results, and it is unlikely that leakage would consistently depend only on the amount of bimetallic catalyst in the reactor. Butadiene and other diolefins are known poisons of noble metal catalysts. This small difference apparently led to the rapid deactivation of the dual catalyst bed reactor when the noble metal was contacted first. However, the
1192
Langmuir, Vol. 20, No. 4, 2004
Figure 4. 1-Butene conversion in the dual-feed reactor, 40 °C. (a) Activity increases dramatically (the point at 40 mg is with no FeCe/Gr in the reactor) upon the first addition of FeCe/ Grafoil to the reactor and (b) selectivity toward 2-butene, cis+ trans-, increases slightly as the amount of FeCe/Grafoil increases. Open squares are the control study, adding blank Grafoil to the upper section of the reactor.
synergistic effect was maintained and the rate of deactivation was dramatically reduced when the bimetallic bed was contacted first. The dual-bed, dual-feed reactor exhibits no activity when no material or when blank Grafoil is present above the noble metal catalyst. In contrast, the reactor exhibits high activity, approximately 0.25 mol/min g-Pd, when FeCe/ Grafoil is present, as seen in Figure 4. As with the U-tube reactor, an activity plateau occurs as the amount of FeCe/ Grafoil in the bed increases. Unlike the U-tube reactor, though, the selectivity of the reactor is low, 30-40%. The control studies, which show no activity with just noble metal catalyst in the reactor, make it clear that all the conversion occurs in the bimetallic portion of the bed. However, the noble metal is clearly activating the bimetallic catalyst, since the bimetallic catalyst by itself has no activity at these temperatures. Hydrogen spillover provides the best explanation for the activation. Discussion The present work represents a new approach to designing catalytic reactors. Instead of choosing a single catalyst that has the best combination of activity, selectivity, and stability, a combination of catalysts can be selected to yield better performance. This research continues recent work that demonstrated that hydrogen spillover greatly accelerated isomerization and selective hydrogenation of 1-butene and 1,3-butadiene.3,4 Those
Weigle and Phillips
studies demonstrated that atomic hydrogen generated at noble metal sites spilled over to and activated bimetallic sites toward isomerization. The activity of the catalyst mixtures was up to 4 times greater than the activity of the noble metal alone. The current study was designed to test the impact of hydrogen spillover in reactors with distinct zones of different catalytic materials, as opposed to the well-mixed beds previously studied. Several significant results were found. Synergy between the staged beds occurred during the isomerization/hydrogenation reactions, suggesting hydrogen spillover occurred over macroscopic lengths. It was also possible to use spillover to create a catalytic guard bed. That is, the bimetallic catalysts removed 1,3-butadiene before it contacted the noble metal catalyst. This effect reduced the long-term deactivation rate of the catalyst system. Finally, the fact that the observed synergism can be observed in the dual-bed reactor for either bed order eliminates the suggestion that the activating species is a gas-phase species. In contrast, the hypothesis that the activating species is a surface diffusing hydrogen atom is completely consistent with this observation. The results of the dual-bed, dual-feed reactor clearly demonstrate that the noble metal catalyst activates the bimetallic catalyst. The reactor design eliminates hydrocarbon contact with the noble metal catalyst. The only possibility for contact is if the hydrocarbon back-diffuses into the noble metal zone. The lack of activity measured during the control studies, when noble metal catalyst was present but bimetallic catalyst was not, indicates that significant back-diffusion does not occur. The results of this study further support the earlier postulate that spillover explains the activity and selectivity synergy observed during alkene isomerization (doublebond shift) over trimetallic catalysts and physical mixtures of catalysts.3,4,8 The polyfunctional model, the most likely alternative explanation for catalyst synergy, cannot explain the experimental observations. The polyfunctional model requires a stable intermediate between 1-butene and 2-butene, but no plausible intermediate can be proposed. Butane might be considered but can be quickly eliminated because of its thermodynamics. The standard Gibbs free energy for the hydrogenation of 1-butene is -86.9 kJ/mol, while the standard Gibbs free energy for the dehydrogenation of butane to cis- and trans-2-butene is 83.7 and 80.7 kJ/mol, respectively. Butane formation is overwhelmingly favored in all reactions, so it is unlikely to act as an intermediate. The second argument against the polyfunctional model and for surface diffusion of the activating species results from testing the U-tube reactor with the bimetallic catalyst located upstream of the noble metal catalyst. Adding bimetallic catalyst to the reactor clearly increases the overall activity of the bed, while adding blank support to the reactor does not. Since the bimetallic catalyst by itself has no activity at these temperatures, it is reasonable to say that the noble metal catalyst produces an intermediate or activating species. For an intermediate to form at a noble metal site and then react at a bimetallic site, the intermediate’s bulk diffusion rate would need to be comparable to its convective transport rate. The Peclet number for the reactor was calculated to be approximately 1000, indicating that bulk-phase diffusion is negligible compared to convection. In contrast, surface diffusion of an activating species is not impacted by considerations of diffusion resistance in the gas phase. Moreover, the polyfunctional model requires that the hydrocarbon
Reactor Design Using Hydrogen Spillover
contact two distinct catalyst sites. In the dual-bed, dualfeed reactor, the hydrocarbon contacts only the bimetallic catalyst. In contrast to the deficiencies of the polyfunctional model, the significant activity synergy is readily explained by hydrogen spillover. Many researchers have investigated hydrogen spillover, and various hypotheses exist concerning the state of the hydrogen present on the surfaces. The presence of H+, H•, H-, and vibrationally excited H2 have all been proposed.11-15 The generally accepted mechanisms of alkene isomerization and hydrogenation16-20 require atomic hydrogen. The hydrogen atom then adds to the alkene, creating a metastable intermediate. Two pathways are open at this point. A second hydrogen atom can add to the intermediate, forming butane, or a hydrogen atom can be extracted from the intermediate, forming 2-butene. Thus, H• is the most likely surface species. It is clear from all previous mechanisticstudies,16-20 including deuterium exchange studies, that atomic hydrogen is essential to the reaction. These hydrogen atoms are formed on the noble metal surfaces and then are transported through the bed via surface diffusion to the bimetallic surfaces. Further, recent inelastic neutron scattering experiments directly demonstrated the ability of atomic hydrogen to diffuse across carbon surfaces.21 (11) Roessner, F.; Roland, U. J. Mol. Catal. A 1996, 112, 401-412. (12) Roland, U.; Braunschweig, T.; Roessner, F. J. Mol. Catal. A 1997, 127, 61-84. (13) Baumgarten, E.; Niemeyer, I. React. Kinet. Catal. Lett. 2000, 70, 371-377. (14) Baumgarten, E.; Meyer, G. React. Kinet. Catal. Lett. 2000, 71, 325-333. (15) Baumgarten, E.; Maschke, L. Appl. Catal. A 2000, 202, 171177. (16) Wells, P. B.; Wilson, G. R. J. Catal. 1967, 9, 70-75. (17) MacNab, J. I.; Webb, G. J. Catal. 1968, 10, 19-26. (18) Mellor, S. D.; Wells, P. B. Trans. Faraday Soc. 1969, 65, 18731882. (19) Mellor, S. D.; Wells, P. B. Trans. Faraday Soc. 1969, 65, 18831890. (20) Burwell, R. L.; Schrage, K. J. Am. Chem. Soc. 1965, 87, 52535254.
Langmuir, Vol. 20, No. 4, 2004 1193
The deactivation rate was a strong function of bed configuration. The activity stabilizes after 2 h when the bimetallic catalyst is upstream of the noble metal catalyst, but it never stabilizes when the bed order is reversed. This result provides additional evidence of hydrogen spillover. Deactivation may be caused by diolefin impurities in the feed poisoning the noble metal surface. Another study4 demonstrated that the bimetallic catalyst selectively hydrogenates and isomerizes 1,3-butadiene to 2-butene. Thus, it can be hypothesized that spillover hydrogen activates the bimetallic catalyst such that it removes the butadiene impurity. When the bimetallic catalyst is located upstream of the noble metal, all the butadiene is removed before it reaches the noble metal catalyst. Thus, it cannot poison the noble metal surfaces. When the bimetallic catalyst is downstream of the noble metal catalyst, the butadiene is fed directly to the noble metal and the protective effect is eliminated. A similar protective effect during desulfurization is described by Delmon.22 In that study, spillover oxygen prevents deactivation caused by unwanted solid-state reactions. In summary, this research represents a new approach for using a combination of catalysts instead of using a single catalyst for the isomerization or hydrogenation of alkenes. It also demonstrates the ability of spillover species to migrate macroscopic distances, apparently by surface diffusion. The results suggest a new method for designing catalytic reactors in which noble metal catalysts are separated from the main catalyst bed. The function of the noble metal is simply to generate spillover species, and the reaction of interest occurs in a separate section of the reactor. Membrane reactors are one option for large-scale reactors capable of exploiting long-range spillover. LA034400M (21) Mitchell, P. C. H.; Ramirez-Cuesta, A. J.; Parker, S. F.; Tomkinson, J.; Thompsett, D. J. Phys. Chem. B 2003, 107, 6838-6845. (22) Delmon, B. In Catalyst Deactivation 1994; Delmon, B., Froment, G. F., Eds.; Elsevier: Amsterdam, 1994; pp 113-128.
