A Multistep Surface Mechanism for Ethane Oxidative Dehydrogenation on Pt- and Pt/Sn-Coated Monoliths

Transcription

1 Ind. Eng. Chem. Res. 2005, 44, A Multistep Surface Mechanism for Ethane Oxidative Dehydrogenation on Pt- and Pt/Sn-Coated Monoliths Francesco Donsì, Kenneth A. Williams, and Lanny D. Schmidt* Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, Minnesota A computational study of ethane oxidative dehydrogenation to ethylene on Pt- and Pt/Sn-coated monoliths is presented as an improvement to previous kinetic models in reproducing experimental findings over a wide range of feed conditions. The multistep surface mechanism containing 20 reversible reactions among 11 surface species is based on published reaction steps for hydrogen and methane oxidation combined with lumped steps for ethane surface chemistry and coupled with an established homogeneous mechanism to form the detailed chemistry model. Simulation results at 1 atm are in good agreement with experimental data obtained on Pt at variable C 2 H 6 / O 2 and C 2 H 6 /O 2 /H 2 ratios and predict experimentally observed phenomena such as ignition temperatures and homogeneous ethylene formation. The model is also used to predict Pt monolith performance over an industrially relevant range of space velocities ( h -1 ) and pressures (1-10 atm). Furthermore, the Pt mechanism is extended to a Pt/Sn catalyst by changing two parameters in the H and CO oxidation steps, and agreement with experiments is obtained with and without H 2 addition. * To whom correspondence should be addressed. Tel.: Fax: schmi001@umn.edu. Current address: Dipartimento Ingegneria Chimica, Università di Napoli Federico II, Piazzale Tecchio 80, I Napoli, Italy. 1. Introduction Ethane oxidative dehydrogenation (ODH) has been investigated as an alternative to ethane dehydrogenation for ethylene production. 1,2 However, none of the catalysts investigated for ethane ODH in conventional reactors are suitable and profitable for large-scale applications because of the low yields and coking that occur by decreasing feed dilution and the ethane-tooxygen ratio when shifting from laboratory scale to industrial scale. 3 Ethane ODH can be carried out in short contact time reactors ( 5 ms residence time) while obtaining ethane conversion as high as 80% and an ethylene selectivity of 70%. 4 Improved ethylene yield has been attained by tuning catalyst composition and feed conditions. 5-7 A Pt/Sn catalyst with hydrogen addition gives ethane conversion and ethylene selectivity comparable to those of the present industrial process. 7 Adding H 2 decreases C 2 H 6 conversion to 70% but increases C 2 H 4 selectivity to 85%. Originally, the process was believed to occur through purely heterogeneous chemistry on the basis of the experimental observation of large differences in synthesis gas yield among various noble metals during methane partial oxidation. 8 Accordingly, a heterogeneous model was proposed to describe ethane dehydrogenation on a Pt catalyst in 23 steps. 9 All chemistry was assumed to occur on the catalyst surface, and gas-phase reactions were ignored. However, later experimental results showed that significant homogeneous reactions occur at the high temperatures reached in the reactor ( 1000 C). Indeed, experimental data has suggested that only CO x and H 2 O are formed on the surface of a Pt catalyst at low temperatures; at higher temperatures ethylene yield observed at the exit of the reactor was attributed entirely to gas-phase reactions. Variable bed length experiments with effluent sampling downstream of the catalytic monolith showed that most of the ethylene is formed in the gas phase. 13 The finding that performance similar to that of noble metals could be obtained on a completely different catalyst (e.g., LaMnO 3 ) further excluded the direct involvement of the catalyst in ethylene formation. 14,15 Nevertheless, even if ethylene formation occurs entirely in the gas phase, the catalyst is extremely important in igniting and sustaining gas-phase reactions 16,17 and making the process feasible at short contact times. On the basis of experimental evidence of the role of gas-phase reactions, Huff and co-workers proposed a kinetic model for ethane ODH (Huff model), assuming that the catalyst s role was to oxidize a fraction of the fuel to CO x and H 2 O (heat released on the surface then drove the homogeneous formation of ethylene). 18 Although in good agreement with the experimental results without H 2 addition, the Huff model seems to fail in the case of H 2 addition and was based on the assumption of dissociative adsorption of ethane to C and H followed by oxidation of C and H. The rate constants of hydrogen and oxygen adsorption/desorption and ethane dissociative adsorption were considered adjustable parameters. Additionally, it was assumed that oxygen could adsorb noncompetitively on the catalytic surface while all other species adsorbed competitively in order to reduce computing time. Experimental and theoretical studies have shown that Pt step sites bind oxygen atoms more strongly than terrace sites, which leads to a much higher level of O 2 dissociative adsorption in steps than terraces In addition, experimental evidence of noncompetitive O 2 adsorption was reported in a study of CO oxidation on supported Pt particles. 25 Recently, Zerkle and co-workers proposed a detailed model for ethane oxidation on a Pt catalyst. 26 The ethane surface chemistry was based on the model of /ie CCC: $ American Chemical Society Published on Web 04/05/2005

2 3454 Ind. Eng. Chem. Res., Vol. 44, No. 10, 2005 Wolf et al. 27 for nonoxidative methane conversion on Pt and was combined with oxidative steps. This coupling led to a heterogeneous mechanism consisting of 19 surface species and 82 elementary reactions. The heterogeneous mechanism was implemented together with a detailed homogeneous mechanism in a full 2D model, since the authors claimed that mass and heat transfer are extremely significant factors in the short contact time reactor. In particular, reactant mass transfer from the bulk of the gas phase to the catalyst surface and transfer of heat from the surface to the gas phase were considered crucial points. The model predicted the amount of ethylene formed on the catalyst to be dependent on feed composition; ethylene was formed mainly on the surface when H 2 was added to the reacting mixture. These findings of purely heterogeneous ethylene formation are not in agreement with experimental findings In this work, the surface chemistry of the detailed kinetic model is based on well-established reaction rates for hydrogen and methane oxidation to CO x and H 2 O. These steps are combined with lumped steps for ethane surface adsorption and dissociation to obtain a simple and flexible mechanism that can be used to predict process performance over a wide range of experimental conditions. Creation of the mechanism was motivated by experimental evidence for gas-phase ethylene formation. 11,13 Additional effort is devoted to predict the catalyst s performance outside the temperature range of interest for ODH and in particular the catalyst ignition temperature and product distribution before the threshold temperature for gas-phase reactions is reached. 2. Computational Methods 2.1. Reactor Models. As a first approximation, the system was modeled as a plug-flow reactor. The plugflow assumption is an extremely useful tool for building a kinetic mechanism considering the tradeoff between the extremely low computing time (seconds) and the accuracy of the model. A computational study of CH 4 surface oxidation in short contact time reactors showed that the predicted differences between a plug-flow and a complete 2D model are limited to entrance effects for honeycomb monoliths. 28 It must be emphasized that during CH 4 oxidation under atmospheric pressure homogeneous reactions account for less than 10% of fuel consumption. In contrast, during ethane ODH probably only 10% of the fuel is consumed on the surface, and the remaining ethane undergoes gas-phase reactions. Hence the validity of the assumptions of infinitely fast heat and mass transfer is critical only in a short fraction of the modeled reactor length. Furthermore, the ideal Reynolds number (<100) and the very high length-todiameter ratio ( 20 for 45 pore per linear inch (ppi) monolith, d c ) 0.42 mm, and L ) 10 mm) suggest that axial dispersion is extremely low. In reality, foam monoliths unlike honeycomb monoliths possess pore tortuosity that enhances turbulence and radial transport that may limit boundary layer development. 29,30 These considerations led to the conclusion that the plug-flow model was valid for initial estimates of the mechanism. The foam monolith reactor was described through an idealized single pore model 18,26,31 (laminar flow) since surface area per volume is the major parameter in determining reactor performance. Final tuning of the surface mechanism should be carried out with a model that takes into account a detailed flow description (e.g., complete 2D flow model); however, the foam monolith geometry is ill-defined for a rigorous computational fluid dynamics study. The mass and energy balance equations for a plugflow reactor can be written respectively as K g dw i Fv z A( h i gas dw K g i Fv z A dz + w i a i s iw i ) W i (s ia i + ω ia) (1) gas dz + c dt dv z p dz + v z dz ) + K g h i w i + 1 ( gas 2 v z 2) a i K g s iw i ) a e Q e (2) Numerical simulations were performed using the Chemkin PLUG code 32 for a single pore of a 45 ppi, 92 wt % Al 2 O 3 foam monolith (d c ) 0.42 mm 33 ). Even though the catalytic monolith was 1 cm long, the total length of the modeled pore was 3 cm to capture the effects of any gas-phase chemistry occurring after the catalyst. The gas hourly space velocity was h -1, corresponding to a total flow rate of 5 standard liters per minute (slpm) for an entire monolith diameter of 17 mm. The inlet pressure was always 1.2 atm, dilution was 30 vol % N 2, and the C 2 H 6 /O 2 ratio was varied between 1.5 and 2. When added to C 2 H 6 as secondary fuel, H 2 was added in appropriate amounts holding other flow rates constant so that the total flow rate increased and N 2 dilution decreased. Simulation input conditions (temperature and inlet species compositions) were taken from previous experimental studies, 5,7,34 with which numerical simulations are compared. Entrance effects in the catalytic monolith were studied using a boundary-layer model with the Chemkin code CRESLAF. 32 The mass, momentum, species, and energy conservation equations for boundary-layer flow in a cylindrical channel can be written respectively as v z Fv z z +Fv v z r and the applicable components of mass flux vector J are given in eq 7: The contribution of thermal diffusion to species diffusion was considered but found negligible. Auxiliary conditions were as follows and are described in detail elsewhere: 28 gas Fv z z + 1 r Fv r ) 0 (3) r r r )- P z + 1 r w i Fv z z +Fv w i r r ) ω i W i + 1 r J i,r r r Fc p( v T z z + v T r r) ) 1 r r( r) λr T - i c p,i J i,r T J i,r )-F W i Wh D Y i i,m r r(µr v z r ) (4) r - i (5) h i ω iw i (6) (7)

3 Ind. Eng. Chem. Res., Vol. 44, No. 10, z ) 0, r: v z ) V 0, v r ) 0, T ) T in, w i ) w i,in (8) z, r ) z, r ) 0: d c v z r ) v r ) T r ) w i r ) 0 (9) 2 : v z ) v r ) 0, J i,r +Fw i v st ) s i W i, 1 r r( r) λr T )- h i s iw i (10) Results along the radius at various axial locations were integrated according to eq 11 to calculate mass-average species and temperature profiles: d c /2 Fvz φr dr 0 φ av ) d c /2 Fvz 0 r dr 2.2. Energy Balance. Temperature measurements on experimental reactors 5,7,34 show that the system is not adiabatic. However, use of a heat transfer coefficient along a single pore was avoided, since intrapore heat transfer is a complex function of radial and axial distance in the foam. Consequently, heat losses from the reactor were lumped and estimated using experimental data. Since the heat loss term is coupled with the plug-flow model, the heat transfer resistance is implicitly assumed to be concentrated on the external side of the reactor. Considering the entire monolith as the system boundary, heat losses were calculated by comparing the measured gas temperature at the exit of the monolith to the adiabatic outlet temperature calculated on the basis of measured conversion and product distribution: Reactor heat loss varied with the inlet mixture composition. For various i feed mixtures used in ethane ODH experiments, 34 the overall heat transfer coefficient was evaluated using eq 13 with an external temperature of 25 C: The heat transfer coefficient U i was always in the range of J/(s m 2 K) for any i. An average value of 400 J/(s m 2 K) (U) was assumed and kept constant for all calculations. To determine the average heat flux from a single channel, the adiabatic temperature profile along the axial coordinate z was obtained from the numerical simulation of an adiabatic reactor, and heat flux from the reactor was calculated as a function of z: The heat flux, q(z), which applies to the entire monolith, was then divided by the product of the monolith s specific geometric surface area ( 8000 m 2 /m 3 based on the volume of the entire monolith 29 ) and length (0.01 m) in order to refer the heat flux to the surface area of a single hypothetical pore as a function of the axial coordinate. This function was implemented in the plugflow model energy balance as a heat loss term. i (11) Q i )Fv z c p A(T adiab - T meas ) (12) U i ) Q i A e (T adiab,i - T ext ) (13) q(z) ) U(T adiab (z) - T ext ) (14) 2.3. Heterogeneous Mechanism. The multistep surface reaction mechanism for ethane ODH on Pt contains 20 reversible reactions among 11 surface species. Hydrogen surface chemistry on Pt has been previously described The Chou mechanism, 38 which can predict the ignition of a H 2 /O 2 mixture in the low temperature range, was employed with only minor modifications designed to make the mechanism enthalpically consistent. Carbon chemistry for methane combustion and partial oxidation has been discussed by several authors The mechanism proposed by Veser 39 was chosen for simplicity since CH 4 decomposition to C and H is lumped in a single step. Once C is formed on the surface, it is oxidized to CO that either desorbs into the gas phase or is further oxidized to CO 2. The mechanism resulting from the combination of hydrogen and methane oxidation steps was coupled with ethane chemistry. As discussed, none of the available ethane ODH surface mechanisms 9,18,26 appear to suitably predict experimentally observed ethane ODH phenomena with and without H 2 addition. In this work, noncompetitive oxygen adsorption on Pt step sites (X*, assumed site density ) mol/cm 2 ) was preserved from the Huff model, 18 and the X* site density was treated as a fitting parameter to describe ignition; the competitive adsorption of all the other species occurs on Pt terrace sites (S*, site density ) mol/ cm 2 ). 26 The 78/22 terrace to step site ratio is in good agreement with experimental findings from temperature programmed desorption spectra suggesting an 80/ 20 ratio for a polycrystalline Pt foil. 23 However, the Huff mechanism 18 does not take into account a surface route for ethylene formation. Even though ethylene formation primarily through gas-phase chemistry was confirmed by experiments, 10,11,13,16,17 a heterogeneous pathway for ethylene formation was included in the present mechanism by incorporating steps that were conceived to produce ethylene purely through surface chemistry. 9 Ethane decomposition and oxidation were described through a lumped mechanism (Table 1). Following the energetic schemes of Figure 1, ethane can adsorb on 2 vacant sites (R5) and then dehydrogenate to ethyl through a straight (R25) or an oxygen-assisted (R27) dissociation. The oxygen-assisted step (R27) is important in describing ignition at the correct temperature, while at higher temperatures ( C) the straight dehydrogenation step R25 becomes important. Adsorbed ethyl radical can either decompose to C and H via a lumped exothermic step (R29) or undergo further endothermic dehydrogenation leading to C 2 H 4 formation (R31). C 2 H 4 can then either desorb (R15) or undergo C-C bond cleavage to C and CH 4 (R33). Ethylene adsorption (R6) was described by a sticking coefficient (0.015) significantly higher than that for ethane adsorption (0.003) in agreement with experimental observations. 34 Methane can adsorb on the surface (R7) and then undergo decomposition to C and H (R35). 39 Reaction R36 was added to account for surface methane formation from C and H. Experimental evidence for methane formation was found during synthesis gas oxidation on noble metal catalysts. 44,45 Oxygen adsorbs dissociatively on Pt (R1) and is mainly involved in H oxidation to OH (R19) and C oxidation to CO (R37) and CO 2 (R39) in addition to ethane ODH (R27). The adsorption of H 2 (R2) was considered first order in Pt. 26 When present in the feed, hydrogen is adsorbed and readily oxidized to H 2 Oina path competitive with CO x formation but favored at

4 3456 Ind. Eng. Chem. Res., Vol. 44, No. 10, 2005 Table 1. Surface Reactions for Ethane Oxidative Dehydrogenation on a Pt Monolith a reaction S 0 (cm, mol, s) k 0 lower temperature (R19 and R21). The adsorption and desorption of OH (R3 and R12 as proposed by Chou and co-workers 38 ) were also included, even though these steps do not have a determining influence on reaction progression under the conditions investigated. Hydrogen and carbon chemistry of methane down to C and H was not adjusted further after enthalpic consistency was imposed. Note that in this heterogeneous scheme the adsorption/desorption of radicals between the catalyst surface and the gas phase was neglected as well as the reaction of gas-phase intermediates expected in low concentration, such as acetylene. In the future development of the mechanism and the extension of its validity to other reaction processes and conditions, these steps need to be implemented. Consequently, the main adjustable mechanism parameters correspond to a few reactions. The sticking coefficient for the ethane adsorption step (described above) was reduced from proposed by Zerkle 26 to to predict ignition properly and not changed further to fit the experimental product distributions. Additionally, the preexponential factor of reaction R27 (oxygen-assisted ethane dehydrogenation) was adjusted in order to describe ignition at the experimentally E a (kj/mol) adsorption R1 O 2(g) + 2X* f 2O* Chou et al., 2000 R2 H 2(g) + 2S* f 2H* Chou et al., 2000 c R3 OH(g) + S* f OH* Chou et al., 2000 R4 H 2O(g) + S* f H 2O* Chou et al., 2000 R5 C 2H 6(g) + 2S* f C 2H 6* this work R6 C 2H 4(g) + S* f C 2H 4* Zerkle et al., 2000 R7 CH 4(g) + S* f CH 4* this work R8 CO(g) + S* f CO* Veser et al., 2000 R9 CO 2(g) + S* f CO 2* Veser et al., 2000 desorption R10 2O* f O 2(g) + 2X* Chou et al., 2000 R11 2H* f H 2(g) + 2S* Chou et al., 2000 R12 OH* f OH(g) + S* Chou et al., 2000 R13 H 2O* f H 2O(g) + S* Chou et al., 2000 b R14 C 2H 6* f C 2H 6(g) + 2S* Zerkle et al., 2000 R15 C 2H 4* f C 2H 4(g) + S* Zerkle et al., 2000 R16 CH 4*f CH 4(g) + S* Veser et al., 2000 R17 CO* f CO(g) + S* Veser et al., 2000 R18 CO 2* f CO 2(g) + S* Veser et al., 2000 surface reactions R19 H* + O* f OH* + X* Bond et al., 1996 R20 OH* + X* f H* + O* θ 0 Chou et al., Zerkle et al., 2000 b,e R21 H* + OH* f H 2O* + S* Chou et al., 2000 R22 2OH* + X*f H 2O* + O* + S* Bond et al., 1996 f R23 H 2O* + S* f OH* + H* Chou et al., 2000 b,e R24 H 2O* + O* +S* f 2OH* + X* Zerkle et al., 2000 b,e,g R25 C 2H 6* f C 2H 5* + H* Wolf et al., 1999 R26 C 2H 5* + H* f C 2H 6* Wolf et al., 1999 R27 C 2H 6* + O* f C 2H 5* + OH* + X* Zerkle et al., 2000 e R28 C 2H 5* + OH* + X* f C 2H 6* + O* Zerkle et al., 2000 b,f R29 C 2H 5* + 6S* f 2C* + 5H* this work c R30 2C* + 5H* f C 2H 5* + 6S* this work d R31 C 2H 5* + S* f C 2H 4* + H* Huff and Schmidt, 1996 e R32 C 2H 4* + H* f C 2H 5* + S* Zerkle et al., 2000 b,e R33 C 2H 4* + S* f CH 4* + C* Zerkle et al., 2000 b,e R34 CH 4* + C* f C 2H 4* + S* Zerkle et al., 2000 b,e R35 CH 4* + 4S* f C* + 4H* Veser et al., 2000 b,c,e R36 C* + 4H* f CH 4* + 4S* this work d R37 C* + O* f CO* + X* Chou et al., 2000 e R38 CO* + X* f C* + O* Veser et al., 2000 e R39 CO* + O* f CO 2* + X* Veser et al., 2000 e R40 CO 2* + X* f CO* + O* Veser et al., 2000 b,e a X* site density ) mol cm -2 ; S* site density ) mol cm -2. Preexponential units are s -1 for first-order reactions and cm 2 mol -1 s -1 for second-order reactions. b E a modified in order to match enthalpic constraints at 298 K. c First order in S*. d First order in C and H. e Preexponential factor changed in order to fit experimental data. f Zero order in X*. g Zero order in S*. measured temperature. Selected parameters for reactions R29 through R34 (3 reversible reactions leading to C and H formation on the surface) were tuned in order to fit the experimental data and maintain the mechanism s enthalpic consistency (See Table 1). In the kinetic mechanism the contribution of reactions R31 and R33 to ethyl decomposition via ethylene and methane is small. The lumped step R29 accounting for a parallel path of C and H formation from ethyl is very fast. Nevertheless, R31 and R33 were included for flexibility in order to extend the mechanism to other catalysts where the catalytic step for ethylene formation may be important. Reaction R19 (H oxidation to OH) was found to be the most relevant in determining system performance and was adjusted to describe the experimental data on a different catalyst (Pt/Sn). The enthalpic consistency of the surface mechanism at 298 K is shown in Figure 1 where the difference in enthalpy of the gas species is preserved through the formation of surface species. Ethylene and water formation from ethane and oxygen (ODH) is exothermic ( kj/mol) and can occur through the formation of surface species of decreasing energy from ethane to ethyl (righthand branch of Figure 1a) in addition to endothermic ref

5 Ind. Eng. Chem. Res., Vol. 44, No. 10, Figure 1. Energy level diagrams (not drawn to scale, kj/mol) of surface reactions at 298 K (shown in Table 1) for reaction paths leading to (a) CH 4/C 2H 4 and (b) CO x. ethane dehydrogenation. Ethylene formation via nonoxidadative dehydrogenation ( kj/mol) and ethylene dehydrogenation to acetylene ( kj/mol) are endothermic reactions that can occur only at high temperature. Methane can be formed from ethane and hydrogen via an exothermic path (Figure 1a). In comparison to ODH, synthesis gas formation produces similar amounts of energy ( kj/mol). Total oxidation of ethane to CO 2 and H 2 O is highly exothermic ( kj/mol) and can also occur on the catalyst surface (Figure 1b). In order to produce heat on the catalyst it is desirable to oxidize a small amount of ethane to CO 2 and H 2 O to drive ODH in the gas phase Homogeneous Mechanism. Several mechanisms for gas-phase hydrocarbon oxidation were examined for ethane ODH. The GRI mechanism 41 was not adequate for the conditions investigated, since it was derived for oxygen-rich mixtures. The mechanism proposed by Mims and co-workers 42 derived for oxidative coupling of methane and the Marinov mechanism 43 for hydrodrocarbons up to C 4 in oxygen-poor conditions were substantially similar. Numerical simulations in this work reflect the homogeneous mechanism proposed by Mims and co-workers 42 without adjustment. Note that the parameters used in the surface reaction steps are indirectly dependent on the gas-phase chemistry; therefore, any errors in the homogeneous mechanism inevitably affect the optimized kinetic parameters for the surface chemistry. 3. Results 3.1. Surface Mechanism Sensitivity Analysis. To tune the mechanism, a sensitivity analysis was per-

6 3458 Ind. Eng. Chem. Res., Vol. 44, No. 10, 2005 Figure 2. Sensitivity coefficients for (a) C 2H 6 conversion, (b) C 2H 4 selectivity, and (c) CO selectivity for a C 2H 6/O 2 ) 2/1 mixture (black bars) and for a C 2H 6/O 2/H 2 ) 2/1/2 mixture (gray bars). formed on the surface reactions for both C 2 H 6 /O 2 and C 2 H 6 /O 2 /H 2 feed mixtures. Reactions with the largest influence on C 2 H 6 conversion and C 2 H 4 and CO selectivities were identified and adjusted in order to fit experimental data. 34 The sensitivity coefficient (S) is defined as the ratio of the relative increase of the quantity analyzed (x) by the relative increase of the preexponential factor (k 0 ) for the reaction of interest: S ) ( k 0 x )( x k 0 ) (15) Figure 2 shows sensitivity coefficients of the most critical reactions for C 2 H 6 conversion (Figure 2a), C 2 H 4 (Figure 2b), and CO selectivity (Figure 2c) for a C 2 H 6 / O 2 ) 2/1 mixture and a C 2 H 6 /O 2 /H 2 ) 2/1/2 mixture. Reaction parameters considered in the sensitivity analysis were typically varied by 50% of their original value. Although reactions R1 (O 2 adsorption), R2 (H 2 adsorption), and R11 (H 2 desorption) show very high sensitivity, the corresponding sticking coefficients/preexponential were not changed from values reported for Pt that fit H 2 chemistry well. 38 The sticking coefficient of reaction R5 (C 2 H 6 adsorption) was changed from to 0.003, a value capable of describing experimental ignition temperatures satisfactorily. Even though Table 2. Modified Surface Reaction Steps for Ethane Oxidative Dehydrogenation on a Pt/Sn Monolith reaction k 0 (J, mol, s) E a (kj/mol) surface reactions R19 H* + O* f OH* + X* this work R39 CO* + O* f CO 2* + X* this work reactions R5 (only adjusted to describe experimental ignition temperatures), R6, and the corresponding desorption reactions (R14 and R15) displayed large sensitivity, they were not adjusted to fit experimental conversion/selectivity data. CO adsorption and desorption reactions (R8 and R17) also show large sensitivities, especially in the absence of H 2 in the feed, but were not adjusted since their rates seem to be well established. C 2 H 4 and CO selectivities are highly sensitive to the oxidation reaction of H to OH (R19). Nevertheless, once it was verified that the values proposed by Bond 36 for the preexponential factor and activation energy were satisfactorily accurate, this reaction was not further adjusted to fit the data, but was instead considered as the key parameter in the extension of the mechanism to the Pt/Sn catalyst (see Table 2). In contrast, other surface reactions involved in hydrogen oxidation, such as R21 and R22, exhibit very low sensitivity and were not modified. ref

7 Ind. Eng. Chem. Res., Vol. 44, No. 10, Figure 3. Simulation results (lines) and experimental data (black symbols, 34 white symbols 5 )forac 2H 6/O 2 mixture on a Pt catalyst (5 slpm feed flow rate and 30 vol % N 2 dilution): C 2H 6 conversion and C-atom selectivity for C 2H 4 (a), C-atom selectivity for CO and CO 2 (b), C-atom selectivity for CH 4 and C 2H 2 (c), and H-atom selectivity (d) at varying C 2H 6/O 2 ratios. Both ethane to ethyl dehydrogenation reactions (R25 and R27) have low sensitivity because they are fast in comparison to other reactions. Parameters for reaction R25 were not adjusted, while the preexponential factor of R27 was adjusted to attain ignition at the experimentally verified temperature. Also, reaction R29 (C 2 H 5 decomposition to C and H) has a very fast reaction rate and consequently a very low sensitivity (Figure 2) and was not adjusted to tune the mechanism. C 2 H 5 decomposes mainly through this route (R29). Reactions R31 through R34 (reversible steps of C 2 H 4 and CH 4 formation from C 2 H 5 ) show a low sensitivity coefficient, but were adjusted from reported values 9,26 in order to suppress ethylene and methane formation on the catalyst surface at low temperatures as suggested by experimental findings. 10,11 Thus, ethyl decomposition through these reactions is only a minor path in comparison with R29. Sensitivity analysis shows a large sensitivity for reaction R36, especially for H 2 addition, when large amounts of H are adsorbed on the catalyst surface. R36 parameters were adjusted in order to fit the observed methane selectivity under various experimental conditions. Ethylene selectivity is very sensitive to reaction R36, since methane formation competes with C 2 H 4 formation. Conversion and selectivities are very sensitive to reaction R37 (C oxidation to CO). Its preexponential factor was adjusted from the value proposed by Veser and Frauhammer 39 in order to fit experimental data. The oxidation of CO to CO 2 (R39) has a low sensitivity and was changed only when switching from Pt to Pt/Sn to gain a better fit of the experimental data (see Table 2) Ethane ODH on Pt. The developed mechanism was compared to experimental data with and without H 2 addition. Although the experimental findings of Henning 34 were used to tune the mechanism, additional literature data sets 5,7 are included in applicable figures for comparison. For a C 2 H 6 /O 2 feed, the experimental temperature measured at the front face of the monolith was 350 C with minor oscillations ((50 C), while it was 100 C lower (250 C) for H 2 addition (C 2 H 6 /O 2 /H 2 feed). These temperatures were used as inputs in the respective simulations. ForaC 2 H 6 /O 2 feed, model simulations agree well with experimental data for a C 2 H 6 /O 2 feed ratio from 1.5 to 2 (Figure 3), which are typical operating conditions for ethylene production. In particular, the trend of experimental ethane conversion and ethylene selectivity are quantitatively reproduced by the model for all C 2 H 6 /O 2 ratios considered (Figure 3a). Experimental CO and CO 2 selectivities are well reproduced by the model (Figure 3b). However, from a qualitative standpoint, the formation of CO x species predicted by the mechanism is less sensitive to the mixture s oxygen content than that shown by experiments. CH 4 experimental selectivity is qualitatively reproduced by the model (Figure 3c) with a significant contribution from surface reaction, where CH 4 is desorbed after recombination of adsorbed C and H atoms. In contrast, acetylene is produced exclusively in the gas phase, and its computed selectivity is in good agreement with the experimental data (Figure 3c), as are H 2 and H 2 O selectivities (Figure 3d). H 2 O is mainly produced on the catalytic surface due to fuel oxidation at the reactor inlet, while H 2 is a substantial product of gas-phase reactions. Oxygen conversion (not shown) was always complete both in the simulations and in the experimental results. For H 2 addition, agreement of the simulations with experimental data remains good (Figure 4). The H 2 /O 2 ratio was varied in the same range as in experiments, between 0 (corresponding to C 2 H 6 /O 2 ) 2 of Figure 3) and 3. Experiments and simulations exhibit the same

8 3460 Ind. Eng. Chem. Res., Vol. 44, No. 10, 2005 Figure 4. Simulation results (lines) and experimental data (black symbols, 34 white symbols 7 ) for a Pt catalyst with H 2 addition (C 2H 6/O 2 ) 2/1, variable feed flow rate and N 2 dilution): ethane conversion and C-atom selectivity for C 2H 4 (a), C-atom selectivity for CO and CO 2, C-atom selectivity for CH 4 and C 2H 2 (c), and moles of H 2 formed per mole of O 2 fed (d) at varying H 2/O 2 ratios. trend with an increase in C 2 H 4 selectivity and a corresponding decrease in C 2 H 6 conversion with increasing H 2 /O 2 ratio (Figure 4a). The trend of CO selectivity is overpredicted and CO 2 selectivity is underpredicted (Figure 4b). Consequently, the model predicts the formation of larger amounts of water (not shown) and lower amounts of CH 4 than experiments show. Simulated methane selectivity (Figure 4c) shows an increasing trend with H 2 /O 2 ratio because the increased H surface coverage increases the rate of R36. C 2 H 2 is constant and well predicted by the gas-phase mechanism. Simulated H 2 formation lies between experimental values (Figure 4d) Ethane ODH on Pt/Sn. In order to test the surface mechanism s ability to predict ethane decomposition on a Pt/Sn catalyst, two steps in the surface mechanism (R19 and R39) were modified to fit experimental data (Table 2). Experimental observations suggest that higher ethylene selectivity occurs on Pt/Sn than on Pt because less C and more H are sacrificed to oxidation products; consequently, CO x selectivity is globally lower and H 2 O selectivity is larger on Pt/Sn. 5,7,34 To predict this behavior, one major change was made to the Pt mechanism by increasing the preexponential factor for reaction R19 (H surface oxidation) by 1 order of magnitude. Additionally, reaction R39 was modified (the preexponential factor was reduced by 2 orders of magnitude from its original value); however, its very low sensitivity prohibits this change from having a significant effect on product distribution and conversion. With only minor modifications to the Pt mechanism, which were dictated by the phenomenological nature of the Pt/Sn catalyst, a very good fit of the experimental data is obtained (Figure 5). The agreement of simulated C 2 H 6 conversion and C 2 H 4 selectivity (Figure 5a) with experimental data is excellent. In addition, the computed distribution of the other products for various C 2 H 6 /O 2 ratios is in agreement with the experimental data (Figure 5). For H 2 addition to the feed, when H oxidation on the catalyst surface becomes largely important, good agreement between simulation and experimental data is obtained. Simulation results are in agreement with experimentally measured ethane conversion and product distributions for H 2 /O 2 ratios between 0 and 3 (Figure 6). In particular, upon addition of H 2 (H 2 /O 2 ) 1) a significant system response is observed, with an increase in C 2 H 4 selectivity of around 10% and a concurrent drop in CO and CO 2 selectivities, even though ethane conversion remains constant. For further addition of H 2 (H 2 /O 2 > 1), the effect on ethylene selectivity is less significant with only a moderate increase, while ethane conversion starts to drop. These features are well predicted by the model (Figure 6a,b). Simulated methane and acetylene selectivities (Figure 6c) show similar trends to the experimental data, with CH 4 increasing and C 2 H 2 constant with increasing H 2 / O 2. Simulated H 2 formation lies between experimental values (Figure 6d) CH 4 and H 2 Chemistry on Pt. The present mechanism was developed in part from an assembly of different surface mechanisms developed for H 2 /O 2 and CH 4 /O 2 chemistry. Since the mechanism was tuned on the basis of ethane ODH experiments, the performance of CH 4 and H 2 chemistry subsets was also compared to methane partial oxidation experiments to ensure that the carbon and hydrogen chemistry was still working properly. Specifically, the model was compared to experimental data obtained in short contact time reactors with the same reactor morphology (45 ppi ceramic foam monoliths), same catalyst (Pt), and lower flow rate (4 slpm

9 Ind. Eng. Chem. Res., Vol. 44, No. 10, Figure 5. Simulation results (lines) and experimental data (black symbols, 34 white symbols 5 )forac 2H 6/O 2 mixture on a Pt/Sn catalyst (5 slpm feed flow rate and 30 vol % N 2 dilution): ethane conversion and C-atom selectivity for C 2H 4 (a), C-atom selectivity for CO and CO 2 (b), C-atom selectivity for CH 4 and C 2H 2 (c), and H-atom selectivity (d) at varying C 2H 6/O 2 ratios. Figure 6. Simulation results (lines) and experimental data (black symbols, 34 white symbols 7 ) for a Pt/Sn catalyst with H 2 addition (C 2H 6/O 2 ) 2/1, variable feed flow rate and N 2 dilution): ethane conversion and C-atom selectivity for C 2H 4 (a), C-atom selectivity for CO and CO 2 (b), C-atom selectivity for CH 4 and C 2H 2 (c), and moles of H 2 formed per mole of O 2 fed (d) at varying H 2/O 2 ratios. instead of 5 slpm). 8,46 The front face catalyst temperature and the heat losses from the experimental reactor were not measured in experiments and were estimated instead as fitting parameters. An inlet temperature of 350 C and heat losses 25% higher than in the ethane ODH reactor were assumed. ForaCH 4 /O 2 feed ranging from 1.6 to 2, which are typical conditions for syngas production, model simula-

10 3462 Ind. Eng. Chem. Res., Vol. 44, No. 10, 2005 Figure 7. Simulation results (lines) and experimental data (black symbols, 46 white symbols 8 )forach 4/O 2 mixture on a Pt catalyst (4 slpm feed flow rate and 10% N 2 dilution): methane conversion (a) and C-atom selectivity for CO and H-atom selectivity for H 2 and H 2O (b) at varying CH 4/O 2 ratios. tions agree well with experimental data (Figure 7). Methane conversion is well reproduced, and predicted CO selectivity agrees well with experimental data. Predicted H 2 selectivity is lower than the experimental values, especially at a low CH 4 /O 2 ratio, due to the onset of gas-phase reactions partially consuming CH 4 in radical reactions leading to coupling products Temperature and Species Profiles for Pt. In this section, the computed profiles of temperature and major species conversion and yield are reported along the axial coordinate for the Pt heterogeneous mechanism together with the heterogeneous and homogeneous reaction rates for a feed with and without H 2 addition. In order to have comparable values between heterogeneous and homogeneous reaction rates, surface reaction rates were converted from an area basis [mol/(cm 2 s)] to a volume basis [mol/(cm 3 s)] by using the monolith s specific geometric surface area ( 8000 m 2 /m 3 based on the volume of the entire monolith 29 ) C 2 H 6 /O 2 Feed Mixtures. For a feed with a C 2 H 6 /O 2 ratio of 2, the computed profiles of temperature, ethane and oxygen conversion, and the carbon and hydrogen yields of the main products are shown in Figure 8. The net rates of both gas-phase and surface chemistry for major species along the axial coordinate are shown in Figure 9. Experimental temperature data are also presented in Figure 8a, and good qualitative agreement between simulated and experimental temperatures is observed. However, direct quantitative comparison between experimental and predicted simulation temperatures is inappropriate since there are fundamental differences between the model and experimental geometries and heat transfer behavior. For example, temperature data shown for variable bed length experiments 13 (black circles) correspond to a fixed bed of spheres and a variable catalyst length instead of Figure 8. Simulated axial species and temperature profiles for a Pt catalyst (C 2H 6/O 2 ) 2/1, 5 slpm feed flow rate, 30 vol % N 2 dilution): (a) C 2H 6 and O 2 conversion and temperature; (b) yield to C 2H 4, CO, and CO 2; (c) yield to CH 4 and C 2H 2; and (d) H-atom yields to H 2,H 2O, and hydrocarbons (HC). Experimental temperature data (black circles/gray triangles 13 and square at 1 cm axial distance 7 ) are also presented in panel a for qualitative comparison with the simulations. Direct quantitative comparison between experimental and predicted simulation temperatures is inappropriate since there are differences between model and experimental geometries and heat transfer properties.

11 Ind. Eng. Chem. Res., Vol. 44, No. 10, Figure 9. Predicted net rates of homogeneous (dashed lines) and heterogeneous (solid lines) chemistry of the major species from simulations shown in Figure 8. a 1 cm foam monolith in simulations. From Figure 8a, it is evident that C 2 H 6 conversion increases with a more gradual slope in the first 2 mm of the catalyst than in the remaining length. The model predicts that only surface reactions of C 2 H 6 and O 2 consumption and CO 2, H 2 O, and CO formation take place in the first 2 mm (Figure 9). In agreement with experiments, 12,14 CO formation occurs at higher temperature than CO 2 (Figure 8b). The formation of oxidation products, such as CO 2, some CO (Figure 8b), and H 2 O (Figure 8d) with the consumption of about 10% of ethane and 50% of oxygen in the first 2 mm effects a strong temperature increase from the inlet temperature (350 C) to above 800 C (Figure 8a), which can be considered the approximate threshold temperature of gas-phase reactions. 11,12 Above 800 C, a change in slope of ethane conversion can be observed (Figure 8a), as ethane consumption in the gas phase is much faster than on the surface (Figure 9). The onset of gas-phase reactions after 2 mm is also revealed by the formation of products, such as ethylene (Figure 8b), CH 4 (Figure 8c), and H 2 (Figure 8d). C 2 H 2 (Figure 8c) is also formed, but for longer residence times and in lower amounts. Once ignited in the presence of oxygen, gas-phase reactions become much faster than surface reactions, and the catalyst does not seem to play a crucial role except favoring the formation of some CO and H 2 O. About 85% of the converted C 2 H 6 is reacted in the gas phase toward the formation of C 2 H 4,H 2, and CO (Figure 9). A reaction path analysis (Figure 10a) shows that ethane is consumed mainly via gas-phase cracking reactions ( 80%) and to a minor extent through the gasphase O 2 -assisted dehydrogenation ( 20%). All C 2 H 4 is

12 3464 Ind. Eng. Chem. Res., Vol. 44, No. 10, 2005 Figure 10. Major gas-phase reactions contributing to net rates of species production shown in Figure 9 for (a) C 2H 6 and (b) C 2H 4. The main path leading to ethylene from ethane goes through mainly dehydrogenation steps not involving oxygen. Steps involving oxygenated compounds account for 20% of gas-phase ethane consumption and less than 10% of ethylene formation. formed by homogeneous reactions, while heterogeneous reactions consume it to a small extent. Homogeneous steps involving oxygenated compounds account for less than 10% of ethylene formation (Figure 10b). H 2 is formed exclusively in the gas phase. CO is formed mainly on the surface ( 80%) under conditions of higher temperature and lower oxygen content than at the reactor inlet, where CO 2 is preferentially produced. Moreover, simulations suggest that very little CO 2 is formed in the gas phase and that CH 4 is formed only at high temperatures both in the gas phase and on the surface when a large surface coverage of C and H is attained. Because of endothermic reactions and heat losses along the reactor, temperature rapidly decreases after the first 2 mm, quenching other possible reactions and preserving ethylene (Figure 8a). Hence, all the products reach stable concentration within 1 cm from the inlet C 2 H 6 /O 2 /H 2 Feed Mixtures. The results from H 2 addition for a feed mixture with C 2 H 6 /O 2 /H 2 equal to 2/1/2 are shown in Figure 11, where axial species profiles are reported, and in Figure 12, where axial profiles of homogeneous and heterogeneous reactions rates are shown. Upon H 2 addition, heat production occurs mainly from oxidation of H 2 to H 2 O on the catalyst surface (Figures 11d and 12). H 2 consumption and H 2 O formation (up to 70% of the final value) in the first 3 mm of the reactor correspond to an O 2 conversion of 60% and an extremely low ethane conversion (very close to 0%). As temperature rises to above 900 C, effected by H 2 conversion, homogeneous reactions are ignited that lead to ethylene formation (Figure 11b) and the re-formation of H 2 (Figure 11d). At the ignition of homogeneous reactions, C 2 H 6 is consumed primarily through homogeneous chemistry (Figure 12). C 2 H 4 formation occurs exclusively in the gas phase (Figure 12) and is accompanied by formation of H 2, CO, and H 2 O. H 2 is consumed by surface reactions and is subsequently formed through homogeneous chemistry. CO 2 yield is extremely low because little O is available for deep C oxidation on the catalyst surface for the high fuel-to-oxygen ratios investigated, and C is instead oxidized to CO at temperatures higher than the inlet temperature (Figure 11b). Methane formation occurs in the gas phase and on the catalyst surface at high temperatures with a slightly higher yield than in the absence of H 2. A reaction path analysis for the H 2 addition case indicates very similar results for C 2 H 6 consumption and C 2 H 4 formation as the previous case with no hydrogen addition (Figure 10). C 2 H 6 is consumed and C 2 H 4 is formed mainly through nonoxidative gas-phase cracking (data not shown) System Performance at Higher Flow Rates and Pressure. In addition to gaining a better understanding of the kinetics involved in ethane ODH at short contact times, this mechanism was also used as a tool to predict system performance under experimental conditions not easily achieved. Specifically, the predicted effects of flow rate and pressure were explored. In both cases, a reaction mixture with C 2 H 6 /O 2 of 2/1 over a Pt catalyst was investigated with 30 vol % N 2 dilution. The heat transfer coefficient of the reactor was assumed not to change significantly with varying flow rate or pressure, since it is mainly dependent on the system geometry. Thus, the same heat loss function was implemented as in the previous simulations. It should be noted that the simulations used to explore higher pressures should not be considered exact quantitative predictions for two main reasons: (1) catalytic reactions and adsorption/desorption steps of most gasphase radicals and intermediates are not considered in the present work, and (2) some rate constants in the homogeneous mechanism for chemically activated reactions and hydrocarbon radical oxidation are strictly valid only at 1 atm. 42 Instead, results from higher pressure simulations are meant only as a qualitative guide to understand the product distribution from ethane ODH at high vs low pressure Effect of Flow Rate. Increasing flow rate has little effect on simulated ethane conversion over a range of space velocities ( h -1 ). This result can be explained by considering that very little of the catalytic channel (2 mm) is required to ignite and sustain gas-phase reactions. Increasing flow rate mainly moves the reaction front downstream and increases peak temperature (data not shown). However, the increase in the axial temperature profile does not significantly affect product distribution. C 2 H 4 selectivity is constant, and CO and CO 2 selectivities change by less than 1%. Other species selectivities are not significantly affected by an increase in flow rate. Therefore, larger flow rates can be safely fed to the reactor without a decrease in ethylene selectivity, until the reaction front is pushed outside of the reactor leading to extinction (in agreement with the experimental findings on Pt 13 ).

13 Ind. Eng. Chem. Res., Vol. 44, No. 10, Figure 11. Simulated axial species and temperature profiles for a Pt catalyst (C 2H 6/O 2/H 2 ) 2/1/2, 7.33 slpm feed flow rate, 20.5 vol % N 2 dilution): (a) C 2H 6 and O 2 conversion and temperature; (b) yield to C 2H 4 and CO; (c) yield to CH 4,C 2H 2, and CO 2; (d) moles of H 2 and H 2O per mole of H 2 fed. Experimental temperature data (black circles/gray triangles 13 and square at 1 cm axial distance 7 ) are also presented in panel a for qualitative comparison with the simulations. Direct quantitative comparison between experimental and predicted simulation temperatures is inappropriate since there are differences between model and experimental geometries and heat transfer properties Effect of Pressure. An increase in pressure of about 1 order of magnitude (from 1.2 to 10 atm) gives rise to an increase in ethane conversion of about 10% and a concurrent decrease in ethylene selectivity of about 3% (Figure 13a). In this case, the increased system heat formation due to increased partial pressure of the reactants leads to a higher system temperature and higher ethane conversion. At the catalyst back face, the temperature at 10 atm is about 50 C higher than at 1 atm, which has a negative effect on ethylene formation. Above 875 C, ethylene is further decomposed to CH 4 (CH 4 selectivity is increased from 4 to 8%) and to C 2 H 2 (Figure 13c). While CO 2 selectivity remains constant, pressure increase has a negative effect on CO selectivity, which decreases 8% over the pressure range considered in these simulations (Figure 13b). H 2 and H 2 O selectivities slightly decrease (Figure 13d), since larger amounts of CH 4 and C 2 H 2 and other C 3 - C 4 hydrocarbons (not shown) are formed at higher temperatures. These predictions agree qualitatively with experimental H 2 adition data on Pt that show increasing the pressure from 2 to 4 bar does not significantly influence catalytic combustion but lowers cracking selectivity to ethylene Discussion Previous kinetic models for ethane ODH on Pt monoliths are limited in that they either predict that ethylene formation occurs primarily on the catalyst surface 9,26 or cannot properly simulate co-feeding of H To address these issues, a surface mechanism was derived for Pt-based catalysts from published kinetic data for H 2 and CO oxidation on Pt combined with lumped steps for ethane decomposition and combined with a published homogeneous C 1 -C 2 reaction mechanism for rich oxidation conditions. 42 Through minor modifications, the surface mechanism (although not parametrically unique) was validated by experimental sets reported in the literature 5,7,34 for ethane/oxygen feeds with and without H 2 addition on two different catalysts. In addition, qualitative observations concerning the threshold temperature of surface reactions and product distribution at low temperatures (before the ignition of gas-phase reactions) were validated by published experimental data. 10,12 Specifically, experimental observations that only CO 2 and H 2 O are formed on Pt at low temperatures, while larger amounts of CO are produced at higher temperatures with larger O 2 consumption, are in agreement with the reaction rates reported in Figures 9 and The Role of Surface Reactions in Ethane ODH. Ethane ODH on Pt foams is a process driven mainly by gas-phase reactions; the catalyst ignites the feed mixture and sustains the homogeneous reactions leading to ethylene formation. C 2 H 4 formation reaches completion within 1 cm past the monolith exit and occurs mostly while O 2 is still present. Not all O 2 is consumed on the catalyst surface. Instead, a fraction of it speeds up homogeneous reactions. As argued previously, 14,48,49 oxygen is necessary in the gas phase to speed radical propagation reactions 50 and achieve quick ethylene formation, even though the main paths leading to ethylene are nonoxidative. Consistently, experiments showed that at the exit of the 1 cm long catalytic monolith, O 2 conversion and ethylene formation are not complete. 13 These results indicate that only 2 mm of catalyst may be sufficient to sustain the ODH process at the flow rates investigated. Nevertheless, surface