Global Oxidation Kinetics for Propylene, CO, H2, and NO: Determination and Optimization, Lab Reports of Health sciences

Download Global Oxidation Kinetics for Propylene, CO, H2, and NO: Determination and Optimization and more Lab Reports Health sciences in PDF only on Docsity! Global Kinetics for Platinum Diesel Oxidation Catalysts Chaitanya S. Sampara,*,† Edward J. Bissett,‡ Matthew Chmielewski,§ and Dennis Assanis† Department of Mechanical Engineering, UniVersity of Michigan, 2032 W.E. Lay AutomotiVe Laboratory, 1231 Beal AVenue, Ann Arbor, Michigan 48109, General Motors R&D, 30500 Mound Road, Warren, Michigan 48090, and Aerotek, 26211 Central Park BouleVard, Southfield, Michigan 48076 Global oxidation kinetics for propylene (C3H6), CO, H2, and NO were determined over a platinum (Pt) catalyst with simulated diesel exhaust between 200 and 415 °C over wide concentration ranges. An integral reactor with high space velocity capability (up to 2 million h-1) was used to generate low and moderate conversion data for the rate-generation process. First-order concentration dependency for all the reactants involved in the C3H6, CO, and H2 oxidation reactions captured the experimental behavior very well. For the NO-NO2 reaction, the rate was found to be first order with respect to NO and 0.5 order with respect to O2. An overall inhibition term including only the effects of CO and NO for all the reactions was found to be adequate over the range of conditions examined in this study. A simplified 1D reactor code was used to interpret the data and predict exit concentrations. An objective function was defined for the optimization process, which is sensitive to model predictions at all conversion levels. An optimization strategy was also developed to systematically simplify the form of rate expressions and to generate proper initial guesses for each of the intermediate steps. The final rate forms were compared with light-off curves generated on a full-scale reactor mounted on a 1.7 L Isuzu engine at University of Michigan. Introduction Diesel engines are gaining increasing popularity because of their increased fuel economy and better performance charac- teristics among available propulsion systems. Stringent emission regulations have forced researchers to devise strategies that combine advanced combustion modes and a combination of aftertreatment components to overcome the perennial soot-NOx tradeoff in the diesel engine. A diesel oxidation catalyst (DOC) is an integral component of the diesel aftertreatment architecture.1 While it is used in conjunction with a diesel particulate filter (DPF) to generate an exotherm that is used to oxidize the soot in the DPF, it can also be used to generate an equimolar mixture of NO-NO2 for the efficient operation of the selective catalytic reduction (SCR). An overriding purpose, however, is to oxidize the increased total hydrocarbons (THC) and CO, which are generated by uncon- ventional combustion strategies due to lean and rich premixed compression ignition (PCI) and exhaust gas recirculation. The reactions of interest in a DOC are THC, CO, H2, and NO oxidation under lean conditions. Optimization of the exhaust system architecture needed for such strategies requires efficient reactor models that can accurately predict the conversion levels of various species. While 1D models, which have been proven to effectively model aftertreatment components, well-describe heat and mass transfer effects, the biggest problem with their reliability and accuracy lies in the reaction kinetics. Voltz et al.2 reported oxidation kinetics for propylene (C3H6) and CO for slightly lean exhaust for temperatures between 200 and 370 °C. The experiments for the kinetics study were carried out with simulated gasoline engine exhaust with C3H6 as the representative hydrocarbon. A Langmuir-Heinshelwood-type rate expression with CO and C3H6 inhibition terms was proposed for both CO and C3H6 oxidation reactions. Their reaction orders with respect to the corresponding reactants were found to be 1. NO, which is mostly inert under the oxidizing conditions considered, was found to inhibit the respective reactions. They proposed an empirical term to represent its effect on the rates. Yu Yao3 and Morooka and Ozaki4 studied CO + O2 and C3H6 + O2 reactions at 300 °C and between 122 and 200 °C, respectively. The reaction rates that were reported in the form of a power law exhibited postive order with respect to O2 and negative order with respect to C3H6. Yentekakis et al.5 studied the C3H6 + NO + O2 system and reported that NO had significant inhibition on C3H6 oxidation for a platinum catalyst. Since our feed is more complicated (C3H6 + CO + NO + O2), we allow for the possibility of our data exhibiting rate forms that are different from the usual convention. Olsson et al.6 and Després et al.7 reported microkinetics for NO oxidation on Pt and Pt-BaO catalysts, respectively. The feed stream in either case did not include any reductants such as THC, CO, or H2. Majewski et al.8 reported that platinum catalysts support NO-NO2 reaction rather than other reactions involving NO such as C3H6 + NO, CO + NO, etc. Differential data under isothermal conditions is ideally suited for global rate generation since it provides direct simultaneous measurement of the concentrations, the temperature, and the rates themselves. However, the necessary experiments to generate such differential data might require extreme space velocities, and simultaneously non-negligible but differential rates for multiple reactions are not possible when the rates are too different. It is a common practice in the literature to use the rate forms from Voltz et al. and recalibrate the constants with light-off curves or Federal test procedure (FTP) data. The rates obtained using light-off curves will be best, by construction, at nominal inlet concentrations and near the light-off temperature, and they may not be adequate when used to model other reactor operating conditions. FTP data is influenced by additional transient and transport-influenced processes that make it unnecessarily com- plex for the sole purpose of generating kinetics. In this paper, we propose a systematic way of generating global reaction rates * Corresponding author. E-mail: csampara@umich.edu. † University of Michigan. ‡ General Motors R&D. § Aerotek. 7993Ind. Eng. Chem. Res. 2007, 46, 7993-8003 10.1021/ie070642w CCC: $37.00 © 2007 American Chemical Society Published on Web 10/06/2007 and apply the method to generate oxidation kinetics for the species observed in a DOC. C3H6 was used to represent all the hydrocarbons in the diesel exhaust. While we realize that this might be an oversimplification, developing kinetics for this HC model is important to understand and evaluate the importance of this assumption. A more complicated HC speciated system is studied in our subsequent work.9 Experimental Section Concentration/Temperature Domain. An important first step in generating useful kinetic rates is to properly define the ranges of concentrations and temperatures for which the resulting rates are intended and choose the actual test points within this domain that will be used to inform the optimizer of the measured rates. While a domain that is too small does not encompass all intended applications, one that is too large may force global rates to reflect unnecessarily complex behavior not observed in engine applications. The upper bounds for inlet concentrations of reactants that encompass practical operation were determined. Test conditions for lower concentrations were established by stepping down in factors of three based on our previous experience.10 Since our kinetic rates are intended to be used locally throughout our DOC reactors, it is also important to include the small concentrations expected in the downstream portions of the reactor as well as the expected inlet concentrations. The HC and CO concentra- tions were chosen to be high enough to describe engine output during PCI combustion, but not so high to describe what could be achieved with postinjection of fuel. For all tests, water concentration was held constant at 8.7%, and CO2 concentration was held constant at 10%. Four temperature levels were chosen based on typical DOC operation. The temperature spacing was equidistant in 1/T when the temperature is in K. The various species concentrations and temperature levels are given in Table 1. A full factorial covering all six variables would result in 640 tests. However, such a full matrix is not representative of the diesel exhaust in general, and hence, we impose additional constraints, given in Table 2, to obtain a more realistic set. Approximately 25 test points (concentration combinations) at each temperature were then randomly selected for testing. Because NO oxidation rate was greatly suppressed in the presence of other species, separate experiments with the test matrix given in Table 3 were conducted to infer the NO rate. While NO2 was not in the inlet feed for lower temperatures (200 and 255 °C), where the reverse reaction is negligible, it was added to the inlet stream for higher-temperature cases (325 and 415 °C). Pretesting Catalyst Preparation. A Pt loading of 7.7 g/ft3 on a γ-Al2O3 washcoat was used for all the experiments. The monolith supported catalyst (400 cells/in.2, wall thickness of 0.007 in.) was hydrothermally aged in a furnace at 650 °C for 16 h. A constant 2.2 L/min flow of 10% H2O in air was fed to the furnace for the entire 16 h aging period. For CO chemi- sorption, the catalyst cores were ground into powder and reduced in H2 at 300 °C for 20 min. After cooling the sample to room temperature, CO was pulsed over the samples to determine CO uptake. The data reported for CO chemisorption are based on the average of two measurements. The active site density determined based on this method was 0.080 mol-site/m3, corresponding to a Pt dispersion of 5.7%. We assumed the ratio of CO to active Pt site was 1 for the purpose of calculation of the surface site density. Reactor Setup. Experiments were carried out in a well- insulated vertical stainless steel tubular reactor containing either 1 in. o.d. for 3/4 in. diameter samples or 2 in. o.d. for 1.5 in. diameter samples. Samples were held in place using a com- pressible ceramic paper wrap that also ensured that reactor flow passed through the catalyst channels. Volumetric space velocities between 50 000 and 2 000 000 h-1 were achieved by varying the reactor flow rates and the catalyst size. The full range of feed concentrations for C3H6, CO, H2, and NO was achieved by using two concentration levels of BOC certified compressed gas bottles. Heating of the catalyst was achieved by flowing air, N2, CO2, and vaporized water through two inline heaters. Reactor section temperatures were monitored using three Type K thermocouples. The first thermocouple was place just above the inlet face of the catalyst. The second was placed 3/4 in. below the first. The third could be adjusted axially, to accom- modate catalysts of varying length, and was placed just after the outlet face of the catalyst. The reactor system pressure was maintained at 1.6 atm to ensure proper flow through the Fourier transform infrared (FTIR). Analysis. The reactor inlet and outlet flows were analyzed using a MKS MultiGas 2030 process stream FTIR for CO, CO2, NO, NO2, N2O, NH3, C3H6, formaldehyde, CH4, and water. Hydrogen (H2) was analyzed using a mass spectometer. There a few important details regarding our experiments that we would like to highlight. • Because of the presence of excess amounts of oxygen, as is typical for diesel exhaust, we assumed that the oxidation reactions were more dominant than other side reactions. • Monolith samples with thin washcoats (∼20 µm) were used to minimize diffusion resistance within the washcoat. • Monoliths are known to have excellent but not perfect heat and mass transfer characteristics. Just as in actual applications, when the surface reaction rates are sufficiently fast, such as at higher temperatures, total reaction rate will be limited by the rate of transport from the gas to the catalyst surface and so will become insensitive to the surface kinetics. For this reason, we do not even measure kinetics in transport-limited regimes. • Individual experiments that essentially produce either complete or no conversion of a particular species are locally insensitive to reaction kinetics and so are of little quantitative value for our methods of kinetics extraction. Therefore, a wide range of space velocities available with our experimental setup was exploited by adjusting the space velocity of each individual experiment to avoid conversion extremes. Table 1. Concentration and Temperature Levels for Various Species HC (C3), ppm CO, ppm O2, % NO, ppm H2, ppm T, °C 2000 900 13 400 200 200 600 300 4 100 70 255 200 100 325 60 30 415 20 Table 2. Concentration Constraints for Test Matrix HC < 3CO HC > CO/3 H2 < CO H2 > CO/10 Table 3. NO, NO2, and O2 Concentration Levels NO, ppm NO2, ppm O2, % T, ppm 450 150 13 200 150 50 5 255 325 415 7994 Ind. Eng. Chem. Res., Vol. 46, No. 24, 2007 equally at each temperature and each temperature and species contributes equally to the overall norm calculation. The norm definition is given in eq 19, where the sum of the two log terms is interpreted as taking one term or the other depending on the conversion for that species. It should be noted that the data points likely to be strongly influenced by mass transport effects (for example, high tem- peratures) would be expected to contribute little to the norm since we have good confidence in the model’s ability to describe transport behavior. The errors in our measurements were incorporated in the norm by removing from the summation any test points for which ∆xg,i e or xg,i e was <5 ppm (3 ppm for H2). This was based on the accuracy of our FTIR/mass spectometer measurements. Results and Discussion C3H6, CO, and H2 Oxidation. As we described earlier, the general procedure for generation of global rates includes first assuming particular forms for the relevant rates and then optimizing for the rate parameters to best fit the experimental data. Our evaluation of how well the rate form represents the data is based on the norm or the objective function, which we obtain at the end of the optimization. For the scope of this study, we start from conventional Langmuir-Hinshelwood forms that have been used for similar purposes in the literature, and we successively add or remove inhibition/enhancement terms to check if there is a significant improvement in our norm value. Adjusting the degrees of freedom available to the optimizer in this additive fashion will always result in improved fits with more terms, so we exercised subjective judgment in determining “significant” improvements based on inspection of the corre- sponding parity plots. We also placed a premium on simplicity of the rate forms and on precedents in the literature. As discussed in the Introduction section, the oxidation rates for C3H6 and CO from Voltz et al., as written by Oh and Cavendish,11 are widely used throughout the literature for inferring these rates. Oh and Cavendish also assumed that the H2 rate is similar to that of CO. The rate form as given by Oh and Cavendish is where Although we understand that we are not generating microki- netics, we loosely refer to the constants in the denominator as adsorption constants. All rate and adsorption constants have Arrhenius form, A e-E/RT. We start with a modified version of this. First, we convert the rates to functions of molar concentrations per unit volume (mol/m3) instead of mole fractions. When obtaining rates from elementary steps, one naturally arrives at global rates in terms of concentrations and not mole fractions. The pressure effect is then automatically absorbed in the rates, and they can be expected to apply at any pressure, in particular, at the elevated 1.6 atm of our reactor experiments. This can be done by multiplying and dividing the mole fractions with c ) p/RT, the molar concentration. Half of these c’s are absorbed into cs,i ) cxs,i. The other half, along with the temperature, T, in the inhibition term of eq 21 are absorbed into the new rate constants below. To do this, p (at the 1 atm of Voltz) and R are constants, so that the new approximate rate constants below are formally Arrhenius forms multiplied by various factors of T. These approximate rate constants are forced into pure Arrhenius form by replotting at a few representative temperatures and refitting. Next, we remove the second term in the inhibition. Voltz et al.2 stated that this term was included for empirical reasons only, Figure 1. xg comparison between model and experiment. Figure 2. ∆xg comparison between model and experiment. norm2 ) 1 nsp′ ∑ i)1 nsp′ 1 nT ∑ j)1 nT 1 nj ∑ k)1 nj [log2[∆xg,im∆xg,ie ]conve50% + log2[xg,imxg,ie ]conv>50%] (19) ri ) khixs,ixs,O2 G1 [ molcmPt2‚s] (20) G1 ) T(1 + Kh COxs,CO + Kh C3H6xs,C3H6) 2(1 + Kh CO‚C3H6xs,CO 2 xs,C3H6 2)(1 + Kh NOxs,NO 0.7) (21) Ind. Eng. Chem. Res., Vol. 46, No. 24, 2007 7997 so we discarded this term to initially minimize the degrees of freedom. Finally, the exponent on the concentration of NO in the inhibition term was set to unity. Although this term is empirical, we initially retained it here to capture the effect of NO on the rates (if any). However, if the exponent of the NO concentration is <1, the resulting rate is not differentiable at zero, making the rate infinitely sensitive to the concentration of NO as the NO concentration approaches zero (dri/d[NO] f ∞ at [NO] ) 0). While NO is relatively inert compared to the oxidation of CO and C3H6, as in the case Voltz considered, there are regimes of interest for DOC operation where depletion of NO is possible. The rate forms that we used as a starting point for our optimizations are, therefore, given by eq 22. where 1. Initial Guess for Optimization. Global optimization methods generally have the advantage of being less sensitive to the choice of initial guess compared to local optimization methods. For our approach, we prefer a careful, controlled, incremental improvement of the rates, so we restricted the optimization to a local method and accepted the more stringent demands for a good initial guess. Working from the modified rate constants inferred from Oh and Cavendish was not generally effective, at least partially because we are dealing with a much different concentration regime. Therefore, we developed an alternative approach to generating an initial guess described below. The division of the temperature regime into discrete bins aids the process of generating a proper initial guess in our case. For generating rate and adsorption constants for C3H6, CO, and H2, we assumed NO was linearly varying between inlet and exit. Since, in most of the cases, there was very little NO conversion, the linear approximation for NO is well-justified. The optimiza- tion is performed on individual ki and Ki (rate and adsorption constants) for the rate forms given in eq 22 at a fixed temperature. Since the scaling ensures that the variables (scaled ki and Ki) are O(1), any number that is O(1) is suitable as a starting guess. The results from the optimization at individual temperatures are plotted in an Arrhenius plot (log ki vs 1/T) to generate proper initial guesses for all pre-exponentials and activation energies. A typical Arrhenius plot generated for the initial guess of the rate constant for C3H6 is shown in Figure 3. NO Oxidation. The oxidative environment promotes the NO oxidation reaction rather than the NO reduction reactions that occur over three-way catalysts (CO + NO, C3H6 + NO, etc.). Our measurements commonly indicate that the net NO oxidation reaction is greatly suppressed in the presence of reductants in the stream. Hence, we conducted separate experiments to infer the NO h NO2 rate with only NO, NO2, and O2 present in the feed. We note that this reaction is reversible and is limited by equilibrium. All the data obtained from these sets of experiments gave us differential data. We started initially with data that contained only NO and O2 in the inlet stream to infer the forward rate. Simple log-log plots of NO oxidation rate vs concentration at individual temperatures indicated that a simple power law formulation was not adequate but that a Langmuir-Hinshelwood rate form involving NO inhibition effects was well-suited. The oxygen dependence on the rate remained nearly constant for this exercise. The choice for stoichiometric coefficients for the reactant concentrations in the rate was based on the power law results from individual temperatures and the simplest, commonly used algebraic form that ensures a vanishing global rate at equilibrium. From these observations, the following rate form was proposed for NO oxidation. This form is physically reasonable, and it vanishes at equilibrium. Also, the apparent reaction orders in the numerator respect the stoichiometric coefficients of the reaction, and the term in the denominator appropriately captures the self-inhibition of NO that we observed in the preliminary analysis. cR is the concentration at the reference pressure of 1 atm. Keq is the equilibrium constant, based on the free energy of the NO oxidation reaction. kNO and KNO represent the rate constant and adsorption constants, respectively, with Arrhenius forms of temperature dependency. We first optimize the rate expression given in eq 24 for data obtained with feeds containing NO, NO2, and O2. A parity plot obtained with the optimized parameters appears in Figure 4. Note that we plot only ∆xg (xg in - xg out) when comparing experimentally measured values with model predicted values with optimized parameters because all the data generated for NO + NO2 + O2 in the stream had conversions less than 50%. We conclude that this model captures the experimental data well over the entire concentration and temperature regimes. We realize, however, that this rate expression should predict small NO conversions in the presence of reductant in the stream since we do not observe any significant net oxidation reaction for NO when reductant is present. To evaluate this rate, we plot in Figure 5 the experimental ∆xNO against model ∆xNO for the cases in the original test matrix that contained C3H6, CO, and H2 (reductants) along with NO in the stream. Figure 3. Initial guess for AC3H6 and EC3H6 from optimization at individual temperatures. ri ) kics,ics,O2 G2 [ mol mol-site‚s] (22) G2 ) (1 + KC3H6cs,C3H6 + KCOcs,CO) 2(1 + KNOcs,NO) (23) rNO ) kNOcs,NO xcs,O2 (1 + KNOcs,NO) [1 - cs,NO2Keqcs,NO x cRcs,O2] (24) Keq ) 1.5 × 10-4 e6864/T (25) 7998 Ind. Eng. Chem. Res., Vol. 46, No. 24, 2007 Clearly, the model overpredicts the NO conversions by an order of magnitude in comparison to the experimental data when reductants are present in the feed. The rate expression, hence, needs to be modified when being used for these cases. We take advantage of not just the similarity in the NO inhibition terms between eqs 23 and 24 but also their equality when reductants are absent (cs,C3H6 ) cs,CO ) 0) to generalize eq 24 to eq 26, which can be used when reductant is also present. where G2 is given by eq 23. Final optimization of the rate constant kNO in this last form is performed when the optimization of the full problem is considered in the next section. Optimization of the Full Problem. With the rate forms (eqs 22 and 26) and initial guesses known, we optimize for all the reactions (C3H6, CO, H2, NO, and NO2) for the entire concentration and temperature domains. The objective function (norm) value at the end of the optimization was 0.4870. The next step is to systematically add or remove terms accounting for inhibition/enhancement to see if any improvement is obtained in the norm value. For the first pass, we remove one term at a time, namely, KCOcs,CO, KC3H6cs,C3H6, or KNOcs,NO, to estimate which term, when removed and a subsequent reoptimization of A’s and E’s is performed, results in a solution with the least perturbation to the norm. Table 4 gives the values of the norm when each of the inhibition terms for C3H6, CO, and NO were removed individually. These values show that the removal of the C3H6 term from the denominator resulted in the smallest perturbation in the norm, and that the perturbation is small enough to justify removal of this term in the denominator. That is, we judged that the gain in simplicity of the rate form outweighed the nearly imperceptible loss in ability to represent the data. Removal of any terms in addition to C3H6, namely, CO or NO, is pointless since the values in Table 4 demonstrate that, in either case, the reoptimized norm was significantly increased even with one more degree of freedom (presence of C3H6). H2 was present in the feed stream, and hence, we studied a possible enhancement effect of H2 on the other rates. We included a term, (1 + KH2cs,H2), in the numerator of all the rates to represent any H2 enhancement. Rerunning the optimization gave a very marginal improvement in the overall norm value. The final norm value with the inclusion of the H2 enhancement term was 0.4868. Since this was not a substantial improvement in the norm value, we concluded that the H2 effect is not worth including in the rate expressions. The absence of the terms for HC inhibition or H2 enhance- ment should not be interpreted as denying the existence of these effects. Detailed experiments designed specifically to elucidate these details for these individual reactions could very likely reveal these commonly observed effects. Rather, we claim that these effects are not large enough to require representation within our global reaction scheme to describe rates over several orders of magnitude in concentration and in our temperature regime, especially when it is desired to keep the total degrees of freedom represented by our kinetic parameters small. While our results are designed to capture the overall trends quantita- tively, the parity plots below show that the predictions for some of the individual cases can still contain substantial errors that could mask additional chemical effects. The final rate forms we used for the kinetic parameter optimization are given by eqs 27 and 28. Figure 4. Comparison of ∆xNO between model and experiment for cases that have only NO, NO2, and O2 in the feed stream (no reductant). Figure 5. Comparison of ∆xNO between model and experiment (all temperatures) in the presence of reductants in the feed stream using the rate form of eq 24. rNO ) kNOcs,NO xcs,O2 G2 [1 - cs,NO2Keqcs,NO x cRcs,O2] (26) Table 4. Final Norm Values after Various Terms in the Inhibition Terms Were Removed term removed final norm value KCOcs,CO 0.9345 KC3H6cs,C3H6 0.4873 KNOcs,NO 0.6932 Table 5. Rate Constants as a Result of the Final Optimization of the Full Problem pre-exponential value activation energy value AHC 1.123 × 109 EHC 5.156 × 104 ACO 3.725 × 106 ECO 2.213 × 104 AH2 1.335 × 107 EH2 3.032 × 104 ANO 1.086 × 103 AaCO 10.57 EaCO -9.709 × 103 AaNO 32.19 EaNO -1.901 × 104 Ind. Eng. Chem. Res., Vol. 46, No. 24, 2007 7999 reactor studies. A series of degreening tests were conducted on the engine-mounted DOC before the light-off data were taken. There are several reasons behind the disagreement of the light- off curves over the entire range of temperatures for total hydrocarbons (THC) and CO. • In the engine exhaust, only the gas-phase temperatures upstream and downstream of the DOC were measured. As in the kinetic analysis, we used these two temperatures to generate a linear interpolant for the temperatures inside the DOC. Since this procedure ignores any potential local sharp temperature rise produced by the rapid exotherms within the reactor, we may underestimate the temperature sensitivity of reactor response to light-off conditions. • The assumption that C3H6 represents the THC of the entire diesel exhaust is a very crude one. Diesel exhaust, we believe, should be split into at least two categories: one that consists of heavier hydrocarbons (e.g., unburned fuel components) and the other that consists of partially oxidized lighter hydrocarbons (e.g., combustion products). • Since the NO concentrations and conversions were not entirely reliable, and since all the rates are a function of NO through the inhibition, the rates would not be able to capture the experimental behavior fully. • Finally, the aging procedure followed for the catalyst used in the reactor (16 h hydrothermal aging) was different from that followed for the catalyst used with the engine tests (using engine exhaust as reported by Knafl et al.13). Conclusions Global oxidation reaction rates for C3H6, CO, H2, and NO in the presence of excess O2 were developed over wide temperature and concentration ranges. A common inhibition term, containing only factors for CO and NO for all the rates, was found to satisfactorily represent our data. An attempt to simply and directly capture potential enhancement due to H2 in our rate expressions showed that this effect did not significantly affect this representation. We used the following necessary machinery for global reaction rate generation: careful choice of temperature and concentration domains, random sampling of data for reactor measurements, a reactor capable of high space velocities, optimization routines, reactor codes, and, finally, an optimization methodology to generate proper initial guesses and successively improve the rate form. An objective function that critically evaluates all model vs experiment comparisons is defined. All the rates gave reasonable agreement with the laboratory experimental data. Attempts to validate the kinetic model against the results of engine tests with a full-size DOC show that the measured CO and THC conversions increase more rapidly with temperature during catalyst light-off. Further work would be required to clarify the origins of this discrepancy. Acknowledgment We thank Calvin Koch for lending his experimental expertise and for helping to generate the test matrix. We thank Richard J. Blint for helping ensure that we had the resources and materials to carry out this work. We thank Alexander Knafl for answering any questions regarding the engine data. Finally, we thank Se H. Oh for his technical reviews and advice, which greatly improved the quality of this work. Notation aj ) active site density for reaction j, mol-site m-3 A ) face area, m2 Ai ) pre-exponential for rate constant, m6 mol-1 mol-site-1 s-1 Aai ) pre-exponential for adsorption constant, mol-1 m-3 c ) total molar concentration of gas, mol m-3 cR ) total molar concentration at 1 atm, mol m-3 cs, cs,i ) vector and component, respectively, of molar concen- tration of trace species at catalyst surface, mol m-3 Dh ) hydraulic diameter of channel, m Di,m ) binary diffusion coefficient of species i in the mixture, m2 s-1 Ei ) activation energy for rate constant, J mol-1 Eai ) activation energy for adsorption constant, J mol-1 G, G1, G2 ) inhibition terms in rate expressions ∆G ) free energy for the NO-NO2 reaction, J mol-1 ki ) rate constant, varying units k̂i ) rate constant for rates in ref 11, mol K-1 cm-2 s-1 km,i ) mass transfer coefficient for species i, mol m-2 s-1 Ki ) adsorption constant for species i, m3 mol-1 Kh i ) adsorption constant for rates in ref 11 Keq ) equilibrium constant for NO-NO2 reaction L ) length of reactor, m Mi ) molecular weight of species i, kg mol-1 Sh ) Sherwood number for mass transfer p ) total pressure, N m-2 R ) gas constant, J mol-1 K-1 rj ) rate of production of reaction j, mol mol-site-1 s-1 si,j ) stoichiometric coefficient of species i in reaction j S ) surface area per reactor volume, m-1 Σi ) diffusion volume of species i T ) temperature of solid/gas phase, K Tr ) reference temperature, K V ) volume of reactor, m3 w ) molar flow rate, mol s-1 wr ) reference molar flow rate, mol s-1 ∆xg ) change in mole fraction across the reactor xi 0 ) inlet mole fraction of species i xg,i ) mole fraction of species i in bulk gas phase xi,r ) reference mole fraction of species i Figure 11. CO validation with engine-generated light-off curves. 8002 Ind. Eng. Chem. Res., Vol. 46, No. 24, 2007 xs, xs,i ) mole fractions of species in gas at catalyst surface z ) axial position, m Literature Cited (1) Blakeman, P. G.; Chiffey, A. F.; Phillips, P. R.; Twigg, M. V.; Walker, A. P. DeVelopments in diesel emission aftertreatment technology; SAE 2003-01-3753; Society of Automotive Engineers: Warrendale, PA, 2003. (2) Voltz, S. E.; Morgan, C. R.; Liederman, D.; Jacob, S. M. Kinetic study of Carbon Monoxide and Propylene oxidation on platinum catalysts. Ind. Eng. Chem. Prod. Res. DeV. 1973, 12 (4), 294. (3) Yu Yao, Y.-F. The oxidation of CO and hydrocarbons over noble metal catalysts. J. Catal. 1984, 87, 152. (4) Morooka, Y.; Ozaki, A. Regularities in catalytic properties of metal oxides in propylene oxidation. J. Catal. 1966, 5, 116. (5) Yentekakis, I. V.; Tellou, V.; Botzolaki, G.; Rapakousios, I. A. A comparative study of the C3H6 + NO + O2, C3H6 + O2 and NO + O2 reactions in excess oxygen over Na-modified Pt/γ-Al2O3 catalysts. Appl. Catal., B 2005, 56, 229. (6) Olsson, L.; Persson, H.; Fridell, E.; Skoglundh, M.; Andersson, B. A kinetic study of NO oxidation and NOx storage on Pt/Al2O3 and Pt/BaO/ Al2O3. J. Phys. Chem. B 2001, 105, 6895. (7) Després, J.; Elsener, M.; Koebel, M.; Kröcher, O.; Schnyder, B.; Wokaun, A. Catalytic oxidation of nitrogen monoxide over Pt/SiO2. Appl. Catal., B 2004, 50, 73. (8) Majewski, A. W.; Ambs, J. L.; Bickel, K. Nitrogen oxides reactions in diesel oxidation catalysts; SAE 950374; Society of Automotive Engi- neers: Warrendale, PA, 1995. (9) Sampara, C. S.; Bissett, E. J.; Chmielewski, M. Global kinetics for a commercial diesel oxidation catalyst with two exhaust hydrocarbons. Ind. Eng. Chem. Res., submitted for publication. (10) Bissett, E. J.; Oh, S. H.; Sinkevitch, R. M. Pt surface kinetics for a PrOx reactor for fuel cell feed stream procesing. Chem. Eng. Sci. 2005, 60, 4709. (11) Oh, S. H.; Cavendish, J. C. Transients of monolith catalytic converters: Response to step changes in feedstream temperature as related to controlling automobile emissions. Ind. Eng. Chem. Prod. Res. DeV. 1982, 21, 29. (12) Fuller, E. N.; Schettler, P. D.; Giddings, C. J. A new method for prediction of binary gas-phase diffusion coefficients. Ind. Eng. Chem. 1966, 58 (5), 19. (13) Knafl, A.; Busch, S. B.; Han, M.; Bohac, S. V.; Assanis, D. N.; Szymkowicz, P. G.; Blint, R. D. Characterizing light-off behaVior and species-resolVed conVersion efficiencies during in-situ diesel oxidation catalyst degreening; SAE 2006-01-0209; Society of Automotive Engi- neers: Warrendale, PA, 2006. ReceiVed for reView May 7, 2007 ReVised manuscript receiVed July 1, 2007 Accepted August 3, 2007 IE070642W Ind. Eng. Chem. Res., Vol. 46, No. 24, 2007 8003