Ind. Eng. Chem. Res. 1994,33, 1669-1673
1669
Diffusional Effects and Intrinsic Kinetics for NO Reduction by CO over Pt-Rhlr-AlzO3 Monolithic Catalysts Vassiliki Kyriacopoulou, Apostolos Psyllos, and Constantine Philippopoulos' Department of Chemical Engineering, Laboratory of Chemical Process Engineering, National Technical University of Athens, GR 15780 Athens, Greece
Experimental kinetic data have been obtained for the CO-NO reaction system over laboratory made platina-rhodium monolithic catalysts with different thicknesses of catalytic active substrate on a monolithic ceramic support. A synthetic gas mixture which consisted of CO, NO, and Nz was the feed of an integral tubular reactor between 200 and 350 "C. A first-order kinetic rate expression with respect to NO concentration, which includes a n inhibition term of second order, was found to fit the experimental data via a model which also accounts for internal transport resistances. The reaction rate constants, the activation energy of the reaction, the enthalpy of the adsorption process, and the effective diffusivity of NO were estimated. T h e obtained kinetic and diffusional data can be used in the design of a catalytic converter with improved efficiency. Introduction Automotive three-way catalysts consist mainly of platinum and rhodium. The reaction of nitric oxide (NO) with carbon monoxide (CO) is a removal pathway for nitrogen oxides and is enhanced by the existence of rhodium in catalytic converters. It is also well-known that the performance of these catalytic converters is inhibited by internal mass transfer of the reactants. Therefore, in addition to the experimental estimation of the intrinsic kinetics of the reduction of NO by CO it is important to evaluate the internal diffusional effects in order to develop catalysts with improved NO removal efficiency. The catalytic reduction of NO by CO over platinum on alumina and over base metal oxides was studied by Force and Ayen (1972). Their experiments were carried out at temperatures between 240 and 500 "C, and the NO concentrations ranged from 50 to 5000 ppm. At a first approximation their kinetics seemed to be of first order with respect to both NO and CO. A higher approximation showed that the mechanism must be a combination of absorbed NO and CO on adjacent sites. Their kinetic rate equation is shown in Table 1,where the units of P a r e atm. Bettman and Otto (1983) reported first-order reaction kinetics with respect to NO concentration, for a CO-NO reaction carried out with a commercial three-way catalyst, in the absence of oxygen. Astrong inhibition factor which depended on CO concentration is included in their kinetic model which proposes a negative apparent activation energy for the reaction. This result is due to diffusion falsification of the kinetic data by the pore diffusion. Subramanian and Varma (1985) reported a kinetic rate equation for NO reduction using their kinetic data fitted with a model which accounts for pore diffusion and external transport resistances. In their experiments a gas mixture consisting of CO, 02,NO, Nz, and H20 was the feed in a tubular reactor which contained a Pt-Rhlr-AlZOa catalyst. They deduced that the NO reduction kinetics were of first order with respect to CO with an inhibition effect that depended on CO and NO. All the kinetic rate equations mentioned above are shown in Table 1. The activity and kinetic performance of cerium alumina supported rhodium catalysts on the CO-NO reaction have been presented by Oh (1990) in order to determine with
* To whom correspondence should be addressed.
Table 1. Kinetic Rate Equations for NO Reduction Force et al. (1972)
where
[
k, = 15.7 exp -
F] [
7650 KNO = 0.034 exp 3771 Kco = 3.15 exp[=]2900 Varma et al. (1986)
kl[COl'~4[0210~3[N010~'3 RNO
[l + k,[C0Il2
where
k, = 1.00 X lo" exp
-l&)O1
k, = 4.16 X 106exp[7i?;-] 1300 Bettman et al. (1991) RNO
=
8.84[CO] [NO] [1+ O.l94[CO] exp(5.26 X 1O4/Rrr?l2
a simple kinetic analysis and temperature-programmed desorption experiments the influence of cerium addition on NO reduction activity at low temperatures. An estimation of the effective diffusion coefficient of nitrogen oxide is given in a recent work by Beeckman (1991). The estimation was obtained by using a straightforward technique based on the NO flux measurement through the porous walls of a monolith-type ceramic catalyst in the absence of chemical reaction. In the present work, the CO-NO reaction system is investigated over laboratory made platina-rhodium monolithic catalysts with different thicknesses of a catalytic active substrate on a monolithic ceramic support, in order to elucidate the reaction kinetics and internal masstransfer effects. These catalysts were prepared by successive deposition of catalytic layers on a monolithic ceramic support from a slurry containing the precious metals, the additives, and alumina. A mathematical model
0 1994 American Chemical Society OSSS-5SS5/94/2633-1669~~4.50/0
1670 Ind.
Eng.Chem. Rea., Vol. 33, No.7.1994
which takesaccount of reaction kineticsand pore diffusion resistance was used, in order to estimate the key kinetic and diffusional parameters. Experimental Section Catalysts. The catalysts used in the experiments consisted of, first, amonolithicsubstrateof cordieritewith 400cells in." (62 cells cm-2), a wall thickness of 0.17 mm, and a porosity of 0.19 mL g1(measured by mercury penetration porosimetry) which had been supplied by Corning;second,athincoatingwhichis thecatalyticactive layer containing mainly alumina and the precious metals (platina and rhodium). The thin coating was loaded on the monolithic substrate by a wet impregnation procedure. The ceramic substrate was dipped into an aqueous slurry which was the product of a wet milling process (Blachou et al., 1992), with a solids content of 42 w t % and a mean particle diameter of the solids content of 2.8 pm deduced from a Rosin-Rammler distribution. The solidsconsisted of y -A1203.87.43%;Pt, 1.36 %; Rh, 0.14%; CeO2,1.82%; and LazO3,9.25% in dry basis. The alumina used in our experiments was prepared by a calcination process of hydrated alumina and consisted of y-alumina and boehmite, accordmgtoX-ray diffractionanalysis. The specific surface area of the alumina thus produced was derived as 176 m2 g-' by application of the BET equation, and it pmessed a pore volume of 0.68 mL gl obtained from mercury porosimetry. An incipient wetness impregnation method was used to load the precious metals and the additives from the corresponding water-soluble salts on the alumina. The mixture was then dried at 110 "C for 1 h, calcined at 600 OC for 2 h, and then reduced in a hydrogen flow of 200 mL min-' a t 400 OC for 2 h. This mixture, prepared and pretreated in the previous way, was the raw material for the wet milling process. Inordertoprepareacatalyst,theceramicsupportcoated with the wet slurry is blown with air to remove excess slurry and subsequently dried and calcined. Five samples of catalysts were prepared, where the catalytic active Substrate on the monolithic ceramic support was 11.45, 16.61,19.79,25.01,and34.84 % w/w,respectively. Catalyst layers were deposited one after the other until the desired total catalyst layer weight was achieved. By the above mentioned procedure, the prepared catalysts consist of thesamePt,Rh,Ce,andLacompositionofactivematerial hut they differ in substrate thicknesses. Scanningelectron microscopy (SEM) observation of the prepared catalysts reveals that the thickness of the wash coating was almost uniform except for the comers of the square monolithic channels. In addition, no distribution of the active components across the active substrate was revealed. Experimental Setup. The experimental setup is shown in Figure 1. The reactor is a quartz tubular reactor with an inner diameter of 13mm encased in an electrically heated furnace. The monolithic catalyst, 5 mm in length and having a diameter equal to that of the reactor, was placed in the middle of the heated zone. The reactor feed was a gas mixture, viz., CO, NO, and N2. The feed gas temperature was measured usinga thermocouple adjacent to the inlet of the bed section, and a PID controller was used for temperature adjustment. Samples from the reactant and product streams were analyzed using a combination of a continuous COIN0 analyzer (IMR-GmbH, 2OOO-P), and a gas chromatography a p paratus (HP-5710B) enables us to quantitatively analyze the following compounds CO, NO, COP,and NP. The gas chromatograph was equipped with two columns, a silica gel 80-100 mesh and a molecular sieves 5A 80-100 mesh, connected in series with a thermal conductivity detector.
c n i L.--
Figure 1. Experimental setup: PI, pressure indicator; TI, temperature indicator; S.C., Mmpling v a l v a
Experimental Conditions. The experimental conditions employed in the steady-state kinetic studies were a total flow rate of 37.7 mg 8-I. consisting of NO and CO with concentrations of 1200 and 2900 ppm, respectively (balanced with Na),atemperature range between 220and 350 OC. The reador pressure was maintained close to atmospheric. Experimental Data. Kinetic data were obtained as outlet concentrations of CO and NO for a set of reactor gas temperatures. By measuring the inlet and outlet concentrations of CO and NO the material balance was verified and the conversionof NO reduction was calculated. The experimental data are presented in Figure 2. An examination of the experimental results leads to the conclusion that for the same reaction temperature an increase of the active material does not raise the NO conversion for all catalysts. Generally, an increase in the catalytic active material (or the catalytic substrate in our case) is expected to cause an increase in the observed chemical reaction rate or an equivalent increase in the reactant conversion under the same operating conditions, provided that the entire amount of the catalyst participates in the process under investigation. The nonmonotonous change of the achieved conversion associated with an increase of the substrate thickness (or the equivalent total substrate mass), which has been observed in our experimental runs, is attributed to a kinetic model involving a second-order inhibition term, in conjunction with the reactant concentration decline due to pore diffusion limitations. I t should be noted here that reaction kinetics with a first-order inhibition term lead to a monotonous
Ind. Eng. Chem. Res., Vol. 33, No. 7, 1994 1671 100,
-
I
I
A dCNOs -
z=6
dz
Washcoat Content
‘NO4 = ‘N0,f Equation 3 was made dimensionless by introducing the following parameters:
z=O
Hence, the following equation (4) was generated: (4) 240 280 320 TEMPERATURE ( C) Figure 2. Experimental data of NO reduction. 160
200
360
rate decrease in contrast those with a second-order term which causes the rate to pass through a maximum.
with boundary conditions y = l at l = 1
and the effectiveness factor was given by the equation ( 5 ) below:
Data Analysis The model used for the experimental data fitting included a mass balance equation for an integral reactor with heterogeneous kinetics and catalyst pore diffusion and reaction. The model was based on the following assumptions: (a) an irreversible reaction takes place, that is, the reduction of NO by CO, (b) the model is one dimensional, and the reactants concentration is varied across the reactor axis, (c) pore diffusion takes place in the porous catalyst substrate, (d) the catalytic process is controlled by reaction and pore diffusion, and (e) the effectiveness factor is assumed for slab geometry and firstorder reaction with second-order inhibition. The kinetic rate equation which is assumed to contain a second-order inhibition term is
where CNOand Cco are the reactants concentration. The material balance for an integral tubular reactor over an infinitesimal distance dx along the monolithic catalyst yields the following equation:
where a is the mass of catalyst per reaction surface area, u is the circumference of a monolith channel, and q is the effectiveness factor, with the following boundary conditions: Assuming that the active substrate temperature is radially uniform, the mass balance equation for the reactant across the substrate depth ( 2 ) is
&lr(Y) dY (5)
1=
So1Wdy Method of Correlation. A Marquardt-Levenberg (Marquardt, 1963) method was adopted for parametric fitting. This is an optimization method which uses a steepest descent algorithm in order to find a set of parameters that is near the maximum (or minimum) of a given function. The parameters for the present correlation were ko, kl, E, AHr, Do, and Ed, and the function to be minimized was the least squares deviation between calculated and experimental reactor exit NO conversions. The second order differential equation (4) was solved by a collocation algorithm where eq 4 was discretized by the means of appropriate orthogonal polynomials and selected interpolation grid points. A system of algebraic equations instead of eq 4 was solved iteratively until convergence was attained. Then, by the use of eq 5, the effectiveness factor, q, was calculated. The integration of eq 5 was performed with the method of weighted residuals. The result is a predicted value for q, which is used in eq 2. The value of q as a function of the Thiele modulus and /9 is presented also as a parametric nomogram in Figure 3. An appropriate set of initial values for the parameters was determined, eq 2 was solved via a Gear algorithm (Gear, 19711, and the rate of reaction was calculated by means of interphase concentrations. At every point on the reactor axis the value for the bulk NO concentration must be equal to the value of the bulk NO concentration used in the effectiveness factor calculation. If the difference between them exceeds a preset tolerance then the whole procedure is repeated with a new value of the bulk NO concentration and a new predicted value for q. The least square deviation between calculated and experimental reactor exit NO conversions was calculated in every iteration, and if it was greater than the required tolerance, the parameters were then adjusted until convergence was reached. Results and Discussion The experimental runs were fit by the method described above, and the plotted points are scattered close to the 4 5 O line, indicating a good quality of fit as shown in Figure
with the following boundary conditions:
4.
1672 Ind. Eng. Chem. Res., Vol. 33, No. 7, 1994
(Smith,1983): C 1
L"
8
U. v)
8
0.1
C
Q)
6
Ew
0.01
0.001 10 -I
1 10 10 ' 10 ' Thiele modulus,@ Figure 3. Effectiveness factor with inhibition kinetics (isothermal
DK = 9700a(T/M1'2 where DK is the Knudsen diffusion coefficient (cm2/s),a is the catalyst pore radius (cm), T is the catalyst pore and M is the molecular weight of the temperature (K), diffusing substance. Application of the formula above with a mean pore radius of 35 A, T = 520 K, and M = 30 gives for the Knudsen diffusion a value DK = 1.41 X le2 cm2/s. The value of the pore radius was deduced from mercury porosimetry data on samples of prepared catalysts over the pore radius range of 18-1200 A. The equation for the effective diffusion coefficient above gives (for T = 520 2 X 1k2cm2/s,and the tortuosity K) 8Value Of D e , ~ &=~ 2.17 factor is estimated equal to 1.5 (generally the tortuosity factor ranges between 1 and 5). The values of the Knudsen and the effective diffusion coefficients calculated above are of the same order of magnitude, so it can be stated that Knudsen diffusion is the dominant diffusion process mechanism.
case).
Conclusions
7
C
60
50
" * i 40
5 v)
W
*p/.
Washcoat Content
.rj
o 11.5% n 16.6%
20
25.0%
34.8%
10
Experimental NO conversion (93)
Figure 4. Comparison of reaction rates calculated using the
optimized kinetic parameterawith the experimentallyobservedrates. The kinetic equation with the parameters obtained from the NO reduction data was estimated kO RNO
exp
[&]
cNOcCO
=
whereko = 2.26 X 1013cm3(~rn~/mol)/(sg,~,~),E = 12900 cal/mol, kl = 5.70 X 106cms/mol, and AH = 2400 cal/mol. From the kinetic rate equation for NO reduction which fits our experimental data it can be deduced that the NO reduction kinetics follows an almost first-order kinetics law in the case of low NO concentration since the inhibition term is rather small, that is, 8 = kl eXp(AH/Rg/T)CNo,B