Beruflich Dokumente
Kultur Dokumente
KYRIACOS ZYGOURAKIS*
Department of Chemical Engineering,Rice University,Houston, TX 77251, U.S.A.
and
RUTHERFORD ARIS
Department of Chemical Engineeringand Materials Science, University of Minnesota, Minneapolis,
MN 55455, U.S.A.
Abstract-Oxidation reactions of mobile exhaust gas in a catalyticmonolith are considered in this paper.
The exhaustgas is deseribed by four concentrations, namely those of the carbon monoxide, fast- and
slow-oxidizing hydrocarbons and hydrogen. The problem of multiple reactions and diffusion in the
catalytic layer of the monolith walls is studied first. The diffusional resistance of the catalytic layer is often
significant in spite of its thinness and it seemsimportant to allow for both the hydrogen and hydrocarbon
oxidation reactions in analyzing the system for steady-state. The influence of hydrogenon the multiplicity
of steady-statewall temperaturesis also studiedand it is found to be important, because hydrogenreacts
rapidly at lower temperatures.
733
734 K. Z~cioua~~rs and R. ARIS
platinum catalysts have been studied extensively (see The boundary conditions (5) indicate that the
for example [S-lo]), there is but a limited amount of ceramic forming the monolith walls is assumed to be
kinetic data for the oxidation of H&O or impermeable to matter.
Hz-CO-hydrocarbon mixtures. We will assume in Equations (4&(6) become in dimensionless form:
this study that the reaction rate expression for the
oxidation of hydrogen is given by:
d’yi &?(I -_yXl -y,)RCy,,yJ
drz = -
(7)
r, = - k,Y,Y,fR’ (3)
l-_YY,=&--E,Y, (10)
3. SOLUTION OF THE SYSTEM OF COUPLED BVPS The monotone iterations method is a successive
AND THE QUESTION OF MULTIPLICITY approximation scheme. Starting from an
The steady state problem of multiple reactions upper-upper solution 6+“,yz0) = (1, 1) as an initial
taking place in the catalytic layer requires the solu- approximation, we compute a sequence of iterates
tion of the system of boundary value problems (v,“, yr”) by successively solving linear and uncoupled
(BVPs) described by eqns (7HlO). Obviously, the BVPs of the form
two BVPs for the concentration of carbon monoxide
and fast-oxidizing hydrocarbons (i = 1, 2) are cou- d?,
pled and highly nonlinear, because of the term -@ - &Y: = -JXY,” - ‘9 Y,“_ ‘)
(E,, - E,yr)RoJ,, yr)_ Once this system is solved for __Q:-‘; j= 1,2
y,(l) and y,(c), the two remaining BVPs for the (14)
slow-oxidizing hydrocarbons and hydrogen become
dyl
-=O @r-O; i-l,2
linear in y,(r) and y4(t) respectively and can be easily (15)
solved. de
We now turn our attention to the system of
coupled and nonlinear BVPs: ~+a,;=0 f&r =I; i = 1,2. (16)
d’Yi
p = -n_Y,.Y2); i = 1, 2 (11) It has been shown [22] that this sequence of approx-
imations decreases uniformly in the interval [0, l] and
monotonically until it converges to the maximal
dyi
-=O @t=O; i=l,2 (12) solution (y:, y_$) of the system. Another sequence of
d< iterates (w,“. wz”) can be generated starting from
(w,O, wrO)= (0,O) as an initial approximation and
dyi
,g+g$;=o @i=l; i=l,2. (13) solving successively the BVPs defmed by eqns
(14)-(16). The latter sequence increases uniformly
and monotonically until it converges to the minimal
It was first attempted to solve this system using a
solution (*y,, eyr). If now (JQ,yr) is any other solution
global collocation scheme. Newton-Raphson’s and
of the system (llH13), then
Powell’s method[20] were used to solve the resulting
system of nonlinear algebraic equations. However,
cyiIy,Iy:; i = 1,2 for all c in [0, 11. (17)
significant numerical difficulties were encountered
(see[21-231 for a detailed discussion). The nonlinear
Therefore, if the computed maximal and minimal
iterations diverged even for moderate temperatures
solutions are identical, the BVP system (llH13)
(80&900”K) and the accuracy of the approximations
has a unique solution. If they are not identical, then
to the profiles y,(c) and y*(c) deteriorated rapidly
the problem has at least two positive solutions and
with increasing temperature, to the point that they
any other solution will be bounded by these two. The
would become unacceptable for the solution of the
high accuracy of our numerical algorithm[23] makes
BVP which would yield the hydrogen profile.
such a determination possible.
The major drawback of the nonlinear iteration
The nonlinearitiesf,(y,, y2) andf,(y,. y2) in eqn (11)
methods, however, lies in the fact that they do not
have been studied in detail[21]. We have found that,
answer the question of multiplicity of steady-state
for all values of temperatures and concentrations in
solutions. We should note here that the question of
the range of practical interest, they satisfy the
multiplicity has not been answered convincingly even
conditions
in the case when a single reaction (CO oxidation) was
considered to take place in the catalytic layer of %>-A. i=1,2 (18)
monolith reactors[l4,25] (see also [16] and [24]). ay, ”
The multiplicity question is all the harder to answer
when multiple reactions are taking place in the a
-s-O; i#j, i,j = 1,2 (19)
catalytic layer. The use of a searching technique in
8Y,
conjunction with nonlinear iterations would require
prohibitively long computation times. For all these for all values of y,, yr, where A,, A, are sutliciently
reasons, the monotone iterations scheme with adap- large positive constants. Thus, the convergence of the
tive collocation developed in companion nonlinear iterations defined by eqns (14H16) is al-
publications[22,23] was used to solve this system of ways guaranteed, according to the results presented
coupled BVPs. This method provides a priori unique- in [22].
ness criteria as well as a multiplicity criterion, that The monotone iterations method is indeed a pow-
can be numerically implemented. In addition, it pro- erful tool and has enabled us to investigate the
vides guaranteed convergence for all values of Thiele problem of solution multiplicity. The following range
moduli and computes the solutions with very high of parameters was completely characterized:
accuracy by using an adaptive collocation
scheme 1231. T = 500-1200X, Y,, = O-0.06, Ybz= O-2000 ppm.
736 K. ZY~~UIUKISand R. Aars
10-l
:1 :1
--_
102 lc+
1
J
A
-3
10
z 001 0.02
Y
bl
0.03 0.04 la36
0
Ybl
0.04
Fig. 4. The effect of HCl bulk concentration on the Fig. 4. The effect of temperature on the apparent rate of
apparent rate of reaction i, for CO (including diffusional reaction ;, for CO (including diffusional limitations).
limitations) -- Interphase mass transfer limited rate.
tration, the CO oxidation apparent rate reaches its temperature the qssumption of intraphase diffusion
maximum at about Yb, = 0.05 and then starts de- control introduces an error in the computation of the
creasing, in accordance with the negative order kinet- apparent reaction rate i, that can be as high as 10%
ics of the reaction. for Y,, = 0.04 and Y, = IOOOppm. We note, how-
Although the rates of both reactions are very fast, ever, that at these high temperatures the hydro-
the reactions do not reach the intraphase diffusion carbons exhibit negligible inhibition effects. The ex-
control regime even at 9OO”K, as Fig. 7 shows. At this planation is presented in Fig. 8, where the
100 _
T q SO0 OK
1.2
Ti
.8 tt /
lf //
//’ / ,/ /
/ / /
Yb2-. 400 ppm
I
// / / / / /’
4
/
0.01 0.02 0.03 0.04 0.05
Ybl ’ bl
Fig. 5 Effectiveness factors for CO (-_) and I-ICI (- - -) Fig. 7. Apparent rate of reaction i, for CO - high tem-
oxidation-the effect of temperature. perature case. -- interphase mass transfer limited rate.
K. Z~ooua~~rs and R. Ams
concentration profiles of CO and HCl are plotted. 5. STEADY-STATE TEMPERATURES FOR THE
Since the hydrocarbons react much faster than the MONOLlTH WALLS
CO at this temperature, their dimensionless surface In the previous section we have shown that the
concentration is almost 2.5 times smaller than that of hydrocarbon content of exhaust gases affects
CO and drops quickly to zero as the reaction pro- significantly the rate of oxidation of carbon monox-
ceeds in the catalytic layer. ide. Also, the heats of combustion of the hydro-
In all the above computations we have considered carbons have very large values (4195 kJ/gmol for
a comercially available catalyst with an alumina hexane vs 283 kJ/gmol for carbon monoxide). There-
support. The physical properties of the support were fore even very low concentrations of hydrocarbons
given by Hegedus et aZ.[24]. The elfective diffusivities can contribute a great deal to the heat generated in
of the reacting species in the catalytic layer were the reactor. Many investigators[2], [27] have also
computed according to the random pore model of suggested that if the exhaust gas contains much
Wakao and Smith[l3], which is applicable to cata- hydrogen, then the maximum wall temperature may
bc much higher than the equilibrium adiabatic flame
temperature of the gas. Hegedus[27] has actually
measured temperatures higher than the adiabatic one
during the oxidation of H, in a single passage mono-
T ~700 OK
lith reactor.
Our objective here will be to determine the effect of
Yb,Z 0.02 the hydrogen and hydrocarbon oxidation reactions
Yb2Z 400 ppm on the steady state temperatures predicted for the
.2 monolith walls. We will consider a thin slab of
/ ceramic substrate, coated on both sides by a thin
yi
/ layer of platinum supported catalyst and exposed to
‘bi
two streams of exhaust gas. Four gaseous species
oxidize within the catalytic layer: carbon monoxide,
hydrogen and fast- and slow-oxidizing hydrocarbons
(fractions HCl and HC2 respectively). The two gas
streams may have different bulk temperatures and
compositions.
Since the catalytic layers are very thin when com-
pared to the monolith walls, we will assume that the
reactions are actually taking place on the wall-gas
.6 .8 interface. The geometry of the system is shown in Fig.
t 10.
Fig. 9. Concentration profiles of CO, HCl and H, in the If we neglect axial conduction, the overall heat
catalytic layer. balance equation for the slab becomes
Multiple oxidation reactions and diffusion in the catalytic layer of monolith reactors 739
W =w.+(w,-wJ[. (26)
CERAMIC I
SUBSTRATE I
I
Cl-_ -----___ t
I
L--y2 % = zL(w‘J (29)
where
The computations were done for four groups of In Fig. 11 the steady state temperature predictions
data, thought to represent typical conditions and of Models (I), (II) and (III) are compared for data
compositions of exhaust gas for different driving group A. Model (I) predicts a single steady state over
modes of the automobile. These data are given in the entire temperature range considered. Model (II)
Table 1. In the basic case considered, the values of predicts ignition’ at about 600°K and after a very
heat transfer coefficients were computed from the narrow and sharp multiplicity region the wall tem-
limiting Nusselt number for flow between parallel perature jumps to high values.
plates. Computations were also carried out for Model (III) predicts three unmistakable steady
different passage geometries. The formulas proposed states for bulk temperatures between 420 and 580°K.
by Hawthorn[ll] were used for the calculation of The steady states are widely spaced and an analysis
heat and mass transfer coefficients. The apparent showed that they are not due to numerical errors.
reaction rates ii were computed according to the When the bulk temperatures increase, the model
method discussed in Section 3. predicts ignition at about 580°K. The wall tem-
The results of the computations are shown in Figs. peratures attain very high values afterwards, which
11 to 13, where the predicted wall steady state exceed significantIy the adiabatic reaction tem-
temperatures are plotted vs the bulk temperature of peratures predicted for the given concentrations of
the gas streams. reactants.
1400-
DATA3 A
r
400 600 aoc
Tf [“K]
400 800
Tf,*K Fig. 12. Model III steady state temperature predictions for
various automobile driving modes (A -Deceleration, lou
Fig. 11. Wall steady state temperatures T predicted by hydrocacban concentration, E-Deceleration, high hydra
Models I, II and III vs bulk fluid temperature T/. carbon concentration, C-Idling, D-Acceleration).
Multiple oxidation reactions and diffusion in the catalytic layer of monolith reactors 741
included in monolith calculations. The effect of hy- [l] Kno J. C. W., Morgan C. R. and Larsen H. G., Sot.
Auromo~iue Engrs. 1971, Paper No. 71-0289.
drogen is especially important since it was found that
[2] Wei J., Ado. Catalysis 1975 24 57.
it can significantly enlarge the multiplicity region by 133 Jackson M. W., Sot. Automotitre Engrs. 1966, Paper
lowering dramatically the extinction temperature. It No. 66-0404
also results in higher wall temperatures than those [4] Margolis L. Ya., Adu. Catalysis 1963 14 429.
predicted by considering only the CO oxidation. [5] Patterson W. R. and Kemball C., J. Catalysis 1963 2
465
When the oxidation of hydrocarbons is also consid- [6] Schwartz A., Holbrook L. L. and Wise H., J. Catalysis
ered, the wall temperature obtains still higher values 1971 21 199.
and the ignition temperature increases. [7l Voltz S. E., Morgan C. R., Liederman D. and Jacob S.
M., Ind. Engng Chem. Prod. Res. Develop. 1973 12 294.
[S] Bliessner J., Ph.D. Thesis, University of Minnesota,
1979.
NOTATION
[9] Hanson F. V. and Boudart M., J. Catalysis 1978 53 56.
mass transfer Biot number for component i, [lo] Boudart M., Collins D. M., Hanson F. V. and Spicer
N. E., J. Voc. Sci. Tech. 1977 14 441.
k&ID,
1111
. _ Hawthorn R. D.. A.I.Ch.E. Svrw. . _ Series 1973 70 137,
heat transfer Biot number, h,L,/k
428.
mass heat capacity of solid, J/kg”K [12] Dabill D. W., Gentry S. J., Holland H. B. and Jones
effective diffusivity of component i, m2/s A., J. Catalysis 1978 53 164.
constants defined by eqn (10) [13] Smith I. M., Chemical Engineering, Kinetics, 2nd Edn.
nonlinear functions of y, and y2 defined by McGraw-Hill. New York 1970.
1141
_ _ Wei J. and J&ker E. R.. Adv.Chemistrv Series 1975 143
eqn (7) 116.
functions defined by eqn (31) [15] Lee S. T. and A+ R., Chem. Engng Sci 1977 32 827.
mass flow rate, kg/(m*)(s) [16] Smith T. G., Zahradnik J. and Carberry J. J., Chem.
heat transfer coefficient, J/sm*“K Enen@ Sci. 1975 30 763.
[17] V&&ba J., Mikus O., Ngwen K., Hlavacek V. and
thermal conductivity of solid, J/sm”K
Skrivanek J., Chem. Engng Sri. 1975 30 201.
mass transfer coefficient for component i, m/s (181 Kays W. M. and London A. L., Compact FIeat Ex-
thickness of catalytic layer, m changers, 2nd Edn. McGraw-Hill, New York 1964.
thickness of monolith wall, m [19] Bird R. B., Stewart W. E. and Lightfoot E. N., Trans-
apparent rate of reaction i, gmol/sm* port Phenomena. Wiley, New York 1960.
[20] Powell M. J. D., In Numerical Methods fir Nonlinear
time, s Algebraic Eyuations, (edited by P. Rabinowitz), p. 115.
temperature, “K Gordon & Breach, New York 1970.
dimensionless solid temperature, T/T, [21] Zygourakis K., Ph.D. Thesis, University of Minnesota
length coordinate, m 1980.
dimensionless concentration of component i [22] Zygourakis K. and Aris R., J. Nonlinear Analysis:
Theorv. h&rho& & AD& 1982 6 555.
mole fraction of component i [23] Zygo&kis K. and A& R., Comput. Chem. Engng in
press 1982.
[24] Hegedus L. L., Oh S. H. and Baron K., A.I.Ch:E:J.
Greek symbols 19-c 23 632.
AH, heat of reaction, J/gmol [25] Young L. G. and Finlayson B. A., A.1.Ch.E.J. 1976 22
343.
i dimensionless length coordinate in monolith
[26] Heck R. H., Wei J. and Katzer J. R., A.I.Ch.E.J. 1976
wall, x/L, 22 477.
effectiveness factor for reaction i [27] Hegedus L. L., A.I.Ch.E.J. 1975 21 849.
dimensionless fluid temperature, TJT,; i = a [28] Hlavacek V. and Vortruba J., Ado. Catalysis 1978 27
59.
or b
[29] Aronson D. G. and P&tier L. A., Arch. Ration. iUech.
dimensionless length coordinate in catalytic Anal. 1974 54 175.
layer, xfl [30] Jones A., Firth J. G. and Jones T. A., 1. Phys. E:
solid density, kg/m3 Scientific Znstr. 1975 8 37.
dimensionless time, kt/pc,L,* [31] Acres G. J. K., Platinum Metals Rev. 1960 10 60.
[32] FisherG. B. and Gland J. L., General Motors Research
Thiele moduli, &I2 Y,,/D,i Publication GMR-2944R. PCP-92R.
k, = k,exp [E*/Rr]k,,
where
R = (1 + k,,(C0))2 (C) C
The constants k,, and k,, were taken from the kinetic 460 560
T OK
model proposed by Voltz et a1.[7]. The effectiveness factors
for the two reactions were considered to be equal to unity, Fig. Al. Heat released during the oxidation of Hz-CO
and the catalytic layer was assumed to be isothermal. Then mixtures over supported platinum catalyst for k. = 1.1 and
simple mass balances for the two reactants yield
E* = 1600°K. (- * (Hz) = 0.005, (CO) = 0; - (H3 = 0.01,
(CO) = 0.005; -- (H,) = 0.005, (CO) = 0.01; -.- (H2) = 0,
kp,( Y,. - Y,) = lr,, @) (CO) = 0.01).
where i stands for CO or H,, Y, and Y,, are the respective
bulk and surface concentrations of the two reactants and r,;
are the reaction rates expressed per volume of catalyst.
The system of nonlinear algebraic equations (D) is solved
to give the surface concentrations of the reactants. In order
to compare. our results to those of Dabill et al. [12], we have
computed the total heat generated by the two reactions as
effect of the parameters on the oxidation of hydrogen is for by our simple model. Experimental invcstigations[32]
shown in Fig. AZ. We should note that the minimum have shown that water adsorbs on platinum less strongly
oxidation temperature of hydrogen does not depend than carbon monoxide. One would expect therefore that CO
strongly on the values of k. pnd E*. is a much stronger inhibitor of these reactions than H,O.
The possible inhibition e.Rects of water vapor on the The validity of this assumption, however, should be checked
oxidation of hydrogen on platimm[31] are not accounted exDerimentally.