Peri, J. B., J. Phys. Chem., 69, 220 (1965). Pines, H., Manassen. J., Adv. Catal., 16, 49 (1966). Pines, H., Haag, W., J. Am. Chem. SOC.,82, 2471 (1960). Sandell, F. B., "Colorimetric Determinations of Traces of Metils", Interscience, New York, N.Y., 1959. Santacesaria, E.,Carra, S.,Adami, I., lnd. Eng. Chem., Prod. Res. Dev.. 16, 41 (1977). Schwab, G. M.. Kral, H., Proc. lnt. Congr. Cat. 3rd. Amsterdam, 20 (1964).
Tanabe, K., "Solid Acids and Bases", Academic Press, New York, N.Y., 1970. Yamadaya, M., Shimomura, K., Huckida, H., Shokubay (Tokio), 7 (3), 313 (1965).
Receiued for reuiew June 21, 1976 Accepted November 5 , 1976
Kinetics of Ternary Nitric Oxide Reduction on Ruthenium Ronald Schleppy, Jr., and Yatish T. Shah' Chemical and Petroleum Engineering Department, University of Pittsburgh, Pittsburgh, Pennsylvania 1526 7
The reduction of nitric and nitrous oxide by mixtures of carbon monoxide and hydrogen in helium over a fiber glass supported Ruthenium catalyst has been studied. The rate data from a differential tubular reactor, operating at atmospheric pressure and at temperatures ranging from 210 to 360 OC,were correlated with power law models. Feed concentrations of the NO and N20 were varied from 0.2 to 1.7 mol YO,while those of the CO and H2 were varied from 1.O to 5.0 mol % . N20 was found to be a significant intermediate product of NO reduction at low conversions. The relative rates of reaction and the dependence of reaction rate upon feed concentration were in general agreement with previous studies of the same reaction systems using different supports.
Introduction Recently Schleppy and Shah (1976a) presented a kinetic study of nitric oxide reduction by carbon monoxide and hydrogen in binary mixtures over fiber glass supported ruthenium and platinum catalysts. At low conversions it was found that the NO was reduced primarily to both N2 and N2O. NH:j formation was generally less than 10% except for NOHP/Pt which was not investigated. The low conversion data were correlated into power law rate expressions. Attempts to find a Langmuir-Hinshelwood type expression which would successfully correlate the data were unsuccessful. The present paper will extend that study to include the kinetics of NO reduction over a fiber glass supported Ru catalyst in a ternary mixture (NO-CO-Hz). Also considered are the kinetics of N20 reduction, the inclusion of which will provide some explanation for the N20 intermediacy phenomena. The principal reactions occurring in a catalytic converter between NO, CO, HY, and NzO are
-
+ CO 2 N 0 + CO NO + H2 NO
---*
---*
2N0
+ Hz
+ COn N20 + COZ f/2N2+ HzO '/2N2
+ H20 2NH3 + 2H20
(2) (3)
-
(5)
---*
(7)
---*
N20
+ 5H2 N20 + Hz N2 + HnO N20 + CO N2 + COS CO + H20 e CO2 + H i
2N0
(1)
(4)
(6)
(8)
Literature not cited in the previous work includes a study of the dual state behavior of Ru catalysts by Voorhoeve and Trimble (1975). The primary difference they noted between the two states during NO reduction was in the amount of NH3 formed. Only small differences were observed in any of t h e . reactions considered in this work. Sugi et al. (1975) studied N20 intermediacy in the reduction of NO by CO over an Fe203
catalyst. Their data indicated that the NO was much more strongly adsorbed than the N20, thereby displacing the newly formed N20 from the catalyst surface before it had time to react with the CO. At higher temperatures, however, the NzO did show an increasing tendency to react before being displaced. Studies of NO reduction in multicomponent systems over noble metal catalysts contain some interesting observations. Kobylinski and Taylor (1974) report that H2 is the predominant reducing agent in the H*-CO-NO/Ru system, but that the overall rate of reduction is closer to that of the CO-NO/Ru binary system. Klimisch and Barnes (1972) claim that the H20 present in the exhaust gas reacts with the CO via the water gas shift reaction to produce H2 which thereby becomes the dominant reducing species. Voorhoeve and Trimble (1975) note that CO is the main reducing agent a t low temperatures in the Hz-CO-NO-H20-C02/Ru system, whereas H 2 tends to predominate a t higher temperatures. According to Taylor and Klimisch (1973) this is due to the fact that the H2-NO reaction is faster a t higher temperatures than the CO-NO reaction. Bauerle and Nobe (1974) found that the presence of HzO reduced the selectivity of Ru toward Nz. Bauerle et al. (1972), Klimisch and Taylor (1973), and Shelef and Gandhi (1972) in separate studies on the effects of 0 2 on NO reduction all agree that although the presence of 0 2 slightly reduces the amount of NH3 produced, it has little effect on the overall NO reduction rate provided that a significant excess of reducing agent is present.
Experimental Section Details of the equipment used, a description of the catalyst, and the experimental procedure are given by Schleppy and Shah (1976a). During the present study two different Ru catalyst beds were used. No reduction was carried out over a bed made up of 28 0.4% Ru impregnated fiber glass cloth disks (2.35 g, 2.6 cm) "sandwiched" out with neutral fiber glass to a final bed length of 5.9 cm. For the NzO reduction runs an Ind. Eng. Chem., Prod.
Res. Dev., Vol. 16, No. 1, 1977
47
Table I. Binary Power Law Correlations ___System
NO-CO/Pt NO-CO/Ru
Rxn 1 2 1
2
NO-HJ~U N?O-CO/Ru NIO-H,/RU
"s
3 4 7 6
ko, g-mol/min/g of cat.
1.12 x 1.22 x 1.54 X 5.56 X 3.03 X 2.98 X 3.26 x
104 1010' lo6 lox lofi lob
-
EX
-_
cal/g-mol
H1
22.5 30.5 23.0 18.8 26.3 23.6 33.4
Power law exponent CO NO
s, 90"
g-mol/min/g of cat.
-
21.9 22.0 26.0 22.4 17.0 13.3 24.8
6.94 4.65 36.8' 35.1' 97.0' 32.9' 23.2 >17 000
-
0.69 1.08 -0.62 -0.31 -0.36 -0.22
--
-1.16
-
1.03
-0.66
-1.13 0.81
0.59
-
0.87 0.97
-
-
= Standard deviation of % error between rate data and correlation.
0.5% N O or 0.50/0N10 at 300 "C.
rx
NrO --
--
-_
Predicted from correlations for 2.0% CO or 2.0% H? and
Rate predicted for the N?O/Ru catalyst.
unsandwiched bed of 36 0.15% R u impregnated disks (2.50 g, 3.0 cm) was used. In all runs helium was used as the carrier gas a t a total gas flow rate of 4250 cm 'bin a t STP. The 10.3% NO/He and the 10.4% NLO/He mixtures used as feed gases were found to contain N? as an impurity in amounts of 0.089% and 0.011%, respectively. During the calculations the measured values of product N r were corrected accordingly.
r
Results a n d Discussion Method of Analysis. As described in the previous paper, all concentrations were obtained directly or indirectly from the gas chromatograph. Conversions were generally kept below 20% to justify the assumption of a differential reactor. Again the data were fitted to the appropriate power law model for the reaction under consideration. The most general power law expression for the reactions considered here is r = ko exp(-EIRT)PaNoP"c.oP1 ~
~ P ~ h > (9) g
Obviously, the partial pressure terms for those species not present in a given reaction system were not used in the correlation. In analyzing the ternary data N H 3 formation was neglected since mass balances revealed that it always accounted for less than 10%of the N O reduced and since there was no practical method available to analyze for it. This left reactions 1 through 4 to account for the reduction of the NO itself. Because of the stoichiometry of these four reactions, it was impossible to determine the individual extents of all of them from balances on the products. For this reason only two reaction combinations were correlated: N? formation, (1) and (3); and N,O formation, (2) and (4). Examination of the asymptotic standard deviations of the parameters provided by the nonlinear regression analyses used in the previous paper showed that the greater part of the difference between model and data was attributable to uncertainty in the value of ko (which can be viewed as the activity of the catalyst) rather than to variations in the dependence of the rate upon temperature (E) or upon the concentrations (a, b, c); 95% confidence limits generally placed the true value of ko within f20-30% of its mean value while the true values of E , a, b, and c were placed within f l - 2 % of their mean values. Furthermore, it was noted that the catalyst activity changed very little during a given sequence of uninterrupted runs. This means that rather good estimates of E, a, b, c, and d can be obtained with a minimum of runs provided that the runs over which a given parameter is varied are taken consecutively without interruption. All data presented herein were taken in groups of 4-5 consecutive runs. In each such group of runs only one parameter was varied over its range of interest. Other concentrations were held as close as possible. T h e temperature was sometimes adjusted so as to keep the % conversion within tolerable limits. 48
Ind. Eng. Chem., Prod. Res. Dev., Vol. 16, No. 1, 1977
1 .65
1.70
1.75
1.80
1.85
l Q Q Q / T . GEG K Figure 1. Reaction rate vs. 1/T, CO-NzOIRu. Data points adjusted to: 2.0%CO, 0.5% NZO. Line is the best line drawn through all data points.
Linear least squares fits were then used to obtain the separate parameters. First, E was calculated and then used, if necessary, to correct the other groups to some arbitrary average temperature. As they were obtained, the values of a, b, c, and d were used to correct for minor fluctuations in the supposedly constant compositions. This algorithm was continued for 5 interations. B i n a r y Kinetics. In an effort to better understand NO reduction in general and to come closer to a working kinetic model, data have been taken and correlated for NjX-CO/Ru. Data for N10-Hi/Ru were not obtainable since 100% conversion was observed a t the lowest controllable bed inlet temperature, 200 "C. The NzO-CO/Ru temperature data are illustrated in Figure 1. Similarly, the N.0 and CO groups are shown in Figure 2. Overall, the temperature varied from 270 to 341 "C. In each of the Figures 1 through 9, the data points shown have been adjusted slightly, using the final correlation parameters, to the arbitrary conditions given in each separate figure. This adjustment parallels the process used in obtaining the correlation parameters in that it corrects for the known fluctuations in the supposedly constant parameters. The straight lines in each plot represent the final correlation, eq 9, from all the data groups for that system. The gap between the points and the line is thus a measure of the difference between the catalyst activity a t the time of the particular group and the system average value of ho from which the line was drawn. Complete numerical data for all runs are available from Schleppy (1976).
i
% H,
bibi\@~ 5 3 i 5 6 i 8 9 % eo Figure 2. Reaction rate vs. % CO and % N20, CO-NrO/Ru. oh CO data points adjusted to: 0.5%NiO, 320 "C. % N20 data points adjusted to: 2.0% CO, 320 "C. Lines are derived from the best fit through all data. '19-1
h
3
k
5
Figure 4. Reaction rate vs. % Hs,Hs-CO-NO/Ru. Data points adjusted to: 1.5%CO, 0.5%NO, 235 O C . Lines are obtained from eq 9 and Table 11.
'm
o
-
N2
a
m
o
0
D
o
o
X X
'm
1
1.92
'a 1.96
2 .a0
10BQ/T,
2.04
2.06
DEG K
Figure 3. Reaction rate vs. 1/T,Hg-CO-NO/Ru. Data points adjusted to: 1.5% H?, 1.5%CO, 0.5% NO. Lines are obtained from eq 9 and Tahle 11.
The resulting power law parameters for the N20-CO/Ru system as well as those from the previous NO reduction study are presented in Table I for comparison purposes. Values of the reaction rate, r , a t an arbitrary set of given conditions for each system studied are given in Tables I and 11. For the Ru systems, these values have all been calculated for the 0.15% Ru catalyst used in the NzO runs. Sample runs of all the NO reactions were taken on the new catalyst to determine its activity. Values of ho reported, however, are for the catalyst that the correlation runs were actually performed on. An examination of the Ru results in Table I leads to several interesting conclusions. If similar site Langmuir-Hinshelwood type behavior is assumed for these reactions; i.e., the apparent order of reaction for each reactant is determined by its relative adsorptive bond strength, then the order of bond strength on
Ru is NO > CO > N 2 0 . This clearly follows from the observation that NO inhibits the NO-CO/Ru reactions whereas CO inhibits the reaction between N20 and CO. Similar conclusions were made in part from adsorption data by Taylor and Klimisch (1973) and by Otto and Shelef (1973) among others. In addition, the relative reaction rates and activation energies indicate that less NzO will be formed a t higher temperatures since El > E2 and E: E d .Although reaction 7 is actually slower than reactions 1 and 2 a t 300 "C, its much greater activation energy points to faster NzO reduction and the gradual disappearance of N20 as a n intermediate product a t higher temperatures. These observations are all consistent with those in the literature which were cited above and in the previous paper. I t is clear t h a t the correlations presented here suffer from Ind. Eng. Chem., Prod. Res. Dev., Vol. 16, No. 1, 1977
49
Table 11. Ternary Power Law Correlations System
Prod.
NO/Ru
NL
NJO NLO/RU H i 0 COL
ko, g-mol/min/g of cat.
E x lo-?, cal/g-mol
Hl
2.04 x 104 5.28 X lo5 7.04 x 10" 4.96 X lofi
20.2 23.4 46.6 30.6
0.36 0.26 0.20 -0.20
Power law exponent CO NO -0.32 -0.11
-0 038
-0.091 -1.80
-
-0.48
NzO
s , %"
-
12.56 9.25
-
1.34 0.74
r x 10-6,b g-mol/min/g of cat.
11.1
135.0' 73.7' 5.27
6.56
3.09
s = Standard deviation of % error between rate data and correlation. Predicted from correlations for 2.0% CO, HP,and 0.5% NO or NlO at 300 "C. ' Rate predicted for the NJO/RUcatalyst.
0 P-
+- aU
m-
e
- *\
z
: :E, : : : I
o X
N2°
x
\
1
o
E
m
-i
cs
b
9 '10'
2
3
%
% NO
Figure 6. Reaction rate vs. % NO, H*-CO-NO/Ru. Data points adjusted to: 1.5%Hz, 1.5%CO, 235 "C. Lines are obtained from eq 9 and Table 11.
5
4
6
H,
Figure 8. Reaction rate vs. % Hn, Ha-CO-N20/Ru. Data points adjusted to: 2.0% CO, 0.5%N20,320 "C. Lines are obtained from eq 9 and Table 11.
m
0
m
\
P
lo-
rn-
*I-
u
m-
0
0 -..
N-
I e
--
E-.. '= J
o r
-
W
mm-
PLo-
+- m U
E
*-
I
-'m
1.58
1.68
1.63
1 .65
1.68
1.78
+ e i \ g
2
5
4
5
5
l B B O / T . DEG K Figure 7. Reaction rate vs. 1/T, H~-CO-N~O/RU. Data points adjusted to: 2.0%Ha, 2.0% CO, 0.5% NzO. Lines are obtained from eq 9 and Table 11.
co Figure 9. Reaction rate versus % CO, H2-CO-N20/Ru. Data points adjusted to: 2.0%H2,0.5%N20,320 "C. Lines are obtained from eq 9 and Table 11.
a lack of data at high (400-600 "C) temperatures and at low NO concentrations ( E1 and E4 > E2 coupled with the ternary N20 results in Table I1 where E6 > E7 suggests that Hz will be the dominant reducing agent a t higher temperatures, independent of the water gas shift reaction. I t can be easily seen from concentration dependency plots shown in Figures 4-6 and 8-10 or from the power law exponents in Table I1 that NO inhibits the rate of NO reduction whereas CO does not. Similarly, CO inhibits the rate of N20 reduction whereas Hz does not. As with the binary results, the relative order of inhibition found in the ternary results is quite consistent with the hypothesis that the order of adsorptive
bond strength of these reactants on Ru is NO > CO > H2 > N 2 0 . Finally it should also be mentioned that although the rate of ternary NO reduction is greater than that for NzO a t 300 "C, and although E N ~ > O EN^, N20 is still expected t o disappear as a measurable intermediate product a t much higher temperatures because of the much higher activation energies associated with N20 reduction. In summary, the results of this study provide a quantitative measure of the effects of temperature and feed concentrations upon NO reduction in the presence of both Hz and CO over fiber glass supported ruthenium. I t is not known whether these results can be extrapolated much above 400 "C. I t is also evident that a t very low NO concentrations the rates of the NO reduction reactions will not increase without bound as predicted by these correlations.
Acknowledgment The financial help of PPG Industries for one of the authors (R.S.) is gratefully appreciated. We are also very grateful to Dr. I. Zlochower for the preparation of the catalyst support and many valuable discussions and suggestions. Nomenclature a, b, c, d = power law model exponents for NO, CO, HP, and NzO, respectively, eq 9, dimensionless E = Arrhenius activation energy, cal/g-mol K , = equilibrium constant for water gas shift reaction (8), dimensionless ho = frequency factor, eq 9, g-mol/min/g of cat. P = partial pressure, atm R = universal gas constant, 1.987 g cal/g-mol/K r = reaction rate, g-mol/min/g of cat. s = standard deviation of % error in rate correlations, % T = temperature, K Literature Cited Bauerle, G. L., Nobe, K., Ind. Eng. Chem., Prod. Res. Dev., 13, 185 (1974). Bauerle, G.L., Service, G. R., Nobe, K., Ind. Eng. Chem., Prw! Res. Dev., 11,
54 (1972). Klimisch, R . L., Barnes, G. J., Environ. Sci. Techno/., 6, 543 (1972). Klimisch, R. L.. Taylor, K. C., Environ. Sci. Techno/., 7, 127 (1973). Kobylinski, T. P., Taylor, B. W., J. Catal., 33, 376 (1974). Otto, K., Shelef, M., Z.Phys. Chem. (Franankfu~amMain),85, 308 (1973). Schleppy, R.. Jr., PhD Thesis, University of Pittsburgh, Pittsburgh, Pa., 1976. Schleppy, R., Jr.. Shah, Y. T., Ind. Eng. Chem., Prod. Res. Dev., 15, 172
(1976a). Schleppy, R., Jr.. Shah, Y. T., Chem. Eng. Sci. (in press) (1976b). Shelef, M., Gandhi, H. S., Ind. Eng. Chem., Prod. Res. Dev., 11, 393 (1972). sugi, y.,Todo, N., Toshio, S.,Bull. Chem. SOC.~ p n .48, , 337 (1975). Taylor, K. C.,Klimisch, R. L., J. Catal., 30,478 (1973). Voorhoeve, R . J. H., Trimble, L. E.. J. Catal., 38, 80 (1975).
Receiued f o r reuiew October 26, 1976 Accepted December 13, 1976 Ind. Eng. Chem., Prod. Res. Dev., Vol. 16, No. 1, 1977
51