The Oxidation of Carbon Monoxide on Supported Platinum

A STUDY OF HONEYCOMB CATALYST REACTION KINETICS

By R. C. Shishu Ford Motor Company, Dearborn, Michigan and Professor L. S. Kowalczyk Chemical Engineering Department, University of Detroit

Reliable kinetic rate data have not previously been available for the oxidation of carbon monoxide on platinum catalysts used in the control of automobile exhaust emissions. This paper discusses results of studies using an isothermal diferential catalyst bed and confirms that mass transfer and pore difusion did not control the reaction rate, so that the observed kinetics represent the true surface reaction rates.

A number of investigations on oxidation isothermal reactor. In this paper only the of carbon monoxide over platinum catalyst first phase will be discussed. The second have been reported (1-8). The conclusions phase involving a non-isothermal bed will be of various investigators regarding the kinetics reported later. of this reaction are rather conflicting. Accord- A schematic diagram of the experimental ing to one group both carbon monoxide and apparatus is shown in Fig. I. The flow rates oxygen must be adsorbed on the catalyst of nitrogen carrier gas, oxygen and diluted surface to form the reaction product. On the carbon monoxide were measured with cali- other hand, others contend that only one brated flowmeters. The gases were passed reactant need be adsorbed on the catalyst through a stainless steel manifold used as a while the other must originate from the gas mixing chamber and then through a master phase. There is general agreement, however, flowmeter to measure the total flow rate. that the oxidation mechanism changes with The composition of the feed gas was deter- temperature. At low temperatures the reac- mined by passing the gas mixture through tion is inhibited by carbon monoxide (I). carbon monoxide and oxygen analysers. Once At higher temperatures the reaction is limited the composition was determined, the gas was largely by bulk gas-phase diffusion (1,3,5). passed through the catalyst bed placed in a The temperature at which this occurs lies quartz tube reactor. At the steady state the between 572°F (300°C)(3) and 960°F (545°C) extent of conversion was measured. A (1). Beckman model IR-215 non-dispersive infra- The purpose of this investigation was to red analyser wTas used to analyse carbon obtain a reliable kinetic rate expression for monoxide. Oxygen was measured by a carbon monoxide oxidation on a platinum Beckman Model F3 paramagnetic analyser. catalyst in an isothermal differential bed over Platinised honeycomb having the charac- the concentration ranges prevailing in auto- teristics listed in Table I was used for this mobile exhaust gases, and to demonstrate study. It was difficult to prepare test samples the validity of this rate expression for a non- of the honeycomb with a round cross-section

Platinum Metals Rev., 1974, 18, (2), 58-64 58 VENT

PRESSURE GAUGE BY-PA 5 S MASTER ROTAMETER

HOL CO AIR GAS

N2 02 GAS SPAN GAS Fig. 1 Diagram nf the apparatzbs used to study the oxidation of carbon monoxide. The platinised catalyst was mounted in the reactor, which is shown in more detail in Fig. 2 on the next page

such that the sample would fit in a round tube furnace was used to preheat the gas mixture snugly enough to avoid appreciable gas by- and to maintain the catalyst section at the passing near the walls. Therefore, a square desired temperature level. quartz tube reactor, shown schematically in The catalyst bed was a thin slice (& in deep) Fig. 2, was used. The pre-heating and after- of platinised honeycomb. Two inert support cooling sections were filled with an inert cubes of honeycomb channel were used to packing of pyrex beads (3 mm diameter) and provide calming zones before and after the vermiculite. This packing was found to be catalyst section. Temperatures at several inert in a previous study. A Lindberg locations on both faces of the catalyst were measured to ascertain that essentially isothermal conditions were obtained. Since the Table I chemical reaction takes place on the surface Characteristics of Platinised Honeycomb of the catalyst, the thermocouples were Catalyst cemented to the surface. The cement was found to be inert for carbon monoxide Weight 0.0026 I b oxidation. Chromel-Alumel thermocouples were used. Three thermocouples were used Dimension xEx$in on each face: one at the centre, one at an edge Pt content 0.14% (by weight) and one at a corner (see Fig. 3). Isothermal Support material Mullite coated with conditions were maintained in the reactor alumina both axially and radially within f3”F. Support type Hexagonal channels A flow rate was established such that film Nominal channel size $ in diffusion was not rate limiting. The initial Hydraulic mean concentration of carbon monoxide was varied diameter of channel 0.133 in from 0.2 to 2.0 per cent while that of oxygen Bulk density 40.8 Ibjft” was varied from 0.13 to 2.1 per cent. Several Void volume 55% possible combinations of carbon monoxide Est. geometric surface 300 ft2/ft3 and oxygen concentrations were studied. BET surface area 12 m21g These correspond to rich, lean and stoichio- metric conditions. The lowest temperature

Platinum Metals Rev., 1974, 18, (2) 59 QUARTZ SPHERE + VERMICULITE PACKING COOLING JACKET \ \ WATER + PRODUCTS

I REAC TANT 5 THERMOCOUPLES Fig. 2 The honeycomb catalyst sample was di&cult to prepare with a round cross-section to jit a cylindrical reactor, and so a reactor with a square cross-section was fabricated from quartz

studied for each set of carbon monoxide and The correlation of the reaction rate oxygen concentrations corresponded to that data was attempted with a simple power- temperature which resulted in the lowest law rate model and a number of reaction detectable conversion. The upper tempera- rate expressions based on Langmuir- ture was limited to that which allowed main- Hinshelwood’s mechanism. The model tenance of the isothermicity of the bed. parameters were estimated using a non-linear Temperatures higher than that yielded least squares technique suggested by higher conversions, and rendered the bed Levenberg, Meiron and Fraser (11, 12, 13). non-isothermal because of high heat of reaction. The range of temperature studied was 460 to 800°F. r5 Experimental data obtained with the isothermal differential reactor are given in Table 11. This includes only those data which have a maximum conversion of carbon monoxide up to 10per cent so that they could be treated as differential bed data for the analysis. The complete set of experimental data is given elsewhere (9). The rate of carbon monoxide oxidation was calculated from the design equation, -Ax A(W/F) (1) where r is the rate of reaction, AX is the conversion, W is the catalyst weight and F Fig. 3 Three thermocouples were used on is the molal flow rate of carbon monoxide. each face of the catalyst bed, which consisted of a & in. deep and x 9 in. square slice Calculations similar to those made by Potter of platinised honeycomb. Thermocouples 1 and Baron (10) indicated that mass transfer and 2 were $xed to the centres of each face, thermocouples 3 and 4 were $xed on either effects were negligible. Estimated values of side of one corner and thermocouples 5 and Thiele’s modulus indicated that pore diffu- 6 were fixed on either side of one edge sion hindrance was also negligible.

Platinum Metals Rev., 1974, 18, (2) 60 Table II Experimental Results for Catalytic Oxidation of Carbon Monoxide on Platinum Catalyst-Isothermal Differential Bed Data Table II Experimental Results for Catalytic Oxidation of Carbon Monoxide Run on Platinum Catalyst-Isothermal Differential Bed Data

17 35.8 0.40 1.O Total5.7 Reactor1.07 Bed502.1 Flow Rate Feed Composition Conversion Pressure Tem peratu re Run19 35.8 0.41 !.O 8.4 1.07 523.5 SCFH %CO %O* % atm "F 74 35.7 1.04 1.97 3.4 1.08 504.3 17 75 35.835.7 0.401.04 1.O 1.97 5.7 4.4 1.071.08 502.1524.4 19 85 35.844.9 0.410.21 !.O 0.50 8.46.8 1.071.13 523.5494.8 74 79 35.743.9 1.041.02 1.97 2.12 3.46.4 1.081 .I1 504.3548.6 75 89 35.744.2 1.041.02 1.972.0 4.4 7.4 1.081.13 524.4555.3 85 90 44.944.2 0.211.02 0.502.1 6.84.4 1.131.13 494.8531.7 79 25 43.935.7 1.020.21 2.12 0.53 6.4 8.7 1 .I11.08 548.6481.6 89 26 44.235.7 1.020.22 2.0 0.55 7.4 7.0 1.131.08 555.3479.0 90 32 44.235.7 1.020.20 2.1 0.51 4.4 5.5 1.13I .08 531.7461.4 25 7 35.726.8 0.210.40 0.530.21 8.7 7.2 1.081.05 481.6581.2 26 33 35.726.7 0.220.40 0.550.21 7.0 5.2 1.081.04 479.0577.2 32 57 35.735.7 0.201.98 0.511 .o 5.5 3.3 1.081.08 461.4620.1 7 58 26.835.7 0.401.98 0.211 .o 7.2 4.6 1.051.08 581.2634.9 33 13 26.735.6 0.400.39 0.210.2 5.27.6 1.041.07 577.2601.3 57 92 35.735.9 1.980.39 1 .o 0.2 3.3 7.1 1.081.09 620.1594.2 58 67 35.715.0 1.981 .o 1.o 0.1 3 4.6 3.5 1.081.01 634.9653.3 13 68 35.615.C 0.391 .o 0.2 0.13 7.6 6.5 1.071.01 601.3714.0 92 64 35.915.0 0.392.0 0.2 0.21 7.1 5.0 1.091.01 594.2719.1 67 65 15.015.0 1 .o 2.0 0.1 0.213 3.5 5.7 1.011.01 653.3753.2 68 66 15.C15.0 1.o 2.0 0.130.21 6.5 7.5 1.011 .Ol 714.0790.0 64 54 15.026.8 2.0 2.0 0.210.22 5.0 3.0 1.011.05 719.1708.5 65 55 15.026.8 2.0 2.0 0.210.22 5.7 5.0 1.011.05 753.2764.5 66 56 15.026.8 2.0 2.0 0.210.24 7.5 6.5 1.Ol 1.05 790.0804.0 54 26.8 2.0 0.22 3.0 1.05 708.5 55 26.8 2.0 0.22 5.0 1.05 764.5 Power-law56 26.8 Model 2.0 0.24 vestigators6.5 (8). An1.05 Arrhenius plot804.0 is shown All the data were correlated by the rate in Fig. 4. A graphical comparison of predicted expression and observed conversion is shown in Fig. 5. (2) Power-law Model vestigatorsIt has been (8). An postulated Arrhenius that plot the is CO shown in- =2.25 r05 e-ZHH~OIRT kmoles CO/h.lb cat. All the data were (Po,)correlated " by the rate inhibition Fig. 4. Aeffect graphical on the comparison reaction rate of predictedis due to expressionwhere the standard deviation is 1.87 x IO-~ andfast observedand strong conversion adsorption is of showncarbon in monoxide Fig. 5. moles CO/h.lb cat. The orders of reaction(2) thatIt hasinhibits been the postulated oxygen adsorption. that the CO Several in- =2.25 r05 e-ZHH~OIRT &%.- moles CO/h.lb cat. of 1.0and -0.5 with(Po,) respect " to 0, and CO hibitionpossible effect rate controllingon the reaction steps ratecould is bedue pos- to whererespectively the standard are in agreementdeviation withis 1.87 other x IO-~ in- fasttulated. and strong The data adsorption of the present of carbon investigation monoxide moles CO/h.lb cat. The orders of reaction that inhibits the oxygen adsorption. Several of 1.0and -0.5 with respect to 0, and CO possible rate controlling steps could be pos- respectively are in agreement with other in- tulated. The data of the present investigation Platinum Metals Rev., 1974, 18, (2) 61

Platinum Metals Rev., 1974, 18, (2) 61 The power-law model, as shown in equation (2) has a limitation because of the rate being inversely proportional to the 0.5 power of partial pressure of carbon monoxide. This predicts an infinite reaction rate as the partial pressure of carbon monoxide approaches zero. Intuitively, one would expect that the reaction rate becomes zero as the partial pressure of carbon monoxide approaches zero. Therefore, the extrapolation of the model beyond the range of experimental conditions may be erroneous. Langmuir -Hinshelwood Model Attempts were also made to fit the data to Langmuir-Hinshelwood models, each of which is based on some plausible mechanism of reaction. Out of several possible models considered, a satisfactory correlation was achieved only by the rate equation shown in equation (3). (3) r- kKcoKozpcopo' moles C0jh.lb cat. (I + Kcopco + Ko,poJZ This model is generally referred to as the dual is not sufficient to specify indisputably a site model. Estimated parameters are : particular reaction mechanism' However> k ~ 2.31 Iog e--4S100/RT moles CO/h.lb cat. the same argument of fast and strong ad- Kco --2.16 &UfiOO'RT atm-l sorotion of carbon monoxide could be invoked. KO, = 1.30 x 10-I eZaR3OinTam-' where the standard deviation is 1.78 x IO-~ moles CO/h.lb cat. The logarithms of these constants were plotted against reciprocal absolute temperatures in Fig. 6. Fig. 7 compares calculated and experimental conversion. This model does not have the limitation as stated about the power-law model. The denominator remains a finite quantity as the partial pressure of CO approaches zero. The derivation of the dual-site rate equation has been given by Corrigan (14)and others, based on a model of surface reaction between two different adsorbed reactants on adjacent sites. However, it should be emphasised here that, in this study, engineering analysis of the data has been made and rate equation (3) should be considered an empirical rate model.

Platinum Metals Rev., 1974, 18, (2) 62 model failed to predict the behaviour of the non-isothermal bed under conditions of low initial carbon monoxide concentration and high conversion. The temperature and concentration profiles in the non-isothermal bed are predicted satisfactorily by the dual-site rate model. These results will be reported in detail in a later paper. Conclusions The experimental results and calculations show that for the conditions of the experiments and type of catalyst used in this work the bulk phase mass transfer and pore diffusion resistance do not control the rate of catalytic oxidation of carbon monoxide on platinum. Therefore the observed kinetics represent the true surface reaction As shown in Figs. 5 and 7, the agreement rates. In agreement with other investigators between experiments and calculations for the oxidation rate showed an inhibition effect both power-law and dual-site models is due to carbon monoxide. reasonably good, indicating satisfactory cor- The differential isothermal bed data could relation of the data. The use of these rate be correlated equally well by a simple models was made to predict the behaviour power-law model and by a Langmuir- of a non-isothermal reactor. The power-law Hinshelwood dual-site model. The orders of the reaction in the power-law model are +I.O with respect to 0, and -0.5 to CO. The power-law model has a limitation at low partial pressure of carbon monoxide (pco) and high conversion as the rate becomes infinite for pco approaching zero. The dual- site model does not have this limitation. Acknowledgements The authors thank R. L. Gealer and E. C. Su for several helpful discussions throughout the course of this work, and T. E. Sharp for assisting in writing the computer program. References i I. Langmuir, Trans. Faraday SOC.,1922, 17, 62 I 2 P. V. McKinney, J. Am. Chem. SOC.,1934, 56,2557 3 L. S. Solov’eva, Russ. J. Phys. Chern., 1960, 34, (61, 582 4 H. Heyne and F. C.Tomkins, Proc. Rqy. SOC., 1966,292A3 (1431),460

Platinum Metals Rev., 1974, 18, (2) 63 5 M. B. Syzdykbaeva, N. M. Popova and D. V. 9 R. C. Shishu, D.Eng. Thesis, University of Sokol'skii, Izv. Akad. Nauk. Kaz. S.S.R., Detroit, 1972 sere Khim., 1967, 17, (41, 37 10 C. Potter and S. Baron, Chem. Engng. Prog., 6 C. D. Scott, U.S. Rept. ORNL-3043, UC-4- 195r, 47Y 473 Chemistry II K. Levenberg, Q. Appl. Math., 1944,2,164 7 A. V. Sklyarov, I. I. Tret'yakov, B. R. Shub 12 J. Merion,J. Opt. SOC.Am., 1965, 55, 1105 and S. Z.Roginskii, Proc. Acad. Scb. U.S.S.R., 13 R. D. B. Fraser, Anal. Chem., 1966, 38, I770 Phys. Chem. Sect., 1969, 189, 829 14 T. E. Corrigan, Chem. Engng., 1954, 61 8 R. A. Sills, Thesis, M.I.T., I970 (Nov.), 236

A New Mining Area for Rustenburg

For many years now Rustenburg 14 vertical shafts are in operation as well Platinum Mines has been the largest as a great number of incline winzes. The producer of platinum and its allied metals new mine, to be known as the Amandel- in the world and successive stages of bult Section, lies some 30 kilometres to expansion have been reported here from the north-east of the Union Section. At time to time. Further moves to increase the present time three vertical shafts are capacity to over 1,500,000 ounces of being sunk to extract ore from the deeper platinum a year are now under way with areas as well as 21 incline winzes, one of the opening of a new mining area with an which is illustrated below, designed to additional capacity initially of 225,000 work the shallower parts of the deposit. ounces a year. The two mining areas at The treatment and refining plants of present being exploited are at Rustenburg Matthey Rustenburg Refiners are, of itself and at the Union Section about 100 course, being expanded to cope with the kilometres to the north, where altogether increased output from the mines.

Platinum Metals Rev., 1974, 18, (2) 64