The several isothermal examples treated in this paper should be sufficient illustration of both the power and relative sim- plicity of the recycle model as a device whereby complex, yield-sensitive, reaction networks may be profitably analyzed as regards the influence of backmixing upon yield. Other systems of interest-e.g., equilibrium reactions, Langmuir- Hinshelwood, Hougen-Watson rate equations-may also be treated in the fashion outlined here.

For the adiabatic reactor, the recycle concept extends the Douglas-Eagleton adiabatic PFR analytical solution to a domain free of the plugflow restriction. Backmixing is not, per se, always deleterious to yield ( 7 7 ) ) nor does conversion necessarily suffer under its influence, as the adiabatic reactor analyses presented here demonstrate.

The recycle model does reduce to the nonsegregated, micro- mixedness condition for nonlinear reaction kinetics, as R is increased to values above 20. As Eagleton notes, since the PFR ( R = 0) is totally segregated, it is likely that intermediate values of R physically describe a partially segregated system. For linear kinetics, this is of no consequence, while in cases involving nonlinearity the distinction is of import. Thus the results set forth here are qualitatively sound for nonlinear kinetics and quantitatively sound for linear systems.

Application of the model is now being extended to adiabatic yield sensitive schemes as well as to emulsion phase mixing in fluidized bed reactors in which the pumping action of the bubble phase might be a suspected agent in promoting actual emulsion phase recycle.

Acknowledgment

We are indebted to L. C. Eagleton for his invaluable review of the manuscript and, particularly, for bringing the work of G. R. Worrell to our a?te:ntion.

Nomenclature

A , B, C, D = molecular species, or concentration = heat capacity = axial mixing coefficient

= exponential = reduced concentration, A / A , = total flow rate to PFR = Q + q

c, D, E = activation energy

F g = functiona1,ity -AH = reaction enthalpy change K = rate constant ratio k = rate constant L = reactor length

rexp

n = number of CSTR’s P Q q R = recycle ratio, q / Q

= functionality = volumetric flow rate to recycle system = recycle flow rate

= gas constant = temperature

U = velocity in reactor

x = distance along reactor

RT. V = reactor volume

X, Y, Z CR = rate of reaction a = 4-K P o e = holding time, V / Q Y

P = density x = axial thermal conductivity

SUBSCRIPTS

= defined by Equation 26

= defined by Equation 24

= defined by Equation 26 7 = holding time, V / Q + q

0 = initial or system feed condition = feed condition to PFR 1

literature Cited

(1) Almasy, G., Acta Chim. Acad. Sci. Hung. 24, 197 (1960). (2) Zbid., 25, 243 (1960). (3) Benson, Sidney, “Foundations of Chemical Kinetics,” p. 43,

McGraw-Hill, New York. 1960. (4) Carberry, J.’J., A . Z. Ch: E . J . 4, 13 M (1958). (5) Carberry, J. J., Can. J . Chem. Eng. 36,207 (1958). (6) Carberry, J. J., Wendel, M. M., A . Z. Ch. E. J . 9, 129 (1963). (7) Cholette, A., Can. J . Chem. Eng. 39, 192 (1961). (8) Coste, J., Amundson, N. R., Rudd, D., Zbid., 39, 149 (1961). (9) de Maria, F., Longfield, J. E., Butler, G., Znd. Eng. Chem. 53,

(10) Douglas, J. M., Eagleton, L. C., IND. ENG. CHEM. FUNDA-

(11) Gillespie, B. M., Carberry, J. J., Chem. Eng. Sci., 21, No. 5,

(12) Kramers, H., Westerterp, K. R., “Chemical Reactor Design

(1 3) Levenspiel, O., “Chemical Reactor Engineering,” Wiley,

(14) Parts, A. G., Australian J . Chem. 11, 251 (1958). (15) Smith, J. M., “Chemical Engineering Kinetics,” p. 128,

(16) Wehner, J. F., Wilhelm, R. H., Chem. Eng. Sci. 6,89 (1956). (17) Worrell, G. R., Ph.D. thesis, University of Pennsylvania,

259 (1961).

MENTALS 1, 116 (1962).

(May 1966).

and Operation,” Academic Press, New York, 1963.

New York, 1964.

McGraw-Hill, New York, 1956.

1963. (18) Worrell, G. R., Eagleton, L. C., Can. J . Chem. Eng. 42, 254

(19) Zwietering, T. N., Chem. Eng. Sci. 11, l(1959). (1964).

RECEIVED for review July 6, 1965 ACCEPTED February 3, 1966

Work supported in part by a grant from the National Science Foundation.

HETEROGENEOUS CATALYSIS IN A

CONTINUOUS STIRRED TANK REACTOR D . G . T A J B L , ’ J . B . S I M O N S , A N D J A M E S J . C A R B E R R Y

Department of Chemical Engineering, University of Notre Dame, Notre Dame, Znd.

THE procurement of precise kinetic data for solid-catalyzed gaseous reactions poses a number of problems in that

transport phenomena (interparticle and inter-intraphase) often intrude upon the surface reaction, tending to falsify the data and thus frustrate both the physical chemist seeking surface

rate laws and the chemical engineer who seeks surface rate models which become bases for reactor scale-up.

The often employed integral catalytic reactor can rarely be operated isothermally and differentiation of integral reactor data further complicates analysis. The differential reactor, operating a t extremely small conversion levels, does provide

1 Present address, Institute of Gas Technology, Chicago, 111. point rate data : However, extremely precise analytical

VOL. 5 NO. 2 M A Y 1 9 6 6 171

A continuous stirred tank catalytic reactor has been designed, evaluated as a perfect mixer, and employed to determine the kinetics of CO oxidation with oxygen as catalyzed by 0.5 weight % Pd on a-alumina pellets. The rate is found to be proportional to the Oz/CO ratio and the apparent activation energy is 28.5 kcal. per mole in the temperature range of 205” to 234” C. at a total pressure of 1 atm.

methods are then required if precise rate laws or models are to be derived from the raw data. Clearly, the ideal laboratory catalytic reactor is one which operates isothermally (negligible interparticle, interphase, and intraparticle gradients of heat and mass) while delivering integral reactor conversions. The continuous recycle reactor (8) and its equivalent, the continu- ous stirred tank catalytic reactor (CSTCR) provide, in princi- ple, the transport-gradient-free character of the differential reactor and, a t the same time, permit operation at finite (integ- ral reactor) conversion levels (3). Trotter and Wilhelm (72) employed a reactor capable of utilizing fluidizable catalysts, while Ford and Perlmutter (6) designed one in which the reactor wall itself served as the catalytic surface. Our concern is directed toward a reactor which will accommodate com- mercial size pellets and extruded catalysts (3).

The design, construction, and operation of such a catalytic CSTR are described in this paper. The region of operation (r.p.m. and flow rate) in which “perfect mixing” prevails was specified by pulse testing and kinetics of palladium-catalyzed oxidation of CO with 0 2 determined over a temperature range of from 200” to 234’ C. Catalysis by other transition metals was also studied and will be reported later ( 7 7 ) .

Details of Reactor

The continuous stirred tank catalytic reactor, constructed of stainless steel, is a cylindrical vessel, 3 inches in i.d. and 3 inches high, provided with inlet and effluent ports and a cooled agitator shaft and seal as detailed in Figure 1. Four vertical baffles are placed 90” apart on the inner wall. The catalyst pellets are contained in a single layer, in four thin baskets constructed of low mesh stainless steel screen, as shown in Figure 2. The four-basket unit (each basket set a t a right angle to the other) is affixed to the agitator shaft. Two single propellers, one above and the other below the catalyst basket unit, are also secured to the shaft. In operation, the feed gas enters beneath the rotating shaft and leaves, above the swirling catalyst, a t the top of the reactor just off-center from the shaft seal. A thermocouple well is situated within the reactor. In operation, the catalyst pellets are swept through the con- tinuously fed reacting gas stream.

Mixing Characteristics of Reactor

To exploit the lumped parameter advantage of the CSTCR, it must be demonstrated that the reactor behaves as a perfect mixer. The mixing characteristics of the reactor described above were determined by measuring the response of the non- reacting system to a pulse input of helium injected into a steadily flowing air stream at various volumetric flow rates and agitation speeds. Our interest was confined to a determina- tion of the operating region (agitator speed, r.p.m., and feed rate) in which apparent perfect mixing prevailed as judged by obedience of the effluent pulse concentration to the theoretical response predictable for a single CSTR. In Figure 3 are shown sample data points secured in the pulse testing study. The solid line is the theoretical response for perfect mixing. The term “perfect mixing” is used here in the phenomenologi- cal sense, as no attempt was made to investigate the micro scale character of mixing within the reactor. The results of the mixing study for a wide range of agitation and feed rates are presented in an operating diagram for our particular reactor in

Figure 4. The cross-hatched zone designates the apparant transition boundary between the perfectly mixed regime of operation (above the boundary) and that region (below the boundary) wherein the pulse response data deviated from the theoretical prediction. Based upon this information, the CSTCR is always operated at an agitator speed above 1600 r.p.m., thus permitting the maximum variation in flow rate (contact time).

Kinetic Study

To demonstrate the operability of the CSTCR, the palla- dium-catalyzed oxidation of C O with 0 2 was selected as a worthy test system because the reaction is highly exothermic, the system is of interest per se, and CO oxidation merits atten- tion in view of apprehensions regarding smog and its abate- ment by oxidation of automotive exhaust fumes. The fact that product analysis poses no major difficulties also inspired selec- tion of this oxidation.

Prior Work

Surprisingly little systematic research has been focused upon A recent review (4) CO oxidation as catalyzed by palladium.

,&&E-- Pul l e y S h e a v e

G l y c e r o l Sea - B e a r i n g C o o l a n t

T e f l o n B e a r i n g

,Product g a s

P r o p e I I o r

I - 1 Figure 1 . Continuous stirred tank catalytic reactor

A g i t a t o r S h a f t

A%

Figure 2. Detail of catalyst basket unit

172 I & E C F U N D A M E N T A L S

dCO 0 2 - k p dt

cites the work of Schwab ( 9 ) in which kinetics were formulated for CO oxidized by 0 2 over a palladium wire. Schwab found

which model might suggest that the rate-controlling step is oxygen chemisorption. More formally

(2) kOz - dCO - __ -

, i t (1 + kCO)

which, for high adsorption of CO, reduces to Equation 1. Schwab‘s observed rate law may also suggest surface reaction between oxygen and CO as rate-controlling. For if CO is far more strongly absorbed than 0 2 , then

dCO kOz.CO (3) - - __ -

c’t (1 + K C 0 7

which obviously reduces to Equation 1 a t high values of KCO (KCO>>l). In another study, Modell (7) measured the oxidation kinetics over Pd and found a rate law similar to Schwab’s.

In conflict with Schwab’s results, Eley and Daglish (5) found, for the reaction catalyzed by pure Pd (lOOyo), the rate equation

I l ! R l

Figure 3. T3ypical pulse-testing data in mixing study

l 2 w I I I I 1 I

P e r f e c t

M i x i n g

\ / ”\

0 400 8 0 0 12oo 1600

A g i t a t o r S p e e d , RPM.

Figure 4. catalytic reactor

Operating diagram for continuous stirred tank

Regime in which perfect mixing occurs CIS inferred from mixing srwdier

(4)

which might suggest the dissociative chemisorption of 0 2 as rate controlling.

Pertinent to these kinetic studies is the recent work of Stephens (70), who studied the CO-Oe system in the presence of Pd metal film. The very high absorptivity of CO on Pd was established in this work, in conformity with the general implica- tions of Equations 1, 2, 3, and 4. Chemisorption rates of 0 2 on Pd in the presence of CO are not to be found, as reaction is then inevitable, thus frustrating measurements of adsorption per se. O n the other hand, 0 2 adsorption rates in the absence of C O would be of questionable value, as the Pd surface is clearly affected, with respect to its adsorptive characteristics, by the presence or absence of adsorbed CO.

Thus while prior studies shed light upon CO oxidation kinetics on wires and films, no data exist on Pd deposited upon a commercial support. Aside from the practical interest associated with supported catalysts, a reasoned interest exists among catalytic chemists in the relation, if any, which may exist in the catalytic activity (kinetics) of a given metal as used in diverse forms, such as a film, wire, and a deposited entity.

Description of Catalyst

In this particular phase of the work we employed 0.5 weight 70 of Pd deposited upon 0-alumina in the form of 3/16-inch cylinders, as freely supplied by M. Arnold, The Girdler Catalyst Corp. The method of Pd deposition \\as such that the Pd is superficially deposited upon the shell of the 3j16-inch alumina pellet. The actual depth of Pd pene- tration into the cylinder is less than 1 mm.

The reported BET surface area of the catalyst pellet here employed is 7 to 10 sq. meters per gram and the pore volume is 0.04 cc. per gram.

Preparations and Procedure

Reactants C O and 0 2 are each metered individually using soap-bubble meters and then mixed, dried, and passed into the CSTCR. An effluent bypass permits sampling of the stream for analysis in a Fisher partitioner (Model 25V). The partitioner resolves 0 2 , N2, CO, and CO?. The reactor (Figure 1) is heated by electrical heating tape, temperature control being maintained within 1’ C. by a Thermolyne step- less input control.

Flowmeters were calibrated on a frequent basis, as was the partitioner, using known mixtures.

A run was initiated by simply admitting the reactant mix- ture to the CSTCR, allowing time for both flow and catalyst equilibration, and then sampling the effluent for CO? deter- mination. Temperature, a function of extent of reaction and feed and effluent heat capacities as well as heater input, was adjusted to a constant and knonn value prior to sampling for rate determination. For the CSTCR the rate is simply equal to the difference between input and exit mole quantities (of COz) divided by holding time determined by total flow rate and reactor volume. For a solid-catalyzed reaction, rates are more appropriately expressed in terms of moles per time, grams of catalyst, and are so reported here.

Analysis of Data

For the reaction

the rate of COZ formation in a well stirred, continuously fed reactor is

VOL. 5 NO. 2 M A Y 1 9 6 6 173

for no COZ in the feed, where Q is the effluent flow rate, Vis the free reactor volume, (CO,) is concentration, and p is grams of catalyst per unit volume of reactor. Q is clearly a simple func- tion of reactor temperature, pressure, and extent of conversion. Analysis of partitioner samples provided COZ concentration (found to be linear in peak height) which, since C O and 0 2

analyses were also provided, agreed with reactant consumption data within i l % .

For a given temperature, r , the rate as determined by Equa- tion 6 was measured as a function of effluent and, therefore, reactor concentrations of CO, 0 2 , and CO,; these species variations were realized by contact time and/or feed com- position variation for a given temperature. In this study conversion varied from 2 to 1570, in the temperature range 200' to 234OC. The amount of catalyst used was 10 grams (100 pellets). In a few runs agitation rate was purposely re- duced to zero and the reaction rate was then observed to drop, suggesting reactant bypassing-Le., nonideal mixing.

Results

As shown in Figure 5, a t several temperature levels of opera- tion, the measured rate proved to be linear in the O2jCO ratio. This arithmetic plot represents rate of reaction as a function of 0 2 / C O ; thus the slopes provide the rate constant, k . In Figure 6, an Arrhenius plot of log k us. 1 / T is displayed, from which an apparent activation energy of 28.5 kcal. per mole is derived.

The results are compared with those of other investigations in Table I.

Discussion

I t is impossible to suggest an unambiguous surface rate model on the basis of the kinetic equation found to fit CO oxidation by 0, as catalyzed by supported Pd. Clearly the same kinetic form prevails for Pd fashioned as a wire filament and as a metal deposited (and dispersed) upon a support. Small differences in activation energy are apparent as a function of the form assumed by the catalyst.

Diffusional influences were apparently absent from our study. Specifically, intraparticle mass and thermal diffusional gradients would be determined by a modified Thiele modulus in which the actual catalyst "thickness" is less than 1 mm., although the support is "16 inch in diameter (and length). Calculations demonstrate that the appropriate Thiele modulus for the highest rate encountered is less than unity, assuring an effectiveness of virtually unity a t the highest rate of reaction

Interphase gradients are subject to firmer assessment, for the CSTR operates a t a uniform flux of mass and heat. In con- sequence, if interphase heat and mass transport coefficients can be calculated, interphase concentration and temperature differences may be directly calculated ( 7 ) . Values of h and k, can be estimated, since the position and agitator speed allow an estimate of particle velocity through the gas. Precise deter-

(2).

Table 1. Results

Investigator Catalyst C. E, Kcal. T:mP.,

Schwab ( 9 ) Pd wire 250-320 22,2" Eley and Daglish ( 5 ) Pd wire 95-130 28.3b Model1 (7) Pd wire 140-150 28.7" This work Pd on a-alumina 200-234 28.5" zk 2.7

a Based upon rate model, k.Oz/CO. Based upon rate model, k.On/ (Coy .

m 0

X L

-

I / I

3

0 I 02/20

Figure 5. CO ratio at various temperature levels

Reaction rate as a function of 0 2 /

0.

+ NC6

0.0 Y

0 . 0 0 1 ~

I

\ E,= 2 8 . 5 k c a l . 31

L 2 .05 2.1 2 15

Figure 6. Specific rate constant, k vs. 1 / T o K., in CO oxidation over Pd on a-alumina

mination of this velocity is not possible, as the pellets are not swept through a "still" gas, but rather a gas environment in unknown motion. Nevertheless, a minimum interphase AT can be computed a firiori and, as shown in Figure 6 for two of the 234' C. runs, finite temperature differences were calculated to exist between particle surface and the gas, thus accounting for the nonlinearity of these higher rate data secured a t 234OC. Given the high activation energy of nearly 30 kcal., even small interphase A T ' S can significantly enhance the rate over the values anticipated on the basis of fluid phase temperature measurements.

Conclusions

A continuous stirred tank catalytic reactor (CSTCR) has been designed, constructed, evaluated as a perfect mixer, and

174 l & E C F U N D A M E N T A L S

demonstrated to be an effective device for the procurement of catalytic rate data.

Specifically, the oxidation kinetics of CO as catalyzed by supported Pd in the presence of 0 2 has been studied and the rate equation and apparent activation energy have been de- termined.

Because of the uniform flux resulting from the perfect mixing realized in thc CSTCR, interphase temperature and concen- tration gradients can tle readily estimated a priori for a given reaction rate, enthalpy change, and agitator-catalyst r.p.m.

Nomenclature C = concentration h k = rate constant K = adsorption coefficient k , Q = volumetric flow rate r.p.m. = agitator speed, revolutions per minute r = rate of reaction AT = interphase temperature difference V = reactor volume P

= interphase heat transfer coefficient

= interphase mass transport coefficient

= grams of catalyst/volume of reactor

literature Cited

(1) Carberry, J. J., A . Z . Ch. E. J . 6,460(1960). (2) Ibid., 7 , 350 (1961). (3) Carberry, J. J., Znd. Eng. Chem. 56, No. 11, 39 (1964). (4 ) Dixon. J. K.. Longfield. J. E.. Catalysis 7 . 281 (1960). (5) Eley, ,D. D.,‘ Dadish, A. G.,‘ 2nd ‘Interkational Congress on

(6) Ford, F. E., Perlmutter, D. D., Chem. Eng. Sci. 19, 371 (1964). (7) Modell, M., D. Sc. thesis, Massachusetts Institute of Tech-

Catalysis, Vol. 11, pp. 1615-24, 1961.

nology, 1964. (8) Perkins, T. K., Rase, H. P., A . I . Ch. E. J . 4, 351 (1958). (9) Schwab, G. M., Gossner, K., Z. Physik. Chem. Neue Folge 16, . . . . - .

39 (1958): (10) Stephens, S. J., J . Phys. Chem. 63, 188 (1959). (11) Tajbl, D. G., Ph.D. thesis, University of Notre Dame, 1966. (12) Trotter, Ide, Wilhelm, R. H., forthcoming publication.

RECEIVED for review July 16, 1965 ACCEPTED February 1, 1966

Work supported by a Grant from the Petroleum Research Fund of the American Chemical Society. Grateful acknowledgment is made to the donors of the fund. The kinetic studies are a part of the dissertation of D. G. Tajbl submitted to the Graduate School, University of Notre Dame, in partial fulfillment of the require- ments for the degree of doctor of philosophy.

RATE AND

0x1 DATION

MECHANISM OF GAS-PHASE

0 F PARTSPER-M I LLlO N CONCENTRATIONS OF NITRIC OXIDE M I L T O N E . M I O R R I S O N , l R O B E R T G . R I N K E R , * A N D W I L L I A M H . C O R C O R A N

California Institute of Technology, Pasadena, Calif.

Rdtes of the air oxidation of parts-per-million concentrations of nitric oxide were studied homogeneously at atmospheric pressure and ambient temperatures in a constant-volume batch reactor. The initial concen- tration of nitric oxide was varied from 2 to 75 p.p.m., while the oxygen concentration ranged from 3 to 2 5 volume 9;. The initial order of the oxidation reaction in the absence of nitrogen dioxide was deter- mined to be ;!.OO =k 0.09 for nitric oxide and 0.97 f 0.1 1 for oxygen. From initial rate data at 26.5” C., a third-order rate constant of (1.297 * 0.051) X 1 O4 (liter)2/(g. mole)2(sec.) was obtained. The addition of nitrogen dioxide increased the initial oxidation rate, and that compound showed an autocatalytic effect throughout the course of the reaction. A nonlinear least-squares analysis was used to develop a mechc nism involving six reactions, with NO$, Nz03, and NzOs as intermediates. Use of that mechanism gave a minimum standard deviation of 1.6 p.p.m. for the predicted concentrations of nitric oxide relative to the experimental data.

OLLUTION of the air in metropolitan areas is an increasing problem. Nitric oxide and unburned hydrocarbons,

formed by high-compre.ssion, internal-combustion engines and industrial plants, are of major concern in air pollution. Nitric oxide reacts in the atmosphere with molecular oxygen to form nitrogen dioxide, which oxidizes hydrocarbons photochemically to ketones, aldehydes, and alcohols. These compounds, nitro- gen oxides and oxidized hydrocarbons, are the constituents of so-called “smog” and are detrimental to the health of both the human and plant population. The basic chemical reactions

Present address, E. I. du Pont de Nemours & Co., Inc.,

Present address, University of California, Santa Barbara, Chattanooga, Tenn.

Calif.

which occur in these atmospheres, however, are not well under- stood. Because of the small amount of quantitative work which has been carried out on the air oxidation of nitric oxide in parts-per-million concentrations (p.p.m., defined as mole fraction X 106), rates of reaction of nitric oxide with oxygen were measured, and a mechanism is proposed.

The mechanism of the oxidation of nitric oxide to nitrogen dioxide has been a subject of controversy ever since Raschig (20) found that the reaction was third order. He reported that the oxidation of nitric oxide was second order in nitric oxide and first order in oxygen. A number of other investiga- tors (3-5, 76,24,28, 33) after Raschig also studied the system. Although they generally agreed that a t total pressures below 50 mm. of mercury the reaction was second order in nitric oxide

VOL. 5 NO. 2 M A Y 1 9 6 6 175