I
R. H. PRICE and D. B. SCHIEWETZ' University of Cincinnati, Cincinnati, Ohio
Catalytic Liquid-Phase Hydrogenation This i s the first reporton a seriesof investigations that may expose new and profitable areas for industrial development in applied chemical kinetics B m A u s E OF THEIR economic advantage in continuous processing, much interest in applied chemical kinetics has centered on vapor-phase catalytic reactions. Concepts originating in this work, particularly re-emphasis on mass transfer, have become valuable tools for the engineer. Heterogeneous reactions in solution, however, although investigated in the laboratory for decades, have been exploited only to a limited extent. This is understandable-complexities of liquids and solutions themselves, along with obscure hydrodynamic features of batch systems, present real obstacles. Yet, stirred-tank reactors, for example, are widely used industrially, and there is need to extend kinetic concepts to such systems. This investigation, the first in a series for studying liquid-phase kinetics, is not intended to be specific on any facet of the problem, but rather to expose those areas requiring detailed study. With classical kinetic conce@s in mind, an experimental approach was employed wherein effects of all known variables were determined. No attempt is made to refer exhaustively to the extensive
Present address, Yerkes Research Labo-
ratory, E. I. du Pont de Nemours & Co., Inc., Buffalo, N. Y.
literature on the general subject. Little is directly applicable to the type of system employed, and only qualitative treatment is possible now. The reaction studied was the liquidphase hydrogenation of cyclohexene in the presence of a supported platinum catalyst. Its choice was arbitrary, and while of little practical importance, the reaction is representative of a common type carried out in batch processing. Certain experimental advantages were gained-hydrogenation proceeds readily at normal temperature and pressure, has no possibility of side reactions, and, with an over-all equilibrium constant of 2 X 108 at 25' C., the reverse reaction is negligible. The analytical procedures were also elementary. Apparatus
A semiflow system was employed where reactant gas was bubbled through a solution of cyclohexene in cyclohexane, and the catalyst suspended in the solution by mechanical agitation. This type of process represents a compromise between the autoclave-type reactor and a completely continuous flow system. The present investigation was conducted entirely in laboratory glass equipment. VENT
Experimental apparatus T-C = thermocouple
CONSTANT TEMP.
8ATH
The reactor consisted of a 1-liter threenecked Morton flask. This flask has four equally spaced perpendicular indentations about its periphery and a concave bottom, which aid the action of the agitator. A four-bladed stirrer driven by a variable-speed motor extended into the flask through a packing gland in the center neck. Tubes were sealed into one side neck for hydrogen delivery, sample withdrawal, and a thermocouple well. An ice water condenser on the exit neck minimized solvent escape by entrainment or vaporization. Hydrogen was dispersed into the reacting solution by a 20-mm. sintered-glass tube extending to the bottom of the flask. The entire assembly was immersed in a constant temperature bath maintained at 1 0 . l o C . However, because of the exothermic nature of the reaction, a temperature rise of 1' to 2' C. occurred in the reactor. Electrolytic hydrogen was purified by passing through a bed of copper shot heated to 150' C. and subsequently through activated-alumina drying tubes. Gas flow was metered with capillary orifices and pressure control was achieved through a Cartesian manostat. Difficulty was encountered in the early parts of the investigation with the purity of cyclohexene which undergoes rapid autoxidation during storage. It is generally agreed that the initial product of the oxidation is the hydroperoxide, but the subsequent transformations are in doubt (70). I n this case, the probable final product was cyclohexen-3-01. for infrared analysis of stored material indicated the presence of an alcohol. The effect of this impurity seemed to be competitive adsorption on the catalyst, with a slower hydrogenation rate than cyclohexene, rather than that of even a temporary poison. Each series of runs was made over such a short period of time that this effect is not appreciable, although between plots, rates may vary 10 to 15%. Once this factor was recognized, the purity of cyclohexene was ensured in subsequent runs by using freshly distilled material which had previously been refluxed over a 370 sodium amalgam. The fraction employed had a head VOL. 49,
NO. 5
M A Y 1957
807
0
5.24
TIME
Figure 1.
two
moo
am
-
zmo
3ow
SECOHOS
Typical concentration-time relotionship for experimental runs
Ternperdure, 26' C.; preuwe, 746 mm. of mercury; hydrqen h w , 30.7 X d e per r-di mtdtlysiweight, 0.975 gram; stirrer speed, 1 100 r.p.m.; dope, 13.7 X 10 -6 male per liter per remnd
temperature of 82.0' C. at 744 mm. of mercury. Cyclohexane was purified by scrubbing with concentrated sulfuric acid and distilling. The boiling point was 79 6' C. at 744 mm. of mercury. The catalyst was a 5% deposit of platinum on minus 300-mesh powdered activated-alumina carrier. Before use, the catalyst was dried for 48 houri at 120° C. and stored in a desiccator. I t was reclaimed by filtration from the reaction solution, washed with acetone, and redried. Periodic check runs indicated no decrease in activity, and fresh catalyst was required only to replace that last in handling.
Procedura
For a typical run, the system was purged with nitrogen which was also passed through the heated copper shot to remove oxygen. To the reaction flask was then added all but a small portion of a weighed amount of cyclohexane, about 500 ml. The catalyst was weighed by difference from a desiccator into the portion of solvent withheld, to prevent moisture pickup, and then both were transferred quantitatively into the reactor. Hydrogen was passed through the system for a half-hour flushingDeriod to saturate thesolvent and catalysiDurinz this wriod, flow rate. aeitator speed, &d t&pera&res were adjusted to the desired values. The reaction was started by adding the desired amount of cyclohexene, about 25 ml., to the flask through the $amplewithdrawal tube. After allowing 2 to 3 minutes for mixing, the initial sample was taken and additional samples thereafter +t measured time intervals until completion of the reaction. Usq808
ally five or six 3-ml. samples were taken for each run; volume change in the reaction flask was negligible. Samples were analyzed for bulk concentration of cyclohexene. by a method based on bromination of the double bond, and subsequent titration of excess bromine with thiosulfate (8). Tests with known solutions of cyclohexene in cyclohexane demonstrated the procedure to be completely quantitative. Rate of reaction was determined by plotting concentration of cyclohexene with time (Figure 1). Throughout this investigation, a linear relationship was found for all runs, and the over-all reaction rate was calculated directly from the slope of the plot. When a chemical reaction occurs in a heterogeneous system, certain interfacial factors which influence the observed rate of reaction are superimposed on specific chemical effects In this system, for hydrogen in the gas strqim to react catalytically with cyclohcxene and form a product in the main body of the solution, the following steps must be considered: 1. Transfer of hydrogen from gas phase to gas-liquid interface to bulk of solution. 2. Transfer of hydrogen and/or cyclohexene from the bulk of the solution to the catalyst-liquid interface and then adsorption on catalyst surface. 3. Reaction between hydrogen and cyclohexene-either both adsorbed, hydrogen adsorbed and cyclohexene at the catalyst-liquid interface, or cyclohexene adsorbed and hydrogen at the catalyst-liquid interface. 4. Desorption of cyclohexane from catalyst surface. In this system, when the reactants are initially in different phases and the
INDUSTRI*L AND ENGINEERING cHEM16lRY
Catalyst is wetted by one, solution of the second can be of utmost importance. Also, the fact that the reaction product is the solvent removes its diffusion into the bulb of solution from consideration. Adsorption and combination of reactants have heen characterized as chemical processes, while the other s t e p are described as mass trans fer. Generalized expressions for quantitative treatment may be written for each step (S), but they do not include all factors required to describe the system. Under steady-state conditions, the slowest step will determine the observed rate, and heterogeneous reactions are classified on the basis of the controlling step or step. The variables studied may be divided into two categories-those obviously entering the reaction itself, such as concentration of reactants and catalyst, and those detined by the physical syste-e.g., flask shape and rate of agitation. Any of the variables may have affected the steps just described. Effect of cyclohexene concentration has already bcen implied. For the observed rate to be constant during a run, as shown by the linearity in Figure 1, it must be independent of the bulk concentration of cyclohexene, which decreased with time. According to classical kinetics, this is zero order with respect to the hydrogen acceptor, and has been obscrved in many hydrogenations (72). Under conditions used in this work then, diffusion of cyelohexene to the liquidsolid interface, and its adsorption on the catalyst are not the controlling step, for both depend on bulk concentration. Some workers have attributed this independency to strong adsorption of the reactant, wherein concentration on the catalyst surface remaim essentially constant, even at v q f low concentrations. Figure 2, showing the relationship between observed rate and total pressure in the reactor, indicates the effect of hydrogen concentration. Since vapor pressures of solvent and liquid reactant were determined only by temperature, which was held constant, a change of total pressure in the systemcorresponds to a change in the partial pressure of the hydrogen. According to Henry's law, this would effect a linear change in solution wncentration of hydrogen, although in this case, it may not be the equilibrium, but rather a steady-state value. Linearity in Figure 2 could have been predicted' if either rate of solution or rate of diffusion of hydrogen to the liquid-solid. interface were the rate-controlling step. Howwer, cxtrapolation of this line to zero pressure indicates a finite rate of reaction which, of course, cannot be the case. , This discrepancy is further evidenced by the fact that over the limited p m r e range that could be.studied with the present
7
CHEMICAL PROCESSES
P
s
.
*
I)
equipment, the oberved rate increased 54% but according to Henry's law, the equilibrium hydrogen concentration should increase almost 80%. If it '131 assumed these two are proportional, some other factor must be affecting the observed rate. A plausible explanation of these data is size of the bubbles at reduced pressures. The mass rate of flow was held constant during this series of runs, requiring the volume rate of flow to increase as the pressure on the system was decreased. If the rate-controlling step were hydrogen solution, the slower reaction rates expected at reduced pressures would he partially offsit by an increase in gas contact area, Moreover, certain hydrodynamic factors, such as effect of larger bubbles on stirring &ciency, are also involved. The slope of the line could then be decreased, although the linear relationship retained, since gas volurhe is inversely proportional to pressure. If the slope were 10% greater, the line would agree with that predicted by Henry's law. A third factor connected directly with the reaction imelf is that of catalyst concentration. Many hydrogenation studies have reported a linear increase in the observed rate with addition of catalyst. This was true in this work (Figure 3) over only a small range, and the observed rate approached a limiting value. Extension of the curve to the origin is not supposition; in several runs without catalyst under differeat conditions, no change in concentration of cydohexene could be detected after 24 houn. By calculation it can be shown that the initial amount of cydohcxene employed was more than sufficient to cover completely the surface of the largest quantity of catalyst used. One explanation of thelimiting curve might then be that solution of hydrogen was determining the observed rate. That is, despite availability of reactive sites, rate of hydrogen transferral across the gas-liquid interface is limited. However, there is a second possibility. The k e d stirrer was unable to hold the catalyst entirely in suspension. With small amounts of catalyst, the suspension appeared homogeneous. But as the amount of catalyst was increased, some of the solid was deposited on the bottom and sides of the flask. The point at which this was first observed coincided with the maximum in observed reaction rate indicated on the plot. If the limiting effect of catalyst concentration is caused by ability to hold the solid in suspension, then only SUBpended material was dective in catalyzing the reaction. Generally, this is probably true, for the stagnant en. velope around settled material would
lBlOiYTL
Figure 2.
PREISYRL
-
"ILLIY111118
YERCYI"
Effect of pressure on rate of reaction
Temperature, 25-6' C.; hydrogen Row, 27.6 gram; stirrer speed, 1500 r.p.m.
present larger resistance to mass transfer than the turbulent body of solution. However, the exposed surface of the settled catalyst should be most reactive of all, for velocity of the solution at this point would be greater than that of the suspended solid. It is felt that this effect was offset by the fact that the settled material was not stagnant, but in a kind of dynamic equilibrium. Some solid was continuously swept into the fluid while other was deposited, and a fairly constant inventory maintained. This could account for scatter in the data. The method of hydrogen dispeniou used in this work was not the most efficient. A semiflow system was em-
X
mole per remnd; catalyst weight, 0.976
ployed to maintain the steady-state hydrogen concentration essentially constant in the reaction solution, but because of its effect on mixing, the gas flow itself is a variable (Figure 4). At two different pressures, a 20% increment separates the two curves. Calculation of the solubility of hydrogen in cydohexane from literature data indicates a 24% increase in the equilibrium concentration for the same increase in partial pressure of the gas. This is well within experimental error. There are two reasonable explanations for the shape of the curves. One, mentioned previously, is the increase in rate of hydrogen solution with an increase in contact area afforded by the
: rY I
Tenperahve, 25' C.8 p r e ~ ~ v r748 e , mm. of mercury; hydrogen Row, 31.3 X IO-' mole per secmd; diner speed, 1000 ,.p.m. VOL.49,NO.S
MAY 1917
809
I
I
0
2
4
6
FLOW
Figure 4.
RATE
-
8
10
MOLES
PER
12
SEC.
x
14
Effect of hydrogen flow rate on rate of reaction
Temperature, 25-6' C.; pressure, 0 = 5 3 4 mm. and 0.974 gram; stirrer speed, 1500 r.p.m.
bubbling reactant. However, this factor could be partially nullified by a decrease in contact time of the bubbles because they issue at higher velocities. The gas flow rate can also affect the mixing efficiency in the swirling solution and thereby affect the diffusional processes. When no gas is flowing, contents of the flask will be in motion, probably with some degree of a regular pattern. With
STIRRER
Figure
SPEED
-
e
= 741
mm. of mercury;
catalyst weight,
the gas flowing, the stream of bubbles disrupts this pattern and the turbulence is increased. At low flow rates, it was observed that some bubbles were carried along in solution for a short distance, interrupting the flow of fluid as they rose to the surface. At higher flow rates, the bubbles rose so quickly that their effect appeared to be similar to that of a solid tube extending into the fluid
REVOLUTIONS
PER
MINUTE
5. Effect of stirring on rate of reaction
Temperature, 25-7' C.; pressure, 7 4 6 of mm. mercury; hydrogen flow, 0 81.5 X 1 0-6 mole per second; catalyst weight, 0.976 gram
8 10
16
10'
INDUSTRIAL AND ENGINEERING CHEMISTRY
= 29.6 X 1 O - b and
=
pattern; they had a fixed imerruptive influence, and further increases in flow rates had no effect on the stirring efficiency. In this investigation, these two effects could not be separated; both could have been present. Increase in contact area with greater flow would be important only if hydrogen solution were the rate-controlling step. However, this is probably not the case. Because turbulence in the reaction flask offered the best method of studying mass transfer effects, interest in this work centered on the effect of the stirring speed on the observed rate. The Morton-type flask and stirrer design were chosen in the hope that the combined mixing efficiency would permit several rate controlling steps to be observed. Recent studies on vapor phase catalytic reactions have indicated the difficulties of eliminating mass transfer effect ( 6 ) , and it was thought these would be multiplied for reactions in solution. The stirrer motor employed had a variable shaft speed of 600 to 2500 r.p.m. The effect of stirring speed was studied over this range for two gas flow rates (Figure 5). Two similar curves were obtained, the upper being displaced by approximately an equal increment of 22% over the lower. That this increment is maintained at both ends of the plot where nonlinearity is observed-Le., 600 to 1000 r.p.m., and 1700 to 2500 r.p.m.-suggests that effect of gas flow rate is some additive function to turbulence created by the stirrer. Since the rate appears to be approaching some lower constant value with each curve, it is postulated that the effect is one of agitation itself. If measurements could have been extended to slower stirrer speeds, mixing caused by the paddle would have become negligible compared to that created by the bubbling gas. If solution of hydrogen were rate controlling, the increment probably would not be preserved in the linear portions of the curves. Linear dependence of reaction rate on stirrer speed is a classical criterion for mass transfer-controlled reactions and implies that the "effective film" thickness is inversely proportional to the stirrer speed in this range. Mass transfer of cyclohexene to the catalyst surface did not control the observed rate. Thus, in the linear portion of the curves, the reaction rate must correspond to rate of hydrogen diffusion, for no other step in the over-all process would be so affected by stirrer speed. The validity of the postulate can be determined by simple calculation. According to Fick's law, assuming adsorption on the catalyst at the interface is rapid, rate of mass transfer of one component through the bulk of the solution may be given as
CHEMICAL PROCESSES where c is bulk concentration of the component; D, diffusion coefficient; V, volume of solution; A, apparent catalyst area; and B, the path of diffusion or effective film thickness. This is a familiar expression for diffusional processes and is similar to that proposed early by Nernst ( 9 ) from his work with dissolution of solids in liquids. Levich (7) has shown from purely theoretical considerations that B 0: 0 ’ 1 3 Assuming all other variables to be the same,
or rates of mass transfer for hydrogen and cyclohexene will be equal when
z
Using Arnold’s method for estimating coefficients (7) and solubility data of Frolich and others ( 2 ) for hydrogen in cyclohexane, cyclohexene concentration at equal rates was calculated as 0.0143 mole per liter. This is generally less than that of the last sample taken in the experimental run. Thus, diffusion of cyclohexene in the solution must always have been greater than that of the hydrogen which controlled the reaction, at least in the linear portions of Figure 5. Another criterion often applied in classical kinetics is that for reactions controlled by chemical processes-Le., observed rate is independent of the stirring speed. O n this basis, either curve in Figure 5 when considered alone, shows that mass transfer effects had been eliminated at higher speeds. Presented together, the curves invalidate such a conclusion; only the greatest observed rate could be that of the chemical reaction, assuming the effect of flow rate is at least partially one of agitation. Whether or not the highest rate is out of the mass transfer region cannot be determined from these data because the curves are essentially graphical evidence of the mixing efficiency of the system, which is probably limited in itself. In any case, that the observed rate be independent of stirring rate is a necessary, but not sufficient, condition to define chemical control of a reaction. Influence of temperature on the observed rate of reaction is often used to help determine the rate-controlling step. Diffusional processes generally require much lower apparent energies of activation (3) than either surface reactions or adsorption. A series of runs was made between loo and 40’ C. to see if some of the higher observed rates might lie in the transition region of control. Specific velocity constants were calculated with the assumption of a first-order
0 = 1300 r.p.m.
rate dependency on hydrogen concentration, and plotted as a function of temperature in accord with the Arrhenius equation. Hydrogen solubilities in cyclohexane at various temperatures were used as an approximation to concentration. Since solubility data were available only at two temperatures (2, 77), a Clausius-Clapeyron type relationship was used to calculate values at the desired temperatures. It cannot be determined how closely actual hydrogen concentrations in the system approached such equilibrium values (Figure 6). Data for 1300 r.p.m. fall on a straight line except at lower temperatures where the deviation strongly suggests a shift in control to the chemical processes which are influenced more by temperature changes. Points for the higher stirrer speed are somewhat scattered because of turbulent instability at that rate, and the line for higher temperatures is drawn only to suggest an expected shift to mass transfer control. Apparent activation energies calculated from slopes of the solids lines for 1300 and 2500 r.p.m. are 4800 and 12,800 cal. per mole, respectively. The value for 1300 r.p.m. confirms that the observed rate is mass transfer-controlled in this region, for it corresponds to that generally reported for diffusional processes. At 2500 r.p.m., indications are that the reaction was at least in the transition region, if not close to chemical control. The value for the chemical reaction, of course, cannot be determined until other factors confirm this as the controlling step. In any case, these data seem to substantiate that mass transfer effects predominated in the data taken in this work, although they do show it is not impossible to
0 = 2500
r.p.m.
approach conditions where such effects are negligible-if not by increasing stirring efficiency, then by lowering temperature of the reaction. Considerable effort was made in this study to create maximum turbulence in the reaction flask. While high stirring efficiencies were obtained, mass transfer effects could not be completely eliminated. Even reducing the amount of catalyst was ineffective. Data were compared with several runs made in a standard low-pressure Parr hydrogenation apparatus. Proportionate quantities of materials were used in these batch runs and the reactor bottle was shaken at the maximum rate of 350 cycles per minute. Reaction rates were determined from cyclohexene concentration in solution rather than by decrease in hydrogen pressure, since the latter depends on reservoir volume. Specific velocity constants were calculated as for the semiflow system, and found to be comparable at stirring rates of 1500 to 1600 r.p.m., where mass transfer effects predominate. The rate of chemical reactions themselves determines in part the relative importance of mass transfer effects; therefore, cognizance that diffusional processes are likely to be present in such batch systems is essential. Several other factors were briefly investigated in this work. When temperature effects indicated a shift in the rate-controlling step at the higher stirring speeds, it was thought a change in stirrer design might promote further change. A second paddle of four blades was added, so that an end view of this stirrer showed eight blades equally spaced around the shaft, although in two planes. About inch separated the two paddles, and all the blades were VOL. 49, NO. 5
MAY 1957
81 1
a much smaller range of stirring rates than previously, and theaaximum observed rate was much lo
that of adsorption of one ofthe reactants, or of the chemical combination itsclf. Regardm of wbich rate is ohscmd, the rate of combination must be in accord with the law of rnagg action, written as r (ITIIREl)
*ED
.
~"OL",lO*P
Ps"
YIWIS
Figure 7. Hedl of stirrer design on rote of reoction
-
Temperowre, 2 5 6 ' C.; pressure, 746 mm. of ne-^ r-d; c o t d p l rslgM, 0.976 gram; - = ingle ond 0
at an angle of about 10 degrees to the vertical. The depth of the s t i m r below thesolutionsurface was not alped. Although it could he of fundsmenta;l importance to the hydrodynamic8 of the system, this factor was not studied. The effect of stirring speed on the tion was again deokrved rate i m r and is conthat premoualy found. At lower speeds. this stirrer did have a greater mixing efficiency,and a limiting rate was not approached as before. indicatine the relative turbulence created by the g& Aow. At higher stirring speeds, stirrer efficiency fell off r a p
'
wen flow, 81.5 double paddle rtirrar
X
molt, per
idly, and was less than the single paddlc probably because of vortex formation (4). At 2509 r.p.m., a gas e n d o p e m d a c t d y he obsenred around the shaft. Thus, care must be exercised in generalizing stirring effects, even within one system. Mixing efficiency of the sin& paddle in the Mortbn fla& was also w m w d to that in a m a r round-bottomed flask. The Morton, flask corresponds somewhat to a b a & d tank and stiming effects were magnified by its use &ure 8). & c a w of the streamlined 00& obtained in the round-boitomecl flask, the linear relationsliip applies to
-
M. eo
where e represents concentration. The question is whether e is measured in the bulk of solution, 01 face. Where only sorbed, knowledge seems necessary' to them. The case sorbed is generally pmtu genations. &cause mass fects eould not be completely
flow system is u this problem.
5
5