Fundamentals of Mass Transfer and Kinetics for the Hydrogenation of Nitrobenzene to Aniline

Reinaldo M. Machado, Air Products and Chemicals, Inc. No. 01-2007

Introduction The catalytic hydrogenation of nitrobenzene to From a historical perspective, aniline is perhaps one aniline in a continuously mixed slurry reactor is of the more important synthetic organic chemicals a complex chemical process. A number of com- ever manufactured. In 1856, Sir William Henry Perkin, peting mass transfer and kinetic rate processes a student at the Royal College of Chemistry in London, contribute to the overall observed reaction rate. discovered and isolated a purple dye during the oxi- Scale-up and optimization of the process require dation of impure aniline1. The discovery of this dye, that the contributing rate processes are under- known as mauve, created quite a stir and Perkin, stood individually and their impact on the total seeing the value of his discovery, proceeded to scale process is quantified. Laboratory reactors must be up the synthetic process for the production of mauve, operated under conditions that will allow mean- which included the synthesis of aniline. This process ingful process characterization and scale-up. was to become one of the first commercial processes to generate a synthetic organic chemical. Two classical mechanistic routes to aniline from During the last three decades, polyurethane plastics nitrobenzene are possible depending upon pro- have emerged as a growth industry and aniline once cess conditions and the effects of gas/liquid and

ALR Application Note again plays a key role as an industrial intermediate liquid/catalyst mass transfer. Intermediates formed used in the manufacture of MDI, 4,4’-diphenylmeth- during these competing chemical routes can act ane diisocyanate, a key commercial monomer in the as catalyst poisons that can radically change manufacture of polyurethane plastics. reactor performance. Aniline is produced by the reduction of nitrobenzene, This paper will describe and characterize these which is produced from the nitration of benzene in a competing processes for scale-up. mixture of sulfuric and nitric acid. Originally, nitrobenzene was reacted with dispersed iron in the presence Keywords: Mass transfer, kinetics, scale-up, of HCl to generate aniline and an iron oxide sludge. hydrogenation, aniline, nitrobenzene, slurry, agita-

tion, liquid-solid, catalyst, pulse kinetics, reaction NO2 NH2 HCI (1) rate. + 9/4 Fe + H2O + 3/4 Fe3O4

This process generated large quantities of waste and was eventually replaced by the catalytic hydrogenation of nitrobenzene in a three-phase slurry reactor. Aniline can also be produced in the gas phase by the reduction of nitrobenzene with hydrogen over fixed catalysts2. This paper focuses on the characterization CH2, Bulk

of a slurry process for the reduction of nitrobenzene to n aniline. Catalyst NO2 NH2 CNitrobenyine, Bulk Catalyst (2) + 3 H2 + 2 H2O Concentratio

The catalytic hydrogenation of nitrobenzene to aniline Gas Liquid Bulk in a continuously mixed slurry reactor is a complex Film Film Liquid chemical process. dp 0 Distance 2 A number of competing mass transfer and kinetic rate Solid / Liquid Film processes contribute to the overall observed reaction Figure 1 rate. Scale-up and optimization of the process require Typical concentration profile during hydrogenation of nitrobenzene. that the contributing rate processes are understood individually and that their impact on the total process Literature rate be quantified. Laboratory reactors must be operat- A review of the literature can yield a confusing ed under conditions that will allow meaningful process description of the overall process kinetics of nitroben- characterization for scale-up. zene hydrogenation with authors reporting reaction orders for nitrobenzene and hydrogen between zero The key rate processes are illustrated in Figure 1, and one depending upon reaction conditions3-6. which shows schematically typical concentration A general review of the kinetic studies is given in profiles during the hydrogenation of the nitrobenzene: the doctoral thesis by Füsun Yücelen7. Most kinetic first, the rate of hydrogen mass transfer from the gas studies generate global kinetic rate models based on phase to the liquid phase; the rate of hydrogen and hydrogen uptake and overall conversion. However nitrobenzene mass transfer from the bulk liquid phase some detailed studies using analytical methods to to the outer surface of the catalyst; the rate of hydro- identify the individual intermediate species have also gen and nitrobenzene mass transfer into the porous been used to characterize reaction kinetic processes8. catalyst; and finally, the adsorption and kinetic rates of the hydrogen and nitrobenzene on the inner catalytic Additionally, most of the kinetics studies in the litera- surface of the catalyst particle. ture are conducted at relatively low temperatures with solvents. Optimized industrial processes would rather Two classical, mechanistic pathways to aniline from eliminate the use of solvents and their associated nitrobenzene are possible depending upon process separation processes, which can negatively impact conditions and the effects of gas/liquid and liquid/ overall process economics. Environmental pressures catalyst mass transfer. Intermediates formed during also tend to favor processes that use fewer solvents. these competing chemical routes can act as revers- ible catalyst poisons that can radically change reactor It would generally be preferred to use the reaction performance. It is not within the scope of this paper products of the nitrobenzene reduction, i.e., aniline to exhaustively review the literature or to optimize the and water, as the reaction solvent. Unfortunately, ani- process catalysts, temperature or conditions for the line and water are not miscible at low temperatures; reduction of nitrobenzene to aniline; rather, this paper therefore, higher temperatures must be employed will describe the principles, laboratory methods and to maintain a single liquid phase under process con- analysis techniques for characterizing these competing ditions. No literature studies are reported under these processes. conditions. It is hoped that his study will help shed some light on this process while illustrating the general principles required for characterizing a hydrogenation process from the laboratory scale.

ALR Application Note Experimental Nitrobenzene Nitrosobenzene Phenylhydroxylamine Aniline

For this set of studies the global reaction kinetics was H2 H2 H2 NO2 NO NOH NH2 determined by continuously monitoring the reaction H exotherm using the METTLER TOLEDO RC1/HP60 high- pressure reactor fitted with a single baffle and gas + Azoxzbenzene induction agitator supplied by Mettler-Toledo Inc.9. N N The METTLER TOLEDO PIC10 pressure controller was O H operated to allow hydrogen gas into the reactor to 2 maintain the desired pressure setpoint while gas was N N Azobenzene not allowed to exit the reactor. H The reduction of nitrobenzene to aniline is exothermic 2 H and generates –536.6 ± 5.9 kJ/mole as measured in 2 N N 2 NH2 this study, making the calorimetric method ideal for H H this process. Reagents were all purchased through Phenylhydrazine Figure 2 Aldrich. Individual species were monitored using inter- Haber’s reaction pathways for the reduction of nitrobenzene to aniline, Strätz10. nal standard gas chromatographic methods. Sponge ® 200 nickel or Raney type catalysts were used in this A. N. Campbell (11) study at concentrations between 0.05 % and 0.10 %. 180 1 Phase Region All reactions were carried out with the agitator operat- 160 W.Alexejew (12) ing at 1000 rpm, which gave a mixing intensity mea- 140 sured at 1.0 watt/liter. 120 2 Phase Region emperature, Celsius T Chemistry 110

The chemistry for the reduction of nitrobenzene to ani- 100 line was originally elucidated by Haber as reported by 80 Strätz10 and is illustrated in Figure 2. Two paths exist to aniline according to this scheme. 60

The first proceeds from the sequential reduction of Figure 3 Weight % Aniline nitrobenzene: first to nitrosobenzene, next to phenylhy- Aniline/water phase diagram. droxylamine, and finally to aniline. This path, we will see, is favored under conditions aniline. From Figure 3, temperatures in excess of in which the concentrations of nitrobenzene are low, 160 °C would be required to keep this mixture homo- < 0.15 %. If the concentrations of nitrosobenzene geneous. High temperatures can impact byproduct and phenylhydroxylamine are allowed to increase, the chemistry, reducing aniline yields. However, if aniline formation of azoxybenzene proceeds rapidly followed is fed to the reactor system along with the nitroben- by reduction to azobenzene, phenyl hydrazine, and zene or if water is continuously removed via distil- finally, aniline. lation, the concentration of aniline in a continuous The second path, while ultimately yielding aniline, process can be maintained at aniline concentrations is much slower kinetically on nickel catalysts and is higher than that dictated by the stoichiometry. favored under conditions in which the nitrobenzene concentrations are high, > 0.15 %. During high tem- Therefore, lower temperatures are required to keep the perature studies, the only stable intermediates identi- system homogeneous. In this study the continuous fied and monitored were nitrobenzene, azoxybenzene feed to the reactor consisted of 50 % aniline and and azobenzene, as we will see later. 50 % nitrobenzene, which maintained a nominal steady-state composition in the reactor of 14.3 % Aniline/Water Solubility water and 85.7 % aniline. Under this composition, The solubility of aniline and water has been well doc- temperatures in excess of 135 °C are required to umented and is illustrated in Figure 311,12. maintain homogeneity of the liquid phase. The stoichiometric composition of the complete reduction product of nitrobenzene is 27.9 % water/72.1 %

ALR Application Note Gas-Liquid Mass Transfer Many involve the oxidation of aqueous sulfite solu- The primary and most fundamental rate process that tions and the absorption of oxygen. Measured values must be characterized when conducting any hydroge- of kla from these experiments must be translated to nation reaction is the mass transfer of gas into the the system of interest, which requires detailed knowl- liquid reaction phase. In most industrial processes edge of the physical properties of the reaction mixture. using pure hydrogen, the gas film mass transfer An alternative to this method is the use of the batch resistance can be ignored and mass transfer rates are absorption method, which determines the kla in the dominated by the liquid film resistance. The rate of exact reaction composition to be used in the pro- gas mass transfer can then be represented by the cess study under the process conditions of interest. following expression: The advantages of this approach are obvious as no

kl · Ag estimates of physical properties are required and the Rate = _ (3) exact mass transfer coefficient for the system of inter- VL (CH2, sat CH2, bulk) est is obtained directly. where kl = liquid side mass transfer coefficient [m/sec], Ag = gas/liquid interfacial area, VL = reactor Both the gas solubility and the mass transfer coef- liquid volume. The parameters kl·Ag/VL can be com- ficient, kla, can be accurately measured for virtually bined into the familiar mass transfer coefficient, any liquid chemical system by monitoring the batch absorption of a gas into a liquid. This is accomplished kl · Ag = kla (4) by continuously monitoring the pressure drop via a VL fast-response pressure transducer in a batch gas- The mass transfer coefficient, kla [sec-1], can be liquid system. The technique described by Dietrich et viewed as a fundamental parameter describing a par- al. is straight forward to implement19. ticular reactor system. Numerous correlations for the gas mass transfer coefficient in a variety of reactors First, the volume of the reaction/mass-transfer ves- have been documented13,14. sel must be determined. The mass of the liquid and However, these correlations may not be applicable its density at the process temperature must also be to small laboratory reactors that often exhibit unique known. Next, the reactor is filled approximately half mixing dynamics and reactor geometry and operate full with the process liquid to be studied without the with a particular chemistry and a wide spectrum of catalyst. For the solubility measurements to be accu- process physical properties. Often trying to match per- rate, the system must be completely purged of all formance and mixing between small- and largescale gases so that the only gases in the system are vapors reactors is difficult when only relying upon mechani- from the liquid phase. cal measurements such as agitation rate, geometry, agitator design, etc. A more consistent approach to Once the system has been purged, it is sealed and studying process dynamics in small reactors is to equilibrated at the temperature to be studied with measure the gas mass transfer coefficient directly in appropriate agitation. The equilibrium pressure of the the actual reaction system. Establishing a realistic liquid with its vapor is recorded. Next the agitator rota- mass transfer correlation for the laboratory reactor tion is stopped and the circulating liquid in the reactor system will allow «tuning» of the kla to the desired is allowed to stabilize. The gas to be studied is slowly conditions required for the scale-up or scale-down of added to the head space of the reaction vessel to the a process. The strategy for scale-up is to match mass desired pressure. It is important to preheat the gas to transfer coefficients, kla, from the small to the large the process temperature before introduction into the scale rather than trying to match geometry, agitation vessel. During this entire procedure the temperature rates, or other mixing characteristics. and pressure are monitored. The quiescent mixture is allowed to stabilize for 1 or 2 minutes and then the Batch Absorption Method for the Determination of agitation is increased within 1 to 2 seconds to its final kla and Gas Solubility designated value. The pressure from a fast-response Numerous techniques have been applied to the mea- pressure transducer is monitored electronically and surement of gas-liquid mass transfer in slurry reactors recorded using a fast data sampling device or a strip and bubble columns15-18. chart recorder.

ALR Application Note 16.4 Immediately after the agitator is started, the pressure 800 RPM in the vessel begins to decrease as gas from the head 16.2 1000RPM space of the reactor vessel is absorbed into the liquid. , barabs The pressure continues to drop until the saturation 16.0 1200 RPM point is reached. 15.8 1400 RPM A typical example of this experiment is illustrated Final Pressure in Figure 4. It is clear from Figure 4 that the rate of 15.6 absorption is strongly influenced by the agitation 15.4 intensity. From this data both the gas solubility and 15.2

the mass-transfer coefficient, kla, can be determined. Pressure in the reactor

The key results are summarized in the following 0 5 10 15 20 25 equations: Time, seconds Figure 4 PPM − E VG 1 Pressure profile during batch hydrogen absorption in the RC1/HP60 w/induction agitator CPLsat = ( − P E) · · · (5) (14.3 % water in aniline @ 140 °C). PPF − EL VL RT 1.00 RPM kla

PP− PP− ) 1/sec M F M E F ln = kla · · t (6) 800 0.146 -P 0.50

PP− F PPF − E M 1000 0.235

)/(P 0.30 PP− F 1200 0.335 M E (7) ln(α) = − kla · · t 0.20 PPF − E 1400 0.422 where 0.10 Alpha, (P-P CLsat = saturation solubility of gas in the liquid phase [mole/liter or M], 0.05 P = pressure in the vessel [barabs], 0 2 4 6 8 10 12 14 16 18 20 P = maximum pressure after the head space Time, seconds M Figure 5 is charged with gas [barabs], Characterization of gas-liquid mass transfer via batch hydrogen absorbtion (14.3 % water in aniline @ 140 °C). PF = final pressure after gas has saturated the liquid phase [barabs], 0.50 PE = equilibrium pressure of the liquid with its Gas Induction vapor [barabs], Agitator 0.40 14.3 % Water V = volume of the gas phase [liters], in Aniline G @ 140 °C V = volume of the liquid phase [liters], L ..30 R = ideal gas constant, 0.08314 [barabs·liter/mole·K], 0.20 T = temperature in degrees [K], kla, 1/sec kla = mass transfer coefficient [sec-1], 0.10 t = time from the onset of agitation [sec], α = alpha, (P-P )/(P -P ), fraction of pressure F M F 0.00 remaining. 0 2 4 6 8 10 12 14 16 18 20 Rotation Rate, RPM Figure 6 The key assumptions in this analysis are as follows: Effect of agitation on the gas/liquid mass transfer coefficient in the RC1/HP60. first, the gas phase is ideal and follows the ideal gas law. Second, the value of kla is constant during the The effect of the agitation rate on the mass transfer experiment. Finally, the volume of the liquid and gas coefficient is plotted in Figure 6 for this reactor sys- phases is essentially constant during the experiment. tem. The gas mass transfer coefficient can also be a The data illustrated in Figure 4 is plotted according function of the liquid height above the agitator and this to Equation 7 in Figure 5. From the slope of the line, effect can also be studied using this technique. the value of kla can be determined.

ALR Application Note However, for this study an average reactor liquid No experimental method is absolutely unambiguous, volume of 1 liter was used as a basis for comparison. because catalyst concentration can also affect the In most well-designed, hydrogenation stirred tank mass transfer coefficient and complex reaction chem- reactors, gas mass transfer coefficients between 0.05 istry may vary with catalyst concentration. sec-1 and 0.5 sec-1 can be achieved. It is the responsibility of the investigator to explore these variables as required. From Figure 6, a kla of 0.235 sec-1 @ 1000 rpm was chosen, which represents values often found in Solid-Liquid Mass Transfer larger-scale, stirred hydrogenation reaction vessels. Once hydrogen has entered the well-mixed bulk liquid Solubility measurements and estimates for a variety of phase of the stirred reactor and as feed is added to conditions are tabulated in Table 1 for this study. the liquid phase of the reactor, both the hydrogen and nitrobenzene, in this case, must be transported from Hydrogen solubility in 14.3 % water/85.7 % aniline [M] the bulk liquid phase, according to Figure 1, to the external surface of the solid catalyst particle. Total system pressure [barg] This process is designated as liquid/ solid mass Temperature 8.0 11.0 14.0 transfer. When the substrate to be reduced, nitroben- 140 °C 0.0090 0.0138 0.0186 zene, is in high concentration, then only the mass 160 °C 0.0073 0.0122 0.0170 transfer of hydrogen needs to be considered. However, Table 1 when the process is operated in a continuous or Hydrogen solubility measurements and estimates from Equation 5. semibatch manner under conditions in which the feed A Warning about Agitation Studies is added sufficiently slowly to limit its accumulation, Often it is desired to operate a laboratory reactor then its mass transport to the catalyst must be con- under conditions in which mass transfer is sufficiently sidered as well. The rate of liquid to solid mass trans- fast that it does not impact the overall process hydro- fer can be expressed in the following equation: genation rate. Investigators increase the agitation ks · As rate until the process kinetics no longer increase with Rate = (8) _ increasing agitation. This plateau at high rotation rates VL (CS, bulk CH, interface) is assumed to be the gas mass transfer independent where region of operation. Unfortunately, this approach is ks = liquid/solid mass transfer coefficient [m/sec], not rigorous in determining if a process is independent As = total solid/liq uid interfacial area in the reactor, of gas mass transfer. VL = reactor liquid volume, Often in small reactors, especially those that are CS,bulk, CS,interface = molar concentrations of nitro- unbaffled, at high rotation rates, the process liquid benzene or hydrogen in the bulk liquid phase begins to form a vortex and begins to swirl, limiting and at the solid/liquid interface. gas/liquid mixing and mass transfer. Under such conditions the only process that may have been elu- The parameters ks·As/VL can be combined into the cidated is the dependency and limitation of the gas familiar mass transfer coefficient, mass transfer coefficient at high agitation rates. ks · As A more rigorous and consistent way to ensure experi- = ksa (9) mentally that a process is independent of gas mass VL transfer is to determine the effect of changing the kla For small catalyst particles, the mass transfer coef- on the overall hydrogenation rate or to determine ficient, ksa [sec–1], may be viewed as a fundamental the effect of changing the catalyst concentration on parameter describing the catalyst and to a lesser the overall process rate. extent the reactor system itself. For example, if changing the catalyst concentration Unlike the kla, which is primarily a function of the continues to impact a process rate in a linear manner, agitator and reactor design, agitation intensity, and then the process may be assumed to be indepen- the physical properties of the fluid, the ksa in catalytic dent of gas mass transfer. When gas mass transfer systems is often dominated by the physical charac- becomes limiting, increasing the catalyst concentra- teristics of the catalyst, i.e., particle size distribution, tion will have little or no effect on the process rate. particle shape and density.

ALR Application Note While mixing intensity and reactor design can influ- cussion of these theories is beyond the scope of this ence the ksa, for catalytic systems where dpi,cat. paper and we will focus on estimates of Repi that can <50 microns, the effect of reactor design and mixing provide a bound on the liquid/solid mass transfer rate is minimal once a uniform catalyst suspension has process. For catalyst systems with catalyst sizes less been achieved. than 50 microns, often one can let C = 0 and approx- imate Sh = 2. This estimate is especially good with The solid/liquid surface area is determined exclusively particles <10 microns. Not surprising the diffusivity is by the particle size distribution, the total amount of often the most critical parameter impacting estimates catalyst in the system and the catalyst density. of the ksa so that sophisticated methods of calcu-

Assuming that the catalyst particles can be approxi- lating Repi are rendered futile without a reasonable mated as spherical, the total solid/liquid surface area measure or estimate of the diffusivity. In this system can be determined by summing up the surface area the diffusivity was estimated using the Stokes/Einstein associated with each fraction of the particle size distri- equation which can be used when limited information bution according to the following: is available for a system23. The Wilke-Chang estima- tion method can also be used for liquid systems23. AsAs == AsAs (10) ∑Σ i i T D = 1.05 · 10-9 · (13) 1/3 μ · V mi,cat 6 m Asi = · (11) where ρcatdp i,cat Vm = molar volume of diffusing species at its where boiling point, (113 ml/gmole for nitrobenzene, dpi,cat = diameter of catalyst particle, 14.3 ml/gmole for hydrogen), mi,cat = mass of catalyst with the particle diameter, T = temperature in degrees Kelvin, dpi,cat, μ = liquid viscosity, (0.00263 Poise @ 140 °C for Pcat = liquid-filled catalyst density. 14.3 % water in aniline), D = .4·10-5 cm2/sec @ 140 °C for nitrobenzene. A number of reviews and articles describe various correlations used to estimate the liquid/solid mass Kolmogoroff’s theory is based on the turbulent transfer14,20-22. Most liquid/solid mass transfer eddy length, λ, which is given as correlations conform to the following relationship: 3 1 µ 4 λ = (14) a b e ·ρ2 Sh = 2+C · Repi · Sc (12) v where where ev = Mixing energy per unit volume of the reactor Sh ksi · dpi,cat/D, liquid reactor liquid ksi = mass transfer coefficient for particles with diameter dpi,cat, The Repi based on Kolmogoroff’s theory is given as D = diffusivity of hydrogen or substrate in the 1 2 2 reaction liquid mass, Re = e · dp 4·ρ ,for λ >> dp pi v i,cat µ3 i,cat Repi = particle Reynolds Number for a catalyst (15) particle with diameter, dp , 1 i.cat 2 3 Sc = Schmidt Number, μ/( ·D), Re = e · dp 4·ρ ,for λ << dp ρ pi v i,cat µ3 i,cat μ, ρ = liquid-phase viscosity and density, respectively, a, b, C = dimensionless coefficients.

The theories of mass transport differ most funda- mentally in their attempts to define Repi using either terminal velocity-slip velocity theory or Kolmogoroff’s theory of local isotropic turbulence20. A complete dis-

ALR Application Note The Repi based on the terminal velocity-slip velocity a wide range of process conditions can be effectively is given as explored. In our studies we investigated 8 barg, 11 barg and 14 barg total pressure at 100 °C, 120 °C, v Re = dp · ρ · c , [14] (16) 140 °C and 160 °C. pi i,cat µ3 where Mass transfer coefficients based on a=1/2, b=1/3 according to Eqn 12. v = (v 2+v 2+v 2)1/2, c e tp s Normalized ksaw Basis for Repi Mixing Energy C, Eqn 12 [1/(sec·wt % cat.)] ve = neutral particle relative velocity, 0.40 –––––– –––––– 0.00 vtp = terminal particle velocity, vs = slip velocity due to inertial differences between 0.48 Panicle terminal –––––– 0.72 the particles and fluid. velocity, vc = vtp Often for small particles we may assume for estima- 0.61 Kolmogoroff’s 1.0 watt/liter, 0.40 theory agitator tion purposes that vc = vtp. RC1/HP60 @ 1000, The mass transfer coefficient for each population of rpm w/gas the catalyst distribution is given by 0.74 Kolmogoroff’s 10.0 watt/liter 0.40 theory ksa = ks · As (17) Table 2 i i i Estimates of the´nitrobenzene mass transfer coefficient @ 140 °C. and the total mass transfer coefficient can then be Differential calculated from 20 100 Distribution

Cumulative ksa = ksa (18) % Σ i 15 75 Distribution

For convenience, we can define a solid/liquid mass eight transfer coefficient that is normalized to the weight 10 50 percent of catalyst in the reaction mixture. %

5 25 ight ksa ferential W We Cumulative ksa = (19) Dif wt % catalyst 0 0 0 2 5 10 20 50 100 200 500 The particle size distribution for sponge nickel catalyst Catalyst Particle Size, dp, microns Figure 7 is illustrated in Figure 7 and is typical for this class Sponge nickel catalyst particle size distribution (weight/volume average paticle size = of catalyst. In Figure 8 the partial mass transfer coef- 34.2 microns). ficient based on variations of Equation 12 as indicated 0.16 is illustrated. The final net mass transfer coefficients Kolmogoroff’s 0.14 Theory for the catalyst used are tabulated in Table 2 accord- 1 Watt/Liter Net ksaw = ing to Equations 17, 18 and 19. For the terminal-slip 0.12 0.61 1/(sec Wt %) velocity theory example calculated in Table 2, the 0.10 Terminal/Slip velocity, vc, was assumed to be equal to the terminal Velocity Theory 0.08 C = 0.72 settling velocity, v . The terminal velocity was calcu- Net ksaw = tp 0.48 1/(sec Wt %) lated according to Cheremisinoff22. 0.06

0.04 Sh = 2 Net ksaw = 0.40 1/(sec Wt %) Pulse Kinetic Studies 0.02 An effective way of getting a broad perspective on ksai (1/sec)/(Net Wt % Catalyst) 0.00 the kinetics in a particular system is to conduct pulse 0 2 5 10 20 50 100 200 500 experiments with the appropriate feed under the Catalyst Particle Size, dp, microns Figure 8 desired conditions. It is generally easy to control tem- Comparative distributions of the mass transfer coefficient for nitrobenzene and sponge perature and pressure during pulse experiments and nickel catalyst.

ALR Application Note Reaction Unlike most studies that employ hydrogen uptake as a 100 Temperature measure of kinetics, we used isoperibolic calorimetry, i.e., the reactor jacket maintained at the designated atts 80 160 °C temperature setpoint with the reaction exotherm 140 °C 60 monitored continuously in the RC1/HP60 calorimetric 120 °C reactor. In this manner the hydrogen pressure can be 40 100 °C maintained constant during the reaction. Nominal 3 gram pulses of 50 % aniline and 50 % 20 Reaction Rate, W nitrobenzene (1500 ppm or 0.016 M nitrobenzene) 0 were added to the reactor between 5 and 20 seconds, which contained 14.3 % water, 85.7 % aniline and 0 100 200 300 400 500 600 700 approximately 0.05 % catalyst. Time, Seconds Figure 9 Exotherm profiles from feed pulse tests at 11 barg. A sample of the isoperibolic data is illustrated in 0.0 Figure 9. The data was renormalized at the end of the Slope/Wt % Catalyst = k 1/ (sec Wt % ) Temperature -0.5 feed addition and conformed within statistical limits 100 °C to a simple rate model, first order in nitrobenzene -1.0 120 °C and zero order in hydrogen, which is illustrated in Figure 10 for experiments at 11 barg. Based on the -1.5 140 °C 160 °C low overall kinetic rates during the pulse tests, the -2.0 complications of gas/liquid mass transfer can be ignored. The rate data was normalized for the weight -2.5 percent of catalyst used and is illustrated in Table 3. -3.0 In {1-Thermal Coversion}

-3.5 Normalized first order rate constants 0 100 200 300 400 500 600 700 in 14.3 % water/85.7 % aniline [1/(sec wt% cat.)] Time, Seconds Figure 10 Total system pressure [barg] First order rate analysis of nitrobenzene pulse test (at 11.0 barg pressure w/ 0.05 % catalyst). Temperature 8.0 11.0 14.0

100 °C 0.117 0.0958 0.0972 1.0 120 °C 0.146 0.100 0.109 Net System 140 °C 0.467 0.482 0.567 Pressure 8.0 barg 160 °C 0.661 0.690 0.702 0.5 Table 3 11.0 barg Effect of temperature and pressure on nitrobenzene hydrogenation 0.3 rate constants. 14.0 barg 0.2

The rate constants are plotted according to the tradi- 0.1 tional Arrhenius model in Figure 11 and as a function

of hydrogen solubility in Figure 12. There are two -160 °C -140 °C -120 °C -100 °C 0.05 points that should be noted in these plots. 2.3 2.3 2.4 2.5 2.6 2.7

First, the rate constants are not a function of the Rate Constant, 1/(Sec Wt % Catalyst) 1000/T(K) Figure 11 hydrogen solubility as illustrated in Figure 12. Arrhenius plot of normalized nitrobenzene rate constants at different pressures. Secondly, there is a sharp transition in the reaction rate constants between 120 °C and 140 °C. This This low activation energy is consistent with mass transition corresponds to the transition from a single transfer controlled processes. Recall that the normal- homogeneous liquid phase at 140 °C to a two-liquid- ized liquid/solid mass transfer coefficient for this phase region at 120 °C according to Figure 3. In the catalyst given in the previous section is between 0.40 single-phase region, the rate constants are a weak and 0.61 1/(sec · wt % cat.), which is approximately function of temperature and show an activation energy the measured rate constant at high temperature for of 5.4 kcal/mole according to Figure 11. this process.

ALR Application Note 1.00 This suggests that the process rate during the pulse Process experiments is controlled by liquid/solid mass transfer. Temperature 0.80 140 °C

When operating under solid/liquid mass transfer 160 °C control, the effective concentration of nitrobenzene at 0.60 the catalyst solid/liquid interface is essentially zero, according to Equation 8. This corresponds to surface 0.40 reaction kinetics that are infinitely fast compared to the time scales of diffusion. 0.20

0.00 Under the two-phase region at and below 120 °C, the 0.005 0.010 0.015 0.020 0.025 rate was independent of both pressure and tempera- Rate Constant, 1/(Sec Wt % Catalyst) Hydrogen Solubility, Mole/Liter ture and probably controlled by liquid/liquid mass Figure 12 Effect of hydrogen concentration on normalized nitrobenzene rate constants. transfer or nitrobenzene solubility limits in the water phase. Sponge nickel will generally favor the water phase in a two-phase mixture. The low rate observed the reactor volume during the semibatch experiment in the two-phase region suggests that commercial can be assumed constant at the average fill volume, processes should be operated under single-phase Vr(average) = [Vr(initial) + Vr(final)]/2. process conditions. It should also be noted that only Also, if a continuous process is operated at very high nitrobenzene was considered to be an active species conversion, then Qout · Cout << Qin · Cin and Rate · Vr during these experiments and no other intermediates and Qout · Cout ≈ 0. Then under these conditions the were monitored. design equations for the continuous and semibatch processes are virtually identical: Semibatch and Pseudocontinuous Operation and dN = (Q ·C )−(Rate·V) (21) Rate Analysis dt in in r The heat generation for highly exothermic reactions is often difficult to control in batch processes. Therefore, Since it is generally desired to operate a CSTR at high it is preferred to operate these reactions under semi- conversion, this analysis suggests that an effective batch or continuous addition of the feed. The overall method to characterize a continuous process would transient mass balance for species A in a CSTR is be operation of a semibatch process at high, instan- given by the following expression for a simple process taneous conversion, i.e., the concentration of A in the reaction, A B. reactor << Cin, to obtain process results that are simi- Acc. In Out Reaction Rate lar when compared to a continuous operation. This dN mode of operation for a semibatch reactor is called = (Q ·C )−(Q ·C ) −(Rate·V) (18) dt in in out out r pseudocontinuous operation. From an operational perspective in the laboratory, semibatch operations yield where a large volume of information using very little material N = moles of A in the reactor, and generating very little waste and are preferred over Qin, Qout = incoming and exiting volumetric flow true continuous operation whenever possible. rates to the reactor, Cin, Cout = incoming and exiting concentration of A [moles/liter]., Rate = volumetric molar react. rate for A B, Vr = liquid reactor volume, a function of time during semibatch operation.

A reactor operated in semibatch mode requires that

Qout = 0. When a semibatch experiment is operated over a narrow volume range such that Vr(initial) and Vr(final) differ by less than approximately 20 %, then

ALR Application Note 10 320 The application of this concept is best illustrated in Reaction Rate 280 @ 140 °C Figure 13. The RC1/HP60 reactor was filled to approxi- 14 barg, 0.05 % mately 1 liter with 14.3 % water, 85.7 % aniline and 240 Catalyst

atts Theoretical Feed 0.5 grams of catalyst. Six consecutive increasing 200 Controlled Rate feed ramps of 50 % aniline/50 % nitrobenzene Infinitely 160 were imposed. The reaction was run at 140 °C at 14.0 Fast Reaction barg. Two curves are illustrated in Figure 13. The first 120 shows the theoretical maximum exotherm rate based 80 on the feed rates assuming that the reaction rate is 40 Reaction Rate, W infinite and the second curve is the actual reaction 0 rate measured in watts. 0 20 40 60 80 100 120 140 As illustrated, during the first three feed ramps, the Figure 13 Time, Minutes reaction rate initially trailed the step changes in the Effect of feed rate on reaction rate during multiramp feed experiments: feed rate but eventually approached so that at the comparison with theoretical feed controlled rate. rate plateaus, Q · C = Rate · V . During these first in in r 1.4 three step changes, accumulation of nitrobenzene was 240 Reaction Rate 1.2 minimal and the semibatch reactor was operating in 200 Mass Feed Added pseudocontinuous mode. 1.0 160 Note that the ksa estimated for this process is ksa = atts Nitrobenzene 0.8 0.5 1/(sec · wt % cat.) · 0.05 % catalyst = 120 Azobenzene 0.6 0.025 (1/sec). This is one order of magnitude smaller 80 Azoxybenzene than the kla = 0.235 1/sec. 0.4 40

0.2 Figure 14 illustrates the concentration profiles of two 0 Reaction Rate, W 0.0 intermediate species, azoxybenzene and azobenzene, Wt % and nitrobenzene. During the first three steps in feed, 0 20 40 60 80 100 120 140 Intemediates the only species identified was nitrobenzene. Figure 14 Time, Minutes This result is consistent with the first path of the Haber Accumulation of nitrobenzene and formation of intermediates reaction scheme, illustrated in Figure 2, and yields (140 °C, 14 barg, 0.05 % catalyst). high reaction rates when the feed is added continuously to maintain the nitrobenzene at low concentra- The process rate as a function of nitrobenzene con- tions, < 0.1 %. It is also consistent with the pulse centration is illustrated in Figure 15 for the range of test experiments, which demonstrate that under low operation before the appearance of intermediates. nitrobenzene concentration the rate processes are This analysis is similar to that of Lopidana2 who controlled by liquid/solid mass transfer. investigated the hydrogenation of nitrobenzene on platinum catalysts. He shows that the linear portion As the feed ramps continue, the reaction rate begins of the rate curve at low nitrobenzene concentration to deviate significantly from the theoretical rate based is zero order with respect to hydrogen, as found in on the feed rate. As the rate of nitrobenzene addition this study during the pulse tests. In addition, he dem- increases, the reaction rate decreases and the accu- onstrated that the plateau region of the rate curve is mulation of nitrobenzene in the reactor increases obviously zero order with respect to nitrobenzene, yet dramatically. As the rate decreases with increasing first order with respect to hydrogen. nitrobenzene concentration, the emergence of azoxybenzene is revealed in Figure 14. At this point the process no longer conforms to pseudocontinuous operation and the reaction mechanism begins to shift to the second pathway described by Haber illustrated in Figure 2.

ALR Application Note 11 200 The process rates, calculated using Equation 22, from Rate Based on Pulse Test the pulse experiments at 140 °C and 14 barg are plotted in Figure 15. The rates are identical with those 160 atts observed from the pseudocontinuous operation at low Mixed Control Region 120 Interparticle Mass Transfer, nitrobenzene concentration. This demonstrates that Adsorption, Surface Kinetics under the pseudocontinuous operation, the reactor is operating under liquid/solid mass transfer control. 80

Calculation of rates from the pulse experiments: Process Rate, W 40 Solid/Liquid Mass Transfer Control

-1 -1 Rate [watts] = –ΔHreaction [J/mole] · k [sec · wt %cat ] 0 0.0 0.2 0.4 0.6 · V [liter] · C [M] r nitrobenzene (22) Concentration of Nitrobenzene, Wt % Figure 15 Effect of nitrobenzene concentration on overall process rate Note: 0.100 % nitrobenzene @ 140 °C = 0.00813 M (140 °C, 14 barg, 0.05 % catalyst). where

ΔHreaction = Heat of reaction based on Scale-up Calculations nitrobenzene The maximum process rates may be estimated for (–536.6 ± 5.9 kJ/mole). a variety of simplified cases. For example, the maxi- k(sec–1·wt % cat.–1) = Process rate constants from mum rate based on the gas mass transfer assumes pulse experiments. that the bulk hydrogen concentration is equal to zero. Although this situation is generally not desired in a Eventually, the feed is stopped for the run in Figures real process, it does give an upper limit on the maxi- 13 and 14 and the process rate achieves a new yet mum rate possible in a reactor system based on the much lower overall value. At this point the reactor is gas mixing. We can define the Ma ximum Gas Mass operating in a «batch» mode. The low rate is appar- Transfer Rate, MGMTR, as ently caused by the affinity of nitrobenzene, azoxybenzene and perhaps azobenzene for the catalyst and (23) MGMTR = kla · CH2,sat their competition with hydrogen for surface sites. This «poisoning» of the catalyst by process intermedi- In our system, at 140 °C and 1000 rpm, –1 ates and nitrobenzene must be avoided to maintain MGMTR = 0.235 sec · 0.0186 M, high production rates. Not until the concentrations MGMTR = 0.00437 moles hydrogen/(liter · sec). of nitrobenzene and the intermediates have been reduced does the rate begin to recover, increasing at In terms of reactor energy production, the end of the process. (24) MGMTR = kla · CH2,sat · ß–1 · –ΔHreaction Without the sensitive measurements of the heat generation curves, the overall reaction rates observed in where this regime may appear to be constant using conven- ß = Stoichiometric factor, moles hydrogen/ tional rate measurement techniques. This may explain moles substrate (ß = 3 for nitrobenzene). why many investigators report that under batch condi- MGMTR = 782 watts/liter in this study at 140 °C tions at constant pressure, the reaction order is zero at @ 1000 rpm. with respect to nitrobenzene or conversion. From the The hydrogen in the bulk phase during the reaction heat generation curves and concentration profiles, can be calculated by combining Equations 3 and 23 this conclusion is obviously an oversimplification of to get the following: the rate process. These lower process rates at high nitrobenzene concentration indicate that the process (25) CH2,bulk = CH2,sat · (1–Rate/MGMTR) has changed from liquid/solid mass transfer control to a more complex, mixed control regime impacted most Consider operating the reduction of nitrobenzene in a likely by adsorption, interparticle mass transfer and large-scale reactor continuously with sufficient catalyst surface kinetics. and feed rate to give a steady-state energy production

ALR Application Note 12 rate of 300 watts/liter. If the reactor has a MGMTR of The catalyst requirement can be calculated: 600 watts/liter as calculated from Equation 23, then wt% catalyst = 0.000932 moles aniline/ the hydrogen concentration in the bulk at steady-state (liter·sec)/(0.467 sec–1 · wt % cat.–1 can be estimated to be half of the saturation concen- · 0.004065 M nitrobenzene) tration based on Equation 25. To operate a laboratory = 0.49 % reactor at the same bulk hydrogen concentration conditions, either the pressure, the mass transfer coef- Conclusions ficient or both must be adjusted according to equation The hydrogenation of nitrobenzene is a complex 25 to match so that the following condition is met: process impacted by the competing effects of mass transfer and kinetics. These processes, however, CH · (1-Rate /MGMTR ) = 2,sat,1 1 1 (26) can be individually analyzed to reveal the controlling CH2,sat,2 · (1-Rate2/MGMTR2) mechanisms under different process conditions. The reaction pathway according to Haber is deter- where the subscript 1 designates the scaled-up mined by reactor process conditions. At high tempera- reactor at pressure P1 and subscript 2 designates the ture and low nitrobenzene concentration, solid/liquid laboratory reactor at P2. mass transfer was found to dominate, yielding a process that could be operated at high rates. However, at To achieve similar catalyst interfacial concentrations higher nitrobenzene concentration intermediates form, of hydrogen and nitrobenzene in both the laboratory which along with nitrobenzene poison the catalyst and reactor and the plant reactor, the reactor should be lead to low process rates. operated at the same volumetric production rate and catalyst concentration. As long as the bulk hydrogen concentrations in both reactors are the same, then References the concentration profiles in the large reactor and the 1. McMurry, J., Organic Chemistry, Brooks/Cole laboratory will be equivalent. Publishing Company, Monterey, Calif., p. 961 (1984). Example: 2. Rihani, D.N., T.K. Narayanan and L.K. It is desired to operate a large scale-up CSTR at Doraiswamy, «Kinetics of Catalytic Vapor Phase 500 watts/liter, which corresponds to an aniline Hydrogenation of Nitrobenzene to Aniline» production rate of 0.932 moles aniline/(m3·sec) or I&EC Process Design and Development, Vol. 4, 0.000932 moles aniline/(liter · sec). No. 4, pp. 403-410 (1965). The reactor is operated continuously at 140 °C, 3. Loppidana, M.A.A., et al., «Hydrogenation of 8 barg, with a volume of 1 m3, a kla of 0.5 sec–1 and Aromatic Nitro Compounds with a a feed of 50 % aniline/50 % nitrobenzene. Calculate Pt/SiO2-AlPO4 Catalyst» Bull. Chem. Soc. Jpn., the bulk hydrogen concentration and the required 60, pp. 3415-3419 (1987). catalyst. 4. Schorlemmer, C., «Reaction Kinetics of Hydrogenation of Nitrobenzene to Aniline in the The MGMTR is calculated from Equations 23 and 24: Liquid Phase» Chem. Techn., Vol. 33, No. 1, MGMTR = 0.5 sec–1·0.0090 M pp. 24-28 (1981). = 0.0045 moles hydrogen/(sec liter) 5. Yao, H.C., and P.H. Emmett, «Kinetics of Liquid MGMTR = 0.0045 moles hydrogen/(sec·liter)/3 Phase Hydrogenation. II. Hydrogenation of · 536,600 J/mole nitrobenzene Aromatic and Aliphatic Nitrocompounds Over a = 805 watts/liter Colloidal Platinum Catalyst» J. Am. Chem. Soc.,

CH2,bulk = (1–500/805) · 0.0090 M = 0.0034 M Vol. 83, pp. 796-799 (1961). 6. Yao, H.C., and P.H. Emmett, «Kinetics of Liquid To maintain operation in the low nitrobenzene range, Phase Hydrogenation. II. Hydrogenation of we chose to operate at 0.05 wt % or 0.004065 M Aromatic and Aliphatic Nitroconpounds Over a nitrobenzene in the reactor. Colloidal Platinum Catalyst» J. Am. Chem. Soc., The process rate constant in this region of operation is Vol. 83, pp. 796-799 (1961). 0.467 sec–1 · wt % cat.–1, according to Table 3.

ALR Application Note 13 References 7. Yücelen, F., Thesis Dissertation, Effects of Mass 17. Kato, Y., et al., «Gas Holdup and Overall Transfer in Liquid Phase Catalytic Consecutive Volumetric Absorption Coefficients in Bubble Hydrogenation of 2,6-Dinitrotoluene, Swiss Columns with Suspended Solid Particles. Federal Institute of Technology, Zürich (1984). Absorption Rate of Oxygen by an Aqueous 8. Smith, G.V., R. Song, M. Gasior, R.E. Malz, Solution of Sodium Sulfite», Intenrnational Jr., «Hydrogenation of Nitrosobenzene Over Chemical Engineering, Vol. 13, No. 3 (1973). Palladium Catalysts» Preprints of Pagers and 18. Murugesan, T., and T. E. Degaleesan, «Hold- Posters of the 14th Conference on Catalysis up, Interfacial Area and Power Requirements in of Organic Reactions, edited by J. R. Kosak and Turbine Agitated Gas-Liquid Contactors», Chem. T. A Johnson, Albuquerque, New Mexico, Eng. Comm., Vol. 117. pp. 263-278 (1992). April 27-29, 1992. 19. Dietrich, E., C. Mathieu, H. Delmas and 9. Mettler Toledo Inc., 69 Princeton-Heightstown J. Jenck, «Raney-Nickel Catalyzed Hydroge- Rd., P.O. Box 71, Heightstown, nations: Gas-Liquid Mass Transfer in Gas- N.J. 08520-0071. Induced Stirred Slurry Reactors», Chemical 10. Strätz, A. M., Catalysis of Organic Reactions, Engineering Science, Vol. 47, No. 13/14 (1992). Vol. 18, Chapter 17, p. 337, Dekker (1984). 20. Nienow, A.W., Mixing in the Process Industries, 11. Campbell, A. N., L Am. Chem. Soc., 67, Butterworths Series in Chemical Engineering, pg. 981 (1945). Chapter 18, London (1985). 12. Alexejew, W., Ann. Phvs. (Leipzig), Neue Folge 21. Beenackers, A.A.C.M., and W.P.M. Van Swaaij, 28, pg. 305 (1886). «Mass Transfer in Gas-Liquid Slurry Reactors» 13. Tatterson, G.B., Fluid Mixing and Gas Dispersion Chemical Engineering Science, Vol. 48, No. 18, in Agitated Tanks, McGraw-Hill, pp. 3109-3139 (1993). New York (1991). 22. Cheremisinoff, N.P., (Editor), Encyclopedia of 14. Ramachandran, P.A., and R.V. Chaudhari, Three- Fluid Mechanics, Vol. 5, Slurry Flow Technology, Phase Catalytic Reactors, Topics in Chemical Gulf Publishing Company, Houston, pp. 685-716 Engineering, Vol. 2, Chapters 9, Gordon and and pp. 197-209 (1986). Breach Science Publishers, Philadelphia (1983). 23. Reid, R.C., J.M. Prausnitz and T.K. Sherwood, 15. Yagi, H., and F. Yoshida, «Gas Absorption The Properties of Gases and Liquids, 3rd ed., by Newtonian and Non-Newtonian Fluids in McGraw-Hill Book Company, New York, Sparged Agitated Vessels», Ind. Eng. Chem. pg. 567 (1977). Process Des. Dev., Vol. 14, No. 4 (1975). 16. Yoshida, F., and K. Akita, «Performance of Gas Bubble Columns: Volumetric Liquid Phase Mass Transfer Coefficient and Gas Hold-up», AIChE This lecture was held at the 7th RC User Forum USA Journal, Vol. 11, No. I (1965). in St. Petersburg Beach, Florida, in October 1994.

Mettler-Toledo AutoChem Inc. www.mt.com/AutoChem 7075 Samuel Morse Drive For more Information Columbia, MD 21046, USA Phone +1-410 910 8500 Fax +1-410 910 8600

Mettler-Toledo AG, AutoChem Sonnenbergstrasse 74 CH-8603 Schwerzenbach, Switzerland Quality certificate. Development, production Phone +41-44 806 77 11 and testing according to ISO 9001. Fax +41-44 806 72 90

Internet www.mt.com/autochem Environmental management system E-Mail [email protected] according to ISO 14001.

Subject to technical changes. ©11/2007 Mettler-Toledo AG European conformity. The CE conformity Printed in Switzerland, mark provides you with the assurance that Marketing ALR our products comply with the EU directives.