Ind. Eng. Chem. Res.

2005, 44, 261-266 261

KINETICS, CATALYSIS, AND REACTION ENGINEERING

Kinetics of p-Xylene Liquid-Phase Catalytic Oxidation to

Terephthalic Acid
Qinbo Wang,* Xi Li, Lijun Wang, Youwei Cheng, and Gang Xie
Department of Chemical Engineering, Zhejiang University, Hangzhou,
310027 Zhejiang, People’s Republic of China

A fractional kinetic model for the liquid-phase oxidation of p-xylene to terephthalic acid catalyzed
by cobalt acetate and manganese acetate and promoted by hydrogen bromide was proposed.
The developed model parameters were determined in a nonlinear optimization, minimizing the
difference between the simulated and experimental time evolution of the product composition
obtained in a semibatch oxidation reactor where the gas and liquid phases were well mixed.
The experiments included four values of the initial concentration of p-xylene. Further, the effects
of the oxygen partial pressure, reaction temperature, catalyst concentration, and promoter
concentration on the time evolution of the experimental product distribution and kinetic constants
of the developed model were investigated.

Introduction posed the lumped kinetic model as shown in Figure 1

for the liquid-phase oxidation of PX to TA,4 where the
As one of the most important aromatic compounds, reactions of PX to TALD and PT to 4-CBA involve the
terephthalic acid (TA) is widely used in organic synthe- addition of 1O2 and the reactions of TALD to PT and
sis, particularly in the polyester industry. Commercially, 4-CBA to TA involve the addition of 1/2O2. Cao and
the majority of TA is produced by the air oxidation of Cincotti studied the kinetic scheme at low temperature
p-xylene (PX) in acetic acid (HAc) in the temperature and assumed that all the reactions were zeroth order
range from 180 to 205 °C, catalyzed by cobalt acetate with respect to gaseous reactants under the condition
(Co(Ac)2) and manganese acetate (Mn(Ac)2) and pro- that the oxygen partial pressure was higher than a
moted by hydrogen bromide (HBr).1 To gain better minimum value and first order with respect to liquid
insight into the reaction mechanism and identify the reactants.5-8 Yan studied the oxidation kinetics at high
effects of different parameters on the progress of the temperature and also assumed that all the reactions
reaction, it is essential to study the kinetics. Further, were 0.65th order with respect to PX and first order with
the rational design, optimization, control, and analysis respect to the other liquid reactants, and the effect of
of the oxidation of PX to TA process also require oxygen partial pressure on the reaction rate was ne-
knowledge of the kinetics. glected.4 However, when the nth reaction kinetics was
The oxidation of PX to TA occurs through radical used in an industrial oxidation process, there was a
chain elementary reactions, which involve a very large significant difference between the simulated results and
number of radicals as well as molecular species. Eman- the industrial operating results. It was mainly because
uel proposed a detailed radical chain mechanism of the the kinetics previously obtained was empirical and did
liquid-phase cobalt catalyzed oxidation of PX.2 However, not reveal the reaction mechanism sufficiently. Re-
the complexity of this reacting system clearly prevents cently, Wang developed the kinetic model as shown in
the evaluation of the individual values of the kinetic eq 1 from the radical chain reaction mechanisms.9
constant by direct fitting of the model results against
the experimental data, mostly because of the inability Cj
of measuring the concentration of the radical species. r j ) kj j ) 1-4 (1)
The most common approach is to lump the detailed 4
mechanism into a set of global reactions which involve
only molecular species, whose concentration can be, in
( ∑
i)1
diCi + θ)
βj

principle, easily monitored as a function of time.3

Without involving a formal procedure of general validity When eq 1 was used to describe the effect of the
but simply including the minimum number of reactions reaction factors on the reactions in the industrial
to describe the behavior of all the species of interest, reactor, the predicted oxidation results agreed with the
various lumped kinetic schemes for the homolytic oxida- industrial results satisfactorily.10 In this work, the
tion process have been developed in the literature for experiments to determine the kinetic model constants
PX oxidation. By accounting for the most important of eq 1 were reported. The experiments included four
intermediates and final products of the process, that is, values of the initial concentration of p-xylene. Further,
p-tolualdehyde (TALD), p-toluic acid (PT), 4-carboxy- the effects of the oxygen partial pressure, reaction
benzaldehyde (4-CBA), and TA, Yan has recently pro- temperature, catalyst concentration, and promoter con-
10.1021/ie049372x CCC: $30.25 © 2005 American Chemical Society
Published on Web 12/23/2004
262 Ind. Eng. Chem. Res., Vol. 44, No. 2, 2005

Semibatch Reactor Model

By taking into account only the formation of the
molecular species that represented the most important
intermediate and final product, the lumped kinetic
scheme as shown in Figure 1 and the kinetic model of
eq 1 was used. The description of the diffusion and
Figure 1. Lumped kinetic scheme for the oxidation of p-xylene
reaction processes at the gas-liquid interface was
to terephthalic acid. neglected because of the elimination of mass transfer
influence. Since all experiments were performed under
Table 1. Operating Conditions for the Experimental the kinetic regime, the following mass balances were
Runs used:
CPX,0 CCo CMn CBr
(mol/ P O2 (104 kg/ (104 kg/ (104 kg/ dCj 4
run T (°C) kgHAc) (kPa) kgHAc) kgHAc) kgHAc)
dt
) ∑
i)1
νi,jrj j ) 1-4 (2)
1 191 3.145 40.0 350 326 475
2 191 1.887 40.0 350 326 475
3 191 0.943 40.0 350 326 475 along with the initial conditions
4 191 0.472 40.0 350 326 475
5 191 0.943 12.0 350 326 475 Cj ) Cj,0 j ) 1-4 (3)
6 191 0.943 20.0 350 326 475
7 191 0.943 28.0 350 326 475
8 185 0.943 40.0 350 326 475
The energy balance was not considered here, since the
9 188 0.943 40.0 350 326 475 reactor had been operated under isothermal conditions.
10 194 0.943 40.0 350 326 475 The kinetic parameters were determined in a non-
11 197 0.943 40.0 350 326 475 linear optimization, minimizing the difference between
12 191 0.943 40.0 250 233 339 the simulated and the experimental time evolution of
13 191 0.943 40.0 450 419 610 the product composition of experimental runs 1-4. The
14 191 0.943 40.0 550 512 746
15 191 0.943 40.0 350 326 237
fourth-order Runge-Kutta method was used to solve
16 191 0.943 40.0 350 326 712 eq 2, and the simplex method was used in the nonlinear
17 191 0.943 40.0 350 326 949 optimizations. The method is implemented in the Mat-
lab Optimization Toolbox (The Mathworks). The simu-
centration on the time evolution of the experimental lation was written in Matlab to use their optimization
product distribution and kinetic constants of the devel- routine.
oped model were investigated.
Results and Discussion
Experimental Setup and Procedure Kinetic Parameter Evaluation. The time evolution
of the experimental product composition along with the
The experimental apparatus, sampling apparatus, simulated value for experimental runs 1-4 is sum-
and analytical methods used in this work were described marized in Figure 2. A comparison between the simu-
elsewhere.11 Briefly, it consisted of a 500 mL titanium lated results and the experimental data is also shown
reactor maintained at the desired temperature through in Figure 2, where it can be seen that the obtained
forced circulation of diathermic oil and wall electric agreement is in general satisfactory. The average rela-
heating. Both the liquid and gaseous phases were tive deviation of the simulated value and the experi-
continuously stirred. The system was equipped with mental data is <5%. This model is able to predict the
three condensers in order to ensure complete condensa- reactor behavior as a function of the concentration of
tion and recycling of the evaporated compounds. The the liquid reactants. The obtained kinetic parameters
reactor temperature was continuously monitored during are given in Table 2.
the experimental runs with a thermal resistance ther- Effect of Oxygen Partial Pressure. In the litera-
mometer. In a typical experimental run, the reactor was ture, when the PX oxidation process was studied, the
charged with 300 mL mixtures of proposed PX, HAc, oxygen partial pressure was kept at a high value so that
and catalyst. After the temperature reached the desired the effect of the oxygen partial pressure on the oxidation
value, the gas (air) was then continuously fed through process could be neglected. Cao and Cincotti studied the
the liquid. The experimental runs were carried out PX oxidation process with air and pure oxygen catalyzed
under the conditions summarized in Table 1. The by the cobalt naphthenate in a semibatch reactor,5-8 but
reproducibility of the experimental runs was verified by the studies were carried out at a low temperature of
repeating each of them at least twice. The experimental 80-130 °C and atmospheric pressure. They assumed the
runs reported in Table 1 were performed at 900 rpm, PX oxidation process was zeroth order with respect to
since the influence of stirring speed of this value on the the oxygen partial pressure. Digruov found that the
product distribution was found to be negligible. A special oxidation of PX was independent of the oxygen partial
sampler was used to sample the reaction products. The pressure at a temperature of 70-105 °C and an oxygen
low boiling temperature components such as PX, HAC, partial pressure of 20-100 kPa.12 Unfortunately, in the
and TALD were analyzed by gas chromatography (GC) industrial reactors, the volume fraction of oxygen in the
( Shimadzu GC-9A), and the high boiling temperature offgas is controlled within 4-5% for safety10 and the
components such as PT, 4-CBA, and TA were analyzed corresponding oxygen partial pressure is 10-20 kPa. At
by high pressure liquid chromatography (HPLC) (Shi- such a low oxygen partial pressure, the reaction is no
madzu-6A). Isopropyl benzene was used as the internal longer zeroth order with respect to the oxygen partial
standard substance to correlate the data obtained from pressure. To obtain a kinetic model which can be used
GC and HPLC analyses. in industrial processing, experiments of different oxygen
 Ind. Eng. Chem. Res., Vol. 44, No. 2, 2005 263

Figure 2. Evolution of the experimental and simulated product composition for experimental runs 1-4. The parameters in the simulation
are from the optimization, Table 2.

Figure 3. Evolution of the experimental and simulated product composition for experimental runs 3 and 5-7. The parameters in the
simulation are from the optimization.

Table 2. Optimal Parameters from Optimizations with Table 3. Optimal Reaction Rate Constants for
Experimental Runs 1-4 Experimental Runs 3 and 5-17
ki (min-1) di (kgHAc/mol) β θ run k1 (min-1) k2 (min-1) k3 (min-1) k4 (min-1)
i)1 0.1716 1.4247 0 0.0146 3 0.176 0.725 0.0361 0.338
i)2 0.7002 0 0.5254 5 0.138 0.547 0.0276 0.257
i)3 0.0353 0 0 6 0.162 0.656 0.0329 0.307
i)4 0.3296 4.8419 0.8111 7 0.176 0.723 0.0366 0.339
8 0.139 0.590 0.0256 0.248
9 0.153 0.650 0.0303 0.288
partial pressures were performed. The experimental 10 0.189 0.762 0.0409 0.386
conditions are listed in Table 1. The experimental 11 0.216 0.863 0.0479 0.436
results obtained from experimental runs 3 and 5-7 are 12 0.119 0.513 0.0250 0.253
13 0.211 0.841 0.0472 0.433
illustrated in Figure 3, and the rate constants are given 14 0.217 0.908 0.0568 0.521
in Table 3. The results show that the effect of the oxygen 15 0.099 0.507 0.0232 0.220
partial pressure on the reactions decreases with the 16 0.205 0.837 0.0473 0.461
increase of partial pressure over a threshold value 17 0.219 0.855 0.0578 0.531
of 28.0 kPa. In the range 10-20 kPa of industrial to model the industrial reactor, correction coefficients
reaction conditions, the oxygen partial pressure influ- must be introduced into the rate constants.
ences the reaction significantly. That is one of the Effect of Temperature. PX liquid-phase oxidation
reasons that when the kinetic model of Yan was used at 185, 188, 191, 194, and 197 °C was carried out to
264 Ind. Eng. Chem. Res., Vol. 44, No. 2, 2005

Figure 4. Evolution of the experimental and simulated product composition for experimental runs 3 and 8-11. The parameters in the
simulation are from the optimization.

Figure 5. Evolution of the experimental and simulated product composition for experimental runs 3 and 12-14. The parameters in the
simulation are from the optimization.

Table 4. Optimized Kinetic Constants and Activity ity energy for each oxidation step are shown in Table
Energy 4. The obtained activation energy of different reaction
ki i)1 i)2 i)3 i)4 steps ranges from 54.9 to 92.8 kJ/mol; among them, the
Ei (kJ/mol) 65.5 54.9 92.8 84.9 activation energy of the PT to 4-CBA step is the highest
ki,0 (min-1) 4.07 × 106 1.08 × 106 9.80 × 108 1.17 × 109 (92.8 kJ/mol) and that of the 4-CBA to TA step is the
variance (R2) 0.998 0.995 0.999 0.997 second highest (84.9 kJ/mol); these are remarkably
higher than the values of the other steps (54.9-65.5 kJ/
investigate the temperature effect. Plots of Cj/CPX,0 as
mol). This fact shows that oxidation of the second
a function of time for experimental runs 3 and 8-11
are shown in Figure 4.The kinetic model of eq 1 is methyl group of PX is more sensitive to the temperature
employed to fit the data. A comparison between the variation than the first one.
simulated results and the experimental data is also Effect of Catalyst Concentration. Plots of the
shown in Figure 4. The optimal rate constants at each quantity Cj/CPX,0 as a function of time for each compo-
temperature are shown in Table 3. From Figure 4 and nent at various catalyst concentrations are shown in
Table 3, it can be seen that the reaction temperature Figure 5, while fixing the mass ratio of [Co2+]/[Mn2+]/
has a strong effect on the reaction rate of each step and [Br-] at 1. The values of the kinetic constants appearing
the reaction rate was found to increase with an increase in the lumped kinetic scheme shown in Figure 1 are
in the reaction temperature. The rate constants ob- then estimated by fitting the time evolution of the
tained can be well represented by the Arrhenius rela- experimental product composition of runs 3 and 12-14
tionship and the optimized kinetic constants and activ- in Table 1 through an optimization procedure using the
 Ind. Eng. Chem. Res., Vol. 44, No. 2, 2005 265

Figure 6. Evolution of the experimental and simulated product composition for experimental runs 3 and 15-17. The parameters in the
simulation are from the optimization.

model given by eqs 1-3, and they are shown in Table and the corrosion of the equipment, it is not a wise
3. A comparison between the model results and the choice to run the oxidation process at a high level of
experimental data is also shown in Figure 5, where it [Br-]. In industries, the selected [Br-]/[Co2+] ratio is ∼1/
also may be seen that the obtained agreement is 2-1.
generally satisfactory. This allows us to confirm that
the lumped kinetic scheme retains a level of process Conclusions
description detailed enough to characterize the distribu-
tion of the most important products when changing Some kinetic aspects of the air oxidation of PX in HAc
catalyst concentration within a relatively wide range. in the temperature range from 184 to 197 °C, catalyzed
From Table 3, we can conclude that the reaction rates by Co(Ac)2 and Mn(Ac)2 and promoted by HBr, are
are much more concentration-sensitive at a low catalyst examined in this work from the experimental point of
concentration than at a high catalyst concentration. It view. The effect of various operating variables, that is,
is seen that the rate constants k1 and k2 increase temperature, oxygen partial pressure, initial PX con-
eventually with the increasing catalyst concentration, centration, catalyst concentration, and promoter con-
reaching a plateau, but the rate constants k3 and k4 centration, is analyzed. A model for the description of
increase almost linearly with the increasing catalyst the semibatch gas-liquid reactor for PX oxidation has
concentration. been developed. The proposed reactor model is based
Effect of Promoter. If the catalyst system only on a lumping kinetic scheme and a fractional-like kinetic
including [Co2+] and [Mn2+] is used in the catalytic model of the complex radical chain mechanism of the
oxidation of PX, not only the oxidation rate but also the oxidation process, where only the most important
yield of TA will be very low. A promoter must be compounds detected experimentally, that is, TALD, PT,
introduced into the catalytic system.13 HBr is a common 4-CBA, and TA, are taken into account. The model
promoter widely used in the industrial oxidation of PX parameters and rate constants are determined in a
to TA. It is necessary to study the effect of [Br-] on the nonlinear optimization, minimizing the difference be-
oxidation rate. Plots of the quantity Cj/CPX,0 as a tween the simulated and experimental time evolution
function of time for experimental runs 3 and 15-17 in of the product composition. A comparison between model
Table 1 are shown in Figure 6. The optimal values of prediction and experimental data is also presented, and
the kinetic constants are shown in Table 3. A compari- the agreement is quite satisfactory. The obtained results
son between the simulated results and the experimental confirm the fractional kinetic model.
data is also shown in Figure 6, where it also may be
seen that the obtained agreement is generally satisfac- Acknowledgment
tory. The financial support by the National Science Foun-
A concentration effect the same as the catalyst dation of China (no. 20076039) and SINOPEC (no.
concentration effect on the oxidation rate is found for X500029) is gratefully acknowledged.
[Br-]. The reaction rates of PX to TALD and TALD to
PT increase by 25%; also the reaction rates of PT to
4-CBA and 4-CBA to TA increase by more than 60% Notation
when [Br-]/[Co2+] increases from 1 to 2, which clearly Ci ) concentration of ith component, mol/kgHAc
shows that [Br-] accelerates the reaction rate signifi- di ) model parameters in eq 1
cantly. The higher [Br-] is, the faster the reaction will ki ) the rate constants, min-1
be. However, considering the discharge of the pollutants ri ) rate of the ith step reaction, mol/(kgHAc‚min)
266 Ind. Eng. Chem. Res., Vol. 44, No. 2, 2005

[i] ) mass concentration of substance i, kg/kgHAc (4) Yan, X.; Du, W.; Qian, F. Development of a Kinetic Model
for Industrial Oxidation of p-Xylene by RBF-PLS and CCA. AIChE
Greek Letters J. 2004, 50 (6), 1169-1176.
β and θ ) kinetic parameters (5) Cao, G.; Pisu, M.; Morbidelli, M. A Lumped Kinetic Model
for Liquid-phase Catalytic Oxidation of p-Xylene to Terephthalic
νij ) stoichiometric coefficient of the jth component in the Acid. Chem. Eng. Sci. 1994, 49 (24), 5775-5788.
ith reaction (6) Cao, G.; Servida, A.; Pisu, M. Kinetics of p-Xylene Liquid-
Subscripts phase Catalytic Oxidation. AIChE J. 1994, 40 (7), 1156-1166.
(7) Cincotti, A.; Orrù, R.; Bori, A.; et al. Effect of Catalyst
0 ) initial conditions Concentration and Simulation of Precipitation Processes on
Liquid-phase Catalytic Oxidation of p-Xylene to Terephthalic Acid.
Abbreviations Chem. Eng. Sci. 1997, 52 (21), 4205-4213.
4-CBA ) 4-carboxybenzaldehyde (8) Cincotti, A.; Orrù, R.; Cao, G. Kinetics and Related Engi-
Co(Ac)2 ) cobalt acetate neering Aspects of Catalytic Liquid-phase Oxidation of p-Xylene
to Terephthalic Acid. Catal. Today 1999, 52, 331-347.
HAc ) acetic acid (9) Wang, L. Studies on the Kinetics of p-Xylene Oxidation and
HBr ) hydrogen bromide the Reactor Simulation. MS Thesis, Zhejiang University, China,
Mn(Ac)2 ) manganese acetate 2001.
O2 ) oxygen (10) Wang, Q.; Li, X.; Xie, G.; et al. Modeling of p-Xylene
PT ) p-toluic acid Oxidation Reactor-Condenser. J. Zhejiang Univ.: Eng. Sci. 2004,
PX ) p-xylene 38 (8), 1029-1034.
TA ) terephthalic acid (11) Cheng, Y.; Zhang, L.; Xie, G.; et al. Experiment Technique
TALD ) p-tolualdehyde of p-Xylene Liquid-phase Oxidation. Chem. React. Eng. Technol.
2003, 19 (2), 182-186.
(12) Digruov, N. G.; Dyakonov, J. A.; Lebedev, N. N.; et al.
Literature Cited Kinetics of Liquid-Phase Oxidation of p-Xylene into Terephthalic
Acid with a Cobalt Bromide Catalyst. Izv. Vyssh. Uchebn. Zaved.,
(1) Raghavendrachar, P.; Ramachadran, S. Liquid-Phase Cata- Khim. Khim. Tekhnol. 1970, 13, 407-412.
lytic Oxidation of p-Xylene. Ind. Eng. Chem. Res. 1992, 31 (2), (13) Astarita, G.; Ocone, R. Chemical Reaction Engineering of
453-462. Complex Mixtures. Chem. Eng. Sci. 1992, 47 (9), 2135-2147.
(2) Emanuel, N. M.; Gal, D. Modelling of Oxidation Processes; Received for review July 18, 2004
Akademiai Kiado: Budapest, Hungary, 1986. Revised manuscript received October 24, 2004
(3) Cavalieri d’Oro, P.; Danoczy, E.; Roffia, P. On the Low Accepted November 9, 2004
Temperature Oxidation of p-Xylene. Oxid. Commun. 1980, 1, 153-
155. IE049372X