Você está na página 1de 13

ARTICLE

pubs.acs.org/IECR

Kinetics of Toluene Disproportionation: Modeling and Experiments


Marcos W. N. Lob~ao, Andre L. Alberton, Slvio A. B. V. Melo, Marcelo Embiruc-u,
Jose L. F. Monteiro, and Jose Carlos Pinto*,

Programa de Engenharia Industrial, Escola Politecnica, Universidade Federal da Bahia, Rua Aristides Novis, no 2, Federac-~ao,
Salvador, 40210-630 BA, Brasil

Pontifcia Universidade Catolica do Rio de Janeiro, Rua Marqu^es de S~ao Vicente 225-371 L, Rio de Janeiro, 22451-900 RJ, Brasil

Programa de Engenharia Qumica/COPPE, Universidade Federal do Rio de Janeiro, Cidade Universitaria, CP: 68502,
Rio de Janeiro, 21941-972 RJ, Brasil
ABSTRACT: This work presents modeling and experimental studies on the kinetics of toluene disproportionation in operation
ranges that include real industrial operation conditions. The inuence of reaction temperature, reactor pressure, feed composition,
and residence time on conversion of reactants and product selectivity was investigated. Experiments were performed according to a
sequential experimental design strategy, in order to provide maximum accuracy for model predictions. Statistical treatment of
parameter estimates and model adequacy was performed with the help of maximum likelihood principles. Excellent agreement
between model predictions and available experimental data was obtained in the full ranges of investigated experimental conditions.

derived from mechanisms that consider toluene adsorption3,6,19


or surface reaction3,20 as the RDS. In all these studies, the catalysts
used to promote toluene disproportionation were based on
Y-zeolites,3,15,20 ZSM-5 zeolites,4,5,10,12 and more commonly
H-mordenite.68,10,11,13,14,1618
As briey reviewed above, although the kinetics of toluene
disproportionation has been investigated by several researches,
distinct reaction mechanisms have been proposed and distinct
kinetic models have been derived. Discrepancies can possibly be
ascribed to the dierent analyzed catalyst types, operating
conditions, conversion levels, and coking of catalysts; nevertheless, it is certain that agreement has yet to be reached
regarding the kinetics of toluene disproportionation over heterogeneous catalysts.
It is important to emphasize that many works published in
the literature were performed at low pressures and not at
the relatively high pressures practiced in real industrial
environments.47,12,15,19,20 When the operation conditions
resembled the actual industrial operation conditions, simplied models based on pseudohomogeneous approach were
proposed, specially for the study of catalyst deactivation.810
However, the use of simplied kinetic models leads to a
poorer ts, when compared to the LHHW and ER expressions. 19 Besides, as one can observe in Table 1, many
studies were performed at low toluene conversions, reducing
the reliability of the models for prediction of real industrial
operation conditions.
Based on the previous paragraphs, the use of LHHW models
to t experimental data obtained at operation conditions that
resemble the actual industrial conditions seems appealing. For
this reason, the main objective of the present work is to

1. INTRODUCTION
Toluene disproportionation is an important chemical transformation for most reneries and chemical complexes, constituting the heart of economically relevant commercial
processes used for direct production of aromatic products
and indirect manufacture of polymer resins, synthetic bers,
and plasticizers.1 The toluene disproportionation reaction can
be represented as
2T a B X

R1

where T is toluene, B is benzene, and X is xylenes. Toluene


disproportionation is performed commercially with the help of
heterogeneous catalysts.1,2 One of the most important processes is the Tatoray process, where typical industrial conditions include temperatures of 350530 C, pressures of
1050 bar, and H2/aromatic ratios between 5 and 12/1.1,2
The high hydrogen to aromatic ratio is necessary to avoid the
catalyst deactivation.3
Reliable kinetic studies are very important for design, analysis,
and control of the industrial process. Although most toluene
disproportionation technologies are mature, kinetics of toluene
disproportionation over heterogeneous catalysts is not completely understood even in ranges of experimental conditions
normally employed by industrial processes. Table 1 presents a
summarized review about the kinetic models proposed in the
literature, including the experimental conditions and the conversion ranges investigated by the authors.
Dierent kinetic models have been used so far to describe
the catalytic disproportionation of toluene. Many authors
proposed a pseudohomogeneous kinetics of rst order4,5
or second order610 with respect to the toluene partial pressure in order to describe the reaction rates. Good ts have
been reported for LangmuirHinshelwood-Hougen-Watson
(LHHW) kinetic rate expressions, derived from mechanisms
that consider surface reactions1118 as the rate determining
step (RDS), and EleyRideal (ER) kinetic rate expressions,
r 2011 American Chemical Society

Received: July 17, 2011


Accepted: November 7, 2011
Revised:
October 11, 2011
Published: November 07, 2011
171

dx.doi.org/10.1021/ie2015526 | Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Table 1. Kinetic Studies of Toluene Disproportionation over Zeolites Catalyst (Adapted from Marques17)
year

ref

catalyst

P (bar)

T (C)

kinetic modelsa

XT (%)

1979

(3)

HY/AlF3/Cu

211.1

400500

e22

E-R reaction surface or toluene adsorption as rds

1981

(7)

modied mordenite

350450

<45

pseudohomogeneous 2nd order

1981

(11)

mordenite

919

400

e10

L-H reaction surface as rds

1985

(12)

HZSM-5

240300

e10

L-H surface reaction as rds

1987

(6)

mordenite

350450

e25

pseudohomogeneous 2nd order/E-R


toluene adsorption as rds

1987

(13)

mordenite

530

360430

<45

L-H reaction surface as rds

1990
1990

(20)
(4)

HY and HY-Ni
HZSM-5

3
1

340500
502562

e10
e10

E-R reaction surface as rds


pseudohomogeneous 1st order

1990

(14)

mordenite

5.140.5

-c

e10

L-H surface reaction as rds

1993

(15)

HY

240320

e8

L-H reaction surface as rds

1993

(16)

mordenite

19.629.4

430

e33

L-H reaction surface as rds

1994

(5)

modied HZSM-5

450500

e15

pseudohomogeneous 1st order

1994

(17)

mordenite

528

390434

<38

L-H reaction surface as rds

2000

(19)

HZSM-5

475

e25

E-R toluene adsorption as rds

2004
2004

(8)
(9)

mordenite
-c

28.59
31.2d

397451
347352d

<55
-c

pseudohomogeneous 2nd order


pseudohomogeneous 2nd order

2007

(10)

mordenite and HZSM5

-c

300400

e30

pseudohomogeneous 2nd order

2010

(18)b

mordenite

530

300380

<30

L-H reaction surface as rds

E-R: EleyRideal, L-H: LangmuirHinshelwood. b Also evaluating xylenes disproportionation. c Not reported. d Other investigated conditions not
reported.

investigate the kinetics of toluene disproportionation over a


commercial H-mordenite catalyst in a broader range of
experimental conditions, including the conditions normally
employed to perform the reaction in industrial sites. The
obtained experimental data are then used to build a kinetic
model in order to represent reaction rates in a wider range of
experimental conditions. A sequential experimental design
technique is used to select the experiments, to accelerate the
model building process, and to provide maximum accuracy
for model predictions.

estimated through minimization of the following objective


function (F)21,22
0
 12
ipred i
iexp
Nexp Ny
yj xi ,  yj
x ,
B
C
F
1
@
A
i
i1 j1
y, j

where Nexp represents the number of experiments, Ny is the


number of response variables, y(i)
j is the response variable j for
(i)
(i)
is the standard deviation of variable y(i)
experiment i, y,j
j , and x
is the set of experimental design variables for experiment i. The
superscripts exp and pred represent the experimental value and
the value calculated with the model, respectively.
After minimization of the objective function and estimation of
the model parameters, it is possible to infer the uncertainties of
the parameter estimates. If the experimental errors are suciently small, the covariance matrix of parameter uncertainties,
V, can be calculated21,22
2
!T
!31
Nexp
 1 yi
i
y
i
5
V 4
2
3 Vy
3

i1

2. PARAMETER ESTIMATION AND EXPERIMENTAL


DESIGN
Let us assume that a set of experiments has been performed.
The set of independent experimental conditions dened by the
analyst is represented here by x and named as the set of design
variables (reactor design, mass of catalyst, feed ow rates, reactor
temperature, and reactor pressure). The set of dependent
experimental conditions obtained by the analyst as the experimental results is represented here by y and named as the set of
response variables (outlet compositions). Variables x and y are
related to each other through the mass balance equations and the
kinetic model.
In order to analyze the available experimental data, the following hypotheses are assumed to be valid: (i) experimental
measurements are subject to random experimental errors, which
follow the normal distribution; (ii) experiments are well done, so
that gross and persistent errors are not present; (iii) the proposed
mathematical model is perfect, so that experimental measurements uctuate around model predictions; (iv) design variables
are well controlled and are not subject to signicant experimental
errors; and (v) experimental errors are not correlated to each
other. In this case, the set of kinetic parameters can be

where the matrix Vy(i) is the covariance matrix of experimental


uncertainties for experiment i. As it was assumed that experimental errors are independent
2
6
6
6

V i
y
6
4

172

y1 2

0
l
0

y2 2
l
0

333

333

0
l

333

yNy 2

3
7
7
7
7
5

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Table 2. Properties of the H-Mordenite Catalyst


parameter

value

acidity (mol NH3 3 g1)

0.225

pores volume (cm3 3 g1)

0.66

density (g 3 cm3)

surface area (m2/g)

1.33
330

Chemical composition
Si

18

Al

28

Fe, Zn, Ni, and Cu


Ca, Mg, K, Na

<0.2
<0.3

humidity

8.6

The diagonal elements of V represent the parameter uncertainties, while the nondiagonal elements of V represent the parameter correlations. Finally, the parameter uncertainties are transformed into uncertainties of model predictions as21,22
 
 T
y
y
pred
Vy

4
V
3 3
Sequential experimental design strategies are used to select
experimental conditions that allow for discrimination of rival
models and/or estimation of model parameters with maximum
accuracy.21,2326 In short, experimental design techniques make
use of the available models to predict future experimental data
and infer the experimental conditions that can lead to the most
precise set of parameter estimates, avoiding the costs of performing experiments in regions where the information content is
low.2123 After selection of the experimental conditions, the
experiment must be performed and the model parameters must
be re-estimated. The procedure must be repeated iteratively until
attainment of the desired parameter precision or selection of the
best model candidate.
Several criteria can be used for design of experiments for
estimation of precise model parameters. The y-trace criterion
proposed by Pinto et al.27 is used in the present work. According
to this criterion, the optimal experimental condition is the one
that allows for minimization of the trace of matrix Vypred
presented in eq 4. As explained by Pinto et al.,27 this criterion
is focused on the model performance and does not necessarily
leads to the most precise set of model parameters but to the
lowest prediction uncertainties. As the trace of Vypred must be
calculated through model simulations, the eciency of the
experimental design technique depends strongly on the ability
of the model to predict the experimental data.

Figure 1. (A) Experimental apparatus and (B) reactor scheme.

into the reactor with the help of a metering pump (Whitey 1/3
HP), with ow rates in the range between 0 and 10 mL 3 min1.
The liquid stream was vaporized with the help of external
electrical resistances and mixed with the hydrogen stream. The
hydrogen/aromatic ratio (RHC) was set to 6 (in molar basis) in
all experiments.
A xed bed reactor was used to perform the reactions, as
illustrated in Figure 1B. The reactor was lled with three distinct
particle layers. The catalyst bed was placed between two beds of
silica of similar length. Silica particles presented the same average
diameter of catalyst particles. Experiments performed without
the catalyst (blank experiments) never led to any signicant
conversion of the reactants. The internal reactor temperature was
monitored and controlled with 3 type-J thermocouples placed at
the three solid layers, as shown in Figure 1B. The reactor was
placed in a vessel containing uidized alumina pellets and
electrical resistances for improved control of the reactor temperature. Additional details about the experimental apparatus can
be found elsewhere.13

3. METHODOLOGY
3.1. Experimental Setup and Procedure. The acidity, pore
volume, surface area, and chemical composition of the catalyst
are presented in Table 2. The catalyst used in the present work is
a commercial H-mordenite grade presenting low concentrations
of Ca, K, and Na, which indicate that hydrogen constitutes the
only compensation cation in the catalytic system. The catalyst
presents a low atomic Si/Al ratio of 0.62. Additional details on
the catalyst properties are described elsewhere.13
Reaction experiments were carried out in a laboratory unit, as
illustrated in Figure 1A. Aromatic compounds (toluene, benzene,
xylenes) kept in the liquid phase in a storage reservoir were fed
173

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

where s represents a catalyst site and RDS indicates the rate


determining steps. According to the mechanism presented in R3,
the following rate expressions could be derived for toluene and
xylenes disproportionation reactions

The outlet stream was condensed, and samples of the liquid


stream were analyzed with the help of a CG-25 gas chromatograph equipped with a thermal conductivity detector (TCD).
Hydrogen was used as the carrier gas. The chromatograph
columns were made out of bentona 34 on chromosorb P, suitable
for separation of aromatic compounds.28
A reference experiment was performed every three experiments in order to guarantee that no signicant deactivation of the
catalyst had occurred. Whenever observed conversion deviations
were larger than 5% for the reference experimental condition, the
catalyst bed was replaced and the previous three experiments
were discarded. Experiments performed with catalysts of dierent sizes and with dierent ow-rates13,29 indicated that the
external and internal resistance to mass transfer were not
signicant at the analyzed operation conditions.
3.2. Modeling. The starting point of this work was the study
developed by Krahl,13 who investigated the gas phase disproportionation of toluene on a commercial H-mordenite catalyst
and proposed LHHW kinetic rate expressions to fit the obtained
experimental data. As observed experimentally, the product
distribution could be described very well by the kinetic model
derived from the following reaction mechanism
2T a B X

R1

2X a T TMB

R2

r1

r2

k2 3 KX2 3 PX2  PT 3 PTMB =KEX


1 KT 3 PT KX 3 PX KB 3 PB KTMB 3 PTMB 2

where r1 and r2 are the reaction rates for toluene and xylenes,
respectively; k1 and k2 are kinetic rate parameters; Kj is the
adsorption equilibrium parameter for component j; Pj is the
partial pressure of component j (j = T, X, B, TMB); and KET and
KEX are the reaction equilibrium constants for toluene and
xylenes disproportionation, respectively.30 The kinetic rate parameters (k1 and k2) and adsorption equilibrium parameters (Kj) can
be described as function of the reactor temperature

where TMB is a pool of trimethylbenzenes. The author proposed


several reaction mechanisms and derived a number of LHHW
and ER kinetic models to explain the available data, but the best
results were obtained with the following reaction scheme
T s a Ts
2Ts a Bs Xs RDS
Bs a B s
Xs a X s
2Xs a Ts TMBs RDS
TMBs a TMB s

k1 3 KT2 3 PT2  PB 3 PX =KET


1 KT 3 PT KX PX KB 3 PB KTMB 3 PTMB 2

ki PFi 3 eEi =R 3 TK , i 1, 2

Kj
eSj =R  Hj =R 3 TK , j T, X, B, TMB
bar 1

where PFi is the pre-exponential factor, and Ei is the activation


energy for reaction i; Sj is the entropy and Hj is the enthalpy
of adsorption for component j, and TK is the reactor temperature. Correlation between pre-exponential factors and activation energies can be minimized if the rate expressions are
reparameterized as3133

R3


TK  Tr
TK
2
e
r1
3 PT  PB3 PX =KET 
!0






12
TK  Tr
TK  Tr
TK  Tr
TK  Tr
mol
A

B
A

B
A

B
A

B
T
T3
X
X3
B
B3
TMB
TMB 3
B
C
TK
TK
TK
TK
gcat 3 h 3 bar 2
B
C
e
e
e
B1 e
C
P
P
P
P

T
X
B
TMB C
3
3
3
3
B
bar
bar
bar
bar
@
A
A1 B1 3

9



TK  Tr
TK
2
e
r2
3 PTMB =KEX
3 PX  PT 
!0






12
TK  Tr
TK  Tr
TK  Tr
TK  Tr
mol
A

B
A

B
A

B
A

B
T
T3
X
X3
B
B3
TMB
TMB 3
B
C
TK
TK
TK
TK
gcat 3 h 3 bar 2
B
C
e
e
e
B1 e
C

P
P
P
P
T
X
B
TMB
3
3
3
B
C
bar
bar
bar
bar
@
A
A2 B 2 3

10
where A1, B1, A2, B2, AT, BT, AX, BX, AB, BB, ATMB, and BTMB are
the parameters that must be estimated and Tr is a reference

temperature, made equal to 666.9 K in order to minimize the


correlation between the parameter estimates.3133 The values of
174

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Table 3. Sequence of Designed Experiments in the Temperature of 360 C, with y-Trace Criterion
aromatic molar fractions in the feed
P (bar)

(gcat 3 h/garomatic)

31.19

0.0563

30.54

0.1000

30.82

0.2300

30.93

yT

yB

aromatic molar fractions outlet


yX

RHC

yT

yX

7.06

0.9749

0.0117

6.43

0.9524

0.0228

6.15

0.8991

0.0519

0.3630

6.21

0.8697

0.0685

30.87

0.3900

6.53

0.8587

0.0671

30.14

0.1497

6.16

0.0797

0.8422

30.22

0.2500

6.03

0.1211

0.7616

30.22
10.18

0.0970
0.1496

0
0

0
0

1
1

5.98
6.03

0.0574
0.024

0.8863
0.9519

10.18

0.2503

5.94

0.0314

0.9375

5.30

0.1994

5.94

0.0128

0.9736

5.20

0.2989

5.97

0.0199

0.9603

5.30

0.4000

6.07

0.0264

0.9469

30.22

0.0996

0.6081

0.3919

6.12

0.5968

0.0050

30.12

0.1991

0.6081

0.3919

5.98

0.5840

0.0113

30.12

0.2988

0.6081

0.3919

6.05

0.5733

0.0165

the kinetic parameters presented in eqs 9 and 10 can be recovered


from the estimated parameter values as
PF1 eA1  2 3 AT B1  2 3 BT

11a

E1 B1  2 3 BT 3 R 3 Tr

11b

PF2 eA2  2 3 AX B2  2 3 BX

11c

E2 B2  2 3 BX 3 R 3 Tr

11d

Sj R 3 Aj Bj , j T, X, B, TMB

11e

Hj Bj 3 R 3 Tr , j T, X, B, TMB

11f

3.3. Numerical Procedures. The set of ordinary eqs 12a-12d


was solved numerically with a fourth-order RungeKutta
method.35 The relative precision of the integration was always
better than 1.0  105. Model parameters were estimated as
described in Section 2, using the well-known GaussNewton
method.21,22 Distinct initial guesses were used in order to avoid
local minima. The relative precision of the estimation step was
always better than 1.0  104. All numerical procedures were
implemented in Fortran.36
3.4. Experimental Design. As the number of model parameters to be estimated was high, it was convenient to group the
model parameters in terms of the reactor temperature. Analyzed
reaction temperatures were equal to 360, 380, 404, and 431 C,
based on the work of Krahl.13 For the temperature level of
360 C, sixteen experiments were carried out, including the data
reported by Krahl,13 which were generated without the proposition of statistical experimental designs. For temperatures of 380,
404, and 431 C, the following experimental design procedure
was adopted:
(i) a set of initial experiments was selected, including data
reported by Krahl;13
(ii) the set of experiments was performed;
(iii) model parameters were estimated and the results evaluated;
(iv) for a given temperature, the experimental design procedure was interrupted when suciently low prediction
errors had been obtained;
(v) otherwise, a new experiment was designed, as described in
Section 2, before returning to step (ii).
Some experiments were also performed for feeds containing
mixtures of toluene and benzene, in order to estimate the
parameters KB and KX independently. Experimental conditions
are presented in Tables 36.
In order to perform step (iv), it was assumed that reaction
conditions could be manipulated in the following ranges:
the total pressure (P) could assume the values of 5, 10, and
30 bar, which are typical values of real industrial operations;1,2

The reactor was described as a standard plug ow reactor


(PFR), which was supported by the high Peclet number (above
80) and the large ratio between the reactor diameter and the
catalyst pellet diameter (higher than 20).34 Thus, the individual mass balance equations could be written as
1 dyT
r2
 r1
M 3 d
2

12

1 dyX
r1
 r2
3
M d
2
 0

3 3 yB 2 3 y0T y0X  2 3 yT  yX
yB
3

12b

yTMB 1  yB  yT  yX

12d

12c

where y T, yX, yB, and yTMB are the molar fractions of species T,
X, B, and TMB, respectively, considering only the aromatics
species; is the spatial time (in gcat 3 h/garomatic); and M is the
mean molecular weight of the aromatic feed. The terms y 0T, y0X,
and y0B represent the feed conditions of yT, y X and yB.
175

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Table 4. Sequence of Designed Experiments in the Temperature of 380 C, with y-Trace Criterion
aromatic molar fractions in the feed
P (bar)

(gcat 3 h/garomatic)

30.55

0.3590

30.68

0.1900

30.42

0.1070

30.97

0.0516

30.97

yT

yB

aromatic molar fractions outlet


yX

RHC

yT

yX

5.82

0.7472

0.1182

5.91

0.8369

0.0802

6.51

0.8954

0.0509

6.61

0.9496

0.0236

0.0475

5.96

0.9468

0.0258

30.24

0.2667

0.6041

0.3959

6.61

0.5516

0.0361

30.14

0.1005

0.6041

0.3959

6.66

0.5795

0.0169

5.399
5.399

0.3992
0.1999

1
1

0
0

0
0

6.25
6.10

0.9685
0.9810

0.0167
0.0099

10.28

0.2975

5.89

0.9562

0.0219

10.28

0.1051

6.53

0.9796

0.0104

30.14

0.2574

6.09

0.2215

0.5558

30.14

0.0998

6.09

0.1308

0.7429

30.24

0.1497

6.09

0.1599

0.6833

10.38
30.12

0.1497
0.0500

0
0

0
0

1
1

6.02
6.04

0.0756
0.0624

0.8525
0.8775

Initial Set of Experiments

Designed Experiments

Table 5. Sequence of Designed Experiments in the Temperature of 404 C, with y-Trace Criterion
aromatic molar fractions in the feed
P (bar)

(gcat 3 h/garomatic)

30.14

0.4060

31.05

0.3590

30.93

0.2350

30.90
31.54

0.0977
0.0512

30.14
30.24

yT

yB

aromatic molar fractions outlet


yX

RHC

yT

yX

6.56

0.6216

0.1713

5.90

0.6379

0.1693

6.49

0.7074

0.1410

1
1

0
0

0
0

5.99
6.20

0.8354
0.9013

0.0820
0.0499

0.1008

0.6046

0.3954

6.09

0.5449

0.0343

0.2608

0.6092

0.3908

6.35

0.5015

0.0586

5.50

0.3992

6.24

0.9442

0.0289

5.40

0.1995

6.08

0.9688

0.0152

10.28

0.2961

5.96

0.9077

0.0464

10.28

0.1009

6.14

0.9575

0.0212

30.14
30.24

0.2514
0.1499

0
0

0
0

1
1

6.06
6.13

0.2493
0.2309

0.4790
0.5298

30.14

0.0988

6.07

0.1880

0.6224

10.78

0.1494

6.07

0.1378

0.7301

30.22

0.0499

6.06

0.1282

0.7477

Initial Set of Experiments

Designed Experiments

the spatial time () could assume the values of 0.05, 0.1,


0.15, 0.2, and 0.25 gcat 3 h 3 garomatic1, in order to allow for
wide variation of conversions;
feed could be composed of pure toluene or xylenes or mixed
charges of toluene/benzene with molar ratios of 9/1, 8/2, 7/
3, 6/4, 5/5 and toluene/xylenes with molar ratios of 7/3, 5/
5 in order to allow for estimation of parameters associated

with the adsorption constants of chemical species


independently.
Only new experimental conditions could be designed in step
(iv), which means that selection of replicates was forbidden. It is
also important to emphasize that experiments performed with a
pure xylenes feed can allow for improved estimation of the
parameters associated to the products.
176

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Table 6. Sequence of Designed Experiments in the Temperature of 431 C, with y-Trace Criterion
aromatic molar fractions in the feed
P (bar)

(gcat 3 h/garomatic)

30.88

0.0911

20.61

0.1650

15.27

0.3990

31.30

0.0298

30.76

yT

yB

aromatic molar fractions outlet


yX

RHC

yT

yX

5.94

0.7164

0.1361

6.23

0.6978

0.1410

6.10

0.6058

0.1783

5.77

0.8709

0.0612

0.2500

5.96

0.5676

0.1943

20.86

0.0387

5.60

0.8939

0.0507

20.57

0.2860

6.27

0.6078

0.1772

5.59
5.63

0.0590
0.3790

1
1

0
0

0
0

6.35
5.62

0.9691
0.8216

0.0149
0.0886

15.35

0.0490

6.01

0.8943

0.0515

15.21

0.2320

5.64

0.6910

0.1485

10.69

0.0540

6.27

0.9288

0.0365

5.67

0.2200

6.11

0.8930

0.0533

10.36

0.3770

6.51

0.6964

0.1479

10.53

0.1950

5.46

0.7946

0.0968

30.14
30.14

0.1518
0.2328

0.5
0.5

0.5
0.5

0
0

6.12
6.14

0.4187
0.3985

0.0485
0.0602

30.14

0.2007

0.6011

0.3989

6.06

0.4597

0.0770

30.12

0.1000

6.05

0.2476

0.4883

30.12

0.1495

6.03

0.2546

0.4675

10.18

0.0998

6.07

0.1692

0.6653

30.22

0.0499

6.09

0.2051

0.5872

5.20
10.18

0.0996
0.1505

0
0

0
0

1
1

5.96
6.03

0.0991
0.2095

0.8086
0.5843

5.20

0.1497

5.96

0.1265

0.7538

Initial Set of Experiments

Designed Experiments

4. RESULTS
First, replicated experiments were performed in order to
determine the experimental uncertainty of response variables
(conversions of toluene and xylenes). For all experimental runs,
three samples were collected for chromatographic analysis. Besides, for one predened standard condition of = 0.08 gcat 3 h 3
garomatic1, T = 431 C, P = 30 bar and pure feed of toluene,
replicates of the whole experiment were performed. Based on the
previous steps, it was observed that a suitable value for the
standard deviations for toluene and xylenes molar fractions were
approximately equal to T X 1  102. It must be
emphasized that the experimental error is of fundamental
importance for evaluating model adequacy and parameter signicance, as described in Section 2. It is true that experimental
errors may depend on the experimental conditions;37 however,
the determination of experimental uncertainties in all investigated conditions is not possible in most kinetic studies.38 Besides,
as shown below, the excellent agreement between predicted and
observed values strongly suggests that the hypothesis of constant
experimental variability is appropriate.
Tables 36 present the experimental results for each analyzed
temperature level, including the initial set of experiments and the
sequentially designed experiments. Although the main focus of
the present work was the analysis of the toluene disproportionation, the xylenes disproportionation was also investigated

because the y-trace criterion indicated that feeding of pure


xylenes could lead to improved estimation of model parameters
(such as the adsorption constants KX and KTMB). As one can also
see in Tables 36, sequentially designed experiments led to
experimentation at distinct pressure, residence time, and feed
composition levels, allowing for good exploration of the experimental region.
Although experiments were grouped in terms of the reactor
temperatures, obtained model parameters and experiments were
used afterward for more involving estimation of parameters for
the nonisothermal case. Parameter values and their respective
standard deviations are presented in Tables 79. Table 7 shows
the parameter values obtained at each particular temperature
level after the end of the sequential design procedure.
The trajectories of model parameters for the designed experiments are presented in Figure 2. It can be observed that the
determinants of the covariance matrix of parameter uncertainties
tend to decrease as additional experimental results are accumulated, indicating the improved precision of parameter estimates.
The oscillatory behavior of the determinant values can be
associated with signicant changes of the parameter estimates,
as dierent experimental conditions are inserted into the experimental set. Figure 2 also indicate that parameter uncertainties
approach minimum limiting asymptotic values, justifying the
interruption of the sequential experimental design.
177

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Table 7. Estimated Kinetics Parameters for Individual


Temperatures
parameter

mean

standard value

Table 8. Estimated Kinetic Parameters and Respective Standard Deviations and Coecients of Variation

coecient of variation (%)a

parameter

Temperature of 360 C
k1KT2

3.359  104

2.907  104

86.54

KT
k2KX2

9.094  103
2.675  103

1.150  101
2.191  104

1265
8.191

KX

2.379  101

2.909  102

12.23

KB

3.167  101

2.374  101

74.96

KTMB

4.555  101

2.654  101

58.27

Temperature of 380 C
k1KT2

1.154  103

1.996  104

17.30

KT

6.438  102

3.025  102

46.99

2

1.781  103

12.53

KX
KB

4.751  101
2.044  101

5.192  102
6.993  102

10.93
34.21

KTMB

4.466  101

1.329  101

29.77

k2KX

1.422  10

2.80  10

2.68  104

9.60

1.08  101

2.15  102

19.90

2

2.64  10

2.40  103

9.09

KX

3.75  101

3.98  102

10.63

KB
KTMB

1.45  101
9.42  101

4.01  102
1.34  101

27.56
14.24

k1KT
KT
k2KX

k1KT

1.358  10

2

6.140  10

4

A1

5.980

0.071

1.18

B1
AT

30.700
1.960

1.450
0.132

4.72
6.73

BT

12.900

2.760

21.40

A2

4.160

0.057

1.37

B2

27.200

1.180

4.34

AB

1.290

0.151

11.71

BB

7.410

3.350

45.21

AX

1.130

0.075

6.67

BX
ATMB

0.021
0.584

1.470
0.176

7067.31
30.14

BTMB

17.800

3.010

16.91

Table 9. Pre-Exponential Factors, Activation Energies, and


Entropies of Adsorption and Enthalpies of Adsorption Recovered from the Parameter Estimates Presented in Table 5
parameter
PF1 (mol 3 gcat1 3 h1 3 bar2)
E1 (J 3 mol1)
1

1

ST (J 3 K

3 mol )
HT (J 3 mol1)

0.05

KT

2.983  101

1.959  102

0.07

PF2 (mol 3 gcat1 3 h1 3 bar2)

k2KX2

6.248  102

3.774  103

0.06

E2 (J 3 mol )

KX

2.963  101

5.865  102

0.20

1

2

KB

4.484  10

KTMB

1.324

4.113  10

1.745  101

coecient of variation (%)a

Coecient of variation is calculated as the standard deviation divided


by the mean value multiplied by 100.

Temperature of 431 C
2

standard deviation

Temperature of 404 C
3

mean

1

1

1

SB (J 3 K

3 mol )
HB (J 3 mol1)
SX (J 3 K1 3 mol1)

0.09
0.13

HX (J 3 mol1)

Coecient of variation is calculated as the standard deviation divided


by the mean value  100.

STMB (J 3 K1 3 mol1)


HTMB (J 3 mol1)

As one can see in Table 7, all parameters could be obtained


with good precision (except KT at 360 C). Precision of model
parameters was better at higher temperatures, probably as a
consequence of the higher sensibility of response variables with
respect to the experimental conditions presented in Tables 36.
Besides, the number of experimental runs was higher at 431 C.
Table 8 presents the values and precisions of parameters
estimated with all available experimental data. As one can
observe, most parameters could be estimated with very good
precision. Good estimation was obtained for parameters A1 and
B1, related to k1 3 KT2; A2 and B2, related to k2 3 KX2; AT and BT,
related to KT; and ATMB and BTMB, related to KTMB, whose
coecients of variation fell between 1.18% to 21.40%. Good
estimation was obtained for parameter AB, but BB was estimated
with coecient of variation of 45%. Parameter AX was estimated
precisely, but parameter BX could not be estimated with good
precision. Figure 3 presents the ratio between the kinetic
and adsorption constants at temperatures of 431 and 360 C.
One can observe the much larger variations of the kinetic rate
constants, showing the higher sensitivity of kinetic rate constants
to temperature (and the highest precision as a consequence) as
compared to the adsorption constants. On the other hand, the

mean
17.12
27.17  103
90.96
71.53  103
101.33  109
151.04  103
50.88
41.09  103
9.57
115.33
143.13
98.69  103

parameter KB was not very sensitive to the experimental conditions, which explains the large uncertainty of BX. Since the
parameter KX did not vary signicantly in the analyzed experimental range, a single constant value could be adopted for this
parameter in the temperature range investigated here.
Table 9 presents the parameters PFi, Ei (i = 1,2), Sj, and Hj
(j = T,B,X,TMB) that correspond to the values presented in
Table 8. It can be observed in Table 9 that there are some
apparent thermodynamic inconsistencies in the estimated entropy and enthalpy of adsorption for several compounds. According to the literature,39 spontaneous adsorption must be
exothermic (H < 0) as the entropy of adsorption is negative
(S < 0). Apparent thermodynamic inconsistencies have been
reported in other kinetic studies40,41 and used to discard the
investigated models.4246 It must be pointed out, however, that
constraints can be imposed on the estimation problems in order
to avoid such thermodynamic inconsistencies,18 which means
that numerical procedures can be used to force the proposed models
to obey the imposed constraints. Nevertheless, as extensively
discussed by Pinto et al.,38 there may be no fundamental theoretical
basis for implementation of similar procedures, since the kinetic
178

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Figure 2. Determinant of covariance matrix of parameter uncertainties


as function of the number of designed experiments (for T = 431 C, the
rst design experiment was 18).

Figure 4. Experimental output molar fractions for (A) toluene feed and
(B) xylenes feed.

eventually obtained for all estimated model parameters when


simple reaction mechanisms and usual adsorption models are
proposed.
The results presented so far illustrate typical problems of
kinetic studies related to the precise estimation of the model
parameters. Moreover, obtained results suggest that the evaluation of phenomenological aspects based solely on the parameter
values of models tted to available data should be performed very
carefully. Thermodynamic inconsistency strongly suggests that
the proposed model is likely to be incorrect, from the phenomenological point of view; nonetheless, even if all parameters are
thermodynamic consistent, this does not guarantee that the
proposed model is phenomenologically correct. Actually, if one
is not interested in the fundamental phenomenological aspects of
the proposed model, it seems reasonable to focus primarily on
the predictive capacity of the proposed models.
It can be observed in Table 10 that activation energies reported
in the literature lie in the range between 43 and 118 kJ/mol, while
the value obtained in the present work is close to 27 kJ/mol.
However, for complex models, such as the LHHW model used
here, parameter estimates are correlated to each other, making
comparison with previously published material questionable, as
extensively discussed by Pinto et al.38 Even when very simple
models are compared to each other, dierent parameter values can
be found, depending on the analyzed experimental range and
operation conditions. For example, Gnep and Guisnet11 found
that the kinetics of disproportionation depended strongly on the
catalyst type and on its pretreatment. Particularly, it was found that
the reaction order decreased with the wet-air or dry-air pretreatment of the mordenite catalyst prepared from uorinated alumina,
making the interpretation of kinetic orders doubtful.

Figure 3. Ratio between model parameters at the maximum temperature of 431 C and the minimum temperature of 360 C.

model is a necessary simplication of reality. Thus, the proposition of kinetic models is almost always the result of arbitrary
assumptions, so that thermodynamic consistency can eventually
constitute a numerical trick of the kinetic formalism, especially
when it can be corrected with the help of numerical techniques at
the expense of the quality of the model t. In this sense, one could
always argue that the quality of model tting is more important
than the thermodynamic consistency of estimated parameters if
the extensive investigation of the experimental region has been
carried out. This point of view is adopted in this work. Therefore,
the proposed modeling procedure privileges the model t and
the regression analysis of the model parameters. In fact, excellent
agreement is reached between model predictions and the experimental data, as shown in Figure 4 in wide ranges of toluene and
xylenes conversions. One can also observe that systematic
deviations and outliers cannot be identied in Figure 4, as
assumed in Section 2.
It must be clear, though, that thermodynamic consistency
should be expected in exact phenomenological kinetic models,
which are rarely available, especially when the reaction mechanisms are complex. If thermodynamic consistency is not obtained
for a given kinetics, either the model hypotheses are not adequate
(reaction network, choice of the rate determining step, etc.) or the
adopted adsorption theory is not the most appropriate. In such
case, other models might be tested by the analyst, if the analyst is
not satised with the obtained model performance, although it
cannot be guaranteed that thermodynamic consistency will be
179

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Table 10. Activation Energies for Toluene


Disproportionation
ref

activation energy (kJ/mol)

this work

27.17

(3)
(4)

88.00
118.71

(5)

64.79

(7)

60.61

(8)

104.50

(9)

102.00

(10)

54.22

(12)

84.85

(13)
(14)

117.04
43.47

(19)

99.00

(20)

87.78

Figure 7. Experimental (symbols) and simulated (line) toluene conversions with feed of toluene+H2 at 10 bar as a function of spatial time
for several temperatures.

Figure 8. Experimental (symbols) and simulated (line) toluene conversions with feed of toluene+H2 at 5 bar as a function of spatial time
for several temperatures.
Figure 5. Experimental (symbols) and simulated (line) toluene conversions with feed of toluene+H2 at (A) 30 bar as a function of spatial
time for several temperatures.

Figure 9. Experimental (symbols) and simulated (line) conversions at


431 C with feed of toluene+H2 at several pressures.
Figure 6. Experimental (symbols) and simulated (line) toluene conversions with feed of toluene+benzene+H2 at 30 bar as a function of
spatial time for several temperatures. (For T = 431, yT = 0.5, and yBZ =
0.5 and for other temperatures yT = 0.6 and yBZ = 0.4).

accurate matrix inversions demanded by the parameter estimation procedure is feasible only when the reparametrization is
performed. Particularly, pre-exponential factors and activations
energies of the Arrhenius equation are correlated strongly,
demanding reparametrization for achievement of good parameter estimation.3133 In the present problem, the estimation
of model parameters without reprarametrization was not possible, so that it was not possible to determine the standard
deviation of the original model parameters, such as the activation
energy and pre-exponential factors presented in Table 9. The
alternative technique of error propagation could be used but

It is important to emphasize that reparametrization makes the


estimation of model parameters easier, since the correlations
between parameters can be signicantly reduced (and sometimes
removed). However, reparameterization does not lead to improvement of model predictions, as the model remains essentially the same. Despite that, reparameterization is appealing for
numerical and statistical reasons.3133 In fact, in some cases
180

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

Figure 12. Experimental (symbols) and simulated (line) conversions as


a function of spatial time for several temperatures at 5 bar with feed of
xylenes+H2.

Figure 10. Experimental (symbols) and simulated (line) conversions as


a function of spatial time for several temperatures at 30 bar with feed of
xylenes+H2.

of toluene conversions with respect to pressure decreases. This


can be explained in terms of eqs 9 and 10, as reaction rates
increase with the square of the system pressure when the partial
pressures are close to zero and are insensitive to pressure changes
when the partial pressures grow to innity. Similar results are
observed in Figures 58.
Similarly, Figures 1012 present xylenes conversions for
experiments carried out with pure xylenes in the feed. As one
can observe in Figures 1012, the eects of operation conditions
on xylenes disproportionation are very similar to the eects
observed for toluene disproportionation over H-mordenite. The
comparison between Figures 5 and 10, Figures 7 and 11, and
Figures 8 and 12 indicates that the rates of xylenes disproportionation are always higher than the rates of toluene disproportionation for similar experimental conditions. This can only be
explained in terms of the parameter values in the investigated
experimental ranges, given the relative complexity of the kinetic
rate expressions and the thermodynamic reversibility of the
reactions. A possible mechanistic explanation can be related to
a simple statistical eect, as xylene molecules present two methyl
groups available for disproportionation, while toluene molecules
present only one methyl group. A second eect can be related to
the well-known electron-donor eect of methyl groups, which
can lead to more eective adsorption of xylene molecules onto
the catalyst.
Based on the available experimental results, it is possible to
perform sensitivity analyses regarding the inuence of the
analyzed experimental conditions on the conversion of toluene
and/or xylenes in specied experimental regions. For instance,
when Figures 5 and 9 are compared, an increase of 10 bar has
approximately the same impact on toluene conversion as an
increase of 50 C, at the contact time of 0.3 (h 3 gcat/gAromatic).
However, it is important to emphasize that the sensitivity to
experimental conditions change in the analyzed experimental
range. For this reason, the kinetic rate expressions obtained in the
present work can be used in the near future for more involving
optimization and sensitivity analysis studies.
As shown in Figures 412, excellent model ts to experimental data can be obtained in all cases. Therefore, the model
presented here can be used successfully for representation of
obtained experimental data in the investigated operation range.
Therefore, the model can be used with condence for simulation,
process design, process control, and interpretation of industrial
and laboratorial reactors.

Figure 11. Experimental (symbols) and simulated (line) conversions as


a function of spatial time for several temperatures at 10 bar with feed of
xylenes+H2.

would lead to unrealistic estimation of parameter uncertainties,


since the original nonlinear transformations in eq 11 must be
linearized when error propagation is performed. For this reason,
it is dicult to compare the data presented in Table 10 based on
rigorous statistical grounds.
Figures 58 show toluene conversions as functions of the
spatial time at dierent temperatures and pressures. As shown in
Figures 58, the increase of temperature signicantly increases
toluene conversions in all the investigated experimental conditions. A comparison between Figures 5 and 6 for similar pressures
and temperatures reveals that the presence of benzene reduces
the conversion of toluene very signicantly. According to eq 9,
the increase of the partial benzene pressure leads to reduction of
the reaction rates because of both the adsorption term in the
denominator and the thermodynamic equilibrium term in the
numerator. According to Figure 6, the inuence of temperature is
much less pronounced when benzene is present in the feed.
Particularly, when temperatures are in the range between 404
and 431 C, toluene conversions are essentially constant because
of the higher amounts of benzene in the feed in the experiment
performed at 431 C.
The inuence of pressure on toluene conversions can be
visualized in Figure 9, where the toluene conversions at 431 C
are presented as functions of the spatial time at dierent
pressures. Figure 9 clearly shows that the increase of the pressure
leads to signicant increase of toluene conversions. As observed
in Figure 9, with the increase of reaction pressure, the sensitivity
181

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

4. CONCLUSIONS
The kinetics of toluene and xylenes disproportionation over a
commercial H-mordenite catalyst has been investigated with the
help of sequential experimental design procedures in a wide
range of operation conditions. The inuence of reaction temperature, reactor pressure, feed composition, and residence time
on conversion of reactants and product selectivity was investigated. Experiments were performed according to a sequential
experimental design strategy, in order to provide maximum
accuracy for model predictions. Statistical treatment of parameter estimates and model adequacy was performed with the
help of maximum likelihood principles. Excellent agreement
between model predictions and available experimental data was
obtained in the full ranges of investigated experimental conditions, although thermodynamic consistency of parameter estimates was not observed in some cases.

(12) Beltrame, P.; Beltrame, P. L.; Carniti, P.; Forni, L.; Zuretti, G.
Toluene disproportionation catalyzed by various zeolites. Zeolites 1985,
5, 400.
(13) Krahl, C. A. Cinetica de desproporcionamento de toluene sobre
mordenita. M.Sc. Dissertation, Universidade Federal do Rio de Janeiro,
Rio de Janeiro, Brazil, 1987 (in Portuguese).
(14) Shanker, U.; Rawat, D. S.; Bawa, J. S.; Dabral, R. P.; Bhattacharya,
K. K. Kinetics of vapor-phase disproportionation of toluene over H-mordenite. Erdol Kohle Erdgas P. 1990, 43, 489.
(15) Corma, A.; Llopis, F. E.; Monton, J. B. Inuence of the structural
parameters of Y zeolite on the transalkylation of alkylaromatics. J. Catal.
1993, 140, 384.
(16) Barbosa, L. A. M. M. Efeito da press~ao parcial de hidrog^enio na
cinetica de desproporcionamento do tolueno, M.Sc. Thesis, Universidade
Federal do Rio de Janeiro, Rio de Janeiro, Brazil, 1993 (in Portuguese).
(17) Marques, N. A. Estudo cinetico da reac-~ao de desproporcionamento de tolueno, M.Sc. Dissertation, Universidade Federal do Rio de
Janeiro, Rio de Janeiro, Brazil, 2000 (in Portuguese).
(18) Henriques, C. A. Isomerizac-~ao de xilenos sobre mordenitas:
estudo cinetico da reac-~ao e caracterizac-~ao do coque formado, D.Sc. Thesis,
Universidade Federal do Rio de Janeiro, Brazil, 1994 (in Portuguese).
(19) Uguina, M. A.; Sotelo, J. L.; Serrano, D. P. Kinetics of toluene
disproportionation over fresh and coked H-mordenite. Catal. Today
2004, 97, 297.
(20) Dooley, K. M.; Brignac, S. D.; Price, G. L. Kinetics of zeolitecatalyzed toluene disproportionation. Ind. Eng. Chem. Res. 1990, 29, 789.
(21) Bard, Y. Nonlinear parameter estimation; Academic Press: New York,
1974.
(22) Schwaab, M.; Pinto, J. C. Analise de dados experimentais I:
fundamentos de estatstica e estimac-~ao de par^ametros; E-Papers, Rio de
Janeiro, 2007 (in Portuguese).
(23) Atkinson, A. C.; Donev, A. N.; Tobias, R. D. Optimum experimental designs, with SAS; Clarendon Press: Oxford, 2007.
(24) Schwaab, M.; Queipo, C. Q.; Silva, F. M.; Barreto, A. G., Jr.;
Nele, M.; Pinto, J. C. A new approach for sequential experimental design
for model discrimination. Chem. Eng. Sci. 2006, 61, 5791.
(25) Schwaab, M.; Monteiro, J. L.; Pinto, J. C. Sequential experimental design for model discrimination: Taking into account the
posterior covariance matrix of dierences between model predictions.
Chem. Eng. Sci. 2008, 63, 2408.
(26) Alberton, A. L.; Schwaab, M.; Lob~ao, M. W. N.; Pinto, J. C.
experimental design for the joint model discrimination and precise
parameter estimation through information measures. Chem. Eng. Sci.
2011, 66, 1940.
(27) Pinto, J. C.; Lob~ao, M. W.; Monteiro, J. L. Sequential experimental
design for parameter estimation: a dierent approach. Chem. Eng. Sci. 1990,
45, 883.
(28) Cieplinski, E. W. Gas chromatographic analysis of aromatic hydrocarbons with modied bentonite columns. Eect of bentone 34 concentration on performance of packed columns. Anal. Chem. 1965, 37, 1160.
(29) Sattereld, C. N., Sherwood, T. K. The role of diusion in catalysis;
Addison-Wesley Publishing Company: 1963.
(30) Alberty, R. Equilibrium disproportionation and isomerization
of alkylbenzenes. Ind. Eng. Chem. Fundam. 1986, 25, 211.
(31) Schwaab, M.; Pinto, J. C. Optimum reference temperature for
reparameterization of the Arrhenius equation. Part 1: Problems involving one kinetic constant. Chem. Eng. Sci. 2007, 62, 2750.
(32) Schwaab, M.; Lemos, L. P.; Pinto, J. C. Optimum reference
temperature for reparameterization of the Arrhenius equation. Part 2:
Problems involving multiple reparameterizations. Chem. Eng. Sci. 2008,
63, 2895.
(33) Schwaab, M.; Pinto, J. C. Optimum reparameterization of
power function models. Chem. Eng. Sci. 2008, 63, 4631.
(34) Levenspiel, O. Chemical Reaction Engineering, 2nd ed.; McGraw
Hill: New York, 1979.
(35) Sperandio, D.; Mendes, J. T.; Silva, L. H. M. Calculo numerico Caracteristicas matematicas e computacionais dos metodos numericos;
Prentice Hall: S~ao Paulo, 2003 (in Portuguese).

AUTHOR INFORMATION
Corresponding Author

*Phone: 55-21-25628337. Fax: 55-21-25628300. E-mail: pinto@


peq.coppe.ufrj.br.

ACKNOWLEDGMENT
The authors thank CNPq  Conselho Nacional de Desenvolvimento Cientco e Tecnologico, CAPES  Coordenac-~ao de
Aperfeic-oamento de Pessoal de Nvel Superior, and FAPERJ 
Fundac-~ao Carlos Chagas Filho de Apoio a Pesquisa do Estado do
Rio de Janeiro, for providing scholarships and supporting this work.
REFERENCES
(1) Speight, J. G. Chemical and process design handbook; McGraw-Hill:
New York, 2002.
(2) Cannella, W. J. Xylenes and ethylbenzene. Kirk-Othmer Encyclopedia of Chemical Technology, 2000.
(3) Aneke, L. E.; Gerritsen, L. A.; Eilers, J.; Trion, R. The disproprotionation of toluene over a HY/-AlF3/Cu catalyst: 2. Kinetics. J. Catal.
1979, 59, 37.
(4) Bhaskar, G. V.; Do, D. D. Toluene disproportionation reaction
over HZSM-5 zeolites: kinetics and mechanism. Ind. Eng. Chem. Res.
1990, 29, 355.
(5) Das, J.; Bhat, Y. S.; Halgeri, A. B. Selective toluene disproportionation over pore size controlled MFI zeolite. Ind. Eng. Chem. Res. 1994,
33, 246.
(6) Bharati, S. P.; Bhatia, S. Deactivation kinetics of toluene disproportionation over hydrogen mordenite catalyst. Ind. Eng. Chem. Res. 1987,
26, 1854.
(7) Bhavikatti, S. S.; Patwardhan, S. R. Toluene disproportionation
over nickel-loaded aluminum-decient mordenite. 2. Kinetics. Ind. Eng.
Chem. Prod. Res. Dev. 1981, 20, 106.
(8) Tsai, T. C.; Chen, W. H.; Lai, C. S.; Liu, S. B.; Wang, I.; Ku, C. S.
Kinetics of toluene disproportionation over fresh and coked H-mordenite. Catal. Today 2004, 97, 297.
(9) Xu, O.; Su, H.; Ji, J.; Jin, X.; Chu, J. Kinetic model and simulation
analysis for toluene disproportionation and C9-aromatics transalkylation. Chin. J. Chem. Eng. 2007, 15, 326.
(10) Waziri, S. M.; Aitani, A. M.; Al-Khattaf, S. Transformation of
toluene and 1,2,4-trimethylbenzene over zsm-5 and mordenite catalysts:
a comprehensive kinetic model with reversibility. Ind. Eng. Chem. Res.
2010, 49, 6376.
(11) Gnep, N. S.; Guisnet, M. Toluene disproportionation over
mordenites - II: Kinetic study. Appl. Catal. 1981, 1, 329.
182

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Industrial & Engineering Chemistry Research

ARTICLE

(36) Noronha, F. B.; Pinto, J. C.; Monteiro, J. L.; Lob~ao, M. W.; Santos,
T. J. ESTIMA: Um pacote computacional para estimac-~ao de par^ametros de
projeto de experimentos; Relatorio Tecnico, PEQ/COPPE, Universidade
Federal do Rio de Janeiro: Rio de Janeiro, 1993 (in Portuguese).
(37) Alberton, A. L.; Schwaab, M.; Schmal, M.; Pinto, J. C. Experimental errors in kinetic tests and its inuence on the precision of
estimated parameters. Part IAnalysis of rst-order reactions. Chem.
Eng. J. 2009, 155, 816.
(38) Pinto, J. C.; Lob~ao, M. W.; Alberton, A. L.; Schwaab, M.;
Embiruc-u, M.; Vieira de Melo, S. Critical analysis of kinetic modeling
procedures. Int. J. Chem. React. Eng 2011, 9, A87.
(39) Vannice, M. A. Kinetic of Catalytic Reactions; Springer Science:
New York, 2005.
(40) Peterson, T. I.; Lapidus, L. Nonlinear estimation analysis of the
kinetics of catalytic ethanol dehydrogenation. Chem. Eng. Sci. 1966, 21, 655.
(41) Xu, J.; Froment, G. F. Methane steam reforming, methanation
and water-gas shift: I. Intrinsic kinetics. AIChE J. 1989, 35, 88.
(42) Cavalieri, M. Estudo cinetico da reac-~ao de desidrogenac-~ao de
etilbenzeno, M.Sc. Thesis, Universidade Federal do Rio de Janeiro,
Brazil, 1992 (in Portuguese).
(43) Van Trimpont, P. A.; Marin, G. B.; Froment, G. F. Kinetics of
methylcyclohexane dehydrogenation on sulded commercial platinum/
alumina and plantinum-rhenium/alumina catalysts. Ind. Eng. Chem.
Fundam. 1986, 25, 544.
(44) Gut, G.; Jaeger, R. Kinetics of the catalytic dehydrogenation of
cyclohexanol to cyclohexanone on a zinc oxide catalyst in a gradientless
reactor. Chem. Eng. Sci. 1982, 37, 319.
(45) Lee, W. J.; Froment, G. F. Ethylbenzene Dehydrogenation into
styrene: kinetic modeling and reactor simulation. Ind. Eng. Chem. Res.
2008, 47, 9183.
(46) Specchia, S.; Conti, F.; Specchia, V. Kinetic studies on Pd/
CexZr1-xO2 catalyst for methane combustion. Ind. Eng. Chem. Res. 2010,
49, 11101.

183

dx.doi.org/10.1021/ie2015526 |Ind. Eng. Chem. Res. 2012, 51, 171183

Você também pode gostar