1824 Ind. Eng. Chem. Res.

1991,30, 1824-1828
y = response Draper, N. R.; Smith, H. Applied Regression Analysis, 2nd ed.;
Y = yield of MA Wiley: New York, 1981; pp 1-125.
Y = vector of responses Hodnett, B. K. Vanadium-Phosphorus Oxide Catalysts for the Se-
= least-squares estimate lective Oxidation of C4 Hydrocarbons to Maleic Anhydride. Catal.
Rev. Sci. Eng. 1985, 27 (3), 373-424.
Registry No. PO, 1314-56-3; MA, 108-31-6;Mo, 7439-98-7; Hodnett, B. K.; Delmon, B. Factors Influencing the Selectivity of
Ce, 7440-45-1; H(CH2),H, 106-97-8;V, 7440-62-2. Vanadium-Phosphorus Oxide Catalysts for n-Butane Oxidation
to Maleic Anhydride. Appl. Catal. 1985, 15, 141-150.
Hucknall, D. J. Selective Oxidation of Hydrocarbons; Academic
Press: London, 1974; Chapter 4.
Literature Cited Katsumoto, K.; Marquis, D. M. US. Patent 4 132 670, 1979.
Khuri, A. I.; Cornell, J. A. Response Surfaces Designs and Analyses;
Box, G. E. P.; Hunter, W. C.;Hunter, J. S. Statistics for Experi- Marcel Dekker: New York, 1987; Chapters 1, 2, 3, and 5.
menters; Wiley: New York, 1978; Chapters 10 and 12.
Buchanan, J. S.; Sundaresan, S. Kinetics and Redox Properties of Kittrell, J. R.; Erjavec, J. Response Surface Methods in Heteroge-
neous Kinetic Modelling. Ind. Eng. Chem. Process Des. Deu.
Vanadium Phosphate Catalysts for Butane Oxidation. Appl. 1968, 7 (3),321-327.
Catal. 1986, 26, 211-226.
Cavani, F.; Centi, G.; Trifiro, F. The Chemistry of Catalysts Based Moser, T. P.; Schrader, G. L. Selective Oxidation of n-Butane to
on Vanadium-Phosphorus Oxides Note IV Catalytic Behaviour Maleic Anhydride by Model V-P-0 Catalysts. J . Catal. 1985,92,
of Catalysts Prepared in Organic Medium in the Oxidation of C4 216-231.
Fraction. Appl. Catal. 1984,9, 191-202. Rao, M. S.; Iyengar, S. S. In Computer Modelling of Complex Bio-
Cavani, F.; Centi, G.; Manenti, I.; Trifiro, F. Catalytic Conversion logical Systems; Iyengar, S. s.,Ed.; CRC Press: Boca Raton, FL,
of C, Hydrocarbons on Vanadium-Phosphorus Oxide: Factors 1984; pp 29-53.
Influencing the Selectivity of l-Butene Oxidation. Ind. Eng. Smith, J. M. Chemical Engineering Kinetics, 3rd ed.; McGraw-Hill:
Chem. Prod. Res. Deu. 1985a, 24, 221-226. New York, 1981; Chapter 11.
Cavani, F.; Centi, G.; Trifiro, F. Study of n-Butane Oxidation to Varma, R. L.; Saraf, D. N. Oxidation of Butene to Maleic Anydride
Maleic Anhydride in a Tubular Flow Stacked-Pellet Reactor. I. Kinetics and Mechanism. J . Catal. 1978,55, 361-372.
Influence of Phosphorus on the Selectivity. Appl. Catal. 1985b, Varma, R. L.; Saraf, D. N. Selective Oxidation of C, Hydrocarbons
15, 151-160. to Maleic Anhydride. Ind. Eng. Chem. Prod. Res. Deu. 1979, 18
Centi, G.; Fornasari, G.; Trifiro, F. On the Mechanism of n-Butane (11, 7-13.
Oxidation to Maleic Anhydride: Oxidation in Oxygen-Stiochiom- Wellauer, T. P.; Cresswell, D. L.; Newson, E. J. Optimal Policies in
etry-Controlled Condtions. J . Catal. 1984a, 89, 44-51. Maleic Anhydride Production through Detailed Reactor Modell-
Centi, G.; Manenti, I.; Riva, A.; Trifiro, F. The Chemistry of Cata- ing. Chem. Eng. Sci. 1986, 41 ( 4 ) , 765-772.
lysts Based on Vanadium-Phosphorus Oxides Note I11 Catalytic Wenig, R. W.; Scharder, G. L. Vanadium-Phosphorus-Oxygen In-
Behaviour of Different Phases in 1-Butene Oxidation to Maleic dustrial Catalysts for n-Butane Oxidation: Characterization and
Anhydride. Appl. Catal. 1984b,9, 177-190. Kinetic Measurements. Ind. Eng. Chem. Fundam. 1986, 25,
Centi, G.; Trifiro, F.; Ebner, J. R.; Franchetti, V. M. Mechanistic 612-620.
Aspects of Maleic Anhydride Synthesis from C4 Hydrocarbons
over Phosphorus Vanadium Oxide. Chem. Reu. 1988,88,55-80. Receiued for review July 2, 1990
Davies, 0. L. The Design and Analysis of Industrial Experiments; Reuised manuscript received February 6, 1991
Longman Group: New York, 1978; Chapter 11. Accepted February 12,1991

Selective Oxidation of 11 -Butane to Maleic Anhydride. 2. Identification

of Rate Expression for the Reaction
Shyamal K. Bej and Musti S. Rao*
Department of Chemical Engineering, Indian Institute of Technology, K a n p u r 208 016, U.P.,
India

The modeling of the selective oxidation of n-butane t o MA has been done in a systematic manner.
Both Langmuir-Hinshelwood and redox models have been proposed and tested with differential
rate data. Four redox models have been fitted to the experimental data. Discrimination among
these models has been achieved with the use of a sequential experimental design. T h e redox model
having first-order dependency on n-butane partial pressure and zero-order dependency on oxygen
partial pressure was found to be the most appropriate.

1. Introduction Buchanan and Sundaresan (1986) have proposed different

forms of redox models for this reaction. An adsorption
The selective oxidation of n-butane to maleic anhydride model has been proposed by Lerou and Weiher (1986).
(MA) is a complex reaction from the mechanistic and Though a number of studies are available on the modeling
modeling point of view. Besides the main reaction of of this reaction, there is some doubt regarding the order
selective formation of MA from n-butane, two other side of reaction with respect to oxygen and also the activation
reactions, viz., the complete oxidation of n-butane and of energy. In this present work we have tried to clarify some
MA to oxides of carbon, take place simultaneously. of these doubts by systematic modeling studies.
Different types of models have been proposed by dif- The reaction has been studied over a Mo- and Ce-pro-
ferent investigators (e.g., Wohlfahrt and Hofmann (1980), moted VPO catalyst (P/V atomic ratio = 1.0s; weight ratio
Centi et al. (19851, Buchanan and Sundaresan (1986), of Mo/V = 0.06; weight ratio of Ce/V = 0.0221, the op-
Lerou and Weiher (1986), and Schneider et al. (1987)). timum composition of which was established in part 1 of
Wohlfahrt and Hofmann (1980), Centi et al. (1985), and this series. The catalyst had a surface area of 20 m2/g.
The details of catalyst preparation were also given in part
* T o whom correspondence should be addressed. 1 of this series.
0888-5885/91/2630-1824$02.50/0 0 1991 American Chemical Society
 Ind. Eng. Chem. Res., Vol. 30, No. 8, 1991 1825
Table I. Concentrations of the Components Involved in the Rate Equations and Rate of -Butane Depletion
I)

rate of n-butane
inlet concn, vol % outlet concn, vol % concn*'01 % depletion, g-mol/(g.h)
temp, O C n-butane oxygen n-butane ~
oxygen ~~
n-butane oxygen x 103
370 0.60 21.95 0.54 21.71 0.57 21.83 1.2114
0.94 19.77 0.86 19.47 0.90 19.62 1.8055
1.20 17.68 1.14 17.42 1.17 17.55 2.1209
1.50 15.58 1.42 15.24 1.46 15.41 2.4604
1.80 13.57 1.70 13.15 1.75 13.36 2.9357
390 0.59 21.78 0.55 21.63 0.57 21.71 2.3296
0.90 19.68 0.84 19.42 0.87 19.55' 2.7238
1.21 17.80 1.13 17.44 1.17 17.62 4.0466
1.54 15.80 1.40 15.28 1.47 15.54 4.7088
1.80 13.47 1.72 13.18 1.76 13.40 5.0342
410 0.60 21.68 0.56 21.50 0.58 21.59 3.7777
0.90 19.55 0.84 19.29 0.87 19.42 4.6764
1.22 17.50 1.14 17.16 1.18 17.33 6.2457
1.50 15.44 1.44 15.14 1.47 15.29 7.0716
1.80 13.26 1.70 12.84 1.75 13.05 7.5654

2. Experimental Section Table 11. Values of Parameters for Models 1-4

The setup used for the present study was the same as model K," K," RSS
that described in part 1 of this series except that oxygen Temperature = 370 "C
and nitrogen gases were used separately instead of air. The 1 0.3060 0.2169 0.2254 X
2 0.0985 0.2214 0.1980 X
experiments were performed and the products were ana- 3 0.0553 0.2348 0.1689 X lo-'
lyzed according to procedures described in part l of this 4 0.0308 0.2445 0.1407 X
series.
Temperature = 390 "C
1 0.4878 0.3990 0.2007 X 10"
3. Results and Discussion 2 0.1619 0.4174 0.2077 X 10"
3 0.0937 0.4298 0.2141 X 10"
The general scheme of the reaction can be represented 4 0.0523 0.4467 0.2212 x 10"
by a combination of series and parallel reactions as follows:

n-butane - (1)
MA (1)
1
2
3
Temperature = 410 O C
0.5447
0.1791
0.1022
0.6848
0.7356
0.7222
0.1432 X 10"
0.1315 X IO"
0.1347 X 10"
4 0.0575 0.8219 0.1473 X 10"
oxides of carbon
a Units: g-mol/(g.h.atm).
In the present investigation we have conducted detailed
modeling studies of the fmt reaction (i). For that we have
conducted experiments under such conditions that MA is order of the reaction with respect to n-butane.
the only product of the reaction. From preliminary ex- Similarly, the rate a t which oxygen is consumed at the
periments it is observed that the reaction carried out below site R for the oxidation of R to X is given by
410 O C and at low conversion of n-butane leads to very high -dp02/dt = KIP8PR (5)
selectivity (>96%) toward MA. Under such conditions,
the above complex reaction is reduced to a simplified form where po, is the partial pressure of oxygen, dR is the
as
n-butane MA-
Both Langmuir-Hinshelwood models and redox models
(2)
fraction of surface occupied by R, and n is the order of
reaction with respect to oxygen.
If a moles of oxygen are required for the oxidation of
1 mol of n-butane to MA, then a t steady state oxygen
can be proposed for the reaction. We have derived four balance on site X gives
redox models and three Langmuir-Hinshelwood models aK&eX = KIP6$R (6)
for this reaction.
The derivation of the redox models is based on the or
general principles as discussed by Thomas and Thomas
(1967). The reaction. can be written in the following aK$Bex = K&(l - 8x1 (7)
stepwise manner: Rearrangement of the above equation gives the following
KI final form:
O,+R-X
KIKzPBP6,
= (8)
B + xL !- MA + R
-rg
(3) KlP8, + 'YK2PB
where B, MA, R, and X stand for n-butane, maleic anhy- There is very little doubt regarding the value of m. Most
dride, reduced state of the catalyst, and oxidized state of of the previous investigators, viz., Centi et al. (1985),Bu-
the catalyst, respectively. chanan and Sundaresan (1986),and Centi et al. (19881,
The rate a t which n-butane is depleted to form MA on have established the value of m as 1.0. However, there is
site X is given by much doubt regarding the value of n. In the present in-
-dpe/dt = K & Y x (4) vestigation we have assumed the value of m as 1.0 but four
different values of n, viz., 1.0, 0.5,0.25, and 0 for models
where p B is the partial pressure of n-butane a t a time t, 1, 2, 3, and 4,respectively.
8x is the fraction of surface occupied by X, and m is the The Langmuir-Hinuhelwood models have been derived
1826 Ind. Eng. Chem. Res., Vol. 30, No. 8, 1991

Table 111. Posterior Probabilities of Models (Initial Prior Probability for Each Model = 0.25)
posterior probability of
temD, O C discrim stage total no. of exDts model 1 model 2 model 3 model 4
370 1 5 0.210 0.240 0.260 0.290
2 6 0.010 0.110 0.310 0.570
3 I 0.108 0.305 0.587
4 8 0.100 0.294 0.605
5 9 0.095 0.290 0.614
6 10 0.090 0.280 0.630
390 1 5 0.256 0 252 0.248 0.244
2 6 0.002 0.130 0.350 0.520
3 i 0.120 0.340 0.540
4 8 0.117 0.325 0.560
5 9 0.110 0.320 0.570
6 10 U.106 0.315 0.580
410 1 9 0.246 0.257 0.254 0.243
2 6 0.001 0 080 13.320 0.590
c
3 0.070 0.320 0.610
4 8 0 060 0.310 0.630
5 9 0.056 0.306 0.637
6 10 0 0% 0.306 0.655

on the lines the modeling of the propylene oxidation re- The values of the parameters for models 1-4, as esti-
action in Tan et al. (1988). mated by least-squares technique at various temperatures,
Model 5. For this model, a single site mechanism is are given in Table 11. From the values of parameters (not
considered. It is assumed that all sites have affinities for shown here) it appeared that the Langmuir-Hinshelwood
n-butane and oxygen. Then we get the following form of models did not fit the rate data well since the estimated
the Langmuir-Hinshelwood model. parameters were found to be negative. On the other hand,
all the redox models seem to fit the experimental results
well. This is quite reasonable since only lattice oxygen
rather than adsorbed oxygen of the catalyst is involved in
the reaction. This has also been reported by Srivastava
In this model, it is also assumed that each adsorbed (1988) for other selective oxidation reactions.
molecule occupies only one site. The values of residual sum of squares (RSS) for all the
Model 6. In this model two different sites-one with redox models are also given in Table 11. Since the RSS
affinity for n-butane and the other with affinity for are very close to each other, it is not possible to discrim-
oxygen-are considered. The occupancy of one site by a inate among the redox models based on minimum RSS
single molecule is also assumed here. With these as- criterion.
sumptions, the Langmuir-Hinshelwood model becomes In order to discriminate among the four redox models,
K2KAKRPBP02 we have applied the Bayesian method of discrimination
-rB = (10) among rival models. The details of this method are given
(1 -k KAPo1)(1 + KRPB) by several authors (e.g., Box and Hill (1967), Roth (19651,
Model 7. Here, besides the assumption of two different and Prasad and Rao (1977)).
sites as in model 6, the occupancy of a pair of adjacent sites The Bayes' theorem states that
by each oxygen molecule has been considered. Then the
Langmuir-Hinshelwood equation becornes (13)

:=1

where Ai (i = 1,..., r ) denotes the ith model, B denotes the

data, P(Ai)denotes the prior probability of the ith model,
The notations used for developing the Langmuir-Hin- and P(B/A,) denotes the likelihood for the ith model. A
shelwood models are as follows: KA is the oxygen ad- knowledge of prior probabilities for various models is re-
sorption equilibrium constant, KR is the n-butane ad- quired. If there is no information regarding the prior
sorption equilibrium constant, and K2 is the rate constant probabilities, equal probabilities for all the models are
for n-butane oxidation. assigned. A knowledge of the error structure is required
Experiments were carried out at temperatures of 370, for calculating the likelihood. Once the observations and
390, and 410 "C under differential conditions. The ex- design matrix are substituted into the expression for the
perimental conditions and responses of the experiments probability density function, (p.d.f.1, the resulting ex-
are given in Table I. The rates are calculated by use of pression, which is a function of k and u2, is called the
the relation likelihood.
-rB = A x / ( A ( W / F ) ) (12) With the use of the estimated parameters of Table 11,
the likelihoods were calculated. Equal prior probabilities
Where AX and A( WIF) represent conversion and contact were assigned and the posterior probabilities were calcu-
time, respectively. The value of concentration used in the lated at three different temperatures (370, 390, and 410
rate equation is the average of the inlet and outlet con- " C ) on the basis of the results of five experiments a t each
centrations of the respective component. The difference temperature. From the values of the posterior probabilities
between the inlet and outlet concentrations ranged up to after five preliminary runs, it was difficult to discriminate
1070,substantiating the differential operation of the re- among the models on the basis of the existing data.
actor. With these values, all the above seven models have When it is not possible to discriminate among rival
been tested. models from the available data, it is necessary to design
 Ind. Eng. Chem. Res., Vol. 30, No. 8, 1991 1827
Table IV. Values of Parameters of Model 4 Based on All - Partial pressure of n-butane,atm
Runs 5 O-Ol75 0.02 0.0225 0025 0.0275 003
temperature, OC 370 390 410
2 0.0.4
-@
K1,g-mol/(gatm.h) 0.03362 0.0586 0.0643 E
K2,g-mol/(gatmh) 0.2383 0.4339 0.7841 8 0.06.
c

Table V. Comparison of Activation Energies with Those

Reported in the Literatum
value of
activation energy,
source kcal/mol 01.2-
our investigation El 12.60 Figure 1. Experimental region and settings for experiments for
E2 25.90 sequential model discrimination at 370 OC.
Buchanan and Sundaresan (1986) E2 = 27.71
Centi et al. (1985) E2 = 10.77
Partial pressure of n-butane , a t m
Schneider et al. (1987) E2 17.24 e
0.02
~-o,o~0175 0.0225 0.025 0.0275 0.03
further experiments that will provide maximum discrim- c
b)
ination. While several criteria are available, we have E2 0.06
chosen Box and Hill's criterion (Box and Hill, 1967) for c
the design of experiments for model discrimination. The 0

design criterion to be maximized is given by $0.06

I
2
-.-po.10
- 0

Figure 2. Experimental region and settings for experiments for

sequential model discrimination at 390 O C .

Partial pressure ot n - b u t a n e d m
After N experimental runs, one selects the ( N + 1)th run
g>,&0175 002 04225 0025 0.0275 0,03
(&@
a t those operating conditions that maximize D. In the C
L
above model, T stands for probability, 2 for the common
variance of the N observations, u t and u? are variances
for the predicted values of +1 under models i and j ,
respectively, and 91f)+land fEl are the predicted values
of YN+1 under models i and J, respectively. After con-
ducting the ( N + 1)th run, the current posterior proba-
bilities for various models are calculated from the Bayes'
theorem by using the following probability density function
for a single observation Figure 9. Experimental region and settings for experiments for
sequential model discrimination at 410 "C.

a constant n-butane partial pressure of 0.03 atm, but

varying partial pressures of oxygen from 0.04 to 0.10 atm,
support model 4. Finally, parameters for this particular
where yN+1 is the value of the response obtained from model (model 4) are estimated considering all runs for each
actual experiment under the ( N + 1)th experimental temperature. This is given in Table IV. Activation en-
conditions. The variance for the predicted values of yN+l ergies are also calculated on the basis of an Arrhenius plot
under model i is calculated by the procedure given by Box (Table V). The activation energy for the oxidation of
and Hill (1967). After conducting ( N + 1)th run, the n-butane to MA obtained in the present study is compa-
posterior probabilities are calculated and if discrimination rable to the value reported by Buchanan and Sundaresan
is not achieved successfully, further experiments are de- (1986).
signed sequentially.
The experimental settings of the discriminatory runs for 4. Conclusions
three temperatures, viz., 370,390, and 410 "C,are shown
in Figures 1,2, and 3, respectively. The current values of The selective oxidation of n-butane to MA can be better
posterior probabilities for the models after each discrim- described by redox models over Langmuir-Hinshelwood
inatory run are shown in Table 111. models. The experimental data best fit a redox model
From the results it is evident that the reaction is de- having zero-order dependency on oxygen partial pressure
pendent on a very low power of oxygen concentration. and first-order dependency on n-butane partial pressure.
Models having first-order and half-order dependency on The model is given by
oxygen concentration can be safely discarded for all the KIKflB
temperatures. Moreover, though models 3 and 4 are closely -rB =
K1+ ffKflB
competing with each other, model 4 is superior to model
3 for all the temperatures. Therefore, model 4 having In previous studies there was much confusion regarding
zero-order dependency on oxygen concentration seems to the order of the reaction with respect to oxygen. The
be the most appropriate one for the reaction. As seen from estimated rate constants based on this particular model
Figure 1, it is confirmed that experiments conducted at showed an Arrhenius dependency on temperature. The
1828 Ind. Eng. Chem. Res., Vol. 30, No. 8, 1991

calculated activation energy for the oxidation of n-butane Registry No. MA,108-31-6;H(CH2),H, 106-9743
to MA is 25.9 kcal/mol, which is comparable to the value
reported by Buchanan and Sundaresan (1986). The ac-
tivation energy for the oxidation of the reduced catalyst Literature Cited
is 12.6 kcal/mol. The value of this activation energy was
not available in the literature. Box, G. E. P.; Hill, W. J. Discrimination Among Mechanistic Models.
Technometrics 1967,9 (l),57-71.
Buchanan, J. S.;Sundaresan, S.Kinetics and Redox Properties of
Nomenclature Vanadium Phosphate Catalysts for Butane Oxidation. Appl.
CataE. 1986,26,211-226.
D = Box and Hill’s design criterion Centi, G.; Fornasari, G.; Trifiro, F. n-Butane Oxidation to Maleic
f = function Anhydride on Vanadium-Phosphorus Oxides: Kinetic Analysis
k = parameters with a Tubular Flow Stacked-Pellet Reactor. Ind. Eng. Chem.
K1 = rate constant for the oxidation of reduced catalyst Prod. Res. Dev. 1985,24,32-37.
K 2 = rate constant for n-butane oxidation Centi, G.; Trifiro, F.; Ebner, J. R.; Franchetti, V. M. Mechanistic
KA = oxygen adsorption equilibrium constant Aspects of Maleic Anhydride Synthesis from C, Hydrocarbons
KR = n-butane adsorption equilibrium constant over Phosphorus Vanadium Oxide. Chem. Rev. 1988,88,55-80.
m = order of reaction with respect to n-butane Lerou, J. J.; Weiher, J. F. Paper presented a t Pittsburgh/Cleveland
n = order of reaction with respect to oxygen Catalysis Society Meeting, 1986.
N = experimental trial Prasad, K. B. S.; Rao, M. S.Use of expected likelihood in sequential
model discrimination in multiresponse systems. Chem. Eng. Sci.
P(A,) = prior probability for the ith model 1977,32, 1411-1418.
P(B/A,) = likelihood for the ith model Roth, P. M. Design of Experiments for Discriminating Among Rival
p , = partial pressure of ith component Models. Ph.D. Thesis, Princeton University, Princeton, NJ, 1965.
P, = probability density function under model i Schneider, P.; Emig, G.; Hofmann, H. Kinetic Investigation and
r = number of models Reactor Simulation for the Catalytic Gas-phase Oxidation of n-
r g = rate of reaction of n-butane Butane to Maleic Anhydride. Ind. Eng. Chem. Res. 1987,26,
R = reduced state of catalyst 2236-2241.
W / F = contact time Srivastava, R. D. Heterogeneous Catalytic Science; CRC Press:
X = oxidized state of catalyst Boca Raton, FL, 1988; pp 41-49.
Tan, H. S.; Downie, J.; Bacon, D. W. The Kinetics of the Oxidation
X = conversion in eq 12 of Propylene Over a Bismuth Molybdate Catalyst. Can. J.Chem.
y = predicted value of response Eng. 1988,66,611-618.
Greek Symbols Thomas, J. M.; Thomas, W. J. Introduction to the Principles of
Heterogeneous Catalysis; Academic Press: London, 1967;Chap-
OR = fraction of surface occupied by R ter 8.
8x = fraction of surface occupied by X Wohlfahrt, K.; Hofmann, H. Kinetics of the synthesis of maleic
x = probability anhydride from n-butane. Chem. Ing. Tech. 1980, 52 (lo),
x,, = posterior probability of ith model (i = 1-4) after con- 811-814.
ducting j experiments
g2 = common variance of n observations Received for review July 2, 1990
u,*= variance of predicted values of observations under model Revised manuscript received February 6, 1991
1 Accepted February 12, 1991