You are on page 1of 13

6730 Ind. Eng. Chem. Res.

2003, 42, 6730-6742

Butane Oxidation to Maleic Anhydride: Kinetic Modeling and


Byproducts
Marı́a J. Lorences,†,‡ Gregory S. Patience,*,§,| Fernando V. Dı́ez,† and José Coca†
Department of Chemical Engineering and Environmental Technology, University of Oviedo,
Julian Claverı́a s/n, Oviedo, Asturias, Spain 33209, and DuPont Ibérica SA, Tamón, Avilés, Spain 33469

Reactor technology for maleic anhydride continues its evolution. New processes achieve higher
yields with lower investment by operating in a net reducing environment where the oxygen
concentration is lower than that required stoichiometrically to react all of the butane. In this
paper, we examined a wide range of operating conditions to quantify the effect of a reducing
environment on maleic anhydride selectivity, byproduct acid productivity, and reaction rates.
The experiments were carried out with a vanadium phosphorus oxide catalyst in a fluidized-
bed reactor and a novel feed gas manifold. Oxygen, carbon monoxide, butane, and acid
concentrations were measured online at a frequency of about 1 Hz. Acetic and acrylic acids were
the predominant byproduct acids, but fumaric, methacrylic, and phthalic acids were also detected.
Under reducing conditions, carbon adsorbed on the catalyst surface, byproduct acid yields
increased, and both the selectivity and reaction rates decreased. A redox kinetic model was
developed to account for the experimental observations and included both V5+ and V4+ oxidation
states and a “VC4” complex, which represented carbon adsorption.

1. Introduction condensable gases. The feed stream surpasses the 1.8


vol % butane limit and operates at concentrations near
Maleic anhydride is synthesized commercially by 4% with oxygen concentrations in the range of 10%.
partially oxidizing n-butane over vanadium phosphorus Although several processes under development oper-
oxide (VPO) catalysts. Over the last 10 years, its price ate with high inlet butane concentrations, most kinetic
on the merchant market has dropped significantly, and studies published in the literature are restricted to
the decline has been due to catalyst improvements, highly oxidizing reaction conditions typical of conven-
process innovations, and economies of scale. Early tional fixed beds (see work by Hutchings4). However,
technology was entirely based on fixed-bed reactors with under reducing conditions, neither is oxygen a limiting
benzene not n-butane as a feedstock. The fluidized-bed reagent nor is catalyst reoxidation a rate-limiting step.
process was commercialized in the late 1980s and has Therefore, the kinetic expressions have limited useful-
several advantages including superior heat transfer, ness for either fluidized-bed processes, CFB technology,
more concentrated product streams, and larger scale. membrane reactors, or fixed-bed processes where oxygen
In the mid-1990s, circulating fluidized-bed (CFB) tech- is the limiting reactant.
nology was commercialized in which catalyst is shuttled Many recent studies5-7 have been devoted to charac-
between an oxidizing environment and a reducing terizing the catalyst performance under fuel-rich condi-
(butane-rich) environment. This process has good heat- tions and under a cyclic mode where the catalyst is
transfer characteristics but has the advantage of even exposed to successive pulses of a net reducing and
larger scale than conventional fluidized beds as well as oxidizing environment.8-11 The former work was de-
an even more concentrated product stream.1 voted to evaluating the potential of membrane reactors,
Concentrated product streams and high butane feed whereas the latter was pertinent to CFBs. The reducing
concentrations translate into reduced vessel sizes (cata- conditions expected at the inlet of a membrane reactor
lyst inventories) and thus superior economics. Mon- lead to a rapid reduction of the catalyst surface and,
santo2 have operated a fixed-bed reactor in the flam- therefore, formation of butenes and COx may be favored.
mability region (C4H10 > 1.8 vol % in air) and claimed Mallada et al.12 report that, in fixed-bed reactors
that the hot spot can be overcome using 40% dilution operating at high butane concentrations, the catalyst
in the first part of the reactor. Recently, Pantochim3 near the effluent becomes less selective as the gas-phase
claims to have improved fixed-bed process economics by oxygen is depleted: the COx selectivity increases, and
feeding pure O2 instead of air and recycling the non- the overall activity declines.
In this paper, we present new experimental data
* To whom correspondence should be addressed. E-mail: under both butane-rich and -lean conditions. All experi-
Gregory.Patience@INVISTA.com. ments were run in a fluidized-bed reactor with a

University of Oviedo. Fax: +34-98-510-3434. E-mail: fds@ capacity of up to 0.5 kg. The gas manifold was specifi-
correo.uniovi.es. cally designed to operate either under steady-state

Current address: HALDOR TOPSØE A/S, Nymøllevej 55,
2800 Lyngby, Denmark. Tel.: +45-45272330. E-mail: mlp@
feeding conditions or under unsteady-state cycling
topsoe.dk. conditions to monitor transients and simulate typical
§
DuPont Ibérica SA. CFB operation. Together with the wide range of butane
|
Current address: INVISTA (International) SA, European and oxygen concentrations, we evaluated the catalytic
Technical Centre 146, route du Nant-d’Avril, CH-1217 Meyrin, performance at various temperatures, gas flow rates,
Geneva, Switzerland. Tel.: +41-22-7176996. and water vapor feed rates.
10.1021/ie0302948 CCC: $25.00 © 2003 American Chemical Society
Published on Web 11/20/2003
Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003 6731

electrically heated sand bath (30 A) to maintain near-


isothermal conditions. The axial temperature gradient
was monitored with a 10-point thermocouple positioned
equidistant along the height. The temperature was
constant along the bed height but dropped substantially
in the freeboard section, indicating that gas-phase
product decomposition was minimal. Air, nitrogen, and
butane feed streams are premixed outside of the bed.
As illustrated in Figures 1 and 2, the feed gas is fed
through a tube from the top of the heated sand bath.
The tube is coiled around the circumference of the
reactor in the sand bath to preheat the gas. It entered
the bottom of the reactor into a plenum and then
through a sintered metal frit that distributed it evenly
across the diameter. We tested various superficial gas
velocities, but most experiments were conducted in the
Figure 1. Fluidized-bed reactor in detail. bubbling fluidized-bed regime with a solids bed height
between 100 and 500 mm. A flanged upper section
The objectives of the paper are to quantify the effect minimized solids entrainment to the top of the reactor,
of a reducing inlet gas composition on the byproduct acid and a sintered metal filter retained all catalyst in the
profile and evaluate the kinetic reaction rates. We reactor.
compare several kinetic models with our experimental Feed/Effluent Section. Figure 2 illustrates the feed
data. Many models predict butane conversion well but reactor manifold as well as the effluent analysis con-
not selectivity. Most models are entirely empirical and figuration. Both nitrogen and air flow rates were
therefore have limited use in predicting the reactor controlled with Tylan FC-2900V-4S flow controllers.
performance outside the scope of experimental condi- Accurate measurement of the butane flow was prob-
tions. To develop a more rigorous model, we consider lematic, and we eventually succeeded in feeding butane
both steady-state and transient experiments. Under at a constant rate with a Brooks 5850S mass flow
cyclic operation, carbon may adsorb strongly to the controller heating the line and the flowmeter to 60 °C.
surface and only desorb when oxygen is fed to the An eight- and four-way valve system was developed
catalyst. The new kinetic expression characterizes this either to cycle butane/N2 or oxygen/N2 (simulating CFB
phenomenon and at the same time correctly character- operation) or to feed them together.
izes the effect of butane and oxygen concentrations
under steady-state conditions. Further work is required The reactor effluent was maintained at 200 °C to
to incorporate the byproduct acids as well as the product prevent product condensation in the line, and it was
split between CO and CO2. scrubbed with water recycled from the reactor quench.
The accumulated acids in the quench/absorber were
sampled frequently and analyzed offline by high-
2. Experimental Section
performance liquid chromatography (HPLC; Hewlett-
All catalytic testing was carried out in a Hastelloy Packard 1050) equipped with a variable-wavelength UV
C-276 vessel of 0.04 m diameter and 0.79 m height. Its detector. An ION-300 column was used to separate the
design pressure and temperature were 51 bar and -30/ different acids, which included acetic, acrylic, succinic,
620 °C, respectively. The reactor was immersed in an fumaric, propionic, butyric, methacrylic, and phthalic

Figure 2. Experimental setup: (1) compressor; (2) electroneumatic valve; (3) mass flow controller; (4) check valve; (5) heater; (6) 10-
point thermocouple; (7) fluidized sand bath; (8) peristaltic pump; (9) absorber; (10) oxygen paramagnetic analyzer; (11) gas chromatograph;
(12) integrator; (13) butane and CO IR analyzers; (14) data acquisition system; (15) computer.
6732 Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003

Figure 3. Evolution of the oxygen and acids concentration under


experimental conditions.
Figure 4. SEM image of the VPO fluidized-bed catalyst.
acids. In addition to the frequent sample analysis, the
quench liquid conductivity was measured continuously Table 1. Particle Properties of the VPO Catalyst
online to monitor the acid production rate. At the end
dp < 10 0 100 < dp < 125 11.1
of each experiment, the absorber was washed three 10 < dp < 20 0.5 125 < dp < 150 7.6
times. These samples were also analyzed by HPLC and 20 < dp < 25 2.0 150 < dp < 200 8.9
included in the mass balance calculations. 25 < dp < 32 5.2 200 < dp < 250 3.9
Downstream of the absorber, a slip stream of the 32 < dp < 44 12.7 dp,sv 64
effluent gas passed through an ice trap and the non- 44 < dp < 66 22.9 Fp 1890
condensables were measured with various online ana- 66 < dp < 88 17.0 Fsk 2990
88 < dp < 100 7.1 Umf 0.0056
lytical instruments. The rest of the exit gas was bubbled
in water, and both liquid samples were analyzed by
forces and abrasion typical of fluidized-bed reactors.13
HPLC. A relatively small amount of carbon was col-
The scanning electron microscopy (SEM) image (Figure
lected by the ice trap and in the purge gas. The
4) shows that the particles are spherical and the surface
efficiency of the absorber was greater than 99%.
is smooth with very few asperities. After several hun-
A Hewlett-Packard 5890 gas chromatograph equipped
dred hours of exposure to reaction conditions, we began
with both a flame ionization detector and a thermal
the kinetic experiments. We frequently returned to a
conductivity detector measured the concentrations of
base case condition and found that the catalyst activity
CO, CO2, O2, N2, and C4H10. Oxygen was also monitored
was constant throughout the several months required
online (in real time) with a Siemens Oxymat 5F
to complete the study.
paramagnetic analyzer that had a response time of 1 s
Table 1 shows the particle size distribution of the VPO
for 80% of full scale. ABB model 501B IR analyzers
catalyst analyzed with a Coulter instrument and some
monitored both CO and C4H10 gases (online and in real
other particle properties relevant to fluidized-bed reac-
time).
tors. The mean particle diameter (dp,sv) was calculated
Figure 3 shows an example of the oxygen concentra-
according to the following expression:
tion of the effluent gas and the quench liquid conductiv-
ity of a typical experiment. During the first 5 min,
1

( )
oxygen rises sharply from 0% to over 7%. Thereafter, it dsv ) (1)
xi
reaches a constant value over 7.5% after 10 min. The
conductivity increases rapidly during the first several ∑ dpi
minutes and then increases with time linearly. (Note
that the acid concentration and conductivity are related The particle and skeletal densities were measured by
by a second-order polynomial function.) helium pycnometry, and the minimum fluidization
To reduce error related to the transient, all mass velocity was measured experimentally.14
balance calculations were based on acid and GC analysis
measured after 18 min. A typical experiment lasted from 3. Results and Discussion
70 to 180 min and depended ultimately on the repro-
ducibility of the GC traces; a minimum of three were 3.1. Experimental Results. Three different sets of
collected for each experiment. steady-state experiments were carried out to assess the
To ensure safe operation, the reactor was controlled influence of operating conditions on the reactor perfor-
with a PLC and instrumented with interlocks for high mance and to develop a suitable model for the butane
reactor pressure and temperature. Power failure or low to maleic anhydride reaction kinetics. The first set of
flow rates would also trigger interlocks, and the fail safe experiments was conducted at atmospheric pressure
mode was to switch air/butane feeds to nitrogen. In the and different flows, temperatures, and feed concentra-
case of low flow and high temperature or pressure, tions. The second set was carried out with half of the
power to the electrical heaters would shut off. amount of catalyst of that used in the first series.
Catalyst. A VPO catalyst used in this study was Finally, a statistical design of experiments was per-
prepared in an “organic medium” and encapsulated in formed with the butane concentration, temperature, and
a porous silica shell in order to make it resistant to reactor gas velocity as factors. Table 2 details the target
Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003 6733

Table 2. Experimental Conditions (Variables and Levels)


W W1, 2W1
P 1
F 0.76, 1.8, 3.6
T 623, 653, 693
C4H10 inlet, vol % 2, 5, 9
O2 inlet, vol % 4, 10, 20

value of the operating conditions, and Table 3 sum-


marizes a selection of the raw data.
A few experiments were conducted at 2 and 4 barg,
but the results obtained were not fully satisfactory
because of the unreliability of the feeding butane at
these pressures. It would condense in the line, and thus
its flow rate control was poor. Another set of experi- Figure 5. Conversion of butane vs selectivity to maleic anhydride
ments was also run, introducing different amounts of at different reaction temperatures and feed concentrations.
water in the feed, but the results were not conclusive.
Experimental data where the error of either carbon or However, at these conditions, maleic acid production
oxygen mass was greater than 5% were rejected. rates were 60% of those recorded with 4% butane and
Figures 5-7 demonstrate the effect of both temper- 9% oxygen!
ature and feed gas concentrations on butane conversion Maleic selectivity is greatest at low temperatures,
and maleic anhydride and main byproduct acid selectiv- indicating that its activation energy is lower than that
ity. At all temperature and feed gas concentrations, the for the reaction to combustion products. The maximum
maleic anhydride selectivity decreases with increasing is in the range of 70% and is achieved at 350 °C
butane conversion. Buchanan and Sundaresan15 re- regardless of the inlet gas composition. Figure 5 shows
ported a maximum in maleic yield at around 90% that there is a near-linear relationship between the
conversion, whereas Centi et al.16 observed a maximum maleic anhydride selectivity and conversion: For each
at about 80% conversion operating at 300 °C on a high inlet gas composition, the selectivity declines steadily
surface area fresh VPO catalyst. In this study, maleic with increasing conversion. The slope of the curve is
yields (defined as conversion multiplied by selectivity) almost vertical under the most reducing conditions and
are highest at about 80% with 1-2% butane in the feed. progressively becomes less steep as the environment
Table 3. Experimental Results
Q inlet, C4H10 inlet, O2 inlet,
T, K cm3/min vol % vol % X, % SMA, % Sacet, % Sacry, % Sfum, % Smacr, % Sphthal, % SCO, % SCO2, %
623 1800 1.30 4.06 35.61 66.32 4.74 0.86 0.08 0.03 0.00 12.69 15.27
623 1800 1.30 10.03 45.34 64.42 5.94 0.86 0.06 0.04 0.00 13.06 15.63
623 3580 1.61 3.96 22.35 64.68 8.87 1.15 0.04 0.08 0.00 13.88 11.30
623 3580 1.61 9.94 27.84 67.17 5.85 1.11 0.04 0.07 0.00 12.88 12.87
623 1790 4.67 3.95 12.75 64.15 5.04 1.45 0.05 0.11 0.53 12.76 15.90
623 1790 4.68 9.95 20.46 66.16 7.21 1.31 0.06 0.09 0.19 11.30 13.68
623 3590 4.81 3.94 8.14 68.59 0.92 1.53 0.05 0.15 0.54 14.17 14.06
623 3585 4.82 9.93 11.65 66.10 4.74 1.31 0.05 0.12 0.22 13.43 14.03
623 1790 9.05 3.95 7.33 54.60 10.71 1.52 0.06 0.19 1.33 13.34 18.24
623 1800 9.00 9.89 13.09 64.92 6.12 0.03 0.08 0.12 0.43 13.64 14.66
623 3590 9.02 3.94 4.62 60.79 4.77 1.46 0.04 0.25 1.01 14.99 16.69
623 3600 9.00 9.89 8.04 64.61 6.03 1.54 0.05 0.17 0.41 13.23 13.96
653 1800 1.50 4.05 43.08 56.65 5.09 0.84 0.10 0.06 0.10 18.94 18.22
653 1800 1.50 10.01 56.71 52.09 8.56 0.77 0.06 0.08 0.16 20.15 18.14
653 3580 1.61 3.96 29.66 60.21 9.60 1.06 0.06 0.09 0.15 14.19 14.64
653 3585 1.61 9.94 42.39 51.88 6.81 1.17 0.04 0.12 0.27 23.07 16.64
653 1790 4.67 3.95 16.25 48.74 6.13 1.59 0.05 0.08 0.82 21.48 21.10
653 1790 4.68 9.95 31.11 55.59 8.12 1.41 0.06 0.08 0.13 17.91 16.69
653 3590 4.81 3.94 12.30 54.72 8.76 1.64 0.06 0.10 0.74 16.20 17.78
653 3585 4.82 9.93 23.84 54.46 5.86 1.37 0.06 0.20 0.12 21.31 16.61
653 1790 9.05 3.95 8.73 41.69 9.52 1.92 0.11 0.16 0.01 20.49 26.11
653 1800 9.00 9.89 18.95 47.69 5.34 2.14 0.07 0.12 0.00 23.82 20.81
653 3590 9.02 3.94 6.34 49.85 5.74 1.84 0.07 0.18 0.01 21.27 21.04
653 3600 9.00 9.89 13.73 52.42 6.91 1.56 0.07 0.23 0.01 22.00 16.80
653 3615 2.00 19.81 35.53 62.87 2.77 1.23 0.06 0.07 0.04 17.30 15.66
653 1805 1.84 19.84 60.30 49.43 3.18 0.76 0.09 0.04 0.06 25.37 21.08
693 1805 1.50 4.05 50.44 51.87 3.39 0.68 0.04 0.00 0.00 22.01 22.01
693 1800 1.50 10.01 68.48 44.60 7.01 1.62 0.05 0.03 0.00 23.83 22.86
693 3580 1.61 3.96 35.39 54.96 4.42 0.82 0.03 0.04 0.00 21.77 17.96
693 3585 1.61 9.94 47.04 52.98 4.09 0.85 0.36 0.04 0.00 21.55 20.12
693 1790 4.67 3.95 19.08 35.01 3.92 1.54 0.48 0.11 1.19 29.34 28.41
693 1790 4.68 9.95 41.52 49.85 4.46 1.08 0.08 0.14 0.36 21.75 22.27
693 3590 4.81 3.94 14.65 45.03 3.74 0.02 0.05 0.12 1.01 26.03 23.99
693 3585 4.82 9.93 29.53 53.59 4.50 1.26 0.06 0.06 0.18 21.24 19.09
693 1790 9.05 3.95 8.90 28.91 4.23 2.32 0.13 0.19 1.76 27.18 35.28
693 1800 9.00 9.89 23.82 42.75 5.35 1.66 0.06 0.09 0.00 25.11 24.98
693 3595 9.02 3.94 8.98 37.28 5.91 2.00 0.13 0.15 0.11 27.18 27.24
693 1735 2.49 19.71 64.06 47.00 3.27 0.86 0.08 0.05 0.03 24.53 24.18
410 3450 1.80 19.85 48.86 51.10 3.62 1.06 0.05 0.08 0.02 22.93 21.14
6734 Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003

Table 4. Product Distribution at 633 K in the Exit Gas


Line
t ) 20 min mg/min t )75 min mg/min
C4H10 3.14% 3.19%
CO2 2.83% 1.69%
CO 1.90% 1.70%
maleic acid 1075 ppm 26.9 12900 ppm 108
acetic acid 86 ppm 2.2 592 ppm 4.6
fumaric acid 1.95 ppm 0.05 11 ppm 0.08
acrylic acid 276 ppm 6.9 475 ppm 1.81
methacrylic acid 5.4 ppm 0.27 5.5 ppm 0.0009
phthalic acid 0 ppm 0 25 ppm 0.23

conditions, it was generally lower than 0.04%. The origin


of phthalic anhydride is unclear, but one possibility
Figure 6. Conversion of butane vs selectivity to acetic acid at
would be impurities in the n-butane feed: n-pentane is
different reaction temperatures and feed concentrations.
known to react over VPO to produce phthalic anhydride.
However, the butane purity was 99.5% with less than
0.4% i-butane, less than 0.1% propane, and only 0.05%
other hydrocarbons; therefore, it would seem unlikely
that n-pentane would be the origin of this acid. Several
other unidentified peaks were present on the chromato-
gram, but because the response is so highly dependent
upon the carbon-carbon bonds, estimating their relative
abundance was not attempted.
Figure 6 plots butane conversion versus acetic acid
selectivity at different temperatures and inlet gas
compositions. The tendency is similar to that for maleic
acid: selectivity is generally higher at lower tempera-
tures and decreases with increasing conversion. Under
Figure 7. Conversion of butane vs selectivity to acrylic acid at reducing conditions, the selectivity drops very rapidly
different reaction temperatures and feed concentrations. with small increases in the conversion. Under oxidizing
becomes more oxidizing. On average, a 30 °C temper- conditions, acetic acid selectivity is somewhat lower but
ature rise drops the selectivity by 10%; at low butane is less sensitive to butane conversion; it is an order of
concentrations (2%), it drops by 8% for a 30 °C rise, and magnitude lower than maleic acid and about 5 times
it drops by 12% at the highest butane concentration higher than acrylic acid.
tested (Figure 5). Centi et al.16 reported a decrease in Under reducing conditions, the acrylic acid selectivity
the selectivity related to the ratio of residual concentra- can reach as high as 50% of the acetic acid selectivity
tion of oxygen to n-butane (operating at 0.32, 0.7, and but it is generally on the order of 20%. At high oxygen
1.1% butane in air). This ratio is proportional to the partial pressures, the acrylic acid selectivity is generally
oxidizing power of the mixture. It is suggested that at below 1% and it reaches 2.5% at high butane partial
the end of the catalytic bed the maleic anhydride pressures. It increases with an increase in both the
produced is not stable as a result of the overoxidation temperature and conversion, as shown in Figure 7,
of the catalyst surface. Under these low feed butane which is opposite to the trend for both maleic and acetic
concentrations, an increase of the V5+ concentration at acids. This might indicate that its formation follows a
the end of the catalyst bed was confirmed by chemical completely independent pathway or that it may be a
and spectroscopic analyses of the spent catalyst.17 decomposition product of maleic acid.
The relationship between the reaction conditions and Some experimental data suggest that acrylic acid is
conversion is more complex. Under highly oxidizing indeed a maleic anhydride decomposition product. The
conditions (2% butane and air), butane is the limiting effluent line was generally maintained at a temperature
reagent and, therefore, conversion is limited by the of 200 °C in order to avoid product condensation. In one
contact time and temperature. Under reducing condi- test, it was heated to 360 °C for 30 min and brought
tions (9% butane/4% oxygen), oxygen is the limiting back down to the standard temperature afterward.
reagent. The temperature or contact time has little Acids were collected for an additional 55 min under
impact on the absolute value of the butane conversion identical operating conditions (5% C4, 10% O2, 410 °C,
but affects the selectivity considerably. For example, by 3600 mL/min). Two different liquid samples were taken
increasing the temperature by 60 °C, the butane con- after 20 and 75 min and were analyzed by HPLC. The
version increased from 7% to 9% but the selectivity data are summarized in Table 4 and show some re-
dropped by over 25%. markable differences in product distribution. Exit bu-
Buchanan and Sundaresan15 showed that acetic and tane concentrations were equivalent at both times,
acrylic acids are the principle byproduct acids, and they which would indicate that it was essentially operating
also detected trace amounts of ethylene. Together with at steady state. However, both the CO and CO2 concen-
these acids, we also detected low concentrations of trations were higher when the heating tape was at 360
phthalic, methacrylic, and fumaric acids. The average °C; the CO2 concentration was twice as high, which
selectivities were 0.3%, 0.1%, and 0.08%, respectively, indicates that some of the acids were combusting. On
and the data are summarized in Table 3. Selectivity of the basis of the change in acid production rates, clearly
phthalic acid reached as high as 1.7% at high temper- both the maleic anhydride and acetic acid combust:
atures and a reducing atmosphere. Under oxidizing their formation rates are 4 and 2 times lower during
Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003 6735

with a surface oxygen and the subsequently reduced


catalyst is reoxidized by molecular oxygen. Such redox
mechanisms can be formally described by the kinetic
equations of Mars-van Krevelen or Eley-Rideal.
Figure 8. Simplified butane oxidation reaction network.
Several authors18,21-23 report the presence of different
the time at which the heating tape was maintained at vanadium oxidation states under reaction conditions,
the higher temperature. and several reaction mechanisms have been proposed.
Some of these assume both the selective oxidation of
3.2. Kinetic Modeling. In general, the experimental
butane to maleic anhydride and maleic anhydride
evidence in the published literature suggests that
decomposition taking place on the same active site,18,22
vanadium pyrophosphate [(VO)2P2O7 (VPO)] is the
while poisoning studies revealed24 that these reactions
active phase for selectively reacting butane to maleic
occur on two different sites corresponding to V4+and V5+
anhydride. The reaction is very complex and involves
oxidation states. Although it is now generally accepted
abstraction of eight hydrogen atoms and insertion of
that the presence of both V4+and V5+ is required during
three oxygen atoms into the four-carbon skeleton: 14
the reaction, the role of the different oxidation states of
electron transfer. Although this phase has been well
vanadium is still unclear.25 Hutchings et al.26 demon-
characterized, its performance may depend on the
strate a distinct parallel between improving the cata-
preparation method, and many patents have been
lytic performance and a decrease in the amount of V5+
granted declaring an optimal manufacturing method.
present.
Together with the precursor recipe (reducing agent
and solvent), other factors that influence the VPO Many intrinsic rate expressions have been developed
catalyst performance include the P/V ratio, promoters, to describe both the selective and nonselective reaction
vanadium oxidation state, and calcination/activation kinetics. Some of the expressions are entirely empirical,
procedure. The relation between these parameters and while others are based upon Langmuir-Hinshelwood
the catalytic performance was reviewed by Hodnett,18 and redox-type mechanisms. Table 5 summarizes a
and it can be concluded by the different studies that number of different kinetic models proposed in the
the strength of the reducing agent determines the literature, as well as the reactor and catalyst type and
average oxidation state of the precursor and therefore the experimental conditions.
determines the activation temperature. The P/V ratio Hydrodynamic Modeling. To accurately assess the
influences the redox properties of the catalyst, the phase reaction kinetics, an adequate transport model for the
composition, and the distribution of vanadium oxidation gas and solid phases is essential. Both steady-state and
states in the catalyst. The VPO precursor is prepared transient techniques are commonly employed to quan-
in either an aqueous or organic medium. The latter tify the gas-phase hydrodynamics. In this work, we
usually results in a lower oxidation potential of V4+ to employed a transient step response method. A constant
V5+, higher surface area, and higher maleic anhydride flow of nitrogen was fed to the reactor, and then it was
yields.19 The P/V ratio strongly influences the oxidation switched to air. The oxygen concentration was moni-
state as well as the selectivity to maleic anhydride. P/V tored online at a frequencey of 1 Hz. This step change
ratios on the order of 1.1:1 and bulk catalyst oxidation to the input function is referred to as the heavyside unit
states near 4+ are considered optimal. Phosphorus step function and is shown in Figure 9. Time “0” is the
appears to modulate the catalyst activity, and as it is moment at which the valve is turned. As the air sweeps
depleted with time, the activity generally increases and across the catalyst, the oxygen concentration increases
the selectivity decreases. gradually to its steady-state value of approximately
A simple triangular reaction scheme is frequently 21%. At low gas velocities, the reactor is operating in
adopted to describe the reaction mechanism, as shown the fixed-bed fluid regime and axial dispersion is quite
in Figure 8. Butane may react to form either maleic low. At 6 mm/s, the bed operates between the minimum
anhydride or CO/CO2 (COx). Maleic anhydride may also fluidization and minimum bubbling fluidization re-
combust to form COx. This scheme clearly neglects gimes. The bed is in the bubbling fluidized regime at
byproduct acid evolution, which may account for as 12 and 18 mm/s, and these were the velocities at which
much as 10% of the total under certain conditions. most kinetic experiments were conducted.
Moreover, it oversimplifies the variation of the carbon Cold flow experiments (atmospheric pressure and
oxide selectivity with reaction conditions. For example, room temperature) were made in a quartz reactor (4.1
under oxidizing conditions, ratios of CO/CO2 may run cm i.d.) at different gas velocities in the fixed-bed and
as high as 1.6:1.8. A similar CO/CO2 ratio was also bubbling regimes, a second set of experiments were
observed by Buchanan and Sundaresan15 using a VPO made in the Hastelloy reactor at temperatures of 350,
catalyst with P/V equal to 1. They reported a decreasing 380, and 410 °C, pressures of 1, 3, and 5 bar abs, and
of the CO/CO2 ratio with increasing conversion and different amounts of catalyst, and a third set was
decreasing temperature. A CO/CO2 higher than 1 was conducted to evaluate the transport effects due to the
also observed by Wohlfahrt and Hofmann.20 In our case, connecting tubing and empty reactor volume. All tests
the ratio drops to below 1.2 as the conditions become were repeated three times at each gas velocity, and the
more reducing and tend to 0.9 at a butane concentration plots are superimposed in Figure 9. Reproducibility was
of 5-10%. Lumping CO and CO2 together when model- very good.
ing the kinetics is not strictly correct, but because the Figure 10 shows the first derivative of the step
carbon-surface oxide reaction mechanism is so compli- response experiments and represents the Dirac δ input
cated, it is convenient to ignore side reactions and lump function. In the case of ideal gas flow (fixed beds), the
the carbon oxides. response curve is symmetrical around the average
It is generally accepted that the surface activation of residence time. At low gas velocities, 0.3 cm/s, the
butane is the rate-limiting step and that a redox model response curve is nearly symmetrical. At higher gas
best represents the reaction mechanism: Butane reacts velocities, the curves are less symmetrical and have a
6736 Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003

Table 5. Kinetic Models for the Selective Oxidation of Butane to Maleic Anhydride over VPO Catalyst
catalyst T, K O2 C4H10 kinetic expressions
k1,2CB
r1,2 )
K1CB K2CMA
organic mediuma,15 663-713 21% 0.62-1.55% 1+ +
CO2 CO 2

k3CMA
r3 )
K1CB K2CMA
1+ +
CO 2 CO 2
k1KBCBCO2R
r1 )
1 + KBCB
organic mediuma,16 573-613 O2/C4H10 ) 15 0.32-1.1% r2 ) k2CO2β
CO2γ
r3 ) k3CMA
CBδ

k1pBR1
r1 )
1 + K2CMA
industrial catalyst?a,27 573-653 21% e3% r2 ) k2pBR2
k3pM
r3 )
(1 + K2CMA)2

organic mediuma,28-30 693-713 13% 0.56-2.38% expressions in section 3.2


r1 ) k1pB
V2O5/P2O3a,31,32 673-753 21% <0.95% r2 ) k2pB
r3 ) k3pMA

k1(K1pO2)1/2
r1 ) pB
1 + (K1pO2)1/2
k2K2pO2
aqueous mediumb,33 693-753 5-30% 2.5-5.5% r2 ) pB
1 + K 2p O 2

k3K2pO2
r3 ) pB
1 + K 2p O 2

k1KBpBpO2
r1 )
1 + KBpB + KMApMA + KH2OpH2O

k2KBpBpO2
r2 )
1 + KBpB + KMApMA + KH2OpH2O

k3KBpBpO2
organic mediumc,34 603-723 7-13% 0.6-3.5% r3 )
1 + KBpB + KMApMA + KH2OpH2O

k4KBpBpO2
r4 )
1 + KBpB + KMApMA + KH2OpH2O

k5KBpBpO2
r5 )
1 + KBpB + KMApMA + KH2OpH2O

k1pBpO20.5
r1 )
pB
1 + KB 0.5 + KCpO20.5
pO2

organic mediumb,6 648-723 1-20% 10-30% k2pBpO20.5


r2 )
pB
1 + KB 0.5 + KCpO20.5
pO2

k3pB
r3 )
pB
1 + KB 0.5
+ KCpO20.5
pO2
Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003 6737

Table 6. n-CSTR in Series Model Parameters


Ug σ2 tm n Pe D
0.003 385 118 36 70 0.00001
0.006 339 61 11 20 0.00008
0.012 129 29 7 11 0.0003
0.018 67 20 6 10 0.0005

gives the following expression:

Q n
) (3)
Vj t m

where n is the number of tanks and tm the mean


residence time. The residence time distribution for the
whole system is obtained from a mass balance for n
Figure 9. Oxygen breakthrough curves as a function of the gas
tanks:35

[ ()
velocity (F curve).
nt 1 nt 2
F(t) ) 1 - e-nt/tm 1 + + +
tm 2! tm
... +
1
() ]
nt
(n - 1)! tm
n-1
(4)

tn-1
E(t) ) e-t/tmi (5)
(n - 1)!tmin

Figure 10 compares the model predictions and the


experimental data and shows very good agreement.
The parameters obtained for different gas velocities
are given in Table 6. The number of tanks varies
between 36 at low gas velocities (fixed-bed operation)
and 6 at high gas velocities (bubbling bed). The signifi-
Figure 10. n-CSTR in series model predictions. cance of the assumed number of tanks in series has
profound implications when the conversion of the limit-
steep leading edge and a significant tail. Most hydro- ing reagent approaches 100%. At conversions in the
dynamic studies of fluid beds divide the bed into two range of 35%, predictions of a plug-flow model (infinite
phases: an emulsion phase with a high concentration number of tanks in series) differ from those of a CSTR
of solids and a bubble phase. Bubbles form at the bottom (one tank in series) by only 6%.
of the reactor and ascend through the bed at high Kinetic Parameter Estimation. The kinetic con-
velocities. The bubbles may break apart and coalesce, stants are calculated based on a normalized reference
and this action increases the mass-transfer rate between temperature of 653 K as follows:

[ ( )]
the two phases. The leading edge may represent the fast
bubbles, and the tail may result from a certain degree -Eaj 1 1
kj ) kj,653 - (6)
of backmixing. R T 653
In general, the best-fit parameters for the kinetic
constants were reasonably insensitive to the assumed Kinetic parameters were estimated by minimizing the
degree of staging of the solid phase. Therefore, all following objective function, where R2 represents the
kinetics are based on the assumption that the solids coefficient of determination:
were completely backmixed.
n-CSTR in Series Model. Two single-phase models
Φ ) 3 - RX,C4H102 - RS,MA2 - RO22 (7)
may be used to characterize the step response experi-
ments: plug flow with axial dispersion and a series of where
perfectly mixed tanks.
We have adopted the tanks-in-series model as a first 2
∑(XC H 4 10,calc
- XC4H10,exp)2
RX,C4H10 )1- (8)
∑(XC H
approximation for the gas-phase hydrodynamics. For
small deviations from plug flow, this model agrees well 4 10,exp
h C4H10,exp)2
-X
with the axial-dispersion model.
A tracer mass balance around tank j yields the
RS,MA2 ) 1 -
∑(SMA,calc - SMA,exp)2 (9)
following equation:
∑(SMA,exp - Sh MA,exp)2
Cj )
Q -tQ/Vj
Vj
e ∫0tCj-1etQ/V dt
j
(2)
2 ∑(O2,calc - O2,exp)2
RO2 )1- (10)
∑(O2,exp - Oh 2,exp)2
where Q is the flow velocity and Vj the volume of tank
j. Assuming that all of the tanks have the same volume Butane conversion (X,C4H10), maleic anhydride selectiv-
6738 Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003

Table 7. Kinetic Parametersa Obtained Using the Model


Reported by Buchanan12
Buchanan (1986) Buchanan (1986)
P/V ) 1.0 P/V ) 1.1 present work
k1,653K (s-1) 1.96 0.61 2.17
k2,653K (s-1) 0.86 0.30 1.34
k3,653K (s-1) 0.07 0.04 0.19
Ea1 30 28 13
Ea2 35 31 25
Ea3 43 33 16
K1 59 20 14
K2 26 12 208
a RX,C4H102 ) 0.94; RS,MA2 ) 0.51; RO22 ) 0.97; Φ ) 0.58.

ity (S,MA), and the exit oxygen concentration were


chosen as the three factors to minimize in the objective
function.
Single Site: Model of Buchanan and Sundare-
san.15 Several kinetic models are published in the
literature, and the first step in this study was to adapt
these models to the kinteic data shown in Table 3. The
Buchanan and Sundaresan15 expression has the form Figure 11. Conversion of butane (a) and selectivity to maleic
of a single-site redox model and follows the reaction anhydride (b) predicted by Bej and Rao’s model.
network shown in Figure 8. At very low butane concen-
trations, the reaction rate approaches first-order with involved in the model are
respect to butane. Butane competes for active sites at k1
higher concentrations, as does maleic anhydride. Yields C4H10 + RV5+ 98 MA + RV4+ + 4H2O (11)
are maximized by maintaining a high partial pressure
of oxygen in the gas phase. k2
MA + βV4+ 98 4COx + βV3+ + H2O (12)
Rate constants and activation energies for two cata-
lysts tested by Buchanan and Sundaresan15 with dif- k3
ferent P/V ratios are given in Table 7. Both selec- C4H10 + γV4+ 98 COx + γV3+ + 4H2O (13)
tive and nonselective reaction rates drop by a factor of
k4
about 3 with the higher phosphorus content. Phospho- O2 + V4+ 98 V5+ (14)
rus has a much less pronounced effect on maleic
anhydride combustion where the reaction rate drops by k5
50%. O2 + V3+ 98 V4+ (15)
The last column of Table 7 summarizes the best-fit
Assuming that the sum of the active sites is equal to 1
rate constants and activation energies for the catalyst and that under steady-state conditions the oxidation
tested in this study. Reaction rates are on the same rates for each site equals the reduction rate, expressions
order of magnitude as the catalyst with a 1:1 P/V ratio, for the fraction of the catalyst in each oxidation state
although the rate constants for the nonselective parallel are as follows:
reaction to carbon oxides (k2) as well as the product
combustion (k3) are higher by 50% and 300%, respec- 3+ 1
θV ) (16)
tively. Active energies are significantly lower. Product 1 + K2 + K1K2
inhibition of maleic anhydride is predicted to be con-
siderably higher for the catalyst tested in this study 4+ K2
compared to those in the Buchanan and Sundaresan15 θV ) (17)
1 + K2 + K1K2
study while the butane inhibition is lower.
The objective function, Φ, which represents the good- 5+ K1K2
θV ) (18)
ness of fit between the model predictions and experi- 1 + K2 + K1K2
mental data, equals 0.58. The model characterizes
butane conversion extremely well (RX,C4H102 ) 0.94) but where
not so well for maleic anhydride selectivity (RS,MA2 )
0.51). In this model, selectivity is largely determined k4[O2]
by the ratio of the rate constants of the parallel reactions K1 ) (19)
Rk1[C4]
(k1 and k2).
Two Sites: Model of Bej and Rao.30 The Bej and k5[O2]
K2 ) (20)
Rao30 model is based on two active sites. The selective βk2[MA] + γk3[C4]
oxidation takes place on V5+ sites, and both maleic
anhydride and butane combust on V4+ sites. The reac- We fit the experimental data to this kinetic model, but
tion pathway is shown in Figure 8, and the steps overall the agreement was slightly poorer than for the
Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003 6739

single-site model, Φ ) 0.61. The model adequately Table 8. Kinetic Parameters Estimated by the Two-Site
characterizes the butane conversion as shown in Figure Modela
11 with RX,C4H102 ) 0.90. However, the coefficient of i k0i, s-1 Eai i k0i, s-1 Eai
determination for maleic selectivity was only 0.55, and 1 0.42 15 4 0.28 17
the graph shows considerable scatter between the 2 0.02 18 5 0.32 11
predicted and experimental values. Note that the best- a RX,C4H102 ) 0.96; RS,MA2 ) 0.68; RO22 ) 0.98; Φ ) 0.39.
fit parameter for butane combustion on V4+ sites was
zero. Values greater than zero increased the objective
function.
Two-Site Redox Kinetic Model. We developed a
two-site model similar to Bej and Rao’s and assumed
that the two catalytic sites were either selective (V5I)
or nonselective (V5II) and that they cycled between the
V5+ and V4+ oxidation states. Butane and maleic
anhydride may react with either. This mechanism is
supported by experimental evidence showing low selec-
tivity when butane is fed to a catalyst exposed to air
for an extended period of time.36 The proposed model is
shown:
k1
C4H10 + RV5I 98 MA + RV4I + 4H2O Figure 12. Evolution of the reaction products during the reduc-
tion (2% C4/N2) and the oxidation of the catalyst.
5I
r1 ) k1[C4]θ (21)
On the basis of a sensitivity analysis, we concluded that
5II
k2
4II the experimental data collected were insufficient to
C4H10 + βV 98 4COx + βV + 5H2O warrant two reoxidation rate constants and, therefore,
r2 ) k2[C4]θ5II (22) one value for both selective and nonselective reoxida-
tions was used (k5 ) k6). On the basis of the results of
k3 the parameter estimation, the model was further sim-
MA + γV5I 98 4COx + γV4I + H2O plified and combustion of maleic anhydride on selective
r3 ) k3[MA]θ5I (23) sites was dropped. (The best-fit value for combustion of
maleic anhydride on nonselective V5+ sites, k3, ap-
k4 proached zero). Table 8 summarizes the parameter
MA + γV5II 98 4COx + γV4II + H2O estimates. The modeling suggests that maleic anhydride
combusts to the same degree on selective and nonselec-
r4 ) k4[MA]θ5II (24)
tive V5+ sites and that the parallel decomposition of
k5
butane on nonselective sites is low.
O2 + V4I 98 V5I r5 ) k5[O2]θ4I (25) Although this model has few fitted parameters com-
pared to the two discussed above, it characterizes maleic
k6 anhydride selectivity somewhat better with RS,MA2 )
O2 + V4II 98 V5II r6 ) k6[O2]θ4II (26) 0.68. This value is still unsatisfactory because there are
considerable differences between the predicted and
An explicit expression for the oxidation states is calcu- calculated selectivities.
lated by combining a site balance with the reaction rates Two-Site Redox Kinetic Model with Carbon
and assuming that the sum of the fraction of the reduced Adsorption. Many of the models proposed in the
and oxidized selective and nonselective sites equals 1. literature can adequately capture the relationship be-
At steady state, the reoxidation rate equals the sum of tween operating conditions and butane conversion.
the reduction rates However, work is required to characterize selectivity,
which is more sensitive to the assumed mechanism. On
r1 + r3 ) r5 (27) the basis of unsteady-state experiments, we reported
that catalyst may become over oxygenated, and these
r2 + r4 ) r6 (28) sites may combust maleic anhydride to form COx. Figure
12 shows another phenomenon impossible to identify
Combining these questions leads to the following ex- with steady-state experimental conditions: carbon ab-
pressions: sorption.
A catalyst that was previously oxidized by air was
k5[O2]
θ5I ) (29) exposed to a reducing atmosphere (2% butane in nitro-
k1[C4] + k3[MA] + k5[O2] gen) for 40 min. It was purged with nitrogen for 10 min,
and then air was fed to the catalyst for an additional
θ4I ) 1 - k1θ5I (30) 40 min. The nitrogen purge and air treatment were
repeated and used as a basis to calculate oxygen uptake
and during the first oxidation step. Butane and carbon
monoxide concentrations were measured online with IR
k6[O2] detectors, and oxygen was measured with a paramag-
θ5II ) (31) netic analyzer. The acid evolution was monitored with
k2[C4] + k4[MA] + k6[O2]
a conductivity meter in the liquid quench.
During the first 6 min under reducing conditions, the
θ4II ) 1 - k2θ5II (32) liquid conductivity increases substantially and levels off
6740 Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003

thereafter. The carbon monoxide peaks after 3 min and Table 9. Kinetic Parameters Estimated by the Two-Site
then tails off to zero after 12 min. During the first Model with Carbon Adsorptiona
oxygen treatment, the liquid conductivity increases i k0i, s-1 Eai i k0i, s-1 Eai
slightly but the evolution of carbon monoxide is signifi- 1 0.33 15 4 0.57 9
cant: it peaks immediately as the oxygen contacts the 2 0.37 14 5 0.06 28
catalyst and after 6 min drops to zero. 3 0.05 17
This experiment demonstrates that butane reacts a RX,C4H102 ) 0.96; RS,MA2 ) 0.83; RO22 ) 0.98; Φ ) 0.23.
with catalyst lattice oxygen. Furthermore, it shows that
carbonaceous species may adsorb to the catalyst surface
and are desorbed as maleic anhydride or combusted
when exposed to gas-phase oxygen.
The kinetic models discussed above ignore the pos-
sibility of carbon adsorbing to the surface. Surface
carbon could severely reduce the reaction rates and at
the same time account for the discrepancies in the
maleic anhydride selectivity predictions. Figure 12
shows that a little maleic anhydride evolves from the
catalyst during the first oxidation step.
To account for this phenomenon, we modified the two-
site model to include a third site that we designate as
VC4. (The C4 designation is arbitrary because we have
no convincing experimental evidence that the butane
skeleton is intact. Some maleic anhydride does evolve,
and for this reason, we adopt this nomenclature.)
The model shown below assumes that butane and
maleic anhydride may react with V5+ and that butane
may further react with V4+ sites to form the VC4
complex. The adsorbed butane reacts with oxygen to
form V4+ and carbon oxides, and oxygen also oxidizes
V4+ to V5+.
k1 Figure 13. Conversion of butane (a) and selectivity to maleic
C4H10 + RV5+ 98 MA + RV4+ + 4H2O (33) anhydride (b) predicted by the two-site model with carbon adsorp-
tion.
k2
MA + βV5+ 98 4COx + βV4+ + H2O (34)
dize the catalyst with adsorbed carbon is significantly
k3 greater than that for the other reactions. However, the
C4H10 + 4V4+ 98 4VC4 (35) rate constant for the carbon deposition reaction is low,
as is the reoxidation rate.
k4 Figure 13 illustrates the good agreement between the
O2 + V4+ 98 V5+ (36) model predictions and expeimental data. The objective
function, Φ, equals 0.23, and RX,C4H102 equals 0.96 and
k5
γO2 + 4VC4 98 4COx + 4V4+ + 5H2O (37) RS,MA2 0.83, which is a substantial improvement over
the four models previously discussed.
The general form is quite similar to the Bej and Rao Figure 14 shows the fraction of V4+, V5+, and VC4 sites
two-site model. One major difference is that the MA calculated as a function of the oxygen concentration at
reacts with oxidized sites and not reduced sites. An the exit of the reactor. It is observed how, at exit oxygen
explicit expression for the reaction rates is derived concentrations below 0.5%, most of the sites would be
assuming that the sum of the active sites equals 1 and VC4. Less than 10% would be V5+ sites and less than
that under steady-state conditions the oxidation rate for 20% V4+. As the exit oxygen concentration increases,
each site equals the reduction rate. The resultant set the amount of V5+ increases considerably. The amount
of equations for the fraction of VC4, V4+, and V5+ sites of V4+ increases at the same time although not so
is identical with the Bej and Rao expressions (eqs 16- drastically. At an oxygen concentration higher than 10%
18, respectively). The major difference is that maleic at the exit, 80% of the sites are in the 5+ state. Note
anhydride combusts on V5+ sites instead of V4+ sites, that, under oxidizing conditions, the number of VC4 sites
and thus the coefficients become K1 and K2 approaches zero and the kinetic expression becomes
identical with the first expression in Table 5: a single-
k4[O2] site redox model.
K1 ) (38)
Rk1[C4] + βk2[MA]
4. Conclusions
and Reactor technology is driving toward increasing bu-
tane partial pressures in order to maximize maleic
γk5[O2]
K2 ) (39) anhydride productivity. As a consequence, gas-phase
4k3[C4] oxygen concentrations must be reduced to minimize the
hazards. The combination of high butane and low
The kinetic parameters obtained with this model are oxygen can result in lower selectivity and higher
summarized in Table 9. The activation energy to reoxi- byproduct acid selectivity. Together with acetic and
Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003 6741

Acknowledgment
The authors gratefully acknowledge the financial and
technical support by DuPont Ibérica SA and the as-
sistance during the experimental work by Mónica
Alonso Carreño.

Nomenclature
Ci ) concentration of species i, mol‚m-3
[C4] ) butane concentration, mol‚m-3
dp ) particle diameter, µm
dp,sv ) surface volume particle diameter, µm
D ) axial dispersion, m2‚s-1
Eai ) activation energy, kcal‚mol-1
F ) total gas flow rate, NL‚min-1
Hmf ) minimum fluidization height, m
ki ) reaction rate constant, various units
k0i ) preexponential factor, various units
[MA] ) maleic anhydride concentration, mol‚m-3
n ) number of tanks in the n-CSTR in series model
P ) reaction pressure, barg
Pe ) Peclet number of axial dispersion
ri ) reaction rate, mol‚s-1‚m-3
RX,C4H102 ) coefficient of determination for butane conver-
sion
RS,MA2 ) coefficient of determination for maleic anhydride
Figure 14. Fraction of V5+, V4+, and VC4 sites as a function of selectivity
the exit oxygen concentration.
RO22 ) coefficient of determination exit oxygen concentra-
tion
acrylic acids, we have identified significant quantities Si ) selectivity to compound i
of fumaric, methacrylic, and phthalic acids. The product T ) reaction temperature, K
profile of the acids changes with operating conditions, tm) mean residence time, s
which indicates that their reaction pathways are dif- Umf ) minimum fluidization velocity, m‚s-1
ferent. Ug ) gas linear velocity, m‚s-1
Published kinetic models characterize the relationship Vi+ ) oxidation state of the catalyst
between butane conversion and operating conditions W ) weight of the catalyst, g
reasonably well but poorly predict maleic anhydride xi ) fraction of particles with diameter i
selectivity. We show that butane may irreversibly X ) butane conversion, %
adsorb onto the catalyst surface from a complex that Greek Symbols
lowers the reaction rate. Under very oxidizing condi-
tions, the surface concentration of this complex is R ) moles of oxygen required to oxidize 1 mol of n-butane
to MA
essentially zero, but it increases as the gas-phase
β ) moles of oxygen required to oxidize 1 mol of n-butane
environment becomes more reducing. Carbon oxides
to COx
evolve when reacted with oxygen. We have developed a γ ) moles of oxygen required to oxidize 1 mol of MA to
new redox kinetic model to account for these observa- COx
tions and show good agreement between the model Fp ) particle density, kg‚m-3
predictions and experimental data for both butane Fsk ) skeletal density, kg‚m-3
conversion and maleic anhydride selectivity. Under Φ ) objective function defined in eq 7
oxidizing conditions, the model is equivalent to the θi ) fraction of surface covered by the ith site
Buchanan and Sundaresan expression; it has the same
form as the Bej and Rao kinetic model except that Literature Cited
maleic anhydride reacts with oxidized sites (V5+) instead
of reduced sites (V4+). (1) Contractor, R. M.; Garnett, D. I.; Horowitz, H. S.; Bergna,
H. E.; Patience, G. S.; Schwartz, J. T.; Sisler, G. M. A new
Maleic anhydride production rates were greatest at commercial scale process for n-butane oxidation to maleic anhy-
butane and oxygen concentrations of 4% and 9%, dride using a circulating fluid bed reactor. In New developments
respectively. However, the data reported show that in selective oxidation; Corberán, V. C., Bellon, S. V., Eds.; Elsevier
byproduct acid production also increases, which may Science BV: Amsterdam, The Netherlands, 1994; pp C3-1-C.3-
7.
have important commercial implications with respect (2) Kwentus, G. K.; Suda, M. U.S. Patent 4501907.
to downstream purification equipment. In addition, (3) Parkinson, G.; D’Aquino, R. O2 Breathes New Line Into
carbon adsorption may become problematic for processes Processes. Chem. Eng. 1999, 28-31.
that rely on shuttling the catalyst from oxidizing (4) Hutchings, G. J. Effect of promoters and reactant concen-
environments. The design of strippers is critical to tration on the selective oxidation of n-butane to maleic anhydride
minimize carbon carryover to the regeneration step. using vanadium phosphorus oxide catalysts. Appl. Catal. 1991,
72, 1-32.
More work is required both to establish the relationship (5) Mota, S.; Abon, M.; Volta, J. C.; Dalmon, J. A. Selective
between CO and CO2 and to identify the other byproduct Oxidation of n-butane on a VPO Catalyst: Study under Fuel-Rich
acids detected. Conditions. J. Catal. 2000, 193, 308-318.
6742 Ind. Eng. Chem. Res., Vol. 42, No. 26, 2003

(6) Mallada, R. Estudio de la Oxidación Selective de Butano a (23) Cavani, F.; Centi, G.; Manenti, I.; Trifiro, F. Catalytic
Anhı́drido Maleico en Reactores de Membrana. Ph.D. Thesis, Conversion of C4 Hydrocarbons on Vanadium-Phosphorus Ox-
University of Zaragoza, Zaragoza, Spain, 1999. ides: Factors Influencing the Selectivity of 1-Butene Oxidation.
(7) Alonso, M.; Lorences, M. J.; Pina, M. P.; Patience, G. S. Ind. Eng. Chem. Prod. Res. Dev. 1985, 24, 221-226.
Butane partial oxidation in an externally fluidized bed-membrane (24) Centi, G.; Golinelli, G.; Trifiro, F. Nature of the active sites
reactor. Catal. Today 2001, 67, 151-157. of (VO)2P2O7 in the Selective Oxidation of n-Butane. Evidence from
(8) Huang, X. F.; Chen, B. H.; Liu, B. J.; Silveston, P. L.; Li, C. doping experiments. Appl. Catal. 1989, 48, 13-24.
Y. Reoxidation kinetics of a VPO catalyst. Catal. Today 2002, 74, (25) Centi, G. Vanadyl PyrophosphatesA Critical Overview.
121-130. Catal. Today 1993, 16, 1-4.
(9) Wang, D.; Barteau, M. A. Oxidation kinetics of partially
reduced vanadyl pyrophosphate catalyst. Appl. Catal. A 2002, 223, (26) Hutchings, G. J.; Burrows, A.; Sajip, S.; Kiely, C. J.; Bere,
205-214. K. E.; Volta, J.; Tuel, A.: Abon, M. Understanding the microstruc-
(10) Wang, D.; Barteau, M. A. Kinetics of Butane Oxidation tural transformation mechanism which takes place during the
by a Vanadyl Pyrophosphate Catalyst. J. Catal. 2001, 197, 17- activation of vanadium phosphorus oxide catalysts. In 3rd World
25. Congress on Oxidation Catalysis; Grasselli, R. K., Oyama, S. T.,
(11) Mills, P. L.; Randall, H. T.; McCracken, J. S. Redox kinetics Gaffney, A. M., Lyons, J. E., Eds.; Elsevier Science BV: Amster-
of VOPO4 with butane and oxygen using the TAP reactor system. dam, The Netherlands, 1997.
Chem. Eng. Sci. 1999, 54, 3709-3721. (27) Sharma, R. K.; Cresswell, D. L. Kinetics and fixed bed
(12) Mallada, R.; Sajip, S.; Kiely, C. J.; Menéndez, M.; San- reactor modelling of butane oxidation to maleic anhydride. AIChE
tamarı́a, J. Influence of the Reaction Atmosphere on the Charac- J. 1991, 37 (1), 39-47.
teristics and Performance of VPO catalyst. J. Catal. 2000, 196, (28) Bej, S. K.; Rao, M. S. Selective Oxidation of n-Butane to
1-7. Maleic Anhydride. 1. Optimization studies. Ind. Eng. Chem. Res.
(13) Bergna, H. E. (assigned to E. I. du Pont de Nemours & 1991, 30, 1819-1824.
Co.). U.S. Patent 4,679,477, June 30, 1987. (29) Bej, S. K.: Rao, M. S. Selective Oxidation of n-Butane to
(14) Lorences, M. J.; Patience, G. S.; Dı́ez, F. V.; Coca, J. Fines Maleic Anhydride. 2. Identification of rate expression for the
effects on collapsing fluidized beds. Powder Technol. 2003, 131, reaction. Ind. Eng. Chem. Res. 1991, 30, 1824-1828.
234-240.
(30) Bej, S. K.; Rao, M. S. Selective Oxidation of n-Butane to
(15) Buchanan, J. S.; Sundaresan, S. Kinetics and Redox
Maleic Anhydride. 3. Modelling studies. Ind. Eng. Chem. Res.
properties of Vanadium Phosphate catalysts for Butane oxidation.
1991, 30, 1829-1832.
Appl. Catal. 1985, 19, 65-75.
(16) Centi, G.; Fornasari, G.; Trifiro, F. n-Butane Oxidation to (31) Escardino, A.; Sola, C.; Ruiz, F. Oxidación Catalı́tica de
Maleic Anhydride on vanadium-phosphorous oxides: Kinetic butano a anhı́drido maleico. I. Mecanismo de reacción. An. Quim.
analysis with a tubular stacked-pellet reactor. Ind. Eng. Chem. 1973, 69, 385-396.
Prod. Res. Dev. 1985, 24, 32-37. (32) Escardino, A.; Sola, C.; Ruiz, F. Oxidación Catalı́tica de
(17) Cavani, F.; Centi, G.; Trifiro, F. Study of n-Butane Oxida- butano a anhı́drido maleico. II. Estudio cinético del proceso (para
tion to Maleic Anhydride in a tubular flow stacked-pellet reactor. presiones parciales de hidrocarburo en la alimentación inferiors
Influence of Phosphorus on the selectivity. Appl. Catal. 1985, 15 a 7.5 mmHg). An. Quim. 1973, 69, 1157-1169.
(1), 151-160. (33) Schneider, P.; Emig, G.; Hofmann, H. Kinetic Investigation
(18) Hodnett, B. K. Vanadium-Phosphorous Oxide Catalysts and Reactor Simulation for the Catalytic Gas-phase Oxidation of
for the Selective Oxidation of C4 Hydrocarbons to Maleic Anhy- n-Butane to Maleic Anhydride. Ind. Eng. Chem. Res. 1987, 26,
dride. Catal. Rev. 1985, 27 (3), 373-424. 2236-2241.
(19) Cavani, F.; Centi, G.; Manenti, I.; Riva, A.; Trifiro, F. (34) Lerou, J. J.; Mills, P. L. DuPont Butane Oxidation Process.
Proceedings of the 187th National Meeting of the American Precis. Process Technol. 1993, 175-195.
Chemical Society. E. V. Murphee Award Symposium, St. Louis,
MO, Apr 1984. (35) Chen, L.; Hu, G. Applications of a Statistical Theory in
(20) Wohlfahrt, K.; Hofmann, H. Kinetics of the synthesis of Residence Time Distributions. AIChE J. 1993, 39 (9), 1558-1561.
Maleic Anhydride from n-Butane. Chem.-Ing.-Tech. 1980, 52, 811- (36) Lorences, M. J. Oxidación Selectiva de n-Butano to An-
814. hı́drido Maleico en un reactor de lecho fluidizado. Ph.D. Thesis,
(21) Centi, G.; Fornasari, G.; Trifiro, F. On the Mechanism of University of Oviedo, Oviedo, Spain, 2000.
n-Butane Oxidation to Maleic Anhydride: Oxidation in Oxygen-
Stoichiometry-Controlled Conditions. J. Catal. 1984, 89, 44-51. Received for review April 7, 2003
(22) Centi, G.; Trifiro, F.; Ebner, J. R.; Franchetti, V. Mecha- Revised manuscript received September 22, 2003
nistic Aspects of Maleic Anhydride Synthesis from C4 Hydrocar- Accepted September 25, 2003
bons over Phosphorous Vanadium Oxide. Chem. Rev. 1988, 88,
55-80. IE0302948

You might also like