Kinetics of citraconic anhydride formation via condensation of formaldehyde and succinates

Article abstract of DOI:10.1021/op020210r

Formation of citraconic anhydride via condensation of succinic acid and its derivatives with formaldehyde is carried out over γ-alumina catalyst in a continuous fixed-bed reactor. Dimethyl succinate and Formalin (37 wt % formaldehyde, 10 wt % methanol in water) are the preferred feed materials for the reaction; catalyst activity is sustained with Formalin relative to that with other formaldehyde sources such as trioxane or Formcel, because the water in Formalin inhibits coke formation. With this feed combination, a total citraconate yield of 31% of theoretical with 72% selectivity is achieved at a weight hour space velocity of 0.9 kg of succinate/kg of catalyst/h, a succinate to formaldehyde molar feed ratio of 1:2, and a temperature of 380 °C. The reaction is free from mass transfer limitations at these conditions. A kinetic model is presented that describes product distributions and reactant conversion as a function of space velocity and temperature. The reaction system is part of an overall process to produce itaconic acid from renewable resource-based succinic acid.

Full text of DOI:10.1021/op020210r

Organic Process Research & Development 2002, 6, 611−617
Kinetics of Citraconic Anhydride Formation via Condensation of Formaldehyde
and Succinates
Dushyant Shekhawat, Kirthivasan Nagarajan,† James E. Jackson,† and Dennis J. Miller*
Department of Chemical Engineering, Michigan State UniVersity, East Lansing, Michigan 48824, U.S.A.
Scheme 1
Abstract:
Formation of citraconic anhydride via condensation of succinic
acid and its derivatives with formaldehyde is carried out over
γ-alumina catalyst in a continuous fixed-bed reactor. Dimethyl
succinate and Formalin (37 wt % formaldehyde, 10 wt %
methanol in water) are the preferred feed materials for the
reaction; catalyst activity is sustained with Formalin relative
to that with other formaldehyde sources such as trioxane or
Formcel, because the water in Formalin inhibits coke formation.
With this feed combination, a total citraconate yield of 31% of
theoretical with 72% selectivity is achieved at a weight hour
space velocity of 0.9 kg of succinate/kg of catalyst/h, a succinate
to formaldehyde molar feed ratio of 1:2, and a temperature of
380 °C. The reaction is free from mass transfer limitations at
these conditions. A kinetic model is presented that describes
product distributions and reactant conversion as a function of
space velocity and temperature. The reaction system is part of
an overall process to produce itaconic acid from renewable
resource-based succinic acid.
technology matures, making it attractive as a feedstock for
biomass-based chemical production.
The Stobbe condensation,7 discovered in 1894, offers a
potentially efficient route for the production of substituted
succinates via condensation of esters of succinic acid with
aldehydes and ketones. The reaction is essentially unique to
succinic acid esters and generally takes place in alcohol
solution in the presence of a strong base (alkoxide or sodium
hydride) to give the half-ester as a product.8
The classic Stobbe condensation does not occur to any
practical extent when formaldehyde is used as the aldehyde.
However, the vapor-phase catalytic (heterogeneous) con-
densation of succinic acid, succinic anhydride, or alkyl
succinate esters with formaldehyde does take place to give
citraconic anhydride (CAN), an isomer of itaconic anhydride
(Scheme 1). Several patents describe catalytic routes to CAN
formation from succinates,9-12 with transient yields as high
as 70% of theoretical claimed.9 The motivation for CAN
formation is to produce itaconic acid via hydrolysis of CAN
to citraconic acid and isomerization to itaconic acid. Itaconic
acid is a carboxylated analogue of the important monomer
methacrylic acid and as such is able to take part in addition
polymerization, giving polymers with many free carboxylic
acid groups that confer advantageous wettability and ion-
exchange properties.13
Itaconic acid is currently produced commercially (8 ×
106 kg/yr) by the fermentation of glucose using Aspergillus
terreus.13 This fungal fermentation is carried out in batch
processes requiring dilute solutions (∼10 wt % glucose) and
extended processing times (6-10 days per batch). Itaconic
acid yields are on the order of 50-60% of theoretical; this
process results in an itaconic acid selling price of ∼$2/lb
I. Introduction
Succinic acid (1,4-butanedioic acid) and its alkyl esters
are reactive species that find applications in industrial and
consumer products and as intermediates for specialty and
fine chemicals production.1,2 Recent advances in fermentation
technologies for succinic acid production,3-6 arising both
from genetic modification of the microorganism and im-
proved separations, have led to yields of succinic acid as
high as 1.1 kg of succinic acid/kg of glucose (with CO2
incorporation) and have nearly eliminated acetic acid as a
coproduct.6 A significant decrease in the price of the acid
as a raw material is thus expected as the manufacturing
* To whom correspondence should be addressed. Department of Chemical
Engineering, 2527 Engineering Building, Michigan State University, East
Lansing, MI 48824. Telephone: (517) 353-3928. Fax: (517) 432-1105. E-mail:
millerd@egr.msu.edu.
† Department of Chemistry, Michigan State University, East Lansing, MI
48824.
(1) Fumagalli, C.; Spa, L. In Kirk-Othmer Encyclopedia of Chemical Technol-
ogy, 4th ed; Kroschwitz, J. I., Howe-Grant, M., Eds.; John Wiley and
Sons: New York, 1997; 22, 1074-1102.
(2) Szmant, H. H. Organic Building Blocks of the Chemical Industry; John
Wiley and Sons: New York, 1989.
(7) Stobbe, H. Ber. Dtsch. Chem. Ges. 1893, 26, 2312.
(8) Johnson, W. S.; Daub, D. B. In Organic Reactions; Adams, R., Ed.; John
Wiley and Sons: New York, 1951; VI, 2-73.
(9) Tate, B.; Berg. R. (Pfizer). U.S. Patent 3,835,162, 1974.
(10) Fujii, C.; Kibayashi, I.; Hasegawa, Y. (denki Kagaku Kogyo K. K.). Japanese
Patent 50-049223, 1975.
(3) Guettler, M. V.; Jain, M. K.; Soni, B. K. (Michigan Biotechnology Institute)-
. U.S. Patent 5,504,004, 1994.
(4) Datta, R.; Glassner, D. A.; Jain, M. K.; Vick Roy, J. R. (Michigan
Biotechnology Institute). U.S. Patent 5,168,055, 1992.
(5) Tsao, G. T.; Cao, N. J.; Gong, C. S. AdV. Biochem. Eng. Biotechnol. 1999,
65, 243-280.
(6) Nghiem, N. P.; Donnelly, M.; Millard, C. S.; Stols, L. (Lockheed Martin
Energy Research Corporation). U.S. Patent 5,869,301, 1999.
(11) Fujii, C.; Shimizui, T. (denki Kagaku Kogyo K. K.). Japanese Patent 49-
101326, 1974.
(12) Fujii, C.; Shimizui, T. (denki Kagaku Kogyo K. K.). Japanese Patent 49-
101327, 1974.
(13) Luskin, L. S. In Functional Monomers; Yocum, R. H., Nyquist, E. B., Eds.;
Marcel Dekker: New York, 1974; Vol. 2, pp 465-500.
10.1021/op020210r CCC: $22.00 © 2002 American Chemical Society
Published on Web 06/08/2002
Vol. 6, No. 5, 2002 / Organic Process Research & Development
•
611

and is not expected to facilitate expanded production in the
foreseeable future.
(DMS) as feeds, a single pump was used with the two
reactants mixed in a single 100-mL buret serving as a feed
reservoir. For TO:DMS molar ratios above 1:1, the feed
buret, pump, and feed line were heated to 70 °C using heating
tapes and a variable autotransformer to ensure that a single
liquid phase was present. For DMS and Formalin as feeds,
two HPLC pumps were used to deliver the feeds to the
reactor. The use of succinic anhydride (SAN) and TO as
feeds required the use of a syringe pump and steam-traced
feed lines maintained at 120 °C, because both materials are
solids at room temperature. For all feeds, the 60 cm of feed
tubing just before the reactor inlet was heated to 250 °C
using electrical heating tape. To aid in vaporization of feed
and to sweep the vaporized feed into the reactor, helium
(AGA, 99.99%) was introduced to the reactor through a
separate1/8 in. tube, heat-traced to 250 °C.
In an earlier publication we identified several intermediate
surface area γ-aluminas as attractive catalysts for the
formation of CAN from succinates and formaldehyde.14 In
this contribution, process conditions for the reaction are
evaluated. First, several sources of succinates and formal-
dehyde in different combinations are investigated as potential
feedstocks for CAN production. Then, reactor temperature,
pressure, weight hourly space velocity (WHSV), feed
composition, and catalyst particle size are evaluated to
optimize the yield of CAN and minimize catalyst deactivation
with time-on-stream. On the basis of the results of these
parametric studies, a kinetic model is developed to predict
reactant conversions and product yields versus WHSV and
temperature.
Reactor effluent passed through heat-traced tubing and
was directed via a heated six-port valve (Valco, Inc.) to one
of two 25-mL stainless steel product collection traps. Traps
were kept in ice water if the feed was DMS and Formalin,
to collect methanol and formaldehyde, and in warm water
(40 °C) for DMS/TO feeds to prevent solidification of
unreacted TO.
High-performance liquid chromatography (HPLC) and gas
chromatography (GC) were used to analyze reaction prod-
ucts. For liquid chromatography, 20 wt % acetonitrile in 5
mM H2SO4 aqueous solution was used as the mobile phase
at a flow rate of 0.4 mL/min. Raw product samples were
diluted 20-fold with mobile phase and injected onto a 15
cm × 0.5 cm i.d. BioRad HPX-87H column at 40 °C;
products concentrations were determined using a refractive
index detector with oxalic acid as the internal standard. Gas
chromatography was performed on a Varian 3300 using a
large-bore capillary column (SPB1, Supelco, 0.53 mm i.d.
× 30 m) with flame ionization detector (FID) and helium as
a carrier gas. Methyl lactate was used as an internal standard.
Outlet gases from the reactor were analyzed directly using
CO and CO2 IR meters (Riken, Inc.).
II. Experimental Section
Feed and Catalyst Materials. Two forms of succinate
were used as feedstocks in these investigations: dimethyl
succinate (DMS) (bp 203 °C; Aldrich Chemical Co, 98%)
and succinic anhydride (SAN) (mp 118 °C, bp 269 °C;
Aldrich, 98%). Formaldehyde was used in one of three
forms: trioxane (TO) (mp 64 °C, bp 115 °C; Aldrich, 98%),
the trimeric form of formaldehyde; Formalin (JT Baker), a
commercially available source containing 37 wt % formal-
dehyde and 10 wt % methanol in aqueous solution, and
Formcel (Celanese), another commercial source composed
of 55 wt % formaldehyde, 35 wt % methanol, and 10 wt %
water.
The catalyst used in this study was a γ-alumina (Norpro
SA3177). This material has a N2 BET surface area of 100
m2/g, acidic site density of 0.25 mmol/g, and basic site
density of 0.05 mmol/g as determined by temperature-
programmed desorption of ammonia and carbon dioxide,
respectively, and a Hammett acid strength constant Ho )
-0.2 to +1.1. The alumina was ground and sieved to 30-
60 mesh using standard sieve trays and calcined in air for 6
h at 500 °C before loading into the reactor.
Product Hydrolysis. The reactor effluent contains signifi-
cant quantities of dimethyl and monomethyl esters of
citraconic and succinic acids as well as the free acid and
anhydride forms. In HPLC analyses, the coelution of
dimethyl and monomethyl citraconate esters with their
analogue succinates made accurate conversion and yield
determinations impossible. To overcome this obstacle, the
product mixture was hydrolyzed by adding a small amount
of sulfuric acid to the product solution and refluxing for 2-3
h to recover all succinate and citraconate species as free acids.
Because hydrolysis was time-consuming, only selected
samples were hydrolyzed; unless otherwise stated, yields and
conversions are reported for unhydrolyzed products. Typi-
cally, about 20% of total citraconates formed were present
as monomethyl or dimethyl esters; therefore, reported yields
for unhydrolyzed mixtures are typically lower than the actual
values by this amount.
Apparatus and Experimental Conditions. The fixed-
bed reactor used in this study was described in detail in a
previous publication.14 Briefly, the reactor consists of a 10-
mL 316 stainless steel cone closure pressure vessel (Auto-
clave Engineers) surrounded by a clamshell heater. Reactor
temperature was controlled by a programmable temperature
controller with the control thermocouple reading the tem-
perature of the external reactor surface. The reactor was
equipped with a rupture disk to prevent catastrophic rupture
in case of plugging of sample lines (which can happen in
the presence of succinic anhydride and paraformaldehyde,
both solids at room temperature). Typical catalyst charge to
the reactor was 5.0 ( 0.1 g.
Feed materials were introduced into the reactor, using
HPLC pumps (Bio-Rad, Inc.) or a syringe pump (PDC, Inc.).
The choice of feed species dictated the configuration of the
feed system. For trioxane (TO) and dimethyl succinate
Product Yield and SelectiVity Calculations. Fractional
conversion is reported either in terms of dimethyl succinate
or succinic anhydride converted to any product including
(14) Shekhawat, D.; Kirthivasan, N.; Jackson, J. E.; Miller, D. J. Appl. Catal.,
A 2001, 223, 261-273.
612
•
Vol. 6, No. 5, 2002 / Organic Process Research & Development

Table 1. Reaction conditions
are summarized in Table 2. All reactions were conducted
with a succinate:formaldehyde molar ratio of 1:2; succinate
mole fraction for each feed is given in Table 2. The
conversion and yield reported are at the time of maximum
citraconic anhydride yield, which occurs between 1 and 2 h
after steady state is achieved in the reactor.
All combinations of succinate feeds and formaldehyde
gave significant yields of citraconic anhydride at base-case
conditions, suggesting that the C1 and C4 carbons of succinate
do not participate in the condensation. The difference in
performance between the various feed combinations resides
primarily in the extent of CO2 formation (cracking) and coke
deposition that results in catalyst deactivation.
With succinic anhydride (SAN) and trioxane, which were
fed in molten form with no solvent added, extensive coking
occurred, and deactivation was very rapid. Citraconic acid
yield reached a maximum of 44% after 1-1.5 h, and then
rapidly declined to near zero by 5 h of reaction. Large
quantities of carbon dioxide were produced. We attribute this
behavior in part to the high boiling point of SAN (269 °C)
and its high concentration in the feed, which made complete
vaporization of the feed difficult. Reactor operation was also
difficult with this feed combination, as SAN condensed and
plugged the reactor at any location that was not adequately
heat traced.
With DMS and trioxane (TO), the first hour of reaction
was dominated by CO2 formation. Citraconic acid yield
reached a maximum of 35% after 2 h and then fell off to
approximately 22% after 5 h reaction time.14 Similar behavior
was observed with the DMS/Formcel feed combination;
initial yields and DMS conversion were higher than with
DMS/TO feed but declined more rapidly.
base-case
value
parameter
range
reactor temperature (°C)
reactor pressure (MPa)
feed zone preheat temperature (°C)
Formaldehyde to succinate molar ratio
liquid feed flow rate (mL/min)
helium flow rate (mL(STP)/min)
350-410
0.5-3.5
200-250
0.5-4
0.10-0.45
27-82
380
0.5
250
2
0.15
27
WHSV (kg of succinate/kg of catalyst/h) 0.45-2.70
0.90
other succinates, or as the fraction of all succinate species
converted to nonsuccinate products. Product yield is defined
as the percentage of theoretical based on dimethyl succinate
or succinic anhydride fed to the reactor; product selectivity
is calculated as product yield divided by the conversion of
all succinate species to nonsuccinate products. These defini-
tions are appropriate because succinates in any form can react
or be recovered, reconverted to feed material, and recycled
back to the process. Carbon dioxide yields are based on
succinate fed but are divided by two for DMS feed to account
for the fact that each mole of DMS can give 2 mol of CO2
upon cracking. A molar balance on total succinate (C4)
carbon was done for each sample to assess reliability of the
experiments; fractional recovery of the succinate carbon fed
is reported as the measure of material balance closure for
the experiment.
III. Results and Discussion
Most experiments were carried out for 5 h at steady state,
with samples taken in 30-min intervals. A summary of
reaction conditions is given in Table 1; the “base-case”
conditions were used as a starting point in most evaluations.
In our earlier publication,14 we showed that catalyst activity
declines with time-on-stream for this reaction system because
of coking. Thus, reactant conversion and product yields are
reported as a function of reaction time or as the maximum
observed in the experiment. We have shown14 that catalyst
activity can be entirely recovered by burning in air at 500
°C, so decline in catalyst activity with time-on-stream is not
a barrier to development.
The use of Formalin as the formaldehyde source with
DMS gave strikingly different behavior: citraconate yields
and succinate conversion reached constant values after 30
min of reaction and then continued essentially unchanged
over 5 h of reaction. The yield of CO2 was low throughout
the reaction period as well. Similar results were obtained
with succinic anhydride and trioxane dissolved in excess
methanol (to form a monomethyl succinate/trioxane feed,
MMS/TO in Table 2). Addition of methanol or water as a
solvent greatly reduced catalyst coking and led to citraconic
Evaluation of Feedstocks. Succinate conversion and
product yields for different succinate/formaldehyde feed
combinations over SA3177 alumina at base-case conditions
Table 2. Summary of results from different feedstocks over SA-3177 alumina
before hydrolysis
after hydrolysis
yield of conv
feedstock
conv.
yield
yield of yield of
yield
yield
carbon
(feed mol fraction) succinates of CAN
DMS
MMS
of SAN of CO2 recovery citraconates succinates selectivity
DMS + TO
83
76
74
80
81
26
43
23
26
30
-
19
8
-
9
18
16
15
8
92
90
99
90
89
35
44
26
31
34
48
67
40
43
56
73
70
78
72
61
(0.22) (0.44)
SAN + TO
0
7
(0.24) (0.48)
MMS + TO
(0.15) (0.29)
DMS + Formalin
(0.1) (0.19)
DMS + Formcel
(0.15) (0.30)
30
-
-
28
8
-
18
8
10
Vol. 6, No. 5, 2002 / Organic Process Research & Development
•
613

anhydride yield that was nearly constant with reaction time
out to 5 h. The presence of a solvent appeared to facilitate
vaporization and dispersion of the feed materials, thus
minimizing the cracking and polymerization reactions that
led to coke formation on the alumina surface.
The extent of hydrolysis or esterification of the succinate
feedstock used depended greatly on the quantity of methanol
or water in the reaction stream. For succinic anhydride/
trioxane feed, the extent of esterification was minimal,
because there is little methanol present. The only esterifi-
cation product formed was monomethyl succinate, formed
from methanol liberated via the Cannizzaro reaction, which
in turn can take place as water is evolved in citraconic
anhydride formation. In contrast, with Formalin as a feed
significant hydrolysis of DMS to MMS and free succinic
acid took place in the reactor. Fortunately, since all forms
of succinate are active for reaction, interconversion between
diester, monoester, and anhydride forms did not apparently
affect overall performance for CAN formation.
Figure 1. Effect of temperature on the yield of CAN [350 °C
(b); 380 °C (9); 410 °C (2)] and conversion of DMS [350 °C
(O); 380 °C (0); 410 °C (4)] (Feed: DMS + TO, base-case
conditions except temperature, unhydrolyzed product).
The use of DMS/Formalin or MMS/TO is thus preferred
for the formation of citraconic anhydride. Both of these feed
combinations provide good yield and selectivity to the desired
product with relatively little loss of succinate or formalde-
hyde to cracking. Coke deposition still takes place with these
feed materials, but the rate of coke deposition is much slower
than with the other feeds, and product yields are nearly
unaffected over the 5-h time period of reaction. The only
drawback to the use of succinic anhydride in methanol is
extensive methanol dehydration to dimethyl ether over the
acidic alumina; recovery of the dimethyl ether and hydrolysis
back to methanol would require expensive process equipment
because of its very low boiling point. This, coupled with
the low cost and ready availability of Formalin, makes the
DMS/Formalin feed combination the preferred choice for
citraconic anhydride formation.
Figure 2. Effect of feed flow rate on the yield of CAN [WHSV
) 0.9 h-1 (b); WHSV ) 1.8 h-1 (9); WHSV ) 2.7 h-1 (2)]
and conversion of DMS [WHSV ) 0.9 h-1 (O); WHSV ) 1.8
h-1 (0); WHSV ) 2.7 h-1 (4)] (Feed: DMS + Formcel, base-
case conditions except WHSV, unhydrolyzed product).
Reaction Conditions. The effects of reaction temperature,
pressure, WHSV, feed composition, and catalyst particle size
have been investigated for the purpose of identifying
optimum conditions for CAN formation.
The conversion of DMS and yield of citraconates were not
significantly affected by reactor pressure. This result is
somewhat surprising and as yet unexplained, particularly in
light of the kinetic model described later.
Temperature. The effect of temperature on CAN yield
and DMS conversion is given Figure 1 for DMS/TO feed.
Conversion of DMS increases with increasing temperature,
but selectivity decreases with temperature. A maximum in
the CAN yield is thus observed at 380 °C in the laboratory
reactor. At higher temperatures, cracking predominates,
resulting in high conversion of succinates and low selectivity
to CAN. At lower temperatures, selectivity to CAN is very
high, but conversion is low. Similar results were achieved
with DMS/Formalin feed (not shown), with selectivity to
CAN at 350 °C as high as 87% (versus 70% at 380 °C).
The ultimate choice of reactor temperature for a process will
depend on economic constraints: operation at low temper-
ature and low space velocity may prove most attractive from
an economic standpoint.
Space Velocity. The condensation reaction of DMS and
Formcel (at base-case conditions) was carried out at dif-
ferent liquid feed flow rates (9 mL/h to 27 mL/h) over a 5-g
catalyst bed to give WHSVs ranging from 0.90 to 2.70 kg
of DMS/kg of catalyst/h. The flow rate of helium was
changed in the same proportion as liquid feed flow rate to
keep the molar composition of the feed constant. The effect
of WHSV (kg of succinate/kg of catalyst/h) on CAN yield
and DMS conversion from Formcel and DMS feed is
reported in Figure 2.
The conversion of DMS at the lowest liquid feed rate
(WHSV ) 0.9/h) was always ∼10% higher than that at
WHSV ) 1.8/h. DMS conversion declines further, but by a
lesser amount, upon increasing the liquid feed flow to give
WHSV ) 2.7/h. Yields of CAN and CO2 at different liquid
feed flow rates followed the DMS conversion trends. Yields
of MMS and SAN were not affected by changing space
Pressure. Most experiments were run at a reactor pressure
of 0.5 MPa, but repeated experiments were conducted at 2.7
MPa to check the dependence of yield and selectivity on
pressure (DMS/TO feed at otherwise base-case conditions).
614
•
Vol. 6, No. 5, 2002 / Organic Process Research & Development

Figure 3. Effect of WHSV (kg of succinate/kg of catalyst/h)
on the yield of CAN [shorter reactor bed (WHSV ) 0.90 h-1
)
Figure 4. Effect of feed molar ratio (formaldehyde to succi-
nate) on the yield of CAN [4:1 (b); 2:1 (9); 1:1 (2); 0.5:1
([)] and conversion of DMS [4:1 (O); 2:1 (0); 1:1 (4); 0.5:1
(])] (Feed: DMS + Formcel, unhydrolyzed product, base-case
conditions except WHSV ) 2.1 kg of succinate/kg of catalyst/
h).
(b); longer reactor bed (WHSV ) 0.45 h-1) (9); longer reactor
bed (WHSV ) 0.90 h-1) (2)] and conversion of DMS [shorter
reactor bed (WHSV ) 0.90 h-1) (O); longer reactor bed (WHSV
) 0.45 h-1) (0); longer reactor bed (WHSV ) 0.90 h-1) (4)]
(Feed: DMS + Formlin, base-case conditions except WHSV,
unhydrolyzed product).
slightly with increasing feed molar ratio. As with tempera-
ture, the final choice of feed molar ratio resides in process
economics; the higher yields achieved with greater excess
of formaldehyde will be offset by increased costs of
separating and recycling unreacted formaldehyde.
It is noteworthy that the extent of coking decreased at
lower formaldehyde levels. This suggests that part of coke
formation during reaction comes from formaldehyde, or that
added water and methanol are responsible for reduced coking
by aiding in vaporization of organic species or gasifying the
coke (with steam) as it forms.
Particle Size. The typical catalyst particle size of 30-60
mesh (dp ) 0.3-0.5 mm) was reduced to 60-100 mesh (dp
) 0.16-0.3 mm) to ascertain the effects of mass transport
on the reaction. Identical results were observed with both
particle sizes, suggesting that the condensation reaction is
not mass-transfer limited.
velocity, suggesting that the esterification/hydrolysis reac-
tions reach equilibrium in the reactor. The selectivity to CAN
was not affected by changing WHSV.
Additional evaluation of the effect of space velocity on
product yields were conducted with DMS and Formalin,
using a catalyst bed twice the usual size (10 g) in several
experiments. Results are given in Figure 3; increasing bed
length at the same liquid rate increases yield of CAN and
CO2 and conversion of DMS. Maintaining the same WHSV
in the longer bed by doubling flow rate results in identical
conversion of DMS, which is expected if the reaction is
behaving predictably. However, the yield of CAN decreased
slightly with the larger bed (2 vs b in Figure 3). Whether
this difference lies within the range of experimental uncer-
tainty or the reactor is exhibiting some nonideal behavior
cannot be ascertained at this time.
Evaluation of the observable modulus ηφ2 ) RobsL2/CoDe
further verified the absence of mass-transfer limitations. In
the modulus, Co is the inlet DMS concentration, De ) ꢀ2DDMS
) 9.1 × 10-6 cm2/s is the effective diffusivity calculated
Feed Molar Ratio. Several different molar ratios of
formaldehyde to DMS (4:1, 2:1, 1:1, and 0.5:1) were
evaluated using DMS/Formcel feed. Reaction conditions
were same as the base case except that a liquid feed flow
rate of 0.30 mL/min and a gas flow rate of 55 mL(STP)/
min was used. The inlet mole fraction of DMS, the liquid
flow rate, and hence WHSV (kg of DMS/kg of catalyst/h),
were kept the same for all feed molar ratios; this was done
by replacing formaldehyde with a mixture of methanol and
water (in the same proportion as in Formcel) as formaldehyde
concentration was reduced. The effect of feed molar ratio
on the yield of CAN and conversion of DMS is given in
Figure 4.
Yield of CAN increased as the ratio of formaldehyde to
DMS in the feed increased. A maximum CAN yield of 26%
at 66% DMS conversion was observed at a 4:1 molar ratio;
in contrast, for 0.5:1 only an 11% CAN yield was obtained.
Surprisingly, the conversion of DMS was essentially unaf-
fected by the quantity of formaldehyde present. Thus,
selectivity of CAN increased with increasing feed molar ratio
of formaldehyde to DMS. Yield of CO2 also increased
using ꢀ ) 0.4 and the Chapman-Enskog equation for DDMS
,
L ) dp/6, and Robs is the average consumption rate of DMS
over the reactor volume. The largest value of ηφ2 observed
in any reaction was less than 0.1; thus, according to the
Weisz-Prater criterion the reaction is not mass-transfer
limited.
CO2 as a Carrier Gas. Reaction of DMS and Formalin
over SA-3177 was carried out with CO2 as a carrier gas
instead of helium. There were two reasons behind the use
of CO2 as a carrier for the formation of CAN. First, it was
thought that coke deposited on the catalyst could be removed
via in situ oxidation by CO2. Second, it was thought that
CO2 might neutralize the basic sites on the alumina surface,
thus reducing the extent to which formaldehyde is consumed
by the Cannizzaro reaction. The experiment shows that CAN
yield, succinate conversion, and extent of coke deposition
were the same with CO2 or with helium as a carrier gas.
Vol. 6, No. 5, 2002 / Organic Process Research & Development
•
615

Table 3. Reactions of the kinetic model
reaction
description
reaction
constant
1
condensation of succinates
(DMS, MMS, SA) with
formaldehyde to form
CH3OOC(CH2)2COOCH3 + HCHO f C5H4O3 + 2CH3OH
CH3OOC(CH2)2COOH + HCHO f C5H4O3 + MeOH + H2O
HOOC(CH2)2COOH + HCHO f C5H4O3 + 2 H2O
k1
citraconic anhydride
2
3
4
5
citraconic anhydride cracking
dimethyl succinate hydrolysis
monomethyl succinate hydrolysis
succinate (DMS, MMS, SA)
decomposition
C5H4O3 f CO2 + coke + H2O
k2
CH3OOC(CH2)2COOCH3 + H2O ) CH3OOC(CH2)2COOH + MeOH
CH3OOC(CH2)2COOH + H2O ) HOOC(CH2)2COOH + MeOH
CH3OOC(CH2)2COOCH3 f 2 CO2 + coke
CH3OOC(CH2)2COOH f 2 CO2 + coke
HOOC(CH2)2COOH f 2 CO2 + coke
Ke3
Ke4
k5
6
7
Canizzaro reaction
formaldehyde decomposition
2HCHO + H2O f CH3OH + CO2 + H2
HCHO f CO + H2
k6
k7
dC
Kinetic Model. A steady-state kinetic model for the
formation of CAN from DMS and formaldehyde over
γ-alumina (SA3177) has been developed using simple nth
order kinetics in the model equations. Because of the
complexity of the reaction system, we do not believe that a
Langmuir-Hinshelwood kinetic model containing additional
adjustable parameters is warranted for this system. The
absence of mass-transport resistances facilitates calculation
of intrinsic kinetic rate constants. Thermodynamic analysis
of the reactions using experimental and predicted (via
molecular modeling) heats and free energies of formation
indicates that all reactions are thermodynamically favorable
enough to be considered irreversible.
The seven reactions included in the kinetic model are
listed in Table 3. Reaction 1 accounts for CAN production
from succinates and formaldehyde via second-order pathways
(k1). The second and fifth reactions describe first-order
cracking reactions of CAN (k2) and succinates (k5) into CO2.
The third and fourth reactions describe succinate hydrolysis
and esterification; these reactions are assumed to be at
equilibrium in the reactor.
CAN:
c ) k1CsuCf - k2Cc
(3)
dτ
dC
succinate:
formaldehyde:
MeOH:
water:
-
-
su ) k1CsuCf + k5Csu
(4)
(5)
dτ
dCf
dτ
) k1CsuCf + 2k6Cf2Cw + k7Cf
dCoh
dτ
) 2k1CdCf + k1CmCf + k6Cf2Cw (6)
) -k6Cf2Cw + k1CmCf + 2k1CsaCf (7)
dCw
dτ
dC
CO:
CO ) k7Cf
(8)
(9)
dτ
dC
CO2:
CO2 ) k5Csu + k6Cf2Cw + k2Cc
dτ
Where Cc, Cd, Cf, Cw, Cm, Coh, Csa, Csu, CCO, and CCO2
are steady-state vapor concentrations (kmol/m3) of CAN,
DMS, formaldehyde, water, MMS, methanol, SA, total
succinate, CO, and CO2, respectively, and τ is contact time
defined as catalyst volume/feed volumetric vapor flow rate
(m3 of catalyst/m3 s-1 of feed). For the 2:1 molar Formalin:
DMS feed mixture at 653 K and 0.5 MPa, WHSV is related
to τ via
CmCOH
Ke3
Ke4
)
)
(1)
(2)
CdCw
C
SACOH
CmCw
WHSV (kg of succinate/kg of catalyst/h) ) 2.97/τ (s) (10)
This assumption is justified by the high temperatures of
reaction and significant acidity of the alumina catalyst, both
of which facilitate hydrolysis/esterification, and because
outlet stream concentrations satisfy eqs 1 and 2 over a wide
range of conditions. The equilibrium constants, Ke3 and Ke4,
used in the model are average values calculated from reactor
outlet concentrations in all experiments.
Finally, reaction 6 describes the Canizzaro reaction of
formaldehyde (k6; second-order reaction with respect to
formaldehyde and first-order reaction with respect to water)
and reaction 7 describes formaldehyde decomposition to CO
and H2 (k7).
Control runs feeding DMS only and CAN only were
performed over γ-alumina (SA-3177) at base-case conditions
to obtain first-order rate constants for the cracking reactions.
We assume that CO2 is formed stoichiometrically from the
cracking reaction of succinates (2 mol CO2/mol succinate)
and CAN (1 mol CO2/mol CAN). Rate constants for the
cracking reactions (k2 and k5) were determined by solving
the first-order differential equations in DMS and CAN.
After determining and incorporating the rate constants k2
and k5 and the equilibrium constants Ke3 and Ke4 into the
model differential equations, the rate constants k1, k6, and k7
were fit by nonlinear regression of the experimental data
using Polymath (CACHE, Inc.) to minimize the sum of the
squares of differences in predicted and experimental values
of concentrations of all species except methanol and water
in the reaction system. The experimental concentrations were
The one-dimensional steady-state molar balances in an
integral, tubular reactor are given below for all species
involved in the reactor system, with the rate expressed on a
unit catalyst volume basis:
616
•
Vol. 6, No. 5, 2002 / Organic Process Research & Development

Table 4. Parameters of kinetic modela
Table 5. Comparison of predicted and experimental outlet
concentrations (10-3 kmol/m3) from reactor
rate constant
activation
energy
E (kJ/mol)
or equilibrium
constant
preexponential factor
τ ) 5.8 s
τ ) 3.6 s
τ ) 2.9 s
A
species exptl predicted exptl predicted exptl predicted
k1
k2
Ke3
Ke4
k5
k6
k7
(6.7 ( 0.8) × 105 m3kmol-1 s-1 67.0 ( 10.0
6.0 ( 1.0 s-1
30.6 ( 7.5
35.6 ( 9.2
84.2 ( 21.0
48.6 ( 26.0
CAN
DMS
HCHO 11.96
MMS
SA
MeOH
CO
1.96
2.49
1.92
2.69
13.37
2.94
1.19
14.71
1.25
1.33
2.51
10.95
2.48
1.29
9.86
0.70
0.58
1.13
2.61
14.21
2.88
1.17
12.28
0.78
1.19
2.54
14.62
2.64
1.04
10.24
0.44
1.09
2.95
15.19
3.24
1.31
12.99
0.67
(1.9 ( 0.4) × 102
(5.2 ( 0.9) × 105
(3.1 ( 1.0) × 101 s-1
3.14
0.92
9.44
1.10
1.01
(1.5 ( 0.5) × 103 m6 kmol-2 s-1 26.4 ( 10.5
(1.6 ( 0.4) × 105 s-1
88.4 ( 33.5
CO2
0.92
0.51
0.56
0.47
a k ) A exp(-E/RT) or Ke ) A exp(-E/RT).
3.6 × 10-3 kmol/m3 at τ ) 25 s; this maximum corresponds
to a yield of 41% with a selectivity of 64%. This represents
an initial estimate of the preferred condition for reactor
operation in an optimized process design.
taken as average values for each species exiting the reactor
over the last 2.5 h of reaction. This was done at 350
and 380 °C. The resulting activation energy and preexpo-
nential factor for each of the rate constants in the kinetic
model, along with uncertainty in each value, are presented
in Table 4.
IV. Summary
Several succinate and formaldehyde substrates have been
investigated for the formation of citraconic anhydride.
Dimethyl succinate and Formalin are an attractive feed
combination for the reaction, as reasonable yields of citra-
conate were obtained with very good selectivity and the
catalyst activity was stable for an extended time period. After
hydrolysis of raw products, a maximum 31% yield of
citraconates at 43% conversion of succinates was observed
from the reaction of DMS and Formalin over SA3177
alumina at the base-case conditions. Higher formaldehyde-
to-DMS molar ratios gave slightly better yield and selectivity
to CAN. Yield of CAN decreased as WHSV was increased,
either by changing bed length or liquid flow rate, but
selectivity remained unchanged. Yield of CAN increased
with increasing reaction temperature, but selectivity declined
due to more extensive succinate cracking at elevated tem-
peratures. DMS is not prone to cracking at elevated tem-
perature as is SAN and, hence, is a more favorable feed
material. The decay in catalyst activity was much slower
with Formalin than with other formaldehyde sources, likely
because the water vapor present cleans the catalyst during
the reaction.
The uncertainty associated with each rate constant in
Table 4 is based on the experimental standard deviation for
the exiting species concentrations. This uncertainty in species
concentration arises mainly from scatter in the chromato-
graphic data and from difficulty in recovering all material
in the sample collection process. The formaldehyde (as
paraformaldehyde), citraconic anhydride, and succinic an-
hydride, all major components of the reactor effluent stream,
solidify below 150 °C and thus are challenging to recover.
Nevertheless, succinate carbon balances over the course of
experiment (Table 2) were typically at 90+% of closure and
the overall total mass balances (including all species) were
always within (2% of closure.
Uncertainty in values of k2 and k5 was thus calculated
using the observed standard deviation of two percentage
points (e.g., 40 ( 2%) in succinate and citraconate conver-
sion in the independent cracking experiments. Uncertainties
in k1, Ke3, and Ke4 were also estimated on the basis of a
standard deviation for citraconate yield and succinate conver-
sion of (2 percentage points in condensation reactions. For
reactions 6 and 7, the uncertainty in CO and CO2 outlet
concentrations were taken as (5% of their absolute con-
centrations as measured by the IR meters in the effluent gas
line.
Predicted product stream concentrations are calculated by
applying the rate constants into eqs 1-9 above; predicted
values are compared with experimental concentrations at 380
°C in Table 5. In general, the model predicts outlet
concentrations for the key species quite well. The primary
exception is methanol (MeOH) concentration, which is
predicted rather poorly because it was not included in the
least-squares fit and because of formation of dimethyl ether
during reaction that is not accounted for in the kinetic model.
Extending the prediction of concentrations to higher values
of τ gives a broad maximum in CAN concentration of
The condensation reaction of DMS with Formalin over
intermediate surface area alumina is not mass-transfer
limited. The kinetic model provides insight into reaction
kinetics and provides direction for future scale-up and process
design and simulation.
Acknowledgment
The support of Applied Carbochemicals, Inc., and the
Crop Bioprocessing Center, State of Michigan Research
Excellence Fund, is greatly appreciated.
Received for review January 16, 2002.
OP020210R
Vol. 6, No. 5, 2002 / Organic Process Research & Development
•
617