Chemical Engineering Journal 173 (2011) 198–209

Contents lists available at ScienceDirect

Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej

A biphasic model describing soybean oil epoxidation with H2 O2 in a fed-batch

reactor
E. Santacesaria a,∗ , R. Tesser a , M. Di Serio a , R. Turco a , V. Russo a , D. Verde b
a
University of Naples “Federico II”, Department of Chemistry, Complesso Universitario Monte S. Angelo, Via Cintia 4, IT 80126 Naples, Italy
b
Mythen SpA – Zona Industriale Località Macchia, IT75013 Ferrandina (MT), Italy

a r t i c l e i n f o a b s t r a c t

Article history: In the present work, the kinetics, mass transfer and heat transfer of soybean oil epoxidation with H2 O2
Received 9 February 2011 have been studied in a fed and pulse-fed-batch reactor. The reaction has been performed with peroxy-
Received in revised form 28 April 2011 formic acid (PFA), generated in situ, by reacting concentrated hydrogen peroxide (60 wt.%) with formic
Accepted 6 May 2011
acid (FA), in the presence of sulphuric or phosphoric acid as catalysts. The kinetic study also considers
two important aspects occurring simultaneously with the epoxidation reaction, namely: the degradation
Keywords:
resulting from the opening of the oxirane rings, and the hydrogen peroxide decomposition. Epoxidation
Epoxidation
is a highly exothermic reaction and the evolution of the temperature in the reactor over a period of time
Soybean oil
Hydrogen Peroxide
is strongly dependent on the amounts and way in which a mixture of H2 O2 and formic acid is added to the
Kinetics mixture of oil and catalyst. In this paper, a biphasic kinetic model has been developed considering all of
Liquid-liquid reaction system the occurring reactions in each phase, the partition of reagents and products between the phases and the
evolution of any involved chemical specie along the time. Different kinetic runs have been successfully
simulated after the evaluation, by mathematical regression analysis or by independent means, of all the
kinetic and thermodynamic parameters of the model. The heat transfer properties of the used reactor
have been determined following different approaches. In addition, the evolution of the temperature of
the reacting mixture during the time has also been simulated with the developed mathematical model.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Epoxidized oils are important chemicals, normally used as

plasticizers and stabilizers for PVC resins. Nowadays, it is very
important to increase the epoxidized soybean oil (ESBO) produc-
tivity, because this substance is a good substitute of phthalates as
a plasticizer, since phthalates have been banned in many countries
for their negative effects on health [1,2]. The epoxidation of veg-
etable oils, for example soybean oil, is industrially carried out by
reacting the double bonds of the oil with a peroxyacid (generally
peroxyacetic or peroxyformic acid) generated in situ by reacting
concentrated hydrogen peroxide with acetic or formic acid in the (2)
presence of a mineral acid as a catalyst. The peroxyacid formation
This reaction is highly exothermic (H = −55 kcal/mol for each dou-
occurs in the aqueous phase according to the following reaction:
ble bond) and an excessive increase of the temperature in the
industrial reactors is prevented by adding a limited amount of a
H2 O2 + RCOOH ↔ RCOOOH + H2 O (1) mixture of H2 O2 and formic or acetic acid to the mixture of oil
and acid catalyst [3]. The reaction normally requires 8–10 h to be
completed, keeping the temperature between 60 and 75 ◦ C. Finally,
Subsequently, performic acid migrates into the oil phase and it must be pointed out that the epoxidation reaction, initially,
spontaneously reacts according to the following scheme: seems very fast and the temperature rise is high and steep, in the
subsequent additions of reactants the reaction rate is slower and
the temperature rise decreases. Therefore, an optimization and/or
eventual process intensification will require detailed information
∗ Corresponding author. about both the kinetics of the occurring reactions and the effects
E-mail address: elio.santacesaria@unina.it (E. Santacesaria). of heat and mass transfer. In the last decade, some papers have

1385-8947/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.cej.2011.05.018
 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209 199

that, according to our experimental observations, is enhanced by

Nomenclature the stainless steel surface of the industrial reactors [10]. In the
present work, the kinetics of the soybean oil epoxidation by per-
C C unsaturated group (double bond) oxyformic acid (PFA), generated in situ, by reacting concentrated
FA formic acid H2 O2 (60 wt.%) with formic acid, in the presence of sulphuric or
PFA performic acid phosphoric acid as catalysts, has been studied in a fed-batch reac-
Epox epoxide group tor. In particular, in this kinetic study we have considered: (i) the
DEGi decomposed oxirane group by: i = H2 O2 , FA (formic partition between the phases of reagents and products; (ii) the
acid), PFA (performic acid), H2 O, H (proton) kinetics of the reactions occurring in the two different phases; (iii)
ri reaction rate [mol L−1 min−1 ] the degradation of the oxirane rings with related reactions; (iv) the
ki kinetic constant (the units depends on the reaction) hydrogen peroxide decomposition; (v) any effect of mass transfer
JJi mass transfer rate for the component J in the phase in limiting the reaction rates; (vi) the three occurring thermal con-
I [mol L−1 min−1 ] tributions related, respectively to the heat released by reactions,
ˇJi product between the mass transfer coefficient and the heat removed by exchange with the thermostatting fluid and
the interfacial surface area for the component J in the heat dispersed with the external environment. As it is difficult
the phase i to maintain isothermal conditions with a so extreme exothermic
aq∗ org∗
HJ partition coefficients defined as [C]j /[C]j reaction, kinetic runs have been performed in thermal dynamic
ai Specific interfacial area for the phase I [dm2 /dm3 ] conditions. Some kinetic runs have been made in pulse-fed-batch
S Absolute interfacial area for the liquid phase [dm2 ] conditions by adding the oxidant reagents in different subsequent
U global thermal exchange coefficient steps, other have been performed in fed-batch conditions by adding
[cal m−2 min−1 ◦ C−1 ] a continuous flow rate of the oxidant reagent. Knowing the amount
A exchange area [m2 ] of heat generated by the reaction, the heat exchanged with the re-
ni moles of the ith compound [mol] circulating fluid and the heat dispersed, it is possible to relate the
Fi molar flow-rate of the compound i [mol min−1 ] thermal profile directly to the double bond conversion. An effort
Vi reaction volume referred to the phase i [L] has also been made to estimate the partition coefficients of reagents
cP,i specific heat capacity at constant pressure of the and products in the two reacting phases using the SPARC method
specie i [cal g−1 ◦ C−1 ] [11].
Ti temperature of the fluid i [◦ C] The main scopes of this work are: (i) to verify the kinetic laws of
PMi molecular weight of the ith compound [g mol−1 ] all the involved reactions and to evaluate the related parameters on
q heat flow [cal min−1 ] the basis of our experimental runs, (ii) to evaluate the role of heat
Hi enthalpy of the ith reaction [cal mol−1 ] and mass transfer in the reaction and (iii) to develop a model useful
for modelling a continuous reactor operating in safe conditions.
Subscripts and superscripts
H proton 2. Experimental
DB double bond
Epox epoxide group 2.1. Materials, apparatus and methods
FA formic acid
PFA performic acid 2.1.1. Materials
0 initial value The used soybean oil, with Iodine Number 128 (gI2 /100 gsample )
aq aqueous phase was purchased in a local food-store (the fatty acid composition
org organic phase of this oil, determined by gas-chromatographic analysis, was (%,
J jacket w/w): palmitic = 11, stearic = 4, oleic = 23, linoleic = 56, linolenic = 5,
P process others = 1).
r epoxidation reaction Hydrogen peroxide (60 wt.%) was supplied by Mythen SpA.
ox H2 O2 decomposition Formic acid (96%, w/w), sulphuric acid (97%, w/w), phosphoric acid
deg ring opening reaction (85%, w/w) and all others employed reagents were supplied by
add adding Aldrich at the highest level of purity available (>99.9%) and were
used as received without further purification.

been published concerning the kinetics of the epoxidation of veg-
etable oil [4–6]. Normally, all these papers have considered, for the
purposes of simplification, the reaction occurring in a monophasic
system (pseudo-homogeneous models). On the contrary, Rangara-
jan et al. [7] first proposed a two-phase model, considering a local
phase concentration to calculate the reaction rate and the mass
transfer effect. Campanella et al. [8,9] have studied the degrada-
tion of the oxirane rings of epoxidized vegetable oils by hydrolysis
and by H2 O2 attack. Moreover, Campanella et al. [4], also reported a
study on the kinetics of both epoxidation and oxirane rings opening
reaction, taking into account the presence of two reacting immis-
cible liquids. The monophasic approach is clearly oversimplified,
while, the few biphasic approaches, reported by the literature,
need confirmation of the effectiveness of the adopted models and
related parameters. Finally, all of the kinetic studies previously
proposed, did not consider the hydrogen peroxide decomposition
2.1.2. Apparatus
The kinetic runs were carried out in a cylindrical jacketed glass
reactor (500 mL) with three necks, equipped with a thermocouple,
a reflux condenser and a glass dripping funnel for adding the aque-
ous reactants. The reacting mixture was stirred with a magnetic
bar (750 rpm). Temperature control was made by re-circulating
thermostatted water inside the reactor jacket. The temperatures
of the water entering and exiting the reactor jacket and the tem-
perature of the reacting mixture were continuously monitored
and recorded using a data acquisition module purchased from the
National Instruments Co.
2.1.3. Methods
Epoxidation of soybean oil was carried out with peroxyformic
acid generated in situ by reacting H2 O2 (60 wt.%) and formic acid
(95 wt.%) in the presence of sulphuric (98 wt.%) or phosphoric acid
200 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209

Table 1
List of the performed runs and relative reaction conditions.

Run Oil (g) H2 O2 (g) (60 wt.%) Formic acid (g) (95 wt.%) Catalyst Mass of catalyst (g) Operating modality

1 100 36.7 5.38 H2 SO4 0.64 Pulse-fed-batch

2 100 36.7 5.38 H2 SO4 0.64 Fed-batch
3 100 42.0 10.76 H2 SO4 0.84 Fed-batch
4 100 33.4 2.69 H2 SO4 0.57 Fed-batch
5 100 36.7 5.38 H2 SO4 0.64 Pulse-fed-batch
6 100 36.7 5.38 H2 SO4 0.32 Pulse-fed-batch
7 100 36.7 5.38 H2 SO4 0.64 Pulse-fed-batch
8 100 36.7 5.38 H2 SO4 1.28 Pulse-fed-batch
9 100 36.7 5.38 H3 PO4 0.98 Pseudo-fed-batch
10 80 29.5 4.41 H3 PO4 0.78 Pulse-fed-batch

(86 wt.%) as catalysts. The main adopted reaction conditions are 80

summarized in Table 1. 78
Two typical procedures were followed, differing in the modal-
76
ity of the oxidizing aqueous phase addition, i.e., a semi-continuous
fed-batch in the first case and a pulse-fed-batch in the second. Fed- 74

Temperature (°C)
batch runs (runs 2–4 of Table 1) are characterised by a continuous 72
flow rate of 0.3 cm3 /min of an oxidizing mixture of hydrogen perox-
ide and formic acid, the composition of which is reported in Table 1, 70

that is added to 100 g of well stirred soybean oil containing the 68

amount of catalyst reported in Table 1. The oil was kept at about
65 ◦ C, while, the oxidizing mixture was fed at room temperature.
As the reaction is strongly exothermic, the temperature in the reac-
tor increased, reaching a value oscillating between 68 and 72 ◦ C.
However, the temperature profile, over time, was continuously
monitored and registered. The temperature of the thermostatting
fluid was changed to obtain an approximately constant reaction
temperature. Small samples were taken at different times for the
analysis of the Iodine Number (for determining double bonds con-
version) and Epoxide Number (for evaluating selectivity and yield).
Thermal profiles obtained for the runs 2–4 are reported in Fig. 1.
These runs were performed with the scope of estimating the role
of formic acid concentration and evaluating the kinetic constants
for a given temperature.
For the pulse-fed-batch procedure: 100 g of soybean oil
(0.504 mol of double bonds unsaturation) were loaded in the glass
reactor and heated under constant magnetic stirring (750 rpm), at
80
75
70
65
60
55
Temperature (°C)
Experimental temperature profile
66
Calculated temperature profile
64
62
Jacket temperature profile
60
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75
Time (minutes)
Fig. 2. Thermal behaviour of the run 1 of Table 1.
55 ◦ C. At this temperature, the catalyst was added and the heating
was continued. At about 60 ◦ C, a mixture of hydrogen peroxide and
formic acid, the composition of which is reported in Table 1, was
added in a limited amount. The increase of the temperature of the
reacting mixture was continuously recorded by a data acquisition
system. The water re-circulating in the reactor jacket cools the reac-
tion mixture until the initial starting temperature or other higher
predetermined temperature is reached. At this point, another lim-
ited amount of hydrogen peroxide/formic acid mixture was added
to the oil/catalyst mixture and again the temperature progres-
sively increased for the epoxidation reaction and decreased due
Run 2
to the cooling effect of the re-circulating thermostatted water. The
TP described sequence was repeated several times (in general 7–10
TJ

50
75
70 35
65 Run 3
Oxidising reactants adding (cm )
3

60 TP 30

55 TJ 25
50
75 20

70 15 Run 1
65 Run 5
Run 4 10 Run 6
60 TP Run 7
55 5 Run 8
TJ
50
0
0 25 50 75 100 125 150 175 200 225 250 275 300 0 5 10 15 20 25 30 35 40 45 50 55 60 65
Time (minutes) Time (minutes)

Fig. 1. Thermal profiles of runs 2, 3 and 4 of Table 1. Tp and TJ refer, respectively, to Fig. 3. Reactants mixture successive additions (H2 O2 and formic acid in the ratio
the reaction environment and jacket temperature. reported in Table 1) for epoxidation reaction runs with H2 SO4 as catalyst.
 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209 201

100 Temperature inside reactor -Run 5 45

Temperature of recirculating fluid-Run 5

Cumulative oxidising reactants adding (cm )

3
Temperature inside reactor -Run 6 40
Temperature of recirculating fluid-Run 6
90
35

30
Temperature (°C)

80
25

70 20

15
60
Adding period Run 6 Digestion period Run 6 10 Run 9
Adding period Run 5 Run 10
Digestion period Run 5 5
50
0 10 20 30 40 50 60 70 80 90 100 0
Time (minutes) 0 30 60 90 120 150 180 210 240 270 300
Time (minutes)
Fig. 4. Thermal profiles obtained in the epoxidation of soybean oil by following
different adding times and modalities (see Table 1, run 5, run 6 and Fig. 3). Fig. 5. Reactants mixture successive additions (H2 O2 and formic acid in the ratio
reported in Table 1) for epoxidation reaction runs with H3 PO4 as catalyst (runs 9
and 10 of Table 1).

additions) and Fig. 2 reports the thermal effect registered related

to run 1 of Table 1. tion of the oxidizing reagents. Also in these cases thermal profiles
The sequence and modality of the reactant additions was consis- were carefully monitored and registered. Examples of such profiles
tently changed for each pulse fed batch run, listed in Table 1, with related to run 5 and 6 are reported in Fig. 4. As can be seen from
the scope of evaluating the effect of increasing or decreasing the this figure, the addition of 7 mL of reactants (first adding of run 5)
amount of reactants added in each step. The added amounts were caused a more significant increase in the temperature than when
also changed within a single run starting, normally, with smaller using 3 mL (first adding in run 6). The thermal profiles reported in
amounts for the first steps and increasing the amounts in the sub- Fig. 4 suggest that in the first part (adding period) the epoxidation
sequent steps. The pulse-fed-batch reactants adding modalities for reaction rate is much faster, as the observed increase in the tem-
run 1, 5–8 of Table 1 are reported in Fig. 3 (catalyst sulphuric acid). In perature would appear to show. This behaviour can be related to
run 1, samples to be analyzed were taken, after each pulse addition, the dilution of both the catalyst and the H2 O2 , occurring during
to evaluate activities and selectivities achieved during the oxidiz- the reaction, also as a consequence of the large amount of water
ing additions. In this case, after each addition we waited the time formed in the epoxidation reaction, in the ring opening reactions
necessary for lowering the temperature to a predetermined value and in the H2 O2 decomposition. The temperature peaks observed
before a further pulse addition. With this modality, the entire oper- after the adding time were due to our intervention in heating the
ation had a duration of about 1 h. This run was performed with the reaction mixture to investigate the behaviour of the reaction dur-
scope of determining the overall heat transfer coefficient U and ing the digestion period at higher temperatures in order to better
validating the kinetic model and related parameters. estimate the ring opening reaction rates. In these runs (runs 5–8 of
In the other pulse-fed-batch runs (5–8) we have to distinguish Table 1) the first sample for the analyses was taken at the end of the
two different reaction periods, i.e., the addition period that had a addition period. Other samples were then taken at different times
duration of 24–36 min (see Fig. 3) and a digestion period in which during the digestion period with the aim of better evaluating the
the reaction continued for about 2–3 h without any further addi- contribution of the epoxide ring opening reactions in lowering the

74 74
A B
72 72

70 70
Temperature (°C)

Temperature (°C)

68 68

66 66

64 64

62 62

60 Temperature inside the reactor 60 Temperature inside the reactor

Temperature of the recirculating fluid Temperature of the recirculating fluid
58 58
0 60 120 180 240 300 360 420 480 540 0 60 120 180 240 300 360 420 480 540
Time (minutes) Time (minutes)

Fig. 6. Thermal behaviour of runs 9 of Table 1 (reported in A) characterised by a pseudo-continuous addition of the oxidizing reactants (H2 O2 and formic acid in the ratio
reported in Table 1) and of the pulse-fed-batch run 10 of Table 1 (reported in B).
202 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209

90
120 Run 6: 0.32g cat
Run 7: 0.64g cat
110 85 Run 8: 1.28g cat
100
80
Temperature (°C)

90

Temperature (°C)
H3PO4 0.98g - ΔT= 13°C
H2SO4 0.64g - ΔT= 65°C 75
80

70 70

60
65
50

40 60
0 5 10 15 20 25 30 35 40 45
Time (minutes) 55
1.0 1.5 2.0 2.5 3.0 3.5 4.0
Fig. 7. Thermal profiles related to the first addition of 17 mL of the oxidizing reac-
tants (solution of H2 O2 and formic acid) by using, respectively H2 SO4 (run 5 of Time (minutes)
Table 1) and H3 PO4 (run 9), as catalysts.
Fig. 8. Experimental thermal profile of the first reactants addition in epoxidation of
soybean oil at different catalyst concentration (see Table 1, runs 6, 7 and 8).

selectivity to the desired product. These runs were also performed

with the scope of determining the activation energies considering
the significant temperature increases obtained during these types
of runs.
Other examples of pulse-fed-batch runs, related to phosphoric
acid (runs 9 and 10) are reported in Fig. 5. As can be seen in Fig. 5,
run 9 was performed by adding very small amounts of the reactants
thus approaching a fed-batch behaviour. The corresponding ther-
mal behaviour for such run is reported in Fig. 6A, while, the thermal
profile for run 10 is reported in Fig. 6B.
As already mentioned for the runs 5–8 and 10, when the reac-
tants addition steps were completed, the reaction was continued
for several hours, at a constant temperature. Periodically, sam-
ples were taken from the reactor and quickly cooled; then the oil
phase was separated by the aqueous phase, by centrifugation at
3500 rpm, for 30 min in 50 mL vials. The organic phase was dis-
solved in ethyl acetate and washed with a solution of sodium
bicarbonate in water (5 wt.%) and then, with a solution of NaCl
(5 wt.%), until the complete elimination of the acidity was achieved.
The final product was then dried in a rotary evaporator. The sam-
ples taken were analyzed to determine the Iodine Number, the
oxirane number and the residual unreacted hydrogen peroxide,
according to the analytical methods reported by the literature
[12–15].
In order to study the heat exchange of the used reactor, several
runs were carried out by heating or cooling, using the re-circulating
thermostatted water, a known amount of soybean oil put in the
reactor. The observed changes in temperature of the oil and of the
heating/cooling water, at respectively the inlet and the outlet of the
reactor jacket, were recorded and registered using the data acqui-
sition system. These temperatures data allow an estimation of both
the heat exchange rate and the overall heat exchange coefficient in
the used device.

Table 2
Kinetic data related to the runs performed in isothermal conditions (runs 2–4 of Table 1) the mean volume of each withdrawn sample was about 4 cm3 .

Run Time (min) Epoxide Number Iodine Number Conversion (%) Yield (%) Selectivity (%)

Exp Calc Exp Calc Exp Calc

2 15 0.37 0.42 110.0 111.0 14.1 4.6 32.6 34.0

30 1.17 1.37 98.0 98.2 23.4 14.5 62.0 63.0
60 2.42 3.00 79.0 68.1 38.3 30.0 78.3 81.0
90 3.66 3.69 55.0 43.6 57.0 45.4 79.6 82.7
120 4.97 5.30 32.0 26.1 75.0 61.6 82.1 82.6
180 5.57 5.78 12.0 11.7 90.6 69.1 76.3 78.9
240 5.72 5.65 7.8 7.0 94.0 70.9 75.4 74.0

3 15 0.52 0.53 108.6 110.0 15.1 6.4 42.4 50.0

30 1.26 1.42 100.2 96.8 21.7 15.6 71.4 72.3
60 2.81 3.10 66.1 66.0 48.3 34.8 72.0 79.3
90 3.54 4.46 50.6 40.0 60.4 43.9 72.0 80.4
120 5.17 5.38 17.8 21.0 86.1 64.1 74.4 80.0
180 5.77 5.85 7.8 5.0 93.9 71.6 76.2 75.4
240 5.35 5.48 2.5 1.6 98.0 66.3 67.6 68.8
300 2.00 4.8 2.0 0.6 98.4 64.2 65.0 62.2

4 15 0.29 0.34 114.0 115.0 10.9 3.6 32.9 44.0

30 0.80 1.07 107.0 103.0 16.4 9.9 60.5 69.8
60 2.09 2.54 83.0 79.5 35.2 25.9 73.7 75.8
90 3.41 3.73 58.0 56.4 54.7 42.3 77.3 80.6
120 4.54 4.47 37.0 40.3 71.1 56.3 79.2 81.0
180 5.08 4.95 26.0 29.7 79.7 63.0 79.1 79.9
240 5.28 4.95 22.0 23.4 82.8 65.5 79.1 75.6
300 5.38 4.80 19.0 20.1 85.2 66.7 78.4 71.0
 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209
2.2. Description of the kinetic runs
The experimental runs with, respectively, H2 SO4 and H3 PO4 as
catalysts, showed that the two catalysts have different activities,
sulphuric acid being much more active than phosphoric acid. This
fact is evident in Fig. 7, where the thermal profiles, obtained by
adding an equal amount of the oxidant mixture, in the presence
of respectively H2 SO4 and H3 PO4 , are reported. As can be seen,
using H2 SO4 an increase of the temperature of more than 60 ◦ C
was observed, while, with H3 PO4 the increase was only 13 ◦ C. The
epoxidation performed in the presence of H2 SO4 is faster than that
in the presence of H3 PO4 . As a consequence, the stepwise reactants
addition periods of pulse-fed-batch modality can consistently be
shortened, maintaining a satisfactory level of the Oxirane Number.
For this reason, we have focused our work mainly on the description
of the runs performed with H2 SO4 . The epoxidation reaction rate,
seems to depend linearly on the sulphuric acid concentration, as
can be seen by observing the initial temperature profiles reported
in Fig. 8, obtained in the presence of different amounts of catalyst,
but maintaining constant the ratio between the reagents and the
volume of added oxidant.
As reported in Table 1, 10 different kinetic runs were performed.
In Table 2, data collected for the isothermal runs (2–4) are reported.
It is interesting to observe that at low conversions the selectivity
and yield are unexpectedly low. This means that the ring opening
reaction contribution, in the initial part of the reaction, is very high
and this aspect must be taken into account in the kinetic model.
This result is further confirmed by observing the data reported in
Table 3, related to run 1 of Table 1. In this case, samples were taken
after each addition of the oxidant solution. The thermal profiles
of almost all the mentioned runs have already been shown in a
previous section of the work (Figs. 1, 2, 4, 6A and B).
2.3. Theoretical approach for the development of the kinetic
model
2.3.1. Description of the reaction scheme and of the kinetic rate
laws
As mentioned in a previous section, the epoxidation reaction
occurs in two successive steps: one, promoted by an acid catalyst,
occurring in the aqueous solution between formic acid and hydro-
gen peroxide to give performic acid, and the other occurring in the
organic phase between performic acid and the double bonds con-
tained in the soybean oil molecules. Obviously, the second reaction
can occur only if performic acid partially migrates from the aque-
ous to the organic phase. This implies, for a correct interpretation
of the experimental data, the knowledge of performic and formic
acids partition coefficients and of the related mass transfer coeffi-
cients. Then, it is necessary to take into account that other parasitic
reactions can occur, opening the epoxide ring. On the basis of the
suggestions coming from the literature [4–9] a complete scheme of
all the possible occurring reactions can be written as it follows:
Aqueous phase:
H2 O2 + HCOOH ↔ HCOOOH + H2 O
H2 O2 → H2 O + 0.5O2
Organic phase:
HCOOOH + C C(i) → Epox(i) + HCOOH
Epox(i) + H2 O → DEGH2 O
Epox(i) + FA → DEGFA
Epox(i) + PFA → DEGPFA
Epox(i) + H2 O2 → DEGH2 O2
203
Then, it must pointed out that reaction (5), corresponding to the
reaction between the double bonds and performic acid, to give the
epoxidized desired product, according to La Scala and Wool [16],
occurs with rates that are different for, respectively, trienes, dienes
or monoenes. In particular, La Scala and Wool [16] demonstrated
that trienes are much more reactive than dienes and monoenes,
and perhaps also that the ring opening reaction rate would be sus-
ceptible to the characteristics of the molecular structure. This will
be ascertained through our experimental runs and we have taken
into account these aspects in our kinetic model. According to the
reported scheme (3–9) there are four reactions concurring in low-
ering the yield by opening the oxirane rings. For this purpose, it is
opportune to point out that the only works describing the kinetics
of these reactions [8,9,17], show that the pH of the aqueous solution
is determinant in promoting all these reactions. Moreover, accord-
ing to the same authors the ring opening reactions mainly occur
at the water–oil interphase. In this case, the reaction scheme can
be simplified by considering only one ring opening reaction pro-
moted by the protonic attack of the epoxide rings at the interphase
followed by the formation of a carbocation that can promptly react
with hydrogen peroxide, water, formic and performic acids. If the
ring opening to carbocation step is rate determining, the kinetic law
in this case can simply be described by the product of the aqueous
protonic concentration multiplied by the oxirane group concentra-
tion at the interphase that we have assumed proportional to that
of bulk concentration. On the other hand:
Epox(i) + H3 O+ → DEGH (10)
In this work we have considered both the hypothesis of four
independent ring opening reactions, each with its own kinetic law,
and a single acid promoted ring opening reaction obtaining com-
parable results in the runs simulations.
On the basis of both our observations and the information taken
from the literature, the following kinetic expressions have been
adopted, all expressed in [mol L−1 min−1 ]:
Aqueous phase:
 
aq + 1
rb = kb [H3 O ] [H2 O2 ]aq [FA]aq − [PFA]aq [H2 O]aq (11)
Keq
aq
ra = ka [H2 O2 ]2.5
aq (12)
Organic phase:
org
rc (i) = kc (i)[DB(i)]org [PFA]org (13)
Liquid–liquid interphase:
rd (i) = kd (i)[Epox(i)]org [H3 O+ ]aq
org
(14)
The index (i) is related to the possibility of a different reactivity
of the double bonds, respectively present in trienes, dienes and
monoenes fatty acid chains.
(3) The kinetic expression related to the hydrogen peroxide decom-
position considers a 2.5 order as our previous researches have
(4)
demonstrated [10].
The kinetic law used to describe the peroxyacid formation (Eq.
(11)) is taken from the literature. As a matter of fact, De Filippis
(5)
and Scarsella [18] have deeply studied the kinetics of this reaction
(6) in the presence of a soft acid environment (formic acid) obtaining,
in those conditions, an activation energy of 10.40 kcal/mol and an
(7) equilibrium constant, at 25 ◦ C, of 0.8. The kinetic parameters found
(8) by De Filippis and Scarsella have been adopted by us as first attempt
parameters for submitting our experimental data to mathematical
(9) regression analysis.
204 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209

Table 3
Kinetic data related to run 1 of Table 1. The mean volume of each withdrawn sample was about 2.3 cm3 .

Time (min) Epoxide Number Iodine Number Conversion (%) Yield (%) Selectivity (%)

Exp Calc Exp Calc Exp Calc

4.0 0.37 0.24 109.5 120.0 14.5 4.6 31.7 47.9

13.0 0.78 0.77 103.8 108.0 18.9 9.7 51.3 61.4
22.0 1.45 1.60 91.2 93.1 28.7 18.0 62.6 72.7
30.0 1.92 2.08 82.4 83.6 35.6 23.8 66.8 74.2
39.0 2.63 2.80 68.6 70.0 46.4 32.6 70.2 76.7
51.0 3.70 3.92 51.2 49.3 60.0 45.9 76.5 79.1
56.0 4.23 4.24 40.3 43.1 68.5 52.5 76.6 79.2
61.0 4.46 4.51 37.5 37.7 70.7 55.3 78.2 79.1

2.3.2. Theoretical approach for the estimation of partition binary parameters for percarboxylic acids, in particular performic
coefficients, mass transfer rates and overall heat exchange acid. For this reason, both UNIQUAC and UNIFAC cannot be con-
As mentioned before, epoxidation reaction requires that per- fidently applied to describe the partition of all the chemical
formic acid, formed in the aqueous phase, migrates in the organic species constituting the reactive system. Concerning the partition
phase (see reaction (5)). In the meantime, reaction (5) gives formic of formic–performic acid, this aspect has been tackled only by Cam-
acid that returns back to the aqueous phase. Partition equilibria panella et al. [4] using UNIFAC, but the groups’ contributions for
for these two compounds allow the development of both the reac- performic are not reported in the current literature and Campanella
tions (3) and (5), provided that the interface surface area is large et al. don’t explain how they estimated these values. For this reason,
enough to warrant a mass transfer faster than the reaction demand. in the present work, the partition coefficients have been calculated
A picture of what occurs in the two phases and across the interface using the SPARC on-line calculator [11]; the following equations
water–oil is shown in Fig. 9. represent the obtained values and related dependence on the tem-
For a rigorous kinetic approach, it is necessary to consider perature:
together the partition and mass transfer phenomena related to both
1
formic and performic acids. Moreover, water and hydrogen per- = 9.0E − 07Tp2 − 5.0E − 05TP + 0.0046, R2 = 0.9782 (21)
HFA
oxide are also partitioned even though their solubilities are very
low. 1
= 4.0E − 06Tp2 − 4.0E − 05TP + 0.0402, R2 = 0.9899 (22)
Concerning mass transfer rates, we adopted the Whitman’s two HPFA
films theory [19] in which the gradients are confined into the 1
boundary layers of the two liquids sides. In this case, we can write = 9.0E − 08Tp2 − 6.0E − 06TP + 0.0005, R2 = 0.9592 (23)
HH2 O2
for a generic j specie the following mass balance equations:
In Fig. 10, a plot of the oil/water partition coefficients for, respec-
JFA = ˇFA ([FA]aq − HFA [FA]∗org )
aq aq
(15) tively, formic, performic acids and hydrogen peroxide are reported
JFA = ˇFA ([FA]∗org − [FA]org )
org org
(16) as a function of the temperature. As water is always present in great
excess, its concentration in oil can be considered constant and equal
JPFA = ˇPFA ([PFA]aq − HPFA [PFA]∗org ) to an average value of 0.28 g of water for 100 g of oil, at 70 ◦ C. From
aq aq
(17)
the data reported in Fig. 10 we can observe that solubilities are in
ˇPFA ([PFA]∗org
org org
JPA = − [PFA]org ) (18) the order PFA > FA > H2 O2 .
Concerning ˇJi , this parameter represents the product between
([H2 O2 ]aq − HH2 O2 [H2 O2 ]∗org )
aq aq
JH = ˇH (19)
2 O2 2 O2 the mass transfer coefficient and the interfacial surface area for
ˇH O ([H2 O2 ]∗org
org org the component J in the phase i. In particular, we have adopted,
JH O = − [H2 O2 ]org ) (20)
2 2 2 2
as a first approximation, that this parameter is constant for the
aq∗ org∗
where HJ are the partition coefficients defined as [C]j /[C]j .
In order to calculate the partition coefficients, useful for describ- 0.09
ing the studied system, some authors have applied, on the basis Performic Acid
0.08 Formic Acid
of experimental measurements on soybean oil/acetic acid/water
or epoxidized soybean oil/acetic acid/water systems, the UNIQUAC 0.07
Hydrogen Peroxide
approach, matching the experimental data [20,21]. Unfortunately,
0.06
there is not enough experimental data available related to the
0.05

0.04
1/Hj

0.0075

0.0050

0.0025

0.0000
20 30 40 50 60 70 80 90 100 110
Temperature (°C)

Fig. 10. Oil/water partition coefficients for formic, performic acids and hydrogen
Fig. 9. Simplified scheme of both the partition equilibria and the reactions involved peroxide, as a function of the temperature. The coefficients have been obtained by
in the two phases aqueous and organic. using the SPARC on-line calculator.
 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209 205

Table 4 Table 5
Average values of density, specific heat and reaction enthalpy. Kinetic constants values determined at the reference temperature of 70 ◦ C by math-
ematical regression analysis on the experimental data and numerical value of other
Density (g cm−3 ) Specific heats (cal g−1 ◦ C−1 ) Enthalpy (cal mol−1 ) parameters of the model.
H2 O2 = 1.44 cP,H2 O2 = 0.662 Hr = −55,000 [7]
Parameter Value Unit
oil = 0.89 cP,oil = 0.50 Hox = −23,440 [23]
A = 1.16 cP,A = 0.531 Hdeg = −21,000 [24] ka 1.04E−03 ± 1.6E−0.4 L1.5,aq mol−1.5 min−1
H2 O = 0.985 cP,H2 O = 1 kb 6.56E−01 ± 2.0E−02 L2,org mol−2 min−1
Keq 5.17 ± 0.8 –
kcmono,a 2.72 ± 0.5 Lorg mol−1 min−1
aq
aqueous phase (ˇJ ). In fact, as the liquid volume (Vaq ) increases kcmono,b 2.00 ± 0.4 Lorg mol−1 min−1
kctri 133.9 ± 9.8 Lorg mol−1 min−1
during the experimental runs, the generated absolute interface area
kdmono,a 2.94E−03 ± 4.6E−04 Laq mol−1 min−1
(S) increases too, so the ratio between these two parameters, that kdmono,b 7.97E−04 ± 5.7E−05 Laq mol−1 min−1
represents the specific interfacial area for the liquid phase (aaq ), kdtri 0.40 ± 0.05 Laq mol−1 min−1
can be considered roughly constant (aaq = S/Vaq ). On the contrary, ˇaq 250 ± 20 min−1
considering the organic phase, the interfacial area increases with Eab 11.39 ± 0.6 kcal mol−1
org Eac 24.89 ± 0.2 kcal mol−1
the aqueous volume. Therefore, it is possible to obtain ˇJ from
Ead 8.86 ± 0.4 kcal mol−1
the following expression: U0 4571 ± 10 cal m−2 min−1 ◦ C−1
˛ −32.39 ± 0.2 cal m−2 min−1 ◦ C−1
org V aq aq V
aq
ˇj = kL aorg = kL aaq = ˇj (24) Parameters valid for the runs performed with H2 SO4 . Parameters valid for the runs
V org V org
performed with H3 PO4 .
aq
We have assumed that ˇJ
is the same for all the involved compo-
nents.
Starting from expressions (21)–(24), it is possible to calculate runs by considering the difference between the temperature of the
the concentrations of all the mentioned compounds at the interface, thermostatting fluid flowing in the jacket and the reaction temper-
by stating that in steady-state conditions, the following balance is ature inside the reactor (see Figs. 2,3,5, and 8A and B). However,
always valid: an approximated linear dependence of U with the conversion has
aq org
been found:
JJ Vaq = JJ Vorg (25)
U = U0 + ˛XDB (31)
As a consequence the following expressions for calculating the
interphase concentrations, can be derived: 2.4. Mass and energy balances
aq org
Vaq ˇFA [FA]aq + Vorg ˇFA [FA]org
[FA]∗org = org aq (26) 2.4.1. Mass balance
Vorg ˇFA + Vaq ˇFA HFA
The mass balance has been written as follows:
aq org
Vaq ˇPFA [PFA]aq + Vorg ˇPFA [PFA]org
[PFA]∗org = (27) accumulation = inlet + reacted[=]mol min−1 (32)
org aq
Vorg ˇPFA + Vaq ˇPFA HPFA
The inlet term represents the adding of the reactive mixture of
aq org
Vaq ˇH [H2 O2 ]aq + Vorg ˇH [H2 O2 ]org H2 O2 and formic acid; for this purpose, in the case of pulse fed
[H2 O2 ]∗org = 2 O2
org aq
2 O2
(28)
Vorg ˇH + Vaq ˇH HH2 O2 batch runs, it is necessary to introduce a square pulse function to
2 O2 2 O2
simulate the additions. This function can be written as the sum of
Temperature profiles obtained by heating or cooling soybean each adding volume divided by the time necessary for the adding
oil in the absence of the reaction, have been interpreted using the itself. It is then possible to write the mole balance equation for each
following heat balance equation: component in (mol min−1 ).
dTP UA(Tj − TP ) ◦ Aqueous phase:
= [ C min−1 ] (29) aq
dt Cpoil Woil dnH
2 O2 aq
= FH2 O2 + (−ra − rb − JH )Vaq (33)
where TP is the temperature of the oil inside the reactor and Tj is dt 2 O2

the temperature of water flowing in the external jacket. In order aq

dnH
to apply this equation it is necessary to find an expression for the 2O
= FH2 O + (ra + rb )Vaq (34)
specific heat capacity of the oil. For this purpose, it is possible to dt
find in the literature [22] the following relationship: aq
dnFA aq
−4 −1 ◦ −1
= FFA + (−rb − JFA )Vaq (35)
cP,oil = 4.777 × 10 Tp + 0.4688[cal g C ] (30) dt
aq
From all the data collected, in different experimental conditions, dnPFA aq
= FPFA + (+rb − JPFA )Vaq (36)
we have estimated for the overall heat exchange coefficient (U), in dt
the range of 60–75 ◦ C (temperature range of the reaction), a value Organic phase:
falling in the range of 2550–3000 cal m−2 min−1 ◦ C−1 .
org
It must be considered that as a consequence of the epoxidation dnH O org
reaction the oil phase composition changes with the reaction time.
2 2
= (+JH )Vorg (37)
dt 2 O2
As a matter of fact, epoxidized soybean oil is much more viscous org
dnFA org
than the unreacted substrate (respectively 450cP for epoxidized = (+rc (i)) + JFA Vorg (38)
soybean oil and 110cP for soybean oil at 25 ◦ C) and we can fore- dt
see that the overall heat exchange coefficient U will change with org
dnPFA org
the conversion degree. Therefore, the values estimated with inde- = (−rc (i) + JPFA )Vorg (39)
dt
pendent runs in the absence of the reaction and of the aqueous
org
solution can only be considered first attempt values. A more pre- dnDB (i)
= −rc (i)Vorg (40)
cise parameter can be obtained by regression analysis on the kinetic dt
206 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209

140 8 140 8
A B
120 120
6 6
100 100

Oxirane Number
Iodine Number

Oxirane Number
Iodine Number
80 80
4 4
60 60

40 40
2 2
20 20

0 0 0 0
0 50 100 150 200 250 0 50 100 150 200 250
Time (minutes) Time (minutes)

140 8
C
120
6

Oxirane Number
100
Iodine Number

80
4
60

40
2
20

0 0
0 50 100 150 200 250
Time (minutes)

Fig. 11. Experimental and simulated iodine and oxirane numbers against the epoxidation reaction time reaction. Dots are experimental data curves correspond to model
prediction. Plots in (A)–(C) correspond to runs 5–7 of Table 1. The mean volume of each withdrawn sample was about 9.2 cm3 .

org
dnEpox (i) Therefore, the heat balance can be interpreted with the follow-
= (+rc (i) − rd (i))Vorg (41) ing relation:
dt
org
dnDEG (i) qaccumulation = qinlet + qreaction − qexchange [=]cal mol−1 (45)
H
= +rd (i)Vorg (42)
dt and the terms of the heat balance can be written as follows:
In order to integrate these equations the following initial conditions dTp
qaccumulation = c̄p Mtot (46)
have been considered. First of all, the total number of the double dt
bonds present in the oil is given by the following expression: qinlet = (WH2 O2 cpH2 O2 + WA cpA + WH2 O cpH2 O )(Tadd − Tp ) (47)
0
Niodine Moil
nTOT,0
DB
= 0
where Niodine = 128 qreaction = [Hr (−rc (i)) + Hdeg (−rd (i))]Vorg
PMI2 100
+ [HOX (−ra )]Vaq (48)
(gI2 in 100 g of substrate) (43)

Moreover, as it has been seen, it is important to consider sep- qexchange = UA(Tp − TJ ) (49)
arately the double bonds in fatty acid chains of, respectively,
The overall thermal exchange coefficient (U) and its dependence
monoenes, dienes and trienes. The initial values of the mentioned
on double bond conversion, has been evaluated by regression anal-
double bonds are calculated starting from the molar fraction of each
ysis. All the densities, specific heats and enthalpy change values
kind of double bond present in the soybean oil ˚i . In this way, the
necessary for the calculation are reported in Table 4.
initial value of each double bond is calculated as follows:
0
nDB (i) = ˚i nTOT,0
DB
(44) 2.5. Determination of the model parameters, runs simulation and
discussion
The other initial concentrations are zero before adding the oxi-
dizing reagents. Each kinetic run, reported in Table 1, is characterised by a num-
ber of 6–8 samples taken at different times. Three different analyses
2.4.2. Energy balance have been carried out on these samples to determine the Iodine
The second step of the experimental data elaboration regards Number (to evaluate the double bonds conversion), the Oxirane
the simulation of the thermal profiles using an opportune enthalpy Number (to evaluate the epoxidation reaction yields) and the resid-
balance. As a first approximation, we have considered that the ual hydrogen peroxide after the epoxidation reaction and eventual
heat released by the reactions is instantaneously distributed on the decomposition. Finally, the evolution of the temperature along the
entire mass of the reaction. time in each run has been monitored by a data acquisition system
 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209 207

for each second of reaction time, that is, for the temperature we 105
A
have thousands experimental points. All of this experimental data 100
has been subjected to mathematical regression analysis and the
95
best fittings have been obtained with the parameters reported in
Table 5. The dependence of the kinetic constants with the temper- 90

Temperature (°C)
ature has been described with the following relationship: 85
   80
Eai 1 1
ki = kiref exp − (50) 75
R T ref Tp
70
65
kiref being the kinetic constant chosen at a fixed temperature (here
70 ◦ C). The activation energies values, when known, have been 60 Experimental temperature inside reactor
taken, as a first approximation, by the literature [7–9,16,18]. 55 Calculated temperature inside reactor
As reported in the previous Section 2.3.1, and in agreement Temperature of reciculating fluid
50
with the literature findings [16] trienes are more reactive than 0 30 60 90 120 150 180 210 240 270 300
dienes and monoenes in the epoxidation reaction. Our experimen- Time (minutes)
tal results also suggest that the ring opening reaction rate is higher
105
for trienes and the selectivity at low conversions is much lower B
as a consequence. Hence, in order to simulate the experimental 100
data it would be necessary to introduce different rates and related 95
parameters for trienes, dienes and monoenes. As a matter of fact,
90
we observed that in order to obtain a satisfactory agreement it is

Temperature (°C)
enough to distinguish only the greater activities in both the reac- 85
tions (epoxidation and ring opening) for only one double bond of 80
the trienes (about 5% of the total double bonds in soybean oil). All 75
of the other double bonds have approximately the same reactivity
70
in both the reactions. In conclusion, we have two kinds of dou-
ble bonds, one extremely reactive in both the reactions (5% of the 65
total), and less reactive ones. For each type of double bond we have 60 Experimental temperature inside reactor
two different equations to integrate (epoxidation and ring open- Calculated temperature inside reactor
55
ing rates) and the related parameters are reported in Table 5. We Temperature of reciculating fluid
have obtained a reactivity ratio between the most reactive double 50
bonds and the others of approximately 50 compared to the value 0 30 60 90 120 150 180 210 240 270 300
of 4 that can be estimated from the work of La Scala and Wool Time (minutes)
[16] but these authors have obtained their results operating at a
Fig. 12. Experimental and simulated temperature of the reacting system for the run
lower temperature (19 ◦ C) in the absence of sulphuric acid catalyst 5 (graph A) and 7 (graph B) of Table 1. Continuous curve is related to experimental
and interpreted their results with a monophasic model. The pres- data, dotted curve correspond to the model prediction.
ence of sulphuric acid probably favours the double bonds shifting,
increasing the formation of more reactive conjugated double bonds. Fig. 14(A and B) on the contrary, show the experimental data and
The agreements obtained with the described model for the related simulations for the runs 9, 10 of Table 1, performed in the
isothermal runs 2–4 can be appreciated in Table 2, where, exper- presence of H3 PO4 , as a catalyst, while, Fig. 15(A and B), shows ther-
imental and calculated Epoxide Numbers, Iodine Numbers and mal profiles related to runs 9 and 10. It is interesting to observe that
selectivities are reported for comparison. As can be seen in Fig. 1, the kinetic constants of the epoxidation reaction for, respectively,
these runs have shown an increase in the reactor temperature of phosphoric and sulphuric acid, are comparable, but the ring open-
about 5–6 ◦ C as a consequence of the reaction. This aspect has been
considered and the values of U (see relation 31), the overall thermal
exchange coefficient that can be calculated with the parameters 60
reported in Table 5, effectively reproduce the observed thermal
gradients.
The agreement obtained in describing run 1 can be appreciated
in Table 3 concerning Iodine Numbers, Epoxide Numbers and Selec-
40
tivities and in Fig. 2 for the thermal profile. Also in this case the
H2O2 (wt.%)

gradient between the temperature of the reaction and that of the

thermostatting fluid, flowing in the reactor jacket, is well simulated
as can be seen in Fig. 2.
The model developed in this work is able to simulate, with 20
satisfactory agreement, all of the kinetic runs performed. For this
purpose, Fig. 11(A–C) shows the experimental data and related sim-
ulations for the runs 5–7 of Table 1, performed in the presence of
sulphuric acid, as a catalyst.
The simulation of the thermal profiles for the runs 5 and 7 are 0
0 50 100 150 200 250
reported, as examples, in Fig. 12(A and B). The model is also able to
Time (minutes)
simulate the hydrogen peroxide consumption during the time (for
both the reactions of epoxidation and decomposition), see Fig. 13 Fig. 13. Experimental and simulated H2 O2 (wt.%) residual against the reaction time.
as an example, corresponding to run 5 of Table 1. Dots are experimental data, continuous curve is model prediction for run 5 of Table 1.
208 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209

140 8 140 8
A B
120 120
6 6
100 100

Oxirane Number

Oxirane Number
Iodine Number
Iodine Number 80 80
4 4
60 60

40 40
2 2
20 20

0 0 0 0
0 50 100 150 200 250 300 350 400 450 500 550 0 200 400 600 800 1000 1200 1400 1600
Time (minutes) Time (minutes)

Fig. 14. Experimental and simulated Iodine and Oxirane Numbers against the epoxidation time reaction. Dots are experimental data curves are model prediction. Plot in (A)
and (B) correspond to run 9 and 10 of Table 1. The mean volume of each withdrawn sample was about 8 cm3 .

ing kinetic constants are different even considering the difference
in the acidity for the two different systems. On the contrary, the
activation energies found by regression analysis for both the cata-
lysts and reactions are very similar and an average value is reported
in Table 5.
In the examined runs we have changed the temperature, the
adding oxidant solution modality, the type and catalyst concentra-
tion and the concentration of formic acid. The hydrogen peroxide
concentration and iodine and Epoxide Number change with time
during each run. Therefore, we have all of the possible factors that
can affect the reaction under control. It must pointed out that, nor-
mally, mass transfer does not affect epoxidation rate in the adopted
90
A rate is quite low as can be appreciated by comparing the data
85
Temperature (°C)
conditions with the exception of those runs in which very high
temperatures are reached.
Another observation is related to the identification of the rate
determining step (rds). By plotting the ratio between the epoxida-
org(i) aq
tion and the performic acid formation rates rc /rb it is possible
to observe a value less than 1. The greatest value obtained for
monoenes is approximately 0.2. This means that epoxidation reac-
tion in the oil phase is rds in disagreement with the assumption
generally made in the previous literature. On the other hand, it
is known that the formation of performic acid occurs faster than
the formation of other peracids [25]. Another observation, deriving
from the previous one is that despite the fact that formic acid inter-
venes with a first order kinetic law in the performic acid formation
rate the influence of its concentration on the overall epoxidation
Experimental temperature inside the reactor
Calculated temperature inside the reactor
reported in Table 2. As can be seen, the evolution of the conversion
Temperature of recirculating fluid
with time for the three runs is quite similar.

80
3. Conclusions
75
The epoxidation of soybean oil with hydrogen peroxide
70 (60 wt.%) has been studied in the presence of, respectively,
sulphuric and phosphoric acid, as catalysts. We observed, in agree-
65 ment with other authors, that the reaction is strongly affected
mainly by the temperature and, in the initial steps, also by the cat-
60 alyst concentration. As it is very difficult to keep the temperature
0 100 200 300 400 500 600 700 of the reacting mixture constant due to the strong exothermicity
Time (minutes) of the main reaction, we have studied simultaneously the kinetics
of all the involved reactions (main and side reactions) in a ther-
90 Experimental temperature inside the reactor
B mal dynamic way, by measuring accurately the evolution of the
Calculated temperature inside the reactor temperature gradient between the thermostatting fluid flowing
85 Temperature of recirculating fluid
in the reactor jacket and the liquid bulk inside the reactor dur-
ing the time. A biphasic kinetic model considering all the reactions
Temperature (°C)

80 occurring, respectively in the aqueous phase (oxidation of formic

to performic acid and hydrogen peroxide decomposition), in the oil
75 phase (epoxidation) and at the water/oil interphase (epoxide ring
opening reaction) has been developed. The partition equilibria of
70 reactants and products, between the two phases, have been esti-
mated using the SPARC method, because there is no experimental
65
data available for the involved substances. The model also consid-
ers the possibility of mass transfer limitation for the diffusion of a
reactant, for example performic acid, from the aqueous to the oil
60
phase or formic acid, formed in the oil phase diffusing from oil to
0 100 200 300 400 500 600 700 water. Kinetic laws have been identified for each reaction with the
Time (minutes) help of the information found in the literature. All of the kinetic
runs have been subjected to regression analysis, thus determining
Fig. 15. Experimental and simulated temperature of the reacting system for the run
5 (graph A) and 7 (graph B) of Table 1. Continuous curve is related to experimental all of the parameters of the model. The model leads to a very satis-
data, dotted curve correspond to the model prediction. factory agreement. Some relevant aspects, never observed before,
 E. Santacesaria et al. / Chemical Engineering Journal 173 (2011) 198–209
have been identified. For example, in contrast with the current lit-
erature on the subject, which asserts that performic acid formation
is rate determining step in the epoxidation, we have demonstrated
that the epoxidation occurring in the oil phase is approximately
five times slower than the oxidation of formic to performic acid.
Another observation arising from the analysis of our experimental
data is that some double bonds (about 5% of the total) are more
reactive in both epoxidation and ring opening reaction. As a conse-
quence, the selectivity to epoxide is low (35–40%) at low conversion
(10–15%) then increases for higher conversions reaching a maxi-
mum value of 75–80% and then decreases again for the intervention
of the ring opening reaction. This last reaction has been considered
in our model as a single reaction whose rate determining step is the
protonic attack to the epoxide ring followed by the formation of a
carbocation that can readily react with hydrogen peroxide, formic
and performic acids, or water. This assumption strongly reduces
the number of parameters of the model without decreasing the
fitting agreement. As water is always present in excess, the acid cat-
alyzed hydrolysis of the epoxide rings is probably the most favoured
reaction. This model has been developed with the aim of simulat-
ing the behaviour of the industrial reactors currently working in
pulse fed-batch conditions but in perspective for modelling and
simulating continuous reactors, this being the basis for the process
intensification.
Acknowledgements
This study was funded by EC VII Framework Programme CP-IP
228853-2 COPIRIDE. Thanks are also due to MYTHEN SpA for the
initial financial support.
References
[1] M.A. Babich, PhD Thesis, Review of Exposure end Toxicity Data for Phtha-
late Substitutes; U.S. Consumer Product Safety Commission 4330, East West
Highway, Bethesda, MD 20814 (2010).
[2] O. Fenollar, D. Garcìa, L. Sànchez, G. Lopez, R. Balart, Optimization of the curing
conditions of PVC plastisols based on the use o fan epoxidized fatty acid ester
plasticizer, Eur. Polym. J. 45 (2009) 2674–2684.
[3] MYTHEN SpA Company; Private communication.
[4] A. Campanella, C. Fontanini, M.A. Baltanas, High yield epoxidation of fatty acid
methyl esters with performic acid generated in situ, Chem. Eng. J. 144 (2008)
466–475.
209
[5] L.H. Gan, S.H. Goh, K.S. Ooi, Kinetic studies of epoxidation and oxirane cleavage
of palm olein. Methyl esters, J. Am. Oil Chem. Soc. 69 (1992) 347–351.
[6] S. Sinadinovic-Fiser, M. Jankovic, Z.S. Petrovic, Kinetic of in situ of soybean oil in
bulk catalyzed by ion exchange resin, J. Am. Oil Chem. Soc. 78 (2001) 725–731.
[7] B. Rangarajan, A. Havey, E.A. Grulke, P.D. Culnan, Kinetic parameters of a two-
phase model for in situ epoxidation of soybean oil, J. Am. Oil Chem. Soc. 72
(1995) 1161–1169.
[8] A. Campanella, M.A. Baltanas, Degradation of the oxirane ring of epoxidized
vegetable oils in liquid-liquid systems. I. Hydrolysis and attack by H2 O2 , Latin
Am. Appl. Res. 35 (2005) 205–210.
[9] A. Campanella, M.A. Baltanas, Degradation of the oxirane ring of epoxidized
vegetable oils in liquid–liquid systems. II. Reactivity with solvated acetic and
peracetic acid, Latin Am. Appl. Res. 35 (2005) 211–216.
[10] E. Santacesaria, M. Di Serio, R. Tesser, V. Russo, R. Turco, A new simple
microchannel divide to test process intensification, Ind. Eng. Chem. Res. 50
(2011) 2569–2575.
[11] L.A. Carreira, S. Hilal, S.W. Karickhoff, in: P. Politzer, J.S. Murray (Eds.), Theoret-
ical and Computational Chemistry, Quantitative Treatment of Solute/Solvent
Interactions, Elsevier Publishers, 1994, http://sparc.chem.uga.edu/sparc/.
[12] Method NGD C 32-1976.
[13] E.C. Deaborne, R.M. Fuoss, A.K. Mackenzie, R.G. Shephered Jr., Epoxy resins
from bis-, tris-, and tetrakis-glycidyl ethers, Ind. Eng. Chem. 45 (12)
(1953) 2716.
[14] C. Paquot, A. Hautfenne, Commission on Oils Fats and Derivatives: Standard
Methods for the Analysis of Oils, Fats and Derivatives, Blackwell Scientific
Publications, IUPAC, Applied Chemistry Division, London, 1987.
[15] I.M. Kolthoff, Sandel, Mecchan, Treatise Analytical Chemistry, vol. 2, 888.
[16] J. La Scala, R.P. Wool, Effect of FA composition on epoxidation kinetics of TAG,
J. Am. Oil Chem. Soc. 79 (2002) 373–378.
[17] F.A.M. Zaher, H. El Mallah, M.M. El-Hefnawy, Kinetics of oxirane cleavages in
epoxidized soybean oil, J. Am. Oil Chem. Soc. 66 (5) (1989) 668–670.
[18] P. De Filippis, M. Scarsella, N. Verdone, Peroxyformic acid formation: a kinetic
study, Ind. Eng. Chem. Res. 48 (3) (2009) 1372–1375.
[19] W.G. Whitman, Preliminary experimental confirmation of the two-film theory
of gas absorption, Chem. Metall. Eng. 29 (1923) 146–148.
[20] M. Jankovic, S. Sinadinovic-Fiser, M. Lamshoeft, Liquid–liquid equilibrium con-
stant for acetic acid in an epoxidized soybean oil–acetic acid–water system, J.
Am. Oil Chem. Soc. (2009), doi:10.1007/S11746-009-1531-Z.
[21] S. Sinadinovic-Fiser, M. Jankovic, Prediction of the partition coefficient for acetic
acid in two phase system soybean oil–water, J. Am. Oil Chem. Soc. 84 (2007)
669–674.
[22] A.A. Noor Azian Morad, F. Mustafa Kamal, T.W. Panau, Yew liquid specific heat
capacity estimation for fatty acids triacylglycerols, and vegetable oils based
on their fatty acid composition, J. Am. Oil Chem. Soc. 77 (9) (2000) 1001–
1005.
[23] C.W. Jones, Applications of Hydrogen Peroxide and Derivatives, Series Ed., RSC
Clean Technology Monographs, Cambridge, 1999, p. 14.
[24] Information from: http://www.ethyleneoxide.com.
[25] S. Leveneur, PhD dissertation under the supervision of D.Y. Murzin, T. Salmi, L.
Estel of the Abo Akademi University (INSA) 2009.