ARTICLE pubs.acs.org/IECR
Biphasic Model Describing Soybean Oil Epoxidation with H2O2 in Continuous Reactors E. Santacesaria,*,† A. Renken,‡ V. Russo,† R. Turco,† R. Tesser,† and M. Di Serio† †
University of Naples “Federico II”, Department of Chemistry, Complesso Universitario Monte S. Angelo, Via Cintia 4, IT 80126 Naples, Italy ‡ Ecole Polytechnique Federale de Lausanne (EPFL), EPFL-ISIC, Station 6, CH-1015 Lausanne, Switzerland ABSTRACT: Epoxidized soybean oil (ESBO) is produced in industry by reacting soybean oil, at 6070 °C, with hydrogen peroxide in the presence of formic or acetic acid. A small amount of sulphuric or phosphoric acid is necessary to catalyze the oxidation of the carboxylic acid to the corresponding per-carboxylic acid. Per-carboxylic acid, formed in situ, migrates from the aqueous phase to the oil phase where it spontaneously reacts with the double bonds to give an oxirane ring. This reaction is extremely exothermic (ΔH = 55 kcal/mol) and must be kept under thermal control. Two undesired side reactions can occur: the oxirane ring-opening and the hydrogen peroxide decomposition. In a previous work, a biphasic kinetic model for describing the epoxidation of soybean oil in fed or pulse-fed-batch reactors has been developed and the parameters of the model have been determined by mathematical regression analysis. In the present paper, the model has been adapted to simulate also kinetic runs performed in two continuous tubular reactors of different sizes, filled with spheres of stainless steel (AISI 316) used as static mixer. The agreement found, in simulating the continuous runs, validates the developed biphasic kinetic model. This model constitutes a valid base for the design of a continuous process and for promoting the process intensification.
1. INTRODUCTION Epoxidized oils are mainly used as plasticizers and stabilizers for PVC resins. In particular, epoxidized soybean oil (ESBO) is a good substitute for phthalates, as plasticizer, these last being banned in many countries for their negative effects on health.1,2 This means that the ESBO demand is destined to increase very much in the future. The epoxidation of vegetable oils is carried out in industry 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 presence of a mineral acid as catalyst. Then, peroxyacid migrates into the oil phase and spontaneously reacts with the double bonds of the fatty acids chains giving an oxirane ring. This reaction is extremely exothermic (ΔH = 55 kcal/mol for each double bond) and the excessive increase of temperature in industrial reactors can be prevented by adding a limited amount of the oxidant reagents (H2O2 mixed with formic or acetic acid) to a mixture of oil and acid catalyst3 in a fed-batch mode. Normally, the reaction requires 810 h at 6075 °C to reach the target values of 6.5 for oxirane number (ON) and 2 for residual iodine number (IN), corresponding to an oxirane yield of about 80% and a double bond conversion of 98%. To increase the ESBO production, for satisfying the demand of the market, it is imperative to shift from fed-batch to continuous operation. To design a continuous process under safe conditions, detailed information about both the kinetics of the reactions occurring and the heat and mass transfer are required. By examining the literature, it is possible to find some papers concerning the kinetics of vegetable oil epoxidation.46 All these papers considered, as simplification, the reaction occurring in a monophasic system (pseudohomogeneous models). Rangarajan et al.7 first proposed a two-phase model. Then, Campanella et al.4 reported a study on r 2011 American Chemical Society
the kinetics of both epoxidation and oxirane rings degradation, taking into account the presence of two reacting immiscible liquids. Campanella et al.8,9 have also studied in detail the degradation of the oxirane rings of epoxidized vegetable oils by hydrolysis and by H2O2 attack. In a recent paper, we have studied the kinetics of the soybean oil epoxidation, in the presence of formic acid, using both sulphuric and phosphoric acids as catalysts.10 In that paper, different kinetic runs have been performed, in dynamic conditions, in fed-batch and pulse-fed-batch reactors, by evaluating the evolution with time of reagents and products and monitoring, at any instant, the change of the temperature. All the kinetic runs have been interpreted with a biphasic kinetic model taking into account (i) the kinetic laws of all the reactions occurring in the two immiscible phases, (ii) the partition of all the components involved in the reactions, (iii) the mass transfer from one phase to the other, and (iv) the heat transfer affecting the reactor temperature. A set of kinetic parameters has been determined by mathematical regression analysis on the different performed runs. In the present paper the biphasic model, developed in the mentioned previous work,10 has been adjusted to describe the behavior of a tubular packed-bed continuous reactor. The reactor packing is constituted of small spheres of stainless steel with the scope to act as static mixer for favoring the local micromixing. The partition coefficients of the reagents and products, in the two reacting Special Issue: CAMURE 8 and ISMR 7 Received: July 26, 2011 Accepted: December 6, 2011 Revised: December 3, 2011 Published: December 06, 2011 8760
dx.doi.org/10.1021/ie2016174 | Ind. Eng. Chem. Res. 2012, 51, 8760–8767
Industrial & Engineering Chemistry Research
ARTICLE
phases, have been estimated with the SPARC method11 as in the previous work. The scope of this work is to validate the previously developed biphasic model on continuous runs verifying in particular the rate law equations and related parameters. This model is clearly a useful base for modeling, optimizing, and intensifying the epoxidation reaction in continuous reactors, representing an improvement of a previous approach based on a monophasic model.12 For this purpose, an application of the model to design a two-step continuous reactor operation with the aim to reach the industrial target limits of ON and IN will be reported in the discussion section. The model could be the basis for developing the soybean oil epoxidation process intensification, following the route recently described by Hessel et al.13
2. EXPERIMENTAL SECTION 2.1. Materials, Apparatus, and Methods. 2.1.1. Materials. The used soybean oil, with iodine number 128 (gI2/100 gsample) was purchased in a local food store (the fatty acid composition of this oil, determined by gas-cromatographic analysis, was (% w/w) palmitic = 11, stearic = 4, oleic = 23, linoleic = 56, linolenic = 5, others = 1). Hydrogen peroxide (60 wt.%) was supplied by Mythen SpA. Formic acid (96% w/w), sulphuric acid (97% w/w), phosphoric acid (85% w/w), and all other employed reagents were supplied by Aldrich at the highest level of purity available (>99.9%) and were used as received without further purification. 2.1.2. Apparatus. The kinetic runs were carried out in two jacketed glass tubular reactors of different sizes, completely filled with spheres of stainless steel (AISI 316) of 2.2 mm diameter. Reactor 1 was a tube of 1.0 cm diameter and 27 cm length, and volume of the empty reactor was 22 cm3, while the void volume of the reactor filled with the mentioned spheres was 8.3 cm3 (void fraction = 0.37). Reactor 2 was a tube of 1 cm diameter and 46 cm length, and the volume of the empty reactor was 38 cm3, while the void volume of the reactor filled with the mentioned spheres was 14 cm3 (void fraction = 0.37). A scheme of the used laboratory plant is reported in Figure 1. 2.1.3. Methods. A preheated stream of oil was fed to the reactor, while a mixture of hydrogen peroxide, formic, and sulphuric acid, taken at room temperature, was mixed with the hot oil at the inlet of the continuous reactor. Periodically, samples were withdrawn at the reactor outlet and quickly cooled; then the oil phase was separated from the aqueous one by centrifugation at 3500 rpm for 30 min in 50-mL vials. The organic phase was dissolved 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 conversion and yield of the withdrawn samples was monitored through the determination of, respectively, the iodine and the oxirane numbers, according to the analytical methods reported by the literature.1417 The conversions have been determined as Conversion ¼
½Iodine NumberINLET ½Iodine NumberOUTLET 100 ½Iodine NumberINLET
The yields to oxirane have been determined as Oxirane Yield ¼
½Oxirane NumberOUTLET 253:8 100 ½Iodine NumberINLET 16
Figure 1. Scheme of the used laboratory plant. TIN and TOUT are respectively the temperatures at the inlet and at the outlet of the tubular reactor, R is the reactor, C is a cooler, and TIC is a temperature controller.
with the iodine number being reported as g of I2 in 100 g of oil and the oxirane number being g of oxygen reacted with double bonds per 100 g of oil. The yields can be limited by the occurrence of the ring-opening reactions.
3.0. RESULTS AND DISCUSSION 3.1. Description of the Kinetic Runs. In Table 1, a list of the kinetic runs performed in continuous tubular packed bed reactors is reported together with the adopted operative conditions. As it can be seen in this table we have operated by changing the inlet temperature, the residence time, and the oxidant/oil ratio. The composition of the oxidant mixture has been kept constant, corresponding to 85.9 wt % of H2O2 (with concentration of 60 wt %), 12.6 wt % of formic acid (with concentration of 96 wt %), and 1.5 wt % of sulphuric acid (with concentration of 97 wt %). 3.2. Theoretical Approach for the Development of the Kinetic Model. The epoxidation reaction, as it has been reported in our recent paper,10 occurs in two successive steps: one, promoted by an acid catalyst, occurring in the aqueous solution between formic acid and hydrogen peroxide to give performic acid, and the other, occurring in the organic phase, between performic acid and the double bonds of the soybean oil molecules. Obviously, the second reaction can occur only if performic acid partially migrates from the aqueous to the organic phase. This implies, for a correct interpretation of the experimental data, the knowledge of the performic and formic acids partition coefficients and of the related mass transfer coefficients. It is then necessary to take into account that other competitive reactions can occur, such as the epoxide rings opening and the hydrogen peroxide decomposition. The complete scheme of all the occurring reactions can be written as follows: Aqueous phase:
H2 O2 þ HCOOH T HCOOOH þ H2 O 8761
ð1Þ
dx.doi.org/10.1021/ie2016174 |Ind. Eng. Chem. Res. 2012, 51, 8760–8767
Industrial & Engineering Chemistry Research
ARTICLE
Table 1. Operative Conditions, Experimental Results, and Simulations Related to the Continuous Epoxidation Runs Performed in This Work in Reactor 1 (Entries 15) and Reactor 2 (Entries 617)a TOUT (°C) run
Qoil/Qox
τ (min)
TJ (°C)
1
4/1
1.66
70.0
2
4/1
1.66
77.0
3 4
4/1 3/2
1.66 1.66
5
3/2
6
8/2
7
IN
T
(°C)
iodine number
oxirane number
EXP
SIM
EXP
SIM
EXP
SIM
66.0
70.0
72.1
125.0
121.0
0.18
0.40
78.0
77.0
79.3
117.0
119.0
0.65
0.51
93.0 67.0
83.0 66.0
93.0 67.0
95.6 69.4
118.0 113.6
114.9 116.3
0.54 0.88
0.71 0.61
1.66
78.0
78.0
78.0
81.4
102.4
102.4
1.58
1.38
1.40
72.5
71.0
74.0
75.3
116.0
113.5
0.55
0.65
8/2
1.40
82.5
80.3
93.0
86.2
101.1
101.1
1.10
1.34
8
9/1
1.40
72.5
70.4
76.0
73.9
119.2
121.6
0.44
0.33
9
9/1
1.40
82.5
80.2
87.0
83.9
117.6
119.3
0.42
0.42
10
4/1
2.80
72.5
70.4
79.0
73.5
108.8
110.2
1.01
0.85
11 12
14/1 14/1
0.93 0.93
72.5 82.5
70.2 80.5
75.0 84.0
73.8 84.2
121.9 121.3
125.1 124.2
0.26 0.25
0.16 0.19
13
14/1
0.93
92.5
90.0
97.5
93.8
121.0
123.5
0.18
0.22
14
28/1
0.47
92.5
90.4
92.5
95.7
123.7
123.7
0.20
0.21
15
29/1
0.47
92.5
90.3
94.0
94.1
125.9
126.9
0.08
0.06
16
26/4
0.47
72.5
70.0
76.5
77.1
115.6
121.6
0.66
0.30
17
24/6
0.47
72.5
70.0
78.0
78.7
110.4
118.9
0.98
0.37
a
The composition of the oxidant mixture has been kept constant corresponding to 85.9 wt % of H2O2 (with concentration of 60 wt %), 12.6 wt % of formic acid (with concentration of 96 wt %), and 1.5 wt % of sulphuric acid (with concentration 97 wt %)
H2 O2 f H2 O þ 0:5O2
ð2Þ
Liquidliquid interphase: þ r d org ðiÞ ¼ kd ðiÞ 3 aaq sp 3 ½EpoxðiÞorg 3 ½H3 O aq
Organic phase: HCOOOH þ C ¼ CðiÞ f Epox ðiÞ þ HCOOH
ð3Þ
Liquidliquid interphase:
In the ring-opening reaction, the formation of carbocation is the rate determining step, as shown in our previous work.10 The carbocation can react then with any nucleophilic molecule (Nu) such as water, hydrogen peroxide, formic, or performic acid with a probability roughly proportional to the related concentrations. The kinetic expressions of the different occurring reactions, in [mol L1 min1], are reported below: Aqueous phase:
The index (i) is related to the possibility of a different reactivity of the double bonds contained in the oil as, respectively, trienes, dienes, and monoenes, according to the findings by La Scala et al.18 and our previous experimental observations.10 For what concerns mass transfer rates, we adopted Whitman’s two films theory19 in which the gradients are confined into the boundary layers of the two liquid sides. In this case, we can write the following mass transfer equations: ð9Þ J FA aq ¼ βFA aq 3 ½FAaq H FA 3 ½FAorg
( ) 1 ½PFA ½H r b aq ¼ kb 3 ½H3 Oþ 3 ½H2 O2 aq 3 ½FAaq O 2 aq 3 aq K 3 eq
ð5Þ r a aq ¼ ka 3 ½H2 O2 2:5 aq
ð6Þ
Organic phase: r c org ðiÞ ¼ kc ðiÞ 3 ½DBðiÞorg 3 ½PFAorg
ð8Þ
ð7Þ
J FA org ¼ βFA org 3 ½FAorg ½FAorg
ð10Þ
J PFA aq ¼ βPFA aq 3 ½PFAaq H PFA 3 ½PFAorg
ð11Þ
J PFA org ¼ βPFA org 3 ½PFAorg ½PFAorg
ð12Þ
Mass transfer related to hydrogen peroxide and water have been neglected because these components are poorly soluble in oil and are always in excess with respect to formic and performic acids. HJ appearing in relations 912 are the partition coefficients defined as [C]*aq/[C]*org and calculated by using the SPARC online calculator.11 The following equations represent the 8762
dx.doi.org/10.1021/ie2016174 |Ind. Eng. Chem. Res. 2012, 51, 8760–8767
Industrial & Engineering Chemistry Research
ARTICLE
obtained values and related dependence on the temperature: 1=H FA ¼ 9:0 10 7 3 T 2P 5:0 10 5 3 T P þ 0:0046,
R 2 ¼ 0:9782
ð13Þ
1=H PFA ¼ 4:0 10 6 3 T 2P 4:0 10 5 3 T P þ 0:0402,
R 2 ¼ 0:9899
ð14Þ
1=H H2 O2 ¼ 9:0 10 8 3 T 2P 6:0 10 6 3 T P þ 0:0005,
R 2 ¼ 0:9592
ð15Þ
As water is present always in a great excess, its concentration in oil can be considered constant and equal to an average value of 0.28 g of water for 100 g of oil at 70 °C. From the results reported in the already mentioned previous work10 we can observe that solubilities are in the order PFA > FA > H2O2. For what concerns βki, this parameter represents the product between the mass transfer coefficient and the interfacial surface area for the component k in the phase i. At this purpose, on the basis of experimental evidence, we have separated the β values in two different contributions (kL and asp), by considering a dependence of the specific interfacial area on both the overall volumetric flowrate and the ratio between the oxidizing mixture and oil according to a polynomial correlation of the type: ! !2 Q Q Q Q aq aq tot tot aaq þ δ2 3 sp ¼ δ1 3 ref ref Qorg 3 Qtot Qorg 3 Qtot þ δ3 3
Qaq Qtot ref Qorg 3 Qtot
!3 ð16Þ
3 where Qref tot is equal to 1 cm /minute. Concerning the specific surface area, we have observed that the most important effect, influencing both the conversion and the oxirane selectivity, is related to the overall volumetric flow-rate. By applying relation 16 we observed that δ2 and δ3 contributions can be neglected. With the specific interfacial area being the ratio between the interface and the volume of the relative phase, we can correlate the specific areas of respectively aqueous and organic phase as follows:
aaq sp org
asp
¼
Qorg Qaq aq a f aorg sp ¼ Qaq Qorg 3 sp
ð17Þ
the physical properties contained in that relation (viscosity, molecular weight, etc.) we have tried to evaluate the ratio (for formic and performic acids) between the mass transfer coefficients in respectively water and oil. The evaluation of this ratio clearly reduces the number of adjustable parameters being the mass transfer in water related to the one in oil for the same substance. The correction coefficient σ has been introduced, because, in this approach we considered a pseudobinary mixture. On the contrary, the reaction mixture is not composed simply of oil and water but is much more complex. We assumed a constant value of σ along the reactor although we know that the viscosity of the epoxidized soybean oil is much greater than the one of the fresh oil. This approximation is reasonable in our case because conversion and yields were relatively low. All chemical and physical properties have been considered at an average temperature as a reasonable approximation. However, we have checked that this approximation does not significantly affect the agreement between calculated and experimental results. Starting from expressions 912, it is possible to calculate the concentrations of all the mentioned compounds at the interface, by imposing that in steady-state conditions, the following balance is always valid: J k aq 3 Q aq ¼ J k org 3 Q org
ð19Þ
As a consequence the following expressions for calculating the interface concentrations can be derived:
½FAorg ¼
½PFAorg ¼
Qaq 3 βFA aq 3 ½FAaq þ Qorg 3 βFA org 3 ½FAorg Qorg 3 βFA org þ Qaq 3 βFA aq 3 HFA
Qaq 3 βPFA aq 3 ½PFAaq þ Qorg 3 βPFA org 3 ½PFAorg Qorg 3 βPFA org þ Qaq 3 βPFA aq 3 HPFA
ð20Þ
ð21Þ
Concerning the heat transfer, it must be considered that as a consequence of the epoxidation reaction, the oil phase composition changes along the reactor’s axis. As a matter of fact, epoxidized soybean oil is much more viscous than the unreacted substrate (450 Cp for epoxidized soybean oil and 110 Cp for soybean oil at 25 °C) and we can foresee that the overall heat exchange coefficient U will change with the conversion degree, considering an approximated linear dependence of U with the conversion: U ¼ Uo þ α 3 X DB
ð22Þ
The validity of this relation has already been verified in a previous work.10 3.2.1. Mass and Energy Balances. The balance equations for each component can be written as follows: Aqueous phase:
Moreover, the kL values for two immiscibile liquid phases can be roughly estimated by using the Wilke and Chang correlation.20 Starting from their expression, it is possible to write the ratio between the mass transfer coefficients related to the two liquid films, as follows: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ϕaq 3 PMaq μorg kL, aq ¼ σ3 ð18Þ kL, org ϕorg 3 PMorg 3 μaq where jaq = 2.6 for water, jorg = 1 for oil, μ is the viscosity, and σ is a correction coefficient determined from the experimental results. The WilkeChang relation is useful to estimate, in a first approximation, the mass transfer coefficients of the components in respectively the aqueous and the oil phase. Then, considering 8763
d½H2 O2 aq ra rb ¼ dV Q tot
ð23Þ
d½H2 Oaq þra þ rb ¼ dV Q tot
ð24Þ
d½FAaq rb JFA aq ¼ dV Q tot
ð25Þ
d½PFAaq þrb J PFA aq ¼ dV Q tot
ð26Þ
dx.doi.org/10.1021/ie2016174 |Ind. Eng. Chem. Res. 2012, 51, 8760–8767
Industrial & Engineering Chemistry Research
ARTICLE
Table 2. Average Values of Density, Specific Heat, and Reaction Enthalpy density (g cm3) FH2O2 = 1.44
specific heats (cal g1 °C1)
Table 3. Kinetic Constants Values Determined in Previous Work from Runs Performed in Fed-Batch Conditions10
enthalpy (cal mol‑1)
cpH2O2 = 0.622
ΔHr = 55000
Foil = 0.89 FA = 1.16
cpoil = 0.50 cpA = 0.531
ΔHox = 23440 ΔHdeg = 2100022
FH2O = 0.985
cpH2O = 1
parameter
7
( 1.6 10
.4
kb Keq
6.56 10 ( 2.0 10 5.17 ( 0.8
2
kcmono
Organic phase: ð27Þ
d½PFAorg rc ðiÞ þ J PFA org ¼ dV Q tot
ð28Þ
d½DBorg ðiÞ rc ðiÞ ¼ dV Q tot
ð29Þ
d½Epoxorg ðiÞ þrc ðiÞ rd ðiÞ ¼ dV Q tot
ð30Þ
d½DEGH org ðiÞ þrd ðiÞ ¼ dV Q tot
ð31Þ
Concerning the energy balance, we have considered, as a first approximation, that the heat released by reaction is instantaneously distributed on all the mass of the reaction. Therefore, the heat balance can be interpreted with the following relation: qaccumulation ¼ þ qreaction qexchange ½¼ cal 3 L1 min1
ð32Þ
and the terms of the heat balance can be written as follows: dTP qaccumulation ¼ cp 3 Wtot 3 dV
ð33Þ
qreaction ¼ ½ΔHr 3 ð rc ðiÞÞ þ ΔH deg 3 ð rd ðiÞÞ þ ½ΔH ox 3 ð ra Þ
ð34Þ
qexchange ¼ U 3
A ðTP TJ Þ V3
ð35Þ
The thermodynamic properties of the different components the reaction are reported in Table 2. 3.2.2. Experimental Runs Simulation. The kinetic runs, reported in Table 1, have been simulated with the described model using the same intrinsic kinetic parameters determined in a previous work10 for describing fed-batch and pulse-fed-batch runs. Those parameters are reported, as a useful reminder, in Table 3. The only adjustable parameters of the model for describing the whole set of continuous runs are the mass and heat transfer coefficients (kL, σ, and Uo) with these being dependent on the adopted fluid dynamic conditions, and the interfacial area (δ1) affecting both the overall mass transfer coefficient β and the ringopening reaction rate. It is important to point out that the adopted fluid dynamic conditions can greatly affect the ringopening reaction rate through the interfacial surface area as it can be deduced by observing the rate law (8). In our previous work10
3 1
unit 1.5,aq
L
2,org
L -
mol1.5 min1 mol2 min1
2.72 ( 0.5
Lorg mol1 min1
tri
133.9 ( 9.8
Lorg mol1 min1
Eab
11.39 ( 0.6
kcal mol1
Eac
24.89 ( 0.2
kcal mol1
Ead
8.86 ( 0.4
kcal mol1
kc
d½FAorg þrc ðiÞ þ J FA org ¼ dV Q tot
1.04 10
ka 21
value ( confidence interval
we have considered as kinetic parameter the product kd(i) 3 aaq sp , while in this paper we have tried to evaluate both the contributions of kd(i) and separately aaq sp being the first constant independent of the flow rate and the second dependent according to the relation 16. The mathematical regressions have been performed on three experimental points for each run that are IN, ON, and the outlet temperature of the reaction mixture. For example, for reactor 1 we obtained 15 experimental points that we have described with 5 parameters, and for reactor 2 we have processed 36 experimental data with the same set of parameters, except Uo, because the two reactors were characterized by different wall thicknesses. These parameters are reported in Table 4. In Table 1, together with the experimental data, the calculated values of both IN and ON are reported for comparison. As it can be seen, the agreement is satisfactory considering all the approximations introduced and the passage from fed batch to continuous reactor for such an exothermic reaction. Therefore, the proposed biphasic kinetic model and related parameters can be considered reliable. An example of the calculated profiles along the reactor for the ON, IN, and temperature has been reported in Figure 2a, whereas in Figure 2b the formic and performic profiles along the reactor, in the two phases, are reported. This example is referred to the run 7 of Table 1 that is the run with the highest reached conversion. It is interesting to show the role of the overall flow rate and of the oxidant/oil ratio on, respectively, the double bond conversion (Figure 3) and the yield (Figure 4). As it can be seen the effect of the oxidant/oil ratio is relevant in both cases, while the effect of the overall flow rate is less relevant. 3.3. Discussion. The performed continuous runs have confirmed the validity of the biphasic kinetic model developed in a previous work10 for describing many different fed batch experimental runs. As a matter of fact, we have shown that continuous runs can be described by using the same rate laws and kinetic parameters with the exception of (i) the mass and heat transfer coefficient that must be adapted to the new fluid dynamic conditions and (ii) the ring-opening reaction for which rate depends on the interfacial area that is relatively high in the described tubular reactors containing small stainless steel spheres as static mixer elements. This occurs for the intense local micromixing, characterizing these reactors, leading to a high interfacial surface area. By comparing, for example, β and the product kd(i) 3 aaq sp , both containing the interfacial area, in respectively the continuous and fed batch reactor, this last used in a previous work10 it resulted that the continuous reactors show an interfacial area that is from 3 to 18 times the ones of the fed-batch stirred reactor. In the continuous reactors the interfacial area changes 8764
dx.doi.org/10.1021/ie2016174 |Ind. Eng. Chem. Res. 2012, 51, 8760–8767
Industrial & Engineering Chemistry Research
ARTICLE
Table 4. Adjustable Additional Parameters for Describing Continuous Runs parameter
reactor 1
reactor 2
unit
U0
9900 ( 40
6000 ( 50
cal/(m2 min °C)
α
32.0 ( 2.1
32.0 ( 2.1
cal/(m2 min °C)
δ1
580 ( 50
641 ( 40
cm2/cm3
σ
3.06 ( 0.36
3.06 ( 0.36
-
kL,aq
1.26 ( 0.13
1.26 ( 0.13
cm/min
kL,org
0.56 ( 0.04
0.56 ( 0.04
cm/min
kdmono
2.42 ( 0.22 105
2.42 ( 0.22 105
kdtri
3.35 ( 0.15 103
3.35 ( 0.15 103
(1/1000) 3 L2,aq 3 mol1 3 cm2 3 min1 (1/1000) 3 L2,aq 3 mol1 3 cm2 3 min1
Figure 2. Simulation of both aqueous and oil phase compositions related to Run 7 of Table 1. In plot A are reported the IN, ON, and temperature profiles. In plot B are reported the profiles of the main components in the two phases.
with the overall flow rate, while in the stirred fed batch reactor the area was always constant because of the constant stirring rate. This last point is very important, because the increase of the interfacial area is highly positive for the initial conversion, because, initially the epoxidation reaction is very fast, but in the meantime the ringopening reaction rate, strongly affected by the interfacial area, too, limits the selectivity of the reaction in particular for a prolonged
residence time. This behavior is clearly shown in Figure 5, where the simulations of the ON and IN at different temperatures (70, 80, 90 °C) are reported, by assuming the oil/oxidant volumetric ratio equals 8:2.5, a catalyst concentration of 0.19 mol/Laq, and isothermal conditions. As it can be seen, by operating with one of the described packed bed tubular reactor (reactor 2), the target values required by the industrial operation 8765
dx.doi.org/10.1021/ie2016174 |Ind. Eng. Chem. Res. 2012, 51, 8760–8767
Industrial & Engineering Chemistry Research
Figure 3. Dependence of the double bonds conversion on the overall flow rate and the ratio between the oxidant solution and oil. The simulation has been performed at 80 °C.
Figure 4. Dependence of the oxirane yield on the overall flow rate and the ratio between the oxidant solution and oil. The simulation has been performed at 80 °C.
are never reached. This occurs because the ring-opening reaction rate is too high and lowers the oxirane yields. Another observation is that at higher temperature the epoxidation yields are higher at lower residence time. Therefore, in order to obtain the best performances it is necessary to operate for a short time at high temperature and interface area, then it is necessary to lower the ring-opening reaction rate by limiting the available interfacial area. Hence, a two-step reaction would be adopted using two different reactors, one characterized by a high interface area and another with a lower interface area using, in this last case, a less efficient static mixer. In Figure 6 is reported the simulation of two reactors in series in which the first reactor is similar to the previously described reactor 2 but imposing 10 min of residence time, plus another reactor having an interfacial area of 1 /10 with respect to the first one, both operating at 90 °C. In this case, as it can be seen, the industrial target values are reached for both ON (≈ 6.5) and IN (e 2.0) in about 45 min instead of 68 h, the time required in industry for the fed-batch operation.
ARTICLE
Figure 5. Simulation of continuous reactors working at different temperature, imposing the same β value found in the experimental runs and an oil/oxidant flow-rate ratio of 8/2.5.
Figure 6. Simulation of two continuous reactors put in series. The first reactor works imposing a residence time of 10 min and the same β value found in the experimental runs, while the second reactor works with a β value corresponding to 10% with respect to the first reactor. Both the reactors work with an oil/oxidant flow-rate ratio of 8/2.5 and a temperature of 90 °C.
4.0. CONCLUSIONS In the present work a biphasic kinetic model with related parameters, developed for describing the soybean oil epoxidation reaction, performed in fed-batch reactors, has been tested by applying it to describe also the kinetic runs performed in continuous tubular reactors packed with stainless steel spheres, used as static mixer elements, at the scope of favoring an intense local micromixing. This model could be used for modeling continuous reactors and would be the base for promoting the process intensification. It has been seen that by putting two reactors with different interfacial areas in series can achieve the required industrial specimens in less than 45 min instead of the actual 68 h required for the pulse-fed-batch operation. ’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]. 8766
dx.doi.org/10.1021/ie2016174 |Ind. Eng. Chem. Res. 2012, 51, 8760–8767
Industrial & Engineering Chemistry Research
’ ACKNOWLEDGMENT This study was funded by EC VII Framework Programme CP-IP 228853-2 COPIRIDE. ’ LIST OF SYMBOLS CdC/DB = Unsatured group (Double bond) FA = Formic acid PFA = Performic acid Epox = Epoxide group DEGH = Decomposed oxirane group by H (proton) Nu = Nucleophilic molecule ri = Reaction rate [mol L1 min1] ki = Kinetic constant (the units depends on the reaction) Jki = Mass transfer rate for the component k in the phase i [mol L1 min1] i βk = Product between the mass transfer coefficient and the interfacial surface area for the component k in the phase i aspi = Specific interfacial area for the phase i [cm2/cm3] δ = Polynomial parameter for the specific interfacial area [cm2/cm3] kL,i = Mass transfer coefficient for the phase i [cm/min] Hj = Partition coefficients defined as [C]aq*/[C]org* σ = Wilke and Chang correction parameter [-] U = Global thermal exchange coefficient [cal m2 min1 °C1] Uo, α = Parameters that express the dependence of the global thermal exchange coefficient with the double bonds conversion degree [cal m2 min1 °C1] A = Exchange area [m2] V = Reactor’s volume [L] Qi = Volumetric flow-rate: aq = aqueous phase, org = organic phase, tot = overall [cm3 min1] ref Qtot = Reference overall flow-rate imposed equal to 1 cm3 min1 Wtot = Overall mass flow-rate [g min1] cP,i = Specific heat capacity at constant pressure of the species i [cal g1 °C1] Ti = Temperature of the fluid i [°C] Tj = Temperature of the jacket [°C] TIN = Inlet temperature of the tubular reactor [°C] TOUT = Outlet temperature of the tubular reactor [°C] PMi = Molecular weight of the i-th compound [g mol1] q = Heat flow [cal min1] ΔHi = Enthalpy of the i-th reaction [cal mol1] XDB = Double bonds conversion degree [-] ’ SUBSCRIPTS AND SUPERSCRIPTS H Proton FA Formic acid PFA Performic acid aq Aqueous phase org Organic phase J Jacket P Process r Epoxidation reaction ox H2O2 decomposition deg Ring-Opening reaction (degradation) IN Inlet OUT Outlet
ARTICLE
(2) Fenollar, O.; Garcia, D.; Sanchez, L.; Lopez, G.; Balart, R. Optimization of the curing conditions of PVC plastisols based on the use of an epoxidized fatty acid ester plasticizer. Eur. Polym. J. 2009, 45, 2674–2684. (3) MYTHEN SpA Company; Private communication. (4) Campanella, A.; Fontanini, C.; Baltanas, M. A. High yield epoxidation of fatty acid methyl esters with performic acid generated in situ. Chem. Eng. J. 2008, 144, 466–475. (5) Gan, L. H.; Goh, S. H.; Ooi, K. S. Kinetic Studies of Epoxidation and Oxirane Cleavage of Palm Olein Methyl Esters. J. Am. Oil Chem. Soc. 1992, 69, 347–351. (6) Sinadinovic-Fiser, S.; Jankovic, M.; Petrovic, Z. S. Kinetic of in-situ of soybean oil in bulk catalyzed by ion exchange resin. J. Am. Oil Chem. Soc. 2001, 78, 725–731. (7) Rangarajan, B.; Havey, A.; Grulke, E. A.; Culnan, P. D. Kinetic parameters of a two-phase model for in situ epoxidation of soybean oil. J. Am. Oil Chem. Soc. 1995, 72, 1161–1169. (8) Campanella, A.; Baltanas, M. A. Degradation of the oxirane ring of epoxidized vegetable oils in liquid-liquid systems: I. Hydrolysis and attack by H2O2. Lat. Am. Appl. Res. 2005, 35, 205–210. (9) Campanella, A.; Baltanas, M. A. Degradation of the oxirane ring of epoxidized vegetable oils in liquid-liquid systems: II. Reactivity with solvated acetic and peracetic acid. Lat. Am. Appl. Res. 2005, 35, 211–216. (10) Santacesaria, E.; Tesser, R.; Di Serio, M.; Turco, R.; Russo, V.; Verde, D. A biphasic model describing soybean oil epoxidation with H2O2 in a fed-batch reactor. Chem. Eng. J. 2011, 173 (1), 198–209. (11) Carreira, L. A., Hilal, S.; Karickhoff, S. W. Quantitative Treatment of Solute/Solvent Interactions. Vol. 1 in Theoretical and Computational Chemistry Series; Politzer, P., Murray, J. S., Eds.; Elsevier Publishers: New York, 1994; http://sparc.chem.uga.edu/sparc/. (12) Hessel, V.; Cortese, B.; De Croon, M. H. J. M. Novel Process Windows. Chem. Eng. Sci. 2011, 66 (7), 1426–1448. (13) Cortese, B.; De Croon, M. H. J. M.; Hessel, V. High temperature epoxidation of soybean oil in flow - speeding up elemental reactions wanted and not wanted. Ind. Eng. Chem. Res. 2011No. 10.1021 ie 200868w. (14) Norme Grassi e Derivati NGD C 32; Stazione Sperimentale Oli e Grassi, Ed.; Milan, Italy, 1976. (15) Deaborne, E. C.; Fuoss, R. M.; Mackenzie, A. K.; Shephered, R. G., Jr. Epoxy Resins from Bis-, Tris-, and Tetrakis-Glycidyl Ethers. Ind. Eng. Chem. 1953, 45 (12), 2716. (16) Paquot, C.; Hautfenne, A. Commission on Oils Fats and Derivatives: Standard Methods for the Analysis of Oils, Fats and Derivatives; IUPAC, Applied Chemistry Division; Blackwell Scientific Publications: London, 1987. (17) Kolthoff, I. M.; Sandel, M. Treatise Analytical Chemistry; John Wiley & Sons Inc.: New York, 1993; Vol. 2, p 888. (18) La Scala, J.; Wool, R. P. Effect of FA Composition on epoxidation kinetics of TAG. J. Am. Oil Chem. Soc. 2002, 79, 373–378. (19) Whitman, W. G. Preliminary Experimental Confirmation of the Two-Film Theory of Gas Absorption. Chem. Metall. Eng. 1923, 29, 146–148. (20) Wilke, C. R.; Chang, P. Correlation of diffusion coefficients in dilute solutions. AIChE J. 1955, 264–270. (21) Jones, C. W. Applications of Hydrogen Peroxide and Derivatives Series; Clean Technology Monographs; RSC: Cambridge, 1999; pp 14. (22) American Chemistry Council. http://www.ethyleneoxide.com.
’ REFERENCES (1) Babich, M. A.; PhD Thesis, Review of Exposure end Toxicity Data for Phthalate Substitutes; U.S. Consumer Product Safety Commission, 4330 East West Highway, Bethesda, MD 20814, 2010. 8767
dx.doi.org/10.1021/ie2016174 |Ind. Eng. Chem. Res. 2012, 51, 8760–8767
