Vous êtes sur la page 1sur 10

3160

Ind. Eng. Chem. Res. 2010, 49, 31603169

Production of Biodiesel at the Kinetic Limit in a Centrifugal Reactor/Separator


Joanna McFarlane,* Costas Tsouris, Joseph F. Birdwell, Jr., Denise L. Schuh, Hal L. Jennings, Amy M. Palmer Boitrago, and Sarah M. Terpstra
Oak Ridge National Laboratory, PO Box 2008, Oak Ridge, Tennessee 37831-6181

The kinetics of the transesterication of soybean oil has been investigated in a centrifugal contactor reactor/ separator at temperatures from 45 to 80 C and pressures up to 2.6 bar. The high shear force and turbulent mixing achieved in the contactor minimized the effect of diffusion on the apparent reaction rate, and hence it could be assumed that the transesterication rate was limited by the reaction kinetics. The yields of product methyl esters were quantied using gas chromatography ame ionization detection (GC-FID), infrared (IR) spectroscopy, proton nuclear magnetic resonance (H1NMR), and viscosity measurements and typically were found to achieve 90% of complete conversion within 2 min. However, to meet American Society for Testing and Materials (ASTM) specications with one pass through the reactor, a minimum 22-min residence time at 80 C was needed. Performance was improved by stepwise processing, allowing separation of byproduct glycerine and injection of additional small aliquots of methanol at each step. The chemical kinetics was successfully modeled using a three-step mechanism of reversible reactions, and employing activation energies from the literature, with some modication in pre-exponential factors. The mechanism correctly predicted the exponential decline in reaction rate as increasing methyl ester and glycerine concentrations allow reverse reactions to occur at signicant rates.
Introduction A variety of alternatives to petroleum are being developed as vehicular fuels to reduce dependence on gasoline and diesel fuel.1 Biodiesel, a product of the transesterication of fats and oils, is commercially available for blending with standard diesel fuel. Industrial production of biodiesel from oil of low fattyacid content follows homogeneous base-catalyzed transesterication, a sequential reaction of the parent triglyceride with an alcohol, usually methanol, into methyl ester and glycerol products. Although the price of diesel fuel has increased, economical production of biodiesel is a challenge because of (1) the increasing price of oil feedstocks and reagent methanol, (2) a distributed supply of feedstocks that reduces the potential for economies of scale, (3) processing conditions that include pressures and temperatures above ambient, and (4) multiple processing steps to reduce contaminants to ASTM specication D6751 limits.2,3 Commercially produced biodiesel is made in batch reactors to achieve high enough yields. Producers and investigators have focused on the kinetics of transesterication to see if conversions to methyl ester are limited by mass transfer effects or by slow kinetics.4,5 Much of the cost of biodiesel production is related to the conversion of the oil to the methyl ester; hence, the process would be improved by the use of a continuous rather than batch process, with energy savings generated by combined reaction and separation, online analysis, and reagent methanol added by titration as needed to produce ASTM specication grade fuel. Hence, Oak Ridge National Laboratory (ORNL) has been investigating the use of a centrifugal reactor/separator to increase mass transfer and reduce the transesterication processing time to that limited by chemical kinetics. Centrifugal contactors were initially developed as solvent extraction devices for use in actinide recovery from spent nuclear reactor fuel in place of mixer-settlers.6,7 Because of its ease of operation, rapid attainment of steady state, high mass transfer
* To whom correspondence should be addressed. Tel.: 865-574-4941. Fax: 865-241-4829. E-mail: mcfarlanej@ornl.gov.

and phase separation efciencies, and compact size, the centrifugal contactor was chosen for intensication of the biodiesel production process. In these experiments the immiscible liquid reagents were introduced into the reactor as two separate feed streams: a mixture of methanol and base catalyst, and vegetable oil. Following reaction, the immiscible glycerol and methyl ester products were separated by centrifugal force in the same vessel. A commercial unit was modied to increase the residence time from a few seconds to a few minutes by achieving hold-up in the mixing zone.8,9 The advantage of this device over previously described centrifugal reactors (e.g., by Peterson and co-workers10) is that the increased residence time is integrated into the unit design and additional delay loops and processing are not required. Through understanding of the chemical kinetics of transesterication in the centrifugal reactor/separator, the objective was to develop an efcient, high-yield, continuous process for basecatalyzed biodiesel production. The model developed to describe the data can be used to optimize the process for maximum throughput or minimum residence time and minimal use of methanol, key factors in the viability of biodiesel production from triglyceride. Experimental Section Centrifugal Contactor. Transesterication of commercial grade soy oil was carried out in a modied centrifugal contactor. The distinctive feature of the centrifugal contactor apparatus is the use of a rotor within a stationary cylinder (the housing) to accomplish the intimate mixing of two immiscible liquids to produce a dispersion, which is then subjected to a centrifugal force used to separate the two liquid phases.11 In solvent extraction applications, the generation of a nely divided dispersion promotes the transfer of one or more solutes from one liquid phase into another by maximizing interfacial area. Similarly, in the case of a chemical reaction between immiscible reactants, the creation of a dispersion minimizes the diffusion pathway of reactants to the interfacial boundary. In both cases,

10.1021/ie901229x 2010 American Chemical Society Published on Web 03/02/2010

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010

3161

Figure 1. Schematic showing uid ows in a centrifugal contactor. The reagents can be introduced separately through one or more of the solution inlets. The ow-through apparatus includes additional side ports, allowing recirculation of the mixture through the mixing zone.

dispersion formation is accomplished by means of Couette mixing, in which shear forces are created in the annulus between a stationary outer cylinder and, in this case, a 5-cm-diameter rotating inner cylinder, as shown in Figure 1.8 As the dispersion ows downward inside the housing, it enters the rotor (i.e., the separation zone) through an opening located at the bottom end of the rotor at the centerline. Angular inertia created by the rotor causes the liquid to be pushed against the inner rotor wall, with uid slip being prevented by axial vanes that are attached to the inner wall. The resulting centrifugal force coalesces the dispersion into its constituent phases, which become layered axially based on their relative densities. Weirs located at the upper end of the rotor control both the ow through the rotor and the axial location of the interface between the liquid phases; hence the separation of liquid phases passing through the rotor is based on density difference. Although the main goal of the project was to evaluate a continuous production of methyl ester in the reactor/separator, kinetics studies were also carried out in batch mode to facilitate sampling at frequent intervals, and batch experiments were undertaken at above ambient pressure in the centrifugal contactor reaction vessel to increase the reaction yield. Hence, transesterication kinetics was studied in three congurations: (1) in a stirred glass reaction vessel to enable visual monitoring of the uids and in a 200-mL centrifugal reactor/separator run in (2) batch and (3) ow-through mode. Materials. Reagents were soybean oil, methanol, and 30% methanol/methylate (all from Nu-Energie, LLC, http://www.nuenergie.com) at volumetric phase ratios from 4:1 to 6:1 oil/ methanol, or mole ratios with an excess of methanol, from 5.62 to 4.71. These phase ratios were used for both the static and ow-through experiments. The benchmark methanol-to-oil mole ratio was the same as the one used at the Nu-Energie plant, that is, 4.81. All reagents were used without further purication to simulate conditions at the manufacturing facility. The acid number of the feedstock oil was determined by titration with alcoholic KOH to be 0.06 ( 0.01 mg KOH/g. In the batch experiments, the oil was preheated to the desired temperature before the addition of the methanol/methylate mixture. Kinetics Experiments in Stirred Glass Vessel. A magnetically stirred (100 rpm) glass reaction vessel (500 mL) was

heated to temperatures between 45 and 60 C using ow from a VWR thermostatted bath through a temperature-controlled water jacket. Temperatures were measured using a Type K thermocouple. Reaction mixtures of total volumes between 100 and 200 mL, methanol-to-oil molar phase ratios of 4.8 to 5.3, were heated separately, then combined in the reaction vessel at the initial time of the experiment. Sampling from the reaction vessel was done by pipetting a portion from the mixture into vials containing 1 mL of 1 M HCl to stop the transesterication process. Sampling was performed at several intervals after injection of the methanol: 15, 30, 45, 60, 90, 120, 180, 240, 300, 420, 600, and 900 s, to within (5 s. Samples were analyzed by gas chromatography (GC). Flow-Through Experiments in Centrifugal Contactor. Base-catalyzed biodiesel synthesis and simultaneous separation of methyl ester and glycerol products were carried out in a modied centrifugal contactor.9 The reactor was operated in continuous mode at temperatures from 50 to 60 C, with resistive heating controlled with a Thermolyne controller (model 45500). The methanol-to-oil molar phase ratio, nominally 4.87, was controlled through the volumetric ow rates pumped (Fluid Metering, Inc.) to the reactor: the oil ow rate ranging from 50.0 to 150.0 ( 0.5 mL min-1, and the methanol with base catalyst ow rate ranging from 10.0 to 30.0 ( 0.5 mL min-1. Samples of product were taken from the lighter-ow and heavier-ow discharges of the reactor/separator. The total volumetric ow was set so as to obtain a mean reactor residence time of 1-3 min for a total volume of 180 mL. A time of 5 min was needed to establish stable ows through the contactor as the glycerine phase was only one-tenth of the volume of the methyl ester phase. Samples were taken at intervals between 1 and 20 min after ow was initiated. These samples were also acidied by contact with 1 M HCl to prevent further reaction and were analyzed using GC. Observations made during testing using a contactor with a transparent Lucite housing and dyed feed solution indicated that the residence time distribution was broad. Although the desired average residence time was set as 1 min, a small amount of material exited the apparatus less than 10 s after introduction because of pulsing in the mixing zone. This effect was reduced in later experiments by lengthening the contactor housing.

3162

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010

Elevated Pressure Tests in Contactor Done with Single Charge of Reagents. An important performance goal for biodiesel production as a fuel is the achievement of an ASTMdened level of conversion from the triglycerides. Tests were performed at 50 and 60 C as well as temperatures above the boiling point of methanol in a sealed vessel. For tests at ambient pressure, the procedure followed was very similar to that described for tests in the stirred glass vessel. In testing performed at elevated (i.e., above ambient) pressure, however, the contactor had to be operated in batch mode because pressurizable feed and product tanks were not available. The oil was heated in the contactor to the desired temperature, 80 C, under a blanket of about 1.5 bar N2 with the rotor at full speed. Methanol and base were then injected under pressure, at 2 bar, into the housing. During the reaction, the pressure in the vessel increased, usually to 2.6 bar. In these tests, the device was congured so as to prevent transfer of uids from the reaction zone into the separation zone. Reaction products were removed directly from the reaction zone at the completion of each test and analyzed using infrared spectroscopy (IR), viscosity measurements, and nuclear magnetic resonance spectroscopy (NMR). The lengths of batch-mode runs were varied to determine the progress of the reaction under pressures up to 2.6 bar and temperatures up to 80 C. In some cases the glycerine was separated from the oil/methyl ester mixture, the latter being again reacted in the contactor to see if removing the byproduct could improve the yield of methyl ester. Chemical Composition Analysis. Gas chromatography was used to determine changes in composition as a function of time; however, because the sample preparation removed the product glycerine and the unreacted triglyceride peaks were not detectable, these data could not be used for quantication of the fraction of bound glycerine. For these measurements, IR, viscosity, and NMR analyses were used. For the kinetics data, reactions were halted by contact of the sample with an equal volume of 1 M HCl. The samples were centrifuged, and the organic layer was separated and rinsed with twice the volume of deionized water to remove much of the unreacted methanol. Although most of the methanol was associated with the glycerine phase, a signicant fraction remained in the methyl ester, and had to be removed prior to analysis. Reaction products were analyzed by GC FID, with a HewlettPackard 5890 II GC. The analysis procedure followed ASTM D6854.12 The precision in the measured peak areas was estimated to be (10% from the injection of reference standards. Infrared (IR) analysis of the products was performed using both a Thermo-Nicolet Magna-IR 560 Fourier transform infrared (FTIR) spectrometer and a Bruker FTIR spectrometer. The Bruker analyses were conducted off-site by using the Cognis model QTA system (http://www.cognis.com/products/ Business+Units/AgroSolutions/Grain+Analysis/) through NuEnergie LLC, and were reported with a precision as low as ( 0.01 wt % in the triglyceride, diglyceride, and monoglyceride components of the bound glycerine fraction. (Bound glycerine refers to the wt% of the glycerine backbone in the acylglyceride molecule.) However, the accuracy of the results was estimated to be (0.1 wt % as a result of uncertainties introduced during sample preparation. For instance, it was found that the analytical results were biased if unreacted methanol was not removed by washing either with deionized water or with 1 M HCl. Changes in viscosity, measured with a Brookeld model DV-E viscometer, were correlated with reaction yield and were used to monitor progress toward the ASTM standard in bound

Figure 2. Viscosity as a function of bound glycerine in biodiesel at 25 C. The weight fraction of bound glycerine (9) is not identical to the amount of unreacted oil ([) as the latter does not include the diglycerides and monoglycerides. Table 1. NMR Peak Shifts Observed from Samples of Methyl Esters and Soybean Oil shift relative to TMS 0.8 1.3 2.0 2.3 2.8 3.6 3.7 4.1-4.3 5.3

splitting multiplet doublet multiplet triplet triplet singlet singlet multiplet multiplet

assignment Terminal CH3 -CH2-CH2- beta to CdO -CH2- alpha to CdO -H to double bond free-glycerol H CH3-OH on glycerine backbone R-H to double bond

sample soybean oil, methyl soybean oil, methyl soybean oil, methyl soybean oil, methyl soybean oil, methyl methyl ester methyl ester soybean oil soybean oil, methyl ester ester ester ester ester

ester

glycerine concentration (<0.24 wt %). Although the main purpose for conversion of the triglyceride to methyl ester is to reduce viscosity, this type of analysis is not standard for monitoring reaction progress and so is discussed here in some detail. A calibration curve for viscosity measurements was developed using the Nu-Energie, LLC, commercial methyl ester product mixed with unreacted-soy oil, washed, and heated to remove any residual methanol in the same manner as the samples from the experiments (Figure 2). All viscosities were measured at 25 C, as established by using a water jacket on the viscometer fed by a thermostatted water bath. The biodiesel calibration samples were found to degrade over time. Hence, the overall accuracy of the viscosity analysis was estimated to be (1 wt % unreacted oil. Samples were analyzed by H1NMR in a Bruker Avance-400 NMR after dissolution in CDCl3 (Aldrich, lot no. 00808TH, 0.03% v/v tetramethylsilane, TMS). Peak shifts are given in Table 1 and compared to the Sadtler compilation.13 Because of the variety of constituent fatty acid chains, the peaks were generally quite broad, with the sharpest being the methyl group on the methyl-ester product, at ) 3.7. The peaks that were uniquely attributed to the acylglyceride were present at a shift between 4.1 and 4.3 ppm. A small sidebar on the 3.7 peak could be attributed to hydrogen on the C-O-H of the glycerine backbone of the partially reacted acylglyceride. Results The focus of the study was to demonstrate the continuous production of biodiesel in the reactor/separator. Samples from the ow-through contactor experiments (phase ratios ranging from 4.1 to 4.9 mols methanol-to-oil) were analyzed using gas

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010

3163

Figure 3. Typical gas chromatogram of transesterication product from either centrifugal reactor or open stirred reaction vessel. This particular analysis was done on product from an open stirred vessel after 1-min reaction time, 60 C, and methanol-to-oil mole ratio of 4.9. Unlabeled peaks include the solvent (4 min) and free glycerine (8.6 min).

chromatography (Figures 3 and 4) and IR spectroscopy. Figure 3 shows a typical gas chromatograph for samples taken from a methanolysis reaction. Gas chromatographic data from ow through experiments are shown in Figure 4, where results are presented in a bar chart showing the analysis of ows from the heavy and light phase ports of the contactor, taken by summing GC peak areas corresponding to C16, C18, C20, C22, and C24 methyl esters. Figure 4 shows the relative change in composition as a function of time after initiating ow through the contactor, hence the units on the ordinate are arbitrary. The gas chromatograms show signicant production of methyl esters with little residual acylglyceride. However, results from the IR analysis indicate some contamination of the methyl ester with oil. This was thought to have occurred because some of the reagents experienced a residence time much shorter than that predicted by the average ow rate through the reactor. A fairly broad distribution in residence time was conrmed by visualization experiments in the clear-housing contactor, where some of the dyed solution passed through the reactor in a few seconds. Hence, although qualitative proof-of-principle demonstration of the reactor/separator was obtained in these experiments, they did not give data that could be used to develop or conrm a kinetic model. After a couple of minutes of operation, steady ows were established through the reactor outlets. The higher the methanol-to-oil molar ratio and the greater the rotor speed was, which was varied between 3600 and 4200 rpm in the tests, the greater was the reaction yield as demonstrated by steady ows of products from both outlets of the contactor. With the exception of the ow at the mole ratio of 4.9 and rotor speed of 3600 rpm, all of the tests indicated steady ows being achieved within 5 min. In Figure 4, it can be seen that it took longer to establish a ow through the heavy side port than the light side port meaning that at rst the ow was mainly methyl ester product. Per unit volume, glycerine was produced at a much lower rate than methyl ester. This fact, coupled with the high viscosity of glycerine, meant that several minutes were required to establish ows through the reactor. In addition, the weir in the dispersion zone was deliberately set so as to produce as pure a methyl ester product as possible, leaving residual unreacted oil in the glycerine stream. All of the factors discussed above suggested

that a quantitative kinetic analysis and comparison with a model be done using well-stirred batch reactors rather than in the owthrough conguration. Typical kinetic results from the glass stirred vessel at 45, 52, and 60 C as analyzed by gas chromatography are shown in Figure 5. In these tests a steep rise in the peak area attributed to the C16 (methyl palmitate) and C18 (methyl oleate, methyl linolate, and methyl linoleate) esters was observed in the rst minute of residence time, and then a leveling off, or even decrease, in the signal occurred. The relative signal intensity was corrected by a naphthalene-d8 internal standard; hence, these data have arbitrary units. As expected, the initial rise in signal was steeper at 60 C than at 45 C because of increasing reaction rates. Because conversion to methyl ester in the reactor appeared to plateau after an initially rapid increase, this behavior suggested that the reaction rate was not limited by mass transfer to the interface. In fact, a slow degradation of product over time could be seen in decreased and broadened methyl ester peaks. Discoloration and darkening was also observed. Thermal degradation is expected to occur through beta scission adjacent to the carbonyl group and the unsaturated bonds in the fatty acid chains.14,15 As the triglycerides that comprise soybean oil are signicantly unsaturated, it is possible that this mechanism of decomposition occurred when the oil was heated for several minutessup to half an hour when including the preheating period before the addition of methanol. Kinetic results from batch ambient-pressure tests in the reactor/separator done at 50 C and at a methanol-to-oil mole ratio of 5.74 are presented in Figure 6. Because the system was not pressurized, sampling was done online through a port at the base of the vessel. Data were taken from IR measurements using the Bruker spectrometer, starting at 3 min after the reagents were rst mixed in the reactor. Infrared data at earlier times were not available because the commercial spectrometer had been calibrated only for acylglyceride concentrations less than 7 wt % triglyceride, 5 wt % diglyceride, and 3 wt % monoglyceride. At a time of 3 min after the reaction was initiated the amount of triglyceride was less than 6 wt % (Figure 6) and decreasing exponentially (Figure 7), and the ratios between the diglycerides and monoglycerides were fairly constant, within the reproducibility of the data, or (0.2 (80 C). These fairly static concentration ratios between the acylglycerides are discussed later in terms of the chemical kinetic model and used to explain the difculty in accelerating the production of biodiesel. In the batch tests, data were not collected beyond 10 min as the increase in reaction yield had tapered off. An experiment was conducted at a slow rotor speed, 3000 rpm at 50 C, and showed a reduced rate of conversion to methyl ester. However, even at higher rotor speeds, the overall reaction yield at temperatures up to 60 C did not go above 90%, indicating that a substantial increase in reaction temperature was needed to reach ASTM grade fuel in a reasonable time. Increased temperature required operation at above ambient pressure to prevent vaporization and loss of the methanol reagent. Results from the pressurized reactor tests from both IR analysis and viscosity measurements are given in Table 2. Conditions of the reaction are also shown in the table, including the maximum pressure and temperature reached. For tests done under pressure, the operating condition that worked best was to slightly pressurize the reactor containing the oil with dry nitrogen as it was heated to approximately 80 C. The pressure increased rapidly after injection of the methanol, indicative of the progress of the transesterication reaction, and dropped again

3164

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010

Figure 4. Gas chromatographic data showing ows from (A) light and (B) heavy ports of the centrifugal contactor (ow-through experiment with methanol-to-oil mole ratio of 4.9, 60 C, 3600 rpm), at intervals of 1, 3, 6, and 10 min after the ows of reagents were initiated.

Figure 5. Gas chromatographic results for production of C16 and C18 methyl esters as a function of reaction time at 45 and 60 C under ambient pressure in a stirred glass vessel. In both cases, the methanol-to-oil mole ratio was 4.9.

Figure 7. Ratio of triglyceride concentration to initial concentration as a function of time, taken under batch conditions in the centrifugal contactor, rotor speed of 3600 rpm, 4.87 methanol-to-oil mole ratio. In the case of the reaction at 50 C (upper line), a close-to-linear t suggests that a pseudorst-order removal process may be assumed. Deviations from linearity at 60 C ([), suggest that a pseudo-rst-order assumption that reagent methanol is in excess is not valid under these conditions.

Figure 6. Concentration of reagents and intermediates ((0.1 wt %) as a function of time in batch contactor tests (3600 rpm, 5.73 methanol-to-oil mole ratio, 50 C): triglyceride (), diglyceride (0), monoglyceride (]), total wt % bound glycerine (O), calculated from adding the contributions from the glycerine backbone of each of the acylglycerides. The dashed line rising to the right refers to the extent of the reaction or fraction of bound glycerine converted to methyl esters.

after a few minutes as the system equilibrated between the acylglycerides. Sampling in this case could only be done at the end of the test, as it involved depressurizing the vessel, so each of the data points came from a separate run. The difference in the bound glycerine concentrations from the two types of measurement, IR and viscosity analysis, shows that above the detection limit for viscosity, the two sets of data agreed within the reproducibility of the viscosity analysis, or (0.2%, for short reaction times up to 10 min. Signicant discrepancy was observed at longer times, which indicated possible degradation of the biodiesel, as observed earlier in the GC analysis.

Results from viscosity measurements are also presented graphically in Figure 8, showing the approach to ASTM standard for bound glycerine (<0.24 wt %). The graph shows a slow increase in triglyceride conversion from 90% completion at 2 min. From these data the minimum time for achievement of ASTM standard in one pass appears to be 22 min at 80 C. Reaction yields from IR spectroscopy are presented in Table 2, columns 5-8 and 10, and from viscosity measurements, column 9. As can be seen from table for entries corresponding to a total reaction time of 10 min, the reaction yield was high (<95% conversion) for mixtures that were reacted twice at 5 min, which can be compared with 96% for one pass at 15 min. In some cases a small amount of methanol, less than 1 mL, was injected into the second reaction stage; however, the resulting increase in methyl ester yield was not signicant. The variability in the results was indicative of the sensitivity of the analysis to small errors in sampling, such as contamination of sampling lines in the contactor with residual oil. Hence, the reported uncertainties in Table 2 were considerably greater than the measurement errors discussed in the Experimental Section. Methanolysis yields at above ambient pressure and temperature (temperature of 80 C at 2.6 bar) suggest that a 2 min residence time giving a 90% reaction yield would be optimal for the design of a commercial reactor/separator. Thus, to achieve ASTM specication fuel, that is, <0.24 wt % total glycerine, two or three cross-current reactors would be needed in series, with the byproduct glycerine being separated from the triglyceride/methyl ester mixture in stages. This experiment

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010


Table 2. Pressurized Transesterication, 3600 rpm total reaction time ((0.08 min) 5 5 5 10b 10b 10b 10b 10c 10c 15 15 15 15 15 30 45 45 methanol-to-oil molar ratio 4.8 4.8 4.8 5.0 5.1 4.9 5.9 4.8 4.8 4.8 4.8 4.8 4.8 4.8 4.8 4.8 4.8 max temp ((1 C) 75.0 82.0 82.0 80.0 79.5 82 82 79.5 79.5 80 60 80 80 80 80 80 78.9 max pressure ((0.03 bar) 2.52 2.52 2.52 1.97 2.52 2.17 2.17 2.52 2.52 2.48 1.00 2.52 1.93 2.52 2.41 2.48 2.59 MONO wt % 0.375 0.127 0.121 0.255 0.164 0.120 0.124 0.120 0.119 0.159 0.115 0.152 0.238 0.179 0.139 0.179 0.136 DI wt % 0.251 0.132 0.140 0.269 0.159 0.082 0.088 0.107 0.120 0.118 0.182 0.120 0.352 0.340 0.070 0.210 0.105 TRI wt % 0.157 0.073 0.078 0.171 0.165 0.050 0.050 0.050 0.070 0.070 0.226 0.048 0.194 0.086 0.031 0.117 0.087 total bound glycerine ((0.01 wt %) 0.783 0.332 0.339 0.696 0.489 0.253 0.261 0.277 0.309 0.346 0.523 0.321 0.784 0.604 0.239 0.506 0.328 bound glycerine by viscosity ((0.2 wt %)a 0.69 NM NM 0.72 0.51 NM NM NM NM 1.03 0.84 0.43 2.07 1.13 0.07 1.4 0.66

3165

free glycerine ((0.001 wt %) 0.001 0.000 0.045 0.001 0.018 0.000 0.098 0.000 0.034 0.000 0.008 0.000 0.000 0.078 0.000 0.041 0.000

a NM ) not measured because the bound glycerine was below the detection limit of the analytical method. b Two 5-min residence periods in the reactor with methanol added to the second batch. c Two 5-min residence periods in the reactor, with no addition of methanol.

Figure 8. Yield of batch transesterication reaction in terms of the weight percent of triglyceride reacted (2) and remaining total bound glycerine () as a function of reaction time (80 C, above ambient pressure to 2.6 bar, 3600 rpm rotor speed). The arrows indicate the conversion goal of <0.24 wt % bound glycerine or 97.8% conversion of acylglyceride.

Figure 9. Bound glycerine (wt %) is plotted as a function of time (min) for the staged production of biodiesel. Both experimental data (80 C, 3600 rpm, pressure up to 2.6 bar, methanol-to-oil mole ratio of 4.81) are presented, along with simulations from a chemical kinetic analysis.

was carried out with analysis by NMR, and the results are presented in Figure 9. In these tests, a small injection of methanol into the second and third stages further increased the reaction yield beyond that predicted for a model of a single stage reactor. The experimental acylglyceride concentration at 2 min was lower than predicted, which could have arisen from sampling errors, but could also be explained by an expected initial increase in the amount of bound glycerine as the triglyceride is converted into di- and monoglycerides. The different acylglycerides could not be separated in the proton NMR spectra. Kinetic Analysis and Discussion The biodiesel production experiments used angular momentum to maximize process intensication, a principle espoused by Ramshaw and co-workers,16 and to reduce the effect of mass transfer at the kinetic limit. In these experiments, the quality of the biodiesel produced, or the yield of the reaction, reached a maximum at 3600 rpm. At higher rpm (4800) the yield decreased, as it also did at lower rotational speeds (e.g., 3000 rpm). This can be explained from simulation of Taylor-Couette

ow, characteristic of annular centrifugal extractors as done by Deshmukh and co-workers17 using computational uid dynamics (CFD). At high turbulence, CFD calculations show that intravortex transport is fast relative to intervortex transport; thus, increasing rotational speed does not necessarily increase mixing in the reactor. As the contact time between reagents is governed by mixing, higher angular momentum does not necessarily translate to higher yields. This effect has also been modeled by Yacoub and Maron18 for a reactive systemssequential oxidation of liquids in an annular reactor. Hence, a kinetic analysis ignoring transport effects was performed using a three-step reaction (including back reactions) as proposed by Noureddini and Zhu19 as well as Freedman and co-workers20 for soybean oil and by Karmee and co-workers21 for Pongamia oil and computed using the kinetics code CHEMKIN.22 In addition to parameters describing the rate equations, thermodynamic data are needed for the computation of changes of state in the reactor, namely, heat capacities, enthalpies, and entropies, over the temperature range of the simulation. Derivation of functions describing these state variables is discussed below.

3166

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010


Table 4. Reaction Mechanism Values for CHEMKIN Analysis rate constants (cm3 mole-1 s-1) k1 k-1 k2 k-2 k3 k-3 Ea/R (K) 6844 4923 9445 5625 2607 4969 A (cm3 mole-1 s-1) 1.33 109 7.64 106 1.80 1013 7.48 108 1.29 104 5.59 105 k (80 C) (cm3 mole-1 s-1) 5.04 6.70 43.0 89.9 8.01 0.431

Table 3. Triglyceride Composition of Soybean Oil and Associated Thermodynamic Properties normalized Ln L O P mole (linolenic) (linoleic) (oleic) (palmitic) fraction C18:3 C18:2 C18:1 C16:0 0.18 0.34 0.27 0.21 278 545 1.13 -0.2946 1 0 0 0 0.083 278 547 1.07 2 3 2 2 0.75 280 547 1.13 0 0 0 0 1 0 0 1 0.083 0.083 282 256 546 1.18 526 1.11

triglyceride LLLn LLL OLL LLP mole fraction molecular weight, g Tc, C correction factor, J g-1 K-1

Chemical Kinetics Reaction Scheme. C57H92O6 (TRI) + CH4O (methanol) y z C39H64O5 (DI) + \
k-1 k1

C19H32O2 (ME) \ C39H64O5 + CH4O y z C21H36O4 (MONO) + C19H32O2


k-2 k3 k2

C21H36O4 + CH4O y z C3H8O3 (glycerine) + C19H32O2 \


k-3

Assumptions had to be made about the composition of the soybean oil as the fatty acid composition was not available for the commercial samples used in this experiment. However, the kinetic mechanism used to model the transesterication process was not expected to be sensitive to small changes in triglyceride composition as reected in the calculated physical properties of the mixture. Hence, the oil composition was assumed to be similar to that analyzed using high-performance liquid chromatography by Holcapek and co-workers.23 Their analysis showed that the soybean oil comprised 52 substituent triglycerides, of which the 4 most prevalent were used in our prediction of thermodynamic properties (substituent chains including Ln ) linolenic, L ) linoleic, O ) oleic, and P ) palmitic acids). The relative concentrations of the fatty chains in soybean oil constituent triglycerides are listed in Table 3, along with the predictions of critical temperature, Tc, and acentric factor, , following the procedure of Morad and colleagues.24 Although heat capacities have been measured for selected triglycerides,25 the data for this particular mixture for soybean oil are not available. Hence, these thermodynamic properties, along with ideal gas heat capacities calculated according to the method of Rihani and Dorasiwamy,26 were used to predict the heat capacity as a function of temperature following the method given by Morad and colleagues24,27 for a mixture of triglycerides. The average composition of the triglyceride was estimated to be C57H92O6. A correction factor was applied to the liquid specic heat capacity as suggested by Morad et al. Similar predictions were carried out for the diglyceride and monoglyceride intermediates, as well as the methyl esters, using nominal compositions of C29H64O5, C21H36O4, and C19H32O2, respectively. Over the temperature range of the experiments being simulated, the heat capacities were calculated to vary linearly with temperature. Previous work28 has indicated that the thermodynamic properties of the various triglycerides do not vary greatly between chain lengths of (2, and hence these predicted properties were considered to be sufciently accurate for the kinetic model used in this analysis. Thermodynamic data for

glycerine, methanol, and water were taken from the National Institute of Standards and Technology (NIST) compilation.29 The second-order rate constants, k1 through k-3, were assumed to have no temperature dependence in the pre-exponential factor, following the standard Arrhenius expression: k ) ATne-Ea/(RT), where n ) 0. The shunt reaction proposed by Noureddini was not incorporated into the model. The activation energies Ea and pre-exponential factors A were taken from Noureddini and Zhu19 for a well-mixed system (Reynolds number ) 12 400), and are reported in Table 4, along with rate constants for the system at 80 C. The rate constants for the reactions involving the diglyceride were reported to be an order of magnitude greater than the reactions involving the triglyceride and monoglyceride. Although Noureddini does not explicitly discuss this, an argument could be made that the formation of methyl ester at the reaction site improves the local miscibility of the reagents, thus increasing the rate of reaction. Mechanisms proposed by a different group20 do not show the same discrepancy, but include tting parameters, such as shunt reactions, to better simulate their data. Results from the kinetic model are presented in Figures 9-11. Overall, the long-time predicted behavior approximated the approach to limiting methyl ester yield observed in the pressurized batch contactor reactions, also shown earlier in Figure 9. The overall rate of the production of methyl ester can be written as d[ME] ) k1[TRI][CH3OH] - k-1[DI][ME] + dt k2[DI][CH3OH] - k-2[MONO][ME] + k3[MONO][CH3OH] - k-3[ME][C3H8O3] - k4[ME][H2O] The mechanism above also includes the reaction for the hydrolysis of methyl ester, k4. When the rates of forward and

Figure 10. Experimental mole ratios of diglyceride/triglyceride (+) and monoglyceride/diglyceride (/), at 80 C and 3600 rpm in the pressurized batch reactor, and simulation results for the diglyceride/triglyceride ratio (solid line) and the monoglyceride/diglyceride ratio (dashed line). Simulations shown on the graph use rate constants from Noureddini and Zhu,19 brown lines, and modications to k1 and k3 (blue lines), as specied in the text.

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010

3167

Figure 11. Relative sensitivity coefcients (arbitrary units) for production of methyl ester on the rate of conversion of triglyceride, diglyceride, monoglyceride, and hydrolysis of methyl ester over the time of the simulation (10 min). The simulation was carried out for a reaction at 80 C, 2.6 bar, similar to the kinetic tests in the pressurized batch reactor.

backward reactions become equivalent in each of the steps of the methanolysis mechanism, the rate of production of methyl ester approaches zero, which is what was observed in these experiments. A sensitivity analysis was performed on the rate of production of methyl ester with respect to the rates of conversion of triglyceride, diglyceride, and monoglyceride (Figure 11). First order sensitivity coefcients for species concentration and reaction rate on model parameters were simultaneously computed along with the solution of the differential equations describing the reaction mechanism.22 At early times or up to 2.5 min, the rate of production of methyl ester was most sensitive to the rate of the decomposition of the triglyceride (the blue line), while at later times, the rate of conversion of the monoglyceride to methyl ester (the green line) became more important. These sensitivity results were used to select which rate constants to vary to achieve better simulation of the data. Although the simulation shown in Figure 11 is for transesterication at 80 C, similar results were observed at higher and lower temperatures, with the time scale being shortened or lengthened, respectively. Results of the sensitivity analysis were used in further modeling of the transesterication process to optimize the rate constants and reproduce the experimental data. The brown lines in Figure 10 were calculated using the rate constants given in Table 4. The blue lines show results of simulations in which the pre-exponential factor for k1 and k3 had each been increased by a factor of 2 to better t the experimentally determined ratio of diglyceride/triglyceride. The reason for the better t with higher pre-exponential factors is not understood, but could be due to differences in soy oil composition or to differences in mass transfer between this experiment and those done by Noureddini and co-workers,19 even though an attempt was made by us and them to minimize the effect of mixing. The kinetic calculations performed here included saponication, or the hydrolysis of methyl ester. Activation energies were derived from Smith and Levenson30 for the saponication of the ethyl ester, as a comparable value for the methyl ester was not available. A reaction was also included to model the preferential partitioning of methanol into glycerine and away from the oil and methyl ester. The rate constant was estimated from values for hydrocarbons measured by Castells and colleagues.31 However, inclusion of these reactions was found to

have little effect on the overall yield of methyl ester, although small amounts of soap (hydrolysis product) were predicted when water was present in the mixture. Similar steady-state behavior has been observed by Karmee and co-workers21 in the transesterication of Pongamia oil, but the latter experiments were conducted for hundreds of minutes. Hence, with the high degree of mixing in the contactor system, the yield of the reaction appears to be driven by the kinetics up to the thermodynamic limit. In particular, the transesterication process is impeded by the build up of products in the reactor, including glycerine that can recombine with methyl ester. Hence, if the reaction mixture enters the rotor after a couple of minutes, and methyl esters are separated from byproduct glycerine, it should be possible to reduce the rate of back reactions and drive the process to >98% conversion. As discussed earlier, the gas chromatographic results (Figure 5) indicated a degradation of product with heating time, likely arising from beta ssion of the alkyl chain. Degradation reactions were not included in the kinetics mechanism as specic degradation products were not identied in the gas chromatogram, rather, a broadening of the peaks was attributed to generation of free fatty acids and fragments along with a decreased signal-to-noise ratio. The kinetic analysis of transesterication in the centrifugal contactor will be employed to explain and optimize future experiments where two or three contactors will be tagged together in a cross-ow conguration. For instance, reaction yield in single-stage reactors at longer residence times can be compared with yields in reactors at multiple stages with short residence times. Also, yields can be calculated with differing amounts of methanol added between the reactor stages. Finally, the transesterication reactions actually take place at the interface between the immiscible reagents, thus the system is inhomogeneous. A model that accounts for uid ow and mass transfer is needed before the reactor is scaled-up to a pilotscale or commercial sized plant. Conclusions Kinetics experiments were carried out to understand the process of transesterication in a centrifugal contactor. As the results describe, the ASTM specication for bound acylglycerides was achieved only at long reaction times in a singlestage batch reactor, greater than 22 min, even when conducted at elevated temperature and pressure. The approach to equilibrium between the acylglycerides limits the overall rate of production of methyl ester. After the reactants were in a wellmixed reactor for a few minutes, the centrifugal contactor for example, their overall conversion to methyl esters was found to be only dependent on the temperature. Hence, in the single-pass conguration, the residence time required to achieve ASTM specication gives little throughput advantage over the current batch reaction process. In addition, achieving a residence time of several minutes is not feasible in the contactor as it is currently engineered. Aside from temperature, the limitation to achieving complete reaction seems to be the presence of the products, including glycerine, which hinder complete conversion because of reversible reactions. Signicant improvement in quality was indicated after a second and third pass, where product and unreacted oil from the rst and second stages were collected and separated from the glycerine. The separated mixture was further reacted with a minor addition of methanol. The kinetic analysis suggests that the explanation for this improvement is the reduction in the rate of the back reactions that hinder the overall conversion to methyl esters.

3168

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010


(2) Vasudevan, P. T.; Briggs, M. Biodiesel ProductionsCurrent State of the Art and Challenges. J. Ind. Microbiol. Biotechnol. 2008, 35, 421. (3) ASTM Method D6751. Standard Specication for Biodiesel Fuel Blend Stock (B100) for Middle Distillate Fuels; ASTM International: West Conshohocken, PA, 2008. (4) Karmee, S. K.; Mahesh, P.; Ravi, R.; Chadha, A. Kinetic Study of the Base-Catalyzed Transesterication of Monoglycerides from Pongamia Oil. J. Am. Oil Chem. Soc. 2004, 81, 425. (5) Darnoko, D.; Cheryan, M. Kinetics of Palm Oil Transesterication in a Batch Teactor. J. Am. Oil Chem. Soc. 2000, 77, 1263. (6) Geeting, M. W.; Brass, E. A.; Brown, S. J.; Campbell, S. G. Scaleup of Caustic-Side Solvent Extraction Process for Removal of Cesium at Savannah River Site. Sep. Sci. Technol. 2008, 43, 2786. (7) Taylor, R. J.; May, I. Advances in Actinide and Technetium Kinetics for Applications in Process Flowsheet Modeling. Sep. Sci. Technol. 2001, 36, 1225. (8) Tsouris, C.; McFarlane, J.; Birdwell, J. F., Jr.; Jennings, H. L. Continuous Production of Biodiesel via an Intensied Reactive/Extraction Process. In SolVent Extraction. Fundamentals to Industrial Applications, Proceedings of ISEC, International Solvent Extraction Conference, 2008; Moyer, B. A., Ed.; Canadian Institute of Mining Metallurgy and Petroleum: Montreal, Canada, 2008; pp 923-930. (9) Birdwell, J. F., Jr.; Jennings, H. L.; McFarlane, J.; Tsouris, C. Integrated reactor and centrifugal separator and uses thereof. U.S. Patent Application 12/128157, 2008. (10) Peterson, C. L.; Cook, J. L.; Thompson, J. C.; Taberski, J. S. Continuous Flow Biodiesel Production. Appl. Eng. Agric. 2001, 18, 5. (11) Leonard, R. A.; Bernstein, G. J.; Ziegler, A. A.; Pelto, R. H. Annular Centrifugal Contactors for Solvent-Extraction. Sep. Sci. Technol. 1980, 15, 925. (12) Standard Test Method for the Determination of Free and Total Glycerin in B-100 Biodiesel Methyl Esters by Gas Chromatography; ASTM Method D6854; ASTM International: West Conshohocken, PA; 2007. (13) Proton NMR Collection; Sadtler Research Laboratories, Inc: Philadelphia, PA, 1980; pp 2708-2825. (14) Nawar, W. W.; Dubravcic, M. F. Thermal Decomposition of Methyl Oleate. J. Am. Oil Chem. Soc. 1968, 45, 100. (15) Osmont, A., Catoire, L. Dagaut, P. Thermodynamic Data for the Modeling of the Thermal Decomposition of Biodiesel. 1. Saturated and Monounsaturated FAMEs, J. Phys. Chem. A, published online Aug. 20, 2009, http://dx.doi.org/10.1021/jp904896r. (16) Jachuck, R. J.; Lee, J.; Kolokotsa, D.; Ramshaw, C.; Valachis, P.; Yanniotis, S. Process Intensication for Energy Saving. Appl. Therm. Eng. 1987, 17, 861. (17) Deshmukh, S. S.; Sathe, M. J.; Jyeshtharaj, B. J.; Koganti, S. B. Residence Time Distribution and Flow Patterns in the Single-Phase Annular Region of Annular Centrifugal Extractor. Ind. Eng. Chem. Res. 2009, 48, 37. (18) Yacoub, N.; Maron, D. M. Analysis of Centrifugal Annular Reactor with Radial Flow. Chem. Eng. Sci. 1984, 39, 313. (19) Noureddini, H.; Zhu, D. Kinetics of Transesterication of Soybean Oil. J. Am. Oil Chem. Soc. 1997, 74, 1457. (20) Freedman, B.; Buttereld, R. O.; Pryde, E. H. Transesterication Kinetics of Soybean Oil. J. Am. Oil Chem. Soc. 1986, 63, 1375. (21) Karmee, S. K.; Chandna, D.; Ravi, R.; Chadha, A. Kinetics of BaseCatalyzed Transesterication of Triglycerides from Pongamia Oil. J. Am. Oil Chem. Soc. 2006, 83, 873. (22) Lutz, A. E.; Kee, R. J.; Miller, J. A. SENKIN: A Fortran Program for Predicting Homogeneous Gas Phase Chemical Kinetics with SensitiVity Analysis. Sandia National Laboratory Report, SAND87-8248; Sandia National Laboratory: Livermore, CA, February, 1988. (23) Holeapek, M.; Jandera, P; Zderadieka, P.; Hruba, L. Characterization ` ` of Triacylglycerol and Diacylglycerol Composition of Plant Oils Using HighPerformance Liquid ChromatographysAtmospheric Pressure Chemical Ionization Mass Spectrometry. J. Chromatogr. A 2003, 1010, 195. (24) Morad, N. A.; Kamal, A. A. M.; Panau, F.; Yew, T. W. Liquid Specic Heat Capacity Estimation for Fatty Acids. Triacylglycerols, and Vegetable Oils Based on Their Fatty Acid Composition. J. Am. Oil Chem. Soc. 2000, 77, 1001. (25) Charbonnet, G. H.; Singleton, W. S. Thermal Properties of Fats and Oils. VI. Heat Capacity, Heats of Fusion and Transition, and Entropy of Trilaurin, Trimyristin, Tripalmitin, and Tristearin. J. Am. Oil Chem. Soc. 1947, 24, 140. (26) Rihani, D. N.; Doraiswamy, L. K. Estimation of Heat Capacity of Organic Compounds from Group Contributions. Ind. Eng. Chem. Fundam. 1965, 4, 17. (27) Morad, N. A.; Idrees, M.; Hasan, A. A. Specic Heat Capacities of Pure Triglycerides by Heat-Flux Differential Scanning Calorimetry. J. Therm. Anal. 1995, 45, 1449.

These ndings could be used to develop a pilot scale system for the continuous production of biodiesel. A second or third stage will increase the capital investment required to implement the process improvement, but the increase should be offset by reduced operating costs. Acknowledgment The authors thank Nu-Energie (particularly Brian Hullette and Joel Day) for providing the reaction materials and help with the analysis of the product, and Elizabeth Ashby for her assistance with the chemical kinetic modeling. Preliminary research was sponsored by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U.S. Department of Energy. Funding for this project was provided in part by the Department of Energys Ofce of Energy Efciency and Renewable Energys Technology Commercialization and Deployment Programs Technology Commercialization Fund and by Nu-Energie, LLC, under CRADA No. 01377. Supporting Information Available: Experimental details and chemical kinetics calculations, namely, input les formatted for CHEMKIN III. This material is available free of charge via the Internet at http://pubs.acs.org. Appendix
AbbreViations ASTM ) American Society for Testing and Materials CFD ) computational uid dynamics CRADA ) Cooperative Research and Development Agreement DI ) diglyceride FT-IR ) Fourier transform infrared spectroscopy FT-Raman ) Fourier transform Raman spectroscopy GC-FID ) gas chromatography ame ionization detection 1 H NMR ) proton nuclear magnetic resonance spectroscopy IR ) infrared spectroscopy (includes FTIR, and FT-Raman) L ) linoleic, -O(CdO)C17H29 LLC ) limited liability company Ln ) linolenic, -O(CdO)C17H27 ME ) methyl ester MONO ) monoglyceride NIST ) National Institute of Standards and Technology O ) oleic, -O(CdO)C17H31 ORNL ) Oak Ridge National Laboratory P ) palmitic, -O(CdO)C15H31 TMS ) tetramethylsilane TRI ) triglyceride UT ) University of Tennessee Symbols A ) preexponential factor C16, etc. ) the number of carbons in the fatty acid fragment of the methyl ester. Ea ) activation energy k ) second order rate constant R ) gas constant T, Tc ) temperature, critical temperature ) acentric factor

Literature Cited
(1) Van Gerpen, J.; Shanks, B.; Pruszko, R.; Clements, D.; Knothe, G. Biodiesel Production Technology. Subcontractor Report, NREL/SR-51036244, Contract DE-AC36-99-GO10337; National Renewable Energy Laboratory: Golden, CO, 2004.

Ind. Eng. Chem. Res., Vol. 49, No. 7, 2010


(28) Charkravarthy, K.; McFarlane, J.; Daw, S.; Ra, Y.; Reitz, R.; Grifn, J. Physical Properties of Soy Bio-Diesel and n-Heptane: Implications for Use of Bio-Diesel in Diesel Engines. Soc. Automot. Eng. Trans. J. Fuels Lubr. 2007, 116, 885. (29) National Institute of Standards and Technology. NIST Standard Reference Database 69: NIST Chemistry WebBook. U.S. Secretary of Commerce on behalf of the United States of America, 2008. http:// webbook.nist.gov/chemistry (accessed June 2009). (30) Smith, H. A.; Levenson, H. S. Kinetics of the Saponication of the Ethyl Esters of Normal Aliphatic Acids. J. Am. Chem. Soc. 1939, 61, 1172.

3169

(31) Castells, R. C.; Arancibia, E. L.; Nardillo, A. M. Solution and Adsorption of Hydrocarbons in Glycerol As Studied by Gas-Liquid Chromatography. J. Phys. Chem. 1982, 86, 4456.

ReceiVed for reView August 4, 2009 ReVised manuscript receiVed January 26, 2010 Accepted February 10, 2010 IE901229X

Vous aimerez peut-être aussi