9718
Ind. Eng. Chem. Res. 2005, 44, 9718-9726
Performing Esterification Reactions by Combining Heterogeneous Catalysis and Pervaporation in a Batch Process Damjan Nemec† and Robert van Gemert* Akzo Nobel Chemicals b.v., P.O. Box 9300, NL-6800 SB Arnhem, The Netherlands
A pragmatic approach is presented for first screening and design purposes when considering performing esterifications in a batch process, which combines heterogeneous catalysis and pervaporation for shifting the reaction toward product formation by water removal. A relatively slow esterification between tartaric acid and ethanol was chosen as a model system. Realistic performance of currently commercially available catalysts and ceramic membranes is taken into account by available experimental data. Three different process configurations, which differ in the degree of coupling of reaction and pervaporation, are analyzed. For the esterification reaction system studied, it is shown that a multifunctional membrane reactor is not a viable concept, since its operation does not proximate to the optimal requirements for the two integrated coprocesses. When it comes to configurations where catalytic reaction and separation are performed in separate units, the end design will usually depend on practical issues such as membrane stability and ease of operation. 1. Introduction Integration of reaction and separation may lead to substantial savings in capital as well as operating costs, which can come as a result of higher conversion and selectivity, smaller equipment size, energy integration, etc. As such, the topic attracts the interest of academia and industry alike.1-3 To take advantage of the potential benefits of integrating reaction and separation, an appropriate chemical reaction has to be identified. Esterification reactions are typically referred to in this respect since they are limited by thermodynamic equilibrium. Commercially, they are carried out either by using a large excess of one of the reactants or by removing one of the products by distillation or vacuum stripping. The former is a relatively inefficient approach, since it requires a large reactor volume and a separation step afterward. Distillation, on the other hand, favorably shifts the equilibrium by removing one of the products. However, it is an energy-intensive operation, which besides higher operating costs can also cause problems when dealing with temperature-sensitive products or by increasing the yield of byproducts through side reactions.4 Azeotrope formation is, in a number of cases, also a problem with distillation. Pervaporation has been proven to be a successful technique for selectively removing a certain component from a liquid mixture because of its low energy requirements and its ability to separate azeotropic mixtures.5 As such, the technique is being seriously considered for performing esterifications by reactive separations,6,7 especially since the emergence of ceramic (silica or zeolite) membranes, which selectively remove water as one of the esterification products and are believed to be resistant to elevated temperatures and relatively harsh * Corresponding author. Tel.: +31-26-366-5753. Fax: +31-26-366-5871. E-mail: robert.vangemert@ akzonobel-chemicals.com. † Current address: BIA Separations, Slovenia. E-mail:
[email protected].
chemical environments. However, esterifications are usually catalyzed by homogeneous catalysts, which are typically strong acids such as sulfuric and methanesulfonic acid. These have been found to attack the ceramic membranes, which, as a result, quickly lose their performance. Therefore, heterogeneous (solid acid) catalysts such as ion-exchange resins (Amberlyst and Nafion as typical commercially available products) have been considered in order to prevent the degradation of ceramic membranes. Furthermore, application of solid catalysis offers other potential benefits, like easy separation of catalyst from product. This means easy recovery of catalyst and/or absence of need for product cleanup as well as the absence of waste streams that come with neutralization steps. Solid catalysts are often found to yield higher selectivity as well. There exists a number of possibilities for how to combine catalytic reaction and pervaporation to achieve a desired result. Lim et al.8 have already discussed the issue. However, they have only done so from a theoretical perspective, whereas the present work deals with a pragmatic approach by considering commercially available materials (catalyst and membranes) and is supported by available experimental data. The study focuses on the production of speciality chemicals, which are, due to moderate capacities ( 1 × 105) (14)
However, the region of practical interest for the design of a pervaporation module with currently available membranes (103 < Re < 105) is not covered by these correlations. Therefore, kL values were determined from the water flux data available (Figure 2) with the help of eq 12. The intrinsic membrane resistance (RM(p) ) 13.5 bar‚m2‚s/mol) was estimated from the data obtained at the highest linear velocities also via eq 12, whereby it was assumed that the resistance due to the liquid film is negligible. The results of experimentally determined liquid-side mass transfer coefficients (kL) are shown in Figure 5. The obtained results seem to be in good agreement with the general trends dictated by the correlations for laminar and turbulent flow; therefore, a new correlation was developed for that specific region for the purpose of using it for the simulation of the whole
(103 < Re < 104)
(15)
As can be seen from Figures 2 and 3, the predictive value of the membrane model (eq 12) and the correlation for the liquid-side mass transfer coefficient (eq 15) is very good indeed and can be used for reliable simulations of the pervaporation process. 5. Process Configuration The catalyst (Amberlyst 15) is commercially available only in particle form and does not really possess high mechanical stability. This, in essence, brings down the number of choices to a single reactor type: the packed bed. On the other hand, there exists a number of options for how to combine the reactor with the pervaporation unit within the whole batch system. Figure 6 shows three possibilities. First, there is the parallel configuration, where the reaction mixture is recirculated through the packed-bed reactor and pervaporation module in two separate loops. The second configuration, which introduces a higher degree of coupling, is the socalled configuration in series, where the mixing tank, packed bed, and pervaporation module are connected with one single loop. A possible recirculation of feed over the pervaporation module can be introduced if that is desired because of different hydrodynamic requirements in the reactor and pervaporation units (i.e., lower flow rates in the reactor due to pressure drop limitations and higher flow rates in the pervaporation unit to avoid liquid-side mass transfer limitations). The third possible configuration is the fully integrated (or multifunctional) reactor, which performs the reaction and pervaporation within a single vessel. The multifunctional membrane reactor was envisioned as being incorporated into the commercially available PERVAP SMS membrane module offered by Sulzer Chemtech (also shown in Figure 6). The module is originally meant for heat exchange and mass transfer (in this case, pervaporation), but its use could be extended to incorporate a catalytic reaction by adding a solid catalyst within the annular space surrounding the membrane. The multifunctional reactor obviously introduces the highest degree of coupling because degrees of freedom are lost due to the coupling
9722
Ind. Eng. Chem. Res., Vol. 44, No. 25, 2005
Figure 8. Amounts of ethanol converted and diethyl ester produced over time as functions of the recirculation rate of the reaction solution through the packed-bed reactor (values correspond to 10 m3 batch volume, 3.1 m3 packed bed, and 110 m2 of membrane area at 85 °C)
Figure 7. Typical concentration profiles of components for the esterification between tartaric acid and ethanol at 85 °C: (a) stoichiometric ratio of reactants and (b) 10% excess of ethanol.
of the amount of catalyst with the membrane area as well as a single flow rate that facilitates the reaction and pervaporation process. The performance of the three possible process configurations has been studied by computer simulations. The built models were kept simple but realistic by incorporating the available experimental findings for catalyst and membrane performance. More details are given with the discussion on each of the three process configurations. 5.1. Parallel Configuration. Plug flow is assumed in the packed-bed reactor as well as in the pervaporation unit. Because of a high flow rate through the membrane module, the composition of the feed hardly changes during a single pass; therefore, composition differences in the radial direction are negligible. In the mixing tank, complete and instantaneous mixing is assumed. All three units are heated to the same temperature. We recall that the maximum allowable pressure drop over the packed bed for Amberlyst catalyst is 1 bar; therefore, monitoring this parameter is important. This was determined by applying the well-known Ergun equation. Figure 7 shows typical concentration profiles of components for the esterification between tartaric acid and ethanol. A realistic goal for an industrial batch process is to process a batch within 24 h. This would mean that the reaction would need to be completed in ∼18 h and a further 6 h would be used for product aftertreatment (if needed), empting and cleaning the vessels, and charging the new batch and heating it up to the desired temperature. Figure 7a indicates that, when the desired product is a relatively “clean” diethyl ester, this is not possible to achieve within this time frame. Despite the fact that almost all of the water has been removed from the solution, there is still a substantial amount of
monoethyl ester present, and a lot more time would be needed for it to transform into the diethyl form. Removal of “traces” of water from the reaction mixture with pervaporation is very slow because the driving force is low. Therefore, an excess of ethanol is needed if we are to achieve relative full conversion within the time frame of 18 h (Figure 7b). The excess of ethanol can easily be removed from the product by stripping; however, this results in prolonged production time and additional expenses. For reaction equilibrium shift, pervaporation offers no real advantage over distillation, where a small excess of one of the reactants is also employed in practice; on the other hand, it is much less energy intensive because of a selective water permeation, and pervaporation also avoids a difficult azeotrope separation. A parameter worth paying attention to in a batch process is the recirculation rate of the mixture through the packed-bed reactor. As can be seen from Figure 8, the higher the flow through the reactor (the shorter the residence time of a unit volume), the higher the conversion is over time. This, at first, goes against one’s intuition; however, the reason for this lies in the fact that the longer a unit volume spends in the reactor, the more products are formed, which slows down the reaction. This means that the catalyst is less effectively used. Therefore, it turns out that it is better to achieve only a small conversion per pass, which is then followed by a “fresh” supply of reaction mixture at a faster rate, since in this way the catalyst is utilized better. In fact, it is desirable to have as fast a recirculation as possible, with the only limit being the allowable pressure drop over the bed. 5.2. Configuration in Series. There are no particularities in modeling the configuration in series in comparison to what has already been mentioned with the parallel configuration. When comparing the parallel configuration with the configuration in series, it has to be said that the latter, in principle, performs better, as is shown in Figure 9. The difference in performance is not big and has to do with the dilution of the water that is formed in the packed-bed reactor, which takes place in the mixing vessel. The diluted stream (in terms of water content) then enters the pervaporation module, and as a consequence, the driving force for water removal is lower than what it could have been (if compared to the configuration in series).
Ind. Eng. Chem. Res., Vol. 44, No. 25, 2005 9723
Figure 9. Comparison of process performance in terms of ethanol and water concentration profiles for the configuration in series and parallel configuration at identical operating conditions.
Figure 10. Comparison of process performance in terms of ethanol, diethyl ester, and water concentration profiles for two different modes of operation of the parallel configuration: 1-stage process ) recirculation of reaction mixture through the pervaporation module from the start; 2-stage process ) recirculation of reaction mixture through the pervaporation module only after 3 h.
However, the time to reach the specified end conversion is not necessarily the determining factor in process design. An important issue can also, in some cases, be the stability of membranes. With the current process, the membranes represent the bulk of the investment cost (∼3000 euro/m2); therefore, an operation mode that preserves the membranes as long as possible should be found. From our experience, currently available ceramic membranes lose their performance (selectivity) after a certain period of time. Although the mechanism is not completely clear, it is speculated that the harsh acidic environment may be responsible. If this does occur, the idea is to perform the esterification in two sequential stages. In the first part, the reaction mixture is only recirculated through the packed-bed reactor. At this point, the reaction will be driven by the high starting concentrations of reactants. After this first stage, the bulk of the acid will have reacted and, therefore, should pose less threat to the membranes. This two-stage process was simulated and compared with a single-stage process. The results are shown in Figure 10. During the first 3 h, no water was removed by pervaporation. Despite this, the difference in ethanol conversion and diethyl ester production only starts appearing after ∼2.5 h. Of course, there is a large difference in water concentration; however, as speculated before, this has no real effect on the conversion
since, at the beginning, the reaction is driven toward product formation mainly because of high concentrations of reactants at the start of the reaction. Once the water concentration builds up, the reaction will inevitably slow and, at this point (after 3 h of operation), the second stage is started by switching on the recirculation of the half-reacted mixture through the pervaporation module. It would be reasonable to expect that the time for completion of the reaction will be prolonged somewhat because of water concentration buildup; however, as it turns out, much of this “lost” time is made up in the course of the second stage because of this same high water content. Because the driving force is high, the water flux through the membrane (water removal) is also fast. In the end (after 18 h), the differences in conversion are marginal. On the other hand, when stability of membranes is an issue, this mode of operation can offer a considerable benefit to the membranes, since acid concentrations to which the membranes are exposed are significantly lower. This could mean that the lifetime of the membranes will be prolonged several times. The above mode of operation is only possible with the parallel configuration, which will be the preferred process setup whenever concerns exist over membrane stability. One could argue that membranes could also be preserved with the configuration-in-series setup if the conversion per single pass would be high enough. That is true if the reaction is fast and/or if the residence time is long enough. However, when we recall the results presented in Figure 8, it has to be said that one would pay the price in one way or the other; either conversion per pass would be low and membranes would suffer or the conversion per pass would be high enough, but because of low recirculation rates, the overall conversion would suffer. 5.3. Multifunctional Reactor. The Sulzer PERVAP SMS membrane module is designed for Sulzer-type silica membranes which have the selective silica layer deposited on the outside, unlike the ones of Pervatech with the selective layer on the inside of the tubular membrane; however, we will assume that their performance is comparable. The geometric characteristics of the tube elements in the module are shown in Figure 11, with the catalyst packed in the annular space between the membrane on the inside and the heat exchanger tube on the outside. For such a case, the following balance can be written
v(r)
( )
dci 1 ∂ ∂ci + FbRi ) -Dr dz r ∂r ∂r
(16)
with the following boundary conditions:
dci ) 0 @ r ) Dout/2 dr
(17)
for all components,
dci ) 0 @ r ) Din/2 dr
(18)
for rejected components, and
Ji(m) )
(xiγipisat - yipvac) @ r ) Din/2 Hi 1 (p) + RM Mi cTOT kL
(
)
(19)
9724
Ind. Eng. Chem. Res., Vol. 44, No. 25, 2005
Figure 11. Geometrical characteristics of the Sulzer PERVAP SMS module: Din ) 14 mm, Dout ) 23 mm, and x ) (Dout - Din)/2 ) 4.5 mm.
for selectively removed components. Although a velocity profile in the radial direction would need to be estimated, we assume for the first appoximation that it is uniform over the whole cross section. When it comes to estimating the mass transfer coefficient, one has to take into account that the correlations for an empty tube are no loger valid; therefore, new correlations for packed beds are needed. Again, assuming the analogy between heat and mass transfer is valid, the correlations of Dixon and Cresswell13 for fluid-to-wall heat transfer can be adapted as follows:
Sh ) 0.6Re0.5Sc0.33 Sh ) 0.2Re0.8Sc0.33
(1 < Re < 40) (40 < Re < 2 × 105)
(20) (21)
The linear dimension for the Sherwood and Reynolds numbers for packed beds is the particle diameter, unlike that for empty tubes, where the linear dimension is the diameter of the tube. A comparison between the mass transfer coefficients obtained for an empty tube and a tube packed with particles reveals that the latter is ∼1 order of magnitude higher for the same superficial velocities. However, because of the pressure drop limitation of 1 bar, the maximum velocity was found to be ∼3 cm/s, which is 2 orders of magnitude lower than what is typically employed in a pervaporation unit (see Figure 3). This means that the low obtainable liquid-side mass transfer coefficient is the major contributor to the resistance for the pervaporation process, and the corresponding rate of water removal is much lower compared to what can be achieved when catalytic reaction and pervaporation are performed in separate units, because there the operating conditions can be set independently. Another problem with the multifunctional reactor is the coupling of the amount of catalyst and the membrane area. It is very important to ensure a sufficient amount of catalyst to get the reaction going in the first place. Figure 12 demonstrates this by comparing three cases: (1) multifunctional reactor; (2) configuration in series, equal amount of catalyst but higher mass transfer coefficient compared to the multifunctional reactor case; (3) configuration in series, higher amount of catalyst and higher mass transfer coefficient compared to the multifunctional reactor case. When cases (1) and (2) are compared, there is no real difference in the amount of reacted ethanol (Figure 12a), despite the fact that the water removal (Figure 12b) is much better for case (2). Only when the amount of catalyst available is substantially increased, case (3), we see a significant difference in overall process performance. This means that the space available for the catalyst is far too small for our model esterification reaction, which, admittedly, is fairly slow to begin with. Nevertheless, this points out just another possible disadvantage of a multifunctional reactor concept.
Figure 12. Comparison of process performance in terms of ethanol and water concentration profiles. Case (1): multifunctional reactor, Vcat ) 0.65 m3, kL ) 0.4 × 10-5 m/s; Case (2): configuration in series, Vcat ) 0.65 m3, kL ) 4.1 × 10-5 m/s; Case (3): configuration in series, Vcat ) 2.65 m3, kL ) 4.1 × 10-5 m/s.
A few final words on the multifunctional reactor: even if one would find a suitable (fast) reaction, a desired membrane performance, and an operational window (where the coupling of catalyst amount and membrane surface as well as the flow rates would not hinder its performance), it is still hard to imagine the concept would see light in the industry. It is simply not practical. One should just consider reactor preparation by having to carefully pack over 1000 annular tubes! Furthermore, the particles in direct contact with the membrane will inevitably increase the possibility of fouling. Even if fouling would have occurred anyway (due to deposits from the feed), maintenance with cleaning-in-place procedures is questionable because of the possibility of catalyst deactivation. 6. Conclusions A pragmatic approach toward the design and evaluation of a batch reaction process combining heterogeneous catalysis and pervaporation has been shown. Although the modeling was kept simple, it does include
Ind. Eng. Chem. Res., Vol. 44, No. 25, 2005 9725
realistic, experimentally determined values for the currently available commercial ceramic membranes. When it comes to esterifications, pervaporation can be used as an alternative to today’s conventional technology, distillation. However, to reach sufficient conversion, an excess of one of the reactants (up to 10%) is needed, which means an after-treatment of the product (like stripping) still needs to be applied. On the other hand, pervaporation is much less energy intensive compared to distillation. The exact configuration of a process combining catalytic reaction and separation will depend on the rates of reaction and water removal as well as more practical issues such as membrane stability and ease of operation and maintenance. In principle, the higher the degree of coupling of the two-unit operations, the less flexibility there exists in design and operation of the process. For the study case chosen, the multifunctional reactor does not perform up to the level of configurations where catalytic reaction and separation are performed in separate units, since its operation does not proximate to the optimal requirements for the two integrated coprocesses, which is a general problem with multifunctional reactors.14 Furthermore, it is not a practical process concept from the point of view of reactor preparation and maintenance. The reaction between tartaric acid and ethanol is fairly slow, requiring a substantial amount of catalyst in packed-bed form, and as a consequence, a number of limitations of a multifunctional reactor were revealed. For a much faster reaction, where the membranes could be impregnated with a thin layer of catalyst, such a reactor concept could prove to be more viable. Multifunctional membrane reactors would prove especially advantageous in cases where selectivity issues are of major concern, as a product could be removed in situ before it reacts further. This, however, does not really apply to esterification reactions where the product is water, which only plays a role in the thermodynamic limitations.
Pi ) permeance of ith component through the membrane (kg/m2‚s‚bar) pi(sat) ) saturation pressure of ith component (bar) p(vac) ) pressure on the permeate side (bar) r ) radial distance (m) Rg ) gas constant (8.314 J/mol‚K) Ri ) reaction rate (mol/m3‚s) Re ) Reynolds number, Fvdh/µ (/) RM(c) ) intrinsic resistance of membrane, concentration based (s/m) RM(p) ) intrinsic resistance of membrane, pressure based (bar‚m2‚s/mol) Sc ) Schmidt number, µi/FiDi (/) Sh ) Sherwood number, kLdh/Di (/) T ) temperature (K) t ) time (s) v ) linear velocity of flowing liquid (m/s) xi ) mole fraction of ith component in liquid phase (/) yi ) mole fraction of ith component in the permeate (/) Greek Letters γi ) activity coefficient of ith component (/) δL ) thickness of boundary layer (m) δM ) thickness of membrane (m) µ ) viscosity of reaction mixture (Pa‚s) F ) density of reaction mixture (kg/m3) Fb ) bulk density of catalyst bed (kg/m3) Subscripts 1 ) first reaction 2 ) second reaction b ) bulk in ) inner int ) liquid-membrane interface out ) outer p ) permeate Abbreviations EtOH ) ethanol TaEt2 ) diethyl tartarate TaH2 ) tartaric acid TaHEt ) ethyl tartarate
Acknowledgment This research has been supported by a Marie Curie Fellowship of the European Community program Industry Host Fellowship under Contract No. HPMI-CT2002-00191. The authors thank Ellie Renkema and Lian Nabuurs for performing the kinetic experiments and Antoon ten Kate for providing the UNIFAC calculations. Nomenclature ci ) mole concentration of ith component (mol/m3) cn ) normalized mole concentration (/) cTOT ) total molar concentration of components (mol/m3) Di ) diffusivity of ith component (m2/s) dh ) hydraulic diameter (m) Ji ) mass flux of ith component through the membrane (kg/m2‚s) (m) Ji ) mole flux of ith component through the membrane (mol/m2‚s) Hi ) Henry constant of ith component (bar) k ) reaction rate constant (m3/kgcat‚mol‚s) ka ) adsorption constant (m3/mol) Keq ) equilibrium constant (/) kL ) liquid-side mass transfer coefficient (m/s) L ) membrane tube length (m) Mcat ) mass of catalyst (kg) Mi ) molar weight of ith component (kg/mol)
Literature Cited (1) Stankiewicz, A. Reactive Separations for Process Intensification: An Industrial Perspective. Chem. Eng. Process. 2003, 42, 137. (2) Krishna, R. Reactive Separations: More Ways to Skin a Cat. Chem. Eng. Sci. 2002, 57, 1491. (3) Armor, J. N. Overcoming Equilibrium Limitations in Chemical Processes. Appl. Catal., A 2001, 222, 91. (4) Keurentjes J. T. F.; Janssen, G. H. R.; Gorissen, J. J. The Esterification of Tartaric Acid with Ethanol: Kinetics and Shifting the Equilibrium by Means of Pervaporation. Chem. Eng. Sci. 1994, 24A, 4681. (5) Jonquieres, A.; Clement, R.; Lochon P.; Neel, J.; Dresch, M.; Chretien, B. Industrial State-of-the-Art of Pervaporation and Vapour Permeation in the Western Countries. J. Membr. Sci. 2002, 206, 87. (6) Benedict, J. D.; Parulekar, S. J.; Tsai, S.-P. Esterification of Lactic Acid and Ethanol with/without Pervaporation. Ind. Eng. Chem. Res. 2003, 42, 2282. (7) Peters, T. A.; Fontalvo, J.; Vorstman M. A. G.; Keurentjes, J. T. F. Design Directions for Composite Catalytic Hollow Fibre Membranes for Condensation Reactions. Trans. Inst. Chem. Eng. 2004, 82 (A2), 220 (8) Lim, S. Y.; Park, B.; Hung, F.; Sahimi, M.; Tsotsis, T. T. Design Issues of Pervaporation Membrane Reactors for Esterification. Chem. Eng. Sci. 2002, 57, 4933. (9) Sharma, M. M. Some Novel Aspects of Cationic IonExchange Resins as Catalysts. React. Funct. Polym. 1995, 26, 3. (10) Levenspiel, O. Chemical Reaction Engineering, 2nd ed.; Wiley: New York, 1972.
9726
Ind. Eng. Chem. Res., Vol. 44, No. 25, 2005
(11) Wijmans, J. G.; Baker, R. W. A Simple Predictive Treatment of the Permeation Process in Pervaporation. J. Membr. Sci. 1993, 79, 101. (12) Karlsson, H. O. E.; Tra¨ga˚rdh, G. Pervaporation of Dilute Organic-Water Mixtures. A Literature Review on Modelling Studies and Applications to Aroma Compound Recovery. J. Membr. Sci. 1993, 76, 121. (13) Dixon, A. G.; Cresswell, D. L. Theoretical Prediction of Effective Heat Transfer Parameters in Packed Beds. AIChE J. 1979, 25, 663.
(14) Stitt, H. Multifunctional Reactors? “Up to a Point Lord Copper”. Trans. Inst. Chem. Eng. 2004, 82 (A2), 129.
Received for review March 1, 2005 Revised manuscript received July 23, 2005 Accepted July 26, 2005 IE050283+
