Ind. Eng. Chem. Res. 1987,26,696-699
696
inadequate number of manipulative variables and inadequate capacity to tolerate disturbances that increase the load on the process, can be identified from steady-state considerations. There are a number of alternative ways of restoring the steady-state operability, and short-cut procedures can be used to estimate the cost penalties. A t the preliminary stage of a process design, we normally look for a low-cost alternative, rather than the optimum solution. However, a dynamic study is necessary to make definitive conclusions about operability. Acknowledgment We are grateful to the Department of Energy for supporting this work (DOE Contract DE-AC02-8ER10938). Nomenclature A = heat-exchanger area, ft2 C = heat capacity, BTU/(lb O F ) lf= flow rate, lb/h Q = heat duty, BTU/h AT,, = log-mean temperature difference, OF U = overall heat-transfer coefficient, BTU/(h ft2 O F ) Literature Cited Douglas, J. M. Presented a t the Proceedings of the Engineering Foundation Conference on Chemical Process Control 11, Sea Island, GA, 1981; p 497.
Douglas, J. M.; Reiff, E. W.; Kittrell, J. R. Chem. Eng. Sci. 1980,35, 322. Fisher, W. R.; Doherty, M. F.; Douglas, J. M. submitted for publication in Ind. Eng. Chem. Res. 1987a. Fisher, W. R.; Doherty, M. F.; Douglas J. M. submitted for publication in Ind. Eng. Chem. Res. 1987b. Grossmann, I. E.; Morari, M. Presented at the Proceedings of the 2nd International Conference on Foundations of Computer-Aided Process Design, Snowmass, CO, June 19-24, 1983. Halemane, K. P.; Grossmann, I. E. AIChE J . 1983, 29, 425. Kittrell, J. R.; Watson, C. C. Chem. Eng. Prog. 1966, 62(4), 79. Linnhoff, B.; Turner, J. A. Chem. Eng. (London) 1980, 363, 742. Marselle, D. F.; Morari, M.; Rudd, D. F. Chem. Eng. Sci. 1982, 37, 259. Nishida, N.; Ichikawa, A.; Tazaki, E. Ind. Eng. Chem. Process Des. Dev. 1974, 13, 209. Saboo, A. K.; Morari, M. Chem. Eng. Sci. 1984 39(3), 579. Saboo, A. K.; Morari, M.; Woodcock, D. C. Chem. Eng. Sci. 1985,40, 1552. Shinskey, F. G. Distillation Control for Productivity and Energy Conservation; McGraw-Hill: New York, 1977. Swaney, R. E.; Grossmann, I. E. AIChE J . 1985a, 31, 621. Swaney, R. E.; Grossmann, I. E. AIChE J . 1985b, 31, 631. Terrill, D. L. Ph.D. Dissertation, University of Massachusetts, Amherst, MA, 1985. Terrill, D. L.; Douglas, J. M. Ind. Eng. Chem. Res. 1987, preceding paper in this issue. Townsend, D. W.; Linnhoff, B. AIChE J . 1983, 29, 742.
Received f o r review August 8, 1985 Revised manuscript received May 19, 1986 Accepted J u n e 13, 1986
Extraction of Ethanol from Aqueous Solution. 1. Solvent Less Volatile than Ethanol: 2-Ethylhexanol Francisco Ruiz,* Vicente Gomis, and Rogelio F. Botella DivisiBn de Ingenieria Q u h i c a , Universidad de Alicante, Alicante, Spain
Liquid-liquid equilibrium data for the ternary system water-ethanol-2-ethylhexanol have been determined experimentally at 25 "C and correlated simultaneously together with vapor-liquid equilibrium data by using the UNIQUAC model. A suitable extraction process for separating ethanol and water using 2-ethylhexanol as the solvent has been chosen, and the design calculations have been carried out t o determine the energetic requirements. The properties which another solvent should offer t o decrease these energetic requirements have been studied. The fermentation process and its recent developments have led to the efficient production of dilute alcohol-water mixtures. The conventional method, distillation and azeotropic distillation, for recovering anhydrous ethanol from the fermentation broth consumes 50-80% of the energy used in a typical fermentation ethanol manufacturing process and is frequently cited in criticizing the potential of biomass-derived ethanol as a liquid fuel (Ladisch and Dyck, 1979). However, new technologies for separating alcohol from water solutions soon may lower significantly the cost of producing ethyl alcohol by fermentation. Liquid-liquid extraction is one possible means of accomplishing this separation. Several investigators have obtained experimental data of distribution coefficients and separation factors at high dilution of ethanol for a wide range of solvents (e.g.: Roddy, 1981; Roddy and Coleman, 1981; Munson and King, 1984). However, the design calculations cannot be done using only these data of one tie line for each system, because they depend on the ethanol concentration.
In this work, liquid-liquid equilibrium data covering the whole range of the heterogeneous region of a ternary system water-ethanol-solvent have been determined experimentally. The new data have been integrated into the total needed data set in order to carry out the calculation of the solvent circulation rates and energetic requirements of a suitable extraction process and to examine what properties the solvent should offer in order to decrease these requirements. Depending upon the volability of the solvent, these extraction processes can be divided into two types: those using a solvent less volatile than ethanol (part 1) and those using a solvent more volatile than ethanol (part 2). The solvent chosen in part 1 is 2-ethylhexanol. Ternary liquid-liquid equilibrium (LLE) data for the water (W)-ethanol (E)-2-ethylhexanol (EH) system have been measured at 25 "C and correlated simultaneously together with vapor-liquid equilibrium (VLE) data for the systems water-ethanol and ethanol-2-ethylhexanol. The results obtained allow the design calculations of an extraction process of the type reported by Munson and King (1984). 0 1987 American Chemical Society
Ind. Eng. Chem. Res., Vol. 26, No. 4,1987 697 Table I. Mutual Solubility Data (wt 70)for Water (W)-Ethanol (E)-2-Ethylhexanol (EH) at 25 OC
o
G ~ o s h r I N I T I Uf i i r r u a i
e Eirarnr
OF
0 soL"s,Llr*
7.8 13.4 19.0 23.8 29.2
22.7 32.9 37.8 40.2 41.4
69.5 53.7 43.2 36.0 29.4
35.1 40.7 47.4 55.0
41.6 41.5 40.6 39.0
TIE
' ,*e
POIN-
23.3 17.8 12.0 6.0
Table 11. Tie Line Data (wt % ) for Water (W)-Ethanol (E)-2-Ethylhexanol (EH) at 25 OC initial mixture aqueous phase organic phase La Xw XE XEH Xw XE XEH 0.0 99.9 0.0 0.085 2.42 0.0 97.6 5.0 93.9 6.03 0.10 3.00 4.00 93.0 88.8 11.1 0.12 10.0 3.75 8.60 87.7 15.0 82.5 17.3 0.17 4.63 13.0 82.4 78.1 21.7 0.24 20.0 5.90 18.3 75.8 25.0 73.2 26.4 0.45 7.93 24.0 68.1 30.0 68.7 30.4 10.6 0.88 29.5 59.9 35.0 63.0 35.0 2.03 15.3 35.0 49.7 40.0 54.7 39.1 6.16 23.2 40.3 36.5 "Ethanol level (defined as L = XE in the global initial mixture).
The properties that another solvent should offer to decrease the energetic requirements of the extraction process are studied.
Experimental Section All chemicals (Merck) were used as supplied. The contents of volatile impurities were determined by gas chromatographic analysis: all compounds contained negligible amounts of impurities (less than 0.2 wt %). Data for the binodal curve of the ternary system were determined by using the cloud-point method. The experimental device was that used by Ruiz and Prats (1983). Equilibrium data were obtained by preparing mixtures of known overall composition by weighing the components, stirring intensely, and setting for 2 h at constant temperature (25 f 0.1 "C). At the end of each experiment, samples were taken from both phases and analyzed by means of gas chromatography. Good separation of the three components was obtained on a 2-m X 1/8-in.column packed with Chromosorb 101 100/120. The column temperature was 190 OC, and detection was carried out by thermal conductivity for the organic phases and by flame ionization for the aqueous phases. The helium flow rate was 40 mL/min. To obtain quantitative results, we applied the internal standard method, 1-propanol being the standard compound used for this purpose. Furthermore, the addition of 1-propanol prevents phase separation effects. The relative accuracy of the weight fraction measurements was 2 %. The methodology applied in selecting the points to be determined experimentally was as reported in a previous paper (Ruiz et al., 1984). Initial mixtures were selected such that Xw = XEH,Xibeing the weight percentage of component i; the ethanol levels L were increased stepwise until the homogeneous region was reached. Phase-Equilibrium Data Table I shows the mutal solubility data (weight percent) for the system water (W)-ethanol (E)-2-ethylhexanol (EH) at 25 "C. The tie lines for this ternary system appear in Table 11. The values of L for the initial mixtures are also included. In this way, the overall compositions are known and the accuracy of the measurements can be checked easily since the two equilibrium compositions and the overall composition have to lie on a straight line. Figure
W
H
Figure 1. Phase diagram (wt %) for the ternary system water (W)-ethanol (E)-2-ethylhexanol (EH)at 25 "C.
1shows the phase diagram (weight percent) for this ternary system.
Correlation of LLE VLE Data The simultaneous representation of both vapor-liquid and liquid-liquid equilibria for the ternary system water (W)-ethanol (E)-2-ethylhexanol (EH) is necessary for the design of the separation operation of the components of the extract, since, for at least part of the operation, it is possible that there exist two liquid phases in addition to a vapor phase. Thus, the correlation of the experimental LLE data obtained together with VLE data is needed. The VLE data for the system water-ethanol at 1.013 bar were those obtained by Stabnikov et al. (1972). VLE data for the system ethanol-2-ethylhexanol at 1.013 bar could not be located in the literature and were predicted by using the UNIFAC method developed by Fredenslund et al. (1977). The ethanol molecule has been subdivided into its UNIFAC groups as CH,, CH,, and OH. 2-Ethylhexanol consists of two groups of CH3, five groups of CH2, one group of CH, and one group of OH. The UNIFAC group volume and surface parameters for these groups and the interaction parameters between groups are those given by Gmehling et al. (1982). The UNIQUAC equation (Abrams and Prausnitz, 1975) was used to simultaneously correlate the present LLE and VLE data. The pure component molecular structure constants for the UNIQUAC equation are those given by Prausnitz et al. (1980). The parameter estimation was carried out by minimizing using the Nelder-Mead minimization procedure (1965), with the objective function NL3 2
F = Q1/"LCCC(xijk k
l
i
NV
- Zijk)' + Qz/NVC(P, m
- pJz
(1)
where NV is the number of VLE data points, NL is the number of LLE data points, i denotes the phase (i = 1and 2), j denotes the component 0' = 1-3), k denotes the tie line (k = 1, 2, ...,NL), m denotes the VLE data ( m = l2 2, ..., NV), P represents the experimental pressure, P represents the calculated pressure, x is the experimental composition (mole fraction), f is the calculated composition, and Q1 and Qz are weight factors associated with LLE or VLE. In this work, they were chosen such that the value of the objective function corresponding to LLE was similar to that correspondingto VLE. A ratio Q1/Q2= loo00 was used when the pressure was expressed in bar. For the VLE data, the pressures were calculated assuming ideal vapor phase and setting the pure component
698 Ind. Eng. Chem. Res., Vol. 26, No. 4, 1987 Table 111. Constants of component water ethanol 2-ethylhexanol
the Antoine Eauation A B 1730.63 8.0713 8.1122 1592.86 6.9147 1339.70
C 233.43 226.18 147.81
Table IV. UNIQUAC Interaction Parameters ( K ) for the Ternary LLE and VLE Data: Water (W)-Ethanol (E)-2-Ethylhexanol (EH) W E EH 25.3 w 0.0 154.8 0.0 -7.2 E 3.0 EH 488.2 152.5 0.0
'. Figure 3. Flow sheet of the separation process: P = extraction column; Q = dehydration column; R = solvent recovery distillation column; S, T, and U = heat exchangers. The location of the equilibrium plates is shown.
Figure 2. Comparison between the correlated and calculated LLE and VLE data for the system water (W)-ethanol (E)-2-ethylhexanol (EH). Points represent the correlated data and lines the calculated ones.
fugacities equal to the vapor pressures. Vapor pressures ( P ,in mm Hg) of the three components were calculated by the Antoine equation log P = A - B / ( C + T) (2) where T i s the temperature ("C). The Antoine constants for the three components are listed in Table 111. For the LLE data, the calculated compositions were predicted by setting the concentration in the global initial mixture equal to that in the middle of the corresponding experimental tie line such as reported in a previous paper (Ruiz and Gomis, 1986). The estimated UNIQUAC binary interaction parameters for the ternary system water (W)-ethanol (E)-2-ethylhexanol (EH) are presented in Table IV. In Figure 2, the calculated data using these parameters are represented. As a comparison, the same figure shows the data used in the correlation (experimental data for LLE of waterethanol-2-ethylhexanol (W-E-EH) and VLE of W-E and predicted ones by UNIFAC for VLE of E-EH). The results indicate that the UNIQUAC equation gives a good correlation of the data. The obtained UNIQUAC parameters allow the calculation of VLE, LLE, and vapor-liquid-liquid equilibrium (VLE) data for this ternary system. To carry out these calculations, the conventional methods for two phases (dew-point, bubble-point, and two-liquid-phase flash calculation methodologies) and for three phases (combinated two-liquid-phase flash, dew-point, and bubble-point calculation methodologies) have been used. Separation Process Example The UNIQUAC correlations of the VLE and LLE data were used to prepare and calculate a conceptual flow sheet
which is shown in Figure 3. The flow sheet was described by Munson and King (1984), and it is similar to those for recovery of acetic acid by extraction with high-boiling
Table V. Flow Rate (mol/s) and Compositions (mol % ) of All the Streams of Figure 3 flow rate, stream mol/s W, mol 70 E, mol % EH, mol % 5 80 20 4.83 22 28 50 4 99.97 0.02 0.01 2.45 100 5.93 100 75 25 5.36 1.41 75 25 8 100 2.20 9 1 100 10 6.93 14.4 85.6 11 100 8.35
solvents reported by Othmer (1958) and Brown (1963). Stream 1 containing 20 mol % ethanol (about 40% alcohol by volume) and coming from the beer still of a typical fermentation ethanol manufacturing plant is fed to extraction column P, which produces a raffinate (stream 3) containing less than 0.0002 mol % ethanol. The output extract (stream 21, after being heated close to its bubble point in heat exchanger S, goes to extractive distillation column Q where it is dehydrated. The bottoms stream from column Q (stream 10) containing only ethanol and solvent is transferred to distillation column R where it is separated into a overhead stream containing only ethanol (stream 9) and a bottoms stream containing the solvent (stream 11). A portion of this (stream 4) is recycled to extraction column P, and the other portion (stream 5) is cooled to its bubble-point temperature and recycled to the upper section of extractive distillation column Q to keep water more volatile than ethanol on the upper stages. Heat exchangers T and U cool streams 4 and 7 to 25 "C which are fed to extraction column P. Process calculations were carried out by using the McCabe-Thiele graphical design procedures for both dehydration column Q and solvent recovery distillation column R. Bubble-point feeds and constant molar overflow were assumed. An operating reflux ratio of 1.5 times the minimum reflux ratio was used. Since the volatility of 2-ethylhexanol is very low compared to water and ethanol, the system water-ethanol-2-ethylhexanol in dehydration column Q can be considered as pseudobinary. The mole percentages of 2-ethylhexanol are constant in both the rectifying section (60%) and the stripping section (57%) of column Q. The pseudobinary equilibrium curves were obtained by using the UNIQUAC model and the parameters obtained in the correlation. VLE and VLLE data were used. For extraction column P, the graphical method of Hunter and Nash (1932) was used. The flow
I n d . Eng. Chem. Res. 1987,26, 699-705
rates and compositions of all the streams of the process are summarized in Table V. The calculated number of equiIibrium stages and the feed plate location are shown in Figure 3 on each column. The greatest energetic requirements of the process are the latent heat for the vaporization of stream 6 in dehydration column Q (about 5.36 X 40 = 224 kJ/mol of ethanol obtained; the latent heats for water and ethanol are about 40 kJ/mol) and in solvent recovery distillation column R (about 2.2 X 40 = 88 kJ/mol) and a sensible heat of stream 4 (about 0.26 X 2.45 X (184 - 25) = 101 kJ/mol; the heat capacity and boiling temperature of 2-ethylhexanol are 0.26 kJ/(mol "C) and 184 OC,respectively). In addition a significant fraction of the required latent and sensible heat input can be recycled through heat exchangers which are not shown. These energetic requirements represent an important fraction of the energy that would be liberated if the alcohol is burnt as a fuel (1360 kJ/mol). However, it can be possible to find another solvent with properties more favorable than 2-ethylhexanol. These properties should be those decreasing the flow rates of streams 4, 6, and 8. (1) A greater distribution coefficient for ethanol is needed to decrease the solvent-to-water ratio and so decrease stream 4. (2) A greater separation factor is needed to decrease the reflux ratio in dehydration column Q and so decrease stream 6. (3) The enhancement by the solvent of the relative volatility of water to ethanol should be greater in order to decrease both the reflux ratio in the dehydration column (and so decrease stream 6) and stream 5, since less flow rate of 2-ethylhexanol is necessary to keep the water more volatile than ethanol. (4) The solvent must have less heat capacity and a lower boiling point to decrease the enthalpy of stream 11 (and therefore the enthalpy of stream 4). As proposed by Munson and King (19841, a diluent can be incorporated
699
into the solvent so as to provide volatility in the reboiler of solvent recovery distillation column R. However, the solvent volatility must be sufficiently less than that of ethanol to facilitate the separation in this column. (5) The solubility of the solvent in water must be low enough so that it is not necessary to remove and recover the residual solvent from the aqueous raffinate (stream 3) leaving the extractor. Registry No. H 3 C C H 2 0 H , 64-17-5; H3C(CH2)3CH(CHZCHJCHZOH, 104-76-7.
Literature Cited Abrams, D. S.; Prausnitz, J. M. AIChE J. 1975,21, 116. Brown, W. V. Chem. Eng. Prog. 1963, 59(10), 65. Fredenslund, A.; Gmehling, J.; Rasmussen, P. Vapor-Liquid Equilibria Using UNZFAC; Elsevier: Amsterdam, 1977. Gmehling, J.; Rasmussen, P.; Fredenslund, A. Znd. Eng. Chem. Process Des. Dev. 1982, 21, 127. Hunter, T. J.; Nash, A. W. J . SOC.Chem. Znd. 1932, 51, 285T. Ladisch, M. R.; Dyck K. Science (Washington,D. C.) 1979,205,898. Munson, C. L.; King, C. J. Ind. Eng. Chem. Process Des. Dev. 1984, 23, 109.
Nelder, J. A.; Mead, R. Comput. J . 1965, 7, 308. Othmer, D. F. Ind. Eng. Chem. 1958, 50(3),60A. Prausnitz, J.; Anderson, T.; Grens, E.; Eckert, C.; Hsieh, R.; O'Connell, J. Computer Calculations for Multicomponent Vapor-Liquid and Liquid-Liquid Equilibria; Prentice Hall: Englewood Cliffs, NJ, 1980. Roddy, J. W. Ind. Eng. Chem. Process Des. Deu. 1981, 20, 104. Roddy, J. W.; Coleman, C. F. Ind. Eng. Chem. Fundam. 1981,20, 250. Ruiz, F.; Gomis, V. Ind. Eng. Chem. Process Des. Dev. 1986,25(1), 216. Ruiz, F.; Prats, D. Fluid Phase Equilib. 1983, 10, 77. Ruiz, F.; Prats, D.; Gomis, V. J . Chem. Eng. Data 1984, 29, 147. Stabnikov, V. N.; Matyushev, B. Z.; Protsyuk, T. B.; Yushchenko, N. M. Pishch. Prom. 1972, 15, 49.
Received f o r review October 25, 1985 Revised manuscript received September 22, 1986 Accepted December 6, 1986
Progressing Batch Hydrolysis Reactor J o h n D. Wright,* Paul W. Bergeron, and Pamela J. Werdene Solar Energy Research Institute, Golden, Colorado 80401-3393
In the dilute acid hydrolysis of lignocellulose to produce fermentable sugars, conditions severe enough to hydrolyze crystalline cellulose are also severe enough to degrade the product sugars. Countercurrent flow of liquids and solids minimizes sugar degradation and product dilution by removing the sugars from the reaction zone before substantial degradation can occur. The progressing batch reactor uses several percolation reactors in series to simulate countercurrent flows while retaining the simplicity of the percolation reactor. In all dilute acid hydrolysis processes for glucose production, conditions severe enough to hydrolyze crystalline cellulose to glucose are also severe enough to degrade the glucose into undesirable compounds such as hydroxymethylfurfural (HMF), levulinic acid, and formic acid. One way to minimize the sugar degradation is to remove the sugars from the reaction zone before substantial degradation occurs. Sugars are most efficiently removed by a reactor system that uses countercurrent flow of liquids and solids, which allows simultaneous achievement of high yields and high sugar concentrations. The progressing batch hydrolysis process, invented and now under development at the Solar Energy Research Institute (SERI), 0888-5885/87 / 2626-0699$01.50/ 0
uses several percolation reactors in series to simulate countercurrent flow of liquids and solids. In this way, the advantages of countercurrent flow are achieved, and the mechanical and operational simplicity of the percolation reactor is retained. This paper describes the theory and operation of the progressing batch hydrolysis reactor and presents the results of our mathematical modeling of the system. Theory of Operation Hydrolysis Kinetics. The high-temperature acid hydrolysis of the carbohydrate portion of lignocellulose can 0 1987 American Chemical Society