col
Content
Desalination 223 (2008) 417–426
Separation of alcohol–water solutions by distillation through hollow fibers
Guoliang Zhanga,b*, Lan Linb, Qin Menga, Youyi Xua
a
College of Materials Science and Chemical Engineering, Zhejiang University, Hangzhou 310027, P.R. China b College of Biological and Environmental Engineering, Zhejiang University of Technology, Hangzhou 310014, P.R. China Tel/Fax +86(571)88320863; email: [email protected]
Received 24 December 2006; accepted 3 January 2007
Abstract New results of investigation on distillation in hollow fiber structured packing are discussed. The membranes used here for separation of methanol, ethanol and isopropanol-water solutions are non-porous hollow fibers, unlike those of membrane distillation. The fibers are non-selective with coating and have little resistance to mass transfer. Because liquid flows inside the lumens of the fibers and vapor flows counter currently outside fibers, the distillation process can avoid flooding effectively even at very high flows. As a result, a mole fraction of alcohol equal to 0.04 was prepared as feed in all of the three solutions, and an attractive low value of the height of mass transfer unit (HTU) less than 0.1 m could be obtained at total reflux. Although the distillate concentration drops as the heating rate rises, as noted before, the decreasing rate of concentration and HTU in the same module seems to change inversely with the molecular weight of alcohols. Compared with normal structured packing, the ordinate data of common capacity factor of hollow fiber column calculated on the Eckert version of GPDC correlation were 3–10 times higher above flooding. All these offered the possibility of distillation with better, more productive separations. Keywords: Distillation; Hollow fiber membrane; Alcohol–water solution; Separation
1. Introduction This paper continues to explore the distillation process with new structured packing column, a new type of membrane contactors. As is well
*Corresponding author.
known, conventional packed distillation columns always have some obstacles in selection of packing material and uniform distribution of fluids before they enter the packed bed. The interdependence of the two-fluid phases which are contacted very closely often leads to operational problems such as emulsions, foaming, unloading and flooding [1–3]. Using hollow fiber module is
Presented at the conference on Desalination and the Environment. Sponsored by the European Desalination Society and Center for Research and Technology Hellas (CERTH), Sani Resort, Halkidiki, Greece, April 22–25, 2007.
0011-9164/06/$– See front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.desal.0000.00.000
418
G. Zhang et al. / Desalination 223 (2008) 417–426
an alternative effective way to overcome these disadvantages and the relative experimental results are proved to be promising according to Zhang and Cussler [4,5]. Unlike most packed towers, fluids are contacted flow on opposite sides of membrane and the liquid/liquid interface forms at the mouth of each pore. Further, the hollow fibers are supposed to be non-selective by choosing suitable membranemaking and coating materials and therefore be different with earlier method called “membrane distillation” [6–8]. The hollow fiber distillation process shown in Fig. 1 can be potential route for faster separation, even if there is no direct commercial value now. Such a hollow fiber column has three advantages over conventional packed towers. First, it offers substantially more interfacial area than conventional approaches by non-dispersive contacting via a microporous membrane. A typical interfacial area per volume is 201 ft2/ft3, almost triples the value of Sulzer Mellapak structured packing which is only 76 ft2/ft3. Second, because liquid flows inside the fibers and vapor flows counter currently outside fibers, the two-fluid flows are independent and can be easily kept uniform distribution to avoid flooding effectively even at very high flows. Moreover, the flows of liquid and vapor are now not around submerged objects and only a very small pressure drop across the membrane is required to ensure that the liquid/vapor interface remains immobilized at the mouth of the pore. Third, unlike traditional packed towers, no density difference between fluids is required and theoretically hollow fiber modules can be operated in any orientation. Our aim is to improve product purity and separation efficiency by using different non-selective hollow fiber membranes in different systems. As usual, we hope to maximize mass transfer rate by optimizing our design and process operation. Here, the big interfacial area will be beneficial to make more efficient separation, according to Bennet and Kovak [9] and Kellehar and Fair [10],
Condenser
Gauged cylinder
Hollow fiber contactor
Heater XD
Liquid Vapor Reboiler
XW
Fig. 1. Schematic process for hollow fiber structured packing distillation. The hollow fiber module replaces the conventional packed tower used for differential distillation of alcohol–water systems. The structure of new module tower in which liquid flows inside fibers and vapor rises counter currently outside fibers leads to very different operational characteristics.
which means the columns of hollow fiber structured packing will perform better, with the height of transfer unit (HTU) as small as possible. In order to do this, we will take three alcohol systems including methanol, ethanol and isopropanol–water solutions for separation experiments. We want to see whether the performance of the modules worked formerly can keep well or not and find how different alcohols affect the separation.
G. Zhang et al. / Desalination 223 (2008) 417–426
419
2. Theoretical considerations The theoretical analysis is based on the mass balances and transfer unit theory basic to the operation of a conventional differential distillation column. We begin our work with the same data on estimates of vapor-liquid equilibrium as differential and staged distillation [3,11]. For simplicity, the column of hollow fibers is operated at total reflux, which means the total vapor and liquid fluxes are equal in the column. The relative balances on the more volatile species in both vapor and liquid give the operating line y=x (1)
theory has been used for many years to design absorption columns in industry, the challenge has always been to design an appropriate height for the packed column. The height of an absorption column l is given by l = NTU × HTU (4)
where HTU is the height of a transfer unit. In our experiment, since the length of modules is set, the true value of HTU can be easily estimated by HTU = l/NTU (5)
where y and x are the mole fraction of the more volatile species in the vapor and the liquid, respectively. Then we can get a balance on the more volatile species in the gas phase.
0 = − vG
dy − Ka ( y − y*) dz
(2)
where vG is the gas velocity, y and y* are the gas phase mole fractions in the actual vapor and in the vapor which is in equilibrium with the liquid, z is the distance measured from the bottom of the column, K is the overall mass transfer coefficient based on the vapor side, and a is the interfacial area per volume between vapor and liquid. Integrate this equation, we can easily find the number of transfer units (NTU) for our column
NTU =
∫
yl y0
dy Kal = y − y* vG
(3)
where y0 and yl are the specified vapor compositions at the bottom (z = 0) and at the top (z = l) of the column, l is the length of the hollow fiber module. As usual, the NTU calculated by integration of the experimental data is a good measure of the difficulty of the distillation and shows how high the separation efficiency is. A larger value means an easier separation, and a smaller one signals a more difficult separation. As transfer unit
Since the height of a transfer unit HTU can be defined as the height of a packed section required to accomplish a change in concentration equal to the average driving force in that section, the smaller the HTU is, the better design should be. HTU normally is a measure of industrial tower or column’s separation efficiency. A still smaller HTU means more efficient and more rapid separation process, and that’s what we dream about. The technology route of a differential distillation tower structurally packed by hollow fibers is similar to that of conventional one, as Kister [3] described, but the process is very different. As noted above, we start our experiment with a very low feed concentration, and the desired concentrations of distillate and bottom. The liquid and vapor flows are very independent because of the non-porous hollow fiber walls. The twophase flow and maldistribution in conventional process can then be effectively prevented. The liquid and vapor can be pumped very freely with almost no pressure drop coming from fluid friction, and still not be bothered by loading and flooding. Although we have little experience to guide our choice on column designing and material choosing, we think it important to choose the structure and dimension of the column’s internals, and we will specify the relative difficulty of the separation of different alcohol systems as NTU. We can analyze our experiment using the same
420
G. Zhang et al. / Desalination 223 (2008) 417–426
equations as given above. After measuring the concentrations y0 and yl and calculating the NTU value, we can estimate mass transfer coefficients from the NTU and hollow fiber geometry and find if they are different in different separation systems or with empirical values of membrane contactors elsewhere. 3. Experimental The microporous polyether sulfone hollow fibers (Porous Media, MN, USA and RDCWTT, Hangzhou, China) used in this work have an inside diameter of 0.07 cm and an outside diameter of 0.11 cm. A thin layer of 5 µm polydimethylsiloxane was coated on the membrane for stopping convection. Thus liquid flow could not pass through the wall to the vapor side while vapor could flee out freely. From the earlier results of Zhang and Cussler [4,5], we chose 25 fibers in a glass shell with inside diameter of 1.40 cm as module design basis. Each module with two nozzles had an effective mass transfer length of 25 cm. The module properties are compared in Table 1 with commercial structured packings – Sulzer Mellapak [12]. The void fractions of the hollow fiber modules are lower than those of the conventional structured packing, while the areas per volume are larger, according to Yang and Cussler [13] and Kistler and Cussler [14]. The hollow fibers were potted with SG-856 epoxy
(Jinpeng Chem., Zhejiang, China) and HC 703 silicone (Xiguang, Wuxi, China). All distillation experiments were run at total reflux. In a typical experiment, all solvents including methanol, ethanol and isopropanol (Sinopharm Chemical Reagent Co., Ltd, Shanghai, China) were at least HPLC-UV grade and used as received, and an alcohol solution with a molar fraction of 0.04 was prepared as feed with doubly distilled water. About 800 cm3 of this solution was added to the reboiler and heated in a 350 w 98-1-C digital auto-control heating mantle (Taisite Co., Tianjin, China). The heating rate was adjusted by a FATO SVC-500 VA automatic voltage regulator and measured by a JL 4006B electric multimeter (Junling Co., Hangzhou, China). Then the vapor generated flew up the shell side with liquid reflux from condenser dropped down inside fibers concurrently (Fig. 1). All flows were measured volumetrically. Solution concentrations were measured as 1 cm3 samples taken from the reboiler and from the distillate, respectively. The concentrations were determined by injecting 2 μL samples into a SF GC-1102 gas chromatograph equipped with a temperature conduct detector (TCD) and a Super Porapak Q steel column. The carrier gas was hydrogen of purity 99.99%. Each concentration was measured at least in duplicate. To begin the experiment, we usually ran the column at a heating rate for about 1–2 h to approach steady state, when the first sample was taken. Subsequent samples
Table 1 Properties of hollow fiber modules and structured packing Module packing Module 1 Module 2 Sulzer Mellapak Sulzer BX Raschig Rings Raschig Rings Size or type 25 (number of fibers) 50 (number of fibers) 250Y BX 500 1 inch 1/2 inch Void fraction, e 0.85 0.71 0.95 0.90 0.78 0.70 Area per volume, a (ft2/ft3) 201 484 76 150 58 101 Packing factor, FP (a/e3) 327 1352 89 206 179 580
G. Zhang et al. / Desalination 223 (2008) 417–426
421
were then taken at 15–30 min intervals to ensure steady state. The calculation of the NTU from these data is complicated because the slope of each equilibrium line for different alcohol systems is not constant, so that m is not constant. Over the concentration range of 0.1 < x < 0.9, the vapor–liquid equilibrium fit the polynomials and a data relationship can be expressed mathematically as below [15]. Equation for methanol (no azeotropic point), 0.1 < x ≤ 0.3 y* = 0.1754 + 2.8545x − 4.0810x2 0.3 < x < 0.9 y* = 0.4823 + 0.6679x − 0.1571x2
(6)
(7)
Equation for ethanol (azeotropic concentration x = 0.881) 0.17 < x ≤ 0.33 y* = 0.348 + 1.1821x − 1.4292x2 0.33 < x < 0.68 y* = 0.4977 + 0.1791x − 0.2599x2 (8)
in different alcohol systems, finding some regularity and achieving the lowest value of the height of a transfer unit as possible as we can. As before, we can easily operate hollow fiber at flows greater than that causing flooding, as shown in Fig. 2. The relationship between the flooding gas velocity and other physical properties of the system is presented in the form of an empirical, experimental correlation for pressure drop in packed columns. This method was proposed by Sherwood in 1938 and improved by Eckert in 1970 to determine the flood point, which basically remains the standard procedure today [1,3,11]. We plot the normal flow parameter on the abscissa, and the common capacity factor on the ordinate. The solid curve is the empirically determined limit of flooding. For conventional packing, the gas velocity in an operating column must obviously be lower than the flooding velocity and is often chosen as one-half the predicted
1
(9)
Y = (uo2FPρ Gf { ρ L} { μ L})/(gρ H O(L))
2
Equation for isopropanol (azeotropic concentration x = 0.675) 0.13 < x ≤ 0.33 y* = 0.4438 + 0.5449x − 0.6984x2 0.33 < x < 0.79 y* = 0.5678 − 0.2586x + 0.6278x2 (10)
0.1
0.01 Methanol Ethanol Isopropanol
(11)
Using these relations and Eq. (3), the NTUs and HTUs can be easily found for comparison with different separation systems operated in the same module. The results are given below. 4. Results and discussion This section reports distillation separations using a structure packing of hollow fibers. We are especially interested in confirming its behavior
0.001 0.01
0.1 X = (LML/GMG) (ρG/ρL)
1
0.5
10
Fig. 2. Hollow fiber distillation operate above flooding. Like conventional distillation process, the flow parameter on the abscissa is plotted vs. the capacity factor on the ordinate. All the data of methanol, ethanol and isopropanol–water systems are at fluxes high above the normal flooding limit, shown as the solid line.
422
G. Zhang et al. / Desalination 223 (2008) 417–426
flooding velocity obtained from a generalized correlation. So points above this solid curve cannot be reached to ensure safe operation. In Fig. 2, we also show the data points of our experiments with hollow fibers in different alcohol solutions. Without exception again, all the data points of methanol, ethanol and isopropanol–water systems are high above the limit solid curve where flooding normally occurs. The highest data is 3–5 times above the solid flooding line. Since we always ran at total reflux, these data only cover a narrow range of abscissa in which they proceed in the order isopropanol > ethanol > methanol. But the results clearly show how easily hollow fiber structured packing can operate above the normal flooding point. In conventional cases, people must choose a velocity far from the flooding velocity but not so low as to require a much larger column which will consume more cost of power, and lower gas–liquid velocities often lead to a nearly proportional reduction in mass transfer rate. In our experiments, we could potentially operate at still higher flow rates since vapor and liquid are in contact only across the hole of membrane walls. There is no resistance for a bubble rising through liquid, and the falling liquid is not slowed by the rising vapor, just as other membrane contactors do. That is where the hollow fiber columns always have an advantage on. In order to compare the operational performance of hollow fiber with normal structural packing in distillation process, the F-factor is plotted vs. the NTU per meter (Fig. 3). The data points here for same hollow fiber module in different alcohol–water systems are compared with a type of Sulzer structured packing (Sulzer Chemtech), shown as the solid lines. The abscissa shows a flow parameter F, which tends to cover about the same range for different total pressures in conventional distillation, since the flooding velocity varies with rG1/2. Parameter F, generally used to characterize performance of distillation packing and known as the F-factor, see Refs. [3,12], is defined as
F = uG r G1/2
(12)
where uG is the vapor velocity, rG is the vapor density. The data in Fig. 3 show how the NTU increase as the F-factor gets smaller. This figure also shows some of the results under conditions where the normal packing will flood, and the separation efficiency estimated from NTU can strongly compete with some excellent industrial structured packing such as Sulzer Mellapak. We next turn to find how the hollow fiber module works as flows changes from high to low. From Eq. (3), the reciprocal of the NTU (1/NTU) should be proportional to the vapor velocity vG. This is partly supported by the data in Fig. 4 when vapor velocity becomes high for all the three systems. At low vapor velocity, the NTU keeps a relative high value for some period. We are not sure whether there are some limits with the geometry of our column design or it is just
4.00
3.00
NTU (1/m)
2.00
Methanol Ethanol 1.00 Isopropanol Mellapak
0.00 0.1
1
10
Ff at or = uGρ G1/2 (m/s(kg/m3)1/2)
Fig. 3. Operational performance of hollow fiber and normal structural packing in distillation process. The F-factor is plotted vs. the number of transfer units. The data for same hollow fiber module in different alcohol– water systems are compared with a Sulzer structured packing, shown as the solid line.
G. Zhang et al. / Desalination 223 (2008) 417–426
2.0 Methanol Ethanol Isopropanol 1.6 2.0 Methanol Ethanol Isopropanol y = 309.97x + 0.0688 y = 132.49x + 0.1304 y = 127.58x + 0.2367
423
1/NTU
1.2
1/NTU
0.8 0.4 80 120 160 200 240 0 0 0.001 0.002 0.003 0.004 0.005 0.006
1.0
0.0
Vapor Velocity (cm/s)
Fig. 4. The number of transfer units (NTU) vs. vapor velocity. The slower the vapor flows, the higher NTU value can be got, which means more efficient separation. Further, the data curve for lower molecular weight alcohol such as methanol is steeper than ethanol and isopropanol, and very close to linear when vapor velocity surpass 140 cm/s.
Liquid Flow (mol/s)
Fig. 5. The number of transfer units (NTU) vs. liquid flow. The reciprocal of the NTU is proportional to the molar liquid flow. Higher NTU value can be got when the liquid flow becomes slower. The slope of the linear relationship is in the order of methanol > ethanol > isopropanol.
caused by heat losses due to the simple jacket insulation. But the trend does mean that we can obtain faster separation and higher purity of alcohols in our hollow fiber column simply by reducing the flows. We can use slow flows without reducing the interfacial area just by making sure the liquid inside fibers are full and can run continuously. Further, the NTU for lower molecular weight alcohol such as methanol system separation increases faster than ethanol and isopropanol, but at same vapor velocity, the NTU for higher molecular weight alcohol like ethanol and isopropanol looks larger, which means more efficient separation. Since the vapor velocity in our experiments is calculated directly from heat balance, as described by Poling et al. [16], and there is some insulation problem which may cause heat loss, especially when the counter current flow is slow, we use data read from gauged cylinder and plot Fig. 5 to show change of the NTU vs. liquid flow more precisely. From Eq. (4), we have a similar form of the NTU rewritten as
NTU =
∫
x1 x0
dx ⎡K a⎤ = ⎢ x ⎥l * x − x ⎣ L ⎦
(13)
where Kx is an overall mass transfer coefficient based on a liquid side. Now the results in Fig. 5 meet our expectation from Eq. (13) very well, the reciprocal of the NTU is proportional to the molar liquid flow. As above, higher NTU value can be got when the liquid flow becomes slower, and the slope of the linear relationship is in the order of methanol > ethanol > isopropanol. This gives the overall mass transfer coefficient Kx for methanol, ethanol and isopropanol of 2.45 × 10−5, 5.73 × 10−5 and 5.95 × 10−5, respectively, in accordance with the same order as former results, according to Zhang and Cussler [4,5] and Chung et al. [17]. The intercept on this plot may be dependent on the properties of flow, module configuration and the heat loss of process. To reassure our results, we show the variation of distillate concentration with heating rate in Fig. 6. The higher distillate concentrations
424
1.0
G. Zhang et al. / Desalination 223 (2008) 417–426
0.40
0.8
Methanol
0.30
Distillate Concentration (mol)
Ethanol Isopropanol
HTU (m) Methanol Ethanol Isopropanol
0.6
0.20
0.4
0.10
0.2
0.00
0 80 100 120 140 Heat Rate (W) 160 180 200
80
100
120
140 Heat Rate (W)
160
180
200
Fig. 6. Distillate concentration vs. heating rate. As the heating rate increases, the distillate concentration at the top of module decreases.
Fig. 7. The height of a transfer unit (HTU) vs. heating rate. Because the heating rate is proportional to residence time in the column, the height of a transfer unit drops as the heating rate rises.
observed at the lower heating rate are consequence of better mass transfer in vapor/liquid interface, and the relative sensitivity of distillate concentration change with heating rate is methanol > ethanol > isopropanol. As we know, there is an azeotropic point for both ethanol and isopropanol solution, such as isopropanol–water azeotrope forming at an isopropanol mole fraction of 0.675, and at present we cannot break through it with our owned modules in the experiments, that means separation of methanol solution without azeotrope formation will be simpler and easier. Finally, the height of a transfer unit is plotted vs. heating rate in Fig. 7. Because the heating rate is proportional to residence time in the column, the height of a transfer unit drops obviously as the heating rate rises while the contacting time becomes shorter. In our experiments, a low-mole fraction of alcohol as low as 0.04 was prepared as feed in all of the three solutions, and an attractive low value of the HTU less than 0.1 m could be easily obtained at total reflux. The data in Fig. 6 also show the same behavior as the above, the
decreasing rate of the height of mass transfer unit in the same hollow fiber module seems to change inversely with the molecular weight of alcohols. Although the above results have already shown a promising future for fast and efficient separation in hollow fibers, we hope to do better. In conventional distillation, packing always wins by high surface area per unit volume and high void fraction, so the hollow fiber can go farther with better characteristics on it. As before, the physical properties of the system, the liquid/gas flow and the packing controls the performance of conventional packed column. That means the geometry of packing, usually characterized as ‘Packing factor’, is critical for process designing, see refs. [3,9,11]. Packing factor FP is defined as
FP =
a e3
(14)
where e is the void fraction of the packing. To get a bigger value of this packing factor, the packing in a set column should be chosen smaller because FP is roughly proportional to the packing
G. Zhang et al. / Desalination 223 (2008) 417–426
425
size (see Table 1). But even the smallest conventional packing cannot get a higher packing factor than hollow fibers whose interfacial area per volume is large. In our experiments, a same size module gives a packing factor of 327 ft−1 with 25 fibers and 1352 ft−1 with 50 fibers, the amplifier factor is very big. That means we still have huge space to optimize our separation in hollow fibers. We look forward to more productive distillation over a wider operational range and provide more significant energy savings in the near future. 5. Conclusions New results have been introduced which show the separation of methanol, ethanol and isopropanol–water solutions with distillation using hollow fiber structured packing. The membranes used here for are non-porous hollow fibers, unlike those of membrane distillation. The fibers are non-selective with coating and have little resistance to mass transfer. Because liquid flows inside the lumens of the fibers and vapors flows counter currently outside fibers, the distillation process can avoid flooding effectively even at 3–10 times higher flows than conventional cases. With a mole fraction of alcohol as low as 0.04 prepared as feed in all of the three solutions, an attractive low value of the HTU less than 0.1 m can be easily obtained at total reflux. Moreover, while the distillate concentration drops as the heating rate rises, just as noted before, the decreasing rate of distillate concentration and HTU in the same module seems to change inversely with the molecular weight of alcohols. Compared with conventional structured packing, the hollow fibers have a better geometry of packing and can realize a more productive distillation and will save more energy in separation. Further experiments will focus on the optimization of module structure and process design and using more systems to validate our investigation.
Acknowledgments The authors benefited from conversation with Dr. E.L. Cussler during visiting CEMS, UMN. This work was primarily funded by the Research Fund of the National Natural Science Foundation of China (grant 20476096). Other financial support came from the Zhejiang Provincial Bureau of Science and Technology, China (grants 2005C33040 and 2006C23067).
References
[1] [2] J.-D. Seader and E.J. Henley, Separation Process Principles, Wiley, New York, 1998. J.-L. Humphrey and G.-E. Keller, Separation Process Technology, McGraw-Hill, New York, 1997. H.-Z. Kister, Distillation Design, McGraw-Hill, New York, 1992. G. Zhang and E.-L. Cussler, Distillation in hollow fibers, AIChE J., 49 (2003) 2344–2351. G. Zhang and E.-L. Cussler, Hollow fibers as structured distillation packing, J. Membr. Sci., 215 (2003) 185–193. K.-W. Lawson and D.-R. Lloyd, Membrane distillation, J. Membr. Sci., 124 (2000) 1–25. G. Alan and S.-T. Hwang, Hollow fiber membrane contactors, J. Membr. Sci., 159 (1999) 61–106. M.-A. Izquierdo-Gil and G. Jonsson, Factors affecting flux and ethanol separation performance in vacuum membrane distillation, J. Membr. Sci., 214 (2003) 113–130. D.-L. Bennet and K.-W. Kovak, Optimize distillation columns, Chem. Eng. Prog., 19 (2000) 84–96. T. Kellehar and J.-R. Fair, Distillation studies in a high gravity contactor, Ind. Eng. Chem. Res., 35 (1996) 4646–4655. E.-L. Cussler, Diffusion, Cambridge University Press, Cambridge, 1997. Sulzer Chemtech, Structural packings for distillation and absorption, Product Bulletin 22,13,06 40–111, (2001) 00–70. M.-C. Yang and E.-L. Cussler, Designing hollow fiber contactors, AIChE J., 32 (1986) 1910–1920.
[3] [4] [5]
[6] [7] [8]
[9]
[10]
[11] [12]
[13]
426
G. Zhang et al. / Desalination 223 (2008) 417–426
[16] B.-E. Poling, J.-M. Prausnitz and J.-P. O’Connell, Gases and Liquids, 5th edn., McGraw-Hill, New York, 2000. [17] J.-B. Chung, J.-P. Debrocher and E.-L. Cussler, Distillation with nanoporous or coated hollow fibers, J. Membr. Sci., 257 (2005) 3–10.
[14] K.-A. Kistler and E.-L. Cussler, Membrane modules for building ventilation, Trans. Inst. Chem. Eng., 80 (2002) 53–64. [15] J. Gmehling and U. Onken, Vapor-Liquid Equilibrium Data Collection, Vol. 1, Part 1, DECHEMA, Frankfurt, 1977.
Separation of alcohol–water solutions by distillation through hollow fibers
Guoliang Zhanga,b*, Lan Linb, Qin Menga, Youyi Xua
a
College of Materials Science and Chemical Engineering, Zhejiang University, Hangzhou 310027, P.R. China b College of Biological and Environmental Engineering, Zhejiang University of Technology, Hangzhou 310014, P.R. China Tel/Fax +86(571)88320863; email: [email protected]
Received 24 December 2006; accepted 3 January 2007
Abstract New results of investigation on distillation in hollow fiber structured packing are discussed. The membranes used here for separation of methanol, ethanol and isopropanol-water solutions are non-porous hollow fibers, unlike those of membrane distillation. The fibers are non-selective with coating and have little resistance to mass transfer. Because liquid flows inside the lumens of the fibers and vapor flows counter currently outside fibers, the distillation process can avoid flooding effectively even at very high flows. As a result, a mole fraction of alcohol equal to 0.04 was prepared as feed in all of the three solutions, and an attractive low value of the height of mass transfer unit (HTU) less than 0.1 m could be obtained at total reflux. Although the distillate concentration drops as the heating rate rises, as noted before, the decreasing rate of concentration and HTU in the same module seems to change inversely with the molecular weight of alcohols. Compared with normal structured packing, the ordinate data of common capacity factor of hollow fiber column calculated on the Eckert version of GPDC correlation were 3–10 times higher above flooding. All these offered the possibility of distillation with better, more productive separations. Keywords: Distillation; Hollow fiber membrane; Alcohol–water solution; Separation
1. Introduction This paper continues to explore the distillation process with new structured packing column, a new type of membrane contactors. As is well
*Corresponding author.
known, conventional packed distillation columns always have some obstacles in selection of packing material and uniform distribution of fluids before they enter the packed bed. The interdependence of the two-fluid phases which are contacted very closely often leads to operational problems such as emulsions, foaming, unloading and flooding [1–3]. Using hollow fiber module is
Presented at the conference on Desalination and the Environment. Sponsored by the European Desalination Society and Center for Research and Technology Hellas (CERTH), Sani Resort, Halkidiki, Greece, April 22–25, 2007.
0011-9164/06/$– See front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.desal.0000.00.000
418
G. Zhang et al. / Desalination 223 (2008) 417–426
an alternative effective way to overcome these disadvantages and the relative experimental results are proved to be promising according to Zhang and Cussler [4,5]. Unlike most packed towers, fluids are contacted flow on opposite sides of membrane and the liquid/liquid interface forms at the mouth of each pore. Further, the hollow fibers are supposed to be non-selective by choosing suitable membranemaking and coating materials and therefore be different with earlier method called “membrane distillation” [6–8]. The hollow fiber distillation process shown in Fig. 1 can be potential route for faster separation, even if there is no direct commercial value now. Such a hollow fiber column has three advantages over conventional packed towers. First, it offers substantially more interfacial area than conventional approaches by non-dispersive contacting via a microporous membrane. A typical interfacial area per volume is 201 ft2/ft3, almost triples the value of Sulzer Mellapak structured packing which is only 76 ft2/ft3. Second, because liquid flows inside the fibers and vapor flows counter currently outside fibers, the two-fluid flows are independent and can be easily kept uniform distribution to avoid flooding effectively even at very high flows. Moreover, the flows of liquid and vapor are now not around submerged objects and only a very small pressure drop across the membrane is required to ensure that the liquid/vapor interface remains immobilized at the mouth of the pore. Third, unlike traditional packed towers, no density difference between fluids is required and theoretically hollow fiber modules can be operated in any orientation. Our aim is to improve product purity and separation efficiency by using different non-selective hollow fiber membranes in different systems. As usual, we hope to maximize mass transfer rate by optimizing our design and process operation. Here, the big interfacial area will be beneficial to make more efficient separation, according to Bennet and Kovak [9] and Kellehar and Fair [10],
Condenser
Gauged cylinder
Hollow fiber contactor
Heater XD
Liquid Vapor Reboiler
XW
Fig. 1. Schematic process for hollow fiber structured packing distillation. The hollow fiber module replaces the conventional packed tower used for differential distillation of alcohol–water systems. The structure of new module tower in which liquid flows inside fibers and vapor rises counter currently outside fibers leads to very different operational characteristics.
which means the columns of hollow fiber structured packing will perform better, with the height of transfer unit (HTU) as small as possible. In order to do this, we will take three alcohol systems including methanol, ethanol and isopropanol–water solutions for separation experiments. We want to see whether the performance of the modules worked formerly can keep well or not and find how different alcohols affect the separation.
G. Zhang et al. / Desalination 223 (2008) 417–426
419
2. Theoretical considerations The theoretical analysis is based on the mass balances and transfer unit theory basic to the operation of a conventional differential distillation column. We begin our work with the same data on estimates of vapor-liquid equilibrium as differential and staged distillation [3,11]. For simplicity, the column of hollow fibers is operated at total reflux, which means the total vapor and liquid fluxes are equal in the column. The relative balances on the more volatile species in both vapor and liquid give the operating line y=x (1)
theory has been used for many years to design absorption columns in industry, the challenge has always been to design an appropriate height for the packed column. The height of an absorption column l is given by l = NTU × HTU (4)
where HTU is the height of a transfer unit. In our experiment, since the length of modules is set, the true value of HTU can be easily estimated by HTU = l/NTU (5)
where y and x are the mole fraction of the more volatile species in the vapor and the liquid, respectively. Then we can get a balance on the more volatile species in the gas phase.
0 = − vG
dy − Ka ( y − y*) dz
(2)
where vG is the gas velocity, y and y* are the gas phase mole fractions in the actual vapor and in the vapor which is in equilibrium with the liquid, z is the distance measured from the bottom of the column, K is the overall mass transfer coefficient based on the vapor side, and a is the interfacial area per volume between vapor and liquid. Integrate this equation, we can easily find the number of transfer units (NTU) for our column
NTU =
∫
yl y0
dy Kal = y − y* vG
(3)
where y0 and yl are the specified vapor compositions at the bottom (z = 0) and at the top (z = l) of the column, l is the length of the hollow fiber module. As usual, the NTU calculated by integration of the experimental data is a good measure of the difficulty of the distillation and shows how high the separation efficiency is. A larger value means an easier separation, and a smaller one signals a more difficult separation. As transfer unit
Since the height of a transfer unit HTU can be defined as the height of a packed section required to accomplish a change in concentration equal to the average driving force in that section, the smaller the HTU is, the better design should be. HTU normally is a measure of industrial tower or column’s separation efficiency. A still smaller HTU means more efficient and more rapid separation process, and that’s what we dream about. The technology route of a differential distillation tower structurally packed by hollow fibers is similar to that of conventional one, as Kister [3] described, but the process is very different. As noted above, we start our experiment with a very low feed concentration, and the desired concentrations of distillate and bottom. The liquid and vapor flows are very independent because of the non-porous hollow fiber walls. The twophase flow and maldistribution in conventional process can then be effectively prevented. The liquid and vapor can be pumped very freely with almost no pressure drop coming from fluid friction, and still not be bothered by loading and flooding. Although we have little experience to guide our choice on column designing and material choosing, we think it important to choose the structure and dimension of the column’s internals, and we will specify the relative difficulty of the separation of different alcohol systems as NTU. We can analyze our experiment using the same
420
G. Zhang et al. / Desalination 223 (2008) 417–426
equations as given above. After measuring the concentrations y0 and yl and calculating the NTU value, we can estimate mass transfer coefficients from the NTU and hollow fiber geometry and find if they are different in different separation systems or with empirical values of membrane contactors elsewhere. 3. Experimental The microporous polyether sulfone hollow fibers (Porous Media, MN, USA and RDCWTT, Hangzhou, China) used in this work have an inside diameter of 0.07 cm and an outside diameter of 0.11 cm. A thin layer of 5 µm polydimethylsiloxane was coated on the membrane for stopping convection. Thus liquid flow could not pass through the wall to the vapor side while vapor could flee out freely. From the earlier results of Zhang and Cussler [4,5], we chose 25 fibers in a glass shell with inside diameter of 1.40 cm as module design basis. Each module with two nozzles had an effective mass transfer length of 25 cm. The module properties are compared in Table 1 with commercial structured packings – Sulzer Mellapak [12]. The void fractions of the hollow fiber modules are lower than those of the conventional structured packing, while the areas per volume are larger, according to Yang and Cussler [13] and Kistler and Cussler [14]. The hollow fibers were potted with SG-856 epoxy
(Jinpeng Chem., Zhejiang, China) and HC 703 silicone (Xiguang, Wuxi, China). All distillation experiments were run at total reflux. In a typical experiment, all solvents including methanol, ethanol and isopropanol (Sinopharm Chemical Reagent Co., Ltd, Shanghai, China) were at least HPLC-UV grade and used as received, and an alcohol solution with a molar fraction of 0.04 was prepared as feed with doubly distilled water. About 800 cm3 of this solution was added to the reboiler and heated in a 350 w 98-1-C digital auto-control heating mantle (Taisite Co., Tianjin, China). The heating rate was adjusted by a FATO SVC-500 VA automatic voltage regulator and measured by a JL 4006B electric multimeter (Junling Co., Hangzhou, China). Then the vapor generated flew up the shell side with liquid reflux from condenser dropped down inside fibers concurrently (Fig. 1). All flows were measured volumetrically. Solution concentrations were measured as 1 cm3 samples taken from the reboiler and from the distillate, respectively. The concentrations were determined by injecting 2 μL samples into a SF GC-1102 gas chromatograph equipped with a temperature conduct detector (TCD) and a Super Porapak Q steel column. The carrier gas was hydrogen of purity 99.99%. Each concentration was measured at least in duplicate. To begin the experiment, we usually ran the column at a heating rate for about 1–2 h to approach steady state, when the first sample was taken. Subsequent samples
Table 1 Properties of hollow fiber modules and structured packing Module packing Module 1 Module 2 Sulzer Mellapak Sulzer BX Raschig Rings Raschig Rings Size or type 25 (number of fibers) 50 (number of fibers) 250Y BX 500 1 inch 1/2 inch Void fraction, e 0.85 0.71 0.95 0.90 0.78 0.70 Area per volume, a (ft2/ft3) 201 484 76 150 58 101 Packing factor, FP (a/e3) 327 1352 89 206 179 580
G. Zhang et al. / Desalination 223 (2008) 417–426
421
were then taken at 15–30 min intervals to ensure steady state. The calculation of the NTU from these data is complicated because the slope of each equilibrium line for different alcohol systems is not constant, so that m is not constant. Over the concentration range of 0.1 < x < 0.9, the vapor–liquid equilibrium fit the polynomials and a data relationship can be expressed mathematically as below [15]. Equation for methanol (no azeotropic point), 0.1 < x ≤ 0.3 y* = 0.1754 + 2.8545x − 4.0810x2 0.3 < x < 0.9 y* = 0.4823 + 0.6679x − 0.1571x2
(6)
(7)
Equation for ethanol (azeotropic concentration x = 0.881) 0.17 < x ≤ 0.33 y* = 0.348 + 1.1821x − 1.4292x2 0.33 < x < 0.68 y* = 0.4977 + 0.1791x − 0.2599x2 (8)
in different alcohol systems, finding some regularity and achieving the lowest value of the height of a transfer unit as possible as we can. As before, we can easily operate hollow fiber at flows greater than that causing flooding, as shown in Fig. 2. The relationship between the flooding gas velocity and other physical properties of the system is presented in the form of an empirical, experimental correlation for pressure drop in packed columns. This method was proposed by Sherwood in 1938 and improved by Eckert in 1970 to determine the flood point, which basically remains the standard procedure today [1,3,11]. We plot the normal flow parameter on the abscissa, and the common capacity factor on the ordinate. The solid curve is the empirically determined limit of flooding. For conventional packing, the gas velocity in an operating column must obviously be lower than the flooding velocity and is often chosen as one-half the predicted
1
(9)
Y = (uo2FPρ Gf { ρ L} { μ L})/(gρ H O(L))
2
Equation for isopropanol (azeotropic concentration x = 0.675) 0.13 < x ≤ 0.33 y* = 0.4438 + 0.5449x − 0.6984x2 0.33 < x < 0.79 y* = 0.5678 − 0.2586x + 0.6278x2 (10)
0.1
0.01 Methanol Ethanol Isopropanol
(11)
Using these relations and Eq. (3), the NTUs and HTUs can be easily found for comparison with different separation systems operated in the same module. The results are given below. 4. Results and discussion This section reports distillation separations using a structure packing of hollow fibers. We are especially interested in confirming its behavior
0.001 0.01
0.1 X = (LML/GMG) (ρG/ρL)
1
0.5
10
Fig. 2. Hollow fiber distillation operate above flooding. Like conventional distillation process, the flow parameter on the abscissa is plotted vs. the capacity factor on the ordinate. All the data of methanol, ethanol and isopropanol–water systems are at fluxes high above the normal flooding limit, shown as the solid line.
422
G. Zhang et al. / Desalination 223 (2008) 417–426
flooding velocity obtained from a generalized correlation. So points above this solid curve cannot be reached to ensure safe operation. In Fig. 2, we also show the data points of our experiments with hollow fibers in different alcohol solutions. Without exception again, all the data points of methanol, ethanol and isopropanol–water systems are high above the limit solid curve where flooding normally occurs. The highest data is 3–5 times above the solid flooding line. Since we always ran at total reflux, these data only cover a narrow range of abscissa in which they proceed in the order isopropanol > ethanol > methanol. But the results clearly show how easily hollow fiber structured packing can operate above the normal flooding point. In conventional cases, people must choose a velocity far from the flooding velocity but not so low as to require a much larger column which will consume more cost of power, and lower gas–liquid velocities often lead to a nearly proportional reduction in mass transfer rate. In our experiments, we could potentially operate at still higher flow rates since vapor and liquid are in contact only across the hole of membrane walls. There is no resistance for a bubble rising through liquid, and the falling liquid is not slowed by the rising vapor, just as other membrane contactors do. That is where the hollow fiber columns always have an advantage on. In order to compare the operational performance of hollow fiber with normal structural packing in distillation process, the F-factor is plotted vs. the NTU per meter (Fig. 3). The data points here for same hollow fiber module in different alcohol–water systems are compared with a type of Sulzer structured packing (Sulzer Chemtech), shown as the solid lines. The abscissa shows a flow parameter F, which tends to cover about the same range for different total pressures in conventional distillation, since the flooding velocity varies with rG1/2. Parameter F, generally used to characterize performance of distillation packing and known as the F-factor, see Refs. [3,12], is defined as
F = uG r G1/2
(12)
where uG is the vapor velocity, rG is the vapor density. The data in Fig. 3 show how the NTU increase as the F-factor gets smaller. This figure also shows some of the results under conditions where the normal packing will flood, and the separation efficiency estimated from NTU can strongly compete with some excellent industrial structured packing such as Sulzer Mellapak. We next turn to find how the hollow fiber module works as flows changes from high to low. From Eq. (3), the reciprocal of the NTU (1/NTU) should be proportional to the vapor velocity vG. This is partly supported by the data in Fig. 4 when vapor velocity becomes high for all the three systems. At low vapor velocity, the NTU keeps a relative high value for some period. We are not sure whether there are some limits with the geometry of our column design or it is just
4.00
3.00
NTU (1/m)
2.00
Methanol Ethanol 1.00 Isopropanol Mellapak
0.00 0.1
1
10
Ff at or = uGρ G1/2 (m/s(kg/m3)1/2)
Fig. 3. Operational performance of hollow fiber and normal structural packing in distillation process. The F-factor is plotted vs. the number of transfer units. The data for same hollow fiber module in different alcohol– water systems are compared with a Sulzer structured packing, shown as the solid line.
G. Zhang et al. / Desalination 223 (2008) 417–426
2.0 Methanol Ethanol Isopropanol 1.6 2.0 Methanol Ethanol Isopropanol y = 309.97x + 0.0688 y = 132.49x + 0.1304 y = 127.58x + 0.2367
423
1/NTU
1.2
1/NTU
0.8 0.4 80 120 160 200 240 0 0 0.001 0.002 0.003 0.004 0.005 0.006
1.0
0.0
Vapor Velocity (cm/s)
Fig. 4. The number of transfer units (NTU) vs. vapor velocity. The slower the vapor flows, the higher NTU value can be got, which means more efficient separation. Further, the data curve for lower molecular weight alcohol such as methanol is steeper than ethanol and isopropanol, and very close to linear when vapor velocity surpass 140 cm/s.
Liquid Flow (mol/s)
Fig. 5. The number of transfer units (NTU) vs. liquid flow. The reciprocal of the NTU is proportional to the molar liquid flow. Higher NTU value can be got when the liquid flow becomes slower. The slope of the linear relationship is in the order of methanol > ethanol > isopropanol.
caused by heat losses due to the simple jacket insulation. But the trend does mean that we can obtain faster separation and higher purity of alcohols in our hollow fiber column simply by reducing the flows. We can use slow flows without reducing the interfacial area just by making sure the liquid inside fibers are full and can run continuously. Further, the NTU for lower molecular weight alcohol such as methanol system separation increases faster than ethanol and isopropanol, but at same vapor velocity, the NTU for higher molecular weight alcohol like ethanol and isopropanol looks larger, which means more efficient separation. Since the vapor velocity in our experiments is calculated directly from heat balance, as described by Poling et al. [16], and there is some insulation problem which may cause heat loss, especially when the counter current flow is slow, we use data read from gauged cylinder and plot Fig. 5 to show change of the NTU vs. liquid flow more precisely. From Eq. (4), we have a similar form of the NTU rewritten as
NTU =
∫
x1 x0
dx ⎡K a⎤ = ⎢ x ⎥l * x − x ⎣ L ⎦
(13)
where Kx is an overall mass transfer coefficient based on a liquid side. Now the results in Fig. 5 meet our expectation from Eq. (13) very well, the reciprocal of the NTU is proportional to the molar liquid flow. As above, higher NTU value can be got when the liquid flow becomes slower, and the slope of the linear relationship is in the order of methanol > ethanol > isopropanol. This gives the overall mass transfer coefficient Kx for methanol, ethanol and isopropanol of 2.45 × 10−5, 5.73 × 10−5 and 5.95 × 10−5, respectively, in accordance with the same order as former results, according to Zhang and Cussler [4,5] and Chung et al. [17]. The intercept on this plot may be dependent on the properties of flow, module configuration and the heat loss of process. To reassure our results, we show the variation of distillate concentration with heating rate in Fig. 6. The higher distillate concentrations
424
1.0
G. Zhang et al. / Desalination 223 (2008) 417–426
0.40
0.8
Methanol
0.30
Distillate Concentration (mol)
Ethanol Isopropanol
HTU (m) Methanol Ethanol Isopropanol
0.6
0.20
0.4
0.10
0.2
0.00
0 80 100 120 140 Heat Rate (W) 160 180 200
80
100
120
140 Heat Rate (W)
160
180
200
Fig. 6. Distillate concentration vs. heating rate. As the heating rate increases, the distillate concentration at the top of module decreases.
Fig. 7. The height of a transfer unit (HTU) vs. heating rate. Because the heating rate is proportional to residence time in the column, the height of a transfer unit drops as the heating rate rises.
observed at the lower heating rate are consequence of better mass transfer in vapor/liquid interface, and the relative sensitivity of distillate concentration change with heating rate is methanol > ethanol > isopropanol. As we know, there is an azeotropic point for both ethanol and isopropanol solution, such as isopropanol–water azeotrope forming at an isopropanol mole fraction of 0.675, and at present we cannot break through it with our owned modules in the experiments, that means separation of methanol solution without azeotrope formation will be simpler and easier. Finally, the height of a transfer unit is plotted vs. heating rate in Fig. 7. Because the heating rate is proportional to residence time in the column, the height of a transfer unit drops obviously as the heating rate rises while the contacting time becomes shorter. In our experiments, a low-mole fraction of alcohol as low as 0.04 was prepared as feed in all of the three solutions, and an attractive low value of the HTU less than 0.1 m could be easily obtained at total reflux. The data in Fig. 6 also show the same behavior as the above, the
decreasing rate of the height of mass transfer unit in the same hollow fiber module seems to change inversely with the molecular weight of alcohols. Although the above results have already shown a promising future for fast and efficient separation in hollow fibers, we hope to do better. In conventional distillation, packing always wins by high surface area per unit volume and high void fraction, so the hollow fiber can go farther with better characteristics on it. As before, the physical properties of the system, the liquid/gas flow and the packing controls the performance of conventional packed column. That means the geometry of packing, usually characterized as ‘Packing factor’, is critical for process designing, see refs. [3,9,11]. Packing factor FP is defined as
FP =
a e3
(14)
where e is the void fraction of the packing. To get a bigger value of this packing factor, the packing in a set column should be chosen smaller because FP is roughly proportional to the packing
G. Zhang et al. / Desalination 223 (2008) 417–426
425
size (see Table 1). But even the smallest conventional packing cannot get a higher packing factor than hollow fibers whose interfacial area per volume is large. In our experiments, a same size module gives a packing factor of 327 ft−1 with 25 fibers and 1352 ft−1 with 50 fibers, the amplifier factor is very big. That means we still have huge space to optimize our separation in hollow fibers. We look forward to more productive distillation over a wider operational range and provide more significant energy savings in the near future. 5. Conclusions New results have been introduced which show the separation of methanol, ethanol and isopropanol–water solutions with distillation using hollow fiber structured packing. The membranes used here for are non-porous hollow fibers, unlike those of membrane distillation. The fibers are non-selective with coating and have little resistance to mass transfer. Because liquid flows inside the lumens of the fibers and vapors flows counter currently outside fibers, the distillation process can avoid flooding effectively even at 3–10 times higher flows than conventional cases. With a mole fraction of alcohol as low as 0.04 prepared as feed in all of the three solutions, an attractive low value of the HTU less than 0.1 m can be easily obtained at total reflux. Moreover, while the distillate concentration drops as the heating rate rises, just as noted before, the decreasing rate of distillate concentration and HTU in the same module seems to change inversely with the molecular weight of alcohols. Compared with conventional structured packing, the hollow fibers have a better geometry of packing and can realize a more productive distillation and will save more energy in separation. Further experiments will focus on the optimization of module structure and process design and using more systems to validate our investigation.
Acknowledgments The authors benefited from conversation with Dr. E.L. Cussler during visiting CEMS, UMN. This work was primarily funded by the Research Fund of the National Natural Science Foundation of China (grant 20476096). Other financial support came from the Zhejiang Provincial Bureau of Science and Technology, China (grants 2005C33040 and 2006C23067).
References
[1] [2] J.-D. Seader and E.J. Henley, Separation Process Principles, Wiley, New York, 1998. J.-L. Humphrey and G.-E. Keller, Separation Process Technology, McGraw-Hill, New York, 1997. H.-Z. Kister, Distillation Design, McGraw-Hill, New York, 1992. G. Zhang and E.-L. Cussler, Distillation in hollow fibers, AIChE J., 49 (2003) 2344–2351. G. Zhang and E.-L. Cussler, Hollow fibers as structured distillation packing, J. Membr. Sci., 215 (2003) 185–193. K.-W. Lawson and D.-R. Lloyd, Membrane distillation, J. Membr. Sci., 124 (2000) 1–25. G. Alan and S.-T. Hwang, Hollow fiber membrane contactors, J. Membr. Sci., 159 (1999) 61–106. M.-A. Izquierdo-Gil and G. Jonsson, Factors affecting flux and ethanol separation performance in vacuum membrane distillation, J. Membr. Sci., 214 (2003) 113–130. D.-L. Bennet and K.-W. Kovak, Optimize distillation columns, Chem. Eng. Prog., 19 (2000) 84–96. T. Kellehar and J.-R. Fair, Distillation studies in a high gravity contactor, Ind. Eng. Chem. Res., 35 (1996) 4646–4655. E.-L. Cussler, Diffusion, Cambridge University Press, Cambridge, 1997. Sulzer Chemtech, Structural packings for distillation and absorption, Product Bulletin 22,13,06 40–111, (2001) 00–70. M.-C. Yang and E.-L. Cussler, Designing hollow fiber contactors, AIChE J., 32 (1986) 1910–1920.
[3] [4] [5]
[6] [7] [8]
[9]
[10]
[11] [12]
[13]
426
G. Zhang et al. / Desalination 223 (2008) 417–426
[16] B.-E. Poling, J.-M. Prausnitz and J.-P. O’Connell, Gases and Liquids, 5th edn., McGraw-Hill, New York, 2000. [17] J.-B. Chung, J.-P. Debrocher and E.-L. Cussler, Distillation with nanoporous or coated hollow fibers, J. Membr. Sci., 257 (2005) 3–10.
[14] K.-A. Kistler and E.-L. Cussler, Membrane modules for building ventilation, Trans. Inst. Chem. Eng., 80 (2002) 53–64. [15] J. Gmehling and U. Onken, Vapor-Liquid Equilibrium Data Collection, Vol. 1, Part 1, DECHEMA, Frankfurt, 1977.
