Pergamon P l h S0043-1354(97)00342-5
Wat. Res, Vol. 31, No. II, pp. 2878-2884, 1997 ~(') 1997 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0043-1354/97 $17.00 + 0.00
FT30 MEMBRANES OF CHARACTERIZED POROSITIES IN THE REVERSE OSMOSIS ORGANICS REMOVAL FROM AQUEOUS SOLUTIONS L. KASTELAN-KUNST, K. KO~UTIC, V. DANANI(~ and B. KUNST* Faculty of Chemical Engineering and Technology, University of Zagreb, 10000 Zagreb, P.O. Box 177, Croatia (Received Januao, 1995; accepted in revised [orm Februao: 1997)
Abstract--Removal of some organic pollutants from aqueous solutions by reverse osmosis using commercial FT30 membranes of characterized porosities was investigated. The pore size distribution analysis and calculation of the effective number of pores have shown the presence of numerous narrow unimodally distributed pores in the upper selective membrane layer. The organic pollutant removal by reverse osmosis is governed mostly by mutual interactions between the membrane material, solute and water molecules. © 1997 Elsevier Science Ltd Key words--reverse osmosis, membrane porosity, organic pollutants
NOMENCLATURE f= N= (PRo~p)~7= (PR0xp)3, = Y= Ea* = c=
solute separation coefficient number of pores in a membrane membrane productivity at 17 bar (M/T) membrane productivity at 34 bar (M/T) pore size distribution function (L ~) Taft number solution concentration, ppm (M/M -3) INTRODUCTION
The applicability of membrane separations for water and wastewater treatment is well known. Membrane separation technology has been widely used for desalination and inorganic salts withdrawal from salty and wastewaters (Belfort, 1984; Sourirajan and Matsuura, 1985a). The efficient removal of organic substances from aqueous solutions and wastewaters by reverse osmosis has also received considerable attention (Matsuura and Sourirajan, 1972; Dickson et al., 1975; Bhattarcharyya et al., 1987; Williams et al., 1990) due to environmental reasons. Although there are examples of successful and economically viable use of the reverse osmosis process for separation of organic substances (Argo 1984), a lot of work still has to be done to understand the interrelations between the membrane characteristics, such as its porosity and performance, and to develop reliable and competitive membrane separation systems for organic pollutant removal. This paper deals specifically with reverse osmosis separation of some organic chemicals of industrial and environmental significance appearing in low *Author to whom all correspondence should be addressed [Fax: 385 01 45 97 250].
concentrations (less than 100 ppm) in wastewaters from chemical and petrochemical plants. The reverse osmosis performance parameters, separation coefficient and permeate flux, of such a separation are known to be influenced by operating variables such as solute concentration and operating pressure, by the physicochemical properties of the system, such as the polar and steric characteristics of the solute and membrane, and by membrane porosity parameters. In order to determine and control the latter group of parameters, i.e. the membrane's pore size, pore size distribution (PSD) and effective number of pores, we have chosen in this work samples of the often used commercial reverse osmosis membrane type, Dow FilmTec's FT30. By determining the porosity parameters of the membrane samples its role in the organic pollutant removal process from aqueous solution is better understood. Recently Kastelan-Kunst et al. (1996) proposed a method that enabled the determination of pore size and pore size distribution in the membrane's upper layer (skin). The concept and the procedure are based on the surface force-pore flow model of material transport through a membrane (Matsuura et al., 1981) and experimentally obtained solute separations for a number of markers. In the course of the calculation, pertinent quantities such as permeation volume, permeation rate, solute concentration at the pore outlet and solute separation for a single pore were defined. By adding up these quantities for a collection of pores the permeation rate and the solute separation values of various markers were computed as functions of the pore size distribution through the collection of active pores of the membrane's upper
Reverse osmosis membranes in organics removal Table h Some properties of FT30 membrane Property Sodium chloride separation, f Permeate flow rate Maximum operating pressure Maximum operating temperature pH range Free chlorine tolerance
Value > 0.991 ~ 6301 m-: h -~ 69 bar 45-'C 2-11 < 0.1 ppm
layer. The results were finally c o m p a r e d with experimental data to give the best fit pore size distribution in the m e m b r a n e ' s skin. In a d d i t i o n to pore dimensions a n d pore size distributions, the effective n u m b e r o f pores in the m e m b r a n e ' s skin could have been calculated using this method. The m e t h o d has already been successfully examined with l a b o r a t o r y m a n u f a c t u r e d cellulose triacetate asymmetric m e m b r a n e ' s , a n d this work offers a n o p p o r t u n i t y to apply the m e t h o d to commercial membranes. EXPERIMENTAL
Commercial FT30 membranes manufactured by the FilmTec Corporation of Dow Chemical Company were used throughout the work. The FT30 membranes are of the thin-film composite type and their performance and
properties have been listed in the Dow-FilmTec technical bulletin (Dow, 1994). In Table 1 some of the membrane characteristics relevant to this work are given. The membranes were tested in a simple laboratory reverse osmosis apparatus illustrated schematically in Fig. l(a). Five reverse osmosis cells like the one shown in Fig. l(b) were connected in series. The surface area of the membranes investigated was 13.2 cm-'. The reverse osmosis experiments with sodium chloride, the organic reference solutes (markers) and chosen organic pollutant solutions were of the short-run type, each lasting for about 3 h; they were carried out at the laboratory temperature and at pressures of 17.0 and 34.0 bar using a feed flow rate of 350 ml min-L The pH values of the solutions were kept at 6.8-7.0. The concentration of the feed in the case of sodium chloride solution was 3500 ppm, and in the cases of the markers and the organic contaminant solutions, 100 ppm. The product rates (PR) referring to the membrane permeated solution corrected to 25%C, as well as the solute separation coefficient, .f, defined as f=
cr~d --Cp,~,, Creed
c in ppm,
were determined in each experiment. Five organic reference solutes (markers) were used for pore size distribution analysis. All the markers are listed in Table 2 together with the organic pollutants used in the reverse osmosis separation examinations. The sodium chloride concentrations were determined by conductivity measurements, and those for the markers and
I ~ ~ ~ ~ ~ ~ ~
I~7~/.~/~j/~/////~~-'.//~/'/.//'/~./~Membrane Permeate Fig. 1.(a) Reverse osmosis set-up: H, hold-up tank; P, high pressure pump; M, manometer; RO, reverse osmosis cells; R, pressure regulator. (b) Cross-section of reverse osmosis cell.
L. Kagtelan-Kunst et al. Table 2. Organic reference solutes (markers) and organic pollutants used in experiments Name Formula Molecular weight CAS number Markers Ethanol C2H~OH 46.07 64-17-5 1,3-Dioxolane C~H602 74.08 646-06-0 1,4-Dioxane C4H802 88.11 123-91-1 12-Crown-4 C8HI604 176.21 294-93-9 (1,4,7,10-tetraoxacyclododecane) 18-Crown-6 C,2H2406 264.32 17455-13-9 ( 1,4,7,I0,13,16-hexaoxacyclooctadecane) Organic pollutants Formaldehyde CH20 30.03 50-00-0 CH3CO2C:H5 88.11 141-78-6 Ethyl acetate CH3COC2H5 72. I 1 78-93-3 2-Butanone 2-Ethoxy ethanol HOCH:CH2OC2H5 90.12 110-80-5 1,2-Ethanediol HOCH2CH:OH 62.07 107-21- 1 (Ethylene glycol) Tetrahydrofuran C4HsO 72.11 109-99-9 Neopentylglycol HOCH2C(CH3)2CH2OH 104.15 126-30-7 (2,2-dimethyl-1,3-propanediol)
the selected organic pollutants using a carbon analyzer (lonics., Inc., model 1555). The method of PSD determination and the calculation of the effective number of membrane pores (Kastelan-Kunst et al., 1996) in this work were slightly modified by measuring the separation of markers at two different pressures. In this way the number of experiments was increased and the reliability of the calculation procedure improved. The experimental data on the separation of the markers used throughout the PSD calculation are given in Tables A I and A2 in the Appendix. RESULTS AND DISCUSSION Table 3 gives the data obtained using the usual membrane characterization method with sodium chloride solution at two pressures. All five membrane samples were shown to be highly selective to sodium chloride as was expected from the manufacturer's data. The producitivities of the five membrane samples were rather different, M5 showing the highest and M2 the lowest value and those of the other membranes having intermediate productivity values. The results of the pore size and the pore size distribution determinations for five samples of the composite FT-30 membranes (Fig. 2) show that pores in the skin were almost the same size, and a similar narrow unimodal pore size distribution. Maxima of the PSD of the FT30 membranes were between 5.5 and 6.5 A, designating the samples as typical reverse osmosis membranes. Laboratory prepared phase inversion membranes, on the other hand, usually show a wider and most often bimodal or even Table 3. Results of membrane characterization using sodium chloride solution M1 M2 M3 M4 M5 P = 17 bar PR* (g h- ') 11.2 10.2 12.4 12.6 16.3 0.995 0.991 0.982 0.995 0.986 f P = 34 bar 22.1 20.0 24.0 24.2 31.2 PR* (gh -~) 0.995 0.994 0.987 0.995 0.994 f *Membrane surface area = 13.2cm2.
multimodal PSD. For comparison this is illustrated in Fig. 3 for a cellulose triacetate membrane. The unimodality of the pore size distribution is a feature which good commercial membranes are supposed to possess, and the results obtained confirming this seem reasonable from the standpoint of the thin-film composite membrane fabrication procedures. Its last step, the formation of a thin aromatic polyamide layer by interfacial polycondensation (Petersen and Cadotte, 1990) is supposed to equalize properties such as porosity of the upper membrane layer. The effective number of pores in the upper layer of the investigated membranes, calculated in the course of the pore size and PSD determination, to some extent follows the productivity values of the membrane samples (Table 4). The largest number of
l MI ......
.......... M 5
Fig. 2. Pore size distribution of the FT30 membrane samples.
Reverse osmosis membranes in organics removal 0.6
...... . . . . . . . . . .
--Iil~ ~ ///~,~
17 b a r
...... 3 4
0.1 0.0 ~
0.7 "5t~ 0.6
Pore size, [A]
Fig. 3. Comparison of pore size distributions of FT30 and laboratory manufactured cellulose triacetate membranes.
Fig. 4. Correlation between the average separation data of the examined organic pollutants and their Taft numbers.
Table 4. Membrane productivities obtained using sodium chloride solution and effectivenumber of pores in upper membrane layer (PR~xphr* (PR,~p)34* Membrane (gh -t) (gh ~) Nx l0 -~ I I 1.2 22. l 1.299 2 10.2 20.0 1.384 3 12.4 24.0 1.766 4 12.6 24.2 2.576 5 16.3 31.2 2.878 *Membrane surface area = 13.2cm2.
5.7 ,~, the smallest average pore dimensions of all the membranes, and for the same productivity more such small pores must exist in the skin of this membrane. Additional parameters such as membrane thickness, tortuosity factor, etc. may also significantly affect the number of pores in the membrane's skin. The similar pore dimensions and PSD found during the porosity characterization of several FT-30 membrane samples make these membranes suitable for studies of other factors affecting separation performance. In further work such membranes with characterized porosities were used for reverse osmosis separations of selected chemically different solutes appearing in wastewaters of petrochemical and organic chemical industries. The membrane separation coefficients, f , obtained in the experiments at the lower operating pressure (17 bar) are tabulated in Table 5, together with the corresponding molecular weight o f the solutes and their Taft parameters E~*, a measure of the polar character of solute molecules. It is immediately evident that a correlation does not exist between the pollutant separation coefficients, f , and the molecular weight data. This
effective pores in membrane M5 explains its highest productivity value. Such reasoning would lead one to expect sample M2 to have the least number of pores. This is not the case, however; the number of pores in sample M2 is small, but not the smallest. There are evidently additional factors affecting the membrane productivity. One o f them is the pore size distribution itself. Its influence is well illustrated by membranes M3 and M4 which exhibit equal productivity and a markedly different number o f pores. This result can be explained by the fact that two samples do not have quite the same pore size distribution curves. The pore size distribution of membrane M4 is located around
Table 5. Reverse osmosis separation coefficientsand Taft parameters for different organic pollutants at 17 bar Substance Molecular weight f~,= f~,-so 5"a* Formaldehyde 30 0.440-0.467 0.451 + 0.490 Ethyl acetate 88.1 0.765-0.853 0.812 0 2-Butanone 72. I 0.828-0.877 0.852 - 0. I0 2-Ethoxy ethanol 90 0.853-0.901 0.875 - 0.20 Ethylene glycol 62.1 0.881-0.944 0.921 - 0.20 Tetrahydrofuran 72. I 0.916-0.944 0.926 - 0.25 Neopentyl glycol 104 0.971-0.988 0.980 - 0.33
means that the membrane separation cannot be thought of as simple filtration, i.e. a "sieving" process. Such a conclusion makes the frequently used concept of a "molecular weight cut-off' of a membrane inappropriate and somewhat misleading. It seems worthwhile to emphasize that the idea of membrane separation as a sieving process is applicable for microfiltration membranes with large pores; it is less applicable for ultrafiltration membranes with smaller pores, and cannot be justified at all in the case of nanofiltration and reverse osmosis membranes with pore dimensions of the order of magnitude of between 5 and 30A (0.5-3 nm). The pore size and PSD in the membrane's skin naturally influence the separation performance of reverse osmosis and nanofiltration membranes, but they are not the only decisive factors. Interactions between solute molecules, the solvent and membrane material should also be taken into account as important parameters affecting the separation characteristics of a membrane. The results presented in Table 5 and by the full line in Fig. 4 show a smooth relationship between the average (for five membranes) separation data of the organic pollutants and their Taft numbers, Eo*, taken from the literature or calculated (Taft, 1956; Matsuura and Sourirajan, 1973). The plotting of the average separation coefficient values for all five membranes in this figure can be justified by the virtually identical correlations obtained for each investigated membrane and for the average separation values (columns 2-5 in Table 6). A decrease of the Ea* values is followed by a regular increase of the solute separation coefficient, f. Such a result could be explained by the notion that the organic solute separation value,f, should be connected primarily to nonionic interactions between the membrane material, organic solute and solvent molecules. This notion is in agreement with the results of the fundamental work of Sourirajan and Matsuura (1985b) giving a physicochemical basis for reverse osmosis separations of nonionized organic solutes. According to their findings the membrane separation effect manifested by a decreased solute concentration in the permeate is the consequence of a net preferential sorption of water at the membrane-solution interface. The preferential sorption of water is the overall result of mutual interactions between the membrane material, organic T a b l e 6. C o n s t a n t s A a n d B a n d c o r r e l a t i o n d a t a o f l i n e a r r e l a t i o n f vs. ~ztr* f o r e a c h m e m b r a n e a n d f o r t h e a v e r a g e v a l u e s
MI M2 M3 M4 M5
0.7906 0.7867 0.7806 0.7673 0.7617
0.6153 0.6548 0.6304 0.6493 0.6358
Correlation coefficient, r -
0.9877 0.9774 0.9839 0.9953 0.9992
0.0293 0.0425 0.0344 0.0189 0.0075
nonionic solute and water molecules. Nonionic mutual interactions arise from short-range forces depending on the individual contributions of the polar, steric, and nonpolar character of the solute, solvent and membrane material involved. Water is a polar solvent; a membrane surface made from aromatic polyamide chains has both polar groups and a less polar backbone; and the nonionized organic molecule may be polar or nonpolar, or both. In a system like ours consisting of the same solvent, the same membrane material and different pollutant molecules, the polarities of the last entity will indicate the relative numerical values of mutual interactions. There are several ways to quantitatively express solute molecule polarity (Sourirajan and Matsuura, 1985b). These are: l, the solute molecule's hydrogen bonding ability as represented by its acidity or basicity; 2, the solute's dissociation constant represented by its p/~ value; and 3, the Taft or Hammett numbers for the solute molecule. The Taft number concept has a firm thermodynamic base and it is generally applicable. The Taft number values of the substituent groups in a molecule are additive, and extensive numerical data on Y-a*, measured and/or calculated, are available in the literature (Taft, 1956). A lower Taft number value for the organic molecule expresses its lower proton-donating power. In the presence of such a molecule the preferential sorption of water at the membrane-solution interface will be stronger and the solute separation coefficient, f, will be higher. Our results show that such reasoning is also valid in the case where the solutes used belong to various classes of organic substances. The limitations are that the solute concentration in the feed solution should be low ( < 200 ppm), and the operating pressure high enough ( > 15 bar) so that osmotic pressure effects on the membrane performance can be neglected. The results obtained at a higher operating pressure (34 bar) are given in Table 7. The tabulated data show that at higher pressure a similar correlation to that at 17 bar between the solute separation, f, and E6* values exists, although the data are slightly more scattered than those obtained at lower pressure. The relationship is again linear (dotted line in Fig. 4), showing that the two-fold increase of the operating pressure did not change the mechanism of the solute separation. The slope is less steep than that for 17 bar, showing the reduced effect of the Taft parameter on the solute separation; a result that could have been predicted. At higher pressures the mechanical force acting in the membrane pores prevails much more strongly over other, pressure-independent forces, such as steric repulsion of the nonionized solute from the pore wall and van der Waals attraction force. Consequently, at higher pressures, the influence of the short-range forces appearing in the solute-solvent-membrane system is weaker.
Reverse osmosis membranes in organics removal
Table 7. Reverse osmosis separation coelficients and Taft parameters for different organic pollutants at 34 bar Substance Molecular weight f,,,~, f~v,,~, E~* Formaldehyde 30 0.562--0.631 0.601 + 0.490 Ethyl acetate 88.1 0.8764).930 0.897 0 2-Butanone 72.1 0.833-0.902 0.882 - 0.10 2-Ethoxy ethanol 90 0.917-0.936 0.926 - 0.20 Ethylene glycol 62.1 0.870~.934 0.910 - 0.20 Tetrahydrofuran 72. I 0.897-0.935 0.924 - 0.25 Neopentyl glycol 104 0.964-0.978 0.970 - 0.33
CONCLUSIONS 1) The pores in the upper layer (skin) o f the FT-30 m e m b r a n e samples have almost the same dimensions; their pore size distribution is also very similar being n a r r o w a n d u n i m o d a l with a m a x i m u m between 5.5 a n d 6.5 A. 2) The calculated effective n u m b e r o f pores in the skin o f the tested m e m b r a n e s follows, to some extent, the m e m b r a n e productivity (permeability) values. The deviations can be attributed to slight variations in P S D a n d some o t h e r effects. 3) T h e r e is no correlation between the organic p o l l u t a n t separation coefficient a n d the molecular weight data, showing that reverse osmosis m e m b r a n e separation c a n n o t be t h o u g h t o f as a simple sieving process. Therefore the term " m o l e c u l a r weight cut-off" with respect to a reverse osmosis m e m b r a n e is i n a p p r o p r i a t e a n d should n o t be used. 4) A regular relationship between the separation d a t a o f different organic pollutants a n d their Taft n u m b e r s (as the m e a n s of expressing the solute molecule polarity) exists, showing that mutual interaction between the m e m b r a n e material, organic n o n i o n i c solute a n d water molecules are the basic p a r a m e t e r s in the reverse osmosis separation process.
Argo D. R. (1984) Use of lime clarification and reverse osmosis in water reclamation. WPCF 56(12), 1238. Belfort G. (1984) Synthetic Membrane Processes, Fundamentals and Water Applications. Academic Press Inc., Orlando, FL. Bhattacharyya D., Adams R. and Williams M. (1989) Separation of selected organics and inorganic solutes by
low pressure reverse osmosis membranes. In Biological and Synthetic Membranes (Edited by Liss A. R.). D. Butterfield, New York. Dickson J. M., Matsuura T., Blais P, and Sourirajan S. 0975) Reverse osmosis separations of some organic and inorganic solutes in aqueous solutions using aromatic polyamide membranes. J. Appl. Polym. Sci. 19, 801. Dow (1994) FilmTec Membranes, Products and Specifications, Technical Bulletin CH 172-044-G-1294R, The Dow Chemical Company. Minneapolis, MN. Kastelan-Kunst L., Dananic V., Kunst B. and Kosutic K. (1996) Preparation and porosity of cellulose triacetate reverse osmosis membranes. J. Membrane Sci. 109, 223. Matsuura T. and Sourirajan S. (1972) Physicochemical criteria for reverse osmosis separation of aldehydes, ketones, ethers, esters and amines in aqueous solutions using porous cellulose acetate membranes. J. Appl. Polym. Sci. 16, 1663. Matsuura T. and Sourirajan S. (1973) Physicochemical criteria for reverse osmosis separation of monohydric and polyhydric alcohols in aqueous solutions using porous cellulose acetate membranes. J. Appl. Polym. Sci. 17, 1043. Matsuura T. and Sourirajan S. (1981) Reverse osmosis transport through capillary pores under the influence of surface forces. Ind. Eng. Chem. Process Des. Dev. 20, 273. Petersen R. J. and Cadotte J. E. (1990) Thin film composite reverse osmosis membranes, Chap. 5. In Handbook of Industrial Membrane Technology (Edited by Porter M. C.). Noyes Publications, Park Ridge, NJ. Sourirajan S. and Matsuura T. (1985a) Chap. 8, pp. 848-884. In Reverse Osmosis/Ultrafiltration Process Principles. National Research Council Canada, Ottawa. Sourirajan S. and Matsuura T. (1985b) Chaps 1 and 2. In Reverse Osmosis[Ultrafiltration Process Principles. National Research Council Canada, Ottawa. Taft R. W. Jr (1956) pp. 556-675. In Steric Effects in Organic Chemistry (Edited by Newman M. S.). Wiley, New York. Williams M., Deshmukh R. and Bhattacharyya D. (1990) Separation of hazardous organics by reverse osmosis membranes. Environ. Prog. 9, 118.
Table A I. Reverse osmosis productivity and separation data for referenceorganic solutes (markers) at 17 bar Ethanol 1,3-Dioxolane 1,4-Dioxane 12-Crown-5 18-Crown-6 Ml PR* (gh -~) 11.7 12.5 ll.0 II.7 12.4 f 0.607 0.809 0.921 0.966 0.95 I M2 PR* (gh ~) 10.2 10.9 9.7 10.3 ll.0 f 0.621 0.786 0.912 0.967 0.951 M3 PR* (gh -~) 12.7 13.5 12.6 12.9 13.6 f 0.572 0.790 0.903 0.972 0.953 M4 PR* (gh -~) 12.8 13.9 12.1 13.1 13.5 f 0.554 0.759 0.881 0.974 0.952 M5 PR* (gh -~) 17.0 18.9 16.7 17.2 17.8 f 0.507 0.667 0.917 0.985 0.953 Membrane surface area = 13.2cm-'.
L. K a g t e l a n - K u n s t et al.
Table A2. Reverse osmosis productivity and separation data for reference organic solutes (markers) at 34 bar 1,4-Dioxane
PR* (g h -~) 0.681
PR* ( g h ~) 0.706
PR* ( g h ~) 0.688
PR* ( g h ~) 0.639
f M5 f
PR* ( g b ~) 0.661
Membrane surface area = 13.2 cm 2.
18-Crown-6 24.3 0.970 21.3 0.975 26.3 0.975 26.5 0.989 34.7 0.974