Ind. Eng. Chem. Prod. Res. Dev. 1983, 22, 77-85 Granstein, D. L.; Williams, H. J. J. Appl. folym. Scl. 1974, 78, 1. Kraus, G. J. Appl. folym. Sci. 1963, 7 , 861. Kraus, 0.I n “Reinforcement of Elastomers”; Kraus, G., Ed.; Intersclence: New York, 1965; Chapter 4, p 125. Lange, F. F. Am. Ceram. SOC.Bull. 1970, 49, 390. Leldner, J.; Woodhams, R . T. 11th Ann. Techn. Conf., Reinforced Plastics/ Composites Institute, SPI, 1976, Sec. 8F, 1. Markin, C.; Williams, H. L. J. Appl. folym. Sci. 1974, 78, 21. Migiiaresi. C.; Nlcoials, L.;Nicodeme, L.;DiBenedetto, A. T. folym. Composites 1971, 2 , 29. Mullins, L. I n “The Chemistry and Physics of Rubberlike Substances”; Bateman, L., Ed.; Maclaren: London, 1983; p 301. Nicolais, L.; Narkis, M. folym. Eng. Sci. 1971, 1 7 , 194. Nleisen, L. E. J. Appl. folym. Scl. 1966, IO, 97. Nlelsen, L. E. I n “Mechanical Properties of Polymers and Composites”; Dekker: New York, 1974; Vol. 2, p 413. Oberth, A. E. Rubber Chem. Techno/. 1967, 4 0 , 1337.
77
Payne, A. R. I n “Composite Materials”; Hollday, L., Ed., Elsevier: New York, 1966; Chapter 7, p 23. Preston, 0.; Grant, N. J. Trans. AIM€, 1961, 227, 612. Rehner, J., Jr. I n “Reinforcement of Elastomers”; Kraus, G., Ed.; Interscience: New York, 1965; Chapter 5, p 153. Runge, M. L.; Dreyfuss, P. J. folym. Scl., folym. Chem. Ed. 1979, 77, 1067. Schwarz, W. W.; Lowrey, R. D. J. Appl. folym. Sci. 1967, 7 7 , 533. Schwarzl, F. R.; Bree, H. W.; Nederveen, C. Y. I n “Proceedings, Fourth International Congress on Rheology”; Lee, E. H., Ed.; Interscience: New York, 1965; Part 3, p 241. Smith, T. L. Trans. SOC.Rheol. 1959, 3 , 113. Wong, S. G. J. Rubber Res. Inst. 1979, 2 7 , 24.
Received for reuiew May 19, 1982 Accepted September 15, 1982
Effect of Membrane Materials and Average Pore Sizes on Reverse Osmosis Separation of Dyes Llu Tlnghul, Takeshl Matsuura, and S. Sourlrajan’ Division of Chemlstty, National Research Council of Canada, Ottawa, Ontarlo K I A OR9 Canada
Reverse osmosis separations of 13 dyes were conducted by using cellulose acetate and aromatic polyamidohydrazide membranes of various average pore sizes at operating pressures in the range 689 to 1723 kPag. High performance liquid chromatography (HPLC) experiments were also performed with respect to all dyes studied using chromatography columns filled with powders of membrane polymer materials. Interfacial force constants A, B, and D, which are associated with the electrostatic force, van der Waals force, and steric hindrance, respectively, were generated by applying the transport equations to the data from reverse osmosis and HPLC experiments. These force constants enable the prediction of the dye solute separation for a given average pore radius of the membrane. This paper illustrates how the science of reverse osmosis offersa basis for the choice of membrane materiais for use in reverse osmosis applications involving separation of dyes in aqueous solutions.
Introduction The reverse osmosis separation of dye compounds from aqueous solutions is of both practical and fundamental interest. The dyeing of textile products is a process which consumes a considerable amount of water. The effluent water from the process often contains residual dyes which may be recovered (Aurich et al., 1972). Reverse osmosis offers an efficient method of separating dyes from processing water. From a fundamental point of view the reverse osmosis separations of dye molecules are particularly interesting because of their ionic, polar organic, and nonpolar organic characters. The ionic character originates from the electrolytic structure, while polar and nonpolar organic characters are caused by polar functional groups such as hydroxyl and amino groups and aromatic structures of the dye molecule. Consequently, both electrostatic and nonelectrolytic (van der Waals) interaction forces are expected to work between the dye molecule and the membrane polymer material (Vickerstaff, 1954). Therefore, the comparison of the separation data of the dye compounds with those of electrolyte solutes without organic components and of undissociated organic solutes without electrolytic components may be expected to reveal the effect of combined electrostatic and van der Waals forces on reverse osmosis separations. A total of 13 dye compounds listed in Table I is considered in this work. These dyes are generally classified as basic (dye number 1 , 2 ) , acidic (dye number 3-8), and others (dye number 9-13). While basic dyes possess cat0196-4321/83/1222-0077$01.50/0
ionic properties originating from the positively charged nitrogen or sulfur ions, all acidic, and other dyes possess anionic nature due to negatively charged sulfonate group. In this regard all dyes belonging to the last two classes have essentially the same ionic properties, and different terms apply to the classification by application rather than the chemical structure (Color Index, 1971). It may be noted that basic dyes with positive charges are so named since they have affinity to basic textile material with net negative charges, while acidic dyes with negative charges are so named since they have affinity to acidic textile materials with net positive charges. Thus, the change in the reverse osmosis performance data with the change in the ionic nature of dye molecules is itself another subject worth studying. Despite practical and fundamental interest, only few works on the dye separation by reverse osmosis and ultrafiltration have been documented in the literature. Cooper and Booth (1979) studied the separation of polymeric dyes using Romicon ultrafiltration membranes covering the range of pore size 30 to 66 X m. Though the product permeation rate was analyzed applying the gel model, the gel layer resistances calculated for membranes of different pore sizes were not constant, suggesting that a simple series resistance model is not applicable and some interaction between the gel layer and the membrane surface must occur. Kukushkina et al. (1977) investigated the separation of nine different dyes by using cellulose acetate membranes of average diameter ranging 48 to 98 X @ 1983 American Chemical Society
78
Ind. Eng. Chem. Prod. Res. Dev., Vol. 22, No. 1, 1983
Table I.
Dye Compounds Used in This Study
dye no.
name of dye
color index
mol wt
structural formula
Basic Dyes r
1
Acridine Orange
46005
2
Methylene Blue
52015
Acidic Dyes H;
3
Orange I1
15510
4
Acid Blue
15706
5
Indigo Carmine
73015
6
Amaranth
16185
7
Coomassie Blue
42655
8
Naphthol Green B
10020
830.9
a
Other Dyes
9
Alizarine Red S
58005
319.26
10
Eriochromp Black T
14645
438.38
11
Alizarol Cyanine RC
43820
467.41
12
Congo Red
22120
650.68
13
Chlorazol Black E
30235
735.75
Ind. Eng. Chem. Prod. Res. Dev., Vol. 22, No. 1, 1983 79
m and concluded that high separations of dye solutes are favored by the aggregation of dye solutes in the aqueous solution, which is facilitated by the sorption of dye molecules on cellulose acetate membrane material. Though they found that the affinity of dye solutes to the cellulose acetate material depends on the chemical nature of dye molecules, they gave no quantitative data on the strength of such affmity. Brandon et al. (1979) studied the practical application of reverse osmosis process for treating the wastewater from textile dyeing operation. Probably the most systematic and comprehensive work on the reverse osmosis separation of dyes was conducted by Masuda et al. (Masuda, 1976; Masuda et al., 1980a,b). Using feed solutions containing sodium chloride, and acidic and other dyes, they found that the product rate decreased and the separation of sodium chloride increased, while separations of dyes decreased with increase in sodium chloride concentration in the feed solution. These experimental observations were attributed to the adsorption of dye solutes onto the membrane surface, which was confirmed by independent adsorption measurements. They concluded also that the membrane material had no effect on the dye separation on the basis that the dye separation by a polysulfone ultrafiltration membrane was nearly equal to that by a cellulose acetate membrane. The latter conclusion, however, is groundless since the comparison of separations by membranes of different polymer materials was made without any reference to the pore size and the pore size distribution of the membranes used. Two membrane materials, namely, cellulose acetate and aromatic polyamidohydrazide PPPH 8273, are considered in this work. From our previous works (Matsuura and Sourirajan, 1971; Dickson et al., 1975; Matsuura et al., 1974b, 1975,1977) on the separation of various organic and inorganic solutes, it was concluded that cellulose acetate membranes are basic, while aromatic polyamide and amidohydrazide membranes are acidic in nature. Consequently, cellulose acetate material should exhibit a stronger affinity to basic dyes than to acidic dyes, while in the case of aromatic polyamidohydrazide material, the order should be reversed. Moreover, the change in the interaction force has to be reflected in the membrane performance data. In this paper the effect of the polymer material on the interaction force is studied by using the retention volume data of liquid chromatography, while the effect on reverse osmosis separation is investigated by comparing the solute separation data from membranes of the same average pore size. The method for the above study is illustrated and the results are reported and discussed.
Theoretical Section The theoretical analysis of combined high performance liquid chromatography (HPLC) and reverse osmosis data in this paper is the same as described in detail in the previous work (Matsuura et al., 1981a,b) based on the surface force-pore flow model for reverse osmosis transport (Matsuura and Sourirajan, 1981). The chromatography data on retention volumes enable one to calculate the interfacial equilibrium distribution coefficient of a solute A, designated as KA', by KA'
=
([VR'IA
- [VR'lmin)/([VR'lwater
t 101
DYE no. 1 t I
1
I
CELLULOSE ACETATE
I
a '1.0
9
L
I
P=
-1.0
'1 ~
-'.O
'"1 1
AROMATIC POLYAMIDOHYDRAZ IDE PPPH8273
b
DYE no. 5
I
I
CELLULOSE
"1 ~
ACETATE
1 1
AROMATIC POLYAMIDOHYDRAZIDE PPPH8273
t
Figure 1. Potential curve of interfacial force, (a) System dye no. m3;D = 3.56 X 1-CA: I, A = 1.8 X m; B = 42.72 X m; hypothetical systems; 11, A = 1.8 X m; B = 0; D = 3.56 X m3; D = 3.56 X m; 111, A = 0; B = 42.72 X m. (b) System dye no. 1-PPPH 8273: I, A = 0.9 X m, B = 97.66 X lo-%m3;D = 5.12 X m; hypothetical systems: 11, A = 0.9 X m; B = 0; D = 5.12 X m; 111, A = 0; B = 97.66 X m3; D = 5.12 X m. (c) System dye no. 5-CA: A = 2.0 X 10-lom; B = 119.2 X m3; D = 5.48 X m. (d) System dye no. 5-PPPH 8273: A = 1.0 X lo-'' m; B = 155.2 X m3;D = 4.60 x W0m.
above should be further related to the force constants involved in the potential function associated with the interaction force working on the solute molecule from the polymer material. Since a dye molecule contains an ionic part, which causes an electrostatic repulsive force, and one or more aromatic rings, which cause van der Waals attractive force, the form of potential function should involve both terms of electrostatic repulsion and van der Waals attraction. Such a potential function has the form (Matsuura and Sourirajan, 1981; Matsuura et al., 1981a) very large (when d ID) (24 A B $J = -RT - -RT (when d > D) (2b) d d3 where A and B are the force constants associated with the strength of electrostatic repulsion and van der Waals attraction, respectively, and D is the distance associated with steric repulsion. Some typical potential functions of dye molecules are illustrated in Figure 1. When a nonelectrolyte such as glycerol solute is considered, A is equal to zero and the term expressing the electrostatic force vanishes. Using the Maxwell-Boltzmann distribution law, the concentration profile at the polymer solution interfacial region may be written as $J =
C A ~= CAbe-9/RT
= 0 (d 5
D)
(34
- [VR'lmin) (l)
where [ VR']A, [ VR']w&, and [ VR']- denote the retention volume of A, that of water which is considered to be equal to that of DzO, and that of a reference solute which demonstrates the least retention volume among all solutes tested, respectively (Matsuura and Sourirajan, 1978). Raffinose was chosen as the latter reference solute for both CA and PPPH 8273 columns. The value of K A ' obtained
where cAi and CAb denote the concentration of solute A at the interfacial region and bulk solution, respectively. Calculating the surface excess of the solute on the basis of the interfacial concentration profile given by eq 3 and assuming that excessive amount of solute is compressed in the region of interfacial water, the thickness of which,
80
-
Ind. Eng. Chem. prod. Res. Dev., Vol. 22. No. 1. 1983
t , was obtained in the previous work for both cellulose acetate and aromatic polyamidohydrazide materials as 9.5 X lO-'O m and 6.8 X m, respectively (Matsuura et al., 1981h),the average solute concentration at the interfacial water phase, cAi,can he obtained hy averaging the surface excess over the thickness of the interfacial water layer. When cAiis divided by the concentration in the hulk solution cAb,we obtain the equilibrium distribution coefficient as
K,' = Figure 2. Schematical description of radii R. and Rb.
(Matsuura et al., 1981a,h). In eq 4 D, denotes the radius of water molecule which is taken as 0.087 nm in this work. The analytical techniques involving liquid chromatography data on the basis of the above theory are documented in the literature (Chan et al., 1982; Farnand et al., 1983; Matsuura and Sourirajan, 1978; Matsuura et al., 1981a,h,c,, 1982a,b; Taketani et al., 1982a,h). In the earlier work (Matsuura and Sourirajan, 1981; Matsuura et al., 1981a) describing the transport in a cylindrical pore under the influence of interaction forces, solute separation based on the boundary concentration was expressed as
where a ( p ) is a dimensionless velocity profile in the membrane pore as a function of dimensionless radius p . which is obtained solving a differential equation involving all forces working on the solvent fluid under the operating conditions of reverse osmosis. The quantities b and Q, representing frictional function and potential function, respectively, of interfacial force working on the solute molecule from the membrane pore wall, are given by b = 1/(1- 2.104X + 2.09Xs - 0.95h5) (when X 50.22) (64 b = 44.57
- 416%
sumed to develop. Between R, and R,, the relationship is Rb = R, + Dwam,since in the coaxial region (illustrated as shadowed region in Figure 2) the center of water molecule cannot exist. The symbol r denotes any radial distance and the dimensionless quantity p is defined as r/Ra. From the solute separation f'hased on the boundary concentration, the solute separation f based on the feed concentration can he calculated by 1 (8) = f' + (1 - f l exp(u./k) where u, is the linear velocity of the permeate solution through the membrane which is practically the same as u,* = AP/c for dilute solutions, where A denotes pure water permeation constant calculated by
A=
(PWP)
(9)
3600*S*M~*P
Furthermore, the mass transfer coefficient in the boundary phase for sodium chloride solute, kNaCI.is presented as a function of pure water permeability constant, A, by the relation kN.cl = 83.60A 3.00 X 10" (when A 5 0.623 X 10") (loa)
+
kNaCI
=
23.81A
+ 40.05 X 10"
(when A >0.623 x
lo")
(lob)
which was produced in this work for the reverse osmosis cell and the feed flow rate used hy applying the basic equations developed by Kimura and Sourirajan (Sourirajan, 1970). For the solutes used in this study, mass transfer coefiicient k can he obtained from the relation (Matsuura et al.. 1974a)
+ 934.9X2 +
302.4X3 (when 1 > X > 0.22) (6b)
where X is defined as D/Rb, and Q = very large (when (R,/R.) - p 5 (D/R.))
(7a)
Equation 6 was empirically derived in the earlier work (Matsuura et al., 1981a) and it demonstrates a very steep increase in the quantity b with the increase in A. Equation 7 can he obtained from eq 2 by replacing d by Rb - r and rearranging quantities involved into dimensionless forms. The meaning of symhols used in eq 7 is as illustrated in Figure 2. Rb is the radius of the pore. R, is the radius of water channel where a laminar flow of solvent is as-
where Dm and (DAB)NacI denote the diffusivity of the dilute solution of the solute and of sodium chloride solute. When eq 4,5,6 and 7 are examined, it is found that both f'and KA' are functions of quantities A, B, D, and Rb for given reverse osmosis operating conditions, since the dimensionless velocity profile a ( p ) is uniquely determined for a uarticular oueratinp- condition (Matsuura and Sourirajan, 1981). Usine ea 4-7. RL is determined in the first s t e of ~ calculatioi foi each Gemhrane used in this study. 'For this purpose glycerol was chosen as a reference solute and its Stokes' law radius, rA, was regarded as D for glycerol. Since glycerol has no charge, A is considered to he equal to zero. Values for A and D being known, the value of B for glycerol solute with respect to the cellulose acetate material-aqueous solution system can he obtained hy a p plying eq 4 and an experimental KA' value obtained from
Ind. Eng. Chem. Prod. Res. Dev., Vol. 22, No. 1, 1983 81
chromatography data of glycerol solute and cellulose acetate column. Then, from the reverse osmosis experimental data on f with respect to reference glycerol solute and a given cellulose acetate membrane, f'is calculated by using eq 8-11. Finally, the quantity f'thus calculated is used in eq 5,6, and 7 to solve for Rb. The calculation of Rb was applied to all cellulose acetate membranes used in this study. Rb values for all PPPH 8273 membranes were calculated by the identical method. The results of the calculation are given in Table I1 together with the experimental data for glycerol. For analyzing data from experiments with a given dye solute and CA membrane material, data on f'from three CA membranes of different average pore sizes and KA' for the particular dye solute and CA material are used. Since Rb is known to each one of the membranes used, there are only three parameters A, B, and D left to be determined in this system. It should be noted that parameters A, B, and D are dependent only on a membrane material-solvent-solute system and independent of the pore structure of the membrane. Therefore, a nonlinear regression analysis enables one to solve these three parameters on the basis of four data, including three f'values and one KA' value. This calculation method is applied to all dye compounds under consideration. The same method is also applicable to PPPH 8273 membrane material. With known values of A, B, and D, then, the calculation of the separation for any given pore radius, Rb, becomes possible.
Experimental Section HPLC Experiment. A liquid chromatograph, Model ALC202 of Waters Associates, fitted with a differential refractometer was used in this work. The method of column preparation and the general experimental technique used were the same as those reported earlier (Matsuura et al., 1976). Briefly, glycerol and dye solutes were injected into the solvent water stream which flows through a column packed with membrane polymer powder. The particle size was kept in the range 38-53 pm by sieving and the column length was 60 cm. The chromatographic property of cellulose acetate Eastman E-400 material was represented by that of cellulose acetate Eastman E-398 material. This approximation is justified on the basis that the acetyl contents in both polymers are not too far different. Cellulose acetate material in the chromatography is abbreviated below as CA. The solvent water flow rate through the column was fixed at 0.275 cm3/min. Ten microliters of sample solution (1wt % solute) was injected into the column and the retention volume was determined. In the case of heavy water, 10 wt 5% solution was used as the sample. Reverse Osmosis Membranes. Cellulose acetate Eastman E-398 (CA-398) membranes of two different surface porosities made in the laboratory by the method described by Pageau and Sourirajan (1972) were used in this work. In this method the average pore size of the membrane was controlled by changing the shrinkage temperature. A cellulose acetate Eastman E-400 (CA-400) membrane which was made by the method described by Kutowy et al. (1978) was also used in this work. Aromatic polyamidohydrazide (PPPH 8273) membranes of different surface porosities were also made in the laboratory by the method described in the earlier work (Matsuura et al., 1977), in which the porosity of the membrane was controlled by changing the evaporation time of the solvent. The details of the membrane formation process are summarized in Table 111. Reverse Osmosis Experiments. Each membrane was subjected to an initial pure water pressure of 2068 kPa
0
w
4
0
c
3
cu
m
:
82
Ind. Eng. Chem. Prod. Res. Dev., Vol. 22, No. 1, 1983
Table 111. Film Casting Details Cellulose Acetate Membranes film no. CA-1
CA-2
acetone water magnesium perchlorate casting conditions casting solution temperature, "C temperature of casting atmosphere, "C solvent evaporation time gelation medium composition
E-398-3 17.0 69.2 12.35 1.45
E-398-3 17.0 69.2 12.35 1.45
E-400-25 14.8 63.0 19.9 2.3
10 30 1 min ice water
10 30 1 min ice water
shrinkage temperature, "C
70
70
23-25 23-25 3s 30 wt % ethanol-water solution, 1 "C no shrinkage
~~~~~
CA-3
~
casting solution composition, wt % cellulose acetate polymer
PPPH 8273 Membranes ~
film no. PPPH 8273-1 casting solution composition, wt % lithium nitrate PPPH 8273 polymer N,N-dimethyl acetamide casting conditions solvent evaporation temperature, "C solvent evaporation time, min gelation medium
PPPH 8273-2
PPPH 8273-3
4.76 14.29 80.95
4.76 14.29 80.95
4.76 14.29 80.95
95 7 ice water
95 5 ice water
90 6 ice water
gauge (= 300 psig) for about 2 h prior to subsequent use in reverse osmosis experiments all of which were carried out in the operating pressure range of 1724 kPa gauge (= 250 psig) and at the laboratory temperature (23-25 "C). For the purpose of membrane specifications in terms of pure water permeability constant A (in kg-mol of H20/m2 s kPa) and solute transport parameter (DM/KS) (treated as a single quantity, m/s), aqueous feed solutions containing 3500 ppm of NaCl were used (Sourirajan, 1970). Data on A and (DM/KB)thus obtained are listed in Table I1 with respect to all membranes used in this work together with experimental reverse osmosis data of sodium chloride solute. In all other experiments the solute concentrations in the aqueous feed solutions were so low (
