Mechanism of Formation of Reverse Osmosis Membranes Precipitation of Cellulose Acetate Membranes in Aqueous Solutions Moshe A. Frommer’, Reuven Matz, and Uri Rosenthal Hydronautics-Israel, Ltd., Kiriat W e i z m a n n , RehoGot, Israel, and Hydronautics, Inc., Pindell School R d . , Laurel, M d . 20810 The phenomena involved in the precipitation of cellulose acetate membranes by aqueous solutions were studied. The density of reverse osmosis cellulose acetate membranes precipitated in aqueous solutions increases with the decrease of the water activity of the leaching bath. The desalination characteristics of these membranes can be controlled by suitably choosing the gelling media; “salt gelling” can be used to reduce or even eliminate the need for the annealing stage. The density of membranes gelled in aqueous solutions is determined by the rates of flow of liquids into and out of the cast solution during leaching. By lowering the water activity in the precipitation bath, the rate of flow of water into the cast solution i s lowered. The rate of flow of solvent out of the cast solution hardly i s affected by the composition of the aqueous solution. Consequently, the polymer solution concentrates prior to its precipitation by the nonsolvent, and the resultant membrane is denser.
R e c e n t l y , Bloch and Frommer (1970) and Frommer et al. (1970) showed that the structure and desalination properties of cellulose acetate membranes cast from various binary solutions and leached in an ice water bath are very significantly affected by processes taking place during precipitation of the polymer in the nonsolvent bath. More specifically, the density of the main bulk of the membrane can be correlated to: The concentration of nonsolvent required for precipitating the cellulose acetate from the casting solution; and the direction and magnitude of osmotic flows of liquids into and out of the cast solution during leaching. However, until recently, the parameters used for controlling the desalination properties of reverse osmosis membranes were primarily related t o the composition of the casting solution, or to postprecipitation treatments such as hot water annealing. This paper shows that even by utilizing the same casting solution, the density of the porous substructure of cellulose acetate membrane may be varied by varying the composition of the leaching bath from pure water to aqueous solution. Moreover, the influence of solutes on the porosity of the resultant membrane can be explained in terms of the abovementioned mechanism. Experimental
Materials. The cellulose acetates used were Eastman’s types E-398-3 and E-398-10. Solvents [acetone, dioxane, acetic acid, formamide, dimethyl formamide ( D M F ), dimethyl sulfoxide (DMSO), and triethyl phosphate (TEP)]were either of analytical grade purity or redistilled from technical grade liquids. Salts and sucrose were of analytical grade purities. Procedures. Porous cellulose acetate membranes used
T o whom correspondence should be addressed.
for measuring the direction and magnitude of osmotic flows were cast from 2 0 5 dried E-398-3 cellulose acetate in T E P , leached immediately in an ice water bath, annealed 30 min a t 80”C , and stored in water. Annealing eliminated pinholes and imperfections possibly existing in the nonannealed membrane. Reverse osmosis membranes were cast a t 18-mils thickness from 2 5 2 dried E-398-10 cellulose acetate in 45 to 30 (wt 6 ) acetone-formamide solution, leached at room temperature for one hr in an aqueous solution, washed in running water for one hr, and then tested for thickness, porosity, and desalination properties by standard procedures (Bloch and Frommer, 1970; Matz e t al., 1970). The procedure for determining the concentration of water or aqueous solution required for precipitation of cellulose acetate from a 20% solution was identical to that described previously (Frommer et al., 1970). The direction and magnitude of volume flow through a porous cellulose acetate membrane separating water or a salt solution from a 2:3 (weight ratio) water:solvent solution were measured a t 30.0” i 0.1” C in a polypropylene cell described in a previous paper (Frommer et al., 1970). However, since the authors were interested in elucidating qualitative effects, the measuring capillaries were placed vertically and changes in the levels of the liquids were determined for approximately one hour. Results
Figure 1 describes the influences of various aqueous solutions on the “partial shrinkage” of the membrane, namely on the ratio (Ad d , ) of the difference ( A d ) between the thickness of the cast layer ( d o )and the final thickness of the membranes ( d ) to the casting thickness ( d o ) . The thickness of the cast layer generally is lower than the opening of the casting knife. I n this case, the ratio between the thickness of cast layer to the opening of the casting knife was 0.87. as calculated from the measured thickness Ind. Eng. Chem. Prod. Res. Develop., Vol. 10, No. 2, 1971
193
of an air-dried membrane, the measured 10% porosity of such a membrane, and the known volume fraction of the components in the casting solution. The maximal possible “partial shrinkage” is 0.84 and not 0.75 since the volume occupied by the 26% cellulose acetate in the casting solution is only about 1670 of the total volume, owing to its higher density. Figure 1 shows that the partial shrinkage of the membrane increases with the decrease in the activity of the water in the leaching bath (almost irrespective of the nature of the solute in the aqueous solution). This dependence is especially sharp in the range of water activities from 0.7 to 1.0. Water activities of the various solutions were taken from Robinson and Stokes (1968). That the density of cellulose acetate membranes precipitated in concentrated aqueous solutions increases with the decrease in the activity of the nonsolvent (water) in the precipitating bath can be exemplified further by the data of Figure 2 . Again it is obvious that the porosity (as measured from the water content) of the leached membranes is mainly dependent on the activity of the 0.5
~
~~
Table I. Concentrations ( W t O h ) of Water and Aqueous NaCl Solutions Required to Precipitate Cellulose Acetdte from 20 W t YOSolutions a t Room Tem pera t u rea Nonsolvent Solvent
H?O
1N NaCl
Acetone Dioxane Acetic acid DMF DMSO 3 Acetonei2 Formamide
28 30 31 12 15 15
24 < 15 30 11 15 14
2N NaCl
< 10 14
Figures are meaningful within 2% error. Table II. Influence of Salt on Direction and Magnitude of Volume Flows through a Porous Cellulose Acetate Membrane Separating Water or Salt Solution from a 2:3 Water-Solvent Solution” Solvent in the 2:3 water-solvent solution
Dioxane Acetic acid DMSO
Volume flow NaCl concn in water compartment
Direction
0 4N 0 4N 0 4N
Solution to water Solution to water Not detectable Solution to water Water to solution Water to solution
Magnitude, PI/ cm’ min 100 microns
0.08 0.9
...
0.2 4 0.12
“ T h e capillary tubings in the measuring cell were vertical, and the changes in the levels of the liquids were determined 20 to 60 min after introducing the liquids into the cell. 0.E 0 CoCl
h
210” W
N o C l SOLUTIONS
A
L i C l SATURATED
e “,NO3
W
SATURATED
0 KN03 SATURATED
(3
5Z
SOLUTIONS
0
0.7
*
B
I
VI A
SUCROSE SOLUTIONS WATER AIRDRIED
ri
k
2
0.6
0.5 0
0.2
0.4
0.6
0.8
1 .o
WATER ACTIVITY
Figure 1. Dependence of the ”partial shrinkage” - Ad/d,, of cellulose acetate membranes precipitated, at room temperature, in various aqueous solutions, on the water activity of these solutions Water activities were taken from Robinson and Stokes (1968). The “airdried” membrane was allowed to evaporate over 48 hr
194
Ind. Eng. Chem. Prod. Res. Develop., Vol. 10, No. 2, 1971
water in the leaching bath, and much less sensitive to the nature of the solute in the aqueous solution, either electrolyte or sucrose. An important practical outcome of these observations is the possibility of controlling the desalination characteristics of reverse osmosis membranes by suitably choosing the gelling media. As shown in Figure 3, the flow rates through cellulose acetate membranes cast from Loeb-Sourirajan “modified” solution decrease, and the salt rejections of these membranes increase with the increase in the concentration of NaCl in the precipitation bath. “Salt gelling,” therefore, can be employed to reduce the annealing temperature, or even eliminate the annealing stage. More detailed data on the influence of various electrolytes in the leaching bath on the desalination characteristics of cellulose acetate membranes were reported recently (Zisner and Loeb, 1970). T o test whether these observations can be explained in terms of a mechanism suggested previously (Frommer et al., 1970) for explaining the structure of cellulose acetate membranes cast from binary solutions and leached in water the authors have determined the concentrations of salt solutions required for precipitating the polymer from the casting solution; and estimated the directions and magnitudes of flows of liquids into or out of the cast solution during leaching. The influence of salt on the concentratioti of water required for precipitating cellulose acetate is given in Table I. The concentration of 1N NaCl aqueous solution was chosen because, on the one hand, it is the lowest salt concentration which can be used practically for eliminating annealing of desalination membranes (Zisner and Loeb,
0
0.80
CaCI, SOLUTIONS
o N a C l SOLUTIONS
./
L i C l SATURATED
A
0.70
v ",NO,
SATURATED
0 K N 0 3 SATURATED
0.60
SUCROSE SOLUTIONS
*
2
9
0.50
0
AIR DRIED A N D WATER SWELLED
PI UJ
g
WATER
AIR DRIED 0
0.40
-I
5 !c
h
0.30 A
0.20
0.10
,
0
I
I
I
I
,
4
WATER ACTIVITY Figure 2. Dependence of the partial pore volume of cellulose acetate membranes precipitated, a t room temperature, in various aqueous solutions, on the water activity of these solutions 8C
---_
FLUX REJECTION
90
70
z
e
z 3
s
\
5c
X 2
U
30 70
/ 20
d
IC
/
/'
1
60
0 0.2
0.3
0.4
0.5
0.6
0.7
0.0
0.9
1.0
1.1
1.2
1.3
1.4
1.5
$ N a C I I N VREClPlTAllON 8ATH
Figure 3. Dependence of the flow rate through, and the salt rejection of, cellulose acetate membranes precipitated, at room temperature in aqueous NaCl solutions, on the concentration of NaCl in these solutions Desalination characteristics determined on 0.1 N NaCl solutions a t 800 psi
1970), and on the other hand, it is the most common high salt concentration which can be prepared in the mixed water-organic solvent solutions. Table I shows that, in most cases, the addition of NaCl to the water only slightly lowers the water concentration required for precipitation. Dioxane represents the only solvent, having high solvation power for cellulose acetate, which is very dramatically affected by the addition of NaCl to the nonsolvent, and this, presumably, is because dioxane is the only watersoluble nonpolar solvent (as judged by its zero permanent dipole moment) for cellulose acetate. The influence of salt in the water compartment on the direction and magnitude of osmotic flows through a porous cellulose acetate membrane separating water or salt solution from a 2:3 water:solvent solution is given in Table 11. The figures represent rates of volume flows from 20 to 60 min after introduction of the liquid into the cell, and are lower than those given previously (Frommer et al., 1970) both because they are not initial rates, and because of the formation of a small counter hydrostatic pressure. For solvent in the so1vent:water solution, dioxane, acetic acid, or DMSO was used. Frommer et al. (1970) showed that these systems may represent the three possible directions of volume flows. In the case of dioxane, there is a net volume flow from the solution into the water compartment. When acetic acid is used as the solvent in the water-solvent solution, there is no volume flow through the membrane to either direction, whereas, when DMSO is contained in the solution, there is a net volume flow from the water into the solution compartment. The data of Table I1 show that the replacement of water in the water compartment by 4N NaCl solution Ind. Eng. Chern. Prod. Res. Develop., Vol. 10, No. 2, 1971
195
considerably affects the magnitude of the total volume flow. In the dioxane system, the addition of salt into the water compartment caused a significant increase in the rate of volume flow from the solution compartment into that containing water. When acetic acid composes the solvent in the water-solvent solution, there is no volume flow in either direction. However, when salt is added to the water compartment, a net volume flow from solution to water is created. In the DMSO system, the addition of salt into the water compartment results in a reduction of the rate of volume flow from water to solution. Discussion
The results described above can be summarized in three basic observations: (1) The density of a cellulose acetate membrane precipitated in an aqueous solution is higher than that of a membrane precipitated in pure water. This is reflected by the lower thickness, lower porosity, lower permeability to water, and higher salt rejection of the “salt-gelled’’ membrane. The increase in density is determined primarily by the decrease in the activity of the water in the leaching bath rather than by the nature of the solute. (2) The concentration of salt solution required for precipitating cellulose acetate from a 2070 binary solution is essentially identical to the concentration of pure water required for the same purpose. (3) Concerning the direction and magnitude of volume flows through a porous cellulose acetate membrane separating water from a 2-to-3 water-solvent solution, the addition of salt to the water compartment considerably affects the direction and the magnitude of osmotic flows through the membrane. Observations 1 and 2 and, independently, observation 3 indicate that the lowering of the water activity in the precipitation bath considerably lowers the rate of penetration of water (nonsolvent) into the cast solution, but hardly affects the rate of flow out of the solvent. This conclusion may be tested against the data presented in Table 11. In the dioxane system, the rate of solvent flow from the solution to the water compartment is greater than the net water flow into the solution. If it is correct to assume that the lowering of the water activity lowers the net water flow into the solution but hardly affects the rate of dioxane flow, then the overall volume flow from solution to water should increase. This was found experimentally. Similarly, in the acetic acid system, there is no volume flow in either direction when pure water is in the second compartment. If, as assumed, the salt affects mainly the rate of water flow, its addition to the water compartment will result in a net volume flow from the solution into the 4N NaCl compartment, since the net water flow is now lower than the magnitude of solvent flow out of the solution compartment. Similar considerations can explain the behavior of the DMSO system. The same conclusion-i.e., that the existence of solute in the precipitation bath lowers the rate of penetration of water into the cast solution but hardly affects the rate of flow of solvent out of it-is indicated also by observations 1 and 2. If the hypothesis that the density of the membrane is determined by the density of the corresponding polymer solution a t the precipitation point is accepted, then the factors which determine the density of a membrane prepared by the “casting-leaching” method 196
Ind. Eng. Chem. Prod. Res. Develop., Vol. 10, No. 2, 1971
are: The concentration of nonsolvent required for precipitating the polymer from the cast solution; and the relative magnitudes of flows of solvent out of the cast solution and nonsolvent into the cast solution. The authors have, however, shown (observation 2) that the concentration of salt solution required for precipitating cellulose acetate from 20% binary solution is essentially identical to the concentration of pure water required for the same purpose. That membranes precipitated in concentrated aqueous solutions are denser than those precipitated in pure water (observation 1) indicates, therefore, that the cast polymer solution concentrated prior to its precipitation by the nonsolvent. The lowering of the water activity in the leaching bath lowers the rate of its penetration into the cast solution. The rate of solvent flow out of the cast solution is much less affected by the existence of solute in the leaching bath. For this reason, the density of the membrane is determined primarily by the activity of the water in the leaching bath rather than by the nature of the solute. Understanding the mechanism of “salt gelling” is of great practical importance as it enables the regulation of the desalination characteristics of reverse osmosis membranes by suitably choosing the composition of the leaching bath. Salt gelling can be employed for reducing the annealing temperature as shown in Figure 3, and possibly for eliminating the annealing step altogether, for certain membrane applications. Detailed data on the influence of various electrolytes in the leaching bath on the desalination characteristics of cellulose acetate membranes were reported recently (Zisner and Loeb, 1970). I t must be recalled, however, that the permeability to water and the selectivity to salt of reverse osmosis membranes are determined much less by the average porosity of the membranes than by the nature of the “skin” (the thin, dense selective layer). The differences between the rates of water flows and the salt rejections of membranes precipitated in aqueous solutions containing various electrolytes but displaying the same water activity, therefore, might be attributed to differences in the nature of the “skin” induced by the specific electrolyte. A more detailed study of the relations between the composition of the leaching bath and the structure of the resultant membrane as observed by scanning electron microscope will be reported soon. Acknowledgment
The authors acknowledge the stimulating discussions with Ora Kedem and Rene Bloch. Literature Cited
Bloch, R., Frommer, M. A., Desalination, 7, 259 (1970). Frommer, M. A,, Feiner, I., Kedem, O., Bloch, R., ibid., 7, 393 (1970). Matz, R., Tulin, M. P., Gollan, A,, Preiser, H. S., Alcalay, H., U.S. Dept. of the Interior, Office of Saline Water, R and D Progress Rept. No. 542, 1970. Robinson, R. A., Stokes, R. H., “Electrolyte Solutions,” Butterworths, London, England, 1968, pp 476-510. Zisner, E , , Loeb, S., Fresh Water Sea, Proc. I n t . S y m p . , 3rd., (Dubrovnik), 2,615 (1970). RECEIVED for review August 17, 1970 ACCEPTED January 15, 1971 Work supported by a research grant, No. 14-30-2529, from the Office of Saline Water, U. S.Dept. of the Interior.