Ind. Eng. Chem. Process Des. Dev. lQ83, 22, 625-632
HDN; however, this effect was not due solely to the bassanite, which was the major mineral in the lignite. A 20% feldspar-KY No. 9 mixture showed an increase of 38% in HDN activity over the pure mineral matter sample. Finally, it was observed that HC1 treatment prior to activity testing destroys the catalytic ability of the KY No. 9 sample by loss of iron and titanium but does not affwt that of the KY No. 11. Similar treatment with water and with a KC1 solution were also deleterious to KY No. 11mineral matter. Acknowledgment The experimental work of David Klueh is gratefully acknowledged. This work was supported by US DOE under Contract No. EX-7642-01-2233. Literature Cited
625
Guin, J. A.; Tamer, A. R.; Lee, J. M.; Lo, L.; Curtls, C. W. Ind. Eng. Chem. Process Des. Dev. 1979, 18, 371. Johannes, A. H.; Hamrin, C. E.; Jr. Fuel Process. Techno/. 1983, in press. Katzer, J. R.; Shrasubramanian, R. Catel. Rev. Sci. fng. 1979, 20, 155. Liu, K. H; D.; Johannes, A. H.; Hamrin, C. E., Jr. Fuel 1983, in press. Lycourghiotis, A.; Katsanos, N. A.; Vattls. D. J . Chem. Soc.. Faraday Trans. 11979, 75, 2401. McIlwied, H. G. Ind. Eng. Chem. Process Des. Dev. 1971. 10, 125. Morodta, S.; Hamrin, C. E.,Jr. Fue/ 1978. 57, 776. Sakata, Y.; Hamrln, C. E., Jr. Fuel 1989a, 62, 508. Sakata, Y.; Hamrin, C. E., Jr. Ind. Eng. Chem. Prod. Res. Dev. 1983b, 22, 250. Satterfidd, C. N.; COCChettO, J. F. AIChE J . 1975, 21, 1107. Smlth, H. A. "Catalysis", Vol. V., pp 231-234; P. H. Emmett, Ed.; Reinhold, New York, 1957. Sonnemans, J.; Neyens, W. J.; Mars, P. J . Catel. 1974, 34, 230. VaMyanathan, B. S. M.S. Thesis, Unlversky of Kentucky, Lexington, Kentucky, 1977. Yavorsky, P. M.; Akhtar, S.; Friedman, S. Preprint, 65th Annual A I C M Meeting, New York, 1972.
Received for review July 22, 1982 Revised manuscript received January 20, 1983 Accepted January 29, 1983
Cocchetto, J. F.; Satterfield, C. N. Ind. Eng. Chem. Process Des. Dev. 1978, 75, 272. Garg, D.; Givens, E. N. Ind. Eng. Chem. Process Des. Dev. 1982, 21, 113.
Aromatic Hydrocarbon-Water Separations by a Pressure-Driven Membrane Separation Process J. M. Dlckson,+ Mehrdad Babel-Plrouz, and Douglas R. Lloyd' Department of Chemical Engineering, The Unlvershy of Texas at Austin, Austin, Texas 78712
This project is concerned with the removal of aromatic hydrocarbons from water by a pressuredriven membrane
separation process. In this paper, the reverse osmosis separation of single solute aqueous solutions is reported. Data are presented for three solutes (benzene, toluene, and cumene) at feed concentrations ranging from 5 to 260 ppm under four applied pressures and using six cellulose acetate membranes of different surface porosity. The results are explained in terms of the magnitude and direction of solute-membrane interactions.
Introduction This paper deals with the removal of aromatic hydrocarbon solutes from water by the pressure-driven membrane separation process commonly referred to as reverse osmosis. Under the influence of economic and environmental considerations, the efficient removal and recovery of such organic compounds from dilute aqueous solutions has received considerable attention recently. If membrane separation processes are to compete with existing technologies, it will be necessary to develop a thorough understanding of the phenomena which govern such separations. In the reverse osmosis process a membrane is mounted in a cell and a high-pressure feed solution is circulated over the surface of the membrane. A portion of this solution is transported through the membrane and emerges as a permeate stream. The rate at which solution components permeate through the membrane, and thus the ability to separate these species, depends on both the nature and the strength of the interactions between the system components (solvent, solute, and membrane material) as well as the morphology of the membrane under the specific operating conditions. The interactive forces between system components may be such that either the 'Department of Chemical Engineering, McMaster University, Hamilton, Ontario, Canada. 0196-4305/83/1122-0625$01.50/0
solvent or the solute has a stronger affinity for the membrane. The differences between these two situations lead to markedly different separation behavior in pressuredriven membrane separation processes. The solutes under consideration in this paper are ones which, for various physicochemical reasons, exhibit a strong affinity for the cellulose acetate (CA) membrane material; that is, the solute experiences a stronger attraction to the membrane surface than does the solvent water. This is in contrast to the more familiar case of sodium chloride in water where the water-membrane affinity is greater than is the solute-membrane affinity. (Actually in this case the NaCl is repelled by the CA). The aromatic hydrocarbons investigated here are representative of a large number of organic solutes of industrial importance which show strong attractive interaction with the commercially available membrane materials. Thus, it is essential that a thorough understanding be gained about these systems and the controlling factors. In this paper, the manner in which the performance criteria (separation and flux)are influenced by operating variables (operating gauge pressure and feed solution concentration), the physicochemical properties (such as the nonpolar characteristic of the solute), and membrane pore size is investigated. The study focuses on the system cellulose acetate (membrane material)-water (solvent)-aromatic hydrocarbon (solute). In particular, single-solute dilute 0 1983 American Chemical Society
626
Ind. Eng. Chem. Process Des. Dev., Vol. 22, No. 4, 1983
aqueous solutions of benzene, toluene, and cumene are dealt with as model systems. Solute selection was based on considerations of range of nonpolar or dispersive character (as indicated by Small’s molar attraction constants), water solubility, and industrial as well as environmental significance. The selection of membrane material was based on the wealth of knowledge available in the membrane literature on this material and on the widespread usage of cellulose acetate as the membrane material in commercial units. A number of models have been developed to describe transport and separation in reverse osmosis (Lonsdale et al., 1965,1967; Spiegler and Kedem, 1966; Merten, 1966; Sourirajan, 1970a;Jonsson and Boesen, 1975; Pusch, 1977). Attempts to apply these models to situations in which the solute-membrane affinity dominates are reviewed by Dickson and Lloyd (1981); further developments can be found in papers by Burghoff et al. (1980) and Jonsson (1980). Work in our laboratory directed toward developing a qualitative and quantitative description of the reverse osmosis performance of systems dominated by solutemembrane affinity (Dickson and Lloyd, 1981) has indicated that of the available models, the one most successful in dealing with this situation is an interactive pore flow model proposed by Sourirajan (1970a) and referred to as the “preferential sorption-capillary flow” model. According to this model, if the membrane has an appropriately porous structure and if there is a difference in either the magnitude or the direction of the interactive forces between the membrane and the constituents of the feed solution, then the permeate solution will differ in concentration from the feed solution. It is important to understand that both the pore size as well as the nature and strength of the interaction between solution components and the membrane are critical in determining the separation and flux. In a single solute solution both the solvent (water) and the solute will experience either net forces of attraction to or repulsion from the membrane material. In the case where either the solute-membrane repulsion or the water-membrane affinity dominates (referred to as “water preferential sorption”), an essentially pure molecular layer of water can be considered to exist in the region immediately adjacent to the membrane. By applying pressure to the system, this molecular layer of water is transported through the membrane pores and immediately another layer of water replaces the previous one. The result is the separation of the solution components by the membrane system and the formation of a water-rich permeate. Quantitatively, separation (f) is defined as
where ml and m3 are the feed and permeate molalities, respectively. Water preferential sorption is the more familiar case; it includes most inorganic and many polar or ionized organic solutes with cellulose acetate membranes and the solvent water. These systems have been reported extensively in the literature. Some general statements may be made concerning the reverse osmosis transport. (i) Increasing the operating pressure usually increases separation. (ii) The decrease in permeate flux with increasing feed concentration is due to the osmotic pressure of the solution. (iii) Positive separation is observed. However, when the solute-membrane affiiity dominates (referred to as ”solute preferential sorption”), these statements do not hold true. Systems involving solute preferential sorption (Matsuura and Sourirajan, 1973) are
Table I. Physicochemical Dataa
solute
structure
benzene
C,H,-H C,H,-CH, C,H,CH(CH3 )*
toluene cumene a
molar molar attraction solubility in water constant, (calcm3)’/*/ x lo3 at 25 ‘C, g-mol kmol/m3
825 949 1191
22.79 5.59 0.42
Reprinted from Matsuura and Sourirajan (1973).
characterized by the following. (i) Increasing the operating pressure tends to decrease the separation. (ii) The permeate flux is significantly less than the pure water permeation flux, even when osmotic pressure effects are negligible. (iii) The separation may be positive, zero, or negative depending on the specific operating conditions. The physicochemical criteria for the reverse osmosis separations of several classes of organic solutes in aqueous solutions using Loeb-Sourirajan type porous CA membranes have been extensively discussed (Sourirajan and Matsuura, 1977a,b; Matsuura et al., 1976). Cellulose acetate membranes have both polar and nonpolar characteristics. The polar character of the membrane is attributed to the carboxyl and ester groups present in the polymer, and the nonpolar character is due to the backbone carbon chain. Therefore, both the polar solvent (water) and nonpolar solute (hydrocarbon) may have affinity toward the membrane material. The relative affinity of the components present in the solute-solvent system toward the membrane determines the extent of the preferential sorption. The nonpolar or dispersive character of the aromatic hydrocarbon solute can be represented by the “molar attraction constant” (Small, 1953). An increase in the dispersive character of a series of homologous compounds is indicated by an increase in the molar attraction constant. Values of the molar attraction constant plus other data for benzene, toluene, and cumene are presented in Table I. A mathematical analysis of the preferential sorptioncapillary flow model of reverse osmosis has been developed (Kimura and Sourirajan, 1967) for the case of water preferential sorption and serves as a useful starting place for the analysis of solute preferential sorption situations. The Kimura-Sourirajan analysis gives rise to the following basic equations relating the pure water permeability coefficient (A),the molar flux of solute (NA)and of solvent (NB),the pure water mass flux (nJ,the solute transport parameter (Dm/Kd, and the mass transfer coefficient ( k ) on the high-pressure side of the membrane under operating conditions of constant temperature (Sourirajan, 1970b)
(3)
where the symbols are defined in the Nomenclature section.
Ind. Eng. Chem. Process Des. Dev., Vol. 22, No. 4, 1983
627
Table 11. Characterization and Performance of the Cellulose Acetate Membranes and of the Radial Flow Cells" film and cell number 2
1
pure water permeability coefficient, 0.7068 A, x l o 7 , (kmol of HzO)/(mzs kPa) 97.9 NaCl-water separation, % 7.57 total solution mass flux, n T , x l o 3 , kg/m2 s solute transport parameter, ( D A MlK7 )NaC1 1.388 x 10',m/s In C * N a C l -12.555 43.0 mass transfer coefficient, k, X l o 6 ,m/s
0.8635 97.3 9.25
3 1.214 96.0 12.84
2.157
4.166
-12.114 50.9
-11.456 49.9b
4
1.818
5 2.222
6 2.447
91.1 19.06
78.9 23.48
58.5 24.88
12.87
39.05
63.94
-9.218 49.bb
-8.725 25.6'
-10.3!8 51.8
mz ;operating pressure, 6900 kPa; feed concentration, 1 0 000 ppm NaCl; temperature, 25 "C; a Film area, 1.443 X m/s is used. m/s is used. ' A value of k = 25.0 X An average value of k = 49.0 x feed flow rate 400 mL/min.
Because of the assumptions inherent in the KimuraSourirajan analysis, these equations are not directly applicable to the solute preferential sorption case. For instance, the Kimura-Sourirajan analysis assumes that the solution flux is less than the flux obtained with a pure water feed at similar operating conditions solely because of the osmotic pressure effects. However, for solute preferential sorption cases, there is a decrease in flux even for dilute feed solutions. In this case, the flux decrease cannot be attributed to osmotic pressure effects, but is considered to be a result of pore blocking by the solute. Previously (Dickson and Lloyd, 1981; Dickson et al., 1979),it was found that for solute preferential sorption the quantity (1 - nT/np) serves as a useful parameter demonstrating the influence of preferentially sorbed solute on flux. The quantity is taken to be representative of the "extent of pore blocking". The term %/np indicates how much the mass flux (h) of the permeating solution (solute plus solvent) differs from the pure water mass flux (np) due to solute pore blocking. As the difference between nT and np becomes greater, (1- nT/np) increases from zero (no pore blocking) to a limit of 1.0 (totally blocked pore). The boundary layer concentration can be related to the quantity (1- nT/nP)by the following empirical equation (1 - nT/nP) = ZX,Zr
(7)
where 2 and are constants dependent upon several factors including the strength of the sorption forces at the membrane-solution interface and the porous structure of the membrane surface. From the above discussion it is apparent that a quantitative measure of the membrane pore size is needed for the purposes of correlating the separation data to the membrane porosity. A quantity representing the average pore size on the membrane (on a relative basis) has been established (Matsuura et al., 1975). This "pore size parameter" (given the symbol In C*NaCI) is calculated from the solute transport parameter via the following equation (Matsuura et al., 1975) In C*NaCl = In ( D ~ / K (cm/s)) T - 1.37 (8) This quantity is used for membrane specification and is based on the data obtained from the NaCl experiments. Briefly, an increase in the value of In C*NaC1 indicates an increase in the average pore size. Experimental Section In this study cellulose acetate membranes, made by the general Loeb-Sourirajan technique (Pageau and Sourirajan, 1972) and cast from a solution of 17.0 wt % cellulose acetate (Eastman E39&3), 69.2 w t % acetone, 12.35 w t 70 water, and 1.45 wt % magnesium perchlorate were employed. The casting solution was cooled to 0 "C and then cast, with a Gardner Knife, on a glass plate at room tem-
perature (25 "C). An evaporation period of 1 min was allowed before the membrane was gelled by immersing it in ice water (0 "C). After 1h the membrane was removed from the glass plate and cut into circles of the appropriate diameter for mounting in the reverse osmosis cell. Prior to mounting, the membrane was heat treated in water for 10 min. This final step reduced the size of the surface pores, which increases the sodium chloride-water separation for the membrane. The higher the temperature used the greater the reduction in pore size. This method produced a series of six membranes with differing average pore sizes on the membrane surface. After mounting, the membranes were subjected to prepressurization treatment at 12000 kPa for 2 h to stabilize the behavior. The solutes used were all reagent grade (purchased from Fisher Scientific Co.) and were used without further purification. Feed solutions were made with distilled water. Reverse Osmosis Experiments. The apparatus used consists of six radial flow cells connected in series, a high-pressure diaphragm metering pump, a surge tank, a back-pressure regulator, and a pressure gauge. The feed solution was pumped through six cells, and after passing through the regulator it was returned to the feed reservoir. The temperature of the entire system was controlled at 25 f 1 "C. Sodium chloride experiments (at 10000 ppm and 6900 kPa) were repeated periodically to characterize the membranes and to monitor any changes in the membrane behavior over the duration of the studies. The range of feed concentrations in the organic solute studies was from 5 to 260 ppm depending on the limits of solubility and the sensitivity of the chemical analysis. The feed flow rate was kept constant at -400 mL/min and four different operating pressures of 690,1725,3450, and 6900 kPa were used. In the experiments, the volume of solution which permeates through each membrane is small relative to the total volume of feed solution. Thus, average values of the feed concentration and the permeate concentration over the duration of the experiment are used in calculating separation via eq 1. The sodium chloride concentrations were measured with a YSI Conductivity Bridge (Model 31) and the organic solutions were analyzed with an Oceanography International Corporation Total Carbon Analyzer (Model 524C).
Results and Discussion Membrane and Cell Characterization. The six cells and corresponding cellulose acetate membranes were characterized according to the pure water and sodium chloride performance data. These data are presented in Table 11. Actual characterization experiments were repeated at regular intervals in order to monitor the membrane change, and the illustrated data represent the average of 17 tests.
628
Ind. Eng. Chem. Process Des. Dev., Vol. 22, No. 4, 1983
BENZENE
N 0
(II
53.4
ppn
0 116.1 ppn A 124.8 p p m d 221.0 PP*
138.2 P P ~ 4 259.5 p p m A
U
u
0
252.8 p p m
:I 9
-
I . I /
-13
-12
-11
-io
-9
-8
-13
-12
-11
-io .
A -9
-e
~
I
-13
-12
-11
-10
-9
-8
Figure 1. Effect of feed concentration on the reverse osmosis performance for the benzenewater system at four different pressures. Membrane material = cellulose acetate; In C*N.CIobtained at 6900 kPa; feed flow rate = 400 mL/min; temperature = 25 "C.
The pure water permeability coefficient (a measure of the overall membrane porosity) tended to decrease over the period of several experiments. This decrease, which was attributed to membrane compaction, varied from a 10% decrease for the membranes of the largest pore size to a 5% decrease for the membranes of smallest pore size. The rate of membrane compaction during exposure to these hydrocarbon solutions was slightly higher than would normally be observed for salt solution experiments. This accelerated decrease in the value of A may be the result of the detrimental effect that high interfacial concentrations of organics have on the cellulose acetate. The membrane in cell 6 was damaged during the course of this study and was replaced by a similar membrane. The remainder of the experiments were performed with this new membrane. Since these two membranes were similar in performance, for simplicity they are treated as one membrane (called membrane 6). The average separation and the solution mass flux for the aqueous NaCl solutions measured under the indicated conditions are listed in Table 11. The six membrane samples covered the range of sodium chloride separation from 58% to -98% (corresponding to a range of the pore size from -12.5 to -8.7). Since the solute parameter In C*NaC1 transport parameter for sodium chloride, and hence In C*Na~l, remained wentially constant over the experimental time period it can be assumed that the membrane pore size remained constant. Values of the sodium chloride mass transfer coefficient listed in Table I1 are reasonably constant and an average value of k = 49 X 10+ m/s is used for cells 1 to 5. However, cell 6 has a marginally larger flow channel for feed which results in lower feed solution velocities and a lower k value. For cell 6 the lower k value of 25 X lo* m/s is used for this study. Hydrocarbon Solute Reverse Osmosis System. The experimentally determined performance for the separation of benzene and water at four different pressures is illustrated in Figure 1. The data obtained at each operating pressure indicate that as the pore size increases the benzenewater separation decreases, reaches a minimum,
then increases. For each pressure, the separation was shown via an analysis of covariance to have no concentration dependence at a 0.99 confidence level. Comparing separation data, over the range of pressures studied, it is observed that as the operating pressure increases the separation decreases. At each pressure and for each concentration studied, as the pore size increases the extent of pore blocking decreases. The data have been fit to a first-order relationship. A number of points are worthy of note. For any given operating pressure, the slope is an increasing function with respect to feed concentration. This observation is in contrast to the previously stated apparent concentration independence of separation. By way of comparison, if a plot of (1 - nT/nP) vs. In C*NaCIwere made for sodium chloride-water systems in the concentration range discussed here, the result would be a horizontal line at (1nT/nP) equals zero. At higher concentrations of sodium chloride in water, there would be a reduction in the ratio nT/nP due to osmotic pressure effects rather than the blocking of pores by solute molecules. However, the results would not necessarily follow a simple relationship such as that shown in Figure 1. The effect of operating pressure on the extent of pore blocking is not immediately evident from Figure 1 due to the additional influence of concentration. The influence of pressure on extent of pore blocking is discussed below. The data for the single solute systems of toluene-water and cumene-water are illustrated in Figures 2 and 3, respectively. These results are similar to those for the benzene-water system. The shape and direction of the separation curves as a function of the pore size parameter In C*NacI are similar to those for benzene. Separation again decreases with increasing pressure and remains independent of concentration. Similarly, the decrease in extent of pore blocking with both increasing pore size and decreasing concentration are again observed. Since the cumene is much less soluble in water than either toluene or benzene (see Table I), the concentrations of cumene used are much lower. Even when these concentration differences are taken into consideration, differences are
Ind. Eng. Chem. Process Des. Dev., Vol. 22, No. 4, 1983 629
TOLUENE
m N
6900 k P a
17.2
cl 11.2 ppm 0 32.k p p n A 76.7 p p m 0 153.9 p p n
Ln
c 2.
0 18.5 A 35. k 0 85.1
.
+ 97.5
0 -
+J 0'
26. I 0 55.8 [I1
ppn ppn
A 136.7 ppn 4 161.2 P P ~
-
ppm
ppn ppi 211.4 ppn
A 93.3
.
Q
:Q-L. ::%; $
d
L E
d
X 214. I
X
0 Ln
k5.k 0 51.6 [II
ppn ppn ppn ppn ppn ppm
+
0'
a
*
az
&
--------- -Q---
m o
----
---
-- - - ai--.
N m
n
'-13
-12
n
-8
-13
-12
-IO
-11
-9
-8 -13
-12
-11
Lpm
:. L
-9
-10
-11
@
0 6.8
ppn
0 9.9
0 8.0 A 15.9
PP"
0 12.2
ppm
A
16.2 4 25.9
ppm ppm ppm ppm
j
-
-10
7.7 0 10.2 A 16.0 e 26.5 [II
-9
-8 -23
-I2
-10
-11
cl 1.5
ppm ppn ppm ppm
0 8.1 A
.
Q
12.7 24.6
-9
-8
ppn ppm ppn ppn
Ln
-I-,
0
0-
L Z
d
0'
a
uz
"
w: Q
8 4
c-----------------
(no
+."" N m
observed between the three solutewater systems. For any given pressure, concentration, and membrane, the separation observed for the benzene-water system is less than that of the toluene-water system, which is less than the corresponding value for cumene-water. For similar molar concentrations, the extent of pore blocking increases in the order benzene < toluene < cumene, and the slope of the (1 - %/np) vs. In C*Nacl plot increases in the same order. For all three solutes there is a minimum in the separation vs. In C*NacI curves. This minimum appears to occur, within experimental error, at an In C*NaC1 value of -9.75 for all solutes. The rate at which separation decreases and then increases with increasing pore size once again is or-
-----____________
,
---a--
dered benzene < toluene < cumene. By taking the concentration independent average separation obtained for a given solute and membrane, it is possible to plot the data for all membranes and solutes as a function of pressure. These plots for membranes 3 to 6 are similar to each other and only 1,2,3, and 6 are shown in Figure 4. For the three solutes, all membranes exhibited a decrease in separation as pressure is increased. As discussed above, this decrease in separation with increasing pressure is characteristic of solute preferential sorption. It is particularly interesting to note that the dependence of separation on pressure decreases as the pressure increases. From a practical point of view it is desirable to
630
Ind. Eng. Chem. Process Des. Dev., Vol. 22, No. 4, 1983
Figure 6. Correlation of extent of pore blocking, [l - (nT/np)], v d boundary layer concentration of toluene, Xu,for two representative membranes of different pore size: (0) 690 kPa; (0) 1725 kPa; (A) 3450 kPa; ( 0 ) 6900 kPa; (-) membrane 1; (- - -) membrane 6. Figure 4. Effect of operating pressure on the separation of benzene ( O ) , toluene ( O ) , and cumene (A)in single solute aqueous systems for four representative membranes of different pore size. N 3
b 2
x105
Figure 5. Correlation of extent of pore blocking, [l - (nT/np)],and boundary layer concentration of benzene, Xu,for two representative membranes of different pore size: (0) 690 kPa; (0)1725 kPa, (A) 3450 kPa, ( 0 ) 6900 kPa; (-) membrane 1; ( - - - ) membrane 6.
operate at low pressures to increase the separation; however, this decrease in pressure will also result in a decrease in flux. The apparent anomalous behavior of the cumene-water-membrane 1 (smallest average pore size) system will require further investigation before an explanation can be offered. The blocking of the pores on the membrane surface by sorbed solute can be treated in a more quantitative manner than presented above. The appropriate k value for the particular solute of interest was estimated for each cell (Matsuura et ai., 1974) and the effective boundary layer concentration, X A 2 , was calculated for each experiment (Dickson et al., 1979). For two typical membranes, the extent of pore blocking (1 - n ~ / n p )vs. X A 2 is shown in Figures 5 to 7 for three solutes. These plots show that the extent of pore blocking appears to be independent of operating pressure. Similar results were obtained for all membranes. Thus, pressure changes affect the same relative change in the solution flux and the pure water flux. This plot permits a more quantitative treatment of the pore blocking data. As suggested in the Introduction and
Figure 7. Correlation of extent of pore blocking, [ l - (nT/n,)], and boundary layer concentration of cumene, XA2,for two representative membranes of different pore size: (n)690 kPa; (0) 1725 kPa; (A) 3450 kPa; ( 0 ) 6900 kPa; (-) membrane 1; ( - - - ) membrane 6. Table 111. Proportionality Constants from the Correlation of XA and the Extent of Pore Blocking membrane 1 2 3 4 5
6
z benzene
toluene
cumene
2944 2501 2122 1696 1348 935
7538 6535 5525 4527 3592 2183
30895 27566 23120 18721 13403 8072
in previous publications (Dickson and Lloyd, 1981; Dickson et al., 1979) it is possible to relate the extent of pore blocking to the boundary concentration XA2by eq 7. This equation, which has the form of the Freundlich sorption isotherm, is consistent with the sorption phenomena taking place between the solute and the membrane surface. By taking the natural logarithm of both sides of eq 7, a linear equation is produced. Using linear least squares, the coefficients 2 and {were calculated. Since the { values were reasonably close to 1.0, the 2 values were recalculated with { set equal to 1.0. Setting { = 1.0 is equivalent to saying that the (1 - nT/nP) vs. x A 2 relationship can be represented by a straight line through the origin as indicated in Figures 5 to 7. These new 2 values are given for
Ind. Eng. Chem. Process Des. Dev., Vol. 22, No. 4, 1983 631
0 BENZENE 0 TOLUENE A
CUMENE
Inc x N a C 1 Figure.8. Correlation of proportionality constant (2 from eq 7) and pore size parameter In C*NaCI.
all membranes in Table 111. The data in Table I11 demonstrate not only the dependence of 2 on the average pore size but also the dependence of 2 on the nonpolar or dispersive character of the solute. With the limited amount of data available, it is not possible to speculate on the exact nature of the relationship between Z and the dispersive character other than to state that 2 increases with increasing solute dispersive character. The physical interpretation of this trend is that at identical concentrations benzene blocks the pore the least while cumene has the greatest ability to block the pore. There are sufficient data to obtain a functionality between 2 and In C * N ~ CIn~Figure 8, this functionality is shown to be adequately represented by a linear relationship on a semilogarithmic basis. Within experimental error the slopes of these three lines appear to be identical. Interpretation of the Hydrocarbon Solute Data. The trends in Figures 1 to 4 are consistent with the qualitative features of solute preferential sorption as discussed earlier in thispaper. The solution flux is lower than the pure water flux due to pore blocking rather than any significant osmotic pressure effects. This effect is enhanced by either decreasing the pore size or increasing the feed Concentration. Both of these factors lead to a relative increase in the solute content of the pore and thus to a restricted solution transport through the pore. In the highly concentrated region of solution near the membrane surface (referred to here as the interfacial region), the solute-membrane attraction forces decrease with distance from the membrane surface. Thus, solute mobility increases and solute concentration decreases with increasing distance from the membrane surface. In the limit, a point is reached a t which the solute mobility and concentration can be considered to be no longer under the influence of the interfacial attractive forces. When a portion of this interfacial layer is transported through the pores of a membrane, it is the average solute mobility and average solute concentration of the transported portion that determines the flux and separation. If the average pore on the membrane surface is extremely small, the portion of solution transported is that region of high concentration and low mobility. As the average pore size is increased, the portion of the interfacial layer being transported increases to include less concentrated-more mobile solution. Thus, the average concentration in the pore decreases and the average solute mobility increases. Decreasing average concentration of the transported solution indicates increasing separation. Increasing solute mobility of the transported solution indicates decreasing separation. These two factors present opposing influences on separation as a function of pore size. Note that the
solute in the region near the membrane is strongly attracted to the membrane and is relatively immobilized while the water remains relatively mobile. However, the velocity of the solution passing through the pore creates a shear force which strips the sorbed solute molecules from the surface and transports them. The factors discussed above can be used to explain the results observed in the present study. The opposing factors of decreasing concentration and increasing mobility with increasing distance from the surface suggest that separation as a function of pore size may exhibit a local minimum as observed in Figures 1through 3. The initial section of decreasing separation with increasing pore size represents the transport of the lower portions of the interfacial layer. In this region, the mobility of the solute is a stronger function of distance than is the concentration; thus, solute mobility determines the separation. The section of the curve with increasing separation as pore size increases represents the transport of a larger portion of the interfacial layer. In this region, the solute concentration is a stronger function of distance than is the mobility; thus, the average solute concentration in the pore determines the separation. Between these two sections of the curves there is a minimum where the two opposing'factors cancel. For a membrane with a given average porosity, increasing the pressure increases the velocity of the solution through the pore resulting in an increased flux. The increased velocity causes increased solute mobility due to she& of the concentrated interfacial layer; thus, separation decreases with increasing pressure as observed in Figures 1 to 4. The solute-water separation may be positive, negative, or zero depending upon the system and specific operating conditions. When these factors are such that the solute-membrane interaction is relatively strong, the solute is immobilized and the permeating water leads to positive separation. When these factors are such that the solute-membrane interactions are relatively weak, the solute enriched solution in the interfacial region is carried though the pore with less resistance resulting in negative separation. Balancing these two factors produces zero separation. The solution flux is significantly less than the pure water flux because the solute molecules sorbed to the pore wall tend to decrease the effective pore diameter. This phenomenon is known as the pore blocking effect. In comparing the present work with aromatic hydrocarbon-water-cellulose acetate studies reported in the literature (Matsuura and Sourirajan, 1973), a number of points are of interest. The present findings are in agreement with the previously reported results in that for any given system the separation is pressure dependent. However, Matsuura and Sourirajan (1973) also report a pressure dependence of the extent of pore blocking; in the present study no such pressure dependence was found (see Figures 5 to 7). Instead, the extent of pore blocking was found to be concentration dependent (see Figures 1 to 3). The discrepancy can be explained by the fact that Matsuura and Sourirajan (1973) treated a range of feed concentrations (36 to 64 ppm benzene, 19 to 53 ppm cumene) as one concentration. Evidently, a different concentration was used for each pressure study; thus, the concentration effect was incorrectly interpreted as a pressure effect. The present study demonstrates the significant role that concentration plays in determining membrane performance in systems showing strong solute-membrane affinity. Since the extent of pore blocking is primarily a function of the physicochemical interaction of the solute and the membrane, it is more logical that the extent of pore blocking be concentration dependent as reported here
632 Ind. Eng. Chem. Process Des. Dev., Vol. 22, No. 4, 1983
rather than pressure dependent as reported earlier (Matsuura and Sourirajan, 1973). The dependence of the extent of pore blocking on the physicochemical nature of the solute is evident when Figures 5 through 7 are compared. All other parameters being equal, the extent of pore blocking is least with the more mobile solute (benzene) and greatest with the least mobile solute (cumene). In comparing the results in previous sections certain differences in behavior were noted for the three solutes. The order of these differences is always benzene < toluene < cumene. This order is the same as the order of the dispersive character of the solute and therefore the order of solute-membrane affinity. In future papers, this trend will be treated in more detail. Conclusions The ability of reverse osmosis to separate from water the solutes benzene, toluene, and cumene in a single-solute dilute aqueous solution has been demonstrated. The dependence of solute-water separation and solution flux on the operating pressure, feed concentration, solute dispersive or nonpolar character, and membrane pore size were determined. For each system studied, the solute-water separation decreased with increasing pressure and was independent of feed concentration. The separation decreased, went through a minimum, and then increased with increasing pore size. The solution flux increased with increasing pressure, decreasing feed concentration, and increasing pore size. The influence of the solute on flux is readily illustrated when the solution mass flux (9) is combined with the pure water mass flux (np) in the quantity (1 - nT/np), which is taken to represent the extent of pore blocking. The extent of pore blocking was found to increase with both increasing feed and boundary layer concentrations and decreasing pore size while remaining independent of operating pressure. Separation and extent of pore blocking were shown to increase with the nonpolar character of the solute. The nonpolar character of the solute as represented by the molar attraction constant increased in the order benzene < toluene < cumene. The observed results can be qualitatively justified in terms of the preferential sorption-capillary flow mechanism of reverse osmosis. According to this model, for the solutes under consideration, the solute and solvent are both attracted to the membrane material. Within the pore, the flux of the water and solute depends on the mobility of these species which, in turn, is dependent upon the degree of attraction to the pore wall. In the case of aromatic hydrocarbons in water, the membrane-hydrocarbon attraction is stronger than the membrane-water attraction. Thus, in the pore, the water is more mobile than the solute and the permeate has a lower solute concentration than that of the feed concentration; that is, positive separation is achieved. Acknowledgment The authors wish to gratefully acknowledge the support of this project from the Office of Water Research and Technology (grant number 14-34-0001-0522) and the scholarship for J.M.D. from the Natural Sciences and
Engineering Research Council of Canada. Nomenclature A = pure water permeability coefficient, kmol of H20/(m2s kPa) C. = molar density of solution j , kmol/m3 = membrane pore size parameter DAB = diffusivity of solute A in solvent B, m2/s DAM/KT = solute transport parameter, m/s DAM= diffusivity of solute in membrane, m2/s K = partition coefficient 7 = effective membrane thickness, m f = separation k = mass transfer coefficient, m/s M i= molecular weight of species i , kg/kmol m . = concentration of solution j , molality i( = molar flux of species i, kmol/(m2 s) np = pure water mass flux, kg/(m2 s) nT = total solution mass flux, kg/(m2 s) P = operating pressure, kPa gauge X i j = mole fraction of component i in solution j 2 = proportionality constant defined in eq 7 { = exponent defined by eq 7 H A , = osmotic pressure of solution j due to solute A
d*
Subscripts
1 = bulk solution on the high-pressure side of the membrane 2 = concentrated boundary solution on the high-pressure side
of the membrane
3 = permeated product solution on the low-pressure side of
the membrane A = solute B = solvent M = membrane P = pure water T = total solution (solvent plus solute) Registry No. Cumene, 98-82-8; cellulose acetate, 9004-35-7; benzene, 71-43-2; toluene, 108-88-3. Literature Cited Burghoff, H . 4 . ; Lee, K. L.; Pusch, W. J. Appl. Polym. Sci. 1980, 25, 323-347. Dickson, J. M.; Matsuura, T.; Sourirajan, S . Ind. Eng. Chem. Process Des. Dev. 1979, 18, 641-647. Dickson, J. M.; Lloyd, D. R. I n "Synthetic Membranes", Vol. 11, Turbak, A., Ed.; ACS Symposium Serles No. 154, American Chemical Society, Washington, DC, 1981; Chapter 18. Jonsson, G.; Boesen, C. E. Desalination 1975, 17, 145-165. Jonsson, G. Desallnation 1960, 35, 21-38. Kimura, S.; Sourirajan, S.AlChE J. 1987, 13, 497-503. Lonsdale, H. K.; Merten, U.; Riley, R. L. J. Appl. Polym. Sci. 1965, 9 , 1341-1362. Lonsdale, H. K.; Merten, U.; Tagaml, M. J. Appl. Polym, Sci. 1967, 7 1 , 1807-1820. Matsuura, T.; Sourirajan, S. J . Appl. Polym. Sci. 1973, 77, 3663-3708. Matsuura, T.; Bednas, M. E.; Sourirajan, S.J. Appl. Po/ym. Sci. 1974, 18, 567-588. Matsuura, T.; Pageau, L.; Sourirajan, S. J. Appl. Po/ym. Sci. 1975, 19, 179-198. Matsuura, T.; Dickson, J. M.; Sourirajan, S. Ind. Eng. Chem. Process D e s . Dev. 1976, 15, 149-161. Merten, U., Ed. "Desalination by Reverse Osmosis", M.I.T. Press: Cambridge, MA, 1966; pp 15-54. Pageau, L.; Sourirajan, S. J. Appl. Polym. Sci. 1972, 16, 3185-3206. Pusch, W.Ber. Bunsenges. Phys. Chem. 1977, 8 1 , 269-276. Small, P. A. J. Appl. Chem. 1953, 3 , 71-80. Sourirajan, S. "Reverse Osmosis", Academic Press: New York. 1970a; Chapter 1; 1970b; Chapter 3. Sourirajan, S.; Matsuura, T. I n "Reverse Osmosis and Synthetic Membranes", Sowlrajan, S.,Ed.; National Research Council of Canada: Ottawa, 1977a; Chapter 2; 1977b; Chapter 3. Spiegler, K. S . ; Kedem, 0. Desalination 1966, 7 , 31 1-326.
Received for review August 4, 1982 Revised manuscript received January 21, 1983 Accepted February 7, 1983
