Ind. Eng. Chem. Res. 1987,26, 2385-2389
2385
Effect of Polyvinylpyrrolidone Additive on the Performance of Polyethersulfone Ultrafiltration Membraned Lucie Y. Lafrenigret and Frank D. F. Talbot Department of Chemical Engineering, University of Ottawa, Ottawa KIN 6N5, Canada
Takeshi Matsuura* and Srinivasa Sourirajad Division of Chemistry, National Research Council of Canada, Ottawa K 1 A OR6, Canada
The effect of polyvinylpyrrolidone additive on the performance of polyethersulfone ultrafiltration (UF) membranes was studied using various compositions for film casting solutions. The concentration of polyethersulfone polymer was changed from 15 t o 30 w t %, while keeping the weight ratio of polyvinylpyrrolidone t o polyethersulfone in the range 0-2.5. N-Methylpyrrolidone was used for solvent. T h e performance data of U F membranes and the pore size and the pore size distribution on the membrane surface have also been investigated. It was found from viscoelastic and other observations t h a t the interaction between polyethersulfone and polyvinylpyrrolidone is strongest when their weight ratio is unity. T h e structure of the film casting solution under such a strong interaction force causes an increase in the size of the largest pores involved in the pore size distribution and, consequently, increases the permeation rate through the U F membrane. This paper is presented as one of a series of works in which the effect of the structure of the polymer in the membrane casting solution on the performance of the resulting membrane is investigated. In our previous works (Nguyen et al., 1985, 1987a; Matsuura and Sourirajan, 1985) the study was focused on the aromatic polyamide poly(m-phenylene-iso(70)-cotere(30)phthalamide). The effects of polymer molecular weight and of casting solution composition on polymer structure in the casting solution was investigated by using viscoelastic experiments. The results were related to the average pore size and the pore size distribution on the surface of the resulting membranes. The above approach has also been applied to study the effect of the nonsolvent swelling agent on the casting solution structure of aromatic polyamide (Nguyen and Matsuura, 1986; Nguyen et al., 1987b). The objective of this work is to conduct a similar investigation on the structure of the polymer in casting solutions prepared for the formation of ultrafiltration (UF) membranes from polyethersulfone (PES) Victrex material. The latter polymeric material was chosen because it has excellent characteristics for UF membrane applications, which include chemical and thermal stability and mechanical strength. Several features differ between casting solutions prepared from aromatic polyamide and polyethersulfone polymers. In the case of aromatic polyamide, the nonsolvent swelling agents added to the casting solution consist mostly of inorganic electrolytes. These agents form ion-polymer complex, bringing several polymer molecules into one supermolecular polymer aggregate. In the case of polyethersulfone polymer, on the other hand, the additives are polymeric compounds, such as polyvinylpyrrolidone (PVP) (Cabasso et al., 1976), the function of which is expected to be different from that of electrolytic 'Present address: Esso Petroleum Canada, Sarnia, Ontario
N7T 7M1,Canada. t Present address: Industrial Membrane Research Institute, Department of Chemical Engineering, University of Ottawa, Ottawa, Ontario K1N 6N5, Canada. f Issued as NRC No. 28034. 0888-5885/87/2626-2385$01.5Q/O
additives. Therefore, it is the goal of this paper to elucidate the state of PES polymer in the casting solution in the presence of a polymeric additive and to investigate its effect on the surface pore structure and hence the performance of polyethersulfone UF membranes. Though there have been several investigations on the preparation of polyethersulfone UF membranes in the literature (Kai et al., 1985; Tweddle et al., 1983), this work is unique in this respect.
Experimental Section Materials. Polyethersulfone Victrex (200P) supplied by Imperial Chemical Industries was used. The molecular weight of polymer was about 30 000 as determined by gel permeation chromatography. A purified grade N methylpyrrolidone (NMP) was used as solvent. Polyvinylpyrrolidone of molecular weight 10000 was used as nonsolvent swelling agent. Membrane Preparation. The membranes were laboratory prepared following the standard method of phase inversion technique which yields membranes with asymmetric pore structures. Twenty-one solution compositions, each one of which is positioned in Figure 1on a triangular diagram of polymer (P)-solvent (S)-nonsolvent swelling agent (N), were chosen for the study. Membranes so prepared are coded according to the solution composition; i.e., PES-15-1.0 indicates that PES concentration in the casting solution is 15 wt % and PVP/PES weight ratio is 1.0. The polymer solution was cast on a smooth glass plate to the thickness of 0.025 cm. The solution temperature and the temperature of the casting atmosphere were both ambient. After 1min of solvent evaporation in ambient atmosphere, the membranes were immersed in ice-cold water for more than 15 min. The apparatus and the experimental procedure used were the same as those reported earlier (Hsieh et al., 1979). As organic reference solutes for UF experiments, polyethylene glycols (PEG) of molecular weights from 300 to 15000 were used. All experiments were performed a t laboratory temperature, a t the operating pressure of 345 kPag (50 psig), and a t the solution feed flow rate of 2200 mL/min. In each experiment, flux and solute rejection
Published 1987 by the American Chemical Society
2386 Ind. Eng. Chem. Res., Vol. 26, No. 11, 1987 Polymer (PES1
Table I. Interfacial Interaction Force Parameters Pertinent to Some PEG Solutes MW of PEG i030B,m3 lO’OD, m
A
300 9 000 15 000
-42.551 118052 197 238
4.72 28.24 33.65
such pore size distributions, the distribution function of the ith component given as
and a quantity defined as Nonsolvent (PVPI
Solvent INMPI
Figure 1. Casting solution compositions of P (polyethersu1fone)-S (N-methylpyrro1idone)-N (polyvinylpyrrolidone)system represented on a triangular diagram.
V, were determined. The flux data are represented by pure water permeation rate (PWP) and product rate (PR) in grams per hour per given area of film surface (14.5 cm2), and they were corrected to 25 “C using relative viscosity and density data for pure water. The terms “product” and “product rate” refer to membrane permeated solutions. The fraction solute rejection ( f ) was calculated from
f=
solute (ppm) in feed - solute (ppm) in product (1) solute (ppm) in feed
The concentrations of PEG solute in feed and product solutions were determined by using a Beckman total carbon analyzer Model 915B. Intrinsic Viscosity Measurement. The viscosity measurement of the polymer solution was conducted by means of a Cannon-Ubbelohde viscometer; the capillary size was chosen so that kinetic energy corrections were not necessary. The temperature was controlled at 25 f 0.2 OC. The initially charged solution was progressively diluted with more solvent and the flux time measured after each dilution. For each flux time, check determinations were made until three exact repeats were obtained, or the standard deviation was less than 0.2 s. The total polymer concentration, including PES and PVP in the initial solution, ranged from 1.0 to 2.5 g/dL of NMP solvent. Intrinsic viscosities were determined by extrapolation of qsp/c to zero concentration. The plots were linear with correlation coefficients of more than 0.99. Maximum standard deviation of the intrinsic viscosity was 0.005. Only specific viscosity values between 0.2 and 1.0 were considered, since outside these limits the qBP/cversus c relationship becomes nonlinear.
Theoretical Section Method of the Determination of the Pore Size Distribution. The pore size distribution of the membrane is determined by applying transport equations developed on the basis of the surface force-pore flow (SFPF) model (Sourirajan and Matsuura, 1985) to UF separation data of chosen reference solutes and finding the best pore size distribution to optimize the agreement of calculated and experimental separation data. The details of the above procedure are described in the literature. The method is briefly outlined below. The pore size distribution is expressed in terms of one or more Gaussian normal distributions. For describing
hi = ni/nl
(3)
are necessary, where Rb,l, ut, and n,denote the average pore size (radius), standard deviation, and the number of pores which belong to the ith distribution (Chan et al., 1982). We also define that Rb,$ becomes progressively greater as i increases. The interaction force working between an organic solute and the membrane pore wall in the aqueous solution is expressed by interaction force constants defined by the interfacial potential function such as when d ID 4(d) = very large B = --RT when d > D d3
(4)
where d is the distance between the pore wall and the center of the solute molecule, D is a constant associated with the distance of steric repulsion, and B expresses the nature and magnitude of the van der Waals force. Among the above quantities, the interaction force parameters B and D can be determined from the specific surface excess data, rA/cA,b, obtainable from chromatographic retention volume data. In this work, the data for the system PESPEG available in the literature (Sourirajan and Matsuura, 1985) are used for PEG whose molecular weight is from 600 to 6000. For PEG-300, -9000, and -15000, data were generated in this work by the extrapolation of available literature data, and the results are reported in Table I. The parameters associated with the pore size distribution, Rbs, u,, and h,, are determined so that chosen parameters together with the above B and D values optimize the agreement of calculated and experimental separation data of reference solutes.
Results and Discussion
UF Experimental Data. The experimental results for PEG-6000 solute separation and the product rate are illustrated in Figure 2 for different PES concentrations and PVP/PES weight ratios. The following observations have been made from Figure 2. 1. The product rate decreases with increasing PES concentration, except for PVP/PES ratio of 1.0, where a maximum in the above correlation is observed. 2. There is a maximum in solute separation as PES concentration increases. 3. The product rate passes through a maximum when PVP/PES weight ratio is 1.0. The last conclusion is more obvious in Figure 3. Pore Size Distribution. The best fit data of calculated and experimental PEG solute separations are shown in Figure 4 for PES-15-0.2 and PES-20-0.2 membranes as
Ind. Eng. Chem. Res., Vol. 26, No. 11, 1987 2387 PVPlPES WEIGHT RATIO
8
i 80 -
0 I-
2 70-
//
d
,
60W
t3
6
50-
m
3d00
60b0 9dOO 12600 PEG MOLECULAR WEIGHT
15000
Figure 4. Agreement between calculated and experimental separation of PEG solutes with various molecular weights. Operating conditions the same as Figure 2; symbols are experimental values.
: 1 2
"15
v,
20
25
30
PES CONCENTRATION, wt%
Figure 2. Effect of polymer concentration in the casting solution psig); on membrane performance: operating pressure, 345 Wag (4% solute, PEG-6000; concentration in feed, 200 ppm; effective f i i area, 14.5 cm2; feed flow rate, 2200 mL/min.
-lo!
t
I
I
I
I
2 1 PVPIPES WEIGHT RATIO
I
I
0
Figure 5. Correlation between a,/Rb,Z and PVP/PES weight ratio.
I-
3
2
400
0.5
1 1.5 2 2.5 PVP/PES WEIGHT RATIO
3
Figure 3. Effect of PVP/PES weight ratio of the casting solution on membrane performance. Operating conditions the same as Figure n z.
typical examples. The agreement of experimental and calculated data is very good. On the basis of the best fit data, the parameters characterizing the pore size distribution, i.e., Rb,2, ul,u2, and h2,have been generated, and the results are reported in Table 11. The average deviation is also included in the table. Bimodal distributions were considered since a single distribution did not fit the data satisfactorily (Chan et al., 1982). Apparently there is no definitive correlation between the casting solution composition and the above parameters. Figure 5 shows, however, that the u2/Rb,2ratio passes through a maximum when it is correlated with the PVP/PES ratio. The position of the maximum ratio is either 0.5 (for PES concentration of 15 wt %) or 1.0 (for PES concentration of 20 and 25 wt % ) and coincides with the position of the maximum [PR] in Figure 3.
Table 11. Parameters Characterizing Bimodal Pore Size Distributions Obtained from U F Experimentsn parameters characterizing pore size distribution PVP/PES IO'ORb,l, IO''U1, 1O1O8b,2, IO'OUz, av wt ratio m m m m h, dev, % PES Concentration, 15 wt % 0.0 12.4 4.1 51.9 0.093 0.0084 5.6 1.8 55.0 0.22 0.0020 3.4 0.2 12.8 3.4 48.4 4.4 0.0006 5.0 0.5 12.3 48.0 0.96 0.0010 7.3 13.0 4.2 1.0 0.27 0.0047 4.6 4.8 33.2 2.0 14.3 7.7 0.003 0.0194 2.7 16.8 7.0 30.0 0.0 0.2 0.5 1.0 1.5 2.0
PES Concentration, 20 wt % 43.5 0.0044 0.0115 11.8 5.2 47.0 0.12 0.0020 11.1 1.6 3.0 44.3 0.98 0.0030 12.4 0.0154 11.5 1.4 33.3 1.1 0.94 0.0031 2.7 45.3 13.2 46.3 0.0046 0.0328 9.4 4.2
6.2 1.5 9.1 4.4 6.3 11.5
0.2 1.0 1.4
PES 12.4 12.9 5.05
6.6 7.6 10.9
Concentration, 25 w t % 0.078 0.0050 4.8 30.0 0.264 0.0071 5.8 32.7 0.0043 0.0278 2.0 42.8
a Average deviation = [xy-:l(fi,exptl - f ~ , c ~ c ~ ) 2 1 ' /where 2 / ~ ~ ni l = number of reference solutes.
2388 Ind. Eng. Chem. Res., Vol. 26, No. 11, 1987
- 0 1
0
02 0 4 06 08 WEIGHT FRACTION, PESIiPES'PVPI
10
Figure 6. Intrinsic viscosity as a function of weight fraction PES/(PES + PVP) in the casting solution. Intrinsic viscosity data determined in N-methylpyrrolidone solvent at 25 "C.
With an increase in the C T ~ / ratio, R ~ , ~the distribution extends to larger values in the pore size. Though the number of these pores (of larger pore size) is small, their contribution to the volumetric flow rate through the membrane may be significant due to the fourth-power effect of the pore size according to the Poiseuille law. Hence, it seems reasonable that both the c 2 / R b , 2 ratio and [PR] pass through a maximum at a nearly identical PVP/PES ratio. Viscoelastic Behavior of the Polymer Solution. The intrinsic viscosity data are shown in Figure 6 as a function of PES weight fraction in PES-PVP mixture. There is an inflection point in the PES weight fraction range of 0.5-0.7. If there is no interaction between PES and PVP, the intrinsic viscosity has to be linear with respect to the PES weight fraction. The anomality indicates interaction between PES and PVP polymers. Interestingly, the above range of PES weight fractions in PES-PVP mixture corresponds to PVP/PES weight ratio of 0.4-1.0, which, again, nearly coincides with the position where the maximum in [PR] occurs. Furthermore, the ratio of the number of repeat units of PVP to that of PES can be computed from the PVP/PES weight ratio, and it has been found that 0.9-2.1 of the former ratio corresponds to 0.4-1.0 of the latter ratio. This implies that the interaction between PVP and PES polymers is the strongest when from one to two O=C-N< functional groups in the PVP polymer interacts with each O=S=O group which is apparently the most active polar functional group in the PES polymer. The interaction might as well be of doner/acceptor nature between O=C-N< and an aromatic ring. The strongest interaction between PVP and PES polymers a t a PVP/PES weight ratio of unity was also evidenced while the PES membranes were prepared in the laboratory. Usually in the gelation process, water in the gelation bath turns white because of the release of some low molecular weight PES from the film surface (Wumans and Smolders, 1983). The degree of coloration of the gelation media, however, depends on the PVP/PES weight ratio. The coloration was least, which means the gelation media remained practically clear, when the above ratio was equal to one, particularly with respect to casting solutions containing 20 w t '30PES polymer. This indicates that the strongest interaction between two polymers at a PVP/PES ratio of unity inhibits the release of PES into the gelation media. As a result of the above observation, a different view to the roll of PVP as nonsolvent swelling agent has emerged. Polyvinylpyrrolidone may be entrapped in the PES network and form an integral part of the polymeric structure, providing the polymer surface with a more hydrophilic nature than the surface of PES alone, which in turn causes
a higher water flux through the membrane. The PVP/ PES weight ratio where the strongest interaction takes place corresponds precisely to the point where the maximum water flux occurs. This view, however, has been denied by a chemical analysis of the membrane. The analysis has revealed that PVP/PES weight ratio in the actual membrane sheet is less than 0.04 even when the above ratio is from 0.4 to 1.0 in the casting solution. This analytical result indicates that PVP is leached out into water almost completely while the membrane is stored or pressure-treated. Therefore, it is concluded that the primary effect of PVP in the casting solution is on the structure of the casting solution and, as a consequence, on the pore size and the pore size distribution of the membrane.
Conclusion The following has been observed while studying the formation of polyethersulfone UF membranes. 1. The highest product permeation rate is obtained at a PVP/PES weight ratio of unity when the PES concentration in the casting solution is 15 30 wt % . Particularly, when PES concentration is 20 wt % , a permeation rate of 280 g/h (membrane surface area, 14.5 cm2)at a operating pressure of 345 kPag (=50 psig) was achieved. The rejection of PEG-6000 was 87 'YO. 2. A wider distribution of the pore size extending into greater values in the pore size is obtainable at the above PVP/PES ratio. 3. An anormality in the intrinsic viscosity and PVP/ PES weight ratio correlation occurs when the latter ratio is close to unity. From the above observation, it has been concluded that the interaction between PES and PVP is strongest when the PVP/PES weight ratio is unity. The solution structure formed under such strong interaction forces causes an increase in the sue of the largest pores involved in the pore size distribution and consequently increases the permeation rate.
-
Acknowledgment We thank T. A. Tweddle for his technical assistance and 0. Kutowy, T. D. Nguyen, and A. Y. Tremblay for enlightening discussions on the subject. We are also indebted to Dr. S. Coulombe for the determination of the molecular weight of polyethersulfone polymer.
Nomenclature B = constant characterizing the van der Waals attraction force, m3 D = constant characterizing the steric repulsion at the interface, m d = distance between polymer material surface and the center of solute molecule, m f = fraction solute rejection based on the feed concentration hi = ni/nl ni = number of pores belonging to the ith distribution [PR] = membrane permeated product rate for given area of membrane surface, kg/h [PWP] = pure water permeation rate for given area of membrane surface, kg/h Rb = membrane pore radius, m Rb = ayerage of Rb,m Rb,i = R bof the ith distribution, m R = gas constant T = absolute temperature Yi(&,) = pore size distribution function of the ith normal distribution, l / m
Ind. Eng. Chem. Res. 1987,26, 2389-2393 Greek Symbol ci = standard deviation of the ith normal distribution, m
Registry No. PEG-6000, 25322-68-3; victrex (200P), 2566742-9; poly(vinylpyrrolidone),9003-39-8.
Literature Cited Cabasso, I.; Klein, E.; Smith, J. K. J . Appl. Polym. Sci. 1976, 20, 2377. Chan. K.: Matsuura, T.: Souriraian, S. Ind. Eng. Chem. Prod. Res. Dev. 1982, 21, 605. Hsieh. F.-H.: Matsuura. T.: Souriraian. S. Znd. Enz. Chem. Process Des. Dev.’1979, 18, 414.‘ Kai, M.; Ishii, K.; Tsugaya, H.; Miyano, T. In Reverse Osmosis and Ultrafiltration; Sourirajan, S., Matsuura, T., Eds.; ACS Symposium Series 281; American Chemical Society: Washington, DC, 1985; pp 21-33. Matsuura, T.; Sourirajan, S. In Reverse Osmosis and Ultrafiltration; Sourirajan, S., Matsuura, T., Eds.; ACS Symposium Series 281; American Chemical Society: Washington, DC, 1985; pp 1-19. “
I
-
2389
Nguyen, T. D.; Chan, K.; Matsuura, T.; Sourirajan, S. Ind. Eng. Chem. Prod. Res. Dev. 1985,24,655. Nguyen, T. D.; Matauura, T. In Proceedings of International Membrane Conference;Malaiyandi, M., Talbot, F. D. F., Kutowy, o., Eds.; National Research Council of Canada: Ottawa, 1986; pp 99-114. Nguyen, T. D.; Matsuura, T.; Sourirajan, S. Chem. Eng. Commun. 1987a,in press. Nguyen, T. D.; Matsuura, T.; Sourirajan, S. Chem. Eng. Commun. 1987b,in press. Sourirajan, S.; Matsuura, T. Reverse Osmosis/ Ultrafiltration Process Fundamentals; National Research Council of Canada: Ottawa, 1985; Chapter 4. Tweddle, T. A.; Kutowy, 0.;Thayer, W. L.; Sourirajan, S. Znd. Eng. Chem. Prod. Res. Dev. 1983, 22, 320. Wumans, J. G.; Smolders, C. A. Eur. Polym. J . 1983, 19, 1143.
Received for review April 18, 1987 Revised manuscript received June 14, 1987 Accepted July 27, 1987
Performance Test of Absorption Air-Cooling Unit Using NH4Br-NH3 and NH41-NH3 Systems Hideki Yamamoto, Seiji Sanga, and Junji Tokunaga* Department of Chemical Engineering, Faculty of Engineering, Kansai University, Suita, Osaka 564, Japan
An absorption air-cooling apparatus using NH4Br-NH3 and NH41-NH3 systems was designed and tested in order to most effectively utilize low energy (i.e., solar energy and hot drain). This apparatus used liquid ammonia as the refrigerant and ammonium halides (NH4Bror NH41)as the absorbent, and this absorption refrigerator permitted air cooling and refrigeration. Furthermore, a rectifying tower for recovery of refrigerant from ammonia solution was not needed. Coefficients of performance (C.O.P.) of the apparatus using these two ammonia solution systems were determined for the same operating conditions, and these values are compared and discussed. The effect of Tg, T,,and N on C.O.P. was investigated under various conditions. These values were also compared with the ones calculated from the enthalpy-concentration charts for both the NH4Br-NH3 and the NH41-NH3 systems. In a recent analysis of advanced thermal energy storage systems for heating or cooling utilizing solar energy, the process of using chemical reaction of an anhydrous salt and ammonia is proposed and discussed for its practicality (Fujiwara and Sato, 1985; Jeager and Fox, 1981). The reaction products from anhydrous salt and ammonia are referred to as ammoniated salts or amine complexes whose state is either a solid or liquid. If the liquid phase is used as a working system, the rate of absorption and heat transfer is activated by agitator. The absorption cooling model proposed here consists of a generator (or absorber), a condenser, and an evaporator, and the cycles involving the generating-condensingprocess and the evaporating-absorbing one are carried out alternately. Liquid-phase systems of NH4Br-NH3 and NHJ-NH, systems (Toyoda et al., 1983) were used, where ammonium halide was the absorbent and liquid ammonia was the refrigerant. In these systems, air cooling is possible and the rectifying tower for recovery of the refrigerant from solution is not needed. This paper is concerned with the performance test of the absorption air-cooling unit using liquid phase of NH,Br-NH, and NH41-NH3 systems.
Absorption Air-Cooling Unit Used Absorption Air-Cooling Unit by Two Batch Processes. As solar heat and hot drain are low and intermittent energies, these heat sources are not suitable for an ordinary type of cooling system. Then, corresponding
to the properties of these low-energy sources, the cooling unit was designed for the intermittent operation. This absorption air-cooling unit shown in Figure 1 was driven by the alternate operation of two batch processes. One is the generating-condensing process (I) shown in Figure 2. In this process, the strong ammonia solution (where liquid ammonia is rich) in the generator heated by hot water and ammonia gas of about 100% purity is generated under high pressure. Then, ammonia gas moved to the condenser is condensed by cooling water, and liquid ammonia is stored in the condenser in this process. Finally, the pressure in the generator approaches the saturated vapor pressure of ammonia solution at the temperature of heating water. The other process is the evaporating-absorbing process (11)shown in Figure 3. The liquid ammonia in the condenser is reduced through the expansion valve, and the vaporization of ammonia occurs in the evaporator. Then, the refrigerant evaporated under low temperature and low pressure flows to the absorber and dissolves in the weak ammonia solution in the absorber under constant pressure (0.49 MPa). These two processes are repeated alternately.
Evaluation for Coefficient of Performance of This Unit The ratio of QR to QH is very important for practicality of these units. In order to compare the efficiency of these systems, coefficient of performance, E , was defined as the ratio of QR to QH, that is, 6
= (Q R /Q H )~ O O
0888-5885/87/2626-2389~01.~0~00 1987 American Chemical Society
(1)