Experimental Study of Water and Salt Fluxes through Reverse

Mar 26, 2005 - It was found that, for both DS 5 and DDS 930 membranes, the salt flux data also followed the power relation with (rjc0). The transport ...
150 downloads 7 Views 172KB Size
Download PDF
Environ. Sci. Technol. 2005, 39, 3382-3387

Experimental Study of Water and Salt Fluxes through Reverse Osmosis Membranes WENWEN ZHOU AND LIANFA SONG* Centre for Water Research, Department of Civil Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260

Water flux and salt rejection rate, which are the two most important parameters in evaluating the performance of a reverse osmosis membrane process, are desirable to be directly related to the membrane properties and operating conditions. However, the membrane transport theories in their general forms are unable to describe the membrane performance satisfactorily. In this study, water and salt fluxes through reverse osmosis membranes were carefully examined with a cross-flow filtration cell under various operating conditions. Experimental results showed that a notable permeate flux was detected when the driving pressure was smaller than the feed osmotic pressure. Water flux increased with the driving pressure nonlinearly before approaching a linear relation with the pressure. In addition, salt transport was highly dependent on both operating pressure and feed salt concentration. A power relationship between salt flux and concentration was correlated well with the experimental data. The equations for water and salt fluxes obtained from this work would provide a facile and accurate means for predicting the membrane performance in design and optimization of reverse osmosis processes.

Introduction Membrane processes have emerged recently as one of the most promising technologies for advanced water reclamation and wastewater treatment (1, 2). As one of the widely used membrane processes, reverse osmosis (RO) is proven to be a good barrier for inorganic ions (3, 4) as well as the emerging organic pollutants, such as disinfection byproducts (DBPs) and endocrine-disrupting chemicals (EDCs) (2, 5). The performance of a membrane as a permselective barrier is determined by two parameters, namely, water flux and salt rejection rate (6-11). Water flux is defined as the volume of permeate produced per unit membrane area per unit time; while salt rejection rate is related to the ratio of salt concentration in permeate to that in the feed. Process economics requires a RO system to have a high permeate flux and a low permeate salt concentration. Therefore, to understand water and salt transports and to quantify water and salt fluxes are the most important concerns in RO processes. The driving force for water transport across the membrane is the pressure difference between the driving pressure and the osmotic pressure, while the driving force for salt passage is the concentration difference between the feed and the permeate sides (2-4, 6-11). Both water flux and salt rejection are also highly dependent on membrane properties, solution chemistry, and operating conditions (12). In the literature, * Corresponding author phone: +65-6874-8796; fax: +65-68742890; e-mail: [email protected]. 3382

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 9, 2005

a linear relationship was commonly agreed to exist between water flux and driving pressure (4, 13-16). The nonlinear behavior of water flux observed experimentally, especially for high salt concentration, was considered abnormal and unable to be explained within the existing theoretical frame (4, 17). Song (18) pointed out that such nonlinearity could be theoretically explained with an alternative expression of the driving force. On the other hand, salt flux was usually linearly related to its driving forces (i.e., concentration differences) (7, 13, 19). A power relationship, however, was also reported to exist between salt flux and concentration (14). It was widely noted that salt rejection generally increased with pressure but decreased with feed salt concentration nonlinearly (4, 20, 21). However, current transport theories and models, although correlating with experimental results well, are inconvenient to use to predict salt rejection because of the dependence of main equation coefficients on feed salt concentration (4, 17). The objective of this study was to investigate the water and salt transport behaviors from the experimental point of view and to propose more accurate and practical equations for water and salt fluxes through RO membranes. The laboratory-scale cross-flow membrane filtration tests were conducted to study the membrane performance under different operating conditions. On the basis of the experimental data, transport equations for water and salt fluxes were proposed, which can be easily used for design and optimization purposes. Excellent results were obtained when the published experimental data in the literatures were correlated with the proposed equations.

Materials and Methods Experimental Setup. The stainless steel Sepa CF membrane unit (Osmonics, Minnetonka, MN) was used in the experiments. Feed stream was pumped to the membrane unit from the feed tank, in which a heater/chiller was used to maintain the feedwater temperature at a constant room temperature. Feed inflow rate was adjusted by a bypass valve. The combination of the bypass valve and the pressure gauge on the concentrate outlet allowed a fine control over a wide range of driving pressures and cross-flow velocities within the membrane channel, which has a cross-sectional flow area of 1.64 × 10-4 m2. The flux and conductivity of the permeate were continuously measured by a digital flow meter (Signet, El Monte, CA) and a conductivity meter (WTW, Weilheim, Germany), respectively, and recorded in real time by a personal computer. Membrane Specifications. The membrane used in this study was a commercial flat-sheet RO membrane, CE membrane, provided by Osmonics Inc. (Minnetonka, MN). It was made of cellulose acetate and had a specified salt rejection up to 97%. The membrane was cut to dimensions of 19.1 cm × 14.0 cm before installation. Prior to usage, the membrane was equilibrated with distilled water for at least 24 h. Subsequently, the membrane was subjected to daily mechanical washing and stored in the 5 °C refrigerator when not in use during the course of the experiments. Experimental Procedure. To avoid the effect of membrane compaction on flux, the test membrane was equilibrated by filtering DI water first at constant pressure till the invariant flux was obtained. Six salt solutions were then prepared by adding sodium chloride (NaCl) into DI water at concentrations of 100, 500, 1000, 5000, 10 000, and 20 000 ppm, respectively. 10.1021/es0403561 CCC: $30.25

 2005 American Chemical Society Published on Web 03/26/2005

FIGURE 1. Dependence of water flux on operating pressures under different feed salt concentrations through a reverse osmosis membrane. Nonlinear behaviors were observed at high salt concentrations. Each solution was fed to the experimental setup at a fixed inflow rate of 2.8 L/min and at different pressures with initial pressure of 800 psi (∼5520 kPa). The total permeate and concentrate flow rates were regularly checked to ensure that the sum was equal to that of the feed inflow rate. Salt concentration in the permeate was measured using a conductivity meter after the membrane system had continuously run for about 2.0 h to achieve a stable performance. The whole procedure was repeated for lower pressures of 700, 600, 500, 400, 300, 200, 100, and 50 psi (∼4830, 4140, 3450, 2760, 2070, 1380, 690, and 345 kPa). Salt rejection at each combination of feed salt concentration and operating pressure was then calculated from the initial feed salt concentration and measured permeate salt concentration.

FIGURE 2. Effect of operating pressure on salt rejection of a reverse osmosis membrane at different feed salt concentrations. Salt rejection increased with pressure for all concentrations. high. The salt rejection for high feed salt concentration showed a low salt rejection at low pressure but increased at a higher rate with pressure. According to the popular solution-diffusion model, the salt flux was commonly expressed as a linear function of concentration difference between the feed and the permeate (4, 7, 13, 19):

Js ) k(c0 - cp)

where Js is the salt flux; c0 and cp are the feed and permeate salt concentrations, respectively; and k is the salt permeability constant. The permeate salt concentration is determined by

cp )

Results and Discussions Transport Behaviors under Different Operating Conditions. The dependence of water flux on operating pressure at different feed salt concentrations is plotted in Figure 1. It was found that there was a strong dependence of water flux on feed concentration. Water flux decreased with increasing feed salt concentration, which was well explained with the classic membrane transport theories because the elevated osmotic pressure reduced the net driving pressure. Figure 1 showed that the water flux increased linearly with the operating pressure only for deionized water and that the nonlinearity of the water flux increased with feed salt concentrations, although such nonlinearity was not very obvious for low feed salt concentrations (e.g., 100, 500, or 1000 ppm). There are basically two distinguished pressure ranges separated by the feed osmotic pressure (∆π0). Water flux increases linearly with driving pressure when the pressure is higher than the feed osmotic pressure. It is commonly assumed that there is no permeate production when the driving pressure is lower than the feed osmotic pressure. However, it was found in this study that there was a notable water flux and that the flux increased slowly and nonlinearly with the operating pressure. This behavior of water flux has been observed before as reported in the literature, but no satisfactory explanation was provided. The effect of operating pressure on salt rejection was shown in Figure 2. The general trend in Figure 2 for all feed salt concentrations indicated that salt rejection increased with increasing operating pressure. For a given salt concentration, salt rejection increased faster when pressure was low, and the increasing rate decreased as the pressure became

(1)

Js Jv

(2)

where Jv (.Js) is the water flux. With eqs 1 and 2, a commonly used parameter for salt transport, the salt rejection, can be derived as follows:

rj ) 1 -

or

( )

cp Js k(c0 - cp) cp k )1)1)1- 1) c0 Jvc0 Jvc0 Jv c0 k 1 - rj Jv 1 1 )k +1 rj Jv

(3)

Therefore, if the linear relationship as stated in eq 1 was true, the plot of 1/rj vs 1/Jv should be a straight line. However, the experimental salt rejection data for CE membrane produced a curve rather than a straight line as plotted in Figure 3 and low R 2 values from regression (curve-fitting) as shown in Table 1. The experimental data demonstrated that both water and salt fluxes through the CE membrane might not be adequately described with the linear relationships of the corresponding driving forces. More appropriate equations for water and salt fluxes will be proposed and discussed in the following sections of this paper. Water Flux. In RO process, one popular equation for process design and system operation is based on irreversible thermodynamics, from which the water flux is calculated by the following equation:

Jv ) A(∆p - σ∆π0) VOL. 39, NO. 9, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

(4) 9

3383

FIGURE 3. Nonlinear dependence of the experimental salt rejection data on the driving force (concentration difference).

FIGURE 4. Correlation of permeate flux with the net driving force determined with eq 6 at different feed concentrations. The inclusion of salt rejection in the water flux equation results a better linearity.

where A is the water permeability constant of the membrane; ∆π0 is the feed osmotic pressure, which is a function of feed salt concentration c0; and σ (0 e σ e 1) is the reflection coefficient. The reflection coefficient is a vaguely defined parameter of RO membranes and usually assumed a constant value (6, 7, 9, 10). It is commonly regarded that no permeate would be produced when the net driving pressure (∆p - σ∆π0) is negative. Since the value of σ is very close to unity for many RO membranes, the condition for nil water flux becomes that the driving pressure is smaller than the feed osmotic pressure. The experimental results as discussed before however indicate that the eq 4 may cause confusion and be misleading in describing the water flux under different driving pressure, especially with high feed salt concentration. A better expression for water flux can be derived starting with the net driving force(∆p - ∆π) for water flux, with the osmotic pressure difference being calculated by

20 000 ppm in Figure 4. The osmotic coefficient was determined by the van’t Hoff equation in this study. It could be seen that, for all the concentrations being tested, the permeate flux was linearly dependent on its net driving force. The permeability coefficient value of water permeability constant A in eq 6 obtained by linear regression was 3.93 × 10-9 m s-1 kPa-1. Unlike the transport coefficients in KedemKatchalsky or solution-diffusion models (6, 7, 9, 10), which are highly dependent on concentration, the permeability coefficient A is independent of operating conditions, such as pressure and concentration. This feature makes eq 6 more desirable or suitable for calculating water flux through RO membranes. Since membrane properties determine the value of the permeability coefficient (A), eq 6 can be used in characterization of membrane water transport features for RO processes. For instance, for cellulose acetate CE membranes used in this study, the permeate flux increases linearly with its net driving force, in which the actual variation in salt rejection is considered in the calculation of osmotic pressure. The transport coefficient A of 3.93 × 10-9 m s-1 kPa-1 can be used for the CE membranes in the application of brackish or seawater desalination with the NaCl concentration up to 20 000 ppm. Salt Flux or Salt Rejection. From the previous discussion, it was found that the salt flux was not linearly related to the concentration difference across the membrane. We speculated that the salt flux might be proportional to the concentration difference to a power greater than one (R > 1), which is reflected in the following equation with

∆π ) π0 - πp ) fos(c0 - cp) ) rj∆π0

Js ) B(c0 - cp)R ) Bc0RrjR

TABLE 1. Summary of Regression Coefficient k and R 2 Values with Eqation 3 feed concn (ppm)

k (m/s)

R 2 value

100 500 1000 5000 10 000 20 000

1.24 × 10-7 1.90 × 10-7 2.25 × 10-7 1.26 × 10-7 1.07 × 10-7 5.56 × 10-8

0.8346 0.7935 0.9235 0.8241 0.8720 0.8578

(5)

(7)

where fos is the osmotic pressure coefficient. Then water flux can be linearly related to the net driving force (pressure) with the following equation:

where B is a correlation coefficient. With eqs 2 and 7, the salt rejection can be derived as

Jv ) A(∆p - rj∆π0)

B rj ) 1 - c0R - 1rjR Jv

(6)

Although eq 6 is similar to eq 4 in form, the advantage of the former over the latter is obvious. The main reason is that the salt rejection in eq 6 is well-defined and easily measurable. The well-known dependence of salt rejection on driving pressure can provide a natural explanation to the nonlinearity of water flux with the pressure. To examine the validity of eq 6 for water flux, permeate flux (Jv) was plotted versus its net driving force (∆p - rj∆π0) under different feed concentrations ranging from 100 to 3384

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 9, 2005

(8)

Such an equation is not convenient to describe the dependence of salt rejection on feed concentration, and the transport coefficients B and R cannot be easily obtained from it. Rearranging eq 8, it gives

Jv(1 - rj)c0 ) B(c0rj)R

(9)

The coefficients B and R in eq 9 can be easily evaluated from

FIGURE 6. Correlation of the whole set of experimental salt rejection data under all pressures with eq 9. The values of salt transport coefficients B and r for CE membrane are determined from plotting.

FIGURE 5. Regression of experimental salt rejection data at different salt concentrations with eq 9 under constant pressures. Good correlation was obtained for each pressure.

TABLE 2. Summary of Regression Coefficients B, r, and R 2 Values with Equation 9 operating pressure (psi)

B

r

R 2 value

100 200 300 400 500 600 700 800

1.49 × 10-7 2.39 × 10-7 2.85 × 10-7 2.46 × 10-7 2.10 × 10-7 1.76 × 10-7 2.20 × 10-7 2.25 × 10-7

1.02 1.03 1.02 1.06 1.09 1.11 1.09 1.10

0.9714 0.9808 0.9952 0.9964 0.9976 0.9976 0.9917 0.9949

a parameter estimation method, such as the corresponding linear regression of {log[Jv(1 - rj)c0]} with{log(rjc0)}. The plot of {log[Jv(1 - rj)c0]} versus {log(rjc0)} under different operating pressures was shown in Figure 5. The parameter B and R values and R 2 values for salt transport were summarized in Table 2. The high R 2 values indicated that the proposed power law for salt transport correlates the experimental data very well. For different operating pressures,the value of transport coefficient B basically kept in a

FIGURE 7. Water transport equation agrees well with experimental data from Pusch’s experiments (17) with CA-90 and CA-85 membranes. narrow range of 2.10 × 10-7 to 2.50 × 10-7 with a few

TABLE 3. Summary of Conditions for the Cited Experiments membrane name material type tested salt testing cell refs

DS 5 polysulfone- polyamide RO NaCl cross-flow 22

DDS 930 cellulose acetate RO NaCl cross-flow 9

CA-90 cellulose acetate RO NaCl cross-flow 17

CA-85 cellulose acetate RO NaCl cross-flow 17

VOL. 39, NO. 9, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3385

FIGURE 8. Salt transport equation agrees well with experimental data from Pusch’s experiments (17) with CA-90 and CA-85 membranes. exceptions. It could also be found from Table 2 that the value of R fell in the range of 1.02-1.11 and that the value of R showed an increasing trend with salt concentration. It indicated that the nonlinearity of the salt flux increased with salt concentration. Since operating condition-independent transport coefficients are more desirable in membrane transport for design and optimization purpose, the data in Figure 5 for different pressures were plotted altogether in Figure 6. It could be roughly estimated that the B and R values independent of pressure and feed salt concentration for the CE membrane were around 2.10 × 10-7 and 1.07, respectively. Correlation of the Reported Data in the Literature. Membrane transport was experimentally investigated in the past and reported in the literature. Some of the experimental data (summarized in Table 3) were used to further evaluate the validity of the two equations (6 and 7) proposed in this study. Figure 7 shows the relationship between water flux from Pusch’s classic experimental data for (a) CA-90 and (b) CA-85 membranes. It was found from Figure 7 that the experimental data fit the proposed water flux eq 6 extremely well. The water permeability constant A was dependent only on the membrane properties. For the different concentration solution tested, the CA-90 membrane had an A value of 1.64 × 10-9 m s-1 kPa-1, which was found to be lower than that of the CA-85 membranes (i.e., 3.22 × 10-9 m s-1 kPa-1). Both the CA-90 and CA-85 membranes had a lower water permeability coefficient than the CE membrane used in this study. Similarly, salt rejection data from Pusch’s experiments (17) was examined with the proposed equation and shown in Figure 8. It was noted that the proposed power relationship between the salt flux and concentration difference correlated the literature data satisfactorily. The value of salt transport coefficient R was about 1.45 and 1.51 for CA-90 and CA-85 3386

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 9, 2005

FIGURE 9. Correlation of salt rejections from experiments with (a) DS 5 membrane (from Schirg and Widmer; 22 and (b) DDS 930 membrane (from Jonsson; 9). membranes, respectively, which was much higher than that for CE membrane in this study. More experimental data from Schirg and Widmer (22) and Jonsson (9) are plotted in Figure 9, panels a and b, respectively. It was found that, for both DS 5 and DDS 930 membranes, the salt flux data also followed the power relation with (rjc0). The transport coefficient (R) was found to be 1.55 and 1.12 for DS 5 and DDS 930 membranes, respectively. It indicated that the salt transport characteristics of the CE membrane were close to that of the DDS 930 membrane, while both CA membranes were close to the DS 5 membrane. Finally, although the exact mechanism to determine the value of the coefficient R of a membrane is yet unknown, the parameter can be a useful indicator of the nonlinearity of salt transport properties of RO membranes.

Nomenclature A

water permeability constant (m s-1 kPa-1)

B

coefficient for salt transport introduced in eq 7

c0

feed salt concentration (mg/L)

cp

permeate salt concentration (mg/L)

fos

osmotic pressure coefficient (kPa mg-1 L-1)

Js

solute flux (g m-2 s-1)

Jv

water flux (m/s)

k

salt permeability constant (m/s)

rj

salt rejection rate

∆p

driving pressure across the membrane (kPa)

∆π

osmotic pressure across the membrane (kPa)

∆π0

osmotic pressure of the feed (kPa)

R

transport parameter in equation for solute flux

σ

reflection coefficient

Literature Cited (1) Jacangelo, J. G.; Chellam, S.; Trussell, R. R. The membrane treatment. Civ. Eng. 1998, 68, 42-45. (2) Taylor, J. S.; Jacobs, E. P. Reverse osmosis and nanofiltration. In Water Treatment: Membrane Processes; American Water Works Association Research Foundation; McGraw-Hill: New York, 1996. (3) Pusch, W. Determination of transport parameters of synthetic membrane by hyperfiltration experiment. Part I: Derivation of transport relationship from the linear relations of thermodynamics of irreversible processes. Ber. Bunsen-Ges. Phys. Chem. 1977, 81 (3), 269-276. (4) Soltanieh, M.; Gill, W. N. Review of reverse osmosis membranes and transport models. Chem. Eng. Commun. 1981, 12, 279363. (5) Nghiem, L. D.; Schafer, A. I.; Elimelech, M. Removal of natural hormones by nanofiltration membranes: Measurement, modeling, and mechanisms. Environ. Sci. Technol. 2004, 38 (6), 18881896. (6) Kedem, O.; Katchalsky, A. Thermodynamic analysis of the permeability of biological membranes to non-electrolytes. Biochim. Biophys. Acta 1958, 27, 229-246. (7) Lonsdale, H. K.; Merten, U.; Riley, R. L. Transport properties of cellulose acetate membranes to selected solutes. J. Appl. Polym. Sci. 1965, 9, 1341-1362. (8) Rosenbaum, S.; Skeins, W. E. Concentration and pressure dependence of rate of membrane permeation. J. Appl. Polym. Sci. 1968, 12, 2169-2181. (9) Jonsson, G. Overview of theories for water and solute transport in UF/RO membranes. Desalination 1980, 35, 21-38. (10) Mason, E. A.; Lonsdale, H. K. Statistical-mechanical theory of membrane transport. J. Membr. Sci. 1990, 51, 1-81. (11) Mulder, M. Basic Principles of Membrane Technology, 2nd ed.; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1996.

(12) Van Gauwbergen, D.; Baeyens, Assessment of the design parameters for wastewater treatment by reverse osmosis. Water Sci. Technol. 1999, 40 (4-5), 269-276. (13) Wijmans, J. G.; Baker, R. W. The solution-diffusion model: a review. J. Membr. Sci. 1995, 107, 1-21. (14) Levenstein, R.; Hasson, D.; Semiat, R. Utilization of the Donnan effect for improving electrolyte separation with nanofiltration membranes. J. Membr. Sci. 1996, 116, 77-92. (15) Van Gauwbergen, D.; Baeyens, J.; Creemers, C. Modeling osmotic pressures for aqueous solutions for 2-1 and 2-2 electrolytes. Desalination 1997, 109, 57-65. (16) Van Gauwbergen, D.; Baeyens, J. Modeling reverse osmosis by irreversible thermodynamics. Sep. Purif. Technol. 1998, 13, 117128. (17) Pusch, W. Determination of transport parameters of synthetic membrane by hyperfiltration experiment, Part II: Membrane transport parameters Independent of pressure and/or pressure difference. Ber. Bunsen-Ges. Phys. Chem. 1977, 81 (3), 854-864. (18) Song, L. Thermodynamic modeling of solute transport through reverse osmosis membrane. Chem. Eng. Commun. 2000, 180, 145-167. (19) Merten, U. Transport properties of osmotic membranes. In Desalination by Reverse Osmosis; MIT Press: Cambridge, MA. 1966. (20) Peeters, J. M. M.; Boom, J. P.; Mulder, M. H. V.; Strathmann, H. Retention measurements of nanofiltration membranes with electrolyte solutions. J. Membr. Sci. 1998, 145, 199-209. (21) Ong, S. L.; Zhou, W. W.; Song, L.; Ng, W. J. Evaluation of feed concentration effects on salt/ion transport through RO/NF membranes with Nernst-Planck-Donnan model. Environ. Eng. Sci. 2002, 19 (6), 429-439. (22) Schirg, P.; Widmer, F. Characterisation of nanofiltration membranes for the separation of aqueous dye-salt solutions. Desalination 1992, 89, 89-107.

Received for review February 12, 2004. Revised manuscript received February 9, 2005. Accepted February 11, 2005. ES0403561

VOL. 39, NO. 9, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3387