Determination of membrane pore size distribution using the fractional rejection of nonionic and charged macromolecules

Determination of membrane pore size distribution using the fractional rejection of nonionic and charged macromolecules

Journal of Membrane Science 201 (2002) 191–201 Determination of membrane pore size distribution using the fractional rejection of nonionic and charge...

434KB Sizes 0 Downloads 41 Views

Journal of Membrane Science 201 (2002) 191–201

Determination of membrane pore size distribution using the fractional rejection of nonionic and charged macromolecules Sangyoup Lee a , Gunyoung Park a , Gary Amy b , Seung-Kwang Hong c , Seung-Hyeon Moon a , Duck-Hee Lee d , Jaeweon Cho a,∗ a

c

Department of Environmental Science and Engineering, Kwangju Institute of Science and Technology, 1 Oryong-dong, Buk-gu, Gwangju 500-712, South Korea b Civil & Environmental Engineering Department, University of Colorado, Boulder, CO 80309, USA Civil & Environmental Engineering Department, University of Central Florida, P.O. Box 162450, Orlando, FL 32816-2450, USA d Young Lin Instrument Co. Ltd., Shin-sa dong, Kang nam-gu, Seoul 135-120, South Korea Received 31 March 2001; received in revised form 31 March 2001; accepted 22 October 2001

Abstract The objective of this study was to develop a new measurement technique for the determination of pore size distributions (PSDs) of polymeric and ceramic membranes, including NF, UF, and MF membranes. The proposed method uses the fractional rejection (FR) concept of a solute in membrane pores. Experimental measurements were conducted using a high performance liquid chromatography (HPLC) equipped with size exclusion chromatography (SEC) columns and a refractive index (RI) detector. A specially designed membrane filtration unit was also used. Two different macromolecules, including nonionic polyethylene glycols (PEG) and natural organic matter (NOM) with ionizable functional (carboxylic and phenolic) groups, were used as solutes. Membrane PSDs, determined with PEG and NOM, can be defined as absolute and effective membrane PSDs, respectively. Two different types of membranes (flat-sheet polymeric and tubular ceramic) were used in this work. Experimental procedures include three major steps: (1) measurements of relative molecular mass (RMM) distributions of solutes included in the membrane feed and corresponding permeate, (2) the calculation of solute FR, and (3) PSD determination. The main results and advantages of this method are: (1) the PSD of various membranes with different pore sizes can be measured using a relatively easy method without significant limitations of pore size and membrane type; (2) various factors that affect membrane PSD, including pH, ionic strength, ion binding, and hydrodynamics, can also be evaluated; (3) the effective PSD of membranes with negatively-charged surfaces, and which exhibit significant shifts in PSD towards the lower RMM region can also be determined. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Pore size distribution; Fractional rejection; Nonionic polyethylene glycol

1. Introduction and background Size exclusion plays a major role in the solute rejection of a membrane based on its pore size and ∗ Corresponding author. Tel.: +82-62-970-2443; fax: +82-62-970-2434. E-mail address: [email protected] (J. Cho).

the solute molecular size. The pore size and its distribution have been measured using various methods including the bubble point method, liquid displacement, solute probe techniques, and many others [1,2]. Each method has its specific characteristics for various membranes with different pore sizes, and also exhibits some limitations in membrane type and pore size. In this work, a characterization technique was developed

0376-7388/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 7 3 8 8 ( 0 1 ) 0 0 7 2 9 - 3

192

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

to determine pore size distribution (PSD), using the fractional rejection (FR) of nonionic polyethylene glycol (PEG), of different membrane materials and types (spiral wound, hollow fiber, and tubular types). We also employed charged macromolecules to obtain an effective PSD, resulting from charge interactions between the membrane surface and charged natural organic matter (NOM) components. Nonionic PEG macromolecules have been widely used for the determination of the nominal molecular weight cutoff (MWCO) and PSD of membranes [2,3]. Singh et al. [2] described membrane characteristics using solute transport and atomic force microscopy (AFM). They derived a relation between solute rejection by a membrane and solute diameter. Using this relation, mean pore size could be calculated. A probability density function was then used to determine membrane PSD using mean and standard deviation values of pore size that were obtained from solute rejection tests and AFM images. In these studies, however, the influence of the steric and hydrodynamic interaction between PEG and pore sizes on solute rejection were ignored. It was found by Braghetta et al. [4], and Cho et al. [5] that the rejection of PEGs by charged membranes was influenced by solution chemistry parameters, such as pH and ionic strength. The percentage rejections of PEG by charged membranes increased at higher ionic strength and with the addition of calcium, compared to an ambient solution at a lower ionic strength. This presents evidence that membrane pore size is effectively reduced by double layer compaction due to increased ionic strength and cation binding. The nominal MWCO values of membranes, provided by manufacturers, are generally calculated by rejection tests with a PEG with an average relative molecular mass (RMM). However, every PEG solution exhibits RMM distributions with higher monodispersivity than expected. The RMM distribution of PEGs have been measured using the fractionation method with various UF membranes with different nominal MWCO values [4]. In this work, the RMM distribution of PEGs was measured using high performance size exclusion chromatography (HPSEC), and high performance liquid chromatography (HPLC) with a refractive index (RI) detector. The added calcium (for ion binding to the membrane surface) and increased ionic strength can reduce the double layer thickness of negative-charged mem-

branes; subsequently, the flexible membrane matrix can be shrunk due to reduced charge repulsion between the ionizable functional groups in the membrane polymeric matrix [4]. Reduced pore sizes result in an increase in water permeability and a decrease in neutral solute rejection by porous membranes, during which solute transport is controlled by convection as opposed to diffusion [6]. Even though it is generally assumed that the distance between the adjacent polymer matrices is fixed and not dependent upon charge interaction, water (solvent) transport through pores may be enhanced when the double layer is expanded [4]. This increased water permeability can reduce neutral solute rejection when the ionic strength is low. When a negatively-charged membrane was used for the rejection of a neutral solute (like PEG), it became clear that water permeability and solute rejection can be affected by the solution pH and ionic strength, as well as the flexibility of polymer membrane matrix. When negative-charged membranes are used for the rejection of negative-charged molecules, charge interactions between the membrane and solute may play a major role in solute rejection. The nominal MWCO of a membrane under these conditions is no longer an indicator of solute rejection. Effective MWCO was a concept introduced to predict ionic solute rejections by charged membranes, and these values were much smaller than the manufacturer’s nominal membrane MWCOs, and those determined by PEG rejection tests [4,7]. 2. Experimental 2.1. Nonionic and charged macromolecules as solutes PEG solutions (Aldrich) with a wide range of average RMM (100–10,000) were purchased, and each PEG solution was prepared at a concentration of approximate 50.0 mg/l using deionized (DI) water of conductivity 0.8 ␮S/cm. The monodispersivity (the ratio of weight-average RMM to number-average RMM) of the PEGs were reported to be in the range of 1.1–1.2 by the manufacturer. The molecular mass distributions were measured by size exclusion chromatography (SEC) with a PEG separation column (Waters, Ultrahydrogel 120, Japan) and an RI detector (Waters 410, Milford, US). DI water with a low organic

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

concentration (less than 500 ␮g/l of total organic carbon (TOC)) was prepared through a series of processes, which included two mixed beds of anion/cation exchange resins, an activated carbon cartridge, and a reverse osmosis membrane filter. TOC was measured using the UV/oxidants oxidation method (Dohrman, DC-180, US) with a TOC analyzer and an autosampler. Nakdong river surface water (NR-SW) was sampled from the Bansong water treatment plant located at Changwon city (Korea), and immediately filtered through a 0.45 ␮m filter and stored in a refrigerator at 5 ◦ C. Dissolved organic carbon (DOC) was measured, and NOM in NR-SW was separated into three different fractions: hydrophobic NOM, transphilic NOM, and hydrophilic NOM using the XAD-8 and XAD-4 isolation method [8,9]. Hydrophobic and transphilic NOM were mainly comprised of hydrophobic and hydrophilic acids, respectively [10]. Each fraction of the three NOM components were characterized by mass measurements, and the carboxylic and phenolic acidities of the hydrophobic and hydrophilic acids were determined using a micro-titrator (Metrohm, CH-910). Fifty milliliters samples were taken and acidified with 5N HCl (pH < 3.0), and sparged with nitrogen gas for at least 15 min to remove inorganic carbonate species. Incremental volumes of 2.0–25.0 ␮l of 0.05N NaOH were then added by micro-titrator to increase the pH to 10.0. The amount of 0.05N NaOH added allowed the determination of carboxylic (pH 3–8) and phenolic acidities (twice the amount between pH 8–10) [11]. The charge density of the NOM acids, as determined by the titration method, can be employed to demonstrate charge interactions between charged NOM acids and a negatively-charged membrane. 2.2. Membrane materials Two different polymeric and four different ceramic membranes were used for PSD determinations. The nominal MWCOs used for each membrane were as stated by the manufacturers. We also determined the nominal MWCO of the membranes using PEG rejection tests and the resulting fractional rejections; the nominal MWCO of the membrane may be defined as the relative molecular mass of the component that is rejected by 90%. The fractional rejection of the PEG solute by the membrane was calculated using the fol-

193

lowing equation [12]: WMi (feed) − WMi (perm)(1 − Roverall ) RMi = WMi (feed)

(1)

where RMi is the fractional rejection of a certain RMM “i”. WMi is the mass fraction of that RMM in the specific stream and Roverall is the overall amount of solute rejected by the membrane, based on DOC measurements. The membrane surface charge was measured using zeta potential measurement and ionizable functional groups analysis. Zeta potential measurements of polymer membranes were performed by an electrophoretic method using a commercial instrument (ELS-8000, Otsuka Electronics, Japan) with a latex solution. The zeta potential was correlated with electrical mobility as measured by laser light scattering. Titania (TiO2 ), the material that the ceramic membranes were composed of, has a point of zero charge (pHpzc ) of 6.25 [13]. We also measured the zeta potential values at various pH levels (4–8) to determine the iso-electric point (i.e.p.) of the ceramic membrane, with ground fine particles from a tubular type of ceramic membrane. The charge density of the membrane surface can also be measured using a potentiometric titration, similar to the method used for NOM acidity measurement, as described. The active layer of the polymeric membrane sample (surface area = 58.9 cm2 ) was cut into many small pieces, which were then placed in a titration vessel and potentiometrically titrated versus 0.05N NaOH to determine the presence of ionizable functional groups, quantitatively. The ceramic membrane was ground into fine particles, and the fine particles were put into a titration vessel for the potentiometric titration. Membrane charge density was calculated from the amount of 0.05N NaOH added, the units of functionality at a certain pH are expressed as milli-equivalents per gram dried membrane. The characteristics of membranes tested are shown in Table 1. 2.3. Membrane filtration apparatus and operation The membrane filtration unit accommodated active filtration areas of 60.0 and 95.2 cm2 for polymeric and ceramic membranes, respectively, and consisted of a membrane holder, pump with gear type pump head, needle valves (for the feed, retentate, and permeate streams), and pressure and flow-rate gauges. A

194

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

Table 1 Membrane characteristics Membrane

Material

Nominal MWCO provided by manufacturers

Zeta potential at pH 7 (mV)

Acidity at pH 7 (meq./g membrane)

i.e.p.

ESNA GM T-1000 T-3000 T-5000 T-8000

Thin film composite Thin film composite Titania Titania Titania Titania

250 8000 1000 3000 5000 8000

−9.90 −31.5

0.057 0.077

4.80 None (pH 3–10)

−25.7

0.057

3.70

schematic diagram of the filtration unit is shown in Fig. 1; it consists of two different membrane holders, for polymeric or tubular ceramic membranes, which may be exchanged for a particular membrane type. The feed flow-rate and the resulting cross-flow velocity may be adjusted by varying the pump head rpm. Trans-membrane pressure was controlled by a needle valve located in the retentate line; the trans-membrane pressure was in the range of 50 psi (344.7 kPa) ∼ 100 psi (689.4 kPa). The water temperature was maintained at room temperature (21–23 ◦ C). The test macromolecules, PEGs and NOM, exhibited different molecular sizes, and the mass transfer coefficients (k, cm/s) were estimated using the suggested laminar flow relationships (Re = 90–140) in a rectangular channel for polymer membranes, as described by Porter [14]:  k = 1.62

kdh = 1.62(Re Sc dh /L)0.33 D for ceramic membrane

Sh =

(3)

The mass transfer coefficient can be combined with the initial permeate flux (J0 : permeate product (cm3 /s)/membrane active area (cm2 )) to estimate the J0 /k ratio, which is a measure of the relative transport motion (convection through membrane pores versus the back-diffusion away from the membrane) of solutes [5]. 2.4. PSD determination procedures

0.33 UD2

dh L for polymer membrane filtration

For a tubular ceramic membrane, the mass transfer coefficient is estimated from a correlation between the Sherwood (Sh), Reynolds (Re), and Schmidt (Sc) numbers for laminar flow in a tube [12]:

(2)

The PSD of polymeric and ceramic membranes was determined using the following procedure.

where U is the average velocity of feed fluid (cm/s), D the solute diffusion coefficient (cm2 /s), dh equivalent to the hydraulic diameter (cm), L is the channel length (cm).

2.4.1. Step 1 The solution containing either PEG or NOM was made using DI water. Prior to the filtration of a solution, the membrane was filtered with DI water for 3 h,

Fig. 1. Schematic of PSD measurement apparatus.

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

195

and the J0 /k was adjusted to 4.5 by controlling either the pump speed or the needle valve in the retentate stream. Either PEG or NOM solution was then placed in the filtration unit for the membrane filtration tests, and the permeate sample was collected after a period of 3 h. The RMM distributions of solutes in the feed and permeate samples were measured using a HPSEC method. 2.4.2. Step 2 By comparing the RMM distributions, solute fractional rejection may be calculated using Eq. (1), i.e. if a particular size of a solute (either PEG or NOM) is considered, Eq. (1) can provide the specific rejection value of the solute with the given particular molecular size. Using this result obtained from either the PEG or NOM filtration, the nominal and effective MWCOs of a membrane were determined. The difference in nominal and effective MWCO is attributed to the effect of charge interactions between the membrane surface and the NOM acids on NOM rejection. 2.4.3. Step 3 Fractional rejection increases as RMM increases, which is similar to the trend of an accumulated rejection curve. It can be inferred that the slope of the fractional rejection curve at a certain RMM is indicative of the incidence of membrane pores relative to this RMM (i.e. a higher slope corresponding to a particular molecular size means more existence of pores with that size). Thus, the more a pore with a certain RMM exists, the more the increment in the rejection of a solute with that RMM increases.

3. Results and discussion 3.1. Nominal and effective MWCO determination Figs. 2–7 illustrate how to determine the effective MWCO of the polymeric and ceramic membranes using PEG dissolved in DI water either with or without the addition of 10 mM NaCl (or Ca2+ from Ca(NO3 )2 ); (a) RMM distributions of PEGs included in the three different feeds (PEG in DI water, and PEG with the addition of either Ca2+ or NaCl) and corresponding permeate samples (see Figs. 2(a)–7(a)),

Fig. 2. GM membrane: (a) RMM distributions (by SEC) of the PEGs contained in membrane feed (thin solid line), the permeate from the feed (PEG dissolved in DI water, dashed line), the permeate from the feed (PEG dissolved in DI water with an addition of NaCl, 䊏 and solid line), and the permeate from the feed (PEG dissolved in DI water with an addition of Ca2+ , thick solid line); (b) calculated FR of specific RMMs (symbols and lines are the same as those in (a)).

and (b) different fractional rejection curves drawn from the RMM distributions of the three different feeds and the corresponding permeate samples (see Figs. 2(b)–7(b)). RMM distributions of PEGs in the permeates from the feed solution with either Ca2+ or NaCl exhibited greater shifts toward lower RMM, compared to those without the addition of either Ca2+ or NaCl. Using these RMM shifts, the FR of each membrane and the nominal MWCOs were determined. The nominal MWCO of a membrane, in this article, is defined as that RMM which was 90% rejected by the membrane.

196

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

Fig. 3. ESNA membrane: (a) RMM distributions; (b) calculated FR of specific RMMs (symbols and lines are the same as those in Fig. 2).

Fig. 4. T-8000 membrane: (a) RMM distributions; (b) calculated FR of specific RMMs (symbols and lines are the same as those in Fig. 2).

Comparisons between experimentally determined MWCOs (defined as PEG-nominal MWCO) (with PEG solutions in DI water) and nominal MWCOs, as provided by manufacturers, are shown in Table 2. Both MWCOs exhibited similar values, thus, we can infer that manufacturers used a similar method for determining the nominal MWCOs using PEGs with various RMMs or any other nonionic macromolecule. Because PEG is a nonionizable macromolecule, possible factors that affect PEG rejection by a charged membrane include: (1) the ionic strength of the PEG solution, and its influence upon membrane surface charge; (2) an increase in divalent cation (i.e. Ca2+ ) concentration, which results in a decrease of

the membrane surface charge through cation binding with negatively-charged functional groups on the membrane surface; (3) hydrodynamic operating conditions such as the J0 /k ratio that may alter solute transport near the membrane surface. For all the experiments in this paper, a J0 /k value of 4.5 was used, to provide equivalent hydrodynamic conditions. The PEG-nominal MWCOs (based on DI + NaCl (or Ca2+ )) of the two relatively large pore membranes (GM and T-8000), and the ESNA membrane were slightly smaller than both the PEG-nominal (based on DI) and manufacturer’s nominal MWCOs. This result can be explained by two possible phenomena: (1) for a polymeric membrane, membrane pore sizes

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

Fig. 5. T-5000 membrane: (a) RMM distributions; (b) calculated FR of specific RMMs (symbols and lines are the same as those in Fig. 2).

are actually reduced due to charge repulsion reduction within the membrane matrices, as a result of double layer compaction; (2) for both polymeric and ceramic membranes, the double layer of the membrane surface inside the pores was compacted due to increased ionic strength (upon the addition of NaCl) and cation binding, and that the compaction of this double layer reduced the pore permeation area available for water transport [4]. The second explanation is based on the assumption that the tested UF and NF membranes are mostly controlled by convection (as opposed to diffusion) for solute transport through the membrane pores. It can be verified that the Peclet number for all of the filtration tests in this article were much higher than 1.0.

197

Fig. 6. T-3000 membrane: (a) RMM distributions; (b) calculated FR of specific RMMs (symbols and lines are the same as those in Fig. 2).

3.2. Absolute PSD The PSD of each membrane may be determined by interpreting the slopes of the fractional rejection curve of the membrane as incidences of particular pore sizes. Continuous PSD of polymeric and ceramic membranes are shown in Figs. 8 and 9, respectively. It should be noted, that as the membrane nominal MWCO increases, it exhibits a wider PSD pattern. For example, GM and T-8000 (MWCO = 8000 mass units) contain many pores with RMMs between 6000 and 8000, suggesting that a solute with a RMM of 7000 can pass through the membrane pores to some extent. However, the T-3000 membrane (MWCO =

198

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

Fig. 7. T-1000 membrane: (a) RMM distributions; (b) calculated FR of specific relative molecular masses (symbols and lines are the same as those in Fig. 2).

Fig. 8. PSDs of: (a) the GM; (b) the ESNA membranes which were determined with membrane permeates from various feed solutions (PEG dissolved in DI water (dashed line), PEG dissolved in DI water with an addition of NaCl (䊏 and solid line), and PEG dissolved in DI water with an addition of Ca2+ (thick solid line)).

3000 mass units) does not seem to allow a solute with a RMM of 3000 to pass through the membrane pores. Even though the nominal MWCO of each membrane may indicate the relative size exclusion potential, it is advantageous to obtain a membrane PSD to more fully describe the solute size exclusion mechanism. Charge interaction effects on membrane PSD were more distinct for polymeric membranes than for ceramic membranes. The two polymeric membranes were influenced by cation binding and ionic strength increase with respect to charge interactions. Effective pore sizes of the GM membrane were somewhat reduced by both divalent cation binding and increased ionic strength. These results can also be explained by actual shrinkage of pores and/or reduced available

permeation areas in pores, which were discussed in the previous section on nominal and effective MWCO. The ESNA membrane differed from the GM membranes in terms of effective PSD, i.e. increased ionic strength caused a slight increases in pore size, however, this difference is almost negligible in terms of the total RMM scale (almost 10–20 RMM units). Generally, ceramic membranes do not appear to be affected by charge interactions in terms of effective PSD. Only the T-8000 membrane was similar to the polymeric membranes in this respect, the other ceramic membranes were not influenced by charge interactions with respect to pore size distribution. This may be understood on the basis that ceramic membranes are composed of rigid membrane materials, which

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

199

Table 2 Experimentally determined and nominal MWCO values Membranes tested

Experimentally determined MWCO PEG-nominal MWCO, based on DI

Comparison of experimentally determined and nominal MWCO values GM 7830 ESNA 230 T-8000 7660 T-5000 5230 T-3000 3360 T-1000 1040

PEG-nominal MWCO, based on DI + NaCl (DI + Ca2+ ) 7740 220 7480 5230 3360 1040

Nominal MWCO, as provided by manufacturer

8000 250 8000 5000 3000 1000

Percentage variations between experimentally determined and nominal MWCO values GM 2.0 3.0 ESNA 8.0 12.0 T-8000 4.3 6.5 T-5000 4.6 4.6 T-3000 12.0 12.0 T-1000 4.0 4.0

Fig. 9. PSDs of four different ceramic membranes: (a) T-8000; (b) T-5000; (c) T-3000; (d) T-1000 (symbols and lines are the same as those in Fig. 8).

200

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

surround the pores, thus, ceramic membrane pores are resistant to shrinkage, even though charge repulsion between the membrane pore surface is reduced due to either cation binding or increased ionic strength. A reduction of water permeation favored area (i.e. polar permeation area in pores influenced by surface charge), resulting from double layer compaction due to increased ionic strength, is believed to occur in ceramic membranes; however, this did not affect convective solute transport through the membrane. 3.3. Effective PSD Up to this point, the membrane PSD was determined by the FR-PEG method with nonionizable PEG. The PSDs obtained are not dependent upon charge interactions between the membrane pore surface and the solute. When solute rejection tests of a charged membrane were performed with NOM containing NOM acids, charge repulsion effects did increase NOM rejection by the membrane. The presence of negatively-charged NOM acids and their charge densities were determined by both XAD isolation and the potentiometric titration method. The charge densities (in terms of carboxylic and phenolic acidities) of the two isolated NOM acids from feed water, are represented in Table 3, along with the acidities of the NOM acids associated with the membrane permeates. The membrane permeates contained less NOM acidity than the membrane feed, which suggested that negative-charged membranes preferentially reject NOM acids (hydrophobic acids + hydrophilic acids) rather than noncharged NOM (primarily the hydrophilic NOM fraction). Increased solute rejection due to preferential rejection of NOM acids, subsequently resulted in much smaller PSDs (i.e. effective Table 3 NOM acidities of raw and membrane treated samples Sample

Feed GM ESNA

Hydrophobic acid of hydrophobic NOM (36.2% of total NOM as C, meq./g C)

Hydrophilic acid of transphilic NOM (19.6% of total NOM as C, meq./g C)

–COOH

–OH

–COOH

–OH

36.6 19.9 3.5

17.4 12.1 4.9

26.3 15.4 4.1

19.9 9.8 5.2

Fig. 10. Effective PSD of the GM and ESNA membranes which were determined with NOM containing negatively-ionizable functional groups such as carboxylic and phenolic groups: (a) the molecular mass distribution of feed NOM of NR-SW; (b) the effective pore size distribution of the GM membrane; (c) the effective pore size distribution of the ESNA membrane.

PSD) compared to the PSDs determined with PEG (see Fig. 10). This result implies that the nominal MWCO of membrane in this situation is not an informative parameter for the rejection of solutes with ionizable functional groups. The effective MWCO and PSD of

S. Lee et al. / Journal of Membrane Science 201 (2002) 191–201

membranes enable us to predict solute rejection based on a size exclusion and charge interactions mechanisms, and to demonstrate the effects of chemical and operating conditions on membrane performance. 4. Conclusions In this article, a new measurement technique for the determination of PSD was proposed. This proposed a relatively easy method, and did not include significant limitations. Using it, the PSD of various membranes ranging from NF to MF could be measured. This method uses the FR concept with nonionic (PEG-like) and charged solutes (for example NOM); the FR was estimated from the RMM distributions of solutes, which were measured by SEC. This method could provide membrane PSDs in terms of relative molecular mass (mass units), which allows information obtained to be used directly to predict membrane solute rejection. This PSD determination method can be applied for both polymeric and ceramic membranes with a wide range of pore sizes under different chemical and hydrodynamic conditions. When solution chemistry (such as pH and ionic strength) is altered, determined PSD of a membrane differ from absolute PSD (determined using deionized water). The “effective PSD of membrane” is new membrane information, which is significantly different from the currently quoted absolute PSD. The effective PSD could be determined for charged macromolecules (like NOM acids), and charge interactions that enhanced solute rejection were taken into account, which provides a significantly different PSD from that quoted at present and facilitates the accurate prediction of membrane molecular weight dependent rejection. It is anticipated that when membrane-related research studies are performed, this suggested technique can be used to determine membrane PSD prior to actual filtration tests, because membrane MWCO, as currently quoted, is too approximate a measure of the true membrane rejection characteristics. Acknowledgements This work was supported by the Korea Science and Engineering Foundation (KOSEF) through the

201

Advanced Environmental Monitoring Research Center (ADEMRC) at Kwangju Institute of Science and Technology. This work was also supported by the National Research Laboratory (NRL) Program (Cleaner Separation Lab.) of Korea Institute of Science and Technology Evaluation and Planning (KISTEP) (Project No. 2000-N-NL-01-C-185). References [1] S.I. Nakao, Review: determination of pore size and pore size distribution. 3. Filtration membranes, J. Membr. Sci. 96 (1994) 131–165. [2] S. Singh, K.C. Khulbe, T. Matsuura, P. Ramamurthy, Membrane characterization by solute transport and atomic force microscopy, J. Membr. Sci. 142 (1998) 117–127. [3] C.M. Tam, A.Y. Tremblay, Membrane pore characterization— comparison between single and multicomponent solute probe techniques, J. Membr. Sci. 57 (1991) 271–287. [4] A. Braghetta, F.A. DiGiano, W.P. Ball, Nanofiltration of natural organic matter: pH and ionic strength effects, J. Environ. Eng. ASCE 123 (7) (1997) 628–641. [5] J. Cho, G. Amy, J. Pellegrino, Membrane filtration of natural organic matter: factors and mechanisms affecting rejection and flux decline with charged ultrafiltration (UF) membrane, J. Membr. Sci. 164 (1–2) (2000) 89–110. [6] A. Tandon, S.K. Gupta, G.P. Agarwal, Modeling of protein transmission through ultrafiltration membranes, J. Membr. Sci. 97 (1994) 83–90. [7] J. Cho, G. Amy, J. Pellegrino, Membrane filtration of natural organic matter: initial comparison of rejection and flux decline characteristics with ultrafiltration and nanofiltration membranes, Water Res. 33 (1999) 2517–2526. [8] G.R. Aiken, D.M. McKnight, K.A. Thorn, E.M. Thurman, Isolation of hydrophilic organic acids from water using nonionic macroporous resins, Org. Geochem. 18 (4) (1992) 567–573. [9] E. Thurman, R. Malcolm, Preparative isolation of aquatic humic substances, Environ. Sci. Technol. 15 (1981) 463– 466. [10] J.A. Leeheer, T.I. Noyes, A Filtration and Column Adsorption System for On Site Concentration and Fractionation of Organic Substances from Large Volumes of Water, US Geological Survey on Water-Supply, Paper 2230, 1984. [11] E.M. Thurman, Organic Geochemistry of Natural Waters, Martinus Nijhoff/Dr W. Junk Publishers, Dordrecht/ Boston/Lancaster, 1985. [12] M. Mulder, Basic Principles of Membrane Technology, Kluwer Academic Publishers, Dordrecht, 1996. [13] W. Stumm, Chemistry of the Solid–Water Interface: Processes at the Mineral–Water and Particle–Water Interface in Natural Systems, Wiley, New York, 1992. [14] M.C. Porter, Concentration polarization with membrane ultrafiltration, Ind. Eng. Chem. Prod. Res. Dev. 11 (1972) 234–248.