Mi Elimelech JMS 2008

download Mi Elimelech JMS 2008

of 11

Transcript of Mi Elimelech JMS 2008

  • 7/28/2019 Mi Elimelech JMS 2008

    1/11

    Journal of Membrane Science 320 (2008) 292302

    Contents lists available at ScienceDirect

    Journal of Membrane Science

    j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / m e m s c i

    Chemical and physical aspects of organic fouling of forward osmosis membranes

    Baoxia Mi , Menachem Elimelech

    Department of Chemical Engineering, Environmental Engineering Program, P.O. Box 208286, Yale University, New Haven, CT 06520-8286, USA

    a r t i c l e i n f o

    Article history:

    Received 16 December 2007

    Received in revised form 21 March 2008

    Accepted 8 April 2008

    Available online 22 April 2008

    Keywords:

    Forward osmosis

    Osmosis

    Organic fouling

    Alginate

    Humic acid

    Bovine serum albumin

    Intermolecular adhesion force

    Foulantfoulant interaction

    Pressure retarded osmosis

    a b s t r a c t

    The growing attention to forward osmosis (FO) membrane processes from various disciplines raises the

    demand for systematic research on FO membrane fouling. This study investigates the role of various

    physical and chemical interactions, such as intermolecular adhesion forces, calcium binding, initial per-meate flux,and membraneorientation,in organic fouling of forward osmosis membranes. Alginate, bovine

    serum albumin (BSA), and Aldrich humic acid (AHA) were chosenas model organic foulants.Atomic force

    microscopy (AFM) was used to quantify the intermolecular adhesion forces between the foulant and

    the clean or fouled membrane in order to better understand the fouling mechanisms. A strong correla-

    tion between organic fouling and intermolecular adhesion was observed, indicating that foulantfoulant

    interaction plays an importantrole in determining therate and extent of organic fouling. Thefouling data

    showedthat FO fouling is governed by thecoupled influence of chemical andhydrodynamic interactions.

    Calcium binding, permeation drag, and hydrodynamic shear force are the major factors governing the

    development of a fouling layer on the membrane surface. However, the dominating factors controlling

    membrane fouling vary from foulant to foulant. With stronger intermolecular adhesion forces, hydrody-

    namic conditions for favorable foulant deposition leading to cake formation are more readily attained.

    Before a compact cake layer is formed, the fouling rate is affected by both the intermolecular adhe-

    sion forces and hydrodynamic conditions. However, once the cake layer forms, all three foulants have

    very similar flux decline rates, and further changes in hydrodynamic conditions do not influence fouling

    behavior.

    2008 Elsevier B.V. All rights reserved.

    1. Introduction

    Pressure-driven membrane processes microfiltration (MF),

    ultrafiltration (UF), nanofiltration (NF), and reverse osmosis (RO)

    have been employed heavily in the field of water purifica-

    tion, wastewater reclamation, and desalination. These membrane

    processes use hydraulic pressure as the driving force for water

    transport through the membrane. Significant hydraulic pressure

    is often required for the operation of such processes, especially

    in RO desalination, which results in extensive use of prime (elec-

    tric) energy. This drawback, coupled with limitations on feed water

    recovery, has led to the investigation of alternative approaches towater desalination.

    Forward osmosis (FO), a potential alternative to pressure-driven

    membrane processes such as RO in certain applications, has been

    gaining popularity in recent years. FO uses a concentrated draw

    solution to generate high osmotic pressure, which pulls water

    across a semi-permeable membrane from the feed solution. The

    draw solute is then separated from the diluted draw solution to

    Corresponding author. Tel.: +1 203 432 4333; fax: +1 203 432 2881.

    E-mail address: [email protected] (B. Mi).

    recycle thesolute, as well as to produce clean product water.FO has

    been explored for use in seawater desalination [1,2], wastewater

    reclamation [35], industrial wastewater treatment [6], and liq-

    uid food processing [7]. Pressure-retarded osmosis (PRO), a closely

    related process, also utilizes osmotic pressure as the driving force

    for water permeation through a semi-permeable membrane. PRO

    has been evaluated as a potential process for generating electric-

    ity by utilizing the osmotic pressure difference between saline and

    fresh waters [8,9] or between a working fluid and a draw solution

    in a closed loop [7,10].

    The growing interest in FO from various disciplines calls for

    more fundamental research that can lead to a better understand-ing of the FO process and further advances in the technology.

    Efforts have been made to develop a draw solution that can induce

    high osmotic pressure, be easily separated from the clean product

    water, and require low energy for regeneration [2,11]. Other stud-

    ies have aimed at developing an understanding of water transport

    phenomena through FO membranesfor instance, the influence of

    internal concentration polarization, membrane structure/material,

    and membrane orientation on membrane flux [1215]. These stud-

    ies provide importantinformationthat can leadto the development

    of new FO membranes with reduced internal concentration polar-

    ization and high water permeability.

    0376-7388/$ see front matter 2008 Elsevier B.V. All rights reserved.

    doi:10.1016/j.memsci.2008.04.036

    http://www.sciencedirect.com/science/journal/03767388mailto:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_14/dx.doi.org/10.1016/j.memsci.2008.04.036http://localhost/var/www/apps/conversion/tmp/scratch_14/dx.doi.org/10.1016/j.memsci.2008.04.036mailto:[email protected]://www.sciencedirect.com/science/journal/03767388
  • 7/28/2019 Mi Elimelech JMS 2008

    2/11

    B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302 293

    One notable research area that hasbeen overlooked so farin this

    emerging technology is membrane fouling. It is well known that

    membrane fouling is a major obstacle to the efficient application

    of membrane technologyin applications involving seawater desali-

    nation, wastewater reuse, and water treatment. Numerous studies

    have been conducted with pressure-driven membrane processes

    to understand the causes of membrane fouling and to develop

    strategies for fouling control. However, very few publications have

    addressed the problem of FO membrane fouling [6,16].

    One potential advantage of using FO is that it operates with

    no hydraulic pressure, which may result in lower membrane foul-

    ing propensity than pressure-driven membrane processes due to

    lessercake layer compaction. Hollowayet al. [6] compared the foul-

    ing behavior of FO and RO for wastewater centrate treatment, and

    demonstrated a slower flux decline rate in FO than in RO. However,

    the lack of systematic, controlled studies on FO membrane fouling

    makes it impossible to fully explain the different fouling behaviors

    of the two processes. Membrane fouling in RO has been well stud-

    ied and the physicochemical mechanisms involved are relatively

    understood. For instance, the effects of intermolecular forces, diva-

    lent cations, and hydrodynamic conditions (initial flux, cross-flow

    velocity) on RO membrane fouling have been studied extensively

    [1721]. To date, however, it is not clear how these physical and

    chemical interactions impact FO membrane fouling.

    The objective of this study is to understand the role of sev-

    eral key physical and chemical interactions in organic fouling of

    forwardosmosismembranes. Threeorganicmacromolecules algi-

    nate, bovine serum albumin (BSA), and Aldrich humic acid (AHA)

    were chosen as model foulants. Atomic force microscopy (AFM)

    was used to quantify the intermolecular adhesion forces between

    the foulant and the clean or fouled membrane. The adhesion forces

    were correlated with the membrane fouling behavior to elucidate

    thefouling mechanisms of theFO membraneat themolecular level.

    The influence of other physical and chemical factors, for instance,

    calcium binding, initial permeate flux, and membrane orientation,

    was also investigated.

    2. Materials and methods

    2.1. Forward osmosis membrane

    Theforwardosmosis membraneused in this studywas provided

    by Hydration Technologies, Inc. (Albany, OR). It has an asym-

    metric structure and is made of cellulose acetate supported by

    embedded polyester mesh. The total thickness of the membrane

    is approximately 50m based on examination of the membrane

    cross-section by scanning electron microscopy. Other characteris-

    tics of the membrane are given in McCutcheon et al. [2].

    2.2. Organic foulants

    We used bovine serum albumin (BSA), sodium alginate, and

    Aldrich humic acid (AHA) (SigmaAldrich, St. Louis, MO) as model

    organic foulants. According to the manufacturer, the molecular

    weightof theBSA isapproximately66 kDa. Thesodium alginatewas

    extracted from brown seaweed, and its molecular weight ranges

    from 12 to 80 kDa. The organic foulants were received in powder

    form. Stocksolutions forBSA andalginate(10g/L)were preparedby

    dissolving the foulant in deionized (DI) water. Mixing of the stock

    solution was performed for over 24h to ensure complete dissolu-

    tion of the foulant. The stock solution was stored in sterilized glass

    bottles at 4 C without further purification.

    AHA stock solution was purified to decrease ash content and

    remove bound iron following the procedure described by Hong

    and Elimelech [17]. Briefly, AHAsolution (10 g/L) was first prepared

    by dissolving AHA powder in DI water. The pH of the AHA solu-

    tion was adjusted to approximately 1 by addition of 1 M HCl. Then,

    precipitation of AHA took place for 10min. The AHA solution was

    then centrifuged at 628.34 rad/s (6000 rpm) for 10min. After cen-

    trifugation, the supernatant was discarded and the precipitate was

    resuspended in 1 M HClsolution. Theabove cleaning procedurewas

    repeated five times. After acid precipitation, the AHA was further

    purified by dialysis against DI water. Finally, the solid content of

    purified AHA suspension was determined by weighing the mass

    after freeze-drying the sample. The concentration of the AHA solu-

    tion was adjusted to 10g/L, and the stock was stored in a sterilized

    bottle at 4 C.

    2.3. Test solutions

    The feedsolutionfor fouling experiments contained50 mM NaCl

    and200 mg/L foulant; some of the solutions also contained 0.5mM

    CaCl2. The ambientpH of the BSA, alginate, and AHA feed solutions

    was 6.3, 5.8, and 6.2, respectively. The test solution for AFM force

    measurement contained 50 mM NaCl and 20 mg/L foulant, with or

    without 0.5 mM CaCl2. However, the AFM tests used lower foulant

    concentration than fouling experiments, because a high concen-

    tration of foulant interferes with the force measurement. The draw

    solution for the fouling experiments was composed of 1.5 or 4 M

    NaCl. A commercial software from OLI System Inc. (Morris Plains,

    NJ) was used to calculate the osmotic pressures of draw and feed

    solutions.

    2.4. Bench-scale forward osmosis fouling experiments

    The FO fouling experiments were performed with a bench-scale

    membrane system as depicted in Fig. 1. The cross-flow membrane

    cell was custom built with equally structured channels on both

    sides of the membrane. The dimensions of the channels are 77mm

    long by 26 mm wide by 3 mm deep. No spacer was used in the

    channel to accelerate membranefouling.Co-current cross-flowwasused to minimize strain on the suspended membrane. Variable

    speed gear pumps (Micropump, Vancouver, WA) were usedto gen-

    erate cross-flows, forming separate closed loops for the feed and

    draw solutions. The draw solution tank was placed on a digital

    scale (Denver Instruments, Denver, CO) and weight changes were

    monitored by a computer to record the permeate flux. A constant

    Fig.1. A schematic diagramof thelaboratory-scale forwardosmosis(FO) systemfor

    fouling experiments.

  • 7/28/2019 Mi Elimelech JMS 2008

    3/11

    294 B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302

    feed and draw solution temperature of 201 C was maintained

    by a water bath (Neslab, Newington, NH). Heat transfer took place

    through submerged stainless steel heat exchanger coils within the

    water bath.

    The protocol for all fouling experiments comprised the follow-

    ing steps. First, a new membrane coupon was placed in the unit

    before each experiment. Since the membrane has an asymmet-

    ric structure, the fouling behaviors at both membrane orientations

    were tested. When the membrane active layer is placed against

    the feed solution, we refer to this orientation as forward osmosis

    (FO) mode, whereas when the membrane active layer is against

    the draw solution, we refer to this orientation as pressure-retarded

    osmosis (PRO) mode. Next, 2 L feed solution without foulant and

    2 L draw solution were added to the feed and draw solution tanks,

    respectively. Cross-flows of feed and draw solutions were run for

    1 h in their respective closed loops, without passing flow through

    the membrane cell, to stabilize the temperature of the system. The

    temperature was maintained at 201 C for all experiments. After

    reaching the desired temperature, the bypass valves of both cross-

    flows were closed to allow flow of feed and draw solution through

    both sides of the membrane. The cross-flow velocity for both the

    feed and draw solution sides was fixed at 8.5 cm/s. After the initial

    fluxstabilized, which took about 1 h, 200mg/L of foulant was added

    to the feed solution and the fouling experiment was continued for

    2024 h. A computer was used to continuously monitor water flux

    throughout the fouling experiment.

    Baseline experiments were conducted to quantify the flux

    decline due to the decrease in the osmotic driving force during the

    fouling experiments as the draw solution is continuously diluted by

    the permeate water. The baseline experiments followed the same

    protocol as that for the fouling experiments except that no foulant

    was added to the feed solution.

    2.5. Interfacial force measurement

    Atomic force microscopy (AFM) was used to measure thefoulantfoulant and foulantmembrane interfacial forces, follow-

    ing the procedures described by Li and Elimelech [19]. The force

    measurements were performed with a colloid probe on a Multi-

    mode AFM(VeecoMetrology Group, Santa Barbara, CA). The colloid

    probe was made by attaching a carboxylate modified latex (CML)

    particle (Interfacial Dynamics Corp., Portland, OR) to a commercial

    SiN AFM cantilever (Veeco Metrology Group, Santa Barbara, CA).

    The CML particle (4.0m in diameter) was attached by Norland

    Optical adhesive (NorlandProducts, Inc.,Cranbury, NJ) to the tipless

    SiN cantilever and cured under UV light for 20min.

    The AFM adhesion force measurements were performed in a

    fluid cell filled with the test solution. For each force measurement

    experiment, a piece of fresh membrane was set up in the fluid

    cell and rinsed with DI water. Then the test solution, containing

    50 mM NaCl and 20mg/L foulant, with or without 0.5mM CaCl2,

    was injected to fully displace the DI water in the fluid cell. The

    test solution was left to equilibrate with the FO membrane for

    3045 min to allow foulants to adsorb on the membrane surface

    as well as on the colloidal probe. The force measurements were

    conducted at five different locations on the membrane. 30 force

    measurements were taken at each location to minimize inherent

    variability in the force data, which is mainly attributed to the het-

    erogeneity of the membrane surface. Only the retracting (pull-off)

    force curves were processed and converted. For each solution con-

    dition, both theaverages of alladhesionforces at differentlocations

    as well as the force distributions are presented.

    3. Results and discussion

    3.1. Forward osmosis membrane characteristics

    Membrane surface morphology is known to play a role in mem-

    brane fouling. Therefore, we used AFM to characterize the surface

    properties of the FO membrane. As shown in Fig. 2a, the mem-brane has some bumpy areas on the surface, which is primarily

    caused by the embedded polyester mesh. The more localized sur-

    face morphology,shown in Fig.2b, depicts a roughnesson theorder

    of several tens of nanometers. The surface roughness of the FO

    membrane does not differ much from a typical RO/NF membrane

    [22].

    Fig. 3 shows the frequency distribution of the adhesion forces

    between theAFM particle (CML)probeand thecleanFO membrane.

    The CML particle contains a highly charged layer of carboxylate

    functional groups, thereby serving as a surrogate for carboxy-

    late rich foulants, such as humic acid and alginate [19]. Since the

    membrane surface is clean during the initial stage of fouling, the

    adhesion forces between the foulant and the clean membrane

    surface determine the initial fouling rate. The adhesion force, F,normalized by the radius of the particle, R, is proportional to the

    energy per unit area required to separate the particle and the flat

    surface by an infinite distance [19]. Therefore, F/R can be viewed as

    a measure of theenergy required topreventa foulant from accumu-

    lating on the membrane surface, which makes it a good indicator

    for the membrane fouling potential.

    The frequency distribution is obtained from a total of 150 force

    measurements at five different locations. The distribution plot is

    presented to illustrate the spread of intermolecularadhesion forces

    obtained during force measurements. Fig. 3 shows that the adhe-

    sion forcesbetweenthe foulant andthe clean membraneare spread

    over a wide range, from 0.1 to 0.9 mN/m, which is most likely due

    Fig. 2. AFM images of the FO membrane active layer. The images were taken in 50 mM NaCl solution. The scanned areas are: (left) 20m20m and (right) 2m2m.

    Note also the difference in the vertical scales of the images.

  • 7/28/2019 Mi Elimelech JMS 2008

    4/11

    B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302 295

    Fig. 3. Frequency distribution of the adhesion forces between the AFM particle

    probe and the clean FO membrane (representing foulantmembrane interaction).

    Test solution: 50 mM NaCl and pH 7. The force measurements were performed at 5

    different locations on the membrane, with 30 measurements at each location (i.e.,

    total 150 measurements).

    to the surface roughness of the forward osmosis membrane. The

    results indicate that membrane surface morphology greatly influ-

    ences the foulantmembrane interactions.

    3.2. Baseline experiments

    Since the fouling experiments were performed in batch mode,

    the osmotic driving force for water flux kept decreasing due to the

    dilution of draw solution and concentration of feed solution. There-

    fore, the flux decline in the fouling experiments is caused not only

    by membrane fouling but also by the decrease in osmotic driving

    force. In order to separate the effects of fouling and decrease of

    osmotic driving force, we conducted baseline experiments to quan-tify the flux decline due to the progressive decrease in the osmotic

    driving force.

    As shown in Fig. 4, the baseline experiments were conducted

    in both FO and PRO modes under conditions corresponding to

    the fouling experiments. Fig. 4a shows the permeate flux decline

    as a function of accumulated permeate volume. The water flux

    behavior in these experiments as a function of the corresponding

    draw solution osmotic pressure, as calculated based on the dilu-

    tion of draw solution by permeate flow, is presented in Fig. 4b.

    The osmotic pressures of the 50 mM NaCl feed and 1.5/4M NaCl

    draw solutionsare 233047.5 Pa(2.3 atm)and 7498050/26040525Pa

    (74/257 atm), respectively. The FO membrane pure water perme-

    ability coefficient obtained under hydraulic pressure (RO mode) is

    3.61012 m/(s Pa). Note,however, that theactual water fluxof the

    FO membraneis much lowerthan thevalue calculated based on the

    water permeability coefficient and the osmotic driving force. This

    behavior is attributed to internal concentration polarization effects

    as described in detail in our recent publications [12,13].

    The fluxdecline curves shown in Fig. 4 wereused as baselines to

    normalize the flux decline curves obtained from the fouling exper-

    iments. The baseline flux was first divided by its corresponding

    initial fluxto obtain a normalization factor.Then, theflux data from

    the fouling experiment was divided by the normalization factor to

    obtain the correctedflux. Toavoidconfusion,the flux decline curves

    from the fouling experiments presented in this paper use the cor-

    rected, normalized flux instead of the actualobserved flux. In other

    words, the flux decline curves shown in Figs. 512 solely represent

    the effect of membrane fouling.

    3.3. Chemical aspects of fouling: Role of calcium and

    intermolecular forces

    3.3.1. Alginate

    To study the effect of calcium on alginate fouling, fouling exper-

    iments were performed with feed solutions of 50mM NaCl with or

    without Ca2+ (0.5mM). The flux decline curves obtained for each

    fouling condition (Fig. 5a) are compared with the corresponding

    adhesion forces (Fig.5b and c). Note, however, thatthe alginate con-

    centrations are lower in the force measurements than the fouling

    experiments (20 and 200 mg/L, respectively). This is because high

    foulant concentration is needed for accelerated fouling, whereas

    high alginate concentration interferes with intermolecular interac-

    tions in the AFM force measurements.

    Calcium ions have been shown to enhance alginate fouling in

    reverse osmosis membrane systems [8,9]. Similarly, for the FO

    membrane, a much more severe flux decline is observed duringalginate fouling in the presence of calcium ions compared to that

    in the absence of calcium ions (Fig. 5a). Consistently, the aver-

    age adhesion force in the presence of calcium ions is 0.66 mN/m,

    about twice the average force without calcium ions. The results

    demonstrate that the presence of calcium ions enhances the inter-

    molecular adhesion between alginate molecules, resulting in more

    severe membrane fouling.

    The greater adhesion between alginate molecules in the pres-

    ence of calcium ions is attributed to intermolecular bridging by

    Fig. 4. Baseline tests with the clean FO membrane in FO and PRO modes. The permeate flux decline is plotted against (a) the accumulated permeate volume and (b) the

    corresponding osmotic pressure of the draw solution. The feed and draw solutions initially contain 50 mM NaCl and 4.0/1.5M NaCl, respectively. Other test conditions:

    cross-flow velocity of 8.5 cm/s, pH of 7.0, and temperature of 201

    C. Note that a water flux of 10m/s corresponds to 36.0 L m

    2 h

    1 or 21.2gal ft

    2 d

    1 (gfd).

  • 7/28/2019 Mi Elimelech JMS 2008

    5/11

    296 B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302

    Fig. 5. Effectof calcium on alginatefoulingof theFO membrane(FO mode).Experi-

    mental conditions forfoulingexperiments: draw solutioncontaining 4 M NaCl; feed

    solution containing 200 mg/L alginate and 50 mM NaCl, with/without 0.5mM Ca2+;

    cross-flow velocity of 8.5 cm/s; pH of 5.8; temperature of 201 C. (a) Flux decline

    curves with alginatein thepresence andabsenceof Ca2+; water fluxes arecorrected

    to account for the dilution of draw solution and reduction of the osmotic driving

    force during the fouling run. (b and c) Frequency distribution of the foulantfoulant

    adhesion forces for alginate in the absence of Ca2+ and in the presence of 0.5 mM

    Ca2+, respectively. The test solutions for the force measurements had a pH of 5.8,

    contained20 mg/Lalginate and 50mM NaCl, with/without calcium. The membrane

    surface area for the fouling tests is 0.002m2. Note that a water flux of 10m/s

    corresponds to 36.0 L m2 h1 or 21.2gal ft2 d1 (gfd).

    Fig. 6. TEM cross-sections of alginate fouling layer formed in a fouling experiment

    with 200 mg/L alginate, 50 mM NaCl, and 0.5mM Ca2+. The alginate gel layer was

    peeled off the membrane surface after the fouling experiment.

    calcium ions, resulting in the formation of a cross-linked alginate

    gellayer on themembrane surface. Alginate is a linearcopolymer of

    mannuronic and guluronic acids that contain abundant carboxylicfunctional groups. Calcium ions bind preferentially to the carboxy-

    late groups of alginate in a highly cooperative manner and form

    bridges between neighboring alginate molecules, leading to the

    formation of a gel network [18]. The alginate gel layer formed in

    the fouling experiment with 200 mg/L alginate, 50 mM NaCl, and

    0.5mM Ca2+ was peeled off the membrane surface and examined

    under TEM. As demonstrated in Fig. 6, the alginate gel layer is com-

    posed of cross-linked long chain molecules forming a relatively

    thick network structure.

    3.3.2. BSA

    The effect of calcium on BSA fouling is also studied by per-

    forming fouling experiments and force measurements with and

    without Ca2+. The flux decline curves obtained from each foulingrun (Fig. 7a) are compared with the adhesion forces measured by

    AFM (Figs. 7b and c). The BSA concentrations in the force mea-

    surements and the fouling experiments were 20 and 200 mg/L,

    respectively.

    Fig. 7a demonstrates that, unlike alginate, BSA fouling is not

    affected by the presence of calcium ions. Consistently, the adhe-

    sion forces measured in the absence and presence of calcium ions

    are comparable, which is also roughly the same as the adhesion

    force with alginate in theabsence of calcium ions. Our results again

    demonstratethat theintermolecular force is a good indicator of the

    fouling behavior of organic foulants. The reason that calcium does

    not affect BSA fouling could be due tothe low content of carboxylic

    groups in BSA molecules, which greatly reduces the possibility of

    forming complexes and a cross-linked foulant layer with calcium

  • 7/28/2019 Mi Elimelech JMS 2008

    6/11

    B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302 297

    Fig.7. Effect ofcalciumon BSAfouling ofthe FOmembrane(FO mode).Experimental

    conditionsforfoulingexperiments:drawsolutioncontaining4 MNaCl;feedsolution

    containing 200mg/L BSA and 50 mM NaCl, with/without 0.5mM Ca2+; cross-flow

    velocity of 8.5 cm/s; pH of 6.3; and temperature of 201 C. (a) Flux decline curves

    ofBSA inthe presenceandabsenceof Ca2+; water fluxes arecorrected to account for

    the dilution of draw solution and reduction of the osmotic driving force during the

    fouling run. (b and c) Frequency distribution of the foulantfoulant adhesion forces

    for BSA in the absence of Ca2+ and in the presence of 0.5mM Ca2+, respectively. The

    test solutionsforthe force measurementshad a pHof 6.3,contained20 mg/LBSAand

    50 mM NaCl, with/without calcium. Note that a water flux of 10m/s c orresponds

    to 36.0L m2 h1 or 21.2gal ft2 d1 (gfd).

    Fig. 8. Effect of calcium on AHA fouling of the FO membrane (FO mode). Exper-

    imental conditions for fouling experiments: draw solution containing 4 M NaCl;

    feedsolution containing 200 mg/LAHA and50 mMNaCl,with/without0.5 mMCa2+;

    cross-flow velocity of 8.5 cm/s; pH of 6.2; temperature of 201 C. (a) Flux decline

    curves of AHA in the presence and absence of Ca2+; water fluxes are corrected

    to account for the dilution of draw solution and reduction of the osmotic driving

    force during the fouling run.(b and c) Frequency distribution of the foulantfoulant

    adhesion forces for AHA in the absence of Ca 2+ and the presence of 0.5 mM Ca2+,

    respectively. The test solutions for the force measurements had a pH of 6.2, con-

    tained 20 mg/L AHA and 50 mM NaCl, with/without calcium. Note that a water flux

    of 10m/s corresponds to 36.0 L m2 h1 or 21.2gal ft2 d1 (gfd).

  • 7/28/2019 Mi Elimelech JMS 2008

    7/11

    298 B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302

    ions. The concentrations of carboxylic functional groups in various

    foulants, represented by carboxylic acidity, have been determined

    by potentiometric titration in previous studies [23]. The carboxylic

    acidity of BSA is around 1 meq/g, much lower than that of alginate,

    3.5meq/g.

    3.3.3. Humic acid

    The flux decline curves obtained with AHA are shown in Fig. 8a.The correspondingadhesion forces measuredwith andwithout cal-

    cium ions are shown in Figs. 8b and c. The AHA concentrations

    in the force measurements and fouling experiments were 20 and

    200 mg/L, respectively. The flux decline curves demonstrate that

    AHA fouling is enhanced by the presence of calcium ions, although

    not to the large extent shown for alginate. This behavior is con-

    sistent with AHA fouling behavior observed in RO systems [20].

    However, the measured adhesion forces are not consistent with

    the fouling data. Fig. 8b shows that the adhesion forces measured

    in the absence of calcium ions are unexpectedly large and are dis-

    tributed over a very wide range. This observation is attributable to

    the presence of colloidal aggregates of AHA of varying morphology

    and size (from a few nanometers to hundreds of nanometers) [24].

    The colloidal aggregates interfere with the intermolecular force

    measurements, thus producing false results of adhesion forces.

    3.3.4. Correlation between fouling and intermolecular adhesion

    force

    Fig. 9 compares the fouling behaviors of the three foulants: BSA,

    AHA, and alginate. In the absence of calcium ions (Fig. 9a), the flux

    decline rates for the three foulants are relatively slow and are only

    slightly different. Upon addition of calcium ions, however, the dif-

    ferences in the fouling rates with the three foulants become more

    significant, where alginate fouls the membrane much faster than

    AHA and BSA (Fig. 9b).

    The average adhesion forces (F/R) corresponding to each fouling

    experiment are also listed in Fig. 9. We observe a strong correlation

    between organic fouling rate and intermolecular adhesion forces.

    A stronger adhesion force is generally associated with a faster fluxdecline. The correlation between organic fouling and intermolec-

    ular adhesion confirms our previous finding in RO systems that

    foulantfoulant interaction plays an important role in determining

    the rate and extent of organic fouling [19]. Strong foulantfoulant

    adhesion causes faster accumulation of foulant on the membrane

    surface, thereby resulting in more severe membrane fouling. Our

    results suggest that the foulant adhesion force can serve as a good

    indicator of the fouling potential in FO membrane systems.

    Fig. 9. Comparison of the fouling behaviors of BSA, AHA, and alginate in the FO mode. (a and b) Absence and presence of Ca 2+, respectively. Water fluxes are corrected to

    account for the dilution of draw solution and reduction of the osmotic driving force during the fouling run. Experimental conditions for fouling experiments: draw solution

    containing 4M NaCl; feed solution containing 200 mg/L foulant and 50mM NaCl, with/without 0.5mM Ca2+; cross-flow velocity of 8.5cm/s; temperature of 201 C. Note

    that a water flux of 10m/s corresponds to 36.0 L m2

    h1

    or 21.2gal ft2

    d1

    (gfd).

    The strong correlation between organic fouling and intermolec-

    ular adhesion also provides a mechanistic understanding at the

    molecular level for the different effects of calcium on the foul-

    ing behaviors of the three foulants. Calcium ions are known to

    form complexes with the carboxylic functional groups in organic

    macromolecules, therefore increasing foulantfoulant intermolec-

    ular adhesion. The carboxylic acidities for BSA, AHA, and alginate

    are 1.0, 3.4, and 3.5 meq/g, respectively, as determined by poten-

    tiometric titration in our previous studies [17,23]. The carboxylic

    acidity of BSA is the lowest among the three foulants, indicating

    that BSA molecules have the least opportunity to form complexes

    with calcium. Therefore, the impact of calcium on the adhesion

    force between BSA molecules is the smallest among the three

    foulants, resulting in the slowest fouling rate. For AHA and algi-

    nate, although they have similar carboxylic acidity, the adhesion

    force of alginate is much higher than AHA in the presence of cal-

    cium ions. The difference is attributed to the unique alginate gel

    forming mechanism. The alginate gel formation not only requires

    carboxylate functional groups to form complexes with calcium,

    but also needs certain structural characteristics to allow effec-

    tive intermolecular bridging. For example, in the linear copolymer

    structure of alginate, the polyguluronic acid blocks bind much

    more effectively with calcium ions than the polymannuronic acid

    blocks. Also, it was found that only polyguluronic acid blocks

    over a certain size can be involved in calcium cross-linking, with

    the larger being more effective in cross-linking [25]. Therefore,

    alginate structural characteristics areimportant forforming a cross-

    linked network by intermolecular bridging, which explains why

    alginate has a much stronger force andcorrespondingly faster foul-

    ing rate than AHA, even though the two have similar carboxylic

    acidity.

    3.4. Effect of membrane orientation

    The absence of hydraulic pressure in FO processes allows the

    membrane to be operated in two orientations. When the mem-

    brane active layer is placed against the feed solution, we refer tothis orientation as FO mode, whereas when the membrane active

    layer is against the draw solution, we refer to this orientation as

    pressure-retarded osmosis (PRO) mode. Since internal concentra-

    tion polarizationis more pronounced in FO mode than in PRO mode

    [17], which significantly diminishes the effective osmotic driving

    force for water flux,the draw solution usedin FOmode has a higher

    concentration (4 M NaCl) than in PRO mode (1.5 M NaCl) in order

    to obtain the same initial flux.

  • 7/28/2019 Mi Elimelech JMS 2008

    8/11

    B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302 299

    Fig. 10. Effect of membrane orientation on FO membrane fouling by alginate, BSA,

    and AHA. Thefluxdeclinecurvesof alginate, BSA, andAHAare shownin (a), (b), and

    (c), respectively. Water fluxes arecorrected to account for the dilution of draw solu-

    tion and reduction of the osmotic driving force during the fouling run. Membrane

    orientations are represented by two modes: FO (membrane active layer facing feed

    solution) and PRO (membrane active layer facing draw solution). The feed solution

    contained 200mg/L foulant, 50mM NaCl, and 0.5mM Ca2+. The draw solution con-

    tained 4 M NaCl for FO mode and 1.5M NaCl for PRO mode. The cross-flow velocity

    and temperature are similar to those in Fig. 5. Note that a water flux of 10m/s

    corresponds to 36.0L m2 h1 or 21.2gal ft2 d1 (gfd).

    The effects of membrane orientation on the fouling of alginate,

    BSA, and AHA are shown in Fig. 10. We observe that membrane

    orientation has different effects on the three foulants. Fouling with

    alginate is not affected by membrane orientation, with a similar

    flux decline obtained in FO and PRO modes. In contrast, for BSA and

    AHA, the flux decline is more severe in the PRO mode than in the

    FO mode. This observation is most notable with AHA (Fig. 10c). The

    different membrane orientation effects suggest that the different

    organic foulants have different fouling mechanisms.

    Membrane fouling is generally governed by the coupled influ-

    ence of chemical and hydrodynamic interactions [20]. Chemical

    interactions, such as calcium binding, have been found to affect

    intermolecular adhesion forces and membrane fouling behavior.

    Hydrodynamic interactions, such as permeation drag resulting

    from convective flow toward the membrane and shear force

    resulting from cross-flow parallel to the membrane, influence

    the deposition and accumulation of foulant molecules on the

    membrane surface. The relative importance of these factors in con-

    trolling organic fouling of FO membranes depends on the foulant

    type and membrane orientation.

    Since the FO membrane has an asymmetric structure, charac-

    terized by a dense active layer on top of a porous support layer,

    membrane fouling occurs on different surfaces in FO and PRO

    modes. In FO mode, with the membrane active layer against the

    feed solution, foulant deposition/accumulation occurs on top of

    the active layer. Foulant deposition is affected by both permeation

    drag and shear force, resulting from the permeate flux and bulk

    cross-flow, respectively. In PROmode, however, withthe membrane

    porous support layer against the feed solution, foulant deposition

    takes place within the porous structure of the membrane. Since

    cross-flow velocity vanishes within the porous support layer, the

    influence of hydrodynamic shear forces is absent at the initial stage

    of fouling in PRO mode.

    As the two membrane orientations provide different hydro-

    dynamic conditions during membrane fouling, the effect of

    membrane orientation can be used as an indicator for the role

    of hydrodynamic conditions in membrane fouling. Since mem-brane orientation shows no influence on flux decline with alginate

    (Fig. 10a), we conclude that hydrodynamic interactions do not play

    a dominant role in alginate fouling. Instead, chemical interactions

    (calcium binding) play a more important role. Calcium binding

    results in a highly structured gel layer,which is relatively unaffected

    bychanges in hydrodynamicconditions. ForBSA andAHA, however,

    thecalcium-binding effects are less significant, andthe influence of

    hydrodynamic interactions becomes moreimportant. In PROmode,

    the absence of cross-flow within the membrane porous support

    layer precludes shear force as a mechanism to drive foulant away

    from the membrane. Therefore, fouling with BSA and AHA is more

    severe in the PRO mode than in the FO mode. The marked flux

    decline with AHA in the PRO mode is attributed to cake layer for-

    mation due to lack of shear force as well as hindered back diffusionof AHA aggregates in the porous structure.

    3.5. Effect of initial permeate flux

    The effects of initial flux on fouling of alginate, BSA, and AHA

    are shown in Fig. 11 for fouling in the PRO mode. For each fouling

    experiment, the flux decline is plotted in two different forms: the

    permeate flux versus permeate volume (Figs. 11ac) and the corre-

    sponding normalized flux versus permeate volume (Figs. 11df). A

    faster flux decline is generally noticed for all three foulants at the

    early stages of filtration when the initial permeate flux is high, but

    the degree of influence varies from foulant to foulant.

    For alginate and AHA, the influence of initial flux is relatively

    small, but for BSA, the higher initial flux causes significantly

  • 7/28/2019 Mi Elimelech JMS 2008

    9/11

    300 B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302

    Fig. 11. Effect of initial flux on membrane fouling in the PRO mode (membrane active layer facing draw solution). (a, b, c) The flux decline curves for alginate, BSA, and AHA,

    respectively. (d, e, f) The corresponding normalized flux for the three foulants. Water fluxes are corrected to account for the dilution of draw solution and reduction of the

    osmotic driving force during the fouling run. The draw solution contained 4 M or 1.5M NaCl. The feed solution contained 200 mg/L foulant, 50mM NaCl, and 0.5mM Ca2+.

    The cross-flow velocity and temperature are similar to those in Fig. 5. Note that a water flux of 10m/s corresponds to 36.0 L m2 h1 or 21.2gal ft2 d1 (gfd).

    higher flux decline. Since the effect of initial flux on membrane

    fouling is attributed mainly to the permeation drag resulting from

    convective flow toward the membrane, our results indicate that

    the stronger permeation drag exerted on BSA molecules results in

    the formation of a cake layer. The large difference in the fouling

    behavior with BSA for low and high permeation drags reflects a

    transition in the fouling layer from a loose fouling layer structureto a much more compact cake layer. Changes in permeation drag

    appear to have less effect on alginate and AHA, because the fouling

    layers under both conditions are already in the form of a cake layer.

    Once a cake layer forms, fouling becomes much less sensitive to

    changes in hydrodynamic conditions.

    3.6. Coupled influence of intermolecular adhesion and

    hydrodynamic forces

    Our previous discussion has shown that intermolecular adhe-

    sion and hydrodynamic forces, mainly permeation drag and shear

    force, are the major factors governing the development of a fouling

    layer on the membrane surface. However, the dominating fac-

    tors that control membrane fouling vary significantly with organic

    foulant type. Membrane fouling with alginate is rapid due to inter-

    molecular binding of foulants by calcium, but is relatively insen-

    sitive to hydrodynamic conditions (cross-flow and initial flux). In

    contrast, BSA fouling is more subject to hydrodynamic interactions

    than to calcium effects. The relative influence of intermolecular

    adhesion and hydrodynamic forces on fouling in the FO and PRO

    modes is conceptually illustrated in Fig. 12. The corresponding nor-malized flux decline curves are also shown in this figure.

    Fig. 12 shows a clear trend in the coupled influence of inter-

    molecular adhesion and hydrodynamic interactions on membrane

    fouling. With stronger intermolecular adhesion forces, hydrody-

    namic conditions for favorable foulant deposition leading to cake

    formation aremore readily attained. This is demonstratedin Fig. 12

    for alginate, which forms a cake layer under all three tested hydro-

    dynamic conditions: FO, PRO, and PRO at high initial flux. For AHA,

    the foulant with moderate intermolecular adhesion force, the for-

    mation of a cake layer is absent in the FO mode (the least favorable

    depositionconditions), but is formed when the shearforce vanishes

    in the PRO mode (more favorable deposition conditions). For the

    foulant with very weak intermolecular adhesion force (BSA), cake

    layer forms only in thePRO mode at high initial flux(the most favor-

  • 7/28/2019 Mi Elimelech JMS 2008

    10/11

    B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302 301

    Fig.12. Schematic illustration of the coupled influenceof intermolecularadhesion and hydrodynamic forces on membrane fouling by alginate,AHA, and BSA.The horizontal

    axis compares the different foulants with respect to intermolecular adhesion forces (from strong, left, to weak, right). The vertical axis represents the different hydrody-

    namic/deposition conditions, from more favorable (bottom) to less favorable (top) deposition conditions. The corresponding normalized flux decline curves are shown on

    top. For these experiments, the feed solution contained 200 mg/L foulant, 50mM NaCl, and 0.5 mM Ca2+. Other test conditions are similar to those in Fig. 5.

    able deposition conditions). The favorable deposition conditions

    leading to cakeformation areindicatedin Fig.12 by theshaded area.

    Once a cake layer forms, a rapid flux decline is observed, but fur-ther changes in hydrodynamic conditions do not affect the fouling

    behavior. In addition, when a cake layer forms, the intermolecu-

    lar adhesion force has no effect on fouling. This is demonstrated

    by the similar flux decline curves with the different foulants in the

    PRO mode at high initial flux.

    4. Conclusion

    A strong correlation between organic fouling and intermolecu-

    lar adhesion force was observed, indicating that foulantfoulant

    interaction plays an important role in determining the rate and

    extent of organic fouling. Thefouling data showedthat FO fouling is

    governed by the coupled influence of chemical and hydrodynamic

    interactions. Calcium binding, permeation drag, and hydrodynamic

    shear force are the major factors governing the development of

    a fouling layer on the membrane surface. However, the dominat-

    ing factors controlling membrane fouling can vary from foulant tofoulant. With stronger intermolecular adhesion forces, hydrody-

    namic conditions for favorable deposition and cake formation are

    more readily attained. Alginate has the strongest intermolecular

    interactions due to calcium binding, and it forms cake layer under

    all tested hydrodynamic conditions. On the contrary, having weak

    intermolecular interactions, BSA forms cake layer only at the most

    favorable hydrodynamic conditions. AHA behavior lies in between

    that of alginate and BSA. Before forming a compact cake layer,

    fouling is sensitive to intermolecular interactions and changes in

    hydrodynamic conditions. However, once cake layer forms, a rapid

    flux decline is observed, and changes in hydrodynamic conditions

    or intermolecular adhesion have little effect on the fouling behav-

    ior. All three foulants exhibit very similar flux decline curves when

    cake layers form.

  • 7/28/2019 Mi Elimelech JMS 2008

    11/11

    302 B. Mi, M. Elimelech / Journal of Membrane Science 320 (2008) 292302

    Acknowledgement

    The authors are grateful for the financial support received from

    the California Department of Water Resources under Award Num-

    ber 4600 007446.

    References

    [1] R.E.Kravath,J.A. Davis,Desalination of seawaterby direct osmosis,Desalination16 (1975) 151155.

    [2] J.R.McCutcheon, R.L.McGinnis,M. Elimelech,A novel ammonia-carbondioxideforward (direct) osmosis desalination process, Desalination 174 (2005) 111.

    [3] J.L. Cartinella, T.Y. Cath, M.T. Flynn, G.C. Miller, K.W. Hunter, A.E. Childress,Removal of natural steroid hormones from wastewater using membrane con-tactor processes, Environ. Sci. Technol. 40 (2006) 73817386.

    [4] T.Y. Cath, D. Adams, A.E. Childress, Membrane contactor processes for wastew-ater reclamation in space II. Combined direct osmosis,osmoticdistillation,andmembrane distillation for treatment of metabolic wastewater, J. Membr. Sci.257 (2005) 111119.

    [5] T. Cath, S. Gormly, E. Beaudry, M. Flynn, V. Adams, A. Childress, Membranecontactor processes for wastewater reclamation in space Part I. Direct osmoticconcentration as pretreatment for reverse osmosis, J. Membr. Sci. 257 (2005)8598.

    [6] R.W. Holloway, A.E. Childress, K.E.Dennett, T.Y. Cath, Forward osmosis for con-centration of anaerobic digester centrate, Water Res. 41 (2007) 40054014.

    [7] B. Jiao, A. Cassano, E. Drioli, Recent advances on membrane processes for the

    concentration of fruit juices: a review, J. Food Eng. 63 (2004) 303324.[8] R.J. Aaberg, Osmotic powera new and powerful renewable energy source,

    ReFocus 4 (2003) 4850.[9] K.L. Lee, R.W. Baker, H.K. Lonsdale, Membranes for power generation by

    pressure-retarded osmosis, J. Membr. Sci. 8 (1981) 141171.[10] R.L.McGinnis,J.R. McCutcheon,M. Elimelech,A novel ammonia-carbondioxide

    osmotic heat engine for power generation, J. Membr. Sci. 305 (2007) 1319.[11] J.R. McCutcheon, R.L. McGinnis,M. Elimelech,Desalination by ammonia-carbon

    dioxide forward osmosis: influence of draw and feed solution concentrationson process performance, J. Membr. Sci. 278 (2006) 114123.

    [12] G.T. Gray, J.R. McCutcheon, M. Elimelech, Internal concentration polarizationin forward osmosis: role of membrane orientation, Desalination 197 (2006)18.

    [13] J.R.McCutcheon,M. Elimelech, Influence of concentrative and dilutiveinternalconcentration polarization on flux behavior in forward osmosis, J. Membr. Sci.284 (2006) 237247.

    [14] H.Y. Ng, W. Tang, W.S. Wong, Performance of forward (direct) osmosis pro-cess: membrane structure and transport phenomenon, Environ. Sci. Technol.40 (2006) 24082413.

    [15] K.Y. Wang, T.S. Chuang, J.J. Qin, Polybenzimidazole (PBI) nanofiltration hollow

    fibermembranesappliedin forwardosmosisprocess, J. Membr.Sci. 300 (2007)612.

    [16] T. Cath, A. Childress, M. Elimelech, Forward osmosis: principles, applications,and recent developments, J. Membr. Sci. 281 (2006) 7087.

    [17] S.K. Hong, M. Elimelech, Chemical and physical aspects of natural organicmatter (NOM) fouling of nanofiltration membranes, J. Membr. Sci. 132 (1997)159181.

    [18] S. Lee,M. Elimelech,Relating organic fouling of reverseosmosismembranes tointermolecular adhesion forces, Environ. Sci. Technol. 40 (200 6) 980987.

    [19] Q. Li, M. Elimelech, Organic fouling and chemical cleaning of nanofiltrationmembranes: measurements and mechanisms, Environ. Sci. Technol. 38 (2004)46834693.

    [20] A. Seidel, M. Elimelech, Coupling between chemical and physical interactionsin natural organic matter (NOM) fouling of nanofiltration membranes: impli-cations for fouling control, J. Membr. Sci. 203 (2002) 245255.

    [21] C.Y. Tang, Q.S. Fu, C.S. Criddle, J.O. Leckie, Effect of flux (transmembrane pres-sure) and membrane properties on fouling and rejection of reverse osmosisand nanofiltration membranes treating perfluorooctane sulfonate containing

    wastewater, Environ. Sci. Technol. 41 (2007) 20082014.[22] E.M. Vrijenhoek,S.K. Hong, M. Elimelech, Influenceof membrane surfaceprop-erties on initial rate of colloidal fouling of reverse osmosis and nanofiltrationmembranes, J. Membr. Sci. 188 (2001) 115128.

    [23] W. Ang, M. Elimelech, Protein (BSA) fouling of reverse osmosis membranes:implications for wastewater reclamation, J. Membr. Sci. 296 (2007) 8392.

    [24] A.R.Costa, M.N.de Pinho,M. Elimelech,Mechanismsof colloidalnaturalorganicmatter fouling in ultrafiltration, J. Membr. Sci. 281 (2006) 716725.

    [25] H.H. Tonnesen, J. Karlsen, Alginate in drug delivery systems, Drug Dev. Ind.Pharm. 28 (2002) 621630.