Performance of Thin-Layer Hydrogel Polyethersulfone Composite

May 9, 2012 - Membranes during Dead-End Ultrafiltration of Various Protein. Solutions ... proteins during microfiltration (MF)7,8 and UF.5,9−11 It h...
0 downloads 0 Views 3MB Size
Article pubs.acs.org/IECR

Performance of Thin-Layer Hydrogel Polyethersulfone Composite Membranes during Dead-End Ultrafiltration of Various Protein Solutions Polina D. Peeva, Thomas Knoche, Thorsten Pieper, and Mathias Ulbricht* Lehrstuhl für Technische Chemie II, Universität Duisburg-Essen, Universitätsstr. 7, 45117 Germany S Supporting Information *

ABSTRACT: Low-fouling thin-layer hydrogel composite membranes were prepared during UV initiated grafting from of the hydrophilic monomer poly(ethylene glycol) methacrylate onto polyethersulfone (PES) ultrafiltration (UF) membranes. The selectivity of the functionalized membranes was adjusted by varying the UV irradiation dose and applying the cross-linking agent N,N′-methylene bisacrylamide. Virgin and composite membranes were tested in short (20-fold volume reduction) and long (24 h) dead-end (DE) filtration experiments of various protein solutions and the performance improvement by the membrane hydrophilization was evaluated. Moreover, the effects of membrane molecular weight cutoff, solute size, and solute charge (as function of pH) as well as cleaning were evaluated. The dominating fouling mechanisms were identified using the classical model equation proposed by Hermia [Hermia, J. Constant Pressure Blocking Filtration LawsApplication to Power Law NonNewtonian Fluids. Trans. Inst. Chem. Eng. 1982, 60, 183] for DE filtration mode. The results showed that the surface functionalization improved the membrane performance during filtration of protein solutions. Moreover, the cleanability of functionalized membranes with water was much more effective compared to unmodified PES membranes. The performed fouling mechanism study clarified the occurring processes during filtrations with virgin and composite membranes.

1. INTRODUCTION Membranes for ultrafiltration (UF) are widely used in the industry for concentration or fractionation of valuable solutions. The main problem during such processes is the flux decline and loss of selectivity due to the deposition of solutes in the membranes surface and within the pores. Flux decline in membrane filtration is caused by the increase in membrane resistance due to the formation of a concentration polarization (CP) layer and the development of another resistance layer by deposition of feed solutes (mainly in terms of pore blockage and subsequent cake formation). Strong flux decline is attributed to the quick blocking of membrane pores by the retained molecules/particles.1 The subsequent cake formation and growth decrease the permeate flux further. In the case of dead-end (DE) filtration mode (closed system), the permeate flux may tend to zero as a result of the increasing osmotic pressure.2 Consequently, filtration processes become expensive because of the short lifetime of the membranes, the necessity of frequent cleanings, modest fluxes, and the extra energy required for circulating the solution in an attempt to control fouling.3 Fouling can occur in two ways: adsorption of foulant (irreversible, cannot be removed by physical cleaning) and cake formation (generally reversible by water washing or back flush).4 In the early stage of the filtration process, hydrophobic solute/particle−membrane interactions govern the fouling behavior (deposition on the membrane surface), in later stages solute/particle−solute interaction determine the membrane performance (interactions of the bulk solutes with the deposited).5,6 The larger and stronger the fouling, the more intense is the required cleaning. The wide membrane application in biotechnology and water purification increased the interest in understanding and © 2012 American Chemical Society

controlling fouling. Many authors studied the fouling with proteins during microfiltration (MF)7,8 and UF.5,9−11 It has been found that the fouling mechanisms are governed by the solute/particle properties, solute−solute interactions, the membrane characteristics, solute−membrane interactions, and the hydrodynamic conditions during filtration. The relationship between pore size and solute size plays an important role. From this relationship, the dominating fouling mechanism can be estimated. Here, three cases of fouling can be distinguished.12,13 I. When the membrane pores are much larger than the solute/particle (dsolute ≪ dpore), pore narrowing will occur, governed by adsorption. The CP can increase the driving force for adsorption. It has been discussed that pore narrowing and subsequent pore plugging can cause stronger flux decline than outer surface fouling.14 II. Pore blockage occurs when the solute size is similar to the membrane pore size. In this case, the solute can plug the pores leading to a flux reduction. III. If high solute retention is provided by the membrane (e.g., dsolute ≫ dpore), the extent of CP may lead to the formation of gel layer, which can further lead to cake layer formation due to solutes deposition. Membrane properties contribute significantly to the overall membrane performance during filtration. Important characteristics are the molecular weight cutoff (MWCO), hydrophilicity, surface charge, and morphology (pore structure, porosity, and roughness) of the membranes. The effects of hydrophilicity and Received: Revised: Accepted: Published: 7231

December 9, 2011 May 6, 2012 May 9, 2012 May 9, 2012 dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

Table 1. Overview of the Used Membranes, Their Basic Characteristics, and the Operating Parameters in Long Dead-End Filtrations membrane virgin

PES 50 PEGMA/MBAA

PES 50a PES 30a PES 10a 40/0 5 J/cm2 a 40/1 5 J/cm2 a 40/4 5 J/cm2 a 40/0 8 J/cm2

water permeability [L/h m2 bar]35

MWCO with dextran [kDa]35

apparent pore diam. from MWCOb (nominal pore diam.c) [nm]

± ± ± ± ± ± ±

95 90 42 30 28 22 32

12.2 (15c) 11.8 (11c) 7.8 (8.5c) 6.4 6.2 5.4 6.7

500 200 100 200 130 60 100

53 22 13 25 15 7 11

appl. pressure [bar] 0.2 0.5 1 0.5 0.77 1.7 1

± ± ± ± ± ± ±

0.02 0.06 0.13 0.06 0.08 0.18 0.01

These membranes also were used in short dead-end filtrations with 20-fold volume reduction at 1 bar and pH = 6. bApparent pore size diameter calculated from the evaluated MWCO values.35 cNominal pore diameter of virgin membranes given by the membrane producer.36 a

PEGs are fouling resistant to proteins is their kosmotropic properties.33 Hence, thin-film composite membranes with grafted antifouling layers made of such kosmotropic polymer hydrogels are very promising materials in the field of membrane separations in aquatic systems. When a hydrogel layer is applied on the membrane surface (and in the pore openings), the membrane surface is shielded from the foulants, so that they cannot reach the hydrophobic material and deposit. Susanto and Ulbricht30 showed that the prepared PES-based composite UF membranes with grafted thin hydrogel layers from kosmotropic polymers exhibited very high fouling resistance to myoglobin during static adsorption and DE filtration experiments over 2 h. In this study, thin-layer hydrogel composite membranes were prepared via UV-initiated “grafting” from copolymerization of poly(ethylene glycol) methacrylate (PEGMA) onto PES UF membranes. The thin-layer architecture was adjusted using the cross-linking monomer N,N′-methylene bisacrylamide (MBAA), affecting the hydrogel mesh size (which influenced the membrane rejection). The performance of these new membranes was tested in DE filtration of various protein solutions. Based on the obtained results from short time DE filtrations, the aim of this work was to evaluate the performance improvement due to the applied hydrogel and the effects of membrane pore size (MWCO), solute size and charge, and pH on the membranes fouling and rejection behavior during 24 h long DE filtration experiments. The occurring fouling processes were evaluated using the classical model equation proposed by Hermia.1 Furthermore, cleaning tests delivered information about the reversibility of the fouling.

surface charge have been comprehensively studied. Hydrophobic materials tend to be fouled preferentially by many biocompounds (e.g., proteins.15,16) When membranes with varied hydrophilicity have been studied, more hydrophilic membranes showed higher fouling resistance during filtration in aquatic solutions.17−19 Feed properties that influence fouling include solute nature, size, and concentration, solute charge (density), hydrophobicity and functional groups, pH, and ionic strength of the solution. The solute charge is an important characteristic because it determines the solute−solute and membrane−solute interactions. The electrostatic effect has been studied by many authors.9−11,20 It has been shown that if repulsion interactions are present (due to the same charge of membrane and solute), the antifouling behavior can be improved. In general, attraction and repulsion forces between membrane surface and solute can govern the membrane selectivity; that is, transmission will be enhanced for solutes that are oppositely charged compared to the membrane.20,21 Furthermore, an understanding of fouling causes and mechanisms is necessary to control fouling more efficiently. Fouling mechanisms have been widely studied.22−24 To describe the dominating fouling mechanisms during filtration, several models have been proposed, such as the standard pore blocking model,1 the combined pore blocking-cake formation model,25,26 the unifying model for CP, gel-layer formation and particle deposition,27 etc. Cleaning the membrane can remove the deposited layer and increase membrane flux. Mechanical and chemical cleaning steps are often applied. Mechanical cleaning procedures are membrane outer surface rinsing and back wash with water. Here, only the reversible part of fouling can be removed. To increase the effectiveness of cleaning, chemicals are applied. Disadvantages of these processes are the necessity of interrupting the filtration process to perform the cleaning and the reduced membrane lifetime by the impact of the chemicals. Most of the membranes for MF and UF for industrial applications are prepared from polyethersulfone (PES), polysulfone, and polyvinylidene fluoride. Due to their rather hydrophobic characters and their consequent strong fouling tendency, the materials are usually modified via bulk polymer modification28 and surface modification.29,30 Increasing the membrane hydrophilicity at the surface has been shown to decrease fouling during protein filtration.28,29,31 The most studied modifier is poly(ethylene glycol) (PEG), which is a charge-neutral polymer that can interact with water via hydrogen bonds creating an energetic barrier to the adsorption of biomolecules at the membrane surface.32 The reason why

2. EXPERIMENTAL SECTION 2.1. Materials. PES UF membranes with nominal MWCO of 10, 30, and 50 kDa from Sartorius-Stedim Biotech GmbH, Germany, were used in this study. Prior to use, the membranes were washed in ethanol (p.a., VWR Germany) for 30 min and kept in deionized water (Milli-Q system from Millipore, U.S.A.) for equilibration. 50 kDa membranes were modified with 40 g/ L PEGMA 400 from Polysciences Inc., U.S.A. (the number indicates the MW of the PEG residue in g/mol) and varied amounts of MBAA as cross-linking agent (Sigma-Aldrich, U.S.A.). Based on the results from a previous work that membrane modification with 5 J/cm2 was sufficient to obtain good low-fouling properties in short DE filtration of proteins, the membranes were irradiated with at least 5 J/cm2 at 5 mW/ cm2 UV intensity using the UV irradiation system UVA Cube 2000, Hönle AG, Germany, equipped with a 20 cm long 7232

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

for performing short DE filtrations with 20-fold volume reduction in order to characterize the effect of feed concentration on membrane fouling (flux) and rejection. The Amicon cell was filled with 10 mL of test solution, and the filtration was performed until 9.5 mL permeate were collected. A balance was connected to a computer and the mass of permeate was taken every 30 s so that the permeate flux over time could be calculated. The volume reduction was calculated as follows (eq 2):

mercury lamp. Detailed information about the functionalization process was already given in previous paper.34 An overview of the used virgin and modified membranes, amounts of MBAA, UV irradiation dose, membrane water permeability, MWCO, apparent pore diameter calculated from MWCO (with dextran) obtained in previous study,35 and nominal pore diameter of virgin membranes given by the producer36 is presented in Table 1. Modified membranes are labeled as 40/x y J/cm2, where 40 is the concentration of PEGMA, x is the concentration of the cross-linker MBAA in the modifier solution, and y is the applied UV irradiation dose (cf. Table 1). The membranes were tested in DE filtrations of the proteins myoglobin (Sigma-Aldrich, U.S.A.), bovine serum albumin (BSA; type HV from Gerbu Biotechnik GmbH, Germany), and their equimolar mixture as model systems for processes in the biotechnology. The basic characteristics of the test systems (molecular weight (MW), isoelectric point (IEP), concentration (c), and pH during the performed experiments) are summarized in Table 2.

volume reduction =

solute myoglobin BSA equimolar mixture a

MW [kg/mol]

IEP

17 67

7 ∼5

long dead-end

c [g/L]

pH

c [g/L]

pH

1 1 1

6 6 6

0.1 0.1 0.1

4; 6; 8 4; 6; 8 4; 6; 8

(Vfeed,0/Vpermeate)

, [−] (2)

where Vfeed,0 is the initial feed volume and Vpermeate is the volume of collected permeate. After the process was stopped, samples from the collected permeate were taken and analyzed via UV−vis measurement using the Cary 50 Probe UV−visible spectrophotometer from Varian Inc., U.S.A., and via total organic carbon (TOC) measurement using the TOC-Vcpn638 equipment from Shimadzu-Siemens. Myoglobin was detected at 509 nm wavelength, while single BSA was measured at 280 nm. In order to obtain the amount of myoglobin in the protein mixture, the myoglobin concentration measured via UV−vis was subtracted from the total protein concentration from TOC. The apparent rejection was calculated using eq 3:

Table 2. Characteristics of the Used Test Solutions short dead-enda

Vfeed,0

R = (1 − c p/c f ) × 100, [%]

(3)

where cf and cp are the protein concentrations in the feed and permeate, respectively. 2.2.4. Long-Term Dead-End Filtrations. Long-term DE filtrations were done with 450 mL test solution using the same setup as that described in section 2.2.2. The filtration process was started at an initial flux of 100 L/hm2 for all tested membranes by adjusting the operation pressure (cf. Table 1). This allowed the direct comparison of the membrane productivity. The mass of permeate was taken every 30 s for period of 24 h, and the permeate flux over time was calculated. After the process was stopped, samples from permeate and feed were taken and their concentration was measured by TOC and UV−vis, as described in section 2.2.3, in order to calculate the apparent total solute rejection (eq 3). The relative flux at the end of the filtration process compared to the initial water flux was calculated (eq 4):

20-fold volume reduction.

All solutions were prepared in 0.01 M phosphate buffer in ultrapure water (consisting of KH2PO4 and Na2HPO4·2H2O, from Fluka Chemie AG, Germany) and filtered through 0.22 μm cellulose membrane filter (Millipore, U.S.A.) before use. NaOH (1 M solution from Waldeck, Germany) was used during membrane cleaning tests. 2.2. Methods. 2.2.1. Dynamic Light Scattering. The particle size distribution (PSD) of the proteins in the solutions at varied pH was analyzed via dynamic light scattering (DLS) measurement with the Zetasizer equipment from Malvern, Germany. The results were expressed in hydrodynamic diameter distribution by numbers. 2.2.2. Water Permeability. Prior to the tests, the membranes were compacted by filtering deionized water at high pressure (3 bar) for at least 30 min using an Amicon cell 8010 from Millipore, U.S.A. equipped with additional feed reservoir (400 mL). The water permeability measurements were done under stirring conditions (300 min−1) at room temperature, pressurizing the water with nitrogen. The mass of permeate collected for 5 min was measured using a balance, and the permeability was calculated according to eq 1: m Lp = , [L/h m 2 bar] ρtAp (1)

relative flux =

Lp,end Lp,start

, [−] (4)

Lp,start and Lp,end are the water permeability at the start and the end of the filtration, respectively. 2.2.5. Fouling Mechanisms. In the following, the classical model equation proposed by Hermia1 for DE filtration mode will be described (eq 5: ⎛ dt ⎞ n d2t ⎜ ⎟ k = ⎝ dV ⎠ dV 2

where m is the mass of permeate, with density ρ, collected for time t through membrane area A at pressure p. Only membranes with a flux deviation within 15% relative to the average values for the respective membrane type were used for further characterization. 2.2.3. Short 20-Fold Volume Reduction Dead-End Filtrations. The same setup as that described above was used

(5)

where t is the time, V is the permeate volume, k is a fouling coefficient, and n is a dimensionless filtration constant representing the filtration mode. The value of n = 0 corresponds to cake formation, n = 1 to intermediate blocking, n = 1.5 to pore constriction, and n = 2 indicates the pore blocking regime.1,25 The differentials are defined as follows (eqs 6 and 7): 7233

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

Figure 1. Effect of pH on the hydrodynamic diameter of BSA, myoglobin, and BSA/myoglobin mixture; (a) pH = 4; (b) pH = 6; (c) pH = 8.

Figure 2. Effect of the volume reduction during dead-end filtration with virgin and composite membranes: (a) permeate flux of virgin PES 30 and PES 50 40/0 5 J/cm2 during 20-fold volume reduction in DE filtration of BSA; (b) apparent rejection of myoglobin as single solute in 2-fold and 20fold volume reduction filtrations.

dt 1 = dV JA d2t 1 dJ =− 3 2 2 dV J A dt

flux recovery =

(6)

Lp,actual Lp,0

× 100, [%] (8)

Lp,0 and Lp,actual are the initial water permeability and the water permeability after every cleaning step, respectively.

(7)

3. RESULTS AND DISCUSSION 3.1. Effect of the pH on the Protein Characteristics. Due to the fact that BSA has its IEP at pH ∼ 537 and myoglobin at pH = 7,38 their net charges vary in the performed experiments (with pH variation). It is expected that at pH = 4 and pH = 8 both proteins have the same charge, both are positive or negative, respectively. At pH = 6, BSA is negatively charged, whereas myoglobin shows positive charge. Thus, attraction forces between the oppositely charged proteins can occur, which would lead to the presence of agglomerates in the protein mixture. The results from DLS measurements (Figure 1) confirm this expectation. At pH = 4 (Figure 1a) and pH = 8 (Figure 1c), myoglobin has the smallest hydrodynamic diameter and BSA has the largest, whereas the mixture is characterized by a medium hydrodynamic diameter. The apparent molecule size in mixture can be explained by contributions of both proteins to light scattering during the measurement (since the hydrodynamic diameter distributions of the protein fractions are relatively close, the result for the protein mixture resembles an overlapping). In contrast, at pH = 6, the protein mixture

The flux change over time dJ/dt was directly calculated from the collected flux data. It should be noticed that this model has been developed by assuming uniform, nonconnected pores,1 which is not the case with PES UF membranes. 2.2.6. Cleaning. Cleaning tests were performed after longterm DE filtrations (cf. section 2.2.4). The effect of mechanical (external washing and back wash with water) and chemical cleaning (with NaOH at pH = 13 and room temperature (21 °C)) on the membrane water flux was studied. After the filtration procedure, the protein solution was removed from the filtration system and the membrane water flux was measured. Thereafter, the membrane surface was cleaned by filling the Amicon cell with water and stirring for 30 s twice; the water flux was then determined again. Subsequently, the membrane was turned with the skin side down and water was filtered for 10 min at 1 bar. The membrane was placed in its initial position to measure the water flux again. At last, NaOH solution (pH = 13) was filtered for 10 min at 1 bar and water flux was measured once again. The flux recovery from every cleaning step was calculated by comparing the actual water flux to the initial using eq 8: 7234

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

Figure 3. Effect of the nominal MWCO of virgin membranes on the permeate flux, rejection, and the fouling mechanism during dead-end filtration at pH = 6: (a) BSA; (b) myoglobin.

low rejection of these two membranes. In this case, CP is less pronounced, and therefore, it affects the solute permeation to a lower extent. 3.3. Long-Term Dead-End Filtrations. In 24 h long DE filtration experiments, virgin and composite membranes were tested. The aim was the evaluation of the effect of membrane MWCO, solute size, and pH value (with influence onto electrostatic interactions) on the membrane performance. The performance improvement by functionalized membranes will be shown in terms of permeate flux and cleanability. Here, only data that are representative for the all collected results will be presented. The relative fluxes and apparent rejections for all tested membranes are summarized in Table 1 in Supporting Information. 3.3.1. Effect of the Membranes MWCO and Solute Size on Filtration Flux and Fouling Mechanism. The effect of the ratio MWCO/solute size will be presented and discussed. First, the results for virgin membranes will be explained; thereafter, a comparison of virgin and functionalized membranes with similar characteristics (water permeability and MWCO) will be done. Virgin Membranes. The obtained results from permeate flux during 24 h and apparent total rejection of BSA and myoglobin as single solutes in solution during filtration through virgin membranes with varied MWCO at pH = 6 are presented in Figure 3. The figure also contains results from the performed fouling mechanism study on the basis of the classical model equation proposed by Hermia.1 In Figure 3a, the data from filtration of BSA are presented. Strong flux decline was measured as a result of the BSA deposition on the membrane surface (very strong flux decrease in the first 1 h) and the further cake layer growth (further decrease in permeate flux). Obviously, the smaller the pores are, the stronger the flux decline is. To describe the processes occurring during filtration, the data from the fouling mechanism study can be discussed. In the insert graphic, it can be seen that the initial slope of the curves is n > 2, which does not correspond to the proposed model. In the literature, cases of n > 2 have been reported.26,39 This observation results from the resistance of the support structure of the studied composite membrane causing fluid to flow through the open pores, leading to more rapid pore blockage.40 The tested anisotropic PES membranes here may exhibit similar behavior to the described composite membranes. Furthermore, it can be seen that the initial slope n of the curves increases with

exhibits an apparent hydrodynamic diameter similar to BSA. The reason could be the agglomeration of the oppositely charged protein molecules as a result of electrostatic attraction. Nevertheless, an increase in the hydrodynamic diameter for the mixture was not detected by DLS. If agglomeration of oppositely charged molecules is assumed, the net charge of the agglomerate would differ from the charges of both proteins (expected to be less negatively charged compared to BSA). This will probably lead to another shape and hydrodynamic size, which can be similar to the size of BSA. It should also be considered that, in mixtures, there is dynamic equilibrium between agglomerated and single proteins. At pH = 6, this equilibrium is shifted to the agglomerate state, while at pH = 4 and pH = 8 it is shifted more to the unimer state. 3.2. Short 20-Fold Volume Reduction Dead-End Filtrations. For high solute rejection, the volume reduction leads to an increase in the feed concentration. The increase in feed concentration may result in stronger CP and fouling or cause enhanced solute transport through the membrane. Selected results for permeate flux of BSA during volume reduction and the comparison of the myoglobin rejection in 2fold and 20-fold volume reduction experiments are presented in Figure 2. Figure 2a shows more stable flux for the concentration of BSA for functionalized PES 50 40/0 5 J/cm2 compared to virgin PES 30 as a result of the weak interactions between BSA and the hydrogel. Here, it can be seen that the applied membrane functionalization prevents the solute deposition on the membrane surface resulting in higher flux. In Figure 2b, the rejection of myoglobin from DE filtrations with 2-fold and 20-fold volume reduction in DE filtration is presented. The data from 2-fold volume reduction DE filtration tests were already published.34 The functionalized membranes were irradiated with 5 J/cm2. The rejection values increased in accordance to the decreasing nominal MWCO of virgin membranes and the increasing MBAA amount used for preparation of the composite membranes. For membranes exhibiting high myoglobin rejection (virgin PES 30, virgin PES 10, PES 50 40/1, and PES 50 40/4), lower rejection of myoglobin was observed during 20-fold volume reduction in comparison to 2-fold reduction, caused by the higher CP.3 Due to the strong accumulation of solute inside the membrane, pore narrowing may be the reason for the obtained opposite effect with virgin PES 50. Membrane PES 50 40/0, modified without a cross-linker, exhibited similar behavior. The reason can be the 7235

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

Figure 4. Effect of the membrane pore size on solutes deposition during filtration at pH = 6: (a) BSA; (b) myoglobin.

increasing nominal MWCO (The first data point was taken 30 s after beginning the filtration. For PES 10 and PES 30 it seems that already during this time the transition to cake filtration is reached; that is, the pore blocking regime characterized by fast flux decline is not detected, since that process occurs very fast). From here, it can be inferred that the flux decline at the initial filtration stages is stronger and faster with decreasing pore size. After the plateau is reached, slopes of n < 0 are obtained. n is more negative for membranes with smaller nominal MWCO, which means that the transition to cake formation regime occurs faster for these membranes. Different is the situation when the solute is not completely rejected by the membrane. In this case, if membranes with different pore sizes, that is, different solute rejection, are compared, the fouling mechanisms will change with changing pore size (MWCO). In the following, the observed data from the filtration of myoglobin through membranes with varied nominal MWCO will be discussed. Figure 3b presents the permeate fluxes through these membranes and the total rejection during DE filtration (samples for rejection measurements were taken from the collected permeate after the filtration was finished, cf. section 2.2.3). The flux curves behave differently; that is, there is strong flux decline for all tested membranes at the beginning of the process, but there is different flux behavior in the later stages. PES 50 showed the strongest flux decline, while PES 30 exhibited the highest permeate flux. The initial slopes of the curves from the fouling mechanism analysis indicate the faster pore narrowing of PES 30 compared to that of PES 50. Here, the fouling mechanisms are the key for the explanation. In case of dsolute ≪ dpore (cf. section 1), pore narrowing occurs. The pores of PES 30 are narrowed by the deposits in the pores and may undergo complete blocking, whereas PES 50 also undergoes pore narrowing, but the pores may be not closed (cf. rejection data during cross-flow (CF) filtration presented in ref 41). The rejection data presented here are apparent total values for the respective batches. At the beginning of the filtration process, high amounts of myoglobin may pass the membrane combined with high flux, that is, a high amount of permeate. In the later stages, when the rejection is expected to decrease due to the narrowed and/or blocked pores, lower flux is measured; that is, the permeate is diluted but not to such extent that the total rejection is affected. Since the pores of PES

50 are not completely blocked, its inner pore surface has more contact with myoglobin molecules than those of PES 10 and PES 30. At pH = 6, myoglobin is attracted to the membrane surface, and in combination with the previous considerations, this can result in lower flux through PES 50, which is more permeable for solutes. The positive charge of myoglobin facilitates the solutes’ transport through the membrane. PES 10 ends up with higher permeate flux than PES 50 as a result of the presence of only outer surface fouling (complete rejection of myoglobin is observed during CF filtration.41 To clarify the presented results, Figure 4 shows schematically the buildup of fouling deposits of big (BSA, Figure 4a) and small molecules (myoglobin, Figure 4b) depending on the membrane pore size. When the solute is completely rejected by the membranes (Figure 4a), the CP and cake layers that are built on the top of the membranes would have similar characteristics, even on membranes that have variable pore sizes but still exhibit complete rejection. Thus, the membrane pore size/porosity, that is, the membrane hydraulic resistance, will determine the flux. This explains the stronger flux decline for membranes with smaller pores at pH = 6. In contrast, at pH = 8, PES 50 exhibits lower flux than PES 10 (cf. Table 1 in the Supporting Information). If the solute molecules have stronger charge, for example, at pH = 8, the layers on the top of the membrane would be less dense, meaning that the flux will also increase. Indeed, the permeate fluxes of both membranes increase compared to experiments at pH = 6 but to various extents. Here, the balance between electrostatic repulsion and hydrophobic interactions is of great importance, that is, the protein deposition on the membrane surface caused by hydrophobic interaction should be also taken into account. For smaller pores, the cake layer porosity would affect the flux, whereas pore plugging of membranes with higher nominal MWCO (the pore size of PES 50 is similar to the size of BSA) would still play a role in combination with the cake layer properties. Consequently, this would lead to more reduced flux of PES 50 compared to membranes with smaller pore size. Similar results have been found by Martin et al.42 The lower values of relative flux for BSA solutions at pH = 4 with increasing membrane pore size can be explained by the membrane−solute interactions. BSA is attracted by the membrane surface due to the opposite charge, causing protein 7236

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

Table 3. Comparison of the Relative Permeate Flux after 24 h Dead-End Filtration of Virgin and Modified Membranes with Similar Water Flux and MWCO: An Overview BSA membrane pair Lp [L/h m2 bar] MWCO [kDa] membrane pair Lp [L/h m2 bar] MWCO [kDa] membrane pair Lp [L/h m2 bar] MWCO [kDa] a

myoglobin

mixture

virgin

functionalized

pH

virgin

functionalized

virgin

functionalized

virgin

functionalized

PES 30 200 ± 22 90 PES 10 100 ± 13 42 PES 10 100 ± 13 42

PES 50 40/0 5 J/cm2 200 ± 25 30 PES 50 40/4 5 J/cm2 60 ± 7 22 PES 50 40/0 8 J/cm2 100 ± 11 32

8 6 4 8 6 4 8 4

0.32 0.25 0.20 0.31 0.20 n.d. 0.31 0.30

0.53 0.51 0.61 0.53 0.45

0.24 0.25 n.d. 0.20 0.21 0.16 0.20 0.16

0.55 0.39

0.30 0.20 n.d.a n.d. 0.20 n.d. n.d. n.d.

0.69 0.40

0.70 0.90

0.51 0.32 0.50 0.66 0.38

0.42

n.d.: at least one of the experiments not done.

Figure 5. Comparison of permeate flux, rejection, and fouling mechanism during dead-end filtrations of virgin PES 30 and PES 50 modified with PEGMA only (40/0) and 5 J/cm2 at pH = 6: (a) BSA; (b) myoglobin.

deposition. If dsolute ∼ dpore, pore blocking occurs, resulting in stronger disruption of the flux rather than by deposition (dsolute ≫ dpore). As already mentioned, when solutes with smaller size are filtered (Figure 4b), the fouling mechanism would change with the varied solute rejection. In case of relatively large pores, fouling occurs on the outer membrane surface, as well as in the pore openings, leading to pore narrowing and also pore blocking. Thus, both fouling mechanisms would affect the permeate flux. When the pores are small enough so that there is minimum permeation of solute, mainly outer surface fouling may be governing the membrane performance. At pH = 8, the flux decline increases with decreasing pore sizes (cf. relative flux in Table 1, Supporting Information). At this pH, myoglobin is repelled by the membrane. Thus, the rejection of myoglobin increases for all membrane nominal MWCOs. First, the increase in rejection leads to increased CP near the membrane surface, i.e., the higher the rejection the more strongly the permeate flux would be reduced by the CP layer. Second, since myoglobin is not attracted by the membrane surface, no strong interactions will be present in the membrane interior. Thus, PES 10 has the lowest permeate flux due to the strongest CP. Considering the increased rejection of myoglobin by PES 10 and PES 30 at pH = 8, more pronounced CP could be the reason for the lower permeate fluxes in comparison to pH = 6. In contrast, the rejection by PES 50 is also higher at pH = 8, but still 40%. At pH = 4, PES 10 and PES 50 behave similarly during filtration of myoglobin. Here, myoglobin exhibits strong

positive charge, which can result in strong solute deposition on the membrane surface. Moreover, at pH = 4, the lowest relative fluxes are measured (see Supporting Information) indicating the strongest fouling. Hence, both membranes exhibit similarly poor performance. At all pH values, the effect of pore size on the flux decline during filtration of equimolar mixture of BSA and myoglobin followed the trend found for single BSA solution. Due to the presence of myoglobin in the mixture, combined fouling can be expected, which results in lower fluxes. Comparison of Virgin and Composite Membranes. Protein solutions were filtered through functionalized membranes. The obtained relative fluxes after 24 h of DE filtration and apparent total rejection values are shown in Table 1 in the Supporting Information. Furthermore, analogous to the results in Figure 3, flux, rejection, and fouling mechanism data from filtration of BSA and myoglobin through membranes modified with 5 J/cm2 and varied amount of MBAA are presented in Figure I (Supporting Information). For evaluation of the membrane performance improvement by functionalized membranes, comparison of virgin and modified membranes with similar basic characteristics should be done. Table 3 contains the obtained relative flux data for three membrane pairs. In each pair, the membranes exhibit similar water permeability and nearly comparable MWCO (exception is the first pair (virgin PES 30 and PES 50 40/0 5 J/ cm2), where the MWCO of the virgin PES 30 is much higher). As it can be seen, in all filtration conditions, functionalized membranes exhibited higher relative fluxes. Comparing the 7237

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

Figure 6. Effect of the hydrogel layer on solutes deposition during filtration at pH = 6: (a) BSA; (b) myoglobin.

consequently, flux decline may be enhanced to such extent that the membrane flux goes to zero. By application of a hydrophilic layer on the membrane surface (and within the pores), the membrane−solute interactions can be reduced because the water structure close to the surface is similar to the bulk water structure and no attraction between surface and solutes exists43 (Figure 6a, right). Nevertheless, CP will still affect the permeate flux. When small solutes reach the membrane surface, depending on the relationship solute size/pore size (dsolute/dpore), deposition in the pore openings and membrane inner surface resulting in pore narrowing as well as pore plugging (as shown in Figure 6b, left) may occur. Further cake layer growth is possible. Depending on the solute permeability of the membrane, CP would influence the flux to varied extent. Since the membrane is modified with hydrophilic polyPEGMA, the strong hydration of the hydrogel reduces the membrane− solute interactions, leading to reduced fouling and enhanced flux (Figure 6b right). Nevertheless, if the hydrogel mesh size is similar or larger relative to the solute size, solutes may be caught in the hydrogel network, and this would affect the membrane flux and selectivity (e.g., in Table 3, PES 50 40/0 5 J/cm2 (apparent pore diameter = 6.4 nm, cf. Table 1) showed lower relative flux after DE filtration of myoglobin (median molecule diameter ∼4 nm, cf. Figure 1) compared to PES 50 40/4 5 J/cm2 (apparent pore diameter = 5.4 nm, cf. Table 1)). Therefore, the hydrogel network has to be well adapted to the solutes to be filtered. CP would influence the permeate flux, since it cannot be affected by the membrane modification (for same rejection). 3.3.2. Effect of pH. In all performed filtration experiments, the tested membranes are negatively charged to varied extents (ζ-potential results for similar membranes have been presented by Susanto et al.44). Thus, BSA attraction to the membrane surface can be predicted at pH = 4, and repulsion at pH = 6 and pH = 8. In contrast, myoglobin will be attracted by the membrane surface at pH = 4 and pH = 6, and repelled at pH = 8. More complicated is the case of filtration experiments with mixtures of both proteins. First, effects of sterical hindrance can occur (BSA is much bigger), which will affect the protein transmission. Moreover, at pH = 6, more complex interactions can be expected. If solute agglomeration occurs partially, myoglobin and BSA will be present in the solution as both unimers and agglomerates, where each species will interact

rejection results for these membranes (summarized in Table 1, Supporting Information), in all cases functionalized membranes showed less rejection than the corresponding virgin membranes. Both facts result from the solute deposition on the virgin membrane surfaces, while functionalized surfaces are shielded by the applied hydrogel layer. Exemplarily, experimental data (permeate flux, total rejection and fouling mechanism analysis) from filtration of BSA and myoglobin through virgin PES 30 and modified PES 50 40/0 5 J/cm2 are shown in Figure 5. It can be seen that the permeate flux of modified PES 50 40/ 0 during filtration of BSA (Figure 5a) is much higher than the flux obtained with virgin PES 30. An initial strong flux drop was not measured with functionalized membranes, which indicates the effective shielding of the membrane surface by the applied hydrogel layer. Nevertheless, the permeate flux decreased as a consequence of the occurring CP effects. From the fouling mechanism analysis, it can be concluded that the transition mode from pore blocking to cake formation (n < 0)1 occurs faster for the functionalized membrane (more negative slope), which leads to lower d2t/dV2 values. The more negative slope of the curve, that is, the fast decrease in d2t/dV2, reflects a large reduction in the rate of flux decline25 leading to higher permeate fluxes compared to virgin membranes. In the case of myoglobin filtration (Figure 5b), the flux improvement by functionalized membranes is less pronounced. In both cases, virgin membranes do not reach steady state of the curve d2t/ dV2 vs t, indicating that the cake filtration regime is not reached during the 24 h of analysis. To describe the principal of (low-)fouling behavior in the upper described cases, a schematic view of the occurring membrane−solute interactions is presented in Figure 6a for BSA and Figure 6b for myoglobin filtration. Here, virgin membranes with relatively small pores are compared with functionalized membranes exhibiting similar characteristics. For ultrafiltration of BSA, there is very high or complete solute rejection. In the case of virgin membranes, hydrophobic interactions govern the foulant deposition on the membrane surface. The reason is the water structure close to hydrophobic surfaces, described as “less-dense”; hence, water molecules close to the surface can be replaced by solutes.43 Further, due to solute−solute interactions, a cake layer is built (Figure 6a, left). The CP deteriorates further the permeate flux decline. Even more, when appropriate conditions are present, fouling and, 7238

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

Figure 7. Effect of the pH on the permeate flux and rejection during dead-end filtration of BSA through (a) virgin PES 50; (b) PES 50 modified with PEGMA only (40/0) and 5 J/cm2.

Repeated exposure to chemical cleaning with NaOH of hydrogel composite membranes without exposure to protein solutions in between showed no change in the membrane flux, and this was taken as an evidence for the stability of the barrier properties imparted by the grafted hydrogel layer under these cleaning conditions. Earlier, the stability of similar PEGMAbased thin-layer hydrogel composite membranes under cleaning conditions had been investigated by ATR-IR spectroscopy, and no changes had been observed over 8 cycles.30 The effect of membrane nominal MWCO (for virgin membranes) and hydrogel cross-linking degree (for composite membranes) on the flux recovery in each step of cleaning after filtration of BSA and myoglobin is presented in Figures III and IV in the Supporting Information. Virgin membranes did not recover flux from cleaning with water significantly; the flux reached only 30−40% of the initial water flux. Here, a slight effect of the membrane nominal MWCO could be found; that is, membranes with smaller MWCO recovered less flux. Chemical cleaning at pH = 13 influenced the flux recovery. At these cleaning conditions, virgin membranes recovered about 60− 70% of their initial water flux. In contrast, cleaning only with water of functionalized membranes increased the membrane flux recovery to nearly 90%. Furthermore, the cross-linking degree affected the cleaning efficiency; that is, higher fluxes were obtained for membranes functionalized with more crosslinker MBAA. In cleaning after BSA filtrations, external cleaning was more effective, whereas after myoglobin filtrations, back wash recovered more flux. Chemical cleaning increased the membrane flux to another 10% recovery flux to nearly 100% for some membranes. Exemplary flux recovery data from cleaning after filtrations of BSA and myoglobin through the membrane pair virgin PES 30 vs PES 50 40/0 5 J/cm2 are compared in Figure 8. As it can be seen, functionalized membranes increased their flux after cleaning with water, while the effect of such external washing on the flux of virgin membranes was negligible. Chemical cleaning was able to increase the flux of virgin membranes but not to very high extent. In contrast, cleaning at pH = 13 did not further influence the flux recovery of composite membranes significantly because large recovery could be achieved already with water, by external washing for BSA and by backwashing for myoglobin The different behavior for the two proteins can also be explained very well. The (almost) completely rejected larger protein BSA forms a weakly attached fouling layer on the outer membrane surface. The

differently with the membrane surface, that is, attraction (myoglobin), low repulsion (agglomerate), and repulsion (BSA). Results for such systems are also included in the Supporting Information, but this paper focuses on filtration of single compound solutions; filtration of protein mixtures in CF mode are further discussed in another study.41 The influence of solute−membrane interactions due to electrostatic effects will be first explained on the basis of results from filtration of BSA, where very high or complete rejection was observed. Exemplarily, Figure 7 presents the permeate flux and apparent total rejection of BSA for UF through virgin PES 30 (Figure 7a) and PES 50 40/0 5 J/cm2 (Figure 7b) at varied pH values. (Figure II in the Supporting Information shows the permeate flux of myoglobin through these membranes.) The results from all tested membranes are summarized in Table I (Supporting Information). It is easy to see that the permeate flux of BSA solution through virgin PES 30 decreases more strongly when strong attractive solute−membrane interactions are present, that is, with decreasing pH stronger flux decline was measured. The reason is the strong fouling (solutes deposition) of BSA. In contrast, functionalized membranes exhibited more stable permeate flux as a consequence of the reduced solute− membrane interactions by the applied hydrogel layer. Regarding the rejection data for these membranes, when repulsion forces were present (at pH = 8), the rejection of BSA was 100%. If attraction forces dominated (at pH = 4), transport of BSA through the membrane was obtained (the rejection decreased slightly). This effect was more pronounced for the functionalized membranes; this can be explained because fouling of virgin membranes affected the rejection (higher rejection due to fouling). Interesting was the rejection during filtration of myoglobin. From Table I in Supporting Information, it can be taken that virgin PES 30 rejected myoglobin to 76% at pH = 8, to 40% at pH = 6, and to 85% at pH = 4. At pH = 8, repulsion led to the relatively high rejection, while at pH = 4 attraction forces caused strong fouling and consequently also high rejection. Similar results were obtained with virgin PES 10, whereas virgin PES 50 showed still low rejection at pH = 4 possibly due to the larger barrier pore size. Membranes functionalized with 5 J/cm2 exhibited relatively similar myoglobin rejection at the studied pH values. 3.3.3. Cleaning. The reversibility of fouling was tested during mechanical and chemical cleaning experiments. 7239

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

further adaptation or optimization of composite membrane structure and operation conditions. It can be concluded that the membrane hydrophilization did improve the membrane performance during filtration of proteins and cleaning tests. Moreover, the electrostatic interactions, which play an important role during filtration of proteins, became negligible for filtrations with thin-layer hydrogel composite membranes based on neutral kosmotropic polymers such as PEG.



ASSOCIATED CONTENT

S Supporting Information *

Relative flux and rejection of all tested virgin and composite membranes after 24 h of dead-end filtration; effect of the crosslinking degree and pH on the permeate flux, rejection, and fouling mechanism during dead-end filtration with virgin and modified PES 50; effect of cleaning on membrane flux. This material is available free of charge via the Internet at http:// pubs.acs.org.

Figure 8. Comparison of the flux recovery of virgin PES 30 and PES 50 modified with PEGMA only (40/0) after filtrations at pH = 6: (a) BSA; (b) myoglobin.



(partially) permeable smaller protein myoglobin is, to some extent, accumulated within the hydrogel layer; however, because the interactions with the polyPEGMA network are very weak, so that the protein is just mechanically entrapped and the contact with hydrophobic PES is essentially prevented, easy removal by flow reversal is possible. Overall, with the thinlayer hydrogel composite membranes, mechanical cleaning with water under appropriate conditions may be sufficient and used as single cleaning step without applying chemicals. Data about the effect of pH on the flux recovery during cleaning of virgin PES 50 and PES 50 40/0 5 J/cm2 can be found in the Supporting Information (Figures V and VI, respectively). It was found that lower pH during UF caused less flux recovery for virgin membranes, whereas a different effect was found for composite membranes. After the filtration of BSA, cleaning was not influenced by the BSA solution pH. In contrast, myoglobin removal was more successful after UF at lower pH. This effect may be a consequence of the rejection behavior at varied pH combined with possible myoglobin molecule conformation change in the vicinity of pores (since the environment in the membrane surface region and pores is different than in the bulk solution).

AUTHOR INFORMATION

Corresponding Author

*Fax.: 49-201-183 3147. E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS



REFERENCES

The donation of the PES UF membranes by Sartorius-Stedim Biotech GmbH is gratefully acknowledged. PDP would like to thank the Deutsche Bundesstiftung Umwelt for providing her a PhD scholarship.

(1) Hermia, J. Constant Pressure Blocking Filtration Laws Application to Power Law Non-Newtonian Fluids. Trans. Inst. Chem. Eng. 1982, 60, 183. (2) Song, L.; Elimelech, M. Theory of Concentration Polarization in Cross-Flow Filtration. J. Chem. Soc., Faraday Trans. 1995, 91, 3389. (3) Baker, R. W. Membrane Technology and Applications; John Wiley & Sons Ltd.: Chichester, 2004. (4) Asatekin, A.; Kang, S.; Elimelech, M.; Mayes, A. M. Anti-Fouling Ultrafiltration Membranes Containing Polyacrylonitrile-Graft-Poly(Ethylene Oxide) Comb Copolymer Additives. J. Membr. Sci. 2007, 298, 136. (5) Huisman, I. H.; Prádanos, P.; Hernández, A. The Effect of Protein−Protein and Protein−Membrane Interactions on Membrane Fouling in Ultrafiltration. J. Membr. Sci. 2000, 179, 79. (6) Kelly, S. T.; Zydney, A. L. Mechanisms for BSA Fouling During Microfiltration. J. Membr. Sci. 1995, 107, 115. (7) Chan, R.; Chen, V. The Effects of Electrolyte Concentration and pH on Protein Aggregation and Deposition: Critical Flux and Constant Flux Membrane Filtration. J. Membr. Sci. 2001, 185, 177. (8) Velasco, C.; Ouammou, M.; Calvo, J. I.; Hernández, A. Protein Fouling in Microfiltration: Deposition Mechanism as a Function of Pressure for Different pH. J. Colloid Interface Sci. 2003, 266, 148. (9) Fane, A. G.; Fell, C. J. D.; Suki, A. The Effect of pH and Ionic Environment on the Ultrafiltration of Protein Solutions with Retentive Membranes. J. Membr. Sci. 1983, 16, 195. (10) Rohani, M. M.; Zydney, A. L. Role of Electrostatic Interactions During Protein Ultrafiltration. Adv. Colloids Interface Sci. 2010, 160, 40. (11) Salgin, S. Effects of Ionic Environments on Bovine Serum Albumin Fouling in a Cross-Flow Ultrafiltration System. Chem. Eng. Technol. 2007, 30, 255.

4. CONCLUSIONS Low-fouling thin-layer composite membranes were prepared via photografting PEGMA onto PES UF membranes. During UF at varied conditions, it was found that the composite membranes behaved much more stable with better performance than the unmodified PES membranes. From long-term DE filtration tests, the main fouling mechanisms and the effect of membrane MWCO and solute size were evaluated. Pore narrowing and further pore plugging and cake formation were obtained in cases where dsolute ≪ dpore and the processes occurred faster with decreasing the MWCO. Comparing virgin and functionalized membranes with similar basic characteristics, different fouling behavior was found indicating better performance of functionalized membranes. Electrostatic interactions were important during filtrations through virgin membranes, while for functionalized membranes, when high rejection was obtained, the pH (and hence changed electrostatics) did not influence the membrane performance. Cleaning with water was efficient only for the hydrogel composite membranes; it can be expected that cleaning without or with much less chemicals may be used with satisfying effects. Of course, this will require 7240

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241

Industrial & Engineering Chemistry Research

Article

(12) Bacchin, P.; Aimar, P.; Field, R. W. Critical and Sustainable Fluxes: Theory, Experiments, and Applications. J. Membr. Sci. 2006, 281, 42. (13) Belfort, G.; Pimbley, J. M.; Greiner, A.; Chung, K. Y. Diagnosis of Membrane Fouling Using a Rotating Annular Filter. 1. Cell Culture Media. J. Membr. Sci. 1993, 77, 1. (14) Marshall, A. D.; Munro, P. A.; Trägårdh, G. The Effect of Protein Fouling in Microfiltration and Ultrafiltration on Permeate Flux, Protein Retention, and Selectivity: A Literature Review. Desalination 1993, 91, 65. (15) Koehler, J. A.; Ulbricht, M.; Belfort, G. Intermolecular Forces between a Protein and a Hydrophilic Modified Polysulfone Film with Relevance to Filtration. Langmuir 2000, 16, 10419. (16) Möckel, D.; Staude, E.; Guiver, M. D. Static Protein Adsorption, Ultrafiltration Behavior, and Cleanability of Hydrophilized Polysulfone Membranes. J. Membr. Sci. 1999, 158, 63. (17) Ochoa, N. A.; Masuelli, M.; Marchese, J. Effect of Hydrophilicity on Fouling of an Emulsified Oil Wastewater with PVDF/PMMA Membranes. J. Membr. Sci. 2003, 226, 203. (18) Susanto, H.; Franzka, S.; Ulbricht, M. Dextran Fouling of Polyethersulfone Ultrafiltration MembranesCauses, Extent and Consequences. J. Membr. Sci. 2007, 296, 147. (19) Su, Y.; Mu, C.; Li, C.; Jiang, Z. Antifouling Property of a Weak Polyelectrolyte Membrane Based on Poly(Acrylonitrile) During Protein Ultrafiltration. Ind. Eng. Chem. Res. 2009, 48, 3136. (20) Burns, D. B.; Zydney, A. L. Contributions to Electrostatic Interactions on Protein Transport in Membrane Systems. AlChE J. 2001, 47, 1101. (21) Burns, D. B.; Zydney, A. L. Effect of Solution pH on Protein Transport through Ultrafiltration Membranes. Biotechnol. Bioeng. 1999, 64, 27. (22) Costa, A. R.; de Pinho, M. N.; Elimelech, M. Mechanisms of Colloidal Natural Organic Matter Fouling in Ultrafiltration. J. Membr. Sci. 2006, 281, 716. (23) Taniguchi, M.; Kilduff, J. E.; Belfort, G. Modes of Natural Organic Matter Fouling During Ultrafiltration. Environ. Sci. Technol. 2003, 37, 1676. (24) Waite, T. D.; Schäfer, A. I.; Fane, A. G.; Heuer, A. Colloidal Fouling of Ultrafiltration Membranes: Impact of Aggregate Structure and Size. J. Colloid Interface Sci. 1999, 212, 264. (25) Ho, C.-C.; Zydney, A. L. A Combined Pore Blockage and Cake Filtration Model for Protein Fouling During Microfiltration. J. Colloid Interface Sci. 2000, 232, 389. (26) Ho, C.-C.; Zydney, A. L. Protein Fouling of Asymmetric and Composite Microfiltration Membranes. Ind. Eng. Chem. Res. 2001, 40, 1412. (27) Bacchin, P.; Si-Hassen, D.; Starov, V.; Clifton, M. J.; Aimar, P. A Unifying Model for Concentration Polarization, Gel-Layer Formation, and Particle Deposition in Cross-Flow Membrane Filtration of Colloidal Suspensions. Chem. Eng. Sci. 2002, 57, 77. (28) Su, Y.; Liang, Y.; Mu, C.; Jiang, Z. Improved Performance of Poly(Vinylidene Fluoride) Microfiltration Membranes Prepared by Freeze and Immersion Precipitation Coupling Method. Ind. Eng. Chem. Res. 2011, 50, 10525. (29) Sun, M.; Su, Y.; Mu, C.; Jiang, Z. Improved Antifouling Property of Pes Ultrafiltration Membranes Using Additive of Silica−PVP Nanocomposite. Ind. Eng. Chem. Res. 2009, 49, 790. (30) Susanto, H.; Ulbricht, M. Photografted Thin Layer Polymer Hydrogel Layers on PES Ultrafiltration Membranes: Characterization, Stability, and Influence on Separation Performance. Langmuir 2007, 23, 7818. (31) van Reis, R.; Zydney, A. Membrane Separations in Biotechnology. Curr. Opin. Biotechnol. 2001, 12, 208. (32) Asatekin, A.; Menniti, A.; Kang, S.; Elimelech, M.; Morgenroth, E.; Mayes, A. M. Antifouling Nanofiltration Membranes for Membrane Bioreactors from Self-Assembling Graft Copolymers. J. Membr. Sci. 2006, 285, 81.

(33) Kane, R. S.; Deschatelets, P.; Whitesides, G. M. Kosmotropes Form the Basis of Protein-Resistant Surfaces. Langmuir 2003, 19, 2388. (34) Peeva, P. D.; Pieper, T.; Ulbricht, M. Tuning the Ultrafiltration Properties of Anti-Fouling Thin-Layer Hydrogel Polyethersulfone Composite Membranes by Suited Crosslinker Monomers and PhotoGrafting Conditions. J. Membr. Sci. 2010, 362, 560. (35) Peeva, P. D.; Million, N.; Ulbricht, M. Factors Affecting the Sieving Behavior of Anti-Fouling Thin-Layer Cross-Linked Hydrogel Polyethersulfone Composite Ultrafiltration Membranes. J. Membr. Sci. 2011, 390−391, 99−112. (36) Sartorius-Stedim Biotech, Göttingen, Personal Correspondence, 2008. (37) Böhme, U.; Scheler, U. Effective Charge of Bovine Serum Albumin Determined by Electrophoresis NMR. Chem. Phys. Lett. 2007, 435, 342. (38) Figueiredo, K.; Ferraz, H.; Borges, C.; Alves, T. Structural Stability of Myoglobin in Organic Media. Protein J. 2009, 28, 224. (39) Zydney, A. L.; Ho, C.-C.; Yuan, W. In New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes, Membrane Science and Technology, 8; Bhattacharyya, D., Butterfield, D. A., Eds.; Elsevier: Amsterdam, 2003. (40) Loh, S.; Beuscher, U.; Poddar, T. K.; Porter, A. G.; Wingard, J. M.; Husson, S. M.; Wickramasinghe, S. R. Interplay among Membrane Properties, Protein Properties and Operating Conditions on Protein Fouling During Normal-Flow Microfiltration. J. Membr. Sci. 2009, 332, 93. (41) Peeva, P. D.; Knoche, T.; Pieper, T.; Ulbricht, M. Cross-Flow Ultrafiltration of Protein Solutions through Unmodified and Surface Functionalized Polyethersulfone MembranesEffect of Process Conditions on Separation Performance. Sep. Purif. Technol. 2011, 92, 83−92. (42) Martín, A.; Martínez, F.; Calvo, J. I.; Prádanos, P.; Palacio, L.; Hernández, A. Protein Adsorption onto an Inorganic Microfiltration Membrane: Solute-Solid Interactions and Surface Coverage. J. Membr. Sci. 2002, 207, 199. (43) Vogler, E. A. Water and the Acute Biological Response to Surfaces. J. Biomater. Sci., Polym. Ed. 1999, 10, 1015. (44) Susanto, H.; Balakrishnan, M.; Ulbricht, M. Via Surface Functionalization by Photograft Copolymerization to Low-Fouling Polyethersulfone-Based Ultrafiltration Membranes. J. Membr. Sci. 2007, 288, 157.

7241

dx.doi.org/10.1021/ie202893e | Ind. Eng. Chem. Res. 2012, 51, 7231−7241