Capillary pore membranes with grafted diblock copolymers showing reversibly changing ultrafiltration properties with independent response to ions and temperature

Journal of Membrane Science 514 (2016) 510–517

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Capillary pore membranes with grafted diblock copolymers showing reversibly changing ultrafiltration properties with independent response to ions and temperature Martyna Gajda a,b, Mathias Ulbricht a,b,n a b

Lehrstuhl für Technische Chemie II, Universität Duisburg-Essen, 45117 Essen, Germany CENIDE – Center for Nanointegration Duisburg-Essen, 47057 Duisburg, Germany

art ic l e i nf o

a b s t r a c t

Article history: Received 19 February 2016 Received in revised form 29 April 2016 Accepted 1 May 2016 Available online 10 May 2016

Track-etched polyethylene terephthalate membranes with a pore size of 110 nm have been modified by a grafted diblock copolymer structure in order to obtain dual-responsive ultrafiltration membranes. The temperature-responsive poly-N-isopropylacrylamide (PNIPAAm) and the ion-responsive poly-N,N-dimethyl-N-methacryloyloxyethyl-N-(3-sulfopropyl) ammonium betaine (PSPE) were grafted from the pore walls by sequential surface-initiated atom transfer radical polymerization. To achieve a well controlled pore functionalization, the polymerization rate had been adjusted to a very slow chain growth. A successful grafting of PSPE as first block and of PNIPAAm as second block could be demonstrated. To study the temperature and ion responsivity, hydraulic permeability and dextran diffusion experiments at 25 °C and 40 °C with different salt solutions had been carried out. It could be shown that such membranes can change their barrier pore size from a more open to a more closed state in dependency of temperature as well as kind of ions and their concentration in the feed. The grafted layer thickness showed a particularly strong increase in the presence of KClO4. Like expected for a well defined polymer brush, an expanded or collapsed state for each of the individual blocks could be obtained. In presence of KClO4 at 25 °C both blocks are expanded and the pores are in a more closed state. If one of these stimuli is changed, for example when temperature is increased to 40 °C or ions are removed, the block copolymer brushes collapse partially what leads to more open pores. However, if both stimuli are applied together, i.e. if temperature is 40 °C and no ions are in solution, diblock copolymers collapse fully what leads to fully opened pores. It had also been demonstrated that this dual responsivity and the magnitude of the effects depend very strongly on the absolute degree of grafting and the ratio between the two different responsive polymer blocks. Furthermore, the ability of changing effective membrane barrier pore size by ions and temperature has been investigated in more detail by dextran diffusion experiments, and very pronounced and reversible changes of molecular sieving curve and molecular weight cut-off could be obtained. & 2016 Elsevier B.V. All rights reserved.

Keywords: Surface functionalization Track-etched membrane Grafted diblock copolymer Stimuli-responsive membrane Ultrafiltration

1. Introduction Stimuli responsive materials, in particular polymers, had gained a lot of interest in recent years due to their unique ability of a large, rapid and reversible change of their properties in response to a small variation of the environmental conditions [1]. This concept had also been applied to obtain stimuli-responsive membranes [2], and already reported materials can be sensitive towards, for example, temperature [3], pH [4], ionic strength [5], n Corresponding author at: Lehrstuhl für Technische Chemie II, Universität Duisburg-Essen, 45117 Essen, Germany. E-mail address: [email protected] (M. Ulbricht).

http://dx.doi.org/10.1016/j.memsci.2016.05.001 0376-7388/& 2016 Elsevier B.V. All rights reserved.

specific ions [6], even larger molecules such as proteins [7], as well as light irradiation [8] or magnetic fields [9]. A commonly used mechanism to obtain responsive barrier properties is based on a reversible swelling/deswelling transition of polymeric hydrogels confined on the membrane surface or in membrane pores. However, many of these membranes comprise not well controllable functional polymer structures because conventional radical polymerization had been used for synthesis or fabrication. Therefore, the full potential of advanced functional polymeric architectures cannot be utilized. For this work, atom transfer radical polymerization (ATRP) had been chosen what allows the easy control of polymer chain length and functionality, and synthesis parameters can be readily adjusted

M. Gajda, M. Ulbricht / Journal of Membrane Science 514 (2016) 510–517

so that it can be optimized to many different polymers and specific demands [10]. Moreover, this technique makes it possible to achieve well-defined macromolecular architectures like diblock- [11,12] or multiblock structures [13,14], with controlled molecular weight and low polydispersity of each block. In the recent decades, this technique had also been used for surface functionalization of materials [10]. Several groups have already reported about the modification of different membrane materials by surface initiated (SI) ATRP for various applications. For instance, Fu et al. have grafted polystyrene derivatives on polyphosphazene membranes leading to novel ion exchange membranes for fuel cell technologies [15]. Singh et al. have prepared high-capacity macroporous membrane adsorbers by grafting of the weak polyelectrolyte poly(acrylic acid) (PAA) on the surface of regenerated cellulose membranes [16]. Other reports had been about improved antifouling properties by grafting of poly(ethylene glycol) (PEG)-containing or polyzwitterionic brushes on membrane surfaces [17,18]. Furthermore, Singh et al. and Wandera et al. reported about modifications of ultrafiltration membranes for simultaneous changing of sieving and improving antifouling properties; the key to success had been using SI-ATRP conditions enabling very slow chain growth [19,20]. Due to their “smart”, tunable permeability and separation properties many groups focused on stimuli-responsive membranes prepared by SI-ATRP. The stimuli-responsive polymers can be introduced into the membranes as blend with the bulk material or grafted on the surface and on the pore walls what leads to a pore size tunable by external conditions [21]. Friebe and Ulbricht had obtained temperature-responsive membranes by grafting poly (N-isopropylacrylamide) (PNIPAAm) layers with a thickness of up to 80 nm on the walls of capillary pores in track-etched polyethylene terephthalate (PET) membranes with pore diameters of about 700 nm, what lead to reversible swelling and deswelling of the layer within the pore, at temperatures below and above the lower critical solution temperature (LCST) of this polymer (32 °C) [22]. Yang and Ulbricht published about analogous PET-based capillary pore membranes, but with grafted poly(N,N-dimethyl-Nmethacryloxyethyl-N-(3-sulfopropyl) ammonium betain) (PSPE) which were able to change the permeate flux in dependency of the ion type and concentration [23]. Li et al. had used PAA for the grafting on nylon membranes what led to pH-responsive membranes [24]. However, mostly membranes with barrier pores larger than 400 nm had been modified. For modification of pores which are smaller than 100 nm, very well controlled polymerization conditions leading to very slow chain growth are required because otherwise the polymer distribution will be uneven and this may even lead to clogged pores. That had been demonstrated by Alem et al. when modifying a track-etched membrane with 80 nm pores with PNIPAAm in an aqueous system. They observed a flux decrease after raising the temperature what might be due to the long polymer chains on the outer surface which build a dense layer onto of the pores when collapsing [25]. Frost and Ulbricht have proven the importance of a very well controlled polymerization by modifying PET track-etched membranes with a pore diameter of 110 nm with PNIPAAm and obtaining well-defined ultrafiltration membranes which reversibly open or close their pores at high or low temperature, respectively [26]. There have been also several reports about grafting of diblock copolymers on microfiltration membrane surface what leads to dual- or even multi-responsive materials with novel properties and possible applications. For example, temperature- and pH-responsive microfiltration membranes had been obtained by grafting of PNIPAAm as a first block and PAA as the second block [27] or by using the reverse sequence [28]. There are also reports about pH- and temperature-responsive membranes by grafting of blocks of poly(diethylene glycol)methyl ether methacrylate) or PNIPAAm, in combination with either poly (methacrylic acid) or poly(N,N-dimethylaminoethyl methacrylate)

511

Fig. 1. Schematic depiction of grafted diblock copolymer brushes with PSPE as first block and PNIPAAm as second block in a cylindrical pore. The block structure leads to four different pore diameters due to four different chain expansion states in dependency of the stimuli salt type and concentration ([S]) and temperature ([T]).

[29,30]. Kumar et al. reported about a membrane for release of dye in response to temperature, pH or exposure to UV light [31]. However, for the preparation of ultrafiltration membranes with grafted diblock structures within the pores a very well controlled polymerization is required, and this had not yet been reported. Besides a multi-step adjustable molecular weight cut-off in the ultrafiltration range or advanced control of macromolecular drug release through a membrane, such materials could also be developed toward logic gates [32]. Here we report about dual-responsive ultrafiltration membranes with a tailored functional polymer structure and architecture and tunable pore size. A track-etched PET membrane has been functionalized by grafting of a diblock copolymer made of PSPE and PNIPAAm, leading to materials where pore size can be reversibly manipulated by ion concentration or kind of ions and temperature like it is shown in Fig. 1. First, SI-ATRP conditions had been adjusted to a very slow polymer chain growth what leads to well controlled polymer brush length and only few termination reactions. This had then been utilized to prepare membranes with well defined grafted diblock copolymer brushes on the pore walls at different overall degree of grafting and block ratios. The responsivity of the membranes towards ions and temperature has finally been carefully evaluated by hydraulic permeability and dextran diffusion experiments at varied solution conditions and temperature. 2. Experimental section 2.1. Materials and chemicals Polyethylene terephthalate (PET) tracked-etched (TE) membranes with a nominal pore diameter of 80 nm, a porosity of 10% and a thickness of 8 mm were purchased from Oxyphen AG (Switzerland). Potassium permanganate (499%), 2-aminoethanol (499%), 4-(dimethylamino) pyridine (498%), triethylamine (499.5%), potassium perchlorate (KClO4;498.5%), and aluminum chloride (AlCl3;499%) were purchased from Fluka. Sulfuric acid (495%), methanol (99.99%), acetonitrile (99.99%), acetone (99.99%), and diethyl ether (99.99%) were from Fisher Scientific. Hydrochloric acid (37%) aqueous solution and N,N-dimethylformamide (DMF;499.8%) were from Bernd Kraft (Germany). N,N-Diisopropylcarbodiimide (497%) and copper(II) bromide

512

M. Gajda, M. Ulbricht / Journal of Membrane Science 514 (2016) 510–517

(CuBr2) (499%) were from Acros Organics. 1-Hydroxybenzotriazole hydrate (497%), α-bromoisobutyryl bromide (498%), copper(I) bromide (CuBr) (498%), 2,2′-bipyridine (Bpy) (499%), [2-(methacryloyloxy)ethyl]dimethyl-(3-sulfopropyl)ammonium hydroxide (SPE;497%), N-isopropylacryamide (NIPAAm; 97%), inhibitor remover and the dextrans were from Sigma-Aldrich. Potassium chloride (KCl; 99%) was purchased from AppliChem and calcium chloride (CaCl2;494%) was from Roth. All chemicals were used as received; only NIPAAm was recrystallized from hexane in order to remove the inhibitor. The water which was used for all experiments was provided by a MilliQ system (Millipore).

cell was connected to a water reservoir which was under nitrogen pressure what enabled the adjustment of trans-membrane pressure which was set to a maximum of 0.3 bar. Permeate was collected in a beaker for a certain time and the amount of water was determined gravimetrically. For temperature adjustment, the cell was placed into a temperature-controlled water bath. The permeate flux, J, was calculated according to Eq. (2), with the mass of permeate, m, in a particular measuring time, Δt, the water or salt solution density at a specific temperature T, ρT, and the effective membrane (filter) area, A (V is the permeate volume).

J= 2.2. Syntheses 2.2.1. Pre-functionalization of the membranes Before grafting, membrane sheets had been pre-modified according to Friebe and Ulbricht [22]. First, they were oxidized with potassium permanganate what leads to an increased density of carboxylic acid groups. After activation of the carboxylic acid groups with diisopropylcarbodiimide and N-hydroxybenzotriazole, an animation with ethanol amine in DMF was carried out. In the next step, the ATRP initiator α-bromoisobutyrylbromide was immobilized; for that, 4 membrane sheets (10
15 cm) were put in 50 mL solution of 80 mmol/L α-bromoisobutyrylbromide, 100 mmol/L triethylamine and 5 mmol/L 4-(N′N′-dimethylamino)pyridine in dry acetonitrile for 2 h. Afterwards they were washed three times in pure acetonitrile and twice in methanol and dried at 50 °C for two hours. Before using them for grafting, the membrane sheets were cut into smaller squares (3
3 cm). 2.2.2. Grafting of the diblock copolymer on the membrane surface For grafting of the first PSPE block, the procedure of Yang and Ulbricht [23] had been used as basis and modified in some points. As reaction vessel, 100 mL 3-neckflasks with one membrane (3
3 cm) inside and protected by argon atmosphere were used. It had been found that a solution with 0.2 mol/L SPE, 0.002 mol/L CuBr, 0.0008 mol/L CuBr2 and 0.004 mol/L Bpy in a 1/1 methanol/water mixture worked best for the modification of the PET membrane with 110 nm pores. This corresponds to a molar ratio of [SPE]/[CuBr]/[CuBr2]/[Bpy] ¼[50]:[0.5]:[0.2]:[1]. For the preparation of the solution, Bpy and CuBr2 were dissolved in water/methanol mixture and purged with argon for 30 min; thereafter CuBr was added. In a separate flask, the SPE solution was prepared and purged with argon under ice bath cooling for 30 min This SPE solution was added to the reaction solution after CuBr had been dissolved. Then, the solution was transferred into the flasks with the membranes under argon atmosphere where the polymerization took place. After the desired polymerization times which were here 15, 30, 45 and 60 min, the polymerization was stopped by adding 10 mL of a solution of 500 mg CuBr2 and 1.25 mL PMDETA in 100 mL water/methanol. Then the membranes were removed, washed and dried at 50 °C for two hours. The grafting of the PNIPAAm block was carried out like described in a previous work [26]. A degassed solution with 11.3 g re-crystallized NIPAAm, 11 mg CuCl2, 83 mg CuCl and 344 mL Me6TREN in 100 mL DMF was added under argon atmosphere to the flasks containing one membrane (3
3 cm). This corresponds to a molar ratio of [NIPAAm]/[Me6Tren]/[CuCl]/[CuCl2] ¼[120]:[1.5]:[1]:[0.1]. For stopping the polymerization, 10 mL of a solution of 500 mg CuBr2 and 1.25 mL PMDETA in 100 mL DMF was added. Either membranes already grafted with PSPE were used as macro-initiator or the procedure was used in reference experiments where PNIPAAm was directly grafted on the ATRP-initiator immobilized membranes. 2.3. Membrane characterization 2.3.1. Gas flow/pore dewetting permporometry The pore size of the base membrane had been determined to be about 110 nm by gas flow/pore dewetting permporometry using the PMI capillary flow permporometer (Porous Materials, Inc., USA) as described in detail in previous publications [22] (see Supporting information, Fig. S1). 2.3.2. Determination of degree of grafting The membranes were weighed on a micro balance (Genius, Sartorius) before and after grafting. From the differences of the weight before grafting, m0, and after grafting, mgr, the degree of grafting, DG, which corresponds to the polymer weight per specific surface area of the used sample, Asp (cm2), has been calculated according to Eq. (1).

DG=

(mgr −m 0 ) Asp

(1)

The specific surface area of the base membrane had been determined to be 6.5 m2/g by nitrogen adsorption isotherm measurement and BET analysis. 2.3.3. Flux measurements and determination of hydrodynamic layer thickness of grafted polymer layer on the pore wall For determination of water permeability, a dead-end setup with a 25 mL Amicon cell and a membrane diameter of 25 mm (Millipore) had been used. The

dm

m V = ρ T ⋅A⋅∆t A⋅∆t

(2)

According to Hagen-Poiseuille law (3), the hydrodynamic radius rm or diameter of the modified membrane pore is correlated to the flux through that pore, J′.

J′ =

∆P⋅π⋅rm4 ∆P⋅π⋅d m4 = 8⋅η⋅l 128⋅η⋅l

(3)

where ΔP is the applied trans-membrane pressure, η is the viscosity of the used solution and l is the length of the pore represented by the membrane thickness. As Hagen-Poiseuille Eq. (3) is valid only for one single pore, it is used to describe the flux J through the entire membrane by multiplying it with the number of pores Np per active membrane area A; see Eq. (4). 1

⎛ 128⋅V ⋅η⋅l ⎞ 4 ⎟⎟ d m=⎜⎜ ⎝ ∆P⋅π⋅A⋅∆t⋅Np ⎠

(4)

For the determination of the pore number NP, the flux through one single pore has been calculated by Eq. (3) because the pore diameter of the base membrane db is known from gas flow/pore dewetting measurements (cf. Section 2.3.1) first. Secondly, the real flux through the base membrane at the same trans-membrane pressure ΔP has been measured. The number of pores NP is then obtained by dividing the measured flux through membrane by the calculated flux for one pore. The hydrodynamic layer thickness, Lh, can then be determined according to Eq. (5). L h=

db −d m 2

(5)

with db, from gas flow/pore dewetting permporometry and dm from the flux data for the respective membrane under specific conditions. This is analogous to the procedure used in previous works [3,22,26,33].

2.3.4. Dextran diffusion experiments For diffusion experiments, a circular membrane sample with an effective diameter of 20 mm was placed between two glass chambers, which contained the feed and permeate solutions. As feed, a solution of different dextrans (average molecular weight of 50,000, 100,000, 200,000, 500,000 and 2,000,000 g/mol at a total concentration of 10 g/L (2 g/L of each dextran) in water was used. As initial permeate, water was used. To protect the membranes from microbial fouling, 0.01 mol/L NaN3 has been added to the initial permeate and the feed solutions. In cases where diffusion was carried out with dextran in 1000 mmol/L KCl or 100 mmol/L KClO4 solutions, the same salt concentration had been added to feed and to permeate. The diffusion cell was placed into a temperature-controlled water bath, and the temperature was adjusted to 25 or 40 °C. After diffusion times of 1, 2, 3, 4, and 5 days a small volume was taken from feed and permeate compartments, and the carbon content was analyzed via TOC measurement (TOC-VCPH/CPN, Shimadzu). After 7 days, samples were taken and analyzed by high-performance gel permeation chromatography (GPC; 2 SUPREMA columns from PSS; detector RI-101 from Shodex; elution with 0.01 mol/L NaN3 in water at 25 °C; calibration with dextran standards from PSS), and the rejection as function of molecular weight, R(M), was calculated according to Eq. (6). R (M )=1 −

2⋅cP cF +cR

(6)

where cP, cF, and cR are the concentrations of dextran of a specific molecular weight, M, in the permeate, the feed and the retentate, respectively. In the cases where permeate contained added KCl or KClO4, it was removed via dialysis before GPC analyses. Therefore, the permeate was filled into a dialysis tube (ZelluTrans, Roth, MWCO¼ 3500 g/mol) and placed into a water bath with stirrer. The water was changed to pure water four more times until conductivity reached the value of pure water. Thereafter, NaN3 was added so that a concentration of 0.01 mol/L was obtained. From the calculated molecular weight cut-off (MWCO), i.e. the M value where R is 0.9, the cut-off pore size of the membrane could be estimated. The corresponding test solute size has been calculated using a well established relationship between the molecular weight of dextran and its Stokes radius which also refers to its hydrodynamic radius [34].

2.3.5. Reference ultrafiltration experiments Polyethersulfone (PES) ultrafiltration membranes with a nominal molecular

M. Gajda, M. Ulbricht / Journal of Membrane Science 514 (2016) 510–517

513

weight cut-off of 150,000 g/mol were purchased from Microdyn-Nadir (Germany). As described in Section 2.3.3, an Amicon cell was used for filtration. As feed, solutions of dextran 168,000 g/mol at a concentration of 1 g/L, either in water or in 100 mmol/l KClO4 in water, were used. A trans-membrane pressure of 0.05 bar was applied. The permeate and feed were then analyzed via TOC measurements and the rejection, R, calculated according to Eq. (7).

R=(1 −

cP )⋅100% cF

(7)

where cP and cF are the dextran concentrations of permeate and feed, respectively.

3. Results and discussion 3.1. Synthesis of PET-g-PNIPAAm-b-PSPE capillary pore membranes by SI-ATRP First, a well controlled growth of polymer chains by SI-ATRP had to be established, what is crucial for a successful grafting of diblock copolymers. ATRP is based on the reversible cleavage of carbon-halide bonds, initially in the ATRP initiator group and later on the end of the growing polymer chain. This reaction is typically catalyzed by complexes of Cu(I) ions which are oxidized to Cu(II) upon radical formation and are regenerated after re-formation of the carbon-halide bond [10]. For membranes with pores in the ultrafiltration range, efficient polymerization initiation and very slow chain growth are required; such well adjusted reaction conditions lead to an even polymer growth on the outer surface and also on the pore walls. Only very few termination reactions are allowed to occur, what is the basis for grafting of a second block because active carbon-halide groups at the chain end are required for a re-initiation of a second polymerization. In earlier work [26], the same base membrane as in present work (TE PET with nominal pore diameter of 80 nm) had been used, and the density of immobilized ATRP active C-Br had been estimated to be about 1.0 nm  2. Considering a typical value for grafting efficiency, i.e., 10%, this would lead to a grafted chain density of  0.1 nm  2 (cf. [26]). An overview on the reaction steps is provided in Scheme 1. Reaction conditions according to Yang and Ulbricht [23] which had worked well for PET track-etched membranes with pore diameters in the range of 600– 800 nm lead to fast increase of degree of grafting (DG) and after 1 h to very high values what caused blocking of the pores. When a short polymerization period was applied, no linear growth of DG with time could be observed. This indicated poor reaction control, so that no adjustment of the DG could been obtained. The rate of polymer growth was then decelerated by adjusting the ratio between Cu(I) and Cu (II), i.e. reducing the fraction of activator relative to deactivator species. Fig. 2 clearly shows that under such conditions the DG was linearly growing with increasing polymerization time. The ratio of [CuBr(I)]:[CuBr2(II)] ¼[5]:[2] led to a well controlled polymerization, and the relatively low grafted PSPE layer thickness made the resulting membranes suited for hydraulic permeability tests. The membranes which had been synthesized by these conditions had a DG in the range from 0.12 to 0.33 mg/cm2. For further grafting functionalization, the membrane type obtained after 15 min reaction time and with a DG of about 0.15 mg/cm2 was chosen. That DG corresponds to a hydrodynamic layer thickness of 3 nm, what had been calculated from water permeability data via Hagen-Poiseuille equation (cf. Sections 2.3.3 and 3.3). For low grafting density, this layer thickness is a reasonable estimate of hydrodynamic diameter of the grafted chains; for high grafting density, the actual hydrodynamic diameter would be over-estimated when using hydrodynamic layer thickness. Considering the brush criterion, i.e. that average distance between grafting sites should be smaller than hydrodynamic diameter of the grafted macromolecules, and that the estimate of 10% grafting efficiency (cf. above) is

Fig. 2. Degree of grafting of PSPE as first block on capillary PET membranes as function of polymerization time. conservative, the obtained grafted structure is probably in between the “brush” and the “mushroom” regime (cf. Refs. [22,26,27]). Hence, it is in the range of grafting density where mutual repulsion between grafted chains has some influence but grafted chains are not yet in fully stretched conformation. Such small layer thickness offers enough space for grafting of a second block and for subsequent swelling of the copolymer layer within the pore without completely blocking it. However, the used ATRP system is very sensitive. If the reaction solution has contact to only very small amounts of oxygen, it may become completely inactivated by oxidation of Cu(I) to Cu(II), but if the concentration of Cu (I) is a slightly too high, the polymerization proceeds too fast and uncontrolled. Because of this very high sensitivity of the reaction system, it was difficult to obtain always reproducible results with respect to the grafted mass of PSPE on the membrane. Only batches of membranes prepared under these conditions and having a DG in the range of 0.15–0.35 mg/cm2 and water permeabilities between 730 and 135 L/m2 h bar had been used for further experiments. According to the conditions established by Frost and Ulbricht [26], PNIPAAm had been grafted on the PSPE layer on PET membranes, and also on membranes only with PET-immobilized ATRP initiator (cf. Scheme 1); results are summarized in Fig. 3. Because of the lower reactivity of NIPAAm compared to SPE, the more reactive CuCl instead of CuBr (used for PSE) had been used as catalyst in these experiments. There was no significant difference between the DG for grafting directly on the membrane pore walls or on pores with PSPE chains as first block. In both cases, the DG of PNIPAAm was growing linearly with the polymerization time, what is an indication for well controlled polymerization conditions. Obviously, only very few termination reactions had occurred during preparation of the first block, otherwise there would have been a significantly lower density of initiator groups and, consequently, a lower DG would have been achieved. The grafting of the second block was efficient for low and high DG values of PNIPAAm, and this data indicate that a diblock structure had indeed been achieved. This can be explained by two main reasons. First, the initiator density on the membrane had been high so that under the used conditions mutual repulsion of neighbored grafted chains has influence on the degree of swelling (cf. above). Nevertheless, it must be kept in mind that the actual grafting density is unknown (cf. Section 3.4); its experimental determination for a copolymer covalently grafted to another polymer (here PET) is extremely challenging and typically not feasible. Second, the grafting of PNIPAAm as second block was performed in DMF which is a poor solvent for PSPE and what consequently leads to a collapse of the grafted PSPE. Therefore, the ATRP initiator

Scheme 1. Immobilization of ATRP initiator groups on the pre-modified PET membrane surface featuring a high density of carboxylic acid groups, SI-ATRP with SPE to yield the 1st block (PSPE), and subsequent SI-ATRP with NIPAAm to yield the 2nd block (PNIPAAm).

514

M. Gajda, M. Ulbricht / Journal of Membrane Science 514 (2016) 510–517

Fig. 3. Degree of grafting of PNIPAAm as function of polymerization time, either on PET membranes which had before been functionalized with PSPE as first block (DG 0.15 mg/cm2), or directly on ATRP initiator-immobilized PET membrane. groups on the PET surface which may have remained un-reacted in the first grafting reaction are covered by the collapsed PSPE brushes. Hence, these initiator groups cannot be activated and react with NIPAAm. Further, as will be shown in the next sections, pronounced stimuli responsivity towards both, temperature and ions, could be observed from permeability measurements. If both kinds of polymer chains would be grafted directly on the membrane surface next to each other, the longer chains would determine the responsivity as the shorter ones would be hidden in between the longer ones, and a collapse of those would have no influence on the permeability (cf. Ref. [27]). Consequently, the formation of co-grafted brushes, i.e., a layer architecture where PSPE and PNIPAAm are both tethered directly to the pore surface is very unlikely. Because of the very low polymerization rates (cf. Figs. 2 and 3), the low degrees of functionalization (corresponding to only a few percent of the total mass of monomer added to the pore at the start of the reaction), and the fast diffusion of monomer from the additional solution supplied on both sides of the membrane (cf. Section 2.2.2), the reaction conditions are identical for the outer membrane surface and the entire pore surface so that grafting will most likely be uniform (cf. also Ref. [26]). 3.2. Influence of the concentration of particular ions and of temperature on the grafted PSPE-b-PNIPAAm layer thickness in PET membrane pores The responsivity of the PET-g-PSPE-b-PNIPAAm membranes to ions and temperature had been analyzed in detail by measuring the permeability at 25 °C and 40 °C with KClO4 solution, which in the preliminary experiments had caused a big increase of hydrodynamic layer thickness and with KCl which had caused a smaller increase. Analogous observations had been made before in studies with PET-g-PSPE membranes of larger pore size [23]. For comparison, also the permeabilities of membranes only with grafted PNIPAAm and PSPE homopolymer chains had been analyzed. For these measurements, a diblock copolymer grafted membrane with a DG of 0.33 mg/cm2 for PSPE and 0.89 mg/cm2 for PNIPAAm had been chosen, what corresponds to a hydrodynamic layer thickness of  9 nm for PSPE and  16 nm for PNIPAAm. The base membranes barrier pore size had been measured to be about 110 nm by gas flow/pore dewetting permporometry (cf. Fig. S1), so that after grafting it should be reduced by up to 50 nm. That still should enable permeability measurements at low pressure. As high flow rates might deform the polymer chains due to their flexible structure [35], trans-membrane pressures of up to a maximum of 0.3 bar, what with the flux through the unmodified membrane corresponds to a pore wall shear rate of 7*104 s  1, has been used for permeability measurements. All results are summarized in Fig. 4. From Fig. 4 it can be clearly seen that membranes with grafted pure PNIPAAm were only responsive to temperature and showed no change of layer thickness in different salt solutions. Only in a solution containing 1000 mmol/L KCl, the layers collapsed what can be explained by the “salting-out” effect of this hydrogel which occurs at high salt concentrations and might also be linked to a reduced LCST of PNIPAAm at this concentration [36]. In contrast, membranes with grafted pure PSPE brushes showed a large responsivity towards salt solutions, but they were also responsive to temperature changes. The latter effect is linked to the upper critical solution temperature (UCST) of PSPE which causes a better solubility and expansion of PSPE chains at higher temperature. The salt responsivity, in particular the larger effects of “chaotropic” ClO4  compared to Cl  ions, could be explained by the empirical Hofmeister series [23,37]. This is in line with the concept of “matching water affinities” which had been introduced by Collins [38]. Generally, monovalent cations have much less specific effects compared to anions. ClO4  ions and ammonium groups in PSPE have a high tendency of building ion pairs what leads to increased net charge in the zwitterionic side groups, thus causing a chain expansion with increasing concentration of salt. This is the so called “anti polyelectrolyte” effect, and the magnitude is stronger for ClO4  than for Cl- ions which matches less

Fig. 4. Hydrodynamic layer thickness of PSPE-b-PNIPAAm layers (DG for PSPE 0.33 mg/cm2; DG for PNIPAAm 0.89 mg/cm2) as well as the respective grafted homopolymers (DG for PNIPAAm 0.58 mg/cm2 and DG for PSPE 0.32 mg/cm2) in PET membrane pores in dependency of different salt solutions and temperatures. well to the ammonium groups in PSPE. However, not only the ion type has an effect on the expansion but also the total salt concentration. High concentrations of KClO4 (100 mmol/l, close to the solubility limit) lead to an increase of the copolymer layer thickness (“anti polyelectrolyte effect”), while very high concentrations of KCl (1000 mmol/L solution) lead to a decrease (“salting out”). For the PET-g-PSPE-b-PNIPAAm membranes one can see a combination of the properties of membranes comprising grafted pure PSPE and pure PNIPAAm. For all solutions there was a decrease of layer thickness when the temperature was raised to 40 °C, what is caused by the collapse of the PNIPAAm chains. Obviously, the collapse of PNIPAAm antagonized the expansion of the PSPE chains at high temperature. That might be due to the ratio of the two grafted polymers; the DG was much higher for PNIPAAm. The intensity of layer expansion depended on the kind of salt and the concentration. The salt KClO4 lead to a stronger expansion than KCl, and higher salt concentration lead also to higher swelling, unless concentration was so high that a “salting-out” of the PNIPAAm block occurred, like in case of 1000 mmol/L KCl. For grafted diblock copolymer layers with a micro-segregated morphology (cf. Fig. 1) access of ions to the lower layer (here PSPE) might also have an influence on the response. Considering spacing of the hydrophilic chains due to grafting density (cf. Section 3.1) and the much smaller size of hydrated ions diffusion into the layer and contact to the PSPE region should not be hindered. However, an analysis of data shown in Fig. 4 from this perspective reveals that the increase of layer thickness of pure PSPE layers upon addition of salt is larger than that of the PSPE-bPNIPAAm layers (see Supporting information, Fig. S2). One reason may be that this difference is caused by the PNIPAAm segments although the pure PNIPAAm layers show no sensitivity to ions at those concentrations (cf. Fig. 4). The limits of the analyses in terms of hydrodynamic layer thickness will be discussed critically later (see Section 3.4). Nevertheless, the hypothesis that this difference might be caused by limited contact between ions and PSPE segments layer in case of PSPE-b-PNIPAAm membranes may indeed be tested more in more detail, also by synthesis of membranes with PNIPAAm-b-PSPE structure. In Fig. 5 the responsivity of PET-g-PNIPAAm-b-PSPE is summarized by showing the effective pore diameter of the membrane in dependency of the two different stimuli. It can be clearly seen that at 25 °C and in presence of strong chaotropic ions, the strongest layer expansion and therefore the smallest pore diameter can be

Fig. 5. Hydrodynamic pore diameter of PET-g-PSPE-b-PNIPAAm capillary pore membranes (DG for PSPE 0.33 mg/cm2; DG for PNIPAAm 0.89 mg/cm2) as function of salt (10 mmol/L and 100 mmol/L KClO4) and temperature.

M. Gajda, M. Ulbricht / Journal of Membrane Science 514 (2016) 510–517

Fig. 6. Change of hydrodynamic layer thickness of PSPE-b-PNIPAAm in PET membrane pores, with varied DG of PNIPAAm at constant DG of PSPE (0.12 mg/cm2), measured in different solutions (100 mmol/L KClO4 and 1000 mmol/L KCl) at two different temperatures, all relative to the layer thickness in water at 25 °C. obtained. However, if the temperature is raised to 40 °C or if no salt is present either the PNIPAAm block or the PSPE block collapse, and the total layer thickness decreases and pore diameter increases. In case that both of these stimuli are active the total layer thickness decreases to its minimum so that the pores are opened to the largest extent. 3.3. Stimuli-responsivity of the PET-g-PSPE-b-PNIPAAm capillary pore membranes in dependency of different block ratios In further experiments, the influence of the PNIPAAm block length on the responsivity of PET-g-PSPE-b-PNIPAAm membranes had been evaluated by retrieving the hydrodynamic layer thickness from permeability measurements. For this series, the DG of PSPE had been kept constant at about 0.12 mg/cm2 (variation o 0.01 mg/cm2) and the DG of PNIPAAm had been varied from 0 to 1.01 mg/cm2, what corresponds to a hydrodynamic layer thickness from 3 to 23 nm measured for PET-g-PNIPAAm in pure water. For better comparison not the absolute thickness is shown in Fig. 6, but the change of layer thickness compared to that which is measured in pure water at 25 °C. With increasing fraction of PNIPAAm, there was a clear tendency of stronger reduction of layer thickness at high temperature while the ion responsivity decreased. This increase of temperature responsivity was valid for pure water but also for both salt solutions, with KCl and KClO4. Up to a DG of 0.36 mg/cm2 for PNIPAAm, the properties of PSPE dominated the responsitivity of the PET-g-PSPE-b-PNIPAAm membrane, i.e., no decrease of the layer thickness at 40 °C compared to water at 25 °C was observed. Hence, the “anti-polyelectrolyte” effect of PSPE over-compensated the LCST effect of PNIPAAm. Consequently, at 25 °C, a thickness increase by both salts, somewhat larger for KClO4 compared to KCl, could be observed. At a DG of 0.46 mg/cm2 for PNIPAAm, what corresponds to a mass ratio [PSPE]:[PNIPAAm]  1:3, high ion responsitivity and high temperature responsivity could be obtained. At 40 °C, the layer thickness decreased stronger in the presence of KCl what seemed to be dominated by the total salt concentration and caused by the “salting-out” effect (for PNIPAAm). At a DG of 1.01 mg/cm2 for PNIPAAm, the temperature responsivity of PNIPAAm outweighed the ion sensitivity of PSPE. At 40 °C, in a solution of 1000 mmol/L KCl the strongest collapse of the layer could be observed because of the combination of “salting-out” and temperature effects. However, the small block of PSPE lead still to a strong increase of the layer thickness in presence of KClO4 what shows that even at such high mass ratio ([PSPE]:[PNIPAAm]¼ 1:6.7) there is still a significant responsivity towards ions. Hence, the ion responsivity decreased with increasing ratio of PNIPAAm to PSPE, but even at a high ratio the independent responsivity towards the chaotropic salt could be observed. All these effects strongly support the assumption already discussed in Section 3.1, that the grafted diblock copolymer layer (due to its relatively high grafting density) adopts to a large extent a micro-segregated layered architecture (cf. Fig. 1). 3.4. Molecular sieving in diffusion experiments through PSPE-b-PNIPAAm grafted PET membranes at different temperatures and salt conditions The hydraulic permeability measurements already indicated that the effective pore size of the PET-g-PSPE-b-PNIPAAm membranes was changing in dependency of the temperature and the kind of ions. In order to proof that this in indeed also related to the actual pore diameter in the sub-100 nm range, diffusion experiments with solutions of dextran mixtures (average molecular weights between 50,000 and 2,000,000 g/mol) had been done. Again, such conditions had been chosen because under these experimental conditions no pressure is used so that the polymer layers cannot be deformed by convective flow and macromolecules are not forced through hydrogel-lined deformable pores. To evaluate the responsitivity of the membranes, pure water, as well as KCl and KClO4 solutions had been used at 25

515

Fig. 7. Rejection curves from diffusion experiments for PET-g-PSPE-b-PNIPAAm membranes (DG PSPE 0.33 mg/cm2; DG PNIPAAm 0.89 mg/cm2) in different solutions (100 mmol/L KClO4 and 1000 mmol/L KCl) and at different temperatures, measured by GPC.

and 40 °C. The diffusion measurements had been done for 7 days; during the first 5 days the rate of permeation through the membrane had been analyzed based on TOC data. After 7 days the feed and permeate had been analyzed by GPC, yielding data on of the fraction of rejected macromolecules of a particular size. The sieving curves which are shown in Fig. 7 and in Supporting information (Fig. S3) reveal, which percentage of molecules with a particular molecular weight is rejected by the grafted membrane in different solutions at different temperatures. The molecular weight cut-off (MWCO; i.e. M for R ¼ 0.9) could also be estimated from this data. In pure water at 25 °C, the membrane showed the highest rejection (MWCO¼ 40,000 g/mol). However, if the temperature was increased to 40 °C, the rejection curve shifted to very high molecular weights (MWCO ¼ 350,000 g/mol) what points to an increase of the effective pore diameter. For KCl and KClO4 solutions at 40 °C and for KClO4 solution at 25 °C, the rejection curves were similar; apparently larger macromolecules than in pure water at 25 °C but smaller ones than in water at 40 °C were rejected (MWCO 115,000 g/mol). The rejection curve for KCl at 25 °C was shifted significantly to bigger macromolecules than in pure water at 25 °C (MWCO¼ 78,000 g/mol) but smaller ones than for same salt at higher temperature. Analogous results were obtained for membranes with different PNIPAAm:PSPE ratios (cf. Supporting information; Fig. S3). Overall, it can be clearly seen that membrane pores open when temperature is increased what can be explained by the collapse of PNIPAAm segments due to exceeding the LCST. The KCl solution leads, due to the dominating “salting out” effect, also to enlarged pores. At 40 °C this effect is stronger than at 25 °C because the LCST of PNIPAAm is exceeded. However, the sieving curves indicated that the pores are enlarged in presence of KClO4 what is unexpected as the PSPE layer should expand in presence of ions; and this effect, i.e. pore narrowing, has been observed consistently when analyzing hydraulic permeability data (cf. Sections 3.2 and 3.3). The results for dextran diffusion rates for all different block copolymer grafted membranes which are presented in Supporting Information (Figs. S4 and S5) also pointed to a reduced pore size in presence of KClO4. Results of an additional ultrafiltration experiment (Supporting information, Table S1) provided strong evidence that the hydrodynamic diameter of dextran becomes smaller in 100 mmol/L KClO4 solution compared to pure water: the rejection of the same dextran with the same ultrafiltration membrane, as measured by total organic carbon, was significantly reduced in KClO4 solution compared to pure water (more detailed description in Supporting Information). Such effect is not observed for KCl solutions so that the reason is probably also linked to the chaotropic properties of ClO4  ions. Because the complementary analysis of dextran in the MWCO analysis had been done by GPC in water as mobile phase, the reduction of dextran size occurred only during the diffusion experiments in KClO4 solution. This can explain the shift of the rejection curve to higher molecular weights, but this data cannot be linked to a larger membrane pore size. From the MWCO data, the corresponding hydrodynamic size of dextran has been calculated (cf. Section 2.3.4), which correlates to effective barrier pore diameter. Here, this value is only used as rough estimation and more interesting is its change as function of applied stimuli as well as the comparison with pore diameters estimated from hydraulic permeability (see Table 1). Because of the additional effects of ClO4  ions on dextran (see above), only data with KCl are used to assess the salt effects. Note that because KCl had been used at a concentration of 1000 mmol/l, the effects are different from the salt effects with KClO4  where the effective pore size became smaller when increasing salt concentration up to 100 mmol/l (cf. Fig. 5). From hydraulic permeability and diffusion experiments the same qualitative trends concerning the pore size can be observed. In water at 25 °C the smallest pores can be found. In KCl solution of high concentration at 25 °C the pores open

516

M. Gajda, M. Ulbricht / Journal of Membrane Science 514 (2016) 510–517

Table 1 Comparison of effective membrane pore size estimated from hydraulic permeability and dextran diffusion (PET-g-PSPE-b-PNIPAAm membranes; DG PSPE 0.33 mg/cm2; DG PNIPAAm 0.89 mg/cm2). Pore size [nm] from

Water 25 °C KCl 25 °C Water 40 °C KCl 40 °C

Hydraulic permeability

Dextran diffusion

63 71 80 73

8 11 25 14

slightly. At 40 °C in water the largest pores can be found for both experiments. By adding KCl at 40 °C the pores close slightly but are still larger than at 25 °C. However, the relative change of pore size between 25 and 40 °C is much more distinct in diffusion experiments. This is because the absolute pore diameters of the modified membrane calculated from hydraulic permeability data are much larger than those obtained from dextran diffusion. The surprisingly large differences between estimations from two independent experiments can only be related to the different experimental conditions, with convective liquid flow in one case and only solute diffusion in the other case. First of all, from diffusion date, the pore barrier pore diameter is underestimated because it had been calculated from MWCO (i.e. at 90% rejection), and because solute sieving is additionally hindered by pore entrance effects. Further, in hydraulic permeability measurements water could flow not only through the open space of the pore but partially also through the free space between the polymer chains, i.e., through the grafted hydrogel layer [39]. This hydrogel filled fraction of the pore space is not accessible for large dextran molecules and it will impose significant hindrance for smaller dextrans. Therefore, the effective pore size for dextran transport is smaller than that for (bulk) water. Hence, these quantitative data also indicate that the assumption of a high grafting density, leading to a “brush” structure with strong mutual repulsion between grafted block copolymer chains may not be valid (cf. Section 3.1). If that would be the case, the hydraulic permeability results could in addition be influenced by a deformation of polymer chains due to the relatively high shear rates imposed during the flux measurements [35]. A maximum shear rate of 7*104 s  1 had been estimated for unmodified base membrane under the used experimental conditions (cf. Section 3.1). For grafted membranes, the same trans-membrane pressure leads to lower flux, but this is because the pore diameter is reduced by the hydrogel layer on the pore wall. Hence, a deformation of this hydrogel layer, having high water content, due to shear is also realistic. And the dextran diffusion data are certainly again not influenced by such convection-specific effect. Therefore, one can conclude that the grafted diblock copolymer layer is probably not in the “brush” regime but has still a significant degree of vertical segregation of the two grafted polymers which leads to pronounced independent responses to specific stimuli (cf. Fig. 1). And this very interesting behavior is even more pronounced, as seen by the larger relative changes of pore size, under conditions without convective flow through the pores (cf. [39]).

4. Conclusions In this work ion- and temperature-responsive membranes have been synthesized starting with PET track-etched membranes with a pore size of 110 nm. Ion- and temperature-responsive properties have been introduced by using zwitterionic PSPE with “anti polyelectrolyte” behavior including specific effects of chaotropic salts, and PNIPAAm with dominating LCST effects, respectively. SI-ATRP conditions had been adjusted to a very slow, controlled chain growth from the pore surface for each of the two functional copolymers. By that, successful grafting of PSPE as first block and PNIPAAm as second block has been obtained. The reaction conditions favored the formation of a diblock instead of a co-grafted architecture. All experiments concerning the membrane barrier properties under varied solution conditions support that a vertical segregation of the two functional blocks in the grafted hydrogel layer had been achieved because the swelling or deswelling of the respective blocks could be triggered independently by different stimuli. Hydraulic permeability measurements reveal that compared to Cl  ions, ClO4  ions cause a stronger increase of layer thickness of grafted PSPE-b-PNIPAAm. Therefore ClO4  is a more effective stimulus toward closing the pores. Independently, the

membrane pore size can be tuned by temperature, so that the smallest pores are obtained at a temperature below LCST of PNIPAAm and in presence of salt (but not too high concentration, see below), and the largest pores at temperature above LCST without salt. However, it has also been shown that not only the ion type but also the salt concentration dictates the polymer layer thickness. Too high salt concentration (shown for Cl  ) will promote collapse of PNIPAAm even at low temperature. Furthermore, also the ratio between ion- and temperature responsive blocks is important for the overall functionality of the membrane. Significant differences between barrier pore size estimates from hydraulic permeability and dextran diffusion had been observed, which can be related to partial permeation of water through the grafted hydrogel layer and grafted layer deformation under convective conditions. Most important, the macromolecular solute diffusion data confirmed the opening and closing mechanism in response to ions and temperature, and even more pronounced differences between the various states have been obtained because only solute size exclusion is involved as selective barrier mechanism. The data for dextran at high ClO4  concentration indicate that additional effects of chaotropic salts on size of macromolecules in solution may need to be considered as well. Overall, this work had provided significant progress for design and synthesis of stimuli-responsive membranes. The demonstrated pronounced and independent responses to two stimuli could enable various functions as logic gates (cf. [32]). It should also be noted that the conversion of an external signal such as high frequency electromagnetic field into thermo-responsivity of capillary pore membranes with help of localized nanoparticle heaters and pore confined PNIPAAm had also already been demonstrated [33,40] and could be combined with specific ion response. Novel dual-responsive ultrafiltration membranes based on the principles established here could be applied in advanced sensors for detection of particular ions or for the controlled release of drugs.

Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.memsci.2016.05. 001.

References [1] M.A.C. Stuart, W.T.S. Huck, J. Genzer, M. Muller, C. Ober, M. Stamm, G. B. Sukhorukov, I. Szleifer, V.V. Tsukruk, M. Urban, F. Winnik, S. Zauscher, I. Luzinov, S. Minko, Emerging applications of stimuli-responsive polymer materials, Nat. Mater. 9 (2010) 101–113. [2] D. Wandera, S.R. Wickramasinghe, S.M. Husson, Stimuli-responsive membranes, J. Membr. Sci. 357 (2010) 6–35. [3] C. Geismann, A. Yaroshchuk, M. Ulbricht, Permeability and electrokinetic characterization of poly(ethylene terephthalate) capillary pore membranes with grafted temperature-responsive polymers, Langmuir 23 (2007) 76–83. [4] M. Ulbricht, Photograft-polymer-modified microporous membranes with environment-sensitive permeabilities, React. Funct. Polym. 31 (1996) 165–177. [5] M. Birkner, M. Ulbricht, Ultrafiltration membranes with markedly different pH- and ion-responsivity by photografted zwitterionic polysulfobetain or polycarbobetain, J. Membr. Sci. 494 (2015) 57–67. [6] T. Ito, T. Hioki, T. Yamaguchi, T. Shinbo, S. Nakao, S. Kimura, Development of a molecular recognition ion gating membrane and estimation of its pore size control, J. Am. Chem. Soc. 124 (2002) 7840–7846. [7] Y. Sugawara, T. Tamaki, T. Yamaghuchi, Development of an aptamer-functionalized molecular recognition gating membrane targeting a specific protein on the basis of the aggregation phenomena of DNA-PNIPAM, Polymer 62 (2015) 86–93. [8] A. Nayak, H.W. Liu, G. Belfort, An optically reversible switching membrane surface, Angew. Chem. -Int. Ed. 45 (2006) 4094–4098. [9] H.H. Himstedt, Q. Yang, X.H. Qian, S.R. Wickramasinghe, M. Ulbricht, Toward remote-controlled valve functions via magnetically responsive capillary pore membranes, J. Membr. Sci. 423 (2012) 257–266. [10] K. Matyjaszewski, J.H. Xia, Atom transfer radical polymerization, Chem. Rev.

M. Gajda, M. Ulbricht / Journal of Membrane Science 514 (2016) 510–517

101 (2009) 2921–2990. [11] Y.J. Shih, Y. Chang, A. Deratani, D. Quemener, “Schizophrenic” hemocompatible copolymers via switchable thermoresponsive transition of nonionic/zwitterionic block self-assembly in human blood, Biomacromolecules 13 (2012) 2849–2858. [12] G. Masci, M. Diociaiuti, V. Crescenzi, ATRP synthesis and association properties of thermo responsive anionic block copolymers, J. Polym. Sci. Pol. Chem. 46 (2008) 4830–4842. [13] X.L. Zhao, W.G. Liu, D.Y. Chen, X.Z. Lin, W.W. Lu, Effect of block order of ABAand BAB-type NIPAAm/HEMA triblock copolymers on thermoresponsive behavior of solutions, Macromol. Chem. Phys. 208 (2007) 1773–1781. [14] F.J. Xu, E.T. Kang, K.G. Neoh, pH- and temperature-responsive hydrogels from crosslinked triblock copolymers prepared via consecutive atom transfer radical polymerizations, Biomaterials 27 (2006) 2787–2797. [15] F.Y. Fu, H.L. Xu, Y. Dong, M.L. He, Z.Y. Zhang, T.W. Luo, Y.X. Zhang, X. Hao, C. J. Zhu, Design of polyphosphazene-based graft copolystyrenes with alkylsulfonate branch chains for proton exchange membranes, J. Membr. Sci. 489 (2015) 119–128. [16] N. Singh, J. Wang, M. Ulbricht, S.R. Wickramasinghe, S.M. Husson, Surfaceinitiated atom transfer radical polymerization: a new method for preparation of polymeric membrane adsorbers, J. Membr. Sci. 309 (2008) 64–72. [17] W.W. Yue, H.J. Li, T. Xiang, H. Qin, S.D. Sun, C.S. Zhao, Grafting of zwitterion from polysulfone membrane via surface-initiated ATRP with enhanced antifouling property and biocompatibility, J. Membr. Sci. 446 (2013) 79–91. [18] J.K. Koh, Y.W. Kim, S.H. Ahn, B.R. Min, J.H. Kim, Antifouling poly(vinylidene fluoride) ultrafiltration membranes containing amphiphilic comb polymer additive, J. Polym. Sci. Part B-Polym. Phys. 48 (2010) 183–189. [19] N. Singh, Z. Chen, N. Tomer, S.R. Wickramasinghe, N. Soice, S.M. Husson, Modification of regenerated cellulose ultrafiltration membranes by surfaceinitiated atom transfer radical polymerization, J. Membr. Sci. 311 (2008) 225–234. [20] D. Wandera, S.R. Wickramasinghe, S.M. Husson, Modification and characterization of ultrafiltration membranes for treatment of produced water, J. Membr. Sci. 373 (2011) 178–188. [21] J. Ran, L. Wu, Z.H. Zhang, T.W. Xu, Atom transfer radical polymerization (ATRP): a versatile and forceful tool for functional membranes, Progress. Polym. Sci. 39 (2014) 124–144. [22] A. Friebe, M. Ulbricht, Controlled pore functionalization of poly(ethylene terephthalate) track-etched membranes via surface-initiated atom transfer radical polymerization, Langmuir 23 (2007) 10316–10322. [23] Q. Yang, M. Ulbricht, Novel membrane adsorbers with grafted zwitterionic polymers synthesized by surface-initiated ATRP and their salt-modulated permeability and protein binding properties, Chem. Mater. 24 (2012) 2943–2951. [24] C.Y. Li, W.C. Wang, F.J. Xu, L.Q. Zhang, W.T. Yang, Preparation of pH-sensitive membranes via dopamine-initiated atom transfer radical polymerization, J.

517

Membr. Sci. 367 (2011) 7–13. [25] H. Alem, A.S. Duwez, P. Lussis, P. Lipnik, A.M. Jonas, S. Demoustier-Champagne, Microstructure and thermo-responsive behavior of poly (N-isopropylacrylamide) brushes grafted in nanopores of track-etched membranes, J. Membr. Sci. 308 (2008) 75–86. [26] S. Frost, M. Ulbricht, Thermoresponsive ultrafiltration membranes for the switchable permeation and fractionation of nanoparticles, J. Membr. Sci. 448 (2013) 1–11. [27] A. Friebe, M. Ulbricht, Cylindrical pores responding to two different stimuli via surface-initiated atom transfer radical polymerization for synthesis of grafted diblock copolymers, Macromolecules 42 (2009) 1838–1848. [28] K. Pan, X.W. Zhang, R.M. Ren, B. Cao, Double stimuli-responsive membranes grafted with block copolymer by ATRP method, J. Membr. Sci. 356 (2010) 133–137. [29] S. Yamamoto, J. Pietrasik, K. Matyjaszewski, Temperature- and pH-responsive dense copolymer brushes prepared by ATRP, Macromolecules 41 (2008) 7013–7020. [30] Z.B. Zhang, X.L. Zhu, F.J. Xu, K.G. Neoh, E.T. Kang, Temperature- and pH-sensitive nylon membranes prepared via consecutive surface-initiated atom transfer radical graft polymerizations, J. Membr. Sci. 342 (2009) 300–306. [31] S. Kumar, Y.L. Dory, M. Lepage, Y. Zhao, Surface-grafted stimuli-responsive block copolymer brushes for the thermo-, photo- and ph-sensitive release of dye molecules, Macromolecules 44 (2011) 7385–7393. [32] H. Ohashi, S. Ebina, T. Yamaguchi, Logistic gate-like permeable property of gating membrane with ion-recognition polyampholyte, Polymer 55 (2014) 1412–1419. [33] X. Lin, R. Huang, M. Ulbricht, Novel magneto-responsive membrane for remote control switchable molecular sieving, J. Mater. Chem. B 4 (2016) 867–879. [34] 〈http://www.dextran.net/about-dextran/dextran-chemistry/physical-proper ties.aspx〉 (accessed: 29.04.16). [35] R.P. Castro, H.G. Monbouquette, Y. Cohen, Shear-induced permeability changes in a polymer grafted silica membrane, J. Membr. Sci. 179 (2000) 207–220. [36] C.A. Naini, M. Thomas, S. Franzka, S. Frost, M. Ulbricht, N. Hartmann, Hofmeister effect of sodium halides on the switching energetics of thermoresponsive polymer brushes, Macromol. Rapid Commun. 34 (2013) 417–422. [37] P. Lo Nostro, B.W. Ninham, Hofmeister phenomena: an update on ion specificity in biology, Chem. Rev. 112 (2012) 2286–2322. [38] K.D. Collins, Ions from the Hofmeister series and osmolytes: effects on proteins in solution and in the crystallization process, Methods 34 (2004) 300–311. [39] C. Geismann, F. Tomicki, M. Ulbricht, Block copolymer photo-grafted poly (ethylene terephthalate) capillary pore membranes distinctly switchable by two different stimuli, Sep. Sci. Technol. 44 (2009) 3312–3329. [40] A. Gajda, M. Ulbricht, Magnetic Fe3O4 nanoparticle heaters in smart porous membrane valves, J. Mater. Chem. B 2 (2014) 1317–1326.