Permeability and Electrokinetic Characterization of Poly(ethylene

Poly(ethylene terephthalate) (PET) track-etched membranes with average pore diameters of 692 and 1629 nm were functionalized using the monomer ...
0 downloads 0 Views 186KB Size
76

Langmuir 2007, 23, 76-83

Permeability and Electrokinetic Characterization of Poly(ethylene terephthalate) Capillary Pore Membranes with Grafted Temperature-Responsive Polymers† Christian Geismann, Andriy Yaroshchuk,‡ and Mathias Ulbricht* Lehrstuhl fu¨r Technische Chemie II, UniVersita¨t Duisburg-Essen, 45117 Essen, Germany ReceiVed February 8, 2006. In Final Form: May 30, 2006 Poly(ethylene terephthalate) (PET) track-etched membranes with average pore diameters of 692 and 1629 nm were functionalized using the monomer N-isopropylacrylamide (NIPAAm) and a photoinitiated “grafting-from” approach in which a surface-selective reaction has been most efficiently achieved by combinations of the unmodified PET surface with benzophenone and, alternatively, of an aminated PET surface with benzophenone carboxylic acid. Consistent estimations of the pore diameters of the base PET membranes and of the effective grafted polyNIPAAm layer thicknesses on the PET pore walls were possible only on the basis of the permeabilities measured with aqueous solutions of higher ionic strength (e.g., 0.1 M NaCl). However, the permeabilities measured with ultrapure water indicated that the “electroviscous effect” was significant for both base membranes. The influences of membrane pore diameter, surface charge, and solution ionic strength could be interpreted in the framework of the space-charge model. Functionalized membranes with collapsed grafted polymer hydrogel layer thicknesses of a few nanometers exhibited almost zero values of the zeta potential estimated from the trans-membrane streaming potential measurements. This was caused by a “hydrodynamic screening” of surface charge by the neutral hydrogel. Very pronounced changes in permeability as a function of temperature were measured for PET membranes with grafted polyNIPAAm layers, and the effective layer thickness in the swollen stateshere up to ∼300 nmscorrelated well with the degree of functionalization. The subtle additional effects of solution ionic strength on the hydrodynamic layer thickness at 25 °C were different from the effects for the base PET membranes and could be explained by a variation in the degree of swelling, resembling a “salting-out” effect. Overall, it had been demonstrated that the functionalized capillary pore membranes are well suited for a detailed and quantitative evaluation of the relationships between the synthesis, the structure, and the function of grafted stimuli-responsive polymer layers.

1. Introduction Surface functionalization of polymeric materials has drawn much attention in recent years. In this field, modifications with stimuli-responsive polymers are of particular interest because of their ability to undergo well-defined changes in their properties (e.g., shape or volume).1 This effect can be utilized, for example, to create novel valves in microfluidic channels,2-6 which could be alternatives to conventional (i.e., mechanical) systems. Functionalized polymer membrane pores can act as model systems for such actuators. In addition, some porous membranes have already been investigated as modules in microfluidic systems, for example, as separators or flow-through reactors7 or as gateable “nanofluidic” interconnects.8 A widely used method for the surface functionalization of porous membranes is “grafting-from” (i.e., a heterogeneous copolymerization of functional monomers) via UV irradiation, and this approach has already been used to † Part of the Stimuli-Responsive Materials: Polymers, Colloids, and Multicomponent Systems special issue. * Corresponding author. E-mail: [email protected]. Fax: +49 201 183-3147. ‡ Current affiliation: Paul Scherrer Institut, Villigen, Switzerland.

(1) Hoffman, A. S. Macromol. Symp. 1995, 98, 645. (2) Arndt, K. F.; Kuckling, D.; Richter, A. Polym. AdV. Technol. 2000, 11, 496. (3) Beebe, D. J.; Moore, J. S.; Bauer, J. M.; Yu, Q.; Liu, R. H.; Devadoss, C.; Jo, B. H. Nature 2000, 404, 588. (4) Zhao, B.; Moore, J. S. Langmuir 2001, 17, 4758. (5) Harmon, M. E.; Tang, M.; Frank, C. W. Polymer 2003, 44, 4547. (6) Yu, C.; Mutlu, S.; Selvaganapathy, P.; Mastrangelo, C. H.; Svec, F.; Fre´chet, J. M. J. Anal. Chem. 2003, 75, 1958. (7) Wang, P. C.; DeVoe, D. L.; Lee, C. S. Electrophoresis 2001, 22, 3857. (8) Kuo, T. C.; Cannon, D. M.; Chen, Y.; Tulock, J. J.; Shannon, M. A.; Sweedler, J. V.; Bohn, P. W. Anal. Chem. 2003, 75, 1861.

prepare membranes with pH-responsive9-12 or temperatureresponsive permeability.13 Here we report on the photograft copolymerization of N-isopropylacrylamide (NIPAAm) on poly(ethylene terephthalate) (PET) track-etched membranes. PolyNIPAAm is a wellknown thermoresponsive polymer with a lower critical solution temperature (LCST) of 32 °C in aqueous solution.14,15 By using track-etched membranes with capillary pores having a narrow size distribution and functionalization conditions that lead to an even surface coverage of the pore walls,12 a characterization of the grafted polymer layers, in particular, of their effective thickness and of its variation with temperature, should be possible by means of permeability measurements and calculations based on simple pore flow models. To facilitate a surface-selective grafting-from functionalization, two variants of photoinitiator precoating have been selected in this study: combinations of the as-received PET surface (having moderate surface carboxyl density) with benzophenone and of an aminated PET surface with benzophenone carboxylic acid.12 However, when characterizing surfaces in contact with aqueous solutions, it is also necessary to consider the additional interactions caused by the surface charge.16 A very useful method for electrokinetic investigations is the determination of the pH(9) Ulbricht, M. React. Funct. Polym. 1996, 31, 165. (10) Peng, T.; Cheng, Y. L. J. Appl. Polym. Sci. 1998, 70, 2133. (11) Yang, B.; Yang, W. J. Macromol. Sci., Pure Appl. Chem. 2003, 40, 309. (12) Geismann, C.; Ulbricht, M. Macromol. Chem. Phys. 2005, 206, 268. (13) Liang, L.; Feng, X.; Peurrung, L.; Vishwanathan, V. J. Membr. Sci. 1999, 162, 235. (14) Zhang, X. Z.; Yang, Y. Y.; Chung, T. S.; Ma, K. X. Langmuir 2001, 17, 6094. (15) Mao, H.; Li, C.; Zhang, Y.; Bergbreiter, D. E.; Cremer, P. S. J. Am. Chem. Soc. 2003, 125, 2850.

10.1021/la0603774 CCC: $37.00 © 2007 American Chemical Society Published on Web 07/01/2006

Characterization of Capillary Pore Membranes

dependent zeta potential (ζ), which gives valuable information on the surface charge and its dependency on the surface functionality and solution conditions.16-19 In capillaries and in membrane pores, these effects are more complex because of the overlap of electrical double layers due to the small dimensions. Such effects have already been analyzed theoretically and experimentally,20-27 but the influence of additional surface-grafted polymer hydrogel layers has not yet been investigated in detail. In this work, we show that nanoscale phenomena with a significant impact on the system’s properties, such as the “electroviscous effect” (refs 20-22), which could be interpreted in the framework of the space-charge model,28 and the saltingout effect, need to be considered when characterizing the grafted polymer layers. The synthesized membranes turned out to be an excellent system to use in investigating the correlations between the synthesis and the structure of macromolecular architectures on the micro- and nanoscale. 2. Experimental Section 2.1. Materials. PET track-etched capillary pore membranes with nominal pore diameters of 400 and 1000 nm were purchased from Oxyphen (Dresden, Germany). Solvents and reagents of analytical grade were commercial products and were used as received. N-Isopropylacrylamide (NIPAAm) was purchased from Acros (Geel, Belgium) and used without further purification. Diethylamine, benzophenone (BP), and 4-benzoylbenzoic acid (BPC) were from Fluka (Steinheim, Germany) and were of the highest purity available. Water purified with a Milli-Q system from Millipore (Eschborn, Germany) was used for all of the experiments. 2.2. Photoinitiated Graft Copolymerization. Premodification of membranes (yielding aminated PET) and the grafting-from procedure have been described in detail previously.12 Aqueous NIPAAm monomer solutions of 2.5, 5, and 10 wt % were used to synthesize functionalized unmodified and aminated PET membranes. Preadsorbed photoinitiators were BP for the unmodified membranes and BPC for the aminated PET membranes. The photoinitiator concentration (50 mmol‚L-1), adsorption time (60 min), and irradiation time (15 min) were kept constant in all experiments. The thus-functionalized membranes were labeled according to the nominal pore size (400, 1000), the membrane/ photoinitiator combination (U-BP, A-BPC), and the monomer concentration (2.5, 5, 10). The degree of functionalization (DG) for the polyNIPAAm grafted membranes was determined via gravimetry.12 2.3. Chemical Capping of PET Membrane Surface Charge. The procedure for the amidation of carboxyl groups on the asreceived PET surface was similar to the membrane premodification toward an aminated PET surface described before.12 Unmodified PET-400 membrane samples were immersed in a solution of 0.5 g of N,N′-diisopropylcarbodiimide and 1.0 g of N-hydroxybenzotriazole (16) Werner, C.; Ko¨rber, H.; Zimmermann, R.; Dukhin, S.; Jacobasch, H. J. J. Colloid Interface Sci. 1998, 208, 329. (17) Zimmermann, R.; Dukhin, S.; Werner, C. J. Phys. Chem. B 2001, 105, 8544. (18) Chan, Y. H. M.; Schweiss, R.; Werner, C.; Grunze, M. Langmuir 2003, 19, 7380. (19) Meinhold, D.; Schweiss, R.; Zschoche, S.; Janke, A.; Baier, A.; Simon, F.; Dorschner, H.; Werner, C. Langmuir 2004, 20, 396. (20) Levine, S.; Marriott, J. R.; Neale, G.; Epstein, N. J. Colloid Interface Sci. 1975, 52, 136. (21) Bowen, W. R.; Jenner, F. J. Colloid Interface Sci. 1995, 173, 388. (22) Huisman, I. H.; Tra¨gårdh, G.; Tra¨gårdh, C.; Pihlajama¨ki, A. J. Membr. Sci. 1998, 147, 187. (23) Szymczyk, A.; Aoubiza, B.; Fievet, P.; Pagetti, J. J. Colloid Interface Sci. 1999, 216, 285. (24) Vainshtein, P.; Gutfinger, C. J. Micromech. Microeng. 2002, 12, 252. (25) Yaroshchuk, A.; Ribitsch, V. Langmuir 2002, 18, 2036. (26) Yaroshchuk, A.; Zhukova, O.; Ulbricht, M.; Ribitsch, V. Langmuir 2005, 21, 6872. (27) Chun, M. S.; Lee, T. S.; Choi, N. W. J. Micromech. Microeng. 2005, 15, 710. (28) Yaroshchuk, A. E. AdV. Colloid Interface Sci. 1995, 60, 1.

Langmuir, Vol. 23, No. 1, 2007 77 in 50 mL of N,N-dimethylformamide (DMF) and shaken for 30 min at room temperature. After washing with DMF, the samples were shaken for 3 h in a 0.5 wt % solution of diethylamine in DMF, rinsed two times with DMF and ethanol, respectively, and dried at 45 °C overnight. The thus-obtained diethylamine-modified membranes are referred to as 400-U-D. The quantitative carboxyl group determinations for the U and U-D samples were carried out as described earlier.12 2.4. Membrane Characterization. 2.4.1. Pore Size Distribution. The pore size distributions were analyzed with a PMI capillary flow porometer (Porous Materials, Inc., Ithaca, NY). The gas flow was measured as a function of the trans-membrane pressure, first through a dry membrane and then after wetting the membrane with 1,1,2,3,3,3hexafluoropropene (“Galwick”, PMI, surface tension 16 dyn‚cm-1). The pore size distribution was then deduced from the comparison of the two experiments by using the PMI software. The morphology of PET membranes was studied with a LEICA S420 scanning electron microscope (Leica-Cambridge Instruments, Cambridge, U.K.). Prior to the measurement, the membrane pieces (5 × 5 mm2) were fixed on a stub with a graphite adhesive and then sputtered with gold at 15 mA and an electrode distance of ca. 25 mm (Emitech Automatic Coater K550, EM Technologies Ltd., Ashford, Kent, U.K.). 2.4.2. Zeta Potential from Streaming Potential. The cell and experimental setup for the measurement of the trans-membrane streaming potential has been described before.29,30 Experiments were always started in the range of pH 5-6 with a 10-3 M KCl solution prepared with ultrapure water; the other pH values were adjusted by the addition of dilute KOH or HCl solutions. The zeta potential ζ was calculated with the Helmholtz-Smoluchowski equation ζ)

κη ∆E r0 ∆P

(1)

where κ is the liquid conductivity, η is the liquid viscosity, r is the liquid permittivity, 0 is the permittivity of free space, ∆E is the streaming potential, and ∆P is the hydrodynamic pressure difference. 2.4.3. Permeability. Flux measurements were made using stirred cells with 3 mL volume and 1.77 cm2 active membrane area (Amicon Model 8003, Millipore), and the trans-membrane pressure was adjusted by the height of a water reservoir above the membrane (25 cm). The temperature of the reservoir contents was maintained at 25 and 45 °C, respectively, using a thermostat (Julabo, Seelbach, Germany), and the Amicon cell was kept in a water bath also kept at the same temperature. The measurements were carried out with ultrapure water, with the pH adjusted to 2 using HCl, with 10-4 and 10-3 M KCl solutions and with 0.1 M NaCl solution. Permeability resulted from dividing the flux by the trans-membrane pressure, and the pore sizes of the membranes were calculated from the permeability data by using the equation of Hagen-Poiseuille31a V π∆Pr4 ) ∆t 8ηL

(2)

where V is the volume of the permeate relating to a single cylindrical membrane pore, ∆t is the time, ∆P is the trans-membrane pressure, r is the pore radius, η is the viscosity of water, and L is the capillary length (i.e., the membrane thickness). This equation is valid assuming a cylindrical geometry of the membrane pores and equal sizes of all of the pores. The first result is the effective hydrodynamic pore diameter, dh () 2r). The thickness of the grafted layer on the pore wall, lh, can be calculated, assuming an even functionalization, as the difference between the pore radius of the unmodified or premodified membrane and the pore radius of the functionalized membrane. (29) Rodemann, K.; Staude, E.; J. Membr. Sci. 1995, 104, 147. (30) Lettmann, C.; Mo¨ckel, D.; Staude, E. J. Membr. Sci. 1999, 159, 243. (31) Mulder, M. Basic Principles of Membrane Technology, 2nd ed.; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1996; (31a): p 224; (31b): p 181.

78 Langmuir, Vol. 23, No. 1, 2007

Geismann et al. Table 1. Degrees of Functionalization Relative to the Outer Membrane Surface and Specific Surface Area as Well as Resultant Dry Layer Thicknesses of PolyNIPAAm-Functionalized Membranes

membrane 400-U-BP 2.5 400-U-BP 5 400-U-BP 10 400-A-BPC 2.5 400-A-BPC 5 400-A-BPC 10 1000-U-BP 2.5 1000-U-BP 5 1000-A-BPC 2.5 1000-A-BPC 5

Figure 1. Pore size distribution of PET track-etched membranes (nominal pore diameters 400 and 1000 nm) obtained from the pore dewetting as a function of trans-membrane pressure (permporometry; sample graphs for each type of membrane).

Figure 2. SEM image of the outer surface of an unmodified PET 1000 membrane (10 000-fold magnification).

3. Results 3.1. Pore Size Distribution. Inconsistencies between the nominal and real pore sizes have been observed and discussed before for the PET 400 membrane.12 Therefore, for both unmodified PET membranes used in this study, the pore sizes were measured with an independent method based on the dewetting of the membrane pores filled with a liquid of a known surface tension as a function of pressure.31b Six different samples of each membrane were analyzed, and the sample results are shown in Figure 1. It can clearly be seen that both membrane types had very narrow pore size distributions but the pore sizes from the permporometry (PET 400: 692 ( 17 nm; PET 1000: 1629 ( 130 nm) exceeded the nominal pore sizes by more than 60%. An SEM analysis of an unmodified PET 1000 membrane confirmed these results: pore diameters between 1500 and 1700 nm had been observed (Figure 2). Furthermore, the cylindrical pore shape and the very narrow pore size distribution were also confirmed. A few double pores were also seen. Very similar results were earlier obtained for the PET 400 membranes.12 Thus, it is necessary to take into account this systematic deviation in the pore size from the manufacturer data in further calculations of parameters that are based on the membranes’

degree of degree of functionalization calculated dry functionalization per specific surface layer thickness DG (µg/cm2) area DGspec (µg/cm2) lgr (nm) 7 16 31 25 57 214 20 59 26 66

0.2 0.4 0.8 0.6 1.4 5.4 1.8 5.3 2.3 5.9

1.5 3.7 7.2 5.7 13.2 49.2 16.0 48.0 21.3 53.3

pore size, such as the specific membrane surface area, Aspec, and the effective hydrodynamic pore diameter, dh. Assuming pore sizes of 692 and 1629 nm for the unmodified PET 400 and PET 1000, respectively, and by using the corresponding hydraulic permeabilities with negligible nanoscale effects (i.e., those measured with 0.1 M NaCl solution, see section 4.3.1), the pore densities have been estimated to be 7.6 × 107 cm-2 (PET 400) and 8.5 × 106 cm-2 (PET 1000). 3.2. Degree of Functionalization. The degrees of functionalization were estimated from the gravimetry data, and the dry layer thicknesses of the polyNIPAAm layers were calculated, assuming an even coverage of the total surface area of the membrane, isocylindrical pore geometry, and a dry polyNIPAAm density of 1.1 g‚cm-3. This assumption is justified because previous results for analogous photoinitiated functionalizations of porous polymer membranes show that a complete and relatively even coverage of the entire specific surface area can be achieved (e.g. refs 9 and 12). The results are shown in Table 1. For 400-U-BP membranes, the DG value was doubled by increasing the monomer concentration by a factor of 2, whereas for 400-A-BPC membranes this increase was larger for the highest monomer concentration. For both kinds of PET 1000 membranes, doubling the monomer concentration resulted in a 2.5-3 times higher DG. A series of membranes with varying average grafted polyNIPAAm layer thickness in the range of up to 60 nm, in two different pore diameters, was available for the further studies. 3.3. Zeta Potential. Zeta potentials were determined as function of pH in 10-3 M KCl solution in order to evaluate the surface properties of initial and functionalized membrane samples. Figure 3 shows the zeta potential versus pH curves for selected membranes. Zeta potentials at characteristic pH values are given in Table 2. From Figure 3 and Table 2, it can be seen that the surface charge of the unmodified PET 400 membrane was clearly negative at pH 3.6 with a tendency to reach a value of zero (isoelectric point) at still lower pH (approximately 1.8). At higher pH, it became increasingly negative, reaching a plateau value of -23 mV at pH >8. The aminated PET 400 membrane had a positive charge with a zeta potential approaching 35 mV at the lowest pH value. An increase in pH caused a decrease in the zeta potential, and a plateau value of -30 mV was reached between pH 8 and 9. The isoelectric point of this membrane was determined at pH 5.7. The results for the PET 1000 membranes were similar to those for the membranes with a smaller pore size. Only the aminated 1000 nm membrane deviated somewhat from its 400 nm counterpart, but the trends in the graphs were similar. Both polyNIPAAm-functionalized PET 400 membranes showed only very low absolute values of the zeta potential in

Characterization of Capillary Pore Membranes

Langmuir, Vol. 23, No. 1, 2007 79

Figure 3. Zeta potential vs pH for selected PET 400 membranes. Table 2. Zeta Potential of PET Membranes at pH 3.6 and pH 5-6 (10-3 M KCl) as Well as Isoelectric Pointsa membrane 400-U 400-A 400-U-BP 5 400-A-BPC 2.5 1000-U 1000-A 1000-A-BPC 2.5

ζ at pH 3.6 (mV) ζ at pH 5-6 (mV) isoelectric -12.7 35.1 -0.2 0.2 -13.0 20.1 0.3

-17.9 2.8 -0.8 -0.1 -21.5 -15.2 -0.9

1.8 5.7 n/a n/a 2.4 4.9 n/a

a Data beyond the analyzed pH range have been estimated by extrapolation via curve fitting.

the investigated pH range. At pH values above 6, all membranes had a negative zeta potential. However, at low pH a very small but significant difference was seen: the polyNIPAAm grafted membranes based on the unmodified PET had a negative value and the ones based on the aminated PET had a positive value (Table 2). 3.4. Pure Water Permeability. As expected considering the specifications, the membrane permeabilities were high (e.g., for the unmodified PET 400 55 600 L‚m-2‚h-1‚bar-1 and for the unmodified PET 1000 249 500 L‚m-2‚h-1‚bar-1, both at 25 °C). The effective hydrodynamic pore diameters of modified PET 400 and PET 1000 membranes, calculated by using the HagenPoiseuille equation from the water permeabilities measured with ultrapure water at 25 and 45 °C, reveal a very pronounced temperature-responsive switching effect (Table 3). Indeed, the effective hydrodynamic pore diameter of each unmodified membrane had the same value at 25 and 45 °C, respectively. On the contrary, both modified membranes had a significantly smaller pore diameter at 25 than at 45 °C. At 45 °C, the pore diameters of the polyNIPAAm-functionalized membranes were of the same magnitude as their precursor membrane, but an unusual effect was the increase for 400-U-BP 10 membrane as compared to that for 400-U. Also for the other

membranes with grafted polyNIPAAm layers, the pore sizes at 45 °C seemed to increase instead of becoming systematically smaller with increasing amounts of the grafted polymer. For the 400-U-BP 2.5 membrane with a very low DG, such an effect was even observed at 25 °C. The origin of this effect will be analyzed in section 3.5. 3.5. Influence of Membrane Surface Charge and Solution Ionic Strength on Permeability. 3.5.1. Base Membranes for Grafting and the Diethylamine-Modified Membrane. Unexpected differences in the permeabilities had been observed for the base membranes when using ultrapure water or aqueous solutions (Table 4). Unmodified membranes were reacted with diethylamine in order to eliminate fixed surface charges. Carboxyl groups on the surface of the membranes before and after such a “capping” reaction were characterized by reversible ionic dye binding, and pore sizes were calculated from the permeability data measured with different solutions (Table 4). Unmodified PET 400 membranes contained a significant number of carboxyl groups. This number was considerably decreased (to about 14%) by the amidation reaction with diethylamine. No carboxylic groups could be detected on membranes after amination. The hydrodynamic pore diameters calculated from the permeabilities measured with pure water were consistently smaller than those from measurements with pH 2 and 0.1 M NaCl solutions. The same effect but of a smaller magnitude had also been observed for the unmodified PET 1000 membrane. For all three membranes with a nominal pore diameter of 400 nm, the data for pH 2 and 0.1 M NaCl were identical within the range of error. It should be noted that, because of the addition of hydrochloric acid for pH adjustment, the ion strength of the pH 2 solution had been ca. 10% of that of the 0.1 M NaCl solution. Therefore, the effect of solution pH could not be separately evaluated, and these data may be considered to be characteristic of moderate ionic strength. However, the pore diameters of both the unmodified (carboxylic) and the aminated PET membranes estimated from the pure water permeability were significantly smaller than the ones for the diethylamine-modified (carboxyl-capped) membrane. 3.5.2. Grafted Membranes at High Salt Concentration. The permeabilities at 25 and 45 °C of the membranes grafted with polyNIPAAm had been measured using aqueous solutions with a salt concentration of 0.1 M. Calculated pore sizes as a function of pore size reduction from DG are shown in Figure 4. In contrast to the data obtained with pure water (section 3.4), the pore sizes became consistently smaller with increasing DG. Furthermore, the curves were identical for the grafted membranes prepared using the two different (unmodified and aminated) base membranes. Similar data were obtained for the PET 1000 membranes (see Discussion, section 4.3). 3.5.3. Base and Grafted Membranes at Varying Salt Concentration. To further elucidate the effects of salt concentration on the permeability and hence on the hydrodynamic pore diameter, the KCl concentration was varied (the same electrolyte was used

Table 3. Permeabilities and Effective Hydrodynamic Pore Diameters of Unmodified and Two Selected PolyNIPAAm Modified Membranes (PET 400 and PET 1000, Respectively) at 25 and 45 °C (Pure Water)

membrane

permeability at 25 °C J/p (L/m2 h bar)

permeability at 45 °C J/p (L/m2 h bar)

hydrodynamic pore diameter at 25 °C dh,25 (nm)

hydrodynamic pore diameter at 45 °C dh,45 (nm)

400-U 400-U-BP 10 1000-U 1000-U-BP 5

55 567 15 571 249 527 32 191

81 776 95 626 373 254 303 129

642 467 1598 958

639 665 1598 1517

80 Langmuir, Vol. 23, No. 1, 2007

Geismann et al.

Table 4. Carboxyl Group Concentrations and Hydrodynamic Pore Diameters (Calculated from the Permeability Data) of Unmodified and Diethylamine-Modified Membranes

a

membrane

carboxyl group concentration cCOOH (pmol/cm2)

hydrodynamic pore diameter at 45 °C, H2O, dh,water (nm)

hydrodynamic pore diameter at 45 °C, pH 2, dh,pH2 (nm)

hydrodynamic pore diameter at 45 °C, NaCl 0.1 M, dh,NaCl (nm)

400-U 400-U-D 400-A 1000-U

42 6 0a 47

639 ( 4 674 ( 4 635 ( 5 1598 ( 32

691 ( 4 679 ( 4 685 ( 4 n.a.

692 ( 4 680 ( 5 688 ( 4 1629 ( 31

Note that the fixed surface charge is different from zero. The surface amino group concentration of this membrane is 348 pmol/cm2.

Figure 4. Pore size from permeabilities with 0.1 M NaCl solution vs pore size reduction calculated from DG.

Figure 5. Hydrodynamic pore diameters of selected PET 400 membranes at different salt concentrations at 25 °C.

for the streaming potential measurements; see Figure 3 and Table 2), and the results are summarized in Figure 5. The dependency on the salt concentration was different for the base PET and the g-polyNIPAAm PET membranes. For the unmodified (carboxyl) and the aminated membranes, the apparent pore size was significantly smaller only for pure water; the values for the other salt concentrations were within the range of uncertainty. For the carboxyl-capped membranes (PET 400-UD), the variations were somewhat larger, but the value for ultrapure water was significantly higher than for the precursor membrane PET 400-U (Table 4) and no significant trend as a function of salt concentration was seen. For the g-polyNIPAAm PET membranes at a lower DG (400-U-BP 5; Table 1), only very small effects were detected (the maximum difference in pore diameter was 13 nm). However, at a higher DG (400-A-BPC 5) a systematic decrease in hydrodynamic pore diameter was observed with decreasing salt concentration (from 260 nm at 0.1 M to 223 nm for pure water).

monomer concentration. This was expected and is consistent with the previous results.12 The premodification of the base polymer with a multifunctional alkylamine considerably enhanced the efficiency of the subsequent graft copolymerization via UVinitiated hydrogen abstraction because of the significantly increased number of accessible reactive groups. However, in this work a linear dependency of DG on the monomer concentration was found for the U-BP membranes, and a larger (exponential) increase was found for the A-BPC membranes (Table 1). This can be explained by a less distinct competition between the respective photoinitiator and the neutral monomer, compared to that in earlier work,12 where the photoinitiator could be desorbed from the surface because of the ion exchange with deprotonated acrylic acid. Furthermore, under identical functionalization conditions, the DG relative to the specific surface area was always much larger (by a factor of 4 to 13) for the PET 1000 membranes as compared to that for the PET 400 membranes. This can be explained as a pore size effect. In a larger pore volume, a larger absolute amount of monomer is present and therefore immediately available for the growth of the grafted polymer chains. Considering the different surface-to-volume ratios for the two membranes and the relatively short UV irradiation (reaction) time, a stronger limitation of the grafting reaction due to a reduced local monomer concentration and slow mass transfer (monomer diffusion into the pores) could occur for the PET 400 membranes. 4.2. Zeta Potential. The pore surface of the unmodified PET membrane was negatively charged, as shown in Figure 3 and Table 2. This can be explained by the significant concentration of surface carboxyl groups (Table 4). This surface charge was changed in an expected way by the amination premodification. However, at high pH values, both surfaces exhibited about the same negative zeta potential. This can be explained by the only moderate carboxyl density for the unmodified membranes and by the effect of anion (hydroxyl) adsorption17-19,32 onto the pore walls of the membranes. By completely covering the pore surface with g-polyNIPAAm, the absolute values of the zeta potentials became very small. Such an effect, a “hydrodynamic screening” of the surface, had also been observed before for other surfaces and neutral hydrogels. It had been explained by either a reduction in the number of adsorption sites for the anions or/and a reduced attractive force for this adsorption18,33 or rather by a complete hydrodynamic immobilization within the hydrogel layer. However, it should be noted that at low pH values a subtle but significant difference was seen between the polyNIPAAm grafted layers on the unmodified (carboxyl) and the aminated PET surfaces (Table 2). Therefore, despite the very effective hydrodynamic shielding by the (thin) grafted polymer hydrogel layers that reduced the absolute zeta potential values to less than 1% relative to the respective base material, a fixed surface charge (carboxyl vs amino group) on that base material could still be

4. Discussion 4.1. Degree of Functionalization. For all of the membranes, an increase in DG was observed as a consequence of increasing

(32) Zangi, R.; Engberts, J. B. F. N. J. Am. Chem. Soc. 2005, 127, 2272. (33) Kreuzer, H. J.; Wang, R. L. C.; Grunze, M. J. Am. Chem. Soc. 2003, 125, 8384.

Characterization of Capillary Pore Membranes

Langmuir, Vol. 23, No. 1, 2007 81

Table 5. Apparent Increase in Viscosity at 25 and 45 °C for Unmodified and Aminated PET 400 and Unmodified PET 1000

membrane 400-U 400-A 1000-U

increase in viscosity at 25 °C (%)

increase in viscosity at 45 °C (%)

34 27 4

37 38 8

detected. This can be explained by the initiation mechanism of the graft copolymerization. The carboxyl and amino groups on the PET, “hidden” below the grafted neutral polymer hydrogel, had to be essentially preserved.12 4.3. Effective Hydrodynamic Pore Diameters and Grafted Polymer Layer Thicknesses. 4.3.1. Base Membranes. To elucidate the reasons for the obvious inconsistencies when calculating pore diameters from permeabilities measured with ultrapure water (section 3.4), the impact of the membrane surface charge and the solutions’ ionic strength on this characterization method has been investigated. A PET membrane surface with “capped” carboxyl groups has been prepared by the reaction of the unmodified PET membrane with diethylamine. The results of permeability measurements for the 400-U and the 400-U-D membranes (Table 4 and Figure 5) show clearly that both membrane surface charge and the ionic strength of the solution have a significant impact on the permeability data. Similar effects have already been observed in our earlier study while investigating the effect of solution pH, but the interpretations were focused on the pH sensitivity of the grafted polymer poly(acrylic acid) on the pore walls of the PET 400 membrane. However, the permeability of the unmodified PET membrane also increased when the pH decreased from 7 to 2, and that effect was tentatively related to a reversible swelling of PET.12 In this study, a significant (and almost identical) increase in permeability was seen either with ultrapure water when the surface charge had been reduced by capping surface carboxyl groups (note that the capping efficiency was only about 86%; Table 4) or when the NaCl concentration was increased to 0.1 M. Moreover, the salt concentration threshold to reach the maximum permeability was very low (10-4 M; Figure 5). Considering the magnitude of the effect and the parameters having an influence, a change in the pore geometry of the PET membranes can be excluded. A more plausible explanation is the so-called “electroviscous effect”.20-23 In pure water at pH >5, the carboxyl groups on the unmodified PET membrane surface are mainly deprotonated, and the surface charge is negative (Figure 3 and Table 2). A solution with low ionic strength has a large Debye length. The consequence for the porous membrane is an increase in the apparent viscosity22 and thus a reduction in the pore size calculated from the permeability. The apparent pore size of the carboxyl-capped membrane in ultrapure water is close to the pore size estimated with 0.1 M NaCl solution, where a much lower Debye length causes a decrease in the apparent viscosity to a value that cannot be distinguished from that of the bulk solution. According to ref 22, the pore diameters calculated from the permeability measurements with a high salt solution (∼0.1 M) represent the real pore sizes best because all chargebased interactions with the surface are minimized. The magnitude of the electroviscous effect had been expressed as an apparent viscosity increase that had been calculated from the ratio of the permeabilities for the ultrapure water and 0.1 M NaCl solution (Table 5). Figure 6 shows the results of sample calculations of relative electroviscosity in NaCl solution within the scope of the standard space-charge model (identical straight cylindrical pores).28 It is

Figure 6. Results of sample calculations of relative electroviscosity in NaCl solution within the scope of the standard space-charge model (identical straight cylindrical pores with diameters of 692 and 1629 nm, respectively, see section 3.1, and a surface carboxylate concentration of 45 pmol/cm2, see Table 4).

seen that in very dilute solutions the experimentally observed phenomenon’s magnitude (∼37% relative difference in the hydraulic permeabilities in very dilute vs more concentrated solutions) is reached and even slightly exceeded. It is also seen that at 10-4 M concentration the phenomenon becomes too weak to be reliably observed experimentally, in agreement with our measurements in 10-4 M solutions. The Figure also shows the relative electroviscosity calculated for the membrane with the larger pores (assuming the same surface charge density as for the smaller pores). In this case at a concentration where a value of 37% is reached for the membrane with smaller pores, the estimated relative electroviscosity turns out to be essentially larger than the experimentally observed value (some 21% instead of 8%). One of the possible explanations for this is the fact that the pore size distribution was considerably broader in the case of a membrane with larger pores (standard deviation of 8% vs 2.5% in the case of a membrane with smaller pores). Indeed, the electroviscosity is caused by electro-osmosis in the field of streaming potential. The latter becomes smaller as a result of the electrical “short-cut” through smaller pores, which have smaller hydraulic permeabilities but larger electric conductivities. 4.3.2. PET Membranes with Grafted PolyNIPAAm. The expected temperature-responsive switching effect was very clearly observed (Table 3). With a monomer concentration of 5 wt % in the functionalization step for the PET 1000 membrane, a somewhat larger effect (i.e., a reduction in the effective pore size by ∼37%) has been achieved as compared with that of the 10 wt % monomer used for the PET 400 membrane (reduction by ∼30%). This higher efficiency, however, is due to the 2-timessmaller inner surface area of the PET 1000 membrane (calculated from the permporometry results and the NaCl 0.1 M permeability data, section 4.3.1), which leads to an approximately 2-foldlarger DG normalized by the specific surface of the PET 1000 membrane (Table 1). Therefore, the extent of reversible pore blocking in the range of several hundred nanometers could be very well adjusted by the functionalization conditions adapted to the pore geometry of the respective membrane. However, a more detailed analysis of the permeability data with respect to this effective layer thickness is necessary. As shown in Figure 4, at 0.1 M salt concentration an unambiguous correlation between the DG and the pore diameter reduction had been achieved. More specifically, a clear correlation

82 Langmuir, Vol. 23, No. 1, 2007

Geismann et al.

Figure 7. Correlation between the dry layer thickness from DG and the wet layer thickness of g-polyNIPAAm in the swollen and collapsed states, respectively, from the permeability measurements with 0.1 M NaCl solution.

between the dry layer thickness, calculated from the gravimetrically determined degree of functionalization, and the effective hydrodynamic layer thicknesses from the 0.1 M NaCl solution permeability data could be obtained for both types of base membranes, irrespective of the method used for the surface functionalization (Figure 7). Membrane 400-A-BPC 10 was excluded from this analysis because the extent of pore blocking was too large. The slopes of the graphs in Figure 7 represent the swelling ratios of the grafted polymer layer compared to the dry state. It can be seen that for PET 400 membranes the slope for the grafted polyNIPAAm in the wet collapsed state appears to be about 1. This could be explained by a relatively low grafting density of the polyNIPAAm chains, and considering the photoinitiation mechanism, this is a plausible assumption.12,34 Note also that the initiation of the graft copolymerization would not consume the surface functional groups such as carboxyl or amino. This low density should give rise to dry polymer globules that, upon contact with water, would expand in both the lateral and vertical directions. This would be different from a polymer “brush” that would preferentially expand in the vertical direction. A consequence of such a looser “mushroom” structure would be a higher susceptibility to shear than for a “brush” structure (e.g., ref 35). However, for PET 1000 membranes the slope is even lower than 1. An explanation for this could be the much smaller ratio of the inner surface area to the outer surface area of this membrane, as compared to that for PET 400 (6 vs 27). Hence, an uneven distribution of grafted polyNIPAAm (less on the pore walls as compared with that on the outer, UV-exposed surface) would have larger consequences for PET 1000, and this could explain the swelling that is too low as deduced from the permeability data. However, when measuring the changes in the permeability for identical samples, these effects have no impact on the accuracy of the temperature-responsive swelling ratios (below). The swelling ratio increased with decreasing temperature by a factor of 16 (wet vs wet) for PET 400 membranes and a factor of 12 (wet vs wet) for the PET 1000 membranes. This clearly indicates the large flexibility and significant (adjustable) average chain length of the grafted polyNIPAAm. These properties can again be related to the initiation mechanism of the graft (34) Ulbricht, M.; Oechel, A.; Lehmann, C.; Tomaschewski, G.; Hicke, H. G. J. Appl. Polym. Sci. 1995, 55, 1707. (35) Castro, R. P.; Monbouquette, H. G.; Cohen, Y. J. Membr. Sci. 2000, 179, 207.

Figure 8. Correlation between the wet layer thickness of gpolyNIPAAm in the swollen state (from the permeability measurements with 0.1 M NaCl solution) and the swelling ratio for this layer (the wet layer thickness in pure water was calculated with the hydrodynamic pore diameter of the unmodified membrane obtained with 0.1 M NaCl solution as a reference value). Data for grafted PET 400 membranes at 25 °C.

copolymerization because the surface-selective photoinitiation clearly favors the formation of linear grafted polymer at a minimized interference by the polymerization in solution, and under the conditions used, the average chain length increases with increasing monomer concentration. Furthermore, the g-polyNIPAAm layer thickness in the swollen state is also influenced by the salt concentration in solution (Figure 5). However, this effectsa systematic decrease in thickness with increasing salt concentrationsis different from the influence of salt concentration on the base membranes’ permeability where a threshold concentration of 10-4 M had been found, which was considered to be characteristic of the electroviscous effect (section 4.3.1). Extreme values (under our experimental conditions) for this swelling of the grafted polyNIPAAm layer (expressed as its thickness in water vs its thickness in 0.1 M NaCl solution) are shown for the PET 400 membranes (Figure 8). One can see that the absolute values are between ∼1.7 and ∼1. Hence, this effect is much smaller than the temperatureinduced volume transition, where an average ratio of 16 was

Characterization of Capillary Pore Membranes

found (Figure 7). A reduced degree of swelling with increasing salt concentration for the nonionic polyNIPAAm resembles the salting-out effect that had been related to the changes in the inter- and intramolecular hydrogen bonding and in polar interactions as well as in hydrophobic interactions due to the changes in the solvent (water) structure.36 For polymers with a LCST, this salt effect is typically much smaller than the temperature-induced volume transition, and almost no systematic studies of this phenomenon have yet been performed. However, very recently and by using a special setup for precise temperature measurements, a significant effect of the salt concentration and especially of the type of anion, described by the Hofmeister series, on the LCST of polyNIPAAm has been observed.37 Among the salts studied, NaCl had a moderate effect (decrease by 1 K for an increase of 0.2 M). The data in Figure 7 suggest that such effects could also be studied with grafted polyNIPAAm on the pore walls of PET membranes. Additionally, the decrease in the observed swelling ratio with increasing DG may indicate that the conformational flexibility of the grafted chains in the hydrogel layer (which decreases with increasing layer thickness) could have a significant effect as well. Significant differences in the swelling ratios in the expanded state can also be seen for the different base membrane specifications: larger values are found for the PET 400 membranes (Figure 7). It can be concluded that the membrane pore size also has an influence on the structure of the grafted polymer in the swollen state. The representative values for the linear flow rate through the pores under the experimental conditions (section 2.4.3), calculated for the unmodified membranes at 25 °C, are 0.4 cm‚s-1 for the PET 400 membranes and 1.0 cm‚s-1 for the PET 1000 membranes. Therefore, the effective layer thickness at the same surface density of grafted polymer may be smaller for the PET 1000 membranes, presumably because the shear stress imposed during the flow measurement is larger in the larger pores. (36) Park, T. G.; Hoffman, A. S. Macromolecules 1993, 26, 5045. (37) Zhang, Y.; Furyk, S.; Bergbreiter, D. E.; Cremer, P. S. J. Am. Chem. Soc. 2005, 127, 14505.

Langmuir, Vol. 23, No. 1, 2007 83

5. Conclusions Track-etched capillary pore membranes with two different pore sizes were functionalized with a temperature-responsive polymer via UV-initiated graft polymerization and characterized with established methods. The results were qualitatively comparable to the results obtained in the previous work, where pHsensitive membranes had been created. A primary functionalization (amination) of PET 400 membranes in combination with a photoinitiator of opposite charge led to higher graft copolymerization efficiencies. For PET 1000, this efficiency increase was also found but was less distinct. This study revealed clearly that electrokinetic and shear effects must be considered when interpreting the results of permeability measurements in terms of pore diameters. Further investigations of these effects are being carried out in ongoing work. Along with a precise characterization of the isoporous structure and the pore diameter of the base membranes using complementary methods, it is possible to determine the effective thickness of the grafted polymer layers on the pore walls and their response to stimuli (in this study temperature) as well as solvent conditions (in this study, the salt concentration). Those quantitative investigations will be extended to other polymer systems as well as other stimuli and solvent conditions. Overall, the feasibility of the capillary pore membranes for a detailed evaluation of the relationships between the synthesis, the structure, and the function of grafted polymer layers has been proven again, here for a temperature-responsive polymer with dimensions between a few nanometers and several hundred nanometers. Acknowledgment. The financial support of the Deutsche Forschungsgemeinschaft (DFG; grant Du 128/14-1) via a Mercator Guest Professorship for A.Y. at the Universita¨t Duisburg-Essen is gratefully acknowledged. LA0603774