Examining the electrochemical properties of a nanofiltration
Journal of Membrane Science 276 (2006) 286–294
Examining the electrochemical properties of a nanofiltration membrane with atomic force microscopy Jonathan A. Brant a,∗ , Kelly M. Johnson b , Amy E. Childress b a
b
Rice University, Department of Civil and Environmental Engineering, MS 317, Houston, TX 77251, USA University of Nevada, Reno Department of Civil and Environmental Engineering, MS 258, Reno, NV 89557, USA Received 21 May 2005; received in revised form 23 September 2005; accepted 3 October 2005 Available online 2 November 2005
Abstract In this investigation, two methods for characterizing membrane surface potential are investigated. Results from atomic force microscopy (AFM) analyses are compared with streaming potential measurements. In calculating surface potential from AFM force measurements, assumptions of constant charge and constant potential were both considered for modeling electrostatic interactions. For a ceramic mica surface, the constant charge assumption was found to be most appropriate while for a polymeric membrane surface, the constant potential assumption provided results that agreed better with theoretical expectations. For both the mica and membrane surfaces, results from AFM agreed with the measured values determined from streaming potential analysis. The advantage of AFM is that in addition to determining the mean surface potential value for membrane surfaces, this technique provides a spatially resolved measure of charge distribution. One drawback of the technique is that it is sensitive to surface roughness, as the measured charge distribution increased with increasing surface roughness. © 2005 Elsevier B.V. All rights reserved. Keywords: AFM; Zeta potential; Membrane; Streaming potential; Nanofiltration
1. Introduction Nanofiltration (NF) membranes are used in a wide range of drinking water, wastewater, and industrial applications [1,2]. Separation by NF membranes occurs primarily due to size exclusion and electrostatic interactions [1–3,4]. For colloids and uncharged molecules, sieving or size exclusion is most responsible for separation; for ions and charged organics, electrostatic interactions are responsible for separation [2,5,6]. For all applications, membrane charge characteristics play a significant role in the transport of water and solute molecules through the membrane. Additionally, the interaction of colloids and charged macromolecules with the membrane, and subsequent fouling of the membrane, is dependent on the charge properties [7]. Because of this, the availability of a simple, reproducible, standardized method of measuring membrane charge properties is of critical importance. In experimental investigations of membrane charge, streaming potential measurements have typically been used to calcu-
∗
Corresponding author. Tel.: +1 713 348 3374; fax: +1 713 348 5203. E-mail address: [email protected] (J.A. Brant).
0376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2005.10.002
late zeta potential. Streaming potential is the potential induced when an electrolyte solution is pumped across a stationary, charged surface. Streaming potential can be used to calculate zeta potential using the Helmholtz–Smoluchowski equation. Several works on streaming potential measurements of NF membranes (e.g. [1,4,8–12]) have appeared in the literature. Although streaming potential measurements are the most frequently used method for evaluating charge properties, they have also been criticized. Results from prior studies reveal uncertainty in individual measurements as well as data scatter [13]. The Helmholtz–Smoluchowski relationship used to calculate zeta potential breaks down at very high or low ionic strengths [14]. Differences in instrument design and the lack of a calibration standard for streaming potential analyzers makes comparison of data among laboratories challenging. An additional concern is that membrane surfaces are heterogeneous—both physically and chemically and for rough or chemically heterogeneous surfaces, surface potential calculated from streaming potential measurements may provide an incomplete description of the surface’s charge characteristics [15]. For example, because zeta potential is an average value of the potential at some distance away from the surface, hydrodynamic effects due to surface roughness may distort zeta potential results [16,17]. Furthermore, streaming
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
potential measurements cannot account nor describe the charge distribution across a surface as it only provides an average zeta potential value for the whole surface. Nevertheless, the mean value provided by streaming potential measurements serves as a reasonable basis from which to compare values determined using other techniques. A dramatic advancement in membrane research is the ability to directly measure the interfacial interaction between a probe and a membrane surface using atomic force microscopy (AFM) [18]. AFM was originally developed as an imaging tool with atomic-level resolution; however, its operating principle has always been based on measuring the force between a surface and a small probe [18,19]. Thus, AFM may be used to measure the interaction force as a function of separation distance between two surfaces [20–22]. The chemical interactions thought to be important in colloid–surface interactions are generally assessed using the classical Derjaguin–Landau– Verwey–Overbeek (DLVO) theory [23,24], which describes the interaction between a colloid and a surface as the interplay between Lifshitz–van der Waals (LW) attractive interactions and electrostatic double layer (EL) repulsive interactions. In the last decade, several studies have focused on comparing surface charge data obtained from AFM to surface potential data determined from electrokinetic measurements (e.g. [25–30]). However, only one of these studies involved polymeric membranes. Bowen et al. [30] modeled the interaction between a colloid probe and a commercial NF membrane surface using membrane zeta potential as a fitting parameter. At low ionic strength (10−3 M), AFM surface potential measurements were in good agreement with the zeta potential calculated from streaming potential. However, at higher ionic strength (10−1 M), when the surface roughness approached the Debye length, poor agreement was seen between the zeta potential derived from the AFM data and that determined from streaming potential. Surface roughness was identified as one possible source of error when comparing zeta potential results determined from AFM force curves and streaming potential measurements. The advantage of the AFM method is that it is possible to directly measure the impact of surface roughness on the zeta potential across the surface, which cannot be done through measuring streaming potential. Additionally, as AFM measurements are carried out at different locations across a surface, having relatively small areas of interaction, they may be used to characterize surface charge distribution. This represents a potentially valuable advancement over streaming potential measurements. The observed discrepancy between the AFM data and the streaming potential results observed by Bowen et al. [30] may also be attributed to the presence of non-DLVO interactions at shorter separation distances. Such additional types of interactions at the membrane surface may include short-ranged acid–base (AB) (electron donor/electron acceptor) interactions that occur between two surfaces immersed in a polar solvent [31]. These interactions may be accounted for using an extended form of the DLVO model (i.e., the XDLVO approach) [32]. In recent investigations, the XDLVO approach has been shown to be successful in describing membrane-colloid interactions [31,33,34]. The XDLVO approach describes the total interac-
287
tion between two surfaces in aqueous media according to [32]: U XDLVO = U LW + U EL + U AB
(1)
where UXDLVO is the total interaction energy between the membrane and colloid immersed in water, ULW is the Lifshitz–van der Waals interaction term, UEL is the electrostatic interaction term, and UAB is the acid–base interaction term. From a thermodynamic standpoint, attraction or adhesion between two interacting surfaces occurs when UXDLVO is negative; repulsion occurs when UXDLVO is positive. The total interaction energy is often evaluated as a function of separation distance between the interacting surfaces using interaction energy profiles. Interaction energy profiles illustrate the type of interaction (attractive or repulsive) occurring as a colloid approaches a membrane surface and the separation distance at which the individual interaction energies are dominant. The main objective of this investigation was to determine if AFM is a feasible method for measuring the surface potential of a polymeric membrane surface. A secondary objective was to determine the most appropriate assumption for modeling electrostatic interactions for the surface within the framework of an XDLVO model. In doing this, short-range interactions that have not previously been accounted for were considered so that a more accurate zeta potential could be determined from AFM analyses. First, the interactions between an AFM probe having a well-defined surface chemistry and geometry and a relatively smooth NF membrane were studied; second, surface potentials determined from the AFM force measurements were compared with those determined from streaming potential measurements. In parallel, the use of AFM to determine the surface potential of mica was also evaluated. The purpose of this was to compare surface potential results for a relatively smooth, homogeneous surface with results for a heterogeneous membrane surface as well as to determine the variability of surface charge across the membrane surface. Because variation in charge density across a membrane surface is often cited as a reason for preferential deposition of foulants on membranes [10,35], a spatially resolved measurement of membrane charge provides valuable insight into membrane fouling by determining whether the variation is substantially large as to alter membrane-solute interactions compared to the mean charge value. 2. Materials and methods 2.1. Surfaces Two surfaces were chosen for study: muscovite mica and a relatively smooth NF membrane. Mica was chosen because its surface is molecularly smooth and its chemistry is well characterized [27,29,36]. The muscovite mica used in this investigation was research grade (SPI Supplies, West Chester, PA). It was stored in a vacuum dessicator and was used with no preparation. The NF membrane was a thin film composite polyamide membrane, designated HL by the manufacturer (Osmonics, Vista, CA). The membrane was stored at 5 ◦ C in ultrapure water obtained from a Millipore (Burlington, MA) water purification
288
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
system. Individual membrane samples were equilibrated in test solutions for 24 h prior to AFM and streaming potential measurements. 2.2. Colloid probes Colloid probes (NovaScan Technologies, Ames, IA) were formed by gluing glass microspheres to the end of tipless silicon nitride cantilevers [37]. The glass microspheres had a diameter of 6 m, which was verified using a scanning electron microscope (SEM). The glass probes were functionalized with carboxyl (COOH) groups arranged in a monolayer. The result was a carboxylated colloid probe having a known surface chemistry and a well-defined geometry. 2.3. XDLVO interaction profiles Contact angles for the mica and membrane surfaces were measured with three well-characterized probe liquids (ultrapure water, formamide, and bromonaphthalene). Triplicate contact angles were measured on no less than five different samples for each surface–probe liquid combination. Contact angle values for a monolayer terminating in carboxyl groups were taken from Drelich [38] and were used to calculate the surface energy properties for the carboxylated AFM probe. Solid surface energies were calculated from the measured contact angles using the Young–Dupr´e equation [32]: γ1 (1 + cos θ) = 2 (2) γlLW γsLW + γl+ γs− + γl− γs+ where θ is the measured contact angle; γ LW is the van der Waals free energy component; γ + is the electron-acceptor component; γ − is the electron-donor component; and the subscripts l and s designate the liquid and solid phases, respectively. The values for the van der Waals and acid–base components were then used to calculate ULW and UAB in terms of force normalized to probe radius as a function of separation distance. Detailed expressions for these equations and their derivations have been outlined elsewhere (e.g. [31,32]). The electrostatic component, UEL , was calculated using the limiting assumptions of constant charge and constant potential as outlined in Elimelech et al. [39]. In calculating UEL , the solution ionic strength and the zeta potential of the colloid probe surface were used as input parameters, while the zeta potential of the sample surface served as the only variable. In this regard, the zeta potential of the sample surface was the fitting parameter for fitting the XDLVO interaction profile to the AFM force curves. Thus, the zeta potential of the sample surface was determined at each location where an AFM force curve was collected.
IA) was 0.12 N/m. This value was used for converting cantilever deflection to interaction force values. Force curves are initially reported as force as a function of scanner position by the AFM instrumentation. Scanner position was converted to separation distance (h) by determining the onset of constant compliance between the scanner position and cantilever deflection (i.e., where cantilever deflection becomes a linear function of piezo scanner position) and subtracting this value from all other scanner position values [40]. Measurements were conducted in a liquid cell at 20 ◦ C with −2 10 and 10−3 M KCl at pH 10. Prior to the measurements, the AFM liquid cell was cleaned with acetone, rinsed with ultrapure water, and blown dry using a high velocity stream of nitrogen. Force curves were collected at a minimum of three different locations across each sample surface. At least five scans (i.e., approach and retract cycles) were collected at each location. The AFM experiments and XDLVO simulations were all performed at pH 10 because at this pH, the functional groups on the colloid probe and the membrane surfaces are fully dissociated [41] and the surfaces carry a net negative charge; this leads to a repulsive interaction, which makes the electrostatic contribution to the force curve easier to distinguish. In the AFM force curves the separation distance at which the interaction becomes either repulsive or attractive was identified as the point where the measured force is either positive or negative, respectively, relative to y = 0 on the plot. At separation distances greater than this value no force was considered to be acting on the colloid probe and was designated as the zero force region of the plot. Furthermore, to be considered as a measurable interaction (attractive or repulsive) it had to be greater in magnitude than the noise in the baseline (y = 0) of the AFM force curve. Goodness of fit was determined by minimizing the average of the difference of normalized force (AF/R) (where R is the probe radius) between the theoretical and experimental curves as a function of probe-tosurface separation distance (h) for a minimum of fifteen curves under each set of conditions. 2.5. Streaming potential measurements Tangential streaming potential measurements were carried out on the mica and membrane surface using a commercial streaming potential analyzer (ZetaCAD, CAD Instrumentation, Les Essarts Le Roi, France). Measurements were conducted using 1 and 10 mM KCl over a pH range of 3–9 at 20 ◦ C. The Ag/Ag/Cl electrode plating becomes unstable above pH 9, so the streaming potential value at pH 10 was extrapolated from the experimental data using a best-fit relationship. Zeta potential was calculated from the measured streaming potential using the Helmholtz–Smoluchowski equation with the Fairbrother and Mastin substitution [15].
2.4. AFM force curves AFM analyses were performed with an atomic force microscope (Park Scientific Instruments, Sunnyvale, CA) equipped with a liquid cell. Force measurements were performed using the colloid probe technique. The cantilever spring constant reported by the manufacturer (Novascan Technologies, Ames,
3. Results and discussion 3.1. Surface characterization The pH dependence of zeta potentials for the mica and membrane are shown in Fig. 1 for two ionic strengths. Both surfaces
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
289
Table 2 Measured surface energy properties for the mica, membrane, and carboxyl surfaces Surface
γ LW (mJ/m2 )
γ + (mJ/m2 )
γ − (mJ/m2 )
GSWS (mJ/m2 )
Mica Membrane COOH
40.6 25.7 27.4
1.8 7.1 0.1
51.5 27.6 41.8
148.1 1.6 26.3
pH 5.6 and T = 20 ◦ C.
Surface energy properties for the mica, membrane, and carboxylated colloid probe are reported in Table 2. For each surface, the van der Waals (γ LW ) component, electron acceptor (γ + ) component, electron donor (γ − ) component, and free energy of cohesion (Gsws) are reported. A positive value for Gsws indicates a hydrophilic surface while a negative value indicates a hydrophobic one [43]. Thus, the mica and carboxylated probe are strongly hydrophilic and the membrane is weakly hydrophilic. Moreover, each of the three surfaces is monopolar, or characterized by comparatively large γ − components [44]. 3.2. Force curve analysis
Fig. 1. Zeta potential determined from streaming potential for the (a) mica and (b) membrane as a function of pH and ionic strength.
were negatively charged over the entire pH range investigated. The mica (Fig. 1a) exhibited an isoelectric point between 2.5 and 3 and the membrane (Fig. 1b) exhibited an isoelectric point between pH 3 and 3.5 for 1 and 10 mM KCl, respectively. For the pH and ionic strengths studied, the mica surface was more negatively charged than the membrane surface. The absolute magnitude of the zeta potential for both surfaces decreased with increasing ionic strength, in accordance with classical theories on the relationship between electric double layer compression and zeta potential [42]. The zeta potential values for both surfaces at pH 10, as calculated from best-fit polynomial relationships, are reported in Table 1. Table 1 Zeta potential values determined from streaming potential analysis for the mica and membrane surfaces at pH 10 Surface
Ionic strength of KCl (M)
ζ (mV)
Mica
10−2 10−3
−34 −47
Membrane
10−2 10−2
−28 −36
T = 20 ◦ C; n = 3.
AFM force curves for the carboxylated probe approaching the mica surface at pH 10 and ionic strengths of 10−2 and 10−3 M KCl are reported in Fig. 2a and b, respectively. The average AFM force curve was calculated from the total curve population for each condition and is subsequently reported for the respective ionic strengths. As the colloid probe approaches the mica surface, it encountered a long-range repulsive interaction due to repulsive electrostatic interactions between the similarly charged surfaces [15]. The range of the repulsive interaction was lower at the higher ionic strength (Fig. 2a) due to compression of the electrical double layer [14]. Specifically, at 10−2 M KCl, the interaction became repulsive at a separation distance of 22 nm (average value) and at 10−3 M KCl, the interaction became repulsive at a distance of approximately 45 nm (average value). These distances were determined based on analysis of the average AFM force curve using a graphical analysis software package, and represent the surface-to-surface separation distance where a positive deflection or repulsion first occurs. XDLVO interaction profiles were generated using the mean zeta potential values determined from fits to the AFM data assuming both constant charge and constant potential. The fit for zeta potential was mostly based on the range of the repulsive interaction, which is defined as the separation distance at which the interaction first becomes repulsive. In other words, the primary objective in fitting the XDLVO interaction profiles with the AFM force curves was to match the point of the onset of repulsion for the two curves. A main reason that the fit was carried out this way was because it has been shown [31] that the shape of XDLVO interaction profiles generally do not match with the shape of AFM force curves. This is a result of the mechanical nature, bending characteristics, and other associated complexities involved with the AFM cantilever [40]. XDLVO interaction profiles for the carboxylated probe approaching the mica surface at ionic strengths of 10−2 and 10−3 M KCl and for both constant charge and constant poten-
290
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
at the smaller separation distances [45]. At small separation distances, the constant charge assumption predicted a greater repulsive interaction than the constant potential assumption. In the constant charge assumption, the potential is assumed to be sufficiently small in the region between the two interacting surfaces so that the overall potential is the sum of the contribution from each surface. In the constant potential assumption, the potential in the vicinity of each surface is assumed to be due to that surface alone [15]. Thus, interactions predicted by the constant charge assumption are always greater than those predicted by the constant potential assumption. Also, at the lower ionic strength, the range of the interaction was greater than it was for the higher ionic strength. Specifically the fit was performed such that, at 10−2 M KCl, the interaction became repulsive at separation distances of 22 and 26 nm for the constant potential and constant charge assumptions, respectively, and at 10−3 M KCl the interaction became repulsive at separation distances of 43 and 51 nm for the constant potential and constant charge assumptions, respectively. AFM force curves for a carboxylated probe approaching the membrane surface at pH 10 and ionic strengths of 10−2 and 10−3 M KCl are reported in Fig. 4. When comparing the entire population of force curves with the average force curve,
Fig. 2. AFM force curves between a COOH-functionalized colloid probe and mica in aqueous (a) 10−2 M and (b) 10−3 M KCl solutions at 25 ◦ C and pH 10 shown with their associated average curves.
tial assumptions are shown in Fig. 3. Similar to the AFM force curves, the XDLVO interactions profiles demonstrated a purely repulsive interaction regime. The difference between the constant charge and constant potential assumptions is most apparent
Fig. 3. XDLVO-calculated interaction profiles for a carboxylated probe and mica in aqueous 10−2 and 10−3 M KCl solutions at 25 ◦ C and pH 10. The interaction profiles were calculated using the mean zeta potential determined from XDLVO fits to the AFM data and are reported for both assumptions of constant charge and constant potential.
Fig. 4. AFM force curves between a COOH-functionalized colloid probe and the membrane in aqueous (a) 10−2 M and (b) 10−3 M KCl solutions at 25 ◦ C and pH 10 shown with their associated average curves.
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
Fig. 5. XDLVO-calculated interaction profiles for a carboxylated probe and the membrane in aqueous 10−2 and 10−3 M KCl solutions at 25 ◦ C and pH 10. The interaction profiles were calculated using the mean zeta potential determined from XDLVO fits to the AFM data and are reported for both assumptions of constant charge and constant potential.
it can be seen that greater variability exists in the magnitude and range than when the carboxylated probe approached the mica surface. The greater variability for the membrane surface may be attributed to physical and chemical heterogeneities and/or contamination of the probe. Similar to what was reported for the carboxylated probe–mica interaction, as the carboxylated probe approaches the membrane surface, it encountered a longrange repulsion and the range of the repulsion decreased as the ionic strength increased. At an ionic strength of 10−3 M KCl the interaction became repulsive at 35 nm (Fig. 4a) and at an ionic strength of 10−2 M KCl the interaction became repulsive at 15 nm (Fig. 4b). XDLVO interaction profiles for the carboxylated probe approaching the membrane surface at ionic strengths of 10−2 and 10−3 M KCl and for both constant charge and constant potential assumptions are shown in Fig. 5. The XDLVO interactions profiles demonstrated a purely repulsive interaction regime, the constant charge assumption predicted a greater repulsive interaction than the constant potential assumption, and at the lower ionic strength, the range of the interaction was greater than it was for the higher ionic strength. The fit was performed such that, at 10−3 M KCl the interaction became repulsive at separation distances of 23 and 47 nm for the constant potential and constant charge assumptions, respectively, and at 10−2 M KCl, the interaction became repulsive at separation distances of 12 and 20 nm, for the constant potential and constant charge assumptions, respectively. The sensitivity of the XDLVO fits to changes in the input variable (zeta potential) was approximately 1 mV for fitting the AFM curves measured here. Thus, with regard to precision, this method is comparable to values determined from streaming potential measurements. The zeta potential values for mica and the membrane determined from the XDLVO fits to the AFM data are shown in Table 3. For the mica, the zeta potential determined assuming constant surface charge was in better agreement with that calculated from streaming potential (∆ = 28 and 19% for I = 10−2 and 10−3 M KCl, respectively) than zeta potential determined
291
assuming constant surface potential (∆ = 55 and 53% at I = 10−2 and 10−3 M KCl, respectively). This would indicate that the constant charge assumption might be more suitable when evaluating the electrostatic characteristics of mica surfaces. The net negative surface charge for mica results from the loss of potassium ions to solution (i.e., isomorphic replacement) [46]. In cases where isomorphic replacement is responsible for the charge of a surface, the constant charge assumption is preferred over the constant potential assumption [47]. On the contrary, for the membrane, results generated assuming constant potential were in better agreement with zeta potential calculated from the streaming potential (∆ = 36 and 14% at I = 10−2 and 10−3 M KCl, respectively) than zeta potential calculated using the constant charge assumption (∆ = 51 and 46% at I = 10−2 and 10−3 M KCl, respectively). Competitive adsorption and proton interactions with the membrane material are thought to be the main charging mechanisms for a thin film composite membrane, thus the potential is likely to remain constant as the interacting electrical double layers begin to overlap [8]. Consequently, the constant potential assumption seems to be more suitable for evaluating the electrostatic characteristics of the membrane surface. However, it is important to note that the innate differences between the two limiting assumptions make it very difficult to give either a cart blanch application to either the mica or the membrane. For example, when comparing the zeta potential values obtained with the constant potential and constant charge assumptions (Table 3), it can be seen that the constant charge assumption consistently resulted in a lower standard deviation for the XDLVO-derived data. This is due to the fact that the constant charge assumption more accurately describes the longrange portion of the electrostatic interaction while the constant potential assumption is more accurate at shorter separations [48]. At shorter separations, the interactions are a complex balance between the mechanical properties of the cantilever, short-range chemical interactions, and roughness effects that are not as pronounced at the longer separations. This causes the variation in the AFM force curves (in terms of slope) to increase as separation distance decreases (see Figs. 2a and b and 4a and b), resulting in the higher standard deviation for the model fits to these data sets using the constant potential assumption. The actual electrostatic interaction falls somewhere between the two limiting assumptions of constant charge and constant potential [42]. This
Table 3 Zeta potential determined from AFM measurements for the mica and membrane surfaces Surface
Ionic strength of KCl (M)
Constant potential ζ (mV)
Constant charge ζ (mV)
Mica
10−2 10−3
−76 ± 28.4 −100 ± 27.1
−47 ± 3.9 −58 ± 2.3
Membrane
10−2 10−3
−18 ± 4.4 −31 ± 5.2
−57 ± 1.0 −67 ± 0.7
pH 10, T = 20 ◦ C, n = 15. The reported values represent the mean of all the XDLVO fits to the AFM data and are reported with the standard deviation of the data set.
292
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
Table 4 AFM-generated surface roughness statistics for mica and membrane
Mica Membrane
RMS roughness (nm)
Average roughness (nm)
SAD (%)
0.02 10.07
0.02 7.25
0.001 0.09
presents a source of error when fitting the XDLVO model to the AFM force curves as it relies on the ability of the model used to accurately characterize the measured interaction force. Therefore, to most accurately derive zeta potential from AFM force curves, both limiting assumptions should be considered. One potential advantage of measuring surface charge from AFM force curves instead of from streaming potential measurements is the ability to quantify the variation in charge across the surface. Surface roughness and charge distribution (i.e., changes in charge density and functionality) are the principle sources of variability in surface charge [17,49–53]. Although the contributions of each cannot be easily distinguished in AFM analysis, the overall variation in charge may be quantified. AFM analyses of the mica and membrane surfaces revealed varying degrees of roughness for the two surfaces (Table 4). The mica surface was found to be smooth almost to the atomic scale (RA = 0.02 nm), while the membrane surface was rougher (RA = 10.07 nm). The variation in charge across the two surfaces is best demonstrated by the standard deviations in the XDLVO fit data reported in Table 3. From the standard deviation values reported in Table 3 for the (constant charge) mica and (constant potential) membrane, it can be seen that the mica coefficients of variation (standard deviation divided by mean) are less than 4 and 8% for 10−3 and 10−2 M KCl, respectively, and the membrane coefficients of variation are 17 and 24% for 10−3 and 10−2 M KCl, respectively. The mean and standard deviations may further be used to calculate the range of values for the resulting interactions with a carboxylated probe according to the XDLVO approach for the mica and membrane surfaces (Figs. 6 and 7, respectively). The curves in Figs. 6 and 7 were calculated using the assumption of constant charge for the mica and constant potential for the membrane. Changes in zeta potential generate observable differences in the XDLVO force profiles for both surfaces (Figs. 6 and 7). In these cases the changes do not reverse the types of interaction that are present (e.g., from repulsive to attractive) but they do alter the range and magnitude of the respective interactions. As ionic strength increases, surface roughness features begin to approximate the Debye length (a measure of the double layer thickness). For instance, the Debye lengths for the 10−2 and 10−3 M KCl solutions were 3.06 and 9.66 nm, respectively, which is in the range of the average roughness of the membrane (Table 4). Thus, as ionic strength increases, the impact of roughness on the measured charge variation should also increase. However, this was not the case here as similar charge variations were measured at the two ionic strengths (Table 3). This is most likely due to the fact that even at the low ionic strength (10−3 M KCl) the Debye length was on a similar scale as the roughness features. The impact of ionic strength on the AFM derived zeta potential may be more substantial for other
Fig. 6. XDLVO fits to the AFM measured force curves using zeta potential as the fitting parameter at (a) 10−2 M and (b) 10−3 M KCl for mica. The three curves were calculated using the mean fit zeta potential and the high and low value as determined from the standard deviation in the data and were determined using the constant charge assumption.
polyamide membranes, which have considerable more roughness (RA ∼ 30 − 50 nm). Although the membrane surface was determined to be more rough than the mica the calculated surface area difference (SAD) for the membrane was relatively low (Table 4). This indicates that although peak features were present on the membrane surface they were sparsely distributed. The similar variations in zeta potential for the mica and membrane surfaces also suggest that factors other than roughness (e.g., charge density) are contributing to the measured variation in charge. The contribution of charge density to the variation in the measured zeta potential remains a subject of future study. Nevertheless, the ability of this method to measure the spatial variability of zeta potential has been demonstrated and the changes have been shown to significantly affect the subsequent force profiles. The ability of the AFM technique to characterize the variability in zeta potential across a membrane surface has substantial ramifications when describing membrane performance [10,21,31–33]. This information will allow membrane manufacturers and users to account for and modify the charge distribution across membrane surfaces in order to optimize performance
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
293
than may be obtained with streaming potential measurements alone. References
Fig. 7. XDLVO fits to the AFM measured force curves using zeta potential as the fitting parameter at (a) 10−2 M and (b) 10−3 M KCl for the membrane. The three curves were calculated using the mean fit zeta potential and the high and low value as determined from the standard deviation in the data and were determined using the constant potential assumption.
through the elimination of favorable deposition sites (i.e., areas of lower charge and weaker repulsive interactions). 4. Conclusions Utilizing the XDLVO approach, a quantitative estimate of zeta potential was achieved from AFM force measurements. The mean zeta potential values calculated from the AFM force curves closely agreed with those determined from streaming potential measurements. Results for mica were more closely modeled by the constant charge assumption while those for the polyamide membrane were more closely approximated by the constant potential assumption. The data clearly show that AFM compares favorably to established electrokinetic techniques as a tool for measuring membrane surface potential. Although streaming potential measurements are generally the accepted technique for determining membrane surface charge, they do not provide a spatially resolved measure of charge distribution. The AFM technique can account for this shortcoming by characterizing the charge distribution, as well as a mean charge value, for polymeric membrane surfaces. Thus providing a more comprehensive assessment of membrane surface charge characteristics
[1] M. Nystr¨om, L. Kaipia, S. Luque, Fouling and retention of nanofiltration membranes, J. Membr. Sci. 98 (1995) 249–262. [2] P. Berg, G. Hagmeyer, R. Gimbel, Removal of pesticides and other micropollutants by nanofiltration, Desalination 113 (1997) 205–208. [3] P. Eriksson, Water and salt transport through two types of polyamide composite membranes, J. Membr. Sci. 36 (1988) 297–313. [4] J.M.M. Peeters, M.H.V. Mulder, H. Strathmann, Streaming potential measurements as a characterization method for nanofiltration membranes, Colloids Surf. A 150 (1999) 247–259. [5] Y. Garba, S. Taha, N. Gondrexon, G. Dorange, Ion transport modelling through nanofiltration membranes, J. Membr. Sci. 160 (1999) 187–200. [6] T. Tsuru, X.L. Wang, S.I. Nakao, S. Kimura, M. Togoh, Transport of organic electrolytes with electrostatic and steric-hindrance effects through nanofiltration membranes, J. Chem. Eng. Jpn. 28 (1995) 372–382. [7] M.R. Wiesner, S. Chellam, The promise of membrane technologies, Environ. Sci. Technol. 33 (1999) 360–366. [8] G. Hagmeyer, R. Gimbel, Modelling the rejection of nanofiltration membranes using zeta potential measurements, Sep. Purif. Technol. 15 (1999) 19–30. [9] R. Blank, K.-H. Muth, S. Proske-Gerhards, E. Staude, Electrokinetic investigations of charged porous membranes, Colloids Surf. A 140 (1998) 3–11. [10] A.E. Childress, M. Elimelech, Relating nanofiltration membrane performance to membrane charge (electrokinetic) characteristics, Environ. Sci. Technol. 34 (2000) 3710–3716. [11] A.E. Childress, M. Elimelech, Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanofiltration membranes, J. Membr. Sci. 119 (1996) 253–268. [12] M. Ernst, A. Bismarck, SpringerF J., M. Jekel, Zeta-potential and rejection rates of a polyethersulfone nanofiltration membrane in single salt solutions, J. Membr. Sci. 165 (2000) 251–259. [13] W.M. Chapman, S. Delagah, J. Pellegrino, Variance of streaming potential measurements, J. Membr. Sci. 161 (1999) 247–261. [14] K.A. Hunter, M.W. Leonard, Colloid stability and aggregation in estuaries. 1. Aggregation kinetics of riverine dissolved iron after mixing with seawater, Geochim. Cosmochim. Acta 52 (1988) 1123–1130. [15] M. Elimelech, J. Gregory, X. Jia, R.A. Williams, Particle Deposition and Aggregation: Measurement, Modelling and Simulation, Butterworth–Heinemann, Oxford, 1995. [16] M.J.B. Ariza, E. Rodriguez-Castellion, L. Palacio, Effect of hydration of polyamide membranes on the surface electrokinetic parameters: surface characterization by X-ray photoelectronic spectroscopy and atomic force microscopy, J. Colloid Interf. Sci. 247 (2002) 149–158. [17] W.R. Bowen, T.A. Doneva, Atomic force microscopy studies of membranes: effect of surface roughness on double-layer interactions and particle adhesion, J. Colloid Interf. Sci. 229 (2000) 544–549. [18] T.J. Senden, Force microscopy and surface interactions, Curr. Opin. Colloid Interf. Sci. 6 (2001) 95–101. [19] S. Lee, W.M. Sigmund, Repulsive van der Waals forces for silica and alumina, J. Colloid Interf. Sci. 243 (2001) 365–369. [20] W.R. Bowen, N. Hilal, R.W. Lovitt, A.O. Sharif, P.M. Williams, Atomic force microscope studies of membranes: force measurement and imaging in electrolyte solutions, J. Membr. Sci. 126 (1997) 77. [21] W.R. Bowen, N. Hilal, R.W. Lovitt, C.J. Wright, A new technique for membrane characterisation: direct measurement of the force of adhesion of a single particle using an atomic force microscope, J. Membr. Sci. 139 (1998) 269. [22] W.R. Bowen, N. Hilal, R.W. Lovitt, C.J. Wright, An atomic force microscopy study of the adhesion of a silica sphere to a silica surfaceeffects of surface cleaning, Colloid. Surf. A: Physicochem. Eng. Asp. 157 (1999) 117–125.
294
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
[23] B.V. Derjaguin, L.D. Landau, Theory of the stability of strongly charged lyophobic sols and of the adhesion of strongly charged particles in solutions of electrolytes, Acta Physicochim. URSS 14 (1941) 733–762. [24] E.J.W. Verwey, J.T.G. Overbeek, Theory of the Stability of Lyophobic Colloids, Elsevier, Amsterdam, 1948. [25] T. Abraham, D. Christendat, K. Karan, Z.H. Xu, J. Masliyah, Asphaltene-silica interactions in aqueous solutions: direct force measurements combined with electrokinetic studies, Industr. Eng. Chem. Res. 41 (2002) 2170–2177. [26] C.E. McNamee, A. Matsumoto, P.G. Hartley, P. Mulyaney, K. Tsujii, Interaction forces and zeta potentials of cationic polyelectrolyte coated silica surfaces in water and ethanol: effects of chain length and concentration of perfluorinated anionic surfactants on their binding to the surface, Langmuir 17 (2001) 6220–6227. [27] G. Toikka, R.E. Hayes, Direct measurements of colloidal forces between mica and silica in aqueous electrolyte, J. Colloid Interf. Sci. 191 (1997) 102–109. [28] B. Zhmud, A. Meurk, L. Bergstr¨om, Evaluation of surface ionization parameters from AFM data, J. Colloid Interf. Sci. 1998 (1998) 332– 343. [29] P.G. Hartley, I. Larson, P. Scales, Electrokinetic and direct force measurements between silica and mica surfaces in dilute electrolyte solutions, Langmuir 13 (1997) 2207–2214. [30] W.R. Bowen, T.A. Doneva, J.A.G. Stoton, The use of atomic force microscopy to quantify membrane surface electrical properties, Colloid. Surf. A: Physicochem. Eng. Asp. 201 (2002) 73–83. [31] J.A. Brant, A.E. Childress, Membrane-colloid interactions: comparison of extended DLVO predictions with AFM force measurements, Environ. Eng. Sci. 19 (2002) 413–427. [32] C.J. van Oss, Interfacial Forces in Aqueous Media, Marcel Dekker, New York, 1994. [33] J.A. Brant, A.E. Childress, Colloidal adhesion to hydrophilic membrane surfaces, J. Membr. Sci. 241 (2004) 235–248. [34] J.A. Brant, A.E. Childress, Assessing short-range membrane-colloid interactions using surface energetics, J. Membr. Sci. 203 (2002) 257–273. [35] L. Meagher, C. Klauber, R.M. Pashley, The influence of surface forces on the fouling of polypropylene microfiltration membranes, Colloid. Surf. A: Physicochem. Eng. Asp. 106 (1996) 63–81. [36] T.J. Senden, C.J. Drummond, Surface chemistry and tip-sample interactions in atomic force microscopy, Colloid. Surf. A: Physicochem. Eng. Asp. 94 (1995) 29–51.
[37] W.A. Ducker, T.J. Senden, R.M. Pashley, Direct measurement of colloidal forces using an atomic force microscope, Nature 353 (1991) 239–241. [38] J. Drelich, J.L. Wilbur, J.D. Miller, G.M. Whitesides, Contact angles for liquid drops at a model heterogeneous surface consisting of alternating and parallel hydrophobic/hydrophilic strips, Langmuir 12 (1996) 1913–1922. [39] M. Elimelech, J. Gregory, X. Jia, R.A. Williams, Particle Deposition and Aggregation: Measurement, Modeling and Simulation, Butterworth–Heinemann, Oxford, 1995. [40] B. Cappella, G. Dietler, Force–distance curves by atomic force microscopy, Surf. Sci. Rep. 34 (1999) 1–104. [41] J. Schaep, C. Vandecasteele, Evaluating the charge of nanofiltration membranes, J. Membr. Sci. 188 (2001) 129–136. [42] R.J. Hunter, Foundations of Colloid Science, Clarendon Press, Oxford, 1989. [43] C.J. van Oss, Acid–base interfacial interactions in aqueous media, Colloid. Surf. A 78 (1993) 1. [44] C.J. van Oss, R.F. Giese, W. Wu, On the predominant electron-donicity of polar solid surfaces, J. Adhes. 63 (1997) 71–88. [45] R.J. Hunter, Foundations of Colloid Science, vol. I, Oxford University Press, New York, NY, 1986. [46] W.D. Neese, Introduction to Mineralogy, Oxford University Press, New York, 2000. [47] R.J. Hunter, Zeta Potential in Colloid Science: Principles and Applications, Academic Press, London, 1981. [48] J.N. Israelachvili, Intermolecular and Surface Forces, 2nd ed., Academic Press Harcourt Brace Jovanovich, London, 1992. [49] E.M.V. Hoek, S. Bhattacharjee, M. Elimelech, Effect of membrane surface roughness on colloid-membrane DLVO interactions, Langmuir 19 (2003) 4836–4847. [50] L.W. Suresh, Direct measurement of the effect of surface roughness on the colloidal forces between a particle and flat plate, J. Colloid. Interf. Sci. 196 (1997) 177–190. [51] L.W. Suresh, Effect of surface roughness on the interaction energy between a colloidal sphere and a flat plate, J. Colloid. Interf. Sci. 183 (1996) 199–213. [52] S. Bhattacharjee, C.-H. Ko, M. Elimelech, DLVO interaction between rough surfaces, Langmuir 14 (1998) 3365–3375. [53] J.F.L. Duval, F.A.M. Leermakers, H.P. van Leeuwen, Electrostatic interactions between double layers: influence of surface roughness, regulation, and chemical heterogeneity, Langmuir 20 (2004) 5052–5065.
Examining the electrochemical properties of a nanofiltration membrane with atomic force microscopy Jonathan A. Brant a,∗ , Kelly M. Johnson b , Amy E. Childress b a
b
Rice University, Department of Civil and Environmental Engineering, MS 317, Houston, TX 77251, USA University of Nevada, Reno Department of Civil and Environmental Engineering, MS 258, Reno, NV 89557, USA Received 21 May 2005; received in revised form 23 September 2005; accepted 3 October 2005 Available online 2 November 2005
Abstract In this investigation, two methods for characterizing membrane surface potential are investigated. Results from atomic force microscopy (AFM) analyses are compared with streaming potential measurements. In calculating surface potential from AFM force measurements, assumptions of constant charge and constant potential were both considered for modeling electrostatic interactions. For a ceramic mica surface, the constant charge assumption was found to be most appropriate while for a polymeric membrane surface, the constant potential assumption provided results that agreed better with theoretical expectations. For both the mica and membrane surfaces, results from AFM agreed with the measured values determined from streaming potential analysis. The advantage of AFM is that in addition to determining the mean surface potential value for membrane surfaces, this technique provides a spatially resolved measure of charge distribution. One drawback of the technique is that it is sensitive to surface roughness, as the measured charge distribution increased with increasing surface roughness. © 2005 Elsevier B.V. All rights reserved. Keywords: AFM; Zeta potential; Membrane; Streaming potential; Nanofiltration
1. Introduction Nanofiltration (NF) membranes are used in a wide range of drinking water, wastewater, and industrial applications [1,2]. Separation by NF membranes occurs primarily due to size exclusion and electrostatic interactions [1–3,4]. For colloids and uncharged molecules, sieving or size exclusion is most responsible for separation; for ions and charged organics, electrostatic interactions are responsible for separation [2,5,6]. For all applications, membrane charge characteristics play a significant role in the transport of water and solute molecules through the membrane. Additionally, the interaction of colloids and charged macromolecules with the membrane, and subsequent fouling of the membrane, is dependent on the charge properties [7]. Because of this, the availability of a simple, reproducible, standardized method of measuring membrane charge properties is of critical importance. In experimental investigations of membrane charge, streaming potential measurements have typically been used to calcu-
∗
Corresponding author. Tel.: +1 713 348 3374; fax: +1 713 348 5203. E-mail address: [email protected] (J.A. Brant).
0376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2005.10.002
late zeta potential. Streaming potential is the potential induced when an electrolyte solution is pumped across a stationary, charged surface. Streaming potential can be used to calculate zeta potential using the Helmholtz–Smoluchowski equation. Several works on streaming potential measurements of NF membranes (e.g. [1,4,8–12]) have appeared in the literature. Although streaming potential measurements are the most frequently used method for evaluating charge properties, they have also been criticized. Results from prior studies reveal uncertainty in individual measurements as well as data scatter [13]. The Helmholtz–Smoluchowski relationship used to calculate zeta potential breaks down at very high or low ionic strengths [14]. Differences in instrument design and the lack of a calibration standard for streaming potential analyzers makes comparison of data among laboratories challenging. An additional concern is that membrane surfaces are heterogeneous—both physically and chemically and for rough or chemically heterogeneous surfaces, surface potential calculated from streaming potential measurements may provide an incomplete description of the surface’s charge characteristics [15]. For example, because zeta potential is an average value of the potential at some distance away from the surface, hydrodynamic effects due to surface roughness may distort zeta potential results [16,17]. Furthermore, streaming
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
potential measurements cannot account nor describe the charge distribution across a surface as it only provides an average zeta potential value for the whole surface. Nevertheless, the mean value provided by streaming potential measurements serves as a reasonable basis from which to compare values determined using other techniques. A dramatic advancement in membrane research is the ability to directly measure the interfacial interaction between a probe and a membrane surface using atomic force microscopy (AFM) [18]. AFM was originally developed as an imaging tool with atomic-level resolution; however, its operating principle has always been based on measuring the force between a surface and a small probe [18,19]. Thus, AFM may be used to measure the interaction force as a function of separation distance between two surfaces [20–22]. The chemical interactions thought to be important in colloid–surface interactions are generally assessed using the classical Derjaguin–Landau– Verwey–Overbeek (DLVO) theory [23,24], which describes the interaction between a colloid and a surface as the interplay between Lifshitz–van der Waals (LW) attractive interactions and electrostatic double layer (EL) repulsive interactions. In the last decade, several studies have focused on comparing surface charge data obtained from AFM to surface potential data determined from electrokinetic measurements (e.g. [25–30]). However, only one of these studies involved polymeric membranes. Bowen et al. [30] modeled the interaction between a colloid probe and a commercial NF membrane surface using membrane zeta potential as a fitting parameter. At low ionic strength (10−3 M), AFM surface potential measurements were in good agreement with the zeta potential calculated from streaming potential. However, at higher ionic strength (10−1 M), when the surface roughness approached the Debye length, poor agreement was seen between the zeta potential derived from the AFM data and that determined from streaming potential. Surface roughness was identified as one possible source of error when comparing zeta potential results determined from AFM force curves and streaming potential measurements. The advantage of the AFM method is that it is possible to directly measure the impact of surface roughness on the zeta potential across the surface, which cannot be done through measuring streaming potential. Additionally, as AFM measurements are carried out at different locations across a surface, having relatively small areas of interaction, they may be used to characterize surface charge distribution. This represents a potentially valuable advancement over streaming potential measurements. The observed discrepancy between the AFM data and the streaming potential results observed by Bowen et al. [30] may also be attributed to the presence of non-DLVO interactions at shorter separation distances. Such additional types of interactions at the membrane surface may include short-ranged acid–base (AB) (electron donor/electron acceptor) interactions that occur between two surfaces immersed in a polar solvent [31]. These interactions may be accounted for using an extended form of the DLVO model (i.e., the XDLVO approach) [32]. In recent investigations, the XDLVO approach has been shown to be successful in describing membrane-colloid interactions [31,33,34]. The XDLVO approach describes the total interac-
287
tion between two surfaces in aqueous media according to [32]: U XDLVO = U LW + U EL + U AB
(1)
where UXDLVO is the total interaction energy between the membrane and colloid immersed in water, ULW is the Lifshitz–van der Waals interaction term, UEL is the electrostatic interaction term, and UAB is the acid–base interaction term. From a thermodynamic standpoint, attraction or adhesion between two interacting surfaces occurs when UXDLVO is negative; repulsion occurs when UXDLVO is positive. The total interaction energy is often evaluated as a function of separation distance between the interacting surfaces using interaction energy profiles. Interaction energy profiles illustrate the type of interaction (attractive or repulsive) occurring as a colloid approaches a membrane surface and the separation distance at which the individual interaction energies are dominant. The main objective of this investigation was to determine if AFM is a feasible method for measuring the surface potential of a polymeric membrane surface. A secondary objective was to determine the most appropriate assumption for modeling electrostatic interactions for the surface within the framework of an XDLVO model. In doing this, short-range interactions that have not previously been accounted for were considered so that a more accurate zeta potential could be determined from AFM analyses. First, the interactions between an AFM probe having a well-defined surface chemistry and geometry and a relatively smooth NF membrane were studied; second, surface potentials determined from the AFM force measurements were compared with those determined from streaming potential measurements. In parallel, the use of AFM to determine the surface potential of mica was also evaluated. The purpose of this was to compare surface potential results for a relatively smooth, homogeneous surface with results for a heterogeneous membrane surface as well as to determine the variability of surface charge across the membrane surface. Because variation in charge density across a membrane surface is often cited as a reason for preferential deposition of foulants on membranes [10,35], a spatially resolved measurement of membrane charge provides valuable insight into membrane fouling by determining whether the variation is substantially large as to alter membrane-solute interactions compared to the mean charge value. 2. Materials and methods 2.1. Surfaces Two surfaces were chosen for study: muscovite mica and a relatively smooth NF membrane. Mica was chosen because its surface is molecularly smooth and its chemistry is well characterized [27,29,36]. The muscovite mica used in this investigation was research grade (SPI Supplies, West Chester, PA). It was stored in a vacuum dessicator and was used with no preparation. The NF membrane was a thin film composite polyamide membrane, designated HL by the manufacturer (Osmonics, Vista, CA). The membrane was stored at 5 ◦ C in ultrapure water obtained from a Millipore (Burlington, MA) water purification
288
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
system. Individual membrane samples were equilibrated in test solutions for 24 h prior to AFM and streaming potential measurements. 2.2. Colloid probes Colloid probes (NovaScan Technologies, Ames, IA) were formed by gluing glass microspheres to the end of tipless silicon nitride cantilevers [37]. The glass microspheres had a diameter of 6 m, which was verified using a scanning electron microscope (SEM). The glass probes were functionalized with carboxyl (COOH) groups arranged in a monolayer. The result was a carboxylated colloid probe having a known surface chemistry and a well-defined geometry. 2.3. XDLVO interaction profiles Contact angles for the mica and membrane surfaces were measured with three well-characterized probe liquids (ultrapure water, formamide, and bromonaphthalene). Triplicate contact angles were measured on no less than five different samples for each surface–probe liquid combination. Contact angle values for a monolayer terminating in carboxyl groups were taken from Drelich [38] and were used to calculate the surface energy properties for the carboxylated AFM probe. Solid surface energies were calculated from the measured contact angles using the Young–Dupr´e equation [32]: γ1 (1 + cos θ) = 2 (2) γlLW γsLW + γl+ γs− + γl− γs+ where θ is the measured contact angle; γ LW is the van der Waals free energy component; γ + is the electron-acceptor component; γ − is the electron-donor component; and the subscripts l and s designate the liquid and solid phases, respectively. The values for the van der Waals and acid–base components were then used to calculate ULW and UAB in terms of force normalized to probe radius as a function of separation distance. Detailed expressions for these equations and their derivations have been outlined elsewhere (e.g. [31,32]). The electrostatic component, UEL , was calculated using the limiting assumptions of constant charge and constant potential as outlined in Elimelech et al. [39]. In calculating UEL , the solution ionic strength and the zeta potential of the colloid probe surface were used as input parameters, while the zeta potential of the sample surface served as the only variable. In this regard, the zeta potential of the sample surface was the fitting parameter for fitting the XDLVO interaction profile to the AFM force curves. Thus, the zeta potential of the sample surface was determined at each location where an AFM force curve was collected.
IA) was 0.12 N/m. This value was used for converting cantilever deflection to interaction force values. Force curves are initially reported as force as a function of scanner position by the AFM instrumentation. Scanner position was converted to separation distance (h) by determining the onset of constant compliance between the scanner position and cantilever deflection (i.e., where cantilever deflection becomes a linear function of piezo scanner position) and subtracting this value from all other scanner position values [40]. Measurements were conducted in a liquid cell at 20 ◦ C with −2 10 and 10−3 M KCl at pH 10. Prior to the measurements, the AFM liquid cell was cleaned with acetone, rinsed with ultrapure water, and blown dry using a high velocity stream of nitrogen. Force curves were collected at a minimum of three different locations across each sample surface. At least five scans (i.e., approach and retract cycles) were collected at each location. The AFM experiments and XDLVO simulations were all performed at pH 10 because at this pH, the functional groups on the colloid probe and the membrane surfaces are fully dissociated [41] and the surfaces carry a net negative charge; this leads to a repulsive interaction, which makes the electrostatic contribution to the force curve easier to distinguish. In the AFM force curves the separation distance at which the interaction becomes either repulsive or attractive was identified as the point where the measured force is either positive or negative, respectively, relative to y = 0 on the plot. At separation distances greater than this value no force was considered to be acting on the colloid probe and was designated as the zero force region of the plot. Furthermore, to be considered as a measurable interaction (attractive or repulsive) it had to be greater in magnitude than the noise in the baseline (y = 0) of the AFM force curve. Goodness of fit was determined by minimizing the average of the difference of normalized force (AF/R) (where R is the probe radius) between the theoretical and experimental curves as a function of probe-tosurface separation distance (h) for a minimum of fifteen curves under each set of conditions. 2.5. Streaming potential measurements Tangential streaming potential measurements were carried out on the mica and membrane surface using a commercial streaming potential analyzer (ZetaCAD, CAD Instrumentation, Les Essarts Le Roi, France). Measurements were conducted using 1 and 10 mM KCl over a pH range of 3–9 at 20 ◦ C. The Ag/Ag/Cl electrode plating becomes unstable above pH 9, so the streaming potential value at pH 10 was extrapolated from the experimental data using a best-fit relationship. Zeta potential was calculated from the measured streaming potential using the Helmholtz–Smoluchowski equation with the Fairbrother and Mastin substitution [15].
2.4. AFM force curves AFM analyses were performed with an atomic force microscope (Park Scientific Instruments, Sunnyvale, CA) equipped with a liquid cell. Force measurements were performed using the colloid probe technique. The cantilever spring constant reported by the manufacturer (Novascan Technologies, Ames,
3. Results and discussion 3.1. Surface characterization The pH dependence of zeta potentials for the mica and membrane are shown in Fig. 1 for two ionic strengths. Both surfaces
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
289
Table 2 Measured surface energy properties for the mica, membrane, and carboxyl surfaces Surface
γ LW (mJ/m2 )
γ + (mJ/m2 )
γ − (mJ/m2 )
GSWS (mJ/m2 )
Mica Membrane COOH
40.6 25.7 27.4
1.8 7.1 0.1
51.5 27.6 41.8
148.1 1.6 26.3
pH 5.6 and T = 20 ◦ C.
Surface energy properties for the mica, membrane, and carboxylated colloid probe are reported in Table 2. For each surface, the van der Waals (γ LW ) component, electron acceptor (γ + ) component, electron donor (γ − ) component, and free energy of cohesion (Gsws) are reported. A positive value for Gsws indicates a hydrophilic surface while a negative value indicates a hydrophobic one [43]. Thus, the mica and carboxylated probe are strongly hydrophilic and the membrane is weakly hydrophilic. Moreover, each of the three surfaces is monopolar, or characterized by comparatively large γ − components [44]. 3.2. Force curve analysis
Fig. 1. Zeta potential determined from streaming potential for the (a) mica and (b) membrane as a function of pH and ionic strength.
were negatively charged over the entire pH range investigated. The mica (Fig. 1a) exhibited an isoelectric point between 2.5 and 3 and the membrane (Fig. 1b) exhibited an isoelectric point between pH 3 and 3.5 for 1 and 10 mM KCl, respectively. For the pH and ionic strengths studied, the mica surface was more negatively charged than the membrane surface. The absolute magnitude of the zeta potential for both surfaces decreased with increasing ionic strength, in accordance with classical theories on the relationship between electric double layer compression and zeta potential [42]. The zeta potential values for both surfaces at pH 10, as calculated from best-fit polynomial relationships, are reported in Table 1. Table 1 Zeta potential values determined from streaming potential analysis for the mica and membrane surfaces at pH 10 Surface
Ionic strength of KCl (M)
ζ (mV)
Mica
10−2 10−3
−34 −47
Membrane
10−2 10−2
−28 −36
T = 20 ◦ C; n = 3.
AFM force curves for the carboxylated probe approaching the mica surface at pH 10 and ionic strengths of 10−2 and 10−3 M KCl are reported in Fig. 2a and b, respectively. The average AFM force curve was calculated from the total curve population for each condition and is subsequently reported for the respective ionic strengths. As the colloid probe approaches the mica surface, it encountered a long-range repulsive interaction due to repulsive electrostatic interactions between the similarly charged surfaces [15]. The range of the repulsive interaction was lower at the higher ionic strength (Fig. 2a) due to compression of the electrical double layer [14]. Specifically, at 10−2 M KCl, the interaction became repulsive at a separation distance of 22 nm (average value) and at 10−3 M KCl, the interaction became repulsive at a distance of approximately 45 nm (average value). These distances were determined based on analysis of the average AFM force curve using a graphical analysis software package, and represent the surface-to-surface separation distance where a positive deflection or repulsion first occurs. XDLVO interaction profiles were generated using the mean zeta potential values determined from fits to the AFM data assuming both constant charge and constant potential. The fit for zeta potential was mostly based on the range of the repulsive interaction, which is defined as the separation distance at which the interaction first becomes repulsive. In other words, the primary objective in fitting the XDLVO interaction profiles with the AFM force curves was to match the point of the onset of repulsion for the two curves. A main reason that the fit was carried out this way was because it has been shown [31] that the shape of XDLVO interaction profiles generally do not match with the shape of AFM force curves. This is a result of the mechanical nature, bending characteristics, and other associated complexities involved with the AFM cantilever [40]. XDLVO interaction profiles for the carboxylated probe approaching the mica surface at ionic strengths of 10−2 and 10−3 M KCl and for both constant charge and constant poten-
290
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
at the smaller separation distances [45]. At small separation distances, the constant charge assumption predicted a greater repulsive interaction than the constant potential assumption. In the constant charge assumption, the potential is assumed to be sufficiently small in the region between the two interacting surfaces so that the overall potential is the sum of the contribution from each surface. In the constant potential assumption, the potential in the vicinity of each surface is assumed to be due to that surface alone [15]. Thus, interactions predicted by the constant charge assumption are always greater than those predicted by the constant potential assumption. Also, at the lower ionic strength, the range of the interaction was greater than it was for the higher ionic strength. Specifically the fit was performed such that, at 10−2 M KCl, the interaction became repulsive at separation distances of 22 and 26 nm for the constant potential and constant charge assumptions, respectively, and at 10−3 M KCl the interaction became repulsive at separation distances of 43 and 51 nm for the constant potential and constant charge assumptions, respectively. AFM force curves for a carboxylated probe approaching the membrane surface at pH 10 and ionic strengths of 10−2 and 10−3 M KCl are reported in Fig. 4. When comparing the entire population of force curves with the average force curve,
Fig. 2. AFM force curves between a COOH-functionalized colloid probe and mica in aqueous (a) 10−2 M and (b) 10−3 M KCl solutions at 25 ◦ C and pH 10 shown with their associated average curves.
tial assumptions are shown in Fig. 3. Similar to the AFM force curves, the XDLVO interactions profiles demonstrated a purely repulsive interaction regime. The difference between the constant charge and constant potential assumptions is most apparent
Fig. 3. XDLVO-calculated interaction profiles for a carboxylated probe and mica in aqueous 10−2 and 10−3 M KCl solutions at 25 ◦ C and pH 10. The interaction profiles were calculated using the mean zeta potential determined from XDLVO fits to the AFM data and are reported for both assumptions of constant charge and constant potential.
Fig. 4. AFM force curves between a COOH-functionalized colloid probe and the membrane in aqueous (a) 10−2 M and (b) 10−3 M KCl solutions at 25 ◦ C and pH 10 shown with their associated average curves.
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
Fig. 5. XDLVO-calculated interaction profiles for a carboxylated probe and the membrane in aqueous 10−2 and 10−3 M KCl solutions at 25 ◦ C and pH 10. The interaction profiles were calculated using the mean zeta potential determined from XDLVO fits to the AFM data and are reported for both assumptions of constant charge and constant potential.
it can be seen that greater variability exists in the magnitude and range than when the carboxylated probe approached the mica surface. The greater variability for the membrane surface may be attributed to physical and chemical heterogeneities and/or contamination of the probe. Similar to what was reported for the carboxylated probe–mica interaction, as the carboxylated probe approaches the membrane surface, it encountered a longrange repulsion and the range of the repulsion decreased as the ionic strength increased. At an ionic strength of 10−3 M KCl the interaction became repulsive at 35 nm (Fig. 4a) and at an ionic strength of 10−2 M KCl the interaction became repulsive at 15 nm (Fig. 4b). XDLVO interaction profiles for the carboxylated probe approaching the membrane surface at ionic strengths of 10−2 and 10−3 M KCl and for both constant charge and constant potential assumptions are shown in Fig. 5. The XDLVO interactions profiles demonstrated a purely repulsive interaction regime, the constant charge assumption predicted a greater repulsive interaction than the constant potential assumption, and at the lower ionic strength, the range of the interaction was greater than it was for the higher ionic strength. The fit was performed such that, at 10−3 M KCl the interaction became repulsive at separation distances of 23 and 47 nm for the constant potential and constant charge assumptions, respectively, and at 10−2 M KCl, the interaction became repulsive at separation distances of 12 and 20 nm, for the constant potential and constant charge assumptions, respectively. The sensitivity of the XDLVO fits to changes in the input variable (zeta potential) was approximately 1 mV for fitting the AFM curves measured here. Thus, with regard to precision, this method is comparable to values determined from streaming potential measurements. The zeta potential values for mica and the membrane determined from the XDLVO fits to the AFM data are shown in Table 3. For the mica, the zeta potential determined assuming constant surface charge was in better agreement with that calculated from streaming potential (∆ = 28 and 19% for I = 10−2 and 10−3 M KCl, respectively) than zeta potential determined
291
assuming constant surface potential (∆ = 55 and 53% at I = 10−2 and 10−3 M KCl, respectively). This would indicate that the constant charge assumption might be more suitable when evaluating the electrostatic characteristics of mica surfaces. The net negative surface charge for mica results from the loss of potassium ions to solution (i.e., isomorphic replacement) [46]. In cases where isomorphic replacement is responsible for the charge of a surface, the constant charge assumption is preferred over the constant potential assumption [47]. On the contrary, for the membrane, results generated assuming constant potential were in better agreement with zeta potential calculated from the streaming potential (∆ = 36 and 14% at I = 10−2 and 10−3 M KCl, respectively) than zeta potential calculated using the constant charge assumption (∆ = 51 and 46% at I = 10−2 and 10−3 M KCl, respectively). Competitive adsorption and proton interactions with the membrane material are thought to be the main charging mechanisms for a thin film composite membrane, thus the potential is likely to remain constant as the interacting electrical double layers begin to overlap [8]. Consequently, the constant potential assumption seems to be more suitable for evaluating the electrostatic characteristics of the membrane surface. However, it is important to note that the innate differences between the two limiting assumptions make it very difficult to give either a cart blanch application to either the mica or the membrane. For example, when comparing the zeta potential values obtained with the constant potential and constant charge assumptions (Table 3), it can be seen that the constant charge assumption consistently resulted in a lower standard deviation for the XDLVO-derived data. This is due to the fact that the constant charge assumption more accurately describes the longrange portion of the electrostatic interaction while the constant potential assumption is more accurate at shorter separations [48]. At shorter separations, the interactions are a complex balance between the mechanical properties of the cantilever, short-range chemical interactions, and roughness effects that are not as pronounced at the longer separations. This causes the variation in the AFM force curves (in terms of slope) to increase as separation distance decreases (see Figs. 2a and b and 4a and b), resulting in the higher standard deviation for the model fits to these data sets using the constant potential assumption. The actual electrostatic interaction falls somewhere between the two limiting assumptions of constant charge and constant potential [42]. This
Table 3 Zeta potential determined from AFM measurements for the mica and membrane surfaces Surface
Ionic strength of KCl (M)
Constant potential ζ (mV)
Constant charge ζ (mV)
Mica
10−2 10−3
−76 ± 28.4 −100 ± 27.1
−47 ± 3.9 −58 ± 2.3
Membrane
10−2 10−3
−18 ± 4.4 −31 ± 5.2
−57 ± 1.0 −67 ± 0.7
pH 10, T = 20 ◦ C, n = 15. The reported values represent the mean of all the XDLVO fits to the AFM data and are reported with the standard deviation of the data set.
292
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
Table 4 AFM-generated surface roughness statistics for mica and membrane
Mica Membrane
RMS roughness (nm)
Average roughness (nm)
SAD (%)
0.02 10.07
0.02 7.25
0.001 0.09
presents a source of error when fitting the XDLVO model to the AFM force curves as it relies on the ability of the model used to accurately characterize the measured interaction force. Therefore, to most accurately derive zeta potential from AFM force curves, both limiting assumptions should be considered. One potential advantage of measuring surface charge from AFM force curves instead of from streaming potential measurements is the ability to quantify the variation in charge across the surface. Surface roughness and charge distribution (i.e., changes in charge density and functionality) are the principle sources of variability in surface charge [17,49–53]. Although the contributions of each cannot be easily distinguished in AFM analysis, the overall variation in charge may be quantified. AFM analyses of the mica and membrane surfaces revealed varying degrees of roughness for the two surfaces (Table 4). The mica surface was found to be smooth almost to the atomic scale (RA = 0.02 nm), while the membrane surface was rougher (RA = 10.07 nm). The variation in charge across the two surfaces is best demonstrated by the standard deviations in the XDLVO fit data reported in Table 3. From the standard deviation values reported in Table 3 for the (constant charge) mica and (constant potential) membrane, it can be seen that the mica coefficients of variation (standard deviation divided by mean) are less than 4 and 8% for 10−3 and 10−2 M KCl, respectively, and the membrane coefficients of variation are 17 and 24% for 10−3 and 10−2 M KCl, respectively. The mean and standard deviations may further be used to calculate the range of values for the resulting interactions with a carboxylated probe according to the XDLVO approach for the mica and membrane surfaces (Figs. 6 and 7, respectively). The curves in Figs. 6 and 7 were calculated using the assumption of constant charge for the mica and constant potential for the membrane. Changes in zeta potential generate observable differences in the XDLVO force profiles for both surfaces (Figs. 6 and 7). In these cases the changes do not reverse the types of interaction that are present (e.g., from repulsive to attractive) but they do alter the range and magnitude of the respective interactions. As ionic strength increases, surface roughness features begin to approximate the Debye length (a measure of the double layer thickness). For instance, the Debye lengths for the 10−2 and 10−3 M KCl solutions were 3.06 and 9.66 nm, respectively, which is in the range of the average roughness of the membrane (Table 4). Thus, as ionic strength increases, the impact of roughness on the measured charge variation should also increase. However, this was not the case here as similar charge variations were measured at the two ionic strengths (Table 3). This is most likely due to the fact that even at the low ionic strength (10−3 M KCl) the Debye length was on a similar scale as the roughness features. The impact of ionic strength on the AFM derived zeta potential may be more substantial for other
Fig. 6. XDLVO fits to the AFM measured force curves using zeta potential as the fitting parameter at (a) 10−2 M and (b) 10−3 M KCl for mica. The three curves were calculated using the mean fit zeta potential and the high and low value as determined from the standard deviation in the data and were determined using the constant charge assumption.
polyamide membranes, which have considerable more roughness (RA ∼ 30 − 50 nm). Although the membrane surface was determined to be more rough than the mica the calculated surface area difference (SAD) for the membrane was relatively low (Table 4). This indicates that although peak features were present on the membrane surface they were sparsely distributed. The similar variations in zeta potential for the mica and membrane surfaces also suggest that factors other than roughness (e.g., charge density) are contributing to the measured variation in charge. The contribution of charge density to the variation in the measured zeta potential remains a subject of future study. Nevertheless, the ability of this method to measure the spatial variability of zeta potential has been demonstrated and the changes have been shown to significantly affect the subsequent force profiles. The ability of the AFM technique to characterize the variability in zeta potential across a membrane surface has substantial ramifications when describing membrane performance [10,21,31–33]. This information will allow membrane manufacturers and users to account for and modify the charge distribution across membrane surfaces in order to optimize performance
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
293
than may be obtained with streaming potential measurements alone. References
Fig. 7. XDLVO fits to the AFM measured force curves using zeta potential as the fitting parameter at (a) 10−2 M and (b) 10−3 M KCl for the membrane. The three curves were calculated using the mean fit zeta potential and the high and low value as determined from the standard deviation in the data and were determined using the constant potential assumption.
through the elimination of favorable deposition sites (i.e., areas of lower charge and weaker repulsive interactions). 4. Conclusions Utilizing the XDLVO approach, a quantitative estimate of zeta potential was achieved from AFM force measurements. The mean zeta potential values calculated from the AFM force curves closely agreed with those determined from streaming potential measurements. Results for mica were more closely modeled by the constant charge assumption while those for the polyamide membrane were more closely approximated by the constant potential assumption. The data clearly show that AFM compares favorably to established electrokinetic techniques as a tool for measuring membrane surface potential. Although streaming potential measurements are generally the accepted technique for determining membrane surface charge, they do not provide a spatially resolved measure of charge distribution. The AFM technique can account for this shortcoming by characterizing the charge distribution, as well as a mean charge value, for polymeric membrane surfaces. Thus providing a more comprehensive assessment of membrane surface charge characteristics
[1] M. Nystr¨om, L. Kaipia, S. Luque, Fouling and retention of nanofiltration membranes, J. Membr. Sci. 98 (1995) 249–262. [2] P. Berg, G. Hagmeyer, R. Gimbel, Removal of pesticides and other micropollutants by nanofiltration, Desalination 113 (1997) 205–208. [3] P. Eriksson, Water and salt transport through two types of polyamide composite membranes, J. Membr. Sci. 36 (1988) 297–313. [4] J.M.M. Peeters, M.H.V. Mulder, H. Strathmann, Streaming potential measurements as a characterization method for nanofiltration membranes, Colloids Surf. A 150 (1999) 247–259. [5] Y. Garba, S. Taha, N. Gondrexon, G. Dorange, Ion transport modelling through nanofiltration membranes, J. Membr. Sci. 160 (1999) 187–200. [6] T. Tsuru, X.L. Wang, S.I. Nakao, S. Kimura, M. Togoh, Transport of organic electrolytes with electrostatic and steric-hindrance effects through nanofiltration membranes, J. Chem. Eng. Jpn. 28 (1995) 372–382. [7] M.R. Wiesner, S. Chellam, The promise of membrane technologies, Environ. Sci. Technol. 33 (1999) 360–366. [8] G. Hagmeyer, R. Gimbel, Modelling the rejection of nanofiltration membranes using zeta potential measurements, Sep. Purif. Technol. 15 (1999) 19–30. [9] R. Blank, K.-H. Muth, S. Proske-Gerhards, E. Staude, Electrokinetic investigations of charged porous membranes, Colloids Surf. A 140 (1998) 3–11. [10] A.E. Childress, M. Elimelech, Relating nanofiltration membrane performance to membrane charge (electrokinetic) characteristics, Environ. Sci. Technol. 34 (2000) 3710–3716. [11] A.E. Childress, M. Elimelech, Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanofiltration membranes, J. Membr. Sci. 119 (1996) 253–268. [12] M. Ernst, A. Bismarck, SpringerF J., M. Jekel, Zeta-potential and rejection rates of a polyethersulfone nanofiltration membrane in single salt solutions, J. Membr. Sci. 165 (2000) 251–259. [13] W.M. Chapman, S. Delagah, J. Pellegrino, Variance of streaming potential measurements, J. Membr. Sci. 161 (1999) 247–261. [14] K.A. Hunter, M.W. Leonard, Colloid stability and aggregation in estuaries. 1. Aggregation kinetics of riverine dissolved iron after mixing with seawater, Geochim. Cosmochim. Acta 52 (1988) 1123–1130. [15] M. Elimelech, J. Gregory, X. Jia, R.A. Williams, Particle Deposition and Aggregation: Measurement, Modelling and Simulation, Butterworth–Heinemann, Oxford, 1995. [16] M.J.B. Ariza, E. Rodriguez-Castellion, L. Palacio, Effect of hydration of polyamide membranes on the surface electrokinetic parameters: surface characterization by X-ray photoelectronic spectroscopy and atomic force microscopy, J. Colloid Interf. Sci. 247 (2002) 149–158. [17] W.R. Bowen, T.A. Doneva, Atomic force microscopy studies of membranes: effect of surface roughness on double-layer interactions and particle adhesion, J. Colloid Interf. Sci. 229 (2000) 544–549. [18] T.J. Senden, Force microscopy and surface interactions, Curr. Opin. Colloid Interf. Sci. 6 (2001) 95–101. [19] S. Lee, W.M. Sigmund, Repulsive van der Waals forces for silica and alumina, J. Colloid Interf. Sci. 243 (2001) 365–369. [20] W.R. Bowen, N. Hilal, R.W. Lovitt, A.O. Sharif, P.M. Williams, Atomic force microscope studies of membranes: force measurement and imaging in electrolyte solutions, J. Membr. Sci. 126 (1997) 77. [21] W.R. Bowen, N. Hilal, R.W. Lovitt, C.J. Wright, A new technique for membrane characterisation: direct measurement of the force of adhesion of a single particle using an atomic force microscope, J. Membr. Sci. 139 (1998) 269. [22] W.R. Bowen, N. Hilal, R.W. Lovitt, C.J. Wright, An atomic force microscopy study of the adhesion of a silica sphere to a silica surfaceeffects of surface cleaning, Colloid. Surf. A: Physicochem. Eng. Asp. 157 (1999) 117–125.
294
J.A. Brant et al. / Journal of Membrane Science 276 (2006) 286–294
[23] B.V. Derjaguin, L.D. Landau, Theory of the stability of strongly charged lyophobic sols and of the adhesion of strongly charged particles in solutions of electrolytes, Acta Physicochim. URSS 14 (1941) 733–762. [24] E.J.W. Verwey, J.T.G. Overbeek, Theory of the Stability of Lyophobic Colloids, Elsevier, Amsterdam, 1948. [25] T. Abraham, D. Christendat, K. Karan, Z.H. Xu, J. Masliyah, Asphaltene-silica interactions in aqueous solutions: direct force measurements combined with electrokinetic studies, Industr. Eng. Chem. Res. 41 (2002) 2170–2177. [26] C.E. McNamee, A. Matsumoto, P.G. Hartley, P. Mulyaney, K. Tsujii, Interaction forces and zeta potentials of cationic polyelectrolyte coated silica surfaces in water and ethanol: effects of chain length and concentration of perfluorinated anionic surfactants on their binding to the surface, Langmuir 17 (2001) 6220–6227. [27] G. Toikka, R.E. Hayes, Direct measurements of colloidal forces between mica and silica in aqueous electrolyte, J. Colloid Interf. Sci. 191 (1997) 102–109. [28] B. Zhmud, A. Meurk, L. Bergstr¨om, Evaluation of surface ionization parameters from AFM data, J. Colloid Interf. Sci. 1998 (1998) 332– 343. [29] P.G. Hartley, I. Larson, P. Scales, Electrokinetic and direct force measurements between silica and mica surfaces in dilute electrolyte solutions, Langmuir 13 (1997) 2207–2214. [30] W.R. Bowen, T.A. Doneva, J.A.G. Stoton, The use of atomic force microscopy to quantify membrane surface electrical properties, Colloid. Surf. A: Physicochem. Eng. Asp. 201 (2002) 73–83. [31] J.A. Brant, A.E. Childress, Membrane-colloid interactions: comparison of extended DLVO predictions with AFM force measurements, Environ. Eng. Sci. 19 (2002) 413–427. [32] C.J. van Oss, Interfacial Forces in Aqueous Media, Marcel Dekker, New York, 1994. [33] J.A. Brant, A.E. Childress, Colloidal adhesion to hydrophilic membrane surfaces, J. Membr. Sci. 241 (2004) 235–248. [34] J.A. Brant, A.E. Childress, Assessing short-range membrane-colloid interactions using surface energetics, J. Membr. Sci. 203 (2002) 257–273. [35] L. Meagher, C. Klauber, R.M. Pashley, The influence of surface forces on the fouling of polypropylene microfiltration membranes, Colloid. Surf. A: Physicochem. Eng. Asp. 106 (1996) 63–81. [36] T.J. Senden, C.J. Drummond, Surface chemistry and tip-sample interactions in atomic force microscopy, Colloid. Surf. A: Physicochem. Eng. Asp. 94 (1995) 29–51.
[37] W.A. Ducker, T.J. Senden, R.M. Pashley, Direct measurement of colloidal forces using an atomic force microscope, Nature 353 (1991) 239–241. [38] J. Drelich, J.L. Wilbur, J.D. Miller, G.M. Whitesides, Contact angles for liquid drops at a model heterogeneous surface consisting of alternating and parallel hydrophobic/hydrophilic strips, Langmuir 12 (1996) 1913–1922. [39] M. Elimelech, J. Gregory, X. Jia, R.A. Williams, Particle Deposition and Aggregation: Measurement, Modeling and Simulation, Butterworth–Heinemann, Oxford, 1995. [40] B. Cappella, G. Dietler, Force–distance curves by atomic force microscopy, Surf. Sci. Rep. 34 (1999) 1–104. [41] J. Schaep, C. Vandecasteele, Evaluating the charge of nanofiltration membranes, J. Membr. Sci. 188 (2001) 129–136. [42] R.J. Hunter, Foundations of Colloid Science, Clarendon Press, Oxford, 1989. [43] C.J. van Oss, Acid–base interfacial interactions in aqueous media, Colloid. Surf. A 78 (1993) 1. [44] C.J. van Oss, R.F. Giese, W. Wu, On the predominant electron-donicity of polar solid surfaces, J. Adhes. 63 (1997) 71–88. [45] R.J. Hunter, Foundations of Colloid Science, vol. I, Oxford University Press, New York, NY, 1986. [46] W.D. Neese, Introduction to Mineralogy, Oxford University Press, New York, 2000. [47] R.J. Hunter, Zeta Potential in Colloid Science: Principles and Applications, Academic Press, London, 1981. [48] J.N. Israelachvili, Intermolecular and Surface Forces, 2nd ed., Academic Press Harcourt Brace Jovanovich, London, 1992. [49] E.M.V. Hoek, S. Bhattacharjee, M. Elimelech, Effect of membrane surface roughness on colloid-membrane DLVO interactions, Langmuir 19 (2003) 4836–4847. [50] L.W. Suresh, Direct measurement of the effect of surface roughness on the colloidal forces between a particle and flat plate, J. Colloid. Interf. Sci. 196 (1997) 177–190. [51] L.W. Suresh, Effect of surface roughness on the interaction energy between a colloidal sphere and a flat plate, J. Colloid. Interf. Sci. 183 (1996) 199–213. [52] S. Bhattacharjee, C.-H. Ko, M. Elimelech, DLVO interaction between rough surfaces, Langmuir 14 (1998) 3365–3375. [53] J.F.L. Duval, F.A.M. Leermakers, H.P. van Leeuwen, Electrostatic interactions between double layers: influence of surface roughness, regulation, and chemical heterogeneity, Langmuir 20 (2004) 5052–5065.
