ARTICLE pubs.acs.org/est
Depth Heterogeneity of Fully Aromatic Polyamide Active Layers in Reverse Osmosis and Nanofiltration Membranes Orlando Coronell,†,^ Benito J. Mari~nas,*,‡,^ and David G. Cahill§,^ †
Department of Environmental Sciences and Engineering, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599, United States ‡ Department of Civil and Environmental Engineering, §Department of Materials Science and Engineering, and ^Science and Technology Center of Advanced Materials for the Purification of Water with Systems, University of Illinois at UrbanaChampaign, Urbana, Illinois 61801, United States
bS Supporting Information ABSTRACT: We studied the depth heterogeneity of fully aromatic polyamide (PA) active layers in commercial reverse osmosis (RO) and nanofiltration (NF) membranes by quantifying near-surface (i.e., top ∼6 nm) and volume-averaged properties of the active layers using X-ray photoelectron spectrometry (XPS) and Rutherford backscattering spectrometry (RBS), respectively. Some membranes (e.g., ESPA3 RO) had active layers that were depth homogeneous with respect to the concentration and pKa distribution of carboxylic groups, degree of polymer cross-linking, concentration of barium ion probe that associated with ionized carboxylic groups, and steric effects experienced by barium ion. Other membranes (e.g., NF90 NF) had active layers that were depth heterogeneous with respect to the same properties. Our results therefore support the existence of both depth-homogeneous and depth-heterogeneous active layers. It remains to be assessed whether the depth heterogeneity consists of gradually changing properties throughout the active layer depth or of distinct sublayers with different properties.
’ INTRODUCTION Reverse osmosis (RO) and nanofiltration (NF) membranes are attractive technologies for water treatment because they are capable of removing a broad range of contaminants in one treatment step. In RO/NF membranes, contaminants are rejected as a result of differences between their permeation rate and that of water through an active layer (∼50200 nm) that sits on a porous polysulfone support (∼50 μm) backed by a nonwoven polyester fabric (∼200 μm).1 Fully aromatic crosslinked polyamide (PA) is the polymer of choice for the active layers of most RO and many NF membranes.13 Fully aromatic PA active layers are produced by interfacial polymerization. The polysulfone support is first soaked with an aqueous solution of m-phenylenediamine (MPD), and then brought into contact with a solution of trimesoyl chloride (TMC) in organic solvent (e.g., hexane).1 The resulting active layer structure has been proposed to be heterogeneous throughout its depth.1,46 For example, Wamser and Gilbert 5 proposed on the basis of contact angle titration studies that carboxylic and amine groups predominate at the PA surfaces that interface with the TMC and MPD solutions respectively during polymerization. Also, on the basis of a mathematical model describing the kinetics of interfacial r 2011 American Chemical Society
polymerization,6 Freger and Srebnik proposed that the carboxylicrich and amine-rich sublayers are loosely cross-linked and separated by a densely cross-linked core that serves as the true selective barrier. Such studies indicate that the properties of the surface of PA active layers, on which phenomena at the membrane-bulk feed interface depend, are different from the properties of the inner region of the active layer, on which watermembrane and solutemembrane interactions depend during water and solute transport from the feed to the permeate side. As a result, to improve the current understanding of permeation phenomena during RO and NF membrane filtration, there is a need to characterize as a function of depth the physicochemical properties and interactions with contaminants in active layers. Unfortunately, there are no experimental techniques capable of resolving heterogeneity in active layers as a function depth; however, we have shown 3,79 that Rutherford backscattering spectrometry (RBS) can be used to quantify volume-averaged properties of active layers, and in this study we show that X-ray Received: January 2, 2011 Accepted: March 22, 2011 Revised: March 21, 2011 Published: April 13, 2011 4513
dx.doi.org/10.1021/es200007h | Environ. Sci. Technol. 2011, 45, 4513–4520
Environmental Science & Technology photoelectron spectroscopy (XPS) can be used to characterize corresponding near-surface properties. Accordingly, the objective of this study was to compare the nearsurface (∼6 nm) and volume-averaged values of several physicochemical properties of the fully aromatic PA active layers of eight commercial RO/NF membranes. The properties investigated included: (i) elemental composition,7,10 (ii) concentration of carboxylic groups,8 (iii) degree of polymer cross-linking,3,8 and (iv) concentration of barium ion (Ba2þ) that associated with deprotonated carboxylic groups (RCOO) at pH ∼10.3,9 Additionally, we compared the effect of pH on two active layer properties for one RO and one NF membrane: (v) ionization behavior of carboxylic groups;8 and (vi) steric effects experienced by Ba2þ.3,9
’ MATERIALS AND METHODS (SEE EXTENDED VERSION IN THE SUPPORTING INFORMATION) Membranes. Experiments were performed with coupons (2.5 5.0 cm2) of the following membranes: ESPA3 RO, SWC5 RO, ESPAB RO and ESNA NF (Hydranautics, Oceanside, CA), FT30 RO and NF90 NF (Dow Liquid Separation, Midland, MI), TFCS NF (Koch Membrane Systems, Wilmington, MA), and LF10 RO (Nitto Denko, Japan). Attenuated total reflectance-Fourier transformed infrared spectroscopy (ATRFTIR) analyses 2,3 revealed (data not shown) that all membranes had fully aromatic PA active layers. While FT30 RO and LF10 RO membranes have a hydrophilic, nitrogen-free coating on top of their active layers,3,8,11 no coating was detected on the other membranes using ATR-FTIR 2 and RBS analyses 3 (data not shown). Prior to sample preparation, membrane coupons were rinsed with and stored in nanopure (g17.8 MΩ 3 cm) water as detailed in our previous work.9 Experiments were performed at room temperature (21 ( 2 C), and at least two replicates were prepared for each experimental condition (see Table 1 for details). After sample preparation, a piece (≈ 2.5 1.0 cm2) was cut from each sample for XPS analyses and the rest of the sample was analyzed by RBS. Ion-Probe Solutions and Chemicals. Ion-probe solutions were prepared with nanopure water and ACS grade chemicals of 99%þ purity. Silver nitrate (AgNO3) and barium chloride dihydrate (BaCl2 3 2H2O) were used as sources of silver (Agþ) and barium (Ba2þ) ions respectively in probing experiments; barium nitrate (Ba(NO3)2) was used in AgþBa2þ displacement tests. The concentrations of silver (106103 M) and barium (1060.32 M) in solution were below their solubility limits.12,13 The pH of ion-probe solutions during Ba2þ-probing experiments was adjusted with HCl or NaOH; for experiments involving silver, HNO3 was used instead of HCl. Silver solutions were prepared and used under dim red light to avoid photoreactivity. Ion Probing with Silver (Agþ) and Barium (Ba2þ). Ion probing with Agþ and Ba2þ was used to study the concentration of RCOO and the concentration of Ba2þ that associates with accessible RCOO respectively in active layers. Details on sample preparation procedures can be found elsewhere.8,9 In brief, for probing with Agþ, coupons were immersed in concentrated AgNO3 solution at the pH of interest to saturate RCOO with Agþ. The silver concentration in solution and duration of membrane immersion ensured negligible kinetic limitations for the saturation of accessible RCOO with Agþ. Next, immersions were performed in dilute (106 M) AgNO3 solution at the same pH of the saturation step to lower the concentration of Agþ not ionically associated with RCOO below the detection levels of XPS and RBS. Then, the samples were dried using filter paper, and
ARTICLE
air-dried at room temperature for g24 h. Ion-probing with Ba2þ was performed in a similar fashion to probing with Agþ, but using an ionprobe solution of BaCl2 3 2H2O. ESPA3 RO and NF90 NF membranes were probed in the pH range 3.4310.33. All other membranes were probed only at pH ∼10. AgþBa2þ Displacement Tests. The stoichiometry of association between Ba2þ and RCOO in the active layer of the ESPA3 RO membrane was quantified using ion-displacement tests in which first Agþ was used to saturate RCOO groups and then Ba2þ was used to displace Agþ. Details on experimental procedures can be found elsewhere.9 Whereas the concentration of Agþ displaced by Ba2þ from the active layer (Δ[Agþ]) provided the concentration of sites neutralized by Ba2þ, the concentration of Ba2þ neutralizing those sites was given by the Ba2þ concentration ([Ba2þ]) after ion displacement. Accordingly, the stoichiometry of association between RCOO groups and Ba2þ at a concentration Δ[Agþ] of neutralized sites was calculated as NN = Δ[Agþ]/[Ba2þ], where NN is referred to as the neutralization number. AgþBa2þ displacement tests were performed in the pH range 8.0210.33. XPS Analyses. We used XPS to quantify membrane surface elemental compositions including surface elemental fractions of ion probes. Prior to XPS analyses, samples were vacuum-dried for ∼72 h. XPS analyses were performed with a photoelectron spectrometer employing an MgKR X-ray source (1253.6 eV). High-resolution (0.1 eV) scans were collected for carbon (C1s), oxygen (O1s), nitrogen (N1s), chlorine (Cl2p), silver (Ag3d), and barium (Ba3d). Data were collected from a circular area with a diameter of 0.22 mm. From our experimental settings it can be calculated (section S2 of the Supporting Information) that ∼95% of the XPS signal was collected from within ∼6 nm of the membrane sample surface. Accordingly, XPS results describe the near-surface region of the PA active layer for uncoated membranes, and a combination of the near-surface regions of the coating and PA active layer for coated membranes. RBS Analyses. We used RBS analyses to quantify volumeaveraged elemental compositions 7,10 and concentrations of ion probes 3,8 in active layers. For coated membranes, analyses of RBS data took into account the presence of the coat, that is, we obtained the properties of the PA active layer alone. The differences between XPS and RBS results were attributed to depth heterogeneity in the active layer only for uncoated membranes; for coated membranes, these differences describe the dissimilarities between the near-surface properties of the membrane (i.e., a combination of coating and active layer) and the volume-averaged properties of the active layer. Details on RBS experimental procedures and data analysis were reported previously.79,14,15 RBS results for ESPA3 RO and NF90 NF membranes were reproduced from our previous studies.3 RBS results for all other membranes were obtained for this study. Atomic Force Microscopy (AFM) Analyses. AFM was used to characterize membrane surface roughness. Samples were air-dried for ≈48 h before analyses. An AFM microscope equipped with BudgetSensors Tap300Al tips (Sofia, Bulgaria) was used in tapping mode to scan at least duplicate 10 10 μm2 areas on the surface of each membrane. Roughness was calculated as root-mean-square roughness (Rrms).16
’ RESULTS AND DISCUSSION Elemental Composition. Table 1 shows the average elemental compositions obtained by XPS and RBS. Elemental percentages exclude hydrogen because hydrogen is not quantifiable by XPS and 4514
dx.doi.org/10.1021/es200007h |Environ. Sci. Technol. 2011, 45, 4513–4520
Environmental Science & Technology
ARTICLE
Table 1. Summary of Results from XPS, RBS, Agþ-Probing/XPS/RBS, and AFM Analyses of Uncoated ESPA3 RO, SWC5 RO, TFCS NF, ESNA NF, NF90 NF and ESPAB RO Membranes, and Coated FT30 RO and LF10 RO Membranes; Uncertainties Represent Standard Deviation a between Samples Tested membrane
ESPA3 RO
SWC5 RO
TFCS NF
ESNA NF
ESPAB RO
NF90 NF
FT30b RO
LF10b RO
ratingc
BW RO
SW RO
NF
NF
BW RO
NF
SW RO
BW RO
coating
no
no
no
no
no
no
yes
yes
XPSi
74.2 ( 0.4
77.8 ( 0.6
73.0 ( 0.9
73.6 ( 0.7
73.6 ( 0.6
76.0 ( 0.4
71.0 ( 1.6
77.9 ( 2.8
RBSi
71.6 ( 1.1
72.7 ( 1.3
72.5 ( 0.4
71.8 ( 0.7
73.7 ( 1.5
74.2 ( 1.8
71.0 ( 1.2
71.1 ( 1.5
XPS2j
74.3 ( 1.0
XPSi
13.9 ( 0.5
11.3 ( 0.7
15.2 ( 0.8
13.7 ( 0.5
15.6 ( 0.6
13.3 ( 0.2
26.2 ( 1.7
20.2 ( 2.8
RBSi
14.2 ( 0.5
13.0 ( 0.5
13.9 ( 0.1
13.2 ( 0.4
13.1 ( 0.7
13.7 ( 0.9
17.4 ( 0.4
18.1 ( 1.0
XPS2j XPSi
14.2 ( 0.8 11.1 ( 0.1
10.3 ( 0.3
10.9 ( 0.3
11.3 ( 0.2
15.2 ( 0.9 10.7 ( 0.1
10.6 ( 0.2
2.8 ( 1.7
1.7 ( 0.2
RBSi
12.9 ( 0.7
12.8 ( 0.7
12.4 ( 0.3
13.1 ( 0.4
13.1 ( 0.9
12.4 ( 1.1
11.5 ( 0.6
9.3 ( 0.6
XPS2j
10.7 ( 0.4
%C
%O
%N
%Cl
O:N ratiod
%RCOOe
73.9 ( 0.9
10.8 ( 0.3
XPSi
0.8 ( 0.0
0.6 ( 0.1
0.9 ( 0.1
1.3 ( 0.1
0.1 ( 0.0
0.1 ( 0.1
0.0 ( 0.0
0.2 ( 0.0
RBSi
1.3 ( 0.1
1.5 ( 0.1
1.2 ( 0.1
1.9 ( 0.1
0.2 ( 0.1
0.2 ( 0.1
0.1 ( 0.0
1.4 ( 0.0
XPS2j
0.8 ( 0.1
0.1 ( 0.0
XPSi
1.26 ( 0.04
1.09 ( 0.09
1.39 ( 0.07
1.21 ( 0.05
1.46 ( 0.07
1.26 ( 0.03
16 ( 14
12 ( 2
RBSi XPS2j
1.10 ( 0.05 1.33 ( 0.09
1.01 ( 0.02
1.12 ( 0.02
1.00 ( 0.00
1.00 ( 0.04 1.41 ( 0.10
1.06 ( 0.06
1.52 ( 0.05
1.95 ( 0.08
XPSk
0.69 ( 0.02
0.42 ( 0.04
0.54 ( 0.09
0.52 ( 0.03
0.43 ( 0.01
0.57 ( 0.01
0.24 ( 0.07
0.32 ( 0.05
RBS l
0.71 ( 0.01
0.41 ( 0.01
0.55 ( 0.01
0.54 ( 0.01
0.23 ( 0.02
0.43 ( 0.01
0.48 ( 0.02
0.53 ( 0.01
(%RCOO)/(%N) XPSk 0.063 ( 0.002 0.040 ( 0.003 0.049 ( 0.008 0.046 ( 0.003 0.040 ( 0.001 0.054 ( 0.001 0.092 ( 0.048 0.178 ( 0.013 RBSl
0.054 ( 0.004 0.031 ( 0.001 0.043 ( 0.002 0.042 ( 0.000 0.018 ( 0.003 0.034 ( 0.005 0.040 ( 0.003 0.059 ( 0.006
XPS k
94.1 ( 0.1
96.2 ( 0.3
95.3 ( 0.7
95.6 ( 0.3
96.1 ( 0.1
94.9 ( 0.0
91.7 ( 3.9
RBS l
94.9 ( 0.3
97.0 ( 0.1
95.9 ( 0.1
96.0 ( 0.0
98.3 ( 0.3
96.7 ( 0.4
96.2 ( 0.2
94.4 ( 0.5
ng
XPS k RBS l
0.62 ( 0.01 0.68 ( 0.02
0.76 ( 0.02 0.81 ( 0.005
0.70 ( 0.05 0.74 ( 0.01
0.73 ( 0.02 0.75 ( 0.001
0.76 ( 0.01 0.89 ( 0.02
0.68 ( 0.003 0.79 ( 0.02
0.45 ( 0.29 0.76 ( 0.01
0.00 ( 0.08 0.64 ( 0.02
xg
XPS k
0.38 ( 0.01
0.24 ( 0.02
0.30 ( 0.05
0.27 ( 0.02
0.24 ( 0.01
0.32 ( 0.00
0.55 ( 0.29
1.00 ( 0.08
RBS l
0.32 ( 0.02
0.19 ( 0.00
0.26 ( 0.01
0.25 ( 0.00
0.11 ( 0.02
0.21 ( 0.02
0.24 ( 0.01
0.36 ( 0.02
98.7 ( 22.1
100 ( 0.7
21.8 ( 0.7
43.3 ( 2.8
140.3 ( 33.9
58.2 ( 2.8
75.5 ( 3.8
71.1 ( 11.8
DPCpH≈10f
Rrmsh (nm)
84.9 ( 0.9
a
For experimental conditions where only duplicates (A and B) were run, the standard deviation was calculated as the difference between either A or B and their average value. b We previously reported 3 the PA active layers of FT30 RO and LF10 RO membranes to have coats of approximate composition and thickness C0.32O0.12H0.56 and 25 ( 35 nm, and C0.28O0.12H0.60Cl0.001 and 58 ( 33 nm, respectively. c RO, reverse osmosis; NF, nanofiltration; BW, brackish water; SW, seawater d Average elemental O:N percent ratios were calculated as the average between the O:N ratios of the replicates (not as the ratio between the average %O and the average %N). e %RCOO = %Ag values measured at pH ∼10. f DPC (degree of polymer cross-linking) ≈ DPCpH∼10 was calculated from eq 4 using the %RCOO/%N values at pH ∼10 shown in the previous row of this table. g n and x represent the fraction of fully aromatic polyamide repeating units that are fully cross-linked, and that contain a carboxylic group respectively, and were calculated assuming that the concentration of amine groups in the active layer is negligible compared to that of carboxylic groups (i.e., n þ x = 1). h The root-mean-square roughness (Rrms) value for each membrane is the average of at least two replicates, each of which was obtained by scanning a 10 10 μm2 surface area with the atomic force microscope (AFM) in tapping mode. In particular, five replicates were obtained for brackish water RO membranes ESPA3 and ESPAB for which the corresponding roughness values were found to be highly dependent on the location of the spot analyzed. i Values in this row are the average of 46 replicates. j Values in this row are the average of 50 replicates. k Values in this row are the average of 24 replicates. l Values in this row are the average of 23 replicates.
does not affect the XPS signal intensity of other elements. XPS results for ESPA3 RO and ESPAB RO membranes show that for uncoated membranes there was no significant difference between elemental compositions obtained with 4 replicates (i.e., XPS rows in Table 1) and 50 replicates (i.e., XPS2 rows in Table 1). Consequently, the lower number of replicates was used for XPS analyses of all other membranes. Table 1 also shows that the nitrogen content measured in the near-surface region of uncoated active layers by XPS was 1420% lower than the corresponding volume-averaged value measured by RBS. Similarly, even though the difference between oxygen contents measured by XPS and RBS did not show any
specific trend, the O:N ratio measured by XPS was 846% higher than the RBS value. Given that in PA active layers the nitrogen content contributed by amide links is g99% of the total nitrogen,3,8 then the lower nitrogen content (and higher O:N ratios) obtained by XPS suggests a reduced surface concentration of amide links. Since there is limited information about protocols used for membrane production, it is not possible to provide a definite explanation of why XPS results show a higher O:N ratio than RBS results. Table 1 shows that chlorine content was also lower in the nearsurface region of the PA active layers. We previously 3,7,8 showed that chlorine content in the PA active layers of commercial membranes is 4515
dx.doi.org/10.1021/es200007h |Environ. Sci. Technol. 2011, 45, 4513–4520
Environmental Science & Technology
ARTICLE
not of ionic nature. Also, it has been reported that chlorination of PA membranes results in chlorine substitution of the amidic hydrogen and/or chlorine addition to the meta and ortho positions in the nitrogen-bonded benzene rings.1,17,18 Accordingly, membrane chlorination should result in lower concentrations of permanently bound chlorine in regions of the active layers with lower concentrations of amide links, which is consistent with the lower nitrogen and chlorine contents found in the near-surface region of the active layers. Because previous studies 18,19 show that, under appropriate chlorine exposure conditions, chlorination of PA membranes increases water flux without significant loss of salt rejection, we suggest that covalentbonded chlorine in PA active layers results from controlled treatment of the membranes with hypochlorite by manufacturers. We attribute the relatively higher standard deviations in elemental compositions obtained by XPS for the coated FT30 RO and LF10 RO membranes, compared to those obtained for uncoated membranes, to the uneven thicknesses of their coatings (i.e., δFT30 COAT = 25 ( 35 nm and δLF10 COAT = 58 ( 33 nm 3). The significantly high standard deviations of the coating thicknesses are consistent with transmission electron microscope (TEM) images reported by Tang et al. 11 of a coated FT30 RO membrane that showed uneven coverage of the PA active layer by the coating. Table 1 also shows that, compared to the uncoated membranes, the coated membranes had (i) significantly lower surface contents of chlorine and nitrogen, and (ii) significantly higher surface O:N ratios. Both findings are consistent with the use of cross-linked polyvinyl alcohol as coating material.1,20 The higher variability of the surface nitrogen content for the FT30 RO membrane compared to for the LF10 RO membrane is consistent with the thicker coating of the latter membrane; a thicker coating likely provides a more even coverage of the PA layer. Concentration of Ionized RCOO Groups. In our previous studies,3,8 we measured the volume-averaged concentration of ionized carboxylic groups in PA active layers by probing the RCOO with Agþ, and quantifying the resulting Agþ content ([Agþ]) in the active layers by RBS. It was assumed that [Agþ] = [RCOO] based on: (i) the one-to-one correspondence between Agþ and RCOO, and (ii) the fact that the Agþ is smaller than the smallest pores reported for PA active layers.8 In this study, we used the Agþ-probing procedure together with XPS and RBS to quantify near-surface concentrations and volume-averaged concentrations, respectively, of ionized carboxylic groups in active layers. We chose pH ∼10 for sample preparation because our previous research3,8 indicates that at pH = 10 the degree of ionization of carboxylic groups in PA active layers is g90%; therefore, performing Agþ-probing/XPS/RBS experiments at pH ≈ 10 allowed us to simultaneously assess the depth heterogeneity of carboxylic groups and quantify the approximate total content of carboxylic groups in active layers. The results for the content of RCOO are presented in Table 1 (molar units) and part a of Figure 1 (elemental percentages). Molar concentrations were calculated as (section S3 of the Supporting Information) 0
½RCOO ¼ εAg
∑
C, O, N, Cl
FPA 0 0 ðεi Mi Þ þ 0:67 εC MH
ð1Þ where ε0i , Mi, and FPA = 1.24 g/cm3 21 are the elemental fraction (excluding hydrogen) of element i, molar mass of element i, and dry density of polyamide, respectively, and subscripts Ag, C, and H
Figure 1. Concentrations of (a) ionized carboxylic groups ([RCOO]) and (b) barium ions ([Ba2þ]) that associate with accessible carboxylic groups at pH ∼10 in the polyamide active layers of eight commercial RO/NF membranes near the active layer surface (i.e., XPS data, solid bars) and in average throughout the volume of the active layer (i.e., RBS data, cross-hatched bars). Error bars indicate standard deviation. LF10 RO and FT30 RO membranes are coated, all other membranes are uncoated.
indicate respective silver, carbon, and hydrogen. Results show that the content of RCOO in the active layer was depth heterogeneous for the uncoated membranes ESPAB RO and NF90 NF with an RCOO content >33% higher in the near-surface region compared to the volume-averaged value. In contrast, the content of RCOO was depth homogeneous in the remaining uncoated membranes (i.e., ESPA3 RO, SWC5 RO, TFCS NF, and ESNA NF). Consequently, even though all six uncoated active layers had lower nitrogen contents in the near-surface region, for four uncoated active layers there was no significant difference between the near-surface and volume-averaged values obtained for the concentration of carboxylic groups. For coated membranes, all charge density was attributed to RCOO in the PA active layer based on reports 1,20 indicating that coatings are made of nonionic polymers. The measured surface charge indicates that the RCOO content in the surface was >40% lower than the average content in the PA active layer, consistent with the partial coverage of the PA active layer by the coating. Ohshima and Kondo 22 proposed a model that predicts that the electric potential in the vicinity of a flat charged film permeable to 4516
dx.doi.org/10.1021/es200007h |Environ. Sci. Technol. 2011, 45, 4513–4520
Environmental Science & Technology
ARTICLE
The depth heterogeneity of the total concentration (CT,RCOOH) and pKa distribution of carboxylic groups in ESPA3 RO and NF90 NF membranes was studied by fitting the XPS and RBS data for the RCOO concentration as a function of pH to 3 ! n 10pKa, i ½RCOO ¼ CT , RCOOH wi pH 10 þ 10pKa, i i¼1
∑
ð2Þ
Figure 2. Concentrations of ionized carboxylic groups ([RCOO]) and barium ions ([Ba2þ]) that associate with accessible ionized carboxylic groups as a function of pH in the polyamide active layers of (a) ESPA3 RO and (b) NF90 NF membranes near the active layer surface (i.e., XPS data, solid symbols) and in average throughout the volume of the active layer (i.e., RBS data, open symbols). Error bars indicate standard deviation. (Red) Continuous and (green) dashed lines represent the fit to eq 2 of the XPS and RBS data respectively for the concentration of ionized carboxylic groups. XPS data was obtained for this study, and RBS data was reproduced from our previous work.3
ions is determined by the charge density in the top ∼2 nm of the film. Assuming that the OhshimaKondo model is correct, then because 95% of the silver signal in our XPS analyses comes from within ≈5 nm of the membrane surface (section S2 of the Supporting Information), our results indicate that for some PA active layers the near-surface region controlling the electric exclusion of ionic contaminants has a different charge density than the volume-averaged value of the active layer. RO/NF membranes, however, are not flat as assumed in the Ohshima-Kondo model. In fact, the roughness scale of PA active layers (Table 1) is significantly greater than the double-layer thickness (
