Transport of small and neutral solutes through reverse osmosis 1 membranes: Role of skin layer conformation of 2 the polyamide film

21 The polyamide skin layer of reverse osmosis (RO) membranes was characterised using 22 advanced and complementary analytical techniques to investigate the mechanisms underlying 23 the permeation of contaminants of emerging concern in potable water reuse – N - 24 nitrosodimethylamine (NDMA) and N -nitrosomethylethylamine (NMEA). This study used 25 five RO membrane samples with similar membrane properties. The five RO membrane samples 26 spanned over a large range of water permeance (0.9–5.8 L/m 2 hbar) as well as permeation of 27 NDMA (9–66%) and NMEA (3–29%). Despite such distinctthese differences among the five 28 RO membranes, characterisations of the skin layer using positron annihilation lifetime 29 spectroscopy, atomic force microscopy and field emission scanning electron microscopy 30 revealed almost no variation difference in their free-volume hole-radius (0.270–0.275 nm), 31 effective surface area (198–212%) and thickness (30–35 nm) of the skin layer. The results 32 suggest that there could be other RO skin layer properties, such as the interconnectivity of the 33 protuberances within the polyamide skin layer additional to the free-volume hole-size and 34 thickness of the skin layer, which can also govern water and solute permeation.


39
N-nitrosodimethylamine (NDMA; C2H6N2O) and N-nitrosomethylethylamine (NMEA; 40 C3H8N2O) are micropollutants of significant concern in potable water reuse since they are 41 probable carcinogenic chemicals [1]. With a molecular weight of 74 g/mol, NDMA is the 42 smallest in the N-nitrosamine group. NDMA and NMEA are neutral and hydrophilic 43 compounds at environmental pH (i.e. pH 6-8). Although reverse osmosis (RO) membrane 44 separation can achieve excellent rejection of a range of impurities in reclaimed water including 45 salts, macro-organics, and many micropollutants, the rejection of NDMA, NMEA and several 46 other N-nitrosamines is low and highly variable because of its small molecular size and lack of 47 charge [2][3][4][5]. Thus, they are often detectable in RO permeate at concentrations higher than 48 guideline or target values set by water authorities around the world. For example, California 49 has established a notification level of 10 ng/L for NDMA and a public health goal of 3 ng/L 50 [6]. Similarly, in Australia, the guideline value of NDMA in water intended for potable reuse 51 has been also set at 10 ng/L [7]. The low and highly variable separation performance of RO 52 with respect to NDMA rejection necessitates post treatment by advanced oxidation (UV 53 irradiation and H2O2) [8]. Recent research [9] suggests that NDMA rejection by RO 54 membranes varies significantly amongst the many RO membranes available on the market. 55 Thus, further insights which lead to better membrane selection and improvement of the 56 separation performance of RO for N-nitrosamine removal can directly contribute to the 57 economic viability and public safety of potable water reuse. 58 Given the significant importance of low molecular weight micropollutants in potable reuse, 59 numerous previous studies have been conducted to reveal the permeation mechanisms of 60 micropollutants through RO membranes [10][11][12][13]. The significance of steric (size) interaction 61 between solutes and the free-volume holes within the RO membrane active skin layer has been 62 The free-volume holesholes in the membrane skin layer in polymeric matrixare thought 69 to play an important role in water and solute transport through the RO membrane.

Membranes and membrane treatment system
116 Two commercially available RO membranesnamely ESPA2 and ESPABand a prototype 117 RO membrane were obtained as flat sheet samples from Hydranautics/Nitto (Osaka, Japan). 118 The active skin layers of these membranes have similar chemical ingredients although the 119 detailed information is proprietary. The ESPA2 membrane has been employed in many potable 120 water reuse schemes [14], while the ESPAB membrane is designed for boron removal and has 121 been widely used in the second pass of RO seawater desalination plants. In addition, samples 122 of the ESPAB and Prototype membranes were also subjected to heat treatment to alter the 123 physical properties. These heat-treated samples are designated as heated ESPAB and heated 124 Prototype, respectively. Thus, in total, five different membrane samples were used in this 125 investigation. 126

Experimental protocols
where r (< 1nm) is approximated as a spherical shape. Positron irradiation was carried out 161 under vacuum (~ 10 -5 Pa) and about 2 × 10 6 positron annihilation events were collected for the 162 positron lifetime spectrum of each sample. Spectra were analysed using a non-linear least-163 squares fitting program. Unless otherwise stated, the incident energy (Ein) was set at 1.0 keV, 164 which corresponds to a mean implantation depth of 31 nm from the top (implantation depth 165 range = 0-90 nm) with a material density of 1.3 g/cm 3 (Fig. 1)  pre-treatment steps involving the replacement of water in the membranes with tert-Butyl 177 alcohol followed by freeze drying. Images were obtained in air using tapping mode with a 178 silicate cantilever. The scanning area was 5 μm × 5 μm. The effective surface area of each 179 membrane was calculated based on the data of three samples. Effective surface area here was 180 defined as a ratio between the actual (measured) area and the sample area as described in the 181 following formula: 182 permeance as well as solute passage with respect to both NDMA and NEMA (Fig. 2). Heat 195 treatment was effective to reduce solute passage and water permeance. After heat treatment, 196 NDMA passage through the ESPAB and the Prototype membranes decreased from 56 to 37% 197 and from 18 to 9%, respectively. The pure water permeance of these membrane also 198 proportionally decreased as can be seen from Fig. 2. A strong linear correlation between solute 199 passage with respect to both NDMA and NMEA and water permeance can be confirmed in   Prototype, respectively. This indicates that these RO membranes have a similar chemical 220

Characterisations of the RO skin layer
propertycomposition. In contrast,It is noted that heat treatment increased the peak intensity 221 ratio from 0.21 to 0.30 and from 0.17 to 0.18 for ESPAB and Prototype membrane, respectively. 222 The cause of the changes in the peak intensity for ESPAB after heat treatment remains unclear, 223 but it will be in the scope of our future study.

Free-volume hole-radius 225
The mean free-volume hole-radius of the selected RO membranes was determined at a mean 226 implantation depth of 31 nm using τo-Ps values (pick-off annihilation lifetime of o-Ps) from 227 PALS analysis (Table S3). The free-volume hole-radius of the three unheated RO membranes 228 (i.e. ESPA2, ESPAB and Prototype) was almost identical, ranging from 0.270 to 0.275 nm (Fig.  229   3). Heat treatment did not show any discernible impact on the free-volume hole-radius. It is 230 noteworthy that PALS analysis at other implantation depths (i.e. 10 and 59 nm) of the ESPAB 231 membrane did not show any significant variation in the free-volume hole-radius due to heat 232 treatment (Fig. S4). It is noted that current PALS technique cannot confirm a small difference 233 in free-volume hole-radius of RO membranes less than 0.01 nm due to the inherent errors in 234 PALS and the inhomogeneity of the membrane samples. Thus, the free-volume hole-radius of 235 all five membrane samples in Fig. 3 are considered to be similar. molecules, a variation of 0.01 nm 2 in free-volume hole-area among the five RO membranes 244 may still be an important factor. However, there was no observable correlation between the 245 measured free-volume hole-radius and the passage of NDMA and NMEA (Fig. S5). Given the 246 similar free-volume hole-size of the five membrane samples, these results suggest that a factor 247 other than the free-volume hole-size can also govern the permeation of NDMA and NMEA by 248 these RO membranes. 249

Effective surface area 250
The effective membrane surface area was determined by taking into account the topography of 251 the RO skin layer at the microscopic level (i.e. surface roughness) using AFM. Indeed, at the 252 microscopic level, the effective membrane surface area can differ considerably from the surface 253 area normally used to calculate the permeate flux [9]. It is noted that permeate flux 254 considerably influences NDMA permeation [13]. Since the skin layer can play an important 255 role in solute permeation through the RO membrane as proposed in literature [30,38], it is 256 important to take into account the effective membrane surface area when assessing solute 257 permeation. 258 Despite the large variation in the visualized "ridge-and-valley" structure among the three 259 different types of RO membranes (i.e. ESPA2, ESPAB and Prototype), their effective surface 260 area was almost identical, ranging from 198 and 212% (Fig. 4). In other words, the effective 261 membrane surface area at the microscopic level is approximately two times the plain area. 262 Likewise, heat treatment did not cause any discernible changes in the effective surface area. 263 Results from Fig. 4 confirm that separation experiments in this study were also at the same 264 permeate flux for a systematic comparison of solute permeation among all selected RO 265 membranes. More importantly, the observation of the large variation in permeance (Fig. 2) and 266 almost identical effective surface area (Fig. 4)

Thickness 276
The thickness of the skin layer of the selected RO membranes was evaluated using a cross-277 sectional view obtained by FE-SEM. The FE-SEM analysis (Fig. 5)  and 34 nm, respectively (Fig. 6). Results in Fig. 6 indicate that there was no discernible 306 variation in thickness among the three RO membranes in this study. In addition, heat treatment 307 did not appear to alter the crumpled film thickness. Indeed, differences in the crumpled film 308 thickness between heated and unheated samples were within the measurement error margin (i.e. 309 standard deviation of two samples of the same membrane). As a result, in this study, variation 310 in water flux and the passage of NDMA and NMEA cannot be attributed to the measured 311 crumpled film thickness (Fig. S7), suggesting that other skin layer properties such as 312 conformation of the crumpled polyamide films could also govern their permeation.

318
Comprehensive analysis of the skin layer of five RO membrane samples using PALS, FE-SEM 319 and AFM revealed that there could be other RO skin layer properties besides the free-volume 320 hole-radius and thickness of the crumpled film that can govern water and solute permeation. 321 This is a significant finding in membrane transport, because the free-volume hole-radius and 322 thickness of the crumpled film have often been considered the only membrane properties 323 governing the membrane transport. 324 FE-SEM images obtained here identified that the free-volume hole-radius analysed by PALS 325 was likely to result from the crumpled polyamide film. According to the distribution of positron 326 implantation depth (Fig. 1) In addition to the swelling effects, chemical properties of the internal skin layer could be a 362 major contributor to a variation in diffusion coefficient and sorption coefficient of water and 363 solutes, which ultimately leads to a variation in their water permeance and separation 364 performance. Typically, increases in the degree of polyamide cross-linking can cause less water 365 property other than the free-volume hole-size and thickness that can also govern the transport 379 of water and small and neutral solutes such as NDMA and NMEA that are of signifincant 380 concern in potable water reuse. Such property is likely to be the protuberance conformation or 381 interconnectivity of the protuberance within the membrane polyamide skin layer. In addition, 382 FE-SEM data also reveal that current PALS technique may not be suitable for determining 383 free-volume hole-radius of the flat polyamide film located at the interface between the 384 polyamide skin and the polysulfone supporting layer beneath the crumpled polyamide films. 385 Technology (MEXT), Japan. We also thank Hydranautics/Nitto for providing RO