High Performance Nanofiltration Membrane for Effective Removal of

May 31, 2018 - Environmental Science & Technology Letters. Li, Karanikola, Zhang, Wang, and Elimelech. 2018 5 (5), pp 266–271. Abstract: Conventiona...
3 downloads 0 Views 4MB Size
Subscriber access provided by Universiteit Utrecht

Environmental Processes

High Performance Nanofiltration Membrane for Effective Removal of Perfluoroalkyl Substances at High Water Recovery Chanhee Boo, Yunkun Wang, Ines Zucker, Youngwoo Choo, Chinedum O. Osuji, and Menachem Elimelech Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.8b01040 • Publication Date (Web): 31 May 2018 Downloaded from http://pubs.acs.org on May 31, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 29

Environmental Science & Technology

1 2

High Performance Nanofiltration Membrane for Effective Removal of Perfluoroalkyl Substances at High Water Recovery

3 4 5 6 7 8 9 10 11

Chanhee Boo,† Yunkun Wang,§†* Ines Zucker,† Youngwoo Choo,†

12

Chinedum O. Osuji,† and Menachem Elimelech†*

13 14 15 16 17 18 19

§

Shandong Key Laboratory of Water Pollution Control and Resource Reuse, School of Environmental Science and Engineering, Shandong University, Jinan 250100, China



Department of Chemical and Environmental Engineering, Yale University, New Haven, Connecticut 06520-8286

20 21 22 23 24 25 26 27 §†*

Corresponding author: email: [email protected]; Tel: +86 (531) 88365400

28 29

†*

Corresponding author: email: [email protected]; Tel: +1 (203) 432-2789

1 ACS Paragon Plus Environment

Environmental Science & Technology

30

ABSTRACT

31

We demonstrate the fabrication of a loose, negatively charged nanofiltration (NF) membrane

32

with tailored selectivity for the removal of perfluoroalkyl substances with reduced scaling

33

potential. A selective polyamide layer was fabricated on top of a polyethersulfone support via

34

interfacial polymerization of trimesoyl chloride and a mixture of piperazine and bipiperidine.

35

Incorporating high molecular weight bipiperidine during the interfacial polymerization enables

36

the formation of a loose, nanoporous selective layer structure. The fabricated NF membrane

37

possessed a negative surface charge and had a pore diameter of ~1.2 nm, much larger than a

38

widely used commercial NF membrane (i.e., NF270 with pore diameter of ~0.8 nm). We

39

evaluated the performance of the fabricated NF membrane for the rejection of different salts (i.e.,

40

NaCl, CaCl2, and Na2SO4) and perfluorooctanoic acid (PFOA). The fabricated NF membrane

41

exhibited a high retention of PFOA (~90%) while allowing high passage of scale-forming

42

cations (i.e., calcium). We further performed gypsum scaling experiments to demonstrate lower

43

scaling potential of the fabricated loose porous NF membrane compared to NF membranes

44

having a dense selective layer under solution conditions simulating high water recovery. Our

45

results demonstrate that properly designed NF membranes are a critical component of a high

46

recovery NF system, which provide an efficient and sustainable solution for remediation of

47

groundwater contaminated with perfluoroalkyl substances.

48 49 50

TOC Art

51 52 53 54 2 ACS Paragon Plus Environment

Page 2 of 29

Page 3 of 29

Environmental Science & Technology

55

INTRODUCTION

56

Perfluoroalkyl substances (PFASs) are emerging contaminants that are persistent in the

57

environment, bio-accumulative, and toxic even at trace concentrations. 1-5 PFASs are used in

58

production of firefighting foams and their widespread use, particularly at airports, oil refineries,

59

and military bases, has led to severe groundwater contamination. 6-9 Because of their unique

60

physicochemical properties, such as strong carbon-fluorine bond and low vapor pressure, PFASs

61

are highly resistant to degradation by chemical or biological processes. 10-12

62

Nanofiltration (NF) is a low-pressure membrane-based separation process widely used in

63

water and wastewater treatment.13, 14 NF has the potential to provide an effective solution for the

64

removal of perfluorooctanoic acid (PFOA) — an important class of PFASs with significant

65

environmental concerns due to widespread industrial use 15 and frequent detection in drinking

66

water resources7 — from groundwater at high water recoveries.16, 17 Separation of PFOA by NF

67

is based on size (steric) exclusion and electrostatic interactions. 18 PFOA has relatively high

68

molecular weights (414 g/mol) and thus can be retained by NF membranes via size exclusion.

69

PFOA removal can be further enhanced via electrostatic repulsion by tailoring the NF membrane

70

surface and pore charge, because PFOA with a low dissociation constant (pKa of –0.1) is

71

negatively charged at the pH of natural waters.19 Several studies have investigated the removal of

72

PFOA by commercial NF membranes and reported relatively high retention rates ~90%. 20, 21

73

Minimizing waste (brine) streams by achieving high water recovery is crucial for successful

74

application of NF to treat PFOA contaminated groundwaters. Current NF systems are limited to

75

low recoveries (75-80%) mainly due to inorganic scaling because groundwaters contain high

76

levels of scale-forming inorganic species, including calcium, magnesium, and silica in the form

77

of silicic acid.22,

78

environmental impacts associated with brine management and, thus, enhance the feasibility and

79

sustainability of NF technology for groundwater remediation. 24, 25

23

There is a critical need to increase water recovery to reduce costs and

80

Nanofiltration membranes have a thin selective layer with pores at the nanometer scale. 26

81

While NF membranes poorly reject monovalent salt, they exhibit high retention rates for scale-

82

forming species such as calcium, sulfate, and silicic acid.23, 27 Such high rejection brings about a

83

significant concern for membrane scaling, especially when designing high water recovery NF

84

systems.28 Therefore, the development of novel NF membranes that provide high PFOA removal 3 ACS Paragon Plus Environment

Environmental Science & Technology

85

and reduced scaling potential is of paramount importance to make NF an efficient and

86

sustainable technology for groundwater remediation.

87

In this study, we demonstrate the fabrication of a loose, negatively charged NF membrane

88

with tailored rejection of perfluoroalkyl substances and scale-forming cations from feed

89

solutions simulating groundwater. The performance of the fabricated membrane for salt and

90

perfluorooctanoic acid removal was evaluated and compared to commercial NF membranes. The

91

selective polyamide layer of the fabricated NF membrane was extensively characterized to relate

92

membrane performance to the physicochemical properties of the selective layer. We further

93

demonstrated a low gypsum scaling potential of the fabricated NF membranes, highlighting the

94

potential application of NF membranes with tailored selectivity for high recovery NF systems to

95

treat PFAS-contaminated groundwaters.

96

MATERIALS AND METHODS

97

Materials and Chemicals. Polyethersulfone (PES) ultrafiltration membranes (LX-300K,

98

Synder FiltrationTM, Vacaville, CA) with a molecular weight cut-off of 300 kDa were used as a

99

substrate. The PES membranes were prewetted by immersing in deionized (DI) water for at least

100

24 h prior to use. Trimesoyl chloride (1,3,5-benzenetricarbonyl trichloride, TMC, 98%),

101

piperazine (PIP, 99%), 4,4′-Bipiperidyl dihydrochloride (BP, 97%), triethylamine (TEA, ≥99%),

102

and sodium hydroxide (NaOH) from Sigma-Aldrich (Sigma-Aldrich, St. Louis, MO) and hexane

103

from J.T. Baker (ACS reagent, Phillipsburg, NJ) were used for interfacial polymerization (IP) of

104

the polyamide (PA) selective layer.

105

The performance of the fabricated NF membranes was compared to a commercial NF

106

membrane, denoted NF270 by the manufacturer (FilmTec Corp., Minneapolis, MN). As

107

indicated by the manufacturer, the NF270 membrane comprises a semi-aromatic piperazine-

108

based, polyamide layer on top of a microporous polysulfone (PSf) support. The membranes were

109

received as flat sheet samples. They were gently washed with deionized (DI) water to remove

110

any preservatives and were stored in DI water at 4 °C.

111

Thin-Film Composite Nanofiltration Membrane Fabrication. TFC NF membranes

112

were fabricated by forming a polyamide (PA) selective layer on top of the PES support

113

membrane via interfacial polymerization (IP). For obtaining a dense PA layer structure, an 4 ACS Paragon Plus Environment

Page 4 of 29

Page 5 of 29

Environmental Science & Technology

114

aqueous solution of 1.0 wt % PIP was used while a mixture of 0.5 wt % PIP and 0.5 wt % BP

115

was employed to construct a loose, porous selective layer during the interfacial polymerization.

116

The NF membranes fabricated using 1.0 wt % PIP and a mixture of 0.5 wt % PIP and 0.5 wt %

117

BP are hereafter denoted as PIP and PIP + BP, respectively. To catalyze the reaction, 0.5 wt %

118

TEA and 0.15 wt % NaOH were added to both aqueous amine solutions. For IP reaction, the PES

119

support membrane (10 cm  15 cm) was tightly taped onto a clean glass plate with a water-proof

120

tape (Fisherbrand Colored Labeling Tape, Fisher Scientific, Inc., MA), with the skin layer facing

121

upward. Approximately 15 mL of aqueous amine solution was dispensed onto the surface of the

122

support membrane and allowed to contact for 90 s. The amine solution was then poured off, and

123

residual amine solution was removed from the membrane surface using an air knife. Next, the

124

monomer-saturated support membrane was immersed in 0.15 wt % TMC in hexane for 30 s,

125

resulting in the formation of a thin-film composite polyamide layer. The composite membranes

126

were then air dried for 120 s, rinsed thoroughly, and stored in DI at 4 °C.

127

Membrane Characterization. Zeta potential of membrane surface was evaluated by a

128

streaming potential analyzer utilizing an asymmetric clamping cell (EKA, Brookhaven

129

Instruments, Holtsville, NY) as described elsewhere.29 Measurements were performed with a

130

solution containing 1 mM KCl and 0.1 mM KHCO3. Electrolyte solution flows into the cell were

131

generated by pressure ranging from 0 to 300 mbar driven by a mechanical pump. The induced

132

streaming potential was measured using Ag/AgCl electrodes mounted at each end of the

133

clamping cell.

134

The carboxylic group density of the polyamide selective layer was measured using the silver

135

elution method.30 First, a 10-mM stock silver nitrate solution (ACS reagent, ≥99% from Sigma)

136

was prepared in 1 wt % nitric acid (trace-metal grade from Sigma) and further diluted to 40 M

137

and 1 M in DI water to obtain solutions for silver binding and rinsing, respectively. Then,

138

solution pH was adjusted to 7 and 10.5 using sodium hydroxide (0.1 M) and nitric acid (1 wt %).

139

Membranes were first wetted in DI water for at least 24 h prior to measurement after which

140

circular membrane samples (2.0 cm2) were punched and the fabric backings were physically

141

removed. The remaining polyamide/support (i.e., either PES or PSf) films were immersed twice,

142

each for 10 min, in 10 mL of the 40 M silver nitrate solution at pH 7 or 10.5 to bind silver one-

143

to-one with ionized carboxyl groups. After the binding step, isolated polyamide/support films 5 ACS Paragon Plus Environment

Environmental Science & Technology

144

were immersed four times, each time for 7 min, in 10 mL of 1  silver nitrate solution at the

145

same pH used during binding to rinse off unbound silver. After each step, film samples were

146

gently touched against Kim wipes to minimize solution carryover. Following the wash steps,

147

isolated polyamide/support films were immersed in 5 mL of 1 wt % nitric acid for 30 min to

148

protonate the carboxyl groups and elute the bound silver ions. After removing films, the silver

149

ion concentration in the solution was determined using ICP-MS.

150

Elemental mapping of the PA selective layer was performed by energy-dispersive X-ray

151

analysis (EDX) in a scanning transmission electron microscope (STEM). The NF membrane

152

samples were embedded in epoxy resin (SPI-PON 812, Structure Probe, Inc.), followed by

153

curing at 60 °C overnight. The samples were then cross-sectioned by ultramicrotome (Leica EM

154

UC7) with a diamond knife (Diatome Ultra 45). A thin specimen of ~100 nm thick cross-

155

sectional film was floated on the water trough of the diamond knife and then transferred onto

156

lacey carbon grids (Ted Pella). STEM-EDX analysis was conducted on TEM (FEI Tecnai Osiris)

157

with an accelerating voltage of 200 kV, equipped with a Super-X EDS detector.

158

The thickness of the selective layer was determined by imaging the isolated PA film using

159

atomic force microscopy (AFM).31-33 The NF membranes were cut into a small piece (~ 1  1

160

cm2) and the fabric backings were physically removed. The isolated polyamide/support films

161

were placed on a silicon wafer with two ends of the film being pressed by a glass slide to avoid

162

film folding and wrinkling during sample preparation. Then, a few drops of dimethylformamide

163

(DMF) were placed on the film to dissolve the PES or PSf support. After a few minutes, the PES

164

or PSf dissolved and DMF solution was carefully wiped off using Kim wipes without touching

165

the film. This procedure was repeated several times to achieve a complete dissolution of the

166

support and to have a transparent free polyamide film tightly fixed on the silicon wafer.

167

Prior to AFM imaging, multiple parallel dents were developed on the isolated PA films using

168

a metal precision glide needle (18 G  1 in, Becton Dickinson & Co., NJ) without damaging the

169

silicon wafer. A border region (10  10 m) between the PA film and the silicon wafer was

170

scanned by AFM using a Bruker Dimension FastScan AFM with a FastScan-B tip (5 nm tip

171

radius, Bruker, Billerica, MA) in tapping mode at a scan rate of 3 Hz. The obtained AFM images

172

were analyzed using the Section function of Nanoscope Analysis v1.5 (Bruker) to determine the

173

PA layer thickness. 6 ACS Paragon Plus Environment

Page 6 of 29

Page 7 of 29

Environmental Science & Technology

174

Estimation of NF Membrane Pore Size. The average pore size of the NF membrane

175

was determined based on a pore transport model that incorporates steric (size) exclusion and

176

hindered convection and diffusion from the rejection data of reference inert organic solutes. 34

177

Neutrally charged, low molecular weight organic molecules — erythritol, xylose, and dextrose

178

(≥99%, Sigma-Aldrich, St. Louis, MO) — were used as the reference organic tracers. 35 Prior to

179

NF experiments, the membrane was pre-compacted under 13.8 bar (200 psi) hydraulic pressure

180

with DI water as the feed for 4 h. After compaction, each organic solute was injected to the feed

181

to obtain a concentration of 40 mg/L (as total organic carbon, TOC) and subsequent experiments

182

were conducted at 4.1, 6.2, 8.3, and 10.3 bar with a crossflow velocity of 21.4 cm/s and a feed

183

temperature of 25.0  0.5 °C. The permeate flux was recorded after the system was run for 1 h at

184

each pressure. Permeate and feed samples were taken for TOC analysis (TOC V-CSH, Shimadzu

185

Corp., Japan) to determine the rejection of the reference organic solutes. Details on the pore size

186

estimation based on the pore transport model using the measured rejection of the reference

187

organic solutes are provided in the Supporting Information.

188

Salt and PFOA Rejection Tests. Single salt rejections were evaluated using 2, 10, and

189

20 mM of sodium chloride (NaCl), calcium chloride (CaCl 2), and sodium sulfate (Na2SO4)

190

solutions. Ion rejections were measured from NF experiments using a salt solution prepared by

191

mixing 5 mM NaCl, 5 mM CaCl2, and 5 mM Na2SO4 with a total ionic strength of 35 mM.

192

Retention of PFOA (Perfluorooctanoic acid, 96%, Sigma-Aldrich, St. Louis, MO) by NF

193

membranes was evaluated using feed solutions containing 1 mg/L of PFOA in DI water and in

194

the combined salt solutions. Feed solution pH was adjusted to 7.0  0.1 during NF experiments.

195

Prior to each experiment, the membranes were compacted for at least 4 h at 6.9 bar (100 psi)

196

using DI water until there was no variation in permeate flux. The feed solution temperature was

197

kept constant at 25.0  0.5 °C throughout the experiment. To compare rejection under the same

198

permeate flux conditions, the hydraulic pressures were set to approximately 6.9 bar (100 psi) for

199

the PIP + BP and PIP membranes and 3.1 bar (45 psi) for the NF270 membrane. Permeate and

200

feed samples were collected after the system was equilibrated for 1 h at each condition and

201

analyzed as described below. Rejection performance of NF membranes was determined by

202

comparing the species concentration in the feed (Cf) and permeate (Cp) samples (i.e., Robs = 1-

203

Cp/Cf). 7 ACS Paragon Plus Environment

Environmental Science & Technology

204

Concentrations of NaCl, CaCl2, and Na2SO4 in the feed and permeate were measured using a

205

calibrated conductivity meter (Oakton Instruments, Vernon Hills, IL). Ion chromatography (IC,

206

Dionex, Sunnyvale, CA) equipped with CS14 and AS14A IonPac separation columns was used

207

to quantify cation and anion concentrations, respectively. PFOA concentration was determined

208

using a high-performance liquid chromatograph (HPLC, Agilent 1290 Infinity Series) coupled

209

with a mass spectrometer (Agilent 6550A iFunnel Q-TOF MS). A sample volume of 1 µL was

210

injected into a C18 column (Eclipse Plus, 1.8 m, 4.6  50 mm) using a mobile phase gradient of

211

solvent comprising 40% methanol and 60% water with 20 mM ammonium acetate at a flow rate

212

of 0.5 mL min-1. Methanol rate was increased from 40% to 98% in 5 min, held for 1 min,

213

dropped back to 40%, followed by a 1-min stabilization. The Q-TOF MS was operated using an

214

electrospray ionization (ESI) interface in negative mode. Linearity was determined from 0.1 µg

215

L-1 to 100 g L-1 using calibration curves (R2 ≥ 0.99), with a limit of quantification (LOQ) of 10

216

ng L-1.

217

Scaling Experiments. The protocol for scaling experiments comprised the following steps.

218

First, a new membrane coupon was placed in the NF cross-flow unit and compacted under 13.8

219

bar (200 psi) hydraulic pressure with DI water as the feed for 4 h. Crossflow velocity and feed

220

solution temperature were maintained at 21.4 cm/s and 25.0  0.5 °C, respectively. After

221

compaction, the system was depressurized while allowing the feed flows to the membrane by

222

maintaining the crossflow velocity of 21.4 cm/s. Calcium chloride (1 M), calcium sulfate (1 M),

223

and sodium chloride (5 M) stock solutions were prepared in advance and filtered with a 0.45 m

224

cellulose acetate membrane filter (Corning, Tewksbury, MA). Stock solutions were added to the

225

feed reservoir to obtain a scaling solution comprising 15 mM CaCl 2, 10 mM Na2SO4, and 10 mM

226

NaCl, with a gypsum (CaSO4·2H2O) saturation index (SI) of 0.54. The feed solution pH was

227

adjusted to 7.0  0.1. The scaling solution was fully mixed and equilibrated for 1 h without

228

applying pressure (i.e., no permeate flux). After equilibrium, the hydraulic pressures were set to

229

about 7.2, 12.4, and 6.6 bar for the PIP + BP, PIP, and NF270 membranes, respectively, to

230

achieve an identical initial permeate flux of ~80 L m -2 h-1. The gypsum scaling experiments were

231

conducted for 20 h at a feed crossflow velocity of 15.0 cm/s and temperature of 25.0  0.5 °C in

232

a recycling mode (i.e., the permeate was recycled back to the feed solution such that the gypsum

233

saturation index was kept constant over time). 8 ACS Paragon Plus Environment

Page 8 of 29

Page 9 of 29

Environmental Science & Technology

234

RESULTS AND DISCUSSION

235

Polyamide Selective Layer Characteristics. We target the fabrication of NF membranes

236

having a selective layer with a loose, nanoporous structure and a negative surface charge to

237

facilitate passage of scale-forming cations and to enhance rejection of negatively charged PFOA

238

via Donnan (charge) exclusion. Interfacial polymerization is the state-of-the-art technique

239

allowing the synthesis of a thin, highly cross-linked selective polyamide layer. 36 Typical NF

240

membranes fabricated via interfacial polymerization of piperazine (PIP, MW of 86.1 g/mol) and

241

trimesoyl chloride exhibit fairly high salt rejection, especially for salts with divalent ions, such as

242

CaCl2, MgCl2, and Na2SO4.37, 38 To achieve our goal of fabricating NF membranes with reduced

243

cation rejection, we incorporated bipiperidine (BP) having a higher molecular weight (168.3

244

g/mol) than PIP during interfacial polymerization. Bipiperidine comprises two piperidine units

245

bound to each other by a carbon-carbon single bond, allowing chain extension and/or

246

crosslinking polymerization to form a loose polyamide structure as described in Figure S1. 39, 40

247

Specifically, we used a mixture of 0.5 wt % PIP and 0.5 wt % BP to obtain a loose, nanoporous

248

selective layer, while a dense polyamide layer was constructed employing 1.0 wt % PIP; these

249

membranes are denoted as PIP + BP and PIP, respectively.

250

To verify the structural properties of the selective layer, we evaluated the average pore size

251

of the PIP + BP membrane and compared it to the PIP and commercial NF270 membranes. The

252

real (intrinsic) retention of inert organic tracers (Rr) was obtained from the observed retention

253

(Ro) by accounting for the effect of concentration polarization using eq S3. Real retentions of the

254

organic tracers by the PIP + BP, PIP, and NF270 membranes at different permeate water fluxes

255

(i.e., transmembrane pressures) are presented in Figure S2.

256

The obtained real retentions were used to estimate the NF membrane average pore size using

257

the membrane pore transport model described earlier. The estimated pore radii are consistent for

258

the different organic reference solutes as summarized in Table S1 of Supporting Information.

259

Based on these results, we conclude that the PIP + BP membrane has an average pore radius of

260

0.61 nm, which is much larger than the radii of the PIP (0.47 nm) and NF270 (0.41 nm)

261

membranes. The larger pore size of the PIP + BP membrane compared to that of the PIP and

262

NF270 membranes demonstrates that incorporating BP during the interfacial polymerization

263

forms a polyamide network with a relatively loose, nanoporous structure. 9 ACS Paragon Plus Environment

Environmental Science & Technology

264

Rejection of charged solutes by electrostatic (Donnan) exclusion is directly related to the

265

charge characteristics of NF membranes.41 We investigated the charge properties of the PIP + BP,

266

PIP, and NF270 membranes by determining the zeta potential and carboxyl group density of the

267

polyamide selective layer. As shown in Figure 1a, all NF membranes display a negative surface

268

charge at solution pH of 7 and 10; the magnitude of the negative zeta potentials follows the order

269

of NF270 > PIP > PIP + BP membranes. The polyamide selective layer of desalination

270

membranes fabricated via interfacial polymerization of trimesoyl chloride (TMC) and amine-

271

based monomers (i.e., BP or PIP) inherently possesses an outer layer of negative fixed charges

272

resulting from hydrolysis of unreacted acyl chlorides of TMC to carboxyl groups. 42 The carboxyl

273

groups of the polyamide film have a dissociation constant (pKa value) of ~5.2.43, 44 The observed

274

negative zeta potential of the PIP + BP, PIP, and NF270 membranes at the investigated pH range

275

is attributed to the ionized surface carboxyl groups.

276

FIGURE 1

277

Streaming potential analysis allows qualitative evaluation of the membrane surface charge

278

characteristics. However, this method does not quantify the fixed charge density but rather

279

determines the zeta (electrokinetic) potential, which is determined by the combined effects of

280

charged groups on the surface as well as ions near the membrane surface. 45 In addition, zeta

281

potential does not account for functional groups buried within the polyamide polymer network or

282

inside the pores,30,

283

mechanisms of NF membranes.

45

although such information is critical to understand the separation

284

To better quantify NF membrane charge characteristics, we measured the carboxylic group

285

density of the PA layer using the recently developed silver elution method. 30 This method allows

286

for the quantification of areal carboxyl density based on the assumption that small, cationic silver

287

ions are bound one-to-one with ionized carboxyl groups that are present on the membrane

288

surface as well as those buried within the polyamide network. 44 As presented in Figure 1b, the

289

areal carboxyl group densities of the PIP + BP, PIP, and NF270 membranes measured at pH 7

290

were in the range of 0.7 – 1.7 sites/nm2, while much higher values were obtained at pH 10.5

291

ranging from 3.8 – 13.0 sites/nm2. These values are relatively small compared to those reported

292

for commercial reverse osmosis (RO) membranes,30 likely due to the thinner polyamide film of

293

NF membranes,46,

47

which has lower degree of cross-linking formed by the semi-aromatic 10 ACS Paragon Plus Environment

Page 10 of 29

Page 11 of 29

Environmental Science & Technology

294

piperazine-based monomer compared to that formed on RO membranes by the fully aromatic m-

295

phenylenediamine monomer.

296

The carboxylic group density measured at pH 7 follows the order of NF270 > PIP > PIP +

297

BP membranes, which correlates well with the results obtained from zeta potential measurements

298

(Figures 1a-b). At a solution pH 10.5, the measured carboxylic group densities were much higher

299

than those measured at pH 7 for all membranes. Additionally, the PIP membrane was found to

300

have much higher carboxylic group density than the PIP + BP and NF270 membranes at a

301

solution pH of 10.5. Such a substantial difference in carboxyl group densities at low (pH 7) and

302

high (pH 10.5) solution pH is consistent with our previous observation of a series of commercial

303

RO membranes.30 We attribute the results to the likely increase in acidity within the polyamide

304

film during the silver binding process. Considering the relatively high carboxylic group

305

concentration within the polyamide film, which is estimated to be 0.2–0.7 M, 44, 48 a substantial

306

decrease in local pH within the polyamide film is postulated; that is, the local pH within the

307

polyamide film would be much lower compared to the bulk solution pH employed during the

308

silver binding procedure. Thus, when a bulk solution pH of 10.5 was employed, the majority of

309

carboxylic groups within the polyamide film remain deprotonated/ionized (COO -), while only a

310

fraction of carboxyl groups exist in ionized form within the film at bulk solution pH of 7.0. It can

311

be concluded that the areal density measured at pH 7.0 corresponds to the carboxyl groups on the

312

membrane surface, while that measured at pH 10.5 corresponds to groups buried within the

313

polyamide network.30 The remarkably high density of carboxylic groups of the PIP membrane

314

observed at pH 10.5 (Figure 1b) based on this proposed mechanism is further discussed in the

315

following subsections.

316

We conducted STEM-EDX elemental mapping to depict the cross-section images of the PA

317

films. The polyamide selective layer is rich in nitrogen while the underlying PSf or PES support

318

layer contains sulfur but no nitrogen, providing elemental contrast in STEM-EDX analysis

319

(Figures 2a-1 to a-3). Although the STEM-EDX elemental mapping was limited to characterize

320

the distinct shape of the protruding polyamide layer nodules,49 it allowed visualization of the

321

relative polyamide layer thickness for the PIP + BP, PIP, and NF270 membranes. The PA layer

322

thickness of the NF270 membrane is likely in the range of 20–30 nm (Figure 2a-3), similar to

323

values reported from other high-resolution microscopy and surface characterization studies. 31, 50 11 ACS Paragon Plus Environment

Environmental Science & Technology

324

A relatively high and thicker signal of nitrogen was detected for the PIP membrane (Figure 2a-2),

325

indicating that the PIP membrane may have a thicker PA selective layer than the other two

326

membranes.

327

FIGURE 2

328

To achieve a more quantitative analysis of the selective layer thickness, the PA film was

329

imaged by atomic force microscopy (AFM). The isolated PA film on the silicon wafer was

330

scratched using a precision glide needle (Figure S4). A border region between the PA film and

331

the silicon wafer was scanned using AFM to determine the PA layer thickness as shown in

332

Figures 2b-1 to b-3. The average PA layer thicknesses of the PIP + BP, PIP, and NF270

333

membranes were 43  2, 61  7, and 29  4 nm, respectively. Consistent with the previous

334

STEM-EDX elemental mapping analysis, the PIP membrane had a thicker PA layer than the PIP

335

+ BP and NF270 membranes. In particular, the thickness of the PIP membrane PA layer was

336

more than two times higher than that of the NF270 membrane. The measured polyamide layer

337

thickness explains the previously observed higher carboxylic group density of the PIP membrane

338

at pH 10.5 compared to that of the other two membranes (Figure 1b). The thicker PIP membrane

339

PA layer contains more carboxylic groups buried within the polyamide network or inside the

340

pores.

341

The investigated structural and charge properties of the PIP + BP, PIP, and NF270

342

membranes are schematically visualized in Figures 2c-1 to c-3. As we intended, the PIP + BP

343

membrane had a pore diameter of ~1.2 nm, much larger than the pore diameters of the PIP

344

(~0.94 nm) and NF270 (~0.80 nm) membranes. All membranes possess negative surface and

345

pore charge, which is attributed to the carboxylic groups of the polyamide film. The thickness

346

normalized carboxyl group density, indicated as “COO-” in Figures 2c-1 to c-3, was calculated

347

by dividing the total carboxylic group density at pH 10.5 (shown in Figure 1b) by the average PA

348

layer thickness measured using AFM (shown in Figure 2b). It is assumed that the carboxyl areal

349

density shown in Figure 1b is based on a dimensionless unit square value. The thickness

350

normalized carboxylic group density is the highest for the PIP membrane (0.21 mn -1), followed

351

by the NF270 (0.19 nm-1) and PIP + BP (0.11 nm-1) membranes.

12 ACS Paragon Plus Environment

Page 12 of 29

Page 13 of 29

Environmental Science & Technology

352

In the following subsections, we relate the polyamide selective layer characteristics discussed

353

above to the salt and PFOA rejection behavior of the NF membranes. We further depict the

354

underlying rejection and scaling mechanisms of the fabricated membranes.

355

Single Salt Rejection Performance. Salt rejection behavior of the PIP + BP, PIP, and

356

NF270 membranes was evaluated with sodium chloride, calcium chloride, and sodium sulfate at

357

concentrations of 2, 10, and 20 mM (Figure 3). We employed hydraulic pressure of 6.9 bar (100

358

psi) for the PIP + BP and PIP membranes and 3.1 bar (45 psi) for the NF270 membrane during

359

the NF experiments to compare the rejection performance of all membranes that had different

360

permeabilities (Figure S3), at the same permeate water flux. As shown in Figure 3a, the rejection

361

of NaCl decreased with increasing salt concentration for all three membranes, which is typically

362

observed for negatively charged NF membranes.51, 52 For symmetric salts like NaCl, electrostatic

363

repulsion between co-ions (ions with the same charge as the membrane) and the membrane

364

governs the NF rejection mechanism.18 For this reason, NF membranes exhibit higher NaCl

365

rejection with increasing solution pH because more carboxylic groups on the PA selective layer

366

are deprotonated (COO-) at higher pH, thereby enhancing Cl- rejection by charge repulsion.53

367

Similarly, at higher salt concentration, screening of the charge on the PA selective layer reduces

368

the effectiveness of electrostatic repulsion of Cl - ions, resulting in a decrease of NaCl rejection.

369

We also observe that the PIP + BP membrane showed much lower NaCl rejections of ~50% and

370

~25% at low (2 mM) and high (20 mM) concentrations, respectively, than the PIP and NF270

371

membranes.

372

FIGURE 3

373

It is interesting to note that NaCl rejection of the PIP membrane was higher than that of the

374

NF270 membrane at all salt concentrations tested, despite the PIP membrane having a larger

375

pore size (Table S1) and less negative zeta potential than the NF270 membrane (Figure 1a). The

376

observed higher NaCl rejection of the PIP membrane can be explained by the thickness of the

377

selective polyamide layer. As described in Figures 2c-2 and c-3, the PIP membrane has a

378

polyamide layer much thicker (61 nm) than the NF270 membrane (29 nm), which induces higher

379

resistance to ion transport through the selective layer. This result highlights the fact that in

380

addition to membrane pore size and charge properties, the PA selective layer thickness plays a

13 ACS Paragon Plus Environment

Environmental Science & Technology

381

critical role in governing ion rejection/selectivity as well as water permeability (Figure S3) of NF

382

membranes.

383

Observed trends of CaCl2 rejection as a function of salt concentration were opposite to those

384

observed with NaCl, where the rejection of CaCl 2 increased at higher concentrations for all NF

385

membranes (Figure 3b). Charge-based separation of asymmetric salts, such as CaCl 2 and Na2SO4,

386

is dominated by the ions of higher valency (Ca2+ for CaCl2 and SO42- for Na2SO4) because the

387

electrostatic repulsive or attractive interactions between such ions and the membrane control the

388

rate of salt transport.18, 53 Hence, CaCl2 rejection by the negatively charged PIP + BP, PIP, and

389

NF270 membranes is mainly controlled by the rate of Ca 2+ permeation through the PA selective

390

layer. At a relatively low CaCl2 concentration (2 mM), electrostatic attraction between Ca2+ and

391

the negative membrane charge facilitates the transport of Ca 2+, resulting in a low CaCl2 rejection.

392

However, cation transport through the negatively charged membrane is reduced at higher salt

393

concentrations (20 mM in Figure 3b) due to effective charge neutralization and screening of the

394

membrane carboxyl groups, resulting in reduced electrostatic attraction and thus higher rejection

395

of CaCl2. When the effect of charge interaction diminishes at high ionic strengths, size exclusion

396

has a stronger effect on Ca2+ retention. Considering the size of the hydrated Ca2+ ion (radius of

397

~0.41 nm)54 and the estimated pore radii of the PIP (~0.47 nm) and NF270 (~0.41 nm)

398

membranes (Table S1), size exclusion of Ca2+ can be effective at high salt concentrations. In

399

contrast, the PIP + BP membrane showed a much lower CaCl 2 rejection (