The antifouling performance of an ultrafiltration membrane with

pre-deposited carbon nanofiber layers for water treatment

Ting Liu1,2, Yuanlong Lian1, Nigel Graham2, Wenzheng Yu3,2*, Kening Sun1*

1 School of Chemistry and Chemical Engineering, Beijing Institute of Technology,

Beijing 100081, China (liuting@bit.edu.cn, sunkening@bit.edu.cn)

2 Department of Civil and Environmental Engineering, Imperial College London,

South Kensington Campus, London SW7 2AZ, UK (n.graham@imperial.ac.uk)

3 Key Laboratory of Drinking Water Science and Technology, Research Center for

Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China

(w.yu@imperial.ac.uk, wzyu@rcees.ac.cn)

1

Abstract: In order to improve the performance of the ultrafitration (UF) membrane

process in drinking water treatment, in terms of permeate flux and natural organic

matter (NOM) removal, a new form of carbon nanofiber (CNF) layer derived from

bacterial cellulose (BC) was prepared and applied as a pre-deposited coating on the

UF membrane surface. Using bench-scale, dead-end filtration tests, both CNF and

CNF modified by ethanol treatment (M-CNF), were evaluated for the treatment of

two model NOM solutions, namely bovine serum albumin (BSA) and sodium alginate

(SA). The results showed that both types of coating were effective in mitigating

membrane fouling (lower flux decline), with the mitigation increasing with the

coating quantity, and also enhanced the removal of BSA and SA. In particular, the M-

2

1

2

3

4

5

6

7

8

9

10

CNF layer at the greater loading (24 g/m2) was able to reduce membrane fouling to a

very substantial degree and achieve >90% removal of BSA and SA. Characterization

of the CNF and M-CNF layers showed significant differences in their morphological

and structural properties which may explain the observed differences in their ability to

reduce membrane fouling; protection of the UF membrane by the carbon nanofiber

layers may be attributed to both physical separation and surface adsorption of the

NOM biopolymers.

Keywords: water treatment; ultrafiltration; carbon nanofiber; membrane fouling;

antifouling layer

1. Introduction

In the past decade, ultrafiltration (UF) membrane technology has been increasingly

applied in water treatment plants for drinking water supply in China and many other

countries. Although various approaches to enhancing the performance of UF have

received attention, fouling of the membrane still remains the principal operational

limitation, affecting the process reliability and cost-effectiveness. Among the

approaches being considered is the development of new, more efficient membrane

materials, but progress so far has been slow [1]. Among the common materials used

for water treatment membranes have been polyether sulfone (PES) [2, 3],

polyvinylidene fluoride (PVDF) [4, 5], cellulose acetate (CA) [6], cross-linked

polyamide (PA) [7], polycarbonate [8], and others. In many cases, pre-treatment

processes are needed to remove the organic matter in the influent/raw water prior to 3

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

the membrane to mitigate the membrane fouling.

As an alternative to conventional coagulation, floc separation and sand filtration,

the use of powdered activated carbon (PAC) coupled with ultrafiltration (PAC-UF) is

an emerging technology for the removal of NOM (natural organic matter) for drinking

water [9-11], particularly for proteinaceous substances, and some fractions of humic-

type substances [12]. As a sole pretreatment method, various studies have shown that

PAC can remove some of the organic matter responsible for membrane fouling [13,

14]. Yu and co-workers also found that PAC adsorption together with coagulation

could decrease the amount of dissolved organic matter (DOM) reaching the

membrane surface and the extent of internal membrane fouling [12]. However, some

studies have indicated that the addition of PAC may induce a greater degree of

membrane fouling in long-term operation [12]. Similarly, research conducted using

other types of adsorbent particles added to UF systems to remove NOM and other

contaminants from the raw water [15], showed contrary effects on membrane fouling;

in some cases fouling was reduced, while in others it was exacerbated [16]. Also, a

layer of either cellulose or (especially) polysulfone nanofibers on the surface of

ultrafiltration membranes could improve the fouling resistance [17]. In other studies,

particular nano-absorption materials such as heated iron or aluminum oxide particles

[13, 14, 16, 18], were pre-deposited on membrane surfaces in order to achieve NOM

removal on and within the particle layer, and thereby reduced the quantity of foulants

reaching the membrane surface [15, 19, 20].

Materials with high specific surface area and three-dimensional (3D) structures

4

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

have great potential to be used as a pre-adsorptive layer on a membrane surface to

protect the membrane from rapid fouling. Bacterial cellulose (BC), a common

biomass nano-material, is composed of interconnected 3D networks of natural

cellulose nanofibers with a high Young’s modulus and a large number of hydrogen

bonds. Due to BC’s high surface area-to-volume ratio, remarkable mechanical

properties, sustainable source, and low cost, BC-derived materials are promising

alternatives to be used as a membrane coating in the UF membrane system. However,

to-date there has been little reported in the way of related research in this field of

application. BC-derived carbon nanofibers (CNF), which are made from BC after

freezing and pyrolyzing, could be utilized as a pre-adsorption coating material on a

UF membrane for water treatment, owing to its highly favourable physical and

chemical properties. In addition, the CNF material is capable of being regenerated and

reused by reheating to remove adsorbed organic matter, which is potentially another

important advantage for its use in practice as an adsorbent UF coating.

In this study we have prepared and tested a BC-derived CNF layer that was pre-

deposited on a UF membrane surface to investigate its contribution to NOM removal

and fouling control. In addition, a modified CNF (M-CNF), using ethanol in the

preparation to alter the surface chemistry of the CNF, was also evaluated in terms of

NOM removal and fouling mitigation. The tests involved the use of model NOM

solutions containing known foulants and compared the flux performance of the UF

systems with CNF and M-CNF as pre-deposited layers on the membrane surface and

changes in the nature (micromorphology) of the layers. The results described

5

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

subsequently showed that the CNF and M-CNF layers provided an enhanced

antifouling performance compared to uncoated UF with the model NOM foulants.

2. Materials and reagents

2.1. Model NOM solutions

All chemicals were analytical regents except as indicated subsequently. Sodium

alginate (SA) and bovine serum albumin (BSA) (Sinopharm Chemical Reagent Co.,

Ltd, China) were obtained as reagent grade chemicals. Both chemicals have been used

in many studies previously as models for polysaccharides and proteins, respectively,

found in surface waters serving as sources for drinking water supply (e.g. Yu et al.,

2017). BSA used here was described as shallow grass yellow ball or needle crystal,

and its MW was approximately 68k Dalton (data provided by supplier, Sigma). The

molecular formula of SA is (C6H7NaO6)N, and its MW was approximately 200k

Dalton (provided by Sigma). Fresh solutions of both chemicals were prepared at a

total concentration of 3 g/L using deionized (DI) water (Millipore Milli-Q). Each

stock solution was stored in the dark at 4 oC and brought to room temperature shortly

before the preparation of working solutions and use in tests, and the stock solution

was used within 3 days. Final solutions of 10 mg/L SA or BSA were prepared with 5

mM NaHCO3 and the pH was adjusted to 7.0 using either 0.1 M NaOH or 0.1 M HCl,

before applying the solution to the membrane system. Throughout the experimental

period the temperature was constant at 25±0.5 oC.

6

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

2.2. Preparation of CNF/M-CNF

The BC hydrogel (Hainan Yeguo Foods Co., Ltd., China) was immersed in DI

water for 48 h, and then rinsed with DI water 2-3 times before use. The BC hydrogel

was cut into patches of about 1.0×1.5×2.5 cm3 in size. After freezing using liquid

nitrogen, the frozen hydrogel BC was freeze-dried (Beijing boyikang Laboratory

Instrument Co. Ltd., China) and transformed into BC aerogel. The resulting BC

aerogel was subsequently pyrolyzed under a flowing N2 atmosphere at 900 oC for 2 h

to produce a black carbon nanofiber (CNF) framework/material. Finally, the CNF

material was ground into very small pieces (grinding machine MG200, Grinder,

China) for membrane pre-coating (sieved to around 200 mesh), corresponding to a

size smaller than 74 μm. For comparison, a modified CNF (M-CNF) material was

prepared, involving the CNF being immersed in ethanol for 12 h and then air-dried for

12 h.

2.3. Preparation of CNF/M-CNF pre-coated layers and UF experiments

The UF membrane used in the experiments was made of polyvinylidene fluoride

(PVDF) (Beijing Ande Membrane Separation Technology & Engineering Co., Ltd,

China) with a nominal molecular weight (MW) cutoff of 100 kDa (~ 10-20 nm). Its

other properties included a working temperature range of -10~45 oC, a pH range of

3~11, a working pressure of 0~4 bar, and its diameter for the tests was 79 cm. Prior to

use, each membrane was immersed in DI water for at least 24 h and subjected to

ultrasonic treatment for 5 min to remove any impurities and production deposits on

7

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

the membrane surfaces. Then 0.05 g (12 g/m2) and 0.1 g (24 g/m2) CNF and M-CNF

were dispersed in the DI water, and deposited onto the UF membrane surface by

filtration, thereby forming a pre-coated layer. Before the UF tests with NOM

solutions, the permeate flux of the membrane was determined by filtering DI water

until the flux reached a constant value; details of the method are given below.

Each of the UF tests involved three filtration cycles, with backwashing and re-

coating between each cycle. Thus, the operational procedures were as follows: 1) 300

mL model solution (SA or BSA) was filtered by a UF membrane which had been pre-

coated with CNF or M-CNF; 2) at the end of the filtration period the fouled CNF or

M-CNF layer was scraped from the membrane, and then the membrane was reversed

and backwashed with 50 mL DI water; 3) after the backwash the membrane was

reversed again (back to its original arrangement) and a new layer of CNF or M-CNF

added to the membrane surface (at the given concentration) before commencing the

next filtration cycle. After the completion of the series of filtration cycles, the CNF or

M-CNF layer on the membrane surface was scraped off with an obtuse plastic sheet

for subsequent characterization.

The variation (decline) of normalized flux J/J0 with filtration time was used to

describe the process of UF membrane fouling, where J0 is the initial membrane flux.

The UF filtration tests were conducted using a stirred cell apparatus (Millipore,

Amicon 8400, USA) at a constant filtration pressure (0.1 MPa) applied by nitrogen

gas, and the measurement of permeate volume was determined by an electronic

balance (ML4002, Mettler-Toledo International Inc., USA) connected to a computer

8

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

for continuous data logging at one second intervals. The fouling and treatment

performance of the membranes with or without pre-deposited CNF/M-CNF layers

were studied by filtering 10 mg/L SA or BSA solutions, in a similar manner to that

reported previously [21-24].

2.4. Characterization of CNF, M-CNF and un-coated, UF membranes at the

completion of filtration cycles

At the completion of the membrane cycles, the fouled CNF and M-CNF layers on

the membrane surface were carefully scraped off with a plastic sheet for subsequent

analysis. The methods used in the analysis of these, and other, samples are described

as follows. Samples of BC aerogel and the CNF/M-CNF layers were coated with Au

using a Precision Etching Coating System (Quorum, Model 150R ES) and imaged

with a FEG250 field emission scanning electron microscope (SEM, FEI, America)

equipped with a thermal field emission gun at 25kV accelerating voltage.

Thermogravimetric differential thermal analysis (TG-DTA) (TG/DTA 6200, SII Nano

Technology Inc., Japan) was conducted in a corundum crucible from 50 oC to 600 oC

with a heating rate of 20 oC/min under a nitrogen flow rate of 200 mL/min to evaluate

the weight loss of the freeze-dried BSA and SA in samples of the membrane influent

and effluent. The specific surface area and porosity of the CNF and M-CNF layers

before and after UF tests were determined by the Brunauer Emmett Teller (BET)

method, involving nitrogen adsorption-desorption isotherms measured at 77 K on a

TriStar II3020 Micrometrics apparatus (USA). Fourier transform infrared

9

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

spectroscopy (FTIR, Thermo Scientific, Nicolet iS10) and X-Ray diffraction (XRD,

Rigaku, Ultima IV, Cu-Κα radiation, 45 kV, 50 mA, Japan) were used to characterize

the components and crystallinity of BC, CNF and M-CNF samples. Raman spectra

were obtained using a Micro-Raman spectrometer (Renishaw, UK) with a 532 nm

wavelength incident laser. Total organic carbon (TOC) of 0.45 μm filtered solutions

was determined by a total organic carbon analyzer (TOC-VCPH, Shimadzu, Japan).

An ultraviolet-visible light spectrophotometer (UV-2450, Shimadzu, Japan) was used

to determine the content of BSA in the membrane influent and effluent solutions.

3. Results and discussions

3.1. SEM images

BC aerogel, CNF and M-CNF samples were examined by SEM (Figure 1) to

investigate the mechanism of fouling reduction caused by the pre-deposited CNF and

M-CNF layers. The size distribution and average size of the fiber diameter are of

importance in understanding the fouling control process since they have a significant

impact on the properties of the pre-deposition layers on the membrane surface, and for

a given mass loading they determine the porosity and thickness of the layers. The

SEM images showed that the average diameter of fibers of the BC aerogel was

relatively larger than that of the CNF and M-CNF fibers, which was attributed to the

effects of carbonization and dehydration. The distribution of fiber diameters was

obtained by a statistical method using the ‘Smile View’ software [25], which indicated

that the fiber diameter of BC aerogel was mainly distributed in the range of 20-40 nm. 10

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

The average diameter of CNF fibers decreased from around 20 nm (23 nm) without

modification by ethanol to approximately 10 nm (13.5 nm) for the M-CNF fibers.

Based on these values, and the measurements of surface area and pore volume (Table

S1), described later, the approximate thickness of the layers were calculated to be:

11.5 µm (12 g/m2) and 23 µm (24 g/m2) for the CNF, and 9.7 µm (12 g/m2) and 19.4

µm (24 g/m2) for the M-CNF. Thus, the M-CNF layers were thinner and more

compact than the CNF layers. This was supported by cross-sectional images of CNF

and M-CNF layers (at 6 g/m2 loading), shown in Figure 1 (d - g), which indicated the

thicknesses of the CNF and M-CNF layers as 5.25 μm and 4.95 μm, respectively.

Therefore, the deposited layers have the capability to mitigate membrane fouling as

reported elsewhere with other layer types [17, 26], and shown here in the following

results and discussion.

11

187

188

189

190

191

192

193

194

195

196

197

198

12

Figure 1. SEM images of (A) BC aerogel, (B) CNF and (C) M-CNF; and inserts are

statistical results of size distribution of the (A) BC nanofibers, (B) CNF and (C) M-

CNF (by Smile View software; other conditions: voltage = 25 kV; magnification =

50,000×), and SEM cross-section images of CNF (D and E) and M-CNF (F and G) 13

199

200

201

202

layers (with 6 g/m2 loading).

3.2. Structure of CNF and M-CNF layers

In order to further explain the superior performance of the UF membrane with the

M-CNF layer, compared to that with the CNF layer, in terms of membrane fouling,

the nature of the M-CNF and CNF layers were investigated by means of XRD, FTIR

and TG-DTA analyses. Figure 2A displays the XRD patterns of samples of the BC

aerogel, and CNF and M-CNF layers. In the case of the BC aerogel, three broad peaks

located at 14.7°, 17.0°, and 23.0° were observed, which was due to the partial

crystallization of pristine BC [27]. In marked contrast, there was only one wide XRD

peak for the CNF and M-CNF samples, indicating a significant decrease in

crystallinity and that the composition had undergone an apparent change (Figure 2).

The reasons may be as follows: it has been reported that the heat treatment of

cellulose to over 260 oC induces an irreversible crystal transformation from cellulose

Iα to cellulose Iβ accompanying the shift of the peak 110 to the wider angle on the

XRD patterns [28]. In addition, the Raman spectra of CNF and M-CNF samples are

shown in Figure 2B. In these spectra, the D-band at 1330 cm-1 was induced by the

defective structures of the carbon material, while the G-band at 1590 cm−1 was related

to E2g graphite mode [29]. Hence, the intensity ratio of the D band to G band (ID/IG)

reflects the defective extent and the structure quality of CNF semi-quantitatively. The

greater intensity ratio of the sample suggests the poorer graphite structure of the

carbon material. It can be seen that the intensity ratio (ID/IG) of the D band and G band

14

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

of M-CNF is about 1.11, while the ID/IG of CNF is 1.15. It can be explained that the

presence of unrepaired defects that remained after the removal of large amounts of

oxygen-containing functional groups and its modification by ethanol, resulted in the

variation of ID/IG intensity ratios of the Raman spectra [29, 30]. This was further

confirmed by the FTIR analysis of CNF and M-CNF, as described in the following

section.

Figure 2. (A) XRD patterns of BC aerogel, CNF and M-CNF samples; (B) Raman

spectra of CNF and M-CNF samples.

3.3. Performance of the membrane processes

The effect of pre-deposited CNF and M-CNF layers with different thicknesses

(loadings) on the development of membrane fouling, as indicated by flux decline, was

investigated using 10 mg/L BSA and SA solutions as model substrates for NOM-rich

influent water. In the filtration experiments, the initial fluxes (J0) of the membrane did

not change significantly with the different thicknesses of pre-deposited layers and

were equal to 205±8 L/m2h. The flux-decline curves for the membranes with CNF and

15

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

M-CNF layers obtained in the filtration tests are shown in Figure 3, with those for a

pristine UF membrane (only PVDF).

For both the SA and BSA solutions, the presence of the CNF layer substantially

reduced the extent of membrane fouling (Figures 3A and 3B). Thus, for the SA

solution, the final J/J0 of the UF with 12 g/m2 CNF increased, relative to the uncoated

membrane, from 0.30 to 0.46 after three filtration-backwash cycles, and increased to

0.48 with 24 g/m2 CNF (Figure 3A). However, larger J/J0 values were obtained when

treating BSA solution, as shown in Figure 3B, where the J/J0 increased from 0.66

(only PVDF) to 0.72 (12 g/m2 CNF) and 0.74 (24 g/m2 CNF) after the first cycle and

from 0.59 (only PVDF) to 0.65 (12 g/m2 CNF) and 0.69 (24 g/m2 CNF) after the

second cycle. Even after the third filtration cycle the final J/J0 was significantly

greater at 0.67 for 24 g/m2 CNF compared to 0.57 for the uncoated (only PVDF)

membrane. It was evident from these results that the SA solution induced a higher

normalized flux decline than that for the BSA solution, which was the case without, or

with, the CNF pre-coated layer. This difference in fouling behavior may be a

consequence of the size of the organic substrates (viz: ~20 nm SA, ~7.5 nm BSA; Yu

et al., 2017) relative to the pore sizes of the PVDF membrane and coatings; this will

be discussed subsequently.

The results in Figure 3A also show that for the SA solution the J/J0 recovery for

the 12 g/m2 and 24 g/m2 CNF pre-coated layers after backwash was 96% and 93%,

respectively, which were only slightly less than that of a new, uncoated membrane.

For the second recovery after backwash, the results were similar. Therefore, it can be

16

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

concluded that the membrane fouling was mainly determined by the cake layer on the

surface of the membrane and the CNF pre-coated layer can mitigate the flux decline

for the SA solution. In contrast, as shown in Figure 3B, relatively serious irreversible

fouling occurred with the BSA solution in the absence of the CNF layer. It was also

evident that with a greater CNF loading, the extent of irreversible membrane fouling

was reduced.

The performance of the CNF layer as a pre-filtration/adsorption layer for the SA

and BSA solutions, and the impact on membrane fouling, were compared with that for

the M-CNF layer (Figures 3C and 3D). The results showed that the flux decline with

the M-CNF pre-coated layer (12 g/m2 and 24 g/m2) was remarkably less than that of

the CNF pre-coated layer (12 g/m2 and 24 g/m2) for the treatment of both SA and BSA

solutions. In both cases the flux decline with the 24 g/m2 M-CNF layer was less than

that with the 12 g/m2 M-CNF layer, especially for the SA solution, and the greater

loading (24 g/m2 M-CNF) was able to mitigate the membrane fouling to a low level.

The differences in the characteristics of the layers (e.g. smaller average diameter of

fibers, and narrower pores and channels between the fibers, of the M-CNF layer) may

explain the different fouling behavior and NOM removal performance shown in

Figure 3. Also the mechanism of membrane fouling will be discussed subsequently

based on the changes of structure and chemical bonds within the CNF and M-CNF

layers during filtration.

17

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

Figure 3. Normalized flux decline of an uncoated, and CNF coated, PVDF membrane

in the treatment of 10 mg/L SA (A) and 10 mg/L BSA (B) solutions for three

filtration-backwash cycles (‘down’ arrows indicate backwash events), and the

normalized flux decline of uncoated and CNF/M-CNF coated membranes in the

treatment of 10 mg/L SA (C) and 10 mg/L BSA (D) (first filtration cycle); removal of

18

283

284

285

286

287

288

SA (E) and BSA (F) by the CNF/M-CNF coated membranes in terms of TOC (first

filtration cycle); the initial membrane flux for both BSA and SA solutions was

approximately 205±8 L/m2h.

The results of the comparative treatment performance for the UF, CNF, M-CNF

membranes in terms of the removal of SA and BSA are shown in Figure 3E and 3F.

These indicate that there was an enhanced, but significantly different, effectiveness of

organic matter removal by the presence of a CNF or M-CNF coating on the

membrane surface when treating the different NOM solutions. The results indicated

that the uncoated UF membrane could dramatically remove SA by more than 90%

(Figure 3E), reflecting a little greater size of the SA macromolecules relative to the

UF pores, while the additional coating of a CNF or M-CNF layer further improved the

removal efficiency, but to a minor extent. For the BSA solution, the coating of

CNF/M-CNF was able to substantially improve the removal efficiency from around

55% to more than 80%, and for the greater loading (thickness) of the coated layer (24

g/m2) the BSA removal increased to more than 90% (Figure 3F); for both SA and BSA

the removal efficiency of the M-CNF was consistently greater than that of the CNF.

SEM images of the CNF and M-CNF layers on the UF membrane after NOM

filtration are given in Figure 4, and show the comparative extent of blockage by the

BSA and SA. It can be seen that few large pores were evident on the M-CNF layer

and the apparent number of pores was smaller (Figure 4C) compared to the CNF layer

(Figure 4A) from the treatment of BSA solution. In addition to the blockage of pores,

19

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

the adsorption of organic matter within the M-CNF layer was greater than that of the

CNF layer, as indicated by the accumulation of BSA. The comparative extent of

blockage of the CNF and M-CNF layers by SA was very similar, and more pores of

the M-CNF layer appeared to have been blocked (Figure 4D). These observations are

consistent with more NOM penetration to the surface of the PVDF membrane, and the

greater flux decline, for the UF membrane with CNF layer, as reported previously

(Figure 3).

Figure 4. SEM images of 24 g/m2 pre-coated CNF (A) and M-CNF (C) layers,

20

2 um2 um

(D)(C)

2 um2 um

(B)(A)

310

311

312

313

314

315

316

317

318

respectively, fouled by BSA; and SEM images of 24 g/m2 pre-coated CNF (B) and M-

CNF (D) layers, respectively, fouled by SA (other conditions: voltage = 25kv;

magnification = 50,000×).

3.4. FTIR analysis

The FTIR spectra of the pristine and fouled CNF and M-CNF membrane layers, as

well as the BSA and SA, were obtained in order to investigate structural differences in

the layers and confirm the adsorption of BSA and SA by the CNF/M-CNF layers;

these are shown in Figure 5. The infrared spectra of BSA (protein) exhibited five

characteristic bands in the range of 1000~1700 cm-1, three of which are known as

amide bands [31]. The Amide I band, which has the strongest absorption of infrared

light, is found between 1600 and 1700 cm-1. It is primarily caused by stretching

vibrations of C＝O, coupled weakly with C-N stretch and N-H bending [31]. The

Amide II band is mainly caused by the C-N stretch along with N-H in-plane bending

and occurs at 1500-1600 cm-1 [32]. This band of the BSA in our study shows at 1553

cm-1. The Amide II band of the CNF layer after adsorbing BSA shifted from 1591 cm-1

to 1595 cm-1. The Amide III band is found at 1200-1300 cm-1, and this band of the

BSA showed here at 1245 cm-1. However, when BSA was adsorbed on CNF, its FTIR

spectra became less obvious (Figure 5A). A similar phenomenon can also be observed

at 2964 cm-1, which corresponded to CH2 bonds.

In the case of CNF with SA (Figure 5B), the FTIR spectra had a dominating signal

21

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

at 3200-3500 cm-1 from the presence of hydroxyl groups in CNF and hydrogen bonds

in both CNF and SA [33, 34]. The absorption bands at 1596 cm-1 and 1419 cm-1,

corresponded to asymmetric COO stretching vibration and symmetric COO stretching

vibration of SA, respectively [35]. Because of the superposition of bands 1061 cm-1

(CNF) and 1127 cm-1 (SA), absorption bands for CNF after treating SA solution were

visible between them, which confirmed the adsorption of SA by the CNF layer.

Figure 5. FTIR spectra of coated UF membranes: 24 g/m2 and 12 g/m2 CNF (A, B)

and M-CNF (C, D) layers after adsorbing/filtering BSA and SA respectively (the

spectra for pristine CNF and M-CNF, and BSA and SA are also shown for reference in

the corresponding figure).

22

340

341

342

343

344

345

346

347

348

349

350

Also the FTIR spectra of M-CNF fouled by BSA and SA are shown in Figures 5C

and 5D. The bands at about 1400 cm-1 (M-CNF) can be attributed to C-H in-plane

bending vibration [36]; the observed bands at 1603 cm-1 are attributed to C=O or C=C

stretching vibrations, indicating respectively the existence of furanic and aromatic

groups [37-39]; the bands at 629 cm−1 are ascribed to deformation vibration of C-H

planes [36, 40]. The bands of M-CNF, 1049 and 1087 cm-1, decreased which

correspond to alcoholic C-O stretching vibration for BSA and SA. The peaks present

at 880 and 3127 cm-1, corresponded to the CH out-of-plane deformations in aromatic

rings and =C–H stretching modes, respectively [10]. CH bonds give rise to the peaks

at 2925/2890 cm-1 (C–H aliphatic stretch), which indicate the adsorption of BSA or

SA. In general, the FTIR results indicate that the surfaces of the M-CNF samples were

more abundant in hydroxyl, carbonyl, carboxyl, and aromatic groups compared to the

CNF samples. These surface groups provide possibilities for their further

functionalization, making the materials more hydrophilic [36, 39] and thus able to

adsorb more BSA (1620 cm-1) and SA (1127 cm-1) (Figure 5C and Figure 5D). This

may be the reason that the M-CNF layer showed the better flux performance than the

CNF layer (Figure 3), whereby the surface groups provided more adsorptive sites for

the M-CNF to adsorb the NOM foulants (BSA and SA), and preventing them from

reaching the UF surface and pores, and fouling them.

3.5. Thermogravimetric analysis

The results of thermogravimetric analysis (TGA) and differential

23

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

thermogravimetric analysis (DTG) of BSA in the influent and effluent before and after

the membrane filtration with CNF or M-CNF layers are presented in Figure 6. The

weight loss curves shown in Figures 6A and 6C represent BSA in the influent and

effluents of the membrane coated by CNF/M-CNF layers with two thicknesses. It is

clear that the weight loss occurred between 100 and 150 oC, which is attributed to the

evaporation of different MW organic substances of BSA. It is believed that physically

adsorbed and hydrogen bond-linked water molecules were lost at this first stage of

heating [41, 42], while the larger MW components of BSA were adsorbed or retained

by the CNF/M-CNF layers. The weight loss was about 29.5%, 28.0% and 23.6%

(Table S2) respectively for BSA influent, BSA effluent of 12 g/m2 and 24 g/m2 CNF

pre-coating layers, with the corresponding maximum thermal decomposition

occurring at 131.4 oC, 130.5 oC and 127.1 oC (Figure 6B); the corresponding peaks for

the M-CNF layers occurred at 131.4 oC, 128.6 oC and 125.3 oC (Figure 6D). Since a

higher maximum decomposition temperature corresponds to larger MW of organic

matter in the influent and effluent, this indicated that more BSA of larger MW was

adsorbed by the M-CNF coated membrane. In addition, the BSA content in the

influent and filtered effluent (by the membrane with 12 g/m2 and 24 g/m2 CNF/M-

CNF pre-coating layer) was also evaluated in terms of UV absorbance, and the

corresponding spectra as shown in Figures 6E and 6F. These showed a lower UV

absorbance for the M-CNF layer effluent, which indicated a greater BSA removal

(Figure 3F) by the M-CNF layer compared to the CNF layer, in agreement with the

TGA results.

24

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

25

395

396

Figure 6. The weight loss and DTG curves of BSA in the influent and effluents of the

UF membrane with CNF layer (A, B) or M-CNF layer (C, D); the UV spectra of BSA

in the influent and effluent of UF membrane with 12 g/m2 (E) and 24 g/m2 (F),

respectively, of CNF and M-CNF layers (There is no UV absorbance of SA).

4. Conclusions

In the study described in this paper, the performance of two novel coatings (CNF

and M-CNF) on a PVDF UF membrane has been evaluated, with the coatings

prepared from bacterial cellulose via freezing and pyrolyzing. Both types of coated

membrane were compared with a conventional PVDF membrane in terms of

mitigating membrane fouling and organic substrate retention when treating solutions

of bovine serum albumin (BSA) and sodium alginate (SA) as representative NOM

constituents of raw surface waters. The principal results of the study were as follows:

(1) Both CNF and M-CNF coatings on the UF membrane significantly decreased

membrane fouling, especially M-CNF, which was able to decrease flux decline to a

very low level. It was found that an increase in the mass loading of CNF or M-CNF,

and thus, coating thickness, decreased the membrane flux decline (fouling).

26

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

(2) Both CNF and M-CNF coatings gave enhanced removal of SA, and particularly,

BSA, achieving 90% overall removal at the highest fiber loading. Although the CNF

and M-CNF layers had similar surface areas, the M-CNF exhibited greater removals,

possibly due to its more compact structure. The removal of BSA and SA increased

with the CNF and M-CNF loading.

(3) The M-CNF fiber surfaces appeared to contain more functional groups than those

of the CNF, which may have contributed to the superior performance of the former

coating. Thus, the observed protection of the UF membrane to NOM fouling by the

CNF/M-CNF layers may be attributed to both physical separation (size exclusion) and

surface adsorption effects.

Author information

Corresponding Author

*Phone: +86-10-68918696; fax: +86-10-68918696; e-mail:bitkeningsun@163.com

(K.S.); w.yu@imperial.ac.uk (W.Y.).

Notes

The authors declare no competing financial interest.

Acknowledgements

This work was financially supported by the National Natural Science Foundation of

China (51308043).

27

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

References

[1]. Ma H., Burger C., Hsiao B. S., Chu B., 2011. Ultra-fine cellulose nanofibers: new

nano-scale materials for water purification. Journal of Materials Chemistry 21(21),

7507.

[2]. Cai Z. X., Kim J. S., Benjamin M. M., 2008. NOM removal by adsorption and

membrane filtration using heated aluminum oxide particles. Environ Sci Technol

42(2), 619-623.

[3]. Petersen R. J., 1993. Composite Reverse-Osmosis and Nanofiltration Membranes.

J Membrane Sci 83(1), 81-150.

[4]. Barona G. N. B., Cha B. J., Jung B., 2007. Negatively charged poly(vinylidene

fluoride) microfiltration membranes by sulfonation. Journal Of Membrane Science

290(1-2), 46-54.

[5]. Fan L. H., Nguyen T., Roddick F. A., Harris J. L., 2008. Low-pressure membrane

filtration of secondary effluent in water reuse: Pre-treatment for fouling reduction.

Journal Of Membrane Science 320(1-2), 135-142.

[6]. Chen G. Q., Kanehashi S., Doherty C. M., Hill A. J., Kentish S. E., 2015. Water

vapor permeation through cellulose acetate membranes and its impact upon

membrane separation performance for natural gas purification. Journal of Membrane

Science 487, 249-255.

[7]. Jarusutthirak C., Amy G., Croue J. P., 2002. Fouling characteristics of wastewater

effluent organic matter (EfOM) isolates on NF and UF membranes. Desalination

145(1-3), 247-255.

[8]. Kim J. S., Shi W., Yuan Y. P., Benjamin M. M., 2007. A serial filtration 28

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

investigation of membrane fouling by natural organic matter. J Membrane Sci 294(1-

2), 115-126.

[9]. Tomaszewska M., Mozia, S., 2002. Removal of organic matter from water by

PAC/UF system. Water Research 36(16), 4137-4143.

[10]. Grzyb B., Hildenbrand C., Berthon-Fabry S., Bégin D., Job N., Rigacci A.,

Achard P., 2010. Functionalisation and chemical characterisation of cellulose-derived

carbon aerogels. Carbon 48(8), 2297-2307.

[11]. Schideman L. C., Mariñas B. J., Snoeyink V. L., Qi S., Campos C., 2007.

Three-Component Adsorption Modeling to Evaluate and Improve Integrated

Sorption−Membrane Processes. Environmental Science & Technology 41(18), 6547-

6553.

[12]. Yu W. Z., Xu L., Qu J. H., Graham N., 2014. Investigation of pre-coagulation

and powder activate carbon adsorption on ultrafiltration membrane fouling. Journal

Of Membrane Science 459, 157-168.

[13]. Cai Z. X., Wee C., Benjamin M. M., 2013. Fouling mechanisms in low-

pressure membrane filtration in the presence of an adsorbent cake layer. Journal Of

Membrane Science 433, 32-38.

[14]. Kang S. K., Choo K. H., 2010. Why does a mineral oxide adsorbent control

fouling better than powdered activated carbon in hybrid ultrafiltration water

treatment? J Membrane Sci 355(1-2), 69-77.

[15]. Nomura T., Zhu C., Sheng N., Saito G., Akiyama T., 2015.

Microencapsulation of metal-based phase change material for high-temperature

29

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

thermal energy storage. Sci Rep-Uk 5, 9117.

[16]. Zhang M., Li C., Benjamin M. M., Chang Y., 2003. Fouling and natural

organic matter removal in adsorbent/membrane systems for drinking water treatment.

Environ Sci Technol 38(7), 1663–1669.

[17]. Dobosz K. M., Kuo-Leblanc C. A., Martin T. J., Schiffman J. D., 2017.

Ultrafiltration Membranes Enhanced with Electrospun Nanofibers Exhibit Improved

Flux and Fouling Resistance. Ind Eng Chem Res 56(19), 5724-5733.

[18]. Shi W., Benjamin M. M., 2008. Membrane interactions with NOM and an

adsorbent in a vibratory shear enhanced filtration process (VSEP) system. J

Membrane Sci 312(1-2), 23-33.

[19]. Kim J. S., Cai Z. X., Benjamin M. M., 2008. Effects of adsorbents on

membrane fouling by natural organic matter. J Membrane Sci 310(1-2), 356-364.

[20]. Kim J., Cai Z. X., Benjamin M. M., 2010. NOM fouling mechanisms in a

hybrid adsorption/membrane system. J Membrane Sci 349(1-2), 35-43.

[21]. Saha N. K., Balakrishnan M., Ulbricht M., 2006. Polymeric membrane

fouling in sugarcane juice ultrafiltration: role of juice polysaccharides. Desalination

189(1-3), 59-70.

[22]. Katsoufidou K., Yiantsios S. G., Karabelas A. J., 2007. Experimental study of

ultrafiltration membrane fouling by sodium alginate and flux recovery by

backwashing. Journal Of Membrane Science 300(1-2), 137-146.

[23]. Susanto H., Arafat H., Janssen E. M. L., Ulbricht M., 2008. Ultrafiltration of

polysaccharide-protein mixtures: Elucidation of fouling mechanisms and fouling

control by membrane surface modification. Sep Purif Technol 63(3), 558-565.30

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

[24]. Herrero C., Pradanos P., Calvo J. I., Tejerina F., Hernandez A., 1997. Flux

decline in protein microfiltration: Influence of operative parameters. J Colloid Interf

Sci 187(2), 344-351.

[25]. Yu W. Z., Graham N., Yang Y. J., Zhou Z. Q., Campos L. C., 2015. Effect of

sludge retention on UF membrane fouling: The significance of sludge crystallization

and EPS increase. Water Res 83, 319-328.

[26]. Shukla A. K., Alam J., Alhoshan M., Dass L. A., Muthumareeswaran M. R.,

2017. Development of a nanocomposite ultrafiltration membrane based on

polyphenylsulfone blended with graphene oxide. Sci Rep-Uk 7.

[27]. Watanabe K., Tabuchi M., Morinaga Y., Yoshinaga F., 1998. Structural

Features and Properties of Bacterial Cellulose Produced in Agitated Culture. Cellulose

5(3), 187-200.

[28]. Yamamoto H., Horii F., Odani H., 1989. Structural changes of native

cellulose crystals induced by annealing in aqueous alkaline and acidic solutions at

high temperatures. Macromolecules 22(10), 4130-4132.

[29]. Ma J., Yang M., Yu F., Zheng J., 2015. Water-enhanced removal of

ciprofloxacin from water by porous graphene hydrogel. Sci Rep-Uk 5(1), 13578.

[30]. Fan Z., Wang K., Wei T., Yan J., Song L., Shao B., 2010. An environmentally

friendly and efficient route for the reduction of graphene oxide by aluminum powder.

Carbon 48(5), 1686-1689.

[31]. Glassford S. E., Byrne B., Kazarian S. G., 2013. Recent applications of ATR

FTIR spectroscopy and imaging to proteins. Biochimica et Biophysica Acta (BBA) -

31

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

Proteins and Proteomics 1834(12), 2849-2858.

[32]. Haris P. I., Severcan F., 1999. FTIR spectroscopic characterization of protein

structure in aqueous and non-aqueous media. J Mol Catal B-Enzym 7(1-4), 207-221.

[33]. Feng Y., Zhang X., Shen Y., Yoshino K., Feng W., 2012. A mechanically

strong, flexible and conductive film based on bacterial cellulose/graphene

nanocomposite. Carbohydrate Polymers 87(1), 644-649.

[34]. Wan Y. Z., Huang Y., Yuan C. D., Raman S., Zhu Y., Jiang H. J., He F., Gao

C., 2007. Biomimetic synthesis of hydroxyapatite/bacterial cellulose nanocomposites

for biomedical applications. Materials Science and Engineering: C 27(4), 855-864.

[35]. Chiew C. S. C., Poh P. E., Pasbakhsh P., Tey B. T., Yeoh H. K., Chan E. S.,

2014. Physicochemical characterization of halloysite/alginate bionanocomposite

hydrogel. Applied Clay Science 101, 444-454.

[36]. Ren Y., Xu Q., Zhang J., Yang H., Wang B., Yang D., Hu J., Liu Z., 2014.

Functionalization of biomass carbonaceous aerogels: selective preparation of

MnO2@CA composites for supercapacitors. Acs Appl Mater Inter 6(12), 9689-97.

[37]. Sun X., Li Y., 2004. Colloidal carbon spheres and their core/shell structures

with noble-metal nanoparticles. Angewandte Chemie International Edition 43(5), 597-

601.

[38]. Jin Y., Chen H., Chen M., Liu N., Li Q., 2013. Graphene-patched

CNT/MnO2 nanocomposite papers for the electrode of high-performance flexible

asymmetric supercapacitors. Acs Appl Mater Inter 5(8), 3408-16.

[39]. White R. J., Budarin V., Luque R., Clark J. H., Macquarrie D. J., 2009.

Tuneable porous carbonaceous materials from renewable resources. Chem Soc Rev 32

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

38(12), 3401.

[40]. Titirici M. M., Antonietti M., 2010. Chemistry and materials options of

sustainable carbon materials made by hydrothermal carbonization. Chem Soc Rev

39(1), 103-16.

[41]. Pavlidou S., Papaspyrides C. D., 2008. A review on polymer–layered silicate

nanocomposites. Progress in Polymer Science 33(12), 1119-1198.

[42]. Sajjan A. M., Jeevan Kumar B. K., Kittur A. A., Kariduraganavar M. Y.,

2013. Novel approach for the development of pervaporation membranes using sodium

alginate and chitosan-wrapped multiwalled carbon nanotubes for the dehydration of

isopropanol. Journal of Membrane Science 425-426, 77-88.

Supporting Information

The antifouling performance of an ultrafiltration membrane with

33

551

552

553

554

555

556

557

558

559

560

561

pre-deposited carbon nanofiber layers for water treatment

Ting Liu1,2, Yuanlong Lian1, Nigel Graham2, Wenzheng Yu2,3*, Kening Sun1*

1 School of Chemistry and Chemical Engineering, Beijing Institute of Technology,

Beijing 100081, China (liuting@bit.edu.cn, sunkening@bit.edu.cn)

2 Department of Civil and Environmental Engineering, Imperial College London,

South Kensington Campus, London SW7 2AZ, UK (n.graham@imperial.ac.uk)

3 State Key Laboratory of Environmental Aquatic Chemistry, Research Center for

Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China

(w.yu@imperial.ac.uk, wzyu@rcees.ac.cn)

Table S1. The BET surface area and BJH pore diameter distribution of pristine CNF

and M-CNF, and M-CNF fouled by BSA and SA (with 24 g/m2 and 12 g/m2 M-CNF

loadings).

Table S2. The weight loss percentage of BSA inflow and effluent at different

temperatures

Figure S1. Nitrogen adsorption-desorption isotherms of samples of the pristine CNF

and M-CNF layers, and 24 g/m2 and 12 g/m2 M-CNF layers fouled by BSA and SA

34

Table S1. The BET surface area and BJH pore diameter distribution of pristine CNF

and M-CNF, and M-CNF fouled by BSA and SA (with 24 g/m2 M-CNF loadings).

Physical properties of CNF and M-CNF (without and with adsorbed BSA or SA).

SampleBET Surface

Area (m2/g)

BJH adsorption BJH desorption

Pore Volume

(cc/g)

Pore Diameter

(nm)

Pore Volume

(cc/g)

Pore Diameter

(nm)

CNF 129.0 0.220 1.43 0.215 1.41

M-CNF 104.9 0.453 1.94 0.449 1.82

24 g/m2 M-

CNF-BSA81.5 0.198 2.06 0.207 1.94

24 g/m2 M-

CNF-SA136.2 0.495 1.93 0.486 1.70

35

Table S2. The weight loss percentage of BSA inflow and effluent at different

temperatures.

Temp(◦C)

Weight loss (%)

Sample

100 153.7 200 300 400 456 500 600

BSA inflow1.6

825.91

26.2

327.68

28.7

329.12 29.3 29.45

　 　 　 153.3 　 　 　 　 　 　

12g/m2 CNF effluent1.8

923.96 24.7 25.81

27.0

127.36

27.5

928.04

　 　 　 150.2 　 　 　 　 　 　

24g/m2 CNF effluent2.0

720.76

21.3

522.46

23.1

523.41

23.5

723.58

36

Figure S1. Nitrogen adsorption-desorption isotherms of samples of the pristine CNF

and M-CNF layers, and 24 g/m2 and 12 g/m2 M-CNF layers fouled by BSA and SA.

37