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 ([email protected], [email protected])
2 Department of Civil and Environmental Engineering, Imperial College London,
South Kensington Campus, London SW7 2AZ, UK ([email protected])
3 Key Laboratory of Drinking Water Science and Technology, Research Center for
Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China
([email protected], [email protected])
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:[email protected]
(K.S.); [email protected] (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 ([email protected], [email protected])
2 Department of Civil and Environmental Engineering, Imperial College London,
South Kensington Campus, London SW7 2AZ, UK ([email protected])
3 State Key Laboratory of Environmental Aquatic Chemistry, Research Center for
Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China
([email protected], [email protected])
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