Accepted Manuscript Hydrophilic fraction of natural organic matter causing irreversible fouling of microfiltration and ultrafiltration membranes Hiroshi Yamamura, Kenji Okimoto, Katsuki Kimura, Yoshimasa Watanabe PII:

S0043-1354(14)00056-6

DOI:

10.1016/j.watres.2014.01.024

Reference:

WR 10427

To appear in:

Water Research

Received Date: 6 July 2013 Revised Date:

25 December 2013

Accepted Date: 13 January 2014

Please cite this article as: Yamamura, H., Okimoto, K., Kimura, K., Watanabe, Y., Hydrophilic fraction of natural organic matter causing irreversible fouling of microfiltration and ultrafiltration membranes, Water Research (2014), doi: 10.1016/j.watres.2014.01.024. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT 1 Hydrophilic fraction of natural organic matter causing irreversible fouling of microfiltration and ultrafiltration 2 membranes

4 Hiroshi Yamamura a *, Kenji Okimotob, Katsuki Kimura b, Yoshimasa Watanabec 5

RI PT

3

SC

6 a Department of Integrated Science and Engineering for Sustainable Society, Chuo University, 1-12-27 Kasuga,

M AN U

7 Bunkyo-ku, Tokyo 112-8551, Japan

8 b Division of Built Environment, Graduate School of Engineering, Hokkaido University, N13W8, Sapporo 060-8628, 9 Japan, [email protected]

TE D

10 c Research and Development Initiative, Chuo University, 1-12-27 Kasuga, Bunkyo-ku, Tokyo 112-8551, Japan, 11 [email protected]

13 *Corresponding author.

EP

12

AC C

14 Phone number: +81-3-3817-7257

15 E-mail address: [email protected] (Hiroshi YAMAMURA). 16 Postal address: 1-12-27 Kasuga, Bunkyo-ku, Tokyo 112-8551, Japan 17 18 19 1

ACCEPTED MANUSCRIPT 20 Abstract

21 Although membrane filtration is a promising technology in the field of drinking water treatment, persistent membrane

RI PT

22 fouling remains a major disadvantage. For more efficient operation, causative agents of membrane fouling need to be 23 identified. Membrane fouling can be classified into physically reversible and irreversible fouling on basis of the

SC

24 removability of the foulants by physical cleaning. Four types of natural organic matter (NOM) in river water used as a 25 source of drinking water were fractionated into hydrophobic and hydrophilic fractions, and their potential to develop

M AN U

26 irreversible membrane fouling was evaluated by a bench-scale filtration experiment together with spectroscopic and 27 chromatographic analyses. In this study, only dissolved NOM was investigated without consideration of interactions 28 of NOM fractions with particulate matter. Results demonstrated that despite identical total organic carbon (TOC),

TE D

29 fouling development trends were significantly different between hydrophilic and hydrophobic fractions. The 30 hydrophobic fractions did not increase membrane resistance, while the hydrophilic fractions caused severe loss of

EP

31 membrane permeability. These results were identical with the case when the calcium was added to hydrophobic and

AC C

32 hydrophilic fractions. The largest difference in NOM characteristics between hydrophobic and hydrophilic fractions 33 was the presence or absence of macromolecules; the primary constituent causing irreversible fouling was inferred to 34 be “biopolymers,” including carbohydrates and proteins. In addition, the results demonstrated that the extent of 35 irreversible fouling was considerably different depending on the combination of membrane materials and NOM 36 characteristics. Despite identical nominal pore size (0.1 µm), a polyvinylidene fluoride (PVDF) membrane was found 37 to be more rapidly fouled than a PE membrane. This is probably explained by the generation of strong hydrogen 2

38 bonding between hydroxyl groups of biopolymers and fluorine of the PVDF membrane. On the basis of these findings, ACCEPTED MANUSCRIPT 39 it was suggested that the higher fouling potential of the hydrophilic fraction of the dissolved NOMs from various 40 natural water sources are mainly attributed to macromolecules, or biopolymers.

RI PT

41 42 Keywords: natural organic matter; hydrophilic organic matter; microfiltration; ultrafiltration; membrane fouling 43

SC

44

M AN U

45 1. Introduction

46 Recently, low-pressure filtration through semipermeable membranes (microfiltration (MF) and ultrafiltration (UF)) has 47 received considerable attention. This type of filtration is used all over the world in drinking water treatment processes to

TE D

48 remove pathogenic microorganisms (Jacangelo et al. 1995). However, membrane fouling increases the amount of 49 energy needed for filtration because it decreases water permeability to a large extent (AWWA Membrane Technology

EP

50 Research Comittee 2005). Although the membrane is periodically cleaned by air-scrubbing or backwashing to reduce 51 fouling, some organic and inorganic substances accumulate in and on the membrane during long-term operations.

AC C

52 These deposits, which can only be removed by chemical cleaning, result in irreversible fouling, which is the persistent 53 loss of membrane permeability (Kimura et al. 2004, Yamamura et al. 2007a). Because the chemical cleaning process 54 creates problems, such as disposal of the spent chemical cleaner and shortening of membrane lifetime (Malmrose et al. 55 2003), frequency of chemical cleaning should be minimized. An alternative approach is to gain more knowledge about 56 what causes irreversible fouling and formulate countermeasures to prevent or reduce fouling. 57 In various previous studies on membrane fouling in drinking water treatment, natural organic matter (NOM) has been 3

58 suggested as a major constituent causing ACCEPTED significant loss of membrane permeability (Cho et al. 2000, Cho et al. 1999, MANUSCRIPT 59 Howe and Clark 2002, Huang et al. 2007, Katsoufidou et al. 2010, Lee et al. 2006a, Yamamura et al. 2007a, Yuan and 60 Zydney 2000, Zularisam et al. 2011). NOM is composed of various types of non-biodegradable organic compounds

RI PT

61 having size distribution from nanometer to micrometer (Leenheer 2009). Owing to the diversity of NOM composition, 62 identification of the key fraction causing irreversible fouling has not been investigated successfully, and no single 63 cause of irreversible fouling could be identified (Huang et al. 2007, Katsoufidou et al. 2010, Zularisam et al. 2011).

SC

64 NOM can be operationally classified into two fractions: hydrophobic and hydrophilic (Aiken et al. 1992). In early

M AN U

65 studies, many researchers focused on fouling caused by hydrophobic organic matter. In this context, the hydrophobic 66 fractions are mainly composed of humic and fulvic acids that are metabolized by natural or biological degradation, 67 and their structures are known to be rich in aromatic carbon and carboxyl groups (Krasner et al. 1996, Leenheer 2009,

TE D

68 Ma et al. 2001). Commercially available humic acid has often been used as a hydrophobic organic matter surrogate 69 when investigating fouling potential (Katsoufidou et al. 2005, Tian et al. 2013a, Yuan and Zydney 2000). Yuan et al.

EP

70 and Tian et al., for example, used humic substances purchased from Aldrich chemicals and evaluated the fouling 71 potentials that occurred in MF and UF membranes. However, because the composition and molecular weight

AC C

72 distribution of such organic reagents substantially differ from the organic matter extracted from natural water samples 73 (Kimura et al. 2006, Leenheer 2009), the results obtained in their studies would be difficult to apply to fouling that 74 occurs in an actual water treatment operation. 75 In recent years, fouling caused by hydrophilic organic matter has received considerable attention. The hydrophilic 76 fraction is mainly composed of polysaccharides and proteins, and their structure is rich in aliphatic carbons and 77 hydroxyl groups (Krasner et al. 1996, Leenheer 2009, Ma et al. 2001). Most of the filtration experiments were 4

78 performed using natural water samples,ACCEPTED and some advanced analytical tools including Fourier-transform infrared MANUSCRIPT 79 spectroscopy (FTIR) or liquid chromatography with organic carbon detector (LC-OCD) were often used to identify 80 fouling due to the hydrophilic fraction (Cho et al. 1998, Kimura et al. 2004, Schmitt and Flemming 1998, Tian et al.

RI PT

81 2013b). Cho et al. and Kimura et al. compared the FTIR signatures of fouled and clean membranes and showed that the 82 fouled membrane was covered with more polysaccharide-like substances than with humic substances (Cho et al. 1998, 83 Kimura et al. 2004).

SC

84 More recently, synergistic effects of particulate matter and different organic fractions (i.e., humic substance and

M AN U

85 biopolymers) have also been recognized through the studies using model substances (Jermann et al. 2008, Jermann et 86 al. 2007, Xiao et al. 2013) or natural waters (Tian et al. 2013b). Jermann et al. (2007) proposed that humic substances 87 first adsorbed on the membrane and could function as a bridge between alginate and membrane, which lead to more

TE D

88 stable, thereby, less reversible fouling layer than the homogenous alginate cake, while Tian et al. (2013b) suggested 89 that the presence of biopolymer helped to accumulate the particles by forming fouling layer on the surface of

EP

90 membrane.

91 On the other hand, many researchers have also pointed out that calcium in the feed water would play an important role

AC C

92 in the development of fouling caused by NOM. The calcium may induce the pore blocking by coagulating the humic 93 acid and may provide a bridge binding between negatively charged NOM and negatively charged membrane surface. 94 Although Jermann et al. (2007) and Tian et al. (2013a) demonstrated the acceleration of fouling with the addition of 95 calcium to NOM solution, there have been contradictory results as well. Lee et al. (2006) investigated the influence of 96 calcium on the fouling development by the filtration using hydrophobic NOMs collected from natural water, and 97 found that there was no difference in the trend in fouling development with and without the calcium addition. 5

98

ACCEPTED MANUSCRIPT

99 We speculate that those controversies over the decades are probably attributed to (1) ambiguity of the definitions of 100 “irreversible” fouling and (2) the variation of characteristics of the NOM used for the filtration experiment depending

RI PT

101 on when (Her et al. 2000) or where (Lee et al. 2004) the original water samples were taken. Although the importance 102 of irreversible fouling has been recognized, most studies described above were based on relatively short-term 103 experiments without physical cleanings, thereby, did not clearly differentiate irreversible fouling and reversible

SC

104 fouling. To identify the key fraction that causes irreversible fouling, experiments that include physical cleaning over

M AN U

105 relatively long periods of operation is indispensable. In addition, due to the diverse characteristics of NOM found in 106 different water samples, it is necessary to use various types of NOM in the same series of filtration experiments. 107

TE D

108 In the present study, we investigated the key fraction of NOM that causes irreversible fouling in MF and UF 109 membranes, using the comprehensive combinations of four types of NOM sources and four types of MF and UF

EP

110 membranes. The dissolved organic matter was further fractionated into hydrophobic and hydrophilic fractions using 111 XAD resins. Then, a bench-scale MF and UF filtration experiment was carried out by using the fractionated samples

AC C

112 in the presence of routine physical cleaning to determine the potential of physically irreversible fouling of each 113 NOM fraction. In addition, influence of calcium on the fouling caused by hydrophobic and hydrophilic NOMs was 114 determined by a comparison of fouling trend in the presence and absence of calcium. By integrating the fouling 115 potential with the dissolved NOM characteristics of each fraction, the key fraction contributing to irreversible fouling 116 and the influence of water sources on the causes of irreversible fouling was further discussed. As well, interactions 117 with colloidal or particulate matter were not investigated as they were outside the scope of this study. 6

118

ACCEPTED MANUSCRIPT

119

RI PT

120 2. Materials and Methods

M AN U

SC

121 2.1. Study sites

TE D

122

AC C

124

EP

123 Fig. 1 Location of study sites

125 The NOMs used in this study were obtained from four distinctive drinking water sources in Japan: Toyohira River in 126 Hokkaido, Lake Inbanuma in Chiba, Kushiro River in Hokkaido, and Yodo River in Osaka. The Toyohira River water 127 (Ty-RW) was sampled at the Moiwa water purification plant, Sapporo city, Hokkaido. This river is about 72.5 km long 128 and drains an area of 904.8 km2. The level of organic matter in the Toyohira River is very low compared to the other 129 three water sources (total organic carbon (TOC): 0.8 mg-C/L, specific UV absorbance (SUVA):2.8 L/(m·mg)). Water

7

130 from Lake Inbanuma (In-LW) was sampled at the intake of the Inbanuma plant, Sakura city, Chiba. This lake is ACCEPTED MANUSCRIPT 131 approximately 40 km northeast of the Tokyo metropolitan area and its catchment area (which includes seven cities, six 132 towns, and two community villages) is approximately 514 km2. Owing to chronic pollution by human sewage and

RI PT

133 agricultural fertilization, heavy eutrophication (TOC: 5.7 mg-C/L, SUVA: 2.1 L/(m·mg)) has recently become a 134 serious problem. Kushiro River water (Ks-RW) was sampled at the Aikoku water purification plant, Kushiro city, 135 Hokkaido. This river is located in the eastern part of Hokkaido and has a catchment area of 2,204 km2. The river flows

SC

136 into the Kushiro Mire, which has been protected under the Ramsar Convention and designated as a national park.

M AN U

137 Nevertheless, water quality has recently worsened (TOC: 0.9 mg-C/L, SUVA: 4.2 L/(m·mg)) because surrounding 138 areas have been used for urban and agricultural purposes. The Yodo River water (Yd-RW) was sampled at the Kunijima 139 water purification plant, Osaka city, Osaka. This river is 148.31 km long and drains an area of 8,420 km2. The original

TE D

140 source of the Yodo River is Lake Biwa in Shiga. Because the river runs through metropolitan areas, including Shiga and 141 Kyoto cities, half of the water in the river contains wastewater discharges (TOC: 1.8 mg-C/L, SUVA: 2.0 L/(m·mg)).

EP

142 2.2. Concentration and fractionation of sampled NOMs

AC C

143 The raw water samples were fractionated into hydrophobic, transphilic, and hydrophilic fractions after concentration 144 using a reverse osmosis (RO) membrane. Eight hundred liters of collected water was immediately filtered using three 145 cartridge filters in a step-by-step manner (10 µm, 2 µm, and 0.45 µm) to remove suspended matter and further applied 146 to a cation exchange softener to remove divalent cations. The removal of divalent cations mitigates the formation of 147 scaling in the following RO concentration process. The organic content of the samples was then concentrated using a 1 148 inch spiral wounded RO membrane module (TFC-SW, Koch membrane systems, inc,, USA; salt rejection rate 99.75% 8

149 at 800 psi) until the TOC of the retentateACCEPTED became greater than 10 mg-C/L. The RO system was operated in cross-flow MANUSCRIPT 150 mode under trans-membrane pressure (TMP) of 10 bar; the permeate flow rate was not reduced to a large extent 151 during the operation (Sun et al. 1995). The overall recovery of organic matter defined by the following equation was

RI PT

152 95% for Ty-RW, 81% for In-LW, 80% for Ks-RW, and 91% for Yd-RW. 153 154 Recovery of organic matter =

(DOC of concentrated sample) × (Volume of concentrated sample) (DOC of pretreated sample) × (Volume of pretreated sample)

SC

155

M AN U

156 Subsequently, 40 L of concentrates (dissolved organic carbon (DOC): 10.0 mg-C/L) was fractionated into three groups 157 using Amberlite DAX-8 and Amberlite XAD-4 resins: hydrophobic fraction (HPO-f), transphilic fraction (TPI-f), and 158 hydrophilic fraction (HPI-f). Fractionation was conducted based on the method established by Aiken et al. (1992). In

TE D

159 brief, the concentrated NOMs were first acidified to pH 2 with HCl and then pumped through DAX-8 and XAD-4 160 columns at 200 mL/min in series. The portion that passed through both DAX-8 and XAD-4 resins without retention was

EP

161 designated as the HPI-f, the portion that was retained on the DAX-8 resin was designated as the HPO-f, and the portion 162 that was retained on the XAD-4 resin was designated as the TPI-f. The ratios of carbon content for HPI-f were

AC C

163 calculated by multiplying DOC and volume of HPI-f. The value of TPI-f was determined by subtracting the total 164 carbon mass of HPI-f from that of the samples passing through the DAX-8 resin. The value of HPO-f was determined 165 by subtracting the total carbon mass of the concentrates from the sum of the total mass of HPO-f and HPI-f. The 166 HPO-f and TPI-f were eluted from the resins by washing the resins with 0.1N NaOH at 100 mL/min (Aiken et al. 1992). 167 The three fractions were deionized by electrodialysis membrane (SELEMION, AGC Engineering co., Ltd., Tokyo, 168 Japan) operation until electric conductivity was less than 0.5 mS/Ω and were diluted to a concentration of 2.0 mg-C/L

9

169 with Mill-Q water before use. Most ofACCEPTED the diluted samples were used for the filtration experiment and NOM MANUSCRIPT 170 characterization, but some of them were further lyophilized and used for FTIR and NMR spectroscopy. The above 171 fractionation process was repeated twice to check the reproducibility of our procedure.

RI PT

172 2.3. Bench-scale filtration experiment 173 HPI-f and HPO-f which were diluted to a concentration of 2.0 mg TOC/L were used as feed water for the filtration

SC

174 experiments. In the case of TPI-f, we could not conduct any filtration experiments and chemical analysis since its 175 recovery rate from XAD-4 was low in some surface water samples and a sufficient amount of organic matter could not

M AN U

176 be obtained by the present fractionation method. Figure 2 shows a schematic diagram of the experimental apparatus 177 used in this study. Tiny membrane modules (membrane area = 1.44 × 10−3 m2) were prepared using hollow-fiber 178 membranes made from four different MF and UF membranes. The two MF membranes had the same nominal pore

TE D

179 size of 0.1µm but were made from different polymers. One MF was made from polyethylene (PE; Mitsubishi Rayon, 180 Tokyo, Japan) and contact angle and surface charge are 32 and -21 mV respectively. The other MF was made from

EP

181 polyvinylidenefluoride (PVDF; Asahikasei Chemicals, Tokyo, Japan) and contact angle and surface charge are 64

AC C

182 and -11 mV respectively. The two UF had the same molecular weight cut-off of 100,000 Da but were made from 183 different polymers. One UF was made from polyacrylonitrile (PAN; Toray Industries, Tokyo, Japan) and contact angle 184 and surface charge are 20 and -13 mV respectively. The other UF was made from polyether sulfone (PES; Daicen 185 membrane-systems, Tokyo, Japan) and contact angle and surface charge are 79 and -20 mV respectively. Membrane 186 filtration was carried out using a peristaltic pump (MP-3, EYELA, Tokyo, Japan), and a constant flow rate mode of 187 operation was used. Membrane permeate flux was fixed at 1.5 m3/m2/d (62.5 L/(m2·h)) for all filtration experiments. 10

188 During filtration, hydraulic backwashing ACCEPTED was carried out every 30 min and neither sparging nor draining was applied. MANUSCRIPT 189 The backwashing duration was 30 s and the pressure was 50 kPa. This filtration cycle was repeated for 12 h. Under the 190 physical cleaning condition of this study, the reversible foulants seems not to be taken off completely because of the

RI PT

191 absence of air scrubbing. To remove the contaminants accumulating on the surface of membrane completely, at the 192 termination of the filtration, membrane module was taken out from the submersion tank and was wiped gently with 193 soft sponge. Then, the degree of irreversible fouling was evaluated according to the value of trans-membrane pressure

TE D

M AN U

SC

194 filtering Milli-Q water under constant flux of 1.5 m/day.

EP

195

196

AC C

Fig. 2 Bench-scale apparatus used for filtration experiments

197 2.4. Analytical methods

198 DOC was measured using a TOC analyzer (TOC-5000A, Shimadzu, Kyoto, Japan) with a standard method 5310B 199 (WEF 1998), and humic substances were spectrophotometrically quantified by ultraviolet (UV) absorbance at 260 nm 200 (U-2000, Hitachi, Tokyo, Japan) and a quartz cuvette with a 1-cm path length (Tambo and Kamei 1998). Before UV 201 and DOC measurements were taken, the samples were filtered through 0.45-µm hydrophilic polytetrafluoroethylene 11

202 (PTFE) membranes. The specific UV absorbance (SUVA)MANUSCRIPT was calculated by dividing the UV absorbance at 260 nm by ACCEPTED 203 the DOC concentration (Tambo and Kamei 1998). Fluorescence excitation emission matrix (FEEM) was determined by 204 a fluorescence spectrophotometer equipped with a 150-W ozone-free xenon lamp (RF-5300PC, Shimadzu, Kyoto,

RI PT

205 Japan). The spectrometer displayed a maximum emission intensity of 1000 arbitrary units (AU) and excitation and 206 emission slits were set to a 10-nm band-pass. Both emission and excitation wavelengths were varied stepwise by 5 nm 207 from 220 nm to 550 nm. To quantify the obtained spectra, we classified the FEEM map into five regions as reported

SC

208 by Chen et al. and fluorescence regional integration (FRI) was applied to each FEEM map. The calculation method for

M AN U

209 FRI is described elsewhere (Chen et al. 2003). The concentrations of carbohydrate were determined by the 210 phenol-sulfuric acid method (Dubois et al. 1956, Whistler and Wolfrom 1962) using glucose as standards. The 211 detection limit of phenol-sulfuric acid method is 0.4 mg-glucose/L (Matsuhiro et al. 2009). The molecular weight

TE D

212 distribution of the DOC was determined by size-exclusion high-performance liquid chromatography (HPLC-SEC) with 213 UV detector (Hitachi, Tokyo, Japan) and Sievers turbo total organic carbon analyzer (800 Turbo). Two columns,

EP

214 TSK-gel G3000SW (Tosoh, Tokyo, Japan) and Toyopearl HW-65S (Tosoh, Tokyo, Japan), connected in series were 215 used to cover a molecular weight range from 10 to 107 Da (Yamamura et al. 2007a). For FTIR studies, KBr pellets

AC C

216 containing 0.25% of the sample were prepared and examined in an FTIR spectrophotometer (FT/IR-8400S, Shimadzu, 217 Kyoto, Japan) at a resolution of 4 cm−1. Solid-state cross polarization magic angle spinning carbon-13 (CPMAS 218 13C) NMR spectra of the membrane foulant were obtained with a Brucker MSL300 spectrometer at 75.47 219 MHz with a spin rate of 8 kHz and a pulse width of 4.5 µs for the 90o pulse. Contact time was set to 1 ms. 220 Acquisition time and recycle delay were 30 ms and 4 s, respectively. 221 12

222 3. Results

ACCEPTED MANUSCRIPT

223 3.1. Characteristics of hydrophobic and hydrophilic organic matter used for membrane filtration experiments 224 3.1.1. Characteristics of sampled natural water

RI PT

225 Four collected drinking water sources showed distinctive water quality in terms of hydrophobicity and organic 226 composition. SUVA, which is a ratio of UV absorbance divided by DOC, has been used as an indicator of organic 227 hydrophobicity. The value of SUVA showed that Ty-RW (2.8 L/(m·mg)) and Ks-RW (4.2 L/(m·mg)) was more

SC

228 hydrophobic than In-LW (2.1 L/(m·mg)) and Yd-RW (2.0 L/(m·mg)). The relative abundances of carbon molecular

M AN U

229 descriptors, which were obtained from the spectra of (CPMAS 13C) NMR, can provide the difference in structure of 230 organic matter contained in four different drinking water sources more detail (see Table 1). As seen in Table 1, In-LW 231 (35.2%) and Yd-RW (29.3%) was rich in carbohydrate-like organic matters with the feature of the high contents of

TE D

232 C-Al-O, but In-LW (7.9%) was rich in anom-C compared to Yd-RW (6.1). This may suggests the structure of 233 carbohydrate was different between Yd-RW and In-LW. On the other hand, both Ty-RW (13.9) and Ks-RW (11.1)

EP

234 showed high contents of aromatic carbon (C-ar-CH) in their structure, which was in accordance with their relatively 235 high value of SUVA. Comparing the contents of methoxy carbon (C-O-CH3), the value was higher for Ks-RW (8.9)

AC C

236 than for Ty-RW (6.6). This may indicate that the Ks-RW contained more microorganism derived organic carbon than 237 the Ty-RW. The HPO/TPI/HPI composition obtained for four different drinking water sources support the results 238 obtained in in (CPMAS 13C) NMR spectra. Ty-RW, which showed aromatic feature in (CPMAS 13C) NMR spectra, 239 exhibited the highest percentage of hydrophobic fraction. In-LW, which showed high contents of carbohydrate-like 240 structure in (CPMAS 13C) NMR spectra, exhibited relatively high percentage of hydrophilic fraction. Ks-RW, which 241 shows aliphatic rich features in (CPMAS 13C) NMR, exhibited similar composition with In-LW and indicated the 13

242 highest percentage of hydrophobic fraction. Yd-RW, which showed the significant sign of proteins (C-COO) in ACCEPTED MANUSCRIPT 243 (CPMAS 13C) NMR, exhibited the highest percentage of transphilic fraction. 244

246

RI PT

245 Table 1 The relative abundances of carbon molecular descriptors obtained for four different drinking water sources

C-COO

C-ar-O

C-ar-CH

C-anom

C-Al-O

C-OCH3

C-Al-CH

Ty-RW

3.6

13.7

0.0

13.9

6.7

25.1

6.6

30.2

In-LW

0.1

14.6

0.0

10.8

7.9

35.2

8.1

23.0

Ks-RW

4.0

6.7

0.0

11.1

5.4

28.4

8.9

35.5

Yd-RW

4.2

14.6

0.0

8.9

6.1

29.3

7.5

29.4

M AN U

C-C=O

SC

and HPO/TPI/HPI composition of NOMs sampled from four different water sources.

HPO-f

Ty-RW

60

In-LW

49

Ks-RW

50

Yd-RW

53

HPI-f

20

20

23

28

20

30

31

16

TE D

247

TPI-f

248

EP

249

250 3.1.2. Fundamental characteristics of fractionated samples

252 fractions

AC C

251 Table 2 Fundamental characteristics of organic matter fractionated into hydrophobic and hydrophilic

14

ACCEPTED Hydrophobic fraction (HPO-f)MANUSCRIPT (a)

(mg-C/L) UV (1/cm) SUVA (1/m/mg/L) Carbohydrate (mg-Glu/mg-C)

(b)

In-LW

(c)

Ks-RW

(d)

Yd-RW

(a)

Ty-RW

(b)

In-LW

(c)

Ks-RW

(d)

Yd-RW

1.6

2.4

2.4

2.2

1.8

1.7

1.5

1.7

0.055

0.078

0.078

0.063

0.042

0.025

0.025

0.024

3.7

3.3

3.0

2.9

2.4

9.8

9.1

20.3

10.0

45.5

1.5

1.8

1.4

52.7

27.2

46.5

SC

253

RI PT

DOC

Ty-RW

Hydrophilic fraction (HPI-f)

M AN U

254 The fundamental characteristics of the NOM used for membrane filtration are presented in Table 2. Proteins could not 255 be measured as the Lowry method was not sensitive enough (>10mg/L). The TOC of all fractionated samples was 256 adjusted to the same level (2.0 ± 0.5 mg-C/L) to discriminate the influence of the amount of organic matter on the

TE D

257 fouling development. Although the fractionation was successful, some differences were seen in the properties of the 258 NOM. Regarding to the hydrophilic fraction, the value of SUVA was relatively high for Ty-RW (2.4 mg/(L·m)) and

EP

259 Ks-RW (1.8 mg/(L·m)). This would be explained by their high contents of C-ar-O (see Table 1). In addition, In-LW, 260 Yd-RW and Ty-RW showed rich in carbohydrate-like substances of 52.7 mg-Glu/mg-C, 46.5 mg-Glu/mg-C and 45.5

AC C

261 mg-Glu/mg-C, respectively. This was corresponding to large contents of anomeric-C of these three sources. Regarding 262 to the hydrophobic fraction, Ty-RW showed relatively hydrophobic nature, with the sign of high SUVA value of 3.7 263 mg/(L·m). This is probably explained by large contents of C-ar-CH as listed in Table 1. Interestingly, Ks-RW showed 264 the highest concentration of carbohydrate of 20.3 mg-Glu/mg-C in hydrophobic fraction, regardless of the smallest 265 concentration in hydrophilic fraction. To provide more precise features of eight fractionated organic matter, the 266 following section will present spectroscopic and chromatographic analyses on basis of fluorescence EEM, infrared 15

267 spectroscopy, and size-exclusion chromatography. ACCEPTED MANUSCRIPT 268 269 3.1.2. Fluorescence EEMs

RI PT

270 FEEM spectra of the organic matter fractionated into hydrophobic and hydrophilic organic matter are presented in 271 Figures 3 and 4. FEEM spectroscopy can identify the presence of fluorescence organic matter in water samples as 272 small as 5 mL. The vertical and horizontal axes of the representative 3-D map shows excitation and emission

SC

273 wavelengths, respectively. The organic matter (e.g., humic acid, fulvic acid, or protein) can be categorized by

M AN U

274 analyzing shapes and peaks of the contour map. Chen et al. (2003) reviewed previous applications of FEEM analysis to 275 natural water samples and suggested that the FEEM map could be divided into five regions depending on the nature of 276 fluorescent organic matter: Region I includes peaks for aromatic protein I, Region II for aromatic protein II, Region III

TE D

277 for fulvic acid-like (FA-like) substances, Region IV for soluble microbial by-product-like (SMP-like) substances, and 278 Region V for humic acid-like (HA-like) substances. The comparison of the organic composition was achieved by

EP

279 fluorescence regional integration (FRI) method provided by Chen et al (2003). The distribution of FRI in eight

AC C

280 fractionated organic matter was also illustrated in Figure 5.

16

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

281

Fig. 3 Fluorescence excitation–emission matrix of hydrophobic organic matter fractionated from four

283

different drinking water sources

AC C

EP

TE D

282

284 17

Fig. 4 Fluorescence excitation–emission matrix of hydrophilic organic matter fractionated from four ACCEPTED MANUSCRIPT

286

different drinking water sources

M AN U

SC

RI PT

285

TE D

287

288 Fig. 5 Distribution of FRI in hydrophobic and hydrophilic organic matter fractionated from four different drinking

290

water sources

EP

289

AC C

291 Figure 3 illustrated that the shape of the spectra was generally similar but the location of the peaks was marginally 292 different among the HPO-f samples. In Region V, In-LW and Yd-RW had a maximum at approximately 450/260 nm 293 (Ex/Em) and a shoulder at Ex= 420-440 nm and Em=270-330 nm. In addition, only for In-LW and Yd-RW, some 294 microbial-derived organic substances (or SMP-like substances) were found as a broad shoulder at approximately 295 300/270 nm (Ex/Em) in Region IV. The feature of SMP was also identified as higher FRI distribution of Region IV for 296 In-LW and Yd-RW (Figure 5). Lee et al. (2006) showed that FEEM of algal organic matter exhibits three maximums 18

297 at around 430/350, 450/260, and 330/270ACCEPTED nm and the locations of these three maximums were well consistent with MANUSCRIPT 298 that of FEEMs of In-LW and Yd-RW. This suggests that HPO-f of In-LW and Yd-RW has similar feature with the 299 algal organic matter. On the other hand, HPO-f of Ty-RW and Ks-RW had two maximums at 430/350 nm and 250/260

RI PT

300 nm, respectively. Generally, the humic acid exhibits a fluorescence intensity maximum at higher excitation 301 wavelengths compared to fulvic acid, the maximum at 430/350 and 250/260 in Ty-RW and Ks-RW would correspond 302 to the fulvic acid form and humic acid form of humic substances, respectively. Comparing FEEM of HPO-f between

SC

303 Ks-RW and Ty-RW, the intensity of two maximums was three times higher for Ks-RW than for Ty-RW, regardless that

M AN U

304 the same level of organic carbon was applied. This was probably due to the presence of electron-donating functional 305 groups (e.g., amine, hydroxyl) in the organic matter fractionated from Ks-RW (see Figure 6). With regard to 306 fluorescence characteristics, two types of organic matter were seen in the hydrophobic samples: (type A) the In-LW

TE D

307 and Yd-RW groups have features of algal organic matter, and (type B) the Ks-RW and Ty-RW group has features of 308 both fulvic and humic acid.

EP

309

310 In Figure 4, there were peaks around Ex=370-430 and Em=320-350. As discussed above, the maxima at relatively

AC C

311 higher excitation wavelengths show the presence of humic acid, and therefore HPI-f was found to contain humic acid. 312 Interestingly, HPI-f of Ks-RW showed relatively higher FRI distribution of Region V compared to the other HPI-f. 313 FEEM demonstrated that HPI-f of Ks-RW had specific feature in terms of organic structure. 314 In addition to the sign of humic acid (i.e., peaks around Ex=370-430 and Em=320-350), all HPI-f had a maximum or 315 shoulder in Region IV. The peak or shoulder in Region IV indicates the presence of SMP-like organic matter and it 316 was well consistent with the previous studies that determined the FEEMs of hydrophilic fraction of natural water 19

317 samples (Chen et al. 2003, Lee et al. 2006b). Figure 5 MANUSCRIPT supported the SMP-like feature of HPI-f with higher FRI ACCEPTED 318 distribution of Region IV (more than30%) than the HPO-f (less than 15%). Comparing the intensity of the maximum 319 of SMP-like substances, the value for the HPI-f was in the order In-LW (10), Ks-RW (3.1), Yd-RW (2.4) and Ty-RW

RI PT

320 (1.7). Based on the fact that the relative amount of SMP-like organic matter was almost similar except for Ks-RW and 321 level of organic carbon was almost identical among four hydrophilic organic matters, the difference in peak intensity 322 of SMP-like substances may be due to the difference in organic characteristics, as discussed in the difference in

SC

323 intensity of peak maximum for humic acid between Ty-RW and Ks-RW. With regard to fluorescence characteristics,

M AN U

324 two types of organic matter were seen in the hydrophilic samples: (type C) the Ks-RW group was rich in hydrophobic 325 organic matter as well as SMP-like organic matter; (type D) the Ty-RW and In-LW and Yd-RW groups was rich in 326 SMP-like organic matter.

TE D

327 328 3.1.3. FTIR spectra

AC C

EP

329

20

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

330

Fig. 6 FTIR spectra of hydrophobic (left) and hydrophilic (right) organic matter fractionated from four

332

different drinking water sources

TE D

331

333

EP

334 FTIR spectra of the organic matter fractionated into hydrophobic and hydrophilic fractions are shown in Figure 6. By 335 comparing the HPO-f spectra, we can observe some similarities in the location of peaks: 1140, 1380, 1620, and 1660

AC C

336 cm−1 are assigned as humic-like substances; 1080 and 1110 cm−1 are assigned as carbohydrate-like substances; and 337 1580 cm−1 is assigned as an amide-II structure (Leenheer 2009). Comparison of the peak height of the HPO-f reveals 338 that Ty-RW and Ks-RW contained abundant phenolic-like organic matter (1140 and 1620 cm−1), YD-RW was rich in 339 carbohydrate-like substances (1080 and 1100 cm−1), and In-LW contained significant organic matter features common 340 to amide-like substances (1580 cm−1). These characteristics were well consistent with the features of the EEM maps. 341 In addition, for the In-LW spectra, there were clear indications of the presence of ammonium chlorides with a distinct 21

342 peak at 1440 cm−1 and the lipids with broad peaks at approximately 2960, 2940, and 1380 cm−1 (Leenheer 2009). ACCEPTED MANUSCRIPT 343 These reflect the pollution of In-LW by effluent organic matter and algae-organic matter. 344

RI PT

345 The spectra of HPI-f exhibited clear differences in composition and structure of the NOM from the four sample 346 collection sites. All spectra had large and broad peaks near 1080 cm−1 and 3000–3300 cm−1, which were designated as 347 the presence of carbohydrate-like substances (Leenheer 2009). In addition, the interfusion of humic-like substances to

SC

348 the hydrophilic fractions was identified as a shoulder or a peak near 1150 and 1620 cm−1, indicating the presence of a

M AN U

349 phenolic-like substance and C=C bonding, respectively (Leenheer 2009). This supports the EEM maps that 350 demonstrated the presence of humic substances in four hydrophilic samples. We also observed a small but sharp peak 351 at 1380 cm−1 in the hydrophilic samples. This is attributed to the presence of CH3, or the samples were found to

TE D

352 contain N-acetyl amino acid, lipids, or hydrocarbons (Leenheer 2009). Comparison of the peak height between 1080 353 and 1620 cm−1 shows that Ty-RW and Yd-RW were rich in carbohydrate-like structures. In-LW contained significant

EP

354 amounts of both humic-like and carbohydrate-like structures (Leenheer 2009). These findings are in good agreement 355 with the EEM maps that show clear peaks of humic substances for In-LW. In addition, there was a shoulder near 1580

357

AC C

356 cm−1 unique to Ks-RW, indicating the presence of a form of amide-I in the skeleton of the NOM.

358 3.1.4. Molecular size distribution

22

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

359

Fig. 7 Molecular weight distribution of hydrophobic (left) and hydrophilic (right) organic matter

361

fractionated from four different drinking water sources

TE D

360

EP

362

363 The characteristics of hydrophobic and hydrophilic organic matter were then assessed using size exclusion

AC C

364 chromatography. Our LC-OCD system is comparable to DOC-Labor developed by Huber (2011) except for the 365 column selection. Although our LC-OCD is less sensitive than DOC-Labor with five times higher detection limit (i.e., 366 20 ppm), the fractionation range is wider from hundreds Da to ten million Da. From our previous study of molecular 367 size distribution of NOMs, organic matter in natural river water are known to have two identical chromatogram peaks 368 at 10–15 min and 17–30 min. The former was designated as large molecules and had the features of low UV 369 absorbance, and the latter was designated as small molecules and had the feature of high UV absorbance. Based on the 23

370 definition made by Huber et al. (2011), the large molecularMANUSCRIPT can be referred to as a biopolymer and small molecules can ACCEPTED 371 be referred as humic acid (Her et al. 2002, Yamamura et al. 2007a). The molecular size distribution of hydrophobic 372 and hydrophilic organic matter is illustrated in Figure 7. The chromatograms of HPO-f show one significant humic

RI PT

373 acid peak between 17 min and 25 min in retention time and no significant biopolymer peaks. This finding suggests 374 that the SMP-like substances identified in the FEEM and FT-IR spectra analysis would be relatively small in size. 375 On the other hand, the chromatograms of HPI-f showed two peaks of both biopolymer and humic acid. This indicates

SC

376 that the biopolymer mainly existed in the hydrophilic samples but not in the hydrophobic samples or that most of

M AN U

377 biopolymer was categorized as the hydrophilic fraction. Comparing the amount of biopolymer in HPI-f, it was found 378 to be in the order In-LW (1.33 mg/L) >> Yd-RW (0.57 mg/L) > Ty-RW (0.22 mg/L) > Ks-RW (0.16 mg/L). This is 379 well consistent with the order of carbohydrate concentration shown in Table 2 but the relationship between biopolymer

TE D

380 concentration and carbohydrate concentration was not liner (R-Squared value = 0.55). 381

EP

382 3.2. Fouling potentials of hydrophobic and hydrophilic organic matter 383 3.2.1. Development of physically irreversible fouling designated in TMP-operation time diagrams

AC C

384 The time-course change of TMP is shown in Figure S1. Under a constant flow rate, the TMPs increased in an arc with 385 time owing to the development of fouling and sharply decreased immediately after the regular backwashing of 386 membranes (every 30 min of operation). In Figure S1, the difference of TMP values before and after backwashing (A in 387 Figure S1) indicates the degree of physically reversible fouling. Although membrane module was wiped gently with 388 soft sponge at the termination of operation to remove the contaminants accumulating on the surface of membrane 389 completely, the cleaning achieved no mitigation of fouling for all case because of high back-wash efficiency. 24

390 Therefore, the line connecting TMP values just after every backwashing indicates the development of physically ACCEPTED MANUSCRIPT 391 irreversible fouling (B in Figure S1) and the difference of final and initial TMP values (C in Figure S1) designates the 392 degree or potential of physically irreversible fouling. The following series of figures that compare the fouling

RI PT

393 development of fractionated organic matter only show the development of irreversible fouling (B in Figure S1) 394 because this type of fouling is more important in practical operations. 395

SC

396 3.2.2. Development of physically irreversible fouling during filtering of hydrophobic and hydrophilic fractions

398

AC C

EP

TE D

M AN U

397 of NOM with PE and PVDF membranes

399 Fig. 8 Development of physically irreversible fouling during filtering of hydrophobic (left) and hydrophilic 400

(right) organic matter fractionated from four different drinking water sources using a PE membrane

401 25

402 The changes in TMP during filtering of hydrophobic and MANUSCRIPT hydrophilic organic matter using a PE membrane are shown ACCEPTED 403 in Figure 8. There is a large gap in the development of physically irreversible fouling between HPO-f and HPI-f 404 despite equal levels of applied organic matter (2.0±0.5 mg-C/L). Regardless of the source of the NOM, TMP for

RI PT

405 filtering HPO organic matter increased by less than 7 kPa. However, the filtration of HPI organic matter caused a rapid 406 increase of greater than 30 kPa in TMP. This demonstrates that the major cause of physically irreversible fouling in the 407 PE membrane was hydrophilic organic matter. The development of physically irreversible fouling was similar among

SC

408 the four types of hydrophilic organic matter at the initial stage of operation (approximately 400 min); however, there is

M AN U

409 some discrepancy in the degree of physically irreversible fouling. At the termination of operation, the degree of 410 physically irreversible fouling was found to be in the order Yd-RW (33 kPa) > Ty-RW (29 kPa) > Ks-RW (23 kPa) >

AC C

EP

TE D

411 In-LW (15 kPa).

412 413 Fig. 9 Development of physically irreversible fouling during filtering of hydrophobic (left) and hydrophilic (right) 414 organic matter fractionated from four different drinking water sources using a PVDF membrane 26

415

ACCEPTED MANUSCRIPT

416 The changes in TMPs during filtering of hydrophobic and hydrophilic organic matter with a PVDF membrane are 417 shown in Figure 9. Similar to PE membrane filtering, severe fouling was observed only when the hydrophilic organic

RI PT

418 matter was filtered using a PVDF membrane. However, the development of physically irreversible fouling differs 419 slightly between PVDF and PE filtering, particularly at the initial stage of operation. The PVDF membrane was 420 rapidly fouled by HPI-f just after starting operation, and the degree of physically irreversible fouling which was

SC

421 evaluated at the end of the operation was in the order In-LW (42 kPa) > Yd-RW (39 kPa) > Ks-RW (34 kPa) ~ Ty-RW

M AN U

422 (34 kPa), which differs from the order of the degree of fouling that occurred with the PE membrane. On the basis of 423 the results described above, it was clearly demonstrated that the major cause of physically irreversible fouling was 424 hydrophilic organic matter and that the degree of irreversible fouling was strongly influenced by the type of membrane

TE D

425 material.

426 4. Discussion

EP

427 4.1. Summary of the characteristics of organic matter comprising the four different drinking water sources

AC C

428 In the present study, four types of water sources were selected as the distinguishing sampling points: non-polluted 429 river water (Ty-RW), pond water polluted by wastewater effluents and algae organic matter (In-LW), peat river water 430 (Ks-RW), and downstream water affected by wastewater effluents from upstream cities (Yd-RW). These water sources 431 are currently used as sources of drinking water. In future, it is possible that the water from these sources will be treated 432 with a MF/UF membrane for producting drinking water. On the basis of the investigation of the organic characteristics 433 of the fractionated NOM, clear differences in the feature of the organic matter were observed. The characteristics of 27

434 hydrophobic and hydrophilic isolates wereACCEPTED summarized asMANUSCRIPT follows. 435 Based on the FEEM analysis (Figures 3 and 5), HPO-f was categorized into two types; (type A) the In-LW and 436 Yd-RW group has features of algal organic matter and (type B) the Ks-RW and Ty-RW group has a feature of both

RI PT

437 fulvic and humic acid. The most significant difference in the two types of HPO-f was the proportion of SMP-like 438 substances (see Figure 5). The type A isolates showed 5% higher ratio of Region 4 than the type B isolates. With 439 respect to the size distribution of hydrophobic fraction (Figure 7), there were no significant differences in the

SC

440 concentration of both humic substances and biopolymer. This implies that the SMP-like substances detected by

M AN U

441 FEEMs in HPO-f has relatively small molecular size around ~thousand Da.

442 Based on the FEEM analysis (Figures 4 and 5), the HPI-f was categorized into two types; (type C) the Ks-RW group 443 was rich in hydrophobic organic matter as well as SMP-like organic matter and (type D) the Ty-RW, In-LW and

TE D

444 Yd-RW groups was rich in SMP-like organic matter. The category was also identified in carbohydrate concentration 445 (Table 2). The concentration of carbohydrate in HPI-f of Ks-RW was half of that in the other three hydrophilic organic

EP

446 matters. The distribution of FRI (see Figure 5) also illustrated the difference in feature of Ks-RW among four 447 hydrophilic fractions. The portion of Region IV (SMP-like substances) in HPI-f was 13 % higher for Ks-RW than for

AC C

448 the other source waters, although the intensity of maximum of SMP-like substances (Ex=300-350 /Em=270-300) was 449 high secondarily among HPI-f. With respect to the size distribution of hydrophilic fraction (Figure 7), all hydrophilic 450 fractions were found to have both biopolymer and humic acid. The amount of biopolymer concentration was in the 451 order In-LW (1.33 mg/L) > Yd-RW (0.57 mg/L) > Ty-RW (0.22 mg/L) > Ks-RW (0.16 mg/L), and the order was well 452 consistent with that of carbohydrate concentration. Assuming that the protein concentration was proportional to the 453 peak intensity of SMP-like substances in EEM map (Figure 4), the biopolymer concentration was found to have strong 28

454 relationship with the summation of carbohydrate and protein concentrations. ACCEPTED MANUSCRIPT 455 456 4.2. Major fraction of natural organic matter causing irreversible fouling in MF membrane

RI PT

457 The results obtained in this study simulate the initial filtration stage of actual operation because filtration duration (700 458 min) of this study is far shorter than the actual operation (more than 1 week). Under such experimental condition, the 459 adsorption will have more significant role than cake accumulation, and the combination of NOM characteristics,

SC

460 solution and membrane materials seems to have large impact on the degree of irreversible fouling. As described above,

M AN U

461 the water samples used for the membrane filtration experiment were considerably different in terms of both functional 462 properties and molecular weight distribution. However, in spite of such significant differences in organic 463 characteristics, the development of irreversible fouling was almost identical. All filtrations showed that even though

TE D

464 the same amount of organic matter was applied, hydrophobic organic matter did not participate in the development of 465 physically irreversible fouling, while the hydrophilic organic matter did. The conclusions obtained in the present study

EP

466 were almost similar with the case when 12 mg/L of calcium was added to the feeds (Figure 10) although Tian et al. 467 (2013a) demonstrated significant increase in irreversible resistance when the calcium was added to BSA or

AC C

468 commercially available humic solutions. This discrepancy in impact of calcium on the irreversible fouling was 469 probably explained by the difference in NOM characteristics between natural water and artificial water. Based on the 470 study carried out by Yuan et al. (1999), the molecular size distribution was far larger for commercially available humic 471 acid than for natural water. Because the larger particles were easy to be coagulated by the addition of divalent or 472 trivalent cations, sub-micron particles which easily induce the physically irreversible fouling would forms with the 473 addition of calcium. The results obtained in Figure 8-10 clearly indicate that regardless of the water source, the major 29

474 cause of irreversible fouling is hydrophilic rather thanMANUSCRIPT hydrophobic organic matter, although the possibility of ACCEPTED 475 synergetic effect between hydrophilic and hydrophobic organic matters including changing the hydrophobicity or 476 membrane charge by adsorbing to the surface of the membrane, as suggested by Jermann et al. (2007), should be

RI PT

477 considered. The adsorption of the HPO-f on the membrane surface would indirectly affect hydrophilic organic matter’s 478 ability to accumulate on the surface of the membrane by changing the affinity for adhesion to the membrane surface 479 (Jermann et al. 2007, Li and Elimelech 2004).

SC

480 The reason hydrophobic organic matter did not contribute to the physically irreversible fouling is considered to be

M AN U

481 owing to (1) relatively small molecular size compared to the pore size of the membrane and/or (2) low affinity to the 482 membrane surface (Yamamura et al. 2008). The influence of molecular size on the development of physically 483 irreversible fouling was investigated by the determination of the fouling potential of UF membranes (Figure 11). The

TE D

484 UF membrane used in the present study had a cut-off value of 100 kDa, which was small enough to remove some of 485 the hydrophobic organic matter; therefore, the UF of hydrophobic molecules was anticipated to cause irreversible

EP

486 fouling to some extent. However, both a hydrophobic polyethersulfone (PES) membrane and a hydrophilic 487 polyacrylonitrile (PAN) membrane showed no increase in TMPs when the hydrophobic organic matter was filtrated,

AC C

488 while severe fouling was observed only for the filtration of hydrophilic organic matter using a hydrophobic PES 489 membrane. These observations clearly indicate that the size of organic matter does not influence the development of 490 irreversible fouling, but the affinity with the surface of the membrane affects the development of physically 491 irreversible fouling. It was speculated that the hydrophobic organic matter did not have a large affinity with the surface 492 of the membrane and consequently passed through the membrane pores without adsorption or was washed out by 493 routine physical cleaning. 30

SC

RI PT

ACCEPTED MANUSCRIPT

M AN U

494

495 Fig. 10 Development of physically irreversible fouling during filtering of hydrophobic (left) and hydrophilic 496 (right) organic matter fractionated from four different drinking water sources in the presence of 12 mg/L of

TE D

497 calcium using a PVDF membrane

AC C

EP

498

31

SC

RI PT

ACCEPTED MANUSCRIPT

M AN U

499

500 Fig. 11 Development of physically irreversible fouling during filtering of hydrophobic and hydrophilic organic 501 matter fractionated from Ty-RW using PES (left) and PAN (right) membranes

TE D

502

503 It is interesting to note that the degree of irreversible fouling caused by HPI-f was fairly different depending on the

EP

504 combination of membrane material and water sources. In general, the rate of development of irreversible fouling was 505 higher for PVDF membranes than PE membranes (Kimura et al. 2006, Yamamura et al. 2008, Yamamura et al. 2007a,

AC C

506 Yamamura et al. 2007b). This trend is in good accordance with our previous study that included a pilot-scale filtration 507 experiment using PE and PVDF membranes in parallel. Because both PE and PVDF membranes have the same 508 nominal pore size (0.1 µm), the difference in development of irreversible fouling may be owing to the difference in 509 interaction forces between the organic matter and the membrane (Li and Elimelech 2004, Yamamura et al. 2008). 510 Additionally, we found that fouling potential of each membrane seems to have relationship with the amount of 511 biopolymer concentration of fractionated NOM. Figure 12 shows the correlation of biopolymer concentration with 32

512 degree of physically irreversible fouling ACCEPTED in PVDF, PE PES and PAN membranes. For the PVDF membranes, fouling MANUSCRIPT 513 potential was proportional to the logarithm of biopolymer concentration and it did not changed even when the calcium 514 was added to feed water (r2=0.81). The plot for PES membrane was on the line for PVDF membrane, which suggests

RI PT

515 that the fouling trend caused by biopolymer would be similar between PVDF and PES membrane. On the other hand, 516 PE and PAN membranes showed different trend in relation between fouling potential and biopolymer concentration 517 with PVDF and PES membranes. At any biopolymer concentration, the PE membrane showed lower fouling potential

SC

518 than the PVDF membrane. On the basis of our previous study that quantified the affinity between carbohydrate-like

M AN U

519 substances and the surfaces of PVDF and PE membranes using atomic force microscopy, enormous interaction forces 520 were observed only for the PVDF membrane because it was probable that hydrogen bonding occurred between 521 fluorine in the PVDF and oxygen in the hydroxyl group (Yamamura et al. 2008). Similarly, in the present study, it

TE D

522 could be considered that the carbohydrate-like substances in the biopolymer of hydrophilic organic matter adsorbed 523 more on the surface of PVDF membrane than on the PE membrane by hydrogen bonding and caused irreversible loss

AC C

525

EP

524 of permeation.

33

SC

RI PT

ACCEPTED MANUSCRIPT

526

M AN U

527 Fig. 12 Correlation of biopolymer concentration with degree of physically irreversible fouling: degree of physically 528 irreversible fouling was determined by subtracting initial TMP value from final TMP value. 529

TE D

530 We should be aware of the fact that the feed water for the filtration experiments was not original natural water but the 531 fractionated water. In unaltered natural water, other factors (i.e., interactions among NOM fractions) and interactions

EP

532 between particular matter and NOM fractions may play a role for fouling development as well. But investigation of 533 these factors was outside the scope of the present paper. There is evidence in the literature that biopolymer

AC C

534 concentration in natural water is not necessarily linear correlated to non-back-washable irreversible fouling. This 535 indicates that, as far as irreversible fouling concerned, the concentration of biopolymers in water cannot be considered 536 as the determining factor. In the present study, we have clarified that the biopolymers in the HPI fractions have strong 537 correlation with the fouling potential of MF and UF membranes. Further investigations will be needed to evaluate the 538 contribution of the HPI biopolymer to the irreversible fouling quantitatively in the presence of other fractions or in the 539 actual condition. 34

540

ACCEPTED MANUSCRIPT

541 5. Conclusions 542 In this study, we compared the fouling potential of individual hydrophilic and hydrophobic fractions of NOM, which

RI PT

543 were isolated from four characteristic water sources, and identified the key components causing irreversible fouling in 544 MF and UF membranes. The conclusions are summarized as follows:

The key components causing irreversible fouling were in the hydrophilic fraction of the organic matter regardless

546

of the water source and membrane used.

547 2.

The major constituents causing irreversible fouling were biopolymers, which were composed of hydrophilic

548

organic matter including carbohydrates and proteins.

549 3.

Hydrophobic NOM did hardly cause any irreversible fouling.

550 4.

The extent of irreversible fouling caused by hydrophilic NOM was strongly affected by both membrane material

551

and NOM characteristics.

552 5.

The extent of irreversible fouling that occurred in a PVDF membrane was governed by the amount of

553

biopolymers in the hydrophilic NOM.

EP

AC C

554

TE D

M AN U

SC

545 1.

555 Acknowledgements

556 This work was supported by JSPS Grant-in-Aid (1852032) for JSPS Fellows. 557 558 References 559 Aiken, G.R., Mcknight, D.M., Thorn, K.A. and Thurman, E.M. (1992) Isolation of Hydrophilic 35

560 Organic-Acids from Water Using Nonionic Macroporous Resins. Org Geochem 18(4), 567-573. ACCEPTED MANUSCRIPT 561 AWWA Membrane Technology Research Comittee (2005) Recent advances and research needs in membrane 562 fouling. J Am Water Works Ass, 79-89.

RI PT

563 Chen, W., Westerhoff, P., Leenheer, J.A. and Booksh, K. (2003) Fluorescence excitation - Emission matrix 564 regional integration to quantify spectra for dissolved organic matter. Environ Sci Technol 37(24), 565 5701-5710.

SC

566 Cho, J., Amy, G. and Pellegrino, J. (2000) Membrane filtration of natural organic matter: factors and

M AN U

567 mechanisms affecting rejection and flux decline with charged ultrafiltration (UF) membrane. J Membrane 568 Sci 164(1-2), 89-110.

569 Cho, J.W., Amy, G. and Pellegrino, J. (1999) Membrane filtration of natural organic matter: Initial

571 Water Res 33(11), 2517-2526.

TE D

570 comparison of rejection and flux decline characteristics with ultrafiltration and nanofiltration membranes.

EP

572 Cho, J.W., Amy, G., Pellegrino, J. and Yoon, Y.M. (1998) Characterization of clean and natural organic

574 101-108.

AC C

573 matter (NOM) fouled NF and UF membranes, and foulants characterization. Desalination 118(1-3),

575 Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A. and Smith, F. (1956) Colorimetric method for 576 determination of sugars and related substances. Anal Chem 28, 350-356. 577 Her, N., Amy, G., Foss, D., Cho, J., Yoon, Y. and Kosenka, P. (2002) Optimization of method for detecting 578 and characterizing NOM by HPLC-size exclusion chromatography with UV and on-line DOC detection. 579 Environ Sci Technol 36(5), 1069-1076. 36

580 Her, N., Amy, G. and Jarusutthirak, C. (2000) MANUSCRIPT Seasonal variations of nanofiltration (NF); foulants ACCEPTED 581 identification and control. Desalination 132, 143-160. 582 Howe, K.J. and Clark, M.M. (2002) Fouling of microfiltration and ultrafiltration membranes by natural

RI PT

583 waters. Environ Sci Technol 36(16), 3571-3576. 584 Huang, H., Lee, N., Young, T., Gary, A., Lozier, J.C. and Jacangelo, J.G. (2007) Natural organic matter 585 fouling of low-pressure, hollow-fiber membranes: Effects of NOM source and hydrodynamic conditions.

SC

586 Water Res 41(17), 3823-3832.

M AN U

587 Jacangelo, J.G., Adham, S.S. and Laine, J.-M. (1995) Mechanism of Cryptosporidium, Giardia, and MS2 588 virus removal by MF and UF. J Am Water Works Ass 87(9).

589 Jermann, D., Pronk, W. and Boller, M. (2008) Mutual influences between natural organic matter and

TE D

590 inorganic particles and their combined effect on ultrafiltration membrane fouling. Environ Sci Technol 591 42(24), 9129-9136.

EP

592 Jermann, D., Pronk, W., Meylan, S. and Boller, M. (2007) Interplay of different NOM fouling mechanisms 593 during ultrafiltration for drinking water production. Water Res 41(8), 1713-1722.

AC C

594 Katsoufidou, I.S., Sioutopoulos, D.C., Yiantsios, S.G. and Karabelas, A.J. (2010) UF membrane fouling by 595 mixtures of humic acids and sodium alginate Fouling mechanisms and reversibility. Desalination 264(3), 596 220-227.

597 Katsoufidou, K., Yiantsios, S.G. and Karabelas, A.J. (2005) A study of ultrafiltration membrane fouling by 598 humic acids and flux recovery by backwashing: Experiments and modeling. J Membrane Sci 266(1-2), 599 40-50. 37

600 Kimura, K., Hane, Y., Watanabe, Y., Amy, G. and Ohkuma, N. (2004) Irreversible membrane fouling during ACCEPTED MANUSCRIPT 601 ultrafiltration of surface water. Water Res 38(14-15), 3431-3441. 602 Kimura, K., Yamamura, H. and Watanabe, Y. (2006) Irreversible fouling in MF/UF membranes caused by

RI PT

603 natural organic matters (NOMs) isolated from different origins. Separ Sci Technol 41(7), 1331-1344. 604 Krasner, S.W., Croue, J., Buffle, J. and Perdue, E.M. (1996) Three approaches for characterizing NOM. J 605 Am Water Works Ass 88(6), 66-79.

SC

606 Lee, N., Amy, G. and Croué, J.-P. (2006a) Low-pressure membrane (MF/UF) fouling associated with

M AN U

607 allochthonous versus autochthonous natural organic matter. Water Res 40(12), 2357-2368. 608 Lee, N., Amy, G. and Croue, J.P. (2006b) Low-pressure membrane (MF/UF) fouling associated with 609 allochthonous versus autochthonous natural organic matter. Water Res 40(12), 2357-2368.

TE D

610 Lee, N., Amy, G., Croue, J.P. and Buisson, H. (2004) Identification and understanding of fouling in 611 low-pressure membrane (MF/UF) filtration by natural organic matter (NOM). Water Res 38(20), 4511-4523.

EP

612 Leenheer, J.A. (2009) Systematic Approaches to Comprehensive Analyses of Natural Organic Matter. 613 Annals of Environmental Science 3, 1-130.

AC C

614 Li, Q.L. and Elimelech, M. (2004) Organic fouling and chemical cleaning of nanofiltration membranes: 615 Measurements and mechanisms. Environ Sci Technol 38(17), 4683-4693. 616 Ma, H.Z., Allen, H.E. and Yin, Y.J. (2001) Characterization of isolated fractions of dissolved organic matter 617 from natural waters and a wastewater effluent. Water Res 35(4), 985-996. 618 Malmrose, P., Lozier, J., Marie, J., Mickley, M., Reiss, R., Russell, J., Schaefer, J., Sethi, S. and Worley, J. 619 (2003) Committee Report: Residuals Management for Low-pressure Membranes. J Am Water Works Ass 38

620 95(6), 68.

ACCEPTED MANUSCRIPT

621 Matsuhiro, B., Torres, R., Zuniga, E.A., Aguirre, M.J., Mendoza, L. and Isaacs, M. (2009) 622 DETERMINATION OF LOW MOLECULAR WEIGHT CARBOHYDRATES IN CABERNET

RI PT

623 SAUVIGNON RED WINES. Journal of the Chilean Chemical Society 54(4), 405-407. 624 Schmitt, J. and Flemming, H.C. (1998) FTIR-spectroscopy in microbial and material analysis. Int Biodeter 625 Biodegr 41(1), 1-11.

M AN U

627 Surface and Ground Waters. Water Res 29(6), 1471-1477.

SC

626 Sun, L., Perdue, E.M. and Mccarthy, J.F. (1995) Using Reverse-Osmosis to Obtain Organic-Matter from

628 Tambo, N. and Kamei, T. (1998) Coagulation and flocculation on water quality matrix. Water Sci Technol 629 37(10), 31-41.

TE D

630 Tian, J.-y., Ernst, M., Cui, F. and Jekel, M. (2013a) Effect of different cations on UF membrane fouling by 631 NOM fractions. Chem Eng J 223(0), 547-555.

EP

632 Tian, J.Y., Ernst, M., Cui, F.Y. and Jekel, M. (2013b) Correlations of relevant membrane foulants with UF 633 membrane fouling in different waters. Water Res 47(3), 1218-1228.

AC C

634 WEF, A. (1998) Standard Methods for the Examination of Water And Wastewater, American Public Health 635 Association, Washington, DC.

636 Whistler, R.L. and Wolfrom, W.L. (1962) Methods in carbohydrate chemistry, Academic Press. 637 Xiao, F., Xiao, P., Zhang, W.J. and Wang, D.S. (2013) Identification of key factors affecting the organic 638 fouling on low-pressure ultrafiltration membranes. J Membrane Sci 447, 144-152. 639 Yamamura, H., Kimura, K., Okajima, T., Tokumoto, H. and Watanabe, Y. (2008) Affinity of functional 39

640 groups for membrane surfaces: implications for physically irreversible fouling. Environ Sci Technol 42(14), ACCEPTED MANUSCRIPT 641 5310-5315. 642 Yamamura, H., Kimura, K. and Watanabe, Y. (2007a) Mechanism involved in the evolution of physically

RI PT

643 irreversible fouling in microfiltration and ultrafiltration membranes used for drinking water treatment. 644 Environ Sci Technol 41(19), 6789-6794.

645 Yamamura, H., Okimoto, K., Kimura, K. and Watanabe, Y. (2007b) Influence of calcium on the evolution of

SC

646 irreversible fouling in microfiltration/ultrafiltration membranes. Journal of Water Supply: Research and

M AN U

647 Technology—AQUA 56(6–7), 425.

648 Yuan, W. and Zydney, A.L. (2000) Humic acid fouling during ultrafiltration. Environ Sci Technol 34(23), 649 5043-5050.

TE D

650 Zularisam, A.W., Ahmad, A., Sakinah, M., Ismail, A.F. and Matsuura, T. (2011) Role of natural organic 651 matter (NOM), colloidal particles, and solution chemistry on ultrafiltration performance. Sep Purif Technol

654

AC C

653

EP

652 78(2), 189-200.

40

Table 1

The relative abundances of carbon molecular descriptors obtained for four different drinking

ACCEPTED MANUSCRIPT

water sources and HPO/TPI/HPI composition of NOMs sampled from four different water sources.

C-COO

C-ar-O

C-ar-CH

C-anom

C-Al-O

C-OCH3

C-Al-CH

Ty-RW

3.6

13.7

0.0

13.9

6.7

25.1

6.6

30.2

In-LW

0.1

14.6

0.0

10.8

7.9

35.2

8.1

23.0

Ks-RW

4.0

6.7

0.0

11.1

5.4

28.4

8.9

35.5

Yd-RW

4.2

14.6

0.0

8.9

6.1

29.3

7.5

29.4

TPI-f

Ty-RW

60

20

In-LW

49

23

Ks-RW

50

20

Yd-RW

53

31

AC C

EP

TE D

M AN U

All the value listed in Table 1 has the unit of %.

HPI-f

SC

HPO-f

RI PT

C-C=O

20 28 30 16

Table 2 Fundamental characteristics of organic matter fractionated into hydrophobic and hydrophilic

ACCEPTED MANUSCRIPT

Hydrophobic fraction (HPO-f) (a)

(1/m/mg/L) Carbohydrate (mg-Glu/mg-C)

(d)

Yd-RW

(a)

Ty-RW

1.6

2.4

2.4

2.2

1.8

0.055

0.078

0.078

0.063

0.042

3.7

3.3

3.0

2.9

2.4

9.8

9.1

20.3

10.0

(b)

In-LW

1.7

45.5

(c)

Ks-RW

1.5

(d)

Yd-RW

1.7

0.025

0.025

0.024

1.5

1.8

1.4

27.2

46.5

SC

SUVA

Ks-RW

M AN U

(1/cm)

(c)

TE D

UV

In-LW

EP

(mg-C/L)

(b)

AC C

DOC

Ty-RW

Hydrophilic fraction (HPI-f)

RI PT

fractions

52.7

ACCEPTED MANUSCRIPT

Highlights Highlights The key components causing irreversible fouling were generally in the HPI fraction




The major irreversible foulants were biopolymers in HPI fraction




HPO fraction was not the sole contributor to irreversible fouling




Linear relationship was seen between potential of irreversible fouling and amount

RI PT




AC C

EP

TE D

M AN U

SC

of biopolymer

ACCEPTED MANUSCRIPT

Supporting Information for Hydrophilic fraction of natural organic matter causing irreversible fouling of microfiltration and ultrafiltration membranes

RI PT

Hiroshi Yamamura*, Kenji Okimoto, Katsuki Kimura, Yoshimasa Watanabe

Department of Integrated Science and Engineering for Sustainable Society,

M AN U

1 Pages, 1 Figures

SC

Chuo University, 1-12-27 Kasuga, Bunkyo-ku, Tokyo 112-8551, Japan

*Authors to whom correspondence should be addressed.

AC C

EP

TE D

H. Yamamura: telephone: +81 (3) 3817-7257; Fax: +81 (3) 3817-7257; email: [email protected]

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Figure S1: An example of typical time-course change of TMP