Accepted Manuscript

The feasibility of nanofiltration membrane bioreactor (NF-MBR)+reverse osmosis

(RO) process for water reclamation: Comparison with ultrafiltration membrane
bioreactor (UF-MBR)+RO process

Ming Feng Tay, Chang Liu, Emile R. Cornelissen, Bing Wu, Tzyy Haur Chong

PII: S0043-1354(17)30929-6
DOI: 10.1016/j.watres.2017.11.013
Reference: WR 13343

To appear in: Water Research

Received Date: 20 August 2017

Revised Date: 30 October 2017
Accepted Date: 5 November 2017

Please cite this article as: Tay, M.F., Liu, C., Cornelissen, E.R., Wu, B., Chong, T.H., The feasibility of
nanofiltration membrane bioreactor (NF-MBR)+reverse osmosis (RO) process for water reclamation:
Comparison with ultrafiltration membrane bioreactor (UF-MBR)+RO process, Water Research (2017),
doi: 10.1016/j.watres.2017.11.013.

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

At 0% Recovery
7
NF Fed with UF-MBR permeate,

PT
dTMP/dt ~0.11 bar/day
Feed Permeate
Ca2+

TMP(bar)
6

RI
Dissolved Fed with NF-MBR permeate,

SC
organics 5 dTMP/dt ~0.03 bar/day

Mg2+

U Reverse Osmosis (RO)

(a)
4

AN
0 10 20 30
NF-MBR Time (day)

M
D
At 90% Recovery

TE
10
Fed with UF-MBR permeate,
UF CP
dTMP/dt ~0.30 ba r/da y
EP
Feed Permeate 8

TMP(bar)
C

Fed with NF-MBR permeate,

dTMP/dt ~0.09 bar/da y
6
AC

Fouling
(b)
4
0 5 10 15
Time (day)
UF-MBR
 ACCEPTED MANUSCRIPT

2 The Feasibility of Nanofiltration Membrane Bioreactor (NF-MBR)+Reverse Osmosis (RO)

3 Process for Water Reclamation: Comparison with Ultrafiltration Membrane Bioreactor (UF-

4 MBR)+RO Process

PT
5

RI
6 Ming Feng Tay a,b, Chang Liub, Emile R. Cornelissenb,c, Bing Wu b*, Tzyy Haur Chong a,b*

SC
7

U
a
8 School of Civil and Environmental Engineering, Nanyang Technological University, Singapore
AN
9 639798, Singapore

b
10 Singapore Membrane Technology Centre, Nanyang Environment and Water Research Institute,
M

11 Nanyang Technological University, Singapore 637141, Singapore

D

c
12 KWR Watercycle Research Institute, 3433 PE Nieuwegein, Netherlands
TE

13

14
EP

15 * Corresponding authors:
C

16 Tzyy Haur Chong, Email: [email protected], Tel: +65 6513 8126, Fax: +65 6791 0676
AC

17 Bing Wu, Email: [email protected]

18

19

20

1
 ACCEPTED MANUSCRIPT

21 Abstract

22 This study examines the feasibility of a novel nanofiltration membrane bioreactor (NF-MBR)

23 followed by reverse osmosis (RO) process for water reclamation at 90% recovery and using an

24 ultrafiltration MBR (UF-MBR)+RO as baseline for comparison. Both MBRs adopted the same

PT
25 external hollow fiber membrane configurations and operating conditions. The collected permeates

26 of the MBRs were subsequently fed to the respective RO systems. The results showed that the NF-

RI
27 MBR (operated at a constant flux of 10 L/m2h) achieved superior MBR permeate quality due to

SC
28 enhanced biodegradation and high rejection capacity of the NF membrane, leading to lower RO

29 fouling rates (~3.3 times) as compared to the UF-MBR. Further analysis indicated that the cake

U
30 layer fouling that caused the cake-enhanced osmotic pressure (CEOP) effect contributed
AN
31 predominantly to the transmembrane pressure (TMP) increase in the NF-MBR, while irreversible

32 pore fouling was the major reason for UF membrane fouling. Furthermore, it was found that the
M

33 biopolymers (i.e., organics with MW >10 kDa) were the main components present in the foulants of
D

34 the NF/UF membranes and RO membranes. The analysis indicated that the NF-MBR+RO system at
TE

35 recovery of 90% has comparable energy consumption as the UF-MBR+RO system at recovery of

36 75%. Our findings proved the feasibility of the NF-MBR+RO for water reclamation at a high
EP

37 recovery rate.

38 Key words: Energy consumption; Dissolved organic substances; Membrane fouling;

C

39 Nanofiltration; Reverse osmosis; Water reclamation

AC

40

41

42

2
 ACCEPTED MANUSCRIPT

43 1. Introduction

44 Water scarcity resulted from global population growth, rapid urbanization and growing economic

45 activity is one of the major problems in the 21st century. Besides surface water, groundwater, and

46 seawater, water reclamation has become an important alternative for indirect or direct potable use

PT
47 (Bartels et al., 2005).

RI
48 Universally, water reclamation refers to the treatment of used or wastewater to the quality suitable

49 for water recovery by reverse osmosis (RO) process. For example, the NEWater concept in

SC
50 Singapore involves the pretreatment of effluent from conventional activated sludge (AS) process by

51 a low-pressure membrane process such as microfiltration/ultrafiltration (MF/UF) prior to RO (Qin

52
U
et al., 2006). Alternatively, the effluent from a membrane bioreactor (MBR) can be directly used as
AN
53 feed water to RO (Chen et al., 2012; Qin et al., 2006). MBRs have become a more viable option for
M

54 wastewater treatment compared to AS process due to its advantages in terms of better effluent

55 quality, smaller footprint, better disinfection, and reduced sludge production (Yamamoto, 2001). In
D

56 addition, the successful integration of MBR and RO demonstrated a more consistent removal
TE

57 performance and lower RO fouling rates compared to the AS+MF/UF+RO process for water

58 reclamation (Qin et al., 2006).

EP

59 Typically, in wastewater reclamation processes, the recovery levels of RO are 75 to 85% using two
C

60 or more stages of RO in series (Wetterau et al., 2011). Although promising technologies are
AC

61 available to achieve over 90% of recovery, such as semi-batch RO process or known as closed

62 circuit desalination (CCD) (Septon and Efraty, 2016; Warsinger et al., 2016; Werber et al., 2017),

63 energy-efficient reverse osmosis (EERO) (Chong Loo and Krantz, 2015; Chong Loo Fane et al.,

64 2015), the practical recovery level is limited by membrane fouling, including inorganic scaling such

65 as calcium phosphate, calcium carbonate and silica, organic fouling and biofouling (Antony et al.,

3
 ACCEPTED MANUSCRIPT

66 2011; Kent et al., 2011; Moreno et al., 2013; Ridgway et al., 1984; Zhao et al., 2010). Usually

67 antiscalants or disinfecting agents are added to RO feed water to reduce RO fouling, however, some

68 antiscalants could enhance the biofouling potential in RO systems (Antony et al., 2011) and

69 disinfecting agents may eventually cause the degradation of membranes in long run (Antony et al.,

PT
70 2010). Emerging technology such as flow reversal (FR) technology is an attractive option for scale

71 control which works by periodically switching the flow direction in the RO pressure vessel arrays to

RI
72 disturb the kinetics of nucleation/crystallization (i.e. scaling) process (Lu et al., 2012). It has been

73 demonstrated to be effective for calcium sulfate and calcium carbonate scaling, but its efficiency on

SC
74 other types of fouling is not well established.

75
U
One avenue to prevent RO membrane fouling is to improve the quality of RO feed water. Despite
AN
76 the use of MF/UF or MF/UF-MBR, dissolved organic carbons (DOCs) and scale forming

77 solutes/ions can easily pass through the membrane owing to the relatively large pore sizes of
M

78 MF/UF membranes. In particular, conventional MBR has its own limitations in removing a wide
D

79 range of trace organic carbons (TrOCs) such as pesticides, endocrine disrupting chemicals (EDCs),
TE

80 hormones and pharmaceutically active compounds (PhACs) (Tadkaew et al., 2011). Several studies

81 have found that hydrophobic TrOCs can be adsorbed to the biomass, resulting in longer retention
EP

82 time in the bioreactor, thus increasing the removal efficiency. In contrast, hydrophilic TrOCs such

83 as carbamazepine and diclofenac were found to be persistent in both AS and MBR process
C

84 (Radjenović et al., 2009; Tadkaew et al., 2011; Wijekoon et al., 2013). Furthermore, these
AC

85 compounds are almost not retained by MF/UF membranes.

86 To achieve high quality permeate, the integration of high retention membrane separation processes,

87 such as membrane distillation (MD), forward osmosis (FO) and nanofiltration (NF) with biological

88 reactor to form MD-MBR, FO-MBR (known as OMBR) and NF-MBR configurations, respectively,

4
 ACCEPTED MANUSCRIPT

89 has been explored in recent years (Choi et al., 2002; Cornelissen et al., 2008; Goh et al., 2013; Luo

90 et al., 2015; Qiu et al., 2016). The key differences between high retention MBRs and conventional

91 MBRs are summarized in Table 1.

92 Insert Table 1

PT
93 Such high retention MBRs have enhanced biodegradation of refractory organics due to prolonged

94 residence time as well as effective rejection of inorganic compounds (i.e. Ca2+). However, there

RI
95 remain several technical challenges in high retention MBRs, such as salt accumulation, low

96 permeate flux (i.e. < 10 L/m2 h), and membrane stability (W. C. L. Lay et al., 2010; Luo et al., 2016;

SC
97 Luo et al., 2014). Specifically, in the FO-MBRs, the setbacks include (i) reversed salt diffusion

U
98 effect that results in the loss of draw solutes and requires constant topping up of draw solutes to
AN
99 maintain the flux; (ii) salinity increase in the bioreactor due to salt accumulation and reverse salt

100 diffusion as well as concentration decrease of the draw solution due to solute diffusion would
M

101 reduce the net driving force for water permeation, thus causing the flux to decline over time; (iii)

102 elevated salts could increase the soluble microbial products (SMP) and extracellular polymeric
D

103 substances (EPS) in the mixed liquor that cause greater membrane fouling (Luo et al., 2017); (iv)
TE

104 requires additional steps (i.e. RO or other processes) to separate the draw solution and water. More

105 importantly, the combined FO-MBR and RO process cannot reduce the energy requirement for
EP

106 water production compare to the conventional MBR+RO process due to the trade-off between FO
C

107 and draw solution recovery processes (Shaffer et al., 2015). While in MD-MBR, the requirement of
AC

108 waste heat and membrane wetting are still the bottlenecks.

109 Compared to MD-MBRs and FO-MBRs, NF-MBRs are more feasible. This is attributed to the facts

110 that (i) NF is a pressure driven process similar to MF/UF, which is less complicated compared to

111 FO or MD; (ii) NF has lower monovalent salt rejection, thus resulting in a lower salt accumulation

112 in the bioreactor. The maximum salt concentration factor can be controlled by regulating the ratio of

5
 ACCEPTED MANUSCRIPT

113 sludge retention time/hydraulic retention time (SRT/HRT, refer to Figure S1 in the Supplementary

114 Data for details) (W. C. Lay et al., 2010); (iii) NF application for surface and groundwater treatment

115 is rather mature with a large number of full-scale installations. However, it shall be noted that the

116 NF-MBR for wastewater treatment is still in its infancy. In previous attempts on NF-MBRs, an

PT
117 extremely low flux of 0.042 L/m2h or equivalent to 0.5 L/m2h·bar was attained with a commercial

118 cellulose acetate NF (Choi Fukushi et al., 2007; Choi et al., 2002; Choi et al., 2006; Choi Lee et al.,

RI
119 2007). The low permeability or high operating pressure required to obtain a decent flux has made

120 the NF-MBR less attractive. However, the novel low-pressure hollow fiber NF membrane,

SC
121 previously developed by the Singapore Membrane Technology Centre (SMTC) for water softening

U
122 application, may be applied to overcome this challenge. The novel NF membrane possesses a
AN
123 positively charged thin film selective layer with a pure water permeability of ~ 17 L/m2h·bar,

124 molecular weight cut off (MWCO) of < 500 Da, and rejections of Mg2+ and Ca2+ ~ 90% and Na+ <
M

125 15% (Liu et al., 2015; Rajabzadeh et al., 2014). Most of the commercial membranes in the market

126 (as summarized in Table S1 in the Supplementary Data) require operating pressures of > 5 bar, but
D

127 our membrane only requires < 2 bar to achieve this performance. The very low rejection of Na+ ion
TE

128 while maintaining high rejection of the divalent ions is the key to achieve such a low operating

129 pressure condition.

EP

130 In this study, we aim to develop a high-flux NF-MBR (i.e., 10 L/m2h) using the SMTC’s in-house
C

131 fabricated low-pressure hollow fiber NF membrane (with inside-out configuration), and integrate
AC

132 the NF-MBR with a RO process for high recovery water reclamation (i.e., 90% of recovery). The

133 main objectives of this study are (1) to evaluate the performance of an NF-MBR+RO for high

134 recovery water reclamation, (2) to illustrate the membrane fouling mechanisms and examine foulant

135 characteristics in the NF-MBR and RO membranes, and (3) to estimate the energy consumption of

136 the NF-MBR+RO at different recovery levels. An UF-MBR+RO system was used as a baseline for

6
 ACCEPTED MANUSCRIPT

137 comparison. This study provides meaningful information for further improvement and scale-up of

138 the NF-MBR+RO system.

139 2. Materials and methods

140 2.1. UF-MBR, NF-MBR and RO set-ups and operations

PT
141 Two identical lab-scale moving bed biofilm membrane bioreactors (hereinafter referred as MBR)

RI
142 employing side-stream ultrafiltration (defined as UF-MBR) and nanofiltration (defined as NF-

SC
143 MBR) membrane modules respectively were set up as shown in Figure S2 (Supplementary Data).

144 Each bioreactor had a total effective working volume of 7.0 L, and filled with 117 pieces of K3 bio-

U
145 carriers (i.e., filling ratio of 10%). The dimension of a bio-carrier is 12 × Ø 25 (mm, W × D) with a
AN
146 surface area of 500 m2/m3 (Ping Yin Chemical Packing, China). No activated sludge was inoculated

147 in the bioreactors. The aeration rate of bioreactors was set at 3 L/min. The sludge retention time
M

148 (SRT), hydraulic retention time (HRT), pH and temperature of bioreactors were kept at 30 d (i.e.,

149 0.233 L of sludge was discharged daily, accounting for only 3% of the feed flow, so the recovery of
D

150 NF-MBR was assumed to be ~100%), 22 h, 6.8 ± 0.5, 20 ± 0.5oC, respectively. The bioreactors
TE

151 were continuously with municipal wastewater collected from Ulu Pandan wastewater treatment
EP

152 plant of Singapore (after sieving with a 1 mm opening mesh). The permeate flux of both MBRs

153 was maintained at 10 L/m2h via a mass flow controller installed on the permeate stream. The
C

154 pressures of feed, retentate and permeate were recorded by digital pressure transducers (Ashcroft,
AC

155 USA) connected to a data logging system (LabVIEW, National Instrument, USA). The operating

156 parameters of UF-MBR and NF-MBR and the characteristics of the municipal wastewater are

157 summarized in Tables 2 and 3, respectively.

158 Insert Table 2

7
 ACCEPTED MANUSCRIPT

159 Insert Table 3

160 In this study, SMTC’s in-house fabricated polyethersulfone (PES) hollow fiber UF membrane and

161 low-pressure hollow fiber NF membrane with PES as substrate were used and their characteristics

162 are summarized in Table 4. Each membrane module consisted of 25 hollow fibers (0.46 m in length;

PT
163 a total effective area of 0.031 m2) and filtration was performed in an inside (lumen)-out (shell)

164 configuration. Periodic physical membrane cleaning was performed when the transmembrane

RI
165 pressure (TMP) reached the pre-set upper limit due to membrane fouling. After flushing the lumen

SC
166 side of the hollow fiber membrane with tap water at 500 mL/min for 2 min, backwashing was

167 performed (15 min) by using distilled water at a pressure of 0.5 and 2 bar for UF and NF

U
168 membrane, respectively. When the desired TMP was not achieved after membrane cleaning, the
AN
169 fouled membrane module was replaced with a new module. In the NF-MBR, three new NF

170 membrane modules were used in series, hereinafter, defined as NF1, NF2, and NF3; while in the
M

171 UF-MBR, two new UF membrane modules were used in series, hereinafter defined as UF1 and
D

172 UF2.
TE

173 Insert Table 4

174 Two identical RO systems were used to study the fouling potential of the permeates from the two
EP

175 MBRs. The details of the RO system (Figure S3, Supplementary Data) were described previously
C

176 (Suwarno et al., 2016; Wu et al., 2013). Briefly, the RO setup comprised of a feed tank, a high-
AC

177 pressure pump (Hydra Cell D-03-S, USA), a cooling device equipped with temperature controller

178 (PolyScience 9106, USA), a flat sheet RO test cell, a back pressure regulator (SS-4R3A, Swagelok),

179 a mass flow controller (LIQUI-FLOW L30, Bronkhorst, Netherlands), pressure transducers

180 (EUTECH Instruments, USA), and conductivity meters (EUTECH Instruments, USA). The RO

181 membrane (BW-30, DOW FilmTec, USA) with an effective membrane area of 0.0186 m2 was

8
 ACCEPTED MANUSCRIPT

182 employed in the RO test cell. The crossflow rate of the RO system was maintained at 18 L/h (i.e.,

183 equivalent to a crossflow velocity of 0.10 m/s) and the permeate flux was kept at 20 L/m2h by the

184 mass flow controller. In this study, the RO concentrate and RO permeate were recirculated to feed

185 tank and the test solution in the feed tanks was replenished daily.

PT
186 2.2. Analytical measurement

RI
187 2.2.1. Water quality parameters

SC
188 The feed water, mixed liquor and permeate of the MBRs were sampled twice weekly for analysis.

189 The supernatant of the feed water and mixed liquor was obtained by centrifuging for 10 min at 4000

U
190 ×g and then filtering through 0.45 µm membrane filter before analysis. Total organic carbon (TOC)
AN
191 and total nitrogen (TN) were determined by a TOC/TN-V analyzer (Shimadzu, Japan). Chemical

192 oxygen demand (COD), ammonia, nitrite, nitrate and total phosphorus (TP) were measured using
M

193 the colorimetric method with a spectrophotometer (DR 3900, HACH, USA). Concentrations of

194 cations (i.e. Ca2+, Mg2+) were determined by an inductively coupled plasma optical emission
D

195 spectrometry (ICP-OES, Optima 8000, Perkin Elmer, USA). The extracellular polymeric substances
TE

196 (EPS) content was determined by analyzing the concentrations of polysaccharides and proteins in
EP

197 the MBR permeate. The polysaccharides concentration was determined according to the phenol-

198 sulfuric acid method with glucose as a standard and absorbance measurement at 490 nm using a
C

199 spectrometer (HACH, USA) (Dubois et al., 1956). The protein concentration was determined by
AC

200 Bradford method with bovine serum albumin (BSA) as standard and absorbance measurement at

201 595 nm (Bradford, 1976).

202 Liquid chromatography-organic carbon detector (LC-OCD) analyzer (LC-OCD Model 8, DOC-

203 LABOR, Germany), a size-exclusion chromatography system coupled with organic carbon

204 detection and organic nitrogen detection was used to quantify the dissolved organic fractions in the
9
 ACCEPTED MANUSCRIPT

205 sample. The organic matter was separated into major fractions based on their molecular weight and

206 assign them into different fraction group such as biopolymers (typically MW > 10 kDa), humic

207 substances (MW ~ 500 Da), building blocks (MW ~ 300 Da), low molecular weight organics acids

208 (MW ~ 200–250 Da), and neutrals (MW < 200 Da). Details of the LC-OCD system and analysis

PT
209 were described in the literature (Huber et al., 2011).

210 2.2.2. Total biomass concentration

RI
211 Total biomass concentration in the bioreactor was defined as the sum of mixed liquor suspended

SC
212 solids (MLSS) and total attached growth biomass on the bio-carriers. The attached growth biomass

U
213 on the bio-carrier was extracted by sonication (10 min) and vortex (1 min). The total attached
AN
214 growth biomass was estimated by multiplying the attached growth biomass extracted per bio-carrier

215 and the total number of bio-carriers in the bioreactor. The MLSS and biomass concentration were
M

216 measured according to Standard Methods (American Public Health et al., 2005).
D

217 2.2.3. Membrane autopsy

TE

218 At the end of the membrane filtration tests, membrane modules were removed from the setup for

219 autopsy studies. The foulants were detached from the membrane surfaces by soaking the
EP

220 membranes in 30 mL of Milli-Q water and then ultrasonication for 30 min and vortex for 1 min

221 (Jeong et al., 2013). The foulant solution was centrifuged for 10 min at 4000 ×g and then filtered
C

222 through 0.45 µm membrane filter, and the filtrates (i.e., soluble foulants) were used for EPS and
AC

223 LC-OCD analysis.

224 3. Results and Discussion

10
 ACCEPTED MANUSCRIPT

225 3.1. MBR performance

226 3.1.1. Overall reactor performance

227 Both NF-MBR and UF-MBR were operated for 60-80 days. The accumulation of dissolved

PT
228 organics and divalent ions was observed in the NF-MBR (Figure 1). For example, after 60 days of

229 operation (i.e., both reactors tended to be stable based on the constant total organic removals as

RI
230 shown in Figure 2), the concentrations of DOC (~116 mg/L), conductivity (~ 2212 µs/cm), Ca2+

231 (~267 mg/L), Mg2+ (~108 mg/L) and PO43- (~71 mg/L) in the NF-MBR were approximately 5.3,

SC
232 2.4, 8.3, 13.5, and 2.8 times of that in the UF-MBR (DOC at ~22 mg/L, conductivity at ~ 930

µs/cm, Ca2+ at ~32 mg/L, Mg2+ at ~8 mg/L, and PO43-at ~25 mg/L), respectively. The concentration

U
233

of Ca2+, Mg2+, and PO43- increased gradually and reached an equilibrium level at around day 10 due
AN
234

235 to high rejection properties of the NF membranes (NF1). On day 38 a new NF membrane (NF2)
M

236 resulted in a further increase of conductivity, Ca2+, Mg2+, and phosphate concentrations, and no

237 equilibrium was reached before day 52. On day 52, a new NF membrane with higher rejection
D

238 properties, NF3, was used. This has resulted in a further build-up of Ca2+, Mg2+, and PO43-
TE

239 concentrations in the NF-MBR. The retention and accumulation of slowly-biodegradable and non-

240 biodegradable organic compounds were inevitable in the high retention NF-MBRs (Figure 1). This
EP

241 finding was similar to the reported studies in the NF-MBR, FO-MBR and MD-MBR (Choi Lee et

242 al., 2007; Phattaranawik et al., 2008; Qiu et al., 2016).

C
AC

243 Insert Figure 1

244 The total attached growth biomass on the bio-carriers in NF-MBR and UF-MBR were maintained at

245 399±96 and 407±79 mg/L, respectively; whereas the MLSS in the reactors were kept at 503±69 and

246 513±96 mg/L, respectively (Figure S4 and S5, Supplementary Data). Despite the total biomass was

11
 ACCEPTED MANUSCRIPT

247 relatively low (~1 g/L), the NF-MBR and UF-MBR achieved excellent biodegradation of DOC at

248 80.5±8.2% and 76.7±9.5%, respectively (detailed mass balance calculation was presented in the

249 Supplementary Data). This implied that the elevated salts level (i.e., Ca2+, Mg2+) in the NF-MBR

250 did not show significant negative impact on the biodegradation of organics; instead slightly greater

PT
251 biodegradation rate was observed (Figure S6, Supplementary Data). This could be attributed to two

252 factors. First, previous studies had pointed out the microorganisms derived from municipal

RI
253 wastewater could tolerate high salt concentration of up to 10 g/L without acclimation (Jang et al.,

254 2013). Interestingly, the presence of additional salt (<10 g/L) could improve the biodegradability of

SC
255 organic carbon due to stimulatory effect on microorganisms (Ng et al., 2005). Second, a longer

U
256 retention time of the slowly-biodegradable organic substances could enhance their degradation.
AN
257 Figures 2a and 2b show the COD and DOC levels in the permeates and removals of both MBRs,

258 respectively. The NF-MBR produced the permeate containing 0.5 to 2.5 mg/L of DOC, much lower
M

259 than the permeate of UF-MBR, i.e., in the range between 7.0 to 9.5 mg/L. The overall COD and
D

260 DOC removal rates in the NF-MBR were 99.6±0.8% and 97.5±1.8%, significantly higher than those
TE

261 in the UF-MBR (93.4±3.0% and 81.3±6.9%, respectively).

262 Insert Figure 2

EP

263 In addition, the permeate qualities of NF-MBR and UF-MBR were compared as shown in Figure 3.
C

264 In general, due to better retention of the NF membrane, the permeate of NF-MBR had lower
AC

265 concentrations of proteins (~12 times lower), polysaccharides (~1.6 times lower), Ca2+ (~2.3 times

266 lower), Mg2+ (~5.2 times lower), and PO43- (~1.4 times lower) compared to that of the UF-MBR.

267 The improved quality of the NF-MBR permeate (i.e., RO feed water) is expected to contribute to

268 less fouling in the subsequent RO process.

12
 ACCEPTED MANUSCRIPT

269 Insert Figure 3

270 3.1.2. TMP profiles of MBRs

271 Figures 4a and 4b illustrate the TMP profiles of the NF-MBR and UF-MBR, respectively, under a

PT
272 continuous filtration mode with periodic water flushing and backwashing. In the NF-MBR, from

273 Day 0 to 38, the periodical cleaning of NF membrane (i.e., NF1) was performed when the TMP

RI
274 reached to 1.5 times of the initial TMP. During the early stage of operation (Day 0-12), a gradual

275 increase in TMP was observed. This could be attributed to (i) an increase in osmotic pressure due to

SC
276 the build-up of predominantly rejected inorganic divalent ions (Ca2+, Mg2+, PO43-, etc.) and organic

U
277 molecules in the bioreactor (Figure 1); and (ii) the cake formation due to the deposition of
AN
278 suspended and colloidal particles, soluble microbial products and microorganisms on the membrane

279 surface that increased the overall resistance and possibly the cake enhanced osmotic pressure
M

280 (CEOP) effect (Tang et al., 2011).

D

281 Insert Figure 4

TE

282 In detail, after first backwashing cleaning of NF1 on Day 7, the TMP could not return to the original

283 value of ~100 kPa, mainly due to a continuous increase in osmotic pressure (i.e., as indicated by the
EP

284 increase of conductivity, Figure 1). After an additional 6 days of filtration, the second backwashing

285 was performed when the TMP reached ~220 kPa, the TMP could not return to its original value and
C

286 was even higher than the value after first cleaning. In the third cycle (Day 13-19), the TMP rise was
AC

287 more notable, hence, pure water flushing was performed to remove the loosely bound cake layer

288 before backwashing. Nevertheless, a rapid fouling rate of ~ 150 kPa/day was observed in the 4th

289 cycle (Day 20-22) and the TMP reached the upper limit of ~ 350 kPa. Meanwhile, the constant flux

290 could not be attained and dropped from 10 L/m2h to 8.5 L/m2h (data not shown). The duration of 2

13
 ACCEPTED MANUSCRIPT

291 subsequent cycles became shorter (3-4 days). Also, it was observed that when the TMP exceeded

292 ~250 kPa, the fouling rates increased exponentially (for the last 3 cycles). This could be attributed

293 to the formation of a more compressed cake layer at a high operating pressure of ~ 400 kPa

294 (Teychene et al., 2011), which was not effectively removed by the employed cleaning method.

PT
295 A new cleaning protocol in an attempt to prolong the NF membrane lifespan was adopted in the

296 operation of NF2. It was hypothesized that more frequent hydraulic physical cleaning was

RI
297 beneficial in removing the limited cake layer (i.e., the TMP rise was low) by shear and uplift forces.

SC
298 To prevent the cake compression or collapse under a high-pressure condition, a combination of

299 water flushing and backwashing (i.e., pure water flushing for 2 min before backwashing for 15 min)

300
U
was performed when the TMP reached an upper limit of 200 kPa. After implementing the new
AN
301 cleaning protocol, the cycle duration was reduced to around ~1.5-2.5 days. Unexpectedly, a sudden

302 jump in TMP to ~320 kPa at Day 12 occurred due to unattended operation over the weekend. Thus,
M

303 the membrane was replaced. In NF3, frequent hydraulic physical cleaning was performed, i.e., cycle
D

304 duration of 1 to 3 days (when TMP reached an upper limit of 200 kPa). The lifespan of NF3 could
TE

305 be prolonged to ~26 days without any chemical cleaning.

306 On the other hand, in the UF-MBR with a periodically water flushing and backwashing, the TMP
EP

307 increased from an initial value of ~10 kPa to 40 kPa within 25 days during start-up period (for UF1)

308 and ~ 37 days during the stabilization period (For UF2). The TMP profile of UF2 was compared to
C
AC

309 the TMP profile of NF3 (from 150 kPa to ~250 kPa in 26 days). In general, the UF membrane was

310 cleaned every 3-5 days, which had lower cleaning frequency compared to the NF membrane.

311 3.1.3. Fouling mechanism

14
 ACCEPTED MANUSCRIPT

312 To further clarify the membrane fouling mechanisms in the NF-MBR and UF-MBR, the membrane

313 pure water permeability (PWP) was measured (at a fixed pressure of 100 kPa) after each physical

314 cleaning cycle (Figure 5a). A gradual decline of the NF membrane PWP over time was noticed,

315 which means predominance of cake layer fouling and slow build-up of irreversible fouling. On the

PT
316 other hand, the UF membrane experienced (i) a dramatic drop in the PWP in the first cycle, and (ii)

317 a continuing loss in the PWP in the following cycles. Overall, the irreversible fouling has caused

RI
318 90% loss in PWP after 7 cycles. These observations implied that the membrane fouling mechanisms

319 in the NF-MBR and UF-MBR appeared to be dissimilar. The fouling experienced by the NF-MBR

SC
320 was highly reversible by physical cleaning due to the formation of cake layers on the NF membrane

U
321 (it is noted that the NF membrane with pore size distribution ranging from 200-300 Da). In contrast,
AN
322 irreversible fouling was predominant in the UF-MBR as internal pore blocking and pore plugging

323 were likely to take place in the UF membrane (Fu et al., 2016). Despite the NF-MBR had greater
M

324 amount of potential foulants than UF-MBR (Table 5) and they had different fouling mechanisms,

325 both membranes exhibited approximately similar final value of irreversible fouling resistance (i.e.,
D

326 Rirr = Rafter physical cleaning – Rmembrane) as shown in Figure 5b. The results suggest that both membranes
TE

327 may have very similar amount of attached foulants that cannot be easily removed by the employed

328 physical cleaning. Upon examining the TMP and resistance data, the additional pressure (i.e.
EP

329 ∆TMP, Figure S7 in the Supplementary Data) required in the UF-MBR was mainly to overcome the
C

330 increase in irreversible fouling. Conversely, the main contribution for additional pressure in NF-
AC

331 MBR was due to the CEOP effect.

332 Insert Figure 5

333 Nevertheless, effective periodically physical cleaning strategies (backwashing and flushing) could

334 help extend the operation lifespan of NF membrane by preventing the collapse and compression of

15
 ACCEPTED MANUSCRIPT

335 cake layer, which may eventually lead to irreversible fouling. This finding allows us to further

336 improve NF-MBR performance by optimizing physical fouling control methods, such as employing

337 two-phase flow cleaning protocol using periodical air and water cleaning (Wibisono et al., 2014).

338 3.1.4. Organic foulant analysis

PT
339 Table 5 describes the organic fractions in the feed, mixed liquor, permeate of both MBRs. The

RI
340 contribution ratio of each organic fraction to the DOC was presented in Figure 6. In general, the

341 biopolymer fraction (typically MW >10 kDa), accounting for ~24 % of the total DOC in the feed,

SC
342 was removed effectively in both MBRs (>98.5%). The UF-MBR removed less humic substances

U
343 (59%), building blocks (70%), and LMWs (76%) than the NF-MBR (all >95%). It was noted that
AN
344 humic substances (MW ~500 Da), building blocks (MW ~300 Da), LMWs (MW ~200-250 Da) had

345 smaller sizes than the UF membrane pore size (~50 kDa). Partial removal of these substances by the
M

346 UF membrane could be due to the retention by cake layer formed on the membrane or adsorption in

347 the membrane pores (Aslam et al., 2017). Furthermore, humic substances are the second abundant
D

348 organic fractions (~19%) in the NF foulants. The humic substances are relatively stable and could
TE

349 chelate with calcium ions to form calcium enhanced organic fouling layer. Such fouling tends to be

350 more difficult to be removed by physical cleaning (Sutzkover-Gutman et al., 2010). The high
EP

351 rejection of organic substances by the NF membrane resulted in their accumulations in the NF-

352 MBR with concentration factors of 2.1, 6.0, 4.9, and 2.5 for biopolymers, humic substances,
C
AC

353 building blocks, and LMWs, respectively. The difference in the concentration factors could be due

354 to the difference in biodegradability of these compounds. For example, biopolymers and LMWs

355 were more easily biologically degraded than humic substances and building blocks. This finding

356 was similar to the observation in the high retention FO-MBR (Qiu et al., 2016).

357 Insert Table 5

16
 ACCEPTED MANUSCRIPT

358 Insert Figure 6

359 3.1.5 Correlation between organic fractions in mixed liquor and membrane fouling

360 In the UF-MBR, majority of the foulants were biopolymers (75.7% or 159.2 mg/m2) due to the high

PT
361 retention property of the UF membrane towards large MW biopolymers. However, it shall be noted

362 that despite the smaller amount of the low MW fractions, i.e., building blocks and LMW, in the

RI
363 foulant, the impact of these components towards the overall resistance could be more critical than

364 the biopolymer fraction due to the effect of pore plugging that resulted in irreversible fouling (Chen

SC
365 et al., 2016). On the other hand, in the NF-MBR, the foulants were comprised of 56.6%

biopolymers (128.1 mg/m2), 19.2% humic substances (43.4 mg/m2), 7.6% building blocks (17.2

U
366

mg/m2), and 16.7% LMWs (37.8 mg/m2). In addition, by comparing the ratio of organic fractions in
AN
367

368 the mixed liquor and foulants, it was noticed that the NF membrane was more prone to fouling
M

369 caused by biopolymers than other organic fractions. Furthermore, both extracted foulants from the

370 UF2 and NF3 membranes have comparable amount of DOC and organic fractions, which was
D

371 consistent with the fouling resistance tendency (Figure 5).

TE

372 3.2. RO performance

EP

373 3.2.1 TMP Profiles of ROs

C

374 After about 20 days of operation of both MBRs, the permeates of UF-MBR and NF-MBR were
AC

375 collected and directly fed to the respective RO systems to assess the RO performance. At recovery

376 of ~ 0% (i.e., the permeate was fully returned to the RO feed tank), the permeate of UF-MBR

377 caused more severe fouling on the RO membrane (~3.7 times higher) compared to the permeate of

378 NF-MBR (Figure 7a). The higher fouling rate observed in the RO system fed with UF-MBR

379 permeate was corroborated by the presence of higher amount of soluble organic compounds (Figure

17
 ACCEPTED MANUSCRIPT

380 3 and Table 5). This could result in more organic fouling by adsorption of organics onto the

381 membrane surface (i.e., formation of conditioning film), which subsequently initiated the bacterial

382 deposition (Suwarno et al., 2016). Furthermore, these organic compounds could serve as the carbon

383 sources to promote the biofilm formation on the RO membrane.

PT
384 In another attempt to simulate the membrane fouling at a high recovery of 90%, the collected MBR

385 permeate was concentrated by discharging the RO permeate until the concentration equivalent to

RI
386 90% recovery was attained. Subsequently, the test was performed with a full recycle of permeate

SC
387 back to the feed tank to assess the membrane fouling. As shown in Figure 7b, the initial TMP

388 increased from 5.3 to 8.5 bar and from 4.5 to 7.0 bar when the recovery ratio was increased from 0

389
U
to 90% for the UF-MBR and NF-MBR, respectively. The lower initial operating pressure required
AN
390 for the concentration process in the NF-MBR could be attributed to its superior permeate quality

391 with less divalent ions and dissolved organics (Figure 3 and Table 5). At a high recovery rate of
M

392 90%, TMP increase rate (dTMP/dt) for the RO system fed with the UF-MBR and NF-MBR
D

393 permeate was at ~ 0.30 and 0.094 bar/day, respectively. Obviously, the membrane fouling at 90% of
TE

394 recovery ratio is more severe compared to that at 0% of recovery ratio. Nevertheless, with the

395 improved quality of the RO feed water, the fouling rate of the RO membrane in the NF-MBR+RO
EP

396 was ~ 3.3 times lower than that in the UF-MBR+RO. The lower fouling tendency could translate

397 into energy saving and operation cost saving (e.g., chemical cleaning) associated with fouling.
C
AC

398 Insert Figure 7

399 3.2.2. Foulant analysis

400 Shown in Figure 6, in the RO fouling test with UF-MBR permeate, the ratio of biopolymers in the

401 RO foulant was 43.9%, significantly higher than that in the permeate (1.7%), which suggested a

18
 ACCEPTED MANUSCRIPT

402 greater tendency of their accumulation onto the RO membrane. Meanwhile, the ratios of building

403 blocks (24.6%) and LMWs (31.5%) were comparable to those in the permeate (22.1% and 29.0%

404 respectively). In addition, on the RO membranes, no humic substance (i.e., non-detectable) were

405 observed. Similarly, in the NF-MBR+RO system, a high ratio of biopolymers (40.7%) in the RO

PT
406 foulants was observed. The biopolymers appeared to be the major component of the RO foulants

407 despite both permeates had different compositions and very low amount of biopolymers. The

RI
408 possible mechanism is the transformation of smaller organic molecules to the larger biopolymers

409 through aggregation or bio-conversion (such as producing extracellular polymeric substances)

SC
410 (Jeong et al., 2013). Future studies on this aspect will be performed.

411
U
3.2.3. Correlation between organic fractions in MBR permeate and RO fouling
AN
412 It was found that the amounts of biopolymers, building blocks, and LMWs were ~ 5 to 6 times
M

413 higher in the RO foulant with the UF-MBR permeate than those with the NF-MBR permeate (Table

414 5). The autopsy results agreed well with the TMP profiles in Figure 7. Thus, the removal of organic
D

415 fractions was crucial to the sustainable operation of downstream RO processes. In fact, in a recent
TE

416 work (Katsuki Kimura et al., 2016), where an NF membrane was used to pretreat the MF/UF-MBR

417 effluent prior the RO process, has shown that (1) severe membrane fouling occurred only in the co-
EP

418 presence of organics and silica; (2) the removal of the organic matter from the RO feed water could

419 significantly reduce the RO membrane fouling.

C
AC

420 3.3. Comparison of energy consumption of NF-MBR+RO and UF-MBR+RO

421 Energy consumption is an important aspect in evaluating the feasibility of NF-MBR+RO process

422 for high recovery water reclamation. In MBR+RO processes, aeration energy and pump energy are

423 practically the two major contributors to the total energy demand. The total aeration energy is the

19
 ACCEPTED MANUSCRIPT

424 sum of power consumption of blowers for membrane (only for submerged membrane configuration)

425 and biological process. Based on the typical values applied in the aerobic MBRs, normalized

426 biological aeration demand of 10 Nm3 air/m3 product (Qin et al., 2007) and a specific membrane

427 aeration demand of 0.3 Nm3/m2h (Judd and Judd, 2006) were chosen. A power consumption of

PT
428 0.019 kWh/Nm3 air for coarse bubble aeration was taken for calculation (Maere et al., 2011). Pump

429 energy usage was evaluated using the pump power requirement equation (Judd and Judd, 2006):

RI
Q ρgH
430 P= (1)
1000 ξ

SC
431 where P is the pump energy demand (kW), Q is the volumetric liquid flow rate (m3/s), ρ is the

432
U
pumped fluid density (kg/m3), g is gravitational field strength (9.8 N/kg), H is water head (m, which
AN
433 is equivalent to ~10 times of the feed pressure (bar)), and ξ is pump efficiency (assuming 80% in

434 this study).

M

435 Table 6 summarizes the total energy consumption (kWh/m3 product) of the submerged UF-
D

436 MBR+RO and side-stream NF-MBR+RO at a RO recovery rate of 75% or 90% respectively. It is
TE

437 noted that if the RO membrane is operated at recovery of 90% in the submerged UF-MBR+RO

438 systems, additional RO pre-treatment (i.e., MBR post-treatment, such as activated carbon
EP

439 adsorption process (Wang et al., 2014) or low-pressure NF process (K. Kimura et al., 2016)) is
C

440 generally necessary in order to achieve consistent RO performance. Thus, this scenario is not
AC

441 discussed in this study.

442 In the NF-MBR+RO system, increasing the RO recovery rate from 75% to 90%, the energy

443 consumption only inclines from 0.732 to 0.739 kWh/m3, both are slightly lower than that of the

444 submerged UF-MBR+RO (0.770 kWh/m3) at 75% of recovery rate. This signifies the feasibility of

445 NF-MBR+RO for high recovery water reclamation process. It is worth noting that the capital cost of

20
 ACCEPTED MANUSCRIPT

446 NF-MBR+RO may be higher than UF-MBR+RO due to relatively lower NF permeate flux (10

447 L/m2h) and larger NF membrane area is required. A decrease in capital cost of NF-MBR+RO could

448 be foreseen with improved permeability by developing novel NF membranes and advanced NF

449 membrane fouling control strategies.

PT
450 Insert Table 6

RI
451 4. Conclusions

SC
452 In this study, a novel NF-MBR+RO system for high recovery water reclamation was developed and

453 compared with a UF-MBR+RO system. The following conclusions can be drawn:

454
U
(1) The NF-MBR achieved higher organic and inorganic removal than the UF-MBR. The extended
AN
455 retention time could enhance the biodegradation of retained organics in the reactor.
M

456 (2) The fouling experienced by the NF-MBR was highly reversible in which the PWP can be

457 restored by physical cleaning, while irreversible fouling was predominant in the UF-MBR.
D
TE

458 (3) The subsequent RO membrane fed with the NF-MBR permeate displayed better performance

459 than that with the UF-MBR permeate due to lower organic/inorganic substances present in the NF-
EP

460 MBR permeate.

C

461 (4) In both NF-MBR+RO and UF-MBR+RO, biopolymers were identified as the predominant
AC

462 organic foulants in the MBR and RO processes.

463 (5) The calculated energy consumption of the NF-MBR+RO system at 90% recovery was

464 comparable to that of the UF-MBR+RO at 75% recovery, which proves the feasibility of NF-

465 MBR+RO for high recovery water reclamation.

21
 ACCEPTED MANUSCRIPT

466 Acknowledgements

467 The funding support from the Ministry of Education (MOE) of Singapore is acknowledged. The

468 funding support from the Economic Development Board (EDB) of Singapore to the Singapore

469 Membrane Technology Centre (SMTC) is also acknowledged.

PT
470 References

RI
471 American Public Health, A., Eaton, A. D., American Water Works, A., Water Environment, F.,
472 2005. Standard methods for the examination of water and wastewater Washington, D.C.,

SC
473 APHA-AWWA-WEF.
474 Antony, A., Fudianto, R., Cox, S., Leslie, G., (2010). Assessing the oxidative degradation of
475 polyamide reverse osmosis membrane-Accelerated ageing with hypochlorite exposure,
476 Journal of Membrane Science, 347 (1-2), pp. 159-164.

U
477 Antony, A., Low, J. H., Gray, S., Childress, A. E., Le-Clech, P., Leslie, G., (2011). Scale formation
478 and control in high pressure membrane water treatment systems: A review, Journal of
AN
479 Membrane Science, 383 (1), pp. 1-16.
480 Aslam, M., Charfi, A., Lesage, G., Heran, M., Kim, J., (2017). Membrane bioreactors for
481 wastewater treatment: A review of mechanical cleaning by scouring agents to control
M

482 membrane fouling, Chemical Engineering Journal, 307, pp. 897-913.

483 Bartels, C. R., Wilf, M., Andes, K., Iong, J. (2005) Design considerations for wastewater treatment
484 by reverse osmosis. Vol. 51. Water Science and Technology (pp. 473-482).
D

485 Bradford, M. M., (1976). A rapid and sensitive method for the quantitation of microgram quantities
486 of protein utilizing the principle of protein-dye binding, Analytical Biochemistry, 72 (1), pp.
TE

487 248-254.
488 Chen, F., Bi, X., Ng, H. Y., (2016). Effects of bio-carriers on membrane fouling mitigation in
489 moving bed membrane bioreactor, Journal of Membrane Science, 499, pp. 134-142.
490 Chen, J. C., Ng, W. J., Luo, R., Mu, S., Zhang, Z., Andersen, M., Jørgensen, P. E., (2012).
EP

491 Membrane Bioreactor Process Modeling and Optimization: Ulu Pandan Water Reclamation
492 Plant, Journal of Environmental Engineering, 138 (12), pp. 1218-1226.
493 Choi, J.-H., Fukushi, K., Yamamoto, K., (2007). A submerged nanofiltration membrane bioreactor
494 for domestic wastewater treatment: the performance of cellulose acetate nanofiltration
C

495 membranes for long-term operation, Separation and Purification Technology, 52 (3), pp.
496 470-477.
AC

497 Choi, J.-H., Dockko, S., Fukushi, K., Yamamoto, K., (2002). A novel application of a submerged
498 nanofiltration membrane bioreactor (NF MBR) for wastewater treatment, Desalination, 146
499 (1–3), pp. 413-420.
500 Choi, J.-H., Fukushi, K., Ng, H. Y., Yamamoto, K., (2006). Evaluation of a long-term operation of
501 a submerged nanofiltration membrane bioreactor (NF MBR) for advanced wastewater
502 treatment, Water Science and Technology, 53 (6), pp. 131-136.
503 Choi, J.-H., Lee, S. H., Fukushi, K., Yamamoto, K., (2007). Comparison of sludge characteristics
504 and PCR–DGGE based microbial diversity of nanofiltration and microfiltration membrane
505 bioreactors, Chemosphere, 67 (8), pp. 1543-1550.

22
 ACCEPTED MANUSCRIPT

506 Chong, T. H., Loo, S.-L., Krantz, W. B., (2015). Energy-efficient reverse osmosis desalination
507 process, Journal of Membrane Science, 473, pp. 177-188.
508 Chong, T. H., Loo, S.-L., Fane, A. G., Krantz, W. B., (2015). Energy-efficient reverse osmosis
509 desalination: Effect of retentate recycle and pump and energy recovery device efficiencies,
510 Desalination, 366, pp. 15-31.
511 Cornelissen, E. R., Harmsen, D., de Korte, K. F., Ruiken, C. J., Qin, J. J., Oo, H., Wessels, L. P.,
512 (2008). Membrane fouling and process performance of forward osmosis membranes on
513 activated sludge, Journal of Membrane Science, 319 (1-2), pp. 158-168.

PT
514 Fu, C., Bi, X. J., Ng, H. Y., (2016). Effects of bio-carriers on membrane fouling mitigation in
515 moving bed membrane bioreactor, Journal of Membrane Science, 499, pp. 134-142.
516 Goh, S., Zhang, J., Liu, Y., Fane, A. G., (2013). Fouling and wetting in membrane distillation (MD)

RI
517 and MD-bioreactor (MDBR) for wastewater reclamation, Desalination, 323 (Supplement
518 C), pp. 39-47.
519 Huber, S. A., Balz, A., Abert, M., Pronk, W., (2011). Characterisation of aquatic humic and non-
520 humic matter with size-exclusion chromatography – organic carbon detection – organic

SC
521 nitrogen detection (LC-OCD-OND), Water Research, 45 (2), pp. 879-885.
522 Jang, D., Hwang, Y., Shin, H., Lee, W., (2013). Effects of salinity on the characteristics of biomass
523 and membrane fouling in membrane bioreactors, Bioresource Technology, 141, pp. 50-56.

U
524 Jeong, S., Kim, S. J., Kim, L. H., Shin, M. S., Vigneswaran, S., Nguyen, T. V., Kim, I. S., (2013).
525 Foulant analysis of a reverse osmosis membrane used pretreated seawater, Journal of
AN
526 Membrane Science, 428, pp. 434-444.
527 Judd, S., Judd, C. (Eds.). (2006). The MBR book : principles and applications of membrane
528 bioreactors in water and wastewater treatment. Amsterdam ; Boston ; London: Elsevier.
529 Kent, F. C., Farahbakhsh, K., Mahendran, B., Jaklewicz, M., Liss, S. N., Zhou, H., (2011). Water
M

530 reclamation using reverse osmosis: Analysis of fouling propagation given tertiary membrane
531 filtration and MBR pretreatments, Journal of Membrane Science, 382 (1–2), pp. 328-338.
532 Kimura, K., Okazaki, S., Ohashi, T., Watanabe, Y., (2016). Importance of the co-presence of silica
D

533 and organic matter in membrane fouling for RO filtering MBR effluent, J. Membrane Sci.,
534 501, pp. 60-67.
TE

535 Kimura, K., Okazaki, S., Ohashi, T., Watanabe, Y., (2016). Importance of the co-presence of silica
536 and organic matter in membrane fouling for RO filtering MBR effluent, Journal of
537 Membrane Science, 501 (Supplement C), pp. 60-67.
538 Lay, W. C., Liu, Y., Fane, A. G., (2010). Impacts of salinity on the performance of high retention
EP

539 membrane bioreactors for water reclamation: A review, Water Res., 44 (1), pp. 21-40.
540 Lay, W. C. L., Liu, Y., Fane, A. G., (2010). Impacts of salinity on the performance of high retention
541 membrane bioreactors for water reclamation: A review, Water Research, 44 (1), pp. 21-40.
C

542 Liu, C., Shi, L., Wang, R., (2015). Crosslinked layer-by-layer polyelectrolyte nanofiltration hollow
543 fiber membrane for low-pressure water softening with the presence of SO42− in feed water,
AC

544 Journal of Membrane Science, 486, pp. 169-176.

545 Lu, X., Kujundzic, E., Mizrahi, G., Wang, J., Cobry, K., Peterson, M., Gilron, J., Greenberg, A. R.,
546 (2012). Ultrasonic sensor control of flow reversal in RO desalination—Part 1: Mitigation of
547 calcium sulfate scaling, Journal of Membrane Science, 419, pp. 20-32.
548 Luo, W., Xie, M., Hai, F. I., Price, W. E., Nghiem, L. D., (2016). Biodegradation of cellulose
549 triacetate and polyamide forward osmosis membranes in an activated sludge bioreactor:
550 Observations and implications, Journal of Membrane Science, 510, pp. 284-292.
551 Luo, W., Hai, F. I., Kang, J., Price, W. E., Nghiem, L. D., Elimelech, M., (2015). The role of
552 forward osmosis and microfiltration in an integrated osmotic-microfiltration membrane
553 bioreactor system, Chemosphere, 136 (Supplement C), pp. 125-132.

23
 ACCEPTED MANUSCRIPT

554 Luo, W., Hai, F. I., Price, W. E., Guo, W., Ngo, H. H., Yamamoto, K., Nghiem, L. D., (2014). High
555 retention membrane bioreactors: Challenges and opportunities, Bioresource Technology,
556 167, pp. 539-546.
557 Luo, W., Phan, H. V., Xie, M., Hai, F. I., Price, W. E., Elimelech, M., Nghiem, L. D., (2017).
558 Osmotic versus conventional membrane bioreactors integrated with reverse osmosis for
559 water reuse: Biological stability, membrane fouling, and contaminant removal, Water
560 Research, 109, pp. 122-134.
561 Maere, T., Verrecht, B., Moerenhout, S., Judd, S., Nopens, I., (2011). BSM-MBR: A benchmark

PT
562 simulation model to compare control and operational strategies for membrane bioreactors,
563 Water Res., 45 (6), pp. 2181-2190.
564 Moreno, J., Monclús, H., Stefani, M., Cortada, E., Aumatell, J., Adroer, N., De Lamo-Castellví, S.,

RI
565 Comas, J., (2013). Characterisation of RO fouling in an integrated MBR/RO system for
566 wastewater reuse, Water Science and Technology, 67 (4), pp. 780-788.
567 Ng, H. Y., Ong, S. L., Ng, W. J., (2005). Effects of sodium chloride on the performance of a
568 sequencing batch reactor, Journal of Environmental Engineering (ASCE), 131 (11), pp.

SC
569 Phattaranawik, J., Fane, A. G., Pasquier, A. C. S., Bing, W., (2008). A novel membrane bioreactor
570 based on membrane distillation, Desalination, 223 (1-3), pp. 386-395.
571 Qin, J.-J., Kekre, K. A., Tao, G., Oo, M. H., Wai, M. N., Lee, T. C., Viswanath, B., Seah, H.,

U
572 (2006). New option of MBR-RO process for production of NEWater from domestic sewage,
573 Journal of Membrane Science, 272 (1–2), pp. 70-77.
AN
574 Qin, J. J., Wai, M. N., Tao, G., Kekre, K. A., Seah, H., (2007). Membrane bioreactor study for
575 reclamation of mixed sewage mostly from industrial sources, Sep. Purif. Technol., 53 (3),
576 pp. 296-300.
577 Qiu, G., Zhang, S., Srinivasa Raghavan, D. S., Das, S., Ting, Y.-P., (2016). The potential of hybrid
M

578 forward osmosis membrane bioreactor (FOMBR) processes in achieving high throughput
579 treatment of municipal wastewater with enhanced phosphorus recovery, Water Research,
580 105, pp. 370-382.
D

581 Radjenović, J., Petrović, M., Barceló, D., (2009). Fate and distribution of pharmaceuticals in
582 wastewater and sewage sludge of the conventional activated sludge (CAS) and advanced
TE

583 membrane bioreactor (MBR) treatment, Water Research, 43 (3), pp. 831-841.
584 Rajabzadeh, S., Liu, C., Shi, L., Wang, R., (2014). Preparation of low-pressure water softening
585 hollow fiber membranes by polyelectrolyte deposition with two bilayers, Desalination, 344,
586 pp. 64-70.
EP

587 Ridgway, H. F., Justice, C. A., Whittaker, C., Argo, D. G., Olson, B. H., (1984). Biofilm Fouling of
588 RO Membranes&#x2014;Its Nature and Effect on Treatment of Water for Reuse, Journal
589 (American Water Works Association), 76 (6), pp. 94-102.
C

590 Septon, J., Efraty, A., (2016). CCD series no-17: application of the BWRO-CCD technology for
591 high-recovery low-energy desalination of domestic effluents, Desalination and Water
AC

592 Treatment, 57 (21), pp. 9585-9601.

593 Shaffer, D. L., Werber, J. R., Jaramillo, H., Lin, S., Elimelech, M., (2015). Forward osmosis: Where
594 are we now?, Desalination, 356, pp. 271-284.
595 Sutzkover-Gutman, I., Hasson, D., Semiat, R., (2010). Humic substances fouling in ultrafiltration
596 processes, Desalination, 261 (3), pp. 218-231.
597 Suwarno, S. R., Hanada, S., Chong, T. H., Goto, S., Henmi, M., Fane, A. G., (2016). The effect of
598 different surface conditioning layers on bacterial adhesion on reverse osmosis membranes,
599 Desalination, 387, pp. 1-13.

24
 ACCEPTED MANUSCRIPT

600 Tadkaew, N., Hai, F. I., McDonald, J. A., Khan, S. J., Nghiem, L. D., (2011). Removal of trace
601 organics by MBR treatment: The role of molecular properties, Water Research, 45 (8), pp.
602 2439-2451.
603 Tang, C. Y., Chong, T. H., Fane, A. G., (2011). Colloidal interactions and fouling of NF and RO
604 membranes: A review, Advances in Colloid and Interface Science, 164 (1), pp. 126-143.
605 Teychene, B., Guigui, C., Cabassud, C., (2011). Engineering of an MBR supernatant fouling layer
606 by fine particles addition: A possible way to control cake compressibility, Water Research,
607 45 (5), pp. 2060-2072.

PT
608 Wang, G. H., Fan, Z., Wu, D. X., Qin, L., Zhang, G. L., Gao, C. J., Meng, Q., (2014).
609 Anoxic/aerobic granular active carbon assisted MBR integrated with nanofiltration and
610 reverse osmosis for advanced treatment of municipal landfill leachate, Desalination, 349,

RI
611 pp. 136-144.
612 Warsinger, D. M., Tow, E. W., Nayar, K. G., Maswadeh, L. A., Lienhard V, J. H., (2016). Energy
613 efficiency of batch and semi-batch (CCRO) reverse osmosis desalination, Water Research,
614 106, pp. 272-282.

SC
615 Werber, J. R., Deshmukh, A., Elimelech, M., (2017). Can batch or semi-batch processes save
616 energy in reverse-osmosis desalination?, Desalination, 402, pp. 109-122.
617 Wetterau, G., Liu, P., Chalmers, B., Richardson, T., VanMeter, H. B., (2011). Optimizing RO

U
618 Design Criteria for Indirect Potable Reuse, IDA Journal of Desalination and Water Reuse, 3
619 (4), pp. 40-45.
AN
620 Wibisono, Y., Cornelissen, E. R., Kemperman, A. J. B., van der Meer, W. G. J., Nijmeijer, K.,
621 (2014). Two-phase flow in membrane processes: A technology with a future, Journal of
622 Membrane Science, 453, pp. 566-602.
623 Wijekoon, K. C., Hai, F. I., Kang, J., Price, W. E., Guo, W., Ngo, H. H., Nghiem, L. D., (2013).
M

624 The fate of pharmaceuticals, steroid hormones, phytoestrogens, UV-filters and pesticides
625 during MBR treatment, Bioresource Technology, 144, pp. 247-254.
626 Wu, B., Kitade, T., Chong, T. H., Uemura, T., Fane, A. G., (2013). Impact of membrane bioreactor
D

627 operating conditions on fouling behavior of reverse osmosis membranes in MBR–RO

628 processes, Desalination, 311, pp. 37-45.
TE

629 Yamamoto, K. (2001). Membrane bioreactor: An advanced wastewater treatment/reclamation

630 technology and its function in excess-sludge minimization A2 - Matsuo, Tomonori. In K.
631 Hanaki, S. Takizawa & H. Satoh (Eds.), Advances in Water and Wastewater Treatment
632 Technology (pp. 229-237). Amsterdam: Elsevier Science B.V.
EP

633 Zhao, Y., Song, L., Ong, S. L., (2010). Fouling behavior and foulant characteristics of reverse
634 osmosis membranes for treated secondary effluent reclamation, Journal of Membrane
635 Science, 349 (1), pp. 65-74.
C

636
637
AC

638

25
 ACCEPTED MANUSCRIPT
Table 1. A comparison of conventional and high retention MBRs (Luo et al., 2014)

Table 2. Operating parameters of the UF-MBR and NF-MBR

Table 3. Characteristics of the municipal wastewater (n=25)

Table 4. Properties of the UF and NF membranes (Liu et al., 2015)

Table 5. Concentrations of dissolved organic fractions in the feed, mixed liquor,

PT
permeate and foulants of the MBRs and ROs.

Table 6. A comparison of energy consumption (kWh/m3 product) for NF-MBR+RO and

RI
UF-MBR+RO at different recovery ratios.

U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
Table 1. A comparison of conventional and high retention MBRs (Luo et al., 2014)

Conventional MBR High retention MBR

MF/UF-MBR MD-MBR FO-MBR NF-MBR
Membrane Hydrophilic porous Hydrophobic Dense FO (usually Dense NF (usually
MF/UF porous MF TFC type) TFC type)
Driving force Hydraulic pressure Vapour pressure Osmotic pressure Hydraulic pressure
NaCl rejection (%) Negligible ~ 100 ~ 100 40 – 90
TOC in permeate (mg/L) 3 – 10 < 0.8 <3 1–4
Organic micropollutants low high high high

PT
rejection
Permeate flux (L/m2h) 10 – 30 1.2 – 15 < 10 < 2.5
Additional step - Heat source is water/draw solute -
required separation

RI
U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
Table 2. Operating parameters of the UF-MBR and NF-MBR

Parameter MBRs
HRT (h) 22
SRT (d) 30
Flux (L/m2h) 10
Organic loading Rate (g COD/L.day) 0.43
F/M Ratio (g COD/g MLSS.d) 0.43

PT
Crossflow velocity of side stream 0.20
membrane unit (m/s)
pH 6.3 – 7.3
Aeration (L/min) 3.0

RI
Dissolved oxygen level (mg/L) 5.5 – 6.5
Biocarrier filling ratio (%) 10

U SC
AN
M
D
TE
CEP
AC
 ACCEPTED MANUSCRIPT
Table 3. Characteristics of the municipal wastewater (n=25)

Parameter Value
tCOD (mg/L) 389.8 ± 169.9
sCOD (mg/L) 124.9 ± 45.1
DOC (mg/L) 48.3 ± 13.9
Ca2+ (mg/L) 28.9 ± 3.5
Mg2+ (mg/L) 7.9 ± 1.2
Na+ (mg/L) 105.2 ± 6.9

PT
TN (mg/L) 40.7 ± 6.1
NH4-N (mg/L) 40.9 ± 4.5
NO3—N (mg/L) 0.3 ± 0.2
NO2—N (mg/L)

RI
N.D.
PO43- (mg/L) 23.6 ± 3.2
Conductivity (µs/cm) 859.5 ± 18.9

SC
Note: tCOD: total chemical oxygen demand; sCOD: soluble chemical oxygen demand;
DOC: dissolved organic carbon; TN: total nitrogen; TP: total phosphorus; N.D.: not
detectable.

U
AN
M
D
TE
CEP
AC
 ACCEPTED MANUSCRIPT
Table 4. Properties of the UF and NF membranes (Liu et al., 2015)

UF membrane NF membrane
Material PES Glutaraldehyde (GA) crosslinked layer-by-
layer polyelectrolyte with PES UF as substrate
Pore size (MWCO, 50,000 200-300
Da)
Active Layer N.A. Lumen side

PT
ID/OD (mm) 0.9/1.4 0.9/1.4
Ca2+ Rejection (%) N.A. 91.8
Mg2+ Rejection (%) N.A. 98.4
Note: MWCO: Molecular weight cut-off; ID: inner diameter; OD: outer diameter; N.A.: not applicable.

RI
U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
Table 5. Concentrations of dissolved organic fractions in the feed, mixed liquor,
permeate and foulants of the MBRs and ROs.

Humic Building
DOC Biopolymers LMW
substances blocks
Feed (mg/L) 35.6 8.4 7.8 5.0 8.2
Mixed Liquor
16.7 7.7 4.1 2.0 2.8
(mg/L)
UF permeate

PT
8.0 0.1 3.2 1.5 2.0
UF-MBR + (mg/L)
a
RO UF foulant
201.2 159.2 N.D.e 37.6 13.4
(mg/m2)

RI
RO foulantb
104.5 45.3 N.D. 25.4 32.5
(mg/m2)
Mixed Liquor

SC
109.5 17.8 46.9 24.3 20.9
(mg/L)
NF permeate
0.9 N.D. 0.1 0.1 0.4
NF-MBR + (mg/L)
RO NF foulantc

U
198.3 128.1 43.4 17.2 37.8
(mg/m2)
RO foulantd
AN
22.6 7.0 N.D. 4.6 5.6
(mg/m2)
a
UF2 membrane was removed for autopsy at the end of cycle.
b
RO membrane was operated for ~8 days, after which a TMP of 6.5 bar was reached
M

(equivalent to ~ 20% increase in TMP).

c
NF3 membrane was removed for autopsy at the end of cycle.
d
RO membrane was operated for ~26 days, after which a TMP of 6.5 bar was reached
D

(equivalent to ~ 20% increase in TMP).

e
N.D. represents "not detectable".
TE
C EP
AC
 ACCEPTED MANUSCRIPT
Table 6. A comparison of energy consumption (kWh/m3 product) for NF-MBR+RO and
UF-MBR+RO at different recovery ratios.

Submerged UF- Side-stream NF- Side-stream NF-

MBR+RO (75% MBR+RO (75% MBR+RO (90%
recovery) recovery) recovery)
Feed pump 0.005 0.005 0.004
Waste pump 0.0001 0.0001 0.0001
MBR UF/NF crossflow pump - 0.231a 0.193

PT
UF permeate pump 0.012 - -
Aeration (biological) - 0.190 0.190
Aeration (membrane) 0.447b - -
0.306c 0.352d

RI
RO High-pressure pump 0.306
Total 0.770 0.732 0.739
a
Feed pressure of NF at 100% of recovery was set at 2.5 bar.
b
UF flux was set at 17 L/m2h.

SC
c
Feed pressure of RO at 75% of recovery was set at 6.62 bar.
d
Feed pressure of RO at 90% of recovery was set at 9.12 bar.

U
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
Figure 1. DOC, Conductivity, Ca2+, Mg2+, and phosphate concentrations in the NF-
MBR.

Figure 2. COD (a) and DOC (B) in the permeate of NF-MBR and UF-MBR.

Figure 3. Comparison of permeate quality of NF-MBR and UF-MBR.

Figure 4. TMP profiles in the (a) NF-MBR and (b) UF-MBR.

PT
Figure 5. (a) Membrane water permeability and (b) irreversible fouling resistance after
each cleaning cycle for UF2 and NF3. Note that the test period is 37 and 26 days for UF2
and NF3, respectively.

RI
Figure 6. Contribution ratios of dissolved organic fractions in feed, mixed liquor,
permeate and foulants of UF-MBR+RO and NF-MBR+RO.

SC
Figure 7. TMP profiles of the RO systems fed with permeate from UF-MBR and NF-
MBR at a concentration equivalent to (a) 0% recovery (b) 90% recovery. CP represents
the concentration process to increase the concentration of the solution from 0 to 90%

U
recovery equivalent.
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 1. DOC, Conductivity, Ca2+, Mg2+, and phosphate concentrations in the NF-
M

MBR.
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

Permeate (NF-MBBR) Permeate (UF-MBBR)

Permeate (NF-MBBR) Permeate (UF-MBBR) Removal (NF-MBBR) Removal (UF-MBBR)
Removal (NF-MBBR) Removal (UF-MBBR)
40 100 40 100
Permeate COD ( mg/L)

Permeate DOC (mg/L)

Removal ratio ( %)
(a) (b)

Removal ratio ( %)
80 80
30 30
60 60
20 20

PT
40 40
10 20 10
20

RI
0 0 0 0
0 20 40 60 80 0 20 40 60 80
Time (day) Time (day)

U SC
Figure 2. COD (a) and DOC (B) in the permeate of NF-MBR and UF-MBR.
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

40

Conductivity (× 100 µS/cm)

Concentration (mg/L) UF-MBR NF-MBR
30

20

PT
10

RI
0

U SC
Figure 3. Comparison of permeate quality of NF-MBR and UF-MBR.
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

500
(a) NF1 NF2 NF3
400
TMP ( kPa)

300

200

PT
100

RI
0 10 20 30 40 50 60 70 80
Time (day)

SC
50
(b) UF1 UF2
40

U
TMP ( kPa)

30
AN
20

10
M

0
D

0 10 20 30 40 50 60 70 80
Time (day)
TE

Figure 4. TMP profiles in the (a) NF-MBR and (b) UF-MBR.

C EP
AC
 ACCEPTED MANUSCRIPT

UF2 NF3 UF2 NF3

200 5.5 10
(a) (b)
Permeability (mL/min)

Permeability (mL/min)
160 5.0 8

Rirr (×10 12 m-1)

120 4.5 6

80 4.0 4

PT
40 3.5 2

0 3.0 0

RI
0 5 10 15 20 0 5 10 15 20
No. of cleaning cycle
No. of cleaning cycle

SC
Figure 5. (a) Membrane water permeability and (b) irreversible fouling resistance after
each cleaning cycle for UF2 and NF3. Note that the test period is 37 and 26 days for UF2

U
and NF3, respectively.
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

Biopolymer HS Building block LMW

UF-MBR+RO NF-MBR+RO
100

80
Ratio (%)

60

PT
40

RI
20

SC
UF foulant

NF foulant
Feed

Mixed Liquor

Mixed Liquor
RO foulant

RO foulant
Permeate

Permeate
U
AN
Figure 6. Contribution ratios of dissolved organic fractions in feed, mixed liquor,
permeate and foulants of UF-MBR+RO and NF-MBR+RO.
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

7
Fed with UF-MBR permea te,
dTMP/dt ~0.11 ba r/da y

TMP(bar) 6

Fed with NF-MBR permea te,

PT
dTMP/dt ~0.03 ba r/da y
5

RI
(a)
4
0 10 20 30

SC
Time (day)
10

U
Fed with UF-MBR permeate,
CP
dTMP/dt ~0.30 bar/day
AN
8
TMP(bar)

Fed with NF-MBR permeate,

dTMP/dt ~0.09 bar/day
6
D

Fouling
(b)
TE

4
0 5 10 15
Time (day)
C EP

Figure 7. TMP profiles of the RO systems fed with permeate from UF-MBR and NF-
MBR at a concentration equivalent to (a) 0% recovery (b) 90% recovery. CP represents
AC

the concentration process to increase the concentration of the solution from 0 to 90%
recovery equivalent.
 ACCEPTED MANUSCRIPT
Highlights:

• A novel NF-MBR+RO system for high recovery water reclamation was developed

• The high-flux NF-MBR achieved excellent permeate quality in treating domestic

wastewater

PT
• Fouling mechanisms and foulant characteristics in the NF-MBR+RO and UF-

MBR+RO system were illustrated

RI
• Energy consumption between the NF-MBR+RO and UF-MBR+RO system was

SC
compared

U
AN
M
D
TE
C EP
AC