Accepted Manuscript Optimization of Gravity-Driven Membrane (GDM) Filtration Process for Seawater Pretreatment Bing Wu, Florian Hochstrasser, Ebrahim Akhondi, Noemi Ambaun, Lukas Techirren, Michael Burkhardt, Anthony G. Fane, Wouter Pronk PII:
S0043-1354(16)30080-X
DOI:
10.1016/j.watres.2016.02.021
Reference:
WR 11834
To appear in:
Water Research
Received Date: 7 December 2015 Revised Date:
10 February 2016
Accepted Date: 11 February 2016
Please cite this article as: Wu, B., Hochstrasser, F., Akhondi, E., Ambaun, N., Techirren, L., Burkhardt, M., Fane, A.G., Pronk, W., Optimization of Gravity-Driven Membrane (GDM) Filtration Process for Seawater Pretreatment, Water Research (2016), doi: 10.1016/j.watres.2016.02.021. 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.
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT Optimization of Gravity-Driven Membrane (GDM) Filtration Process for Seawater
2
Pretreatment
3
Bing Wu1,*, Florian Hochstrasser2, Ebrahim Akhondi1, Noemi Ambaun3, Lukas Techirren2,
4
Michael Burkhardt2, Anthony G. Fane1,4, Wouter Pronk5,*
RI PT
1
5 6
1
Singapore Membrane Technology Centre, Nanyang Environment and Water Research Institute, Nanyang Technological University, 1 Cleantech Loop,
8
CleanTech One #06-08, Singapore 637141
Sciences Rapperswil, Oberseestrasse 10, 8640 Rapperswil, Switzerland 3
11
4
Eawag, Swiss Federal Institute of Aquatic Science and Technology, Überlandstrasse 133, 8600 Duebendorf, Switzerland
EP
16
5
* Corresponding authors:
AC C
15
School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798
13
14
Institute of Environmental Engineering, ETH Zürich, 8093 Zürich, Switzerland
TE D
10
12
Institute of Environmental and Process Engineering, HSR University of Applied
M AN U
2
9
SC
7
17
Wouter Pronk, Email:
[email protected], Tel: 41-58-7655381, Fax: 41-58-7655802
18
Bing Wu, Email:
[email protected], Tel: 65-9125-8929, Fax: 65-6791-0756
19
20
21
1
ACCEPTED MANUSCRIPT Abstract
23
Seawater pretreatment by gravity-driven membrane (GDM) filtration at 40 mbar has been
24
investigated. In this system, a beneficial biofilm develops on the membrane that helps to
25
stabilize flux. The effects of membrane type, prefiltration and system configuration on stable
26
flux, biofilm layer properties and dissolved carbon removal were studied. The results show
27
that the use of flat sheet PVDF membranes with pore sizes of 0.22 and 0.45 µm in GDM
28
filtration achieved higher stabilized permeate fluxes (7.3-8.4 L/m2h) than that of flat sheet
29
PES 100 kD membranes and hollow fibre PVDF 0.1 µm membranes. Pore constriction and
30
cake filtration were identified as major membrane fouling mechanisms, but their relative
31
contributions varied with filtration time for the various membranes. Compared to raw
32
seawater, prefiltering of seawater with meshes at sizes of 10, 100 and 1000 µm decreased the
33
permeate flux, which was attributed to removal of beneficial eukaryotic populations. Optical
34
coherence tomography (OCT) showed that the porosity of the biofouling layer was more
35
significantly related with permeate flux development rather than its thickness and roughness.
36
To increase the contact time between the biofilm and the dissolved organics, a hybrid
37
biofilm-submerged GDM reactor was evaluated, which displayed significantly higher
38
permeate fluxes than the submerged GDM reactor. Although integrating the biofilm reactor
39
with the membrane system displayed better permeate quality than the GDM filtration cells, it
40
could not effectively reduce dissolved organic substances in the seawater. This may be
41
attributed to the decomposition/degradation of solid organic substances in the feed and
42
carbon fixation by the biofilm. Further studies of the dynamic carbon balance are required.
43
Key words: Biofilm layer; Gravity-driven membranes; Optical Coherence Tomography
44
(OCT); Seawater pretreatment; Stabilized flux
AC C
EP
TE D
M AN U
SC
RI PT
22
45
2
ACCEPTED MANUSCRIPT 1.
Introduction
47
Seawater reverse osmosis (SWRO) desalination as an alternative supply of high-quality
48
drinking water is receiving global attention. However, compared to surface and groundwater
49
RO processes, SWRO requires much more energy (Pearce, 2008). The energy consumption
50
of RO membrane filtration is associated with the osmotic pressure of the feed and the
51
recovery (the ratio of product flow and feed flow). Thermodynamics dictate the minimum
52
energy for desalination, and major efforts have been made in SWRO to approach this
53
minimum. In the desalination process, additional energy is required for intake, pretreatment,
54
posttreatment, and brine discharge stages. It has been noted that the energy demand for
55
pretreatment of raw seawater accounts for the majority of the ancillary energy used
56
(Elimelech and Phillip, 2011) and provides an opportunity for meaningful reduction of the
57
overall system energy.
58
The primary goal of seawater pretreatment is to remove particles and reduce organics in the
59
seawater feed, which will cause less fouling in the RO process. Although conventional
60
pretreatment processes (e.g., coagulation, dissolved air flotation, and media filtration) have
61
been widely used for SWRO due to their moderate energy consumption, they typically
62
consume chemicals and produce quality of the treated seawater that often is not satisfactory.
63
This can lead to higher cleaning frequency and replacement of RO membranes. In addition,
64
stable RO operation may not be guaranteed with poor or variable quality seawater (Greenlee
65
et al., 2009; Pearce, 2008). The improved pretreatment achievable by low pressure
66
membranes has been recognized for some time (Pearce, 2008), and includes microfiltration
67
(MF) (Bae et al., 2011), ultrafiltration (UF) (Huang et al., 2011), while even nanofiltration
68
(NF) was considered as a suitable pretreatment (Kaya et al., 2015). However, these pressure-
69
driven membrane filtration processes also require effective fouling control strategies (such as
AC C
EP
TE D
M AN U
SC
RI PT
46
3
ACCEPTED MANUSCRIPT cross-flow, backwashing, and/or air scouring etc.) (Akhondi et al., 2014), which contribute
71
significant energy demand to the overall desalination process (Knops et al., 2007).
72
Gravity-driven membrane filtration (GDM), which was initially developed as a low energy
73
process to treat different surface waters and diluted wastewaters (Derlon et al., 2013; Derlon
74
et al., 2012; Peter-Varbanets et al., 2012; Peter-Varbanets et al., 2010; Peter-Varbanets et al.,
75
2011; Peter-Varbanets et al., 2009), has also shown promise as a seawater pre-treatment
76
process with less energy and chemical cleaning (Akhondi et al., 2015). In the GDM filtration
77
process, the feed is under a gravity pressure of 40-100 mbar and is applied without crossflow
78
to the membranes. Over time a biofilm develops on the membrane that arrests further fouling
79
and leads to a stabilized permeate flux. It has been shown that the bacterial community
80
utilizes the organic substances in the feed seawater for their growth and augmentation to form
81
biofouling layers on the membrane, while the movement and predation behaviour of
82
eukaryotic organisms in the biofouling layer produce an open and spatially heterogeneous
83
structure (Akhondi et al., 2015; Derlon et al., 2013; Derlon et al., 2012; Klein et al., 2016).
84
The GDM process consumes only about 3-10% of the energy used in conventional UF
85
pretreatment (Akhondi et al., 2015), however, its permeate flux (3.6 to 7.3 L/m2h) is almost
86
an order of magnitude less than that of conventional UF pretreatment (Xu et al., 2012). The
87
trade-off lies in more membranes required to achieve the same productivity. Therefore, there
88
is an incentive to seek improvements in permeate flux without significant added energy input
89
and capital cost.
90
Our previous work on GDM filtration of seawater (Akhondi et al., 2015) showed the
91
following: (i) higher temperatures improved permeate flux during GDM filtration of
92
seawater. This can be attributed to lower viscosity, increased bacterial metabolism and the
93
presence of eukaryotic organisms that have greater predation activity at higher temperatures,
94
leading to the formation of more porous biofouling layers. (ii) A higher hydrostatic pressure
AC C
EP
TE D
M AN U
SC
RI PT
70
4
ACCEPTED MANUSCRIPT also provides an increased permeate flux thanks to the greater driving force. However, this
96
strategy is limited if the goal is to avoid increased energy demand and capital cost. The
97
influences of other operating parameters on the permeate flux of the GDM filtration process
98
for seawater pretreatment have not been investigated to date.
99
In this study, we focus on optimization of the GDM process by assessing the productivity and
100
permeate quality over a range of operating conditions (such as membrane type and
101
prefiltration strategy). To further improve permeate flux and quality, we also investigated a
102
combined biofilm reactor with a submerged GDM filtration system, in which biofilm carriers
103
facilitated the increase in biovolume and retention time of organic substances. This study
104
provides key information for achieving sustainable operation of GDM as seawater
105
pretreatment.
106 107
2.
108
2.1. Raw seawater
109
Raw seawater was collected from the Tuas Spring Desalination plant, Singapore. The pH,
110
conductivity, dissolved oxygen of raw seawater are given in Table 1.
111
2.2. Experimental setup and operating conditions
112
2.2.1. GDM filtration cell system
113
The GDM filtration cell setup was described previously (Akhondi et al., 2015). In summary,
114
the setup was located inside a dark container and all filtration cells, tubings, and tanks were
115
covered by aluminium foil in order to prevent algal growth. Feed seawater was added to the
116
storage tank periodically, and pumped to the feed tank. The water level in the feed tank was
117
kept constant by overflow, which was connected to the storage tank. The feed seawater
AC C
EP
TE D
Materials and Methods
M AN U
SC
RI PT
95
5
ACCEPTED MANUSCRIPT flowed from the feed tank to membrane filtration cells, which were located 40 cm below the
119
water level of the feed tank (i.e., a hydrostatic pressure of 40 mbar). The permeate of the
120
filtration cell was collected using a plastic bottle and the weight of permeate was measured
121
using an electric balance (OHAUS, USA) on a daily basis. The membrane filtration cell
122
comprised bottom and top parts. The top had a glass window that allowed observation of the
123
biofouling layer morphology. Before the clean membranes were put in the filtration cell, they
124
were soaked in distilled water for 24 h to remove impurities.
125
The GDM filtration cells were used to investigate the effects of three different flat-sheet
126
membranes and one hollow fibre membrane and different pre-treated seawaters on GDM
127
filtration performance. In the first group of experiments, the four GDM filtration systems
128
were operated with four types of membranes fed with raw seawater. The membrane
129
properties are listed in Table 1. The hollow fibre membranes had an internal diameter of 0.6
130
mm and external diameter of 1.2 mm. The hollow fibre membranes were horizontally
131
orientated and the operation mode was outside-in. In the second group of experiments, three
132
prefiltration meshes with sizes of 10 µm, 100 µm, and 1000 µm were employed to pretreat the
133
raw seawater. For each prefiltered seawater, two GDM filtration experiments (flat sheet PES
134
membrane, 100 kD, Microdyn-Nadir, Germany) were performed in parallel. All the
135
experiments were performed at room temperature (23±1°C).
136
2.1.2. Hybrid biofilm-submerged GDM reactor
137
The hybrid biofilm-submerged GDM filtration system is illustrated in Figure 1. Raw seawater
138
in the feed tank was pumped to the biofilm reactor (3.6 L) filled with Kaldnes K3 biofilter
139
media (China). The feed flow rate was regulated according to the permeate flow rate, which
140
ensured a total hydraulic retention time of ~40 h. The effluent from the biofilm reactor was
141
delivered to the GDM tank (8 L) and the effluent from the GDM tank was returned back to
AC C
EP
TE D
M AN U
SC
RI PT
118
6
ACCEPTED MANUSCRIPT the biofilm reactor via a two-channel peristaltic pump (Cole-Parmer, US) at an average
143
recirculation rate of 150 ml/min. The constant effluent level of the GDM tank ensured a
144
hydrostatic pressure of 40 mbar. As a control, a submerged GDM reactor was operated
145
without a biofilter column. Two sandwich type membrane modules (PES, 100 kDa,
146
Microdyn-Nadir, Germany, Figure S1), each with a membrane area of 0.0198 m2 were
147
submerged in the GDM tank and the respective permeate was collected in a beaker. All the
148
experiments were performed at room temperature (23±1°C).
149
2.3. Analytical measurements
150
The dissolved organic carbon (DOC) of the raw seawater (after being filtered through a 0.45
151
µm hydrophilic filter) was monitored using a TOC analyzer (Shimadzu, Japan). The total
152
suspended solids content (TSS) was measured according to Standard Methods (APHA,
153
1998). Turbidity of the seawater was examined using a turbidity meter (Hach, US).
154
Size-exclusion chromatography, in combination with organic carbon and nitrogen detection
155
(LC-OCD-OND) was used to quantify the major soluble organic fractions with different sizes
156
and chemical functions in the feed and permeate waters (after being filtered through a 0.45
157
µm hydrophilic filter) (Akhondi et al., 2015; Huber et al., 2011). On-line purified mobile
158
phase was delivered by an HPLC pump (S-100, Knauer, Berlin, Germany) at a flow rate of
159
1.1 mL/min to an autosampler (MLE, Dresden, Germany) and the chromatographic column
160
(TSK HW 50S, 3000 theoretical plates, Toso, Japan). The first detector after chromatographic
161
separation was a non-destructive, fixed wavelength UV-detector (UVD 254 nm, type S-200,
162
Knauer, Berlin, Germany) for analyzing organic carbon. For nitrogen detection, a side stream
163
was diverted after UVD with a restricted flow rate of 0.1 mL/min (back pressure-driven).
164
Assimilable organic carbon (AOC) measurement was performed using flow cytometry (BD
165
Biosciences, US) as described by Hammes et al.(Hammes et al., 2008; Hammes and Egli,
AC C
EP
TE D
M AN U
SC
RI PT
142
7
ACCEPTED MANUSCRIPT 2005). The samples were filtered into 40 ml vials with PTFE covers (EPA ASM Vial Kit,
167
Thermo Scientific, USA) using a 0.22 µm filter (Millex GP, Millipore, USA) and then
168
sterilized in a water bath at 70 °C for at least 30 min. The sample was inoculated with raw
169
seawater and incubated at 30 °C. After 72 h, 1 ml of sample and 10 µl of 0.2 M EDTA
170
(Kanto Chemical, Japan) were mixed and incubated at 37°C for 10 min. The sample was
171
stained with 10 µl of SYBR green (Molecular Probes, USA) that was diluted 100 times in
172
dimethyl sulfoxide (Sigma-Aldrich, USA). After incubation at 37°C for 20 min, the sample
173
was transferred to a flat bottom well plate for flow cytometry analysis. The unstained sample
174
was used as control. The results were processed using proprietary software (CSampler, BD,
175
USA). Counts in a defined region of a density plot were converted to an AOC concentration
176
according to the relationship (107 cell/mL equivalent to 1 µg AOC/mL) as described by
177
Hammes and Egli (Hammes and Egli, 2005).
178
Transparent exopolymer particles (TEP) content was measured using the method first
179
introduced by Passow and Alldredge (Passow and Alldredge, 1995). In summary, the sample
180
was filtered through a polycarbonate filter (Millipore, USA) with a pore size of 0.1 µm. The
181
accumulated TEP on the membrane was subsequently stained with 3 mL of 0.02% aqueous
182
solution of alcian blue in 0.06% acetic acid (Aldrich-Sigma, USA). Then the membrane was
183
washed with distilled water to remove the excess dye. Six millilitre of 80% H2SO4 solution
184
(Honeywell, Korea) was used to remove the complex of TEP and alcian blue from the
185
polycarbonate membrane and to dissolve it for 2 h. The absorption of the solution was
186
measured at a wavelength of 787 nm using a spectrometer (Hach, USA) and the
187
concentration of TEP was calculated based on a calibration with gum xanthan (Sigma-
188
Aldrich, USA) as a standard.
189
2.4. OCT observation of biofouling layer morphology
AC C
EP
TE D
M AN U
SC
RI PT
166
8
ACCEPTED MANUSCRIPT An optical coherence tomography (OCT, Thorlab, US) was used as a non-invasive approach
191
to investigate the in situ biofouling layer structure at the supramicronscale. After the
192
experiment, the filtration cell was carefully removed from the system. OCT images of the
193
cross-sectional membrane and biofouling layer were collected through the cover glass
194
window of the filtration cell. The physical properties of the biofouling layer such as mean
195
biofouling layer thickness, absolute roughness and relative roughness were calculated
196
according to previously reported references (Akhondi et al., 2015; Derlon et al., 2012). The
197
porosity of the biofouling layer was calculated according to the Carman-Kozeny equation and
198
Darcy’s Law (Derlon et al., 2012; Katsoufidou et al., 2005):
M AN U
SC
RI PT
190
ܴ௨ ௬ =
199
ܴ௧௧ =
ଶఌయ మ
∆
H
Eq. [1]
Eq. [2]
ఓ
TE D
200
ହ (ଵିఌ)మ
ܴ௧௧ = ܴ + ܴ௨ ௬ + ܴ௩௦
201
Eq. [3]
Where ε (dimensionless) is the biofouling layer porosity, rp (m) is the characteristic radius of
203
the particles forming the cake layer, H (m) corresponds to the thickness of the biofouling
204
layer, ∆P (Pa) is the transmembrane pressure (40 mbar in this study), µ is the viscosity (Pa s)
205
of the filterate, J is the stabilized flux (L/m2 h), Rtotal is the total resistance (m-1), Rmembrane is
206
resistance of the pristine membrane, Rbiofouling
207
Rirreversible is the irreversible fouling resistance. In this study, Rbiofouling layer was considered as
208
95% of (Rbiofouling layer + Rirreversible) (Derlon et al., 2012) and rp was 3 nm (Akhondi et al.,
209
2015).
AC C
EP
202
210 211
3.
Results and Discussion 9
layer
is the biofouling layer resistance and
ACCEPTED MANUSCRIPT 3.1.
Impact of different membranes on GDM performance
213
The performance of GDM filtration using different membranes was examined over an
214
operation period of 55 days. All four types of membranes in the GDM filtration cells
215
displayed a similar flux development profile (Figure 2). Specifically, a rapid drop in flux
216
during the initial filtration period was observed, followed by a slow decline. Although the
217
clean water fluxes were significantly different (Table 1), their respective permeate fluxes
218
were relatively similar when raw seawater had been filtered for 10 days. A slight recovery of
219
permeate fluxes was noticed after 20 days and stabilized permeate fluxes were achieved after
220
about 40 days. Such a four-stage flux development profile was similar to the observations in
221
the previous study (Akhondi et al., 2015).
222
After Day 40, the fluxes for flat sheet PES with a pore size of 100 kD, flat sheet PVDF with a
223
pore size of 0.22 and 0.45 µm, and hollow fibre PVDF with a pore size of 0.1 µm stabilized
224
at 4.5±0.1, 8.4±0.1, 7.3±0.4, and 3.6±0.3 (relatively not stable) L/m2h, respectively. For the
225
flat sheet membranes tested, the two MF PVDF membranes (0.22 and 0.45 µm) achieved
226
comparable permeate fluxes, which were significantly higher than UF PES membrane (100
227
kD). For the MF PVDF membranes tested, the flat sheet membranes (0.22 and 0.45 µm) had
228
better flux performance than the hollow fibre membrane (0.1 µm).
229
Hermia’s model (Hermia, 1985) was employed to elucidate the membrane fouling
230
mechanisms in the dead-end GDM filtration cell, which was explained in detail in the
231
previous study (Akhondi et al., 2015). The experimental data in Figure 2 have been tested
232
with the linearized forms of Hermia’s model, in which the pore constriction and cake
233
filtration models were found to describe the experimental data the best (Figure 3). Obviously,
234
the membrane fouling mechanisms for the four types of tested membranes were relatively
235
dissimilar. Throughout the duration of the experiment with the UF PES membrane (100 kD),
AC C
EP
TE D
M AN U
SC
RI PT
212
10
ACCEPTED MANUSCRIPT pore constriction and cake filtration were simultaneously predominant, similarly to the
237
observations reported in our previous study (Akhondi et al., 2015). For the MF membrane,
238
pore constriction is more predominant during the initial filtration due to the relatively large
239
pore size, while after that, pore constriction and cake filtration were simultaneously
240
predominant. In addition, Figure 3 predicts that the permeability of the hollow fibre MF
241
membrane would continue to decrease with accumulated permeate volume (Figure 3a) and
242
filtration time (Figure 3b). This implies that the hollow fibre membrane may need more time
243
to achieve a stabilized flux compared to the flat sheet membrane.
244
The OCT images in Figure 4 depict the biofouling layer morphology for the four types of
245
membranes at Day 55 and Table 2 summarizes the characteristics of the biofouling layers. It
246
is worth noting that the biofouling layers on the flat sheet PVDF microfiltration membranes
247
were obviously thicker and more porous than that on either PES UF membrane or hollow
248
fibre PVDF membrane. However, the biofouling layers formed on the PES membrane
249
appeared to be rougher than those on the PVDF MF membranes. Compared to the MF
250
membrane, the UF membrane tended to reject more organic substances, which would have
251
promoted bacterial growth and theoretically led to more predation. This would result in
252
higher porosity and roughness of the biofouling layer (Akhondi et al., 2015; Derlon et al.,
253
2013; Derlon et al., 2012). However, the excess organic substances could occupy the void
254
spaces in the biofilm matrix to increase the biofouling layer compactness (i.e, less porosity).
255
On the other hand, the organic molecules having comparable sizes to UF membrane pores
256
could pass through MF membrane, but tended to block UF membranes. It should be also
257
noted that different membrane materials may display different adsorption capabilities of
258
organics. Thus, we propose that the combined effects of these complicated phenomena
259
determined the higher permeate flux of the PVDF MF membranes than that of the PES UF
260
membrane.
AC C
EP
TE D
M AN U
SC
RI PT
236
11
ACCEPTED MANUSCRIPT The feed and permeate DOC (note: this means organic substances less than 0.45 µm) levels,
262
monitored during the flux stabilization period (day 40-55), are shown in Figure S2. The
263
permeate from the flat sheet membrane (0.22 and 0.45 µm PVDF and 100 kD PES
264
membrane) had almost similar DOC content to that of the raw seawater feed. However, the
265
hollow fibre membrane produced the permeate with an increased DOC content. Several
266
mechanisms could explain the increase of DOC in the permeate. First, as shown in Tables 3
267
and 5, the raw seawater contained relatively large amounts of suspended solids (TSS: 39.67
268
mg/L; Turbidity: 5.27±2.88 NTU). These suspended solids will be deposited on the
269
membrane and become integrated into the biofouling layer. Due to biological activity in the
270
biofouling layer, a portion of this deposited material can be hydrolysed or converted,
271
resulting in the formation of dissolved organic compounds, which can permeate the
272
membrane. Similar trends have been observed in the GDM filtration of creek water (Chomiak
273
et al., 2014). The hydrolysis of suspended organic material within the biofilm was suggested
274
to lead to elevated DOC levels in the permeate. Another possible explanation for the elevated
275
DOC content is the carbon fixation activity within the microbial biofilm. Although light was
276
excluded as far as possible, some diffuse light may have penetrated the experimental setup,
277
leading to carbon fixation, transforming carbon dioxide to organic substances and thus,
278
contributing to permeate DOC (Akhondi et al., 2015). Therefore, not surprisingly, the
279
permeate DOC was relatively close to or even higher than the feed DOC.
280
3.2.
281
As discussed in Section 3.1, the suspended solids in the raw seawater (Table 3) may
282
contribute to permeate DOC by decomposition of suspended organics. By prefiltering raw
283
seawater through various meshes, greater-sized suspended solids were expected to be
AC C
EP
TE D
M AN U
SC
RI PT
261
Impact of prefiltration on GDM performance
12
ACCEPTED MANUSCRIPT removed before it was fed to the GDM filtration system. Whether such prefiltration could
285
benefit membrane performance and permeate quality was tested in this study.
286
Table 3 shows that after prefiltering raw seawater with meshes, the TSS contents in the
287
seawater dramatically decreased by 4~5 times. In addition, the prefiltered seawater had less
288
TSS when a finer mesh was used. The flux developments during 55-day GDM filtration (Flat
289
sheet PES membrane, 100 kD) fed with various prefiltered seawaters are presented in Figure
290
S3 and the stabilized fluxes are presented in Table 3. Surprisingly, when the GDM was fed
291
with prefiltered seawater, the stabilized permeate flux dropped by 10-20% compared to that
292
of GDM filtration using raw seawater (Table 2). Although a smaller-sized mesh removed
293
more particulate organic substances, the stabilized permeate flux was lower than that of GDM
294
using the greater-sized mesh, or no mesh. This finding agrees with the observations of Derlon
295
et al. (Derlon et al., 2013) using GDM to treat raw and prefiltered river waters.
296
The OCT images (Figure 5 and Table 4) show that the biofouling layer thickness was 32, 49,
297
and 55 µm and roughness was 1, 10, 10 µm when GDM was fed with the prefiltered seawater
298
at a mesh size of 10, 100, and 1000 µm respectively, which was thinner and of lower
299
roughness than that in the case of raw seawater (Figure 4a and Table 2). In addition, less
300
porosity of the biofouling layer on the membrane was noticed using the smaller-sized mesh as
301
the prefilter compared to using greater-sized mesh or no mesh (Tables 2 and 4). Thus, the
302
permeate flux was more strongly correlated with porous nature of the biofouling layer, rather
303
than the TSS concentration in the feed seawater or the biofilm thickness and roughness.
304
Previous studies have proven that the movement and predation activity of the eukaryotic
305
organisms led to the formation of heterogeneous and porous biofouling layer (Derlon et al.,
306
2013; Derlon et al., 2012). The microscopic observation of biofouling layers on the
307
membrane fed with raw seawater (Figure S4) shows that some eukaryotic organisms had
AC C
EP
TE D
M AN U
SC
RI PT
284
13
ACCEPTED MANUSCRIPT greater size from a few hundred microns to thousand microns. When the seawater was
309
prefiltered, the eukaryotic organisms whose sizes were greater than the mesh size (Figure S4)
310
could be removed from the seawater. Therefore, the extent of movements and predation
311
behaviour by the eukaryotic organisms would be reduced, and this would decrease the
312
heterogeneous nature of the biofouling layer, especially the porosity.
313
To further investigate transport and transmission of the dissolved organic substances, the
314
compositions of dissolved organic substances were analyzed by LC-OCD in the feed and
315
permeate as summarized in Table S1. Similar to the findings in Section 3.1, the permeate
316
DOC was higher than the feed DOC. In the feed and permeate seawater, low molecular
317
weight substances were the major DOC fraction, almost 1.6-1.8 and 2.3-2.6 times of the sum
318
of biopolymers, humic substances, and building blocks respectively. Compared to the feed
319
seawater, the permeate contained more building blocks (300-500 Da) and low molecular
320
weight neutrals and acids ( 0.1 µm) in the raw seawater was removed by the
346
hybrid biofilm-GDM reactor, among which 15% was degraded by the biofilm and 85% was
347
rejected by the membrane. Although the contents of the dissolved organic carbon and major
348
components in the raw seawater, reactor effluent, and permeate were comparably indifferent
349
in the hybrid biofilm-GDM reactor (Table 5), the permeate quality in this reactor was much
350
better than in the filtration cells using prefiltered seawater (Table S1). This suggests that an
351
increased hydraulic retention time was beneficial to improve the removal of the dissolved
352
organic substances during GDM filtration.
353
Assimilable organic carbon (AOC) has been identified as an advanced indicator for
354
assessment of biofouling potential of the seawater (Naidu et al., 2013). The effluent of the
355
hybrid biofilm-GDM reactor contained 203±55 µg/L of AOC, which was significantly higher
AC C
EP
TE D
M AN U
SC
RI PT
332
15
ACCEPTED MANUSCRIPT than that in the raw seawater (73±37 µg/L). Subsequently, the AOC was reduced by almost
357
50% in the GDM system (98±43 µg/L in the permeate). Unfortunately, the hybrid biofilm-
358
GDM reactor under the operating conditions in this study could not effectively remove AOC
359
substances. A detailed study on the degradation, transformation, and permeation of the total
360
organic substances (particulate and dissolved) in GDM filtration is necessary in order to
361
quantitatively describe the carbon balance of the total organic substances and optimize
362
conditions in the GDM process for subsequent seawater RO.
SC
363
RI PT
356
4.
Conclusions
365
In this study we investigated possible measures to optimize the membrane performance and
366
water quality from a GDM filtration system for the treatment of sea water as a pretreatment
367
before RO. The flux development, permeate quality, and biofouling layer structure in the
368
GDM process were examined. Flat sheet MF membranes displayed better performance than
369
either flat sheet UF membrane or hollow fibre MF membranes. Prefiltered seawater did not
370
improve permeate flux of the GDM filtration although organic substances were partly
371
removed. This was attributed to the formation of less porous biofouling layers because some
372
eukaryotic organisms were also removed by prefiltration. A hybrid system comprising a
373
biofilm reactor and a submerged GDM was evaluated. The hybrid biofilm-submerged GDM
374
reactor significantly improved permeate flux compared to the filtration cells. However, GDM
375
filtration could not reduce dissolved organic substances and assimilable organic substances in
376
the seawater, which was possibly attributed to the decomposition of suspended organic
377
material and carbon fixation by biofilm. Further studies of the dynamic organic carbon
378
balance in GDM systems are necessary.
AC C
EP
TE D
M AN U
364
379
16
ACCEPTED MANUSCRIPT Acknowledgements
381
The authors are thankful to Dr Linbo Liu, Dr Xin Liu, Dr Weiyi Li, Jun Gu for their help in
382
OCT measurement, Dr Frederik Hammes for his generous sharing on AOC measurement
383
method, and Weikang Lim for his kindly help in taking seawater from the desalination plant.
384
The Economic Development Board is acknowledged for funding the Singapore Membrane
385
Technology Centre (SMTC), Nanyang Technological University.
386
References:
387 388 389 390
Akhondi, E., Wicaksana, F., Fane, A. G., 2014. Evaluation of fouling deposition, fouling reversibility and energy consumption of submerged hollow fiber membrane systems with periodic backwash. J. Membrane Sci. 452, 319-331. http://dx.doi.org/10.1016/j.memsci.2013.10.031
391 392 393 394 395
Akhondi, E., Wu, B., Sun, S. Y., Marxer, B., Lim, W. K., Gu, J., Liu, L. B., Burkhardt, M., McDougald, D., Pronk, W., Fane, A. G., 2015. Gravity-driven membrane filtration as pretreatment for seawater reverse osmosis: Linking biofouling layer morphology with flux stabilization. Water Res. 70, 158-173. http://dx.doi.org/10.1016/j.watres.2014.12.001
396 397
APHA (Ed.). (1998). Standard methods for the examination of water and wastewater. Washington, D.C.: American Public Health Association (APHA).
398 399 400 401
Bae, H., Kim, H., Jeong, S., Lee, S., 2011. Changes in the relative abundance of biofilmforming bacteria by conventional sand-filtration and microfiltration as pretreatments for seawater reverse osmosis desalination. Desalination 273 (2-3), 258-266. http://dx.doi.org/10.1016/j.desal.2010.12.030
402 403 404 405
Chomiak, A., Mimoso, J., Koetzsch, S., Sinnet, B., Pronk, W., Derlon, N., Morgenroth, E., 2014. Biofilm formation and permeate quality improvement in Gravity Driven Membrane ultrafiltration. Water Sci. Tech.-W. Sup. 14 (2), 274-282. http://dx.doi.org/10.2166/ws.2013.197
406 407 408 409
Derlon, N., Koch, N., Eugster, B., Posch, T., Pernthaler, J., Pronk, W., Morgenroth, E., 2013. Activity of metazoa governs biofilm structure formation and enhances permeate flux during Gravity-Driven Membrane (GDM) filtration. Water Res. 47 (6), 2085-2095. http://dx.doi.org/10.1016/j.watres.2013.01.033
410 411 412
Derlon, N., Peter-Varbanets, M., Scheidegger, A., Pronk, W., Morgenroth, E., 2012. Predation influences the structure of biofilm developed on ultrafiltration membranes. Water Res. 46 (10), 3323-3333. http://dx.doi.org/10.1016/j.watres.2012.03.031
AC C
EP
TE D
M AN U
SC
RI PT
380
17
ACCEPTED MANUSCRIPT Elimelech, M., Phillip, W. A., 2011. The Future of Seawater Desalination: Energy, Technology, and the Environment. Science 333 (6043), 712-717. http://dx.doi.org/10.1126/science.1200488
416 417 418
Greenlee, L. F., Lawler, D. F., Freeman, B. D., Marrot, B., Moulin, P., 2009. Reverse osmosis desalination: Water sources, technology, and today's challenges. Water Res. 43 (9), 2317-2348. http://dx.doi.org/10.1016/j.watres.2009.03.010
419 420 421 422
Hammes, F., Berney, M., Wang, Y. Y., Vital, M., Koster, O., Egli, T., 2008. Flow-cytometric total bacterial cell counts as a descriptive microbiological parameter for drinking water treatment processes. Water Res. 42 (1-2), 269-277. http://dx.doi.org/10.1016/j.watres.2007.07.009
423 424 425
Hammes, F. A., Egli, T., 2005. New method for assimilable organic carbon determination using flow-cytometric enumeration and a natural microbial consortium as inoculum. Environ. Sci. Technol. 39 (9), 3289-3294. http://dx.doi.org/10.1021/Es048277c
426 427 428
Hermia, J., 1985. Blocking filtration----Application to non-newtonian fluids, in: A. Rushton (ed.) Mathematical models and design methods in solid-liquid separation, Dordrecht, Martinus Nijhoff Publishers.
429 430 431
Huang, H. O., Cho, H., Schwab, K., Jacangelo, J. G., 2011. Effects of feedwater pretreatment on the removal of organic microconstituents by a low fouling reverse osmosis membrane. Desalination 281, 446-454. http://dx.doi.org/10.1016/j.desal.2011.08.018
432 433 434 435
Huber, S. A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography - organic carbon detection organic nitrogen detection (LC-OCD-OND). Water Res. 45 (2), 879-885. http://dx.doi.org/10.1016/j.watres.2010.09.023
436 437 438 439
Katsoufidou, K., Yiantsios, S. G., Karabelas, A. J., 2005. A study of ultrafiltration membrane fouling by humic acids and flux recovery by backwashing: Experiments and modeling. Journal of Membrane Science 266 (1-2), 40-50. http://dx.doi.org/DOI 10.1016/j.memsci.2005.05.009
440 441 442
Kaya, C., Sert, G., Kabay, N., Arda, M., Yuksel, M., Egemen, O., 2015. Pre-treatment with nanofiltration (NF) in seawater desalination-Preliminary integrated membrane tests in Urla, Turkey. Desalination 369, 10-17. http://dx.doi.org/10.1016/j.desal.2015.04.029
443 444 445
Klein, T., Zihlmann, D., Derlon, N., Isaacson, C., Szivak, I., Weissbrodt, D. G., Pronk, W., 2016. Biological control of biofilms on membranes by metazoans. Water Res. 88 2029. http://dx.doi.org/10.1016/j.watres.2015.09.050
446 447 448
Knops, F., van Hoof, S., Futselaar, H., Broens, L., 2007. Economic evaluation of a new ultrafiltration membrane for pretreatment of seawater reverse osmosis. Desalination 203 (1-3), 300-306. http://dx.doi.org/10.1016/j.desal.2006.04.013
449 450 451
Naidu, G., Jeong, S., Vigneswaran, S., Rice, S. A., 2013. Microbial activity in biofilter used as a pretreatment for seawater desalination. Desalination 309, 254-260. http://dx.doi.org/10.1016/j.desal.2012.10.016
AC C
EP
TE D
M AN U
SC
RI PT
413 414 415
18
ACCEPTED MANUSCRIPT Passow, U., Alldredge, A. L., 1995. A dye-binding assay for the spectrophotometric measurement of transparent exopolymer particles (TEP). Limnol. Oceanogr. 40 (7), 1326-1335.
455 456 457
Pearce, G. K., 2008. UF/MF pre-treatment to RO in seawater and wastewater reuse applications: a comparison of energy costs. Desalination 222 (1-3), 66-73. http://dx.doi.org/10.1016/j.desa1.2007.05.029
458 459 460
Peter-Varbanets, M., Gujer, W., Pronk, W., 2012. Intermittent operation of ultra-low pressure ultrafiltration for decentralized drinking water treatment. Water Res. 46 (10), 32723282. http://dx.doi.org/10.1016/j.watres.2012.03.020
461 462 463
Peter-Varbanets, M., Hammes, F., Vital, M., Pronk, W., 2010. Stabilization of flux during dead-end ultra-low pressure ultrafiltration. Water Res. 44 (12), 3607-3616. http://dx.doi.org/10.1016/j.watres.2010.04.020
464 465 466
Peter-Varbanets, M., Margot, J., Traber, J., Pronk, W., 2011. Mechanisms of membrane fouling during ultra-low pressure ultrafiltration. J. Membrane Sci. 377 (1-2), 42-53. http://dx.doi.org/10.1016/j.memsci.2011.03.029
467 468 469
Peter-Varbanets, M., Zurbrugg, C., Swartz, C., Pronk, W., 2009. Decentralized systems for potable water and the potential of membrane technology. Water Res. 43 (2), 245265. http://dx.doi.org/10.1016/j.watres.2008.10.030
470 471 472
Xu, J., Ruan, L. G., Wang, X., Jiang, Y. Y., Gao, L. X., Gao, J. C., 2012. Ultrafiltration as pretreatment of seawater desalination: Critical flux, rejection and resistance analysis. Sep. Purif. Technol. 85, 45-53. http://dx.doi.org/10.1016/j.seppur.2011.09.038
SC
M AN U
TE D EP
475
AC C
473 474
RI PT
452 453 454
19
ACCEPTED MANUSCRIPT Table 1. The properties of the seawater and membranes used in this study
Raw seawater*
pH
Conductivity (mS/cm)
DO (mg/L)
7.93±0.16
47.44±1.13
5.49±1.24 Hollow fibre membrane
RI PT
Flat sheet membrane PES
PVDF
PVDF
PVDF
Brand
Microdyn-Nadir
Millipore
Millipore
GE
Pore size
100 kD
0.22 µm
0.45 µm
0.1 µm
Area (m2)
0.0023
0.0023
0.0023
0.0011
Contact angle (º)**
81±3
77±1
74±2
76±2
Clean water flux at 40 mbar (L/m2 h)
27.3±0.4
227.1±1.5
679.2±46.2
795.0±18.3
M AN U
SC
Materials
* The pH and conductivity of the feed and permeate water were determined by a pH/conductivity meter (Mettler Toledo, Switerzland). The dissolved oxygen (DO) in the feed tank was determined by a DO meter (YSI, USA).
AC C
EP
TE D
** The contact angle was measured by contact angle measurement systems (Dataphysics, Germany);
ACCEPTED MANUSCRIPT
Table 2. Characteristics of biofouling layer on the different membranes in GDM filtration cells (Day 55). Averaged thickness (µm)
Absolute roughness (µm)
Relative roughness (µm)
Flat sheet PES (100 kD)
62
14
0.20
Flat sheet PVDF (0.22 µm)
127
12
0.09
Flat sheet PVDF (0.45 µm)
111
4
0.04
94.2
7.3
Hollow fibre PVDF (0.1 µm)
77
8
0.08
90.8
3.6*
AC C
EP
Stable flux (L/m2 h)
91.4
4.5
95.0
8.4
SC
M AN U
TE D
* not stable.
Calculated porosity (%)
RI PT
Membrane type
ACCEPTED MANUSCRIPT Table 3. The performance of GDM after seawater prefiltration with different meshes After prefiltration with mesh Mesh size at 10 µm
Mesh size at 100 µm
Mesh size at 1000 µm
7.65
8.25
Feed TSS (mg/L)
39.67
6.25
Stable flux (LMH)*
4.5±0.1
3.6±0.8
RI PT
Raw seawater
3.9±0.5
4.0±0.2
AC C
EP
TE D
M AN U
SC
* The GDM filtration was performed for 55 days. The GDM flux presented here was the averaged value of the fluxes during day 40-55.
ACCEPTED MANUSCRIPT Table 4. Characteristics of of biofouling layers in GDM filtration cells fed with different prefiltrated seawaters at Day 55. Averaged thickness (µm)
Absolute roughness (µm)
Relative roughness (µm)
Calculated porosity (%)
Mesh size at 10 µm
32
1
0.03
87.2
Mesh size at 100 µm
49
10
0.20
89.7
Mesh size at 1000 µm
55
10
0.18
90.3
AC C
EP
TE D
M AN U
SC
RI PT
After filtration with mesh
ACCEPTED MANUSCRIPT
Effluent
Permeate
Turbidity (NTU)
5.27±2.88
1.13±0.44
0.13±0.04
TEP (mg/L)
1.44±0.55
1.22±0.19
0.69±0.19
DOC (mg/L)
2.34±0.46
2.34±0.51
2.39±0.72
Biopolymers (mg/L)
0.08±0.02
0.13±0.04
0.09±0.02
Humic substances (mg/L)
0.48±0.05
0.52±0.05
0.53±0.03
Building blocks (mg/L)
0.20±0.02
0.21±0.04
0.21±0.02
LWM neutrals (mg/L)
1.48±0.44
1.39±0.45
1.49±0.76
LWM acids (mg/L)
n.q.
n.q.
n.q.
AOC (µg/L)
73±37
M AN U
RI PT
Raw seawater feed
SC
Table 5. Comparison of organic substances in feed, effluent, and permeate in the hybrid biofilm-GDM reactor (n=7).
AC C
EP
TE D
n.q. = not quantifiable (< 1ppb calculated)
203±55
98±43
ACCEPTED MANUSCRIPT
Submerged GMD
Biofilm reactor
RI PT
0.4 m
Kaldness carrier
Membrane module
M AN U
Permeate tank
SC
Feed tank
AC C
EP
TE D
Figure 1. Schematic diagram of the hybrid biofilm-submerged GDM reactor.
ACCEPTED MANUSCRIPT
30 Flat sheet PES (100 kD)
Flat sheet PVDF (0.22 µm)
Flat shhet PVDF (0.45 µm)
Hollow fibre PVDF (0.1 µm)
1000
800
RI PT
Flux (LMH)
600 400 200 0
10
0
0
10
0.5 1 Time (Day)
1.5
SC
0
M AN U
Flux (LMH)
20
20
30
40
50
60
TE D
Time (Day)
AC C
EP
Figure 2. Flux developments during GDM filtration using different membranes.
ACCEPTED MANUSCRIPT
a
n=0 (cake filtration)
RI PT
4
SC
2
M AN U
Time/Volume, t/V (Day/L)
6
Flat sheet PES (100kD) Flat sheet PVDF (0.22 µm) Flat sheet PVDF (0.45 µm) Hollow fibre PVDF (0.1 µm)
0 0
10
20
30
Volume, V (L)
b
TE D
n=3/2 (pore constriction)
EP
4
2
AC C
Time/Volume, t/V (Day/L)
6
Flat sheet PES (100kD) Flat sheet PVDF (0.22 µm) Flat sheet PVDF (0.45 µm) Hollow fibre PVDF (0.1 µm)
0
0
20
40
60
Time, t (Day) Figure 3. Experimentally determined t/V data plotted as a function of filtration volume (a) and filtration time (b).
ACCEPTED MANUSCRIPT
(b) Flat sheet PVDF (0.22 µm)
RI PT
(a) Flat sheet PES (100 kD)
TE D
M AN U
SC
(c) Flat sheet PVDF (0.45 µm) (d) Hollow fibre PVDF (0.1 µm)
AC C
EP
Figure 4. Morphologies of biofouling layers on the different membranes in GDM filtration cells (Day 55).
ACCEPTED MANUSCRIPT
(b)
(c)
RI PT
(a)
AC C
EP
TE D
M AN U
SC
Figure 5. Morphologies of biofouling layers in GDM filtration cells fed with different prefiltrated seawater at Day 55. (a) Mesh size at 10 µm; (b) Mesh size at 100 µm; (c) Mesh size at 1000 µm.
ACCEPTED MANUSCRIPT
20 Membrane module 1
Membrane module 2
RI PT
Flux (LMH)
15
10
0 0
40
M AN U
SC
5
80 Time (Day)
120
160
AC C
EP
TE D
Figure 6. Flux developments in the hybrid biofilm-submerged GDM reactor.
ACCEPTED MANUSCRIPT > The use of MF in GDM filtration achieved higher stabilized flux than that of UF. > Pore constriction and cake filtration were major fouling mechanisms of GDM system. > Prefiltering of seawater decreased flux due to removal of eukaryotes. > More porous biofouling layer led to higher stabilized flux.
AC C
EP
TE D
M AN U
SC
RI PT
> Integration of biofilm reactor with submerged GDM reactor improved flux.