Journal of Membrane Science 382 (2011) 41–49

Contents lists available at ScienceDirect

Journal of Membrane Science

journal homepage: www.elsevier.com/locate/memsci

Impact of membrane configuration on fouling in anaerobic membrane bioreactors

I. Martin-Garcia a , V. Monsalvo b , M. Pidou c , P. Le-Clech d , S.J. Judd a , E.J. McAdam a,∗ , B. Jefferson a
a
Cranfield Water Science Institute, Building 39, Cranfield University, Bedfordshire MK43 0AL, UK
b
Dpto. Ingeniería Química, Universidad Autónoma de Madrid, Madrid 28049, Spain
c
Advanced Water Management Centre, The University of Queensland, Brisbane, Australia
d
UNESCO Centre for Membrane Science and Technology, School of Chemical Engineering, The University of New South Wales, Sydney, Australia

a r t i c l e i n f o a b s t r a c t

Article history: The filtration performance of flocculated and granulated configured anaerobic membrane bioreactors
Received 14 April 2011 (MBR) treating domestic wastewater has been evaluated and compared to conventional aerobic MBR.
Received in revised form 19 July 2011 Immersed hollow fibre (HF) and external tubular membrane geometries were additionally compared
Accepted 26 July 2011
with the latter operated in both pumped and gas-lift mode. After 200 d of operation, both granular
Available online 4 August 2011
and flocculated anaerobic MBR (AnMBR) suspensions were characterised by an increased population
of colloidal particles while the aerobic MBR retained a unimodal particle size distribution with a d50 of
Keywords:
20 ␮m. Consequently, the flocculated AnMBR supernatant was characterised by a soluble microbial prod-
Bubbling
Scouring
uct (SMP) concentration ca. 500% higher than the aerobic MBR, such that the lowest critical fluxes for
Critical flux both HF and tubular membranes were recorded for the AnMBR. In comparison, the granulated AnMBR
Crossflow sludge was characterised by a low mixed liquor suspended solids concentration and an SMP concen-
Submerged tration below 50% that of the flocculated anaerobic MBR. Consequently, similar fluxes to those of the
Sidestream aerobic MBR were achieved with the granulated anaerobic sludge using immersed HF membranes. Oper-
ating external tubular membranes in gas-lift appeared less effective for the granular AnMBR than the
Aerobic MBR. However, critical fluxes >40 L m−2 h−1 were achieved using pumped mode. Results suggest
granular AnMBR systems to be most suited to domestic wastewater treatment using either immersed HF
membranes or external tubular membranes in pumped crossflow mode.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction in external tubular membranes operated under pumped [2–5] or

Anaerobic treatment of domestic wastewater is constrained by
the low organic strength of the wastewater, the quality of the avail-
able carbon and the high Ks value associated with the anaerobic
community which impair bacterial growth and effective treatment.
Anaerobic membrane bioreactor (AnMBR) technology can enhance
effluent quality from increased rejection of solids and colloidal
organic matter, and also achieve higher biomass concentrations by
minimising washout. AnMBR thus increases the viability of anaer-
obic treatment, offering treatment at reduced energy demand and
sludge yield over that of aerobic processes, for which the aeration
energy demand is significant. However, AnMBR studies undertaken
to date have generally found that membrane fouling is significant
due to the high concentration of colloidal material, inclusive of pro-
tein and polysaccharide, the concentrations of which are further
exacerbated at extended solids retention times (SRTs) [1].
Previous AnMBR fouling studies have primarily investigated
the effect of pumped crossflow velocity on filtration performance
∗ Corresponding author. Tel.: +44 1234 754546.
E-mail address: e.mcadam@cranfield.ac.uk (E.J. McAdam).
gas-lift [6–10] condition. While gas sparging in immersed mem-
brane systems is known to present the most cost effective means
of fouling control in aerobic wastewater treatment applications,
permeability data for anaerobic immersed flat-sheet and hollow
fibre (HF) geometries is more limited [11–14]. Comparisons of liq-
uid pumping and gas sparging in external tubular membranes [15]
and gas sparging in immersed flat-sheet and HF membranes have
been undertaken in AnMBR [11]; however, a direct comparison of
external and immersed configured AnMBRs has yet to be made.
Direct comparison of current AnMBR studies also remains
challenging as two reactor configurations now predominate: con-
ventional completely mixed flocculated reactors [8–10] and upflow
anaerobic sludge blanket (UASB) reactors [10,12,14]. In the latter
case, membrane filtration is sited at the head of the UASB or in
a subsequent stage, thus the membrane is challenged with the
sludge supernatant. While UASB based AnMBR are less studied,
lower irreversible fouling than completely mixed reactors has been
demonstrated in a comparative study on the treatment of black-
water [10]. The aim of this present study is therefore to provide a
direct comparison of external and immersed membrane operation
in both flocculated and UASB configured AnMBR treating domestic
wastewater. Previous AnMBR studies primarily utilised synthetic

0376-7388/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2011.07.042
42 I. Martin-Garcia et al. / Journal of Membrane Science 382 (2011) 41–49
Table 1
Module characteristics for the tubular and hollow fibre membranes tested.
Parameter Unit Experimental reactor
Pilot plant
Geometry Tubular HF
Filtration area m2 0.022 0.93
Material – PVDF PVDF
Pore size ␮m 0.03 0.04
Module length m 1 0.7
Fibre OD/LD mm 8 1.9/0.8
CSA m2 – 0.0074
Packing density m2 /m3 – 300
ID, internal diameter; OD, outer diameter; LD, lumen diameter; HF, hollow fibre.
wastewaters comprising of soluble carbon sources [9,11,14]. While
such studies present significant insight, real wastewaters present
more complex colloidal and particulate matrices. Furthermore, to
date most published studies have used an external thermal source
to stabilise reactor operation to mesophilic or thermophilic condi-
tions [2–4]. Energetic modelling has demonstrated that methane
generated from AnMBR treating domestic wastewater is insuffi-
cient to support heating the influent wastewater due to the low
organic strength and high liquid flow rate [16]. Consequently,
to provide environmentally relevant conditions, the flocculated
AnMBR, granular AnMBR (G-AnMBR) and a reference aerobic MBR
(AeMBR) were operated on real domestic wastewater and without
external temperature control.
2. Materials and methods
2.1. Pilot plants
Three MBR comprising suspended growth (flocculated) aer-
obic and anaerobic MBR and a granular anaerobic MBR
were operated in parallel for 250 d fed with wastewater
having mean chemical oxygen demand (COD), biochemical
oxygen demand (BOD5 ), suspended solids (SS) and ammonia
concentrations of 338 mg COD L−1 (range 197–553 mg COD L−1 ),
167 mg BOD L−1 (range 155–285 mg BOD L−1 ), 84 mg SS L−1 (range
51–186 mg SS L−1 ) and 35 mg NH4 -N L−1 (range 15–48 mg NH4 -
N L−1 ) respectively. The AeMBR and AnMBR pilot plants comprised
two tanks of 1.5 m3 total volume divided between the biological
(1 m3 ) and membrane (0.5 m3 ) compartments. Wastewater was
introduced in the biological tank through a floating valve which
controlled the level of sludge at a height of 1.5 m, creating a total
working volume of 1.2 m3 . The AeMBR was continuously aerated
at a flow rate of 50–100 L min−1 through cylindrical fine bubble
diffusers at the biological tank base. The AnMBR tank was sealed
with a PVC lid. Biomass was cycled between both compartments
through external pumps, and an additional pump was operated
in cycles of 15 min on/15 min off to mix the reactor contents by
recycling biomass through venturi nozzles located at the base of
this chamber. Both flocculated MBRs were fitted with polyvinyld-
ifluoride (PVDF) HF membranes of 12.5 m2 surface area, 0.08 ␮m
nominal pore size, 1 m length, 0.0177 m2 cross sectional area and
a packing density of 710 m2 m−3 (Table 1). Permeate was continu-
ously extracted using peristaltic pumps (620 Du, Watson Marlow,
Falmouth) to provide an instantaneous flux of 6 L m−2 h−1 . Mem-
brane fouling was controlled by continuous gas sparging at gas
velocities ranging from 0.02 m s−1 to 0.078 m s−1 in both aerobic
and anaerobic systems, the latter employing nitrogen enriched air
(>99% N2 ) generated from an industrial gas membrane nitrogen
generator unit (Atlas Copco, Herts, UK) fed with compressed air.
During the 250-d trial, 12 L of sludge per day were withdrawn from
the suspended aerobic and anaerobic MBRs in order to set SRT based
on bioreactor volume at 100 d. This value could be taken as repre-
MBR pilot plant
AeMBR AnMBR G-AnMBR
HF HF HF
12.5 12.5 0.93
PVDF PVDF PVDF
0.08 0.08 0.04
1 1 0.7
1.3 1.3 1.9/0.8
0.0177 0.0177 0.0074
710 710 300
sentative for the mean residence time in the anaerobic system since
no sludge accumulation was observed. For the AeMBR, which was
started with a low MLSS concentration of 1 g MLSS L−1 , the dynamic
SRT reached a value of 80–85 d for the last 100 d of operation tak-
ing into account biomass accumulation within the reactor for the
estimation of the sludge age.
The granular anaerobic MBR (G-AnMBR) consisted of a 85 L per-
spex cylindrical vessel (1.75 m depth × 0.25 m diameter) seeded
with 40 L of granular sludge from an UASB sited at a sugar refinery
(British Sugar, Suffolk, UK), with a 38 L perspex cylinder housing
the membrane (Fig. 1). The granular UASB contactor was operated
in the expanded mode (EGSB) using an external recirculation pump
(620s, Watson Marlow, Falmouth, UK) to maintain a superficial
upflow velocity (Vup ) of less than 1 m h−1 . Effluent from the granule
contactor was recirculated through the membrane tank during per-
meation. At the fixed upflow velocity, the granule bed expanded to
a depth of 0.60 m, or approx. 30% of the column depth; the effluent
entering into the membrane tank was characterised by a relatively
low solids fraction compared to the flocculated reactors. A PVDF
hollow-fibre module with a surface area of 0.93 m2 and nominal
pore size 0.04 ␮m, 1 m length, cross sectional area of 0.0074 m2
and a packing density of 300 m2 m−3 was used in this study at a set
instantaneous flux (J) of 6 L m−2 h−1 to provide a biotank hydraulic
retention time (HRT) of 16 h. In the G-AnMBR membrane fouling
was controlled by gas sparging nitrogen at gas velocities ranging
from 0 to 0.057 m s−1 . During the 250-d trial, no granular biomass
was withdrawn from the biotank; samples were only collected from
the membrane tank for analysis.
2.2. Short-term fouling experiments
Short term fouling experiments for the AeMBR and AnMBR
were conducted in an external 38 L cylindrical tank (0.20 m diam-
eter × 1.2 m height) using a 30 L slurry from either reactor (Fig. 2).
For immersed trials, an identical PVDF HF membrane to that used
in the granular AnMBR was tested (Table 1). A 0.022 m2 tubu-
lar PVDF membrane with a nominal pore size of 0.03 ␮m was
used for external (sidestream) experiments. The external tubu-
lar membrane was operated in both gas lift and pumped mode.
During immersed and gas lift operation, nitrogen-enriched air for
anaerobic experiments or natural air for aerobic experiments were
employed for membrane scouring and biomass mixing. For gas
lift experiments, the base of the membrane inlet was connected
through a ‘T’ junction to provide a port for gas injection. The same
range of specific gas demand, 0.2–1.2 m3 m−2 h−1 , was used for
both membrane geometries. For pumped crossflow operation, a
centrifugal pump (EHEIM Gmbh and Co., Deizisau, Germany) was
employed to generate a cross flow velocity (CFV) controlled by a
throttling valve sited upstream of the membrane module to val-
ues between 0.4 and 2 m s−1 . For all studies both the retentate
mixed liquor and permeate product streams from the membrane
module were returned to the tank to maintain a constant mixed
 I. Martin-Garcia et al. / Journal of Membrane Science 382 (2011) 41–49 43

Fig. 1. Schematic representation of the MBR pilot plants. The aerobic MBR (AeMBR, top) as well as the anaerobic MBR (AnMBR, middle) and granular anaerobic MBR (G-
AnMBR, bottom) which comprised of separate membrane and biological tanks. In all cases membrane pumps were employed to cycle sludge between both chambers and in
the case of the suspended AnMBR to mix reactor contents. Nitrogen gas was used as the sparge gas for anaerobic environments.

liquor suspended solids (MLSS) concentration. The transmembrane
pressure (TMP) in the immersed system was recorded from pres-
sure transducers (Druck Ltd., Leicester) installed in the permeate
line. For the tubular membrane, additional pressure transducers
were installed upstream and downstream of the retentate chan-
nel to provide accurate TMP determination. The pressure output
signal (4–20 mA) was logged via a 16-bit analogue to digital con-
verter (Pico Technology Ltd., St Neots, Cambridgeshire) onto a
personal computer. Fouling propensity was evaluated by compar-
ing fouling rates and critical fluxes obtained from the flux-step
method [17], in which membrane flux was incrementally increased
in steps of 2–4 L m−2 h−1 with step durations of 15 min and the
change in transmembrane pressure (dP/dt) monitored. The criti-
cal flux was estimated as the highest flux at which no increase in
transmembrane pressure was observed (dP/dt = 0) [8]. The sensi-
tivity of the pressure-transducer used in this study resided below
0.25% of range in accordance with externally validated calibra-
tion. Therefore while the ‘strict form’ of critical flux was applied,
a reading of absolute zero (dP/dt = 0) did not guarantee zero depo-
sition but rather the point of minimum deposition. Between each
critical flux test, test membranes were cleaned overnight with
1% sodium hypochlorite and the clean water permeability (K)
measured to evaluate permeability recovery prior to the next
trial.
44 I. Martin-Garcia et al. / Journal of Membrane Science 382 (2011) 41–49

10
(a) AeMBR

Volume fraction (%)

8
200 days
6 30 days

0
0.1 1 10 100 1000 10000

Particle size, dP (µm)

10
(b)

Volume Fraction (%)

AnMBR
8
200 days
6
30 days
Fig. 2. Experimental reactor used for short-term experiments. Two membrane con-
figurations were used; an externally configured (side-stream) tubular membrane 4
and an immersed hollow-fibre membrane. Nitrogen gas was used as the sparge gas
for anaerobic sludge during immersed and tubular gas-lift membrane trials. 2

0
0.1 1 10 100 1000 10000
2.3. Analytical procedures
Particle size, dP (µm)
Suspended solids, MLSS, mixed liquor volatile suspended solids
(MLVSS) and BOD5 were measured according to Standard Meth-
ods [18]. Ammoniacal nitrogen (NH4 -N) and COD were analysed 10
(c) G-AnMBR
using Merck pre-prepared cell tests (VWR International, Lutter-
Volume fraction (%)

worth, UK). Particle size was determined using an integrated laser
diffractor by recirculating suspensions through the optical unit of
a Malvern Mastersizer 2000 (Malvern Instrument Ltd., Worcester-
shire, UK), yielding floc sizes as d10 , d50 and d90 values respectively
defining the equivalent diameter at which 10%, 50% and 90% of the
suspended particles are smaller. A coefficient of variance (CV) of
0.02 was determined for a fixed particle size of 0.5 ␮m (Latex micro-
spheres, Duke Scientific, US). Sludge supernatant was obtained by
centrifuging the biomass at 10,000 × g for 15 min followed by filtra-
tion of decanted liquid with a 1 ␮m rated filter [1]. Serial molecular
weight fractionation of the supernatant was conducted using a
400 mL Amicon 8400 dead-end stirred cell (Millipore, USA) with
76 mm diameter polyethersulphone UF membranes of nominal
molecular weights 10, 100, and 500 kDa (Millipore, USA). Frac-
tionations were conducted at a constant shear (through magnetic
stirring) under an applied pressure of 2 Barg and a filtrate/retentate
ratio of 0.4, using laboratory grade nitrogen gas (>99%). Protein and
carbohydrate concentrations were determined according to the
respective adapted methods of [19] and [20]. Bovine serum albumin
(BSA) and d-glucose were used as standards for the determination
of protein and carbohydrate respectively. At a fixed concentration
of 20 mg L−1 , CV of <0.01 and 0.02 were determined for d-glucose
and BSA respectively.
3. Results and discussion
3.1. Overall bioreactor performance
The AnMBR was operated at a HRT and sludge retention time
(SRT) of 16 h and 100 d respectively (Table 2). The G-AnMBR
operated at the same HRT but infinite SRT, with no wasting of
the granular biomass during the 250-d trial. Influent wastewater
consisted of settled sewage. Particle size in the feed wastewa-
ter was characterised by a d50 of 91 ␮m. In the AeMBR, started
without external seed material, COD removal improved from 85%
to over 95% as MLSS concentration increased from 1 g MLSS L−1
to the final average concentration of 8.7 g MLSS L−1 , while BOD
8 200 days
30 days
6
Feed
4
2
0
0.1 1 10 100 1000 10000
Particle size, dP (µm)
Fig. 3. Particle size distributions following 30 d and 200 d operation of the (a) aerobic
MBR (AeMBR), (b) anaerobic MBR (AnMBR) and (c) granular anaerobic MBR (G-
AnMBR). Particle size distribution of the feedwater is included for reference (c).
removal was consistently >95%. Both anaerobic systems presented
similar behaviour in terms of treatment performance, with COD
removal increasing from 80% initially following inoculation with
no acclimation up to 90% by the end of the study. Effluent BOD
ranged between 5 mg BOD L−1 and 15 mg BOD L−1 , yielding a sta-
ble 90–95% removal. In the G-AnMBR, the Vup was maintained
below 1 m h−1 to reduce the concentration of solids in the over-
flow of the biological tank and subsequently minimise solids
loading onto the membrane tank. However, solids concentra-
tion in the membrane tank gradually increased from an initial
concentration of approximately 0.1 g SS L−1 , commensurate with
the feedwater, up to 0.6 g MLSS L−1 by the end of the study. In
the seeded AnMBR, while MLSS concentration ranged between
6.6 g MLSS L−1 and 9.6 g MLSS L−1 , the MLSS concentration sta-
bilised at 7.7 g MLSS L−1 .
3.2. Evolution of particle size distribution
The particle measurement after 30 d of operation indicated
unimodal distributions for both AeMBR and AnMBRs with corre-
sponding d50 values of 32 ␮m and 43 ␮m respectively (Fig. 3 and
Table 2). In contrast, a bimodal distribution was recorded for the
sludge in the G-AnMBR membrane chamber with principal peaks
 I. Martin-Garcia et al. / Journal of Membrane Science 382 (2011) 41–49 45

Table 2
Summary of operational parameters, bioreactor performance and bulk characterisation data from the AeMBR, AnMBR and G-AnMBR.

Parameter Unit AeMBR AnMBR G-AnMBR

Bioreactor operation and treatment performance

SRT d 100 100 –
HRT h 16 16 16
COD removal % 95 84 86
BOD removal % 97 93 95
Bulk characterisation
MLSS g l−1 8.7 7.7 0.1–0.6
PSD
d10 ␮m a
11.8 /5.0 b
15.1 /1.4 a b
1.1a /0.2b
d50 ␮m 32.4a /14.0b 42.8a /7.6b 60.3a /1.4b
d90 ␮m 69.2a /34.7b 91.2a /19.9b 181.9a /91.2b
SMPCOD mg l−1 99 ± 28 598 ± 150 198 ± 73
SMP fractionation
<1 ␮m–500 kDa % 43 87 –
500–100 kDa % 2 0 –
100–10 kDa % 2 4 –
<10 kDa % 53 9 –
SMPc mg l−1 18 ± 9 47 ± 14 18 ± 12
SMPp mg l−1 18 ± 13 108 ± 27 50 ± 21
SMP (P/C) – 1.0 2.3 2.8
a
Particle size recorded at 30 d operation.
b
Particle size recorded at 200 d operation.

at 138 ␮m and 0.6 ␮m, similar to the feedwater which was char-
acterised by peaks at 138 ␮m and 1.1 ␮m. The similarity can be
attributed to the limited solids retention exhibited by the fluidised
granular bed.
Differences in transient aggregate size were observed between
the AeMBR and AnMBR systems. The quasi-normal distribution ini-
tially recorded at 30 d in the AeMBR was maintained at 220 d of
operation, while the median size decreased from a d50 of 32 ␮m
to 14 ␮m. While SRT had not stabilised at 30 d of operation, this
finding is in agreement with the study of Huang et al. [21] where
lower aggregate sizes have been recorded in anaerobic MBRs oper-
ated at longer SRT. Following 200 d of flocculated AnMBR operation,
a bimodal distribution was clearly demonstrated with peaks cen-
tred at 11.5 ␮m and 1.7 ␮m particle size, the latter being indicative
of colloidal generation linked to either floc erosion or poor floc-
culation [22]; the accumulation of high molecular weight organic
degradation by-products could therefore partially contribute to the
skew observed. The bimodal particle size distribution recorded in
the G-AnMBR also indicated lower aggregate sizes; the distribu-
tion recorded at 220 d operation was dominated by a peak of >5%
centred at 1.7 ␮m. However, the higher colloidal content in the
AnMBR and G-AnMBR in comparison to the AeMBR commonly
derives from the higher degree of dispersive growth, frequently
reported in anaerobic systems [15,23] and is associated with high
fouling propensities [12,15,24]. A recent study comparing the per-
formance of a VFA and glucose/VFA-fed anaerobic MBR revealed the
latter to provide a higher fouling propensity as a result of the higher
growth of single cell acidogenic bacteria, was evidenced by a reduc-
tion in measured particle size and supported by light microscopy
observations [8].
3.3. SMP concentration and composition
108 ± 27 mg SMPP L−1 and 47 ± 14 mg SMPC L−1 . Speciation of the
supernatant produced from the G-AnMBR indicated lower organic
concentrations in the SMPCOD , SMPP and SMPC fractions compared
to the flocculated AnMBR at 198 ± 73 mg L−1 , 50 ± 21 mg L−1
and 18 ± 12 mg L−1 respectively. Huang et al. [21] identified that
by operating anaerobic MBR at long SRT, higher protein and
carbohydrate SMP concentrations developed which resulted in
increased fouling. The SMP values in this study are higher than
the range of 145–150 mg SMPCOD L−1 reported previously for the
treatment of low-strength synthetic domestic wastewaters in both
flocculated and granular AnMBRs [11,14]. However, since both
studies extracted sludge supernatant by filtration with 0.45 ␮m
membranes the SMP concentration may have been lower than for
this study due to the higher size exclusion offered by the filter.
Further speciation of the SMPCOD fraction by serial fractionation
indicated that while for the AnMBR greater than 80% of SMP
was between 1 ␮m and 500 kDa (Table 2), for the AeMBR only
approximately 43% was present in that fraction, further supporting
the reported particle size distribution data.
3.4. Critical fluxes in gas lift tubular membranes
The injection ratios (ε) obtained from measuring liquid dis-
placement by gas lift for the different gas sparging rates (Table 3)
passed through minima of 0.26 and 0.18 for the AeMBR and AnMBR
sludge respectively, corresponding to a superficial gas velocity (VG )
of 0.02 m s−1 after which they increased to maxima of 0.43 and
0.64 for VG = 0.21 m s−1 . Injection ratios in the range 0.2–0.9 corre-
Table 3
Impact of gas velocity (VG , m s−1 ) on liquid velocity (VL , m s−1 ) in the externally
configured (sidestream) tubular membrane during filtration of the AeMBR, AnMBR
and G-AnMBR sludge.

The concentration of soluble microbial products (SMP) VG (m s−1 ) AeMBR AnMBR G-AnMBR
determined in the supernatant as SMPCOD , SMP pro- VL (m s−1
) ε a
VL (m s−1
) εa
VL (m s−1 ) εa
tein (SMPP ) and SMP carbohydrate (SMPC ) in the AeMBR
0.02 0.07 0.22 0.09 0.18 0.26 0.07
averaged 99 ± 28 mg SMPCOD L−1 , 18 ± 13 mg SMPP L−1 and
0.05 0.13 0.28 0.12 0.29 0.36 0.12
18 ± 9 mg SMPC L−1 respectively (Table 2). The recorded SMPCOD 0.09 0.20 0.31 0.17 0.35 0.50 0.15
data are in agreement with previously reported results for 0.14 0.21 0.39 0.21 0.40 0.62 0.18
an AeMBR operated under similar operational conditions 0.18 0.25 0.42 0.24 0.43 0.71 0.20
0.21 0.28 0.43 0.12 0.64 0.35 0.38
(15 h HRT) and wastewater characteristics [25]. Data from
the suspended growth AnMBR were 598 ± 150 mg SMPCOD L−1 , a
ε = VG /(VG + VL ).
46 I. Martin-Garcia et al. / Journal of Membrane Science 382 (2011) 41–49

25 (a) 50
Fouling rate, dP/dt (mbar min-1)

AeMBR
VG (m s-1) 0.02 0.05 0.09 0.14
20 40 AnMBR

Critical flux (L m -2 h-1)

AnMBR
G-AnMBR G-AnMBR
15 AeMBR 30

10 20

5 10

0 0
0 5 10 15 20 25 30 35 0 0.5 1 1.5 2 2.5

Membrane flux, J (l m-2 h-1) Cross-flow velocity, CFV (m s -1)

Fig. 4. Impact of increased gas velocity (VG 0.02–0.14 m s−1 ) on fouling rate (b) 50
(dP/dt, mbar min−1 ) under tubular gas-lift mode at assigned membrane fluxes of AeMBR
4–28 L m−2 h−1 . Experiments conducted between 180 d and 240 d operation.
40 AnMBR

Critical flux (L m -2 h-1)

G-AnMBR
spond to slug-flow [26]. For the G-AnMBR sludge, the superficial 30
liquid velocity (VL ) induced by VG was approximately 300% of the
VL recorded for flocculated sludges; for a VG of 0.02 m s−1 , a VL 20
of 0.26 m s−1 was recorded for the G-AnMBR sludge compared to
0.07 m s−1 and 0.09 m s−1 for the AeMBR and AnMBR respectively.
10
Consequently for the G-AnMBR sludge a ε below 0.20 was cal-
culated for VG < 0.18 m s−1 , corresponding to bubble-flow which
0
provides less intense membrane scouring than slug flow [26].
0 4000 8000 12000 16000
Analysis of fouling rates (dP/dt) obtained during flux step
experiments and for a VG of 0.02–0.21 m s−1 indicated that when Reynolds number, Re
applying gas lift to tubular membranes, critical flux (Jc ) is more
Fig. 5. Impact of increased (a) crossflow velocity (CFV 0.4–2 m s−1 ) or (b) Reynold’s
influenced by sludge character than by gas sparging intensity; number (Re 500–15,000) on attainable critical flux (Jc ) during filtration in the
Jc in the AeMBR, AnMBR and G-AnMBRs remained at 8, 4 and externally configured tubular membrane operated in pumped mode. Experiments
4 L m−2 h−1 respectively independent of VG (Fig. 4). However, while conducted between 180 d and 240 d operation.
Jc remained relatively constant, a reduction in dP/dt was observed
from 11.8 mbar min−1 to 0.26 mbar min−1 for the AnMBR sludge as
reported for a tubular gas-lift membrane with a VFA and glucose
VG increased from 0.02 to 0.21 m s−1 at a flux (J) of 8 L m−2 h−1 . Since
feed respectively at VG values higher than those used in the cur-
shear-induced diffusivity is a less effective back-transport mecha-
rent study [15]. In this study, the colloidal nature of fouling in an
nism for low radius particles [27,28], it is postulated that the limited
anaerobic gas-lift MBR was corroborated by replacing the sludge
influence of VG on improving Jc arises from the contribution of col-
supernatant with tap water which resulted in a increase of Jc from
loid and soluble organics to fouling for gas-lift operation. Colloids
10 to 20 L m−2 h−1 . This result is consistent with the highest J being
are predominantly controlled by Brownian motion; consequently
obtained for the AeMBR, with a corresponding lowest concentra-
while a high shear environment is created by slug-flow, colloids
tion of soluble and colloidal organics, while the anaerobic systems
and soluble organics continue to preferentially migrate toward the
presented the highest SMP concentration and demonstrated the
membrane surface. Similarly, permeate flux has been observed to
lowest Jc . Earlier studies demonstrated significant flux promotion
be dependent only on applied TMP when increasing VG by an order
in AnMBRs through increasing VG [6,23]. An increase in J from 25 to
of magnitude from 0.001 m s−1 during ultrafiltration of 260 kDa
100 L m−2 h−1 following increasing VG from 0.2 m s−1 to 8.9 m s−1
molecular weight Dextran solutions [29]. Gas sparging has been
has been reported [23], albeit an appreciable decline in flux at
shown to have no effect on permeability of an air lift AeMBR due
higher gas sparging intensities attributed to the absence of a pro-
to the higher contribution of internal fouling (77%) compared to
tective cake layer which allows deposition of small particles on
cake layer formation (23%) [26]. The reduction in dP/dt recorded for
the membrane surface [31]. These studies employed ceramic mem-
the AnMBR upon increasing VG in the current study does however
branes, known to provide a higher organic fouling resistance than
indicate that, at high SMP concentrations, high shear rates partially
the polymeric materials used in the present study [5,32].
attenuate colloidal deposition.
Previous research has demonstrated that decreasing biomass
concentration in anaerobic MBRs can increase flux and reduce the 3.5. Critical fluxes in pumped tubular membranes
gas sparging demand [10,30]. However, in this study the lower
MLSS and SMP concentrations of the G-AnMBR sludge did not On using fluid pumping to generate the CFV in the tubular mem-
present higher Jc compared to the AnMBR system. The limited brane, Jc increased from 4 to 41 L m−2 h−1 in the G-AnMBR and from
impact of VG in the granular system can be attributed to the more 4 to 19 L m−2 h−1 in the suspended AnMBR for a CFV increase from
disperse gas phase at the lower ε values recorded, creating limited 0.4 to 2 m s−1 (Fig. 5). Traditionally, gas bubbles have been thought
turbulence at the membrane wall. Interestingly the dP/dt observed to enhance permeation by introducing localised liquid flow and
for J > Jc was significantly lower for the G-AnMBR compared to shear transients [1,33]. However, in this study, increasing CFV had a
the AnMBR: for a J of 11–12 L m−2 h−1 , the AnMBR dP/dt ranged greater impact on permeability than gas sparging in anaerobic sys-
between 8 and 25 mbar min−1 compared to 1–2 mbar min−1 for the tems. At CFV > 0.8 m s−1 , the Jc attained for the AeMBR was between
G-AnMBR. Similar AnMBR Jc values of 8 and 4 L m−2 h−1 have been that of the two anaerobic MBR, with a maximum Jc of 29 L m−2 h−1
 I. Martin-Garcia et al. / Journal of Membrane Science 382 (2011) 41–49 47

at the highest CFV of 2 m s−1 . At a CFV of 0.4 m s−1 in the AeMBR

system, a Jc of approx. 8 L m−2 h−1 was attained, similar to the Jc
recorded for the lowest VG of 0.02 m s−1 . This suggests that although
the influence of VG on dP/dt is not substantial, the shear intro-
duced at the membrane surface by gas bubble flow even at a low VG
(ε > 0.22) can incur a considerable shear effect in the AeMBR. Sim-
ilar trends have been reported elsewhere [34,35], although in this
case increasing VG was shown to increase Jc . It should be noted that
for both anaerobic sludges, Jc values equivalent to those obtained
in gas-lift more were measured at the lowest CFV of 0.4 m s−1 . It
has further been illustrated that at high CFV the back-transport
velocity of sub-micron particles is adequately described by both
shear-induced diffusion and Brownian diffusion [36]. Where the
main reactors to have been operated as tubular MBR rather than
HF-MBR, the colloidal fraction in the flocculated systems may have
been more dominant due to the augmented shear [37]. However,
this study does illustrate that increased CFV can provide enhanced
suppression of concentration polarisation [23] in colloid dominated
systems.
The impact of CFV on Jc was more prominent for the G-AnMBR
(Fig. 5). The higher concentration of colloidal foulants in the G-
AnMBR with respect to the AeMBR may be compensated by the
higher shear resulting from its lower viscosity. Comparison of filtra-
tion performance between G-AnMBR and AnMBR systems operated
at low MLSS concentrations indicated that for a CFV of 1 m s−1 , sim-
ilar J values of 50 L m−2 h−1 were obtained [4,38]. In this study,
based on a viscosity of 1 mPa s−1 for the G-AnMBR supernatant,
a Re range from 3200 to 16,000 corresponds to CFV from 0.4 to
2 m s−1 . In the AeMBR and AnMBR systems, based on approxi-
mated rheological parameters [39,40], the Re ranges from 900 to
9000. Consequently, when comparing system efficacy based on
normalised turbulence (Re, Fig. 5), higher Jc values are obtained in
the AeMBR than the G-AnMBR: increasing Re from 3200 to 6170 and
from 3350 to 6400 in the G-AnMBR and AeMBR, respectively pro-
vided increases in Jc from 4 to 12 L m−2 h−1 and 16 to 29 L m−2 h−1 Fig. 6. Impact of specific gas demand per unit membrane area (SGDm
in the AeMBR. 0.19–1.16 m3 m−2 h−1 ) on attainable critical flux (Jc ) for AeMBR (a), AnMBR (b) and
G-AnMBR sludge (c) with an immersed hollow-fibre membrane. Experiments con-
ducted between 120 d and 150 d operation.

3.6. Critical fluxes in immersed HF membranes

An increase in gas intensity from 0.007 to 0.027 m s−1 (N2 gas
flow rate [m3 s−1 ] normalised to diffuser surface area in [m2 ])
resulted in similar increases in Jc from 3 L m−2 h−1 to 14 L m−2 h−1
in the AeMBR and G-AnMBR respectively. Previous studies have
reported a threshold gas sparging rate after which no further
improvement in J is obtained [28,41]. In this study, the thresh-
old was demonstrated for the G-AnMBR at a gas intensity of
0.041 m s−1 , corresponding to a flux of 14 L m−2 h−1 . However, for
the gas intensities studied, a threshold was not observed for the
AeMBR, with the highest Jc of 17 L m−2 h−1 recorded at the max-
imum aeration studied (0.057 m s−1 ). An increase in gas intensity
from 0.007 to 0.041 m s−1 in the AnMBR yielded a small increase in
Jc from 2 L m−2 h−1 to 5 L m−2 h−1 . These results indicate that the G-
AnMBR system requires approximately 50% lower gas demand than
the AnMBR to achieve an equivalent Jc . Furthermore, at higher gas
intensities, a 200% increase in Jc can be achieved using G-AnMBR
rather than AnMBR. Similar findings have been reported from com-
paring AnMBR and G-AnMBR for blackwater treatment [10], where
it was asserted that a higher permeability could be achieved due to
the lower colloidal concentration present in the granular reactor.
Similar assumptions regarding the impact of the low viscosity on
increased shear intensity arise for gas sparged systems; the approx-
imate velocity gradient associated with membrane scouring (G)
ranges from 75 to 200 s−1 and from 87 to 208 s−1 in the aerobic and
anaerobic MBRs respectively, while for the same range of scouring
rate in the G-AnMBR, G varied between 230 s−1 and 670 s−1 . How-
ever, comparative specific energy demands for the immersed and
pumped crossflow configurations of 0.3 kWh m−3 and 3.7 kWh m−3
of permeate produced respectively indicate that immersed mem-
branes present the most economically appropriate means for shear
induction.
The gas flows applied in tubular and HF geometries yielded
similar specific gas demands (SGDm , the gas flow rate [m3 h−1 ]
normalised against membrane surface area [m2 ]) ranging from 0.2
to 1.6 m3 m−2 h−1 . Comparison of results between the two gas-
membrane systems indicate that at low SGDm values of 0.2 and
0.4 m3 m−2 h−1 , the tubular module produced the higher Jc of 4 and
8 L m−2 h−1 in the AeMBR and AnMBR respectively (Fig. 6). At low
applied gas velocities, fouling by aggregates dominates over solu-
ble/colloidal fouling. Consequently, gas sparging is more effective
in tubular membranes at the lower velocities due to the slug flow
regime induced, whereas in HF systems the energy introduced with
bubbling is dissipated across the receiving water column. However,
above this SGDm threshold, the Jc attainable in flocculated systems
are higher with the immersed HF module. For the G-AnMBR, the HF
membrane yielded a higher Jc over the entire SGDm range tested.
The limited impact of gas-lift shear in the low solids environment
provided by the G-AnMBR sludge supports the concept of solids-
dominated convective transport at low velocities. Previous reports
have indicated the diminishing effect observed following increased
gas sparging in tubular membranes to be due to the increased fre-
quency of gas slugs which has a slightly deleterious impact on the
dynamic turbulence [29]. In contrast, the requirement for higher
48 I. Martin-Garcia et al. / Journal of Membrane Science 382 (2011) 41–49
SGDm in HF systems is intuitive since the hydrodynamic environ-
ment is poorly defined. However, the greater effect of gas sparging
on Jc obtained with HF membranes suggests that mechanisms such
as fibre movement, which physically removes the mass transfer
boundary layer and induces liquid flows transverse to the fibres
[33], may be more significant in removing colloidal fouling than
the shear induced in the liquid membrane interface by gas bubbling
inside the tubular membrane.
4. Conclusions
The filtration performance of a flocculated AnMBR and granular
AnMBR treating domestic wastewater has been evaluated and com-
pared to a conventional aerobic MBR. The flocculated AnMBR sludge
was characterised by a similar MLSS concentration to an aerobic
MBR, but with approximately 500% greater supernatant colloidal
fraction. In comparison, the granulated AnMBR sludge was char-
acterised by a low MLSS concentration and, while the supernatant
colloidal fraction was higher than the aerobic MBR, the SMP con-
centration was below 50% of the flocculated AnMBR. The externally
configured tubular membranes operated in gas-lift mode provided
equivalent critical fluxes for both the granular and flocculated
AnMBR, though the fouling rate (dP/dt) was markedly lower for
the granular sludge. In contrast, when operating the tubular geom-
etry in pumped flow mode, considerably higher critical fluxes were
attainable with the granular AnMBR. At a CFV > 0.8 m s−1 , critical
fluxes achieved with the granular AnMBR exceeded those for both
aerobic and anaerobic flocculated MBR, suggesting that the shear
conditions created were effective in limiting concentration polari-
sation within colloid dominated matrices. Higher critical fluxes of
approximately 200% were also achieved for the granular anaerobic
MBR with an immersed hollow fibre membrane, compared to the
anaerobic flocculated matrix. The higher fluxes measured for the
granular anaerobic MBR during gas sparging of immersed mem-
branes or pumped crossflow of external membranes reflects the
impact of lower viscosity on shear. These results indicate that gran-
ular rather than flocculated AnMBR systems appear most suitable
for domestic wastewater treatment. However, comparative specific
energy demands for the immersed and pumped crossflow config-
urations of 0.3 kWh m−3 and 3.7 kWh m−3 of permeate produced
respectively indicate that immersed membranes present the most
economically attractive option for granular anaerobic systems.
Acknowledgements
This research project has been supported by a Marie Curie Early
Stage Research Training Fellowship of the European Community’s
Sixth Framework Programme under contract number MEST-CT-
2005-021050. The authors would also like to thank our industrial
sponsors Severn Trent Water and Yorkshire Water for their finan-
cial and technical support. Dr. I.G. Martin is now at the University
of Valencia, Spain.
References
[1] S. Judd, The MBR Book: Principles and Applications in Water and Wastewater
Treatment, Elsevier Science, Amsterdam, 2006.
[2] A. Beaubien, M. Bâty, F. Jeannot, E. Francoeur, J. Manem, Design and operation of
anaerobic membrane bioreactors: development of a filtration testing strategy,
J. Membr. Sci. 109 (1996) 173–184.
[3] K.H. Choo, C.H. Lee, Membrane fouling mechanisms in the membrane-coupled
anaerobic bioreactor, Water Res. 30 (1996) 1771–1780.
[4] S. Elmaleh, L. Abdelmoumni, Cross-flow filtration of an anaerobic
methanogenic suspension, J. Membr. Sci. 131 (1997) 261–274.
[5] W.R. Ghyoot, W.H. Verstraete, Coupling membrane filtration to anaerobic pri-
mary sludge digestion, Environ. Technol. 18 (1997) 569–580.
[6] E. Kayawake, Y. Narukami, M. Yamagata, Anaerobic digestion by a ceramic
membrane enclosed reactor, J. Ferment. Bioeng. 71 (1991) 122–125.
[7] T. Imasaka, H. So, K. Matsushita, T. Furukawa, N. Kanekuni, Application of
gas–liquid two-phase cross-flow filtration to pilot-scale methane fermenta-
tion, Drying Technol. 11 (1993) 769–785.
[8] D. Jeison, J.B. van Lier, Cake formation and consolidation: main factors
governing the applicable flux in anaerobic submerged membrane bioreac-
tors (AnSMBR) treating acidified wastewaters, Sep. Purif. Technol. 56 (2007)
71–78.
[9] Z. Huang, S.L. Ong, H.Y. Ng, Feasibility of submerged anaerobic membrane biore-
actor (SAMBR) for treatment of low-strength wastewater, Water Sci. Technol.
58 (2008) 1925–1931.
[10] E. van Voorthuizen, A. Zwijnenburg, W. van der Meer, H. Temmink, Biological
black water treatment combined with membrane separation, Water Res. 42
(2008) 4334–4340.
[11] A.Y. Hu, D.C. Stuckey, Treatment of dilute wastewaters using a novel submerged
anaerobic membrane bioreactor, J. Environ. Eng. 132 (2006) 190–198.
[12] H.J. Lin, K. Xie, B. Mahendran, D.M. Bagley, K.T. Leung, S.N. Liss, B.Q. Liao, Sludge
properties and their effects on membrane fouling in submerged anaerobic
membrane bioreactors (SAnMBRs), Water Res. 43 (2009) 3827–3837.
[13] C. Wen, X. Huang, Y. Qian, Domestic wastewater treatment using an anaero-
bic bioreactor coupled with membrane filtration, Process Biochem. 35 (1999)
335–340.
[14] L.B. Chu, F.L. Yang, X.W. Zhang, Anaerobic treatment of domestic wastewater
in a membrane-coupled expended granular sludge bed (EGSB) reactor under
moderate to low temperature, Process Biochem. 40 (2005) 1063–1070.
[15] D. Jeison, J.B. van Lier, Thermophilic treatment of acidified and partially
acidified wastewater using an anaerobic submerged MBR: factors affecting
long-term operational flux, Water Res. 41 (2007) 3868–3879.
[16] I.G. Martin, M. Pidou, A. Soares, S. Judd, B. Jefferson, Modelling the energy
demands of aerobic and anaerobic membrane bioreactors for wastewater treat-
ment, Environ. Technol. 32 (9) (2011) 921–932.
[17] P. Le-Clech, B. Jefferson, I.S. Chang, S.J. Judd, Critical flux determination by the
flux-step method in a submerged membrane bioreactor, J. Membr. Sci. 227
(2003) 81–93.
[18] American Public Health Association (APHA), Standard Methods for the Exami-
nation of Water and Wastewater, 20th ed., American Public Health Association,
Washington, DC, USA, 1998.
[19] O.H. Lowry, N.J. Rosebrough, A.L. Farr, R.J. Randall, Protein measurement with
the Folin phenol reagent, J. Biol. Chem. 193 (1951) 265–275.
[20] M. Dubois, K.A. Gilles, J.K. Hamilton, P.A. Rebers, F. Smith, Colorimetric method
for determination of sugars and related substances, Anal. Chem. 28 (1956)
350–356.
[21] Z. Huang, S.L. Ong, H.Y. Ng, Submerged anaerobic membrane bioreactor for
low-strength wastewater treatment: effect of HRT and SRT on treatment per-
formance and membrane fouling, Water Res. 45 (2011) 705–713.
[22] B.-M. Wilén, B. Jin, P. Lant, The influence of key chemical constituents in acti-
vated sludge on surface and flocculating properties, Water Res. 37 (2003)
2127–2139.
[23] T. Imasaka, N. Kanekuni, H. So, S. Yoshino, Cross-flow filtration of methane
fermentation broth by ceramic membranes, J. Ferment. Bioeng. 68 (1989)
200–206.
[24] E.J. McAdam, S.J. Judd, E. Cartmell, B. Jefferson, Influence of substrate on fouling
in anoxic immersed membrane bioreactors, Water Res. 41 (2007) 3859–3867.
[25] M. Speı̌randio, A. Masse, M.C. Espinosa-Bouchot, C. Cabassud, Characterization
of sludge structure and activity in submerged membrane bioreactor, Water Sci.
Technol. 52 (2005) 401–408.
[26] I.S. Chang, S.J. Judd, Air sparging of a submerged MBR for municipal wastewater
treatment, Process Biochem. 37 (2002) 915–920.
[27] E.J. McAdam, S.J. Judd, Optimisation of dead-end filtration conditions for an
immersed anoxic membrane bioreactor, J. Membr. Sci. 325 (2008) 940–946.
[28] E.J. McAdam, E. Cartmell, S.J. Judd, Comparison of dead-end and continuous
filtration conditions in a denitrification membrane bioreactor, J. Membr. Sci.
369 (2011) 167–173.
[29] Z.F. Cui, S.R. Bellara, P. Homewood, Airlift crossflow membrane filtration — a
feasibility study with dextran ultrafiltration, J. Membr. Sci. 128 (1997) 83–91.
[30] D. Jeison, J.B. van Lier, Cake layer formation in anaerobic submerged membrane
bioreactors (AnSMBR) for wastewater treatment, J. Membr. Sci. 284 (2006)
227–236.
[31] J. Lee, W.Y. Ahn, C.H. Lee, Comparison of the filtration characteristics between
attached and suspended growth microorganisms in submerged membrane
bioreactor, Water Res. 35 (2001) 2435–2445.
[32] I.J. Kang, S.H. Yoon, C.H. Lee, Comparison of the filtration characteristics of
organic and inorganic membranes in a membrane-coupled anaerobic biore-
actor, Water Res. 36 (2002) 1803–1813.
[33] Z.F. Cui, S. Chang, A.G. Fane, The use of gas bubbling to enhance membrane
processes, J. Membr. Sci. 221 (2003) 1–35.
[34] P. Le-Clech, B. Jefferson, S.J. Judd, A comparison of submerged and
sidestream tubular membrane bioreactor configurations, Desalination 173
(2005) 113–122.
[35] M. Pidou, S.A. Parsons, G. Raymond, P. Jeffrey, T. Stephenson, B. Jefferson, Foul-
ing control of a membrane coupled photocatalytic process treating greywater,
Water Res. 43 (2009) 3932–3939.
[36] E. Tardieu, A. Grasmick, V. Geaugey, J. Manem, Hydrodynamic control of biopar-
ticle deposition in a MBR applied to wastewater treatment, J. Membr. Sci. 147
(1998) 1–12.
[37] J. Ho, S. Sung, Effects of solid concentrations and cross-flow hydrodynamics on
microfiltration of anaerobic sludge, J. Membr. Sci. 345 (2009) 142–147.
 I. Martin-Garcia et al. / Journal of Membrane Science 382 (2011) 41–49 49

[38] B.D. Cho, A.G. Fane, Fouling transients in nominally sub-critical flux
operation of a membrane bioreactor, J. Membr. Sci. 209 (2002) 391–
403.
[39] G. Laera, C. Giordano, A. Pollice, D. Saturno, G. Mininni, Membrane bioreac-
tor sludge rheology at different solid retention times, Water Res. 41 (2007)
4197–4203.
[40] A. Pevere, G. Guibaud, E. van Hullebusch, P. Lens, Identification of rheological
parameters describing the physico-chemical properties of anaerobic sulphido-
genic sludge suspensions, Enzyme Microb. Technol. 40 (2007) 547–554.
[41] E.H. Bouhabila, R.B. Aïm, H. Buisson, Microfiltration of activated sludge using
submerged membrane with air bubbling (application to wastewater treat-
ment), Desalination 118 (1998) 315–322.