Bioresource Technology 138 (2013) 329–338

Contents lists available at SciVerse ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Harvesting microalgal biomass using a magnetically induced membranevibration (MMV) system: Filtration performance and energyconsumption

0960-8524/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.biortech.2013.03.175

⇑ Corresponding author. Tel.: +32 16 321594; fax: +32 16 321998.

M.R. Bilad a, V. Discart a, D. Vandamme b, I. Foubert b, K. Muylaert c, Ivo F.J. Vankelecom a,⇑a Centre for Surface Chemistry and Catalysis, Faculty of Bioscience Engineering, KU Leuven, Kasteelpark Arenberg 23, Box 2461, 3001 Leuven, Belgiumb Lab Aquatic Biology, Microbial en Molecular Systems, KU Leuven Kulak, E. Sabbelaan 53, B-8500 Kortrijk, Belgiumc Lab Aquatic Biology, Faculty of Biology, KU Leuven Kulak, E. Sabbelaan 53, B-8500 Kortrijk, Belgium

h i g h l i g h t s

� A vibrated membrane (MMV) wasapplied to up-concentrate microalgalbroths.� Filterability of both broths was

assessed and compared.� Membranes with higher porosities

offer better performances.� The MMV offers good fouling control

and proved to be economicallyattractive.� The MMV showed a low energy

consumption in microalgaeharvesting.

g r a p h i c a l a b s t r a c t

Tel: 016/32.15.94 Fax: 016/32.19.98

Permeate recirculation

Pressuregauges

Sample collection

Multi-channel Peristaltic pump

Membranes

Air source for m

ixing

Filtration tankC

ontaining microalgal broth

P

PMagneticVibrating engine

Water circulation

Vibrating driver

EEnergymeter

a r t i c l e i n f o

Article history:Received 7 January 2013Received in revised form 25 March 2013Accepted 26 March 2013Available online 4 April 2013

Keywords:Algae dewateringMembrane foulingMicroalgae harvestingSubmerged filtrationP. tricornutum

a b s t r a c t

This study was performed to investigate the effectiveness of submerged microfiltration to harvest both amarine diatom Phaeodactylum tricornutum and a Chlorella vulgaris in a recently developed magneticallyinduced membrane vibrating (MMV) system. We assess the filtration performance by conducting theimproved flux step method (IFM), fed-batch concentration filtrations and membrane fouling autopsyusing two lab-made membranes with different porosity. The full-scale energy consumption was also esti-mated. Overall results suggest that the MMV offers a good fouling control and the process was proven tobe economically attractive. By combining the membrane filtration (15� concentration) with centrifuga-tion to reach a final concentration of 25% w/v, the energy consumption to harvest P. tricornutum and C.vulgaris was, respectively, as low as 0.84 and 0.77 kW h/m3, corresponding to 1.46 and 1.39 kW h/kg ofthe harvested biomass.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction ments, natural pigments and polyunsaturated fatty acids (Raja

Harvesting microalgal biomass from cultivation broths is one ofthe major challenges that limits the widespread and full-scaleapplication of microalgae as raw material for many different end-products, especially biofuel (Greenwell et al., 2010). Microalgae-based-products are currently only economically feasible on a lim-ited scale, primarily for high-value products, such as food supple-

et al., 2008). When aiming at the biomass, harvesting is normallythe bottleneck of the downstream processing. However, when justsome of the desired components (mostly high value products) needto be extracted from the biomass, the extraction procedure andefficiency is of a more vital interest (Šoštaric M, 2012). The mostcommon harvesting technique is centrifugation, but few othersare also suggested because of economical reason, including floccu-lation, electro-coagulation-flocculation and membrane filtration.

Micro- and ultrafiltration have been proven to be effective toharvest microalgae, both in cross-flow and submerged configura-

Table 1The growth media for the microalgae cultivation. The solutions were dissolved indemineralized water for the C. vulgaris. Marine salts (NaCl) were added for P.tricornutum.

Component Concentrationin the final medium (M)

Tris buffer 3.78 � 10�3

NaNO3 1 � 10�3

CaCl2�2H2O 2.5 � 10�3

MgSO4�7H2O 1.5 � 10�4

NaHCO3 1.5 � 10�4

Na2SiO3�9H2O 1.00 � 10�4

K2HPO4 5.00 � 10�5

Trace metals solutionNa2EDTA�2H2O 3.78 � 10�3

FeCl3�6H2O 1 � 10�3

CuSO4�5H2O 2.5 � 10�3

ZnSO4�7H2O 1.5 � 10�4

CoCl2�6H2O 1.5 � 10�4

MnCl2�4H2O 1.00 � 10�4

Na2MoO4�2H2O 5.00 � 10�5

H3BO3 1.65 � 10�5

Vitamins solutionThiamine HCl (vitamin B1) 2.96 � 10�7

Biotin (vitamin H) 2.05 � 10�9

Cyanocobalamin (vitamin B12) 3.69 � 10�10

330 M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338

tion (Rossignol et al., 1991; Zhang et al., 2010; Bilad et al., 2012).Membrane filtration has many advantages in this context: an al-most complete retention of biomass, potential disinfection viaremoval of microorganisms bigger then the nominal pore sizes,limited addition of chemicals into the system, thus preventingthem from accumulating in the system, and a low energy con-sumption (Mouchet and Bonnelye, 1998; Judd and The, 2006).The cross-flow filtration offers a high productivity (high appliedfluxes) due to the high cross-flow velocity and high shear ratesexposed onto the membrane surfaces. However, this consumesconsiderable amounts of energy due to the sometimes high ap-plied pressures and liquid velocities, especially for ultrafiltrationmembranes (Le-Clech et al., 2006). In addition, over-exposure ofmicroalgal biomass to shear, especially in the intake and pump-ing systems, may break microalgal cells to form smaller parti-cles, colloids and algal organic matters (AOMs) or promoterelease of exopolymeric substances (EPS) (Babel and Takizawa,2010; Ladner et al., 2010). Furthermore, cell breakage may alsolead to the loss of targeted products from the cell interior. Onthe other hand, the submerged aerated system applies lowerpressures, brings limited exposure of microalgal cell to enhancedshear rates (hence reducing EPS release), and thus offers a lowerenergy consumption and reduced investment cost than thecross-flow system (Judd and The, 2006; Le-Clech et al., 2006; Ba-bel and Takizawa, 2010). However, these membrane processesalso have some disadvantages, such as the relatively low fluxesthat increase the capital costs, when large volumes have to betreated. Membrane scouring via air bubbling is less effectivethan cross flow operation, thus increasing the maintenancecleaning frequency and down-times (Mouchet and Bonnelye,1998).

The required shear at the membrane surface in a submergedsystem is generally provided by coarse air bubbles which can, inthe case of algae cultivation, also act to supply the required carbondioxide (CO2) from the air. However, this approach produces a rel-atively weak shear rate, so that only moderate membrane fluxescan be applied. It is also often difficult to ensure a homogeneousbubble distribution and a ‘‘plateau’’ in terms of flux improvementis reached beyond which no further improvement can be achievedby increasing the bubbles flow (Cui et al., 2003; Genkin et al.,2006).

The enhancement of membrane shear-rates has long been rec-ognized as one of the most efficient factors for fouling control. Itis implemented either by moving the fluid or the membrane. Themembrane can be moved in a circular rotation, a torsional vibra-tion or in vertical and horizontal oscillation systems (Jaffrin,2008; Bilad et al., 2012). Application of a rotating disk system foralgal harvesting, recently reported by Frappart et al. (2011) andRios et al. (2011), showed that it almost doubled the membraneproductivity compared to a reference cross-flow system, ascribedto the high shear-rates at the liquid-membrane interface. However,Ladner et al. (2010) found a very significant impact of enhancedshear on the microalgal cells. The AOM released from sheared mic-roalgal cells caused increased membrane pore blocking. This phe-nomenon was not observed in the other studies (Frappart et al.,2011; Rios et al., 2011), probably due to different types of microal-gae (cell wall), type of pumps, filtration experimental designs(shorter time-frame), etc. Therefore, a process that would maintaina high shear-rate only at the liquid-membrane interface, and not inthe whole bulk, would be beneficial to achieve an efficient filtra-tion process.

In this study, we applied a magnetically induced membranevibrating (MVM) system to harvest both a marine diatom P. tricor-nutum and a C. vulgaris. Both are promising species for the produc-tion of microalgal biomass for food, feed, or fuel, and are currentlyintensively studied (Greenwell et al., 2010; Raja et al., 2008). Their

sizes are also large enough to be retained by the appliedmembranes.

The effect of membrane porosity is also investigated here bytesting two lab-made polyvinylidene fluoride (PVDF) membranes.The filtration performances are evaluated using the improvedflux-step method (IFM) (van der Marel et al., 2009) and fed-batchconcentration filtrations. The membrane fouling was evaluatedby observing scanning electron microscopy (SEM) images. The en-ergy consumption for a full-scale MVM application in algae har-vesting was also tentatively estimated by extrapolating the dataobtained from a related full-scale submerged MBR for wastewatertreatment.

2. Methods

2.1. Cultivation and characterization of microalgae

The tests were conducted using two different microalgal brothsnamely a Phaeodactylum tricornutum and Chlorella vulgaris. A smallcontamination during the experiment of the latter broth was ob-served under light microscopy by Scenedesmus sp. However, thedegree of contamination was assumed to be too small to affectthe results. Both broths were cultivated using a 25 L lab-scale pho-tobioreactor operated in batch-wise. The growth media are givenin Table 1. The photobioreactor design and operation are givenby Vandamme et al. (2011). The filtration experiments were per-formed for one week within the stationary growth phase, whichwas achieved after 7 days of cultivation. The fresh biomass concen-trations were 0.25 and 0.21 g/l for P. tricornutum and C. vulgaris,respectively. All feeds for each broth were produced in one batchcultivation to avoid discrepancies that often occur on the feedproperties among different batches.

2.2. Membrane preparation, properties and module potting

Two different flat-sheet membranes with different porositywere prepared from 9% and 12% w/w PVDF, (Mw�534,000)/N,N-dimethylformamide (DMF) solutions by phase inversion using 3%w/w of polyvinylpyrrolidone (PVP, Mw�10,000) as an additive.The solutions were cast on a polypropylene non-woven support(Novatexx 2471, kindly supplied by Freudenberg, Germany) witha 250 lm wet thickness and casting speed of 2.25 cm/s, and then

M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338 331

brought into contact with demineralized water as the non-solvent.In the non-solvent bath, the polymer phase separates and solidifiesto form a thin porous membrane. Both membranes were cast at21 �C and 40% relative humidity. The membranes were then storedin tap water before being potted. The polymer concentration in thecasting solution of each membrane is further used to index themembranes (PVDF-9 and PVDF-12). Two identical sets of mem-branes were prepared and each set was used for each microalgalculture. Fresh membranes were used only for the first filtrationand the membranes were chemically cleaned in between subse-quent filtration tests. Prior to use, all membranes were potted toform modules with an effective membrane area of 0.016 m2

according to Bilad et al. (2011).The microstructure of the freshmembranes was observed with SEM (Philips SEM XL30 FEG withEDX dx-4i system). The membrane pore size and surface porositywere estimated from SEM images using imageJ (NIH, USA).

The flux (J), the permeance (L), relative permeance (LR) and spe-cific filtration volume (VC) of the membrane and filtration were cal-culated by using Eqs. (1)–(4) respectively.

J ¼ VAtðl=m2 hÞ ð1Þ

J ¼ JTMP

ðl=m2 hbarÞ ð2Þ

LR ¼LL0ðdimensionlessÞ ð3Þ

VC ¼VAðm3=m2Þ ð4Þ

where V is volume (l or m3), t time (h), A effective filtration area(m2) and TMP trans-membrane pressure (bar or kPa), L0 permeanceof fresh membrane (l/m2 h bar).The harvesting efficiency of micro-algal biomass (g) was determined based upon the decrease in opti-cal density of the microalgal suspension between feed and

Permeate recirculation

Pressuregauges

Sample collection

Multi-channel Peristaltic pump

Membranes

Air source for m

ixing

Filtration tank

(A) (B)

P

PMagneticVibrating engine

Water circulation

Vibrating driver

EEnergymeter

Vibrating driver

V

Fig. 1. Experimental set-up for (A) the improved flux stepping filtration method (IFM) tenarrow edges of the two vibrating membranes, i.e., the alignment of the two membranesurface view of the vibrating membranes.

permeate, as measured at 550 nm with a spectrophotometer. Theg (%) was subsequently calculated using Eq. (5).

g ¼ ODf � ODp

ODf� 100 ð%Þ ð5Þ

where ODf is the optical density of the feed and ODp the optical den-sity of the permeate.

2.3. Experimental set-up and magnetic vibrating module

The filtration performance tests were performed in a lab-scalefiltration set-up, illustrated in Fig. 1. The MMV engine was placedat the top of the tank, and it was connected to the membrane mod-ules inside the tank via the module frame, to which two membranemodules were fixed as shown in Fig. 1A.

In MMV, the membrane is fixed onto a vibration frame that isdirectly connected to an engine. A magnetic force generates vibra-tion through the engine via a pull-and-push mode that is derivedfrom a programmable sonic device, thus offering a high degree offlexibility (Bilad et al., 2012). In the MMV system, only the li-quid-membrane interface experiences shear stress and not thebulk liquid, thus minimizing energy consumption and the secre-tion of AOMs. This process has proven to be efficient in submergedmembrane bioreactor (MBRs) (Bilad et al., 2012) and in the directsewage filtration (Mezohegyi et al., 2012). The MVM generated amagnetically induced vibration in order to provide an enhancedshear-rates at the liquid-membrane interface. The MMV system in-cludes a vibration driver, electric wire, vibration engine and theactually vibrating part (including the membrane). The vibrationdriver (Test Tone Generator software) provides the vibrating signal.The vibration itself was created in the vibrations by magneticattraction/repulsion forces in ‘‘push and pull’’ mode. The move-ment orientation of the vibrating part faced the narrow face ofthe module in order to both prevent the bumping of liquid ontothe membrane and minimize the associated energy dissipation.

Permeatecollection

Algae storage

Water circulation

ibrating directionsAi

r sou

rce

for m

ixin

g

st in a total permeate recycle filtration mode, also showing the parallel view of themodules, and (B) the fed-batch concentration filtration showing the set-up in a full

Feed 1 (25 L) Retentate 2(1.67 L)

Permeate 2

Membrane

Membrane

Perm

eate

1

Retentate 1(5 L)

Filtration 1

Membrane

Filtration 2

Additional filtration

Fig. 2. The general scheme of the filtration tests. An additional filtration was required to speed-up the process. Filtrations 1 and 2 were performed at a fixed flux (J) of 32 l/m2 h. Retentate from the additional filtration was later added to the retentate of filtration 1. Retentate 1 was later used as feed for filtration 2. The improved flux stepping(IFM) test was performed using feed 1, retentate 1 and 2.

332 M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338

The vibration moved the membrane linearly to the left and right(see Fig. 1B) through a sinusoidal pattern signal derived from thevibration driver. The real vibration power was later calculated fromthe electric current and electrostatic potential measured on theelectrical wire by two Volt-Ohm meter (DVM 890-Velleman). Thefiltration was performed at a fixed vibration frequency of 45 Hzand a vibrating power of 6.4 Watt and was operated in a continu-ous vibration method without relaxation.

The filtration tank has a working volume of 1.6 L and isequipped with an air bubble aeration to both mix the culture andsupply the CO2 (contained in the air) for the microalgae duringthe experimental duration. The aeration is located at the bottomcorner of the tank, as shown in Fig. 1. The aeration was intention-ally not aimed towards the membrane to exclude the scouring ef-fect of these air bubbles, like in an aerated system (Bilad et al.,2012). The mixing would have been better with an impeller, butthis was technically not feasible in this study, since it would re-quire a larger filtration tank and thus more broth medium. A par-allel filtration was applied during the test to achieve a correctcomparison of membrane performances by applying exactly iden-tical operating and microalgal broth conditions for all tested mem-branes. Such parallel operation is essential to avoid discrepanciesin the feed composition, due to the dynamic behavior of the micro-algae over the testing period. Inside the filtration tank, each perme-ate line was connected to a separated channel using isoprenemanifold tubes through an individual line and an individual vac-uum gauge in a multi-channel peristaltic pump. The filtrationfluxes were adjusted in this case by changing the pump rotationalspeed.

2.4. Membrane filtration

The filterability of the continuous and fed-batch processes wasassessed via IFM and fed-batch filtration (Fig. 1). Two types of fil-tration tests were conducted during the experiments (Fig. 2):fed-batch concentration filtration and the IFM. A final filtration ofretentate 2 at a fixed flux of 32 l/m2 h for 3 h was performed atthe end of the test. The aim of this extra filtration was to foul themembranes and to study the effectiveness of membrane chemicalcleaning. The fouled and the cleaned membrane samples were alsoobserved using SEM.

2.4.1. Fed-batch filtrationIn the fed-batch filtration, 1.6 L of feed was put in the filtration

tank and the remaining was left in the storage tank. The microalgalbroth was concentrated by filtering the broth in the filtration tank,

which was continuously fed with fresh broth from a storage tank atthe same flow rate as the total membrane flux. This way, the vol-ume of broth in the filtration tank was kept constant to ensure con-stant tank hydrodynamics. The filtration was stopped when allmicroalgal broth from the storage tank was finished.

Two stages of fed-batch filtrations (filtrations 1 and 2) wereperformed for both broth samples at a fixed flux of 32 l/m2 h. Thefresh broth, the retentate of filtration 1 and the retentate of filtra-tion 2 are further referred to as feed 1, retentate 1 and retentate 2,respectively. For filtration 1, 8 L of feed 1 was concentrated to pro-duce 1.6 L of retentate 1 for 6 h, hence giving a 5� volumetric con-centration factor. For filtration 2, 5 L of retentate 1 wasconcentrated to produce 1.6 L of retentate 2 for 3.5 h, hence givinga 3� volumetric concentration factor, and 15� overall concentra-tion factor. To make-up the remaining 3.4 L of retentate 1, an addi-tional filtration was included (Fig. 2). This was done sinceperforming filtration 1 at very high fluxes was not possible dueto an instant deterioration of filterability as a result of acute foul-ing. In contrary, filtration at low fluxes would have resulted in anextended filtration time, which would in turns have prolongedthe experimental duration beyond the limit (one week), afterwhich the ageing microalgal broths would dramatically have chan-ged in properties. Such ageing won’t be an issue in a full-scale pro-cess, since the broth should be produced continuously.

2.4.2. Improved flux-stepping test filtrationThe feed filterability was assessed using the IFM test. The

description and the importance of this test was discussed earlierelsewhere (van der Marel et al., 2009). This method applies suc-cessive membrane fluxes (JH) with increasing level, and includesan intermediate flux decrease to a reference low flux (JL = 7 l/m2 h) after each JH step. JH was started from 10 l/m2 h and step-wise increased by 5 l/m2 h at step durations of 10 min until themaximum pump capacity (50 l/m2 h) was reached. Two types ofJCs could be obtained from each test, namely the common criticalflux (JC) and the critical flux for irreversibility (JCir). The values ofJC and JCir were defined as the flux step where a first minimumTMP increase (DTMPmin) of 10 Pa/min was reached for JH and JL,respectively. The method for DTMPmin calculation is describedelsewhere (Bilad et al., 2012). The performance of the membraneswas evaluated based on both their JC and JCir values. Higher JC’s

and JCir’s, respectively indicate a lower total fouling and lowerirreversible fouling propensity of the corresponding membrane/feed combination. The IFM tests were performed for feed 1, reten-tate 1 and retentate 2, sandwiched by the fed-batch concentrationfiltrations. It was practically not possible to apply multiple IFM

M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338 333

tests or to replicate the measurements due to the limited periodof sample usage (one week). Therefore, a statistical analysis froma repeated test could not be provided in this study. The interpre-tation of the data will thus only be done in rather general termsfor the clear trends only.

2.5. Membrane cleaning and foulant characterization

In between the series of filtration tests, the membranes werecleaned by flushing them with tap water. They were then soakedconsecutively into 1000 ppm sodium hypochlorite (NaClO) follow-ing by 1000 ppm of citric acid (C6H8O7) solution, each for 1.5 h. Thecitric acid cleaning was incorporated, since a severe inorganic foul-ing was found in our earlier study when only applying NaOCl (Biladet al., 2012). A special cleaning procedure was applied after the fi-nal filtration test. All the fouled membranes were cut into threeseparated pieces. The first piece was used as the fouled membranesample. The second piece was chemically cleaned using 1000 ppm(NaOCl) for 3 h. And the third piece was chemically cleaned using1000 ppm NaOCl for 1.5 h, following by 1000 ppm of citric acid for1.5 h. Afterward, all samples were dried at room temperature,coated with a sputtered gold layer and their microstructures wereobserved using SEM.

2.6. Energy consumption

The estimation of the energy consumption for the filtration wasadapted from the energy consumption map of a full-scale sub-merged MBR (250 m3/h) applied in municipal wastewater treat-ment (Fenu et al., 2010). This capacity might be too-large, butstill reasonable for bulk-scale microalgal biomass production, likeas raw material of biofuel. To make a complete estimation, follow-ing considerations were taken into account:

(1) The calculation was based on a similar volumetric capacity.(2) Energy consumptions associated to the bioreactor operation

solely (fine bubble aerations, sludge mixing and disposal,pre-treatment and tank recycle) were excluded.

(3) The energy consumptions of the equivalent unit operationswere assumed to be similar. In this case included influent(Pin) and permeate pumping (PP).

(4) Energy consumptions of the unit operations related to mem-brane area, i.e., cleaning in place (CIP) were calculated basedon the membrane area ratios of the microalgae harvestingunit and the referred MBR plant.

(5) In contrast to the MBR application, the energy related tomembrane aeration was not entirely required in the micro-algae filtration, but was still needed for membrane tank mix-ing and CO2 supply purposes. The amount was assumed as10% of the membrane aeration energy of coarse bubble aer-ation (AC) of the referred MBR system, thus also reducing theair compression energy demand (Ca) to a similar magnitude.This assumption is quite sensible since in the referred full-scale plant (Fenu et al., 2010), only 7.3% of the whole energydemand was used for internal circulation, chemical dosingpump and mixing. It corresponds to 19% of the Ca and AC.

Since in the referenced MBR, the operational flux was set at asub-critical value (Jref, 22 l/m2 h), the applied flux (J, l/m2 h) formicroalgae harvesting was selected as 85% of its corresponding JC

(Eq. (6)), which was assumed as its operational net-flux. Therefore,the ratio (rA) of membrane area needed (An, m2) to the referencedmunicipal MBR area (Aref, m2) can be defined as in Eq. (7). The en-ergy consumptions can thus be calculated using Eqs. (8) and (9).

J ¼ 0:85Jc ð1=m2hÞ ð6Þ

rA ¼An

Aref¼ Tref

Jð�Þ ð7Þ

Ev ¼ Pin þ Pp þ rAð0:1ðAc þ CaÞ þ CIPþ EVibÞ ðkW h=m3Þ ð8Þ

Ew ¼Ev

gaqðkW h=kgÞ ð9Þ

Where Ev is the estimated energy consumption required to per-form a concentration to a certain level based on the amount of per-meate volume (kW h/m3), Ew the estimated energy consumptionbased on dry weight of the harvested biomass (kW h/kg) and qthe solid concentration of microalgae in the feed stream (kg/m3)and EVib energy consumption of vibrating engine (kW h/m3).

The specific energy consumption for the vibration engine wasestimated to be 0.002 kW h/m2 membrane. This value is obtainedby assuming that one vibrating engine (1 kW power) is used to vi-brate a single train of 125 membrane elements containing 125 ele-ments. Each element has a 4 m2 effective filtration area (twosurface faces, having 1 m width and 2 m height), resulting in500 m2 membrane area/train. By calculating the total membranearea of the assumed plant capacity, the EVib could thus beestimated.

3. Results

3.1. Membranes

The PVDF-9 and PVDF-12 have a nominal pore size, surfaceporosity and permeance of 0.036 lm, 29% and 6291 ± 900 l/m2 -h bar, and 0.013 lm, 19.2%, and 5345 ± 426 l/m2 h bar, respectively(see Fig. S1 in Supplementary material). The permeance is quitehigh compared to most commercial MF membranes (Judd, 2006).It is mainly because of their very high surface porosity, which im-plies a very low hydraulic membrane resistance. From the SEMimages, it is clear that the microalgal cells size is much bigger thatthe membrane pore, which theoretically ensures complete biomassretention. No obvious membrane defects observed from Fig. S1.However, due to difficulties to control the constant and homoge-neous condition during lab-scale membrane preparation, some de-fect can still exist.

3.2. Critical flux and critical flux for irreversibility

Fig. 3 shows the JC and JCir for different membranes, differentfeeds and different microalgal broths. The upward arrows (") indi-cate that the critical flux of the membranes exceeded the practicallimit of 50 l/m2 h set by the pump. The IFM measurements wereconducted three times for different feeds and for both microalgalbroths. Results indicate that membrane porosity, feed type andmicroalgal species affect the filterability of the algal broth. The ef-fect of the aforementioned parameters is discussed in Sections3.2.1–3.2.2. Since the results show that JC or JCir values often exceedthe maximum applicable flux, the build-up TMP during the filtra-tions will also be used directly for comparison (Fig. 4).

3.2.1. Role of membrane pore size and surface porosityThe JC and JCir values increase with increasing membrane pore

size and/or surface porosity, i.e., from PVDF-12 to PVDF-9. Intui-tively, one can expect that a lower local flux (i.e., amount of liquidpassing through a single pore) and lower feed retentions are ex-pected for larger pores. This trend was also observed for equivalentmembranes tested in the aerated filtration system (Bilad et al.,2012). The similar role of membrane porosity was also proven byvan der Marel et al. (2010) for the filtration of activated sludge.For most applied conditions, all JCir’s exceed the maximum JH

Fig. 3. Critical flux (JC) and critical flux for irreversibility (JCir) of differentmembranes for different feed solutions. " indicates that the value is more than50 l/m2 h.

334 M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338

(>50 l/m2 h), indicating a low irreversible fouling tendency. A lowirreversible fouling rate is quite reasonable since the filtrationwas performed in rather short filtration times. During the JL, themajority of foulants, especially the ones that have a weak affinityfor the membrane surface, were removed via vibration. A moreaccurate measurement is expected by performing a longer-termfiltration or by conducting multiple IFMs in absence of any clean-ing inbetween the measurements, as demonstrated by van derMarel et al. (2009). However, those approaches were not applicablein this study because of the aforementioned reasons. The JC’s of

0

5

10

15

20

0 50 100 150 200

PVDF-9

PVDF-12

PVDF-9

PVDF-12

0 50 10

Time (m

0

5

10

15

20

TMP

(kP

a)

Feed 1 Rete

Fig. 4. The TMP profile during the IFM measurement. The figures share the y-axis. The fuwith open symbol represent the TMP at low fluxes (JL).

retentate 2 for the C. vulgaris broths were surprisingly high anddo not follow the main trend. This cannot be explained with theavailable data in this study.

3.2.2. Concentration level and microalgal speciesA clear influence of the concentration level on the JC and JCir is

shown in Fig. 3. Except for PVDF-12 used on the retentate 2 of C.vulgaris, all data show a consistent trend. The more concentratedthe microalgae solution, the higher the membrane fouling ten-dency, independent upon microalgae species, as reflected in thelower JC’s and JCir’s. The result is reasonable since at higher concen-tration, more solids and AOMs exist in the feed as a result of theprior concentration stage. Apart from that, concentrating the bio-mass also means accumulating the AOMs, and in particular thetransparent exopolymeric particles (TEP) in the retentate that willrender the membrane fouling more severe. The presence of thoseaforementioned macromolecules and their general role in mem-brane fouling has been reported elsewhere (Bilad et al., 2012; Yinet al., 2012). No visible trends for JCir for PVDF-9 and PVDF-12can be given since no significant TMP build-up is observed inFig. 4. The high JCir reflects the low degree of irreversible fouling.

This result also reveals a trade-off between the effectiveness ofmembrane filtration and the concentration level. A too high brothconcentration has a too high fouling tendency that diminishes theadvantage of applying membrane technology for further concen-tration. Therefore, the concept of applying a hybrid process, wheremembrane technology is used as the primary concentration stepfollowed by (mostly) a centrifugation process as secondary step(Bilad et al., 2012) seems recommended and should be imple-mented. The optimum primary concentration level to be achievedvia the membrane filtration should be determined, depending onthe microalgal species, the membrane properties and the opera-tional parameters.

Both microalgal broths used in this study have an almost similarfilterability judging from both their JCs and JCirs. Microalgal specieshaving a rigid cell wall structure and round shape, such as C. vulga-ris, usually have a better filterability compared to those with a softcell walls and fusiform shape, such as P. tricornutum (Lewin et al.,1985). However, the relative difference in filterability between

0 150 200 0 50 100 150 200

in)

P. t

ricor

nutu

mC

. vul

garis

Retentate 2ntate 1

ll lines with full symbols represent the TMP at high fluxes (JH) and the dotted lines

M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338 335

both considered broths cannot clearly be observed in this study,possibly due to a too low solid concentration.

3.3. Fed-batch concentration filtration

The preliminary results from the fed-batch filtrations are con-sistent with the IFM tests in Section 3.2. Filtration of the feed withlower starting biomass concentration using the more porous mem-brane (PVDF-9) offers significant advantages, as judged from itsbetter filterability, as shown in Fig. 5. A lower relative permeancemeans that the membrane performance deviates more from its ori-gin (fresh membranes). The relative permeance profiles of PVDF-9on filtration of both retentates are always higher than that ofPVDF-12. It confirms the results obtained in the IFM tests: mem-branes with a lower JC experience a higher fouling rate when theyare operated at a similar flux, and vice versa. It is widely acceptedthat operating a filtration above JC leads to the formation of a cakeor gel, thus rapid fouling is expected. Occurrence of the severe foul-ing can be seen clearly from a sharp decline of relative permeanceduring the filtration.

The filtration flux of 32 l/m2 h was initially chosen as an arbi-trary value. It was selected to be modest enough to prevent quickand severe fouling while high enough to ensure sufficient produc-tivity. The fouling rates observed are quite severe. It means that amore frequent maintenance cleaning will be required in a long-term filtration. After an (unintentional) overnight relaxation, onlya relatively small permeance recovery was observed for the P. tri-cornutum broth filtration, as can be seen from the small increaseof relative permeance just after relaxation. It is worth noting thatthe filtration performance presented here was achieved withoutany systematic optimization of the vibration parameters, and with-out involving any regular relaxation or backwashing in the filtra-tion cycle. It is thus fair still to expect a significant improvementon the filterability after substantial optimization of these parame-ters in future studies. In common practice, relaxation and/or back-washing are always applied to maintain a good filterability (Yanet al., 2012).

The effectiveness of a maintenance cleaning can be seen fromthe initial relative permeance of Filtration 2 that gets very closeto the initial relative permeance of Filtration 1. Although the ap-plied maintenance cleaning was effective, it seems that its fre-quency should be rather high, taking into account the rate ofpermeance loss during the fed-batch filtration. A high maintenancecleaning frequency means more down time, more costs for clean-ing agents and generally a shorter membrane life time (more costsfor membrane replacement). However, a more systematic and

Specific filtratio

0.00

0.04

0.08

0.12

0.16

0.20

0.00 0.06 0.12 0.18

Rel

ativ

e pe

rmea

nce

PVDF-9 PVDF-12PVDF-9 PVDF-12

P. tricornutum

Fig. 5. Filtration profile of the fed-batch concentration filtration. The dashed lines with opwere both performed at a fixed flux of 32 l/m2 h for 6 and 3.5 h, respectively. An overni

comprehensive study is still needed, especially in long-term to as-sess the relation between fouling and chemical cleaning.

3.4. Scanning electron microscopy and membrane cleaning

The SEM images of the fresh, fouled and cleaned membranes areshown in Fig. S1. The fouled membrane samples were taken after3 h of extra filtration (Section 2.4). A clear difference in surfaceproperties is observed for PVDF-9 and PVDF-12. The pore size orboth membranes is much smaller than the microalgal cell size(>2 lm), thus a complete biomass retention is expected. Comparedto the fresh membranes, some visible small defects (few cracks) onthe membrane surface and some pore enlargement were observed,most probably partly originating from the SEM sample preparation(drying) or as a side effect of the repeated and aggressive chemicalcleanings (pore enlargement). Before being analyzed, the mem-brane samples had experienced six chemical cleanings. It was re-ported that aggressive chemical cleaning enlarges the membranepore size and also deteriorates membrane chemical strength (Pus-pitasaria et al., 2010; Zhang et al., 2011).

The deposited P. tricornutum cells on the fouled PVDF-9 mem-brane surface are easily recognized (Fig. S1). The major fractionof foulants the fouled membrane samples were most probablywashed out during the SEM sample preparation, despite a seriesof formaldehyde dryings to preserve their structures. They seemedto loosely bind to the membrane surface. Smaller particles andmacromolecules are also involved in the membrane fouling, whichpartially cover the membrane surface and block the membranepores, as shown especially for PVDF-12 in the C. vulgaris filtrations.The partial coverage of the membrane surface restricts the perme-ation and increases the local flux (Discart et al., 2012).

The SEM images of the cleaned membranes in Fig. S1 also showthat the majority of the foulants could be removed from the mem-brane surface. However, small fractions of foulant still remainedafter the NaOCl cleaning (irremovable foulants), especially for themembranes applied in the P. tricornutum broth. This finding cannotbe explained in the present study due to limited supporting data.Cleaning with a combination of NaOCl and citric acid was alwaysbetter, irrespective the broths or the membranes, thus providinga strong indication of the effectiveness of combined chemicalscleaning (i.e., NaOCl and citric acid). As an oxidant, NaOCl removesorganics, but normally shows little impact on inorganic scaling.Citric acid can be used to remove the latter to ensure better clean-ing performances. In many cases, organics cannot always be re-moved from the membrane surface because they are bound tothe membrane polymer matrix via cationic bonds, in which multi-

0.00 0.06 0.12 0.18

n volume (m3/m2)

C. vulgaris

en symbols and the full lines with closed symbols represent filtration 1 and 2, whichght relaxation was performed after 3 h of filtration 1.

Table 2Harvesting efficiency and energy consumption for algae harvesting using submerged filtrations, expressed per volume of permeate and per weight of microalgae harvested. It isworth noting that very low filtration efficiency was obtained for the P. tricornutum filtration of feed 1 using PVDF-9, which most probably because of membrane defects. The fluxes(J) here, is the expected operational fluxes in an up-scaled set-up calculated using Eq. (6).

Algae species Feed q (kg/m3) Harvesting efficiency EV (kW h/m3) EW (kW h/kg)

g J (l/m2 h)

PVDF-9 (%) PVDF-12 (%) PVDF-9 PVDF-12 PVDF-9 PVDF-12 PVDF-9 PVDF-12

P. tricornutum Feed 1 0.25 54 85 42.50 42.50 Fresh broth (without concentration)Retentate 1 0.79 77 97 38.25 34.00 0.20 0.24 0.26 0.31Retentate 2 1.52 84 98 25.50 21.25 0.26 0.33 0.17 0.22

C. vulgaris Feed 1 0.21 90 100 42.50 34.00 Fresh broth (without concentration)Retentate 1 0.73 94 98 34.00 25.50 0.22 0.29 0.16 0.40Retentate 2 1.43 85 98 29.75 34.00 0.23 0.24 0.34 0.17

Table 3Energy consumption of complete microalgae harvesting using three different scenarios. The biomass density (q) after membrane filtration were recalculated based on the result ofPVDF-12, which has the higher biomass retention (always >97%). The effect of density change on the centrifugation energy consumption is not considered in this calculation.

Microalgaespecies

Processscheme

Membrane filtration Centrifugation RelativeEV

reduction (%)Relative EW

reduction (%)h(%)

q(kg/m3)

EV

(kW h/m3)

EW

(kW h/kg)

h(%)

g(%)

q(kg/m3)

EV

(kW h/m3)

EW

(kW h/kg)

EV total(kW h/m3)

EW total(kW h/kg)

P. tricornutumI 99.9 84 250 7.99 38.06 7.99 0.00 38.06 0.00II 80 1.21 0.24 1.02 19.2 1.60 1.11 1.84 76.96 2.13 94.41

III93.3

3.68 0.31 1.28 6.6 0.53 0.17 0.84 89.46 1.46 96.18

C. vulgaris I 99.9 84 250 7.99 45.32 7.99 0.00 45.32 0.00II 80 1.03 0.29 1.44 19.3 1.59 1.88 1.88 76.43 3.28 92.75III 93.3 3.09 0.24 1.19 6.6 0.53 0.20 0.77 90.40 1.39 96.93

Process scheme (I) is direct harvesting using centrifugation that requiring 99.9% of volumetric reduction (h), (II) a hybrid process involving a 5� concentration via membranefiltration, followed by centrifugation and (III) a hybrid process involving a 15� concentration via membrane filtration, followed by centrifugation. The centrifugationperformance parameters were adapted from Ref. Uduman et al. (2010). The h is the required volume to be reduced to meet the concentration objective.

336 M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338

valent cations act as bridging agents (Lee and Kim, 2009). Only alimited number of crystal precipitates were observed in the SEMimages and EDX spectra (data not shown) respectively, indicatingefficient inorganic removal using citric acid.

3.5. Harvesting efficiency and energy consumption

The harvesting efficiency and energy consumption of the differ-ent membrane filtrations performed in this study are shown in Ta-ble 2. A higher separation efficiency (P97%) was obtained forPVDF-12 than for PVDF-9, which is normal based on their pore size.A low biomass retention was found for the filtration of the P. tricor-nutum feed 1 broth using PVDF-9. This result is very surprising,considering the relative difference of membrane effective pore sizeand microalgal cell size, as also discussed in Section 3.4. Since suchlow efficiency is not acceptable (because of significant biomasswashed-out, thus diminishing yield), only results from PVDF-12are used in the energy consumption estimation. The observedlow efficiency of <90% is probably caused by occasional membranedefects that appeared during membrane potting. The gradualimprovement in retention efficiency of the later filtration (filtration2) is most probably due to the blockage of this limited number ofdefects by the microalgal cells.

The biomass concentrations after 5� and 15� volumetric con-centrations had increased by factor of 3.2 and 6.1 for P. tricornutumand 3.5 and 6.8 for C. vulgaris, respectively. These values are indeedmuch lower than anticipated from the volumetric concentrationfactors (5� and 15�). A significant biomass loss could have oc-curred via the cake layer on the fouled membrane, which was re-moved during chemical cleaning, together with some drainingvia the membrane or potting defects.

Table 3 presents the estimated energy consumption of a full-scale process of microalgae harvesting, extrapolated from the dataobtained in this lab-scale study. It is clearly seen that membraneswith better filterability (represented by their realized fluxes, recal-culated using Eq. (6)), always offer a lower EV. This result suggeststhe importance of appropriate membrane material/properties hav-ing better fouling resistance and permeation properties to meet theharvesting objectives.

In general, the submerged microfiltration using the MMV sys-tem offers a relatively low energy consumption. The lowest energyconsumption obtained was 0.20 and 0.22 kW h/m3 for P. tricornu-tum and C. vulgaris, respectively. This value is slightly better thanthe one achieved using aerated filtration of 0.25 kW h/m3, suggest-ing a better energy saving offered by the MMV system (Bilad et al.,2012). However, a direct comparison using exactly the same feed isnecessary to observe the real advantages of the vibrated over theaerated systems, as will be done in a further study.

To demonstrate the effectiveness of a submerged membrane fil-tration as a first step in the microalgae harvesting, more compre-hensive energy consumption estimations are presented inTable 3. Three different scenarios of microalgae harvesting arecompared, namely (I) direct harvesting using centrifugation, (II) ahybrid process involving a 5� volumetric concentration via mem-brane filtration, followed by centrifugation and (III) a hybrid pro-cess involving a 15� volumetric concentration via membranefiltration, followed by centrifugation. The objective of harvestingat the end of each process scheme is to obtain a 250 kg/m3 micro-algal slurry (suitable for the next step of final dewatering (Udumanet al., 2010). Those magnitudes of concentrations of membrane fil-tration of 5� and 15� and centrifugation up to 25% w/w, are pro-ven feasible in this study and from literature (Uduman et al., 2010),respectively. The harvesting efficiency and energy consumption of

M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338 337

membrane filtration data used were based on Table 2 and from lit-erature data for the centrifugation step (EV = 8 kW h/m3, g = 84%)(Mohn, 1980). The initial concentrations for both broths are as-sumed to be similar to feed 1, and the biomass concentrations afterthe concentration step were recalculated using PVDF-12 data(>97% biomass retentions). The effect of density change on the cen-trifugation energy consumption was not considered in thiscalculation.

For scheme I (Table 3), a volumetric reduction factor (h) of>99.9% is required to achieve the harvesting objective. It meansthat the majority of the water should be eliminated from the freshfeed broth using high-cost centrifugation. It corresponds to a totalenergy consumption of 7.99 kW h/m3 feed, or 38.06 and45.32 kW h/kg of harvested P. tricornutum and C. vulgaris, respec-tively. Those values will only be economically feasible for thehigh-valued microalgal products, but not for the low-value deriva-tive products, such as biofuel.

For scheme II, a h-value of 80% is achieved by membrane filtra-tion to reach a density of 1.21 and 1.03 kg/m3 respectively for P. tri-cornutum and C. vulgaris broths, leaving only 19.19% forcentrifugation to reach the final q of 250 kg/m3. Therefore, a signif-icant EV reduction of >77% and EW reduction of >92% can be real-ized in scheme II. An even more promising result was obtainedfor Scheme III. Since a h-value of 93.33% is achieved via membranefiltration to reach densities of 3.68 and 3.09 kg/m3, respectively forP. tricornutum and C. vulgaris broths, only 6.60% is left for centrifu-gation. A more than 90% and 96% reduction of EV and EW can beachieved. This is mainly because the major part of the requiredpre-harvesting is realized via low-cost submerged vibrated mem-brane filtration.

The fed-batch concentration level significantly affects the totalEV and EW. It is clear that the higher the required concentration le-vel, the higher the direct energy consumption of harvesting pro-cess. However, a too high concentration by means of membranefiltration could lead to severe membrane fouling, as discussed ear-lier. As indicated in Table 2, increasing the concentration level sig-nificantly lowers the applied flux, which on its turn increases theenergy consumption associated with the membrane filtration. Athigh biomass concentration, the viscosity of the broth slurry be-comes very high, thus reducing its filterability. The optimum valuefor the concentration level thus varies per microalgae species.

Comparing the energy consumption of different unit operationsavailable in literature is very difficult, due to the differences in mic-roalgal species, initial and final biomass concentrations, scale,assumptions, calculation methods, etc. It is also important to no-tice that each process can not necessarily be applied to other feeds.For instance, pressure and vacuum filters are only applicable forlarge microalgal cells, such as C. proboscideum and Scenedesmus,but not for P. tricornutum or C. vulgaris. The results in Table 3 areconsiderably lower compared to literature data for most studiedharvesting technologies (Danquah et al., 2009; Poelman et al.,1997).

4. Conclusions

This study reveals the advantage of using submerged vibratedmicrofiltration as an energy efficient technology for microalgaepre-harvesting, thus forming a hybrid process with the well-estab-lished centrifugation. The IFM and fed-batch filtration results sug-gest a moderate degree of membrane fouling, allowing the systemto operate at high fluxes. Combining NaOCl and citric acid was effi-cient to remove the majority of foulants. However, its impact onoperational (chemical costs and operational complexity) andinvestment (membrane replacement) will still need to be ad-dressed in future studies.

Acknowledgements

KU Leuven for support in the frame of the OT (OT 11/061), IDOfinancing (IDO 06/008), and the Flemish Government for theMethusalem and FWO funding (G.0808.10N) and the Federal Gov-ernment for an IAP Grant. Muhammad R. Bilad acknowledges KULeuven for a PDM postdoctoral fellowship.

Institute for the promotion of Innovation by Science and Tech-nology-Strategic Basic Research (IWT-SBO) project Sunlight andthe KU Leuven Research Coordination Office-Industrial ResearchFund (DOC-IOF) project Algae-Tech and bioethanol project (IOF-KP/10/002).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.biortech.2013.03.175.

References

Greenwell, H.C., Laurens, L.M.L., Shields, R.J., Lovitt, R.W., Flynn, K.J., 2010. Placingmicroalgae on the biofuels priority list: a review of the technological challenges.J. R. Soc. Interface 7, 703–726.

Raja, R., Hemaiswarya, S., Kumar, N.A., Sridhar, S., Rengasamy, R., 2008. Aperspective on the biotechnological potential of microalgae. Crit. Rev.Microbiol. 34, 77–88.

Maja Šoštaric, M., Klinar, D., Bricelj, M., Golob, J., Berovic, M., Likozar, B., 2012.Growth, lipid extraction and thermal degradation of the microalga Chlorellavulgaris. New Biotechnol. 29, 325–331.

Rossignol, N., Vandajon, L., Jaouen, P., Quemeneur, F., 1991. Membrane technologyfor the continuous separation microalgae/culture medium: comparedperformances of cross-flow microfiltration and ultrafiltration. Aquacut. Eng.20, 191.

Zhang, X., Hu, Q., Sommerfeld, M., Puruhito, E., Chen, Y., 2010. Harvesting algalbiomass for biofuels using ultrafiltration membranes. Bioresour. Technol. 101,5297–5304.

Bilad, M.R., Vandamme, D., Foubert, I., Muylaert, K., Vankelecom, I.F.J., 2012.Harvesting microalgal biomass using submerged microfiltration process.Bioresour. Technol. 111, 343–352.

Mouchet, P., Bonnelye, V., 1998. Solving algae problems: French expertise andworldwide applications. J. Water SRT – Aqua 3, 125–141.

Judd, S., The, M.B.R., 2006. Book: Principles and Applications of MembraneBioreactors in Water and Wastewater Treatment, first ed. Elsevier, Amsterdam.

Le-Clech, P., Chen, V., Fane, A.G., 2006. Fouling in membrane bioreactors used inwastewater treatment. J. Membr. Sci. 284, 17–53.

Babel, S., Takizawa, S., 2010. Microfiltration membrane fouling and cake behaviourduring algal filtration. Desalination 261, 46–51.

Ladner, D.A., Vardonaand, D.R., Clark, M.M., 2010. Effects of shear on microfiltrationand ultrafiltration fouling by marine bloom-forming algae. J. Membr. Sci. 356,33–43.

Cui, Z.F., Changm, S., Fane, A.G., 2003. The use of gas bubbling to enhance membraneprocesses. J. Membr. Sci. 221, 1–35.

Genkin, G., Waite, T.D., Fane, A.G., Chang, S., 2006. The effect of vibration andcoagulant addition on the filtration performance of submerged hollow fibremembranes. J. Membr. Sci. 281, 726–734.

Jaffrin, M.Y., 2008. Dynamic shear-enhanced membrane filtration: a review ofrotating disks, rotating membranes and vibrating systems. J. Membr. Sci. 328,7–25.

Bilad, M.R., Mezohegyi, G., Declerck, P., Vankelecom, I.F.J., 2012. Novel magneticallyinduced membrane vibration (MMV) for fouling control in membranebioreactors. Water Res. 46, 63–72.

Frappart, M., Masse, A., Jaffrin, M.Y., Pruvost, J., Jaouen, P., 2011. Influence ofhydrodynamics in tangential and dynamic ultrafiltration systems formicroalgae separation. Desalination 263, 279–283.

Rios, S.D., Clavero, E., Salvado, J., Farriol, X., Torras, C., 2011. Dynamic microfiltrationin microalgae harvesting for biodiesel production. Ind. Eng. Chem. Res. 50,2455–2460.

van der Marel, P., Zwijnenburg, A., Kemperman, A.J.B., Wessling, M., Temmink, H.,van der Meer, W.G.J., 2009. An improved flux-step method to determine thecritical flux and the critical flux for irreversibility in a membrane bioreactor. J.Membr. Sci. 332, 24–29.

Vandamme, D., Pontes, S.C., Goiris, K., Foubert, I., Pinoy, L.J., Muylaert, K., 2011.Evaluation of electro-coagulation–flocculation for harvesting marine andfreshwater microalgae. Biotechnol. Bioeng. 108, 2320–2329.

Bilad, M.R., Declerck, P., Piasecka, A., Vanysacker, L., Yan, X., Vankelecom, I.F.J., 2011.Development and validation of a high-throughput membrane bioreactor (HT-MBR). J. Membr. Sci. 379, 146–153.

338 M.R. Bilad et al. / Bioresource Technology 138 (2013) 329–338

Mezohegyi, G., Bilad, M.R., Vankelecom, I.F.J., 2012. Direct sewage up-concentrationby submerged aerated and vibrated membranes. Bioresour. Technol. 118, 1–7.

Fenu, A., Roels, J., Wambecq, T., De Gussem, K., Thoeye, C., De Gueldre, G., Van DeSteene, B., 2010. Energy audit of a full scale MBR system. Desalination 262, 121–128.

van der Marel, P., Zwijnenburg, A., Kemperman, A., Wessling, M., Temmink, H., vander Meer, W., 2010. Influence of membrane properties on fouling in submergedmembrane bioreactors. J. Membr. Sci. 348, 66–74.

Yin, Y., Hong-Ying, H., Xin, L., Yin-Hu, W., Zhang, Z., Sheng-Lan, J., 2012.Accumulation characteristics of soluble algal products (SAP) by a fresh watermicroalga Scenedesmus sp. LX1 during batch cultivation for biofuel production.Bioresour. Technol. 110, 184–189.

Lewin, J.C., Lewin, R.A., Philpotdt, E., 1985. Observations on Phaeodactylumtricornutum. J. Gen. Microbiol. 18, 418–426.

Yan, X., Bilad, M.R., Gerards, R., Vriens, L., Piasecka, A., Vankelecom, I.F.J., 2012.Comparison of MBR performance and membrane cleaning in a single-stageactivated sludge system and a two-stage anaerobic/aerobic (A/A) system fortreating synthetic molasses wastewater. J. Membr. Sci. 394–395, 49–56.

Puspitasaria, V., Granville, A., Clech, P.L., Chen, V., 2010. Cleaning and ageing effectof sodium hypochlorite on polyvinylidene fluoride (PVDF) membrane. Sep.Purif. Technol. 72, 301–308.

Zhang, Y., Tian, J., Liang, H., Nan, J., Chen, Z., Li, G., 2011. Chemical cleaning of fouledPVC membrane during ultrafiltration of algal-rich water. J. Environ. Sci. 23, 529–536.

Discart, V., Bilad, M.R., Vandamme, D., Foubert, I., Muylaert, K., Vankelecom, I.F.J.,2012. Role of transparent exopolymeric particles on membrane fouling:Chlorella vulgaris broth filtration. Bioresour. Technol. 129, 18–25.

Lee, M., Kim, J., 2009. Membrane autopsy to investigate CaCO3 scale formation inpilot-scale, submerged membrane bioreactor treating calcium-rich wastewater.J. Chem. Technol. Biotechnol. 84, 1397–1404.

Uduman, N., Qi, Y., Danquah, M.K., Forde, G.M., Hoadley, A., 2010. Dewatering ofmicroalgal cultures: a major bottleneck to algae-based fuels. J. Renew. Sustain.Energ. 2, 1–15.

Mohn, F.H., 1980. Experiences and strategies in the recovery of biomass from masscultures of microalgae in algae biomass. In: Schelef, G., Soeder, C.J. (Ed.),Elsevier, Amsterdam, pp. 547–571.

Danquah, M.K., Ang, L., Uduman, N., Moheimani, N., Forde, G.M., 2009. Dewateringof microalgal culture for biodiesel production: exploring polymer flocculationand tangential flow filtration. J. Chem. Technol. Biotechnol. 84, 1078–1083.

Poelman, E., De Pauw, N., Jeurissen, B., 1997. Potential of electrolytic flocculation forrecovery of microalgae. Resour. Conserv. Recycl. 19, 1–10.